Bug 109020

Could you include a copy of the /sys/devices/pci0000:00/0000:00:02.0/drm/card0/error file? Thanks.

For later reference on this bug, the mesa version is : Mesa 19.0.0-devel - padoka PPA

(In reply to Lionel Landwerlin from comment #1)
> Could you include a copy of the
> /sys/devices/pci0000:00/0000:00:02.0/drm/card0/error file? Thanks.
> 
> For later reference on this bug, the mesa version is : Mesa 19.0.0-devel -
> padoka PPA

How can I do that if my system freezes instantly and after reboot I just get: "No error state collected".

(In reply to Lionel Landwerlin from comment #1)
> Could you include a copy of the
> /sys/devices/pci0000:00/0000:00:02.0/drm/card0/error file? Thanks.
> 
> For later reference on this bug, the mesa version is : Mesa 19.0.0-devel -
> padoka PPA

"tail -f" and output to file don't work

(In reply to Nikolay Borodin from comment #3)
> (In reply to Lionel Landwerlin from comment #1)
> > Could you include a copy of the
> > /sys/devices/pci0000:00/0000:00:02.0/drm/card0/error file? Thanks.
> > 
> > For later reference on this bug, the mesa version is : Mesa 19.0.0-devel -
> > padoka PPA
> 
> "tail -f" and output to file don't work

I assume the machine isn't available on the network anymore? You can't remotely copy the file?

Maybe try to run the following command as root, trying to copy the file every 1 second : 

while true; do sleep 1; lines=`wc -l /sys/devices/pci0000:00/0000:00:02.0/drm/card0/error | cut -d ' ' -f 1`; if [ $lines -gt 5 ]; then cp /sys/devices/pci0000:00/0000:00:02.0/drm/card0/error /home/; fi; done

(In reply to Lionel Landwerlin from comment #4)
> (In reply to Nikolay Borodin from comment #3)
> > (In reply to Lionel Landwerlin from comment #1)
> > > Could you include a copy of the
> > > /sys/devices/pci0000:00/0000:00:02.0/drm/card0/error file? Thanks.
> > > 
> > > For later reference on this bug, the mesa version is : Mesa 19.0.0-devel -
> > > padoka PPA
> > 
> > "tail -f" and output to file don't work
> 
> I assume the machine isn't available on the network anymore? You can't
> remotely copy the file?
> 
> Maybe try to run the following command as root, trying to copy the file
> every 1 second : 
> 
> while true; do sleep 1; lines=`wc -l
> /sys/devices/pci0000:00/0000:00:02.0/drm/card0/error | cut -d ' ' -f 1`; if
> [ $lines -gt 5 ]; then cp
> /sys/devices/pci0000:00/0000:00:02.0/drm/card0/error /home/; fi; done

It doesn't make any sense because my system freezes before it will copy a file and it also fails to reset GPU.

Game Over.

My experience is the system doesn't immediately become unresponsive.
Proof is, the i915 driver is detecting the hang by using a timer after which no operation currently running has made any progress.

So there is a pretty good chance you'll be able to capture the file.
That loop might need an additional sync to ensure the file makes it to the disk.

(In reply to Lionel Landwerlin from comment #6)
> My experience is the system doesn't immediately become unresponsive.
> Proof is, the i915 driver is detecting the hang by using a timer after which
> no operation currently running has made any progress.
> 
> So there is a pretty good chance you'll be able to capture the file.
> That loop might need an additional sync to ensure the file makes it to the
> disk.

I forgot this but adding i915.reset=0 to the command line of the kernel might also help so that the i915 driver doesn't attempt to touch anything once the hang is detected.

Created attachment 142796 [details]
error file

(In reply to Lionel Landwerlin from comment #7)
> (In reply to Lionel Landwerlin from comment #6)
> > My experience is the system doesn't immediately become unresponsive.
> > Proof is, the i915 driver is detecting the hang by using a timer after which
> > no operation currently running has made any progress.
> > 
> > So there is a pretty good chance you'll be able to capture the file.
> > That loop might need an additional sync to ensure the file makes it to the
> > disk.
> 
> I forgot this but adding i915.reset=0 to the command line of the kernel
> might also help so that the i915 driver doesn't attempt to touch anything
> once the hang is detected.

That's better.

https://bugs.freedesktop.org/attachment.cgi?id=142796

I would like to see if I can reproduce to test a potential fix.
Unfortunately using Mesa 19.0.0-devel I'm running into a warning the prevents me from running the game : "DX11 feature level 10.0 is required to run the engine"

Can you share the setup you're using?
I assumed you were launching that game from Steam.

(In reply to Lionel Landwerlin from comment #10)
> I would like to see if I can reproduce to test a potential fix.
> Unfortunately using Mesa 19.0.0-devel I'm running into a warning the
> prevents me from running the game : "DX11 feature level 10.0 is required to
> run the engine"
> 
> Can you share the setup you're using?
> I assumed you were launching that game from Steam.

- Can you provide me your glxinfo output?
- What version of dxvk do you use?
- Do you have vulkan libraries installed?

(In reply to Nikolay Borodin from comment #11)
> (In reply to Lionel Landwerlin from comment #10)
> > I would like to see if I can reproduce to test a potential fix.
> > Unfortunately using Mesa 19.0.0-devel I'm running into a warning the
> > prevents me from running the game : "DX11 feature level 10.0 is required to
> > run the engine"
> > 
> > Can you share the setup you're using?
> > I assumed you were launching that game from Steam.
> 
> - Can you provide me your glxinfo output?
> - What version of dxvk do you use?

I'm using Proton 3.16 which seems to be using dxvk : 20353f6f62f802bab9618b19940748af19cc1812 dxvk (v0.90-141-g20353f6)

I'm guessing that's the issue.


> - Do you have vulkan libraries installed?

Yes, latest loader & mesa master.

(In reply to Lionel Landwerlin from comment #10)
> I would like to see if I can reproduce to test a potential fix.
> Unfortunately using Mesa 19.0.0-devel I'm running into a warning the
> prevents me from running the game : "DX11 feature level 10.0 is required to
> run the engine"
> 
> Can you share the setup you're using?
> I assumed you were launching that game from Steam.

Btw, can I test this potential fix too?

(In reply to Nikolay Borodin from comment #13)
> (In reply to Lionel Landwerlin from comment #10)
> > I would like to see if I can reproduce to test a potential fix.
> > Unfortunately using Mesa 19.0.0-devel I'm running into a warning the
> > prevents me from running the game : "DX11 feature level 10.0 is required to
> > run the engine"
> > 
> > Can you share the setup you're using?
> > I assumed you were launching that game from Steam.
> 
> Btw, can I test this potential fix too?

Sure, if you can just cherry pick that commit for Mesa : https://github.com/djdeath/mesa/commit/02da2df87f59ac784117727d986a40df173df0ae

Unfortunately I haven't managed to compile a newer version of proton (including the right version of wine & dxvk) that actually manages to start the game yet.

Created attachment 142839 [details]
error file V2

(In reply to Lionel Landwerlin from comment #14)
> (In reply to Nikolay Borodin from comment #13)
> > (In reply to Lionel Landwerlin from comment #10)
> > > I would like to see if I can reproduce to test a potential fix.
> > > Unfortunately using Mesa 19.0.0-devel I'm running into a warning the
> > > prevents me from running the game : "DX11 feature level 10.0 is required to
> > > run the engine"
> > > 
> > > Can you share the setup you're using?
> > > I assumed you were launching that game from Steam.
> > 
> > Btw, can I test this potential fix too?
> 
> Sure, if you can just cherry pick that commit for Mesa :
> https://github.com/djdeath/mesa/commit/
> 02da2df87f59ac784117727d986a40df173df0ae
> 
> Unfortunately I haven't managed to compile a newer version of proton
> (including the right version of wine & dxvk) that actually manages to start
> the game yet.

The crash is still present:

https://bugs.freedesktop.org/attachment.cgi?id=142839

I did these commands:

export LIBGL_DRIVERS_PATH=/home/monsterovich/Distfiles/compile/mesa/lib
export LD_LIBRARY_PATH=/home/monsterovich/Distfiles/compile/mesa/lib

glxinfo reported that valid driver is loaded.

OpenGL version string: 3.0 Mesa 19.0.0-devel (git-cdf5e03545)

(In reply to Nikolay Borodin from comment #16)
> (In reply to Lionel Landwerlin from comment #14)
> > (In reply to Nikolay Borodin from comment #13)
> > > (In reply to Lionel Landwerlin from comment #10)
> > > > I would like to see if I can reproduce to test a potential fix.
> > > > Unfortunately using Mesa 19.0.0-devel I'm running into a warning the
> > > > prevents me from running the game : "DX11 feature level 10.0 is required to
> > > > run the engine"
> > > > 
> > > > Can you share the setup you're using?
> > > > I assumed you were launching that game from Steam.
> > > 
> > > Btw, can I test this potential fix too?
> > 
> > Sure, if you can just cherry pick that commit for Mesa :
> > https://github.com/djdeath/mesa/commit/
> > 02da2df87f59ac784117727d986a40df173df0ae
> > 
> > Unfortunately I haven't managed to compile a newer version of proton
> > (including the right version of wine & dxvk) that actually manages to start
> > the game yet.
> 
> The crash is still present:
> 
> https://bugs.freedesktop.org/attachment.cgi?id=142839
> 
> I did these commands:
> 
> export LIBGL_DRIVERS_PATH=/home/monsterovich/Distfiles/compile/mesa/lib
> export LD_LIBRARY_PATH=/home/monsterovich/Distfiles/compile/mesa/lib
> 
> glxinfo reported that valid driver is loaded.
> 
> OpenGL version string: 3.0 Mesa 19.0.0-devel (git-cdf5e03545)

Thanks a lot for the update, that's really helpful.

Looking at the error state, I think there could be a disconnect between the shader and how the samplers/dataports are setup.
Are you sure you run the game with the validation layers?
I would expect some errors.

(In reply to Lionel Landwerlin from comment #17)
> (In reply to Nikolay Borodin from comment #16)
> > (In reply to Lionel Landwerlin from comment #14)
> > > (In reply to Nikolay Borodin from comment #13)
> > > > (In reply to Lionel Landwerlin from comment #10)
> > > > > I would like to see if I can reproduce to test a potential fix.
> > > > > Unfortunately using Mesa 19.0.0-devel I'm running into a warning the
> > > > > prevents me from running the game : "DX11 feature level 10.0 is required to
> > > > > run the engine"
> > > > > 
> > > > > Can you share the setup you're using?
> > > > > I assumed you were launching that game from Steam.
> > > > 
> > > > Btw, can I test this potential fix too?
> > > 
> > > Sure, if you can just cherry pick that commit for Mesa :
> > > https://github.com/djdeath/mesa/commit/
> > > 02da2df87f59ac784117727d986a40df173df0ae
> > > 
> > > Unfortunately I haven't managed to compile a newer version of proton
> > > (including the right version of wine & dxvk) that actually manages to start
> > > the game yet.
> > 
> > The crash is still present:
> > 
> > https://bugs.freedesktop.org/attachment.cgi?id=142839
> > 
> > I did these commands:
> > 
> > export LIBGL_DRIVERS_PATH=/home/monsterovich/Distfiles/compile/mesa/lib
> > export LD_LIBRARY_PATH=/home/monsterovich/Distfiles/compile/mesa/lib
> > 
> > glxinfo reported that valid driver is loaded.
> > 
> > OpenGL version string: 3.0 Mesa 19.0.0-devel (git-cdf5e03545)
> 
> Thanks a lot for the update, that's really helpful.
> 
> Looking at the error state, I think there could be a disconnect between the
> shader and how the samplers/dataports are setup.
> Are you sure you run the game with the validation layers?
> I would expect some errors.

Are you talking about this?:

export VK_INSTANCE_LAYERS=VK_LAYER_LUNARG_standard_validation

(In reply to Nikolay Borodin from comment #18)
> (In reply to Lionel Landwerlin from comment #17)
> > (In reply to Nikolay Borodin from comment #16)
> > > (In reply to Lionel Landwerlin from comment #14)
> > > > (In reply to Nikolay Borodin from comment #13)
> > > > > (In reply to Lionel Landwerlin from comment #10)
> > > > > > I would like to see if I can reproduce to test a potential fix.
> > > > > > Unfortunately using Mesa 19.0.0-devel I'm running into a warning the
> > > > > > prevents me from running the game : "DX11 feature level 10.0 is required to
> > > > > > run the engine"
> > > > > > 
> > > > > > Can you share the setup you're using?
> > > > > > I assumed you were launching that game from Steam.
> > > > > 
> > > > > Btw, can I test this potential fix too?
> > > > 
> > > > Sure, if you can just cherry pick that commit for Mesa :
> > > > https://github.com/djdeath/mesa/commit/
> > > > 02da2df87f59ac784117727d986a40df173df0ae
> > > > 
> > > > Unfortunately I haven't managed to compile a newer version of proton
> > > > (including the right version of wine & dxvk) that actually manages to start
> > > > the game yet.
> > > 
> > > The crash is still present:
> > > 
> > > https://bugs.freedesktop.org/attachment.cgi?id=142839
> > > 
> > > I did these commands:
> > > 
> > > export LIBGL_DRIVERS_PATH=/home/monsterovich/Distfiles/compile/mesa/lib
> > > export LD_LIBRARY_PATH=/home/monsterovich/Distfiles/compile/mesa/lib
> > > 
> > > glxinfo reported that valid driver is loaded.
> > > 
> > > OpenGL version string: 3.0 Mesa 19.0.0-devel (git-cdf5e03545)
> > 
> > Thanks a lot for the update, that's really helpful.
> > 
> > Looking at the error state, I think there could be a disconnect between the
> > shader and how the samplers/dataports are setup.
> > Are you sure you run the game with the validation layers?
> > I would expect some errors.
> 
> Are you talking about this?:
> 
> export VK_INSTANCE_LAYERS=VK_LAYER_LUNARG_standard_validation

Yes. I just saw that Proton got upgrade with DXVk 0.94 so will give another try directly on Steam.

> > Are you talking about this?:
> > 
> > export VK_INSTANCE_LAYERS=VK_LAYER_LUNARG_standard_validation
> 
> Yes. I just saw that Proton got upgrade with DXVk 0.94 so will give another
> try directly on Steam.

I always had it enabled.

It is clear what is the cause of the bug? Can I help with anything else?

I encountered this GPU hang today and I think it's related to the way the Intel HD driver allocates VRAM for Vulkan in Linux. I tested the game Darksiders 3 with DXVK 1.3 (as of writing).

I have 2 systems both running a freshly installed Xubuntu 19.04.

1st System: Desktop with i3-7100 CPU and 8GB RAM (DDR4 2400MHZ) - The game freezes soon after Launch with no possibility to switch to any tty's and close it. A hard reset was required.

2nd System: Laptop with i3-5005U and 16GB RAM (DDR3 1600MHZ) - The game runs and plays without issues. See Screenshot: https://i.imgur.com/R3DLG9N.jpg


With these results I believe that the Intel HD allocates 1GB V-RAM on a system with 8GB Ram whereas with 16GB it allocates 2GB V-RAM.

Further inspection of this I ran the game on my main gaming system with a GTX1060 6GB, the game was already hitting 1GB V-RAM.

Unfortunately I lack the skills to go deeper in this.

(In reply to Jonathan Farrugia from comment #22)
> With these results I believe that the Intel HD allocates 1GB V-RAM on a
> system with 8GB Ram whereas with 16GB it allocates 2GB V-RAM.

Please disregard this part from my previous comment, I was wrong.


I found out something else though from my 2nd test tonight - After some gameplay on my 16GB laptop, DXVK allocated 3.9GB just for V-RAM, (pictured here: https://i.imgur.com/JzCY74f.jpg). According to intel's ark spec sheet, the GPU can allocate as much as 16GB for V-RAM.

Secondly I checked the specs on my desktop motherboard where the i3-7100 is installed (Link: https://www.asrock.com/mb/Intel/H110M-STX/index.asp) and at the "Graphics" part you can see it clearly states 'Max. shared memory 1024MB'. 

For comparison purposes I checked other motherboard specs and most of them list equal, lower and higher values and in some cases it's not listed (presumably as much as there is system ram).

I will run further tests tomorrow and also try to find an explanation of why the game doesn't crash with WineD3D.

Today I removed one of the two 8GB sticks on my laptop to rule out the ram disparity between the 2 test systems. The game ran fine.


On the desktop side I ran more games which would allocate a fair amount of V-RAM, one of test cases here was Guild Wars 2 which ran fine and this pretty much rules out it's not a memory issue as I was initially keen to believe. Screenshot: https://i.imgur.com/qSVVZ7Y.jpg  


I think the issue here is specific to Gen9 graphics.

Any suggestions how I can proceed and provide additional information?

hello guys.
I found out that current issue with the game is the same to
https://bugs.freedesktop.org/show_bug.cgi?id=110228

and the fix was landed into kernel here https://bugs.freedesktop.org/show_bug.cgi?id=110998

I tested it and confirmed that Darksides 3 apitrace stopped hang the system after that patch. So I would say that current issue also can be closed.

I can happily confirm that the patch referred by Denis (https://bugs.freedesktop.org/show_bug.cgi?id=110998) fixes the issue with Darksiders 3 (UE4 game) on Kabylake GPU.

I think this bug may safely be considered fixed. Marking it as a duplicate to ticket 110998.

*** This bug has been marked as a duplicate of bug 110998 ***

Created attachment 145338 [details]
Darksiders 3 DXVK Kabylake Intel HD

Screenshot of the game running with DXVK on Kabylake i3-7100 using Intel HD GPU.