diff options
Diffstat (limited to 'src')
| -rw-r--r-- | src/backend/access/brin/brin.c | 2 | ||||
| -rw-r--r-- | src/backend/access/brin/brin_pageops.c | 10 | ||||
| -rw-r--r-- | src/backend/access/heap/hio.c | 42 | ||||
| -rw-r--r-- | src/backend/access/heap/vacuumlazy.c | 17 | ||||
| -rw-r--r-- | src/backend/access/transam/xact.c | 14 | ||||
| -rw-r--r-- | src/backend/storage/freespace/README | 38 | ||||
| -rw-r--r-- | src/backend/storage/freespace/freespace.c | 301 | ||||
| -rw-r--r-- | src/backend/storage/freespace/indexfsm.c | 6 | ||||
| -rw-r--r-- | src/bin/pg_upgrade/info.c | 16 | ||||
| -rw-r--r-- | src/bin/pg_upgrade/pg_upgrade.h | 6 | ||||
| -rw-r--r-- | src/bin/pg_upgrade/relfilenode.c | 63 | ||||
| -rw-r--r-- | src/include/storage/freespace.h | 9 | ||||
| -rw-r--r-- | src/test/regress/expected/fsm.out | 73 | ||||
| -rw-r--r-- | src/test/regress/parallel_schedule | 2 | ||||
| -rw-r--r-- | src/test/regress/serial_schedule | 1 | ||||
| -rw-r--r-- | src/test/regress/sql/fsm.sql | 66 |
16 files changed, 49 insertions, 617 deletions
diff --git a/src/backend/access/brin/brin.c b/src/backend/access/brin/brin.c index 5c2b0c76358..aba234c0af8 100644 --- a/src/backend/access/brin/brin.c +++ b/src/backend/access/brin/brin.c @@ -1152,7 +1152,7 @@ terminate_brin_buildstate(BrinBuildState *state) freespace = PageGetFreeSpace(page); blk = BufferGetBlockNumber(state->bs_currentInsertBuf); ReleaseBuffer(state->bs_currentInsertBuf); - RecordPageWithFreeSpace(state->bs_irel, blk, freespace, InvalidBlockNumber); + RecordPageWithFreeSpace(state->bs_irel, blk, freespace); FreeSpaceMapVacuumRange(state->bs_irel, blk, blk + 1); } diff --git a/src/backend/access/brin/brin_pageops.c b/src/backend/access/brin/brin_pageops.c index 2e83aa42f78..fcc468ced99 100644 --- a/src/backend/access/brin/brin_pageops.c +++ b/src/backend/access/brin/brin_pageops.c @@ -310,7 +310,7 @@ brin_doupdate(Relation idxrel, BlockNumber pagesPerRange, if (extended) { - RecordPageWithFreeSpace(idxrel, newblk, freespace, InvalidBlockNumber); + RecordPageWithFreeSpace(idxrel, newblk, freespace); FreeSpaceMapVacuumRange(idxrel, newblk, newblk + 1); } @@ -461,7 +461,7 @@ brin_doinsert(Relation idxrel, BlockNumber pagesPerRange, if (extended) { - RecordPageWithFreeSpace(idxrel, blk, freespace, InvalidBlockNumber); + RecordPageWithFreeSpace(idxrel, blk, freespace); FreeSpaceMapVacuumRange(idxrel, blk, blk + 1); } @@ -654,7 +654,7 @@ brin_page_cleanup(Relation idxrel, Buffer buf) /* Measure free space and record it */ RecordPageWithFreeSpace(idxrel, BufferGetBlockNumber(buf), - br_page_get_freespace(page), InvalidBlockNumber); + br_page_get_freespace(page)); } /* @@ -703,7 +703,7 @@ brin_getinsertbuffer(Relation irel, Buffer oldbuf, Size itemsz, /* Choose initial target page, re-using existing target if known */ newblk = RelationGetTargetBlock(irel); if (newblk == InvalidBlockNumber) - newblk = GetPageWithFreeSpace(irel, itemsz, true); + newblk = GetPageWithFreeSpace(irel, itemsz); /* * Loop until we find a page with sufficient free space. By the time we @@ -895,7 +895,7 @@ brin_initialize_empty_new_buffer(Relation idxrel, Buffer buffer) * pages whose FSM records were forgotten in a crash. */ RecordPageWithFreeSpace(idxrel, BufferGetBlockNumber(buffer), - br_page_get_freespace(page), InvalidBlockNumber); + br_page_get_freespace(page)); } diff --git a/src/backend/access/heap/hio.c b/src/backend/access/heap/hio.c index 69a7a23874f..d41d318eef9 100644 --- a/src/backend/access/heap/hio.c +++ b/src/backend/access/heap/hio.c @@ -246,14 +246,8 @@ RelationAddExtraBlocks(Relation relation, BulkInsertState bistate) * Immediately update the bottom level of the FSM. This has a good * chance of making this page visible to other concurrently inserting * backends, and we want that to happen without delay. - * - * Since we know the table will end up with extraBlocks additional - * pages, we pass the final number to avoid possible unnecessary - * system calls and to make sure the FSM is created when we add the - * first new page. */ - RecordPageWithFreeSpace(relation, blockNum, freespace, - firstBlock + extraBlocks); + RecordPageWithFreeSpace(relation, blockNum, freespace); } while (--extraBlocks > 0); @@ -390,9 +384,20 @@ RelationGetBufferForTuple(Relation relation, Size len, * We have no cached target page, so ask the FSM for an initial * target. */ - targetBlock = GetPageWithFreeSpace(relation, - len + saveFreeSpace, - false); + targetBlock = GetPageWithFreeSpace(relation, len + saveFreeSpace); + + /* + * If the FSM knows nothing of the rel, try the last page before we + * give up and extend. This avoids one-tuple-per-page syndrome during + * bootstrapping or in a recently-started system. + */ + if (targetBlock == InvalidBlockNumber) + { + BlockNumber nblocks = RelationGetNumberOfBlocks(relation); + + if (nblocks > 0) + targetBlock = nblocks - 1; + } } loop: @@ -499,13 +504,6 @@ loop: { /* use this page as future insert target, too */ RelationSetTargetBlock(relation, targetBlock); - - /* - * In case we used an in-memory map of available blocks, reset it - * for next use. - */ - FSMClearLocalMap(); - return buffer; } @@ -565,12 +563,9 @@ loop: /* * Check if some other backend has extended a block for us while - * we were waiting on the lock. We only check the FSM -- if there - * isn't one we don't recheck the number of blocks. + * we were waiting on the lock. */ - targetBlock = GetPageWithFreeSpace(relation, - len + saveFreeSpace, - true); + targetBlock = GetPageWithFreeSpace(relation, len + saveFreeSpace); /* * If some other waiter has already extended the relation, we @@ -675,8 +670,5 @@ loop: */ RelationSetTargetBlock(relation, BufferGetBlockNumber(buffer)); - /* This should already be cleared by now, but make sure it is. */ - FSMClearLocalMap(); - return buffer; } diff --git a/src/backend/access/heap/vacuumlazy.c b/src/backend/access/heap/vacuumlazy.c index 2aa729817cb..0f70dc883d1 100644 --- a/src/backend/access/heap/vacuumlazy.c +++ b/src/backend/access/heap/vacuumlazy.c @@ -153,7 +153,7 @@ static BufferAccessStrategy vac_strategy; static void lazy_scan_heap(Relation onerel, VacuumParams *params, LVRelStats *vacrelstats, Relation *Irel, int nindexes, bool aggressive); -static void lazy_vacuum_heap(Relation onerel, LVRelStats *vacrelstats, BlockNumber nblocks); +static void lazy_vacuum_heap(Relation onerel, LVRelStats *vacrelstats); static bool lazy_check_needs_freeze(Buffer buf, bool *hastup); static void lazy_vacuum_index(Relation indrel, IndexBulkDeleteResult **stats, @@ -780,7 +780,7 @@ lazy_scan_heap(Relation onerel, VacuumParams *params, LVRelStats *vacrelstats, pgstat_progress_update_multi_param(2, hvp_index, hvp_val); /* Remove tuples from heap */ - lazy_vacuum_heap(onerel, vacrelstats, nblocks); + lazy_vacuum_heap(onerel, vacrelstats); /* * Forget the now-vacuumed tuples, and press on, but be careful @@ -919,7 +919,7 @@ lazy_scan_heap(Relation onerel, VacuumParams *params, LVRelStats *vacrelstats, Size freespace; freespace = BufferGetPageSize(buf) - SizeOfPageHeaderData; - RecordPageWithFreeSpace(onerel, blkno, freespace, nblocks); + RecordPageWithFreeSpace(onerel, blkno, freespace); } } continue; @@ -963,7 +963,7 @@ lazy_scan_heap(Relation onerel, VacuumParams *params, LVRelStats *vacrelstats, } UnlockReleaseBuffer(buf); - RecordPageWithFreeSpace(onerel, blkno, freespace, nblocks); + RecordPageWithFreeSpace(onerel, blkno, freespace); continue; } @@ -1381,7 +1381,7 @@ lazy_scan_heap(Relation onerel, VacuumParams *params, LVRelStats *vacrelstats, * taken if there are no indexes.) */ if (vacrelstats->num_dead_tuples == prev_dead_count) - RecordPageWithFreeSpace(onerel, blkno, freespace, nblocks); + RecordPageWithFreeSpace(onerel, blkno, freespace); } /* report that everything is scanned and vacuumed */ @@ -1443,7 +1443,7 @@ lazy_scan_heap(Relation onerel, VacuumParams *params, LVRelStats *vacrelstats, /* Remove tuples from heap */ pgstat_progress_update_param(PROGRESS_VACUUM_PHASE, PROGRESS_VACUUM_PHASE_VACUUM_HEAP); - lazy_vacuum_heap(onerel, vacrelstats, nblocks); + lazy_vacuum_heap(onerel, vacrelstats); vacrelstats->num_index_scans++; } @@ -1517,10 +1517,9 @@ lazy_scan_heap(Relation onerel, VacuumParams *params, LVRelStats *vacrelstats, * Note: the reason for doing this as a second pass is we cannot remove * the tuples until we've removed their index entries, and we want to * process index entry removal in batches as large as possible. - * Note: nblocks is passed as an optimization for RecordPageWithFreeSpace(). */ static void -lazy_vacuum_heap(Relation onerel, LVRelStats *vacrelstats, BlockNumber nblocks) +lazy_vacuum_heap(Relation onerel, LVRelStats *vacrelstats) { int tupindex; int npages; @@ -1557,7 +1556,7 @@ lazy_vacuum_heap(Relation onerel, LVRelStats *vacrelstats, BlockNumber nblocks) freespace = PageGetHeapFreeSpace(page); UnlockReleaseBuffer(buf); - RecordPageWithFreeSpace(onerel, tblk, freespace, nblocks); + RecordPageWithFreeSpace(onerel, tblk, freespace); npages++; } diff --git a/src/backend/access/transam/xact.c b/src/backend/access/transam/xact.c index 8522a2f51c5..0a34651581e 100644 --- a/src/backend/access/transam/xact.c +++ b/src/backend/access/transam/xact.c @@ -48,7 +48,6 @@ #include "replication/walsender.h" #include "storage/condition_variable.h" #include "storage/fd.h" -#include "storage/freespace.h" #include "storage/lmgr.h" #include "storage/md.h" #include "storage/predicate.h" @@ -2587,12 +2586,6 @@ AbortTransaction(void) pgstat_report_wait_end(); pgstat_progress_end_command(); - /* - * In case we aborted during RelationGetBufferForTuple(), clear the local - * map of heap pages. - */ - FSMClearLocalMap(); - /* Clean up buffer I/O and buffer context locks, too */ AbortBufferIO(); UnlockBuffers(); @@ -4880,13 +4873,6 @@ AbortSubTransaction(void) pgstat_report_wait_end(); pgstat_progress_end_command(); - - /* - * In case we aborted during RelationGetBufferForTuple(), clear the local - * map of heap pages. - */ - FSMClearLocalMap(); - AbortBufferIO(); UnlockBuffers(); diff --git a/src/backend/storage/freespace/README b/src/backend/storage/freespace/README index 0d3cd29772e..e7ff23b76f7 100644 --- a/src/backend/storage/freespace/README +++ b/src/backend/storage/freespace/README @@ -8,41 +8,7 @@ free space to hold a tuple to be stored; or to determine that no such page exists and the relation must be extended by one page. As of PostgreSQL 8.4 each relation has its own, extensible free space map stored in a separate "fork" of its relation. This eliminates the disadvantages of the former -fixed-size FSM. There are two exceptions: - -1. Hash indexes never have a FSM. -2. For very small tables, a 3-page relation fork would be relatively large -and wasteful, so to save space we refrain from creating the FSM if the -heap has HEAP_FSM_CREATION_THRESHOLD pages or fewer. - -To locate free space in the latter case, we simply try pages directly without -knowing ahead of time how much free space they have. To maintain good -performance, we create a local in-memory map of pages to try, and only mark -every other page as available. For example, in a 3-page heap, the local map -would look like: - -ANAN -0123 - -Pages 0 and 2 are marked "available", and page 1 as "not available". -Page 3 is beyond the end of the relation, so is likewise marked "not -available". First we try page 2, and if that doesn't have sufficient free -space we try page 0 before giving up and extending the relation. There may -be some wasted free space on block 1, but if the relation extends to 4 pages: - -NANA -0123 - -We not only have the new page 3 at our disposal, we can now check page 1 -for free space as well. - -Once the FSM is created for a heap we don't remove it even if somebody deletes -all the rows from the corresponding relation. We don't think it is a useful -optimization as it is quite likely that relation will again grow to the same -size. - -FSM data structure ------------------- +fixed-size FSM. It is important to keep the map small so that it can be searched rapidly. Therefore, we don't attempt to record the exact free space on a page. @@ -226,3 +192,5 @@ TODO ---- - fastroot to avoid traversing upper nodes with just 1 child +- use a different system for tables that fit into one FSM page, with a + mechanism to switch to the real thing as it grows. diff --git a/src/backend/storage/freespace/freespace.c b/src/backend/storage/freespace/freespace.c index c3ed4242e2d..eee82860575 100644 --- a/src/backend/storage/freespace/freespace.c +++ b/src/backend/storage/freespace/freespace.c @@ -76,14 +76,6 @@ #define FSM_ROOT_LEVEL (FSM_TREE_DEPTH - 1) #define FSM_BOTTOM_LEVEL 0 -/* Status codes for the local map. */ - -/* Either already tried, or beyond the end of the relation */ -#define FSM_LOCAL_NOT_AVAIL 0x00 - -/* Available to try */ -#define FSM_LOCAL_AVAIL 0x01 - /* * The internal FSM routines work on a logical addressing scheme. Each * level of the tree can be thought of as a separately addressable file. @@ -97,32 +89,6 @@ typedef struct /* Address of the root page. */ static const FSMAddress FSM_ROOT_ADDRESS = {FSM_ROOT_LEVEL, 0}; -/* - * For small relations, we don't create FSM to save space, instead we use - * local in-memory map of pages to try. To locate free space, we simply try - * pages directly without knowing ahead of time how much free space they have. - * - * Note that this map is used to the find the block with required free space - * for any given relation. We clear this map when we have found a block with - * enough free space, when we extend the relation, or on transaction abort. - * See src/backend/storage/freespace/README for further details. - */ -typedef struct -{ - BlockNumber nblocks; - uint8 map[HEAP_FSM_CREATION_THRESHOLD]; -} FSMLocalMap; - -static FSMLocalMap fsm_local_map = -{ - 0, - { - FSM_LOCAL_NOT_AVAIL - } -}; - -#define FSM_LOCAL_MAP_EXISTS (fsm_local_map.nblocks > 0) - /* functions to navigate the tree */ static FSMAddress fsm_get_child(FSMAddress parent, uint16 slot); static FSMAddress fsm_get_parent(FSMAddress child, uint16 *slot); @@ -141,14 +107,10 @@ static Size fsm_space_cat_to_avail(uint8 cat); /* workhorse functions for various operations */ static int fsm_set_and_search(Relation rel, FSMAddress addr, uint16 slot, uint8 newValue, uint8 minValue); -static void fsm_local_set(Relation rel, BlockNumber cur_nblocks); static BlockNumber fsm_search(Relation rel, uint8 min_cat); -static BlockNumber fsm_local_search(void); static uint8 fsm_vacuum_page(Relation rel, FSMAddress addr, BlockNumber start, BlockNumber end, bool *eof); -static bool fsm_allow_writes(Relation rel, BlockNumber heapblk, - BlockNumber nblocks, BlockNumber *get_nblocks); /******** Public API ********/ @@ -165,46 +127,13 @@ static bool fsm_allow_writes(Relation rel, BlockNumber heapblk, * amount of free space available on that page and then try again (see * RecordAndGetPageWithFreeSpace). If InvalidBlockNumber is returned, * extend the relation. - * - * For very small heap relations that don't have a FSM, we try every other - * page before extending the relation. To keep track of which pages have - * been tried, initialize a local in-memory map of pages. */ BlockNumber -GetPageWithFreeSpace(Relation rel, Size spaceNeeded, bool check_fsm_only) +GetPageWithFreeSpace(Relation rel, Size spaceNeeded) { uint8 min_cat = fsm_space_needed_to_cat(spaceNeeded); - BlockNumber target_block, - nblocks; - /* First try the FSM, if it exists. */ - target_block = fsm_search(rel, min_cat); - - if (target_block == InvalidBlockNumber && - (rel->rd_rel->relkind == RELKIND_RELATION || - rel->rd_rel->relkind == RELKIND_TOASTVALUE) && - !check_fsm_only) - { - nblocks = RelationGetNumberOfBlocks(rel); - - if (nblocks > HEAP_FSM_CREATION_THRESHOLD) - { - /* - * If the FSM knows nothing of the rel, try the last page before - * we give up and extend. This avoids one-tuple-per-page syndrome - * during bootstrapping or in a recently-started system. - */ - target_block = nblocks - 1; - } - else if (nblocks > 0) - { - /* Initialize local map and get first candidate block. */ - fsm_local_set(rel, nblocks); - target_block = fsm_local_search(); - } - } - - return target_block; + return fsm_search(rel, min_cat); } /* @@ -215,47 +144,16 @@ GetPageWithFreeSpace(Relation rel, Size spaceNeeded, bool check_fsm_only) * also some effort to return a page close to the old page; if there's a * page with enough free space on the same FSM page where the old one page * is located, it is preferred. - * - * For very small heap relations that don't have a FSM, we update the local - * map to indicate we have tried a page, and return the next page to try. */ BlockNumber RecordAndGetPageWithFreeSpace(Relation rel, BlockNumber oldPage, Size oldSpaceAvail, Size spaceNeeded) { - int old_cat; - int search_cat; + int old_cat = fsm_space_avail_to_cat(oldSpaceAvail); + int search_cat = fsm_space_needed_to_cat(spaceNeeded); FSMAddress addr; uint16 slot; int search_slot; - BlockNumber nblocks = InvalidBlockNumber; - - /* First try the local map, if it exists. */ - if (FSM_LOCAL_MAP_EXISTS) - { - Assert((rel->rd_rel->relkind == RELKIND_RELATION || - rel->rd_rel->relkind == RELKIND_TOASTVALUE) && - fsm_local_map.map[oldPage] == FSM_LOCAL_AVAIL); - - fsm_local_map.map[oldPage] = FSM_LOCAL_NOT_AVAIL; - return fsm_local_search(); - } - - if (!fsm_allow_writes(rel, oldPage, InvalidBlockNumber, &nblocks)) - { - /* - * If we have neither a local map nor a FSM, we probably just tried - * the target block in the smgr relation entry and failed, so we'll - * need to create the local map. - */ - fsm_local_set(rel, nblocks); - return fsm_local_search(); - } - - /* Normal FSM logic follows */ - - old_cat = fsm_space_avail_to_cat(oldSpaceAvail); - search_cat = fsm_space_needed_to_cat(spaceNeeded); /* Get the location of the FSM byte representing the heap block */ addr = fsm_get_location(oldPage, &slot); @@ -278,45 +176,21 @@ RecordAndGetPageWithFreeSpace(Relation rel, BlockNumber oldPage, * Note that if the new spaceAvail value is higher than the old value stored * in the FSM, the space might not become visible to searchers until the next * FreeSpaceMapVacuum call, which updates the upper level pages. - * - * Callers have no need for a local map. */ void -RecordPageWithFreeSpace(Relation rel, BlockNumber heapBlk, - Size spaceAvail, BlockNumber nblocks) +RecordPageWithFreeSpace(Relation rel, BlockNumber heapBlk, Size spaceAvail) { - int new_cat; + int new_cat = fsm_space_avail_to_cat(spaceAvail); FSMAddress addr; uint16 slot; - BlockNumber dummy; - - if (!fsm_allow_writes(rel, heapBlk, nblocks, &dummy)) - /* No FSM to update and no local map either */ - return; /* Get the location of the FSM byte representing the heap block */ addr = fsm_get_location(heapBlk, &slot); - new_cat = fsm_space_avail_to_cat(spaceAvail); fsm_set_and_search(rel, addr, slot, new_cat, 0); } /* - * Clear the local map. We must call this when we have found a block with - * enough free space, when we extend the relation, or on transaction abort. - */ -void -FSMClearLocalMap(void) -{ - if (FSM_LOCAL_MAP_EXISTS) - { - fsm_local_map.nblocks = 0; - memset(&fsm_local_map.map, FSM_LOCAL_NOT_AVAIL, - sizeof(fsm_local_map.map)); - } -} - -/* * XLogRecordPageWithFreeSpace - like RecordPageWithFreeSpace, for use in * WAL replay */ @@ -330,31 +204,6 @@ XLogRecordPageWithFreeSpace(RelFileNode rnode, BlockNumber heapBlk, BlockNumber blkno; Buffer buf; Page page; - bool write_to_fsm; - - /* This is meant to mirror the logic in fsm_allow_writes() */ - if (heapBlk >= HEAP_FSM_CREATION_THRESHOLD) - write_to_fsm = true; - else - { - /* Open the relation at smgr level */ - SMgrRelation smgr = smgropen(rnode, InvalidBackendId); - - if (smgrexists(smgr, FSM_FORKNUM)) - write_to_fsm = true; - else - { - BlockNumber heap_nblocks = smgrnblocks(smgr, MAIN_FORKNUM); - - if (heap_nblocks > HEAP_FSM_CREATION_THRESHOLD) - write_to_fsm = true; - else - write_to_fsm = false; - } - } - - if (!write_to_fsm) - return; /* Get the location of the FSM byte representing the heap block */ addr = fsm_get_location(heapBlk, &slot); @@ -1055,141 +904,3 @@ fsm_vacuum_page(Relation rel, FSMAddress addr, return max_avail; } - -/* - * For heaps, we prevent creation of the FSM unless the number of pages - * exceeds HEAP_FSM_CREATION_THRESHOLD. For tables that don't already have - * a FSM, this will save an inode and a few kB of space. - * - * XXX The API is a little awkward -- if the caller passes a valid nblocks - * value, it can avoid invoking a system call. If the caller passes - * InvalidBlockNumber and receives a false return value, it can get an - * up-to-date relation size from get_nblocks. This saves a few cycles in - * the caller, which would otherwise need to get the relation size by itself. - */ -static bool -fsm_allow_writes(Relation rel, BlockNumber heapblk, - BlockNumber nblocks, BlockNumber *get_nblocks) -{ - bool skip_get_nblocks; - - if (heapblk >= HEAP_FSM_CREATION_THRESHOLD) - return true; - - /* Non-heap rels can always create a FSM. */ - if (rel->rd_rel->relkind != RELKIND_RELATION && - rel->rd_rel->relkind != RELKIND_TOASTVALUE) - return true; - - /* - * If the caller knows nblocks, we can avoid a system call later. If it - * doesn't, maybe we have relpages from a previous VACUUM. Since the table - * may have extended since then, we still have to count the pages later if - * we can't return now. - */ - if (nblocks != InvalidBlockNumber) - { - if (nblocks > HEAP_FSM_CREATION_THRESHOLD) - return true; - else - skip_get_nblocks = true; - } - else - { - if (rel->rd_rel->relpages != InvalidBlockNumber && - rel->rd_rel->relpages > HEAP_FSM_CREATION_THRESHOLD) - return true; - else - skip_get_nblocks = false; - } - - RelationOpenSmgr(rel); - if (smgrexists(rel->rd_smgr, FSM_FORKNUM)) - return true; - - if (skip_get_nblocks) - return false; - - /* last resort */ - *get_nblocks = RelationGetNumberOfBlocks(rel); - if (*get_nblocks > HEAP_FSM_CREATION_THRESHOLD) - return true; - else - return false; -} - -/* - * Initialize the local map of blocks to try, for when there is no FSM. - * - * When we initialize the map, the whole heap is potentially available to - * try. Testing revealed that trying every block can cause a small - * performance dip compared to when we use a FSM, so we try every other - * block instead. - */ -static void -fsm_local_set(Relation rel, BlockNumber cur_nblocks) -{ - BlockNumber blkno, - cached_target_block; - - /* The local map must not be set already. */ - Assert(!FSM_LOCAL_MAP_EXISTS); - - /* - * Starting at the current last block in the relation and working - * backwards, mark alternating blocks as available. - */ - blkno = cur_nblocks - 1; - while (true) - { - fsm_local_map.map[blkno] = FSM_LOCAL_AVAIL; - if (blkno >= 2) - blkno -= 2; - else - break; - } - - /* Cache the number of blocks. */ - fsm_local_map.nblocks = cur_nblocks; - - /* Set the status of the cached target block to 'unavailable'. */ - cached_target_block = RelationGetTargetBlock(rel); - if (cached_target_block != InvalidBlockNumber && - cached_target_block < cur_nblocks) - fsm_local_map.map[cached_target_block] = FSM_LOCAL_NOT_AVAIL; -} - -/* - * Search the local map for an available block to try, in descending order. - * As such, there is no heuristic available to decide which order will be - * better to try, but the probability of having space in the last block in the - * map is higher because that is the most recent block added to the heap. - * - * This function is used when there is no FSM. - */ -static BlockNumber -fsm_local_search(void) -{ - BlockNumber target_block; - - /* Local map must be set by now. */ - Assert(FSM_LOCAL_MAP_EXISTS); - - target_block = fsm_local_map.nblocks; - do - { - target_block--; - if (fsm_local_map.map[target_block] == FSM_LOCAL_AVAIL) - return target_block; - } while (target_block > 0); - - /* - * If we didn't find any available block to try in the local map, then - * clear it. This prevents us from using the map again without setting it - * first, which would otherwise lead to the same conclusion again and - * again. - */ - FSMClearLocalMap(); - - return InvalidBlockNumber; -} diff --git a/src/backend/storage/freespace/indexfsm.c b/src/backend/storage/freespace/indexfsm.c index 9d8f43d3739..58cedeaa9f7 100644 --- a/src/backend/storage/freespace/indexfsm.c +++ b/src/backend/storage/freespace/indexfsm.c @@ -37,7 +37,7 @@ BlockNumber GetFreeIndexPage(Relation rel) { - BlockNumber blkno = GetPageWithFreeSpace(rel, BLCKSZ / 2, true); + BlockNumber blkno = GetPageWithFreeSpace(rel, BLCKSZ / 2); if (blkno != InvalidBlockNumber) RecordUsedIndexPage(rel, blkno); @@ -51,7 +51,7 @@ GetFreeIndexPage(Relation rel) void RecordFreeIndexPage(Relation rel, BlockNumber freeBlock) { - RecordPageWithFreeSpace(rel, freeBlock, BLCKSZ - 1, InvalidBlockNumber); + RecordPageWithFreeSpace(rel, freeBlock, BLCKSZ - 1); } @@ -61,7 +61,7 @@ RecordFreeIndexPage(Relation rel, BlockNumber freeBlock) void RecordUsedIndexPage(Relation rel, BlockNumber usedBlock) { - RecordPageWithFreeSpace(rel, usedBlock, 0, InvalidBlockNumber); + RecordPageWithFreeSpace(rel, usedBlock, 0); } /* diff --git a/src/bin/pg_upgrade/info.c b/src/bin/pg_upgrade/info.c index 902bfc647e9..2f925f086c3 100644 --- a/src/bin/pg_upgrade/info.c +++ b/src/bin/pg_upgrade/info.c @@ -200,8 +200,6 @@ create_rel_filename_map(const char *old_data, const char *new_data, map->old_db_oid = old_db->db_oid; map->new_db_oid = new_db->db_oid; - map->relpages = old_rel->relpages; - map->relkind = old_rel->relkind; /* * old_relfilenode might differ from pg_class.oid (and hence @@ -420,7 +418,6 @@ get_rel_infos(ClusterInfo *cluster, DbInfo *dbinfo) char *nspname = NULL; char *relname = NULL; char *tablespace = NULL; - char *relkind = NULL; int i_spclocation, i_nspname, i_relname, @@ -428,9 +425,7 @@ get_rel_infos(ClusterInfo *cluster, DbInfo *dbinfo) i_indtable, i_toastheap, i_relfilenode, - i_reltablespace, - i_relpages, - i_relkind; + i_reltablespace; char query[QUERY_ALLOC]; char *last_namespace = NULL, *last_tablespace = NULL; @@ -499,7 +494,7 @@ get_rel_infos(ClusterInfo *cluster, DbInfo *dbinfo) */ snprintf(query + strlen(query), sizeof(query) - strlen(query), "SELECT all_rels.*, n.nspname, c.relname, " - " c.relfilenode, c.reltablespace, c.relpages, c.relkind, %s " + " c.relfilenode, c.reltablespace, %s " "FROM (SELECT * FROM regular_heap " " UNION ALL " " SELECT * FROM toast_heap " @@ -530,8 +525,6 @@ get_rel_infos(ClusterInfo *cluster, DbInfo *dbinfo) i_relname = PQfnumber(res, "relname"); i_relfilenode = PQfnumber(res, "relfilenode"); i_reltablespace = PQfnumber(res, "reltablespace"); - i_relpages = PQfnumber(res, "relpages"); - i_relkind = PQfnumber(res, "relkind"); i_spclocation = PQfnumber(res, "spclocation"); for (relnum = 0; relnum < ntups; relnum++) @@ -563,11 +556,6 @@ get_rel_infos(ClusterInfo *cluster, DbInfo *dbinfo) curr->relname = pg_strdup(relname); curr->relfilenode = atooid(PQgetvalue(res, relnum, i_relfilenode)); - curr->relpages = atoi(PQgetvalue(res, relnum, i_relpages)); - - relkind = PQgetvalue(res, relnum, i_relkind); - curr->relkind = relkind[0]; - curr->tblsp_alloc = false; /* Is the tablespace oid non-default? */ diff --git a/src/bin/pg_upgrade/pg_upgrade.h b/src/bin/pg_upgrade/pg_upgrade.h index baeb8ff0f85..2f67eee22b8 100644 --- a/src/bin/pg_upgrade/pg_upgrade.h +++ b/src/bin/pg_upgrade/pg_upgrade.h @@ -147,8 +147,6 @@ typedef struct char *tablespace; /* tablespace path; "" for cluster default */ bool nsp_alloc; /* should nspname be freed? */ bool tblsp_alloc; /* should tablespace be freed? */ - int32 relpages; /* # of pages -- see pg_class.h */ - char relkind; /* relation kind -- see pg_class.h */ } RelInfo; typedef struct @@ -175,10 +173,6 @@ typedef struct */ Oid old_relfilenode; Oid new_relfilenode; - - int32 relpages; /* # of pages -- see pg_class.h */ - char relkind; /* relation kind -- see pg_class.h */ - /* the rest are used only for logging and error reporting */ char *nspname; /* namespaces */ char *relname; diff --git a/src/bin/pg_upgrade/relfilenode.c b/src/bin/pg_upgrade/relfilenode.c index dd3c8cefe42..0c78073f0eb 100644 --- a/src/bin/pg_upgrade/relfilenode.c +++ b/src/bin/pg_upgrade/relfilenode.c @@ -14,12 +14,10 @@ #include <sys/stat.h> #include "catalog/pg_class_d.h" #include "access/transam.h" -#include "storage/freespace.h" static void transfer_single_new_db(FileNameMap *maps, int size, char *old_tablespace); static void transfer_relfile(FileNameMap *map, const char *suffix, bool vm_must_add_frozenbit); -static bool new_cluster_needs_fsm(FileNameMap *map); /* @@ -176,8 +174,7 @@ transfer_single_new_db(FileNameMap *maps, int size, char *old_tablespace) /* * Copy/link any fsm and vm files, if they exist */ - if (new_cluster_needs_fsm(&maps[mapnum])) - transfer_relfile(&maps[mapnum], "_fsm", vm_must_add_frozenbit); + transfer_relfile(&maps[mapnum], "_fsm", vm_must_add_frozenbit); if (vm_crashsafe_match) transfer_relfile(&maps[mapnum], "_vm", vm_must_add_frozenbit); } @@ -281,61 +278,3 @@ transfer_relfile(FileNameMap *map, const char *type_suffix, bool vm_must_add_fro } } } - -/* - * new_cluster_needs_fsm() - * - * Return false for small heaps if we're upgrading across PG 12, the first - * version where small heap relations don't have FSMs by default. - */ -static bool -new_cluster_needs_fsm(FileNameMap *map) -{ - char old_primary_file[MAXPGPATH]; - struct stat statbuf; - - /* fsm/vm files added in PG 8.4 */ - Assert(GET_MAJOR_VERSION(old_cluster.major_version) >= 804); - - if (!(GET_MAJOR_VERSION(old_cluster.major_version) <= 1100 && - GET_MAJOR_VERSION(new_cluster.major_version) >= 1200)) - return true; - - /* Always transfer FSMs of non-heap relations. */ - if (map->relkind != RELKIND_RELATION && - map->relkind != RELKIND_TOASTVALUE) - return true; - - /* - * If pg_class.relpages falsely reports that the heap is above the - * threshold, we will transfer a FSM when we don't need to, but this is - * harmless. - */ - if (map->relpages > HEAP_FSM_CREATION_THRESHOLD) - return true; - - /* Determine path of the primary file. */ - snprintf(old_primary_file, sizeof(old_primary_file), "%s%s/%u/%u", - map->old_tablespace, - map->old_tablespace_suffix, - map->old_db_oid, - map->old_relfilenode); - - /* - * If pg_class.relpages falsely reports that the heap is below the - * threshold, a FSM would be skipped when we actually need it. To guard - * against this, we verify the size of the primary file. - */ - if (stat(old_primary_file, &statbuf) != 0) - { - pg_fatal("error while checking for file existence \"%s.%s\" (\"%s\"): %s\n", - map->nspname, map->relname, old_primary_file, strerror(errno)); - - /* Keep compiler quiet. */ - return false; - } - else if (statbuf.st_size > HEAP_FSM_CREATION_THRESHOLD * BLCKSZ) - return true; - else - return false; -} diff --git a/src/include/storage/freespace.h b/src/include/storage/freespace.h index dbaae651c58..8b000334382 100644 --- a/src/include/storage/freespace.h +++ b/src/include/storage/freespace.h @@ -18,20 +18,15 @@ #include "storage/relfilenode.h" #include "utils/relcache.h" -/* Only create the FSM if the heap has greater than this many blocks */ -#define HEAP_FSM_CREATION_THRESHOLD 4 - /* prototypes for public functions in freespace.c */ extern Size GetRecordedFreeSpace(Relation rel, BlockNumber heapBlk); -extern BlockNumber GetPageWithFreeSpace(Relation rel, Size spaceNeeded, - bool check_fsm_only); +extern BlockNumber GetPageWithFreeSpace(Relation rel, Size spaceNeeded); extern BlockNumber RecordAndGetPageWithFreeSpace(Relation rel, BlockNumber oldPage, Size oldSpaceAvail, Size spaceNeeded); extern void RecordPageWithFreeSpace(Relation rel, BlockNumber heapBlk, - Size spaceAvail, BlockNumber nblocks); -extern void FSMClearLocalMap(void); + Size spaceAvail); extern void XLogRecordPageWithFreeSpace(RelFileNode rnode, BlockNumber heapBlk, Size spaceAvail); diff --git a/src/test/regress/expected/fsm.out b/src/test/regress/expected/fsm.out deleted file mode 100644 index 698d4b0be56..00000000000 --- a/src/test/regress/expected/fsm.out +++ /dev/null @@ -1,73 +0,0 @@ --- --- Free Space Map test --- -SELECT current_setting('block_size')::integer AS blocksize, -current_setting('block_size')::integer / 8 AS strsize -\gset -CREATE TABLE fsm_check_size (num int, str text); --- Fill 3 blocks with one record each -ALTER TABLE fsm_check_size SET (fillfactor=15); -INSERT INTO fsm_check_size SELECT i, rpad('', :strsize, 'a') -FROM generate_series(1,3) i; --- There should be no FSM -VACUUM fsm_check_size; -SELECT pg_relation_size('fsm_check_size', 'main') / :blocksize AS heap_nblocks, -pg_relation_size('fsm_check_size', 'fsm') / :blocksize AS fsm_nblocks; - heap_nblocks | fsm_nblocks ---------------+------------- - 3 | 0 -(1 row) - --- The following operations are for testing the functionality of the local --- in-memory map. In particular, we want to be able to insert into some --- other block than the one at the end of the heap, without using a FSM. --- Fill most of the last block -ALTER TABLE fsm_check_size SET (fillfactor=100); -INSERT INTO fsm_check_size SELECT i, rpad('', :strsize, 'a') -FROM generate_series(101,105) i; --- Make sure records can go into any block but the last one -ALTER TABLE fsm_check_size SET (fillfactor=30); --- Insert large record and make sure it does not cause the relation to extend -INSERT INTO fsm_check_size VALUES (111, rpad('', :strsize, 'a')); -VACUUM fsm_check_size; -SELECT pg_relation_size('fsm_check_size', 'main') / :blocksize AS heap_nblocks, -pg_relation_size('fsm_check_size', 'fsm') / :blocksize AS fsm_nblocks; - heap_nblocks | fsm_nblocks ---------------+------------- - 3 | 0 -(1 row) - --- Extend table with enough blocks to exceed the FSM threshold -DO $$ -DECLARE curtid tid; -num int; -BEGIN -num = 11; - LOOP - INSERT INTO fsm_check_size VALUES (num, 'b') RETURNING ctid INTO curtid; - EXIT WHEN curtid >= tid '(4, 0)'; - num = num + 1; - END LOOP; -END; -$$; -VACUUM fsm_check_size; -SELECT pg_relation_size('fsm_check_size', 'fsm') / :blocksize AS fsm_nblocks; - fsm_nblocks -------------- - 3 -(1 row) - --- Add long random string to extend TOAST table to 1 block -INSERT INTO fsm_check_size -VALUES(0, (SELECT string_agg(md5(chr(i)), '') - FROM generate_series(1, :blocksize / 100) i)); -VACUUM fsm_check_size; -SELECT pg_relation_size(reltoastrelid, 'main') / :blocksize AS toast_nblocks, -pg_relation_size(reltoastrelid, 'fsm') / :blocksize AS toast_fsm_nblocks -FROM pg_class WHERE relname = 'fsm_check_size'; - toast_nblocks | toast_fsm_nblocks ----------------+------------------- - 1 | 0 -(1 row) - -DROP TABLE fsm_check_size; diff --git a/src/test/regress/parallel_schedule b/src/test/regress/parallel_schedule index 3b7759b6faa..312b7c22c50 100644 --- a/src/test/regress/parallel_schedule +++ b/src/test/regress/parallel_schedule @@ -20,7 +20,7 @@ test: boolean char name varchar text int2 int4 int8 oid float4 float8 bit numeri # strings depends on char, varchar and text # numerology depends on int2, int4, int8, float4, float8 # ---------- -test: strings numerology point lseg line box path polygon circle date time timetz timestamp timestamptz interval inet macaddr macaddr8 tstypes fsm +test: strings numerology point lseg line box path polygon circle date time timetz timestamp timestamptz interval inet macaddr macaddr8 tstypes # ---------- # Another group of parallel tests diff --git a/src/test/regress/serial_schedule b/src/test/regress/serial_schedule index 43523c6fc62..9d44c313298 100644 --- a/src/test/regress/serial_schedule +++ b/src/test/regress/serial_schedule @@ -40,7 +40,6 @@ test: inet test: macaddr test: macaddr8 test: tstypes -test: fsm test: geometry test: horology test: regex diff --git a/src/test/regress/sql/fsm.sql b/src/test/regress/sql/fsm.sql deleted file mode 100644 index 7295e0439b2..00000000000 --- a/src/test/regress/sql/fsm.sql +++ /dev/null @@ -1,66 +0,0 @@ --- --- Free Space Map test --- -SELECT current_setting('block_size')::integer AS blocksize, -current_setting('block_size')::integer / 8 AS strsize -\gset - -CREATE TABLE fsm_check_size (num int, str text); - --- Fill 3 blocks with one record each -ALTER TABLE fsm_check_size SET (fillfactor=15); -INSERT INTO fsm_check_size SELECT i, rpad('', :strsize, 'a') -FROM generate_series(1,3) i; - --- There should be no FSM -VACUUM fsm_check_size; -SELECT pg_relation_size('fsm_check_size', 'main') / :blocksize AS heap_nblocks, -pg_relation_size('fsm_check_size', 'fsm') / :blocksize AS fsm_nblocks; - --- The following operations are for testing the functionality of the local --- in-memory map. In particular, we want to be able to insert into some --- other block than the one at the end of the heap, without using a FSM. - --- Fill most of the last block -ALTER TABLE fsm_check_size SET (fillfactor=100); -INSERT INTO fsm_check_size SELECT i, rpad('', :strsize, 'a') -FROM generate_series(101,105) i; - --- Make sure records can go into any block but the last one -ALTER TABLE fsm_check_size SET (fillfactor=30); - --- Insert large record and make sure it does not cause the relation to extend -INSERT INTO fsm_check_size VALUES (111, rpad('', :strsize, 'a')); - -VACUUM fsm_check_size; -SELECT pg_relation_size('fsm_check_size', 'main') / :blocksize AS heap_nblocks, -pg_relation_size('fsm_check_size', 'fsm') / :blocksize AS fsm_nblocks; - --- Extend table with enough blocks to exceed the FSM threshold -DO $$ -DECLARE curtid tid; -num int; -BEGIN -num = 11; - LOOP - INSERT INTO fsm_check_size VALUES (num, 'b') RETURNING ctid INTO curtid; - EXIT WHEN curtid >= tid '(4, 0)'; - num = num + 1; - END LOOP; -END; -$$; - -VACUUM fsm_check_size; -SELECT pg_relation_size('fsm_check_size', 'fsm') / :blocksize AS fsm_nblocks; - --- Add long random string to extend TOAST table to 1 block -INSERT INTO fsm_check_size -VALUES(0, (SELECT string_agg(md5(chr(i)), '') - FROM generate_series(1, :blocksize / 100) i)); - -VACUUM fsm_check_size; -SELECT pg_relation_size(reltoastrelid, 'main') / :blocksize AS toast_nblocks, -pg_relation_size(reltoastrelid, 'fsm') / :blocksize AS toast_fsm_nblocks -FROM pg_class WHERE relname = 'fsm_check_size'; - -DROP TABLE fsm_check_size; |