Re: pxeloader problem: BTX halted

On Fri, Feb 20, 2004 at 10:56:06AM +0200, Ruslan Ermilov wrote:
> On Thu, Feb 19, 2004 at 09:30:53AM -0500, John Baldwin wrote:
> > On Thursday 19 February 2004 06:21 am, Ruslan Ermilov wrote:
> > > On Wed, Feb 18, 2004 at 08:59:59AM -0500, John Baldwin wrote:
> > > > On Sunday 15 February 2004 05:54 am, Bjoern A. Zeeb wrote:
> > > > > Hi,
> > > > >
> > > > > I have a problem with latest pxeboot which seems to "crash". The only
> > > > > thing I could read on the screen (no serial console at that point) had
> > > > > been
> > > > >
> > > > > <some register dump>
> > > > > BTX halted
> > > > >
> > > > > this seems to be in sys/boot/i386/btx/btx/btx.S
> > > > >
> > > > > I have set following in make.conf:
> > > > >
> > > > > LOADER_TFTP_SUPPORT=	YES
> > > > > LOADER_BZIP2_SUPPORT=	YES
> > > > > NOFORTH=		YES
> > > > >
> > > > > There is no keyboard attached.
> > > > >
> > > > > The last pxeboot I had been using had been from 20040125.
> > > > >
> > > > > anyone any ideas - any patch to try ?
> > > >
> > > > Try again with newer sources as I think ru_at_ has fixed a bug in his
> > > > renaming of the asm files in the bootstrap.
> > >
> > > That wasn't a bug, just a small timeframe between repocopies
> > > and necessary changes.  There was a bug in kgzldr though, where
> > > I mistakenly cause the kgzldr.o to be stripped when installing
> > > to /usr/lib.  This has been causing problems for release snapshots
> > > where loader(8) is kgzipped.
> > 
> > Well, the above report is a bug that needs to be fixed then. :(  *sigh*
> > 
> Guys, I've verified that with my changes to sys/boot/ makefiles
> I get the same binaries as before (modulo three byte difference
> from the vers.c stuff) for both loader{.sym,.bin,} and pxeboot,
> after doing some minor mods to current versions of makefiles so
> that btxcrt.o and vers.o are linked in the same order as they
> did before.  So should this issue exist (sorry, I cannot verify
> the PXE booting at the moment), it must be elsewhere and doesn't
> relate to my work in sys/boot/ makefiles.
> 
> Bjoern, I recall John was doing some work in libstand recently.
> Can you see if reverting to the 2004/01/15 (GMT) version of
> libstand cures your problem?

This commit broke PXE booting:

-----
ru          2004/02/09 06:11:58 PST

  FreeBSD src repository

  Modified files:
    sys/boot/i386        Makefile.inc
    sys/boot/i386/boot0  Makefile
    sys/boot/i386/boot2  Makefile
    sys/boot/i386/btx/btx Makefile
    sys/boot/i386/btx/btxldr Makefile
    sys/boot/i386/btx/lib Makefile
    sys/boot/i386/cdboot Makefile
    sys/boot/i386/kgzldr Makefile
    sys/boot/i386/loader Makefile
    sys/boot/i386/mbr    Makefile
    sys/boot/i386/pxeldr Makefile
  Log:
  - Factor out -nostdlib to an upper level Makefile.inc.
 
  - Now that bsd.prog.mk deals with programs linked with -nostdlib
    better, and has a notion of an "internal" program, use PROG
    where possible.  This has a good impact on the contents of
    .depend files and causes programs to be linked with cc(1).

  XXX: boot2 couldn't be converted as it's actually two programs.

  Tested on:      i386, amd64
-----

I verified this with a binary search using perforce syncing against a
local checkout of //depot/vendor/freebsd/src/.  At 2004/02/09:06:15
pxeldr works.  At 2004/02/09:06:16 it dies.  The only files that change
in this time frame are these files in sys/boot.

Specifically the problem is rev 1.73 of sys/boot/i386/loader/Makefile.  A
tip of tree checkout with rev 1.73 removed PE boots.

-- Brooks

-- 
Any statement of the form "X is the one, true Y" is FALSE.
PGP fingerprint 655D 519C 26A7 82E7 2529  9BF0 5D8E 8BE9 F238 1AD4