diff options
author | Chris Wilson <chris@chris-wilson.co.uk> | 2016-10-28 15:58:37 +0300 |
---|---|---|
committer | Chris Wilson <chris@chris-wilson.co.uk> | 2016-10-28 22:53:47 +0300 |
commit | 1233e2db199dea015391db03d3478b3392201c41 (patch) | |
tree | 264eee9509dcecd6cdbeeb44200e5ce1af54b0c6 /drivers/gpu/drm/i915/i915_gem_tiling.c | |
parent | 03ac84f1830ec0b90f622500591eb3cc554ee479 (diff) | |
download | linux-1233e2db199dea015391db03d3478b3392201c41.tar.xz |
drm/i915: Move object backing storage manipulation to its own locking
Break the allocation of the backing storage away from struct_mutex into
a per-object lock. This allows parallel page allocation, provided we can
do so outside of struct_mutex (i.e. set-domain-ioctl, pwrite, GTT
fault), i.e. before execbuf! The increased cost of the atomic counters
are hidden behind i915_vma_pin() for the typical case of execbuf, i.e.
as the object is typically bound between execbufs, the page_pin_count is
static. The cost will be felt around set-domain and pwrite, but offset
by the improvement from reduced struct_mutex contention.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20161028125858.23563-14-chris@chris-wilson.co.uk
Diffstat (limited to 'drivers/gpu/drm/i915/i915_gem_tiling.c')
-rw-r--r-- | drivers/gpu/drm/i915/i915_gem_tiling.c | 2 |
1 files changed, 2 insertions, 0 deletions
diff --git a/drivers/gpu/drm/i915/i915_gem_tiling.c b/drivers/gpu/drm/i915/i915_gem_tiling.c index 34d5ada49720..6608799ee1f9 100644 --- a/drivers/gpu/drm/i915/i915_gem_tiling.c +++ b/drivers/gpu/drm/i915/i915_gem_tiling.c @@ -259,6 +259,7 @@ i915_gem_set_tiling(struct drm_device *dev, void *data, if (!err) { struct i915_vma *vma; + mutex_lock(&obj->mm.lock); if (obj->mm.pages && obj->mm.madv == I915_MADV_WILLNEED && dev_priv->quirks & QUIRK_PIN_SWIZZLED_PAGES) { @@ -267,6 +268,7 @@ i915_gem_set_tiling(struct drm_device *dev, void *data, if (!i915_gem_object_is_tiled(obj)) __i915_gem_object_pin_pages(obj); } + mutex_unlock(&obj->mm.lock); list_for_each_entry(vma, &obj->vma_list, obj_link) { if (!vma->fence) |