Skip to contentSkip to conversation

powerpc/mm: Add memory barrier in __hugepte_alloc()

PATCH Applied Sukadev Bhattiprolu 4 participants 5 messages Apr 5, 2016 – Jul 22, 2016

Subsystems (2)
lore
Share & tools
lore
#0 lore raw

PATCH 1/1 powerpc/mm: Add memory barrier in __hugepte_alloc()

Quoted 1 line

From f7b73c6b4508fe9b141a43d92be2f9dd7d3c4a58 Mon Sep 17 00:00:00 2001

From: Sukadev Bhattiprolu <[email protected]>
Date: Thu, 24 Mar 2016 02:07:57 -0400
Subject: [PATCH 1/1] powerpc/mm: Add memory barrier in __hugepte_alloc()

__hugepte_alloc() uses kmem_cache_zalloc() to allocate a zeroed PTE and proceeds to use the newly allocated PTE. Add a memory barrier to make sure that the other CPUs see a properly initialized PTE.

Based on a fix suggested by James Dykman.

---
Note:
The bug was encountered and fix tested on an older version of the kernel. Forward porting to mainline.

---
 arch/powerpc/mm/hugetlbpage.c | 7 +++++++
 1 file changed, 7 insertions(+)
arch/powerpc/mm/hugetlbpage.c +7 -0
index d991b9e..081f679 100644
@@ -81,6 +81,13 @@ static int __hugepte_alloc(struct mm_struct *mm, hugepd_t *hpdp,
 	if (! new)
 		return -ENOMEM;
 
+	/*
+	 * Make sure other cpus find the hugepd set only after a
+	 * properly initialized page table is visible to them.
+	 * For more details look for comment in __pte_alloc().
+	 */
+	smp_wmb();
+
 	spin_lock(&mm->page_table_lock);
 #ifdef CONFIG_PPC_FSL_BOOK3E
 	/*
Signature · 1 line
1.8.3.1
#1 lore raw

On Tue 05-04-16 12:05:47, Sukadev Bhattiprolu wrote:
[...]

Quoted 5 lines
arch/powerpc/mm/hugetlbpage.c +0 -0
index d991b9e..081f679 100644

arch/powerpc/mm/hugetlbpage.c static int __hugepte_alloc(struct mm_struct *mm, hugepd_t *hpdp, · line 90

Quoted 10 lines
arch/powerpc/mm/hugetlbpage.c +7 -0
@@ -81,6 +81,13 @@ static int __hugepte_alloc(struct mm_struct *mm, hugepd_t *hpdp,
 	if (! new)
 		return -ENOMEM;
 
+	/*
+	 * Make sure other cpus find the hugepd set only after a
+	 * properly initialized page table is visible to them.
+	 * For more details look for comment in __pte_alloc().
+	 */
+	smp_wmb();
+

what is the pairing memory barrier?

Quoted 5 lines
arch/powerpc/mm/hugetlbpage.c +0 -0
 	spin_lock(&mm->page_table_lock);
 #ifdef CONFIG_PPC_FSL_BOOK3E
 	/*
Signature · 1 line
1.8.3.1
Signature · 8 lines
Michal Hocko
SUSE Labs

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to [email protected].  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"[email protected]"> [email protected] </a>
#2 lore raw
Quoted 11 lines · Michal Hocko

[ text/plain ]
On Tue 05-04-16 12:05:47, Sukadev Bhattiprolu wrote:
[...]

Quoted 8 lines
arch/powerpc/mm/hugetlbpage.c +0 -0
index d991b9e..081f679 100644
@@ -81,6 +81,13 @@ static int __hugepte_alloc(struct mm_struct *mm, hugepd_t *hpdp,
 	if (! new)
 		return -ENOMEM;
 
Quoted 13 lines

arch/powerpc/mm/hugetlbpage.c static int __hugepte_alloc(struct mm_struct *mm, hugepd_t *hpdp, · line 90

Quoted 7 lines
arch/powerpc/mm/hugetlbpage.c +7 -0
+	/*
+	 * Make sure other cpus find the hugepd set only after a
+	 * properly initialized page table is visible to them.
+	 * For more details look for comment in __pte_alloc().
+	 */
+	smp_wmb();
+

what is the pairing memory barrier?

Quoted 3 lines
arch/powerpc/mm/hugetlbpage.c +0 -0
 	spin_lock(&mm->page_table_lock);
 #ifdef CONFIG_PPC_FSL_BOOK3E
 	/*

This is documented in __pte_alloc(). I didn't want to repeat the same here.

/*
* Ensure all pte setup (eg. pte page lock and page clearing) are
* visible before the pte is made visible to other CPUs by being
* put into page tables.
*
* The other side of the story is the pointer chasing in the page
* table walking code (when walking the page table without locking;
* ie. most of the time). Fortunately, these data accesses consist
* of a chain of data-dependent loads, meaning most CPUs (alpha
* being the notable exception) will already guarantee loads are
* seen in-order. See the alpha page table accessors for the
* smp_read_barrier_depends() barriers in page table walking code. */
smp_wmb(); /* Could be smp_wmb__xxx(before|after)_spin_lock */

-aneesh

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

#3 lore raw
Quoted 30 lines · Aneesh Kumar K.V
Quoted 11 lines · Michal Hocko

[ text/plain ]
On Tue 05-04-16 12:05:47, Sukadev Bhattiprolu wrote:
[...]

Quoted 8 lines
arch/powerpc/mm/hugetlbpage.c +0 -0
index d991b9e..081f679 100644
@@ -81,6 +81,13 @@ static int __hugepte_alloc(struct mm_struct *mm, hugepd_t *hpdp,
 	if (! new)
 		return -ENOMEM;
 
Quoted 13 lines

arch/powerpc/mm/hugetlbpage.c static int __hugepte_alloc(struct mm_struct *mm, hugepd_t *hpdp, · line 90

Quoted 7 lines
arch/powerpc/mm/hugetlbpage.c +7 -0
+	/*
+	 * Make sure other cpus find the hugepd set only after a
+	 * properly initialized page table is visible to them.
+	 * For more details look for comment in __pte_alloc().
+	 */
+	smp_wmb();
+

what is the pairing memory barrier?

Quoted 3 lines
arch/powerpc/mm/hugetlbpage.c +0 -0
 	spin_lock(&mm->page_table_lock);
 #ifdef CONFIG_PPC_FSL_BOOK3E
 	/*

This is documented in __pte_alloc(). I didn't want to repeat the same here.

Quoted 14 lines

/*
* Ensure all pte setup (eg. pte page lock and page clearing) are
* visible before the pte is made visible to other CPUs by being
* put into page tables.
*
* The other side of the story is the pointer chasing in the page
* table walking code (when walking the page table without locking;
* ie. most of the time). Fortunately, these data accesses consist
* of a chain of data-dependent loads, meaning most CPUs (alpha
* being the notable exception) will already guarantee loads are
* seen in-order. See the alpha page table accessors for the
* smp_read_barrier_depends() barriers in page table walking code. */
smp_wmb(); /* Could be smp_wmb__xxx(before|after)_spin_lock */

OK, I have missed the reference to __pte_alloc. My bad!

Signature · 8 lines
Michal Hocko
SUSE Labs

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to [email protected].  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"[email protected]"> [email protected] </a>
#4 lore raw
In reply to Sukadev Bhattiprolu
Quoted 9 lines · Sukadev Bhattiprolu
Quoted 1 line

From f7b73c6b4508fe9b141a43d92be2f9dd7d3c4a58 Mon Sep 17 00:00:00 2001

From: Sukadev Bhattiprolu <[email protected]>
Date: Thu, 24 Mar 2016 02:07:57 -0400
Subject: [PATCH 1/1] powerpc/mm: Add memory barrier in __hugepte_alloc()

__hugepte_alloc() uses kmem_cache_zalloc() to allocate a zeroed PTE and proceeds to use the newly allocated PTE. Add a memory barrier to make sure that the other CPUs see a properly initialized PTE.

Quoted 6 lines

Based on a fix suggested by James Dykman.

Applied to powerpc next, thanks.

https://git.kernel.org/powerpc/c/0eab46be21449f1612791201aa

cheers