Hi,
I'm trying to get Fedora running on an MVME 2700 single board computer. This little board has a 750 PowerPC processor. I've got Debian 3.1 running happily atop the 2.6.11.7 kernel. It runs like a champ, but I'd like to try running Fedora instead of Debian. This single board computer (we've actually got ~35 of them) has no CDROM drive, and I had to do quite a bit of horsing around with the kernel configuration to get it to boot anyway (via NFS), so there's no way for me to just boot off of the iso image CDROMs and install Fedora normally. Is there any way I can install from the CDROM images (which, of course, I can mount with the "-o loop" option) without booting from the CDROMs? Is there a script somewhere that will install a minimal set of RPMs somewhere in my files system, so that I can later mount that area as root? Is there any way to do a network install without first booting off a CDROM?
Thanks for any info!
Ken Young Smithsonian Astrophysical Observatory 60 Garden Street, MS 78 Phoney Dolson has a secret name. Cambridge, MA 02138 USA (617) 495-7330 kyoung@cfa.harvard.edu
On Mon, 2005-04-25 at 19:38, Ken Young wrote:
Hi,
I'm trying to get Fedora running on an MVME 2700 single board computer. This little board has a 750 PowerPC processor. I've got Debian 3.1 running happily atop the 2.6.11.7 kernel. It runs like a champ, but I'd like to try running Fedora instead of Debian. This single board computer (we've actually got ~35 of them) has no CDROM drive, and I had to do quite a bit of horsing around with the kernel configuration to get it to boot anyway (via NFS), so there's no way for me to just boot off of the iso image CDROMs and install Fedora normally. Is there any way I can install from the CDROM images (which, of course, I can mount with the "-o loop" option) without booting from the CDROMs? Is there a script somewhere that will install a minimal set of RPMs somewhere in my files system, so that I can later mount that area as root? Is there any way to do a network install without first booting off a CDROM?
While not familiar with debian, I have done countless no media installs with various hardware configs. You should be able to boot the install kernel and initrd using your bootloader and walk through the install. Once the install is done Fedora leaves a kickstart config with the info needed to replicate the install.
A glance at the specs does not show a network bootable ethernet but you can boot the install from the install kernel.
Since you have linux running already, with FC4t2 you can copy vmlinuz and initrd.img from the install cd /images/pxeboot to a bootable location and setup grub or whatever to boot them. These are the same files as used in the boot.iso but saves the work getting to them.
I typically place all the files on my installation server and do http installs but you can of course do nfs installs with just the isos in an exportable dir.
HTH
Bret
On Mon, 25 Apr 2005, Bret Hughes wrote:
On Mon, 2005-04-25 at 19:38, Ken Young wrote:
Hi,
I'm trying to get Fedora running on an MVME 2700 single board computer. This little board has a 750 PowerPC processor. I've got Debian 3.1 running happily atop the 2.6.11.7 kernel. It runs like a champ, but I'd like to try running Fedora instead of Debian. This single board computer (we've actually got ~35 of them) has no CDROM drive, and I had to do quite a bit of horsing around with the kernel configuration to get it to boot anyway (via NFS), so there's no way for me to just boot off of the iso image CDROMs and install Fedora normally. Is there any way I can install from the CDROM images (which, of course, I can mount with the "-o loop" option) without booting from the CDROMs? Is there a script somewhere that will install a minimal set of RPMs somewhere in my files system, so that I can later mount that area as root? Is there any way to do a network install without first booting off a CDROM?
While not familiar with debian, I have done countless no media installs with various hardware configs. You should be able to boot the install kernel and initrd using your bootloader and walk through the install.
Thanks very much for your quick reply. Booting the install kernel is precisely what I'm trying to avoid. I've got a working kernel, and I had to tweek quite a few parameters in the .config in the kernel tree to build something that would work in this weird little computer (I initially cross-compiled on an x86 workstation, but now I can compile the kernel natively under Debian). I don't think I can boot the install kernel, so what I'm trying to do is see if there's some way I can jump into the installation process at a later step. For example, can I manually invoke the initrd from the CDROM after booting my kernel? Or can I issue commands that would be equivalent to what initrd does? Or is there just some list of RPMs I could load into some area that would constitute a minimal distribution of Fedora which I could subsequently mount as root?
Thanks again for your help.
Ken
On Mon, 2005-04-25 at 22:01, Ken Young wrote:
On Mon, 25 Apr 2005, Bret Hughes wrote:
On Mon, 2005-04-25 at 19:38, Ken Young wrote:
Hi,
I'm trying to get Fedora running on an MVME 2700 single board computer. This little board has a 750 PowerPC processor. I've got Debian 3.1 running happily atop the 2.6.11.7 kernel. It runs like a champ, but I'd like to try running Fedora instead of Debian. This single board computer (we've actually got ~35 of them) has no CDROM drive, and I had to do quite a bit of horsing around with the kernel configuration to get it to boot anyway (via NFS), so there's no way for me to just boot off of the iso image CDROMs and install Fedora normally. Is there any way I can install from the CDROM images (which, of course, I can mount with the "-o loop" option) without booting from the CDROMs? Is there a script somewhere that will install a minimal set of RPMs somewhere in my files system, so that I can later mount that area as root? Is there any way to do a network install without first booting off a CDROM?
While not familiar with debian, I have done countless no media installs with various hardware configs. You should be able to boot the install kernel and initrd using your bootloader and walk through the install.
Thanks very much for your quick reply. Booting the install kernel is precisely what I'm trying to avoid. I've got a working kernel, and I had to tweek quite a few parameters in the .config in the kernel tree to build something that would work in this weird little computer (I initially cross-compiled on an x86 workstation, but now I can compile the kernel natively under Debian). I don't think I can boot the install kernel, so what I'm trying to do is see if there's some way I can jump into the installation process at a later step. For example, can I manually invoke the initrd from the CDROM after booting my kernel? Or can I issue commands that would be equivalent to what initrd does? Or is there just some list of RPMs I could load into some area that would constitute a minimal distribution of Fedora which I could subsequently mount as root?
hmm. I really do not know what is in the install kernel that enables the launch of the installation. I suppose, as you indicate that it is the initscripts in the initrd. I would try the redhat anaconda list. Anaconda being the installation program for redhat. The guys on that list and the kickstart list are extremely knowledgeable about the entire process.
have you tried to boot the kernel? At some point I think you are going to bang into module version discrepancy issues and will have to rebuild a kernel anyway. I suspect that there is a way to rebuild anaconda and the install kernel so that you will have the options needed for your hardware.
FWIW I believe that there is also a redhat embedded list that probably has some guys using some pretty funky hardware. Here is a list of all the redhat lists:
http://www.redhat.com/mailman/listinfo/
looks like anaconda-devel-list is a decent bet.
Bret
Sorry I can't be more help.
On Mon, 2005-04-25 at 23:54 -0500, Bret Hughes wrote:
have you tried to boot the kernel? At some point I think you are going to bang into module version discrepancy issues and will have to rebuild a kernel anyway. I suspect that there is a way to rebuild anaconda and the install kernel so that you will have the options needed for your hardware.
You have to rebuild the initrd
A good resource on "customising" the installer is at: http://rau.homedns.org/twiki/bin/view/Anaconda/AnacondaDocumentationProject
On Mon, 2005-04-25 at 20:38 -0400, Ken Young wrote:
Hi,
I'm trying to get Fedora running on an MVME 2700 single board computer. This little board has a 750 PowerPC processor. I've got Debian 3.1
750 would be a G3 based one? /me is trying to remember running off memory
running happily atop the 2.6.11.7 kernel. It runs like a champ, but I'd like to try running Fedora instead of Debian. This single board computer (we've actually got ~35 of them) has no CDROM drive, and I had to do quite a bit of horsing around with the kernel configuration to get it to boot anyway (via NFS), so there's no way for me to just boot off of the iso image CDROMs and install Fedora normally. Is there any way I can
You could use tftpboot from the OF directly, no? Then use the regular Fedora ISOs
Is there any way to do a network install without first booting off a CDROM?
OpenFirmware is your friend