mm,oom: do not loop !__GFP_FS allocation if the OOM killer is disabled

[uclinux-h8/linux.git] / mm / page_alloc.c

diff --git a/mm/page_alloc.c b/mm/page_alloc.c
index 3583f71..a762be5 100644 (file)
--- a/mm/page_alloc.c
+++ b/mm/page_alloc.c
@@ -2858,8 +2858,12 @@ __alloc_pages_may_oom(gfp_t gfp_mask, unsigned int order,
                         * XXX: Page reclaim didn't yield anything,
                         * and the OOM killer can't be invoked, but
                         * keep looping as per tradition.
+                        *
+                        * But do not keep looping if oom_killer_disable()
+                        * was already called, for the system is trying to
+                        * enter a quiescent state during suspend.
                         */
-                       *did_some_progress = 1;
+                       *did_some_progress = !oom_killer_disabled;
                        goto out;
                }
                if (pm_suspended_storage())