[PATCH v2 0/3] drm/panthor: Fix regressions introduced recently

[PATCH v2 0/3] drm/panthor: Fix regressions introduced recently

[Boris Brezillon]

Hello,

This is a set of fixes for regressions noticed while testing the
drm-misc-next branch against mesa-ci.

Regards,

Boris

Boris Brezillon (3):
  drm/panthor: Drop a WARN_ON() in group_free_queue()
  drm/panthor: Make sure caches are flushed/invalidated when an AS is
    recycled
  drm/panthor: Unlock the locked region before disabling an AS

 drivers/gpu/drm/panthor/panthor_mmu.c   | 39 +++++++++++++++++++++----
 drivers/gpu/drm/panthor/panthor_sched.c |  5 ++--
 2 files changed, 35 insertions(+), 9 deletions(-)

-- 
2.51.1

[PATCH v2 1/3] drm/panthor: Drop a WARN_ON() in group_free_queue()

[Boris Brezillon]

It appears the timeout can still be enabled when we reach that point,
because of the asynchronous progress check done on queues that resets
the timer when jobs are still in-flight, but progress was made.
We could add more checks to make sure the timer is not re-enabled when
a group can't run anymore, but we don't have a group to pass to
queue_check_job_completion() in some context.

It's just as safe (we just want to be sure the timer is stopped before
we destroy the queue) and simpler to drop the WARN_ON() in
group_free_queue().

v2:
- Collect R-bs

Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Liviu Dudau <liviu.dudau@arm.com>
Reviewed-by: Chia-I Wu <olvaffe@gmail.com>
---
 drivers/gpu/drm/panthor/panthor_sched.c | 5 ++---
 1 file changed, 2 insertions(+), 3 deletions(-)

diff --git a/drivers/gpu/drm/panthor/panthor_sched.c b/drivers/gpu/drm/panthor/panthor_sched.c
index 389d508b3848..203f6a0a6b9a 100644
--- a/drivers/gpu/drm/panthor/panthor_sched.c
+++ b/drivers/gpu/drm/panthor/panthor_sched.c
@@ -893,9 +893,8 @@ static void group_free_queue(struct panthor_group *group, struct panthor_queue *
 	if (IS_ERR_OR_NULL(queue))
 		return;
 
-	/* This should have been disabled before that point. */
-	drm_WARN_ON(&group->ptdev->base,
-		    disable_delayed_work_sync(&queue->timeout.work));
+	/* Disable the timeout before tearing down drm_sched components. */
+	disable_delayed_work_sync(&queue->timeout.work);
 
 	if (queue->entity.fence_context)
 		drm_sched_entity_destroy(&queue->entity);
-- 
2.51.1

[PATCH v2 2/3] drm/panthor: Make sure caches are flushed/invalidated when an AS is recycled

[Boris Brezillon]

When we re-assign a slot to a different VM, we need to make sure the
old VM caches are flushed before doing the switch. Specialize
panthor_mmu_as_disable() so we can skip the slot programmation while
still getting the cache flushing, and call this helper from
panthor_vm_active() when an idle slot is recycled.

v2:
- Collect R-bs

Fixes: 6e2d3b3e8589 ("drm/panthor: Add support for atomic page table updates")
Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Liviu Dudau <liviu.dudau@arm.com>
Reviewed-by: Chia-I Wu <olvaffe@gmail.com>
---
 drivers/gpu/drm/panthor/panthor_mmu.c | 26 ++++++++++++++++++++------
 1 file changed, 20 insertions(+), 6 deletions(-)

diff --git a/drivers/gpu/drm/panthor/panthor_mmu.c b/drivers/gpu/drm/panthor/panthor_mmu.c
index 8ba5259e3d28..3644af1a8e56 100644
--- a/drivers/gpu/drm/panthor/panthor_mmu.c
+++ b/drivers/gpu/drm/panthor/panthor_mmu.c
@@ -585,7 +585,8 @@ static int panthor_mmu_as_enable(struct panthor_device *ptdev, u32 as_nr,
 	return as_send_cmd_and_wait(ptdev, as_nr, AS_COMMAND_UPDATE);
 }
 
-static int panthor_mmu_as_disable(struct panthor_device *ptdev, u32 as_nr)
+static int panthor_mmu_as_disable(struct panthor_device *ptdev, u32 as_nr,
+				  bool recycle_slot)
 {
 	int ret;
 
@@ -595,6 +596,12 @@ static int panthor_mmu_as_disable(struct panthor_device *ptdev, u32 as_nr)
 	if (ret)
 		return ret;
 
+	/* If the slot is going to be used immediately, don't bother changing
+	 * the config.
+	 */
+	if (recycle_slot)
+		return 0;
+
 	gpu_write64(ptdev, AS_TRANSTAB(as_nr), 0);
 	gpu_write64(ptdev, AS_MEMATTR(as_nr), 0);
 	gpu_write64(ptdev, AS_TRANSCFG(as_nr), AS_TRANSCFG_ADRMODE_UNMAPPED);
@@ -714,6 +721,11 @@ int panthor_vm_active(struct panthor_vm *vm)
 
 		drm_WARN_ON(&ptdev->base, refcount_read(&lru_vm->as.active_cnt));
 		as = lru_vm->as.id;
+
+		ret = panthor_mmu_as_disable(ptdev, as, true);
+		if (ret)
+			goto out_unlock;
+
 		panthor_vm_release_as_locked(lru_vm);
 	}
 
@@ -853,7 +865,7 @@ static void panthor_vm_declare_unusable(struct panthor_vm *vm)
 	vm->unusable = true;
 	mutex_lock(&ptdev->mmu->as.slots_lock);
 	if (vm->as.id >= 0 && drm_dev_enter(&ptdev->base, &cookie)) {
-		panthor_mmu_as_disable(ptdev, vm->as.id);
+		panthor_mmu_as_disable(ptdev, vm->as.id, false);
 		drm_dev_exit(cookie);
 	}
 	mutex_unlock(&ptdev->mmu->as.slots_lock);
@@ -1780,7 +1792,7 @@ static void panthor_mmu_irq_handler(struct panthor_device *ptdev, u32 status)
 			ptdev->mmu->as.slots[as].vm->unhandled_fault = true;
 
 		/* Disable the MMU to kill jobs on this AS. */
-		panthor_mmu_as_disable(ptdev, as);
+		panthor_mmu_as_disable(ptdev, as, false);
 		mutex_unlock(&ptdev->mmu->as.slots_lock);
 
 		status &= ~mask;
@@ -1809,7 +1821,8 @@ void panthor_mmu_suspend(struct panthor_device *ptdev)
 		struct panthor_vm *vm = ptdev->mmu->as.slots[i].vm;
 
 		if (vm) {
-			drm_WARN_ON(&ptdev->base, panthor_mmu_as_disable(ptdev, i));
+			drm_WARN_ON(&ptdev->base,
+				    panthor_mmu_as_disable(ptdev, i, false));
 			panthor_vm_release_as_locked(vm);
 		}
 	}
@@ -1930,7 +1943,7 @@ static void panthor_vm_free(struct drm_gpuvm *gpuvm)
 		int cookie;
 
 		if (drm_dev_enter(&ptdev->base, &cookie)) {
-			panthor_mmu_as_disable(ptdev, vm->as.id);
+			panthor_mmu_as_disable(ptdev, vm->as.id, false);
 			drm_dev_exit(cookie);
 		}
 
@@ -2790,7 +2803,8 @@ void panthor_mmu_unplug(struct panthor_device *ptdev)
 		struct panthor_vm *vm = ptdev->mmu->as.slots[i].vm;
 
 		if (vm) {
-			drm_WARN_ON(&ptdev->base, panthor_mmu_as_disable(ptdev, i));
+			drm_WARN_ON(&ptdev->base,
+				    panthor_mmu_as_disable(ptdev, i, false));
 			panthor_vm_release_as_locked(vm);
 		}
 	}
-- 
2.51.1

[PATCH v2 3/3] drm/panthor: Unlock the locked region before disabling an AS

[Boris Brezillon]

An AS can be disabled in the middle of a VM operation (VM being
evicted from an AS slot, for instance). In that case, we need the
locked section to be unlocked before releasing the slot.

v2:
- Add an lockdep_assert_held() in panthor_mmu_as_disable()
- Collect R-bs

Fixes: 6e2d3b3e8589 ("drm/panthor: Add support for atomic page table updates")
Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Liviu Dudau <liviu.dudau@arm.com>
Reviewed-by: Chia-I Wu <olvaffe@gmail.com>
---
 drivers/gpu/drm/panthor/panthor_mmu.c | 13 +++++++++++++
 1 file changed, 13 insertions(+)

diff --git a/drivers/gpu/drm/panthor/panthor_mmu.c b/drivers/gpu/drm/panthor/panthor_mmu.c
index 3644af1a8e56..493e6428d478 100644
--- a/drivers/gpu/drm/panthor/panthor_mmu.c
+++ b/drivers/gpu/drm/panthor/panthor_mmu.c
@@ -588,14 +588,27 @@ static int panthor_mmu_as_enable(struct panthor_device *ptdev, u32 as_nr,
 static int panthor_mmu_as_disable(struct panthor_device *ptdev, u32 as_nr,
 				  bool recycle_slot)
 {
+	struct panthor_vm *vm = ptdev->mmu->as.slots[as_nr].vm;
 	int ret;
 
+	lockdep_assert_held(&ptdev->mmu->as.slots_lock);
+
 	/* Flush+invalidate RW caches, invalidate RO ones. */
 	ret = panthor_gpu_flush_caches(ptdev, CACHE_CLEAN | CACHE_INV,
 				       CACHE_CLEAN | CACHE_INV, CACHE_INV);
 	if (ret)
 		return ret;
 
+	if (vm && vm->locked_region.size) {
+		/* Unlock the region if there a lock pending. */
+		ret = as_send_cmd_and_wait(ptdev, vm->as.id, AS_COMMAND_UNLOCK);
+		if (ret)
+			return ret;
+
+		vm->locked_region.start = 0;
+		vm->locked_region.size = 0;
+	}
+
 	/* If the slot is going to be used immediately, don't bother changing
 	 * the config.
 	 */
-- 
2.51.1

Re: [PATCH v2 3/3] drm/panthor: Unlock the locked region before disabling an AS

[Boris Brezillon]

On Wed,  3 Dec 2025 12:05:27 +0100
Boris Brezillon <boris.brezillon@collabora.com> wrote:

> An AS can be disabled in the middle of a VM operation (VM being
> evicted from an AS slot, for instance). In that case, we need the
> locked section to be unlocked before releasing the slot.
> 
> v2:
> - Add an lockdep_assert_held() in panthor_mmu_as_disable()
> - Collect R-bs
> 
> Fixes: 6e2d3b3e8589 ("drm/panthor: Add support for atomic page table updates")
> Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
> Reviewed-by: Liviu Dudau <liviu.dudau@arm.com>
> Reviewed-by: Chia-I Wu <olvaffe@gmail.com>
> ---
>  drivers/gpu/drm/panthor/panthor_mmu.c | 13 +++++++++++++
>  1 file changed, 13 insertions(+)
> 
> diff --git a/drivers/gpu/drm/panthor/panthor_mmu.c b/drivers/gpu/drm/panthor/panthor_mmu.c
> index 3644af1a8e56..493e6428d478 100644
> --- a/drivers/gpu/drm/panthor/panthor_mmu.c
> +++ b/drivers/gpu/drm/panthor/panthor_mmu.c
> @@ -588,14 +588,27 @@ static int panthor_mmu_as_enable(struct panthor_device *ptdev, u32 as_nr,
>  static int panthor_mmu_as_disable(struct panthor_device *ptdev, u32 as_nr,
>  				  bool recycle_slot)
>  {
> +	struct panthor_vm *vm = ptdev->mmu->as.slots[as_nr].vm;
>  	int ret;
>  
> +	lockdep_assert_held(&ptdev->mmu->as.slots_lock);
> +
>  	/* Flush+invalidate RW caches, invalidate RO ones. */
>  	ret = panthor_gpu_flush_caches(ptdev, CACHE_CLEAN | CACHE_INV,
>  				       CACHE_CLEAN | CACHE_INV, CACHE_INV);
>  	if (ret)
>  		return ret;
>  
> +	if (vm && vm->locked_region.size) {
> +		/* Unlock the region if there a lock pending. */
> +		ret = as_send_cmd_and_wait(ptdev, vm->as.id, AS_COMMAND_UNLOCK);
> +		if (ret)
> +			return ret;
> +
> +		vm->locked_region.start = 0;
> +		vm->locked_region.size = 0;

Actually, resetting the locked_region range triggers the WARN_ON()s in
the vm_[un]map_pages() path, so I'll get rid of that in v3.

> +	}
> +
>  	/* If the slot is going to be used immediately, don't bother changing
>  	 * the config.
>  	 */