mm: check that we have the right vma in __access_remote_vm()
authorMichael Ellerman <michael@ellerman.id.au>
Thu, 14 Apr 2011 22:22:10 +0000 (15:22 -0700)
committerLinus Torvalds <torvalds@linux-foundation.org>
Thu, 14 Apr 2011 23:06:55 +0000 (16:06 -0700)
commitfe936dfc23fed3475b11067e8d9b70553eafcd9e
treeb45ad916853194b26bfe4504879e0bff64a43bf7
parent4471a675dfc7ca676c165079e91c712b09dc9ce4
mm: check that we have the right vma in __access_remote_vm()

In __access_remote_vm() we need to check that we have found the right
vma, not the following vma before we try to access it.  Otherwise we
might call the vma's access routine with an address which does not fall
inside the vma.

It was discovered on a current kernel but with an unreleased driver,
from memory it was strace leading to a kernel bad access, but it
obviously depends on what the access implementation does.

Looking at other access implementations I only see:

  $ git grep -A 5 vm_operations|grep access
  arch/powerpc/platforms/cell/spufs/file.c- .access = spufs_mem_mmap_access,
  arch/x86/pci/i386.c- .access = generic_access_phys,
  drivers/char/mem.c- .access = generic_access_phys
  fs/sysfs/bin.c- .access = bin_access,

The spufs one looks like it might behave badly given the wrong vma, it
assumes vma->vm_file->private_data is a spu_context, and looks like it
would probably blow up pretty quickly if it wasn't.

generic_access_phys() only uses the vma to check vm_flags and get the
mm, and then walks page tables using the address.  So it should bail on
the vm_flags check, or at worst let you access some other VM_IO mapping.

And bin_access() just proxies to another access implementation.

Signed-off-by: Michael Ellerman <michael@ellerman.id.au>
Reviewed-by: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Cc: Hugh Dickins <hughd@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
mm/memory.c