Bug 309070

Reporter, I think if you read bug 218122 you may be able to install without hwprobe=-modules.pata. I've been waiting on an answer to comment20 there before trying.

I just found out this same bug was already reported in comment #11 of Bug #305095. I suggest to close mine as duplicate, but I'll leave that to the Assignee.

I can not install without "hwprobe=-modules.pata", doesn't work. From comments in Bug #218122 I assume this detection should be automatic and I should not have to enter any parameters.

This is normal when you have a IDE disk with more than 15 partitions.

The detection is not automatic. You need to reboot and add the line as
suggested to you boot options.

As already reported, those options do not work. Please read again my report! :-/

Ie: after rebooting with that option, it asks for CD1, and does never find it. I'm using DVD.


Plus: according to Bug #218122, it is/will be automatic.

So these things need to be fixed in linuxrc.
Reassigning to Steffen.

What is written in bug #218122 has nothing to do with installation.
So it is not relevant for this case.

No, linuxrc works fine. It's _yast_ that doesn't see the CD (see comment 0).

But this very much looks like a dup of bug 303799. Carlos, could you
check whether 'lshal' works (on any console the moment yast complains
about the missing CD)?

Carlos, no need to check. See my comment in bug 303799.

*** This bug has been marked as a duplicate of bug 303799 ***

(In reply to comment #5 from Thomas Fehr)
> What is written in bug #218122 has nothing to do with installation.
> So it is not relevant for this case.
 
My reading of https://bugzilla.novell.com/show_bug.cgi?id=218122#c18 is that reporter can, because of the bug 303799 problem, as an alternative to using hwprobe=-modules.pata to start the installer, do as specified in that comment - if the fix for that bug was checked in prior to beta3 freeze.

(In reply to comment #7 from Steffen Winterfeldt)
> Carlos, no need to check. See my comment in bug 303799.

Too late, I had already rebooted, had lunch, tried, and rebooted back to 10.2 to report back to you.

I got on 'lshal':


error: dbus_bus_get: org.freedesktop.DBus.Error.FileNotFound: Failed to connect to socket /var/run/dbus/system_bus_socket: No such file or directory

However, I can mount manually the dvd or a HD partition.

I have the y2log file of that test if you want it.

(In reply to comment #8 from Felix Miata)

> My reading of https://bugzilla.novell.com/show_bug.cgi?id=218122#c18 is that
> reporter can, because of the bug 303799 problem, as an alternative to using
> hwprobe=-modules.pata to start the installer, do as specified in that comment -
> if the fix for that bug was checked in prior to beta3 freeze.
> 

Sorry, Felix, but if I were to try that, I'd need a "guide for dummies" O:-)

What I'm afraid is that people with more than 16 partitions on PATA will not be able to install 10.3 GM at all, and maybe not even 10.4 as pata support will be removed completely by then :-(

Installation on 10.3 will be possible as soon as the problem of detection of
CD/DVD drive named hd* instead of sd* is fixed. Steffen is working on this.

On the long term you are right. You either have to consolidate your disk
to only use up to 15 partitions or switch to a distribution that pledges to
support plain old IDE driver forever.

16 partitions is a ridiculous limit, a step back to worse; present day bigger disks are likely to have more partitions than ever. Testers, for instance, need several partitions per tested OS (like 10.2, 10.3, ubuntu, windows, etc). Even some applications prefer their own partitions.

There are people using more than 40 (see Bug 218122#c8). A solution has to be found. Perhaps use something different than scsi. According to http://lists.opensuse.org/opensuse/2007-08/msg01741.html, that is indeed what some think and are designing:

] Libata being a part of scsi is considered a major shortcoming.  The                              
] plan according to the lkml-ide list is to eventually seperate it out                             
] into its own infrastructure.  Maybe use a /dev/diskX type of naming                              
] convention.                                                                                      

Since the limit for minor device numbers to be only up to 255 is gone since
years, it would not be a problem to support more than 15 partitions on any SCSI
disk. This could even be compatible to current major/minor layout of SCSI
disk. Just use current minor number for partitions up to 15 as is and use
minor numbers above 255 for partition numbers above 15.

But as long as this is not done in the kernel there is not much userland
can do to work around this. The plan to use kpartx to create DM maps
instead of kernel partitions create IMHO more problems than it solves.