In heapam_relation_copy_for_cluster(), begin_heap_rewrite() sets
rwstate->rs_new_rel->rd_smgr correctly but next line tuplesort_begin_cluster()
get called which cause the system cache invalidation and due to CCA setting,
wipe out rwstate->rs_new_rel->rd_smgr which wasn't restored for the subsequent
operations and causes segmentation fault.
By calling RelationOpenSmgr() before calling smgrimmedsync() in
end_heap_rewrite() would fix the failure. Did the same in the attached patch.
Regards,
Amul
On Mon, Mar 22, 2021 at 11:53 AM Neha Sharma
<neha.sha...@enterprisedb.com> wrote:
>
> Hello,
>
> While executing the below test case server crashed with Segfault 11 on master
> branch.
> I have enabled the CLOBBER_CACHE_ALWAYS in src/include/pg_config_manual.h
>
> Issue is only reproducing on master branch.
>
> Test Case:
> CREATE TABLE sm_5_323_table (col1 numeric);
> CREATE INDEX sm_5_323_idx ON sm_5_323_table(col1);
>
> CLUSTER sm_5_323_table USING sm_5_323_idx;
>
> \! /PGClobber_build/postgresql/inst/bin/clusterdb -t sm_5_323_table -U edb -h
> localhost -p 5432 -d postgres
>
> Test case output:
> edb@edb:~/PGClobber_build/postgresql/inst/bin$ ./psql postgres
> psql (14devel)
> Type "help" for help.
>
> postgres=# CREATE TABLE sm_5_323_table (col1 numeric);
> CREATE TABLE
> postgres=# CREATE INDEX sm_5_323_idx ON sm_5_323_table(col1);
> CREATE INDEX
> postgres=# CLUSTER sm_5_323_table USING sm_5_323_idx;
> CLUSTER
> postgres=# \! /PGClobber_build/postgresql/inst/bin/clusterdb -t
> sm_5_323_table -U edb -h localhost -p 5432 -d postgres
> clusterdb: error: clustering of table "sm_5_323_table" in database "postgres"
> failed: server closed the connection unexpectedly
> This probably means the server terminated abnormally
> before or while processing the request.
>
> Stack Trace:
> Core was generated by `postgres: edb postgres 127.0.0.1(50978) CLUSTER
> '.
> Program terminated with signal SIGSEGV, Segmentation fault.
> #0 0x000055e5c85ea0b4 in mdopenfork (reln=0x0, forknum=MAIN_FORKNUM,
> behavior=1) at md.c:485
> 485 if (reln->md_num_open_segs[forknum] > 0)
> (gdb) bt
> #0 0x000055e5c85ea0b4 in mdopenfork (reln=0x0, forknum=MAIN_FORKNUM,
> behavior=1) at md.c:485
> #1 0x000055e5c85eb2f0 in mdnblocks (reln=0x0, forknum=MAIN_FORKNUM) at
> md.c:768
> #2 0x000055e5c85eb61b in mdimmedsync (reln=0x0,
> forknum=forknum@entry=MAIN_FORKNUM) at md.c:930
> #3 0x000055e5c85ec6e5 in smgrimmedsync (reln=<optimized out>,
> forknum=forknum@entry=MAIN_FORKNUM) at smgr.c:662
> #4 0x000055e5c81ae28b in end_heap_rewrite (state=state@entry=0x55e5ca5d1d70)
> at rewriteheap.c:342
> #5 0x000055e5c81a32ea in heapam_relation_copy_for_cluster
> (OldHeap=0x7f212ce41ba0, NewHeap=0x7f212ce41058, OldIndex=<optimized out>,
> use_sort=<optimized out>, OldestXmin=<optimized out>,
> xid_cutoff=<optimized out>, multi_cutoff=0x7ffcba6ebe64,
> num_tuples=0x7ffcba6ebe68, tups_vacuumed=0x7ffcba6ebe70,
> tups_recently_dead=0x7ffcba6ebe78) at heapam_handler.c:984
> #6 0x000055e5c82f218a in table_relation_copy_for_cluster
> (tups_recently_dead=0x7ffcba6ebe78, tups_vacuumed=0x7ffcba6ebe70,
> num_tuples=0x7ffcba6ebe68, multi_cutoff=0x7ffcba6ebe64,
> xid_cutoff=0x7ffcba6ebe60, OldestXmin=<optimized out>,
> use_sort=<optimized out>, OldIndex=0x7f212ce40670, NewTable=0x7f212ce41058,
> OldTable=0x7f212ce41ba0)
> at ../../../src/include/access/tableam.h:1656
> #7 copy_table_data (pCutoffMulti=<synthetic pointer>, pFreezeXid=<synthetic
> pointer>, pSwapToastByContent=<synthetic pointer>, verbose=<optimized out>,
> OIDOldIndex=<optimized out>,
> OIDOldHeap=16384, OIDNewHeap=<optimized out>) at cluster.c:908
> #8 rebuild_relation (verbose=<optimized out>, indexOid=<optimized out>,
> OldHeap=<optimized out>) at cluster.c:604
> #9 cluster_rel (tableOid=<optimized out>, indexOid=<optimized out>,
> params=<optimized out>) at cluster.c:427
> #10 0x000055e5c82f2b7f in cluster (pstate=pstate@entry=0x55e5ca5315c0,
> stmt=stmt@entry=0x55e5ca510368, isTopLevel=isTopLevel@entry=true) at
> cluster.c:195
> #11 0x000055e5c85fcbc6 in standard_ProcessUtility (pstmt=0x55e5ca510430,
> queryString=0x55e5ca50f850 "CLUSTER public.sm_5_323_table;",
> context=PROCESS_UTILITY_TOPLEVEL, params=0x0,
> queryEnv=0x0, dest=0x55e5ca510710, qc=0x7ffcba6ec340) at utility.c:822
> #12 0x000055e5c85fd436 in ProcessUtility (pstmt=pstmt@entry=0x55e5ca510430,
> queryString=<optimized out>, context=context@entry=PROCESS_UTILITY_TOPLEVEL,
> params=<optimized out>,
> queryEnv=<optimized out>, dest=dest@entry=0x55e5ca510710,
> qc=0x7ffcba6ec340) at utility.c:525
> #13 0x000055e5c85f6148 in PortalRunUtility
> (portal=portal@entry=0x55e5ca570d70, pstmt=pstmt@entry=0x55e5ca510430,
> isTopLevel=isTopLevel@entry=true,
> setHoldSnapshot=setHoldSnapshot@entry=false,
> dest=dest@entry=0x55e5ca510710, qc=qc@entry=0x7ffcba6ec340) at pquery.c:1159
> #14 0x000055e5c85f71a4 in PortalRunMulti (portal=portal@entry=0x55e5ca570d70,
> isTopLevel=isTopLevel@entry=true, setHoldSnapshot=setHoldSnapshot@entry=false,
> dest=dest@entry=0x55e5ca510710, altdest=altdest@entry=0x55e5ca510710,
> qc=qc@entry=0x7ffcba6ec340) at pquery.c:1305
> #15 0x000055e5c85f8823 in PortalRun (portal=portal@entry=0x55e5ca570d70,
> count=count@entry=9223372036854775807, isTopLevel=isTopLevel@entry=true,
> run_once=run_once@entry=true,
> dest=dest@entry=0x55e5ca510710, altdest=altdest@entry=0x55e5ca510710,
> qc=0x7ffcba6ec340) at pquery.c:779
> #16 0x000055e5c85f389e in exec_simple_query (query_string=0x55e5ca50f850
> "CLUSTER public.sm_5_323_table;") at postgres.c:1185
> #17 0x000055e5c85f51cf in PostgresMain (argc=argc@entry=1,
> argv=argv@entry=0x7ffcba6ec670, dbname=<optimized out>, username=<optimized
> out>) at postgres.c:4415
> #18 0x000055e5c8522240 in BackendRun (port=<optimized out>, port=<optimized
> out>) at postmaster.c:4470
> #19 BackendStartup (port=<optimized out>) at postmaster.c:4192
> #20 ServerLoop () at postmaster.c:1737
> #21 0x000055e5c85237ec in PostmasterMain (argc=<optimized out>,
> argv=0x55e5ca508fe0) at postmaster.c:1409
> #22 0x000055e5c811a2cf in main (argc=3, argv=0x55e5ca508fe0) at main.c:209
>
> Thanks.
> --
> Regards,
> Neha Sharma
diff --git a/src/backend/access/heap/rewriteheap.c b/src/backend/access/heap/rewriteheap.c
index 8241ba8f312..b44343a0166 100644
--- a/src/backend/access/heap/rewriteheap.c
+++ b/src/backend/access/heap/rewriteheap.c
@@ -339,7 +339,10 @@ end_heap_rewrite(RewriteState state)
* wrote before the checkpoint.
*/
if (RelationNeedsWAL(state->rs_new_rel))
+ {
+ RelationOpenSmgr(state->rs_new_rel);
smgrimmedsync(state->rs_new_rel->rd_smgr, MAIN_FORKNUM);
+ }
logical_end_heap_rewrite(state);
