drm/i915: Skip object locking around a no-op set-domain ioctl

If we are already in the desired write domain of a set-domain ioctl,
then there is nothing for us to do and we can quickly return back to
userspace, avoiding any lock contention. By recognising that the
write_domain is always a subset of the read_domains, and excluding the
no-op case of requiring 0 read_domains in the ioctl, we can infer if the
current write_domain matches the target read_domains, there is nothing
for us to do.

Secondary aspect of this is that we undo the arbitrary fetching and
potential flushing of all pages for a set-domain(.write=CPU) call on a
fresh object -- which was introduced simply because we do the get-pages
before taking the struct_mutex.

References: 40e62d5d6b ("drm/i915: Acquire the backing storage outside of struct_mutex in set-domain")
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Cc: Matthew Auld <matthew.william.auld@gmail.com>
Reviewed-by: Matthew Auld <matthew.william.auld@gmail.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20190321161908.8007-2-chris@chris-wilson.co.uk
This commit is contained in:
Chris Wilson 2019-03-21 16:19:08 +00:00
parent a679f58d05
commit 754a254427
1 changed files with 23 additions and 3 deletions

View File

@ -1484,17 +1484,37 @@ i915_gem_set_domain_ioctl(struct drm_device *dev, void *data,
if ((write_domain | read_domains) & I915_GEM_GPU_DOMAINS)
return -EINVAL;
/* Having something in the write domain implies it's in the read
/*
* Having something in the write domain implies it's in the read
* domain, and only that read domain. Enforce that in the request.
*/
if (write_domain != 0 && read_domains != write_domain)
if (write_domain && read_domains != write_domain)
return -EINVAL;
if (!read_domains)
return 0;
obj = i915_gem_object_lookup(file, args->handle);
if (!obj)
return -ENOENT;
/* Try to flush the object off the GPU without holding the lock.
/*
* Already in the desired write domain? Nothing for us to do!
*
* We apply a little bit of cunning here to catch a broader set of
* no-ops. If obj->write_domain is set, we must be in the same
* obj->read_domains, and only that domain. Therefore, if that
* obj->write_domain matches the request read_domains, we are
* already in the same read/write domain and can skip the operation,
* without having to further check the requested write_domain.
*/
if (READ_ONCE(obj->write_domain) == read_domains) {
err = 0;
goto out;
}
/*
* Try to flush the object off the GPU without holding the lock.
* We will repeat the flush holding the lock in the normal manner
* to catch cases where we are gazumped.
*/