[Bug,107328] radeon_gart_table_vram_pin takes 473 ms during ACPI S3 resume
diff mbox series

Message ID bug-107328-502-ux2PPnxKD2@http.bugs.freedesktop.org/
State New
Headers show
Series
  • [Bug,107328] radeon_gart_table_vram_pin takes 473 ms during ACPI S3 resume
Related show

Commit Message

bugzilla-daemon@freedesktop.org July 24, 2018, 9:26 a.m. UTC
https://bugs.freedesktop.org/show_bug.cgi?id=107328

Paul Menzel <pmenzel+bugs.freedesktop@molgen.mpg.de> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
         Resolution|WONTFIX                     |INVALID

--- Comment #2 from Paul Menzel <pmenzel+bugs.freedesktop@molgen.mpg.de> ---
Just as a follow-up, the long time seems to have been caused by ftrace.

With the patch below on top of drm-tip from this morning, it only takes 7 ms. 

                radeon_gart_tlb_flush(rdev);
+               DRM_INFO("GART: Done restoring entries\n");
        }

        return r;

Patch
diff mbox series

diff --git a/drivers/gpu/drm/radeon/radeon_gart.c
b/drivers/gpu/drm/radeon/radeon_gart.c
index 1cef155cc933..f79a2f8b8d8e 100644
--- a/drivers/gpu/drm/radeon/radeon_gart.c
+++ b/drivers/gpu/drm/radeon/radeon_gart.c
@@ -175,10 +175,13 @@  int radeon_gart_table_vram_pin(struct radeon_device
*rdev)
                /* We might have dropped some GART table updates while it
wasn't
                 * mapped, restore all entries
                 */
+               DRM_INFO("GART: Restore entries: num cpu pages %u, num gpu
pages %u\n",
+                               rdev->gart.num_cpu_pages,
rdev->gart.num_gpu_pages);
                for (i = 0; i < rdev->gart.num_gpu_pages; i++)
                        radeon_gart_set_page(rdev, i,
rdev->gart.pages_entry[i]);
                mb();