[Intel-gfx] [PATCH] drm/i915: Kick rcu harder to free objects

[Tvrtko Ursulin]

On 08/09/2022 15:32, Das, Nirmoy wrote:
> Hi Ville,
> 
> 
> I fixed a similar issue in DII but I couldn't reproduce it in drm
> 
> http://intel-gfx-pw.fi.intel.com/patch/228850/?series=15910&rev=2.
> 
> I wonder if that fixes the problem you are facing then I can send that 
> to drm.
> 
> diff --git a/drivers/gpu/drm/i915/i915_gem.c 
> b/drivers/gpu/drm/i915/i915_gem.c
> index 7809be3a6840..5438e9277924 100644
> --- a/drivers/gpu/drm/i915/i915_gem.c
> +++ b/drivers/gpu/drm/i915/i915_gem.c
> @@ -1213,7 +1213,7 @@  void i915_gem_init_early(struct drm_i915_private 
> *dev_priv)
> 
>   void i915_gem_cleanup_early(struct drm_i915_private *dev_priv)
>   {
> -    i915_gem_drain_freed_objects(dev_priv);
> +    i915_gem_drain_workqueue(dev_priv);
>       GEM_BUG_ON(!llist_empty(&dev_priv->mm.free_list));
>       GEM_BUG_ON(atomic_read(&dev_priv->mm.free_count));
>       drm_WARN_ON(&dev_priv->drm, dev_priv->mm.shrink_count);

Yes why not, more black magic (count to three) but if it works... :) I 
also spy the general area has been a bit neglected. Like:

i915_gem_driver_remove:
...
   i915_gem_drain_workqueue
   i915_gem_drain_freed_objects

While i915_gem_drain_workqueue:
...
   i915_gem_drain_freed_objects

So i915_gem_drain_freed_objects in i915_gem_driver_remove is redundant 
already.

Should i915_gem_drain_freed_objects be unexported and all callers made 
just call i915_gem_drain_workqueue after your patch? Or if "drain free 
objects" is considered more self descriptive it could be made as an 
alias to i915_gem_drain_workqueue.

Regards,

Tvrtko

> 
> 
> Regards,
> 
> Nirmoy
> 
> On 9/6/2022 7:46 PM, Ville Syrjala wrote:
>> From: Ville Syrjälä <ville.syrjala at linux.intel.com>
>>
>> On gen3 the selftests are pretty much always tripping this:
>> <4> [383.822424] pci 0000:00:02.0: drm_WARN_ON(dev_priv->mm.shrink_count)
>> <4> [383.822546] WARNING: CPU: 2 PID: 3560 at 
>> drivers/gpu/drm/i915/i915_gem.c:1223 i915_gem_cleanup_early+0x96/0xb0 
>> [i915]
>>
>> Looks to be due to the status page object lingering on the
>> purge_list. Call synchronize_rcu() ahead of it to make more
>> sure all objects have been freed.
>>
>> Signed-off-by: Ville Syrjälä <ville.syrjala at linux.intel.com>
>> ---
>>   drivers/gpu/drm/i915/i915_gem.c | 1 +
>>   1 file changed, 1 insertion(+)
>>
>> diff --git a/drivers/gpu/drm/i915/i915_gem.c 
>> b/drivers/gpu/drm/i915/i915_gem.c
>> index 0f49ec9d494a..5b61f7ad6473 100644
>> --- a/drivers/gpu/drm/i915/i915_gem.c
>> +++ b/drivers/gpu/drm/i915/i915_gem.c
>> @@ -1098,6 +1098,7 @@ void i915_gem_drain_freed_objects(struct 
>> drm_i915_private *i915)
>>           flush_delayed_work(&i915->bdev.wq);
>>           rcu_barrier();
>>       }
>> +    synchronize_rcu();
>>   }
>>   /*