From: Chris Wright <firstname.lastname@example.org>
To: Arjan van de Ven <email@example.com>
Cc: utz lehmann <firstname.lastname@example.org>,
Chris Wright <email@example.com>
Subject: Re: [PATCH] scheduling priorities with rlimit
Date: Sun, 9 Jan 2005 12:34:41 -0800 [thread overview]
Message-ID: <20050109123441.O469@build.pdx.osdl.net> (raw)
In-Reply-To: <firstname.lastname@example.org>; from email@example.com on Sun, Jan 09, 2005 at 08:06:37PM +0100
* Arjan van de Ven (firstname.lastname@example.org) wrote:
> On Sun, 2005-01-09 at 18:15 +0100, utz lehmann wrote:
> > Hi
> > I really like the idea of controlling the maximum settable scheduling
> > priorities via rlimit. See the Realtime LSM thread. I want to give users
> > the right to raise the priority of previously niced jobs.
> > I have modified Chris Wright's patch (against 2.6.10):
> > (http://marc.theaimsgroup.com/?l=linux-kernel&m=110513793228776&w=2)
> > - allow always to increase nice levels (lower priority).
> > - set the default for RLIMIT_PRIO to 0.
> > - add the other architectures.
> > With this the default is compatible with the old behavior.
> > With RLIMIT_PRIO > 0 a user is able to raise the priority up to the
> > value. 0-39 for nice levels 19 .. -20, 40-139 for realtime priorities
> > (0 .. 99).
> this is a bit of an awkward interface don't you think?
Yes it is. But I didn't think of a better one.
> I much rather have the rlimit match the exact nice values we communicate
> to userspace elsewhere, both to be consistent and to not expose
> scheduler internals to userpsace.
The problem is the numbers are inconsistent between user interfaces already.
RT priorities are [0, 99], nice vaules are [-20, 19]. Perhaps it'd be
simpler to break it down to just three values for the rlimit.
0: Same as now, raise nice value only.
1: Can lower nice value.
2: Can set RT policy (this includes any priority [1, 99], or optionally
max out at something lower than 99, reserving full CAP_SYS_NICE to 99).
Each level inherits the permissions of the lower level, and none of them
allow the CAP_SYS_NICE ability to affect processes other than your own.
> Also I like the idea of allowing sysadmins to make certain users/groups
> nice levels 5 and higher (think a university machine that makes all
> students nice 5 and higher only, while giving staff 0 and higher, and
> the sysadmin -5 and higher ;)
This is a separate issue. It's about setting the default during login
which can be done with setpriority (still could be done via pam).
Linux Security Modules http://lsm.immunix.org http://lsm.bkbits.net
next prev parent reply other threads:[~2005-01-09 20:35 UTC|newest]
Thread overview: 6+ messages / expand[flat|nested] mbox.gz Atom feed top
2005-01-09 17:15 [PATCH] scheduling priorities with rlimit utz lehmann
2005-01-09 19:06 ` Arjan van de Ven
2005-01-09 19:23 ` utz lehmann
2005-01-09 20:34 ` Chris Wright [this message]
2005-01-10 18:01 ` utz lehmann
2005-01-10 15:15 ` Horst von Brand
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).