http://bugs.winehq.org/show_bug.cgi?id=24164
Summary: Wine is not data-race detector friendly Product: Wine Version: 1.2 Platform: x86 OS/Version: Linux Status: UNCONFIRMED Severity: normal Priority: P2 Component: -unknown AssignedTo: wine-bugs@winehq.org ReportedBy: bart.vanassche@gmail.com
Running Wine under a data-race detector triggers huge numbers of data-race reports on Wine functions. This makes it impossible to analyze data races in application that are running under Wine with Valgrind. An example (Valgrind trunk r11292):
$ valgrind --trace-children=yes --vex-iropt-precise-memory-exns=yes --gen-suppressions=all --tool=drd --show-confl-seg=no wine notepad [ ... ] ==1331== Conflicting store by thread 1 at 0x7f00006c size 4 ==1331== at 0x476BD36: RtlEnterCriticalSection (in /usr/lib/wine/ntdll.dll.so ) ==1331== by 0x47811DC: RtlAllocateHeap (in /usr/lib/wine/ntdll.dll.so) ==1331== by 0x4B44408: scmdatabase_autostart_services (in /usr/lib/wine/servi ces.exe.so) ==1331== by 0x4B45C9E: main (in /usr/lib/wine/services.exe.so) ==1331== Allocation context: unknown. [ ... ] ==1331== Conflicting load by thread 1 at 0x7f000128 size 4 ==1331== at 0x477C80F: list_next (in /usr/lib/wine/ntdll.dll.so) ==1331== by 0x477EF7E: HEAP_FindFreeBlock (in /usr/lib/wine/ntdll.dll.so) ==1331== by 0x47812D6: RtlAllocateHeap (in /usr/lib/wine/ntdll.dll.so) ==1331== by 0x4B44408: scmdatabase_autostart_services (in /usr/lib/wine/services.exe.so) ==1331== by 0x4B45C9E: main (in /usr/lib/wine/services.exe.so) ==1331== Allocation context: unknown. [ ... ] ==1331== Conflicting load by thread 1 at 0x7f0047e0 size 4 ==1331== at 0x477C7EC: list_remove (in /usr/lib/wine/ntdll.dll.so) ==1331== by 0x478138D: RtlAllocateHeap (in /usr/lib/wine/ntdll.dll.so) ==1331== by 0x4B44408: scmdatabase_autostart_services (in /usr/lib/wine/services.exe.so) ==1331== by 0x4B45C9E: main (in /usr/lib/wine/services.exe.so) ==1331== Allocation context: unknown. [ ... ]
http://bugs.winehq.org/show_bug.cgi?id=24164
Austin English austinenglish@gmail.com changed:
What |Removed |Added ---------------------------------------------------------------------------- Keywords| |download, source
http://bugs.winehq.org/show_bug.cgi?id=24164
--- Comment #1 from Bart Van Assche bart.vanassche@gmail.com 2010-08-26 11:06:41 --- Note: I have not yet analyzed whether the reported conflicting memory accesses are harmless or not.
http://bugs.winehq.org/show_bug.cgi?id=24164
Dan Kegel dank@kegel.com changed:
What |Removed |Added ---------------------------------------------------------------------------- Status|UNCONFIRMED |NEW CC| |dank@kegel.com Ever Confirmed|0 |1
--- Comment #2 from Dan Kegel dank@kegel.com 2010-08-26 17:13:25 --- Not surprising. We'll probably need a mondo suppression file like the one we use for memcheck.
http://bugs.winehq.org/show_bug.cgi?id=24164
--- Comment #3 from Bart Van Assche bart.vanassche@gmail.com 2010-08-27 05:15:35 --- (In reply to comment #2)
Not surprising. We'll probably need a mondo suppression file like the one we use for memcheck.
Suppressing everything that is reported by DRD on Wine itself might work but doesn't seem a good idea to me. Not only might this hide real Wine threading issues, but this will also cause slow operation of Valgrind because of the high frequency at which Valgrind has to perform stack matching.
http://bugs.winehq.org/show_bug.cgi?id=24164
--- Comment #4 from Dan Kegel dank@kegel.com 2010-08-27 08:57:32 --- It's not especially good that we need a suppression file for memcheck for wine either, but you have to start somewhere.
Somebody needs to start sweeping with the data-race detector and clean up all the errors just like we did with memcheck. (We're still not done with memcheck - we never cleaned up wineboot.) It's a big job, but maybe a few days would help a lot.
http://bugs.winehq.org/show_bug.cgi?id=24164
--- Comment #5 from Bart Van Assche bart.vanassche@gmail.com 2010-08-29 12:37:19 CDT --- Created an attachment (id=30470) --> (http://bugs.winehq.org/attachment.cgi?id=30470) Patch for suppressing the most common false positive data race reports
Is the style of the attached patch acceptable ? With that patch applied on top of the Wine git repository no false positive data race reports are produced by Helgrind or DRD when starting applications like Notepad or Wordpad. Notes: - The attached patch has only been tested lightly so far. - The patch only does something meaningful if Wine is built with the Valgrind 3.6.0 headers. Since Valgrind version 3.6.0 has not yet been released, this means the trunk. - The client requests inserted by the patch are only supported by the Helgrind and DRD tools that will be included in Valgrind 3.6.0.
http://bugs.winehq.org/show_bug.cgi?id=24164
Bart Van Assche bart.vanassche@gmail.com changed:
What |Removed |Added ---------------------------------------------------------------------------- Attachment #30470|0 |1 is obsolete| |
--- Comment #6 from Bart Van Assche bart.vanassche@gmail.com 2010-09-01 05:23:17 CDT --- Created an attachment (id=30505) --> (http://bugs.winehq.org/attachment.cgi?id=30505) Patch for suppressing the most common false positive data race reports, version 2
Fixes one real race in heap.c by moving the heap_set_debug_flags() call up. The access pattern of npc->pipe (rpcrt4) should be examined further - I'm not sure it is safe.
http://bugs.winehq.org/show_bug.cgi?id=24164
Wylda wylda@volny.cz changed:
What |Removed |Added ---------------------------------------------------------------------------- Keywords| |patch CC| |wylda@volny.cz
--- Comment #7 from Wylda wylda@volny.cz 2010-09-01 05:28:54 CDT ---
It's better to send any patches to wine-devel for review. Adding keyword anyway.
http://bugs.winehq.org/show_bug.cgi?id=24164
--- Comment #8 from Alexandre Julliard julliard@winehq.org 2010-09-01 06:11:23 CDT --- (In reply to comment #7)
It's better to send any patches to wine-devel for review. Adding keyword anyway.
You certainly don't want to add annotations all over the place. If there are really that many false positives it sounds like the tool needs improving.
http://bugs.winehq.org/show_bug.cgi?id=24164
--- Comment #9 from Bart Van Assche bart.vanassche@gmail.com 2010-09-01 10:09:03 CDT --- (In reply to comment #8)
You certainly don't want to add annotations all over the place. If there are really that many false positives it sounds like the tool needs improving.
I'm afraid that the above statement means you are not familiar with data race detection tools. The annotations I added are necessary for every existing data race detection tool - whether it is DRD, Helgrind or VTune.
As an example, the global variable 'server_start_time' is updated by multiple threads without protection by a critical section or any other mutual exclusion primitive. The only way to avoid that data race detection tools complain on such an access pattern is to add a suppression for that access pattern.
http://bugs.winehq.org/show_bug.cgi?id=24164
Bart Van Assche bart.vanassche@gmail.com changed:
What |Removed |Added ---------------------------------------------------------------------------- Attachment #30505|0 |1 is obsolete| |
--- Comment #10 from Bart Van Assche bart.vanassche@gmail.com 2010-09-02 10:25:08 CDT --- Created an attachment (id=30522) --> (http://bugs.winehq.org/attachment.cgi?id=30522) Patch for suppressing the most common false positive data race reports
Version three of this patch.
http://bugs.winehq.org/show_bug.cgi?id=24164
Austin English austinenglish@gmail.com changed:
What |Removed |Added ---------------------------------------------------------------------------- CC| |austinenglish@gmail.com
https://bugs.winehq.org/show_bug.cgi?id=24164
--- Comment #11 from Ken Sharp imwellcushtymelike@gmail.com --- Is this still an issue in Wine 1.7.45 or later?
https://bugs.winehq.org/show_bug.cgi?id=24164
joaopa jeremielapuree@yahoo.fr changed:
What |Removed |Added ---------------------------------------------------------------------------- CC| |jeremielapuree@yahoo.fr
--- Comment #12 from joaopa jeremielapuree@yahoo.fr --- Does the bug still occur with wine-6.19?
https://bugs.winehq.org/show_bug.cgi?id=24164
--- Comment #13 from joaopa jeremielapuree@yahoo.fr --- Looks like the bug still occurs with wine-6.19
==11923== Conflicting load by thread 4 at 0x04e92228 size 4 ==11923== at 0x471D6F5: NtTerminateThread (in /opt/wine-devel/lib/wine/i386-unix/ntdll.so) ==11923== by 0x47045C8: ??? (in /opt/wine-devel/lib/wine/i386-unix/ntdll.so) ==11923== by 0xFFFFFFFD: ??? ==11923== by 0x474733F: ??? ==11923== Allocation context: unknown. ==11923== {
-
wine-bugs＠winehq.org -
WineHQ Bugzilla