[PATCH] drm/amdgpu/sriov: reject kms open if TDR not finished or failed

[Deng, Emily]

Sorry, I don't get your point, when the GPU is doing a reset, even let the dkms open, the userspace will
still fail, so during the reset, any driver open kms is meaningless.

Best Wishes,
Emily Deng

> -----Original Message-----
> From: Michel Dänzer [mailto:michel at daenzer.net]
> Sent: Thursday, April 26, 2018 5:01 PM
> To: Deng, Emily <Emily.Deng at amd.com>; Liu, Monk <Monk.Liu at amd.com>
> Cc: amd-gfx at lists.freedesktop.org
> Subject: Re: [PATCH] drm/amdgpu/sriov: reject kms open if TDR not finished
> or failed
> 
> On 2018-04-26 09:57 AM, Deng, Emily wrote:
> >>>> This seems like a bad idea, as it would cause basically any
> >>>> userspace which wants to use the GPU to fail to start during a GPU
> >>>> reset.
> >>> [Emily] Yes, this is what the change want to do, when driver is
> >>> doing gpu recover or hardware is doing reset, it doesn't want to be
> >>> interrupted, and during the reset any driver open kms is
> >>> meaningless.
> >>
> >> Applications randomly failing to start up during a GPU reset would be
> >> surprising and confusing for the user. The driver needs to handle
> >> this transparently.
> > [Emily] Yes, you are right, how about to sleep for some time in here
> > to wait the GPU reset successfully. After sleep, if it still in gpu
> > reset, then return error to APP?
> 
> I think the only case where returning an error here *might* be appropriate is
> when the driver has tried resetting the GPU, but it has definitely and
> irrecoverably failed. But even in that case, it might be better to let the open
> succeed, and let userspace figure out what happened by other means.
> 
> 
> --
> Earthling Michel Dänzer               |               http://www.amd.com
> Libre software enthusiast             |             Mesa and X developer