Re: [PATCH v3] mm/vmscan: add a fatal signals check in drop_slab_node

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



On Tue 15-09-20 19:40:01, zangchunxin@xxxxxxxxxxxxx wrote:
> From: Chunxin Zang <zangchunxin@xxxxxxxxxxxxx>
> 
> On our server, there are about 10k memcg in one machine. They use memory
> very frequently. We have observed that drop_caches can take a
> considerable amount of time, and can't stop it.
> 
> There are two reasons:
> 1. There is somebody constantly generating more objects to reclaim
>    on drop_caches, result the 'freed' always bigger than 10.
> 2. The process has no chance to process signals.
> 
> We can get the following info through 'ps':
> 
>   root:~# ps -aux | grep drop
>   root  357956 ... R    Aug25 21119854:55 echo 3 > /proc/sys/vm/drop_caches
>   root 1771385 ... R    Aug16 21146421:17 echo 3 > /proc/sys/vm/drop_caches
>   root 1986319 ... R    18:56 117:27 echo 3 > /proc/sys/vm/drop_caches
>   root 2002148 ... R    Aug24 5720:39 echo 3 > /proc/sys/vm/drop_caches
>   root 2564666 ... R    18:59 113:58 echo 3 > /proc/sys/vm/drop_caches
>   root 2639347 ... R    Sep03 2383:39 echo 3 > /proc/sys/vm/drop_caches
>   root 3904747 ... R    03:35 993:31 echo 3 > /proc/sys/vm/drop_caches
>   root 4016780 ... R    Aug21 7882:18 echo 3 > /proc/sys/vm/drop_caches
> 
> Use bpftrace follow 'freed' value in drop_slab_node:
> 
>   root:~# bpftrace -e 'kprobe:drop_slab_node+70 {@ret=hist(reg("bp")); }'
>   Attaching 1 probe...
>   ^B^C
> 
>   @ret:
>   [64, 128)        1 |                                                    |
>   [128, 256)      28 |                                                    |
>   [256, 512)     107 |@                                                   |
>   [512, 1K)      298 |@@@                                                 |
>   [1K, 2K)       613 |@@@@@@@                                             |
>   [2K, 4K)      4435 |@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@|
>   [4K, 8K)       442 |@@@@@                                               |
>   [8K, 16K)      299 |@@@                                                 |
>   [16K, 32K)     100 |@                                                   |
>   [32K, 64K)     139 |@                                                   |
>   [64K, 128K)     56 |                                                    |
>   [128K, 256K)    26 |                                                    |
>   [256K, 512K)     2 |                                                    |

I am not sure this is very helpful for this patch but whatever.

> We need one path to stop the process.

I would use the following instead
"
Add a bail out on the fatal signals in the main loop so that the
operation can be terminated by userspace.
"

> 
> Signed-off-by: Chunxin Zang <zangchunxin@xxxxxxxxxxxxx>
> Signed-off-by: Muchun Song <songmuchun@xxxxxxxxxxxxx>

Acked-by: Michal Hocko <mhocko@xxxxxxxx>

> ---
> 
> 	changelogs in v3: 
> 	1) update the description of the patch.
> 		v2 named: mm/vmscan: fix infinite loop in drop_slab_node
> 
> 	changelogs in v2: 
> 	1) via check fatal signal break loop.
> 
>  mm/vmscan.c | 3 +++
>  1 file changed, 3 insertions(+)
> 
> diff --git a/mm/vmscan.c b/mm/vmscan.c
> index b6d84326bdf2..6b2b5d420510 100644
> --- a/mm/vmscan.c
> +++ b/mm/vmscan.c
> @@ -704,6 +704,9 @@ void drop_slab_node(int nid)
>  	do {
>  		struct mem_cgroup *memcg = NULL;
>  
> +		if (signal_pending(current))
> +			return;
> +
>  		freed = 0;
>  		memcg = mem_cgroup_iter(NULL, NULL, NULL);
>  		do {
> -- 
> 2.11.0
> 

-- 
Michal Hocko
SUSE Labs




[Index of Archives]     [Linux ARM Kernel]     [Linux ARM]     [Linux Omap]     [Fedora ARM]     [IETF Annouce]     [Bugtraq]     [Linux OMAP]     [Linux MIPS]     [eCos]     [Asterisk Internet PBX]     [Linux API]

  Powered by Linux