[3626] in Release_7.7_team

home help back first fref pref prev next nref lref last post

Re: 9.1.19 blocking on Linux kernel issue

daemon@ATHENA.MIT.EDU (Greg Hudson)
Sat Dec 14 11:04:44 2002

From: Greg Hudson <ghudson@MIT.EDU>
To: Greg Hudson <ghudson@mit.edu>
Cc: release-team@mit.edu
In-Reply-To: <200212102150.QAA00514@error-messages.mit.edu>
Content-Type: text/plain
Content-Transfer-Encoding: 7bit
Date: 14 Dec 2002 11:04:40 -0500
Message-Id: <1039881880.22111.14.camel@error-messages.mit.edu>
Mime-Version: 1.0

A status update:

On Tue, 2002-12-10 at 16:50, Greg Hudson wrote:
> Some conceivable ways out:

>   * If we could identify a source patch for the problem

This avenue has not been going well.  There is one bug in Red Hat
bugzilla which is similar
(https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=76237), and it
hasn't been analyzed or fixed by Red Hat.  And the problem isn't
sufficiently reproduceable that we could just fire up a kernel debugger
and go to it; it seems to happen once almost every 9.1.18 machine, but
after that it's very sporadic.

>   * We could (temporarily) install a Red Hat kernel RPM which contains
>     a modified /sbin/mkinitrd, which produces the initrd in some other
>     way.  The "some other way" is hard to identify

There are two options: we could build a cramfs image (cramfs is a simple
read-only filesystem; a cramfs image can be created by a simple
user-mode tool without any kernel help), or we can use a ramdisk device
instead of a loop device.  Unfortunately, Red Hat builds cramfs as a
module in their default kernel config, rendering it useless for an
initrd containing bootstrap modules.  Using ramdisk devices is a little
dicey--there's no way to check if a ramdisk device is in use, although
you can check if one is mounted--but as a result, I think no one else
uses them for anything/  So I don't think we should encounter any
problems.

I have created a modified initrd script which uses a ramdisk device, and
verified that it works on one virgin 9.1.18 machine.  More testing is
required.  I will work on the update script magic to ensure that the
modified initrd is used by the kernel update.


home help back first fref pref prev next nref lref last post