From: Arne R. <ag...@po...> - 2006-03-30 19:40:32
|
Hi Nicholas, Am Donnerstag, den 30.03.2006, 11:07 -0800 schrieb Nicholas A. Bellinger: > Greetings Arne, > > Sorry for the delay on this, I have been backlogged recently.. > > On Tue, 2006-03-28 at 17:44 +0200, Arne Redlich wrote: > > Am Dienstag, den 28.03.2006, 10:04 -0500 schrieb Ming Zhang: > > > @Arne, could you help me to check this as well. > > > > > > i guess ietd.conf does not contain this parameter, so it is left to > > > default and thus iet does not declare it. but core-ini does not use > > > default value 8192 as the maxrecvdsl for ini, i guess it uses what it > > > declared. > > > > Yep, that's rather likely. At least core-iscsi 1.6.2.2 had this bug: > > http://groups.google.com/group/Core-iSCSI/browse_frm/thread/a0b66ade9417b738/79c039a6a2f7a940#79c039a6a2f7a940 > > > > Skimming through the changelogs of 1.6.2.{3,4}, it doesn't look as if it > > has been fixed. Nicholas? > > > > Correct, I was waiting for a bit of additional feedback when the patch > was originally posted by Tony Battersby a few months back, and Bill > Studenmund recently gave a confirmation on this. I have not applied the > patch in its current form (because it was using CPP defines) but I will > take a look at saving both sides declared values and using these instead > of CONN_OPS(conn)->MaxRecvDataSegmentLength in the transmit cases. Yep, I noticed that the patch wasn't mergable as it was, but during my initial reading of the thread I must have somehow fallen victim to the misconception that you actually agreed it was a bug and were going to fix it - otherwise I would have pinged you earlier. ;-) Re-reading it now, I don't know how I got that impression ... Arne |