>-----Original Message----- >From: Mike Christie [mailto:[EMAIL PROTECTED] >Sent: Tuesday, August 12, 2008 2:22 PM >To: Dev, Vasu; [email protected] >Subject: Re: [Open-FCoE] [RFC PATCH 1/2] libfc: removed checks for >timer_pending and then updating exch ref count. > >Dev, Vasu wrote: >>>> >>>> Also for the fc_fcp_timeout->fc_timeout_error path, if the seq_ptr is >>>> set we never call exch_done to clean up the response handler. >> >> Abort handling in fc_exch.c will complete the exchange in this case but > >Oh yeah, Vasu what are you talking about here? I was not worried about >the exhange getting cleaned up.
You are right, please ignore above note was redundant in this case. > I was worried about the ep->resp calling >into the fc_fcp.c with a freed fsp. There was nothing in the existing >code to prevent that right? Correct, I was also talking about for both issues of dangling ep->resp and double exch_done() in fc_fcp.c abort handling. So kind of we were level set on issues but I think I got lost on multiple ep->resp functions but now I understand that you were talking about calling same ep->resp either for abort or normal scsi IO and I was also suggesting same. I was also looking for that so called hack (ref check in release) beyond abort paths for any junk frame in RX for the exch which is about to free, though that is very unlikely once abort or scsi IO is completed. I think call back from fc_exch.c to free fsp can be helpful here in some cases or for at least junk frame case. So at what stage you are in with these changes? _______________________________________________ devel mailing list [email protected] http://www.open-fcoe.org/mailman/listinfo/devel
