Commit 508b9f8e authored by MinChan Kim's avatar MinChan Kim Committed by Linus Torvalds
Browse files

mm: fix mlocked page counter mismatch



When I tested following program, I found that the mlocked counter
is strange.  It cannot free some mlocked pages.

It is because try_to_unmap_file() doesn't check real
page mappings in vmas.

That is because the goal of an address_space for a file is to find all
processes into which the file's specific interval is mapped.  It is
related to the file's interval, not to pages.

Even if the page isn't really mapped by the vma, it returns SWAP_MLOCK
since the vma has VM_LOCKED, then calls try_to_mlock_page.  After this the
mlocked counter is increased again.

COWed anon page in a file-backed vma could be a such case.  This patch
resolves it.

-- my test program --

int main()
{
       mlockall(MCL_CURRENT);
       return 0;
}

-- before --

root@barrios-target-linux:~# cat /proc/meminfo | egrep 'Mlo|Unev'
Unevictable:           0 kB
Mlocked:               0 kB

-- after --

root@barrios-target-linux:~# cat /proc/meminfo | egrep 'Mlo|Unev'
Unevictable:           8 kB
Mlocked:               8 kB
Signed-off-by: default avatarMinChan Kim <minchan.kim@gmail.com>
Acked-by: default avatarLee Schermerhorn <lee.schermerhorn@hp.com>
Acked-by: default avatarKOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Tested-by: default avatarLee Schermerhorn <lee.schermerhorn@hp.com>
Cc: Hugh Dickins <hugh@veritas.com>
Signed-off-by: default avatarAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: default avatarLinus Torvalds <torvalds@linux-foundation.org>
parent 02ac597c
......@@ -1072,7 +1072,8 @@ static int try_to_unmap_file(struct page *page, int unlock, int migration)
spin_lock(&mapping->i_mmap_lock);
vma_prio_tree_foreach(vma, &iter, &mapping->i_mmap, pgoff, pgoff) {
if (MLOCK_PAGES && unlikely(unlock)) {
if (!(vma->vm_flags & VM_LOCKED))
if (!((vma->vm_flags & VM_LOCKED) &&
page_mapped_in_vma(page, vma)))
continue; /* must visit all vmas */
ret = SWAP_MLOCK;
} else {
......
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment