mm/smaps: don't access young/dirty bit if pte unpresent
authorPeter Xu <peterx@redhat.com>
Fri, 5 Aug 2022 16:00:03 +0000 (12:00 -0400)
committerGreg Kroah-Hartman <gregkh@linuxfoundation.org>
Wed, 31 Aug 2022 15:16:37 +0000 (17:16 +0200)
commitd28f319043f08ee4049f1dd28239ee6a350aada9
treee3cb5e45690ff63212a932dda9b324e1fe8c6034
parent28dccc4eaf9864f194b0eb0c44b1e5cbf9935971
mm/smaps: don't access young/dirty bit if pte unpresent

[ Upstream commit efd4149342db2df41b1bbe68972ead853b30e444 ]

These bits should only be valid when the ptes are present.  Introducing
two booleans for it and set it to false when !pte_present() for both pte
and pmd accountings.

The bug is found during code reading and no real world issue reported, but
logically such an error can cause incorrect readings for either smaps or
smaps_rollup output on quite a few fields.

For example, it could cause over-estimate on values like Shared_Dirty,
Private_Dirty, Referenced.  Or it could also cause under-estimate on
values like LazyFree, Shared_Clean, Private_Clean.

Link: https://lkml.kernel.org/r/20220805160003.58929-1-peterx@redhat.com
Fixes: b1d4d9e0cbd0 ("proc/smaps: carefully handle migration entries")
Fixes: c94b6923fa0a ("/proc/PID/smaps: Add PMD migration entry parsing")
Signed-off-by: Peter Xu <peterx@redhat.com>
Reviewed-by: Vlastimil Babka <vbabka@suse.cz>
Reviewed-by: David Hildenbrand <david@redhat.com>
Reviewed-by: Yang Shi <shy828301@gmail.com>
Cc: Konstantin Khlebnikov <khlebnikov@openvz.org>
Cc: Huang Ying <ying.huang@intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>
fs/proc/task_mmu.c