Hi,
With (guest) kernel 5.5+ with qemu/kvm on arm64, I get lots of display corruptions leading to this kind of screen: https://openqa.opensuse.org/tests/1174521#step/yast2_i/24
I git bisected it to commit c66df701e783bc666593e6e665f13670760883ee ********** drm/virtio: switch from ttm to gem shmem helpers
virtio-gpu basically needs a sg_table for the bo, to tell the host where the backing pages for the object are. So the gem shmem helpers are a perfect fit. Some drm_gem_object_funcs need thin wrappers to update the host state, but otherwise the helpers handle everything just fine.
Once the fencing was sorted the switch was surprisingly easy and for the most part just removing the ttm code.
v4: fix drm_gem_object_funcs name.
Signed-off-by: Gerd Hoffmann kraxel@redhat.com Acked-by: Daniel Vetter daniel.vetter@ffwll.ch Reviewed-by: Chia-I Wu olvaffe@gmail.com Link: http://patchwork.freedesktop.org/patch/msgid/20190829103301.3539-15-kraxel@r... **********
I also tested kernel 5.6-rc2 which has the same bug. Without kvm, the display is fine.
Regards, Guillaume
On Thu, Feb 20, 2020 at 4:44 AM Guillaume Gardet Guillaume.Gardet@arm.com wrote:
Hi,
With (guest) kernel 5.5+ with qemu/kvm on arm64, I get lots of display corruptions leading to this kind of screen: https://openqa.opensuse.org/tests/1174521#step/yast2_i/24
Looking at the screenshot, it seems cacheline-related?
There was a change of memory type
https://lists.freedesktop.org/archives/dri-devel/2019-August/233456.html
While the guest memory type is ignored on Intel, it is honored on ARM. This attempt to fix it
https://lists.freedesktop.org/archives/dri-devel/2019-December/248271.html
does not seem to land.
I git bisected it to commit c66df701e783bc666593e6e665f13670760883ee
drm/virtio: switch from ttm to gem shmem helpers
virtio-gpu basically needs a sg_table for the bo, to tell the host where the backing pages for the object are. So the gem shmem helpers are a perfect fit. Some drm_gem_object_funcs need thin wrappers to update the host state, but otherwise the helpers handle everything just fine.
Once the fencing was sorted the switch was surprisingly easy and for the most part just removing the ttm code.
v4: fix drm_gem_object_funcs name.
Signed-off-by: Gerd Hoffmann kraxel@redhat.com Acked-by: Daniel Vetter daniel.vetter@ffwll.ch Reviewed-by: Chia-I Wu olvaffe@gmail.com Link: http://patchwork.freedesktop.org/patch/msgid/20190829103301.3539-15-kraxel@r...
I also tested kernel 5.6-rc2 which has the same bug. Without kvm, the display is fine.
Regards, Guillaume
Hi,
-----Original Message----- From: Chia-I Wu olvaffe@gmail.com Sent: 20 February 2020 19:41 To: Guillaume Gardet Guillaume.Gardet@arm.com Cc: dri-devel@lists.freedesktop.org; Gerd Hoffmann kraxel@redhat.com; Daniel Vetter daniel.vetter@ffwll.ch; Catalin Marinas Catalin.Marinas@arm.com; nd nd@arm.com Subject: Re: [Bug] virtio-gpu broken with qemu/kvm on arm64 on kernel 5.5+
On Thu, Feb 20, 2020 at 4:44 AM Guillaume Gardet Guillaume.Gardet@arm.com wrote:
Hi,
With (guest) kernel 5.5+ with qemu/kvm on arm64, I get lots of display
corruptions leading to this kind of screen:
Looking at the screenshot, it seems cacheline-related?
It could be.
There was a change of memory type
https://lists.freedesktop.org/archives/dri-devel/2019-August/233456.html
While the guest memory type is ignored on Intel, it is honored on ARM. This attempt to fix it
https://lists.freedesktop.org/archives/dri-devel/2019-December/248271.html
does not seem to land.
I applied this patch on top of 5.5.4, but it does not fix the problem. Maybe more similar changes are required?
Regards, Guillaume
I git bisected it to commit c66df701e783bc666593e6e665f13670760883ee
drm/virtio: switch from ttm to gem shmem helpers
virtio-gpu basically needs a sg_table for the bo, to tell the host where the backing pages for the object are. So the gem shmem helpers are a perfect fit. Some drm_gem_object_funcs need thin wrappers to update the host state, but otherwise the helpers handle everything just fine.
Once the fencing was sorted the switch was surprisingly easy and for the most part just removing the ttm code.
v4: fix drm_gem_object_funcs name.
Signed-off-by: Gerd Hoffmann kraxel@redhat.com Acked-by: Daniel Vetter daniel.vetter@ffwll.ch Reviewed-by: Chia-I Wu olvaffe@gmail.com Link: http://patchwork.freedesktop.org/patch/msgid/20190829103301.3539-15-kr axel@redhat.com
I also tested kernel 5.6-rc2 which has the same bug. Without kvm, the display is fine.
Regards, Guillaume
On Fri, Feb 21, 2020 at 2:06 AM Guillaume Gardet Guillaume.Gardet@arm.com wrote:
Hi,
-----Original Message----- From: Chia-I Wu olvaffe@gmail.com Sent: 20 February 2020 19:41 To: Guillaume Gardet Guillaume.Gardet@arm.com Cc: dri-devel@lists.freedesktop.org; Gerd Hoffmann kraxel@redhat.com; Daniel Vetter daniel.vetter@ffwll.ch; Catalin Marinas Catalin.Marinas@arm.com; nd nd@arm.com Subject: Re: [Bug] virtio-gpu broken with qemu/kvm on arm64 on kernel 5.5+
On Thu, Feb 20, 2020 at 4:44 AM Guillaume Gardet Guillaume.Gardet@arm.com wrote:
Hi,
With (guest) kernel 5.5+ with qemu/kvm on arm64, I get lots of display
corruptions leading to this kind of screen:
Looking at the screenshot, it seems cacheline-related?
It could be.
There was a change of memory type
https://lists.freedesktop.org/archives/dri-devel/2019-August/233456.html
While the guest memory type is ignored on Intel, it is honored on ARM. This attempt to fix it
https://lists.freedesktop.org/archives/dri-devel/2019-December/248271.html
does not seem to land.
I applied this patch on top of 5.5.4, but it does not fix the problem. Maybe more similar changes are required?
The patch looks legit. Maybe the memory type is not the root cause?
On Fri, Feb 21, 2020 at 10:41 AM Chia-I Wu olvaffe@gmail.com wrote:
On Fri, Feb 21, 2020 at 2:06 AM Guillaume Gardet Guillaume.Gardet@arm.com wrote:
Hi,
-----Original Message----- From: Chia-I Wu olvaffe@gmail.com Sent: 20 February 2020 19:41 To: Guillaume Gardet Guillaume.Gardet@arm.com Cc: dri-devel@lists.freedesktop.org; Gerd Hoffmann kraxel@redhat.com; Daniel Vetter daniel.vetter@ffwll.ch; Catalin Marinas Catalin.Marinas@arm.com; nd nd@arm.com Subject: Re: [Bug] virtio-gpu broken with qemu/kvm on arm64 on kernel 5.5+
On Thu, Feb 20, 2020 at 4:44 AM Guillaume Gardet Guillaume.Gardet@arm.com wrote:
Hi,
With (guest) kernel 5.5+ with qemu/kvm on arm64, I get lots of display
corruptions leading to this kind of screen:
Looking at the screenshot, it seems cacheline-related?
It could be.
There was a change of memory type
https://lists.freedesktop.org/archives/dri-devel/2019-August/233456.html
While the guest memory type is ignored on Intel, it is honored on ARM. This attempt to fix it
https://lists.freedesktop.org/archives/dri-devel/2019-December/248271.html
does not seem to land.
I applied this patch on top of 5.5.4, but it does not fix the problem. Maybe more similar changes are required?
The patch looks legit. Maybe the memory type is not the root cause?
Perhaps try the entire series?
-----Original Message----- From: Gurchetan Singh gurchetansingh@chromium.org Sent: 22 February 2020 01:47 To: Guillaume Gardet Guillaume.Gardet@arm.com Cc: dri-devel@lists.freedesktop.org; Gerd Hoffmann kraxel@redhat.com; Daniel Vetter daniel.vetter@ffwll.ch; Catalin Marinas Catalin.Marinas@arm.com; nd nd@arm.com; Chia-I Wu olvaffe@gmail.com Subject: Re: [Bug] virtio-gpu broken with qemu/kvm on arm64 on kernel 5.5+
On Fri, Feb 21, 2020 at 10:41 AM Chia-I Wu olvaffe@gmail.com wrote:
On Fri, Feb 21, 2020 at 2:06 AM Guillaume Gardet Guillaume.Gardet@arm.com wrote:
Hi,
-----Original Message----- From: Chia-I Wu olvaffe@gmail.com Sent: 20 February 2020 19:41 To: Guillaume Gardet Guillaume.Gardet@arm.com Cc: dri-devel@lists.freedesktop.org; Gerd Hoffmann kraxel@redhat.com; Daniel Vetter daniel.vetter@ffwll.ch; Catalin Marinas Catalin.Marinas@arm.com; nd nd@arm.com Subject: Re: [Bug] virtio-gpu broken with qemu/kvm on arm64 on kernel 5.5+
On Thu, Feb 20, 2020 at 4:44 AM Guillaume Gardet Guillaume.Gardet@arm.com wrote:
Hi,
With (guest) kernel 5.5+ with qemu/kvm on arm64, I get lots of display
corruptions leading to this kind of screen:
Looking at the screenshot, it seems cacheline-related?
It could be.
There was a change of memory type
https://lists.freedesktop.org/archives/dri-devel/2019-August/23345 6.html
While the guest memory type is ignored on Intel, it is honored on ARM. This attempt to fix it
https://lists.freedesktop.org/archives/dri-devel/2019-December/248 271.html
does not seem to land.
I applied this patch on top of 5.5.4, but it does not fix the problem. Maybe more similar changes are required?
The patch looks legit. Maybe the memory type is not the root cause?
Perhaps try the entire series?
Applied entire series on top of 5.5.6, but still the same problem.
Guillaume
Hi,
Perhaps try the entire series?
Latest version is at: https://git.kraxel.org/cgit/linux/log/?h=drm-virtio-no-wc
Applied entire series on top of 5.5.6, but still the same problem.
Can you double-check? Cherry-picking the shmem + virtio patches ...
694c824491f1 drm/shmem: add support for per object caching flags. 3f4d10f853fb drm/virtio: fix mmap page attributes
... into 5.6-rc3 fixes the corruption for me.
cheers, Gerd
Hi,
-----Original Message----- From: Gerd Hoffmann kraxel@redhat.com Sent: 26 February 2020 07:46 To: Guillaume Gardet Guillaume.Gardet@arm.com Cc: Gurchetan Singh gurchetansingh@chromium.org; dri- devel@lists.freedesktop.org; Daniel Vetter daniel.vetter@ffwll.ch; Catalin Marinas Catalin.Marinas@arm.com; nd nd@arm.com; Chia-I Wu olvaffe@gmail.com Subject: Re: [Bug] virtio-gpu broken with qemu/kvm on arm64 on kernel 5.5+
Hi,
Perhaps try the entire series?
Latest version is at: https://git.kraxel.org/cgit/linux/log/?h=drm-virtio-no-wc
Applied entire series on top of 5.5.6, but still the same problem.
Can you double-check? Cherry-picking the shmem + virtio patches ...
694c824491f1 drm/shmem: add support for per object caching flags. 3f4d10f853fb drm/virtio: fix mmap page attributes
... into 5.6-rc3 fixes the corruption for me.
I tried those 2 patches on top of kernel 5.6-rc3 and they indeed fix the problem.
I applied them on top of 5.5.6 and they also fix the problem! Could we get those 2 patches applied to stable 5.5, please?
Thanks, Guillaume
cheers, Gerd
Hi,
... into 5.6-rc3 fixes the corruption for me.
I tried those 2 patches on top of kernel 5.6-rc3 and they indeed fix the problem.
I applied them on top of 5.5.6 and they also fix the problem! Could we get those 2 patches applied to stable 5.5, please?
Series just re-posted. Reviews welcome. With some luck this makes 5.6-rc4, 5.6-rc5 otherwise, and should hit stable shortly thereafter,
cheers, Gerd
dri-devel@lists.freedesktop.org
-
Chia-I Wu -
Gerd Hoffmann -
Guillaume Gardet -
Gurchetan Singh