Re: [9fans] Venti Install

Re: [9fans] Venti Install

[YAMANASHI Takeshi]

> > > > > Installing a Venti system with plan9.iso sucks the big one. I hear
> > > > > that Russ has magic code which makes it not suck. Can this PLEASE get
> > > > > put in so that the next time I do this, I don't have to wait N hours?

If you install venti next time, give it a small arena partition.
400MB should be fine for the dump right after the installation.
You can format another bigger partition and add it later.

# am installing venti on 200GB disk now
-- 

Re: [9fans] Venti Install

[Bruce Ellis]

that's a smart hint.  tho going to the pub while it does its stuff
is tastier. my limbo venti is "format on demand" which seems
more sensible.  who cares if the rest of the arena space contains
crap?  i don't - as long as it's ready for use when needed.

brucee

On 3/26/07, YAMANASHI Takeshi <9.nashi@gmail.com> wrote:
> > > > > > Installing a Venti system with plan9.iso sucks the big one. I hear
> > > > > > that Russ has magic code which makes it not suck. Can this PLEASE get
> > > > > > put in so that the next time I do this, I don't have to wait N hours?
>
> If you install venti next time, give it a small arena partition.
> 400MB should be fine for the dump right after the installation.
> You can format another bigger partition and add it later.
>
> # am installing venti on 200GB disk now
> --
>
>

Re: [9fans] Venti Install

[Charles Forsyth]

several other people have made helpful suggestions about
coping with the time to format, but i thought i'd include
a reminder to ensure dma and rwm are switched on.
that might require editing the switch cases in sdata.c(?).
that upgrades the original PC controller scheme to work nearly as well as an RL02.
the few times i've formatted venti partitions, i've found that
made a big difference to the elapsed time, in the sense that i
went from `how many days will this take??' to `not too bad'.

Re: [9fans] Venti Install

[Bruce Ellis]

yeah well i chose O(1) over O(n) because n is getting really big.

brucee

On 3/26/07, Charles Forsyth <forsyth@terzarima.net> wrote:
> several other people have made helpful suggestions about
> coping with the time to format, but i thought i'd include
> a reminder to ensure dma and rwm are switched on.
> that might require editing the switch cases in sdata.c(?).
> that upgrades the original PC controller scheme to work nearly as well as an RL02.
> the few times i've formatted venti partitions, i've found that
> made a big difference to the elapsed time, in the sense that i
> went from `how many days will this take??' to `not too bad'.
>

Re: [9fans] Venti Install

[YAMANASHI Takeshi]

Yeah.  DMA helps a lot on venti formatting.
However, I saw this ominous message on my console when
installing another tip9ug host on VMware ESX 3.0 Server
which only supports Buslogic SCSI adaptor.

	init: starting /bin/rc
	echo: write error: bad process or channel control request
	echo: write error: bad process or channel control request

I guess someone tried to turn DMA on for the SCSI drive and failed.
-- 

Re: [9fans] Venti Install

[erik quanstrom]

are you sure this is a DMA problem.  there are lots of 
places in the kernel that return Ebadctl.

pc/devether.c:266: 		error(Ebadctl);
pc/ethersink.c:31: 	error(Ebadctl);
pc/ethervgbe.c:1082: 		error(Ebadctl);
pc/sdata.c:2202: 			error(Ebadctl);
pc/sdata.c:2208: 			error(Ebadctl);
pc/sdata.c:2212: 			error(Ebadctl);
pc/sdata.c:2218: 			error(Ebadctl);
pc/sdata.c:2223: 			error(Ebadctl);
pc/sdata.c:2227: 				error(Ebadctl);
pc/sdata.c:2232: 			error(Ebadctl);
pc/sdata.c:2236: 			error(Ebadctl);
pc/sdata.c:2242: 			error(Ebadctl);
pc/sdata.c:2245: 		error(Ebadctl);
pc/sdmv50xx.c:1170: 	cmderror(cb, Ebadctl);
pc/wavelan.c:1120: 		error(Ebadctl);
port/devpnp.c:615: 			error(Ebadctl);
port/devpnp.c:623: 			error(Ebadctl);
port/devproc.c:1056: 			error(Ebadctl);
port/devproc.c:1357: 			error(Ebadctl);
port/devproc.c:1367: 			error(Ebadctl);
port/devproc.c:1372: 			error(Ebadctl);
port/devproc.c:1379: 			error(Ebadctl);
port/devsd.c:97: 				error(Ebadctl);
port/devsd.c:155: 		error(Ebadctl);
port/devsd.c:1230: 			error(Ebadctl);
port/devsd.c:1255: 			error(Ebadctl);
port/devsd.c:1258: 				error(Ebadctl);
port/devsd.c:1267: 				error(Ebadctl);
port/devsd.c:1273: 			error(Ebadctl);
port/devsegment.c:387: 			error(Ebadctl);
port/error.h:25: extern char Ebadctl[];		/* bad process or channel control request */
ppc/devether.c:256: 		error(Ebadctl);


YAMANASHI Takeshi <9.nashi@gmail.com> writes

| 
| Yeah.  DMA helps a lot on venti formatting.
| However, I saw this ominous message on my console when
| installing another tip9ug host on VMware ESX 3.0 Server
| which only supports Buslogic SCSI adaptor.
| 
| 	init: starting /bin/rc
| 	echo: write error: bad process or channel control request
| 	echo: write error: bad process or channel control request
| 
| I guess someone tried to turn DMA on for the SCSI drive and failed.

Re: [9fans] Venti Install

[Russ Cox]

The lines that were quoted:

       init: starting /bin/rc
       echo: write error: bad process or channel control request
       echo: write error: bad process or channel control request

are almost always caused by trying to enable dma
when it's not available.

In the case of VMware, I believe DMA is never available,
since they handle the REP INSL instruction as efficiently
as they could simulate DMA anyway.  I know this is the
case for the simulated IDE.  I am not as sure about
simulated SCSI.

Russ

[9fans] Venti Install

[Devon H. O'Dell]

Hi,

Installing a Venti system with plan9.iso sucks the big one. I hear
that Russ has magic code which makes it not suck. Can this PLEASE get
put in so that the next time I do this, I don't have to wait N hours?

Thanks,

Devon

Re: [9fans] Venti Install

[Russ Cox]

> Installing a Venti system with plan9.iso sucks the big one. I hear
> that Russ has magic code which makes it not suck. Can this PLEASE get
> put in so that the next time I do this, I don't have to wait N hours?

At the moment, you have the choice between slow, stable, tested code
and fast, unstable, less tested code.  The slow code is in the distribution.
I'm sorry venti installs are slow, but the new code still has problems
that need to be worked out.

Russ

Re: [9fans] Venti Install

[Devon H. O'Dell]

2007/3/19, Russ Cox <rsc@swtch.com>:
> > Installing a Venti system with plan9.iso sucks the big one. I hear
> > that Russ has magic code which makes it not suck. Can this PLEASE get
> > put in so that the next time I do this, I don't have to wait N hours?
>
> At the moment, you have the choice between slow, stable, tested code
> and fast, unstable, less tested code.  The slow code is in the distribution.
> I'm sorry venti installs are slow, but the new code still has problems
> that need to be worked out.

Do you have an ISO I can try?

> Russ

--dho

Re: [9fans] Venti Install

[Russ Cox]

On 3/19/07, Devon H. O'Dell <devon.odell@gmail.com> wrote:
> 2007/3/19, Russ Cox <rsc@swtch.com>:
> > > Installing a Venti system with plan9.iso sucks the big one. I hear
> > > that Russ has magic code which makes it not suck. Can this PLEASE get
> > > put in so that the next time I do this, I don't have to wait N hours?
> >
> > At the moment, you have the choice between slow, stable, tested code
> > and fast, unstable, less tested code.  The slow code is in the distribution.
> > I'm sorry venti installs are slow, but the new code still has problems
> > that need to be worked out.
>
> Do you have an ISO I can try?

No, the code only exists in p9p.
I have not brought it up on Plan 9 proper.

Russ

Re: [9fans] Venti Install

[Devon H. O'Dell]

2007/3/19, Russ Cox <rsc@swtch.com>:
> On 3/19/07, Devon H. O'Dell <devon.odell@gmail.com> wrote:
> > 2007/3/19, Russ Cox <rsc@swtch.com>:
> > > > Installing a Venti system with plan9.iso sucks the big one. I hear
> > > > that Russ has magic code which makes it not suck. Can this PLEASE get
> > > > put in so that the next time I do this, I don't have to wait N hours?
> > >
> > > At the moment, you have the choice between slow, stable, tested code
> > > and fast, unstable, less tested code.  The slow code is in the distribution.
> > > I'm sorry venti installs are slow, but the new code still has problems
> > > that need to be worked out.
> >
> > Do you have an ISO I can try?
>
> No, the code only exists in p9p.
> I have not brought it up on Plan 9 proper.
>
> Russ

Well that's a pity. I think it would get more testing as an option in
the install image. Anyway, I guess that's something for later. It
finished, but there was some i/o error, so I may be reinstalling
anyway in case it's something that can be put in some ISO for us
random users to test between now and then.

--dho

Re: [9fans] Venti Install

[john]

> Hi,
> 
> Installing a Venti system with plan9.iso sucks the big one. I hear
> that Russ has magic code which makes it not suck. Can this PLEASE get
> put in so that the next time I do this, I don't have to wait N hours?
> 
> Thanks,
> 
> Devon

When you talk about it sucking, are you referring to the time spent
formatting Venti partitions, or what? I know that takes forever and
really really sucks, but is it any different if you do it, say, on an
already-installed system?


John, not a Venti guru