drm/panthor: Fix race with suspend during unplug

[PATCH] drm/panthor: Fix race with suspend during unplug

Posted by Ketil Johnsen 3 months, 2 weeks ago

There is a race between panthor_device_unplug() and
panthor_device_suspend() which can lead to IRQ handlers running on a
powered down GPU. This is how it can happen:
- unplug routine calls drm_dev_unplug()
- panthor_device_suspend() can now execute, and will skip a lot of
  important work because the device is currently marked as unplugged.
- IRQs will remain active in this case and IRQ handlers can therefore
  try to access a powered down GPU.

The fix is simply to take the PM ref in panthor_device_unplug() a
little bit earlier, before drm_dev_unplug().

Signed-off-by: Ketil Johnsen <ketil.johnsen@arm.com>
Fixes: 5fe909cae118a ("drm/panthor: Add the device logical block")
---
 drivers/gpu/drm/panthor/panthor_device.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/drivers/gpu/drm/panthor/panthor_device.c b/drivers/gpu/drm/panthor/panthor_device.c
index 81df49880bd87..962a10e00848e 100644
--- a/drivers/gpu/drm/panthor/panthor_device.c
+++ b/drivers/gpu/drm/panthor/panthor_device.c
@@ -83,6 +83,8 @@ void panthor_device_unplug(struct panthor_device *ptdev)
 		return;
 	}
 
+	drm_WARN_ON(&ptdev->base, pm_runtime_get_sync(ptdev->base.dev) < 0);
+
 	/* Call drm_dev_unplug() so any access to HW blocks happening after
 	 * that point get rejected.
 	 */
@@ -93,8 +95,6 @@ void panthor_device_unplug(struct panthor_device *ptdev)
 	 */
 	mutex_unlock(&ptdev->unplug.lock);
 
-	drm_WARN_ON(&ptdev->base, pm_runtime_get_sync(ptdev->base.dev) < 0);
-
 	/* Now, try to cleanly shutdown the GPU before the device resources
 	 * get reclaimed.
 	 */
-- 
2.47.2

Re: [PATCH] drm/panthor: Fix race with suspend during unplug

Posted by Liviu Dudau 3 months ago

On Wed, Oct 22, 2025 at 12:32:41PM +0200, Ketil Johnsen wrote:
> There is a race between panthor_device_unplug() and
> panthor_device_suspend() which can lead to IRQ handlers running on a
> powered down GPU. This is how it can happen:
> - unplug routine calls drm_dev_unplug()
> - panthor_device_suspend() can now execute, and will skip a lot of
>   important work because the device is currently marked as unplugged.
> - IRQs will remain active in this case and IRQ handlers can therefore
>   try to access a powered down GPU.
> 
> The fix is simply to take the PM ref in panthor_device_unplug() a
> little bit earlier, before drm_dev_unplug().
> 
> Signed-off-by: Ketil Johnsen <ketil.johnsen@arm.com>
> Fixes: 5fe909cae118a ("drm/panthor: Add the device logical block")

Pushed to drm-misc-next.

Best regards,
Liviu

> ---
>  drivers/gpu/drm/panthor/panthor_device.c | 4 ++--
>  1 file changed, 2 insertions(+), 2 deletions(-)
> 
> diff --git a/drivers/gpu/drm/panthor/panthor_device.c b/drivers/gpu/drm/panthor/panthor_device.c
> index 81df49880bd87..962a10e00848e 100644
> --- a/drivers/gpu/drm/panthor/panthor_device.c
> +++ b/drivers/gpu/drm/panthor/panthor_device.c
> @@ -83,6 +83,8 @@ void panthor_device_unplug(struct panthor_device *ptdev)
>  		return;
>  	}
>  
> +	drm_WARN_ON(&ptdev->base, pm_runtime_get_sync(ptdev->base.dev) < 0);
> +
>  	/* Call drm_dev_unplug() so any access to HW blocks happening after
>  	 * that point get rejected.
>  	 */
> @@ -93,8 +95,6 @@ void panthor_device_unplug(struct panthor_device *ptdev)
>  	 */
>  	mutex_unlock(&ptdev->unplug.lock);
>  
> -	drm_WARN_ON(&ptdev->base, pm_runtime_get_sync(ptdev->base.dev) < 0);
> -
>  	/* Now, try to cleanly shutdown the GPU before the device resources
>  	 * get reclaimed.
>  	 */
> -- 
> 2.47.2
> 

-- 
====================
| I would like to |
| fix the world,  |
| but they're not |
| giving me the   |
 \ source code!  /
  ---------------
    ¯\_(ツ)_/¯

Re: [PATCH] drm/panthor: Fix race with suspend during unplug

Posted by Boris Brezillon 3 months, 2 weeks ago

On Wed, 22 Oct 2025 12:32:41 +0200
Ketil Johnsen <ketil.johnsen@arm.com> wrote:

> There is a race between panthor_device_unplug() and
> panthor_device_suspend() which can lead to IRQ handlers running on a
> powered down GPU. This is how it can happen:
> - unplug routine calls drm_dev_unplug()
> - panthor_device_suspend() can now execute, and will skip a lot of
>   important work because the device is currently marked as unplugged.
> - IRQs will remain active in this case and IRQ handlers can therefore
>   try to access a powered down GPU.
> 
> The fix is simply to take the PM ref in panthor_device_unplug() a
> little bit earlier, before drm_dev_unplug().
> 
> Signed-off-by: Ketil Johnsen <ketil.johnsen@arm.com>
> Fixes: 5fe909cae118a ("drm/panthor: Add the device logical block")

Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>

> ---
>  drivers/gpu/drm/panthor/panthor_device.c | 4 ++--
>  1 file changed, 2 insertions(+), 2 deletions(-)
> 
> diff --git a/drivers/gpu/drm/panthor/panthor_device.c b/drivers/gpu/drm/panthor/panthor_device.c
> index 81df49880bd87..962a10e00848e 100644
> --- a/drivers/gpu/drm/panthor/panthor_device.c
> +++ b/drivers/gpu/drm/panthor/panthor_device.c
> @@ -83,6 +83,8 @@ void panthor_device_unplug(struct panthor_device *ptdev)
>  		return;
>  	}
>  
> +	drm_WARN_ON(&ptdev->base, pm_runtime_get_sync(ptdev->base.dev) < 0);
> +
>  	/* Call drm_dev_unplug() so any access to HW blocks happening after
>  	 * that point get rejected.
>  	 */
> @@ -93,8 +95,6 @@ void panthor_device_unplug(struct panthor_device *ptdev)
>  	 */
>  	mutex_unlock(&ptdev->unplug.lock);
>  
> -	drm_WARN_ON(&ptdev->base, pm_runtime_get_sync(ptdev->base.dev) < 0);
> -
>  	/* Now, try to cleanly shutdown the GPU before the device resources
>  	 * get reclaimed.
>  	 */

Re: [PATCH] drm/panthor: Fix race with suspend during unplug

Posted by Liviu Dudau 3 months, 2 weeks ago

On Wed, Oct 22, 2025 at 12:32:41PM +0200, Ketil Johnsen wrote:
> There is a race between panthor_device_unplug() and
> panthor_device_suspend() which can lead to IRQ handlers running on a
> powered down GPU. This is how it can happen:
> - unplug routine calls drm_dev_unplug()
> - panthor_device_suspend() can now execute, and will skip a lot of
>   important work because the device is currently marked as unplugged.
> - IRQs will remain active in this case and IRQ handlers can therefore
>   try to access a powered down GPU.
> 
> The fix is simply to take the PM ref in panthor_device_unplug() a
> little bit earlier, before drm_dev_unplug().
> 
> Signed-off-by: Ketil Johnsen <ketil.johnsen@arm.com>
> Fixes: 5fe909cae118a ("drm/panthor: Add the device logical block")

Reviewed-by: Liviu Dudau <liviu.dudau@arm.com>

Best regards,
Liviu

> ---
>  drivers/gpu/drm/panthor/panthor_device.c | 4 ++--
>  1 file changed, 2 insertions(+), 2 deletions(-)
> 
> diff --git a/drivers/gpu/drm/panthor/panthor_device.c b/drivers/gpu/drm/panthor/panthor_device.c
> index 81df49880bd87..962a10e00848e 100644
> --- a/drivers/gpu/drm/panthor/panthor_device.c
> +++ b/drivers/gpu/drm/panthor/panthor_device.c
> @@ -83,6 +83,8 @@ void panthor_device_unplug(struct panthor_device *ptdev)
>  		return;
>  	}
>  
> +	drm_WARN_ON(&ptdev->base, pm_runtime_get_sync(ptdev->base.dev) < 0);
> +
>  	/* Call drm_dev_unplug() so any access to HW blocks happening after
>  	 * that point get rejected.
>  	 */
> @@ -93,8 +95,6 @@ void panthor_device_unplug(struct panthor_device *ptdev)
>  	 */
>  	mutex_unlock(&ptdev->unplug.lock);
>  
> -	drm_WARN_ON(&ptdev->base, pm_runtime_get_sync(ptdev->base.dev) < 0);
> -
>  	/* Now, try to cleanly shutdown the GPU before the device resources
>  	 * get reclaimed.
>  	 */
> -- 
> 2.47.2
> 

-- 
====================
| I would like to |
| fix the world,  |
| but they're not |
| giving me the   |
 \ source code!  /
  ---------------
    ¯\_(ツ)_/¯

Re: [PATCH] drm/panthor: Fix race with suspend during unplug

Posted by Steven Price 3 months, 2 weeks ago

On 22/10/2025 11:32, Ketil Johnsen wrote:
> There is a race between panthor_device_unplug() and
> panthor_device_suspend() which can lead to IRQ handlers running on a
> powered down GPU. This is how it can happen:
> - unplug routine calls drm_dev_unplug()
> - panthor_device_suspend() can now execute, and will skip a lot of
>   important work because the device is currently marked as unplugged.
> - IRQs will remain active in this case and IRQ handlers can therefore
>   try to access a powered down GPU.
> 
> The fix is simply to take the PM ref in panthor_device_unplug() a
> little bit earlier, before drm_dev_unplug().
> 
> Signed-off-by: Ketil Johnsen <ketil.johnsen@arm.com>
> Fixes: 5fe909cae118a ("drm/panthor: Add the device logical block")

Reviewed-by: Steven Price <steven.price@arm.com>

> ---
>  drivers/gpu/drm/panthor/panthor_device.c | 4 ++--
>  1 file changed, 2 insertions(+), 2 deletions(-)
> 
> diff --git a/drivers/gpu/drm/panthor/panthor_device.c b/drivers/gpu/drm/panthor/panthor_device.c
> index 81df49880bd87..962a10e00848e 100644
> --- a/drivers/gpu/drm/panthor/panthor_device.c
> +++ b/drivers/gpu/drm/panthor/panthor_device.c
> @@ -83,6 +83,8 @@ void panthor_device_unplug(struct panthor_device *ptdev)
>  		return;
>  	}
>  
> +	drm_WARN_ON(&ptdev->base, pm_runtime_get_sync(ptdev->base.dev) < 0);
> +
>  	/* Call drm_dev_unplug() so any access to HW blocks happening after
>  	 * that point get rejected.
>  	 */
> @@ -93,8 +95,6 @@ void panthor_device_unplug(struct panthor_device *ptdev)
>  	 */
>  	mutex_unlock(&ptdev->unplug.lock);
>  
> -	drm_WARN_ON(&ptdev->base, pm_runtime_get_sync(ptdev->base.dev) < 0);
> -
>  	/* Now, try to cleanly shutdown the GPU before the device resources
>  	 * get reclaimed.
>  	 */