NetBSD on Digital's VAX series of machines (NetBSD/vax)

David Brownlee | 22 Jun 2012 23:12

HEADs up - gcc 4.1 reversion & MSCP CD changes in current

gcc:

Pending someone having the time and expertise to look into the gcc 4.5
vax codegen issues I've reverted -current vax to the same gcc 4.1 used
in netbsd-6. This should at least make current usable for people in
the (probably quite long :) meantime.

MSCP CD drives:

These should now show up as racd* devices. Remember to run MAKEDEV &
update fstab. CDs without disklabels will now default to a single
iso9660 partition for the whole disk (similar to SCSI cd* devices).

sysinst:

Has been updated to take advantage of the racd* split to offer racd0a
as the default CD drive on systems with an MSCP CD drive, and cd0a on
systems with a native SCSI CD drive. (On systems with both the first
one in sysctl hw.disknames will be used).

Dave McGuire | 22 Jun 2012 23:18

Re: HEADs up - gcc 4.1 reversion & MSCP CD changes in current


  Sweet!  Thank you for your efforts!

            -Dave

On 06/22/2012 05:12 PM, David Brownlee wrote:
> gcc:
> 
> Pending someone having the time and expertise to look into the gcc 4.5
> vax codegen issues I've reverted -current vax to the same gcc 4.1 used
> in netbsd-6. This should at least make current usable for people in
> the (probably quite long :) meantime.
> 
> MSCP CD drives:
> 
> These should now show up as racd* devices. Remember to run MAKEDEV &
> update fstab. CDs without disklabels will now default to a single
> iso9660 partition for the whole disk (similar to SCSI cd* devices).
> 
> sysinst:
> 
> Has been updated to take advantage of the racd* split to offer racd0a
> as the default CD drive on systems with an MSCP CD drive, and cd0a on
> systems with a native SCSI CD drive. (On systems with both the first
> one in sysctl hw.disknames will be used).

--

-- 
Dave McGuire, AK4HZ
New Kensington, PA

(Continue reading)

David Brownlee | 23 Jun 2012 20:27

Picon

Re: HEADs up - gcc 4.1 reversion & MSCP CD changes in current

On 22 June 2012 22:18, Dave McGuire <mcguire <at> neurotica.com> wrote:
>
>  Sweet!  Thank you for your efforts!

I was just doing some housekeeping while the serious devs are away :)

Unfortunately I managed to introduce two separate bugs in the sysinst
changes which Christos & Martin fixed :/ Not so impressive.

Tom Ivar Helbekkmo | 24 Jun 2012 13:01

Picon

Re: HEADs up - gcc 4.1 reversion & MSCP CD changes in current

David Brownlee <abs <at> absd.org> writes:

> Pending someone having the time and expertise to look into the gcc 4.5
> vax codegen issues I've reverted -current vax to the same gcc 4.1 used
> in netbsd-6.

I've been trying to build -current for VAX today, from scratch, with
your changes, but after several iterations of updating old-style
function definitions (here's the latest:

/usr/src/lib/libc/gen/alarm.c:55: warning: old-style function definition

), I'm still not past libc, and I'm starting to worry that modernizing
all of these calls is maybe not the preferred solution...?

-tih
--

-- 
"The market" is a bunch of 28-year-olds who don't know anything. --Paul Krugman

David Brownlee | 25 Jun 2012 08:47

Re: HEADs up - gcc 4.1 reversion & MSCP CD changes in current

On 24 June 2012 12:01, Tom Ivar Helbekkmo <tih <at> hamartun.priv.no> wrote:
> David Brownlee <abs <at> absd.org> writes:
>
>> Pending someone having the time and expertise to look into the gcc 4.5
>> vax codegen issues I've reverted -current vax to the same gcc 4.1 used
>> in netbsd-6.
>
> I've been trying to build -current for VAX today, from scratch, with
> your changes, but after several iterations of updating old-style
> function definitions (here's the latest:
>
> /usr/src/lib/libc/gen/alarm.c:55: warning: old-style function definition
>
> ), I'm still not past libc, and I'm starting to worry that modernizing
> all of these calls is maybe not the preferred solution...?

It looks like gcc 4.5 changed behaviour to "helpfully" not warn about
old style definitions for system headers.

Updating all of the reported old-style function definitions is
definitely the final preferred solution, but to get a build right now
you may want to add

NOGCCERROR=yes

to your /etc/mk.conf

Tom Ivar Helbekkmo | 25 Jun 2012 09:24

Picon

Re: HEADs up - gcc 4.1 reversion & MSCP CD changes in current

David Brownlee <abs <at> absd.org> writes:

> Updating all of the reported old-style function definitions is
> definitely the final preferred solution, [...]

Ah, well, I'm in no hurry, so if I do a "-k" run, and fix them all, and
submit the patch, that would be useful?  If so, I might just do that...

-tih
--

-- 
"The market" is a bunch of 28-year-olds who don't know anything. --Paul Krugman

David Brownlee | 25 Jun 2012 10:12

Re: HEADs up - gcc 4.1 reversion & MSCP CD changes in current

On 25 June 2012 08:24, Tom Ivar Helbekkmo <tih <at> hamartun.priv.no> wrote:
> David Brownlee <abs <at> absd.org> writes:
>
>> Updating all of the reported old-style function definitions is
>> definitely the final preferred solution, [...]
>
> Ah, well, I'm in no hurry, so if I do a "-k" run, and fix them all, and
> submit the patch, that would be useful?  If so, I might just do that...

Absolutely - though I knocked off about 1/4 of them over breakfast
this morning and was planning on finishing them off over lunch :)

Tom Ivar Helbekkmo | 25 Jun 2012 10:24

Picon

Re: HEADs up - gcc 4.1 reversion & MSCP CD changes in current

David Brownlee <abs <at> absd.org> writes:

> Absolutely - though I knocked off about 1/4 of them over breakfast
> this morning and was planning on finishing them off over lunch :)

Ah, then you'll beat me to it, so I'll just wait for your commit.  :)

-tih
--

-- 
"The market" is a bunch of 28-year-olds who don't know anything. --Paul Krugman

David Brownlee | 25 Jun 2012 19:38

Re: HEADs up - gcc 4.1 reversion & MSCP CD changes in current

On 25 June 2012 09:24, Tom Ivar Helbekkmo <tih <at> hamartun.priv.no> wrote:
> David Brownlee <abs <at> absd.org> writes:
>
>> Absolutely - though I knocked off about 1/4 of them over breakfast
>> this morning and was planning on finishing them off over lunch :)
>
> Ah, then you'll beat me to it, so I'll just wait for your commit.  :)

This is turning out to be something of a trek. So far have fixed:
- -Wl,--fatal-warnings being turned off for gcc-4.5 but not 4.1
- pdisk casting a uint32_t pointer to (long *)
- a (valid) uninitialized use warning in nd6_rtr.c which gcc 4.1
warned about but 4.5 did not
- 126 lib files with old-style definitions

... and now the kernel has failed to link as rfc6056.c has been removed.

Going to run cvs update again and restart a clean build as soon as I
get home. Hopefully the nth time is the lucky charm :)

David Brownlee | 26 Jun 2012 00:56

Re: HEADs up - gcc 4.1 reversion & MSCP CD changes in current

On 25 June 2012 18:38, David Brownlee <abs <at> absd.org> wrote:
> On 25 June 2012 09:24, Tom Ivar Helbekkmo <tih <at> hamartun.priv.no> wrote:
>> David Brownlee <abs <at> absd.org> writes:
>>
>>> Absolutely - though I knocked off about 1/4 of them over breakfast
>>> this morning and was planning on finishing them off over lunch :)
>>
>> Ah, then you'll beat me to it, so I'll just wait for your commit.  :)
>
> This is turning out to be something of a trek. So far have fixed:
> - -Wl,--fatal-warnings being turned off for gcc-4.5 but not 4.1
> - pdisk casting a uint32_t pointer to (long *)
> - a (valid) uninitialized use warning in nd6_rtr.c which gcc 4.1
> warned about but 4.5 did not
> - 126 lib files with old-style definitions
>
> ... and now the kernel has failed to link as rfc6056.c has been removed.
>
> Going to run cvs update again and restart a clean build as soon as I
> get home. Hopefully the nth time is the lucky charm :)

Test built vax & amd64 and committed.

Currently testing increasing MAXPARTITION from 8 to 16 for vax :)

Matt Thomas | 26 Jun 2012 02:21

Re: HEADs up - gcc 4.1 reversion & MSCP CD changes in current


On Jun 25, 2012, at 3:56 PM, David Brownlee wrote:

> 
> Currently testing increasing MAXPARTITION from 8 to 16 for vax :)

That will need compat for the old disklabel since you will to
increase MAXMAXPARTITIONS

David Brownlee | 1 Jul 2012 22:50

Re: HEADs up - gcc 4.1 reversion & MSCP CD changes in current

On 26 June 2012 01:21, Matt Thomas <matt <at> 3am-software.com> wrote:
> On Jun 25, 2012, at 3:56 PM, David Brownlee wrote:
>>
>> Currently testing increasing MAXPARTITION from 8 to 16 for vax :)
> That will need compat for the old disklabel since you will to
> increase MAXMAXPARTITIONS

I'm possibly a little confused - my plan was to
- Change MAXPARTITIONS from 8 to 16
- Add OLDMAXPARTITIONS 8
- Define __HAVE_OLD_DISKLABEL and the DISKUNIT & DISKPART dance
- Probably increase inodes on install media
- Check for port specific magic boot block or disklabe limitations

MAXMAXPARTITIONS seems to be a MI value? Help!

Tom Ivar Helbekkmo | 26 Jun 2012 12:07

Picon

Re: HEADs up - gcc 4.1 reversion & MSCP CD changes in current

David Brownlee <abs <at> absd.org> writes:

> Test built vax & amd64 and committed.

Seems to be missing at least one adjustment, still:

[...]
--- install-../gnu/lib/libstdc++-v3_4 ---
/usr/tools/bin/vax--netbsdelf-install  -N /usr/src/etc -l s -r libstdc++.so.7.0  /usr/arena/vax/usr/lib/libstdc++.so.7
[...]
obsolete fix:
        Removed obsolete file /usr/arena/vax//usr/lib/libstdc++.so.7.0
[...]
======  1 missing files in DESTDIR  ========
Files in flist but missing from DESTDIR.
File wasn't installed ?
------------------------------------------
./usr/lib/libstdc++.so.7.0
========  end of 1 missing files  ==========
[...]

-tih
--

-- 
"The market" is a bunch of 28-year-olds who don't know anything. --Paul Krugman

Tom Ivar Helbekkmo | 26 Jun 2012 12:35

Picon

Re: HEADs up - gcc 4.1 reversion & MSCP CD changes in current

Tom Ivar Helbekkmo <tih <at> Hamartun.Priv.NO> writes:

> Seems to be missing at least one adjustment, still:

My bad - it did that because I had an old libstdc++.so.7.1 sitting
there.  I removed that, and reran the build, and all is fine.

-tih
--

-- 
"The market" is a bunch of 28-year-olds who don't know anything. --Paul Krugman

Tom Ivar Helbekkmo | 26 Jun 2012 19:37

Picon

Re: HEADs up - gcc 4.1 reversion & MSCP CD changes in current

David Brownlee <abs <at> absd.org> writes:

> Test built vax & amd64 and committed.

I've tried booting a brand new kernel on my 4000/700A, which has NetBSD
5.1 installed, and it still fails in the same way it's been doing with
other, newer than 5.1, kernels.  It's got one MSCP disk (actually SCSI),
and one TMSCP TK70 tape station (genuine) installed, and, at boot, says:

mtc0 at uba0: ...
mscpbus0 at mtc0: ...
uda0 at uba0: ...
mscpbus1 at uda0: ...
mt0 at mscpbus0 drive 0: TK70
mt1 at mscpbus0 drive 0: TK70
ra0 at mscpbus1 drive 0: RA90
ra1 at mscpbus1 drive 0: RA90

Being confused about its disk, it fails to boot.

Under 5.1, the devices show up properly.

-tih
--

-- 
"The market" is a bunch of 28-year-olds who don't know anything. --Paul Krugman

David Brownlee | 26 Jun 2012 19:57

Re: HEADs up - gcc 4.1 reversion & MSCP CD changes in current

On 26 June 2012 18:37, Tom Ivar Helbekkmo <tih <at> hamartun.priv.no> wrote:
> David Brownlee <abs <at> absd.org> writes:
>
>> Test built vax & amd64 and committed.
>
> I've tried booting a brand new kernel on my 4000/700A, which has NetBSD
> 5.1 installed, and it still fails in the same way it's been doing with
> other, newer than 5.1, kernels.  It's got one MSCP disk (actually SCSI),
> and one TMSCP TK70 tape station (genuine) installed, and, at boot, says:
>
> mtc0 at uba0: ...
> mscpbus0 at mtc0: ...
> uda0 at uba0: ...
> mscpbus1 at uda0: ...
> mt0 at mscpbus0 drive 0: TK70
> mt1 at mscpbus0 drive 0: TK70
> ra0 at mscpbus1 drive 0: RA90
> ra1 at mscpbus1 drive 0: RA90
>
> Being confused about its disk, it fails to boot.
>
> Under 5.1, the devices show up properly.

OK, so I've not introduced any new breakage with my racd changes, but
something is broken.

Hmm, could you send a full dmesg from 5.1 and current. What is the
actual failure message after probing - does it just hang?

(Continue reading)

Tom Ivar Helbekkmo | 26 Jun 2012 21:13

Picon

Re: HEADs up - gcc 4.1 reversion & MSCP CD changes in current

David Brownlee <abs <at> absd.org> writes:

> OK, so I've not introduced any new breakage with my racd changes, but
> something is broken.

Yeah, this has been like this for a couple of months at least.

> Hmm, could you send a full dmesg from 5.1 and current. What is the
> actual failure message after probing - does it just hang?

Here's the dmesg output booting 5.1.1:

Copyright (c) 1996, 1997, 1998, 1999, 2000, 2001, 2002, 2003, 2004, 2005,
    2006, 2007, 2008, 2009, 2010, 2011, 2012
    The NetBSD Foundation, Inc.  All rights reserved.
Copyright (c) 1982, 1986, 1989, 1991, 1993
    The Regents of the University of California.  All rights reserved.

NetBSD 5.1.1 (GENERIC) #0: Thu Jan  5 02:17:53 UTC 2012
	builds <at> b7.netbsd.org:/home/builds/ab/netbsd-5-1-1-RELEASE/vax/201201040549Z-obj/home/builds/ab/netbsd-5-1-1-RELEASE/src/sys/arch/vax/compile/GENERIC
VAX 4000/700A
total memory = 127 MB
avail memory = 121 MB
timecounter: Timecounters tick every 10.000 msec
mainbus0 (root)
cpu0 at mainbus0KA692, NVAX (ucode rev 2)
ze0 at mainbus0
ze0: hardware address 08:00:2b:2b:67:a6
shac at mainbus0 not configured
uba0 at mainbus0: Q22

(Continue reading)

Michael L. Hitch | 26 Jun 2012 22:54

Re: HEADs up - gcc 4.1 reversion & MSCP CD changes in current

On Tue, 26 Jun 2012, Tom Ivar Helbekkmo wrote:

> David Brownlee <abs <at> absd.org> writes:
>
>> OK, so I've not introduced any new breakage with my racd changes, but
>> something is broken.
>
> Yeah, this has been like this for a couple of months at least.
...
> With a completely current kernel, things get reordered slightly after
> the Qbus is detected (I'm copying by hand, now, so I'm skipping the
> new version announcement, and the unchanged stuff):
>
> mtc0 at uba0 csr 174500 vec 774 ipl 17
> mscpbus0 at mtc0: version 4 model 14
> mscpbus0: DMA burst size set to 4
> uda0 at uba0 csr 172150 vec 770 ipl 17
> mscpbus1 at uda0: version 6 model 13
> mscpbus1: DMA burst size set to 4
> mt0 at mscpbus0 drive 0: TK70
> mt1 at mscpbus0 drive 0: TK70
> ra0 at mscpbus1 drive 0: RA90
> ra1 at mscpbus1 drive 0: RA90
>
> Notice how it's now not announcing any of the devices until after it's
> got both buses detected, while 5.1.1 showed them immediately after the
> bus had been detected, before going on to the next controller.

   I wonder if this might be due to the change in mscp.c that does the 
device configuration in a worker thread.  I see that there is a static

(Continue reading)

Return

Return to gmane.os.netbsd.ports.vax.

Search Archive

Language

Change language

Options

Current view: Threads only / Showing only 20 lines / Not hiding cited text.
Change to All messages, whole messages, or hide cited text.

Post a message
NNTP Newsgroup
Classic Gmane web interface
XML

XML

RSS Feed
List Information

About Gmane