5 files changed, 256 insertions, 30 deletions

diff --git a/sys/dev/mps/mps.c b/sys/dev/mps/mps.c
index c1c1f6c9992d..149e5760a810 100644
--- a/sys/dev/mps/mps.c
+++ b/sys/dev/mps/mps.c
@@ -140,6 +140,7 @@ mps_diag_reset(struct mps_softc *sc,int sleep_flag)
 {
 	uint32_t reg;
 	int i, error, tries = 0;
+	uint8_t first_wait_done = FALSE;
 
 	mps_dprint(sc, MPS_TRACE, "%s\n", __func__);
 
@@ -182,15 +183,32 @@ mps_diag_reset(struct mps_softc *sc,int sleep_flag)
 
 	/* Wait up to 300 seconds in 50ms intervals */
 	error = ETIMEDOUT;
-	for (i = 0; i < 60000; i++) {
-		/* wait 50 msec */
-		if (mtx_owned(&sc->mps_mtx) && sleep_flag == CAN_SLEEP)
-			msleep(&sc->msleep_fake_chan, &sc->mps_mtx, 0,
-			    "mpsdiag", hz/20);
-		else if (sleep_flag == CAN_SLEEP)
-			pause("mpsdiag", hz/20);
-		else
-			DELAY(50 * 1000);
+	for (i = 0; i < 6000; i++) {
+		/*
+		 * Wait 50 msec. If this is the first time through, wait 256
+		 * msec to satisfy Diag Reset timing requirements.
+		 */
+		if (first_wait_done) {
+			if (mtx_owned(&sc->mps_mtx) && sleep_flag == CAN_SLEEP)
+				msleep(&sc->msleep_fake_chan, &sc->mps_mtx, 0,
+				    "mpsdiag", hz/20);
+			else if (sleep_flag == CAN_SLEEP)
+				pause("mpsdiag", hz/20);
+			else
+				DELAY(50 * 1000);
+		} else {
+			DELAY(256 * 1000);
+			first_wait_done = TRUE;
+		}
+		/*
+		 * Check for the RESET_ADAPTER bit to be cleared first, then
+		 * wait for the RESET state to be cleared, which takes a little
+		 * longer.
+		 */
+		reg = mps_regread(sc, MPI2_HOST_DIAGNOSTIC_OFFSET);
+		if (reg & MPI2_DIAG_RESET_ADAPTER) {
+			continue;
+		}
 		reg = mps_regread(sc, MPI2_DOORBELL_OFFSET);
 		if ((reg & MPI2_IOC_STATE_MASK) != MPI2_IOC_STATE_RESET) {
 			error = 0;
@@ -236,7 +254,7 @@ mps_transition_ready(struct mps_softc *sc)
 	sleep_flags = (sc->mps_flags & MPS_FLAGS_ATTACH_DONE)
 					? CAN_SLEEP:NO_SLEEP;
 	error = 0;
-	while (tries++ < 5) {
+	while (tries++ < 1200) {
 		reg = mps_regread(sc, MPI2_DOORBELL_OFFSET);
 		mps_dprint(sc, MPS_INIT, "Doorbell= 0x%x\n", reg);
 
@@ -656,6 +674,9 @@ int
 mps_reinit(struct mps_softc *sc)
 {
 	int error;
+	struct mpssas_softc *sassc;
+
+	sassc = sc->sassc;
 
 	MPS_FUNCTRACE(sc);
 
@@ -736,6 +757,8 @@ mps_reinit(struct mps_softc *sc)
 	mps_dprint(sc, MPS_INFO, "%s finished sc %p post %u free %u\n", 
 	    __func__, sc, sc->replypostindex, sc->replyfreeindex);
 
+	mpssas_release_simq_reinit(sassc);
+
 	return 0;
 }
 
@@ -2510,6 +2533,7 @@ int
 mps_request_polled(struct mps_softc *sc, struct mps_command *cm)
 {
 	int error, timeout = 0, rc;
+	struct timeval cur_time, start_time;
 
 	error = 0;
 
@@ -2517,22 +2541,33 @@ mps_request_polled(struct mps_softc *sc, struct mps_command *cm)
 	cm->cm_complete = NULL;
 	mps_map_command(sc, cm);
 
+	getmicrotime(&start_time);
 	while ((cm->cm_flags & MPS_CM_FLAGS_COMPLETE) == 0) {
 		mps_intr_locked(sc);
 
-		DELAY(50 * 1000);
-		if (timeout++ > 1000) {
+		if (mtx_owned(&sc->mps_mtx))
+			msleep(&sc->msleep_fake_chan, &sc->mps_mtx, 0,
+			    "mpspoll", hz/20);
+		else
+			pause("mpsdiag", hz/20);
+
+		/*
+		 * Check for real-time timeout and fail if more than 60 seconds.
+		 */
+		getmicrotime(&cur_time);
+		timeout = cur_time.tv_sec - start_time.tv_sec;
+		if (timeout > 60) {
 			mps_dprint(sc, MPS_FAULT, "polling failed\n");
 			error = ETIMEDOUT;
 			break;
 		}
 	}
-	
+
 	if (error) {
 		mps_dprint(sc, MPS_FAULT, "Calling Reinit from %s\n", __func__);
 		rc = mps_reinit(sc);
-		mps_dprint(sc, MPS_FAULT, "Reinit %s\n", 
-				(rc == 0) ? "success" : "failed");
+		mps_dprint(sc, MPS_FAULT, "Reinit %s\n", (rc == 0) ? "success" :
+		    "failed");
 	}
 
 	return (error);
diff --git a/sys/dev/mps/mps_sas.c b/sys/dev/mps/mps_sas.c
index 83c5c3b1d004..d2ca8888c56f 100644
--- a/sys/dev/mps/mps_sas.c
+++ b/sys/dev/mps/mps_sas.c
@@ -191,6 +191,16 @@ mpssas_startup_increment(struct mpssas_softc *sassc)
 }
 
 void
+mpssas_release_simq_reinit(struct mpssas_softc *sassc)
+{
+	if (sassc->flags & MPSSAS_QUEUE_FROZEN) {
+		sassc->flags &= ~MPSSAS_QUEUE_FROZEN;
+		xpt_release_simq(sassc->sim, 1);
+		mps_dprint(sassc->sc, MPS_INFO, "Unfreezing SIM queue\n");
+	}
+}
+
+void
 mpssas_startup_decrement(struct mpssas_softc *sassc)
 {
 	MPS_FUNCTRACE(sassc->sc);
@@ -1003,7 +1013,7 @@ mpssas_action(struct cam_sim *sim, union ccb *ccb)
 		    cts->ccb_h.target_id));
 		targ = &sassc->targets[cts->ccb_h.target_id];
 		if (targ->handle == 0x0) {
-			mpssas_set_ccbstatus(ccb, CAM_SEL_TIMEOUT);
+			mpssas_set_ccbstatus(ccb, CAM_DEV_NOT_THERE);
 			break;
 		}
 
@@ -1120,6 +1130,14 @@ mpssas_complete_all_commands(struct mps_softc *sc)
 			wakeup(cm);
 			completed = 1;
 		}
+
+		if (cm->cm_sc->io_cmds_active != 0) {
+			cm->cm_sc->io_cmds_active--;
+		} else {
+			mps_dprint(cm->cm_sc, MPS_INFO, "Warning: "
+			    "io_cmds_active is out of sync - resynching to "
+			    "0\n");
+		}
 		
 		if ((completed == 0) && (cm->cm_state != MPS_CM_STATE_FREE)) {
 			/* this should never happen, but if it does, log */
@@ -1647,14 +1665,14 @@ mpssas_action_scsiio(struct mpssas_softc *sassc, union ccb *ccb)
 	if (targ->handle == 0x0) {
 		mps_dprint(sc, MPS_ERROR, "%s NULL handle for target %u\n", 
 		    __func__, csio->ccb_h.target_id);
-		mpssas_set_ccbstatus(ccb, CAM_SEL_TIMEOUT);
+		mpssas_set_ccbstatus(ccb, CAM_DEV_NOT_THERE);
 		xpt_done(ccb);
 		return;
 	}
 	if (targ->flags & MPS_TARGET_FLAGS_RAID_COMPONENT) {
 		mps_dprint(sc, MPS_ERROR, "%s Raid component no SCSI IO "
 		    "supported %u\n", __func__, csio->ccb_h.target_id);
-		mpssas_set_ccbstatus(ccb, CAM_TID_INVALID);
+		mpssas_set_ccbstatus(ccb, CAM_DEV_NOT_THERE);
 		xpt_done(ccb);
 		return;
 	}
@@ -1685,13 +1703,16 @@ mpssas_action_scsiio(struct mpssas_softc *sassc, union ccb *ccb)
 
 	if ((sc->mps_flags & MPS_FLAGS_SHUTDOWN) != 0) {
 		mps_dprint(sc, MPS_INFO, "%s shutting down\n", __func__);
-		mpssas_set_ccbstatus(ccb, CAM_TID_INVALID);
+		mpssas_set_ccbstatus(ccb, CAM_DEV_NOT_THERE);
 		xpt_done(ccb);
 		return;
 	}
 
 	cm = mps_alloc_command(sc);
-	if (cm == NULL) {
+	if (cm == NULL || (sc->mps_flags & MPS_FLAGS_DIAGRESET)) {
+		if (cm != NULL) {
+			mps_free_command(sc, cm);
+		}
 		if ((sassc->flags & MPSSAS_QUEUE_FROZEN) == 0) {
 			xpt_freeze_simq(sassc->sim, 1);
 			sassc->flags |= MPSSAS_QUEUE_FROZEN;
@@ -2170,6 +2191,18 @@ mpssas_scsiio_complete(struct mps_softc *sc, struct mps_command *cm)
 		}
 	}
 
+	/*
+	 * If this is a Start Stop Unit command and it was issued by the driver
+	 * during shutdown, decrement the refcount to account for all of the
+	 * commands that were sent.  All SSU commands should be completed before
+	 * shutdown completes, meaning SSU_refcount will be 0 after SSU_started
+	 * is TRUE.
+	 */
+	if (sc->SSU_started && (csio->cdb_io.cdb_bytes[0] == START_STOP_UNIT)) {
+		mps_dprint(sc, MPS_INFO, "Decrementing SSU count.\n");
+		sc->SSU_refcount--;
+	}
+
 	/* Take the fast path to completion */
 	if (cm->cm_reply == NULL) {
 		if (mpssas_get_ccbstatus(ccb) == CAM_REQ_INPROG) {
@@ -2999,7 +3032,7 @@ mpssas_action_smpio(struct mpssas_softc *sassc, union ccb *ccb)
 			mps_dprint(sc, MPS_ERROR,
 				   "%s: handle %d does not have a valid "
 				   "parent handle!\n", __func__, targ->handle);
-			mpssas_set_ccbstatus(ccb, CAM_REQ_INVALID);
+			mpssas_set_ccbstatus(ccb, CAM_DEV_NOT_THERE);
 			goto bailout;
 		}
 #ifdef OLD_MPS_PROBE
@@ -3010,7 +3043,7 @@ mpssas_action_smpio(struct mpssas_softc *sassc, union ccb *ccb)
 			mps_dprint(sc, MPS_ERROR,
 				   "%s: handle %d does not have a valid "
 				   "parent target!\n", __func__, targ->handle);
-			mpssas_set_ccbstatus(ccb, CAM_REQ_INVALID);
+			mpssas_set_ccbstatus(ccb, CAM_DEV_NOT_THERE);
 			goto bailout;
 		}
 
@@ -3020,7 +3053,7 @@ mpssas_action_smpio(struct mpssas_softc *sassc, union ccb *ccb)
 				   "%s: handle %d parent %d does not "
 				   "have an SMP target!\n", __func__,
 				   targ->handle, parent_target->handle);
-			mpssas_set_ccbstatus(ccb, CAM_REQ_INVALID);
+			mpssas_set_ccbstatus(ccb, CAM_DEV_NOT_THERE);
 			goto bailout;
 
 		}
@@ -3033,7 +3066,7 @@ mpssas_action_smpio(struct mpssas_softc *sassc, union ccb *ccb)
 				   "%s: handle %d parent %d does not "
 				   "have an SMP target!\n", __func__,
 				   targ->handle, targ->parent_handle);
-			mpssas_set_ccbstatus(ccb, CAM_REQ_INVALID);
+			mpssas_set_ccbstatus(ccb, CAM_DEV_NOT_THERE);
 			goto bailout;
 
 		}
@@ -3042,7 +3075,7 @@ mpssas_action_smpio(struct mpssas_softc *sassc, union ccb *ccb)
 				   "%s: handle %d parent handle %d does "
 				   "not have a valid SAS address!\n",
 				   __func__, targ->handle, targ->parent_handle);
-			mpssas_set_ccbstatus(ccb, CAM_REQ_INVALID);
+			mpssas_set_ccbstatus(ccb, CAM_DEV_NOT_THERE);
 			goto bailout;
 		}
 
@@ -3055,7 +3088,7 @@ mpssas_action_smpio(struct mpssas_softc *sassc, union ccb *ccb)
 		mps_dprint(sc, MPS_INFO,
 			   "%s: unable to find SAS address for handle %d\n",
 			   __func__, targ->handle);
-		mpssas_set_ccbstatus(ccb, CAM_REQ_INVALID);
+		mpssas_set_ccbstatus(ccb, CAM_DEV_NOT_THERE);
 		goto bailout;
 	}
 	mpssas_send_smpcmd(sassc, ccb, sasaddr);
@@ -3366,6 +3399,20 @@ mpssas_check_eedp(struct mps_softc *sc, struct cam_path *path,
 	}
 
 	xpt_path_string(local_path, path_str, sizeof(path_str));
+
+	/*
+	 * If this is a SATA direct-access end device,
+	 * mark it so that a SCSI StartStopUnit command
+	 * will be sent to it when the driver is being
+	 * shutdown.
+	 */
+	if ((cgd.inq_data.device == T_DIRECT) && 
+		(target->devinfo & MPI2_SAS_DEVICE_INFO_SATA_DEVICE) &&
+		((target->devinfo & MPI2_SAS_DEVICE_INFO_MASK_DEVICE_TYPE) ==
+		MPI2_SAS_DEVICE_INFO_END_DEVICE)) {
+		lun->stop_at_shutdown = TRUE;
+	}
+
 	mps_dprint(sc, MPS_INFO, "Sending read cap: path %s handle %d\n",
 	    path_str, target->handle);
 
diff --git a/sys/dev/mps/mps_sas.h b/sys/dev/mps/mps_sas.h
index dac49058d8f0..0a3ddf158382 100644
--- a/sys/dev/mps/mps_sas.h
+++ b/sys/dev/mps/mps_sas.h
@@ -35,6 +35,7 @@ struct mpssas_lun {
 	lun_id_t	lun_id;
 	uint8_t		eedp_formatted;
 	uint32_t	eedp_block_size;
+	uint8_t		stop_at_shutdown;
 };
 
 struct mpssas_target {
diff --git a/sys/dev/mps/mps_sas_lsi.c b/sys/dev/mps/mps_sas_lsi.c
index c219be14bb13..583632947e34 100644
--- a/sys/dev/mps/mps_sas_lsi.c
+++ b/sys/dev/mps/mps_sas_lsi.c
@@ -120,6 +120,9 @@ int mpssas_get_sas_address_for_sata_disk(struct mps_softc *sc,
     u64 *sas_address, u16 handle, u32 device_info);
 static int mpssas_volume_add(struct mps_softc *sc,
     u16 handle);
+static void mpssas_SSU_to_SATA_devices(struct mps_softc *sc);
+static void mpssas_stop_unit_done(struct cam_periph *periph,
+    union ccb *done_ccb);
 
 void
 mpssas_evt_handler(struct mps_softc *sc, uintptr_t data,
@@ -910,6 +913,138 @@ out:
 }
 
 /**
+ * mpssas_SSU_to_SATA_devices 
+ * @sc: per adapter object
+ *
+ * Looks through the target list and issues a StartStopUnit SCSI command to each
+ * SATA direct-access device.  This helps to ensure that data corruption is
+ * avoided when the system is being shut down.  This must be called after the IR
+ * System Shutdown RAID Action is sent if in IR mode.
+ *
+ * Return nothing.
+ */
+static void
+mpssas_SSU_to_SATA_devices(struct mps_softc *sc)
+{
+	struct mpssas_softc *sassc = sc->sassc;
+	union ccb *ccb;
+	path_id_t pathid = cam_sim_path(sassc->sim);
+	target_id_t targetid;
+	struct mpssas_target *target;
+	struct mpssas_lun *lun;
+	char path_str[64];
+	struct timeval cur_time, start_time;
+
+	/*
+	 * For each LUN of each target, issue a StartStopUnit command to stop
+	 * the device.
+	 */
+	sc->SSU_started = TRUE;
+	sc->SSU_refcount = 0;
+	for (targetid = 0; targetid < sc->facts->MaxTargets; targetid++) {
+		target = &sassc->targets[targetid];
+		if (target->handle == 0x0) {
+			continue;
+		}
+
+		SLIST_FOREACH(lun, &target->luns, lun_link) {
+			ccb = xpt_alloc_ccb_nowait();
+			if (ccb == NULL) {
+				mps_dprint(sc, MPS_FAULT, "Unable to alloc CCB "
+				    "to stop unit.\n");
+				return;
+			}
+
+			/*
+			 * The stop_at_shutdown flag will be set if this LUN is
+			 * a SATA direct-access end device.
+			 */
+			if (lun->stop_at_shutdown) {
+				if (xpt_create_path(&ccb->ccb_h.path,
+				    xpt_periph, pathid, targetid,
+				    lun->lun_id) != CAM_REQ_CMP) {
+					mps_dprint(sc, MPS_FAULT, "Unable to "
+					    "create LUN path to stop unit.\n");
+					xpt_free_ccb(ccb);
+					return;
+				}
+				xpt_path_string(ccb->ccb_h.path, path_str,
+				    sizeof(path_str));
+
+				mps_dprint(sc, MPS_INFO, "Sending StopUnit: "
+				    "path %s handle %d\n", path_str,
+				    target->handle);
+			
+				/*
+				 * Issue a START STOP UNIT command for the LUN.
+				 * Increment the SSU counter to be used to
+				 * count the number of required replies.
+				 */
+				mps_dprint(sc, MPS_INFO, "Incrementing SSU "
+				    "count\n");
+				sc->SSU_refcount++;
+				ccb->ccb_h.target_id =
+				    xpt_path_target_id(ccb->ccb_h.path);
+				ccb->ccb_h.target_lun = lun->lun_id;
+				ccb->ccb_h.ppriv_ptr1 = sassc;
+				scsi_start_stop(&ccb->csio,
+				    /*retries*/0,
+				    mpssas_stop_unit_done,
+				    MSG_SIMPLE_Q_TAG,
+				    /*start*/FALSE,
+				    /*load/eject*/0,
+				    /*immediate*/FALSE,
+				    MPS_SENSE_LEN,
+				    /*timeout*/10000);
+				xpt_action(ccb);
+			}
+		}
+	}
+
+	/*
+	 * Wait until all of the SSU commands have completed or time has
+	 * expired (60 seconds).  pause for 100ms each time through.  If any
+	 * command times out, the target will be reset in the SCSI command
+	 * timeout routine.
+	 */
+	getmicrotime(&start_time);
+	while (sc->SSU_refcount) {
+		pause("mpswait", hz/10);
+		
+		getmicrotime(&cur_time);
+		if ((cur_time.tv_sec - start_time.tv_sec) > 60) {
+			mps_dprint(sc, MPS_FAULT, "Time has expired waiting "
+			    "for SSU commands to complete.\n");
+			break;
+		}
+	}
+}
+
+static void
+mpssas_stop_unit_done(struct cam_periph *periph, union ccb *done_ccb)
+{
+	struct mpssas_softc *sassc;
+	char path_str[64];
+
+	sassc = (struct mpssas_softc *)done_ccb->ccb_h.ppriv_ptr1;
+
+	xpt_path_string(done_ccb->ccb_h.path, path_str, sizeof(path_str));
+	mps_dprint(sassc->sc, MPS_INFO, "Completing stop unit for %s\n",
+	    path_str);
+
+	if (done_ccb == NULL)
+		return;
+
+	/*
+	 * Nothing more to do except free the CCB and path.  If the command
+	 * timed out, an abort reset, then target reset will be issued during
+	 * the SCSI Command process.
+	 */
+	xpt_free_path(done_ccb->ccb_h.path);
+	xpt_free_ccb(done_ccb);
+}
+
+/**
  * mpssas_ir_shutdown - IR shutdown notification
  * @sc: per adapter object
  *
@@ -933,7 +1068,7 @@ mpssas_ir_shutdown(struct mps_softc *sc)
 
 	/* is IR firmware build loaded? */
 	if (!sc->ir_firmware)
-		return;
+		goto out;
 
 	/* are there any volumes?  Look at IR target IDs. */
 	// TODO-later, this should be looked up in the RAID config structure
@@ -958,11 +1093,11 @@ mpssas_ir_shutdown(struct mps_softc *sc)
 	}
 
 	if (!found_volume)
-		return;
+		goto out;
 
 	if ((cm = mps_alloc_command(sc)) == NULL) {
 		printf("%s: command alloc failed\n", __func__);
-		return;
+		goto out;
 	}
 
 	action = (MPI2_RAID_ACTION_REQUEST *)cm->cm_req;
@@ -978,4 +1113,7 @@ mpssas_ir_shutdown(struct mps_softc *sc)
 	 */
 	if (cm)
 		mps_free_command(sc, cm);
+
+out:
+	mpssas_SSU_to_SATA_devices(sc);
 }
diff --git a/sys/dev/mps/mpsvar.h b/sys/dev/mps/mpsvar.h
index 50e93f00ae64..dead1dbd840d 100644
--- a/sys/dev/mps/mpsvar.h
+++ b/sys/dev/mps/mpsvar.h
@@ -32,7 +32,7 @@
 #ifndef _MPSVAR_H
 #define _MPSVAR_H
 
-#define MPS_DRIVER_VERSION	"16.00.00.00-fbsd"
+#define MPS_DRIVER_VERSION	"18.00.00.00-fbsd"
 
 #define MPS_DB_MAX_WAIT		2500
 
@@ -417,6 +417,10 @@ struct mps_softc {
 
 	char				exclude_ids[80];
 	struct timeval			lastfail;
+
+	/* StartStopUnit command handling at shutdown */
+	uint32_t			SSU_refcount;
+	uint8_t				SSU_started;
 };
 
 struct mps_config_params {
@@ -759,6 +763,7 @@ struct mpssas_target * mpssas_find_target_by_handle(struct mpssas_softc *, int,
 void mpssas_realloc_targets(struct mps_softc *sc, int maxtargets);
 struct mps_command * mpssas_alloc_tm(struct mps_softc *sc);
 void mpssas_free_tm(struct mps_softc *sc, struct mps_command *tm);
+void mpssas_release_simq_reinit(struct mpssas_softc *sassc);
 
 SYSCTL_DECL(_hw_mps);