linux-kernel - Re: [PATCH v4 2/2] io_uring: Add support for napi_busy

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite for Android: free password hash cracker in your pocket

[<prev] [next>] [<thread-prev] [day] [month] [year] [list]

Message-ID: <40be865d-17c0-80a3-e434-73317c5bff70@linux.alibaba.com>
Date:   Thu, 3 Mar 2022 15:12:01 +0800
From:   Hao Xu <haoxu@...ux.alibaba.com>
To:     Olivier Langlois <olivier@...llion01.com>,
        Jens Axboe <axboe@...nel.dk>,
        Pavel Begunkov <asml.silence@...il.com>
Cc:     io-uring <io-uring@...r.kernel.org>,
        linux-kernel <linux-kernel@...r.kernel.org>
Subject: Re: [PATCH v4 2/2] io_uring: Add support for napi_busy_poll


On 3/3/22 06:03, Olivier Langlois wrote:
> On Wed, 2022-03-02 at 14:38 +0800, Hao Xu wrote:
>>>>
>>>> If that is what you suggest, what would this info do for the
>>>> caller?
>>>>
>>>> IMHO, it wouldn't help in any way...
>>> Hmm, I'm not sure, you're probably right based on that ENOMEM here
>>> shouldn't
>>>
>>> fail the arm poll, but we wanna do it, we can do something like
>>> what
>>> we do for
>>                               ^---but if we wanna do it
> My position is that being able to perform busy poll is a nice to have
> feature if the necessary resources are available. If not the request
> will still be handled correctly so nothing special should be done in
> case of mem alloc problem.
Exactly what I meant.
>
> but fair enough, lets wait for Jens and Pavel to chime him if they
> would like to see something to be done here.
Agree.
>
> Beside that, all I need to know is if napi_list needs to be protected
> in __io_sq_thread with regards to io worket threads to start working on
> a v5.

Sorry for the delay, was stuck in other things. We definitely need

lock in this case too. It should be several lines code, super appreciate

if you could add it.


Thanks,

Hao

>
> I'll look into this question too...