huge tmpfs: fix Committed_AS leak
authorHugh Dickins <hughd@google.com>
Sat, 24 Sep 2016 03:24:23 +0000 (20:24 -0700)
committerLinus Torvalds <torvalds@linux-foundation.org>
Sat, 24 Sep 2016 18:20:01 +0000 (11:20 -0700)
Under swapping load on huge tmpfs, /proc/meminfo's Committed_AS grows
bigger and bigger: just a cosmetic issue for most users, but disabling
for those who run without overcommit (/proc/sys/vm/overcommit_memory 2).

shmem_uncharge() was forgetting to unaccount __vm_enough_memory's
charge, and shmem_charge() was forgetting it on the filesystem-full
error path.

Fixes: 800d8c63b2e9 ("shmem: add huge pages support")
Signed-off-by: Hugh Dickins <hughd@google.com>
Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
mm/shmem.c

index aec5b492d94754de80033942be8a6d1197013e03..971fc83e6402858dd558d0f04609ddad38373eb4 100644 (file)
@@ -270,7 +270,7 @@ bool shmem_charge(struct inode *inode, long pages)
                info->alloced -= pages;
                shmem_recalc_inode(inode);
                spin_unlock_irqrestore(&info->lock, flags);
-
+               shmem_unacct_blocks(info->flags, pages);
                return false;
        }
        percpu_counter_add(&sbinfo->used_blocks, pages);
@@ -291,6 +291,7 @@ void shmem_uncharge(struct inode *inode, long pages)
 
        if (sbinfo->max_blocks)
                percpu_counter_sub(&sbinfo->used_blocks, pages);
+       shmem_unacct_blocks(info->flags, pages);
 }
 
 /*