auto-reboot "broken" in 2.6.23.1 (from 2.6.22.12)

Previous thread: x86_64: Make sparsemem/vmemmap the default memory model by Christoph Lameter on Monday, November 12, 2007 - 4:52 pm. (14 messages)

Next thread: [PATCH 0/5][RFC] Physical PCI slot objects by Alex Chiang on Monday, November 12, 2007 - 5:08 pm. (69 messages)
From: Linda Walsh
Date: Monday, November 12, 2007 - 4:36 pm

Hello,
I upgraded the kernel on a multi-cpu Dell workstation (690) to
2.6.23.1.  It had been running at 2.6.20.  I moved to the new kernel via 
"make oldconfig" -- choosing defaults.  That didn't work.

To narrow things down I tried it under 2.6.21.1 (made
oldconfig from working 2.6.20/proc/config.gz).  This worked.
Downloaded and made 2.6.22.12 (w/oldconfig from the
working 2.6.21.1 config).

Tried deriving a new config from the 2.6.22.12/proc/config.gz
that worked (using make oldconfig).  This kernel won't reboot.
It goes through the entire shutdown process, and says it is
about to reboot the system.  Then it just stops. 

One of the new questions on ".23", was suspend-to-ram support.
That defaulted to 'yes' in oldconfig -- tried 'no' (thinking
maybe it was related), but that made no difference.

I assume something is unique to my configuration, but don't
know what would have changed in machine shutdown that would
have affected my system.  I also tried compiling in APM
and telling it to use real-mode to try to power-off the machine
(thinking, maybe, that flag might also affect rebooting; made
no difference).

I can post the config if that helps, but didn't want to include
a bunch of superfluous info if it wasn't needed.

Thanks,
Linda


-

From: Linda Walsh
Date: Monday, November 12, 2007 - 7:26 pm

More info, the last program executed to reboot was:
"reboot -df" (in the "reboot" system script").

That should reboot w/no disk sync and no-wtmp entry (they
are done earlier in the program, explicitly).

I looked at the reboot code in the kernel for the i386
(arch/i386/kernel/reboot.c).

A new "special case" was added in 2.6.23.1 (vs. working
2.6.22.12) for a "Dell Inc.","Optiplex 745". Was that a new
machine, or was it included because of some new reboot
problem affecting the machine in 2.6.23?  If it is the latter,
maybe the problem is affecting my platform as well (Dell
Precision-690).  There is nothing looking like a Precision
workstation on the special Dell list in the reboot code.

If I include "reboot=b" on the kernel command line, reboot "works".
However, this was not needed in the previous kernel(s).
It's an easy workaround for me, but worrisome to have it
break for an unknown reason.

Anyone have any ideas?
Meanwhile, guess I'll go add reboot=b in my boot menu file. ;^/


-

Previous thread: x86_64: Make sparsemem/vmemmap the default memory model by Christoph Lameter on Monday, November 12, 2007 - 4:52 pm. (14 messages)

Next thread: [PATCH 0/5][RFC] Physical PCI slot objects by Alex Chiang on Monday, November 12, 2007 - 5:08 pm. (69 messages)