Movatterモバイル変換

[0]ホーム
URL:

Skip to content

Navigation Menu

Sign in
Appearance settings

Search code, repositories, users, issues, pull requests...

Provide feedback

We read every piece of feedback, and take your input very seriously.

Saved searches

Use saved searches to filter your results more quickly

 Sign up
Appearance settings

Commit93078e6

Browse files
committed
Maintain valid md.c state when FileClose() fails.
FileClose() failure ordinarily causes a PANIC. Suppose the userdisables that PANIC via data_sync_retry=on. After mdclose() issued aFileClose() that failed, calls into md.c raised SIGSEGV. This fix addsrepalloc() calls during mdclose(); update a comment about ignoringrepalloc() cost. The rate of relation segment count change is a minorfactor; more relevant to overall performance is the rate of mdclose()and subsequent re-opening of segments. Back-patch to v10, where commit45e191e introduced the bug.Reviewed by Kyotaro Horiguchi.Discussion:https://postgr.es/m/20191222091930.GA1280238@rfd.leadboat.com
1 parent2e89a12 commit93078e6

File tree

1 file changed

+6
-14
lines changed
  • src/backend/storage/smgr

1 file changed

+6
-14
lines changed

‎src/backend/storage/smgr/md.c

Lines changed: 6 additions & 14 deletions
Original file line numberDiff line numberDiff line change
@@ -491,18 +491,10 @@ mdclose(SMgrRelation reln, ForkNumber forknum)
491491
{
492492
MdfdVec*v=&reln->md_seg_fds[forknum][nopensegs-1];
493493

494-
/* if not closed already */
495-
if (v->mdfd_vfd >=0)
496-
{
497-
FileClose(v->mdfd_vfd);
498-
v->mdfd_vfd=-1;
499-
}
500-
494+
FileClose(v->mdfd_vfd);
495+
_fdvec_resize(reln,forknum,nopensegs-1);
501496
nopensegs--;
502497
}
503-
504-
/* resize just once, avoids pointless reallocations */
505-
_fdvec_resize(reln,forknum,0);
506498
}
507499

508500
/*
@@ -1025,10 +1017,10 @@ _fdvec_resize(SMgrRelation reln,
10251017
else
10261018
{
10271019
/*
1028-
* It doesn't seem worthwhile complicating the codeby having a more
1029-
*aggressive growth strategy here; the number of segments doesn't
1030-
*grow that fast, and the memory context internally will sometimes
1031-
*avoiddoing an actual reallocation.
1020+
* It doesn't seem worthwhile complicating the codeto amortize
1021+
*repalloc() calls. Those are far faster than PathNameOpenFile() or
1022+
*FileClose(), and the memory context internally will sometimes avoid
1023+
* doing an actual reallocation.
10321024
*/
10331025
reln->md_seg_fds[forknum]=
10341026
repalloc(reln->md_seg_fds[forknum],

0 commit comments

Comments
 (0)
[8]ページ先頭
©2009-2025 Movatter.jp