Re: [RFC PATCH v2 1/3] misc_cgroup: add support for nofile limit

[brookxu <brookxu.cn-Re5JQEeQqe8AvxtiuMwx3w@public.gmane.org>]

Thanks for your time.

Tejun Heo wrote on 2021/7/27 5:27:
> On Thu, Jul 22, 2021 at 11:20:17PM +0800, brookxu wrote:
>> From: Chunguang Xu <brookxu-1Nz4purKYjRBDgjK7y7TUQ@public.gmane.org>
>>
>> Since the global open files are limited, in order to avoid the
>> abnormal behavior of some containers from generating too many
>> files, causing other containers to be unavailable, we need to
>> limit the open files of some containers.
>>
>> v2: fix compile error while CONFIG_CGROUP_MISC not set.
>>
>> Signed-off-by: Chunguang Xu <brookxu-1Nz4purKYjRBDgjK7y7TUQ@public.gmane.org>
>> Reported-by: kernel test robot <lkp-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>
> 
> This is different from pid in that there's no actual limit on how many open
> files there can be in the system other than the total amount of available
> memory. I don't see why this would need a separate limit outside of memory
> control. A couple machines I looked at all have file-max at LONG_MAX by
> default too.

According to files_maxfiles_init(), we only allow about 10% of free memory to
create filps, and each filp occupies about 1K of cache. In this way, on a 16G
memory machine, the maximum usable filp is about 1,604,644. In general
scenarios, this may not be a big problem, but if the task is abnormal, it will
very likely become a bottleneck and affect other modules. 

> Thanks.
>