[PATCH 1/2] drm/amdkfd: Fix DQM asserts on Hawaii

[PATCH 1/2] drm/amdkfd: Fix DQM asserts on Hawaii

[Felix Kuehling]

start_nocpsch would never set dqm->sched_running on Hawaii due to an
early return statement. This would trigger asserts in other functions
and end up in inconsistent states.

Bug: https://github.com/RadeonOpenCompute/ROCm/issues/1624
Signed-off-by: Felix Kuehling <Felix.Kuehling@amd.com>
---
 drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c | 9 ++++++---
 1 file changed, 6 insertions(+), 3 deletions(-)

diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c b/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c
index dd0b952f0173..104b70e61ba0 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c
@@ -1004,14 +1004,17 @@ static void uninitialize(struct device_queue_manager *dqm)
 
 static int start_nocpsch(struct device_queue_manager *dqm)
 {
+	int r = 0;
+
 	pr_info("SW scheduler is used");
 	init_interrupts(dqm);
 	
 	if (dqm->dev->adev->asic_type == CHIP_HAWAII)
-		return pm_init(&dqm->packet_mgr, dqm);
-	dqm->sched_running = true;
+		r = pm_init(&dqm->packet_mgr, dqm);
+	if (!r)
+		dqm->sched_running = true;
 
-	return 0;
+	return r;
 }
 
 static int stop_nocpsch(struct device_queue_manager *dqm)
-- 
2.32.0

[PATCH 2/2] drm/amdkfd: Make KFD support on Hawaii experimental

[Felix Kuehling]

Hawaii support is mostly untested these days. ROCm user mode also
depends on custom firmware for AQL packet processing, that was never
pushed upstream due to quality regressions in graphics driver testing.

Signed-off-by: Felix Kuehling <Felix.Kuehling@amd.com>
---
 drivers/gpu/drm/amd/amdkfd/kfd_device.c | 6 +++++-
 1 file changed, 5 insertions(+), 1 deletion(-)

diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_device.c b/drivers/gpu/drm/amd/amdkfd/kfd_device.c
index 267668b96456..facc28f58c1f 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_device.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_device.c
@@ -147,7 +147,11 @@ struct kfd_dev *kgd2kfd_probe(struct amdgpu_device *adev, bool vf)
 #ifdef CONFIG_DRM_AMDGPU_CIK
 	case CHIP_HAWAII:
 		gfx_target_version = 70001;
-		if (!vf)
+		if (!amdgpu_exp_hw_support)
+			pr_info(
+	"KFD support on Hawaii is experimental. See modparam exp_hw_support\n"
+				);
+		else if (!vf)
 			f2g = &gfx_v7_kfd2kgd;
 		break;
 #endif
-- 
2.32.0

RE: [PATCH 2/2] drm/amdkfd: Make KFD support on Hawaii experimental

[Russell, Kent]

[AMD Official Use Only]

Reviewed-by: Kent Russell <kent.russell@amd.com>



> -----Original Message-----
> From: amd-gfx <amd-gfx-bounces@lists.freedesktop.org> On Behalf Of Felix Kuehling
> Sent: Wednesday, December 8, 2021 3:26 AM
> To: amd-gfx@lists.freedesktop.org
> Subject: [PATCH 2/2] drm/amdkfd: Make KFD support on Hawaii experimental
>
> Hawaii support is mostly untested these days. ROCm user mode also
> depends on custom firmware for AQL packet processing, that was never
> pushed upstream due to quality regressions in graphics driver testing.
>
> Signed-off-by: Felix Kuehling <Felix.Kuehling@amd.com>
> ---
>  drivers/gpu/drm/amd/amdkfd/kfd_device.c | 6 +++++-
>  1 file changed, 5 insertions(+), 1 deletion(-)
>
> diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_device.c
> b/drivers/gpu/drm/amd/amdkfd/kfd_device.c
> index 267668b96456..facc28f58c1f 100644
> --- a/drivers/gpu/drm/amd/amdkfd/kfd_device.c
> +++ b/drivers/gpu/drm/amd/amdkfd/kfd_device.c
> @@ -147,7 +147,11 @@ struct kfd_dev *kgd2kfd_probe(struct amdgpu_device *adev, bool
> vf)
>  #ifdef CONFIG_DRM_AMDGPU_CIK
>       case CHIP_HAWAII:
>               gfx_target_version = 70001;
> -             if (!vf)
> +             if (!amdgpu_exp_hw_support)
> +                     pr_info(
> +     "KFD support on Hawaii is experimental. See modparam exp_hw_support\n"
> +                             );
> +             else if (!vf)
>                       f2g = &gfx_v7_kfd2kgd;
>               break;
>  #endif
> --
> 2.32.0

Re: [PATCH 1/2] drm/amdkfd: Fix DQM asserts on Hawaii

[Felix Kuehling]

Ping.

On 2021-12-08 3:25 a.m., Felix Kuehling wrote:

> start_nocpsch would never set dqm->sched_running on Hawaii due to an
> early return statement. This would trigger asserts in other functions
> and end up in inconsistent states.
>
> Bug: https://github.com/RadeonOpenCompute/ROCm/issues/1624
> Signed-off-by: Felix Kuehling <Felix.Kuehling@amd.com>
> ---
>   drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c | 9 ++++++---
>   1 file changed, 6 insertions(+), 3 deletions(-)
>
> diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c b/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c
> index dd0b952f0173..104b70e61ba0 100644
> --- a/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c
> +++ b/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c
> @@ -1004,14 +1004,17 @@ static void uninitialize(struct device_queue_manager *dqm)
>   
>   static int start_nocpsch(struct device_queue_manager *dqm)
>   {
> +	int r = 0;
> +
>   	pr_info("SW scheduler is used");
>   	init_interrupts(dqm);
>   	
>   	if (dqm->dev->adev->asic_type == CHIP_HAWAII)
> -		return pm_init(&dqm->packet_mgr, dqm);
> -	dqm->sched_running = true;
> +		r = pm_init(&dqm->packet_mgr, dqm);
> +	if (!r)
> +		dqm->sched_running = true;
>   
> -	return 0;
> +	return r;
>   }
>   
>   static int stop_nocpsch(struct device_queue_manager *dqm)

RE: [PATCH 1/2] drm/amdkfd: Fix DQM asserts on Hawaii

[Russell, Kent]

[AMD Official Use Only]

Reviewed-by: Kent Russell <kent.russell@amd.com>


> -----Original Message-----
> From: amd-gfx <amd-gfx-bounces@lists.freedesktop.org> On Behalf Of Felix Kuehling
> Sent: Monday, January 10, 2022 7:13 PM
> To: amd-gfx@lists.freedesktop.org
> Subject: Re: [PATCH 1/2] drm/amdkfd: Fix DQM asserts on Hawaii
>
> Ping.
>
> On 2021-12-08 3:25 a.m., Felix Kuehling wrote:
>
> > start_nocpsch would never set dqm->sched_running on Hawaii due to an
> > early return statement. This would trigger asserts in other functions
> > and end up in inconsistent states.
> >
> > Bug:
> https://nam11.safelinks.protection.outlook.com/?url=https%3A%2F%2Fgithub.com%2FRad
> eonOpenCompute%2FROCm%2Fissues%2F1624&amp;data=04%7C01%7Ckent.russell%40a
> md.com%7C44c423a1e21b4676d29c08d9d4972868%7C3dd8961fe4884e608e11a82d994e18
> 3d%7C0%7C0%7C637774567959648449%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjA
> wMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C3000&amp;sdata=IYVH4ZU
> UOL1cVzCLZfvoFkRO5%2FKlHsSd6H8RRUP73Nk%3D&amp;reserved=0
> > Signed-off-by: Felix Kuehling <Felix.Kuehling@amd.com>
> > ---
> >   drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c | 9 ++++++---
> >   1 file changed, 6 insertions(+), 3 deletions(-)
> >
> > diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c
> b/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c
> > index dd0b952f0173..104b70e61ba0 100644
> > --- a/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c
> > +++ b/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c
> > @@ -1004,14 +1004,17 @@ static void uninitialize(struct device_queue_manager *dqm)
> >
> >   static int start_nocpsch(struct device_queue_manager *dqm)
> >   {
> > +   int r = 0;
> > +
> >     pr_info("SW scheduler is used");
> >     init_interrupts(dqm);
> >
> >     if (dqm->dev->adev->asic_type == CHIP_HAWAII)
> > -           return pm_init(&dqm->packet_mgr, dqm);
> > -   dqm->sched_running = true;
> > +           r = pm_init(&dqm->packet_mgr, dqm);
> > +   if (!r)
> > +           dqm->sched_running = true;
> >
> > -   return 0;
> > +   return r;
> >   }
> >
> >   static int stop_nocpsch(struct device_queue_manager *dqm)