All of lore.kernel.org
 help / color / mirror / Atom feed
* [PATCH] drm/amdgpu/gfx11: don't enable gfx on pipe1 with tmz
@ 2026-11-19 18:56 Alex Deucher
  2026-11-19 19:10 ` sashiko-bot
  0 siblings, 1 reply; 2+ messages in thread
From: Alex Deucher @ 2026-11-19 18:56 UTC (permalink / raw)
  To: amd-gfx; +Cc: Alex Deucher

FW doesn't support tmz on pipe one.  Only use pipe 0 if
tmz is enabled.

Fixes: fec62b5e1425 ("drm/amdgpu/gfx11: enable gfx pipe1 hardware support")
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
---
 drivers/gpu/drm/amd/amdgpu/gfx_v11_0.c | 11 ++++++++---
 1 file changed, 8 insertions(+), 3 deletions(-)

diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v11_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v11_0.c
index 6283adec811a3..69ec1d13c4a88 100644
--- a/drivers/gpu/drm/amd/amdgpu/gfx_v11_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/gfx_v11_0.c
@@ -5334,10 +5334,15 @@ static int gfx_v11_0_early_init(struct amdgpu_ip_block *ip_block)
 		adev->gfx.num_compute_rings = 0;
 	} else {
 		/* rs64 only supports one gfx pipe */
-		if (adev->gfx.rs64_enable)
+		if (adev->gfx.rs64_enable) {
 			adev->gfx.num_gfx_rings = 1;
-		else
-			adev->gfx.num_gfx_rings = GFX11_NUM_GFX_RINGS;
+		} else {
+			/* FW doesn't support TMZ on gfx pipe1 */
+			if (adev->gmc.tmz_enabled)
+				adev->gfx.num_gfx_rings = 1;
+			else
+				adev->gfx.num_gfx_rings = GFX11_NUM_GFX_RINGS;
+		}
 		adev->gfx.num_compute_rings = min(amdgpu_gfx_get_num_kcq(adev),
 						  AMDGPU_MAX_COMPUTE_RINGS);
 	}
-- 
2.56.1


^ permalink raw reply related	[flat|nested] 2+ messages in thread

* Re: [PATCH] drm/amdgpu/gfx11: don't enable gfx on pipe1 with tmz
  2026-11-19 18:56 [PATCH] drm/amdgpu/gfx11: don't enable gfx on pipe1 with tmz Alex Deucher
@ 2026-11-19 19:10 ` sashiko-bot
  0 siblings, 0 replies; 2+ messages in thread
From: sashiko-bot @ 2026-11-19 19:10 UTC (permalink / raw)
  To: Alex Deucher, amd-gfx

Thank you for your contribution! Sashiko AI review found 2 potential
issues to consider:
- [High] drm/amdgpu/gfx11: gfx_v11_0_early_init() reads tmz_enabled
  before init
- [High] drm/amdgpu/gfx11: gfx_v11_0_sw_init() enables pipe1 with TMZ
  and crashes
--

commit 3a6ea4e2362ddd6d94fcfcb16809c1b42fa126ef
Author: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu/gfx11: don't enable gfx on pipe1 with tmz

This commit limits adev->gfx.num_gfx_rings to 1 in gfx_v11_0_early_init()
when Trusted Memory Zone (TMZ) is enabled, because GFX11 firmware does not
support TMZ on GFX pipe 1.

> diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v11_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v11_0.c
> index 5d068a9d9c72c..16c8dff858c5d 100644
> --- a/drivers/gpu/drm/amd/amdgpu/gfx_v11_0.c
> +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v11_0.c
> @@ -5302,10 +5302,15 @@ static int gfx_v11_0_early_init(struct amdgpu_ip_block *ip_block)
>  		adev->gfx.num_compute_rings = 0;
>  	} else {
>  		/* rs64 only supports one gfx pipe */
> -		if (adev->gfx.rs64_enable)
> +		if (adev->gfx.rs64_enable) {
>  			adev->gfx.num_gfx_rings = 1;
> -		else
> -			adev->gfx.num_gfx_rings = GFX11_NUM_GFX_RINGS;
> +		} else {
> +			/* FW doesn't support TMZ on gfx pipe1 */
> +			if (adev->gmc.tmz_enabled)
> +				adev->gfx.num_gfx_rings = 1;
> +			else
> +				adev->gfx.num_gfx_rings = GFX11_NUM_GFX_RINGS;
> +		}
>  		adev->gfx.num_compute_rings = min(amdgpu_gfx_get_num_kcq(adev),
>  						  AMDGPU_MAX_COMPUTE_RINGS);
>  	}

[Severity: High]
Is adev->gmc.tmz_enabled already initialized when gfx_v11_0_early_init()
runs?

In amdgpu_device_init(), amdgpu_device_ip_early_init() runs before
amdgpu_gmc_tmz_set():

drivers/gpu/drm/amd/amdgpu/amdgpu_device.c:amdgpu_device_init() {
	...
	/* early init functions */
	r = amdgpu_device_ip_early_init(adev);
	if (r)
		return r;
	...
	/* Enable TMZ based on IP_VERSION */
	amdgpu_gmc_tmz_set(adev);
	...
}

Because adev->gmc.tmz_enabled is still false during early_init,
adev->gfx.num_gfx_rings is always set to GFX11_NUM_GFX_RINGS (2) and GFX
pipe 1 remains enabled when TMZ is active.

[Severity: High]
Should gfx_v11_0_sw_init() also check adev->gmc.tmz_enabled when setting
adev->gfx.me.num_pipe_per_me?

Commit fec62b5e1425 ("drm/amdgpu/gfx11: enable gfx pipe1 hardware
support") enabled pipe 1 in both gfx_v11_0_early_init() and
gfx_v11_0_sw_init():

drivers/gpu/drm/amd/amdgpu/gfx_v11_0.c:gfx_v11_0_sw_init() {
	...
		if (adev->gfx.rs64_enable)
			adev->gfx.me.num_pipe_per_me = 1;
		else
			adev->gfx.me.num_pipe_per_me = 2;
	...
}

When num_gfx_rings is 1 in gfx_v11_0_early_init(),
gfx_v11_0_set_ring_funcs() only initializes adev->gfx.gfx_ring[0].funcs
and leaves gfx_ring[1].funcs NULL.  If gfx_v11_0_sw_init() still sets
num_pipe_per_me to 2, gfx_v11_0_me_init() ->
amdgpu_gfx_graphics_queue_acquire() acquires a queue on pipe 1 and resets
num_gfx_rings to 2, and the ring initialization loop in
gfx_v11_0_sw_init() then calls gfx_v11_0_gfx_ring_init() ->
amdgpu_ring_init() on gfx_ring[1], causing a NULL pointer dereference on
ring->funcs->type.

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20261119185657.4071721-1-alexander.deucher@amd.com?part=1

^ permalink raw reply	[flat|nested] 2+ messages in thread

end of thread, other threads:[~2026-11-19 19:10 UTC | newest]

Thread overview: 2+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2026-11-19 18:56 [PATCH] drm/amdgpu/gfx11: don't enable gfx on pipe1 with tmz Alex Deucher
2026-11-19 19:10 ` sashiko-bot

This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.