mm: softdirty: unmapped addresses between VMAs are clean
authorPeter Feiner <pfeiner@google.com>
Thu, 9 Oct 2014 22:28:32 +0000 (15:28 -0700)
committerLinus Torvalds <torvalds@linux-foundation.org>
Fri, 10 Oct 2014 02:25:58 +0000 (22:25 -0400)
commit81d0fa623c5b8dbd5279d9713094b0f9b0a00fb4
tree840455ffeafaef31a53b94dc74df80d2561c713e
parent3193913ce62c63056bc67a6ae378beaf494afa66
mm: softdirty: unmapped addresses between VMAs are clean

If a /proc/pid/pagemap read spans a [VMA, an unmapped region, then a
VM_SOFTDIRTY VMA], the virtual pages in the unmapped region are reported
as softdirty.  Here's a program to demonstrate the bug:

int main() {
const uint64_t PAGEMAP_SOFTDIRTY = 1ul << 55;
uint64_t pme[3];
int fd = open("/proc/self/pagemap", O_RDONLY);;
char *m = mmap(NULL, 3 * getpagesize(), PROT_READ,
               MAP_ANONYMOUS | MAP_SHARED, -1, 0);
munmap(m + getpagesize(), getpagesize());
pread(fd, pme, 24, (unsigned long) m / getpagesize() * 8);
assert(pme[0] & PAGEMAP_SOFTDIRTY);    /* passes */
assert(!(pme[1] & PAGEMAP_SOFTDIRTY)); /* fails */
assert(pme[2] & PAGEMAP_SOFTDIRTY);    /* passes */
return 0;
}

(Note that all pages in new VMAs are softdirty until cleared).

Tested:
Used the program given above. I'm going to include this code in
a selftest in the future.

[n-horiguchi@ah.jp.nec.com: prevent pagemap_pte_range() from overrunning]
Signed-off-by: Peter Feiner <pfeiner@google.com>
Cc: "Kirill A. Shutemov" <kirill@shutemov.name>
Cc: Cyrill Gorcunov <gorcunov@openvz.org>
Cc: Pavel Emelyanov <xemul@parallels.com>
Cc: Jamie Liu <jamieliu@google.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Naoya Horiguchi <n-horiguchi@ah.jp.nec.com>
Signed-off-by: Naoya Horiguchi <n-horiguchi@ah.jp.nec.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
fs/proc/task_mmu.c