diff options
Diffstat (limited to 'src/backend/storage/buffer/buf_init.c')
| -rw-r--r-- | src/backend/storage/buffer/buf_init.c | 407 |
1 files changed, 210 insertions, 197 deletions
diff --git a/src/backend/storage/buffer/buf_init.c b/src/backend/storage/buffer/buf_init.c index 20f8195d1e9..4ce064d6713 100644 --- a/src/backend/storage/buffer/buf_init.c +++ b/src/backend/storage/buffer/buf_init.c @@ -1,13 +1,13 @@ /*------------------------------------------------------------------------- * * buf_init.c-- - * buffer manager initialization routines + * buffer manager initialization routines * * Copyright (c) 1994, Regents of the University of California * * * IDENTIFICATION - * $Header: /cvsroot/pgsql/src/backend/storage/buffer/buf_init.c,v 1.10 1997/07/28 00:54:33 momjian Exp $ + * $Header: /cvsroot/pgsql/src/backend/storage/buffer/buf_init.c,v 1.11 1997/09/07 04:48:15 momjian Exp $ * *------------------------------------------------------------------------- */ @@ -35,98 +35,103 @@ #include "utils/dynahash.h" #include "utils/hsearch.h" #include "utils/memutils.h" -#include "executor/execdebug.h" /* for NDirectFileRead */ +#include "executor/execdebug.h" /* for NDirectFileRead */ #include "catalog/catalog.h" /* - * if BMTRACE is defined, we trace the last 200 buffer allocations and - * deallocations in a circular buffer in shared memory. + * if BMTRACE is defined, we trace the last 200 buffer allocations and + * deallocations in a circular buffer in shared memory. */ #ifdef BMTRACE -bmtrace *TraceBuf; -long *CurTraceBuf; -#define BMT_LIMIT 200 -#endif /* BMTRACE */ -int ShowPinTrace = 0; - -int NBuffers = NDBUFS; /* NDBUFS defined in miscadmin.h */ -int Data_Descriptors; -int Free_List_Descriptor; -int Lookup_List_Descriptor; -int Num_Descriptors; - -BufferDesc *BufferDescriptors; -BufferBlock BufferBlocks; +bmtrace *TraceBuf; +long *CurTraceBuf; + +#define BMT_LIMIT 200 +#endif /* BMTRACE */ +int ShowPinTrace = 0; + +int NBuffers = NDBUFS; /* NDBUFS defined in miscadmin.h */ +int Data_Descriptors; +int Free_List_Descriptor; +int Lookup_List_Descriptor; +int Num_Descriptors; + +BufferDesc *BufferDescriptors; +BufferBlock BufferBlocks; + #ifndef HAS_TEST_AND_SET -long *NWaitIOBackendP; +long *NWaitIOBackendP; + #endif -extern IpcSemaphoreId WaitIOSemId; +extern IpcSemaphoreId WaitIOSemId; + +long *PrivateRefCount;/* also used in freelist.c */ +long *LastRefCount; /* refcounts of last ExecMain level */ +long *CommitInfoNeedsSave; /* to write buffers where we have + * filled in */ -long *PrivateRefCount; /* also used in freelist.c */ -long *LastRefCount; /* refcounts of last ExecMain level */ -long *CommitInfoNeedsSave; /* to write buffers where we have filled in */ - /* t_tmin (or t_tmax) */ + /* t_tmin (or t_tmax) */ /* * Data Structures: - * buffers live in a freelist and a lookup data structure. - * + * buffers live in a freelist and a lookup data structure. + * * * Buffer Lookup: - * Two important notes. First, the buffer has to be - * available for lookup BEFORE an IO begins. Otherwise - * a second process trying to read the buffer will - * allocate its own copy and the buffeer pool will - * become inconsistent. + * Two important notes. First, the buffer has to be + * available for lookup BEFORE an IO begins. Otherwise + * a second process trying to read the buffer will + * allocate its own copy and the buffeer pool will + * become inconsistent. * * Buffer Replacement: - * see freelist.c. A buffer cannot be replaced while in - * use either by data manager or during IO. + * see freelist.c. A buffer cannot be replaced while in + * use either by data manager or during IO. * * WriteBufferBack: - * currently, a buffer is only written back at the time - * it is selected for replacement. It should - * be done sooner if possible to reduce latency of - * BufferAlloc(). Maybe there should be a daemon process. + * currently, a buffer is only written back at the time + * it is selected for replacement. It should + * be done sooner if possible to reduce latency of + * BufferAlloc(). Maybe there should be a daemon process. * * Synchronization/Locking: * - * BufMgrLock lock -- must be acquired before manipulating the - * buffer queues (lookup/freelist). Must be released - * before exit and before doing any IO. + * BufMgrLock lock -- must be acquired before manipulating the + * buffer queues (lookup/freelist). Must be released + * before exit and before doing any IO. * * IO_IN_PROGRESS -- this is a flag in the buffer descriptor. - * It must be set when an IO is initiated and cleared at - * the end of the IO. It is there to make sure that one - * process doesn't start to use a buffer while another is - * faulting it in. see IOWait/IOSignal. + * It must be set when an IO is initiated and cleared at + * the end of the IO. It is there to make sure that one + * process doesn't start to use a buffer while another is + * faulting it in. see IOWait/IOSignal. * - * refcount -- A buffer is pinned during IO and immediately - * after a BufferAlloc(). A buffer is always either pinned - * or on the freelist but never both. The buffer must be - * released, written, or flushed before the end of - * transaction. + * refcount -- A buffer is pinned during IO and immediately + * after a BufferAlloc(). A buffer is always either pinned + * or on the freelist but never both. The buffer must be + * released, written, or flushed before the end of + * transaction. * * PrivateRefCount -- Each buffer also has a private refcount the keeps - * track of the number of times the buffer is pinned in the current - * processes. This is used for two purposes, first, if we pin a - * a buffer more than once, we only need to change the shared refcount - * once, thus only lock the buffer pool once, second, when a transaction - * aborts, it should only unpin the buffers exactly the number of times it - * has pinned them, so that it will not blow away buffers of another - * backend. + * track of the number of times the buffer is pinned in the current + * processes. This is used for two purposes, first, if we pin a + * a buffer more than once, we only need to change the shared refcount + * once, thus only lock the buffer pool once, second, when a transaction + * aborts, it should only unpin the buffers exactly the number of times it + * has pinned them, so that it will not blow away buffers of another + * backend. * */ -SPINLOCK BufMgrLock; +SPINLOCK BufMgrLock; -long int ReadBufferCount; -long int ReadLocalBufferCount; -long int BufferHitCount; -long int LocalBufferHitCount; -long int BufferFlushCount; -long int LocalBufferFlushCount; +long int ReadBufferCount; +long int ReadLocalBufferCount; +long int BufferHitCount; +long int LocalBufferHitCount; +long int BufferFlushCount; +long int LocalBufferFlushCount; /* @@ -138,111 +143,121 @@ long int LocalBufferFlushCount; void InitBufferPool(IPCKey key) { - bool foundBufs,foundDescs; - int i; - - /* check padding of BufferDesc and BufferHdr */ - /* we need both checks because a sbufdesc_padded > PADDED_SBUFDESC_SIZE - will shrink sbufdesc to the required size, which is bad */ - if (sizeof(struct sbufdesc) != PADDED_SBUFDESC_SIZE || - sizeof(struct sbufdesc_unpadded) > PADDED_SBUFDESC_SIZE) - elog(WARN,"Internal error: sbufdesc does not have the proper size, " - "contact the Postgres developers"); - if (sizeof(struct sbufdesc_unpadded) <= PADDED_SBUFDESC_SIZE/2) - elog(WARN,"Internal error: sbufdesc is greatly over-sized, " - "contact the Postgres developers"); - - Data_Descriptors = NBuffers; - Free_List_Descriptor = Data_Descriptors; - Lookup_List_Descriptor = Data_Descriptors + 1; - Num_Descriptors = Data_Descriptors + 1; - - SpinAcquire(BufMgrLock); - + bool foundBufs, + foundDescs; + int i; + + /* check padding of BufferDesc and BufferHdr */ + + /* + * we need both checks because a sbufdesc_padded > + * PADDED_SBUFDESC_SIZE will shrink sbufdesc to the required size, + * which is bad + */ + if (sizeof(struct sbufdesc) != PADDED_SBUFDESC_SIZE || + sizeof(struct sbufdesc_unpadded) > PADDED_SBUFDESC_SIZE) + elog(WARN, "Internal error: sbufdesc does not have the proper size, " + "contact the Postgres developers"); + if (sizeof(struct sbufdesc_unpadded) <= PADDED_SBUFDESC_SIZE / 2) + elog(WARN, "Internal error: sbufdesc is greatly over-sized, " + "contact the Postgres developers"); + + Data_Descriptors = NBuffers; + Free_List_Descriptor = Data_Descriptors; + Lookup_List_Descriptor = Data_Descriptors + 1; + Num_Descriptors = Data_Descriptors + 1; + + SpinAcquire(BufMgrLock); + #ifdef BMTRACE - CurTraceBuf = (long *) ShmemInitStruct("Buffer trace", - (BMT_LIMIT * sizeof(bmtrace)) + sizeof(long), - &foundDescs); - if (!foundDescs) - memset(CurTraceBuf, 0, (BMT_LIMIT * sizeof(bmtrace)) + sizeof(long)); - - TraceBuf = (bmtrace *) &(CurTraceBuf[1]); + CurTraceBuf = (long *) ShmemInitStruct("Buffer trace", + (BMT_LIMIT * sizeof(bmtrace)) + sizeof(long), + &foundDescs); + if (!foundDescs) + memset(CurTraceBuf, 0, (BMT_LIMIT * sizeof(bmtrace)) + sizeof(long)); + + TraceBuf = (bmtrace *) & (CurTraceBuf[1]); #endif - - BufferDescriptors = (BufferDesc *) - ShmemInitStruct("Buffer Descriptors", - Num_Descriptors*sizeof(BufferDesc),&foundDescs); - - BufferBlocks = (BufferBlock) - ShmemInitStruct("Buffer Blocks", - NBuffers*BLCKSZ,&foundBufs); - + + BufferDescriptors = (BufferDesc *) + ShmemInitStruct("Buffer Descriptors", + Num_Descriptors * sizeof(BufferDesc), &foundDescs); + + BufferBlocks = (BufferBlock) + ShmemInitStruct("Buffer Blocks", + NBuffers * BLCKSZ, &foundBufs); + #ifndef HAS_TEST_AND_SET - { - bool foundNWaitIO; - - NWaitIOBackendP = (long *)ShmemInitStruct("#Backends Waiting IO", - sizeof(long), - &foundNWaitIO); - if (!foundNWaitIO) - *NWaitIOBackendP = 0; - } + { + bool foundNWaitIO; + + NWaitIOBackendP = (long *) ShmemInitStruct("#Backends Waiting IO", + sizeof(long), + &foundNWaitIO); + if (!foundNWaitIO) + *NWaitIOBackendP = 0; + } #endif - - if (foundDescs || foundBufs) { - - /* both should be present or neither */ - Assert(foundDescs && foundBufs); - - } else { - BufferDesc *buf; - unsigned long block; - - buf = BufferDescriptors; - block = (unsigned long) BufferBlocks; - - /* - * link the buffers into a circular, doubly-linked list to - * initialize free list. Still don't know anything about - * replacement strategy in this file. - */ - for (i = 0; i < Data_Descriptors; block+=BLCKSZ,buf++,i++) { - Assert(ShmemIsValid((unsigned long)block)); - - buf->freeNext = i+1; - buf->freePrev = i-1; - - CLEAR_BUFFERTAG(&(buf->tag)); - buf->data = MAKE_OFFSET(block); - buf->flags = (BM_DELETED | BM_FREE | BM_VALID); - buf->refcount = 0; - buf->buf_id = i; + + if (foundDescs || foundBufs) + { + + /* both should be present or neither */ + Assert(foundDescs && foundBufs); + + } + else + { + BufferDesc *buf; + unsigned long block; + + buf = BufferDescriptors; + block = (unsigned long) BufferBlocks; + + /* + * link the buffers into a circular, doubly-linked list to + * initialize free list. Still don't know anything about + * replacement strategy in this file. + */ + for (i = 0; i < Data_Descriptors; block += BLCKSZ, buf++, i++) + { + Assert(ShmemIsValid((unsigned long) block)); + + buf->freeNext = i + 1; + buf->freePrev = i - 1; + + CLEAR_BUFFERTAG(&(buf->tag)); + buf->data = MAKE_OFFSET(block); + buf->flags = (BM_DELETED | BM_FREE | BM_VALID); + buf->refcount = 0; + buf->buf_id = i; #ifdef HAS_TEST_AND_SET - S_INIT_LOCK(&(buf->io_in_progress_lock)); + S_INIT_LOCK(&(buf->io_in_progress_lock)); #endif + } + + /* close the circular queue */ + BufferDescriptors[0].freePrev = Data_Descriptors - 1; + BufferDescriptors[Data_Descriptors - 1].freeNext = 0; } - - /* close the circular queue */ - BufferDescriptors[0].freePrev = Data_Descriptors-1; - BufferDescriptors[Data_Descriptors-1].freeNext = 0; - } - - /* Init the rest of the module */ - InitBufTable(); - InitFreeList(!foundDescs); - - SpinRelease(BufMgrLock); - + + /* Init the rest of the module */ + InitBufTable(); + InitFreeList(!foundDescs); + + SpinRelease(BufMgrLock); + #ifndef HAS_TEST_AND_SET - { - int status; - WaitIOSemId = IpcSemaphoreCreate(IPCKeyGetWaitIOSemaphoreKey(key), - 1, IPCProtection, 0, 1, &status); - } + { + int status; + + WaitIOSemId = IpcSemaphoreCreate(IPCKeyGetWaitIOSemaphoreKey(key), + 1, IPCProtection, 0, 1, &status); + } #endif - PrivateRefCount = (long *) calloc(NBuffers, sizeof(long)); - LastRefCount = (long *) calloc(NBuffers, sizeof(long)); - CommitInfoNeedsSave = (long *) calloc(NBuffers, sizeof(long)); + PrivateRefCount = (long *) calloc(NBuffers, sizeof(long)); + LastRefCount = (long *) calloc(NBuffers, sizeof(long)); + CommitInfoNeedsSave = (long *) calloc(NBuffers, sizeof(long)); } /* ----------------------------------------------------- @@ -255,43 +270,41 @@ InitBufferPool(IPCKey key) int BufferShmemSize() { - int size = 0; - int nbuckets; - int nsegs; - int tmp; - - nbuckets = 1 << (int)my_log2((NBuffers - 1) / DEF_FFACTOR + 1); - nsegs = 1 << (int)my_log2((nbuckets - 1) / DEF_SEGSIZE + 1); - - /* size of shmem binding table */ - size += MAXALIGN(my_log2(BTABLE_SIZE) * sizeof(void *)); /* HTAB->dir */ - size += MAXALIGN(sizeof(HHDR)); /* HTAB->hctl */ - size += MAXALIGN(DEF_SEGSIZE * sizeof(SEGMENT)); - size += BUCKET_ALLOC_INCR * - (MAXALIGN(sizeof(BUCKET_INDEX)) + - MAXALIGN(BTABLE_KEYSIZE) + - MAXALIGN(BTABLE_DATASIZE)); - - /* size of buffer descriptors */ - size += MAXALIGN((NBuffers + 1) * sizeof(BufferDesc)); - - /* size of data pages */ - size += NBuffers * MAXALIGN(BLCKSZ); - - /* size of buffer hash table */ - size += MAXALIGN(my_log2(NBuffers) * sizeof(void *)); /* HTAB->dir */ - size += MAXALIGN(sizeof(HHDR)); /* HTAB->hctl */ - size += nsegs * MAXALIGN(DEF_SEGSIZE * sizeof(SEGMENT)); - tmp = (int)ceil((double)NBuffers/BUCKET_ALLOC_INCR); - size += tmp * BUCKET_ALLOC_INCR * - (MAXALIGN(sizeof(BUCKET_INDEX)) + - MAXALIGN(sizeof(BufferTag)) + - MAXALIGN(sizeof(Buffer))); - + int size = 0; + int nbuckets; + int nsegs; + int tmp; + + nbuckets = 1 << (int) my_log2((NBuffers - 1) / DEF_FFACTOR + 1); + nsegs = 1 << (int) my_log2((nbuckets - 1) / DEF_SEGSIZE + 1); + + /* size of shmem binding table */ + size += MAXALIGN(my_log2(BTABLE_SIZE) * sizeof(void *)); /* HTAB->dir */ + size += MAXALIGN(sizeof(HHDR)); /* HTAB->hctl */ + size += MAXALIGN(DEF_SEGSIZE * sizeof(SEGMENT)); + size += BUCKET_ALLOC_INCR * + (MAXALIGN(sizeof(BUCKET_INDEX)) + + MAXALIGN(BTABLE_KEYSIZE) + + MAXALIGN(BTABLE_DATASIZE)); + + /* size of buffer descriptors */ + size += MAXALIGN((NBuffers + 1) * sizeof(BufferDesc)); + + /* size of data pages */ + size += NBuffers * MAXALIGN(BLCKSZ); + + /* size of buffer hash table */ + size += MAXALIGN(my_log2(NBuffers) * sizeof(void *)); /* HTAB->dir */ + size += MAXALIGN(sizeof(HHDR)); /* HTAB->hctl */ + size += nsegs * MAXALIGN(DEF_SEGSIZE * sizeof(SEGMENT)); + tmp = (int) ceil((double) NBuffers / BUCKET_ALLOC_INCR); + size += tmp * BUCKET_ALLOC_INCR * + (MAXALIGN(sizeof(BUCKET_INDEX)) + + MAXALIGN(sizeof(BufferTag)) + + MAXALIGN(sizeof(Buffer))); + #ifdef BMTRACE - size += (BMT_LIMIT * sizeof(bmtrace)) + sizeof(long); + size += (BMT_LIMIT * sizeof(bmtrace)) + sizeof(long); #endif - return size; + return size; } - - |