tomof at acm.org wrote on Sat, 18 Aug 2007 14:18 +0900: > From: Pete Wyckoff <pw at osc.edu> > Subject: Re: [Stgt-devel] [PATCH] save stag & va for a scsi command on the task object > Date: Fri, 17 Aug 2007 17:58:09 -0400 > > > erezz at voltaire.com wrote on Thu, 16 Aug 2007 17:44 +0300: > > > When a command PDU is received, the iSER header contains the read/write > > > va & stag. This data was saved on the conn_info object. If another command PDU > > > was received before this data was used, its va & stag were written over the > > > data for the previous command. This fix saves the va & stag on the task object. > > > > Good stuff. I checked it in with minor modifications to keep trans > > data closer to the rest of the task fields and tweak debugging > > messages a bit: > > > > http://git.osc.edu/?p=tgt.git;a=commitdiff;h=e2be218b262e77776c681602320d598aa854e5e8 > > With this, the iser patchset is ready for inclusion? Sorry, I've not > read it but if it can't handle multiple outsanding commands, surely > it's not ready. > > If it works for all the iSER people (not only OSD people), please > resubmit a new version of the iSER patchset. And please the patch with > the patch style checker (scripts/checkpatch.pl) prior to submission. It's closer now. That was a big problem that Erez fixed. He pointed out off-list that we should run the open-iscsi regression suite against the iser code. That is good advice. Nice addition of checkpatch. I'll see what it turns up. -- Pete |