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: Windows password security audit tool. GUI, reports in PDF.
[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Message-ID: <73673ab9-9ec7-611f-b1f0-5394cd7a172c@kernel.dk>
Date:   Thu, 3 Jun 2021 09:46:17 -0600
From:   Jens Axboe <axboe@...nel.dk>
To:     Justin Forbes <jmforbes@...uxtx.org>,
        Justin Forbes <jforbes@...hat.com>
Cc:     asml.silence@...il.com, io-uring@...r.kernel.org,
        Linux Kernel Mailing List <linux-kernel@...r.kernel.org>
Subject: Re: [PATCH] io_uring: Remove CONFIG_EXPERT

On 6/3/21 9:43 AM, Justin Forbes wrote:
> On Thu, May 27, 2021 at 11:01 AM Justin Forbes <jforbes@...hat.com> wrote:
>>
>> On Thu, May 27, 2021 at 9:19 AM Jens Axboe <axboe@...nel.dk> wrote:
>>>
>>> On 5/27/21 8:12 AM, Justin Forbes wrote:
>>>> On Thu, May 27, 2021 at 8:43 AM Jens Axboe <axboe@...nel.dk> wrote:
>>>>>
>>>>> On 5/26/21 4:34 PM, Justin M. Forbes wrote:
>>>>>> While IO_URING has been in fairly heavy development, it is hidden behind
>>>>>> CONFIG_EXPERT with a default of on.  It has been long enough now that I
>>>>>> think we should remove EXPERT and allow users and distros to decide how
>>>>>> they want this config option set without jumping through hoops.
>>>>>
>>>>> The whole point of EXPERT is to ensure that it doesn't get turned off
>>>>> "by accident". It's a core feature, and something that more and more
>>>>> apps or libraries are relying on. It's not something I intended to ever
>>>>> go away, just like it would never go away for eg futex or epoll support.
>>>>>
>>>>
>>>> I am not arguing with that, I don't expect it will go away. I
>>>> certainly do not have an issue with it defaulting to on, and I didn't
>>>> even submit this with intention to turn it off for default Fedora. I
>>>> do think that there are cases where people might not wish it turned on
>>>> at this point in time. Hiding it behind EXPERT makes it much more
>>>> difficult than it needs to be.  There are plenty of config options
>>>> that are largely expected default and not hidden behind EXPERT.
>>>
>>> Right there are, but not really core kernel features like the ones
>>> I mentioned. Hence my argument for why it's correct as-is and I
>>> don't think it should be changed.
>>>
>>
>> Honestly, this is fair, and I understand your concerns behind it. I
>> think my real issue is that there is no simple way to override one
>> EXPERT setting without having to set them all.  It would be nice if
>> expert were a "visible if" menu, setting defaults if not selected,
>> which allows direct override with a config file. Perhaps I will try to
>> fix this in kbuild.
>>
> 
> So it turns out that untangling this in kbuild is very difficult
> without getting very unexpected results. Given the audit and security
> discussions around io_uring lately, I am inclined to believe the
> proper action is still to remove 'if EXPERT'.

I'm still going to disagree with that. In terms of security, io_uring
is in a much better spot than it was in the past. Arguably it should have
been easier to turn off when it went in, but that ship has sailed a long
time ago and the need now isn't really there imho. The lack of audit
isn't new and is something that'll be solved for the next release.

-- 
Jens Axboe

Powered by blists - more mailing lists

Powered by Openwall GNU/*/Linux Powered by OpenVZ