Merge raven.il.steeleye.com:/home/jejb/BK/scsi-error-2.5

into raven.il.steeleye.com:/home/jejb/BK/scsi-misc-2.5

drivers/scsi/hosts.c

@@ -397,6 +397,7 @@ struct Scsi_Host * scsi_register(Scsi_Host_Template *shost_tp, int xtr_bytes)
 	spin_lock_init(&shost->default_lock);
 	scsi_assign_lock(shost, &shost->default_lock);
 	INIT_LIST_HEAD(&shost->my_devices);
+	INIT_LIST_HEAD(&shost->eh_cmd_q);
 
 	init_waitqueue_head(&shost->host_wait);
 	shost->dma_channel = 0xff;
@@ -634,22 +635,9 @@ void scsi_host_busy_dec_and_test(struct Scsi_Host *shost, Scsi_Device *sdev)
 	spin_lock_irqsave(shost->host_lock, flags);
 	shost->host_busy--;
 	sdev->device_busy--;
-	if (shost->in_recovery && (shost->host_busy == shost->host_failed)) {
-		up(shost->eh_wait);
-		SCSI_LOG_ERROR_RECOVERY(5, printk("Waking error handler"
-					  " thread\n"));
-	}
-	spin_unlock_irqrestore(shost->host_lock, flags);
-}
-
-void scsi_host_failed_inc_and_test(struct Scsi_Host *shost)
-{
-	unsigned long flags;
-
-	spin_lock_irqsave(shost->host_lock, flags);
-	shost->in_recovery = 1;
-	shost->host_failed++;
-	if (shost->host_busy == shost->host_failed) {
+	if (shost->in_recovery && shost->host_failed &&
+	    (shost->host_busy == shost->host_failed))
+	{
 		up(shost->eh_wait);
 		SCSI_LOG_ERROR_RECOVERY(5, printk("Waking error handler"
 					  " thread\n"));

drivers/scsi/hosts.h

@@ -384,6 +384,7 @@ struct Scsi_Host
     spinlock_t		  default_lock;
     spinlock_t		  *host_lock;
 
+    struct list_head	eh_cmd_q;
     struct task_struct    * ehandler;  /* Error recovery thread. */
     struct semaphore      * eh_wait;   /* The error recovery thread waits on
                                           this. */
@@ -587,7 +588,6 @@ extern void scsi_host_init(void);
  */
 extern void scsi_host_busy_inc(struct Scsi_Host *, Scsi_Device *);
 extern void scsi_host_busy_dec_and_test(struct Scsi_Host *, Scsi_Device *);
-extern void scsi_host_failed_inc_and_test(struct Scsi_Host *);
 
 /**
  * scsi_find_device - find a device given the host

drivers/scsi/scsi.c

@@ -790,13 +790,9 @@ static void scsi_softirq(struct softirq_action *h)
 				if ((status_byte(SCpnt->result) & CHECK_CONDITION) != 0) {
 					SCSI_LOG_MLCOMPLETE(3, print_sense("bh", SCpnt));
 				}
-				if (SCpnt->device->host->eh_wait != NULL) {
-					scsi_eh_eflags_set(SCpnt, SCSI_EH_CMD_FAILED | SCSI_EH_CMD_ERR);
-					SCpnt->owner = SCSI_OWNER_ERROR_HANDLER;
-					SCpnt->state = SCSI_STATE_FAILED;
 
-					scsi_host_failed_inc_and_test(SCpnt->device->host);
-				} else {
+				if (!scsi_eh_scmd_add(SCpnt, 0))
+				{
 					/*
 					 * We only get here if the error
 					 * recovery thread has died.
@@ -1306,6 +1302,44 @@ void scsi_device_put(struct scsi_device *sdev)
 	module_put(sdev->host->hostt->module);
 }
 
+/**
+ * scsi_set_device_offline - set scsi_device offline
+ * @sdev:	pointer to struct scsi_device to offline. 
+ *
+ * Locks:	host_lock held on entry.
+ **/
+void scsi_set_device_offline(struct scsi_device *sdev)
+{
+	struct scsi_cmnd *scmd;
+	int	cmds_active = 0;
+	unsigned long flags;
+
+	sdev->online = FALSE;
+
+	spin_lock_irqsave(&sdev->list_lock, flags);
+	list_for_each_entry(scmd, &sdev->cmd_list, list) {
+		if (scmd->request && scmd->request->rq_status != RQ_INACTIVE) {
+			/*
+			 * If we are unable to remove the timer, it means
+			 * that the command has already timed out or
+			 * finished.
+			 */
+			if (!scsi_delete_timer(scmd)) {
+				continue;
+			}
+
+			++cmds_active;
+
+			scsi_eh_scmd_add(scmd, SCSI_EH_CANCEL_CMD);
+		}
+	}
+	spin_unlock_irqrestore(&sdev->list_lock, flags);
+
+	if (!cmds_active) {
+		/* FIXME: Send online state change hotplug event */
+	}
+}
+
 /*
  * Function:	scsi_slave_attach()
  *

drivers/scsi/scsi.h

@@ -455,6 +455,7 @@ extern int scsi_slave_attach(struct scsi_device *);
 extern void scsi_slave_detach(struct scsi_device *);
 extern int scsi_device_get(struct scsi_device *);
 extern void scsi_device_put(struct scsi_device *);
+extern void scsi_set_device_offline(struct scsi_device *);
 extern void scsi_done(Scsi_Cmnd * SCpnt);
 extern void scsi_finish_command(Scsi_Cmnd *);
 extern int scsi_retry_command(Scsi_Cmnd *);
@@ -726,6 +727,7 @@ struct scsi_cmnd {
 
 	struct list_head list;  /* scsi_cmnd participates in queue lists */
 
+	struct list_head eh_entry; /* entry for the host eh_cmd_q */
 	int eh_state;		/* Used for state tracking in error handlr */
 	int eh_eflags;		/* Used by error handlr */
 	void (*done) (struct scsi_cmnd *);	/* Mid-level done function */
@@ -850,6 +852,7 @@ struct scsi_cmnd {
  */
 #define SCSI_MLQUEUE_HOST_BUSY   0x1055
 #define SCSI_MLQUEUE_DEVICE_BUSY 0x1056
+#define SCSI_MLQUEUE_EH_RETRY    0x1057
 
 /*
  * old style reset request from external source
@@ -960,13 +963,13 @@ static inline Scsi_Cmnd *scsi_find_tag(Scsi_Device *SDpnt, int tag) {
 /*
  * Scsi Error Handler Flags
  */
-#define SCSI_EH_CMD_ERR	0x0001	/* Orig cmd error'd */
-#define SCSI_EH_CMD_FAILED	0x0002	/* Orig cmd error type failed */
-#define SCSI_EH_CMD_TIMEOUT	0x0004	/* Orig cmd error type timeout */
-#define SCSI_EH_REC_TIMEOUT	0x0008	/* Recovery cmd timeout */
+#define SCSI_EH_CANCEL_CMD	0x0001	/* Cancel this cmd */
+#define SCSI_EH_REC_TIMEOUT	0x0002	/* EH retry timed out */
 
 #define SCSI_SENSE_VALID(scmd) ((scmd->sense_buffer[0] & 0x70) == 0x70)
 
+extern int scsi_eh_scmd_add(struct scsi_cmnd *, int);
+
 int scsi_set_medium_removal(Scsi_Device *dev, char state);
 
 extern int scsi_device_register(struct scsi_device *);

drivers/scsi/scsi_lib.c

@@ -117,7 +117,7 @@ int scsi_queue_insert(struct scsi_cmnd *cmd, int reason)
 	 */
 	if (reason == SCSI_MLQUEUE_HOST_BUSY)
 		host->host_blocked = host->max_host_blocked;
-	else
+	else if (reason == SCSI_MLQUEUE_DEVICE_BUSY)
 		device->device_blocked = device->max_device_blocked;
 
 	/*

drivers/scsi/scsi_syms.c

@@ -80,6 +80,7 @@ EXPORT_SYMBOL(scsi_slave_attach);
 EXPORT_SYMBOL(scsi_slave_detach);
 EXPORT_SYMBOL(scsi_device_get);
 EXPORT_SYMBOL(scsi_device_put);
+EXPORT_SYMBOL(scsi_set_device_offline);
 
 /*
  * This symbol is for the highlevel drivers (e.g. sg) only.