database/postgres
1
0
Fork 0
mirror of https://github.com/postgres/postgres.git synced 2025-10-15 05:46:52 +03:00
Code Activity

Fix EPQ crash from missing partition pruning state in EState

Browse Source 
Commit bb3ec16e14 moved partition pruning metadata into PlannedStmt.
At executor startup this metadata is used to initialize the EState
fields es_part_prune_infos, es_part_prune_states, and
es_part_prune_results.  EvalPlanQualStart() failed to copy those
fields into the child EState, causing NULL dereference when Append
ran partition pruning during a recheck. This can occur with DELETE
or UPDATE on partitioned tables that use runtime pruning, e.g. with
generic plans.

Fix by copying all partition pruning state into the EPQ estate.

Add an isolation test that reproduces the crash with concurrent
UPDATE and DELETE on a partitioned table, where the DELETE session
hits the crash during its EPQ recheck after the UPDATE commits.

Bug: #19056
Reported-by: Fei Changhong <feichanghong@qq.com>
Diagnozed-by: Fei Changhong <feichanghong@qq.com>
Author: David Rowley <dgrowleyml@gmail.com>
Co-authored-by: Amit Langote <amitlangote09@gmail.com>
Discussion: https://postgr.es/m/19056-a677cef9b54d76a0%40postgresql.org
This commit is contained in:
Amit Langote
2025-09-19 11:38:29 +09:00
parent 3cd3a039da
commit 8741e48e5d
3 changed files with 26 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files

9
src/backend/executor/execMain.c
View File

@@ -3084,6 +3084,15 @@ EvalPlanQualStart(EPQState *epqstate, Plan *planTree)
*/
rcestate->es_unpruned_relids = parentestate->es_unpruned_relids;
/*
* Also make the PartitionPruneInfo and the results of pruning available.
* These need to match exactly so that we initialize all the same Append
* and MergeAppend subplans as the parent did.
*/
rcestate->es_part_prune_infos = parentestate->es_part_prune_infos;
rcestate->es_part_prune_states = parentestate->es_part_prune_states;
rcestate->es_part_prune_results = parentestate->es_part_prune_results;
/*
* Initialize private state information for each SubPlan. We must do this
* before running ExecInitNode on the main query tree, since

9
src/test/isolation/expected/eval-plan-qual.out
View File

@@ -1457,3 +1457,12 @@ step sysmerge2:
step c1: COMMIT;
step sysmerge2: <... completed>
step c2: COMMIT;
starting permutation: s1pp1 s2pp1 s2pp2 s2pp3 c1 c2
step s1pp1: UPDATE another_parttbl SET b = b + 1 WHERE a = 1;
step s2pp1: SET plan_cache_mode TO force_generic_plan;
step s2pp2: PREPARE epd AS DELETE FROM another_parttbl WHERE a = $1;
step s2pp3: EXECUTE epd(1); <waiting ...>
step c1: COMMIT;
step s2pp3: <... completed>
step c2: COMMIT;

8
src/test/isolation/specs/eval-plan-qual.spec
View File

@@ -204,6 +204,7 @@ step sys1 {
UPDATE pg_class SET reltuples = 123 WHERE oid = 'accounts'::regclass;
}
step s1pp1 { UPDATE another_parttbl SET b = b + 1 WHERE a = 1; }
session s2
setup { BEGIN ISOLATION LEVEL READ COMMITTED; }
@@ -312,6 +313,10 @@ step sysmerge2 {
step c2 { COMMIT; }
step r2 { ROLLBACK; }
step s2pp1 { SET plan_cache_mode TO force_generic_plan; }
step s2pp2 { PREPARE epd AS DELETE FROM another_parttbl WHERE a = $1; }
step s2pp3 { EXECUTE epd(1); }
session s3
setup { BEGIN ISOLATION LEVEL READ COMMITTED; }
step read { SELECT * FROM accounts ORDER BY accountid; }
@@ -415,3 +420,6 @@ permutation simplepartupdate_noroute complexpartupdate_doesnt_route c1 c2 read_p
permutation sys1 sysupd2 c1 c2
permutation sys1 sysmerge2 c1 c2
# Exercise run-time partition pruning code in an EPQ recheck
permutation s1pp1 s2pp1 s2pp2 s2pp3 c1 c2