linux-kernel - Re: [REGRESSION] RLIMIT

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

Date:   Fri, 16 Sep 2016 16:58:52 -0700
From:   Linus Torvalds <torvalds@...ux-foundation.org>
To:     Sam Varshavchik <mrsam@...rier-mta.com>,
        Ingo Molnar <mingo@...nel.org>, Joe Perches <joe@...ches.com>
Cc:     Laura Abbott <labbott@...hat.com>, Brent <fix@...realm.com>,
        Konstantin Khlebnikov <koct9i@...il.com>,
        Andrew Morton <akpm@...ux-foundation.org>,
        Cyrill Gorcunov <gorcunov@...nvz.org>,
        Christian Borntraeger <borntraeger@...ibm.com>,
        "linux-mm@...ck.org" <linux-mm@...ck.org>,
        Linux Kernel Mailing List <linux-kernel@...r.kernel.org>
Subject: Re: [REGRESSION] RLIMIT_DATA crashes named

On Fri, Sep 16, 2016 at 3:30 PM, Sam Varshavchik <mrsam@...rier-mta.com> wrote:
>>
>> Sam, do you end up seeing the kernel warning in your logs if you just
>> go back earlier in the boot?
>
> Yes, I found it.
>
> Sep 10 07:36:29 shorty kernel: mmap: named (1108): VmData 52588544 exceed
> data ulimit 20971520. Update limits or use boot option ignore_rlimit_data.
>
> Now that I know what to search for: this appeared about 300 lines earlier in
> /var/log/messages.

Ok, so that's a pretty strong argument that we shouldn't just warn once.

Maybe the warning happened at bootup, and it is now three months
later, and somebody notices that something doesn't work. It might not
be *critical* (three months without working implies it isn't), but it
sure is silly for the kernel to say "yeah, I already warned you, I'm
not going to tell you why it's not working any more".

So it sounds like if the kernel had just had a that warning be
rate-limited instead of happening only once, there would never have
been any confusion about the RLIMIT_DATA change.

Doing a grep for "pr_warn_once()", I get the feeling that we could
just change the definition of "once" to be "at most once per minute"
and everybody would be happy.

Maybe we could change all the "pr_xyz_once()" to consider "once" to be
a softer "at most once per minute" thing. After all, these things are
*supposed* to be very uncommon to begin with, but when they do happen
we do want the user to be aware of them.

Here's a totally untested patch. What do people say?

                 Linus

View attachment "patch.diff" of type "text/plain" (1579 bytes)