informations about boot sequence (Re: F15 - mysql start problem)

[Michał Piotrowski]

Hi,

2011/5/10 Miloslav Trmač <mitr at volny.cz>:
> On Tue, May 10, 2011 at 5:10 PM, Lennart Poettering
> <mzerqung at 0pointer.de> wrote:
>> On Tue, 10.05.11 02:17, Miloslav Trmač (mitr at volny.cz) wrote:
>>
>>> On Tue, May 10, 2011 at 1:33 AM, Lennart Poettering
>>> <mzerqung at 0pointer.de> wrote:
>>> > Countermeasures for the /dev/shm issue? I don't know of any. tmpfs
>>> > doesn't do quota. That's the key problem here.
>>>
>>> mount options, file permissions, SELinux.  Perhaps not something that
>>> you'd want to do on a general-purpose desktop, but quite reasonable
>>> for a single-purpose server.
>>
>> No. mount options, file permissions, SELinux don't allow you to fix the
>> quota issue with /dev/shm.
>
> There is no "quota issue".  There is a DoS threat.

The point is that you can write to /run/user/<login>/ to fill entire
tmpfs space. When tmpfs is full, there is no space for other things
like pid's and lock's. So it is exactly "quota issue".

With /dev/shm/ you can do similar, creative things.

>  Mounting with
> size=X, where X is large enough to accommodate the applications I care
> about, and small enough that the system won't run out of memory and
> swap space, is a countermeasure to the DoS.[1]  Using file permissions
> or SELinux to only allow users I care about to use /dev/shm is a
> countermeasure to DoS by other users.
>
> Quota is also (only) a way to avoid the DoS[2].  Sure, it's the most
> generally applicable one.
>
> Really, all I wanted to do in this subthread is to support the request
> for a release note.
>    Mirek
>
> [1] How many users of Fedora would notice if the /dev/shm tmpfs was
> limited to 512MB by default? And no, I'm _not_ advocating making this
> the default configuration.
>
> [2] ... like many DoS countermeasures, by intentionally adding a
> different, more predictable, kind of DoS, always denying users the
> service of allocating /dev/shm over a certain limit.  Making a system
> ideally DoS-proof is probably impossible (after all there's little
> outside difference between a DoS and an user that legitimately needs
> to use 101% of a resource); nevertheless if users are willing to
> sacrifice capacity or features, they can often get some level of DoS
> resistance.
> --
> devel mailing list
> devel at lists.fedoraproject.org
> https://admin.fedoraproject.org/mailman/listinfo/devel



-- 
Best regards,
Michal

http://eventhorizon.pl/