Re: SCSI module eata no longer loading automatically from initrd on Sid on i386
On Wed, 24 Mar 2010 07:05:13 -0400 (EDT), Arthur Marsh wrote:
> Why shouldn't the eata driver be loaded once the PCI bus has been
> scanned and a device that the eata driver knows about [1044:a400] in
> this case is detected?
>
> 00:08.0 SCSI storage controller [0100]: Adaptec (formerly DPT)
> SmartCache/Raid I
> -IV Controller [1044:a400] (rev 02)
> Flags: bus master, medium devsel, latency 160, IRQ 11
> BIST result: 00
> I/O ports at d400 [size=32]
> Expansion ROM at ee000000 [disabled] [size=32K]
Because the hotplug system doesn't know that eata is the driver for this
board. Once the eata driver gets loaded, it can detect that the board
supports the EATA/DMA protocol, and therefore it grabs the board.
But the driver has to get loaded first. And the driver will only be
loaded automatically by the hotplug system if a module alias (internal
or external) matches the device identification of the card. For
comparison purposes, let's look at the 3w-9xxx driver. This is from
a Lenny system on the i386 architecture:
----------
debian:~# modinfo 3w-9xxx
filename: /lib/modules/2.6.26-2-686/kernel/drivers/scsi/3w-9xxx.ko
version: 2.26.02.010
license: GPL
description: 3ware 9000 Storage Controller Linux Driver
author: AMCC
srcversion: 81BF6C7871BF808D126874D
alias: pci:v000013C1d00001005sv*sd*bc*sc*i*
alias: pci:v000013C1d00001004sv*sd*bc*sc*i*
alias: pci:v000013C1d00001003sv*sd*bc*sc*i*
alias: pci:v000013C1d00001002sv*sd*bc*sc*i*
depends: scsi_mod
vermagic: 2.6.26-2-686 SMP mod_unload modversions 686
debian:~#
----------
Do you see those four "alias" lines above? These are four internal aliases
defined for the module. They represent four different PCI device
identifications: 13c1:1005, 13c1:1004, 13c1:1003, and 13c1:1002. If the
hotplug system finds any of these PCI devices in the system, it will
load the 3w-9xxx module, unless a
blacklist 3w-9xxx
statement is found in a file in the /etc/modprobe.d directory. By contrast,
look at the output of the "modinfo" command for the eata module:
----------
debian:~# modinfo eata
filename: /lib/modules/2.6.26-2-686/kernel/drivers/scsi/eata.ko
description: EATA/DMA SCSI Driver
license: GPL
author: Dario Ballabio
depends: scsi_mod
vermagic: 2.6.26-2-686 SMP mod_unload modversions 686
parm: eata: equivalent to the "eata=..." kernel boot option.
Example: modprobe eata "eata=0x7410,0x230,lc:y,tm:0,mq:4,ep:n" (string)
----------
Do you see the difference? There are no "alias" entries for this module.
For whatever reason, the author/maintainer of the eata module elected
not to define internal aliases for each board that the driver supports.
I'm guessing that this is because there are so many boards that the driver
supports that maintaining the list would be a never-ending job. Also,
some of the boards may be better supported by a more hardware-specific driver,
and if the eata module listed the board then the hotplug system would load
both drivers: the specific one for that board and eata. And the two
drivers would interfere with each other.
I have that problem on my IBM
ThinkPad 600. By default, the hotplug system wants to load three different
sound drivers: snd-cs4232, snd-cs4236, and snd-wavefront, because all three
of them have an alias called "pnp:dCSC0000*" which matches the ISA plug-and-
play device id of my CS4237B sound chip. I only want one of them: snd-cs4236.
I have to manually blacklist snd-cs4232 and snd-wavefront to keep the
hotplug system from loading them. On the other hand, the mwave module,
which is the correct driver for the ISA plug-and-play device pnp:dIBM3760,
has no internal aliases defined. I have to define an external alias for
it to get it to load. The mwave driver was written before Linux had a
hotplug system, and IBM didn't bother to go back and add the alias
retroactively.
In the case of eata, the best way to handle the situation is to
create an external alias.
What I would do is to create a file called /etc/modprobe.d/local. In it
I would put the following statement:
alias pci:v00001044d0000A400* eata
Then, remove the "eata" entry from /etc/initramfs-tools/modules.
Also make sure that eata is not listed in /etc/modules. Then,
run update-initramfs again. Check to make sure that the eata module is
still included in your initial RAM filesystem. Re-run lilo, if you're
using lilo, then shutdown and reboot. If I have the syntax right for
the alias command above, the hotplug system should load the eata module
at the proper time for SCSI adapter device drivers to be loaded, but
only if the card is actually found in the system. If you were to
shutdown, remove the board, and boot again, the eata driver would not
be loaded, since the board was not found. In this sense, it is
automatic. (But only after you define the external alias.)
I'm not sure if the "A" in the device id needs to be upper case or
lower case. Based on the vendor id for the 3w-9xxx driver, which
uses an upper-case "C", I assumed upper case. You might have to
experiment a little to get a correct match.
If you want to file a bug report that the eata driver should define
an internal alias for every board that it supports, you can give it
a try. But I'm not holding my breath. Good luck. ;-)
--
.''`. Stephen Powell <zlinuxman@wowway.com>
: :' :
`. `'`
`-
Reply to: