On Wed, 30 Jul 2025 at 19:07, Japin Li <japi...@hotmail.com> wrote: > On Tue, 29 Jul 2025 at 06:57, Peter Smith <smithpb2...@gmail.com> wrote: >> Here are the latest v15 patches. >> >> Changes include: >> >> PATCH 0002. >> - README now says user should not tamper with VCI internal relations >> - fixes/test the VACUUM bug -- fix provided by Japin [1] >> - fixes/tests the reported segv for attempted REFRESH of VCI internal >> relation -- see [2 comment#1] >> - fixes/tests VCI internal relation dependency on the indexed table >> - simplifies code for PG_TEMP_FILES_DIR -- see [2 comment#2] >> > > Hi Peter, > > Thanks for updating the patches. > > 1. > I've identified another TRAP failure. Here are the reproduction steps: > > rm -rf demo > initdb -D demo > cat <<EOF >>demo/postgresql.auto.conf > shared_preload_libraries = 'vci' > max_worker_processes = '20' > EOF > > pg_ctl -D demo start > > cat <<EOF | psql postgres > CREATE EXTENSION vci; > CREATE TABLE t (id int, info text); > CREATE INDEX ON t USING vci (id); > INSERT INTO t SELECT id, md5(random()::text) FROM generate_series(1, 1000) id; > REINDEX TABLE t; > REINDEX TABLE t; > EOF > > The current VCI design doesn't support REINDEX, which is expected. But I've > discovered an unexpected issue: running REINDEX on a table a second time > causes > an assertion failure. >
It appears VCI doesn't support reindexing; add_reindex_index_hook() returns false, and reindex_index() exits without restoring the security context. PFA. -- Regards, Japin Li ChengDu WenWu Information Technology Co., Ltd.
>From a142adc318d9ca423b4c8f98502c1aa0a60cb8f7 Mon Sep 17 00:00:00 2001 From: Japin Li <japi...@hotmail.com> Date: Thu, 31 Jul 2025 16:09:13 +0800 Subject: [PATCH] Restore security context for add_reindex_index_hook --- contrib/vci/expected/bugs.out | 7 +++++++ contrib/vci/sql/bugs.sql | 9 +++++++++ src/backend/catalog/index.c | 6 ++++++ 3 files changed, 22 insertions(+) diff --git a/contrib/vci/expected/bugs.out b/contrib/vci/expected/bugs.out index a32ea0ef983..4ebbde8a80f 100644 --- a/contrib/vci/expected/bugs.out +++ b/contrib/vci/expected/bugs.out @@ -66,3 +66,10 @@ SELECT relname FROM pg_class WHERE relname ~ '^vci_*'; --------- (0 rows) +-- Bug reported by Japin Li that forgot to restore security context +-- See https://www.postgresql.org/message-id/ME0P300MB0445827B6E9CC04E0FAEE446B624A%40ME0P300MB0445.AUSP300.PROD.OUTLOOK.COM +CREATE TABLE t5 (id int, info text); +CREATE INDEX t5_idx ON t5 USING vci (id); +REINDEX TABLE t5; +REINDEX TABLE t5; +DROP TABLE t5; diff --git a/contrib/vci/sql/bugs.sql b/contrib/vci/sql/bugs.sql index 2675f514c21..54d9c321cac 100644 --- a/contrib/vci/sql/bugs.sql +++ b/contrib/vci/sql/bugs.sql @@ -36,3 +36,12 @@ CREATE INDEX t4_idx ON t4 USING vci (id); SELECT relname FROM pg_class WHERE relname ~ '^vci_*' ORDER BY relname; DROP TABLE t4; SELECT relname FROM pg_class WHERE relname ~ '^vci_*'; + +-- Bug reported by Japin Li that forgot to restore security context +-- See https://www.postgresql.org/message-id/ME0P300MB0445827B6E9CC04E0FAEE446B624A%40ME0P300MB0445.AUSP300.PROD.OUTLOOK.COM + +CREATE TABLE t5 (id int, info text); +CREATE INDEX t5_idx ON t5 USING vci (id); +REINDEX TABLE t5; +REINDEX TABLE t5; +DROP TABLE t5; diff --git a/src/backend/catalog/index.c b/src/backend/catalog/index.c index dd67530fdf5..fedbece872a 100644 --- a/src/backend/catalog/index.c +++ b/src/backend/catalog/index.c @@ -3704,6 +3704,12 @@ reindex_index(const ReindexStmt *stmt, Oid indexId, { RemoveReindexPending(RelationGetRelid(iRel)); + /* Roll back any GUC changes */ + AtEOXact_GUC(false, save_nestlevel); + + /* Restore userid and security context */ + SetUserIdAndSecContext(save_userid, save_sec_context); + /* Close rels, but keep locks */ index_close(iRel, NoLock); table_close(heapRelation, NoLock); -- 2.43.0
