headers
der Mouse | 3 Aug 2010 20:32

4.0.1 versus DEBUG/DIAGNOSTIC

I've been running 4.0.1 on a few of my i386 boxen for quite a while
now, and I've had it on a Tadpole SPARCbook 3GX for a bit less time but
still some months.  Now, I've finally found motivation and lesiure
together for trying it on other SPARCs.  Specifically, my guinea-pig
machine is an LX.

I routinely use kernels with DEBUG and DIAGNOSTIC enabled.  On the
Tadpole, I started with the TADPOLE3GX kernel, and I find that I did
not turn on DEBUG and DIAGNOSTIC.  However, the kernel I'm using on the
LX *does* have DEBUG and DIAGNOSTIC turned on, and I'm seeing the
printf at pmap.c line 4207 (pmap.c,v 1.307) tripping:

pmap_chk: wired count 472

The number varies; I've also seen 575.  Oddly, once it starts
happening, every process fork or exit appears to provoke another such
message - but the numbers, while they vary, do not seem to exhibit an
upward trend; they just bounce around among a relatively few values.
This started after I ran a program that calls mmap() and then fork() to
get memory shared between two processes, in case that's relevant - I
have trouble seeing how it could be, but it's the only thing I can
think of that might be tickling something unusual in pmap.

Does this indicate a real problem with the 4.0.1 SPARC pmap?  Or is it
a case of DEBUG reporting a condition that is not actually a problem
but was of interest to someone at some point?  Or what?

I'm basically wondering what to do: comment out the message? backport a
patch? stop trying to use 4.0.1 on sparc? file a PR and hope someone
with SPARC pmap clue is motivated to look at an issue with a moderately
(Continue reading)

Permalink | Reply | 0 comments |
headers
David Brownlee | 3 Aug 2010 22:38
Picon

Re: Severe problems with sun4

On 23 July 2010 12:42, Christian Corti
<corti <at> informatik.uni-stuttgart.de> wrote:
> On Fri, 16 Jul 2010, Christian Corti wrote:
>>
>> is it possible that the sun4 port has some issues? Fact is that support of
>> VME bus SUNs (a 4/260 in my case) seems to be utterly broken. The main
>> problem I have is with the Xylogics drivers (xy, but xd seems to be almost
>> identical), they always create kernel panics while attaching the controller.
>
> More problems:
> First I fixed the xy driver (which was painful due to the uncomprehensible
> code) and now it seems that NetBSD is broken in respect with VME bus
> interrupts. This also explains why I can't use any SCSI disk, too. The
> kernel panics when a VME interrupt is generated:
> [...]
> si0 at vme0 addr 200000 irq 2 vector 40
> si0: options=7<RESELECT,DMA_INTR,DMA>
> scsibus0 at si0: 8 targets, 8 luns per target
> timecounter: Timecounter "clockinterrupt" frequency 100 Hz quality 0
> scsibus0: waiting 2 seconds for devices to settle...
> sd0 at scsibus0 target 0 lun 0: <MAXTOR, XT-4380S, B3C> disk fixed
> vmeintr4: level 2 vector 0x40 [**]
> trap type 0x7: pc=0xf014492c npc=0xf0144930 psr=8000c6<S,PS>
> kernel: alignment fault trap
> Stopped at      netbsd:vmeintr4+0x90:   ld              [%i0 + %g0], %g1
> [...]
>
> [**] this output has been introduced by me in vme_machdep.c in function
>     vmeintr4
>
(Continue reading)

Permalink | Reply | 2 comments |
headers
Christian Corti | 4 Aug 2010 11:27
Picon
Picon
Favicon

Re: Severe problems with sun4

On Tue, 3 Aug 2010, David Brownlee wrote:
> On 23 July 2010 12:42, Christian Corti
>> More problems:
>> First I fixed the xy driver (which was painful due to the uncomprehensible
>> code) and now it seems that NetBSD is broken in respect with VME bus
>> interrupts. This also explains why I can't use any SCSI disk, too. The
[...]
> If you have time it could well be worth test booting older NetBSD
> releases until you find one that works and then to check how the
> interrupt code works there :)

Ok, some news with 4.0.1 (the oldest I had tried was 1.6 and that didn't 
work, too). I've found the reason why interrupts didn't work: A VMEbus 
interrupt caused an infinite interrupt loop in vmeintr4 (file 
vme_machdep.c) because of the statement
[...]
   splx(ihp->ih.ih_classipl);
[...]

Each time the kernel executed that statement it reentered vmeintr4, and so 
on. So in fact a loop was created which then was terminated by the 
watchdog reset. Removing that statement made the VMEbus interrupt system 
work. So now both the SCSI controller and the Xylogics 450 are working.
BTW the onboard I/O devices (e.g. ethernet, serial ports) weren't 
affected.

I will submit the updated Xylogics driver source when I have finished 
with enhancing (notably the missing low-level ioctls which I need), 
testing and cleaning up the code.
(Continue reading)

Permalink | Reply | 1 comment |
headers
David Brownlee | 5 Aug 2010 14:15
Picon

Re: Severe problems with sun4

On 4 August 2010 10:27, Christian Corti
<corti <at> informatik.uni-stuttgart.de> wrote:

> Ok, some news with 4.0.1 (the oldest I had tried was 1.6 and that didn't
> work, too). I've found the reason why interrupts didn't work: A VMEbus
> interrupt caused an infinite interrupt loop in vmeintr4 (file vme_machdep.c)
> because of the statement
> [...]
>  splx(ihp->ih.ih_classipl);
> [...]
>
> Each time the kernel executed that statement it reentered vmeintr4, and so
> on. So in fact a loop was created which then was terminated by the watchdog
> reset. Removing that statement made the VMEbus interrupt system work. So now
> both the SCSI controller and the Xylogics 450 are working.
> BTW the onboard I/O devices (e.g. ethernet, serial ports) weren't affected.

Good catch. Its quite possible the last person to test didn't have any
physical disks attached...

> I will submit the updated Xylogics driver source when I have finished with
> enhancing (notably the missing low-level ioctls which I need), testing and
> cleaning up the code.

Looking forward to that :)
Permalink | Reply | 0 comments |
headers
peasthope | 6 Aug 2010 19:37
Picon
Favicon

Re (2): installing NetBSD on a Sparcstation 2.

Date:	Mon, 12 Jul 2010 17:31:39 -0700
From:	"Aaron J. Grier" <agrier <at> poofygoof.com> 
> is there a reason netboot is not being mentioned?

OK; with the 4.0.1 netboot I get to sysinst.  It reports 
"I can not find any hard disks for use by Net-BSD.";   
but I know there is one internal and one external disk.
Going to /bin/sh in the Utilities and issueing "fdisk -l" 
gives nothing.  Tips?

Thanks,            ... Peter E.

--

-- 
Carnot netboots *.SUN4C; installation of NetBSD 5 pending.
Personal site works;  http://members.shaw.ca/peasthope/ .
Permalink | Reply | 0 comments |
headers
Jon Buller | 7 Aug 2010 04:48
Favicon

Re: Re (2): installing NetBSD on a Sparcstation 2.

On Fri, 2010-08-06 at 10:37 -0700, peasthope <at> shaw.ca wrote:
> Date:	Mon, 12 Jul 2010 17:31:39 -0700
> From:	"Aaron J. Grier" <agrier <at> poofygoof.com> 
> > is there a reason netboot is not being mentioned?
> 
> OK; with the 4.0.1 netboot I get to sysinst.  It reports 
> "I can not find any hard disks for use by Net-BSD.";   
> but I know there is one internal and one external disk.
> Going to /bin/sh in the Utilities and issueing "fdisk -l" 
> gives nothing.  Tips?

Don't use fdisk on a NetBSD/sparc, that is what disklabel and newfs is
for.  Watch the boot messages for the disk names and hand the one(s) you
want of to disklabel to partition them.  If your netbooting, it might be
simpler to do a whole diskless install, and use that as a
backup/bootstrap for your final destination.

Sometimes the installed "fluff" can be handy, like using factor and bc
to figure out a fake cylinder size to use all the sectors on a disk,
instead of just using cyl * trk/cyl * sec/trk which is usually a bit
short of the whole disk.  All those numbers should be in the boot
messages as well.  Or use real editors for all the config files
installed on the /mnt mounted local soon-to-be-boot-disk, it's sometimes
a lot easier than the tools squeezed into an install image.

I'd give you more specific details, but it's been a while since I had a
NetBSD-4 installation, and that was sun4m anyway, not sun4c.  (Not that
there is any significant install difference I can think of.)

Jon
(Continue reading)

Permalink | Reply | 7 comments |
headers
der Mouse | 7 Aug 2010 06:02

Re: Re (2): installing NetBSD on a Sparcstation 2.

>> Going to /bin/sh in the Utilities and issueing "fdisk -l" gives
>> nothing.  Tips?
> Don't use fdisk on a NetBSD/sparc, that is what disklabel and newfs
> is for.

Well, that's what disklabel and newfs are for on i386 too.

More to the point, NetBSD/sparc doesn't use peecee-style MBR labels,
which are what fdisk works with, so it has no use for fdisk.

If sysinst can't find any disks for use with NetBSD, it's pretty much
certain they're not being found by the boot sequence for some reason.
What I'd suggest, if you can manage it, is to capture the boot messages
somehow.  One way would be to do a diskless install and run dmesg from
that.  Another would be to use serial console.  If all else fails, a
ten-finger copy of what you see on the screen when you boot -s and it's
sitting at the "give me a shell name" prompt would help some.

What sort of disk do you _think_ the machine has?  In particular, how
is it connected, powered, enclosed, etc?

> Watch the boot messages for the disk names and hand the one(s) you
> want of to disklabel to partition them.

That will work only if autoconf finds them.  I don't think it is or, I
believe, sysinst would be listing them.

/~\ The ASCII				  Mouse
\ / Ribbon Campaign
 X  Against HTML		mouse <at> rodents-montreal.org
(Continue reading)

Permalink | Reply | 6 comments |
headers
William Barnett-Lewis | 7 Aug 2010 07:18
Picon

Re: Re (2): installing NetBSD on a Sparcstation 2.

(Let's try this again...)
On Fri, Aug 6, 2010 at 11:02 PM, der Mouse <mouse <at> rodents-montreal.org> wrote:
>>> Going to /bin/sh in the Utilities and issueing "fdisk -l" gives
>>> nothing.  Tips?

Hell, I've got a U5 that I can not get NBSD to install on. I have no
clue why but it hangs half way through every time at the same freaking
place. I've pretty much given up on the *BSDs for Sun systems.

Is there a current FAQ on the road bumps to getting these systems installed?

William

--

-- 
Live like you will never die, love like you've never been hurt, dance
like no-one is watching.
                Alex White
Permalink | Reply | 3 comments |
headers
Craig Dewick | 7 Aug 2010 09:51
Picon
Picon

re: upgrading from old 2.4.x to 5.0.x

On Sun, 1 Nov 2009, matthew green wrote:

>
> plain GENERIC should work fine if the system has multiple CPUs.
> if that fails, please file a PR or at least show some more info?
>
> anyone want to help look at the SMP problem?

Sorry for the ultra-long delay with this. Finally I have the target 
machine running again so I'm going to investigate more about upgrading it. 
If need be I can take out a CPU module and run with a single module. That 
will save quite a bit of power too which is good for my electricity bill. 
8-)

Craig.

--

-- 
Craig Dewick - Adelaide Australia - Web: http://lios.apana.org.au/~cdewick"
Email: cdewick <at> nos.lios.apana.org.au - SunShack: http://www.sunshack.org
Galleries: http://www.sunshack.org/gallery2 - Also lots of tech data, etc.
Sun Microsystems webring at http://n.webring.com/hub?ring=sunmicrosystemsu
Permalink | Reply | 1 comment |
headers
Martin Husemann | 7 Aug 2010 12:44
Picon

Re: Re (2): installing NetBSD on a Sparcstation 2.

On Sat, Aug 07, 2010 at 12:18:12AM -0500, William Barnett-Lewis wrote:
> Hell, I've got a U5 that I can not get NBSD to install on.

Just to make sure it is not something very obvious: you did try to install
NetBSD/sparc64 on the U5, not NetBSD/sparc?

For sparc64 there is no trick, just boot the install CD (or netboot the install
kernel) and sysinst will do the install for you.

For NetBSD/sparc you will need a few tricks, but probably you do not want
to run it on that machine anyway.

Martin
Permalink | Reply | 0 comments |

Gmane