Bug 249873

Summary: grub and lilo fail to install on sata drive
Product: [openSUSE] openSUSE 10.3 Reporter: darren winter <sgt-d>
Component: InstallationAssignee: Olaf Dabrunz <odabrunz>
Status: RESOLVED DUPLICATE QA Contact: Jiri Srain <jsrain>
Severity: Major    
Priority: P5 - None CC: coolo, duwe, fabrice29, hamishm, sgt-d, stefan.fent
Version: Alpha 1   
Target Milestone: ---   
Hardware: Other   
OS: Other   
Whiteboard:
Found By: Other Services Priority:
Business Priority: Blocker: ---
Marketing QA Status: --- IT Deployment: ---
Attachments: YaST logs
yast2 logs....

Description darren winter 2007-02-28 17:43:48 UTC
suse 10.3 alpha 1 64 bit

mobo: asus a8n32-sli

drive: sata wd raptors (74gb versions), in raid 0

sata connector on mobo: nforce4 / silicone graphics / nv raid

setup runs perfectly until right after all of the packages finally install.

(of course getting to this point takes a full hour)

at this point it attempts to install grub which always fails with:

grub error:
stage2 .. and .. dmraid
grub> setup --stage2=boot/grub/stage2 (hd0) (/dec/hda,1)
Error 23: Error while parsing number
grub> quit

lilo was also attempted but also fails to install.

example:

http://www.linuxquestions.org/questions/showthread.php?t=511076
Comment 1 Cyril Hrubis 2007-03-02 16:25:10 UTC
Please attach y2logs from this installation. If you are in doubt follow:

http://en.opensuse.org/Bugs/YaST

Thanks!
Comment 2 darren winter 2007-03-03 16:50:01 UTC
yast2 logs are physically impossible to retrieve.

there is no valid bootable os after grub fails.

we were able to hack in later, but it only worked one time, and we had no way to access other drives or hardware, so there would have been no way for us to copy logs off of that machine.

i do not know if this matters or not, but this is a well known problem.
Comment 3 Fabrice LE CORRE 2007-03-20 21:48:59 UTC
I voted for this bug as I had the same problem with 10.2 final and will probably with 10.3alpha but I'm stuck before reaching this step
Comment 4 Stefan Fent 2007-04-10 15:41:10 UTC
This should be fixed with the latest perl-Bootloader.
Other than that - there is an easy workaround, thus no blocker for a alpha version, decreasing severity.
Comment 5 Stefan Fent 2007-05-10 15:41:38 UTC
still happening with Alpha3?
Comment 6 Fabrice LE CORRE 2007-05-17 16:53:02 UTC
I still had the problem with alpha4 (new install). I fixed it adding manually 
(hd1) /dev/sda in my device.map because yast forgot it.

my config is:
1 IDE drive with Opensuse detected as sda  
1 SATA with XP detected as sdb

Grub is installed in the MBR. I also had to map (hd0) (hd1) map (hd1) (hd0) to cheat windows.

I leave it as NEEDINFO as I'm not the original info provider (darren winter)  
Comment 7 darren winter 2007-05-17 23:50:15 UTC
thanks for the follow ups.  i will try to get my brother to test the newest alpha this weekend or next.
Comment 8 Christoph Thiel 2007-08-08 10:23:19 UTC
Darren, any news on this bug?
Comment 9 darren winter 2007-08-08 12:55:37 UTC
up to and including opensuse 10.3 alpha 5 (32-bit), grub and lilo would not successfully install to the hard drive unless we disabled the nvidia raid chip in bios.  i know we tried sata, but i'm not sure if we tested ide.

anyway, he finally got alpha 5 to install on an ide once we disabled the nvidia raid chip in bios.

we shouldn't have had to do this though, nforce has been out for a relatively long time and their page shows that most major distros already include drivers.  suse may have the drivers but maybe they aren't loading in the kernel, i'm not sure.
Comment 10 Hamish Morrison 2007-08-09 11:00:25 UTC
I also have this problem with openSUSE 10.2.
Although I managed to fix it by editing grub's configuration files manually.

Yast tried to install GRUB to (/dev/mapper/asr_,4) rather than (hd0,4).
I also had to add (hd0) /dev/hda to /boot/grub/device.map
Comment 11 Stefan Fent 2007-08-09 11:27:19 UTC
There is no way we can solve this w / o the logs, as it works here since 10.2 GA
on the machines we have, eg on my workstation (with nVidia SATA RAID).

Comment 12 Hamish Morrison 2007-08-10 09:45:05 UTC
Created attachment 156731 [details]
YaST logs

These logs should provide the info needed, but I've not removed NEEDINFO as I'm not the original info provider.
Comment 13 darren winter 2007-08-13 12:39:01 UTC
my brother (asus a8n32-sli) finally successfully installed to his sata drives and setup a software raid without any problems...   but he had to manually disable his nvidia raid in his bios.  until he did that, suse always attempted to install to /dev/mapper devices and grub (and lilo) always failed miserably.  once nvidia chip was disabled, suse no longer saw or tried to use /dev/mapper devices.

similarly, i (asus a8n-sli deluxe) made sure my nvidia raid setting was disabled in bios, which it has been since i first got my computer (i disabled it immediately), however, suse must still see my nvidia raid chip (somehow - even thought it IS disabled) because every time i try to install to my sata it instantly shows /dev/mapper devices, and several installation attempts have failed.  i am back on ide as we speak.  i just can't get past the /dev/mapper junk.  all i get is:  Error 23: Error while parsing number.

i can mount that drive and get any logs you need, just tell me the file names that you want me to include.

basically, if the nvidia chip is enabled (and some cases even if it is disabled, asd in my case), suse fails to properly setup grub.  sadly, the installer doesn't test the grub installation during the most obvious time (when you are setting up your boot loader and partions) - instead we have to wait 45+ minutes for the software installation to complete just to find out that the os is fully installed but completely unbootable.
Comment 14 darren winter 2007-08-13 13:00:58 UTC
Created attachment 157061 [details]
yast2 logs....

yast2 logs...
Comment 15 darren winter 2007-08-13 13:01:55 UTC
logs attached
Comment 16 Stefan Fent 2007-08-23 15:54:50 UTC
Sorry, but the installation part of the yast2logs is in 
/var/log/YaST2/y2log-1

could you please attach this as well?
Comment 17 darren winter 2007-08-30 17:57:43 UTC
/var/log/YaST2/y2log-1

that file did not exist.

i attempted 2x to install beta 3 x86_64 on the same sata drive, but the dvd kept messing up.

i almost guarantee that grub still wont work though because during the installation attempts it was still trying to use /dev/mapper - which it shouldn't because linux shouldn't see nvidia raid chip if it is disabled in bios like mine is.

anyway, maybe tonight or tomorrow i will make a final sata install attempt.

incidentally, i did post a follow up after your comment, on the same day, but i guess the novell bug page was acting up.  originally i could see the message, but today when i came to reply, my last message was gone.
Comment 18 darren winter 2007-09-02 23:49:19 UTC
ok, the installation keeps failing, over and over again, with the same error, but the log file you mentioned is never there.

anyway, it is still a problem, and i'm not the only one seeing it.  i see reports of the dreaded "error 23" problem at least once every day or two in #suse.

it's not going away, and it's not getting better.

again, it might just be related to my hardware, i don't know.

guess i'm stuck with ide for now.
Comment 19 Richard Creighton 2007-09-03 02:11:41 UTC
(In reply to comment #18 from darren winter)
> ok, the installation keeps failing, over and over again, with the same error,
> but the log file you mentioned is never there.
> 
This problem is directly related to 

https://bugzilla.novell.com/show_bug.cgi?id=304657

Where they are similar is that they both involve mixed hardware, ie, IDE and SATA drives.   In their infinite wisdom, someone decided that the age old idea of calling IDE drives HDxx and SCSI drives which somehow equated to SATA would be called SDxx.  Then someone decided to call ALL drives SDxx including IDE drives and made the change during the kernel builds somewhere after 2.6.18.  Beta versions 10 10.3 are past that point and if you notice, your IDE drives are called SDAx and your SAATA drives designations are bumped up one letter.  Needless to say, a lot of programs, installers, drivers, you name it, needed to be updated and many including GRUB and LILO apparantly were not completely modified, nor was YAST2 - repair which was the subject of the above mentioned bug.  All of the 'fixes' and workarounds mentioned above are efforts to force the system to not have mixed hardware, ie, only SATA or only IDE in which case, the software works until you again put the mixture back in place, then it is iffy again.   Until the real, underlying problem is addressed, no amount of log files is going to help...they are always on the wrong drive from where you expect them to be.  If you need logs, I uploaded some examples for that bug retrieved off of my supposedly immune 10.2 drive which was not supposed to be part of the 10.3 installation.   It is no longer available for me to re-upload it for this bug but it should be of value to prove a point if you go fetch it.  Again, the problem is the renaming of IDE drives to SDxx without allowing for it througout the entire Linux sytem of drivers and install programs.
Comment 20 darren winter 2007-09-04 13:59:16 UTC
thanks for the explanation...  what you've said makes a lot of sense.

while i am running mixed drives after the installation, i always install linux with only one drive attached.  i refuse to let any distro even provide the potential to damage existing data.

so, during all of my failed sata installation attempts, there has never been a mixed hardware issue since there is only one drive attached, and it has only messed up when it's a sata drive.  ide's seem to work fine.

i'm really curious if your explanation applies here or not since i'm only installing to one drive.
Comment 21 darren winter 2007-09-04 15:10:02 UTC
scratch some of what i said.  my dvd drives are ide.  i guess it still sees my other ide chains too even though no drives are attached.
Comment 22 Richard Creighton 2007-09-04 15:51:40 UTC
Check and see if your drive map in /boot/grub doesn't reflect the CDRom occupying space.   Mine is HDA1 and my 10.2 installation is HDA2 and my 1st SATA drive with the 10.3beta is SDB1 not the expected SDA1..   I too disconnected the IDE drive with 10.2 to protect it but left the DVD and the drive mapping still changed.   However, once changed and the OS installed and stabilized, future upgrades go smoothly....thankfully.
Comment 23 Stephan Kulow 2007-09-08 06:15:39 UTC
NEEDINFO outdated
Comment 24 darren winter 2007-09-10 13:29:32 UTC
incidentally, i have found a solution to this problem:

don't install to sata.

it seems to be working out quite well for me at least.

j/k really, but in my case there is no workaround.

no matter what i do, suse keeps using /dev/mapper - and every install fails with the same error that has been reported for months.

for some stupid reason...  someone thought it would be a good idea to force us to install all of our software packages before installing a boot loader.  so we have to wait 30-50 minutes before we get to find out if the boot loader will work or not.  what brainiac came up with that idea?  the boot loader should be installed and fully tested immediately following the partitioning and formatting operations.  period.

final is under a month away, looks like my install will be ide.
Comment 25 Stephan Kulow 2007-09-14 08:24:07 UTC
this seems to be a somewhat exotic setup, so I'd like to downgrade it, especially as there is a workaround (manually fix the bootloader config) and doesn't affect too many.
Comment 26 Torsten Duwe 2007-09-14 11:15:43 UTC

*** This bug has been marked as a duplicate of bug 309837 ***
Comment 27 Hamish Morrison 2007-09-14 15:06:31 UTC
This bug is definitely not a duplicate of bug 309837, if you had taken time to read, you wouldn't have made that mistake.

Stephan Kulow: I do not believe you are right, many people new to Linux won't be capable of manually fixing the bootloader and would give them them a very good first impression of Linux.
Comment 28 Torsten Duwe 2007-09-14 15:14:11 UTC
If you had taken time to read, you wouldn't have reopened it before 309837 is fixed. :-( 

*** This bug has been marked as a duplicate of bug 309837 ***
Comment 29 Torsten Duwe 2007-09-19 09:56:23 UTC
Bug 309837 is fixed now in 10.3 RC1. As the problems some of you have seen may have been side effects thereof, please retest with RC1 and reopen if your problem persists. Thanks a lot for your patience and support!
Comment 30 darren winter 2007-09-21 14:27:45 UTC
i will test rc1 over the weekend.
Comment 31 darren winter 2007-09-23 21:40:58 UTC
OMFG!!!!

MY SATA PROBLEM IS FIXED IN RC1!!!  \o/

bug 249873 seems fixed in rc1, at least for me.

my problem is unrelated to bug 309837.

at the partitioning stage, it no longer says "/dev/mapper", instead it shows "/dev/sda1", etc.

i am running mixed ide/sata...  ide is my dvd drives that i use to install suse, and sata was where i've been trying to install to.

in all of the alpha and betas this was not possible, but it worked perfectly on the first try in rc1.

keep in mind though, i have the nvidia raid chip disabled in bios.

i would consider this problem solved.  looks like 10.3 final will be on sata after all :)

is anyone else seeing success with rc1 on sata where previous 10.3 installs failed?