Red Hat Bugzilla – Bug 114218
FC1 test install fails after pxeboot+hard disk install
Last modified: 2007-11-30 17:10:35 EST
From Bugzilla Helper:
User-Agent: Mozilla/5.0 (X11; U; Linux i586; en-US; rv:1.4) Gecko/20030701
Description of problem:
Installing from hd fails after booting w/pxeboot setup. Error is
"Fedora tree at... does not seem to match boot image", but the boot
images (vmlinuz, initrd.img) were fetched from matching pxeboot
directory within that very same tree.
Version-Release number of selected component (if applicable):
Steps to Reproduce:
1.populate tftpboot on server with vmlinux and initrd.img and
configure pxelinux.cfg/default so as to allow booting
append initrd=initrd.img text devfs=nomount ramdisk_size=9216
2. Download iso images to /dev/hdd1
3. Boot with PXE and try harddisk install
Actual Results: Error message about mismatch between boot and install
messages on Alt-F4 suggest that it is trying to open the cdrom and
mount something from loop device
Expected Results: proceed with install
Severity set to high since it basically prevents me getting anywhere
My disk setup is slightly weird: hde-hdh are on an Adaptec 1210SA
controller which doesn't work with the install kernel, hdd is an old
drive that I borrowed and attached to a vacant (slave) plug. Hence the
disabling of hde and hdg.
I had a similar problem with net install from www.mirror.ac.uk. Also
(I think) temporarily from download.fedora.redhat.com, but at some
point it decided to go on to disk partitioning etc. then crashed
trying to obtain package information. The two sites had different
netstg2.img files when I checked. Of the two sites, the latter is
about ten times slower, so I'm not really motivated to download from
The images must match exactly. It sounds like you're just seeing
mirror lag (ie boot image bits get to the mirror before netstg2.img etc)
Notice that this also applies to HARD DISK installs where hdstg2 is
sitting inside the .iso files. The actual PXE boot files (vmlinuz,
initrd.img) does not appear to differ between source and mirror.
Besides: all of these files and directories have dates at least one
week old. How is one supposed to know if things have gotten out of
step? (Preferably before hauling 2GB over the 'net!) And shouldn't
there be a way of bypassing the check?
Or are you telling me that there is a correct kernel and initrd image
sitting inside the .iso files, differing from the extracted tree?
Oops, thought this was devel, not test1. *sigh*
Justin -- do you know of any problems with hard drive installs?
Things should all match up.
And the check is very intentionally not bypassable. If your images
don't match, then you very likely don't have the same kernel version
and will then be missing all of the kernel modules in the second stage
which means no ext3 and numerous other problems.
I have not seen this issue with hard disk installs from iso images.
Is it possible the pxeboot environment was not sending the disk1
Confirmed: The pxeboot and isolinux dirs inside the yarrow*disc1.iso
both contain an initrd.img that is different from the one in the
FTP/HTTP dir on download.fedora.redhat.com
linux:~ # md5sum /mnt2/images/pxeboot/*
linux:~ # logout
Connection to 192.168.1.10 closed.
[root@butterfly tftpboot]# md5sum initrd.img vmlinuz
And the obvious workaround works!
This is fine with the final release AFAIK.