diff mbox series

mm: memcg: use READ_ONCE()/WRITE_ONCE() to access stock->nr_pages

Message ID 20240501095420.679208-1-leitao@debian.org (mailing list archive)
State New
Headers show
Series mm: memcg: use READ_ONCE()/WRITE_ONCE() to access stock->nr_pages | expand

Commit Message

Breno Leitao May 1, 2024, 9:54 a.m. UTC 
A memcg pointer in the per-cpu stock can be accessed by drain_all_stock()
and consume_stock() in parallel, causing a potential race.

KCSAN shows this data-race clearly in the splat below:

	BUG: KCSAN: data-race in drain_all_stock.part.0 / try_charge_memcg

	write to 0xffff88903f8b0788 of 4 bytes by task 35901 on cpu 2:
	try_charge_memcg (mm/memcontrol.c:2323 mm/memcontrol.c:2746)
	__mem_cgroup_charge (mm/memcontrol.c:7287 mm/memcontrol.c:7301)
	do_anonymous_page (mm/memory.c:1054 mm/memory.c:4375 mm/memory.c:4433)
	__handle_mm_fault (mm/memory.c:3878 mm/memory.c:5300 mm/memory.c:5441)
	handle_mm_fault (mm/memory.c:5606)
	do_user_addr_fault (arch/x86/mm/fault.c:1363)
	exc_page_fault (./arch/x86/include/asm/irqflags.h:37
		        ./arch/x86/include/asm/irqflags.h:72
			arch/x86/mm/fault.c:1513
			arch/x86/mm/fault.c:1563)
	asm_exc_page_fault (./arch/x86/include/asm/idtentry.h:623)

	read to 0xffff88903f8b0788 of 4 bytes by task 287 on cpu 27:
	drain_all_stock.part.0 (mm/memcontrol.c:2433)
	mem_cgroup_css_offline (mm/memcontrol.c:5398 mm/memcontrol.c:5687)
	css_killed_work_fn (kernel/cgroup/cgroup.c:5521 kernel/cgroup/cgroup.c:5794)
	process_one_work (kernel/workqueue.c:3254)
	worker_thread (kernel/workqueue.c:3329 kernel/workqueue.c:3416)
	kthread (kernel/kthread.c:388)
	ret_from_fork (arch/x86/kernel/process.c:147)
	ret_from_fork_asm (arch/x86/entry/entry_64.S:257)

	value changed: 0x00000014 -> 0x00000013

This happens because drain_all_stock() is reading stock->nr_pages, while
consume_stock() might be updating the same address, causing a potential
data-race.

Make the shared addresses bulletproof regarding to reads and writes,
similarly to what stock->cached_objcg and stock->cached.
Annotate all accesses to stock->nr_pages with READ_ONCE()/WRITE_ONCE().

Signed-off-by: Breno Leitao <leitao@debian.org>
---
 mm/memcontrol.c | 24 +++++++++++++++---------
 1 file changed, 15 insertions(+), 9 deletions(-)

Comments

Shakeel Butt May 1, 2024, 8:47 p.m. UTC | #1 
On Wed, May 01, 2024 at 02:54:20AM -0700, Breno Leitao wrote:
> A memcg pointer in the per-cpu stock can be accessed by drain_all_stock()
> and consume_stock() in parallel, causing a potential race.
> 
> KCSAN shows this data-race clearly in the splat below:
> 
> 	BUG: KCSAN: data-race in drain_all_stock.part.0 / try_charge_memcg
> 
> 	write to 0xffff88903f8b0788 of 4 bytes by task 35901 on cpu 2:
> 	try_charge_memcg (mm/memcontrol.c:2323 mm/memcontrol.c:2746)
> 	__mem_cgroup_charge (mm/memcontrol.c:7287 mm/memcontrol.c:7301)
> 	do_anonymous_page (mm/memory.c:1054 mm/memory.c:4375 mm/memory.c:4433)
> 	__handle_mm_fault (mm/memory.c:3878 mm/memory.c:5300 mm/memory.c:5441)
> 	handle_mm_fault (mm/memory.c:5606)
> 	do_user_addr_fault (arch/x86/mm/fault.c:1363)
> 	exc_page_fault (./arch/x86/include/asm/irqflags.h:37
> 		        ./arch/x86/include/asm/irqflags.h:72
> 			arch/x86/mm/fault.c:1513
> 			arch/x86/mm/fault.c:1563)
> 	asm_exc_page_fault (./arch/x86/include/asm/idtentry.h:623)
> 
> 	read to 0xffff88903f8b0788 of 4 bytes by task 287 on cpu 27:
> 	drain_all_stock.part.0 (mm/memcontrol.c:2433)
> 	mem_cgroup_css_offline (mm/memcontrol.c:5398 mm/memcontrol.c:5687)
> 	css_killed_work_fn (kernel/cgroup/cgroup.c:5521 kernel/cgroup/cgroup.c:5794)
> 	process_one_work (kernel/workqueue.c:3254)
> 	worker_thread (kernel/workqueue.c:3329 kernel/workqueue.c:3416)
> 	kthread (kernel/kthread.c:388)
> 	ret_from_fork (arch/x86/kernel/process.c:147)
> 	ret_from_fork_asm (arch/x86/entry/entry_64.S:257)
> 
> 	value changed: 0x00000014 -> 0x00000013
> 
> This happens because drain_all_stock() is reading stock->nr_pages, while
> consume_stock() might be updating the same address, causing a potential
> data-race.
> 
> Make the shared addresses bulletproof regarding to reads and writes,
> similarly to what stock->cached_objcg and stock->cached.
> Annotate all accesses to stock->nr_pages with READ_ONCE()/WRITE_ONCE().
> 
> Signed-off-by: Breno Leitao <leitao@debian.org>

Acked-by: Shakeel Butt <shakeel.butt@linux.dev>
Roman Gushchin May 1, 2024, 11:18 p.m. UTC | #2 
On Wed, May 01, 2024 at 02:54:20AM -0700, Breno Leitao wrote:
> A memcg pointer in the per-cpu stock can be accessed by drain_all_stock()
> and consume_stock() in parallel, causing a potential race.
> 
> KCSAN shows this data-race clearly in the splat below:
> 
> 	BUG: KCSAN: data-race in drain_all_stock.part.0 / try_charge_memcg
> 
> 	write to 0xffff88903f8b0788 of 4 bytes by task 35901 on cpu 2:
> 	try_charge_memcg (mm/memcontrol.c:2323 mm/memcontrol.c:2746)
> 	__mem_cgroup_charge (mm/memcontrol.c:7287 mm/memcontrol.c:7301)
> 	do_anonymous_page (mm/memory.c:1054 mm/memory.c:4375 mm/memory.c:4433)
> 	__handle_mm_fault (mm/memory.c:3878 mm/memory.c:5300 mm/memory.c:5441)
> 	handle_mm_fault (mm/memory.c:5606)
> 	do_user_addr_fault (arch/x86/mm/fault.c:1363)
> 	exc_page_fault (./arch/x86/include/asm/irqflags.h:37
> 		        ./arch/x86/include/asm/irqflags.h:72
> 			arch/x86/mm/fault.c:1513
> 			arch/x86/mm/fault.c:1563)
> 	asm_exc_page_fault (./arch/x86/include/asm/idtentry.h:623)
> 
> 	read to 0xffff88903f8b0788 of 4 bytes by task 287 on cpu 27:
> 	drain_all_stock.part.0 (mm/memcontrol.c:2433)
> 	mem_cgroup_css_offline (mm/memcontrol.c:5398 mm/memcontrol.c:5687)
> 	css_killed_work_fn (kernel/cgroup/cgroup.c:5521 kernel/cgroup/cgroup.c:5794)
> 	process_one_work (kernel/workqueue.c:3254)
> 	worker_thread (kernel/workqueue.c:3329 kernel/workqueue.c:3416)
> 	kthread (kernel/kthread.c:388)
> 	ret_from_fork (arch/x86/kernel/process.c:147)
> 	ret_from_fork_asm (arch/x86/entry/entry_64.S:257)
> 
> 	value changed: 0x00000014 -> 0x00000013
> 
> This happens because drain_all_stock() is reading stock->nr_pages, while
> consume_stock() might be updating the same address, causing a potential
> data-race.

Btw there is an actual data race, which is periodically causing a needless
draining of a per-cpu stock.

> 
> Make the shared addresses bulletproof regarding to reads and writes,
> similarly to what stock->cached_objcg and stock->cached.
> Annotate all accesses to stock->nr_pages with READ_ONCE()/WRITE_ONCE().
> 
> Signed-off-by: Breno Leitao <leitao@debian.org>

Reviewed-by: Roman Gushchin <roman.gushchin@linux.dev>
Michal Hocko May 3, 2024, 11:28 a.m. UTC | #3 
On Wed 01-05-24 02:54:20, Breno Leitao wrote:
> A memcg pointer in the per-cpu stock can be accessed by drain_all_stock()
> and consume_stock() in parallel, causing a potential race.
> 
> KCSAN shows this data-race clearly in the splat below:
> 
> 	BUG: KCSAN: data-race in drain_all_stock.part.0 / try_charge_memcg
> 
> 	write to 0xffff88903f8b0788 of 4 bytes by task 35901 on cpu 2:
> 	try_charge_memcg (mm/memcontrol.c:2323 mm/memcontrol.c:2746)
> 	__mem_cgroup_charge (mm/memcontrol.c:7287 mm/memcontrol.c:7301)
> 	do_anonymous_page (mm/memory.c:1054 mm/memory.c:4375 mm/memory.c:4433)
> 	__handle_mm_fault (mm/memory.c:3878 mm/memory.c:5300 mm/memory.c:5441)
> 	handle_mm_fault (mm/memory.c:5606)
> 	do_user_addr_fault (arch/x86/mm/fault.c:1363)
> 	exc_page_fault (./arch/x86/include/asm/irqflags.h:37
> 		        ./arch/x86/include/asm/irqflags.h:72
> 			arch/x86/mm/fault.c:1513
> 			arch/x86/mm/fault.c:1563)
> 	asm_exc_page_fault (./arch/x86/include/asm/idtentry.h:623)
> 
> 	read to 0xffff88903f8b0788 of 4 bytes by task 287 on cpu 27:
> 	drain_all_stock.part.0 (mm/memcontrol.c:2433)
> 	mem_cgroup_css_offline (mm/memcontrol.c:5398 mm/memcontrol.c:5687)
> 	css_killed_work_fn (kernel/cgroup/cgroup.c:5521 kernel/cgroup/cgroup.c:5794)
> 	process_one_work (kernel/workqueue.c:3254)
> 	worker_thread (kernel/workqueue.c:3329 kernel/workqueue.c:3416)
> 	kthread (kernel/kthread.c:388)
> 	ret_from_fork (arch/x86/kernel/process.c:147)
> 	ret_from_fork_asm (arch/x86/entry/entry_64.S:257)
> 
> 	value changed: 0x00000014 -> 0x00000013
> 
> This happens because drain_all_stock() is reading stock->nr_pages, while
> consume_stock() might be updating the same address, causing a potential
> data-race.
> 
> Make the shared addresses bulletproof regarding to reads and writes,
> similarly to what stock->cached_objcg and stock->cached.
> Annotate all accesses to stock->nr_pages with READ_ONCE()/WRITE_ONCE().
> 
> Signed-off-by: Breno Leitao <leitao@debian.org>

Acked-by: Michal Hocko <mhocko@suse.com>

It is worth mentioning that the race is harmless.

> ---
>  mm/memcontrol.c | 24 +++++++++++++++---------
>  1 file changed, 15 insertions(+), 9 deletions(-)
> 
> diff --git a/mm/memcontrol.c b/mm/memcontrol.c
> index fabce2b50c69..d3befe3b62fa 100644
> --- a/mm/memcontrol.c
> +++ b/mm/memcontrol.c
> @@ -2310,6 +2310,7 @@ static void memcg_account_kmem(struct mem_cgroup *memcg, int nr_pages)
>  static bool consume_stock(struct mem_cgroup *memcg, unsigned int nr_pages)
>  {
>  	struct memcg_stock_pcp *stock;
> +	unsigned int stock_pages;
>  	unsigned long flags;
>  	bool ret = false;
>  
> @@ -2319,8 +2320,9 @@ static bool consume_stock(struct mem_cgroup *memcg, unsigned int nr_pages)
>  	local_lock_irqsave(&memcg_stock.stock_lock, flags);
>  
>  	stock = this_cpu_ptr(&memcg_stock);
> -	if (memcg == READ_ONCE(stock->cached) && stock->nr_pages >= nr_pages) {
> -		stock->nr_pages -= nr_pages;
> +	stock_pages = READ_ONCE(stock->nr_pages);
> +	if (memcg == READ_ONCE(stock->cached) && stock_pages >= nr_pages) {
> +		WRITE_ONCE(stock->nr_pages, stock_pages - nr_pages);
>  		ret = true;
>  	}
>  
> @@ -2334,16 +2336,18 @@ static bool consume_stock(struct mem_cgroup *memcg, unsigned int nr_pages)
>   */
>  static void drain_stock(struct memcg_stock_pcp *stock)
>  {
> +	unsigned int stock_pages = READ_ONCE(stock->nr_pages);
>  	struct mem_cgroup *old = READ_ONCE(stock->cached);
>  
>  	if (!old)
>  		return;
>  
> -	if (stock->nr_pages) {
> -		page_counter_uncharge(&old->memory, stock->nr_pages);
> +	if (stock_pages) {
> +		page_counter_uncharge(&old->memory, stock_pages);
>  		if (do_memsw_account())
> -			page_counter_uncharge(&old->memsw, stock->nr_pages);
> -		stock->nr_pages = 0;
> +			page_counter_uncharge(&old->memsw, stock_pages);
> +
> +		WRITE_ONCE(stock->nr_pages, 0);
>  	}
>  
>  	css_put(&old->css);
> @@ -2380,6 +2384,7 @@ static void drain_local_stock(struct work_struct *dummy)
>  static void __refill_stock(struct mem_cgroup *memcg, unsigned int nr_pages)
>  {
>  	struct memcg_stock_pcp *stock;
> +	unsigned int stock_pages;
>  
>  	stock = this_cpu_ptr(&memcg_stock);
>  	if (READ_ONCE(stock->cached) != memcg) { /* reset if necessary */
> @@ -2387,9 +2392,10 @@ static void __refill_stock(struct mem_cgroup *memcg, unsigned int nr_pages)
>  		css_get(&memcg->css);
>  		WRITE_ONCE(stock->cached, memcg);
>  	}
> -	stock->nr_pages += nr_pages;
> +	stock_pages = READ_ONCE(stock->nr_pages) + nr_pages;
> +	WRITE_ONCE(stock->nr_pages, stock_pages);
>  
> -	if (stock->nr_pages > MEMCG_CHARGE_BATCH)
> +	if (stock_pages > MEMCG_CHARGE_BATCH)
>  		drain_stock(stock);
>  }
>  
> @@ -2428,7 +2434,7 @@ static void drain_all_stock(struct mem_cgroup *root_memcg)
>  
>  		rcu_read_lock();
>  		memcg = READ_ONCE(stock->cached);
> -		if (memcg && stock->nr_pages &&
> +		if (memcg && READ_ONCE(stock->nr_pages) &&
>  		    mem_cgroup_is_descendant(memcg, root_memcg))
>  			flush = true;
>  		else if (obj_stock_flush_required(stock, root_memcg))
> -- 
> 2.43.0
diff mbox series

Patch

diff --git a/mm/memcontrol.c b/mm/memcontrol.c
index fabce2b50c69..d3befe3b62fa 100644
--- a/mm/memcontrol.c
+++ b/mm/memcontrol.c
@@ -2310,6 +2310,7 @@  static void memcg_account_kmem(struct mem_cgroup *memcg, int nr_pages)
 static bool consume_stock(struct mem_cgroup *memcg, unsigned int nr_pages)
 {
 	struct memcg_stock_pcp *stock;
+	unsigned int stock_pages;
 	unsigned long flags;
 	bool ret = false;
 
@@ -2319,8 +2320,9 @@  static bool consume_stock(struct mem_cgroup *memcg, unsigned int nr_pages)
 	local_lock_irqsave(&memcg_stock.stock_lock, flags);
 
 	stock = this_cpu_ptr(&memcg_stock);
-	if (memcg == READ_ONCE(stock->cached) && stock->nr_pages >= nr_pages) {
-		stock->nr_pages -= nr_pages;
+	stock_pages = READ_ONCE(stock->nr_pages);
+	if (memcg == READ_ONCE(stock->cached) && stock_pages >= nr_pages) {
+		WRITE_ONCE(stock->nr_pages, stock_pages - nr_pages);
 		ret = true;
 	}
 
@@ -2334,16 +2336,18 @@  static bool consume_stock(struct mem_cgroup *memcg, unsigned int nr_pages)
  */
 static void drain_stock(struct memcg_stock_pcp *stock)
 {
+	unsigned int stock_pages = READ_ONCE(stock->nr_pages);
 	struct mem_cgroup *old = READ_ONCE(stock->cached);
 
 	if (!old)
 		return;
 
-	if (stock->nr_pages) {
-		page_counter_uncharge(&old->memory, stock->nr_pages);
+	if (stock_pages) {
+		page_counter_uncharge(&old->memory, stock_pages);
 		if (do_memsw_account())
-			page_counter_uncharge(&old->memsw, stock->nr_pages);
-		stock->nr_pages = 0;
+			page_counter_uncharge(&old->memsw, stock_pages);
+
+		WRITE_ONCE(stock->nr_pages, 0);
 	}
 
 	css_put(&old->css);
@@ -2380,6 +2384,7 @@  static void drain_local_stock(struct work_struct *dummy)
 static void __refill_stock(struct mem_cgroup *memcg, unsigned int nr_pages)
 {
 	struct memcg_stock_pcp *stock;
+	unsigned int stock_pages;
 
 	stock = this_cpu_ptr(&memcg_stock);
 	if (READ_ONCE(stock->cached) != memcg) { /* reset if necessary */
@@ -2387,9 +2392,10 @@  static void __refill_stock(struct mem_cgroup *memcg, unsigned int nr_pages)
 		css_get(&memcg->css);
 		WRITE_ONCE(stock->cached, memcg);
 	}
-	stock->nr_pages += nr_pages;
+	stock_pages = READ_ONCE(stock->nr_pages) + nr_pages;
+	WRITE_ONCE(stock->nr_pages, stock_pages);
 
-	if (stock->nr_pages > MEMCG_CHARGE_BATCH)
+	if (stock_pages > MEMCG_CHARGE_BATCH)
 		drain_stock(stock);
 }
 
@@ -2428,7 +2434,7 @@  static void drain_all_stock(struct mem_cgroup *root_memcg)
 
 		rcu_read_lock();
 		memcg = READ_ONCE(stock->cached);
-		if (memcg && stock->nr_pages &&
+		if (memcg && READ_ONCE(stock->nr_pages) &&
 		    mem_cgroup_is_descendant(memcg, root_memcg))
 			flush = true;
 		else if (obj_stock_flush_required(stock, root_memcg))