mm: cma: don't replace lowmem pages with highmem
authorRabin Vincent <rabin@rab.in>
Thu, 5 Jul 2012 10:22:23 +0000 (15:52 +0530)
committerMarek Szyprowski <m.szyprowski@samsung.com>
Fri, 6 Jul 2012 09:07:04 +0000 (11:07 +0200)
The filesystem layer expects pages in the block device's mapping to not
be in highmem (the mapping's gfp mask is set in bdget()), but CMA can
currently replace lowmem pages with highmem pages, leading to crashes in
filesystem code such as the one below:

  Unable to handle kernel NULL pointer dereference at virtual address 00000400
  pgd = c0c98000
  [00000400] *pgd=00c91831, *pte=00000000, *ppte=00000000
  Internal error: Oops: 817 [#1] PREEMPT SMP ARM
  CPU: 0    Not tainted  (3.5.0-rc5+ #80)
  PC is at __memzero+0x24/0x80
  ...
  Process fsstress (pid: 323, stack limit = 0xc0cbc2f0)
  Backtrace:
  [<c010e3f0>] (ext4_getblk+0x0/0x180) from [<c010e58c>] (ext4_bread+0x1c/0x98)
  [<c010e570>] (ext4_bread+0x0/0x98) from [<c0117944>] (ext4_mkdir+0x160/0x3bc)
   r4:c15337f0
  [<c01177e4>] (ext4_mkdir+0x0/0x3bc) from [<c00c29e0>] (vfs_mkdir+0x8c/0x98)
  [<c00c2954>] (vfs_mkdir+0x0/0x98) from [<c00c2a60>] (sys_mkdirat+0x74/0xac)
   r6:00000000 r5:c152eb40 r4:000001ff r3:c14b43f0
  [<c00c29ec>] (sys_mkdirat+0x0/0xac) from [<c00c2ab8>] (sys_mkdir+0x20/0x24)
   r6:beccdcf0 r5:00074000 r4:beccdbbc
  [<c00c2a98>] (sys_mkdir+0x0/0x24) from [<c000e3c0>] (ret_fast_syscall+0x0/0x30)

Fix this by replacing only highmem pages with highmem.

Reported-by: Laura Abbott <lauraa@codeaurora.org>
Signed-off-by: Rabin Vincent <rabin@rab.in>
Acked-by: Michal Nazarewicz <mina86@mina86.com>
Signed-off-by: Marek Szyprowski <m.szyprowski@samsung.com>
mm/page_alloc.c

index 44030096da631b5f49b4ee1a81d0b9ace59f5c7e..4a4f9219683f63d8594c143fa24f27d4daa6a8ff 100644 (file)
@@ -5635,7 +5635,12 @@ static struct page *
 __alloc_contig_migrate_alloc(struct page *page, unsigned long private,
                             int **resultp)
 {
-       return alloc_page(GFP_HIGHUSER_MOVABLE);
+       gfp_t gfp_mask = GFP_USER | __GFP_MOVABLE;
+
+       if (PageHighMem(page))
+               gfp_mask |= __GFP_HIGHMEM;
+
+       return alloc_page(gfp_mask);
 }
 
 /* [start, end) must belong to a single zone. */