[PATCH] mm, soft offline: split thp at the beginning of soft_offline_page()

[PATCH] mm, soft offline: split thp at the beginning of soft_offline_page()

[Naoya Horiguchi]

When we try to soft-offline a thp tail page, put_page() is called on the
tail page unthinkingly and VM_BUG_ON is triggered in put_compound_page().
This patch splits thp before going into the main body of soft-offlining.

The interface of soft-offlining is open for userspace, so this bug can
lead to DoS attack and should be fixed immedately.

Signed-off-by: Naoya Horiguchi <n-horiguchi@ah.jp.nec.com>
Cc: stable@vger.kernel.org
---
 mm/memory-failure.c | 8 ++++++++
 1 file changed, 8 insertions(+)

diff --git v3.7-rc7.orig/mm/memory-failure.c v3.7-rc7/mm/memory-failure.c
index 8fe3640..e48e235 100644
--- v3.7-rc7.orig/mm/memory-failure.c
+++ v3.7-rc7/mm/memory-failure.c
@@ -1548,9 +1548,17 @@ int soft_offline_page(struct page *page, int flags)
 {
 	int ret;
 	unsigned long pfn = page_to_pfn(page);
+	struct page *hpage = compound_trans_head(page);
 
 	if (PageHuge(page))
 		return soft_offline_huge_page(page, flags);
+	if (PageTransHuge(hpage)) {
+		if (PageAnon(hpage) && unlikely(split_huge_page(hpage))) {
+			pr_info("soft offline: %#lx: failed to split THP\n",
+				pfn);
+			return -EBUSY;
+		}
+	}
 
 	ret = get_any_page(page, pfn, flags);
 	if (ret < 0)
-- 
1.7.11.7

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

Re: [PATCH] mm, soft offline: split thp at the beginning of soft_offline_page()

[Andi Kleen]

Naoya Horiguchi <n-horiguchi@ah.jp.nec.com> writes:

> When we try to soft-offline a thp tail page, put_page() is called on the
> tail page unthinkingly and VM_BUG_ON is triggered in put_compound_page().
> This patch splits thp before going into the main body of soft-offlining.

Looks good.

>
> The interface of soft-offlining is open for userspace, so this bug can
> lead to DoS attack and should be fixed immedately.

The interface is root only and root can do everything anyways, so it's
not really a security issue.

-Andi

-- 
ak@linux.intel.com -- Speaking for myself only

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

Re: [PATCH] mm, soft offline: split thp at the beginning of soft_offline_page()

[Naoya Horiguchi]

On Tue, Nov 27, 2012 at 01:08:38PM -0800, Andi Kleen wrote:
> Naoya Horiguchi <n-horiguchi@ah.jp.nec.com> writes:
> 
> > When we try to soft-offline a thp tail page, put_page() is called on the
> > tail page unthinkingly and VM_BUG_ON is triggered in put_compound_page().
> > This patch splits thp before going into the main body of soft-offlining.
> 
> Looks good.
> 
> >
> > The interface of soft-offlining is open for userspace, so this bug can
> > lead to DoS attack and should be fixed immedately.
> 
> The interface is root only and root can do everything anyways, so it's
> not really a security issue.

OK, this description had better not be here. I replace with attached one.

(.. and I forgot to disable suppress-cc for stable@vger.kernel.org in
the initial post so did it now, sorry.)

---
From: Naoya Horiguchi <n-horiguchi@ah.jp.nec.com>
Date: Tue, 27 Nov 2012 15:38:23 -0500
Subject: [PATCH] mm, soft offline: split thp at the beginning of
 soft_offline_page()

When we try to soft-offline a thp tail page, put_page() is called on the
tail page unthinkingly and VM_BUG_ON is triggered in put_compound_page().
This patch splits thp before going into the main body of soft-offlining.

Signed-off-by: Naoya Horiguchi <n-horiguchi@ah.jp.nec.com>
Cc: stable@vger.kernel.org
---
 mm/memory-failure.c | 8 ++++++++
 1 file changed, 8 insertions(+)

diff --git a/mm/memory-failure.c b/mm/memory-failure.c
index 8fe3640..e48e235 100644
--- a/mm/memory-failure.c
+++ b/mm/memory-failure.c
@@ -1548,9 +1548,17 @@ int soft_offline_page(struct page *page, int flags)
 {
 	int ret;
 	unsigned long pfn = page_to_pfn(page);
+	struct page *hpage = compound_trans_head(page);
 
 	if (PageHuge(page))
 		return soft_offline_huge_page(page, flags);
+	if (PageTransHuge(hpage)) {
+		if (PageAnon(hpage) && unlikely(split_huge_page(hpage))) {
+			pr_info("soft offline: %#lx: failed to split THP\n",
+				pfn);
+			return -EBUSY;
+		}
+	}
 
 	ret = get_any_page(page, pfn, flags);
 	if (ret < 0)
-- 
1.7.11.7

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

Re: [PATCH] mm, soft offline: split thp at the beginning of soft_offline_page()

[Andrew Morton]

On Tue, 27 Nov 2012 16:05:31 -0500
Naoya Horiguchi <n-horiguchi@ah.jp.nec.com> wrote:

> When we try to soft-offline a thp tail page, put_page() is called on the
> tail page unthinkingly and VM_BUG_ON is triggered in put_compound_page().
> This patch splits thp before going into the main body of soft-offlining.
> 
> The interface of soft-offlining is open for userspace, so this bug can
> lead to DoS attack and should be fixed immedately.
> 
> Signed-off-by: Naoya Horiguchi <n-horiguchi@ah.jp.nec.com>
> Cc: stable@vger.kernel.org
> ---
>  mm/memory-failure.c | 8 ++++++++
>  1 file changed, 8 insertions(+)
> 
> diff --git v3.7-rc7.orig/mm/memory-failure.c v3.7-rc7/mm/memory-failure.c
> index 8fe3640..e48e235 100644
> --- v3.7-rc7.orig/mm/memory-failure.c
> +++ v3.7-rc7/mm/memory-failure.c
> @@ -1548,9 +1548,17 @@ int soft_offline_page(struct page *page, int flags)
>  {
>  	int ret;
>  	unsigned long pfn = page_to_pfn(page);
> +	struct page *hpage = compound_trans_head(page);
>  
>  	if (PageHuge(page))
>  		return soft_offline_huge_page(page, flags);
> +	if (PageTransHuge(hpage)) {
> +		if (PageAnon(hpage) && unlikely(split_huge_page(hpage))) {
> +			pr_info("soft offline: %#lx: failed to split THP\n",
> +				pfn);
> +			return -EBUSY;
> +		}
> +	}

We can use PageTransCompound() here, as we know it isn't a hugetlbfs
page.  This will then permit the PageAnon() test to be omitted, methinks?

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

Re: [PATCH] mm, soft offline: split thp at the beginning of soft_offline_page()

[Naoya Horiguchi]

On Tue, Nov 27, 2012 at 01:54:58PM -0800, Andrew Morton wrote:
> On Tue, 27 Nov 2012 16:05:31 -0500
> Naoya Horiguchi <n-horiguchi@ah.jp.nec.com> wrote:
> 
> > When we try to soft-offline a thp tail page, put_page() is called on the
> > tail page unthinkingly and VM_BUG_ON is triggered in put_compound_page().
> > This patch splits thp before going into the main body of soft-offlining.
> > 
> > The interface of soft-offlining is open for userspace, so this bug can
> > lead to DoS attack and should be fixed immedately.
> > 
> > Signed-off-by: Naoya Horiguchi <n-horiguchi@ah.jp.nec.com>
> > Cc: stable@vger.kernel.org
> > ---
> >  mm/memory-failure.c | 8 ++++++++
> >  1 file changed, 8 insertions(+)
> > 
> > diff --git v3.7-rc7.orig/mm/memory-failure.c v3.7-rc7/mm/memory-failure.c
> > index 8fe3640..e48e235 100644
> > --- v3.7-rc7.orig/mm/memory-failure.c
> > +++ v3.7-rc7/mm/memory-failure.c
> > @@ -1548,9 +1548,17 @@ int soft_offline_page(struct page *page, int flags)
> >  {
> >  	int ret;
> >  	unsigned long pfn = page_to_pfn(page);
> > +	struct page *hpage = compound_trans_head(page);
> >  
> >  	if (PageHuge(page))
> >  		return soft_offline_huge_page(page, flags);
> > +	if (PageTransHuge(hpage)) {
> > +		if (PageAnon(hpage) && unlikely(split_huge_page(hpage))) {
> > +			pr_info("soft offline: %#lx: failed to split THP\n",
> > +				pfn);
> > +			return -EBUSY;
> > +		}
> > +	}
> 
> We can use PageTransCompound() here, as we know it isn't a hugetlbfs
> page.  This will then permit the PageAnon() test to be omitted, methinks?

Using PageTransCompound() is OK for me. But without this PageAnon() test,
we can trigger BUG_ON(!PageAnon) inside split_huge_page() when soft
offlining is kicked on non-huge compound pages.

The point about telling thp from hugetlbfs pages and non-huge compound
pages was discussed last month over 7a71932d5676b7410 ("kpageflags: fix
wrong KPF_THP on non-huge compound pages"), where we used a bit tricky
workaround of using PageTransCompound(page) && PageLRU(compound_trans_head(page)).

Thanks,
Naoya

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>