Bug 21559
| Summary: | BTX loader sometime show registers | ||||||
|---|---|---|---|---|---|---|---|
| Product: | Base System | Reporter: | vova <vova> | ||||
| Component: | i386 | Assignee: | freebsd-bugs (Nobody) <bugs> | ||||
| Status: | Closed FIXED | ||||||
| Severity: | Affects Only Me | ||||||
| Priority: | Normal | ||||||
| Version: | 4.1-STABLE | ||||||
| Hardware: | Any | ||||||
| OS: | Any | ||||||
| Attachments: |
|
||||||
Responsible Changed From-To: freebsd-bugs->rnordier I'll look at this. I'm having this exact same problem with about a dozen machines. All using ATA/IDE drives. Chipsets range from old generic Intel 440BX or 440GX to ServerWorks, as well as VIA with a K6, an IBM ThinkPad 600, and SIS630 w/Celeron. Problem started Nov 9th. Using /boot/loader from before Nov 9th fixes the problem. Any /boot/loader built since Nov 9th craps out either showing registers as described in the PR or a repeating: ... forth not found builtin: not found builtins not found forth not found builtin: not found builtins not found forth not found builtin: not found builtins not found only not found can't load 'kernel' Type '?' for a list of commands, 'help' for more detailed help. ok _ I'm experiencing this problem with both 5.0-CURRENT and 4.2-STABLE. Since the problem started on all the machenes I do buildworld on at about the same time and since there's such a different variety of hardware involved I suspect the problem to be in some sort of change to the tool chain used to build /boot/loader at around Nov 9th. However other people seem to have started experiencing this same problem much earlier. In any case I need help with this. If need be I can supply coppies of both a working and non working /boot/loader. I do buildworld with both '-march=pentiumpro -O2' optimization settings on 686 hardware where it seems to work despite being officially unsupported and '-march=pentium -O' on K6/Pentium hardware. Same results with /boot/loader in all cases. -- Yarema Yarema wrote: > The following reply was made to PR i386/21559; it has been noted by GNATS. > > From: Yarema <yds@dppl.com> > To: freebsd-gnats-submit@FreeBSD.org, vova@express.ru, > dwstevens@conduit-it.com, rnordier@FreeBSD.org > Cc: > Subject: Re: i386/21559: BTX loader sometime show registers > Date: Thu, 30 Nov 2000 02:34:06 -0500 > > I'm having this exact same problem with about a dozen machines. All > using ATA/IDE drives. Chipsets range from old generic Intel 440BX or > 440GX to ServerWorks, as well as VIA with a K6, an IBM ThinkPad 600, and > SIS630 w/Celeron. Problem started Nov 9th. Using /boot/loader from > before Nov 9th fixes the problem. Any /boot/loader built since Nov 9th > craps > out either showing registers as described in the PR or a repeating: > > ... > forth not found > builtin: not found > builtins not found > forth not found > builtin: not found > builtins not found > forth not found > builtin: not found > builtins not found > only not found > > can't load 'kernel' > > Type '?' for a list of commands, 'help' for more detailed help. > ok _ > > I'm experiencing this problem with both 5.0-CURRENT and 4.2-STABLE. > Since the problem started on all the machenes I do buildworld on at > about the same time and since there's such a different variety of > hardware involved I suspect the problem to be in some sort of change to > the tool chain used to build /boot/loader at around Nov 9th. However > other people seem to have started experiencing this same problem much > earlier. > > In any case I need help with this. If need be I can supply coppies of > both a working and non working /boot/loader. > > I do buildworld with both '-march=pentiumpro -O2' optimization settings > on 686 hardware where it seems to work despite being officially > unsupported and '-march=pentium -O' on K6/Pentium hardware. Same > results with /boot/loader in all cases. If you can come up with a non-working /boot/loader built with standard CFLAGS, we can certainly help get to the bottom of this. (If so, please send along a binary together with instructions on how to recreate it.) However, if it is the gcc optimization settings that's causing the problem, you really just need to stop using them. :) -- Robert Nordier rnordier@nordier.com rnordier@FreeBSD.org Robert Nordier wrote:
>
> Yarema wrote:
>
> > The following reply was made to PR i386/21559; it has been noted by GNATS.
> >
> > From: Yarema <yds@dppl.com>
> > To: freebsd-gnats-submit@FreeBSD.org, vova@express.ru,
> > dwstevens@conduit-it.com, rnordier@FreeBSD.org
> > Cc:
> > Subject: Re: i386/21559: BTX loader sometime show registers
> > Date: Thu, 30 Nov 2000 02:34:06 -0500
> >
> > I'm having this exact same problem with about a dozen machines. All
> > using ATA/IDE drives. Chipsets range from old generic Intel 440BX or
> > 440GX to ServerWorks, as well as VIA with a K6, an IBM ThinkPad 600, and
> > SIS630 w/Celeron. Problem started Nov 9th. Using /boot/loader from
> > before Nov 9th fixes the problem. Any /boot/loader built since Nov 9th
> > craps
> > out either showing registers as described in the PR or a repeating:
> >
> > ...
> > forth not found
> > builtin: not found
> > builtins not found
> > forth not found
> > builtin: not found
> > builtins not found
> > forth not found
> > builtin: not found
> > builtins not found
> > only not found
> >
> > can't load 'kernel'
> >
> > Type '?' for a list of commands, 'help' for more detailed help.
> > ok _
> >
> > I'm experiencing this problem with both 5.0-CURRENT and 4.2-STABLE.
> > Since the problem started on all the machenes I do buildworld on at
> > about the same time and since there's such a different variety of
> > hardware involved I suspect the problem to be in some sort of change to
> > the tool chain used to build /boot/loader at around Nov 9th. However
> > other people seem to have started experiencing this same problem much
> > earlier.
> >
> > In any case I need help with this. If need be I can supply coppies of
> > both a working and non working /boot/loader.
> >
> > I do buildworld with both '-march=pentiumpro -O2' optimization settings
> > on 686 hardware where it seems to work despite being officially
> > unsupported and '-march=pentium -O' on K6/Pentium hardware. Same
> > results with /boot/loader in all cases.
>
> If you can come up with a non-working /boot/loader built with
> standard CFLAGS, we can certainly help get to the bottom of this.
> (If so, please send along a binary together with instructions on
> how to recreate it.)
>
> However, if it is the gcc optimization settings that's causing the
> problem, you really just need to stop using them. :)
OK, I commented out my CFLAGS setting in /etc/make.conf rebuilt and
reinstalled the loader and same results. Dunno what else to say about
reproducing this.
# cd /usr/src && make world
gives me a bad /boot/loader regardless whether I use any optimizations
in CFLAGS or not. Attached is the latest bad /boot/loader gzipped.
Biult with the default -O optimization setting.
--
Yarema
Yarema wrote: > > If you can come up with a non-working /boot/loader built with > > standard CFLAGS, we can certainly help get to the bottom of this. > > (If so, please send along a binary together with instructions on > > how to recreate it.) > > > > However, if it is the gcc optimization settings that's causing the > > problem, you really just need to stop using them. :) > > OK, I commented out my CFLAGS setting in /etc/make.conf rebuilt and > reinstalled the loader and same results. Dunno what else to say about > reproducing this. > > # cd /usr/src && make world > > gives me a bad /boot/loader regardless whether I use any optimizations > in CFLAGS or not. Attached is the latest bad /boot/loader gzipped. > Biult with the default -O optimization setting. From the look of it, you may still be linking with a libstand that was built with more than standard optimisation. However, the closest I can get to your binary and your results is by enabling LOADER_TFTP_SUPPORT Then I also get a crash or nothing is found at the prompt. Do you have TFTP turned on? I would help if you sent the loader.sym file that matches the non-working loader binary (best send the pair). -- Robert Nordier rnordier@nordier.com rnordier@FreeBSD.org Robert Nordier wrote:
>
> Yarema wrote:
>
> > > If you can come up with a non-working /boot/loader built with
> > > standard CFLAGS, we can certainly help get to the bottom of this.
> > > (If so, please send along a binary together with instructions on
> > > how to recreate it.)
> > >
> > > However, if it is the gcc optimization settings that's causing the
> > > problem, you really just need to stop using them. :)
> >
> > OK, I commented out my CFLAGS setting in /etc/make.conf rebuilt and
> > reinstalled the loader and same results. Dunno what else to say about
> > reproducing this.
> >
> > # cd /usr/src && make world
> >
> > gives me a bad /boot/loader regardless whether I use any optimizations
> > in CFLAGS or not. Attached is the latest bad /boot/loader gzipped.
> > Biult with the default -O optimization setting.
>
> >From the look of it, you may still be linking with a libstand that
> was built with more than standard optimisation. However, the closest
> I can get to your binary and your results is by enabling
>
> LOADER_TFTP_SUPPORT
>
> Then I also get a crash or nothing is found at the prompt.
>
> Do you have TFTP turned on? I would help if you sent the loader.sym
> file that matches the non-working loader binary (best send the pair).
Robert,
Thanks for your help. The lapse in time between finding and
uncommenting LOADER_TFTP_SUPPORT for no good reason in /etc/make.conf
and actually finding myself with a half dozen different machines with a
broken loader led me to jump to conclusions. Perhaps that option should
be removed or a comment added that it's broken for now. I had the same
results with both 4.x-STABLE and 5.0-CURRENT /boot/loader. It's all
good now.
--
Yarema
Responsible Changed From-To: rnordier->freebsd-bugs Give someone else a chance to attend to this. State Changed From-To: open->closed Originator verified that this was no longer a problem. According to Hiten Pandya who did the legwork on this the originator was not using the correct libstand. |
While loading loader shows: BTX loader 1.00 BTX version is 1.01 Console: internal video/keyboard BIOS drive A: is disk0 BIOS drive C: is disk1 BIOS 638kB/523200kB available memory FreeBSD/i386 bootstrap loader, Revision 0.8 (root@vbook.express.ru, Tue Sep 26 13:23:17 MSD 2000) Loading /boot/defaults/loader.conf int=00000006 err=00000000 efl=00010203 eip=680001d1 eax=680001d1 ebx=000075cc ecx=00000011 edx=0000765d esi=00007648 edi=00000002 ebp=00093ff8 esp=00093fcc cs=002b ds=0033 es=0033 fs=0033 gs=0033 ss=0033 cs:eip=ff ff ff ff ff ff ff ff-ff ff ff ff ff ff ff ff ss:esp=cb d8 00 00 cc 75 00 00-48 76 00 00 01 00 00 00 BTX halted not always, but often, no keyboard present on this machine I am useing comconsole Fix: don't use loader How-To-Repeat: try boot 4.1-stable via comconsole some times