mm/rmap: map_pte() was not handling private ZONE_DEVICE page properly
authorRalph Campbell <[email protected]>
Tue, 30 Oct 2018 22:04:11 +0000 (15:04 -0700)
committerLinus Torvalds <[email protected]>
Wed, 31 Oct 2018 15:54:11 +0000 (08:54 -0700)
commitaab8d0520e6e7c2a61f71195e6ce7007a4843afb
treedeffb905900498c2b41705e7da8510457111c86b
parentf813f21971b96f61a789dd48151f92220fdd2e0a
mm/rmap: map_pte() was not handling private ZONE_DEVICE page properly

Private ZONE_DEVICE pages use a special pte entry and thus are not
present.  Properly handle this case in map_pte(), it is already handled in
check_pte(), the map_pte() part was lost in some rebase most probably.

Without this patch the slow migration path can not migrate back to any
private ZONE_DEVICE memory to regular memory.  This was found after stress
testing migration back to system memory.  This ultimatly can lead to the
CPU constantly page fault looping on the special swap entry.

Link: http://lkml.kernel.org/r/[email protected]
Signed-off-by: Ralph Campbell <[email protected]>
Signed-off-by: Jérôme Glisse <[email protected]>
Reviewed-by: Balbir Singh <[email protected]>
Cc: Andrew Morton <[email protected]>
Cc: Kirill A. Shutemov <[email protected]>
Cc: <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
mm/page_vma_mapped.c