i386/74008: IBM eServer x225 cannot boot any v5.x - endless dump scrolling in bootloader.

[Bjoern A. Zeeb]

The following reply was made to PR i386/74008; it has been noted by GNATS.

From: "Bjoern A. Zeeb" <bzeeb-lists at lists.zabbadoz.net>
To: Kostik Belousov <kostikbel at gmail.com>
Cc: bug-followup at freebsd.org, kib at freebsd.org
Subject: Re: i386/74008: IBM eServer x225 cannot boot any v5.x - endless dump
 scrolling in bootloader.
Date: Tue, 13 Nov 2007 11:37:47 +0000 (UTC)

 Hi,
 
 it's me again.
 
 Summary: I have two identical machines, behaving different.
           I have two machines which are both fine with a RELENG_7
           loader from Nov 7th (post jhb MFCs).
 
 Ok, unluckily there was a second identical machine like the first one.
 That second one was only tried with the good working 7 setup up to
 today.
 
 I will name the machine that I had been talking in this PR up to now
 m#1 and the second, new one, m#2.
 
 Bot machines work fine with the "new 7" loader so from now on I am
 explaining what I tried with the old one:
 
 Trying an old loader from October (where all this started) on
 
 m#1 now gives the BTX version line and a "hang" which as I now know after
      all the debugging is because the old loader is trying CHS access by
      default and not EDD. CHS does not seem to work in this setup.
      I am no way able to reproduce the BTX halted + stack trace on this
      machine even with the old original loader.
 
 m#2 gives the BTX halted + stack trace with the old original loader,
      the same thing I had seen on m#1 and am no longer able to reproduce.
 
 So the next thing tried was updating the BIOS on m#2 like it was done
 on m#1 before. reboot and direct power off and unplug (had to
 disconnect the CD-ROm and reconnect the Flash).
 
 Directly try to boot the freebsd from flash got me the same stack
 trace.
 
 I have done multiple rounds through BIOS setup + try + sometimes power
 off/unplug + coffee(not really) with no or small BIOS changes each
 time.
 
 No change to get away with the stack trace.
 
 In the end I had both machines on serial console and checked the
 (visible) BIOS settings were 1:1 identical (modulo UUID, MAC-Address
 and IPMI IP).
 
 No change. m#1 still "hung" and m#2 still gave the backtrace.
 
 I checked both LSI BIOS options to be identical and checked on serial
 console that all output was identical for both machines (modulo the
 obvious differences).
 
 No change. m#1 still "hung" and m#2 still gave the backtrace.
 
 
 I moved the loader to loader.orig on the flash and put my last debug
 loader in which both had some printfs + used EDD by default.
 
 Both m#1 and m#2 booted without a hickup. Both should the same
 register dumps that I had in.
 
 I restored the loader.orig to loader and both were back to the old
 behavior. m#1 still "hung" and m#2 still gave the backtrace.
 
 
 I last tried the up-to-date stock 7 setup on both machines and they
 were both booting fine.
 
 I am wondering if I should dd nvram to HDD and see if there are any
 obvious differences but I do not know if that's worth.
 
 In the end I have two identical IBM machines behaving differntly
 and I guess that one of my many debugging sessions or the fact that
 the machine was lying around for a week must have poked something that
 is not visible to the outside so that the stack trace went away.
 
 As said before that CHS does not work is strange but them announcing
 an inofficial EDD bit makes on wonder as well.
 
 
 Conclusion: update your loader to RELENG_7 and see if it works and
 report back.
 
 -- 
 Bjoern A. Zeeb                                 bzeeb at Zabbadoz dot NeT
 Software is harder than hardware  so better get it right the first time.