Regression in kernel-3.4.1-1.fc17.x86_64

Josh Boyer jwboyer at redhat.com
Tue Jun 5 17:44:17 UTC 2012


On Tue, Jun 05, 2012 at 11:47:52AM -0500, Bruno Wolff III wrote:
> On Tue, Jun 05, 2012 at 12:43:15 -0400,
>   Josh Boyer <jwboyer at redhat.com> wrote:
> >On Tue, Jun 05, 2012 at 10:56:44AM -0500, Bruno Wolff III wrote:
> >>kernel-3.4.1-1.fc17.x86_64 hasn't had a bodhi request yet, so I
> >>can't give it negative karma yet. On the one machine where I tried
> >>to login with a graphical login (gdm) I was not able to enter text
> >>using my USB keyboard. Rebooting into 3.4.0-1.fc17.x86_64 got things
> >>working again.
> >
> >That doesn't make much sense.  3.4.0-1 actually contains almost all of
> >the patches found in 3.4.1 already (if not all of them).  Is that the
> >only package you updated at that time?  If not, the initramfs for 3.4.1
> >might have picked up some userspace change.
> >
> >josh
> 
> I think some other packages were updated between the last reboot
> before updating to 3.4.1 and that reboot. However nothing was
> updated between testing 3.4.1 and falling back to 3.4.0.
> 
> However, I only tried 3.4.1 once, so it's certainly possible that it
> was some other intermittant problem that just happened to fail on
> that boot.

OK.  I ran through the patches that were included in 3.4.0-1 and what
wound up in 3.4.1.  The only additional patch in 3.4.1 is:

commit 577bd3ef6eed18ea3c866b2f25b14aa5e139b41f
Author: Bjørn Mork <bjorn at mork.no>
Date:   Wed May 9 13:53:22 2012 +0200

    USB: cdc-wdm: cannot use dev_printk when device is gone
    
    commit 6b0b79d38806481c1c8fffa7c5842f3c83679a42 upstream.
    
    We cannot dereference a removed USB interface for
    dev_printk. Use pr_debug instead where necessary.


which seems unrelated to login issues.

3.4.0-1 had 3 additional intel wireless related fixes, but again
unrelated to login issues.

I suspect something in the initramfs changed if it's persistent.  You
might try rebuilding the initramfs (save the current one first) for
3.4.0-1 with whatever you have for current userspace on your box and see
if it also fails.

josh


More information about the kernel mailing list