After configuring DHCP and PXE server on one system. The client boots up, got IP from dhcp server and start remote installation of Linux. It found the boot server, then it prompt "Press any key to enter kernel parameters.." If you let it default, it would down loaded the kernel image and initrd image. "Downloading Linux kernel image... Downloading initrd image... Uncompressing Linux... Ok, booting the kernel." It hangs when it try to boot the kernel. If you do enter the kernel parameter by typing in "ks" It would boot the kernel and locate the kickstart file to complete the headless installation.
We grabbed from rawhide and that does not fix the problem. This is preventing us from getting a working version of RHL to our WW sales force and customers. We need a binary conatining this fix, verified by RH, posted on a RH support site with instructions on how to apply the fix to RHL 6.2. Problem Details below: -----Original Message----- From: Erik Troan [ewt] Sent: Thursday, June 15, 2000 4:07 PM To: 'kickstart-list' Subject: RE: kickstart - bootnet.img kernel hang Sounds like a bug in the loader -- I suggest grabbing the pxelinux boot loader from rawhide and trying it (the Intel PXE kernel loader has known problems). Erik On Thu, 15 Jun 2000, Krig, Steve B wrote: > "Downloading Linux kernel image... > Downloading initrd image... > Uncompressing Linux... Ok, booting the kernel." > > Hangs here. No more console output. > > -----Original Message----- > From: Erik Troan [ewt] > Sent: Thursday, June 15, 2000 12:28 PM > To: 'kickstart-list' > Subject: Re: kickstart - bootnet.img kernel hang > > > > What's hang mean? If the kernel starting up at all? > > Erik > > On Thu, 15 Jun 2000, Krig, Steve B wrote: > > > > > with RHL 6.2, we are not able to use kickstart for unattended installs as > > the kernel is hanging. > > > > We are seeing the bootnet kernel hang after it is loaded/uncompressed via > > PXE. > > > > > mount -o loop -t vfat /redhat-cdrom/images/bootnet.img /mnt/floppy > > > cp /mnt/floppy/vmlinuz /tftpboot/X86PC/UNDI/linux-install/linux.1 > > > cp /mnt/floppy/initrd.img /tftpboot/X86PC/UNDI/linux- install/linux.2 > > > > Can we get a kernel image fix? > > > > thanks > > -steve > > > >
did you try entering "linux console=ttyS0" as boot parameters?
If I understand it correctly, Redhat suggeted trying to: send the client machine pxelinux.bin, not the kernel image itself." My assumption is that they're suggesting to replace /tftpboot/X86PC/UNDI/linux- install/linux.1 with PXELINUX.BIN from the SYSLINUX package. Tried this. With the Chardonnay, it attemps to grab PXELINUX.BIN and execute it but the system hangs. Thinking that PXELINUX.BIN might be the bootloader replacement, I also tried replacing the file /tftpboot/X86PC/UNDI/linux-install/linux.0 with PXELINUX.BIN and the Chardonnay hangs Using a "NetPC" client which contains the "LANDesk Service Agent v0.99j" boot agent, I'm able to connect to the PXE server. By replacing the bstrap.0 file with pxelinxu.bin the client was able grab PXELINUX.BIN and execute it. So it appears that PXELINUX.BIN does not work with the Chardonney clients (which PXE-2.0) Tried a client with addin PRO100+ NIC (contains PXE-1.0) with same unsucessful result. Also I tried to send the pxelinux.bin image over to the client through a TFTP server. This method works with the netpc client but don't work with Chardonnay.
There are major problems with the bootloader that comes in the pxe package. When using initrd in several memory configurations, data is overwritten before the kernel boots. This causes lockups and unusual behavior after booting the kernel. We use the pxelinux bootloader for network booting here. You need the tftp server and pxe server from rawhide in order to make it work. I can provide packages built for a 6.2 system if needed. pxelinux is not a drop-in replacement for the loader that comes with the Intel pxe server. Read the pxelinux documentation for details.
*** Bug 12530 has been marked as a duplicate of this bug. ***
The test box we have here plugged into our test OS environment and pxe booted into the installer on the first try.
A few things that are hard to relay on the phone: Using the 6.2 pxe server, pxelinux will not get the "myip" part of the dhcp reply packet. This means that pxelinux will say, "My IP address seems to be 0.0.0.0". It will then try to load its configuration file via tftp by loading: pxelinux.cfg/00000000 pxelinux.cfg/0000000 pxelinux.cfg/000000 pxelinux.cfg/00000 pxelinux.cfg/0000 pxelinux.cfg/00 pxelinux.cfg/00 pxelinux.cfg/0 pxelinux.cfg/default This is fine if you want to use the same pxelinux configuration for ever machine, you just make one configuration file named 00000000. Only if you want per-IP configuration files do you need a modified pxe server. Be sure to configure the pxe server properly: [Service_Types] 10,pxelinux 0,BStrap # Menu string that will be displayed on the client screen # after F8 is pressed. [X86PC/UNDI/MENU] 10,Linux Install 0,Local Boot [X86PC/UNDI/pxelinux/ImageFile_Name] 0 2 pxelinux # Image file name for BStrap boot server # format : <min layer #> <max layer #> <base file name> [X86PC/UNDI/BStrap/ImageFile_Name] 0 0 bstrap --- In /tftpboot/X86PC/UNDI/pxelinux/ you should have: pxelinux.0 <-- pxelinux binary file pxelinux.cfg/ <-- directory for config files pxelinux.cfg/00000000 <-- default configuration file In our configuration, we also have the following in our pxelinux/ directory: msw/vmlinuz msw/initrd.img These are the images that I use for my install testing. In our configuration there is one directory per developer. Our default pxelinux configuration (pxelinux.cfg/default) looks like this: default linux prompt 1 timeout 50 label linux kernel msw/vmlinuz append initrd=msw/initrd.img ks Now, the last piece is the in.tftpd server. I've placed a RPM of the server we use on our pxe server at: http://people.redhat.com/msw/updates/i386/tftp-server-0.16-5.i386.rpm You must modify the /etc/inetd.conf to read: tftp dgram udp wait root /usr/sbin/tcpd in.tftpd -s /tftpboot (default 6.2 reads:) tftp dgram udp wait root /usr/sbin/tcpd in.tftpd This gives you a side effect of having a secure chroot() tftp serving space, good for the security conscious administrator. I double checked our mtftp configuration and it seems that we don't use mtftp to serve the pxelinux.0. Sorry for sending you down the wrong path on that one.
With a few modification to the configuration notes above, we were able to implement pxelinux with Redhat 6.2 SBE2 to do a complete headless installation. Configuration notes: The stock 6.2 tftpd does not work with pxelinux. We need to down load the tftp-hpa-0.13.tar.gz from http://www.kernel.org/pub/software/network/tftp/ Set the prompt and timeout to 0 in the default pxelinux configuration file. Our default configuration file look like this: default linux prompt 0 timeout 0 label linux kernel vmlinuz append initrd=msw/initrd.img ks
With additional modifications to the pxe.conf and /etc/services. We were able to use pxe to do a full headless and handfree installation of Redhat 6.2 SBE 2. We have completed headless installation using PXE and PXElinuxl.