[Intel-gfx] [PATCH 4/5] drm/i915: Support for pread/pwrite from/to non shmem backed objects

[Chris Wilson]

On Wed, Apr 29, 2015 at 11:19:18AM +0100, Chris Wilson wrote:
> On Wed, Apr 29, 2015 at 03:01:58PM +0530, ankitprasad.r.sharma at intel.com wrote:
> > +	while (remain > 0) {
> > +		/* Operation in this page
> > +		 *
> > +		 * page_base = page offset within aperture
> > +		 * page_offset = offset within page
> > +		 * page_length = bytes to copy for this page
> > +		 */
> > +		page_base = offset & PAGE_MASK;
> > +		page_offset = offset_in_page(offset);
> > +		page_length = remain;
> > +		if ((page_offset + remain) > PAGE_SIZE)
> > +			page_length = PAGE_SIZE - page_offset;
> > +
> > +		/* This is a slow read/write as it tries to read from
> > +		 * and write to user memory which may result into page
> > +		 * faults
> > +		 */
> 
> Hmm, I am in the process of rewriting this function (GTT pwrite) to reduce
> aperture pressure (which will also be important here). However for your
> use case, we do need to drop the locks around slow_user_access() so that
> we can use non-atomic copy functions and *pagefault*. We do need the full
> level of complexity like i915_gem_shmem_pread()

It would be a good exercise to try and reproduce the fautling failure condition
with igt/gem_stolen.
-Chris

-- 
Chris Wilson, Intel Open Source Technology Centre