On Fri, Feb 07, 2014 at 01:43:29PM -0800, Andrew Morton wrote:
> On Fri, 07 Feb 2014 16:16:04 -0500 Naoya Horiguchi <n-horiguchi@xxxxxxxxxxxxx> wrote:
>
> > # Resending due to sending failure. Sorry if you received twice.
> > ---
> > mce-test detected a test failure when injecting error to a thp tail page.
> > This is because we take page refcount of the tail page in madvise_hwpoison()
> > while the fix in commit a3e0f9e47d5e ("mm/memory-failure.c: transfer page
> > count from head page to tail page after split thp") assumes that we always
> > take refcount on the head page.
> >
> > When a real memory error happens we take refcount on the head page where
> > memory_failure() is called without MF_COUNT_INCREASED set, so it seems to me
> > that testing memory error on thp tail page using madvise makes little sense.
> >
> > This patch cancels moving refcount in !MF_COUNT_INCREASED for valid testing.
> >
> > ...
> >
> > --- v3.14-rc1.orig/mm/memory-failure.c
> > +++ v3.14-rc1/mm/memory-failure.c
> > @@ -1042,8 +1042,10 @@ static int hwpoison_user_mappings(struct page *p, unsigned long pfn,
> > * to it. Similarly, page lock is shifted.
> > */
> > if (hpage != p) {
> > - put_page(hpage);
> > - get_page(p);
> > + if (!(flags && MF_COUNT_INCREASED)) {
>
> s/&&/&/
>
> Please carefully retest this, make sure that both cases are covered?
Sorry for silly mistake, flags is MF_COUNT_INCREASED from madvise()
and 0 from /sys/devices/system/memory/soft_offline_page, so the above
logic accidentally made no difference from the correct one in comparing
with these two cases.
I retested it and confirmed that replaced one fixes the problem.
Thanks,
Naoya
---
From: Naoya Horiguchi <n-horiguchi@xxxxxxxxxxxxx>
Date: Fri, 7 Feb 2014 06:33:30 -0500
Subject: [PATCH v2] mm/memory-failure.c: move refcount only in
!MF_COUNT_INCREASED
mce-test detected a test failure when injecting error to a thp tail page.
This is because we take page refcount of the tail page in madvise_hwpoison()
while the fix in commit a3e0f9e47d5e ("mm/memory-failure.c: transfer page
count from head page to tail page after split thp") assumes that we always
take refcount on the head page.
When a real memory error happens we take refcount on the head page where
memory_failure() is called without MF_COUNT_INCREASED set, so it seems to me
that testing memory error on thp tail page using madvise makes little sense.
This patch cancels moving refcount in !MF_COUNT_INCREASED for valid testing.
Signed-off-by: Naoya Horiguchi <n-horiguchi@xxxxxxxxxxxxx>
Cc: stable@xxxxxxxxxxxxxxx # 3.9+: a3e0f9e47d5e
---
mm/memory-failure.c | 6 ++++--
1 file changed, 4 insertions(+), 2 deletions(-)
diff --git a/mm/memory-failure.c b/mm/memory-failure.c
index ab55d489eb05..b68d14d59784 100644
--- a/mm/memory-failure.c
+++ b/mm/memory-failure.c
@@ -1042,8 +1042,10 @@ static int hwpoison_user_mappings(struct page *p, unsigned long pfn,
* to it. Similarly, page lock is shifted.
*/
if (hpage != p) {
- put_page(hpage);
- get_page(p);
+ if (!(flags & MF_COUNT_INCREASED)) {
+ put_page(hpage);
+ get_page(p);
+ }
lock_page(p);
unlock_page(hpage);
*hpagep = p;
--
1.8.5.3
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@xxxxxxxxx. For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@xxxxxxxxx";> email@xxxxxxxxx </a>