static void
ginVacuumPostingTree(GinVacuumState *gvs, BlockNumber rootBlkno)
{
Buffer rootBuffer = InvalidBuffer;
DataPageDeleteStack root,
*ptr,
*tmp;
if (ginVacuumPostingTreeLeaves(gvs, rootBlkno, TRUE, &rootBuffer) == FALSE)
{
Assert(rootBuffer == InvalidBuffer);
return;
}
memset(&root, 0, sizeof(DataPageDeleteStack));
root.leftBlkno = InvalidBlockNumber;
root.isRoot = TRUE;
vacuum_delay_point();
ginScanToDelete(gvs, rootBlkno, TRUE, &root, InvalidOffsetNumber);
ptr = root.child;
while (ptr)
{
tmp = ptr->child;
pfree(ptr);
ptr = tmp;
}
UnlockReleaseBuffer(rootBuffer);
}
/*
* Post-VACUUM cleanup.
*
* Result: a palloc'd struct containing statistical info for VACUUM displays.
*/
IndexBulkDeleteResult *
blvacuumcleanup(IndexVacuumInfo *info, IndexBulkDeleteResult *stats)
{
Relation index = info->index;
BlockNumber npages,
blkno;
BlockNumber totFreePages;
if (info->analyze_only)
return stats;
if (stats == NULL)
stats = (IndexBulkDeleteResult *) palloc0(sizeof(IndexBulkDeleteResult));
/*
* Iterate over the pages: insert deleted pages into FSM and collect
* statistics.
*/
npages = RelationGetNumberOfBlocks(index);
totFreePages = 0;
for (blkno = BLOOM_HEAD_BLKNO; blkno < npages; blkno++)
{
Buffer buffer;
Page page;
vacuum_delay_point();
buffer = ReadBufferExtended(index, MAIN_FORKNUM, blkno,
RBM_NORMAL, info->strategy);
LockBuffer(buffer, BUFFER_LOCK_SHARE);
page = BufferGetPage(buffer, NULL, NULL, BGP_NO_SNAPSHOT_TEST);
if (BloomPageIsDeleted(page))
{
RecordFreeIndexPage(index, blkno);
totFreePages++;
}
else
{
stats->num_index_tuples += BloomPageGetMaxOffset(page);
stats->estimated_count += BloomPageGetMaxOffset(page);
}
UnlockReleaseBuffer(buffer);
}
IndexFreeSpaceMapVacuum(info->index);
stats->pages_free = totFreePages;
stats->num_pages = RelationGetNumberOfBlocks(index);
return stats;
}
/*
* lazy_vacuum_heap() -- second pass over the heap
*
* This routine marks dead tuples as unused and compacts out free
* space on their pages. Pages not having dead tuples recorded from
* lazy_scan_heap are not visited at all.
*
* 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.
*/
static void
lazy_vacuum_heap(Relation onerel, LVRelStats *vacrelstats)
{
MIRROREDLOCK_BUFMGR_DECLARE;
int tupindex;
int npages;
PGRUsage ru0;
pg_rusage_init(&ru0);
npages = 0;
tupindex = 0;
/* Fetch gp_persistent_relation_node information that will be added to XLOG record. */
RelationFetchGpRelationNodeForXLog(onerel);
while (tupindex < vacrelstats->num_dead_tuples)
{
BlockNumber tblk;
Buffer buf;
Page page;
vacuum_delay_point();
tblk = ItemPointerGetBlockNumber(&vacrelstats->dead_tuples[tupindex]);
/* -------- MirroredLock ---------- */
MIRROREDLOCK_BUFMGR_LOCK;
buf = ReadBufferWithStrategy(onerel, tblk, vac_strategy);
LockBufferForCleanup(buf);
tupindex = lazy_vacuum_page(onerel, tblk, buf, tupindex, vacrelstats);
/* Now that we've compacted the page, record its available space */
page = BufferGetPage(buf);
lazy_record_free_space(vacrelstats, tblk,
PageGetHeapFreeSpace(page));
UnlockReleaseBuffer(buf);
MIRROREDLOCK_BUFMGR_UNLOCK;
/* -------- MirroredLock ---------- */
npages++;
}
ereport(elevel,
(errmsg("\"%s\": removed %d row versions in %d pages",
RelationGetRelationName(onerel),
tupindex, npages),
errdetail("%s.",
pg_rusage_show(&ru0))));
}
/*
* lazy_vacuum_heap() -- second pass over the heap
*
* This routine marks dead tuples as unused and compacts out free
* space on their pages. Pages not having dead tuples recorded from
* lazy_scan_heap are not visited at all.
*
* 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.
*/
static void
lazy_vacuum_heap(Relation onerel, LVRelStats *vacrelstats)
{
int tupindex;
int npages;
PGRUsage ru0;
pg_rusage_init(&ru0);
npages = 0;
tupindex = 0;
while (tupindex < vacrelstats->num_dead_tuples)
{
BlockNumber tblk;
Buffer buf;
Page page;
Size freespace;
vacuum_delay_point();
tblk = ItemPointerGetBlockNumber(&vacrelstats->dead_tuples[tupindex]);
buf = ReadBufferExtended(onerel, MAIN_FORKNUM, tblk, RBM_NORMAL,
vac_strategy);
if (!ConditionalLockBufferForCleanup(buf))
{
ReleaseBuffer(buf);
++tupindex;
continue;
}
tupindex = lazy_vacuum_page(onerel, tblk, buf, tupindex, vacrelstats);
/* Now that we've compacted the page, record its available space */
page = BufferGetPage(buf);
freespace = PageGetHeapFreeSpace(page);
UnlockReleaseBuffer(buf);
RecordPageWithFreeSpace(onerel, tblk, freespace);
npages++;
}
ereport(elevel,
(errmsg("\"%s\": removed %d row versions in %d pages",
RelationGetRelationName(onerel),
tupindex, npages),
errdetail("%s.",
pg_rusage_show(&ru0))));
}
/*
* Helper function to perform deletion of index entries from a bucket.
*
* This function expects that the caller has acquired a cleanup lock on the
* primary bucket page, and will return with a write lock again held on the
* primary bucket page. The lock won't necessarily be held continuously,
* though, because we'll release it when visiting overflow pages.
*
* It would be very bad if this function cleaned a page while some other
* backend was in the midst of scanning it, because hashgettuple assumes
* that the next valid TID will be greater than or equal to the current
* valid TID. There can't be any concurrent scans in progress when we first
* enter this function because of the cleanup lock we hold on the primary
* bucket page, but as soon as we release that lock, there might be. We
* handle that by conspiring to prevent those scans from passing our cleanup
* scan. To do that, we lock the next page in the bucket chain before
* releasing the lock on the previous page. (This type of lock chaining is
* not ideal, so we might want to look for a better solution at some point.)
*
* We need to retain a pin on the primary bucket to ensure that no concurrent
* split can start.
*/
void
hashbucketcleanup(Relation rel, Bucket cur_bucket, Buffer bucket_buf,
BlockNumber bucket_blkno, BufferAccessStrategy bstrategy,
uint32 maxbucket, uint32 highmask, uint32 lowmask,
double *tuples_removed, double *num_index_tuples,
bool split_cleanup,
IndexBulkDeleteCallback callback, void *callback_state)
{
BlockNumber blkno;
Buffer buf;
Bucket new_bucket PG_USED_FOR_ASSERTS_ONLY = InvalidBucket;
bool bucket_dirty = false;
blkno = bucket_blkno;
buf = bucket_buf;
if (split_cleanup)
new_bucket = _hash_get_newbucket_from_oldbucket(rel, cur_bucket,
lowmask, maxbucket);
/* Scan each page in bucket */
for (;;)
{
HashPageOpaque opaque;
OffsetNumber offno;
OffsetNumber maxoffno;
Buffer next_buf;
Page page;
OffsetNumber deletable[MaxOffsetNumber];
int ndeletable = 0;
bool retain_pin = false;
bool clear_dead_marking = false;
vacuum_delay_point();
page = BufferGetPage(buf);
opaque = (HashPageOpaque) PageGetSpecialPointer(page);
/* Scan each tuple in page */
maxoffno = PageGetMaxOffsetNumber(page);
for (offno = FirstOffsetNumber;
offno <= maxoffno;
offno = OffsetNumberNext(offno))
{
ItemPointer htup;
IndexTuple itup;
Bucket bucket;
bool kill_tuple = false;
itup = (IndexTuple) PageGetItem(page,
PageGetItemId(page, offno));
htup = &(itup->t_tid);
/*
* To remove the dead tuples, we strictly want to rely on results
* of callback function. refer btvacuumpage for detailed reason.
*/
if (callback && callback(htup, callback_state))
{
kill_tuple = true;
if (tuples_removed)
*tuples_removed += 1;
}
else if (split_cleanup)
{
/* delete the tuples that are moved by split. */
bucket = _hash_hashkey2bucket(_hash_get_indextuple_hashkey(itup),
maxbucket,
highmask,
lowmask);
/* mark the item for deletion */
if (bucket != cur_bucket)
{
/*
* We expect tuples to either belong to current bucket or
* new_bucket. This is ensured because we don't allow
* further splits from bucket that contains garbage. See
* comments in _hash_expandtable.
*/
Assert(bucket == new_bucket);
kill_tuple = true;
}
}
if (kill_tuple)
{
/* mark the item for deletion */
deletable[ndeletable++] = offno;
}
else
{
/* we're keeping it, so count it */
if (num_index_tuples)
*num_index_tuples += 1;
}
}
/* retain the pin on primary bucket page till end of bucket scan */
if (blkno == bucket_blkno)
retain_pin = true;
else
retain_pin = false;
blkno = opaque->hasho_nextblkno;
/*
* Apply deletions, advance to next page and write page if needed.
*/
if (ndeletable > 0)
{
/* No ereport(ERROR) until changes are logged */
START_CRIT_SECTION();
PageIndexMultiDelete(page, deletable, ndeletable);
bucket_dirty = true;
/*
* Let us mark the page as clean if vacuum removes the DEAD tuples
* from an index page. We do this by clearing
* LH_PAGE_HAS_DEAD_TUPLES flag.
*/
if (tuples_removed && *tuples_removed > 0 &&
H_HAS_DEAD_TUPLES(opaque))
{
opaque->hasho_flag &= ~LH_PAGE_HAS_DEAD_TUPLES;
clear_dead_marking = true;
}
MarkBufferDirty(buf);
/* XLOG stuff */
if (RelationNeedsWAL(rel))
{
xl_hash_delete xlrec;
XLogRecPtr recptr;
xlrec.clear_dead_marking = clear_dead_marking;
xlrec.is_primary_bucket_page = (buf == bucket_buf) ? true : false;
XLogBeginInsert();
XLogRegisterData((char *) &xlrec, SizeOfHashDelete);
/*
* bucket buffer needs to be registered to ensure that we can
* acquire a cleanup lock on it during replay.
*/
if (!xlrec.is_primary_bucket_page)
XLogRegisterBuffer(0, bucket_buf, REGBUF_STANDARD | REGBUF_NO_IMAGE);
XLogRegisterBuffer(1, buf, REGBUF_STANDARD);
XLogRegisterBufData(1, (char *) deletable,
ndeletable * sizeof(OffsetNumber));
recptr = XLogInsert(RM_HASH_ID, XLOG_HASH_DELETE);
PageSetLSN(BufferGetPage(buf), recptr);
}
END_CRIT_SECTION();
}
/* bail out if there are no more pages to scan. */
if (!BlockNumberIsValid(blkno))
break;
next_buf = _hash_getbuf_with_strategy(rel, blkno, HASH_WRITE,
LH_OVERFLOW_PAGE,
bstrategy);
/*
* release the lock on previous page after acquiring the lock on next
* page
*/
if (retain_pin)
LockBuffer(buf, BUFFER_LOCK_UNLOCK);
else
_hash_relbuf(rel, buf);
buf = next_buf;
}
/*
* lock the bucket page to clear the garbage flag and squeeze the bucket.
* if the current buffer is same as bucket buffer, then we already have
* lock on bucket page.
*/
if (buf != bucket_buf)
{
_hash_relbuf(rel, buf);
LockBuffer(bucket_buf, BUFFER_LOCK_EXCLUSIVE);
}
/*
* Clear the garbage flag from bucket after deleting the tuples that are
* moved by split. We purposefully clear the flag before squeeze bucket,
* so that after restart, vacuum shouldn't again try to delete the moved
* by split tuples.
*/
if (split_cleanup)
{
HashPageOpaque bucket_opaque;
Page page;
page = BufferGetPage(bucket_buf);
bucket_opaque = (HashPageOpaque) PageGetSpecialPointer(page);
/* No ereport(ERROR) until changes are logged */
START_CRIT_SECTION();
bucket_opaque->hasho_flag &= ~LH_BUCKET_NEEDS_SPLIT_CLEANUP;
MarkBufferDirty(bucket_buf);
/* XLOG stuff */
if (RelationNeedsWAL(rel))
{
XLogRecPtr recptr;
XLogBeginInsert();
XLogRegisterBuffer(0, bucket_buf, REGBUF_STANDARD);
recptr = XLogInsert(RM_HASH_ID, XLOG_HASH_SPLIT_CLEANUP);
PageSetLSN(page, recptr);
}
END_CRIT_SECTION();
}
/*
* If we have deleted anything, try to compact free space. For squeezing
* the bucket, we must have a cleanup lock, else it can impact the
* ordering of tuples for a scan that has started before it.
*/
if (bucket_dirty && IsBufferCleanupOK(bucket_buf))
_hash_squeezebucket(rel, cur_bucket, bucket_blkno, bucket_buf,
bstrategy);
else
LockBuffer(bucket_buf, BUFFER_LOCK_UNLOCK);
}
/*
* Move tuples from pending pages into regular GIN structure.
*
* This can be called concurrently by multiple backends, so it must cope.
* On first glance it looks completely not concurrent-safe and not crash-safe
* either. The reason it's okay is that multiple insertion of the same entry
* is detected and treated as a no-op by gininsert.c. If we crash after
* posting entries to the main index and before removing them from the
* pending list, it's okay because when we redo the posting later on, nothing
* bad will happen. Likewise, if two backends simultaneously try to post
* a pending entry into the main index, one will succeed and one will do
* nothing. We try to notice when someone else is a little bit ahead of
* us in the process, but that's just to avoid wasting cycles. Only the
* action of removing a page from the pending list really needs exclusive
* lock.
*
* vac_delay indicates that ginInsertCleanup is called from vacuum process,
* so call vacuum_delay_point() periodically.
* If stats isn't null, we count deleted pending pages into the counts.
*/
void
ginInsertCleanup(GinState *ginstate,
bool vac_delay, IndexBulkDeleteResult *stats)
{
Relation index = ginstate->index;
Buffer metabuffer,
buffer;
Page metapage,
page;
GinMetaPageData *metadata;
MemoryContext opCtx,
oldCtx;
BuildAccumulator accum;
KeyArray datums;
BlockNumber blkno;
metabuffer = ReadBuffer(index, GIN_METAPAGE_BLKNO);
LockBuffer(metabuffer, GIN_SHARE);
metapage = BufferGetPage(metabuffer);
metadata = GinPageGetMeta(metapage);
if (metadata->head == InvalidBlockNumber)
{
/* Nothing to do */
UnlockReleaseBuffer(metabuffer);
return;
}
/*
* Read and lock head of pending list
*/
blkno = metadata->head;
buffer = ReadBuffer(index, blkno);
LockBuffer(buffer, GIN_SHARE);
page = BufferGetPage(buffer);
LockBuffer(metabuffer, GIN_UNLOCK);
/*
* Initialize. All temporary space will be in opCtx
*/
opCtx = AllocSetContextCreate(CurrentMemoryContext,
"GIN insert cleanup temporary context",
ALLOCSET_DEFAULT_MINSIZE,
ALLOCSET_DEFAULT_INITSIZE,
ALLOCSET_DEFAULT_MAXSIZE);
oldCtx = MemoryContextSwitchTo(opCtx);
initKeyArray(&datums, 128);
ginInitBA(&accum);
accum.ginstate = ginstate;
/*
* At the top of this loop, we have pin and lock on the current page of
* the pending list. However, we'll release that before exiting the loop.
* Note we also have pin but not lock on the metapage.
*/
for (;;)
{
if (GinPageIsDeleted(page))
{
/* another cleanup process is running concurrently */
UnlockReleaseBuffer(buffer);
break;
}
/*
* read page's datums into accum
*/
processPendingPage(&accum, &datums, page, FirstOffsetNumber);
if (vac_delay)
vacuum_delay_point();
/*
* Is it time to flush memory to disk? Flush if we are at the end of
* the pending list, or if we have a full row and memory is getting
* full.
*
* XXX using up maintenance_work_mem here is probably unreasonably
* much, since vacuum might already be using that much.
*/
if (GinPageGetOpaque(page)->rightlink == InvalidBlockNumber ||
(GinPageHasFullRow(page) &&
(accum.allocatedMemory >= maintenance_work_mem * 1024L)))
{
ItemPointerData *list;
uint32 nlist;
Datum key;
GinNullCategory category;
OffsetNumber maxoff,
attnum;
/*
* Unlock current page to increase performance. Changes of page
* will be checked later by comparing maxoff after completion of
* memory flush.
*/
maxoff = PageGetMaxOffsetNumber(page);
LockBuffer(buffer, GIN_UNLOCK);
/*
* Moving collected data into regular structure can take
* significant amount of time - so, run it without locking pending
* list.
*/
ginBeginBAScan(&accum);
while ((list = ginGetBAEntry(&accum,
&attnum, &key, &category, &nlist)) != NULL)
{
ginEntryInsert(ginstate, attnum, key, category,
list, nlist, NULL);
if (vac_delay)
vacuum_delay_point();
}
/*
* Lock the whole list to remove pages
*/
LockBuffer(metabuffer, GIN_EXCLUSIVE);
LockBuffer(buffer, GIN_SHARE);
if (GinPageIsDeleted(page))
{
/* another cleanup process is running concurrently */
UnlockReleaseBuffer(buffer);
LockBuffer(metabuffer, GIN_UNLOCK);
break;
}
/*
* While we left the page unlocked, more stuff might have gotten
* added to it. If so, process those entries immediately. There
* shouldn't be very many, so we don't worry about the fact that
* we're doing this with exclusive lock. Insertion algorithm
* guarantees that inserted row(s) will not continue on next page.
* NOTE: intentionally no vacuum_delay_point in this loop.
*/
if (PageGetMaxOffsetNumber(page) != maxoff)
{
ginInitBA(&accum);
processPendingPage(&accum, &datums, page, maxoff + 1);
ginBeginBAScan(&accum);
while ((list = ginGetBAEntry(&accum,
&attnum, &key, &category, &nlist)) != NULL)
ginEntryInsert(ginstate, attnum, key, category,
list, nlist, NULL);
}
/*
* Remember next page - it will become the new list head
*/
blkno = GinPageGetOpaque(page)->rightlink;
UnlockReleaseBuffer(buffer); /* shiftList will do exclusive
* locking */
/*
* remove readed pages from pending list, at this point all
* content of readed pages is in regular structure
*/
if (shiftList(index, metabuffer, blkno, stats))
{
/* another cleanup process is running concurrently */
LockBuffer(metabuffer, GIN_UNLOCK);
break;
}
Assert(blkno == metadata->head);
LockBuffer(metabuffer, GIN_UNLOCK);
/*
* if we removed the whole pending list just exit
*/
if (blkno == InvalidBlockNumber)
break;
/*
* release memory used so far and reinit state
*/
MemoryContextReset(opCtx);
initKeyArray(&datums, datums.maxvalues);
ginInitBA(&accum);
}
else
{
blkno = GinPageGetOpaque(page)->rightlink;
UnlockReleaseBuffer(buffer);
}
/*
* Read next page in pending list
*/
CHECK_FOR_INTERRUPTS();
buffer = ReadBuffer(index, blkno);
LockBuffer(buffer, GIN_SHARE);
page = BufferGetPage(buffer);
}
ReleaseBuffer(metabuffer);
/* Clean up temporary space */
MemoryContextSwitchTo(oldCtx);
MemoryContextDelete(opCtx);
}
/*
* compute_tsvector_stats() -- compute statistics for a tsvector column
*
* This functions computes statistics that are useful for determining @@
* operations' selectivity, along with the fraction of non-null rows and
* average width.
*
* Instead of finding the most common values, as we do for most datatypes,
* we're looking for the most common lexemes. This is more useful, because
* there most probably won't be any two rows with the same tsvector and thus
* the notion of a MCV is a bit bogus with this datatype. With a list of the
* most common lexemes we can do a better job at figuring out @@ selectivity.
*
* For the same reasons we assume that tsvector columns are unique when
* determining the number of distinct values.
*
* The algorithm used is Lossy Counting, as proposed in the paper "Approximate
* frequency counts over data streams" by G. S. Manku and R. Motwani, in
* Proceedings of the 28th International Conference on Very Large Data Bases,
* Hong Kong, China, August 2002, section 4.2. The paper is available at
* http://www.vldb.org/conf/2002/S10P03.pdf
*
* The Lossy Counting (aka LC) algorithm goes like this:
* Let s be the threshold frequency for an item (the minimum frequency we
* are interested in) and epsilon the error margin for the frequency. Let D
* be a set of triples (e, f, delta), where e is an element value, f is that
* element's frequency (actually, its current occurrence count) and delta is
* the maximum error in f. We start with D empty and process the elements in
* batches of size w. (The batch size is also known as "bucket size" and is
* equal to 1/epsilon.) Let the current batch number be b_current, starting
* with 1. For each element e we either increment its f count, if it's
* already in D, or insert a new triple into D with values (e, 1, b_current
* - 1). After processing each batch we prune D, by removing from it all
* elements with f + delta <= b_current. After the algorithm finishes we
* suppress all elements from D that do not satisfy f >= (s - epsilon) * N,
* where N is the total number of elements in the input. We emit the
* remaining elements with estimated frequency f/N. The LC paper proves
* that this algorithm finds all elements with true frequency at least s,
* and that no frequency is overestimated or is underestimated by more than
* epsilon. Furthermore, given reasonable assumptions about the input
* distribution, the required table size is no more than about 7 times w.
*
* We set s to be the estimated frequency of the K'th word in a natural
* language's frequency table, where K is the target number of entries in
* the MCELEM array plus an arbitrary constant, meant to reflect the fact
* that the most common words in any language would usually be stopwords
* so we will not actually see them in the input. We assume that the
* distribution of word frequencies (including the stopwords) follows Zipf's
* law with an exponent of 1.
*
* Assuming Zipfian distribution, the frequency of the K'th word is equal
* to 1/(K * H(W)) where H(n) is 1/2 + 1/3 + ... + 1/n and W is the number of
* words in the language. Putting W as one million, we get roughly 0.07/K.
* Assuming top 10 words are stopwords gives s = 0.07/(K + 10). We set
* epsilon = s/10, which gives bucket width w = (K + 10)/0.007 and
* maximum expected hashtable size of about 1000 * (K + 10).
*
* Note: in the above discussion, s, epsilon, and f/N are in terms of a
* lexeme's frequency as a fraction of all lexemes seen in the input.
* However, what we actually want to store in the finished pg_statistic
* entry is each lexeme's frequency as a fraction of all rows that it occurs
* in. Assuming that the input tsvectors are correctly constructed, no
* lexeme occurs more than once per tsvector, so the final count f is a
* correct estimate of the number of input tsvectors it occurs in, and we
* need only change the divisor from N to nonnull_cnt to get the number we
* want.
*/
static void
compute_tsvector_stats(VacAttrStats *stats,
AnalyzeAttrFetchFunc fetchfunc,
int samplerows,
double totalrows)
{
int num_mcelem;
int null_cnt = 0;
double total_width = 0;
/* This is D from the LC algorithm. */
HTAB *lexemes_tab;
HASHCTL hash_ctl;
HASH_SEQ_STATUS scan_status;
/* This is the current bucket number from the LC algorithm */
int b_current;
/* This is 'w' from the LC algorithm */
int bucket_width;
int vector_no,
lexeme_no;
LexemeHashKey hash_key;
TrackItem *item;
/*
* We want statistics_target * 10 lexemes in the MCELEM array. This
* multiplier is pretty arbitrary, but is meant to reflect the fact that
* the number of individual lexeme values tracked in pg_statistic ought to
* be more than the number of values for a simple scalar column.
*/
num_mcelem = stats->attr->attstattarget * 10;
/*
* We set bucket width equal to (num_mcelem + 10) / 0.007 as per the
* comment above.
*/
bucket_width = (num_mcelem + 10) * 1000 / 7;
/*
* Create the hashtable. It will be in local memory, so we don't need to
* worry about overflowing the initial size. Also we don't need to pay any
* attention to locking and memory management.
*/
MemSet(&hash_ctl, 0, sizeof(hash_ctl));
hash_ctl.keysize = sizeof(LexemeHashKey);
hash_ctl.entrysize = sizeof(TrackItem);
hash_ctl.hash = lexeme_hash;
hash_ctl.match = lexeme_match;
hash_ctl.hcxt = CurrentMemoryContext;
lexemes_tab = hash_create("Analyzed lexemes table",
num_mcelem,
&hash_ctl,
HASH_ELEM | HASH_FUNCTION | HASH_COMPARE | HASH_CONTEXT);
/* Initialize counters. */
b_current = 1;
lexeme_no = 0;
/* Loop over the tsvectors. */
for (vector_no = 0; vector_no < samplerows; vector_no++)
{
Datum value;
bool isnull;
TSVector vector;
WordEntry *curentryptr;
char *lexemesptr;
int j;
vacuum_delay_point();
value = fetchfunc(stats, vector_no, &isnull);
/*
* Check for null/nonnull.
*/
if (isnull)
{
null_cnt++;
continue;
}
/*
* Add up widths for average-width calculation. Since it's a
* tsvector, we know it's varlena. As in the regular
* compute_minimal_stats function, we use the toasted width for this
* calculation.
*/
total_width += VARSIZE_ANY(DatumGetPointer(value));
/*
* Now detoast the tsvector if needed.
*/
vector = DatumGetTSVector(value);
/*
* We loop through the lexemes in the tsvector and add them to our
* tracking hashtable.
*/
lexemesptr = STRPTR(vector);
curentryptr = ARRPTR(vector);
for (j = 0; j < vector->size; j++)
{
bool found;
/*
* Construct a hash key. The key points into the (detoasted)
* tsvector value at this point, but if a new entry is created, we
* make a copy of it. This way we can free the tsvector value
* once we've processed all its lexemes.
*/
hash_key.lexeme = lexemesptr + curentryptr->pos;
hash_key.length = curentryptr->len;
/* Lookup current lexeme in hashtable, adding it if new */
item = (TrackItem *) hash_search(lexemes_tab,
(const void *) &hash_key,
HASH_ENTER, &found);
if (found)
{
/* The lexeme is already on the tracking list */
item->frequency++;
}
else
{
/* Initialize new tracking list element */
item->frequency = 1;
item->delta = b_current - 1;
item->key.lexeme = palloc(hash_key.length);
memcpy(item->key.lexeme, hash_key.lexeme, hash_key.length);
}
/* lexeme_no is the number of elements processed (ie N) */
lexeme_no++;
/* We prune the D structure after processing each bucket */
if (lexeme_no % bucket_width == 0)
{
prune_lexemes_hashtable(lexemes_tab, b_current);
b_current++;
}
/* Advance to the next WordEntry in the tsvector */
curentryptr++;
}
/* If the vector was toasted, free the detoasted copy. */
if (TSVectorGetDatum(vector) != value)
pfree(vector);
}
/* We can only compute real stats if we found some non-null values. */
if (null_cnt < samplerows)
{
int nonnull_cnt = samplerows - null_cnt;
int i;
TrackItem **sort_table;
int track_len;
int cutoff_freq;
int minfreq,
maxfreq;
stats->stats_valid = true;
/* Do the simple null-frac and average width stats */
stats->stanullfrac = (double) null_cnt / (double) samplerows;
stats->stawidth = total_width / (double) nonnull_cnt;
/* Assume it's a unique column (see notes above) */
stats->stadistinct = -1.0 * (1.0 - stats->stanullfrac);
/*
* Construct an array of the interesting hashtable items, that is,
* those meeting the cutoff frequency (s - epsilon)*N. Also identify
* the minimum and maximum frequencies among these items.
*
* Since epsilon = s/10 and bucket_width = 1/epsilon, the cutoff
* frequency is 9*N / bucket_width.
*/
cutoff_freq = 9 * lexeme_no / bucket_width;
i = hash_get_num_entries(lexemes_tab); /* surely enough space */
sort_table = (TrackItem **) palloc(sizeof(TrackItem *) * i);
hash_seq_init(&scan_status, lexemes_tab);
track_len = 0;
minfreq = lexeme_no;
maxfreq = 0;
while ((item = (TrackItem *) hash_seq_search(&scan_status)) != NULL)
{
if (item->frequency > cutoff_freq)
{
sort_table[track_len++] = item;
minfreq = Min(minfreq, item->frequency);
maxfreq = Max(maxfreq, item->frequency);
}
}
Assert(track_len <= i);
/* emit some statistics for debug purposes */
elog(DEBUG3, "tsvector_stats: target # mces = %d, bucket width = %d, "
"# lexemes = %d, hashtable size = %d, usable entries = %d",
num_mcelem, bucket_width, lexeme_no, i, track_len);
/*
* If we obtained more lexemes than we really want, get rid of those
* with least frequencies. The easiest way is to qsort the array into
* descending frequency order and truncate the array.
*/
if (num_mcelem < track_len)
{
qsort(sort_table, track_len, sizeof(TrackItem *),
trackitem_compare_frequencies_desc);
/* reset minfreq to the smallest frequency we're keeping */
minfreq = sort_table[num_mcelem - 1]->frequency;
}
else
num_mcelem = track_len;
/* Generate MCELEM slot entry */
if (num_mcelem > 0)
{
MemoryContext old_context;
Datum *mcelem_values;
float4 *mcelem_freqs;
/*
* We want to store statistics sorted on the lexeme value using
* first length, then byte-for-byte comparison. The reason for
* doing length comparison first is that we don't care about the
* ordering so long as it's consistent, and comparing lengths
* first gives us a chance to avoid a strncmp() call.
*
* This is different from what we do with scalar statistics --
* they get sorted on frequencies. The rationale is that we
* usually search through most common elements looking for a
* specific value, so we can grab its frequency. When values are
* presorted we can employ binary search for that. See
* ts_selfuncs.c for a real usage scenario.
*/
qsort(sort_table, num_mcelem, sizeof(TrackItem *),
trackitem_compare_lexemes);
/* Must copy the target values into anl_context */
old_context = MemoryContextSwitchTo(stats->anl_context);
/*
* We sorted statistics on the lexeme value, but we want to be
* able to find out the minimal and maximal frequency without
* going through all the values. We keep those two extra
* frequencies in two extra cells in mcelem_freqs.
*
* (Note: the MCELEM statistics slot definition allows for a third
* extra number containing the frequency of nulls, but we don't
* create that for a tsvector column, since null elements aren't
* possible.)
*/
mcelem_values = (Datum *) palloc(num_mcelem * sizeof(Datum));
mcelem_freqs = (float4 *) palloc((num_mcelem + 2) * sizeof(float4));
/*
* See comments above about use of nonnull_cnt as the divisor for
* the final frequency estimates.
*/
for (i = 0; i < num_mcelem; i++)
{
TrackItem *item = sort_table[i];
mcelem_values[i] =
PointerGetDatum(cstring_to_text_with_len(item->key.lexeme,
item->key.length));
mcelem_freqs[i] = (double) item->frequency / (double) nonnull_cnt;
}
mcelem_freqs[i++] = (double) minfreq / (double) nonnull_cnt;
mcelem_freqs[i] = (double) maxfreq / (double) nonnull_cnt;
MemoryContextSwitchTo(old_context);
stats->stakind[0] = STATISTIC_KIND_MCELEM;
stats->staop[0] = TextEqualOperator;
stats->stacoll[0] = DEFAULT_COLLATION_OID;
stats->stanumbers[0] = mcelem_freqs;
/* See above comment about two extra frequency fields */
stats->numnumbers[0] = num_mcelem + 2;
stats->stavalues[0] = mcelem_values;
stats->numvalues[0] = num_mcelem;
/* We are storing text values */
stats->statypid[0] = TEXTOID;
stats->statyplen[0] = -1; /* typlen, -1 for varlena */
stats->statypbyval[0] = false;
stats->statypalign[0] = 'i';
}
}
else
{
/* We found only nulls; assume the column is entirely null */
stats->stats_valid = true;
stats->stanullfrac = 1.0;
stats->stawidth = 0; /* "unknown" */
stats->stadistinct = 0.0; /* "unknown" */
}
/*
* We don't need to bother cleaning up any of our temporary palloc's. The
* hashtable should also go away, as it used a child memory context.
*/
}
/*
* Assumes that the segment file lock is already held.
* Assumes that the segment file should be compacted.
*/
static bool
AOCSSegmentFileFullCompaction(Relation aorel,
AOCSInsertDesc insertDesc,
AOCSFileSegInfo *fsinfo,
Snapshot snapshot)
{
const char *relname;
AppendOnlyVisimap visiMap;
AOCSScanDesc scanDesc;
TupleDesc tupDesc;
TupleTableSlot *slot;
int compact_segno;
int64 movedTupleCount = 0;
ResultRelInfo *resultRelInfo;
MemTupleBinding *mt_bind;
EState *estate;
bool *proj;
int i;
AOTupleId *aoTupleId;
int64 tupleCount = 0;
int64 tuplePerPage = INT_MAX;
Assert(Gp_role == GP_ROLE_EXECUTE || Gp_role == GP_ROLE_UTILITY);
Assert(RelationIsAoCols(aorel));
Assert(insertDesc);
compact_segno = fsinfo->segno;
if (fsinfo->varblockcount > 0)
{
tuplePerPage = fsinfo->total_tupcount / fsinfo->varblockcount;
}
relname = RelationGetRelationName(aorel);
AppendOnlyVisimap_Init(&visiMap,
aorel->rd_appendonly->visimaprelid,
aorel->rd_appendonly->visimapidxid,
ShareLock,
snapshot);
elogif(Debug_appendonly_print_compaction,
LOG, "Compact AO segfile %d, relation %sd",
compact_segno, relname);
proj = palloc0(sizeof(bool) * RelationGetNumberOfAttributes(aorel));
for (i = 0; i < RelationGetNumberOfAttributes(aorel); ++i)
{
proj[i] = true;
}
scanDesc = aocs_beginrangescan(aorel,
snapshot, snapshot,
&compact_segno, 1, NULL, proj);
tupDesc = RelationGetDescr(aorel);
slot = MakeSingleTupleTableSlot(tupDesc);
mt_bind = create_memtuple_binding(tupDesc);
/*
* We need a ResultRelInfo and an EState so we can use the regular
* executor's index-entry-making machinery.
*/
estate = CreateExecutorState();
resultRelInfo = makeNode(ResultRelInfo);
resultRelInfo->ri_RangeTableIndex = 1; /* dummy */
resultRelInfo->ri_RelationDesc = aorel;
resultRelInfo->ri_TrigDesc = NULL; /* we don't fire triggers */
ExecOpenIndices(resultRelInfo);
estate->es_result_relations = resultRelInfo;
estate->es_num_result_relations = 1;
estate->es_result_relation_info = resultRelInfo;
while (aocs_getnext(scanDesc, ForwardScanDirection, slot))
{
CHECK_FOR_INTERRUPTS();
aoTupleId = (AOTupleId *) slot_get_ctid(slot);
if (AppendOnlyVisimap_IsVisible(&scanDesc->visibilityMap, aoTupleId))
{
AOCSMoveTuple(slot,
insertDesc,
resultRelInfo,
estate);
movedTupleCount++;
}
else
{
/* Tuple is invisible and needs to be dropped */
AppendOnlyThrowAwayTuple(aorel,
slot,
mt_bind);
}
/*
* Check for vacuum delay point after approximatly a var block
*/
tupleCount++;
if (VacuumCostActive && tupleCount % tuplePerPage == 0)
{
vacuum_delay_point();
}
}
SetAOCSFileSegInfoState(aorel, compact_segno,
AOSEG_STATE_AWAITING_DROP);
AppendOnlyVisimap_DeleteSegmentFile(&visiMap,
compact_segno);
/* Delete all mini pages of the segment files if block directory exists */
if (OidIsValid(aorel->rd_appendonly->blkdirrelid))
{
AppendOnlyBlockDirectory_DeleteSegmentFile(aorel,
snapshot,
compact_segno,
0);
}
elogif(Debug_appendonly_print_compaction, LOG,
"Finished compaction: "
"AO segfile %d, relation %s, moved tuple count " INT64_FORMAT,
compact_segno, relname, movedTupleCount);
AppendOnlyVisimap_Finish(&visiMap, NoLock);
ExecCloseIndices(resultRelInfo);
FreeExecutorState(estate);
ExecDropSingleTupleTableSlot(slot);
destroy_memtuple_binding(mt_bind);
aocs_endscan(scanDesc);
pfree(proj);
return true;
}
/*
* Process the pending-TID list between pages of the main scan
*/
static void
spgprocesspending(spgBulkDeleteState *bds)
{
Relation index = bds->info->index;
spgVacPendingItem *pitem;
spgVacPendingItem *nitem;
BlockNumber blkno;
Buffer buffer;
Page page;
for (pitem = bds->pendingList; pitem != NULL; pitem = pitem->next)
{
if (pitem->done)
continue; /* ignore already-done items */
/* call vacuum_delay_point while not holding any buffer lock */
vacuum_delay_point();
/* examine the referenced page */
blkno = ItemPointerGetBlockNumber(&pitem->tid);
buffer = ReadBufferExtended(index, MAIN_FORKNUM, blkno,
RBM_NORMAL, bds->info->strategy);
LockBuffer(buffer, BUFFER_LOCK_EXCLUSIVE);
page = (Page) BufferGetPage(buffer);
if (PageIsNew(page) || SpGistPageIsDeleted(page))
{
/* Probably shouldn't happen, but ignore it */
}
else if (SpGistPageIsLeaf(page))
{
if (SpGistBlockIsRoot(blkno))
{
/* this should definitely not happen */
elog(ERROR, "redirection leads to root page of index \"%s\"",
RelationGetRelationName(index));
}
/* deal with any deletable tuples */
vacuumLeafPage(bds, index, buffer, true);
/* might as well do this while we are here */
vacuumRedirectAndPlaceholder(index, buffer);
SpGistSetLastUsedPage(index, buffer);
/*
* We can mark as done not only this item, but any later ones
* pointing at the same page, since we vacuumed the whole page.
*/
pitem->done = true;
for (nitem = pitem->next; nitem != NULL; nitem = nitem->next)
{
if (ItemPointerGetBlockNumber(&nitem->tid) == blkno)
nitem->done = true;
}
}
else
{
/*
* On an inner page, visit the referenced inner tuple and add all
* its downlinks to the pending list. We might have pending items
* for more than one inner tuple on the same page (in fact this is
* pretty likely given the way space allocation works), so get
* them all while we are here.
*/
for (nitem = pitem; nitem != NULL; nitem = nitem->next)
{
if (nitem->done)
continue;
if (ItemPointerGetBlockNumber(&nitem->tid) == blkno)
{
OffsetNumber offset;
SpGistInnerTuple innerTuple;
offset = ItemPointerGetOffsetNumber(&nitem->tid);
innerTuple = (SpGistInnerTuple) PageGetItem(page,
PageGetItemId(page, offset));
if (innerTuple->tupstate == SPGIST_LIVE)
{
SpGistNodeTuple node;
int i;
SGITITERATE(innerTuple, i, node)
{
if (ItemPointerIsValid(&node->t_tid))
spgAddPendingTID(bds, &node->t_tid);
}
}
else if (innerTuple->tupstate == SPGIST_REDIRECT)
{
/* transfer attention to redirect point */
spgAddPendingTID(bds,
&((SpGistDeadTuple) innerTuple)->pointer);
}
else
elog(ERROR, "unexpected SPGiST tuple state: %d",
innerTuple->tupstate);
nitem->done = true;
}
}
/*
* Move tuples from pending pages into regular GIN structure.
*
* On first glance it looks completely not crash-safe. But if we crash
* after posting entries to the main index and before removing them from the
* pending list, it's okay because when we redo the posting later on, nothing
* bad will happen.
*
* fill_fsm indicates that ginInsertCleanup should add deleted pages
* to FSM otherwise caller is responsible to put deleted pages into
* FSM.
*
* If stats isn't null, we count deleted pending pages into the counts.
*/
void
ginInsertCleanup(GinState *ginstate, bool full_clean,
bool fill_fsm, IndexBulkDeleteResult *stats)
{
Relation index = ginstate->index;
Buffer metabuffer,
buffer;
Page metapage,
page;
GinMetaPageData *metadata;
MemoryContext opCtx,
oldCtx;
BuildAccumulator accum;
KeyArray datums;
BlockNumber blkno,
blknoFinish;
bool cleanupFinish = false;
bool fsm_vac = false;
Size workMemory;
bool inVacuum = (stats == NULL);
/*
* We would like to prevent concurrent cleanup process. For that we will
* lock metapage in exclusive mode using LockPage() call. Nobody other
* will use that lock for metapage, so we keep possibility of concurrent
* insertion into pending list
*/
if (inVacuum)
{
/*
* We are called from [auto]vacuum/analyze or gin_clean_pending_list()
* and we would like to wait concurrent cleanup to finish.
*/
LockPage(index, GIN_METAPAGE_BLKNO, ExclusiveLock);
workMemory =
(IsAutoVacuumWorkerProcess() && autovacuum_work_mem != -1) ?
autovacuum_work_mem : maintenance_work_mem;
}
else
{
/*
* We are called from regular insert and if we see concurrent cleanup
* just exit in hope that concurrent process will clean up pending
* list.
*/
if (!ConditionalLockPage(index, GIN_METAPAGE_BLKNO, ExclusiveLock))
return;
workMemory = work_mem;
}
metabuffer = ReadBuffer(index, GIN_METAPAGE_BLKNO);
LockBuffer(metabuffer, GIN_SHARE);
metapage = BufferGetPage(metabuffer);
metadata = GinPageGetMeta(metapage);
if (metadata->head == InvalidBlockNumber)
{
/* Nothing to do */
UnlockReleaseBuffer(metabuffer);
UnlockPage(index, GIN_METAPAGE_BLKNO, ExclusiveLock);
return;
}
/*
* Remember a tail page to prevent infinite cleanup if other backends add
* new tuples faster than we can cleanup.
*/
blknoFinish = metadata->tail;
/*
* Read and lock head of pending list
*/
blkno = metadata->head;
buffer = ReadBuffer(index, blkno);
LockBuffer(buffer, GIN_SHARE);
page = BufferGetPage(buffer);
LockBuffer(metabuffer, GIN_UNLOCK);
/*
* Initialize. All temporary space will be in opCtx
*/
opCtx = AllocSetContextCreate(CurrentMemoryContext,
"GIN insert cleanup temporary context",
ALLOCSET_DEFAULT_MINSIZE,
ALLOCSET_DEFAULT_INITSIZE,
ALLOCSET_DEFAULT_MAXSIZE);
oldCtx = MemoryContextSwitchTo(opCtx);
initKeyArray(&datums, 128);
ginInitBA(&accum);
accum.ginstate = ginstate;
/*
* At the top of this loop, we have pin and lock on the current page of
* the pending list. However, we'll release that before exiting the loop.
* Note we also have pin but not lock on the metapage.
*/
for (;;)
{
Assert(!GinPageIsDeleted(page));
/*
* Are we walk through the page which as we remember was a tail when
* we start our cleanup? But if caller asks us to clean up whole
* pending list then ignore old tail, we will work until list becomes
* empty.
*/
if (blkno == blknoFinish && full_clean == false)
cleanupFinish = true;
/*
* read page's datums into accum
*/
processPendingPage(&accum, &datums, page, FirstOffsetNumber);
vacuum_delay_point();
/*
* Is it time to flush memory to disk? Flush if we are at the end of
* the pending list, or if we have a full row and memory is getting
* full.
*/
if (GinPageGetOpaque(page)->rightlink == InvalidBlockNumber ||
(GinPageHasFullRow(page) &&
(accum.allocatedMemory >= workMemory * 1024L)))
{
ItemPointerData *list;
uint32 nlist;
Datum key;
GinNullCategory category;
OffsetNumber maxoff,
attnum;
/*
* Unlock current page to increase performance. Changes of page
* will be checked later by comparing maxoff after completion of
* memory flush.
*/
maxoff = PageGetMaxOffsetNumber(page);
LockBuffer(buffer, GIN_UNLOCK);
/*
* Moving collected data into regular structure can take
* significant amount of time - so, run it without locking pending
* list.
*/
ginBeginBAScan(&accum);
while ((list = ginGetBAEntry(&accum,
&attnum, &key, &category, &nlist)) != NULL)
{
ginEntryInsert(ginstate, attnum, key, category,
list, nlist, NULL);
vacuum_delay_point();
}
/*
* Lock the whole list to remove pages
*/
LockBuffer(metabuffer, GIN_EXCLUSIVE);
LockBuffer(buffer, GIN_SHARE);
Assert(!GinPageIsDeleted(page));
/*
* While we left the page unlocked, more stuff might have gotten
* added to it. If so, process those entries immediately. There
* shouldn't be very many, so we don't worry about the fact that
* we're doing this with exclusive lock. Insertion algorithm
* guarantees that inserted row(s) will not continue on next page.
* NOTE: intentionally no vacuum_delay_point in this loop.
*/
if (PageGetMaxOffsetNumber(page) != maxoff)
{
ginInitBA(&accum);
processPendingPage(&accum, &datums, page, maxoff + 1);
ginBeginBAScan(&accum);
while ((list = ginGetBAEntry(&accum,
&attnum, &key, &category, &nlist)) != NULL)
ginEntryInsert(ginstate, attnum, key, category,
list, nlist, NULL);
}
/*
* Remember next page - it will become the new list head
*/
blkno = GinPageGetOpaque(page)->rightlink;
UnlockReleaseBuffer(buffer); /* shiftList will do exclusive
* locking */
/*
* remove read pages from pending list, at this point all content
* of read pages is in regular structure
*/
shiftList(index, metabuffer, blkno, fill_fsm, stats);
/* At this point, some pending pages have been freed up */
fsm_vac = true;
Assert(blkno == metadata->head);
LockBuffer(metabuffer, GIN_UNLOCK);
/*
* if we removed the whole pending list or we cleanup tail (which
* we remembered on start our cleanup process) then just exit
*/
if (blkno == InvalidBlockNumber || cleanupFinish)
break;
/*
* release memory used so far and reinit state
*/
MemoryContextReset(opCtx);
initKeyArray(&datums, datums.maxvalues);
ginInitBA(&accum);
}
else
{
blkno = GinPageGetOpaque(page)->rightlink;
UnlockReleaseBuffer(buffer);
}
/*
* Read next page in pending list
*/
vacuum_delay_point();
buffer = ReadBuffer(index, blkno);
LockBuffer(buffer, GIN_SHARE);
page = BufferGetPage(buffer);
}
UnlockPage(index, GIN_METAPAGE_BLKNO, ExclusiveLock);
ReleaseBuffer(metabuffer);
/*
* As pending list pages can have a high churn rate, it is desirable to
* recycle them immediately to the FreeSpace Map when ordinary backends
* clean the list.
*/
if (fsm_vac && fill_fsm)
IndexFreeSpaceMapVacuum(index);
/* Clean up temporary space */
MemoryContextSwitchTo(oldCtx);
MemoryContextDelete(opCtx);
}
/*
* btvacuumpage --- VACUUM one page
*
* This processes a single page for btvacuumscan(). In some cases we
* must go back and re-examine previously-scanned pages; this routine
* recurses when necessary to handle that case.
*
* blkno is the page to process. orig_blkno is the highest block number
* reached by the outer btvacuumscan loop (the same as blkno, unless we
* are recursing to re-examine a previous page).
*/
static void
btvacuumpage(BTVacState *vstate, BlockNumber blkno, BlockNumber orig_blkno)
{
IndexVacuumInfo *info = vstate->info;
IndexBulkDeleteResult *stats = vstate->stats;
IndexBulkDeleteCallback callback = vstate->callback;
void *callback_state = vstate->callback_state;
Relation rel = info->index;
bool delete_now;
BlockNumber recurse_to;
Buffer buf;
Page page;
BTPageOpaque opaque = NULL;
restart:
delete_now = false;
recurse_to = P_NONE;
/* call vacuum_delay_point while not holding any buffer lock */
vacuum_delay_point();
/*
* We can't use _bt_getbuf() here because it always applies
* _bt_checkpage(), which will barf on an all-zero page. We want to
* recycle all-zero pages, not fail. Also, we want to use a nondefault
* buffer access strategy.
*/
buf = ReadBufferExtended(rel, MAIN_FORKNUM, blkno, RBM_NORMAL,
info->strategy);
LockBuffer(buf, BT_READ);
page = BufferGetPage(buf);
if (!PageIsNew(page))
{
_bt_checkpage(rel, buf);
opaque = (BTPageOpaque) PageGetSpecialPointer(page);
}
/*
* If we are recursing, the only case we want to do anything with is a
* live leaf page having the current vacuum cycle ID. Any other state
* implies we already saw the page (eg, deleted it as being empty).
*/
if (blkno != orig_blkno)
{
if (_bt_page_recyclable(page) ||
P_IGNORE(opaque) ||
!P_ISLEAF(opaque) ||
opaque->btpo_cycleid != vstate->cycleid)
{
_bt_relbuf(rel, buf);
return;
}
}
/* Page is valid, see what to do with it */
if (_bt_page_recyclable(page))
{
/* Okay to recycle this page */
RecordFreeIndexPage(rel, blkno);
vstate->totFreePages++;
stats->pages_deleted++;
}
else if (P_ISDELETED(opaque))
{
/* Already deleted, but can't recycle yet */
stats->pages_deleted++;
}
else if (P_ISHALFDEAD(opaque))
{
/* Half-dead, try to delete */
delete_now = true;
}
else if (P_ISLEAF(opaque))
{
OffsetNumber deletable[MaxOffsetNumber];
int ndeletable;
OffsetNumber offnum,
minoff,
maxoff;
/*
* Trade in the initial read lock for a super-exclusive write lock on
* this page. We must get such a lock on every leaf page over the
* course of the vacuum scan, whether or not it actually contains any
* deletable tuples --- see nbtree/README.
*/
LockBuffer(buf, BUFFER_LOCK_UNLOCK);
LockBufferForCleanup(buf);
/*
* Remember highest leaf page number we've taken cleanup lock on; see
* notes in btvacuumscan
*/
if (blkno > vstate->lastBlockLocked)
vstate->lastBlockLocked = blkno;
/*
* Check whether we need to recurse back to earlier pages. What we
* are concerned about is a page split that happened since we started
* the vacuum scan. If the split moved some tuples to a lower page
* then we might have missed 'em. If so, set up for tail recursion.
* (Must do this before possibly clearing btpo_cycleid below!)
*/
if (vstate->cycleid != 0 &&
opaque->btpo_cycleid == vstate->cycleid &&
!(opaque->btpo_flags & BTP_SPLIT_END) &&
!P_RIGHTMOST(opaque) &&
opaque->btpo_next < orig_blkno)
recurse_to = opaque->btpo_next;
/*
* Scan over all items to see which ones need deleted according to the
* callback function.
*/
ndeletable = 0;
minoff = P_FIRSTDATAKEY(opaque);
maxoff = PageGetMaxOffsetNumber(page);
if (callback)
{
for (offnum = minoff;
offnum <= maxoff;
offnum = OffsetNumberNext(offnum))
{
IndexTuple itup;
ItemPointer htup;
itup = (IndexTuple) PageGetItem(page,
PageGetItemId(page, offnum));
htup = &(itup->t_tid);
/*
* During Hot Standby we currently assume that
* XLOG_BTREE_VACUUM records do not produce conflicts. That is
* only true as long as the callback function depends only
* upon whether the index tuple refers to heap tuples removed
* in the initial heap scan. When vacuum starts it derives a
* value of OldestXmin. Backends taking later snapshots could
* have a RecentGlobalXmin with a later xid than the vacuum's
* OldestXmin, so it is possible that row versions deleted
* after OldestXmin could be marked as killed by other
* backends. The callback function *could* look at the index
* tuple state in isolation and decide to delete the index
* tuple, though currently it does not. If it ever did, we
* would need to reconsider whether XLOG_BTREE_VACUUM records
* should cause conflicts. If they did cause conflicts they
* would be fairly harsh conflicts, since we haven't yet
* worked out a way to pass a useful value for
* latestRemovedXid on the XLOG_BTREE_VACUUM records. This
* applies to *any* type of index that marks index tuples as
* killed.
*/
if (callback(htup, callback_state))
deletable[ndeletable++] = offnum;
}
}
/*
* Apply any needed deletes. We issue just one _bt_delitems_vacuum()
* call per page, so as to minimize WAL traffic.
*/
if (ndeletable > 0)
{
/*
* Notice that the issued XLOG_BTREE_VACUUM WAL record includes
* all information to the replay code to allow it to get a cleanup
* lock on all pages between the previous lastBlockVacuumed and
* this page. This ensures that WAL replay locks all leaf pages at
* some point, which is important should non-MVCC scans be
* requested. This is currently unused on standby, but we record
* it anyway, so that the WAL contains the required information.
*
* Since we can visit leaf pages out-of-order when recursing,
* replay might end up locking such pages an extra time, but it
* doesn't seem worth the amount of bookkeeping it'd take to avoid
* that.
*/
_bt_delitems_vacuum(rel, buf, deletable, ndeletable,
vstate->lastBlockVacuumed);
/*
* Remember highest leaf page number we've issued a
* XLOG_BTREE_VACUUM WAL record for.
*/
if (blkno > vstate->lastBlockVacuumed)
vstate->lastBlockVacuumed = blkno;
stats->tuples_removed += ndeletable;
/* must recompute maxoff */
maxoff = PageGetMaxOffsetNumber(page);
}
else
{
/*
* If the page has been split during this vacuum cycle, it seems
* worth expending a write to clear btpo_cycleid even if we don't
* have any deletions to do. (If we do, _bt_delitems_vacuum takes
* care of this.) This ensures we won't process the page again.
*
* We treat this like a hint-bit update because there's no need to
* WAL-log it.
*/
if (vstate->cycleid != 0 &&
opaque->btpo_cycleid == vstate->cycleid)
{
opaque->btpo_cycleid = 0;
MarkBufferDirtyHint(buf, true);
}
}
/*
* If it's now empty, try to delete; else count the live tuples. We
* don't delete when recursing, though, to avoid putting entries into
* freePages out-of-order (doesn't seem worth any extra code to handle
* the case).
*/
if (minoff > maxoff)
delete_now = (blkno == orig_blkno);
else
stats->num_index_tuples += maxoff - minoff + 1;
}
if (delete_now)
{
MemoryContext oldcontext;
int ndel;
/* Run pagedel in a temp context to avoid memory leakage */
MemoryContextReset(vstate->pagedelcontext);
oldcontext = MemoryContextSwitchTo(vstate->pagedelcontext);
ndel = _bt_pagedel(rel, buf);
/* count only this page, else may double-count parent */
if (ndel)
stats->pages_deleted++;
MemoryContextSwitchTo(oldcontext);
/* pagedel released buffer, so we shouldn't */
}
else
_bt_relbuf(rel, buf);
/*
* This is really tail recursion, but if the compiler is too stupid to
* optimize it as such, we'd eat an uncomfortably large amount of stack
* space per recursion level (due to the deletable[] array). A failure is
* improbable since the number of levels isn't likely to be large ... but
* just in case, let's hand-optimize into a loop.
*/
if (recurse_to != P_NONE)
{
blkno = recurse_to;
goto restart;
}
}
/*
* compute_tsvector_stats() -- compute statistics for a tsvector column
*
* This functions computes statistics that are useful for determining @@
* operations' selectivity, along with the fraction of non-null rows and
* average width.
*
* Instead of finding the most common values, as we do for most datatypes,
* we're looking for the most common lexemes. This is more useful, because
* there most probably won't be any two rows with the same tsvector and thus
* the notion of a MCV is a bit bogus with this datatype. With a list of the
* most common lexemes we can do a better job at figuring out @@ selectivity.
*
* For the same reasons we assume that tsvector columns are unique when
* determining the number of distinct values.
*
* The algorithm used is Lossy Counting, as proposed in the paper "Approximate
* frequency counts over data streams" by G. S. Manku and R. Motwani, in
* Proceedings of the 28th International Conference on Very Large Data Bases,
* Hong Kong, China, August 2002, section 4.2. The paper is available at
* http://www.vldb.org/conf/2002/S10P03.pdf
*
* The Lossy Counting (aka LC) algorithm goes like this:
* Let D be a set of triples (e, f, d), where e is an element value, f is
* that element's frequency (occurrence count) and d is the maximum error in
* f. We start with D empty and process the elements in batches of size
* w. (The batch size is also known as "bucket size".) Let the current batch
* number be b_current, starting with 1. For each element e we either
* increment its f count, if it's already in D, or insert a new triple into D
* with values (e, 1, b_current - 1). After processing each batch we prune D,
* by removing from it all elements with f + d <= b_current. Finally, we
* gather elements with largest f. The LC paper proves error bounds on f
* dependent on the batch size w, and shows that the required table size
* is no more than a few times w.
*
* We use a hashtable for the D structure and a bucket width of
* statistics_target * 10, where 10 is an arbitrarily chosen constant,
* meant to approximate the number of lexemes in a single tsvector.
*/
static void
compute_tsvector_stats(VacAttrStats *stats,
AnalyzeAttrFetchFunc fetchfunc,
int samplerows,
double totalrows)
{
int num_mcelem;
int null_cnt = 0;
double total_width = 0;
/* This is D from the LC algorithm. */
HTAB *lexemes_tab;
HASHCTL hash_ctl;
HASH_SEQ_STATUS scan_status;
/* This is the current bucket number from the LC algorithm */
int b_current;
/* This is 'w' from the LC algorithm */
int bucket_width;
int vector_no,
lexeme_no;
LexemeHashKey hash_key;
TrackItem *item;
/* We want statistics_target * 10 lexemes in the MCELEM array */
num_mcelem = stats->attr->attstattarget * 10;
/*
* We set bucket width equal to the target number of result lexemes. This
* is probably about right but perhaps might need to be scaled up or down
* a bit?
*/
bucket_width = num_mcelem;
/*
* Create the hashtable. It will be in local memory, so we don't need to
* worry about initial size too much. Also we don't need to pay any
* attention to locking and memory management.
*/
MemSet(&hash_ctl, 0, sizeof(hash_ctl));
hash_ctl.keysize = sizeof(LexemeHashKey);
hash_ctl.entrysize = sizeof(TrackItem);
hash_ctl.hash = lexeme_hash;
hash_ctl.match = lexeme_match;
hash_ctl.hcxt = CurrentMemoryContext;
lexemes_tab = hash_create("Analyzed lexemes table",
bucket_width * 4,
&hash_ctl,
HASH_ELEM | HASH_FUNCTION | HASH_COMPARE | HASH_CONTEXT);
/* Initialize counters. */
b_current = 1;
lexeme_no = 1;
/* Loop over the tsvectors. */
for (vector_no = 0; vector_no < samplerows; vector_no++)
{
Datum value;
bool isnull;
TSVector vector;
WordEntry *curentryptr;
char *lexemesptr;
int j;
vacuum_delay_point();
value = fetchfunc(stats, vector_no, &isnull);
/*
* Check for null/nonnull.
*/
if (isnull)
{
null_cnt++;
continue;
}
/*
* Add up widths for average-width calculation. Since it's a
* tsvector, we know it's varlena. As in the regular
* compute_minimal_stats function, we use the toasted width for this
* calculation.
*/
total_width += VARSIZE_ANY(DatumGetPointer(value));
/*
* Now detoast the tsvector if needed.
*/
vector = DatumGetTSVector(value);
/*
* We loop through the lexemes in the tsvector and add them to our
* tracking hashtable. Note: the hashtable entries will point into
* the (detoasted) tsvector value, therefore we cannot free that
* storage until we're done.
*/
lexemesptr = STRPTR(vector);
curentryptr = ARRPTR(vector);
for (j = 0; j < vector->size; j++)
{
bool found;
/* Construct a hash key */
hash_key.lexeme = lexemesptr + curentryptr->pos;
hash_key.length = curentryptr->len;
/* Lookup current lexeme in hashtable, adding it if new */
item = (TrackItem *) hash_search(lexemes_tab,
(const void *) &hash_key,
HASH_ENTER, &found);
if (found)
{
/* The lexeme is already on the tracking list */
item->frequency++;
}
else
{
/* Initialize new tracking list element */
item->frequency = 1;
item->delta = b_current - 1;
}
/* We prune the D structure after processing each bucket */
if (lexeme_no % bucket_width == 0)
{
prune_lexemes_hashtable(lexemes_tab, b_current);
b_current++;
}
/* Advance to the next WordEntry in the tsvector */
lexeme_no++;
curentryptr++;
}
}
/* We can only compute real stats if we found some non-null values. */
if (null_cnt < samplerows)
{
int nonnull_cnt = samplerows - null_cnt;
int i;
TrackItem **sort_table;
int track_len;
int minfreq,
maxfreq;
stats->stats_valid = true;
/* Do the simple null-frac and average width stats */
stats->stanullfrac = (double) null_cnt / (double) samplerows;
stats->stawidth = total_width / (double) nonnull_cnt;
/* Assume it's a unique column (see notes above) */
stats->stadistinct = -1.0;
/*
* Determine the top-N lexemes by simply copying pointers from the
* hashtable into an array and applying qsort()
*/
track_len = hash_get_num_entries(lexemes_tab);
sort_table = (TrackItem **) palloc(sizeof(TrackItem *) * track_len);
hash_seq_init(&scan_status, lexemes_tab);
i = 0;
while ((item = (TrackItem *) hash_seq_search(&scan_status)) != NULL)
{
sort_table[i++] = item;
}
Assert(i == track_len);
qsort(sort_table, track_len, sizeof(TrackItem *),
trackitem_compare_frequencies_desc);
/* Suppress any single-occurrence items */
while (track_len > 0)
{
if (sort_table[track_len - 1]->frequency > 1)
break;
track_len--;
}
/* Determine the number of most common lexemes to be stored */
if (num_mcelem > track_len)
num_mcelem = track_len;
/* Generate MCELEM slot entry */
if (num_mcelem > 0)
{
MemoryContext old_context;
Datum *mcelem_values;
float4 *mcelem_freqs;
/* Grab the minimal and maximal frequencies that will get stored */
minfreq = sort_table[num_mcelem - 1]->frequency;
maxfreq = sort_table[0]->frequency;
/*
* We want to store statistics sorted on the lexeme value using
* first length, then byte-for-byte comparison. The reason for
* doing length comparison first is that we don't care about the
* ordering so long as it's consistent, and comparing lengths
* first gives us a chance to avoid a strncmp() call.
*
* This is different from what we do with scalar statistics --
* they get sorted on frequencies. The rationale is that we
* usually search through most common elements looking for a
* specific value, so we can grab its frequency. When values are
* presorted we can employ binary search for that. See
* ts_selfuncs.c for a real usage scenario.
*/
qsort(sort_table, num_mcelem, sizeof(TrackItem *),
trackitem_compare_lexemes);
/* Must copy the target values into anl_context */
old_context = MemoryContextSwitchTo(stats->anl_context);
/*
* We sorted statistics on the lexeme value, but we want to be
* able to find out the minimal and maximal frequency without
* going through all the values. We keep those two extra
* frequencies in two extra cells in mcelem_freqs.
*/
mcelem_values = (Datum *) palloc(num_mcelem * sizeof(Datum));
mcelem_freqs = (float4 *) palloc((num_mcelem + 2) * sizeof(float4));
for (i = 0; i < num_mcelem; i++)
{
TrackItem *item = sort_table[i];
mcelem_values[i] =
PointerGetDatum(cstring_to_text_with_len(item->key.lexeme,
item->key.length));
mcelem_freqs[i] = (double) item->frequency / (double) nonnull_cnt;
}
mcelem_freqs[i++] = (double) minfreq / (double) nonnull_cnt;
mcelem_freqs[i] = (double) maxfreq / (double) nonnull_cnt;
MemoryContextSwitchTo(old_context);
stats->stakind[0] = STATISTIC_KIND_MCELEM;
stats->staop[0] = TextEqualOperator;
stats->stanumbers[0] = mcelem_freqs;
/* See above comment about two extra frequency fields */
stats->numnumbers[0] = num_mcelem + 2;
stats->stavalues[0] = mcelem_values;
stats->numvalues[0] = num_mcelem;
/* We are storing text values */
stats->statypid[0] = TEXTOID;
stats->statyplen[0] = -1; /* typlen, -1 for varlena */
stats->statypbyval[0] = false;
stats->statypalign[0] = 'i';
}
}
else
{
/* We found only nulls; assume the column is entirely null */
stats->stats_valid = true;
stats->stanullfrac = 1.0;
stats->stawidth = 0; /* "unknown" */
stats->stadistinct = 0.0; /* "unknown" */
}
/*
* We don't need to bother cleaning up any of our temporary palloc's. The
* hashtable should also go away, as it used a child memory context.
*/
}
Datum
ginvacuumcleanup(PG_FUNCTION_ARGS)
{
IndexVacuumInfo *info = (IndexVacuumInfo *) PG_GETARG_POINTER(0);
IndexBulkDeleteResult *stats = (IndexBulkDeleteResult *) PG_GETARG_POINTER(1);
Relation index = info->index;
bool needLock;
BlockNumber npages,
blkno;
BlockNumber totFreePages;
GinState ginstate;
GinStatsData idxStat;
/*
* In an autovacuum analyze, we want to clean up pending insertions.
* Otherwise, an ANALYZE-only call is a no-op.
*/
if (info->analyze_only)
{
if (IsAutoVacuumWorkerProcess())
{
initGinState(&ginstate, index);
ginInsertCleanup(&ginstate, true, stats);
}
PG_RETURN_POINTER(stats);
}
/*
* Set up all-zero stats and cleanup pending inserts if ginbulkdelete
* wasn't called
*/
if (stats == NULL)
{
stats = (IndexBulkDeleteResult *) palloc0(sizeof(IndexBulkDeleteResult));
initGinState(&ginstate, index);
ginInsertCleanup(&ginstate, true, stats);
}
memset(&idxStat, 0, sizeof(idxStat));
/*
* XXX we always report the heap tuple count as the number of index
* entries. This is bogus if the index is partial, but it's real hard to
* tell how many distinct heap entries are referenced by a GIN index.
*/
stats->num_index_tuples = info->num_heap_tuples;
stats->estimated_count = info->estimated_count;
/*
* Need lock unless it's local to this backend.
*/
needLock = !RELATION_IS_LOCAL(index);
if (needLock)
LockRelationForExtension(index, ExclusiveLock);
npages = RelationGetNumberOfBlocks(index);
if (needLock)
UnlockRelationForExtension(index, ExclusiveLock);
totFreePages = 0;
for (blkno = GIN_ROOT_BLKNO; blkno < npages; blkno++)
{
Buffer buffer;
Page page;
vacuum_delay_point();
buffer = ReadBufferExtended(index, MAIN_FORKNUM, blkno,
RBM_NORMAL, info->strategy);
LockBuffer(buffer, GIN_SHARE);
page = (Page) BufferGetPage(buffer);
if (GinPageIsDeleted(page))
{
Assert(blkno != GIN_ROOT_BLKNO);
RecordFreeIndexPage(index, blkno);
totFreePages++;
}
else if (GinPageIsData(page))
{
idxStat.nDataPages++;
}
else if (!GinPageIsList(page))
{
idxStat.nEntryPages++;
if (GinPageIsLeaf(page))
idxStat.nEntries += PageGetMaxOffsetNumber(page);
}
UnlockReleaseBuffer(buffer);
}
/* Update the metapage with accurate page and entry counts */
idxStat.nTotalPages = npages;
ginUpdateStats(info->index, &idxStat);
/* Finally, vacuum the FSM */
IndexFreeSpaceMapVacuum(info->index);
stats->pages_free = totFreePages;
if (needLock)
LockRelationForExtension(index, ExclusiveLock);
stats->num_pages = RelationGetNumberOfBlocks(index);
if (needLock)
UnlockRelationForExtension(index, ExclusiveLock);
PG_RETURN_POINTER(stats);
}
/*
* Bulk deletion of all index entries pointing to a set of heap tuples.
* The set of target tuples is specified via a callback routine that tells
* whether any given heap tuple (identified by ItemPointer) is being deleted.
*
* Result: a palloc'd struct containing statistical info for VACUUM displays.
*/
Datum
rtbulkdelete(PG_FUNCTION_ARGS)
{
Relation rel = (Relation) PG_GETARG_POINTER(0);
IndexBulkDeleteCallback callback = (IndexBulkDeleteCallback) PG_GETARG_POINTER(1);
void *callback_state = (void *) PG_GETARG_POINTER(2);
IndexBulkDeleteResult *result;
BlockNumber num_pages;
double tuples_removed;
double num_index_tuples;
IndexScanDesc iscan;
tuples_removed = 0;
num_index_tuples = 0;
/*
* Since rtree is not marked "amconcurrent" in pg_am, caller should have
* acquired exclusive lock on index relation. We need no locking here.
*/
/*
* XXX generic implementation --- should be improved!
*/
/* walk through the entire index */
iscan = index_beginscan(NULL, rel, SnapshotAny, 0, NULL);
/* including killed tuples */
iscan->ignore_killed_tuples = false;
while (index_getnext_indexitem(iscan, ForwardScanDirection))
{
vacuum_delay_point();
if (callback(&iscan->xs_ctup.t_self, callback_state))
{
ItemPointerData indextup = iscan->currentItemData;
BlockNumber blkno;
OffsetNumber offnum;
Buffer buf;
Page page;
blkno = ItemPointerGetBlockNumber(&indextup);
offnum = ItemPointerGetOffsetNumber(&indextup);
/* adjust any scans that will be affected by this deletion */
/* (namely, my own scan) */
rtadjscans(rel, RTOP_DEL, blkno, offnum);
/* delete the index tuple */
buf = ReadBuffer(rel, blkno);
page = BufferGetPage(buf);
PageIndexTupleDelete(page, offnum);
WriteBuffer(buf);
tuples_removed += 1;
}
else
num_index_tuples += 1;
}
index_endscan(iscan);
/* return statistics */
num_pages = RelationGetNumberOfBlocks(rel);
result = (IndexBulkDeleteResult *) palloc0(sizeof(IndexBulkDeleteResult));
result->num_pages = num_pages;
result->num_index_tuples = num_index_tuples;
result->tuples_removed = tuples_removed;
PG_RETURN_POINTER(result);
}
/*
* lazy_scan_heap() -- scan an open heap relation
*
* This routine sets commit status bits, builds lists of dead tuples
* and pages with free space, and calculates statistics on the number
* of live tuples in the heap. When done, or when we run low on space
* for dead-tuple TIDs, invoke vacuuming of indexes and heap.
*
* If there are no indexes then we just vacuum each dirty page as we
* process it, since there's no point in gathering many tuples.
*/
static void
lazy_scan_heap(Relation onerel, LVRelStats *vacrelstats,
Relation *Irel, int nindexes, List *updated_stats)
{
MIRROREDLOCK_BUFMGR_DECLARE;
BlockNumber nblocks,
blkno;
HeapTupleData tuple;
char *relname;
BlockNumber empty_pages,
vacuumed_pages;
double num_tuples,
tups_vacuumed,
nkeep,
nunused;
IndexBulkDeleteResult **indstats;
int i;
int reindex_count = 1;
PGRUsage ru0;
/* Fetch gp_persistent_relation_node information that will be added to XLOG record. */
RelationFetchGpRelationNodeForXLog(onerel);
pg_rusage_init(&ru0);
relname = RelationGetRelationName(onerel);
ereport(elevel,
(errmsg("vacuuming \"%s.%s\"",
get_namespace_name(RelationGetNamespace(onerel)),
relname)));
empty_pages = vacuumed_pages = 0;
num_tuples = tups_vacuumed = nkeep = nunused = 0;
indstats = (IndexBulkDeleteResult **)
palloc0(nindexes * sizeof(IndexBulkDeleteResult *));
nblocks = RelationGetNumberOfBlocks(onerel);
vacrelstats->rel_pages = nblocks;
vacrelstats->nonempty_pages = 0;
lazy_space_alloc(vacrelstats, nblocks);
for (blkno = 0; blkno < nblocks; blkno++)
{
Buffer buf;
Page page;
OffsetNumber offnum,
maxoff;
bool tupgone,
hastup;
int prev_dead_count;
OffsetNumber frozen[MaxOffsetNumber];
int nfrozen;
vacuum_delay_point();
/*
* If we are close to overrunning the available space for dead-tuple
* TIDs, pause and do a cycle of vacuuming before we tackle this page.
*/
if ((vacrelstats->max_dead_tuples - vacrelstats->num_dead_tuples) < MaxHeapTuplesPerPage &&
vacrelstats->num_dead_tuples > 0)
{
/* Remove index entries */
for (i = 0; i < nindexes; i++)
lazy_vacuum_index(Irel[i], &indstats[i], vacrelstats);
reindex_count++;
/* Remove tuples from heap */
lazy_vacuum_heap(onerel, vacrelstats);
/* Forget the now-vacuumed tuples, and press on */
vacrelstats->num_dead_tuples = 0;
vacrelstats->num_index_scans++;
}
/* -------- MirroredLock ---------- */
MIRROREDLOCK_BUFMGR_LOCK;
buf = ReadBufferWithStrategy(onerel, blkno, vac_strategy);
/* We need buffer cleanup lock so that we can prune HOT chains. */
LockBufferForCleanup(buf);
page = BufferGetPage(buf);
if (PageIsNew(page))
{
/*
* An all-zeroes page could be left over if a backend extends the
* relation but crashes before initializing the page. Reclaim such
* pages for use.
*
* We have to be careful here because we could be looking at a
* page that someone has just added to the relation and not yet
* been able to initialize (see RelationGetBufferForTuple). To
* protect against that, release the buffer lock, grab the
* relation extension lock momentarily, and re-lock the buffer. If
* the page is still uninitialized by then, it must be left over
* from a crashed backend, and we can initialize it.
*
* We don't really need the relation lock when this is a new or
* temp relation, but it's probably not worth the code space to
* check that, since this surely isn't a critical path.
*
* Note: the comparable code in vacuum.c need not worry because
* it's got exclusive lock on the whole relation.
*/
LockBuffer(buf, BUFFER_LOCK_UNLOCK);
MIRROREDLOCK_BUFMGR_UNLOCK;
/* -------- MirroredLock ---------- */
LockRelationForExtension(onerel, ExclusiveLock);
UnlockRelationForExtension(onerel, ExclusiveLock);
/* -------- MirroredLock ---------- */
MIRROREDLOCK_BUFMGR_LOCK;
LockBufferForCleanup(buf);
if (PageIsNew(page))
{
ereport(WARNING,
(errmsg("relation \"%s\" page %u is uninitialized --- fixing",
relname, blkno)));
PageInit(page, BufferGetPageSize(buf), 0);
/* must record in xlog so that changetracking will know about this change */
log_heap_newpage(onerel, page, blkno);
empty_pages++;
lazy_record_free_space(vacrelstats, blkno,
PageGetHeapFreeSpace(page));
}
MarkBufferDirty(buf);
UnlockReleaseBuffer(buf);
MIRROREDLOCK_BUFMGR_UNLOCK;
/* -------- MirroredLock ---------- */
continue;
}
if (PageIsEmpty(page))
{
empty_pages++;
lazy_record_free_space(vacrelstats, blkno,
PageGetHeapFreeSpace(page));
UnlockReleaseBuffer(buf);
MIRROREDLOCK_BUFMGR_UNLOCK;
/* -------- MirroredLock ---------- */
continue;
}
/*
* Prune all HOT-update chains in this page.
*
* We count tuples removed by the pruning step as removed by VACUUM.
*/
tups_vacuumed += heap_page_prune(onerel, buf, OldestXmin,
false, false);
/*
* Now scan the page to collect vacuumable items and check for tuples
* requiring freezing.
*/
nfrozen = 0;
hastup = false;
prev_dead_count = vacrelstats->num_dead_tuples;
maxoff = PageGetMaxOffsetNumber(page);
for (offnum = FirstOffsetNumber;
offnum <= maxoff;
offnum = OffsetNumberNext(offnum))
{
ItemId itemid;
itemid = PageGetItemId(page, offnum);
/* Unused items require no processing, but we count 'em */
if (!ItemIdIsUsed(itemid))
{
nunused += 1;
continue;
}
/* Redirect items mustn't be touched */
if (ItemIdIsRedirected(itemid))
{
hastup = true; /* this page won't be truncatable */
continue;
}
ItemPointerSet(&(tuple.t_self), blkno, offnum);
/*
* DEAD item pointers are to be vacuumed normally; but we don't
* count them in tups_vacuumed, else we'd be double-counting (at
* least in the common case where heap_page_prune() just freed up
* a non-HOT tuple).
*/
if (ItemIdIsDead(itemid))
{
lazy_record_dead_tuple(vacrelstats, &(tuple.t_self));
continue;
}
Assert(ItemIdIsNormal(itemid));
tuple.t_data = (HeapTupleHeader) PageGetItem(page, itemid);
tuple.t_len = ItemIdGetLength(itemid);
tupgone = false;
switch (HeapTupleSatisfiesVacuum(onerel, tuple.t_data, OldestXmin, buf))
{
case HEAPTUPLE_DEAD:
/*
* Ordinarily, DEAD tuples would have been removed by
* heap_page_prune(), but it's possible that the tuple
* state changed since heap_page_prune() looked. In
* particular an INSERT_IN_PROGRESS tuple could have
* changed to DEAD if the inserter aborted. So this
* cannot be considered an error condition.
*
* If the tuple is HOT-updated then it must only be
* removed by a prune operation; so we keep it just as if
* it were RECENTLY_DEAD. Also, if it's a heap-only
* tuple, we choose to keep it, because it'll be a lot
* cheaper to get rid of it in the next pruning pass than
* to treat it like an indexed tuple.
*/
if (HeapTupleIsHotUpdated(&tuple) ||
HeapTupleIsHeapOnly(&tuple))
nkeep += 1;
else
tupgone = true; /* we can delete the tuple */
break;
case HEAPTUPLE_LIVE:
/* Tuple is good --- but let's do some validity checks */
if (onerel->rd_rel->relhasoids &&
!OidIsValid(HeapTupleGetOid(&tuple)))
elog(WARNING, "relation \"%s\" TID %u/%u: OID is invalid",
relname, blkno, offnum);
break;
case HEAPTUPLE_RECENTLY_DEAD:
/*
* If tuple is recently deleted then we must not remove it
* from relation.
*/
nkeep += 1;
break;
case HEAPTUPLE_INSERT_IN_PROGRESS:
/* This is an expected case during concurrent vacuum */
break;
case HEAPTUPLE_DELETE_IN_PROGRESS:
/* This is an expected case during concurrent vacuum */
break;
default:
elog(ERROR, "unexpected HeapTupleSatisfiesVacuum result");
break;
}
if (tupgone)
{
lazy_record_dead_tuple(vacrelstats, &(tuple.t_self));
tups_vacuumed += 1;
}
else
{
num_tuples += 1;
hastup = true;
/*
* Each non-removable tuple must be checked to see if it needs
* freezing. Note we already have exclusive buffer lock.
*/
if (heap_freeze_tuple(tuple.t_data, &FreezeLimit,
InvalidBuffer, false))
frozen[nfrozen++] = offnum;
}
} /* scan along page */
/*
* If we froze any tuples, mark the buffer dirty, and write a WAL
* record recording the changes. We must log the changes to be
* crash-safe against future truncation of CLOG.
*/
if (nfrozen > 0)
{
MarkBufferDirty(buf);
/* no XLOG for temp tables, though */
if (!onerel->rd_istemp)
{
XLogRecPtr recptr;
recptr = log_heap_freeze(onerel, buf, FreezeLimit,
frozen, nfrozen);
PageSetLSN(page, recptr);
}
}
/*
* If there are no indexes then we can vacuum the page right now
* instead of doing a second scan.
*/
if (nindexes == 0 &&
vacrelstats->num_dead_tuples > 0)
{
/* Remove tuples from heap */
lazy_vacuum_page(onerel, blkno, buf, 0, vacrelstats);
/* Forget the now-vacuumed tuples, and press on */
vacrelstats->num_dead_tuples = 0;
vacuumed_pages++;
}
/*
* If we remembered any tuples for deletion, then the page will be
* visited again by lazy_vacuum_heap, which will compute and record
* its post-compaction free space. If not, then we're done with this
* page, so remember its free space as-is. (This path will always be
* taken if there are no indexes.)
*/
if (vacrelstats->num_dead_tuples == prev_dead_count)
{
lazy_record_free_space(vacrelstats, blkno,
PageGetHeapFreeSpace(page));
}
/* Remember the location of the last page with nonremovable tuples */
if (hastup)
vacrelstats->nonempty_pages = blkno + 1;
UnlockReleaseBuffer(buf);
MIRROREDLOCK_BUFMGR_UNLOCK;
/* -------- MirroredLock ---------- */
}
/* save stats for use later */
vacrelstats->rel_tuples = num_tuples;
vacrelstats->tuples_deleted = tups_vacuumed;
/* If any tuples need to be deleted, perform final vacuum cycle */
/* XXX put a threshold on min number of tuples here? */
if (vacrelstats->num_dead_tuples > 0)
{
/* Remove index entries */
for (i = 0; i < nindexes; i++)
lazy_vacuum_index(Irel[i], &indstats[i], vacrelstats);
reindex_count++;
/* Remove tuples from heap */
lazy_vacuum_heap(onerel, vacrelstats);
vacrelstats->num_index_scans++;
}
/* Do post-vacuum cleanup and statistics update for each index */
for (i = 0; i < nindexes; i++)
lazy_cleanup_index(Irel[i], indstats[i], vacrelstats, updated_stats);
/* If no indexes, make log report that lazy_vacuum_heap would've made */
if (vacuumed_pages)
ereport(elevel,
(errmsg("\"%s\": removed %.0f row versions in %u pages",
RelationGetRelationName(onerel),
tups_vacuumed, vacuumed_pages)));
ereport(elevel,
(errmsg("\"%s\": found %.0f removable, %.0f nonremovable row versions in %u pages",
RelationGetRelationName(onerel),
tups_vacuumed, num_tuples, nblocks),
errdetail("%.0f dead row versions cannot be removed yet.\n"
"There were %.0f unused item pointers.\n"
"%u pages contain useful free space.\n"
"%u pages are entirely empty.\n"
"%s.",
nkeep,
nunused,
vacrelstats->tot_free_pages,
empty_pages,
pg_rusage_show(&ru0))));
}
Datum
ginvacuumcleanup(PG_FUNCTION_ARGS)
{
MIRROREDLOCK_BUFMGR_DECLARE;
IndexVacuumInfo *info = (IndexVacuumInfo *) PG_GETARG_POINTER(0);
IndexBulkDeleteResult *stats = (IndexBulkDeleteResult *) PG_GETARG_POINTER(1);
Relation index = info->index;
bool needLock;
BlockNumber npages,
blkno;
BlockNumber totFreePages,
nFreePages,
*freePages,
maxFreePages;
BlockNumber lastBlock = GIN_ROOT_BLKNO,
lastFilledBlock = GIN_ROOT_BLKNO;
/* Set up all-zero stats if ginbulkdelete wasn't called */
if (stats == NULL)
stats = (IndexBulkDeleteResult *) palloc0(sizeof(IndexBulkDeleteResult));
/*
* XXX we always report the heap tuple count as the number of index
* entries. This is bogus if the index is partial, but it's real hard to
* tell how many distinct heap entries are referenced by a GIN index.
*/
stats->num_index_tuples = info->num_heap_tuples;
/*
* If vacuum full, we already have exclusive lock on the index. Otherwise,
* need lock unless it's local to this backend.
*/
if (info->vacuum_full)
needLock = false;
else
needLock = !RELATION_IS_LOCAL(index);
if (needLock)
LockRelationForExtension(index, ExclusiveLock);
npages = RelationGetNumberOfBlocks(index);
if (needLock)
UnlockRelationForExtension(index, ExclusiveLock);
maxFreePages = npages;
if (maxFreePages > MaxFSMPages)
maxFreePages = MaxFSMPages;
totFreePages = nFreePages = 0;
freePages = (BlockNumber *) palloc(sizeof(BlockNumber) * maxFreePages);
for (blkno = GIN_ROOT_BLKNO + 1; blkno < npages; blkno++)
{
Buffer buffer;
Page page;
vacuum_delay_point();
// -------- MirroredLock ----------
MIRROREDLOCK_BUFMGR_LOCK;
buffer = ReadBuffer(index, blkno);
LockBuffer(buffer, GIN_SHARE);
page = (Page) BufferGetPage(buffer);
if (GinPageIsDeleted(page))
{
if (nFreePages < maxFreePages)
freePages[nFreePages++] = blkno;
totFreePages++;
}
else
lastFilledBlock = blkno;
UnlockReleaseBuffer(buffer);
MIRROREDLOCK_BUFMGR_UNLOCK;
// -------- MirroredLock ----------
}
lastBlock = npages - 1;
if (info->vacuum_full && nFreePages > 0)
{
/* try to truncate index */
int i;
for (i = 0; i < nFreePages; i++)
if (freePages[i] >= lastFilledBlock)
{
totFreePages = nFreePages = i;
break;
}
if (lastBlock > lastFilledBlock)
RelationTruncate(
index,
lastFilledBlock + 1,
/* markPersistentAsPhysicallyTruncated */ true);
stats->pages_removed = lastBlock - lastFilledBlock;
}
RecordIndexFreeSpace(&index->rd_node, totFreePages, nFreePages, freePages);
stats->pages_free = totFreePages;
if (needLock)
LockRelationForExtension(index, ExclusiveLock);
stats->num_pages = RelationGetNumberOfBlocks(index);
if (needLock)
UnlockRelationForExtension(index, ExclusiveLock);
PG_RETURN_POINTER(stats);
}
Datum
ginbulkdelete(PG_FUNCTION_ARGS)
{
MIRROREDLOCK_BUFMGR_DECLARE;
IndexVacuumInfo *info = (IndexVacuumInfo *) PG_GETARG_POINTER(0);
IndexBulkDeleteResult *stats = (IndexBulkDeleteResult *) PG_GETARG_POINTER(1);
IndexBulkDeleteCallback callback = (IndexBulkDeleteCallback) PG_GETARG_POINTER(2);
void *callback_state = (void *) PG_GETARG_POINTER(3);
Relation index = info->index;
BlockNumber blkno = GIN_ROOT_BLKNO;
GinVacuumState gvs;
Buffer buffer;
BlockNumber rootOfPostingTree[BLCKSZ / (sizeof(IndexTupleData) + sizeof(ItemId))];
uint32 nRoot;
/* first time through? */
if (stats == NULL)
stats = (IndexBulkDeleteResult *) palloc0(sizeof(IndexBulkDeleteResult));
/* we'll re-count the tuples each time */
stats->num_index_tuples = 0;
gvs.index = index;
gvs.result = stats;
gvs.callback = callback;
gvs.callback_state = callback_state;
initGinState(&gvs.ginstate, index);
// -------- MirroredLock ----------
MIRROREDLOCK_BUFMGR_LOCK;
buffer = ReadBuffer(index, blkno);
/* find leaf page */
for (;;)
{
Page page = BufferGetPage(buffer);
IndexTuple itup;
LockBuffer(buffer, GIN_SHARE);
Assert(!GinPageIsData(page));
if (GinPageIsLeaf(page))
{
LockBuffer(buffer, GIN_UNLOCK);
LockBuffer(buffer, GIN_EXCLUSIVE);
if (blkno == GIN_ROOT_BLKNO && !GinPageIsLeaf(page))
{
LockBuffer(buffer, GIN_UNLOCK);
continue; /* check it one more */
}
break;
}
Assert(PageGetMaxOffsetNumber(page) >= FirstOffsetNumber);
itup = (IndexTuple) PageGetItem(page, PageGetItemId(page, FirstOffsetNumber));
blkno = GinItemPointerGetBlockNumber(&(itup)->t_tid);
Assert(blkno != InvalidBlockNumber);
LockBuffer(buffer, GIN_UNLOCK);
buffer = ReleaseAndReadBuffer(buffer, index, blkno);
}
/* right now we found leftmost page in entry's BTree */
for (;;)
{
Page page = BufferGetPage(buffer);
Page resPage;
uint32 i;
Assert(!GinPageIsData(page));
resPage = ginVacuumEntryPage(&gvs, buffer, rootOfPostingTree, &nRoot);
blkno = GinPageGetOpaque(page)->rightlink;
if (resPage)
{
START_CRIT_SECTION();
PageRestoreTempPage(resPage, page);
MarkBufferDirty(buffer);
xlogVacuumPage(gvs.index, buffer);
UnlockReleaseBuffer(buffer);
END_CRIT_SECTION();
}
else
{
UnlockReleaseBuffer(buffer);
}
vacuum_delay_point();
for (i = 0; i < nRoot; i++)
{
ginVacuumPostingTree(&gvs, rootOfPostingTree[i]);
vacuum_delay_point();
}
if (blkno == InvalidBlockNumber) /* rightmost page */
break;
buffer = ReadBuffer(index, blkno);
LockBuffer(buffer, GIN_EXCLUSIVE);
}
MIRROREDLOCK_BUFMGR_UNLOCK;
// -------- MirroredLock ----------
PG_RETURN_POINTER(gvs.result);
}
/*
* Helper function to perform deletion of index entries from a bucket.
*
* This function expects that the caller has acquired a cleanup lock on the
* primary bucket page, and will return with a write lock again held on the
* primary bucket page. The lock won't necessarily be held continuously,
* though, because we'll release it when visiting overflow pages.
*
* It would be very bad if this function cleaned a page while some other
* backend was in the midst of scanning it, because hashgettuple assumes
* that the next valid TID will be greater than or equal to the current
* valid TID. There can't be any concurrent scans in progress when we first
* enter this function because of the cleanup lock we hold on the primary
* bucket page, but as soon as we release that lock, there might be. We
* handle that by conspiring to prevent those scans from passing our cleanup
* scan. To do that, we lock the next page in the bucket chain before
* releasing the lock on the previous page. (This type of lock chaining is
* not ideal, so we might want to look for a better solution at some point.)
*
* We need to retain a pin on the primary bucket to ensure that no concurrent
* split can start.
*/
void
hashbucketcleanup(Relation rel, Bucket cur_bucket, Buffer bucket_buf,
BlockNumber bucket_blkno, BufferAccessStrategy bstrategy,
uint32 maxbucket, uint32 highmask, uint32 lowmask,
double *tuples_removed, double *num_index_tuples,
bool split_cleanup,
IndexBulkDeleteCallback callback, void *callback_state)
{
BlockNumber blkno;
Buffer buf;
Bucket new_bucket PG_USED_FOR_ASSERTS_ONLY = InvalidBucket;
bool bucket_dirty = false;
blkno = bucket_blkno;
buf = bucket_buf;
if (split_cleanup)
new_bucket = _hash_get_newbucket_from_oldbucket(rel, cur_bucket,
lowmask, maxbucket);
/* Scan each page in bucket */
for (;;)
{
HashPageOpaque opaque;
OffsetNumber offno;
OffsetNumber maxoffno;
Buffer next_buf;
Page page;
OffsetNumber deletable[MaxOffsetNumber];
int ndeletable = 0;
bool retain_pin = false;
bool curr_page_dirty = false;
vacuum_delay_point();
page = BufferGetPage(buf);
opaque = (HashPageOpaque) PageGetSpecialPointer(page);
/* Scan each tuple in page */
maxoffno = PageGetMaxOffsetNumber(page);
for (offno = FirstOffsetNumber;
offno <= maxoffno;
offno = OffsetNumberNext(offno))
{
ItemPointer htup;
IndexTuple itup;
Bucket bucket;
bool kill_tuple = false;
itup = (IndexTuple) PageGetItem(page,
PageGetItemId(page, offno));
htup = &(itup->t_tid);
/*
* To remove the dead tuples, we strictly want to rely on results
* of callback function. refer btvacuumpage for detailed reason.
*/
if (callback && callback(htup, callback_state))
{
kill_tuple = true;
if (tuples_removed)
*tuples_removed += 1;
}
else if (split_cleanup)
{
/* delete the tuples that are moved by split. */
bucket = _hash_hashkey2bucket(_hash_get_indextuple_hashkey(itup),
maxbucket,
highmask,
lowmask);
/* mark the item for deletion */
if (bucket != cur_bucket)
{
/*
* We expect tuples to either belong to curent bucket or
* new_bucket. This is ensured because we don't allow
* further splits from bucket that contains garbage. See
* comments in _hash_expandtable.
*/
Assert(bucket == new_bucket);
kill_tuple = true;
}
}
if (kill_tuple)
{
/* mark the item for deletion */
deletable[ndeletable++] = offno;
}
else
{
/* we're keeping it, so count it */
if (num_index_tuples)
*num_index_tuples += 1;
}
}
/* retain the pin on primary bucket page till end of bucket scan */
if (blkno == bucket_blkno)
retain_pin = true;
else
retain_pin = false;
blkno = opaque->hasho_nextblkno;
/*
* Apply deletions, advance to next page and write page if needed.
*/
if (ndeletable > 0)
{
PageIndexMultiDelete(page, deletable, ndeletable);
bucket_dirty = true;
curr_page_dirty = true;
}
/* bail out if there are no more pages to scan. */
if (!BlockNumberIsValid(blkno))
break;
next_buf = _hash_getbuf_with_strategy(rel, blkno, HASH_WRITE,
LH_OVERFLOW_PAGE,
bstrategy);
/*
* release the lock on previous page after acquiring the lock on next
* page
*/
if (curr_page_dirty)
{
if (retain_pin)
_hash_chgbufaccess(rel, buf, HASH_WRITE, HASH_NOLOCK);
else
_hash_wrtbuf(rel, buf);
curr_page_dirty = false;
}
else if (retain_pin)
_hash_chgbufaccess(rel, buf, HASH_READ, HASH_NOLOCK);
else
_hash_relbuf(rel, buf);
buf = next_buf;
}
/*
* lock the bucket page to clear the garbage flag and squeeze the bucket.
* if the current buffer is same as bucket buffer, then we already have
* lock on bucket page.
*/
if (buf != bucket_buf)
{
_hash_relbuf(rel, buf);
_hash_chgbufaccess(rel, bucket_buf, HASH_NOLOCK, HASH_WRITE);
}
/*
* Clear the garbage flag from bucket after deleting the tuples that are
* moved by split. We purposefully clear the flag before squeeze bucket,
* so that after restart, vacuum shouldn't again try to delete the moved
* by split tuples.
*/
if (split_cleanup)
{
HashPageOpaque bucket_opaque;
Page page;
page = BufferGetPage(bucket_buf);
bucket_opaque = (HashPageOpaque) PageGetSpecialPointer(page);
bucket_opaque->hasho_flag &= ~LH_BUCKET_NEEDS_SPLIT_CLEANUP;
}
/*
* If we have deleted anything, try to compact free space. For squeezing
* the bucket, we must have a cleanup lock, else it can impact the
* ordering of tuples for a scan that has started before it.
*/
if (bucket_dirty && IsBufferCleanupOK(bucket_buf))
_hash_squeezebucket(rel, cur_bucket, bucket_blkno, bucket_buf,
bstrategy);
else
_hash_chgbufaccess(rel, bucket_buf, HASH_WRITE, HASH_NOLOCK);
}
/*
* Bulk deletion of all index entries pointing to a set of heap tuples and
* check invalid tuples left after upgrade.
* The set of target tuples is specified via a callback routine that tells
* whether any given heap tuple (identified by ItemPointer) is being deleted.
*
* Result: a palloc'd struct containing statistical info for VACUUM displays.
*/
IndexBulkDeleteResult *
gistbulkdelete(IndexVacuumInfo *info, IndexBulkDeleteResult *stats,
IndexBulkDeleteCallback callback, void *callback_state)
{
Relation rel = info->index;
GistBDItem *stack,
*ptr;
/* first time through? */
if (stats == NULL)
stats = (IndexBulkDeleteResult *) palloc0(sizeof(IndexBulkDeleteResult));
/* we'll re-count the tuples each time */
stats->estimated_count = false;
stats->num_index_tuples = 0;
stack = (GistBDItem *) palloc0(sizeof(GistBDItem));
stack->blkno = GIST_ROOT_BLKNO;
while (stack)
{
Buffer buffer;
Page page;
OffsetNumber i,
maxoff;
IndexTuple idxtuple;
ItemId iid;
buffer = ReadBufferExtended(rel, MAIN_FORKNUM, stack->blkno,
RBM_NORMAL, info->strategy);
LockBuffer(buffer, GIST_SHARE);
gistcheckpage(rel, buffer);
page = (Page) BufferGetPage(buffer);
if (GistPageIsLeaf(page))
{
OffsetNumber todelete[MaxOffsetNumber];
int ntodelete = 0;
LockBuffer(buffer, GIST_UNLOCK);
LockBuffer(buffer, GIST_EXCLUSIVE);
page = (Page) BufferGetPage(buffer);
if (stack->blkno == GIST_ROOT_BLKNO && !GistPageIsLeaf(page))
{
/* only the root can become non-leaf during relock */
UnlockReleaseBuffer(buffer);
/* one more check */
continue;
}
/*
* check for split proceeded after look at parent, we should check
* it after relock
*/
pushStackIfSplited(page, stack);
/*
* Remove deletable tuples from page
*/
maxoff = PageGetMaxOffsetNumber(page);
for (i = FirstOffsetNumber; i <= maxoff; i = OffsetNumberNext(i))
{
iid = PageGetItemId(page, i);
idxtuple = (IndexTuple) PageGetItem(page, iid);
if (callback(&(idxtuple->t_tid), callback_state))
todelete[ntodelete++] = i;
else
stats->num_index_tuples += 1;
}
stats->tuples_removed += ntodelete;
if (ntodelete)
{
START_CRIT_SECTION();
MarkBufferDirty(buffer);
PageIndexMultiDelete(page, todelete, ntodelete);
GistMarkTuplesDeleted(page);
if (RelationNeedsWAL(rel))
{
XLogRecPtr recptr;
recptr = gistXLogUpdate(buffer,
todelete, ntodelete,
NULL, 0, InvalidBuffer);
PageSetLSN(page, recptr);
}
else
PageSetLSN(page, gistGetFakeLSN(rel));
END_CRIT_SECTION();
}
}
else
{
/* check for split proceeded after look at parent */
pushStackIfSplited(page, stack);
maxoff = PageGetMaxOffsetNumber(page);
for (i = FirstOffsetNumber; i <= maxoff; i = OffsetNumberNext(i))
{
iid = PageGetItemId(page, i);
idxtuple = (IndexTuple) PageGetItem(page, iid);
ptr = (GistBDItem *) palloc(sizeof(GistBDItem));
ptr->blkno = ItemPointerGetBlockNumber(&(idxtuple->t_tid));
ptr->parentlsn = BufferGetLSNAtomic(buffer);
ptr->next = stack->next;
stack->next = ptr;
if (GistTupleIsInvalid(idxtuple))
ereport(LOG,
(errmsg("index \"%s\" contains an inner tuple marked as invalid",
RelationGetRelationName(rel)),
errdetail("This is caused by an incomplete page split at crash recovery before upgrading to PostgreSQL 9.1."),
errhint("Please REINDEX it.")));
}
}
UnlockReleaseBuffer(buffer);
ptr = stack->next;
pfree(stack);
stack = ptr;
vacuum_delay_point();
}
return stats;
}
/*
* Bulk deletion of all index entries pointing to a set of heap tuples.
* The set of target tuples is specified via a callback routine that tells
* whether any given heap tuple (identified by ItemPointer) is being deleted.
*
* Result: a palloc'd struct containing statistical info for VACUUM displays.
*/
IndexBulkDeleteResult *
hashbulkdelete(IndexVacuumInfo *info, IndexBulkDeleteResult *stats,
IndexBulkDeleteCallback callback, void *callback_state)
{
Relation rel = info->index;
double tuples_removed;
double num_index_tuples;
double orig_ntuples;
Bucket orig_maxbucket;
Bucket cur_maxbucket;
Bucket cur_bucket;
Buffer metabuf;
HashMetaPage metap;
HashMetaPageData local_metapage;
tuples_removed = 0;
num_index_tuples = 0;
/*
* Read the metapage to fetch original bucket and tuple counts. Also, we
* keep a copy of the last-seen metapage so that we can use its
* hashm_spares[] values to compute bucket page addresses. This is a bit
* hokey but perfectly safe, since the interesting entries in the spares
* array cannot change under us; and it beats rereading the metapage for
* each bucket.
*/
metabuf = _hash_getbuf(rel, HASH_METAPAGE, HASH_READ, LH_META_PAGE);
metap = HashPageGetMeta(BufferGetPage(metabuf));
orig_maxbucket = metap->hashm_maxbucket;
orig_ntuples = metap->hashm_ntuples;
memcpy(&local_metapage, metap, sizeof(local_metapage));
_hash_relbuf(rel, metabuf);
/* Scan the buckets that we know exist */
cur_bucket = 0;
cur_maxbucket = orig_maxbucket;
loop_top:
while (cur_bucket <= cur_maxbucket)
{
BlockNumber bucket_blkno;
BlockNumber blkno;
bool bucket_dirty = false;
/* Get address of bucket's start page */
bucket_blkno = BUCKET_TO_BLKNO(&local_metapage, cur_bucket);
/* Exclusive-lock the bucket so we can shrink it */
_hash_getlock(rel, bucket_blkno, HASH_EXCLUSIVE);
/* Shouldn't have any active scans locally, either */
if (_hash_has_active_scan(rel, cur_bucket))
elog(ERROR, "hash index has active scan during VACUUM");
/* Scan each page in bucket */
blkno = bucket_blkno;
while (BlockNumberIsValid(blkno))
{
Buffer buf;
Page page;
HashPageOpaque opaque;
OffsetNumber offno;
OffsetNumber maxoffno;
OffsetNumber deletable[MaxOffsetNumber];
int ndeletable = 0;
vacuum_delay_point();
buf = _hash_getbuf_with_strategy(rel, blkno, HASH_WRITE,
LH_BUCKET_PAGE | LH_OVERFLOW_PAGE,
info->strategy);
page = BufferGetPage(buf);
opaque = (HashPageOpaque) PageGetSpecialPointer(page);
Assert(opaque->hasho_bucket == cur_bucket);
/* Scan each tuple in page */
maxoffno = PageGetMaxOffsetNumber(page);
for (offno = FirstOffsetNumber;
offno <= maxoffno;
offno = OffsetNumberNext(offno))
{
IndexTuple itup;
ItemPointer htup;
itup = (IndexTuple) PageGetItem(page,
PageGetItemId(page, offno));
htup = &(itup->t_tid);
if (callback(htup, callback_state))
{
/* mark the item for deletion */
deletable[ndeletable++] = offno;
tuples_removed += 1;
}
else
num_index_tuples += 1;
}
/*
* Apply deletions and write page if needed, advance to next page.
*/
blkno = opaque->hasho_nextblkno;
if (ndeletable > 0)
{
PageIndexMultiDelete(page, deletable, ndeletable);
_hash_wrtbuf(rel, buf);
bucket_dirty = true;
}
else
_hash_relbuf(rel, buf);
}
/* If we deleted anything, try to compact free space */
if (bucket_dirty)
_hash_squeezebucket(rel, cur_bucket, bucket_blkno,
info->strategy);
/* Release bucket lock */
_hash_droplock(rel, bucket_blkno, HASH_EXCLUSIVE);
/* Advance to next bucket */
cur_bucket++;
}
/* Write-lock metapage and check for split since we started */
metabuf = _hash_getbuf(rel, HASH_METAPAGE, HASH_WRITE, LH_META_PAGE);
metap = HashPageGetMeta(BufferGetPage(metabuf));
if (cur_maxbucket != metap->hashm_maxbucket)
{
/* There's been a split, so process the additional bucket(s) */
cur_maxbucket = metap->hashm_maxbucket;
memcpy(&local_metapage, metap, sizeof(local_metapage));
_hash_relbuf(rel, metabuf);
goto loop_top;
}
/* Okay, we're really done. Update tuple count in metapage. */
if (orig_maxbucket == metap->hashm_maxbucket &&
orig_ntuples == metap->hashm_ntuples)
{
/*
* No one has split or inserted anything since start of scan, so
* believe our count as gospel.
*/
metap->hashm_ntuples = num_index_tuples;
}
else
{
/*
* Otherwise, our count is untrustworthy since we may have
* double-scanned tuples in split buckets. Proceed by dead-reckoning.
* (Note: we still return estimated_count = false, because using this
* count is better than not updating reltuples at all.)
*/
if (metap->hashm_ntuples > tuples_removed)
metap->hashm_ntuples -= tuples_removed;
else
metap->hashm_ntuples = 0;
num_index_tuples = metap->hashm_ntuples;
}
_hash_wrtbuf(rel, metabuf);
/* return statistics */
if (stats == NULL)
stats = (IndexBulkDeleteResult *) palloc0(sizeof(IndexBulkDeleteResult));
stats->estimated_count = false;
stats->num_index_tuples = num_index_tuples;
stats->tuples_removed += tuples_removed;
/* hashvacuumcleanup will fill in num_pages */
return stats;
}
/*
* VACUUM cleanup: update FSM
*/
IndexBulkDeleteResult *
gistvacuumcleanup(IndexVacuumInfo *info, IndexBulkDeleteResult *stats)
{
Relation rel = info->index;
BlockNumber npages,
blkno;
BlockNumber totFreePages;
bool needLock;
/* No-op in ANALYZE ONLY mode */
if (info->analyze_only)
return stats;
/* Set up all-zero stats if gistbulkdelete wasn't called */
if (stats == NULL)
{
stats = (IndexBulkDeleteResult *) palloc0(sizeof(IndexBulkDeleteResult));
/* use heap's tuple count */
stats->num_index_tuples = info->num_heap_tuples;
stats->estimated_count = info->estimated_count;
/*
* XXX the above is wrong if index is partial. Would it be OK to just
* return NULL, or is there work we must do below?
*/
}
/*
* Need lock unless it's local to this backend.
*/
needLock = !RELATION_IS_LOCAL(rel);
/* try to find deleted pages */
if (needLock)
LockRelationForExtension(rel, ExclusiveLock);
npages = RelationGetNumberOfBlocks(rel);
if (needLock)
UnlockRelationForExtension(rel, ExclusiveLock);
totFreePages = 0;
for (blkno = GIST_ROOT_BLKNO + 1; blkno < npages; blkno++)
{
Buffer buffer;
Page page;
vacuum_delay_point();
buffer = ReadBufferExtended(rel, MAIN_FORKNUM, blkno, RBM_NORMAL,
info->strategy);
LockBuffer(buffer, GIST_SHARE);
page = (Page) BufferGetPage(buffer);
if (PageIsNew(page) || GistPageIsDeleted(page))
{
totFreePages++;
RecordFreeIndexPage(rel, blkno);
}
UnlockReleaseBuffer(buffer);
}
/* Finally, vacuum the FSM */
IndexFreeSpaceMapVacuum(info->index);
/* return statistics */
stats->pages_free = totFreePages;
if (needLock)
LockRelationForExtension(rel, ExclusiveLock);
stats->num_pages = RelationGetNumberOfBlocks(rel);
if (needLock)
UnlockRelationForExtension(rel, ExclusiveLock);
return stats;
}
/*
* Scan through posting tree, delete empty tuples from leaf pages.
* Also, this function collects empty subtrees (with all empty leafs).
* For parents of these subtrees CleanUp lock is taken, then we call
* ScanToDelete. This is done for every inner page, which points to
* empty subtree.
*/
static bool
ginVacuumPostingTreeLeaves(GinVacuumState *gvs, BlockNumber blkno, bool isRoot)
{
Buffer buffer;
Page page;
bool hasVoidPage = FALSE;
MemoryContext oldCxt;
buffer = ReadBufferExtended(gvs->index, MAIN_FORKNUM, blkno,
RBM_NORMAL, gvs->strategy);
page = BufferGetPage(buffer);
ginTraverseLock(buffer, false);
Assert(GinPageIsData(page));
if (GinPageIsLeaf(page))
{
oldCxt = MemoryContextSwitchTo(gvs->tmpCxt);
ginVacuumPostingTreeLeaf(gvs->index, buffer, gvs);
MemoryContextSwitchTo(oldCxt);
MemoryContextReset(gvs->tmpCxt);
/* if root is a leaf page, we don't desire further processing */
if (GinDataLeafPageIsEmpty(page))
hasVoidPage = TRUE;
UnlockReleaseBuffer(buffer);
return hasVoidPage;
}
else
{
OffsetNumber i;
bool hasEmptyChild = FALSE;
bool hasNonEmptyChild = FALSE;
OffsetNumber maxoff = GinPageGetOpaque(page)->maxoff;
BlockNumber *children = palloc(sizeof(BlockNumber) * (maxoff + 1));
/*
* Read all children BlockNumbers. Not sure it is safe if there are
* many concurrent vacuums.
*/
for (i = FirstOffsetNumber; i <= maxoff; i++)
{
PostingItem *pitem = GinDataPageGetPostingItem(page, i);
children[i] = PostingItemGetBlockNumber(pitem);
}
UnlockReleaseBuffer(buffer);
for (i = FirstOffsetNumber; i <= maxoff; i++)
{
if (ginVacuumPostingTreeLeaves(gvs, children[i], FALSE))
hasEmptyChild = TRUE;
else
hasNonEmptyChild = TRUE;
}
pfree(children);
vacuum_delay_point();
/*
* All subtree is empty - just return TRUE to indicate that parent
* must do a cleanup. Unless we are ROOT an there is way to go upper.
*/
if (hasEmptyChild && !hasNonEmptyChild && !isRoot)
return TRUE;
if (hasEmptyChild)
{
DataPageDeleteStack root,
*ptr,
*tmp;
buffer = ReadBufferExtended(gvs->index, MAIN_FORKNUM, blkno,
RBM_NORMAL, gvs->strategy);
LockBufferForCleanup(buffer);
memset(&root, 0, sizeof(DataPageDeleteStack));
root.leftBlkno = InvalidBlockNumber;
root.isRoot = TRUE;
ginScanToDelete(gvs, blkno, TRUE, &root, InvalidOffsetNumber);
ptr = root.child;
while (ptr)
{
tmp = ptr->child;
pfree(ptr);
ptr = tmp;
}
UnlockReleaseBuffer(buffer);
}
/* Here we have deleted all empty subtrees */
return FALSE;
}
}
Datum
ginbulkdelete(PG_FUNCTION_ARGS)
{
IndexVacuumInfo *info = (IndexVacuumInfo *) PG_GETARG_POINTER(0);
IndexBulkDeleteResult *stats = (IndexBulkDeleteResult *) PG_GETARG_POINTER(1);
IndexBulkDeleteCallback callback = (IndexBulkDeleteCallback) PG_GETARG_POINTER(2);
void *callback_state = (void *) PG_GETARG_POINTER(3);
Relation index = info->index;
BlockNumber blkno = GIN_ROOT_BLKNO;
GinVacuumState gvs;
Buffer buffer;
BlockNumber rootOfPostingTree[BLCKSZ / (sizeof(IndexTupleData) + sizeof(ItemId))];
uint32 nRoot;
gvs.index = index;
gvs.callback = callback;
gvs.callback_state = callback_state;
gvs.strategy = info->strategy;
initGinState(&gvs.ginstate, index);
/* first time through? */
if (stats == NULL)
{
/* Yes, so initialize stats to zeroes */
stats = (IndexBulkDeleteResult *) palloc0(sizeof(IndexBulkDeleteResult));
/* and cleanup any pending inserts */
ginInsertCleanup(&gvs.ginstate, true, stats);
}
/* we'll re-count the tuples each time */
stats->num_index_tuples = 0;
gvs.result = stats;
buffer = ReadBufferExtended(index, MAIN_FORKNUM, blkno,
RBM_NORMAL, info->strategy);
/* find leaf page */
for (;;)
{
Page page = BufferGetPage(buffer);
IndexTuple itup;
LockBuffer(buffer, GIN_SHARE);
Assert(!GinPageIsData(page));
if (GinPageIsLeaf(page))
{
LockBuffer(buffer, GIN_UNLOCK);
LockBuffer(buffer, GIN_EXCLUSIVE);
if (blkno == GIN_ROOT_BLKNO && !GinPageIsLeaf(page))
{
LockBuffer(buffer, GIN_UNLOCK);
continue; /* check it one more */
}
break;
}
Assert(PageGetMaxOffsetNumber(page) >= FirstOffsetNumber);
itup = (IndexTuple) PageGetItem(page, PageGetItemId(page, FirstOffsetNumber));
blkno = GinGetDownlink(itup);
Assert(blkno != InvalidBlockNumber);
UnlockReleaseBuffer(buffer);
buffer = ReadBufferExtended(index, MAIN_FORKNUM, blkno,
RBM_NORMAL, info->strategy);
}
/* right now we found leftmost page in entry's BTree */
for (;;)
{
Page page = BufferGetPage(buffer);
Page resPage;
uint32 i;
Assert(!GinPageIsData(page));
resPage = ginVacuumEntryPage(&gvs, buffer, rootOfPostingTree, &nRoot);
blkno = GinPageGetOpaque(page)->rightlink;
if (resPage)
{
START_CRIT_SECTION();
PageRestoreTempPage(resPage, page);
MarkBufferDirty(buffer);
xlogVacuumPage(gvs.index, buffer);
UnlockReleaseBuffer(buffer);
END_CRIT_SECTION();
}
else
{
UnlockReleaseBuffer(buffer);
}
vacuum_delay_point();
for (i = 0; i < nRoot; i++)
{
ginVacuumPostingTree(&gvs, rootOfPostingTree[i]);
vacuum_delay_point();
}
if (blkno == InvalidBlockNumber) /* rightmost page */
break;
buffer = ReadBufferExtended(index, MAIN_FORKNUM, blkno,
RBM_NORMAL, info->strategy);
LockBuffer(buffer, GIN_EXCLUSIVE);
}
PG_RETURN_POINTER(gvs.result);
}
static ArrayTuple
gistVacuumUpdate(GistVacuum *gv, BlockNumber blkno, bool needunion)
{
ArrayTuple res = {NULL, 0, false};
Buffer buffer;
Page page,
tempPage = NULL;
OffsetNumber i,
maxoff;
ItemId iid;
int lenaddon = 4,
curlenaddon = 0,
nOffToDelete = 0,
nBlkToDelete = 0;
IndexTuple idxtuple,
*addon = NULL;
bool needwrite = false;
OffsetNumber offToDelete[MaxOffsetNumber];
BlockNumber blkToDelete[MaxOffsetNumber];
ItemPointerData *completed = NULL;
int ncompleted = 0,
lencompleted = 16;
vacuum_delay_point();
buffer = ReadBufferWithStrategy(gv->index, blkno, gv->strategy);
LockBuffer(buffer, GIST_EXCLUSIVE);
gistcheckpage(gv->index, buffer);
page = (Page) BufferGetPage(buffer);
maxoff = PageGetMaxOffsetNumber(page);
if (GistPageIsLeaf(page))
{
if (GistTuplesDeleted(page))
needunion = needwrite = true;
}
else
{
completed = (ItemPointerData *) palloc(sizeof(ItemPointerData) * lencompleted);
addon = (IndexTuple *) palloc(sizeof(IndexTuple) * lenaddon);
/* get copy of page to work */
tempPage = GistPageGetCopyPage(page);
for (i = FirstOffsetNumber; i <= maxoff; i = OffsetNumberNext(i))
{
ArrayTuple chldtuple;
bool needchildunion;
iid = PageGetItemId(tempPage, i);
idxtuple = (IndexTuple) PageGetItem(tempPage, iid);
needchildunion = (GistTupleIsInvalid(idxtuple)) ? true : false;
if (needchildunion)
elog(DEBUG2, "gistVacuumUpdate: need union for block %u",
ItemPointerGetBlockNumber(&(idxtuple->t_tid)));
chldtuple = gistVacuumUpdate(gv, ItemPointerGetBlockNumber(&(idxtuple->t_tid)),
needchildunion);
if (chldtuple.ituplen || chldtuple.emptypage)
{
/* update tuple or/and inserts new */
if (chldtuple.emptypage)
blkToDelete[nBlkToDelete++] = ItemPointerGetBlockNumber(&(idxtuple->t_tid));
offToDelete[nOffToDelete++] = i;
PageIndexTupleDelete(tempPage, i);
i--;
maxoff--;
needwrite = needunion = true;
if (chldtuple.ituplen)
{
Assert(chldtuple.emptypage == false);
while (curlenaddon + chldtuple.ituplen >= lenaddon)
{
lenaddon *= 2;
addon = (IndexTuple *) repalloc(addon, sizeof(IndexTuple) * lenaddon);
}
memcpy(addon + curlenaddon, chldtuple.itup, chldtuple.ituplen * sizeof(IndexTuple));
curlenaddon += chldtuple.ituplen;
if (chldtuple.ituplen > 1)
{
/*
* child was split, so we need mark completion
* insert(split)
*/
int j;
while (ncompleted + chldtuple.ituplen > lencompleted)
{
lencompleted *= 2;
completed = (ItemPointerData *) repalloc(completed, sizeof(ItemPointerData) * lencompleted);
}
for (j = 0; j < chldtuple.ituplen; j++)
{
ItemPointerCopy(&(chldtuple.itup[j]->t_tid), completed + ncompleted);
ncompleted++;
}
}
pfree(chldtuple.itup);
}
}
}
Assert(maxoff == PageGetMaxOffsetNumber(tempPage));
if (curlenaddon)
{
/* insert updated tuples */
if (gistnospace(tempPage, addon, curlenaddon, InvalidOffsetNumber, 0))
{
/* there is no space on page to insert tuples */
res = vacuumSplitPage(gv, tempPage, buffer, addon, curlenaddon);
tempPage = NULL; /* vacuumSplitPage() free tempPage */
needwrite = needunion = false; /* gistSplit already forms
* unions and writes pages */
}
else
/* enough free space */
gistfillbuffer(gv->index, tempPage, addon, curlenaddon, InvalidOffsetNumber);
}
}
/*
* If page is empty, we should remove pointer to it before deleting page
* (except root)
*/
if (blkno != GIST_ROOT_BLKNO && (PageIsEmpty(page) || (tempPage && PageIsEmpty(tempPage))))
{
/*
* New version of page is empty, so leave it unchanged, upper call
* will mark our page as deleted. In case of page split we never will
* be here...
*
* If page was empty it can't become non-empty during processing
*/
res.emptypage = true;
UnlockReleaseBuffer(buffer);
}
else
{
/* write page and remove its childs if it need */
START_CRIT_SECTION();
if (tempPage && needwrite)
{
PageRestoreTempPage(tempPage, page);
tempPage = NULL;
}
/* Empty index */
if (PageIsEmpty(page) && blkno == GIST_ROOT_BLKNO)
{
needwrite = true;
GistPageSetLeaf(page);
}
if (needwrite)
{
MarkBufferDirty(buffer);
GistClearTuplesDeleted(page);
if (!gv->index->rd_istemp)
{
XLogRecData *rdata;
XLogRecPtr recptr;
char *xlinfo;
rdata = formUpdateRdata(gv->index->rd_node, buffer,
offToDelete, nOffToDelete,
addon, curlenaddon, NULL);
xlinfo = rdata->next->data;
recptr = XLogInsert(RM_GIST_ID, XLOG_GIST_PAGE_UPDATE, rdata);
PageSetLSN(page, recptr);
PageSetTLI(page, ThisTimeLineID);
pfree(xlinfo);
pfree(rdata);
}
else
PageSetLSN(page, XLogRecPtrForTemp);
}
END_CRIT_SECTION();
if (needunion && !PageIsEmpty(page))
{
res.itup = (IndexTuple *) palloc(sizeof(IndexTuple));
res.ituplen = 1;
res.itup[0] = PageMakeUnionKey(gv, buffer);
}
UnlockReleaseBuffer(buffer);
/* delete empty children, now we havn't any links to pointed subtrees */
for (i = 0; i < nBlkToDelete; i++)
gistDeleteSubtree(gv, blkToDelete[i]);
if (ncompleted && !gv->index->rd_istemp)
gistxlogInsertCompletion(gv->index->rd_node, completed, ncompleted);
}
for (i = 0; i < curlenaddon; i++)
pfree(addon[i]);
if (addon)
pfree(addon);
if (completed)
pfree(completed);
if (tempPage)
pfree(tempPage);
return res;
}
Datum
gistvacuumcleanup(PG_FUNCTION_ARGS)
{
IndexVacuumInfo *info = (IndexVacuumInfo *) PG_GETARG_POINTER(0);
GistBulkDeleteResult *stats = (GistBulkDeleteResult *) PG_GETARG_POINTER(1);
Relation rel = info->index;
BlockNumber npages,
blkno;
BlockNumber totFreePages,
nFreePages,
*freePages,
maxFreePages;
BlockNumber lastBlock = GIST_ROOT_BLKNO,
lastFilledBlock = GIST_ROOT_BLKNO;
bool needLock;
/* Set up all-zero stats if gistbulkdelete wasn't called */
if (stats == NULL)
{
stats = (GistBulkDeleteResult *) palloc0(sizeof(GistBulkDeleteResult));
/* use heap's tuple count */
Assert(info->num_heap_tuples >= 0);
stats->std.num_index_tuples = info->num_heap_tuples;
/*
* XXX the above is wrong if index is partial. Would it be OK to just
* return NULL, or is there work we must do below?
*/
}
/* gistVacuumUpdate may cause hard work */
if (info->vacuum_full)
{
GistVacuum gv;
ArrayTuple res;
/* note: vacuum.c already acquired AccessExclusiveLock on index */
gv.index = rel;
initGISTstate(&(gv.giststate), rel);
gv.opCtx = createTempGistContext();
gv.result = stats;
gv.strategy = info->strategy;
/* walk through the entire index for update tuples */
res = gistVacuumUpdate(&gv, GIST_ROOT_BLKNO, false);
/* cleanup */
if (res.itup)
{
int i;
for (i = 0; i < res.ituplen; i++)
pfree(res.itup[i]);
pfree(res.itup);
}
freeGISTstate(&(gv.giststate));
MemoryContextDelete(gv.opCtx);
}
else if (stats->needFullVacuum)
ereport(NOTICE,
(errmsg("index \"%s\" needs VACUUM FULL or REINDEX to finish crash recovery",
RelationGetRelationName(rel))));
/*
* If vacuum full, we already have exclusive lock on the index. Otherwise,
* need lock unless it's local to this backend.
*/
if (info->vacuum_full)
needLock = false;
else
needLock = !RELATION_IS_LOCAL(rel);
/* try to find deleted pages */
if (needLock)
LockRelationForExtension(rel, ExclusiveLock);
npages = RelationGetNumberOfBlocks(rel);
if (needLock)
UnlockRelationForExtension(rel, ExclusiveLock);
maxFreePages = npages;
if (maxFreePages > MaxFSMPages)
maxFreePages = MaxFSMPages;
totFreePages = nFreePages = 0;
freePages = (BlockNumber *) palloc(sizeof(BlockNumber) * maxFreePages);
for (blkno = GIST_ROOT_BLKNO + 1; blkno < npages; blkno++)
{
Buffer buffer;
Page page;
vacuum_delay_point();
buffer = ReadBufferWithStrategy(rel, blkno, info->strategy);
LockBuffer(buffer, GIST_SHARE);
page = (Page) BufferGetPage(buffer);
if (PageIsNew(page) || GistPageIsDeleted(page))
{
if (nFreePages < maxFreePages)
freePages[nFreePages++] = blkno;
totFreePages++;
}
else
lastFilledBlock = blkno;
UnlockReleaseBuffer(buffer);
}
lastBlock = npages - 1;
if (info->vacuum_full && nFreePages > 0)
{ /* try to truncate index */
int i;
for (i = 0; i < nFreePages; i++)
if (freePages[i] >= lastFilledBlock)
{
totFreePages = nFreePages = i;
break;
}
if (lastBlock > lastFilledBlock)
RelationTruncate(rel, lastFilledBlock + 1);
stats->std.pages_removed = lastBlock - lastFilledBlock;
}
RecordIndexFreeSpace(&rel->rd_node, totFreePages, nFreePages, freePages);
pfree(freePages);
/* return statistics */
stats->std.pages_free = totFreePages;
if (needLock)
LockRelationForExtension(rel, ExclusiveLock);
stats->std.num_pages = RelationGetNumberOfBlocks(rel);
if (needLock)
UnlockRelationForExtension(rel, ExclusiveLock);
PG_RETURN_POINTER(stats);
}
/*
* Bulk deletion of all index entries pointing to a set of heap tuples and
* check invalid tuples after crash recovery.
* The set of target tuples is specified via a callback routine that tells
* whether any given heap tuple (identified by ItemPointer) is being deleted.
*
* Result: a palloc'd struct containing statistical info for VACUUM displays.
*/
Datum
gistbulkdelete(PG_FUNCTION_ARGS)
{
IndexVacuumInfo *info = (IndexVacuumInfo *) PG_GETARG_POINTER(0);
GistBulkDeleteResult *stats = (GistBulkDeleteResult *) PG_GETARG_POINTER(1);
IndexBulkDeleteCallback callback = (IndexBulkDeleteCallback) PG_GETARG_POINTER(2);
void *callback_state = (void *) PG_GETARG_POINTER(3);
Relation rel = info->index;
GistBDItem *stack,
*ptr;
/* first time through? */
if (stats == NULL)
stats = (GistBulkDeleteResult *) palloc0(sizeof(GistBulkDeleteResult));
/* we'll re-count the tuples each time */
stats->std.num_index_tuples = 0;
stack = (GistBDItem *) palloc0(sizeof(GistBDItem));
stack->blkno = GIST_ROOT_BLKNO;
while (stack)
{
Buffer buffer = ReadBufferWithStrategy(rel, stack->blkno, info->strategy);
Page page;
OffsetNumber i,
maxoff;
IndexTuple idxtuple;
ItemId iid;
LockBuffer(buffer, GIST_SHARE);
gistcheckpage(rel, buffer);
page = (Page) BufferGetPage(buffer);
if (GistPageIsLeaf(page))
{
OffsetNumber todelete[MaxOffsetNumber];
int ntodelete = 0;
LockBuffer(buffer, GIST_UNLOCK);
LockBuffer(buffer, GIST_EXCLUSIVE);
page = (Page) BufferGetPage(buffer);
if (stack->blkno == GIST_ROOT_BLKNO && !GistPageIsLeaf(page))
{
/* only the root can become non-leaf during relock */
UnlockReleaseBuffer(buffer);
/* one more check */
continue;
}
/*
* check for split proceeded after look at parent, we should check
* it after relock
*/
pushStackIfSplited(page, stack);
/*
* Remove deletable tuples from page
*/
maxoff = PageGetMaxOffsetNumber(page);
for (i = FirstOffsetNumber; i <= maxoff; i = OffsetNumberNext(i))
{
iid = PageGetItemId(page, i);
idxtuple = (IndexTuple) PageGetItem(page, iid);
if (callback(&(idxtuple->t_tid), callback_state))
{
todelete[ntodelete] = i - ntodelete;
ntodelete++;
stats->std.tuples_removed += 1;
}
else
stats->std.num_index_tuples += 1;
}
if (ntodelete)
{
START_CRIT_SECTION();
MarkBufferDirty(buffer);
for (i = 0; i < ntodelete; i++)
PageIndexTupleDelete(page, todelete[i]);
GistMarkTuplesDeleted(page);
if (!rel->rd_istemp)
{
XLogRecData *rdata;
XLogRecPtr recptr;
gistxlogPageUpdate *xlinfo;
rdata = formUpdateRdata(rel->rd_node, buffer,
todelete, ntodelete,
NULL, 0,
NULL);
xlinfo = (gistxlogPageUpdate *) rdata->next->data;
recptr = XLogInsert(RM_GIST_ID, XLOG_GIST_PAGE_UPDATE, rdata);
PageSetLSN(page, recptr);
PageSetTLI(page, ThisTimeLineID);
pfree(xlinfo);
pfree(rdata);
}
else
PageSetLSN(page, XLogRecPtrForTemp);
END_CRIT_SECTION();
}
}
else
{
/* check for split proceeded after look at parent */
pushStackIfSplited(page, stack);
maxoff = PageGetMaxOffsetNumber(page);
for (i = FirstOffsetNumber; i <= maxoff; i = OffsetNumberNext(i))
{
iid = PageGetItemId(page, i);
idxtuple = (IndexTuple) PageGetItem(page, iid);
ptr = (GistBDItem *) palloc(sizeof(GistBDItem));
ptr->blkno = ItemPointerGetBlockNumber(&(idxtuple->t_tid));
ptr->parentlsn = PageGetLSN(page);
ptr->next = stack->next;
stack->next = ptr;
if (GistTupleIsInvalid(idxtuple))
stats->needFullVacuum = true;
}
}
UnlockReleaseBuffer(buffer);
ptr = stack->next;
pfree(stack);
stack = ptr;
vacuum_delay_point();
}
PG_RETURN_POINTER(stats);
}
/*
* btvacuumpage --- VACUUM one page
*
* This processes a single page for btvacuumscan(). In some cases we
* must go back and re-examine previously-scanned pages; this routine
* recurses when necessary to handle that case.
*
* blkno is the page to process. orig_blkno is the highest block number
* reached by the outer btvacuumscan loop (the same as blkno, unless we
* are recursing to re-examine a previous page).
*/
static void
btvacuumpage(BTVacState *vstate, BlockNumber blkno, BlockNumber orig_blkno)
{
MIRROREDLOCK_BUFMGR_DECLARE;
IndexVacuumInfo *info = vstate->info;
IndexBulkDeleteResult *stats = vstate->stats;
IndexBulkDeleteCallback callback = vstate->callback;
void *callback_state = vstate->callback_state;
Relation rel = info->index;
bool delete_now;
BlockNumber recurse_to;
Buffer buf;
Page page;
BTPageOpaque opaque;
restart:
delete_now = false;
recurse_to = P_NONE;
/* call vacuum_delay_point while not holding any buffer lock */
vacuum_delay_point();
/*
* We can't use _bt_getbuf() here because it always applies
* _bt_checkpage(), which will barf on an all-zero page. We want to
* recycle all-zero pages, not fail. Also, we want to use a nondefault
* buffer access strategy.
*/
// -------- MirroredLock ----------
MIRROREDLOCK_BUFMGR_LOCK;
buf = ReadBufferWithStrategy(rel, blkno, info->strategy);
LockBuffer(buf, BT_READ);
page = BufferGetPage(buf);
opaque = (BTPageOpaque) PageGetSpecialPointer(page);
if (!PageIsNew(page))
_bt_checkpage(rel, buf);
/*
* If we are recursing, the only case we want to do anything with is a
* live leaf page having the current vacuum cycle ID. Any other state
* implies we already saw the page (eg, deleted it as being empty). In
* particular, we don't want to risk adding it to freePages twice.
*/
if (blkno != orig_blkno)
{
if (_bt_page_recyclable(page) ||
P_IGNORE(opaque) ||
!P_ISLEAF(opaque) ||
opaque->btpo_cycleid != vstate->cycleid)
{
_bt_relbuf(rel, buf);
MIRROREDLOCK_BUFMGR_UNLOCK;
// -------- MirroredLock ----------
return;
}
}
/* Page is valid, see what to do with it */
if (_bt_page_recyclable(page))
{
/* Okay to recycle this page */
if (vstate->nFreePages < vstate->maxFreePages)
vstate->freePages[vstate->nFreePages++] = blkno;
vstate->totFreePages++;
stats->pages_deleted++;
}
else if (P_ISDELETED(opaque))
{
/* Already deleted, but can't recycle yet */
stats->pages_deleted++;
}
else if (P_ISHALFDEAD(opaque))
{
/* Half-dead, try to delete */
delete_now = true;
}
else if (P_ISLEAF(opaque))
{
OffsetNumber deletable[MaxOffsetNumber];
int ndeletable;
OffsetNumber offnum,
minoff,
maxoff;
/*
* Trade in the initial read lock for a super-exclusive write lock on
* this page. We must get such a lock on every leaf page over the
* course of the vacuum scan, whether or not it actually contains any
* deletable tuples --- see nbtree/README.
*/
LockBuffer(buf, BUFFER_LOCK_UNLOCK);
LockBufferForCleanup(buf);
/*
* Check whether we need to recurse back to earlier pages. What we
* are concerned about is a page split that happened since we started
* the vacuum scan. If the split moved some tuples to a lower page
* then we might have missed 'em. If so, set up for tail recursion.
* (Must do this before possibly clearing btpo_cycleid below!)
*/
if (vstate->cycleid != 0 &&
opaque->btpo_cycleid == vstate->cycleid &&
!(opaque->btpo_flags & BTP_SPLIT_END) &&
!P_RIGHTMOST(opaque) &&
opaque->btpo_next < orig_blkno)
recurse_to = opaque->btpo_next;
/*
* Scan over all items to see which ones need deleted according to the
* callback function.
*/
ndeletable = 0;
minoff = P_FIRSTDATAKEY(opaque);
maxoff = PageGetMaxOffsetNumber(page);
if (callback)
{
for (offnum = minoff;
offnum <= maxoff;
offnum = OffsetNumberNext(offnum))
{
IndexTuple itup;
ItemPointer htup;
itup = (IndexTuple) PageGetItem(page,
PageGetItemId(page, offnum));
htup = &(itup->t_tid);
if (callback(htup, callback_state))
deletable[ndeletable++] = offnum;
}
}
/*
* Apply any needed deletes. We issue just one _bt_delitems() call
* per page, so as to minimize WAL traffic.
*/
if (ndeletable > 0)
{
_bt_delitems(rel, buf, deletable, ndeletable, true);
stats->tuples_removed += ndeletable;
/* must recompute maxoff */
maxoff = PageGetMaxOffsetNumber(page);
}
else
{
/*
* If the page has been split during this vacuum cycle, it seems
* worth expending a write to clear btpo_cycleid even if we don't
* have any deletions to do. (If we do, _bt_delitems takes care
* of this.) This ensures we won't process the page again.
*
* We treat this like a hint-bit update because there's no need to
* WAL-log it.
*/
if (vstate->cycleid != 0 &&
opaque->btpo_cycleid == vstate->cycleid)
{
opaque->btpo_cycleid = 0;
SetBufferCommitInfoNeedsSave(buf);
}
}
/*
* If it's now empty, try to delete; else count the live tuples. We
* don't delete when recursing, though, to avoid putting entries into
* freePages out-of-order (doesn't seem worth any extra code to handle
* the case).
*/
if (minoff > maxoff)
delete_now = (blkno == orig_blkno);
else
stats->num_index_tuples += maxoff - minoff + 1;
}
if (delete_now)
{
MemoryContext oldcontext;
int ndel;
/* Run pagedel in a temp context to avoid memory leakage */
MemoryContextReset(vstate->pagedelcontext);
oldcontext = MemoryContextSwitchTo(vstate->pagedelcontext);
ndel = _bt_pagedel(rel, buf, NULL, info->vacuum_full);
/* count only this page, else may double-count parent */
if (ndel)
stats->pages_deleted++;
/*
* During VACUUM FULL it's okay to recycle deleted pages immediately,
* since there can be no other transactions scanning the index. Note
* that we will only recycle the current page and not any parent pages
* that _bt_pagedel might have recursed to; this seems reasonable in
* the name of simplicity. (Trying to do otherwise would mean we'd
* have to sort the list of recyclable pages we're building.)
*/
if (ndel && info->vacuum_full)
{
if (vstate->nFreePages < vstate->maxFreePages)
vstate->freePages[vstate->nFreePages++] = blkno;
vstate->totFreePages++;
}
MemoryContextSwitchTo(oldcontext);
/* pagedel released buffer, so we shouldn't */
}
else
_bt_relbuf(rel, buf);
MIRROREDLOCK_BUFMGR_UNLOCK;
// -------- MirroredLock ----------
/*
* This is really tail recursion, but if the compiler is too stupid to
* optimize it as such, we'd eat an uncomfortably large amount of stack
* space per recursion level (due to the deletable[] array). A failure is
* improbable since the number of levels isn't likely to be large ... but
* just in case, let's hand-optimize into a loop.
*/
if (recurse_to != P_NONE)
{
blkno = recurse_to;
goto restart;
}
}
/*
* Process one page during a bulkdelete scan
*/
static void
spgvacuumpage(spgBulkDeleteState *bds, BlockNumber blkno)
{
Relation index = bds->info->index;
Buffer buffer;
Page page;
/* call vacuum_delay_point while not holding any buffer lock */
vacuum_delay_point();
buffer = ReadBufferExtended(index, MAIN_FORKNUM, blkno,
RBM_NORMAL, bds->info->strategy);
LockBuffer(buffer, BUFFER_LOCK_EXCLUSIVE);
page = (Page) BufferGetPage(buffer);
if (PageIsNew(page))
{
/*
* We found an all-zero page, which could happen if the database
* crashed just after extending the file. Initialize and recycle it.
*/
SpGistInitBuffer(buffer, 0);
SpGistPageSetDeleted(page);
/* We don't bother to WAL-log this action; easy to redo */
MarkBufferDirty(buffer);
}
else if (SpGistPageIsDeleted(page))
{
/* nothing to do */
}
else if (SpGistPageIsLeaf(page))
{
if (SpGistBlockIsRoot(blkno))
{
vacuumLeafRoot(bds, index, buffer);
/* no need for vacuumRedirectAndPlaceholder */
}
else
{
vacuumLeafPage(bds, index, buffer, false);
vacuumRedirectAndPlaceholder(index, buffer);
}
}
else
{
/* inner page */
vacuumRedirectAndPlaceholder(index, buffer);
}
/*
* The root pages must never be deleted, nor marked as available in FSM,
* because we don't want them ever returned by a search for a place to put
* a new tuple. Otherwise, check for empty/deletable page, and make sure
* FSM knows about it.
*/
if (!SpGistBlockIsRoot(blkno))
{
/* If page is now empty, mark it deleted */
if (PageIsEmpty(page) && !SpGistPageIsDeleted(page))
{
SpGistPageSetDeleted(page);
/* We don't bother to WAL-log this action; easy to redo */
MarkBufferDirty(buffer);
}
if (SpGistPageIsDeleted(page))
{
RecordFreeIndexPage(index, blkno);
bds->stats->pages_deleted++;
}
else
bds->lastFilledBlock = blkno;
}
SpGistSetLastUsedPage(index, buffer);
UnlockReleaseBuffer(buffer);
}
/*
* lazy_scan_heap() -- scan an open heap relation
*
* This routine sets commit status bits, builds lists of dead tuples
* and pages with free space, and calculates statistics on the number
* of live tuples in the heap. When done, or when we run low on space
* for dead-tuple TIDs, invoke vacuuming of indexes and heap.
*
* If there are no indexes then we just vacuum each dirty page as we
* process it, since there's no point in gathering many tuples.
*/
static void
lazy_scan_heap(Relation onerel, LVRelStats *vacrelstats,
Relation *Irel, int nindexes, bool scan_all)
{
BlockNumber nblocks,
blkno;
HeapTupleData tuple;
char *relname;
BlockNumber empty_pages,
vacuumed_pages;
double num_tuples,
tups_vacuumed,
nkeep,
nunused;
IndexBulkDeleteResult **indstats;
int i;
PGRUsage ru0;
Buffer vmbuffer = InvalidBuffer;
BlockNumber next_not_all_visible_block;
bool skipping_all_visible_blocks;
pg_rusage_init(&ru0);
relname = RelationGetRelationName(onerel);
ereport(elevel,
(errmsg("vacuuming \"%s.%s\"",
get_namespace_name(RelationGetNamespace(onerel)),
relname)));
empty_pages = vacuumed_pages = 0;
num_tuples = tups_vacuumed = nkeep = nunused = 0;
indstats = (IndexBulkDeleteResult **)
palloc0(nindexes * sizeof(IndexBulkDeleteResult *));
nblocks = RelationGetNumberOfBlocks(onerel);
vacrelstats->rel_pages = nblocks;
vacrelstats->scanned_pages = 0;
vacrelstats->nonempty_pages = 0;
vacrelstats->latestRemovedXid = InvalidTransactionId;
lazy_space_alloc(vacrelstats, nblocks);
/*
* We want to skip pages that don't require vacuuming according to the
* visibility map, but only when we can skip at least SKIP_PAGES_THRESHOLD
* consecutive pages. Since we're reading sequentially, the OS should be
* doing readahead for us, so there's no gain in skipping a page now and
* then; that's likely to disable readahead and so be counterproductive.
* Also, skipping even a single page means that we can't update
* relfrozenxid, so we only want to do it if we can skip a goodly number
* of pages.
*
* Before entering the main loop, establish the invariant that
* next_not_all_visible_block is the next block number >= blkno that's not
* all-visible according to the visibility map, or nblocks if there's no
* such block. Also, we set up the skipping_all_visible_blocks flag,
* which is needed because we need hysteresis in the decision: once we've
* started skipping blocks, we may as well skip everything up to the next
* not-all-visible block.
*
* Note: if scan_all is true, we won't actually skip any pages; but we
* maintain next_not_all_visible_block anyway, so as to set up the
* all_visible_according_to_vm flag correctly for each page.
*/
for (next_not_all_visible_block = 0;
next_not_all_visible_block < nblocks;
next_not_all_visible_block++)
{
if (!visibilitymap_test(onerel, next_not_all_visible_block, &vmbuffer))
break;
vacuum_delay_point();
}
if (next_not_all_visible_block >= SKIP_PAGES_THRESHOLD)
skipping_all_visible_blocks = true;
else
skipping_all_visible_blocks = false;
for (blkno = 0; blkno < nblocks; blkno++)
{
Buffer buf;
Page page;
OffsetNumber offnum,
maxoff;
bool tupgone,
hastup;
int prev_dead_count;
OffsetNumber frozen[MaxOffsetNumber];
int nfrozen;
Size freespace;
bool all_visible_according_to_vm;
bool all_visible;
bool has_dead_tuples;
if (blkno == next_not_all_visible_block)
{
/* Time to advance next_not_all_visible_block */
for (next_not_all_visible_block++;
next_not_all_visible_block < nblocks;
next_not_all_visible_block++)
{
if (!visibilitymap_test(onerel, next_not_all_visible_block,
&vmbuffer))
break;
vacuum_delay_point();
}
/*
* We know we can't skip the current block. But set up
* skipping_all_visible_blocks to do the right thing at the
* following blocks.
*/
if (next_not_all_visible_block - blkno > SKIP_PAGES_THRESHOLD)
skipping_all_visible_blocks = true;
else
skipping_all_visible_blocks = false;
all_visible_according_to_vm = false;
}
else
{
/* Current block is all-visible */
if (skipping_all_visible_blocks && !scan_all)
continue;
all_visible_according_to_vm = true;
}
vacuum_delay_point();
vacrelstats->scanned_pages++;
/*
* If we are close to overrunning the available space for dead-tuple
* TIDs, pause and do a cycle of vacuuming before we tackle this page.
*/
if ((vacrelstats->max_dead_tuples - vacrelstats->num_dead_tuples) < MaxHeapTuplesPerPage &&
vacrelstats->num_dead_tuples > 0)
{
/* Log cleanup info before we touch indexes */
vacuum_log_cleanup_info(onerel, vacrelstats);
/* Remove index entries */
for (i = 0; i < nindexes; i++)
lazy_vacuum_index(Irel[i],
&indstats[i],
vacrelstats);
/* Remove tuples from heap */
lazy_vacuum_heap(onerel, vacrelstats);
/*
* Forget the now-vacuumed tuples, and press on, but be careful
* not to reset latestRemovedXid since we want that value to be
* valid.
*/
vacrelstats->num_dead_tuples = 0;
vacrelstats->num_index_scans++;
}
buf = ReadBufferExtended(onerel, MAIN_FORKNUM, blkno,
RBM_NORMAL, vac_strategy);
/* We need buffer cleanup lock so that we can prune HOT chains. */
LockBufferForCleanup(buf);
page = BufferGetPage(buf);
if (PageIsNew(page))
{
/*
* An all-zeroes page could be left over if a backend extends the
* relation but crashes before initializing the page. Reclaim such
* pages for use.
*
* We have to be careful here because we could be looking at a
* page that someone has just added to the relation and not yet
* been able to initialize (see RelationGetBufferForTuple). To
* protect against that, release the buffer lock, grab the
* relation extension lock momentarily, and re-lock the buffer. If
* the page is still uninitialized by then, it must be left over
* from a crashed backend, and we can initialize it.
*
* We don't really need the relation lock when this is a new or
* temp relation, but it's probably not worth the code space to
* check that, since this surely isn't a critical path.
*
* Note: the comparable code in vacuum.c need not worry because
* it's got exclusive lock on the whole relation.
*/
LockBuffer(buf, BUFFER_LOCK_UNLOCK);
LockRelationForExtension(onerel, ExclusiveLock);
UnlockRelationForExtension(onerel, ExclusiveLock);
LockBufferForCleanup(buf);
if (PageIsNew(page))
{
ereport(WARNING,
(errmsg("relation \"%s\" page %u is uninitialized --- fixing",
relname, blkno)));
PageInit(page, BufferGetPageSize(buf), 0);
empty_pages++;
}
freespace = PageGetHeapFreeSpace(page);
MarkBufferDirty(buf);
UnlockReleaseBuffer(buf);
RecordPageWithFreeSpace(onerel, blkno, freespace);
continue;
}
if (PageIsEmpty(page))
{
empty_pages++;
freespace = PageGetHeapFreeSpace(page);
if (!PageIsAllVisible(page))
{
PageSetAllVisible(page);
SetBufferCommitInfoNeedsSave(buf);
}
LockBuffer(buf, BUFFER_LOCK_UNLOCK);
/* Update the visibility map */
if (!all_visible_according_to_vm)
{
visibilitymap_pin(onerel, blkno, &vmbuffer);
LockBuffer(buf, BUFFER_LOCK_SHARE);
if (PageIsAllVisible(page))
visibilitymap_set(onerel, blkno, PageGetLSN(page), &vmbuffer);
LockBuffer(buf, BUFFER_LOCK_UNLOCK);
}
ReleaseBuffer(buf);
RecordPageWithFreeSpace(onerel, blkno, freespace);
continue;
}
/*
* Prune all HOT-update chains in this page.
*
* We count tuples removed by the pruning step as removed by VACUUM.
*/
tups_vacuumed += heap_page_prune(onerel, buf, OldestXmin, false,
&vacrelstats->latestRemovedXid);
/*
* Now scan the page to collect vacuumable items and check for tuples
* requiring freezing.
*/
all_visible = true;
has_dead_tuples = false;
nfrozen = 0;
hastup = false;
prev_dead_count = vacrelstats->num_dead_tuples;
maxoff = PageGetMaxOffsetNumber(page);
for (offnum = FirstOffsetNumber;
offnum <= maxoff;
offnum = OffsetNumberNext(offnum))
{
ItemId itemid;
itemid = PageGetItemId(page, offnum);
/* Unused items require no processing, but we count 'em */
if (!ItemIdIsUsed(itemid))
{
nunused += 1;
continue;
}
/* Redirect items mustn't be touched */
if (ItemIdIsRedirected(itemid))
{
hastup = true; /* this page won't be truncatable */
continue;
}
ItemPointerSet(&(tuple.t_self), blkno, offnum);
/*
* DEAD item pointers are to be vacuumed normally; but we don't
* count them in tups_vacuumed, else we'd be double-counting (at
* least in the common case where heap_page_prune() just freed up
* a non-HOT tuple).
*/
if (ItemIdIsDead(itemid))
{
lazy_record_dead_tuple(vacrelstats, &(tuple.t_self));
all_visible = false;
continue;
}
Assert(ItemIdIsNormal(itemid));
tuple.t_data = (HeapTupleHeader) PageGetItem(page, itemid);
tuple.t_len = ItemIdGetLength(itemid);
tupgone = false;
switch (HeapTupleSatisfiesVacuum(tuple.t_data, OldestXmin, buf))
{
case HEAPTUPLE_DEAD:
/*
* Ordinarily, DEAD tuples would have been removed by
* heap_page_prune(), but it's possible that the tuple
* state changed since heap_page_prune() looked. In
* particular an INSERT_IN_PROGRESS tuple could have
* changed to DEAD if the inserter aborted. So this
* cannot be considered an error condition.
*
* If the tuple is HOT-updated then it must only be
* removed by a prune operation; so we keep it just as if
* it were RECENTLY_DEAD. Also, if it's a heap-only
* tuple, we choose to keep it, because it'll be a lot
* cheaper to get rid of it in the next pruning pass than
* to treat it like an indexed tuple.
*/
if (HeapTupleIsHotUpdated(&tuple) ||
HeapTupleIsHeapOnly(&tuple))
nkeep += 1;
else
tupgone = true; /* we can delete the tuple */
all_visible = false;
break;
case HEAPTUPLE_LIVE:
/* Tuple is good --- but let's do some validity checks */
if (onerel->rd_rel->relhasoids &&
!OidIsValid(HeapTupleGetOid(&tuple)))
elog(WARNING, "relation \"%s\" TID %u/%u: OID is invalid",
relname, blkno, offnum);
/*
* Is the tuple definitely visible to all transactions?
*
* NB: Like with per-tuple hint bits, we can't set the
* PD_ALL_VISIBLE flag if the inserter committed
* asynchronously. See SetHintBits for more info. Check
* that the HEAP_XMIN_COMMITTED hint bit is set because of
* that.
*/
if (all_visible)
{
TransactionId xmin;
if (!(tuple.t_data->t_infomask & HEAP_XMIN_COMMITTED))
{
all_visible = false;
break;
}
/*
* The inserter definitely committed. But is it old
* enough that everyone sees it as committed?
*/
xmin = HeapTupleHeaderGetXmin(tuple.t_data);
if (!TransactionIdPrecedes(xmin, OldestXmin))
{
all_visible = false;
break;
}
}
break;
case HEAPTUPLE_RECENTLY_DEAD:
/*
* If tuple is recently deleted then we must not remove it
* from relation.
*/
nkeep += 1;
all_visible = false;
break;
case HEAPTUPLE_INSERT_IN_PROGRESS:
/* This is an expected case during concurrent vacuum */
all_visible = false;
break;
case HEAPTUPLE_DELETE_IN_PROGRESS:
/* This is an expected case during concurrent vacuum */
all_visible = false;
break;
default:
elog(ERROR, "unexpected HeapTupleSatisfiesVacuum result");
break;
}
if (tupgone)
{
lazy_record_dead_tuple(vacrelstats, &(tuple.t_self));
HeapTupleHeaderAdvanceLatestRemovedXid(tuple.t_data,
&vacrelstats->latestRemovedXid);
tups_vacuumed += 1;
has_dead_tuples = true;
}
else
{
num_tuples += 1;
hastup = true;
/*
* Each non-removable tuple must be checked to see if it needs
* freezing. Note we already have exclusive buffer lock.
*/
if (heap_freeze_tuple(tuple.t_data, FreezeLimit,
InvalidBuffer))
frozen[nfrozen++] = offnum;
}
} /* scan along page */
/*
* If we froze any tuples, mark the buffer dirty, and write a WAL
* record recording the changes. We must log the changes to be
* crash-safe against future truncation of CLOG.
*/
if (nfrozen > 0)
{
MarkBufferDirty(buf);
if (RelationNeedsWAL(onerel))
{
XLogRecPtr recptr;
recptr = log_heap_freeze(onerel, buf, FreezeLimit,
frozen, nfrozen);
PageSetLSN(page, recptr);
PageSetTLI(page, ThisTimeLineID);
}
}
/*
* If there are no indexes then we can vacuum the page right now
* instead of doing a second scan.
*/
if (nindexes == 0 &&
vacrelstats->num_dead_tuples > 0)
{
/* Remove tuples from heap */
lazy_vacuum_page(onerel, blkno, buf, 0, vacrelstats);
/*
* Forget the now-vacuumed tuples, and press on, but be careful
* not to reset latestRemovedXid since we want that value to be
* valid.
*/
vacrelstats->num_dead_tuples = 0;
vacuumed_pages++;
}
freespace = PageGetHeapFreeSpace(page);
/* Update the all-visible flag on the page */
if (!PageIsAllVisible(page) && all_visible)
{
PageSetAllVisible(page);
SetBufferCommitInfoNeedsSave(buf);
}
/*
* It's possible for the value returned by GetOldestXmin() to move
* backwards, so it's not wrong for us to see tuples that appear to
* not be visible to everyone yet, while PD_ALL_VISIBLE is already
* set. The real safe xmin value never moves backwards, but
* GetOldestXmin() is conservative and sometimes returns a value
* that's unnecessarily small, so if we see that contradiction it just
* means that the tuples that we think are not visible to everyone yet
* actually are, and the PD_ALL_VISIBLE flag is correct.
*
* There should never be dead tuples on a page with PD_ALL_VISIBLE
* set, however.
*/
else if (PageIsAllVisible(page) && has_dead_tuples)
{
elog(WARNING, "page containing dead tuples is marked as all-visible in relation \"%s\" page %u",
relname, blkno);
PageClearAllVisible(page);
SetBufferCommitInfoNeedsSave(buf);
/*
* Normally, we would drop the lock on the heap page before
* updating the visibility map, but since this case shouldn't
* happen anyway, don't worry about that.
*/
visibilitymap_clear(onerel, blkno);
}
LockBuffer(buf, BUFFER_LOCK_UNLOCK);
/* Update the visibility map */
if (!all_visible_according_to_vm && all_visible)
{
visibilitymap_pin(onerel, blkno, &vmbuffer);
LockBuffer(buf, BUFFER_LOCK_SHARE);
if (PageIsAllVisible(page))
visibilitymap_set(onerel, blkno, PageGetLSN(page), &vmbuffer);
LockBuffer(buf, BUFFER_LOCK_UNLOCK);
}
ReleaseBuffer(buf);
/* Remember the location of the last page with nonremovable tuples */
if (hastup)
vacrelstats->nonempty_pages = blkno + 1;
/*
* If we remembered any tuples for deletion, then the page will be
* visited again by lazy_vacuum_heap, which will compute and record
* its post-compaction free space. If not, then we're done with this
* page, so remember its free space as-is. (This path will always be
* taken if there are no indexes.)
*/
if (vacrelstats->num_dead_tuples == prev_dead_count)
RecordPageWithFreeSpace(onerel, blkno, freespace);
}
/* save stats for use later */
vacrelstats->scanned_tuples = num_tuples;
vacrelstats->tuples_deleted = tups_vacuumed;
/* now we can compute the new value for pg_class.reltuples */
vacrelstats->new_rel_tuples = vac_estimate_reltuples(onerel, false,
nblocks,
vacrelstats->scanned_pages,
num_tuples);
/* If any tuples need to be deleted, perform final vacuum cycle */
/* XXX put a threshold on min number of tuples here? */
if (vacrelstats->num_dead_tuples > 0)
{
/* Log cleanup info before we touch indexes */
vacuum_log_cleanup_info(onerel, vacrelstats);
/* Remove index entries */
for (i = 0; i < nindexes; i++)
lazy_vacuum_index(Irel[i],
&indstats[i],
vacrelstats);
/* Remove tuples from heap */
lazy_vacuum_heap(onerel, vacrelstats);
vacrelstats->num_index_scans++;
}
/* Release the pin on the visibility map page */
if (BufferIsValid(vmbuffer))
{
ReleaseBuffer(vmbuffer);
vmbuffer = InvalidBuffer;
}
/* Do post-vacuum cleanup and statistics update for each index */
for (i = 0; i < nindexes; i++)
lazy_cleanup_index(Irel[i], indstats[i], vacrelstats);
/* If no indexes, make log report that lazy_vacuum_heap would've made */
if (vacuumed_pages)
ereport(elevel,
(errmsg("\"%s\": removed %.0f row versions in %u pages",
RelationGetRelationName(onerel),
tups_vacuumed, vacuumed_pages)));
ereport(elevel,
(errmsg("\"%s\": found %.0f removable, %.0f nonremovable row versions in %u out of %u pages",
RelationGetRelationName(onerel),
tups_vacuumed, num_tuples,
vacrelstats->scanned_pages, nblocks),
errdetail("%.0f dead row versions cannot be removed yet.\n"
"There were %.0f unused item pointers.\n"
"%u pages are entirely empty.\n"
"%s.",
nkeep,
nunused,
empty_pages,
pg_rusage_show(&ru0))));
}
/*
* VACUUM cleanup: update FSM
*/
Datum
gistvacuumcleanup(PG_FUNCTION_ARGS)
{
IndexVacuumInfo *info = (IndexVacuumInfo *) PG_GETARG_POINTER(0);
GistBulkDeleteResult *stats = (GistBulkDeleteResult *) PG_GETARG_POINTER(1);
Relation rel = info->index;
BlockNumber npages,
blkno;
BlockNumber totFreePages;
BlockNumber lastBlock = GIST_ROOT_BLKNO,
lastFilledBlock = GIST_ROOT_BLKNO;
bool needLock;
/* No-op in ANALYZE ONLY mode */
if (info->analyze_only)
PG_RETURN_POINTER(stats);
/* Set up all-zero stats if gistbulkdelete wasn't called */
if (stats == NULL)
{
stats = (GistBulkDeleteResult *) palloc0(sizeof(GistBulkDeleteResult));
/* use heap's tuple count */
stats->std.num_index_tuples = info->num_heap_tuples;
stats->std.estimated_count = info->estimated_count;
/*
* XXX the above is wrong if index is partial. Would it be OK to just
* return NULL, or is there work we must do below?
*/
}
if (stats->needReindex)
ereport(NOTICE,
(errmsg("index \"%s\" needs VACUUM FULL or REINDEX to finish crash recovery",
RelationGetRelationName(rel))));
/*
* Need lock unless it's local to this backend.
*/
needLock = !RELATION_IS_LOCAL(rel);
/* try to find deleted pages */
if (needLock)
LockRelationForExtension(rel, ExclusiveLock);
npages = RelationGetNumberOfBlocks(rel);
if (needLock)
UnlockRelationForExtension(rel, ExclusiveLock);
totFreePages = 0;
for (blkno = GIST_ROOT_BLKNO + 1; blkno < npages; blkno++)
{
Buffer buffer;
Page page;
vacuum_delay_point();
buffer = ReadBufferExtended(rel, MAIN_FORKNUM, blkno, RBM_NORMAL,
info->strategy);
LockBuffer(buffer, GIST_SHARE);
page = (Page) BufferGetPage(buffer);
if (PageIsNew(page) || GistPageIsDeleted(page))
{
totFreePages++;
RecordFreeIndexPage(rel, blkno);
}
else
lastFilledBlock = blkno;
UnlockReleaseBuffer(buffer);
}
lastBlock = npages - 1;
/* Finally, vacuum the FSM */
IndexFreeSpaceMapVacuum(info->index);
/* return statistics */
stats->std.pages_free = totFreePages;
if (needLock)
LockRelationForExtension(rel, ExclusiveLock);
stats->std.num_pages = RelationGetNumberOfBlocks(rel);
if (needLock)
UnlockRelationForExtension(rel, ExclusiveLock);
PG_RETURN_POINTER(stats);
}
