AMD General
Hi Vitaly,
We should only return the cached VBLANK counter in task context for
get_vblank_counter(), based on two key considerations:
1. It ensures the reset rwsem is never acquired or released from hardirq
or softirq contexts.
2. It allows the VBLANK IRQ handler to continue reading the live hardware
counter even while GPU recovery holds the semaphore. Otherwise, falling back to
the cached counter on a failed trylock would cause DRM's VBLANK count to stall.
Regards,
Prike
> -----Original Message-----
> From: Prosyak, Vitaly <[email protected]>
> Sent: Thursday, November 12, 2026 10:52 AM
> To: Liang, Prike <[email protected]>; [email protected]; dri-
> [email protected]
> Cc: Deucher, Alexander <[email protected]>; Koenig, Christian
> <[email protected]>; Wentland, Harry <[email protected]>; Li,
> Sun peng (Leo) <[email protected]>; [email protected]; Prosyak, Vitaly
> <[email protected]>; Lin, Ping Lei <[email protected]>
> Subject: Re: [PATCH RESEND 5/5] drm/amdgpu: keep reading the vblank counter in
> IRQ context during reset
>
> Hi Prike,
>
> Thanks for the fix. Would !in_hardirq() be better than in_task() here?
>
> The vblank-disable timer also reaches this callback via
> vblank_disable_fn() -> drm_vblank_disable_and_save() ->
> drm_update_vblank_count(). In softirq context, in_task() is false, so the
> proposed
> check also bypasses the reset semaphore for timer reads.
> Using !in_hardirq() would protect that path while preserving the hard-IRQ
> bypass.
>
> What do you think?
>
> Thanks, Vitaly
>
> On 2026-11-11 03:55, Prike Liang wrote:
> > amdgpu_get_vblank_counter_kms() returns the cached vblank->last
> > instead of reading the hardware counter whenever it can't take
> > reset_domain->sem for reading. The vblank interrupt reads the counter
> > through the same callback via drm_crtc_handle_vblank(), so once GPU
> > recovery holds the semaphore for writing, every vblank interrupt sees
> > diff == 0 and the DRM vblank count stops advancing. The CRTCs keep
> > scanning out until dm_suspend() turns vblank off and stops the
> > streams, though, which only happens after the rest of the recovery
> > preparation.
> >
> > Vblank waits in that window don't see a vblank anymore. They only end
> > when dm_suspend() turns vblank off, or after a second with the "vblank
> > wait timed out" warning if that takes longer. The fbdev damage worker
> > is likely to be waiting already, because recovery prints "GPU reset
> > begin!" right before it takes the semaphore.
> >
> > Vblank interrupts only arrive while a CRTC is scanning out, so the
> > interrupt path can read the hardware counter.
> > amdgpu_device_skip_hw_access() likewise only checks the reset
> > semaphore in task context. Only back off in task context, so the
> > interrupt path keeps the count moving until DM turns vblank off.
> >
> > Fixes: f2bc71d16e73 ("drm/amdgpu: serialize vblank counter reads
> > against GPU reset")
> > Cc: Vitaly Prosyak <[email protected]<mailto:[email protected]>>
> > Cc: James Lin <[email protected]<mailto:[email protected]>>
> > Cc: Leo Li <[email protected]<mailto:[email protected]>>
> > Assisted-by: Copilot:claude-opus-5.5
> > Signed-off-by: Prike Liang <[email protected]<mailto:[email protected]>>
> > ---
> > drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c | 25
> > ++++++++++++++++++++++---
> > 1 file changed, 22 insertions(+), 3 deletions(-)
> >
> > diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c
> > b/drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c
> > index 14a72a406d37..06e539c4c3ea 100644
> > --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c
> > +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c
> > @@ -1692,6 +1692,7 @@ u32 amdgpu_get_vblank_counter_kms(struct drm_crtc
> *crtc)
> > struct drm_device *dev = crtc->dev;
> > unsigned int pipe = crtc->index;
> > struct amdgpu_device *adev = drm_to_adev(dev);
> > + bool locked = false;
> > int vpos, hpos, stat;
> > u32 count;
> >
> > @@ -1699,9 +1700,26 @@ u32 amdgpu_get_vblank_counter_kms(struct
> drm_crtc *crtc)
> > DRM_ERROR("Invalid crtc %u\n", pipe);
> > return -EINVAL;
> > }
> > - if (!down_read_trylock(&adev->reset_domain->sem))
> > - return drm_crtc_vblank_count(crtc);
> >
> > + /*
> > + * A GPU reset is in progress and the display hardware cannot be read
> > + * safely. Return the hardware counter value DRM already cached rather
> > + * than the software vblank count: the core feeds this callback's
> > + * result back into vblank->last, and a software count lives in a
> > + * different numeric domain, which later trips
> > + * drm_WARN_ON_ONCE(cur_vblank != vblank->last) in
> > + * drm_update_vblank_count().
> > + *
> > + * Only task context has to back off. Vblank interrupts only arrive
> > + * while a CRTC is scanning out, so the interrupt path can read the
> > + * hardware, and it has to keep counting: vblank waits that already
> > + * sleep when a reset takes the semaphore never complete otherwise.
> > + */
> > + if (in_task()) {
> > + if (!down_read_trylock(&adev->reset_domain->sem))
> > + return drm_crtc_vblank_crtc(crtc)->last;
> > + locked = true;
> > + }
> >
> > /* The hw increments its frame counter at start of vsync, not at start
> > * of vblank, as is required by DRM core vblank counter handling.
> > @@ -1747,7 +1765,8 @@ u32 amdgpu_get_vblank_counter_kms(struct drm_crtc
> *crtc)
> > DRM_DEBUG_VBL("NULL mode info! Returned count may be
> wrong.\n");
> > }
> >
> > - up_read(&adev->reset_domain->sem);
> > + if (locked)
> > + up_read(&adev->reset_domain->sem);
> > return count;
> > }
> >