Thread: BUG #15105: OpenTransientFile() should be paired withCloseTransientFile() rather than close()
BUG #15105: OpenTransientFile() should be paired withCloseTransientFile() rather than close()
From
PG Bug reporting form
Date:
The following bug has been logged on the website: Bug reference: 15105 Logged by: Pan Bian Email address: bianpan2016@163.com PostgreSQL version: 10.3 Operating system: Linux Description: File: src/backend/storage/ipc/dsm_impl.c Function: dsm_impl_mmap() Details: The handler opened with OpenTransientFile() should be closed with CloseTransientFile(). However, in function dsm_impl_mmap(), on a certain path, the return value of OpenTransientFile() (at line 885) is passed to close() (at line 926). It is better to use CloseTransientFile() here, as done at line 909. For your convenience, I paste the related source code as follows: 845 static bool 846 dsm_impl_mmap(dsm_op op, dsm_handle handle, Size request_size, 847 void **impl_private, void **mapped_address, Size *mapped_size, 848 int elevel) 849 { ... 883 /* Create new segment or open an existing one for attach or resize. */ 884 flags = O_RDWR | (op == DSM_OP_CREATE ? O_CREAT | O_EXCL : 0); 885 if ((fd = OpenTransientFile(name, flags, 0600)) == -1) 886 { 887 if (errno != EEXIST) 888 ereport(elevel, 889 (errcode_for_dynamic_shared_memory(), 890 errmsg("could not open shared memory segment \"%s\": %m", 891 name))); 892 return false; 893 } 894 895 /* 896 * If we're attaching the segment, determine the current size; if we are 897 * creating or resizing the segment, set the size to the requested value. 898 */ 899 if (op == DSM_OP_ATTACH) 900 { 901 struct stat st; 902 903 if (fstat(fd, &st) != 0) 904 { 905 int save_errno; 906 907 /* Back out what's already been done. */ 908 save_errno = errno; 909 CloseTransientFile(fd); 910 errno = save_errno; 911 912 ereport(elevel, 913 (errcode_for_dynamic_shared_memory(), 914 errmsg("could not stat shared memory segment \"%s\": %m", 915 name))); 916 return false; 917 } 918 request_size = st.st_size; 919 } 920 else if (*mapped_size > request_size && ftruncate(fd, request_size)) 921 { 922 int save_errno; 923 924 /* Back out what's already been done. */ 925 save_errno = errno; 926 close(fd); 927 if (op == DSM_OP_CREATE) 928 unlink(name); 929 errno = save_errno; 930 931 ereport(elevel, 932 (errcode_for_dynamic_shared_memory(), 933 errmsg("could not resize shared memory segment \"%s\" to %zu bytes: %m", 934 name, request_size))); 935 return false; 936 } ... 1038 *mapped_address = address; 1039 *mapped_size = request_size; 1040 CloseTransientFile(fd); 1041 1042 return true; 1043 } Thanks!
Re: BUG #15105: OpenTransientFile() should be paired withCloseTransientFile() rather than close()
From
Michael Paquier
Date:
On Fri, Mar 09, 2018 at 03:29:25AM +0000, PG Bug reporting form wrote: > Details: The handler opened with OpenTransientFile() should be closed with > CloseTransientFile(). However, in function dsm_impl_mmap(), on a certain > path, the return value of OpenTransientFile() (at line 885) is passed to > close() (at line 926). It is better to use CloseTransientFile() here, as > done at line 909. Good catch! This is visibly a copy-paste error coming from dsm_impl_posix(). As a patch it gives the attached. I am adding also Robert in CC for awareness. -- Michael
Attachment
Re: BUG #15105: OpenTransientFile() should be paired with CloseTransientFile() rather than close()
From
Tom Lane
Date:
Michael Paquier <michael@paquier.xyz> writes: > On Fri, Mar 09, 2018 at 03:29:25AM +0000, PG Bug reporting form wrote: >> Details: The handler opened with OpenTransientFile() should be closed with >> CloseTransientFile(). However, in function dsm_impl_mmap(), on a certain >> path, the return value of OpenTransientFile() (at line 885) is passed to >> close() (at line 926). It is better to use CloseTransientFile() here, as >> done at line 909. > Good catch! This is visibly a copy-paste error coming from > dsm_impl_posix(). As a patch it gives the attached. I am adding also > Robert in CC for awareness. Presumably, this would have been found sooner if AtEOXact_Files acted like most other end-of-xact cleanup functions and whined about leaked resources (in the commit case). I wonder why it doesn't. regards, tom lane
Re: BUG #15105: OpenTransientFile() should be paired with CloseTransientFile() rather than close()
From
Tom Lane
Date:
I wrote: > Michael Paquier <michael@paquier.xyz> writes: >> On Fri, Mar 09, 2018 at 03:29:25AM +0000, PG Bug reporting form wrote: >>> Details: The handler opened with OpenTransientFile() should be closed with >>> CloseTransientFile(). However, in function dsm_impl_mmap(), on a certain >>> path, the return value of OpenTransientFile() (at line 885) is passed to >>> close() (at line 926). It is better to use CloseTransientFile() here, as >>> done at line 909. >> Good catch! This is visibly a copy-paste error coming from >> dsm_impl_posix(). As a patch it gives the attached. I am adding also >> Robert in CC for awareness. Now that the commitfest crunch is over, I've checked and pushed this. > Presumably, this would have been found sooner if AtEOXact_Files acted > like most other end-of-xact cleanup functions and whined about leaked > resources (in the commit case). I wonder why it doesn't. On closer inspection, such a cross-check wouldn't have helped find this particular mistake, because it was in an error exit path that's likely never been exercised by anybody, and certainly isn't hit in the regression tests. Still, I think it'd be a good thing to add, and hence propose the attached patch. (I've verified that the warning doesn't fire in "make check-world".) regards, tom lane diff --git a/src/backend/access/transam/xact.c b/src/backend/access/transam/xact.c index 948733c..a2d5981 100644 --- a/src/backend/access/transam/xact.c +++ b/src/backend/access/transam/xact.c @@ -2123,7 +2123,7 @@ CommitTransaction(void) AtEOXact_on_commit_actions(true); AtEOXact_Namespace(true, is_parallel_worker); AtEOXact_SMgr(); - AtEOXact_Files(); + AtEOXact_Files(true); AtEOXact_ComboCid(); AtEOXact_HashTables(true); AtEOXact_PgStat(true); @@ -2401,7 +2401,7 @@ PrepareTransaction(void) AtEOXact_on_commit_actions(true); AtEOXact_Namespace(true, false); AtEOXact_SMgr(); - AtEOXact_Files(); + AtEOXact_Files(true); AtEOXact_ComboCid(); AtEOXact_HashTables(true); /* don't call AtEOXact_PgStat here; we fixed pgstat state above */ @@ -2603,7 +2603,7 @@ AbortTransaction(void) AtEOXact_on_commit_actions(false); AtEOXact_Namespace(false, is_parallel_worker); AtEOXact_SMgr(); - AtEOXact_Files(); + AtEOXact_Files(false); AtEOXact_ComboCid(); AtEOXact_HashTables(false); AtEOXact_PgStat(false); diff --git a/src/backend/postmaster/autovacuum.c b/src/backend/postmaster/autovacuum.c index 3b90b16..02e6d81 100644 --- a/src/backend/postmaster/autovacuum.c +++ b/src/backend/postmaster/autovacuum.c @@ -531,7 +531,7 @@ AutoVacLauncherMain(int argc, char *argv[]) } AtEOXact_Buffers(false); AtEOXact_SMgr(); - AtEOXact_Files(); + AtEOXact_Files(false); AtEOXact_HashTables(false); /* diff --git a/src/backend/postmaster/bgwriter.c b/src/backend/postmaster/bgwriter.c index b7813d7..d5ce685 100644 --- a/src/backend/postmaster/bgwriter.c +++ b/src/backend/postmaster/bgwriter.c @@ -198,7 +198,7 @@ BackgroundWriterMain(void) /* we needn't bother with the other ResourceOwnerRelease phases */ AtEOXact_Buffers(false); AtEOXact_SMgr(); - AtEOXact_Files(); + AtEOXact_Files(false); AtEOXact_HashTables(false); /* diff --git a/src/backend/postmaster/checkpointer.c b/src/backend/postmaster/checkpointer.c index 4b452e7..0950ada 100644 --- a/src/backend/postmaster/checkpointer.c +++ b/src/backend/postmaster/checkpointer.c @@ -282,7 +282,7 @@ CheckpointerMain(void) /* we needn't bother with the other ResourceOwnerRelease phases */ AtEOXact_Buffers(false); AtEOXact_SMgr(); - AtEOXact_Files(); + AtEOXact_Files(false); AtEOXact_HashTables(false); /* Warn any waiting backends that the checkpoint failed. */ diff --git a/src/backend/postmaster/walwriter.c b/src/backend/postmaster/walwriter.c index 4fa3a3b..688d5b5 100644 --- a/src/backend/postmaster/walwriter.c +++ b/src/backend/postmaster/walwriter.c @@ -179,7 +179,7 @@ WalWriterMain(void) /* we needn't bother with the other ResourceOwnerRelease phases */ AtEOXact_Buffers(false); AtEOXact_SMgr(); - AtEOXact_Files(); + AtEOXact_Files(false); AtEOXact_HashTables(false); /* diff --git a/src/backend/storage/file/fd.c b/src/backend/storage/file/fd.c index f772dfe..07e1e48 100644 --- a/src/backend/storage/file/fd.c +++ b/src/backend/storage/file/fd.c @@ -314,7 +314,7 @@ static bool reserveAllocatedDesc(void); static int FreeDesc(AllocateDesc *desc); static void AtProcExit_Files(int code, Datum arg); -static void CleanupTempFiles(bool isProcExit); +static void CleanupTempFiles(bool isCommit, bool isProcExit); static void RemovePgTempFilesInDir(const char *tmpdirname, bool missing_ok, bool unlink_all); static void RemovePgTempRelationFiles(const char *tsdirname); @@ -2902,17 +2902,19 @@ AtEOSubXact_Files(bool isCommit, SubTransactionId mySubid, /* * AtEOXact_Files * - * This routine is called during transaction commit or abort (it doesn't - * particularly care which). All still-open per-transaction temporary file - * VFDs are closed, which also causes the underlying files to be deleted - * (although they should've been closed already by the ResourceOwner - * cleanup). Furthermore, all "allocated" stdio files are closed. We also - * forget any transaction-local temp tablespace list. + * This routine is called during transaction commit or abort. All still-open + * per-transaction temporary file VFDs are closed, which also causes the + * underlying files to be deleted (although they should've been closed already + * by the ResourceOwner cleanup). Furthermore, all "allocated" stdio files are + * closed. We also forget any transaction-local temp tablespace list. + * + * The isCommit flag is used only to decide whether to emit warnings about + * unclosed files. */ void -AtEOXact_Files(void) +AtEOXact_Files(bool isCommit) { - CleanupTempFiles(false); + CleanupTempFiles(isCommit, false); tempTableSpaces = NULL; numTempTableSpaces = -1; } @@ -2926,12 +2928,15 @@ AtEOXact_Files(void) static void AtProcExit_Files(int code, Datum arg) { - CleanupTempFiles(true); + CleanupTempFiles(false, true); } /* * Close temporary files and delete their underlying files. * + * isCommit: if true, this is normal transaction commit, and we don't + * expect any remaining files; warn if there are some. + * * isProcExit: if true, this is being called as the backend process is * exiting. If that's the case, we should remove all temporary files; if * that's not the case, we are being called for transaction commit/abort @@ -2939,7 +2944,7 @@ AtProcExit_Files(int code, Datum arg) * also clean up "allocated" stdio files, dirs and fds. */ static void -CleanupTempFiles(bool isProcExit) +CleanupTempFiles(bool isCommit, bool isProcExit) { Index i; @@ -2979,6 +2984,11 @@ CleanupTempFiles(bool isProcExit) have_xact_temporary_files = false; } + /* Complain if any allocated files remain open at commit. */ + if (isCommit && numAllocatedDescs > 0) + elog(WARNING, "%d temporary files and directories not closed at end-of-transaction", + numAllocatedDescs); + /* Clean up "allocated" stdio files, dirs and fds. */ while (numAllocatedDescs > 0) FreeDesc(&allocatedDescs[0]); diff --git a/src/include/storage/fd.h b/src/include/storage/fd.h index 484339b..548a832 100644 --- a/src/include/storage/fd.h +++ b/src/include/storage/fd.h @@ -123,7 +123,7 @@ extern void SetTempTablespaces(Oid *tableSpaces, int numSpaces); extern bool TempTablespacesAreSet(void); extern int GetTempTablespaces(Oid *tableSpaces, int numSpaces); extern Oid GetNextTempTableSpace(void); -extern void AtEOXact_Files(void); +extern void AtEOXact_Files(bool isCommit); extern void AtEOSubXact_Files(bool isCommit, SubTransactionId mySubid, SubTransactionId parentSubid); extern void RemovePgTempFiles(void);
Re: BUG #15105: OpenTransientFile() should be paired withCloseTransientFile() rather than close()
From
Michael Paquier
Date:
On Tue, Apr 10, 2018 at 06:40:43PM -0400, Tom Lane wrote: > On closer inspection, such a cross-check wouldn't have helped find this > particular mistake, because it was in an error exit path that's likely > never been exercised by anybody, and certainly isn't hit in the regression > tests. Still, I think it'd be a good thing to add, and hence propose the > attached patch. (I've verified that the warning doesn't fire in "make > check-world".) Good idea. +1 for your proposed patch. -- Michael