The pre-release of Hard Hat Linux for SGI...

The pre-release of Hard Hat Linux for SGI...

[Alex deVries]

The non-alpha, full version of Linux for many Indy's is available at:

ftp://ftp.linux.sgi.com/pub/redhat/hardhat-sgi-5.1.tar.gz

(It is uploading as we speak; please wait until the file size is 267015800
bytes.  Yes, this is big.  Make sure you match the md5sum in
hardhat-sgi-5.1.tar.gz.asc. You can also just wait a few weeks until it is
on CD.) 

If there are any significant problems, I may have an oportunity to fix 
small things before it gets pressed to CD. It would be nice if people
tried installing it before then.  I'm afraid all I have is a 500MB disk to
do installs, so I've never been able to do a full install (which is
something like 650MB).

This has a lot of new features in it:

- many more packages; things like glibc, egcs, gcc, XFree libs, kernel
sources, too many to mention here. 

- a better install that doesn't display funny messages, doesn't complain
about a) broken dependancies [1] b) that certain packages can't install
because they're missing glibc[2] and c) the error that you must have a
swap partition, but offers no way of creating one. 

- fdisk does something, although I'm not sure what.  It is untested.

- a much better kernel that has things like the neato power button and LED
drivers

- glibc updates so that Ralf's up and coming Mozilla will run on it

Stuff that is in there but untested:
- new fdisk from Oliver
- the upgrade feature

Stuff that didn't make it in:
- my initrd stuff; this means you still have to install from another
machine with NFS.  I'm sorry, I know it is painful.
- Mike's graphics fixes.
- gdb
- emacs (sorry, I lost the source RPM)

Enjoy, and let me know if you do have the bandwidth do download and use
it.

Two other significant accomplishments this week that may have not been
touted enough:

- Mike Shaver got some nice coloured bars to display on the screen.  I
have good feelings about this.

- Ralf got Mozilla built, parts of gdb and a lot of other stuff.

Good work, guys!

Also, some thanks to SGI's support for replacing the internal SCSI cable
that was causing me problems on such short notice.

- Alex

[1] Man, was that a pain.  I have a new respect for the packaging folk at
RedHat. Rebuilding things like tetex is a pain.
[2] I rebuilt everything, since the alpha1 packages were built with a
broken rpm.

-- 
Alex deVries, puffin on LinuxNet.
http://www.engsoc.carleton.ca/~adevries/ .

Re: The pre-release of Hard Hat Linux for SGI...

[ralf]

On Tue, Jul 14, 1998 at 07:38:32PM -0400, Alex deVries wrote:

> - Mike Shaver got some nice coloured bars to display on the screen.  I
> have good feelings about this.
> 
> - Ralf got Mozilla built, parts of gdb and a lot of other stuff.
> 
> Good work, guys!

There is even more going on this week.  The DECstation guys have managed
to run their first user task, Thomas managed to get X up on the Magnum 4000 /
Olivetti.  I'd say a pretty good week.

Having done enough packaging work myself I know what pain you had to go
through and I'd like to thank you in the name of us all.  Having the first
reasonle difficult to install distribution is a quantum leap in every
Linux port's history.

Time to put up a webpage and register users :-)

  Ralf

Re: The pre-release of Hard Hat Linux for SGI...

[Alex deVries]

On Wed, 15 Jul 1998 ralf@uni-koblenz.de wrote:
> There is even more going on this week.  The DECstation guys have managed
> to run their first user task, Thomas managed to get X up on the Magnum 4000 /
> Olivetti.  I'd say a pretty good week.

Absolutely!

> Time to put up a webpage and register users :-)

Yup.  I updated the WWW and FTP site a bit.  It'd be really good if we
could maybe put a direct link from the main page to the Manhattan page.
Also, it'd be good to have some sort of hardware compatibility page.  I
know there will be people wondering about CPU compatibility, something I'm
a bit confused about. Honza, could you help us out with that?

I know it's a bitch to download 270MB, but I'd really like it if someone
would give this dsitribution a shot.  Red Hat's going to press this on a
CD later this week, and we need to make sure that we get it right.
Actually, even Linux/SGI coasters would be pretty cool...

We're now at 455 packages, which is something like 70 short of i386.  I
think we're at the point now where we could do another pass over the
packages and get a couple more out. Things like emacs and gdb are just
around the corner.  

I think this is worth an article on slashdot.org.

When Hard Hat is released, is SGI interested in doing a press release?

This is really excellent work on everyone's part. Especially in the last
month, we've really pulled everything together.  There's been some long
hours for some of the developers, and it's very satisfying to see a
distribution come out of it.


- Alex

Re: The pre-release of Hard Hat Linux for SGI...

[Honza Pazdziora]

> 
> > Time to put up a webpage and register users :-)
> 
> Yup.  I updated the WWW and FTP site a bit.  It'd be really good if we

OK, I've updated other links to manhattan/. Is the name Hard Hat 5.1
official?

> could maybe put a direct link from the main page to the Manhattan page.
> Also, it'd be good to have some sort of hardware compatibility page.  I
> know there will be people wondering about CPU compatibility, something I'm
> a bit confused about. Honza, could you help us out with that?

Yes, I'll try to create the page. Basically R4400, R4600, R5000,
PC and SC, anyrate are OK? How about R4000? Other hardware: XGE, XZ
(for consoles), 10BaseT, any reasonable SCSI? How about the Challenge S?

> I know it's a bitch to download 270MB, but I'd really like it if someone
> would give this dsitribution a shot.  Red Hat's going to press this on a
> CD later this week, and we need to make sure that we get it right.
> Actually, even Linux/SGI coasters would be pretty cool...

I have question about the ftp part (some of them based on the fact
that we are little bit short of space on our mirror, but I'd like to
mirror as much as possible). How about the RPM's in the
/src/ftp/pub/redhat/old directory? Shouldn't the be in some
subdirectory or gone? Also, I suggest renaming
/src/ftp/pub/redhat/old/redhat-5.1 to something like
redhat-5.1-alpha1 to be consistent with the name reflect the name of
redhat-5.1alpha1.tgz.

Also, /src/ftp/pub/test/vml.tar.gz could be renamed to something like
vmlinux-indy-2.1.99.tar.gz (is it 2.1.99?), and perhaps we could even
put the kernels to some directory other than test (maybe kernel?) and
also have a link there from software.html.

Anyway, our mirror already has 56 MB of that 270, let's hope it will
go on ;-)

------------------------------------------------------------------------
 Honza Pazdziora | adelton@fi.muni.cz | http://www.fi.muni.cz/~adelton/
                   I can take or leave it if I please
------------------------------------------------------------------------

Re: The pre-release of Hard Hat Linux for SGI...

[Honza Pazdziora]

> 
> > Time to put up a webpage and register users :-)
> 
> Yup.  I updated the WWW and FTP site a bit.  It'd be really good if we

OK, I've updated other links to manhattan/. Is the name Hard Hat 5.1
official?

> could maybe put a direct link from the main page to the Manhattan page.
> Also, it'd be good to have some sort of hardware compatibility page.  I
> know there will be people wondering about CPU compatibility, something I'm
> a bit confused about. Honza, could you help us out with that?

Yes, I'll try to create the page. Basically R4400, R4600, R5000,
PC and SC, anyrate are OK? How about R4000? Other hardware: XGE, XZ
(for consoles), 10BaseT, any reasonable SCSI? How about the Challenge S?

> I know it's a bitch to download 270MB, but I'd really like it if someone
> would give this dsitribution a shot.  Red Hat's going to press this on a
> CD later this week, and we need to make sure that we get it right.
> Actually, even Linux/SGI coasters would be pretty cool...

I have question about the ftp part (some of them based on the fact
that we are little bit short of space on our mirror, but I'd like to
mirror as much as possible). How about the RPM's in the
/src/ftp/pub/redhat/old directory? Shouldn't the be in some
subdirectory or gone? Also, I suggest renaming
/src/ftp/pub/redhat/old/redhat-5.1 to something like
redhat-5.1-alpha1 to be consistent with the name reflect the name of
redhat-5.1alpha1.tgz.

Also, /src/ftp/pub/test/vml.tar.gz could be renamed to something like
vmlinux-indy-2.1.99.tar.gz (is it 2.1.99?), and perhaps we could even
put the kernels to some directory other than test (maybe kernel?) and
also have a link there from software.html.

Anyway, our mirror already has 56 MB of that 270, let's hope it will
go on ;-)

------------------------------------------------------------------------
 Honza Pazdziora | adelton@fi.muni.cz | http://www.fi.muni.cz/~adelton/
                   I can take or leave it if I please
------------------------------------------------------------------------

Re: The pre-release of Hard Hat Linux for SGI...

[ralf]

On Wed, Jul 15, 1998 at 11:00:47AM +0200, Honza Pazdziora wrote:

> Yes, I'll try to create the page. Basically R4400, R4600, R5000,
> PC and SC, anyrate are OK? How about R4000? Other hardware: XGE, XZ
> (for consoles), 10BaseT, any reasonable SCSI? How about the Challenge S?

10BaseT - never tested. I use 10Base2 but I assume the mediaselection works
properly.

XGE, XZ graphics are not supported yet.  We might however try to use the
ARC firmware to support anything that we don't support yet as a console.

We only support the the builtin WD33C93 hostadapter.  Extra GIO boards,
especially the WD33C95 (afaik available in the Challenge S), are not
supported yet.

Challenge S isn't supported yet.

The cache issues seems to be confusing to many people, so here a simple
explanation of the issue.  The actual issues are a bit more complex
especially in the case of R4000SC, R4000MC, R4400SC, R4400MC.

Processor types: R4000PC / R4400PC, R4600, R5000.  We've got support for
the second level cache as available on the R4600 and R5000 CPU _boards_,
known as R4600SC and R5000SC.  We do not support the CPU controlled
cache as in R4000SC, R4400SC and R5000.

Why does the second level cache make a difference?  The second level
cache on MIPS systems is ``visible'' to software and the kernel needs
appropriate support for each of them.  We do have this support code for
the external cache controllers on the Indy R4600 and R5000 modules.

R4000SC, R4000MC, R4400SC, R4400MC have integrated cache controllers
which we don't support.  The R5000 also has one which we don't support.
However the Indy R5000SC modules use an external cache controller and
that one is supported.

There are more MIPS CPU types that we support - some of them even
don't exist yet as silicon - but these types were never shipped in Indys.

  Ralf

Re: The pre-release of Hard Hat Linux for SGI...

[Alex deVries]

On Wed, 15 Jul 1998, Honza Pazdziora wrote:
> > Yup.  I updated the WWW and FTP site a bit.  It'd be really good if we
> OK, I've updated other links to manhattan/. Is the name Hard Hat 5.1
> official?

Yes.

> Yes, I'll try to create the page. Basically R4400, R4600, R5000,
> PC and SC, anyrate are OK? How about R4000? Other hardware: XGE, XZ
> (for consoles), 10BaseT, any reasonable SCSI? How about the Challenge S?

I think you need to list individual CPUs.

The SCSI on the Indy, yes, but the one on the Challenge S is different, I
thought.

> I have question about the ftp part (some of them based on the fact
> that we are little bit short of space on our mirror, but I'd like to
> mirror as much as possible). How about the RPM's in the
> /src/ftp/pub/redhat/old directory? Shouldn't the be in some
> subdirectory or gone? Also, I suggest renaming
> /src/ftp/pub/redhat/old/redhat-5.1 to something like
> redhat-5.1-alpha1 to be consistent with the name reflect the name of
> redhat-5.1alpha1.tgz.

I've cleaned that up a bit.

> Also, /src/ftp/pub/test/vml.tar.gz could be renamed to something like
> vmlinux-indy-2.1.99.tar.gz (is it 2.1.99?), and perhaps we could even
> put the kernels to some directory other than test (maybe kernel?) and
> also have a link there from software.html.

Done.

There is a kernel packaged as a binary in the RPMS...

> Anyway, our mirror already has 56 MB of that 270, let's hope it will
> go on ;-)

Let us know how that install goes...

- Alex

Re: The pre-release of Hard Hat Linux for SGI...

[ralf]

On Wed, Jul 15, 1998 at 02:05:15AM -0400, Alex deVries wrote:

> We're now at 455 packages, which is something like 70 short of i386.  I
> think we're at the point now where we could do another pass over the
> packages and get a couple more out. Things like emacs and gdb are just
> around the corner.  
> 
> I think this is worth an article on slashdot.org.

It's already on /.

> When Hard Hat is released, is SGI interested in doing a press release?

Thought I doubt that'll happen it would be really nice.

  Ralf

Re: The pre-release of Hard Hat Linux for SGI...

[Honza Pazdziora]

OK, I finally got the tar.gz here. Let's look:

I've started installing it with networked WS, development, and some of
the servers (NFS, news).

The fdisk seems to work, I was able to change the ID of the partition.
There is still the funny thing when you are selecting the root
partition -- the free space is negative (-522MB). Interesting thing is
that this dialog shows the partitions as Linux native and Linux swap,
even if the fdisk would show them as EFS and raw.

I tried to mkfs the root with check for bad blocks, this killed the
installer with Install exited abnormally. Without the checks it run
OK.

The dialog sees the swap partition as Swap001 but then you still have
to say Continue without swap (I never got stuck in the middle of the
menus with Alpha 1 -- I always got that Continue option anyway).

Some of the packages still complain about failing, on console 4 Unable
to load interpreter. During the install of 230 packages there were
about 20 messages

	ldconfig: Exception at [<88109534>] (8810960c)

(don't kill me if I retyped the numbers wrong). These numbers were
exactly the same in all installs I did.

About 3 packages before end, the install failed with Out of memory,
Install exited abnormally, messages about signal, etc.

So I rebooted and tried just with networked WS and development, this
failed in about 60 % with Install exited abnormally.

The third attempt I did was only 126 packages of base and networked
WS, this finished fine.

The steps afterwards, that failed, were /usr/sbin/timeconfig and
/usr/sbin/ntsysv. But I'm pretty sure that with RH 5.1 A 1 and some
wider selection of packages, these run OK. So they might just be in
some packages I did not select.

After booting the new kernel, I get Unable to open an initial console.

I will try tomorrow to boot the old kernel to get the prompt and
perhaps install and start sshd there to see the results of the new
kernel.

> Enjoy, and let me know if you do have the bandwidth do download and use
> it.

Alex, if you make any changes or so, please make them available as
patch or packages, not as a new whole tar.gz -- the network was
really slow from U.S. to Europe today.

Thanks and I'm looking forward to more experiments,

------------------------------------------------------------------------
 Honza Pazdziora | adelton@fi.muni.cz | http://www.fi.muni.cz/~adelton/
                   I can take or leave it if I please
------------------------------------------------------------------------

Re: The pre-release of Hard Hat Linux for SGI...

[Honza Pazdziora]

OK, I finally got the tar.gz here. Let's look:

I've started installing it with networked WS, development, and some of
the servers (NFS, news).

The fdisk seems to work, I was able to change the ID of the partition.
There is still the funny thing when you are selecting the root
partition -- the free space is negative (-522MB). Interesting thing is
that this dialog shows the partitions as Linux native and Linux swap,
even if the fdisk would show them as EFS and raw.

I tried to mkfs the root with check for bad blocks, this killed the
installer with Install exited abnormally. Without the checks it run
OK.

The dialog sees the swap partition as Swap001 but then you still have
to say Continue without swap (I never got stuck in the middle of the
menus with Alpha 1 -- I always got that Continue option anyway).

Some of the packages still complain about failing, on console 4 Unable
to load interpreter. During the install of 230 packages there were
about 20 messages

	ldconfig: Exception at [<88109534>] (8810960c)

(don't kill me if I retyped the numbers wrong). These numbers were
exactly the same in all installs I did.

About 3 packages before end, the install failed with Out of memory,
Install exited abnormally, messages about signal, etc.

So I rebooted and tried just with networked WS and development, this
failed in about 60 % with Install exited abnormally.

The third attempt I did was only 126 packages of base and networked
WS, this finished fine.

The steps afterwards, that failed, were /usr/sbin/timeconfig and
/usr/sbin/ntsysv. But I'm pretty sure that with RH 5.1 A 1 and some
wider selection of packages, these run OK. So they might just be in
some packages I did not select.

After booting the new kernel, I get Unable to open an initial console.

I will try tomorrow to boot the old kernel to get the prompt and
perhaps install and start sshd there to see the results of the new
kernel.

> Enjoy, and let me know if you do have the bandwidth do download and use
> it.

Alex, if you make any changes or so, please make them available as
patch or packages, not as a new whole tar.gz -- the network was
really slow from U.S. to Europe today.

Thanks and I'm looking forward to more experiments,

------------------------------------------------------------------------
 Honza Pazdziora | adelton@fi.muni.cz | http://www.fi.muni.cz/~adelton/
                   I can take or leave it if I please
------------------------------------------------------------------------

Re: The pre-release of Hard Hat Linux for SGI...

[Alex deVries]

These are very big problems; I will have a look in a couple of hours.

-- 
Alex deVries, puffin on LinuxNet.
http://www.engsoc.carleton.ca/~adevries/ .



On Wed, 15 Jul 1998, Honza Pazdziora wrote:

> Date: Wed, 15 Jul 1998 22:17:42 +0200 (MET DST)
> From: Honza Pazdziora <adelton@informatics.muni.cz>
> To: Alex deVries <adevries@engsoc.carleton.ca>
> Cc: linux@cthulhu.engr.sgi.com
> Subject: Re: The pre-release of Hard Hat Linux for SGI...
> 
> 
> OK, I finally got the tar.gz here. Let's look:
> 
> I've started installing it with networked WS, development, and some of
> the servers (NFS, news).
> 
> The fdisk seems to work, I was able to change the ID of the partition.
> There is still the funny thing when you are selecting the root
> partition -- the free space is negative (-522MB). Interesting thing is
> that this dialog shows the partitions as Linux native and Linux swap,
> even if the fdisk would show them as EFS and raw.
> 
> I tried to mkfs the root with check for bad blocks, this killed the
> installer with Install exited abnormally. Without the checks it run
> OK.
> 
> The dialog sees the swap partition as Swap001 but then you still have
> to say Continue without swap (I never got stuck in the middle of the
> menus with Alpha 1 -- I always got that Continue option anyway).
> 
> Some of the packages still complain about failing, on console 4 Unable
> to load interpreter. During the install of 230 packages there were
> about 20 messages
> 
> 	ldconfig: Exception at [<88109534>] (8810960c)
> 
> (don't kill me if I retyped the numbers wrong). These numbers were
> exactly the same in all installs I did.
> 
> About 3 packages before end, the install failed with Out of memory,
> Install exited abnormally, messages about signal, etc.
> 
> So I rebooted and tried just with networked WS and development, this
> failed in about 60 % with Install exited abnormally.
> 
> The third attempt I did was only 126 packages of base and networked
> WS, this finished fine.
> 
> The steps afterwards, that failed, were /usr/sbin/timeconfig and
> /usr/sbin/ntsysv. But I'm pretty sure that with RH 5.1 A 1 and some
> wider selection of packages, these run OK. So they might just be in
> some packages I did not select.
> 
> After booting the new kernel, I get Unable to open an initial console.
> 
> I will try tomorrow to boot the old kernel to get the prompt and
> perhaps install and start sshd there to see the results of the new
> kernel.
> 
> > Enjoy, and let me know if you do have the bandwidth do download and use
> > it.
> 
> Alex, if you make any changes or so, please make them available as
> patch or packages, not as a new whole tar.gz -- the network was
> really slow from U.S. to Europe today.
> 
> Thanks and I'm looking forward to more experiments,
> 
> ------------------------------------------------------------------------
>  Honza Pazdziora | adelton@fi.muni.cz | http://www.fi.muni.cz/~adelton/
>                    I can take or leave it if I please
> ------------------------------------------------------------------------
> 

Re: The pre-release of Hard Hat Linux for SGI...

[Alex deVries]

On Wed, 15 Jul 1998, Honza Pazdziora wrote:
> The fdisk seems to work, I was able to change the ID of the partition.
> There is still the funny thing when you are selecting the root
> partition -- the free space is negative (-522MB). Interesting thing is
> that this dialog shows the partitions as Linux native and Linux swap,
> even if the fdisk would show them as EFS and raw.
> I tried to mkfs the root with check for bad blocks, this killed the
> installer with Install exited abnormally. Without the checks it run
> OK.

Okay.  mkfs is messed; we should document that.  I don't intend on fixing
it anytime really soon.

Fine, swap doesn't work either.  I don't mind skipping that for now.

> Some of the packages still complain about failing, on console 4 Unable
> to load interpreter. During the install of 230 packages there were
> about 20 messages
> 	ldconfig: Exception at [<88109534>] (8810960c)

_THAT_ worries me a lot.

If you have an existing install of linux on your SGI, do this:
- install the ldconfig rpm that's in the distribution you have
- run ldconfig

Does that still give the exception?

> About 3 packages before end, the install failed with Out of memory,
> Install exited abnormally, messages about signal, etc.

That's a problem.  How much memory do you have?

> So I rebooted and tried just with networked WS and development, this
> failed in about 60 % with Install exited abnormally.

What package was installing at the time the install broke?

> The third attempt I did was only 126 packages of base and networked
> WS, this finished fine.

Good.

> After booting the new kernel, I get Unable to open an initial console.

Did you pass a root= variable to the kernel when booting?  You should
normally have to do something like:

boot vmlinux root=/dev/sdb1

Does the kernel see the partitions on your disk?

> Alex, if you make any changes or so, please make them available as
> patch or packages, not as a new whole tar.gz -- the network was
> really slow from U.S. to Europe today.

Definitely.

These problems worry me a lot.  Please, if you have had the same or other
experiences, I want to know about them.

- Alex

Re: The pre-release of Hard Hat Linux for SGI...

[Honza Pazdziora]

> 
> > About 3 packages before end, the install failed with Out of memory,
> > Install exited abnormally, messages about signal, etc.
> 
> That's a problem.  How much memory do you have?

32 MB.

> > So I rebooted and tried just with networked WS and development, this
> > failed in about 60 % with Install exited abnormally.
> 
> What package was installing at the time the install broke?

ncurses4, I believe.

> > The third attempt I did was only 126 packages of base and networked
> > WS, this finished fine.
> 
> Good.
> 
> > After booting the new kernel, I get Unable to open an initial console.
> 
> Did you pass a root= variable to the kernel when booting?  You should
> normally have to do something like:
> 
> boot vmlinux root=/dev/sdb1

Yes, sure. I do $linux ;-)

> Does the kernel see the partitions on your disk?

Yes, upto there everything is fine. The Unable to open an initial
console message comes after the message about freeing kernel memory
-- 44kB. The system however seems to be doing something after that
-- at least accoring to the sound of the disk. It works and then it
goes silent. When I boot init=/bin/bash, I again get the Unable to
open an initial console message and then nothing, but the while
before the disk goes silent is shorter. So I assume the prompt is
somewhere out there.

Interesting thing (for me, anyway ;-): when I tried to boot the old
kernel (RH 5.1 A 1, 2.1.99), I got the same message about initial
console. So there must be something wrong on the root filesystem.

During the isntallation, I was not able to run a single process (like
mount or cat, all end with Segmentation fault). When I booted to
Upgrade, I was able to work with the /mnt filesystem just fine. I
thought the initial console could be caused by a missing /etc/inittab
but it's there. What could be the case? The upgrade fails with signal
8 when it should start installing the first selected package, but
before that I can work rather well (just tried ls, cat, mount) with
the filesystem, so I could fix the initial console problem, if I knew
how ;-)

------------------------------------------------------------------------
 Honza Pazdziora | adelton@fi.muni.cz | http://www.fi.muni.cz/~adelton/
                   I can take or leave it if I please
------------------------------------------------------------------------

Re: The pre-release of Hard Hat Linux for SGI...

[Honza Pazdziora]

> 
> > About 3 packages before end, the install failed with Out of memory,
> > Install exited abnormally, messages about signal, etc.
> 
> That's a problem.  How much memory do you have?

32 MB.

> > So I rebooted and tried just with networked WS and development, this
> > failed in about 60 % with Install exited abnormally.
> 
> What package was installing at the time the install broke?

ncurses4, I believe.

> > The third attempt I did was only 126 packages of base and networked
> > WS, this finished fine.
> 
> Good.
> 
> > After booting the new kernel, I get Unable to open an initial console.
> 
> Did you pass a root= variable to the kernel when booting?  You should
> normally have to do something like:
> 
> boot vmlinux root=/dev/sdb1

Yes, sure. I do $linux ;-)

> Does the kernel see the partitions on your disk?

Yes, upto there everything is fine. The Unable to open an initial
console message comes after the message about freeing kernel memory
-- 44kB. The system however seems to be doing something after that
-- at least accoring to the sound of the disk. It works and then it
goes silent. When I boot init=/bin/bash, I again get the Unable to
open an initial console message and then nothing, but the while
before the disk goes silent is shorter. So I assume the prompt is
somewhere out there.

Interesting thing (for me, anyway ;-): when I tried to boot the old
kernel (RH 5.1 A 1, 2.1.99), I got the same message about initial
console. So there must be something wrong on the root filesystem.

During the isntallation, I was not able to run a single process (like
mount or cat, all end with Segmentation fault). When I booted to
Upgrade, I was able to work with the /mnt filesystem just fine. I
thought the initial console could be caused by a missing /etc/inittab
but it's there. What could be the case? The upgrade fails with signal
8 when it should start installing the first selected package, but
before that I can work rather well (just tried ls, cat, mount) with
the filesystem, so I could fix the initial console problem, if I knew
how ;-)

------------------------------------------------------------------------
 Honza Pazdziora | adelton@fi.muni.cz | http://www.fi.muni.cz/~adelton/
                   I can take or leave it if I please
------------------------------------------------------------------------

Re: The pre-release of Hard Hat Linux for SGI...

[Oliver Frommel]

> 
> Fine, swap doesn't work either.  I don't mind skipping that for now.
>

hmm. _something_ seems to work here :)

Starting kswapd v 1.5 
Adding Swap: 50368k swap-space (priority -1)


             total       used       free     shared    buffers     cached
Mem:         28332      25716       2616       1676       4900      15956
-/+ buffers/cache:       4860      23472
Swap:        50368        988      49380


o. 

Re: The pre-release of Hard Hat Linux for SGI...

[Honza Pazdziora]

Alright, here are the steps I needed to do in order to have machine that
would allow me to log in. To repeat what I stared with, I installed
just the newtorked WS, did not select any aditional packages, did not
remove any from selection, just base and network.

The first problem: Unable to open initial console. Since you are able
to boot Upgrade from network and have the second console available
(do not start the upgrade, just boot it and hit twice Enter to get rid
of those messages about mount failing), you can create the entries
in the dev directory, that are missing and cause the trouble. I did

	cd /dev
	tar cf - * | ( cd /mnt/dev ; tar xvf - )

so you take the devices from the installation root (there are just
about 8 of them, but there is console ther and that's you need).
The install.log shows that the installation of the dev RPM failed,
so that's why we need to create the devices by hand (obviously, when
it says it fails, it really fails in some way ;-) So you have
console but do not have devices for disks, so Linux won't be able
to remount root, etc.

I've found device /tmp/sdb2 still on the second console of Upgrade,
numbers 8/18, so I created /mnt/dev/sdb2 and /mnt/dev/sdb1 with 8/18
and 8/17. You might be installing to different drives, but anyway,
look around and you'll find the numbers.

If you do not have /etc/passwd like I didn't, you can do

	echo 'root::0:0:root:/:/bin/bash' > /mnt/etc/passwd

to be able to log in after you reboot.

I think it's time to reboot, so do sync and reboot.

I had to boot with init=/bin/bash because first I forgot to create the
/etc/passwd, and second, the /bin/login said something about libraries
(it just blinked and returned the prompt but I saw something like .so
so I think that was it). So once you have bash# prompt, do
/sbin/ldconfig.

> > Some of the packages still complain about failing, on console 4 Unable
> > to load interpreter.

P.S.: I maybe did not stressed out that some of the packages failed
_and_ some gave the following message -- I do not know which gave the
message about ldconfig, but certainly not all that failed.

		During the install of 230 packages there were
> > about 20 messages
> > 	ldconfig: Exception at [<88109534>] (8810960c)
> 
> _THAT_ worries me a lot.
> 
> If you have an existing install of linux on your SGI, do this:
> - install the ldconfig rpm that's in the distribution you have

It was installed fine already ...

> - run ldconfig

... so I did run ldconfig ...

> Does that still give the exception?

... and no, no exception, just some messages about some libraries
that I decided are harmless. Anyway, after this, I was able to log in
on prompt and start ncftp to get the dev RPM and install it and install
ssh to get to my mail and send this report.

Do you want the install.log, Alex? I thought you said that the problem
with failing packages was caused by csh and that it was harmless
(just a message). I'm affraid it really means some problem, because
it did not create the /dev entries and did not create the
/etc/passwd.

More later,

------------------------------------------------------------------------
 Honza Pazdziora | adelton@fi.muni.cz | http://www.fi.muni.cz/~adelton/
                   I can take or leave it if I please
------------------------------------------------------------------------

Re: The pre-release of Hard Hat Linux for SGI...

[Honza Pazdziora]

Alright, here are the steps I needed to do in order to have machine that
would allow me to log in. To repeat what I stared with, I installed
just the newtorked WS, did not select any aditional packages, did not
remove any from selection, just base and network.

The first problem: Unable to open initial console. Since you are able
to boot Upgrade from network and have the second console available
(do not start the upgrade, just boot it and hit twice Enter to get rid
of those messages about mount failing), you can create the entries
in the dev directory, that are missing and cause the trouble. I did

	cd /dev
	tar cf - * | ( cd /mnt/dev ; tar xvf - )

so you take the devices from the installation root (there are just
about 8 of them, but there is console ther and that's you need).
The install.log shows that the installation of the dev RPM failed,
so that's why we need to create the devices by hand (obviously, when
it says it fails, it really fails in some way ;-) So you have
console but do not have devices for disks, so Linux won't be able
to remount root, etc.

I've found device /tmp/sdb2 still on the second console of Upgrade,
numbers 8/18, so I created /mnt/dev/sdb2 and /mnt/dev/sdb1 with 8/18
and 8/17. You might be installing to different drives, but anyway,
look around and you'll find the numbers.

If you do not have /etc/passwd like I didn't, you can do

	echo 'root::0:0:root:/:/bin/bash' > /mnt/etc/passwd

to be able to log in after you reboot.

I think it's time to reboot, so do sync and reboot.

I had to boot with init=/bin/bash because first I forgot to create the
/etc/passwd, and second, the /bin/login said something about libraries
(it just blinked and returned the prompt but I saw something like .so
so I think that was it). So once you have bash# prompt, do
/sbin/ldconfig.

> > Some of the packages still complain about failing, on console 4 Unable
> > to load interpreter.

P.S.: I maybe did not stressed out that some of the packages failed
_and_ some gave the following message -- I do not know which gave the
message about ldconfig, but certainly not all that failed.

		During the install of 230 packages there were
> > about 20 messages
> > 	ldconfig: Exception at [<88109534>] (8810960c)
> 
> _THAT_ worries me a lot.
> 
> If you have an existing install of linux on your SGI, do this:
> - install the ldconfig rpm that's in the distribution you have

It was installed fine already ...

> - run ldconfig

... so I did run ldconfig ...

> Does that still give the exception?

... and no, no exception, just some messages about some libraries
that I decided are harmless. Anyway, after this, I was able to log in
on prompt and start ncftp to get the dev RPM and install it and install
ssh to get to my mail and send this report.

Do you want the install.log, Alex? I thought you said that the problem
with failing packages was caused by csh and that it was harmless
(just a message). I'm affraid it really means some problem, because
it did not create the /dev entries and did not create the
/etc/passwd.

More later,

------------------------------------------------------------------------
 Honza Pazdziora | adelton@fi.muni.cz | http://www.fi.muni.cz/~adelton/
                   I can take or leave it if I please
------------------------------------------------------------------------

Re: The pre-release of Hard Hat Linux for SGI...

[Alex deVries]

Okay, this queston is simple:

are you 100% sure that you're not installing with old packages, and that
the huge .tar.gz has the same md5sum ?

- alex

-- 
Alex deVries, puffin on LinuxNet.
http://www.engsoc.carleton.ca/~adevries/ .

Re: The pre-release of Hard Hat Linux for SGI...

[Honza Pazdziora]

> 
> Okay, this queston is simple:
> 
> are you 100% sure that you're not installing with old packages, and that
> the huge .tar.gz has the same md5sum ?

I might be doing some stupid mistake somewhere but I just checked the
checksum and it's the same, the installer says Hard Hat, it installs
kernel 2.1.100 and there are 458 packages in the mipseb/RedHat/RPMS
directory that came from the tar. And my nfsroot argument to
$installinux point to tftpboot and from there the link goes to this
mipseb subdirectory (you can actually check the unpacked tar at our
mirror ftp://ftp.fi.muni.cz/pub/linux/sgi/redhat/mipseb, I install
from that machine). So I'm like at 98 % that it's not the old
version ;-)

------------------------------------------------------------------------
 Honza Pazdziora | adelton@fi.muni.cz | http://www.fi.muni.cz/~adelton/
                   I can take or leave it if I please
------------------------------------------------------------------------

Re: The pre-release of Hard Hat Linux for SGI...

[Honza Pazdziora]

> 
> Okay, this queston is simple:
> 
> are you 100% sure that you're not installing with old packages, and that
> the huge .tar.gz has the same md5sum ?

I might be doing some stupid mistake somewhere but I just checked the
checksum and it's the same, the installer says Hard Hat, it installs
kernel 2.1.100 and there are 458 packages in the mipseb/RedHat/RPMS
directory that came from the tar. And my nfsroot argument to
$installinux point to tftpboot and from there the link goes to this
mipseb subdirectory (you can actually check the unpacked tar at our
mirror ftp://ftp.fi.muni.cz/pub/linux/sgi/redhat/mipseb, I install
from that machine). So I'm like at 98 % that it's not the old
version ;-)

------------------------------------------------------------------------
 Honza Pazdziora | adelton@fi.muni.cz | http://www.fi.muni.cz/~adelton/
                   I can take or leave it if I please
------------------------------------------------------------------------

Installing noarch packages

[Honza Pazdziora]

Hello,

when I try to install automake-1.3-2.noarch.rpm, I get

package automake-1.3-2 is for a different architecture
error: automake-1.3-2.noarch.rpm cannot be installed

This is with rpm version 2.5.1 from HH 5.1 PR distribution. I even
tried to build the automake rpm from src, got rpm fine but then
ended with the same message. So I assume there is something wrong
not with the noarch packages but with the rpm.

BTW: the swap really is recognized and mounted during boot. So
probably the only fix that is needed is to get rid of that menu
Repartition / Continue (or change conditions upon which it is
displayed).

------------------------------------------------------------------------
 Honza Pazdziora | adelton@fi.muni.cz | http://www.fi.muni.cz/~adelton/
                   I can take or leave it if I please
------------------------------------------------------------------------

Installing noarch packages

[Honza Pazdziora]

Hello,

when I try to install automake-1.3-2.noarch.rpm, I get

package automake-1.3-2 is for a different architecture
error: automake-1.3-2.noarch.rpm cannot be installed

This is with rpm version 2.5.1 from HH 5.1 PR distribution. I even
tried to build the automake rpm from src, got rpm fine but then
ended with the same message. So I assume there is something wrong
not with the noarch packages but with the rpm.

BTW: the swap really is recognized and mounted during boot. So
probably the only fix that is needed is to get rid of that menu
Repartition / Continue (or change conditions upon which it is
displayed).

------------------------------------------------------------------------
 Honza Pazdziora | adelton@fi.muni.cz | http://www.fi.muni.cz/~adelton/
                   I can take or leave it if I please
------------------------------------------------------------------------

Re: Installing noarch packages

[Honza Pazdziora]

> 
> This is with rpm version 2.5.1 from HH 5.1 PR distribution. I even
> tried to build the automake rpm from src, got rpm fine but then
> ended with the same message. So I assume there is something wrong
> not with the noarch packages but with the rpm.

I forgot to say that you can of course install it with --ignorearch.
Still, this might be the problem why the devices installfailed,
couldn't it?

------------------------------------------------------------------------
 Honza Pazdziora | adelton@fi.muni.cz | http://www.fi.muni.cz/~adelton/
                   I can take or leave it if I please
------------------------------------------------------------------------

Re: Installing noarch packages

[Honza Pazdziora]

> 
> This is with rpm version 2.5.1 from HH 5.1 PR distribution. I even
> tried to build the automake rpm from src, got rpm fine but then
> ended with the same message. So I assume there is something wrong
> not with the noarch packages but with the rpm.

I forgot to say that you can of course install it with --ignorearch.
Still, this might be the problem why the devices installfailed,
couldn't it?

------------------------------------------------------------------------
 Honza Pazdziora | adelton@fi.muni.cz | http://www.fi.muni.cz/~adelton/
                   I can take or leave it if I please
------------------------------------------------------------------------

Message while mounting NFS

[Honza Pazdziora]

Hello,

These messages run accross console and go to syslog when I mount
NFS disk from remote host:

Jul 16 17:04:33 nemesis kernel: lockd_up: no pid, 2 users??
Jul 16 17:04:38 nemesis kernel: portmap: RPC call returned error 146
Jul 16 17:04:38 nemesis kernel: RPC: task of released request still queued!
Jul 16 17:04:38 nemesis kernel: RPC: (task is on xprt_pending)
Jul 16 17:04:43 nemesis kernel: portmap: RPC call returned error 146
Jul 16 17:04:43 nemesis kernel: RPC: task of released request still queued!
Jul 16 17:04:43 nemesis kernel: RPC: (task is on xprt_pending)
Jul 16 17:04:43 nemesis kernel: lockd_up: makesock failed, error=-146
Jul 16 17:04:48 nemesis kernel: portmap: RPC call returned error 146
Jul 16 17:04:48 nemesis kernel: RPC: task of released request still queued!
Jul 16 17:04:48 nemesis kernel: RPC: (task is on xprt_pending)

The filesystem is however mounted and further works OK. The remote
system is Solaris 2.5.1.

Yours,

------------------------------------------------------------------------
 Honza Pazdziora | adelton@fi.muni.cz | http://www.fi.muni.cz/~adelton/
                   I can take or leave it if I please
------------------------------------------------------------------------

Re: Message while mounting NFS

[Alex deVries]

Yup... this, I'm told, is because the NFS client in 2.1.100 is a bit sick.
that won't get fixed until we do a bit of kernel work.

- Alex

-- 
Alex deVries, puffin on LinuxNet.
http://www.engsoc.carleton.ca/~adevries/ .



On Thu, 16 Jul 1998, Honza Pazdziora wrote:

> Date: Thu, 16 Jul 1998 19:09:42 +0200 (MET DST)
> From: Honza Pazdziora <adelton@informatics.muni.cz>
> To: linux@cthulhu.engr.sgi.com
> Subject: Message while mounting NFS
> 
> 
> Hello,
> 
> These messages run accross console and go to syslog when I mount
> NFS disk from remote host:
> 
> Jul 16 17:04:33 nemesis kernel: lockd_up: no pid, 2 users??
> Jul 16 17:04:38 nemesis kernel: portmap: RPC call returned error 146
> Jul 16 17:04:38 nemesis kernel: RPC: task of released request still queued!
> Jul 16 17:04:38 nemesis kernel: RPC: (task is on xprt_pending)
> Jul 16 17:04:43 nemesis kernel: portmap: RPC call returned error 146
> Jul 16 17:04:43 nemesis kernel: RPC: task of released request still queued!
> Jul 16 17:04:43 nemesis kernel: RPC: (task is on xprt_pending)
> Jul 16 17:04:43 nemesis kernel: lockd_up: makesock failed, error=-146
> Jul 16 17:04:48 nemesis kernel: portmap: RPC call returned error 146
> Jul 16 17:04:48 nemesis kernel: RPC: task of released request still queued!
> Jul 16 17:04:48 nemesis kernel: RPC: (task is on xprt_pending)
> 
> The filesystem is however mounted and further works OK. The remote
> system is Solaris 2.5.1.
> 
> Yours,
> 
> ------------------------------------------------------------------------
>  Honza Pazdziora | adelton@fi.muni.cz | http://www.fi.muni.cz/~adelton/
>                    I can take or leave it if I please
> ------------------------------------------------------------------------
> 

Message while mounting NFS

[Dong Liu]

Honza Pazdziora writes:
 > 
 > Hello,
 > 
 > These messages run accross console and go to syslog when I mount
 > NFS disk from remote host:
 > 
 > Jul 16 17:04:33 nemesis kernel: lockd_up: no pid, 2 users??
 > Jul 16 17:04:38 nemesis kernel: portmap: RPC call returned error 146
 > Jul 16 17:04:38 nemesis kernel: RPC: task of released request still queued!
 > Jul 16 17:04:38 nemesis kernel: RPC: (task is on xprt_pending)
 > Jul 16 17:04:43 nemesis kernel: portmap: RPC call returned error 146
 > Jul 16 17:04:43 nemesis kernel: RPC: task of released request still queued!
 > Jul 16 17:04:43 nemesis kernel: RPC: (task is on xprt_pending)
 > Jul 16 17:04:43 nemesis kernel: lockd_up: makesock failed, error=-146
 > Jul 16 17:04:48 nemesis kernel: portmap: RPC call returned error 146
 > Jul 16 17:04:48 nemesis kernel: RPC: task of released request still queued!
 > Jul 16 17:04:48 nemesis kernel: RPC: (task is on xprt_pending)
 > 
 > The filesystem is however mounted and further works OK. The remote
 > system is Solaris 2.5.1.
 > 

I think this is a bug of red hat system (BTW, where I can send them this bug
report?), portmapper is started after nfsfs, it should be started before
nfsfs.

Dong

Re: Message while mounting NFS

[Alex deVries]

On Thu, 16 Jul 1998, Dong Liu wrote:
> I think this is a bug of red hat system (BTW, where I can send them this bug
> report?), portmapper is started after nfsfs, it should be started before
> nfsfs.

Does starting portmapper before nfsfs actually work for you?  I still had
NFS client problems on my HH install (although the client works somewhat
better on my totally custom and hacked install).

You don't mail them to redhat, you mail them to me or the list... redhat
doesn't really have anything to do with it apart from selling this and us
using their source.

- Alex

Re: Message while mounting NFS

[Honza Pazdziora]

> 
> On Thu, 16 Jul 1998, Dong Liu wrote:
> > I think this is a bug of red hat system (BTW, where I can send them this bug
> > report?), portmapper is started after nfsfs, it should be started before
> > nfsfs.
> 
> Does starting portmapper before nfsfs actually work for you?  I still had

Thanks to all. Yes, installing portmap-4.0-11.mipseb.rpm and moving
S40portmap do S14portmap fixes the problem.

------------------------------------------------------------------------
 Honza Pazdziora | adelton@fi.muni.cz | http://www.fi.muni.cz/~adelton/
                   I can take or leave it if I please
------------------------------------------------------------------------

Re: Message while mounting NFS

[Honza Pazdziora]

> 
> On Thu, 16 Jul 1998, Dong Liu wrote:
> > I think this is a bug of red hat system (BTW, where I can send them this bug
> > report?), portmapper is started after nfsfs, it should be started before
> > nfsfs.
> 
> Does starting portmapper before nfsfs actually work for you?  I still had

Thanks to all. Yes, installing portmap-4.0-11.mipseb.rpm and moving
S40portmap do S14portmap fixes the problem.

------------------------------------------------------------------------
 Honza Pazdziora | adelton@fi.muni.cz | http://www.fi.muni.cz/~adelton/
                   I can take or leave it if I please
------------------------------------------------------------------------

Re: Message while mounting NFS

[Alex deVries]

I'll create appropriate updates to fix this.  That should be a piece of
cake.

- Alex

-- 
Alex deVries, puffin on LinuxNet.
http://www.engsoc.carleton.ca/~adevries/ .



On Fri, 17 Jul 1998, Honza Pazdziora wrote:

> Date: Fri, 17 Jul 1998 09:50:24 +0200 (MET DST)
> From: Honza Pazdziora <adelton@informatics.muni.cz>
> To: Alex deVries <adevries@engsoc.carleton.ca>
> Cc: dliu@npiww.com, adelton@informatics.muni.cz,
>     linux@cthulhu.engr.sgi.com
> Subject: Re: Message while mounting NFS
> 
> > 
> > On Thu, 16 Jul 1998, Dong Liu wrote:
> > > I think this is a bug of red hat system (BTW, where I can send them this bug
> > > report?), portmapper is started after nfsfs, it should be started before
> > > nfsfs.
> > 
> > Does starting portmapper before nfsfs actually work for you?  I still had
> 
> Thanks to all. Yes, installing portmap-4.0-11.mipseb.rpm and moving
> S40portmap do S14portmap fixes the problem.
> 
> ------------------------------------------------------------------------
>  Honza Pazdziora | adelton@fi.muni.cz | http://www.fi.muni.cz/~adelton/
>                    I can take or leave it if I please
> ------------------------------------------------------------------------
> 

Re: Message while mounting NFS

[Dong Liu]

Alex deVries writes:
 > 
 > On Thu, 16 Jul 1998, Dong Liu wrote:
 > > I think this is a bug of red hat system (BTW, where I can send them this bug
 > > report?), portmapper is started after nfsfs, it should be started before
 > > nfsfs.
 > 
 > Does starting portmapper before nfsfs actually work for you?  I still had
 > NFS client problems on my HH install (although the client works somewhat
 > better on my totally custom and hacked install).

Yes, it works for me, the only thing I changed is the "chkconfig" line of 
/etc/rc.d/init.d/{portmap,nfsfs} and run "chkconfig reset", no more
kernel error message and long delay of mounting nfs. 

Here is another thought, in order to mount /usr as nfs, should portmap
be under /sbin instead of /usr/sbin?

BTW, speaking of nfs, I have observed another problem, when I first
login to a user whose home directory is nfs mounted from an HP server,
sgi linux kernel prints

NFS: server xxxx, readdir reply truncated
NFS: nr=156, slots=2, len=10

All the i386 linux boxes don't exhibit this problem


Dong

Re: Message while mounting NFS

[Matthew Kirkwood]

On Thu, 16 Jul 1998, Honza Pazdziora wrote:

> These messages run accross console and go to syslog when I mount
> NFS disk from remote host:
> 
> Jul 16 17:04:33 nemesis kernel: lockd_up: no pid, 2 users??
> Jul 16 17:04:38 nemesis kernel: portmap: RPC call returned error 146
> Jul 16 17:04:38 nemesis kernel: RPC: task of released request still queued!
> Jul 16 17:04:38 nemesis kernel: RPC: (task is on xprt_pending)

[snip]

I think this is because the 2.1 NFS client does not like running
without a local portmap.

Matthew.

Re: Installing noarch packages

[Alex deVries]

there's a problem with /usr/lib/rpmrc then.  It should have a line like:
arch_compat: mipseb: noarch

If that doesn't work, try adding:
arch_compat: mips: noarch

in it.  Hm.  I thought I'd fixed that.

- Alex

-- 
Alex deVries, puffin on LinuxNet.
http://www.engsoc.carleton.ca/~adevries/ .



On Thu, 16 Jul 1998, Honza Pazdziora wrote:

> Date: Thu, 16 Jul 1998 18:17:57 +0200 (MET DST)
> From: Honza Pazdziora <adelton@informatics.muni.cz>
> To: Honza Pazdziora <adelton@informatics.muni.cz>
> Cc: adevries@engsoc.carleton.ca, adelton@informatics.muni.cz,
>     linux@cthulhu.engr.sgi.com
> Subject: Installing noarch packages
> 
> 
> Hello,
> 
> when I try to install automake-1.3-2.noarch.rpm, I get
> 
> package automake-1.3-2 is for a different architecture
> error: automake-1.3-2.noarch.rpm cannot be installed
> 
> This is with rpm version 2.5.1 from HH 5.1 PR distribution. I even
> tried to build the automake rpm from src, got rpm fine but then
> ended with the same message. So I assume there is something wrong
> not with the noarch packages but with the rpm.
> 
> BTW: the swap really is recognized and mounted during boot. So
> probably the only fix that is needed is to get rid of that menu
> Repartition / Continue (or change conditions upon which it is
> displayed).
> 
> ------------------------------------------------------------------------
>  Honza Pazdziora | adelton@fi.muni.cz | http://www.fi.muni.cz/~adelton/
>                    I can take or leave it if I please
> ------------------------------------------------------------------------
> 

Re: Installing noarch packages

[Honza Pazdziora]

> 
> there's a problem with /usr/lib/rpmrc then.  It should have a line like:
> arch_compat: mipseb: noarch

This was there already.

> If that doesn't work, try adding:
> arch_compat: mips: noarch

This fixed the problem. Thanks,

------------------------------------------------------------------------
 Honza Pazdziora | adelton@fi.muni.cz | http://www.fi.muni.cz/~adelton/
                   I can take or leave it if I please
------------------------------------------------------------------------

Re: Installing noarch packages

[Honza Pazdziora]

> 
> there's a problem with /usr/lib/rpmrc then.  It should have a line like:
> arch_compat: mipseb: noarch

This was there already.

> If that doesn't work, try adding:
> arch_compat: mips: noarch

This fixed the problem. Thanks,

------------------------------------------------------------------------
 Honza Pazdziora | adelton@fi.muni.cz | http://www.fi.muni.cz/~adelton/
                   I can take or leave it if I please
------------------------------------------------------------------------

Re: Installing noarch packages

[Alex deVries]

I'll get a new rpm of rpm that fixes this; I think I submitted that fix
for rpm 2.5.2 anyway. 

- alex

-- 
Alex deVries, puffin on LinuxNet.
http://www.engsoc.carleton.ca/~adevries/ .



On Fri, 17 Jul 1998, Honza Pazdziora wrote:

> Date: Fri, 17 Jul 1998 09:16:10 +0200 (MET DST)
> From: Honza Pazdziora <adelton@informatics.muni.cz>
> To: Alex deVries <adevries@engsoc.carleton.ca>
> Cc: adelton@informatics.muni.cz, linux@cthulhu.engr.sgi.com
> Subject: Re: Installing noarch packages
> 
> > 
> > there's a problem with /usr/lib/rpmrc then.  It should have a line like:
> > arch_compat: mipseb: noarch
> 
> This was there already.
> 
> > If that doesn't work, try adding:
> > arch_compat: mips: noarch
> 
> This fixed the problem. Thanks,
> 
> ------------------------------------------------------------------------
>  Honza Pazdziora | adelton@fi.muni.cz | http://www.fi.muni.cz/~adelton/
>                    I can take or leave it if I please
> ------------------------------------------------------------------------
> 

Re: The pre-release of Hard Hat Linux for SGI...

[ralf]

On Thu, Jul 16, 1998 at 03:54:50PM +0200, Honza Pazdziora wrote:

> P.S.: I maybe did not stressed out that some of the packages failed
> _and_ some gave the following message -- I do not know which gave the
> message about ldconfig, but certainly not all that failed.
> 
> 		During the install of 230 packages there were
> > > about 20 messages
> > > 	ldconfig: Exception at [<88109534>] (8810960c)
> > 
> > _THAT_ worries me a lot.

That Exception message is just a debug message.  It means that the kernel
caught a user process passing a bad pointer argument.  This message
doesn't necessarily mean something fatal happend.

> ... and no, no exception, just some messages about some libraries
> that I decided are harmless. Anyway, after this, I was able to log in

Harmless -> symlink pointing to nonexisting files?

  Ralf

Re: The pre-release of Hard Hat Linux for SGI...

[Honza Pazdziora]

> > ... and no, no exception, just some messages about some libraries
> > that I decided are harmless. Anyway, after this, I was able to log in
> 
> Harmless -> symlink pointing to nonexisting files?

Well, something about libraries not existing (libpam and libpwdb),
but when I checked the RPM it was installed OK and I was able to
to log in, so I did not cosider it a problem.

------------------------------------------------------------------------
 Honza Pazdziora | adelton@fi.muni.cz | http://www.fi.muni.cz/~adelton/
                   I can take or leave it if I please
------------------------------------------------------------------------

Some register dumps

[Honza Pazdziora]

The first register dump I got when I tried to rpm -Uvh kernel-headers.
I however believe that it was not caused by the package but by the
overall activity on the system -- on rpm, one make, one lmbench.
I still believe that I do not deserve a register dump ;-) I'm rewriting
the numbers from paper, so there will be mistakes, I'm sure:

Scheduling in interrupt
page fault from irq handler: 0001
$0: 00000000 88140000 00000018 00000009
$4: 88146100 00000001 8899e000 88146100
$8: 00000720 88008000 00000001 00000000
$12: 00000008 000003e2 883749dc 80000000
$16: 89182000 89182000 00000000 89182000
$20: 00000001 0000000b 1fffffff bfb90000
$24: 00000002 8814793c
$28: 89182000 89182568 89182568 880262e4
epc: 880262c4
Status: 3000fc03
Cause 0000000c
Aiee, killing interrupt handler

This run across the scrren for some while, then it stopped. The
previous copies seemed still the same -- at least at a glance.

The second dump I got when compilling ssh, going some ftps and ssh.

Scheduling in interrupt
page fault from irq handler: 0001
$0: 00000000 88140000 00000018 00000009
$4: 88146100 00000001 88700000 88146100
$8: 00000720 88008000 00000001 00000000
$12: 00000008 000003e2 883749dc 80000000
$16: 88ebe000 88ebe000 00000000 88ebe000
$20: 00000001 0000000b 1fffffff bfb90000
$24: 00000000 00000000
$28: 88ebe000 88ebe4e0 88ebe4e0 880s62c4
epc: 880262c4
Status: 3000fc03
Cause 0000000c
Aiee, killing interrupt handler

after which one line

kmem_free: Bad obj addr (objp = 88ebe430, name=signal_queue)

and then another 30 or so lines of

kmem_free: Bad obj addr (objp = 00000007, name=signal_queue)

This is 2.1.100 from HH 5.1 PR distribution, R4600PC, 32 MB.

The register lsitings wouldn't bother me if they did not crash my
machine at the same time ;-)

------------------------------------------------------------------------
 Honza Pazdziora | adelton@fi.muni.cz | http://www.fi.muni.cz/~adelton/
                   I can take or leave it if I please
------------------------------------------------------------------------

Re: The pre-release of Hard Hat Linux for SGI...

[Thomas Bogendoerfer]

On Tue, Jul 14, 1998 at 07:38:32PM -0400, Alex deVries wrote:
> - many more packages; things like glibc, egcs, gcc, XFree libs, kernel
> sources, too many to mention here. 

how did you built egcs ? I just tried it on the Indy and ld dies, when 
building shared libstdc++ (same as on my Olli). Which binutils are you
using ?

Thomas.

PS: Could you please include mipsel also when modifying .spec Files. Most of
the time you need the same fixes for mipseb and mipsel.

Thomas.

-- 
See, you not only have to be a good coder to create a system like Linux,
you have to be a sneaky bastard too ;-)
                   [Linus Torvalds in <4rikft$7g5@linux.cs.Helsinki.FI>]

Re: Message while mounting NFS

[Honza Pazdziora]

> Honza Pazdziora wrote:
> > 
> > Hello,
> > 
> > These messages run accross console and go to syslog when I mount
> > NFS disk from remote host:
> > 
> > Jul 16 17:04:33 nemesis kernel: lockd_up: no pid, 2 users??
> > Jul 16 17:04:38 nemesis kernel: portmap: RPC call returned error 146
> > Jul 16 17:04:38 nemesis kernel: RPC: task of released request still queued!
> > Jul 16 17:04:38 nemesis kernel: RPC: (task is on xprt_pending)

[...]

> Try forcing an rsize and wsize of 8k (8192), make sure you are up to
> date on nfs, mountd and lockd patches, try forcing mounts to NFSv2 and
> if all else fails (my favorite) remove rpcbind from the picture.

Well, just starting portmap before NFS mounting the disks helped. I'm however
not able to say how "system" this solutions is ;-)

> ---------1---------2---------3---------4---------5---------6---------7
> Eric Kimminau                           FTA/RSA
> eak@detroit.sgi.com                     Silicon Graphics, Inc

[CC sent to SGI/Linux list]

------------------------------------------------------------------------
 Honza Pazdziora | adelton@fi.muni.cz | http://www.fi.muni.cz/~adelton/
                   I can take or leave it if I please
------------------------------------------------------------------------