Preventing indirection for IndexPageGetOpaque for known-size page special areas
Started by Matthias van de Meentalmost 4 years ago25 messages
Hi,
I noticed that effectively all indexes use the special region of a
page to store some index-specific data on the page. In all cases I've
noticed, this is a constant-sized struct, located at what is
effectively a fixed offset from the end of the page; indicated on the
page by pd_special; and accessed through PageGetSpecialPointer() (that
is about equal to `page + page->pd_special` modulo typing and
assertions).
Seeing that these indexes effectively always use a constant-sized
struct as the only data in the special region; why does this code
depend on the pd_special pointer to retrieve their PageOpaque pointer?
Wouldn't a constant offset off of (page), based on the size of the
opaque structure and BLCKSZ, be enough?
Sure, in assertion builds this also validates that pd_special indeed
points in the bounds of the page, but PageGetSpecialPointer() does not
validate that the struct itself is in the bounds of the page, so
there's little guarantee that this is actually safe.
Additionally, this introduces a layer of indirection that is not
necessarily useful: for example the_bt_step_left code has no use for
the page's header information other than that it contains the
pd_special pointer (which is effectively always BLCKSZ -
MAXALIGN(sizeof(opaque))). This introduces more data and ordering
dependencies in the CPU, where this should be as simple as a constant
offset over the base page pointer.
Assuming there's no significant reason to _not_ to change this code to
a constant offset off the page pointer and only relying on pd_special
in asserts when retrieving the IndexPageOpaques, I propose the
attached patches:
0001 adds a new macro PageGetSpecialOpaque(page, opaquedatatyp); which
replaces PageGetSpecialPointer for constant sized special area
structures, and sets up the SPGist, GIST and Gin index methods to use
that instead of PageGetSpecialPointer.
0002 replaces manual PageGetSpecialPointer calls & casts in btree code
with a new macro BTPageGetOpaque, which utilizes PageGetSpecialOpaque.
0003 does the same as 0002, but for hash indexes.
A first good reason to do this is preventing further damage when a
page is corrupted: if I can somehow overwrite pd_special,
non-assert-enabled builds would start reading and writing at arbitrary
offsets from the page pointer, quite possibly in subsequent buffers
(or worse, on the stack, in case of stack-allocated blocks).
A second reason would be less indirection to get to the opaque
pointer. This should improve performance a bit in those cases where we
(initially) only want to access the [Index]PageOpaque struct.
Lastly, 0002 and 0003 remove some repetitive tasks of dealing with the
[nbtree, hash] opaques by providing a typed accessor macro similar to
what is used in the GIN and (SP-)GIST index methods; improving the
legibility of the code and decreasing the churn.
Kind regards,
Matthias van de Meent.
Attachments:
v1-0001-Add-known-size-pre-aligned-special-area-pointer-m.patchapplication/octet-stream; name=v1-0001-Add-known-size-pre-aligned-special-area-pointer-m.patchDownload
From c6191d6c52de2604d7809e5216858d5296969778 Mon Sep 17 00:00:00 2001
From: Matthias van de Meent <boekewurm+postgres@gmail.com>
Date: Wed, 16 Feb 2022 00:55:24 +0100
Subject: [PATCH v1 1/3] Add known-size pre-aligned special area pointer macro
This removes 1 layer of indirection for special areas of which we know the
type (=size) and location.
Special area access through the page header is an extra cache line that needs
to be fetched. If might only want to look at the special area, it is much
less effort to calculate the offset to the special area directly instead of
first checking the page header - saving one cache line to fetch.
Assertions are added to check that the page has a correctly sized special
area, and that the page is of the expected size. This detects data corruption,
instead of doing random reads/writes into the page or data allocated next to
the page being accessed.
Additionally, updates the GIN, GIST and SP-GIST [Index]PageGetOpaque macros
with the new pre-aligned special area accessor.
---
src/include/access/ginblock.h | 3 ++-
src/include/access/gist.h | 3 ++-
src/include/access/spgist_private.h | 3 ++-
src/include/storage/bufpage.h | 13 +++++++++++++
4 files changed, 19 insertions(+), 3 deletions(-)
diff --git a/src/include/access/ginblock.h b/src/include/access/ginblock.h
index 9347f464f3..9c2af08a27 100644
--- a/src/include/access/ginblock.h
+++ b/src/include/access/ginblock.h
@@ -108,7 +108,8 @@ typedef struct GinMetaPageData
/*
* Macros for accessing a GIN index page's opaque data
*/
-#define GinPageGetOpaque(page) ( (GinPageOpaque) PageGetSpecialPointer(page) )
+#define GinPageGetOpaque(page) \
+ ( (GinPageOpaque) PageGetSpecialOpaque(page, GinPageOpaqueData) )
#define GinPageIsLeaf(page) ( (GinPageGetOpaque(page)->flags & GIN_LEAF) != 0 )
#define GinPageSetLeaf(page) ( GinPageGetOpaque(page)->flags |= GIN_LEAF )
diff --git a/src/include/access/gist.h b/src/include/access/gist.h
index a3337627b8..987f771c4d 100644
--- a/src/include/access/gist.h
+++ b/src/include/access/gist.h
@@ -164,7 +164,8 @@ typedef struct GISTENTRY
bool leafkey;
} GISTENTRY;
-#define GistPageGetOpaque(page) ( (GISTPageOpaque) PageGetSpecialPointer(page) )
+#define GistPageGetOpaque(page) \
+ ( (GISTPageOpaque) PageGetSpecialOpaque(page, GISTPageOpaqueData) )
#define GistPageIsLeaf(page) ( GistPageGetOpaque(page)->flags & F_LEAF)
#define GIST_LEAF(entry) (GistPageIsLeaf((entry)->page))
diff --git a/src/include/access/spgist_private.h b/src/include/access/spgist_private.h
index eb56b1c6b8..b0bdc7df3e 100644
--- a/src/include/access/spgist_private.h
+++ b/src/include/access/spgist_private.h
@@ -75,7 +75,8 @@ typedef SpGistPageOpaqueData *SpGistPageOpaque;
#define SPGIST_LEAF (1<<2)
#define SPGIST_NULLS (1<<3)
-#define SpGistPageGetOpaque(page) ((SpGistPageOpaque) PageGetSpecialPointer(page))
+#define SpGistPageGetOpaque(page) \
+ ((SpGistPageOpaque) PageGetSpecialOpaque(page, SpGistPageOpaqueData))
#define SpGistPageIsMeta(page) (SpGistPageGetOpaque(page)->flags & SPGIST_META)
#define SpGistPageIsDeleted(page) (SpGistPageGetOpaque(page)->flags & SPGIST_DELETED)
#define SpGistPageIsLeaf(page) (SpGistPageGetOpaque(page)->flags & SPGIST_LEAF)
diff --git a/src/include/storage/bufpage.h b/src/include/storage/bufpage.h
index e9f253f2c8..aa395d3c39 100644
--- a/src/include/storage/bufpage.h
+++ b/src/include/storage/bufpage.h
@@ -329,6 +329,19 @@ PageValidateSpecialPointer(Page page)
(char *) ((char *) (page) + ((PageHeader) (page))->pd_special) \
)
+/*
+ * PageGetSpecialOpaque
+ * Returns pointer to special space on a page, of the given type.
+ * This removes the data dependency on the page header in builds
+ * without asserts enabled.
+ */
+#define PageGetSpecialOpaque(page, OpaqueDataTyp) \
+( \
+ AssertMacro(PageGetPageSize(page) == BLCKSZ && \
+ PageGetSpecialSize(page) == MAXALIGN(sizeof(OpaqueDataTyp))), \
+ (OpaqueDataTyp *) ((char *) (page) + (BLCKSZ - MAXALIGN(sizeof(OpaqueDataTyp)))) \
+)
+
/*
* PageGetItem
* Retrieves an item on the given page.
--
2.30.2
v1-0003-Update-hash-code-to-use-PageGetSpecialOpaque-repl.patchapplication/octet-stream; name=v1-0003-Update-hash-code-to-use-PageGetSpecialOpaque-repl.patchDownload
From e9136b78c3beac355122d9ca7721a7f4ef2f2c1e Mon Sep 17 00:00:00 2001
From: Matthias van de Meent <boekewurm+postgres@gmail.com>
Date: Wed, 16 Feb 2022 01:01:29 +0100
Subject: [PATCH v1 3/3] Update hash code to use PageGetSpecialOpaque; replace
cast+macro
As in the earlier patch, this reduces memory access requirements by directly
reading into the target space instead of needing to access the header of the
page first.
Also makes the code a lot more readable by replacing templated macro+casts
with a single macro specialized for hash page opaques.
---
contrib/pageinspect/hashfuncs.c | 6 +++---
contrib/pgstattuple/pgstatindex.c | 2 +-
contrib/pgstattuple/pgstattuple.c | 2 +-
src/backend/access/hash/hash.c | 6 +++---
src/backend/access/hash/hash_xlog.c | 26 ++++++++++++------------
src/backend/access/hash/hashinsert.c | 6 +++---
src/backend/access/hash/hashovfl.c | 22 ++++++++++----------
src/backend/access/hash/hashpage.c | 30 ++++++++++++++--------------
src/backend/access/hash/hashsearch.c | 12 +++++------
src/backend/access/hash/hashutil.c | 4 ++--
src/include/access/hash.h | 3 +++
11 files changed, 61 insertions(+), 58 deletions(-)
diff --git a/contrib/pageinspect/hashfuncs.c b/contrib/pageinspect/hashfuncs.c
index 090eba4a93..313c8ee252 100644
--- a/contrib/pageinspect/hashfuncs.c
+++ b/contrib/pageinspect/hashfuncs.c
@@ -69,7 +69,7 @@ verify_hash_page(bytea *raw_page, int flags)
(errcode(ERRCODE_INDEX_CORRUPTED),
errmsg("index table contains corrupted page")));
- pageopaque = (HashPageOpaque) PageGetSpecialPointer(page);
+ pageopaque = HashPageGetOpaque(page);
if (pageopaque->hasho_page_id != HASHO_PAGE_ID)
ereport(ERROR,
(errcode(ERRCODE_INVALID_PARAMETER_VALUE),
@@ -151,7 +151,7 @@ static void
GetHashPageStatistics(Page page, HashPageStat *stat)
{
OffsetNumber maxoff = PageGetMaxOffsetNumber(page);
- HashPageOpaque opaque = (HashPageOpaque) PageGetSpecialPointer(page);
+ HashPageOpaque opaque = HashPageGetOpaque(page);
int off;
stat->dead_items = stat->live_items = 0;
@@ -203,7 +203,7 @@ hash_page_type(PG_FUNCTION_ARGS)
type = "unused";
else
{
- opaque = (HashPageOpaque) PageGetSpecialPointer(page);
+ opaque = HashPageGetOpaque(page);
/* page type (flags) */
pagetype = opaque->hasho_flag & LH_PAGE_TYPE;
diff --git a/contrib/pgstattuple/pgstatindex.c b/contrib/pgstattuple/pgstatindex.c
index b9939451b4..e1048e47ff 100644
--- a/contrib/pgstattuple/pgstatindex.c
+++ b/contrib/pgstattuple/pgstatindex.c
@@ -641,7 +641,7 @@ pgstathashindex(PG_FUNCTION_ARGS)
HashPageOpaque opaque;
int pagetype;
- opaque = (HashPageOpaque) PageGetSpecialPointer(page);
+ opaque = HashPageGetOpaque(page);
pagetype = opaque->hasho_flag & LH_PAGE_TYPE;
if (pagetype == LH_BUCKET_PAGE)
diff --git a/contrib/pgstattuple/pgstattuple.c b/contrib/pgstattuple/pgstattuple.c
index a9658d389b..3094566908 100644
--- a/contrib/pgstattuple/pgstattuple.c
+++ b/contrib/pgstattuple/pgstattuple.c
@@ -452,7 +452,7 @@ pgstat_hash_page(pgstattuple_type *stat, Relation rel, BlockNumber blkno,
{
HashPageOpaque opaque;
- opaque = (HashPageOpaque) PageGetSpecialPointer(page);
+ opaque = HashPageGetOpaque(page);
switch (opaque->hasho_flag & LH_PAGE_TYPE)
{
case LH_UNUSED_PAGE:
diff --git a/src/backend/access/hash/hash.c b/src/backend/access/hash/hash.c
index a259a301fa..fd1a7119b6 100644
--- a/src/backend/access/hash/hash.c
+++ b/src/backend/access/hash/hash.c
@@ -515,7 +515,7 @@ loop_top:
_hash_checkpage(rel, buf, LH_BUCKET_PAGE);
page = BufferGetPage(buf);
- bucket_opaque = (HashPageOpaque) PageGetSpecialPointer(page);
+ bucket_opaque = HashPageGetOpaque(page);
/*
* If the bucket contains tuples that are moved by split, then we need
@@ -717,7 +717,7 @@ hashbucketcleanup(Relation rel, Bucket cur_bucket, Buffer bucket_buf,
vacuum_delay_point();
page = BufferGetPage(buf);
- opaque = (HashPageOpaque) PageGetSpecialPointer(page);
+ opaque = HashPageGetOpaque(page);
/* Scan each tuple in page */
maxoffno = PageGetMaxOffsetNumber(page);
@@ -884,7 +884,7 @@ hashbucketcleanup(Relation rel, Bucket cur_bucket, Buffer bucket_buf,
Page page;
page = BufferGetPage(bucket_buf);
- bucket_opaque = (HashPageOpaque) PageGetSpecialPointer(page);
+ bucket_opaque = HashPageGetOpaque(page);
/* No ereport(ERROR) until changes are logged */
START_CRIT_SECTION();
diff --git a/src/backend/access/hash/hash_xlog.c b/src/backend/access/hash/hash_xlog.c
index 55937b9a68..62dbfc3a5d 100644
--- a/src/backend/access/hash/hash_xlog.c
+++ b/src/backend/access/hash/hash_xlog.c
@@ -203,7 +203,7 @@ hash_xlog_add_ovfl_page(XLogReaderState *record)
true);
/* update backlink */
ovflpage = BufferGetPage(ovflbuf);
- ovflopaque = (HashPageOpaque) PageGetSpecialPointer(ovflpage);
+ ovflopaque = HashPageGetOpaque(ovflpage);
ovflopaque->hasho_prevblkno = leftblk;
PageSetLSN(ovflpage, lsn);
@@ -215,7 +215,7 @@ hash_xlog_add_ovfl_page(XLogReaderState *record)
HashPageOpaque leftopaque;
leftpage = BufferGetPage(leftbuf);
- leftopaque = (HashPageOpaque) PageGetSpecialPointer(leftpage);
+ leftopaque = HashPageGetOpaque(leftpage);
leftopaque->hasho_nextblkno = rightblk;
PageSetLSN(leftpage, lsn);
@@ -342,7 +342,7 @@ hash_xlog_split_allocate_page(XLogReaderState *record)
HashPageOpaque oldopaque;
oldpage = BufferGetPage(oldbuf);
- oldopaque = (HashPageOpaque) PageGetSpecialPointer(oldpage);
+ oldopaque = HashPageGetOpaque(oldpage);
oldopaque->hasho_flag = xlrec->old_bucket_flag;
oldopaque->hasho_prevblkno = xlrec->new_bucket;
@@ -465,7 +465,7 @@ hash_xlog_split_complete(XLogReaderState *record)
HashPageOpaque oldopaque;
oldpage = BufferGetPage(oldbuf);
- oldopaque = (HashPageOpaque) PageGetSpecialPointer(oldpage);
+ oldopaque = HashPageGetOpaque(oldpage);
oldopaque->hasho_flag = xlrec->old_bucket_flag;
@@ -488,7 +488,7 @@ hash_xlog_split_complete(XLogReaderState *record)
HashPageOpaque nopaque;
newpage = BufferGetPage(newbuf);
- nopaque = (HashPageOpaque) PageGetSpecialPointer(newpage);
+ nopaque = HashPageGetOpaque(newpage);
nopaque->hasho_flag = xlrec->new_bucket_flag;
@@ -710,7 +710,7 @@ hash_xlog_squeeze_page(XLogReaderState *record)
*/
if (xldata->is_prev_bucket_same_wrt)
{
- HashPageOpaque writeopaque = (HashPageOpaque) PageGetSpecialPointer(writepage);
+ HashPageOpaque writeopaque = HashPageGetOpaque(writepage);
writeopaque->hasho_nextblkno = xldata->nextblkno;
}
@@ -729,7 +729,7 @@ hash_xlog_squeeze_page(XLogReaderState *record)
_hash_pageinit(ovflpage, BufferGetPageSize(ovflbuf));
- ovflopaque = (HashPageOpaque) PageGetSpecialPointer(ovflpage);
+ ovflopaque = HashPageGetOpaque(ovflpage);
ovflopaque->hasho_prevblkno = InvalidBlockNumber;
ovflopaque->hasho_nextblkno = InvalidBlockNumber;
@@ -748,7 +748,7 @@ hash_xlog_squeeze_page(XLogReaderState *record)
XLogReadBufferForRedo(record, 3, &prevbuf) == BLK_NEEDS_REDO)
{
Page prevpage = BufferGetPage(prevbuf);
- HashPageOpaque prevopaque = (HashPageOpaque) PageGetSpecialPointer(prevpage);
+ HashPageOpaque prevopaque = HashPageGetOpaque(prevpage);
prevopaque->hasho_nextblkno = xldata->nextblkno;
@@ -766,7 +766,7 @@ hash_xlog_squeeze_page(XLogReaderState *record)
if (XLogReadBufferForRedo(record, 4, &nextbuf) == BLK_NEEDS_REDO)
{
Page nextpage = BufferGetPage(nextbuf);
- HashPageOpaque nextopaque = (HashPageOpaque) PageGetSpecialPointer(nextpage);
+ HashPageOpaque nextopaque = HashPageGetOpaque(nextpage);
nextopaque->hasho_prevblkno = xldata->prevblkno;
@@ -903,7 +903,7 @@ hash_xlog_delete(XLogReaderState *record)
{
HashPageOpaque pageopaque;
- pageopaque = (HashPageOpaque) PageGetSpecialPointer(page);
+ pageopaque = HashPageGetOpaque(page);
pageopaque->hasho_flag &= ~LH_PAGE_HAS_DEAD_TUPLES;
}
@@ -933,7 +933,7 @@ hash_xlog_split_cleanup(XLogReaderState *record)
page = (Page) BufferGetPage(buffer);
- bucket_opaque = (HashPageOpaque) PageGetSpecialPointer(page);
+ bucket_opaque = HashPageGetOpaque(page);
bucket_opaque->hasho_flag &= ~LH_BUCKET_NEEDS_SPLIT_CLEANUP;
PageSetLSN(page, lsn);
MarkBufferDirty(buffer);
@@ -1024,7 +1024,7 @@ hash_xlog_vacuum_one_page(XLogReaderState *record)
* Mark the page as not containing any LP_DEAD items. See comments in
* _hash_vacuum_one_page() for details.
*/
- pageopaque = (HashPageOpaque) PageGetSpecialPointer(page);
+ pageopaque = HashPageGetOpaque(page);
pageopaque->hasho_flag &= ~LH_PAGE_HAS_DEAD_TUPLES;
PageSetLSN(page, lsn);
@@ -1116,7 +1116,7 @@ hash_mask(char *pagedata, BlockNumber blkno)
mask_page_hint_bits(page);
mask_unused_space(page);
- opaque = (HashPageOpaque) PageGetSpecialPointer(page);
+ opaque = HashPageGetOpaque(page);
pagetype = opaque->hasho_flag & LH_PAGE_TYPE;
if (pagetype == LH_UNUSED_PAGE)
diff --git a/src/backend/access/hash/hashinsert.c b/src/backend/access/hash/hashinsert.c
index faf609c157..4f2fecb908 100644
--- a/src/backend/access/hash/hashinsert.c
+++ b/src/backend/access/hash/hashinsert.c
@@ -95,7 +95,7 @@ restart_insert:
bucket_buf = buf;
page = BufferGetPage(buf);
- pageopaque = (HashPageOpaque) PageGetSpecialPointer(page);
+ pageopaque = HashPageGetOpaque(page);
bucket = pageopaque->hasho_bucket;
/*
@@ -183,7 +183,7 @@ restart_insert:
/* should fit now, given test above */
Assert(PageGetFreeSpace(page) >= itemsz);
}
- pageopaque = (HashPageOpaque) PageGetSpecialPointer(page);
+ pageopaque = HashPageGetOpaque(page);
Assert((pageopaque->hasho_flag & LH_PAGE_TYPE) == LH_OVERFLOW_PAGE);
Assert(pageopaque->hasho_bucket == bucket);
}
@@ -384,7 +384,7 @@ _hash_vacuum_one_page(Relation rel, Relation hrel, Buffer metabuf, Buffer buf)
* check it. Remember that LH_PAGE_HAS_DEAD_TUPLES is only a hint
* anyway.
*/
- pageopaque = (HashPageOpaque) PageGetSpecialPointer(page);
+ pageopaque = HashPageGetOpaque(page);
pageopaque->hasho_flag &= ~LH_PAGE_HAS_DEAD_TUPLES;
metap = HashPageGetMeta(BufferGetPage(metabuf));
diff --git a/src/backend/access/hash/hashovfl.c b/src/backend/access/hash/hashovfl.c
index 4836875196..e34cfc302d 100644
--- a/src/backend/access/hash/hashovfl.c
+++ b/src/backend/access/hash/hashovfl.c
@@ -159,7 +159,7 @@ _hash_addovflpage(Relation rel, Buffer metabuf, Buffer buf, bool retain_pin)
BlockNumber nextblkno;
page = BufferGetPage(buf);
- pageopaque = (HashPageOpaque) PageGetSpecialPointer(page);
+ pageopaque = HashPageGetOpaque(page);
nextblkno = pageopaque->hasho_nextblkno;
if (!BlockNumberIsValid(nextblkno))
@@ -364,7 +364,7 @@ found:
/* initialize new overflow page */
ovflpage = BufferGetPage(ovflbuf);
- ovflopaque = (HashPageOpaque) PageGetSpecialPointer(ovflpage);
+ ovflopaque = HashPageGetOpaque(ovflpage);
ovflopaque->hasho_prevblkno = BufferGetBlockNumber(buf);
ovflopaque->hasho_nextblkno = InvalidBlockNumber;
ovflopaque->hasho_bucket = pageopaque->hasho_bucket;
@@ -516,7 +516,7 @@ _hash_freeovflpage(Relation rel, Buffer bucketbuf, Buffer ovflbuf,
_hash_checkpage(rel, ovflbuf, LH_OVERFLOW_PAGE);
ovflblkno = BufferGetBlockNumber(ovflbuf);
ovflpage = BufferGetPage(ovflbuf);
- ovflopaque = (HashPageOpaque) PageGetSpecialPointer(ovflpage);
+ ovflopaque = HashPageGetOpaque(ovflpage);
nextblkno = ovflopaque->hasho_nextblkno;
prevblkno = ovflopaque->hasho_prevblkno;
writeblkno = BufferGetBlockNumber(wbuf);
@@ -600,7 +600,7 @@ _hash_freeovflpage(Relation rel, Buffer bucketbuf, Buffer ovflbuf,
*/
_hash_pageinit(ovflpage, BufferGetPageSize(ovflbuf));
- ovflopaque = (HashPageOpaque) PageGetSpecialPointer(ovflpage);
+ ovflopaque = HashPageGetOpaque(ovflpage);
ovflopaque->hasho_prevblkno = InvalidBlockNumber;
ovflopaque->hasho_nextblkno = InvalidBlockNumber;
@@ -613,7 +613,7 @@ _hash_freeovflpage(Relation rel, Buffer bucketbuf, Buffer ovflbuf,
if (BufferIsValid(prevbuf))
{
Page prevpage = BufferGetPage(prevbuf);
- HashPageOpaque prevopaque = (HashPageOpaque) PageGetSpecialPointer(prevpage);
+ HashPageOpaque prevopaque = HashPageGetOpaque(prevpage);
Assert(prevopaque->hasho_bucket == bucket);
prevopaque->hasho_nextblkno = nextblkno;
@@ -622,7 +622,7 @@ _hash_freeovflpage(Relation rel, Buffer bucketbuf, Buffer ovflbuf,
if (BufferIsValid(nextbuf))
{
Page nextpage = BufferGetPage(nextbuf);
- HashPageOpaque nextopaque = (HashPageOpaque) PageGetSpecialPointer(nextpage);
+ HashPageOpaque nextopaque = HashPageGetOpaque(nextpage);
Assert(nextopaque->hasho_bucket == bucket);
nextopaque->hasho_prevblkno = prevblkno;
@@ -751,7 +751,7 @@ _hash_initbitmapbuffer(Buffer buf, uint16 bmsize, bool initpage)
_hash_pageinit(pg, BufferGetPageSize(buf));
/* initialize the page's special space */
- op = (HashPageOpaque) PageGetSpecialPointer(pg);
+ op = HashPageGetOpaque(pg);
op->hasho_prevblkno = InvalidBlockNumber;
op->hasho_nextblkno = InvalidBlockNumber;
op->hasho_bucket = InvalidBucket;
@@ -824,7 +824,7 @@ _hash_squeezebucket(Relation rel,
wblkno = bucket_blkno;
wbuf = bucket_buf;
wpage = BufferGetPage(wbuf);
- wopaque = (HashPageOpaque) PageGetSpecialPointer(wpage);
+ wopaque = HashPageGetOpaque(wpage);
/*
* if there aren't any overflow pages, there's nothing to squeeze. caller
@@ -855,7 +855,7 @@ _hash_squeezebucket(Relation rel,
LH_OVERFLOW_PAGE,
bstrategy);
rpage = BufferGetPage(rbuf);
- ropaque = (HashPageOpaque) PageGetSpecialPointer(rpage);
+ ropaque = HashPageGetOpaque(rpage);
Assert(ropaque->hasho_bucket == bucket);
} while (BlockNumberIsValid(ropaque->hasho_nextblkno));
@@ -1005,7 +1005,7 @@ readpage:
wbuf = next_wbuf;
wpage = BufferGetPage(wbuf);
- wopaque = (HashPageOpaque) PageGetSpecialPointer(wpage);
+ wopaque = HashPageGetOpaque(wpage);
Assert(wopaque->hasho_bucket == bucket);
retain_pin = false;
@@ -1076,7 +1076,7 @@ readpage:
LH_OVERFLOW_PAGE,
bstrategy);
rpage = BufferGetPage(rbuf);
- ropaque = (HashPageOpaque) PageGetSpecialPointer(rpage);
+ ropaque = HashPageGetOpaque(rpage);
Assert(ropaque->hasho_bucket == bucket);
}
diff --git a/src/backend/access/hash/hashpage.c b/src/backend/access/hash/hashpage.c
index 28c5297a1d..39206d1942 100644
--- a/src/backend/access/hash/hashpage.c
+++ b/src/backend/access/hash/hashpage.c
@@ -166,7 +166,7 @@ _hash_initbuf(Buffer buf, uint32 max_bucket, uint32 num_bucket, uint32 flag,
if (initpage)
_hash_pageinit(page, BufferGetPageSize(buf));
- pageopaque = (HashPageOpaque) PageGetSpecialPointer(page);
+ pageopaque = HashPageGetOpaque(page);
/*
* Set hasho_prevblkno with current hashm_maxbucket. This value will be
@@ -529,7 +529,7 @@ _hash_init_metabuffer(Buffer buf, double num_tuples, RegProcedure procid,
if (initpage)
_hash_pageinit(page, BufferGetPageSize(buf));
- pageopaque = (HashPageOpaque) PageGetSpecialPointer(page);
+ pageopaque = HashPageGetOpaque(page);
pageopaque->hasho_prevblkno = InvalidBlockNumber;
pageopaque->hasho_nextblkno = InvalidBlockNumber;
pageopaque->hasho_bucket = InvalidBucket;
@@ -693,7 +693,7 @@ restart_expand:
goto fail;
opage = BufferGetPage(buf_oblkno);
- oopaque = (HashPageOpaque) PageGetSpecialPointer(opage);
+ oopaque = HashPageGetOpaque(opage);
/*
* We want to finish the split from a bucket as there is no apparent
@@ -864,7 +864,7 @@ restart_expand:
lowmask = metap->hashm_lowmask;
opage = BufferGetPage(buf_oblkno);
- oopaque = (HashPageOpaque) PageGetSpecialPointer(opage);
+ oopaque = HashPageGetOpaque(opage);
/*
* Mark the old bucket to indicate that split is in progress. (At
@@ -883,7 +883,7 @@ restart_expand:
* initialize the new bucket's primary page and mark it to indicate that
* split is in progress.
*/
- nopaque = (HashPageOpaque) PageGetSpecialPointer(npage);
+ nopaque = HashPageGetOpaque(npage);
nopaque->hasho_prevblkno = maxbucket;
nopaque->hasho_nextblkno = InvalidBlockNumber;
nopaque->hasho_bucket = new_bucket;
@@ -1010,7 +1010,7 @@ _hash_alloc_buckets(Relation rel, BlockNumber firstblock, uint32 nblocks)
*/
_hash_pageinit(page, BLCKSZ);
- ovflopaque = (HashPageOpaque) PageGetSpecialPointer(page);
+ ovflopaque = HashPageGetOpaque(page);
ovflopaque->hasho_prevblkno = InvalidBlockNumber;
ovflopaque->hasho_nextblkno = InvalidBlockNumber;
@@ -1091,11 +1091,11 @@ _hash_splitbucket(Relation rel,
bucket_obuf = obuf;
opage = BufferGetPage(obuf);
- oopaque = (HashPageOpaque) PageGetSpecialPointer(opage);
+ oopaque = HashPageGetOpaque(opage);
bucket_nbuf = nbuf;
npage = BufferGetPage(nbuf);
- nopaque = (HashPageOpaque) PageGetSpecialPointer(npage);
+ nopaque = HashPageGetOpaque(npage);
/* Copy the predicate locks from old bucket to new bucket. */
PredicateLockPageSplit(rel,
@@ -1198,7 +1198,7 @@ _hash_splitbucket(Relation rel,
/* chain to a new overflow page */
nbuf = _hash_addovflpage(rel, metabuf, nbuf, (nbuf == bucket_nbuf));
npage = BufferGetPage(nbuf);
- nopaque = (HashPageOpaque) PageGetSpecialPointer(npage);
+ nopaque = HashPageGetOpaque(npage);
}
itups[nitups++] = new_itup;
@@ -1251,7 +1251,7 @@ _hash_splitbucket(Relation rel,
/* Else, advance to next old page */
obuf = _hash_getbuf(rel, oblkno, HASH_READ, LH_OVERFLOW_PAGE);
opage = BufferGetPage(obuf);
- oopaque = (HashPageOpaque) PageGetSpecialPointer(opage);
+ oopaque = HashPageGetOpaque(opage);
}
/*
@@ -1264,11 +1264,11 @@ _hash_splitbucket(Relation rel,
*/
LockBuffer(bucket_obuf, BUFFER_LOCK_EXCLUSIVE);
opage = BufferGetPage(bucket_obuf);
- oopaque = (HashPageOpaque) PageGetSpecialPointer(opage);
+ oopaque = HashPageGetOpaque(opage);
LockBuffer(bucket_nbuf, BUFFER_LOCK_EXCLUSIVE);
npage = BufferGetPage(bucket_nbuf);
- nopaque = (HashPageOpaque) PageGetSpecialPointer(npage);
+ nopaque = HashPageGetOpaque(npage);
START_CRIT_SECTION();
@@ -1392,7 +1392,7 @@ _hash_finish_split(Relation rel, Buffer metabuf, Buffer obuf, Bucket obucket,
bucket_nbuf = nbuf;
npage = BufferGetPage(nbuf);
- npageopaque = (HashPageOpaque) PageGetSpecialPointer(npage);
+ npageopaque = HashPageGetOpaque(npage);
/* Scan each tuple in new page */
nmaxoffnum = PageGetMaxOffsetNumber(npage);
@@ -1446,7 +1446,7 @@ _hash_finish_split(Relation rel, Buffer metabuf, Buffer obuf, Bucket obucket,
}
npage = BufferGetPage(bucket_nbuf);
- npageopaque = (HashPageOpaque) PageGetSpecialPointer(npage);
+ npageopaque = HashPageGetOpaque(npage);
nbucket = npageopaque->hasho_bucket;
_hash_splitbucket(rel, metabuf, obucket,
@@ -1587,7 +1587,7 @@ _hash_getbucketbuf_from_hashkey(Relation rel, uint32 hashkey, int access,
/* Fetch the primary bucket page for the bucket */
buf = _hash_getbuf(rel, blkno, access, LH_BUCKET_PAGE);
page = BufferGetPage(buf);
- opaque = (HashPageOpaque) PageGetSpecialPointer(page);
+ opaque = HashPageGetOpaque(page);
Assert(opaque->hasho_bucket == bucket);
Assert(opaque->hasho_prevblkno != InvalidBlockNumber);
diff --git a/src/backend/access/hash/hashsearch.c b/src/backend/access/hash/hashsearch.c
index 7ca542a3fb..524af27409 100644
--- a/src/backend/access/hash/hashsearch.c
+++ b/src/backend/access/hash/hashsearch.c
@@ -187,7 +187,7 @@ _hash_readnext(IndexScanDesc scan,
{
*pagep = BufferGetPage(*bufp);
TestForOldSnapshot(scan->xs_snapshot, rel, *pagep);
- *opaquep = (HashPageOpaque) PageGetSpecialPointer(*pagep);
+ *opaquep = HashPageGetOpaque(*pagep);
}
}
@@ -233,7 +233,7 @@ _hash_readprev(IndexScanDesc scan,
LH_BUCKET_PAGE | LH_OVERFLOW_PAGE);
*pagep = BufferGetPage(*bufp);
TestForOldSnapshot(scan->xs_snapshot, rel, *pagep);
- *opaquep = (HashPageOpaque) PageGetSpecialPointer(*pagep);
+ *opaquep = HashPageGetOpaque(*pagep);
/*
* We always maintain the pin on bucket page for whole scan operation,
@@ -258,7 +258,7 @@ _hash_readprev(IndexScanDesc scan,
LockBuffer(*bufp, BUFFER_LOCK_SHARE);
*pagep = BufferGetPage(*bufp);
- *opaquep = (HashPageOpaque) PageGetSpecialPointer(*pagep);
+ *opaquep = HashPageGetOpaque(*pagep);
/* move to the end of bucket chain */
while (BlockNumberIsValid((*opaquep)->hasho_nextblkno))
@@ -352,7 +352,7 @@ _hash_first(IndexScanDesc scan, ScanDirection dir)
PredicateLockPage(rel, BufferGetBlockNumber(buf), scan->xs_snapshot);
page = BufferGetPage(buf);
TestForOldSnapshot(scan->xs_snapshot, rel, page);
- opaque = (HashPageOpaque) PageGetSpecialPointer(page);
+ opaque = HashPageGetOpaque(page);
bucket = opaque->hasho_bucket;
so->hashso_bucket_buf = buf;
@@ -398,7 +398,7 @@ _hash_first(IndexScanDesc scan, ScanDirection dir)
LockBuffer(buf, BUFFER_LOCK_SHARE);
page = BufferGetPage(buf);
- opaque = (HashPageOpaque) PageGetSpecialPointer(page);
+ opaque = HashPageGetOpaque(page);
Assert(opaque->hasho_bucket == bucket);
if (H_BUCKET_BEING_POPULATED(opaque))
@@ -463,7 +463,7 @@ _hash_readpage(IndexScanDesc scan, Buffer *bufP, ScanDirection dir)
Assert(BufferIsValid(buf));
_hash_checkpage(rel, buf, LH_BUCKET_PAGE | LH_OVERFLOW_PAGE);
page = BufferGetPage(buf);
- opaque = (HashPageOpaque) PageGetSpecialPointer(page);
+ opaque = HashPageGetOpaque(page);
so->currPos.buf = buf;
so->currPos.currPage = BufferGetBlockNumber(buf);
diff --git a/src/backend/access/hash/hashutil.c b/src/backend/access/hash/hashutil.c
index edb6fa968f..fe37bc47cb 100644
--- a/src/backend/access/hash/hashutil.c
+++ b/src/backend/access/hash/hashutil.c
@@ -239,7 +239,7 @@ _hash_checkpage(Relation rel, Buffer buf, int flags)
if (flags)
{
- HashPageOpaque opaque = (HashPageOpaque) PageGetSpecialPointer(page);
+ HashPageOpaque opaque = HashPageGetOpaque(page);
if ((opaque->hasho_flag & flags) == 0)
ereport(ERROR,
@@ -574,7 +574,7 @@ _hash_kill_items(IndexScanDesc scan)
buf = _hash_getbuf(rel, blkno, HASH_READ, LH_OVERFLOW_PAGE);
page = BufferGetPage(buf);
- opaque = (HashPageOpaque) PageGetSpecialPointer(page);
+ opaque = HashPageGetOpaque(page);
maxoff = PageGetMaxOffsetNumber(page);
for (i = 0; i < numKilled; i++)
diff --git a/src/include/access/hash.h b/src/include/access/hash.h
index cd7b2a53d8..d3fd8bf239 100644
--- a/src/include/access/hash.h
+++ b/src/include/access/hash.h
@@ -85,6 +85,9 @@ typedef struct HashPageOpaqueData
typedef HashPageOpaqueData *HashPageOpaque;
+#define HashPageGetOpaque(page) \
+ ((HashPageOpaque) PageGetSpecialOpaque((page), HashPageOpaqueData))
+
#define H_NEEDS_SPLIT_CLEANUP(opaque) (((opaque)->hasho_flag & LH_BUCKET_NEEDS_SPLIT_CLEANUP) != 0)
#define H_BUCKET_BEING_SPLIT(opaque) (((opaque)->hasho_flag & LH_BUCKET_BEING_SPLIT) != 0)
#define H_BUCKET_BEING_POPULATED(opaque) (((opaque)->hasho_flag & LH_BUCKET_BEING_POPULATED) != 0)
--
2.30.2
v1-0002-Update-nbtree-code-to-use-PageGetSpecialOpaque-re.patchapplication/octet-stream; name=v1-0002-Update-nbtree-code-to-use-PageGetSpecialOpaque-re.patchDownload
From 2784b1e6365b8f755d815e5354bc00f693c1b75a Mon Sep 17 00:00:00 2001
From: Matthias van de Meent <boekewurm+postgres@gmail.com>
Date: Wed, 16 Feb 2022 00:59:50 +0100
Subject: [PATCH v1 2/3] Update nbtree code to use PageGetSpecialOpaque;
replace cast+macro
As in the earlier patch, this reduces memory access requirements by directly
reading into the target space instead of needing to access the header of the
page first.
Also makes the code a lot more readable by replacing templated macro+casts
with a single macro specialized for BTree page opaques.
---
contrib/amcheck/verify_nbtree.c | 32 +++++++-------
contrib/pageinspect/btreefuncs.c | 6 +--
contrib/pgstattuple/pgstatindex.c | 2 +-
contrib/pgstattuple/pgstattuple.c | 2 +-
src/backend/access/nbtree/nbtdedup.c | 6 +--
src/backend/access/nbtree/nbtinsert.c | 46 ++++++++++----------
src/backend/access/nbtree/nbtpage.c | 58 ++++++++++++-------------
src/backend/access/nbtree/nbtree.c | 2 +-
src/backend/access/nbtree/nbtsearch.c | 34 +++++++--------
src/backend/access/nbtree/nbtsort.c | 12 ++---
src/backend/access/nbtree/nbtsplitloc.c | 2 +-
src/backend/access/nbtree/nbtutils.c | 6 +--
src/backend/access/nbtree/nbtxlog.c | 34 +++++++--------
src/include/access/nbtree.h | 13 ++++--
14 files changed, 131 insertions(+), 124 deletions(-)
diff --git a/contrib/amcheck/verify_nbtree.c b/contrib/amcheck/verify_nbtree.c
index d2510ee648..70278c4f93 100644
--- a/contrib/amcheck/verify_nbtree.c
+++ b/contrib/amcheck/verify_nbtree.c
@@ -691,7 +691,7 @@ bt_check_level_from_leftmost(BtreeCheckState *state, BtreeLevel level)
state->target = palloc_btree_page(state, state->targetblock);
state->targetlsn = PageGetLSN(state->target);
- opaque = (BTPageOpaque) PageGetSpecialPointer(state->target);
+ opaque = BTPageGetOpaque(state->target);
if (P_IGNORE(opaque))
{
@@ -927,7 +927,7 @@ bt_recheck_sibling_links(BtreeCheckState *state,
LockBuffer(lbuf, BT_READ);
_bt_checkpage(state->rel, lbuf);
page = BufferGetPage(lbuf);
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
if (P_ISDELETED(opaque))
{
/*
@@ -951,7 +951,7 @@ bt_recheck_sibling_links(BtreeCheckState *state,
LockBuffer(newtargetbuf, BT_READ);
_bt_checkpage(state->rel, newtargetbuf);
page = BufferGetPage(newtargetbuf);
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
/* btpo_prev_from_target may have changed; update it */
btpo_prev_from_target = opaque->btpo_prev;
}
@@ -1049,7 +1049,7 @@ bt_target_page_check(BtreeCheckState *state)
OffsetNumber max;
BTPageOpaque topaque;
- topaque = (BTPageOpaque) PageGetSpecialPointer(state->target);
+ topaque = BTPageGetOpaque(state->target);
max = PageGetMaxOffsetNumber(state->target);
elog(DEBUG2, "verifying %u items on %s block %u", max,
@@ -1478,7 +1478,7 @@ bt_target_page_check(BtreeCheckState *state)
/* Get fresh copy of target page */
state->target = palloc_btree_page(state, state->targetblock);
/* Note that we deliberately do not update target LSN */
- topaque = (BTPageOpaque) PageGetSpecialPointer(state->target);
+ topaque = BTPageGetOpaque(state->target);
/*
* All !readonly checks now performed; just return
@@ -1552,7 +1552,7 @@ bt_right_page_check_scankey(BtreeCheckState *state)
OffsetNumber nline;
/* Determine target's next block number */
- opaque = (BTPageOpaque) PageGetSpecialPointer(state->target);
+ opaque = BTPageGetOpaque(state->target);
/* If target is already rightmost, no right sibling; nothing to do here */
if (P_RIGHTMOST(opaque))
@@ -1588,7 +1588,7 @@ bt_right_page_check_scankey(BtreeCheckState *state)
CHECK_FOR_INTERRUPTS();
rightpage = palloc_btree_page(state, targetnext);
- opaque = (BTPageOpaque) PageGetSpecialPointer(rightpage);
+ opaque = BTPageGetOpaque(rightpage);
if (!P_IGNORE(opaque) || P_RIGHTMOST(opaque))
break;
@@ -1893,7 +1893,7 @@ bt_child_highkey_check(BtreeCheckState *state,
else
page = palloc_btree_page(state, blkno);
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
/* The first page we visit at the level should be leftmost */
if (first && !BlockNumberIsValid(state->prevrightlink) && !P_LEFTMOST(opaque))
@@ -1971,7 +1971,7 @@ bt_child_highkey_check(BtreeCheckState *state,
else
pivotkey_offset = target_downlinkoffnum;
- topaque = (BTPageOpaque) PageGetSpecialPointer(state->target);
+ topaque = BTPageGetOpaque(state->target);
if (!offset_is_negative_infinity(topaque, pivotkey_offset))
{
@@ -2128,9 +2128,9 @@ bt_child_check(BtreeCheckState *state, BTScanInsert targetkey,
* Check all items, rather than checking just the first and trusting that
* the operator class obeys the transitive law.
*/
- topaque = (BTPageOpaque) PageGetSpecialPointer(state->target);
+ topaque = BTPageGetOpaque(state->target);
child = palloc_btree_page(state, childblock);
- copaque = (BTPageOpaque) PageGetSpecialPointer(child);
+ copaque = BTPageGetOpaque(child);
maxoffset = PageGetMaxOffsetNumber(child);
/*
@@ -2235,7 +2235,7 @@ static void
bt_downlink_missing_check(BtreeCheckState *state, bool rightsplit,
BlockNumber blkno, Page page)
{
- BTPageOpaque opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ BTPageOpaque opaque = BTPageGetOpaque(page);
ItemId itemid;
IndexTuple itup;
Page child;
@@ -2319,7 +2319,7 @@ bt_downlink_missing_check(BtreeCheckState *state, bool rightsplit,
CHECK_FOR_INTERRUPTS();
child = palloc_btree_page(state, childblk);
- copaque = (BTPageOpaque) PageGetSpecialPointer(child);
+ copaque = BTPageGetOpaque(child);
if (P_ISLEAF(copaque))
break;
@@ -2780,7 +2780,7 @@ invariant_l_offset(BtreeCheckState *state, BTScanInsert key,
bool nonpivot;
ritup = (IndexTuple) PageGetItem(state->target, itemid);
- topaque = (BTPageOpaque) PageGetSpecialPointer(state->target);
+ topaque = BTPageGetOpaque(state->target);
nonpivot = P_ISLEAF(topaque) && upperbound >= P_FIRSTDATAKEY(topaque);
/* Get number of keys + heap TID for item to the right */
@@ -2895,7 +2895,7 @@ invariant_l_nontarget_offset(BtreeCheckState *state, BTScanInsert key,
bool nonpivot;
child = (IndexTuple) PageGetItem(nontarget, itemid);
- copaque = (BTPageOpaque) PageGetSpecialPointer(nontarget);
+ copaque = BTPageGetOpaque(nontarget);
nonpivot = P_ISLEAF(copaque) && upperbound >= P_FIRSTDATAKEY(copaque);
/* Get number of keys + heap TID for child/non-target item */
@@ -2954,7 +2954,7 @@ palloc_btree_page(BtreeCheckState *state, BlockNumber blocknum)
memcpy(page, BufferGetPage(buffer), BLCKSZ);
UnlockReleaseBuffer(buffer);
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
if (P_ISMETA(opaque) && blocknum != BTREE_METAPAGE)
ereport(ERROR,
diff --git a/contrib/pageinspect/btreefuncs.c b/contrib/pageinspect/btreefuncs.c
index 03debe336b..4aaff114c6 100644
--- a/contrib/pageinspect/btreefuncs.c
+++ b/contrib/pageinspect/btreefuncs.c
@@ -93,7 +93,7 @@ GetBTPageStatistics(BlockNumber blkno, Buffer buffer, BTPageStat *stat)
Page page = BufferGetPage(buffer);
PageHeader phdr = (PageHeader) page;
OffsetNumber maxoff = PageGetMaxOffsetNumber(page);
- BTPageOpaque opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ BTPageOpaque opaque = BTPageGetOpaque(page);
int item_size = 0;
int off;
@@ -521,7 +521,7 @@ bt_page_items_internal(PG_FUNCTION_ARGS, enum pageinspect_version ext_version)
uargs->offset = FirstOffsetNumber;
- opaque = (BTPageOpaque) PageGetSpecialPointer(uargs->page);
+ opaque = BTPageGetOpaque(uargs->page);
if (!P_ISDELETED(opaque))
fctx->max_calls = PageGetMaxOffsetNumber(uargs->page);
@@ -617,7 +617,7 @@ bt_page_items_bytea(PG_FUNCTION_ARGS)
uargs->offset = FirstOffsetNumber;
- opaque = (BTPageOpaque) PageGetSpecialPointer(uargs->page);
+ opaque = BTPageGetOpaque(uargs->page);
if (P_ISMETA(opaque))
ereport(ERROR,
diff --git a/contrib/pgstattuple/pgstatindex.c b/contrib/pgstattuple/pgstatindex.c
index 6c4b053dd0..b9939451b4 100644
--- a/contrib/pgstattuple/pgstatindex.c
+++ b/contrib/pgstattuple/pgstatindex.c
@@ -281,7 +281,7 @@ pgstatindex_impl(Relation rel, FunctionCallInfo fcinfo)
LockBuffer(buffer, BUFFER_LOCK_SHARE);
page = BufferGetPage(buffer);
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
/*
* Determine page type, and update totals.
diff --git a/contrib/pgstattuple/pgstattuple.c b/contrib/pgstattuple/pgstattuple.c
index c9b8f01f9b..a9658d389b 100644
--- a/contrib/pgstattuple/pgstattuple.c
+++ b/contrib/pgstattuple/pgstattuple.c
@@ -415,7 +415,7 @@ pgstat_btree_page(pgstattuple_type *stat, Relation rel, BlockNumber blkno,
{
BTPageOpaque opaque;
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
if (P_IGNORE(opaque))
{
/* deleted or half-dead page */
diff --git a/src/backend/access/nbtree/nbtdedup.c b/src/backend/access/nbtree/nbtdedup.c
index 4c48554aec..3e11805293 100644
--- a/src/backend/access/nbtree/nbtdedup.c
+++ b/src/backend/access/nbtree/nbtdedup.c
@@ -62,7 +62,7 @@ _bt_dedup_pass(Relation rel, Buffer buf, Relation heapRel, IndexTuple newitem,
minoff,
maxoff;
Page page = BufferGetPage(buf);
- BTPageOpaque opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ BTPageOpaque opaque = BTPageGetOpaque(page);
Page newpage;
BTDedupState state;
Size pagesaving = 0;
@@ -231,7 +231,7 @@ _bt_dedup_pass(Relation rel, Buffer buf, Relation heapRel, IndexTuple newitem,
*/
if (P_HAS_GARBAGE(opaque))
{
- BTPageOpaque nopaque = (BTPageOpaque) PageGetSpecialPointer(newpage);
+ BTPageOpaque nopaque = BTPageGetOpaque(newpage);
nopaque->btpo_flags &= ~BTP_HAS_GARBAGE;
}
@@ -310,7 +310,7 @@ _bt_bottomupdel_pass(Relation rel, Buffer buf, Relation heapRel,
minoff,
maxoff;
Page page = BufferGetPage(buf);
- BTPageOpaque opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ BTPageOpaque opaque = BTPageGetOpaque(page);
BTDedupState state;
TM_IndexDeleteOp delstate;
bool neverdedup;
diff --git a/src/backend/access/nbtree/nbtinsert.c b/src/backend/access/nbtree/nbtinsert.c
index 68628ec000..f6f4af8bfe 100644
--- a/src/backend/access/nbtree/nbtinsert.c
+++ b/src/backend/access/nbtree/nbtinsert.c
@@ -329,7 +329,7 @@ _bt_search_insert(Relation rel, BTInsertState insertstate)
_bt_checkpage(rel, insertstate->buf);
page = BufferGetPage(insertstate->buf);
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
/*
* Check if the page is still the rightmost leaf page and has
@@ -428,7 +428,7 @@ _bt_check_unique(Relation rel, BTInsertState insertstate, Relation heapRel,
InitDirtySnapshot(SnapshotDirty);
page = BufferGetPage(insertstate->buf);
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
maxoff = PageGetMaxOffsetNumber(page);
/*
@@ -733,7 +733,7 @@ _bt_check_unique(Relation rel, BTInsertState insertstate, Relation heapRel,
nbuf = _bt_relandgetbuf(rel, nbuf, nblkno, BT_READ);
page = BufferGetPage(nbuf);
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
if (!P_IGNORE(opaque))
break;
if (P_RIGHTMOST(opaque))
@@ -822,7 +822,7 @@ _bt_findinsertloc(Relation rel,
BTPageOpaque opaque;
OffsetNumber newitemoff;
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
/* Check 1/3 of a page restriction */
if (unlikely(insertstate->itemsz > BTMaxItemSize(page)))
@@ -888,7 +888,7 @@ _bt_findinsertloc(Relation rel,
_bt_stepright(rel, insertstate, stack);
/* Update local state after stepping right */
page = BufferGetPage(insertstate->buf);
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
/* Assume duplicates (if checkingunique) */
uniquedup = true;
}
@@ -972,7 +972,7 @@ _bt_findinsertloc(Relation rel,
_bt_stepright(rel, insertstate, stack);
/* Update local state after stepping right */
page = BufferGetPage(insertstate->buf);
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
}
}
@@ -1030,7 +1030,7 @@ _bt_stepright(Relation rel, BTInsertState insertstate, BTStack stack)
BlockNumber rblkno;
page = BufferGetPage(insertstate->buf);
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
rbuf = InvalidBuffer;
rblkno = opaque->btpo_next;
@@ -1038,7 +1038,7 @@ _bt_stepright(Relation rel, BTInsertState insertstate, BTStack stack)
{
rbuf = _bt_relandgetbuf(rel, rbuf, rblkno, BT_WRITE);
page = BufferGetPage(rbuf);
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
/*
* If this page was incompletely split, finish the split now. We do
@@ -1120,7 +1120,7 @@ _bt_insertonpg(Relation rel,
IndexTuple nposting = NULL;
page = BufferGetPage(buf);
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
isleaf = P_ISLEAF(opaque);
isroot = P_ISROOT(opaque);
isrightmost = P_RIGHTMOST(opaque);
@@ -1296,7 +1296,7 @@ _bt_insertonpg(Relation rel,
if (!isleaf)
{
Page cpage = BufferGetPage(cbuf);
- BTPageOpaque cpageop = (BTPageOpaque) PageGetSpecialPointer(cpage);
+ BTPageOpaque cpageop = BTPageGetOpaque(cpage);
Assert(P_INCOMPLETE_SPLIT(cpageop));
cpageop->btpo_flags &= ~BTP_INCOMPLETE_SPLIT;
@@ -1504,7 +1504,7 @@ _bt_split(Relation rel, BTScanInsert itup_key, Buffer buf, Buffer cbuf,
* only workspace.
*/
origpage = BufferGetPage(buf);
- oopaque = (BTPageOpaque) PageGetSpecialPointer(origpage);
+ oopaque = BTPageGetOpaque(origpage);
isleaf = P_ISLEAF(oopaque);
isrightmost = P_RIGHTMOST(oopaque);
maxoff = PageGetMaxOffsetNumber(origpage);
@@ -1540,7 +1540,7 @@ _bt_split(Relation rel, BTScanInsert itup_key, Buffer buf, Buffer cbuf,
/* Allocate temp buffer for leftpage */
leftpage = PageGetTempPage(origpage);
_bt_pageinit(leftpage, BufferGetPageSize(buf));
- lopaque = (BTPageOpaque) PageGetSpecialPointer(leftpage);
+ lopaque = BTPageGetOpaque(leftpage);
/*
* leftpage won't be the root when we're done. Also, clear the SPLIT_END
@@ -1716,7 +1716,7 @@ _bt_split(Relation rel, BTScanInsert itup_key, Buffer buf, Buffer cbuf,
rightpage = BufferGetPage(rbuf);
rightpagenumber = BufferGetBlockNumber(rbuf);
/* rightpage was initialized by _bt_getbuf */
- ropaque = (BTPageOpaque) PageGetSpecialPointer(rightpage);
+ ropaque = BTPageGetOpaque(rightpage);
/*
* Finish off remaining leftpage special area fields. They cannot be set
@@ -1887,7 +1887,7 @@ _bt_split(Relation rel, BTScanInsert itup_key, Buffer buf, Buffer cbuf,
{
sbuf = _bt_getbuf(rel, oopaque->btpo_next, BT_WRITE);
spage = BufferGetPage(sbuf);
- sopaque = (BTPageOpaque) PageGetSpecialPointer(spage);
+ sopaque = BTPageGetOpaque(spage);
if (sopaque->btpo_prev != origpagenumber)
{
memset(rightpage, 0, BufferGetPageSize(rbuf));
@@ -1952,7 +1952,7 @@ _bt_split(Relation rel, BTScanInsert itup_key, Buffer buf, Buffer cbuf,
if (!isleaf)
{
Page cpage = BufferGetPage(cbuf);
- BTPageOpaque cpageop = (BTPageOpaque) PageGetSpecialPointer(cpage);
+ BTPageOpaque cpageop = BTPageGetOpaque(cpage);
cpageop->btpo_flags &= ~BTP_INCOMPLETE_SPLIT;
MarkBufferDirty(cbuf);
@@ -2139,7 +2139,7 @@ _bt_insert_parent(Relation rel,
BTPageOpaque opaque;
elog(DEBUG2, "concurrent ROOT page split");
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
/*
* We should never reach here when a leaf page split takes place
@@ -2230,7 +2230,7 @@ void
_bt_finish_split(Relation rel, Buffer lbuf, BTStack stack)
{
Page lpage = BufferGetPage(lbuf);
- BTPageOpaque lpageop = (BTPageOpaque) PageGetSpecialPointer(lpage);
+ BTPageOpaque lpageop = BTPageGetOpaque(lpage);
Buffer rbuf;
Page rpage;
BTPageOpaque rpageop;
@@ -2242,7 +2242,7 @@ _bt_finish_split(Relation rel, Buffer lbuf, BTStack stack)
/* Lock right sibling, the one missing the downlink */
rbuf = _bt_getbuf(rel, lpageop->btpo_next, BT_WRITE);
rpage = BufferGetPage(rbuf);
- rpageop = (BTPageOpaque) PageGetSpecialPointer(rpage);
+ rpageop = BTPageGetOpaque(rpage);
/* Could this be a root split? */
if (!stack)
@@ -2320,7 +2320,7 @@ _bt_getstackbuf(Relation rel, BTStack stack, BlockNumber child)
buf = _bt_getbuf(rel, blkno, BT_WRITE);
page = BufferGetPage(buf);
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
if (P_INCOMPLETE_SPLIT(opaque))
{
@@ -2451,7 +2451,7 @@ _bt_newroot(Relation rel, Buffer lbuf, Buffer rbuf)
lbkno = BufferGetBlockNumber(lbuf);
rbkno = BufferGetBlockNumber(rbuf);
lpage = BufferGetPage(lbuf);
- lopaque = (BTPageOpaque) PageGetSpecialPointer(lpage);
+ lopaque = BTPageGetOpaque(lpage);
/* get a new root page */
rootbuf = _bt_getbuf(rel, P_NEW, BT_WRITE);
@@ -2492,11 +2492,11 @@ _bt_newroot(Relation rel, Buffer lbuf, Buffer rbuf)
_bt_upgrademetapage(metapg);
/* set btree special data */
- rootopaque = (BTPageOpaque) PageGetSpecialPointer(rootpage);
+ rootopaque = BTPageGetOpaque(rootpage);
rootopaque->btpo_prev = rootopaque->btpo_next = P_NONE;
rootopaque->btpo_flags = BTP_ROOT;
rootopaque->btpo_level =
- ((BTPageOpaque) PageGetSpecialPointer(lpage))->btpo_level + 1;
+ (BTPageGetOpaque(lpage))->btpo_level + 1;
rootopaque->btpo_cycleid = 0;
/* update metapage data */
@@ -2680,7 +2680,7 @@ _bt_delete_or_dedup_one_page(Relation rel, Relation heapRel,
Buffer buffer = insertstate->buf;
BTScanInsert itup_key = insertstate->itup_key;
Page page = BufferGetPage(buffer);
- BTPageOpaque opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ BTPageOpaque opaque = BTPageGetOpaque(page);
Assert(P_ISLEAF(opaque));
Assert(simpleonly || itup_key->heapkeyspace);
diff --git a/src/backend/access/nbtree/nbtpage.c b/src/backend/access/nbtree/nbtpage.c
index 6b5f01e1d0..20adb602a4 100644
--- a/src/backend/access/nbtree/nbtpage.c
+++ b/src/backend/access/nbtree/nbtpage.c
@@ -85,7 +85,7 @@ _bt_initmetapage(Page page, BlockNumber rootbknum, uint32 level,
metad->btm_last_cleanup_num_heap_tuples = -1.0;
metad->btm_allequalimage = allequalimage;
- metaopaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ metaopaque = BTPageGetOpaque(page);
metaopaque->btpo_flags = BTP_META;
/*
@@ -112,7 +112,7 @@ _bt_upgrademetapage(Page page)
BTPageOpaque metaopaque PG_USED_FOR_ASSERTS_ONLY;
metad = BTPageGetMeta(page);
- metaopaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ metaopaque = BTPageGetOpaque(page);
/* It must be really a meta page of upgradable version */
Assert(metaopaque->btpo_flags & BTP_META);
@@ -148,7 +148,7 @@ _bt_getmeta(Relation rel, Buffer metabuf)
BTMetaPageData *metad;
metapg = BufferGetPage(metabuf);
- metaopaque = (BTPageOpaque) PageGetSpecialPointer(metapg);
+ metaopaque = BTPageGetOpaque(metapg);
metad = BTPageGetMeta(metapg);
/* sanity-check the metapage */
@@ -372,7 +372,7 @@ _bt_getroot(Relation rel, int access)
rootbuf = _bt_getbuf(rel, rootblkno, BT_READ);
rootpage = BufferGetPage(rootbuf);
- rootopaque = (BTPageOpaque) PageGetSpecialPointer(rootpage);
+ rootopaque = BTPageGetOpaque(rootpage);
/*
* Since the cache might be stale, we check the page more carefully
@@ -440,7 +440,7 @@ _bt_getroot(Relation rel, int access)
rootbuf = _bt_getbuf(rel, P_NEW, BT_WRITE);
rootblkno = BufferGetBlockNumber(rootbuf);
rootpage = BufferGetPage(rootbuf);
- rootopaque = (BTPageOpaque) PageGetSpecialPointer(rootpage);
+ rootopaque = BTPageGetOpaque(rootpage);
rootopaque->btpo_prev = rootopaque->btpo_next = P_NONE;
rootopaque->btpo_flags = (BTP_LEAF | BTP_ROOT);
rootopaque->btpo_level = 0;
@@ -534,7 +534,7 @@ _bt_getroot(Relation rel, int access)
{
rootbuf = _bt_relandgetbuf(rel, rootbuf, rootblkno, BT_READ);
rootpage = BufferGetPage(rootbuf);
- rootopaque = (BTPageOpaque) PageGetSpecialPointer(rootpage);
+ rootopaque = BTPageGetOpaque(rootpage);
if (!P_IGNORE(rootopaque))
break;
@@ -598,7 +598,7 @@ _bt_gettrueroot(Relation rel)
metabuf = _bt_getbuf(rel, BTREE_METAPAGE, BT_READ);
metapg = BufferGetPage(metabuf);
- metaopaque = (BTPageOpaque) PageGetSpecialPointer(metapg);
+ metaopaque = BTPageGetOpaque(metapg);
metad = BTPageGetMeta(metapg);
if (!P_ISMETA(metaopaque) ||
@@ -637,7 +637,7 @@ _bt_gettrueroot(Relation rel)
{
rootbuf = _bt_relandgetbuf(rel, rootbuf, rootblkno, BT_READ);
rootpage = BufferGetPage(rootbuf);
- rootopaque = (BTPageOpaque) PageGetSpecialPointer(rootpage);
+ rootopaque = BTPageGetOpaque(rootpage);
if (!P_IGNORE(rootopaque))
break;
@@ -1220,7 +1220,7 @@ _bt_delitems_vacuum(Relation rel, Buffer buf,
* We can clear the vacuum cycle ID since this page has certainly been
* processed by the current vacuum scan.
*/
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
opaque->btpo_cycleid = 0;
/*
@@ -1338,7 +1338,7 @@ _bt_delitems_delete(Relation rel, Buffer buf, TransactionId latestRemovedXid,
* Unlike _bt_delitems_vacuum, we *must not* clear the vacuum cycle ID at
* this point. The VACUUM command alone controls vacuum cycle IDs.
*/
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
/*
* Clear the BTP_HAS_GARBAGE page flag.
@@ -1718,7 +1718,7 @@ _bt_leftsib_splitflag(Relation rel, BlockNumber leftsib, BlockNumber target)
buf = _bt_getbuf(rel, leftsib, BT_READ);
page = BufferGetPage(buf);
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
/*
* If the left sibling was concurrently split, so that its next-pointer
@@ -1773,7 +1773,7 @@ _bt_rightsib_halfdeadflag(Relation rel, BlockNumber leafrightsib)
buf = _bt_getbuf(rel, leafrightsib, BT_READ);
page = BufferGetPage(buf);
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
Assert(P_ISLEAF(opaque) && !P_ISDELETED(opaque));
result = P_ISHALFDEAD(opaque);
@@ -1842,7 +1842,7 @@ _bt_pagedel(Relation rel, Buffer leafbuf, BTVacState *vstate)
for (;;)
{
page = BufferGetPage(leafbuf);
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
/*
* Internal pages are never deleted directly, only as part of deleting
@@ -2099,7 +2099,7 @@ _bt_mark_page_halfdead(Relation rel, Buffer leafbuf, BTStack stack)
IndexTupleData trunctuple;
page = BufferGetPage(leafbuf);
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
Assert(!P_RIGHTMOST(opaque) && !P_ISROOT(opaque) &&
P_ISLEAF(opaque) && !P_IGNORE(opaque) &&
@@ -2154,7 +2154,7 @@ _bt_mark_page_halfdead(Relation rel, Buffer leafbuf, BTStack stack)
* before entering the critical section --- otherwise it'd be a PANIC.
*/
page = BufferGetPage(subtreeparent);
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
#ifdef USE_ASSERT_CHECKING
@@ -2201,7 +2201,7 @@ _bt_mark_page_halfdead(Relation rel, Buffer leafbuf, BTStack stack)
* nbtree/README.
*/
page = BufferGetPage(subtreeparent);
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
itemid = PageGetItemId(page, poffset);
itup = (IndexTuple) PageGetItem(page, itemid);
@@ -2216,7 +2216,7 @@ _bt_mark_page_halfdead(Relation rel, Buffer leafbuf, BTStack stack)
* is set to InvalidBlockNumber.
*/
page = BufferGetPage(leafbuf);
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
opaque->btpo_flags |= BTP_HALF_DEAD;
Assert(PageGetMaxOffsetNumber(page) == P_HIKEY);
@@ -2253,7 +2253,7 @@ _bt_mark_page_halfdead(Relation rel, Buffer leafbuf, BTStack stack)
XLogRegisterBuffer(1, subtreeparent, REGBUF_STANDARD);
page = BufferGetPage(leafbuf);
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
xlrec.leftblk = opaque->btpo_prev;
xlrec.rightblk = opaque->btpo_next;
@@ -2325,7 +2325,7 @@ _bt_unlink_halfdead_page(Relation rel, Buffer leafbuf, BlockNumber scanblkno,
BlockNumber leaftopparent;
page = BufferGetPage(leafbuf);
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
Assert(P_ISLEAF(opaque) && !P_ISDELETED(opaque) && P_ISHALFDEAD(opaque));
@@ -2364,7 +2364,7 @@ _bt_unlink_halfdead_page(Relation rel, Buffer leafbuf, BlockNumber scanblkno,
/* Fetch the block number of the target's left sibling */
buf = _bt_getbuf(rel, target, BT_READ);
page = BufferGetPage(buf);
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
leftsib = opaque->btpo_prev;
targetlevel = opaque->btpo_level;
Assert(targetlevel > 0);
@@ -2391,7 +2391,7 @@ _bt_unlink_halfdead_page(Relation rel, Buffer leafbuf, BlockNumber scanblkno,
{
lbuf = _bt_getbuf(rel, leftsib, BT_WRITE);
page = BufferGetPage(lbuf);
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
while (P_ISDELETED(opaque) || opaque->btpo_next != target)
{
bool leftsibvalid = true;
@@ -2441,7 +2441,7 @@ _bt_unlink_halfdead_page(Relation rel, Buffer leafbuf, BlockNumber scanblkno,
/* step right one page */
lbuf = _bt_getbuf(rel, leftsib, BT_WRITE);
page = BufferGetPage(lbuf);
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
}
}
else
@@ -2450,7 +2450,7 @@ _bt_unlink_halfdead_page(Relation rel, Buffer leafbuf, BlockNumber scanblkno,
/* Next write-lock the target page itself */
_bt_lockbuf(rel, buf, BT_WRITE);
page = BufferGetPage(buf);
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
/*
* Check page is still empty etc, else abandon deletion. This is just for
@@ -2505,7 +2505,7 @@ _bt_unlink_halfdead_page(Relation rel, Buffer leafbuf, BlockNumber scanblkno,
rightsib = opaque->btpo_next;
rbuf = _bt_getbuf(rel, rightsib, BT_WRITE);
page = BufferGetPage(rbuf);
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
if (opaque->btpo_prev != target)
ereport(ERROR,
(errcode(ERRCODE_INDEX_CORRUPTED),
@@ -2528,7 +2528,7 @@ _bt_unlink_halfdead_page(Relation rel, Buffer leafbuf, BlockNumber scanblkno,
if (leftsib == P_NONE && rightsib_is_rightmost)
{
page = BufferGetPage(rbuf);
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
if (P_RIGHTMOST(opaque))
{
/* rightsib will be the only one left on the level */
@@ -2565,12 +2565,12 @@ _bt_unlink_halfdead_page(Relation rel, Buffer leafbuf, BlockNumber scanblkno,
if (BufferIsValid(lbuf))
{
page = BufferGetPage(lbuf);
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
Assert(opaque->btpo_next == target);
opaque->btpo_next = rightsib;
}
page = BufferGetPage(rbuf);
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
Assert(opaque->btpo_prev == target);
opaque->btpo_prev = leftsib;
@@ -2599,7 +2599,7 @@ _bt_unlink_halfdead_page(Relation rel, Buffer leafbuf, BlockNumber scanblkno,
* of that scan.
*/
page = BufferGetPage(buf);
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
Assert(P_ISHALFDEAD(opaque) || !P_ISLEAF(opaque));
/*
@@ -2814,7 +2814,7 @@ _bt_lock_subtree_parent(Relation rel, BlockNumber child, BTStack stack,
parentoffset = stack->bts_offset;
page = BufferGetPage(pbuf);
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
maxoff = PageGetMaxOffsetNumber(page);
leftsibparent = opaque->btpo_prev;
diff --git a/src/backend/access/nbtree/nbtree.c b/src/backend/access/nbtree/nbtree.c
index c9b4964c1e..07600b86fb 100644
--- a/src/backend/access/nbtree/nbtree.c
+++ b/src/backend/access/nbtree/nbtree.c
@@ -1070,7 +1070,7 @@ backtrack:
if (!PageIsNew(page))
{
_bt_checkpage(rel, buf);
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
}
Assert(blkno <= scanblkno);
diff --git a/src/backend/access/nbtree/nbtsearch.c b/src/backend/access/nbtree/nbtsearch.c
index 9d82d4904d..c74543bfde 100644
--- a/src/backend/access/nbtree/nbtsearch.c
+++ b/src/backend/access/nbtree/nbtsearch.c
@@ -134,7 +134,7 @@ _bt_search(Relation rel, BTScanInsert key, Buffer *bufP, int access,
/* if this is a leaf page, we're done */
page = BufferGetPage(*bufP);
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
if (P_ISLEAF(opaque))
break;
@@ -268,7 +268,7 @@ _bt_moveright(Relation rel,
{
page = BufferGetPage(buf);
TestForOldSnapshot(snapshot, rel, page);
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
if (P_RIGHTMOST(opaque))
break;
@@ -347,7 +347,7 @@ _bt_binsrch(Relation rel,
cmpval;
page = BufferGetPage(buf);
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
/* Requesting nextkey semantics while using scantid seems nonsensical */
Assert(!key->nextkey || key->scantid == NULL);
@@ -451,7 +451,7 @@ _bt_binsrch_insert(Relation rel, BTInsertState insertstate)
cmpval;
page = BufferGetPage(insertstate->buf);
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
Assert(P_ISLEAF(opaque));
Assert(!key->nextkey);
@@ -659,7 +659,7 @@ _bt_compare(Relation rel,
OffsetNumber offnum)
{
TupleDesc itupdesc = RelationGetDescr(rel);
- BTPageOpaque opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ BTPageOpaque opaque = BTPageGetOpaque(page);
IndexTuple itup;
ItemPointer heapTid;
ScanKey scankey;
@@ -1536,7 +1536,7 @@ _bt_readpage(IndexScanDesc scan, ScanDirection dir, OffsetNumber offnum)
Assert(BufferIsValid(so->currPos.buf));
page = BufferGetPage(so->currPos.buf);
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
/* allow next page be processed by parallel worker */
if (scan->parallel_scan)
@@ -2007,7 +2007,7 @@ _bt_readnextpage(IndexScanDesc scan, BlockNumber blkno, ScanDirection dir)
so->currPos.buf = _bt_getbuf(rel, blkno, BT_READ);
page = BufferGetPage(so->currPos.buf);
TestForOldSnapshot(scan->xs_snapshot, rel, page);
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
/* check for deleted page */
if (!P_IGNORE(opaque))
{
@@ -2110,7 +2110,7 @@ _bt_readnextpage(IndexScanDesc scan, BlockNumber blkno, ScanDirection dir)
*/
page = BufferGetPage(so->currPos.buf);
TestForOldSnapshot(scan->xs_snapshot, rel, page);
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
if (!P_IGNORE(opaque))
{
PredicateLockPage(rel, BufferGetBlockNumber(so->currPos.buf), scan->xs_snapshot);
@@ -2191,7 +2191,7 @@ _bt_walk_left(Relation rel, Buffer buf, Snapshot snapshot)
BTPageOpaque opaque;
page = BufferGetPage(buf);
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
for (;;)
{
@@ -2216,7 +2216,7 @@ _bt_walk_left(Relation rel, Buffer buf, Snapshot snapshot)
buf = _bt_getbuf(rel, blkno, BT_READ);
page = BufferGetPage(buf);
TestForOldSnapshot(snapshot, rel, page);
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
/*
* If this isn't the page we want, walk right till we find what we
@@ -2243,14 +2243,14 @@ _bt_walk_left(Relation rel, Buffer buf, Snapshot snapshot)
buf = _bt_relandgetbuf(rel, buf, blkno, BT_READ);
page = BufferGetPage(buf);
TestForOldSnapshot(snapshot, rel, page);
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
}
/* Return to the original page to see what's up */
buf = _bt_relandgetbuf(rel, buf, obknum, BT_READ);
page = BufferGetPage(buf);
TestForOldSnapshot(snapshot, rel, page);
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
if (P_ISDELETED(opaque))
{
/*
@@ -2268,7 +2268,7 @@ _bt_walk_left(Relation rel, Buffer buf, Snapshot snapshot)
buf = _bt_relandgetbuf(rel, buf, blkno, BT_READ);
page = BufferGetPage(buf);
TestForOldSnapshot(snapshot, rel, page);
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
if (!P_ISDELETED(opaque))
break;
}
@@ -2329,7 +2329,7 @@ _bt_get_endpoint(Relation rel, uint32 level, bool rightmost,
page = BufferGetPage(buf);
TestForOldSnapshot(snapshot, rel, page);
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
for (;;)
{
@@ -2349,7 +2349,7 @@ _bt_get_endpoint(Relation rel, uint32 level, bool rightmost,
buf = _bt_relandgetbuf(rel, buf, blkno, BT_READ);
page = BufferGetPage(buf);
TestForOldSnapshot(snapshot, rel, page);
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
}
/* Done? */
@@ -2372,7 +2372,7 @@ _bt_get_endpoint(Relation rel, uint32 level, bool rightmost,
buf = _bt_relandgetbuf(rel, buf, blkno, BT_READ);
page = BufferGetPage(buf);
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
}
return buf;
@@ -2418,7 +2418,7 @@ _bt_endpoint(IndexScanDesc scan, ScanDirection dir)
PredicateLockPage(rel, BufferGetBlockNumber(buf), scan->xs_snapshot);
page = BufferGetPage(buf);
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
Assert(P_ISLEAF(opaque));
if (ScanDirectionIsForward(dir))
diff --git a/src/backend/access/nbtree/nbtsort.c b/src/backend/access/nbtree/nbtsort.c
index 8a19de2f66..c074513efa 100644
--- a/src/backend/access/nbtree/nbtsort.c
+++ b/src/backend/access/nbtree/nbtsort.c
@@ -625,7 +625,7 @@ _bt_blnewpage(uint32 level)
_bt_pageinit(page, BLCKSZ);
/* Initialize BT opaque state */
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
opaque->btpo_prev = opaque->btpo_next = P_NONE;
opaque->btpo_level = level;
opaque->btpo_flags = (level > 0) ? 0 : BTP_LEAF;
@@ -1000,9 +1000,9 @@ _bt_buildadd(BTWriteState *wstate, BTPageState *state, IndexTuple itup,
Assert((BTreeTupleGetNAtts(state->btps_lowkey, wstate->index) <=
IndexRelationGetNumberOfKeyAttributes(wstate->index) &&
BTreeTupleGetNAtts(state->btps_lowkey, wstate->index) > 0) ||
- P_LEFTMOST((BTPageOpaque) PageGetSpecialPointer(opage)));
+ P_LEFTMOST(BTPageGetOpaque(opage)));
Assert(BTreeTupleGetNAtts(state->btps_lowkey, wstate->index) == 0 ||
- !P_LEFTMOST((BTPageOpaque) PageGetSpecialPointer(opage)));
+ !P_LEFTMOST(BTPageGetOpaque(opage)));
BTreeTupleSetDownLink(state->btps_lowkey, oblkno);
_bt_buildadd(wstate, state->btps_next, state->btps_lowkey, 0);
pfree(state->btps_lowkey);
@@ -1017,8 +1017,8 @@ _bt_buildadd(BTWriteState *wstate, BTPageState *state, IndexTuple itup,
* Set the sibling links for both pages.
*/
{
- BTPageOpaque oopaque = (BTPageOpaque) PageGetSpecialPointer(opage);
- BTPageOpaque nopaque = (BTPageOpaque) PageGetSpecialPointer(npage);
+ BTPageOpaque oopaque = BTPageGetOpaque(opage);
+ BTPageOpaque nopaque = BTPageGetOpaque(npage);
oopaque->btpo_next = nblkno;
nopaque->btpo_prev = oblkno;
@@ -1125,7 +1125,7 @@ _bt_uppershutdown(BTWriteState *wstate, BTPageState *state)
BTPageOpaque opaque;
blkno = s->btps_blkno;
- opaque = (BTPageOpaque) PageGetSpecialPointer(s->btps_page);
+ opaque = BTPageGetOpaque(s->btps_page);
/*
* We have to link the last page on this level to somewhere.
diff --git a/src/backend/access/nbtree/nbtsplitloc.c b/src/backend/access/nbtree/nbtsplitloc.c
index c46594e1a2..ee01ceafda 100644
--- a/src/backend/access/nbtree/nbtsplitloc.c
+++ b/src/backend/access/nbtree/nbtsplitloc.c
@@ -152,7 +152,7 @@ _bt_findsplitloc(Relation rel,
SplitPoint leftpage,
rightpage;
- opaque = (BTPageOpaque) PageGetSpecialPointer(origpage);
+ opaque = BTPageGetOpaque(origpage);
maxoff = PageGetMaxOffsetNumber(origpage);
/* Total free space available on a btree page, after fixed overhead */
diff --git a/src/backend/access/nbtree/nbtutils.c b/src/backend/access/nbtree/nbtutils.c
index 84164748b3..96c72fc432 100644
--- a/src/backend/access/nbtree/nbtutils.c
+++ b/src/backend/access/nbtree/nbtutils.c
@@ -1774,7 +1774,7 @@ _bt_killitems(IndexScanDesc scan)
}
}
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
minoff = P_FIRSTDATAKEY(opaque);
maxoff = PageGetMaxOffsetNumber(page);
@@ -2474,7 +2474,7 @@ _bt_check_natts(Relation rel, bool heapkeyspace, Page page, OffsetNumber offnum)
{
int16 natts = IndexRelationGetNumberOfAttributes(rel);
int16 nkeyatts = IndexRelationGetNumberOfKeyAttributes(rel);
- BTPageOpaque opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ BTPageOpaque opaque = BTPageGetOpaque(page);
IndexTuple itup;
int tupnatts;
@@ -2662,7 +2662,7 @@ _bt_check_third_page(Relation rel, Relation heap, bool needheaptidspace,
* Internal page insertions cannot fail here, because that would mean that
* an earlier leaf level insertion that should have failed didn't
*/
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
if (!P_ISLEAF(opaque))
elog(ERROR, "cannot insert oversized tuple of size %zu on internal page of index \"%s\"",
itemsz, RelationGetRelationName(rel));
diff --git a/src/backend/access/nbtree/nbtxlog.c b/src/backend/access/nbtree/nbtxlog.c
index 611f412ba8..fba124b940 100644
--- a/src/backend/access/nbtree/nbtxlog.c
+++ b/src/backend/access/nbtree/nbtxlog.c
@@ -115,7 +115,7 @@ _bt_restore_meta(XLogReaderState *record, uint8 block_id)
md->btm_last_cleanup_num_heap_tuples = -1.0;
md->btm_allequalimage = xlrec->allequalimage;
- pageop = (BTPageOpaque) PageGetSpecialPointer(metapg);
+ pageop = BTPageGetOpaque(metapg);
pageop->btpo_flags = BTP_META;
/*
@@ -146,7 +146,7 @@ _bt_clear_incomplete_split(XLogReaderState *record, uint8 block_id)
if (XLogReadBufferForRedo(record, block_id, &buf) == BLK_NEEDS_REDO)
{
Page page = (Page) BufferGetPage(buf);
- BTPageOpaque pageop = (BTPageOpaque) PageGetSpecialPointer(page);
+ BTPageOpaque pageop = BTPageGetOpaque(page);
Assert(P_INCOMPLETE_SPLIT(pageop));
pageop->btpo_flags &= ~BTP_INCOMPLETE_SPLIT;
@@ -292,7 +292,7 @@ btree_xlog_split(bool newitemonleft, XLogReaderState *record)
rpage = (Page) BufferGetPage(rbuf);
_bt_pageinit(rpage, BufferGetPageSize(rbuf));
- ropaque = (BTPageOpaque) PageGetSpecialPointer(rpage);
+ ropaque = BTPageGetOpaque(rpage);
ropaque->btpo_prev = origpagenumber;
ropaque->btpo_next = spagenumber;
@@ -317,7 +317,7 @@ btree_xlog_split(bool newitemonleft, XLogReaderState *record)
* same for the right page.
*/
Page origpage = (Page) BufferGetPage(buf);
- BTPageOpaque oopaque = (BTPageOpaque) PageGetSpecialPointer(origpage);
+ BTPageOpaque oopaque = BTPageGetOpaque(origpage);
OffsetNumber off;
IndexTuple newitem = NULL,
left_hikey = NULL,
@@ -442,7 +442,7 @@ btree_xlog_split(bool newitemonleft, XLogReaderState *record)
if (XLogReadBufferForRedo(record, 2, &sbuf) == BLK_NEEDS_REDO)
{
Page spage = (Page) BufferGetPage(sbuf);
- BTPageOpaque spageop = (BTPageOpaque) PageGetSpecialPointer(spage);
+ BTPageOpaque spageop = BTPageGetOpaque(spage);
spageop->btpo_prev = rightpagenumber;
@@ -473,7 +473,7 @@ btree_xlog_dedup(XLogReaderState *record)
{
char *ptr = XLogRecGetBlockData(record, 0, NULL);
Page page = (Page) BufferGetPage(buf);
- BTPageOpaque opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ BTPageOpaque opaque = BTPageGetOpaque(page);
OffsetNumber offnum,
minoff,
maxoff;
@@ -541,7 +541,7 @@ btree_xlog_dedup(XLogReaderState *record)
if (P_HAS_GARBAGE(opaque))
{
- BTPageOpaque nopaque = (BTPageOpaque) PageGetSpecialPointer(newpage);
+ BTPageOpaque nopaque = BTPageGetOpaque(newpage);
nopaque->btpo_flags &= ~BTP_HAS_GARBAGE;
}
@@ -639,7 +639,7 @@ btree_xlog_vacuum(XLogReaderState *record)
* Mark the page as not containing any LP_DEAD items --- see comments
* in _bt_delitems_vacuum().
*/
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
opaque->btpo_flags &= ~BTP_HAS_GARBAGE;
PageSetLSN(page, lsn);
@@ -699,7 +699,7 @@ btree_xlog_delete(XLogReaderState *record)
PageIndexMultiDelete(page, (OffsetNumber *) ptr, xlrec->ndeleted);
/* Mark the page as not containing any LP_DEAD items */
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
opaque->btpo_flags &= ~BTP_HAS_GARBAGE;
PageSetLSN(page, lsn);
@@ -737,7 +737,7 @@ btree_xlog_mark_page_halfdead(uint8 info, XLogReaderState *record)
BlockNumber rightsib;
page = (Page) BufferGetPage(buffer);
- pageop = (BTPageOpaque) PageGetSpecialPointer(page);
+ pageop = BTPageGetOpaque(page);
poffset = xlrec->poffset;
@@ -768,7 +768,7 @@ btree_xlog_mark_page_halfdead(uint8 info, XLogReaderState *record)
page = (Page) BufferGetPage(buffer);
_bt_pageinit(page, BufferGetPageSize(buffer));
- pageop = (BTPageOpaque) PageGetSpecialPointer(page);
+ pageop = BTPageGetOpaque(page);
pageop->btpo_prev = xlrec->leftblk;
pageop->btpo_next = xlrec->rightblk;
@@ -833,7 +833,7 @@ btree_xlog_unlink_page(uint8 info, XLogReaderState *record)
if (XLogReadBufferForRedo(record, 1, &leftbuf) == BLK_NEEDS_REDO)
{
page = (Page) BufferGetPage(leftbuf);
- pageop = (BTPageOpaque) PageGetSpecialPointer(page);
+ pageop = BTPageGetOpaque(page);
pageop->btpo_next = rightsib;
PageSetLSN(page, lsn);
@@ -848,7 +848,7 @@ btree_xlog_unlink_page(uint8 info, XLogReaderState *record)
page = (Page) BufferGetPage(target);
_bt_pageinit(page, BufferGetPageSize(target));
- pageop = (BTPageOpaque) PageGetSpecialPointer(page);
+ pageop = BTPageGetOpaque(page);
pageop->btpo_prev = leftsib;
pageop->btpo_next = rightsib;
@@ -865,7 +865,7 @@ btree_xlog_unlink_page(uint8 info, XLogReaderState *record)
if (XLogReadBufferForRedo(record, 2, &rightbuf) == BLK_NEEDS_REDO)
{
page = (Page) BufferGetPage(rightbuf);
- pageop = (BTPageOpaque) PageGetSpecialPointer(page);
+ pageop = BTPageGetOpaque(page);
pageop->btpo_prev = leftsib;
PageSetLSN(page, lsn);
@@ -906,7 +906,7 @@ btree_xlog_unlink_page(uint8 info, XLogReaderState *record)
page = (Page) BufferGetPage(leafbuf);
_bt_pageinit(page, BufferGetPageSize(leafbuf));
- pageop = (BTPageOpaque) PageGetSpecialPointer(page);
+ pageop = BTPageGetOpaque(page);
pageop->btpo_flags = BTP_HALF_DEAD | BTP_LEAF;
pageop->btpo_prev = xlrec->leafleftsib;
@@ -948,7 +948,7 @@ btree_xlog_newroot(XLogReaderState *record)
page = (Page) BufferGetPage(buffer);
_bt_pageinit(page, BufferGetPageSize(buffer));
- pageop = (BTPageOpaque) PageGetSpecialPointer(page);
+ pageop = BTPageGetOpaque(page);
pageop->btpo_flags = BTP_ROOT;
pageop->btpo_prev = pageop->btpo_next = P_NONE;
@@ -1097,7 +1097,7 @@ btree_mask(char *pagedata, BlockNumber blkno)
mask_page_hint_bits(page);
mask_unused_space(page);
- maskopaq = (BTPageOpaque) PageGetSpecialPointer(page);
+ maskopaq = BTPageGetOpaque(page);
if (P_ISLEAF(maskopaq))
{
diff --git a/src/include/access/nbtree.h b/src/include/access/nbtree.h
index 9fec6fb1a8..31edddefc0 100644
--- a/src/include/access/nbtree.h
+++ b/src/include/access/nbtree.h
@@ -70,6 +70,13 @@ typedef struct BTPageOpaqueData
typedef BTPageOpaqueData *BTPageOpaque;
+/*
+ * Pre-aligned accessor to the opaque data in the page's special region.
+ * Avoids data dependency on page header in non-assert builds.
+ */
+#define BTPageGetOpaque(page) \
+ ((BTPageOpaque) PageGetSpecialOpaque((page), BTPageOpaqueData))
+
/* Bits defined in btpo_flags */
#define BTP_LEAF (1 << 0) /* leaf page, i.e. not internal page */
#define BTP_ROOT (1 << 1) /* root page (has no parent) */
@@ -241,7 +248,7 @@ BTPageSetDeleted(Page page, FullTransactionId safexid)
PageHeader header;
BTDeletedPageData *contents;
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
header = ((PageHeader) page);
opaque->btpo_flags &= ~BTP_HALF_DEAD;
@@ -263,7 +270,7 @@ BTPageGetDeleteXid(Page page)
/* We only expect to be called with a deleted page */
Assert(!PageIsNew(page));
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
Assert(P_ISDELETED(opaque));
/* pg_upgrade'd deleted page -- must be safe to delete now */
@@ -294,7 +301,7 @@ BTPageIsRecyclable(Page page)
Assert(!PageIsNew(page));
/* Recycling okay iff page is deleted and safexid is old enough */
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
if (P_ISDELETED(opaque))
{
/*
--
2.30.2
Re: Preventing indirection for IndexPageGetOpaque for known-size page special areas
On Wed, Feb 16, 2022 at 10:24:58PM +0100, Matthias van de Meent wrote:
A first good reason to do this is preventing further damage when a
page is corrupted: if I can somehow overwrite pd_special,
non-assert-enabled builds would start reading and writing at arbitrary
offsets from the page pointer, quite possibly in subsequent buffers
(or worse, on the stack, in case of stack-allocated blocks).
Well, pageinspect has proved to be rather careless in this area for a
couple of years, but this just came from the fact that we called
PageGetSpecialPointer() before checking that PageGetSpecialSize() maps
with the MAXALIGN'd size of the opaque area, if the related AM has
one.
I am not sure that we need to worry about corruptions, as data
checksums would offer protection for most cases users usually need to
worry about, at the exception of page corruptions because of memory
for pages already read in the PG shared buffers from disk or even the
OS cache.
A second reason would be less indirection to get to the opaque
pointer. This should improve performance a bit in those cases where we
(initially) only want to access the [Index]PageOpaque struct.
We don't have many cases that do that, do we?
Lastly, 0002 and 0003 remove some repetitive tasks of dealing with the
[nbtree, hash] opaques by providing a typed accessor macro similar to
what is used in the GIN and (SP-)GIST index methods; improving the
legibility of the code and decreasing the churn.
+#define PageGetSpecialOpaque(page, OpaqueDataTyp) \
+( \
+ AssertMacro(PageGetPageSize(page) == BLCKSZ && \
+ PageGetSpecialSize(page) == MAXALIGN(sizeof(OpaqueDataTyp))), \
+ (OpaqueDataTyp *) ((char *) (page) + (BLCKSZ - MAXALIGN(sizeof(OpaqueDataTyp)))) \
+)Did you notice PageValidateSpecialPointer() that mentions MSVC? Could
this stuff a problem if not an inline function.
Perhaps you should do a s/OpaqueDataTyp/OpaqueDataType/.
As a whole, this patch set looks like an improvement in terms of
checks and consistency regarding the special area handling across the
different AMs for the backend code, while reducing the MAXALIGN-ism on
all those sizes, and this tends to be missed. Any other opinions?
--
Michael
Re: Preventing indirection for IndexPageGetOpaque for known-size page special areas
On Mon, 28 Mar 2022 at 06:33, Michael Paquier <michael@paquier.xyz> wrote:
On Wed, Feb 16, 2022 at 10:24:58PM +0100, Matthias van de Meent wrote:
A first good reason to do this is preventing further damage when a
page is corrupted: if I can somehow overwrite pd_special,
non-assert-enabled builds would start reading and writing at arbitrary
offsets from the page pointer, quite possibly in subsequent buffers
(or worse, on the stack, in case of stack-allocated blocks).Well, pageinspect has proved to be rather careless in this area for a
couple of years, but this just came from the fact that we called
PageGetSpecialPointer() before checking that PageGetSpecialSize() maps
with the MAXALIGN'd size of the opaque area, if the related AM has
one.I am not sure that we need to worry about corruptions, as data
checksums would offer protection for most cases users usually need to
worry about, at the exception of page corruptions because of memory
for pages already read in the PG shared buffers from disk or even the
OS cache.
Not all clusters have checksums enabled (boo!, but we can't
realistically fix that), so on-disk corruption could reasonably
propagate to the rest of such system. Additionally, checksums are only
checked on read, and updated when the page is written to disk (in
PageSetChecksumCopy called from FlushBuffer), but this does not check
for signs of page corruption. As such, a memory bug resulting in
in-memory corruption of pd_special would persist and would currently
have the potential to further corrupt neighbouring buffers.
A second reason would be less indirection to get to the opaque
pointer. This should improve performance a bit in those cases where we
(initially) only want to access the [Index]PageOpaque struct.We don't have many cases that do that, do we?
Indeed not many, but I know that at least the nbtree-code has some
cases where it uses the opaque before other fields in the header are
accessed (sometimes even without accessing the header for other
reasons): in _bt_moveright and _bt_walk_left. There might be more; but
these are cases I know of.
Lastly, 0002 and 0003 remove some repetitive tasks of dealing with the
[nbtree, hash] opaques by providing a typed accessor macro similar to
what is used in the GIN and (SP-)GIST index methods; improving the
legibility of the code and decreasing the churn.+#define PageGetSpecialOpaque(page, OpaqueDataTyp) \ +( \ + AssertMacro(PageGetPageSize(page) == BLCKSZ && \ + PageGetSpecialSize(page) == MAXALIGN(sizeof(OpaqueDataTyp))), \ + (OpaqueDataTyp *) ((char *) (page) + (BLCKSZ - MAXALIGN(sizeof(OpaqueDataTyp)))) \ +)Did you notice PageValidateSpecialPointer() that mentions MSVC? Could
this stuff a problem if not an inline function.
I'm not sure; but the CFbot build (using whatever MSVC is used with
latest Window Server 2019, VS 19) failed to complain in this case.
Furthermore, the case mentioned in the comment of
PageValidateSpecialPointer() refers to problems that only happened
after the macro was updated from one MacroAssert to three seperate
MacroAssert()s; this new code currently only has one, so we should be
fine for now.
Perhaps you should do a s/OpaqueDataTyp/OpaqueDataType/.
This has been fixed in attached v2; and apart from tweaks in the
commit messages there have been no other changes.
As a whole, this patch set looks like an improvement in terms of
checks and consistency regarding the special area handling across the
different AMs for the backend code, while reducing the MAXALIGN-ism on
all those sizes, and this tends to be missed.
Thanks!
-Matthias
Attachments:
v2-0001-Add-known-size-pre-aligned-special-area-pointer-m.patchapplication/x-patch; name=v2-0001-Add-known-size-pre-aligned-special-area-pointer-m.patchDownload
From c22128e0e63ab83f9009a1466ea50c46878d7650 Mon Sep 17 00:00:00 2001
From: Matthias van de Meent <boekewurm+postgres@gmail.com>
Date: Mon, 28 Mar 2022 14:53:43 +0200
Subject: [PATCH v2 1/3] Add known-size pre-aligned special area pointer macro
This removes 1 layer of indirection for special areas of which we know the
type (=size) and location.
Special area access through the page header is an extra cache line that needs
to be fetched. If might only want to look at the special area, it is much
less effort to calculate the offset to the special area directly instead of
first checking the page header - saving one cache line to fetch.
Assertions are added to check that the page has a correctly sized special
area, and that the page is of the expected size. This detects data corruption,
instead of doing random reads/writes into the page or data allocated next to
the page being accessed.
Additionally, updates the GIN, GIST and SP-GIST [Index]PageGetOpaque macros
with the new pre-aligned special area accessor.
---
src/include/access/ginblock.h | 3 ++-
src/include/access/gist.h | 3 ++-
src/include/access/spgist_private.h | 3 ++-
src/include/storage/bufpage.h | 14 ++++++++++++++
4 files changed, 20 insertions(+), 3 deletions(-)
diff --git a/src/include/access/ginblock.h b/src/include/access/ginblock.h
index 9347f464f3..3680098c98 100644
--- a/src/include/access/ginblock.h
+++ b/src/include/access/ginblock.h
@@ -108,7 +108,8 @@ typedef struct GinMetaPageData
/*
* Macros for accessing a GIN index page's opaque data
*/
-#define GinPageGetOpaque(page) ( (GinPageOpaque) PageGetSpecialPointer(page) )
+#define GinPageGetOpaque(page) \
+ ((GinPageOpaque) PageGetSpecialOpaque(page, GinPageOpaqueData))
#define GinPageIsLeaf(page) ( (GinPageGetOpaque(page)->flags & GIN_LEAF) != 0 )
#define GinPageSetLeaf(page) ( GinPageGetOpaque(page)->flags |= GIN_LEAF )
diff --git a/src/include/access/gist.h b/src/include/access/gist.h
index a3337627b8..51223e9051 100644
--- a/src/include/access/gist.h
+++ b/src/include/access/gist.h
@@ -164,7 +164,8 @@ typedef struct GISTENTRY
bool leafkey;
} GISTENTRY;
-#define GistPageGetOpaque(page) ( (GISTPageOpaque) PageGetSpecialPointer(page) )
+#define GistPageGetOpaque(page) \
+ ((GISTPageOpaque) PageGetSpecialOpaque(page, GISTPageOpaqueData))
#define GistPageIsLeaf(page) ( GistPageGetOpaque(page)->flags & F_LEAF)
#define GIST_LEAF(entry) (GistPageIsLeaf((entry)->page))
diff --git a/src/include/access/spgist_private.h b/src/include/access/spgist_private.h
index eb56b1c6b8..b0bdc7df3e 100644
--- a/src/include/access/spgist_private.h
+++ b/src/include/access/spgist_private.h
@@ -75,7 +75,8 @@ typedef SpGistPageOpaqueData *SpGistPageOpaque;
#define SPGIST_LEAF (1<<2)
#define SPGIST_NULLS (1<<3)
-#define SpGistPageGetOpaque(page) ((SpGistPageOpaque) PageGetSpecialPointer(page))
+#define SpGistPageGetOpaque(page) \
+ ((SpGistPageOpaque) PageGetSpecialOpaque(page, SpGistPageOpaqueData))
#define SpGistPageIsMeta(page) (SpGistPageGetOpaque(page)->flags & SPGIST_META)
#define SpGistPageIsDeleted(page) (SpGistPageGetOpaque(page)->flags & SPGIST_DELETED)
#define SpGistPageIsLeaf(page) (SpGistPageGetOpaque(page)->flags & SPGIST_LEAF)
diff --git a/src/include/storage/bufpage.h b/src/include/storage/bufpage.h
index e9f253f2c8..e96595f338 100644
--- a/src/include/storage/bufpage.h
+++ b/src/include/storage/bufpage.h
@@ -329,6 +329,20 @@ PageValidateSpecialPointer(Page page)
(char *) ((char *) (page) + ((PageHeader) (page))->pd_special) \
)
+/*
+ * PageGetSpecialOpaque
+ * Returns pointer to special space on a page, of the given type.
+ * This removes the data dependency on the page header in builds
+ * without asserts enabled.
+ */
+#define PageGetSpecialOpaque(page, OpaqueDataType) \
+( \
+ AssertMacro(PageGetPageSize(page) == BLCKSZ && \
+ PageGetSpecialSize(page) == MAXALIGN(sizeof(OpaqueDataType))), \
+ (OpaqueDataType *) ((char *) (page) + \
+ (BLCKSZ - MAXALIGN(sizeof(OpaqueDataType)))) \
+)
+
/*
* PageGetItem
* Retrieves an item on the given page.
--
2.30.2
v2-0002-Update-nbtree-code-to-use-PageGetSpecialOpaque-re.patchapplication/x-patch; name=v2-0002-Update-nbtree-code-to-use-PageGetSpecialOpaque-re.patchDownload
From 48db6037b088bf21dc2fd11b13d1d565c8fd9789 Mon Sep 17 00:00:00 2001
From: Matthias van de Meent <boekewurm+postgres@gmail.com>
Date: Mon, 28 Mar 2022 16:44:57 +0200
Subject: [PATCH v2 2/3] Update nbtree code to use PageGetSpecialOpaque;
replacing PageGetSpecialPointer+cast
As in the earlier patch, this reduces memory access requirements by directly
reading into the target space instead of needing to access the header of the
page first.
Also makes the code a lot more readable by replacing templated macro+casts
with a single macro specialized for BTree page opaques.
---
contrib/amcheck/verify_nbtree.c | 32 +++++++-------
contrib/pageinspect/btreefuncs.c | 6 +--
contrib/pgstattuple/pgstatindex.c | 2 +-
contrib/pgstattuple/pgstattuple.c | 2 +-
src/backend/access/nbtree/nbtdedup.c | 6 +--
src/backend/access/nbtree/nbtinsert.c | 46 ++++++++++----------
src/backend/access/nbtree/nbtpage.c | 58 ++++++++++++-------------
src/backend/access/nbtree/nbtree.c | 2 +-
src/backend/access/nbtree/nbtsearch.c | 34 +++++++--------
src/backend/access/nbtree/nbtsort.c | 12 ++---
src/backend/access/nbtree/nbtsplitloc.c | 2 +-
src/backend/access/nbtree/nbtutils.c | 6 +--
src/backend/access/nbtree/nbtxlog.c | 34 +++++++--------
src/include/access/nbtree.h | 13 ++++--
14 files changed, 131 insertions(+), 124 deletions(-)
diff --git a/contrib/amcheck/verify_nbtree.c b/contrib/amcheck/verify_nbtree.c
index d2510ee648..70278c4f93 100644
--- a/contrib/amcheck/verify_nbtree.c
+++ b/contrib/amcheck/verify_nbtree.c
@@ -691,7 +691,7 @@ bt_check_level_from_leftmost(BtreeCheckState *state, BtreeLevel level)
state->target = palloc_btree_page(state, state->targetblock);
state->targetlsn = PageGetLSN(state->target);
- opaque = (BTPageOpaque) PageGetSpecialPointer(state->target);
+ opaque = BTPageGetOpaque(state->target);
if (P_IGNORE(opaque))
{
@@ -927,7 +927,7 @@ bt_recheck_sibling_links(BtreeCheckState *state,
LockBuffer(lbuf, BT_READ);
_bt_checkpage(state->rel, lbuf);
page = BufferGetPage(lbuf);
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
if (P_ISDELETED(opaque))
{
/*
@@ -951,7 +951,7 @@ bt_recheck_sibling_links(BtreeCheckState *state,
LockBuffer(newtargetbuf, BT_READ);
_bt_checkpage(state->rel, newtargetbuf);
page = BufferGetPage(newtargetbuf);
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
/* btpo_prev_from_target may have changed; update it */
btpo_prev_from_target = opaque->btpo_prev;
}
@@ -1049,7 +1049,7 @@ bt_target_page_check(BtreeCheckState *state)
OffsetNumber max;
BTPageOpaque topaque;
- topaque = (BTPageOpaque) PageGetSpecialPointer(state->target);
+ topaque = BTPageGetOpaque(state->target);
max = PageGetMaxOffsetNumber(state->target);
elog(DEBUG2, "verifying %u items on %s block %u", max,
@@ -1478,7 +1478,7 @@ bt_target_page_check(BtreeCheckState *state)
/* Get fresh copy of target page */
state->target = palloc_btree_page(state, state->targetblock);
/* Note that we deliberately do not update target LSN */
- topaque = (BTPageOpaque) PageGetSpecialPointer(state->target);
+ topaque = BTPageGetOpaque(state->target);
/*
* All !readonly checks now performed; just return
@@ -1552,7 +1552,7 @@ bt_right_page_check_scankey(BtreeCheckState *state)
OffsetNumber nline;
/* Determine target's next block number */
- opaque = (BTPageOpaque) PageGetSpecialPointer(state->target);
+ opaque = BTPageGetOpaque(state->target);
/* If target is already rightmost, no right sibling; nothing to do here */
if (P_RIGHTMOST(opaque))
@@ -1588,7 +1588,7 @@ bt_right_page_check_scankey(BtreeCheckState *state)
CHECK_FOR_INTERRUPTS();
rightpage = palloc_btree_page(state, targetnext);
- opaque = (BTPageOpaque) PageGetSpecialPointer(rightpage);
+ opaque = BTPageGetOpaque(rightpage);
if (!P_IGNORE(opaque) || P_RIGHTMOST(opaque))
break;
@@ -1893,7 +1893,7 @@ bt_child_highkey_check(BtreeCheckState *state,
else
page = palloc_btree_page(state, blkno);
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
/* The first page we visit at the level should be leftmost */
if (first && !BlockNumberIsValid(state->prevrightlink) && !P_LEFTMOST(opaque))
@@ -1971,7 +1971,7 @@ bt_child_highkey_check(BtreeCheckState *state,
else
pivotkey_offset = target_downlinkoffnum;
- topaque = (BTPageOpaque) PageGetSpecialPointer(state->target);
+ topaque = BTPageGetOpaque(state->target);
if (!offset_is_negative_infinity(topaque, pivotkey_offset))
{
@@ -2128,9 +2128,9 @@ bt_child_check(BtreeCheckState *state, BTScanInsert targetkey,
* Check all items, rather than checking just the first and trusting that
* the operator class obeys the transitive law.
*/
- topaque = (BTPageOpaque) PageGetSpecialPointer(state->target);
+ topaque = BTPageGetOpaque(state->target);
child = palloc_btree_page(state, childblock);
- copaque = (BTPageOpaque) PageGetSpecialPointer(child);
+ copaque = BTPageGetOpaque(child);
maxoffset = PageGetMaxOffsetNumber(child);
/*
@@ -2235,7 +2235,7 @@ static void
bt_downlink_missing_check(BtreeCheckState *state, bool rightsplit,
BlockNumber blkno, Page page)
{
- BTPageOpaque opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ BTPageOpaque opaque = BTPageGetOpaque(page);
ItemId itemid;
IndexTuple itup;
Page child;
@@ -2319,7 +2319,7 @@ bt_downlink_missing_check(BtreeCheckState *state, bool rightsplit,
CHECK_FOR_INTERRUPTS();
child = palloc_btree_page(state, childblk);
- copaque = (BTPageOpaque) PageGetSpecialPointer(child);
+ copaque = BTPageGetOpaque(child);
if (P_ISLEAF(copaque))
break;
@@ -2780,7 +2780,7 @@ invariant_l_offset(BtreeCheckState *state, BTScanInsert key,
bool nonpivot;
ritup = (IndexTuple) PageGetItem(state->target, itemid);
- topaque = (BTPageOpaque) PageGetSpecialPointer(state->target);
+ topaque = BTPageGetOpaque(state->target);
nonpivot = P_ISLEAF(topaque) && upperbound >= P_FIRSTDATAKEY(topaque);
/* Get number of keys + heap TID for item to the right */
@@ -2895,7 +2895,7 @@ invariant_l_nontarget_offset(BtreeCheckState *state, BTScanInsert key,
bool nonpivot;
child = (IndexTuple) PageGetItem(nontarget, itemid);
- copaque = (BTPageOpaque) PageGetSpecialPointer(nontarget);
+ copaque = BTPageGetOpaque(nontarget);
nonpivot = P_ISLEAF(copaque) && upperbound >= P_FIRSTDATAKEY(copaque);
/* Get number of keys + heap TID for child/non-target item */
@@ -2954,7 +2954,7 @@ palloc_btree_page(BtreeCheckState *state, BlockNumber blocknum)
memcpy(page, BufferGetPage(buffer), BLCKSZ);
UnlockReleaseBuffer(buffer);
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
if (P_ISMETA(opaque) && blocknum != BTREE_METAPAGE)
ereport(ERROR,
diff --git a/contrib/pageinspect/btreefuncs.c b/contrib/pageinspect/btreefuncs.c
index 7651c59bbf..3daa31c84d 100644
--- a/contrib/pageinspect/btreefuncs.c
+++ b/contrib/pageinspect/btreefuncs.c
@@ -93,7 +93,7 @@ GetBTPageStatistics(BlockNumber blkno, Buffer buffer, BTPageStat *stat)
Page page = BufferGetPage(buffer);
PageHeader phdr = (PageHeader) page;
OffsetNumber maxoff = PageGetMaxOffsetNumber(page);
- BTPageOpaque opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ BTPageOpaque opaque = BTPageGetOpaque(page);
int item_size = 0;
int off;
@@ -525,7 +525,7 @@ bt_page_items_internal(PG_FUNCTION_ARGS, enum pageinspect_version ext_version)
uargs->offset = FirstOffsetNumber;
- opaque = (BTPageOpaque) PageGetSpecialPointer(uargs->page);
+ opaque = BTPageGetOpaque(uargs->page);
if (!P_ISDELETED(opaque))
fctx->max_calls = PageGetMaxOffsetNumber(uargs->page);
@@ -622,7 +622,7 @@ bt_page_items_bytea(PG_FUNCTION_ARGS)
(int) MAXALIGN(sizeof(BTPageOpaqueData)),
(int) PageGetSpecialSize(uargs->page))));
- opaque = (BTPageOpaque) PageGetSpecialPointer(uargs->page);
+ opaque = BTPageGetOpaque(uargs->page);
if (P_ISMETA(opaque))
ereport(ERROR,
diff --git a/contrib/pgstattuple/pgstatindex.c b/contrib/pgstattuple/pgstatindex.c
index 6c4b053dd0..b9939451b4 100644
--- a/contrib/pgstattuple/pgstatindex.c
+++ b/contrib/pgstattuple/pgstatindex.c
@@ -281,7 +281,7 @@ pgstatindex_impl(Relation rel, FunctionCallInfo fcinfo)
LockBuffer(buffer, BUFFER_LOCK_SHARE);
page = BufferGetPage(buffer);
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
/*
* Determine page type, and update totals.
diff --git a/contrib/pgstattuple/pgstattuple.c b/contrib/pgstattuple/pgstattuple.c
index c9b8f01f9b..a9658d389b 100644
--- a/contrib/pgstattuple/pgstattuple.c
+++ b/contrib/pgstattuple/pgstattuple.c
@@ -415,7 +415,7 @@ pgstat_btree_page(pgstattuple_type *stat, Relation rel, BlockNumber blkno,
{
BTPageOpaque opaque;
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
if (P_IGNORE(opaque))
{
/* deleted or half-dead page */
diff --git a/src/backend/access/nbtree/nbtdedup.c b/src/backend/access/nbtree/nbtdedup.c
index 4c48554aec..3e11805293 100644
--- a/src/backend/access/nbtree/nbtdedup.c
+++ b/src/backend/access/nbtree/nbtdedup.c
@@ -62,7 +62,7 @@ _bt_dedup_pass(Relation rel, Buffer buf, Relation heapRel, IndexTuple newitem,
minoff,
maxoff;
Page page = BufferGetPage(buf);
- BTPageOpaque opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ BTPageOpaque opaque = BTPageGetOpaque(page);
Page newpage;
BTDedupState state;
Size pagesaving = 0;
@@ -231,7 +231,7 @@ _bt_dedup_pass(Relation rel, Buffer buf, Relation heapRel, IndexTuple newitem,
*/
if (P_HAS_GARBAGE(opaque))
{
- BTPageOpaque nopaque = (BTPageOpaque) PageGetSpecialPointer(newpage);
+ BTPageOpaque nopaque = BTPageGetOpaque(newpage);
nopaque->btpo_flags &= ~BTP_HAS_GARBAGE;
}
@@ -310,7 +310,7 @@ _bt_bottomupdel_pass(Relation rel, Buffer buf, Relation heapRel,
minoff,
maxoff;
Page page = BufferGetPage(buf);
- BTPageOpaque opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ BTPageOpaque opaque = BTPageGetOpaque(page);
BTDedupState state;
TM_IndexDeleteOp delstate;
bool neverdedup;
diff --git a/src/backend/access/nbtree/nbtinsert.c b/src/backend/access/nbtree/nbtinsert.c
index 68628ec000..f6f4af8bfe 100644
--- a/src/backend/access/nbtree/nbtinsert.c
+++ b/src/backend/access/nbtree/nbtinsert.c
@@ -329,7 +329,7 @@ _bt_search_insert(Relation rel, BTInsertState insertstate)
_bt_checkpage(rel, insertstate->buf);
page = BufferGetPage(insertstate->buf);
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
/*
* Check if the page is still the rightmost leaf page and has
@@ -428,7 +428,7 @@ _bt_check_unique(Relation rel, BTInsertState insertstate, Relation heapRel,
InitDirtySnapshot(SnapshotDirty);
page = BufferGetPage(insertstate->buf);
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
maxoff = PageGetMaxOffsetNumber(page);
/*
@@ -733,7 +733,7 @@ _bt_check_unique(Relation rel, BTInsertState insertstate, Relation heapRel,
nbuf = _bt_relandgetbuf(rel, nbuf, nblkno, BT_READ);
page = BufferGetPage(nbuf);
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
if (!P_IGNORE(opaque))
break;
if (P_RIGHTMOST(opaque))
@@ -822,7 +822,7 @@ _bt_findinsertloc(Relation rel,
BTPageOpaque opaque;
OffsetNumber newitemoff;
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
/* Check 1/3 of a page restriction */
if (unlikely(insertstate->itemsz > BTMaxItemSize(page)))
@@ -888,7 +888,7 @@ _bt_findinsertloc(Relation rel,
_bt_stepright(rel, insertstate, stack);
/* Update local state after stepping right */
page = BufferGetPage(insertstate->buf);
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
/* Assume duplicates (if checkingunique) */
uniquedup = true;
}
@@ -972,7 +972,7 @@ _bt_findinsertloc(Relation rel,
_bt_stepright(rel, insertstate, stack);
/* Update local state after stepping right */
page = BufferGetPage(insertstate->buf);
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
}
}
@@ -1030,7 +1030,7 @@ _bt_stepright(Relation rel, BTInsertState insertstate, BTStack stack)
BlockNumber rblkno;
page = BufferGetPage(insertstate->buf);
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
rbuf = InvalidBuffer;
rblkno = opaque->btpo_next;
@@ -1038,7 +1038,7 @@ _bt_stepright(Relation rel, BTInsertState insertstate, BTStack stack)
{
rbuf = _bt_relandgetbuf(rel, rbuf, rblkno, BT_WRITE);
page = BufferGetPage(rbuf);
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
/*
* If this page was incompletely split, finish the split now. We do
@@ -1120,7 +1120,7 @@ _bt_insertonpg(Relation rel,
IndexTuple nposting = NULL;
page = BufferGetPage(buf);
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
isleaf = P_ISLEAF(opaque);
isroot = P_ISROOT(opaque);
isrightmost = P_RIGHTMOST(opaque);
@@ -1296,7 +1296,7 @@ _bt_insertonpg(Relation rel,
if (!isleaf)
{
Page cpage = BufferGetPage(cbuf);
- BTPageOpaque cpageop = (BTPageOpaque) PageGetSpecialPointer(cpage);
+ BTPageOpaque cpageop = BTPageGetOpaque(cpage);
Assert(P_INCOMPLETE_SPLIT(cpageop));
cpageop->btpo_flags &= ~BTP_INCOMPLETE_SPLIT;
@@ -1504,7 +1504,7 @@ _bt_split(Relation rel, BTScanInsert itup_key, Buffer buf, Buffer cbuf,
* only workspace.
*/
origpage = BufferGetPage(buf);
- oopaque = (BTPageOpaque) PageGetSpecialPointer(origpage);
+ oopaque = BTPageGetOpaque(origpage);
isleaf = P_ISLEAF(oopaque);
isrightmost = P_RIGHTMOST(oopaque);
maxoff = PageGetMaxOffsetNumber(origpage);
@@ -1540,7 +1540,7 @@ _bt_split(Relation rel, BTScanInsert itup_key, Buffer buf, Buffer cbuf,
/* Allocate temp buffer for leftpage */
leftpage = PageGetTempPage(origpage);
_bt_pageinit(leftpage, BufferGetPageSize(buf));
- lopaque = (BTPageOpaque) PageGetSpecialPointer(leftpage);
+ lopaque = BTPageGetOpaque(leftpage);
/*
* leftpage won't be the root when we're done. Also, clear the SPLIT_END
@@ -1716,7 +1716,7 @@ _bt_split(Relation rel, BTScanInsert itup_key, Buffer buf, Buffer cbuf,
rightpage = BufferGetPage(rbuf);
rightpagenumber = BufferGetBlockNumber(rbuf);
/* rightpage was initialized by _bt_getbuf */
- ropaque = (BTPageOpaque) PageGetSpecialPointer(rightpage);
+ ropaque = BTPageGetOpaque(rightpage);
/*
* Finish off remaining leftpage special area fields. They cannot be set
@@ -1887,7 +1887,7 @@ _bt_split(Relation rel, BTScanInsert itup_key, Buffer buf, Buffer cbuf,
{
sbuf = _bt_getbuf(rel, oopaque->btpo_next, BT_WRITE);
spage = BufferGetPage(sbuf);
- sopaque = (BTPageOpaque) PageGetSpecialPointer(spage);
+ sopaque = BTPageGetOpaque(spage);
if (sopaque->btpo_prev != origpagenumber)
{
memset(rightpage, 0, BufferGetPageSize(rbuf));
@@ -1952,7 +1952,7 @@ _bt_split(Relation rel, BTScanInsert itup_key, Buffer buf, Buffer cbuf,
if (!isleaf)
{
Page cpage = BufferGetPage(cbuf);
- BTPageOpaque cpageop = (BTPageOpaque) PageGetSpecialPointer(cpage);
+ BTPageOpaque cpageop = BTPageGetOpaque(cpage);
cpageop->btpo_flags &= ~BTP_INCOMPLETE_SPLIT;
MarkBufferDirty(cbuf);
@@ -2139,7 +2139,7 @@ _bt_insert_parent(Relation rel,
BTPageOpaque opaque;
elog(DEBUG2, "concurrent ROOT page split");
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
/*
* We should never reach here when a leaf page split takes place
@@ -2230,7 +2230,7 @@ void
_bt_finish_split(Relation rel, Buffer lbuf, BTStack stack)
{
Page lpage = BufferGetPage(lbuf);
- BTPageOpaque lpageop = (BTPageOpaque) PageGetSpecialPointer(lpage);
+ BTPageOpaque lpageop = BTPageGetOpaque(lpage);
Buffer rbuf;
Page rpage;
BTPageOpaque rpageop;
@@ -2242,7 +2242,7 @@ _bt_finish_split(Relation rel, Buffer lbuf, BTStack stack)
/* Lock right sibling, the one missing the downlink */
rbuf = _bt_getbuf(rel, lpageop->btpo_next, BT_WRITE);
rpage = BufferGetPage(rbuf);
- rpageop = (BTPageOpaque) PageGetSpecialPointer(rpage);
+ rpageop = BTPageGetOpaque(rpage);
/* Could this be a root split? */
if (!stack)
@@ -2320,7 +2320,7 @@ _bt_getstackbuf(Relation rel, BTStack stack, BlockNumber child)
buf = _bt_getbuf(rel, blkno, BT_WRITE);
page = BufferGetPage(buf);
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
if (P_INCOMPLETE_SPLIT(opaque))
{
@@ -2451,7 +2451,7 @@ _bt_newroot(Relation rel, Buffer lbuf, Buffer rbuf)
lbkno = BufferGetBlockNumber(lbuf);
rbkno = BufferGetBlockNumber(rbuf);
lpage = BufferGetPage(lbuf);
- lopaque = (BTPageOpaque) PageGetSpecialPointer(lpage);
+ lopaque = BTPageGetOpaque(lpage);
/* get a new root page */
rootbuf = _bt_getbuf(rel, P_NEW, BT_WRITE);
@@ -2492,11 +2492,11 @@ _bt_newroot(Relation rel, Buffer lbuf, Buffer rbuf)
_bt_upgrademetapage(metapg);
/* set btree special data */
- rootopaque = (BTPageOpaque) PageGetSpecialPointer(rootpage);
+ rootopaque = BTPageGetOpaque(rootpage);
rootopaque->btpo_prev = rootopaque->btpo_next = P_NONE;
rootopaque->btpo_flags = BTP_ROOT;
rootopaque->btpo_level =
- ((BTPageOpaque) PageGetSpecialPointer(lpage))->btpo_level + 1;
+ (BTPageGetOpaque(lpage))->btpo_level + 1;
rootopaque->btpo_cycleid = 0;
/* update metapage data */
@@ -2680,7 +2680,7 @@ _bt_delete_or_dedup_one_page(Relation rel, Relation heapRel,
Buffer buffer = insertstate->buf;
BTScanInsert itup_key = insertstate->itup_key;
Page page = BufferGetPage(buffer);
- BTPageOpaque opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ BTPageOpaque opaque = BTPageGetOpaque(page);
Assert(P_ISLEAF(opaque));
Assert(simpleonly || itup_key->heapkeyspace);
diff --git a/src/backend/access/nbtree/nbtpage.c b/src/backend/access/nbtree/nbtpage.c
index 6b5f01e1d0..20adb602a4 100644
--- a/src/backend/access/nbtree/nbtpage.c
+++ b/src/backend/access/nbtree/nbtpage.c
@@ -85,7 +85,7 @@ _bt_initmetapage(Page page, BlockNumber rootbknum, uint32 level,
metad->btm_last_cleanup_num_heap_tuples = -1.0;
metad->btm_allequalimage = allequalimage;
- metaopaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ metaopaque = BTPageGetOpaque(page);
metaopaque->btpo_flags = BTP_META;
/*
@@ -112,7 +112,7 @@ _bt_upgrademetapage(Page page)
BTPageOpaque metaopaque PG_USED_FOR_ASSERTS_ONLY;
metad = BTPageGetMeta(page);
- metaopaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ metaopaque = BTPageGetOpaque(page);
/* It must be really a meta page of upgradable version */
Assert(metaopaque->btpo_flags & BTP_META);
@@ -148,7 +148,7 @@ _bt_getmeta(Relation rel, Buffer metabuf)
BTMetaPageData *metad;
metapg = BufferGetPage(metabuf);
- metaopaque = (BTPageOpaque) PageGetSpecialPointer(metapg);
+ metaopaque = BTPageGetOpaque(metapg);
metad = BTPageGetMeta(metapg);
/* sanity-check the metapage */
@@ -372,7 +372,7 @@ _bt_getroot(Relation rel, int access)
rootbuf = _bt_getbuf(rel, rootblkno, BT_READ);
rootpage = BufferGetPage(rootbuf);
- rootopaque = (BTPageOpaque) PageGetSpecialPointer(rootpage);
+ rootopaque = BTPageGetOpaque(rootpage);
/*
* Since the cache might be stale, we check the page more carefully
@@ -440,7 +440,7 @@ _bt_getroot(Relation rel, int access)
rootbuf = _bt_getbuf(rel, P_NEW, BT_WRITE);
rootblkno = BufferGetBlockNumber(rootbuf);
rootpage = BufferGetPage(rootbuf);
- rootopaque = (BTPageOpaque) PageGetSpecialPointer(rootpage);
+ rootopaque = BTPageGetOpaque(rootpage);
rootopaque->btpo_prev = rootopaque->btpo_next = P_NONE;
rootopaque->btpo_flags = (BTP_LEAF | BTP_ROOT);
rootopaque->btpo_level = 0;
@@ -534,7 +534,7 @@ _bt_getroot(Relation rel, int access)
{
rootbuf = _bt_relandgetbuf(rel, rootbuf, rootblkno, BT_READ);
rootpage = BufferGetPage(rootbuf);
- rootopaque = (BTPageOpaque) PageGetSpecialPointer(rootpage);
+ rootopaque = BTPageGetOpaque(rootpage);
if (!P_IGNORE(rootopaque))
break;
@@ -598,7 +598,7 @@ _bt_gettrueroot(Relation rel)
metabuf = _bt_getbuf(rel, BTREE_METAPAGE, BT_READ);
metapg = BufferGetPage(metabuf);
- metaopaque = (BTPageOpaque) PageGetSpecialPointer(metapg);
+ metaopaque = BTPageGetOpaque(metapg);
metad = BTPageGetMeta(metapg);
if (!P_ISMETA(metaopaque) ||
@@ -637,7 +637,7 @@ _bt_gettrueroot(Relation rel)
{
rootbuf = _bt_relandgetbuf(rel, rootbuf, rootblkno, BT_READ);
rootpage = BufferGetPage(rootbuf);
- rootopaque = (BTPageOpaque) PageGetSpecialPointer(rootpage);
+ rootopaque = BTPageGetOpaque(rootpage);
if (!P_IGNORE(rootopaque))
break;
@@ -1220,7 +1220,7 @@ _bt_delitems_vacuum(Relation rel, Buffer buf,
* We can clear the vacuum cycle ID since this page has certainly been
* processed by the current vacuum scan.
*/
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
opaque->btpo_cycleid = 0;
/*
@@ -1338,7 +1338,7 @@ _bt_delitems_delete(Relation rel, Buffer buf, TransactionId latestRemovedXid,
* Unlike _bt_delitems_vacuum, we *must not* clear the vacuum cycle ID at
* this point. The VACUUM command alone controls vacuum cycle IDs.
*/
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
/*
* Clear the BTP_HAS_GARBAGE page flag.
@@ -1718,7 +1718,7 @@ _bt_leftsib_splitflag(Relation rel, BlockNumber leftsib, BlockNumber target)
buf = _bt_getbuf(rel, leftsib, BT_READ);
page = BufferGetPage(buf);
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
/*
* If the left sibling was concurrently split, so that its next-pointer
@@ -1773,7 +1773,7 @@ _bt_rightsib_halfdeadflag(Relation rel, BlockNumber leafrightsib)
buf = _bt_getbuf(rel, leafrightsib, BT_READ);
page = BufferGetPage(buf);
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
Assert(P_ISLEAF(opaque) && !P_ISDELETED(opaque));
result = P_ISHALFDEAD(opaque);
@@ -1842,7 +1842,7 @@ _bt_pagedel(Relation rel, Buffer leafbuf, BTVacState *vstate)
for (;;)
{
page = BufferGetPage(leafbuf);
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
/*
* Internal pages are never deleted directly, only as part of deleting
@@ -2099,7 +2099,7 @@ _bt_mark_page_halfdead(Relation rel, Buffer leafbuf, BTStack stack)
IndexTupleData trunctuple;
page = BufferGetPage(leafbuf);
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
Assert(!P_RIGHTMOST(opaque) && !P_ISROOT(opaque) &&
P_ISLEAF(opaque) && !P_IGNORE(opaque) &&
@@ -2154,7 +2154,7 @@ _bt_mark_page_halfdead(Relation rel, Buffer leafbuf, BTStack stack)
* before entering the critical section --- otherwise it'd be a PANIC.
*/
page = BufferGetPage(subtreeparent);
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
#ifdef USE_ASSERT_CHECKING
@@ -2201,7 +2201,7 @@ _bt_mark_page_halfdead(Relation rel, Buffer leafbuf, BTStack stack)
* nbtree/README.
*/
page = BufferGetPage(subtreeparent);
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
itemid = PageGetItemId(page, poffset);
itup = (IndexTuple) PageGetItem(page, itemid);
@@ -2216,7 +2216,7 @@ _bt_mark_page_halfdead(Relation rel, Buffer leafbuf, BTStack stack)
* is set to InvalidBlockNumber.
*/
page = BufferGetPage(leafbuf);
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
opaque->btpo_flags |= BTP_HALF_DEAD;
Assert(PageGetMaxOffsetNumber(page) == P_HIKEY);
@@ -2253,7 +2253,7 @@ _bt_mark_page_halfdead(Relation rel, Buffer leafbuf, BTStack stack)
XLogRegisterBuffer(1, subtreeparent, REGBUF_STANDARD);
page = BufferGetPage(leafbuf);
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
xlrec.leftblk = opaque->btpo_prev;
xlrec.rightblk = opaque->btpo_next;
@@ -2325,7 +2325,7 @@ _bt_unlink_halfdead_page(Relation rel, Buffer leafbuf, BlockNumber scanblkno,
BlockNumber leaftopparent;
page = BufferGetPage(leafbuf);
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
Assert(P_ISLEAF(opaque) && !P_ISDELETED(opaque) && P_ISHALFDEAD(opaque));
@@ -2364,7 +2364,7 @@ _bt_unlink_halfdead_page(Relation rel, Buffer leafbuf, BlockNumber scanblkno,
/* Fetch the block number of the target's left sibling */
buf = _bt_getbuf(rel, target, BT_READ);
page = BufferGetPage(buf);
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
leftsib = opaque->btpo_prev;
targetlevel = opaque->btpo_level;
Assert(targetlevel > 0);
@@ -2391,7 +2391,7 @@ _bt_unlink_halfdead_page(Relation rel, Buffer leafbuf, BlockNumber scanblkno,
{
lbuf = _bt_getbuf(rel, leftsib, BT_WRITE);
page = BufferGetPage(lbuf);
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
while (P_ISDELETED(opaque) || opaque->btpo_next != target)
{
bool leftsibvalid = true;
@@ -2441,7 +2441,7 @@ _bt_unlink_halfdead_page(Relation rel, Buffer leafbuf, BlockNumber scanblkno,
/* step right one page */
lbuf = _bt_getbuf(rel, leftsib, BT_WRITE);
page = BufferGetPage(lbuf);
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
}
}
else
@@ -2450,7 +2450,7 @@ _bt_unlink_halfdead_page(Relation rel, Buffer leafbuf, BlockNumber scanblkno,
/* Next write-lock the target page itself */
_bt_lockbuf(rel, buf, BT_WRITE);
page = BufferGetPage(buf);
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
/*
* Check page is still empty etc, else abandon deletion. This is just for
@@ -2505,7 +2505,7 @@ _bt_unlink_halfdead_page(Relation rel, Buffer leafbuf, BlockNumber scanblkno,
rightsib = opaque->btpo_next;
rbuf = _bt_getbuf(rel, rightsib, BT_WRITE);
page = BufferGetPage(rbuf);
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
if (opaque->btpo_prev != target)
ereport(ERROR,
(errcode(ERRCODE_INDEX_CORRUPTED),
@@ -2528,7 +2528,7 @@ _bt_unlink_halfdead_page(Relation rel, Buffer leafbuf, BlockNumber scanblkno,
if (leftsib == P_NONE && rightsib_is_rightmost)
{
page = BufferGetPage(rbuf);
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
if (P_RIGHTMOST(opaque))
{
/* rightsib will be the only one left on the level */
@@ -2565,12 +2565,12 @@ _bt_unlink_halfdead_page(Relation rel, Buffer leafbuf, BlockNumber scanblkno,
if (BufferIsValid(lbuf))
{
page = BufferGetPage(lbuf);
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
Assert(opaque->btpo_next == target);
opaque->btpo_next = rightsib;
}
page = BufferGetPage(rbuf);
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
Assert(opaque->btpo_prev == target);
opaque->btpo_prev = leftsib;
@@ -2599,7 +2599,7 @@ _bt_unlink_halfdead_page(Relation rel, Buffer leafbuf, BlockNumber scanblkno,
* of that scan.
*/
page = BufferGetPage(buf);
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
Assert(P_ISHALFDEAD(opaque) || !P_ISLEAF(opaque));
/*
@@ -2814,7 +2814,7 @@ _bt_lock_subtree_parent(Relation rel, BlockNumber child, BTStack stack,
parentoffset = stack->bts_offset;
page = BufferGetPage(pbuf);
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
maxoff = PageGetMaxOffsetNumber(page);
leftsibparent = opaque->btpo_prev;
diff --git a/src/backend/access/nbtree/nbtree.c b/src/backend/access/nbtree/nbtree.c
index c9b4964c1e..07600b86fb 100644
--- a/src/backend/access/nbtree/nbtree.c
+++ b/src/backend/access/nbtree/nbtree.c
@@ -1070,7 +1070,7 @@ backtrack:
if (!PageIsNew(page))
{
_bt_checkpage(rel, buf);
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
}
Assert(blkno <= scanblkno);
diff --git a/src/backend/access/nbtree/nbtsearch.c b/src/backend/access/nbtree/nbtsearch.c
index 9d82d4904d..c74543bfde 100644
--- a/src/backend/access/nbtree/nbtsearch.c
+++ b/src/backend/access/nbtree/nbtsearch.c
@@ -134,7 +134,7 @@ _bt_search(Relation rel, BTScanInsert key, Buffer *bufP, int access,
/* if this is a leaf page, we're done */
page = BufferGetPage(*bufP);
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
if (P_ISLEAF(opaque))
break;
@@ -268,7 +268,7 @@ _bt_moveright(Relation rel,
{
page = BufferGetPage(buf);
TestForOldSnapshot(snapshot, rel, page);
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
if (P_RIGHTMOST(opaque))
break;
@@ -347,7 +347,7 @@ _bt_binsrch(Relation rel,
cmpval;
page = BufferGetPage(buf);
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
/* Requesting nextkey semantics while using scantid seems nonsensical */
Assert(!key->nextkey || key->scantid == NULL);
@@ -451,7 +451,7 @@ _bt_binsrch_insert(Relation rel, BTInsertState insertstate)
cmpval;
page = BufferGetPage(insertstate->buf);
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
Assert(P_ISLEAF(opaque));
Assert(!key->nextkey);
@@ -659,7 +659,7 @@ _bt_compare(Relation rel,
OffsetNumber offnum)
{
TupleDesc itupdesc = RelationGetDescr(rel);
- BTPageOpaque opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ BTPageOpaque opaque = BTPageGetOpaque(page);
IndexTuple itup;
ItemPointer heapTid;
ScanKey scankey;
@@ -1536,7 +1536,7 @@ _bt_readpage(IndexScanDesc scan, ScanDirection dir, OffsetNumber offnum)
Assert(BufferIsValid(so->currPos.buf));
page = BufferGetPage(so->currPos.buf);
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
/* allow next page be processed by parallel worker */
if (scan->parallel_scan)
@@ -2007,7 +2007,7 @@ _bt_readnextpage(IndexScanDesc scan, BlockNumber blkno, ScanDirection dir)
so->currPos.buf = _bt_getbuf(rel, blkno, BT_READ);
page = BufferGetPage(so->currPos.buf);
TestForOldSnapshot(scan->xs_snapshot, rel, page);
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
/* check for deleted page */
if (!P_IGNORE(opaque))
{
@@ -2110,7 +2110,7 @@ _bt_readnextpage(IndexScanDesc scan, BlockNumber blkno, ScanDirection dir)
*/
page = BufferGetPage(so->currPos.buf);
TestForOldSnapshot(scan->xs_snapshot, rel, page);
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
if (!P_IGNORE(opaque))
{
PredicateLockPage(rel, BufferGetBlockNumber(so->currPos.buf), scan->xs_snapshot);
@@ -2191,7 +2191,7 @@ _bt_walk_left(Relation rel, Buffer buf, Snapshot snapshot)
BTPageOpaque opaque;
page = BufferGetPage(buf);
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
for (;;)
{
@@ -2216,7 +2216,7 @@ _bt_walk_left(Relation rel, Buffer buf, Snapshot snapshot)
buf = _bt_getbuf(rel, blkno, BT_READ);
page = BufferGetPage(buf);
TestForOldSnapshot(snapshot, rel, page);
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
/*
* If this isn't the page we want, walk right till we find what we
@@ -2243,14 +2243,14 @@ _bt_walk_left(Relation rel, Buffer buf, Snapshot snapshot)
buf = _bt_relandgetbuf(rel, buf, blkno, BT_READ);
page = BufferGetPage(buf);
TestForOldSnapshot(snapshot, rel, page);
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
}
/* Return to the original page to see what's up */
buf = _bt_relandgetbuf(rel, buf, obknum, BT_READ);
page = BufferGetPage(buf);
TestForOldSnapshot(snapshot, rel, page);
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
if (P_ISDELETED(opaque))
{
/*
@@ -2268,7 +2268,7 @@ _bt_walk_left(Relation rel, Buffer buf, Snapshot snapshot)
buf = _bt_relandgetbuf(rel, buf, blkno, BT_READ);
page = BufferGetPage(buf);
TestForOldSnapshot(snapshot, rel, page);
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
if (!P_ISDELETED(opaque))
break;
}
@@ -2329,7 +2329,7 @@ _bt_get_endpoint(Relation rel, uint32 level, bool rightmost,
page = BufferGetPage(buf);
TestForOldSnapshot(snapshot, rel, page);
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
for (;;)
{
@@ -2349,7 +2349,7 @@ _bt_get_endpoint(Relation rel, uint32 level, bool rightmost,
buf = _bt_relandgetbuf(rel, buf, blkno, BT_READ);
page = BufferGetPage(buf);
TestForOldSnapshot(snapshot, rel, page);
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
}
/* Done? */
@@ -2372,7 +2372,7 @@ _bt_get_endpoint(Relation rel, uint32 level, bool rightmost,
buf = _bt_relandgetbuf(rel, buf, blkno, BT_READ);
page = BufferGetPage(buf);
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
}
return buf;
@@ -2418,7 +2418,7 @@ _bt_endpoint(IndexScanDesc scan, ScanDirection dir)
PredicateLockPage(rel, BufferGetBlockNumber(buf), scan->xs_snapshot);
page = BufferGetPage(buf);
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
Assert(P_ISLEAF(opaque));
if (ScanDirectionIsForward(dir))
diff --git a/src/backend/access/nbtree/nbtsort.c b/src/backend/access/nbtree/nbtsort.c
index 8a19de2f66..c074513efa 100644
--- a/src/backend/access/nbtree/nbtsort.c
+++ b/src/backend/access/nbtree/nbtsort.c
@@ -625,7 +625,7 @@ _bt_blnewpage(uint32 level)
_bt_pageinit(page, BLCKSZ);
/* Initialize BT opaque state */
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
opaque->btpo_prev = opaque->btpo_next = P_NONE;
opaque->btpo_level = level;
opaque->btpo_flags = (level > 0) ? 0 : BTP_LEAF;
@@ -1000,9 +1000,9 @@ _bt_buildadd(BTWriteState *wstate, BTPageState *state, IndexTuple itup,
Assert((BTreeTupleGetNAtts(state->btps_lowkey, wstate->index) <=
IndexRelationGetNumberOfKeyAttributes(wstate->index) &&
BTreeTupleGetNAtts(state->btps_lowkey, wstate->index) > 0) ||
- P_LEFTMOST((BTPageOpaque) PageGetSpecialPointer(opage)));
+ P_LEFTMOST(BTPageGetOpaque(opage)));
Assert(BTreeTupleGetNAtts(state->btps_lowkey, wstate->index) == 0 ||
- !P_LEFTMOST((BTPageOpaque) PageGetSpecialPointer(opage)));
+ !P_LEFTMOST(BTPageGetOpaque(opage)));
BTreeTupleSetDownLink(state->btps_lowkey, oblkno);
_bt_buildadd(wstate, state->btps_next, state->btps_lowkey, 0);
pfree(state->btps_lowkey);
@@ -1017,8 +1017,8 @@ _bt_buildadd(BTWriteState *wstate, BTPageState *state, IndexTuple itup,
* Set the sibling links for both pages.
*/
{
- BTPageOpaque oopaque = (BTPageOpaque) PageGetSpecialPointer(opage);
- BTPageOpaque nopaque = (BTPageOpaque) PageGetSpecialPointer(npage);
+ BTPageOpaque oopaque = BTPageGetOpaque(opage);
+ BTPageOpaque nopaque = BTPageGetOpaque(npage);
oopaque->btpo_next = nblkno;
nopaque->btpo_prev = oblkno;
@@ -1125,7 +1125,7 @@ _bt_uppershutdown(BTWriteState *wstate, BTPageState *state)
BTPageOpaque opaque;
blkno = s->btps_blkno;
- opaque = (BTPageOpaque) PageGetSpecialPointer(s->btps_page);
+ opaque = BTPageGetOpaque(s->btps_page);
/*
* We have to link the last page on this level to somewhere.
diff --git a/src/backend/access/nbtree/nbtsplitloc.c b/src/backend/access/nbtree/nbtsplitloc.c
index c46594e1a2..ee01ceafda 100644
--- a/src/backend/access/nbtree/nbtsplitloc.c
+++ b/src/backend/access/nbtree/nbtsplitloc.c
@@ -152,7 +152,7 @@ _bt_findsplitloc(Relation rel,
SplitPoint leftpage,
rightpage;
- opaque = (BTPageOpaque) PageGetSpecialPointer(origpage);
+ opaque = BTPageGetOpaque(origpage);
maxoff = PageGetMaxOffsetNumber(origpage);
/* Total free space available on a btree page, after fixed overhead */
diff --git a/src/backend/access/nbtree/nbtutils.c b/src/backend/access/nbtree/nbtutils.c
index 84164748b3..96c72fc432 100644
--- a/src/backend/access/nbtree/nbtutils.c
+++ b/src/backend/access/nbtree/nbtutils.c
@@ -1774,7 +1774,7 @@ _bt_killitems(IndexScanDesc scan)
}
}
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
minoff = P_FIRSTDATAKEY(opaque);
maxoff = PageGetMaxOffsetNumber(page);
@@ -2474,7 +2474,7 @@ _bt_check_natts(Relation rel, bool heapkeyspace, Page page, OffsetNumber offnum)
{
int16 natts = IndexRelationGetNumberOfAttributes(rel);
int16 nkeyatts = IndexRelationGetNumberOfKeyAttributes(rel);
- BTPageOpaque opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ BTPageOpaque opaque = BTPageGetOpaque(page);
IndexTuple itup;
int tupnatts;
@@ -2662,7 +2662,7 @@ _bt_check_third_page(Relation rel, Relation heap, bool needheaptidspace,
* Internal page insertions cannot fail here, because that would mean that
* an earlier leaf level insertion that should have failed didn't
*/
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
if (!P_ISLEAF(opaque))
elog(ERROR, "cannot insert oversized tuple of size %zu on internal page of index \"%s\"",
itemsz, RelationGetRelationName(rel));
diff --git a/src/backend/access/nbtree/nbtxlog.c b/src/backend/access/nbtree/nbtxlog.c
index 611f412ba8..fba124b940 100644
--- a/src/backend/access/nbtree/nbtxlog.c
+++ b/src/backend/access/nbtree/nbtxlog.c
@@ -115,7 +115,7 @@ _bt_restore_meta(XLogReaderState *record, uint8 block_id)
md->btm_last_cleanup_num_heap_tuples = -1.0;
md->btm_allequalimage = xlrec->allequalimage;
- pageop = (BTPageOpaque) PageGetSpecialPointer(metapg);
+ pageop = BTPageGetOpaque(metapg);
pageop->btpo_flags = BTP_META;
/*
@@ -146,7 +146,7 @@ _bt_clear_incomplete_split(XLogReaderState *record, uint8 block_id)
if (XLogReadBufferForRedo(record, block_id, &buf) == BLK_NEEDS_REDO)
{
Page page = (Page) BufferGetPage(buf);
- BTPageOpaque pageop = (BTPageOpaque) PageGetSpecialPointer(page);
+ BTPageOpaque pageop = BTPageGetOpaque(page);
Assert(P_INCOMPLETE_SPLIT(pageop));
pageop->btpo_flags &= ~BTP_INCOMPLETE_SPLIT;
@@ -292,7 +292,7 @@ btree_xlog_split(bool newitemonleft, XLogReaderState *record)
rpage = (Page) BufferGetPage(rbuf);
_bt_pageinit(rpage, BufferGetPageSize(rbuf));
- ropaque = (BTPageOpaque) PageGetSpecialPointer(rpage);
+ ropaque = BTPageGetOpaque(rpage);
ropaque->btpo_prev = origpagenumber;
ropaque->btpo_next = spagenumber;
@@ -317,7 +317,7 @@ btree_xlog_split(bool newitemonleft, XLogReaderState *record)
* same for the right page.
*/
Page origpage = (Page) BufferGetPage(buf);
- BTPageOpaque oopaque = (BTPageOpaque) PageGetSpecialPointer(origpage);
+ BTPageOpaque oopaque = BTPageGetOpaque(origpage);
OffsetNumber off;
IndexTuple newitem = NULL,
left_hikey = NULL,
@@ -442,7 +442,7 @@ btree_xlog_split(bool newitemonleft, XLogReaderState *record)
if (XLogReadBufferForRedo(record, 2, &sbuf) == BLK_NEEDS_REDO)
{
Page spage = (Page) BufferGetPage(sbuf);
- BTPageOpaque spageop = (BTPageOpaque) PageGetSpecialPointer(spage);
+ BTPageOpaque spageop = BTPageGetOpaque(spage);
spageop->btpo_prev = rightpagenumber;
@@ -473,7 +473,7 @@ btree_xlog_dedup(XLogReaderState *record)
{
char *ptr = XLogRecGetBlockData(record, 0, NULL);
Page page = (Page) BufferGetPage(buf);
- BTPageOpaque opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ BTPageOpaque opaque = BTPageGetOpaque(page);
OffsetNumber offnum,
minoff,
maxoff;
@@ -541,7 +541,7 @@ btree_xlog_dedup(XLogReaderState *record)
if (P_HAS_GARBAGE(opaque))
{
- BTPageOpaque nopaque = (BTPageOpaque) PageGetSpecialPointer(newpage);
+ BTPageOpaque nopaque = BTPageGetOpaque(newpage);
nopaque->btpo_flags &= ~BTP_HAS_GARBAGE;
}
@@ -639,7 +639,7 @@ btree_xlog_vacuum(XLogReaderState *record)
* Mark the page as not containing any LP_DEAD items --- see comments
* in _bt_delitems_vacuum().
*/
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
opaque->btpo_flags &= ~BTP_HAS_GARBAGE;
PageSetLSN(page, lsn);
@@ -699,7 +699,7 @@ btree_xlog_delete(XLogReaderState *record)
PageIndexMultiDelete(page, (OffsetNumber *) ptr, xlrec->ndeleted);
/* Mark the page as not containing any LP_DEAD items */
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
opaque->btpo_flags &= ~BTP_HAS_GARBAGE;
PageSetLSN(page, lsn);
@@ -737,7 +737,7 @@ btree_xlog_mark_page_halfdead(uint8 info, XLogReaderState *record)
BlockNumber rightsib;
page = (Page) BufferGetPage(buffer);
- pageop = (BTPageOpaque) PageGetSpecialPointer(page);
+ pageop = BTPageGetOpaque(page);
poffset = xlrec->poffset;
@@ -768,7 +768,7 @@ btree_xlog_mark_page_halfdead(uint8 info, XLogReaderState *record)
page = (Page) BufferGetPage(buffer);
_bt_pageinit(page, BufferGetPageSize(buffer));
- pageop = (BTPageOpaque) PageGetSpecialPointer(page);
+ pageop = BTPageGetOpaque(page);
pageop->btpo_prev = xlrec->leftblk;
pageop->btpo_next = xlrec->rightblk;
@@ -833,7 +833,7 @@ btree_xlog_unlink_page(uint8 info, XLogReaderState *record)
if (XLogReadBufferForRedo(record, 1, &leftbuf) == BLK_NEEDS_REDO)
{
page = (Page) BufferGetPage(leftbuf);
- pageop = (BTPageOpaque) PageGetSpecialPointer(page);
+ pageop = BTPageGetOpaque(page);
pageop->btpo_next = rightsib;
PageSetLSN(page, lsn);
@@ -848,7 +848,7 @@ btree_xlog_unlink_page(uint8 info, XLogReaderState *record)
page = (Page) BufferGetPage(target);
_bt_pageinit(page, BufferGetPageSize(target));
- pageop = (BTPageOpaque) PageGetSpecialPointer(page);
+ pageop = BTPageGetOpaque(page);
pageop->btpo_prev = leftsib;
pageop->btpo_next = rightsib;
@@ -865,7 +865,7 @@ btree_xlog_unlink_page(uint8 info, XLogReaderState *record)
if (XLogReadBufferForRedo(record, 2, &rightbuf) == BLK_NEEDS_REDO)
{
page = (Page) BufferGetPage(rightbuf);
- pageop = (BTPageOpaque) PageGetSpecialPointer(page);
+ pageop = BTPageGetOpaque(page);
pageop->btpo_prev = leftsib;
PageSetLSN(page, lsn);
@@ -906,7 +906,7 @@ btree_xlog_unlink_page(uint8 info, XLogReaderState *record)
page = (Page) BufferGetPage(leafbuf);
_bt_pageinit(page, BufferGetPageSize(leafbuf));
- pageop = (BTPageOpaque) PageGetSpecialPointer(page);
+ pageop = BTPageGetOpaque(page);
pageop->btpo_flags = BTP_HALF_DEAD | BTP_LEAF;
pageop->btpo_prev = xlrec->leafleftsib;
@@ -948,7 +948,7 @@ btree_xlog_newroot(XLogReaderState *record)
page = (Page) BufferGetPage(buffer);
_bt_pageinit(page, BufferGetPageSize(buffer));
- pageop = (BTPageOpaque) PageGetSpecialPointer(page);
+ pageop = BTPageGetOpaque(page);
pageop->btpo_flags = BTP_ROOT;
pageop->btpo_prev = pageop->btpo_next = P_NONE;
@@ -1097,7 +1097,7 @@ btree_mask(char *pagedata, BlockNumber blkno)
mask_page_hint_bits(page);
mask_unused_space(page);
- maskopaq = (BTPageOpaque) PageGetSpecialPointer(page);
+ maskopaq = BTPageGetOpaque(page);
if (P_ISLEAF(maskopaq))
{
diff --git a/src/include/access/nbtree.h b/src/include/access/nbtree.h
index 9fec6fb1a8..31edddefc0 100644
--- a/src/include/access/nbtree.h
+++ b/src/include/access/nbtree.h
@@ -70,6 +70,13 @@ typedef struct BTPageOpaqueData
typedef BTPageOpaqueData *BTPageOpaque;
+/*
+ * Pre-aligned accessor to the opaque data in the page's special region.
+ * Avoids data dependency on page header in non-assert builds.
+ */
+#define BTPageGetOpaque(page) \
+ ((BTPageOpaque) PageGetSpecialOpaque((page), BTPageOpaqueData))
+
/* Bits defined in btpo_flags */
#define BTP_LEAF (1 << 0) /* leaf page, i.e. not internal page */
#define BTP_ROOT (1 << 1) /* root page (has no parent) */
@@ -241,7 +248,7 @@ BTPageSetDeleted(Page page, FullTransactionId safexid)
PageHeader header;
BTDeletedPageData *contents;
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
header = ((PageHeader) page);
opaque->btpo_flags &= ~BTP_HALF_DEAD;
@@ -263,7 +270,7 @@ BTPageGetDeleteXid(Page page)
/* We only expect to be called with a deleted page */
Assert(!PageIsNew(page));
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
Assert(P_ISDELETED(opaque));
/* pg_upgrade'd deleted page -- must be safe to delete now */
@@ -294,7 +301,7 @@ BTPageIsRecyclable(Page page)
Assert(!PageIsNew(page));
/* Recycling okay iff page is deleted and safexid is old enough */
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
if (P_ISDELETED(opaque))
{
/*
--
2.30.2
v2-0003-Update-hash-code-to-use-PageGetSpecialOpaque-repl.patchapplication/x-patch; name=v2-0003-Update-hash-code-to-use-PageGetSpecialOpaque-repl.patchDownload
From b666d9c52c5d47fa66e453d7e2dbb474956a0bff Mon Sep 17 00:00:00 2001
From: Matthias van de Meent <boekewurm+postgres@gmail.com>
Date: Mon, 28 Mar 2022 16:45:52 +0200
Subject: [PATCH v2 3/3] Update hash code to use PageGetSpecialOpaque; replace
PageGetSpecialPointer+cast
As in the earlier patch, this reduces memory access requirements by directly
reading into the target space instead of needing to access the header of the
page first.
Also makes the code a lot more readable by replacing templated macro+casts
with a single macro specialized for hash page opaques.
---
contrib/pageinspect/hashfuncs.c | 6 +++---
contrib/pgstattuple/pgstatindex.c | 2 +-
contrib/pgstattuple/pgstattuple.c | 2 +-
src/backend/access/hash/hash.c | 6 +++---
src/backend/access/hash/hash_xlog.c | 26 ++++++++++++------------
src/backend/access/hash/hashinsert.c | 6 +++---
src/backend/access/hash/hashovfl.c | 22 ++++++++++----------
src/backend/access/hash/hashpage.c | 30 ++++++++++++++--------------
src/backend/access/hash/hashsearch.c | 12 +++++------
src/backend/access/hash/hashutil.c | 4 ++--
src/include/access/hash.h | 3 +++
11 files changed, 61 insertions(+), 58 deletions(-)
diff --git a/contrib/pageinspect/hashfuncs.c b/contrib/pageinspect/hashfuncs.c
index 6de21d6608..69af7b962f 100644
--- a/contrib/pageinspect/hashfuncs.c
+++ b/contrib/pageinspect/hashfuncs.c
@@ -72,7 +72,7 @@ verify_hash_page(bytea *raw_page, int flags)
(int) MAXALIGN(sizeof(HashPageOpaqueData)),
(int) PageGetSpecialSize(page))));
- pageopaque = (HashPageOpaque) PageGetSpecialPointer(page);
+ pageopaque = HashPageGetOpaque(page);
if (pageopaque->hasho_page_id != HASHO_PAGE_ID)
ereport(ERROR,
(errcode(ERRCODE_INVALID_PARAMETER_VALUE),
@@ -154,7 +154,7 @@ static void
GetHashPageStatistics(Page page, HashPageStat *stat)
{
OffsetNumber maxoff = PageGetMaxOffsetNumber(page);
- HashPageOpaque opaque = (HashPageOpaque) PageGetSpecialPointer(page);
+ HashPageOpaque opaque = HashPageGetOpaque(page);
int off;
stat->dead_items = stat->live_items = 0;
@@ -206,7 +206,7 @@ hash_page_type(PG_FUNCTION_ARGS)
type = "unused";
else
{
- opaque = (HashPageOpaque) PageGetSpecialPointer(page);
+ opaque = HashPageGetOpaque(page);
/* page type (flags) */
pagetype = opaque->hasho_flag & LH_PAGE_TYPE;
diff --git a/contrib/pgstattuple/pgstatindex.c b/contrib/pgstattuple/pgstatindex.c
index b9939451b4..e1048e47ff 100644
--- a/contrib/pgstattuple/pgstatindex.c
+++ b/contrib/pgstattuple/pgstatindex.c
@@ -641,7 +641,7 @@ pgstathashindex(PG_FUNCTION_ARGS)
HashPageOpaque opaque;
int pagetype;
- opaque = (HashPageOpaque) PageGetSpecialPointer(page);
+ opaque = HashPageGetOpaque(page);
pagetype = opaque->hasho_flag & LH_PAGE_TYPE;
if (pagetype == LH_BUCKET_PAGE)
diff --git a/contrib/pgstattuple/pgstattuple.c b/contrib/pgstattuple/pgstattuple.c
index a9658d389b..3094566908 100644
--- a/contrib/pgstattuple/pgstattuple.c
+++ b/contrib/pgstattuple/pgstattuple.c
@@ -452,7 +452,7 @@ pgstat_hash_page(pgstattuple_type *stat, Relation rel, BlockNumber blkno,
{
HashPageOpaque opaque;
- opaque = (HashPageOpaque) PageGetSpecialPointer(page);
+ opaque = HashPageGetOpaque(page);
switch (opaque->hasho_flag & LH_PAGE_TYPE)
{
case LH_UNUSED_PAGE:
diff --git a/src/backend/access/hash/hash.c b/src/backend/access/hash/hash.c
index a259a301fa..fd1a7119b6 100644
--- a/src/backend/access/hash/hash.c
+++ b/src/backend/access/hash/hash.c
@@ -515,7 +515,7 @@ loop_top:
_hash_checkpage(rel, buf, LH_BUCKET_PAGE);
page = BufferGetPage(buf);
- bucket_opaque = (HashPageOpaque) PageGetSpecialPointer(page);
+ bucket_opaque = HashPageGetOpaque(page);
/*
* If the bucket contains tuples that are moved by split, then we need
@@ -717,7 +717,7 @@ hashbucketcleanup(Relation rel, Bucket cur_bucket, Buffer bucket_buf,
vacuum_delay_point();
page = BufferGetPage(buf);
- opaque = (HashPageOpaque) PageGetSpecialPointer(page);
+ opaque = HashPageGetOpaque(page);
/* Scan each tuple in page */
maxoffno = PageGetMaxOffsetNumber(page);
@@ -884,7 +884,7 @@ hashbucketcleanup(Relation rel, Bucket cur_bucket, Buffer bucket_buf,
Page page;
page = BufferGetPage(bucket_buf);
- bucket_opaque = (HashPageOpaque) PageGetSpecialPointer(page);
+ bucket_opaque = HashPageGetOpaque(page);
/* No ereport(ERROR) until changes are logged */
START_CRIT_SECTION();
diff --git a/src/backend/access/hash/hash_xlog.c b/src/backend/access/hash/hash_xlog.c
index 55937b9a68..62dbfc3a5d 100644
--- a/src/backend/access/hash/hash_xlog.c
+++ b/src/backend/access/hash/hash_xlog.c
@@ -203,7 +203,7 @@ hash_xlog_add_ovfl_page(XLogReaderState *record)
true);
/* update backlink */
ovflpage = BufferGetPage(ovflbuf);
- ovflopaque = (HashPageOpaque) PageGetSpecialPointer(ovflpage);
+ ovflopaque = HashPageGetOpaque(ovflpage);
ovflopaque->hasho_prevblkno = leftblk;
PageSetLSN(ovflpage, lsn);
@@ -215,7 +215,7 @@ hash_xlog_add_ovfl_page(XLogReaderState *record)
HashPageOpaque leftopaque;
leftpage = BufferGetPage(leftbuf);
- leftopaque = (HashPageOpaque) PageGetSpecialPointer(leftpage);
+ leftopaque = HashPageGetOpaque(leftpage);
leftopaque->hasho_nextblkno = rightblk;
PageSetLSN(leftpage, lsn);
@@ -342,7 +342,7 @@ hash_xlog_split_allocate_page(XLogReaderState *record)
HashPageOpaque oldopaque;
oldpage = BufferGetPage(oldbuf);
- oldopaque = (HashPageOpaque) PageGetSpecialPointer(oldpage);
+ oldopaque = HashPageGetOpaque(oldpage);
oldopaque->hasho_flag = xlrec->old_bucket_flag;
oldopaque->hasho_prevblkno = xlrec->new_bucket;
@@ -465,7 +465,7 @@ hash_xlog_split_complete(XLogReaderState *record)
HashPageOpaque oldopaque;
oldpage = BufferGetPage(oldbuf);
- oldopaque = (HashPageOpaque) PageGetSpecialPointer(oldpage);
+ oldopaque = HashPageGetOpaque(oldpage);
oldopaque->hasho_flag = xlrec->old_bucket_flag;
@@ -488,7 +488,7 @@ hash_xlog_split_complete(XLogReaderState *record)
HashPageOpaque nopaque;
newpage = BufferGetPage(newbuf);
- nopaque = (HashPageOpaque) PageGetSpecialPointer(newpage);
+ nopaque = HashPageGetOpaque(newpage);
nopaque->hasho_flag = xlrec->new_bucket_flag;
@@ -710,7 +710,7 @@ hash_xlog_squeeze_page(XLogReaderState *record)
*/
if (xldata->is_prev_bucket_same_wrt)
{
- HashPageOpaque writeopaque = (HashPageOpaque) PageGetSpecialPointer(writepage);
+ HashPageOpaque writeopaque = HashPageGetOpaque(writepage);
writeopaque->hasho_nextblkno = xldata->nextblkno;
}
@@ -729,7 +729,7 @@ hash_xlog_squeeze_page(XLogReaderState *record)
_hash_pageinit(ovflpage, BufferGetPageSize(ovflbuf));
- ovflopaque = (HashPageOpaque) PageGetSpecialPointer(ovflpage);
+ ovflopaque = HashPageGetOpaque(ovflpage);
ovflopaque->hasho_prevblkno = InvalidBlockNumber;
ovflopaque->hasho_nextblkno = InvalidBlockNumber;
@@ -748,7 +748,7 @@ hash_xlog_squeeze_page(XLogReaderState *record)
XLogReadBufferForRedo(record, 3, &prevbuf) == BLK_NEEDS_REDO)
{
Page prevpage = BufferGetPage(prevbuf);
- HashPageOpaque prevopaque = (HashPageOpaque) PageGetSpecialPointer(prevpage);
+ HashPageOpaque prevopaque = HashPageGetOpaque(prevpage);
prevopaque->hasho_nextblkno = xldata->nextblkno;
@@ -766,7 +766,7 @@ hash_xlog_squeeze_page(XLogReaderState *record)
if (XLogReadBufferForRedo(record, 4, &nextbuf) == BLK_NEEDS_REDO)
{
Page nextpage = BufferGetPage(nextbuf);
- HashPageOpaque nextopaque = (HashPageOpaque) PageGetSpecialPointer(nextpage);
+ HashPageOpaque nextopaque = HashPageGetOpaque(nextpage);
nextopaque->hasho_prevblkno = xldata->prevblkno;
@@ -903,7 +903,7 @@ hash_xlog_delete(XLogReaderState *record)
{
HashPageOpaque pageopaque;
- pageopaque = (HashPageOpaque) PageGetSpecialPointer(page);
+ pageopaque = HashPageGetOpaque(page);
pageopaque->hasho_flag &= ~LH_PAGE_HAS_DEAD_TUPLES;
}
@@ -933,7 +933,7 @@ hash_xlog_split_cleanup(XLogReaderState *record)
page = (Page) BufferGetPage(buffer);
- bucket_opaque = (HashPageOpaque) PageGetSpecialPointer(page);
+ bucket_opaque = HashPageGetOpaque(page);
bucket_opaque->hasho_flag &= ~LH_BUCKET_NEEDS_SPLIT_CLEANUP;
PageSetLSN(page, lsn);
MarkBufferDirty(buffer);
@@ -1024,7 +1024,7 @@ hash_xlog_vacuum_one_page(XLogReaderState *record)
* Mark the page as not containing any LP_DEAD items. See comments in
* _hash_vacuum_one_page() for details.
*/
- pageopaque = (HashPageOpaque) PageGetSpecialPointer(page);
+ pageopaque = HashPageGetOpaque(page);
pageopaque->hasho_flag &= ~LH_PAGE_HAS_DEAD_TUPLES;
PageSetLSN(page, lsn);
@@ -1116,7 +1116,7 @@ hash_mask(char *pagedata, BlockNumber blkno)
mask_page_hint_bits(page);
mask_unused_space(page);
- opaque = (HashPageOpaque) PageGetSpecialPointer(page);
+ opaque = HashPageGetOpaque(page);
pagetype = opaque->hasho_flag & LH_PAGE_TYPE;
if (pagetype == LH_UNUSED_PAGE)
diff --git a/src/backend/access/hash/hashinsert.c b/src/backend/access/hash/hashinsert.c
index faf609c157..4f2fecb908 100644
--- a/src/backend/access/hash/hashinsert.c
+++ b/src/backend/access/hash/hashinsert.c
@@ -95,7 +95,7 @@ restart_insert:
bucket_buf = buf;
page = BufferGetPage(buf);
- pageopaque = (HashPageOpaque) PageGetSpecialPointer(page);
+ pageopaque = HashPageGetOpaque(page);
bucket = pageopaque->hasho_bucket;
/*
@@ -183,7 +183,7 @@ restart_insert:
/* should fit now, given test above */
Assert(PageGetFreeSpace(page) >= itemsz);
}
- pageopaque = (HashPageOpaque) PageGetSpecialPointer(page);
+ pageopaque = HashPageGetOpaque(page);
Assert((pageopaque->hasho_flag & LH_PAGE_TYPE) == LH_OVERFLOW_PAGE);
Assert(pageopaque->hasho_bucket == bucket);
}
@@ -384,7 +384,7 @@ _hash_vacuum_one_page(Relation rel, Relation hrel, Buffer metabuf, Buffer buf)
* check it. Remember that LH_PAGE_HAS_DEAD_TUPLES is only a hint
* anyway.
*/
- pageopaque = (HashPageOpaque) PageGetSpecialPointer(page);
+ pageopaque = HashPageGetOpaque(page);
pageopaque->hasho_flag &= ~LH_PAGE_HAS_DEAD_TUPLES;
metap = HashPageGetMeta(BufferGetPage(metabuf));
diff --git a/src/backend/access/hash/hashovfl.c b/src/backend/access/hash/hashovfl.c
index 4836875196..e34cfc302d 100644
--- a/src/backend/access/hash/hashovfl.c
+++ b/src/backend/access/hash/hashovfl.c
@@ -159,7 +159,7 @@ _hash_addovflpage(Relation rel, Buffer metabuf, Buffer buf, bool retain_pin)
BlockNumber nextblkno;
page = BufferGetPage(buf);
- pageopaque = (HashPageOpaque) PageGetSpecialPointer(page);
+ pageopaque = HashPageGetOpaque(page);
nextblkno = pageopaque->hasho_nextblkno;
if (!BlockNumberIsValid(nextblkno))
@@ -364,7 +364,7 @@ found:
/* initialize new overflow page */
ovflpage = BufferGetPage(ovflbuf);
- ovflopaque = (HashPageOpaque) PageGetSpecialPointer(ovflpage);
+ ovflopaque = HashPageGetOpaque(ovflpage);
ovflopaque->hasho_prevblkno = BufferGetBlockNumber(buf);
ovflopaque->hasho_nextblkno = InvalidBlockNumber;
ovflopaque->hasho_bucket = pageopaque->hasho_bucket;
@@ -516,7 +516,7 @@ _hash_freeovflpage(Relation rel, Buffer bucketbuf, Buffer ovflbuf,
_hash_checkpage(rel, ovflbuf, LH_OVERFLOW_PAGE);
ovflblkno = BufferGetBlockNumber(ovflbuf);
ovflpage = BufferGetPage(ovflbuf);
- ovflopaque = (HashPageOpaque) PageGetSpecialPointer(ovflpage);
+ ovflopaque = HashPageGetOpaque(ovflpage);
nextblkno = ovflopaque->hasho_nextblkno;
prevblkno = ovflopaque->hasho_prevblkno;
writeblkno = BufferGetBlockNumber(wbuf);
@@ -600,7 +600,7 @@ _hash_freeovflpage(Relation rel, Buffer bucketbuf, Buffer ovflbuf,
*/
_hash_pageinit(ovflpage, BufferGetPageSize(ovflbuf));
- ovflopaque = (HashPageOpaque) PageGetSpecialPointer(ovflpage);
+ ovflopaque = HashPageGetOpaque(ovflpage);
ovflopaque->hasho_prevblkno = InvalidBlockNumber;
ovflopaque->hasho_nextblkno = InvalidBlockNumber;
@@ -613,7 +613,7 @@ _hash_freeovflpage(Relation rel, Buffer bucketbuf, Buffer ovflbuf,
if (BufferIsValid(prevbuf))
{
Page prevpage = BufferGetPage(prevbuf);
- HashPageOpaque prevopaque = (HashPageOpaque) PageGetSpecialPointer(prevpage);
+ HashPageOpaque prevopaque = HashPageGetOpaque(prevpage);
Assert(prevopaque->hasho_bucket == bucket);
prevopaque->hasho_nextblkno = nextblkno;
@@ -622,7 +622,7 @@ _hash_freeovflpage(Relation rel, Buffer bucketbuf, Buffer ovflbuf,
if (BufferIsValid(nextbuf))
{
Page nextpage = BufferGetPage(nextbuf);
- HashPageOpaque nextopaque = (HashPageOpaque) PageGetSpecialPointer(nextpage);
+ HashPageOpaque nextopaque = HashPageGetOpaque(nextpage);
Assert(nextopaque->hasho_bucket == bucket);
nextopaque->hasho_prevblkno = prevblkno;
@@ -751,7 +751,7 @@ _hash_initbitmapbuffer(Buffer buf, uint16 bmsize, bool initpage)
_hash_pageinit(pg, BufferGetPageSize(buf));
/* initialize the page's special space */
- op = (HashPageOpaque) PageGetSpecialPointer(pg);
+ op = HashPageGetOpaque(pg);
op->hasho_prevblkno = InvalidBlockNumber;
op->hasho_nextblkno = InvalidBlockNumber;
op->hasho_bucket = InvalidBucket;
@@ -824,7 +824,7 @@ _hash_squeezebucket(Relation rel,
wblkno = bucket_blkno;
wbuf = bucket_buf;
wpage = BufferGetPage(wbuf);
- wopaque = (HashPageOpaque) PageGetSpecialPointer(wpage);
+ wopaque = HashPageGetOpaque(wpage);
/*
* if there aren't any overflow pages, there's nothing to squeeze. caller
@@ -855,7 +855,7 @@ _hash_squeezebucket(Relation rel,
LH_OVERFLOW_PAGE,
bstrategy);
rpage = BufferGetPage(rbuf);
- ropaque = (HashPageOpaque) PageGetSpecialPointer(rpage);
+ ropaque = HashPageGetOpaque(rpage);
Assert(ropaque->hasho_bucket == bucket);
} while (BlockNumberIsValid(ropaque->hasho_nextblkno));
@@ -1005,7 +1005,7 @@ readpage:
wbuf = next_wbuf;
wpage = BufferGetPage(wbuf);
- wopaque = (HashPageOpaque) PageGetSpecialPointer(wpage);
+ wopaque = HashPageGetOpaque(wpage);
Assert(wopaque->hasho_bucket == bucket);
retain_pin = false;
@@ -1076,7 +1076,7 @@ readpage:
LH_OVERFLOW_PAGE,
bstrategy);
rpage = BufferGetPage(rbuf);
- ropaque = (HashPageOpaque) PageGetSpecialPointer(rpage);
+ ropaque = HashPageGetOpaque(rpage);
Assert(ropaque->hasho_bucket == bucket);
}
diff --git a/src/backend/access/hash/hashpage.c b/src/backend/access/hash/hashpage.c
index 28c5297a1d..39206d1942 100644
--- a/src/backend/access/hash/hashpage.c
+++ b/src/backend/access/hash/hashpage.c
@@ -166,7 +166,7 @@ _hash_initbuf(Buffer buf, uint32 max_bucket, uint32 num_bucket, uint32 flag,
if (initpage)
_hash_pageinit(page, BufferGetPageSize(buf));
- pageopaque = (HashPageOpaque) PageGetSpecialPointer(page);
+ pageopaque = HashPageGetOpaque(page);
/*
* Set hasho_prevblkno with current hashm_maxbucket. This value will be
@@ -529,7 +529,7 @@ _hash_init_metabuffer(Buffer buf, double num_tuples, RegProcedure procid,
if (initpage)
_hash_pageinit(page, BufferGetPageSize(buf));
- pageopaque = (HashPageOpaque) PageGetSpecialPointer(page);
+ pageopaque = HashPageGetOpaque(page);
pageopaque->hasho_prevblkno = InvalidBlockNumber;
pageopaque->hasho_nextblkno = InvalidBlockNumber;
pageopaque->hasho_bucket = InvalidBucket;
@@ -693,7 +693,7 @@ restart_expand:
goto fail;
opage = BufferGetPage(buf_oblkno);
- oopaque = (HashPageOpaque) PageGetSpecialPointer(opage);
+ oopaque = HashPageGetOpaque(opage);
/*
* We want to finish the split from a bucket as there is no apparent
@@ -864,7 +864,7 @@ restart_expand:
lowmask = metap->hashm_lowmask;
opage = BufferGetPage(buf_oblkno);
- oopaque = (HashPageOpaque) PageGetSpecialPointer(opage);
+ oopaque = HashPageGetOpaque(opage);
/*
* Mark the old bucket to indicate that split is in progress. (At
@@ -883,7 +883,7 @@ restart_expand:
* initialize the new bucket's primary page and mark it to indicate that
* split is in progress.
*/
- nopaque = (HashPageOpaque) PageGetSpecialPointer(npage);
+ nopaque = HashPageGetOpaque(npage);
nopaque->hasho_prevblkno = maxbucket;
nopaque->hasho_nextblkno = InvalidBlockNumber;
nopaque->hasho_bucket = new_bucket;
@@ -1010,7 +1010,7 @@ _hash_alloc_buckets(Relation rel, BlockNumber firstblock, uint32 nblocks)
*/
_hash_pageinit(page, BLCKSZ);
- ovflopaque = (HashPageOpaque) PageGetSpecialPointer(page);
+ ovflopaque = HashPageGetOpaque(page);
ovflopaque->hasho_prevblkno = InvalidBlockNumber;
ovflopaque->hasho_nextblkno = InvalidBlockNumber;
@@ -1091,11 +1091,11 @@ _hash_splitbucket(Relation rel,
bucket_obuf = obuf;
opage = BufferGetPage(obuf);
- oopaque = (HashPageOpaque) PageGetSpecialPointer(opage);
+ oopaque = HashPageGetOpaque(opage);
bucket_nbuf = nbuf;
npage = BufferGetPage(nbuf);
- nopaque = (HashPageOpaque) PageGetSpecialPointer(npage);
+ nopaque = HashPageGetOpaque(npage);
/* Copy the predicate locks from old bucket to new bucket. */
PredicateLockPageSplit(rel,
@@ -1198,7 +1198,7 @@ _hash_splitbucket(Relation rel,
/* chain to a new overflow page */
nbuf = _hash_addovflpage(rel, metabuf, nbuf, (nbuf == bucket_nbuf));
npage = BufferGetPage(nbuf);
- nopaque = (HashPageOpaque) PageGetSpecialPointer(npage);
+ nopaque = HashPageGetOpaque(npage);
}
itups[nitups++] = new_itup;
@@ -1251,7 +1251,7 @@ _hash_splitbucket(Relation rel,
/* Else, advance to next old page */
obuf = _hash_getbuf(rel, oblkno, HASH_READ, LH_OVERFLOW_PAGE);
opage = BufferGetPage(obuf);
- oopaque = (HashPageOpaque) PageGetSpecialPointer(opage);
+ oopaque = HashPageGetOpaque(opage);
}
/*
@@ -1264,11 +1264,11 @@ _hash_splitbucket(Relation rel,
*/
LockBuffer(bucket_obuf, BUFFER_LOCK_EXCLUSIVE);
opage = BufferGetPage(bucket_obuf);
- oopaque = (HashPageOpaque) PageGetSpecialPointer(opage);
+ oopaque = HashPageGetOpaque(opage);
LockBuffer(bucket_nbuf, BUFFER_LOCK_EXCLUSIVE);
npage = BufferGetPage(bucket_nbuf);
- nopaque = (HashPageOpaque) PageGetSpecialPointer(npage);
+ nopaque = HashPageGetOpaque(npage);
START_CRIT_SECTION();
@@ -1392,7 +1392,7 @@ _hash_finish_split(Relation rel, Buffer metabuf, Buffer obuf, Bucket obucket,
bucket_nbuf = nbuf;
npage = BufferGetPage(nbuf);
- npageopaque = (HashPageOpaque) PageGetSpecialPointer(npage);
+ npageopaque = HashPageGetOpaque(npage);
/* Scan each tuple in new page */
nmaxoffnum = PageGetMaxOffsetNumber(npage);
@@ -1446,7 +1446,7 @@ _hash_finish_split(Relation rel, Buffer metabuf, Buffer obuf, Bucket obucket,
}
npage = BufferGetPage(bucket_nbuf);
- npageopaque = (HashPageOpaque) PageGetSpecialPointer(npage);
+ npageopaque = HashPageGetOpaque(npage);
nbucket = npageopaque->hasho_bucket;
_hash_splitbucket(rel, metabuf, obucket,
@@ -1587,7 +1587,7 @@ _hash_getbucketbuf_from_hashkey(Relation rel, uint32 hashkey, int access,
/* Fetch the primary bucket page for the bucket */
buf = _hash_getbuf(rel, blkno, access, LH_BUCKET_PAGE);
page = BufferGetPage(buf);
- opaque = (HashPageOpaque) PageGetSpecialPointer(page);
+ opaque = HashPageGetOpaque(page);
Assert(opaque->hasho_bucket == bucket);
Assert(opaque->hasho_prevblkno != InvalidBlockNumber);
diff --git a/src/backend/access/hash/hashsearch.c b/src/backend/access/hash/hashsearch.c
index 7ca542a3fb..524af27409 100644
--- a/src/backend/access/hash/hashsearch.c
+++ b/src/backend/access/hash/hashsearch.c
@@ -187,7 +187,7 @@ _hash_readnext(IndexScanDesc scan,
{
*pagep = BufferGetPage(*bufp);
TestForOldSnapshot(scan->xs_snapshot, rel, *pagep);
- *opaquep = (HashPageOpaque) PageGetSpecialPointer(*pagep);
+ *opaquep = HashPageGetOpaque(*pagep);
}
}
@@ -233,7 +233,7 @@ _hash_readprev(IndexScanDesc scan,
LH_BUCKET_PAGE | LH_OVERFLOW_PAGE);
*pagep = BufferGetPage(*bufp);
TestForOldSnapshot(scan->xs_snapshot, rel, *pagep);
- *opaquep = (HashPageOpaque) PageGetSpecialPointer(*pagep);
+ *opaquep = HashPageGetOpaque(*pagep);
/*
* We always maintain the pin on bucket page for whole scan operation,
@@ -258,7 +258,7 @@ _hash_readprev(IndexScanDesc scan,
LockBuffer(*bufp, BUFFER_LOCK_SHARE);
*pagep = BufferGetPage(*bufp);
- *opaquep = (HashPageOpaque) PageGetSpecialPointer(*pagep);
+ *opaquep = HashPageGetOpaque(*pagep);
/* move to the end of bucket chain */
while (BlockNumberIsValid((*opaquep)->hasho_nextblkno))
@@ -352,7 +352,7 @@ _hash_first(IndexScanDesc scan, ScanDirection dir)
PredicateLockPage(rel, BufferGetBlockNumber(buf), scan->xs_snapshot);
page = BufferGetPage(buf);
TestForOldSnapshot(scan->xs_snapshot, rel, page);
- opaque = (HashPageOpaque) PageGetSpecialPointer(page);
+ opaque = HashPageGetOpaque(page);
bucket = opaque->hasho_bucket;
so->hashso_bucket_buf = buf;
@@ -398,7 +398,7 @@ _hash_first(IndexScanDesc scan, ScanDirection dir)
LockBuffer(buf, BUFFER_LOCK_SHARE);
page = BufferGetPage(buf);
- opaque = (HashPageOpaque) PageGetSpecialPointer(page);
+ opaque = HashPageGetOpaque(page);
Assert(opaque->hasho_bucket == bucket);
if (H_BUCKET_BEING_POPULATED(opaque))
@@ -463,7 +463,7 @@ _hash_readpage(IndexScanDesc scan, Buffer *bufP, ScanDirection dir)
Assert(BufferIsValid(buf));
_hash_checkpage(rel, buf, LH_BUCKET_PAGE | LH_OVERFLOW_PAGE);
page = BufferGetPage(buf);
- opaque = (HashPageOpaque) PageGetSpecialPointer(page);
+ opaque = HashPageGetOpaque(page);
so->currPos.buf = buf;
so->currPos.currPage = BufferGetBlockNumber(buf);
diff --git a/src/backend/access/hash/hashutil.c b/src/backend/access/hash/hashutil.c
index edb6fa968f..fe37bc47cb 100644
--- a/src/backend/access/hash/hashutil.c
+++ b/src/backend/access/hash/hashutil.c
@@ -239,7 +239,7 @@ _hash_checkpage(Relation rel, Buffer buf, int flags)
if (flags)
{
- HashPageOpaque opaque = (HashPageOpaque) PageGetSpecialPointer(page);
+ HashPageOpaque opaque = HashPageGetOpaque(page);
if ((opaque->hasho_flag & flags) == 0)
ereport(ERROR,
@@ -574,7 +574,7 @@ _hash_kill_items(IndexScanDesc scan)
buf = _hash_getbuf(rel, blkno, HASH_READ, LH_OVERFLOW_PAGE);
page = BufferGetPage(buf);
- opaque = (HashPageOpaque) PageGetSpecialPointer(page);
+ opaque = HashPageGetOpaque(page);
maxoff = PageGetMaxOffsetNumber(page);
for (i = 0; i < numKilled; i++)
diff --git a/src/include/access/hash.h b/src/include/access/hash.h
index cd7b2a53d8..d3fd8bf239 100644
--- a/src/include/access/hash.h
+++ b/src/include/access/hash.h
@@ -85,6 +85,9 @@ typedef struct HashPageOpaqueData
typedef HashPageOpaqueData *HashPageOpaque;
+#define HashPageGetOpaque(page) \
+ ((HashPageOpaque) PageGetSpecialOpaque((page), HashPageOpaqueData))
+
#define H_NEEDS_SPLIT_CLEANUP(opaque) (((opaque)->hasho_flag & LH_BUCKET_NEEDS_SPLIT_CLEANUP) != 0)
#define H_BUCKET_BEING_SPLIT(opaque) (((opaque)->hasho_flag & LH_BUCKET_BEING_SPLIT) != 0)
#define H_BUCKET_BEING_POPULATED(opaque) (((opaque)->hasho_flag & LH_BUCKET_BEING_POPULATED) != 0)
--
2.30.2
Re: Preventing indirection for IndexPageGetOpaque for known-size page special areas
On Mon, Mar 28, 2022 at 05:09:10PM +0200, Matthias van de Meent wrote:
Not all clusters have checksums enabled (boo!, but we can't
realistically fix that), so on-disk corruption could reasonably
propagate to the rest of such system. Additionally, checksums are only
checked on read, and updated when the page is written to disk (in
PageSetChecksumCopy called from FlushBuffer), but this does not check
for signs of page corruption. As such, a memory bug resulting in
in-memory corruption of pd_special would persist and would currently
have the potential to further corrupt neighbouring buffers.
Well, if it comes to corruption then pd_special is not the only
problem, just one part of it. A broken pd_lower or pd_lower or even
pd_lsn could also cause AMs to point at areas they should not. There
are many reasons that could make things go wrong depending on the
compiled page size.
A second reason would be less indirection to get to the opaque
pointer. This should improve performance a bit in those cases where we
(initially) only want to access the [Index]PageOpaque struct.We don't have many cases that do that, do we?
Indeed not many, but I know that at least the nbtree-code has some
cases where it uses the opaque before other fields in the header are
accessed (sometimes even without accessing the header for other
reasons): in _bt_moveright and _bt_walk_left. There might be more; but
these are cases I know of.
Even these are marginal as the page is already loaded in the shared
buffers..
While looking at the page, the pieces in 0002 and 0003 that I really
liked are the introduction of the macros to grab the special area for
btree and hash. This brings the code of those APIs closer to GiST,
SpGiST and GIN. And it is possible to move to a possible change in
the checks and/or the shape of all the *GetOpaque macros at once after
the switch to this part is done. So I don't really mind introducing
this part.
PageGetSpecialOpaque() would not have saved the day with the recent
pageinspect changes, and that's just moving the responsability of the
page header to something else. I am not sure if it is a good idea to
have a mention of the opaque page type in bufpage.h actually, as this
is optional. Having an AssertMacro() checking that pd_special is in
line with MAXALIGN(sizeof(OpaqueData)) is attractive, but I'd rather
keep that in each AM's headers per its optional nature, and an index
AM may handle things differently depending on a page type, as well.
--
Michael
Attachments:
0001-Introduce-macros-for-hash-and-btree-to-grab-their-op.patchtext/x-diff; charset=us-asciiDownload
From e12d282461c9e4a1cba89cb7ea8e6dde90c829a6 Mon Sep 17 00:00:00 2001
From: Michael Paquier <michael@paquier.xyz>
Date: Thu, 31 Mar 2022 16:30:52 +0900
Subject: [PATCH] Introduce macros for hash and btree to grab their opaque area
---
src/include/access/hash.h | 2 +
src/include/access/nbtree.h | 8 ++--
src/backend/access/hash/hash.c | 6 +--
src/backend/access/hash/hash_xlog.c | 26 +++++------
src/backend/access/hash/hashinsert.c | 6 +--
src/backend/access/hash/hashovfl.c | 22 +++++-----
src/backend/access/hash/hashpage.c | 30 ++++++-------
src/backend/access/hash/hashsearch.c | 12 ++---
src/backend/access/hash/hashutil.c | 4 +-
src/backend/access/nbtree/nbtdedup.c | 6 +--
src/backend/access/nbtree/nbtinsert.c | 46 ++++++++++----------
src/backend/access/nbtree/nbtpage.c | 58 ++++++++++++-------------
src/backend/access/nbtree/nbtree.c | 2 +-
src/backend/access/nbtree/nbtsearch.c | 34 +++++++--------
src/backend/access/nbtree/nbtsort.c | 12 ++---
src/backend/access/nbtree/nbtsplitloc.c | 2 +-
src/backend/access/nbtree/nbtutils.c | 6 +--
src/backend/access/nbtree/nbtxlog.c | 34 +++++++--------
contrib/amcheck/verify_nbtree.c | 32 +++++++-------
contrib/pageinspect/btreefuncs.c | 6 +--
contrib/pageinspect/hashfuncs.c | 6 +--
contrib/pgstattuple/pgstatindex.c | 4 +-
contrib/pgstattuple/pgstattuple.c | 4 +-
23 files changed, 186 insertions(+), 182 deletions(-)
diff --git a/src/include/access/hash.h b/src/include/access/hash.h
index cd7b2a53d8..a22e328762 100644
--- a/src/include/access/hash.h
+++ b/src/include/access/hash.h
@@ -85,6 +85,8 @@ typedef struct HashPageOpaqueData
typedef HashPageOpaqueData *HashPageOpaque;
+#define HashPageGetOpaque(page) ( (HashPageOpaque) PageGetSpecialPointer(page) )
+
#define H_NEEDS_SPLIT_CLEANUP(opaque) (((opaque)->hasho_flag & LH_BUCKET_NEEDS_SPLIT_CLEANUP) != 0)
#define H_BUCKET_BEING_SPLIT(opaque) (((opaque)->hasho_flag & LH_BUCKET_BEING_SPLIT) != 0)
#define H_BUCKET_BEING_POPULATED(opaque) (((opaque)->hasho_flag & LH_BUCKET_BEING_POPULATED) != 0)
diff --git a/src/include/access/nbtree.h b/src/include/access/nbtree.h
index 9fec6fb1a8..5625a90922 100644
--- a/src/include/access/nbtree.h
+++ b/src/include/access/nbtree.h
@@ -70,6 +70,8 @@ typedef struct BTPageOpaqueData
typedef BTPageOpaqueData *BTPageOpaque;
+#define BTPageGetOpaque(page) ( (BTPageOpaque) PageGetSpecialPointer(page) )
+
/* Bits defined in btpo_flags */
#define BTP_LEAF (1 << 0) /* leaf page, i.e. not internal page */
#define BTP_ROOT (1 << 1) /* root page (has no parent) */
@@ -241,7 +243,7 @@ BTPageSetDeleted(Page page, FullTransactionId safexid)
PageHeader header;
BTDeletedPageData *contents;
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
header = ((PageHeader) page);
opaque->btpo_flags &= ~BTP_HALF_DEAD;
@@ -263,7 +265,7 @@ BTPageGetDeleteXid(Page page)
/* We only expect to be called with a deleted page */
Assert(!PageIsNew(page));
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
Assert(P_ISDELETED(opaque));
/* pg_upgrade'd deleted page -- must be safe to delete now */
@@ -294,7 +296,7 @@ BTPageIsRecyclable(Page page)
Assert(!PageIsNew(page));
/* Recycling okay iff page is deleted and safexid is old enough */
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
if (P_ISDELETED(opaque))
{
/*
diff --git a/src/backend/access/hash/hash.c b/src/backend/access/hash/hash.c
index a259a301fa..fd1a7119b6 100644
--- a/src/backend/access/hash/hash.c
+++ b/src/backend/access/hash/hash.c
@@ -515,7 +515,7 @@ loop_top:
_hash_checkpage(rel, buf, LH_BUCKET_PAGE);
page = BufferGetPage(buf);
- bucket_opaque = (HashPageOpaque) PageGetSpecialPointer(page);
+ bucket_opaque = HashPageGetOpaque(page);
/*
* If the bucket contains tuples that are moved by split, then we need
@@ -717,7 +717,7 @@ hashbucketcleanup(Relation rel, Bucket cur_bucket, Buffer bucket_buf,
vacuum_delay_point();
page = BufferGetPage(buf);
- opaque = (HashPageOpaque) PageGetSpecialPointer(page);
+ opaque = HashPageGetOpaque(page);
/* Scan each tuple in page */
maxoffno = PageGetMaxOffsetNumber(page);
@@ -884,7 +884,7 @@ hashbucketcleanup(Relation rel, Bucket cur_bucket, Buffer bucket_buf,
Page page;
page = BufferGetPage(bucket_buf);
- bucket_opaque = (HashPageOpaque) PageGetSpecialPointer(page);
+ bucket_opaque = HashPageGetOpaque(page);
/* No ereport(ERROR) until changes are logged */
START_CRIT_SECTION();
diff --git a/src/backend/access/hash/hash_xlog.c b/src/backend/access/hash/hash_xlog.c
index 55937b9a68..62dbfc3a5d 100644
--- a/src/backend/access/hash/hash_xlog.c
+++ b/src/backend/access/hash/hash_xlog.c
@@ -203,7 +203,7 @@ hash_xlog_add_ovfl_page(XLogReaderState *record)
true);
/* update backlink */
ovflpage = BufferGetPage(ovflbuf);
- ovflopaque = (HashPageOpaque) PageGetSpecialPointer(ovflpage);
+ ovflopaque = HashPageGetOpaque(ovflpage);
ovflopaque->hasho_prevblkno = leftblk;
PageSetLSN(ovflpage, lsn);
@@ -215,7 +215,7 @@ hash_xlog_add_ovfl_page(XLogReaderState *record)
HashPageOpaque leftopaque;
leftpage = BufferGetPage(leftbuf);
- leftopaque = (HashPageOpaque) PageGetSpecialPointer(leftpage);
+ leftopaque = HashPageGetOpaque(leftpage);
leftopaque->hasho_nextblkno = rightblk;
PageSetLSN(leftpage, lsn);
@@ -342,7 +342,7 @@ hash_xlog_split_allocate_page(XLogReaderState *record)
HashPageOpaque oldopaque;
oldpage = BufferGetPage(oldbuf);
- oldopaque = (HashPageOpaque) PageGetSpecialPointer(oldpage);
+ oldopaque = HashPageGetOpaque(oldpage);
oldopaque->hasho_flag = xlrec->old_bucket_flag;
oldopaque->hasho_prevblkno = xlrec->new_bucket;
@@ -465,7 +465,7 @@ hash_xlog_split_complete(XLogReaderState *record)
HashPageOpaque oldopaque;
oldpage = BufferGetPage(oldbuf);
- oldopaque = (HashPageOpaque) PageGetSpecialPointer(oldpage);
+ oldopaque = HashPageGetOpaque(oldpage);
oldopaque->hasho_flag = xlrec->old_bucket_flag;
@@ -488,7 +488,7 @@ hash_xlog_split_complete(XLogReaderState *record)
HashPageOpaque nopaque;
newpage = BufferGetPage(newbuf);
- nopaque = (HashPageOpaque) PageGetSpecialPointer(newpage);
+ nopaque = HashPageGetOpaque(newpage);
nopaque->hasho_flag = xlrec->new_bucket_flag;
@@ -710,7 +710,7 @@ hash_xlog_squeeze_page(XLogReaderState *record)
*/
if (xldata->is_prev_bucket_same_wrt)
{
- HashPageOpaque writeopaque = (HashPageOpaque) PageGetSpecialPointer(writepage);
+ HashPageOpaque writeopaque = HashPageGetOpaque(writepage);
writeopaque->hasho_nextblkno = xldata->nextblkno;
}
@@ -729,7 +729,7 @@ hash_xlog_squeeze_page(XLogReaderState *record)
_hash_pageinit(ovflpage, BufferGetPageSize(ovflbuf));
- ovflopaque = (HashPageOpaque) PageGetSpecialPointer(ovflpage);
+ ovflopaque = HashPageGetOpaque(ovflpage);
ovflopaque->hasho_prevblkno = InvalidBlockNumber;
ovflopaque->hasho_nextblkno = InvalidBlockNumber;
@@ -748,7 +748,7 @@ hash_xlog_squeeze_page(XLogReaderState *record)
XLogReadBufferForRedo(record, 3, &prevbuf) == BLK_NEEDS_REDO)
{
Page prevpage = BufferGetPage(prevbuf);
- HashPageOpaque prevopaque = (HashPageOpaque) PageGetSpecialPointer(prevpage);
+ HashPageOpaque prevopaque = HashPageGetOpaque(prevpage);
prevopaque->hasho_nextblkno = xldata->nextblkno;
@@ -766,7 +766,7 @@ hash_xlog_squeeze_page(XLogReaderState *record)
if (XLogReadBufferForRedo(record, 4, &nextbuf) == BLK_NEEDS_REDO)
{
Page nextpage = BufferGetPage(nextbuf);
- HashPageOpaque nextopaque = (HashPageOpaque) PageGetSpecialPointer(nextpage);
+ HashPageOpaque nextopaque = HashPageGetOpaque(nextpage);
nextopaque->hasho_prevblkno = xldata->prevblkno;
@@ -903,7 +903,7 @@ hash_xlog_delete(XLogReaderState *record)
{
HashPageOpaque pageopaque;
- pageopaque = (HashPageOpaque) PageGetSpecialPointer(page);
+ pageopaque = HashPageGetOpaque(page);
pageopaque->hasho_flag &= ~LH_PAGE_HAS_DEAD_TUPLES;
}
@@ -933,7 +933,7 @@ hash_xlog_split_cleanup(XLogReaderState *record)
page = (Page) BufferGetPage(buffer);
- bucket_opaque = (HashPageOpaque) PageGetSpecialPointer(page);
+ bucket_opaque = HashPageGetOpaque(page);
bucket_opaque->hasho_flag &= ~LH_BUCKET_NEEDS_SPLIT_CLEANUP;
PageSetLSN(page, lsn);
MarkBufferDirty(buffer);
@@ -1024,7 +1024,7 @@ hash_xlog_vacuum_one_page(XLogReaderState *record)
* Mark the page as not containing any LP_DEAD items. See comments in
* _hash_vacuum_one_page() for details.
*/
- pageopaque = (HashPageOpaque) PageGetSpecialPointer(page);
+ pageopaque = HashPageGetOpaque(page);
pageopaque->hasho_flag &= ~LH_PAGE_HAS_DEAD_TUPLES;
PageSetLSN(page, lsn);
@@ -1116,7 +1116,7 @@ hash_mask(char *pagedata, BlockNumber blkno)
mask_page_hint_bits(page);
mask_unused_space(page);
- opaque = (HashPageOpaque) PageGetSpecialPointer(page);
+ opaque = HashPageGetOpaque(page);
pagetype = opaque->hasho_flag & LH_PAGE_TYPE;
if (pagetype == LH_UNUSED_PAGE)
diff --git a/src/backend/access/hash/hashinsert.c b/src/backend/access/hash/hashinsert.c
index faf609c157..4f2fecb908 100644
--- a/src/backend/access/hash/hashinsert.c
+++ b/src/backend/access/hash/hashinsert.c
@@ -95,7 +95,7 @@ restart_insert:
bucket_buf = buf;
page = BufferGetPage(buf);
- pageopaque = (HashPageOpaque) PageGetSpecialPointer(page);
+ pageopaque = HashPageGetOpaque(page);
bucket = pageopaque->hasho_bucket;
/*
@@ -183,7 +183,7 @@ restart_insert:
/* should fit now, given test above */
Assert(PageGetFreeSpace(page) >= itemsz);
}
- pageopaque = (HashPageOpaque) PageGetSpecialPointer(page);
+ pageopaque = HashPageGetOpaque(page);
Assert((pageopaque->hasho_flag & LH_PAGE_TYPE) == LH_OVERFLOW_PAGE);
Assert(pageopaque->hasho_bucket == bucket);
}
@@ -384,7 +384,7 @@ _hash_vacuum_one_page(Relation rel, Relation hrel, Buffer metabuf, Buffer buf)
* check it. Remember that LH_PAGE_HAS_DEAD_TUPLES is only a hint
* anyway.
*/
- pageopaque = (HashPageOpaque) PageGetSpecialPointer(page);
+ pageopaque = HashPageGetOpaque(page);
pageopaque->hasho_flag &= ~LH_PAGE_HAS_DEAD_TUPLES;
metap = HashPageGetMeta(BufferGetPage(metabuf));
diff --git a/src/backend/access/hash/hashovfl.c b/src/backend/access/hash/hashovfl.c
index 4836875196..e34cfc302d 100644
--- a/src/backend/access/hash/hashovfl.c
+++ b/src/backend/access/hash/hashovfl.c
@@ -159,7 +159,7 @@ _hash_addovflpage(Relation rel, Buffer metabuf, Buffer buf, bool retain_pin)
BlockNumber nextblkno;
page = BufferGetPage(buf);
- pageopaque = (HashPageOpaque) PageGetSpecialPointer(page);
+ pageopaque = HashPageGetOpaque(page);
nextblkno = pageopaque->hasho_nextblkno;
if (!BlockNumberIsValid(nextblkno))
@@ -364,7 +364,7 @@ found:
/* initialize new overflow page */
ovflpage = BufferGetPage(ovflbuf);
- ovflopaque = (HashPageOpaque) PageGetSpecialPointer(ovflpage);
+ ovflopaque = HashPageGetOpaque(ovflpage);
ovflopaque->hasho_prevblkno = BufferGetBlockNumber(buf);
ovflopaque->hasho_nextblkno = InvalidBlockNumber;
ovflopaque->hasho_bucket = pageopaque->hasho_bucket;
@@ -516,7 +516,7 @@ _hash_freeovflpage(Relation rel, Buffer bucketbuf, Buffer ovflbuf,
_hash_checkpage(rel, ovflbuf, LH_OVERFLOW_PAGE);
ovflblkno = BufferGetBlockNumber(ovflbuf);
ovflpage = BufferGetPage(ovflbuf);
- ovflopaque = (HashPageOpaque) PageGetSpecialPointer(ovflpage);
+ ovflopaque = HashPageGetOpaque(ovflpage);
nextblkno = ovflopaque->hasho_nextblkno;
prevblkno = ovflopaque->hasho_prevblkno;
writeblkno = BufferGetBlockNumber(wbuf);
@@ -600,7 +600,7 @@ _hash_freeovflpage(Relation rel, Buffer bucketbuf, Buffer ovflbuf,
*/
_hash_pageinit(ovflpage, BufferGetPageSize(ovflbuf));
- ovflopaque = (HashPageOpaque) PageGetSpecialPointer(ovflpage);
+ ovflopaque = HashPageGetOpaque(ovflpage);
ovflopaque->hasho_prevblkno = InvalidBlockNumber;
ovflopaque->hasho_nextblkno = InvalidBlockNumber;
@@ -613,7 +613,7 @@ _hash_freeovflpage(Relation rel, Buffer bucketbuf, Buffer ovflbuf,
if (BufferIsValid(prevbuf))
{
Page prevpage = BufferGetPage(prevbuf);
- HashPageOpaque prevopaque = (HashPageOpaque) PageGetSpecialPointer(prevpage);
+ HashPageOpaque prevopaque = HashPageGetOpaque(prevpage);
Assert(prevopaque->hasho_bucket == bucket);
prevopaque->hasho_nextblkno = nextblkno;
@@ -622,7 +622,7 @@ _hash_freeovflpage(Relation rel, Buffer bucketbuf, Buffer ovflbuf,
if (BufferIsValid(nextbuf))
{
Page nextpage = BufferGetPage(nextbuf);
- HashPageOpaque nextopaque = (HashPageOpaque) PageGetSpecialPointer(nextpage);
+ HashPageOpaque nextopaque = HashPageGetOpaque(nextpage);
Assert(nextopaque->hasho_bucket == bucket);
nextopaque->hasho_prevblkno = prevblkno;
@@ -751,7 +751,7 @@ _hash_initbitmapbuffer(Buffer buf, uint16 bmsize, bool initpage)
_hash_pageinit(pg, BufferGetPageSize(buf));
/* initialize the page's special space */
- op = (HashPageOpaque) PageGetSpecialPointer(pg);
+ op = HashPageGetOpaque(pg);
op->hasho_prevblkno = InvalidBlockNumber;
op->hasho_nextblkno = InvalidBlockNumber;
op->hasho_bucket = InvalidBucket;
@@ -824,7 +824,7 @@ _hash_squeezebucket(Relation rel,
wblkno = bucket_blkno;
wbuf = bucket_buf;
wpage = BufferGetPage(wbuf);
- wopaque = (HashPageOpaque) PageGetSpecialPointer(wpage);
+ wopaque = HashPageGetOpaque(wpage);
/*
* if there aren't any overflow pages, there's nothing to squeeze. caller
@@ -855,7 +855,7 @@ _hash_squeezebucket(Relation rel,
LH_OVERFLOW_PAGE,
bstrategy);
rpage = BufferGetPage(rbuf);
- ropaque = (HashPageOpaque) PageGetSpecialPointer(rpage);
+ ropaque = HashPageGetOpaque(rpage);
Assert(ropaque->hasho_bucket == bucket);
} while (BlockNumberIsValid(ropaque->hasho_nextblkno));
@@ -1005,7 +1005,7 @@ readpage:
wbuf = next_wbuf;
wpage = BufferGetPage(wbuf);
- wopaque = (HashPageOpaque) PageGetSpecialPointer(wpage);
+ wopaque = HashPageGetOpaque(wpage);
Assert(wopaque->hasho_bucket == bucket);
retain_pin = false;
@@ -1076,7 +1076,7 @@ readpage:
LH_OVERFLOW_PAGE,
bstrategy);
rpage = BufferGetPage(rbuf);
- ropaque = (HashPageOpaque) PageGetSpecialPointer(rpage);
+ ropaque = HashPageGetOpaque(rpage);
Assert(ropaque->hasho_bucket == bucket);
}
diff --git a/src/backend/access/hash/hashpage.c b/src/backend/access/hash/hashpage.c
index 28c5297a1d..39206d1942 100644
--- a/src/backend/access/hash/hashpage.c
+++ b/src/backend/access/hash/hashpage.c
@@ -166,7 +166,7 @@ _hash_initbuf(Buffer buf, uint32 max_bucket, uint32 num_bucket, uint32 flag,
if (initpage)
_hash_pageinit(page, BufferGetPageSize(buf));
- pageopaque = (HashPageOpaque) PageGetSpecialPointer(page);
+ pageopaque = HashPageGetOpaque(page);
/*
* Set hasho_prevblkno with current hashm_maxbucket. This value will be
@@ -529,7 +529,7 @@ _hash_init_metabuffer(Buffer buf, double num_tuples, RegProcedure procid,
if (initpage)
_hash_pageinit(page, BufferGetPageSize(buf));
- pageopaque = (HashPageOpaque) PageGetSpecialPointer(page);
+ pageopaque = HashPageGetOpaque(page);
pageopaque->hasho_prevblkno = InvalidBlockNumber;
pageopaque->hasho_nextblkno = InvalidBlockNumber;
pageopaque->hasho_bucket = InvalidBucket;
@@ -693,7 +693,7 @@ restart_expand:
goto fail;
opage = BufferGetPage(buf_oblkno);
- oopaque = (HashPageOpaque) PageGetSpecialPointer(opage);
+ oopaque = HashPageGetOpaque(opage);
/*
* We want to finish the split from a bucket as there is no apparent
@@ -864,7 +864,7 @@ restart_expand:
lowmask = metap->hashm_lowmask;
opage = BufferGetPage(buf_oblkno);
- oopaque = (HashPageOpaque) PageGetSpecialPointer(opage);
+ oopaque = HashPageGetOpaque(opage);
/*
* Mark the old bucket to indicate that split is in progress. (At
@@ -883,7 +883,7 @@ restart_expand:
* initialize the new bucket's primary page and mark it to indicate that
* split is in progress.
*/
- nopaque = (HashPageOpaque) PageGetSpecialPointer(npage);
+ nopaque = HashPageGetOpaque(npage);
nopaque->hasho_prevblkno = maxbucket;
nopaque->hasho_nextblkno = InvalidBlockNumber;
nopaque->hasho_bucket = new_bucket;
@@ -1010,7 +1010,7 @@ _hash_alloc_buckets(Relation rel, BlockNumber firstblock, uint32 nblocks)
*/
_hash_pageinit(page, BLCKSZ);
- ovflopaque = (HashPageOpaque) PageGetSpecialPointer(page);
+ ovflopaque = HashPageGetOpaque(page);
ovflopaque->hasho_prevblkno = InvalidBlockNumber;
ovflopaque->hasho_nextblkno = InvalidBlockNumber;
@@ -1091,11 +1091,11 @@ _hash_splitbucket(Relation rel,
bucket_obuf = obuf;
opage = BufferGetPage(obuf);
- oopaque = (HashPageOpaque) PageGetSpecialPointer(opage);
+ oopaque = HashPageGetOpaque(opage);
bucket_nbuf = nbuf;
npage = BufferGetPage(nbuf);
- nopaque = (HashPageOpaque) PageGetSpecialPointer(npage);
+ nopaque = HashPageGetOpaque(npage);
/* Copy the predicate locks from old bucket to new bucket. */
PredicateLockPageSplit(rel,
@@ -1198,7 +1198,7 @@ _hash_splitbucket(Relation rel,
/* chain to a new overflow page */
nbuf = _hash_addovflpage(rel, metabuf, nbuf, (nbuf == bucket_nbuf));
npage = BufferGetPage(nbuf);
- nopaque = (HashPageOpaque) PageGetSpecialPointer(npage);
+ nopaque = HashPageGetOpaque(npage);
}
itups[nitups++] = new_itup;
@@ -1251,7 +1251,7 @@ _hash_splitbucket(Relation rel,
/* Else, advance to next old page */
obuf = _hash_getbuf(rel, oblkno, HASH_READ, LH_OVERFLOW_PAGE);
opage = BufferGetPage(obuf);
- oopaque = (HashPageOpaque) PageGetSpecialPointer(opage);
+ oopaque = HashPageGetOpaque(opage);
}
/*
@@ -1264,11 +1264,11 @@ _hash_splitbucket(Relation rel,
*/
LockBuffer(bucket_obuf, BUFFER_LOCK_EXCLUSIVE);
opage = BufferGetPage(bucket_obuf);
- oopaque = (HashPageOpaque) PageGetSpecialPointer(opage);
+ oopaque = HashPageGetOpaque(opage);
LockBuffer(bucket_nbuf, BUFFER_LOCK_EXCLUSIVE);
npage = BufferGetPage(bucket_nbuf);
- nopaque = (HashPageOpaque) PageGetSpecialPointer(npage);
+ nopaque = HashPageGetOpaque(npage);
START_CRIT_SECTION();
@@ -1392,7 +1392,7 @@ _hash_finish_split(Relation rel, Buffer metabuf, Buffer obuf, Bucket obucket,
bucket_nbuf = nbuf;
npage = BufferGetPage(nbuf);
- npageopaque = (HashPageOpaque) PageGetSpecialPointer(npage);
+ npageopaque = HashPageGetOpaque(npage);
/* Scan each tuple in new page */
nmaxoffnum = PageGetMaxOffsetNumber(npage);
@@ -1446,7 +1446,7 @@ _hash_finish_split(Relation rel, Buffer metabuf, Buffer obuf, Bucket obucket,
}
npage = BufferGetPage(bucket_nbuf);
- npageopaque = (HashPageOpaque) PageGetSpecialPointer(npage);
+ npageopaque = HashPageGetOpaque(npage);
nbucket = npageopaque->hasho_bucket;
_hash_splitbucket(rel, metabuf, obucket,
@@ -1587,7 +1587,7 @@ _hash_getbucketbuf_from_hashkey(Relation rel, uint32 hashkey, int access,
/* Fetch the primary bucket page for the bucket */
buf = _hash_getbuf(rel, blkno, access, LH_BUCKET_PAGE);
page = BufferGetPage(buf);
- opaque = (HashPageOpaque) PageGetSpecialPointer(page);
+ opaque = HashPageGetOpaque(page);
Assert(opaque->hasho_bucket == bucket);
Assert(opaque->hasho_prevblkno != InvalidBlockNumber);
diff --git a/src/backend/access/hash/hashsearch.c b/src/backend/access/hash/hashsearch.c
index 7ca542a3fb..524af27409 100644
--- a/src/backend/access/hash/hashsearch.c
+++ b/src/backend/access/hash/hashsearch.c
@@ -187,7 +187,7 @@ _hash_readnext(IndexScanDesc scan,
{
*pagep = BufferGetPage(*bufp);
TestForOldSnapshot(scan->xs_snapshot, rel, *pagep);
- *opaquep = (HashPageOpaque) PageGetSpecialPointer(*pagep);
+ *opaquep = HashPageGetOpaque(*pagep);
}
}
@@ -233,7 +233,7 @@ _hash_readprev(IndexScanDesc scan,
LH_BUCKET_PAGE | LH_OVERFLOW_PAGE);
*pagep = BufferGetPage(*bufp);
TestForOldSnapshot(scan->xs_snapshot, rel, *pagep);
- *opaquep = (HashPageOpaque) PageGetSpecialPointer(*pagep);
+ *opaquep = HashPageGetOpaque(*pagep);
/*
* We always maintain the pin on bucket page for whole scan operation,
@@ -258,7 +258,7 @@ _hash_readprev(IndexScanDesc scan,
LockBuffer(*bufp, BUFFER_LOCK_SHARE);
*pagep = BufferGetPage(*bufp);
- *opaquep = (HashPageOpaque) PageGetSpecialPointer(*pagep);
+ *opaquep = HashPageGetOpaque(*pagep);
/* move to the end of bucket chain */
while (BlockNumberIsValid((*opaquep)->hasho_nextblkno))
@@ -352,7 +352,7 @@ _hash_first(IndexScanDesc scan, ScanDirection dir)
PredicateLockPage(rel, BufferGetBlockNumber(buf), scan->xs_snapshot);
page = BufferGetPage(buf);
TestForOldSnapshot(scan->xs_snapshot, rel, page);
- opaque = (HashPageOpaque) PageGetSpecialPointer(page);
+ opaque = HashPageGetOpaque(page);
bucket = opaque->hasho_bucket;
so->hashso_bucket_buf = buf;
@@ -398,7 +398,7 @@ _hash_first(IndexScanDesc scan, ScanDirection dir)
LockBuffer(buf, BUFFER_LOCK_SHARE);
page = BufferGetPage(buf);
- opaque = (HashPageOpaque) PageGetSpecialPointer(page);
+ opaque = HashPageGetOpaque(page);
Assert(opaque->hasho_bucket == bucket);
if (H_BUCKET_BEING_POPULATED(opaque))
@@ -463,7 +463,7 @@ _hash_readpage(IndexScanDesc scan, Buffer *bufP, ScanDirection dir)
Assert(BufferIsValid(buf));
_hash_checkpage(rel, buf, LH_BUCKET_PAGE | LH_OVERFLOW_PAGE);
page = BufferGetPage(buf);
- opaque = (HashPageOpaque) PageGetSpecialPointer(page);
+ opaque = HashPageGetOpaque(page);
so->currPos.buf = buf;
so->currPos.currPage = BufferGetBlockNumber(buf);
diff --git a/src/backend/access/hash/hashutil.c b/src/backend/access/hash/hashutil.c
index edb6fa968f..fe37bc47cb 100644
--- a/src/backend/access/hash/hashutil.c
+++ b/src/backend/access/hash/hashutil.c
@@ -239,7 +239,7 @@ _hash_checkpage(Relation rel, Buffer buf, int flags)
if (flags)
{
- HashPageOpaque opaque = (HashPageOpaque) PageGetSpecialPointer(page);
+ HashPageOpaque opaque = HashPageGetOpaque(page);
if ((opaque->hasho_flag & flags) == 0)
ereport(ERROR,
@@ -574,7 +574,7 @@ _hash_kill_items(IndexScanDesc scan)
buf = _hash_getbuf(rel, blkno, HASH_READ, LH_OVERFLOW_PAGE);
page = BufferGetPage(buf);
- opaque = (HashPageOpaque) PageGetSpecialPointer(page);
+ opaque = HashPageGetOpaque(page);
maxoff = PageGetMaxOffsetNumber(page);
for (i = 0; i < numKilled; i++)
diff --git a/src/backend/access/nbtree/nbtdedup.c b/src/backend/access/nbtree/nbtdedup.c
index 4c48554aec..3e11805293 100644
--- a/src/backend/access/nbtree/nbtdedup.c
+++ b/src/backend/access/nbtree/nbtdedup.c
@@ -62,7 +62,7 @@ _bt_dedup_pass(Relation rel, Buffer buf, Relation heapRel, IndexTuple newitem,
minoff,
maxoff;
Page page = BufferGetPage(buf);
- BTPageOpaque opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ BTPageOpaque opaque = BTPageGetOpaque(page);
Page newpage;
BTDedupState state;
Size pagesaving = 0;
@@ -231,7 +231,7 @@ _bt_dedup_pass(Relation rel, Buffer buf, Relation heapRel, IndexTuple newitem,
*/
if (P_HAS_GARBAGE(opaque))
{
- BTPageOpaque nopaque = (BTPageOpaque) PageGetSpecialPointer(newpage);
+ BTPageOpaque nopaque = BTPageGetOpaque(newpage);
nopaque->btpo_flags &= ~BTP_HAS_GARBAGE;
}
@@ -310,7 +310,7 @@ _bt_bottomupdel_pass(Relation rel, Buffer buf, Relation heapRel,
minoff,
maxoff;
Page page = BufferGetPage(buf);
- BTPageOpaque opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ BTPageOpaque opaque = BTPageGetOpaque(page);
BTDedupState state;
TM_IndexDeleteOp delstate;
bool neverdedup;
diff --git a/src/backend/access/nbtree/nbtinsert.c b/src/backend/access/nbtree/nbtinsert.c
index 68628ec000..f6f4af8bfe 100644
--- a/src/backend/access/nbtree/nbtinsert.c
+++ b/src/backend/access/nbtree/nbtinsert.c
@@ -329,7 +329,7 @@ _bt_search_insert(Relation rel, BTInsertState insertstate)
_bt_checkpage(rel, insertstate->buf);
page = BufferGetPage(insertstate->buf);
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
/*
* Check if the page is still the rightmost leaf page and has
@@ -428,7 +428,7 @@ _bt_check_unique(Relation rel, BTInsertState insertstate, Relation heapRel,
InitDirtySnapshot(SnapshotDirty);
page = BufferGetPage(insertstate->buf);
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
maxoff = PageGetMaxOffsetNumber(page);
/*
@@ -733,7 +733,7 @@ _bt_check_unique(Relation rel, BTInsertState insertstate, Relation heapRel,
nbuf = _bt_relandgetbuf(rel, nbuf, nblkno, BT_READ);
page = BufferGetPage(nbuf);
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
if (!P_IGNORE(opaque))
break;
if (P_RIGHTMOST(opaque))
@@ -822,7 +822,7 @@ _bt_findinsertloc(Relation rel,
BTPageOpaque opaque;
OffsetNumber newitemoff;
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
/* Check 1/3 of a page restriction */
if (unlikely(insertstate->itemsz > BTMaxItemSize(page)))
@@ -888,7 +888,7 @@ _bt_findinsertloc(Relation rel,
_bt_stepright(rel, insertstate, stack);
/* Update local state after stepping right */
page = BufferGetPage(insertstate->buf);
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
/* Assume duplicates (if checkingunique) */
uniquedup = true;
}
@@ -972,7 +972,7 @@ _bt_findinsertloc(Relation rel,
_bt_stepright(rel, insertstate, stack);
/* Update local state after stepping right */
page = BufferGetPage(insertstate->buf);
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
}
}
@@ -1030,7 +1030,7 @@ _bt_stepright(Relation rel, BTInsertState insertstate, BTStack stack)
BlockNumber rblkno;
page = BufferGetPage(insertstate->buf);
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
rbuf = InvalidBuffer;
rblkno = opaque->btpo_next;
@@ -1038,7 +1038,7 @@ _bt_stepright(Relation rel, BTInsertState insertstate, BTStack stack)
{
rbuf = _bt_relandgetbuf(rel, rbuf, rblkno, BT_WRITE);
page = BufferGetPage(rbuf);
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
/*
* If this page was incompletely split, finish the split now. We do
@@ -1120,7 +1120,7 @@ _bt_insertonpg(Relation rel,
IndexTuple nposting = NULL;
page = BufferGetPage(buf);
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
isleaf = P_ISLEAF(opaque);
isroot = P_ISROOT(opaque);
isrightmost = P_RIGHTMOST(opaque);
@@ -1296,7 +1296,7 @@ _bt_insertonpg(Relation rel,
if (!isleaf)
{
Page cpage = BufferGetPage(cbuf);
- BTPageOpaque cpageop = (BTPageOpaque) PageGetSpecialPointer(cpage);
+ BTPageOpaque cpageop = BTPageGetOpaque(cpage);
Assert(P_INCOMPLETE_SPLIT(cpageop));
cpageop->btpo_flags &= ~BTP_INCOMPLETE_SPLIT;
@@ -1504,7 +1504,7 @@ _bt_split(Relation rel, BTScanInsert itup_key, Buffer buf, Buffer cbuf,
* only workspace.
*/
origpage = BufferGetPage(buf);
- oopaque = (BTPageOpaque) PageGetSpecialPointer(origpage);
+ oopaque = BTPageGetOpaque(origpage);
isleaf = P_ISLEAF(oopaque);
isrightmost = P_RIGHTMOST(oopaque);
maxoff = PageGetMaxOffsetNumber(origpage);
@@ -1540,7 +1540,7 @@ _bt_split(Relation rel, BTScanInsert itup_key, Buffer buf, Buffer cbuf,
/* Allocate temp buffer for leftpage */
leftpage = PageGetTempPage(origpage);
_bt_pageinit(leftpage, BufferGetPageSize(buf));
- lopaque = (BTPageOpaque) PageGetSpecialPointer(leftpage);
+ lopaque = BTPageGetOpaque(leftpage);
/*
* leftpage won't be the root when we're done. Also, clear the SPLIT_END
@@ -1716,7 +1716,7 @@ _bt_split(Relation rel, BTScanInsert itup_key, Buffer buf, Buffer cbuf,
rightpage = BufferGetPage(rbuf);
rightpagenumber = BufferGetBlockNumber(rbuf);
/* rightpage was initialized by _bt_getbuf */
- ropaque = (BTPageOpaque) PageGetSpecialPointer(rightpage);
+ ropaque = BTPageGetOpaque(rightpage);
/*
* Finish off remaining leftpage special area fields. They cannot be set
@@ -1887,7 +1887,7 @@ _bt_split(Relation rel, BTScanInsert itup_key, Buffer buf, Buffer cbuf,
{
sbuf = _bt_getbuf(rel, oopaque->btpo_next, BT_WRITE);
spage = BufferGetPage(sbuf);
- sopaque = (BTPageOpaque) PageGetSpecialPointer(spage);
+ sopaque = BTPageGetOpaque(spage);
if (sopaque->btpo_prev != origpagenumber)
{
memset(rightpage, 0, BufferGetPageSize(rbuf));
@@ -1952,7 +1952,7 @@ _bt_split(Relation rel, BTScanInsert itup_key, Buffer buf, Buffer cbuf,
if (!isleaf)
{
Page cpage = BufferGetPage(cbuf);
- BTPageOpaque cpageop = (BTPageOpaque) PageGetSpecialPointer(cpage);
+ BTPageOpaque cpageop = BTPageGetOpaque(cpage);
cpageop->btpo_flags &= ~BTP_INCOMPLETE_SPLIT;
MarkBufferDirty(cbuf);
@@ -2139,7 +2139,7 @@ _bt_insert_parent(Relation rel,
BTPageOpaque opaque;
elog(DEBUG2, "concurrent ROOT page split");
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
/*
* We should never reach here when a leaf page split takes place
@@ -2230,7 +2230,7 @@ void
_bt_finish_split(Relation rel, Buffer lbuf, BTStack stack)
{
Page lpage = BufferGetPage(lbuf);
- BTPageOpaque lpageop = (BTPageOpaque) PageGetSpecialPointer(lpage);
+ BTPageOpaque lpageop = BTPageGetOpaque(lpage);
Buffer rbuf;
Page rpage;
BTPageOpaque rpageop;
@@ -2242,7 +2242,7 @@ _bt_finish_split(Relation rel, Buffer lbuf, BTStack stack)
/* Lock right sibling, the one missing the downlink */
rbuf = _bt_getbuf(rel, lpageop->btpo_next, BT_WRITE);
rpage = BufferGetPage(rbuf);
- rpageop = (BTPageOpaque) PageGetSpecialPointer(rpage);
+ rpageop = BTPageGetOpaque(rpage);
/* Could this be a root split? */
if (!stack)
@@ -2320,7 +2320,7 @@ _bt_getstackbuf(Relation rel, BTStack stack, BlockNumber child)
buf = _bt_getbuf(rel, blkno, BT_WRITE);
page = BufferGetPage(buf);
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
if (P_INCOMPLETE_SPLIT(opaque))
{
@@ -2451,7 +2451,7 @@ _bt_newroot(Relation rel, Buffer lbuf, Buffer rbuf)
lbkno = BufferGetBlockNumber(lbuf);
rbkno = BufferGetBlockNumber(rbuf);
lpage = BufferGetPage(lbuf);
- lopaque = (BTPageOpaque) PageGetSpecialPointer(lpage);
+ lopaque = BTPageGetOpaque(lpage);
/* get a new root page */
rootbuf = _bt_getbuf(rel, P_NEW, BT_WRITE);
@@ -2492,11 +2492,11 @@ _bt_newroot(Relation rel, Buffer lbuf, Buffer rbuf)
_bt_upgrademetapage(metapg);
/* set btree special data */
- rootopaque = (BTPageOpaque) PageGetSpecialPointer(rootpage);
+ rootopaque = BTPageGetOpaque(rootpage);
rootopaque->btpo_prev = rootopaque->btpo_next = P_NONE;
rootopaque->btpo_flags = BTP_ROOT;
rootopaque->btpo_level =
- ((BTPageOpaque) PageGetSpecialPointer(lpage))->btpo_level + 1;
+ (BTPageGetOpaque(lpage))->btpo_level + 1;
rootopaque->btpo_cycleid = 0;
/* update metapage data */
@@ -2680,7 +2680,7 @@ _bt_delete_or_dedup_one_page(Relation rel, Relation heapRel,
Buffer buffer = insertstate->buf;
BTScanInsert itup_key = insertstate->itup_key;
Page page = BufferGetPage(buffer);
- BTPageOpaque opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ BTPageOpaque opaque = BTPageGetOpaque(page);
Assert(P_ISLEAF(opaque));
Assert(simpleonly || itup_key->heapkeyspace);
diff --git a/src/backend/access/nbtree/nbtpage.c b/src/backend/access/nbtree/nbtpage.c
index 6b5f01e1d0..20adb602a4 100644
--- a/src/backend/access/nbtree/nbtpage.c
+++ b/src/backend/access/nbtree/nbtpage.c
@@ -85,7 +85,7 @@ _bt_initmetapage(Page page, BlockNumber rootbknum, uint32 level,
metad->btm_last_cleanup_num_heap_tuples = -1.0;
metad->btm_allequalimage = allequalimage;
- metaopaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ metaopaque = BTPageGetOpaque(page);
metaopaque->btpo_flags = BTP_META;
/*
@@ -112,7 +112,7 @@ _bt_upgrademetapage(Page page)
BTPageOpaque metaopaque PG_USED_FOR_ASSERTS_ONLY;
metad = BTPageGetMeta(page);
- metaopaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ metaopaque = BTPageGetOpaque(page);
/* It must be really a meta page of upgradable version */
Assert(metaopaque->btpo_flags & BTP_META);
@@ -148,7 +148,7 @@ _bt_getmeta(Relation rel, Buffer metabuf)
BTMetaPageData *metad;
metapg = BufferGetPage(metabuf);
- metaopaque = (BTPageOpaque) PageGetSpecialPointer(metapg);
+ metaopaque = BTPageGetOpaque(metapg);
metad = BTPageGetMeta(metapg);
/* sanity-check the metapage */
@@ -372,7 +372,7 @@ _bt_getroot(Relation rel, int access)
rootbuf = _bt_getbuf(rel, rootblkno, BT_READ);
rootpage = BufferGetPage(rootbuf);
- rootopaque = (BTPageOpaque) PageGetSpecialPointer(rootpage);
+ rootopaque = BTPageGetOpaque(rootpage);
/*
* Since the cache might be stale, we check the page more carefully
@@ -440,7 +440,7 @@ _bt_getroot(Relation rel, int access)
rootbuf = _bt_getbuf(rel, P_NEW, BT_WRITE);
rootblkno = BufferGetBlockNumber(rootbuf);
rootpage = BufferGetPage(rootbuf);
- rootopaque = (BTPageOpaque) PageGetSpecialPointer(rootpage);
+ rootopaque = BTPageGetOpaque(rootpage);
rootopaque->btpo_prev = rootopaque->btpo_next = P_NONE;
rootopaque->btpo_flags = (BTP_LEAF | BTP_ROOT);
rootopaque->btpo_level = 0;
@@ -534,7 +534,7 @@ _bt_getroot(Relation rel, int access)
{
rootbuf = _bt_relandgetbuf(rel, rootbuf, rootblkno, BT_READ);
rootpage = BufferGetPage(rootbuf);
- rootopaque = (BTPageOpaque) PageGetSpecialPointer(rootpage);
+ rootopaque = BTPageGetOpaque(rootpage);
if (!P_IGNORE(rootopaque))
break;
@@ -598,7 +598,7 @@ _bt_gettrueroot(Relation rel)
metabuf = _bt_getbuf(rel, BTREE_METAPAGE, BT_READ);
metapg = BufferGetPage(metabuf);
- metaopaque = (BTPageOpaque) PageGetSpecialPointer(metapg);
+ metaopaque = BTPageGetOpaque(metapg);
metad = BTPageGetMeta(metapg);
if (!P_ISMETA(metaopaque) ||
@@ -637,7 +637,7 @@ _bt_gettrueroot(Relation rel)
{
rootbuf = _bt_relandgetbuf(rel, rootbuf, rootblkno, BT_READ);
rootpage = BufferGetPage(rootbuf);
- rootopaque = (BTPageOpaque) PageGetSpecialPointer(rootpage);
+ rootopaque = BTPageGetOpaque(rootpage);
if (!P_IGNORE(rootopaque))
break;
@@ -1220,7 +1220,7 @@ _bt_delitems_vacuum(Relation rel, Buffer buf,
* We can clear the vacuum cycle ID since this page has certainly been
* processed by the current vacuum scan.
*/
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
opaque->btpo_cycleid = 0;
/*
@@ -1338,7 +1338,7 @@ _bt_delitems_delete(Relation rel, Buffer buf, TransactionId latestRemovedXid,
* Unlike _bt_delitems_vacuum, we *must not* clear the vacuum cycle ID at
* this point. The VACUUM command alone controls vacuum cycle IDs.
*/
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
/*
* Clear the BTP_HAS_GARBAGE page flag.
@@ -1718,7 +1718,7 @@ _bt_leftsib_splitflag(Relation rel, BlockNumber leftsib, BlockNumber target)
buf = _bt_getbuf(rel, leftsib, BT_READ);
page = BufferGetPage(buf);
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
/*
* If the left sibling was concurrently split, so that its next-pointer
@@ -1773,7 +1773,7 @@ _bt_rightsib_halfdeadflag(Relation rel, BlockNumber leafrightsib)
buf = _bt_getbuf(rel, leafrightsib, BT_READ);
page = BufferGetPage(buf);
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
Assert(P_ISLEAF(opaque) && !P_ISDELETED(opaque));
result = P_ISHALFDEAD(opaque);
@@ -1842,7 +1842,7 @@ _bt_pagedel(Relation rel, Buffer leafbuf, BTVacState *vstate)
for (;;)
{
page = BufferGetPage(leafbuf);
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
/*
* Internal pages are never deleted directly, only as part of deleting
@@ -2099,7 +2099,7 @@ _bt_mark_page_halfdead(Relation rel, Buffer leafbuf, BTStack stack)
IndexTupleData trunctuple;
page = BufferGetPage(leafbuf);
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
Assert(!P_RIGHTMOST(opaque) && !P_ISROOT(opaque) &&
P_ISLEAF(opaque) && !P_IGNORE(opaque) &&
@@ -2154,7 +2154,7 @@ _bt_mark_page_halfdead(Relation rel, Buffer leafbuf, BTStack stack)
* before entering the critical section --- otherwise it'd be a PANIC.
*/
page = BufferGetPage(subtreeparent);
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
#ifdef USE_ASSERT_CHECKING
@@ -2201,7 +2201,7 @@ _bt_mark_page_halfdead(Relation rel, Buffer leafbuf, BTStack stack)
* nbtree/README.
*/
page = BufferGetPage(subtreeparent);
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
itemid = PageGetItemId(page, poffset);
itup = (IndexTuple) PageGetItem(page, itemid);
@@ -2216,7 +2216,7 @@ _bt_mark_page_halfdead(Relation rel, Buffer leafbuf, BTStack stack)
* is set to InvalidBlockNumber.
*/
page = BufferGetPage(leafbuf);
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
opaque->btpo_flags |= BTP_HALF_DEAD;
Assert(PageGetMaxOffsetNumber(page) == P_HIKEY);
@@ -2253,7 +2253,7 @@ _bt_mark_page_halfdead(Relation rel, Buffer leafbuf, BTStack stack)
XLogRegisterBuffer(1, subtreeparent, REGBUF_STANDARD);
page = BufferGetPage(leafbuf);
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
xlrec.leftblk = opaque->btpo_prev;
xlrec.rightblk = opaque->btpo_next;
@@ -2325,7 +2325,7 @@ _bt_unlink_halfdead_page(Relation rel, Buffer leafbuf, BlockNumber scanblkno,
BlockNumber leaftopparent;
page = BufferGetPage(leafbuf);
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
Assert(P_ISLEAF(opaque) && !P_ISDELETED(opaque) && P_ISHALFDEAD(opaque));
@@ -2364,7 +2364,7 @@ _bt_unlink_halfdead_page(Relation rel, Buffer leafbuf, BlockNumber scanblkno,
/* Fetch the block number of the target's left sibling */
buf = _bt_getbuf(rel, target, BT_READ);
page = BufferGetPage(buf);
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
leftsib = opaque->btpo_prev;
targetlevel = opaque->btpo_level;
Assert(targetlevel > 0);
@@ -2391,7 +2391,7 @@ _bt_unlink_halfdead_page(Relation rel, Buffer leafbuf, BlockNumber scanblkno,
{
lbuf = _bt_getbuf(rel, leftsib, BT_WRITE);
page = BufferGetPage(lbuf);
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
while (P_ISDELETED(opaque) || opaque->btpo_next != target)
{
bool leftsibvalid = true;
@@ -2441,7 +2441,7 @@ _bt_unlink_halfdead_page(Relation rel, Buffer leafbuf, BlockNumber scanblkno,
/* step right one page */
lbuf = _bt_getbuf(rel, leftsib, BT_WRITE);
page = BufferGetPage(lbuf);
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
}
}
else
@@ -2450,7 +2450,7 @@ _bt_unlink_halfdead_page(Relation rel, Buffer leafbuf, BlockNumber scanblkno,
/* Next write-lock the target page itself */
_bt_lockbuf(rel, buf, BT_WRITE);
page = BufferGetPage(buf);
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
/*
* Check page is still empty etc, else abandon deletion. This is just for
@@ -2505,7 +2505,7 @@ _bt_unlink_halfdead_page(Relation rel, Buffer leafbuf, BlockNumber scanblkno,
rightsib = opaque->btpo_next;
rbuf = _bt_getbuf(rel, rightsib, BT_WRITE);
page = BufferGetPage(rbuf);
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
if (opaque->btpo_prev != target)
ereport(ERROR,
(errcode(ERRCODE_INDEX_CORRUPTED),
@@ -2528,7 +2528,7 @@ _bt_unlink_halfdead_page(Relation rel, Buffer leafbuf, BlockNumber scanblkno,
if (leftsib == P_NONE && rightsib_is_rightmost)
{
page = BufferGetPage(rbuf);
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
if (P_RIGHTMOST(opaque))
{
/* rightsib will be the only one left on the level */
@@ -2565,12 +2565,12 @@ _bt_unlink_halfdead_page(Relation rel, Buffer leafbuf, BlockNumber scanblkno,
if (BufferIsValid(lbuf))
{
page = BufferGetPage(lbuf);
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
Assert(opaque->btpo_next == target);
opaque->btpo_next = rightsib;
}
page = BufferGetPage(rbuf);
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
Assert(opaque->btpo_prev == target);
opaque->btpo_prev = leftsib;
@@ -2599,7 +2599,7 @@ _bt_unlink_halfdead_page(Relation rel, Buffer leafbuf, BlockNumber scanblkno,
* of that scan.
*/
page = BufferGetPage(buf);
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
Assert(P_ISHALFDEAD(opaque) || !P_ISLEAF(opaque));
/*
@@ -2814,7 +2814,7 @@ _bt_lock_subtree_parent(Relation rel, BlockNumber child, BTStack stack,
parentoffset = stack->bts_offset;
page = BufferGetPage(pbuf);
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
maxoff = PageGetMaxOffsetNumber(page);
leftsibparent = opaque->btpo_prev;
diff --git a/src/backend/access/nbtree/nbtree.c b/src/backend/access/nbtree/nbtree.c
index dacf3f7a58..06131f23d4 100644
--- a/src/backend/access/nbtree/nbtree.c
+++ b/src/backend/access/nbtree/nbtree.c
@@ -1070,7 +1070,7 @@ backtrack:
if (!PageIsNew(page))
{
_bt_checkpage(rel, buf);
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
}
Assert(blkno <= scanblkno);
diff --git a/src/backend/access/nbtree/nbtsearch.c b/src/backend/access/nbtree/nbtsearch.c
index 9d82d4904d..c74543bfde 100644
--- a/src/backend/access/nbtree/nbtsearch.c
+++ b/src/backend/access/nbtree/nbtsearch.c
@@ -134,7 +134,7 @@ _bt_search(Relation rel, BTScanInsert key, Buffer *bufP, int access,
/* if this is a leaf page, we're done */
page = BufferGetPage(*bufP);
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
if (P_ISLEAF(opaque))
break;
@@ -268,7 +268,7 @@ _bt_moveright(Relation rel,
{
page = BufferGetPage(buf);
TestForOldSnapshot(snapshot, rel, page);
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
if (P_RIGHTMOST(opaque))
break;
@@ -347,7 +347,7 @@ _bt_binsrch(Relation rel,
cmpval;
page = BufferGetPage(buf);
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
/* Requesting nextkey semantics while using scantid seems nonsensical */
Assert(!key->nextkey || key->scantid == NULL);
@@ -451,7 +451,7 @@ _bt_binsrch_insert(Relation rel, BTInsertState insertstate)
cmpval;
page = BufferGetPage(insertstate->buf);
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
Assert(P_ISLEAF(opaque));
Assert(!key->nextkey);
@@ -659,7 +659,7 @@ _bt_compare(Relation rel,
OffsetNumber offnum)
{
TupleDesc itupdesc = RelationGetDescr(rel);
- BTPageOpaque opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ BTPageOpaque opaque = BTPageGetOpaque(page);
IndexTuple itup;
ItemPointer heapTid;
ScanKey scankey;
@@ -1536,7 +1536,7 @@ _bt_readpage(IndexScanDesc scan, ScanDirection dir, OffsetNumber offnum)
Assert(BufferIsValid(so->currPos.buf));
page = BufferGetPage(so->currPos.buf);
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
/* allow next page be processed by parallel worker */
if (scan->parallel_scan)
@@ -2007,7 +2007,7 @@ _bt_readnextpage(IndexScanDesc scan, BlockNumber blkno, ScanDirection dir)
so->currPos.buf = _bt_getbuf(rel, blkno, BT_READ);
page = BufferGetPage(so->currPos.buf);
TestForOldSnapshot(scan->xs_snapshot, rel, page);
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
/* check for deleted page */
if (!P_IGNORE(opaque))
{
@@ -2110,7 +2110,7 @@ _bt_readnextpage(IndexScanDesc scan, BlockNumber blkno, ScanDirection dir)
*/
page = BufferGetPage(so->currPos.buf);
TestForOldSnapshot(scan->xs_snapshot, rel, page);
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
if (!P_IGNORE(opaque))
{
PredicateLockPage(rel, BufferGetBlockNumber(so->currPos.buf), scan->xs_snapshot);
@@ -2191,7 +2191,7 @@ _bt_walk_left(Relation rel, Buffer buf, Snapshot snapshot)
BTPageOpaque opaque;
page = BufferGetPage(buf);
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
for (;;)
{
@@ -2216,7 +2216,7 @@ _bt_walk_left(Relation rel, Buffer buf, Snapshot snapshot)
buf = _bt_getbuf(rel, blkno, BT_READ);
page = BufferGetPage(buf);
TestForOldSnapshot(snapshot, rel, page);
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
/*
* If this isn't the page we want, walk right till we find what we
@@ -2243,14 +2243,14 @@ _bt_walk_left(Relation rel, Buffer buf, Snapshot snapshot)
buf = _bt_relandgetbuf(rel, buf, blkno, BT_READ);
page = BufferGetPage(buf);
TestForOldSnapshot(snapshot, rel, page);
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
}
/* Return to the original page to see what's up */
buf = _bt_relandgetbuf(rel, buf, obknum, BT_READ);
page = BufferGetPage(buf);
TestForOldSnapshot(snapshot, rel, page);
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
if (P_ISDELETED(opaque))
{
/*
@@ -2268,7 +2268,7 @@ _bt_walk_left(Relation rel, Buffer buf, Snapshot snapshot)
buf = _bt_relandgetbuf(rel, buf, blkno, BT_READ);
page = BufferGetPage(buf);
TestForOldSnapshot(snapshot, rel, page);
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
if (!P_ISDELETED(opaque))
break;
}
@@ -2329,7 +2329,7 @@ _bt_get_endpoint(Relation rel, uint32 level, bool rightmost,
page = BufferGetPage(buf);
TestForOldSnapshot(snapshot, rel, page);
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
for (;;)
{
@@ -2349,7 +2349,7 @@ _bt_get_endpoint(Relation rel, uint32 level, bool rightmost,
buf = _bt_relandgetbuf(rel, buf, blkno, BT_READ);
page = BufferGetPage(buf);
TestForOldSnapshot(snapshot, rel, page);
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
}
/* Done? */
@@ -2372,7 +2372,7 @@ _bt_get_endpoint(Relation rel, uint32 level, bool rightmost,
buf = _bt_relandgetbuf(rel, buf, blkno, BT_READ);
page = BufferGetPage(buf);
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
}
return buf;
@@ -2418,7 +2418,7 @@ _bt_endpoint(IndexScanDesc scan, ScanDirection dir)
PredicateLockPage(rel, BufferGetBlockNumber(buf), scan->xs_snapshot);
page = BufferGetPage(buf);
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
Assert(P_ISLEAF(opaque));
if (ScanDirectionIsForward(dir))
diff --git a/src/backend/access/nbtree/nbtsort.c b/src/backend/access/nbtree/nbtsort.c
index 8a19de2f66..c074513efa 100644
--- a/src/backend/access/nbtree/nbtsort.c
+++ b/src/backend/access/nbtree/nbtsort.c
@@ -625,7 +625,7 @@ _bt_blnewpage(uint32 level)
_bt_pageinit(page, BLCKSZ);
/* Initialize BT opaque state */
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
opaque->btpo_prev = opaque->btpo_next = P_NONE;
opaque->btpo_level = level;
opaque->btpo_flags = (level > 0) ? 0 : BTP_LEAF;
@@ -1000,9 +1000,9 @@ _bt_buildadd(BTWriteState *wstate, BTPageState *state, IndexTuple itup,
Assert((BTreeTupleGetNAtts(state->btps_lowkey, wstate->index) <=
IndexRelationGetNumberOfKeyAttributes(wstate->index) &&
BTreeTupleGetNAtts(state->btps_lowkey, wstate->index) > 0) ||
- P_LEFTMOST((BTPageOpaque) PageGetSpecialPointer(opage)));
+ P_LEFTMOST(BTPageGetOpaque(opage)));
Assert(BTreeTupleGetNAtts(state->btps_lowkey, wstate->index) == 0 ||
- !P_LEFTMOST((BTPageOpaque) PageGetSpecialPointer(opage)));
+ !P_LEFTMOST(BTPageGetOpaque(opage)));
BTreeTupleSetDownLink(state->btps_lowkey, oblkno);
_bt_buildadd(wstate, state->btps_next, state->btps_lowkey, 0);
pfree(state->btps_lowkey);
@@ -1017,8 +1017,8 @@ _bt_buildadd(BTWriteState *wstate, BTPageState *state, IndexTuple itup,
* Set the sibling links for both pages.
*/
{
- BTPageOpaque oopaque = (BTPageOpaque) PageGetSpecialPointer(opage);
- BTPageOpaque nopaque = (BTPageOpaque) PageGetSpecialPointer(npage);
+ BTPageOpaque oopaque = BTPageGetOpaque(opage);
+ BTPageOpaque nopaque = BTPageGetOpaque(npage);
oopaque->btpo_next = nblkno;
nopaque->btpo_prev = oblkno;
@@ -1125,7 +1125,7 @@ _bt_uppershutdown(BTWriteState *wstate, BTPageState *state)
BTPageOpaque opaque;
blkno = s->btps_blkno;
- opaque = (BTPageOpaque) PageGetSpecialPointer(s->btps_page);
+ opaque = BTPageGetOpaque(s->btps_page);
/*
* We have to link the last page on this level to somewhere.
diff --git a/src/backend/access/nbtree/nbtsplitloc.c b/src/backend/access/nbtree/nbtsplitloc.c
index c46594e1a2..ee01ceafda 100644
--- a/src/backend/access/nbtree/nbtsplitloc.c
+++ b/src/backend/access/nbtree/nbtsplitloc.c
@@ -152,7 +152,7 @@ _bt_findsplitloc(Relation rel,
SplitPoint leftpage,
rightpage;
- opaque = (BTPageOpaque) PageGetSpecialPointer(origpage);
+ opaque = BTPageGetOpaque(origpage);
maxoff = PageGetMaxOffsetNumber(origpage);
/* Total free space available on a btree page, after fixed overhead */
diff --git a/src/backend/access/nbtree/nbtutils.c b/src/backend/access/nbtree/nbtutils.c
index 84164748b3..96c72fc432 100644
--- a/src/backend/access/nbtree/nbtutils.c
+++ b/src/backend/access/nbtree/nbtutils.c
@@ -1774,7 +1774,7 @@ _bt_killitems(IndexScanDesc scan)
}
}
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
minoff = P_FIRSTDATAKEY(opaque);
maxoff = PageGetMaxOffsetNumber(page);
@@ -2474,7 +2474,7 @@ _bt_check_natts(Relation rel, bool heapkeyspace, Page page, OffsetNumber offnum)
{
int16 natts = IndexRelationGetNumberOfAttributes(rel);
int16 nkeyatts = IndexRelationGetNumberOfKeyAttributes(rel);
- BTPageOpaque opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ BTPageOpaque opaque = BTPageGetOpaque(page);
IndexTuple itup;
int tupnatts;
@@ -2662,7 +2662,7 @@ _bt_check_third_page(Relation rel, Relation heap, bool needheaptidspace,
* Internal page insertions cannot fail here, because that would mean that
* an earlier leaf level insertion that should have failed didn't
*/
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
if (!P_ISLEAF(opaque))
elog(ERROR, "cannot insert oversized tuple of size %zu on internal page of index \"%s\"",
itemsz, RelationGetRelationName(rel));
diff --git a/src/backend/access/nbtree/nbtxlog.c b/src/backend/access/nbtree/nbtxlog.c
index 611f412ba8..fba124b940 100644
--- a/src/backend/access/nbtree/nbtxlog.c
+++ b/src/backend/access/nbtree/nbtxlog.c
@@ -115,7 +115,7 @@ _bt_restore_meta(XLogReaderState *record, uint8 block_id)
md->btm_last_cleanup_num_heap_tuples = -1.0;
md->btm_allequalimage = xlrec->allequalimage;
- pageop = (BTPageOpaque) PageGetSpecialPointer(metapg);
+ pageop = BTPageGetOpaque(metapg);
pageop->btpo_flags = BTP_META;
/*
@@ -146,7 +146,7 @@ _bt_clear_incomplete_split(XLogReaderState *record, uint8 block_id)
if (XLogReadBufferForRedo(record, block_id, &buf) == BLK_NEEDS_REDO)
{
Page page = (Page) BufferGetPage(buf);
- BTPageOpaque pageop = (BTPageOpaque) PageGetSpecialPointer(page);
+ BTPageOpaque pageop = BTPageGetOpaque(page);
Assert(P_INCOMPLETE_SPLIT(pageop));
pageop->btpo_flags &= ~BTP_INCOMPLETE_SPLIT;
@@ -292,7 +292,7 @@ btree_xlog_split(bool newitemonleft, XLogReaderState *record)
rpage = (Page) BufferGetPage(rbuf);
_bt_pageinit(rpage, BufferGetPageSize(rbuf));
- ropaque = (BTPageOpaque) PageGetSpecialPointer(rpage);
+ ropaque = BTPageGetOpaque(rpage);
ropaque->btpo_prev = origpagenumber;
ropaque->btpo_next = spagenumber;
@@ -317,7 +317,7 @@ btree_xlog_split(bool newitemonleft, XLogReaderState *record)
* same for the right page.
*/
Page origpage = (Page) BufferGetPage(buf);
- BTPageOpaque oopaque = (BTPageOpaque) PageGetSpecialPointer(origpage);
+ BTPageOpaque oopaque = BTPageGetOpaque(origpage);
OffsetNumber off;
IndexTuple newitem = NULL,
left_hikey = NULL,
@@ -442,7 +442,7 @@ btree_xlog_split(bool newitemonleft, XLogReaderState *record)
if (XLogReadBufferForRedo(record, 2, &sbuf) == BLK_NEEDS_REDO)
{
Page spage = (Page) BufferGetPage(sbuf);
- BTPageOpaque spageop = (BTPageOpaque) PageGetSpecialPointer(spage);
+ BTPageOpaque spageop = BTPageGetOpaque(spage);
spageop->btpo_prev = rightpagenumber;
@@ -473,7 +473,7 @@ btree_xlog_dedup(XLogReaderState *record)
{
char *ptr = XLogRecGetBlockData(record, 0, NULL);
Page page = (Page) BufferGetPage(buf);
- BTPageOpaque opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ BTPageOpaque opaque = BTPageGetOpaque(page);
OffsetNumber offnum,
minoff,
maxoff;
@@ -541,7 +541,7 @@ btree_xlog_dedup(XLogReaderState *record)
if (P_HAS_GARBAGE(opaque))
{
- BTPageOpaque nopaque = (BTPageOpaque) PageGetSpecialPointer(newpage);
+ BTPageOpaque nopaque = BTPageGetOpaque(newpage);
nopaque->btpo_flags &= ~BTP_HAS_GARBAGE;
}
@@ -639,7 +639,7 @@ btree_xlog_vacuum(XLogReaderState *record)
* Mark the page as not containing any LP_DEAD items --- see comments
* in _bt_delitems_vacuum().
*/
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
opaque->btpo_flags &= ~BTP_HAS_GARBAGE;
PageSetLSN(page, lsn);
@@ -699,7 +699,7 @@ btree_xlog_delete(XLogReaderState *record)
PageIndexMultiDelete(page, (OffsetNumber *) ptr, xlrec->ndeleted);
/* Mark the page as not containing any LP_DEAD items */
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
opaque->btpo_flags &= ~BTP_HAS_GARBAGE;
PageSetLSN(page, lsn);
@@ -737,7 +737,7 @@ btree_xlog_mark_page_halfdead(uint8 info, XLogReaderState *record)
BlockNumber rightsib;
page = (Page) BufferGetPage(buffer);
- pageop = (BTPageOpaque) PageGetSpecialPointer(page);
+ pageop = BTPageGetOpaque(page);
poffset = xlrec->poffset;
@@ -768,7 +768,7 @@ btree_xlog_mark_page_halfdead(uint8 info, XLogReaderState *record)
page = (Page) BufferGetPage(buffer);
_bt_pageinit(page, BufferGetPageSize(buffer));
- pageop = (BTPageOpaque) PageGetSpecialPointer(page);
+ pageop = BTPageGetOpaque(page);
pageop->btpo_prev = xlrec->leftblk;
pageop->btpo_next = xlrec->rightblk;
@@ -833,7 +833,7 @@ btree_xlog_unlink_page(uint8 info, XLogReaderState *record)
if (XLogReadBufferForRedo(record, 1, &leftbuf) == BLK_NEEDS_REDO)
{
page = (Page) BufferGetPage(leftbuf);
- pageop = (BTPageOpaque) PageGetSpecialPointer(page);
+ pageop = BTPageGetOpaque(page);
pageop->btpo_next = rightsib;
PageSetLSN(page, lsn);
@@ -848,7 +848,7 @@ btree_xlog_unlink_page(uint8 info, XLogReaderState *record)
page = (Page) BufferGetPage(target);
_bt_pageinit(page, BufferGetPageSize(target));
- pageop = (BTPageOpaque) PageGetSpecialPointer(page);
+ pageop = BTPageGetOpaque(page);
pageop->btpo_prev = leftsib;
pageop->btpo_next = rightsib;
@@ -865,7 +865,7 @@ btree_xlog_unlink_page(uint8 info, XLogReaderState *record)
if (XLogReadBufferForRedo(record, 2, &rightbuf) == BLK_NEEDS_REDO)
{
page = (Page) BufferGetPage(rightbuf);
- pageop = (BTPageOpaque) PageGetSpecialPointer(page);
+ pageop = BTPageGetOpaque(page);
pageop->btpo_prev = leftsib;
PageSetLSN(page, lsn);
@@ -906,7 +906,7 @@ btree_xlog_unlink_page(uint8 info, XLogReaderState *record)
page = (Page) BufferGetPage(leafbuf);
_bt_pageinit(page, BufferGetPageSize(leafbuf));
- pageop = (BTPageOpaque) PageGetSpecialPointer(page);
+ pageop = BTPageGetOpaque(page);
pageop->btpo_flags = BTP_HALF_DEAD | BTP_LEAF;
pageop->btpo_prev = xlrec->leafleftsib;
@@ -948,7 +948,7 @@ btree_xlog_newroot(XLogReaderState *record)
page = (Page) BufferGetPage(buffer);
_bt_pageinit(page, BufferGetPageSize(buffer));
- pageop = (BTPageOpaque) PageGetSpecialPointer(page);
+ pageop = BTPageGetOpaque(page);
pageop->btpo_flags = BTP_ROOT;
pageop->btpo_prev = pageop->btpo_next = P_NONE;
@@ -1097,7 +1097,7 @@ btree_mask(char *pagedata, BlockNumber blkno)
mask_page_hint_bits(page);
mask_unused_space(page);
- maskopaq = (BTPageOpaque) PageGetSpecialPointer(page);
+ maskopaq = BTPageGetOpaque(page);
if (P_ISLEAF(maskopaq))
{
diff --git a/contrib/amcheck/verify_nbtree.c b/contrib/amcheck/verify_nbtree.c
index d2510ee648..70278c4f93 100644
--- a/contrib/amcheck/verify_nbtree.c
+++ b/contrib/amcheck/verify_nbtree.c
@@ -691,7 +691,7 @@ bt_check_level_from_leftmost(BtreeCheckState *state, BtreeLevel level)
state->target = palloc_btree_page(state, state->targetblock);
state->targetlsn = PageGetLSN(state->target);
- opaque = (BTPageOpaque) PageGetSpecialPointer(state->target);
+ opaque = BTPageGetOpaque(state->target);
if (P_IGNORE(opaque))
{
@@ -927,7 +927,7 @@ bt_recheck_sibling_links(BtreeCheckState *state,
LockBuffer(lbuf, BT_READ);
_bt_checkpage(state->rel, lbuf);
page = BufferGetPage(lbuf);
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
if (P_ISDELETED(opaque))
{
/*
@@ -951,7 +951,7 @@ bt_recheck_sibling_links(BtreeCheckState *state,
LockBuffer(newtargetbuf, BT_READ);
_bt_checkpage(state->rel, newtargetbuf);
page = BufferGetPage(newtargetbuf);
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
/* btpo_prev_from_target may have changed; update it */
btpo_prev_from_target = opaque->btpo_prev;
}
@@ -1049,7 +1049,7 @@ bt_target_page_check(BtreeCheckState *state)
OffsetNumber max;
BTPageOpaque topaque;
- topaque = (BTPageOpaque) PageGetSpecialPointer(state->target);
+ topaque = BTPageGetOpaque(state->target);
max = PageGetMaxOffsetNumber(state->target);
elog(DEBUG2, "verifying %u items on %s block %u", max,
@@ -1478,7 +1478,7 @@ bt_target_page_check(BtreeCheckState *state)
/* Get fresh copy of target page */
state->target = palloc_btree_page(state, state->targetblock);
/* Note that we deliberately do not update target LSN */
- topaque = (BTPageOpaque) PageGetSpecialPointer(state->target);
+ topaque = BTPageGetOpaque(state->target);
/*
* All !readonly checks now performed; just return
@@ -1552,7 +1552,7 @@ bt_right_page_check_scankey(BtreeCheckState *state)
OffsetNumber nline;
/* Determine target's next block number */
- opaque = (BTPageOpaque) PageGetSpecialPointer(state->target);
+ opaque = BTPageGetOpaque(state->target);
/* If target is already rightmost, no right sibling; nothing to do here */
if (P_RIGHTMOST(opaque))
@@ -1588,7 +1588,7 @@ bt_right_page_check_scankey(BtreeCheckState *state)
CHECK_FOR_INTERRUPTS();
rightpage = palloc_btree_page(state, targetnext);
- opaque = (BTPageOpaque) PageGetSpecialPointer(rightpage);
+ opaque = BTPageGetOpaque(rightpage);
if (!P_IGNORE(opaque) || P_RIGHTMOST(opaque))
break;
@@ -1893,7 +1893,7 @@ bt_child_highkey_check(BtreeCheckState *state,
else
page = palloc_btree_page(state, blkno);
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
/* The first page we visit at the level should be leftmost */
if (first && !BlockNumberIsValid(state->prevrightlink) && !P_LEFTMOST(opaque))
@@ -1971,7 +1971,7 @@ bt_child_highkey_check(BtreeCheckState *state,
else
pivotkey_offset = target_downlinkoffnum;
- topaque = (BTPageOpaque) PageGetSpecialPointer(state->target);
+ topaque = BTPageGetOpaque(state->target);
if (!offset_is_negative_infinity(topaque, pivotkey_offset))
{
@@ -2128,9 +2128,9 @@ bt_child_check(BtreeCheckState *state, BTScanInsert targetkey,
* Check all items, rather than checking just the first and trusting that
* the operator class obeys the transitive law.
*/
- topaque = (BTPageOpaque) PageGetSpecialPointer(state->target);
+ topaque = BTPageGetOpaque(state->target);
child = palloc_btree_page(state, childblock);
- copaque = (BTPageOpaque) PageGetSpecialPointer(child);
+ copaque = BTPageGetOpaque(child);
maxoffset = PageGetMaxOffsetNumber(child);
/*
@@ -2235,7 +2235,7 @@ static void
bt_downlink_missing_check(BtreeCheckState *state, bool rightsplit,
BlockNumber blkno, Page page)
{
- BTPageOpaque opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ BTPageOpaque opaque = BTPageGetOpaque(page);
ItemId itemid;
IndexTuple itup;
Page child;
@@ -2319,7 +2319,7 @@ bt_downlink_missing_check(BtreeCheckState *state, bool rightsplit,
CHECK_FOR_INTERRUPTS();
child = palloc_btree_page(state, childblk);
- copaque = (BTPageOpaque) PageGetSpecialPointer(child);
+ copaque = BTPageGetOpaque(child);
if (P_ISLEAF(copaque))
break;
@@ -2780,7 +2780,7 @@ invariant_l_offset(BtreeCheckState *state, BTScanInsert key,
bool nonpivot;
ritup = (IndexTuple) PageGetItem(state->target, itemid);
- topaque = (BTPageOpaque) PageGetSpecialPointer(state->target);
+ topaque = BTPageGetOpaque(state->target);
nonpivot = P_ISLEAF(topaque) && upperbound >= P_FIRSTDATAKEY(topaque);
/* Get number of keys + heap TID for item to the right */
@@ -2895,7 +2895,7 @@ invariant_l_nontarget_offset(BtreeCheckState *state, BTScanInsert key,
bool nonpivot;
child = (IndexTuple) PageGetItem(nontarget, itemid);
- copaque = (BTPageOpaque) PageGetSpecialPointer(nontarget);
+ copaque = BTPageGetOpaque(nontarget);
nonpivot = P_ISLEAF(copaque) && upperbound >= P_FIRSTDATAKEY(copaque);
/* Get number of keys + heap TID for child/non-target item */
@@ -2954,7 +2954,7 @@ palloc_btree_page(BtreeCheckState *state, BlockNumber blocknum)
memcpy(page, BufferGetPage(buffer), BLCKSZ);
UnlockReleaseBuffer(buffer);
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
if (P_ISMETA(opaque) && blocknum != BTREE_METAPAGE)
ereport(ERROR,
diff --git a/contrib/pageinspect/btreefuncs.c b/contrib/pageinspect/btreefuncs.c
index 7651c59bbf..3daa31c84d 100644
--- a/contrib/pageinspect/btreefuncs.c
+++ b/contrib/pageinspect/btreefuncs.c
@@ -93,7 +93,7 @@ GetBTPageStatistics(BlockNumber blkno, Buffer buffer, BTPageStat *stat)
Page page = BufferGetPage(buffer);
PageHeader phdr = (PageHeader) page;
OffsetNumber maxoff = PageGetMaxOffsetNumber(page);
- BTPageOpaque opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ BTPageOpaque opaque = BTPageGetOpaque(page);
int item_size = 0;
int off;
@@ -525,7 +525,7 @@ bt_page_items_internal(PG_FUNCTION_ARGS, enum pageinspect_version ext_version)
uargs->offset = FirstOffsetNumber;
- opaque = (BTPageOpaque) PageGetSpecialPointer(uargs->page);
+ opaque = BTPageGetOpaque(uargs->page);
if (!P_ISDELETED(opaque))
fctx->max_calls = PageGetMaxOffsetNumber(uargs->page);
@@ -622,7 +622,7 @@ bt_page_items_bytea(PG_FUNCTION_ARGS)
(int) MAXALIGN(sizeof(BTPageOpaqueData)),
(int) PageGetSpecialSize(uargs->page))));
- opaque = (BTPageOpaque) PageGetSpecialPointer(uargs->page);
+ opaque = BTPageGetOpaque(uargs->page);
if (P_ISMETA(opaque))
ereport(ERROR,
diff --git a/contrib/pageinspect/hashfuncs.c b/contrib/pageinspect/hashfuncs.c
index 6de21d6608..69af7b962f 100644
--- a/contrib/pageinspect/hashfuncs.c
+++ b/contrib/pageinspect/hashfuncs.c
@@ -72,7 +72,7 @@ verify_hash_page(bytea *raw_page, int flags)
(int) MAXALIGN(sizeof(HashPageOpaqueData)),
(int) PageGetSpecialSize(page))));
- pageopaque = (HashPageOpaque) PageGetSpecialPointer(page);
+ pageopaque = HashPageGetOpaque(page);
if (pageopaque->hasho_page_id != HASHO_PAGE_ID)
ereport(ERROR,
(errcode(ERRCODE_INVALID_PARAMETER_VALUE),
@@ -154,7 +154,7 @@ static void
GetHashPageStatistics(Page page, HashPageStat *stat)
{
OffsetNumber maxoff = PageGetMaxOffsetNumber(page);
- HashPageOpaque opaque = (HashPageOpaque) PageGetSpecialPointer(page);
+ HashPageOpaque opaque = HashPageGetOpaque(page);
int off;
stat->dead_items = stat->live_items = 0;
@@ -206,7 +206,7 @@ hash_page_type(PG_FUNCTION_ARGS)
type = "unused";
else
{
- opaque = (HashPageOpaque) PageGetSpecialPointer(page);
+ opaque = HashPageGetOpaque(page);
/* page type (flags) */
pagetype = opaque->hasho_flag & LH_PAGE_TYPE;
diff --git a/contrib/pgstattuple/pgstatindex.c b/contrib/pgstattuple/pgstatindex.c
index 6c4b053dd0..e1048e47ff 100644
--- a/contrib/pgstattuple/pgstatindex.c
+++ b/contrib/pgstattuple/pgstatindex.c
@@ -281,7 +281,7 @@ pgstatindex_impl(Relation rel, FunctionCallInfo fcinfo)
LockBuffer(buffer, BUFFER_LOCK_SHARE);
page = BufferGetPage(buffer);
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
/*
* Determine page type, and update totals.
@@ -641,7 +641,7 @@ pgstathashindex(PG_FUNCTION_ARGS)
HashPageOpaque opaque;
int pagetype;
- opaque = (HashPageOpaque) PageGetSpecialPointer(page);
+ opaque = HashPageGetOpaque(page);
pagetype = opaque->hasho_flag & LH_PAGE_TYPE;
if (pagetype == LH_BUCKET_PAGE)
diff --git a/contrib/pgstattuple/pgstattuple.c b/contrib/pgstattuple/pgstattuple.c
index c9b8f01f9b..3094566908 100644
--- a/contrib/pgstattuple/pgstattuple.c
+++ b/contrib/pgstattuple/pgstattuple.c
@@ -415,7 +415,7 @@ pgstat_btree_page(pgstattuple_type *stat, Relation rel, BlockNumber blkno,
{
BTPageOpaque opaque;
- opaque = (BTPageOpaque) PageGetSpecialPointer(page);
+ opaque = BTPageGetOpaque(page);
if (P_IGNORE(opaque))
{
/* deleted or half-dead page */
@@ -452,7 +452,7 @@ pgstat_hash_page(pgstattuple_type *stat, Relation rel, BlockNumber blkno,
{
HashPageOpaque opaque;
- opaque = (HashPageOpaque) PageGetSpecialPointer(page);
+ opaque = HashPageGetOpaque(page);
switch (opaque->hasho_flag & LH_PAGE_TYPE)
{
case LH_UNUSED_PAGE:
--
2.35.1
Re: Preventing indirection for IndexPageGetOpaque for known-size page special areas
On Thu, 31 Mar 2022 at 09:32, Michael Paquier <michael@paquier.xyz> wrote:
On Mon, Mar 28, 2022 at 05:09:10PM +0200, Matthias van de Meent wrote:
Not all clusters have checksums enabled (boo!, but we can't
realistically fix that), so on-disk corruption could reasonably
propagate to the rest of such system. Additionally, checksums are only
checked on read, and updated when the page is written to disk (in
PageSetChecksumCopy called from FlushBuffer), but this does not check
for signs of page corruption. As such, a memory bug resulting in
in-memory corruption of pd_special would persist and would currently
have the potential to further corrupt neighbouring buffers.Well, if it comes to corruption then pd_special is not the only
problem, just one part of it. A broken pd_lower or pd_lower or even
pd_lsn could also cause AMs to point at areas they should not. There
are many reasons that could make things go wrong depending on the
compiled page size.A second reason would be less indirection to get to the opaque
pointer. This should improve performance a bit in those cases where we
(initially) only want to access the [Index]PageOpaque struct.We don't have many cases that do that, do we?
Indeed not many, but I know that at least the nbtree-code has some
cases where it uses the opaque before other fields in the header are
accessed (sometimes even without accessing the header for other
reasons): in _bt_moveright and _bt_walk_left. There might be more; but
these are cases I know of.Even these are marginal as the page is already loaded in the shared
buffers..
I can't really disagree there.
While looking at the page, the pieces in 0002 and 0003 that I really
liked are the introduction of the macros to grab the special area for
btree and hash. This brings the code of those APIs closer to GiST,
SpGiST and GIN. And it is possible to move to a possible change in
the checks and/or the shape of all the *GetOpaque macros at once after
the switch to this part is done. So I don't really mind introducing
this part.PageGetSpecialOpaque() would not have saved the day with the recent
pageinspect changes, and that's just moving the responsability of the
page header to something else. I am not sure if it is a good idea to
have a mention of the opaque page type in bufpage.h actually, as this
is optional. Having an AssertMacro() checking that pd_special is in
line with MAXALIGN(sizeof(OpaqueData)) is attractive, but I'd rather
keep that in each AM's headers per its optional nature, and an index
AM may handle things differently depending on a page type, as well.
PageInit MAXALIGNs the size of the special area that it receives as an
argument; so any changes to the page header that would misalign the
value would be AM-specific; in which case it is quite unlikely that
this is the right accessor for your page's special area.
- Matthias
Re: Preventing indirection for IndexPageGetOpaque for known-size page special areas
On Thu, Mar 31, 2022 at 12:09:35PM +0200, Matthias van de Meent wrote:
PageInit MAXALIGNs the size of the special area that it receives as an
argument; so any changes to the page header that would misalign the
value would be AM-specific; in which case it is quite unlikely that
this is the right accessor for your page's special area.
Right. I'd still be tempted to keep that per-AM rather than making
the checks deeper with one extra macro layer in the page header or
with a different macro that would depend on the opaque type, though.
Like in the attached, for example.
--
Michael
Attachments:
opaque-macros.patchtext/x-diff; charset=us-asciiDownload
diff --git a/src/include/access/ginblock.h b/src/include/access/ginblock.h
index 9347f464f3..050fff80dc 100644
--- a/src/include/access/ginblock.h
+++ b/src/include/access/ginblock.h
@@ -108,7 +108,11 @@ typedef struct GinMetaPageData
/*
* Macros for accessing a GIN index page's opaque data
*/
-#define GinPageGetOpaque(page) ( (GinPageOpaque) PageGetSpecialPointer(page) )
+#define GinPageGetOpaque(page) \
+( \
+ AssertMacro(PageGetSpecialSize(page) == MAXALIGN(sizeof(GinPageOpaqueData))), \
+ (GinPageOpaque) PageGetSpecialPointer(page) \
+)
#define GinPageIsLeaf(page) ( (GinPageGetOpaque(page)->flags & GIN_LEAF) != 0 )
#define GinPageSetLeaf(page) ( GinPageGetOpaque(page)->flags |= GIN_LEAF )
diff --git a/src/include/access/gist.h b/src/include/access/gist.h
index a3337627b8..a10caf8ae9 100644
--- a/src/include/access/gist.h
+++ b/src/include/access/gist.h
@@ -164,7 +164,11 @@ typedef struct GISTENTRY
bool leafkey;
} GISTENTRY;
-#define GistPageGetOpaque(page) ( (GISTPageOpaque) PageGetSpecialPointer(page) )
+#define GistPageGetOpaque(page) \
+( \
+ AssertMacro(PageGetSpecialSize(page) == MAXALIGN(sizeof(GISTPageOpaqueData))), \
+ (GISTPageOpaque) PageGetSpecialPointer(page) \
+)
#define GistPageIsLeaf(page) ( GistPageGetOpaque(page)->flags & F_LEAF)
#define GIST_LEAF(entry) (GistPageIsLeaf((entry)->page))
diff --git a/src/include/access/hash.h b/src/include/access/hash.h
index da372841c4..b442ddc4c8 100644
--- a/src/include/access/hash.h
+++ b/src/include/access/hash.h
@@ -85,7 +85,11 @@ typedef struct HashPageOpaqueData
typedef HashPageOpaqueData *HashPageOpaque;
-#define HashPageGetOpaque(page) ((HashPageOpaque) PageGetSpecialPointer(page))
+#define HashPageGetOpaque(page) \
+( \
+ AssertMacro(PageGetSpecialSize(page) == MAXALIGN(sizeof(HashPageOpaqueData))), \
+ (HashPageOpaque) PageGetSpecialPointer(page) \
+)
#define H_NEEDS_SPLIT_CLEANUP(opaque) (((opaque)->hasho_flag & LH_BUCKET_NEEDS_SPLIT_CLEANUP) != 0)
#define H_BUCKET_BEING_SPLIT(opaque) (((opaque)->hasho_flag & LH_BUCKET_BEING_SPLIT) != 0)
diff --git a/src/include/access/nbtree.h b/src/include/access/nbtree.h
index 93f8267b48..e11839857d 100644
--- a/src/include/access/nbtree.h
+++ b/src/include/access/nbtree.h
@@ -70,7 +70,11 @@ typedef struct BTPageOpaqueData
typedef BTPageOpaqueData *BTPageOpaque;
-#define BTPageGetOpaque(page) ((BTPageOpaque) PageGetSpecialPointer(page))
+#define BTPageGetOpaque(page) \
+( \
+ AssertMacro(PageGetSpecialSize(page) == MAXALIGN(sizeof(BTPageOpaqueData))), \
+ (BTPageOpaque) PageGetSpecialPointer(page) \
+)
/* Bits defined in btpo_flags */
#define BTP_LEAF (1 << 0) /* leaf page, i.e. not internal page */
diff --git a/src/include/access/spgist_private.h b/src/include/access/spgist_private.h
index eb56b1c6b8..d61433b5af 100644
--- a/src/include/access/spgist_private.h
+++ b/src/include/access/spgist_private.h
@@ -75,7 +75,12 @@ typedef SpGistPageOpaqueData *SpGistPageOpaque;
#define SPGIST_LEAF (1<<2)
#define SPGIST_NULLS (1<<3)
-#define SpGistPageGetOpaque(page) ((SpGistPageOpaque) PageGetSpecialPointer(page))
+#define SpGistPageGetOpaque(page) \
+( \
+ AssertMacro(PageGetSpecialSize(page) == MAXALIGN(sizeof(SpGistPageOpaqueData))), \
+ (SpGistPageOpaque) PageGetSpecialPointer(page) \
+)
+
#define SpGistPageIsMeta(page) (SpGistPageGetOpaque(page)->flags & SPGIST_META)
#define SpGistPageIsDeleted(page) (SpGistPageGetOpaque(page)->flags & SPGIST_DELETED)
#define SpGistPageIsLeaf(page) (SpGistPageGetOpaque(page)->flags & SPGIST_LEAF)
Re: Preventing indirection for IndexPageGetOpaque for known-size page special areas
On Fri, 1 Apr 2022 at 07:38, Michael Paquier <michael@paquier.xyz> wrote:
On Thu, Mar 31, 2022 at 12:09:35PM +0200, Matthias van de Meent wrote:
PageInit MAXALIGNs the size of the special area that it receives as an
argument; so any changes to the page header that would misalign the
value would be AM-specific; in which case it is quite unlikely that
this is the right accessor for your page's special area.Right. I'd still be tempted to keep that per-AM rather than making
the checks deeper with one extra macro layer in the page header or
with a different macro that would depend on the opaque type, though.
Like in the attached, for example.
I see. I still would like it better if the access could use this
statically determined offset: your opaque-macros.patch doesn't fix the
out-of-bound read/write scenariofor non-assert builds, nor does it
remove the unneeded indirection through the page header that I was
trying to remove.
Even in assert-enabled builds; with my proposed changes the code paths
for checking the header value and the use of the special area can be
executed independently, which allows for parallel (pre-)fetching of
the page header and special area, as opposed to the current sequential
load order due to the required use of pd_special.
-Matthias
Re: Preventing indirection for IndexPageGetOpaque for known-size page special areas
On Fri, 1 Apr 2022 at 10:50, Matthias van de Meent
<boekewurm+postgres@gmail.com> wrote:
On Fri, 1 Apr 2022 at 07:38, Michael Paquier <michael@paquier.xyz> wrote:
On Thu, Mar 31, 2022 at 12:09:35PM +0200, Matthias van de Meent wrote:
PageInit MAXALIGNs the size of the special area that it receives as an
argument; so any changes to the page header that would misalign the
value would be AM-specific; in which case it is quite unlikely that
this is the right accessor for your page's special area.Right. I'd still be tempted to keep that per-AM rather than making
the checks deeper with one extra macro layer in the page header or
with a different macro that would depend on the opaque type, though.
Like in the attached, for example.
I noticed that you committed the equivalent of 0002 and 0003, thank you.
Here's a new 0001 to keep CFBot happy.
I see. I still would like it better if the access could use this
statically determined offset: your opaque-macros.patch doesn't fix the
out-of-bound read/write scenariofor non-assert builds, nor does it
remove the unneeded indirection through the page header that I was
trying to remove.Even in assert-enabled builds; with my proposed changes the code paths
for checking the header value and the use of the special area can be
executed independently, which allows for parallel (pre-)fetching of
the page header and special area, as opposed to the current sequential
load order due to the required use of pd_special.
As an extra argument for this; it removes the need for a temporary
variable on the stack; as now all accesses into the special area can
be based on offsets off the base page pointer (which is also used
often). This would allow the compiler to utilize the available
registers more effectively.
-Matthias
PS. I noticed that between PageGetSpecialPointer and
PageGetSpecialOpaque the binary size was bigger by ~1kb for the
*Opaque variant when compiled with `CFLAGS="-o2" ./configure
--enable-debug`; but that varied a lot between builds and likely
related to binary layouts. For similar builds with `--enable-cassert`,
the *Opaque build was slimmer by ~ 50kB, which is a suprisingly large
amount.
Attachments:
v3-0001-Add-known-size-pre-aligned-special-area-pointer-m.patchapplication/x-patch; name=v3-0001-Add-known-size-pre-aligned-special-area-pointer-m.patchDownload
From d9e6afdfbc021a699e68529cf37d66b8ff999dbc Mon Sep 17 00:00:00 2001
From: Matthias van de Meent <boekewurm+postgres@gmail.com>
Date: Mon, 28 Mar 2022 14:53:43 +0200
Subject: [PATCH v3] Add known-size pre-aligned special area pointer macro
This removes 1 layer of indirection for special areas of which we know the
type (=size) and location.
Special area access through the page header is an extra cache line that needs
to be fetched. If might only want to look at the special area, it is much
less effort to calculate the offset to the special area directly instead of
first checking the page header - saving one cache line to fetch.
Assertions are added to check that the page has a correctly sized special
area, and that the page is of the expected size. This detects data corruption,
instead of doing random reads/writes into the page or data allocated next to
the page being accessed.
Additionally, updates the GIN, [SP-]GIST, Hash, and BTree
[Index]PageGetOpaque macros with the new pre-aligned special area accessor.
---
src/include/access/ginblock.h | 3 ++-
src/include/access/gist.h | 3 ++-
src/include/access/hash.h | 3 ++-
src/include/access/nbtree.h | 3 ++-
src/include/access/spgist_private.h | 3 ++-
src/include/storage/bufpage.h | 14 ++++++++++++++
6 files changed, 24 insertions(+), 5 deletions(-)
diff --git a/src/include/access/ginblock.h b/src/include/access/ginblock.h
index 9347f464f3..3680098c98 100644
--- a/src/include/access/ginblock.h
+++ b/src/include/access/ginblock.h
@@ -108,7 +108,8 @@ typedef struct GinMetaPageData
/*
* Macros for accessing a GIN index page's opaque data
*/
-#define GinPageGetOpaque(page) ( (GinPageOpaque) PageGetSpecialPointer(page) )
+#define GinPageGetOpaque(page) \
+ ((GinPageOpaque) PageGetSpecialOpaque(page, GinPageOpaqueData))
#define GinPageIsLeaf(page) ( (GinPageGetOpaque(page)->flags & GIN_LEAF) != 0 )
#define GinPageSetLeaf(page) ( GinPageGetOpaque(page)->flags |= GIN_LEAF )
diff --git a/src/include/access/gist.h b/src/include/access/gist.h
index a3337627b8..51223e9051 100644
--- a/src/include/access/gist.h
+++ b/src/include/access/gist.h
@@ -164,7 +164,8 @@ typedef struct GISTENTRY
bool leafkey;
} GISTENTRY;
-#define GistPageGetOpaque(page) ( (GISTPageOpaque) PageGetSpecialPointer(page) )
+#define GistPageGetOpaque(page) \
+ ((GISTPageOpaque) PageGetSpecialOpaque(page, GISTPageOpaqueData))
#define GistPageIsLeaf(page) ( GistPageGetOpaque(page)->flags & F_LEAF)
#define GIST_LEAF(entry) (GistPageIsLeaf((entry)->page))
diff --git a/src/include/access/hash.h b/src/include/access/hash.h
index da372841c4..0ed6ccc40b 100644
--- a/src/include/access/hash.h
+++ b/src/include/access/hash.h
@@ -85,7 +85,8 @@ typedef struct HashPageOpaqueData
typedef HashPageOpaqueData *HashPageOpaque;
-#define HashPageGetOpaque(page) ((HashPageOpaque) PageGetSpecialPointer(page))
+#define HashPageGetOpaque(page) \
+ ((HashPageOpaque) PageGetSpecialOpaque(page, HashPageOpaqueData))
#define H_NEEDS_SPLIT_CLEANUP(opaque) (((opaque)->hasho_flag & LH_BUCKET_NEEDS_SPLIT_CLEANUP) != 0)
#define H_BUCKET_BEING_SPLIT(opaque) (((opaque)->hasho_flag & LH_BUCKET_BEING_SPLIT) != 0)
diff --git a/src/include/access/nbtree.h b/src/include/access/nbtree.h
index 93f8267b48..2d87c8a90f 100644
--- a/src/include/access/nbtree.h
+++ b/src/include/access/nbtree.h
@@ -70,7 +70,8 @@ typedef struct BTPageOpaqueData
typedef BTPageOpaqueData *BTPageOpaque;
-#define BTPageGetOpaque(page) ((BTPageOpaque) PageGetSpecialPointer(page))
+#define BTPageGetOpaque(page) \
+ ((BTPageOpaque) PageGetSpecialOpaque(page, BTPageOpaqueData))
/* Bits defined in btpo_flags */
#define BTP_LEAF (1 << 0) /* leaf page, i.e. not internal page */
diff --git a/src/include/access/spgist_private.h b/src/include/access/spgist_private.h
index eb56b1c6b8..b0bdc7df3e 100644
--- a/src/include/access/spgist_private.h
+++ b/src/include/access/spgist_private.h
@@ -75,7 +75,8 @@ typedef SpGistPageOpaqueData *SpGistPageOpaque;
#define SPGIST_LEAF (1<<2)
#define SPGIST_NULLS (1<<3)
-#define SpGistPageGetOpaque(page) ((SpGistPageOpaque) PageGetSpecialPointer(page))
+#define SpGistPageGetOpaque(page) \
+ ((SpGistPageOpaque) PageGetSpecialOpaque(page, SpGistPageOpaqueData))
#define SpGistPageIsMeta(page) (SpGistPageGetOpaque(page)->flags & SPGIST_META)
#define SpGistPageIsDeleted(page) (SpGistPageGetOpaque(page)->flags & SPGIST_DELETED)
#define SpGistPageIsLeaf(page) (SpGistPageGetOpaque(page)->flags & SPGIST_LEAF)
diff --git a/src/include/storage/bufpage.h b/src/include/storage/bufpage.h
index e9f253f2c8..e96595f338 100644
--- a/src/include/storage/bufpage.h
+++ b/src/include/storage/bufpage.h
@@ -329,6 +329,20 @@ PageValidateSpecialPointer(Page page)
(char *) ((char *) (page) + ((PageHeader) (page))->pd_special) \
)
+/*
+ * PageGetSpecialOpaque
+ * Returns pointer to special space on a page, of the given type.
+ * This removes the data dependency on the page header in builds
+ * without asserts enabled.
+ */
+#define PageGetSpecialOpaque(page, OpaqueDataType) \
+( \
+ AssertMacro(PageGetPageSize(page) == BLCKSZ && \
+ PageGetSpecialSize(page) == MAXALIGN(sizeof(OpaqueDataType))), \
+ (OpaqueDataType *) ((char *) (page) + \
+ (BLCKSZ - MAXALIGN(sizeof(OpaqueDataType)))) \
+)
+
/*
* PageGetItem
* Retrieves an item on the given page.
--
2.30.2
Re: Preventing indirection for IndexPageGetOpaque for known-size page special areas
On Fri, Apr 1, 2022 at 11:12 AM Matthias van de Meent
<boekewurm+postgres@gmail.com> wrote:
Here's a new 0001 to keep CFBot happy.
This seems like it would conflict with the proposal from
/messages/by-id/CA+TgmoaD8wMN6i1mmuo+4ZNeGE3Hd57ys8uV8UZm7cneqy3W2g@mail.gmail.com
which I still hope to advance in some form at an appropriate time.
--
Robert Haas
EDB: http://www.enterprisedb.com
Re: Preventing indirection for IndexPageGetOpaque for known-size page special areas
On Tue, 5 Apr 2022 at 21:45, Robert Haas <robertmhaas@gmail.com> wrote:
On Fri, Apr 1, 2022 at 11:12 AM Matthias van de Meent
<boekewurm+postgres@gmail.com> wrote:Here's a new 0001 to keep CFBot happy.
This seems like it would conflict with the proposal from
/messages/by-id/CA+TgmoaD8wMN6i1mmuo+4ZNeGE3Hd57ys8uV8UZm7cneqy3W2g@mail.gmail.com
which I still hope to advance in some form at an appropriate time.
I can see why what would be the case. However, I don't think that is
of much importance here, as reverting the changes that are yet to be
committed would be trivial.
Also, I can't see why we would allow page-level layout changes in
initdb; that seems like the wrong place to do that. All page layout
currently is at compile-time; even checksums (which can be
enabled/disabled in initdb) have reserved space available in the page
header. Why would it be different for nonces?
I don't think that the 'storing an explicit nonce' patch would
conflict in any meaningful form, other than maybe needing to roll back
from PageGetSpecialOpaque to PageGetSpecialPointer when (if) the nonce
is indeed going to be stored in the special area of each page with a
size defined during initdb. Until then, we could have a nice (but
small) performance boost.
-Matthias
Re: Preventing indirection for IndexPageGetOpaque for known-size page special areas
On Fri, Apr 1, 2022 at 11:12 AM Matthias van de Meent
<boekewurm+postgres@gmail.com> wrote:Here's a new 0001 to keep CFBot happy.
This seems like it would conflict with the proposal from
/messages/by-id/CA+TgmoaD8wMN6i1mmuo+4ZNeGE3Hd57ys8uV8UZm7cneqy3W2g@mail.gmail.com
which I still hope to advance in some form at an appropriate time.
It may be weakly related to the current discussion, but I'd like to note
that the other proposed patch for making XID's 64-bit [1]/messages/by-id/CACG=ezZe1NQSCnfHOr78AtAZxJZeCvxrts0ygrxYwe=pyyjVWA@mail.gmail.com will store
per-page-common xid data in the heap page special area. It doesn't
contradict storing a nonce in this area also, but I'd appreciate it if
we could avoid changing heap page special layout twice.
[1]: /messages/by-id/CACG=ezZe1NQSCnfHOr78AtAZxJZeCvxrts0ygrxYwe=pyyjVWA@mail.gmail.com
/messages/by-id/CACG=ezZe1NQSCnfHOr78AtAZxJZeCvxrts0ygrxYwe=pyyjVWA@mail.gmail.com
Re: Preventing indirection for IndexPageGetOpaque for known-size page special areas
On Thu, Apr 7, 2022 at 4:34 AM Matthias van de Meent
<boekewurm+postgres@gmail.com> wrote:
Also, I can't see why we would allow page-level layout changes in
initdb; that seems like the wrong place to do that. All page layout
currently is at compile-time; even checksums (which can be
enabled/disabled in initdb) have reserved space available in the page
header. Why would it be different for nonces?
Because there's no place to put them in the existing page format. We
jammed checksums into the 2-byte field that had previously been set
aside for the TLI, but that wasn't really an ideal solution because it
meant we ended up with a checksum that is only 16 bits wide. However,
the 2 bytes set aside for the TLI weren't really being used
effectively anyway, so repurposing them was relatively easy, and a
16-bit checksum is better than nothing.
For encryption or integrity verification, you need a lot more space,
like 16-32 bytes. I can't see us adding that unilaterally to every
cluster, because (1) it would break pg_upgrade compatibility, (2) it's
a pretty significant amount of space to set aside in every single
cluster whether it is using those features or not, and (3) the
technology is constantly evolving and we can't predict how many bytes
will be needed in the future. AES in whatever mode we use it requires
whatever number of bytes it does, but AES replaced DES and will in
turn be replaced by something else which can have totally different
requirements.
I do understand that there are significant challenges and performance
concerns around having these kinds of initdb-controlled page layout
changes, so the future of that patch is unclear. However, I don't
think that it's practical to say that a 16-bit checksum should be good
enough for everyone, and I also don't think it's practical to say that
if you want a wider checksum or to use encryption or whatever, you
have to recompile from source. Most users are going to get precompiled
binaries from a vendor, and those binaries need to be able to work in
all the configurations that users want to use. It's OK for
experimental things that we only expect developers to fool with to
require recompiling, but things that users have a legitimate need to
reconfigure need to be initdb-time controllable at worst. Hence
fc49e24fa69a15efacd5b8958115ed9c43c48f9a, for example.
--
Robert Haas
EDB: http://www.enterprisedb.com
Re: Preventing indirection for IndexPageGetOpaque for known-size page special areas
On Thu, Apr 7, 2022 at 7:01 AM Robert Haas <robertmhaas@gmail.com> wrote:
Because there's no place to put them in the existing page format. We
jammed checksums into the 2-byte field that had previously been set
aside for the TLI, but that wasn't really an ideal solution because it
meant we ended up with a checksum that is only 16 bits wide. However,
the 2 bytes set aside for the TLI weren't really being used
effectively anyway, so repurposing them was relatively easy, and a
16-bit checksum is better than nothing.
But if we were in a green-field situation we'd probably not want to
use up several bytes for a nonse anyway. You said so yourself.
I do understand that there are significant challenges and performance
concerns around having these kinds of initdb-controlled page layout
changes, so the future of that patch is unclear.
Why does it need to be at initdb time?
Though I cannot prove it, I suspect that the original intent of the
special area was to support an additional (though typically small)
variable length array, that works a little like the current line
pointer array. This array would have to grow backwards (newer items
get appended at earlier physical offsets), unlike our line pointer
array (which gets appended to at the end, in the simple and obvious
way). Growing backwards like this happens with DB systems, that store
their line pointer array at the end of the page(the traditional
approach from the System R days, I believe).
Supporting a variable-length special area array like this would mean
that any time you add a new item to the variable-sized array in the
special area, the page's entire tuple space has to be memmove()'d
backwards by a couple of bytes to create the required space. And so
the relevant bufpage.c routine would have to adjust the whole line
pointer array such that each lp_off received a compensating
adjustment. The array might only be for some kind of page-level
transaction metadata, something like that -- shifting it around is
pretty expensive (reusing existing slots isn't too expensive, though).
Why can't it work like that? You don't really need to build the full
set of bufpage.c facilities (though it might not be a bad idea to
fully support these variable-length arrays, which seem like they might
come in handy). That seems perfectly compatible with what Matthias
wants to do, provided we're willing to deem the special area struct
(e.g. BTOpaque) as always coming "first" (which is essentially the
same as his current proposal anyway). You can even do the same thing
yourself for the nonse (use a fixed, known offset), with relatively
modest effort. You'd need to have AM-specific knowledge (it would
stack right on top of Matthias's technique), but that doesn't seem all
that hard. There are plenty of remaining status bits in BTOpaque, and
probably all other index AM special areas.
--
Peter Geoghegan
Re: Preventing indirection for IndexPageGetOpaque for known-size page special areas
On Thu, Apr 7, 2022 at 2:43 PM Peter Geoghegan <pg@bowt.ie> wrote:
But if we were in a green-field situation we'd probably not want to
use up several bytes for a nonse anyway. You said so yourself.
I don't know what statement of mine you're talking about here, and
while I don't love using up space for a nonce, it seems to be the way
this encryption stuff works. I don't see that there's a reasonable
alternative, green field or no.
I do understand that there are significant challenges and performance
concerns around having these kinds of initdb-controlled page layout
changes, so the future of that patch is unclear.Why does it need to be at initdb time?
Though I cannot prove it, I suspect that the original intent of the
special area was to support an additional (though typically small)
variable length array, that works a little like the current line
pointer array. This array would have to grow backwards (newer items
get appended at earlier physical offsets), unlike our line pointer
array (which gets appended to at the end, in the simple and obvious
way). Growing backwards like this happens with DB systems, that store
their line pointer array at the end of the page(the traditional
approach from the System R days, I believe).Supporting a variable-length special area array like this would mean
that any time you add a new item to the variable-sized array in the
special area, the page's entire tuple space has to be memmove()'d
backwards by a couple of bytes to create the required space. And so
the relevant bufpage.c routine would have to adjust the whole line
pointer array such that each lp_off received a compensating
adjustment. The array might only be for some kind of page-level
transaction metadata, something like that -- shifting it around is
pretty expensive (reusing existing slots isn't too expensive, though).Why can't it work like that? You don't really need to build the full
set of bufpage.c facilities (though it might not be a bad idea to
fully support these variable-length arrays, which seem like they might
come in handy). That seems perfectly compatible with what Matthias
wants to do, provided we're willing to deem the special area struct
(e.g. BTOpaque) as always coming "first" (which is essentially the
same as his current proposal anyway). You can even do the same thing
yourself for the nonse (use a fixed, known offset), with relatively
modest effort. You'd need to have AM-specific knowledge (it would
stack right on top of Matthias's technique), but that doesn't seem all
that hard. There are plenty of remaining status bits in BTOpaque, and
probably all other index AM special areas.
I'm not really following any of this. You seem to be arguing about
whether it's possible to change the length of the special space
*later* than initdb time. I agree that might have some use for some
purpose, but for encryption it's not necessarily all that helpful
because you have to be able to find the nonce on the page before
you've decrypted it. If you don't know whether there's a nonce or
where it's located, you can't do that. What Matthias and I were
discussing is whether you have to make a decision about appending
stuff to the special space *earlier* than initdb-time i.e. at compile
time.
My position is that if we need some space in every page to put a
nonce, the best place to put it is at the very end of the page, within
the special space and after anything else that is stored in the
special space. Code that only manipulates the line pointer array and
tuple data won't care, because pd_special will just be a bit smaller
than it would otherwise have been, and none of that code looks at any
byte offset >= pd_special. Code that looks at the special space won't
care either, as long as it uses PageGetSpecialPointer to find the
data, and doesn't examine how large the special space actually is.
That corresponds pretty well to how existing users of the special
space work, so it seems pretty good.
If we *didn't* put the nonce at the end of the page, where else would
we put it? It has to be at a fixed offset, because otherwise you can't
find it without decrypting the page first, which would be circular.
You could put it at the beginning of the page, or after the page
header and before the line pointer array, but either of those things
seem likely to affect a lot more code, because there's a lot more
stuff that accesses the line pointer array than the special space.
--
Robert Haas
EDB: http://www.enterprisedb.com
Re: Preventing indirection for IndexPageGetOpaque for known-size page special areas
On Thu, Apr 7, 2022 at 12:11 PM Robert Haas <robertmhaas@gmail.com> wrote:
I don't know what statement of mine you're talking about here, and
while I don't love using up space for a nonce, it seems to be the way
this encryption stuff works. I don't see that there's a reasonable
alternative, green field or no.
I just meant that it wouldn't be reasonable to impose a fixed cost on
every user, even those not using the feature. Which you said yourself.
I'm not really following any of this. You seem to be arguing about
whether it's possible to change the length of the special space
*later* than initdb time. I agree that might have some use for some
purpose, but for encryption it's not necessarily all that helpful
because you have to be able to find the nonce on the page before
you've decrypted it. If you don't know whether there's a nonce or
where it's located, you can't do that.
What if you just encrypt a significant subset of the page, and leave a
small amount of metadata that can be read without decryption? Is that
approach feasible?
I think that you could even encrypt the line pointer array (not just
the tuple space), without breaking anything.
What Matthias and I were
discussing is whether you have to make a decision about appending
stuff to the special space *earlier* than initdb-time i.e. at compile
time.
I got that much, of course. That will work, I suppose, but it'll be
the first and last time that anybody gets to do that (unless we accept
it being incompatible with encryption).
That corresponds pretty well to how existing users of the special
space work, so it seems pretty good.
I think that it still needs to be accounted for in a bunch of places.
In particular anywhere concerned with page-level space management. For
example the definition of "1/3 of a page" used to enforce the limit on
the maximum size of an nbtree index tuple will need to account for the
presence of a nonse, either way.
If we *didn't* put the nonce at the end of the page, where else would
we put it? It has to be at a fixed offset, because otherwise you can't
find it without decrypting the page first, which would be circular.
Immediately before the special area proper (say BTOpaque), which would
"logically come after" the special space under this scheme. You
wouldn't have a simple constant offset into the page, but you'd have
something not too far removed from such a constant. It could work as a
constant with minimal context (just the AM type). Just like with
Matthias' patch.
--
Peter Geoghegan
Re: Preventing indirection for IndexPageGetOpaque for known-size page special areas
On Thu, Apr 7, 2022 at 3:27 PM Peter Geoghegan <pg@bowt.ie> wrote:
I just meant that it wouldn't be reasonable to impose a fixed cost on
every user, even those not using the feature. Which you said yourself.
Unfortunately, I think there's bound to be some cost. We can avoid
using the space in the page for every user, but anything that makes
the page layout variable is going to cost some number of CPU cycles
someplace. We have to measure that overhead and see if it's small
enough that we're OK with it.
I got that much, of course. That will work, I suppose, but it'll be
the first and last time that anybody gets to do that (unless we accept
it being incompatible with encryption).
Yeah.
If we *didn't* put the nonce at the end of the page, where else would
we put it? It has to be at a fixed offset, because otherwise you can't
find it without decrypting the page first, which would be circular.Immediately before the special area proper (say BTOpaque), which would
"logically come after" the special space under this scheme. You
wouldn't have a simple constant offset into the page, but you'd have
something not too far removed from such a constant. It could work as a
constant with minimal context (just the AM type). Just like with
Matthias' patch.
I don't think this would work, because I don't think it would be
practical to always know the AM type. Think about applying an XLOG_FPI
record, for example.
--
Robert Haas
EDB: http://www.enterprisedb.com
Re: Preventing indirection for IndexPageGetOpaque for known-size page special areas
On Thu, Apr 7, 2022 at 12:37 PM Robert Haas <robertmhaas@gmail.com> wrote:
On Thu, Apr 7, 2022 at 3:27 PM Peter Geoghegan <pg@bowt.ie> wrote:
I just meant that it wouldn't be reasonable to impose a fixed cost on
every user, even those not using the feature. Which you said yourself.Unfortunately, I think there's bound to be some cost. We can avoid
using the space in the page for every user, but anything that makes
the page layout variable is going to cost some number of CPU cycles
someplace. We have to measure that overhead and see if it's small
enough that we're OK with it.
I wouldn't go so far as to say that any non-zero overhead is not okay
(that sounds really extreme). I would only say this much: wasting
non-trivial amounts of space on every page must not happen. If the
user opts-in to the feature then it's not just waste, so that's
perfectly okay. (I imagine that we agree on this much already.)
Immediately before the special area proper (say BTOpaque), which would
"logically come after" the special space under this scheme. You
wouldn't have a simple constant offset into the page, but you'd have
something not too far removed from such a constant. It could work as a
constant with minimal context (just the AM type). Just like with
Matthias' patch.I don't think this would work, because I don't think it would be
practical to always know the AM type. Think about applying an XLOG_FPI
record, for example.
There are already some pretty shaky heuristics that are used by tools
like pg_filedump for this exact purpose. But they work! And they're
even supported by Postgres, quasi-officially -- grep for "pg_filedump"
to see what I mean.
There are numerous reasons why we might want to put that on a formal
footing, so that we can reliably detect the AM type starting from only
a page image. I suspect that you're going to end up needing to account
for this index AMs anyway, so going this way isn't necessarily making
your life all that much harder. (I am not really sure about that,
though, since I don't have enough background information.)
--
Peter Geoghegan
Re: Preventing indirection for IndexPageGetOpaque for known-size page special areas
Greetings,
* Robert Haas (robertmhaas@gmail.com) wrote:
On Thu, Apr 7, 2022 at 3:27 PM Peter Geoghegan <pg@bowt.ie> wrote:
I just meant that it wouldn't be reasonable to impose a fixed cost on
every user, even those not using the feature. Which you said yourself.Unfortunately, I think there's bound to be some cost. We can avoid
using the space in the page for every user, but anything that makes
the page layout variable is going to cost some number of CPU cycles
someplace. We have to measure that overhead and see if it's small
enough that we're OK with it.
Agreed that there may be some cost in CPU cycles for folks who don't
initialize the database with an option that needs this, but hopefully
that'll be quite small.
I got that much, of course. That will work, I suppose, but it'll be
the first and last time that anybody gets to do that (unless we accept
it being incompatible with encryption).Yeah.
I don't know that I agree with this being the 'first and last time'..?
If we have two options that could work together and each need some
amount of per-page space, such as a nonce or authentication tag, we'd
just need to be able to track which of those are enabled (eg: through
pg_control) and then know which gets what space. I don't see why we
couldn't add something today and then add something else later on. I do
think there'll likely be cases where we have two options that don't make
sense together and we wouldn't wish to allow that (such as page-level
strong checksums and authenticated encryption, as the latter provides
the former inherently) but there could certainly be things that do work
together too.
If we *didn't* put the nonce at the end of the page, where else would
we put it? It has to be at a fixed offset, because otherwise you can't
find it without decrypting the page first, which would be circular.Immediately before the special area proper (say BTOpaque), which would
"logically come after" the special space under this scheme. You
wouldn't have a simple constant offset into the page, but you'd have
something not too far removed from such a constant. It could work as a
constant with minimal context (just the AM type). Just like with
Matthias' patch.I don't think this would work, because I don't think it would be
practical to always know the AM type. Think about applying an XLOG_FPI
record, for example.
I'm also doubtful about how well this would work, but the other question
is- what would be the advantage to doing it this way? If we could have
it be run-time instead of initdb-time, that'd be great (imagine a
database that's encrypted while another isn't in the same cluster, or
even individual tables, which would all be very cool), but I don't think
this approach would make that possible..?
(sorry for only just seeing this thread getting traction, have been a
bit busy with other things ...)
Thanks,
Stephen
Re: Preventing indirection for IndexPageGetOpaque for known-size page special areas
Greetings,
* Peter Geoghegan (pg@bowt.ie) wrote:
On Thu, Apr 7, 2022 at 12:37 PM Robert Haas <robertmhaas@gmail.com> wrote:
On Thu, Apr 7, 2022 at 3:27 PM Peter Geoghegan <pg@bowt.ie> wrote:
I just meant that it wouldn't be reasonable to impose a fixed cost on
every user, even those not using the feature. Which you said yourself.Unfortunately, I think there's bound to be some cost. We can avoid
using the space in the page for every user, but anything that makes
the page layout variable is going to cost some number of CPU cycles
someplace. We have to measure that overhead and see if it's small
enough that we're OK with it.I wouldn't go so far as to say that any non-zero overhead is not okay
(that sounds really extreme). I would only say this much: wasting
non-trivial amounts of space on every page must not happen. If the
user opts-in to the feature then it's not just waste, so that's
perfectly okay. (I imagine that we agree on this much already.)
Right, the *space* wouldn't ever be wasted- either the user opts-in and
the space is used for what the user asked it to be used for, or ther
user doesn't select that option and we don't reserve that space.
Robert's point was that by allowing this to happen at initdb time, there
may be some CPU cycles that end up getting spent, even if the user
didn't opt-in, that would be saved if this decision was made at compile
time. For my 2c, I'd think that would be our main concern with this,
but I also am hopeful and strongly suspect that the extra CPU cycles
aren't likely to be much and therefore would be acceptable.
A thought that's further down the line would be the question of if we
could get those CPU cycles back (and perhaps more..) by using JIT.
Immediately before the special area proper (say BTOpaque), which would
"logically come after" the special space under this scheme. You
wouldn't have a simple constant offset into the page, but you'd have
something not too far removed from such a constant. It could work as a
constant with minimal context (just the AM type). Just like with
Matthias' patch.I don't think this would work, because I don't think it would be
practical to always know the AM type. Think about applying an XLOG_FPI
record, for example.There are already some pretty shaky heuristics that are used by tools
like pg_filedump for this exact purpose. But they work! And they're
even supported by Postgres, quasi-officially -- grep for "pg_filedump"
to see what I mean.
Shaky heuristics feels like something appropriate for pg_filedump.. less
so for things like TDE.
There are numerous reasons why we might want to put that on a formal
footing, so that we can reliably detect the AM type starting from only
a page image. I suspect that you're going to end up needing to account
for this index AMs anyway, so going this way isn't necessarily making
your life all that much harder. (I am not really sure about that,
though, since I don't have enough background information.)
This seems like it'd be pretty difficult given that AMs can be loaded as
extensions..? Also seems like a much bigger change and I'm still not
sure what the advantage is, at least in terms of what this particular
patch is doing.
Thanks,
Stephen
Re: Preventing indirection for IndexPageGetOpaque for known-size page special areas
On Thu, Apr 7, 2022 at 1:09 PM Stephen Frost <sfrost@snowman.net> wrote:
I got that much, of course. That will work, I suppose, but it'll be
the first and last time that anybody gets to do that (unless we accept
it being incompatible with encryption).Yeah.
I don't know that I agree with this being the 'first and last time'..?
If we have two options that could work together and each need some
amount of per-page space, such as a nonce or authentication tag, we'd
just need to be able to track which of those are enabled (eg: through
pg_control) and then know which gets what space.
Sounds very messy.
I don't see why we
couldn't add something today and then add something else later on.
That's what I'm arguing in favor of, in part.
I'm also doubtful about how well this would work, but the other question
is- what would be the advantage to doing it this way? If we could have
it be run-time instead of initdb-time, that'd be great (imagine a
database that's encrypted while another isn't in the same cluster, or
even individual tables, which would all be very cool), but I don't think
this approach would make that possible..?
That would be the main advantage, yes. But I also tend to doubt that
we should make it completely impossible to know anything at all about
the page without fully decrypting it.
It was just a suggestion. I will leave it at that.
--
Peter Geoghegan
Re: Preventing indirection for IndexPageGetOpaque for known-size page special areas
On Thu, 7 Apr 2022 at 21:11, Robert Haas <robertmhaas@gmail.com> wrote:
On Thu, Apr 7, 2022 at 2:43 PM Peter Geoghegan <pg@bowt.ie> wrote:
But if we were in a green-field situation we'd probably not want to
use up several bytes for a nonse anyway. You said so yourself.I don't know what statement of mine you're talking about here, and
while I don't love using up space for a nonce, it seems to be the way
this encryption stuff works. I don't see that there's a reasonable
alternative, green field or no.I do understand that there are significant challenges and performance
concerns around having these kinds of initdb-controlled page layout
changes, so the future of that patch is unclear.Why does it need to be at initdb time?
Though I cannot prove it, I suspect that the original intent of the
special area was to support an additional (though typically small)
variable length array, that works a little like the current line
pointer array. This array would have to grow backwards (newer items
get appended at earlier physical offsets), unlike our line pointer
array (which gets appended to at the end, in the simple and obvious
way). Growing backwards like this happens with DB systems, that store
their line pointer array at the end of the page(the traditional
approach from the System R days, I believe).Supporting a variable-length special area array like this would mean
that any time you add a new item to the variable-sized array in the
special area, the page's entire tuple space has to be memmove()'d
backwards by a couple of bytes to create the required space. And so
the relevant bufpage.c routine would have to adjust the whole line
pointer array such that each lp_off received a compensating
adjustment. The array might only be for some kind of page-level
transaction metadata, something like that -- shifting it around is
pretty expensive (reusing existing slots isn't too expensive, though).Why can't it work like that? You don't really need to build the full
set of bufpage.c facilities (though it might not be a bad idea to
fully support these variable-length arrays, which seem like they might
come in handy). That seems perfectly compatible with what Matthias
wants to do, provided we're willing to deem the special area struct
(e.g. BTOpaque) as always coming "first" (which is essentially the
same as his current proposal anyway). You can even do the same thing
yourself for the nonse (use a fixed, known offset), with relatively
modest effort. You'd need to have AM-specific knowledge (it would
stack right on top of Matthias's technique), but that doesn't seem all
that hard. There are plenty of remaining status bits in BTOpaque, and
probably all other index AM special areas.I'm not really following any of this. You seem to be arguing about
whether it's possible to change the length of the special space
*later* than initdb time. I agree that might have some use for some
purpose, but for encryption it's not necessarily all that helpful
because you have to be able to find the nonce on the page before
you've decrypted it. If you don't know whether there's a nonce or
where it's located, you can't do that. What Matthias and I were
discussing is whether you have to make a decision about appending
stuff to the special space *earlier* than initdb-time i.e. at compile
time.My position is that if we need some space in every page to put a
nonce, the best place to put it is at the very end of the page, within
the special space and after anything else that is stored in the
special space. Code that only manipulates the line pointer array and
tuple data won't care, because pd_special will just be a bit smaller
than it would otherwise have been, and none of that code looks at any
byte offset >= pd_special. Code that looks at the special space won't
care either, as long as it uses PageGetSpecialPointer to find the
data, and doesn't examine how large the special space actually is.
That corresponds pretty well to how existing users of the special
space work, so it seems pretty good.
Except that reserving space on each page requires recalculation of all
variables that depend on the amount of potential free space available
on a page (for some cases this is less important, for some it is
critical that the value is not wrong). If this is always done at
runtime then that can cause significant overhead.
If we *didn't* put the nonce at the end of the page, where else would
we put it? It has to be at a fixed offset, because otherwise you can't
find it without decrypting the page first, which would be circular.
I think there's no specifically good reason why we'd need to put the
nonce in storage at the same place as where we reserve the space for
the nonce in the unencrypted in-memory format.
You could put it at the beginning of the page, or after the page
header and before the line pointer array, but either of those things
seem likely to affect a lot more code, because there's a lot more
stuff that accesses the line pointer array than the special space.
I'm not too keen on anything related to having no page layout
guarantees. I can understand needing a nonce; but couldn't that be put
somewhere different than smack-dab in the middle of what are
considered AM-controlled areas?
I'm not certain why we need lsn on a page after we've checked that we
flushed all WAL up to that LSN. That is, right now we store the LSN in
the in-memory representation of the page because we need it to check
that the WAL is flushed up to that point when we write out the page,
so that we can recover the data in case of disk write issues.
But after flushing the WAL to disk, this LSN on the page is not needed
anymore, and could thus be replaced with a nonce. When reading such
page, the LSN-now-nonce can be replaced with the latest flushed LSN to
prevent unwanted xlog flushes. Sure, this limits nonce to 8 bytes, but
if you really need more than that IMO you can recompile from scratch
with a bigger PageHeader.
Benefit of using existing pageheader structs is that we could enable
TDE on a relation level and on existing clusters - there's no extra
space needed right now as there's already some space available.
Yes, we'd lose the ability to skip redo on pages, but I think that's a
small price to pay when you enable TDE.
-Matthias
Re: Preventing indirection for IndexPageGetOpaque for known-size page special areas
On Thu, Apr 7, 2022 at 03:11:24PM -0400, Robert Haas wrote:
On Thu, Apr 7, 2022 at 2:43 PM Peter Geoghegan <pg@bowt.ie> wrote:
But if we were in a green-field situation we'd probably not want to
use up several bytes for a nonse anyway. You said so yourself.I don't know what statement of mine you're talking about here, and
while I don't love using up space for a nonce, it seems to be the way
this encryption stuff works. I don't see that there's a reasonable
alternative, green field or no.
Uh, XTS doesn't require a nonce, so why are talking about nonces in this
thread?
--
Bruce Momjian <bruce@momjian.us> https://momjian.us
EDB https://enterprisedb.com
Indecision is a decision. Inaction is an action. Mark Batterson
Re: Preventing indirection for IndexPageGetOpaque for known-size page special areas
Bruce Momjian <bruce@momjian.us> writes:
Uh, XTS doesn't require a nonce, so why are talking about nonces in this
thread?
Because some other proposals do, or could, require a per-page nonce.
After looking through this thread, I side with Robert: we should reject
the remainder of this patch. It gives up page layout flexibility that
we might want back someday. Moreover, I didn't see any hard evidence
offered that there's any actual performance gain, let alone such a
compelling gain that we should give up that flexibility for it.
regards, tom lane
Re: Preventing indirection for IndexPageGetOpaque for known-size page special areas
On Mon, Nov 28, 2022 at 05:31:50PM -0500, Tom Lane wrote:
After looking through this thread, I side with Robert: we should reject
the remainder of this patch. It gives up page layout flexibility that
we might want back someday. Moreover, I didn't see any hard evidence
offered that there's any actual performance gain, let alone such a
compelling gain that we should give up that flexibility for it.
As far as I understand, we are talking about this one:
/messages/by-id/CAEze2Wj9c0abW2aRbC8JzOuUdGurO5av6SJ2H83du6tM+Q1rHQ@mail.gmail.com
After a few months looking at it again, I cannot get much excited
about switching these routines from a logic where we look at the page
header to something where we'd rely on the opaque structure size.
I am wondering if it would be worth adding an AssertMacro() like in
this one, though:
/messages/by-id/YkaP64JvZTMgcHtq@paquier.xyz
This still uses PageGetSpecialPointer() to retrieve the pointer area
from the page header, but it also checks that we have a match with the
structure expected by the index AM.
--
Michael
Re: Preventing indirection for IndexPageGetOpaque for known-size page special areas
Michael Paquier <michael@paquier.xyz> writes:
I am wondering if it would be worth adding an AssertMacro() like in
this one, though:
/messages/by-id/YkaP64JvZTMgcHtq@paquier.xyz
Kind of doubt it. It'd bloat debug builds with a lot of redundant
checks, and probably never catch anything. For catching problems
in production, the right place to do this (and where we already do
do it) is in _bt_checkpage.
If any of the other AMs lack page-read-time sanity checks like
_bt_checkpage, I'd be in favor of adding that. But I don't
think a whole bunch of additional checks afterwards will buy much.
regards, tom lane