Description of problem: it seems that, this time, I can't install Fedora rawhide because /sbin/loader crashes after NetworkManager is launched. The weird thing is that this happens only when I specify a full path to a kickstart file on the boot prompt. It doesn't crash after NetworkManager if I install without passing a kickstart file on the command line. Version-Release number of selected component (if applicable): 12.26 How reproducible: always Steps to Reproduce: 1. Download today's boot.iso (x86_64) from a rawhide development tree mirror (it should contain anaconda 12.26) 2. Put it on a USB pendrive using livecd-iso-to-disk 3. Boot the USB pendrive then edit the first entry to add ks=http://arbiter.fedorapeople.org/fedora.ks 4. Make sure you have one ethernet card, /sbin/loader running NetworkManager 5. Watch it die with a backtrace immediately after hiding the "NetworkManager is configuring eth0..." dialog. Additional info: Screenshot of the backtrace (too tired to copy it here) - http://img11.imageshack.us/img11/5462/63141252.jpg
$ eu-addr2line -e usr/lib/debug/usr/lib/anaconda-runtime/loader/loader.debug 0x40782e /usr/src/debug/anaconda-12.26/loader/loader.c:1725 0x418098 /usr/src/debug/anaconda-12.26/loader/urlinstall.c:118 0x408e11 /usr/src/debug/anaconda-12.26/loader/loader.c:1578
JFYI: This happens on another computer with a different ethernet card. I guess this is not an hardware problem?
note to self to add this to common bugs if it's not fixed for beta release... -- Fedora Bugzappers volunteer triage team https://fedoraproject.org/wiki/BugZappers
Do we support using livecd-iso-to-disk with boot.iso - isn't it only for livecd images? Nevertheless, this bug can be reproduced also this (hardly ever used I think) way: - using pxe - adding stage2= boot parameter with value not parseable as http or ftp url - using ks with url -url command I am working on patch.
"Do we support using livecd-iso-to-disk with boot.iso - isn't it only for livecd images?" Hmm - install.img is intended as an equivalent of boot.iso that can be dd'ed straight onto a USB stick, isn't it? If so, you're right, this is pretty much unsupported procedure. -- Fedora Bugzappers volunteer triage team https://fedoraproject.org/wiki/BugZappers
By using livecd-iso-to-disk I basically get only the kernel and the initrd, which should be the same as CentOS "netinstall" iso. Never tried to dd the install.img on a USB stick though.
(In reply to comment #5) > "Do we support using livecd-iso-to-disk with boot.iso - isn't it only for > livecd > images?" > > Hmm - install.img is intended as an equivalent of boot.iso that can be dd'ed > straight onto a USB stick, isn't it? If so, you're right, this is pretty much > unsupported procedure. > No, install.img is part anaconda's filesystem. You would want to dd the boot.iso to an usb stick.
Found the source of all problems. I tried again with today's boot.iso image (anaconda 12.35). This time I dd'ed the boot.iso image with no success. Anaconda crashed again. I decided to give a look at the logs while it tried to initialize the kickstart procedure and saw this: http://img410.imageshack.us/img410/4650/img0249l.jpg (See the last two lines) So I tried to remove stage2=hd:LABEL:"Fedora" from grub command line and guess what? Anaconda started the installation as usual (by downloading install.img from the server). Just my $0.02: I think it's better to add a "Do Kickstart installation" option in grub menu without stage2=hd:LABEL:"Fedora" Hope this helps.
The usual method with the boot.iso is to use the install.img that is on the cdrom. You could try to edit the bootloader to use stage2=hd:LABEL:Fedora:/images/install.img That should prevent the downloading of stage2 from the net and use the one on the usb key. Think this is the same issue that I outlined at: https://bugzilla.redhat.com/show_bug.cgi?id=524524
The segfault should be fixed in anaconda-12.37-1.
(In reply to comment #9) > [snip] > You could try to edit the bootloader to use > stage2=hd:LABEL:Fedora:/images/install.img With that change it works. (In reply to comment #10) > The segfault should be fixed in anaconda-12.37-1. Confirmed as fixed.
Thanks for the bug report, comments, and testing.