[PATCH 1/2] drm/amdgpu/gfx11: adjust KGQ reset sequence

[PATCH 1/2] drm/amdgpu/gfx11: adjust KGQ reset sequence

[Alex Deucher]

Kernel gfx queues do not need to be reinitialized or
remapped after a reset.  This fixes queue reset failures
on APUs.

Fixes: b3e9bfd86658 ("drm/amdgpu/gfx11: add ring reset callbacks")
Closes: https://gitlab.freedesktop.org/drm/amd/-/issues/4789
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
---
 drivers/gpu/drm/amd/amdgpu/gfx_v11_0.c | 12 ------------
 1 file changed, 12 deletions(-)

diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v11_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v11_0.c
index b5a2d09fc3469..489edf517f91c 100644
--- a/drivers/gpu/drm/amd/amdgpu/gfx_v11_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/gfx_v11_0.c
@@ -6854,18 +6854,6 @@ static int gfx_v11_0_reset_kgq(struct amdgpu_ring *ring,
 			return r;
 	}
 
-	r = gfx_v11_0_kgq_init_queue(ring, true);
-	if (r) {
-		dev_err(adev->dev, "failed to init kgq\n");
-		return r;
-	}
-
-	r = amdgpu_mes_map_legacy_queue(adev, ring, 0);
-	if (r) {
-		dev_err(adev->dev, "failed to remap kgq\n");
-		return r;
-	}
-
 	return amdgpu_ring_reset_helper_end(ring, timedout_fence);
 }
 
-- 
2.52.0

[PATCH 2/2] drm/amdgpu/gfx12: adjust KGQ reset sequence

[Alex Deucher]

Kernel gfx queues do not need to be reinitialized or
remapped after a reset.  Align with other gfx11.

Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
---
 drivers/gpu/drm/amd/amdgpu/gfx_v12_0.c | 12 ------------
 1 file changed, 12 deletions(-)

diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v12_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v12_0.c
index d8ab93c82366d..c87f5040d25a1 100644
--- a/drivers/gpu/drm/amd/amdgpu/gfx_v12_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/gfx_v12_0.c
@@ -5318,18 +5318,6 @@ static int gfx_v12_0_reset_kgq(struct amdgpu_ring *ring,
 			return r;
 	}
 
-	r = gfx_v12_0_kgq_init_queue(ring, true);
-	if (r) {
-		dev_err(adev->dev, "failed to init kgq\n");
-		return r;
-	}
-
-	r = amdgpu_mes_map_legacy_queue(adev, ring, 0);
-	if (r) {
-		dev_err(adev->dev, "failed to remap kgq\n");
-		return r;
-	}
-
 	return amdgpu_ring_reset_helper_end(ring, timedout_fence);
 }
 
-- 
2.52.0

RE: [PATCH 1/2] drm/amdgpu/gfx11: adjust KGQ reset sequence

[Zhang, Jesse(Jie)]

[AMD Official Use Only - AMD Internal Distribution Only]

> -----Original Message-----
> From: amd-gfx <amd-gfx-bounces@lists.freedesktop.org> On Behalf Of Alex
> Deucher
> Sent: Thursday, January 29, 2026 12:40 PM
> To: amd-gfx@lists.freedesktop.org
> Cc: Deucher, Alexander <Alexander.Deucher@amd.com>
> Subject: [PATCH 1/2] drm/amdgpu/gfx11: adjust KGQ reset sequence
>
> Kernel gfx queues do not need to be reinitialized or remapped after a reset.  This
> fixes queue reset failures on APUs.
>
> Fixes: b3e9bfd86658 ("drm/amdgpu/gfx11: add ring reset callbacks")
> Closes: https://gitlab.freedesktop.org/drm/amd/-/issues/4789
> Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
> ---
>  drivers/gpu/drm/amd/amdgpu/gfx_v11_0.c | 12 ------------
>  1 file changed, 12 deletions(-)
>
> diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v11_0.c
> b/drivers/gpu/drm/amd/amdgpu/gfx_v11_0.c
> index b5a2d09fc3469..489edf517f91c 100644
> --- a/drivers/gpu/drm/amd/amdgpu/gfx_v11_0.c
> +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v11_0.c
> @@ -6854,18 +6854,6 @@ static int gfx_v11_0_reset_kgq(struct amdgpu_ring
> *ring,
>                       return r;
>       }
>
> -     r = gfx_v11_0_kgq_init_queue(ring, true);
> -     if (r) {
> -             dev_err(adev->dev, "failed to init kgq\n");
> -             return r;
> -     }
> -
> -     r = amdgpu_mes_map_legacy_queue(adev, ring, 0);
> -     if (r) {
> -             dev_err(adev->dev, "failed to remap kgq\n");
> -             return r;
> -     }
> -
>       return amdgpu_ring_reset_helper_end(ring, timedout_fence);  }
>
If it's reset by mmio, we might need to keep it.
So, how about adding a check like this:
static int gfx_v11_0_reset_kgq(struct amdgpu_ring *ring,
                               struct amdgpu_fence *timedout_fence)
 {
        struct amdgpu_device *adev = ring->adev;
+       bool use_mmio = false;
        int r;

        amdgpu_ring_reset_helper_begin(ring, timedout_fence);

-       r = amdgpu_mes_reset_legacy_queue(ring->adev, ring, vmid, false, 0);
+       r = amdgpu_mes_reset_legacy_queue(ring->adev, ring, vmid, user_mmio, 0);
        if (r) {

                dev_warn(adev->dev, "reset via MES failed and try pipe reset %d\n", r);
@@ -6874,16 +6875,18 @@ static int gfx_v11_0_reset_kgq(struct amdgpu_ring *ring,
                        return r;
        }

-       r = gfx_v11_0_kgq_init_queue(ring, true);
-       if (r) {
-               dev_err(adev->dev, "failed to init kgq\n");
-               return r;
-       }
+       if (use_mmio) {
+               r = gfx_v11_0_kgq_init_queue(ring, true);
+               if (r) {
+                       dev_err(adev->dev, "failed to init kgq\n");
+                       return r;
+               }
             ...
        }
> --
> 2.52.0

Re: [PATCH 1/2] drm/amdgpu/gfx11: adjust KGQ reset sequence

[Timur Kristóf]

On Thursday, January 29, 2026 5:39:35 AM Central European Standard Time Alex 
Deucher wrote:
> Kernel gfx queues do not need to be reinitialized or
> remapped after a reset.  This fixes queue reset failures
> on APUs.
> 
> Fixes: b3e9bfd86658 ("drm/amdgpu/gfx11: add ring reset callbacks")
> Closes: https://gitlab.freedesktop.org/drm/amd/-/issues/4789
> Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

The series is:
Reviewed-by: Timur Kristóf <timur.kristof@gmail.com>
(including with Jesse's suggestion if you choose to implement that).

I notice that gfx_v10_0_reset_kgq() also calls gfx_v10_0_kgq_init_queue(), 
should that call be removed as well or made conditional?

> ---
>  drivers/gpu/drm/amd/amdgpu/gfx_v11_0.c | 12 ------------
>  1 file changed, 12 deletions(-)
> 
> diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v11_0.c
> b/drivers/gpu/drm/amd/amdgpu/gfx_v11_0.c index b5a2d09fc3469..489edf517f91c
> 100644
> --- a/drivers/gpu/drm/amd/amdgpu/gfx_v11_0.c
> +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v11_0.c
> @@ -6854,18 +6854,6 @@ static int gfx_v11_0_reset_kgq(struct amdgpu_ring
> *ring, return r;
>  	}
> 
> -	r = gfx_v11_0_kgq_init_queue(ring, true);
> -	if (r) {
> -		dev_err(adev->dev, "failed to init kgq\n");
> -		return r;
> -	}
> -
> -	r = amdgpu_mes_map_legacy_queue(adev, ring, 0);
> -	if (r) {
> -		dev_err(adev->dev, "failed to remap kgq\n");
> -		return r;
> -	}
> -
>  	return amdgpu_ring_reset_helper_end(ring, timedout_fence);
>  }

Re: [PATCH 1/2] drm/amdgpu/gfx11: adjust KGQ reset sequence

[Alex Deucher]

On Thu, Jan 29, 2026 at 4:52 AM Timur Kristóf <timur.kristof@gmail.com> wrote:
>
> On Thursday, January 29, 2026 5:39:35 AM Central European Standard Time Alex
> Deucher wrote:
> > Kernel gfx queues do not need to be reinitialized or
> > remapped after a reset.  This fixes queue reset failures
> > on APUs.
> >
> > Fixes: b3e9bfd86658 ("drm/amdgpu/gfx11: add ring reset callbacks")
> > Closes: https://gitlab.freedesktop.org/drm/amd/-/issues/4789
> > Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
>
> The series is:
> Reviewed-by: Timur Kristóf <timur.kristof@gmail.com>
> (including with Jesse's suggestion if you choose to implement that).
>
> I notice that gfx_v10_0_reset_kgq() also calls gfx_v10_0_kgq_init_queue(),
> should that call be removed as well or made conditional?

In my testing, gfx10 seems to require it, but I plan to look into it
deeper when I get a chance.

Alex

>
> > ---
> >  drivers/gpu/drm/amd/amdgpu/gfx_v11_0.c | 12 ------------
> >  1 file changed, 12 deletions(-)
> >
> > diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v11_0.c
> > b/drivers/gpu/drm/amd/amdgpu/gfx_v11_0.c index b5a2d09fc3469..489edf517f91c
> > 100644
> > --- a/drivers/gpu/drm/amd/amdgpu/gfx_v11_0.c
> > +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v11_0.c
> > @@ -6854,18 +6854,6 @@ static int gfx_v11_0_reset_kgq(struct amdgpu_ring
> > *ring, return r;
> >       }
> >
> > -     r = gfx_v11_0_kgq_init_queue(ring, true);
> > -     if (r) {
> > -             dev_err(adev->dev, "failed to init kgq\n");
> > -             return r;
> > -     }
> > -
> > -     r = amdgpu_mes_map_legacy_queue(adev, ring, 0);
> > -     if (r) {
> > -             dev_err(adev->dev, "failed to remap kgq\n");
> > -             return r;
> > -     }
> > -
> >       return amdgpu_ring_reset_helper_end(ring, timedout_fence);
> >  }
>
>
>
>