Vivek Goyal <[EMAIL PROTECTED]> writes:

> On Wed, Jun 14, 2006 at 11:13:27AM -0600, Eric W. Biederman wrote:
>
> Somehow my previous mail did not make to the mailing list. Trying it 
> again.
>
>> 
>> > Other than that, I think that yes you should be able to use kexec to
>> > bootstrap into any kind of kernel (or code running in real mode).
>> > Certainly linux->xen is possible (and with patches I have posted
>> > to xen-devel, xen->xen and xen->linux, though thats not really relevant),
>> > so its not restricted to booting linux.
>> 
>> Correct.
>> 
>> Most BIOS's get confused after the kernel has run so after 16bit BIOS
>> calls can be a problem.  I have a test image called kexec_test from
>> when I was really trying to get the 16bit mode with BIOS calls to
>> work. 
>> 
>
> Eric, I got a little unrelated question. Does that mean that we can not
> use BIOS services at during transition to the new kernel. I was thinking
> of using "Int 10" for setting the video mode during transition.

We can't do that reliably in the normal kexec case, and definitely not
in the kdump case.

> Currently no messages appear on the console if we are in X or using some
> frame buffer driver and this becomes a problem in kdump as while second
> kernel is booting no messages are displayed and if there is a problem
> at customer site, one would never know where did kdump fail.

Someone sent me a patch that makes the frame-buffer case work.
I just haven't had a chance to follow up on that.

> Any thoughts upon how this problem can potentially be handled, would
> be helpful.

If it is the normal kexec case we can execute an order shutdown and
everything is good.

If it is the kdump case we probably should not count on manual
intervention.

>> It uses the serial port to report how far it has gotten.
>> Depending on how ambitious you are I have a fairly good clue on
>> what needs to happen to make things work even when the BIOS would
>> muck things up.
>> 
>> I have a vague memory that the --real-mode switch had a bug in
>> some version so that may be the problem rather than the bios.
>> In any event kexec_test will report how far it gets and be a useful
>> starting place for further debugging.
>> 
>
> I had found one bug in --real-mode entry long back. I could dig out the
> patch I had used that time. Attaching the patch with the mail for the
> people who might be interested in using real mode entry on i386.

Thanks.

I keep intending to find time to integrate the new patches,
but something always comes when I think I am going to have time :(

Eric
_______________________________________________
fastboot mailing list
[email protected]
https://lists.osdl.org/mailman/listinfo/fastboot

Reply via email to