Skip to content

Commit 9a82a64

Browse files
committed
Fix EPQ crash from missing partition pruning state in EState
Commit bb3ec16 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
1 parent 144de03 commit 9a82a64

File tree

3 files changed

+26
-0
lines changed

3 files changed

+26
-0
lines changed

src/backend/executor/execMain.c

Lines changed: 9 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -3084,6 +3084,15 @@ EvalPlanQualStart(EPQState *epqstate, Plan *planTree)
30843084
*/
30853085
rcestate->es_unpruned_relids = parentestate->es_unpruned_relids;
30863086

3087+
/*
3088+
* Also make the PartitionPruneInfo and the results of pruning available.
3089+
* These need to match exactly so that we initialize all the same Append
3090+
* and MergeAppend subplans as the parent did.
3091+
*/
3092+
rcestate->es_part_prune_infos = parentestate->es_part_prune_infos;
3093+
rcestate->es_part_prune_states = parentestate->es_part_prune_states;
3094+
rcestate->es_part_prune_results = parentestate->es_part_prune_results;
3095+
30873096
/*
30883097
* Initialize private state information for each SubPlan. We must do this
30893098
* before running ExecInitNode on the main query tree, since

src/test/isolation/expected/eval-plan-qual.out

Lines changed: 9 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -1457,3 +1457,12 @@ step sysmerge2:
14571457
step c1: COMMIT;
14581458
step sysmerge2: <... completed>
14591459
step c2: COMMIT;
1460+
1461+
starting permutation: s1pp1 s2pp1 s2pp2 s2pp3 c1 c2
1462+
step s1pp1: UPDATE another_parttbl SET b = b + 1 WHERE a = 1;
1463+
step s2pp1: SET plan_cache_mode TO force_generic_plan;
1464+
step s2pp2: PREPARE epd AS DELETE FROM another_parttbl WHERE a = $1;
1465+
step s2pp3: EXECUTE epd(1); <waiting ...>
1466+
step c1: COMMIT;
1467+
step s2pp3: <... completed>
1468+
step c2: COMMIT;

src/test/isolation/specs/eval-plan-qual.spec

Lines changed: 8 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -204,6 +204,7 @@ step sys1 {
204204
UPDATE pg_class SET reltuples = 123 WHERE oid = 'accounts'::regclass;
205205
}
206206

207+
step s1pp1 { UPDATE another_parttbl SET b = b + 1 WHERE a = 1; }
207208

208209
session s2
209210
setup { BEGIN ISOLATION LEVEL READ COMMITTED; }
@@ -312,6 +313,10 @@ step sysmerge2 {
312313
step c2 { COMMIT; }
313314
step r2 { ROLLBACK; }
314315

316+
step s2pp1 { SET plan_cache_mode TO force_generic_plan; }
317+
step s2pp2 { PREPARE epd AS DELETE FROM another_parttbl WHERE a = $1; }
318+
step s2pp3 { EXECUTE epd(1); }
319+
315320
session s3
316321
setup { BEGIN ISOLATION LEVEL READ COMMITTED; }
317322
step read { SELECT * FROM accounts ORDER BY accountid; }
@@ -415,3 +420,6 @@ permutation simplepartupdate_noroute complexpartupdate_doesnt_route c1 c2 read_p
415420

416421
permutation sys1 sysupd2 c1 c2
417422
permutation sys1 sysmerge2 c1 c2
423+
424+
# Exercise run-time partition pruning code in an EPQ recheck
425+
permutation s1pp1 s2pp1 s2pp2 s2pp3 c1 c2

0 commit comments

Comments
 (0)