[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: Bug#241497: Critical bug still not addressed: upgrade-i386



On Wed, Aug 11, 2004 at 12:04:30AM -0400, Andres Salomon wrote:
> On Tue, 10 Aug 2004 13:19:22 -0700, Steve Langasek wrote:
> > The upgrade-only kernels would be intended solely as an intermediate step
> > during the upgrade process; users should be encouraged to install kernels
> > from the main archive as well as part of the upgrade process. As such, we
> > can probably dispense with security support for the upgrade-i386 kernel
> > packages if necessary.

> If we force it to be intermediate, that's fine (ie, we force 2.4.24 or
> something upon the user, and then force them to upgrade to the latest 2.4
> kernel during the rest of the sarge upgrade).  However, merely encouraging
> a user means we'll probably end up w/ a bunch of users who keep running
> the intermediate kernels.  I wouldn't feel comfortable installing a kernel
> w/ known security holes on a user's machine; especially during a
> woody->sarge upgrade, as there's a good chance the machine in question is
> in a production environment.  So the question is, can we force a kernel
> upgrade from this intermediate kernel?

One way to force the removal of the intermediate kernel is to make
something else in sarge conflict with it, but I can't think of any good
candidates.  (Kernels are notoriously difficult in this respect.)

Given that no 386 users are going to be adding the necessary line to
their apt sources without having read the 386-specific upgrade
information, and given that this kernel won't be vulnerable to any
exploits that don't also affect our latest woody kernels (and anyone
running a newer hand-built kernel doesn't need our transition kernels),
I think it's sufficient if we use big red letters in the documentation
when telling users that these kernels won't have security support.

In practice, it probably makes sense to continue security support for
that directory as long as the security team does for woody in general,
however long they decide that to be.

> Or even better, can we simply do something similar to what the udeb
> folks do, and keep this intermediate kernel using the image and
> modules from the latest proper 2.4 -386 package that's in sarge?

IIRC, the problem with this approach was that newer kernel packages
depend on newer tools than what's available in woody, which in turn run
up against the C++ ABI change that prompted the whole need for the
kernel emulation patch.

-- 
Steve Langasek
postmodern programmer

Attachment: signature.asc
Description: Digital signature


Reply to: