Re: [PATCH 3/3] drm/panthor: Move cache-flush after UPDATE(UNMAPPED)

[Liviu Dudau]

On Thu, Sep 24, 2026 at 02:30:45PM +0200, Boris Brezillon wrote:
> Right now, there's a theoretical window during which the GPU can populate
> the cache with data from a VM that's about to be evicted through the
> UPDATE(UNMAPPED) command.
> In practice this won't happen (hence the absence of Fixes tag) because
> when panthor_mmu_as_disable() is called, the VM is guaranteed to be idle
> (no active CSG pointing to this VM), but as we say, better safe than
> sorry.
> 
> Signed-off-by: Boris Brezillon <boris.brezillon@xxxxxxxxxxxxx>

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

Best regards,
Liviu

> ---
>  drivers/gpu/drm/panthor/panthor_mmu.c | 19 ++++++++++++-------
>  1 file changed, 12 insertions(+), 7 deletions(-)
> 
> diff --git a/drivers/gpu/drm/panthor/panthor_mmu.c b/drivers/gpu/drm/panthor/panthor_mmu.c
> index 038a092fd08c..bb71274aa36b 100644
> --- a/drivers/gpu/drm/panthor/panthor_mmu.c
> +++ b/drivers/gpu/drm/panthor/panthor_mmu.c
> @@ -631,12 +631,6 @@ static int panthor_mmu_as_disable(struct panthor_device *ptdev, u32 as_nr)
>  	panthor_mmu_irq_disable_events(&ptdev->mmu->irq,
>  				       panthor_mmu_as_fault_mask(ptdev, as_nr));
>  
> -	/* Flush+invalidate RW caches, invalidate RO ones. */
> -	ret = panthor_gpu_flush_caches(ptdev, CACHE_CLEAN | CACHE_INV,
> -				       CACHE_CLEAN | CACHE_INV, 0);
> -	if (ret)
> -		return ret;
> -
>  	if (vm && vm->locked_region.size) {
>  		/* Unlock the region if there's a lock pending. */
>  		ret = as_send_cmd_and_wait(ptdev, vm->as.id, AS_COMMAND_UNLOCK);
> @@ -648,7 +642,18 @@ static int panthor_mmu_as_disable(struct panthor_device *ptdev, u32 as_nr)
>  	gpu_write64(mmu->iomem, AS_MEMATTR(as_nr), 0);
>  	gpu_write64(mmu->iomem, AS_TRANSCFG(as_nr), AS_TRANSCFG_ADRMODE_UNMAPPED);
>  
> -	return as_send_cmd_and_wait(ptdev, as_nr, AS_COMMAND_UPDATE);
> +	ret = as_send_cmd_and_wait(ptdev, as_nr, AS_COMMAND_UPDATE);
> +	if (ret)
> +		return ret;
> +
> +	/* Flush+invalidate RW caches after we've unmapped, to make sure any
> +	 * cacheline eviction is effective before we potentially return
> +	 * memory pointed by this VM to the system. This needs to be done
> +	 * after the UPDATE(UNMAPPED) operation to guarantee that not further
> +	 * PT-walk can pull VM data into the cache.
> +	 */
> +	return panthor_gpu_flush_caches(ptdev, CACHE_CLEAN | CACHE_INV,
> +					CACHE_CLEAN | CACHE_INV, 0);
>  }
>  
>  static u32 panthor_mmu_fault_mask(struct panthor_device *ptdev, u32 value)
> 
> -- 
> 2.55.0
> 

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