[9fans] bleeding edge cd install

[9fans] bleeding edge cd install

[ron minnich]

Hi, I just installed from the bleeding edge ISO, fossil file system, and
have sent up venti as per the WIKI web page on setting up venti.

I have also done this:

venti -w /dev/sdC0/arenas0 venti.conf
so that the conf is in the venti partition.

Finally, in plan9.ini I have set this:
venti=127.1

Boot suicides on the way up with an invalid address trap.

What kind of general thing should I look for at this point?

thanks

ron

Re: [9fans] bleeding edge cd install

[David Presotto]

try it without venti.  I think I'm also running a bleeding edge fossil but
with no venti and mine doesn't go tits up.

Re: [9fans] bleeding edge cd install

[Russ Cox]

> try it without venti.  I think I'm also running a bleeding edge fossil but
> with no venti and mine doesn't go tits up.
>

For what it's worth, I did a bleeding-edge install
two weeks ago and then did Venti and it was fine.

I think you have the Venti variable wrong.
It needs to be

	venti=tcp!127.1!17034 tcp!127.1!80

if I recall correctly.

Russ

Re: [9fans] bleeding edge cd install

[ron minnich]

On Thu, 9 Oct 2003, Russ Cox wrote:

> I think you have the Venti variable wrong.
> It needs to be
>
> 	venti=tcp!127.1!17034 tcp!127.1!80

well, ok, but looking at some of the source it looked like processes were
trying to talk to this:
tcp!$venti!venti

which makes some sense.

Anyway I did just this this exact setting in plan9.ini:
venti=tcp!127.1!17034 tcp!127.1!80
[is that really right?]

same problem.

ron

Re: [9fans] bleeding edge cd install

[Russ Cox]

> Anyway I did just this this exact setting in plan9.ini:
> venti=tcp!127.1!17034 tcp!127.1!80
> [is that really right?]

Actually, no it's not.  I just checked /sys/src/9/boot/local.c.

You want:

	venti=#S/sdC0/venti-conf-disk

where #S/sdC0/venti-conf-disk is the disk you branded
with venti/conf -w.

You can say

	venti=#S/sdC0/venti-conf-disk tcp!127.1!17034 tcp!127.1!8000

but the last two fields are the defaults.

Still, the boot process shouldn't be page faulting,
so this might not solve your problem.
But /sys/src/9/boot/local.c is where you want to look
(and look carefully at any messages before the
page fault).

Russ

Re: [9fans] bleeding edge cd install

[David Presotto]

[-- Attachment #1: Type: text/plain, Size: 354 bytes --]

This new venti string should fix all of ron's problems since,
with a correct venti conf disk, he'll get into the if statement
in the boot process that mounts the loopback interface, which
is his second problem.

When he telnetted to 127.1, he was probably going to whatever machine
his default route points to and connecting to its local interface.

[-- Attachment #2: Type: message/rfc822, Size: 2629 bytes --]

From: "Russ Cox" <rsc@swtch.com>
To: 9fans@cse.psu.edu
Subject: Re: [9fans] bleeding edge cd install
Date: Thu, 09 Oct 2003 21:58:41 -0400
Message-ID: <E1A7mYD-000PZq-8X@t40.swtch.com>

> Anyway I did just this this exact setting in plan9.ini:
> venti=tcp!127.1!17034 tcp!127.1!80
> [is that really right?]

Actually, no it's not.  I just checked /sys/src/9/boot/local.c.

You want:

	venti=#S/sdC0/venti-conf-disk

where #S/sdC0/venti-conf-disk is the disk you branded
with venti/conf -w.

You can say

	venti=#S/sdC0/venti-conf-disk tcp!127.1!17034 tcp!127.1!8000

but the last two fields are the defaults.

Still, the boot process shouldn't be page faulting,
so this might not solve your problem.
But /sys/src/9/boot/local.c is where you want to look
(and look carefully at any messages before the
page fault).

Russ

Re: [9fans] bleeding edge cd install

[ron minnich]

On Thu, 9 Oct 2003, Russ Cox wrote:

> You want:
>
> 	venti=#S/sdC0/venti-conf-disk

hey, yeah, I do. That did it!

I'm putting together a little 'wiki addition' and this goes in there too.

Now I have to figure out how to get fossil and venti from 'P' to 'Drive'.
:-)

Thanks

ron

[9fans] kfscmd atime considered harmful

[Richard Miller]

Using disk/kfscmd atime, which reduces disk activity by preventing the
updating of access times when files and directories are accessed for
reading, has an unfortunate side effect of making the effects of wstat
somewhat ephemeral:

term% ls -l d
--rw-r----- M 8 miller miller 0 Oct 10 10:31 d/one
--rw-r----- M 8 miller miller 0 Oct 10 10:31 d/two
term% disk/kfscmd atime
atimes will not be updated
term% chmod 666 d/one
term% mv d/two d/three
term% ls -l d
--rw-rw-rw- M 8 miller miller 0 Oct 10 10:31 d/one
--rw-r----- M 8 miller miller 0 Oct 10 10:31 d/three
term% du >/dev/null
term% ls -l d
--rw-r----- M 8 miller miller 0 Oct 10 10:31 d/one
--rw-r----- M 8 miller miller 0 Oct 10 10:31 d/two

I'll send a suggested fix to 9trouble@plan9.bell-labs.com

-- Richard

Re: [9fans] bleeding edge cd install

[David Presotto]

It needs the 17034 because until the file system starts up, there is no /lib/ndb/common
to translate venti into 17034.

Of course, this doesn't help since it still doesn't work...

Re: [9fans] bleeding edge cd install

[ron minnich]

On Thu, 9 Oct 2003, ron minnich wrote:

> On Thu, 9 Oct 2003, Russ Cox wrote:
>
> > I think you have the Venti variable wrong.
> > It needs to be
> >
> > 	venti=tcp!127.1!17034 tcp!127.1!80
>
> well, ok, but looking at some of the source it looked like processes were
> trying to talk to this:
> tcp!$venti!venti
>
> which makes some sense.
>
> Anyway I did just this this exact setting in plan9.ini:
> venti=tcp!127.1!17034 tcp!127.1!80
> [is that really right?]
>

hmm.

OK, I can do this after booting without any venti set up in plan9.ini:

venti/venti -w
vac -h 172.15.239.128 ...

but not

vac -h 127.1 ...

Strikes me as odd because I can
telnet tcp!127.1!venti
just fine.

But the vac -h 127.1 hangs in Syn_sent (as shown by netstat) and finally
times out.

Time to go sleep more. Probably a silly /lib/ndb/local mis-settting.

ron

Re: [9fans] bleeding edge cd install

[mirtchov]

> I think you have the Venti variable wrong.
> It needs to be

shouldn't there be a loopback address configured before venti can
start on 127.1?  my boot file for pcfl has the following:

	ipconfig loopback /dev/null 127.1

just before venti...

would this work if put in plan9.ini:

	bootargs=tcp loopback /dev/null 127.1

?

then the other ip address can be configured later via dhcp?  or maybe
it has to be configured immediately because fossil needs it...  ugh,
complicated...

Re: [9fans] bleeding edge cd install

[ron minnich]

On Thu, 9 Oct 2003, David Presotto wrote:

> try it without venti.  I think I'm also running a bleeding edge fossil but
> with no venti and mine doesn't go tits up.

yup, it's working fine minus venti, I've got one bootmenu entry with and
one without, and the one with is the one that always dies.

So this is a known problem and not something I've set up wrong? (well,
in reality, it could be both ...)

ron

Re: [9fans] bleeding edge cd install

[David Presotto]

[-- Attachment #1: Type: text/plain, Size: 211 bytes --]

Not a known problem, I just don't happen to run a venti and was running
a new kernel with a fossil so I figured it was worth a try.  Now you
have to figure out what was wrong before.  A crashing PC would help.

[-- Attachment #2: Type: message/rfc822, Size: 2816 bytes --]

From: ron minnich <rminnich@lanl.gov>
To: 9fans@cse.psu.edu
Subject: Re: [9fans] bleeding edge cd install
Date: Thu, 9 Oct 2003 20:03:53 -0600 (MDT)
Message-ID: <Pine.LNX.4.44.0310092001120.7007-100000@maxroach.lanl.gov>

On Thu, 9 Oct 2003, David Presotto wrote:

> try it without venti.  I think I'm also running a bleeding edge fossil but
> with no venti and mine doesn't go tits up.

yup, it's working fine minus venti, I've got one bootmenu entry with and
one without, and the one with is the one that always dies.

So this is a known problem and not something I've set up wrong? (well,
in reality, it could be both ...)

ron