Hi,
I just tried to boot my token ring devices with 2.05 - to no avail,
it hangs after loading the kernel. I then tried older versions, and found
2.03 booting like a charm, and every 2.04-pre including 2.04 hanging.
Here's the config of the devices:
- Networkstation 2800 with IBM PCI Token Ring card 2
- IBM Thinkpad T21 with IBM PCI Token Ring card 2
Maybe this has to do with the changes regarding blksize. At least
the hang in 2.04-pre1 comes much faster than 2.03 boots the
kernel. I'll try to get some network dump of the boot, but that's
not really easy with token-ring ...
pxelinux.cfg:
prompt 1
default 28t
timeout 30
implicit 0
label 28t
kernel lfs28
append nfsroot=10.1.1.53:/usr/local/linuxnc/2.8-t/img devfs=mount
ip=dhcp
label 28t-check
kernel lfs28
append initrd=initrd28 root=/dev/ram0 init=/linuxrc devfs=mount ip=dhcp
[+/- wraparounds ...]
2.03 messages (all typed off the screen ...):
PXELINUX 2.03 2003-04-10 [..]
Found PXENV+ structure
Old PXE API detected, using PXENV+ structure
UNDI data segment at 0000
UNDI data segment size: 0000
UNDI code segment at: 0000
UNDI code segment size: 0000
PXENV entry point found (we hope) at 9981:0BE0
My IP address seems to be 0A01A8F8 10.1.168.248
ip=10.1.168.248:10.1.1.53:10.1.1.1:255.255.0.0
TFTP prefix: /linuxboot/
Trying to load: pxelinux.cfg/0A01A8F8
boot:
Loading lfs28................Ready.
Failed to free base memory, error 0015-001-99810B47
Uncompressing Linux... Ok, booting the kernel.
[ continues to boot ...]
2.04-pre1 messages:
PXELINUX 2.04 2.04-pre1 [..]
Found PXENV+ structure
PXE API version is 0200
Old PXE API detected, using PXENV+ structure
UNDI data segment at 0000
UNDI data segment size: 0000
UNDI code segment at: 0000
UNDI code segment size: 0000
PXENV entry point found (we hope) at 9981:0BE0
My IP address seems to be 0A01A8F8 10.1.168.248
ip=10.1.168.248:10.1.1.53:10.1.1.1:255.255.0.0
TFTP prefix: /linuxboot/
Trying to load: pxelinux.cfg/0A01A8F8
boot:
Loading lfs28....Ready.
Failed to free base memory, error 0015-001-99810B47
[hangs.]
Regards,
Josef
______________________________________________________________________________
Wo gibt es den besten Spam-Schutz? Laut ComputerBild 15-03 bei
WEB.DE FreeMail - Deutschlands beste E-Mail - http://s.web.de/?mc=021122
Hi, "Josef Siemes" <jsiemes at web.de> schrieb am 29.07.03 20:58:15:> I just tried to boot my token ring devices with 2.05 - to no avail, > it hangs after loading the kernel. I then tried older versions, and found > 2.03 booting like a charm, and every 2.04-pre including 2.04 hanging.Strange. That was booting from AIX, with AIX tftpd. Just tried linux (RH Advanced Server), and it works with 2.04. It's a pretty old version of tftp-hpa, though, and it seems to me that the client prints some more dots downloading the file. Maybe different blksize? Regards, Josef ______________________________________________________________________________ 100 Prozent Gewinnausschuettung keine versteckten Kosten. Gemeinsam garantiert gewinnen: WEB.DE Spielgemeinschaften! https://spielgemeinschaften.web.de/
Hi, "Josef Siemes" <jsiemes at web.de> schrieb am 30.07.03 18:38:29:> "Josef Siemes" <jsiemes at web.de> schrieb am 29.07.03 20:58:15: > > I just tried to boot my token ring devices with 2.05 - to no avail, > > it hangs after loading the kernel. I then tried older versions, and found > > 2.03 booting like a charm, and every 2.04-pre including 2.04 hanging. > > Strange. That was booting from AIX, with AIX tftpd. Just tried linux > (RH Advanced Server), and it works with 2.04. It's a pretty old version of > tftp-hpa, though, and it seems to me that the client prints some more > dots downloading the file. Maybe different blksize?Just found it - at least it seems like that. The AIX machine has a maximum IP blocksize of 1492 bytes configured, so every packet got fragmented at 1492 bytes, and a second packet containing the last 8 tftp bytes followed. pxelinux does various strange things then, for example I got 'Kernel not found' for kernels that were there, the keyboard mapping seemed to go completely crazy etc.pp.. And yes, I know that pxelinux doesn't like fragmented packets at all, since the braindead PXE roms don't support this ... I just tried to change MTU to 1500, but it seems that AIX doesn't like that without taking the device down, since the tftp server doesn't seem to serve the client any more. BTW: This is on an ATM interface, with Token-ring lan emulation. So is there any chance to get blksize configurable, and stick to 512 as long as no config file was loaded? Maybe someone else will have problems like these because the MTU is configured somewhat strange. Regards, Josef ____________________________________________________________________________ ImmobilienScout24 - Uber 170.000 Immobilien - Finden Sie hier Ihre neue Traumwohnung! In Immobilien die Nr.1 im Netz: http://info.is24.de/web.de
Hi again, "Josef Siemes" <jsiemes at web.de> schrieb am 11.08.03 21:54:36:> I just tried to change MTU to 1500, but it seems that AIX doesn't > like that without taking the device down, since the tftp server doesn't seem > to serve the client any more.I've tried a different machine with gigabit ethernet and MTU 1500, this works. This is a separate network, in the other one the MTU is set to 1492 because of a bridge that doesn't like an MTU of 1500. Regards, Josef ______________________________________________________________________________ Wochentlich Gewinnen garantiert! Chance auf Jackpot gewahrt! WEB.DE Spiel- gemeinschaften: Jetzt Anteile sichern! https://spielgemeinschaften.web.de/
Hi, "H. Peter Anvin" <hpa at zytor.com> schrieb am 15.08.03 20:44:21:> Josef Siemes wrote: > > "Josef Siemes" <jsiemes at web.de> schrieb am 11.08.03 21:54:36: > >>I just tried to change MTU to 1500, but it seems that AIX doesn't > >>like that without taking the device down, since the tftp server doesn't seem > >>to serve the client any more. > > > > I've tried a different machine with gigabit ethernet and MTU 1500, this works. > > This is a separate network, in the other one the MTU is set to 1492 because > > of a bridge that doesn't like an MTU of 1500.> OK, that's almost certainly the problem, then. What's happening is that > with the larger blksize you're getting fragmentation, and then your PXE > stack doesn't handle fragment reassembly correctly.Jep, that's what I wrote in an earlier posting. The packet got fragmented at 1492 bytes, and the last 8 bytes were sent in another packet.> Perhaps blksize should be a configure file option. I need to figure out > how to best do that.That's what I already suggested.> In the meantime you can disable blksize on the TFTP server.In the current implementation I can't, since I'm not using tftp-hpa, but the one from AIX. This server can't switch off options. But since 2.00 is still working without problems there's no hurry to replace it with 2.05. Regards, Josef __________________________________________________________________________ Die sicherste Form der Kommunikation: E-Mails verschluesseln, Spam-Filter, Adressverifizierung, digitale Unterschrift: http://freemail.web.de