Re: [memcg] 0f12156dff: will-it-scale.per_process_ops -33.6% regression

[Jens Axboe]

On 9/7/21 10:39 AM, Linus Torvalds wrote:
> On Tue, Sep 7, 2021 at 8:46 AM Jens Axboe <axboe@xxxxxxxxx> wrote:
>>
>> Are we at all worried about these? There's been a number of them
>> reported, basically for all the accounting enablements that have been
>> done in this merge window.
> 
> We are worried about them. I'm considering reverting several of them
> because I think the problems are
> 
>  (a) big
> 
>  (b) nontrivial
> 
> and the patches clearly weren't ready and people weren't aware of this issue.

I think that is prudent. When I first enabled it for io_uring it was a
bit of a shit show in terms of performance degradations, and some work
had to be done before it could get enabled in a sane fashion.

The accounting needs to be more efficient if we're seeing 30-50%
slowdowns simply by enabling it on a kmem cache.

-- 
Jens Axboe