[PATCH] USB: ub 03 Oops with CFQ

The blk_cleanup_queue does not necesserily destroy the queue. When we
destroy the corresponding ub_dev, it may leave the queue spinlock pointer
dangling.

This patch moves spinlocks from ub_dev to static memory. The locking
scheme is not changed. These spinlocks are still separate from the ub_lock.

Signed-off-by: Pete Zaitcev <zaitcev@redhat.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>

1 files changed, 45 insertions, 23 deletions

diff --git a/drivers/block/ub.c b/drivers/block/ub.c
index a05fe58..db4943c 100644
--- a/drivers/block/ub.c
+++ b/drivers/block/ub.c
@@ -355,7 +355,7 @@ struct ub_lun {
  * The USB device instance.
  */
 struct ub_dev {
-	spinlock_t lock;
+	spinlock_t *lock;
 	atomic_t poison;		/* The USB device is disconnected */
 	int openc;			/* protected by ub_lock! */
 					/* kref is too implicit for our taste */
@@ -452,6 +452,10 @@ MODULE_DEVICE_TABLE(usb, ub_usb_ids);
 #define UB_MAX_HOSTS  26
 static char ub_hostv[UB_MAX_HOSTS];
 
+#define UB_QLOCK_NUM 5
+static spinlock_t ub_qlockv[UB_QLOCK_NUM];
+static int ub_qlock_next = 0;
+
 static DEFINE_SPINLOCK(ub_lock);	/* Locks globals and ->openc */
 
 /*
@@ -531,7 +535,7 @@ static ssize_t ub_diag_show(struct device *dev, struct device_attribute *attr,
 		return 0;
 
 	cnt = 0;
-	spin_lock_irqsave(&sc->lock, flags);
+	spin_lock_irqsave(sc->lock, flags);
 
 	cnt += sprintf(page + cnt,
 	    "poison %d reset %d\n",
@@ -579,7 +583,7 @@ static ssize_t ub_diag_show(struct device *dev, struct device_attribute *attr,
 		if (++nc == SCMD_TRACE_SZ) nc = 0;
 	}
 
-	spin_unlock_irqrestore(&sc->lock, flags);
+	spin_unlock_irqrestore(sc->lock, flags);
 	return cnt;
 }
 
@@ -627,6 +631,24 @@ static void ub_id_put(int id)
 }
 
 /*
+ * This is necessitated by the fact that blk_cleanup_queue does not
+ * necesserily destroy the queue. Instead, it may merely decrease q->refcnt.
+ * Since our blk_init_queue() passes a spinlock common with ub_dev,
+ * we have life time issues when ub_cleanup frees ub_dev.
+ */
+static spinlock_t *ub_next_lock(void)
+{
+	unsigned long flags;
+	spinlock_t *ret;
+
+	spin_lock_irqsave(&ub_lock, flags);
+	ret = &ub_qlockv[ub_qlock_next];
+	ub_qlock_next = (ub_qlock_next + 1) % UB_QLOCK_NUM;
+	spin_unlock_irqrestore(&ub_lock, flags);
+	return ret;
+}
+
+/*
  * Downcount for deallocation. This rides on two assumptions:
  *  - once something is poisoned, its refcount cannot grow
  *  - opens cannot happen at this time (del_gendisk was done)
@@ -1083,9 +1105,9 @@ static void ub_urb_timeout(unsigned long arg)
 	struct ub_dev *sc = (struct ub_dev *) arg;
 	unsigned long flags;
 
-	spin_lock_irqsave(&sc->lock, flags);
+	spin_lock_irqsave(sc->lock, flags);
 	usb_unlink_urb(&sc->work_urb);
-	spin_unlock_irqrestore(&sc->lock, flags);
+	spin_unlock_irqrestore(sc->lock, flags);
 }
 
 /*
@@ -1108,10 +1130,10 @@ static void ub_scsi_action(unsigned long _dev)
 	struct ub_dev *sc = (struct ub_dev *) _dev;
 	unsigned long flags;
 
-	spin_lock_irqsave(&sc->lock, flags);
+	spin_lock_irqsave(sc->lock, flags);
 	del_timer(&sc->work_timer);
 	ub_scsi_dispatch(sc);
-	spin_unlock_irqrestore(&sc->lock, flags);
+	spin_unlock_irqrestore(sc->lock, flags);
 }
 
 static void ub_scsi_dispatch(struct ub_dev *sc)
@@ -1754,7 +1776,7 @@ static void ub_reset_task(void *arg)
 	 * queues of resets or anything. We do need a spinlock though,
 	 * to interact with block layer.
 	 */
-	spin_lock_irqsave(&sc->lock, flags);
+	spin_lock_irqsave(sc->lock, flags);
 	sc->reset = 0;
 	tasklet_schedule(&sc->tasklet);
 	list_for_each(p, &sc->luns) {
@@ -1762,7 +1784,7 @@ static void ub_reset_task(void *arg)
 		blk_start_queue(lun->disk->queue);
 	}
 	wake_up(&sc->reset_wait);
-	spin_unlock_irqrestore(&sc->lock, flags);
+	spin_unlock_irqrestore(sc->lock, flags);
 }
 
 /*
@@ -1990,11 +2012,11 @@ static int ub_sync_tur(struct ub_dev *sc, struct ub_lun *lun)
 	cmd->done = ub_probe_done;
 	cmd->back = &compl;
 
-	spin_lock_irqsave(&sc->lock, flags);
+	spin_lock_irqsave(sc->lock, flags);
 	cmd->tag = sc->tagcnt++;
 
 	rc = ub_submit_scsi(sc, cmd);
-	spin_unlock_irqrestore(&sc->lock, flags);
+	spin_unlock_irqrestore(sc->lock, flags);
 
 	if (rc != 0) {
 		printk("ub: testing ready: submit error (%d)\n", rc); /* P3 */
@@ -2052,11 +2074,11 @@ static int ub_sync_read_cap(struct ub_dev *sc, struct ub_lun *lun,
 	cmd->done = ub_probe_done;
 	cmd->back = &compl;
 
-	spin_lock_irqsave(&sc->lock, flags);
+	spin_lock_irqsave(sc->lock, flags);
 	cmd->tag = sc->tagcnt++;
 
 	rc = ub_submit_scsi(sc, cmd);
-	spin_unlock_irqrestore(&sc->lock, flags);
+	spin_unlock_irqrestore(sc->lock, flags);
 
 	if (rc != 0) {
 		printk("ub: reading capacity: submit error (%d)\n", rc); /* P3 */
@@ -2333,7 +2355,7 @@ static int ub_probe(struct usb_interface *intf,
 	if ((sc = kmalloc(sizeof(struct ub_dev), GFP_KERNEL)) == NULL)
 		goto err_core;
 	memset(sc, 0, sizeof(struct ub_dev));
-	spin_lock_init(&sc->lock);
+	sc->lock = ub_next_lock();
 	INIT_LIST_HEAD(&sc->luns);
 	usb_init_urb(&sc->work_urb);
 	tasklet_init(&sc->tasklet, ub_scsi_action, (unsigned long)sc);
@@ -2483,7 +2505,7 @@ static int ub_probe_lun(struct ub_dev *sc, int lnum)
 	disk->driverfs_dev = &sc->intf->dev;
 
 	rc = -ENOMEM;
-	if ((q = blk_init_queue(ub_request_fn, &sc->lock)) == NULL)
+	if ((q = blk_init_queue(ub_request_fn, sc->lock)) == NULL)
 		goto err_blkqinit;
 
 	disk->queue = q;
@@ -2554,7 +2576,7 @@ static void ub_disconnect(struct usb_interface *intf)
 	 * and the whole queue drains. So, we just use this code to
 	 * print warnings.
 	 */
-	spin_lock_irqsave(&sc->lock, flags);
+	spin_lock_irqsave(sc->lock, flags);
 	{
 		struct ub_scsi_cmd *cmd;
 		int cnt = 0;
@@ -2571,7 +2593,7 @@ static void ub_disconnect(struct usb_interface *intf)
 			    "%d was queued after shutdown\n", sc->name, cnt);
 		}
 	}
-	spin_unlock_irqrestore(&sc->lock, flags);
+	spin_unlock_irqrestore(sc->lock, flags);
 
 	/*
 	 * Unregister the upper layer.
@@ -2590,19 +2612,15 @@ static void ub_disconnect(struct usb_interface *intf)
 	}
 
 	/*
-	 * Taking a lock on a structure which is about to be freed
-	 * is very nonsensual. Here it is largely a way to do a debug freeze,
-	 * and a bracket which shows where the nonsensual code segment ends.
-	 *
 	 * Testing for -EINPROGRESS is always a bug, so we are bending
 	 * the rules a little.
 	 */
-	spin_lock_irqsave(&sc->lock, flags);
+	spin_lock_irqsave(sc->lock, flags);
 	if (sc->work_urb.status == -EINPROGRESS) {	/* janitors: ignore */
 		printk(KERN_WARNING "%s: "
 		    "URB is active after disconnect\n", sc->name);
 	}
-	spin_unlock_irqrestore(&sc->lock, flags);
+	spin_unlock_irqrestore(sc->lock, flags);
 
 	/*
 	 * There is virtually no chance that other CPU runs times so long
@@ -2636,6 +2654,10 @@ static struct usb_driver ub_driver = {
 static int __init ub_init(void)
 {
 	int rc;
+	int i;
+
+	for (i = 0; i < UB_QLOCK_NUM; i++)
+		spin_lock_init(&ub_qlockv[i]);
 
 	if ((rc = register_blkdev(UB_MAJOR, DRV_NAME)) != 0)
 		goto err_regblkdev;