I found PCG_ACCT_LRU is copied at splitting in patch 2/4 but it will cause VM_BUG_ON(). The fix for it was in patch 3/4...This set is a corrected one. == Now, when THP is enabled, memcg's counter goes wrong. Moreover, rmdir() may not end. I fixed some races since v1. This series is a fix for obviouse counter breakage. When you test, CONFIG_TRANSPARENT_HUGEPAGE=y CONFIG_TRANSPARENT_HUGEPAGE_ALWAYS=y is appreciated. Tests should be done is: # mount -t cgroup none /cgroup/memory -omemory # mkdir /cgroup/memory/A # mkdir /cgroup/memory/A/B # run some programs under B. # echo 0 > /cgroup/memory/A/B/memory.force_empty and check B's memory.stat shows RSS/CACHE/LRU are all 0. Moving tasks while running is another good test. I know there are another problem when memory cgroup hits limit and reclaim in busy. But I will fix it in another patch. Thanks, -Kame -- To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to majordomo@xxxxxxxxxx For more info on Linux MM, see: http://www.linux-mm.org/ . Fight unfair telecom policy in Canada: sign http://dissolvethecrtc.ca/ Don't email: <a href=mailto:"dont@xxxxxxxxx";> email@xxxxxxxxx </a>