Re: oops in copy_page_rep()

[Hillf Danton]

On Tue, Jan 8, 2013 at 1:34 AM, Linus Torvalds
<torvalds@xxxxxxxxxxxxxxxxxxxx> wrote:
> On Mon, Jan 7, 2013 at 4:24 AM, Hillf Danton <dhillf@xxxxxxxxx> wrote:
>>
>> I take another try with waiting added, take a look please.
>
> Hmm. Is there some reason we never need to worry about it for the
> "pmd_numa()" case just above?
>
> A comment about this all might be a really good idea.
>
Yes Sir, added.
---
From: Hillf Danton <dhillf@xxxxxxxxx>
Subject: [PATCH] mm: restore huge pmd splitting check

Hugh said, it's clear that 3.7 had an important pmd_trans_splitting(orig_pmd)
check there, which went AWOL in
d10e63f29488 "mm: numa: Create basic numa page hinting infrastructure".

It is restored for handling stable page fault, with wait_split_huge_page()
added, as suggested also by Hugh, to avoid reapted faults until the split
has completed.

This work is inspired by the oops reported by Dave Jones at
https://lkml.org/lkml/2013/1/5/115

Signed-off-by: Hillf Danton <dhillf@xxxxxxxxx>
---

--- a/mm/memory.c	Sun Jan  6 19:49:50 2013
+++ b/mm/memory.c	Tue Jan  8 20:28:04 2013
@@ -3710,6 +3710,14 @@ retry:
 				return do_huge_pmd_numa_page(mm, vma, address,
 							     orig_pmd, pmd);

+			/*
+			 * Check if pmd is stable
+			 * (numa pmd is stable, see change_huge_pmd())
+			 */
+			if (pmd_trans_splitting(orig_pmd)) {
+				wait_split_huge_page(vma->anon_vma, pmd);
+				goto retry;
+			}
 			if (dirty && !pmd_write(orig_pmd)) {
 				ret = do_huge_pmd_wp_page(mm, vma, address, pmd,
 							  orig_pmd);
--
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/