[Intel-gfx] [PATCH] [v2] drm/i915: Check for get_pages instead of shmem (filp)

[Dave Gordon]

On 10/02/16 17:39, Ben Widawsky wrote:
> On Wed, Feb 10, 2016 at 04:23:08PM +0000, Chris Wilson wrote:
>> On Wed, Feb 10, 2016 at 07:42:23AM -0800, Ben Widawsky wrote:
>>> Do you guys get the CI mails? This version has regressions. v1 did not. I don't
>>> know what to trust.
>>
>> I didn't even see v2 itself!
>>
>>> On Tue, Feb 09, 2016 at 11:44:12AM -0800, Ben Widawsky wrote:
>>>> This behavior of checking for a shmem backed GEM object was introduced here:
>>>> commit 4c914c0c7c787b8f730128a8cdcca9c50b0784ab
>>>> Author: Brad Volkin <bradley.d.volkin at intel.com>
>>>> Date:   Tue Feb 18 10:15:45 2014 -0800
>>>>
>>>>      drm/i915: Refactor shmem pread setup
>>>>
>>>> It is possible for an object to not be a shmem backed GEM object (for example
>>>> userptr objects). An example of how we hit this failure can be found through
>>>> copy_batch() in the command parser because we allocate a userptr object for the
>>>> batch which contains privileged instructions. Userptr calls
>>>> drm_gem_private_object_init() which explicitly sets the filp to none.
>>>>
>>>> NOTE: I manually retyped this from a test machine. So I haven't even compiled
>>>> this exact patch.
>>>>
>>>> v2: Use same logic as from a2a4f916c2f (Kristian, Dave Gordon)
>>>>
>>>> Cc: Chris Wilson <chris at chris-wilson.co.uk>
>>>> Cc: Kristian Høgsberg <krh at bitplanet.net>
>>>> Cc: Dave Gordon <david.s.gordon at intel.com>
>>>> Signed-off-by: Ben Widawsky <benjamin.widawsky at intel.com>
>>>> Tested-by: Jordan Justen <jordan.l.justen at intel.com> (v1)
>>>> Reviewed-by: Jordan Justen <jordan.l.justen at intel.com> (v1)
>>
>>>> ---
>>>>   drivers/gpu/drm/i915/i915_gem.c | 2 +-
>>>>   1 file changed, 1 insertion(+), 1 deletion(-)
>>>>
>>>> diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c
>>>> index e9b19bc..7fd79b0 100644
>>>> --- a/drivers/gpu/drm/i915/i915_gem.c
>>>> +++ b/drivers/gpu/drm/i915/i915_gem.c
>>>> @@ -489,7 +489,7 @@ int i915_gem_obj_prepare_shmem_read(struct drm_i915_gem_object *obj,
>>>>
>>>>   	*needs_clflush = 0;
>>>>
>>>> -	if (!obj->base.filp)
>>>> +	if (WARN_ON((obj->ops->flags & I915_GEM_OBJECT_HAS_STRUCT_PAGE) == 0))

It looks like the above code can't be reached? 'filp' is tested in 
i915_gem_pread_ioctl(), before i915_gem_obj_prepare_shmem_read() is called!

>> Don't use WARN_ON, there is code (or will be) where we use
>> prepare_shmem_read/write to determine if we can use the shmem paths.
>>
>> Also i915_gem_obj_prepare_shmem_write() requires the same treatment.

No such function, but there's a 'filp' test in i915_gem_pwrite_ioctl().

Also, what about i915_gem_mmap_ioctl() ? Is mmap() also going to be 
legitimate without a file pointer?

> > My apologies I had this patch but appear to have accidentally squashed
>> it whilst rebasing. Thanks!
>> -Chris
>
> So... is someone going to land this fix? We need it.

Looks like Chris wants this to read

     if (!(obj->ops->flags & I915_GEM_OBJECT_HAS_STRUCT_PAGE))
         return -EINVAL;

leaving it to the caller to decide whether to log a complaint.
I'm happy with that.

.Dave.