Thanks Edgar, changing the BIOS setting helped (had to set it to "enhanced" for some reason, also remember to set it back after pxe install).
Now I have one more (hopefully) query. Currently my client install script gets all the way to the systemconfigurator section before it belly-ups. The command line looks like this: "chroot /a/ systemconfigurator --excludesto=/etc/systemimager/systemconfig.local.exclude --configsi -stdin << EOL || shellout" Now after rooting around a bit, I found that "systemconfig.local.exclude " doesnt exist, it has apparently been superceded by "updateclient.local.exclude". Now I havent tried changing this yet, so I dont know if its an issue. Secondly, I know --configsi is supposed to configure the hardware, network and boot sector from the block data below it. It doesnt however (there is only network data in the block in any case). My client's /boot/grub (I use grub loader) directory is empty except for the splash image. (I havent checked the network config files, but I imagine they wont be perfect either.) If someone could send me a working image install script(if it is relevant), or boot loader configuration file so I can figure out what should be happening and/or tell me why the client image build does not create the boot configuration properly, I would be much obliged. (At what point is the boot loader config file created? I dont know if this has any bearing, but during client image creation just after pciutils, there is a repeated "awk: fatal: /etc/fstab file not readable or no such file or directory". Of course it is there and it is readable by all. Very strange.) I imagine it has something to do with the lack of support for the SATA drive in the systemconfigurator's hardware detection routine, or the fact that the partitions on the SATA drive have to be called sda during pxe booting because of the way the libata patch to kernel 2.4.24 handles SATAs as scsi. The question I suppose is, do you need a succesful hardware detection for the harddisk to do a proper boot configuration on the client? So close and yet so far. Eugene ----- Original Message ----- From: "Edgardo Evangelista" <[EMAIL PROTECTED]> To: "eugene" <[EMAIL PROTECTED]> Sent: Friday, March 26, 2004 7:01 PM Subject: RE: [Oscar-users] Yet another DHCP query Eugene, Your issue on the SATA is related to the BIOS. Goto the CMOS settings and change your IDE settings to "S-ATA only" instead of "P-ATA and S-ATA". Edgar E. -----Original Message----- From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED] Behalf Of eugene Sent: Friday, March 26, 2004 9:10 AM To: Frank Crawford; Bernard Li Cc: [EMAIL PROTECTED] Subject: Re: [Oscar-users] Yet another DHCP query Frank, The hardware simply fails to initialise if I try to use the tg3 driver in stead of the BCM5700. Its ok though, I built a 2.4.24 kernel with the BCM5700 patch applied. I used the config file from the systemimager directory and just added the BCM5700 driver into the kernel. Now however I am stuck with this SATA drive. Has anyone ever managed to PXE boot one of these things? I built my libata patched kernel with all the modules included. Here are some highlights: (the section just before the tcp/ip get initialised) " Unifrom Multi-Platform E-IDE driver revision: 7.00beta4-2.4 ide: Assuming 33MHz system bus speed for PI modes; override with idebus=xx hda: Maxtor 6Y200M0, ATA Disk drive ide0 at 0x1f0-0x1f7, 0x3f6 on irq 14 Partition check: hda:end request: I/O error, dev 03:00 (hda), sector 0 end_request: I/O error, dev 03:00 (hda), sector 2 end_request: I/O error, dev 03:00 (hda), sector 4 end_request: I/O error, dev 03:00 (hda), sector 6 end_request: I/O error, dev 03:00 (hda), sector 0 end_request: I/O error, dev 03:00 (hda), sector 2 end_request: I/O error, dev 03:00 (hda), sector 4 end_request: I/O error, dev 03:00 (hda), sector 6 unable to read partition table SCSI subsystem driver revision: 1.00 ata: 0x1f0 IDE port busy ata1: SATA max UDMA/133 cmd 0x170 ctl 0x376 bmda 0xF008 irq 15 ATA: abnormal status 0x7F on port 0x177 scsi0 : ata_p11x" (then a lot of stuff about registering raid personalities) It kinda looks like the sata driver is taking hold, but not properly. When it autodetects the hardware: "Detecting hardware: Loading sd_mod... Assuming sd_mod is compiled into kernel, not needed, or allready loaded Loading ide-disk... Assuming ide_disk is compiled into kernel, not needed, or allready loaded " Why is it not attempting to load scsi drivers under which SATA is kept? Can I override hardware autodetection somehow? When it tries to partition the drive it fails with: "Error could not stat device /dev/hda - no such file or directory." I have tried using "sda" in the partition config, but makes no difference. "Error: could not stat device /dev/sda" I can honestly say I am stumped. Sorry for the long mail. Eugene PS. I have tried using the 2.6.0 kernel which has native SATA support too, bit this does a kernel panic each time it has to mount a initrd.img PPS. Should I try to patch a 2.4.20 kernel with a patch meant for 2.4.24? If so I can patch the distro RH9.0 kernel and try to use that. Frank Crawford wrote: > Eugene, > What features of the BCM5700 driver do you need that tg3 doesn't > supply? You are not going to find a bcm5700 driver in any prebuild PXE > kernels, since it isn't a part of the standard Linux kernel. Adding it > yourself shouldn't be too hard, although most instructions for it cover > how to make it as a module, not compile it into the kernel. > > To make your own kernel, etc, look through the archives from around > Sept/Oct last year for details of how to create such a kernel, but you > need to be very aware of all the devices you need to support in your > kernel. > > Frank > > On Fri, 2004-03-26 at 03:30, eugene wrote: > >>Hello again, >> >>I have figured out where the problem is and what is going wrong. >>The PXE kernel does indeed lack support for the BCM5700 range of NICs >>Unfortunately the kernel you sent me does not work either (It needs the >>BCM5700 driver, the tg3 driver doesnt seem to be sufficient). I found >>another kernel from Jeremy Enos by searching the list that does support >>my NIC, unfortunately something else goes wrong when his ramdisk image >>is mounted and I end up with a kernel panic similar to the one reported >>here: >>http://www.mail-archive.com/[EMAIL PROTECTED]/msg03077.htm l >> >>I'm thinking I am going to have to compile my own kernel with BCM5700 >>support in, or at least do something about the initrd.img. >>Any advice? >> >>Thanks >> >>Eugene >> >>Bernard Li wrote: >> >>>Hey Eugene: >>> >>>Can you look carefully at the error message which says "No DHCPOFFERS >>>received"? Does it say 'lo' or 'ethX'? My feeling is that the network >>>driver wasn't even loaded for your client nodes. >>> >>>The kernel image which SIS use to boot up your client nodes for imaging >>>does not contain the appropriate network drivers - I have an image which >>>Frank compiled a while back which works for the bcm5700 series and if >>>you need it, just leave me a message. >>> >>>Cheers, >>> >>>Bernard >>> >>> >>> >>>>-----Original Message----- >>>>From: [EMAIL PROTECTED] >>>>[mailto:[EMAIL PROTECTED] On Behalf Of eugene >>>>Sent: Wednesday, March 24, 2004 9:29 >>>>To: [EMAIL PROTECTED] >>>>Subject: [Oscar-users] Yet another DHCP query >>>> >>>>Hi >>>> >>>>I think I have a vague idea of the answer to this one: >>>>I have manged to get all the way to network booting my client node. >>>>After loading the kernel and drivers, during the install >>>>process the client fails at "start network" with the >>>>following message: >>>> >>>>"No DHCPOFFERS received" >>>> >>>>What I dont understand is how it can load the kernel and the >>>>autoinstall across the network, but fails to configure the >>>>network card for the latter part of the install. It seems >>>>likely that the driver initialization of the onboard broadCom >>>>BCM5700 is failing despite the fact that the driver modules >>>>were included with the chosen kernel. >>>> >>>>Currently the following network drivers are being loaded, why >>>>I dont know: >>>>Intel Pro/1000 >>>>pcnet32.c >>>>ns83820.c >>>>sk98lin (the dlink module I had to install to get the second >>>>NIC on the master node working) >>>> >>>>Might it have anything to do with the fact that the BCM5700 >>>>is installed as eth1 on the master node, while it is eth0 on >>>>the client? Seems worth trying to swop them. >>>> >>>>Please tell me I dont have to recompile the kernel with the >>>>BCM5700 driver. Also for some mysterious reason, the tg3 >>>>driver does not work on the master node and I have to use the >>>>bcm5700 for the broadcom NIC. >>>> >>>>Thanks >>>> >>>> >>>> >>>>------------------------------------------------------- >>>>This SF.Net email is sponsored by: IBM Linux Tutorials Free >>>>Linux tutorial presented by Daniel Robbins, President and CEO >>>>of GenToo technologies. Learn everything from fundamentals to >>>>system >>>>administration.http://ads.osdn.com/?ad_id=1470&alloc_id=3638&op=click >>>>_______________________________________________ >>>>Oscar-users mailing list >>>>[EMAIL PROTECTED] >>>>https://lists.sourceforge.net/lists/listinfo/oscar-users >>>> ------------------------------------------------------- This SF.Net email is sponsored by: IBM Linux Tutorials Free Linux tutorial presented by Daniel Robbins, President and CEO of GenToo technologies. Learn everything from fundamentals to system administration.http://ads.osdn.com/?ad_id=1470&alloc_id=3638&op=click _______________________________________________ Oscar-users mailing list [EMAIL PROTECTED] https://lists.sourceforge.net/lists/listinfo/oscar-users
