[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Testbed-admins] Booting error in Customizing the Generic Image



Hi Mike,

After used this new freebsd-loader-6.4, the error message looks a little
bit changed. Please refer to the attached picture.

Meanwhile, I will try to compile a loader with general distribution to
see if it works. Thanks a lot.

Best Regards,

Cheng Cui
. Frey Computing Services Center . Baton Rouge, LA 70803

-----Original Message-----
From: Mike Hibler [mailto:mike@flux.utah.edu] 
Sent: Monday, October 12, 2009 10:46 AM
To: Cheng Cui
Cc: 'Mike Hibler'; testbed-admins@flux.utah.edu
Subject: Re: [Testbed-admins] Booting error in Customizing the Generic
Image

I spent some time on Friday trying to figure out how our pxeboot could
be
doing anything differently than the MBR boot, or trying to figure out
what
we might have done that "fixed" the problem for Keith.

The only obvious thing was the memory corruption problem that afflicted
the
newer Dells that you mention.  But that bug is fixed in both the 6.2 and
7.2
pxeboots (whenever you see a message early on about the BIOS SMAP
returning
extra attributes or data, you are encountering the fix).  This bug would
also
show up in the on-disk boot loader, but that would have been fixed in
the
6.2 version of /boot/loader you tried.  (note: I never applied our
changes
to 6.4, just 6.2 and 7.2).

I have encountered another problem in the past where it seemed like the
on-disk loader was "too big" and causing it to hang on me.  I have no
idea
what the magic limit was or whether it even exists, I just made the
loader
smaller on a hunch and came up with a version that worked.  I also spent
some time on Friday looking at the code to see if I could find some
magic
limit that might be exceeded, but didn't find one.

I suspect the real problem is that there is some memory corruption going
on
somewhere in either pxeboot or the standard on-disk loader, and it
either
causes problems or not depending on whether the corruption happens in a
critical part of memory or not.  And that depends on the exact size and
version of the loader.

Okay, here is today's shot in the dark.  Download:

	http://www.emulab.net/downloads/freebsd-loader-6.4

This is a smaller version of the 6.4 /boot/loader which has the only
corruption bug I know of fixed.  It removes a couple of features that
you don't need ("split fs" and NFS support).  Let's see if this at least
changes the nature of the failure.  I doubt it will fix anything, as the
6.2 version of the loader I gave you was even smaller than this one...

On Fri, Oct 09, 2009 at 10:10:51AM -0500, Cheng Cui wrote:
> Hello Mike,
> 
> The good news is that FreeBSD6.4 could be booted through the local
disk,
> if avoided the pxeboot process.
> 
> The bad news is that this pxeboot72 still have the same problem with
> previous pxeboot62a. Please refer to the attached picture.
> 
> Well, it looks similar problem had happened before on some advanced
> boxes like Dell PowerEdge, with such kind of pxeboot.
> 
> Best Regards,
> 
> Cheng Cui
> . Frey Computing Services Center . Baton Rouge, LA 70803
> 

Attachment: pxeboot problem with loader6.4 on FreeBSD64.JPG
Description: JPEG image