diff --git a/contrib/pg_prewarm/autoprewarm.c b/contrib/pg_prewarm/autoprewarm.c index 73485a2323cf..195dc4b773b1 100644 --- a/contrib/pg_prewarm/autoprewarm.c +++ b/contrib/pg_prewarm/autoprewarm.c @@ -41,6 +41,7 @@ #include "storage/latch.h" #include "storage/lwlock.h" #include "storage/procsignal.h" +#include "storage/read_stream.h" #include "storage/smgr.h" #include "tcop/tcopprot.h" #include "utils/guc.h" @@ -75,6 +76,28 @@ typedef struct AutoPrewarmSharedState int prewarmed_blocks; } AutoPrewarmSharedState; +/* + * Private data passed through the read stream API for our use in the + * callback. + */ +typedef struct AutoPrewarmReadStreamData +{ + /* The array of records containing the blocks we should prewarm. */ + BlockInfoRecord *block_info; + + /* + * pos is the read stream callback's index into block_info. Because the + * read stream may read ahead, pos is likely to be ahead of the index in + * the main loop in autoprewarm_database_main(). + */ + int pos; + Oid tablespace; + RelFileNumber filenumber; + ForkNumber forknum; + BlockNumber nblocks; +} AutoPrewarmReadStreamData; + + PGDLLEXPORT void autoprewarm_main(Datum main_arg); PGDLLEXPORT void autoprewarm_database_main(Datum main_arg); @@ -422,6 +445,54 @@ apw_load_buffers(void) apw_state->prewarmed_blocks, num_elements))); } +/* + * Return the next block number of a specific relation and fork to read + * according to the array of BlockInfoRecord. + */ +static BlockNumber +apw_read_stream_next_block(ReadStream *stream, + void *callback_private_data, + void *per_buffer_data) +{ + AutoPrewarmReadStreamData *p = callback_private_data; + + CHECK_FOR_INTERRUPTS(); + + while (p->pos < apw_state->prewarm_stop_idx) + { + BlockInfoRecord blk = p->block_info[p->pos]; + + if (!have_free_buffer()) + { + p->pos = apw_state->prewarm_stop_idx; + return InvalidBlockNumber; + } + + if (blk.tablespace != p->tablespace) + return InvalidBlockNumber; + + if (blk.filenumber != p->filenumber) + return InvalidBlockNumber; + + if (blk.forknum != p->forknum) + return InvalidBlockNumber; + + p->pos++; + + /* + * Check whether blocknum is valid and within fork file size. + * Fast-forward through any invalid blocks. We want p->pos to reflect + * the location of the next relation or fork before ending the stream. + */ + if (blk.blocknum >= p->nblocks) + continue; + + return blk.blocknum; + } + + return InvalidBlockNumber; +} + /* * Prewarm all blocks for one database (and possibly also global objects, if * those got grouped with this database). @@ -429,11 +500,9 @@ apw_load_buffers(void) void autoprewarm_database_main(Datum main_arg) { - int pos; BlockInfoRecord *block_info; - Relation rel = NULL; - BlockNumber nblocks = 0; - BlockInfoRecord *old_blk = NULL; + int i; + BlockInfoRecord blk; dsm_segment *seg; /* Establish signal handlers; once that's done, unblock signals. */ @@ -449,108 +518,140 @@ autoprewarm_database_main(Datum main_arg) errmsg("could not map dynamic shared memory segment"))); BackgroundWorkerInitializeConnectionByOid(apw_state->database, InvalidOid, 0); block_info = (BlockInfoRecord *) dsm_segment_address(seg); - pos = apw_state->prewarm_start_idx; + + i = apw_state->prewarm_start_idx; + blk = block_info[i]; /* * Loop until we run out of blocks to prewarm or until we run out of free * buffers. */ - while (pos < apw_state->prewarm_stop_idx && have_free_buffer()) + while (i < apw_state->prewarm_stop_idx && have_free_buffer()) { - BlockInfoRecord *blk = &block_info[pos++]; - Buffer buf; - - CHECK_FOR_INTERRUPTS(); + Oid tablespace = blk.tablespace; + RelFileNumber filenumber = blk.filenumber; + Oid reloid; + Relation rel; /* - * Quit if we've reached records for another database. If previous - * blocks are of some global objects, then continue pre-warming. + * All blocks between prewarm_start_idx and prewarm_stop_idx should + * belong either to global objects or the same database. */ - if (old_blk != NULL && old_blk->database != blk->database && - old_blk->database != 0) - break; + Assert(blk.database == apw_state->database || blk.database == 0); - /* - * As soon as we encounter a block of a new relation, close the old - * relation. Note that rel will be NULL if try_relation_open failed - * previously; in that case, there is nothing to close. - */ - if (old_blk != NULL && old_blk->filenumber != blk->filenumber && - rel != NULL) - { - relation_close(rel, AccessShareLock); - rel = NULL; - CommitTransactionCommand(); - } + StartTransactionCommand(); - /* - * Try to open each new relation, but only once, when we first - * encounter it. If it's been dropped, skip the associated blocks. - */ - if (old_blk == NULL || old_blk->filenumber != blk->filenumber) + reloid = RelidByRelfilenumber(blk.tablespace, blk.filenumber); + if (!OidIsValid(reloid) || + (rel = try_relation_open(reloid, AccessShareLock)) == NULL) { - Oid reloid; + /* We failed to open the relation, so there is nothing to close. */ + CommitTransactionCommand(); - Assert(rel == NULL); - StartTransactionCommand(); - reloid = RelidByRelfilenumber(blk->tablespace, blk->filenumber); - if (OidIsValid(reloid)) - rel = try_relation_open(reloid, AccessShareLock); + /* + * Fast-forward to the next relation. We want to skip all of the + * other records referencing this relation since we know we can't + * open it. That way, we avoid repeatedly trying and failing to + * open the same relation. + */ + for (; i < apw_state->prewarm_stop_idx; i++) + { + blk = block_info[i]; + if (blk.tablespace != tablespace || + blk.filenumber != filenumber) + break; + } - if (!rel) - CommitTransactionCommand(); - } - if (!rel) - { - old_blk = blk; + /* Time to try and open our newfound relation */ continue; } - /* Once per fork, check for fork existence and size. */ - if (old_blk == NULL || - old_blk->filenumber != blk->filenumber || - old_blk->forknum != blk->forknum) + /* + * We have a relation; now let's loop until we find a valid fork of + * the relation or we run out of free buffers. Once we've read from + * all valid forks or run out of options, we'll close the relation and + * move on. + */ + while (i < apw_state->prewarm_stop_idx && + blk.tablespace == tablespace && + blk.filenumber == filenumber && + have_free_buffer()) { + ForkNumber forknum = blk.forknum; + BlockNumber nblocks; + struct AutoPrewarmReadStreamData p; + ReadStream *stream; + Buffer buf; + /* * smgrexists is not safe for illegal forknum, hence check whether * the passed forknum is valid before using it in smgrexists. */ - if (blk->forknum > InvalidForkNumber && - blk->forknum <= MAX_FORKNUM && - smgrexists(RelationGetSmgr(rel), blk->forknum)) - nblocks = RelationGetNumberOfBlocksInFork(rel, blk->forknum); - else - nblocks = 0; - } + if (blk.forknum <= InvalidForkNumber || + blk.forknum > MAX_FORKNUM || + !smgrexists(RelationGetSmgr(rel), blk.forknum)) + { + /* + * Fast-forward to the next fork. We want to skip all of the + * other records referencing this fork since we already know + * it's not valid. + */ + for (; i < apw_state->prewarm_stop_idx; i++) + { + blk = block_info[i]; + if (blk.tablespace != tablespace || + blk.filenumber != filenumber || + blk.forknum != forknum) + break; + } + + /* Time to check if this newfound fork is valid */ + continue; + } - /* Check whether blocknum is valid and within fork file size. */ - if (blk->blocknum >= nblocks) - { - /* Move to next forknum. */ - old_blk = blk; - continue; - } + nblocks = RelationGetNumberOfBlocksInFork(rel, blk.forknum); - /* Prewarm buffer. */ - buf = ReadBufferExtended(rel, blk->forknum, blk->blocknum, RBM_NORMAL, - NULL); - if (BufferIsValid(buf)) - { - apw_state->prewarmed_blocks++; - ReleaseBuffer(buf); - } + p = (struct AutoPrewarmReadStreamData) + { + .block_info = block_info, + .pos = i, + .tablespace = tablespace, + .filenumber = filenumber, + .forknum = forknum, + .nblocks = nblocks, + }; + + stream = read_stream_begin_relation(READ_STREAM_FULL, + NULL, + rel, + p.forknum, + apw_read_stream_next_block, + &p, + 0); - old_blk = blk; - } + /* + * Loop until we've prewarmed all the blocks from this fork. The + * read stream callback will check that we still have free buffers + * before requesting each block from the read stream API. + */ + while ((buf = read_stream_next_buffer(stream, NULL)) != InvalidBuffer) + { + apw_state->prewarmed_blocks++; + ReleaseBuffer(buf); + } - dsm_detach(seg); + read_stream_end(stream); + + /* Advance i past all the blocks just prewarmed. */ + i = p.pos; + blk = block_info[i]; + } - /* Release lock on previous relation. */ - if (rel) - { relation_close(rel, AccessShareLock); CommitTransactionCommand(); } + + dsm_detach(seg); } /* diff --git a/src/tools/pgindent/typedefs.list b/src/tools/pgindent/typedefs.list index 8f28d8ff28eb..5ac290fae789 100644 --- a/src/tools/pgindent/typedefs.list +++ b/src/tools/pgindent/typedefs.list @@ -175,6 +175,7 @@ AttributeOpts AuthRequest AuthToken AutoPrewarmSharedState +AutoPrewarmReadStreamData AutoVacOpts AutoVacuumShmemStruct AutoVacuumWorkItem