I dont have scsi or ide/ata support compiled into my kernel, only the
libata drivers, now that they're split out from the scsi tree. As far
as I could see from the "help" dialog, this was a valid configuration.
However, I can't load the root fs using any scsi device name or ide
device name. eg, /dev/sda1, /dev/sdb1, /dev/sdc1, /dev/hda1 etc etc. I
tried the LABEL=/ option in the boot args, which should work and that
gave me nothing.
My question is, do I still need to compile in scsi disk/cdrom/generic
support into my kernel to get libata devices to work or is there some
other syntax i'm missing? Libata detects my drives, but as far as I
could see, and it flies by too fast to read, no device nodes were
assigned to them.
If you do need scsi support, why isn't that done automatically when you
select your libata drivers? or at least a pointer in the "Help" dialog
to tell you to enable that ?
> My question is, do I still need to compile in scsi disk/cdrom/generic
> support into my kernel to get libata devices to work or is there some
> other syntax i'm missing?
The initrd image provides the ramdisk image which contains modules not
compiled into the kernel which are necessary to access the root
filesystem. Man mkinitrd should explain how to set it up.
Casey Dahlin
Casey Dahlin wrote:
>> My question is, do I still need to compile in scsi disk/cdrom/generic
>> support into my kernel to get libata devices to work or is there some
>> other syntax i'm missing?
>>
>
> The initrd image provides the ramdisk image which contains modules not
> compiled into the kernel which are necessary to access the root
> filesystem. Man mkinitrd should explain how to set it up.
>
> Casey Dahlin
>
>
I'm not using initrd. All my drivers are compiled in that are needed at
boot. My question was in regards to the vagueness of just what the
Libata drivers require to actually work in a system. The libata drivers
will load and detect devices just fine without scsi compiled into the
kernel, they will however be completely useless since they wont be
assigning the drives detected on the busses to anything.
If libata is to be considered an alternative to scsi and ide (as it
should be, and as it is by it's location in the config now) then why is
it still treated as a subset of scsi in the config, requiring you to
compile in scsi drivers just like you would have had to do anyway if
libata was back in it's normal position under SCSI to make any actual
use of libata devices?
It seems like the move out of the scsi tree was pointless, and just adds
confusion since the driver tree isn't ready and hasn't got the makefiles
and kernel config scripts written to reuse the scsi drivers it requires
when someone wants a libata disk or cdrom or whatever device
transparently. If libata is to be on the same level as ide and scsi in
the config, then I shouldn't need to go into scsi to get a libata drive
to work.
It should be a matter of adding some config options under the libata
section for "disk support" and "cdrom support", etc and kconfig should
auto enable the scsi device options required. Maybe just move those
top level scsi drivers into some shared block layer directory since
eventually we'll be left with usb/libata/scsi and they'll all use those
top level drivers to do their device access.
Ed Sweetman wrote:
> My question is, do I still need to compile in scsi disk/cdrom/generic
You need scsi disk.
--
Norberto Bensa
Cel: 011-5654-9539
Ciudad de Buenos Aires, Argentina
[cc'ing linux-ide]
Ed Sweetman wrote:
> My question is, do I still need to compile in scsi disk/cdrom/generic
> support into my kernel to get libata devices to work or is there some
> other syntax i'm missing? Libata detects my drives, but as far as I
> could see, and it flies by too fast to read, no device nodes were
> assigned to them.
Yes, you do need to select SCSI high level drivers you need - sd and sr
should suffice for libata.
> If you do need scsi support, why isn't that done automatically when you
> select your libata drivers? or at least a pointer in the "Help" dialog
> to tell you to enable that ?
Probably because no other SCSI low level driver selects high level
driver automatically. I agree that we need better help message. Feel
free to submit patch.
Thanks.
--
tejun