Re: [PATCH 1/6] mm, memcg: Prevent memory.high load/store tearing

[Michal Hocko]

On Thu 12-03-20 17:32:51, Chris Down wrote:
> A mem_cgroup's high attribute can be concurrently set at the same time
> as we are trying to read it -- for example, if we are in
> memory_high_write at the same time as we are trying to do high reclaim.

I assume this is a replace all kinda patch because css_alloc shouldn't
really be a subject to races. I am not sure about css_reset but it
sounds like a safe as well.

That being said I do not object because this cannot be harmful but it
would be nice to mention that in the changelog just in case somebody
wonders about this in future.
 
> Signed-off-by: Chris Down <chris@xxxxxxxxxxxxxx>
> Cc: Andrew Morton <akpm@xxxxxxxxxxxxxxxxxxxx>
> Cc: Johannes Weiner <hannes@xxxxxxxxxxx>
> Cc: Roman Gushchin <guro@xxxxxx>
> Cc: Tejun Heo <tj@xxxxxxxxxx>
> Cc: linux-mm@xxxxxxxxx
> Cc: cgroups@xxxxxxxxxxxxxxx
> Cc: linux-kernel@xxxxxxxxxxxxxxx
> Cc: kernel-team@xxxxxx

Acked-by: Michal Hocko <mhocko@xxxxxxxx>

> ---
>  mm/memcontrol.c | 13 +++++++------
>  1 file changed, 7 insertions(+), 6 deletions(-)
> 
> diff --git a/mm/memcontrol.c b/mm/memcontrol.c
> index 63bb6a2aab81..d32d3c0a16d4 100644
> --- a/mm/memcontrol.c
> +++ b/mm/memcontrol.c
> @@ -2228,7 +2228,7 @@ static void reclaim_high(struct mem_cgroup *memcg,
>  			 gfp_t gfp_mask)
>  {
>  	do {
> -		if (page_counter_read(&memcg->memory) <= memcg->high)
> +		if (page_counter_read(&memcg->memory) <= READ_ONCE(memcg->high))
>  			continue;
>  		memcg_memory_event(memcg, MEMCG_HIGH);
>  		try_to_free_mem_cgroup_pages(memcg, nr_pages, gfp_mask, true);
> @@ -2545,7 +2545,7 @@ static int try_charge(struct mem_cgroup *memcg, gfp_t gfp_mask,
>  	 * reclaim, the cost of mismatch is negligible.
>  	 */
>  	do {
> -		if (page_counter_read(&memcg->memory) > memcg->high) {
> +		if (page_counter_read(&memcg->memory) > READ_ONCE(memcg->high)) {
>  			/* Don't bother a random interrupted task */
>  			if (in_interrupt()) {
>  				schedule_work(&memcg->high_work);
> @@ -4257,7 +4257,8 @@ void mem_cgroup_wb_stats(struct bdi_writeback *wb, unsigned long *pfilepages,
>  	*pheadroom = PAGE_COUNTER_MAX;
>  
>  	while ((parent = parent_mem_cgroup(memcg))) {
> -		unsigned long ceiling = min(memcg->memory.max, memcg->high);
> +		unsigned long ceiling = min(memcg->memory.max,
> +					    READ_ONCE(memcg->high));
>  		unsigned long used = page_counter_read(&memcg->memory);
>  
>  		*pheadroom = min(*pheadroom, ceiling - min(ceiling, used));
> @@ -4978,7 +4979,7 @@ mem_cgroup_css_alloc(struct cgroup_subsys_state *parent_css)
>  	if (!memcg)
>  		return ERR_PTR(error);
>  
> -	memcg->high = PAGE_COUNTER_MAX;
> +	WRITE_ONCE(memcg->high, PAGE_COUNTER_MAX);
>  	memcg->soft_limit = PAGE_COUNTER_MAX;
>  	if (parent) {
>  		memcg->swappiness = mem_cgroup_swappiness(parent);
> @@ -5131,7 +5132,7 @@ static void mem_cgroup_css_reset(struct cgroup_subsys_state *css)
>  	page_counter_set_max(&memcg->tcpmem, PAGE_COUNTER_MAX);
>  	page_counter_set_min(&memcg->memory, 0);
>  	page_counter_set_low(&memcg->memory, 0);
> -	memcg->high = PAGE_COUNTER_MAX;
> +	WRITE_ONCE(memcg->high, PAGE_COUNTER_MAX);
>  	memcg->soft_limit = PAGE_COUNTER_MAX;
>  	memcg_wb_domain_size_changed(memcg);
>  }
> @@ -5947,7 +5948,7 @@ static ssize_t memory_high_write(struct kernfs_open_file *of,
>  	if (err)
>  		return err;
>  
> -	memcg->high = high;
> +	WRITE_ONCE(memcg->high, high);
>  
>  	for (;;) {
>  		unsigned long nr_pages = page_counter_read(&memcg->memory);
> -- 
> 2.25.1
> 

-- 
Michal Hocko
SUSE Labs