Hi all,

This query has been doing my head in for days now. I am a Linux newbie and I've been trying to recognise 4 Luns that have been added to the system.

I am working on RHEL 4 AS update 4
uname -r
2.6.9-42.ELsmp

I am using a dual port qla2342 fibre channel hba I can see them under /proc/scsi/scsi

Attached devices:
Host: scsi0 Channel: 00 Id: 02 Lun: 00
Vendor: IBM Model: 2105800 Rev: 4.45
Type: Unknown ANSI SCSI revision: 03
Host: scsi0 Channel: 00 Id: 03 Lun: 00
Vendor: IBM Model: 2105800 Rev: 4.45
Type: Unknown ANSI SCSI revision: 03
Host: scsi0 Channel: 00 Id: 01 Lun: 00
Vendor: IBM Model: 2105800 Rev: 4.45
Type: Unknown ANSI SCSI revision: 03
Host: scsi0 Channel: 00 Id: 00 Lun: 00
Vendor: IBM Model: 2105800 Rev: 4.45
Type: Unknown ANSI SCSI revision: 03

Why is the TYPE here shown as unknown. I need this to be showing as point-to-point.

Also this output confuses me, is this 4 LUNs or 1 Lun with 4 paths to it? If its the latter then why haven't the other 3 Luns shown up?


/proc/scsi/qla2xxx/0
<snip>
SCSI LUN Information:
(Id:Lun) * - indicates lun is not registered with the OS.
( 0: 0): Total reqs 3, Pending reqs 0, flags 0x0*, 0:0:81 00
( 1: 0): Total reqs 3, Pending reqs 0, flags 0x0*, 0:0:82 00
( 2: 0): Total reqs 3, Pending reqs 0, flags 0x0*, 0:0:83 00
( 3: 0): Total reqs 3, Pending reqs 0, flags 0x0*, 0:0:84 00

This shows the Luns as not recognised by the OS. The * goes away after I echo "- - -" > /sys/class/scsi_host/host0/scan
SCSI LUN Information:
(Id:Lun) * - indicates lun is not registered with the OS.
( 0: 0): Total reqs 4, Pending reqs 0, flags 0x0, 0:0:81 00
( 1: 0): Total reqs 4, Pending reqs 0, flags 0x0, 0:0:82 00
( 2: 0): Total reqs 4, Pending reqs 0, flags 0x0, 0:0:83 00
( 3: 0): Total reqs 4, Pending reqs 0, flags 0x0, 0:0:84 00

but they re-appear after a simple reboot.

fdisk -l / fdisk -l /dev/sd* doesn't appear to recognise the Luns even after the scan (prior to the reboot).

Nothing new is added on /proc/partitions

I'm seriously confused here, I usually work on AIX and have never faced a problem of similarity to this.

Any idea's anyone ?

Try booting with the switch 'max_luns=XX' added to the kernel line, where xx is any number between 1 and 16384 (4 seems good in this case). You can do this as a test by pressing any key when the grub splash screen is displayed. Select the Linux boot choice and press the 'e' key. Select the kernel line and press the 'e' key, at the end of the line add a space followed by the switch ( i.e. max_luns=4). When done press the 'Enter' key followed by the 'b' key.

If this works then add the switch to the kernel line permanently in the /boot/grub/grub.conf file.

Thanks for your advice Lenard.

I've tried something similair to what you suggested.

I added the following into /etc/modprobe.conf

options scsi_mod max_luns=10 (10 for good measure)

then rebuilt the ramdisk, and reboot. No difference.

I am afraid I am not that familiar with grub I just had a look at it and I wasn't sure if I should just insert the line you suggested in there then reboot, can you give me more details on this please?

Also I'm working remotely from the server so I'm guessing rebooting and entering the switches and options does not apply.

Thanks

Working remotely is a bit hard to insert the switch, but the grub.conf file can be edited by any text editor. A sample line with the switch is;

kernel /vmlinuz-2.6.21-rc1-git1 ro root=LABEL=/1 quiet max_luns=10

But you way is just as good, maybe try something like in your /etc/modprobe.conf file;

alias scsi_hostadapter0 qla2xxx
options scsi_mod max_scsi_luns=128

Also try using the qla2300 kernel module instead;

alias scsi_hostadapter0 qla2300
options scsi_mod max_scsi_luns=128

That hasn't worked for me either (modifying the /etc/modprobe.conf) as I've tried that several times already.

I have a feeling that for some reason the connection is funny as it comes up as an unknown type in /proc/scsi/scsi (like my 1st post). Is there a way in which I can edit/modify the connection options for the Qlogic card. What I mean is change it from Loop only to point-to-point for example. At the moment I have no idea what it is, I can't find the command that will show me this.

The reason I'am asking this is I found this extract from a pdf document that lists the steps to install the qlogic cards, halfway through (steps 7 & 8) it gives the communication options. So this got me wondering if thats what I need to do. It might not be but its worth a shot.

extract from pdf doc:

Steps: Perform the following steps to install the QLogic QLA23xx adapter card:

1. Install the QLogic QLA23xx adapter card in the host system.
2. Connect the cable to the ESS port.
3. Restart the server.
4. Press Ctrl+Q to get to the FAST!Util menu.
5. From the Configuration Settings menu, select Host Adapter Settings.
6. Set the following parameters and values:
a. Host Adapter BIOS: Disabled
b. Frame size: 2048
c. Loop reset delay: 5 (minimum)
d. Adapter hard loop ID: Disabled
e. Hard loop ID: 0
*7. Scroll down to Connection Options. Press Enter to open the Option and Type of Connection window.
*8. Select the option:
0 Loop only
1 Point-to-point
2 Loop preferred (If you cannot use arbitrated loop, then default to point-to-point.)

Note: If you connect the ESS directly to the host system, the option that you select must match the port connections on the ESS. 9. From the Advanced Adapter Settings menu, press the Down Arrow to highlight LUNs per target, and then press Enter.
10. Set the following parameters and values:
a. Execution throttle: 100
b. LUNs per target: 0 or 128
c. Enable LIP reset: No
d. Enable LIP full login: Yes
e. Enable target reset: Yes
f. Login retry count: 60Note: Set the login retry count to 8 if you are running the SDD.
g. Port down retry count: 60
Note: Set the port down retry count to 8 if you are running the SDD.
11. Press Esc to return to the Configuration Settings menu.
12. To save the changes, click Yes. Press Enter.
13. Restart the server.

My other option to go to site and try this out myself

You might want to think about upgrading to the newly released kernel during your visit. This update and the earlier kernel updates may have a fix for this issue;

https://rhn.redhat.com/errata/RHSA-2007-0085.html

I did find this which suggests that the qla2342 is supported in RHEL4;
http://www.redhat.com/archives/nahan.../msg00078.html

Is this the pdf doc you referring to?;
http://download.qlogic.com/drivers/5..._Linux_v8x.pdf

You might want visit: http://support.qlogic.com/support/os...id=255&osid=26

I got those steps from http://publib.boulder.ibm.com/infoce...ac_1aizo3.html

It didn't do me any good to be honest.I took a look at qlogic website and decided to download the latest driver 8.01.06 (my current driver is 8.01.04). Needless to say I ran into further problems when trying to install.

There is this script build.sh that builds and installs all the driver modules but it keeps failing as it can't find the required directory it needs. Its looking for this build directory which is in /lib/modules/2.6.9-42.EL/build , the build directory happens to be a soft link to /usr/src/kernels/2.6.9-42.EL-i686. Looking further into the file I can see all the drivers for the scsi on


Since I am using 2.6.9-42.ELsmp I checked the /lib/modules/2.6.9-42.ELsmp and realised there was no build directory.

Now maybe its because I am running out of suggestions but could this be the related or the reason I am having problems recognising my Luns?

Is there anyone out there with a similair problem ?

The kernel source directory /usr/src/kernels/2.6.9-42.EL-i686 is the same for all 32-bit kernels, but may only be the headers (kernel-devel rpm packages installed); rpm -qa 'kernel-devel*'. You might need the kernel-smp-devel rpm.

If you need the full kernel source, following the advice here;

http://wiki.centos.org/I_need_the_Kernel_Source