[Open-FCoE] Setup question

Joe Eykholt jre at nuovasystems.com
Thu Sep 11 22:25:55 UTC 2008


charles zhuang wrote:
> Hi,
> I am new to the community and just built the environment yesterday
> (following the quick start guide on target and initiator). Here's couple
> of questions and issues that I will greatly appreciate if I can get some
> help from you. I have read all the open-fcoe web site and arch mail. 
> 1. Once I start the initiator and the target, 
>>From initiator side I saw dmesg like following:
> Sep 11 08:50:39 linux kernel: SCSI subsystem initialized
> Sep 11 08:51:17 linux kernel: scsi0 : FCoE Driver
>  
>>From target side I saw dmesg like following:
> Sep 11 10:58:43 vic17 kernel: [29440]: scst:
> scst_cmd_thread:3296:Processing thread started, PID 29440
> Sep 11 10:58:43 vic17 kernel: [29441]: scst:
> scst_cmd_thread:3296:Processing thread started, PID 29441
> Sep 11 10:58:43 vic17 kernel: [29443]: scst:
> scst_init_cmd_thread:3076:Init thread started, PID 29443
> Sep 11 10:58:43 vic17 kernel: [29445]: scst:
> scst_mgmt_cmd_thread:4636:Task management thread started, PID 29445
> Sep 11 10:58:43 vic17 kernel: [29432]: scst: init_scst:1784:SCST version
> 1.0.1 loaded successfully (max mem for commands 252MB, per device 100MB)
> Sep 11 10:58:43 vic17 kernel: [29432]: scst:
> scst_print_config:1589:Enabled features: EXTRACHECKS, DEBUG
> Sep 11 10:58:43 vic17 kernel: [29446]: scst:
> scst_mgmt_thread:5127:Management thread started, PID 29446
> Sep 11 10:58:43 vic17 kernel: [29432]: scst:
> __scst_register_target_template:235:Target template openfct registered
> successfully
> Sep 11 10:58:44 vic17 kernel: device eth0 entered promiscuous mode
> Sep 11 10:58:44 vic17 kernel:  mtu 1478
> Sep 11 10:58:44 vic17 kernel: fc_local_port_recv_flogi_req: FLOGI from
> port WWPN 20000002b3d90d26
> Sep 11 10:58:44 vic17 kernel: [29464]:
> scst_suspend_activity:477:suspend_count 0
> Sep 11 10:58:44 vic17 kernel: [29464]: scst_susp_wait:456:wait_event()
> returned 0
> Sep 11 10:58:44 vic17 kernel: [29464]: scst_suspend_activity:516:Waiting
> for 0 active commands finally to complete
> Sep 11 10:58:44 vic17 kernel: [29464]: scst_susp_wait:456:wait_event()
> returned 0
> Sep 11 10:58:44 vic17 kernel: [29464]:
> __scst_resume_activity:546:suspend_count 0 left
> Sep 11 10:58:44 vic17 kernel: [29464]: scst: scst_register:357:Target
> <NULL> (f733f980) for template openfct registered successfully
> Sep 11 10:58:46 vic17 kernel: fc_local_port_recv_flogi_req: FLOGI from
> port WWPN 20000002b3d90d26
> Sep 11 10:59:18 vic17 last message repeated 16 times
>  
>>From the wireshark, I saw constant FLOGI command packet send out from
> initiator to ethernet destination address 0e:fc:00:ff:ff:fe. That's the
> only fcoe packet I saw.

The 0e:fc:00 part of the destination address is called the FCOE_OUI
or FC-MAP value.  0xefc00 is the new, correct default used by T11, but
unfortunately, the target code hasn't gotten that patch yet.

If you'd like to try it, edit fcoe/fc_fcoe.h in the target code
to change the values of 0xfcfcfc on lines 38 to 0x0efc00
and on line 46 change 0xfcfcfcfffffe to 0x0fc0e00fffffe.

I haven't tested this recently, so I don't know whether this is the
only issue or not, but it's the first thing stopping you.
It is possible that the initiator may have changed recently in ways that
prevent point-to-point mode (involved here) from working correctly.

> My initiator ethernet addr is 00:02:B3:D9:0D:26, so from the target log,
> looks like it correctly receive the flogi request but fail to generate
> the response. I assume I will get a trace in initiator like
> "fc_local_port_flogi_resp point-to-point mode ..." if indeed the target
> send out the response. So does that means the problem is on the target
> side?

Yes, the address change above should fix this.

>>From the note in target quick start, "Note that the target doesn't know
> it's a target and attempts FLOGIs. If you're not seeing FLOGIs from the
> target you have a problem." Once I do "fcconf eth0 enable", I should see
> FLOGI packet send out from target even at that time there's no initiator
> exist, correct?

Yes.  Both sides send FLOGIs until they get a response.

> For my target setup, I don't have a scsi disk, only ide disk ("lsscsi"
> come out empty). Does that cause this problem? I saw some post that use
> vdisk, is that workable now? If so, can someone post a detail
> instruction on how to do that? Basically I just try to run from my
> current ide disk instead of adding a new scsi disk to the system.

I tried vdisk a while ago and it worked for me, so you shouldn't need
a SCSI drive.

Here's the bash script I used.  It is set up to create more than one, but
for now just makes one vdisk.  First do a dd if=/dev/zero of=/tmp/vdisk1 count=NBLOCKS
to make a file NBLOCKs long (use a bigger block size for better performance).

# line one would be: #! /bin/bash  # but may trigger malware detectors
modprobe scst
modprobe scst_vdisk

sleep 2  # wait for scst to be ready? ... maybe not needed.

let lun=0

for i in 1
do
        vdisk=vdisk$i
        echo "open $vdisk /tmp/$vdisk" >/proc/scsi_tgt/vdisk/vdisk

        echo "add $vdisk $lun" > /proc/scsi_tgt/groups/Default/devices
        let lun=$lun+1
done


# end

Do that before you start the target fcoe.

> Also, on the initiator, I don't have openfc.ko, I only have fcoe.ko,
> libfc.ko, scsi_transport_fc.ko, and scsi_mod.ko. Did I miss anything?

openfc.ko has been replaced with libfc.ko, so the instructions are
a bit out of date.

> 2. By following the quick start on initiator, my initiator system
> doesn't have fcconf. I guess I can use the same procedure on the target
> quick start guide to build/install fcconf and hbaapi. I don't need
> openfctgt and scst. Is it correct?

In the rearch, fcconf isn't used on the initiator.  Just do

# modprobe fcoe
# echo "ethX" > /sys/module/fcoe/create

to start it.

> Thanks very much for your help.
>  
> charles
>  
>  
> _______________________________________________
> devel mailing list
> devel at open-fcoe.org
> http://www.open-fcoe.org/mailman/listinfo/devel




More information about the devel mailing list