[libata] locking rewrite (== fix)

A lot of power packed into a little patch.

This change eliminates the sharing between our controller-wide spinlock
and the SCSI core's Scsi_Host lock.  As the locking in libata was
already highly compartmentalized, always referencing our own lock, and
never scsi_host::host_lock.

As a side effect, this change eliminates a deadlock from calling
scsi_finish_command() while inside our spinlock.

drivers/scsi/libata-core.c

@@ -4089,8 +4089,6 @@ static void ata_host_init(struct ata_port *ap, struct Scsi_Host *host,
 	host->unique_id = ata_unique_id++;
 	host->max_cmd_len = 12;
 
-	scsi_assign_lock(host, &host_set->lock);
-
 	ap->flags = ATA_FLAG_PORT_DISABLED;
 	ap->id = host->unique_id;
 	ap->host = host;

drivers/scsi/libata-scsi.c

@@ -39,6 +39,7 @@
 #include <scsi/scsi.h>
 #include "scsi.h"
 #include <scsi/scsi_host.h>
+#include <scsi/scsi_device.h>
 #include <linux/libata.h>
 #include <linux/hdreg.h>
 #include <asm/uaccess.h>
@@ -2405,8 +2406,12 @@ int ata_scsi_queuecmd(struct scsi_cmnd *cmd, void (*done)(struct scsi_cmnd *))
 	struct ata_port *ap;
 	struct ata_device *dev;
 	struct scsi_device *scsidev = cmd->device;
+	struct Scsi_Host *shost = scsidev->host;
 
-	ap = (struct ata_port *) &scsidev->host->hostdata[0];
+	ap = (struct ata_port *) &shost->hostdata[0];
+
+	spin_unlock(shost->host_lock);
+	spin_lock(&ap->host_set->lock);
 
 	ata_scsi_dump_cdb(ap, cmd);
 
@@ -2429,6 +2434,8 @@ int ata_scsi_queuecmd(struct scsi_cmnd *cmd, void (*done)(struct scsi_cmnd *))
 		ata_scsi_translate(ap, dev, cmd, done, atapi_xlat);
 
 out_unlock:
+	spin_unlock(&ap->host_set->lock);
+	spin_lock(shost->host_lock);
 	return 0;
 }