Marc,
On Wed, Jul 28, 2021 at 07:54:58AM +0200, Salvatore Bonaccorso wrote:
> Hi Marc,
>
> Thanks a lot for your investigation.
>
> On Tue, Jul 27, 2021 at 10:28:24PM +0200, Marc wrote:
> > After bisecting, I get this SHA1 as the first to have fixed the issue
> > (at least, it's not showing as easily as before it). It makes sense as
> > the backtrace shows something in amdgpu and this is a bug fix :)
> >
> > 8<---8<---8<---8<---8<---8<---8<---8<---8<---8<---8<---8<---8<---8<---8<---8<---8<---
> > 89fa15ecdca7eb46a711476b961f70a74765bbe4 is the first nobug commit
> > commit 89fa15ecdca7eb46a711476b961f70a74765bbe4
> > Author: Huang Rui <ray.hu...@amd.com>
> > Date: Sat Jan 30 17:14:30 2021 +0800
> >
> > drm/amdgpu: fix the issue that retry constantly once the buffer is
> > oversize
> >
> > We cannot modify initial_domain every time while the retry starts. That
> > will cause the busy waiting that unable to switch to GTT while the vram
> > is not enough.
> >
> > Fixes: f8aab60422c3 ("drm/amdgpu: Initialise drm_gem_object_funcs for
> > imported BOs")
> >
> > Signed-off-by: Huang Rui <ray.hu...@amd.com>
> > Reviewed-by: Alex Deucher <alexander.deuc...@amd.com>
> > Reviewed-by: Christian König <christian.koe...@amd.com>
> > Signed-off-by: Alex Deucher <alexander.deuc...@amd.com>
> > Cc: sta...@vger.kernel.org
> >
> > drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c | 2 +-
> > 1 file changed, 1 insertion(+), 1 deletion(-)
> > 8<---8<---8<---8<---8<---8<---8<---8<---8<---8<---8<---8<---8<---8<---8<---8<---8<---
> >
> > I also tried to simply cherry-pick it on top of v5.10 tag, and it seems
> > to also fix the issue.
>
> I wonder if this will be really enough. I'm asking since the commit
> you identified, is marked to fix another commit, which was not
> backported into the 5.10.y series.
>
> Now that you identified possible fix for this issue, I would suggest
> to report it to upstream, can you do that and keep the Debian bug into
> the loop?
Sorry should have been more clear here: Not upstream for mainline, as
it is fixed, but for looking forward for inclusion of the bugfix into
the 5.10.y stable series upstream, I guess we need to clarify the
above.
Regards,
Salvatore
