[SCSI] qla4xxx: bug fixes

The included patch fixes the following issues:

1. qla3xxx/qla4xxx co-existence issue which can result in a lockup
when qla3xxx driver is unloaded, or when ifdown; ifup is performed on
one of the interfaces correponding to qla3xxx. This is because qla4xxx
HBA supports one ethernet and iscsi interfaces per port. Both iscsi
and ethernet interfaces share the same state machine. The problem has
to do with synchronizing access to the state machine in the event of a
reset

2. mutex_lock() is sometimes not followed by mutex_unlock() prior to
invoking a msleep() in qla4xxx_mailbox_command()
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>

drivers/scsi/qla4xxx/ql4_def.h

@@ -418,7 +418,6 @@ struct scsi_qla_host {
 	 * concurrently.
 	 */
 	struct mutex  mbox_sem;
-	wait_queue_head_t mailbox_wait_queue;
 
 	/* temporary mailbox status registers */
 	volatile uint8_t mbox_status_count;

drivers/scsi/qla4xxx/ql4_glbl.h

@@ -76,4 +76,5 @@ int qla4xxx_process_ddb_changed(struct scsi_qla_host * ha,
 extern int ql4xextended_error_logging;
 extern int ql4xdiscoverywait;
 extern int ql4xdontresethba;
+extern int ql4_mod_unload;
 #endif /* _QLA4x_GBL_H */

drivers/scsi/qla4xxx/ql4_init.c

@@ -958,25 +958,25 @@ static int qla4xxx_start_firmware_from_flash(struct scsi_qla_host *ha)
 	return status;
 }
 
-int ql4xxx_lock_drvr_wait(struct scsi_qla_host *a)
+int ql4xxx_lock_drvr_wait(struct scsi_qla_host *ha)
 {
-#define QL4_LOCK_DRVR_WAIT	300
-#define QL4_LOCK_DRVR_SLEEP	100
+#define QL4_LOCK_DRVR_WAIT	30
+#define QL4_LOCK_DRVR_SLEEP	1
 
 	int drvr_wait = QL4_LOCK_DRVR_WAIT;
 	while (drvr_wait) {
-		if (ql4xxx_lock_drvr(a) == 0) {
-			msleep(QL4_LOCK_DRVR_SLEEP);
+		if (ql4xxx_lock_drvr(ha) == 0) {
+			ssleep(QL4_LOCK_DRVR_SLEEP);
 			if (drvr_wait) {
 				DEBUG2(printk("scsi%ld: %s: Waiting for "
-					      "Global Init Semaphore...n",
-					      a->host_no,
-					      __func__));
+					      "Global Init Semaphore(%d)...n",
+					      ha->host_no,
+					      __func__, drvr_wait));
 			}
 			drvr_wait -= QL4_LOCK_DRVR_SLEEP;
 		} else {
 			DEBUG2(printk("scsi%ld: %s: Global Init Semaphore "
-				      "acquired.n", a->host_no, __func__));
+				      "acquired.n", ha->host_no, __func__));
 			return QLA_SUCCESS;
 		}
 	}

drivers/scsi/qla4xxx/ql4_isr.c

@@ -433,7 +433,6 @@ static void qla4xxx_isr_decode_mailbox(struct scsi_qla_host * ha,
 					readl(&ha->reg->mailbox[i]);
 
 			set_bit(AF_MBOX_COMMAND_DONE, &ha->flags);
-			wake_up(&ha->mailbox_wait_queue);
 		}
 	} else if (mbox_status >> 12 == MBOX_ASYNC_EVENT_STATUS) {
 		/* Immediately process the AENs that don't require much work.
@@ -686,6 +685,7 @@ irqreturn_t qla4xxx_intr_handler(int irq, void *dev_id)
 			       &ha->reg->ctrl_status);
 			readl(&ha->reg->ctrl_status);
 
+			if (!ql4_mod_unload)
 				set_bit(DPC_RESET_HA_INTR, &ha->dpc_flags);
 
 			break;

drivers/scsi/qla4xxx/ql4_mbx.c

@@ -29,18 +29,30 @@ int qla4xxx_mailbox_command(struct scsi_qla_host *ha, uint8_t inCount,
 	u_long wait_count;
 	uint32_t intr_status;
 	unsigned long flags = 0;
-	DECLARE_WAITQUEUE(wait, current);
-
-	mutex_lock(&ha->mbox_sem);
-
-	/* Mailbox code active */
-	set_bit(AF_MBOX_COMMAND, &ha->flags);
 
 	/* Make sure that pointers are valid */
 	if (!mbx_cmd || !mbx_sts) {
 		DEBUG2(printk("scsi%ld: %s: Invalid mbx_cmd or mbx_sts "
 			      "pointer\n", ha->host_no, __func__));
-		goto mbox_exit;
+		return status;
+	}
+	/* Mailbox code active */
+	wait_count = MBOX_TOV * 100;
+
+	while (wait_count--) {
+		mutex_lock(&ha->mbox_sem);
+		if (!test_bit(AF_MBOX_COMMAND, &ha->flags)) {
+			set_bit(AF_MBOX_COMMAND, &ha->flags);
+			mutex_unlock(&ha->mbox_sem);
+			break;
+		}
+		mutex_unlock(&ha->mbox_sem);
+		if (!wait_count) {
+			DEBUG2(printk("scsi%ld: %s: mbox_sem failed\n",
+				ha->host_no, __func__));
+			return status;
+		}
+		msleep(10);
 	}
 
 	/* To prevent overwriting mailbox registers for a command that has
@@ -73,8 +85,6 @@ int qla4xxx_mailbox_command(struct scsi_qla_host *ha, uint8_t inCount,
 	spin_unlock_irqrestore(&ha->hardware_lock, flags);
 
 	/* Wait for completion */
-	set_current_state(TASK_UNINTERRUPTIBLE);
-	add_wait_queue(&ha->mailbox_wait_queue, &wait);
 
 	/*
 	 * If we don't want status, don't wait for the mailbox command to
@@ -83,8 +93,6 @@ int qla4xxx_mailbox_command(struct scsi_qla_host *ha, uint8_t inCount,
 	 */
 	if (outCount == 0) {
 		status = QLA_SUCCESS;
-		set_current_state(TASK_RUNNING);
-		remove_wait_queue(&ha->mailbox_wait_queue, &wait);
 		goto mbox_exit;
 	}
 	/* Wait for command to complete */
@@ -108,8 +116,6 @@ int qla4xxx_mailbox_command(struct scsi_qla_host *ha, uint8_t inCount,
 		spin_unlock_irqrestore(&ha->hardware_lock, flags);
 		msleep(10);
 	}
-	set_current_state(TASK_RUNNING);
-	remove_wait_queue(&ha->mailbox_wait_queue, &wait);
 
 	/* Check for mailbox timeout. */
 	if (!test_bit(AF_MBOX_COMMAND_DONE, &ha->flags)) {
@@ -155,9 +161,10 @@ int qla4xxx_mailbox_command(struct scsi_qla_host *ha, uint8_t inCount,
 	spin_unlock_irqrestore(&ha->hardware_lock, flags);
 
 mbox_exit:
+	mutex_lock(&ha->mbox_sem);
 	clear_bit(AF_MBOX_COMMAND, &ha->flags);
-	clear_bit(AF_MBOX_COMMAND_DONE, &ha->flags);
 	mutex_unlock(&ha->mbox_sem);
+	clear_bit(AF_MBOX_COMMAND_DONE, &ha->flags);
 
 	return status;
 }

drivers/scsi/qla4xxx/ql4_os.c

@@ -40,6 +40,8 @@ MODULE_PARM_DESC(ql4xextended_error_logging,
 		 "Option to enable extended error logging, "
 		 "Default is 0 - no logging, 1 - debug logging");
 
+int ql4_mod_unload = 0;
+
 /*
  * SCSI host template entry points
  */
@@ -422,6 +424,9 @@ static int qla4xxx_queuecommand(struct scsi_cmnd *cmd,
 		goto qc_host_busy;
 	}
 
+	if (test_bit(DPC_RESET_HA_INTR, &ha->dpc_flags))
+		goto qc_host_busy;
+
 	spin_unlock_irq(ha->host->host_lock);
 
 	srb = qla4xxx_get_new_srb(ha, ddb_entry, cmd, done);
@@ -707,16 +712,12 @@ static int qla4xxx_cmd_wait(struct scsi_qla_host *ha)
 	return stat;
 }
 
-/**
- * qla4xxx_soft_reset - performs soft reset.
- * @ha: Pointer to host adapter structure.
- **/
-int qla4xxx_soft_reset(struct scsi_qla_host *ha)
+static void qla4xxx_hw_reset(struct scsi_qla_host *ha)
 {
-	uint32_t max_wait_time;
-	unsigned long flags = 0;
-	int status = QLA_ERROR;
 	uint32_t ctrl_status;
+	unsigned long flags = 0;
+
+	DEBUG2(printk(KERN_ERR "scsi%ld: %s\n", ha->host_no, __func__));
 
 	spin_lock_irqsave(&ha->hardware_lock, flags);
 
@@ -733,6 +734,20 @@ int qla4xxx_soft_reset(struct scsi_qla_host *ha)
 	readl(&ha->reg->ctrl_status);
 
 	spin_unlock_irqrestore(&ha->hardware_lock, flags);
+}
+
+/**
+ * qla4xxx_soft_reset - performs soft reset.
+ * @ha: Pointer to host adapter structure.
+ **/
+int qla4xxx_soft_reset(struct scsi_qla_host *ha)
+{
+	uint32_t max_wait_time;
+	unsigned long flags = 0;
+	int status = QLA_ERROR;
+	uint32_t ctrl_status;
+
+	qla4xxx_hw_reset(ha);
 
 	/* Wait until the Network Reset Intr bit is cleared */
 	max_wait_time = RESET_INTR_TOV;
@@ -966,10 +981,12 @@ static void qla4xxx_do_dpc(struct work_struct *work)
 	struct scsi_qla_host *ha =
 		container_of(work, struct scsi_qla_host, dpc_work);
 	struct ddb_entry *ddb_entry, *dtemp;
+	int status = QLA_ERROR;
 
 	DEBUG2(printk("scsi%ld: %s: DPC handler waking up."
-		"flags = 0x%08lx, dpc_flags = 0x%08lx\n",
-		ha->host_no, __func__, ha->flags, ha->dpc_flags));
+		"flags = 0x%08lx, dpc_flags = 0x%08lx ctrl_stat = 0x%08x\n",
+		ha->host_no, __func__, ha->flags, ha->dpc_flags,
+		readw(&ha->reg->ctrl_status)));
 
 	/* Initialization not yet finished. Don't do anything yet. */
 	if (!test_bit(AF_INIT_DONE, &ha->flags))
@@ -983,31 +1000,28 @@ static void qla4xxx_do_dpc(struct work_struct *work)
 			test_bit(DPC_RESET_HA, &ha->dpc_flags))
 			qla4xxx_recover_adapter(ha, PRESERVE_DDB_LIST);
 
-		if (test_and_clear_bit(DPC_RESET_HA_INTR, &ha->dpc_flags)) {
+		if (test_bit(DPC_RESET_HA_INTR, &ha->dpc_flags)) {
 			uint8_t wait_time = RESET_INTR_TOV;
-			unsigned long flags = 0;
 
-			qla4xxx_flush_active_srbs(ha);
-
-			spin_lock_irqsave(&ha->hardware_lock, flags);
 			while ((readw(&ha->reg->ctrl_status) &
 				(CSR_SOFT_RESET | CSR_FORCE_SOFT_RESET)) != 0) {
 				if (--wait_time == 0)
 					break;
-
-				spin_unlock_irqrestore(&ha->hardware_lock,
-						       flags);
-
 				msleep(1000);
-
-				spin_lock_irqsave(&ha->hardware_lock, flags);
 			}
-			spin_unlock_irqrestore(&ha->hardware_lock, flags);
-
 			if (wait_time == 0)
 				DEBUG2(printk("scsi%ld: %s: SR|FSR "
 					      "bit not cleared-- resetting\n",
 					      ha->host_no, __func__));
+			qla4xxx_flush_active_srbs(ha);
+			if (ql4xxx_lock_drvr_wait(ha) == QLA_SUCCESS) {
+				qla4xxx_process_aen(ha, FLUSH_DDB_CHANGED_AENS);
+				status = qla4xxx_initialize_adapter(ha,
+						PRESERVE_DDB_LIST);
+			}
+			clear_bit(DPC_RESET_HA_INTR, &ha->dpc_flags);
+			if (status == QLA_SUCCESS)
+				qla4xxx_enable_intrs(ha);
 		}
 	}
 
@@ -1062,7 +1076,7 @@ static void qla4xxx_free_adapter(struct scsi_qla_host *ha)
 
 	/* Issue Soft Reset to put firmware in unknown state */
 	if (ql4xxx_lock_drvr_wait(ha) == QLA_SUCCESS)
-		qla4xxx_soft_reset(ha);
+		qla4xxx_hw_reset(ha);
 
 	/* Remove timer thread, if present */
 	if (ha->timer_active)
@@ -1198,7 +1212,6 @@ static int __devinit qla4xxx_probe_adapter(struct pci_dev *pdev,
 	INIT_LIST_HEAD(&ha->free_srb_q);
 
 	mutex_init(&ha->mbox_sem);
-	init_waitqueue_head(&ha->mailbox_wait_queue);
 
 	spin_lock_init(&ha->hardware_lock);
 
@@ -1665,6 +1678,7 @@ no_srp_cache:
 
 static void __exit qla4xxx_module_exit(void)
 {
+	ql4_mod_unload = 1;
 	pci_unregister_driver(&qla4xxx_pci_driver);
 	iscsi_unregister_transport(&qla4xxx_iscsi_transport);
 	kmem_cache_destroy(srb_cachep);

drivers/scsi/qla4xxx/ql4_version.h

@@ -5,4 +5,4 @@
  * See LICENSE.qla4xxx for copyright and licensing details.
  */
 
-#define QLA4XXX_DRIVER_VERSION	"5.00.07-k"
+#define QLA4XXX_DRIVER_VERSION	"5.00.07-k1"