powerpc/thp: Fix crash on mremap
authorAneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Mon, 13 Jan 2014 06:04:24 +0000 (11:34 +0530)
committerBenjamin Herrenschmidt <benh@kernel.crashing.org>
Wed, 15 Jan 2014 04:46:38 +0000 (15:46 +1100)
commitb3084f4db3aeb991c507ca774337c7e7893ed04f
treec5cad3a6843272a173e95a87a6087d362e19073e
parenta6da83f98267bc8ee4e34aa899169991eb0ceb93
powerpc/thp: Fix crash on mremap

This patch fix the below crash

NIP [c00000000004cee4] .__hash_page_thp+0x2a4/0x440
LR [c0000000000439ac] .hash_page+0x18c/0x5e0
...
Call Trace:
[c000000736103c40] [00001ffffb000000] 0x1ffffb000000(unreliable)
[437908.479693] [c000000736103d50] [c0000000000439ac] .hash_page+0x18c/0x5e0
[437908.479699] [c000000736103e30] [c00000000000924c] .do_hash_page+0x4c/0x58

On ppc64 we use the pgtable for storing the hpte slot information and
store address to the pgtable at a constant offset (PTRS_PER_PMD) from
pmd. On mremap, when we switch the pmd, we need to withdraw and deposit
the pgtable again, so that we find the pgtable at PTRS_PER_PMD offset
from new pmd.

We also want to move the withdraw and deposit before the set_pmd so
that, when page fault find the pmd as trans huge we can be sure that
pgtable can be located at the offset.

Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
arch/powerpc/include/asm/pgtable-ppc64.h
include/asm-generic/pgtable.h
mm/huge_memory.c