Topic: Trying to install SME 6.0.1 on SATA RAID ... BUG in SME ...

[ldkeen]

Jeremy,
I now have 95% of the boot.img file done (including the pcitable entries and modinfo stuff). I'm just waiting on viamraid.o compiled against 2.4.20-18.7BOOT. When you send me that I will recompile the boot.img file and we should be right to test. I've also updated the howto specific to the VIA Raid Controller here http://comnetel.com/viamraid.htm
Lloyd

[jeremy]

Just emailed the viamraid.img driver disk image file to you at Comnetel.

As an aside, I would love to simply get the install to boot on its own.  However, it would be really great if we could get a viable driver disk for this.

A viable driver disk that works for BOOT and Post-BOOT would mean being able to add additional drivers at install time, without having to customize the SME install CD to add drivers.  I think you can see the value in striving for this.

I want to at least get things working, perhaps then you and I can work toward putting together a standard mechanism for creating viable SME driver disks for adding additional drivers.

As a first step, you could do a grab of all the linux commands you executed from the shell to put together the floppy from start to finish.  There is a unix command for starting and stopping a capture session, but I can't think of it right now.

[jeremy]

I just caught your link to the new howto for via raid. (I must be tired.)  Thank you for putting that together.  Note:  That driver also supports the VIA VT8237 controller, which is based on the VIA VT6420 controller.

I have a problem understanding the following line that appears in the via raid howto, as well as in the scsi howto:

You should now have your driver object file in /lib/modules/2.4.20-18.7BOOT/misc/viamraid.o.

Do you mean that the "dot o" (object) file should be placed there?  That implies that one needs to compile that file, but neither howto ever mentions compiling any drivers.

The "make install" command that comes with the VIA Driver Source places a copy of the compiled object file into that directory.

Also, in both howtos, this line:

#cp configs/kernel-source-2.4.20-i386-BOOT.config .config

should be changed to:

#cp configs/kernel-2.4.20-i386-BOOT.config .config

Maybe we can do a merge of my howto which covers the initial steps in detail with your howto which creates a working boot image.

http://Contribs.org/viewtopic.php?t=23760

[ldkeen]

Jeremy,
Can you meet me in the chat?

[ldkeen]

Jeremy,
I've put an image on my server for you and I've fixed the ks.cfg problem.

All my e-mails to you seem to be failing with the following:

Hi. This is the qmail-send program at comnetel.com.
I'm afraid I wasn't able to deliver your message to the following addresses.
This is a permanent error; I've given up. Sorry it didn't work out.

<Jeremy.Jedynak@NetCentricInc.com>:
Sorry, I couldn't find a mail exchanger or IP address. (#5.4.4)

Here is the file http://www.comnetel.com/boot1.img
Regards Lloyd

[jeremy]

I'm not sure what would cause that email problem.  You received my emails right?  I'm getting email from other people just fine.

What did you need to do to correct the ks.cfg issue?  Also, I noticed that in your boot image viamraid.o is set yo ugo+rx, while all the other .o files in modules are set to ugo+rw.

I made a boot disk from the boot image you posted.  I am also burning a new SME install CD with the boot image.

Lets see if it works.

[jeremy]

Well, the SME install CD still doesn't load the viamraid driver, and cannot see any drives without the use of a driver disk.

I updated the SME install CD's boot image by copying the  boot image from Lloyd over images/boot.img  I used that same boot image to create a boot floppy.  

What is very odd is that the first thing that the boot floppy does is load the viamraid driver, and then it goes to SME install and asks for the SME install CD to be inserted.

I just inserted my (boot.img) modified SME install CD and selected "more options", "install".  Lets see if somewhere the driver gets copied from either the boot floppy or the modified CD so that the machine can simply boot from the hard drive(s) and work as expected after the install.

[jeremy]

lloyd,

Install done.  Rebooting after install (with no floppy or cdrom inserted), it looks like it now fails sooner than it did before.

There is also a different (new) error message in place of the old one.  Here it is, word for word:

EXT2_fs: unable to read superblock
isofs_read_super: bread failed, dev=09:01, iso_blknum=16, block=32
Kernel panic: VFS: Unable to mount root fs on 09:01

[ldkeen]

Jeremy,
When re-mastering the SME CD with a new boot.img - it's not just a matter of copying the boot.img file back into the images directory. You have to use mkisofs something like #mkisofs -b myboot.img -c boot.catalogue mynewiso.iso. I havent tried it yet but if you do a man mkisofs or google it you'll get the correct syntax. You HAVE to boot from the modified boot.img for it to work (whether that is dd'd to floppy or mkisofs'd to the CD doesn't matter)

[jeremy]

I used the "buildiso" script which is described in another thread:

http://mirror.contribs.org/smeserver/contribs/dmay/mitel/howto/sme-iso-howto.html

buildiso:

Code: [Select]

#!/bin/sh
# useage = ./buildiso BASENAME-VERSION-RELEASE
# Example: ./buildiso smeserver-6.0.1-01

# set location variables
DESTINATION=/root/smeserver
BUILDROOT=/root/smeserver/cdrom

# remove any previous TRANS.TBL files
/usr/bin/find $BUILDROOT -name TRANS.TBL -exec /bin/rm -f {} \;

# create a rpm list for this build and save to destination
/bin/ls -1 $BUILDROOT/e-smith/RPMS > $DESTINATION/$1.list

# generate the hdlist file
/usr/lib/anaconda-runtime/genhdlist $BUILDROOT

# build the iso, save to destination
/usr/bin/mkisofs -V '$1' \
-b dosutils/autoboot/cdboot.img \
-c e-smith/base/boot.cat \
-J -r -T -pad -o $DESTINATION/$1.iso $BUILDROOT


So, should we be modifying "cdboot.img" instead?

Also, as stated in a previous post, I did boot from the boot floppy, which then asked for the CD-ROM to be  inserted, which then performed an install.  The reboot after that install is when the new error message occured.

So, it is not clear how changing the "cdboot.img" in a similar way will fix anything.

[ldkeen]

I used the "buildiso" script which is described in another thread

Well that should have done it. When you boot if the install autodetects the controller and loads viamraid the you are defintately booting from the modified image.

Also, I noticed that in your boot image viamraid.o is set yo ugo+rx

Maybe that could be our problem?? Try changing the permissions on that file to be same as the others and see what happens. I somehow dont think thats going to do it. Maybe in my adaptec case the driver for 2.4.20-18.7 was already included in the kernel, just wasn't supported at boot time. I'm off all day today and I think I have a box with that controller on it and a spare drive so I'll have a bit of a play and see how we go.

[jeremy]

When booting from the modified CD, the controller is not detected and no drivers are loaded.  When booting from the floppy, the controller IS detected and the driver IS loaded, BUT then the installer kicks in -- so this is useless for booting post install.

The viamraid.o file is ugo+r in both cases, which should be good enough for what its being used for.  I doubt a permissions check is being done.

Based on your comments above ( regarding using the -b myboot.img switch when creating the install iso ) and the lack of driver loading at install (boot from install cd), I believe that we need to be modifying the cdboot.img file.

In any case, we seem to have 2 ways, maybe 3 ways, to get the installer to see the drives, and then install onto them, but we have yet to come up with a single way to get the machine to successfully boot post-install.

This is feeling more an more like a hopeless cause.  SME 6.0.1-1 is a complete failure on modern hardware. We are starting to pray for the kernel update guys to release an install ISO with an updated kernel.

If anyone has any ideas, or leads, or has tried something based on this thread, post it.

[ldkeen]

I know it's a bit of a pain but I think the only way we are going to get that system up and going is to boot from an ide drive in the system load the viamraid module which will give us access to the sata drive, mount the drive and then copy /lib/modules/2.4.20-18.7/misc/viamraid.o across to the sata drive. I checked my system and it already had the adaptec drivers there just wasnt detected during install. Yours is different in that the drivers aren't on the system at all. I remember reading a thread somewhere that some guy had to disable his onboard ide and do an ftp install. It's possible also that you could boot of a Knoppix CD which may load the driver giving you access to the drive and you could copy viamraid across to the sata drive then.

[jeremy]

That would be an act of complete attrition.  It would be far less demoralizing to simply abandon SME for a Microsoft-based solution.

These drivers work with other versions of linux.  Hell, they even work with SME at install time!  Other versions of linux can use these drivers both install to and post install, and some versions use kernels that are older than the one used by SME.

One has to ask what is so special about SME's kernel which each of use can compile, that it refuses to use a driver compiled for it -- even from source.

There is big hole in the collective understanding within this SME community of how drivers are and should be incorporated into linux, and therefore into SME.

The kernel update guys have drivers for SATA RAID for their updated kernels.  It is likely that they too had  to compile (at least some of) them from source.  Maybe someone can grab some their attention and focus it here as well.

The likelihood is high that at some point their new kernels will become obsolete, and that someone will want to use then-modern hardware with SME, and run into the same situation as well.

Solving this issue would move the SME community ahead considerably.

[jeremy]

It looks like this thread has died a slow death.

Its too bad there is no active progress on using SME with modern hardware.