Bug 118941

Created attachment 98763 [details]
Error message on ALT-F1

Created attachment 98764 [details]
Kernel finds drive if you 'mknod /dev/hda' (ALT-F2)

Created attachment 98765 [details]
Where are we in the install when this happens? (ALT-F3)

Created attachment 98766 [details]
Random stuff from ALT-F4

Attaching /tmp/syslog and /tmp/anaconda.log will be far more helpful :)

Undoubtedly, but how? I'm in the middle of a failed install, with no
floppy drive or serial ports. Would a USB pen work at that point? Or
can I perhaps trick the network into coming up and using ssh?

Created attachment 98769 [details]
/mnt/syslog

USB pendrive didn't work, but I got the network to cooperate...

Created attachment 98770 [details]
/tmp/anaconda.log

Created attachment 98951 [details]
anaconda log from HP zv5000z AMD64 nForce3 notebook

Created attachment 98952 [details]
syslog from HP zv5000z AMD64 nForce3 notebook

That looks like the same problem I'm having with my HP zv5000z Athlon
64 notebook (nForce3 chipset, not the VIA chipset the eMachines
notebook uses).  I've attached my anaconda.log and syslog.  I used scp
to copy the files to my Linux server.  Nice to have that available.  I
was trying to install from a boot CD and the four FC2 test2 .iso's on
my server via NFS.  linux idle=poll selinux=0 text (idle=poll works
around the unpatched K8 Errata #93 bug in my HP's BIOS). 

this seems similar to <a
href="http://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=118881
">Bug 118881</a>.

I've tried several ways to overcome the "no IDE drives found" problem
to no success:

1. boot the fedora x86_64 DVD
2. mknod /dev/hda
3. boot with FC1 DVD and change to FC2T2
4. boot with vmlinuz + initrd from the FC2T2 DVD copied into /boot as
   explained in:
   www.redhat.com/archives/fedora-test-list/2004-February/msg00445.html

The previous four methods have already been discussed but didn't work
for me.

4. boot with boot-iso from devel
5. boot with vmlinuz + initrd from the devel tree (same trick than before)

In both cases the hard disk and the cdrom are detected by the kernel,
but anaconda asks for a driver or driver disk.

If I choose Hard Disk, in vt3 it says:
devices[0] is TOSHIBA MK60221GAS - hda using ignore (not loaded)

If I choose Local CDROM, in vt3 it says:
devices[0] is PIONEER DVD-RW-DVR-K12RA - hdc using ignore (not loaded)

*** Bug 119550 has been marked as a duplicate of this bug. ***

I'm seing the same problem on an HP d220 desktop machine.  (This is a
Pentium 4 machine, so I guess the problem isn't limited to Athlon64
machines.)  I'd be more than happy to do more testing and be a guinea
pig if you'd like me to.

(For what it's worth, this is one of those machines that CD1 wouldn't
boot in, even though I verified the md5sum before burning, and
verified the disk after burning through the anaconda-runtime iso
checking mechanism.  Booting from a burned boot.iso worked great,
until it couldn't see any IDE CD-ROMs to install from.)

Created attachment 99029 [details]
FC2 T2 Installing successfully after using blank (no partitions) drive

I had the same problem with my M6805 eMachines notebook and was able
to get things working by installing on a completely clean drive with
no partitions.

Now, this is not the best solution for most people, but if you don't
mind wiping things out, it might work. 

This probably points to something wrong in the way
kudzu/partedUtils/isys reads the partition table. If anyone is
interested in the partition table of the drive that _doesn't_ work,
let me know.

PS - I completely replaced the drive with an old 20gb unit I had
laying around, so the smaller capacity may have had something to do
with it working as well (the original unit was 60gb), but I doubt that
it played a part.

PPS - I did not use any kernel cmdline options. 

I just created a DVD from the rawhide-20040331 tree including the
newer anaconda... No change for me, anaconda still could not find
/dev/hda.

After dd-ing /dev/zero to /dev/hda it did reinitialize the drive and
install, but as mentioned that is sub-optimal... Thanks for the
headsup though, at least I can start using it.

This problem you had/have with partition tables and parted may be
related to https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=113201

I have a system with two identical Maxtor 60GB drives on hda and hdb.
 The hda drive's partition table was initially done by Win2k when it
was installed.  Further partitioning of the disk was done by fdisk on
FC1.  The hdb drive was partitioned by fdisk on FC1.

FC2, the 2.6 kernel and parted do not like the hda partitioning and
complain.  I have to use the kernel boot hda=c,h,s parameter to get
things done right.  There is no complaint about hdb.

Trying the hda=c,h,s doesn't make any difference for me with an
emachines m6805.

Removing all partitions from HD and then installing worked for me as
well. 

I'm on a duel boot machine /dev/hda is WindowsXP /dev/hdb is FC2-t2
Whiping out /dev/hda is not an option I have several gigs of data on
it. This machine had FC1 working with the exact same configuration so
when I tried installing again I was greeted with these errors:
                                                                     
                                                       
Unable to align partition properly. This probably means that another
partition tool generated an incorrect partition table, because it
did'nt have the correct BIOS geometry, It is safe to ignore, but
ignoring may cause (fixable) problems with some boot loaders.
                                                                     
                                                       
(I saw this the first time also but assumed (fixable) was nothing to
worry about and clicked okay. The second time around installing I
clicked cancel and came across a new warning):
                                                                     
                                                       
The partition table on device hda was unreadable to create new
partitions it must be initilized, causing the loss of ALL DATA on this
drive.
This operation will overide any previous installation choices about
which drives to ignore.
would you like to initilize this drive, erasing ALL DATA?
                                                                     
                                                       
#fdisk -l
Disk /dev/hda: 122.9 GB, 122942324736 bytes
16 heads, 63 sectors/track, 238216 cylinders
Units = cylinders of 1008 * 512 = 516096 bytes
                                                                     
                                                       
Disk /dev/hdb: 20.4 GB, 20490559488 bytes
255 heads, 63 sectors/track, 2491 cylinders
Units = cylinders of 16065 * 512 = 8225280 bytes
                                                                     
                                                       
If any other info is needed e-mail me i'll be up all night burning data.

It isn't limited to /dev/hdN devices. I tried installing from a USB
hard disk. Upon choosing the USB Mass Storage option for drivers, I get:

devices[1] is Tekram TR-621 2.5 - sda using ignore (not loaded)

Even while the dmesg screen showed that it was able to open the
partition table. It's a 20 GB disk with a single 16 GB partition.

I ran into the same problem using a Gigabyte GA-K8VT800 motherboard. 
Everything worked fine in AMD64 Core 1. However, it's interesting 
that the kernel could start without the "idle=poll" parameter, 
apparently Core 1 can only start with idle=poll even with the latest 
(F6) BIOS installed.
    

Just tried installing from a tree using
anaconda-9.92-0.20040407220217, but the situation has not improved.
Also I noticed that if I boot the boot.iso and attempt a network
install, the via-rhine driver is loaded, but anaconda claims there is
no network device to install with and presents a list of drivers to
try. via-rhine is not in the list because the module is already
loaded... Is this a udev problem, it's acting like anaconda just has
no /dev entries?

About that via-rhine missing:

In stage 2 it is detected, and can be configured in the interface
configuration just as usual.

1.)AS TO #22 FROM VINCENT:
2.)AS TO #13 FROM PAU:

1.) AS TO #22 FROM VINCENT: while installing FC2-t1 in order to have a
dual boot system (with WinXPpro) I had similar problems like the ones
you described - after using parted in fact I've gone through the whole
installation - and winXP won't boot any more. 
- checking the disk with Partition Magic on another machine I got a
message, that the geometry specified is incorrect - the disk was de
facto unusable. 
I've been given a hint to use the recovery utilities from WinXPpro
setup disk (image for 6xfddisk, can be found on microsoft's site). It
helped - winXP booted again. 

2.) AS TO #13 FROM PAU: I'm experiencing the same problem 
I've got :
mainboard: Epox 4beav
IDE hdd: 80GB Seagate st380021a
several partitions (primary and extended)- done with some MS tool (I
can't remember if msdos fdisk or some other).

I've currently got a dual boot (FC2-t1 + winxpPro); tried to upgrade
to dual boot with FC2-t2+WINXP; 

I get stuck with the same kind of error message Pau does:

devices[0] is SEAGATE ((drive type)) - hda using ignore (not loaded)

Jeremy,

Today's rawhide (20040414) is a *VAST* improvement, Thanks!

The hard drives are detected, as is the video and mouse. (anthough
VESA was used rather than ATI, and the display is not probed.)
Keyboard is an issue, but that's a kernel problem for me, and plugging
in a usb keyboard works for the install. I am backing up my customized
kernel that works with the keyboard and PCMCIA, then I will burn a DVD
and re-install.

Existing installs on the hard drive are detected.

Network installs are still not working, driver is loaded, but anaconda
is not seeing it still. 

Next iteration of this problem.

Burned DVD and began resintall, things worked wonderfully up until it
began to install packages. On the first package I received an error
suggesting it could not unmount the CD, and would I make sure I'm not
cd'd into /mnt/source on VT2. (I was not) VT3 shows:

* Unable to copy comps package
* moving (1) to step installpackages
* switching from iso [1,2,3,4] to [] for fedora-logos-1.1.23-1.noarch
* isys.py:mount()- going to mount /tmp/hdc on /mnt/source
* exception was int argument required for fedora-logos-1.1.23-1
* exception in umountCD: isys.umount() can only unmount by mount point

It might help to mention that /mnt/source does not exist anymore.

Okay, the DVD handling should be fixed.  I got bitten by an API change
in rpm-python (again).  Network installs is a dupe of 120360.  The
original thing here about the hard drive not being detected being
fixed, I'm going to close this as rawhide.