headers
Patrick Finnegan | 3 Nov 2005 00:22
Picon
Favicon

IBM 6015 (40p) won't boot installer.

I've tried the installer for both 1.6.2, 2.0.2, and 2.1.  All three 
freeze at the same place (booting sysinst_com0.fs):

> NetBSD/prep BOOT, Revision 1.5
>> (jmc <at> faith.netbsd.org, Tue Mar 22 21:54:55 UTC 2005)

Boot: Loading in()
3537936+185964=0x38d4e0
start=0x100000

That's the output for the 2.0.2 installer.  1.6.2 and 2.1 have the same 
message with different numbers. :)

Any ideas?

Pat
--

-- 
Purdue University Research Computing --  http://www.itap.purdue.edu/rcac
Permalink | Reply | 502 comments |
headers
Andy Ruhl | 3 Nov 2005 02:42
Picon

Re: IBM 6015 (40p) won't boot installer.

On 11/2/05, Patrick Finnegan <vax <at> purdue.edu> wrote:
> I've tried the installer for both 1.6.2, 2.0.2, and 2.1.  All three
> freeze at the same place (booting sysinst_com0.fs):
>
> > NetBSD/prep BOOT, Revision 1.5
> >> (jmc <at> faith.netbsd.org, Tue Mar 22 21:54:55 UTC 2005)
>
> Boot: Loading in()
> 3537936+185964=0x38d4e0
> start=0x100000
>
> That's the output for the 2.0.2 installer.  1.6.2 and 2.1 have the same
> message with different numbers. :)
>
> Any ideas?

Do you know if those boxes are able to boot NetBSD from other people posting?

How are you booting the kernel by the way? I had trouble initially
until I netbooted. But I have a different box.

Andy
Permalink | Reply | 502 comments |
headers
Patrick Finnegan | 3 Nov 2005 03:50
Picon
Favicon

Re: IBM 6015 (40p) won't boot installer.

Andy Ruhl declared on Wednesday 02 November 2005 20:42:
> On 11/2/05, Patrick Finnegan <vax <at> purdue.edu> wrote:
> > I've tried the installer for both 1.6.2, 2.0.2, and 2.1.  All three
> >
> > freeze at the same place (booting sysinst_com0.fs):
> > > NetBSD/prep BOOT, Revision 1.5
> > >
> > >> (jmc <at> faith.netbsd.org, Tue Mar 22 21:54:55 UTC 2005)
> >
> > Boot: Loading in()
> > 3537936+185964=0x38d4e0
> > start=0x100000
> >
> > That's the output for the 2.0.2 installer.  1.6.2 and 2.1 have the
> > same message with different numbers. :)
> >
> > Any ideas?
>
> Do you know if those boxes are able to boot NetBSD from other people
> posting?

It's listed in the "supported hardware" list for NetBSD/prep, and there's 
a news item on the port page saying that support for the machine was 
added on 2004/06/26, which is why I bothered trying NetBSD anyways.

> How are you booting the kernel by the way? I had trouble initially
> until I netbooted. But I have a different box.

dd image to floppy, boot off floppy.
(Continue reading)

Permalink | Reply | 3 comments |
headers
Klaus Klein | 3 Nov 2005 09:56
Picon

Re: IBM 6015 (40p) won't boot installer.

On Wed, 2 Nov 2005 21:50:08 -0500
Patrick Finnegan <vax <at> purdue.edu> wrote:

> Andy Ruhl declared on Wednesday 02 November 2005 20:42:
> > On 11/2/05, Patrick Finnegan <vax <at> purdue.edu> wrote:
> > > I've tried the installer for both 1.6.2, 2.0.2, and 2.1.  All three
> > >
> > > freeze at the same place (booting sysinst_com0.fs):
> > > > NetBSD/prep BOOT, Revision 1.5
> > > >
> > > >> (jmc <at> faith.netbsd.org, Tue Mar 22 21:54:55 UTC 2005)
> > >
> > > Boot: Loading in()
> > > 3537936+185964=0x38d4e0
> > > start=0x100000
> > >
> > > That's the output for the 2.0.2 installer.  1.6.2 and 2.1 have the
> > > same message with different numbers. :)
> > >
> > > Any ideas?
> >
> > Do you know if those boxes are able to boot NetBSD from other people
> > posting?
> 
> It's listed in the "supported hardware" list for NetBSD/prep, and there's 
> a news item on the port page saying that support for the machine was 
> added on 2004/06/26, which is why I bothered trying NetBSD anyways.

Exactly.  However, the creation of the 2.x release branch predates this;
the first release to support this machine will be 3.0.  Daily snapshots
(Continue reading)

Permalink | Reply | 2 comments |
headers
Patrick Finnegan | 3 Nov 2005 19:41
Picon
Favicon

Re: IBM 6015 (40p) won't boot installer.

On Thursday 03 November 2005 03:56, Klaus Klein wrote:
> On Wed, 2 Nov 2005 21:50:08 -0500
>
> Patrick Finnegan <vax <at> purdue.edu> wrote:
> > It's listed in the "supported hardware" list for NetBSD/prep, and
> > there's a news item on the port page saying that support for the
> > machine was added on 2004/06/26, which is why I bothered trying
> > NetBSD anyways.
>
> Exactly.  However, the creation of the 2.x release branch predates
> this; the first release to support this machine will be 3.0.  Daily
> snapshots of this release branch (netbsd-3) are available from
> ftp://ftp.netbsd.org/pub/NetBSD/NetBSD-daily/netbsd-3/ .

There's at least one problem with all of the images there... they're all 
bigger than a floppy disk.  The sysinst_com0.fs image I was going to 
try is nearly 40kB bigger than a disk.

I guess I could try making a boot harddrive... put a type 0x41 partition 
on a hard drive, dd the image to it, and try booting from the disk.  
Still, it'd be a lot more convenient to be able to use a floppy or 
netboot.

> > > How are you booting the kernel by the way? I had trouble
> > > initially until I netbooted. But I have a different box.
> >
> > dd image to floppy, boot off floppy.
> >
> > I can't netboot, as this machine doesn't have a built-in network
> > interface, and I don't seem to have a compatible card.
(Continue reading)

Permalink | Reply | 1 comment |
headers
Patrick Finnegan | 5 Nov 2005 00:05
Picon
Favicon

Re: IBM 6015 (40p) won't boot installer.

Ok, I'm trying out the 3.0 beta installer, and have found a few problems 
that keep me from installing. (I seem to be good at that lately..)

When I try to set up the disklabel on the disk, I get this as a 
response:
     Status: Command failed
    Command: disklabel -w -r -f /tmp/disktab sd0 'ibm40p'
     Hit enter to continue
--------------------------------------------------------------------------------

NetBSD slice at 32130, partition C at 0
Write outside MBR partition? [n]: y
write: No such file or directory

disktab contents:
# cat /tmp/disktab
ibm40p|NetBSD installation generated:\
        :dt=unknown:ty=winchester:\
        :nc#2857:nt#15:ns#91:\
        :sc#1365:su#3933040:\
        :se#512:\
        :pa#1577940:oa#32130:ta=4.2BSD:ba#8192:fa#1024:ta=4.2BSD:\
        :pb#263445:ob#1610070:tb=swap:\
        :pc#3933040:oc#0:tc=unused:\
        :pd#0:od#0:td=unused:\
        :pe#2046345:oe#1873515:te=4.2BSD:be#8192:fe#1024:ta=4.2BSD:\
        :pf#0:of#0:tf=unused:\
        :pg#0:og#0:tg=unused:\
        :ph#0:oh#0:th=unused:
(Continue reading)

Permalink | Reply | 0 comments |
headers
Anders Gavare | 20 Nov 2005 16:16
Picon

FROMBCD coding style flaw

Hi all,

While experimenting with NetBSD/prep code, I noticed a coding style
flaw. It is probably very unlikely to trigger bugs, but I think it is
worth mentioning it anyway.

A quick googling turned up that this flaw was noticed by Simon Burge in
May 2003 (http://mail-index.netbsd.org/tech-kern/2003/05/07/0004.html).
Unfortunately, there is still code around which is not rewritten.

Basically, the FROMBCD macro is misused if it is given a function call
as an argument, instead of a plain variable.

     #define  FROMBCD(x)      (((x) >> 4) * 10 + ((x) & 0xf))

In sys/arch/prep/isa/mcclock_isa.c, the value from mcclock_isa_read()
can differ between two consecutive reads.

     t1 = FROMBCD(mcclock_isa_read(sc, MC_SEC));

could return 40 or 59, if the RTC's MC_SEC value changes from 0x49 to
0x50 between the two reads. (The value can be either 40 or 59, because
it is not well-defined which of the two (x)s that gets evaluated first.)

In most cases, this would only lead to a scenario where the machine
thinks that the time is a few seconds more/less than what the RTC
actually thinks. I think this is what Simon noticed.

For the PReP port, it is slightly more serious. mcclock and mkclock
check for the existence of a clock chip by 1) reading MC_SEC, 2) doing a
(Continue reading)

Permalink | Reply | 0 comments |
headers
Martin Husemann | 20 Nov 2005 17:44
Picon

Re: FROMBCD coding style flaw

On Sun, Nov 20, 2005 at 04:16:41PM +0100, Anders Gavare wrote:
> Basically, the FROMBCD macro is misused if it is given a function call
> as an argument, instead of a plain variable.

A quick idutils run showed at least the following broken uses:

arch/evbarm/tsarm/tsrtc.c
arch/next68k/next68k/rtc.c
arch/prep/isa/mcclock_isa.c
arch/sh3/sh3/clock.c
dev/ic/mm58167.c

IMHO the callers should be fixed, but the question is how to make sure
we catch them.

Maybe we could define a different version for DIAGNOSTIC or DEBUG kernels
that takes the address of the arg, so it fails compilation for the broken
usage - something like this (untested):

#if defined(__GNUC__) && defined(DIAGNOSTIC)
#define     FROMBCD(x)      ({				\
		__typeof(x) *v = &x;			\
		(((*v) >> 4) * 10 + ((*v) & 0xf));	\
	}
#else
#define     FROMBCD(x)      (((x) >> 4) * 10 + ((x) & 0xf))
#endif

And I think we should eliminate all duplicate definitions of this macro,
no matter what solution we go for, which seem to happen at least in:
(Continue reading)

Permalink | Reply | 0 comments |
headers
Andy Ruhl | 24 Nov 2005 17:21
Picon

Really weird filesystem behavior on prep

I'm trying to upgrade my prep box to the latest 3.0 RC1 binary build,
and I noticed something.

I booted the kernel with the 2.0.2 userland still intact. This worked.

I noticed some different stuff, so I wanted to pull the kernel configs
from the 2.0.2 kernel and the 3.0 RC1 kernel and compare them. So I
used this command straight from the options man page:

strings netbsd | sed -n 's/^_CFG_//p' | unvis

Except I redirected that to a file:

strings netbsd | sed -n 's/^_CFG_//p' | unvis > /root/kernconf.2.0.2

Now, for the weird part.

The directory that I redirected the file into suddenly behaves as if
it has no files in it. That is, ls shows nothing:

root <at> bugs:/root # ls /root
root <at> bugs:/root #

If I reboot, it comes back.

What? How can this be happening?

Any advice would be appreciated.

Thanks.
(Continue reading)

Permalink | Reply | 0 comments |
headers
Andy Ruhl | 24 Nov 2005 19:33
Picon

Re: Really weird filesystem behavior on prep

On 11/24/05, Andy Ruhl <acruhl <at> gmail.com> wrote:
> I'm trying to upgrade my prep box to the latest 3.0 RC1 binary build,
> and I noticed something.
>
> I booted the kernel with the 2.0.2 userland still intact. This worked.
>
> I noticed some different stuff, so I wanted to pull the kernel configs
> from the 2.0.2 kernel and the 3.0 RC1 kernel and compare them. So I
> used this command straight from the options man page:
>
> strings netbsd | sed -n 's/^_CFG_//p' | unvis
>
> Except I redirected that to a file:
>
> strings netbsd | sed -n 's/^_CFG_//p' | unvis > /root/kernconf.2.0.2
>
> Now, for the weird part.
>
> The directory that I redirected the file into suddenly behaves as if
> it has no files in it. That is, ls shows nothing:
>
> root <at> bugs:/root # ls /root
> root <at> bugs:/root #
>
> If I reboot, it comes back.
>
> What? How can this be happening?
>
> Any advice would be appreciated.
(Continue reading)

Permalink | Reply | 0 comments |

Gmane