diff --git a/drivers/scsi/scsi.c b/drivers/scsi/scsi.c
index d81f3cc43ff1..1423cb17fbfd 100644
--- a/drivers/scsi/scsi.c
+++ b/drivers/scsi/scsi.c
@@ -670,14 +670,10 @@ int scsi_dispatch_cmd(struct scsi_cmnd *cmd)
 		return SCSI_MLQUEUE_DEVICE_BUSY;
 	}
 
-	/*
-	 * If SCSI-2 or lower, store the LUN value in cmnd.
-	 */
-	if (cmd->device->scsi_level <= SCSI_2 &&
-	    cmd->device->scsi_level != SCSI_UNKNOWN) {
+	/* Store the LUN value in cmnd, if needed. */
+	if (cmd->device->lun_in_cdb)
 		cmd->cmnd[1] = (cmd->cmnd[1] & 0x1f) |
 			       (cmd->device->lun << 5 & 0xe0);
-	}
 
 	scsi_log_send(cmd);
 
diff --git a/drivers/scsi/scsi_scan.c b/drivers/scsi/scsi_scan.c
index 56675dbbf681..f84c40188478 100644
--- a/drivers/scsi/scsi_scan.c
+++ b/drivers/scsi/scsi_scan.c
@@ -736,6 +736,16 @@ static int scsi_probe_lun(struct scsi_device *sdev, unsigned char *inq_result,
 		sdev->scsi_level++;
 	sdev->sdev_target->scsi_level = sdev->scsi_level;
 
+	/*
+	 * If SCSI-2 or lower, and if the transport requires it,
+	 * store the LUN value in CDB[1].
+	 */
+	sdev->lun_in_cdb = 0;
+	if (sdev->scsi_level <= SCSI_2 &&
+	    sdev->scsi_level != SCSI_UNKNOWN &&
+	    !sdev->host->no_scsi2_lun_in_cdb)
+		sdev->lun_in_cdb = 1;
+
 	return 0;
 }
 
diff --git a/drivers/scsi/scsi_sysfs.c b/drivers/scsi/scsi_sysfs.c
index 8b4105a22ac2..85e36f3a5585 100644
--- a/drivers/scsi/scsi_sysfs.c
+++ b/drivers/scsi/scsi_sysfs.c
@@ -1263,7 +1263,19 @@ void scsi_sysfs_device_initialize(struct scsi_device *sdev)
 	sdev->sdev_dev.class = &sdev_class;
 	dev_set_name(&sdev->sdev_dev, "%d:%d:%d:%llu",
 		     sdev->host->host_no, sdev->channel, sdev->id, sdev->lun);
+	/*
+	 * Get a default scsi_level from the target (derived from sibling
+	 * devices).  This is the best we can do for guessing how to set
+	 * sdev->lun_in_cdb for the initial INQUIRY command.  For LUN 0 the
+	 * setting doesn't matter, because all the bits are zero anyway.
+	 * But it does matter for higher LUNs.
+	 */
 	sdev->scsi_level = starget->scsi_level;
+	if (sdev->scsi_level <= SCSI_2 &&
+			sdev->scsi_level != SCSI_UNKNOWN &&
+			!shost->no_scsi2_lun_in_cdb)
+		sdev->lun_in_cdb = 1;
+
 	transport_setup_device(&sdev->sdev_gendev);
 	spin_lock_irqsave(shost->host_lock, flags);
 	list_add_tail(&sdev->same_target_siblings, &starget->devices);
diff --git a/drivers/usb/storage/usb.c b/drivers/usb/storage/usb.c
index cedb29252a92..bf3f8e2de046 100644
--- a/drivers/usb/storage/usb.c
+++ b/drivers/usb/storage/usb.c
@@ -983,6 +983,14 @@ int usb_stor_probe2(struct us_data *us)
 	if (!(us->fflags & US_FL_SCM_MULT_TARG))
 		us_to_host(us)->max_id = 1;
 
+	/*
+	 * Like Windows, we won't store the LUN bits in CDB[1] for SCSI-2
+	 * devices using the Bulk-Only transport (even though this violates
+	 * the SCSI spec).
+	 */
+	if (us->transport == usb_stor_Bulk_transport)
+		us_to_host(us)->no_scsi2_lun_in_cdb = 1;
+
 	/* Find the endpoints and calculate pipe values */
 	result = get_pipes(us);
 	if (result)
diff --git a/include/scsi/scsi_device.h b/include/scsi/scsi_device.h
index 1a0d1842962e..27ecee73bd72 100644
--- a/include/scsi/scsi_device.h
+++ b/include/scsi/scsi_device.h
@@ -174,6 +174,7 @@ struct scsi_device {
 	unsigned wce_default_on:1;	/* Cache is ON by default */
 	unsigned no_dif:1;	/* T10 PI (DIF) should be disabled */
 	unsigned broken_fua:1;		/* Don't set FUA bit */
+	unsigned lun_in_cdb:1;		/* Store LUN bits in CDB[1] */
 
 	atomic_t disk_events_disable_depth; /* disable depth for disk events */
 
diff --git a/include/scsi/scsi_host.h b/include/scsi/scsi_host.h
index cafb260ef2d3..d0f69a3210df 100644
--- a/include/scsi/scsi_host.h
+++ b/include/scsi/scsi_host.h
@@ -693,6 +693,9 @@ struct Scsi_Host {
 	 */
 	struct workqueue_struct *tmf_work_q;
 
+	/* The transport requires the LUN bits NOT to be stored in CDB[1] */
+	unsigned no_scsi2_lun_in_cdb:1;
+
 	/*
 	 * Value host_blocked counts down from
 	 */