mm/mmap.c: close race between munmap() and expand_upwards()/downwards()
authorKirill A. Shutemov <kirill.shutemov@linux.intel.com>
Fri, 24 Jul 2020 04:15:11 +0000 (21:15 -0700)
committerLinus Torvalds <torvalds@linux-foundation.org>
Fri, 24 Jul 2020 19:42:41 +0000 (12:42 -0700)
commit246c320a8cfe0b11d81a4af38fa9985ef0cc9a4c
tree6921f9754486258ec5498ff84ed6502d005c6d99
parentf37e99aca03f63aa3f2bd13ceaf769455d12c4b0
mm/mmap.c: close race between munmap() and expand_upwards()/downwards()

VMA with VM_GROWSDOWN or VM_GROWSUP flag set can change their size under
mmap_read_lock().  It can lead to race with __do_munmap():

Thread A Thread B
__do_munmap()
  detach_vmas_to_be_unmapped()
  mmap_write_downgrade()
expand_downwards()
  vma->vm_start = address;
  // The VMA now overlaps with
  // VMAs detached by the Thread A
// page fault populates expanded part
// of the VMA
  unmap_region()
    // Zaps pagetables partly
    // populated by Thread B

Similar race exists for expand_upwards().

The fix is to avoid downgrading mmap_lock in __do_munmap() if detached
VMAs are next to VM_GROWSDOWN or VM_GROWSUP VMA.

[akpm@linux-foundation.org: s/mmap_sem/mmap_lock/ in comment]

Fixes: dd2283f2605e ("mm: mmap: zap pages with read mmap_sem in munmap")
Reported-by: Jann Horn <jannh@google.com>
Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Reviewed-by: Yang Shi <yang.shi@linux.alibaba.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: Oleg Nesterov <oleg@redhat.com>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: <stable@vger.kernel.org> [4.20+]
Link: http://lkml.kernel.org/r/20200709105309.42495-1-kirill.shutemov@linux.intel.com
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
mm/mmap.c