SCSI subsystem driver - libdata?

[Birdman]

I've had 3 goes at compiling a 2.6 kernel with SATA support and failed 3 times

The help notes in menuconfig advise that most SATA drives are supported by the SCSI subsytem and use libdata. Now, it just so happens that my default sata.i 2.4 kernel that I use to boot Slackware gives this very message when I examine the output of dmesg. The question is, where do I turn this on in the kernel?

PS. I also note that the early output, before the kernel panic, says "IDE: assuming 33MHz system bus speed". I do not get this message when using my default kernel. What have I missed here?

[Birdman]

Am I missing the obvious here?

I have just been browsing the new edition of the Slackbook (jolly good it is too), and in the Kernel section it mentions that all Slackware kernels come with a copy of the config file used to create that kernel. Now, isn't that a great place to start? But, and isn't there always a but, that particular config file is for a 2.4 kernel. Can I load it in to menuconfig to use as a starting point for my 2.6 kernel? Will it be compatible?

[Birdman]

It worked. I now have a shiney new 2,6 kernel

And now (of cause), the BUT.....

I have a freaky mouse. Take a look at the last few lines of dmesg:

psmouse.c: Explorer Mouse at isa0060/serio1/input0 lost synchronization, throwing 1 bytes away.
psmouse.c: Explorer Mouse at isa0060/serio1/input0 lost synchronization, throwing 2 bytes away.
psmouse.c: Explorer Mouse at isa0060/serio1/input0 lost synchronization, throwing 1 bytes away.
psmouse.c: Explorer Mouse at isa0060/serio1/input0 lost synchronization, throwing 2 bytes away.

There is no control at all - talk about ultra-sensitive. I have to kill X with the keyboard.

The mouse is a logitech optical jobby with a scroll well. I connect to PS2 with a converter plug. Tried it as a straight USB but it behalves in the same way. I have now rebooted using my 2.4 kernel and the mouse is back to normal.

Suggestions to a very puzzled linux user are very welcome.

[nordle]

Birdman, sounds like your very nearly there, a couple of things:

The 2.4 config and 2.6 config are NOT compatible.

The mouse thing can easily be sorted, edit /etc/X11/xorg.conf and have the mouse section look a little like this (assuming the use of ps2 converter):

Identifier and driver

Identifier "Mouse1"
Driver "mouse"

Option "Protocol" "IMPS/2"
Option "ZAxisMapping" "4 5"

Option "Device" "/dev/mouse"


As far as your SATA goes, was it not a Silicon Image chip eg (I've tagged options as (Y) to denote it should be built into the kernel, NOT modules etc)

Device Drivers --> ATA/ATAPI --> Silicon Image Chipset Support (Y)
Device Drivers --> SCSI Device --> SCSI Low Level Drivers --> SATA Support (Y) and Silicon Image SATA (Y)

[Birdman]

I am intrigued. The config files may not be compatible but the file did load, I did edit it and I have a system running. Should I expect something really nasty to happen?

My xorg.conf file does show the setting you give. What I actually did do was give xorgsetup and run and I plugged in via USB. The only difference in the settings are "auto" rather than "IMPS/2" and it did not include the ZAxisMapping. I inserted the latter manually. It was not detrimental but the scroll wheel no longer works. I am now back in XP and found that it will not recognise the mouse when plugged in to USB so I am now back to the converter and will reboot to see what happens in linux.

What fun.

[Birdman]

Well I'm back with 2.4

I guess that config file could have screwed the kernel compile? I'll have to go back to the drawing board and carry trying.

[nordle]

Birdman, which 2.6 kernel exactly are you using, I could post my .config, just so you can have a poke around, you could have it and your config in two terminals using make menuconfig and see where the diferences lie maybe.....
Or just straight try to compile it to see if it would work off the bat, then tweak it afterwards.

[Birdman]

nordle, I have been following your every word and have been using the 2.6.11 kernel source with the ck-10 patch

I will be very happy to give your config file a test drive, given all of the trouble I have had starting from scratch.

It maybe my problem lies not with what I haven't selected but perhaps with what I have selected. I get the impression that there can be incompatabilities with certain selections. I thought I had covered all the bases with SiS chip selections and SCSI coverage but failed on each count to have the SATA drive recognised.

[Birdman]

Well, another new one!

Just when I thought I was getting close I got a boot error at the earliest possible stage:

Invalid compressed format (err=1)

What's all that about?

I am getting worse not better

Time for bed me thinks.

[nordle]

Birdman, I've PM'ed you a config which may or may not work, I hope it does as it compiled ok on my machine, but hopefully you've not given up on it yet as its one of those things that when it clicks, it peels another layer away from the Linux onion of mystery.

I promise I haven't been smoking anything odd this morning!

[Birdman]

Well, nordle, many thanks for all of your efforts. I have used your .config. I went through menuconfig just to check whether there might be any of my chipsets missing. I made two changes:

Activated the 4GB of memory - I have 1.5GB in total.
Changed the CPU type for my Athlon

Went through the procedures, including creating new symlinks for the config and Sytem.map files.

The result:

Same errors as previously reported:

VFS: Cannot open root device "803" or unknown-block(8,3)
Please append a correct "root=" boot option
Kernel panic - not syncing:VFS:Unable to mount root fson unknown block (8,3)

So there we are.

The only "successful" compile I have had is by starting with the 2.4 sata.i config file. The system booted but I had the "freaky" mouse problem + a few other wierd things. So, where now

[nordle]

Birdman, I thought you had an Athlon64-3000? no? As I had set the config line to be:
CONFIG_MK8=y
Which is for Athlon64

The error your getting is because lilo.conf is set to read
image = /boot/vmlinuz
root = /dev/???n

where ??? = hda or sda or hde and n = partition number

This is where my current knowledge is incredibly sketchy, but you will have to change lilo.conf, so that it uses the new partition identifier eg

root=/dev/sda1

EDIT:
Dont forget, with lilo, you have to run /sbin/lilo after each time you edit /etc/lilo.conf otherwise your changes will not take effect.

Oh, and I assumed sda1 ie partition 1, but it might be sda2-3-4 etc etc depending on where you installed it.

[Birdman]

I do have an Athlon64-3000. The section in menuconfig that lets you specify your processor was set for x86 where there is an option to select athlon, so I did.

I have installed lilo on a floppy as I do not wish to compromise my boot set-up as hda is seen as the first disk and is where windows has placed its bootloader in the MBR. I have a /boot partition at /dev/sda1 and /root is at /dev/sda3. My lilo.conf looks like this:

# Linux bootable partition config begins for Slackware24 new kernel 2.4.29
image = /boot/vmlinuz
root = /dev/sda3
label = Slackware24
append="hdd=ide-scsi"
read-only
# Linux bootable partition config begins for new kernel 2.6.11
image = /boot/vmlinuz26
root = /dev/sda3
label = Slackware26
read-only

My /boot directory looks like this:

README.initrd boot.0200 diag1.img vmlinuz-ide-2.4.29
System.map config lost+found vmlinuz26
System.map-ide-2.4.29 config-ide-2.4.29 map
System26.map config26 vmlinuz

As you can see it all ties together and Slackware24 boots fine. I used a similar set-up when I got the 2.6 kernel with the freaky mouse to boot. It doesn't therefore look like a lilo problem to me.

There are about 4 screens of output before the kernel panic and it is clear from that that the bios diagnostics spit out all the hardware spec, including the SATA drive but then linux cannot find the partition.

I need to sleep on this one.

Good night for now.

[nordle]

Birdman, I can see I've been as much help as a chocolate tea pot as your a step ahead of my thinking.

The Athlon thing is weird though, as the config file I sent over clearly sais
CONFIG_MK8=y
Which is correct, thats what it should be.

The ONLY other thing that I've seen, is where some people have disabled the SCSI section, as you mentioned earlier something like some parts can conflict with others. I've not seen this myself, but maybe its worth a try.

You could also try maybe the 2.6.12-ck2, but Im clutching at straws here, and you would need to download the 2.6.12 kernel (patch 2.6.11, if you've still got it). But to be honest, Im not sure thats floggin a dead horse with respect to this particular problem.

It must be something obvious, but I cant think of it at the moment. The ##slackware channel on IRC (freenode) is usually very helpful, likewise the newsnet people at alt.os.linux.slackware are very knowledgable, might be worth asking there before going any further as they know a lot more than myself about setting up lots of different systems.

[Birdman]

We have success - of sorts.

>
Birdman, I can see I've been as much help as a chocolate tea pot as your a step ahead of my thinking.
>

You are so wrong nordle. Your last post made me look at what I had been doing and I realised a fatal mistake. I thought I had been using your .config file but realised that it just wasn't possible given your comments. And you were right, it was something really obvious, when you have the experience to realise it that is. I was running

#make mrproper

prior to running

#make menuconfig

The problem was I had copied the .config file across prior to events and did not realise that mrproper would wipe out the .config file. Having sussed that one I changed the order and built a bootable kernel.

However, I have the freaky mouse back so cannot use my new kernel in graphical mode.

Nordle, I have sent you a PM with my dmesg output plus some info that is relevant to the mouse problem.

Thanks for all of your efforts.