2011/5/9 Lennart Poettering <mzerq...@0pointer.de>:
> On Mon, 09.05.11 23:31, Karel Zak (k...@redhat.com) wrote:
>
>>
>> On Sun, Apr 03, 2011 at 11:45:51PM +0200, Kay Sievers wrote:
>> > 2011/4/3 Lennart Poettering <mzerq...@0pointer.de>:
>> > > On Sun, 03.04.11 23:28, Michał Piotrowski (mkkp...@gmail.com) wrote:
>> > >
>> > >> > But for /dev/shm I see no quick fix... do you?
>> > >>
>> > >> Unfortunately not. No one foresaw that quota support on tmpfs will
>> > >> someday be useful :)
>> > >>
>> > >> >
>> > >> > I think we should fix either both or should wait for the proper fix by
>> > >> > the kernel.
>> > >>
>> > >> Can you temporarily fix one?
>> > >
>> > > Well, of course we could.
>> > >
>> > > But, think about it, what does this help? The vulnerability doesn't go
>> > > away by doing this, and we'd have a temporary hack in there, that we'd
>> > > have to remove later on again.
>> >
>> > Systems who might run into problems with /dev/shm, can just add limits
>> > to /etc/fstab, and systemd will re-mount it and apply them.
>> >
>> > There should really be a _proper_ solution some day, be it quota or
>> > something else. We have way too many /tmp-like dirs, where users can
>> > just leave their crap behind and cause problems. This is really
>> > nothing new with systemd.
>>
>>  Wouldn't be possible to use namespaces (pam_namespace ?) and after
>>  user login create any private tmpfs (with explicitly defined size)?
>>
>>  This allows to use the same path (e.g. /run/user) for all users, make
>>  the content of the directory invisible for other users and
>>  effectively control resources. All this is supported by kernel ;-)
>
> Well that's difficult for /dev/shm, since the way it is defined it is
> system global.
>
> And for /run/users it has been proposed to mount a tmpfs into the
> per-user dirs, and we could do this pretty easily. However, since this
> only fixed part of the problem and having gazillions of tmpfs lying
> around, one for each logged in user we thought we better wait for a
> proper solution which is quota, or something similar.

I secretly hope that someone is already working on this
https://bugzilla.redhat.com/show_bug.cgi?id=693253

F15 users need only to know that there is an simple workaround for services DoS.

>
> Lennart
>
> --
> Lennart Poettering - Red Hat, Inc.
> _______________________________________________
> systemd-devel mailing list
> systemd-devel@lists.freedesktop.org
> http://lists.freedesktop.org/mailman/listinfo/systemd-devel
>



-- 
Best regards,
Michal

http://eventhorizon.pl/
_______________________________________________
systemd-devel mailing list
systemd-devel@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/systemd-devel

Reply via email to