OSDN Git Service

drm/i915: Only attempt to scan the requested number of shrinker slabs
authorChris Wilson <chris@chris-wilson.co.uk>
Mon, 15 Jan 2018 21:24:46 +0000 (21:24 +0000)
committerChris Wilson <chris@chris-wilson.co.uk>
Thu, 18 Jan 2018 09:19:05 +0000 (09:19 +0000)
Since commit 4e773c3a8a69 ("drm/i915: Wire up shrinkctl->nr_scanned"),
we track the number of objects we scan and do not wish to exceed that as
it will overly penalise our own slabs under mempressure. Given that we
now know the target number of objects to scan, use that as our guide for
deciding to shrink as opposed to the number of objects we manage to
shrink (which doesn't correspond to the numbers we report to shrinkctl).

Fixes: 4e773c3a8a69 ("drm/i915: Wire up shrinkctl->nr_scanned")
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20180115212455.24046-2-chris@chris-wilson.co.uk
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
drivers/gpu/drm/i915/i915_gem_shrinker.c

index 9029ed0..0e158f9 100644 (file)
@@ -363,13 +363,13 @@ i915_gem_shrinker_scan(struct shrinker *shrinker, struct shrink_control *sc)
                                I915_SHRINK_BOUND |
                                I915_SHRINK_UNBOUND |
                                I915_SHRINK_PURGEABLE);
-       if (freed < sc->nr_to_scan)
+       if (sc->nr_scanned < sc->nr_to_scan)
                freed += i915_gem_shrink(i915,
                                         sc->nr_to_scan - sc->nr_scanned,
                                         &sc->nr_scanned,
                                         I915_SHRINK_BOUND |
                                         I915_SHRINK_UNBOUND);
-       if (freed < sc->nr_to_scan && current_is_kswapd()) {
+       if (sc->nr_scanned < sc->nr_to_scan && current_is_kswapd()) {
                intel_runtime_pm_get(i915);
                freed += i915_gem_shrink(i915,
                                         sc->nr_to_scan - sc->nr_scanned,