Re: random wedges with 2.6.25-rc*

Re: random wedges with 2.6.25-rc*

[Jeff Garzik]

Pierre Ossman wrote:
> Somewhere post 2.6.24, the kernel started getting very temperamentful. I experience random hangs and wedges very often.
> 
> Primarily, the udev startup locks up. If I abort it, it just locks up on more or less every action after that. Some quick debugging showed that I had a whole bunch of modprobe processes sitting around. The new CONFIG_USB_ANNOUNCE_NEW_DEVICES setting makes every startup of udev lock up consistently on at least two machines. It does not seem to be the root of the problem as disabling the option just makes the bug very unlikely.
> 
> I've also seen a couple of lockups in X. Might be a different expression of the same bug, but the effect here is a complete wedge of the user interface. It seems the rest of the machine also locks up, but I haven't confirmed this.
> 
> Anyone else seeing these problems? Someone should as I've seen the problem on both a Lenovo and a HP laptop here.


I'm definitely seeing lockups here too.  2.6.24 is fine, 2.6.24-rc1 or 
-rc2 locks up a minute or so after a successful boot (either to console 
or to X).

Does the following shed any light?

http://lkml.org/lkml/2008/2/17/78

	Jeff

random wedges with 2.6.25-rc*

[Pierre Ossman]

Somewhere post 2.6.24, the kernel started getting very temperamentful. I experience random hangs and wedges very often.

Primarily, the udev startup locks up. If I abort it, it just locks up on more or less every action after that. Some quick debugging showed that I had a whole bunch of modprobe processes sitting around. The new CONFIG_USB_ANNOUNCE_NEW_DEVICES setting makes every startup of udev lock up consistently on at least two machines. It does not seem to be the root of the problem as disabling the option just makes the bug very unlikely.

I've also seen a couple of lockups in X. Might be a different expression of the same bug, but the effect here is a complete wedge of the user interface. It seems the rest of the machine also locks up, but I haven't confirmed this.

Anyone else seeing these problems? Someone should as I've seen the problem on both a Lenovo and a HP laptop here.

Rgds
-- 
     -- Pierre Ossman

  Linux kernel, MMC maintainer        http://www.kernel.org
  PulseAudio, core developer          http://pulseaudio.org
  rdesktop, core developer          http://www.rdesktop.org

Re: random wedges with 2.6.25-rc*

[Alistair John Strachan]

On Tuesday 19 February 2008 18:10:35 Pierre Ossman wrote:
> Primarily, the udev startup locks up. If I abort it, it just locks up on
> more or less every action after that. Some quick debugging showed that I
> had a whole bunch of modprobe processes sitting around. The new
> CONFIG_USB_ANNOUNCE_NEW_DEVICES setting makes every startup of udev lock up
> consistently on at least two machines. It does not seem to be the root of
> the problem as disabling the option just makes the bug very unlikely.

You're not alone, I'm seeing these intermittent hangs at init time on Debian 
unstable at:

"Waiting for /dev to be fully populated" (udevsettle)

"Loading kernel modules" (which I assume is modprobe)

But, I'm not seeing any hangs in X11 (yet).

-- 
Cheers,
Alistair.

137/1 Warrender Park Road, Edinburgh, UK.

Re: random wedges with 2.6.25-rc*

[David Rees]

On Feb 19, 2008 10:10 AM, Pierre Ossman <drzeus-list@drzeus.cx> wrote:
> Somewhere post 2.6.24, the kernel started getting very
> temperamentful. I experience random hangs and wedges very often.

FWIW, on my Fedora 8 Dell Vostro 1000 notebook, ever since updating to
kernel 2.6.23.15-137 I am getting frequent udev startup lockups on
boot.  Resetting and rebooting will get things past udev startup some
of the time, so it doesn't always hang. I'd guess it hangs about half
the time. Once past udev startup, everything works fine, no hangs. I
haven't seen this on a half dozen other desktops/servers running
Fedora 8.

Aside from that, everything else works fine. The previous kernel was
2.6.23.14-115, so I suspect that if these were related, the changes
were backported into the 2.6.23 stable tree in 2.6.23.15?

You can view the reported list of differences between the two kernels
in the package announcement:
https://www.redhat.com/archives/fedora-package-announce/2008-February/msg00255.html

-Dave

Re: random wedges with 2.6.25-rc*

[Pierre Ossman]

On Tue, 19 Feb 2008 12:55:13 -0500
Jeff Garzik <jeff@garzik.org> wrote:

> Pierre Ossman wrote:
> > 
> > Anyone else seeing these problems? Someone should as I've seen the problem on both a Lenovo and a HP laptop here.
> 
> 
> I'm definitely seeing lockups here too.  2.6.24 is fine, 2.6.24-rc1 or 
> -rc2 locks up a minute or so after a successful boot (either to console 
> or to X).
> 
> Does the following shed any light?
> 
> http://lkml.org/lkml/2008/2/17/78
> 

That would explain the lockups in X as the network has been active at those times. But it would be odd if it affects the startup of udev (most likely module loading).

Rgds
-- 
     -- Pierre Ossman

  Linux kernel, MMC maintainer        http://www.kernel.org
  PulseAudio, core developer          http://pulseaudio.org
  rdesktop, core developer          http://www.rdesktop.org