BUG #17874: Incorrect memory access at gistBuildCallback
The following bug has been logged on the website:
Bug reference: 17874
Logged by: Egor Chindyaskin
Email address: kyzevan23@mail.ru
PostgreSQL version: 15.2
Operating system: Ubuntu 22.04
Description:
With a server compiled with flag -DUSE_VALGRIND the following query:
create table test(a int4, b point);
create index test_index on test using gist(b) with (buffering = on,
fillfactor=50);
drop index test_index;
insert into test (a, b) select g, point(g*10, g*10) from generate_series(1,
10000) g;
vacuum analyze test;
create index test_index on test using gist(b) with (buffering = on,
fillfactor=50);
drop index test_index;
insert into test (a, b) select g, point(g*10, g*10) from generate_series(1,
10000) g;
vacuum analyze test;
create index test_index on test using gist(b) with (buffering = on,
fillfactor=50);
drop index test_index;
insert into test (a, b) select g, point(g*10, g*10) from generate_series(1,
10000) g;
vacuum analyze test;
create index test_index on test using gist(b) with (buffering = on,
fillfactor=50);
drop index test_index;
insert into test (a, b) select g, point(g*10, g*10) from generate_series(1,
10000) g;
vacuum analyze test;
causes an incorrect memory access:
==00:00:00:20.465 256552== Invalid read of size 2
==00:00:00:20.465 256552== at 0x20EE4F: gistBuildCallback
(gistbuild.c:921)
==00:00:00:20.465 256552== by 0x241838: heapam_index_build_range_scan
(heapam_handler.c:1794)
==00:00:00:20.465 256552== by 0x20F308: table_index_build_scan
(tableam.h:1791)
==00:00:00:20.465 256552== by 0x20F308: gistbuild (gistbuild.c:317)
==00:00:00:20.465 256552== by 0x2DC8A9: index_build (index.c:3032)
==00:00:00:20.465 256552== by 0x2DE3EC: index_create (index.c:1264)
==00:00:00:20.465 256552== by 0x3946F5: DefineIndex (indexcmds.c:1164)
==00:00:00:20.465 256552== by 0x5E6565: ProcessUtilitySlow
(utility.c:1552)
==00:00:00:20.465 256552== by 0x5E593C: standard_ProcessUtility
(utility.c:1077)
==00:00:00:20.465 256552== by 0x5E5A01: ProcessUtility (utility.c:530)
==00:00:00:20.465 256552== by 0x5E2DDE: PortalRunUtility
(pquery.c:1158)
==00:00:00:20.465 256552== by 0x5E30DB: PortalRunMulti (pquery.c:1315)
==00:00:00:20.465 256552== by 0x5E34DD: PortalRun (pquery.c:791)
==00:00:00:20.465 256552== Address 0xeab8f1e is 382 bytes inside a block of
size 8,192 alloc'd
==00:00:00:20.465 256552== at 0x4848899: malloc (in
/usr/libexec/valgrind/vgpreload_memcheck-amd64-linux.so)
==00:00:00:20.465 256552== by 0x74FFAD: AllocSetContextCreateInternal
(aset.c:446)
==00:00:00:20.465 256552== by 0x7383F2: hash_create (dynahash.c:385)
==00:00:00:20.465 256552== by 0x20F852: gistInitBuildBuffers
(gistbuildbuffers.c:82)
==00:00:00:20.465 256552== by 0x20D5AE: gistInitBuffering
(gistbuild.c:833)
==00:00:00:20.465 256552== by 0x20EF09: gistBuildCallback
(gistbuild.c:955)
==00:00:00:20.465 256552== by 0x241838: heapam_index_build_range_scan
(heapam_handler.c:1794)
==00:00:00:20.465 256552== by 0x20F308: table_index_build_scan
(tableam.h:1791)
==00:00:00:20.465 256552== by 0x20F308: gistbuild (gistbuild.c:317)
==00:00:00:20.465 256552== by 0x2DC8A9: index_build (index.c:3032)
==00:00:00:20.465 256552== by 0x2DE3EC: index_create (index.c:1264)
==00:00:00:20.465 256552== by 0x3946F5: DefineIndex (indexcmds.c:1164)
==00:00:00:20.465 256552== by 0x5E6565: ProcessUtilitySlow
(utility.c:1552)
==00:00:00:20.465 256552==
{
<insert_a_suppression_name_here>
Memcheck:Addr2
fun:gistBuildCallback
fun:heapam_index_build_range_scan
fun:table_index_build_scan
fun:gistbuild
fun:index_build
fun:index_create
fun:DefineIndex
fun:ProcessUtilitySlow
fun:standard_ProcessUtility
fun:ProcessUtility
fun:PortalRunUtility
fun:PortalRunMulti
fun:PortalRun
}
==00:00:00:20.465 256552==
==00:00:00:20.465 256552== Exit program on first error
(--exit-on-first-error=yes)
This issue reproduces on all supported branches.
Hi Egor,
28.03.2023 19:01, PG Bug reporting form wrote:
The following bug has been logged on the website:
Bug reference: 17874
Logged by: Egor Chindyaskin
Email address: kyzevan23@mail.ru
PostgreSQL version: 15.2
Operating system: Ubuntu 22.04
Description:With a server compiled with flag -DUSE_VALGRIND the following query:
create table test(a int4, b point);
create index test_index on test using gist(b) with (buffering = on,
fillfactor=50);
drop index test_index;
insert into test (a, b) select g, point(g*10, g*10) from generate_series(1,
10000) g;
vacuum analyze test;...
causes an incorrect memory access:
==00:00:00:20.465 256552== Invalid read of size 2
==00:00:00:20.465 256552== at 0x20EE4F: gistBuildCallback
(gistbuild.c:921)
...This issue reproduces on all supported branches.
That issue was reported already as bug #16329:
/messages/by-id/16329-7a6aa9b6fa1118a1@postgresql.org
Maybe it's worth to register the fix proposed there for the next commitfest...
Best regards,
Alexander
On 29 Mar 2023, at 12:00, Alexander Lakhin <exclusion@gmail.com> wrote:
That issue was reported already as bug #16329:
/messages/by-id/16329-7a6aa9b6fa1118a1@postgresql.orgMaybe it's worth to register the fix proposed there for the next commitfest...
Summarizing that thread and registering the proposed patch for the next
commitfest seems like the right thing to do.
--
Daniel Gustafsson
Daniel Gustafsson <daniel@yesql.se> writes:
On 29 Mar 2023, at 12:00, Alexander Lakhin <exclusion@gmail.com> wrote:
That issue was reported already as bug #16329:
/messages/by-id/16329-7a6aa9b6fa1118a1@postgresql.org
Maybe it's worth to register the fix proposed there for the next commitfest...
Summarizing that thread and registering the proposed patch for the next
commitfest seems like the right thing to do.
Yeah. We did fix the lack of test coverage that much of that thread
was about, but it seems like we forgot the original bug (which I
guess indicates that the coverage is still inadequate, since the
valgrind buildfarm animals haven't noticed it...)
regards, tom lane
I wrote:
Daniel Gustafsson <daniel@yesql.se> writes:
Summarizing that thread and registering the proposed patch for the next
commitfest seems like the right thing to do.
Yeah.
I went ahead and reviewed/pushed this, so nevermind about a CF entry.
regards, tom lane