* Kernel boot arguments over network
@ 1999-03-01 15:56 David.DeRidder
1999-03-01 20:58 ` Guy Jennings
1999-03-02 9:45 ` Gabriel Paubert
0 siblings, 2 replies; 11+ messages in thread
From: David.DeRidder @ 1999-03-01 15:56 UTC (permalink / raw)
To: linuxppc-dev
Hi,
in my continueing efforts to get our MVME2306 board running, I ran
across another problem. All Linux (mini-)HOWTOs talk about passing
boot arguments (like 'root=/dev/nfs') to the kernel. Some also
mention the case where one is booting a diskless station (such as
an MVME230x board). However, no document describes *how* to pass these
arguments over the network to the kernel.
I'm starting to doubt whether this is possible. I want the BOOTP/TFTP
server to have the initial boot arguments for each of its clients, and
pass them when a client (i.c. an MVME2306 board) boots over the
network. Other than the file name, BOOTP doesn't seem capable to me
to passing anything as long as a command line - but I may be wrong.
I am aware of the fact that both 'NBO' and 'NIOT' allow setting some
"Argument File Name". Is this supposed to be an ASCII file with the
arguments in them ? The PPC-Bug documentation is rather vague about
this matter. (In my TFTP logs, nothing seems to indicate that the
board is trying to load this 'Argument File'.)
All help and suggestions are welcomed,
+-----------------------------------------------------------------+
| David 'Septimus' De Ridder | "Never underestimate the |
| | power of the soul unleashed." |
| <david.de.ridder@bitsmart.com> | - 'Essence Flare', M:TG |
+-----------------------------------------------------------------+
--
"- Do you think this show constitutes cruelty to animals ?
- Not unless their watching it !"
- 'The Muppet Show'
[[ This message was sent via the linuxppc-dev mailing list. Replies are ]]
[[ not forced back to the list, so be sure to Cc linuxppc-dev if your ]]
[[ reply is of general interest. To unsubscribe from linuxppc-dev, send ]]
[[ the message 'unsubscribe' to linuxppc-dev-request@lists.linuxppc.org ]]
^ permalink raw reply [flat|nested] 11+ messages in thread
* Re: Kernel boot arguments over network
1999-03-01 15:56 Kernel boot arguments over network David.DeRidder
@ 1999-03-01 20:58 ` Guy Jennings
1999-03-02 9:38 ` David De Ridder
1999-03-02 9:45 ` Gabriel Paubert
1 sibling, 1 reply; 11+ messages in thread
From: Guy Jennings @ 1999-03-01 20:58 UTC (permalink / raw)
To: loopback; +Cc: linuxppc-dev@lists.linuxppc.org
Hi,
While I'm not sure it's absolutely what you want, you might find
something useful
from Richard Hirst's site
http://www.sleepie.demon.co.uk/linuxvme/index.html
which describes linux for the M68K-based VME cards. This has a link to
a version of lilo that can be used with diskless systems.
Hope this helps,
Guy Jennings
Argonne National Laboratory
[[ This message was sent via the linuxppc-dev mailing list. Replies are ]]
[[ not forced back to the list, so be sure to Cc linuxppc-dev if your ]]
[[ reply is of general interest. To unsubscribe from linuxppc-dev, send ]]
[[ the message 'unsubscribe' to linuxppc-dev-request@lists.linuxppc.org ]]
^ permalink raw reply [flat|nested] 11+ messages in thread
* Re: Kernel boot arguments over network
1999-03-01 20:58 ` Guy Jennings
@ 1999-03-02 9:38 ` David De Ridder
0 siblings, 0 replies; 11+ messages in thread
From: David De Ridder @ 1999-03-02 9:38 UTC (permalink / raw)
To: Guy Jennings; +Cc: linuxppc-dev@lists.linuxppc.org
At 14:58 1-3-99 -0600, Guy Jennings wrote:
>While I'm not sure it's absolutely what you want, you might find
>something useful
>from Richard Hirst's site
>http://www.sleepie.demon.co.uk/linuxvme/index.html
>which describes linux for the M68K-based VME cards. This has a link to
>a version of lilo that can be used with diskless systems.
Thanks for the URL but...the MVME230x series are all PPC-based :-(
I'll check it out anyway,
+-----------------------------------------------------------------+
David 'Septimus' De Ridder <david.de.ridder@bitsmart.com>
"Philosophy is not a supermarket."
+-----------------------------------------------------------------+
[[ This message was sent via the linuxppc-dev mailing list. Replies are ]]
[[ not forced back to the list, so be sure to Cc linuxppc-dev if your ]]
[[ reply is of general interest. To unsubscribe from linuxppc-dev, send ]]
[[ the message 'unsubscribe' to linuxppc-dev-request@lists.linuxppc.org ]]
^ permalink raw reply [flat|nested] 11+ messages in thread
* Re: Kernel boot arguments over network
1999-03-01 15:56 Kernel boot arguments over network David.DeRidder
1999-03-01 20:58 ` Guy Jennings
@ 1999-03-02 9:45 ` Gabriel Paubert
1999-03-02 13:34 ` David.DeRidder
1999-03-02 14:38 ` David.DeRidder
1 sibling, 2 replies; 11+ messages in thread
From: Gabriel Paubert @ 1999-03-02 9:45 UTC (permalink / raw)
To: loopback; +Cc: linuxppc-dev
On Mon, 1 Mar 1999, David.DeRidder wrote:
> Hi,
>
> in my continueing efforts to get our MVME2306 board running, I ran
> across another problem. All Linux (mini-)HOWTOs talk about passing
> boot arguments (like 'root=/dev/nfs') to the kernel. Some also
> mention the case where one is booting a diskless station (such as
> an MVME230x board). However, no document describes *how* to pass these
> arguments over the network to the kernel.
You don't have to (and can't AFAICT). For now on my MVME2302, I have to
sit at the serial console and add root=/dev/nfs to the LinuxPPC boot
prompt. If you don't have a bootp server correctly settup, you also have
to add other parameters like internet addresses in the ip or nfsaddrs
parameter as explained in Documentation/nfsroot.txt.
When I have time, I'll merge code to take default parameters from the
NVRAM in prepboot. The code exists, but I'm too busy on the VME driver
right now.
> I'm starting to doubt whether this is possible. I want the BOOTP/TFTP
> server to have the initial boot arguments for each of its clients, and
> pass them when a client (i.c. an MVME2306 board) boots over the
> network. Other than the file name, BOOTP doesn't seem capable to me
> to passing anything as long as a command line - but I may be wrong.
The right solution is to use the NVRAM I think, at least for very early
setup since you'll end up with a chicken and egg problem (you have to
tell the kernel root=/dev/nfs so that it attempts to use bootp to
configure its network interface).
BTW: the latest version on my ftp server is a tested version which I have
actually booted.
Gabriel.
[[ This message was sent via the linuxppc-dev mailing list. Replies are ]]
[[ not forced back to the list, so be sure to Cc linuxppc-dev if your ]]
[[ reply is of general interest. To unsubscribe from linuxppc-dev, send ]]
[[ the message 'unsubscribe' to linuxppc-dev-request@lists.linuxppc.org ]]
^ permalink raw reply [flat|nested] 11+ messages in thread
* Re: Kernel boot arguments over network
[not found] <000801be6425$b23751d0$0100a8c0@mikem>
@ 1999-03-02 9:46 ` David De Ridder
0 siblings, 0 replies; 11+ messages in thread
From: David De Ridder @ 1999-03-02 9:46 UTC (permalink / raw)
To: Mike McQuade; +Cc: linuxppc-dev
Mike McQuade, with bold words breaking the horrid silence thus began :
>The params you pass in are at the Linux boot prompt,
>you see this prompt at the serial term right after
>you start the boot process.
But therefore, the kernel needs a serial console at boot -
which it doesn't have. With CONF_SERIAL_ECHO, it is possible
to echo all kernel messages (e.g. the ones you see when linux
boots) to a serial port, but it's one-way traffic. You cannot
type anything before the file system is mounted (because you
need getty or the like to create a serial console).
The MVME230x series boards have NO harddisk, so using LILO or
LoadLin doesn't seem possible.
>I rmember having to modify one of the C files in the kernel
>to make the boot device == NFS, without this mod,
>the kernel would panic.
Doesn't the option 'root=/dev/nfs' do the same ?
>do you get anything out the serial port on startup ?
Yes, the kernel I have (from Gabriel Paubert) gets as far as
'Now booting...', and then hangs. Does anyone know what phase should
be next ?
Regards,
+-----------------------------------------------------------------+
David 'Septimus' De Ridder <david.de.ridder@bitsmart.com>
"Philosophy is not a supermarket."
+-----------------------------------------------------------------+
[[ This message was sent via the linuxppc-dev mailing list. Replies are ]]
[[ not forced back to the list, so be sure to Cc linuxppc-dev if your ]]
[[ reply is of general interest. To unsubscribe from linuxppc-dev, send ]]
[[ the message 'unsubscribe' to linuxppc-dev-request@lists.linuxppc.org ]]
^ permalink raw reply [flat|nested] 11+ messages in thread
* Re: Kernel boot arguments over network
1999-03-02 9:45 ` Gabriel Paubert
@ 1999-03-02 13:34 ` David.DeRidder
1999-03-02 13:45 ` Gabriel Paubert
1999-03-02 14:38 ` David.DeRidder
1 sibling, 1 reply; 11+ messages in thread
From: David.DeRidder @ 1999-03-02 13:34 UTC (permalink / raw)
To: Gabriel Paubert; +Cc: loopback, linuxppc-dev
On Tue, 2 Mar 1999, Gabriel Paubert wrote:
> > an MVME230x board). However, no document describes *how* to pass these
> > arguments over the network to the kernel.
>
> You don't have to (and can't AFAICT). For now on my MVME2302, I have to
> sit at the serial console and add root=/dev/nfs to the LinuxPPC boot
Do you mean adding it to the NBO command ? When I boot the kernel with
NBO, it loads over the network (BOOTP and TFTP work perfectly fine),
but does not ask for any arguments. It does show the arguments it knows
(because they are set in the kernel image), i.e. "console=ttyS0").
> prompt. If you don't have a bootp server correctly settup, you also have
> to add other parameters like internet addresses in the ip or nfsaddrs
> parameter as explained in Documentation/nfsroot.txt.
That part (done by PPC-Bug) works perfectly fine. nfsaddrs should be
a kernel option, no ? Anyway, I also pass it on with some BOOTP
parameter (root directory or so).
> When I have time, I'll merge code to take default parameters from the
> NVRAM in prepboot. The code exists, but I'm too busy on the VME driver
> right now.
What is prepboot ?
> The right solution is to use the NVRAM I think, at least for very early
> setup since you'll end up with a chicken and egg problem (you have to
> tell the kernel root=/dev/nfs so that it attempts to use bootp to
> configure its network interface).
Well, I tried using the "Argument File Name" setting (NIOT in PPC-Bug),
but it resulted in no improvement.
> BTW: the latest version on my ftp server is a tested version which I have
> actually booted.
I will download it and test it. No doubt I will get back with the
results real soon.
Thanks,
+-----------------------------------------------------------------+
| David 'Septimus' De Ridder | "Never underestimate the |
| | power of the soul unleashed." |
| <david.de.ridder@bitsmart.com> | - 'Essence Flare', M:TG |
+-----------------------------------------------------------------+
--
"When a lord goes hunting, it is easier to release his hounds
than to catch them again."
- Murashige
[[ This message was sent via the linuxppc-dev mailing list. Replies are ]]
[[ not forced back to the list, so be sure to Cc linuxppc-dev if your ]]
[[ reply is of general interest. To unsubscribe from linuxppc-dev, send ]]
[[ the message 'unsubscribe' to linuxppc-dev-request@lists.linuxppc.org ]]
^ permalink raw reply [flat|nested] 11+ messages in thread
* Re: Kernel boot arguments over network
1999-03-02 13:34 ` David.DeRidder
@ 1999-03-02 13:45 ` Gabriel Paubert
1999-03-02 14:16 ` David.DeRidder
0 siblings, 1 reply; 11+ messages in thread
From: Gabriel Paubert @ 1999-03-02 13:45 UTC (permalink / raw)
To: loopback; +Cc: linuxppc-dev
On Tue, 2 Mar 1999, David.DeRidder wrote:
> Do you mean adding it to the NBO command ? When I boot the kernel with
> NBO, it loads over the network (BOOTP and TFTP work perfectly fine),
> but does not ask for any arguments. It does show the arguments it knows
> (because they are set in the kernel image), i.e. "console=ttyS0").
No, the console=ttyS0 is actually a prompt which asks you for possible
additional parameters, you have to type the additional root=/dev/nfs. This
is a trick in my current bootloader (which should disappear soon) that
sets "console=ttyS0" as a default command line when it can't find a
keyboard. You may append there additional kernel options.
> That part (done by PPC-Bug) works perfectly fine. nfsaddrs should be
> a kernel option, no ? Anyway, I also pass it on with some BOOTP
> parameter (root directory or so).
No, it is not a kernel option. AFAIR, you could build a kernel with
default root set to NFS but network configuration parameters have to be
set either by bootp or by the nfsaddrs (now ip) kernel options.
> What is prepboot ?
The early part of the boot for PreP machines that I've written.
> Well, I tried using the "Argument File Name" setting (NIOT in PPC-Bug),
> but it resulted in no improvement.
No, this won't work. I don't even remember if it is passed to the loaded
program when you select the "Network PreP boot mode enable".
> I will download it and test it. No doubt I will get back with the
> results real soon.
Ok, waiting for you to test it.
Gabriel
[[ This message was sent via the linuxppc-dev mailing list. Replies are ]]
[[ not forced back to the list, so be sure to Cc linuxppc-dev if your ]]
[[ reply is of general interest. To unsubscribe from linuxppc-dev, send ]]
[[ the message 'unsubscribe' to linuxppc-dev-request@lists.linuxppc.org ]]
^ permalink raw reply [flat|nested] 11+ messages in thread
* Re: Kernel boot arguments over network
1999-03-02 13:45 ` Gabriel Paubert
@ 1999-03-02 14:16 ` David.DeRidder
1999-03-02 15:29 ` Gabriel Paubert
0 siblings, 1 reply; 11+ messages in thread
From: David.DeRidder @ 1999-03-02 14:16 UTC (permalink / raw)
To: Gabriel Paubert; +Cc: linuxppc-dev
On Tue, 2 Mar 1999, Gabriel Paubert wrote:
> On Tue, 2 Mar 1999, David.DeRidder wrote:
>
> > Do you mean adding it to the NBO command ? When I boot the kernel with
> > NBO, it loads over the network (BOOTP and TFTP work perfectly fine),
> > but does not ask for any arguments. It does show the arguments it knows
> > (because they are set in the kernel image), i.e. "console=ttyS0").
>
> No, the console=ttyS0 is actually a prompt which asks you for possible
> additional parameters, you have to type the additional root=/dev/nfs. This
Now I see. I wasn't aware the bootloader was waiting for input - I just
thought it was doing something. There is a timeout to this prompt,
right ?
> is a trick in my current bootloader (which should disappear soon) that
> sets "console=ttyS0" as a default command line when it can't find a
> keyboard. You may append there additional kernel options.
Pretty neat that it can take input from a serial keyboard in that stage
of the boot procedure. I thought it was just some kernel info, echoed
due to CONF_SERIAL_ECHO. But instead, it is prepboot, right ?
> > That part (done by PPC-Bug) works perfectly fine. nfsaddrs should be
> > a kernel option, no ? Anyway, I also pass it on with some BOOTP
> > parameter (root directory or so).
>
> No, it is not a kernel option. AFAIR, you could build a kernel with
> default root set to NFS but network configuration parameters have to be
> set either by bootp or by the nfsaddrs (now ip) kernel options.
My mistake : I meant 'boot argument'. You are completely right.
> > Well, I tried using the "Argument File Name" setting (NIOT in PPC-Bug),
> > but it resulted in no improvement.
>
> No, this won't work. I don't even remember if it is passed to the loaded
> program when you select the "Network PreP boot mode enable".
It seems not. Does anybody out there know what this is for ? The PPC-Bug
manual is very uninformative on this matter.
> > I will download it and test it. No doubt I will get back with the
> > results real soon.
>
> Ok, waiting for you to test it.
Tested it, and it still is stuck at 'Now booting...' Does the kernel
look for its root filesystem before giving any output ? If so, it
might be the case that the kernel doesn't find its root fs (but
I'd expect at least a message reporting that is the case). If not,
there must be another problem...
+-----------------------------------------------------------------+
| David 'Septimus' De Ridder | "Never underestimate the |
| | power of the soul unleashed." |
| <david.de.ridder@bitsmart.com> | - 'Essence Flare', M:TG |
+-----------------------------------------------------------------+
--
"Every dog has its day, and today's mine."
Terry Borer after winning the Masters finals at the New York M:TG Pro Tour
[[ This message was sent via the linuxppc-dev mailing list. Replies are ]]
[[ not forced back to the list, so be sure to Cc linuxppc-dev if your ]]
[[ reply is of general interest. To unsubscribe from linuxppc-dev, send ]]
[[ the message 'unsubscribe' to linuxppc-dev-request@lists.linuxppc.org ]]
^ permalink raw reply [flat|nested] 11+ messages in thread
* Re: Kernel boot arguments over network
1999-03-02 9:45 ` Gabriel Paubert
1999-03-02 13:34 ` David.DeRidder
@ 1999-03-02 14:38 ` David.DeRidder
1 sibling, 0 replies; 11+ messages in thread
From: David.DeRidder @ 1999-03-02 14:38 UTC (permalink / raw)
To: Gabriel Paubert; +Cc: linuxppc-dev
On Tue, 2 Mar 1999, Gabriel Paubert wrote:
> > in my continueing efforts to get our MVME2306 board running, I ran
> > across another problem. All Linux (mini-)HOWTOs talk about passing
> > boot arguments (like 'root=/dev/nfs') to the kernel. Some also
> > mention the case where one is booting a diskless station (such as
> > an MVME230x board). However, no document describes *how* to pass these
> > arguments over the network to the kernel.
>
> You don't have to (and can't AFAICT). For now on my MVME2302, I have to
> sit at the serial console and add root=/dev/nfs to the LinuxPPC boot
For the record, there is no difference as far as the kernel is concerned
between an MPC603 (as in your MVME2302) and an MPC604 (as in our
MVME2306) ?
+-----------------------------------------------------------------+
| David 'Septimus' De Ridder | "Never underestimate the |
| | power of the soul unleashed." |
| <david.de.ridder@bitsmart.com> | - 'Essence Flare', M:TG |
+-----------------------------------------------------------------+
--
"Every dog has its day, and today's mine."
Terry Borer after winning the Masters finals at the New York M:TG Pro Tour
[[ This message was sent via the linuxppc-dev mailing list. Replies are ]]
[[ not forced back to the list, so be sure to Cc linuxppc-dev if your ]]
[[ reply is of general interest. To unsubscribe from linuxppc-dev, send ]]
[[ the message 'unsubscribe' to linuxppc-dev-request@lists.linuxppc.org ]]
^ permalink raw reply [flat|nested] 11+ messages in thread
* Re: Kernel boot arguments over network
1999-03-02 14:16 ` David.DeRidder
@ 1999-03-02 15:29 ` Gabriel Paubert
1999-03-02 16:00 ` David De Ridder
0 siblings, 1 reply; 11+ messages in thread
From: Gabriel Paubert @ 1999-03-02 15:29 UTC (permalink / raw)
To: loopback; +Cc: linuxppc-dev
On Tue, 2 Mar 1999, David.DeRidder wrote:
> Now I see. I wasn't aware the bootloader was waiting for input - I just
> thought it was doing something. There is a timeout to this prompt,
> right ?
Yes. Actually this was (partly) copied from a previous early boot, the
biggest change being an attempt to improve keyboard code.
> Pretty neat that it can take input from a serial keyboard in that stage
> of the boot procedure. I thought it was just some kernel info, echoed
> due to CONF_SERIAL_ECHO. But instead, it is prepboot, right ?
Indeed.
> > No, this won't work. I don't even remember if it is passed to the loaded
> > program when you select the "Network PreP boot mode enable".
>
> It seems not. Does anybody out there know what this is for ? The PPC-Bug
> manual is very uninformative on this matter.
I could never dig any information on this. And anyway, residual data are
absolutely necessary to get the interrupt routing through the OpenPIC in
the Raven.
> > Ok, waiting for you to test it.
>
> Tested it, and it still is stuck at 'Now booting...' Does the kernel
> look for its root filesystem before giving any output ? If so, it
> might be the case that the kernel doesn't find its root fs (but
> I'd expect at least a message reporting that is the case). If not,
> there must be another problem...
No, I would rather think that there is a problem with the universe. Plese
rerun the ENV command and answer 0 or N to all questions about the VMEPCI3
bridge. The default settings might interfere with the reallocation of
PCI resources (I'm working on making the Universe known to the PCI code
to avoid conflicts).
Gabriel.
[[ This message was sent via the linuxppc-dev mailing list. Replies are ]]
[[ not forced back to the list, so be sure to Cc linuxppc-dev if your ]]
[[ reply is of general interest. To unsubscribe from linuxppc-dev, send ]]
[[ the message 'unsubscribe' to linuxppc-dev-request@lists.linuxppc.org ]]
^ permalink raw reply [flat|nested] 11+ messages in thread
* Re: Kernel boot arguments over network
1999-03-02 15:29 ` Gabriel Paubert
@ 1999-03-02 16:00 ` David De Ridder
0 siblings, 0 replies; 11+ messages in thread
From: David De Ridder @ 1999-03-02 16:00 UTC (permalink / raw)
To: Gabriel Paubert; +Cc: linuxppc-dev
On Tue, 2 Mar 1999, Gabriel Paubert wrote:
> On Tue, 2 Mar 1999, David.DeRidder wrote:
>
> > I'd expect at least a message reporting that is the case). If not,
> > there must be another problem...
>
> No, I would rather think that there is a problem with the universe. Plese
> rerun the ENV command and answer 0 or N to all questions about the VMEPCI3
That's it !! There is a setting called "VME3PCI Master Master Enable".
This is the culprit for the kernel not booting. It must be set to 'N'
for the kernel to boot. (That, and setting the PReP mode correctly,
as discussed previously on the list). The kernel now comes up, as
it should !
> bridge. The default settings might interfere with the reallocation of
> PCI resources (I'm working on making the Universe known to the PCI code
> to avoid conflicts).
Well, I'm still having a problem for the kernel to gain access to
the NFS server - but that seems a minor issue. (They talk to each
other, but the NFS server is reluctant in giving access - /etc/exports
seems to be correct).
I'll keep you posted about our progress.
+-----------------------------------------------------------------+
| David 'Septimus' De Ridder | "Never underestimate the |
| | power of the soul unleashed." |
| <david.de.ridder@bitsmart.com> | - 'Essence Flare', M:TG |
+-----------------------------------------------------------------+
--
"Every dog has its day, and today's mine."
Terry Borer after winning the Masters finals at the New York M:TG Pro Tour
[[ This message was sent via the linuxppc-dev mailing list. Replies are ]]
[[ not forced back to the list, so be sure to Cc linuxppc-dev if your ]]
[[ reply is of general interest. To unsubscribe from linuxppc-dev, send ]]
[[ the message 'unsubscribe' to linuxppc-dev-request@lists.linuxppc.org ]]
^ permalink raw reply [flat|nested] 11+ messages in thread
end of thread, other threads:[~1999-03-02 16:00 UTC | newest]
Thread overview: 11+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
1999-03-01 15:56 Kernel boot arguments over network David.DeRidder
1999-03-01 20:58 ` Guy Jennings
1999-03-02 9:38 ` David De Ridder
1999-03-02 9:45 ` Gabriel Paubert
1999-03-02 13:34 ` David.DeRidder
1999-03-02 13:45 ` Gabriel Paubert
1999-03-02 14:16 ` David.DeRidder
1999-03-02 15:29 ` Gabriel Paubert
1999-03-02 16:00 ` David De Ridder
1999-03-02 14:38 ` David.DeRidder
[not found] <000801be6425$b23751d0$0100a8c0@mikem>
1999-03-02 9:46 ` David De Ridder
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).