Bug 67290 - Kudzu/Install don't include i2o_pci
Kudzu/Install don't include i2o_pci
Product: Red Hat Linux
Classification: Retired
Component: kudzu (Show other bugs)
i686 Linux
medium Severity high
: ---
: ---
Assigned To: Alan Cox
Depends On:
  Show dependency treegraph
Reported: 2002-06-21 18:35 EDT by Raj
Modified: 2007-04-18 12:43 EDT (History)
1 user (show)

See Also:
Fixed In Version:
Doc Type: Bug Fix
Doc Text:
Story Points: ---
Clone Of:
Last Closed: 2004-09-28 12:29:16 EDT
Type: ---
Regression: ---
Mount Type: ---
Documentation: ---
Verified Versions:
Category: ---
oVirt Team: ---
RHEL 7.3 requirements from Atomic Host:
Cloudforms Team: ---

Attachments (Terms of Use)

  None (edit)
Description Raj 2002-06-21 18:35:21 EDT
From Bugzilla Helper:
User-Agent: Mozilla/4.0 (compatible; MSIE 6.0; Windows NT 5.0; T312461)

Description of problem:
THe i2o_block module is not working properly. THe i2o block device is not 
accessible. The prestine 2.4.18 kernel works. Loading the i2o_block module 
should load the i2o_pci and i2o_core module but in the 2.4.18-4 & -5 kernels it 
only loads the i2o_core module and does not load the i2o_pci module.

Version-Release number of selected component (if applicable):

How reproducible:

Steps to Reproduce:
1.Create an entry for the device in the modules.conf file and reboot
2.Try to do an fdisk on the block device in /dev/i2o/hda

Actual Results:  THe fdisk returns with cannot open /dev/i2o/hda

Additional info:

The i2o device is a promise sx6000 IDE raid card
Comment 1 Need Real Name 2002-06-28 12:49:15 EDT
I observe the same problem using a Intel Raid Controller SRCU31
Comment 2 Alan Cox 2002-07-17 05:00:34 EDT
This is a bug in the installer/config tools. They should list i2o_pci in
modules.conf to be loaded before i2o_core/i2o_block
Comment 3 Dan Lark 2002-09-08 00:13:59 EDT
Well, I have tried to manually load the modules in the right order, i.e.

1. i2o_core
2. i2o_pci
3. i2o_block

It does discover my card (Promise SX6000), but no RAID volumes attached.

I *REALLY* need to get this going. Is there any chance it will be fixed soon?

BTW, I would settle for Promise's drivers, but the dd disk only goes to RH 7.2 
and I can't get the GPL'd pti_st driver to load.
Comment 4 Alan Cox 2002-09-08 08:53:38 EDT
7.2 has i2o and IDE code that can't handle the newer SX6000 cards. 7.3 you need
to hack up a driver disk because i2o_pci isnt on the disk image. It also won't 
handle the newer SX6000 card (pti_st will also btw clash with the IDE layer on
the newer SX6000 too).

With the current kernels the SX6000 should behave solidly, so hopefully with the
next release we do too.
Comment 5 Dan Lark 2002-09-08 09:55:21 EDT
Will running 2.4.19 (or maybe above) fix the situation? I really need to get 
this card working. I have the system up and running on stock RH 7.3, so I could 
upgrade the kernel to get the SX6000. Am I making a vaild assumption?
Comment 6 Alan Cox 2002-09-08 13:12:55 EDT
It should work reliably on a stock 2.4.19 or on the Red Hat rawhide kernel. If
not I'd really like to know about it.
Comment 7 Dan Lark 2002-09-09 01:20:09 EDT
Well I looked at the rawhide kernel RPM... can you say dependency hell? Does 
the Null beta work with the Promise card w/o a problem? I was, however, able to 
get the system to see the card and work with it somewhat after I up2dated the 
kernel and manually load the i2o modules. But you are 100% correct in the fact 
the IDE layer doesn't play well with the Promise. Oh well...

<rant>I will say, however, that having paid for the distro and to not have it 
work "out of the box" and to be then told "hopefully with the
next release we do (behave solidly) too" is more than a little frustrating. I 
am trying to get a customer of mine off of M$ and to have him see comments like 
this and the hoops that I had to jump through became dicey...</rant>
Comment 8 Alan Cox 2002-09-09 05:38:37 EDT
The problem is that we have to specifically detect the card in the IDE layer
because it appears both as an I2O device and as IDE controllers. The driver we
shipped in 7.3 knows how to do that for the older card but Promise changed the
chips on the board so they have new PCI identifiers. We got exactly zero warning
of that.
Comment 9 Dan Lark 2002-09-09 23:31:19 EDT
Well after much SRPM rebuilding, loaded the Rawhide kernel. Bad news:

Problem 1. The kernel detected the IDE controllers on the card.

Rebooted and did the ide2=0 ide3=0...1e9=0 trick to disable the detction of all 
but motherboard IDE.

Problem 2. i2o_block did not pull in i2o_pci, just pulled in i2o_core and then 

Manually removed the i2o modules. and then loaded by hand i2o_core, i2o_pci and 
i2o_block. I then got:
Problem 3. Got the following error messages.
Sep  9 18:48:18 wefp kernel: I2O: Spurious reply to handler 3
Sep  9 18:48:18 wefp last message repeated 454 times

And the onboard LEDs and warning buzzer started complaining vehemently about a 
bad RAID set (at least acording to the docs). Rebooted and went into the card 
setup. RAID set reported as healthy. rebooted and same result.

I may not be a kernel hacker, but this is beginning to look like a showstopper.
Comment 10 Alan Cox 2002-09-10 09:50:38 EDT
Problem 2 is not a bug. Problem 3 means something still stomped on the IDE Its
ide{n}=off not 0 I believe ?
Comment 11 Dan Lark 2002-09-10 10:03:54 EDT
The ide(n)stuff was a shameless lift from a Promise Fasttrak FAQ. Anyway...

Doe I need to add something like:
        add above i2o_block i2o_pci

to /etc/modules.conf? This is unclear. Any suggestions? I fresh out of them.

Comment 12 Dan Lark 2002-09-10 11:37:54 EDT
BTW, ide2=off, etc. doesn't work.

Also noticed this message:
Sep 10 09:25:12 wefp kernel:  i2o/hda:<6>Device claim called, but dev already
owned by I2O Block OSM!<6>I2O Block: Could not open device

I *REALLY*, *REALLY*, *REALLY* need to get this going.
Comment 13 Dan Lark 2002-09-10 15:22:40 EDT
Stock 2.4.19 exhibits the same behavior for the IDE detection.


Is there a fix available? How am I supposed to get this fscking thing working?
Comment 14 Alan Cox 2002-09-13 11:26:18 EDT
The already owned message is bizarre. I really need to see a full log and also
the lspci -v for that system before I can take more guesses as to what is going on.
Comment 15 Clint Goudie 2004-03-07 20:15:03 EST
FYI: you cannot use the 1.12 promise bios with the i2o drivers. If 
you do i2o wont ever detect the card.
Comment 16 Alan Cox 2004-09-28 12:29:16 EDT
The core i2o stuff is now in rawhide and appears happy.

Note You need to log in before you can comment on or make changes to this bug.