The shrinker is prepared to be called unlock (and with struct_mutex held for DIRECT_RECLAIM) so we can skip acquiring the struct_mutex prior to calling the shrinking during freeze. This improves our ability to shrink as we can be more aggressive when we know the caller isn't holding struct_mutex.
Signed-off-by: Chris Wilson <[email protected]> Cc: Joonas Lahtinen <[email protected]> --- drivers/gpu/drm/i915/i915_gem.c | 9 +++++---- 1 file changed, 5 insertions(+), 4 deletions(-) diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c index 52b403b4f8c2..13cad69df20e 100644 --- a/drivers/gpu/drm/i915/i915_gem.c +++ b/drivers/gpu/drm/i915/i915_gem.c @@ -4886,9 +4886,10 @@ void i915_gem_load_cleanup(struct drm_i915_private *dev_priv) int i915_gem_freeze(struct drm_i915_private *dev_priv) { - mutex_lock(&dev_priv->drm.struct_mutex); + /* Discard all purgeable objects, let userspace recover those as + * required after resuming. + */ i915_gem_shrink_all(dev_priv); - mutex_unlock(&dev_priv->drm.struct_mutex); return 0; } @@ -4913,12 +4914,12 @@ int i915_gem_freeze_late(struct drm_i915_private *dev_priv) * we update that state just before writing out the image. * * To try and reduce the hibernation image, we manually shrink - * the objects as well. + * the objects as well, see i915_gem_freeze() */ - mutex_lock(&dev_priv->drm.struct_mutex); i915_gem_shrink(dev_priv, -1UL, I915_SHRINK_UNBOUND); + mutex_lock(&dev_priv->drm.struct_mutex); for (p = phases; *p; p++) { list_for_each_entry(obj, *p, global_link) { obj->base.read_domains = I915_GEM_DOMAIN_CPU; -- 2.11.0 _______________________________________________ Intel-gfx mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/intel-gfx
