Khalid Aziz [off-list ref] writes:

On Thu, 2005-06-30 at 17:52 -0600, Eric W. Biederman wrote:

I have found that I can not walk reboot_notifier list in all cases
before kexec'ing a new kernel. For instance, when handling INIT on ia64,
we are running in interrupt context and atleast some of the reboot
notifier callbacks call schedule(). Calling schedule() is not gonna work
when we are running in interrupt context. I have the same concern for
when panic gets called in interrupt context. So I added a shutdown
function to e1000 driver instead. Patch is attached. This patch has
worked for me.

Thanks.  I will look at your e1000 driver and see if it works on
the hardware I have here.  You changes are at least to places
that could have fixed the problem I am seeing.

For the panic case calling any of the shutdown/reboot notifier bits
is actually wrong.  That kernel that takes over needs to cope with
hardware left in any state.

As soon as I have all the issues sorted out with kexec'ing on INIT on
ia64, I will post a fully updated kexec patch for ia64. I now have kexec
working solid on INIT with e1000 driver and it can handle multiple back
to back INITs and come up in kexec'd kernel every time. I am now trying
to sort some issues out with tg3 driver (another driver with no shutdown
routine :(

In the normal case (not kexec on panic) user space should down the interfaces
before calling kexec.  Is your user space doing that?


Eric