On Tue, 28 Apr 2009, Mike Christie wrote: > Andrew Vasquez wrote: > > On Tue, 28 Apr 2009, Mike Christie wrote: > > > >> Andrew Vasquez wrote: > >>> After an rport's state has transitioned to FC_PORTSTATE_BLOCKED, > >>> but, prior to making the upcall to 'block' the scsi-target > >>> associated with an rport, queued commands can recycle and > >>> ultimately run out of retries causing failures to propagate to > >>> upper-level drivers. Close this transition-window by returning > >>> the non-'retries' modifying DID_IMM_RETRY status for submitted > >>> I/Os. > >>> > >>> Issue seen during continuous LIP-injection. > >>> > >>> Signed-off-by: Andrew Vasquez <andrew.vasquez@xxxxxxxxxx> > >>> > >>> --- > >>> > >>> diff --git a/include/scsi/scsi_transport_fc.h b/include/scsi/scsi_transport_fc.h > >>> index c9184f7..d189e0e 100644 > >>> --- a/include/scsi/scsi_transport_fc.h > >>> +++ b/include/scsi/scsi_transport_fc.h > >>> @@ -687,6 +687,8 @@ fc_remote_port_chkready(struct fc_rport *rport) > >>> case FC_PORTSTATE_BLOCKED: > >>> if (rport->flags & FC_RPORT_FAST_FAIL_TIMEDOUT) > >>> result = DID_TRANSPORT_FAILFAST << 16; > >>> + else if (rport->flags & FC_RPORT_DEVLOSS_PENDING) > >>> + result = DID_IMM_RETRY << 16; > >>> else > >>> result = DID_TRANSPORT_DISRUPTED << 16; > >> I think you can just remove this DID_TRANSPORT_DISRUPTED. The deletion, > >> role change or re-addition code will do the right thing with the IO when > >> it finishes the transition for this case. > > > > Just to be clear here, you're proposing this as an alternate? > > > > -- av > > > > diff --git a/include/scsi/scsi_transport_fc.h b/include/scsi/scsi_transport_fc.h > > index c9184f7..a53a0fd 100644 > > --- a/include/scsi/scsi_transport_fc.h > > +++ b/include/scsi/scsi_transport_fc.h > > @@ -688,7 +688,7 @@ fc_remote_port_chkready(struct fc_rport *rport) > > if (rport->flags & FC_RPORT_FAST_FAIL_TIMEDOUT) > > result = DID_TRANSPORT_FAILFAST << 16; > > else > > - result = DID_TRANSPORT_DISRUPTED << 16; > > + result = DID_IMM_RETRY << 16; > > Yeah, I think that is what we want. We originally had only > DID_IMM_RETRY. When I added DID_TRANSPORT_DISRUPTED, it initially had > infinite retries like DID_IMM_RETRY so the behavior was not changed. > When I fixed DID_TRANSPORT_DISRUPTED to follow the cmd retries/allowed, > I should have changed this code back to use DID_IMM_RETRY. Ok, here's a final one with an updated commit message. --- fc-transport: Close state transition-window during rport deletion. After an rport's state has transitioned to FC_PORTSTATE_BLOCKED, but, prior to making the upcall to 'block' the scsi-target associated with an rport, queued commands can recycle and ultimately run out of retries causing failures to propagate to upper-level drivers. Close this transition-window by returning the non-'retries' modifying DID_IMM_RETRY status for submitted I/Os. Issue seen during continuous LIP-injection. Mike Christie (michaelc@xxxxxxxxxxx) also notes that this is a partial revert of f46e307da925a7b71a0018c0510cdc6e588b87fc ([SCSI] fc class: Add support for new transport errors), as follow-on transport changes now have DID_TRANSPORT_* statuses follow a command's retries/allowed values. Signed-off-by: Andrew Vasquez <andrew.vasquez@xxxxxxxxxx> -- diff --git a/include/scsi/scsi_transport_fc.h b/include/scsi/scsi_transport_fc.h index c9184f7..a53a0fd 100644 --- a/include/scsi/scsi_transport_fc.h +++ b/include/scsi/scsi_transport_fc.h @@ -688,7 +688,7 @@ fc_remote_port_chkready(struct fc_rport *rport) if (rport->flags & FC_RPORT_FAST_FAIL_TIMEDOUT) result = DID_TRANSPORT_FAILFAST << 16; else - result = DID_TRANSPORT_DISRUPTED << 16; + result = DID_IMM_RETRY << 16; break; default: result = DID_NO_CONNECT << 16; -- To unsubscribe from this list: send the line "unsubscribe linux-scsi" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
