etherboot-developers — Discussion of developer issues

Re: [Etherboot-developers] for gurus ;-) remote boot with tftp'd ramdisk ?

[Matt H. <mbh...@ie...>]

On Sat, Oct 07, 2000 at 04:17:48AM +0200, Pierre-Marie Baty wrote:
> Is it possible to boot a linux kernel over the network __without
> using NFS___, say using a ramdisk downloaded by tftp ? I don't mind
> how difficult it could be (kernel hacking,...), either a linux
> kernel or another *nix btw...  I have to set up a diskless node for
> an embedded motherboard which does not need to save any data to
> disk. If I can get rid of NFS I will be able to set up the boot
> server on any OS, including windows 9x.

Sure.  Using mknbi you can specify a ramdisk to be included in the NBI
and downloaded along with the kernel.  The kernel will then uncompress
and mount the ramdisk as /

No kernel hacking is involved.  Just build a compressed ramdisk
(rootfs.gz) as if you were building a Linux bootdisk (you can make it
bigger in this case) as described in the root filesystem chapter of
the Linux Bootdisk HOWOTO :

http://www.linuxdoc.org/HOWTO/Bootdisk-HOWTO/buildroot.html

Make sure your kernel is set to mount the ramdisk as root:

rdev kernel /dev/ram

Run mknbi with something like:

mknbi kernel rootfs.gz

You're done.  Boot the new NBI...

I use this method to boot my company's linux-based thinclients so I
know it'll work.

Hope that helps,

-Matt

[Etherboot-developers] for gurus ;-) remote boot with tftp'd ramdisk ?

[Pierre-Marie B. <ba...@li...>]

Is it possible to boot a linux kernel over the network __without using =
NFS___, say using a ramdisk downloaded by tftp ? I don't mind how =
difficult it could be (kernel hacking,...), either a linux kernel or =
another *nix btw...
I have to set up a diskless node for an embedded motherboard which does =
not need to save any data to disk. If I can get rid of NFS I will be =
able to set up the boot server on any OS, including windows 9x.

Any ideas would be greatly appreciated.

PM

[Etherboot-developers] Re: [Etherboot-users] NIC Database created

[Neil C. D. <ne...@nc...>]

Not sure what this error means after trying to submit a vote!


Fatal error: Call to undefined function: inlcude() in
/home/groups/etherboot/htdocs/db/vote_db.php on line 87


May be it would be nice to be able to include an email address of the
person who tested it if they want to be contacted.

Neil

Georg Baum wrote:
> 
> Hi,
> 
> the Etherboot Network Interface Card Database is now up and running at
> http://etherboot.sourceforge.net/db/
> 
> It provides technical information about NICs and users experiences in
> order to help users to decide which NIC to buy or to solve problems with
> a particular NIC. The reason to set up this database was that it was
> sometimes difficult to get all the information which is needed to make
> booting off a bootrom work.
> 
> To make a start, I loaded the database with some data which was from the
> etherboot source code and from the cards I have. The rest has to be
> entered by users. I provided a form at
> http://etherboot.sourceforge.net/db/change.php?table=nic&new=1
> There is also the possibility to use an existing card as a template, if
> the new one is rather similar (You will find a link at the bottom of
> every cards page to do that). You can also change the entry for a NIC,
> if something is wrong. To avoid double entries and to prevent that
> somebody deletes everything just for fun, the changes you make dont go
> immediately in the database. Instead, they must be confirmed by a member
> of the etherboot team.
> 
> Please have a look at the database, enter additional information if you
> can, and tell your experiences with the "vote"-Link which accompanies
> every card. Dont hesitate to tell me if you think something could be
> made better or if something does not work.
> 
> Thanks,
> 
> Georg
> _______________________________________________
> Etherboot-users mailing list
> Eth...@li...
> http://lists.sourceforge.net/mailman/listinfo/etherboot-users

[Etherboot-developers] NIC Database created

[Georg B. <Geo...@gm...>]

Hi,

the Etherboot Network Interface Card Database is now up and running at
http://etherboot.sourceforge.net/db/

It provides technical information about NICs and users experiences in
order to help users to decide which NIC to buy or to solve problems with
a particular NIC. The reason to set up this database was that it was
sometimes difficult to get all the information which is needed to make
booting off a bootrom work. 

To make a start, I loaded the database with some data which was from the
etherboot source code and from the cards I have. The rest has to be
entered by users. I provided a form at
http://etherboot.sourceforge.net/db/change.php?table=nic&new=1
There is also the possibility to use an existing card as a template, if
the new one is rather similar (You will find a link at the bottom of
every cards page to do that). You can also change the entry for a NIC,
if something is wrong. To avoid double entries and to prevent that
somebody deletes everything just for fun, the changes you make dont go
immediately in the database. Instead, they must be confirmed by a member
of the etherboot team.

Please have a look at the database, enter additional information if you
can, and tell your experiences with the "vote"-Link which accompanies
every card. Dont hesitate to tell me if you think something could be
made better or if something does not work.

Thanks,

Georg

[Etherboot-developers] Etherboot 4.6.6 released

[Ken Y. <ke...@nl...>]

I have released Etherboot 4.6.6 at http://etherboot.sourceforge.net

This is essentially a backpropagation of most of the features in the
development release 4.7.7, those that were deemed non-disruptive.
Changes from 4.6.5 are:

+ New driver for Winbond W89C840 by Igor V. Kovalenko.

+ Bug in mknbi-dos, last sector omitted in ramdisk image.

+ Use ceil() for computing number of rootdir sectors.

+ --harddisk now works for mknbi-dos also.

+ Should work on FAT16 partitions now.

+ Small bug in osloader.c, j not defined in two places if DELIMITERLINES
defined.

+ Removed -nostdlib from ld flags. Seems to be a vestige from when we
were calling gcc as it's not a ld option.

MD5 sums:

59d06e3d4ec8697ac1a8588fdb13a3b6  etherboot-4.6.6.tar.bz2
e0cb6d2c7748f8fe2fc9edb55ecb25f2  etherboot-4.6.6.tar.gz
e89b8ca1d4d3dbf0fc04807edb4f10a0  patch-4.6.6.gz

[Etherboot-developers] Etherboot 4.7.6 released

[Ken Y. <ke...@nl...>]

I have released Etherboot 4.7.6 at http://etherboot.sourceforge.net
As the odd minor number indicates, this is a developmental release with
a couple of changes that need testing. They are:

+ New driver for Winbond W89C840 by Igor V. Kovalenko.

+ Modification to loader.S by Steve Smith for some PCI BIOSes that don't
handle INT19H well.

There are also some small changes, please see the LOG file.

If they test out well, the changes will be backpropagated to 4.6.6. So
please give this release a go if you can.

MD5 sums:

bc8722890401de9095fd0de5c057a68f  etherboot-4.7.6.tar.bz2
6136a7bb004bc24d89290741f003f01d  etherboot-4.7.6.tar.gz

[Etherboot-developers] Winbond W89c840F chip NICs - a new driver / patch uploaded

[Igor V. K. <gar...@ma...>]

... was uploaded to etherboot patches page on SourceForge.

Re[2]: [Etherboot-developers] Winbond W89c840F chip NICs - a new driver

[éÇÏÒØ ëÏ×Á. <gar...@ma...>]

-----Original Message-----
From: ke...@nl...
To: "Igor V. Kovalenko" <gar...@ma...>
Date: 23 Aug 2000 23:29:41 +1000
Subject: Re: [Etherboot-developers] Winbond W89c840F chip NICs - a new driver

> Quoting "Igor V. Kovalenko" <gar...@ma...>:
> > Hi there!
> >
> > I have a bunch of Compex RL100ATX cards there
> > and I needed an Etherboot driver.
> >
> > So now I have one and it should support
> >
> > Compex RL100ATX card
> > Winbond W89c840 card
> >
> > The driver is based on Donald Becker's Linux kernel
> > driver, as usual :)
> >
> > The main problem was whether to use I/O for register access and
> > how to work with device on second PCI bus. So the patch have
> > +    unsigned short  pci_bus_no;
> > added to pci_device structure to remember PCI bus number of device.
> > On our i810-based motherboards the card is on PCI bus 1 (not 0).
> 
> Excellent work Igor! So you have it working? Send me the driver and any patches
> and I will include it in the next release 4.6.6.
> 
> ------------------------------------------------------
> This mail sent via NLC WebMail: http://www.nlc.net.au/
> 

So there is a diff against 4.6.5 attached.

Re: [Etherboot-developers] Winbond W89c840F chip NICs - a new driver

[<ke...@nl...>]

Quoting "Igor V. Kovalenko" <gar...@ma...>:
> Hi there!
>
> I have a bunch of Compex RL100ATX cards there
> and I needed an Etherboot driver.
>
> So now I have one and it should support
>
> Compex RL100ATX card
> Winbond W89c840 card
>
> The driver is based on Donald Becker's Linux kernel
> driver, as usual :)
>
> The main problem was whether to use I/O for register access and
> how to work with device on second PCI bus. So the patch have
> +    unsigned short  pci_bus_no;
> added to pci_device structure to remember PCI bus number of device.
> On our i810-based motherboards the card is on PCI bus 1 (not 0).

Excellent work Igor! So you have it working? Send me the driver and any patches
and I will include it in the next release 4.6.6.

------------------------------------------------------
This mail sent via NLC WebMail: http://www.nlc.net.au/

[Etherboot-developers] Winbond W89c840F chip NICs - a new driver

[Igor V. K. <gar...@ma...>]

Hi there!

I have a bunch of Compex RL100ATX cards there
and I needed an Etherboot driver.

So now I have one and it should support

Compex RL100ATX card
Winbond W89c840 card

The driver is based on Donald Becker's Linux kernel
driver, as usual :)

The main problem was whether to use I/O for register access and
how to work with device on second PCI bus. So the patch have
+    unsigned short  pci_bus_no;
added to pci_device structure to remember PCI bus number of device.
On our i810-based motherboards the card is on PCI bus 1 (not 0).

--
Have a nice time(3)!

[Etherboot-developers] ATFTP

[Ken Y. <ke...@nl...>]

Here's an Open Source project that might be worth watching:

http://freshmeat.net/projects/atftp/

atftp is a client/server implementation of the TFTP protocol that
implements RFCs 1350, 2090, 2347, 2348, and 2349. The server is
multi-threaded and the client presents a friendly interface using
libreadline. The current server implementation lacks multicast and ipv6
support.

[Multithreaded operation is a prerequisite for multicast. Most current
tftp servers fork a separate process for each client.]

[Etherboot-developers] timer services

[<ke...@nl...>]

One deficiency of the current framework is the lack of a low-overhead timer
service for detecting stuck transmit routines. Currently counting loops are used
but this fails on fast CPUs and makes a mistake when the counter has run out and
transmitter is still busy. The timeofday clock is has too high overhead,
requiring a BIOS call. It seems the 8254 counter can do the job. I am envisaging
something like this:

	start transmit;
	load_timer(# of milliseconds);
	while (transmitter still busy and read_timer(t) > 0)
		;
	if (t == 0)
		printf("Transmitter stuck\n");

It appears that read_timer(t) can be as little as two inb instructions and a
shift.

Does anybody know if the 8254 counters are used for other purposes by the BIOS?
There is no OS to worry about at this point. Does anybody know any sample code
for using the 8254?

------------------------------------------------------
This mail sent via NLC WebMail: http://www.nlc.net.au/

[Etherboot-developers] booting PCI NICs

[<ke...@nl...>]

Steve Smith sent me some changes which are said to fix booting on some PCI
BIOSes that don't hook into INT19H. I attach them. Since they could affect
current users, before I put these changes into 4.6.6, I would like confirmation
that:

1. It doesn't break booting that currently works, and if possible,

2. If you were having problems with some PCI BIOSes that seemed to ignore
Etherboot before, that it now works.


------------------------------------------------------
This mail sent via NLC WebMail: http://www.nlc.net.au/

[Etherboot-developers] mknbi-1.0-4 released

[Ken Y. <ke...@nl...>]

I have released mknbi-1.0-4 at http://etherboot.sourceforge.net

Changes:

+ The last used sector of a floppy image was previously omitted from the
ramdisk and this could cause apps to fail.

+ FAT16 "floppies" should work now.

+ --harddisk now works for both mknbi-fdos and mknbi-dos.

For the moment please fetch mknbi-1.0-4 separately until etherboot-4.6.6
is released. It is available as .tar.gz, .rpm and .srpm.

[Etherboot-developers] Re: mknbi-1.0-4

[Ken Y. <ke...@nl...>]

I forgot to mention that the problems fixed in this release only concern
FreeDOS and DOS users. Linux and FreeBSD users can ignore this release
and relax.

BTW, there is a SSH client for DOS here: http://www.vein.hu/~nagyd/
I have booted this diskless onto a 286 with a 320kB "floppy". Felt like
a 2400 baud terminal but ok if you just want to read mail with pine. :-)
It could be a way of providing a roomful of secure terminals with old
386s.

Re: Etherboot 4.6.4 released

[Tom R. <tb...@ma...>]

On Mon, Jul 24, 2000 at 05:05:38PM -0700, Stuart Lynne wrote:
> On Tue, Jul 25, 2000 at 09:35:59AM +1000, Ken Yap wrote:
> > >I've just spent a few weeks poking and proding lilo and syslinux to make
> > >sure they work with the DOC BIOS extension firmware. 
> > >
> > >Moving somewhere else isn't sufficent. You have to avoid clobbering the
> > >BIOS extension when loading the kernel boot sectors and command line
> > >(usually done at 0x90000). 
> > 
> > Do you know of any documentation on what memory areas are used by DOC?
> 
> Nothing that I can give out without an NDA between you and M-Systems....
> 
> And really not much that will help (at least that I have).
> 
> As far as we can tell it simply grabs top 48kb of memory for the BIOS
> extension and then some variable sized amount of memory for tables (depends
> on the size of the DOC, also may appear to be inconsistant because DOC's
> of the same size may not actually be the same inside, for example 16MB
> DOC may have 1 or 2 flash chips, and this changes amount of RAM allocated).
> 
> We have found that for loading from DOC the only reliable way is:
> 
> 	1. load kernel boot sectors low
> 	2. construct command line low
> 	3. load rest of kernel high (>1mb)
> 	4. use boot protocol 2.02
> 
> I may have missed out something, Tom can provide some more details on
> what we did to lilo.... I've added him to the CC list.

What our patched LILO does now is:
(Note: addrs are segment addrs)

	1. load LILO 1st + 2nd stages 64K lower (0x8??? instead of 0x9???),
	2. load kernel boot sectors 64K lower (0x8000 instead of 0x9000),
	3. construct command line directly above kernel boot sectors
	   (approx 0x8100, depending on number of kernel setup segs),
	4. load rest of kernel high (as before),
	5. uses old boot proto if kernel not 2.02 aware (I don't
	   have a 2.02 aware kernel yet),
	6. launch from 0x8020 (64K lower than before)
	   (This only works for newer kernels, because the newer setup
	   segments make a few more BIOS calls, _then_ move themselves
	   to 0x9000.  If LILO builds the launch data in 0x9000, the
	   way the older kernels need it, one or more of the BIOS calls
	   in the setup segs will call to code that has already been
	   stepped on and the BIOS call will cack or hang.)

Something else to watch out for is that some of the new motherboards seem
to take a 1K chunk for BIOS stuff.  I have two that show 639K (or 607K if
the DOC is present), so you can't use right up to the end of 0x8000 either.
LILO seems to use 0x8000..0x8DA20 so it happens to work out.  YMMV.

-- 
Tom Rushworth

Re: Etherboot 4.6.4 released

[Stuart L. <sl...@li...>]

On Tue, Jul 25, 2000 at 09:35:59AM +1000, Ken Yap wrote:
> >I've just spent a few weeks poking and proding lilo and syslinux to make
> >sure they work with the DOC BIOS extension firmware. 
> >
> >Moving somewhere else isn't sufficent. You have to avoid clobbering the
> >BIOS extension when loading the kernel boot sectors and command line
> >(usually done at 0x90000). 
> 
> Do you know of any documentation on what memory areas are used by DOC?

Nothing that I can give out without an NDA between you and M-Systems....

And really not much that will help (at least that I have).

As far as we can tell it simply grabs top 48kb of memory for the BIOS
extension and then some variable sized amount of memory for tables (depends
on the size of the DOC, also may appear to be inconsistant because DOC's
of the same size may not actually be the same inside, for example 16MB
DOC may have 1 or 2 flash chips, and this changes amount of RAM allocated).

We have found that for loading from DOC the only reliable way is:

	1. load kernel boot sectors low
	2. construct command line low
	3. load rest of kernel high (>1mb)
	4. use boot protocol 2.02

I may have missed out something, Tom can provide some more details on
what we did to lilo.... I've added him to the CC list.

-- 
                                            __O 
Lineo                                     _-\<,_ 
PGP Fingerprint: 28 E2 A0 15 99 62 9A 00 (_)/ (_) 88 EC A3 EE 2D 1C 15 68
Stuart Lynne <sl...@li...>            www.lineo.com         604-461-7532