Threaded Mode | Linear Mode

Torgeir · (This post was last modified: 2012-10-14 13:41 by Torgeir.)

Hi!

I hit the wall "again". Now with a HP z400 Workstation. No network.
With an extra network card (broadcom netXtreme) net on that card is working as it should.

Both (ipxe.pxe, ipxe.lkrn) are based on the same source. iPXE 1.0.0+ (09cc). This time I have not tried with any of the undionly (k/kk/kkk), because ipxe.pxe was the one working for most of our computers.

There is no newer BIOS upgrade available for this model.

Any thoughts/ideas?

edit 2: Was a problem with gigabit link. 100Mbit was working OK though.
A patch brom a Broadcom developer fixed the problem.

edit: all of the undionly.kpxe/kkpxe/kkkpxe is working OK. But with undionly I had troubles with some other stuff...

Torgeir

Is it possible to solve this? Can I have some form of fallback from ipxe.pxe to undinoly.kpxe if no network. With some smart ipxe scripting?

This config is pxelinux-4.05, chainloading ipxe.pxe. I have not tried with "dhcp option user-class iPXE" on a closed LAN (will do tomorrow).

Torgeir

**thomil** · 2012-10-01, 15:44

Which NIC does the machine have? The pci IDs would be good to know( lspci -nn ).
Also, is there any specific reason why you're not using undionly.kpxe? It should work on any machine, even if there's no native iPXE driver available for your card.

Torgeir · 2012-10-01, 16:30

The HP 2570p did not get any network with undionly.kpxe (but worked fine with ipxe.pxe).
We are chainloading ipxe from pxelinux 4.05. Will try out tomorrow on an isolated LAN (without chainloading from pxelinux).

Torgeir

Torgeir · (This post was last modified: 2012-10-03 12:25 by Torgeir.)

I belive my problem is like this: http://forum.ipxe.org/showthread.php?tid=1705 (client identifier).
PXE-iPXE.

I cannot test on our production environment (so will have to setup a test Windows Server 2008 R2 with DHCP Server).

On my linux test environment I do not have the same problem as with our Windows DHCP Servers...

Edit: Oh yes I did. Forgot to test with the HP z400... My linux DHCP is working identical as Windows Server 2008 DHCP server..

Torgeir

***robinsmidsrod*** · 2012-10-02, 09:51

(2012-10-01 15:44)meteger Wrote: Also, is there any specific reason why you're not using undionly.kpxe? It should work on any machine, even if there's no native iPXE driver available for your card.

I seem to recall that he had interaction problems with his vendor PXE and undionly.kpxe on one of his machines, which is why I suggested using native drivers.

Torgeir: You could try with ipxe.kpxe, though that is quite a strange target to use. Since ipxe.* contains the undi driver, and any .kpxe target should retain the vendor PXE stack technically it should try UNDI if no native driver is available. But if a native driver is available (but misbehaving) it won't work for you. It's worth a try, though.

Torgeir · (This post was last modified: 2012-10-04 09:07 by Torgeir.)

Code:

make bin/ipxe.pxe DEBUG=dhcp

When iPXE was loading I got this text in red (debug output)

Code:

DHCP 0xdbd84 entering discovery state

DHCP 0xdbd84 DHCP Discover

DHCP 0xdbd84 DHCP Discover

DHCP 0xdbd84 DHCP Discover

DHCP 0xdbd84 DHCP Discover

... timed out.

ifstat command got this output:

Code:

Link up net0

TX:4 TXE:0 RX:58 RXE:48

RXE: 29 x Operation not supported http://ipxe.org/3c086003

RXE: 14 x Socket not connected http://ipxe.org/380f6001

Tried to set an static IP:

Code:

ifopen net0

set net0/ip xxx.xxx.x.xxx

set net0/netmask xxx.xxx.xxx.x

I could not ping this ip from another computer, on the same network.
After some minutes I checked "ifstat" command again an the packets have increased.

Code:

TX:14 TXE:5 RX:2485 RXE:1669

RXE: Operation not supported http://ipxe.org/3c086003

RXE: Socket not connected http://ipxe.org/380f6001

From the error messages. I made one with "netdevice" as debug. Lot of repeating debug output:

Code:

NETDEV net0 unknown network protocol 86dd

NETDEV net0 failed to receive 0x0

... operation not supported

... socket not connected

-||- over and over again..

until:

Connection timed out

Could not configure net0

But the same machine on my lab network. Got Link up ok. DHCP net0 ok.

Can it be something with our company DHCP configuration iPXE does not like?
How can I debug this further?

Torgeir

***robinsmidsrod*** · 2012-10-04, 09:53

Try a packet trace (http://ipxe.org/howto/pcap) and see if you get unexpected responses to your DHCP requests. Make sure you trace on a place were you can actually see the packets, though. Smile

I also notice you have a lot of RXE and some TXE, which could indicate some kind of network configuration error or misbehaving driver. Try to debug the undi driver (what you type in after DEBUG= is the name of some file from the source tree without the .c extension, separate multiple with just a single comma, no space, like this: DEBUG=scsi,iscsi).

Torgeir · (This post was last modified: 2019-09-07 23:47 by Torgeir.)

Only way I could make an unsencored packet trace was on my lab-network, with the undionly.kpxe and the hp 2570p (since that did not work).

Code:

168.2.250 Windows DHCP server

168.2.102 Linux tftp-server

168.2.52 hp 2570p

http://pastebin.com/bqzs3Etj

Torgeir

Torgeir · 2012-10-09, 14:17

Update!

From the ipxe-devel mailing list (Brandon Penglase):

Code:

I am testing with BCM5764M and while iPXE will function properly on

100Mbit, it fails on DHCP still on a gigabit link. This behaviour is

the same as before this commit.

Thanks Kevin and all others who are working on this. I can provide

any data that may help with resolving this issue.

Brendon

I found out the same problem exist with my HP z400. When I was testing my lab-network I was using an old 100Mbit switch. Thats why it was working on my lab-network and not on our company network Smile

I put the same old 100Mbit between our company network and our z400, got straight through (link up).

This was with the ipxe.pxe

Kevin Tran replyed back with this:

Code:

Hi Brandon,

We were able to reproduce the issue you reported on BCM5764m.  We already resolved this.   I will send you the fixes on other e-mail.  Once you confirm that your issue is fixed, we will submit the changes.

Thanks!

Kevin

So I am very eager to try this out with our z400.

Torgeir

Torgeir · 2012-10-11, 07:06

The tg3.h.patch and tg3_hw.c.patch is now working great with my z400 (BCM5764m).

So my problem is now solved Smile

Torgeir

***robinsmidsrod*** · 2012-10-11, 09:14

I'm glad you solved your problem.

Do stick around and help other people solve their problems if you can. I personally would very much like it.

Sedorox · 2012-10-11, 18:18

From what I heard from the guy at Broadcom, he will be giving the changes to iPXE here shortly for inclusion into the main tree, so anyone else having this issue (Comes down to the one broadcom chipset on a Gigabit link), hold out a little bit longer and a fix will be included. I'll make sure to update this post when it happens.

- Brandon P.

Nicola V · 2013-02-19, 16:56

Hi!
Is this patch in the main git tree already? We're having the same issue with a HP z400 and ipxe compiled approximately 4 weeks ago.
The dhcp timeout issue disappear when placing a 100 Mbit switch in the middle to force the speed link.
Thanks!

Torgeir · (This post was last modified: 2018-06-30 09:44 by Torgeir.)

Hi!

Dont think so... Anyway here are they:

tg3_hw.c.patch
https://www.mail-archive.com/ipxe-devel@...hw.c.patch

tg3.h.patch
https://www.mail-archive.com/ipxe-devel@.../tg3.patch

Edit: updated links to original patch from mailinglist.
Original post: https://www.mail-archive.com/ipxe-devel@...01891.html

Torgeir

Nicola V · 2013-02-20, 13:57

(2013-02-20 13:48)Torgeir Wrote: Hi!

Dont think so... Anyway here are they:

tg3_hw.c.patch
http://pastebit.com/pastie/12050

tg3.h.patch
http://pastebit.com/pastie/12051

Torgeir

Thanks I'll try these out then! We're having a similar behaviour on some HP blades with broadcom chips. I'll keep you updated in case this fixes the issue on these machines as well.

Cheers

Nicola V · 2013-02-21, 09:39

Patches working fine on our z400 as well.
After that, we hit the slow download of the images via WDS, but I guess that's unrelated. I'll test the fix on a HP ProLiant BL465c G6 which exhibits a similar behaviour.

harrison · 2016-09-02, 09:33

Hi Torgeir,
how can i download the patches? If i try to follow the link
it ask me for user/password ?
My User / Password from IPXE doesn't work!

Thanks a lot!
Harrison

(2013-02-20 13:57)Nicola V Wrote:
(2013-02-20 13:48)Torgeir Wrote: Hi!

Dont think so... Anyway here are they:

tg3_hw.c.patch
http://pastebit.com/pastie/12050

tg3.h.patch
http://pastebit.com/pastie/12051

Torgeir

Thanks I'll try these out then! We're having a similar behaviour on some HP blades with broadcom chips. I'll keep you updated in case this fixes the issue on these machines as well.

Cheers

Thony · 2018-06-27, 07:50

Hi Torgeir,
how can i download the patches?
When i try to follow the link, i'm asked for user/password ?

Best regards!
Thony

(2013-02-20 13:57)Nicola V Wrote:
(2013-02-20 13:48)Torgeir Wrote: Hi!

Dont think so... Anyway here are they:

tg3_hw.c.patch
http://pastebit.com/pastie/12050

tg3.h.patch
http://pastebit.com/pastie/12051

Torgeir

Thanks I'll try these out then! We're having a similar behaviour on some HP blades with broadcom chips. I'll keep you updated in case this fixes the issue on these machines as well.

Cheers

Torgeir · 2018-06-30, 09:47

Links updated:

https://www.mail-archive.com/ipxe-devel@...hw.c.patch
https://www.mail-archive.com/ipxe-devel@.../tg3.patch

I believe these has been committed...

(2018-06-27 07:50)Thony Wrote: Hi Torgeir,
how can i download the patches?
When i try to follow the link, i'm asked for user/password ?

Best regards!
Thony

Thanks I'll try these out then! We're having a similar behaviour on some HP blades with broadcom chips. I'll keep you updated in case this fixes the issue on these machines as well.

Cheers

[/quote]