[Intel-gfx] [RFC 2/8] drm/i915: Reference count struct drm_i915_file_private

[Tvrtko Ursulin]

On 19/12/2019 20:43, Chris Wilson wrote:
> Quoting Tvrtko Ursulin (2019-12-19 18:00:13)
>> From: Tvrtko Ursulin <tvrtko.ursulin at intel.com>
>>
>> In the following patches we will develope a need to peek into the client
>> owned data from any potential leftover contexts.
>>
>> To facilitate this add reference counting to file_priv.
>>
>> Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin at intel.com>
>> ---
>>   drivers/gpu/drm/i915/gem/i915_gem_context.c |  2 +-
>>   drivers/gpu/drm/i915/i915_drv.c             |  4 ----
>>   drivers/gpu/drm/i915/i915_drv.h             |  4 +++-
>>   drivers/gpu/drm/i915/i915_gem.c             | 14 +++++++++++++-
>>   4 files changed, 17 insertions(+), 7 deletions(-)
>>
>> diff --git a/drivers/gpu/drm/i915/gem/i915_gem_context.c b/drivers/gpu/drm/i915/gem/i915_gem_context.c
>> index e5a7c6f02a47..b482b2e5f31f 100644
>> --- a/drivers/gpu/drm/i915/gem/i915_gem_context.c
>> +++ b/drivers/gpu/drm/i915/gem/i915_gem_context.c
>> @@ -853,7 +853,7 @@ int i915_gem_context_open(struct drm_i915_private *i915,
>>   void i915_gem_context_close(struct drm_file *file)
>>   {
>>          struct drm_i915_file_private *file_priv = file->driver_priv;
>> -       struct drm_i915_private *i915 = file_priv->dev_priv;
>> +       struct drm_i915_private *i915 = file_priv->i915;
>>          struct i915_gem_context *ctx;
>>          unsigned long idx;
>>   
>> diff --git a/drivers/gpu/drm/i915/i915_drv.c b/drivers/gpu/drm/i915/i915_drv.c
>> index 8b08cfe30151..0c9c93418068 100644
>> --- a/drivers/gpu/drm/i915/i915_drv.c
>> +++ b/drivers/gpu/drm/i915/i915_drv.c
>> @@ -1633,13 +1633,9 @@ static void i915_driver_lastclose(struct drm_device *dev)
>>   
>>   static void i915_driver_postclose(struct drm_device *dev, struct drm_file *file)
>>   {
>> -       struct drm_i915_file_private *file_priv = file->driver_priv;
>> -
>>          i915_gem_context_close(file);
>>          i915_gem_release(dev, file);
>>   
>> -       kfree_rcu(file_priv, rcu);
> 
> As you are moving the kfree_rcu() into the i915_gem_release (via a put),
> I think it also makes sense to move the call for i915_gem_context_close
> on this file. Possibly renaming it to i915_gem_file_close() and
> s/drm_i915_file_private/i915_gem_file/ or i915_gem_client (with
> corresponding name changes) in the process.
> 
> For the basic mechanics of this patch though,
> Reviewed-by: Chris Wilson <chris at chris-wilson.co.uk>
> 
> (Though I still suggest a bit of playing with i915_gem_context_close,
> i915_gem_release to tie them together to the notion of the file better.)

Yes, agreed in principle.

But a) I prefer the release name to match with the fops mindset and b) I 
prefer to leave drm_i915_file_private alone as the top level driver 
private container.

What I am not completely happy with, or say undecided, is whether to 
move the kref into i915_drm_client. I had it like that at one point, 
thinking to only have a smallest needed structure pinned in memory, but 
then I simplified in favour of fewer allocations. Now I think I'd like 
to move the kref into i915_drm_client again. Any opinion here?

In a later patch, when I add the i915_gem_client_get/put helpers they 
are already named like that. Hm okay, I also have a naming confusion 
between struct i915_drm_client and i915_gem_client_get/put(). :)

Regards,

Tvrtko