Even weirder hangs (was: FreeBSD 5.4-RC3 hangs on boot, 5.3 works)

[Arjan Van Leeuwen]

This gets weirder every time I try to do something.

The 5.3-RELEASE CD boots and installs a kernel that doesn't hang.

If I do a binary update (with freebsd-update, '5.3-SECURITY' branch
from Colin) to just after 5.3-RELEASE-p7, that kernel works too, and
doesn't hang.

If I use freebsd-update to get the newest kernel (after
5.3-RELEASE-p9, ifconf vulnerability), that kernel does _not_ work and
hangs at the same place as the one on the 5.4-RC3 CD.

So far so good - I thought that this meant that 5.3-RELEASE-p9 and
newer didn't work. Now for the weird part.

If I manually compile 5.3-RELEASE-p7, that does _not_ work!

So, to summarize:
- 5.3-RELEASE binary install works
- 5.3-SECURITY binary update at the time of 5.3-RELEASE-p7 works
- 5.3-RELEASE-p7 compiled manually does not work
- 5.3-SECURITY binary update at the time of 5.3-RELEASE-p9 does not work

I have absolutely no idea what causes these kernels to work or not
work. I hope someone can help me!

Arjan

On 4/21/05, Arjan Van Leeuwen <avleeuwen at gmail.com> wrote:
> On 4/21/05, Arjan Van Leeuwen <avleeuwen at gmail.com> wrote:
> > On 4/21/05, Arjan Van Leeuwen <avleeuwen at gmail.com> wrote:
> > > I'm trying to install FreeBSD 5.4-RC3 on a Lex Neo system
> > > (http://www.lex.com.tw, a small form factor PC with a 1GHz VIA C3
> > > processor).
> > >
> > > The system hangs when booting the kernel, at this point:
> > >
> > > Timecounter "i8254" frequency 1193182 Hz quality 0
> > > CPU: VIA C3 Nehemiah+RNG+ACE (998.70-MHz 686-class CPU)
> > >   Origin = "CentaurHauls"  Id = 0x698  Stepping = 8
> > >   Features=0x381b93f<FPU,VME,DE,PSE,TSC,MSR,CX8,SEP,MTRR,PGE,CMOV,PAT,MMX,FXSR,S
> > > SE>
> > > real memory  = 125763584 (119 MB)
> > > avail memory = 113414144 (108 MB)
> > > npx0: [FAST]
> > >
> > (...)
> > >
> > > Booting in safe mode, turning off ACPI in the BIOS of the machine and
> > > changing 'PNP OS' to yes or no in the BIOS does not help.
> >
> > I just tried changing hint.npx.0.flags to 0x07 and 0x01, which also
> > doesn't help.
> 
> I found something that looks like a solution for NetBSD on Google Groups here:
> 
> http://groups.google.com/groups?threadm=41555.144.59.164.106.1105346400.squirrel%40144.59.164.106
> 
> where they use
> 
> npx*  at acpi?  # Math coprocessor
> 
> to force npx via the acpi table. Is something like this possible on FreeBSD?
> 
> Arjan
>