本节简单介绍了PostgreSQL缓存管理(Buffer Manager)中的实现函数ReadBuffer_common->BufferAlloc,该函数是ReadBuffer的子过程.处理共享缓存的搜索。 ### 一、数据结构 **BufferDesc** 共享缓冲区的共享描述符(状态)数据 ``` /* * Flags for buffer descriptors * buffer描述器标记 * * Note: TAG_VALID essentially means that there is a buffer hashtable * entry associated with the buffer's tag. * 注意:TAG_VALID本质上意味着有一个与缓冲区的标记相关联的缓冲区散列表条目。 */ //buffer header锁定 #define BM_LOCKED (1U << 22) /* buffer header is locked */ //数据需要写入(标记为DIRTY) #define BM_DIRTY (1U << 23) /* data needs writing */ //数据是有效的 #define BM_VALID (1U << 24) /* data is valid */ //已分配buffer tag #define BM_TAG_VALID (1U << 25) /* tag is assigned */ //正在R/W #define BM_IO_IN_PROGRESS (1U << 26) /* read or write in progress */ //上一个I/O出现错误 #define BM_IO_ERROR (1U << 27) /* previous I/O failed */ //开始写则变DIRTY #define BM_JUST_DIRTIED (1U << 28) /* dirtied since write started */ //存在等待sole pin的其他进程 #define BM_PIN_COUNT_WAITER (1U << 29) /* have waiter for sole pin */ //checkpoint发生,必须刷到磁盘上 #define BM_CHECKPOINT_NEEDED (1U << 30) /* must write for checkpoint */ //持久化buffer(不是unlogged或者初始化fork) #define BM_PERMANENT (1U << 31) /* permanent buffer (not unlogged, * or init fork) */ /* * BufferDesc -- shared descriptor/state data for a single shared buffer. * BufferDesc -- 共享缓冲区的共享描述符(状态)数据 * * Note: Buffer header lock (BM_LOCKED flag) must be held to examine or change * the tag, state or wait_backend_pid fields. In general, buffer header lock * is a spinlock which is combined with flags, refcount and usagecount into * single atomic variable. This layout allow us to do some operations in a * single atomic operation, without actually acquiring and releasing spinlock; * for instance, increase or decrease refcount. buf_id field never changes * after initialization, so does not need locking. freeNext is protected by * the buffer_strategy_lock not buffer header lock. The LWLock can take care * of itself. The buffer header lock is *not* used to control access to the * data in the buffer! * 注意:必须持有Buffer header锁(BM_LOCKED标记)才能检查或修改tag/state/wait_backend_pid字段. * 通常来说,buffer header lock是spinlock,它与标记位/参考计数/使用计数组合到单个原子变量中. * 这个布局设计允许我们执行原子操作,而不需要实际获得或者释放spinlock(比如,增加或者减少参考计数). * buf_id字段在初始化后不会出现变化,因此不需要锁定. * freeNext通过buffer_strategy_lock锁而不是buffer header lock保护. * LWLock可以很好的处理自己的状态. * 务请注意的是:buffer header lock不用于控制buffer中的数据访问! * * It's assumed that nobody changes the state field while buffer header lock * is held. Thus buffer header lock holder can do complex updates of the * state variable in single write, simultaneously with lock release (cleaning * BM_LOCKED flag). On the other hand, updating of state without holding * buffer header lock is restricted to CAS, which insure that BM_LOCKED flag * is not set. Atomic increment/decrement, OR/AND etc. are not allowed. * 假定在持有buffer header lock的情况下,没有人改变状态字段. * 持有buffer header lock的进程可以执行在单个写操作中执行复杂的状态变量更新, * 同步的释放锁(清除BM_LOCKED标记). * 换句话说,如果没有持有buffer header lock的状态更新,会受限于CAS, * 这种情况下确保BM_LOCKED没有被设置. * 比如原子的增加/减少(AND/OR)等操作是不允许的. * * An exception is that if we have the buffer pinned, its tag can't change * underneath us, so we can examine the tag without locking the buffer header. * Also, in places we do one-time reads of the flags without bothering to * lock the buffer header; this is generally for situations where we don't * expect the flag bit being tested to be changing. * 一种例外情况是如果我们已有buffer pinned,该buffer的tag不能改变(在本进程之下), * 因此不需要锁定buffer header就可以检查tag了. * 同时,在执行一次性的flags读取时不需要锁定buffer header. * 这种情况通常用于我们不希望正在测试的flag bit将被改变. * * We can't physically remove items from a disk page if another backend has * the buffer pinned. Hence, a backend may need to wait for all other pins * to go away. This is signaled by storing its own PID into * wait_backend_pid and setting flag bit BM_PIN_COUNT_WAITER. At present, * there can be only one such waiter per buffer. * 如果其他进程有buffer pinned,那么进程不能物理的从磁盘页面中删除items. * 因此,后台进程需要等待其他pins清除.这可以通过存储它自己的PID到wait_backend_pid中, * 并设置标记位BM_PIN_COUNT_WAITER. * 目前,每个缓冲区只能由一个等待进程. * * We use this same struct for local buffer headers, but the locks are not * used and not all of the flag bits are useful either. To avoid unnecessary * overhead, manipulations of the state field should be done without actual * atomic operations (i.e. only pg_atomic_read_u32() and * pg_atomic_unlocked_write_u32()). * 本地缓冲头部使用同样的结构,但并不需要使用locks,而且并不是所有的标记位都使用. * 为了避免不必要的负载,状态域的维护不需要实际的原子操作 * (比如只有pg_atomic_read_u32() and pg_atomic_unlocked_write_u32()) * * Be careful to avoid increasing the size of the struct when adding or * reordering members. Keeping it below 64 bytes (the most common CPU * cache line size) is fairly important for performance. * 在增加或者记录成员变量时,小心避免增加结构体的大小. * 保持结构体大小在64字节内(通常的CPU缓存线大小)对于性能是非常重要的. */ typedef struct BufferDesc { //buffer tag BufferTag tag; /* ID of page contained in buffer */ //buffer索引编号(0开始) int buf_id; /* buffer's index number (from 0) */ /* state of the tag, containing flags, refcount and usagecount */ //tag状态,包括flags/refcount和usagecount pg_atomic_uint32 state; //pin-count等待进程ID int wait_backend_pid; /* backend PID of pin-count waiter */ //空闲链表链中下一个空闲的buffer int freeNext; /* link in freelist chain */ //缓冲区内容锁 LWLock content_lock; /* to lock access to buffer contents */ } BufferDesc; ``` **BufferTag** Buffer tag标记了buffer存储的是磁盘中哪个block ``` /* * Buffer tag identifies which disk block the buffer contains. * Buffer tag标记了buffer存储的是磁盘中哪个block * * Note: the BufferTag data must be sufficient to determine where to write the * block, without reference to pg_class or pg_tablespace entries. It's * possible that the backend flushing the buffer doesn't even believe the * relation is visible yet (its xact may have started before the xact that * created the rel). The storage manager must be able to cope anyway. * 注意:BufferTag必须足以确定如何写block而不需要参照pg_class或者pg_tablespace数据字典信息. * 有可能后台进程在刷新缓冲区的时候深圳不相信关系是可见的(事务可能在创建rel的事务之前). * 存储管理器必须可以处理这些事情. * * Note: if there's any pad bytes in the struct, INIT_BUFFERTAG will have * to be fixed to zero them, since this struct is used as a hash key. * 注意:如果在结构体中有填充的字节,INIT_BUFFERTAG必须将它们固定为零，因为这个结构体用作散列键. */ typedef struct buftag { //物理relation标识符 RelFileNode rnode; /* physical relation identifier */ ForkNumber forkNum; //相对于relation起始的块号 BlockNumber blockNum; /* blknum relative to begin of reln */ } BufferTag; ``` **SMgrRelation** smgr.c维护一个包含SMgrRelation对象的hash表,SMgrRelation对象本质上是缓存的文件句柄. ``` /* * smgr.c maintains a table of SMgrRelation objects, which are essentially * cached file handles. An SMgrRelation is created (if not already present) * by smgropen(), and destroyed by smgrclose(). Note that neither of these * operations imply I/O, they just create or destroy a hashtable entry. * (But smgrclose() may release associated resources, such as OS-level file * descriptors.) * smgr.c维护一个包含SMgrRelation对象的hash表,SMgrRelation对象本质上是缓存的文件句柄. * SMgrRelation对象(如非现成)通过smgropen()方法创建,通过smgrclose()方法销毁. * 注意:这些操作都不会执行I/O操作,只会创建或者销毁哈希表条目. * (但是smgrclose()方法可能会释放相关的资源,比如OS基本的文件描述符) * * An SMgrRelation may have an "owner", which is just a pointer to it from * somewhere else; smgr.c will clear this pointer if the SMgrRelation is * closed. We use this to avoid dangling pointers from relcache to smgr * without having to make the smgr explicitly aware of relcache. There * can't be more than one "owner" pointer per SMgrRelation, but that's * all we need. * SMgrRelation可能会有"宿主",这个宿主可能只是从某个地方指向它的指针而已; * 如SMgrRelationsmgr.c会清除该指针.这样做可以避免从relcache到smgr的悬空指针, * 而不必要让smgr显式的感知relcache(也就是隔离了smgr了relcache). * 每个SMgrRelation不能跟多个"owner"指针关联,但这就是我们所需要的. * * SMgrRelations that do not have an "owner" are considered to be transient, * and are deleted at end of transaction. * SMgrRelations如无owner指针,则被视为临时对象,在事务的最后被删除. */ typedef struct SMgrRelationData { /* rnode is the hashtable lookup key, so it must be first! */ //-------- rnode是哈希表的搜索键,因此在结构体的首位 //关系物理定义ID RelFileNodeBackend smgr_rnode; /* relation physical identifier */ /* pointer to owning pointer, or NULL if none */ //--------- 指向拥有的指针,如无则为NULL struct SMgrRelationData **smgr_owner; /* * These next three fields are not actually used or manipulated by smgr, * except that they are reset to InvalidBlockNumber upon a cache flush * event (in particular, upon truncation of the relation). Higher levels * store cached state here so that it will be reset when truncation * happens. In all three cases, InvalidBlockNumber means "unknown". * 接下来的3个字段实际上并不用于或者由smgr管理, * 除非这些表里在cache flush event发生时被重置为InvalidBlockNumber * (特别是在关系被截断时). * 在这里,更高层的存储缓存了状态因此在截断发生时会被重置. * 在这3种情况下,InvalidBlockNumber都意味着"unknown". */ //当前插入的目标bloc BlockNumber smgr_targblock; /* current insertion target block */ //最后已知的fsm fork大小 BlockNumber smgr_fsm_nblocks; /* last known size of fsm fork */ //最后已知的vm fork大小 BlockNumber smgr_vm_nblocks; /* last known size of vm fork */ /* additional public fields may someday exist here */ //------- 未来可能新增的公共域 /* * Fields below here are intended to be private to smgr.c and its * submodules. Do not touch them from elsewhere. * 下面的字段是smgr.c及其子模块私有的,不要从其他模块接触这些字段. */ //存储管理器选择器 int smgr_which; /* storage manager selector */ /* * for md.c; per-fork arrays of the number of open segments * (md_num_open_segs) and the segments themselves (md_seg_fds). * 用于md.c,打开段(md_num_open_segs)和段自身(md_seg_fds)的数组(每个fork一个) */ int md_num_open_segs[MAX_FORKNUM + 1]; struct _MdfdVec *md_seg_fds[MAX_FORKNUM + 1]; /* if unowned, list link in list of all unowned SMgrRelations */ //如没有宿主,未宿主的SMgrRelations链表的链表链接. struct SMgrRelationData *next_unowned_reln; } SMgrRelationData; typedef SMgrRelationData *SMgrRelation; ``` **RelFileNodeBackend** 组合relfilenode和后台进程ID,用于提供需要定位物理存储的所有信息. ``` /* * Augmenting a relfilenode with the backend ID provides all the information * we need to locate the physical storage. The backend ID is InvalidBackendId * for regular relations (those accessible to more than one backend), or the * owning backend's ID for backend-local relations. Backend-local relations * are always transient and removed in case of a database crash; they are * never WAL-logged or fsync'd. * 组合relfilenode和后台进程ID,用于提供需要定位物理存储的所有信息. * 对于普通的关系(可通过多个后台进程访问),后台进程ID是InvalidBackendId; * 如为临时表,则为自己的后台进程ID. * 临时表(backend-local relations)通常是临时存在的,在数据库崩溃时删除,无需WAL-logged或者fsync. */ typedef struct RelFileNodeBackend { RelFileNode node;//节点 BackendId backend;//后台进程 } RelFileNodeBackend; ``` ### 二、源码解读 BufferAlloc是ReadBuffer的子过程.处理共享缓存的搜索.如果已无buffer可用,则选择一个可替换的buffer并删除旧页面,但注意不要读入新页面. 该函数的主要处理逻辑如下: 1.初始化,根据Tag确定hash值和分区锁定ID 2.检查block是否已在buffer pool中 3.在缓冲区中找到该buffer(buf_id >= 0) 3.1获取buffer描述符并Pin buffer 3.2如PinBuffer返回F,则执行StartBufferIO,如该函数返回F,则设置标记\*foundPtr为F 3.3返回buf 4.在缓冲区中找不到该buffer(buf_id < 0) 4.1释放newPartitionLock 4.2执行循环,寻找合适的buffer 4.2.1确保在自旋锁尚未持有时,有一个空闲的refcount入口(条目) 4.2.2选择一个待淘汰的buffer 4.2.3拷贝buffer flags到oldFlags中 4.2.4Pin buffer,然后释放buffer自旋锁 4.2.5如buffer标记位BM_DIRTY,FlushBuffer 4.2.6如buffer标记为BM_TAG_VALID,计算原tag的hashcode和partition lock ID,并锁定新旧分区锁 否则需要新的分区,锁定新分区锁,重置原分区锁和原hash值 4.2.7尝试使用buffer新的tag构造hash表入口 4.2.8存在冲突(buf_id >= 0),在这里只需要像一开始处理的那样,视为已在缓冲池发现该buffer 4.2.9不存在冲突(buf_id < 0),锁定buffer header,如缓冲区没有变脏或者被pinned,则已找到buf,跳出循环 否则,解锁buffer header,删除hash表入口,释放锁,重新寻找buffer 4.3可以重新设置buffer tag,完成后解锁buffer header,删除原有的hash表入口,释放分区锁 4.4执行StartBufferIO,设置\*foundPtr标记 4.5返回buf ``` /* * BufferAlloc -- subroutine for ReadBuffer. Handles lookup of a shared * buffer. If no buffer exists already, selects a replacement * victim and evicts the old page, but does NOT read in new page. * BufferAlloc -- ReadBuffer的子过程.处理共享缓存的搜索. * 如果已无buffer可用,则选择一个可替换的buffer并删除旧页面,但注意不要读入新页面. * * "strategy" can be a buffer replacement strategy object, or NULL for * the default strategy. The selected buffer's usage_count is advanced when * using the default strategy, but otherwise possibly not (see PinBuffer). * "strategy"可以是缓存替换策略对象,如为默认策略,则为NULL. * 如使用默认读取策略,则选中的缓冲buffer的usage_count会加一,但也可能不会增加(详细参见PinBuffer). * * The returned buffer is pinned and is already marked as holding the * desired page. If it already did have the desired page, *foundPtr is * set true. Otherwise, *foundPtr is set false and the buffer is marked * as IO_IN_PROGRESS; ReadBuffer will now need to do I/O to fill it. * 返回的buffer已pinned并已标记为持有指定的页面. * 如果确实已持有指定的页面,*foundPtr设置为T. * 否则的话,*foundPtr设置为F,buffer标记为IO_IN_PROGRESS,ReadBuffer将会执行I/O操作. * * *foundPtr is actually redundant with the buffer's BM_VALID flag, but * we keep it for simplicity in ReadBuffer. * *foundPtr跟buffer的BM_VALID标记是重复的,但为了ReadBuffer中的简化,仍然保持这个参数. * * No locks are held either at entry or exit. * 在进入或者退出的时候,不需要持有任何的Locks. */ static BufferDesc * BufferAlloc(SMgrRelation smgr, char relpersistence, ForkNumber forkNum, BlockNumber blockNum, BufferAccessStrategy strategy, bool *foundPtr) { //请求block的ID BufferTag newTag; /* identity of requested block */ //newTag的Hash值 uint32 newHash; /* hash value for newTag */ //缓冲区分区锁 LWLock *newPartitionLock; /* buffer partition lock for it */ //选中缓冲区对应的上一个ID BufferTag oldTag; /* previous identity of selected buffer */ //oldTag的hash值 uint32 oldHash; /* hash value for oldTag */ //原缓冲区分区锁 LWLock *oldPartitionLock; /* buffer partition lock for it */ //原标记位 uint32 oldFlags; //buffer ID编号 int buf_id; //buffer描述符 BufferDesc *buf; //是否有效 bool valid; //buffer状态 uint32 buf_state; /* create a tag so we can lookup the buffer */ //创建一个tag,用于检索buffer INIT_BUFFERTAG(newTag, smgr->smgr_rnode.node, forkNum, blockNum); /* determine its hash code and partition lock ID */ //根据Tag确定hash值和分区锁定ID newHash = BufTableHashCode(&newTag); newPartitionLock = BufMappingPartitionLock(newHash); /* see if the block is in the buffer pool already */ //检查block是否已在buffer pool中 LWLockAcquire(newPartitionLock, LW_SHARED); buf_id = BufTableLookup(&newTag, newHash); if (buf_id >= 0) { //---- 在缓冲区中找到该buffer /* * Found it. Now, pin the buffer so no one can steal it from the * buffer pool, and check to see if the correct data has been loaded * into the buffer. * 找到了!现在pin缓冲区,确保没有进程可以从缓冲区中删除 * 检查正确的数据是否已装载到缓冲区中. */ buf = GetBufferDescriptor(buf_id); //Pin缓冲区 valid = PinBuffer(buf, strategy); /* Can release the mapping lock as soon as we've pinned it */ //一旦pinned,立即释放newPartitionLock LWLockRelease(newPartitionLock); //设置返回参数 *foundPtr = true; if (!valid) { //如无效 /* * We can only get here if (a) someone else is still reading in * the page, or (b) a previous read attempt failed. We have to * wait for any active read attempt to finish, and then set up our * own read attempt if the page is still not BM_VALID. * StartBufferIO does it all. * 程序执行到这里原因是(a)有其他进程仍然读入了该page,或者(b)上一次读取尝试失败. * 在这里必须等到其他活动的读取完成,然后在page状态仍然不是BM_VALID时设置读取尝试. * StartBufferIO过程执行这些工作. */ if (StartBufferIO(buf, true)) { /* * If we get here, previous attempts to read the buffer must * have failed ... but we shall bravely try again. */ //上一次尝试读取已然失败,这里还是需要勇敢的再试一次! *foundPtr = false;//设置为F } } //返回buf return buf; } /* * Didn't find it in the buffer pool. We'll have to initialize a new * buffer. Remember to unlock the mapping lock while doing the work. * 没有在缓冲池中发现该buffer. * 这时候不得不初始化一个buffer. * 记住:在执行工作的时候,记得首先解锁mapping lock. */ LWLockRelease(newPartitionLock); /* Loop here in case we have to try another victim buffer */ //循环,寻找合适的buffer for (;;) { /* * Ensure, while the spinlock's not yet held, that there's a free * refcount entry. * 确保在自旋锁尚未持有时,有一个空闲的refcount入口(条目). */ ReservePrivateRefCountEntry(); /* * Select a victim buffer. The buffer is returned with its header * spinlock still held! * 选择一个待淘汰的buffer. * 返回的buffer,仍然持有其header的自旋锁. */ buf = StrategyGetBuffer(strategy, &buf_state); Assert(BUF_STATE_GET_REFCOUNT(buf_state) == 0); /* Must copy buffer flags while we still hold the spinlock */ //在仍持有自旋锁的情况下必须拷贝buffer flags oldFlags = buf_state & BUF_FLAG_MASK; /* Pin the buffer and then release the buffer spinlock */ //Pin buffer,然后释放buffer自旋锁 PinBuffer_Locked(buf); /* * If the buffer was dirty, try to write it out. There is a race * condition here, in that someone might dirty it after we released it * above, or even while we are writing it out (since our share-lock * won't prevent hint-bit updates). We will recheck the dirty bit * after re-locking the buffer header. * 如果buffer已脏,尝试刷新到磁盘上. * 这里有一个竞争条件,那就是某些进程可能在我们在上面释放它(或者甚至在我们正在刷新时)之后使该缓冲区变脏. * 在再次锁定buffer header后,我们会重新检查相应的dirty标记位. */ if (oldFlags & BM_DIRTY) { /* * We need a share-lock on the buffer contents to write it out * (else we might write invalid data, eg because someone else is * compacting the page contents while we write). We must use a * conditional lock acquisition here to avoid deadlock. Even * though the buffer was not pinned (and therefore surely not * locked) when StrategyGetBuffer returned it, someone else could * have pinned and exclusive-locked it by the time we get here. If * we try to get the lock unconditionally, we'd block waiting for * them; if they later block waiting for us, deadlock ensues. * (This has been observed to happen when two backends are both * trying to split btree index pages, and the second one just * happens to be trying to split the page the first one got from * StrategyGetBuffer.) * 需要持有buffer内容的共享锁来刷出该缓冲区. * (否则的话,我们可能会写入无效的数据,原因比如是其他进程在我们写入时压缩page). * 在这里,必须使用条件锁来避免死锁. * 在StrategyGetBuffer返回时虽然buffer尚未pinned, * 其他进程可能已经pinned该buffer并且同时已持有独占锁. * 如果我们尝试无条件的锁定,那么因为等待而阻塞.其他进程稍后又会等待本进程,那么死锁就会发生. * (在实际中,两个后台进程在尝试分裂B树索引pages, * 而第二个正好尝试分裂第一个进程通过StrategyGetBuffer获取的page时,会发生这种情况). */ if (LWLockConditionalAcquire(BufferDescriptorGetContentLock(buf), LW_SHARED)) { //---- 执行有条件锁定请求(buffer内容共享锁) /* * If using a nondefault strategy, and writing the buffer * would require a WAL flush, let the strategy decide whether * to go ahead and write/reuse the buffer or to choose another * victim. We need lock to inspect the page LSN, so this * can't be done inside StrategyGetBuffer. * 如使用非默认的策略,则写缓冲会请求WAL flush,让策略确定如何继续以及写入/重用 * 缓冲或者选择另外一个待淘汰的buffer. * 我们需要锁定,检查page的LSN,因此不能在StrategyGetBuffer中完成. */ if (strategy != NULL) { //非默认策略 XLogRecPtr lsn; /* Read the LSN while holding buffer header lock */ //在持有buffer header lock时读取LSN buf_state = LockBufHdr(buf); lsn = BufferGetLSN(buf); UnlockBufHdr(buf, buf_state); if (XLogNeedsFlush(lsn) && StrategyRejectBuffer(strategy, buf)) { //需要flush WAL并且StrategyRejectBuffer /* Drop lock/pin and loop around for another buffer */ //清除lock/pin并循环到另外一个buffer LWLockRelease(BufferDescriptorGetContentLock(buf)); UnpinBuffer(buf, true); continue; } } /* OK, do the I/O */ //现在可以执行I/O了 TRACE_POSTGRESQL_BUFFER_WRITE_DIRTY_START(forkNum, blockNum, smgr->smgr_rnode.node.spcNode, smgr->smgr_rnode.node.dbNode, smgr->smgr_rnode.node.relNode); FlushBuffer(buf, NULL); LWLockRelease(BufferDescriptorGetContentLock(buf)); ScheduleBufferTagForWriteback(&BackendWritebackContext, &buf->tag); TRACE_POSTGRESQL_BUFFER_WRITE_DIRTY_DONE(forkNum, blockNum, smgr->smgr_rnode.node.spcNode, smgr->smgr_rnode.node.dbNode, smgr->smgr_rnode.node.relNode); } else { /* * Someone else has locked the buffer, so give it up and loop * back to get another one. * 其他进程已经锁定了buffer,放弃,获取另外一个 */ UnpinBuffer(buf, true); continue; } } /* * To change the association of a valid buffer, we'll need to have * exclusive lock on both the old and new mapping partitions. * 修改有效缓冲区的相关性,需要在原有和新的映射分区上持有独占锁 */ if (oldFlags & BM_TAG_VALID) { //----------- buffer标记为BM_TAG_VALID /* * Need to compute the old tag's hashcode and partition lock ID. * XXX is it worth storing the hashcode in BufferDesc so we need * not recompute it here? Probably not. * 需要计算原tag的hashcode和partition lock ID. * 这里是否值得存储hashcode在BufferDesc中而无需再次计算?可能不值得. */ oldTag = buf->tag; oldHash = BufTableHashCode(&oldTag); oldPartitionLock = BufMappingPartitionLock(oldHash); /* * Must lock the lower-numbered partition first to avoid * deadlocks. * 必须首先锁定更低一级编号的分区以避免死锁 */ if (oldPartitionLock < newPartitionLock) { //按顺序锁定 LWLockAcquire(oldPartitionLock, LW_EXCLUSIVE); LWLockAcquire(newPartitionLock, LW_EXCLUSIVE); } else if (oldPartitionLock > newPartitionLock) { //按顺序锁定 LWLockAcquire(newPartitionLock, LW_EXCLUSIVE); LWLockAcquire(oldPartitionLock, LW_EXCLUSIVE); } else { /* only one partition, only one lock */ //只有一个分区,只需要一个锁 LWLockAcquire(newPartitionLock, LW_EXCLUSIVE); } } else { //----------- buffer未标记为BM_TAG_VALID /* if it wasn't valid, we need only the new partition */ //buffer无效,需要新的分区 LWLockAcquire(newPartitionLock, LW_EXCLUSIVE); /* remember we have no old-partition lock or tag */ //不需要原有分区的锁&tag oldPartitionLock = NULL; /* this just keeps the compiler quiet about uninit variables */ //这行代码的目的是让编译器"闭嘴" oldHash = 0; } /* * Try to make a hashtable entry for the buffer under its new tag. * This could fail because while we were writing someone else * allocated another buffer for the same block we want to read in. * Note that we have not yet removed the hashtable entry for the old * tag. * 尝试使用buffer新的tag构造hash表入口. * 这可能会失败,因为在我们写入时其他进程可能已为我们希望读入的同一个block分配了另外一个buffer. * 注意我们还没有删除原有tag的hash表入口. */ buf_id = BufTableInsert(&newTag, newHash, buf->buf_id); if (buf_id >= 0) { /* * Got a collision. Someone has already done what we were about to * do. We'll just handle this as if it were found in the buffer * pool in the first place. First, give up the buffer we were * planning to use. * 存在冲突.某个进程已完成了我们准备做的事情. * 在这里只需要像一开始处理的那样,视为已在缓冲池发现该buffer. * 首先,放弃计划使用的buffer. */ UnpinBuffer(buf, true); /* Can give up that buffer's mapping partition lock now */ //放弃原有的partition lock if (oldPartitionLock != NULL && oldPartitionLock != newPartitionLock) LWLockRelease(oldPartitionLock); /* remaining code should match code at top of routine */ //剩余的代码应匹配上面的处理过程 //详细参见以上代码注释 buf = GetBufferDescriptor(buf_id); valid = PinBuffer(buf, strategy); /* Can release the mapping lock as soon as we've pinned it */ //是否新partition lock LWLockRelease(newPartitionLock); //设置标记 *foundPtr = true; if (!valid) { /* * We can only get here if (a) someone else is still reading * in the page, or (b) a previous read attempt failed. We * have to wait for any active read attempt to finish, and * then set up our own read attempt if the page is still not * BM_VALID. StartBufferIO does it all. */ if (StartBufferIO(buf, true)) { /* * If we get here, previous attempts to read the buffer * must have failed ... but we shall bravely try again. */ *foundPtr = false; } } return buf; } /* * Need to lock the buffer header too in order to change its tag. * 需要锁定缓冲头部,目的是修改tag */ buf_state = LockBufHdr(buf); /* * Somebody could have pinned or re-dirtied the buffer while we were * doing the I/O and making the new hashtable entry. If so, we can't * recycle this buffer; we must undo everything we've done and start * over with a new victim buffer. * 在我们执行I/O和标记新的hash表入口时,某些进程可能已经pinned或者重新弄脏了buffer. * 如出现这样的情况,不能回收该缓冲区;必须回滚我们所做的所有事情,并重新寻找新的待淘汰的缓冲区. */ oldFlags = buf_state & BUF_FLAG_MASK; if (BUF_STATE_GET_REFCOUNT(buf_state) == 1 && !(oldFlags & BM_DIRTY)) //已经OK了 break; //解锁buffer header UnlockBufHdr(buf, buf_state); //删除hash表入口 BufTableDelete(&newTag, newHash); //释放锁 if (oldPartitionLock != NULL && oldPartitionLock != newPartitionLock) LWLockRelease(oldPartitionLock); LWLockRelease(newPartitionLock); UnpinBuffer(buf, true); //重新寻找buffer } /* * Okay, it's finally safe to rename the buffer. * 现在终于可以安全的给buffer重命名了 * * Clearing BM_VALID here is necessary, clearing the dirtybits is just * paranoia. We also reset the usage_count since any recency of use of * the old content is no longer relevant. (The usage_count starts out at * 1 so that the buffer can survive one clock-sweep pass.) * 如需要,清除BM_VALID标记,清除脏标记位. * 我们还需要重置usage_count，因为使用旧内容的recency不再相关. * (usage_count从1开始,因此buffer可以在一个时钟周期经过后仍能存活) * * Make sure BM_PERMANENT is set for buffers that must be written at every * checkpoint. Unlogged buffers only need to be written at shutdown * checkpoints, except for their "init" forks, which need to be treated * just like permanent relations. * 确保标记为BM_PERMANENT的buffer必须在每次checkpoint时刷到磁盘上. * Unlogged缓冲只需要在shutdown checkpoint时才需要写入,除非它们"init" forks, * 这些操作需要类似持久化关系一样处理. */ buf->tag = newTag; buf_state &= ~(BM_VALID | BM_DIRTY | BM_JUST_DIRTIED | BM_CHECKPOINT_NEEDED | BM_IO_ERROR | BM_PERMANENT | BUF_USAGECOUNT_MASK); if (relpersistence == RELPERSISTENCE_PERMANENT || forkNum == INIT_FORKNUM) buf_state |= BM_TAG_VALID | BM_PERMANENT | BUF_USAGECOUNT_ONE; else buf_state |= BM_TAG_VALID | BUF_USAGECOUNT_ONE; UnlockBufHdr(buf, buf_state); if (oldPartitionLock != NULL) { BufTableDelete(&oldTag, oldHash); if (oldPartitionLock != newPartitionLock) LWLockRelease(oldPartitionLock); } LWLockRelease(newPartitionLock); /* * Buffer contents are currently invalid. Try to get the io_in_progress * lock. If StartBufferIO returns false, then someone else managed to * read it before we did, so there's nothing left for BufferAlloc() to do. * 缓冲区内存已无效. * 尝试获取io_in_progress lock.如StartBufferIO返回F,意味着其他进程已在我们完成前读取该缓冲区, * 因此对于BufferAlloc()来说,已无事可做. */ if (StartBufferIO(buf, true)) *foundPtr = false; else *foundPtr = true; return buf; } ``` ### 三、跟踪分析 测试脚本,查询数据表: ``` 10:01:54 (xdb@[local]:5432)testdb=# select * from t1 limit 10; ``` 启动gdb,设置断点 ``` (gdb) b BufferAlloc Breakpoint 1 at 0x8778ad: file bufmgr.c, line 1005. (gdb) c Continuing. Breakpoint 1, BufferAlloc (smgr=0x2267430, relpersistence=112 'p', forkNum=MAIN_FORKNUM, blockNum=0, strategy=0x0, foundPtr=0x7ffcc97fb4f3) at bufmgr.c:1005 1005 INIT_BUFFERTAG(newTag, smgr->smgr_rnode.node, forkNum, blockNum); (gdb) ``` 输入参数 smgr-SMgrRelationData结构体指针 relpersistence-关系是否持久化 forkNum-fork类型,MAIN_FORKNUM对应数据文件,还有fsm/vm文件 blockNum-块号 strategy-buffer访问策略,为NULL \*foundPtr-输出参数 ``` (gdb) p *smgr $1 = {smgr_rnode = {node = {spcNode = 1663, dbNode = 16402, relNode = 51439}, backend = -1}, smgr_owner = 0x7f86133f3778, smgr_targblock = 4294967295, smgr_fsm_nblocks = 4294967295, smgr_vm_nblocks = 4294967295, smgr_which = 0, md_num_open_segs = {0, 0, 0, 0}, md_seg_fds = {0x0, 0x0, 0x0, 0x0}, next_unowned_reln = 0x0} (gdb) p *smgr->smgr_owner $2 = (struct SMgrRelationData *) 0x2267430 (gdb) p **smgr->smgr_owner $3 = {smgr_rnode = {node = {spcNode = 1663, dbNode = 16402, relNode = 51439}, backend = -1}, smgr_owner = 0x7f86133f3778, smgr_targblock = 4294967295, smgr_fsm_nblocks = 4294967295, smgr_vm_nblocks = 4294967295, smgr_which = 0, md_num_open_segs = {0, 0, 0, 0}, md_seg_fds = {0x0, 0x0, 0x0, 0x0}, next_unowned_reln = 0x0} (gdb) ``` 1.初始化,根据Tag确定hash值和分区锁定ID ``` (gdb) n 1008 newHash = BufTableHashCode(&newTag); (gdb) p newTag $4 = {rnode = {spcNode = 1663, dbNode = 16402, relNode = 51439}, forkNum = MAIN_FORKNUM, blockNum = 0} (gdb) n 1009 newPartitionLock = BufMappingPartitionLock(newHash); (gdb) 1012 LWLockAcquire(newPartitionLock, LW_SHARED); (gdb) 1013 buf_id = BufTableLookup(&newTag, newHash); (gdb) p newHash $5 = 1398580903 (gdb) p newPartitionLock $6 = (LWLock *) 0x7f85e5db9600 (gdb) p *newPartitionLock $7 = {tranche = 59, state = {value = 536870913}, waiters = {head = 2147483647, tail = 2147483647}} (gdb) ``` 2.检查block是否已在buffer pool中 ``` (gdb) n 1014 if (buf_id >= 0) (gdb) p buf_id $8 = -1 ``` 4.在缓冲区中找不到该buffer(buf_id < 0) 4.1释放newPartitionLock 4.2执行循环,寻找合适的buffer 4.2.1确保在自旋锁尚未持有时,有一个空闲的refcount入口(条目) ---> ReservePrivateRefCountEntry ``` (gdb) n 1056 LWLockRelease(newPartitionLock); (gdb) 1065 ReservePrivateRefCountEntry(); (gdb) ``` 4.2.2选择一个待淘汰的buffer ``` (gdb) n 1071 buf = StrategyGetBuffer(strategy, &buf_state); (gdb) n 1073 Assert(BUF_STATE_GET_REFCOUNT(buf_state) == 0); (gdb) p buf $9 = (BufferDesc *) 0x7f85e705fd80 (gdb) p *buf $10 = {tag = {rnode = {spcNode = 0, dbNode = 0, relNode = 0}, forkNum = InvalidForkNumber, blockNum = 4294967295}, buf_id = 104, state = {value = 4194304}, wait_backend_pid = 0, freeNext = -2, content_lock = {tranche = 54, state = { value = 536870912}, waiters = {head = 2147483647, tail = 2147483647}}} (gdb) ``` 4.2.3拷贝buffer flags到oldFlags中 ``` (gdb) n 1076 oldFlags = buf_state & BUF_FLAG_MASK; (gdb) ``` 4.2.4Pin buffer,然后释放buffer自旋锁 ``` (gdb) 1079 PinBuffer_Locked(buf); (gdb) ``` 4.2.5如buffer标记位BM_DIRTY,FlushBuffer ``` 1088 if (oldFlags & BM_DIRTY) (gdb) ``` 4.2.6如buffer标记为BM_TAG_VALID,计算原tag的hashcode和partition lock ID,并锁定新旧分区锁 否则需要新的分区,锁定新分区锁,重置原分区锁和原hash值 ``` (gdb) 1166 if (oldFlags & BM_TAG_VALID) (gdb) 1200 LWLockAcquire(newPartitionLock, LW_EXCLUSIVE); (gdb) 1202 oldPartitionLock = NULL; (gdb) 1204 oldHash = 0; (gdb) p oldFlags $11 = 4194304 (gdb) ``` 4.2.7尝试使用buffer新的tag构造hash表入口 ``` (gdb) 1214 buf_id = BufTableInsert(&newTag, newHash, buf->buf_id); (gdb) n 1216 if (buf_id >= 0) (gdb) p buf_id $12 = -1 (gdb) ``` 4.2.9不存在冲突(buf_id < 0),锁定buffer header,如缓冲区没有变脏或者被pinned,则已找到buf,跳出循环 否则,解锁buffer header,删除hash表入口,释放锁,重新寻找buffer ``` (gdb) n 1267 buf_state = LockBufHdr(buf); (gdb) 1275 oldFlags = buf_state & BUF_FLAG_MASK; (gdb) 1276 if (BUF_STATE_GET_REFCOUNT(buf_state) == 1 && !(oldFlags & BM_DIRTY)) (gdb) 1277 break; (gdb) ``` 4.3可以重新设置buffer tag,完成后解锁buffer header,删除原有的hash表入口,释放分区锁 ``` 1301 buf->tag = newTag; (gdb) 1302 buf_state &= ~(BM_VALID | BM_DIRTY | BM_JUST_DIRTIED | (gdb) 1305 if (relpersistence == RELPERSISTENCE_PERMANENT || forkNum == INIT_FORKNUM) (gdb) 1306 buf_state |= BM_TAG_VALID | BM_PERMANENT | BUF_USAGECOUNT_ONE; (gdb) 1310 UnlockBufHdr(buf, buf_state); (gdb) 1312 if (oldPartitionLock != NULL) (gdb) 1319 LWLockRelease(newPartitionLock); (gdb) p *buf $13 = {tag = {rnode = {spcNode = 1663, dbNode = 16402, relNode = 51439}, forkNum = MAIN_FORKNUM, blockNum = 0}, buf_id = 104, state = {value = 2181300225}, wait_backend_pid = 0, freeNext = -2, content_lock = {tranche = 54, state = { value = 536870912}, waiters = {head = 2147483647, tail = 2147483647}}} (gdb) ``` 4.4执行StartBufferIO,设置\*foundPtr标记 ``` (gdb) 1326 if (StartBufferIO(buf, true)) (gdb) n 1327 *foundPtr = false; (gdb) ``` 4.5返回buf ``` (gdb) 1331 return buf; (gdb) 1332 } (gdb) ``` 执行完成 ``` (gdb) ReadBuffer_common (smgr=0x2267430, relpersistence=112 'p', forkNum=MAIN_FORKNUM, blockNum=0, mode=RBM_NORMAL, strategy=0x0, hit=0x7ffcc97fb5eb) at bufmgr.c:747 747 if (found) (gdb) 750 pgBufferUsage.shared_blks_read++; (gdb) ``` DONE! ### 四、参考资料 [PG Source Code](https://doxygen.postgresql.org)