diff --git a/deps/jemalloc-win/projects/jemalloc/proj.win32/vc2013/jemalloc.vcxproj b/deps/jemalloc-win/projects/jemalloc/proj.win32/vc2013/jemalloc.vcxproj index d0158cd0..425a08be 100644 --- a/deps/jemalloc-win/projects/jemalloc/proj.win32/vc2013/jemalloc.vcxproj +++ b/deps/jemalloc-win/projects/jemalloc/proj.win32/vc2013/jemalloc.vcxproj @@ -23,7 +23,7 @@ {8B897E33-6428-4254-8335-4911D179BAD1} Jemalloc_vc2008 Win32Proj - 10.0.14393.0 + 10.0.17134.0 diff --git a/msvs/RedisBenchmark/RedisBenchmark.vcxproj b/msvs/RedisBenchmark/RedisBenchmark.vcxproj index 64284512..45da755b 100644 --- a/msvs/RedisBenchmark/RedisBenchmark.vcxproj +++ b/msvs/RedisBenchmark/RedisBenchmark.vcxproj @@ -22,7 +22,7 @@ {B00D4BB5-44DE-405E-839C-D16F547006CF} Win32Proj RedisBenchmark - 10.0.14393.0 + 10.0.17134.0 diff --git a/msvs/RedisCli/RedisCli.vcxproj b/msvs/RedisCli/RedisCli.vcxproj index fb1d4421..8f372232 100644 --- a/msvs/RedisCli/RedisCli.vcxproj +++ b/msvs/RedisCli/RedisCli.vcxproj @@ -21,7 +21,7 @@ {392BBB91-3934-4A56-AF42-65C5728311E8} RedisCli - 10.0.14393.0 + 10.0.17134.0 diff --git a/msvs/RedisForWindows.aps b/msvs/RedisForWindows.aps index eeeff945..47e5aaca 100644 Binary files a/msvs/RedisForWindows.aps and b/msvs/RedisForWindows.aps differ diff --git a/msvs/RedisServer.vcxproj b/msvs/RedisServer.vcxproj index dfbc94c1..3a3f5f19 100644 --- a/msvs/RedisServer.vcxproj +++ b/msvs/RedisServer.vcxproj @@ -21,7 +21,7 @@ Win32Proj {46842776-68A5-EC98-6A09-1859BBFC73AA} - 10.0.14393.0 + 10.0.17134.0 diff --git a/msvs/geohash/geohash.vcxproj b/msvs/geohash/geohash.vcxproj index 5f291422..9b043a37 100644 --- a/msvs/geohash/geohash.vcxproj +++ b/msvs/geohash/geohash.vcxproj @@ -30,7 +30,7 @@ {2B888C28-52CF-4E94-993C-B43D77E0B91A} Win32Proj geohash - 10.0.14393.0 + 10.0.17134.0 diff --git a/msvs/hiredis/hiredis.vcxproj b/msvs/hiredis/hiredis.vcxproj index 25c8f59a..58bf09d9 100644 --- a/msvs/hiredis/hiredis.vcxproj +++ b/msvs/hiredis/hiredis.vcxproj @@ -22,7 +22,7 @@ {13E85053-54B3-487B-8DDB-3430B1C1B3BF} Win32Proj hiredis - 10.0.14393.0 + 10.0.17134.0 diff --git a/msvs/lua/lua/lua.vcxproj b/msvs/lua/lua/lua.vcxproj index 150603dc..72acdc93 100644 --- a/msvs/lua/lua/lua.vcxproj +++ b/msvs/lua/lua/lua.vcxproj @@ -23,7 +23,7 @@ v4.5 ManagedCProj lua - 10.0.14393.0 + 10.0.17134.0 diff --git a/src/Win32_Interop/Win32_Interop.vcxproj b/src/Win32_Interop/Win32_Interop.vcxproj index 24ddff26..242cec1b 100644 --- a/src/Win32_Interop/Win32_Interop.vcxproj +++ b/src/Win32_Interop/Win32_Interop.vcxproj @@ -108,7 +108,7 @@ link.exe -dll -noentry resources/EventLog.res -out:$(TargetDir)EventLog.dll Win32Proj Win32_Redis Win32_Interop - 10.0.14393.0 + 10.0.17134.0 diff --git a/src/adlist.c b/src/adlist.c index d7e1ffca..02f31e7a 100644 --- a/src/adlist.c +++ b/src/adlist.c @@ -356,7 +356,7 @@ void listJoin(list *l, list *o) { else l->head = o->head; - l->tail = o->tail; + if (o->tail) l->tail = o->tail; l->len += o->len; /* Setup other as an empty list. */ diff --git a/src/adlist.c.bak b/src/adlist.c.bak new file mode 100644 index 00000000..d7e1ffca --- /dev/null +++ b/src/adlist.c.bak @@ -0,0 +1,365 @@ +/* adlist.c - A generic doubly linked list implementation + * + * Copyright (c) 2006-2010, Salvatore Sanfilippo + * All rights reserved. + * + * Redistribution and use in source and binary forms, with or without + * modification, are permitted provided that the following conditions are met: + * + * * Redistributions of source code must retain the above copyright notice, + * this list of conditions and the following disclaimer. + * * Redistributions in binary form must reproduce the above copyright + * notice, this list of conditions and the following disclaimer in the + * documentation and/or other materials provided with the distribution. + * * Neither the name of Redis nor the names of its contributors may be used + * to endorse or promote products derived from this software without + * specific prior written permission. + * + * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" + * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE + * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE + * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE + * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR + * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF + * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS + * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN + * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) + * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE + * POSSIBILITY OF SUCH DAMAGE. + */ +#ifdef _WIN32 +#include "Win32_Interop/Win32_Portability.h" +#include "Win32_Interop/win32_types.h" +#endif + +#include +#include "adlist.h" +#include "zmalloc.h" + +/* Create a new list. The created list can be freed with + * AlFreeList(), but private value of every node need to be freed + * by the user before to call AlFreeList(). + * + * On error, NULL is returned. Otherwise the pointer to the new list. */ +list *listCreate(void) +{ + struct list *list; + + if ((list = zmalloc(sizeof(*list))) == NULL) + return NULL; + list->head = list->tail = NULL; + list->len = 0; + list->dup = NULL; + list->free = NULL; + list->match = NULL; + return list; +} + +/* Remove all the elements from the list without destroying the list itself. */ +void listEmpty(list *list) +{ + PORT_ULONG len; + listNode *current, *next; + + current = list->head; + len = list->len; + while(len--) { + next = current->next; + if (list->free) list->free(current->value); + zfree(current); + current = next; + } + list->head = list->tail = NULL; + list->len = 0; +} + +/* Free the whole list. + * + * This function can't fail. */ +void listRelease(list *list) +{ + listEmpty(list); + zfree(list); +} + +/* Add a new node to the list, to head, containing the specified 'value' + * pointer as value. + * + * On error, NULL is returned and no operation is performed (i.e. the + * list remains unaltered). + * On success the 'list' pointer you pass to the function is returned. */ +list *listAddNodeHead(list *list, void *value) +{ + listNode *node; + + if ((node = zmalloc(sizeof(*node))) == NULL) + return NULL; + node->value = value; + if (list->len == 0) { + list->head = list->tail = node; + node->prev = node->next = NULL; + } else { + node->prev = NULL; + node->next = list->head; + list->head->prev = node; + list->head = node; + } + list->len++; + return list; +} + +/* Add a new node to the list, to tail, containing the specified 'value' + * pointer as value. + * + * On error, NULL is returned and no operation is performed (i.e. the + * list remains unaltered). + * On success the 'list' pointer you pass to the function is returned. */ +list *listAddNodeTail(list *list, void *value) +{ + listNode *node; + + if ((node = zmalloc(sizeof(*node))) == NULL) + return NULL; + node->value = value; + if (list->len == 0) { + list->head = list->tail = node; + node->prev = node->next = NULL; + } else { + node->prev = list->tail; + node->next = NULL; + list->tail->next = node; + list->tail = node; + } + list->len++; + return list; +} + +list *listInsertNode(list *list, listNode *old_node, void *value, int after) { + listNode *node; + + if ((node = zmalloc(sizeof(*node))) == NULL) + return NULL; + node->value = value; + if (after) { + node->prev = old_node; + node->next = old_node->next; + if (list->tail == old_node) { + list->tail = node; + } + } else { + node->next = old_node; + node->prev = old_node->prev; + if (list->head == old_node) { + list->head = node; + } + } + if (node->prev != NULL) { + node->prev->next = node; + } + if (node->next != NULL) { + node->next->prev = node; + } + list->len++; + return list; +} + +/* Remove the specified node from the specified list. + * It's up to the caller to free the private value of the node. + * + * This function can't fail. */ +void listDelNode(list *list, listNode *node) +{ + if (node->prev) + node->prev->next = node->next; + else + list->head = node->next; + if (node->next) + node->next->prev = node->prev; + else + list->tail = node->prev; + if (list->free) list->free(node->value); + zfree(node); + list->len--; +} + +/* Returns a list iterator 'iter'. After the initialization every + * call to listNext() will return the next element of the list. + * + * This function can't fail. */ +listIter *listGetIterator(list *list, int direction) +{ + listIter *iter; + + if ((iter = zmalloc(sizeof(*iter))) == NULL) return NULL; + if (direction == AL_START_HEAD) + iter->next = list->head; + else + iter->next = list->tail; + iter->direction = direction; + return iter; +} + +/* Release the iterator memory */ +void listReleaseIterator(listIter *iter) { + zfree(iter); +} + +/* Create an iterator in the list private iterator structure */ +void listRewind(list *list, listIter *li) { + li->next = list->head; + li->direction = AL_START_HEAD; +} + +void listRewindTail(list *list, listIter *li) { + li->next = list->tail; + li->direction = AL_START_TAIL; +} + +/* Return the next element of an iterator. + * It's valid to remove the currently returned element using + * listDelNode(), but not to remove other elements. + * + * The function returns a pointer to the next element of the list, + * or NULL if there are no more elements, so the classical usage patter + * is: + * + * iter = listGetIterator(list,); + * while ((node = listNext(iter)) != NULL) { + * doSomethingWith(listNodeValue(node)); + * } + * + * */ +listNode *listNext(listIter *iter) +{ + listNode *current = iter->next; + + if (current != NULL) { + if (iter->direction == AL_START_HEAD) + iter->next = current->next; + else + iter->next = current->prev; + } + return current; +} + +/* Duplicate the whole list. On out of memory NULL is returned. + * On success a copy of the original list is returned. + * + * The 'Dup' method set with listSetDupMethod() function is used + * to copy the node value. Otherwise the same pointer value of + * the original node is used as value of the copied node. + * + * The original list both on success or error is never modified. */ +list *listDup(list *orig) +{ + list *copy; + listIter iter; + listNode *node; + + if ((copy = listCreate()) == NULL) + return NULL; + copy->dup = orig->dup; + copy->free = orig->free; + copy->match = orig->match; + listRewind(orig, &iter); + while((node = listNext(&iter)) != NULL) { + void *value; + + if (copy->dup) { + value = copy->dup(node->value); + if (value == NULL) { + listRelease(copy); + return NULL; + } + } else + value = node->value; + if (listAddNodeTail(copy, value) == NULL) { + listRelease(copy); + return NULL; + } + } + return copy; +} + +/* Search the list for a node matching a given key. + * The match is performed using the 'match' method + * set with listSetMatchMethod(). If no 'match' method + * is set, the 'value' pointer of every node is directly + * compared with the 'key' pointer. + * + * On success the first matching node pointer is returned + * (search starts from head). If no matching node exists + * NULL is returned. */ +listNode *listSearchKey(list *list, void *key) +{ + listIter iter; + listNode *node; + + listRewind(list, &iter); + while((node = listNext(&iter)) != NULL) { + if (list->match) { + if (list->match(node->value, key)) { + return node; + } + } else { + if (key == node->value) { + return node; + } + } + } + return NULL; +} + +/* Return the element at the specified zero-based index + * where 0 is the head, 1 is the element next to head + * and so on. Negative integers are used in order to count + * from the tail, -1 is the last element, -2 the penultimate + * and so on. If the index is out of range NULL is returned. */ +listNode *listIndex(list *list, PORT_LONG index) { + listNode *n; + + if (index < 0) { + index = (-index)-1; + n = list->tail; + while(index-- && n) n = n->prev; + } else { + n = list->head; + while(index-- && n) n = n->next; + } + return n; +} + +/* Rotate the list removing the tail node and inserting it to the head. */ +void listRotate(list *list) { + listNode *tail = list->tail; + + if (listLength(list) <= 1) return; + + /* Detach current tail */ + list->tail = tail->prev; + list->tail->next = NULL; + /* Move it as head */ + list->head->prev = tail; + tail->prev = NULL; + tail->next = list->head; + list->head = tail; +} + +/* Add all the elements of the list 'o' at the end of the + * list 'l'. The list 'other' remains empty but otherwise valid. */ +void listJoin(list *l, list *o) { + if (o->head) + o->head->prev = l->tail; + + if (l->tail) + l->tail->next = o->head; + else + l->head = o->head; + + l->tail = o->tail; + l->len += o->len; + + /* Setup other as an empty list. */ + o->head = o->tail = NULL; + o->len = 0; +} diff --git a/src/ae.c b/src/ae.c index ff4662f6..1d2b03af 100644 --- a/src/ae.c +++ b/src/ae.c @@ -51,6 +51,7 @@ #include #include +#include "ae.h" #include "zmalloc.h" #include "config.h" @@ -173,6 +174,10 @@ void aeDeleteFileEvent(aeEventLoop *eventLoop, int fd, int mask) aeFileEvent *fe = &eventLoop->events[fd]; if (fe->mask == AE_NONE) return; + /* We want to always remove AE_BARRIER if set when AE_WRITABLE + * is removed. */ + if (mask & AE_WRITABLE) mask |= AE_BARRIER; + aeApiDelEvent(eventLoop, fd, mask); fe->mask = fe->mask & (~mask); if (fd == eventLoop->maxfd && fe->mask == AE_NONE) { @@ -238,7 +243,10 @@ PORT_LONGLONG aeCreateTimeEvent(aeEventLoop *eventLoop, PORT_LONGLONG millisecon te->timeProc = proc; te->finalizerProc = finalizerProc; te->clientData = clientData; + te->prev = NULL; te->next = eventLoop->timeEventHead; + if (te->next) + te->next->prev = te; eventLoop->timeEventHead = te; return id; } @@ -285,7 +293,7 @@ static aeTimeEvent *aeSearchNearestTimer(aeEventLoop *eventLoop) /* Process time events */ static int processTimeEvents(aeEventLoop *eventLoop) { int processed = 0; - aeTimeEvent *te, *prev; + aeTimeEvent *te; PORT_LONGLONG maxId; time_t now = time(NULL); @@ -306,7 +314,6 @@ static int processTimeEvents(aeEventLoop *eventLoop) { } eventLoop->lastTime = now; - prev = NULL; te = eventLoop->timeEventHead; maxId = eventLoop->timeEventNextId - 1; while (te) { @@ -316,10 +323,12 @@ static int processTimeEvents(aeEventLoop *eventLoop) { /* Remove events scheduled for deletion. */ if (te->id == AE_DELETED_EVENT_ID) { aeTimeEvent *next = te->next; - if (prev == NULL) - eventLoop->timeEventHead = te->next; + if (te->prev) + te->prev->next = te->next; else - prev->next = te->next; + eventLoop->timeEventHead = te->next; + if (te->next) + te->next->prev = te->prev; if (te->finalizerProc) te->finalizerProc(eventLoop, te->clientData); zfree(te); @@ -347,12 +356,10 @@ static int processTimeEvents(aeEventLoop *eventLoop) { processed++; if (retval != AE_NOMORE) { aeAddMillisecondsToNow(retval, &te->when_sec, &te->when_ms); - } - else { + } else { te->id = AE_DELETED_EVENT_ID; } } - prev = te; te = te->next; } return processed; @@ -386,7 +393,7 @@ int aeProcessEvents(aeEventLoop *eventLoop, int flags) if (!(flags & AE_TIME_EVENTS) && !(flags & AE_FILE_EVENTS)) return 0; /* Note that we want call select() even if there are no - * file events to process as PORT_LONG as we want to process time + * file events to process as long as we want to process time * events, in order to sleep until the next time event is ready * to fire. */ if (eventLoop->maxfd != -1 || @@ -412,21 +419,18 @@ int aeProcessEvents(aeEventLoop *eventLoop, int flags) if (ms > 0) { tvp->tv_sec = ms / 1000; tvp->tv_usec = (ms % 1000) * 1000; - } - else { + } else { tvp->tv_sec = 0; tvp->tv_usec = 0; } - } - else { + } else { /* If we have to check for events but need to return * ASAP because of AE_DONT_WAIT we need to set the timeout * to zero */ if (flags & AE_DONT_WAIT) { tv.tv_sec = tv.tv_usec = 0; tvp = &tv; - } - else { + } else { /* Otherwise we can block */ tvp = NULL; /* wait forever */ } @@ -444,19 +448,49 @@ int aeProcessEvents(aeEventLoop *eventLoop, int flags) aeFileEvent *fe = &eventLoop->events[eventLoop->fired[j].fd]; int mask = eventLoop->fired[j].mask; int fd = eventLoop->fired[j].fd; - int rfired = 0; + int fired = 0; /* Number of events fired for current fd. */ - /* note the fe->mask & mask & ... code: maybe an already processed - * event removed an element that fired and we still didn't - * processed, so we check if the event is still valid. */ - if (fe->mask & mask & AE_READABLE) { - rfired = 1; + /* Normally we execute the readable event first, and the writable + * event laster. This is useful as sometimes we may be able + * to serve the reply of a query immediately after processing the + * query. + * + * However if AE_BARRIER is set in the mask, our application is + * asking us to do the reverse: never fire the writable event + * after the readable. In such a case, we invert the calls. + * This is useful when, for instance, we want to do things + * in the beforeSleep() hook, like fsynching a file to disk, + * before replying to a client. */ + int invert = fe->mask & AE_BARRIER; + + /* Note the "fe->mask & mask & ..." code: maybe an already + * processed event removed an element that fired and we still + * didn't processed, so we check if the event is still valid. + * + * Fire the readable event if the call sequence is not + * inverted. */ + if (!invert && fe->mask & mask & AE_READABLE) { fe->rfileProc(eventLoop, fd, fe->clientData, mask); + fired++; } + + /* Fire the writable event. */ if (fe->mask & mask & AE_WRITABLE) { - if (!rfired || fe->wfileProc != fe->rfileProc) + if (!fired || fe->wfileProc != fe->rfileProc) { fe->wfileProc(eventLoop, fd, fe->clientData, mask); + fired++; + } } + + /* If we have to invert the call, fire the readable event now + * after the writable one. */ + if (invert && fe->mask & mask & AE_READABLE) { + if (!fired || fe->wfileProc != fe->rfileProc) { + fe->rfileProc(eventLoop,fd,fe->clientData,mask); + fired++; + } + } + processed++; } } @@ -478,15 +512,13 @@ int aeWait(int fd, int mask, PORT_LONGLONG milliseconds) { if (mask & AE_READABLE) pfd.events |= POLLIN; if (mask & AE_WRITABLE) pfd.events |= POLLOUT; - if ((retval = poll(&pfd, 1, (int) milliseconds)) == 1) { - WIN_PORT_FIX /* cast (int) */ + if ((retval = poll(&pfd, 1, (int) milliseconds)) == 1) { WIN_PORT_FIX /* cast (int) */ if (pfd.revents & POLLIN) retmask |= AE_READABLE; if (pfd.revents & POLLOUT) retmask |= AE_WRITABLE; if (pfd.revents & POLLERR) retmask |= AE_WRITABLE; if (pfd.revents & POLLHUP) retmask |= AE_WRITABLE; return retmask; - } - else { + } else { return retval; } } @@ -510,4 +542,4 @@ void aeSetBeforeSleepProc(aeEventLoop *eventLoop, aeBeforeSleepProc *beforesleep void aeSetAfterSleepProc(aeEventLoop *eventLoop, aeBeforeSleepProc *aftersleep) { eventLoop->aftersleep = aftersleep; -} \ No newline at end of file +} diff --git a/src/ae.c.bak b/src/ae.c.bak new file mode 100644 index 00000000..c44fa388 --- /dev/null +++ b/src/ae.c.bak @@ -0,0 +1,515 @@ +/* A simple event-driven programming library. Originally I wrote this code + * for the Jim's event-loop (Jim is a Tcl interpreter) but later translated + * it in form of a library for easy reuse. + * + * Copyright (c) 2006-2010, Salvatore Sanfilippo + * All rights reserved. + * + * Redistribution and use in source and binary forms, with or without + * modification, are permitted provided that the following conditions are met: + * + * * Redistributions of source code must retain the above copyright notice, + * this list of conditions and the following disclaimer. + * * Redistributions in binary form must reproduce the above copyright + * notice, this list of conditions and the following disclaimer in the + * documentation and/or other materials provided with the distribution. + * * Neither the name of Redis nor the names of its contributors may be used + * to endorse or promote products derived from this software without + * specific prior written permission. + * + * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" + * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE + * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE + * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE + * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR + * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF + * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS + * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN + * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) + * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE + * POSSIBILITY OF SUCH DAMAGE. + */ + +#ifdef _WIN32 +#include "Win32_Interop\win32_types_hiredis.h" +#endif +#include "ae.h" +#include +#include +#ifdef _WIN32 +#include +#include +#include "../src/Win32_Interop/Win32_FDAPI.h" +#include "../src/Win32_Interop/Win32_Service.h" +#else +#include +#include +#include +#endif +#include +#include +#include +#include + +#include "ae.h" +#include "zmalloc.h" +#include "config.h" + + /* Include the best multiplexing layer supported by this system. + * The following should be ordered by performances, descending. */ +#ifdef _WIN32 +#include "ae_wsiocp.c" +#else +#ifdef HAVE_EVPORT +#include "ae_evport.c" +#else +#ifdef HAVE_EPOLL +#include "ae_epoll.c" +#else +#ifdef HAVE_KQUEUE +#include "ae_kqueue.c" +#else +#include "ae_select.c" +#endif +#endif +#endif +#endif + +aeEventLoop *aeCreateEventLoop(int setsize) { + aeEventLoop *eventLoop; + int i; + + if ((eventLoop = zmalloc(sizeof(*eventLoop))) == NULL) goto err; + eventLoop->events = zmalloc(sizeof(aeFileEvent)*setsize); + eventLoop->fired = zmalloc(sizeof(aeFiredEvent)*setsize); + if (eventLoop->events == NULL || eventLoop->fired == NULL) goto err; + eventLoop->setsize = setsize; + eventLoop->lastTime = time(NULL); + eventLoop->timeEventHead = NULL; + eventLoop->timeEventNextId = 0; + eventLoop->stop = 0; + eventLoop->maxfd = -1; + eventLoop->beforesleep = NULL; + eventLoop->aftersleep = NULL; + if (aeApiCreate(eventLoop) == -1) goto err; + /* Events with mask == AE_NONE are not set. So let's initialize the + * vector with it. */ + for (i = 0; i < setsize; i++) + eventLoop->events[i].mask = AE_NONE; + return eventLoop; + +err: + if (eventLoop) { + zfree(eventLoop->events); + zfree(eventLoop->fired); + zfree(eventLoop); + } + return NULL; +} + +/* Return the current set size. */ +int aeGetSetSize(aeEventLoop *eventLoop) { + return eventLoop->setsize; +} + +/* Resize the maximum set size of the event loop. + * If the requested set size is smaller than the current set size, but + * there is already a file descriptor in use that is >= the requested + * set size minus one, AE_ERR is returned and the operation is not + * performed at all. + * + * Otherwise AE_OK is returned and the operation is successful. */ +int aeResizeSetSize(aeEventLoop *eventLoop, int setsize) { + int i; + + if (setsize == eventLoop->setsize) return AE_OK; + if (eventLoop->maxfd >= setsize) return AE_ERR; + if (aeApiResize(eventLoop, setsize) == -1) return AE_ERR; + + eventLoop->events = zrealloc(eventLoop->events, sizeof(aeFileEvent)*setsize); + eventLoop->fired = zrealloc(eventLoop->fired, sizeof(aeFiredEvent)*setsize); + eventLoop->setsize = setsize; + + /* Make sure that if we created new slots, they are initialized with + * an AE_NONE mask. */ + for (i = eventLoop->maxfd + 1; i < setsize; i++) + eventLoop->events[i].mask = AE_NONE; + return AE_OK; +} + +void aeDeleteEventLoop(aeEventLoop *eventLoop) { + aeApiFree(eventLoop); + zfree(eventLoop->events); + zfree(eventLoop->fired); + zfree(eventLoop); +} + +void aeStop(aeEventLoop *eventLoop) { + eventLoop->stop = 1; +} + +int aeCreateFileEvent(aeEventLoop *eventLoop, int fd, int mask, + aeFileProc *proc, void *clientData) +{ + if (fd >= eventLoop->setsize) { + errno = ERANGE; + return AE_ERR; + } + aeFileEvent *fe = &eventLoop->events[fd]; + + if (aeApiAddEvent(eventLoop, fd, mask) == -1) + return AE_ERR; + fe->mask |= mask; + if (mask & AE_READABLE) fe->rfileProc = proc; + if (mask & AE_WRITABLE) fe->wfileProc = proc; + fe->clientData = clientData; + if (fd > eventLoop->maxfd) + eventLoop->maxfd = fd; + return AE_OK; +} + +void aeDeleteFileEvent(aeEventLoop *eventLoop, int fd, int mask) +{ + if (fd >= eventLoop->setsize) return; + aeFileEvent *fe = &eventLoop->events[fd]; + if (fe->mask == AE_NONE) return; + + /* We want to always remove AE_BARRIER if set when AE_WRITABLE + * is removed. */ + if (mask & AE_WRITABLE) mask |= AE_BARRIER; + + aeApiDelEvent(eventLoop, fd, mask); + fe->mask = fe->mask & (~mask); + if (fd == eventLoop->maxfd && fe->mask == AE_NONE) { + /* Update the max fd */ + int j; + + for (j = eventLoop->maxfd - 1; j >= 0; j--) + if (eventLoop->events[j].mask != AE_NONE) break; + eventLoop->maxfd = j; + } +} + +int aeGetFileEvents(aeEventLoop *eventLoop, int fd) { + if (fd >= eventLoop->setsize) return 0; + aeFileEvent *fe = &eventLoop->events[fd]; + + return fe->mask; +} + +static void aeGetTime(PORT_LONG *seconds, PORT_LONG *milliseconds) +{ +#ifdef _WIN32 + struct _timeb tb; + + memset(&tb, 0, sizeof(struct _timeb)); + _ftime_s(&tb); + (*seconds) = tb.time; + (*milliseconds) = tb.millitm; +#else + struct timeval tv; + + gettimeofday(&tv, NULL); + *seconds = tv.tv_sec; + *milliseconds = tv.tv_usec / 1000; +#endif +} + +static void aeAddMillisecondsToNow(PORT_LONGLONG milliseconds, PORT_LONG *sec, PORT_LONG *ms) { + PORT_LONG cur_sec, cur_ms, when_sec, when_ms; + + aeGetTime(&cur_sec, &cur_ms); + when_sec = (PORT_LONG) (cur_sec + milliseconds / 1000); + when_ms = cur_ms + milliseconds % 1000; + if (when_ms >= 1000) { + when_sec++; + when_ms -= 1000; + } + *sec = when_sec; + *ms = when_ms; +} + +PORT_LONGLONG aeCreateTimeEvent(aeEventLoop *eventLoop, PORT_LONGLONG milliseconds, + aeTimeProc *proc, void *clientData, + aeEventFinalizerProc *finalizerProc) +{ + PORT_LONGLONG id = eventLoop->timeEventNextId++; + aeTimeEvent *te; + + te = zmalloc(sizeof(*te)); + if (te == NULL) return AE_ERR; + te->id = id; + aeAddMillisecondsToNow(milliseconds, &te->when_sec, &te->when_ms); + te->timeProc = proc; + te->finalizerProc = finalizerProc; + te->clientData = clientData; + te->prev = NULL; + te->next = eventLoop->timeEventHead; + if (te->next) + te->next->prev = te; + eventLoop->timeEventHead = te; + return id; +} + +int aeDeleteTimeEvent(aeEventLoop *eventLoop, PORT_LONGLONG id) +{ + aeTimeEvent *te = eventLoop->timeEventHead; + while (te) { + if (te->id == id) { + te->id = AE_DELETED_EVENT_ID; + return AE_OK; + } + te = te->next; + } + return AE_ERR; /* NO event with the specified ID found */ +} + +/* Search the first timer to fire. + * This operation is useful to know how many time the select can be + * put in sleep without to delay any event. + * If there are no timers NULL is returned. + * + * Note that's O(N) since time events are unsorted. + * Possible optimizations (not needed by Redis so far, but...): + * 1) Insert the event in order, so that the nearest is just the head. + * Much better but still insertion or deletion of timers is O(N). + * 2) Use a skiplist to have this operation as O(1) and insertion as O(log(N)). + */ +static aeTimeEvent *aeSearchNearestTimer(aeEventLoop *eventLoop) +{ + aeTimeEvent *te = eventLoop->timeEventHead; + aeTimeEvent *nearest = NULL; + + while (te) { + if (!nearest || te->when_sec < nearest->when_sec || + (te->when_sec == nearest->when_sec && + te->when_ms < nearest->when_ms)) + nearest = te; + te = te->next; + } + return nearest; +} + +/* Process time events */ +static int processTimeEvents(aeEventLoop *eventLoop) { + int processed = 0; + aeTimeEvent *te; + PORT_LONGLONG maxId; + time_t now = time(NULL); + + /* If the system clock is moved to the future, and then set back to the + * right value, time events may be delayed in a random way. Often this + * means that scheduled operations will not be performed soon enough. + * + * Here we try to detect system clock skews, and force all the time + * events to be processed ASAP when this happens: the idea is that + * processing events earlier is less dangerous than delaying them + * indefinitely, and practice suggests it is. */ + if (now < eventLoop->lastTime) { + te = eventLoop->timeEventHead; + while (te) { + te->when_sec = 0; + te = te->next; + } + } + eventLoop->lastTime = now; + + te = eventLoop->timeEventHead; + maxId = eventLoop->timeEventNextId - 1; + while (te) { + PORT_LONG now_sec, now_ms; + PORT_LONGLONG id; + + /* Remove events scheduled for deletion. */ + if (te->id == AE_DELETED_EVENT_ID) { + aeTimeEvent *next = te->next; + if (te->prev) + te->prev->next = te->next; + else + eventLoop->timeEventHead = te->next; + if (te->next) + te->next->prev = te->prev; + if (te->finalizerProc) + te->finalizerProc(eventLoop, te->clientData); + zfree(te); + te = next; + continue; + } + + /* Make sure we don't process time events created by time events in + * this iteration. Note that this check is currently useless: we always + * add new timers on the head, however if we change the implementation + * detail, this check may be useful again: we keep it here for future + * defense. */ + if (te->id > maxId) { + te = te->next; + continue; + } + aeGetTime(&now_sec, &now_ms); + if (now_sec > te->when_sec || + (now_sec == te->when_sec && now_ms >= te->when_ms)) + { + int retval; + + id = te->id; + retval = te->timeProc(eventLoop, id, te->clientData); + processed++; + if (retval != AE_NOMORE) { + aeAddMillisecondsToNow(retval, &te->when_sec, &te->when_ms); + } else { + te->id = AE_DELETED_EVENT_ID; + } + } + te = te->next; + } + return processed; +} + +/* Process every pending time event, then every pending file event + * (that may be registered by time event callbacks just processed). + * Without special flags the function sleeps until some file event + * fires, or when the next time event occurs (if any). + * + * If flags is 0, the function does nothing and returns. + * if flags has AE_ALL_EVENTS set, all the kind of events are processed. + * if flags has AE_FILE_EVENTS set, file events are processed. + * if flags has AE_TIME_EVENTS set, time events are processed. + * if flags has AE_DONT_WAIT set the function returns ASAP until all + * if flags has AE_CALL_AFTER_SLEEP set, the aftersleep callback is called. + * the events that's possible to process without to wait are processed. + * + * The function returns the number of events processed. */ +int aeProcessEvents(aeEventLoop *eventLoop, int flags) +{ + int processed = 0, numevents; + +#ifdef _WIN32 + if (ServiceStopIssued() == TRUE) { + aeStop(eventLoop); + } +#endif + + /* Nothing to do? return ASAP */ + if (!(flags & AE_TIME_EVENTS) && !(flags & AE_FILE_EVENTS)) return 0; + + /* Note that we want call select() even if there are no + * file events to process as long as we want to process time + * events, in order to sleep until the next time event is ready + * to fire. */ + if (eventLoop->maxfd != -1 || + ((flags & AE_TIME_EVENTS) && !(flags & AE_DONT_WAIT))) { + int j; + aeTimeEvent *shortest = NULL; + struct timeval tv, *tvp; + + if (flags & AE_TIME_EVENTS && !(flags & AE_DONT_WAIT)) + shortest = aeSearchNearestTimer(eventLoop); + if (shortest) { + PORT_LONG now_sec, now_ms; + + aeGetTime(&now_sec, &now_ms); + tvp = &tv; + + /* How many milliseconds we need to wait for the next + * time event to fire? */ + PORT_LONGLONG ms = + (shortest->when_sec - now_sec) * 1000 + + shortest->when_ms - now_ms; + + if (ms > 0) { + tvp->tv_sec = ms / 1000; + tvp->tv_usec = (ms % 1000) * 1000; + } else { + tvp->tv_sec = 0; + tvp->tv_usec = 0; + } + } else { + /* If we have to check for events but need to return + * ASAP because of AE_DONT_WAIT we need to set the timeout + * to zero */ + if (flags & AE_DONT_WAIT) { + tv.tv_sec = tv.tv_usec = 0; + tvp = &tv; + } else { + /* Otherwise we can block */ + tvp = NULL; /* wait forever */ + } + } + + /* Call the multiplexing API, will return only on timeout or when + * some event fires. */ + numevents = aeApiPoll(eventLoop, tvp); + + /* After sleep callback. */ + if (eventLoop->aftersleep != NULL && flags & AE_CALL_AFTER_SLEEP) + eventLoop->aftersleep(eventLoop); + + for (j = 0; j < numevents; j++) { + aeFileEvent *fe = &eventLoop->events[eventLoop->fired[j].fd]; + int mask = eventLoop->fired[j].mask; + int fd = eventLoop->fired[j].fd; + int rfired = 0; + + /* note the fe->mask & mask & ... code: maybe an already processed + * event removed an element that fired and we still didn't + * processed, so we check if the event is still valid. */ + if (fe->mask & mask & AE_READABLE) { + rfired = 1; + fe->rfileProc(eventLoop, fd, fe->clientData, mask); + } + if (fe->mask & mask & AE_WRITABLE) { + if (!rfired || fe->wfileProc != fe->rfileProc) + fe->wfileProc(eventLoop, fd, fe->clientData, mask); + } + processed++; + } + } + /* Check time events */ + if (flags & AE_TIME_EVENTS) + processed += processTimeEvents(eventLoop); + + return processed; /* return the number of processed file/time events */ +} + +/* Wait for milliseconds until the given file descriptor becomes + * writable/readable/exception */ +int aeWait(int fd, int mask, PORT_LONGLONG milliseconds) { + struct pollfd pfd; + int retmask = 0, retval; + + memset(&pfd, 0, sizeof(pfd)); + pfd.fd = fd; + if (mask & AE_READABLE) pfd.events |= POLLIN; + if (mask & AE_WRITABLE) pfd.events |= POLLOUT; + + if ((retval = poll(&pfd, 1, (int) milliseconds)) == 1) { WIN_PORT_FIX /* cast (int) */ + if (pfd.revents & POLLIN) retmask |= AE_READABLE; + if (pfd.revents & POLLOUT) retmask |= AE_WRITABLE; + if (pfd.revents & POLLERR) retmask |= AE_WRITABLE; + if (pfd.revents & POLLHUP) retmask |= AE_WRITABLE; + return retmask; + } else { + return retval; + } +} + +void aeMain(aeEventLoop *eventLoop) { + eventLoop->stop = 0; + while (!eventLoop->stop) { + if (eventLoop->beforesleep != NULL) + eventLoop->beforesleep(eventLoop); + aeProcessEvents(eventLoop, AE_ALL_EVENTS | AE_CALL_AFTER_SLEEP); + } +} + +char *aeGetApiName(void) { + return aeApiName(); +} + +void aeSetBeforeSleepProc(aeEventLoop *eventLoop, aeBeforeSleepProc *beforesleep) { + eventLoop->beforesleep = beforesleep; +} + +void aeSetAfterSleepProc(aeEventLoop *eventLoop, aeBeforeSleepProc *aftersleep) { + eventLoop->aftersleep = aftersleep; +} diff --git a/src/ae.h b/src/ae.h index 4403b5c8..39bf9ea6 100644 --- a/src/ae.h +++ b/src/ae.h @@ -38,9 +38,14 @@ #define AE_OK 0 #define AE_ERR -1 -#define AE_NONE 0 -#define AE_READABLE 1 -#define AE_WRITABLE 2 +#define AE_NONE 0 /* No events registered. */ +#define AE_READABLE 1 /* Fire when descriptor is readable. */ +#define AE_WRITABLE 2 /* Fire when descriptor is writable. */ +#define AE_BARRIER 4 /* With WRITABLE, never fire the event if the + READABLE event already fired in the same event + loop iteration. Useful when you want to persist + things to disk before sending replies, and want + to do that in a group fashion. */ #define AE_FILE_EVENTS 1 #define AE_TIME_EVENTS 2 @@ -64,7 +69,7 @@ typedef void aeBeforeSleepProc(struct aeEventLoop *eventLoop); /* File event structure */ typedef struct aeFileEvent { - int mask; /* one of AE_(READABLE|WRITABLE) */ + int mask; /* one of AE_(READABLE|WRITABLE|BARRIER) */ aeFileProc *rfileProc; aeFileProc *wfileProc; void *clientData; @@ -78,6 +83,7 @@ typedef struct aeTimeEvent { aeTimeProc *timeProc; aeEventFinalizerProc *finalizerProc; void *clientData; + struct aeTimeEvent *prev; struct aeTimeEvent *next; } aeTimeEvent; diff --git a/src/ae.h.bak b/src/ae.h.bak new file mode 100644 index 00000000..1b5abe4d --- /dev/null +++ b/src/ae.h.bak @@ -0,0 +1,131 @@ +/* A simple event-driven programming library. Originally I wrote this code + * for the Jim's event-loop (Jim is a Tcl interpreter) but later translated + * it in form of a library for easy reuse. + * + * Copyright (c) 2006-2012, Salvatore Sanfilippo + * All rights reserved. + * + * Redistribution and use in source and binary forms, with or without + * modification, are permitted provided that the following conditions are met: + * + * * Redistributions of source code must retain the above copyright notice, + * this list of conditions and the following disclaimer. + * * Redistributions in binary form must reproduce the above copyright + * notice, this list of conditions and the following disclaimer in the + * documentation and/or other materials provided with the distribution. + * * Neither the name of Redis nor the names of its contributors may be used + * to endorse or promote products derived from this software without + * specific prior written permission. + * + * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" + * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE + * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE + * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE + * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR + * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF + * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS + * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN + * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) + * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE + * POSSIBILITY OF SUCH DAMAGE. + */ + +#ifndef __AE_H__ +#define __AE_H__ + +#include + +#define AE_OK 0 +#define AE_ERR -1 + +#define AE_NONE 0 /* No events registered. */ +#define AE_READABLE 1 /* Fire when descriptor is readable. */ +#define AE_WRITABLE 2 /* Fire when descriptor is writable. */ +#define AE_BARRIER 4 /* With WRITABLE, never fire the event if the + READABLE event already fired in the same event + loop iteration. Useful when you want to persist + things to disk before sending replies, and want + to do that in a group fashion. */ + +#define AE_FILE_EVENTS 1 +#define AE_TIME_EVENTS 2 +#define AE_ALL_EVENTS (AE_FILE_EVENTS|AE_TIME_EVENTS) +#define AE_DONT_WAIT 4 +#define AE_CALL_AFTER_SLEEP 8 + +#define AE_NOMORE -1 +#define AE_DELETED_EVENT_ID -1 + +/* Macros */ +#define AE_NOTUSED(V) ((void) V) + +struct aeEventLoop; + +/* Types and data structures */ +typedef void aeFileProc(struct aeEventLoop *eventLoop, int fd, void *clientData, int mask); +typedef int aeTimeProc(struct aeEventLoop *eventLoop, PORT_LONGLONG id, void *clientData); +typedef void aeEventFinalizerProc(struct aeEventLoop *eventLoop, void *clientData); +typedef void aeBeforeSleepProc(struct aeEventLoop *eventLoop); + +/* File event structure */ +typedef struct aeFileEvent { + int mask; /* one of AE_(READABLE|WRITABLE|BARRIER) */ + aeFileProc *rfileProc; + aeFileProc *wfileProc; + void *clientData; +} aeFileEvent; + +/* Time event structure */ +typedef struct aeTimeEvent { + PORT_LONGLONG id; /* time event identifier. */ + PORT_LONG when_sec; /* seconds */ + PORT_LONG when_ms; /* milliseconds */ + aeTimeProc *timeProc; + aeEventFinalizerProc *finalizerProc; + void *clientData; + struct aeTimeEvent *next; +} aeTimeEvent; + +/* A fired event */ +typedef struct aeFiredEvent { + int fd; + int mask; +} aeFiredEvent; + +/* State of an event based program */ +typedef struct aeEventLoop { + int maxfd; /* highest file descriptor currently registered */ + int setsize; /* max number of file descriptors tracked */ + PORT_LONGLONG timeEventNextId; + time_t lastTime; /* Used to detect system clock skew */ + aeFileEvent *events; /* Registered events */ + aeFiredEvent *fired; /* Fired events */ + aeTimeEvent *timeEventHead; + int stop; + void *apidata; /* This is used for polling API specific data */ + aeBeforeSleepProc *beforesleep; + aeBeforeSleepProc *aftersleep; +} aeEventLoop; + +/* Prototypes */ +aeEventLoop *aeCreateEventLoop(int setsize); +void aeDeleteEventLoop(aeEventLoop *eventLoop); +void aeStop(aeEventLoop *eventLoop); +int aeCreateFileEvent(aeEventLoop *eventLoop, int fd, int mask, + aeFileProc *proc, void *clientData); +void aeDeleteFileEvent(aeEventLoop *eventLoop, int fd, int mask); +int aeGetFileEvents(aeEventLoop *eventLoop, int fd); +PORT_LONGLONG aeCreateTimeEvent(aeEventLoop *eventLoop, PORT_LONGLONG milliseconds, + aeTimeProc *proc, void *clientData, + aeEventFinalizerProc *finalizerProc); +int aeDeleteTimeEvent(aeEventLoop *eventLoop, PORT_LONGLONG id); +int aeProcessEvents(aeEventLoop *eventLoop, int flags); +int aeWait(int fd, int mask, PORT_LONGLONG milliseconds); +void aeMain(aeEventLoop *eventLoop); +char *aeGetApiName(void); +void aeSetBeforeSleepProc(aeEventLoop *eventLoop, aeBeforeSleepProc *beforesleep); +void aeSetAfterSleepProc(aeEventLoop *eventLoop, aeBeforeSleepProc *aftersleep); +int aeGetSetSize(aeEventLoop *eventLoop); +int aeResizeSetSize(aeEventLoop *eventLoop, int setsize); + +#endif diff --git a/src/aof.c b/src/aof.c index ec5dab8e..42de79a5 100644 --- a/src/aof.c +++ b/src/aof.c @@ -214,25 +214,20 @@ ssize_t aofRewriteBufferWrite(int fd) { /* Starts a background task that performs fsync() against the specified * file descriptor (the one of the AOF file) in another thread. */ void aof_background_fsync(int fd) { - bioCreateBackgroundJob(BIO_AOF_FSYNC, (void*) (PORT_LONG) fd, NULL, NULL); + bioCreateBackgroundJob(BIO_AOF_FSYNC,(void*)(long)fd,NULL,NULL); } -/* Called when the user switches from "appendonly yes" to "appendonly no" - * at runtime using the CONFIG command. */ -void stopAppendOnly(void) { - serverAssert(server.aof_state != AOF_OFF); - flushAppendOnlyFile(1); - aof_fsync(server.aof_fd); - close(server.aof_fd); +/* Kills an AOFRW child process if exists */ +static void killAppendOnlyChild(void) { - server.aof_fd = -1; - server.aof_selected_db = -1; - server.aof_state = AOF_OFF; - /* rewrite operation in progress? kill it, wait child exit */ - if (server.aof_child_pid != -1) { - POSIX_ONLY(int statloc;) + POSIX_ONLY(int statloc;) - serverLog(LL_NOTICE, "Killing running AOF rewrite child: %Id", WIN_PORT_FIX /* %ld -> %Id */ + /* No AOFRW child? return. */ + if (server.aof_child_pid != -1)return; + + /* Kill AOFRW child, wait for child exit. */ + + serverLog(LL_NOTICE, "Killing running AOF rewrite child: %Id", WIN_PORT_FIX /* %ld -> %Id */ (PORT_LONG) server.aof_child_pid); #ifdef _WIN32 AbortForkOperation(); @@ -246,20 +241,36 @@ void stopAppendOnly(void) { aofRemoveTempFile(server.aof_child_pid); server.aof_child_pid = -1; server.aof_rewrite_time_start = -1; - /* close pipes used for IPC between the two processes. */ + /* Close pipes used for IPC between the two processes. */ aofClosePipes(); - } + +} + +/* Called when the user switches from "appendonly yes" to "appendonly no" + * at runtime using the CONFIG command. */ +void stopAppendOnly(void) { + serverAssert(server.aof_state != AOF_OFF); + flushAppendOnlyFile(1); + aof_fsync(server.aof_fd); + close(server.aof_fd); + + server.aof_fd = -1; + server.aof_selected_db = -1; + server.aof_state = AOF_OFF; + killAppendOnlyChild(); + + } /* Called when the user switches from "appendonly no" to "appendonly yes" * at runtime using the CONFIG command. */ int startAppendOnly(void) { char cwd[MAXPATHLEN]; /* Current working dir path for error messages. */ + int newfd; - server.aof_last_fsync = server.unixtime; - server.aof_fd = open(server.aof_filename, O_WRONLY | O_APPEND | O_CREAT WIN32_ONLY(| _O_BINARY), IF_WIN32(_S_IREAD | _S_IWRITE, 0644)); + newfd = open(server.aof_filename, O_WRONLY | O_APPEND | O_CREAT WIN32_ONLY(| _O_BINARY), IF_WIN32(_S_IREAD | _S_IWRITE, 0644)); serverAssert(server.aof_state == AOF_OFF); - if (server.aof_fd == -1) { + if (newfd == -1) { char *cwdp = IF_WIN32(_getcwd, getcwd)(cwd, MAXPATHLEN); serverLog(LL_WARNING, @@ -273,19 +284,56 @@ int startAppendOnly(void) { if (server.rdb_child_pid != -1) { server.aof_rewrite_scheduled = 1; serverLog(LL_WARNING, "AOF was enabled but there is already a child process saving an RDB file on disk. An AOF background was scheduled to start when possible."); - } - else if (rewriteAppendOnlyFileBackground() == C_ERR) { - close(server.aof_fd); - WIN32_ONLY(server.aof_fd = -1;) - serverLog(LL_WARNING, "Redis needs to enable the AOF but can't trigger a background AOF rewrite operation. Check the above logs for more info about the error."); - return C_ERR; + } else { + /* If there is a pending AOF rewrite, we need to switch it off and + * start a new one: the old one cannot be reused becuase it is not + * accumulating the AOF buffer. */ + if (server.aof_child_pid != -1) { + serverLog(LL_WARNING,"AOF was enabled but there is already an AOF rewriting in background. Stopping background AOF and starting a rewrite now."); + killAppendOnlyChild(); + } + if (rewriteAppendOnlyFileBackground() == C_ERR) { + close(newfd); + serverLog(LL_WARNING, "Redis needs to enable the AOF but can't trigger a background AOF rewrite operation. Check the above logs for more info about the error."); + return C_ERR; + } } /* We correctly switched on AOF, now wait for the rewrite to be complete * in order to append data on disk. */ server.aof_state = AOF_WAIT_REWRITE; + server.aof_last_fsync = server.unixtime; + server.aof_fd = newfd; return C_OK; } +/* This is a wrapper to the write syscall in order to retry on short writes + * or if the syscall gets interrupted. It could look strange that we retry + * on short writes given that we are writing to a block device: normally if + * the first call is short, there is a end-of-space condition, so the next + * is likely to fail. However apparently in modern systems this is no longer + * true, and in general it looks just more resilient to retry the write. If + * there is an actual error condition we'll get it at the next try. */ +ssize_t aofWrite(int fd, const char *buf, size_t len) { + ssize_t nwritten = 0, totwritten = 0; + + while(len) { + nwritten = write(fd, buf, len); + + if (nwritten < 0) { + if (errno == EINTR) { + continue; + } + return totwritten ? totwritten : -1; + } + + len -= nwritten; + buf += nwritten; + totwritten += nwritten; + } + + return totwritten; +} + /* Write the append only file buffer on disk. * * Since we are required to write the AOF before replying to the client, @@ -344,7 +392,7 @@ void flushAppendOnlyFile(int force) { * or alike */ latencyStartMonitor(latency); - nwritten = write(server.aof_fd, server.aof_buf, sdslen(server.aof_buf)); + nwritten = aofWrite(server.aof_fd, server.aof_buf, sdslen(server.aof_buf)); latencyEndMonitor(latency); /* We want to capture different events for delayed writes: * when the delay happens with a pending fsync, or with a saving child @@ -353,11 +401,9 @@ void flushAppendOnlyFile(int force) { * useful for graphing / monitoring purposes. */ if (sync_in_progress) { latencyAddSampleIfNeeded("aof-write-pending-fsync", latency); - } - else if (server.aof_child_pid != -1 || server.rdb_child_pid != -1) { + } else if (server.aof_child_pid != -1 || server.rdb_child_pid != -1) { latencyAddSampleIfNeeded("aof-write-active-child", latency); - } - else { + } else { latencyAddSampleIfNeeded("aof-write-alone", latency); } latencyAddSampleIfNeeded("aof-write", latency); @@ -365,7 +411,7 @@ void flushAppendOnlyFile(int force) { /* We performed the write so reset the postponed flush sentinel to zero. */ server.aof_flush_postponed_start = 0; - if (nwritten != (signed) sdslen(server.aof_buf)) { + if (nwritten != (ssize_t)sdslen(server.aof_buf)) { static time_t last_write_error_log = 0; int can_log = 0; @@ -382,8 +428,7 @@ void flushAppendOnlyFile(int force) { IF_WIN32(wsa_strerror(errno), strerror(errno))); server.aof_last_write_errno = errno; } - } - else { + } else { if (can_log) { serverLog(LL_WARNING, "Short write while writing to " "the AOF file: (nwritten=%lld, " @@ -468,8 +513,7 @@ void flushAppendOnlyFile(int force) { latencyEndMonitor(latency); latencyAddSampleIfNeeded("aof-fsync-always", latency); server.aof_last_fsync = server.unixtime; - } - else if ((server.aof_fsync == AOF_FSYNC_EVERYSEC && + } else if ((server.aof_fsync == AOF_FSYNC_EVERYSEC && server.unixtime > server.aof_last_fsync)) { if (!sync_in_progress) aof_background_fsync(server.aof_fd); server.aof_last_fsync = server.unixtime; @@ -557,8 +601,7 @@ void feedAppendOnlyFile(struct redisCommand *cmd, int dictid, robj **argv, int a cmd->proc == expireatCommand) { /* Translate EXPIRE/PEXPIRE/EXPIREAT into PEXPIREAT */ buf = catAppendOnlyExpireAtCommand(buf, cmd, argv[1], argv[2]); - } - else if (cmd->proc == setexCommand || cmd->proc == psetexCommand) { + } else if (cmd->proc == setexCommand || cmd->proc == psetexCommand) { /* Translate SETEX/PSETEX to SET and PEXPIREAT */ tmpargv[0] = createStringObject("SET", 3); tmpargv[1] = argv[1]; @@ -566,8 +609,7 @@ void feedAppendOnlyFile(struct redisCommand *cmd, int dictid, robj **argv, int a buf = catAppendOnlyGenericCommand(buf, 3, tmpargv); decrRefCount(tmpargv[0]); buf = catAppendOnlyExpireAtCommand(buf, cmd, argv[1], argv[2]); - } - else if (cmd->proc == setCommand && argc > 3) { + } else if (cmd->proc == setCommand && argc > 3) { int i; robj *exarg = NULL, *pxarg = NULL; /* Translate SET [EX seconds][PX milliseconds] to SET and PEXPIREAT */ @@ -583,8 +625,7 @@ void feedAppendOnlyFile(struct redisCommand *cmd, int dictid, robj **argv, int a if (pxarg) buf = catAppendOnlyExpireAtCommand(buf, server.pexpireCommand, argv[1], pxarg); - } - else { + } else { /* All the other commands don't need translation or need the * same translation already operated in the command vector * for the replication itself. */ @@ -703,7 +744,7 @@ int loadAppendOnlyFile(char *filename) { serverLog(LL_NOTICE, "Reading RDB preamble from AOF file..."); if (fseek(fp, 0, SEEK_SET) == -1) goto readerr; rioInitWithFile(&rdb, fp); - if (rdbLoadRio(&rdb, NULL) != C_OK) { + if (rdbLoadRio(&rdb,NULL,1) != C_OK) { serverLog(LL_WARNING, "Error reading the RDB preamble of the AOF file, AOF loading aborted"); goto readerr; } @@ -804,7 +845,7 @@ loaded_ok: /* DB loaded, cleanup and return C_OK to the caller. */ readerr: /* Read error. If feof(fp) is true, fall through to unexpected EOF. */ if (!feof(fp)) { if (fakeClient) freeFakeClient(fakeClient); /* avoid valgrind warning */ - serverLog(LL_WARNING, "Unrecoverable error reading the append only file: %s", IF_WIN32(wsa_strerror(errno), strerror(errno))); + serverLog(LL_WARNING,"Unrecoverable error reading the append only file: %s", strerror(errno)); exit(1); } @@ -1080,6 +1121,22 @@ int rewriteHashObject(rio *r, robj *key, robj *o) { return 1; } +/* Call the module type callback in order to rewrite a data type + * that is exported by a module and is not handled by Redis itself. + * The function returns 0 on error, 1 on success. */ +int rewriteModuleObject(rio *r, robj *key, robj *o) { + RedisModuleIO io; + moduleValue *mv = o->ptr; + moduleType *mt = mv->type; + moduleInitIOContext(io,mt,r); + mt->aof_rewrite(&io,key,mv->value); + if (io.ctx) { + moduleFreeContext(io.ctx); + zfree(io.ctx); + } + return io.error ? 0 : 1; +} + /* This function is called by the child rewriting the AOF file to read * the difference accumulated from the parent into a buffer, that is * concatenated at the end of the rewrite. */ @@ -1738,20 +1795,3 @@ cleanup: if (server.aof_state == AOF_WAIT_REWRITE) server.aof_rewrite_scheduled = 1; } - -void aofProcessDiffRewriteEvents(aeEventLoop* eventLoop) -{ - // only do these checks in the parent process and if an aof rewrite is in progress - if (server.aof_child_pid != -1 && server.aof_pipe_read_ack_from_child != -1) { - //1) check if more data can be written to the child and write it. - // in which case we dont need to send any more diffs to the parent - if (server.aof_stop_sending_diff == 0) { - aofChildWriteDiffData(eventLoop, server.aof_pipe_write_data_to_child, NULL, 0); - } - - //2) check if child has signaled parent to stop sending diffs - if (server.aof_stop_sending_diff == 0) { - aofChildPipeReadable(eventLoop, server.aof_pipe_read_ack_from_child, NULL, 0); - } - } -} \ No newline at end of file diff --git a/src/aof.c.bak b/src/aof.c.bak new file mode 100644 index 00000000..1c5e7283 --- /dev/null +++ b/src/aof.c.bak @@ -0,0 +1,1814 @@ +/* + * Copyright (c) 2009-2012, Salvatore Sanfilippo + * All rights reserved. + * + * Redistribution and use in source and binary forms, with or without + * modification, are permitted provided that the following conditions are met: + * + * * Redistributions of source code must retain the above copyright notice, + * this list of conditions and the following disclaimer. + * * Redistributions in binary form must reproduce the above copyright + * notice, this list of conditions and the following disclaimer in the + * documentation and/or other materials provided with the distribution. + * * Neither the name of Redis nor the names of its contributors may be used + * to endorse or promote products derived from this software without + * specific prior written permission. + * + * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" + * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE + * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE + * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE + * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR + * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF + * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS + * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN + * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) + * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE + * POSSIBILITY OF SUCH DAMAGE. + */ + +#ifdef _WIN32 +#include "Win32_Interop/win32_types.h" +#include "Win32_Interop/Win32_Error.h" +#endif + +#include "server.h" +#include "bio.h" +#include "rio.h" + +#include +#include +#include +#include +#ifndef _WIN32 +#include +#include +#include +#include +#else +#include "Win32_Interop/Win32_QFork.h" +#include +#define MAXPATHLEN 1024 +#endif + +void aofUpdateCurrentSize(void); +void aofClosePipes(void); + +/* ---------------------------------------------------------------------------- + * AOF rewrite buffer implementation. + * + * The following code implement a simple buffer used in order to accumulate + * changes while the background process is rewriting the AOF file. + * + * We only need to append, but can't just use realloc with a large block + * because 'huge' reallocs are not always handled as one could expect + * (via remapping of pages at OS level) but may involve copying data. + * + * For this reason we use a list of blocks, every block is + * AOF_RW_BUF_BLOCK_SIZE bytes. + * ------------------------------------------------------------------------- */ + +#define AOF_RW_BUF_BLOCK_SIZE (1024*1024*10) /* 10 MB per block */ + +typedef struct aofrwblock { + PORT_ULONG used, free; + char buf[AOF_RW_BUF_BLOCK_SIZE]; +} aofrwblock; + +/* This function free the old AOF rewrite buffer if needed, and initialize + * a fresh new one. It tests for server.aof_rewrite_buf_blocks equal to NULL + * so can be used for the first initialization as well. */ +void aofRewriteBufferReset(void) { + if (server.aof_rewrite_buf_blocks) + listRelease(server.aof_rewrite_buf_blocks); + + server.aof_rewrite_buf_blocks = listCreate(); + listSetFreeMethod(server.aof_rewrite_buf_blocks, zfree); +} + +/* Return the current size of the AOF rewrite buffer. */ +PORT_ULONG aofRewriteBufferSize(void) { + listNode *ln; + listIter li; + PORT_ULONG size = 0; + + listRewind(server.aof_rewrite_buf_blocks, &li); + while ((ln = listNext(&li))) { + aofrwblock *block = listNodeValue(ln); + size += block->used; + } + return size; +} + +/* Event handler used to send data to the child process doing the AOF + * rewrite. We send pieces of our AOF differences buffer so that the final + * write when the child finishes the rewrite will be small. */ +void aofChildWriteDiffData(aeEventLoop *el, int fd, void *privdata, int mask) { + listNode *ln; + aofrwblock *block; + ssize_t nwritten; + UNUSED(el); + UNUSED(fd); + UNUSED(privdata); + UNUSED(mask); + + while (1) { + ln = listFirst(server.aof_rewrite_buf_blocks); + block = ln ? ln->value : NULL; + if (server.aof_stop_sending_diff || !block) { +#ifndef _WIN32 + aeDeleteFileEvent(server.el, server.aof_pipe_write_data_to_child, + AE_WRITABLE); +#endif + return; + } + if (block->used > 0) { + nwritten = write(server.aof_pipe_write_data_to_child, + block->buf, block->used); + if (nwritten <= 0) return; + memmove(block->buf, block->buf + nwritten, block->used - nwritten); + block->used -= nwritten; + block->free += nwritten; + } + if (block->used == 0) listDelNode(server.aof_rewrite_buf_blocks, ln); + } +} + +/* Append data to the AOF rewrite buffer, allocating new blocks if needed. */ +void aofRewriteBufferAppend(unsigned char *s, PORT_ULONG len) { + listNode *ln = listLast(server.aof_rewrite_buf_blocks); + aofrwblock *block = ln ? ln->value : NULL; + + while (len) { + /* If we already got at least an allocated block, try appending + * at least some piece into it. */ + if (block) { + PORT_ULONG thislen = (block->free < len) ? block->free : len; + if (thislen) { /* The current block is not already full. */ + memcpy(block->buf + block->used, s, thislen); + block->used += thislen; + block->free -= thislen; + s += thislen; + len -= thislen; + } + } + + if (len) { /* First block to allocate, or need another block. */ + int numblocks; + + block = zmalloc(sizeof(*block)); + block->free = AOF_RW_BUF_BLOCK_SIZE; + block->used = 0; + listAddNodeTail(server.aof_rewrite_buf_blocks, block); + + /* Log every time we cross more 10 or 100 blocks, respectively + * as a notice or warning. */ + numblocks = (int) listLength(server.aof_rewrite_buf_blocks); WIN_PORT_FIX /* cast (int) */ + if (((numblocks + 1) % 10) == 0) { + int level = ((numblocks + 1) % 100) == 0 ? LL_WARNING : + LL_NOTICE; + serverLog(level, "Background AOF buffer size: %Iu MB", WIN_PORT_FIX /* %lu -> %Iu */ + aofRewriteBufferSize() / (1024 * 1024)); + } + } + } +#ifndef _WIN32 + /* Install a file event to send data to the rewrite child if there is + * not one already. */ + if (aeGetFileEvents(server.el, server.aof_pipe_write_data_to_child) == 0) { + aeCreateFileEvent(server.el, server.aof_pipe_write_data_to_child, + AE_WRITABLE, aofChildWriteDiffData, NULL); + } +#endif +} + +/* Write the buffer (possibly composed of multiple blocks) into the specified + * fd. If a short write or any other error happens -1 is returned, + * otherwise the number of bytes written is returned. */ +ssize_t aofRewriteBufferWrite(int fd) { + listNode *ln; + listIter li; + ssize_t count = 0; + + listRewind(server.aof_rewrite_buf_blocks, &li); + while ((ln = listNext(&li))) { + aofrwblock *block = listNodeValue(ln); + ssize_t nwritten; + + if (block->used) { + nwritten = write(fd, block->buf, block->used); + if (nwritten != (ssize_t) block->used) { + if (nwritten == 0) errno = EIO; + return -1; + } + count += nwritten; + } + } + return count; +} + +/* ---------------------------------------------------------------------------- + * AOF file implementation + * ------------------------------------------------------------------------- */ + + /* Starts a background task that performs fsync() against the specified + * file descriptor (the one of the AOF file) in another thread. */ +void aof_background_fsync(int fd) { + bioCreateBackgroundJob(BIO_AOF_FSYNC,(void*)(long)fd,NULL,NULL); +} + +/* Kills an AOFRW child process if exists */ +static void killAppendOnlyChild(void) { + + POSIX_ONLY(int statloc;) + + /* No AOFRW child? return. */ + if (server.aof_child_pid != -1)return; + + /* Kill AOFRW child, wait for child exit. */ + + serverLog(LL_NOTICE, "Killing running AOF rewrite child: %Id", WIN_PORT_FIX /* %ld -> %Id */ + (PORT_LONG) server.aof_child_pid); +#ifdef _WIN32 + AbortForkOperation(); +#else + if (kill(server.aof_child_pid, SIGUSR1) != -1) { + while (wait3(&statloc, 0, NULL) != server.aof_child_pid); + } +#endif + /* reset the buffer accumulating changes while the child saves */ + aofRewriteBufferReset(); + aofRemoveTempFile(server.aof_child_pid); + server.aof_child_pid = -1; + server.aof_rewrite_time_start = -1; + /* Close pipes used for IPC between the two processes. */ + aofClosePipes(); + +} + +/* Called when the user switches from "appendonly yes" to "appendonly no" + * at runtime using the CONFIG command. */ +void stopAppendOnly(void) { + serverAssert(server.aof_state != AOF_OFF); + flushAppendOnlyFile(1); + aof_fsync(server.aof_fd); + close(server.aof_fd); + + server.aof_fd = -1; + server.aof_selected_db = -1; + server.aof_state = AOF_OFF; + killAppendOnlyChild(); + + +} + +/* Called when the user switches from "appendonly no" to "appendonly yes" + * at runtime using the CONFIG command. */ +int startAppendOnly(void) { + char cwd[MAXPATHLEN]; /* Current working dir path for error messages. */ + int newfd; + + newfd = open(server.aof_filename, O_WRONLY | O_APPEND | O_CREAT WIN32_ONLY(| _O_BINARY), IF_WIN32(_S_IREAD | _S_IWRITE, 0644)); + serverAssert(server.aof_state == AOF_OFF); + if (newfd == -1) { + char *cwdp = IF_WIN32(_getcwd, getcwd)(cwd, MAXPATHLEN); + + serverLog(LL_WARNING, + "Redis needs to enable the AOF but can't open the " + "append only file %s (in server root dir %s): %s", + server.aof_filename, + cwdp ? cwdp : "unknown", + IF_WIN32(wsa_strerror(errno), strerror(errno))); + return C_ERR; + } + if (server.rdb_child_pid != -1) { + server.aof_rewrite_scheduled = 1; + serverLog(LL_WARNING, "AOF was enabled but there is already a child process saving an RDB file on disk. An AOF background was scheduled to start when possible."); + } else { + /* If there is a pending AOF rewrite, we need to switch it off and + * start a new one: the old one cannot be reused becuase it is not + * accumulating the AOF buffer. */ + if (server.aof_child_pid != -1) { + serverLog(LL_WARNING,"AOF was enabled but there is already an AOF rewriting in background. Stopping background AOF and starting a rewrite now."); + killAppendOnlyChild(); + } + if (rewriteAppendOnlyFileBackground() == C_ERR) { + close(newfd); + serverLog(LL_WARNING, "Redis needs to enable the AOF but can't trigger a background AOF rewrite operation. Check the above logs for more info about the error."); + return C_ERR; + } + } + /* We correctly switched on AOF, now wait for the rewrite to be complete + * in order to append data on disk. */ + server.aof_state = AOF_WAIT_REWRITE; + server.aof_last_fsync = server.unixtime; + server.aof_fd = newfd; + return C_OK; +} + +/* This is a wrapper to the write syscall in order to retry on short writes + * or if the syscall gets interrupted. It could look strange that we retry + * on short writes given that we are writing to a block device: normally if + * the first call is short, there is a end-of-space condition, so the next + * is likely to fail. However apparently in modern systems this is no longer + * true, and in general it looks just more resilient to retry the write. If + * there is an actual error condition we'll get it at the next try. */ +ssize_t aofWrite(int fd, const char *buf, size_t len) { + ssize_t nwritten = 0, totwritten = 0; + + while(len) { + nwritten = write(fd, buf, len); + + if (nwritten < 0) { + if (errno == EINTR) { + continue; + } + return totwritten ? totwritten : -1; + } + + len -= nwritten; + buf += nwritten; + totwritten += nwritten; + } + + return totwritten; +} + +/* Write the append only file buffer on disk. + * + * Since we are required to write the AOF before replying to the client, + * and the only way the client socket can get a write is entering when the + * the event loop, we accumulate all the AOF writes in a memory + * buffer and write it on disk using this function just before entering + * the event loop again. + * + * About the 'force' argument: + * + * When the fsync policy is set to 'everysec' we may delay the flush if there + * is still an fsync() going on in the background thread, since for instance + * on Linux write(2) will be blocked by the background fsync anyway. + * When this happens we remember that there is some aof buffer to be + * flushed ASAP, and will try to do that in the serverCron() function. + * + * However if force is set to 1 we'll write regardless of the background + * fsync. */ +#define AOF_WRITE_LOG_ERROR_RATE 30 /* Seconds between errors logging. */ +void flushAppendOnlyFile(int force) { + ssize_t nwritten; + int sync_in_progress = 0; + mstime_t latency; + + if (sdslen(server.aof_buf) == 0) return; + + if (server.aof_fsync == AOF_FSYNC_EVERYSEC) + sync_in_progress = bioPendingJobsOfType(BIO_AOF_FSYNC) != 0; + + if (server.aof_fsync == AOF_FSYNC_EVERYSEC && !force) { + /* With this append fsync policy we do background fsyncing. + * If the fsync is still in progress we can try to delay + * the write for a couple of seconds. */ + if (sync_in_progress) { + if (server.aof_flush_postponed_start == 0) { + /* No previous write postponing, remember that we are + * postponing the flush and return. */ + server.aof_flush_postponed_start = server.unixtime; + return; + } + else if (server.unixtime - server.aof_flush_postponed_start < 2) { + /* We were already waiting for fsync to finish, but for less + * than two seconds this is still ok. Postpone again. */ + return; + } + /* Otherwise fall trough, and go write since we can't wait + * over two seconds. */ + server.aof_delayed_fsync++; + serverLog(LL_NOTICE, "Asynchronous AOF fsync is taking too long (disk is busy?). Writing the AOF buffer without waiting for fsync to complete, this may slow down Redis."); + } + } + /* We want to perform a single write. This should be guaranteed atomic + * at least if the filesystem we are writing is a real physical one. + * While this will save us against the server being killed I don't think + * there is much to do about the whole server stopping for power problems + * or alike */ + + latencyStartMonitor(latency); + nwritten = aofWrite(server.aof_fd, server.aof_buf, sdslen(server.aof_buf)); + latencyEndMonitor(latency); + /* We want to capture different events for delayed writes: + * when the delay happens with a pending fsync, or with a saving child + * active, and when the above two conditions are missing. + * We also use an additional event name to save all samples which is + * useful for graphing / monitoring purposes. */ + if (sync_in_progress) { + latencyAddSampleIfNeeded("aof-write-pending-fsync", latency); + } else if (server.aof_child_pid != -1 || server.rdb_child_pid != -1) { + latencyAddSampleIfNeeded("aof-write-active-child", latency); + } else { + latencyAddSampleIfNeeded("aof-write-alone", latency); + } + latencyAddSampleIfNeeded("aof-write", latency); + + /* We performed the write so reset the postponed flush sentinel to zero. */ + server.aof_flush_postponed_start = 0; + + if (nwritten != (ssize_t)sdslen(server.aof_buf)) { + static time_t last_write_error_log = 0; + int can_log = 0; + + /* Limit logging rate to 1 line per AOF_WRITE_LOG_ERROR_RATE seconds. */ + if ((server.unixtime - last_write_error_log) > AOF_WRITE_LOG_ERROR_RATE) { + can_log = 1; + last_write_error_log = server.unixtime; + } + + /* Log the AOF write error and record the error code. */ + if (nwritten == -1) { + if (can_log) { + serverLog(LL_WARNING, "Error writing to the AOF file: %s", + IF_WIN32(wsa_strerror(errno), strerror(errno))); + server.aof_last_write_errno = errno; + } + } else { + if (can_log) { + serverLog(LL_WARNING, "Short write while writing to " + "the AOF file: (nwritten=%lld, " + "expected=%lld)", + (PORT_LONGLONG) nwritten, + (PORT_LONGLONG) sdslen(server.aof_buf)); + } + + if (ftruncate(server.aof_fd, server.aof_current_size) == -1) { + if (can_log) { + serverLog(LL_WARNING, "Could not remove short write " + "from the append-only file. Redis may refuse " + "to load the AOF the next time it starts. " + "ftruncate: %s", IF_WIN32(wsa_strerror(errno), strerror(errno))); + } + } + else { + /* If the ftruncate() succeeded we can set nwritten to + * -1 since there is no longer partial data into the AOF. */ + nwritten = -1; + } + server.aof_last_write_errno = ENOSPC; + } + + /* Handle the AOF write error. */ + if (server.aof_fsync == AOF_FSYNC_ALWAYS) { + /* We can't recover when the fsync policy is ALWAYS since the + * reply for the client is already in the output buffers, and we + * have the contract with the user that on acknowledged write data + * is synced on disk. */ + serverLog(LL_WARNING, "Can't recover from AOF write error when the AOF fsync policy is 'always'. Exiting..."); + exit(1); + } + else { + /* Recover from failed write leaving data into the buffer. However + * set an error to stop accepting writes as PORT_LONG as the error + * condition is not cleared. */ + server.aof_last_write_status = C_ERR; + + /* Trim the sds buffer if there was a partial write, and there + * was no way to undo it with ftruncate(2). */ + if (nwritten > 0) { + server.aof_current_size += nwritten; + sdsrange(server.aof_buf, (int) nwritten, -1); WIN_PORT_FIX /* cast (int) */ + } + return; /* We'll try again on the next call... */ + } + } + else { + /* Successful write(2). If AOF was in error state, restore the + * OK state and log the event. */ + if (server.aof_last_write_status == C_ERR) { + serverLog(LL_WARNING, + "AOF write error looks solved, Redis can write again."); + server.aof_last_write_status = C_OK; + } + } + server.aof_current_size += nwritten; + + /* Re-use AOF buffer when it is small enough. The maximum comes from the + * arena size of 4k minus some overhead (but is otherwise arbitrary). */ + if ((sdslen(server.aof_buf) + sdsavail(server.aof_buf)) < 4000) { + sdsclear(server.aof_buf); + } + else { + sdsfree(server.aof_buf); + server.aof_buf = sdsempty(); + } + + /* Don't fsync if no-appendfsync-on-rewrite is set to yes and there are + * children doing I/O in the background. */ + if (server.aof_no_fsync_on_rewrite && + (server.aof_child_pid != -1 || server.rdb_child_pid != -1)) + return; + + /* Perform the fsync if needed. */ + if (server.aof_fsync == AOF_FSYNC_ALWAYS) { + /* aof_fsync is defined as fdatasync() for Linux in order to avoid + * flushing metadata. */ + latencyStartMonitor(latency); + aof_fsync(server.aof_fd); /* Let's try to get this data on the disk */ + latencyEndMonitor(latency); + latencyAddSampleIfNeeded("aof-fsync-always", latency); + server.aof_last_fsync = server.unixtime; + } else if ((server.aof_fsync == AOF_FSYNC_EVERYSEC && + server.unixtime > server.aof_last_fsync)) { + if (!sync_in_progress) aof_background_fsync(server.aof_fd); + server.aof_last_fsync = server.unixtime; + } +} + +sds catAppendOnlyGenericCommand(sds dst, int argc, robj **argv) { + char buf[32]; + int len, j; + robj *o; + + buf[0] = '*'; + len = 1 + ll2string(buf + 1, sizeof(buf) - 1, argc); + buf[len++] = '\r'; + buf[len++] = '\n'; + dst = sdscatlen(dst, buf, len); + + for (j = 0; j < argc; j++) { + o = getDecodedObject(argv[j]); + buf[0] = '$'; + len = 1 + ll2string(buf + 1, sizeof(buf) - 1, sdslen(o->ptr)); + buf[len++] = '\r'; + buf[len++] = '\n'; + dst = sdscatlen(dst, buf, len); + dst = sdscatlen(dst, o->ptr, sdslen(o->ptr)); + dst = sdscatlen(dst, "\r\n", 2); + decrRefCount(o); + } + return dst; +} + +/* Create the sds representation of an PEXPIREAT command, using + * 'seconds' as time to live and 'cmd' to understand what command + * we are translating into a PEXPIREAT. + * + * This command is used in order to translate EXPIRE and PEXPIRE commands + * into PEXPIREAT command so that we retain precision in the append only + * file, and the time is always absolute and not relative. */ +sds catAppendOnlyExpireAtCommand(sds buf, struct redisCommand *cmd, robj *key, robj *seconds) { + PORT_LONGLONG when; + robj *argv[3]; + + /* Make sure we can use strtoll */ + seconds = getDecodedObject(seconds); + when = strtoll(seconds->ptr, NULL, 10); + /* Convert argument into milliseconds for EXPIRE, SETEX, EXPIREAT */ + if (cmd->proc == expireCommand || cmd->proc == setexCommand || + cmd->proc == expireatCommand) + { + when *= 1000; + } + /* Convert into absolute time for EXPIRE, PEXPIRE, SETEX, PSETEX */ + if (cmd->proc == expireCommand || cmd->proc == pexpireCommand || + cmd->proc == setexCommand || cmd->proc == psetexCommand) + { + when += mstime(); + } + decrRefCount(seconds); + + argv[0] = createStringObject("PEXPIREAT", 9); + argv[1] = key; + argv[2] = createStringObjectFromLongLong(when); + buf = catAppendOnlyGenericCommand(buf, 3, argv); + decrRefCount(argv[0]); + decrRefCount(argv[2]); + return buf; +} + +void feedAppendOnlyFile(struct redisCommand *cmd, int dictid, robj **argv, int argc) { + sds buf = sdsempty(); + robj *tmpargv[3]; + + /* The DB this command was targeting is not the same as the last command + * we appended. To issue a SELECT command is needed. */ + if (dictid != server.aof_selected_db) { + char seldb[64]; + + snprintf(seldb, sizeof(seldb), "%d", dictid); + buf = sdscatprintf(buf, "*2\r\n$6\r\nSELECT\r\n$%Iu\r\n%s\r\n", WIN_PORT_FIX /* %lu -> %Iu */ + (PORT_ULONG)strlen(seldb), seldb); + server.aof_selected_db = dictid; + } + + if (cmd->proc == expireCommand || cmd->proc == pexpireCommand || + cmd->proc == expireatCommand) { + /* Translate EXPIRE/PEXPIRE/EXPIREAT into PEXPIREAT */ + buf = catAppendOnlyExpireAtCommand(buf, cmd, argv[1], argv[2]); + } else if (cmd->proc == setexCommand || cmd->proc == psetexCommand) { + /* Translate SETEX/PSETEX to SET and PEXPIREAT */ + tmpargv[0] = createStringObject("SET", 3); + tmpargv[1] = argv[1]; + tmpargv[2] = argv[3]; + buf = catAppendOnlyGenericCommand(buf, 3, tmpargv); + decrRefCount(tmpargv[0]); + buf = catAppendOnlyExpireAtCommand(buf, cmd, argv[1], argv[2]); + } else if (cmd->proc == setCommand && argc > 3) { + int i; + robj *exarg = NULL, *pxarg = NULL; + /* Translate SET [EX seconds][PX milliseconds] to SET and PEXPIREAT */ + buf = catAppendOnlyGenericCommand(buf, 3, argv); + for (i = 3; i < argc; i++) { + if (!strcasecmp(argv[i]->ptr, "ex")) exarg = argv[i + 1]; + if (!strcasecmp(argv[i]->ptr, "px")) pxarg = argv[i + 1]; + } + serverAssert(!(exarg && pxarg)); + if (exarg) + buf = catAppendOnlyExpireAtCommand(buf, server.expireCommand, argv[1], + exarg); + if (pxarg) + buf = catAppendOnlyExpireAtCommand(buf, server.pexpireCommand, argv[1], + pxarg); + } else { + /* All the other commands don't need translation or need the + * same translation already operated in the command vector + * for the replication itself. */ + buf = catAppendOnlyGenericCommand(buf, argc, argv); + } + + /* Append to the AOF buffer. This will be flushed on disk just before + * of re-entering the event loop, so before the client will get a + * positive reply about the operation performed. */ + if (server.aof_state == AOF_ON) + server.aof_buf = sdscatlen(server.aof_buf, buf, sdslen(buf)); + + /* If a background append only file rewriting is in progress we want to + * accumulate the differences between the child DB and the current one + * in a buffer, so that when the child process will do its work we + * can append the differences to the new append only file. */ + if (server.aof_child_pid != -1) + aofRewriteBufferAppend((unsigned char*) buf, (PORT_ULONG) sdslen(buf)); WIN_PORT_FIX /* cast (PORT_ULONG) */ + + sdsfree(buf); +} + +/* ---------------------------------------------------------------------------- + * AOF loading + * ------------------------------------------------------------------------- */ + + /* In Redis commands are always executed in the context of a client, so in + * order to load the append only file we need to create a fake client. */ +struct client *createFakeClient(void) { + struct client *c = zmalloc(sizeof(*c)); + + selectDb(c, 0); + c->fd = -1; + c->name = NULL; + c->querybuf = sdsempty(); + c->querybuf_peak = 0; + c->argc = 0; + c->argv = NULL; + c->bufpos = 0; + c->flags = 0; + c->btype = BLOCKED_NONE; + /* We set the fake client as a slave waiting for the synchronization + * so that Redis will not try to send replies to this client. */ + c->replstate = SLAVE_STATE_WAIT_BGSAVE_START; + c->reply = listCreate(); + c->reply_bytes = 0; + c->obuf_soft_limit_reached_time = 0; + c->watched_keys = listCreate(); + c->peerid = NULL; + listSetFreeMethod(c->reply, decrRefCountVoid); + listSetDupMethod(c->reply, dupClientReplyValue); + initClientMultiState(c); + return c; +} + +void freeFakeClientArgv(struct client *c) { + int j; + + for (j = 0; j < c->argc; j++) + decrRefCount(c->argv[j]); + zfree(c->argv); +} + +void freeFakeClient(struct client *c) { + sdsfree(c->querybuf); + listRelease(c->reply); + listRelease(c->watched_keys); + freeClientMultiState(c); + zfree(c); +} + +/* Replay the append log file. On success C_OK is returned. On non fatal + * error (the append only file is zero-length) C_ERR is returned. On + * fatal error an error message is logged and the program exists. */ +int loadAppendOnlyFile(char *filename) { + struct client *fakeClient; + FILE *fp = fopen(filename, IF_WIN32("rb", "r")); + struct redis_stat sb; + int old_aof_state = server.aof_state; + PORT_LONG loops = 0; + off_t valid_up_to = 0; /* Offset of latest well-formed command loaded. */ + + if (fp == NULL) { + serverLog(LL_WARNING, "Fatal error: can't open the append log file for reading: %s", strerror(errno)); + exit(1); + } + + /* Handle a zero-length AOF file as a special case. An emtpy AOF file + * is a valid AOF because an empty server with AOF enabled will create + * a zero length file at startup, that will remain like that if no write + * operation is received. */ + if (fp && redis_fstat(fileno(fp), &sb) != -1 && sb.st_size == 0) { + server.aof_current_size = 0; + fclose(fp); + return C_ERR; + } + + /* Temporarily disable AOF, to prevent EXEC from feeding a MULTI + * to the same file we're about to read. */ + server.aof_state = AOF_OFF; + + fakeClient = createFakeClient(); + startLoading(fp); + + /* Check if this AOF file has an RDB preamble. In that case we need to + * load the RDB file and later continue loading the AOF tail. */ + char sig[5]; /* "REDIS" */ + if (fread(sig, 1, 5, fp) != 5 || memcmp(sig, "REDIS", 5) != 0) { + /* No RDB preamble, seek back at 0 offset. */ + if (fseek(fp, 0, SEEK_SET) == -1) goto readerr; + } + else { + /* RDB preamble. Pass loading the RDB functions. */ + rio rdb; + + serverLog(LL_NOTICE, "Reading RDB preamble from AOF file..."); + if (fseek(fp, 0, SEEK_SET) == -1) goto readerr; + rioInitWithFile(&rdb, fp); + if (rdbLoadRio(&rdb,NULL,1) != C_OK) { + serverLog(LL_WARNING, "Error reading the RDB preamble of the AOF file, AOF loading aborted"); + goto readerr; + } + else { + serverLog(LL_NOTICE, "Reading the remaining AOF tail..."); + } + } + + /* Read the actual AOF file, in REPL format, command by command. */ + while (1) { + int argc, j; + PORT_ULONG len; + robj **argv; + char buf[128]; + sds argsds; + struct redisCommand *cmd; + + /* Serve the clients from time to time */ + if (!(loops++ % 1000)) { + loadingProgress((off_t) ftello(fp)); WIN_PORT_FIX /* cast (off_t) */ + processEventsWhileBlocked(); + } + + if (fgets(buf, sizeof(buf), fp) == NULL) { + if (feof(fp)) + break; + else + goto readerr; + } + if (buf[0] != '*') goto fmterr; + if (buf[1] == '\0') goto readerr; + argc = atoi(buf + 1); + if (argc < 1) goto fmterr; + + argv = zmalloc(sizeof(robj*)*argc); + fakeClient->argc = argc; + fakeClient->argv = argv; + + for (j = 0; j < argc; j++) { + if (fgets(buf, sizeof(buf), fp) == NULL) { + fakeClient->argc = j; /* Free up to j-1. */ + freeFakeClientArgv(fakeClient); + goto readerr; + } + if (buf[0] != '$') goto fmterr; + len = strtol(buf + 1, NULL, 10); + argsds = sdsnewlen(NULL, len); + if (len && fread(argsds, len, 1, fp) == 0) { + sdsfree(argsds); + fakeClient->argc = j; /* Free up to j-1. */ + freeFakeClientArgv(fakeClient); + goto readerr; + } + argv[j] = createObject(OBJ_STRING, argsds); + if (fread(buf, 2, 1, fp) == 0) { + fakeClient->argc = j + 1; /* Free up to j. */ + freeFakeClientArgv(fakeClient); + goto readerr; /* discard CRLF */ + } + } + + /* Command lookup */ + cmd = lookupCommand(argv[0]->ptr); + if (!cmd) { + serverLog(LL_WARNING, "Unknown command '%s' reading the append only file", (char*) argv[0]->ptr); + exit(1); + } + + /* Run the command in the context of a fake client */ + fakeClient->cmd = cmd; + cmd->proc(fakeClient); + + /* The fake client should not have a reply */ + serverAssert(fakeClient->bufpos == 0 && listLength(fakeClient->reply) == 0); + /* The fake client should never get blocked */ + serverAssert((fakeClient->flags & CLIENT_BLOCKED) == 0); + + /* Clean up. Command code may have changed argv/argc so we use the + * argv/argc of the client instead of the local variables. */ + freeFakeClientArgv(fakeClient); + fakeClient->cmd = NULL; + if (server.aof_load_truncated) valid_up_to = ftello(fp); + } + + /* This point can only be reached when EOF is reached without errors. + * If the client is in the middle of a MULTI/EXEC, log error and quit. */ + if (fakeClient->flags & CLIENT_MULTI) goto uxeof; + +loaded_ok: /* DB loaded, cleanup and return C_OK to the caller. */ + fclose(fp); + freeFakeClient(fakeClient); + server.aof_state = old_aof_state; + stopLoading(); + aofUpdateCurrentSize(); + server.aof_rewrite_base_size = server.aof_current_size; + return C_OK; + +readerr: /* Read error. If feof(fp) is true, fall through to unexpected EOF. */ + if (!feof(fp)) { + if (fakeClient) freeFakeClient(fakeClient); /* avoid valgrind warning */ + serverLog(LL_WARNING,"Unrecoverable error reading the append only file: %s", strerror(errno)); + exit(1); + } + +uxeof: /* Unexpected AOF end of file. */ + if (server.aof_load_truncated) { + serverLog(LL_WARNING, "!!! Warning: short read while loading the AOF file !!!"); + serverLog(LL_WARNING, "!!! Truncating the AOF at offset %llu !!!", + (PORT_ULONGLONG) valid_up_to); + if (valid_up_to == -1 || truncate(filename, valid_up_to) == -1) { + if (valid_up_to == -1) { + serverLog(LL_WARNING, "Last valid command offset is invalid"); + } + else { + serverLog(LL_WARNING, "Error truncating the AOF file: %s", + IF_WIN32(wsa_strerror(errno), strerror(errno))); + } + } + else { + /* Make sure the AOF file descriptor points to the end of the + * file after the truncate call. */ + if (server.aof_fd != -1 && lseek(server.aof_fd, 0, SEEK_END) == -1) { + serverLog(LL_WARNING, "Can't seek the end of the AOF file: %s", + IF_WIN32(wsa_strerror(errno), strerror(errno))); + } + else { + serverLog(LL_WARNING, + "AOF loaded anyway because aof-load-truncated is enabled"); + goto loaded_ok; + } + } + } + if (fakeClient) freeFakeClient(fakeClient); /* avoid valgrind warning */ + serverLog(LL_WARNING, "Unexpected end of file reading the append only file. You can: 1) Make a backup of your AOF file, then use ./redis-check-aof --fix . 2) Alternatively you can set the 'aof-load-truncated' configuration option to yes and restart the server."); + exit(1); + +fmterr: /* Format error. */ + if (fakeClient) freeFakeClient(fakeClient); /* avoid valgrind warning */ + serverLog(LL_WARNING, "Bad file format reading the append only file: make a backup of your AOF file, then use ./redis-check-aof --fix "); + exit(1); +} + +/* ---------------------------------------------------------------------------- + * AOF rewrite + * ------------------------------------------------------------------------- */ + + /* Delegate writing an object to writing a bulk string or bulk PORT_LONGLONG. + * This is not placed in rio.c since that adds the server.h dependency. */ +int rioWriteBulkObject(rio *r, robj *obj) { + /* Avoid using getDecodedObject to help copy-on-write (we are often + * in a child process when this function is called). */ + if (obj->encoding == OBJ_ENCODING_INT) { + return (int) rioWriteBulkLongLong(r, (PORT_LONG) obj->ptr); WIN_PORT_FIX /* cast (int) */ + } + else if (sdsEncodedObject(obj)) { + return (int) rioWriteBulkString(r, obj->ptr, sdslen(obj->ptr)); WIN_PORT_FIX /* cast (int) */ + } + else { + serverPanic("Unknown string encoding"); + } +} + +/* Emit the commands needed to rebuild a list object. + * The function returns 0 on error, 1 on success. */ +int rewriteListObject(rio *r, robj *key, robj *o) { + PORT_LONGLONG count = 0, items = listTypeLength(o); + + if (o->encoding == OBJ_ENCODING_QUICKLIST) { + quicklist *list = o->ptr; + quicklistIter *li = quicklistGetIterator(list, AL_START_HEAD); + quicklistEntry entry; + + while (quicklistNext(li, &entry)) { + if (count == 0) { + int cmd_items = (items > AOF_REWRITE_ITEMS_PER_CMD) ? + AOF_REWRITE_ITEMS_PER_CMD : items; + if (rioWriteBulkCount(r, '*', 2 + cmd_items) == 0) return 0; + if (rioWriteBulkString(r, "RPUSH", 5) == 0) return 0; + if (rioWriteBulkObject(r, key) == 0) return 0; + } + + if (entry.value) { + if (rioWriteBulkString(r, (char*) entry.value, entry.sz) == 0) return 0; + } + else { + if (rioWriteBulkLongLong(r, entry.longval) == 0) return 0; + } + if (++count == AOF_REWRITE_ITEMS_PER_CMD) count = 0; + items--; + } + quicklistReleaseIterator(li); + } + else { + serverPanic("Unknown list encoding"); + } + return 1; +} + +/* Emit the commands needed to rebuild a set object. + * The function returns 0 on error, 1 on success. */ +int rewriteSetObject(rio *r, robj *key, robj *o) { + PORT_LONGLONG count = 0, items = setTypeSize(o); + + if (o->encoding == OBJ_ENCODING_INTSET) { + int ii = 0; + int64_t llval; + + while (intsetGet(o->ptr, ii++, &llval)) { + if (count == 0) { + int cmd_items = (items > AOF_REWRITE_ITEMS_PER_CMD) ? + AOF_REWRITE_ITEMS_PER_CMD : (int) items; WIN_PORT_FIX /* cast (int) */ + + if (rioWriteBulkCount(r, '*', 2 + cmd_items) == 0) return 0; + if (rioWriteBulkString(r, "SADD", 4) == 0) return 0; + if (rioWriteBulkObject(r, key) == 0) return 0; + } + if (rioWriteBulkLongLong(r, llval) == 0) return 0; + if (++count == AOF_REWRITE_ITEMS_PER_CMD) count = 0; + items--; + } + } + else if (o->encoding == OBJ_ENCODING_HT) { + dictIterator *di = dictGetIterator(o->ptr); + dictEntry *de; + + while ((de = dictNext(di)) != NULL) { + sds ele = dictGetKey(de); + if (count == 0) { + int cmd_items = (items > AOF_REWRITE_ITEMS_PER_CMD) ? + AOF_REWRITE_ITEMS_PER_CMD : (int) items; WIN_PORT_FIX /* cast (int) */ + + if (rioWriteBulkCount(r, '*', 2 + cmd_items) == 0) return 0; + if (rioWriteBulkString(r, "SADD", 4) == 0) return 0; + if (rioWriteBulkObject(r, key) == 0) return 0; + } + if (rioWriteBulkString(r, ele, sdslen(ele)) == 0) return 0; + if (++count == AOF_REWRITE_ITEMS_PER_CMD) count = 0; + items--; + } + dictReleaseIterator(di); + } + else { + serverPanic("Unknown set encoding"); + } + return 1; +} + +/* Emit the commands needed to rebuild a sorted set object. + * The function returns 0 on error, 1 on success. */ +int rewriteSortedSetObject(rio *r, robj *key, robj *o) { + PORT_LONGLONG count = 0, items = zsetLength(o); + + if (o->encoding == OBJ_ENCODING_ZIPLIST) { + unsigned char *zl = o->ptr; + unsigned char *eptr, *sptr; + unsigned char *vstr; + unsigned int vlen; + PORT_LONGLONG vll; + double score; + + eptr = ziplistIndex(zl, 0); + serverAssert(eptr != NULL); + sptr = ziplistNext(zl, eptr); + serverAssert(sptr != NULL); + + while (eptr != NULL) { + serverAssert(ziplistGet(eptr, &vstr, &vlen, &vll)); + score = zzlGetScore(sptr); + + if (count == 0) { + int cmd_items = (items > AOF_REWRITE_ITEMS_PER_CMD) ? + AOF_REWRITE_ITEMS_PER_CMD : (int) items; WIN_PORT_FIX /* cast (int) */ + + if (rioWriteBulkCount(r, '*', 2 + cmd_items * 2) == 0) return 0; + if (rioWriteBulkString(r, "ZADD", 4) == 0) return 0; + if (rioWriteBulkObject(r, key) == 0) return 0; + } + if (rioWriteBulkDouble(r, score) == 0) return 0; + if (vstr != NULL) { + if (rioWriteBulkString(r, (char*) vstr, vlen) == 0) return 0; + } + else { + if (rioWriteBulkLongLong(r, vll) == 0) return 0; + } + zzlNext(zl, &eptr, &sptr); + if (++count == AOF_REWRITE_ITEMS_PER_CMD) count = 0; + items--; + } + } + else if (o->encoding == OBJ_ENCODING_SKIPLIST) { + zset *zs = o->ptr; + dictIterator *di = dictGetIterator(zs->dict); + dictEntry *de; + + while ((de = dictNext(di)) != NULL) { + sds ele = dictGetKey(de); + double *score = dictGetVal(de); + + if (count == 0) { + int cmd_items = (items > AOF_REWRITE_ITEMS_PER_CMD) ? + AOF_REWRITE_ITEMS_PER_CMD : (int) items; WIN_PORT_FIX /* cast (int) */ + + if (rioWriteBulkCount(r, '*', 2 + cmd_items * 2) == 0) return 0; + if (rioWriteBulkString(r, "ZADD", 4) == 0) return 0; + if (rioWriteBulkObject(r, key) == 0) return 0; + } + if (rioWriteBulkDouble(r, *score) == 0) return 0; + if (rioWriteBulkString(r, ele, sdslen(ele)) == 0) return 0; + if (++count == AOF_REWRITE_ITEMS_PER_CMD) count = 0; + items--; + } + dictReleaseIterator(di); + } + else { + serverPanic("Unknown sorted zset encoding"); + } + return 1; +} + +/* Write either the key or the value of the currently selected item of a hash. + * The 'hi' argument passes a valid Redis hash iterator. + * The 'what' filed specifies if to write a key or a value and can be + * either OBJ_HASH_KEY or OBJ_HASH_VALUE. + * + * The function returns 0 on error, non-zero on success. */ +static int rioWriteHashIteratorCursor(rio *r, hashTypeIterator *hi, int what) { + if (hi->encoding == OBJ_ENCODING_ZIPLIST) { + unsigned char *vstr = NULL; + unsigned int vlen = UINT_MAX; + PORT_LONGLONG vll = LLONG_MAX; + + hashTypeCurrentFromZiplist(hi, what, &vstr, &vlen, &vll); + if (vstr) { + return (int) rioWriteBulkString(r, (char*) vstr, vlen); WIN_PORT_FIX /* cast (int) */ + } + else { + return (int) rioWriteBulkLongLong(r, vll); WIN_PORT_FIX /* cast (int) */ + } + } + else if (hi->encoding == OBJ_ENCODING_HT) { + sds value = hashTypeCurrentFromHashTable(hi, what); + return rioWriteBulkString(r, value, sdslen(value)); + } + + serverPanic("Unknown hash encoding"); + return 0; +} + +/* Emit the commands needed to rebuild a hash object. + * The function returns 0 on error, 1 on success. */ +int rewriteHashObject(rio *r, robj *key, robj *o) { + hashTypeIterator *hi; + PORT_LONGLONG count = 0, items = hashTypeLength(o); + + hi = hashTypeInitIterator(o); + while (hashTypeNext(hi) != C_ERR) { + if (count == 0) { + int cmd_items = (int) ((items > AOF_REWRITE_ITEMS_PER_CMD) ? WIN_PORT_FIX /* cast (int) */ + AOF_REWRITE_ITEMS_PER_CMD : items); + + if (rioWriteBulkCount(r, '*', 2 + cmd_items * 2) == 0) return 0; + if (rioWriteBulkString(r, "HMSET", 5) == 0) return 0; + if (rioWriteBulkObject(r, key) == 0) return 0; + } + + if (rioWriteHashIteratorCursor(r, hi, OBJ_HASH_KEY) == 0) return 0; + if (rioWriteHashIteratorCursor(r, hi, OBJ_HASH_VALUE) == 0) return 0; + if (++count == AOF_REWRITE_ITEMS_PER_CMD) count = 0; + items--; + } + + hashTypeReleaseIterator(hi); + + return 1; +} + +/* Call the module type callback in order to rewrite a data type + * that is exported by a module and is not handled by Redis itself. + * The function returns 0 on error, 1 on success. */ +int rewriteModuleObject(rio *r, robj *key, robj *o) { + RedisModuleIO io; + moduleValue *mv = o->ptr; + moduleType *mt = mv->type; + moduleInitIOContext(io,mt,r); + mt->aof_rewrite(&io,key,mv->value); + if (io.ctx) { + moduleFreeContext(io.ctx); + zfree(io.ctx); + } + return io.error ? 0 : 1; +} + +/* This function is called by the child rewriting the AOF file to read + * the difference accumulated from the parent into a buffer, that is + * concatenated at the end of the rewrite. */ +ssize_t aofReadDiffFromParent(void) { + char buf[65536]; /* Default pipe buffer size on most Linux systems. */ + ssize_t nread, total = 0; + + while ((nread = + read(server.aof_pipe_read_data_from_parent, buf, sizeof(buf))) > 0) { + server.aof_child_diff = sdscatlen(server.aof_child_diff, buf, nread); + total += nread; + } + return total; +} + +int rewriteAppendOnlyFileRio(rio *aof) { + dictIterator *di = NULL; + dictEntry *de; + size_t processed = 0; + PORT_LONGLONG now = mstime(); + int j; + + for (j = 0; j < server.dbnum; j++) { + char selectcmd[] = "*2\r\n$6\r\nSELECT\r\n"; + redisDb *db = server.db + j; + dict *d = db->dict; + if (dictSize(d) == 0) continue; + di = dictGetSafeIterator(d); + + /* SELECT the new DB */ + if (rioWrite(aof, selectcmd, sizeof(selectcmd) - 1) == 0) goto werr; + if (rioWriteBulkLongLong(aof, j) == 0) goto werr; + + /* Iterate this DB writing every entry */ + while ((de = dictNext(di)) != NULL) { + sds keystr; + robj key, *o; + PORT_LONGLONG expiretime; + + keystr = dictGetKey(de); + o = dictGetVal(de); + initStaticStringObject(key, keystr); + + expiretime = getExpire(db, &key); + + /* If this key is already expired skip it */ + if (expiretime != -1 && expiretime < now) continue; + + /* Save the key and associated value */ + if (o->type == OBJ_STRING) { + /* Emit a SET command */ + char cmd[] = "*3\r\n$3\r\nSET\r\n"; + if (rioWrite(aof, cmd, sizeof(cmd) - 1) == 0) goto werr; + /* Key and value */ + if (rioWriteBulkObject(aof, &key) == 0) goto werr; + if (rioWriteBulkObject(aof, o) == 0) goto werr; + } + else if (o->type == OBJ_LIST) { + if (rewriteListObject(aof, &key, o) == 0) goto werr; + } + else if (o->type == OBJ_SET) { + if (rewriteSetObject(aof, &key, o) == 0) goto werr; + } + else if (o->type == OBJ_ZSET) { + if (rewriteSortedSetObject(aof, &key, o) == 0) goto werr; + } + else if (o->type == OBJ_HASH) { + if (rewriteHashObject(aof, &key, o) == 0) goto werr; + } + /* TODO: [tporadowski] modules support + else if (o->type == OBJ_MODULE) { + if (rewriteModuleObject(aof, &key, o) == 0) goto werr; + } + */ + else { + serverPanic("Unknown object type"); + } + /* Save the expire time */ + if (expiretime != -1) { + char cmd[] = "*3\r\n$9\r\nPEXPIREAT\r\n"; + if (rioWrite(aof, cmd, sizeof(cmd) - 1) == 0) goto werr; + if (rioWriteBulkObject(aof, &key) == 0) goto werr; + if (rioWriteBulkLongLong(aof, expiretime) == 0) goto werr; + } + /* Read some diff from the parent process from time to time. */ + if (aof->processed_bytes > processed + AOF_READ_DIFF_INTERVAL_BYTES) { + processed = aof->processed_bytes; + aofReadDiffFromParent(); + } + } + dictReleaseIterator(di); + di = NULL; + } + return C_OK; + +werr: + if (di) dictReleaseIterator(di); + return C_ERR; +} + +/* Write a sequence of commands able to fully rebuild the dataset into + * "filename". Used both by REWRITEAOF and BGREWRITEAOF. + * + * In order to minimize the number of commands needed in the rewritten + * log Redis uses variadic commands when possible, such as RPUSH, SADD + * and ZADD. However at max AOF_REWRITE_ITEMS_PER_CMD items per time + * are inserted using a single command. */ +int rewriteAppendOnlyFile(char *filename) { + rio aof; + FILE *fp; + char tmpfile[256]; + char byte; + + /* Note that we have to use a different temp name here compared to the + * one used by rewriteAppendOnlyFileBackground() function. */ + snprintf(tmpfile, 256, "temp-rewriteaof-%d.aof", (int) getpid()); + fp = fopen(tmpfile, IF_WIN32("wb", "w")); + if (!fp) { + serverLog(LL_WARNING, "Opening the temp file for AOF rewrite in rewriteAppendOnlyFile(): %s", IF_WIN32(wsa_strerror(errno), strerror(errno))); + return C_ERR; + } + + server.aof_child_diff = sdsempty(); + rioInitWithFile(&aof, fp); + + if (server.aof_rewrite_incremental_fsync) + rioSetAutoSync(&aof, AOF_AUTOSYNC_BYTES); + + if (server.aof_use_rdb_preamble) { + int error; + if (rdbSaveRio(&aof, &error, RDB_SAVE_AOF_PREAMBLE, NULL) == C_ERR) { + errno = error; + goto werr; + } + } + else { + if (rewriteAppendOnlyFileRio(&aof) == C_ERR) goto werr; + } + + /* Do an initial slow fsync here while the parent is still sending + * data, in order to make the next final fsync faster. */ + if (fflush(fp) == EOF) goto werr; + if (fsync(fileno(fp)) == -1) goto werr; + + /* Read again a few times to get more data from the parent. + * We can't read forever (the server may receive data from clients + * faster than it is able to send data to the child), so we try to read + * some more data in a loop as soon as there is a good chance more data + * will come. If it looks like we are wasting time, we abort (this + * happens after 20 ms without new data). */ + int nodata = 0; + mstime_t start = mstime(); + while (mstime() - start < 1000 && nodata < 20) { + if (aeWait(server.aof_pipe_read_data_from_parent, AE_READABLE, 1) <= 0) + { + nodata++; + continue; + } + nodata = 0; /* Start counting from zero, we stop on N *contiguous* + timeouts. */ + aofReadDiffFromParent(); + } + + /* Ask the master to stop sending diffs. */ + if (write(server.aof_pipe_write_ack_to_parent, "!", 1) != 1) goto werr; +#ifndef _WIN32 + if (anetNonBlock(NULL, server.aof_pipe_read_ack_from_parent) != ANET_OK) + goto werr; +#endif + /* We read the ACK from the server using a 10 seconds timeout. Normally + * it should reply ASAP, but just in case we lose its reply, we are sure + * the child will eventually get terminated. */ + if (syncRead(server.aof_pipe_read_ack_from_parent, &byte, 1, 5000) != 1 || + byte != '!') goto werr; + serverLog(LL_NOTICE, "Parent agreed to stop sending diffs. Finalizing AOF..."); + + /* Read the final diff if any. */ + aofReadDiffFromParent(); + + /* Write the received diff to the file. */ + serverLog(LL_NOTICE, + "Concatenating %.2f MB of AOF diff received from parent.", + (double) sdslen(server.aof_child_diff) / (1024 * 1024)); + if (rioWrite(&aof, server.aof_child_diff, sdslen(server.aof_child_diff)) == 0) + goto werr; + + /* Make sure data will not remain on the OS's output buffers */ + if (fflush(fp) == EOF) goto werr; + if (fsync(fileno(fp)) == -1) goto werr; + if (fclose(fp) == EOF) goto werr; + + /* Use RENAME to make sure the DB file is changed atomically only + * if the generate DB file is ok. */ + if (rename(tmpfile, filename) == -1) { + serverLog(LL_WARNING, "Error moving temp append only file on the final destination: %s", IF_WIN32(wsa_strerror(errno), strerror(errno))); + unlink(tmpfile); + return C_ERR; + } + serverLog(LL_NOTICE, "SYNC append only file rewrite performed"); + return C_OK; + +werr: + serverLog(LL_WARNING, "Write error writing append only file on disk: %s", IF_WIN32(wsa_strerror(errno), strerror(errno))); + fclose(fp); + unlink(tmpfile); + return C_ERR; +} + +/* ---------------------------------------------------------------------------- + * AOF rewrite pipes for IPC + * -------------------------------------------------------------------------- */ + + /* This event handler is called when the AOF rewriting child sends us a + * single '!' char to signal we should stop sending buffer diffs. The + * parent sends a '!' as well to acknowledge. */ +void aofChildPipeReadable(aeEventLoop *el, int fd, void *privdata, int mask) { + char byte; + UNUSED(el); + UNUSED(privdata); + UNUSED(mask); + + if (read(fd, &byte, 1) == 1 && byte == '!') { + serverLog(LL_NOTICE, "AOF rewrite child asks to stop sending diffs."); + server.aof_stop_sending_diff = 1; + if (write(server.aof_pipe_write_ack_to_child, "!", 1) != 1) { + /* If we can't send the ack, inform the user, but don't try again + * since in the other side the children will use a timeout if the + * kernel can't buffer our write, or, the children was + * terminated. */ + serverLog(LL_WARNING, "Can't send ACK to AOF child: %s", + IF_WIN32(wsa_strerror(errno), strerror(errno))); + } + } +#ifndef _WIN32 + /* Remove the handler since this can be called only one time during a + * rewrite. */ + aeDeleteFileEvent(server.el, server.aof_pipe_read_ack_from_child, AE_READABLE); +#endif +} + +/* Create the pipes used for parent - child process IPC during rewrite. + * We have a data pipe used to send AOF incremental diffs to the child, + * and two other pipes used by the children to signal it finished with + * the rewrite so no more data should be written, and another for the + * parent to acknowledge it understood this new condition. */ +int aofCreatePipes(void) { + int fds[6] = { -1, -1, -1, -1, -1, -1 }; + int j; + + if (pipe(fds) == -1) goto error; /* parent -> children data. */ + if (pipe(fds + 2) == -1) goto error; /* children -> parent ack. */ + if (pipe(fds + 4) == -1) goto error; /* children -> parent ack. */ + /* Parent -> children data is non blocking. */ +#ifndef WIN32 + if (anetNonBlock(NULL, fds[0]) != ANET_OK) goto error; + if (anetNonBlock(NULL, fds[1]) != ANET_OK) goto error; + if (aeCreateFileEvent(server.el, fds[2], AE_READABLE, aofChildPipeReadable, NULL) == AE_ERR) goto error; +#else + if (FDAPI_PipeSetNonBlock(fds[0], 1) != 0) goto error; + if (FDAPI_PipeSetNonBlock(fds[1], 1) != 0) goto error; + if (FDAPI_PipeSetNonBlock(fds[2], 1) != 0) goto error; +#endif + + server.aof_pipe_write_data_to_child = fds[1]; + server.aof_pipe_read_data_from_parent = fds[0]; + server.aof_pipe_write_ack_to_parent = fds[3]; + server.aof_pipe_read_ack_from_child = fds[2]; + server.aof_pipe_write_ack_to_child = fds[5]; + server.aof_pipe_read_ack_from_parent = fds[4]; + server.aof_stop_sending_diff = 0; + return C_OK; + +error: + serverLog(LL_WARNING, "Error opening /setting AOF rewrite IPC pipes: %s", + IF_WIN32(wsa_strerror(errno), strerror(errno))); + for (j = 0; j < 6; j++) if (fds[j] != -1) close(fds[j]); + return C_ERR; +} + +void aofClosePipes(void) { + aeDeleteFileEvent(server.el, server.aof_pipe_read_ack_from_child, AE_READABLE); + aeDeleteFileEvent(server.el, server.aof_pipe_write_data_to_child, AE_WRITABLE); + close(server.aof_pipe_write_data_to_child); + close(server.aof_pipe_read_data_from_parent); + close(server.aof_pipe_write_ack_to_parent); + close(server.aof_pipe_read_ack_from_child); + close(server.aof_pipe_write_ack_to_child); + close(server.aof_pipe_read_ack_from_parent); +} + +/* ---------------------------------------------------------------------------- + * AOF background rewrite + * ------------------------------------------------------------------------- */ + + /* This is how rewriting of the append only file in background works: + * + * 1) The user calls BGREWRITEAOF + * 2) Redis calls this function, that forks(): + * 2a) the child rewrite the append only file in a temp file. + * 2b) the parent accumulates differences in server.aof_rewrite_buf. + * 3) When the child finished '2a' exists. + * 4) The parent will trap the exit code, if it's OK, will append the + * data accumulated into server.aof_rewrite_buf into the temp file, and + * finally will rename(2) the temp file in the actual file name. + * The the new file is reopened as the new append only file. Profit! + */ +int rewriteAppendOnlyFileBackground(void) { + pid_t childpid; + PORT_LONGLONG start; + + if (server.aof_child_pid != -1 || server.rdb_child_pid != -1) return C_ERR; + if (aofCreatePipes() != C_OK) return C_ERR; + openChildInfoPipe(); + start = ustime(); + +#ifndef _WIN32 + if ((childpid = fork()) == 0) { +#endif + char tmpfile[256]; + +#ifndef _WIN32 + /* Child */ + closeListeningSockets(0); + redisSetProcTitle("redis-aof-rewrite"); +#endif + snprintf(tmpfile, 256, "temp-rewriteaof-bg-%d.aof", (int) getpid()); +#ifdef _WIN32 + childpid = BeginForkOperation_Aof(server.aof_pipe_write_ack_to_parent, + server.aof_pipe_read_ack_from_parent, + server.aof_pipe_read_data_from_parent, + tmpfile, + &server, + sizeof(server), + dictGetHashFunctionSeed()); +#else + if (rewriteAppendOnlyFile(tmpfile) == C_OK) { + size_t private_dirty = zmalloc_get_private_dirty(-1); + + if (private_dirty) { + serverLog(LL_NOTICE, + "AOF rewrite: %Iu MB of memory used by copy-on-write", WIN_PORT_FIX /* %lu -> %Iu */ + private_dirty / (1024 * 1024)); + } + + server.child_info_data.cow_size = private_dirty; + sendChildInfo(CHILD_INFO_TYPE_AOF); + exitFromChild(0); + } + else { + exitFromChild(1); + } + } + else { +#endif + /* Parent */ + server.stat_fork_time = ustime() - start; + server.stat_fork_rate = (double) zmalloc_used_memory() * 1000000 / server.stat_fork_time / (1024 * 1024 * 1024); /* GB per second. */ + latencyAddSampleIfNeeded("fork", server.stat_fork_time / 1000); + if (childpid == -1) { + closeChildInfoPipe(); + serverLog(LL_WARNING, + "Can't rewrite append only file in background: fork: %s", + IF_WIN32(wsa_strerror(errno), strerror(errno))); + aofClosePipes(); + return C_ERR; + } + serverLog(LL_NOTICE, + "Background append only file rewriting started by pid %d", childpid); + server.aof_rewrite_scheduled = 0; + server.aof_rewrite_time_start = time(NULL); + server.aof_child_pid = childpid; + updateDictResizePolicy(); + /* We set appendseldb to -1 in order to force the next call to the + * feedAppendOnlyFile() to issue a SELECT command, so the differences + * accumulated by the parent into server.aof_rewrite_buf will start + * with a SELECT statement and it will be safe to merge. */ + server.aof_selected_db = -1; + replicationScriptCacheFlush(); + return C_OK; +#ifndef _WIN32 + } +#endif + return C_OK; /* unreached */ +} + +void bgrewriteaofCommand(client *c) { + if (server.aof_child_pid != -1) { + addReplyError(c, "Background append only file rewriting already in progress"); + } + else if (server.rdb_child_pid != -1) { + server.aof_rewrite_scheduled = 1; + addReplyStatus(c, "Background append only file rewriting scheduled"); + } + else if (rewriteAppendOnlyFileBackground() == C_OK) { + addReplyStatus(c, "Background append only file rewriting started"); + } + else { + addReply(c, shared.err); + } +} + +void aofRemoveTempFile(pid_t childpid) { + char tmpfile[256]; + + snprintf(tmpfile, 256, "temp-rewriteaof-bg-%d.aof", (int) childpid); + unlink(tmpfile); +} + +/* Update the server.aof_current_size field explicitly using stat(2) + * to check the size of the file. This is useful after a rewrite or after + * a restart, normally the size is updated just adding the write length + * to the current length, that is much faster. */ +void aofUpdateCurrentSize(void) { + struct redis_stat sb; + mstime_t latency; + +#ifdef _WIN32 + if (server.aof_fd == -1) { + serverLog(LL_NOTICE, "Unable to check the AOF length: %s", "appendfd is -1"); + return; + } +#endif + + latencyStartMonitor(latency); + if (redis_fstat(server.aof_fd, &sb) == -1) { + serverLog(LL_WARNING, "Unable to obtain the AOF file length. stat: %s", + IF_WIN32(wsa_strerror(errno), strerror(errno))); + } + else { + server.aof_current_size = sb.st_size; + } + latencyEndMonitor(latency); + latencyAddSampleIfNeeded("aof-fstat", latency); +} + +/* A background append only file rewriting (BGREWRITEAOF) terminated its work. + * Handle this. */ +void backgroundRewriteDoneHandler(int exitcode, int bysignal) { + if (!bysignal && exitcode == 0) { + int newfd, oldfd; + char tmpfile[256]; + PORT_LONGLONG now = ustime(); + mstime_t latency; + + serverLog(LL_NOTICE, + "Background AOF rewrite terminated with success"); + + /* Flush the differences accumulated by the parent to the + * rewritten AOF. */ + latencyStartMonitor(latency); +#ifdef _WIN32 + snprintf(tmpfile, 256, "temp-rewriteaof-bg-%d.aof", + getpid()); + newfd = open(tmpfile, O_WRONLY | O_APPEND | O_CREAT | _O_BINARY, _S_IREAD | _S_IWRITE); +#else + snprintf(tmpfile, 256, "temp-rewriteaof-bg-%d.aof", + (int) server.aof_child_pid); + newfd = open(tmpfile, O_WRONLY | O_APPEND); +#endif + if (newfd == -1) { + serverLog(LL_WARNING, + "Unable to open the temporary AOF produced by the child: %s", IF_WIN32(wsa_strerror(errno), strerror(errno))); + goto cleanup; + } + + if (aofRewriteBufferWrite(newfd) == -1) { + serverLog(LL_WARNING, + "Error trying to flush the parent diff to the rewritten AOF: %s", IF_WIN32(wsa_strerror(errno), strerror(errno))); + close(newfd); + goto cleanup; + } + latencyEndMonitor(latency); + latencyAddSampleIfNeeded("aof-rewrite-diff-write", latency); + + serverLog(LL_NOTICE, + "Residual parent diff successfully flushed to the rewritten AOF (%.2f MB)", (double) aofRewriteBufferSize() / (1024 * 1024)); + + /* The only remaining thing to do is to rename the temporary file to + * the configured file and switch the file descriptor used to do AOF + * writes. We don't want close(2) or rename(2) calls to block the + * server on old file deletion. + * + * There are two possible scenarios: + * + * 1) AOF is DISABLED and this was a one time rewrite. The temporary + * file will be renamed to the configured file. When this file already + * exists, it will be unlinked, which may block the server. + * + * 2) AOF is ENABLED and the rewritten AOF will immediately start + * receiving writes. After the temporary file is renamed to the + * configured file, the original AOF file descriptor will be closed. + * Since this will be the last reference to that file, closing it + * causes the underlying file to be unlinked, which may block the + * server. + * + * To mitigate the blocking effect of the unlink operation (either + * caused by rename(2) in scenario 1, or by close(2) in scenario 2), we + * use a background thread to take care of this. First, we + * make scenario 1 identical to scenario 2 by opening the target file + * when it exists. The unlink operation after the rename(2) will then + * be executed upon calling close(2) for its descriptor. Everything to + * guarantee atomicity for this switch has already happened by then, so + * we don't care what the outcome or duration of that close operation + * is, as PORT_LONG as the file descriptor is released again. */ +#ifdef _WIN32 + // On Windows, if AOF is disabled and an AOF file already exists, + // we don't need to open it since it will have to be closed anyway + // before renaming the temp AOF. + oldfd = -1; /* We'll set this to the current AOF filedes later. */ + + char tmpfile_win_old[256]; + snprintf(tmpfile_win_old, 256, "temp-rewriteaof-old-%d.aof", + (int) server.aof_child_pid); + + if (server.aof_fd != -1) { + // AOF enabled, close the existing AOF file + close(server.aof_fd); + // Now rename the existing AOF file to allow the new file to be renamed + if (rename(server.aof_filename, tmpfile_win_old) == -1) { + serverLog(LL_WARNING, + "Error trying to rename the existing AOF to old tempfile: %s", IF_WIN32(wsa_strerror(errno), strerror(errno))); + // Let's clean the Windows-specific temp file here + unlink(tmpfile_win_old); + goto cleanup; + } + } + latencyStartMonitor(latency); + // Close the temp AOF file before renaming it + close(newfd); + if (rename(tmpfile, server.aof_filename) == -1) { + serverLog(LL_WARNING, + "Error trying to rename the temporary AOF file %s into %s: %s", + tmpfile, + server.aof_filename, + IF_WIN32(wsa_strerror(errno), strerror(errno))); + if (server.aof_fd != -1) { + if (rename(tmpfile_win_old, server.aof_filename) == -1) { + serverLog(LL_WARNING, + "Error trying to rename the old tempfile %s into the existing AOF file %s: %s", + tmpfile, + server.aof_filename, + IF_WIN32(wsa_strerror(errno), strerror(errno))); + serverLog(LL_WARNING, + "Error trying to rename the existing AOF from old tempfile: %s", IF_WIN32(wsa_strerror(errno), strerror(errno))); + // The Windows-specific temp file couldn't be renamed to + // the configured AOF file, that should never happen but + // if it happens we leave the file behind in case the user + // needs it + } + } + goto cleanup; + } + /* now open the files again with new names */ + newfd = open(server.aof_filename, O_WRONLY | O_APPEND | _O_BINARY, 0); + if (newfd == -1) { + /* Windows fix: More info */ + serverLog(LL_WARNING, "Not able to reopen the temporary AOF file after rename"); + goto cleanup; + } + if (server.aof_fd != -1) { + server.aof_fd = open( + tmpfile_win_old, + O_WRONLY | O_APPEND | O_CREAT | _O_BINARY | _O_TEMPORARY, // _O_TEMPORARY forces delete on close flag in CreateFile call. File will be deleted in BIO_CLOSE_FILE job. + 0644); + } +#else + if (server.aof_fd == -1) { + /* AOF disabled */ + + /* Don't care if this fails: oldfd will be -1 and we handle that. + * One notable case of -1 return is if the old file does + * not exist. */ + oldfd = open(server.aof_filename, O_RDONLY | O_NONBLOCK); + } + else { + /* AOF enabled */ + oldfd = -1; /* We'll set this to the current AOF filedes later. */ + } + + /* Rename the temporary file. This will not unlink the target file if + * it exists, because we reference it with "oldfd". */ + latencyStartMonitor(latency); + if (rename(tmpfile, server.aof_filename) == -1) { + serverLog(LL_WARNING, + "Error trying to rename the temporary AOF file %s into %s: %s", + tmpfile, + server.aof_filename, + strerror(errno)); + close(newfd); + if (oldfd != -1) close(oldfd); + goto cleanup; + } +#endif + latencyEndMonitor(latency); + latencyAddSampleIfNeeded("aof-rename", latency); + + if (server.aof_fd == -1) { + /* AOF disabled, we don't need to set the AOF file descriptor + * to this new file, so we can close it. */ + close(newfd); + } + else { + /* AOF enabled, replace the old fd with the new one. */ + oldfd = server.aof_fd; + server.aof_fd = newfd; + if (server.aof_fsync == AOF_FSYNC_ALWAYS) + aof_fsync(newfd); + else if (server.aof_fsync == AOF_FSYNC_EVERYSEC) + aof_background_fsync(newfd); + server.aof_selected_db = -1; /* Make sure SELECT is re-issued */ + aofUpdateCurrentSize(); + server.aof_rewrite_base_size = server.aof_current_size; + + /* Clear regular AOF buffer since its contents was just written to + * the new AOF from the background rewrite buffer. */ + sdsfree(server.aof_buf); + server.aof_buf = sdsempty(); + } + + server.aof_lastbgrewrite_status = C_OK; + + serverLog(LL_NOTICE, "Background AOF rewrite finished successfully"); + /* Change state from WAIT_REWRITE to ON if needed */ + if (server.aof_state == AOF_WAIT_REWRITE) + server.aof_state = AOF_ON; + + /* Asynchronously close the overwritten AOF. */ + if (oldfd != -1) bioCreateBackgroundJob(BIO_CLOSE_FILE, (void*) (PORT_LONG) oldfd, NULL, NULL); + + serverLog(LL_VERBOSE, + "Background AOF rewrite signal handler took %lldus", ustime() - now); + } + else if (!bysignal && exitcode != 0) { + /* SIGUSR1 is whitelisted, so we have a way to kill a child without + * tirggering an error conditon. */ + if (bysignal != SIGUSR1) + server.aof_lastbgrewrite_status = C_ERR; + serverLog(LL_WARNING, + "Background AOF rewrite terminated with error"); + } + else { + server.aof_lastbgrewrite_status = C_ERR; + + serverLog(LL_WARNING, + "Background AOF rewrite terminated by signal %d", bysignal); + } + +cleanup: + aofClosePipes(); + aofRewriteBufferReset(); + aofRemoveTempFile(IF_WIN32(getpid(), server.aof_child_pid)); + server.aof_child_pid = -1; + server.aof_rewrite_time_last = time(NULL) - server.aof_rewrite_time_start; + server.aof_rewrite_time_start = -1; + /* Schedule a new rewrite if we are waiting for it to switch the AOF ON. */ + if (server.aof_state == AOF_WAIT_REWRITE) + server.aof_rewrite_scheduled = 1; +} + +void aofProcessDiffRewriteEvents(aeEventLoop* eventLoop) +{ + // only do these checks in the parent process and if an aof rewrite is in progress + if (server.aof_child_pid != -1 && server.aof_pipe_read_ack_from_child != -1) { + //1) check if more data can be written to the child and write it. + // in which case we dont need to send any more diffs to the parent + if (server.aof_stop_sending_diff == 0) { + aofChildWriteDiffData(eventLoop, server.aof_pipe_write_data_to_child, NULL, 0); + } + + //2) check if child has signaled parent to stop sending diffs + if (server.aof_stop_sending_diff == 0) { + aofChildPipeReadable(eventLoop, server.aof_pipe_read_ack_from_child, NULL, 0); + } + } +} \ No newline at end of file diff --git a/src/bitops.c b/src/bitops.c index 9ebc549e..68eecf6d 100644 --- a/src/bitops.c +++ b/src/bitops.c @@ -34,7 +34,7 @@ * Helpers and low level bit functions. * -------------------------------------------------------------------------- */ -/* Count number of bits set in the binary array pointed by 's' and PORT_LONG +/* Count number of bits set in the binary array pointed by 's' and long * 'count' bytes. The implementation of this function is required to * work with a input string length up to 512 MB. */ size_t redisPopcount(void *s, PORT_LONG count) { @@ -92,7 +92,7 @@ size_t redisPopcount(void *s, PORT_LONG count) { } /* Return the position of the first bit set to one (if 'bit' is 1) or - * zero (if 'bit' is 0) in the bitmap starting at 's' and PORT_LONG 'count' bytes. + * zero (if 'bit' is 0) in the bitmap starting at 's' and long 'count' bytes. * * The function is guaranteed to return a value >= 0 if 'bit' is 0 since if * no zero bit is found, it returns count*8 assuming the string is zero @@ -167,7 +167,7 @@ PORT_LONG redisBitpos(void *s, PORT_ULONG count, int bit) { /* Last word left, scan bit by bit. The first thing we need is to * have a single "1" set in the most significant position in an - * PORT_ULONG. We don't know the size of the PORT_LONG so we use a + * unsigned long. We don't know the size of the long so we use a * simple trick. */ one = PORT_ULONG_MAX; /* All bits set to 1.*/ one >>= 1; /* All bits set to 1 but the MSB. */ diff --git a/src/bitops.c.bak b/src/bitops.c.bak new file mode 100644 index 00000000..9ebc549e --- /dev/null +++ b/src/bitops.c.bak @@ -0,0 +1,1125 @@ +/* Bit operations. + * + * Copyright (c) 2009-2012, Salvatore Sanfilippo + * All rights reserved. + * + * Redistribution and use in source and binary forms, with or without + * modification, are permitted provided that the following conditions are met: + * + * * Redistributions of source code must retain the above copyright notice, + * this list of conditions and the following disclaimer. + * * Redistributions in binary form must reproduce the above copyright + * notice, this list of conditions and the following disclaimer in the + * documentation and/or other materials provided with the distribution. + * * Neither the name of Redis nor the names of its contributors may be used + * to endorse or promote products derived from this software without + * specific prior written permission. + * + * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" + * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE + * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE + * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE + * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR + * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF + * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS + * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN + * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) + * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE + * POSSIBILITY OF SUCH DAMAGE. + */ + +#include "server.h" + +/* ----------------------------------------------------------------------------- + * Helpers and low level bit functions. + * -------------------------------------------------------------------------- */ + +/* Count number of bits set in the binary array pointed by 's' and PORT_LONG + * 'count' bytes. The implementation of this function is required to + * work with a input string length up to 512 MB. */ +size_t redisPopcount(void *s, PORT_LONG count) { + size_t bits = 0; + unsigned char *p = s; + uint32_t *p4; + static const unsigned char bitsinbyte[256] = {0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,1,2,2,3,2,3,3,4,2,3,3,4,3,4,4,5,1,2,2,3,2,3,3,4,2,3,3,4,3,4,4,5,2,3,3,4,3,4,4,5,3,4,4,5,4,5,5,6,1,2,2,3,2,3,3,4,2,3,3,4,3,4,4,5,2,3,3,4,3,4,4,5,3,4,4,5,4,5,5,6,2,3,3,4,3,4,4,5,3,4,4,5,4,5,5,6,3,4,4,5,4,5,5,6,4,5,5,6,5,6,6,7,1,2,2,3,2,3,3,4,2,3,3,4,3,4,4,5,2,3,3,4,3,4,4,5,3,4,4,5,4,5,5,6,2,3,3,4,3,4,4,5,3,4,4,5,4,5,5,6,3,4,4,5,4,5,5,6,4,5,5,6,5,6,6,7,2,3,3,4,3,4,4,5,3,4,4,5,4,5,5,6,3,4,4,5,4,5,5,6,4,5,5,6,5,6,6,7,3,4,4,5,4,5,5,6,4,5,5,6,5,6,6,7,4,5,5,6,5,6,6,7,5,6,6,7,6,7,7,8}; + + /* Count initial bytes not aligned to 32 bit. */ + while((PORT_ULONG)p & 3 && count) { + bits += bitsinbyte[*p++]; + count--; + } + + /* Count bits 28 bytes at a time */ + p4 = (uint32_t*)p; + while(count>=28) { + uint32_t aux1, aux2, aux3, aux4, aux5, aux6, aux7; + + aux1 = *p4++; + aux2 = *p4++; + aux3 = *p4++; + aux4 = *p4++; + aux5 = *p4++; + aux6 = *p4++; + aux7 = *p4++; + count -= 28; + + aux1 = aux1 - ((aux1 >> 1) & 0x55555555); + aux1 = (aux1 & 0x33333333) + ((aux1 >> 2) & 0x33333333); + aux2 = aux2 - ((aux2 >> 1) & 0x55555555); + aux2 = (aux2 & 0x33333333) + ((aux2 >> 2) & 0x33333333); + aux3 = aux3 - ((aux3 >> 1) & 0x55555555); + aux3 = (aux3 & 0x33333333) + ((aux3 >> 2) & 0x33333333); + aux4 = aux4 - ((aux4 >> 1) & 0x55555555); + aux4 = (aux4 & 0x33333333) + ((aux4 >> 2) & 0x33333333); + aux5 = aux5 - ((aux5 >> 1) & 0x55555555); + aux5 = (aux5 & 0x33333333) + ((aux5 >> 2) & 0x33333333); + aux6 = aux6 - ((aux6 >> 1) & 0x55555555); + aux6 = (aux6 & 0x33333333) + ((aux6 >> 2) & 0x33333333); + aux7 = aux7 - ((aux7 >> 1) & 0x55555555); + aux7 = (aux7 & 0x33333333) + ((aux7 >> 2) & 0x33333333); + bits += ((((aux1 + (aux1 >> 4)) & 0x0F0F0F0F) + + ((aux2 + (aux2 >> 4)) & 0x0F0F0F0F) + + ((aux3 + (aux3 >> 4)) & 0x0F0F0F0F) + + ((aux4 + (aux4 >> 4)) & 0x0F0F0F0F) + + ((aux5 + (aux5 >> 4)) & 0x0F0F0F0F) + + ((aux6 + (aux6 >> 4)) & 0x0F0F0F0F) + + ((aux7 + (aux7 >> 4)) & 0x0F0F0F0F))* 0x01010101) >> 24; + } + /* Count the remaining bytes. */ + p = (unsigned char*)p4; + while(count--) bits += bitsinbyte[*p++]; + return bits; +} + +/* Return the position of the first bit set to one (if 'bit' is 1) or + * zero (if 'bit' is 0) in the bitmap starting at 's' and PORT_LONG 'count' bytes. + * + * The function is guaranteed to return a value >= 0 if 'bit' is 0 since if + * no zero bit is found, it returns count*8 assuming the string is zero + * padded on the right. However if 'bit' is 1 it is possible that there is + * not a single set bit in the bitmap. In this special case -1 is returned. */ +PORT_LONG redisBitpos(void *s, PORT_ULONG count, int bit) { + PORT_ULONG *l; + unsigned char *c; + PORT_ULONG skipval, word = 0, one; + PORT_LONG pos = 0; /* Position of bit, to return to the caller. */ + PORT_ULONG j; + int found; + + /* Process whole words first, seeking for first word that is not + * all ones or all zeros respectively if we are lookig for zeros + * or ones. This is much faster with large strings having contiguous + * blocks of 1 or 0 bits compared to the vanilla bit per bit processing. + * + * Note that if we start from an address that is not aligned + * to sizeof(PORT_ULONG) we consume it byte by byte until it is + * aligned. */ + + /* Skip initial bits not aligned to sizeof(PORT_ULONG) byte by byte. */ + skipval = bit ? 0 : UCHAR_MAX; + c = (unsigned char*) s; + found = 0; + while((PORT_ULONG)c & (sizeof(*l)-1) && count) { + if (*c != skipval) { + found = 1; + break; + } + c++; + count--; + pos += 8; + } + + /* Skip bits with full word step. */ + l = (PORT_ULONG*) c; + if (!found) { + skipval = bit ? 0 : PORT_ULONG_MAX; + while (count >= sizeof(*l)) { + if (*l != skipval) break; + l++; + count -= sizeof(*l); + pos += sizeof(*l)*8; + } + } + + /* Load bytes into "word" considering the first byte as the most significant + * (we basically consider it as written in big endian, since we consider the + * string as a set of bits from left to right, with the first bit at position + * zero. + * + * Note that the loading is designed to work even when the bytes left + * (count) are less than a full word. We pad it with zero on the right. */ + c = (unsigned char*)l; + for (j = 0; j < sizeof(*l); j++) { + word <<= 8; + if (count) { + word |= *c; + c++; + count--; + } + } + + /* Special case: + * If bits in the string are all zero and we are looking for one, + * return -1 to signal that there is not a single "1" in the whole + * string. This can't happen when we are looking for "0" as we assume + * that the right of the string is zero padded. */ + if (bit == 1 && word == 0) return -1; + + /* Last word left, scan bit by bit. The first thing we need is to + * have a single "1" set in the most significant position in an + * PORT_ULONG. We don't know the size of the PORT_LONG so we use a + * simple trick. */ + one = PORT_ULONG_MAX; /* All bits set to 1.*/ + one >>= 1; /* All bits set to 1 but the MSB. */ + one = ~one; /* All bits set to 0 but the MSB. */ + + while(one) { + if (((one & word) != 0) == bit) return pos; + pos++; + one >>= 1; + } + + /* If we reached this point, there is a bug in the algorithm, since + * the case of no match is handled as a special case before. */ + serverPanic("End of redisBitpos() reached."); + return 0; /* Just to avoid warnings. */ +} + +/* The following set.*Bitfield and get.*Bitfield functions implement setting + * and getting arbitrary size (up to 64 bits) signed and unsigned integers + * at arbitrary positions into a bitmap. + * + * The representation considers the bitmap as having the bit number 0 to be + * the most significant bit of the first byte, and so forth, so for example + * setting a 5 bits unsigned integer to value 23 at offset 7 into a bitmap + * previously set to all zeroes, will produce the following representation: + * + * +--------+--------+ + * |00000001|01110000| + * +--------+--------+ + * + * When offsets and integer sizes are aligned to bytes boundaries, this is the + * same as big endian, however when such alignment does not exist, its important + * to also understand how the bits inside a byte are ordered. + * + * Note that this format follows the same convention as SETBIT and related + * commands. + */ + +void setUnsignedBitfield(unsigned char *p, uint64_t offset, uint64_t bits, uint64_t value) { + uint64_t byte, bit, byteval, bitval, j; + + for (j = 0; j < bits; j++) { + bitval = (value & ((uint64_t)1<<(bits-1-j))) != 0; + byte = offset >> 3; + bit = 7 - (offset & 0x7); + byteval = p[byte]; + byteval &= ~(1 << bit); + byteval |= bitval << bit; + p[byte] = byteval & 0xff; + offset++; + } +} + +void setSignedBitfield(unsigned char *p, uint64_t offset, uint64_t bits, int64_t value) { + uint64_t uv = value; /* Casting will add UINT64_MAX + 1 if v is negative. */ + setUnsignedBitfield(p,offset,bits,uv); +} + +uint64_t getUnsignedBitfield(unsigned char *p, uint64_t offset, uint64_t bits) { + uint64_t byte, bit, byteval, bitval, j, value = 0; + + for (j = 0; j < bits; j++) { + byte = offset >> 3; + bit = 7 - (offset & 0x7); + byteval = p[byte]; + bitval = (byteval >> bit) & 1; + value = (value<<1) | bitval; + offset++; + } + return value; +} + +int64_t getSignedBitfield(unsigned char *p, uint64_t offset, uint64_t bits) { + int64_t value; + union {uint64_t u; int64_t i;} conv; + + /* Converting from unsigned to signed is undefined when the value does + * not fit, however here we assume two's complement and the original value + * was obtained from signed -> unsigned conversion, so we'll find the + * most significant bit set if the original value was negative. + * + * Note that two's complement is mandatory for exact-width types + * according to the C99 standard. */ + conv.u = getUnsignedBitfield(p,offset,bits); + value = conv.i; + + /* If the top significant bit is 1, propagate it to all the + * higher bits for two's complement representation of signed + * integers. */ + if (value & ((uint64_t)1 << (bits-1))) + value |= ((uint64_t)-1) << bits; + return value; +} + +/* The following two functions detect overflow of a value in the context + * of storing it as an unsigned or signed integer with the specified + * number of bits. The functions both take the value and a possible increment. + * If no overflow could happen and the value+increment fit inside the limits, + * then zero is returned, otherwise in case of overflow, 1 is returned, + * otherwise in case of underflow, -1 is returned. + * + * When non-zero is returned (oferflow or underflow), if not NULL, *limit is + * set to the value the operation should result when an overflow happens, + * depending on the specified overflow semantics: + * + * For BFOVERFLOW_SAT if 1 is returned, *limit it is set maximum value that + * you can store in that integer. when -1 is returned, *limit is set to the + * minimum value that an integer of that size can represent. + * + * For BFOVERFLOW_WRAP *limit is set by performing the operation in order to + * "wrap" around towards zero for unsigned integers, or towards the most + * negative number that is possible to represent for signed integers. */ + +#define BFOVERFLOW_WRAP 0 +#define BFOVERFLOW_SAT 1 +#define BFOVERFLOW_FAIL 2 /* Used by the BITFIELD command implementation. */ + +int checkUnsignedBitfieldOverflow(uint64_t value, int64_t incr, uint64_t bits, int owtype, uint64_t *limit) { + uint64_t max = (bits == 64) ? UINT64_MAX : (((uint64_t)1< max || (incr > 0 && incr > maxincr)) { + if (limit) { + if (owtype == BFOVERFLOW_WRAP) { + goto handle_wrap; + } else if (owtype == BFOVERFLOW_SAT) { + *limit = max; + } + } + return 1; + } else if (incr < 0 && incr < minincr) { + if (limit) { + if (owtype == BFOVERFLOW_WRAP) { + goto handle_wrap; + } else if (owtype == BFOVERFLOW_SAT) { + *limit = 0; + } + } + return -1; + } + return 0; + +handle_wrap: + { + uint64_t mask = ((uint64_t)-1) << bits; + uint64_t res = value+incr; + + res &= ~mask; + *limit = res; + } + return 1; +} + +int checkSignedBitfieldOverflow(int64_t value, int64_t incr, uint64_t bits, int owtype, int64_t *limit) { + int64_t max = (bits == 64) ? INT64_MAX : (((int64_t)1<<(bits-1))-1); + int64_t min = (-max)-1; + + /* Note that maxincr and minincr could overflow, but we use the values + * only after checking 'value' range, so when we use it no overflow + * happens. */ + int64_t maxincr = max-value; + int64_t minincr = min-value; + + if (value > max || (bits != 64 && incr > maxincr) || (value >= 0 && incr > 0 && incr > maxincr)) + { + if (limit) { + if (owtype == BFOVERFLOW_WRAP) { + goto handle_wrap; + } else if (owtype == BFOVERFLOW_SAT) { + *limit = max; + } + } + return 1; + } else if (value < min || (bits != 64 && incr < minincr) || (value < 0 && incr < 0 && incr < minincr)) { + if (limit) { + if (owtype == BFOVERFLOW_WRAP) { + goto handle_wrap; + } else if (owtype == BFOVERFLOW_SAT) { + *limit = min; + } + } + return -1; + } + return 0; + +handle_wrap: + { + uint64_t mask = ((uint64_t)-1) << bits; + uint64_t msb = (uint64_t)1 << (bits-1); + uint64_t a = value, b = incr, c; + c = a+b; /* Perform addition as unsigned so that's defined. */ + + /* If the sign bit is set, propagate to all the higher order + * bits, to cap the negative value. If it's clear, mask to + * the positive integer limit. */ + if (c & msb) { + c |= mask; + } else { + c &= ~mask; + } + *limit = c; + } + return 1; +} + +/* Debugging function. Just show bits in the specified bitmap. Not used + * but here for not having to rewrite it when debugging is needed. */ +void printBits(unsigned char *p, PORT_ULONG count) { + PORT_ULONG j, i, byte; + + for (j = 0; j < count; j++) { + byte = p[j]; + for (i = 0x80; i > 0; i /= 2) + printf("%c", (byte & i) ? '1' : '0'); + printf("|"); + } + printf("\n"); +} + +/* ----------------------------------------------------------------------------- + * Bits related string commands: GETBIT, SETBIT, BITCOUNT, BITOP. + * -------------------------------------------------------------------------- */ + +#define BITOP_AND 0 +#define BITOP_OR 1 +#define BITOP_XOR 2 +#define BITOP_NOT 3 + +#define BITFIELDOP_GET 0 +#define BITFIELDOP_SET 1 +#define BITFIELDOP_INCRBY 2 + +/* This helper function used by GETBIT / SETBIT parses the bit offset argument + * making sure an error is returned if it is negative or if it overflows + * Redis 512 MB limit for the string value. + * + * If the 'hash' argument is true, and 'bits is positive, then the command + * will also parse bit offsets prefixed by "#". In such a case the offset + * is multiplied by 'bits'. This is useful for the BITFIELD command. */ +int getBitOffsetFromArgument(client *c, robj *o, size_t *offset, int hash, int bits) { + PORT_LONGLONG loffset; + char *err = "bit offset is not an integer or out of range"; + char *p = o->ptr; + size_t plen = sdslen(p); + int usehash = 0; + + /* Handle # form. */ + if (p[0] == '#' && hash && bits > 0) usehash = 1; + + if (string2ll(p+usehash,plen-usehash,&loffset) == 0) { + addReplyError(c,err); + return C_ERR; + } + + /* Adjust the offset by 'bits' for # form. */ + if (usehash) loffset *= bits; + + /* Limit offset to 512MB in bytes */ + if ((loffset < 0) || ((PORT_ULONGLONG)loffset >> 3) >= (512*1024*1024)) + { + addReplyError(c,err); + return C_ERR; + } + + *offset = (size_t)loffset; + return C_OK; +} + +/* This helper function for BITFIELD parses a bitfield type in the form + * where sign is 'u' or 'i' for unsigned and signed, and + * the bits is a value between 1 and 64. However 64 bits unsigned integers + * are reported as an error because of current limitations of Redis protocol + * to return unsigned integer values greater than INT64_MAX. + * + * On error C_ERR is returned and an error is sent to the client. */ +int getBitfieldTypeFromArgument(client *c, robj *o, int *sign, int *bits) { + char *p = o->ptr; + char *err = "Invalid bitfield type. Use something like i16 u8. Note that u64 is not supported but i64 is."; + PORT_LONGLONG llbits; + + if (p[0] == 'i') { + *sign = 1; + } else if (p[0] == 'u') { + *sign = 0; + } else { + addReplyError(c,err); + return C_ERR; + } + + if ((string2ll(p+1,strlen(p+1),&llbits)) == 0 || + llbits < 1 || + (*sign == 1 && llbits > 64) || + (*sign == 0 && llbits > 63)) + { + addReplyError(c,err); + return C_ERR; + } + *bits = llbits; + return C_OK; +} + +/* This is an helper function for commands implementations that need to write + * bits to a string object. The command creates or pad with zeroes the string + * so that the 'maxbit' bit can be addressed. The object is finally + * returned. Otherwise if the key holds a wrong type NULL is returned and + * an error is sent to the client. */ +robj *lookupStringForBitCommand(client *c, size_t maxbit) { + size_t byte = maxbit >> 3; + robj *o = lookupKeyWrite(c->db,c->argv[1]); + + if (o == NULL) { + o = createObject(OBJ_STRING,sdsnewlen(NULL, byte+1)); + dbAdd(c->db,c->argv[1],o); + } else { + if (checkType(c,o,OBJ_STRING)) return NULL; + o = dbUnshareStringValue(c->db,c->argv[1],o); + o->ptr = sdsgrowzero(o->ptr,byte+1); + } + return o; +} + +/* Return a pointer to the string object content, and stores its length + * in 'len'. The user is required to pass (likely stack allocated) buffer + * 'llbuf' of at least LONG_STR_SIZE bytes. Such a buffer is used in the case + * the object is integer encoded in order to provide the representation + * without usign heap allocation. + * + * The function returns the pointer to the object array of bytes representing + * the string it contains, that may be a pointer to 'llbuf' or to the + * internal object representation. As a side effect 'len' is filled with + * the length of such buffer. + * + * If the source object is NULL the function is guaranteed to return NULL + * and set 'len' to 0. */ +unsigned char *getObjectReadOnlyString(robj *o, PORT_LONG *len, char *llbuf) { + serverAssert(o->type == OBJ_STRING); + unsigned char *p = NULL; + + /* Set the 'p' pointer to the string, that can be just a stack allocated + * array if our string was integer encoded. */ + if (o && o->encoding == OBJ_ENCODING_INT) { + p = (unsigned char*) llbuf; + if (len) *len = ll2string(llbuf,LONG_STR_SIZE,(PORT_LONG)o->ptr); + } else if (o) { + p = (unsigned char*) o->ptr; + if (len) *len = sdslen(o->ptr); + } else { + if (len) *len = 0; + } + return p; +} + +/* SETBIT key offset bitvalue */ +void setbitCommand(client *c) { + robj *o; + char *err = "bit is not an integer or out of range"; + size_t bitoffset; + ssize_t byte, bit; + int byteval, bitval; + PORT_LONG on; + + if (getBitOffsetFromArgument(c,c->argv[2],&bitoffset,0,0) != C_OK) + return; + + if (getLongFromObjectOrReply(c,c->argv[3],&on,err) != C_OK) + return; + + /* Bits can only be set or cleared... */ + if (on & ~1) { + addReplyError(c,err); + return; + } + + if ((o = lookupStringForBitCommand(c,bitoffset)) == NULL) return; + + /* Get current values */ + byte = bitoffset >> 3; + byteval = ((uint8_t*)o->ptr)[byte]; + bit = 7 - (bitoffset & 0x7); + bitval = byteval & (1 << bit); + + /* Update byte with new bit value and return original value */ + byteval &= ~(1 << bit); + byteval |= ((on & 0x1) << bit); + ((uint8_t*)o->ptr)[byte] = byteval; + signalModifiedKey(c->db,c->argv[1]); + notifyKeyspaceEvent(NOTIFY_STRING,"setbit",c->argv[1],c->db->id); + server.dirty++; + addReply(c, bitval ? shared.cone : shared.czero); +} + +/* GETBIT key offset */ +void getbitCommand(client *c) { + robj *o; + char llbuf[32]; + size_t bitoffset; + size_t byte, bit; + size_t bitval = 0; + + if (getBitOffsetFromArgument(c,c->argv[2],&bitoffset,0,0) != C_OK) + return; + + if ((o = lookupKeyReadOrReply(c,c->argv[1],shared.czero)) == NULL || + checkType(c,o,OBJ_STRING)) return; + + byte = bitoffset >> 3; + bit = 7 - (bitoffset & 0x7); + if (sdsEncodedObject(o)) { + if (byte < sdslen(o->ptr)) + bitval = ((uint8_t*)o->ptr)[byte] & (1 << bit); + } else { + if (byte < (size_t)ll2string(llbuf,sizeof(llbuf),(PORT_LONG)o->ptr)) + bitval = llbuf[byte] & (1 << bit); + } + + addReply(c, bitval ? shared.cone : shared.czero); +} + +/* BITOP op_name target_key src_key1 src_key2 src_key3 ... src_keyN */ +void bitopCommand(client *c) { + char *opname = c->argv[1]->ptr; + robj *o, *targetkey = c->argv[2]; + PORT_ULONG op, j, numkeys; + robj **objects; /* Array of source objects. */ + unsigned char **src; /* Array of source strings pointers. */ + PORT_ULONG *len, maxlen = 0; /* Array of length of src strings, + and max len. */ + PORT_ULONG minlen = 0; /* Min len among the input keys. */ + unsigned char *res = NULL; /* Resulting string. */ + + /* Parse the operation name. */ + if ((opname[0] == 'a' || opname[0] == 'A') && !strcasecmp(opname,"and")) + op = BITOP_AND; + else if((opname[0] == 'o' || opname[0] == 'O') && !strcasecmp(opname,"or")) + op = BITOP_OR; + else if((opname[0] == 'x' || opname[0] == 'X') && !strcasecmp(opname,"xor")) + op = BITOP_XOR; + else if((opname[0] == 'n' || opname[0] == 'N') && !strcasecmp(opname,"not")) + op = BITOP_NOT; + else { + addReply(c,shared.syntaxerr); + return; + } + + /* Sanity check: NOT accepts only a single key argument. */ + if (op == BITOP_NOT && c->argc != 4) { + addReplyError(c,"BITOP NOT must be called with a single source key."); + return; + } + + /* Lookup keys, and store pointers to the string objects into an array. */ + numkeys = c->argc - 3; + src = zmalloc(sizeof(unsigned char*) * numkeys); + len = zmalloc(sizeof(PORT_LONG) * numkeys); + objects = zmalloc(sizeof(robj*) * numkeys); + for (j = 0; j < numkeys; j++) { + o = lookupKeyRead(c->db,c->argv[j+3]); + /* Handle non-existing keys as empty strings. */ + if (o == NULL) { + objects[j] = NULL; + src[j] = NULL; + len[j] = 0; + minlen = 0; + continue; + } + /* Return an error if one of the keys is not a string. */ + if (checkType(c,o,OBJ_STRING)) { + PORT_ULONG i; + for (i = 0; i < j; i++) { + if (objects[i]) + decrRefCount(objects[i]); + } + zfree(src); + zfree(len); + zfree(objects); + return; + } + objects[j] = getDecodedObject(o); + src[j] = objects[j]->ptr; + len[j] = (PORT_LONG) sdslen(objects[j]->ptr); WIN_PORT_FIX /* cast (PORT_LONG) */ + if (len[j] > maxlen) maxlen = len[j]; + if (j == 0 || len[j] < minlen) minlen = len[j]; + } + + /* Compute the bit operation, if at least one string is not empty. */ + if (maxlen) { + res = (unsigned char*) sdsnewlen(NULL,maxlen); + unsigned char output, byte; + PORT_ULONG i; + + /* Fast path: as far as we have data for all the input bitmaps we + * can take a fast path that performs much better than the + * vanilla algorithm. On ARM we skip the fast path since it will + * result in GCC compiling the code using multiple-words load/store + * operations that are not supported even in ARM >= v6. */ + j = 0; + #ifndef USE_ALIGNED_ACCESS + if (minlen >= sizeof(PORT_ULONG)*4 && numkeys <= 16) { + PORT_ULONG *lp[16]; + PORT_ULONG *lres = (PORT_ULONG*) res; + + /* Note: sds pointer is always aligned to 8 byte boundary. */ + memcpy(lp,src,sizeof(PORT_ULONG*)*numkeys); + memcpy(res,src[0],minlen); + + /* Different branches per different operations for speed (sorry). */ + if (op == BITOP_AND) { + while(minlen >= sizeof(PORT_ULONG)*4) { + for (i = 1; i < numkeys; i++) { + lres[0] &= lp[i][0]; + lres[1] &= lp[i][1]; + lres[2] &= lp[i][2]; + lres[3] &= lp[i][3]; + lp[i]+=4; + } + lres+=4; + j += sizeof(PORT_ULONG)*4; + minlen -= sizeof(PORT_ULONG)*4; + } + } else if (op == BITOP_OR) { + while(minlen >= sizeof(PORT_ULONG)*4) { + for (i = 1; i < numkeys; i++) { + lres[0] |= lp[i][0]; + lres[1] |= lp[i][1]; + lres[2] |= lp[i][2]; + lres[3] |= lp[i][3]; + lp[i]+=4; + } + lres+=4; + j += sizeof(PORT_ULONG)*4; + minlen -= sizeof(PORT_ULONG)*4; + } + } else if (op == BITOP_XOR) { + while(minlen >= sizeof(PORT_ULONG)*4) { + for (i = 1; i < numkeys; i++) { + lres[0] ^= lp[i][0]; + lres[1] ^= lp[i][1]; + lres[2] ^= lp[i][2]; + lres[3] ^= lp[i][3]; + lp[i]+=4; + } + lres+=4; + j += sizeof(PORT_ULONG)*4; + minlen -= sizeof(PORT_ULONG)*4; + } + } else if (op == BITOP_NOT) { + while(minlen >= sizeof(PORT_ULONG)*4) { + lres[0] = ~lres[0]; + lres[1] = ~lres[1]; + lres[2] = ~lres[2]; + lres[3] = ~lres[3]; + lres+=4; + j += sizeof(PORT_ULONG)*4; + minlen -= sizeof(PORT_ULONG)*4; + } + } + } + #endif + + /* j is set to the next byte to process by the previous loop. */ + for (; j < maxlen; j++) { + output = (len[0] <= j) ? 0 : src[0][j]; + if (op == BITOP_NOT) output = ~output; + for (i = 1; i < numkeys; i++) { + byte = (len[i] <= j) ? 0 : src[i][j]; + switch(op) { + case BITOP_AND: output &= byte; break; + case BITOP_OR: output |= byte; break; + case BITOP_XOR: output ^= byte; break; + } + } + res[j] = output; + } + } + for (j = 0; j < numkeys; j++) { + if (objects[j]) + decrRefCount(objects[j]); + } + zfree(src); + zfree(len); + zfree(objects); + + /* Store the computed value into the target key */ + if (maxlen) { + o = createObject(OBJ_STRING,res); + setKey(c->db,targetkey,o); + notifyKeyspaceEvent(NOTIFY_STRING,"set",targetkey,c->db->id); + decrRefCount(o); + } else if (dbDelete(c->db,targetkey)) { + signalModifiedKey(c->db,targetkey); + notifyKeyspaceEvent(NOTIFY_GENERIC,"del",targetkey,c->db->id); + } + server.dirty++; + addReplyLongLong(c,maxlen); /* Return the output string length in bytes. */ +} + +/* BITCOUNT key [start end] */ +void bitcountCommand(client *c) { + robj *o; + PORT_LONG start, end, strlen; + unsigned char *p; + char llbuf[LONG_STR_SIZE]; + + /* Lookup, check for type, and return 0 for non existing keys. */ + if ((o = lookupKeyReadOrReply(c,c->argv[1],shared.czero)) == NULL || + checkType(c,o,OBJ_STRING)) return; + p = getObjectReadOnlyString(o,&strlen,llbuf); + + /* Parse start/end range if any. */ + if (c->argc == 4) { + if (getLongFromObjectOrReply(c,c->argv[2],&start,NULL) != C_OK) + return; + if (getLongFromObjectOrReply(c,c->argv[3],&end,NULL) != C_OK) + return; + /* Convert negative indexes */ + if (start < 0 && end < 0 && start > end) { + addReply(c,shared.czero); + return; + } + if (start < 0) start = strlen+start; + if (end < 0) end = strlen+end; + if (start < 0) start = 0; + if (end < 0) end = 0; + if (end >= strlen) end = strlen-1; + } else if (c->argc == 2) { + /* The whole string. */ + start = 0; + end = strlen-1; + } else { + /* Syntax error. */ + addReply(c,shared.syntaxerr); + return; + } + + /* Precondition: end >= 0 && end < strlen, so the only condition where + * zero can be returned is: start > end. */ + if (start > end) { + addReply(c,shared.czero); + } else { + PORT_LONG bytes = end-start+1; + + addReplyLongLong(c,redisPopcount(p+start,bytes)); + } +} + +/* BITPOS key bit [start [end]] */ +void bitposCommand(client *c) { + robj *o; + PORT_LONG bit, start, end, strlen; + unsigned char *p; + char llbuf[LONG_STR_SIZE]; + int end_given = 0; + + /* Parse the bit argument to understand what we are looking for, set + * or clear bits. */ + if (getLongFromObjectOrReply(c,c->argv[2],&bit,NULL) != C_OK) + return; + if (bit != 0 && bit != 1) { + addReplyError(c, "The bit argument must be 1 or 0."); + return; + } + + /* If the key does not exist, from our point of view it is an infinite + * array of 0 bits. If the user is looking for the fist clear bit return 0, + * If the user is looking for the first set bit, return -1. */ + if ((o = lookupKeyRead(c->db,c->argv[1])) == NULL) { + addReplyLongLong(c, bit ? -1 : 0); + return; + } + if (checkType(c,o,OBJ_STRING)) return; + p = getObjectReadOnlyString(o,&strlen,llbuf); + + /* Parse start/end range if any. */ + if (c->argc == 4 || c->argc == 5) { + if (getLongFromObjectOrReply(c,c->argv[3],&start,NULL) != C_OK) + return; + if (c->argc == 5) { + if (getLongFromObjectOrReply(c,c->argv[4],&end,NULL) != C_OK) + return; + end_given = 1; + } else { + end = strlen-1; + } + /* Convert negative indexes */ + if (start < 0) start = strlen+start; + if (end < 0) end = strlen+end; + if (start < 0) start = 0; + if (end < 0) end = 0; + if (end >= strlen) end = strlen-1; + } else if (c->argc == 3) { + /* The whole string. */ + start = 0; + end = strlen-1; + } else { + /* Syntax error. */ + addReply(c,shared.syntaxerr); + return; + } + + /* For empty ranges (start > end) we return -1 as an empty range does + * not contain a 0 nor a 1. */ + if (start > end) { + addReplyLongLong(c, -1); + } else { + PORT_LONG bytes = end-start+1; + PORT_LONG pos = redisBitpos(p+start,(PORT_ULONG)bytes,(int)bit); WIN_PORT_FIX /* cast (PORT_ULONG), cast (int) */ + + /* If we are looking for clear bits, and the user specified an exact + * range with start-end, we can't consider the right of the range as + * zero padded (as we do when no explicit end is given). + * + * So if redisBitpos() returns the first bit outside the range, + * we return -1 to the caller, to mean, in the specified range there + * is not a single "0" bit. */ + if (end_given && bit == 0 && pos == bytes*8) { + addReplyLongLong(c,-1); + return; + } + if (pos != -1) pos += start*8; /* Adjust for the bytes we skipped. */ + addReplyLongLong(c,pos); + } +} + +/* BITFIELD key subcommmand-1 arg ... subcommand-2 arg ... subcommand-N ... + * + * Supported subcommands: + * + * GET + * SET + * INCRBY + * OVERFLOW [WRAP|SAT|FAIL] + */ + +struct bitfieldOp { + uint64_t offset; /* Bitfield offset. */ + int64_t i64; /* Increment amount (INCRBY) or SET value */ + int opcode; /* Operation id. */ + int owtype; /* Overflow type to use. */ + int bits; /* Integer bitfield bits width. */ + int sign; /* True if signed, otherwise unsigned op. */ +}; + +void bitfieldCommand(client *c) { + robj *o; + size_t bitoffset; + int j, numops = 0, changes = 0; + struct bitfieldOp *ops = NULL; /* Array of ops to execute at end. */ + int owtype = BFOVERFLOW_WRAP; /* Overflow type. */ + int readonly = 1; + PORT_LONG higest_write_offset = 0; + + for (j = 2; j < c->argc; j++) { + int remargs = c->argc-j-1; /* Remaining args other than current. */ + char *subcmd = c->argv[j]->ptr; /* Current command name. */ + int opcode; /* Current operation code. */ + PORT_LONGLONG i64 = 0; /* Signed SET value. */ + int sign = 0; /* Signed or unsigned type? */ + int bits = 0; /* Bitfield width in bits. */ + + if (!strcasecmp(subcmd,"get") && remargs >= 2) + opcode = BITFIELDOP_GET; + else if (!strcasecmp(subcmd,"set") && remargs >= 3) + opcode = BITFIELDOP_SET; + else if (!strcasecmp(subcmd,"incrby") && remargs >= 3) + opcode = BITFIELDOP_INCRBY; + else if (!strcasecmp(subcmd,"overflow") && remargs >= 1) { + char *owtypename = c->argv[j+1]->ptr; + j++; + if (!strcasecmp(owtypename,"wrap")) + owtype = BFOVERFLOW_WRAP; + else if (!strcasecmp(owtypename,"sat")) + owtype = BFOVERFLOW_SAT; + else if (!strcasecmp(owtypename,"fail")) + owtype = BFOVERFLOW_FAIL; + else { + addReplyError(c,"Invalid OVERFLOW type specified"); + zfree(ops); + return; + } + continue; + } else { + addReply(c,shared.syntaxerr); + zfree(ops); + return; + } + + /* Get the type and offset arguments, common to all the ops. */ + if (getBitfieldTypeFromArgument(c,c->argv[j+1],&sign,&bits) != C_OK) { + zfree(ops); + return; + } + + if (getBitOffsetFromArgument(c,c->argv[j+2],&bitoffset,1,bits) != C_OK){ + zfree(ops); + return; + } + + if (opcode != BITFIELDOP_GET) { + readonly = 0; + if (higest_write_offset < bitoffset + bits - 1) + higest_write_offset = bitoffset + bits - 1; + /* INCRBY and SET require another argument. */ + if (getLongLongFromObjectOrReply(c,c->argv[j+3],&i64,NULL) != C_OK){ + zfree(ops); + return; + } + } + + /* Populate the array of operations we'll process. */ + ops = zrealloc(ops,sizeof(*ops)*(numops+1)); + ops[numops].offset = bitoffset; + ops[numops].i64 = i64; + ops[numops].opcode = opcode; + ops[numops].owtype = owtype; + ops[numops].bits = bits; + ops[numops].sign = sign; + numops++; + + j += 3 - (opcode == BITFIELDOP_GET); + } + + if (readonly) { + /* Lookup for read is ok if key doesn't exit, but errors + * if it's not a string. */ + o = lookupKeyRead(c->db,c->argv[1]); + if (o != NULL && checkType(c,o,OBJ_STRING)) return; + } else { + /* Lookup by making room up to the farest bit reached by + * this operation. */ + if ((o = lookupStringForBitCommand(c, + higest_write_offset)) == NULL) return; + } + + addReplyMultiBulkLen(c,numops); + + /* Actually process the operations. */ + for (j = 0; j < numops; j++) { + struct bitfieldOp *thisop = ops+j; + + /* Execute the operation. */ + if (thisop->opcode == BITFIELDOP_SET || + thisop->opcode == BITFIELDOP_INCRBY) + { + /* SET and INCRBY: We handle both with the same code path + * for simplicity. SET return value is the previous value so + * we need fetch & store as well. */ + + /* We need two different but very similar code paths for signed + * and unsigned operations, since the set of functions to get/set + * the integers and the used variables types are different. */ + if (thisop->sign) { + int64_t oldval, newval, wrapped, retval; + int overflow; + + oldval = getSignedBitfield(o->ptr,thisop->offset, + thisop->bits); + + if (thisop->opcode == BITFIELDOP_INCRBY) { + newval = oldval + thisop->i64; + overflow = checkSignedBitfieldOverflow(oldval, + thisop->i64,thisop->bits,thisop->owtype,&wrapped); + if (overflow) newval = wrapped; + retval = newval; + } else { + newval = thisop->i64; + overflow = checkSignedBitfieldOverflow(newval, + 0,thisop->bits,thisop->owtype,&wrapped); + if (overflow) newval = wrapped; + retval = oldval; + } + + /* On overflow of type is "FAIL", don't write and return + * NULL to signal the condition. */ + if (!(overflow && thisop->owtype == BFOVERFLOW_FAIL)) { + addReplyLongLong(c,retval); + setSignedBitfield(o->ptr,thisop->offset, + thisop->bits,newval); + } else { + addReply(c,shared.nullbulk); + } + } else { + uint64_t oldval, newval, wrapped, retval; + int overflow; + + oldval = getUnsignedBitfield(o->ptr,thisop->offset, + thisop->bits); + + if (thisop->opcode == BITFIELDOP_INCRBY) { + newval = oldval + thisop->i64; + overflow = checkUnsignedBitfieldOverflow(oldval, + thisop->i64,thisop->bits,thisop->owtype,&wrapped); + if (overflow) newval = wrapped; + retval = newval; + } else { + newval = thisop->i64; + overflow = checkUnsignedBitfieldOverflow(newval, + 0,thisop->bits,thisop->owtype,&wrapped); + if (overflow) newval = wrapped; + retval = oldval; + } + /* On overflow of type is "FAIL", don't write and return + * NULL to signal the condition. */ + if (!(overflow && thisop->owtype == BFOVERFLOW_FAIL)) { + addReplyLongLong(c,retval); + setUnsignedBitfield(o->ptr,thisop->offset, + thisop->bits,newval); + } else { + addReply(c,shared.nullbulk); + } + } + changes++; + } else { + /* GET */ + unsigned char buf[9]; + PORT_LONG strlen = 0; + unsigned char *src = NULL; + char llbuf[LONG_STR_SIZE]; + + if (o != NULL) + src = getObjectReadOnlyString(o,&strlen,llbuf); + + /* For GET we use a trick: before executing the operation + * copy up to 9 bytes to a local buffer, so that we can easily + * execute up to 64 bit operations that are at actual string + * object boundaries. */ + memset(buf,0,9); + int i; + size_t byte = thisop->offset >> 3; + for (i = 0; i < 9; i++) { + if (src == NULL || i+byte >= (size_t)strlen) break; + buf[i] = src[i+byte]; + } + + /* Now operate on the copied buffer which is guaranteed + * to be zero-padded. */ + if (thisop->sign) { + int64_t val = getSignedBitfield(buf,thisop->offset-(byte*8), + thisop->bits); + addReplyLongLong(c,val); + } else { + uint64_t val = getUnsignedBitfield(buf,thisop->offset-(byte*8), + thisop->bits); + addReplyLongLong(c,val); + } + } + } + + if (changes) { + signalModifiedKey(c->db,c->argv[1]); + notifyKeyspaceEvent(NOTIFY_STRING,"setbit",c->argv[1],c->db->id); + server.dirty += changes; + } + zfree(ops); +} diff --git a/src/cluster.c b/src/cluster.c index 86bb45fe..ca27c30b 100644 --- a/src/cluster.c +++ b/src/cluster.c @@ -118,7 +118,7 @@ int clusterLoadConfig(char *filename) { } /* Parse the file. Note that single lines of the cluster config file can - * be really PORT_LONG as they include all the hash slots of the node. + * be really long as they include all the hash slots of the node. * This means in the worst possible case, half of the Redis slots will be * present in a single line, possibly in importing or migrating state, so * together with the node ID of the sender/receiver. @@ -260,6 +260,7 @@ int clusterLoadConfig(char *filename) { *p = '\0'; direction = p[1]; /* Either '>' or '<' */ slot = atoi(argv[j] + 1); + if (slot < 0 || slot >= CLUSTER_SLOTS) goto fmterr; p += 3; cn = clusterLookupNode(p); if (!cn) { @@ -282,6 +283,8 @@ int clusterLoadConfig(char *filename) { else { start = stop = atoi(argv[j]); } + if (start < 0 || start >= CLUSTER_SLOTS) goto fmterr; + if (stop < 0 || stop >= CLUSTER_SLOTS) goto fmterr; while (start <= stop) clusterAddSlot(n, start++); } @@ -438,6 +441,22 @@ int clusterLockConfig(char *filename) { return C_OK; } +/* Some flags (currently just the NOFAILOVER flag) may need to be updated + * in the "myself" node based on the current configuration of the node, + * that may change at runtime via CONFIG SET. This function changes the + * set of flags in myself->flags accordingly. */ +void clusterUpdateMyselfFlags(void) { + int oldflags = myself->flags; + int nofailover = server.cluster_slave_no_failover ? + CLUSTER_NODE_NOFAILOVER : 0; + myself->flags &= ~CLUSTER_NODE_NOFAILOVER; + myself->flags |= nofailover; + if (myself->flags != oldflags) { + clusterDoBeforeSleep(CLUSTER_TODO_SAVE_CONFIG| + CLUSTER_TODO_UPDATE_STATE); + } +} + void clusterInit(void) { int saveconf = 0; @@ -531,6 +550,7 @@ void clusterInit(void) { server.cluster->mf_end = 0; resetManualFailover(); + clusterUpdateMyselfFlags(); } /* Reset a node performing a soft or hard reset: @@ -1862,6 +1882,18 @@ int clusterProcessPacket(clusterLink *link) { } } + /* Copy the CLUSTER_NODE_NOFAILOVER flag from what the sender + * announced. This is a dynamic flag that we receive from the + * sender, and the latest status must be trusted. We need it to + * be propagated because the slave ranking used to understand the + * delay of each slave in the voting process, needs to know + * what are the instances really competing. */ + if (sender) { + int nofailover = flags & CLUSTER_NODE_NOFAILOVER; + sender->flags &= ~CLUSTER_NODE_NOFAILOVER; + sender->flags |= nofailover; + } + /* Update the node address if it changed. */ if (sender && type == CLUSTERMSG_TYPE_PING && !nodeInHandshake(sender) && @@ -2261,7 +2293,7 @@ void clusterReadHandler(aeEventLoop *el, int fd, void *privdata, int mask) { * from event handlers that will do stuff with the same link later. */ void clusterSendMessage(clusterLink *link, unsigned char *msg, size_t msglen) { if (sdslen(link->sndbuf) == 0 && msglen != 0) - aeCreateFileEvent(server.el, link->fd, AE_WRITABLE, + aeCreateFileEvent(server.el,link->fd,AE_WRITABLE|AE_BARRIER, clusterWriteHandler, link); link->sndbuf = sdscatlen(link->sndbuf, msg, msglen); @@ -2803,6 +2835,8 @@ void clusterSendFailoverAuthIfNeeded(clusterNode *node, clusterMsg *request) { clusterSendFailoverAuth(node); server.cluster->lastVoteEpoch = server.cluster->currentEpoch; node->slaveof->voted_time = mstime(); + clusterDoBeforeSleep(CLUSTER_TODO_SAVE_CONFIG|CLUSTER_TODO_FSYNC_CONFIG); + clusterSendFailoverAuth(node); serverLog(LL_WARNING, "Failover auth granted to %.40s for epoch %llu", node->name, (PORT_ULONGLONG) server.cluster->currentEpoch); } @@ -2831,6 +2865,7 @@ int clusterGetSlaveRank(void) { myoffset = replicationGetSlaveOffset(); for (j = 0; j < master->numslaves; j++) if (master->slaves[j] != myself && + !nodeCantFailover(master->slaves[j]) && master->slaves[j]->repl_offset > myoffset) rank++; return rank; } @@ -2968,7 +3003,9 @@ void clusterHandleSlaveFailover(void) { * of an automatic or manual failover: * 1) We are a slave. * 2) Our master is flagged as FAIL, or this is a manual failover. - * 3) It is serving slots. */ + * 3) We don't have the no failover configuration set, and this is + * not a manual failover. + * 4) It is serving slots. */ if (nodeIsMaster(myself) || myself->slaveof == NULL || (!nodeFailed(myself->slaveof) && !manual_failover) || @@ -3351,6 +3388,9 @@ void clusterCron(void) { handshake_timeout = server.cluster_node_timeout; if (handshake_timeout < 1000) handshake_timeout = 1000; + /* Update myself flags. */ + clusterUpdateMyselfFlags(); + /* Check if we have disconnected nodes and re-establish the connection. * Also update a few stats while we are here, that can be used to make * better decisions in other part of the code. */ @@ -3951,7 +3991,8 @@ static struct redisNodeFlags redisNodeFlagsTable[] = { {CLUSTER_NODE_PFAIL, "fail?,"}, {CLUSTER_NODE_FAIL, "fail,"}, {CLUSTER_NODE_HANDSHAKE, "handshake,"}, - {CLUSTER_NODE_NOADDR, "noaddr,"} + {CLUSTER_NODE_NOADDR, "noaddr,"}, + {CLUSTER_NODE_NOFAILOVER, "nofailover,"} }; /* Concatenate the comma separated list of node flags to the given SDS @@ -4560,8 +4601,7 @@ void clusterCommand(client *c) { clusterDoBeforeSleep(CLUSTER_TODO_UPDATE_STATE | CLUSTER_TODO_SAVE_CONFIG); addReply(c, shared.ok); - } - else if (!strcasecmp(c->argv[1]->ptr, "replicate") && c->argc == 3) { + } else if (!strcasecmp(c->argv[1]->ptr,"replicate") && c->argc == 3) { /* CLUSTER REPLICATE */ clusterNode *n = clusterLookupNode(c->argv[2]->ptr); @@ -5030,17 +5070,18 @@ void migrateCloseTimedoutSockets(void) { dictReleaseIterator(di); } -/* MIGRATE host port key dbid timeout [COPY | REPLACE] +/* MIGRATE host port key dbid timeout [COPY | REPLACE | AUTH password] * * On in the multiple keys form: * - * MIGRATE host port "" dbid timeout [COPY | REPLACE] KEYS key1 key2 ... keyN */ + * MIGRATE host port "" dbid timeout [COPY | REPLACE | AUTH password] KEYS key1 + * key2 ... keyN */ void migrateCommand(client *c) { migrateCachedSocket *cs; - int copy, replace, j; + int copy=0, replace=0, j; + char *password = NULL; PORT_LONG timeout; PORT_LONG dbid; - PORT_LONGLONG ttl, expireat; robj **ov = NULL; /* Objects to migrate. */ robj **kv = NULL; /* Key names. */ robj **newargv = NULL; /* Used to rewrite the command as DEL ... keys ... */ @@ -5053,19 +5094,21 @@ void migrateCommand(client *c) { int first_key = 3; /* Argument index of the first key. */ int num_keys = 1; /* By default only migrate the 'key' argument. */ - /* Initialization */ - copy = 0; - replace = 0; - /* Parse additional options */ for (j = 6; j < c->argc; j++) { + int moreargs = j < c->argc-1; if (!strcasecmp(c->argv[j]->ptr, "copy")) { copy = 1; - } - else if (!strcasecmp(c->argv[j]->ptr, "replace")) { + } else if (!strcasecmp(c->argv[j]->ptr,"replace")) { replace = 1; + } else if (!strcasecmp(c->argv[j]->ptr,"auth")) { + if (!moreargs) { + addReply(c,shared.syntaxerr); + return; } - else if (!strcasecmp(c->argv[j]->ptr, "keys")) { + j++; + password = c->argv[j]->ptr; + } else if (!strcasecmp(c->argv[j]->ptr,"keys")) { if (sdslen(c->argv[3]->ptr) != 0) { addReplyError(c, "When using MIGRATE KEYS option, the key argument" @@ -5124,6 +5167,14 @@ try_again: rioInitWithBuffer(&cmd, sdsempty()); + /* Authentication */ + if (password) { + serverAssertWithInfo(c,NULL,rioWriteBulkCount(&cmd,'*',2)); + serverAssertWithInfo(c,NULL,rioWriteBulkString(&cmd,"AUTH",4)); + serverAssertWithInfo(c,NULL,rioWriteBulkString(&cmd,password, + sdslen(password))); + } + /* Send the SELECT command if the current DB is not already selected. */ int select = cs->last_dbid != dbid; /* Should we emit SELECT? */ if (select) { @@ -5141,7 +5192,9 @@ try_again: ttl = expireat - mstime(); if (ttl < 1) ttl = 1; } - serverAssertWithInfo(c, NULL, rioWriteBulkCount(&cmd, '*', replace ? 5 : 4)); + serverAssertWithInfo(c,NULL, + rioWriteBulkCount(&cmd,'*',replace ? 5 : 4)); + if (server.cluster_enabled) serverAssertWithInfo(c, NULL, rioWriteBulkString(&cmd, "RESTORE-ASKING", 14)); @@ -5207,9 +5260,14 @@ try_again: #endif } + char buf0[1024]; /* Auth reply. */ char buf1[1024]; /* Select reply. */ char buf2[1024]; /* Restore reply. */ + /* Read the AUTH reply if needed. */ + if (password && syncReadLine(cs->fd, buf0, sizeof(buf0), timeout) <= 0) + goto socket_err; + /* Read the SELECT reply if needed. */ if (select && syncReadLine(cs->fd, buf1, sizeof(buf1), timeout) <= 0) goto socket_err; @@ -5226,13 +5284,21 @@ try_again: socket_error = 1; break; } - if ((select && buf1[0] == '-') || buf2[0] == '-') { + if ((password && buf0[0] == '-') || + (select && buf1[0] == '-') || + buf2[0] == '-') + { /* On error assume that last_dbid is no longer valid. */ if (!error_from_target) { cs->last_dbid = -1; - addReplyErrorFormat(c, "Target instance replied with error: %s", - (select && buf1[0] == '-') ? buf1 + 1 : buf2 + 1); + char *errbuf; + if (password && buf0[0] == '-') errbuf = buf0; + else if (select && buf1[0] == '-') errbuf = buf1; + else errbuf = buf2; + error_from_target = 1; + addReplyErrorFormat(c,"Target instance replied with error: %s", + errbuf+1); } } else { @@ -5300,7 +5366,7 @@ try_again: } else { /* On error we already sent it in the for loop above, and set - * the curretly selected socket to -1 to force SELECT the next time. */ + * the currently selected socket to -1 to force SELECT the next time. */ } sdsfree(cmd.io.buffer.ptr); diff --git a/src/cluster.c.bak b/src/cluster.c.bak new file mode 100644 index 00000000..0922bf5e --- /dev/null +++ b/src/cluster.c.bak @@ -0,0 +1,5692 @@ +/* Redis Cluster implementation. + * + * Copyright (c) 2009-2012, Salvatore Sanfilippo + * All rights reserved. + * + * Redistribution and use in source and binary forms, with or without + * modification, are permitted provided that the following conditions are met: + * + * * Redistributions of source code must retain the above copyright notice, + * this list of conditions and the following disclaimer. + * * Redistributions in binary form must reproduce the above copyright + * notice, this list of conditions and the following disclaimer in the + * documentation and/or other materials provided with the distribution. + * * Neither the name of Redis nor the names of its contributors may be used + * to endorse or promote products derived from this software without + * specific prior written permission. + * + * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" + * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE + * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE + * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE + * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR + * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF + * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS + * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN + * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) + * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE + * POSSIBILITY OF SUCH DAMAGE. + */ + +#include "server.h" +#include "cluster.h" +#include "endianconv.h" + +#include +POSIX_ONLY(#include ) +POSIX_ONLY(#include ) +#include +POSIX_ONLY(#include ) +POSIX_ONLY(#include ) +#include +POSIX_ONLY(#include ) +#include + +#ifdef _WIN32 +extern int WSIOCP_QueueAccept(int listenfd); +#include "Win32_Interop/Win32_Error.h" +#endif + +/* A global reference to myself is handy to make code more clear. + * Myself always points to server.cluster->myself, that is, the clusterNode + * that represents this node. */ +clusterNode *myself = NULL; + +clusterNode *createClusterNode(char *nodename, int flags); +int clusterAddNode(clusterNode *node); +void clusterAcceptHandler(aeEventLoop *el, int fd, void *privdata, int mask); +void clusterReadHandler(aeEventLoop *el, int fd, void *privdata, int mask); +void clusterSendPing(clusterLink *link, int type); +void clusterSendFail(char *nodename); +void clusterSendFailoverAuthIfNeeded(clusterNode *node, clusterMsg *request); +void clusterUpdateState(void); +int clusterNodeGetSlotBit(clusterNode *n, int slot); +sds clusterGenNodesDescription(int filter); +clusterNode *clusterLookupNode(char *name); +int clusterNodeAddSlave(clusterNode *master, clusterNode *slave); +int clusterAddSlot(clusterNode *n, int slot); +int clusterDelSlot(int slot); +int clusterDelNodeSlots(clusterNode *node); +int clusterNodeSetSlotBit(clusterNode *n, int slot); +void clusterSetMaster(clusterNode *n); +void clusterHandleSlaveFailover(void); +void clusterHandleSlaveMigration(int max_slaves); +int bitmapTestBit(unsigned char *bitmap, int pos); +void clusterDoBeforeSleep(int flags); +void clusterSendUpdate(clusterLink *link, clusterNode *node); +void resetManualFailover(void); +void clusterCloseAllSlots(void); +void clusterSetNodeAsMaster(clusterNode *n); +void clusterDelNode(clusterNode *delnode); +sds representClusterNodeFlags(sds ci, uint16_t flags); +uint64_t clusterGetMaxEpoch(void); +int clusterBumpConfigEpochWithoutConsensus(void); + +/* ----------------------------------------------------------------------------- + * Initialization + * -------------------------------------------------------------------------- */ + + /* Load the cluster config from 'filename'. + * + * If the file does not exist or is zero-length (this may happen because + * when we lock the nodes.conf file, we create a zero-length one for the + * sake of locking if it does not already exist), C_ERR is returned. + * If the configuration was loaded from the file, C_OK is returned. */ +int clusterLoadConfig(char *filename) { + FILE *fp = fopen(filename, IF_WIN32("rb", "r")); + struct IF_WIN32(_stat64, stat) sb; // TODO: verify for 32-bit + char *line; + int maxline, j; + + if (fp == NULL) { + if (errno == ENOENT) { + return C_ERR; + } + else { + serverLog(LL_WARNING, + "Loading the cluster node config from %s: %s", + filename, IF_WIN32(wsa_strerror(errno), strerror(errno))); + exit(1); + } + } + + /* Check if the file is zero-length: if so return C_ERR to signal + * we have to write the config. */ + if (fstat(fileno(fp), &sb) != -1 && sb.st_size == 0) { + fclose(fp); + return C_ERR; + } + + /* Parse the file. Note that single lines of the cluster config file can + * be really PORT_LONG as they include all the hash slots of the node. + * This means in the worst possible case, half of the Redis slots will be + * present in a single line, possibly in importing or migrating state, so + * together with the node ID of the sender/receiver. + * + * To simplify we allocate 1024+CLUSTER_SLOTS*128 bytes per line. */ + maxline = 1024 + CLUSTER_SLOTS * 128; + line = zmalloc(maxline); + while (fgets(line, maxline, fp) != NULL) { + int argc; + sds *argv; + clusterNode *n, *master; + char *p, *s; + + /* Skip blank lines, they can be created either by users manually + * editing nodes.conf or by the config writing process if stopped + * before the truncate() call. */ + if (line[0] == '\n' || line[0] == '\0') continue; + + /* Split the line into arguments for processing. */ + argv = sdssplitargs(line, &argc); + if (argv == NULL) goto fmterr; + + /* Handle the special "vars" line. Don't pretend it is the last + * line even if it actually is when generated by Redis. */ + if (strcasecmp(argv[0], "vars") == 0) { + for (j = 1; j < argc; j += 2) { + if (strcasecmp(argv[j], "currentEpoch") == 0) { + server.cluster->currentEpoch = + strtoull(argv[j + 1], NULL, 10); + } + else if (strcasecmp(argv[j], "lastVoteEpoch") == 0) { + server.cluster->lastVoteEpoch = + strtoull(argv[j + 1], NULL, 10); + } + else { + serverLog(LL_WARNING, + "Skipping unknown cluster config variable '%s'", + argv[j]); + } + } + sdsfreesplitres(argv, argc); + continue; + } + + /* Regular config lines have at least eight fields */ + if (argc < 8) goto fmterr; + + /* Create this node if it does not exist */ + n = clusterLookupNode(argv[0]); + if (!n) { + n = createClusterNode(argv[0], 0); + clusterAddNode(n); + } + /* Address and port */ + if ((p = strrchr(argv[1], ':')) == NULL) goto fmterr; + *p = '\0'; + memcpy(n->ip, argv[1], strlen(argv[1]) + 1); + char *port = p + 1; + char *busp = strchr(port, '@'); + if (busp) { + *busp = '\0'; + busp++; + } + n->port = atoi(port); + /* In older versions of nodes.conf the "@busport" part is missing. + * In this case we set it to the default offset of 10000 from the + * base port. */ + n->cport = busp ? atoi(busp) : n->port + CLUSTER_PORT_INCR; + + /* Parse flags */ + p = s = argv[2]; + while (p) { + p = strchr(s, ','); + if (p) *p = '\0'; + if (!strcasecmp(s, "myself")) { + serverAssert(server.cluster->myself == NULL); + myself = server.cluster->myself = n; + n->flags |= CLUSTER_NODE_MYSELF; + } + else if (!strcasecmp(s, "master")) { + n->flags |= CLUSTER_NODE_MASTER; + } + else if (!strcasecmp(s, "slave")) { + n->flags |= CLUSTER_NODE_SLAVE; + } + else if (!strcasecmp(s, "fail?")) { + n->flags |= CLUSTER_NODE_PFAIL; + } + else if (!strcasecmp(s, "fail")) { + n->flags |= CLUSTER_NODE_FAIL; + n->fail_time = mstime(); + } + else if (!strcasecmp(s, "handshake")) { + n->flags |= CLUSTER_NODE_HANDSHAKE; + } + else if (!strcasecmp(s, "noaddr")) { + n->flags |= CLUSTER_NODE_NOADDR; + } + else if (!strcasecmp(s, "noflags")) { + /* nothing to do */ + } + else { + serverPanic("Unknown flag in redis cluster config file"); + } + if (p) s = p + 1; + } + + /* Get master if any. Set the master and populate master's + * slave list. */ + if (argv[3][0] != '-') { + master = clusterLookupNode(argv[3]); + if (!master) { + master = createClusterNode(argv[3], 0); + clusterAddNode(master); + } + n->slaveof = master; + clusterNodeAddSlave(master, n); + } + + /* Set ping sent / pong received timestamps */ + if (atoi(argv[4])) n->ping_sent = mstime(); + if (atoi(argv[5])) n->pong_received = mstime(); + + /* Set configEpoch for this node. */ + n->configEpoch = strtoull(argv[6], NULL, 10); + + /* Populate hash slots served by this instance. */ + for (j = 8; j < argc; j++) { + int start, stop; + + if (argv[j][0] == '[') { + /* Here we handle migrating / importing slots */ + int slot; + char direction; + clusterNode *cn; + + p = strchr(argv[j], '-'); + serverAssert(p != NULL); + *p = '\0'; + direction = p[1]; /* Either '>' or '<' */ + slot = atoi(argv[j] + 1); + p += 3; + cn = clusterLookupNode(p); + if (!cn) { + cn = createClusterNode(p, 0); + clusterAddNode(cn); + } + if (direction == '>') { + server.cluster->migrating_slots_to[slot] = cn; + } + else { + server.cluster->importing_slots_from[slot] = cn; + } + continue; + } + else if ((p = strchr(argv[j], '-')) != NULL) { + *p = '\0'; + start = atoi(argv[j]); + stop = atoi(p + 1); + } + else { + start = stop = atoi(argv[j]); + } + while (start <= stop) clusterAddSlot(n, start++); + } + + sdsfreesplitres(argv, argc); + } + /* Config sanity check */ + if (server.cluster->myself == NULL) goto fmterr; + + zfree(line); + fclose(fp); + + serverLog(LL_NOTICE, "Node configuration loaded, I'm %.40s", myself->name); + + /* Something that should never happen: currentEpoch smaller than + * the max epoch found in the nodes configuration. However we handle this + * as some form of protection against manual editing of critical files. */ + if (clusterGetMaxEpoch() > server.cluster->currentEpoch) { + server.cluster->currentEpoch = clusterGetMaxEpoch(); + } + return C_OK; + +fmterr: + serverLog(LL_WARNING, + "Unrecoverable error: corrupted cluster config file."); + zfree(line); + if (fp) fclose(fp); + exit(1); +} + +/* Cluster node configuration is exactly the same as CLUSTER NODES output. + * + * This function writes the node config and returns 0, on error -1 + * is returned. + * + * Note: we need to write the file in an atomic way from the point of view + * of the POSIX filesystem semantics, so that if the server is stopped + * or crashes during the write, we'll end with either the old file or the + * new one. Since we have the full payload to write available we can use + * a single write to write the whole file. If the pre-existing file was + * bigger we pad our payload with newlines that are anyway ignored and truncate + * the file afterward. */ +int clusterSaveConfig(int do_fsync) { + sds ci; + size_t content_size; + struct IF_WIN32(_stat64, stat) sb; // TODO: verify for 32-bit + int fd; + + server.cluster->todo_before_sleep &= ~CLUSTER_TODO_SAVE_CONFIG; + + /* Get the nodes description and concatenate our "vars" directive to + * save currentEpoch and lastVoteEpoch. */ + ci = clusterGenNodesDescription(CLUSTER_NODE_HANDSHAKE); + ci = sdscatprintf(ci, "vars currentEpoch %llu lastVoteEpoch %llu\n", + (PORT_ULONGLONG) server.cluster->currentEpoch, + (PORT_ULONGLONG) server.cluster->lastVoteEpoch); + content_size = sdslen(ci); + + if ((fd = open(server.cluster_configfile, O_WRONLY | O_CREAT, 0644)) + == -1) goto err; + + /* Pad the new payload if the existing file length is greater. */ + if (fstat(fd, &sb) != -1) { + if (sb.st_size > (off_t) content_size) { + ci = sdsgrowzero(ci, sb.st_size); + memset(ci + content_size, '\n', sb.st_size - content_size); + } + } + if (write(fd, ci, sdslen(ci)) != (ssize_t) sdslen(ci)) goto err; + if (do_fsync) { + server.cluster->todo_before_sleep &= ~CLUSTER_TODO_FSYNC_CONFIG; + fsync(fd); + } + + /* Truncate the file if needed to remove the final \n padding that + * is just garbage. */ + if (content_size != sdslen(ci) && ftruncate(fd, content_size) == -1) { + /* ftruncate() failing is not a critical error. */ + } + close(fd); + sdsfree(ci); + return 0; + +err: + if (fd != -1) close(fd); + sdsfree(ci); + return -1; +} + +void clusterSaveConfigOrDie(int do_fsync) { + if (clusterSaveConfig(do_fsync) == -1) { + serverLog(LL_WARNING, "Fatal: can't update cluster config file."); + exit(1); + } +} + +/* Lock the cluster config using flock(), and leaks the file descritor used to + * acquire the lock so that the file will be locked forever. + * + * This works because we always update nodes.conf with a new version + * in-place, reopening the file, and writing to it in place (later adjusting + * the length with ftruncate()). + * + * On success C_OK is returned, otherwise an error is logged and + * the function returns C_ERR to signal a lock was not acquired. */ +int clusterLockConfig(char *filename) { + /* flock() does not exist on Solaris + * and a fcntl-based solution won't help, as we constantly re-open that file, + * which will release _all_ locks anyway + */ +#if !defined(__sun) + /* To lock it, we need to open the file in a way it is created if + * it does not exist, otherwise there is a race condition with other + * processes. */ + int fd = open(filename, O_WRONLY | O_CREAT, 0644); + if (fd == -1) { + serverLog(LL_WARNING, + "Can't open %s in order to acquire a lock: %s", + filename, IF_WIN32(wsa_strerror(errno), strerror(errno))); + return C_ERR; + } + +#ifndef _WIN32 + if (flock(fd, LOCK_EX | LOCK_NB) == -1) { + if (errno == EWOULDBLOCK) { +#else + HANDLE hFile = (HANDLE) FDAPI_get_osfhandle(fd); + OVERLAPPED ovlp; + DWORD size_lower, size_upper; + // start offset is 0, and also zero the remaining members of the struct + memset(&ovlp, 0, sizeof ovlp); + // get file size + size_lower = GetFileSize(hFile, &size_upper); + if (!LockFileEx(hFile, LOCKFILE_EXCLUSIVE_LOCK | LOCKFILE_FAIL_IMMEDIATELY, 0, size_lower, size_upper, &ovlp)) { + DWORD err = GetLastError(); + if (err == ERROR_LOCK_VIOLATION) { +#endif + serverLog(LL_WARNING, + "Sorry, the cluster configuration file %s is already used " + "by a different Redis Cluster node. Please make sure that " + "different nodes use different cluster configuration " + "files.", filename); + } + else { + serverLog(LL_WARNING, + "Impossible to lock %s: %s", filename, IF_WIN32(wsa_strerror(errno), strerror(errno))); + } + close(fd); + return C_ERR; + } + /* Lock acquired: leak the 'fd' by not closing it, so that we'll retain the + * lock to the file as long as the process exists. */ +#endif /* __sun */ + + return C_OK; +} + +void clusterInit(void) { + int saveconf = 0; + + server.cluster = zmalloc(sizeof(clusterState)); + server.cluster->myself = NULL; + server.cluster->currentEpoch = 0; + server.cluster->state = CLUSTER_FAIL; + server.cluster->size = 1; + server.cluster->todo_before_sleep = 0; + server.cluster->nodes = dictCreate(&clusterNodesDictType, NULL); + server.cluster->nodes_black_list = + dictCreate(&clusterNodesBlackListDictType, NULL); + server.cluster->failover_auth_time = 0; + server.cluster->failover_auth_count = 0; + server.cluster->failover_auth_rank = 0; + server.cluster->failover_auth_epoch = 0; + server.cluster->cant_failover_reason = CLUSTER_CANT_FAILOVER_NONE; + server.cluster->lastVoteEpoch = 0; + for (int i = 0; i < CLUSTERMSG_TYPE_COUNT; i++) { + server.cluster->stats_bus_messages_sent[i] = 0; + server.cluster->stats_bus_messages_received[i] = 0; + } + server.cluster->stats_pfail_nodes = 0; + memset(server.cluster->slots, 0, sizeof(server.cluster->slots)); + clusterCloseAllSlots(); + +#ifndef WIN32 // TODO: review this to verify if we can lock the file on Windows + /* Lock the cluster config file to make sure every node uses + * its own nodes.conf. */ + if (clusterLockConfig(server.cluster_configfile) == C_ERR) + exit(1); +#endif + + /* Load or create a new nodes configuration. */ + if (clusterLoadConfig(server.cluster_configfile) == C_ERR) { + /* No configuration found. We will just use the random name provided + * by the createClusterNode() function. */ + myself = server.cluster->myself = + createClusterNode(NULL, CLUSTER_NODE_MYSELF | CLUSTER_NODE_MASTER); + serverLog(LL_NOTICE, "No cluster configuration found, I'm %.40s", + myself->name); + clusterAddNode(myself); + saveconf = 1; + } + if (saveconf) clusterSaveConfigOrDie(1); + + /* We need a listening TCP port for our cluster messaging needs. */ + server.cfd_count = 0; + + /* Port sanity check II + * The other handshake port check is triggered too late to stop + * us from trying to use a too-high cluster port number. */ + if (server.port > (65535 - CLUSTER_PORT_INCR)) { + serverLog(LL_WARNING, "Redis port number too high. " + "Cluster communication port is 10,000 port " + "numbers higher than your Redis port. " + "Your Redis port number must be " + "lower than 55535."); + exit(1); + } + + if (listenToPort(server.port + CLUSTER_PORT_INCR, + server.cfd, &server.cfd_count) == C_ERR) + { + exit(1); + } + else { + int j; + + for (j = 0; j < server.cfd_count; j++) { + if (aeCreateFileEvent(server.el, server.cfd[j], AE_READABLE, + clusterAcceptHandler, NULL) == AE_ERR) + serverPanic("Unrecoverable error creating Redis Cluster " + "file event."); + } + } + + /* The slots -> keys map is a radix tree. Initialize it here. */ + server.cluster->slots_to_keys = raxNew(); + memset(server.cluster->slots_keys_count, 0, + sizeof(server.cluster->slots_keys_count)); + + /* Set myself->port / cport to my listening ports, we'll just need to + * discover the IP address via MEET messages. */ + myself->port = server.port; + myself->cport = server.port + CLUSTER_PORT_INCR; + if (server.cluster_announce_port) + myself->port = server.cluster_announce_port; + if (server.cluster_announce_bus_port) + myself->cport = server.cluster_announce_bus_port; + + server.cluster->mf_end = 0; + resetManualFailover(); +} + +/* Reset a node performing a soft or hard reset: + * + * 1) All other nodes are forget. + * 2) All the assigned / open slots are released. + * 3) If the node is a slave, it turns into a master. + * 5) Only for hard reset: a new Node ID is generated. + * 6) Only for hard reset: currentEpoch and configEpoch are set to 0. + * 7) The new configuration is saved and the cluster state updated. + * 8) If the node was a slave, the whole data set is flushed away. */ +void clusterReset(int hard) { + dictIterator *di; + dictEntry *de; + int j; + + /* Turn into master. */ + if (nodeIsSlave(myself)) { + clusterSetNodeAsMaster(myself); + replicationUnsetMaster(); + emptyDb(-1, EMPTYDB_NO_FLAGS, NULL); + } + + /* Close slots, reset manual failover state. */ + clusterCloseAllSlots(); + resetManualFailover(); + + /* Unassign all the slots. */ + for (j = 0; j < CLUSTER_SLOTS; j++) clusterDelSlot(j); + + /* Forget all the nodes, but myself. */ + di = dictGetSafeIterator(server.cluster->nodes); + while ((de = dictNext(di)) != NULL) { + clusterNode *node = dictGetVal(de); + + if (node == myself) continue; + clusterDelNode(node); + } + dictReleaseIterator(di); + + /* Hard reset only: set epochs to 0, change node ID. */ + if (hard) { + sds oldname; + + server.cluster->currentEpoch = 0; + server.cluster->lastVoteEpoch = 0; + myself->configEpoch = 0; + serverLog(LL_WARNING, "configEpoch set to 0 via CLUSTER RESET HARD"); + + /* To change the Node ID we need to remove the old name from the + * nodes table, change the ID, and re-add back with new name. */ + oldname = sdsnewlen(myself->name, CLUSTER_NAMELEN); + dictDelete(server.cluster->nodes, oldname); + sdsfree(oldname); + getRandomHexChars(myself->name, CLUSTER_NAMELEN); + clusterAddNode(myself); + serverLog(LL_NOTICE, "Node hard reset, now I'm %.40s", myself->name); + } + + /* Make sure to persist the new config and update the state. */ + clusterDoBeforeSleep(CLUSTER_TODO_SAVE_CONFIG | + CLUSTER_TODO_UPDATE_STATE | + CLUSTER_TODO_FSYNC_CONFIG); +} + +/* ----------------------------------------------------------------------------- + * CLUSTER communication link + * -------------------------------------------------------------------------- */ + +clusterLink *createClusterLink(clusterNode *node) { + clusterLink *link = zmalloc(sizeof(*link)); + link->ctime = mstime(); + link->sndbuf = sdsempty(); + link->rcvbuf = sdsempty(); + link->node = node; + link->fd = -1; + return link; +} + +/* Free a cluster link, but does not free the associated node of course. + * This function will just make sure that the original node associated + * with this link will have the 'link' field set to NULL. */ +void freeClusterLink(clusterLink *link) { + if (link->fd != -1) { + aeDeleteFileEvent(server.el, link->fd, AE_WRITABLE); + aeDeleteFileEvent(server.el, link->fd, AE_READABLE); + } + sdsfree(link->sndbuf); + sdsfree(link->rcvbuf); + if (link->node) + link->node->link = NULL; + close(link->fd); + zfree(link); +} + +#define MAX_CLUSTER_ACCEPTS_PER_CALL 1000 +void clusterAcceptHandler(aeEventLoop *el, int fd, void *privdata, int mask) { + int cport, cfd; + int max = MAX_CLUSTER_ACCEPTS_PER_CALL; + char cip[NET_IP_STR_LEN]; + clusterLink *link; + UNUSED(el); + UNUSED(mask); + UNUSED(privdata); + + /* If the server is starting up, don't accept cluster connections: + * UPDATE messages may interact with the database content. */ + if (server.masterhost == NULL && server.loading) { + WIN32_ONLY(WSIOCP_QueueAccept(fd);) + return; + } + + while (max--) { + cfd = anetTcpAccept(server.neterr, fd, cip, sizeof(cip), &cport); + if (cfd == ANET_ERR) { + if (errno != EWOULDBLOCK) + serverLog(LL_VERBOSE, + "Error accepting cluster node: %s", server.neterr); +#ifdef _WIN32 + if (WSIOCP_QueueAccept(fd) == -1) { + serverLog(LL_WARNING, + "acceptTcpHandler: failed to queue another accept."); + } +#endif + return; + } + anetNonBlock(NULL, cfd); + anetEnableTcpNoDelay(NULL, cfd); + + /* Use non-blocking I/O for cluster messages. */ + serverLog(LL_VERBOSE, "Accepted cluster node %s:%d", cip, cport); + /* Create a link object we use to handle the connection. + * It gets passed to the readable handler when data is available. + * Initiallly the link->node pointer is set to NULL as we don't know + * which node is, but the right node is references once we know the + * node identity. */ + link = createClusterLink(NULL); + link->fd = cfd; + aeCreateFileEvent(server.el, cfd, AE_READABLE, clusterReadHandler, link); + } +} + +/* ----------------------------------------------------------------------------- + * Key space handling + * -------------------------------------------------------------------------- */ + + /* We have 16384 hash slots. The hash slot of a given key is obtained + * as the least significant 14 bits of the crc16 of the key. + * + * However if the key contains the {...} pattern, only the part between + * { and } is hashed. This may be useful in the future to force certain + * keys to be in the same node (assuming no resharding is in progress). */ +unsigned int keyHashSlot(char *key, int keylen) { + int s, e; /* start-end indexes of { and } */ + + for (s = 0; s < keylen; s++) + if (key[s] == '{') break; + + /* No '{' ? Hash the whole key. This is the base case. */ + if (s == keylen) return crc16(key, keylen) & 0x3FFF; + + /* '{' found? Check if we have the corresponding '}'. */ + for (e = s + 1; e < keylen; e++) + if (key[e] == '}') break; + + /* No '}' or nothing betweeen {} ? Hash the whole key. */ + if (e == keylen || e == s + 1) return crc16(key, keylen) & 0x3FFF; + + /* If we are here there is both a { and a } on its right. Hash + * what is in the middle between { and }. */ + return crc16(key + s + 1, e - s - 1) & 0x3FFF; +} + +/* ----------------------------------------------------------------------------- + * CLUSTER node API + * -------------------------------------------------------------------------- */ + + /* Create a new cluster node, with the specified flags. + * If "nodename" is NULL this is considered a first handshake and a random + * node name is assigned to this node (it will be fixed later when we'll + * receive the first pong). + * + * The node is created and returned to the user, but it is not automatically + * added to the nodes hash table. */ +clusterNode *createClusterNode(char *nodename, int flags) { + clusterNode *node = zmalloc(sizeof(*node)); + + if (nodename) + memcpy(node->name, nodename, CLUSTER_NAMELEN); + else + getRandomHexChars(node->name, CLUSTER_NAMELEN); + node->ctime = mstime(); + node->configEpoch = 0; + node->flags = flags; + memset(node->slots, 0, sizeof(node->slots)); + node->numslots = 0; + node->numslaves = 0; + node->slaves = NULL; + node->slaveof = NULL; + node->ping_sent = node->pong_received = 0; + node->fail_time = 0; + node->link = NULL; + memset(node->ip, 0, sizeof(node->ip)); + node->port = 0; + node->cport = 0; + node->fail_reports = listCreate(); + node->voted_time = 0; + node->orphaned_time = 0; + node->repl_offset_time = 0; + node->repl_offset = 0; + listSetFreeMethod(node->fail_reports, zfree); + return node; +} + +/* This function is called every time we get a failure report from a node. + * The side effect is to populate the fail_reports list (or to update + * the timestamp of an existing report). + * + * 'failing' is the node that is in failure state according to the + * 'sender' node. + * + * The function returns 0 if it just updates a timestamp of an existing + * failure report from the same sender. 1 is returned if a new failure + * report is created. */ +int clusterNodeAddFailureReport(clusterNode *failing, clusterNode *sender) { + list *l = failing->fail_reports; + listNode *ln; + listIter li; + clusterNodeFailReport *fr; + + /* If a failure report from the same sender already exists, just update + * the timestamp. */ + listRewind(l, &li); + while ((ln = listNext(&li)) != NULL) { + fr = ln->value; + if (fr->node == sender) { + fr->time = mstime(); + return 0; + } + } + + /* Otherwise create a new report. */ + fr = zmalloc(sizeof(*fr)); + fr->node = sender; + fr->time = mstime(); + listAddNodeTail(l, fr); + return 1; +} + +/* Remove failure reports that are too old, where too old means reasonably + * older than the global node timeout. Note that anyway for a node to be + * flagged as FAIL we need to have a local PFAIL state that is at least + * older than the global node timeout, so we don't just trust the number + * of failure reports from other nodes. */ +void clusterNodeCleanupFailureReports(clusterNode *node) { + list *l = node->fail_reports; + listNode *ln; + listIter li; + clusterNodeFailReport *fr; + mstime_t maxtime = server.cluster_node_timeout * + CLUSTER_FAIL_REPORT_VALIDITY_MULT; + mstime_t now = mstime(); + + listRewind(l, &li); + while ((ln = listNext(&li)) != NULL) { + fr = ln->value; + if (now - fr->time > maxtime) listDelNode(l, ln); + } +} + +/* Remove the failing report for 'node' if it was previously considered + * failing by 'sender'. This function is called when a node informs us via + * gossip that a node is OK from its point of view (no FAIL or PFAIL flags). + * + * Note that this function is called relatively often as it gets called even + * when there are no nodes failing, and is O(N), however when the cluster is + * fine the failure reports list is empty so the function runs in constant + * time. + * + * The function returns 1 if the failure report was found and removed. + * Otherwise 0 is returned. */ +int clusterNodeDelFailureReport(clusterNode *node, clusterNode *sender) { + list *l = node->fail_reports; + listNode *ln; + listIter li; + clusterNodeFailReport *fr; + + /* Search for a failure report from this sender. */ + listRewind(l, &li); + while ((ln = listNext(&li)) != NULL) { + fr = ln->value; + if (fr->node == sender) break; + } + if (!ln) return 0; /* No failure report from this sender. */ + + /* Remove the failure report. */ + listDelNode(l, ln); + clusterNodeCleanupFailureReports(node); + return 1; +} + +/* Return the number of external nodes that believe 'node' is failing, + * not including this node, that may have a PFAIL or FAIL state for this + * node as well. */ +int clusterNodeFailureReportsCount(clusterNode *node) { + clusterNodeCleanupFailureReports(node); + return (int) listLength(node->fail_reports); WIN_PORT_FIX /* cast (int) */ +} + +int clusterNodeRemoveSlave(clusterNode *master, clusterNode *slave) { + int j; + + for (j = 0; j < master->numslaves; j++) { + if (master->slaves[j] == slave) { + if ((j + 1) < master->numslaves) { + int remaining_slaves = (master->numslaves - j) - 1; + memmove(master->slaves + j, master->slaves + (j + 1), + (sizeof(*master->slaves) * remaining_slaves)); + } + master->numslaves--; + if (master->numslaves == 0) + master->flags &= ~CLUSTER_NODE_MIGRATE_TO; + return C_OK; + } + } + return C_ERR; +} + +int clusterNodeAddSlave(clusterNode *master, clusterNode *slave) { + int j; + + /* If it's already a slave, don't add it again. */ + for (j = 0; j < master->numslaves; j++) + if (master->slaves[j] == slave) return C_ERR; + master->slaves = zrealloc(master->slaves, + sizeof(clusterNode*)*(master->numslaves + 1)); + master->slaves[master->numslaves] = slave; + master->numslaves++; + master->flags |= CLUSTER_NODE_MIGRATE_TO; + return C_OK; +} + +int clusterCountNonFailingSlaves(clusterNode *n) { + int j, okslaves = 0; + + for (j = 0; j < n->numslaves; j++) + if (!nodeFailed(n->slaves[j])) okslaves++; + return okslaves; +} + +/* Low level cleanup of the node structure. Only called by clusterDelNode(). */ +void freeClusterNode(clusterNode *n) { + sds nodename; + int j; + + /* If the node has associated slaves, we have to set + * all the slaves->slaveof fields to NULL (unknown). */ + for (j = 0; j < n->numslaves; j++) + n->slaves[j]->slaveof = NULL; + + /* Remove this node from the list of slaves of its master. */ + if (nodeIsSlave(n) && n->slaveof) clusterNodeRemoveSlave(n->slaveof, n); + + /* Unlink from the set of nodes. */ + nodename = sdsnewlen(n->name, CLUSTER_NAMELEN); + serverAssert(dictDelete(server.cluster->nodes, nodename) == DICT_OK); + sdsfree(nodename); + + /* Release link and associated data structures. */ + if (n->link) freeClusterLink(n->link); + listRelease(n->fail_reports); + zfree(n->slaves); + zfree(n); +} + +/* Add a node to the nodes hash table */ +int clusterAddNode(clusterNode *node) { + int retval; + + retval = dictAdd(server.cluster->nodes, + sdsnewlen(node->name, CLUSTER_NAMELEN), node); + return (retval == DICT_OK) ? C_OK : C_ERR; +} + +/* Remove a node from the cluster. The functio performs the high level + * cleanup, calling freeClusterNode() for the low level cleanup. + * Here we do the following: + * + * 1) Mark all the slots handled by it as unassigned. + * 2) Remove all the failure reports sent by this node and referenced by + * other nodes. + * 3) Free the node with freeClusterNode() that will in turn remove it + * from the hash table and from the list of slaves of its master, if + * it is a slave node. + */ +void clusterDelNode(clusterNode *delnode) { + int j; + dictIterator *di; + dictEntry *de; + + /* 1) Mark slots as unassigned. */ + for (j = 0; j < CLUSTER_SLOTS; j++) { + if (server.cluster->importing_slots_from[j] == delnode) + server.cluster->importing_slots_from[j] = NULL; + if (server.cluster->migrating_slots_to[j] == delnode) + server.cluster->migrating_slots_to[j] = NULL; + if (server.cluster->slots[j] == delnode) + clusterDelSlot(j); + } + + /* 2) Remove failure reports. */ + di = dictGetSafeIterator(server.cluster->nodes); + while ((de = dictNext(di)) != NULL) { + clusterNode *node = dictGetVal(de); + + if (node == delnode) continue; + clusterNodeDelFailureReport(node, delnode); + } + dictReleaseIterator(di); + + /* 3) Free the node, unlinking it from the cluster. */ + freeClusterNode(delnode); +} + +/* Node lookup by name */ +clusterNode *clusterLookupNode(char *name) { + sds s = sdsnewlen(name, CLUSTER_NAMELEN); + dictEntry *de; + + de = dictFind(server.cluster->nodes, s); + sdsfree(s); + if (de == NULL) return NULL; + return dictGetVal(de); +} + +/* This is only used after the handshake. When we connect a given IP/PORT + * as a result of CLUSTER MEET we don't have the node name yet, so we + * pick a random one, and will fix it when we receive the PONG request using + * this function. */ +void clusterRenameNode(clusterNode *node, char *newname) { + int retval; + sds s = sdsnewlen(node->name, CLUSTER_NAMELEN); + + serverLog(LL_DEBUG, "Renaming node %.40s into %.40s", + node->name, newname); + retval = dictDelete(server.cluster->nodes, s); + sdsfree(s); + serverAssert(retval == DICT_OK); + memcpy(node->name, newname, CLUSTER_NAMELEN); + clusterAddNode(node); +} + +/* ----------------------------------------------------------------------------- + * CLUSTER config epoch handling + * -------------------------------------------------------------------------- */ + + /* Return the greatest configEpoch found in the cluster, or the current + * epoch if greater than any node configEpoch. */ +uint64_t clusterGetMaxEpoch(void) { + uint64_t max = 0; + dictIterator *di; + dictEntry *de; + + di = dictGetSafeIterator(server.cluster->nodes); + while ((de = dictNext(di)) != NULL) { + clusterNode *node = dictGetVal(de); + if (node->configEpoch > max) max = node->configEpoch; + } + dictReleaseIterator(di); + if (max < server.cluster->currentEpoch) max = server.cluster->currentEpoch; + return max; +} + +/* If this node epoch is zero or is not already the greatest across the + * cluster (from the POV of the local configuration), this function will: + * + * 1) Generate a new config epoch, incrementing the current epoch. + * 2) Assign the new epoch to this node, WITHOUT any consensus. + * 3) Persist the configuration on disk before sending packets with the + * new configuration. + * + * If the new config epoch is generated and assigend, C_OK is returned, + * otherwise C_ERR is returned (since the node has already the greatest + * configuration around) and no operation is performed. + * + * Important note: this function violates the principle that config epochs + * should be generated with consensus and should be unique across the cluster. + * However Redis Cluster uses this auto-generated new config epochs in two + * cases: + * + * 1) When slots are closed after importing. Otherwise resharding would be + * too expensive. + * 2) When CLUSTER FAILOVER is called with options that force a slave to + * failover its master even if there is not master majority able to + * create a new configuration epoch. + * + * Redis Cluster will not explode using this function, even in the case of + * a collision between this node and another node, generating the same + * configuration epoch unilaterally, because the config epoch conflict + * resolution algorithm will eventually move colliding nodes to different + * config epochs. However using this function may violate the "last failover + * wins" rule, so should only be used with care. */ +int clusterBumpConfigEpochWithoutConsensus(void) { + uint64_t maxEpoch = clusterGetMaxEpoch(); + + if (myself->configEpoch == 0 || + myself->configEpoch != maxEpoch) + { + server.cluster->currentEpoch++; + myself->configEpoch = server.cluster->currentEpoch; + clusterDoBeforeSleep(CLUSTER_TODO_SAVE_CONFIG | + CLUSTER_TODO_FSYNC_CONFIG); + serverLog(LL_WARNING, + "New configEpoch set to %llu", + (PORT_ULONGLONG) myself->configEpoch); + return C_OK; + } + else { + return C_ERR; + } +} + +/* This function is called when this node is a master, and we receive from + * another master a configuration epoch that is equal to our configuration + * epoch. + * + * BACKGROUND + * + * It is not possible that different slaves get the same config + * epoch during a failover election, because the slaves need to get voted + * by a majority. However when we perform a manual resharding of the cluster + * the node will assign a configuration epoch to itself without to ask + * for agreement. Usually resharding happens when the cluster is working well + * and is supervised by the sysadmin, however it is possible for a failover + * to happen exactly while the node we are resharding a slot to assigns itself + * a new configuration epoch, but before it is able to propagate it. + * + * So technically it is possible in this condition that two nodes end with + * the same configuration epoch. + * + * Another possibility is that there are bugs in the implementation causing + * this to happen. + * + * Moreover when a new cluster is created, all the nodes start with the same + * configEpoch. This collision resolution code allows nodes to automatically + * end with a different configEpoch at startup automatically. + * + * In all the cases, we want a mechanism that resolves this issue automatically + * as a safeguard. The same configuration epoch for masters serving different + * set of slots is not harmful, but it is if the nodes end serving the same + * slots for some reason (manual errors or software bugs) without a proper + * failover procedure. + * + * In general we want a system that eventually always ends with different + * masters having different configuration epochs whatever happened, since + * nothign is worse than a split-brain condition in a distributed system. + * + * BEHAVIOR + * + * When this function gets called, what happens is that if this node + * has the lexicographically smaller Node ID compared to the other node + * with the conflicting epoch (the 'sender' node), it will assign itself + * the greatest configuration epoch currently detected among nodes plus 1. + * + * This means that even if there are multiple nodes colliding, the node + * with the greatest Node ID never moves forward, so eventually all the nodes + * end with a different configuration epoch. + */ +void clusterHandleConfigEpochCollision(clusterNode *sender) { + /* Prerequisites: nodes have the same configEpoch and are both masters. */ + if (sender->configEpoch != myself->configEpoch || + !nodeIsMaster(sender) || !nodeIsMaster(myself)) return; + /* Don't act if the colliding node has a smaller Node ID. */ + if (memcmp(sender->name, myself->name, CLUSTER_NAMELEN) <= 0) return; + /* Get the next ID available at the best of this node knowledge. */ + server.cluster->currentEpoch++; + myself->configEpoch = server.cluster->currentEpoch; + clusterSaveConfigOrDie(1); + serverLog(LL_VERBOSE, + "WARNING: configEpoch collision with node %.40s." + " configEpoch set to %llu", + sender->name, + (PORT_ULONGLONG) myself->configEpoch); +} + +/* ----------------------------------------------------------------------------- + * CLUSTER nodes blacklist + * + * The nodes blacklist is just a way to ensure that a given node with a given + * Node ID is not readded before some time elapsed (this time is specified + * in seconds in CLUSTER_BLACKLIST_TTL). + * + * This is useful when we want to remove a node from the cluster completely: + * when CLUSTER FORGET is called, it also puts the node into the blacklist so + * that even if we receive gossip messages from other nodes that still remember + * about the node we want to remove, we don't re-add it before some time. + * + * Currently the CLUSTER_BLACKLIST_TTL is set to 1 minute, this means + * that redis-trib has 60 seconds to send CLUSTER FORGET messages to nodes + * in the cluster without dealing with the problem of other nodes re-adding + * back the node to nodes we already sent the FORGET command to. + * + * The data structure used is a hash table with an sds string representing + * the node ID as key, and the time when it is ok to re-add the node as + * value. + * -------------------------------------------------------------------------- */ + +#define CLUSTER_BLACKLIST_TTL 60 /* 1 minute. */ + + /* Before of the addNode() or Exists() operations we always remove expired + * entries from the black list. This is an O(N) operation but it is not a + * problem since add / exists operations are called very infrequently and + * the hash table is supposed to contain very little elements at max. + * However without the cleanup during long uptimes and with some automated + * node add/removal procedures, entries could accumulate. */ +void clusterBlacklistCleanup(void) { + dictIterator *di; + dictEntry *de; + + di = dictGetSafeIterator(server.cluster->nodes_black_list); + while ((de = dictNext(di)) != NULL) { + int64_t expire = dictGetUnsignedIntegerVal(de); + + if (expire < server.unixtime) + dictDelete(server.cluster->nodes_black_list, dictGetKey(de)); + } + dictReleaseIterator(di); +} + +/* Cleanup the blacklist and add a new node ID to the black list. */ +void clusterBlacklistAddNode(clusterNode *node) { + dictEntry *de; + sds id = sdsnewlen(node->name, CLUSTER_NAMELEN); + + clusterBlacklistCleanup(); + if (dictAdd(server.cluster->nodes_black_list, id, NULL) == DICT_OK) { + /* If the key was added, duplicate the sds string representation of + * the key for the next lookup. We'll free it at the end. */ + id = sdsdup(id); + } + de = dictFind(server.cluster->nodes_black_list, id); + dictSetUnsignedIntegerVal(de, time(NULL) + CLUSTER_BLACKLIST_TTL); + sdsfree(id); +} + +/* Return non-zero if the specified node ID exists in the blacklist. + * You don't need to pass an sds string here, any pointer to 40 bytes + * will work. */ +int clusterBlacklistExists(char *nodeid) { + sds id = sdsnewlen(nodeid, CLUSTER_NAMELEN); + int retval; + + clusterBlacklistCleanup(); + retval = dictFind(server.cluster->nodes_black_list, id) != NULL; + sdsfree(id); + return retval; +} + +/* ----------------------------------------------------------------------------- + * CLUSTER messages exchange - PING/PONG and gossip + * -------------------------------------------------------------------------- */ + + /* This function checks if a given node should be marked as FAIL. + * It happens if the following conditions are met: + * + * 1) We received enough failure reports from other master nodes via gossip. + * Enough means that the majority of the masters signaled the node is + * down recently. + * 2) We believe this node is in PFAIL state. + * + * If a failure is detected we also inform the whole cluster about this + * event trying to force every other node to set the FAIL flag for the node. + * + * Note that the form of agreement used here is weak, as we collect the majority + * of masters state during some time, and even if we force agreement by + * propagating the FAIL message, because of partitions we may not reach every + * node. However: + * + * 1) Either we reach the majority and eventually the FAIL state will propagate + * to all the cluster. + * 2) Or there is no majority so no slave promotion will be authorized and the + * FAIL flag will be cleared after some time. + */ +void markNodeAsFailingIfNeeded(clusterNode *node) { + int failures; + int needed_quorum = (server.cluster->size / 2) + 1; + + if (!nodeTimedOut(node)) return; /* We can reach it. */ + if (nodeFailed(node)) return; /* Already FAILing. */ + + failures = clusterNodeFailureReportsCount(node); + /* Also count myself as a voter if I'm a master. */ + if (nodeIsMaster(myself)) failures++; + if (failures < needed_quorum) return; /* No weak agreement from masters. */ + + serverLog(LL_NOTICE, + "Marking node %.40s as failing (quorum reached).", node->name); + + /* Mark the node as failing. */ + node->flags &= ~CLUSTER_NODE_PFAIL; + node->flags |= CLUSTER_NODE_FAIL; + node->fail_time = mstime(); + + /* Broadcast the failing node name to everybody, forcing all the other + * reachable nodes to flag the node as FAIL. */ + if (nodeIsMaster(myself)) clusterSendFail(node->name); + clusterDoBeforeSleep(CLUSTER_TODO_UPDATE_STATE | CLUSTER_TODO_SAVE_CONFIG); +} + +/* This function is called only if a node is marked as FAIL, but we are able + * to reach it again. It checks if there are the conditions to undo the FAIL + * state. */ +void clearNodeFailureIfNeeded(clusterNode *node) { + mstime_t now = mstime(); + + serverAssert(nodeFailed(node)); + + /* For slaves we always clear the FAIL flag if we can contact the + * node again. */ + if (nodeIsSlave(node) || node->numslots == 0) { + serverLog(LL_NOTICE, + "Clear FAIL state for node %.40s: %s is reachable again.", + node->name, + nodeIsSlave(node) ? "slave" : "master without slots"); + node->flags &= ~CLUSTER_NODE_FAIL; + clusterDoBeforeSleep(CLUSTER_TODO_UPDATE_STATE | CLUSTER_TODO_SAVE_CONFIG); + } + + /* If it is a master and... + * 1) The FAIL state is old enough. + * 2) It is yet serving slots from our point of view (not failed over). + * Apparently no one is going to fix these slots, clear the FAIL flag. */ + if (nodeIsMaster(node) && node->numslots > 0 && + (now - node->fail_time) > + (server.cluster_node_timeout * CLUSTER_FAIL_UNDO_TIME_MULT)) + { + serverLog(LL_NOTICE, + "Clear FAIL state for node %.40s: is reachable again and nobody is serving its slots after some time.", + node->name); + node->flags &= ~CLUSTER_NODE_FAIL; + clusterDoBeforeSleep(CLUSTER_TODO_UPDATE_STATE | CLUSTER_TODO_SAVE_CONFIG); + } +} + +/* Return true if we already have a node in HANDSHAKE state matching the + * specified ip address and port number. This function is used in order to + * avoid adding a new handshake node for the same address multiple times. */ +int clusterHandshakeInProgress(char *ip, int port, int cport) { + dictIterator *di; + dictEntry *de; + + di = dictGetSafeIterator(server.cluster->nodes); + while ((de = dictNext(di)) != NULL) { + clusterNode *node = dictGetVal(de); + + if (!nodeInHandshake(node)) continue; + if (!strcasecmp(node->ip, ip) && + node->port == port && + node->cport == cport) break; + } + dictReleaseIterator(di); + return de != NULL; +} + +/* Start an handshake with the specified address if there is not one + * already in progress. Returns non-zero if the handshake was actually + * started. On error zero is returned and errno is set to one of the + * following values: + * + * EAGAIN - There is already an handshake in progress for this address. + * EINVAL - IP or port are not valid. */ +int clusterStartHandshake(char *ip, int port, int cport) { + clusterNode *n; + char norm_ip[NET_IP_STR_LEN]; + struct sockaddr_storage sa; + + /* IP sanity check */ + if (inet_pton(AF_INET, ip, + &(((struct sockaddr_in *)&sa)->sin_addr))) + { + sa.ss_family = AF_INET; + } + else if (inet_pton(AF_INET6, ip, + &(((struct sockaddr_in6 *)&sa)->sin6_addr))) + { + sa.ss_family = AF_INET6; + } + else { + errno = EINVAL; + return 0; + } + + /* Port sanity check */ + if (port <= 0 || port > 65535 || cport <= 0 || cport > 65535) { + errno = EINVAL; + return 0; + } + + /* Set norm_ip as the normalized string representation of the node + * IP address. */ + memset(norm_ip, 0, NET_IP_STR_LEN); + if (sa.ss_family == AF_INET) + inet_ntop(AF_INET, + (void*)&(((struct sockaddr_in *)&sa)->sin_addr), + norm_ip, NET_IP_STR_LEN); + else + inet_ntop(AF_INET6, + (void*)&(((struct sockaddr_in6 *)&sa)->sin6_addr), + norm_ip, NET_IP_STR_LEN); + + if (clusterHandshakeInProgress(norm_ip, port, cport)) { + errno = EAGAIN; + return 0; + } + + /* Add the node with a random address (NULL as first argument to + * createClusterNode()). Everything will be fixed during the + * handshake. */ + n = createClusterNode(NULL, CLUSTER_NODE_HANDSHAKE | CLUSTER_NODE_MEET); + memcpy(n->ip, norm_ip, sizeof(n->ip)); + n->port = port; + n->cport = cport; + clusterAddNode(n); + return 1; +} + +/* Process the gossip section of PING or PONG packets. + * Note that this function assumes that the packet is already sanity-checked + * by the caller, not in the content of the gossip section, but in the + * length. */ +void clusterProcessGossipSection(clusterMsg *hdr, clusterLink *link) { + uint16_t count = ntohs(hdr->count); + clusterMsgDataGossip *g = (clusterMsgDataGossip*) hdr->data.ping.gossip; + clusterNode *sender = link->node ? link->node : clusterLookupNode(hdr->sender); + + while (count--) { + uint16_t flags = ntohs(g->flags); + clusterNode *node; + sds ci; + + if (server.verbosity == LL_DEBUG) { + ci = representClusterNodeFlags(sdsempty(), flags); + serverLog(LL_DEBUG, "GOSSIP %.40s %s:%d@%d %s", + g->nodename, + g->ip, + ntohs(g->port), + ntohs(g->cport), + ci); + sdsfree(ci); + } + + /* Update our state accordingly to the gossip sections */ + node = clusterLookupNode(g->nodename); + if (node) { + /* We already know this node. + Handle failure reports, only when the sender is a master. */ + if (sender && nodeIsMaster(sender) && node != myself) { + if (flags & (CLUSTER_NODE_FAIL | CLUSTER_NODE_PFAIL)) { + if (clusterNodeAddFailureReport(node, sender)) { + serverLog(LL_VERBOSE, + "Node %.40s reported node %.40s as not reachable.", + sender->name, node->name); + } + markNodeAsFailingIfNeeded(node); + } + else { + if (clusterNodeDelFailureReport(node, sender)) { + serverLog(LL_VERBOSE, + "Node %.40s reported node %.40s is back online.", + sender->name, node->name); + } + } + } + + /* If from our POV the node is up (no failure flags are set), + * we have no pending ping for the node, nor we have failure + * reports for this node, update the last pong time with the + * one we see from the other nodes. */ + if (!(flags & (CLUSTER_NODE_FAIL | CLUSTER_NODE_PFAIL)) && + node->ping_sent == 0 && + clusterNodeFailureReportsCount(node) == 0) + { + mstime_t pongtime = ntohl(g->pong_received); + pongtime *= 1000; /* Convert back to milliseconds. */ + + /* Replace the pong time with the received one only if + * it's greater than our view but is not in the future + * (with 500 milliseconds tolerance) from the POV of our + * clock. */ + if (pongtime <= (server.mstime + 500) && + pongtime > node->pong_received) + { + node->pong_received = pongtime; + } + } + + /* If we already know this node, but it is not reachable, and + * we see a different address in the gossip section of a node that + * can talk with this other node, update the address, disconnect + * the old link if any, so that we'll attempt to connect with the + * new address. */ + if (node->flags & (CLUSTER_NODE_FAIL | CLUSTER_NODE_PFAIL) && + !(flags & CLUSTER_NODE_NOADDR) && + !(flags & (CLUSTER_NODE_FAIL | CLUSTER_NODE_PFAIL)) && + (strcasecmp(node->ip, g->ip) || + node->port != ntohs(g->port) || + node->cport != ntohs(g->cport))) + { + if (node->link) freeClusterLink(node->link); + memcpy(node->ip, g->ip, NET_IP_STR_LEN); + node->port = ntohs(g->port); + node->cport = ntohs(g->cport); + node->flags &= ~CLUSTER_NODE_NOADDR; + } + } + else { + /* If it's not in NOADDR state and we don't have it, we + * start a handshake process against this IP/PORT pairs. + * + * Note that we require that the sender of this gossip message + * is a well known node in our cluster, otherwise we risk + * joining another cluster. */ + if (sender && + !(flags & CLUSTER_NODE_NOADDR) && + !clusterBlacklistExists(g->nodename)) + { + clusterStartHandshake(g->ip, ntohs(g->port), ntohs(g->cport)); + } + } + + /* Next node */ + g++; + } +} + +/* IP -> string conversion. 'buf' is supposed to at least be 46 bytes. + * If 'announced_ip' length is non-zero, it is used instead of extracting + * the IP from the socket peer address. */ +void nodeIp2String(char *buf, clusterLink *link, char *announced_ip) { + if (announced_ip[0] != '\0') { + memcpy(buf, announced_ip, NET_IP_STR_LEN); + buf[NET_IP_STR_LEN - 1] = '\0'; /* We are not sure the input is sane. */ + } + else { + anetPeerToString(link->fd, buf, NET_IP_STR_LEN, NULL); + } +} + +/* Update the node address to the IP address that can be extracted + * from link->fd, or if hdr->myip is non empty, to the address the node + * is announcing us. The port is taken from the packet header as well. + * + * If the address or port changed, disconnect the node link so that we'll + * connect again to the new address. + * + * If the ip/port pair are already correct no operation is performed at + * all. + * + * The function returns 0 if the node address is still the same, + * otherwise 1 is returned. */ +int nodeUpdateAddressIfNeeded(clusterNode *node, clusterLink *link, + clusterMsg *hdr) +{ + char ip[NET_IP_STR_LEN] = { 0 }; + int port = ntohs(hdr->port); + int cport = ntohs(hdr->cport); + + /* We don't proceed if the link is the same as the sender link, as this + * function is designed to see if the node link is consistent with the + * symmetric link that is used to receive PINGs from the node. + * + * As a side effect this function never frees the passed 'link', so + * it is safe to call during packet processing. */ + if (link == node->link) return 0; + + nodeIp2String(ip, link, hdr->myip); + if (node->port == port && node->cport == cport && + strcmp(ip, node->ip) == 0) return 0; + + /* IP / port is different, update it. */ + memcpy(node->ip, ip, sizeof(ip)); + node->port = port; + node->cport = cport; + if (node->link) freeClusterLink(node->link); + node->flags &= ~CLUSTER_NODE_NOADDR; + serverLog(LL_WARNING, "Address updated for node %.40s, now %s:%d", + node->name, node->ip, node->port); + + /* Check if this is our master and we have to change the + * replication target as well. */ + if (nodeIsSlave(myself) && myself->slaveof == node) + replicationSetMaster(node->ip, node->port); + return 1; +} + +/* Reconfigure the specified node 'n' as a master. This function is called when + * a node that we believed to be a slave is now acting as master in order to + * update the state of the node. */ +void clusterSetNodeAsMaster(clusterNode *n) { + if (nodeIsMaster(n)) return; + + if (n->slaveof) { + clusterNodeRemoveSlave(n->slaveof, n); + if (n != myself) n->flags |= CLUSTER_NODE_MIGRATE_TO; + } + n->flags &= ~CLUSTER_NODE_SLAVE; + n->flags |= CLUSTER_NODE_MASTER; + n->slaveof = NULL; + + /* Update config and state. */ + clusterDoBeforeSleep(CLUSTER_TODO_SAVE_CONFIG | + CLUSTER_TODO_UPDATE_STATE); +} + +/* This function is called when we receive a master configuration via a + * PING, PONG or UPDATE packet. What we receive is a node, a configEpoch of the + * node, and the set of slots claimed under this configEpoch. + * + * What we do is to rebind the slots with newer configuration compared to our + * local configuration, and if needed, we turn ourself into a replica of the + * node (see the function comments for more info). + * + * The 'sender' is the node for which we received a configuration update. + * Sometimes it is not actually the "Sender" of the information, like in the + * case we receive the info via an UPDATE packet. */ +void clusterUpdateSlotsConfigWith(clusterNode *sender, uint64_t senderConfigEpoch, unsigned char *slots) { + int j; + clusterNode *curmaster, *newmaster = NULL; + /* The dirty slots list is a list of slots for which we lose the ownership + * while having still keys inside. This usually happens after a failover + * or after a manual cluster reconfiguration operated by the admin. + * + * If the update message is not able to demote a master to slave (in this + * case we'll resync with the master updating the whole key space), we + * need to delete all the keys in the slots we lost ownership. */ + uint16_t dirty_slots[CLUSTER_SLOTS]; + int dirty_slots_count = 0; + + /* Here we set curmaster to this node or the node this node + * replicates to if it's a slave. In the for loop we are + * interested to check if slots are taken away from curmaster. */ + curmaster = nodeIsMaster(myself) ? myself : myself->slaveof; + + if (sender == myself) { + serverLog(LL_WARNING, "Discarding UPDATE message about myself."); + return; + } + + for (j = 0; j < CLUSTER_SLOTS; j++) { + if (bitmapTestBit(slots, j)) { + /* The slot is already bound to the sender of this message. */ + if (server.cluster->slots[j] == sender) continue; + + /* The slot is in importing state, it should be modified only + * manually via redis-trib (example: a resharding is in progress + * and the migrating side slot was already closed and is advertising + * a new config. We still want the slot to be closed manually). */ + if (server.cluster->importing_slots_from[j]) continue; + + /* We rebind the slot to the new node claiming it if: + * 1) The slot was unassigned or the new node claims it with a + * greater configEpoch. + * 2) We are not currently importing the slot. */ + if (server.cluster->slots[j] == NULL || + server.cluster->slots[j]->configEpoch < senderConfigEpoch) + { + /* Was this slot mine, and still contains keys? Mark it as + * a dirty slot. */ + if (server.cluster->slots[j] == myself && + countKeysInSlot(j) && + sender != myself) + { + dirty_slots[dirty_slots_count] = j; + dirty_slots_count++; + } + + if (server.cluster->slots[j] == curmaster) + newmaster = sender; + clusterDelSlot(j); + clusterAddSlot(sender, j); + clusterDoBeforeSleep(CLUSTER_TODO_SAVE_CONFIG | + CLUSTER_TODO_UPDATE_STATE | + CLUSTER_TODO_FSYNC_CONFIG); + } + } + } + + /* If at least one slot was reassigned from a node to another node + * with a greater configEpoch, it is possible that: + * 1) We are a master left without slots. This means that we were + * failed over and we should turn into a replica of the new + * master. + * 2) We are a slave and our master is left without slots. We need + * to replicate to the new slots owner. */ + if (newmaster && curmaster->numslots == 0) { + serverLog(LL_WARNING, + "Configuration change detected. Reconfiguring myself " + "as a replica of %.40s", sender->name); + clusterSetMaster(sender); + clusterDoBeforeSleep(CLUSTER_TODO_SAVE_CONFIG | + CLUSTER_TODO_UPDATE_STATE | + CLUSTER_TODO_FSYNC_CONFIG); + } + else if (dirty_slots_count) { + /* If we are here, we received an update message which removed + * ownership for certain slots we still have keys about, but still + * we are serving some slots, so this master node was not demoted to + * a slave. + * + * In order to maintain a consistent state between keys and slots + * we need to remove all the keys from the slots we lost. */ + for (j = 0; j < dirty_slots_count; j++) + delKeysInSlot(dirty_slots[j]); + } +} + +/* When this function is called, there is a packet to process starting + * at node->rcvbuf. Releasing the buffer is up to the caller, so this + * function should just handle the higher level stuff of processing the + * packet, modifying the cluster state if needed. + * + * The function returns 1 if the link is still valid after the packet + * was processed, otherwise 0 if the link was freed since the packet + * processing lead to some inconsistency error (for instance a PONG + * received from the wrong sender ID). */ +int clusterProcessPacket(clusterLink *link) { + clusterMsg *hdr = (clusterMsg*) link->rcvbuf; + uint32_t totlen = ntohl(hdr->totlen); + uint16_t type = ntohs(hdr->type); + + if (type < CLUSTERMSG_TYPE_COUNT) + server.cluster->stats_bus_messages_received[type]++; + serverLog(LL_DEBUG, "--- Processing packet of type %d, %Iu bytes", WIN_PORT_FIX /* %lu -> %Iu */ + type, (PORT_ULONG) totlen); + + /* Perform sanity checks */ + if (totlen < 16) return 1; /* At least signature, version, totlen, count. */ + if (totlen > sdslen(link->rcvbuf)) return 1; + + if (ntohs(hdr->ver) != CLUSTER_PROTO_VER) { + /* Can't handle messages of different versions. */ + return 1; + } + + uint16_t flags = ntohs(hdr->flags); + uint64_t senderCurrentEpoch = 0, senderConfigEpoch = 0; + clusterNode *sender; + + if (type == CLUSTERMSG_TYPE_PING || type == CLUSTERMSG_TYPE_PONG || + type == CLUSTERMSG_TYPE_MEET) + { + uint16_t count = ntohs(hdr->count); + uint32_t explen; /* expected length of this packet */ + + explen = sizeof(clusterMsg) - sizeof(union clusterMsgData); + explen += (sizeof(clusterMsgDataGossip)*count); + if (totlen != explen) return 1; + } + else if (type == CLUSTERMSG_TYPE_FAIL) { + uint32_t explen = sizeof(clusterMsg) - sizeof(union clusterMsgData); + + explen += sizeof(clusterMsgDataFail); + if (totlen != explen) return 1; + } + else if (type == CLUSTERMSG_TYPE_PUBLISH) { + uint32_t explen = sizeof(clusterMsg) - sizeof(union clusterMsgData); + + explen += sizeof(clusterMsgDataPublish) - + 8 + + ntohl(hdr->data.publish.msg.channel_len) + + ntohl(hdr->data.publish.msg.message_len); + if (totlen != explen) return 1; + } + else if (type == CLUSTERMSG_TYPE_FAILOVER_AUTH_REQUEST || + type == CLUSTERMSG_TYPE_FAILOVER_AUTH_ACK || + type == CLUSTERMSG_TYPE_MFSTART) + { + uint32_t explen = sizeof(clusterMsg) - sizeof(union clusterMsgData); + + if (totlen != explen) return 1; + } + else if (type == CLUSTERMSG_TYPE_UPDATE) { + uint32_t explen = sizeof(clusterMsg) - sizeof(union clusterMsgData); + + explen += sizeof(clusterMsgDataUpdate); + if (totlen != explen) return 1; + } + + /* Check if the sender is a known node. */ + sender = clusterLookupNode(hdr->sender); + if (sender && !nodeInHandshake(sender)) { + /* Update our curretEpoch if we see a newer epoch in the cluster. */ + senderCurrentEpoch = ntohu64(hdr->currentEpoch); + senderConfigEpoch = ntohu64(hdr->configEpoch); + if (senderCurrentEpoch > server.cluster->currentEpoch) + server.cluster->currentEpoch = senderCurrentEpoch; + /* Update the sender configEpoch if it is publishing a newer one. */ + if (senderConfigEpoch > sender->configEpoch) { + sender->configEpoch = senderConfigEpoch; + clusterDoBeforeSleep(CLUSTER_TODO_SAVE_CONFIG | + CLUSTER_TODO_FSYNC_CONFIG); + } + /* Update the replication offset info for this node. */ + sender->repl_offset = ntohu64(hdr->offset); + sender->repl_offset_time = mstime(); + /* If we are a slave performing a manual failover and our master + * sent its offset while already paused, populate the MF state. */ + if (server.cluster->mf_end && + nodeIsSlave(myself) && + myself->slaveof == sender && + hdr->mflags[0] & CLUSTERMSG_FLAG0_PAUSED && + server.cluster->mf_master_offset == 0) + { + server.cluster->mf_master_offset = sender->repl_offset; + serverLog(LL_WARNING, + "Received replication offset for paused " + "master manual failover: %lld", + server.cluster->mf_master_offset); + } + } + + /* Initial processing of PING and MEET requests replying with a PONG. */ + if (type == CLUSTERMSG_TYPE_PING || type == CLUSTERMSG_TYPE_MEET) { + serverLog(LL_DEBUG, "Ping packet received: %p", (void*) link->node); + + /* We use incoming MEET messages in order to set the address + * for 'myself', since only other cluster nodes will send us + * MEET messages on handshakes, when the cluster joins, or + * later if we changed address, and those nodes will use our + * official address to connect to us. So by obtaining this address + * from the socket is a simple way to discover / update our own + * address in the cluster without it being hardcoded in the config. + * + * However if we don't have an address at all, we update the address + * even with a normal PING packet. If it's wrong it will be fixed + * by MEET later. */ + if ((type == CLUSTERMSG_TYPE_MEET || myself->ip[0] == '\0') && + server.cluster_announce_ip == NULL) + { + char ip[NET_IP_STR_LEN]; + + if (anetSockName(link->fd, ip, sizeof(ip), NULL) != -1 && + strcmp(ip, myself->ip)) + { + memcpy(myself->ip, ip, NET_IP_STR_LEN); + serverLog(LL_WARNING, "IP address for this node updated to %s", + myself->ip); + clusterDoBeforeSleep(CLUSTER_TODO_SAVE_CONFIG); + } + } + + /* Add this node if it is new for us and the msg type is MEET. + * In this stage we don't try to add the node with the right + * flags, slaveof pointer, and so forth, as this details will be + * resolved when we'll receive PONGs from the node. */ + if (!sender && type == CLUSTERMSG_TYPE_MEET) { + clusterNode *node; + + node = createClusterNode(NULL, CLUSTER_NODE_HANDSHAKE); + nodeIp2String(node->ip, link, hdr->myip); + node->port = ntohs(hdr->port); + node->cport = ntohs(hdr->cport); + clusterAddNode(node); + clusterDoBeforeSleep(CLUSTER_TODO_SAVE_CONFIG); + } + + /* If this is a MEET packet from an unknown node, we still process + * the gossip section here since we have to trust the sender because + * of the message type. */ + if (!sender && type == CLUSTERMSG_TYPE_MEET) + clusterProcessGossipSection(hdr, link); + + /* Anyway reply with a PONG */ + clusterSendPing(link, CLUSTERMSG_TYPE_PONG); + } + + /* PING, PONG, MEET: process config information. */ + if (type == CLUSTERMSG_TYPE_PING || type == CLUSTERMSG_TYPE_PONG || + type == CLUSTERMSG_TYPE_MEET) + { + serverLog(LL_DEBUG, "%s packet received: %p", + type == CLUSTERMSG_TYPE_PING ? "ping" : "pong", + (void*) link->node); + if (link->node) { + if (nodeInHandshake(link->node)) { + /* If we already have this node, try to change the + * IP/port of the node with the new one. */ + if (sender) { + serverLog(LL_VERBOSE, + "Handshake: we already know node %.40s, " + "updating the address if needed.", sender->name); + if (nodeUpdateAddressIfNeeded(sender, link, hdr)) + { + clusterDoBeforeSleep(CLUSTER_TODO_SAVE_CONFIG | + CLUSTER_TODO_UPDATE_STATE); + } + /* Free this node as we already have it. This will + * cause the link to be freed as well. */ + clusterDelNode(link->node); + return 0; + } + + /* First thing to do is replacing the random name with the + * right node name if this was a handshake stage. */ + clusterRenameNode(link->node, hdr->sender); + serverLog(LL_DEBUG, "Handshake with node %.40s completed.", + link->node->name); + link->node->flags &= ~CLUSTER_NODE_HANDSHAKE; + link->node->flags |= flags & (CLUSTER_NODE_MASTER | CLUSTER_NODE_SLAVE); + clusterDoBeforeSleep(CLUSTER_TODO_SAVE_CONFIG); + } + else if (memcmp(link->node->name, hdr->sender, + CLUSTER_NAMELEN) != 0) + { + /* If the reply has a non matching node ID we + * disconnect this node and set it as not having an associated + * address. */ + serverLog(LL_DEBUG, "PONG contains mismatching sender ID. About node %.40s added %d ms ago, having flags %d", + link->node->name, + (int) (mstime() - (link->node->ctime)), + link->node->flags); + link->node->flags |= CLUSTER_NODE_NOADDR; + link->node->ip[0] = '\0'; + link->node->port = 0; + link->node->cport = 0; + freeClusterLink(link); + clusterDoBeforeSleep(CLUSTER_TODO_SAVE_CONFIG); + return 0; + } + } + + /* Update the node address if it changed. */ + if (sender && type == CLUSTERMSG_TYPE_PING && + !nodeInHandshake(sender) && + nodeUpdateAddressIfNeeded(sender, link, hdr)) + { + clusterDoBeforeSleep(CLUSTER_TODO_SAVE_CONFIG | + CLUSTER_TODO_UPDATE_STATE); + } + + /* Update our info about the node */ + if (link->node && type == CLUSTERMSG_TYPE_PONG) { + link->node->pong_received = mstime(); + link->node->ping_sent = 0; + + /* The PFAIL condition can be reversed without external + * help if it is momentary (that is, if it does not + * turn into a FAIL state). + * + * The FAIL condition is also reversible under specific + * conditions detected by clearNodeFailureIfNeeded(). */ + if (nodeTimedOut(link->node)) { + link->node->flags &= ~CLUSTER_NODE_PFAIL; + clusterDoBeforeSleep(CLUSTER_TODO_SAVE_CONFIG | + CLUSTER_TODO_UPDATE_STATE); + } + else if (nodeFailed(link->node)) { + clearNodeFailureIfNeeded(link->node); + } + } + + /* Check for role switch: slave -> master or master -> slave. */ + if (sender) { + if (!memcmp(hdr->slaveof, CLUSTER_NODE_NULL_NAME, + sizeof(hdr->slaveof))) + { + /* Node is a master. */ + clusterSetNodeAsMaster(sender); + } + else { + /* Node is a slave. */ + clusterNode *master = clusterLookupNode(hdr->slaveof); + + if (nodeIsMaster(sender)) { + /* Master turned into a slave! Reconfigure the node. */ + clusterDelNodeSlots(sender); + sender->flags &= ~(CLUSTER_NODE_MASTER | + CLUSTER_NODE_MIGRATE_TO); + sender->flags |= CLUSTER_NODE_SLAVE; + + /* Update config and state. */ + clusterDoBeforeSleep(CLUSTER_TODO_SAVE_CONFIG | + CLUSTER_TODO_UPDATE_STATE); + } + + /* Master node changed for this slave? */ + if (master && sender->slaveof != master) { + if (sender->slaveof) + clusterNodeRemoveSlave(sender->slaveof, sender); + clusterNodeAddSlave(master, sender); + sender->slaveof = master; + + /* Update config. */ + clusterDoBeforeSleep(CLUSTER_TODO_SAVE_CONFIG); + } + } + } + + /* Update our info about served slots. + * + * Note: this MUST happen after we update the master/slave state + * so that CLUSTER_NODE_MASTER flag will be set. */ + + /* Many checks are only needed if the set of served slots this + * instance claims is different compared to the set of slots we have + * for it. Check this ASAP to avoid other computational expansive + * checks later. */ + clusterNode *sender_master = NULL; /* Sender or its master if slave. */ + int dirty_slots = 0; /* Sender claimed slots don't match my view? */ + + if (sender) { + sender_master = nodeIsMaster(sender) ? sender : sender->slaveof; + if (sender_master) { + dirty_slots = memcmp(sender_master->slots, + hdr->myslots, sizeof(hdr->myslots)) != 0; + } + } + + /* 1) If the sender of the message is a master, and we detected that + * the set of slots it claims changed, scan the slots to see if we + * need to update our configuration. */ + if (sender && nodeIsMaster(sender) && dirty_slots) + clusterUpdateSlotsConfigWith(sender, senderConfigEpoch, hdr->myslots); + + /* 2) We also check for the reverse condition, that is, the sender + * claims to serve slots we know are served by a master with a + * greater configEpoch. If this happens we inform the sender. + * + * This is useful because sometimes after a partition heals, a + * reappearing master may be the last one to claim a given set of + * hash slots, but with a configuration that other instances know to + * be deprecated. Example: + * + * A and B are master and slave for slots 1,2,3. + * A is partitioned away, B gets promoted. + * B is partitioned away, and A returns available. + * + * Usually B would PING A publishing its set of served slots and its + * configEpoch, but because of the partition B can't inform A of the + * new configuration, so other nodes that have an updated table must + * do it. In this way A will stop to act as a master (or can try to + * failover if there are the conditions to win the election). */ + if (sender && dirty_slots) { + int j; + + for (j = 0; j < CLUSTER_SLOTS; j++) { + if (bitmapTestBit(hdr->myslots, j)) { + if (server.cluster->slots[j] == sender || + server.cluster->slots[j] == NULL) continue; + if (server.cluster->slots[j]->configEpoch > + senderConfigEpoch) + { + serverLog(LL_VERBOSE, + "Node %.40s has old slots configuration, sending " + "an UPDATE message about %.40s", + sender->name, server.cluster->slots[j]->name); + clusterSendUpdate(sender->link, + server.cluster->slots[j]); + + /* TODO: instead of exiting the loop send every other + * UPDATE packet for other nodes that are the new owner + * of sender's slots. */ + break; + } + } + } + } + + /* If our config epoch collides with the sender's try to fix + * the problem. */ + if (sender && + nodeIsMaster(myself) && nodeIsMaster(sender) && + senderConfigEpoch == myself->configEpoch) + { + clusterHandleConfigEpochCollision(sender); + } + + /* Get info from the gossip section */ + if (sender) clusterProcessGossipSection(hdr, link); + } + else if (type == CLUSTERMSG_TYPE_FAIL) { + clusterNode *failing; + + if (sender) { + failing = clusterLookupNode(hdr->data.fail.about.nodename); + if (failing && + !(failing->flags & (CLUSTER_NODE_FAIL | CLUSTER_NODE_MYSELF))) + { + serverLog(LL_NOTICE, + "FAIL message received from %.40s about %.40s", + hdr->sender, hdr->data.fail.about.nodename); + failing->flags |= CLUSTER_NODE_FAIL; + failing->fail_time = mstime(); + failing->flags &= ~CLUSTER_NODE_PFAIL; + clusterDoBeforeSleep(CLUSTER_TODO_SAVE_CONFIG | + CLUSTER_TODO_UPDATE_STATE); + } + } + else { + serverLog(LL_NOTICE, + "Ignoring FAIL message from unknown node %.40s about %.40s", + hdr->sender, hdr->data.fail.about.nodename); + } + } + else if (type == CLUSTERMSG_TYPE_PUBLISH) { + robj *channel, *message; + uint32_t channel_len, message_len; + + /* Don't bother creating useless objects if there are no + * Pub/Sub subscribers. */ + if (dictSize(server.pubsub_channels) || + listLength(server.pubsub_patterns)) + { + channel_len = ntohl(hdr->data.publish.msg.channel_len); + message_len = ntohl(hdr->data.publish.msg.message_len); + channel = createStringObject( + (char*) hdr->data.publish.msg.bulk_data, channel_len); + message = createStringObject( + (char*) hdr->data.publish.msg.bulk_data + channel_len, + message_len); + pubsubPublishMessage(channel, message); + decrRefCount(channel); + decrRefCount(message); + } + } + else if (type == CLUSTERMSG_TYPE_FAILOVER_AUTH_REQUEST) { + if (!sender) return 1; /* We don't know that node. */ + clusterSendFailoverAuthIfNeeded(sender, hdr); + } + else if (type == CLUSTERMSG_TYPE_FAILOVER_AUTH_ACK) { + if (!sender) return 1; /* We don't know that node. */ + /* We consider this vote only if the sender is a master serving + * a non zero number of slots, and its currentEpoch is greater or + * equal to epoch where this node started the election. */ + if (nodeIsMaster(sender) && sender->numslots > 0 && + senderCurrentEpoch >= server.cluster->failover_auth_epoch) + { + server.cluster->failover_auth_count++; + /* Maybe we reached a quorum here, set a flag to make sure + * we check ASAP. */ + clusterDoBeforeSleep(CLUSTER_TODO_HANDLE_FAILOVER); + } + } + else if (type == CLUSTERMSG_TYPE_MFSTART) { + /* This message is acceptable only if I'm a master and the sender + * is one of my slaves. */ + if (!sender || sender->slaveof != myself) return 1; + /* Manual failover requested from slaves. Initialize the state + * accordingly. */ + resetManualFailover(); + server.cluster->mf_end = mstime() + CLUSTER_MF_TIMEOUT; + server.cluster->mf_slave = sender; + pauseClients(mstime() + (CLUSTER_MF_TIMEOUT * 2)); + serverLog(LL_WARNING, "Manual failover requested by slave %.40s.", + sender->name); + } + else if (type == CLUSTERMSG_TYPE_UPDATE) { + clusterNode *n; /* The node the update is about. */ + uint64_t reportedConfigEpoch = + ntohu64(hdr->data.update.nodecfg.configEpoch); + + if (!sender) return 1; /* We don't know the sender. */ + n = clusterLookupNode(hdr->data.update.nodecfg.nodename); + if (!n) return 1; /* We don't know the reported node. */ + if (n->configEpoch >= reportedConfigEpoch) return 1; /* Nothing new. */ + + /* If in our current config the node is a slave, set it as a master. */ + if (nodeIsSlave(n)) clusterSetNodeAsMaster(n); + + /* Update the node's configEpoch. */ + n->configEpoch = reportedConfigEpoch; + clusterDoBeforeSleep(CLUSTER_TODO_SAVE_CONFIG | + CLUSTER_TODO_FSYNC_CONFIG); + + /* Check the bitmap of served slots and update our + * config accordingly. */ + clusterUpdateSlotsConfigWith(n, reportedConfigEpoch, + hdr->data.update.nodecfg.slots); + } + else { + serverLog(LL_WARNING, "Received unknown packet type: %d", type); + } + return 1; +} + +/* This function is called when we detect the link with this node is lost. + We set the node as no longer connected. The Cluster Cron will detect + this connection and will try to get it connected again. + + Instead if the node is a temporary node used to accept a query, we + completely free the node on error. */ +void handleLinkIOError(clusterLink *link) { + freeClusterLink(link); +} + +#ifdef _WIN32 +void clusterWriteDone(aeEventLoop *el, int fd, void *privdata, int written) { + WSIOCP_Request *req = (WSIOCP_Request *) privdata; + clusterLink *link = (clusterLink *) req->client; + UNUSED(el); + UNUSED(fd); + + if (sdslen(link->sndbuf) == written) { + sdsrange(link->sndbuf, written, -1); + aeDeleteFileEvent(server.el, link->fd, AE_WRITABLE); + serverLog(LL_DEBUG, "clusterWriteDone written %d fd %d", written, link->fd); + } +} + +void clusterWriteHandler(aeEventLoop *el, int fd, void *privdata, int mask) { + clusterLink *link = (clusterLink*) privdata; + UNUSED(el); + UNUSED(mask); + + int result = WSIOCP_SocketSend(fd, + (char*) link->sndbuf, + (int) (sdslen(link->sndbuf)), + el, + link, + NULL, + clusterWriteDone); + if (errno == WSA_IO_PENDING) + serverLog(LL_DEBUG, "WSA_IO_PENDING writing to socket fd %d", link->fd); + + if (result == SOCKET_ERROR && errno != WSA_IO_PENDING) { + serverLog(LL_WARNING, "Error writing to socket fd %d", link->fd); + handleLinkIOError(link); + return; + } +} +#else +/* Send data. This is handled using a trivial send buffer that gets + * consumed by write(). We don't try to optimize this for speed too much + * as this is a very low traffic channel. */ +void clusterWriteHandler(aeEventLoop *el, int fd, void *privdata, int mask) { + clusterLink *link = (clusterLink*) privdata; + ssize_t nwritten; + UNUSED(el); + UNUSED(mask); + + nwritten = write(fd, link->sndbuf, sdslen(link->sndbuf)); + if (nwritten <= 0) { + serverLog(LL_DEBUG, "I/O error writing to node link: %s", + strerror(errno)); + handleLinkIOError(link); + return; + } + sdsrange(link->sndbuf, nwritten, -1); + if (sdslen(link->sndbuf) == 0) + aeDeleteFileEvent(server.el, link->fd, AE_WRITABLE); +} +#endif + +/* Read data. Try to read the first field of the header first to check the + * full length of the packet. When a whole packet is in memory this function + * will call the function to process the packet. And so forth. */ +void clusterReadHandler(aeEventLoop *el, int fd, void *privdata, int mask) { + char buf[sizeof(clusterMsg)]; + ssize_t nread; + clusterMsg *hdr; + clusterLink *link = (clusterLink*) privdata; + unsigned int readlen, rcvbuflen; + UNUSED(el); + UNUSED(mask); + + while (1) { /* Read as long as there is data to read. */ + rcvbuflen = (unsigned int) sdslen(link->rcvbuf); WIN_PORT_FIX /* cast (unsigned int) */ + if (rcvbuflen < 8) { + /* First, obtain the first 8 bytes to get the full message + * length. */ + readlen = 8 - rcvbuflen; + } + else { + /* Finally read the full message. */ + hdr = (clusterMsg*) link->rcvbuf; + if (rcvbuflen == 8) { + /* Perform some sanity check on the message signature + * and length. */ + if (memcmp(hdr->sig, "RCmb", 4) != 0 || + ntohl(hdr->totlen) < CLUSTERMSG_MIN_LEN) + { + serverLog(LL_WARNING, + "Bad message length or signature received " + "from Cluster bus."); + handleLinkIOError(link); + return; + } + } + readlen = ntohl(hdr->totlen) - rcvbuflen; + if (readlen > sizeof(buf)) readlen = sizeof(buf); + } + + nread = read(fd, buf, readlen); + if (nread == -1 && errno == EAGAIN) { WIN32_ONLY(WSIOCP_QueueNextRead(fd);) return; } /* No more data ready. */ + + if (nread <= 0) { + /* I/O error... */ + serverLog(LL_DEBUG, "I/O error reading from node link: %s", + (nread == 0) ? "connection closed" : IF_WIN32(wsa_strerror(errno), strerror(errno))); + handleLinkIOError(link); + return; + } + else { + /* Read data and recast the pointer to the new buffer. */ + link->rcvbuf = sdscatlen(link->rcvbuf, buf, nread); + hdr = (clusterMsg*) link->rcvbuf; + rcvbuflen += (unsigned int) nread; WIN_PORT_FIX /* cast (unsigned int) */ + } + + /* Total length obtained? Process this packet. */ + if (rcvbuflen >= 8 && rcvbuflen == ntohl(hdr->totlen)) { + if (clusterProcessPacket(link)) { + sdsfree(link->rcvbuf); + link->rcvbuf = sdsempty(); + } + else { + return; /* Link no longer valid. */ + } + } + } + WIN32_ONLY(WSIOCP_QueueNextRead(fd);) +} + +/* Put stuff into the send buffer. + * + * It is guaranteed that this function will never have as a side effect + * the link to be invalidated, so it is safe to call this function + * from event handlers that will do stuff with the same link later. */ +void clusterSendMessage(clusterLink *link, unsigned char *msg, size_t msglen) { + if (sdslen(link->sndbuf) == 0 && msglen != 0) + aeCreateFileEvent(server.el, link->fd, AE_WRITABLE, + clusterWriteHandler, link); + + link->sndbuf = sdscatlen(link->sndbuf, msg, msglen); + + /* Populate sent messages stats. */ + clusterMsg *hdr = (clusterMsg*) msg; + uint16_t type = ntohs(hdr->type); + if (type < CLUSTERMSG_TYPE_COUNT) + server.cluster->stats_bus_messages_sent[type]++; +} + +/* Send a message to all the nodes that are part of the cluster having + * a connected link. + * + * It is guaranteed that this function will never have as a side effect + * some node->link to be invalidated, so it is safe to call this function + * from event handlers that will do stuff with node links later. */ +void clusterBroadcastMessage(void *buf, size_t len) { + dictIterator *di; + dictEntry *de; + + di = dictGetSafeIterator(server.cluster->nodes); + while ((de = dictNext(di)) != NULL) { + clusterNode *node = dictGetVal(de); + + if (!node->link) continue; + if (node->flags & (CLUSTER_NODE_MYSELF | CLUSTER_NODE_HANDSHAKE)) + continue; + clusterSendMessage(node->link, buf, len); + } + dictReleaseIterator(di); +} + +/* Build the message header. hdr must point to a buffer at least + * sizeof(clusterMsg) in bytes. */ +void clusterBuildMessageHdr(clusterMsg *hdr, int type) { + int totlen = 0; + uint64_t offset; + clusterNode *master; + + /* If this node is a master, we send its slots bitmap and configEpoch. + * If this node is a slave we send the master's information instead (the + * node is flagged as slave so the receiver knows that it is NOT really + * in charge for this slots. */ + master = (nodeIsSlave(myself) && myself->slaveof) ? + myself->slaveof : myself; + + memset(hdr, 0, sizeof(*hdr)); + hdr->ver = htons(CLUSTER_PROTO_VER); + hdr->sig[0] = 'R'; + hdr->sig[1] = 'C'; + hdr->sig[2] = 'm'; + hdr->sig[3] = 'b'; + hdr->type = htons(type); + memcpy(hdr->sender, myself->name, CLUSTER_NAMELEN); + + /* If cluster-announce-ip option is enabled, force the receivers of our + * packets to use the specified address for this node. Otherwise if the + * first byte is zero, they'll do auto discovery. */ + memset(hdr->myip, 0, NET_IP_STR_LEN); + if (server.cluster_announce_ip) { + strncpy(hdr->myip, server.cluster_announce_ip, NET_IP_STR_LEN); + hdr->myip[NET_IP_STR_LEN - 1] = '\0'; + } + + /* Handle cluster-announce-port as well. */ + int announced_port = server.cluster_announce_port ? + server.cluster_announce_port : server.port; + int announced_cport = server.cluster_announce_bus_port ? + server.cluster_announce_bus_port : + (server.port + CLUSTER_PORT_INCR); + + memcpy(hdr->myslots, master->slots, sizeof(hdr->myslots)); + memset(hdr->slaveof, 0, CLUSTER_NAMELEN); + if (myself->slaveof != NULL) + memcpy(hdr->slaveof, myself->slaveof->name, CLUSTER_NAMELEN); + hdr->port = htons(announced_port); + hdr->cport = htons(announced_cport); + hdr->flags = htons(myself->flags); + hdr->state = server.cluster->state; + + /* Set the currentEpoch and configEpochs. */ + hdr->currentEpoch = htonu64(server.cluster->currentEpoch); + hdr->configEpoch = htonu64(master->configEpoch); + + /* Set the replication offset. */ + if (nodeIsSlave(myself)) + offset = replicationGetSlaveOffset(); + else + offset = server.master_repl_offset; + hdr->offset = htonu64(offset); + + /* Set the message flags. */ + if (nodeIsMaster(myself) && server.cluster->mf_end) + hdr->mflags[0] |= CLUSTERMSG_FLAG0_PAUSED; + + /* Compute the message length for certain messages. For other messages + * this is up to the caller. */ + if (type == CLUSTERMSG_TYPE_FAIL) { + totlen = sizeof(clusterMsg) - sizeof(union clusterMsgData); + totlen += sizeof(clusterMsgDataFail); + } + else if (type == CLUSTERMSG_TYPE_UPDATE) { + totlen = sizeof(clusterMsg) - sizeof(union clusterMsgData); + totlen += sizeof(clusterMsgDataUpdate); + } + hdr->totlen = htonl(totlen); + /* For PING, PONG, and MEET, fixing the totlen field is up to the caller. */ +} + +/* Return non zero if the node is already present in the gossip section of the + * message pointed by 'hdr' and having 'count' gossip entries. Otherwise + * zero is returned. Helper for clusterSendPing(). */ +int clusterNodeIsInGossipSection(clusterMsg *hdr, int count, clusterNode *n) { + int j; + for (j = 0; j < count; j++) { + if (memcmp(hdr->data.ping.gossip[j].nodename, n->name, + CLUSTER_NAMELEN) == 0) break; + } + return j != count; +} + +/* Set the i-th entry of the gossip section in the message pointed by 'hdr' + * to the info of the specified node 'n'. */ +void clusterSetGossipEntry(clusterMsg *hdr, int i, clusterNode *n) { + clusterMsgDataGossip *gossip; + gossip = &(hdr->data.ping.gossip[i]); + memcpy(gossip->nodename, n->name, CLUSTER_NAMELEN); + gossip->ping_sent = htonl(n->ping_sent / 1000); + gossip->pong_received = htonl(n->pong_received / 1000); + memcpy(gossip->ip, n->ip, sizeof(n->ip)); + gossip->port = htons(n->port); + gossip->cport = htons(n->cport); + gossip->flags = htons(n->flags); + gossip->notused1 = 0; +} + +/* Send a PING or PONG packet to the specified node, making sure to add enough + * gossip informations. */ +void clusterSendPing(clusterLink *link, int type) { + unsigned char *buf; + clusterMsg *hdr; + int gossipcount = 0; /* Number of gossip sections added so far. */ + int wanted; /* Number of gossip sections we want to append if possible. */ + int totlen; /* Total packet length. */ + /* freshnodes is the max number of nodes we can hope to append at all: + * nodes available minus two (ourself and the node we are sending the + * message to). However practically there may be less valid nodes since + * nodes in handshake state, disconnected, are not considered. */ + int freshnodes = (int) dictSize(server.cluster->nodes) - 2; WIN_PORT_FIX /* cast (int) */ + + /* How many gossip sections we want to add? 1/10 of the number of nodes + * and anyway at least 3. Why 1/10? + * + * If we have N masters, with N/10 entries, and we consider that in + * node_timeout we exchange with each other node at least 4 packets + * (we ping in the worst case in node_timeout/2 time, and we also + * receive two pings from the host), we have a total of 8 packets + * in the node_timeout*2 falure reports validity time. So we have + * that, for a single PFAIL node, we can expect to receive the following + * number of failure reports (in the specified window of time): + * + * PROB * GOSSIP_ENTRIES_PER_PACKET * TOTAL_PACKETS: + * + * PROB = probability of being featured in a single gossip entry, + * which is 1 / NUM_OF_NODES. + * ENTRIES = 10. + * TOTAL_PACKETS = 2 * 4 * NUM_OF_MASTERS. + * + * If we assume we have just masters (so num of nodes and num of masters + * is the same), with 1/10 we always get over the majority, and specifically + * 80% of the number of nodes, to account for many masters failing at the + * same time. + * + * Since we have non-voting slaves that lower the probability of an entry + * to feature our node, we set the number of entires per packet as + * 10% of the total nodes we have. */ + wanted = floor(dictSize(server.cluster->nodes) / 10); + if (wanted < 3) wanted = 3; + if (wanted > freshnodes) wanted = freshnodes; + + /* Include all the nodes in PFAIL state, so that failure reports are + * faster to propagate to go from PFAIL to FAIL state. */ + int pfail_wanted = server.cluster->stats_pfail_nodes; + + /* Compute the maxium totlen to allocate our buffer. We'll fix the totlen + * later according to the number of gossip sections we really were able + * to put inside the packet. */ + totlen = sizeof(clusterMsg) - sizeof(union clusterMsgData); + totlen += (sizeof(clusterMsgDataGossip)*(wanted + pfail_wanted)); + /* Note: clusterBuildMessageHdr() expects the buffer to be always at least + * sizeof(clusterMsg) or more. */ + if (totlen < (int)sizeof(clusterMsg)) totlen = sizeof(clusterMsg); + buf = zcalloc(totlen); + hdr = (clusterMsg*) buf; + + /* Populate the header. */ + if (link->node && type == CLUSTERMSG_TYPE_PING) + link->node->ping_sent = mstime(); + clusterBuildMessageHdr(hdr, type); + + /* Populate the gossip fields */ + int maxiterations = wanted * 3; + while (freshnodes > 0 && gossipcount < wanted && maxiterations--) { + dictEntry *de = dictGetRandomKey(server.cluster->nodes); + clusterNode *this = dictGetVal(de); + + /* Don't include this node: the whole packet header is about us + * already, so we just gossip about other nodes. */ + if (this == myself) continue; + + /* PFAIL nodes will be added later. */ + if (this->flags & CLUSTER_NODE_PFAIL) continue; + + /* In the gossip section don't include: + * 1) Nodes in HANDSHAKE state. + * 3) Nodes with the NOADDR flag set. + * 4) Disconnected nodes if they don't have configured slots. + */ + if (this->flags & (CLUSTER_NODE_HANDSHAKE | CLUSTER_NODE_NOADDR) || + (this->link == NULL && this->numslots == 0)) + { + freshnodes--; /* Tecnically not correct, but saves CPU. */ + continue; + } + + /* Do not add a node we already have. */ + if (clusterNodeIsInGossipSection(hdr, gossipcount, this)) continue; + + /* Add it */ + clusterSetGossipEntry(hdr, gossipcount, this); + freshnodes--; + gossipcount++; + } + + /* If there are PFAIL nodes, add them at the end. */ + if (pfail_wanted) { + dictIterator *di; + dictEntry *de; + + di = dictGetSafeIterator(server.cluster->nodes); + while ((de = dictNext(di)) != NULL && pfail_wanted > 0) { + clusterNode *node = dictGetVal(de); + if (node->flags & CLUSTER_NODE_HANDSHAKE) continue; + if (node->flags & CLUSTER_NODE_NOADDR) continue; + if (!(node->flags & CLUSTER_NODE_PFAIL)) continue; + clusterSetGossipEntry(hdr, gossipcount, node); + freshnodes--; + gossipcount++; + /* We take the count of the slots we allocated, since the + * PFAIL stats may not match perfectly with the current number + * of PFAIL nodes. */ + pfail_wanted--; + } + dictReleaseIterator(di); + } + + /* Ready to send... fix the totlen fiend and queue the message in the + * output buffer. */ + totlen = sizeof(clusterMsg) - sizeof(union clusterMsgData); + totlen += (sizeof(clusterMsgDataGossip)*gossipcount); + hdr->count = htons(gossipcount); + hdr->totlen = htonl(totlen); + clusterSendMessage(link, buf, totlen); + zfree(buf); +} + +/* Send a PONG packet to every connected node that's not in handshake state + * and for which we have a valid link. + * + * In Redis Cluster pongs are not used just for failure detection, but also + * to carry important configuration information. So broadcasting a pong is + * useful when something changes in the configuration and we want to make + * the cluster aware ASAP (for instance after a slave promotion). + * + * The 'target' argument specifies the receiving instances using the + * defines below: + * + * CLUSTER_BROADCAST_ALL -> All known instances. + * CLUSTER_BROADCAST_LOCAL_SLAVES -> All slaves in my master-slaves ring. + */ +#define CLUSTER_BROADCAST_ALL 0 +#define CLUSTER_BROADCAST_LOCAL_SLAVES 1 +void clusterBroadcastPong(int target) { + dictIterator *di; + dictEntry *de; + + di = dictGetSafeIterator(server.cluster->nodes); + while ((de = dictNext(di)) != NULL) { + clusterNode *node = dictGetVal(de); + + if (!node->link) continue; + if (node == myself || nodeInHandshake(node)) continue; + if (target == CLUSTER_BROADCAST_LOCAL_SLAVES) { + int local_slave = + nodeIsSlave(node) && node->slaveof && + (node->slaveof == myself || node->slaveof == myself->slaveof); + if (!local_slave) continue; + } + clusterSendPing(node->link, CLUSTERMSG_TYPE_PONG); + } + dictReleaseIterator(di); +} + +/* Send a PUBLISH message. + * + * If link is NULL, then the message is broadcasted to the whole cluster. */ +void clusterSendPublish(clusterLink *link, robj *channel, robj *message) { + unsigned char buf[sizeof(clusterMsg)], *payload; + clusterMsg *hdr = (clusterMsg*) buf; + uint32_t totlen; + uint32_t channel_len, message_len; + + channel = getDecodedObject(channel); + message = getDecodedObject(message); + channel_len = (uint32_t) sdslen(channel->ptr); WIN_PORT_FIX /* cast (uint32_t) */ + message_len = (uint32_t) sdslen(message->ptr); WIN_PORT_FIX /* cast (uint32_t) */ + + clusterBuildMessageHdr(hdr, CLUSTERMSG_TYPE_PUBLISH); + totlen = sizeof(clusterMsg) - sizeof(union clusterMsgData); + totlen += sizeof(clusterMsgDataPublish) - 8 + channel_len + message_len; + + hdr->data.publish.msg.channel_len = htonl(channel_len); + hdr->data.publish.msg.message_len = htonl(message_len); + hdr->totlen = htonl(totlen); + + /* Try to use the local buffer if possible */ + if (totlen < sizeof(buf)) { + payload = buf; + } + else { + payload = zmalloc(totlen); + memcpy(payload, hdr, sizeof(*hdr)); + hdr = (clusterMsg*) payload; + } + memcpy(hdr->data.publish.msg.bulk_data, channel->ptr, sdslen(channel->ptr)); + memcpy(hdr->data.publish.msg.bulk_data + sdslen(channel->ptr), + message->ptr, sdslen(message->ptr)); + + if (link) + clusterSendMessage(link, payload, totlen); + else + clusterBroadcastMessage(payload, totlen); + + decrRefCount(channel); + decrRefCount(message); + if (payload != buf) zfree(payload); +} + +/* Send a FAIL message to all the nodes we are able to contact. + * The FAIL message is sent when we detect that a node is failing + * (CLUSTER_NODE_PFAIL) and we also receive a gossip confirmation of this: + * we switch the node state to CLUSTER_NODE_FAIL and ask all the other + * nodes to do the same ASAP. */ +void clusterSendFail(char *nodename) { + unsigned char buf[sizeof(clusterMsg)]; + clusterMsg *hdr = (clusterMsg*) buf; + + clusterBuildMessageHdr(hdr, CLUSTERMSG_TYPE_FAIL); + memcpy(hdr->data.fail.about.nodename, nodename, CLUSTER_NAMELEN); + clusterBroadcastMessage(buf, ntohl(hdr->totlen)); +} + +/* Send an UPDATE message to the specified link carrying the specified 'node' + * slots configuration. The node name, slots bitmap, and configEpoch info + * are included. */ +void clusterSendUpdate(clusterLink *link, clusterNode *node) { + unsigned char buf[sizeof(clusterMsg)]; + clusterMsg *hdr = (clusterMsg*) buf; + + if (link == NULL) return; + clusterBuildMessageHdr(hdr, CLUSTERMSG_TYPE_UPDATE); + memcpy(hdr->data.update.nodecfg.nodename, node->name, CLUSTER_NAMELEN); + hdr->data.update.nodecfg.configEpoch = htonu64(node->configEpoch); + memcpy(hdr->data.update.nodecfg.slots, node->slots, sizeof(node->slots)); + clusterSendMessage(link, buf, ntohl(hdr->totlen)); +} + +/* ----------------------------------------------------------------------------- + * CLUSTER Pub/Sub support + * + * For now we do very little, just propagating PUBLISH messages across the whole + * cluster. In the future we'll try to get smarter and avoiding propagating those + * messages to hosts without receives for a given channel. + * -------------------------------------------------------------------------- */ +void clusterPropagatePublish(robj *channel, robj *message) { + clusterSendPublish(NULL, channel, message); +} + +/* ----------------------------------------------------------------------------- + * SLAVE node specific functions + * -------------------------------------------------------------------------- */ + + /* This function sends a FAILOVE_AUTH_REQUEST message to every node in order to + * see if there is the quorum for this slave instance to failover its failing + * master. + * + * Note that we send the failover request to everybody, master and slave nodes, + * but only the masters are supposed to reply to our query. */ +void clusterRequestFailoverAuth(void) { + unsigned char buf[sizeof(clusterMsg)]; + clusterMsg *hdr = (clusterMsg*) buf; + uint32_t totlen; + + clusterBuildMessageHdr(hdr, CLUSTERMSG_TYPE_FAILOVER_AUTH_REQUEST); + /* If this is a manual failover, set the CLUSTERMSG_FLAG0_FORCEACK bit + * in the header to communicate the nodes receiving the message that + * they should authorized the failover even if the master is working. */ + if (server.cluster->mf_end) hdr->mflags[0] |= CLUSTERMSG_FLAG0_FORCEACK; + totlen = sizeof(clusterMsg) - sizeof(union clusterMsgData); + hdr->totlen = htonl(totlen); + clusterBroadcastMessage(buf, totlen); +} + +/* Send a FAILOVER_AUTH_ACK message to the specified node. */ +void clusterSendFailoverAuth(clusterNode *node) { + unsigned char buf[sizeof(clusterMsg)]; + clusterMsg *hdr = (clusterMsg*) buf; + uint32_t totlen; + + if (!node->link) return; + clusterBuildMessageHdr(hdr, CLUSTERMSG_TYPE_FAILOVER_AUTH_ACK); + totlen = sizeof(clusterMsg) - sizeof(union clusterMsgData); + hdr->totlen = htonl(totlen); + clusterSendMessage(node->link, buf, totlen); +} + +/* Send a MFSTART message to the specified node. */ +void clusterSendMFStart(clusterNode *node) { + unsigned char buf[sizeof(clusterMsg)]; + clusterMsg *hdr = (clusterMsg*) buf; + uint32_t totlen; + + if (!node->link) return; + clusterBuildMessageHdr(hdr, CLUSTERMSG_TYPE_MFSTART); + totlen = sizeof(clusterMsg) - sizeof(union clusterMsgData); + hdr->totlen = htonl(totlen); + clusterSendMessage(node->link, buf, totlen); +} + +/* Vote for the node asking for our vote if there are the conditions. */ +void clusterSendFailoverAuthIfNeeded(clusterNode *node, clusterMsg *request) { + clusterNode *master = node->slaveof; + uint64_t requestCurrentEpoch = ntohu64(request->currentEpoch); + uint64_t requestConfigEpoch = ntohu64(request->configEpoch); + unsigned char *claimed_slots = request->myslots; + int force_ack = request->mflags[0] & CLUSTERMSG_FLAG0_FORCEACK; + int j; + + /* IF we are not a master serving at least 1 slot, we don't have the + * right to vote, as the cluster size in Redis Cluster is the number + * of masters serving at least one slot, and quorum is the cluster + * size + 1 */ + if (nodeIsSlave(myself) || myself->numslots == 0) return; + + /* Request epoch must be >= our currentEpoch. + * Note that it is impossible for it to actually be greater since + * our currentEpoch was updated as a side effect of receiving this + * request, if the request epoch was greater. */ + if (requestCurrentEpoch < server.cluster->currentEpoch) { + serverLog(LL_WARNING, + "Failover auth denied to %.40s: reqEpoch (%llu) < curEpoch(%llu)", + node->name, + (PORT_ULONGLONG) requestCurrentEpoch, + (PORT_ULONGLONG) server.cluster->currentEpoch); + return; + } + + /* I already voted for this epoch? Return ASAP. */ + if (server.cluster->lastVoteEpoch == server.cluster->currentEpoch) { + serverLog(LL_WARNING, + "Failover auth denied to %.40s: already voted for epoch %llu", + node->name, + (PORT_ULONGLONG) server.cluster->currentEpoch); + return; + } + + /* Node must be a slave and its master down. + * The master can be non failing if the request is flagged + * with CLUSTERMSG_FLAG0_FORCEACK (manual failover). */ + if (nodeIsMaster(node) || master == NULL || + (!nodeFailed(master) && !force_ack)) + { + if (nodeIsMaster(node)) { + serverLog(LL_WARNING, + "Failover auth denied to %.40s: it is a master node", + node->name); + } + else if (master == NULL) { + serverLog(LL_WARNING, + "Failover auth denied to %.40s: I don't know its master", + node->name); + } + else if (!nodeFailed(master)) { + serverLog(LL_WARNING, + "Failover auth denied to %.40s: its master is up", + node->name); + } + return; + } + + /* We did not voted for a slave about this master for two + * times the node timeout. This is not strictly needed for correctness + * of the algorithm but makes the base case more linear. */ + if (mstime() - node->slaveof->voted_time < server.cluster_node_timeout * 2) + { + serverLog(LL_WARNING, + "Failover auth denied to %.40s: " + "can't vote about this master before %lld milliseconds", + node->name, + (PORT_LONGLONG) ((server.cluster_node_timeout * 2) - + (mstime() - node->slaveof->voted_time))); + return; + } + + /* The slave requesting the vote must have a configEpoch for the claimed + * slots that is >= the one of the masters currently serving the same + * slots in the current configuration. */ + for (j = 0; j < CLUSTER_SLOTS; j++) { + if (bitmapTestBit(claimed_slots, j) == 0) continue; + if (server.cluster->slots[j] == NULL || + server.cluster->slots[j]->configEpoch <= requestConfigEpoch) + { + continue; + } + /* If we reached this point we found a slot that in our current slots + * is served by a master with a greater configEpoch than the one claimed + * by the slave requesting our vote. Refuse to vote for this slave. */ + serverLog(LL_WARNING, + "Failover auth denied to %.40s: " + "slot %d epoch (%llu) > reqEpoch (%llu)", + node->name, j, + (PORT_ULONGLONG) server.cluster->slots[j]->configEpoch, + (PORT_ULONGLONG) requestConfigEpoch); + return; + } + + /* We can vote for this slave. */ + clusterSendFailoverAuth(node); + server.cluster->lastVoteEpoch = server.cluster->currentEpoch; + node->slaveof->voted_time = mstime(); + serverLog(LL_WARNING, "Failover auth granted to %.40s for epoch %llu", + node->name, (PORT_ULONGLONG) server.cluster->currentEpoch); +} + +/* This function returns the "rank" of this instance, a slave, in the context + * of its master-slaves ring. The rank of the slave is given by the number of + * other slaves for the same master that have a better replication offset + * compared to the local one (better means, greater, so they claim more data). + * + * A slave with rank 0 is the one with the greatest (most up to date) + * replication offset, and so forth. Note that because how the rank is computed + * multiple slaves may have the same rank, in case they have the same offset. + * + * The slave rank is used to add a delay to start an election in order to + * get voted and replace a failing master. Slaves with better replication + * offsets are more likely to win. */ +int clusterGetSlaveRank(void) { + PORT_LONGLONG myoffset; + int j, rank = 0; + clusterNode *master; + + serverAssert(nodeIsSlave(myself)); + master = myself->slaveof; + if (master == NULL) return 0; /* Never called by slaves without master. */ + + myoffset = replicationGetSlaveOffset(); + for (j = 0; j < master->numslaves; j++) + if (master->slaves[j] != myself && + master->slaves[j]->repl_offset > myoffset) rank++; + return rank; +} + +/* This function is called by clusterHandleSlaveFailover() in order to + * let the slave log why it is not able to failover. Sometimes there are + * not the conditions, but since the failover function is called again and + * again, we can't log the same things continuously. + * + * This function works by logging only if a given set of conditions are + * true: + * + * 1) The reason for which the failover can't be initiated changed. + * The reasons also include a NONE reason we reset the state to + * when the slave finds that its master is fine (no FAIL flag). + * 2) Also, the log is emitted again if the master is still down and + * the reason for not failing over is still the same, but more than + * CLUSTER_CANT_FAILOVER_RELOG_PERIOD seconds elapsed. + * 3) Finally, the function only logs if the slave is down for more than + * five seconds + NODE_TIMEOUT. This way nothing is logged when a + * failover starts in a reasonable time. + * + * The function is called with the reason why the slave can't failover + * which is one of the integer macros CLUSTER_CANT_FAILOVER_*. + * + * The function is guaranteed to be called only if 'myself' is a slave. */ +void clusterLogCantFailover(int reason) { + char *msg; + static time_t lastlog_time = 0; + mstime_t nolog_fail_time = server.cluster_node_timeout + 5000; + + /* Don't log if we have the same reason for some time. */ + if (reason == server.cluster->cant_failover_reason && + time(NULL) - lastlog_time < CLUSTER_CANT_FAILOVER_RELOG_PERIOD) + return; + + server.cluster->cant_failover_reason = reason; + + /* We also don't emit any log if the master failed no long ago, the + * goal of this function is to log slaves in a stalled condition for + * a long time. */ + if (myself->slaveof && + nodeFailed(myself->slaveof) && + (mstime() - myself->slaveof->fail_time) < nolog_fail_time) return; + + switch (reason) { + case CLUSTER_CANT_FAILOVER_DATA_AGE: + msg = "Disconnected from master for longer than allowed. " + "Please check the 'cluster-slave-validity-factor' configuration " + "option."; + break; + case CLUSTER_CANT_FAILOVER_WAITING_DELAY: + msg = "Waiting the delay before I can start a new failover."; + break; + case CLUSTER_CANT_FAILOVER_EXPIRED: + msg = "Failover attempt expired."; + break; + case CLUSTER_CANT_FAILOVER_WAITING_VOTES: + msg = "Waiting for votes, but majority still not reached."; + break; + default: + msg = "Unknown reason code."; + break; + } + lastlog_time = time(NULL); + serverLog(LL_WARNING, "Currently unable to failover: %s", msg); +} + +/* This function implements the final part of automatic and manual failovers, + * where the slave grabs its master's hash slots, and propagates the new + * configuration. + * + * Note that it's up to the caller to be sure that the node got a new + * configuration epoch already. */ +void clusterFailoverReplaceYourMaster(void) { + int j; + clusterNode *oldmaster = myself->slaveof; + + if (nodeIsMaster(myself) || oldmaster == NULL) return; + + /* 1) Turn this node into a master. */ + clusterSetNodeAsMaster(myself); + replicationUnsetMaster(); + + /* 2) Claim all the slots assigned to our master. */ + for (j = 0; j < CLUSTER_SLOTS; j++) { + if (clusterNodeGetSlotBit(oldmaster, j)) { + clusterDelSlot(j); + clusterAddSlot(myself, j); + } + } + + /* 3) Update state and save config. */ + clusterUpdateState(); + clusterSaveConfigOrDie(1); + + /* 4) Pong all the other nodes so that they can update the state + * accordingly and detect that we switched to master role. */ + clusterBroadcastPong(CLUSTER_BROADCAST_ALL); + + /* 5) If there was a manual failover in progress, clear the state. */ + resetManualFailover(); +} + +/* This function is called if we are a slave node and our master serving + * a non-zero amount of hash slots is in FAIL state. + * + * The gaol of this function is: + * 1) To check if we are able to perform a failover, is our data updated? + * 2) Try to get elected by masters. + * 3) Perform the failover informing all the other nodes. + */ +void clusterHandleSlaveFailover(void) { + mstime_t data_age; + mstime_t auth_age = mstime() - server.cluster->failover_auth_time; + int needed_quorum = (server.cluster->size / 2) + 1; + int manual_failover = server.cluster->mf_end != 0 && + server.cluster->mf_can_start; + mstime_t auth_timeout, auth_retry_time; + + server.cluster->todo_before_sleep &= ~CLUSTER_TODO_HANDLE_FAILOVER; + + /* Compute the failover timeout (the max time we have to send votes + * and wait for replies), and the failover retry time (the time to wait + * before trying to get voted again). + * + * Timeout is MAX(NODE_TIMEOUT*2,2000) milliseconds. + * Retry is two times the Timeout. + */ + auth_timeout = server.cluster_node_timeout * 2; + if (auth_timeout < 2000) auth_timeout = 2000; + auth_retry_time = auth_timeout * 2; + + /* Pre conditions to run the function, that must be met both in case + * of an automatic or manual failover: + * 1) We are a slave. + * 2) Our master is flagged as FAIL, or this is a manual failover. + * 3) It is serving slots. */ + if (nodeIsMaster(myself) || + myself->slaveof == NULL || + (!nodeFailed(myself->slaveof) && !manual_failover) || + myself->slaveof->numslots == 0) + { + /* There are no reasons to failover, so we set the reason why we + * are returning without failing over to NONE. */ + server.cluster->cant_failover_reason = CLUSTER_CANT_FAILOVER_NONE; + return; + } + + /* Set data_age to the number of seconds we are disconnected from + * the master. */ + if (server.repl_state == REPL_STATE_CONNECTED) { + data_age = (mstime_t) (server.unixtime - server.master->lastinteraction) + * 1000; + } + else { + data_age = (mstime_t) (server.unixtime - server.repl_down_since) * 1000; + } + + /* Remove the node timeout from the data age as it is fine that we are + * disconnected from our master at least for the time it was down to be + * flagged as FAIL, that's the baseline. */ + if (data_age > server.cluster_node_timeout) + data_age -= server.cluster_node_timeout; + + /* Check if our data is recent enough according to the slave validity + * factor configured by the user. + * + * Check bypassed for manual failovers. */ + if (server.cluster_slave_validity_factor && + data_age > + (((mstime_t) server.repl_ping_slave_period * 1000) + + (server.cluster_node_timeout * server.cluster_slave_validity_factor))) + { + if (!manual_failover) { + clusterLogCantFailover(CLUSTER_CANT_FAILOVER_DATA_AGE); + return; + } + } + + /* If the previous failover attempt timedout and the retry time has + * elapsed, we can setup a new one. */ + if (auth_age > auth_retry_time) { + server.cluster->failover_auth_time = mstime() + + 500 + /* Fixed delay of 500 milliseconds, let FAIL msg propagate. */ + random() % 500; /* Random delay between 0 and 500 milliseconds. */ + server.cluster->failover_auth_count = 0; + server.cluster->failover_auth_sent = 0; + server.cluster->failover_auth_rank = clusterGetSlaveRank(); + /* We add another delay that is proportional to the slave rank. + * Specifically 1 second * rank. This way slaves that have a probably + * less updated replication offset, are penalized. */ + server.cluster->failover_auth_time += + server.cluster->failover_auth_rank * 1000; + /* However if this is a manual failover, no delay is needed. */ + if (server.cluster->mf_end) { + server.cluster->failover_auth_time = mstime(); + server.cluster->failover_auth_rank = 0; + } + serverLog(LL_WARNING, + "Start of election delayed for %lld milliseconds " + "(rank #%d, offset %lld).", + server.cluster->failover_auth_time - mstime(), + server.cluster->failover_auth_rank, + replicationGetSlaveOffset()); + /* Now that we have a scheduled election, broadcast our offset + * to all the other slaves so that they'll updated their offsets + * if our offset is better. */ + clusterBroadcastPong(CLUSTER_BROADCAST_LOCAL_SLAVES); + return; + } + + /* It is possible that we received more updated offsets from other + * slaves for the same master since we computed our election delay. + * Update the delay if our rank changed. + * + * Not performed if this is a manual failover. */ + if (server.cluster->failover_auth_sent == 0 && + server.cluster->mf_end == 0) + { + int newrank = clusterGetSlaveRank(); + if (newrank > server.cluster->failover_auth_rank) { + PORT_LONGLONG added_delay = + (newrank - server.cluster->failover_auth_rank) * 1000; + server.cluster->failover_auth_time += added_delay; + server.cluster->failover_auth_rank = newrank; + serverLog(LL_WARNING, + "Slave rank updated to #%d, added %lld milliseconds of delay.", + newrank, added_delay); + } + } + + /* Return ASAP if we can't still start the election. */ + if (mstime() < server.cluster->failover_auth_time) { + clusterLogCantFailover(CLUSTER_CANT_FAILOVER_WAITING_DELAY); + return; + } + + /* Return ASAP if the election is too old to be valid. */ + if (auth_age > auth_timeout) { + clusterLogCantFailover(CLUSTER_CANT_FAILOVER_EXPIRED); + return; + } + + /* Ask for votes if needed. */ + if (server.cluster->failover_auth_sent == 0) { + server.cluster->currentEpoch++; + server.cluster->failover_auth_epoch = server.cluster->currentEpoch; + serverLog(LL_WARNING, "Starting a failover election for epoch %llu.", + (PORT_ULONGLONG) server.cluster->currentEpoch); + clusterRequestFailoverAuth(); + server.cluster->failover_auth_sent = 1; + clusterDoBeforeSleep(CLUSTER_TODO_SAVE_CONFIG | + CLUSTER_TODO_UPDATE_STATE | + CLUSTER_TODO_FSYNC_CONFIG); + return; /* Wait for replies. */ + } + + /* Check if we reached the quorum. */ + if (server.cluster->failover_auth_count >= needed_quorum) { + /* We have the quorum, we can finally failover the master. */ + + serverLog(LL_WARNING, + "Failover election won: I'm the new master."); + + /* Update my configEpoch to the epoch of the election. */ + if (myself->configEpoch < server.cluster->failover_auth_epoch) { + myself->configEpoch = server.cluster->failover_auth_epoch; + serverLog(LL_WARNING, + "configEpoch set to %llu after successful failover", + (PORT_ULONGLONG) myself->configEpoch); + } + + /* Take responsability for the cluster slots. */ + clusterFailoverReplaceYourMaster(); + } + else { + clusterLogCantFailover(CLUSTER_CANT_FAILOVER_WAITING_VOTES); + } +} + +/* ----------------------------------------------------------------------------- + * CLUSTER slave migration + * + * Slave migration is the process that allows a slave of a master that is + * already covered by at least another slave, to "migrate" to a master that + * is orpaned, that is, left with no working slaves. + * ------------------------------------------------------------------------- */ + + /* This function is responsible to decide if this replica should be migrated + * to a different (orphaned) master. It is called by the clusterCron() function + * only if: + * + * 1) We are a slave node. + * 2) It was detected that there is at least one orphaned master in + * the cluster. + * 3) We are a slave of one of the masters with the greatest number of + * slaves. + * + * This checks are performed by the caller since it requires to iterate + * the nodes anyway, so we spend time into clusterHandleSlaveMigration() + * if definitely needed. + * + * The fuction is called with a pre-computed max_slaves, that is the max + * number of working (not in FAIL state) slaves for a single master. + * + * Additional conditions for migration are examined inside the function. + */ +void clusterHandleSlaveMigration(int max_slaves) { + int j, okslaves = 0; + clusterNode *mymaster = myself->slaveof, *target = NULL, *candidate = NULL; + dictIterator *di; + dictEntry *de; + + /* Step 1: Don't migrate if the cluster state is not ok. */ + if (server.cluster->state != CLUSTER_OK) return; + + /* Step 2: Don't migrate if my master will not be left with at least + * 'migration-barrier' slaves after my migration. */ + if (mymaster == NULL) return; + for (j = 0; j < mymaster->numslaves; j++) + if (!nodeFailed(mymaster->slaves[j]) && + !nodeTimedOut(mymaster->slaves[j])) okslaves++; + if (okslaves <= server.cluster_migration_barrier) return; + + /* Step 3: Idenitfy a candidate for migration, and check if among the + * masters with the greatest number of ok slaves, I'm the one with the + * smallest node ID (the "candidate slave"). + * + * Note: this means that eventually a replica migration will occurr + * since slaves that are reachable again always have their FAIL flag + * cleared, so eventually there must be a candidate. At the same time + * this does not mean that there are no race conditions possible (two + * slaves migrating at the same time), but this is unlikely to + * happen, and harmless when happens. */ + candidate = myself; + di = dictGetSafeIterator(server.cluster->nodes); + while ((de = dictNext(di)) != NULL) { + clusterNode *node = dictGetVal(de); + int okslaves = 0, is_orphaned = 1; + + /* We want to migrate only if this master is working, orphaned, and + * used to have slaves or if failed over a master that had slaves + * (MIGRATE_TO flag). This way we only migrate to instances that were + * supposed to have replicas. */ + if (nodeIsSlave(node) || nodeFailed(node)) is_orphaned = 0; + if (!(node->flags & CLUSTER_NODE_MIGRATE_TO)) is_orphaned = 0; + + /* Check number of working slaves. */ + if (nodeIsMaster(node)) okslaves = clusterCountNonFailingSlaves(node); + if (okslaves > 0) is_orphaned = 0; + + if (is_orphaned) { + if (!target && node->numslots > 0) target = node; + + /* Track the starting time of the orphaned condition for this + * master. */ + if (!node->orphaned_time) node->orphaned_time = mstime(); + } + else { + node->orphaned_time = 0; + } + + /* Check if I'm the slave candidate for the migration: attached + * to a master with the maximum number of slaves and with the smallest + * node ID. */ + if (okslaves == max_slaves) { + for (j = 0; j < node->numslaves; j++) { + if (memcmp(node->slaves[j]->name, + candidate->name, + CLUSTER_NAMELEN) < 0) + { + candidate = node->slaves[j]; + } + } + } + } + dictReleaseIterator(di); + + /* Step 4: perform the migration if there is a target, and if I'm the + * candidate, but only if the master is continuously orphaned for a + * couple of seconds, so that during failovers, we give some time to + * the natural slaves of this instance to advertise their switch from + * the old master to the new one. */ + if (target && candidate == myself && + (mstime() - target->orphaned_time) > CLUSTER_SLAVE_MIGRATION_DELAY) + { + serverLog(LL_WARNING, "Migrating to orphaned master %.40s", + target->name); + clusterSetMaster(target); + } +} + +/* ----------------------------------------------------------------------------- + * CLUSTER manual failover + * + * This are the important steps performed by slaves during a manual failover: + * 1) User send CLUSTER FAILOVER command. The failover state is initialized + * setting mf_end to the millisecond unix time at which we'll abort the + * attempt. + * 2) Slave sends a MFSTART message to the master requesting to pause clients + * for two times the manual failover timeout CLUSTER_MF_TIMEOUT. + * When master is paused for manual failover, it also starts to flag + * packets with CLUSTERMSG_FLAG0_PAUSED. + * 3) Slave waits for master to send its replication offset flagged as PAUSED. + * 4) If slave received the offset from the master, and its offset matches, + * mf_can_start is set to 1, and clusterHandleSlaveFailover() will perform + * the failover as usually, with the difference that the vote request + * will be modified to force masters to vote for a slave that has a + * working master. + * + * From the point of view of the master things are simpler: when a + * PAUSE_CLIENTS packet is received the master sets mf_end as well and + * the sender in mf_slave. During the time limit for the manual failover + * the master will just send PINGs more often to this slave, flagged with + * the PAUSED flag, so that the slave will set mf_master_offset when receiving + * a packet from the master with this flag set. + * + * The gaol of the manual failover is to perform a fast failover without + * data loss due to the asynchronous master-slave replication. + * -------------------------------------------------------------------------- */ + + /* Reset the manual failover state. This works for both masters and slavesa + * as all the state about manual failover is cleared. + * + * The function can be used both to initialize the manual failover state at + * startup or to abort a manual failover in progress. */ +void resetManualFailover(void) { + if (server.cluster->mf_end && clientsArePaused()) { + server.clients_pause_end_time = 0; + clientsArePaused(); /* Just use the side effect of the function. */ + } + server.cluster->mf_end = 0; /* No manual failover in progress. */ + server.cluster->mf_can_start = 0; + server.cluster->mf_slave = NULL; + server.cluster->mf_master_offset = 0; +} + +/* If a manual failover timed out, abort it. */ +void manualFailoverCheckTimeout(void) { + if (server.cluster->mf_end && server.cluster->mf_end < mstime()) { + serverLog(LL_WARNING, "Manual failover timed out."); + resetManualFailover(); + } +} + +/* This function is called from the cluster cron function in order to go + * forward with a manual failover state machine. */ +void clusterHandleManualFailover(void) { + /* Return ASAP if no manual failover is in progress. */ + if (server.cluster->mf_end == 0) return; + + /* If mf_can_start is non-zero, the failover was already triggered so the + * next steps are performed by clusterHandleSlaveFailover(). */ + if (server.cluster->mf_can_start) return; + + if (server.cluster->mf_master_offset == 0) return; /* Wait for offset... */ + + if (server.cluster->mf_master_offset == replicationGetSlaveOffset()) { + /* Our replication offset matches the master replication offset + * announced after clients were paused. We can start the failover. */ + server.cluster->mf_can_start = 1; + serverLog(LL_WARNING, + "All master replication stream processed, " + "manual failover can start."); + } +} + +/* ----------------------------------------------------------------------------- + * CLUSTER cron job + * -------------------------------------------------------------------------- */ + + /* This is executed 10 times every second */ +void clusterCron(void) { + dictIterator *di; + dictEntry *de; + int update_state = 0; + int orphaned_masters; /* How many masters there are without ok slaves. */ + int max_slaves; /* Max number of ok slaves for a single master. */ + int this_slaves; /* Number of ok slaves for our master (if we are slave). */ + mstime_t min_pong = 0, now = mstime(); + clusterNode *min_pong_node = NULL; + static PORT_ULONGLONG iteration = 0; + mstime_t handshake_timeout; + + iteration++; /* Number of times this function was called so far. */ + + /* We want to take myself->ip in sync with the cluster-announce-ip option. + * The option can be set at runtime via CONFIG SET, so we periodically check + * if the option changed to reflect this into myself->ip. */ + { + static char *prev_ip = NULL; + char *curr_ip = server.cluster_announce_ip; + int changed = 0; + + if (prev_ip == NULL && curr_ip != NULL) changed = 1; + if (prev_ip != NULL && curr_ip == NULL) changed = 1; + if (prev_ip && curr_ip && strcmp(prev_ip, curr_ip)) changed = 1; + + if (changed) { + prev_ip = curr_ip; + if (prev_ip) prev_ip = zstrdup(prev_ip); + + if (curr_ip) { + strncpy(myself->ip, server.cluster_announce_ip, NET_IP_STR_LEN); + myself->ip[NET_IP_STR_LEN - 1] = '\0'; + } + else { + myself->ip[0] = '\0'; /* Force autodetection. */ + } + } + } + + /* The handshake timeout is the time after which a handshake node that was + * not turned into a normal node is removed from the nodes. Usually it is + * just the NODE_TIMEOUT value, but when NODE_TIMEOUT is too small we use + * the value of 1 second. */ + handshake_timeout = server.cluster_node_timeout; + if (handshake_timeout < 1000) handshake_timeout = 1000; + + /* Check if we have disconnected nodes and re-establish the connection. + * Also update a few stats while we are here, that can be used to make + * better decisions in other part of the code. */ + di = dictGetSafeIterator(server.cluster->nodes); + server.cluster->stats_pfail_nodes = 0; + while ((de = dictNext(di)) != NULL) { + clusterNode *node = dictGetVal(de); + + /* Not interested in reconnecting the link with myself or nodes + * for which we have no address. */ + if (node->flags & (CLUSTER_NODE_MYSELF | CLUSTER_NODE_NOADDR)) continue; + + if (node->flags & CLUSTER_NODE_PFAIL) + server.cluster->stats_pfail_nodes++; + + /* A Node in HANDSHAKE state has a limited lifespan equal to the + * configured node timeout. */ + if (nodeInHandshake(node) && now - node->ctime > handshake_timeout) { + clusterDelNode(node); + continue; + } + + if (node->link == NULL) { + int fd; + mstime_t old_ping_sent; + clusterLink *link; + + fd = anetTcpNonBlockBindConnect(server.neterr, node->ip, + node->cport, NET_FIRST_BIND_ADDR); + if (fd == -1) { + /* We got a synchronous error from connect before + * clusterSendPing() had a chance to be called. + * If node->ping_sent is zero, failure detection can't work, + * so we claim we actually sent a ping now (that will + * be really sent as soon as the link is obtained). */ + if (node->ping_sent == 0) node->ping_sent = mstime(); + serverLog(LL_DEBUG, "Unable to connect to " + "Cluster Node [%s]:%d -> %s", node->ip, + node->cport, server.neterr); + continue; + } + link = createClusterLink(node); + link->fd = fd; + node->link = link; + aeCreateFileEvent(server.el, link->fd, AE_READABLE, + clusterReadHandler, link); + /* Queue a PING in the new connection ASAP: this is crucial + * to avoid false positives in failure detection. + * + * If the node is flagged as MEET, we send a MEET message instead + * of a PING one, to force the receiver to add us in its node + * table. */ + old_ping_sent = node->ping_sent; + clusterSendPing(link, node->flags & CLUSTER_NODE_MEET ? + CLUSTERMSG_TYPE_MEET : CLUSTERMSG_TYPE_PING); + if (old_ping_sent) { + /* If there was an active ping before the link was + * disconnected, we want to restore the ping time, otherwise + * replaced by the clusterSendPing() call. */ + node->ping_sent = old_ping_sent; + } + /* We can clear the flag after the first packet is sent. + * If we'll never receive a PONG, we'll never send new packets + * to this node. Instead after the PONG is received and we + * are no longer in meet/handshake status, we want to send + * normal PING packets. */ + node->flags &= ~CLUSTER_NODE_MEET; + + serverLog(LL_DEBUG, "Connecting with Node %.40s at %s:%d", + node->name, node->ip, node->cport); + } + } + dictReleaseIterator(di); + + /* Ping some random node 1 time every 10 iterations, so that we usually ping + * one random node every second. */ + if (!(iteration % 10)) { + int j; + + /* Check a few random nodes and ping the one with the oldest + * pong_received time. */ + for (j = 0; j < 5; j++) { + de = dictGetRandomKey(server.cluster->nodes); + clusterNode *this = dictGetVal(de); + + /* Don't ping nodes disconnected or with a ping currently active. */ + if (this->link == NULL || this->ping_sent != 0) continue; + if (this->flags & (CLUSTER_NODE_MYSELF | CLUSTER_NODE_HANDSHAKE)) + continue; + if (min_pong_node == NULL || min_pong > this->pong_received) { + min_pong_node = this; + min_pong = this->pong_received; + } + } + if (min_pong_node) { + serverLog(LL_DEBUG, "Pinging node %.40s", min_pong_node->name); + clusterSendPing(min_pong_node->link, CLUSTERMSG_TYPE_PING); + } + } + + /* Iterate nodes to check if we need to flag something as failing. + * This loop is also responsible to: + * 1) Check if there are orphaned masters (masters without non failing + * slaves). + * 2) Count the max number of non failing slaves for a single master. + * 3) Count the number of slaves for our master, if we are a slave. */ + orphaned_masters = 0; + max_slaves = 0; + this_slaves = 0; + di = dictGetSafeIterator(server.cluster->nodes); + while ((de = dictNext(di)) != NULL) { + clusterNode *node = dictGetVal(de); + now = mstime(); /* Use an updated time at every iteration. */ + mstime_t delay; + + if (node->flags & + (CLUSTER_NODE_MYSELF | CLUSTER_NODE_NOADDR | CLUSTER_NODE_HANDSHAKE)) + continue; + + /* Orphaned master check, useful only if the current instance + * is a slave that may migrate to another master. */ + if (nodeIsSlave(myself) && nodeIsMaster(node) && !nodeFailed(node)) { + int okslaves = clusterCountNonFailingSlaves(node); + + /* A master is orphaned if it is serving a non-zero number of + * slots, have no working slaves, but used to have at least one + * slave, or failed over a master that used to have slaves. */ + if (okslaves == 0 && node->numslots > 0 && + node->flags & CLUSTER_NODE_MIGRATE_TO) + { + orphaned_masters++; + } + if (okslaves > max_slaves) max_slaves = okslaves; + if (nodeIsSlave(myself) && myself->slaveof == node) + this_slaves = okslaves; + } + + /* If we are waiting for the PONG more than half the cluster + * timeout, reconnect the link: maybe there is a connection + * issue even if the node is alive. */ + if (node->link && /* is connected */ + now - node->link->ctime > + server.cluster_node_timeout && /* was not already reconnected */ + node->ping_sent && /* we already sent a ping */ + node->pong_received < node->ping_sent && /* still waiting pong */ + /* and we are waiting for the pong more than timeout/2 */ + now - node->ping_sent > server.cluster_node_timeout / 2) + { + /* Disconnect the link, it will be reconnected automatically. */ + freeClusterLink(node->link); + } + + /* If we have currently no active ping in this instance, and the + * received PONG is older than half the cluster timeout, send + * a new ping now, to ensure all the nodes are pinged without + * a too big delay. */ + if (node->link && + node->ping_sent == 0 && + (now - node->pong_received) > server.cluster_node_timeout / 2) + { + clusterSendPing(node->link, CLUSTERMSG_TYPE_PING); + continue; + } + + /* If we are a master and one of the slaves requested a manual + * failover, ping it continuously. */ + if (server.cluster->mf_end && + nodeIsMaster(myself) && + server.cluster->mf_slave == node && + node->link) + { + clusterSendPing(node->link, CLUSTERMSG_TYPE_PING); + continue; + } + + /* Check only if we have an active ping for this instance. */ + if (node->ping_sent == 0) continue; + + /* Compute the delay of the PONG. Note that if we already received + * the PONG, then node->ping_sent is zero, so can't reach this + * code at all. */ + delay = now - node->ping_sent; + + if (delay > server.cluster_node_timeout) { + /* Timeout reached. Set the node as possibly failing if it is + * not already in this state. */ + if (!(node->flags & (CLUSTER_NODE_PFAIL | CLUSTER_NODE_FAIL))) { + serverLog(LL_DEBUG, "*** NODE %.40s possibly failing", + node->name); + node->flags |= CLUSTER_NODE_PFAIL; + update_state = 1; + } + } + } + dictReleaseIterator(di); + + /* If we are a slave node but the replication is still turned off, + * enable it if we know the address of our master and it appears to + * be up. */ + if (nodeIsSlave(myself) && + server.masterhost == NULL && + myself->slaveof && + nodeHasAddr(myself->slaveof)) + { + replicationSetMaster(myself->slaveof->ip, myself->slaveof->port); + } + + /* Abourt a manual failover if the timeout is reached. */ + manualFailoverCheckTimeout(); + + if (nodeIsSlave(myself)) { + clusterHandleManualFailover(); + clusterHandleSlaveFailover(); + /* If there are orphaned slaves, and we are a slave among the masters + * with the max number of non-failing slaves, consider migrating to + * the orphaned masters. Note that it does not make sense to try + * a migration if there is no master with at least *two* working + * slaves. */ + if (orphaned_masters && max_slaves >= 2 && this_slaves == max_slaves) + clusterHandleSlaveMigration(max_slaves); + } + + if (update_state || server.cluster->state == CLUSTER_FAIL) + clusterUpdateState(); +} + +/* This function is called before the event handler returns to sleep for + * events. It is useful to perform operations that must be done ASAP in + * reaction to events fired but that are not safe to perform inside event + * handlers, or to perform potentially expansive tasks that we need to do + * a single time before replying to clients. */ +void clusterBeforeSleep(void) { + /* Handle failover, this is needed when it is likely that there is already + * the quorum from masters in order to react fast. */ + if (server.cluster->todo_before_sleep & CLUSTER_TODO_HANDLE_FAILOVER) + clusterHandleSlaveFailover(); + + /* Update the cluster state. */ + if (server.cluster->todo_before_sleep & CLUSTER_TODO_UPDATE_STATE) + clusterUpdateState(); + + /* Save the config, possibly using fsync. */ + if (server.cluster->todo_before_sleep & CLUSTER_TODO_SAVE_CONFIG) { + int fsync = server.cluster->todo_before_sleep & + CLUSTER_TODO_FSYNC_CONFIG; + clusterSaveConfigOrDie(fsync); + } + + /* Reset our flags (not strictly needed since every single function + * called for flags set should be able to clear its flag). */ + server.cluster->todo_before_sleep = 0; +} + +void clusterDoBeforeSleep(int flags) { + server.cluster->todo_before_sleep |= flags; +} + +/* ----------------------------------------------------------------------------- + * Slots management + * -------------------------------------------------------------------------- */ + + /* Test bit 'pos' in a generic bitmap. Return 1 if the bit is set, + * otherwise 0. */ +int bitmapTestBit(unsigned char *bitmap, int pos) { + off_t byte = pos / 8; + int bit = pos & 7; + return (bitmap[byte] & (1 << bit)) != 0; +} + +/* Set the bit at position 'pos' in a bitmap. */ +void bitmapSetBit(unsigned char *bitmap, int pos) { + off_t byte = pos / 8; + int bit = pos & 7; + bitmap[byte] |= 1 << bit; +} + +/* Clear the bit at position 'pos' in a bitmap. */ +void bitmapClearBit(unsigned char *bitmap, int pos) { + off_t byte = pos / 8; + int bit = pos & 7; + bitmap[byte] &= ~(1 << bit); +} + +/* Return non-zero if there is at least one master with slaves in the cluster. + * Otherwise zero is returned. Used by clusterNodeSetSlotBit() to set the + * MIGRATE_TO flag the when a master gets the first slot. */ +int clusterMastersHaveSlaves(void) { + dictIterator *di = dictGetSafeIterator(server.cluster->nodes); + dictEntry *de; + int slaves = 0; + while ((de = dictNext(di)) != NULL) { + clusterNode *node = dictGetVal(de); + + if (nodeIsSlave(node)) continue; + slaves += node->numslaves; + } + dictReleaseIterator(di); + return slaves != 0; +} + +/* Set the slot bit and return the old value. */ +int clusterNodeSetSlotBit(clusterNode *n, int slot) { + int old = bitmapTestBit(n->slots, slot); + bitmapSetBit(n->slots, slot); + if (!old) { + n->numslots++; + /* When a master gets its first slot, even if it has no slaves, + * it gets flagged with MIGRATE_TO, that is, the master is a valid + * target for replicas migration, if and only if at least one of + * the other masters has slaves right now. + * + * Normally masters are valid targerts of replica migration if: + * 1. The used to have slaves (but no longer have). + * 2. They are slaves failing over a master that used to have slaves. + * + * However new masters with slots assigned are considered valid + * migration tagets if the rest of the cluster is not a slave-less. + * + * See https://github.com/antirez/redis/issues/3043 for more info. */ + if (n->numslots == 1 && clusterMastersHaveSlaves()) + n->flags |= CLUSTER_NODE_MIGRATE_TO; + } + return old; +} + +/* Clear the slot bit and return the old value. */ +int clusterNodeClearSlotBit(clusterNode *n, int slot) { + int old = bitmapTestBit(n->slots, slot); + bitmapClearBit(n->slots, slot); + if (old) n->numslots--; + return old; +} + +/* Return the slot bit from the cluster node structure. */ +int clusterNodeGetSlotBit(clusterNode *n, int slot) { + return bitmapTestBit(n->slots, slot); +} + +/* Add the specified slot to the list of slots that node 'n' will + * serve. Return C_OK if the operation ended with success. + * If the slot is already assigned to another instance this is considered + * an error and C_ERR is returned. */ +int clusterAddSlot(clusterNode *n, int slot) { + if (server.cluster->slots[slot]) return C_ERR; + clusterNodeSetSlotBit(n, slot); + server.cluster->slots[slot] = n; + return C_OK; +} + +/* Delete the specified slot marking it as unassigned. + * Returns C_OK if the slot was assigned, otherwise if the slot was + * already unassigned C_ERR is returned. */ +int clusterDelSlot(int slot) { + clusterNode *n = server.cluster->slots[slot]; + + if (!n) return C_ERR; + serverAssert(clusterNodeClearSlotBit(n, slot) == 1); + server.cluster->slots[slot] = NULL; + return C_OK; +} + +/* Delete all the slots associated with the specified node. + * The number of deleted slots is returned. */ +int clusterDelNodeSlots(clusterNode *node) { + int deleted = 0, j; + + for (j = 0; j < CLUSTER_SLOTS; j++) { + if (clusterNodeGetSlotBit(node, j)) { + clusterDelSlot(j); + deleted++; + } + } + return deleted; +} + +/* Clear the migrating / importing state for all the slots. + * This is useful at initialization and when turning a master into slave. */ +void clusterCloseAllSlots(void) { + memset(server.cluster->migrating_slots_to, 0, + sizeof(server.cluster->migrating_slots_to)); + memset(server.cluster->importing_slots_from, 0, + sizeof(server.cluster->importing_slots_from)); +} + +/* ----------------------------------------------------------------------------- + * Cluster state evaluation function + * -------------------------------------------------------------------------- */ + + /* The following are defines that are only used in the evaluation function + * and are based on heuristics. Actaully the main point about the rejoin and + * writable delay is that they should be a few orders of magnitude larger + * than the network latency. */ +#define CLUSTER_MAX_REJOIN_DELAY 5000 +#define CLUSTER_MIN_REJOIN_DELAY 500 +#define CLUSTER_WRITABLE_DELAY 2000 + +void clusterUpdateState(void) { + int j, new_state; + int reachable_masters = 0; + static mstime_t among_minority_time; + static mstime_t first_call_time = 0; + + server.cluster->todo_before_sleep &= ~CLUSTER_TODO_UPDATE_STATE; + + /* If this is a master node, wait some time before turning the state + * into OK, since it is not a good idea to rejoin the cluster as a writable + * master, after a reboot, without giving the cluster a chance to + * reconfigure this node. Note that the delay is calculated starting from + * the first call to this function and not since the server start, in order + * to don't count the DB loading time. */ + if (first_call_time == 0) first_call_time = mstime(); + if (nodeIsMaster(myself) && + server.cluster->state == CLUSTER_FAIL && + mstime() - first_call_time < CLUSTER_WRITABLE_DELAY) return; + + /* Start assuming the state is OK. We'll turn it into FAIL if there + * are the right conditions. */ + new_state = CLUSTER_OK; + + /* Check if all the slots are covered. */ + if (server.cluster_require_full_coverage) { + for (j = 0; j < CLUSTER_SLOTS; j++) { + if (server.cluster->slots[j] == NULL || + server.cluster->slots[j]->flags & (CLUSTER_NODE_FAIL)) + { + new_state = CLUSTER_FAIL; + break; + } + } + } + + /* Compute the cluster size, that is the number of master nodes + * serving at least a single slot. + * + * At the same time count the number of reachable masters having + * at least one slot. */ + { + dictIterator *di; + dictEntry *de; + + server.cluster->size = 0; + di = dictGetSafeIterator(server.cluster->nodes); + while ((de = dictNext(di)) != NULL) { + clusterNode *node = dictGetVal(de); + + if (nodeIsMaster(node) && node->numslots) { + server.cluster->size++; + if ((node->flags & (CLUSTER_NODE_FAIL | CLUSTER_NODE_PFAIL)) == 0) + reachable_masters++; + } + } + dictReleaseIterator(di); + } + + /* If we are in a minority partition, change the cluster state + * to FAIL. */ + { + int needed_quorum = (server.cluster->size / 2) + 1; + + if (reachable_masters < needed_quorum) { + new_state = CLUSTER_FAIL; + among_minority_time = mstime(); + } + } + + /* Log a state change */ + if (new_state != server.cluster->state) { + mstime_t rejoin_delay = server.cluster_node_timeout; + + /* If the instance is a master and was partitioned away with the + * minority, don't let it accept queries for some time after the + * partition heals, to make sure there is enough time to receive + * a configuration update. */ + if (rejoin_delay > CLUSTER_MAX_REJOIN_DELAY) + rejoin_delay = CLUSTER_MAX_REJOIN_DELAY; + if (rejoin_delay < CLUSTER_MIN_REJOIN_DELAY) + rejoin_delay = CLUSTER_MIN_REJOIN_DELAY; + + if (new_state == CLUSTER_OK && + nodeIsMaster(myself) && + mstime() - among_minority_time < rejoin_delay) + { + return; + } + + /* Change the state and log the event. */ + serverLog(LL_WARNING, "Cluster state changed: %s", + new_state == CLUSTER_OK ? "ok" : "fail"); + server.cluster->state = new_state; + } +} + +/* This function is called after the node startup in order to verify that data + * loaded from disk is in agreement with the cluster configuration: + * + * 1) If we find keys about hash slots we have no responsibility for, the + * following happens: + * A) If no other node is in charge according to the current cluster + * configuration, we add these slots to our node. + * B) If according to our config other nodes are already in charge for + * this lots, we set the slots as IMPORTING from our point of view + * in order to justify we have those slots, and in order to make + * redis-trib aware of the issue, so that it can try to fix it. + * 2) If we find data in a DB different than DB0 we return C_ERR to + * signal the caller it should quit the server with an error message + * or take other actions. + * + * The function always returns C_OK even if it will try to correct + * the error described in "1". However if data is found in DB different + * from DB0, C_ERR is returned. + * + * The function also uses the logging facility in order to warn the user + * about desynchronizations between the data we have in memory and the + * cluster configuration. */ +int verifyClusterConfigWithData(void) { + int j; + int update_config = 0; + + /* If this node is a slave, don't perform the check at all as we + * completely depend on the replication stream. */ + if (nodeIsSlave(myself)) return C_OK; + + /* Make sure we only have keys in DB0. */ + for (j = 1; j < server.dbnum; j++) { + if (dictSize(server.db[j].dict)) return C_ERR; + } + + /* Check that all the slots we see populated memory have a corresponding + * entry in the cluster table. Otherwise fix the table. */ + for (j = 0; j < CLUSTER_SLOTS; j++) { + if (!countKeysInSlot(j)) continue; /* No keys in this slot. */ + /* Check if we are assigned to this slot or if we are importing it. + * In both cases check the next slot as the configuration makes + * sense. */ + if (server.cluster->slots[j] == myself || + server.cluster->importing_slots_from[j] != NULL) continue; + + /* If we are here data and cluster config don't agree, and we have + * slot 'j' populated even if we are not importing it, nor we are + * assigned to this slot. Fix this condition. */ + + update_config++; + /* Case A: slot is unassigned. Take responsibility for it. */ + if (server.cluster->slots[j] == NULL) { + serverLog(LL_WARNING, "I have keys for unassigned slot %d. " + "Taking responsibility for it.", j); + clusterAddSlot(myself, j); + } + else { + serverLog(LL_WARNING, "I have keys for slot %d, but the slot is " + "assigned to another node. " + "Setting it to importing state.", j); + server.cluster->importing_slots_from[j] = server.cluster->slots[j]; + } + } + if (update_config) clusterSaveConfigOrDie(1); + return C_OK; +} + +/* ----------------------------------------------------------------------------- + * SLAVE nodes handling + * -------------------------------------------------------------------------- */ + + /* Set the specified node 'n' as master for this node. + * If this node is currently a master, it is turned into a slave. */ +void clusterSetMaster(clusterNode *n) { + serverAssert(n != myself); + serverAssert(myself->numslots == 0); + + if (nodeIsMaster(myself)) { + myself->flags &= ~(CLUSTER_NODE_MASTER | CLUSTER_NODE_MIGRATE_TO); + myself->flags |= CLUSTER_NODE_SLAVE; + clusterCloseAllSlots(); + } + else { + if (myself->slaveof) + clusterNodeRemoveSlave(myself->slaveof, myself); + } + myself->slaveof = n; + clusterNodeAddSlave(n, myself); + replicationSetMaster(n->ip, n->port); + resetManualFailover(); +} + +/* ----------------------------------------------------------------------------- + * Nodes to string representation functions. + * -------------------------------------------------------------------------- */ + +struct redisNodeFlags { + uint16_t flag; + char *name; +}; + +static struct redisNodeFlags redisNodeFlagsTable[] = { + {CLUSTER_NODE_MYSELF, "myself,"}, + {CLUSTER_NODE_MASTER, "master,"}, + {CLUSTER_NODE_SLAVE, "slave,"}, + {CLUSTER_NODE_PFAIL, "fail?,"}, + {CLUSTER_NODE_FAIL, "fail,"}, + {CLUSTER_NODE_HANDSHAKE, "handshake,"}, + {CLUSTER_NODE_NOADDR, "noaddr,"} +}; + +/* Concatenate the comma separated list of node flags to the given SDS + * string 'ci'. */ +sds representClusterNodeFlags(sds ci, uint16_t flags) { + size_t orig_len = sdslen(ci); + int i, size = sizeof(redisNodeFlagsTable) / sizeof(struct redisNodeFlags); + for (i = 0; i < size; i++) { + struct redisNodeFlags *nodeflag = redisNodeFlagsTable + i; + if (flags & nodeflag->flag) ci = sdscat(ci, nodeflag->name); + } + /* If no flag was added, add the "noflags" special flag. */ + if (sdslen(ci) == orig_len) ci = sdscat(ci, "noflags,"); + sdsIncrLen(ci, -1); /* Remove trailing comma. */ + return ci; +} + +/* Generate a csv-alike representation of the specified cluster node. + * See clusterGenNodesDescription() top comment for more information. + * + * The function returns the string representation as an SDS string. */ +sds clusterGenNodeDescription(clusterNode *node) { + int j, start; + sds ci; + + /* Node coordinates */ + ci = sdscatprintf(sdsempty(), "%.40s %s:%d@%d ", + node->name, + node->ip, + node->port, + node->cport); + + /* Flags */ + ci = representClusterNodeFlags(ci, node->flags); + + /* Slave of... or just "-" */ + if (node->slaveof) + ci = sdscatprintf(ci, " %.40s ", node->slaveof->name); + else + ci = sdscatlen(ci, " - ", 3); + + /* Latency from the POV of this node, config epoch, link status */ + ci = sdscatprintf(ci, "%lld %lld %llu %s", + (PORT_LONGLONG) node->ping_sent, + (PORT_LONGLONG) node->pong_received, + (PORT_ULONGLONG) node->configEpoch, + (node->link || node->flags & CLUSTER_NODE_MYSELF) ? + "connected" : "disconnected"); + + /* Slots served by this instance */ + start = -1; + for (j = 0; j < CLUSTER_SLOTS; j++) { + int bit; + + if ((bit = clusterNodeGetSlotBit(node, j)) != 0) { + if (start == -1) start = j; + } + if (start != -1 && (!bit || j == CLUSTER_SLOTS - 1)) { + if (bit && j == CLUSTER_SLOTS - 1) j++; + + if (start == j - 1) { + ci = sdscatprintf(ci, " %d", start); + } + else { + ci = sdscatprintf(ci, " %d-%d", start, j - 1); + } + start = -1; + } + } + + /* Just for MYSELF node we also dump info about slots that + * we are migrating to other instances or importing from other + * instances. */ + if (node->flags & CLUSTER_NODE_MYSELF) { + for (j = 0; j < CLUSTER_SLOTS; j++) { + if (server.cluster->migrating_slots_to[j]) { + ci = sdscatprintf(ci, " [%d->-%.40s]", j, + server.cluster->migrating_slots_to[j]->name); + } + else if (server.cluster->importing_slots_from[j]) { + ci = sdscatprintf(ci, " [%d-<-%.40s]", j, + server.cluster->importing_slots_from[j]->name); + } + } + } + return ci; +} + +/* Generate a csv-alike representation of the nodes we are aware of, + * including the "myself" node, and return an SDS string containing the + * representation (it is up to the caller to free it). + * + * All the nodes matching at least one of the node flags specified in + * "filter" are excluded from the output, so using zero as a filter will + * include all the known nodes in the representation, including nodes in + * the HANDSHAKE state. + * + * The representation obtained using this function is used for the output + * of the CLUSTER NODES function, and as format for the cluster + * configuration file (nodes.conf) for a given node. */ +sds clusterGenNodesDescription(int filter) { + sds ci = sdsempty(), ni; + dictIterator *di; + dictEntry *de; + + di = dictGetSafeIterator(server.cluster->nodes); + while ((de = dictNext(di)) != NULL) { + clusterNode *node = dictGetVal(de); + + if (node->flags & filter) continue; + ni = clusterGenNodeDescription(node); + ci = sdscatsds(ci, ni); + sdsfree(ni); + ci = sdscatlen(ci, "\n", 1); + } + dictReleaseIterator(di); + return ci; +} + +/* ----------------------------------------------------------------------------- + * CLUSTER command + * -------------------------------------------------------------------------- */ + +const char *clusterGetMessageTypeString(int type) { + switch (type) { + case CLUSTERMSG_TYPE_PING: return "ping"; + case CLUSTERMSG_TYPE_PONG: return "pong"; + case CLUSTERMSG_TYPE_MEET: return "meet"; + case CLUSTERMSG_TYPE_FAIL: return "fail"; + case CLUSTERMSG_TYPE_PUBLISH: return "publish"; + case CLUSTERMSG_TYPE_FAILOVER_AUTH_REQUEST: return "auth-req"; + case CLUSTERMSG_TYPE_FAILOVER_AUTH_ACK: return "auth-ack"; + case CLUSTERMSG_TYPE_UPDATE: return "update"; + case CLUSTERMSG_TYPE_MFSTART: return "mfstart"; + } + return "unknown"; +} + +int getSlotOrReply(client *c, robj *o) { + PORT_LONGLONG slot; + + if (getLongLongFromObject(o, &slot) != C_OK || + slot < 0 || slot >= CLUSTER_SLOTS) + { + addReplyError(c, "Invalid or out of range slot"); + return -1; + } + return (int) slot; +} + +void clusterReplyMultiBulkSlots(client *c) { + /* Format: 1) 1) start slot + * 2) end slot + * 3) 1) master IP + * 2) master port + * 3) node ID + * 4) 1) replica IP + * 2) replica port + * 3) node ID + * ... continued until done + */ + + int num_masters = 0; + void *slot_replylen = addDeferredMultiBulkLength(c); + + dictEntry *de; + dictIterator *di = dictGetSafeIterator(server.cluster->nodes); + while ((de = dictNext(di)) != NULL) { + clusterNode *node = dictGetVal(de); + int j = 0, start = -1; + + /* Skip slaves (that are iterated when producing the output of their + * master) and masters not serving any slot. */ + if (!nodeIsMaster(node) || node->numslots == 0) continue; + + for (j = 0; j < CLUSTER_SLOTS; j++) { + int bit, i; + + if ((bit = clusterNodeGetSlotBit(node, j)) != 0) { + if (start == -1) start = j; + } + if (start != -1 && (!bit || j == CLUSTER_SLOTS - 1)) { + int nested_elements = 3; /* slots (2) + master addr (1). */ + void *nested_replylen = addDeferredMultiBulkLength(c); + + if (bit && j == CLUSTER_SLOTS - 1) j++; + + /* If slot exists in output map, add to it's list. + * else, create a new output map for this slot */ + if (start == j - 1) { + addReplyLongLong(c, start); /* only one slot; low==high */ + addReplyLongLong(c, start); + } + else { + addReplyLongLong(c, start); /* low */ + addReplyLongLong(c, j - 1); /* high */ + } + start = -1; + + /* First node reply position is always the master */ + addReplyMultiBulkLen(c, 3); + addReplyBulkCString(c, node->ip); + addReplyLongLong(c, node->port); + addReplyBulkCBuffer(c, node->name, CLUSTER_NAMELEN); + + /* Remaining nodes in reply are replicas for slot range */ + for (i = 0; i < node->numslaves; i++) { + /* This loop is copy/pasted from clusterGenNodeDescription() + * with modifications for per-slot node aggregation */ + if (nodeFailed(node->slaves[i])) continue; + addReplyMultiBulkLen(c, 3); + addReplyBulkCString(c, node->slaves[i]->ip); + addReplyLongLong(c, node->slaves[i]->port); + addReplyBulkCBuffer(c, node->slaves[i]->name, CLUSTER_NAMELEN); + nested_elements++; + } + setDeferredMultiBulkLength(c, nested_replylen, nested_elements); + num_masters++; + } + } + } + dictReleaseIterator(di); + setDeferredMultiBulkLength(c, slot_replylen, num_masters); +} + +void clusterCommand(client *c) { + if (server.cluster_enabled == 0) { + addReplyError(c, "This instance has cluster support disabled"); + return; + } + + if (!strcasecmp(c->argv[1]->ptr, "meet") && (c->argc == 4 || c->argc == 5)) { + /* CLUSTER MEET [cport] */ + PORT_LONGLONG port, cport; + + if (getLongLongFromObject(c->argv[3], &port) != C_OK) { + addReplyErrorFormat(c, "Invalid TCP base port specified: %s", + (char*) c->argv[3]->ptr); + return; + } + + if (c->argc == 5) { + if (getLongLongFromObject(c->argv[4], &cport) != C_OK) { + addReplyErrorFormat(c, "Invalid TCP bus port specified: %s", + (char*) c->argv[4]->ptr); + return; + } + } + else { + cport = port + CLUSTER_PORT_INCR; + } + + if (clusterStartHandshake(c->argv[2]->ptr, (int) port, (int) cport) == 0 && WIN_PORT_FIX /* cast (int) */ + errno == EINVAL) + { + addReplyErrorFormat(c, "Invalid node address specified: %s:%s", + (char*) c->argv[2]->ptr, (char*) c->argv[3]->ptr); + } + else { + addReply(c, shared.ok); + } + } + else if (!strcasecmp(c->argv[1]->ptr, "nodes") && c->argc == 2) { + /* CLUSTER NODES */ + robj *o; + sds ci = clusterGenNodesDescription(0); + + o = createObject(OBJ_STRING, ci); + addReplyBulk(c, o); + decrRefCount(o); + } + else if (!strcasecmp(c->argv[1]->ptr, "myid") && c->argc == 2) { + /* CLUSTER MYID */ + addReplyBulkCBuffer(c, myself->name, CLUSTER_NAMELEN); + } + else if (!strcasecmp(c->argv[1]->ptr, "slots") && c->argc == 2) { + /* CLUSTER SLOTS */ + clusterReplyMultiBulkSlots(c); + } + else if (!strcasecmp(c->argv[1]->ptr, "flushslots") && c->argc == 2) { + /* CLUSTER FLUSHSLOTS */ + if (dictSize(server.db[0].dict) != 0) { + addReplyError(c, "DB must be empty to perform CLUSTER FLUSHSLOTS."); + return; + } + clusterDelNodeSlots(myself); + clusterDoBeforeSleep(CLUSTER_TODO_UPDATE_STATE | CLUSTER_TODO_SAVE_CONFIG); + addReply(c, shared.ok); + } + else if ((!strcasecmp(c->argv[1]->ptr, "addslots") || + !strcasecmp(c->argv[1]->ptr, "delslots")) && c->argc >= 3) + { + /* CLUSTER ADDSLOTS [slot] ... */ + /* CLUSTER DELSLOTS [slot] ... */ + int j, slot; + unsigned char *slots = zmalloc(CLUSTER_SLOTS); + int del = !strcasecmp(c->argv[1]->ptr, "delslots"); + + memset(slots, 0, CLUSTER_SLOTS); + /* Check that all the arguments are parseable and that all the + * slots are not already busy. */ + for (j = 2; j < c->argc; j++) { + if ((slot = getSlotOrReply(c, c->argv[j])) == -1) { + zfree(slots); + return; + } + if (del && server.cluster->slots[slot] == NULL) { + addReplyErrorFormat(c, "Slot %d is already unassigned", slot); + zfree(slots); + return; + } + else if (!del && server.cluster->slots[slot]) { + addReplyErrorFormat(c, "Slot %d is already busy", slot); + zfree(slots); + return; + } + if (slots[slot]++ == 1) { + addReplyErrorFormat(c, "Slot %d specified multiple times", + (int) slot); + zfree(slots); + return; + } + } + for (j = 0; j < CLUSTER_SLOTS; j++) { + if (slots[j]) { + int retval; + + /* If this slot was set as importing we can clear this + * state as now we are the real owner of the slot. */ + if (server.cluster->importing_slots_from[j]) + server.cluster->importing_slots_from[j] = NULL; + + retval = del ? clusterDelSlot(j) : + clusterAddSlot(myself, j); + serverAssertWithInfo(c, NULL, retval == C_OK); + } + } + zfree(slots); + clusterDoBeforeSleep(CLUSTER_TODO_UPDATE_STATE | CLUSTER_TODO_SAVE_CONFIG); + addReply(c, shared.ok); + } + else if (!strcasecmp(c->argv[1]->ptr, "setslot") && c->argc >= 4) { + /* SETSLOT 10 MIGRATING */ + /* SETSLOT 10 IMPORTING */ + /* SETSLOT 10 STABLE */ + /* SETSLOT 10 NODE */ + int slot; + clusterNode *n; + + if (nodeIsSlave(myself)) { + addReplyError(c, "Please use SETSLOT only with masters."); + return; + } + + if ((slot = getSlotOrReply(c, c->argv[2])) == -1) return; + + if (!strcasecmp(c->argv[3]->ptr, "migrating") && c->argc == 5) { + if (server.cluster->slots[slot] != myself) { + addReplyErrorFormat(c, "I'm not the owner of hash slot %u", slot); + return; + } + if ((n = clusterLookupNode(c->argv[4]->ptr)) == NULL) { + addReplyErrorFormat(c, "I don't know about node %s", + (char*) c->argv[4]->ptr); + return; + } + server.cluster->migrating_slots_to[slot] = n; + } + else if (!strcasecmp(c->argv[3]->ptr, "importing") && c->argc == 5) { + if (server.cluster->slots[slot] == myself) { + addReplyErrorFormat(c, + "I'm already the owner of hash slot %u", slot); + return; + } + if ((n = clusterLookupNode(c->argv[4]->ptr)) == NULL) { + addReplyErrorFormat(c, "I don't know about node %s", + (char*) c->argv[4]->ptr); + return; + } + server.cluster->importing_slots_from[slot] = n; + } + else if (!strcasecmp(c->argv[3]->ptr, "stable") && c->argc == 4) { + /* CLUSTER SETSLOT STABLE */ + server.cluster->importing_slots_from[slot] = NULL; + server.cluster->migrating_slots_to[slot] = NULL; + } + else if (!strcasecmp(c->argv[3]->ptr, "node") && c->argc == 5) { + /* CLUSTER SETSLOT NODE */ + clusterNode *n = clusterLookupNode(c->argv[4]->ptr); + + if (!n) { + addReplyErrorFormat(c, "Unknown node %s", + (char*) c->argv[4]->ptr); + return; + } + /* If this hash slot was served by 'myself' before to switch + * make sure there are no longer local keys for this hash slot. */ + if (server.cluster->slots[slot] == myself && n != myself) { + if (countKeysInSlot(slot) != 0) { + addReplyErrorFormat(c, + "Can't assign hashslot %d to a different node " + "while I still hold keys for this hash slot.", slot); + return; + } + } + /* If this slot is in migrating status but we have no keys + * for it assigning the slot to another node will clear + * the migratig status. */ + if (countKeysInSlot(slot) == 0 && + server.cluster->migrating_slots_to[slot]) + server.cluster->migrating_slots_to[slot] = NULL; + + /* If this node was importing this slot, assigning the slot to + * itself also clears the importing status. */ + if (n == myself && + server.cluster->importing_slots_from[slot]) + { + /* This slot was manually migrated, set this node configEpoch + * to a new epoch so that the new version can be propagated + * by the cluster. + * + * Note that if this ever results in a collision with another + * node getting the same configEpoch, for example because a + * failover happens at the same time we close the slot, the + * configEpoch collision resolution will fix it assigning + * a different epoch to each node. */ + if (clusterBumpConfigEpochWithoutConsensus() == C_OK) { + serverLog(LL_WARNING, + "configEpoch updated after importing slot %d", slot); + } + server.cluster->importing_slots_from[slot] = NULL; + } + clusterDelSlot(slot); + clusterAddSlot(n, slot); + } + else { + addReplyError(c, + "Invalid CLUSTER SETSLOT action or number of arguments"); + return; + } + clusterDoBeforeSleep(CLUSTER_TODO_SAVE_CONFIG | CLUSTER_TODO_UPDATE_STATE); + addReply(c, shared.ok); + } + else if (!strcasecmp(c->argv[1]->ptr, "bumpepoch") && c->argc == 2) { + /* CLUSTER BUMPEPOCH */ + int retval = clusterBumpConfigEpochWithoutConsensus(); + sds reply = sdscatprintf(sdsempty(), "+%s %llu\r\n", + (retval == C_OK) ? "BUMPED" : "STILL", + (PORT_ULONGLONG) myself->configEpoch); + addReplySds(c, reply); + } + else if (!strcasecmp(c->argv[1]->ptr, "info") && c->argc == 2) { + /* CLUSTER INFO */ + char *statestr[] = { "ok","fail","needhelp" }; + int slots_assigned = 0, slots_ok = 0, slots_pfail = 0, slots_fail = 0; + uint64_t myepoch; + int j; + + for (j = 0; j < CLUSTER_SLOTS; j++) { + clusterNode *n = server.cluster->slots[j]; + + if (n == NULL) continue; + slots_assigned++; + if (nodeFailed(n)) { + slots_fail++; + } + else if (nodeTimedOut(n)) { + slots_pfail++; + } + else { + slots_ok++; + } + } + + myepoch = (nodeIsSlave(myself) && myself->slaveof) ? + myself->slaveof->configEpoch : myself->configEpoch; + + sds info = sdscatprintf(sdsempty(), + "cluster_state:%s\r\n" + "cluster_slots_assigned:%d\r\n" + "cluster_slots_ok:%d\r\n" + "cluster_slots_pfail:%d\r\n" + "cluster_slots_fail:%d\r\n" + "cluster_known_nodes:%Iu\r\n" WIN_PORT_FIX /* %lu -> %Iu */ + "cluster_size:%d\r\n" + "cluster_current_epoch:%llu\r\n" + "cluster_my_epoch:%llu\r\n" + , statestr[server.cluster->state], + slots_assigned, + slots_ok, + slots_pfail, + slots_fail, + dictSize(server.cluster->nodes), + server.cluster->size, + (PORT_ULONGLONG) server.cluster->currentEpoch, + (PORT_ULONGLONG) myepoch + ); + + /* Show stats about messages sent and received. */ + PORT_LONGLONG tot_msg_sent = 0; + PORT_LONGLONG tot_msg_received = 0; + + for (int i = 0; i < CLUSTERMSG_TYPE_COUNT; i++) { + if (server.cluster->stats_bus_messages_sent[i] == 0) continue; + tot_msg_sent += server.cluster->stats_bus_messages_sent[i]; + info = sdscatprintf(info, + "cluster_stats_messages_%s_sent:%lld\r\n", + clusterGetMessageTypeString(i), + server.cluster->stats_bus_messages_sent[i]); + } + info = sdscatprintf(info, + "cluster_stats_messages_sent:%lld\r\n", tot_msg_sent); + + for (int i = 0; i < CLUSTERMSG_TYPE_COUNT; i++) { + if (server.cluster->stats_bus_messages_received[i] == 0) continue; + tot_msg_received += server.cluster->stats_bus_messages_received[i]; + info = sdscatprintf(info, + "cluster_stats_messages_%s_received:%lld\r\n", + clusterGetMessageTypeString(i), + server.cluster->stats_bus_messages_received[i]); + } + info = sdscatprintf(info, + "cluster_stats_messages_received:%lld\r\n", tot_msg_received); + + /* Produce the reply protocol. */ + addReplySds(c, sdscatprintf(sdsempty(), "$%Iu\r\n", WIN_PORT_FIX /* %lu -> %Iu */ + (PORT_ULONG)sdslen(info))); + addReplySds(c, info); + addReply(c, shared.crlf); + } + else if (!strcasecmp(c->argv[1]->ptr, "saveconfig") && c->argc == 2) { + int retval = clusterSaveConfig(1); + + if (retval == 0) + addReply(c, shared.ok); + else + addReplyErrorFormat(c, "error saving the cluster node config: %s", + IF_WIN32(wsa_strerror(errno), strerror(errno))); + } + else if (!strcasecmp(c->argv[1]->ptr, "keyslot") && c->argc == 3) { + /* CLUSTER KEYSLOT */ + sds key = c->argv[2]->ptr; + + addReplyLongLong(c, keyHashSlot(key, (int) sdslen(key))); WIN_PORT_FIX /* cast (int) */ + } + else if (!strcasecmp(c->argv[1]->ptr, "countkeysinslot") && c->argc == 3) { + /* CLUSTER COUNTKEYSINSLOT */ + PORT_LONGLONG slot; + + if (getLongLongFromObjectOrReply(c, c->argv[2], &slot, NULL) != C_OK) + return; + if (slot < 0 || slot >= CLUSTER_SLOTS) { + addReplyError(c, "Invalid slot"); + return; + } + addReplyLongLong(c, countKeysInSlot((unsigned int) slot)); WIN_PORT_FIX /* cast (unsigned int) */ + } + else if (!strcasecmp(c->argv[1]->ptr, "getkeysinslot") && c->argc == 4) { + /* CLUSTER GETKEYSINSLOT */ + PORT_LONGLONG maxkeys, slot; + unsigned int numkeys, j; + robj **keys; + + if (getLongLongFromObjectOrReply(c, c->argv[2], &slot, NULL) != C_OK) + return; + if (getLongLongFromObjectOrReply(c, c->argv[3], &maxkeys, NULL) + != C_OK) + return; + if (slot < 0 || slot >= CLUSTER_SLOTS || maxkeys < 0) { + addReplyError(c, "Invalid slot or number of keys"); + return; + } + + /* Avoid allocating more than needed in case of large COUNT argument + * and smaller actual number of keys. */ + unsigned int keys_in_slot = countKeysInSlot(slot); + if (maxkeys > keys_in_slot) maxkeys = keys_in_slot; + + keys = zmalloc(sizeof(robj*)*maxkeys); + numkeys = getKeysInSlot((unsigned int) slot, keys, (unsigned int) maxkeys); WIN_PORT_FIX /* cast (unsigned int) */ + addReplyMultiBulkLen(c, numkeys); + for (j = 0; j < numkeys; j++) { + addReplyBulk(c, keys[j]); + decrRefCount(keys[j]); + } + zfree(keys); + } + else if (!strcasecmp(c->argv[1]->ptr, "forget") && c->argc == 3) { + /* CLUSTER FORGET */ + clusterNode *n = clusterLookupNode(c->argv[2]->ptr); + + if (!n) { + addReplyErrorFormat(c, "Unknown node %s", (char*) c->argv[2]->ptr); + return; + } + else if (n == myself) { + addReplyError(c, "I tried hard but I can't forget myself..."); + return; + } + else if (nodeIsSlave(myself) && myself->slaveof == n) { + addReplyError(c, "Can't forget my master!"); + return; + } + clusterBlacklistAddNode(n); + clusterDelNode(n); + clusterDoBeforeSleep(CLUSTER_TODO_UPDATE_STATE | + CLUSTER_TODO_SAVE_CONFIG); + addReply(c, shared.ok); + } + else if (!strcasecmp(c->argv[1]->ptr, "replicate") && c->argc == 3) { + /* CLUSTER REPLICATE */ + clusterNode *n = clusterLookupNode(c->argv[2]->ptr); + + /* Lookup the specified node in our table. */ + if (!n) { + addReplyErrorFormat(c, "Unknown node %s", (char*) c->argv[2]->ptr); + return; + } + + /* I can't replicate myself. */ + if (n == myself) { + addReplyError(c, "Can't replicate myself"); + return; + } + + /* Can't replicate a slave. */ + if (nodeIsSlave(n)) { + addReplyError(c, "I can only replicate a master, not a slave."); + return; + } + + /* If the instance is currently a master, it should have no assigned + * slots nor keys to accept to replicate some other node. + * Slaves can switch to another master without issues. */ + if (nodeIsMaster(myself) && + (myself->numslots != 0 || dictSize(server.db[0].dict) != 0)) { + addReplyError(c, + "To set a master the node must be empty and " + "without assigned slots."); + return; + } + + /* Set the master. */ + clusterSetMaster(n); + clusterDoBeforeSleep(CLUSTER_TODO_UPDATE_STATE | CLUSTER_TODO_SAVE_CONFIG); + addReply(c, shared.ok); + } + else if (!strcasecmp(c->argv[1]->ptr, "slaves") && c->argc == 3) { + /* CLUSTER SLAVES */ + clusterNode *n = clusterLookupNode(c->argv[2]->ptr); + int j; + + /* Lookup the specified node in our table. */ + if (!n) { + addReplyErrorFormat(c, "Unknown node %s", (char*) c->argv[2]->ptr); + return; + } + + if (nodeIsSlave(n)) { + addReplyError(c, "The specified node is not a master"); + return; + } + + addReplyMultiBulkLen(c, n->numslaves); + for (j = 0; j < n->numslaves; j++) { + sds ni = clusterGenNodeDescription(n->slaves[j]); + addReplyBulkCString(c, ni); + sdsfree(ni); + } + } + else if (!strcasecmp(c->argv[1]->ptr, "count-failure-reports") && + c->argc == 3) + { + /* CLUSTER COUNT-FAILURE-REPORTS */ + clusterNode *n = clusterLookupNode(c->argv[2]->ptr); + + if (!n) { + addReplyErrorFormat(c, "Unknown node %s", (char*) c->argv[2]->ptr); + return; + } + else { + addReplyLongLong(c, clusterNodeFailureReportsCount(n)); + } + } + else if (!strcasecmp(c->argv[1]->ptr, "failover") && + (c->argc == 2 || c->argc == 3)) + { + /* CLUSTER FAILOVER [FORCE|TAKEOVER] */ + int force = 0, takeover = 0; + + if (c->argc == 3) { + if (!strcasecmp(c->argv[2]->ptr, "force")) { + force = 1; + } + else if (!strcasecmp(c->argv[2]->ptr, "takeover")) { + takeover = 1; + force = 1; /* Takeover also implies force. */ + } + else { + addReply(c, shared.syntaxerr); + return; + } + } + + /* Check preconditions. */ + if (nodeIsMaster(myself)) { + addReplyError(c, "You should send CLUSTER FAILOVER to a slave"); + return; + } + else if (myself->slaveof == NULL) { + addReplyError(c, "I'm a slave but my master is unknown to me"); + return; + } + else if (!force && + (nodeFailed(myself->slaveof) || + myself->slaveof->link == NULL)) + { + addReplyError(c, "Master is down or failed, " + "please use CLUSTER FAILOVER FORCE"); + return; + } + resetManualFailover(); + server.cluster->mf_end = mstime() + CLUSTER_MF_TIMEOUT; + + if (takeover) { + /* A takeover does not perform any initial check. It just + * generates a new configuration epoch for this node without + * consensus, claims the master's slots, and broadcast the new + * configuration. */ + serverLog(LL_WARNING, "Taking over the master (user request)."); + clusterBumpConfigEpochWithoutConsensus(); + clusterFailoverReplaceYourMaster(); + } + else if (force) { + /* If this is a forced failover, we don't need to talk with our + * master to agree about the offset. We just failover taking over + * it without coordination. */ + serverLog(LL_WARNING, "Forced failover user request accepted."); + server.cluster->mf_can_start = 1; + } + else { + serverLog(LL_WARNING, "Manual failover user request accepted."); + clusterSendMFStart(myself->slaveof); + } + addReply(c, shared.ok); + } + else if (!strcasecmp(c->argv[1]->ptr, "set-config-epoch") && c->argc == 3) + { + /* CLUSTER SET-CONFIG-EPOCH + * + * The user is allowed to set the config epoch only when a node is + * totally fresh: no config epoch, no other known node, and so forth. + * This happens at cluster creation time to start with a cluster where + * every node has a different node ID, without to rely on the conflicts + * resolution system which is too slow when a big cluster is created. */ + PORT_LONGLONG epoch; + + if (getLongLongFromObjectOrReply(c, c->argv[2], &epoch, NULL) != C_OK) + return; + + if (epoch < 0) { + addReplyErrorFormat(c, "Invalid config epoch specified: %lld", epoch); + } + else if (dictSize(server.cluster->nodes) > 1) { + addReplyError(c, "The user can assign a config epoch only when the " + "node does not know any other node."); + } + else if (myself->configEpoch != 0) { + addReplyError(c, "Node config epoch is already non-zero"); + } + else { + myself->configEpoch = epoch; + serverLog(LL_WARNING, + "configEpoch set to %llu via CLUSTER SET-CONFIG-EPOCH", + (PORT_ULONGLONG) myself->configEpoch); + + if (server.cluster->currentEpoch < (uint64_t) epoch) + server.cluster->currentEpoch = epoch; + /* No need to fsync the config here since in the unlucky event + * of a failure to persist the config, the conflict resolution code + * will assign an unique config to this node. */ + clusterDoBeforeSleep(CLUSTER_TODO_UPDATE_STATE | + CLUSTER_TODO_SAVE_CONFIG); + addReply(c, shared.ok); + } + } + else if (!strcasecmp(c->argv[1]->ptr, "reset") && + (c->argc == 2 || c->argc == 3)) + { + /* CLUSTER RESET [SOFT|HARD] */ + int hard = 0; + + /* Parse soft/hard argument. Default is soft. */ + if (c->argc == 3) { + if (!strcasecmp(c->argv[2]->ptr, "hard")) { + hard = 1; + } + else if (!strcasecmp(c->argv[2]->ptr, "soft")) { + hard = 0; + } + else { + addReply(c, shared.syntaxerr); + return; + } + } + + /* Slaves can be reset while containing data, but not master nodes + * that must be empty. */ + if (nodeIsMaster(myself) && dictSize(c->db->dict) != 0) { + addReplyError(c, "CLUSTER RESET can't be called with " + "master nodes containing keys"); + return; + } + clusterReset(hard); + addReply(c, shared.ok); + } + else { + addReplyError(c, "Wrong CLUSTER subcommand or number of arguments"); + } +} + +/* ----------------------------------------------------------------------------- + * DUMP, RESTORE and MIGRATE commands + * -------------------------------------------------------------------------- */ + + /* Generates a DUMP-format representation of the object 'o', adding it to the + * io stream pointed by 'rio'. This function can't fail. */ +void createDumpPayload(rio *payload, robj *o) { + unsigned char buf[2]; + uint64_t crc; + + /* Serialize the object in a RDB-like format. It consist of an object type + * byte followed by the serialized object. This is understood by RESTORE. */ + rioInitWithBuffer(payload, sdsempty()); + serverAssert(rdbSaveObjectType(payload, o)); + serverAssert(rdbSaveObject(payload, o)); + + /* Write the footer, this is how it looks like: + * ----------------+---------------------+---------------+ + * ... RDB payload | 2 bytes RDB version | 8 bytes CRC64 | + * ----------------+---------------------+---------------+ + * RDB version and CRC are both in little endian. + */ + + /* RDB version */ + buf[0] = RDB_VERSION & 0xff; + buf[1] = (RDB_VERSION >> 8) & 0xff; + payload->io.buffer.ptr = sdscatlen(payload->io.buffer.ptr, buf, 2); + + /* CRC64 */ + crc = crc64(0, (unsigned char*) payload->io.buffer.ptr, + sdslen(payload->io.buffer.ptr)); + memrev64ifbe(&crc); + payload->io.buffer.ptr = sdscatlen(payload->io.buffer.ptr, &crc, 8); +} + +/* Verify that the RDB version of the dump payload matches the one of this Redis + * instance and that the checksum is ok. + * If the DUMP payload looks valid C_OK is returned, otherwise C_ERR + * is returned. */ +int verifyDumpPayload(unsigned char *p, size_t len) { + unsigned char *footer; + uint16_t rdbver; + uint64_t crc; + + /* At least 2 bytes of RDB version and 8 of CRC64 should be present. */ + if (len < 10) return C_ERR; + footer = p + (len - 10); + + /* Verify RDB version */ + rdbver = (footer[1] << 8) | footer[0]; + if (rdbver > RDB_VERSION) return C_ERR; + + /* Verify CRC64 */ + crc = crc64(0, p, len - 8); + memrev64ifbe(&crc); + return (memcmp(&crc, footer + 2, 8) == 0) ? C_OK : C_ERR; +} + +/* DUMP keyname + * DUMP is actually not used by Redis Cluster but it is the obvious + * complement of RESTORE and can be useful for different applications. */ +void dumpCommand(client *c) { + robj *o, *dumpobj; + rio payload; + + /* Check if the key is here. */ + if ((o = lookupKeyRead(c->db, c->argv[1])) == NULL) { + addReply(c, shared.nullbulk); + return; + } + + /* Create the DUMP encoded representation. */ + createDumpPayload(&payload, o); + + /* Transfer to the client */ + dumpobj = createObject(OBJ_STRING, payload.io.buffer.ptr); + addReplyBulk(c, dumpobj); + decrRefCount(dumpobj); + return; +} + +/* RESTORE key ttl serialized-value [REPLACE] */ +void restoreCommand(client *c) { + PORT_LONGLONG ttl; + rio payload; + int j, type, replace = 0; + robj *obj; + + /* Parse additional options */ + for (j = 4; j < c->argc; j++) { + if (!strcasecmp(c->argv[j]->ptr, "replace")) { + replace = 1; + } + else { + addReply(c, shared.syntaxerr); + return; + } + } + + /* Make sure this key does not already exist here... */ + if (!replace && lookupKeyWrite(c->db, c->argv[1]) != NULL) { + addReply(c, shared.busykeyerr); + return; + } + + /* Check if the TTL value makes sense */ + if (getLongLongFromObjectOrReply(c, c->argv[2], &ttl, NULL) != C_OK) { + return; + } + else if (ttl < 0) { + addReplyError(c, "Invalid TTL value, must be >= 0"); + return; + } + + /* Verify RDB version and data checksum. */ + if (verifyDumpPayload(c->argv[3]->ptr, sdslen(c->argv[3]->ptr)) == C_ERR) + { + addReplyError(c, "DUMP payload version or checksum are wrong"); + return; + } + + rioInitWithBuffer(&payload, c->argv[3]->ptr); + if (((type = rdbLoadObjectType(&payload)) == -1) || + ((obj = rdbLoadObject(type, &payload)) == NULL)) + { + addReplyError(c, "Bad data format"); + return; + } + + /* Remove the old key if needed. */ + if (replace) dbDelete(c->db, c->argv[1]); + + /* Create the key and set the TTL if any */ + dbAdd(c->db, c->argv[1], obj); + if (ttl) setExpire(c, c->db, c->argv[1], mstime() + ttl); + signalModifiedKey(c->db, c->argv[1]); + addReply(c, shared.ok); + server.dirty++; +} + +/* MIGRATE socket cache implementation. + * + * We take a map between host:ip and a TCP socket that we used to connect + * to this instance in recent time. + * This sockets are closed when the max number we cache is reached, and also + * in serverCron() when they are around for more than a few seconds. */ +#define MIGRATE_SOCKET_CACHE_ITEMS 64 /* max num of items in the cache. */ +#define MIGRATE_SOCKET_CACHE_TTL 10 /* close cached sockets after 10 sec. */ + +typedef struct migrateCachedSocket { + int fd; + PORT_LONG last_dbid; + time_t last_use_time; +} migrateCachedSocket; + +/* Return a migrateCachedSocket containing a TCP socket connected with the + * target instance, possibly returning a cached one. + * + * This function is responsible of sending errors to the client if a + * connection can't be established. In this case -1 is returned. + * Otherwise on success the socket is returned, and the caller should not + * attempt to free it after usage. + * + * If the caller detects an error while using the socket, migrateCloseSocket() + * should be called so that the connection will be created from scratch + * the next time. */ +migrateCachedSocket* migrateGetSocket(client *c, robj *host, robj *port, PORT_LONG timeout) { + int fd; + sds name = sdsempty(); + migrateCachedSocket *cs; + + /* Check if we have an already cached socket for this ip:port pair. */ + name = sdscatlen(name, host->ptr, sdslen(host->ptr)); + name = sdscatlen(name, ":", 1); + name = sdscatlen(name, port->ptr, sdslen(port->ptr)); + cs = dictFetchValue(server.migrate_cached_sockets, name); + if (cs) { + sdsfree(name); + cs->last_use_time = server.unixtime; + return cs; + } + + /* No cached socket, create one. */ + if (dictSize(server.migrate_cached_sockets) == MIGRATE_SOCKET_CACHE_ITEMS) { + /* Too many items, drop one at random. */ + dictEntry *de = dictGetRandomKey(server.migrate_cached_sockets); + cs = dictGetVal(de); + close(cs->fd); + zfree(cs); + dictDelete(server.migrate_cached_sockets, dictGetKey(de)); + } + + /* Create the socket */ + fd = anetTcpNonBlockConnect(server.neterr, c->argv[1]->ptr, + atoi(c->argv[2]->ptr)); + if (fd == -1) { + sdsfree(name); + addReplyErrorFormat(c, "Can't connect to target node: %s", + server.neterr); + return NULL; + } + anetEnableTcpNoDelay(server.neterr, fd); + + /* Check if it connects within the specified timeout. */ + if ((aeWait(fd, AE_WRITABLE, timeout) & AE_WRITABLE) == 0) { + sdsfree(name); + addReplySds(c, + sdsnew("-IOERR error or timeout connecting to the client\r\n")); + close(fd); + return NULL; + } + + /* Add to the cache and return it to the caller. */ + cs = zmalloc(sizeof(*cs)); + cs->fd = fd; + cs->last_dbid = -1; + cs->last_use_time = server.unixtime; + dictAdd(server.migrate_cached_sockets, name, cs); + return cs; +} + +/* Free a migrate cached connection. */ +void migrateCloseSocket(robj *host, robj *port) { + sds name = sdsempty(); + migrateCachedSocket *cs; + + name = sdscatlen(name, host->ptr, sdslen(host->ptr)); + name = sdscatlen(name, ":", 1); + name = sdscatlen(name, port->ptr, sdslen(port->ptr)); + cs = dictFetchValue(server.migrate_cached_sockets, name); + if (!cs) { + sdsfree(name); + return; + } + + close(cs->fd); + zfree(cs); + dictDelete(server.migrate_cached_sockets, name); + sdsfree(name); +} + +void migrateCloseTimedoutSockets(void) { + dictIterator *di = dictGetSafeIterator(server.migrate_cached_sockets); + dictEntry *de; + + while ((de = dictNext(di)) != NULL) { + migrateCachedSocket *cs = dictGetVal(de); + + if ((server.unixtime - cs->last_use_time) > MIGRATE_SOCKET_CACHE_TTL) { + close(cs->fd); + zfree(cs); + dictDelete(server.migrate_cached_sockets, dictGetKey(de)); + } + } + dictReleaseIterator(di); +} + +/* MIGRATE host port key dbid timeout [COPY | REPLACE | AUTH password] + * + * On in the multiple keys form: + * + * MIGRATE host port "" dbid timeout [COPY | REPLACE | AUTH password] KEYS key1 + * key2 ... keyN */ +void migrateCommand(client *c) { + migrateCachedSocket *cs; + int copy=0, replace=0, j; + char *password = NULL; + PORT_LONG timeout; + PORT_LONG dbid; + robj **ov = NULL; /* Objects to migrate. */ + robj **kv = NULL; /* Key names. */ + robj **newargv = NULL; /* Used to rewrite the command as DEL ... keys ... */ + rio cmd, payload; + int may_retry = 1; + int write_error = 0; + int argv_rewritten = 0; + + /* To support the KEYS option we need the following additional state. */ + int first_key = 3; /* Argument index of the first key. */ + int num_keys = 1; /* By default only migrate the 'key' argument. */ + + /* Parse additional options */ + for (j = 6; j < c->argc; j++) { + int moreargs = j < c->argc-1; + if (!strcasecmp(c->argv[j]->ptr, "copy")) { + copy = 1; + } else if (!strcasecmp(c->argv[j]->ptr,"replace")) { + replace = 1; + } else if (!strcasecmp(c->argv[j]->ptr,"auth")) { + if (!moreargs) { + addReply(c,shared.syntaxerr); + return; + } + j++; + password = c->argv[j]->ptr; + } else if (!strcasecmp(c->argv[j]->ptr,"keys")) { + if (sdslen(c->argv[3]->ptr) != 0) { + addReplyError(c, + "When using MIGRATE KEYS option, the key argument" + " must be set to the empty string"); + return; + } + first_key = j + 1; + num_keys = c->argc - j - 1; + break; /* All the remaining args are keys. */ + } + else { + addReply(c, shared.syntaxerr); + return; + } + } + + /* Sanity check */ + if (getLongFromObjectOrReply(c, c->argv[5], &timeout, NULL) != C_OK || + getLongFromObjectOrReply(c, c->argv[4], &dbid, NULL) != C_OK) + { + return; + } + if (timeout <= 0) timeout = 1000; + + /* Check if the keys are here. If at least one key is to migrate, do it + * otherwise if all the keys are missing reply with "NOKEY" to signal + * the caller there was nothing to migrate. We don't return an error in + * this case, since often this is due to a normal condition like the key + * expiring in the meantime. */ + ov = zrealloc(ov, sizeof(robj*)*num_keys); + kv = zrealloc(kv, sizeof(robj*)*num_keys); + int oi = 0; + + for (j = 0; j < num_keys; j++) { + if ((ov[oi] = lookupKeyRead(c->db, c->argv[first_key + j])) != NULL) { + kv[oi] = c->argv[first_key + j]; + oi++; + } + } + num_keys = oi; + if (num_keys == 0) { + zfree(ov); zfree(kv); + addReplySds(c, sdsnew("+NOKEY\r\n")); + return; + } + +try_again: + write_error = 0; + + /* Connect */ + cs = migrateGetSocket(c, c->argv[1], c->argv[2], timeout); + if (cs == NULL) { + zfree(ov); zfree(kv); + return; /* error sent to the client by migrateGetSocket() */ + } + + rioInitWithBuffer(&cmd, sdsempty()); + + /* Authentication */ + if (password) { + serverAssertWithInfo(c,NULL,rioWriteBulkCount(&cmd,'*',2)); + serverAssertWithInfo(c,NULL,rioWriteBulkString(&cmd,"AUTH",4)); + serverAssertWithInfo(c,NULL,rioWriteBulkString(&cmd,password, + sdslen(password))); + } + + /* Send the SELECT command if the current DB is not already selected. */ + int select = cs->last_dbid != dbid; /* Should we emit SELECT? */ + if (select) { + serverAssertWithInfo(c, NULL, rioWriteBulkCount(&cmd, '*', 2)); + serverAssertWithInfo(c, NULL, rioWriteBulkString(&cmd, "SELECT", 6)); + serverAssertWithInfo(c, NULL, rioWriteBulkLongLong(&cmd, dbid)); + } + + /* Create RESTORE payload and generate the protocol to call the command. */ + for (j = 0; j < num_keys; j++) { + PORT_LONGLONG ttl = 0; + PORT_LONGLONG expireat = getExpire(c->db, kv[j]); + + if (expireat != -1) { + ttl = expireat - mstime(); + if (ttl < 1) ttl = 1; + } + serverAssertWithInfo(c,NULL, + rioWriteBulkCount(&cmd,'*',replace ? 5 : 4)); + + if (server.cluster_enabled) + serverAssertWithInfo(c, NULL, + rioWriteBulkString(&cmd, "RESTORE-ASKING", 14)); + else + serverAssertWithInfo(c, NULL, rioWriteBulkString(&cmd, "RESTORE", 7)); + serverAssertWithInfo(c, NULL, sdsEncodedObject(kv[j])); + serverAssertWithInfo(c, NULL, rioWriteBulkString(&cmd, kv[j]->ptr, + sdslen(kv[j]->ptr))); + serverAssertWithInfo(c, NULL, rioWriteBulkLongLong(&cmd, ttl)); + + /* Emit the payload argument, that is the serialized object using + * the DUMP format. */ + createDumpPayload(&payload, ov[j]); + serverAssertWithInfo(c, NULL, + rioWriteBulkString(&cmd, payload.io.buffer.ptr, + sdslen(payload.io.buffer.ptr))); + sdsfree(payload.io.buffer.ptr); + + /* Add the REPLACE option to the RESTORE command if it was specified + * as a MIGRATE option. */ + if (replace) + serverAssertWithInfo(c, NULL, rioWriteBulkString(&cmd, "REPLACE", 7)); + } + + /* Transfer the query to the other node in 64K chunks. */ + errno = 0; + { + sds buf = cmd.io.buffer.ptr; + size_t pos = 0, towrite; + int nwritten = 0; + +#ifdef _WIN32 + while ((towrite = sdslen(buf) - pos) > 0) { + towrite = (towrite > (64 * 1024) ? (64 * 1024) : towrite); + while (nwritten != (signed) towrite) { + nwritten = (int) syncWrite(cs->fd, buf + pos, (ssize_t) towrite, timeout); + if (nwritten != (signed) towrite) { + DWORD err = GetLastError(); + if (err == WSAEWOULDBLOCK) { + // Likely send buffer is full. A short delay or two is sufficient to allow this to work. + serverLog(LL_VERBOSE, "In migrate. WSAEWOULDBLOCK with synchronous socket: sleeping for 0.1s"); + Sleep(100); + } + else { + serverLog(LL_WARNING, "SyncWrite failure toWrite=%d written=%d err=%d timeout=%d ", towrite, nwritten, GetLastError(), timeout); + goto socket_err; + } + } + } + pos += nwritten; + nwritten = 0; + } +#else + while ((towrite = sdslen(buf) - pos) > 0) { + towrite = (towrite > (64 * 1024) ? (64 * 1024) : towrite); + nwritten = syncWrite(cs->fd, buf + pos, towrite, timeout); + if (nwritten != (signed) towrite) { + write_error = 1; + goto socket_err; + } + pos += nwritten; + } +#endif + } + + char buf0[1024]; /* Auth reply. */ + char buf1[1024]; /* Select reply. */ + char buf2[1024]; /* Restore reply. */ + + /* Read the AUTH reply if needed. */ + if (password && syncReadLine(cs->fd, buf0, sizeof(buf0), timeout) <= 0) + goto socket_err; + + /* Read the SELECT reply if needed. */ + if (select && syncReadLine(cs->fd, buf1, sizeof(buf1), timeout) <= 0) + goto socket_err; + + /* Read the RESTORE replies. */ + int error_from_target = 0; + int socket_error = 0; + int del_idx = 1; /* Index of the key argument for the replicated DEL op. */ + + if (!copy) newargv = zmalloc(sizeof(robj*)*(num_keys + 1)); + + for (j = 0; j < num_keys; j++) { + if (syncReadLine(cs->fd, buf2, sizeof(buf2), timeout) <= 0) { + socket_error = 1; + break; + } + if ((password && buf0[0] == '-') || + (select && buf1[0] == '-') || + buf2[0] == '-') + { + /* On error assume that last_dbid is no longer valid. */ + if (!error_from_target) { + cs->last_dbid = -1; + char *errbuf; + if (password && buf0[0] == '-') errbuf = buf0; + else if (select && buf1[0] == '-') errbuf = buf1; + else errbuf = buf2; + + error_from_target = 1; + addReplyErrorFormat(c,"Target instance replied with error: %s", + errbuf+1); + } + } + else { + if (!copy) { + /* No COPY option: remove the local key, signal the change. */ + dbDelete(c->db, kv[j]); + signalModifiedKey(c->db, kv[j]); + server.dirty++; + + /* Populate the argument vector to replace the old one. */ + newargv[del_idx++] = kv[j]; + incrRefCount(kv[j]); + } + } + } + + /* On socket error, if we want to retry, do it now before rewriting the + * command vector. We only retry if we are sure nothing was processed + * and we failed to read the first reply (j == 0 test). */ + if (!error_from_target && socket_error && j == 0 && may_retry && + errno != ETIMEDOUT) + { + goto socket_err; /* A retry is guaranteed because of tested conditions.*/ + } + + /* On socket errors, close the migration socket now that we still have + * the original host/port in the ARGV. Later the original command may be + * rewritten to DEL and will be too later. */ + if (socket_error) migrateCloseSocket(c->argv[1], c->argv[2]); + + if (!copy) { + /* Translate MIGRATE as DEL for replication/AOF. Note that we do + * this only for the keys for which we received an acknowledgement + * from the receiving Redis server, by using the del_idx index. */ + if (del_idx > 1) { + newargv[0] = createStringObject("DEL", 3); + /* Note that the following call takes ownership of newargv. */ + replaceClientCommandVector(c, del_idx, newargv); + argv_rewritten = 1; + } + else { + /* No key transfer acknowledged, no need to rewrite as DEL. */ + zfree(newargv); + } + newargv = NULL; /* Make it safe to call zfree() on it in the future. */ + } + + /* If we are here and a socket error happened, we don't want to retry. + * Just signal the problem to the client, but only do it if we did not + * already queue a different error reported by the destination server. */ + if (!error_from_target && socket_error) { + may_retry = 0; + goto socket_err; + } + + if (!error_from_target) { + /* Success! Update the last_dbid in migrateCachedSocket, so that we can + * avoid SELECT the next time if the target DB is the same. Reply +OK. + * + * Note: If we reached this point, even if socket_error is true + * still the SELECT command succeeded (otherwise the code jumps to + * socket_err label. */ + cs->last_dbid = dbid; + addReply(c, shared.ok); + } + else { + /* On error we already sent it in the for loop above, and set + * the currently selected socket to -1 to force SELECT the next time. */ + } + + sdsfree(cmd.io.buffer.ptr); + zfree(ov); zfree(kv); zfree(newargv); + return; + + /* On socket errors we try to close the cached socket and try again. + * It is very common for the cached socket to get closed, if just reopening + * it works it's a shame to notify the error to the caller. */ +socket_err: +#ifdef _WIN32 + serverLog(LL_WARNING, "syncReadLine failure err=%d timeout=%d ", GetLastError(), timeout); +#endif + /* Cleanup we want to perform in both the retry and no retry case. + * Note: Closing the migrate socket will also force SELECT next time. */ + sdsfree(cmd.io.buffer.ptr); + + /* If the command was rewritten as DEL and there was a socket error, + * we already closed the socket earlier. While migrateCloseSocket() + * is idempotent, the host/port arguments are now gone, so don't do it + * again. */ + if (!argv_rewritten) migrateCloseSocket(c->argv[1], c->argv[2]); + zfree(newargv); + newargv = NULL; /* This will get reallocated on retry. */ + + /* Retry only if it's not a timeout and we never attempted a retry + * (or the code jumping here did not set may_retry to zero). */ + if (errno != ETIMEDOUT && may_retry) { + may_retry = 0; + goto try_again; + } + + /* Cleanup we want to do if no retry is attempted. */ + zfree(ov); zfree(kv); + addReplySds(c, + sdscatprintf(sdsempty(), + "-IOERR error or timeout %s to target instance\r\n", + write_error ? "writing" : "reading")); + return; +} + +/* ----------------------------------------------------------------------------- + * Cluster functions related to serving / redirecting clients + * -------------------------------------------------------------------------- */ + + /* The ASKING command is required after a -ASK redirection. + * The client should issue ASKING before to actually send the command to + * the target instance. See the Redis Cluster specification for more + * information. */ +void askingCommand(client *c) { + if (server.cluster_enabled == 0) { + addReplyError(c, "This instance has cluster support disabled"); + return; + } + c->flags |= CLIENT_ASKING; + addReply(c, shared.ok); +} + +/* The READONLY command is used by clients to enter the read-only mode. + * In this mode slaves will not redirect clients as long as clients access + * with read-only commands to keys that are served by the slave's master. */ +void readonlyCommand(client *c) { + if (server.cluster_enabled == 0) { + addReplyError(c, "This instance has cluster support disabled"); + return; + } + c->flags |= CLIENT_READONLY; + addReply(c, shared.ok); +} + +/* The READWRITE command just clears the READONLY command state. */ +void readwriteCommand(client *c) { + c->flags &= ~CLIENT_READONLY; + addReply(c, shared.ok); +} + +/* Return the pointer to the cluster node that is able to serve the command. + * For the function to succeed the command should only target either: + * + * 1) A single key (even multiple times like LPOPRPUSH mylist mylist). + * 2) Multiple keys in the same hash slot, while the slot is stable (no + * resharding in progress). + * + * On success the function returns the node that is able to serve the request. + * If the node is not 'myself' a redirection must be perfomed. The kind of + * redirection is specified setting the integer passed by reference + * 'error_code', which will be set to CLUSTER_REDIR_ASK or + * CLUSTER_REDIR_MOVED. + * + * When the node is 'myself' 'error_code' is set to CLUSTER_REDIR_NONE. + * + * If the command fails NULL is returned, and the reason of the failure is + * provided via 'error_code', which will be set to: + * + * CLUSTER_REDIR_CROSS_SLOT if the request contains multiple keys that + * don't belong to the same hash slot. + * + * CLUSTER_REDIR_UNSTABLE if the request contains multiple keys + * belonging to the same slot, but the slot is not stable (in migration or + * importing state, likely because a resharding is in progress). + * + * CLUSTER_REDIR_DOWN_UNBOUND if the request addresses a slot which is + * not bound to any node. In this case the cluster global state should be + * already "down" but it is fragile to rely on the update of the global state, + * so we also handle it here. + * + * CLUSTER_REDIR_DOWN_STATE if the cluster is down but the user attempts to + * execute a command that addresses one or more keys. */ +clusterNode *getNodeByQuery(client *c, struct redisCommand *cmd, robj **argv, int argc, int *hashslot, int *error_code) { + clusterNode *n = NULL; + robj *firstkey = NULL; + int multiple_keys = 0; + multiState *ms, _ms; + multiCmd mc; + int i, slot = 0, migrating_slot = 0, importing_slot = 0, missing_keys = 0; + + /* Set error code optimistically for the base case. */ + if (error_code) *error_code = CLUSTER_REDIR_NONE; + + /* We handle all the cases as if they were EXEC commands, so we have + * a common code path for everything */ + if (cmd->proc == execCommand) { + /* If CLIENT_MULTI flag is not set EXEC is just going to return an + * error. */ + if (!(c->flags & CLIENT_MULTI)) return myself; + ms = &c->mstate; + } + else { + /* In order to have a single codepath create a fake Multi State + * structure if the client is not in MULTI/EXEC state, this way + * we have a single codepath below. */ + ms = &_ms; + _ms.commands = &mc; + _ms.count = 1; + mc.argv = argv; + mc.argc = argc; + mc.cmd = cmd; + } + + /* Check that all the keys are in the same hash slot, and obtain this + * slot and the node associated. */ + for (i = 0; i < ms->count; i++) { + struct redisCommand *mcmd; + robj **margv; + int margc, *keyindex, numkeys, j; + + mcmd = ms->commands[i].cmd; + margc = ms->commands[i].argc; + margv = ms->commands[i].argv; + + keyindex = getKeysFromCommand(mcmd, margv, margc, &numkeys); + for (j = 0; j < numkeys; j++) { + robj *thiskey = margv[keyindex[j]]; + int thisslot = keyHashSlot((char*) thiskey->ptr, + (int) sdslen(thiskey->ptr)); WIN_PORT_FIX /* cast (int) */ + + if (firstkey == NULL) { + /* This is the first key we see. Check what is the slot + * and node. */ + firstkey = thiskey; + slot = thisslot; + n = server.cluster->slots[slot]; + + /* Error: If a slot is not served, we are in "cluster down" + * state. However the state is yet to be updated, so this was + * not trapped earlier in processCommand(). Report the same + * error to the client. */ + if (n == NULL) { + getKeysFreeResult(keyindex); + if (error_code) + *error_code = CLUSTER_REDIR_DOWN_UNBOUND; + return NULL; + } + + /* If we are migrating or importing this slot, we need to check + * if we have all the keys in the request (the only way we + * can safely serve the request, otherwise we return a TRYAGAIN + * error). To do so we set the importing/migrating state and + * increment a counter for every missing key. */ + if (n == myself && + server.cluster->migrating_slots_to[slot] != NULL) + { + migrating_slot = 1; + } + else if (server.cluster->importing_slots_from[slot] != NULL) { + importing_slot = 1; + } + } + else { + /* If it is not the first key, make sure it is exactly + * the same key as the first we saw. */ + if (!equalStringObjects(firstkey, thiskey)) { + if (slot != thisslot) { + /* Error: multiple keys from different slots. */ + getKeysFreeResult(keyindex); + if (error_code) + *error_code = CLUSTER_REDIR_CROSS_SLOT; + return NULL; + } + else { + /* Flag this request as one with multiple different + * keys. */ + multiple_keys = 1; + } + } + } + + /* Migarting / Improrting slot? Count keys we don't have. */ + if ((migrating_slot || importing_slot) && + lookupKeyRead(&server.db[0], thiskey) == NULL) + { + missing_keys++; + } + } + getKeysFreeResult(keyindex); + } + + /* No key at all in command? then we can serve the request + * without redirections or errors in all the cases. */ + if (n == NULL) return myself; + + /* Cluster is globally down but we got keys? We can't serve the request. */ + if (server.cluster->state != CLUSTER_OK) { + if (error_code) *error_code = CLUSTER_REDIR_DOWN_STATE; + return NULL; + } + + /* Return the hashslot by reference. */ + if (hashslot) *hashslot = slot; + + /* MIGRATE always works in the context of the local node if the slot + * is open (migrating or importing state). We need to be able to freely + * move keys among instances in this case. */ + if ((migrating_slot || importing_slot) && cmd->proc == migrateCommand) + return myself; + + /* If we don't have all the keys and we are migrating the slot, send + * an ASK redirection. */ + if (migrating_slot && missing_keys) { + if (error_code) *error_code = CLUSTER_REDIR_ASK; + return server.cluster->migrating_slots_to[slot]; + } + + /* If we are receiving the slot, and the client correctly flagged the + * request as "ASKING", we can serve the request. However if the request + * involves multiple keys and we don't have them all, the only option is + * to send a TRYAGAIN error. */ + if (importing_slot && + (c->flags & CLIENT_ASKING || cmd->flags & CMD_ASKING)) + { + if (multiple_keys && missing_keys) { + if (error_code) *error_code = CLUSTER_REDIR_UNSTABLE; + return NULL; + } + else { + return myself; + } + } + + /* Handle the read-only client case reading from a slave: if this + * node is a slave and the request is about an hash slot our master + * is serving, we can reply without redirection. */ + if (c->flags & CLIENT_READONLY && + cmd->flags & CMD_READONLY && + nodeIsSlave(myself) && + myself->slaveof == n) + { + return myself; + } + + /* Base case: just return the right node. However if this node is not + * myself, set error_code to MOVED since we need to issue a rediretion. */ + if (n != myself && error_code) *error_code = CLUSTER_REDIR_MOVED; + return n; +} + +/* Send the client the right redirection code, according to error_code + * that should be set to one of CLUSTER_REDIR_* macros. + * + * If CLUSTER_REDIR_ASK or CLUSTER_REDIR_MOVED error codes + * are used, then the node 'n' should not be NULL, but should be the + * node we want to mention in the redirection. Moreover hashslot should + * be set to the hash slot that caused the redirection. */ +void clusterRedirectClient(client *c, clusterNode *n, int hashslot, int error_code) { + if (error_code == CLUSTER_REDIR_CROSS_SLOT) { + addReplySds(c, sdsnew("-CROSSSLOT Keys in request don't hash to the same slot\r\n")); + } + else if (error_code == CLUSTER_REDIR_UNSTABLE) { + /* The request spawns mutliple keys in the same slot, + * but the slot is not "stable" currently as there is + * a migration or import in progress. */ + addReplySds(c, sdsnew("-TRYAGAIN Multiple keys request during rehashing of slot\r\n")); + } + else if (error_code == CLUSTER_REDIR_DOWN_STATE) { + addReplySds(c, sdsnew("-CLUSTERDOWN The cluster is down\r\n")); + } + else if (error_code == CLUSTER_REDIR_DOWN_UNBOUND) { + addReplySds(c, sdsnew("-CLUSTERDOWN Hash slot not served\r\n")); + } + else if (error_code == CLUSTER_REDIR_MOVED || + error_code == CLUSTER_REDIR_ASK) + { + addReplySds(c, sdscatprintf(sdsempty(), + "-%s %d %s:%d\r\n", + (error_code == CLUSTER_REDIR_ASK) ? "ASK" : "MOVED", + hashslot, n->ip, n->port)); + } + else { + serverPanic("getNodeByQuery() unknown error."); + } +} + +/* This function is called by the function processing clients incrementally + * to detect timeouts, in order to handle the following case: + * + * 1) A client blocks with BLPOP or similar blocking operation. + * 2) The master migrates the hash slot elsewhere or turns into a slave. + * 3) The client may remain blocked forever (or up to the max timeout time) + * waiting for a key change that will never happen. + * + * If the client is found to be blocked into an hash slot this node no + * longer handles, the client is sent a redirection error, and the function + * returns 1. Otherwise 0 is returned and no operation is performed. */ +int clusterRedirectBlockedClientIfNeeded(client *c) { + if (c->flags & CLIENT_BLOCKED && c->btype == BLOCKED_LIST) { + dictEntry *de; + dictIterator *di; + + /* If the cluster is down, unblock the client with the right error. */ + if (server.cluster->state == CLUSTER_FAIL) { + clusterRedirectClient(c, NULL, 0, CLUSTER_REDIR_DOWN_STATE); + return 1; + } + + /* All keys must belong to the same slot, so check first key only. */ + di = dictGetIterator(c->bpop.keys); + if ((de = dictNext(di)) != NULL) { + robj *key = dictGetKey(de); + int slot = keyHashSlot((char*) key->ptr, (int) sdslen(key->ptr)); WIN_PORT_FIX /* cast (int) */ + clusterNode *node = server.cluster->slots[slot]; + + /* We send an error and unblock the client if: + * 1) The slot is unassigned, emitting a cluster down error. + * 2) The slot is not handled by this node, nor being imported. */ + if (node != myself && + server.cluster->importing_slots_from[slot] == NULL) + { + if (node == NULL) { + clusterRedirectClient(c, NULL, 0, + CLUSTER_REDIR_DOWN_UNBOUND); + } + else { + clusterRedirectClient(c, node, slot, + CLUSTER_REDIR_MOVED); + } + dictReleaseIterator(di); + return 1; + } + } + dictReleaseIterator(di); + } + return 0; +} \ No newline at end of file diff --git a/src/cluster.h b/src/cluster.h index 9ad47d21..ff8dc561 100644 --- a/src/cluster.h +++ b/src/cluster.h @@ -16,6 +16,7 @@ #define CLUSTER_DEFAULT_NODE_TIMEOUT 15000 #define CLUSTER_DEFAULT_SLAVE_VALIDITY 10 /* Slave max data age factor. */ #define CLUSTER_DEFAULT_REQUIRE_FULL_COVERAGE 1 +#define CLUSTER_DEFAULT_SLAVE_NO_FAILOVER 0 /* Failover by default. */ #define CLUSTER_FAIL_REPORT_VALIDITY_MULT 2 /* Fail report validity. */ #define CLUSTER_FAIL_UNDO_TIME_MULT 2 /* Undo fail if master is back. */ #define CLUSTER_FAIL_UNDO_TIME_ADD 10 /* Some additional time. */ @@ -55,6 +56,7 @@ typedef struct clusterLink { #define CLUSTER_NODE_NOADDR 64 /* We don't know the address of this node */ #define CLUSTER_NODE_MEET 128 /* Send a MEET message to this node */ #define CLUSTER_NODE_MIGRATE_TO 256 /* Master elegible for replica migration. */ +#define CLUSTER_NODE_NOFAILOVER 512 /* Slave will not try to failver. */ #define CLUSTER_NODE_NULL_NAME "\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000" #define nodeIsMaster(n) ((n)->flags & CLUSTER_NODE_MASTER) @@ -64,6 +66,7 @@ typedef struct clusterLink { #define nodeWithoutAddr(n) ((n)->flags & CLUSTER_NODE_NOADDR) #define nodeTimedOut(n) ((n)->flags & CLUSTER_NODE_PFAIL) #define nodeFailed(n) ((n)->flags & CLUSTER_NODE_FAIL) +#define nodeCantFailover(n) ((n)->flags & CLUSTER_NODE_NOFAILOVER) /* Reasons why a slave is not able to failover. */ #define CLUSTER_CANT_FAILOVER_NONE 0 diff --git a/src/cluster.h.bak b/src/cluster.h.bak new file mode 100644 index 00000000..9ad47d21 --- /dev/null +++ b/src/cluster.h.bak @@ -0,0 +1,270 @@ +#ifndef __CLUSTER_H +#define __CLUSTER_H + +/*----------------------------------------------------------------------------- + * Redis cluster data structures, defines, exported API. + *----------------------------------------------------------------------------*/ + +#define CLUSTER_SLOTS 16384 +#define CLUSTER_OK 0 /* Everything looks ok */ +#define CLUSTER_FAIL 1 /* The cluster can't work */ +#define CLUSTER_NAMELEN 40 /* sha1 hex length */ +#define CLUSTER_PORT_INCR 10000 /* Cluster port = baseport + PORT_INCR */ + +/* The following defines are amount of time, sometimes expressed as + * multiplicators of the node timeout value (when ending with MULT). */ +#define CLUSTER_DEFAULT_NODE_TIMEOUT 15000 +#define CLUSTER_DEFAULT_SLAVE_VALIDITY 10 /* Slave max data age factor. */ +#define CLUSTER_DEFAULT_REQUIRE_FULL_COVERAGE 1 +#define CLUSTER_FAIL_REPORT_VALIDITY_MULT 2 /* Fail report validity. */ +#define CLUSTER_FAIL_UNDO_TIME_MULT 2 /* Undo fail if master is back. */ +#define CLUSTER_FAIL_UNDO_TIME_ADD 10 /* Some additional time. */ +#define CLUSTER_FAILOVER_DELAY 5 /* Seconds */ +#define CLUSTER_DEFAULT_MIGRATION_BARRIER 1 +#define CLUSTER_MF_TIMEOUT 5000 /* Milliseconds to do a manual failover. */ +#define CLUSTER_MF_PAUSE_MULT 2 /* Master pause manual failover mult. */ +#define CLUSTER_SLAVE_MIGRATION_DELAY 5000 /* Delay for slave migration. */ + +/* Redirection errors returned by getNodeByQuery(). */ +#define CLUSTER_REDIR_NONE 0 /* Node can serve the request. */ +#define CLUSTER_REDIR_CROSS_SLOT 1 /* -CROSSSLOT request. */ +#define CLUSTER_REDIR_UNSTABLE 2 /* -TRYAGAIN redirection required */ +#define CLUSTER_REDIR_ASK 3 /* -ASK redirection required. */ +#define CLUSTER_REDIR_MOVED 4 /* -MOVED redirection required. */ +#define CLUSTER_REDIR_DOWN_STATE 5 /* -CLUSTERDOWN, global state. */ +#define CLUSTER_REDIR_DOWN_UNBOUND 6 /* -CLUSTERDOWN, unbound slot. */ + +struct clusterNode; + +/* clusterLink encapsulates everything needed to talk with a remote node. */ +typedef struct clusterLink { + mstime_t ctime; /* Link creation time */ + int fd; /* TCP socket file descriptor */ + sds sndbuf; /* Packet send buffer */ + sds rcvbuf; /* Packet reception buffer */ + struct clusterNode *node; /* Node related to this link if any, or NULL */ +} clusterLink; + +/* Cluster node flags and macros. */ +#define CLUSTER_NODE_MASTER 1 /* The node is a master */ +#define CLUSTER_NODE_SLAVE 2 /* The node is a slave */ +#define CLUSTER_NODE_PFAIL 4 /* Failure? Need acknowledge */ +#define CLUSTER_NODE_FAIL 8 /* The node is believed to be malfunctioning */ +#define CLUSTER_NODE_MYSELF 16 /* This node is myself */ +#define CLUSTER_NODE_HANDSHAKE 32 /* We have still to exchange the first ping */ +#define CLUSTER_NODE_NOADDR 64 /* We don't know the address of this node */ +#define CLUSTER_NODE_MEET 128 /* Send a MEET message to this node */ +#define CLUSTER_NODE_MIGRATE_TO 256 /* Master elegible for replica migration. */ +#define CLUSTER_NODE_NULL_NAME "\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000" + +#define nodeIsMaster(n) ((n)->flags & CLUSTER_NODE_MASTER) +#define nodeIsSlave(n) ((n)->flags & CLUSTER_NODE_SLAVE) +#define nodeInHandshake(n) ((n)->flags & CLUSTER_NODE_HANDSHAKE) +#define nodeHasAddr(n) (!((n)->flags & CLUSTER_NODE_NOADDR)) +#define nodeWithoutAddr(n) ((n)->flags & CLUSTER_NODE_NOADDR) +#define nodeTimedOut(n) ((n)->flags & CLUSTER_NODE_PFAIL) +#define nodeFailed(n) ((n)->flags & CLUSTER_NODE_FAIL) + +/* Reasons why a slave is not able to failover. */ +#define CLUSTER_CANT_FAILOVER_NONE 0 +#define CLUSTER_CANT_FAILOVER_DATA_AGE 1 +#define CLUSTER_CANT_FAILOVER_WAITING_DELAY 2 +#define CLUSTER_CANT_FAILOVER_EXPIRED 3 +#define CLUSTER_CANT_FAILOVER_WAITING_VOTES 4 +#define CLUSTER_CANT_FAILOVER_RELOG_PERIOD (60*5) /* seconds. */ + +/* clusterState todo_before_sleep flags. */ +#define CLUSTER_TODO_HANDLE_FAILOVER (1<<0) +#define CLUSTER_TODO_UPDATE_STATE (1<<1) +#define CLUSTER_TODO_SAVE_CONFIG (1<<2) +#define CLUSTER_TODO_FSYNC_CONFIG (1<<3) + +/* Message types. + * + * Note that the PING, PONG and MEET messages are actually the same exact + * kind of packet. PONG is the reply to ping, in the exact format as a PING, + * while MEET is a special PING that forces the receiver to add the sender + * as a node (if it is not already in the list). */ +#define CLUSTERMSG_TYPE_PING 0 /* Ping */ +#define CLUSTERMSG_TYPE_PONG 1 /* Pong (reply to Ping) */ +#define CLUSTERMSG_TYPE_MEET 2 /* Meet "let's join" message */ +#define CLUSTERMSG_TYPE_FAIL 3 /* Mark node xxx as failing */ +#define CLUSTERMSG_TYPE_PUBLISH 4 /* Pub/Sub Publish propagation */ +#define CLUSTERMSG_TYPE_FAILOVER_AUTH_REQUEST 5 /* May I failover? */ +#define CLUSTERMSG_TYPE_FAILOVER_AUTH_ACK 6 /* Yes, you have my vote */ +#define CLUSTERMSG_TYPE_UPDATE 7 /* Another node slots configuration */ +#define CLUSTERMSG_TYPE_MFSTART 8 /* Pause clients for manual failover */ +#define CLUSTERMSG_TYPE_COUNT 9 /* Total number of message types. */ + +/* This structure represent elements of node->fail_reports. */ +typedef struct clusterNodeFailReport { + struct clusterNode *node; /* Node reporting the failure condition. */ + mstime_t time; /* Time of the last report from this node. */ +} clusterNodeFailReport; + +typedef struct clusterNode { + mstime_t ctime; /* Node object creation time. */ + char name[CLUSTER_NAMELEN]; /* Node name, hex string, sha1-size */ + int flags; /* CLUSTER_NODE_... */ + uint64_t configEpoch; /* Last configEpoch observed for this node */ + unsigned char slots[CLUSTER_SLOTS/8]; /* slots handled by this node */ + int numslots; /* Number of slots handled by this node */ + int numslaves; /* Number of slave nodes, if this is a master */ + struct clusterNode **slaves; /* pointers to slave nodes */ + struct clusterNode *slaveof; /* pointer to the master node. Note that it + may be NULL even if the node is a slave + if we don't have the master node in our + tables. */ + mstime_t ping_sent; /* Unix time we sent latest ping */ + mstime_t pong_received; /* Unix time we received the pong */ + mstime_t fail_time; /* Unix time when FAIL flag was set */ + mstime_t voted_time; /* Last time we voted for a slave of this master */ + mstime_t repl_offset_time; /* Unix time we received offset for this node */ + mstime_t orphaned_time; /* Starting time of orphaned master condition */ + PORT_LONGLONG repl_offset; /* Last known repl offset for this node. */ + char ip[NET_IP_STR_LEN]; /* Latest known IP address of this node */ + int port; /* Latest known clients port of this node */ + int cport; /* Latest known cluster port of this node. */ + clusterLink *link; /* TCP/IP link with this node */ + list *fail_reports; /* List of nodes signaling this as failing */ +} clusterNode; + +typedef struct clusterState { + clusterNode *myself; /* This node */ + uint64_t currentEpoch; + int state; /* CLUSTER_OK, CLUSTER_FAIL, ... */ + int size; /* Num of master nodes with at least one slot */ + dict *nodes; /* Hash table of name -> clusterNode structures */ + dict *nodes_black_list; /* Nodes we don't re-add for a few seconds. */ + clusterNode *migrating_slots_to[CLUSTER_SLOTS]; + clusterNode *importing_slots_from[CLUSTER_SLOTS]; + clusterNode *slots[CLUSTER_SLOTS]; + uint64_t slots_keys_count[CLUSTER_SLOTS]; + rax *slots_to_keys; + /* The following fields are used to take the slave state on elections. */ + mstime_t failover_auth_time; /* Time of previous or next election. */ + int failover_auth_count; /* Number of votes received so far. */ + int failover_auth_sent; /* True if we already asked for votes. */ + int failover_auth_rank; /* This slave rank for current auth request. */ + uint64_t failover_auth_epoch; /* Epoch of the current election. */ + int cant_failover_reason; /* Why a slave is currently not able to + failover. See the CANT_FAILOVER_* macros. */ + /* Manual failover state in common. */ + mstime_t mf_end; /* Manual failover time limit (ms unixtime). + It is zero if there is no MF in progress. */ + /* Manual failover state of master. */ + clusterNode *mf_slave; /* Slave performing the manual failover. */ + /* Manual failover state of slave. */ + PORT_LONGLONG mf_master_offset; /* Master offset the slave needs to start MF + or zero if stil not received. */ + int mf_can_start; /* If non-zero signal that the manual failover + can start requesting masters vote. */ + /* The followign fields are used by masters to take state on elections. */ + uint64_t lastVoteEpoch; /* Epoch of the last vote granted. */ + int todo_before_sleep; /* Things to do in clusterBeforeSleep(). */ + /* Messages received and sent by type. */ + PORT_LONGLONG stats_bus_messages_sent[CLUSTERMSG_TYPE_COUNT]; + PORT_LONGLONG stats_bus_messages_received[CLUSTERMSG_TYPE_COUNT]; + PORT_LONGLONG stats_pfail_nodes; /* Number of nodes in PFAIL status, + excluding nodes without address. */ +} clusterState; + +/* Redis cluster messages header */ + +/* Initially we don't know our "name", but we'll find it once we connect + * to the first node, using the getsockname() function. Then we'll use this + * address for all the next messages. */ +typedef struct { + char nodename[CLUSTER_NAMELEN]; + uint32_t ping_sent; + uint32_t pong_received; + char ip[NET_IP_STR_LEN]; /* IP address last time it was seen */ + uint16_t port; /* base port last time it was seen */ + uint16_t cport; /* cluster port last time it was seen */ + uint16_t flags; /* node->flags copy */ + uint32_t notused1; +} clusterMsgDataGossip; + +typedef struct { + char nodename[CLUSTER_NAMELEN]; +} clusterMsgDataFail; + +typedef struct { + uint32_t channel_len; + uint32_t message_len; + /* We can't reclare bulk_data as bulk_data[] since this structure is + * nested. The 8 bytes are removed from the count during the message + * length computation. */ + unsigned char bulk_data[8]; +} clusterMsgDataPublish; + +typedef struct { + uint64_t configEpoch; /* Config epoch of the specified instance. */ + char nodename[CLUSTER_NAMELEN]; /* Name of the slots owner. */ + unsigned char slots[CLUSTER_SLOTS/8]; /* Slots bitmap. */ +} clusterMsgDataUpdate; + +union clusterMsgData { + /* PING, MEET and PONG */ + struct { + /* Array of N clusterMsgDataGossip structures */ + clusterMsgDataGossip gossip[1]; + } ping; + + /* FAIL */ + struct { + clusterMsgDataFail about; + } fail; + + /* PUBLISH */ + struct { + clusterMsgDataPublish msg; + } publish; + + /* UPDATE */ + struct { + clusterMsgDataUpdate nodecfg; + } update; +}; + +#define CLUSTER_PROTO_VER 1 /* Cluster bus protocol version. */ + +typedef struct { + char sig[4]; /* Siganture "RCmb" (Redis Cluster message bus). */ + uint32_t totlen; /* Total length of this message */ + uint16_t ver; /* Protocol version, currently set to 1. */ + uint16_t port; /* TCP base port number. */ + uint16_t type; /* Message type */ + uint16_t count; /* Only used for some kind of messages. */ + uint64_t currentEpoch; /* The epoch accordingly to the sending node. */ + uint64_t configEpoch; /* The config epoch if it's a master, or the last + epoch advertised by its master if it is a + slave. */ + uint64_t offset; /* Master replication offset if node is a master or + processed replication offset if node is a slave. */ + char sender[CLUSTER_NAMELEN]; /* Name of the sender node */ + unsigned char myslots[CLUSTER_SLOTS/8]; + char slaveof[CLUSTER_NAMELEN]; + char myip[NET_IP_STR_LEN]; /* Sender IP, if not all zeroed. */ + char notused1[34]; /* 34 bytes reserved for future usage. */ + uint16_t cport; /* Sender TCP cluster bus port */ + uint16_t flags; /* Sender node flags */ + unsigned char state; /* Cluster state from the POV of the sender */ + unsigned char mflags[3]; /* Message flags: CLUSTERMSG_FLAG[012]_... */ + union clusterMsgData data; +} clusterMsg; + +#define CLUSTERMSG_MIN_LEN (sizeof(clusterMsg)-sizeof(union clusterMsgData)) + +/* Message flags better specify the packet content or are used to + * provide some information about the node state. */ +#define CLUSTERMSG_FLAG0_PAUSED (1<<0) /* Master paused for manual failover. */ +#define CLUSTERMSG_FLAG0_FORCEACK (1<<1) /* Give ACK to AUTH_REQUEST even if + master is up. */ + +/* ---------------------- API exported outside cluster.c -------------------- */ +clusterNode *getNodeByQuery(client *c, struct redisCommand *cmd, robj **argv, int argc, int *hashslot, int *ask); +int clusterRedirectBlockedClientIfNeeded(client *c); +void clusterRedirectClient(client *c, clusterNode *n, int hashslot, int error_code); + +#endif /* __CLUSTER_H */ diff --git a/src/config.c b/src/config.c index 6ed57992..cb273649 100644 --- a/src/config.c +++ b/src/config.c @@ -372,15 +372,19 @@ void loadServerConfigFromString(char *config) { err = "maxmemory-samples must be 1 or greater"; goto loaderr; } + } else if ((!strcasecmp(argv[0],"proto-max-bulk-len")) && argc == 2) { + server.proto_max_bulk_len = memtoll(argv[1],NULL); + } else if ((!strcasecmp(argv[0],"client-query-buffer-limit")) && argc == 2) { + server.client_max_querybuf_len = memtoll(argv[1],NULL); } else if (!strcasecmp(argv[0],"lfu-log-factor") && argc == 2) { server.lfu_log_factor = atoi(argv[1]); - if (server.maxmemory_samples < 0) { + if (server.lfu_log_factor < 0) { err = "lfu-log-factor must be 0 or greater"; goto loaderr; } } else if (!strcasecmp(argv[0],"lfu-decay-time") && argc == 2) { server.lfu_decay_time = atoi(argv[1]); - if (server.maxmemory_samples < 1) { + if (server.lfu_decay_time < 0) { err = "lfu-decay-time must be 0 or greater"; goto loaderr; } @@ -675,6 +679,14 @@ void loadServerConfigFromString(char *config) { err = "cluster slave validity factor must be zero or positive"; goto loaderr; } + } else if (!strcasecmp(argv[0],"cluster-slave-no-failover") && + argc == 2) + { + server.cluster_slave_no_failover = yesnotoi(argv[1]); + if (server.cluster_slave_no_failover == -1) { + err = "argument must be 'yes' or 'no'"; + goto loaderr; + } } else if (!strcasecmp(argv[0],"lua-time-limit") && argc == 2) { server.lua_time_limit = strtoll(argv[1],NULL,10); } else if (!strcasecmp(argv[0],"slowlog-log-slower-than") && @@ -1082,6 +1094,8 @@ void configSetCommand(client *c) { "repl-diskless-sync",server.repl_diskless_sync) { } config_set_bool_field( "cluster-require-full-coverage",server.cluster_require_full_coverage) { + } config_set_bool_field( + "cluster-slave-no-failover",server.cluster_slave_no_failover) { } config_set_bool_field( "aof-rewrite-incremental-fsync",server.aof_rewrite_incremental_fsync) { } config_set_bool_field( @@ -1312,7 +1326,11 @@ void configGetCommand(client *c) { /* Numerical values */ config_get_numerical_field("maxmemory",server.maxmemory); + config_get_numerical_field("proto-max-bulk-len",server.proto_max_bulk_len); + config_get_numerical_field("client-query-buffer-limit",server.client_max_querybuf_len); config_get_numerical_field("maxmemory-samples",server.maxmemory_samples); + config_get_numerical_field("lfu-log-factor",server.lfu_log_factor); + config_get_numerical_field("lfu-decay-time",server.lfu_decay_time); config_get_numerical_field("timeout",server.maxidletime); config_get_numerical_field("active-defrag-threshold-lower",server.active_defrag_threshold_lower); config_get_numerical_field("active-defrag-threshold-upper",server.active_defrag_threshold_upper); @@ -1371,6 +1389,8 @@ void configGetCommand(client *c) { /* Bool (yes/no) values */ config_get_bool_field("cluster-require-full-coverage", server.cluster_require_full_coverage); + config_get_bool_field("cluster-slave-no-failover", + server.cluster_slave_no_failover); config_get_bool_field("no-appendfsync-on-rewrite", server.aof_no_fsync_on_rewrite); config_get_bool_field("slave-serve-stale-data", @@ -2088,8 +2108,12 @@ int rewriteConfig(char *path) { rewriteConfigStringOption(state,"requirepass",server.requirepass,NULL); rewriteConfigNumericalOption(state,"maxclients",server.maxclients,CONFIG_DEFAULT_MAX_CLIENTS); rewriteConfigBytesOption(state,"maxmemory",server.maxmemory,CONFIG_DEFAULT_MAXMEMORY); + rewriteConfigBytesOption(state,"proto-max-bulk-len",server.proto_max_bulk_len,CONFIG_DEFAULT_PROTO_MAX_BULK_LEN); + rewriteConfigBytesOption(state,"client-query-buffer-limit",server.client_max_querybuf_len,PROTO_MAX_QUERYBUF_LEN); rewriteConfigEnumOption(state,"maxmemory-policy",server.maxmemory_policy,maxmemory_policy_enum,CONFIG_DEFAULT_MAXMEMORY_POLICY); rewriteConfigNumericalOption(state,"maxmemory-samples",server.maxmemory_samples,CONFIG_DEFAULT_MAXMEMORY_SAMPLES); + rewriteConfigNumericalOption(state,"lfu-log-factor",server.lfu_log_factor,CONFIG_DEFAULT_LFU_LOG_FACTOR); + rewriteConfigNumericalOption(state,"lfu-decay-time",server.lfu_decay_time,CONFIG_DEFAULT_LFU_DECAY_TIME); rewriteConfigNumericalOption(state,"active-defrag-threshold-lower",server.active_defrag_threshold_lower,CONFIG_DEFAULT_DEFRAG_THRESHOLD_LOWER); rewriteConfigNumericalOption(state,"active-defrag-threshold-upper",server.active_defrag_threshold_upper,CONFIG_DEFAULT_DEFRAG_THRESHOLD_UPPER); rewriteConfigBytesOption(state,"active-defrag-ignore-bytes",server.active_defrag_ignore_bytes,CONFIG_DEFAULT_DEFRAG_IGNORE_BYTES); @@ -2105,6 +2129,7 @@ int rewriteConfig(char *path) { rewriteConfigYesNoOption(state,"cluster-enabled",server.cluster_enabled,0); rewriteConfigStringOption(state,"cluster-config-file",server.cluster_configfile,CONFIG_DEFAULT_CLUSTER_CONFIG_FILE); rewriteConfigYesNoOption(state,"cluster-require-full-coverage",server.cluster_require_full_coverage,CLUSTER_DEFAULT_REQUIRE_FULL_COVERAGE); + rewriteConfigYesNoOption(state,"cluster-slave-no-failover",server.cluster_slave_no_failover,CLUSTER_DEFAULT_SLAVE_NO_FAILOVER); rewriteConfigNumericalOption(state,"cluster-node-timeout",server.cluster_node_timeout,CLUSTER_DEFAULT_NODE_TIMEOUT); rewriteConfigNumericalOption(state,"cluster-migration-barrier",server.cluster_migration_barrier,CLUSTER_DEFAULT_MIGRATION_BARRIER); rewriteConfigNumericalOption(state,"cluster-slave-validity-factor",server.cluster_slave_validity_factor,CLUSTER_DEFAULT_SLAVE_VALIDITY); diff --git a/src/config.c.bak b/src/config.c.bak new file mode 100644 index 00000000..6ed57992 --- /dev/null +++ b/src/config.c.bak @@ -0,0 +1,2199 @@ +/* Configuration file parsing and CONFIG GET/SET commands implementation. + * + * Copyright (c) 2009-2012, Salvatore Sanfilippo + * All rights reserved. + * + * Redistribution and use in source and binary forms, with or without + * modification, are permitted provided that the following conditions are met: + * + * * Redistributions of source code must retain the above copyright notice, + * this list of conditions and the following disclaimer. + * * Redistributions in binary form must reproduce the above copyright + * notice, this list of conditions and the following disclaimer in the + * documentation and/or other materials provided with the distribution. + * * Neither the name of Redis nor the names of its contributors may be used + * to endorse or promote products derived from this software without + * specific prior written permission. + * + * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" + * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE + * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE + * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE + * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR + * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF + * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS + * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN + * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) + * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE + * POSSIBILITY OF SUCH DAMAGE. + */ + +#ifdef _WIN32 +#include "Win32_Interop/win32_types.h" +#include "Win32_Interop/Win32_EventLog.h" +#include "Win32_Interop/Win32_Error.h" +#include +#endif + +#include "server.h" +#include "cluster.h" + +#include +#include + +/*----------------------------------------------------------------------------- + * Config file name-value maps. + *----------------------------------------------------------------------------*/ + +typedef struct configEnum { + const char *name; + const int val; +} configEnum; + +configEnum maxmemory_policy_enum[] = { + {"volatile-lru", MAXMEMORY_VOLATILE_LRU}, + {"volatile-lfu", MAXMEMORY_VOLATILE_LFU}, + {"volatile-random",MAXMEMORY_VOLATILE_RANDOM}, + {"volatile-ttl",MAXMEMORY_VOLATILE_TTL}, + {"allkeys-lru",MAXMEMORY_ALLKEYS_LRU}, + {"allkeys-lfu",MAXMEMORY_ALLKEYS_LFU}, + {"allkeys-random",MAXMEMORY_ALLKEYS_RANDOM}, + {"noeviction",MAXMEMORY_NO_EVICTION}, + {NULL, 0} +}; + +#ifndef _WIN32 +configEnum syslog_facility_enum[] = { + {"user", LOG_USER}, + {"local0", LOG_LOCAL0}, + {"local1", LOG_LOCAL1}, + {"local2", LOG_LOCAL2}, + {"local3", LOG_LOCAL3}, + {"local4", LOG_LOCAL4}, + {"local5", LOG_LOCAL5}, + {"local6", LOG_LOCAL6}, + {"local7", LOG_LOCAL7}, + {NULL, 0} +}; +#endif + +configEnum loglevel_enum[] = { + {"debug", LL_DEBUG}, + {"verbose", LL_VERBOSE}, + {"notice", LL_NOTICE}, + {"warning", LL_WARNING}, + {NULL,0} +}; + +configEnum supervised_mode_enum[] = { + {"upstart", SUPERVISED_UPSTART}, + {"systemd", SUPERVISED_SYSTEMD}, + {"auto", SUPERVISED_AUTODETECT}, + {"no", SUPERVISED_NONE}, + {NULL, 0} +}; + +configEnum aof_fsync_enum[] = { + {"everysec", AOF_FSYNC_EVERYSEC}, + {"always", AOF_FSYNC_ALWAYS}, + {"no", AOF_FSYNC_NO}, + {NULL, 0} +}; + +/* Output buffer limits presets. */ +clientBufferLimitsConfig clientBufferLimitsDefaults[CLIENT_TYPE_OBUF_COUNT] = { + {0, 0, 0}, /* normal */ + {1024*1024*256, 1024*1024*64, 60}, /* slave */ + {1024*1024*32, 1024*1024*8, 60} /* pubsub */ +}; + +/*----------------------------------------------------------------------------- + * Enum access functions + *----------------------------------------------------------------------------*/ + +/* Get enum value from name. If there is no match INT_MIN is returned. */ +int configEnumGetValue(configEnum *ce, char *name) { + while(ce->name != NULL) { + if (!strcasecmp(ce->name,name)) return ce->val; + ce++; + } + return INT_MIN; +} + +/* Get enum name from value. If no match is found NULL is returned. */ +const char *configEnumGetName(configEnum *ce, int val) { + while(ce->name != NULL) { + if (ce->val == val) return ce->name; + ce++; + } + return NULL; +} + +/* Wrapper for configEnumGetName() returning "unknown" insetad of NULL if + * there is no match. */ +const char *configEnumGetNameOrUnknown(configEnum *ce, int val) { + const char *name = configEnumGetName(ce,val); + return name ? name : "unknown"; +} + +/* Used for INFO generation. */ +const char *evictPolicyToString(void) { + return configEnumGetNameOrUnknown(maxmemory_policy_enum,server.maxmemory_policy); +} + +/*----------------------------------------------------------------------------- + * Config file parsing + *----------------------------------------------------------------------------*/ + +int yesnotoi(char *s) { + if (!strcasecmp(s,"yes")) return 1; + else if (!strcasecmp(s,"no")) return 0; + else return -1; +} + +void appendServerSaveParams(time_t seconds, int changes) { + server.saveparams = zrealloc(server.saveparams,sizeof(struct saveparam)*(server.saveparamslen+1)); + server.saveparams[server.saveparamslen].seconds = seconds; + server.saveparams[server.saveparamslen].changes = changes; + server.saveparamslen++; +} + +void resetServerSaveParams(void) { + zfree(server.saveparams); + server.saveparams = NULL; + server.saveparamslen = 0; +} + +void queueLoadModule(sds path, sds *argv, int argc) { + int i; + struct moduleLoadQueueEntry *loadmod; + + loadmod = zmalloc(sizeof(struct moduleLoadQueueEntry)); + loadmod->argv = zmalloc(sizeof(robj*)*argc); + loadmod->path = sdsnew(path); + loadmod->argc = argc; + for (i = 0; i < argc; i++) { + loadmod->argv[i] = createRawStringObject(argv[i],sdslen(argv[i])); + } + listAddNodeTail(server.loadmodule_queue,loadmod); +} + +void loadServerConfigFromString(char *config) { + char *err = NULL; + int linenum = 0, totlines, i; + int slaveof_linenum = 0; + sds *lines; + + lines = sdssplitlen(config,(int)strlen(config),"\n",1,&totlines); WIN_PORT_FIX /* cast (int) */ + + for (i = 0; i < totlines; i++) { + sds *argv; + int argc; + + linenum = i+1; + lines[i] = sdstrim(lines[i]," \t\r\n"); + + /* Skip comments and blank lines */ + if (lines[i][0] == '#' || lines[i][0] == '\0') continue; + + /* Split into arguments */ + argv = sdssplitargs(lines[i],&argc); + if (argv == NULL) { + err = "Unbalanced quotes in configuration line"; + goto loaderr; + } + + /* Skip this line if the resulting command vector is empty. */ + if (argc == 0) { + sdsfreesplitres(argv,argc); + continue; + } + sdstolower(argv[0]); + + /* Execute config directives */ + if (!strcasecmp(argv[0],"timeout") && argc == 2) { + server.maxidletime = atoi(argv[1]); + if (server.maxidletime < 0) { + err = "Invalid timeout value"; goto loaderr; + } + } else if (!strcasecmp(argv[0],"tcp-keepalive") && argc == 2) { + server.tcpkeepalive = atoi(argv[1]); + if (server.tcpkeepalive < 0) { + err = "Invalid tcp-keepalive value"; goto loaderr; + } + } else if (!strcasecmp(argv[0],"protected-mode") && argc == 2) { + if ((server.protected_mode = yesnotoi(argv[1])) == -1) { + err = "argument must be 'yes' or 'no'"; goto loaderr; + } + } else if (!strcasecmp(argv[0],"port") && argc == 2) { + server.port = atoi(argv[1]); + if (server.port < 0 || server.port > 65535) { + err = "Invalid port"; goto loaderr; + } + } else if (!strcasecmp(argv[0],"tcp-backlog") && argc == 2) { + server.tcp_backlog = atoi(argv[1]); + if (server.tcp_backlog < 0) { + err = "Invalid backlog value"; goto loaderr; + } + } else if (!strcasecmp(argv[0],"bind") && argc >= 2) { + int j, addresses = argc-1; + + if (addresses > CONFIG_BINDADDR_MAX) { + err = "Too many bind addresses specified"; goto loaderr; + } + for (j = 0; j < addresses; j++) + server.bindaddr[j] = zstrdup(argv[j+1]); + server.bindaddr_count = addresses; + } else if (!strcasecmp(argv[0],"unixsocket") && argc == 2) { + server.unixsocket = zstrdup(argv[1]); + } else if (!strcasecmp(argv[0],"unixsocketperm") && argc == 2) { + errno = 0; + server.unixsocketperm = (mode_t)strtol(argv[1], NULL, 8); + if (errno || server.unixsocketperm > 0777) { + err = "Invalid socket file permissions"; goto loaderr; + } + } else if (!strcasecmp(argv[0],"save")) { + if (argc == 3) { + int seconds = atoi(argv[1]); + int changes = atoi(argv[2]); + if (seconds < 1 || changes < 0) { + err = "Invalid save parameters"; goto loaderr; + } + appendServerSaveParams(seconds,changes); + } else if (argc == 2 && !strcasecmp(argv[1],"")) { + resetServerSaveParams(); + } + } else if (!strcasecmp(argv[0],"dir") && argc == 2) { + if (chdir(argv[1]) == -1) { + serverLog(LL_WARNING,"Can't chdir to '%s': %s", + argv[1], IF_WIN32(wsa_strerror(errno), strerror(errno))); + exit(1); + } + } else if (!strcasecmp(argv[0],"loglevel") && argc == 2) { + server.verbosity = configEnumGetValue(loglevel_enum,argv[1]); + if (server.verbosity == INT_MIN) { + err = "Invalid log level. " + "Must be one of debug, verbose, notice, warning"; + goto loaderr; + } +#ifdef _WIN32 + setLogVerbosityLevel(server.verbosity); +#endif + } else if (!strcasecmp(argv[0],"logfile") && argc == 2) { + FILE *logfp; + + zfree(server.logfile); +#ifdef _WIN32 + int length = (int)sdslen(argv[1]); + if ((argv[1][0] == '\'' && argv[1][length-1] == '\'') || + (argv[1][0] == '\"' && argv[1][length-1] == '\"')) { + if (length == 2) { + server.logfile = zstrdup("\0"); + } else { + size_t l = length - 2 + 1; + char *p = zmalloc(l); + memcpy(p, argv[1]+1, l); + server.logfile = p; + } + } else { + server.logfile = zstrdup(argv[1]); + } +#else + server.logfile = zstrdup(argv[1]); +#endif + if (server.logfile[0] != '\0') { + /* Test if we are able to open the file. The server will not + * be able to abort just for this problem later... */ + logfp = fopen(server.logfile,"a"); + if (logfp == NULL) { + err = sdscatprintf(sdsempty(), + "Can't open the log file: %s", IF_WIN32(wsa_strerror(errno), strerror(errno))); + goto loaderr; +#ifdef _WIN32 + } else { + setLogFile(server.logfile); +#endif + } + + fclose(logfp); + } + } else if (!strcasecmp(argv[0],"always-show-logo") && argc == 2) { + if ((server.always_show_logo = yesnotoi(argv[1])) == -1) { + err = "argument must be 'yes' or 'no'"; goto loaderr; + } + } else if (!strcasecmp(argv[0],"syslog-enabled") && argc == 2) { + if ((server.syslog_enabled = yesnotoi(argv[1])) == -1) { + err = "argument must be 'yes' or 'no'"; goto loaderr; + } +#ifdef _WIN32 + setSyslogEnabled(server.syslog_enabled); +#endif + } else if (!strcasecmp(argv[0],"syslog-ident") && argc == 2) { + if (server.syslog_ident) zfree(server.syslog_ident); + server.syslog_ident = zstrdup(argv[1]); +#ifdef _WIN32 + setSyslogIdent(server.syslog_ident); +#endif + } else if (!strcasecmp(argv[0],"syslog-facility") && argc == 2) { +#ifdef _WIN32 + // Skip error - just ignore syslog-facility +#else + server.syslog_facility = + configEnumGetValue(syslog_facility_enum,argv[1]); + if (server.syslog_facility == INT_MIN) { + err = "Invalid log facility. Must be one of USER or between LOCAL0-LOCAL7"; + goto loaderr; + } +#endif + } else if (!strcasecmp(argv[0],"databases") && argc == 2) { + server.dbnum = atoi(argv[1]); + if (server.dbnum < 1) { + err = "Invalid number of databases"; goto loaderr; + } + } else if (!strcasecmp(argv[0],"include") && argc == 2) { + loadServerConfig(argv[1],NULL); + } else if (!strcasecmp(argv[0],"maxclients") && argc == 2) { + server.maxclients = atoi(argv[1]); + if (server.maxclients < 1) { + err = "Invalid max clients limit"; goto loaderr; + } + } else if (!strcasecmp(argv[0],"maxmemory") && argc == 2) { + server.maxmemory = memtoll(argv[1],NULL); + } else if (!strcasecmp(argv[0],"maxmemory-policy") && argc == 2) { + server.maxmemory_policy = + configEnumGetValue(maxmemory_policy_enum,argv[1]); + if (server.maxmemory_policy == INT_MIN) { + err = "Invalid maxmemory policy"; + goto loaderr; + } + } else if (!strcasecmp(argv[0],"maxmemory-samples") && argc == 2) { + server.maxmemory_samples = atoi(argv[1]); + if (server.maxmemory_samples <= 0) { + err = "maxmemory-samples must be 1 or greater"; + goto loaderr; + } + } else if (!strcasecmp(argv[0],"lfu-log-factor") && argc == 2) { + server.lfu_log_factor = atoi(argv[1]); + if (server.maxmemory_samples < 0) { + err = "lfu-log-factor must be 0 or greater"; + goto loaderr; + } + } else if (!strcasecmp(argv[0],"lfu-decay-time") && argc == 2) { + server.lfu_decay_time = atoi(argv[1]); + if (server.maxmemory_samples < 1) { + err = "lfu-decay-time must be 0 or greater"; + goto loaderr; + } + } else if (!strcasecmp(argv[0],"slaveof") && argc == 3) { + slaveof_linenum = linenum; + server.masterhost = sdsnew(argv[1]); + server.masterport = atoi(argv[2]); + server.repl_state = REPL_STATE_CONNECT; + } else if (!strcasecmp(argv[0],"repl-ping-slave-period") && argc == 2) { + server.repl_ping_slave_period = atoi(argv[1]); + if (server.repl_ping_slave_period <= 0) { + err = "repl-ping-slave-period must be 1 or greater"; + goto loaderr; + } + } else if (!strcasecmp(argv[0],"repl-timeout") && argc == 2) { + server.repl_timeout = atoi(argv[1]); + if (server.repl_timeout <= 0) { + err = "repl-timeout must be 1 or greater"; + goto loaderr; + } + } else if (!strcasecmp(argv[0],"repl-disable-tcp-nodelay") && argc==2) { + if ((server.repl_disable_tcp_nodelay = yesnotoi(argv[1])) == -1) { + err = "argument must be 'yes' or 'no'"; goto loaderr; + } + } else if (!strcasecmp(argv[0],"repl-diskless-sync") && argc==2) { + if ((server.repl_diskless_sync = yesnotoi(argv[1])) == -1) { + err = "argument must be 'yes' or 'no'"; goto loaderr; + } + } else if (!strcasecmp(argv[0],"repl-diskless-sync-delay") && argc==2) { + server.repl_diskless_sync_delay = atoi(argv[1]); + if (server.repl_diskless_sync_delay < 0) { + err = "repl-diskless-sync-delay can't be negative"; + goto loaderr; + } + } else if (!strcasecmp(argv[0],"repl-backlog-size") && argc == 2) { + PORT_LONGLONG size = memtoll(argv[1],NULL); + if (size <= 0) { + err = "repl-backlog-size must be 1 or greater."; + goto loaderr; + } + resizeReplicationBacklog(size); + } else if (!strcasecmp(argv[0],"repl-backlog-ttl") && argc == 2) { + server.repl_backlog_time_limit = atoi(argv[1]); + if (server.repl_backlog_time_limit < 0) { + err = "repl-backlog-ttl can't be negative "; + goto loaderr; + } + } else if (!strcasecmp(argv[0],"masterauth") && argc == 2) { + zfree(server.masterauth); + server.masterauth = zstrdup(argv[1]); + } else if (!strcasecmp(argv[0],"slave-serve-stale-data") && argc == 2) { + if ((server.repl_serve_stale_data = yesnotoi(argv[1])) == -1) { + err = "argument must be 'yes' or 'no'"; goto loaderr; + } + } else if (!strcasecmp(argv[0],"slave-read-only") && argc == 2) { + if ((server.repl_slave_ro = yesnotoi(argv[1])) == -1) { + err = "argument must be 'yes' or 'no'"; goto loaderr; + } + } else if (!strcasecmp(argv[0],"rdbcompression") && argc == 2) { + if ((server.rdb_compression = yesnotoi(argv[1])) == -1) { + err = "argument must be 'yes' or 'no'"; goto loaderr; + } + } else if (!strcasecmp(argv[0],"rdbchecksum") && argc == 2) { + if ((server.rdb_checksum = yesnotoi(argv[1])) == -1) { + err = "argument must be 'yes' or 'no'"; goto loaderr; + } + } else if (!strcasecmp(argv[0],"activerehashing") && argc == 2) { + if ((server.activerehashing = yesnotoi(argv[1])) == -1) { + err = "argument must be 'yes' or 'no'"; goto loaderr; + } + } else if (!strcasecmp(argv[0],"lazyfree-lazy-eviction") && argc == 2) { + if ((server.lazyfree_lazy_eviction = yesnotoi(argv[1])) == -1) { + err = "argument must be 'yes' or 'no'"; goto loaderr; + } + } else if (!strcasecmp(argv[0],"lazyfree-lazy-expire") && argc == 2) { + if ((server.lazyfree_lazy_expire = yesnotoi(argv[1])) == -1) { + err = "argument must be 'yes' or 'no'"; goto loaderr; + } + } else if (!strcasecmp(argv[0],"lazyfree-lazy-server-del") && argc == 2){ + if ((server.lazyfree_lazy_server_del = yesnotoi(argv[1])) == -1) { + err = "argument must be 'yes' or 'no'"; goto loaderr; + } + } else if (!strcasecmp(argv[0],"slave-lazy-flush") && argc == 2) { + if ((server.repl_slave_lazy_flush = yesnotoi(argv[1])) == -1) { + err = "argument must be 'yes' or 'no'"; goto loaderr; + } + } else if (!strcasecmp(argv[0],"activedefrag") && argc == 2) { + if ((server.active_defrag_enabled = yesnotoi(argv[1])) == -1) { + err = "argument must be 'yes' or 'no'"; goto loaderr; + } + } else if (!strcasecmp(argv[0],"daemonize") && argc == 2) { + if ((server.daemonize = yesnotoi(argv[1])) == -1) { + err = "argument must be 'yes' or 'no'"; goto loaderr; + } + } else if (!strcasecmp(argv[0],"hz") && argc == 2) { + server.hz = atoi(argv[1]); + if (server.hz < CONFIG_MIN_HZ) server.hz = CONFIG_MIN_HZ; + if (server.hz > CONFIG_MAX_HZ) server.hz = CONFIG_MAX_HZ; + } else if (!strcasecmp(argv[0],"appendonly") && argc == 2) { + int yes; + + if ((yes = yesnotoi(argv[1])) == -1) { + err = "argument must be 'yes' or 'no'"; goto loaderr; + } + server.aof_state = yes ? AOF_ON : AOF_OFF; + } else if (!strcasecmp(argv[0],"appendfilename") && argc == 2) { + if (!pathIsBaseName(argv[1])) { + err = "appendfilename can't be a path, just a filename"; + goto loaderr; + } + zfree(server.aof_filename); + server.aof_filename = zstrdup(argv[1]); + } else if (!strcasecmp(argv[0],"no-appendfsync-on-rewrite") + && argc == 2) { + if ((server.aof_no_fsync_on_rewrite= yesnotoi(argv[1])) == -1) { + err = "argument must be 'yes' or 'no'"; goto loaderr; + } + } else if (!strcasecmp(argv[0],"appendfsync") && argc == 2) { + server.aof_fsync = configEnumGetValue(aof_fsync_enum,argv[1]); + if (server.aof_fsync == INT_MIN) { + err = "argument must be 'no', 'always' or 'everysec'"; + goto loaderr; + } + } else if (!strcasecmp(argv[0],"auto-aof-rewrite-percentage") && + argc == 2) + { + server.aof_rewrite_perc = atoi(argv[1]); + if (server.aof_rewrite_perc < 0) { + err = "Invalid negative percentage for AOF auto rewrite"; + goto loaderr; + } + } else if (!strcasecmp(argv[0],"auto-aof-rewrite-min-size") && + argc == 2) + { + server.aof_rewrite_min_size = memtoll(argv[1],NULL); + } else if (!strcasecmp(argv[0],"aof-rewrite-incremental-fsync") && + argc == 2) + { + if ((server.aof_rewrite_incremental_fsync = + yesnotoi(argv[1])) == -1) { + err = "argument must be 'yes' or 'no'"; goto loaderr; + } + } else if (!strcasecmp(argv[0],"aof-load-truncated") && argc == 2) { + if ((server.aof_load_truncated = yesnotoi(argv[1])) == -1) { + err = "argument must be 'yes' or 'no'"; goto loaderr; + } + } else if (!strcasecmp(argv[0],"aof-use-rdb-preamble") && argc == 2) { + if ((server.aof_use_rdb_preamble = yesnotoi(argv[1])) == -1) { + err = "argument must be 'yes' or 'no'"; goto loaderr; + } + } else if (!strcasecmp(argv[0],"requirepass") && argc == 2) { + if (strlen(argv[1]) > CONFIG_AUTHPASS_MAX_LEN) { + err = "Password is longer than CONFIG_AUTHPASS_MAX_LEN"; + goto loaderr; + } + server.requirepass = zstrdup(argv[1]); + } else if (!strcasecmp(argv[0],"pidfile") && argc == 2) { + zfree(server.pidfile); + server.pidfile = zstrdup(argv[1]); + } else if (!strcasecmp(argv[0],"dbfilename") && argc == 2) { + if (!pathIsBaseName(argv[1])) { + err = "dbfilename can't be a path, just a filename"; + goto loaderr; + } + zfree(server.rdb_filename); + server.rdb_filename = zstrdup(argv[1]); + } else if (!strcasecmp(argv[0],"active-defrag-threshold-lower") && argc == 2) { + server.active_defrag_threshold_lower = atoi(argv[1]); + if (server.active_defrag_threshold_lower < 0) { + err = "active-defrag-threshold-lower must be 0 or greater"; + goto loaderr; + } + } else if (!strcasecmp(argv[0],"active-defrag-threshold-upper") && argc == 2) { + server.active_defrag_threshold_upper = atoi(argv[1]); + if (server.active_defrag_threshold_upper < 0) { + err = "active-defrag-threshold-upper must be 0 or greater"; + goto loaderr; + } + } else if (!strcasecmp(argv[0],"active-defrag-ignore-bytes") && argc == 2) { + server.active_defrag_ignore_bytes = memtoll(argv[1], NULL); + if (server.active_defrag_ignore_bytes <= 0) { + err = "active-defrag-ignore-bytes must above 0"; + goto loaderr; + } + } else if (!strcasecmp(argv[0],"active-defrag-cycle-min") && argc == 2) { + server.active_defrag_cycle_min = atoi(argv[1]); + if (server.active_defrag_cycle_min < 1 || server.active_defrag_cycle_min > 99) { + err = "active-defrag-cycle-min must be between 1 and 99"; + goto loaderr; + } + } else if (!strcasecmp(argv[0],"active-defrag-cycle-max") && argc == 2) { + server.active_defrag_cycle_max = atoi(argv[1]); + if (server.active_defrag_cycle_max < 1 || server.active_defrag_cycle_max > 99) { + err = "active-defrag-cycle-max must be between 1 and 99"; + goto loaderr; + } + } else if (!strcasecmp(argv[0],"hash-max-ziplist-entries") && argc == 2) { + server.hash_max_ziplist_entries = memtoll(argv[1], NULL); + } else if (!strcasecmp(argv[0],"hash-max-ziplist-value") && argc == 2) { + server.hash_max_ziplist_value = memtoll(argv[1], NULL); + } else if (!strcasecmp(argv[0],"list-max-ziplist-entries") && argc == 2){ + /* DEAD OPTION */ + } else if (!strcasecmp(argv[0],"list-max-ziplist-value") && argc == 2) { + /* DEAD OPTION */ + } else if (!strcasecmp(argv[0],"list-max-ziplist-size") && argc == 2) { + server.list_max_ziplist_size = atoi(argv[1]); + } else if (!strcasecmp(argv[0],"list-compress-depth") && argc == 2) { + server.list_compress_depth = atoi(argv[1]); + } else if (!strcasecmp(argv[0],"set-max-intset-entries") && argc == 2) { + server.set_max_intset_entries = memtoll(argv[1], NULL); + } else if (!strcasecmp(argv[0],"zset-max-ziplist-entries") && argc == 2) { + server.zset_max_ziplist_entries = memtoll(argv[1], NULL); + } else if (!strcasecmp(argv[0],"zset-max-ziplist-value") && argc == 2) { + server.zset_max_ziplist_value = memtoll(argv[1], NULL); + } else if (!strcasecmp(argv[0],"hll-sparse-max-bytes") && argc == 2) { + server.hll_sparse_max_bytes = memtoll(argv[1], NULL); + } else if (!strcasecmp(argv[0],"rename-command") && argc == 3) { + struct redisCommand *cmd = lookupCommand(argv[1]); + int retval; + + if (!cmd) { + err = "No such command in rename-command"; + goto loaderr; + } + + /* If the target command name is the empty string we just + * remove it from the command table. */ + retval = dictDelete(server.commands, argv[1]); + serverAssert(retval == DICT_OK); + + /* Otherwise we re-add the command under a different name. */ + if (sdslen(argv[2]) != 0) { + sds copy = sdsdup(argv[2]); + + retval = dictAdd(server.commands, copy, cmd); + if (retval != DICT_OK) { + sdsfree(copy); + err = "Target command name already exists"; goto loaderr; + } + } + } else if (!strcasecmp(argv[0],"cluster-enabled") && argc == 2) { + if ((server.cluster_enabled = yesnotoi(argv[1])) == -1) { + err = "argument must be 'yes' or 'no'"; goto loaderr; + } + } else if (!strcasecmp(argv[0],"cluster-config-file") && argc == 2) { + zfree(server.cluster_configfile); + server.cluster_configfile = zstrdup(argv[1]); + } else if (!strcasecmp(argv[0],"cluster-announce-ip") && argc == 2) { + zfree(server.cluster_announce_ip); + server.cluster_announce_ip = zstrdup(argv[1]); + } else if (!strcasecmp(argv[0],"cluster-announce-port") && argc == 2) { + server.cluster_announce_port = atoi(argv[1]); + if (server.cluster_announce_port < 0 || + server.cluster_announce_port > 65535) + { + err = "Invalid port"; goto loaderr; + } + } else if (!strcasecmp(argv[0],"cluster-announce-bus-port") && + argc == 2) + { + server.cluster_announce_bus_port = atoi(argv[1]); + if (server.cluster_announce_bus_port < 0 || + server.cluster_announce_bus_port > 65535) + { + err = "Invalid port"; goto loaderr; + } + } else if (!strcasecmp(argv[0],"cluster-require-full-coverage") && + argc == 2) + { + if ((server.cluster_require_full_coverage = yesnotoi(argv[1])) == -1) + { + err = "argument must be 'yes' or 'no'"; goto loaderr; + } + } else if (!strcasecmp(argv[0],"cluster-node-timeout") && argc == 2) { + server.cluster_node_timeout = strtoll(argv[1],NULL,10); + if (server.cluster_node_timeout <= 0) { + err = "cluster node timeout must be 1 or greater"; goto loaderr; + } + } else if (!strcasecmp(argv[0],"cluster-migration-barrier") + && argc == 2) + { + server.cluster_migration_barrier = atoi(argv[1]); + if (server.cluster_migration_barrier < 0) { + err = "cluster migration barrier must zero or positive"; + goto loaderr; + } + } else if (!strcasecmp(argv[0],"cluster-slave-validity-factor") + && argc == 2) + { + server.cluster_slave_validity_factor = atoi(argv[1]); + if (server.cluster_slave_validity_factor < 0) { + err = "cluster slave validity factor must be zero or positive"; + goto loaderr; + } + } else if (!strcasecmp(argv[0],"lua-time-limit") && argc == 2) { + server.lua_time_limit = strtoll(argv[1],NULL,10); + } else if (!strcasecmp(argv[0],"slowlog-log-slower-than") && + argc == 2) + { + server.slowlog_log_slower_than = strtoll(argv[1],NULL,10); + } else if (!strcasecmp(argv[0],"latency-monitor-threshold") && + argc == 2) + { + server.latency_monitor_threshold = strtoll(argv[1],NULL,10); + if (server.latency_monitor_threshold < 0) { + err = "The latency threshold can't be negative"; + goto loaderr; + } + } else if (!strcasecmp(argv[0],"slowlog-max-len") && argc == 2) { + server.slowlog_max_len = (PORT_ULONG)(strtol(argv[1],NULL,10)); WIN_PORT_FIX /* cast (PORT_ULONG) */ + } else if (!strcasecmp(argv[0],"client-output-buffer-limit") && + argc == 5) + { + int class = getClientTypeByName(argv[1]); + PORT_ULONGLONG hard, soft; + int soft_seconds; + + if (class == -1 || class == CLIENT_TYPE_MASTER) { + err = "Unrecognized client limit class: the user specified " + "an invalid one, or 'master' which has no buffer limits."; + goto loaderr; + } + hard = memtoll(argv[2],NULL); + soft = memtoll(argv[3],NULL); + soft_seconds = atoi(argv[4]); + if (soft_seconds < 0) { + err = "Negative number of seconds in soft limit is invalid"; + goto loaderr; + } + server.client_obuf_limits[class].hard_limit_bytes = hard; + server.client_obuf_limits[class].soft_limit_bytes = soft; + server.client_obuf_limits[class].soft_limit_seconds = soft_seconds; + } else if (!strcasecmp(argv[0],"stop-writes-on-bgsave-error") && + argc == 2) { + if ((server.stop_writes_on_bgsave_err = yesnotoi(argv[1])) == -1) { + err = "argument must be 'yes' or 'no'"; goto loaderr; + } + } else if (!strcasecmp(argv[0],"slave-priority") && argc == 2) { + server.slave_priority = atoi(argv[1]); + } else if (!strcasecmp(argv[0],"slave-announce-ip") && argc == 2) { + zfree(server.slave_announce_ip); + server.slave_announce_ip = zstrdup(argv[1]); + } else if (!strcasecmp(argv[0],"slave-announce-port") && argc == 2) { + server.slave_announce_port = atoi(argv[1]); + if (server.slave_announce_port < 0 || + server.slave_announce_port > 65535) + { + err = "Invalid port"; goto loaderr; + } + } else if (!strcasecmp(argv[0],"min-slaves-to-write") && argc == 2) { + server.repl_min_slaves_to_write = atoi(argv[1]); + if (server.repl_min_slaves_to_write < 0) { + err = "Invalid value for min-slaves-to-write."; goto loaderr; + } + } else if (!strcasecmp(argv[0],"min-slaves-max-lag") && argc == 2) { + server.repl_min_slaves_max_lag = atoi(argv[1]); + if (server.repl_min_slaves_max_lag < 0) { + err = "Invalid value for min-slaves-max-lag."; goto loaderr; + } + } else if (!strcasecmp(argv[0],"notify-keyspace-events") && argc == 2) { + int flags = keyspaceEventsStringToFlags(argv[1]); + + if (flags == -1) { + err = "Invalid event class character. Use 'g$lshzxeA'."; + goto loaderr; + } + server.notify_keyspace_events = flags; + } else if (!strcasecmp(argv[0],"supervised") && argc == 2) { + server.supervised_mode = + configEnumGetValue(supervised_mode_enum,argv[1]); + + if (server.supervised_mode == INT_MIN) { + err = "Invalid option for 'supervised'. " + "Allowed values: 'upstart', 'systemd', 'auto', or 'no'"; + goto loaderr; + } + } else if (!strcasecmp(argv[0],"loadmodule") && argc >= 2) { + queueLoadModule(argv[1],&argv[2],argc-2); + } else if (!strcasecmp(argv[0],"sentinel")) { + /* argc == 1 is handled by main() as we need to enter the sentinel + * mode ASAP. */ + if (argc != 1) { + if (!server.sentinel_mode) { + err = "sentinel directive while not in sentinel mode"; + goto loaderr; + } + err = sentinelHandleConfiguration(argv+1,argc-1); + if (err) goto loaderr; + } +#ifdef _WIN32 + } else if (!strcasecmp(argv[0], "service-name")) { + // ignore. This is taken care of in the win32_service code. + } else if (!strcasecmp(argv[0], "persistence-available")) { + if (strcasecmp(argv[1], "no") == 0) { + // remove BGSAVE, BGREWRITEAOF and replication commands + // when persistence is disabled + int retval; + sds bgsave; + sds bgrewriteaof; + sds replconf; + sds psync; + sds sync; + + bgsave = sdsnew("bgsave"); + bgrewriteaof = sdsnew("bgrewriteaof"); + replconf = sdsnew("replconf"); + psync = sdsnew("psync"); + sync = sdsnew("sync"); + + retval = dictDelete(server.commands, bgsave); + serverAssert(retval == DICT_OK); + retval = dictDelete(server.commands, bgrewriteaof); + serverAssert(retval == DICT_OK); + retval = dictDelete(server.commands, replconf); + serverAssert(retval == DICT_OK); + retval = dictDelete(server.commands, psync); + serverAssert(retval == DICT_OK); + retval = dictDelete(server.commands, sync); + serverAssert(retval == DICT_OK); + + sdsfree(bgsave); + sdsfree(bgrewriteaof); + sdsfree(replconf); + sdsfree(psync); + sdsfree(sync); + } +#endif + } else { + err = "Bad directive or wrong number of arguments"; goto loaderr; + } + sdsfreesplitres(argv,argc); + } + + /* Sanity checks. */ + if (server.cluster_enabled && server.masterhost) { + linenum = slaveof_linenum; + i = linenum-1; + err = "slaveof directive not allowed in cluster mode"; + goto loaderr; + } + + sdsfreesplitres(lines,totlines); + return; + +loaderr: +#ifdef _WIN32 + serverLog(LL_WARNING, "\n*** FATAL CONFIG FILE ERROR ***\n"); + serverLog(LL_WARNING, "Reading the configuration file, at line %d\n", linenum); + serverLog(LL_WARNING, ">>> '%s'\n", lines[i]); + serverLog(LL_WARNING, "%s\n", err); +#else + fprintf(stderr, "\n*** FATAL CONFIG FILE ERROR ***\n"); + fprintf(stderr, "Reading the configuration file, at line %d\n", linenum); + fprintf(stderr, ">>> '%s'\n", lines[i]); + fprintf(stderr, "%s\n", err); +#endif + exit(1); +} + +/* Load the server configuration from the specified filename. + * The function appends the additional configuration directives stored + * in the 'options' string to the config file before loading. + * + * Both filename and options can be NULL, in such a case are considered + * empty. This way loadServerConfig can be used to just load a file or + * just load a string. */ +void loadServerConfig(char *filename, char *options) { + sds config = sdsempty(); + char buf[CONFIG_MAX_LINE+1]; + + /* Load the file content */ + if (filename) { + FILE *fp; + + if (filename[0] == '-' && filename[1] == '\0') { + fp = stdin; + } else { + if ((fp = fopen(filename,IF_WIN32("rb","r"))) == NULL) { + serverLog(LL_WARNING, + "Fatal error, can't open config file '%s'", filename); + exit(1); + } + } + while(fgets(buf,CONFIG_MAX_LINE+1,fp) != NULL) + config = sdscat(config,buf); + if (fp != stdin) fclose(fp); + } + /* Append the additional options */ + if (options) { + config = sdscat(config,"\n"); + config = sdscat(config,options); + } + loadServerConfigFromString(config); + sdsfree(config); +} + +/*----------------------------------------------------------------------------- + * CONFIG SET implementation + *----------------------------------------------------------------------------*/ + +#define config_set_bool_field(_name,_var) \ + } else if (!strcasecmp(c->argv[2]->ptr,_name)) { \ + int yn = yesnotoi(o->ptr); \ + if (yn == -1) goto badfmt; \ + _var = yn; + +#define config_set_numerical_field(_name,_var,min,max) \ + } else if (!strcasecmp(c->argv[2]->ptr,_name)) { \ + if (getLongLongFromObject(o,&ll) == C_ERR) goto badfmt; \ + if (min != LLONG_MIN && ll < min) goto badfmt; \ + if (max != LLONG_MAX && ll > max) goto badfmt; \ + _var = ll; + +#define config_set_memory_field(_name,_var) \ + } else if (!strcasecmp(c->argv[2]->ptr,_name)) { \ + ll = memtoll(o->ptr,&err); \ + if (err || ll < 0) goto badfmt; \ + _var = ll; + +#define config_set_enum_field(_name,_var,_enumvar) \ + } else if (!strcasecmp(c->argv[2]->ptr,_name)) { \ + int enumval = configEnumGetValue(_enumvar,o->ptr); \ + if (enumval == INT_MIN) goto badfmt; \ + _var = enumval; + +#define config_set_special_field(_name) \ + } else if (!strcasecmp(c->argv[2]->ptr,_name)) { + +#define config_set_else } else + +void configSetCommand(client *c) { + robj *o; + PORT_LONGLONG ll; + int err; + serverAssertWithInfo(c,c->argv[2],sdsEncodedObject(c->argv[2])); + serverAssertWithInfo(c,c->argv[3],sdsEncodedObject(c->argv[3])); + o = c->argv[3]; + + if (0) { /* this starts the config_set macros else-if chain. */ + + /* Special fields that can't be handled with general macros. */ + config_set_special_field("dbfilename") { + if (!pathIsBaseName(o->ptr)) { + addReplyError(c, "dbfilename can't be a path, just a filename"); + return; + } + zfree(server.rdb_filename); + server.rdb_filename = zstrdup(o->ptr); + } config_set_special_field("requirepass") { + if (sdslen(o->ptr) > CONFIG_AUTHPASS_MAX_LEN) goto badfmt; + zfree(server.requirepass); + server.requirepass = ((char*)o->ptr)[0] ? zstrdup(o->ptr) : NULL; + } config_set_special_field("masterauth") { + zfree(server.masterauth); + server.masterauth = ((char*)o->ptr)[0] ? zstrdup(o->ptr) : NULL; + } config_set_special_field("cluster-announce-ip") { + zfree(server.cluster_announce_ip); + server.cluster_announce_ip = ((char*)o->ptr)[0] ? zstrdup(o->ptr) : NULL; + } config_set_special_field("maxclients") { + int orig_value = server.maxclients; + + if (getLongLongFromObject(o,&ll) == C_ERR || ll < 1) goto badfmt; + + /* Try to check if the OS is capable of supporting so many FDs. */ + server.maxclients = (int)ll; WIN_PORT_FIX /* cast (int) */ + if (ll > orig_value) { + adjustOpenFilesLimit(); + if (server.maxclients != ll) { + addReplyErrorFormat(c,"The operating system is not able to handle the specified number of clients, try with %d", server.maxclients); + server.maxclients = orig_value; + return; + } + if ((unsigned int) aeGetSetSize(server.el) < + server.maxclients + CONFIG_FDSET_INCR) + { + if (aeResizeSetSize(server.el, + server.maxclients + CONFIG_FDSET_INCR) == AE_ERR) + { + addReplyError(c,"The event loop API used by Redis is not able to handle the specified number of clients"); + server.maxclients = orig_value; + return; + } + } + } + } config_set_special_field("appendonly") { + int enable = yesnotoi(o->ptr); + + if (enable == -1) goto badfmt; + if (enable == 0 && server.aof_state != AOF_OFF) { + stopAppendOnly(); + } else if (enable && server.aof_state == AOF_OFF) { + if (startAppendOnly() == C_ERR) { + addReplyError(c, + "Unable to turn on AOF. Check server logs."); + return; + } + } + } config_set_special_field("save") { + int vlen, j; + sds *v = sdssplitlen(o->ptr,sdslen(o->ptr)," ",1,&vlen); + + /* Perform sanity check before setting the new config: + * - Even number of args + * - Seconds >= 1, changes >= 0 */ + if (vlen & 1) { + sdsfreesplitres(v,vlen); + goto badfmt; + } + for (j = 0; j < vlen; j++) { + char *eptr; + PORT_LONG val; + + val = strtoll(v[j], &eptr, 10); + if (eptr[0] != '\0' || + ((j & 1) == 0 && val < 1) || + ((j & 1) == 1 && val < 0)) { + sdsfreesplitres(v,vlen); + goto badfmt; + } + } + /* Finally set the new config */ + resetServerSaveParams(); + for (j = 0; j < vlen; j += 2) { + time_t seconds; + int changes; + + seconds = strtoll(v[j],NULL,10); + changes = strtoll(v[j+1],NULL,10); + appendServerSaveParams(seconds, changes); + } + sdsfreesplitres(v,vlen); + } config_set_special_field("dir") { + if (chdir((char*)o->ptr) == -1) { + addReplyErrorFormat(c,"Changing directory: %s", IF_WIN32(wsa_strerror(errno), strerror(errno))); + return; + } + } config_set_special_field("client-output-buffer-limit") { + int vlen, j; + sds *v = sdssplitlen(o->ptr,sdslen(o->ptr)," ",1,&vlen); + + /* We need a multiple of 4: */ + if (vlen % 4) { + sdsfreesplitres(v,vlen); + goto badfmt; + } + + /* Sanity check of single arguments, so that we either refuse the + * whole configuration string or accept it all, even if a single + * error in a single client class is present. */ + for (j = 0; j < vlen; j++) { + PORT_LONG val; + + if ((j % 4) == 0) { + int class = getClientTypeByName(v[j]); + if (class == -1 || class == CLIENT_TYPE_MASTER) { + sdsfreesplitres(v,vlen); + goto badfmt; + } + } else { + val = memtoll(v[j], &err); + if (err || val < 0) { + sdsfreesplitres(v,vlen); + goto badfmt; + } + } + } + /* Finally set the new config */ + for (j = 0; j < vlen; j += 4) { + int class; + PORT_ULONGLONG hard, soft; + int soft_seconds; + + class = getClientTypeByName(v[j]); + hard = strtoll(v[j+1],NULL,10); + soft = strtoll(v[j+2],NULL,10); + soft_seconds = strtoll(v[j+3],NULL,10); + + server.client_obuf_limits[class].hard_limit_bytes = hard; + server.client_obuf_limits[class].soft_limit_bytes = soft; + server.client_obuf_limits[class].soft_limit_seconds = soft_seconds; + } + sdsfreesplitres(v,vlen); + } config_set_special_field("notify-keyspace-events") { + int flags = keyspaceEventsStringToFlags(o->ptr); + + if (flags == -1) goto badfmt; + server.notify_keyspace_events = flags; + } config_set_special_field("slave-announce-ip") { + zfree(server.slave_announce_ip); + server.slave_announce_ip = ((char*)o->ptr)[0] ? zstrdup(o->ptr) : NULL; + + /* Boolean fields. + * config_set_bool_field(name,var). */ + } config_set_bool_field( + "rdbcompression", server.rdb_compression) { + } config_set_bool_field( + "repl-disable-tcp-nodelay",server.repl_disable_tcp_nodelay) { + } config_set_bool_field( + "repl-diskless-sync",server.repl_diskless_sync) { + } config_set_bool_field( + "cluster-require-full-coverage",server.cluster_require_full_coverage) { + } config_set_bool_field( + "aof-rewrite-incremental-fsync",server.aof_rewrite_incremental_fsync) { + } config_set_bool_field( + "aof-load-truncated",server.aof_load_truncated) { + } config_set_bool_field( + "aof-use-rdb-preamble",server.aof_use_rdb_preamble) { + } config_set_bool_field( + "slave-serve-stale-data",server.repl_serve_stale_data) { + } config_set_bool_field( + "slave-read-only",server.repl_slave_ro) { + } config_set_bool_field( + "activerehashing",server.activerehashing) { + } config_set_bool_field( + "activedefrag",server.active_defrag_enabled) { +#ifndef HAVE_DEFRAG + if (server.active_defrag_enabled) { + server.active_defrag_enabled = 0; + addReplyError(c, + "Active defragmentation cannot be enabled: it requires a " + "Redis server compiled with a modified Jemalloc like the " + "one shipped by default with the Redis source distribution"); + return; + } +#endif + } config_set_bool_field( + "protected-mode",server.protected_mode) { + } config_set_bool_field( + "stop-writes-on-bgsave-error",server.stop_writes_on_bgsave_err) { + } config_set_bool_field( + "lazyfree-lazy-eviction",server.lazyfree_lazy_eviction) { + } config_set_bool_field( + "lazyfree-lazy-expire",server.lazyfree_lazy_expire) { + } config_set_bool_field( + "lazyfree-lazy-server-del",server.lazyfree_lazy_server_del) { + } config_set_bool_field( + "slave-lazy-flush",server.repl_slave_lazy_flush) { + } config_set_bool_field( + "no-appendfsync-on-rewrite",server.aof_no_fsync_on_rewrite) { + + /* Numerical fields. + * config_set_numerical_field(name,var,min,max) */ + } config_set_numerical_field( + "tcp-keepalive",server.tcpkeepalive,0,LLONG_MAX) { + } config_set_numerical_field( + "maxmemory-samples",server.maxmemory_samples,1,LLONG_MAX) { + } config_set_numerical_field( + "lfu-log-factor",server.lfu_log_factor,0,LLONG_MAX) { + } config_set_numerical_field( + "lfu-decay-time",server.lfu_decay_time,0,LLONG_MAX) { + } config_set_numerical_field( + "timeout",server.maxidletime,0,LONG_MAX) { + } config_set_numerical_field( + "active-defrag-threshold-lower",server.active_defrag_threshold_lower,0,1000) { + } config_set_numerical_field( + "active-defrag-threshold-upper",server.active_defrag_threshold_upper,0,1000) { + } config_set_memory_field( + "active-defrag-ignore-bytes",server.active_defrag_ignore_bytes) { + } config_set_numerical_field( + "active-defrag-cycle-min",server.active_defrag_cycle_min,1,99) { + } config_set_numerical_field( + "active-defrag-cycle-max",server.active_defrag_cycle_max,1,99) { + } config_set_numerical_field( + "auto-aof-rewrite-percentage",server.aof_rewrite_perc,0,LLONG_MAX){ + } config_set_numerical_field( + "hash-max-ziplist-entries",server.hash_max_ziplist_entries,0,LLONG_MAX) { + } config_set_numerical_field( + "hash-max-ziplist-value",server.hash_max_ziplist_value,0,LLONG_MAX) { + } config_set_numerical_field( + "list-max-ziplist-size",server.list_max_ziplist_size,INT_MIN,INT_MAX) { + } config_set_numerical_field( + "list-compress-depth",server.list_compress_depth,0,INT_MAX) { + } config_set_numerical_field( + "set-max-intset-entries",server.set_max_intset_entries,0,LLONG_MAX) { + } config_set_numerical_field( + "zset-max-ziplist-entries",server.zset_max_ziplist_entries,0,LLONG_MAX) { + } config_set_numerical_field( + "zset-max-ziplist-value",server.zset_max_ziplist_value,0,LLONG_MAX) { + } config_set_numerical_field( + "hll-sparse-max-bytes",server.hll_sparse_max_bytes,0,LLONG_MAX) { + } config_set_numerical_field( + "lua-time-limit",server.lua_time_limit,0,LLONG_MAX) { + } config_set_numerical_field( + "slowlog-log-slower-than",server.slowlog_log_slower_than,0,LLONG_MAX) { + } config_set_numerical_field( + "slowlog-max-len",ll,0,LLONG_MAX) { + /* Cast to unsigned. */ + server.slowlog_max_len = (unsigned)ll; + } config_set_numerical_field( + "latency-monitor-threshold",server.latency_monitor_threshold,0,LLONG_MAX){ + } config_set_numerical_field( + "repl-ping-slave-period",server.repl_ping_slave_period,1,LLONG_MAX) { + } config_set_numerical_field( + "repl-timeout",server.repl_timeout,1,LLONG_MAX) { + } config_set_numerical_field( + "repl-backlog-ttl",server.repl_backlog_time_limit,0,LLONG_MAX) { + } config_set_numerical_field( + "repl-diskless-sync-delay",server.repl_diskless_sync_delay,0,LLONG_MAX) { + } config_set_numerical_field( + "slave-priority",server.slave_priority,0,LLONG_MAX) { + } config_set_numerical_field( + "slave-announce-port",server.slave_announce_port,0,65535) { + } config_set_numerical_field( + "min-slaves-to-write",server.repl_min_slaves_to_write,0,LLONG_MAX) { + refreshGoodSlavesCount(); + } config_set_numerical_field( + "min-slaves-max-lag",server.repl_min_slaves_max_lag,0,LLONG_MAX) { + refreshGoodSlavesCount(); + } config_set_numerical_field( + "cluster-node-timeout",server.cluster_node_timeout,0,LLONG_MAX) { + } config_set_numerical_field( + "cluster-announce-port",server.cluster_announce_port,0,65535) { + } config_set_numerical_field( + "cluster-announce-bus-port",server.cluster_announce_bus_port,0,65535) { + } config_set_numerical_field( + "cluster-migration-barrier",server.cluster_migration_barrier,0,LLONG_MAX){ + } config_set_numerical_field( + "cluster-slave-validity-factor",server.cluster_slave_validity_factor,0,LLONG_MAX) { + } config_set_numerical_field( + "hz",server.hz,0,LLONG_MAX) { + /* Hz is more an hint from the user, so we accept values out of range + * but cap them to reasonable values. */ + if (server.hz < CONFIG_MIN_HZ) server.hz = CONFIG_MIN_HZ; + if (server.hz > CONFIG_MAX_HZ) server.hz = CONFIG_MAX_HZ; + } config_set_numerical_field( + "watchdog-period",ll,0,LLONG_MAX) { + if (ll) + enableWatchdog(ll); + else + disableWatchdog(); + + /* Memory fields. + * config_set_memory_field(name,var) */ + } config_set_memory_field("maxmemory",server.maxmemory) { + if (server.maxmemory) { + if (server.maxmemory < zmalloc_used_memory()) { + serverLog(LL_WARNING,"WARNING: the new maxmemory value set via CONFIG SET is smaller than the current memory usage. This will result in keys eviction and/or inability to accept new write commands depending on the maxmemory-policy."); + } + freeMemoryIfNeeded(); + } + } config_set_memory_field("repl-backlog-size",ll) { + resizeReplicationBacklog(ll); + } config_set_memory_field("auto-aof-rewrite-min-size",ll) { + server.aof_rewrite_min_size = ll; + + /* Enumeration fields. + * config_set_enum_field(name,var,enum_var) */ + } config_set_enum_field( + "loglevel",server.verbosity,loglevel_enum) { +#ifdef _WIN32 + setLogVerbosityLevel(server.verbosity); +#endif + } config_set_enum_field( + "maxmemory-policy",server.maxmemory_policy,maxmemory_policy_enum) { + } config_set_enum_field( + "appendfsync",server.aof_fsync,aof_fsync_enum) { + + /* Everyhing else is an error... */ + } config_set_else { + addReplyErrorFormat(c,"Unsupported CONFIG parameter: %s", + (char*)c->argv[2]->ptr); + return; + } + + /* On success we just return a generic OK for all the options. */ + addReply(c,shared.ok); + return; + +badfmt: /* Bad format errors */ + addReplyErrorFormat(c,"Invalid argument '%s' for CONFIG SET '%s'", + (char*)o->ptr, + (char*)c->argv[2]->ptr); +} + +/*----------------------------------------------------------------------------- + * CONFIG GET implementation + *----------------------------------------------------------------------------*/ + +#define config_get_string_field(_name,_var) do { \ + if (stringmatch(pattern,_name,1)) { \ + addReplyBulkCString(c,_name); \ + addReplyBulkCString(c,_var ? _var : ""); \ + matches++; \ + } \ +} while(0); + +#define config_get_bool_field(_name,_var) do { \ + if (stringmatch(pattern,_name,1)) { \ + addReplyBulkCString(c,_name); \ + addReplyBulkCString(c,_var ? "yes" : "no"); \ + matches++; \ + } \ +} while(0); + +#define config_get_numerical_field(_name,_var) do { \ + if (stringmatch(pattern,_name,1)) { \ + ll2string(buf,sizeof(buf),_var); \ + addReplyBulkCString(c,_name); \ + addReplyBulkCString(c,buf); \ + matches++; \ + } \ +} while(0); + +#define config_get_enum_field(_name,_var,_enumvar) do { \ + if (stringmatch(pattern,_name,1)) { \ + addReplyBulkCString(c,_name); \ + addReplyBulkCString(c,configEnumGetNameOrUnknown(_enumvar,_var)); \ + matches++; \ + } \ +} while(0); + +void configGetCommand(client *c) { + robj *o = c->argv[2]; + void *replylen = addDeferredMultiBulkLength(c); + char *pattern = o->ptr; + char buf[128]; + int matches = 0; + serverAssertWithInfo(c,o,sdsEncodedObject(o)); + + /* String values */ + config_get_string_field("dbfilename",server.rdb_filename); + config_get_string_field("requirepass",server.requirepass); + config_get_string_field("masterauth",server.masterauth); + config_get_string_field("cluster-announce-ip",server.cluster_announce_ip); + config_get_string_field("unixsocket",server.unixsocket); + config_get_string_field("logfile",server.logfile); + config_get_string_field("pidfile",server.pidfile); + config_get_string_field("slave-announce-ip",server.slave_announce_ip); + + /* Numerical values */ + config_get_numerical_field("maxmemory",server.maxmemory); + config_get_numerical_field("maxmemory-samples",server.maxmemory_samples); + config_get_numerical_field("timeout",server.maxidletime); + config_get_numerical_field("active-defrag-threshold-lower",server.active_defrag_threshold_lower); + config_get_numerical_field("active-defrag-threshold-upper",server.active_defrag_threshold_upper); + config_get_numerical_field("active-defrag-ignore-bytes",server.active_defrag_ignore_bytes); + config_get_numerical_field("active-defrag-cycle-min",server.active_defrag_cycle_min); + config_get_numerical_field("active-defrag-cycle-max",server.active_defrag_cycle_max); + config_get_numerical_field("auto-aof-rewrite-percentage", + server.aof_rewrite_perc); + config_get_numerical_field("auto-aof-rewrite-min-size", + server.aof_rewrite_min_size); + config_get_numerical_field("hash-max-ziplist-entries", + server.hash_max_ziplist_entries); + config_get_numerical_field("hash-max-ziplist-value", + server.hash_max_ziplist_value); + config_get_numerical_field("list-max-ziplist-size", + server.list_max_ziplist_size); + config_get_numerical_field("list-compress-depth", + server.list_compress_depth); + config_get_numerical_field("set-max-intset-entries", + server.set_max_intset_entries); + config_get_numerical_field("zset-max-ziplist-entries", + server.zset_max_ziplist_entries); + config_get_numerical_field("zset-max-ziplist-value", + server.zset_max_ziplist_value); + config_get_numerical_field("hll-sparse-max-bytes", + server.hll_sparse_max_bytes); + config_get_numerical_field("lua-time-limit",server.lua_time_limit); + config_get_numerical_field("slowlog-log-slower-than", + server.slowlog_log_slower_than); + config_get_numerical_field("latency-monitor-threshold", + server.latency_monitor_threshold); + config_get_numerical_field("slowlog-max-len", + server.slowlog_max_len); + config_get_numerical_field("port",server.port); + config_get_numerical_field("cluster-announce-port",server.cluster_announce_port); + config_get_numerical_field("cluster-announce-bus-port",server.cluster_announce_bus_port); + config_get_numerical_field("tcp-backlog",server.tcp_backlog); + config_get_numerical_field("databases",server.dbnum); + config_get_numerical_field("repl-ping-slave-period",server.repl_ping_slave_period); + config_get_numerical_field("repl-timeout",server.repl_timeout); + config_get_numerical_field("repl-backlog-size",server.repl_backlog_size); + config_get_numerical_field("repl-backlog-ttl",server.repl_backlog_time_limit); + config_get_numerical_field("maxclients",server.maxclients); + config_get_numerical_field("watchdog-period",server.watchdog_period); + config_get_numerical_field("slave-priority",server.slave_priority); + config_get_numerical_field("slave-announce-port",server.slave_announce_port); + config_get_numerical_field("min-slaves-to-write",server.repl_min_slaves_to_write); + config_get_numerical_field("min-slaves-max-lag",server.repl_min_slaves_max_lag); + config_get_numerical_field("hz",server.hz); + config_get_numerical_field("cluster-node-timeout",server.cluster_node_timeout); + config_get_numerical_field("cluster-migration-barrier",server.cluster_migration_barrier); + config_get_numerical_field("cluster-slave-validity-factor",server.cluster_slave_validity_factor); + config_get_numerical_field("repl-diskless-sync-delay",server.repl_diskless_sync_delay); + config_get_numerical_field("tcp-keepalive",server.tcpkeepalive); + + /* Bool (yes/no) values */ + config_get_bool_field("cluster-require-full-coverage", + server.cluster_require_full_coverage); + config_get_bool_field("no-appendfsync-on-rewrite", + server.aof_no_fsync_on_rewrite); + config_get_bool_field("slave-serve-stale-data", + server.repl_serve_stale_data); + config_get_bool_field("slave-read-only", + server.repl_slave_ro); + config_get_bool_field("stop-writes-on-bgsave-error", + server.stop_writes_on_bgsave_err); + config_get_bool_field("daemonize", server.daemonize); + config_get_bool_field("rdbcompression", server.rdb_compression); + config_get_bool_field("rdbchecksum", server.rdb_checksum); + config_get_bool_field("activerehashing", server.activerehashing); + config_get_bool_field("activedefrag", server.active_defrag_enabled); + config_get_bool_field("protected-mode", server.protected_mode); + config_get_bool_field("repl-disable-tcp-nodelay", + server.repl_disable_tcp_nodelay); + config_get_bool_field("repl-diskless-sync", + server.repl_diskless_sync); + config_get_bool_field("aof-rewrite-incremental-fsync", + server.aof_rewrite_incremental_fsync); + config_get_bool_field("aof-load-truncated", + server.aof_load_truncated); + config_get_bool_field("aof-use-rdb-preamble", + server.aof_use_rdb_preamble); + config_get_bool_field("lazyfree-lazy-eviction", + server.lazyfree_lazy_eviction); + config_get_bool_field("lazyfree-lazy-expire", + server.lazyfree_lazy_expire); + config_get_bool_field("lazyfree-lazy-server-del", + server.lazyfree_lazy_server_del); + config_get_bool_field("slave-lazy-flush", + server.repl_slave_lazy_flush); + + /* Enum values */ + config_get_enum_field("maxmemory-policy", + server.maxmemory_policy,maxmemory_policy_enum); + config_get_enum_field("loglevel", + server.verbosity,loglevel_enum); + config_get_enum_field("supervised", + server.supervised_mode,supervised_mode_enum); + config_get_enum_field("appendfsync", + server.aof_fsync,aof_fsync_enum); +#ifndef _WIN32 + config_get_enum_field("syslog-facility", + server.syslog_facility,syslog_facility_enum); +#endif + + /* Everything we can't handle with macros follows. */ + + if (stringmatch(pattern,"appendonly",1)) { + addReplyBulkCString(c,"appendonly"); + addReplyBulkCString(c,server.aof_state == AOF_OFF ? "no" : "yes"); + matches++; + } + if (stringmatch(pattern,"dir",1)) { + char buf[1024]; + + if (getcwd(buf,sizeof(buf)) == NULL) + buf[0] = '\0'; + + addReplyBulkCString(c,"dir"); + addReplyBulkCString(c,buf); + matches++; + } + if (stringmatch(pattern,"save",1)) { + sds buf = sdsempty(); + int j; + + for (j = 0; j < server.saveparamslen; j++) { + buf = sdscatprintf(buf,"%jd %d", + (intmax_t)server.saveparams[j].seconds, + server.saveparams[j].changes); + if (j != server.saveparamslen-1) + buf = sdscatlen(buf," ",1); + } + addReplyBulkCString(c,"save"); + addReplyBulkCString(c,buf); + sdsfree(buf); + matches++; + } + if (stringmatch(pattern,"client-output-buffer-limit",1)) { + sds buf = sdsempty(); + int j; + + for (j = 0; j < CLIENT_TYPE_OBUF_COUNT; j++) { + buf = sdscatprintf(buf,"%s %llu %llu %Id", WIN_PORT_FIX /* %ld -> %Id */ + getClientTypeName(j), + server.client_obuf_limits[j].hard_limit_bytes, + server.client_obuf_limits[j].soft_limit_bytes, + (PORT_LONG) server.client_obuf_limits[j].soft_limit_seconds); + if (j != CLIENT_TYPE_OBUF_COUNT-1) + buf = sdscatlen(buf," ",1); + } + addReplyBulkCString(c,"client-output-buffer-limit"); + addReplyBulkCString(c,buf); + sdsfree(buf); + matches++; + } + if (stringmatch(pattern,"unixsocketperm",1)) { + char buf[32]; + snprintf(buf,sizeof(buf),"%o",server.unixsocketperm); + addReplyBulkCString(c,"unixsocketperm"); + addReplyBulkCString(c,buf); + matches++; + } + if (stringmatch(pattern,"slaveof",1)) { + char buf[256]; + + addReplyBulkCString(c,"slaveof"); + if (server.masterhost) + snprintf(buf,sizeof(buf),"%s %d", + server.masterhost, server.masterport); + else + buf[0] = '\0'; + addReplyBulkCString(c,buf); + matches++; + } + if (stringmatch(pattern,"notify-keyspace-events",1)) { + robj *flagsobj = createObject(OBJ_STRING, + keyspaceEventsFlagsToString(server.notify_keyspace_events)); + + addReplyBulkCString(c,"notify-keyspace-events"); + addReplyBulk(c,flagsobj); + decrRefCount(flagsobj); + matches++; + } + if (stringmatch(pattern,"bind",1)) { + sds aux = sdsjoin(server.bindaddr,server.bindaddr_count," "); + + addReplyBulkCString(c,"bind"); + addReplyBulkCString(c,aux); + sdsfree(aux); + matches++; + } + setDeferredMultiBulkLength(c,replylen,matches*2); +} + +/*----------------------------------------------------------------------------- + * CONFIG REWRITE implementation + *----------------------------------------------------------------------------*/ + +#define REDIS_CONFIG_REWRITE_SIGNATURE "# Generated by CONFIG REWRITE" + +/* We use the following dictionary type to store where a configuration + * option is mentioned in the old configuration file, so it's + * like "maxmemory" -> list of line numbers (first line is zero). */ +uint64_t dictSdsCaseHash(const void *key); +int dictSdsKeyCaseCompare(void *privdata, const void *key1, const void *key2); +void dictSdsDestructor(void *privdata, void *val); +void dictListDestructor(void *privdata, void *val); + +/* Sentinel config rewriting is implemented inside sentinel.c by + * rewriteConfigSentinelOption(). */ +void rewriteConfigSentinelOption(struct rewriteConfigState *state); + +dictType optionToLineDictType = { + dictSdsCaseHash, /* hash function */ + NULL, /* key dup */ + NULL, /* val dup */ + dictSdsKeyCaseCompare, /* key compare */ + dictSdsDestructor, /* key destructor */ + dictListDestructor /* val destructor */ +}; + +dictType optionSetDictType = { + dictSdsCaseHash, /* hash function */ + NULL, /* key dup */ + NULL, /* val dup */ + dictSdsKeyCaseCompare, /* key compare */ + dictSdsDestructor, /* key destructor */ + NULL /* val destructor */ +}; + +/* The config rewrite state. */ +struct rewriteConfigState { + dict *option_to_line; /* Option -> list of config file lines map */ + dict *rewritten; /* Dictionary of already processed options */ + int numlines; /* Number of lines in current config */ + sds *lines; /* Current lines as an array of sds strings */ + int has_tail; /* True if we already added directives that were + not present in the original config file. */ +}; + +/* Append the new line to the current configuration state. */ +void rewriteConfigAppendLine(struct rewriteConfigState *state, sds line) { + state->lines = zrealloc(state->lines, sizeof(char*) * (state->numlines+1)); + state->lines[state->numlines++] = line; +} + +/* Populate the option -> list of line numbers map. */ +void rewriteConfigAddLineNumberToOption(struct rewriteConfigState *state, sds option, int linenum) { + list *l = dictFetchValue(state->option_to_line,option); + + if (l == NULL) { + l = listCreate(); + dictAdd(state->option_to_line,sdsdup(option),l); + } + listAddNodeTail(l,(void*)(PORT_LONG)linenum); +} + +/* Add the specified option to the set of processed options. + * This is useful as only unused lines of processed options will be blanked + * in the config file, while options the rewrite process does not understand + * remain untouched. */ +void rewriteConfigMarkAsProcessed(struct rewriteConfigState *state, const char *option) { + sds opt = sdsnew(option); + + if (dictAdd(state->rewritten,opt,NULL) != DICT_OK) sdsfree(opt); +} + +/* Read the old file, split it into lines to populate a newly created + * config rewrite state, and return it to the caller. + * + * If it is impossible to read the old file, NULL is returned. + * If the old file does not exist at all, an empty state is returned. */ +struct rewriteConfigState *rewriteConfigReadOldFile(char *path) { + FILE *fp = fopen(path,"r"); + struct rewriteConfigState *state = zmalloc(sizeof(*state)); + char buf[CONFIG_MAX_LINE+1]; + int linenum = -1; + + if (fp == NULL && errno != ENOENT) return NULL; + + state->option_to_line = dictCreate(&optionToLineDictType,NULL); + state->rewritten = dictCreate(&optionSetDictType,NULL); + state->numlines = 0; + state->lines = NULL; + state->has_tail = 0; + if (fp == NULL) return state; + + /* Read the old file line by line, populate the state. */ + while(fgets(buf,CONFIG_MAX_LINE+1,fp) != NULL) { + int argc; + sds *argv; + sds line = sdstrim(sdsnew(buf),"\r\n\t "); + + linenum++; /* Zero based, so we init at -1 */ + + /* Handle comments and empty lines. */ + if (line[0] == '#' || line[0] == '\0') { + if (!state->has_tail && !strcmp(line,REDIS_CONFIG_REWRITE_SIGNATURE)) + state->has_tail = 1; + rewriteConfigAppendLine(state,line); + continue; + } + + /* Not a comment, split into arguments. */ + argv = sdssplitargs(line,&argc); + if (argv == NULL) { + /* Apparently the line is unparsable for some reason, for + * instance it may have unbalanced quotes. Load it as a + * comment. */ + sds aux = sdsnew("# ??? "); + aux = sdscatsds(aux,line); + sdsfree(line); + rewriteConfigAppendLine(state,aux); + continue; + } + + sdstolower(argv[0]); /* We only want lowercase config directives. */ + + /* Now we populate the state according to the content of this line. + * Append the line and populate the option -> line numbers map. */ + rewriteConfigAppendLine(state,line); + rewriteConfigAddLineNumberToOption(state,argv[0],linenum); + + sdsfreesplitres(argv,argc); + } + fclose(fp); + return state; +} + +/* Rewrite the specified configuration option with the new "line". + * It progressively uses lines of the file that were already used for the same + * configuration option in the old version of the file, removing that line from + * the map of options -> line numbers. + * + * If there are lines associated with a given configuration option and + * "force" is non-zero, the line is appended to the configuration file. + * Usually "force" is true when an option has not its default value, so it + * must be rewritten even if not present previously. + * + * The first time a line is appended into a configuration file, a comment + * is added to show that starting from that point the config file was generated + * by CONFIG REWRITE. + * + * "line" is either used, or freed, so the caller does not need to free it + * in any way. */ +void rewriteConfigRewriteLine(struct rewriteConfigState *state, const char *option, sds line, int force) { + sds o = sdsnew(option); + list *l = dictFetchValue(state->option_to_line,o); + + rewriteConfigMarkAsProcessed(state,option); + + if (!l && !force) { + /* Option not used previously, and we are not forced to use it. */ + sdsfree(line); + sdsfree(o); + return; + } + + if (l) { + listNode *ln = listFirst(l); + int linenum = (int)((PORT_LONG) ln->value); WIN_PORT_FIX /* cast (int) */ + + /* There are still lines in the old configuration file we can reuse + * for this option. Replace the line with the new one. */ + listDelNode(l,ln); + if (listLength(l) == 0) dictDelete(state->option_to_line,o); + sdsfree(state->lines[linenum]); + state->lines[linenum] = line; + } else { + /* Append a new line. */ + if (!state->has_tail) { + rewriteConfigAppendLine(state, + sdsnew(REDIS_CONFIG_REWRITE_SIGNATURE)); + state->has_tail = 1; + } + rewriteConfigAppendLine(state,line); + } + sdsfree(o); +} + +/* Write the PORT_LONGLONG 'bytes' value as a string in a way that is parsable + * inside redis.conf. If possible uses the GB, MB, KB notation. */ +int rewriteConfigFormatMemory(char *buf, size_t len, PORT_LONGLONG bytes) { + int gb = 1024*1024*1024; + int mb = 1024*1024; + int kb = 1024; + + if (bytes && (bytes % gb) == 0) { + return snprintf(buf,len,"%lldgb",bytes/gb); + } else if (bytes && (bytes % mb) == 0) { + return snprintf(buf,len,"%lldmb",bytes/mb); + } else if (bytes && (bytes % kb) == 0) { + return snprintf(buf,len,"%lldkb",bytes/kb); + } else { + return snprintf(buf,len,"%lld",bytes); + } +} + +/* Rewrite a simple "option-name " configuration option. */ +void rewriteConfigBytesOption(struct rewriteConfigState *state, char *option, PORT_LONGLONG value, PORT_LONGLONG defvalue) { + char buf[64]; + int force = value != defvalue; + sds line; + + rewriteConfigFormatMemory(buf,sizeof(buf),value); + line = sdscatprintf(sdsempty(),"%s %s",option,buf); + rewriteConfigRewriteLine(state,option,line,force); +} + +/* Rewrite a yes/no option. */ +void rewriteConfigYesNoOption(struct rewriteConfigState *state, char *option, int value, int defvalue) { + int force = value != defvalue; + sds line = sdscatprintf(sdsempty(),"%s %s",option, + value ? "yes" : "no"); + + rewriteConfigRewriteLine(state,option,line,force); +} + +/* Rewrite a string option. */ +void rewriteConfigStringOption(struct rewriteConfigState *state, char *option, char *value, char *defvalue) { + int force = 1; + sds line; + + /* String options set to NULL need to be not present at all in the + * configuration file to be set to NULL again at the next reboot. */ + if (value == NULL) { + rewriteConfigMarkAsProcessed(state,option); + return; + } + + /* Set force to zero if the value is set to its default. */ + if (defvalue && strcmp(value,defvalue) == 0) force = 0; + + line = sdsnew(option); + line = sdscatlen(line, " ", 1); + line = sdscatrepr(line, value, strlen(value)); + + rewriteConfigRewriteLine(state,option,line,force); +} + +/* Rewrite a numerical (PORT_LONGLONG range) option. */ +void rewriteConfigNumericalOption(struct rewriteConfigState *state, char *option, PORT_LONGLONG value, PORT_LONGLONG defvalue) { + int force = value != defvalue; + sds line = sdscatprintf(sdsempty(),"%s %lld",option,value); + + rewriteConfigRewriteLine(state,option,line,force); +} + +/* Rewrite a octal option. */ +void rewriteConfigOctalOption(struct rewriteConfigState *state, char *option, int value, int defvalue) { + int force = value != defvalue; + sds line = sdscatprintf(sdsempty(),"%s %o",option,value); + + rewriteConfigRewriteLine(state,option,line,force); +} + +/* Rewrite an enumeration option. It takes as usually state and option name, + * and in addition the enumeration array and the default value for the + * option. */ +void rewriteConfigEnumOption(struct rewriteConfigState *state, char *option, int value, configEnum *ce, int defval) { + sds line; + const char *name = configEnumGetNameOrUnknown(ce,value); + int force = value != defval; + + line = sdscatprintf(sdsempty(),"%s %s",option,name); + rewriteConfigRewriteLine(state,option,line,force); +} + +#ifndef _WIN32 +/* Rewrite the syslog-facility option. */ +void rewriteConfigSyslogfacilityOption(struct rewriteConfigState *state) { + int value = server.syslog_facility; + int force = value != LOG_LOCAL0; + const char *name = NULL, *option = "syslog-facility"; + sds line; + + name = configEnumGetNameOrUnknown(syslog_facility_enum,value); + line = sdscatprintf(sdsempty(),"%s %s",option,name); + rewriteConfigRewriteLine(state,option,line,force); +} +#endif + +/* Rewrite the save option. */ +void rewriteConfigSaveOption(struct rewriteConfigState *state) { + int j; + sds line; + + /* Note that if there are no save parameters at all, all the current + * config line with "save" will be detected as orphaned and deleted, + * resulting into no RDB persistence as expected. */ + for (j = 0; j < server.saveparamslen; j++) { + line = sdscatprintf(sdsempty(),"save %Id %d", WIN_PORT_FIX /* %ld -> %Id */ + (PORT_LONG) server.saveparams[j].seconds, server.saveparams[j].changes); + rewriteConfigRewriteLine(state,"save",line,1); + } + /* Mark "save" as processed in case server.saveparamslen is zero. */ + rewriteConfigMarkAsProcessed(state,"save"); +} + +/* Rewrite the dir option, always using absolute paths.*/ +void rewriteConfigDirOption(struct rewriteConfigState *state) { + char cwd[1024]; + + if (getcwd(cwd,sizeof(cwd)) == NULL) { + rewriteConfigMarkAsProcessed(state,"dir"); + return; /* no rewrite on error. */ + } + rewriteConfigStringOption(state,"dir",cwd,NULL); +} + +/* Rewrite the slaveof option. */ +void rewriteConfigSlaveofOption(struct rewriteConfigState *state) { + char *option = "slaveof"; + sds line; + + /* If this is a master, we want all the slaveof config options + * in the file to be removed. Note that if this is a cluster instance + * we don't want a slaveof directive inside redis.conf. */ + if (server.cluster_enabled || server.masterhost == NULL) { + rewriteConfigMarkAsProcessed(state,"slaveof"); + return; + } + line = sdscatprintf(sdsempty(),"%s %s %d", option, + server.masterhost, server.masterport); + rewriteConfigRewriteLine(state,option,line,1); +} + +/* Rewrite the notify-keyspace-events option. */ +void rewriteConfigNotifykeyspaceeventsOption(struct rewriteConfigState *state) { + int force = server.notify_keyspace_events != 0; + char *option = "notify-keyspace-events"; + sds line, flags; + + flags = keyspaceEventsFlagsToString(server.notify_keyspace_events); + line = sdsnew(option); + line = sdscatlen(line, " ", 1); + line = sdscatrepr(line, flags, sdslen(flags)); + sdsfree(flags); + rewriteConfigRewriteLine(state,option,line,force); +} + +/* Rewrite the client-output-buffer-limit option. */ +void rewriteConfigClientoutputbufferlimitOption(struct rewriteConfigState *state) { + int j; + char *option = "client-output-buffer-limit"; + + for (j = 0; j < CLIENT_TYPE_OBUF_COUNT; j++) { + int force = (server.client_obuf_limits[j].hard_limit_bytes != + clientBufferLimitsDefaults[j].hard_limit_bytes) || + (server.client_obuf_limits[j].soft_limit_bytes != + clientBufferLimitsDefaults[j].soft_limit_bytes) || + (server.client_obuf_limits[j].soft_limit_seconds != + clientBufferLimitsDefaults[j].soft_limit_seconds); + sds line; + char hard[64], soft[64]; + + rewriteConfigFormatMemory(hard,sizeof(hard), + server.client_obuf_limits[j].hard_limit_bytes); + rewriteConfigFormatMemory(soft,sizeof(soft), + server.client_obuf_limits[j].soft_limit_bytes); + + line = sdscatprintf(sdsempty(),"%s %s %s %s %Id", WIN_PORT_FIX /* %ld -> %Id */ + option, getClientTypeName(j), hard, soft, + (PORT_LONG) server.client_obuf_limits[j].soft_limit_seconds); + rewriteConfigRewriteLine(state,option,line,force); + } +} + +/* Rewrite the bind option. */ +void rewriteConfigBindOption(struct rewriteConfigState *state) { + int force = 1; + sds line, addresses; + char *option = "bind"; + + /* Nothing to rewrite if we don't have bind addresses. */ + if (server.bindaddr_count == 0) { + rewriteConfigMarkAsProcessed(state,option); + return; + } + + /* Rewrite as bind ... */ + addresses = sdsjoin(server.bindaddr,server.bindaddr_count," "); + line = sdsnew(option); + line = sdscatlen(line, " ", 1); + line = sdscatsds(line, addresses); + sdsfree(addresses); + + rewriteConfigRewriteLine(state,option,line,force); +} + +/* Glue together the configuration lines in the current configuration + * rewrite state into a single string, stripping multiple empty lines. */ +sds rewriteConfigGetContentFromState(struct rewriteConfigState *state) { + sds content = sdsempty(); + int j, was_empty = 0; + + for (j = 0; j < state->numlines; j++) { + /* Every cluster of empty lines is turned into a single empty line. */ + if (sdslen(state->lines[j]) == 0) { + if (was_empty) continue; + was_empty = 1; + } else { + was_empty = 0; + } + content = sdscatsds(content,state->lines[j]); + content = sdscatlen(content,"\n",1); + } + return content; +} + +/* Free the configuration rewrite state. */ +void rewriteConfigReleaseState(struct rewriteConfigState *state) { + sdsfreesplitres(state->lines,state->numlines); + dictRelease(state->option_to_line); + dictRelease(state->rewritten); + zfree(state); +} + +/* At the end of the rewrite process the state contains the remaining + * map between "option name" => "lines in the original config file". + * Lines used by the rewrite process were removed by the function + * rewriteConfigRewriteLine(), all the other lines are "orphaned" and + * should be replaced by empty lines. + * + * This function does just this, iterating all the option names and + * blanking all the lines still associated. */ +void rewriteConfigRemoveOrphaned(struct rewriteConfigState *state) { + dictIterator *di = dictGetIterator(state->option_to_line); + dictEntry *de; + + while((de = dictNext(di)) != NULL) { + list *l = dictGetVal(de); + sds option = dictGetKey(de); + + /* Don't blank lines about options the rewrite process + * don't understand. */ + if (dictFind(state->rewritten,option) == NULL) { + serverLog(LL_DEBUG,"Not rewritten option: %s", option); + continue; + } + + while(listLength(l)) { + listNode *ln = listFirst(l); + int linenum = (int)((PORT_LONG) ln->value); WIN_PORT_FIX /* cast (int) */ + + sdsfree(state->lines[linenum]); + state->lines[linenum] = sdsempty(); + listDelNode(l,ln); + } + } + dictReleaseIterator(di); +} + +/* This function overwrites the old configuration file with the new content. + * + * 1) The old file length is obtained. + * 2) If the new content is smaller, padding is added. + * 3) A single write(2) call is used to replace the content of the file. + * 4) Later the file is truncated to the length of the new content. + * + * This way we are sure the file is left in a consistent state even if the + * process is stopped between any of the four operations. + * + * The function returns 0 on success, otherwise -1 is returned and errno + * set accordingly. */ +int rewriteConfigOverwriteFile(char *configfile, sds content) { + int retval = 0; + int fd = open(configfile,O_RDWR|O_CREAT,0644); + int content_size = (int)sdslen(content), padding = 0; WIN_PORT_FIX /* cast (int) */ + struct IF_WIN32(_stat64,stat) sb; // TODO: verify for 32-bit + sds content_padded; + + /* 1) Open the old file (or create a new one if it does not + * exist), get the size. */ + if (fd == -1) return -1; /* errno set by open(). */ + if (fstat(fd,&sb) == -1) { + close(fd); + return -1; /* errno set by fstat(). */ + } + + /* 2) Pad the content at least match the old file size. */ + content_padded = sdsdup(content); + if (content_size < sb.st_size) { + /* If the old file was bigger, pad the content with + * a newline plus as many "#" chars as required. */ + padding = (int)(sb.st_size - content_size); WIN_PORT_FIX /* cast (int) */ + content_padded = sdsgrowzero(content_padded,sb.st_size); + content_padded[content_size] = '\n'; + memset(content_padded+content_size+1,'#',padding-1); + } + + /* 3) Write the new content using a single write(2). */ + if (write(fd,content_padded,strlen(content_padded)) == -1) { + retval = -1; + goto cleanup; + } + + /* 4) Truncate the file to the right length if we used padding. */ + if (padding) { + if (ftruncate(fd,content_size) == -1) { + /* Non critical error... */ + } + } + +cleanup: + sdsfree(content_padded); + close(fd); + return retval; +} + +/* Rewrite the configuration file at "path". + * If the configuration file already exists, we try at best to retain comments + * and overall structure. + * + * Configuration parameters that are at their default value, unless already + * explicitly included in the old configuration file, are not rewritten. + * + * On error -1 is returned and errno is set accordingly, otherwise 0. */ +int rewriteConfig(char *path) { + struct rewriteConfigState *state; + sds newcontent; + int retval; + + /* Step 1: read the old config into our rewrite state. */ + if ((state = rewriteConfigReadOldFile(path)) == NULL) return -1; + + /* Step 2: rewrite every single option, replacing or appending it inside + * the rewrite state. */ + + rewriteConfigYesNoOption(state,"daemonize",server.daemonize,0); + rewriteConfigStringOption(state,"pidfile",server.pidfile,CONFIG_DEFAULT_PID_FILE); + rewriteConfigNumericalOption(state,"port",server.port,CONFIG_DEFAULT_SERVER_PORT); + rewriteConfigNumericalOption(state,"cluster-announce-port",server.cluster_announce_port,CONFIG_DEFAULT_CLUSTER_ANNOUNCE_PORT); + rewriteConfigNumericalOption(state,"cluster-announce-bus-port",server.cluster_announce_bus_port,CONFIG_DEFAULT_CLUSTER_ANNOUNCE_BUS_PORT); + rewriteConfigNumericalOption(state,"tcp-backlog",server.tcp_backlog,CONFIG_DEFAULT_TCP_BACKLOG); + rewriteConfigBindOption(state); + rewriteConfigStringOption(state,"unixsocket",server.unixsocket,NULL); + rewriteConfigOctalOption(state,"unixsocketperm",server.unixsocketperm,CONFIG_DEFAULT_UNIX_SOCKET_PERM); + rewriteConfigNumericalOption(state,"timeout",server.maxidletime,CONFIG_DEFAULT_CLIENT_TIMEOUT); + rewriteConfigNumericalOption(state,"tcp-keepalive",server.tcpkeepalive,CONFIG_DEFAULT_TCP_KEEPALIVE); + rewriteConfigNumericalOption(state,"slave-announce-port",server.slave_announce_port,CONFIG_DEFAULT_SLAVE_ANNOUNCE_PORT); + rewriteConfigEnumOption(state,"loglevel",server.verbosity,loglevel_enum,CONFIG_DEFAULT_VERBOSITY); + rewriteConfigStringOption(state,"logfile",server.logfile,CONFIG_DEFAULT_LOGFILE); + rewriteConfigYesNoOption(state,"syslog-enabled",server.syslog_enabled,CONFIG_DEFAULT_SYSLOG_ENABLED); + rewriteConfigStringOption(state,"syslog-ident",server.syslog_ident,CONFIG_DEFAULT_SYSLOG_IDENT); +#ifndef _WIN32 + rewriteConfigSyslogfacilityOption(state); +#endif + rewriteConfigSaveOption(state); + rewriteConfigNumericalOption(state,"databases",server.dbnum,CONFIG_DEFAULT_DBNUM); + rewriteConfigYesNoOption(state,"stop-writes-on-bgsave-error",server.stop_writes_on_bgsave_err,CONFIG_DEFAULT_STOP_WRITES_ON_BGSAVE_ERROR); + rewriteConfigYesNoOption(state,"rdbcompression",server.rdb_compression,CONFIG_DEFAULT_RDB_COMPRESSION); + rewriteConfigYesNoOption(state,"rdbchecksum",server.rdb_checksum,CONFIG_DEFAULT_RDB_CHECKSUM); + rewriteConfigStringOption(state,"dbfilename",server.rdb_filename,CONFIG_DEFAULT_RDB_FILENAME); + rewriteConfigDirOption(state); + rewriteConfigSlaveofOption(state); + rewriteConfigStringOption(state,"slave-announce-ip",server.slave_announce_ip,CONFIG_DEFAULT_SLAVE_ANNOUNCE_IP); + rewriteConfigStringOption(state,"masterauth",server.masterauth,NULL); + rewriteConfigStringOption(state,"cluster-announce-ip",server.cluster_announce_ip,NULL); + rewriteConfigYesNoOption(state,"slave-serve-stale-data",server.repl_serve_stale_data,CONFIG_DEFAULT_SLAVE_SERVE_STALE_DATA); + rewriteConfigYesNoOption(state,"slave-read-only",server.repl_slave_ro,CONFIG_DEFAULT_SLAVE_READ_ONLY); + rewriteConfigNumericalOption(state,"repl-ping-slave-period",server.repl_ping_slave_period,CONFIG_DEFAULT_REPL_PING_SLAVE_PERIOD); + rewriteConfigNumericalOption(state,"repl-timeout",server.repl_timeout,CONFIG_DEFAULT_REPL_TIMEOUT); + rewriteConfigBytesOption(state,"repl-backlog-size",server.repl_backlog_size,CONFIG_DEFAULT_REPL_BACKLOG_SIZE); + rewriteConfigBytesOption(state,"repl-backlog-ttl",server.repl_backlog_time_limit,CONFIG_DEFAULT_REPL_BACKLOG_TIME_LIMIT); + rewriteConfigYesNoOption(state,"repl-disable-tcp-nodelay",server.repl_disable_tcp_nodelay,CONFIG_DEFAULT_REPL_DISABLE_TCP_NODELAY); + rewriteConfigYesNoOption(state,"repl-diskless-sync",server.repl_diskless_sync,CONFIG_DEFAULT_REPL_DISKLESS_SYNC); + rewriteConfigNumericalOption(state,"repl-diskless-sync-delay",server.repl_diskless_sync_delay,CONFIG_DEFAULT_REPL_DISKLESS_SYNC_DELAY); + rewriteConfigNumericalOption(state,"slave-priority",server.slave_priority,CONFIG_DEFAULT_SLAVE_PRIORITY); + rewriteConfigNumericalOption(state,"min-slaves-to-write",server.repl_min_slaves_to_write,CONFIG_DEFAULT_MIN_SLAVES_TO_WRITE); + rewriteConfigNumericalOption(state,"min-slaves-max-lag",server.repl_min_slaves_max_lag,CONFIG_DEFAULT_MIN_SLAVES_MAX_LAG); + rewriteConfigStringOption(state,"requirepass",server.requirepass,NULL); + rewriteConfigNumericalOption(state,"maxclients",server.maxclients,CONFIG_DEFAULT_MAX_CLIENTS); + rewriteConfigBytesOption(state,"maxmemory",server.maxmemory,CONFIG_DEFAULT_MAXMEMORY); + rewriteConfigEnumOption(state,"maxmemory-policy",server.maxmemory_policy,maxmemory_policy_enum,CONFIG_DEFAULT_MAXMEMORY_POLICY); + rewriteConfigNumericalOption(state,"maxmemory-samples",server.maxmemory_samples,CONFIG_DEFAULT_MAXMEMORY_SAMPLES); + rewriteConfigNumericalOption(state,"active-defrag-threshold-lower",server.active_defrag_threshold_lower,CONFIG_DEFAULT_DEFRAG_THRESHOLD_LOWER); + rewriteConfigNumericalOption(state,"active-defrag-threshold-upper",server.active_defrag_threshold_upper,CONFIG_DEFAULT_DEFRAG_THRESHOLD_UPPER); + rewriteConfigBytesOption(state,"active-defrag-ignore-bytes",server.active_defrag_ignore_bytes,CONFIG_DEFAULT_DEFRAG_IGNORE_BYTES); + rewriteConfigNumericalOption(state,"active-defrag-cycle-min",server.active_defrag_cycle_min,CONFIG_DEFAULT_DEFRAG_CYCLE_MIN); + rewriteConfigNumericalOption(state,"active-defrag-cycle-max",server.active_defrag_cycle_max,CONFIG_DEFAULT_DEFRAG_CYCLE_MAX); + rewriteConfigYesNoOption(state,"appendonly",server.aof_state != AOF_OFF,0); + rewriteConfigStringOption(state,"appendfilename",server.aof_filename,CONFIG_DEFAULT_AOF_FILENAME); + rewriteConfigEnumOption(state,"appendfsync",server.aof_fsync,aof_fsync_enum,CONFIG_DEFAULT_AOF_FSYNC); + rewriteConfigYesNoOption(state,"no-appendfsync-on-rewrite",server.aof_no_fsync_on_rewrite,CONFIG_DEFAULT_AOF_NO_FSYNC_ON_REWRITE); + rewriteConfigNumericalOption(state,"auto-aof-rewrite-percentage",server.aof_rewrite_perc,AOF_REWRITE_PERC); + rewriteConfigBytesOption(state,"auto-aof-rewrite-min-size",server.aof_rewrite_min_size,AOF_REWRITE_MIN_SIZE); + rewriteConfigNumericalOption(state,"lua-time-limit",server.lua_time_limit,LUA_SCRIPT_TIME_LIMIT); + rewriteConfigYesNoOption(state,"cluster-enabled",server.cluster_enabled,0); + rewriteConfigStringOption(state,"cluster-config-file",server.cluster_configfile,CONFIG_DEFAULT_CLUSTER_CONFIG_FILE); + rewriteConfigYesNoOption(state,"cluster-require-full-coverage",server.cluster_require_full_coverage,CLUSTER_DEFAULT_REQUIRE_FULL_COVERAGE); + rewriteConfigNumericalOption(state,"cluster-node-timeout",server.cluster_node_timeout,CLUSTER_DEFAULT_NODE_TIMEOUT); + rewriteConfigNumericalOption(state,"cluster-migration-barrier",server.cluster_migration_barrier,CLUSTER_DEFAULT_MIGRATION_BARRIER); + rewriteConfigNumericalOption(state,"cluster-slave-validity-factor",server.cluster_slave_validity_factor,CLUSTER_DEFAULT_SLAVE_VALIDITY); + rewriteConfigNumericalOption(state,"slowlog-log-slower-than",server.slowlog_log_slower_than,CONFIG_DEFAULT_SLOWLOG_LOG_SLOWER_THAN); + rewriteConfigNumericalOption(state,"latency-monitor-threshold",server.latency_monitor_threshold,CONFIG_DEFAULT_LATENCY_MONITOR_THRESHOLD); + rewriteConfigNumericalOption(state,"slowlog-max-len",server.slowlog_max_len,CONFIG_DEFAULT_SLOWLOG_MAX_LEN); + rewriteConfigNotifykeyspaceeventsOption(state); + rewriteConfigNumericalOption(state,"hash-max-ziplist-entries",server.hash_max_ziplist_entries,OBJ_HASH_MAX_ZIPLIST_ENTRIES); + rewriteConfigNumericalOption(state,"hash-max-ziplist-value",server.hash_max_ziplist_value,OBJ_HASH_MAX_ZIPLIST_VALUE); + rewriteConfigNumericalOption(state,"list-max-ziplist-size",server.list_max_ziplist_size,OBJ_LIST_MAX_ZIPLIST_SIZE); + rewriteConfigNumericalOption(state,"list-compress-depth",server.list_compress_depth,OBJ_LIST_COMPRESS_DEPTH); + rewriteConfigNumericalOption(state,"set-max-intset-entries",server.set_max_intset_entries,OBJ_SET_MAX_INTSET_ENTRIES); + rewriteConfigNumericalOption(state,"zset-max-ziplist-entries",server.zset_max_ziplist_entries,OBJ_ZSET_MAX_ZIPLIST_ENTRIES); + rewriteConfigNumericalOption(state,"zset-max-ziplist-value",server.zset_max_ziplist_value,OBJ_ZSET_MAX_ZIPLIST_VALUE); + rewriteConfigNumericalOption(state,"hll-sparse-max-bytes",server.hll_sparse_max_bytes,CONFIG_DEFAULT_HLL_SPARSE_MAX_BYTES); + rewriteConfigYesNoOption(state,"activerehashing",server.activerehashing,CONFIG_DEFAULT_ACTIVE_REHASHING); + rewriteConfigYesNoOption(state,"activedefrag",server.active_defrag_enabled,CONFIG_DEFAULT_ACTIVE_DEFRAG); + rewriteConfigYesNoOption(state,"protected-mode",server.protected_mode,CONFIG_DEFAULT_PROTECTED_MODE); + rewriteConfigClientoutputbufferlimitOption(state); + rewriteConfigNumericalOption(state,"hz",server.hz,CONFIG_DEFAULT_HZ); + rewriteConfigYesNoOption(state,"aof-rewrite-incremental-fsync",server.aof_rewrite_incremental_fsync,CONFIG_DEFAULT_AOF_REWRITE_INCREMENTAL_FSYNC); + rewriteConfigYesNoOption(state,"aof-load-truncated",server.aof_load_truncated,CONFIG_DEFAULT_AOF_LOAD_TRUNCATED); + rewriteConfigYesNoOption(state,"aof-use-rdb-preamble",server.aof_use_rdb_preamble,CONFIG_DEFAULT_AOF_USE_RDB_PREAMBLE); + rewriteConfigEnumOption(state,"supervised",server.supervised_mode,supervised_mode_enum,SUPERVISED_NONE); + rewriteConfigYesNoOption(state,"lazyfree-lazy-eviction",server.lazyfree_lazy_eviction,CONFIG_DEFAULT_LAZYFREE_LAZY_EVICTION); + rewriteConfigYesNoOption(state,"lazyfree-lazy-expire",server.lazyfree_lazy_expire,CONFIG_DEFAULT_LAZYFREE_LAZY_EXPIRE); + rewriteConfigYesNoOption(state,"lazyfree-lazy-server-del",server.lazyfree_lazy_server_del,CONFIG_DEFAULT_LAZYFREE_LAZY_SERVER_DEL); + rewriteConfigYesNoOption(state,"slave-lazy-flush",server.repl_slave_lazy_flush,CONFIG_DEFAULT_SLAVE_LAZY_FLUSH); + + /* Rewrite Sentinel config if in Sentinel mode. */ + if (server.sentinel_mode) rewriteConfigSentinelOption(state); + + /* Step 3: remove all the orphaned lines in the old file, that is, lines + * that were used by a config option and are no longer used, like in case + * of multiple "save" options or duplicated options. */ + rewriteConfigRemoveOrphaned(state); + + /* Step 4: generate a new configuration file from the modified state + * and write it into the original file. */ + newcontent = rewriteConfigGetContentFromState(state); + retval = rewriteConfigOverwriteFile(server.configfile,newcontent); + + sdsfree(newcontent); + rewriteConfigReleaseState(state); + return retval; +} + +/*----------------------------------------------------------------------------- + * CONFIG command entry point + *----------------------------------------------------------------------------*/ + +void configCommand(client *c) { + /* Only allow CONFIG GET while loading. */ + if (server.loading && strcasecmp(c->argv[1]->ptr,"get")) { + addReplyError(c,"Only CONFIG GET is allowed during loading"); + return; + } + + if (!strcasecmp(c->argv[1]->ptr,"set")) { + if (c->argc != 4) goto badarity; + configSetCommand(c); + } else if (!strcasecmp(c->argv[1]->ptr,"get")) { + if (c->argc != 3) goto badarity; + configGetCommand(c); + } else if (!strcasecmp(c->argv[1]->ptr,"resetstat")) { + if (c->argc != 2) goto badarity; + resetServerStats(); + resetCommandTableStats(); + addReply(c,shared.ok); + } else if (!strcasecmp(c->argv[1]->ptr,"rewrite")) { + if (c->argc != 2) goto badarity; + if (server.configfile == NULL) { + addReplyError(c,"The server is running without a config file"); + return; + } + if (rewriteConfig(server.configfile) == -1) { + serverLog(LL_WARNING,"CONFIG REWRITE failed: %s", IF_WIN32(wsa_strerror(errno), strerror(errno))); + addReplyErrorFormat(c,"Rewriting config file: %s", IF_WIN32(wsa_strerror(errno), strerror(errno))); + } else { + serverLog(LL_WARNING,"CONFIG REWRITE executed with success."); + addReply(c,shared.ok); + } + } else { + addReplyError(c, + "CONFIG subcommand must be one of GET, SET, RESETSTAT, REWRITE"); + } + return; + +badarity: + addReplyErrorFormat(c,"Wrong number of arguments for CONFIG %s", + (char*) c->argv[1]->ptr); +} diff --git a/src/db.c b/src/db.c index 8f2ec8e3..e2b63ffc 100644 --- a/src/db.c +++ b/src/db.c @@ -41,6 +41,15 @@ * C-level DB API *----------------------------------------------------------------------------*/ +/* Update LFU when an object is accessed. + * Firstly, decrement the counter if the decrement time is reached. + * Then logarithmically increment the counter, and update the access time. */ +void updateLFU(robj *val) { + unsigned long counter = LFUDecrAndReturn(val); + counter = LFULogIncr(counter); + val->lru = (LFUGetTimeInMinutes()<<8) | counter; +} + /* Low level key lookup API, not actually called directly from commands * implementations that should instead rely on lookupKeyRead(), * lookupKeyWrite() and lookupKeyReadWithFlags(). */ @@ -57,9 +66,7 @@ robj *lookupKey(redisDb *db, robj *key, int flags) { !(flags & LOOKUP_NOTOUCH)) { if (server.maxmemory_policy & MAXMEMORY_FLAG_LFU) { - PORT_LONG ldt = val->lru >> 8; - PORT_LONG counter = LFULogIncr(val->lru & 255); - val->lru = (ldt << 8) | counter; + updateLFU(val); } else { val->lru = LRU_CLOCK(); } @@ -183,6 +190,9 @@ void dbOverwrite(redisDb *db, robj *key, robj *val) { int saved_lru = old->lru; dictReplace(db->dict, key->ptr, val); val->lru = saved_lru; + /* LFU should be not only copied but also updated + * when a key is overwritten. */ + updateLFU(val); } else { dictReplace(db->dict, key->ptr, val); } @@ -1091,6 +1101,25 @@ void propagateExpire(redisDb *db, robj *key, int lazy) { decrRefCount(argv[1]); } +/* This function is called when we are going to perform some operation + * in a given key, but such key may be already logically expired even if + * it still exists in the database. The main way this function is called + * is via lookupKey*() family of functions. + * + * The behavior of the function depends on the replication role of the + * instance, because slave instances do not expire keys, they wait + * for DELs from the master for consistency matters. However even + * slaves will try to have a coherent return value for the function, + * so that read commands executed in the slave side will be able to + * behave like if the key is expired even if still present (because the + * master has yet to propagate the DEL). + * + * In masters as a side effect of finding a key which is expired, such + * key will be evicted from the database. Also this may trigger the + * propagation of a DEL/UNLINK command in AOF / replication stream. + * + * The return value of the function is 0 if the key is still valid, + * otherwise the function returns 1 if the key is expired. */ int expireIfNeeded(redisDb *db, robj *key) { mstime_t when = getExpire(db,key); mstime_t now; @@ -1100,7 +1129,7 @@ int expireIfNeeded(redisDb *db, robj *key) { /* Don't expire anything while loading. It will be done later. */ if (server.loading) return 0; - /* If we are in the context of a Lua script, we claim that time is + /* If we are in the context of a Lua script, we pretend that time is * blocked to when the Lua script started. This way a key can expire * only the first time it is accessed and not in the middle of the * script execution, making propagation to slaves / AOF consistent. @@ -1148,11 +1177,13 @@ int *getKeysUsingCommandTable(struct redisCommand *cmd,robj **argv, int argc, in keys = zmalloc(sizeof(int)*((last - cmd->firstkey)+1)); for (j = cmd->firstkey; j <= last; j += cmd->keystep) { if (j >= argc) { - /* Modules command do not have dispatch time arity checks, so - * we need to handle the case where the user passed an invalid - * number of arguments here. In this case we return no keys - * and expect the module command to report an arity error. */ - if (cmd->flags & CMD_MODULE) { + /* Modules commands, and standard commands with a not fixed number + * of arugments (negative arity parameter) do not have dispatch + * time arity checks, so we need to handle the case where the user + * passed an invalid number of arguments here. In this case we + * return no keys and expect the command implementation to report + * an arity or syntax error. */ + if (cmd->flags & CMD_MODULE || cmd->arity < 0) { zfree(keys); *numkeys = 0; return NULL; diff --git a/src/db.c.bak b/src/db.c.bak new file mode 100644 index 00000000..273e06ae --- /dev/null +++ b/src/db.c.bak @@ -0,0 +1,1455 @@ +/* + * Copyright (c) 2009-2012, Salvatore Sanfilippo + * All rights reserved. + * + * Redistribution and use in source and binary forms, with or without + * modification, are permitted provided that the following conditions are met: + * + * * Redistributions of source code must retain the above copyright notice, + * this list of conditions and the following disclaimer. + * * Redistributions in binary form must reproduce the above copyright + * notice, this list of conditions and the following disclaimer in the + * documentation and/or other materials provided with the distribution. + * * Neither the name of Redis nor the names of its contributors may be used + * to endorse or promote products derived from this software without + * specific prior written permission. + * + * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" + * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE + * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE + * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE + * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR + * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF + * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS + * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN + * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) + * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE + * POSSIBILITY OF SUCH DAMAGE. + */ + +#include "server.h" +#ifdef _WIN32 +#include "Win32_Interop/Win32_QFork.h" +#endif +#include "cluster.h" +#include "atomicvar.h" + +#include +#include + +/*----------------------------------------------------------------------------- + * C-level DB API + *----------------------------------------------------------------------------*/ + +/* Update LFU when an object is accessed. + * Firstly, decrement the counter if the decrement time is reached. + * Then logarithmically increment the counter, and update the access time. */ +void updateLFU(robj *val) { + unsigned long counter = LFUDecrAndReturn(val); + counter = LFULogIncr(counter); + val->lru = (LFUGetTimeInMinutes()<<8) | counter; +} + +/* Low level key lookup API, not actually called directly from commands + * implementations that should instead rely on lookupKeyRead(), + * lookupKeyWrite() and lookupKeyReadWithFlags(). */ +robj *lookupKey(redisDb *db, robj *key, int flags) { + dictEntry *de = dictFind(db->dict,key->ptr); + if (de) { + robj *val = dictGetVal(de); + + /* Update the access time for the ageing algorithm. + * Don't do it if we have a saving child, as this will trigger + * a copy on write madness. */ + if (server.rdb_child_pid == -1 && + server.aof_child_pid == -1 && + !(flags & LOOKUP_NOTOUCH)) + { + if (server.maxmemory_policy & MAXMEMORY_FLAG_LFU) { + updateLFU(val); + } else { + val->lru = LRU_CLOCK(); + } + } + return val; + } else { + return NULL; + } +} + +/* Lookup a key for read operations, or return NULL if the key is not found + * in the specified DB. + * + * As a side effect of calling this function: + * 1. A key gets expired if it reached it's TTL. + * 2. The key last access time is updated. + * 3. The global keys hits/misses stats are updated (reported in INFO). + * + * This API should not be used when we write to the key after obtaining + * the object linked to the key, but only for read only operations. + * + * Flags change the behavior of this command: + * + * LOOKUP_NONE (or zero): no special flags are passed. + * LOOKUP_NOTOUCH: don't alter the last access time of the key. + * + * Note: this function also returns NULL is the key is logically expired + * but still existing, in case this is a slave, since this API is called only + * for read operations. Even if the key expiry is master-driven, we can + * correctly report a key is expired on slaves even if the master is lagging + * expiring our key via DELs in the replication link. */ +robj *lookupKeyReadWithFlags(redisDb *db, robj *key, int flags) { + robj *val; + + if (expireIfNeeded(db,key) == 1) { + /* Key expired. If we are in the context of a master, expireIfNeeded() + * returns 0 only when the key does not exist at all, so it's safe + * to return NULL ASAP. */ + if (server.masterhost == NULL) return NULL; + + /* However if we are in the context of a slave, expireIfNeeded() will + * not really try to expire the key, it only returns information + * about the "logical" status of the key: key expiring is up to the + * master in order to have a consistent view of master's data set. + * + * However, if the command caller is not the master, and as additional + * safety measure, the command invoked is a read-only command, we can + * safely return NULL here, and provide a more consistent behavior + * to clients accessign expired values in a read-only fashion, that + * will say the key as non exisitng. + * + * Notably this covers GETs when slaves are used to scale reads. */ + if (server.current_client && + server.current_client != server.master && + server.current_client->cmd && + server.current_client->cmd->flags & CMD_READONLY) + { + return NULL; + } + } + val = lookupKey(db,key,flags); + if (val == NULL) + server.stat_keyspace_misses++; + else + server.stat_keyspace_hits++; + return val; +} + +/* Like lookupKeyReadWithFlags(), but does not use any flag, which is the + * common case. */ +robj *lookupKeyRead(redisDb *db, robj *key) { + return lookupKeyReadWithFlags(db,key,LOOKUP_NONE); +} + +/* Lookup a key for write operations, and as a side effect, if needed, expires + * the key if its TTL is reached. + * + * Returns the linked value object if the key exists or NULL if the key + * does not exist in the specified DB. */ +robj *lookupKeyWrite(redisDb *db, robj *key) { + expireIfNeeded(db,key); + return lookupKey(db,key,LOOKUP_NONE); +} + +robj *lookupKeyReadOrReply(client *c, robj *key, robj *reply) { + robj *o = lookupKeyRead(c->db, key); + if (!o) addReply(c,reply); + return o; +} + +robj *lookupKeyWriteOrReply(client *c, robj *key, robj *reply) { + robj *o = lookupKeyWrite(c->db, key); + if (!o) addReply(c,reply); + return o; +} + +/* Add the key to the DB. It's up to the caller to increment the reference + * counter of the value if needed. + * + * The program is aborted if the key already exists. */ +void dbAdd(redisDb *db, robj *key, robj *val) { + sds copy = sdsdup(key->ptr); + int retval = dictAdd(db->dict, copy, val); + + serverAssertWithInfo(NULL,key,retval == DICT_OK); + if (val->type == OBJ_LIST) signalListAsReady(db, key); + if (server.cluster_enabled) slotToKeyAdd(key); + } + +/* Overwrite an existing key with a new value. Incrementing the reference + * count of the new value is up to the caller. + * This function does not modify the expire time of the existing key. + * + * The program is aborted if the key was not already present. */ +void dbOverwrite(redisDb *db, robj *key, robj *val) { + dictEntry *de = dictFind(db->dict,key->ptr); + + serverAssertWithInfo(NULL,key,de != NULL); + if (server.maxmemory_policy & MAXMEMORY_FLAG_LFU) { + robj *old = dictGetVal(de); + int saved_lru = old->lru; + dictReplace(db->dict, key->ptr, val); + val->lru = saved_lru; + /* LFU should be not only copied but also updated + * when a key is overwritten. */ + updateLFU(val); + } else { + dictReplace(db->dict, key->ptr, val); + } +} + +/* High level Set operation. This function can be used in order to set + * a key, whatever it was existing or not, to a new object. + * + * 1) The ref count of the value object is incremented. + * 2) clients WATCHing for the destination key notified. + * 3) The expire time of the key is reset (the key is made persistent). + * + * All the new keys in the database should be craeted via this interface. */ +void setKey(redisDb *db, robj *key, robj *val) { + if (lookupKeyWrite(db,key) == NULL) { + dbAdd(db,key,val); + } else { + dbOverwrite(db,key,val); + } + incrRefCount(val); + removeExpire(db,key); + signalModifiedKey(db,key); +} + +int dbExists(redisDb *db, robj *key) { + return dictFind(db->dict,key->ptr) != NULL; +} + +/* Return a random key, in form of a Redis object. + * If there are no keys, NULL is returned. + * + * The function makes sure to return keys not already expired. */ +robj *dbRandomKey(redisDb *db) { + dictEntry *de; + + while(1) { + sds key; + robj *keyobj; + + de = dictGetRandomKey(db->dict); + if (de == NULL) return NULL; + + key = dictGetKey(de); + keyobj = createStringObject(key,sdslen(key)); + if (dictFind(db->expires,key)) { + if (expireIfNeeded(db,keyobj)) { + decrRefCount(keyobj); + continue; /* search for another key. This expired. */ + } + } + return keyobj; + } +} + +/* Delete a key, value, and associated expiration entry if any, from the DB */ +int dbSyncDelete(redisDb *db, robj *key) { + /* Deleting an entry from the expires dict will not free the sds of + * the key, because it is shared with the main dictionary. */ + if (dictSize(db->expires) > 0) dictDelete(db->expires,key->ptr); + if (dictDelete(db->dict,key->ptr) == DICT_OK) { + if (server.cluster_enabled) slotToKeyDel(key); + return 1; + } else { + return 0; + } +} + +/* This is a wrapper whose behavior depends on the Redis lazy free + * configuration. Deletes the key synchronously or asynchronously. */ +int dbDelete(redisDb *db, robj *key) { + return server.lazyfree_lazy_server_del ? dbAsyncDelete(db,key) : + dbSyncDelete(db,key); +} + +/* Prepare the string object stored at 'key' to be modified destructively + * to implement commands like SETBIT or APPEND. + * + * An object is usually ready to be modified unless one of the two conditions + * are true: + * + * 1) The object 'o' is shared (refcount > 1), we don't want to affect + * other users. + * 2) The object encoding is not "RAW". + * + * If the object is found in one of the above conditions (or both) by the + * function, an unshared / not-encoded copy of the string object is stored + * at 'key' in the specified 'db'. Otherwise the object 'o' itself is + * returned. + * + * USAGE: + * + * The object 'o' is what the caller already obtained by looking up 'key' + * in 'db', the usage pattern looks like this: + * + * o = lookupKeyWrite(db,key); + * if (checkType(c,o,OBJ_STRING)) return; + * o = dbUnshareStringValue(db,key,o); + * + * At this point the caller is ready to modify the object, for example + * using an sdscat() call to append some data, or anything else. + */ +robj *dbUnshareStringValue(redisDb *db, robj *key, robj *o) { + serverAssert(o->type == OBJ_STRING); + if (o->refcount != 1 || o->encoding != OBJ_ENCODING_RAW) { + robj *decoded = getDecodedObject(o); + o = createRawStringObject(decoded->ptr, sdslen(decoded->ptr)); + decrRefCount(decoded); + dbOverwrite(db,key,o); + } + return o; +} + +/* Remove all keys from all the databases in a Redis server. + * If callback is given the function is called from time to time to + * signal that work is in progress. + * + * The dbnum can be -1 if all teh DBs should be flushed, or the specified + * DB number if we want to flush only a single Redis database number. + * + * Flags are be EMPTYDB_NO_FLAGS if no special flags are specified or + * EMPTYDB_ASYNC if we want the memory to be freed in a different thread + * and the function to return ASAP. + * + * On success the fuction returns the number of keys removed from the + * database(s). Otherwise -1 is returned in the specific case the + * DB number is out of range, and errno is set to EINVAL. */ +PORT_LONGLONG emptyDb(int dbnum, int flags, void(callback)(void*)) { + int j, async = (flags & EMPTYDB_ASYNC); + PORT_LONGLONG removed = 0; + + if (dbnum < -1 || dbnum >= server.dbnum) { + errno = EINVAL; + return -1; + } + + for (j = 0; j < server.dbnum; j++) { + if (dbnum != -1 && dbnum != j) continue; + removed += dictSize(server.db[j].dict); + if (async) { + emptyDbAsync(&server.db[j]); + } else { + dictEmpty(server.db[j].dict,callback); + dictEmpty(server.db[j].expires,callback); + } + } + if (server.cluster_enabled) { + if (async) { + slotToKeyFlushAsync(); + } else { + slotToKeyFlush(); + } + } + if (dbnum == -1) flushSlaveKeysWithExpireList(); + return removed; +} + +int selectDb(client *c, int id) { + if (id < 0 || id >= server.dbnum) + return C_ERR; + c->db = &server.db[id]; + return C_OK; +} + +/*----------------------------------------------------------------------------- + * Hooks for key space changes. + * + * Every time a key in the database is modified the function + * signalModifiedKey() is called. + * + * Every time a DB is flushed the function signalFlushDb() is called. + *----------------------------------------------------------------------------*/ + +void signalModifiedKey(redisDb *db, robj *key) { + touchWatchedKey(db,key); +} + +void signalFlushedDb(int dbid) { + touchWatchedKeysOnFlush(dbid); +} + +/*----------------------------------------------------------------------------- + * Type agnostic commands operating on the key space + *----------------------------------------------------------------------------*/ + +/* Return the set of flags to use for the emptyDb() call for FLUSHALL + * and FLUSHDB commands. + * + * Currently the command just attempts to parse the "ASYNC" option. It + * also checks if the command arity is wrong. + * + * On success C_OK is returned and the flags are stored in *flags, otherwise + * C_ERR is returned and the function sends an error to the client. */ +int getFlushCommandFlags(client *c, int *flags) { + /* Parse the optional ASYNC option. */ + if (c->argc > 1) { + if (c->argc > 2 || strcasecmp(c->argv[1]->ptr,"async")) { + addReply(c,shared.syntaxerr); + return C_ERR; + } + *flags = EMPTYDB_ASYNC; + } else { + *flags = EMPTYDB_NO_FLAGS; + } + return C_OK; +} + +/* FLUSHDB [ASYNC] + * + * Flushes the currently SELECTed Redis DB. */ +void flushdbCommand(client *c) { + int flags; + + if (getFlushCommandFlags(c,&flags) == C_ERR) return; + signalFlushedDb(c->db->id); + server.dirty += emptyDb(c->db->id,flags,NULL); + addReply(c,shared.ok); +} + +/* FLUSHALL [ASYNC] + * + * Flushes the whole server data set. */ +void flushallCommand(client *c) { + int flags; + + if (getFlushCommandFlags(c,&flags) == C_ERR) return; + signalFlushedDb(-1); + server.dirty += emptyDb(-1,flags,NULL); + addReply(c,shared.ok); + if (server.rdb_child_pid != -1) { +#ifdef _WIN32 + AbortForkOperation(); +#else + kill(server.rdb_child_pid,SIGUSR1); +#endif + rdbRemoveTempFile(server.rdb_child_pid); + } + if (server.saveparamslen > 0) { + /* Normally rdbSave() will reset dirty, but we don't want this here + * as otherwise FLUSHALL will not be replicated nor put into the AOF. */ + int saved_dirty = server.dirty; + rdbSaveInfo rsi, *rsiptr; + rsiptr = rdbPopulateSaveInfo(&rsi); + rdbSave(server.rdb_filename,rsiptr); + server.dirty = saved_dirty; + } + server.dirty++; +} + +/* This command implements DEL and LAZYDEL. */ +void delGenericCommand(client *c, int lazy) { + int numdel = 0, j; + + for (j = 1; j < c->argc; j++) { + expireIfNeeded(c->db,c->argv[j]); + int deleted = lazy ? dbAsyncDelete(c->db,c->argv[j]) : + dbSyncDelete(c->db,c->argv[j]); + if (deleted) { + signalModifiedKey(c->db,c->argv[j]); + notifyKeyspaceEvent(NOTIFY_GENERIC, + "del",c->argv[j],c->db->id); + server.dirty++; + numdel++; + } + } + addReplyLongLong(c,numdel); +} + +void delCommand(client *c) { + delGenericCommand(c,0); +} + +void unlinkCommand(client *c) { + delGenericCommand(c,1); +} + +/* EXISTS key1 key2 ... key_N. + * Return value is the number of keys existing. */ +void existsCommand(client *c) { + PORT_LONGLONG count = 0; + int j; + + for (j = 1; j < c->argc; j++) { + expireIfNeeded(c->db,c->argv[j]); + if (dbExists(c->db,c->argv[j])) count++; + } + addReplyLongLong(c,count); +} + +void selectCommand(client *c) { + PORT_LONG id; + + if (getLongFromObjectOrReply(c, c->argv[1], &id, + "invalid DB index") != C_OK) + return; + + if (server.cluster_enabled && id != 0) { + addReplyError(c,"SELECT is not allowed in cluster mode"); + return; + } + if (selectDb(c,(int)id) == C_ERR) { WIN_PORT_FIX /* cast (int) */ + addReplyError(c,"DB index is out of range"); + } else { + addReply(c,shared.ok); + } +} + +void randomkeyCommand(client *c) { + robj *key; + + if ((key = dbRandomKey(c->db)) == NULL) { + addReply(c,shared.nullbulk); + return; + } + + addReplyBulk(c,key); + decrRefCount(key); +} + +void keysCommand(client *c) { + dictIterator *di; + dictEntry *de; + sds pattern = c->argv[1]->ptr; + int plen = (int)sdslen(pattern), allkeys; + PORT_ULONG numkeys = 0; + void *replylen = addDeferredMultiBulkLength(c); + + di = dictGetSafeIterator(c->db->dict); + allkeys = (pattern[0] == '*' && pattern[1] == '\0'); + while((de = dictNext(di)) != NULL) { + sds key = dictGetKey(de); + robj *keyobj; + + if (allkeys || stringmatchlen(pattern,plen,key,(int)sdslen(key),0)) { + keyobj = createStringObject(key,sdslen(key)); + if (expireIfNeeded(c->db,keyobj) == 0) { + addReplyBulk(c,keyobj); + numkeys++; + } + decrRefCount(keyobj); + } + } + dictReleaseIterator(di); + setDeferredMultiBulkLength(c,replylen,numkeys); +} + +/* This callback is used by scanGenericCommand in order to collect elements + * returned by the dictionary iterator into a list. */ +void scanCallback(void *privdata, const dictEntry *de) { + void **pd = (void**) privdata; + list *keys = pd[0]; + robj *o = pd[1]; + robj *key, *val = NULL; + + if (o == NULL) { + sds sdskey = dictGetKey(de); + key = createStringObject(sdskey, sdslen(sdskey)); + } else if (o->type == OBJ_SET) { + sds keysds = dictGetKey(de); + key = createStringObject(keysds,sdslen(keysds)); + } else if (o->type == OBJ_HASH) { + sds sdskey = dictGetKey(de); + sds sdsval = dictGetVal(de); + key = createStringObject(sdskey,sdslen(sdskey)); + val = createStringObject(sdsval,sdslen(sdsval)); + } else if (o->type == OBJ_ZSET) { + sds sdskey = dictGetKey(de); + key = createStringObject(sdskey,sdslen(sdskey)); + val = createStringObjectFromLongDouble(*(double*)dictGetVal(de),0); + } else { + serverPanic("Type not handled in SCAN callback."); + } + + listAddNodeTail(keys, key); + if (val) listAddNodeTail(keys, val); +} + +/* Try to parse a SCAN cursor stored at object 'o': + * if the cursor is valid, store it as unsigned integer into *cursor and + * returns C_OK. Otherwise return C_ERR and send an error to the + * client. */ +int parseScanCursorOrReply(client *c, robj *o, PORT_ULONG *cursor) { + char *eptr; + + /* Use strtoul() because we need an *unsigned* PORT_LONG, so + * getLongLongFromObject() does not cover the whole cursor space. */ + errno = 0; + *cursor = strtoul(o->ptr, &eptr, 10); + if (isspace(((char*)o->ptr)[0]) || eptr[0] != '\0' || errno == ERANGE) + { + addReplyError(c, "invalid cursor"); + return C_ERR; + } + return C_OK; +} + +/* This command implements SCAN, HSCAN and SSCAN commands. + * If object 'o' is passed, then it must be a Hash or Set object, otherwise + * if 'o' is NULL the command will operate on the dictionary associated with + * the current database. + * + * When 'o' is not NULL the function assumes that the first argument in + * the client arguments vector is a key so it skips it before iterating + * in order to parse options. + * + * In the case of a Hash object the function returns both the field and value + * of every element on the Hash. */ +void scanGenericCommand(client *c, robj *o, PORT_ULONG cursor) { + int i, j; + list *keys = listCreate(); + listNode *node, *nextnode; + PORT_LONG count = 10; + sds pat = NULL; + int patlen = 0, use_pattern = 0; + dict *ht; + + /* Object must be NULL (to iterate keys names), or the type of the object + * must be Set, Sorted Set, or Hash. */ + serverAssert(o == NULL || o->type == OBJ_SET || o->type == OBJ_HASH || + o->type == OBJ_ZSET); + + /* Set i to the first option argument. The previous one is the cursor. */ + i = (o == NULL) ? 2 : 3; /* Skip the key argument if needed. */ + + /* Step 1: Parse options. */ + while (i < c->argc) { + j = c->argc - i; + if (!strcasecmp(c->argv[i]->ptr, "count") && j >= 2) { + if (getLongFromObjectOrReply(c, c->argv[i+1], &count, NULL) + != C_OK) + { + goto cleanup; + } + + if (count < 1) { + addReply(c,shared.syntaxerr); + goto cleanup; + } + + i += 2; + } else if (!strcasecmp(c->argv[i]->ptr, "match") && j >= 2) { + pat = c->argv[i+1]->ptr; + patlen = (int)sdslen(pat); WIN_PORT_FIX /* cast (int) */ + + /* The pattern always matches if it is exactly "*", so it is + * equivalent to disabling it. */ + use_pattern = !(pat[0] == '*' && patlen == 1); + + i += 2; + } else { + addReply(c,shared.syntaxerr); + goto cleanup; + } + } + + /* Step 2: Iterate the collection. + * + * Note that if the object is encoded with a ziplist, intset, or any other + * representation that is not a hash table, we are sure that it is also + * composed of a small number of elements. So to avoid taking state we + * just return everything inside the object in a single call, setting the + * cursor to zero to signal the end of the iteration. */ + + /* Handle the case of a hash table. */ + ht = NULL; + if (o == NULL) { + ht = c->db->dict; + } else if (o->type == OBJ_SET && o->encoding == OBJ_ENCODING_HT) { + ht = o->ptr; + } else if (o->type == OBJ_HASH && o->encoding == OBJ_ENCODING_HT) { + ht = o->ptr; + count *= 2; /* We return key / value for this type. */ + } else if (o->type == OBJ_ZSET && o->encoding == OBJ_ENCODING_SKIPLIST) { + zset *zs = o->ptr; + ht = zs->dict; + count *= 2; /* We return key / value for this type. */ + } + + if (ht) { + void *privdata[2]; + /* We set the max number of iterations to ten times the specified + * COUNT, so if the hash table is in a pathological state (very + * sparsely populated) we avoid to block too much time at the cost + * of returning no or very few elements. */ + PORT_LONG maxiterations = count*10; + + /* We pass two pointers to the callback: the list to which it will + * add new elements, and the object containing the dictionary so that + * it is possible to fetch more data in a type-dependent way. */ + privdata[0] = keys; + privdata[1] = o; + do { + cursor = dictScan(ht, cursor, scanCallback, NULL, privdata); + } while (cursor && + maxiterations-- && + listLength(keys) < (PORT_ULONG)count); + } else if (o->type == OBJ_SET) { + int pos = 0; + int64_t ll; + + while(intsetGet(o->ptr,pos++,&ll)) + listAddNodeTail(keys,createStringObjectFromLongLong(ll)); + cursor = 0; + } else if (o->type == OBJ_HASH || o->type == OBJ_ZSET) { + unsigned char *p = ziplistIndex(o->ptr,0); + unsigned char *vstr; + unsigned int vlen; + PORT_LONGLONG vll; + + while(p) { + ziplistGet(p,&vstr,&vlen,&vll); + listAddNodeTail(keys, + (vstr != NULL) ? createStringObject((char*)vstr,vlen) : + createStringObjectFromLongLong(vll)); + p = ziplistNext(o->ptr,p); + } + cursor = 0; + } else { + serverPanic("Not handled encoding in SCAN."); + } + + /* Step 3: Filter elements. */ + node = listFirst(keys); + while (node) { + robj *kobj = listNodeValue(node); + nextnode = listNextNode(node); + int filter = 0; + + /* Filter element if it does not match the pattern. */ + if (!filter && use_pattern) { + if (sdsEncodedObject(kobj)) { + if (!stringmatchlen(pat, patlen, kobj->ptr, (int)sdslen(kobj->ptr), 0)) WIN_PORT_FIX /* cast (int) */ + filter = 1; + } else { + char buf[LONG_STR_SIZE]; + int len; + + serverAssert(kobj->encoding == OBJ_ENCODING_INT); + len = ll2string(buf,sizeof(buf),(PORT_LONG)kobj->ptr); + if (!stringmatchlen(pat, patlen, buf, len, 0)) filter = 1; + } + } + + /* Filter element if it is an expired key. */ + if (!filter && o == NULL && expireIfNeeded(c->db, kobj)) filter = 1; + + /* Remove the element and its associted value if needed. */ + if (filter) { + decrRefCount(kobj); + listDelNode(keys, node); + } + + /* If this is a hash or a sorted set, we have a flat list of + * key-value elements, so if this element was filtered, remove the + * value, or skip it if it was not filtered: we only match keys. */ + if (o && (o->type == OBJ_ZSET || o->type == OBJ_HASH)) { + node = nextnode; + nextnode = listNextNode(node); + if (filter) { + kobj = listNodeValue(node); + decrRefCount(kobj); + listDelNode(keys, node); + } + } + node = nextnode; + } + + /* Step 4: Reply to the client. */ + addReplyMultiBulkLen(c, 2); + addReplyBulkLongLong(c,cursor); + + addReplyMultiBulkLen(c, listLength(keys)); + while ((node = listFirst(keys)) != NULL) { + robj *kobj = listNodeValue(node); + addReplyBulk(c, kobj); + decrRefCount(kobj); + listDelNode(keys, node); + } + +cleanup: + listSetFreeMethod(keys,decrRefCountVoid); + listRelease(keys); +} + +/* The SCAN command completely relies on scanGenericCommand. */ +void scanCommand(client *c) { + PORT_ULONG cursor; + if (parseScanCursorOrReply(c,c->argv[1],&cursor) == C_ERR) return; + scanGenericCommand(c,NULL,cursor); +} + +void dbsizeCommand(client *c) { + addReplyLongLong(c,dictSize(c->db->dict)); +} + +void lastsaveCommand(client *c) { + addReplyLongLong(c,server.lastsave); +} + +void typeCommand(client *c) { + robj *o; + char *type; + + o = lookupKeyReadWithFlags(c->db,c->argv[1],LOOKUP_NOTOUCH); + if (o == NULL) { + type = "none"; + } else { + switch(o->type) { + case OBJ_STRING: type = "string"; break; + case OBJ_LIST: type = "list"; break; + case OBJ_SET: type = "set"; break; + case OBJ_ZSET: type = "zset"; break; + case OBJ_HASH: type = "hash"; break; + case OBJ_MODULE: { + moduleValue *mv = o->ptr; + type = mv->type->name; + }; break; + default: type = "unknown"; break; + } + } + addReplyStatus(c,type); +} + +void shutdownCommand(client *c) { + int flags = 0; + + if (c->argc > 2) { + addReply(c,shared.syntaxerr); + return; + } else if (c->argc == 2) { + if (!strcasecmp(c->argv[1]->ptr,"nosave")) { + flags |= SHUTDOWN_NOSAVE; + } else if (!strcasecmp(c->argv[1]->ptr,"save")) { + flags |= SHUTDOWN_SAVE; + } else { + addReply(c,shared.syntaxerr); + return; + } + } + /* When SHUTDOWN is called while the server is loading a dataset in + * memory we need to make sure no attempt is performed to save + * the dataset on shutdown (otherwise it could overwrite the current DB + * with half-read data). + * + * Also when in Sentinel mode clear the SAVE flag and force NOSAVE. */ + if (server.loading || server.sentinel_mode) + flags = (flags & ~SHUTDOWN_SAVE) | SHUTDOWN_NOSAVE; + if (prepareForShutdown(flags) == C_OK) exit(0); + addReplyError(c,"Errors trying to SHUTDOWN. Check logs."); +} + +void renameGenericCommand(client *c, int nx) { + robj *o; + PORT_LONGLONG expire; + int samekey = 0; + + /* When source and dest key is the same, no operation is performed, + * if the key exists, however we still return an error on unexisting key. */ + if (sdscmp(c->argv[1]->ptr,c->argv[2]->ptr) == 0) samekey = 1; + + if ((o = lookupKeyWriteOrReply(c,c->argv[1],shared.nokeyerr)) == NULL) + return; + + if (samekey) { + addReply(c,nx ? shared.czero : shared.ok); + return; + } + + incrRefCount(o); + expire = getExpire(c->db,c->argv[1]); + if (lookupKeyWrite(c->db,c->argv[2]) != NULL) { + if (nx) { + decrRefCount(o); + addReply(c,shared.czero); + return; + } + /* Overwrite: delete the old key before creating the new one + * with the same name. */ + dbDelete(c->db,c->argv[2]); + } + dbAdd(c->db,c->argv[2],o); + if (expire != -1) setExpire(c,c->db,c->argv[2],expire); + dbDelete(c->db,c->argv[1]); + signalModifiedKey(c->db,c->argv[1]); + signalModifiedKey(c->db,c->argv[2]); + notifyKeyspaceEvent(NOTIFY_GENERIC,"rename_from", + c->argv[1],c->db->id); + notifyKeyspaceEvent(NOTIFY_GENERIC,"rename_to", + c->argv[2],c->db->id); + server.dirty++; + addReply(c,nx ? shared.cone : shared.ok); +} + +void renameCommand(client *c) { + renameGenericCommand(c,0); +} + +void renamenxCommand(client *c) { + renameGenericCommand(c,1); +} + +void moveCommand(client *c) { + robj *o; + redisDb *src, *dst; + int srcid; + PORT_LONGLONG dbid, expire; + + if (server.cluster_enabled) { + addReplyError(c,"MOVE is not allowed in cluster mode"); + return; + } + + /* Obtain source and target DB pointers */ + src = c->db; + srcid = c->db->id; + + if (getLongLongFromObject(c->argv[2],&dbid) == C_ERR || + dbid < INT_MIN || dbid > INT_MAX || + selectDb(c,(int)dbid) == C_ERR) WIN_PORT_FIX /* cast (int) */ + { + addReply(c,shared.outofrangeerr); + return; + } + dst = c->db; + selectDb(c,srcid); /* Back to the source DB */ + + /* If the user is moving using as target the same + * DB as the source DB it is probably an error. */ + if (src == dst) { + addReply(c,shared.sameobjecterr); + return; + } + + /* Check if the element exists and get a reference */ + o = lookupKeyWrite(c->db,c->argv[1]); + if (!o) { + addReply(c,shared.czero); + return; + } + expire = getExpire(c->db,c->argv[1]); + + /* Return zero if the key already exists in the target DB */ + if (lookupKeyWrite(dst,c->argv[1]) != NULL) { + addReply(c,shared.czero); + return; + } + dbAdd(dst,c->argv[1],o); + if (expire != -1) setExpire(c,dst,c->argv[1],expire); + incrRefCount(o); + + /* OK! key moved, free the entry in the source DB */ + dbDelete(src,c->argv[1]); + server.dirty++; + addReply(c,shared.cone); +} + +/* Helper function for dbSwapDatabases(): scans the list of keys that have + * one or more blocked clients for B[LR]POP or other list blocking commands + * and signal the keys are ready if they are lists. See the comment where + * the function is used for more info. */ +void scanDatabaseForReadyLists(redisDb *db) { + dictEntry *de; + dictIterator *di = dictGetSafeIterator(db->blocking_keys); + while((de = dictNext(di)) != NULL) { + robj *key = dictGetKey(de); + robj *value = lookupKey(db,key,LOOKUP_NOTOUCH); + if (value && value->type == OBJ_LIST) + signalListAsReady(db, key); + } + dictReleaseIterator(di); +} + +/* Swap two databases at runtime so that all clients will magically see + * the new database even if already connected. Note that the client + * structure c->db points to a given DB, so we need to be smarter and + * swap the underlying referenced structures, otherwise we would need + * to fix all the references to the Redis DB structure. + * + * Returns C_ERR if at least one of the DB ids are out of range, otherwise + * C_OK is returned. */ +int dbSwapDatabases(int id1, int id2) { + if (id1 < 0 || id1 >= server.dbnum || + id2 < 0 || id2 >= server.dbnum) return C_ERR; + if (id1 == id2) return C_OK; + redisDb aux = server.db[id1]; + redisDb *db1 = &server.db[id1], *db2 = &server.db[id2]; + + /* Swap hash tables. Note that we don't swap blocking_keys, + * ready_keys and watched_keys, since we want clients to + * remain in the same DB they were. */ + db1->dict = db2->dict; + db1->expires = db2->expires; + db1->avg_ttl = db2->avg_ttl; + + db2->dict = aux.dict; + db2->expires = aux.expires; + db2->avg_ttl = aux.avg_ttl; + + /* Now we need to handle clients blocked on lists: as an effect + * of swapping the two DBs, a client that was waiting for list + * X in a given DB, may now actually be unblocked if X happens + * to exist in the new version of the DB, after the swap. + * + * However normally we only do this check for efficiency reasons + * in dbAdd() when a list is created. So here we need to rescan + * the list of clients blocked on lists and signal lists as ready + * if needed. */ + scanDatabaseForReadyLists(db1); + scanDatabaseForReadyLists(db2); + return C_OK; +} + +/* SWAPDB db1 db2 */ +void swapdbCommand(client *c) { + PORT_LONG id1, id2; + + /* Not allowed in cluster mode: we have just DB 0 there. */ + if (server.cluster_enabled) { + addReplyError(c,"SWAPDB is not allowed in cluster mode"); + return; + } + + /* Get the two DBs indexes. */ + if (getLongFromObjectOrReply(c, c->argv[1], &id1, + "invalid first DB index") != C_OK) + return; + + if (getLongFromObjectOrReply(c, c->argv[2], &id2, + "invalid second DB index") != C_OK) + return; + + /* Swap... */ + if (dbSwapDatabases(id1,id2) == C_ERR) { + addReplyError(c,"DB index is out of range"); + return; + } else { + server.dirty++; + addReply(c,shared.ok); + } +} + +/*----------------------------------------------------------------------------- + * Expires API + *----------------------------------------------------------------------------*/ + +int removeExpire(redisDb *db, robj *key) { + /* An expire may only be removed if there is a corresponding entry in the + * main dict. Otherwise, the key will never be freed. */ + serverAssertWithInfo(NULL,key,dictFind(db->dict,key->ptr) != NULL); + return dictDelete(db->expires,key->ptr) == DICT_OK; +} + +/* Set an expire to the specified key. If the expire is set in the context + * of an user calling a command 'c' is the client, otherwise 'c' is set + * to NULL. The 'when' parameter is the absolute unix time in milliseconds + * after which the key will no longer be considered valid. */ +void setExpire(client *c, redisDb *db, robj *key, PORT_LONGLONG when) { + dictEntry *kde, *de; + + /* Reuse the sds from the main dict in the expire dict */ + kde = dictFind(db->dict,key->ptr); + serverAssertWithInfo(NULL,key,kde != NULL); + de = dictAddOrFind(db->expires,dictGetKey(kde)); + dictSetSignedIntegerVal(de,when); + + int writable_slave = server.masterhost && server.repl_slave_ro == 0; + if (c && writable_slave && !(c->flags & CLIENT_MASTER)) + rememberSlaveKeyWithExpire(db,key); +} + +/* Return the expire time of the specified key, or -1 if no expire + * is associated with this key (i.e. the key is non volatile) */ +PORT_LONGLONG getExpire(redisDb *db, robj *key) { + dictEntry *de; + + /* No expire? return ASAP */ + if (dictSize(db->expires) == 0 || + (de = dictFind(db->expires,key->ptr)) == NULL) return -1; + + /* The entry was found in the expire dict, this means it should also + * be present in the main dict (safety check). */ + serverAssertWithInfo(NULL,key,dictFind(db->dict,key->ptr) != NULL); + return dictGetSignedIntegerVal(de); +} + +/* Propagate expires into slaves and the AOF file. + * When a key expires in the master, a DEL operation for this key is sent + * to all the slaves and the AOF file if enabled. + * + * This way the key expiry is centralized in one place, and since both + * AOF and the master->slave link guarantee operation ordering, everything + * will be consistent even if we allow write operations against expiring + * keys. */ +void propagateExpire(redisDb *db, robj *key, int lazy) { + robj *argv[2]; + + argv[0] = lazy ? shared.unlink : shared.del; + argv[1] = key; + incrRefCount(argv[0]); + incrRefCount(argv[1]); + + if (server.aof_state != AOF_OFF) + feedAppendOnlyFile(server.delCommand,db->id,argv,2); + replicationFeedSlaves(server.slaves,db->id,argv,2); + + decrRefCount(argv[0]); + decrRefCount(argv[1]); +} + +int expireIfNeeded(redisDb *db, robj *key) { + mstime_t when = getExpire(db,key); + mstime_t now; + + if (when < 0) return 0; /* No expire for this key */ + + /* Don't expire anything while loading. It will be done later. */ + if (server.loading) return 0; + + /* If we are in the context of a Lua script, we claim that time is + * blocked to when the Lua script started. This way a key can expire + * only the first time it is accessed and not in the middle of the + * script execution, making propagation to slaves / AOF consistent. + * See issue #1525 on Github for more information. */ + now = server.lua_caller ? server.lua_time_start : mstime(); + + /* If we are running in the context of a slave, return ASAP: + * the slave key expiration is controlled by the master that will + * send us synthesized DEL operations for expired keys. + * + * Still we try to return the right information to the caller, + * that is, 0 if we think the key should be still valid, 1 if + * we think the key is expired at this time. */ + if (server.masterhost != NULL) return now > when; + + /* Return when this key has not expired */ + if (now <= when) return 0; + + /* Delete the key */ + server.stat_expiredkeys++; + propagateExpire(db,key,server.lazyfree_lazy_expire); + notifyKeyspaceEvent(NOTIFY_EXPIRED, + "expired",key,db->id); + return server.lazyfree_lazy_expire ? dbAsyncDelete(db,key) : + dbSyncDelete(db,key); +} + +/* ----------------------------------------------------------------------------- + * API to get key arguments from commands + * ---------------------------------------------------------------------------*/ + +/* The base case is to use the keys position as given in the command table + * (firstkey, lastkey, step). */ +int *getKeysUsingCommandTable(struct redisCommand *cmd,robj **argv, int argc, int *numkeys) { + int j, i = 0, last, *keys; + UNUSED(argv); + + if (cmd->firstkey == 0) { + *numkeys = 0; + return NULL; + } + + last = cmd->lastkey; + if (last < 0) last = argc+last; + keys = zmalloc(sizeof(int)*((last - cmd->firstkey)+1)); + for (j = cmd->firstkey; j <= last; j += cmd->keystep) { + if (j >= argc) { + /* Modules command do not have dispatch time arity checks, so + * we need to handle the case where the user passed an invalid + * number of arguments here. In this case we return no keys + * and expect the module command to report an arity error. */ + if (cmd->flags & CMD_MODULE) { + zfree(keys); + *numkeys = 0; + return NULL; + } else { + serverPanic("Redis built-in command declared keys positions not matching the arity requirements."); + } + } + keys[i++] = j; + } + *numkeys = i; + return keys; +} + +/* Return all the arguments that are keys in the command passed via argc / argv. + * + * The command returns the positions of all the key arguments inside the array, + * so the actual return value is an heap allocated array of integers. The + * length of the array is returned by reference into *numkeys. + * + * 'cmd' must be point to the corresponding entry into the redisCommand + * table, according to the command name in argv[0]. + * + * This function uses the command table if a command-specific helper function + * is not required, otherwise it calls the command-specific function. */ +int *getKeysFromCommand(struct redisCommand *cmd, robj **argv, int argc, int *numkeys) { + if (cmd->flags & CMD_MODULE_GETKEYS) { + return moduleGetCommandKeysViaAPI(cmd,argv,argc,numkeys); + } else if (!(cmd->flags & CMD_MODULE) && cmd->getkeys_proc) { + return cmd->getkeys_proc(cmd,argv,argc,numkeys); + } else { + return getKeysUsingCommandTable(cmd,argv,argc,numkeys); + } +} + +/* Free the result of getKeysFromCommand. */ +void getKeysFreeResult(int *result) { + zfree(result); +} + +/* Helper function to extract keys from following commands: + * ZUNIONSTORE ... + * ZINTERSTORE ... */ +int *zunionInterGetKeys(struct redisCommand *cmd, robj **argv, int argc, int *numkeys) { + int i, num, *keys; + UNUSED(cmd); + + num = atoi(argv[2]->ptr); + /* Sanity check. Don't return any key if the command is going to + * reply with syntax error. */ + if (num > (argc-3)) { + *numkeys = 0; + return NULL; + } + + /* Keys in z{union,inter}store come from two places: + * argv[1] = storage key, + * argv[3...n] = keys to intersect */ + keys = zmalloc(sizeof(int)*(num+1)); + + /* Add all key positions for argv[3...n] to keys[] */ + for (i = 0; i < num; i++) keys[i] = 3+i; + + /* Finally add the argv[1] key position (the storage key target). */ + keys[num] = 1; + *numkeys = num+1; /* Total keys = {union,inter} keys + storage key */ + return keys; +} + +/* Helper function to extract keys from the following commands: + * EVAL