very common kernel modules slow down the boot process

[Stephen Smalley]

On Wed, 2008-04-09 at 11:24 -0400, Dave Jones wrote:
> On Tue, Apr 08, 2008 at 09:47:32AM -0400, Stephen Smalley wrote:
>  > 
>  > On Tue, 2008-04-08 at 09:03 -0400, Dave Jones wrote:
>  > > On Mon, Apr 07, 2008 at 12:25:53PM -0400, Alan Cox wrote:
>  > >  > On Thu, Apr 03, 2008 at 10:04:31AM -0400, Dave Jones wrote:
>  > >  > >  > Do these older/limited machines do anything better now than they did in 
>  > >  > >  > the 2.4 kernel days?
>  > >  > > 
>  > >  > > Good luck trying to get the installer to run on anything less than 512MB these days.
>  > >  > 
>  > >  > Don't bother: either
>  > >  > 
>  > >  > - Use an existing disk image and upgrade it (works in 128MB)
>  > > 
>  > > Part of the selinux-policy-targeted upgrade does something whih
>  > > munches through stupid amounts of memory.  On any box I've tried
>  > > this on with <512MB, the oom killer kicks in, and then I've been
>  > > left with the mess of a half upgraded box, with lots of rpms
>  > > listed twice in the rpmdb.
>  > 
>  > Is this still true?  There were several memory optimizations implemented
>  > in libsemanage and libsepol in time for Fedora 9, so semodule and
>  > semanage should be much less memory hungry than they were in Fedora 8.
> 
> I'm pretty sure I've still seen this happen on rawhide recently.
> I now f8 is definitly still affected, because it blew up last week
> even after some of Dan's recent changes that we hoped would lower
> mem usage. 

The most important optimization wouldn't have shown up until libsemanage
2.0.23.

-- 
Stephen Smalley
National Security Agency