[valgrind] [Bug 253657] missing some amd64 to let wine/amd64 run on valgrind/amd64

[Daniel Lehman]

https://bugs.kde.org/show_bug.cgi?id=253657

--- Comment #30 from Daniel Lehman  ---
Created attachment 125416
  --> https://bugs.kde.org/attachment.cgi?id=125416=edit
bool to skip msvc x64 code path in non-wine case

> The obvious fix is to keep a boolean indicating whether any Windows unwind
> info has been loaded, and if false, just skip this call.  That would remove 
> the
> perf hit for non-Wine use.

something like the attached?  i based it on FPO_info_present.  the patch is
rebased against the rollup.  it still works with the above test case on 64-bit
wine but i verified it is not called with non-wine (using python -c "import
numpy")

-- 
You are receiving this mail because:
You are watching all bug changes.

[valgrind] [Bug 253657] missing some amd64 to let wine/amd64 run on valgrind/amd64

[Julian Seward]

https://bugs.kde.org/show_bug.cgi?id=253657

Julian Seward  changed:

   What|Removed |Added

  Attachment #59716|0   |1
is obsolete||
 Attachment #122256|0   |1
is obsolete||

--- Comment #29 from Julian Seward  ---
Created attachment 125367
  --> https://bugs.kde.org/attachment.cgi?id=125367=edit
Rollup patch, which is simply the 15 patches merged

Here's a rollup patch, which is just the previous 15 patches merged and
with a bit of reformatting.  It applies to the trunk as at 24 Jan 2020.

I haven't landed it because the change in coregrind/m_stacktrace.c potentially
gives a large performance problem.  If the call to VG_(use_CFI_info) fails,
then
we now try with VG_(use_MSVC_x64_info).  That is guaranteed to fail in "normal"
(non-Wine) use.  But before it fails it could potentially visit the entire
debugInfo
list, which can contains hundreds of DebugInfos for large complex applications.

The obvious fix is to keep a boolean indicating whether any Windows unwind
info has been loaded, and if false, just skip this call.  That would remove the
perf hit for non-Wine use.

-- 
You are receiving this mail because:
You are watching all bug changes.

[valgrind] [Bug 253657] missing some amd64 to let wine/amd64 run on valgrind/amd64

[Daniel Lehman]

https://bugs.kde.org/show_bug.cgi?id=253657

--- Comment #28 from Daniel Lehman  ---
nothing new to add for now

-- 
You are receiving this mail because:
You are watching all bug changes.

[valgrind] [Bug 253657] missing some amd64 to let wine/amd64 run on valgrind/amd64

[Julian Seward]

https://bugs.kde.org/show_bug.cgi?id=253657

--- Comment #27 from Julian Seward  ---
(In reply to Daniel Lehman from comment #25)
> Created attachment 122256 [details]
> patches for wine64 and pdbs

I will try and look at these patches in the next couple of weeks, to
see if they can be integrated.  If you have any more recent version
of them, can you please attach them instead?

-- 
You are receiving this mail because:
You are watching all bug changes.

[valgrind] [Bug 253657] missing some amd64 to let wine/amd64 run on valgrind/amd64

[Tom Hughes]

https://bugs.kde.org/show_bug.cgi?id=253657

Tom Hughes  changed:

   What|Removed |Added

 CC||eekn...@gmail.com

--- Comment #26 from Tom Hughes  ---
*** Bug 414659 has been marked as a duplicate of this bug. ***

-- 
You are receiving this mail because:
You are watching all bug changes.

[valgrind] [Bug 253657] missing some amd64 to let wine/amd64 run on valgrind/amd64

[Daniel Lehman]

https://bugs.kde.org/show_bug.cgi?id=253657

Daniel Lehman  changed:

   What|Removed |Added

 Attachment #114890|0   |1
is obsolete||

--- Comment #25 from Daniel Lehman  ---
Created attachment 122256
  --> https://bugs.kde.org/attachment.cgi?id=122256=edit
patches for wine64 and pdbs

rebased to latest master with fix from:
https://bugs.kde.org/show_bug.cgi?id=400538
and retested with:
- wine 4.13
- vs2017 15.9.15

-- 
You are receiving this mail because:
You are watching all bug changes.

[valgrind] [Bug 253657] missing some amd64 to let wine/amd64 run on valgrind/amd64

[Alex Henrie]

https://bugs.kde.org/show_bug.cgi?id=253657

Alex Henrie  changed:

   What|Removed |Added

 CC||alexhenri...@gmail.com

-- 
You are receiving this mail because:
You are watching all bug changes.

[valgrind] [Bug 253657] missing some amd64 to let wine/amd64 run on valgrind/amd64

[Alex Henrie]

https://bugs.kde.org/show_bug.cgi?id=253657

Alex Henrie  changed:

   What|Removed |Added

   See Also||https://bugs.kde.org/show_b
   ||ug.cgi?id=400538

-- 
You are receiving this mail because:
You are watching all bug changes.

[valgrind] [Bug 253657] missing some amd64 to let wine/amd64 run on valgrind/amd64

[Austin English]

https://bugs.kde.org/show_bug.cgi?id=253657

--- Comment #24 from Austin English  ---
(In reply to Daniel Lehman from comment #23)
> > 0030:err:seh:NtRaiseException Unhandled exception code c01d flags 0 
> > addr 0x7bc95b63
> 
> that's probably the iretq instruction from set_cpu_context
> 
> > What VALGRIND_OPTS are you using?
> 
> -v --trace-children=yes --error-limit=no --log-file=output.txt
> --leak-check=full --leak-resolution=high --show-leak-kinds=all
> 
> > did I miss a step?
> 
> no, sorry, i forgot that i had to create a symlink to find the pdb:
> ln -s $HOME $WINEPREFIX/drive_Y
> because the baked-in path is Y:/stuff/leakage

That's weird..when first trying this, I missed your 'drive_Y' bit and was
symlinking to dosdevices/y:, which failed. With drive_Y, it works (not
valgrind's problem, of course).

I can verify that works for me:
+ /home/austin/src/valgrind/vg-in-place wine64 leakage.exe
preloader: Warning: failed to reserve range 0011-6800
==32694== Conditional jump or move depends on uninitialised value(s)
==32694==at 0x14005F7DE: __strncnt (strncnt.cpp:21)
==32694==by 0x7B9D4E0A8F2C: ???
==32694==  Uninitialised value was created by a stack allocation
==32694==at 0x14004BB17: setSBUpLow (in
/tmp/tmp.sYSJGwj3qz/stuff/leakage/leakage.exe)
==32694== 
7E018520
setframe: 7E20FCD0
7E01C6D0
006e:fixme:kernelbase:AppPolicyGetProcessTerminationMethod 0xfffa,
0x7e20fd10
==32694== 12,345 bytes in 1 blocks are definitely lost in loss record 92 of 93
==32694==at 0x7BC5BB35: initialize_block (heap.c:238)
==32694==by 0x7BC5BB35: RtlAllocateHeap (???:0)
==32694==by 0x140006BDA: a (leakage.c:9)
==32694==by 0x140006BF8: b (leakage.c:14)
==32694==by 0x140006C18: c (leakage.c:19)
==32694==by 0x140006CF8: main (leakage.c:43)
==32694== 
==32694== 23,456 bytes in 1 blocks are definitely lost in loss record 93 of 93
==32694==at 0x7BC5BB35: initialize_block (heap.c:238)
==32694==by 0x7BC5BB35: RtlAllocateHeap (???:0)
==32694==by 0x140006C8C: setframe (leakage.c:28)
==32694==by 0x140006CB8: d (leakage.c:33)
==32694==by 0x140006CD8: e (leakage.c:38)
==32694==by 0x140006D0C: main (leakage.c:44)
==32694== 

==

For our valgrind friends, if you'd like to test Daniel's patch, here's a short
script to do so. Get wine64 from your package manager (4.0 stable should work,
or probably any version), then run the script, no special wine knowledge needed
:)

#!/bin/bash
# Requires wine64 (nothing special needed, get it from your package manager)
# and valgrind, of course

export VALGRIND="${VALGRIND:-valgrind}"
export VALGRIND_OPTS="-q --trace-children=yes --track-origins=yes
--leak-check=full --num-callers=20 
--vex-iropt-register-updates=allregs-at-mem-access"

export WINEPREFIX="${WINEPREFIX:-$HOME/.wine}"
tmpdir="$(mktemp -d)"

cd "$tmpdir"
mkdir -p stuff/leakage
cd stuff/leakage
wget -O stuff.tar.bz2 "https://bugs.kde.org/attachment.cgi?id=118764;
tar xjf stuff.tar.bz2

ln -sf "$tmpdir" "$WINEPREFIX/drive_Y"

# Here we go:
"$VALGRIND" wine64 leakage.exe 

cd
rm -rf "$tmpdir"

-- 
You are receiving this mail because:
You are watching all bug changes.

[valgrind] [Bug 253657] missing some amd64 to let wine/amd64 run on valgrind/amd64

[Daniel Lehman]

https://bugs.kde.org/show_bug.cgi?id=253657

--- Comment #23 from Daniel Lehman  ---
> 0030:err:seh:NtRaiseException Unhandled exception code c01d flags 0 addr 
> 0x7bc95b63

that's probably the iretq instruction from set_cpu_context

> What VALGRIND_OPTS are you using?

-v --trace-children=yes --error-limit=no --log-file=output.txt
--leak-check=full --leak-resolution=high --show-leak-kinds=all

> did I miss a step?

no, sorry, i forgot that i had to create a symlink to find the pdb:
ln -s $HOME $WINEPREFIX/drive_Y
because the baked-in path is Y:/stuff/leakage

-- 
You are receiving this mail because:
You are watching all bug changes.

[valgrind] [Bug 253657] missing some amd64 to let wine/amd64 run on valgrind/amd64

[Austin English]

https://bugs.kde.org/show_bug.cgi?id=253657

--- Comment #22 from Austin English  ---
That said, it's still an improvement. Without the patchset, I get:
austin@laptop:~$ /opt/valgrind/bin/valgrind /opt/oldwow64/wine-4.0-rc1/bin/wine
leakage.exe ; echo $?
preloader: Warning: failed to reserve range 0011-6800
preloader: Warning: failed to reserve range 7f00-8200
preloader: Warning: failed to reserve range 0011-6800
0030:err:seh:segv_handler Got unexpected trap 0
==6737== Invalid write of size 8
==6737==at 0x7BC95B68: ??? (in
/opt/oldwow64/wine-4.0-rc1/lib64/wine/ntdll.dll.so)
==6737==by 0x7BC95B62: ??? (in
/opt/oldwow64/wine-4.0-rc1/lib64/wine/ntdll.dll.so)
==6737==by 0x7BC95C3A: ??? (in
/opt/oldwow64/wine-4.0-rc1/lib64/wine/ntdll.dll.so)
==6737==  Address 0x7e20f4b8 is in a rw- anonymous segment
==6737== 
0030:err:seh:NtRaiseException Unhandled exception code c01d flags 0 addr
0x7bc95b63
29

-- 
You are receiving this mail because:
You are watching all bug changes.

[valgrind] [Bug 253657] missing some amd64 to let wine/amd64 run on valgrind/amd64

[Austin English]

https://bugs.kde.org/show_bug.cgi?id=253657

--- Comment #21 from Austin English  ---
(In reply to Daniel Lehman from comment #20)
> Created attachment 118764 [details]
> leakage.exe and pdb
> 
> attached tested with:
> - Visual Studio 2017 (15.9.8)
> - wine 4.3
> - valgrind-3.15.0.GIT (4816357b5c7ee5284cdf72800a81d2dd1845388f)

Thanks. I wasn't able to reproduce, with same valgrind git commit (+ your
patches) and wine-4.0-rc1 (last wow64 build I have handy, but given that you
used 3.15 before I don't think it matters).

What VALGRIND_OPTS are you using? I tried with:
austin@laptop:~$ export VALGRIND_OPTS="-q --trace-children=yes
--track-origins=yes --leak-check=full --num-callers=20 
--vex-iropt-register-updates=allregs-at-mem-access"

and got:
austin@laptop:~$ ~/src/valgrind/vg-in-place /opt/oldwow64/wine-4.0-rc1/bin/wine
leakage.exe ; echo $?
preloader: Warning: failed to reserve range 0011-6800
preloader: Warning: failed to reserve range 7f00-8200
preloader: Warning: failed to reserve range 0011-6800
==6365== Conditional jump or move depends on uninitialised value(s)
==6365==at 0x14005F7DE: ???
==6365==by 0x51B686B7C272: ???
==6365==  Uninitialised value was created by a stack allocation
==6365==at 0x14004BB17: ???
==6365== 
7E01B5F0
setframe: 7E20FCD0
7E01F7A0
==6365== 12,345 bytes in 1 blocks are definitely lost in loss record 86 of 88
==6365==at 0x7BC5B385: initialize_block (heap.c:238)
==6365==by 0x7BC5B385: RtlAllocateHeap (???:0)
==6365==by 0x140006BDA: ???
==6365== 
==6365== 23,456 bytes in 1 blocks are definitely lost in loss record 88 of 88
==6365==at 0x7BC5B385: initialize_block (heap.c:238)
==6365==by 0x7BC5B385: RtlAllocateHeap (???:0)
==6365==by 0x140006C8C: ???
==6365==by 0x51B686B7CD62: ???
==6365== 
0

did I miss a step?

-- 
You are receiving this mail because:
You are watching all bug changes.

[valgrind] [Bug 253657] missing some amd64 to let wine/amd64 run on valgrind/amd64

[Daniel Lehman]

https://bugs.kde.org/show_bug.cgi?id=253657

--- Comment #20 from Daniel Lehman  ---
Created attachment 118764
  --> https://bugs.kde.org/attachment.cgi?id=118764=edit
leakage.exe and pdb

attached tested with:
- Visual Studio 2017 (15.9.8)
- wine 4.3
- valgrind-3.15.0.GIT (4816357b5c7ee5284cdf72800a81d2dd1845388f)

-- 
You are receiving this mail because:
You are watching all bug changes.

[valgrind] [Bug 253657] missing some amd64 to let wine/amd64 run on valgrind/amd64

[Austin English]

https://bugs.kde.org/show_bug.cgi?id=253657

--- Comment #19 from Austin English  ---
(In reply to Daniel Lehman from comment #18)
> Created attachment 114891 [details]
> simple test case for wine64 pdb

Hi Daniel,

Would you mind attaching a pre-built binary? I'd like to test this, but I don't
currently have Visual Studio set up (and won't be able to for a week or so at
the earliest).

Thanks!
-Austin

-- 
You are receiving this mail because:
You are watching all bug changes.

[valgrind] [Bug 253657] missing some amd64 to let wine/amd64 run on valgrind/amd64

[Daniel Lehman]

https://bugs.kde.org/show_bug.cgi?id=253657

Daniel Lehman  changed:

   What|Removed |Added

 CC||dlehma...@gmail.com

--- Comment #18 from Daniel Lehman  ---
Created attachment 114891
  --> https://bugs.kde.org/attachment.cgi?id=114891=edit
simple test case for wine64 pdb

-- 
You are receiving this mail because:
You are watching all bug changes.

[valgrind] [Bug 253657] missing some amd64 to let wine/amd64 run on valgrind/amd64

[Daniel Lehman]

https://bugs.kde.org/show_bug.cgi?id=253657

--- Comment #17 from Daniel Lehman  ---
Created attachment 114890
  --> https://bugs.kde.org/attachment.cgi?id=114890=edit
patches for wine64 and pdbs

The attached patches get wine64 with pdbs working.  I'm sure additional patches
will be needed, but it's a start.

The attached leakage.c produces the following output on 64-bit Wine with these
changes:
==689== 12,345 bytes in 1 blocks are definitely lost in loss record 70 of 71
==689==at 0x7BC5ACCD: initialize_block (heap.c:238)
==689==by 0x7BC5ACCD: RtlAllocateHeap (???:0)
==689==by 0x140006C0A: a (leakage.c:9)
==689==by 0x140006C28: b (leakage.c:14)
==689==by 0x140006C48: c (leakage.c:19)
==689==by 0x140006D28: main (leakage.c:43)
==689== 
==689== 23,456 bytes in 1 blocks are possibly lost in loss record 71 of 71
==689==at 0x7BC5ACCD: initialize_block (heap.c:238)
==689==by 0x7BC5ACCD: RtlAllocateHeap (???:0)
==689==by 0x140006CBC: setframe (leakage.c:28)
==689==by 0x140006CE8: d (leakage.c:33)
==689==by 0x140006D08: e (leakage.c:38)
==689==by 0x140006D3C: main (leakage.c:44)

leakage.c built 32-bit with pdb fails with and without my changes.  But with
the attached, both 32-bit and 64-bit Wine tests still run under valgrind.

Tested with the following:
- Visual Studio 2017 (15.8.3) (Optimizing Compiler Version 19.15.26726 for x64)
- Wine 3.15
- valgrind-3.14.0.GIT (097b2076013bd6082be189ab55dbdb2eb1e572b7)

-- 
You are receiving this mail because:
You are watching all bug changes.

[valgrind] [Bug 253657] missing some amd64 to let wine/amd64 run on valgrind/amd64

[Austin English]

https://bugs.kde.org/show_bug.cgi?id=253657

Austin English  changed:

   What|Removed |Added

 CC||austinengl...@gmail.com

-- 
You are receiving this mail because:
You are watching all bug changes.

[valgrind] [Bug 253657] missing some amd64 to let wine/amd64 run on valgrind/amd64

[Stas Sergeev via KDE Bugzilla]

https://bugs.kde.org/show_bug.cgi?id=253657

--- Comment #16 from Stas Sergeev  ---
Created attachment 97737
  --> https://bugs.kde.org/attachment.cgi?id=97737=edit
a test case

Sorry for delay.
Here's the segregs test case.

-- 
You are receiving this mail because:
You are watching all bug changes.