[PATCH] drm/amdgpu: Increase reset counter only on success

[Alex Deucher]

On Tue, Jul 15, 2025 at 5:49 AM Lazar, Lijo <lijo.lazar at amd.com> wrote:
>
> <ping>
>
> On 7/14/2025 10:52 AM, Lijo Lazar wrote:
> > Increment the reset counter only if soft recovery succeeded. This is
> > consistent with a ring hard reset behaviour where counter gets
> > incremented only if hard reset succeeded.
> >
> > Signed-off-by: Lijo Lazar <lijo.lazar at amd.com>

Reviewed-by: Alex Deucher <alexander.deucher at amd.com>

> > ---
> >  drivers/gpu/drm/amd/amdgpu/amdgpu_ring.c | 9 +++++++--
> >  1 file changed, 7 insertions(+), 2 deletions(-)
> >
> > diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.c
> > index 426834806fbf..6ac0ce361a2d 100644
> > --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.c
> > +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.c
> > @@ -427,6 +427,7 @@ bool amdgpu_ring_soft_recovery(struct amdgpu_ring *ring, unsigned int vmid,
> >  {
> >       unsigned long flags;
> >       ktime_t deadline;
> > +     bool ret;
> >
> >       if (unlikely(ring->adev->debug_disable_soft_recovery))
> >               return false;
> > @@ -441,12 +442,16 @@ bool amdgpu_ring_soft_recovery(struct amdgpu_ring *ring, unsigned int vmid,
> >               dma_fence_set_error(fence, -ENODATA);
> >       spin_unlock_irqrestore(fence->lock, flags);
> >
> > -     atomic_inc(&ring->adev->gpu_reset_counter);
> >       while (!dma_fence_is_signaled(fence) &&
> >              ktime_to_ns(ktime_sub(deadline, ktime_get())) > 0)
> >               ring->funcs->soft_recovery(ring, vmid);
> >
> > -     return dma_fence_is_signaled(fence);
> > +     ret = dma_fence_is_signaled(fence);
> > +     /* increment the counter only if soft reset worked */
> > +     if (ret)
> > +             atomic_inc(&ring->adev->gpu_reset_counter);
> > +
> > +     return ret;
> >  }
> >
> >  /*
>