Re: [Intel-xe] [PATCH v9 6/8] drm/xe/ct: drop mem_access_get from g2h_worker_func()

[Rodrigo Vivi <rodrigo.vivi@kernel.org>]

On Tue, May 23, 2023 at 10:47:25AM +0100, Matthew Auld wrote:
> In the next patch we track the current task executing the runtime_pm
> callbacks, if any, in order to prevent the runtime_pm code waiting for
> itself and deadlocking when called from the callbacks. However with
> g2h_worker_func() we are processing g2h packets from a workqueue when
> receiving an interrupt. The runtime callbacks needs to talk to the GuC
> and so rely on this, however since these are processed from a workqueue
> the current task will be different and when calling something like
> mem_access_get() we deadlock. Callers should 

> in theory

should we at least add an assert?

> already be
> holding an access.ref when talking to the GuC (CT communication looks to
> be disabled when suspending the GT).
> 
> Signed-off-by: Matthew Auld <matthew.auld@intel.com>
> Cc: Rodrigo Vivi <rodrigo.vivi@intel.com>
> Cc: Thomas Hellström <thomas.hellstrom@linux.intel.com>
> Cc: Matthew Brost <matthew.brost@intel.com>
> ---
>  drivers/gpu/drm/xe/xe_guc_ct.c | 2 --
>  1 file changed, 2 deletions(-)
> 
> diff --git a/drivers/gpu/drm/xe/xe_guc_ct.c b/drivers/gpu/drm/xe/xe_guc_ct.c
> index 7a8c9d6a03f3..6e0d2d06b7f5 100644
> --- a/drivers/gpu/drm/xe/xe_guc_ct.c
> +++ b/drivers/gpu/drm/xe/xe_guc_ct.c
> @@ -1081,7 +1081,6 @@ static void g2h_worker_func(struct work_struct *w)
>  	struct xe_guc_ct *ct = container_of(w, struct xe_guc_ct, g2h_worker);
>  	int ret;
>  
> -	xe_device_mem_access_get(ct_to_xe(ct));
>  	do {
>  		mutex_lock(&ct->lock);
>  		ret = dequeue_one_g2h(ct);
> @@ -1095,7 +1094,6 @@ static void g2h_worker_func(struct work_struct *w)
>  			kick_reset(ct);
>  		}
>  	} while (ret == 1);
> -	xe_device_mem_access_put(ct_to_xe(ct));
>  }
>  
>  static void guc_ctb_snapshot_capture(struct xe_device *xe, struct guc_ctb *ctb,
> -- 
> 2.40.1
>