rainbow and pam

[victor]

Ok; My suggestion would then be for us to test allowing
RTPRIO to 99 (the max value), if possible, so we can
test its impact. Then it will be possible to test lower
priorities to see what would be a reasonable limit.

There has also been a question of testing stronger
RT kernel patches in future releases. If so, the whole
thing will need to be reconsidered then.

Our problem at the moment is that the current kernel
has taken a step backwards in RT performance, which
we are trying to address. Maybe this will not be the case
in the future.

In any case, thanks for helping us sort this one out,
I am actually very pleased that we can discuss the XO's
RT performance. In fact, if we tune this right, we will
place the OS (and the laptop) in a position of advantage 
against its competitors in relation to media applications.

Victor
----- Original Message ----- 
From: "Michael Stone" <michael at laptop.org>
To: "victor" <Victor.Lazzarini at nuim.ie>
Cc: "Jim Gettys" <jg at laptop.org>; <devel at lists.laptop.org>
Sent: Saturday, August 16, 2008 10:39 PM
Subject: Re: rainbow and pam


> On Sat, Aug 16, 2008 at 10:15:08PM +0100, victor wrote:
>> Aren't these priorities the same ones set in /etc/security/limits.conf?
>> Or are they set by other means?
> 
> /etc/security/limits.conf is simply one vehicle (specifically, the one
> used by PAM) for getting a uid-0 process (hence a process w/
> CAP_SYS_RESOURCE) like login or su to call setrlimit() before running
> some user-supplied code. Rainbow can accomplish this directly. 
> 
> Michael
> 
> (I am interested in arguments as to whether Rainbow should be made to
> interact with PAM and the 'login'/'session'/'terminal' concept cluster,
> but I haven't yet gotten around to detailed investigation.)