[9fans] fresh compiled kernels no longer boot..?

[9fans] fresh compiled kernels no longer boot..?

[Axel Belinfante]

I updated  today from sources; somehow the kernels I now
compile no longer boot, i.e. at the end of the printing .gz......
the machine reboots, without printing anything, as far as I can see.

Last time I compiled a kernel (and then it worked)
was (at least) several months ago.
The /386/9pcdisk kernel from sources (from nov) boots fine.

When I boot an old kernel I see as first messages
entry: 80100020
cpu0: 334 MHz AuthenticAMD AMD-K6-2 (cpuid: AX 0x058c DX 0x8021BF)
ELCR: 1400

I found recent messages on the list where lucio
mentioned a similar problem with a 486 and jmk mentioned
relaltively recent (correct?)fixes afecting devarch.c, fns.h and l.s.
Is the same problem biting me?
My versions of those files are identical to those on sources.

Help?

Axel.

Re: [9fans] fresh compiled kernels no longer boot..?

[David Presotto]

The 486 problems shouldn't bother a K6 which is a pentium comnpatible machine.
However, jmk has already fixed that for 486's.  If you want, you could comment
out lines 900 and 901 in 9/pc/devarch.c and see if it helps.

Unfortinately, I don't have a matching processor.  I did make a cpu kernel from
the distribution and it worked on a Xeon but that doesn't tell you much I guess.

Re: [9fans] fresh compiled kernels no longer boot..?

[jmk]

On Sun Jan 25 17:38:41 EST 2004, presotto@closedmind.org wrote:
> The 486 problems shouldn't bother a K6 which is a pentium comnpatible machine.
> However, jmk has already fixed that for 486's.  If you want, you could comment
> out lines 900 and 901 in 9/pc/devarch.c and see if it helps.
>
> Unfortinately, I don't have a matching processor.  I did make a cpu kernel from
> the distribution and it worked on a Xeon but that doesn't tell you much I guess.

The DX value from the cpuid instruction shows the processor has the capabilities
which were missing on Lucio's SiS processor and causing it to reboot, so that shouldn't
be the problem.

Re: [9fans] fresh compiled kernels no longer boot..?

[Axel Belinfante]

jmk wrote:
> On Sun Jan 25 17:38:41 EST 2004, presotto@closedmind.org wrote:
> > The 486 problems shouldn't bother a K6 which is a pentium comnpatible machine.
> > However, jmk has already fixed that for 486's.  If you want, you could comment
> > out lines 900 and 901 in 9/pc/devarch.c and see if it helps.
> >
> > Unfortinately, I don't have a matching processor.  I did make a cpu kernel from
> > the distribution and it worked on a Xeon but that doesn't tell you much I guess.
>
> The DX value from the cpuid instruction shows the processor has the capabilities
> which were missing on Lucio's SiS processor and causing it to reboot, so that shouldn't
> be the problem.

i commented out those lines, without result.
got my serial cable and added some print statements to main.c;
it seems to die in cpuidentify().
I'll try adding some more print statements and see what I learn.

Axel.

Re: [9fans] fresh compiled kernels no longer boot..?

[Axel Belinfante]

i wrote:
> jmk wrote:
> > On Sun Jan 25 17:38:41 EST 2004, presotto@closedmind.org wrote:
> > > The 486 problems shouldn't bother a K6 which is a pentium comnpatible machine.
> > > However, jmk has already fixed that for 486's.  If you want, you could comment
> > > out lines 900 and 901 in 9/pc/devarch.c and see if it helps.
> > >
> > > Unfortinately, I don't have a matching processor.  I did make a cpu kernel from
> > > the distribution and it worked on a Xeon but that doesn't tell you much I guess.
> >
> > The DX value from the cpuid instruction shows the processor has the capabilities
> > which were missing on Lucio's SiS processor and causing it to reboot, so that shouldn't
> > be the problem.
>
> i commented out those lines, without result.
> got my serial cable and added some print statements to main.c;
> it seems to die in cpuidentify().
> I'll try adding some more print statements and see what I learn.

It seems that my machine did not like the line (guarded by an if statement):

rdmsr(0x1B, &apicbase);

near the end of cpuidentify().

Now that I'm looking at sources, that piece of code seems to be gone... :-)

Axel.

Re: [9fans] fresh compiled kernels no longer boot..?

[jmk]

On Sun Jan 25 18:45:41 EST 2004, Axel.Belinfante@cs.utwente.nl wrote:
> i wrote:
> > jmk wrote:
> > > On Sun Jan 25 17:38:41 EST 2004, presotto@closedmind.org wrote:
> > > > The 486 problems shouldn't bother a K6 which is a pentium comnpatible machine.
> > > > However, jmk has already fixed that for 486's.  If you want, you could comment
> > > > out lines 900 and 901 in 9/pc/devarch.c and see if it helps.
> > > >
> > > > Unfortinately, I don't have a matching processor.  I did make a cpu kernel from
> > > > the distribution and it worked on a Xeon but that doesn't tell you much I guess.
> > >
> > > The DX value from the cpuid instruction shows the processor has the capabilities
> > > which were missing on Lucio's SiS processor and causing it to reboot, so that shouldn't
> > > be the problem.
> >
> > i commented out those lines, without result.
> > got my serial cable and added some print statements to main.c;
> > it seems to die in cpuidentify().
> > I'll try adding some more print statements and see what I learn.
>
> It seems that my machine did not like the line (guarded by an if statement):
>
> rdmsr(0x1B, &apicbase);
>
> near the end of cpuidentify().
>
> Now that I'm looking at sources, that piece of code seems to be gone... :-)
>
> Axel.

My fault. That code was in my private copy and I updated the main tree before
taking it out. I took it when looking at your problem when I realised is shouldn't
have been there.

Sorry.

Re: [9fans] fresh compiled kernels no longer boot..?

[Axel Belinfante]

Ok. glad it turned out to be as easy to fix as it was. (i was a bit worried)
(and I'm pleasantly surprised that just adding some print statements
 and taking the serial cable was sufficient to locate the problem -
 though I'm sure in this sense it will not be representative)

Axel.

> My fault. That code was in my private copy and I updated the main tree before
> taking it out. I took it when looking at your problem when I realised is shouldn't
> have been there.
>
> Sorry.

Re: [9fans] fresh compiled kernels no longer boot..?

[boyd, rounin]

> ... (and I'm pleasantly surprised that just adding some print statements
>  and taking the serial cable was sufficient to locate the problem -
>  though I'm sure in this sense it will not be representative)

never underestimate the power of judiciously placed print[f]'s as a
debugging tool.

Re: [9fans] fresh compiled kernels no longer boot..?

[matt]

> never underestimate the power of judiciously placed print[f]'s as a
> debugging tool.

but *please* remember to remove them before commiting ;)

and it would be wise to add a comment in case you don't fix the bug in this session but leave the printf


// printf for debugging the stderr problem
printf(2, "are you sure 2 is always stderr?\n");


m

Re: [9fans] fresh compiled kernels no longer boot..?

[boyd, rounin]

> but *please* remember to remove them before commiting ;)

matt, you know me better than that.