From d0024cd1881447fa7aed58db94df379e593c6630 Mon Sep 17 00:00:00 2001 From: Tom Lane Date: Tue, 20 Dec 2011 15:00:36 -0500 Subject: [PATCH] Avoid crashing when we have problems unlinking files post-commit. smgrdounlink takes care to not throw an ERROR if it fails to unlink something, but that caution was rendered useless by commit 3396000684b41e7e9467d1abc67152b39e697035, which put an smgrexists call in front of it; smgrexists *does* throw error if anything looks funny, such as getting a permissions error from trying to open the file. If that happens post-commit, you get a PANIC, and what's worse the same logic appears in the WAL replay code, so the database even fails to restart. Restore the intended behavior by removing the smgrexists call --- it isn't accomplishing anything that we can't do better by adjusting mdunlink's ideas of whether it ought to warn about ENOENT or not. Per report from Joseph Shraibman of unrecoverable crash after trying to drop a table whose FSM fork had somehow gotten chmod'd to 000 permissions. Backpatch to 8.4, where the bogus coding was introduced. --- src/backend/access/transam/twophase.c | 3 +-- src/backend/access/transam/xact.c | 14 ++++---------- src/backend/catalog/storage.c | 3 +-- src/backend/storage/smgr/md.c | 26 ++++++++++++++------------ 4 files changed, 20 insertions(+), 26 deletions(-) diff --git a/src/backend/access/transam/twophase.c b/src/backend/access/transam/twophase.c index d2fecb1ecb..c5a2541807 100644 --- a/src/backend/access/transam/twophase.c +++ b/src/backend/access/transam/twophase.c @@ -1366,8 +1366,7 @@ FinishPreparedTransaction(const char *gid, bool isCommit) for (fork = 0; fork <= MAX_FORKNUM; fork++) { - if (smgrexists(srel, fork)) - smgrdounlink(srel, fork, false); + smgrdounlink(srel, fork, false); } smgrclose(srel); } diff --git a/src/backend/access/transam/xact.c b/src/backend/access/transam/xact.c index c383011b5f..4177a6c5cf 100644 --- a/src/backend/access/transam/xact.c +++ b/src/backend/access/transam/xact.c @@ -4595,11 +4595,8 @@ xact_redo_commit_internal(TransactionId xid, XLogRecPtr lsn, for (fork = 0; fork <= MAX_FORKNUM; fork++) { - if (smgrexists(srel, fork)) - { - XLogDropRelation(xnodes[i], fork); - smgrdounlink(srel, fork, true); - } + XLogDropRelation(xnodes[i], fork); + smgrdounlink(srel, fork, true); } smgrclose(srel); } @@ -4738,11 +4735,8 @@ xact_redo_abort(xl_xact_abort *xlrec, TransactionId xid) for (fork = 0; fork <= MAX_FORKNUM; fork++) { - if (smgrexists(srel, fork)) - { - XLogDropRelation(xlrec->xnodes[i], fork); - smgrdounlink(srel, fork, true); - } + XLogDropRelation(xlrec->xnodes[i], fork); + smgrdounlink(srel, fork, true); } smgrclose(srel); } diff --git a/src/backend/catalog/storage.c b/src/backend/catalog/storage.c index fff933d403..450e292f40 100644 --- a/src/backend/catalog/storage.c +++ b/src/backend/catalog/storage.c @@ -361,8 +361,7 @@ smgrDoPendingDeletes(bool isCommit) srel = smgropen(pending->relnode, pending->backend); for (i = 0; i <= MAX_FORKNUM; i++) { - if (smgrexists(srel, i)) - smgrdounlink(srel, i, false); + smgrdounlink(srel, i, false); } smgrclose(srel); } diff --git a/src/backend/storage/smgr/md.c b/src/backend/storage/smgr/md.c index a761369d39..7150a00407 100644 --- a/src/backend/storage/smgr/md.c +++ b/src/backend/storage/smgr/md.c @@ -323,7 +323,13 @@ mdcreate(SMgrRelation reln, ForkNumber forkNum, bool isRedo) * number until it's safe, because relfilenode assignment skips over any * existing file. * - * If isRedo is true, it's okay for the relation to be already gone. + * All the above applies only to the relation's main fork; other forks can + * just be removed immediately, since they are not needed to prevent the + * relfilenode number from being recycled. Also, we do not carefully + * track whether other forks have been created or not, but just attempt to + * unlink them unconditionally; so we should never complain about ENOENT. + * + * If isRedo is true, it's unsurprising for the relation to be already gone. * Also, we should remove the file immediately instead of queuing a request * for later, since during redo there's no possibility of creating a * conflicting relation. @@ -351,13 +357,10 @@ mdunlink(RelFileNodeBackend rnode, ForkNumber forkNum, bool isRedo) if (isRedo || forkNum != MAIN_FORKNUM) { ret = unlink(path); - if (ret < 0) - { - if (!isRedo || errno != ENOENT) - ereport(WARNING, - (errcode_for_file_access(), - errmsg("could not remove file \"%s\": %m", path))); - } + if (ret < 0 && errno != ENOENT) + ereport(WARNING, + (errcode_for_file_access(), + errmsg("could not remove file \"%s\": %m", path))); } else { @@ -380,6 +383,9 @@ mdunlink(RelFileNodeBackend rnode, ForkNumber forkNum, bool isRedo) ereport(WARNING, (errcode_for_file_access(), errmsg("could not truncate file \"%s\": %m", path))); + + /* Register request to unlink first segment later */ + register_unlink(rnode); } /* @@ -411,10 +417,6 @@ mdunlink(RelFileNodeBackend rnode, ForkNumber forkNum, bool isRedo) } pfree(path); - - /* Register request to unlink first segment later */ - if (!isRedo && forkNum == MAIN_FORKNUM) - register_unlink(rnode); } /*