Re: BUG #19918: SIGSEGV in auto_explain + EXPLAIN (BUFFERS)

From: Andrey Rachitskiy <pl0h0yp1(at)gmail(dot)com>
To: nakamurajames123(at)gmail(dot)com, pgsql-bugs(at)lists(dot)postgresql(dot)org
Subject: Re: BUG #19918: SIGSEGV in auto_explain + EXPLAIN (BUFFERS)
Date: 2026-11-19 11:55:13
Message-ID: CAB8bMit5FV-cTVPLtyG_B1iu97L_-LieJeouZxO0LM+AF-nmxA@mail.gmail.com
Views: Whole Thread | Raw Message | Download mbox | Resend email
Thread:
Lists: pgsql-bugs

чт, 19 нояб. 2026 г. в 16:16, PG Bug reporting form <noreply(at)postgresql(dot)org
>:

> The following bug has been logged on the website:
>
> Bug reference: 19918
> Logged by: Fin
> Email address: nakamurajames123(at)gmail(dot)com
> PostgreSQL version: 19rc2
> Operating system: wsl linux
> Description:
>
> Reproduction Steps
>
> Server configuration (postgresql.conf):
>
> shared_preload_libraries = 'auto_explain'
> auto_explain.log_min_duration = 0
> auto_explain.log_analyze = on
>
> Execute as a regular user:
>
> CREATE TABLE t (id int PRIMARY KEY, v text);
> INSERT INTO t SELECT g, 'x' FROM generate_series(1, 1000) g;
> ANALYZE t;
> EXPLAIN (BUFFERS) SELECT * FROM t WHERE id = 42;
>
>
> Expected Behavior
>
> Since ANALYZE is not specified in the EXPLAIN command, the query should not
> be executed, and only the estimated execution plan should be displayed
> successfully.
>
> auto_explain should not treat a plan that was never executed as though it
> had been analyzed. If a plan is logged, it should not include actual
> execution statistics.
>
>
> Actual Behavior
>
> The backend crashes with SIGSEGV when the final EXPLAIN (BUFFERS) statement
> is executed. The postmaster terminates all other active sessions and
> initiates crash recovery.
>
> The issue is consistently reproducible whenever the execution plan contains
> an Index Scan.
>
> Both master (20devel, 0cf2d43d326) and PostgreSQL 19.0 (REL_19_STABLE,
> 19aa5713be2) are affected.
>
>
> Logs
>
> 1. psql output:
>
> CREATE TABLE
> INSERT 0 1000
> ANALYZE
> psql:r.sql:5: server closed the connection unexpectedly
> This probably means the server terminated abnormally
> before or while processing the request.
> psql:r.sql:5: error: connection to server was lost
>
>
> 2. Server log (build without sanitizers, cassert enabled, PostgreSQL
> 20devel
> on x86_64-linux, compiled by clang-22.1.8, 64-bit):
>
> 2026-11-19 10:19:10.867 KST [3178278] LOG: client backend (PID 3178297)
> was
> terminated by signal 11: Segmentation fault
> 2026-11-19 10:19:10.867 KST [3178278] DETAIL: Failed process was running:
> EXPLAIN (BUFFERS) SELECT * FROM t WHERE id = 42;
> 2026-11-19 10:19:10.867 KST [3178278] LOG: terminating any other active
> server processes
> 2026-11-19 10:19:10.868 KST [3178278] LOG: all server processes
> terminated;
> reinitializing
> 2026-11-19 10:19:10.898 KST [3178303] LOG: database system was
> interrupted;
> last known up at 2026-11-19 10:19:10 KST
> 2026-11-19 10:19:10.904 KST [3178303] LOG: database system was not
> properly
> shut down; automatic recovery in progress
> 2026-11-19 10:19:10.904 KST [3178303] LOG: redo starts at 0/0175C938
> 2026-11-19 10:19:10.907 KST [3178303] LOG: invalid record length at
> 0/017B7400: expected at least 24, got 0
> 2026-11-19 10:19:10.907 KST [3178303] LOG: redo done at 0/017B7348 system
> usage: CPU: user: 0.00 s, system: 0.00 s, elapsed: 0.00 s
> 2026-11-19 10:19:10.907 KST [3178304] LOG: checkpoint starting:
> end-of-recovery fast wait
> 2026-11-19 10:19:10.908 KST [3178304] LOG: checkpoint complete:
> end-of-recovery fast wait: wrote 59 buffers (0.4%), wrote 3 SLRU buffers; 0
> WAL file(s) added, 0 removed, 0 recycled; write=0.001 s, sync=0.001 s,
> total=0.001 s; sync files=53, longest=0.001 s, average=0.000 s;
> distance=362
> kB, estimate=362 kB
> 2026-11-19 10:19:10.909 KST [3178278] LOG: database system is ready to
> accept connections
>
>
> 3. UBSan stack trace (crash location; source path prefixes have been
> shortened):
>
> src/backend/commands/explain.c:3992:45: runtime error: member access within
> null pointer of type 'IndexScanInstrumentation' (aka 'struct
> IndexScanInstrumentation')
> #0 show_indexscan_info src/backend/commands/explain.c:3992:45
> #1 ExplainNode src/backend/commands/explain.c:1983:4
> #2 ExplainPrintPlan src/backend/commands/explain.c:811:2
> #3 explain_ExecutorEnd contrib/auto_explain/auto_explain.c:457:4
> #4 ExplainOnePlan src/backend/commands/explain.c:669:2
> #5 standard_ExplainOneQuery src/backend/commands/explain.c:380:2
> #6 ExplainOneQuery src/backend/commands/explain.c:317:3
> #7 ExplainQuery src/backend/commands/explain.c:230:4
> #8 standard_ProcessUtility src/backend/tcop/utility.c:868:4
> #9 ProcessUtility src/backend/tcop/utility.c:525:3
> #10 PortalRunUtility src/backend/tcop/pquery.c:1256:2
> #11 FillPortalStore src/backend/tcop/pquery.c:1129:4
> #12 PortalRun src/backend/tcop/pquery.c:863:6
> #13 exec_simple_query src/backend/tcop/postgres.c:1302:10
> #14 PostgresMain src/backend/tcop/postgres.c
> #15 BackendMain src/backend/tcop/backend_startup.c:137:2
> #16 postmaster_child_launch
> src/backend/postmaster/launch_backend.c:268:3
> #17 BackendStartup src/backend/postmaster/postmaster.c:3651:8
> #18 ServerLoop src/backend/postmaster/postmaster.c:1727:6
> #19 PostmasterMain src/backend/postmaster/postmaster.c:1414:11
> #20 main src/backend/main/main.c:232:4
>
> SUMMARY: UndefinedBehaviorSanitizer: undefined-behavior
> src/backend/commands/explain.c:3992:45
>
>
>
>
Hi, Fin!

Thanks for the report.

The problem is that auto_explain sets es->analyze based on non-zero
instrument_options.
EXPLAIN (BUFFERS) has this bit set, but the plan is not executed and
iss_Instrument remains NULL.

--
Regards,
Rachitskiy Andrey

Attachment Content-Type Size
0001-Fix-auto_explain-crash-on-EXPLAIN-BUFFERS.patch text/x-patch 1.6 KB

In response to

Browse pgsql-bugs by date

  From Date Subject
Next Message Kirill Reshke 2026-11-19 12:10:16 Re: BUG #19920: Out-of-Bounds Read on a Zero-Length Array in pg_trgm word_similarity('', '')
Previous Message Daniel Gustafsson 2026-11-19 11:53:22 Re: BUG #19920: Out-of-Bounds Read on a Zero-Length Array in pg_trgm word_similarity('', '')