[Open-FCoE] Fwd: problem while using fcgw

Joe Eykholt jre at nuovasystems.com
Sun Sep 21 07:29:31 UTC 2008


manali baldawa wrote:
> Hi Joe,
> 
>     My setup now has a switch between the fcgw and target. I can now see 
> ACC(PLOGI) followed by many PLOGIs(maybe to the initiator). By the way, 
> I didn't use the patch you sent before.
> 
>    I set Atto card to 1Gbps using tyctl since my switch is of 1Gbps.
> 
>     I even tried your patch for rport that I came across when searching 
> for a solution to this problem.
> http://www.mail-archive.com/devel@open-fcoe.org/msg00300.html. But even 
> this didn't help.
> 
>     Attached are the tshark logs. Please have a look at it.
> 
> Thanks & regards
> Manali.


Hi again, Manali.  I'm glad you have a switch now.  It should work
this way.  I think the problem now is that fcgw is using the old
FCoE OUI of fcfcfc.  I can't quite explain why the FLOGI response
comes from the correct OUI.  Maybe the code uses whatever OUI it
got in the FLOGI request.

The PLOGIs you see are from the switch, ff.fc.01 is the switch
for domain 1, which we're connected to.  It does this just to
find out what kind of device has logged in.

Anyway, I think your problem might be solved if you just do
the 'export FCOE_OUI=0x0efc00' before running fcgw.  Let me
know if that works or doesn't.

	Thanks,
	Joe


> On Fri, Sep 19, 2008 at 9:10 PM, Joe Eykholt <jre at nuovasystems.com 
> <mailto:jre at nuovasystems.com>> wrote:
> 
>     I'm cc'ing the list.  Manali and I have been discussing
>     problems he's been having with initiator point-to-point mode.
> 
> 
>     manali baldawa wrote:
> 
>         Thanks Joe.There is some progress from my side but still there
>         are some problems.
> 
>         After applying this patch to initiator code , I don't see that
>         error.
>         I get following messages at initiator side ,
>          *fc_lport_flogi_resp point-to-point mode*
> 
>         It means that initiator understand the FLOGI and accepts it.But
>         I still don't see any luns to initiator side. Tshark messages
>         show that target is responding with /(PLOGI) LS_RJT/ /Command
>         already in progress/.
> 
>         My target is not connected through switch. It's point to point
>         connection.(Qlogic card(2Gbps) to Atto card(4Gbps)).
> 
> 
>     I understood that.  That's a difference in my usage of fcgw.
>     However, now I think it's also the key difference in using
>     point-to-point mode.
> 
> 
>         As I said earlier , I am using SCST as a backend.
>         I get only these messages continuously on target side :
>         *target:# dmesg [0]: q2t_async_event:Async event 0x8030 occured:
>         clear tgt_db *
> 
>         This 0x8030 is macro in qla_mbx.h file
>         *#define MBA_POINT_TO_POINT 0x8030  /* Point to point mode*/*
> 
>         I am attaching the wireshark output also.
>         Do you think , P2P link is broken or open-fcoe initiator doesn't
>         support P2P connection??
> 
> 
>     Maybe it just isn't correct.  So far, I think it's only been
>     tested against the open-fcoe-target implementation, not a real target.
> 
>     Someone said the open-fcoe point-to-point implementation was not
>     as specified by T11.  Correcting it will probably require changes
>     to both the target and initiator sides, but should be done and
>     shouldn't be too difficult..  I think it should be possible to use it
>     the way you are.
> 
>         is switch necessary??
> 
> 
>     Unless the point-to-point operation is fixed, it seems like you
>     need either a switch or to use FCoE all the way to the target
>     instead of using fcgw.
> 
>     I'll look at the trace later today to see if it's something simple.
>     I think the PLOGI-in-progress indication might be a clue.  That is
>     normal and should be handled but might not be.
> 
> 
>         As far as I know , old architecture  open-fcoe initiator and
>         target only used to work in P2P connection.
> 
> 
>     Yes, you're right.
>            
>            Joe
> 
>         Thanks in Advance
>         Manali
> 
> 
> 
> 
> 
> 
>         On Fri, Sep 19, 2008 at 12:25 AM, Joe Eykholt
>         <jre at nuovasystems.com <mailto:jre at nuovasystems.com>
>         <mailto:jre at nuovasystems.com <mailto:jre at nuovasystems.com>>> wrote:
> 
>            Here's what I see:
> 
>            trace-efc0 shows the tachynet driver uses OUI fcfcfc and does
>         receive
>            the FLOGI accept.  The target has assigned FC_ID 0.  fcgw doesn't
>            really care about the OUI mismatch on receive.
> 
>            trace eth0 shows the initator uses OUI 0efc00, and the FLOGI
>         accept
>            does come back from fcgw, so your fcgw setup works.
> 
>            trace tshark.eth0.out shows both sides using 0efc00, and the
>         accept
>            coming back.
> 
>            trace tshark.fc.out shows the same thing with OUI fcfcfc.
> 
>            I think the problem is simply that the initiator expects the
>            FC_ID assigned to be non-zero, which isn't a valid restriction.
> 
>            If you're up to changing the code, try this patch:
> 
>            diff --git a/drivers/scsi/libfc/fc_lport.c
>            b/drivers/scsi/libfc/fc_lport.c
>            index b13f97b..cdabf9a 100644
>            --- a/drivers/scsi/libfc/fc_lport.c
>            +++ b/drivers/scsi/libfc/fc_lport.c
>            @@ -790,7 +790,7 @@ fc_lport_flogi_resp(struct fc_seq *sp, struct
>            fc_frame *fp, void *lp_arg)
> 
>                   fh = fc_frame_header_get(fp);
>                   did = ntoh24(fh->fh_d_id);
>            -       if (fc_frame_payload_op(fp) == ELS_LS_ACC && did != 0) {
>            +       if (fc_frame_payload_op(fp) == ELS_LS_ACC) {
>                           if (fc_lport_debug)
>                                   FC_DBG("assigned fid %x", did);
>                           fc_lport_lock(lp);
> 
>            If that works, let me know and I'll work on getting it committed.
> 
>                   Joe
> 
> 
> 
>            manali baldawa wrote:
>             > I attaching the files once again .
>             >
>             > Thanks in advance
>             > Manali
>             >
>             >
>             > ---------- Forwarded message ----------
>             > From: manali baldawa <manali.baldawa at gmail.com
>         <mailto:manali.baldawa at gmail.com>
>            <mailto:manali.baldawa at gmail.com
>         <mailto:manali.baldawa at gmail.com>>>
>             > Date: Thu, Sep 18, 2008 at 7:59 PM
>             > Subject: Re: [Open-FCoE] problem while using fcgw
>             > To: Joe Eykholt <jre at nuovasystems.com
>         <mailto:jre at nuovasystems.com> <mailto:jre at nuovasystems.com
>         <mailto:jre at nuovasystems.com>>>
>             > Cc: devel at open-fcoe.org <mailto:devel at open-fcoe.org>
>         <mailto:devel at open-fcoe.org <mailto:devel at open-fcoe.org>>
>             >
>             >
>             >
>             > 1. My target has 2gbps line and gateway has 4gbps.
>             >     So earlier due to this speed difference ,  I could not see
>            efcX  online.
>             >     As you said ,  I used tyctl and configured gateway
>         speed as
>            2gbps.
>             >     After that I was able to see efcX interface online.
>             >     So now there is traffic through efcX interface also.
>             >
>             > 2.I got some weird messages at initiator side :
>             >    * fc_lport_flogi_resp bad FLOGI response*
>             > Do you think that new initiator has some problems?? Do I
>         have to
>            use old
>             > initiator??
>             >
>             > 3. I am running fcgw with -n option.
>             >
>             > 4. My target is not a loop device.
>             >
>             > I am attaching wireshark output of efcX and ethX as well
>         as the
>            tcpdump
>             > output of efcx and ethx.
>             > When I captured these logs , i didn't set  FCOE_OUI=0xefc00. I
>            got the same
>             > error with FCOE_OUI set.
>             >
>             > Thanks in advance.
>             > Manali
>             >
>             >
>             > On Thu, Sep 18, 2008 at 5:51 PM, Joe Eykholt
>            <jre at nuovasystems.com <mailto:jre at nuovasystems.com>
>         <mailto:jre at nuovasystems.com <mailto:jre at nuovasystems.com>>> wrote:
>             >
>             >> manali baldawa wrote:
>             >>
>             >>> I tried exporting  this varible export FCOE_OUI=0xefc00 on
>            gatway side and
>             >>> started fcgw again...
>             >>> but i am still facing problem??
>             >>> can anyone help me in this??
>             >>> have anyone used fcgw?? If yes , how did it work ??
>             >>> Please help me out??
>             >>>
>             >> I use a setup like that but with a switch past the gateway.
>             >> There may be some issues in the current initiator with
>             >> point-to-point mode, going directly from fcgw to a target.
>             >>
>             >> If you send me a binary trace, I may be able to help.
>             >> Capture with:  tcpdump -i ethX -w /tmp/trace -s 300
>             >>
>             >> You should be running fcgw with the -n option, although
>             >> not having that wouldn't cause this problem.
>             >>
>             >> Fcgw just passes frames through, it doesn't respond itself.
>             >> With -n it proxies the FLOGI, but still it only responds
>             >> when it hears a response from the other side.
>             >>
>             >> Do you have FC link?  The tachynet driver doesn't
>             >> negotiate speed so if you have a 1G target, you have
>             >> to set the speed using the tyctl utility that comes with
>             >> the driver:  tyctl efc0 1g
>             >>
>             >> Also, if your target is a loop device (NL port), I suspect
>             >> that's not supported by the tachynet driver.
>             >>
>             >>        Joe
>             >>
>             >>
>             >>  Manali
>             >>> On Fri, Sep 12, 2008 at 3:37 PM, manali baldawa
>            <manali.baldawa at gmail.com <mailto:manali.baldawa at gmail.com>
>         <mailto:manali.baldawa at gmail.com <mailto:manali.baldawa at gmail.com>>
> 
>             >>>> wrote:
>             >>>  I want to use fcgw
>             >>>> I have setup *fcoe initiator  -> fcgw -> ( any fc
>         simulator)*
>             >>>> I have all hardware in place ....
>             >>>> i used all configuration steps which are mentioned on
>             >>>> open-fcoe-wiki......
>             >>>> but i am not able to see luns on initiator side.
>             >>>> I think either i missed out something on fcgw...
>             >>>>
>             >>>> i used following steps to set up fcgw.
>             >>>> 1.I have 2gbps atto card
>             >>>> 2. downloaded tachynet driver from sourceforge.net
>         <http://sourceforge.net>
>            <http://sourceforge.net>(link provided on wiki
> 
>             >>>> page)
>             >>>> 3. modprobe tachynet
>             >>>> 4. ifconfig efc01
>             >>>> 5.ifconfig eth1
>             >>>> 6.fcgw -i eth0 -f efc0 -n &
>             >>>>  gateway is running  now
>             >>>>
>             >>>>
>             >>>> Target setup
>             >>>> i have qlogic  and i am using scst .
>             >>>> luns are created on target
>             >>>>
>             >>>> Initiator setup
>             >>>> I have new open-fcoe code
>             >>>> loaded fcoe module
>             >>>>
>             >>>> 1.modprobe fcoe
>             >>>> 2.echo eth1 > /sys/module/fcoe/create
>             >>>>
>             >>>> Could anyone please help me in this??
>             >>>> is my setup is rite??
>             >>>> am i missing any steps ??
>             >>>>
>             >>>> I studied this using wireshark , initiator is sending FLOGI
>            but i think
>             >>>> gatways is not responsding to it and not even forwarding to
>            target.
>             >>>>
>             >>>>
>             >>>> Thanks & Regards
>             >>>> --
>             >>>> Manali
>             >>>>
>             >>>>
>             >>>>
>             >>>
>             >
>             >
> 
> 
> 
> 
>         -- 
>         Manali Baldawa
>         Senior Development Engineer
>         Calsoft Private Ltd
>         Email: manali.baldawa at calsoftinc.com
>         <mailto:manali.baldawa at calsoftinc.com>
>         <mailto:manali.baldawa at calsoftinc.com
>         <mailto:manali.baldawa at calsoftinc.com>>
> 
> 
> 
> 
> 
> -- 
> Manali Baldawa
> Senior Development Engineer
> Calsoft Private Ltd
> Email: manali.baldawa at calsoftinc.com <mailto:manali.baldawa at calsoftinc.com>




More information about the devel mailing list