From fd861a9a4b5a2de8b1d79c5b70ee03be91e7c354 Mon Sep 17 00:00:00 2001 From: Arseniy Mukhin Date: Mon, 16 Jun 2025 18:11:27 +0300 Subject: [PATCH v10 2/4] amcheck: brin_index_check() - index structure check Adds a new function brin_index_check() for validating BRIN indexes. It incudes next checks: - meta page checks - revmap pointers is valid and points to index tuples with expected range blkno - index tuples have expected format - some special checks for empty_ranges - every index tuple has corresponding revmap item that points to it (optional) --- contrib/amcheck/Makefile | 5 +- contrib/amcheck/amcheck--1.5--1.6.sql | 18 + contrib/amcheck/amcheck.control | 2 +- contrib/amcheck/expected/check_brin.out | 134 ++++ contrib/amcheck/meson.build | 4 + contrib/amcheck/sql/check_brin.sql | 100 +++ contrib/amcheck/t/007_verify_brin.pl | 301 +++++++++ contrib/amcheck/verify_brin.c | 864 ++++++++++++++++++++++++ doc/src/sgml/amcheck.sgml | 33 + 9 files changed, 1458 insertions(+), 3 deletions(-) create mode 100644 contrib/amcheck/amcheck--1.5--1.6.sql create mode 100644 contrib/amcheck/expected/check_brin.out create mode 100644 contrib/amcheck/sql/check_brin.sql create mode 100644 contrib/amcheck/t/007_verify_brin.pl create mode 100644 contrib/amcheck/verify_brin.c diff --git a/contrib/amcheck/Makefile b/contrib/amcheck/Makefile index 1b7a63cbaa4..bdfb274c89c 100644 --- a/contrib/amcheck/Makefile +++ b/contrib/amcheck/Makefile @@ -6,11 +6,12 @@ OBJS = \ verify_common.o \ verify_gin.o \ verify_heapam.o \ - verify_nbtree.o + verify_nbtree.o \ + verify_brin.o EXTENSION = amcheck DATA = amcheck--1.2--1.3.sql amcheck--1.1--1.2.sql amcheck--1.0--1.1.sql amcheck--1.0.sql \ - amcheck--1.3--1.4.sql amcheck--1.4--1.5.sql + amcheck--1.3--1.4.sql amcheck--1.4--1.5.sql amcheck--1.5--1.6.sql PGFILEDESC = "amcheck - function for verifying relation integrity" REGRESS = check check_btree check_gin check_heap diff --git a/contrib/amcheck/amcheck--1.5--1.6.sql b/contrib/amcheck/amcheck--1.5--1.6.sql new file mode 100644 index 00000000000..0354451c472 --- /dev/null +++ b/contrib/amcheck/amcheck--1.5--1.6.sql @@ -0,0 +1,18 @@ +/* contrib/amcheck/amcheck--1.5--1.6.sql */ + +-- complain if script is sourced in psql, rather than via CREATE EXTENSION +\echo Use "ALTER EXTENSION amcheck UPDATE TO '1.6'" to load this file. \quit + + +-- +-- brin_index_check() +-- +CREATE FUNCTION brin_index_check(index regclass, + regularpagescheck boolean default false +) + RETURNS VOID +AS 'MODULE_PATHNAME', 'brin_index_check' +LANGUAGE C STRICT PARALLEL RESTRICTED; + +-- We don't want this to be available to public +REVOKE ALL ON FUNCTION brin_index_check(regclass, boolean) FROM PUBLIC; \ No newline at end of file diff --git a/contrib/amcheck/amcheck.control b/contrib/amcheck/amcheck.control index c8ba6d7c9bc..2f329ef2cf4 100644 --- a/contrib/amcheck/amcheck.control +++ b/contrib/amcheck/amcheck.control @@ -1,5 +1,5 @@ # amcheck extension comment = 'functions for verifying relation integrity' -default_version = '1.5' +default_version = '1.6' module_pathname = '$libdir/amcheck' relocatable = true diff --git a/contrib/amcheck/expected/check_brin.out b/contrib/amcheck/expected/check_brin.out new file mode 100644 index 00000000000..6890fff46bd --- /dev/null +++ b/contrib/amcheck/expected/check_brin.out @@ -0,0 +1,134 @@ +-- helper func +CREATE OR REPLACE FUNCTION random_string(int) RETURNS text AS $$ +SELECT string_agg(substring('0123456789abcdefghijklmnopqrstuvwxyz', ceil(random() * 36)::integer, 1), '') FROM generate_series(1, $1); +$$ LANGUAGE sql; +-- empty table index should be valid +CREATE TABLE brintest (a bigint) WITH (fillfactor = 10); +CREATE INDEX brintest_idx ON brintest USING brin (a); +SELECT brin_index_check('brintest_idx', true); + brin_index_check +------------------ + +(1 row) + +-- cleanup +DROP TABLE brintest; +-- min_max opclass +CREATE TABLE brintest (a bigint) WITH (fillfactor = 10); +CREATE INDEX brintest_idx ON brintest USING brin (a int8_minmax_ops) WITH (pages_per_range = 2); +INSERT INTO brintest (a) SELECT x FROM generate_series(1,100000) x; +-- create some empty ranges +DELETE FROM brintest WHERE a > 20000 AND a < 40000; +SELECT brin_index_check('brintest_idx', true); + brin_index_check +------------------ + +(1 row) + +-- rebuild index +DROP INDEX brintest_idx; +CREATE INDEX brintest_idx ON brintest USING brin (a int8_minmax_ops) WITH (pages_per_range = 2); +SELECT brin_index_check('brintest_idx', true); + brin_index_check +------------------ + +(1 row) + +-- cleanup +DROP TABLE brintest; +-- multi_min_max opclass +CREATE TABLE brintest (a bigint) WITH (fillfactor = 10); +CREATE INDEX brintest_idx ON brintest USING brin (a int8_minmax_multi_ops) WITH (pages_per_range = 2); +INSERT INTO brintest (a) SELECT x FROM generate_series(1,100000) x; +-- create some empty ranges +DELETE FROM brintest WHERE a > 20000 AND a < 40000; +SELECT brin_index_check('brintest_idx', true); + brin_index_check +------------------ + +(1 row) + +-- rebuild index +DROP INDEX brintest_idx; +CREATE INDEX brintest_idx ON brintest USING brin (a int8_minmax_multi_ops) WITH (pages_per_range = 2); +SELECT brin_index_check('brintest_idx', true); + brin_index_check +------------------ + +(1 row) + +-- cleanup +DROP TABLE brintest; +-- bloom opclass +CREATE TABLE brintest (a bigint) WITH (fillfactor = 10); +CREATE INDEX brintest_idx ON brintest USING brin (a int8_bloom_ops) WITH (pages_per_range = 2); +INSERT INTO brintest (a) SELECT x FROM generate_series(1,100000) x; +-- create some empty ranges +DELETE FROM brintest WHERE a > 20000 AND a < 40000; +SELECT brin_index_check('brintest_idx', true); + brin_index_check +------------------ + +(1 row) + +-- rebuild index +DROP INDEX brintest_idx; +CREATE INDEX brintest_idx ON brintest USING brin (a int8_bloom_ops) WITH (pages_per_range = 2); +SELECT brin_index_check('brintest_idx', true); + brin_index_check +------------------ + +(1 row) + +-- cleanup +DROP TABLE brintest; +-- inclusion opclass +CREATE TABLE brintest (id serial PRIMARY KEY, a box); +CREATE INDEX brintest_idx ON brintest USING brin (a box_inclusion_ops) WITH (pages_per_range = 2); +INSERT INTO brintest (a) +SELECT box(point(random() * 1000, random() * 1000), point(random() * 1000, random() * 1000)) +FROM generate_series(1, 10000); +-- create some empty ranges +DELETE FROM brintest WHERE id > 2000 AND id < 4000; +SELECT brin_index_check('brintest_idx', true); + brin_index_check +------------------ + +(1 row) + +-- rebuild index +DROP INDEX brintest_idx; +CREATE INDEX brintest_idx ON brintest USING brin (a box_inclusion_ops) WITH (pages_per_range = 2); +SELECT brin_index_check('brintest_idx', true); + brin_index_check +------------------ + +(1 row) + +-- cleanup +DROP TABLE brintest; +-- multiple attributes +CREATE TABLE brintest (id bigserial, a text) WITH (fillfactor = 10); +CREATE INDEX brintest_idx ON brintest USING brin (id int8_minmax_ops, a text_minmax_ops) WITH (pages_per_range = 2); +INSERT INTO brintest (a) SELECT random_string((x % 100)) FROM generate_series(1,3000) x; +-- create some empty ranges +DELETE FROM brintest WHERE id > 1500 AND id < 2500; +SELECT brin_index_check('brintest_idx', true); + brin_index_check +------------------ + +(1 row) + +-- rebuild index +DROP INDEX brintest_idx; +CREATE INDEX brintest_idx ON brintest USING brin (id int8_minmax_ops, a text_minmax_ops) WITH (pages_per_range = 2); +SELECT brin_index_check('brintest_idx', true); + brin_index_check +------------------ + +(1 row) + +-- cleanup +DROP TABLE brintest; +-- cleanup +DROP FUNCTION random_string; diff --git a/contrib/amcheck/meson.build b/contrib/amcheck/meson.build index 1f0c347ed54..ba816c2faf0 100644 --- a/contrib/amcheck/meson.build +++ b/contrib/amcheck/meson.build @@ -5,6 +5,7 @@ amcheck_sources = files( 'verify_gin.c', 'verify_heapam.c', 'verify_nbtree.c', + 'verify_brin.c' ) if host_system == 'windows' @@ -27,6 +28,7 @@ install_data( 'amcheck--1.2--1.3.sql', 'amcheck--1.3--1.4.sql', 'amcheck--1.4--1.5.sql', + 'amcheck--1.5--1.6.sql', kwargs: contrib_data_args, ) @@ -40,6 +42,7 @@ tests += { 'check_btree', 'check_gin', 'check_heap', + 'check_brin' ], }, 'tap': { @@ -50,6 +53,7 @@ tests += { 't/004_verify_nbtree_unique.pl', 't/005_pitr.pl', 't/006_verify_gin.pl', + 't/007_verify_brin.pl', ], }, } diff --git a/contrib/amcheck/sql/check_brin.sql b/contrib/amcheck/sql/check_brin.sql new file mode 100644 index 00000000000..1c97b370cac --- /dev/null +++ b/contrib/amcheck/sql/check_brin.sql @@ -0,0 +1,100 @@ +-- helper func +CREATE OR REPLACE FUNCTION random_string(int) RETURNS text AS $$ +SELECT string_agg(substring('0123456789abcdefghijklmnopqrstuvwxyz', ceil(random() * 36)::integer, 1), '') FROM generate_series(1, $1); +$$ LANGUAGE sql; + + +-- empty table index should be valid +CREATE TABLE brintest (a bigint) WITH (fillfactor = 10); +CREATE INDEX brintest_idx ON brintest USING brin (a); +SELECT brin_index_check('brintest_idx', true); +-- cleanup +DROP TABLE brintest; + +-- min_max opclass +CREATE TABLE brintest (a bigint) WITH (fillfactor = 10); +CREATE INDEX brintest_idx ON brintest USING brin (a int8_minmax_ops) WITH (pages_per_range = 2); +INSERT INTO brintest (a) SELECT x FROM generate_series(1,100000) x; +-- create some empty ranges +DELETE FROM brintest WHERE a > 20000 AND a < 40000; +SELECT brin_index_check('brintest_idx', true); + +-- rebuild index +DROP INDEX brintest_idx; +CREATE INDEX brintest_idx ON brintest USING brin (a int8_minmax_ops) WITH (pages_per_range = 2); +SELECT brin_index_check('brintest_idx', true); +-- cleanup +DROP TABLE brintest; + + + +-- multi_min_max opclass +CREATE TABLE brintest (a bigint) WITH (fillfactor = 10); +CREATE INDEX brintest_idx ON brintest USING brin (a int8_minmax_multi_ops) WITH (pages_per_range = 2); +INSERT INTO brintest (a) SELECT x FROM generate_series(1,100000) x; +-- create some empty ranges +DELETE FROM brintest WHERE a > 20000 AND a < 40000; +SELECT brin_index_check('brintest_idx', true); + +-- rebuild index +DROP INDEX brintest_idx; +CREATE INDEX brintest_idx ON brintest USING brin (a int8_minmax_multi_ops) WITH (pages_per_range = 2); +SELECT brin_index_check('brintest_idx', true); +-- cleanup +DROP TABLE brintest; + + + +-- bloom opclass +CREATE TABLE brintest (a bigint) WITH (fillfactor = 10); +CREATE INDEX brintest_idx ON brintest USING brin (a int8_bloom_ops) WITH (pages_per_range = 2); +INSERT INTO brintest (a) SELECT x FROM generate_series(1,100000) x; +-- create some empty ranges +DELETE FROM brintest WHERE a > 20000 AND a < 40000; +SELECT brin_index_check('brintest_idx', true); + +-- rebuild index +DROP INDEX brintest_idx; +CREATE INDEX brintest_idx ON brintest USING brin (a int8_bloom_ops) WITH (pages_per_range = 2); +SELECT brin_index_check('brintest_idx', true); +-- cleanup +DROP TABLE brintest; + + +-- inclusion opclass +CREATE TABLE brintest (id serial PRIMARY KEY, a box); +CREATE INDEX brintest_idx ON brintest USING brin (a box_inclusion_ops) WITH (pages_per_range = 2); +INSERT INTO brintest (a) +SELECT box(point(random() * 1000, random() * 1000), point(random() * 1000, random() * 1000)) +FROM generate_series(1, 10000); +-- create some empty ranges +DELETE FROM brintest WHERE id > 2000 AND id < 4000; + +SELECT brin_index_check('brintest_idx', true); + +-- rebuild index +DROP INDEX brintest_idx; +CREATE INDEX brintest_idx ON brintest USING brin (a box_inclusion_ops) WITH (pages_per_range = 2); +SELECT brin_index_check('brintest_idx', true); +-- cleanup +DROP TABLE brintest; + + +-- multiple attributes +CREATE TABLE brintest (id bigserial, a text) WITH (fillfactor = 10); +CREATE INDEX brintest_idx ON brintest USING brin (id int8_minmax_ops, a text_minmax_ops) WITH (pages_per_range = 2); +INSERT INTO brintest (a) SELECT random_string((x % 100)) FROM generate_series(1,3000) x; +-- create some empty ranges +DELETE FROM brintest WHERE id > 1500 AND id < 2500; +SELECT brin_index_check('brintest_idx', true); + +-- rebuild index +DROP INDEX brintest_idx; +CREATE INDEX brintest_idx ON brintest USING brin (id int8_minmax_ops, a text_minmax_ops) WITH (pages_per_range = 2); +SELECT brin_index_check('brintest_idx', true); +-- cleanup +DROP TABLE brintest; + + +-- cleanup +DROP FUNCTION random_string; \ No newline at end of file diff --git a/contrib/amcheck/t/007_verify_brin.pl b/contrib/amcheck/t/007_verify_brin.pl new file mode 100644 index 00000000000..c4073f9bdcc --- /dev/null +++ b/contrib/amcheck/t/007_verify_brin.pl @@ -0,0 +1,301 @@ +# Copyright (c) 2021-2025, PostgreSQL Global Development Group + +use strict; +use warnings FATAL => 'all'; + +use PostgreSQL::Test::Cluster; +use PostgreSQL::Test::Utils; + +use Test::More; + +my $node; +my $blksize; +my $meta_page_blkno = 0; + +# +# Test set-up +# +$node = PostgreSQL::Test::Cluster->new('test'); +$node->init(no_data_checksums => 1); +$node->append_conf('postgresql.conf', 'autovacuum=off'); +$node->start; +$blksize = int($node->safe_psql('postgres', 'SHOW block_size;')); +$node->safe_psql('postgres', q(CREATE EXTENSION amcheck)); + +# Tests + +# Test flow: +# - create all necessary relations and indexes for all test cases +# - stop the node +# - insert corruptions for all test cases +# - start the node +# - assertions +# +# This way we avoid waiting for the node to restart for each test, which speeds up the tests. + +my @tests = ( + { + # invalid meta page type + + find => pack('S', 0xF091), + replace => pack('S', 0xAAAA), + blkno => $meta_page_blkno, + expected => wrap('metapage is corrupted') + }, + { + # invalid meta page magic word + + find => pack('L', 0xA8109CFA), + replace => pack('L', 0xBB109CFB), + blkno => $meta_page_blkno, + expected => wrap('metapage is corrupted'), + }, + { + # invalid meta page index version + + find => pack('L*', 0xA8109CFA, 1), + replace => pack('L*', 0xA8109CFA, 2), + blkno => $meta_page_blkno, + expected => wrap('metapage is corrupted') + }, + { + # pages_per_range below lower limit + + find => pack('L*', 0xA8109CFA, 1, 128), + replace => pack('L*', 0xA8109CFA, 1, 0), + blkno => $meta_page_blkno, + expected => wrap('metapage is corrupted') + }, + { + # pages_per_range above upper limit + + find => pack('L*', 0xA8109CFA, 1, 128), + replace => pack('L*', 0xA8109CFA, 1, 131073), + blkno => $meta_page_blkno, + expected => wrap('metapage is corrupted') + }, + { + # last_revmap_page below lower limit + + find => pack('L*', 0xA8109CFA, 1, 128, 1), + replace => pack('L*', 0xA8109CFA, 1, 128, 0), + blkno => $meta_page_blkno, + expected => wrap('metapage is corrupted'), + }, + { + + # last_revmap_page beyond index relation size + + find => pack('L*', 0xA8109CFA, 1, 128, 1), + replace => pack('L*', 0xA8109CFA, 1, 128, 100), + blkno => $meta_page_blkno, + expected => wrap('metapage is corrupted'), + }, + { + # invalid revmap page type + + find => pack('S', 0xF092), + replace => pack('S', 0xAAAA), + blkno => 1, # revmap page + expected => wrap('revmap page is expected at block 1, last revmap page 1'), + }, + { + # revmap item points beyond index relation size + # replace (2,1) with (100,1) + + find => pack('S*', 0, 2, 1), + replace => pack('S*', 0, 100, 1), + blkno => 1, # revmap page + expected => wrap('revmap item points to a non existing block 100, ' + . 'index max block 2. Range blkno: 0, revmap item: (1,0)') + }, + { + # invalid regular page type + + find => pack('S', 0xF093), + replace => pack('S', 0xAAAA), + blkno => 2, # regular page + expected => wrap('revmap item points to the page which is not regular (blkno: 2). ' + . 'Range blkno: 0, revmap item: (1,0)') + }, + { + # revmap item points beyond regular page max offset + # replace (2,1) with (2,2) + + find => pack('S*', 0, 2, 1), + replace => pack('S*', 0, 2, 2), + blkno => 1, # revmap page + expected => wrap('revmap item offset number 2 is greater than regular page 2 max offset 1. ' + . 'Range blkno: 0, revmap item: (1,0)') + }, + { + # invalid index tuple range blkno + + find => pack('LCC', 0, 0xA8, 0x01), + replace => pack('LCC', 1, 0xA8, 0x01), + blkno => 2, # regular page + expected => wrap('index tuple has invalid blkno 1. Range blkno: 0, revmap item: (1,0), index tuple: (2,1)') + }, + { + # range beyond the table size and is not empty + + find => pack('LCC', 0, 0xA8, 0x01), + replace => pack('LCC', 0, 0x88, 0x01), + blkno => 2, # regular page + expected => wrap('the range is beyond the table size, but is not marked as empty, table size: 0 blocks. ' + . 'Range blkno: 0, revmap item: (1,0), index tuple: (2,1)') + }, + { + # corrupt index tuple data offset + # here 0x00, 0x00, 0x00 is padding and '.' is varlena len byte + + find => pack('LCCCC', 0, 0x08, 0x00, 0x00, 0x00) . '(.)' . 'aaaaa', + replace => pack('LCCCC', 0, 0x1F, 0x00, 0x00, 0x00) . '$1' . 'aaaaa', + blkno => 2, # regular page + table_data => sub { + my ($test_struct) = @_; + return qq(INSERT INTO $test_struct->{table_name} (a) VALUES ('aaaaa');); + }, + expected => qr/index tuple header length 31 is greater than tuple len ..\. \QRange blkno: 0, revmap item: (1,0), index tuple: (2,1)\E/ + }, + { + # empty range index tuple doesn't have null bitmap + + find => pack('LCC', 0, 0xA8, 0x01), + replace => pack('LCC', 0, 0x28, 0x01), + blkno => 2, # regular page + expected => wrap('empty range index tuple doesn\'t have null bitmap. ' + . 'Range blkno: 0, revmap item: (1,0), index tuple: (2,1)') + }, + { + # empty range index tuple all_nulls -> false + + find => pack('LCC', 0, 0xA8, 0x01), + replace => pack('LCC', 0, 0xA8, 0x00), + blkno => 2, # regular page + expected => wrap('empty range index tuple attribute 0 with allnulls is false. ' + . 'Range blkno: 0, revmap item: (1,0), index tuple: (2,1)') + }, + { + # empty range index tuple has_nulls -> true + + find => pack('LCC', 0, 0xA8, 0x01), + replace => pack('LCC', 0, 0xA8, 0x03), + blkno => 2, # regular page + expected => wrap('empty range index tuple attribute 0 with hasnulls is true. ' + . 'Range blkno: 0, revmap item: (1,0), index tuple: (2,1)') + }, + { + # invalid index tuple data + # replace varlena len with FF - should work with any endianness + + find => pack('LCCCC', 0, 0x08, 0x00, 0x00, 0x00) . '.' . 'aaaaa', + replace => pack('LCCCCC', 0, 0x08, 0x00, 0x00, 0x00, 0xFF) . 'aaaaa', + blkno => 2, # regular page + table_data => sub { + my ($test_struct) = @_; + return qq(INSERT INTO $test_struct->{table_name} (a) VALUES ('aaaaa');); + }, + expected => qr/attribute 0 stored value 0 with length -1 ends at offset 127 beyond total tuple length ..\.\Q Range blkno: 0, revmap item: (1,0), index tuple: (2,1)\E/ + }, + { + # orphan index tuple + # replace valid revmap item with (0,0) + + find => pack('S*', 0, 2, 1), + replace => pack('S*', 0, 0, 0), + blkno => 1, # revmap page + table_data => sub { + my ($test_struct) = @_; + return qq(INSERT INTO $test_struct->{table_name} (a) VALUES ('aaaaa');); + }, + expected => wrap("revmap doesn't point to index tuple. Range blkno: 0, revmap item: (1,0), index tuple: (2,1)") + } +); + + +# init test data +my $i = 1; +foreach my $test_struct (@tests) { + + $test_struct->{table_name} = 't' . $i++; + $test_struct->{index_name} = $test_struct->{table_name} . '_brin_idx'; + + my $test_data_sql = ''; + if (exists $test_struct->{table_data}) { + $test_data_sql = $test_struct->{table_data}->($test_struct); + } + + $node->safe_psql('postgres', qq( + CREATE TABLE $test_struct->{table_name} (a TEXT); + $test_data_sql + CREATE INDEX $test_struct->{index_name} ON $test_struct->{table_name} USING BRIN (a); + )); + + $test_struct->{relpath} = relation_filepath($test_struct->{index_name}); +} + +# corrupt index +$node->stop; + +foreach my $test_struct (@tests) { + string_replace_block( + $test_struct->{relpath}, + $test_struct->{find}, + $test_struct->{replace}, + $test_struct->{blkno} + ); +} + +# assertions +$node->start; + +foreach my $test_struct (@tests) { + my ($result, $stdout, $stderr) = $node->psql('postgres', qq(SELECT brin_index_check('$test_struct->{index_name}', true))); + like($stderr, $test_struct->{expected}); +} + + +# Helpers + +# Returns the filesystem path for the named relation. +sub relation_filepath { + my ($relname) = @_; + + my $pgdata = $node->data_dir; + my $rel = $node->safe_psql('postgres', + qq(SELECT pg_relation_filepath('$relname'))); + die "path not found for relation $relname" unless defined $rel; + return "$pgdata/$rel"; +} + +sub string_replace_block { + my ($filename, $find, $replace, $blkno) = @_; + + my $fh; + open($fh, '+<', $filename) or BAIL_OUT("open failed: $!"); + binmode $fh; + + my $offset = $blkno * $blksize; + my $buffer; + + sysseek($fh, $offset, 0) or BAIL_OUT("seek failed: $!"); + sysread($fh, $buffer, $blksize) or BAIL_OUT("read failed: $!"); + + $buffer =~ s/$find/'"' . $replace . '"'/gee; + + sysseek($fh, $offset, 0) or BAIL_OUT("seek failed: $!"); + syswrite($fh, $buffer) or BAIL_OUT("write failed: $!"); + + close($fh) or BAIL_OUT("close failed: $!"); + + return; +} + +sub wrap +{ + my $input = @_; + return qr/\Q$input\E/ +} + +done_testing(); \ No newline at end of file diff --git a/contrib/amcheck/verify_brin.c b/contrib/amcheck/verify_brin.c new file mode 100644 index 00000000000..183d189685c --- /dev/null +++ b/contrib/amcheck/verify_brin.c @@ -0,0 +1,864 @@ +/*------------------------------------------------------------------------- + * + * verify_brin.c + * Functions to check postgresql brin indexes for corruption + * + * Copyright (c) 2016-2025, PostgreSQL Global Development Group + * + * contrib/amcheck/verify_brin.c + *------------------------------------------------------------------------- + */ +#include "postgres.h" + +#include "access/htup_details.h" +#include "access/table.h" +#include "access/tableam.h" +#include "access/transam.h" +#include "access/brin.h" +#include "catalog/index.h" +#include "catalog/pg_am_d.h" +#include "catalog/pg_operator.h" +#include "miscadmin.h" +#include "storage/lmgr.h" +#include "storage/smgr.h" +#include "utils/guc.h" +#include "utils/memutils.h" +#include "access/brin_page.h" +#include "access/brin_revmap.h" +#include "utils/lsyscache.h" +#include "verify_common.h" +#include "utils/builtins.h" +#include "utils/array.h" + + +PG_FUNCTION_INFO_V1(brin_index_check); + +typedef struct BrinCheckState +{ + + /* Check arguments */ + + bool regularpagescheck; + + /* BRIN check common fields */ + + Relation idxrel; + Relation heaprel; + BrinDesc *bdesc; + int natts; + BlockNumber pagesPerRange; + + /* Index structure check fields */ + + BufferAccessStrategy checkstrategy; + BlockNumber idxnblocks; + BlockNumber heapnblocks; + BlockNumber lastRevmapPage; + /* Current range blkno */ + BlockNumber rangeBlkno; + /* Current revmap item */ + BlockNumber revmapBlk; + Buffer revmapbuf; + Page revmappage; + uint32 revmapidx; + /* Current index tuple */ + BlockNumber regpageBlk; + Buffer regpagebuf; + Page regpage; + OffsetNumber regpageoffset; + +} BrinCheckState; + +static void brin_check(Relation idxrel, Relation heaprel, void *callback_state, bool readonly); + +static void check_brin_index_structure(BrinCheckState * pState); + +static void check_meta(BrinCheckState * state); + +static void check_revmap(BrinCheckState * state); + +static void check_revmap_item(BrinCheckState * state); + +static void check_index_tuple(BrinCheckState * state, BrinTuple *tuple, ItemId lp); + +static void check_regular_pages(BrinCheckState * state); + +static bool revmap_points_to_index_tuple(BrinCheckState * state); + +static ItemId PageGetItemIdCareful(BrinCheckState * state); + +static void brin_check_ereport(BrinCheckState * state, const char *fmt); + +static void revmap_item_ereport(BrinCheckState * state, const char *fmt); + +static void index_tuple_ereport(BrinCheckState * state, const char *fmt); + +static void index_tuple_only_ereport(BrinCheckState * state, const char *fmt); + + +Datum +brin_index_check(PG_FUNCTION_ARGS) +{ + Oid indrelid = PG_GETARG_OID(0); + BrinCheckState *state = palloc0(sizeof(BrinCheckState)); + + state->regularpagescheck = PG_GETARG_BOOL(1); + + amcheck_lock_relation_and_check(indrelid, + BRIN_AM_OID, + brin_check, + ShareUpdateExclusiveLock, + state); + + PG_RETURN_VOID(); +} + +/* + * Main check function + */ +static void +brin_check(Relation idxrel, Relation heaprel, void *callback_state, bool readonly) +{ + BrinCheckState *state = (BrinCheckState *) callback_state; + + /* Initialize check common fields */ + state->idxrel = idxrel; + state->heaprel = heaprel; + state->bdesc = brin_build_desc(idxrel); + state->natts = state->bdesc->bd_tupdesc->natts; + + + check_brin_index_structure(state); + + + brin_free_desc(state->bdesc); +} + +/* + * Check that index has expected structure + * + * Some check expectations: + * - we hold ShareUpdateExclusiveLock, so revmap could not be extended (i.e. no evacuation) while check as well as + * all regular pages should stay regular and ranges could not be summarized and desummarized. + * Nevertheless, concurrent updates could lead to new regular page allocations + * and moving of index tuples. + * - if revmap pointer is valid there should be valid index tuple it points to. + * - there are no orphan index tuples (if there is an index tuple, the revmap item points to this tuple also must exist) + * - it's possible to encounter placeholder tuples (as a result of crash) + * - it's possible to encounter new pages instead of regular (as a result of crash) + * - it's possible to encounter pages with evacuation bit (as a result of crash) + * + */ +static void +check_brin_index_structure(BrinCheckState * state) +{ + /* Index structure check fields initialization */ + state->checkstrategy = GetAccessStrategy(BAS_BULKREAD); + + check_meta(state); + + /* Check revmap first, blocks: [1, lastRevmapPage] */ + check_revmap(state); + + + if (state->regularpagescheck) + { + /* Check regular pages, blocks: [lastRevmapPage + 1, idxnblocks] */ + check_regular_pages(state); + } + +} + +/* Meta page check and save some data for the further check */ +static void +check_meta(BrinCheckState * state) +{ + Buffer metabuf; + Page metapage; + BrinMetaPageData *metadata; + + /* Meta page check */ + metabuf = ReadBufferExtended(state->idxrel, MAIN_FORKNUM, BRIN_METAPAGE_BLKNO, RBM_NORMAL, + state->checkstrategy); + LockBuffer(metabuf, BUFFER_LOCK_SHARE); + metapage = BufferGetPage(metabuf); + metadata = (BrinMetaPageData *) PageGetContents(metapage); + state->idxnblocks = RelationGetNumberOfBlocks(state->idxrel); + + + if (!BRIN_IS_META_PAGE(metapage) || + metadata->brinMagic != BRIN_META_MAGIC || + metadata->brinVersion != BRIN_CURRENT_VERSION || + metadata->pagesPerRange < 1 || metadata->pagesPerRange > BRIN_MAX_PAGES_PER_RANGE || + metadata->lastRevmapPage <= BRIN_METAPAGE_BLKNO || metadata->lastRevmapPage >= state->idxnblocks) + { + brin_check_ereport(state, "metapage is corrupted"); + } + + state->lastRevmapPage = metadata->lastRevmapPage; + state->pagesPerRange = metadata->pagesPerRange; + UnlockReleaseBuffer(metabuf); +} + +/* + * This is a main part of the brin index structure check. + * We walk revmap page by page from the beginning and check every revmap item and + * every index tuple pointed from the revmap. + */ +static void +check_revmap(BrinCheckState * state) +{ + Relation idxrel = state->idxrel; + BlockNumber lastRevmapPage = state->lastRevmapPage; + ReadStream *stream; + int stream_flags; + ReadStreamBlockNumberCB stream_cb; + BlockRangeReadStreamPrivate stream_data; + + state->rangeBlkno = 0; + state->regpagebuf = InvalidBuffer; + state->heapnblocks = RelationGetNumberOfBlocks(state->heaprel); + + + /* + * Prepare stream data for revmap walk. It is safe to use batchmode as + * block_range_read_stream_cb takes no locks. + */ + stream_flags = READ_STREAM_SEQUENTIAL | READ_STREAM_USE_BATCHING; + /* First revmap page is right after meta page */ + stream_data.current_blocknum = BRIN_METAPAGE_BLKNO + 1; + stream_data.last_exclusive = lastRevmapPage + 1; + + stream_cb = block_range_read_stream_cb; + stream = read_stream_begin_relation(stream_flags, + GetAccessStrategy(BAS_BULKREAD), + idxrel, + MAIN_FORKNUM, + stream_cb, + &stream_data, + 0); + + /* Walk each revmap page */ + while ((state->revmapbuf = read_stream_next_buffer(stream, NULL)) != InvalidBuffer) + { + state->revmapBlk = BufferGetBlockNumber(state->revmapbuf); + LockBuffer(state->revmapbuf, BUFFER_LOCK_SHARE); + state->revmappage = BufferGetPage(state->revmapbuf); + + /* + * Pages with block numbers in [1, lastRevmapPage] should be revmap + * pages + */ + if (!BRIN_IS_REVMAP_PAGE(state->revmappage)) + { + brin_check_ereport(state, psprintf("revmap page is expected at block %u, last revmap page %u", + state->revmapBlk, + lastRevmapPage)); + } + LockBuffer(state->revmapbuf, BUFFER_LOCK_UNLOCK); + + /* Walk and check all brin tuples from the current revmap page */ + state->revmapidx = 0; + while (state->revmapidx < REVMAP_PAGE_MAXITEMS) + { + CHECK_FOR_INTERRUPTS(); + + /* Check revmap item */ + check_revmap_item(state); + + state->rangeBlkno += state->pagesPerRange; + state->revmapidx++; + } + + elog(DEBUG3, "Complete revmap page check: %d", state->revmapBlk); + + ReleaseBuffer(state->revmapbuf); + } + + read_stream_end(stream); + + if (BufferIsValid(state->regpagebuf)) + { + ReleaseBuffer(state->regpagebuf); + } +} + +/* + * Check revmap item. + * + * We check revmap item pointer itself and if it is ok we check the index tuple it points to. + * + * To avoid deadlock we need to unlock revmap page before locking regular page, + * so when we get the lock on the regular page our index tuple pointer may no longer be relevant. + * So for some checks before reporting an error we need to make sure that our pointer is still relevant and if it's not - retry. + */ +static void +check_revmap_item(BrinCheckState * state) +{ + ItemPointerData *revmaptids; + RevmapContents *contents; + ItemPointerData *iptr; + ItemId lp; + BrinTuple *tup; + Relation idxrel = state->idxrel; + + /* Loop to retry revmap item check if there was a concurrent update. */ + for (;;) + { + LockBuffer(state->revmapbuf, BUFFER_LOCK_SHARE); + + contents = (RevmapContents *) PageGetContents(BufferGetPage(state->revmapbuf)); + revmaptids = contents->rm_tids; + /* Pointer for the range with start at state->rangeBlkno */ + iptr = revmaptids + state->revmapidx; + + /* At first check revmap item pointer */ + + /* + * Tuple pointer is invalid means range isn't summarized, just move + * further + */ + if (!ItemPointerIsValid(iptr)) + { + elog(DEBUG3, "Range %u is not summarized", state->rangeBlkno); + LockBuffer(state->revmapbuf, BUFFER_LOCK_UNLOCK); + break; + } + + /* + * Pointer is valid, it should points to index tuple for the range + * with blkno rangeBlkno. Remember it and unlock revmap page to avoid + * deadlock + */ + state->regpageBlk = ItemPointerGetBlockNumber(iptr); + state->regpageoffset = ItemPointerGetOffsetNumber(iptr); + + LockBuffer(state->revmapbuf, BUFFER_LOCK_UNLOCK); + + /* + * Check if the regpage block number is greater than the relation + * size. To avoid fetching the number of blocks for each tuple, use + * cached value first + */ + if (state->regpageBlk >= state->idxnblocks) + { + /* + * Regular pages may have been added, so refresh idxnblocks and + * recheck + */ + state->idxnblocks = RelationGetNumberOfBlocks(idxrel); + if (state->regpageBlk >= state->idxnblocks) + { + revmap_item_ereport(state, + psprintf("revmap item points to a non existing block %u, index max block %u", + state->regpageBlk, + state->idxnblocks - 1)); + } + } + + /* + * To avoid some pin/unpin cycles we cache last used regular page. + * Check if we need different regular page and fetch it. + */ + if (!BufferIsValid(state->regpagebuf) || BufferGetBlockNumber(state->regpagebuf) != state->regpageBlk) + { + if (BufferIsValid(state->regpagebuf)) + { + ReleaseBuffer(state->regpagebuf); + } + state->regpagebuf = ReadBufferExtended(idxrel, MAIN_FORKNUM, state->regpageBlk, RBM_NORMAL, + state->checkstrategy); + } + + LockBuffer(state->regpagebuf, BUFFER_LOCK_SHARE); + state->regpage = BufferGetPage(state->regpagebuf); + + /* Revmap should always point to a regular page */ + if (!BRIN_IS_REGULAR_PAGE(state->regpage)) + { + revmap_item_ereport(state, + psprintf("revmap item points to the page which is not regular (blkno: %u)", + state->regpageBlk)); + + } + + /* Check item offset is valid */ + if (state->regpageoffset > PageGetMaxOffsetNumber(state->regpage)) + { + + /* If concurrent update moved our tuple we need to retry */ + if (!revmap_points_to_index_tuple(state)) + { + LockBuffer(state->regpagebuf, BUFFER_LOCK_UNLOCK); + continue; + } + + revmap_item_ereport(state, + psprintf("revmap item offset number %u is greater than regular page %u max offset %u", + state->regpageoffset, + state->regpageBlk, + PageGetMaxOffsetNumber(state->regpage))); + } + + elog(DEBUG3, "Process range: %u, iptr: (%u,%u)", state->rangeBlkno, state->regpageBlk, state->regpageoffset); + + /* + * Revmap pointer is OK. It points to existing regular page, offset + * also is ok. Let's check index tuple it points to. + */ + + lp = PageGetItemIdCareful(state); + + /* Revmap should point to NORMAL tuples only */ + if (!ItemIdIsUsed(lp)) + { + + /* If concurrent update moved our tuple we need to retry */ + if (!revmap_points_to_index_tuple(state)) + { + LockBuffer(state->regpagebuf, BUFFER_LOCK_UNLOCK); + continue; + } + + index_tuple_ereport(state, "revmap item points to unused index tuple"); + } + + + tup = (BrinTuple *) PageGetItem(state->regpage, lp); + + /* Check if range block number is as expected */ + if (tup->bt_blkno != state->rangeBlkno) + { + + /* If concurrent update moved our tuple we need to retry */ + if (!revmap_points_to_index_tuple(state)) + { + LockBuffer(state->regpagebuf, BUFFER_LOCK_UNLOCK); + continue; + } + + index_tuple_ereport(state, psprintf("index tuple has invalid blkno %u", tup->bt_blkno)); + } + + /* + * If the range is beyond the table size - the range must be empty. + * It's valid situation for empty table now. + */ + if (state->rangeBlkno >= state->heapnblocks) + { + if (!BrinTupleIsEmptyRange(tup)) + { + index_tuple_ereport(state, + psprintf("the range is beyond the table size, " + "but is not marked as empty, table size: %u blocks", + state->heapnblocks)); + } + } + + /* Check index tuple itself */ + check_index_tuple(state, tup, lp); + + LockBuffer(state->regpagebuf, BUFFER_LOCK_UNLOCK); + break; + } +} + +/* + * Check that index tuple has expected structure. + * + * This function follows the logic performed by brin_deform_tuple(). + * After this check is complete we are sure that brin_deform_tuple can process it. + * + * In case of empty range check that for all attributes allnulls are true, hasnulls are false and + * there is no data. All core opclasses expect allnulls is true for empty range. + */ +static void +check_index_tuple(BrinCheckState * state, BrinTuple *tuple, ItemId lp) +{ + + char *tp; /* tuple data */ + uint16 off; + bits8 *nullbits; + TupleDesc disktdesc; + int stored; + bool empty_range = BrinTupleIsEmptyRange(tuple); + bool hasnullbitmap = BrinTupleHasNulls(tuple); + uint8 hoff = BrinTupleDataOffset(tuple); + uint16 tuplen = ItemIdGetLength(lp); + + + /* Check that header length is not greater than tuple length */ + if (hoff > tuplen) + { + index_tuple_ereport(state, psprintf("index tuple header length %u is greater than tuple len %u", hoff, tuplen)); + } + + /* If tuple has null bitmap - initialize it */ + if (hasnullbitmap) + { + nullbits = (bits8 *) ((char *) tuple + SizeOfBrinTuple); + } + else + { + nullbits = NULL; + } + + /* Empty range index tuple checks */ + if (empty_range) + { + /* Empty range tuple should have null bitmap */ + if (!hasnullbitmap) + { + index_tuple_ereport(state, "empty range index tuple doesn't have null bitmap"); + } + + Assert(nullbits != NULL); + + /* Check every attribute has allnulls is true and hasnulls is false */ + for (int attindex = 0; attindex < state->natts; ++attindex) + { + + /* Attribute allnulls should be true for empty range */ + if (att_isnull(attindex, nullbits)) + { + index_tuple_ereport(state, + psprintf("empty range index tuple attribute %d with allnulls is false", + attindex)); + } + + /* Attribute hasnulls should be false for empty range */ + if (!att_isnull(state->natts + attindex, nullbits)) + { + index_tuple_ereport(state, + psprintf("empty range index tuple attribute %d with hasnulls is true", + attindex)); + } + } + + /* We are done with empty range tuple */ + return; + } + + /* + * Range is marked as not empty so we can have some data in the tuple. + * Walk all attributes and checks that all stored values fit into the + * tuple + */ + + tp = (char *) tuple + BrinTupleDataOffset(tuple); + stored = 0; + off = 0; + + disktdesc = brin_tuple_tupdesc(state->bdesc); + + for (int attindex = 0; attindex < state->natts; ++attindex) + { + BrinOpcInfo *opclass = state->bdesc->bd_info[attindex]; + + /* + * if allnulls is set we have no data for this attribute, move to the + * next + */ + if (hasnullbitmap && !att_isnull(attindex, nullbits)) + { + stored += opclass->oi_nstored; + continue; + } + + /* Walk all stored values for the current attribute */ + for (int datumno = 0; datumno < opclass->oi_nstored; datumno++) + { + CompactAttribute *thisatt = TupleDescCompactAttr(disktdesc, stored); + + if (thisatt->attlen == -1) + { + off = att_pointer_alignby(off, + thisatt->attalignby, + -1, + tp + off); + } + else + { + off = att_nominal_alignby(off, thisatt->attalignby); + } + + /* Check that we are still in the tuple */ + if (hoff + off > tuplen) + { + index_tuple_ereport(state, + psprintf("attribute %u stored value %u with length %d " + "starts at offset %u beyond total tuple length %u", + attindex, datumno, thisatt->attlen, off, tuplen)); + } + + off = att_addlength_pointer(off, thisatt->attlen, tp + off); + + /* Check that we are still in the tuple */ + if (hoff + off > tuplen) + { + index_tuple_ereport(state, + psprintf("attribute %u stored value %u with length %d " + "ends at offset %u beyond total tuple length %u", + attindex, datumno, thisatt->attlen, off, tuplen)); + } + stored++; + } + + } + +} + +/* + * At the moment we should have been already check that every index + * tuple in the regular pages has valid structure and range blkno + * (because every normal index tuple must have pointer in the revmap and + * we followed every such pointer in check_revmap). So here we just want + * to do some additional checks to be sure that there is nothing wrong + * with the regular pages [lastRevmapPage + 1, indexnblocks]: + * - there is a pointer in revmap to each NORMAL index tuple + * (no orphans index tuples) + * - all pages have expected type (REGULAR). We can encounter new pages as + * result of crash, so we just skip such pages. + */ +static void +check_regular_pages(BrinCheckState * state) +{ + ReadStream *stream; + int stream_flags; + ReadStreamBlockNumberCB stream_cb; + BlockRangeReadStreamPrivate stream_data; + + /* reset state */ + state->revmapBlk = InvalidBlockNumber; + state->revmapbuf = InvalidBuffer; + state->revmapidx = -1; + state->regpageBlk = InvalidBlockNumber; + state->regpagebuf = InvalidBuffer; + state->regpageoffset = InvalidOffsetNumber; + state->idxnblocks = RelationGetNumberOfBlocks(state->idxrel); + + + /* + * Prepare stream data for regular pages walk. It is safe to use batchmode + * as block_range_read_stream_cb takes no locks. + */ + stream_flags = READ_STREAM_SEQUENTIAL | READ_STREAM_USE_BATCHING | READ_STREAM_FULL; + /* First regular page is right after the last revmap page */ + stream_data.current_blocknum = state->lastRevmapPage + 1; + stream_data.last_exclusive = state->idxnblocks; + + stream_cb = block_range_read_stream_cb; + stream = read_stream_begin_relation(stream_flags, + GetAccessStrategy(BAS_BULKREAD), + state->idxrel, + MAIN_FORKNUM, + stream_cb, + &stream_data, + 0); + + while ((state->regpagebuf = read_stream_next_buffer(stream, NULL)) != InvalidBuffer) + { + OffsetNumber maxoff; + + state->regpageBlk = BufferGetBlockNumber(state->regpagebuf); + LockBuffer(state->regpagebuf, BUFFER_LOCK_SHARE); + state->regpage = BufferGetPage(state->regpagebuf); + + /* Skip new pages */ + if (PageIsNew(state->regpage)) + { + UnlockReleaseBuffer(state->regpagebuf); + continue; + } + + if (!BRIN_IS_REGULAR_PAGE(state->regpage)) + { + brin_check_ereport(state, psprintf("expected new or regular page at block %u", state->regpageBlk)); + } + + /* Check that all NORMAL index tuples within the page are not orphans */ + maxoff = PageGetMaxOffsetNumber(state->regpage); + for (state->regpageoffset = FirstOffsetNumber; state->regpageoffset <= maxoff; state->regpageoffset++) + { + ItemId lp; + BrinTuple *tup; + BlockNumber revmapBlk; + + lp = PageGetItemIdCareful(state); + + if (ItemIdIsUsed(lp)) + { + tup = (BrinTuple *) PageGetItem(state->regpage, lp); + + /* Get revmap block number for index tuple blkno */ + revmapBlk = ((tup->bt_blkno / state->pagesPerRange) / REVMAP_PAGE_MAXITEMS) + 1; + if (revmapBlk > state->lastRevmapPage) + { + index_tuple_only_ereport(state, psprintf("no revmap page for the index tuple with blkno %u", + tup->bt_blkno)); + } + + /* Fetch another revmap page if needed */ + if (state->revmapBlk != revmapBlk) + { + if (BlockNumberIsValid(state->revmapBlk)) + { + ReleaseBuffer(state->revmapbuf); + } + state->revmapBlk = revmapBlk; + state->revmapbuf = ReadBufferExtended(state->idxrel, MAIN_FORKNUM, state->revmapBlk, RBM_NORMAL, + state->checkstrategy); + } + + state->revmapidx = (tup->bt_blkno / state->pagesPerRange) % REVMAP_PAGE_MAXITEMS; + state->rangeBlkno = tup->bt_blkno; + + /* check that revmap item points to index tuple */ + if (!revmap_points_to_index_tuple(state)) + { + index_tuple_ereport(state, psprintf("revmap doesn't point to index tuple")); + } + + } + } + + UnlockReleaseBuffer(state->regpagebuf); + } + + read_stream_end(stream); + + if (state->revmapbuf != InvalidBuffer) + { + ReleaseBuffer(state->revmapbuf); + } +} + +/* + * Check if the revmap item points to the index tuple (regpageBlk, regpageoffset). + * We have locked reg page, and lock revmap page here. + * It's a valid lock ordering, so no deadlock is possible. + */ +static bool +revmap_points_to_index_tuple(BrinCheckState * state) +{ + ItemPointerData *revmaptids; + RevmapContents *contents; + ItemPointerData *tid; + bool points; + + LockBuffer(state->revmapbuf, BUFFER_LOCK_SHARE); + contents = (RevmapContents *) PageGetContents(BufferGetPage(state->revmapbuf)); + revmaptids = contents->rm_tids; + tid = revmaptids + state->revmapidx; + + points = ItemPointerGetBlockNumberNoCheck(tid) == state->regpageBlk && + ItemPointerGetOffsetNumberNoCheck(tid) == state->regpageoffset; + + LockBuffer(state->revmapbuf, BUFFER_LOCK_UNLOCK); + return points; +} + +/* + * PageGetItemId() wrapper that validates returned line pointer. + * + * itemId in brin index could be UNUSED or NORMAL. + */ +static ItemId +PageGetItemIdCareful(BrinCheckState * state) +{ + Page page = state->regpage; + OffsetNumber offset = state->regpageoffset; + ItemId itemid = PageGetItemId(page, offset); + + if (ItemIdGetOffset(itemid) + ItemIdGetLength(itemid) > + BLCKSZ - MAXALIGN(sizeof(BrinSpecialSpace))) + index_tuple_ereport(state, + psprintf("line pointer points past end of tuple space in index. " + "lp_off=%u, lp_len=%u lp_flags=%u", + ItemIdGetOffset(itemid), + ItemIdGetLength(itemid), + ItemIdGetFlags(itemid) + ) + ); + + /* Verify that line pointer is LP_NORMAL or LP_UNUSED */ + if (!((ItemIdIsNormal(itemid) && ItemIdHasStorage(itemid)) || + (!ItemIdIsUsed(itemid) && !ItemIdHasStorage(itemid)))) + { + index_tuple_ereport(state, + psprintf("invalid line pointer storage in index. " + "lp_off=%u, lp_len=%u lp_flags=%u", + ItemIdGetOffset(itemid), + ItemIdGetLength(itemid), + ItemIdGetFlags(itemid) + )); + } + + return itemid; +} + + +/* Report without any additional info */ +static void +brin_check_ereport(BrinCheckState * state, const char *fmt) +{ + ereport(ERROR, + (errcode(ERRCODE_INDEX_CORRUPTED), + errmsg("Index %s is corrupted - %s", RelationGetRelationName(state->idxrel), fmt))); +} + +/* Report with range blkno, revmap item info, index tuple info */ +void +index_tuple_ereport(BrinCheckState * state, const char *fmt) +{ + Assert(state->rangeBlkno != InvalidBlockNumber); + Assert(state->revmapBlk != InvalidBlockNumber); + Assert(state->revmapidx >= 0 && state->revmapidx < REVMAP_PAGE_MAXITEMS); + Assert(state->regpageBlk != InvalidBlockNumber); + Assert(state->regpageoffset != InvalidOffsetNumber); + + ereport(ERROR, + (errcode(ERRCODE_INDEX_CORRUPTED), + errmsg("Index %s is corrupted - %s. Range blkno: %u, revmap item: (%u,%u), index tuple: (%u,%u)", + RelationGetRelationName(state->idxrel), + fmt, + state->rangeBlkno, + state->revmapBlk, + state->revmapidx, + state->regpageBlk, + state->regpageoffset))); +} + +/* Report with index tuple info */ +void +index_tuple_only_ereport(BrinCheckState * state, const char *fmt) +{ + Assert(state->regpageBlk != InvalidBlockNumber); + Assert(state->regpageoffset != InvalidOffsetNumber); + + ereport(ERROR, + (errcode(ERRCODE_INDEX_CORRUPTED), + errmsg("Index %s is corrupted - %s. Index tuple: (%u,%u)", + RelationGetRelationName(state->idxrel), + fmt, + state->regpageBlk, + state->regpageoffset))); +} + +/* Report with range blkno, revmap item info */ +void +revmap_item_ereport(BrinCheckState * state, const char *fmt) +{ + Assert(state->rangeBlkno != InvalidBlockNumber); + Assert(state->revmapBlk != InvalidBlockNumber); + Assert(state->revmapidx >= 0 && state->revmapidx < REVMAP_PAGE_MAXITEMS); + + ereport(ERROR, + (errcode(ERRCODE_INDEX_CORRUPTED), + errmsg("Index %s is corrupted - %s. Range blkno: %u, revmap item: (%u,%u).", + RelationGetRelationName(state->idxrel), + fmt, + state->rangeBlkno, + state->revmapBlk, + state->revmapidx))); +} diff --git a/doc/src/sgml/amcheck.sgml b/doc/src/sgml/amcheck.sgml index 0aff0a6c8c6..2f76af907ec 100644 --- a/doc/src/sgml/amcheck.sgml +++ b/doc/src/sgml/amcheck.sgml @@ -231,6 +231,39 @@ SET client_min_messages = DEBUG1; + + + brin_index_check(index regclass, regularpagescheck boolean) returns void + + brin_index_check + + + + + + brin_index_check tests that its target BRIN index + has no structural corruptions. A ShareUpdateExclusiveLock + is required on the target index by brin_index_check. + + + The following optional arguments are recognized: + + + + regularpagescheck + + + If true, check does another run over all regular pages and tries to + find some corruptions that was not possible to find during the basic check. + + + Defaults to false. + + + + + + -- 2.43.0