Hi,
I'm running into some trouble installing Fedora 7 using pxeboot. I've searched the forum/mailinglist but did not find a complete answer, that's why i'm posting my question.
My server has two e1000 compatible NICs. I've found some forum posts that explain e1000 issues with Fedora 7 but i could not verify that i'm having the same problem.
I'm seeing the following: After anaconda starts it will request a IP using dhcp. On the dhcp-server i see incoming requests and outgoing acks. For some reason anaconda can't access the network after the dhcp request. This will result in anaconda giving me an error saying that it can't access the installation tree.
It looks like anaconda/dhclient doesn't pick up the ACK, or something's preventing network access. Unfortunately it's too early in the bootstage for me to check. (ALT-F2 doesn't have a shell yet) A ping from the dhcp-server to the acked IP address results in 100% packetloss.
When i use the same parameters in FC6 everything works smoothly.
Has anyone had a similar experience?
Best regards, Peter
Peter Bosgraaf wrote: ...
Has anyone had a similar experience?
Yes, with an HP dc7700.
I use PXE to boot the installation and read the kickstart file, the rest is then read from a DVD simply by replacing the nfs line with a cdrom line in the .ks file.
Someone have suggested to use the e100 driver instead, how do I do this during installation?
Mogens
Mogens Kjaer wrote:
Someone have suggested to use the e100 driver instead, how do I do this during installation?
My best guess would be to adjust the modules.alias inside the initrd.
On the side, can anyone tell me what's causing the problem? Is it the kernel, the e1000.o module or anaconda?
Regards, Peter
On 7/16/07, Peter Bosgraaf peter@bosgraaf.org wrote:
Mogens Kjaer wrote:
Someone have suggested to use the e100 driver instead, how do I do this during installation?
My best guess would be to adjust the modules.alias inside the initrd.
On the side, can anyone tell me what's causing the problem? Is it the kernel, the e1000.o module or anaconda?
I've already confirmed that it works fine post-install. My guess is something wonky in the installer kernel and/or anaconda.
Quote:Someone have suggested to use the e100 driver instead, how do I do this during installation? Did this "someone" confirm that this was successful? I've just tried the exact same thing and got nowhere :
* Unpacked the initrd * Edited modules.alias - replaced all references to "e1000" with "e100" * Rebuilt initrd
Now the installer is in an infinite cycle of asking me to configure TCP/IP - if I select the DHCP option, it simply asks me the same question again; if I select the "manual" option, the installer crashes. The module loads (according to the console messages), but evidently no DHCP request is issued.
Help!
gumboot wrote:
Quote:Someone have suggested to use the e100 driver instead, how do I do this during installation? Did this "someone" confirm that this was successful? I've just tried the exact same thing and got nowhere :
I can confirm that it doesn't work :-(
My suggestion was inspired by:
http://forums1.itrc.hp.com/service/forums/questionanswer.do?threadId=1081719
I did the same, the NIC in this machine has the PCI ID: 8086:104a, and I changed modules.alias so that this ID points to e100 and got the same result as you did.
Mogens
Mogens Kjaer wrote:
gumboot wrote:
Quote:Someone have suggested to use the e100 driver instead, how do I do this during installation? Did this "someone" confirm that this was successful? I've just tried the exact same thing and got nowhere :
I can confirm that it doesn't work :-(
My suggestion was inspired by:
http://forums1.itrc.hp.com/service/forums/questionanswer.do?threadId=1081719
I did the same, the NIC in this machine has the PCI ID: 8086:104a, and I changed modules.alias so that this ID points to e100 and got the same result as you did.
I tried making a respin of f7 x86_64 and it seems to work, however only when the machine is connected to a 100Mbit switch, not with a gigabit switch.
Try to see if you can find a respin somewhere.
Mogens
As it happens I was respinning Fedora 7 when I checked back here! Yes, it would appear that rebuilding the whole distro solves the problem, but I can't help thinking that I'm using a sledgehammer to crack a nut.
This is i386, but I'll need to respin the x86_64 version too at some point (I have over 150 i386 boxes to reimage in the next few months - mostly with e1000 cards - and 30+ x86_64 boxes arriving shortly). I'll keep checking for a simple solution, but until then it's good to know there's *A* solution out there!
Cheers, G
gumboot wrote:
As it happens I was respinning Fedora 7 when I checked back here! Yes, it would appear that rebuilding the whole distro solves the problem, but I can't help thinking that I'm using a sledgehammer to crack a nut.
This is i386, but I'll need to respin the x86_64 version too at some point (I have over 150 i386 boxes to reimage in the next few months - mostly with e1000 cards - and 30+ x86_64 boxes arriving shortly). I'll keep checking for a simple solution, but until then it's good to know there's *A* solution out there!
I've set up the kickstart installation to use the DVD.
I PXE boot, load the kickstart file from an http server, do the actual install from a local DVD.
Net access in the %post section seems to work for some unknown reason, at least I can do an NFS mount and transfer files.
But you'll need to visit every machine and insert the F7 DVD.
Mogens
gumboot wrote:
As it happens I was respinning Fedora 7 when I checked back here! Yes, it would appear that rebuilding the whole distro solves the problem, but I can't help thinking that I'm using a sledgehammer to crack a nut.
I've patched the initrd (for x86_64 only) with the latest e1000 (7.6.5) from Intel. Now the installation works fine.
I've put a copy on: ftp://ftp.crc.dk/pub/f7e1000
The e1000.ko file, however, is nearly 20 times bigger than the original, but it seems to work.
Mogens
Mogens Kjaer wrote:
gumboot wrote:
As it happens I was respinning Fedora 7 when I checked back here! Yes, it would appear that rebuilding the whole distro solves the problem, but I can't help thinking that I'm using a sledgehammer to crack a nut.
I've patched the initrd (for x86_64 only) with the latest e1000 (7.6.5) from Intel. Now the installation works fine.
I've put a copy on: ftp://ftp.crc.dk/pub/f7e1000
Their probably bigger because you didn't strip them?
Patched i386 and x86_64 can be found @ http://www.shells.nl/junk/F7e1000/
Cheers, Peter
Peter Bosgraaf wrote:
Patched i386 and x86_64 can be found @ http://www.shells.nl/junk/F7e1000/
Thanks! I've successfully used this on a Dell OptiPlex GX260 that had the described problem.
Bjørge