Re: [Iscsitarget-devel] iet 0.4.7+linux-iscsi 5.0.0.3rc1-255 errors

[FUJITA T. <to...@ac...>]

From: FUJITA Tomonori <fuj...@la...>
Subject: Re: [Iscsitarget-devel] iet 0.4.7+linux-iscsi 5.0.0.3rc1-255 errors
Date: Wed, 27 Apr 2005 01:19:12 +0900

> > > I confirm that I'm seeing similar prints too. I was digging into this
> > > problem a few months ago and didn't find anything strange on initiator
> > > side. In my case, target was checking ExpStatSN from Initiator, which is
> > > wrong. imho. ExpStatSN is used as an acknowledgments for target
> > > sequencing. Initiator can not promise any ordering in delivery of iSCSI
> > > PDU's since TCP/IP can deliver frames out-of-order.
> > 
> > didn't explain myself right. ignore statement above.
> > exp_statsn is just used for ack's. and both receiver and transmitter
> > involved on. so, out-of-order on *iSCSI* level could happen easily.
> 
> I'm not sure about that. Why could out-of-order stat_sn (stat_sn !=
> exp_stat_sn) happen easily? I think that it should be rare. I have not
> seen this before with sfnet.

Sorry. I was confused about CmdSN and StatSN. The current IET method
to check ExpStatSN is wrong. Scott explained the reason (Thanks,
Scott). I fixed it and uploaded the latest (r1044).

There are no commands with out-of-order ExpStatSN in Harald's log. As
Scott pointed out, the log says that the initiator does not send the
latest ExpStatSN.