[Intel-gfx] [PATCH 07/39] drm/i915: Introduce the i915_user_extension_method

[Tvrtko Ursulin]

On 14/03/2019 14:59, Chris Wilson wrote:
> Quoting Tvrtko Ursulin (2019-03-14 14:52:00)
>>
>> On 13/03/2019 14:43, Chris Wilson wrote:
>>> +int i915_user_extensions(struct i915_user_extension __user *ext,
>>> +                      const i915_user_extension_fn *tbl,
>>> +                      unsigned long count,
>>
>> Could shrink to unsigned int if going to 32-bit "name". And...
>>
>>> +                      void *data)
>>> +{
>>> +     unsigned int stackdepth = 512;
>>> +
>>> +     while (ext) {
>>> +             int i, err;
>>> +             u64 x;
>>
>> ... so it is u32 in this version.
> 
> This was a convenience to use one variable for both; it doesn't affect
> the get_user() as that is determined by the sizeof(*ptr) and not the

My bad, I thought the destination controls the size.

> local. However, it did upset array_index_nospec() on 32bit, so it did
> become a u32. So unsigned int count also makes some sense, I was using
> the natural type for ARRAY_SIZE() so that we didn't accidentally
> truncate, but shrinking the array bound is not as harmful.

Ok.

Regards,

Tvrtko