[Bug 51083] New: Deadlock in llvm-mingw compiled app

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=51083

Bug ID: 51083
           Summary: Deadlock in llvm-mingw compiled app
           Product: Wine
           Version: unspecified
          Hardware: x86-64
                OS: Linux
            Status: UNCONFIRMED
          Severity: normal
          Priority: P2
         Component: -unknown
          Assignee: [e-mail]
          Reporter: [e-mail]
      Distribution: ---

Created attachment 69945
  --> https://bugs.winehq.org/attachment.cgi?id=69945
wine log

I have one test program (pomocni-llvm-mingw.exe) who acts very strangely in
wine git version(Wine-6.7-175-G2Deb8C2825A) while in the Windows10 does not
work at all.

What the most surprise is that although the wine64 call is terminated for the
specified program, wine processes and the program itself remain in memory and
after interruptions:

[root@vlada-kuci tmp 21:17:48]$
/asoft-src/ispp-64/linux/cross-compile/llvm-mingw/pomocni-cross-llvm-mingw/pomocni-cross-mingw64.sh
> /tmp/pomocni-cross-mingw64.log 2>&1

^C[root@vlada-kuci tmp 01:00:13]$
[root@vlada-kuci tmp 01:00:15]$
[root@vlada-kuci tmp 01:00:15]$ ps -ef | grep pomocni
root      419722       1  0 00:57 pts/16   00:00:00 pomocni-llvm-mingw.exe
DRIVER=PostgreSQL;DATABASE=ispp-pprostor-beograd;SERVER=localhost;UID=vlada;PWD=vladakk;MaxVarcharSize=65536;MaxLongVarcharSize=65536;ByteaAsLongVarBinary=1
root      419932    1456  0 01:08 pts/16   00:00:00 grep pomocni
[root@vlada-kuci tmp 01:08:41]$ ps -ef | grep win
root      419631       1  0 00:55 ?        00:00:25
/mnt/WD-Elements-25A1/src/wine-git/install64/bin/wineserver
root      419637       1  0 00:55 ?        00:00:00
C:\windows\system32\services.exe
root      419640       1  0 00:55 ?        00:00:13
C:\windows\system32\winedevice.exe
root      419647       1  0 00:55 ?        00:00:00
C:\windows\system32\plugplay.exe
root      419655       1  1 00:55 ?        00:05:57
C:\windows\system32\winedevice.exe
root      419672       1  0 00:55 ?        00:00:00
C:\windows\system32\svchost.exe -k LocalServiceNetworkRestricted
root      419677       1  0 00:55 ?        00:00:00
C:\windows\system32\rpcss.exe
root      419716       1  0 00:57 pts/16   00:00:00
C:\windows\system32\conhost.exe --unix --width 0 --height 0 --server 0x30
root      422318    1456  0 06:38 pts/16   00:00:00 grep win

The program is made with the latest version of llvm-mingw:Clang Version 11.0.0
(https://github.com/llvm/llvm-project.git
176249bd6732a8044d457092ed922768724a6f06).

That same program when it is made and performed in Linux works correctly as
predicted.

In the attachment, I send the following files:

1. wine log - pomocni-cross-mingw64.log

2. exe characteristics - trt.log

3. test script - pomocni-cross-mingw64.sh

4. test program - testIsppRadenik.cpp

Vladimir Koković, DP senior(70),

Serbia, Belgrade, 1.May 2021

[Bug 51083] Deadlock in llvm-mingw compiled app

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=51083

--- Comment #1 from Vladimir Koković  ---
Created attachment 69946
  --> https://bugs.winehq.org/attachment.cgi?id=69946
exe characteristics

[Bug 51083] Deadlock in llvm-mingw compiled app

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=51083

--- Comment #2 from Vladimir Koković  ---
Created attachment 69947
  --> https://bugs.winehq.org/attachment.cgi?id=69947
test script

[Bug 51083] Deadlock in llvm-mingw compiled app

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=51083

--- Comment #3 from Vladimir Koković  ---
Created attachment 69948
  --> https://bugs.winehq.org/attachment.cgi?id=69948
test program

[Bug 51083] Deadlock in llvm-mingw compiled app

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=51083

--- Comment #4 from Vladimir Koković  ---
During the last test, I got with the last version(wine-6.7-246-gcda4abac985) in
the log file next:
0118:err:ntdll:RtlpWaitForCriticalSection section 000000007BC61540
"../wine/dlls/ntdll/loader.c: loader_section" wait timed out in thread 0118,
blocked by 0114, retrying (60 sec)
0118:err:ntdll:RtlpWaitForCriticalSection section 000000007BC61540
"../wine/dlls/ntdll/loader.c: loader_section" wait timed out in thread 0118,
blocked by 0114, retrying (60 sec)
0118:err:ntdll:RtlpWaitForCriticalSection section 000000007BC61540
"../wine/dlls/ntdll/loader.c: loader_section" wait timed out in thread 0118,
blocked by 0114, retrying (60 sec)
...
But $WINEPREFIX/bin/winedbg --command "info proc" shows:
 pid      threads  executable (all id:s are in hex)
 0000011c 1        'start.exe'
 00000124 1        _ 'conhost.exe'
 00000060 3        'explorer.exe'
 00000038 7        'services.exe'
 000000d4 6        _ 'rpcss.exe'
 000000c0 3        _ 'svchost.exe'
 00000084 7        _ 'winedevice.exe'
 00000068 5        _ 'plugplay.exe'
 00000044 4        _ 'winedevice.exe'
 00000020 1        'start.exe'
 00000100 1        _ 'cmd.exe'
 00000110 2           _ 'pomocni-llvm-mingw.exe'
 000000f8 1        _ 'conhost.exe'

And from this, it is seen that 0118 and 0114 do not exist in the list!

[Bug 51083] Deadlock in llvm-mingw compiled app

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=51083

--- Comment #5 from Vladimir Koković  ---
wine64 version=wine-6.7-315-ga64861fb66b

And now a new moment I didn't expect, the problem with the GCC compiled exe:
0114:err:ntdll:RtlpWaitForCriticalSection section 00000000709DB2E0 "?" wait
timed out in thread 0114, blocked by 0000, retrying (60 sec)
0114:err:ntdll:RtlpWaitForCriticalSection section 00000000709DB2E0 "?" wait
timed out in thread 0114, blocked by 0000, retrying (60 sec)
0114:err:ntdll:RtlpWaitForCriticalSection section 00000000709DB2E0 "?" wait
timed out in thread 0114, blocked by 0000, retrying (60 sec)
...
The question is what does thread 0000 mean ?

[Bug 51083] Deadlock in llvm-mingw compiled app

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=51083

--- Comment #6 from Vladimir Koković  ---
The same problem exists with wine version=wine-6.8

[Bug 51083] Deadlock in llvm-mingw compiled app

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=51083

--- Comment #7 from Vladimir Koković  ---
The same problem exists with wine version=wine-6.9

[Bug 51083] Deadlock in llvm-mingw compiled app

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=51083

Damjan Jovanovic  changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |[e-mail]

--- Comment #8 from Damjan Jovanovic  ---
Please start winedbg during the deadlock, run "bt all" and post the output.
Hopefully we can see what that other thread is doing that's keeping it from
releasing the lock (unless it leaked the lock instead).

[Bug 51083] Deadlock in llvm-mingw compiled app

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=51083

Damjan Jovanovic  changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
           Keywords|                            |download, source, testcase
            Version|unspecified                 |6.7

[Bug 51083] Deadlock in llvm-mingw compiled app

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=51083

--- Comment #9 from Vladimir Koković  ---
(In reply to Damjan Jovanovic from comment #8)
> Please start winedbg during the deadlock, run "bt all" and post the output.
> Hopefully we can see what that other thread is doing that's keeping it from
> releasing the lock (unless it leaked the lock instead).

My primary question is how to start winedbg ?
Here's what the situation is at that moment.

ps -ef | grep pomocni
root      367122    1303  0 07:10 pts/16   00:00:00 /bin/bash
/asoft-src/ispp-64/linux/cross-compile/llvm-mingw/pomocni-cross-llvm-mingw/pomocni-cross-mingw64.sh
root      367187       1  0 07:10 pts/16   00:00:00 C:\windows\system32\cmd.exe
/C z:\tmp\vk-pomocni-cross-llvm-mingw.bat
root      367191       1  0 07:10 pts/16   00:00:00 pomocni-llvm-mingw.exe
DRIVER=PostgreSQL;DATABASE=ispp-pprostor-beograd;SERVER=localhost;UID=vlada;PWD=vladakk;MaxVarcharSize=65536;MaxLongVarcharSize=65536;ByteaAsLongVarBinary=1
root      367218    1305  0 07:14 pts/18   00:00:00 grep pomocni

pid      threads  executable (all id:s are in hex)
 0000011c 1        'start.exe'
 00000124 1        _ 'conhost.exe'
 00000068 3        'explorer.exe'
 00000038 7        'services.exe'
 000000d4 6        _ 'rpcss.exe'
 000000c0 3        _ 'svchost.exe'
 00000080 7        _ 'winedevice.exe'
 00000060 5        _ 'plugplay.exe'
 00000044 4        _ 'winedevice.exe'
 00000020 1        'start.exe'
 00000100 1        _ 'cmd.exe'
 00000110 2           _ 'pomocni-llvm-mingw.exe'
 000000f8 1        _ 'conhost.exe'

0118:err:ntdll:RtlpWaitForCriticalSection section 000000007BC61540
"../wine/dlls/ntdll/loader.c: loader_section" wait timed out in thread 0118,
blocked by 0114, retrying (60 sec)
0118:err:ntdll:RtlpWaitForCriticalSection section 000000007BC61540
"../wine/dlls/ntdll/loader.c: loader_section" wait timed out in thread 0118,
blocked by 0114, retrying (60 sec)
0118:err:ntdll:RtlpWaitForCriticalSection section 000000007BC61540
"../wine/dlls/ntdll/loader.c: loader_section" wait timed out in thread 0118,
blocked by 0114, retrying (60 sec)

0118 and 0114 do not exist !

[Bug 51083] Deadlock in llvm-mingw compiled app

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=51083

--- Comment #10 from Damjan Jovanovic  ---
(In reply to Vladimir Koković from comment #9)
> 0118 and 0114 do not exist !

Yes, because those are thread IDs, not process IDs. Try "info thread".

Please run this:
winedbg --command "bt all"
during the deadlock, and post the output, as well as wine's console output.

[Bug 51083] Deadlock in llvm-mingw compiled app

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=51083

Zebediah Figura  changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |[e-mail]

--- Comment #11 from Zebediah Figura  ---
(In reply to Damjan Jovanovic from comment #10)
> (In reply to Vladimir Koković from comment #9)
> > 0118 and 0114 do not exist !
>
> Yes, because those are thread IDs, not process IDs. Try "info thread".
>
> Please run this:
> winedbg --command "bt all"
> during the deadlock, and post the output, as well as wine's console output.

The sample binary is attached; can you do this yourself?

[Bug 51083] Deadlock in llvm-mingw compiled app

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=51083

--- Comment #12 from Damjan Jovanovic  ---
(In reply to Zebediah Figura from comment #11)
> (In reply to Damjan Jovanovic from comment #10)
> > (In reply to Vladimir Koković from comment #9)
> > > 0118 and 0114 do not exist !
> >
> > Yes, because those are thread IDs, not process IDs. Try "info thread".
> >
> > Please run this:
> > winedbg --command "bt all"
> > during the deadlock, and post the output, as well as wine's console output.
>
> The sample binary is attached; can you do this yourself?

There is no binary, only a source file with missing headers and a shell script.

[Bug 51083] Deadlock in llvm-mingw compiled app

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=51083

--- Comment #13 from Vladimir Koković  ---
(In reply to Damjan Jovanovic from comment #12)
> (In reply to Zebediah Figura from comment #11)
> > (In reply to Damjan Jovanovic from comment #10)
> > > (In reply to Vladimir Koković from comment #9)
> > > > 0118 and 0114 do not exist !
> > >
> > > Yes, because those are thread IDs, not process IDs. Try "info thread".
> > >
> > > Please run this:
> > > winedbg --command "bt all"
> > > during the deadlock, and post the output, as well as wine's console output.
> >
> > The sample binary is attached; can you do this yourself?
>
> There is no binary, only a source file with missing headers and a shell
> script.

man winedbg
SYNOPSIS
       winedbg [ options ] [ program_name [ program_arguments ] | wpid ]
where is thread ?

[Bug 51083] Deadlock in llvm-mingw compiled app

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=51083

--- Comment #14 from Damjan Jovanovic  ---
(In reply to Vladimir Koković from comment #13)
> where is thread ?

$ winedbg --command "info thread"

process  tid      prio (all id:s are in hex)
00000020 start.exe
        00000024    0
00000030 winemenubuilder.exe
        00000034    0

00000020 is the process ID for start.exe, 00000024 is the tid (thread ID)
inside that start.exe process.

But don't worry about that.
winedbg --command "bt all"
gives stack traces for all threads in all processes. That should help here. You
don't need anything else.

[Bug 51083] Deadlock in llvm-mingw compiled app

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=51083

--- Comment #15 from Vladimir Koković  ---
Oh yes, and log file is attached ?

[Bug 51083] Deadlock in llvm-mingw compiled app

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=51083

--- Comment #16 from Vladimir Koković  ---
Created attachment 70062
  --> https://bugs.winehq.org/attachment.cgi?id=70062
bt all

[Bug 51083] Deadlock in llvm-mingw compiled app

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=51083

--- Comment #17 from Damjan Jovanovic  ---
 00000110 4           _ 'pomocni-llvm-mingw.exe'

but:

Can't attach process 0110: error 5

warning: could not attach to 0110
Can't attach process 0110: error 5

warning: could not attach to 0110
Can't attach process 0110: error 5

warning: could not attach to 0110
Can't attach process 0110: error 5

warning: could not attach to 0110

Error 5 is:
#define ERROR_ACCESS_DENIED                                5

Maybe you need to disable selinux, or:
echo 0 > /proc/sys/kernel/yama/ptrace_scope
or something else that's denying you access.

[Bug 51083] Deadlock in llvm-mingw compiled app

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=51083

--- Comment #18 from Vladimir Koković  ---
(In reply to Damjan Jovanovic from comment #14)
> (In reply to Vladimir Koković from comment #13)
> > where is thread ?
>
> $ winedbg --command "info thread"
>
> process  tid      prio (all id:s are in hex)
> 00000020 start.exe
>         00000024    0
> 00000030 winemenubuilder.exe
>         00000034    0
>
> 00000020 is the process ID for start.exe, 00000024 is the tid (thread ID)
> inside that start.exe process.
>
> But don't worry about that.
> winedbg --command "bt all"
> gives stack traces for all threads in all processes. That should help here.
> You don't need anything else.

man winedbg
Stack manipulation
       bt     Print calling stack of current thread.
       bt N   Print calling stack of thread of ID N. Note: this doesn't change
the position of the current frame as manipulated by the up & dn commands).

Where is "bt all" ?

[Bug 51083] Deadlock in llvm-mingw compiled app

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=51083

--- Comment #19 from Damjan Jovanovic  ---
(In reply to Vladimir Koković from comment #18)
> Where is "bt all" ?

$ winedbg
Wine-dbg>help
...
  bt [<tid>|all]                         frame <n>
...

[Bug 51083] Deadlock in llvm-mingw compiled app

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=51083

--- Comment #20 from Vladimir Koković  ---
Now I have another situation when I did kill all wine processes !
After the start "bt all", this process is now waiting and does not open my
process 0110 but is in the "bt all" log file last process is 0104.

In my "pomocni-cross-llvm-mingw" log file is now logged in, after starting "BT
ALL" next:
0118:err:ntdll:RtlpWaitForCriticalSection section 000000007BC61540
"../wine/dlls/ntdll/loader.c: loader_section" wait timed out in thread 0118,
blocked by 0114, retrying (60 sec)
00bc:err:wineusb:event_thread_proc Error handling events: System call
interrupted (perhaps due to signal)
00bc:err:wineusb:event_thread_proc Error handling events: System call
interrupted (perhaps due to signal)
00bc:err:wineusb:event_thread_proc Error handling events: System call
interrupted (perhaps due to signal)
018c:err:ntdll:RtlpWaitForCriticalSection section 000000007BC61540
"../wine/dlls/ntdll/loader.c: loader_section" wait timed out in thread 018c,
blocked by 0114, retrying (60 sec)

When I do Kill "llvm-mingw-Linux>pomocni-llvm-mingw.exe" in "bt all" log file,
you can see the following at the end:
Process of pid=0110 has terminated
Exception c0000005

This is most likely to mean that winedbg cannot do attach to 0110 due to the
interrupt that occurred by the USB component.

[Bug 51083] Deadlock in llvm-mingw compiled app

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=51083

--- Comment #21 from Vladimir Koković  ---
The same problem exists with wine version=wine-6.10, with different numbers:
0130:err:ntdll:RtlpWaitForCriticalSection section 000000007BC61540
"../wine/dlls/ntdll/loader.c: loader_section" wait timed out in thread 0130,
blocked by 012c, retrying (60 sec)

[Bug 51083] Deadlock in llvm-mingw compiled app

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=51083

--- Comment #22 from Vladimir Koković  ---
The same problem exists with wine version=wine-6.11 ...

[Bug 51083] Deadlock in llvm-mingw compiled app

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=51083

--- Comment #23 from Vladimir Koković  ---
The same problem exists with wine version=wine-6.12 ...

[Bug 51083] Deadlock in llvm-mingw compiled app

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=51083

--- Comment #24 from Vladimir Koković  ---
(In reply to Vladimir Koković from comment #23)
> The same problem exists with wine version=wine-6.12 ...

0128:err:ntdll:RtlpWaitForCriticalSection section 000000007BC62540
"../wine/dlls/ntdll/loader.c: loader_section" wait timed out in thread 0128,
blocked by 0124, retrying (60 sec)

Additional information on the basis of 'info thread'
----------------------------------------------------
process  tid      prio (all id:s are in hex)
00000020 start.exe
    00000024    0
00000038 services.exe
    0000003c    0
    00000040    0
    0000004c    0
    00000068    0
    000000a8    0
    000000c0    0
    000000d4    0
00000044 winedevice.exe
    00000048    0
    00000054    0
    00000058    0
    0000005c    0
00000060 winedevice.exe
    00000064    0
    0000006c    0
    00000070    0
    00000074    0
    00000094    0
    00000098    0
    0000009c    0
00000078 explorer.exe
    0000007c    0
    00000084    0
    00000088    0
000000a0 plugplay.exe
    000000a4    0
    000000ac    0
    000000b0    0
    000000b4    0
000000b8 svchost.exe
    000000bc    0
    000000c4    0
    000000c8    0
000000cc rpcss.exe
    000000d0    0
    000000d8    0
    000000dc    0
    000000e0    0
    000000e4    0
    000000e8    0
000000f0 conhost.exe
    000000f4    0
000000f8 cmd.exe
    000000fc    0
00000120 pomocni-llvm-mingw.exe
    00000124    0
    00000128    0
0000012c start.exe
    00000130    0
00000134 conhost.exe
    00000138    0
0000013c cmd.exe
    00000140    0

[Bug 51083] Deadlock in llvm-mingw compiled app

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=51083

--- Comment #25 from Vladimir Koković  ---
The same problem exists with wine version=wine-6.13-76-gc518a5362b9 ...