o When restarting the sequencer, clear any pending

  sequencer interrupt codes.  These codes are only
  relevant to the code that was last being executed
  and that context is cleared when we reset the
  program counter.  This addresses a race condition
  between a sequencer interrupt and any SCSI event
  that causes us to restart the sequencer.

o When running the untagged-Q, we must start the
  timer for any transaction we queue.

o Give the firmware half a millisecond between
  pauses to flush work out.  This should give us
  around half a second of total delay before flagging
  an issue with pausing and flushing controller work.

  Only attempt to clear critical sections if there
  are no pending interrupts in the pause and flush
  loop.  If the sequencer has issued an INTSTAT, we
  may not be able to step out of the critical section.

o Cancel pending transactions on devices that
  respond with a selection timeout.  This decreases
  the duration of timeout recovery when a device
  disappears.

  Don't bother forcing renegotiation on a selection
  timeout now that we use the device reset handler
  to abort any pending commands on the target.
  The device reset handler already takes us down
  to async narrow and forces a renegotiation.

o In the device reset handlers, only send a
  BDR sent async event if the status is not
  CAM_SEL_TIMEOUT.  This avoids sending this
  event in the selection timeout case.

o Modify the Core timeout handler to verify that another
  command has the potential to timeout before passing off
  a command timeout as due to some other command.

1 files changed, 88 insertions, 26 deletions

diff --git a/sys/dev/aic7xxx/aic7xxx.c b/sys/dev/aic7xxx/aic7xxx.c
index ea837a2..ef40938 100644
--- a/sys/dev/aic7xxx/aic7xxx.c
+++ b/sys/dev/aic7xxx/aic7xxx.c
@@ -37,7 +37,7 @@
  * IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
  * POSSIBILITY OF SUCH DAMAGES.
  *
- * $Id: //depot/aic7xxx/aic7xxx/aic7xxx.c#148 $
+ * $Id: //depot/aic7xxx/aic7xxx/aic7xxx.c#155 $
  */
 
 #ifdef __linux__
@@ -231,6 +231,9 @@ static int		ahc_check_patch(struct ahc_softc *ahc,
 					u_int start_instr, u_int *skip_addr);
 static void		ahc_download_instr(struct ahc_softc *ahc,
 					   u_int instrptr, uint8_t *dconsts);
+static int		ahc_other_scb_timeout(struct ahc_softc *ahc,
+					      struct scb *scb,
+					      struct scb *other_scb);
 #ifdef AHC_TARGET_MODE
 static void		ahc_queue_lstate_event(struct ahc_softc *ahc,
 					       struct ahc_tmode_lstate *lstate,
@@ -290,10 +293,19 @@ ahc_restart(struct ahc_softc *ahc)
 		ahc_outb(ahc, SEQ_FLAGS2,
 			 ahc_inb(ahc, SEQ_FLAGS2) & ~SCB_DMA);
 	}
+
+	/*
+	 * Clear any pending sequencer interrupt.  It is no
+	 * longer relevant since we're resetting the Program
+	 * Counter.
+	 */
+	ahc_outb(ahc, CLRINT, CLRSEQINT);
+
 	ahc_outb(ahc, MWI_RESIDUAL, 0);
 	ahc_outb(ahc, SEQCTL, ahc->seqctl);
 	ahc_outb(ahc, SEQADDR0, 0);
 	ahc_outb(ahc, SEQADDR1, 0);
+
 	ahc_unpause(ahc);
 }
 
@@ -365,6 +377,7 @@ ahc_run_untagged_queue(struct ahc_softc *ahc, struct scb_tailq *queue)
 	if ((scb = TAILQ_FIRST(queue)) != NULL
 	 && (scb->flags & SCB_ACTIVE) == 0) {
 		scb->flags |= SCB_ACTIVE;
+		aic_scb_timer_start(scb);
 		ahc_queue_scb(ahc, scb);
 	}
 }
@@ -1177,19 +1190,20 @@ ahc_handle_scsiint(struct ahc_softc *ahc, u_int intstat)
 				       scb_index);
 			}
 #endif
-			/*
-			 * Force a renegotiation with this target just in
-			 * case the cable was pulled and will later be
-			 * re-attached.  The target may forget its negotiation
-			 * settings with us should it attempt to reselect
-			 * during the interruption.  The target will not issue
-			 * a unit attention in this case, so we must always
-			 * renegotiate.
-			 */
 			ahc_scb_devinfo(ahc, &devinfo, scb);
-			ahc_force_renegotiation(ahc, &devinfo);
 			aic_set_transaction_status(scb, CAM_SEL_TIMEOUT);
 			ahc_freeze_devq(ahc, scb);
+
+			/*
+			 * Cancel any pending transactions on the device
+			 * now that it seems to be missing.  This will
+			 * also revert us to async/narrow transfers until
+			 * we can renegotiate with the device.
+			 */
+			ahc_handle_devreset(ahc, &devinfo,
+					    CAM_SEL_TIMEOUT,
+					    "Selection Timeout",
+					    /*verbose_level*/1);
 		}
 		ahc_outb(ahc, CLRINT, CLRSCSIINT);
 		ahc_restart(ahc);
@@ -3766,8 +3780,9 @@ ahc_handle_devreset(struct ahc_softc *ahc, struct ahc_devinfo *devinfo,
 			 /*period*/0, /*offset*/0, /*ppr_options*/0,
 			 AHC_TRANS_CUR, /*paused*/TRUE);
 	
-	ahc_send_async(ahc, devinfo->channel, devinfo->target,
-		       CAM_LUN_WILDCARD, AC_SENT_BDR, NULL);
+	if (status != CAM_SEL_TIMEOUT)
+		ahc_send_async(ahc, devinfo->channel, devinfo->target,
+			       CAM_LUN_WILDCARD, AC_SENT_BDR, NULL);
 
 	if (message != NULL
 	 && (verbose_level <= bootverbose))
@@ -5118,14 +5133,17 @@ ahc_pause_and_flushwork(struct ahc_softc *ahc)
 			 * Give the sequencer some time to service
 			 * any active selections.
 			 */
-			aic_delay(200);
+			aic_delay(500);
 		}
 		ahc_intr(ahc);
 		ahc_pause(ahc);
 		paused = TRUE;
 		ahc_outb(ahc, SCSISEQ, ahc_inb(ahc, SCSISEQ) & ~ENSELO);
-		ahc_clear_critical_section(ahc);
 		intstat = ahc_inb(ahc, INTSTAT);
+		if ((intstat & INT_PEND) == 0) {
+			ahc_clear_critical_section(ahc);
+			intstat = ahc_inb(ahc, INTSTAT);
+		}
 	} while (--maxloops
 	      && (intstat != 0xFF || (ahc->features & AHC_REMOVABLE) == 0)
 	      && ((intstat & INT_PEND) != 0
@@ -6849,6 +6867,58 @@ ahc_timeout(struct scb *scb)
 }
 
 /*
+ * Re-schedule a timeout for the passed in SCB if we determine that some
+ * other SCB is in the process of recovery or an SCB with a longer
+ * timeout is still pending.  Limit our search to just "other_scb"
+ * if it is non-NULL.
+ */
+static int
+ahc_other_scb_timeout(struct ahc_softc *ahc, struct scb *scb,
+		      struct scb *other_scb)
+{
+	u_int	newtimeout;
+	int	found;
+
+	ahc_print_path(ahc, scb);
+	printf("Other SCB Timeout%s",
+ 	       (scb->flags & SCB_OTHERTCL_TIMEOUT) != 0
+	       ? " again\n" : "\n");
+
+	newtimeout = aic_get_timeout(scb);
+	scb->flags |= SCB_OTHERTCL_TIMEOUT;
+	found = 0;
+	if (other_scb != NULL) {
+		if ((other_scb->flags
+		   & (SCB_OTHERTCL_TIMEOUT|SCB_TIMEDOUT)) == 0
+		 || (other_scb->flags & SCB_RECOVERY_SCB) != 0) {
+			found++;
+			newtimeout = MAX(aic_get_timeout(other_scb),
+					 newtimeout);
+		}
+	} else {
+		LIST_FOREACH(other_scb, &ahc->pending_scbs, pending_links) {
+			if ((other_scb->flags
+			   & (SCB_OTHERTCL_TIMEOUT|SCB_TIMEDOUT)) == 0
+			 || (other_scb->flags & SCB_RECOVERY_SCB) != 0) {
+				found++;
+				newtimeout =
+				    MAX(aic_get_timeout(other_scb),
+					newtimeout);
+			}
+		}
+	}
+
+	if (found != 0)
+		aic_scb_timer_reset(scb, newtimeout);
+	else {
+		ahc_print_path(ahc, scb);
+		printf("No other SCB worth waiting for...\n");
+	}
+
+	return (found != 0);
+}
+
+/*
  * ahc_recover_commands determines if any of the commands that have currently
  * timedout are the root cause for this timeout.  Innocent commands are given
  * a new timeout while we wait for the command executing on the bus to timeout.
@@ -6975,17 +7045,9 @@ bus_reset:
 			 */ 
 			active_scb = ahc_lookup_scb(ahc, active_scb_index);
 			if (active_scb != scb) {
-				u_int	newtimeout;
-
-				ahc_print_path(ahc, scb);
-				printf("Other SCB Timeout%s",
-			 	       (scb->flags & SCB_OTHERTCL_TIMEOUT) != 0
-				       ? " again\n" : "\n");
-				scb->flags |= SCB_OTHERTCL_TIMEOUT;
-				newtimeout =
-				    MAX(aic_get_timeout(active_scb),
-					aic_get_timeout(scb));
-				aic_scb_timer_reset(scb, newtimeout);
+				if (ahc_other_scb_timeout(ahc, scb,
+							  active_scb) != 0)
+					goto bus_reset;
 				continue;
 			}