Move xen patchset to new version's subdir.

[people/pmueller/ipfire-2.x.git] / src / patches / suse-2.6.27.31 / patches.arch / s390-04-08-cio-ungroup-race-fix.patch

From: Gerald Schaefer <geraldsc@de.ibm.com>
Subject: cio: ccwgroup online vs. ungroup race condition
References: bnc#445100,LTC#50092

Symptom:     Kernel oops when a ccwgroup ungroup operation is triggered while
             online operation is still in progress.
Problem:     Atomicity of ungroup operation is not guaranteed because of
             missing locking.
Solution:    Introduce locking to ungroup operation.

Acked-by: John Jolly <jjolly@suse.de>

---
 drivers/s390/cio/ccwgroup.c |   18 +++++++++++++-----
 1 file changed, 13 insertions(+), 5 deletions(-)

Index: linux-sles11/drivers/s390/cio/ccwgroup.c
===================================================================
--- linux-sles11.orig/drivers/s390/cio/ccwgroup.c
+++ linux-sles11/drivers/s390/cio/ccwgroup.c
@@ -89,15 +89,23 @@ ccwgroup_ungroup_store(struct device *de
 
 	gdev = to_ccwgroupdev(dev);
 
-	if (gdev->state != CCWGROUP_OFFLINE)
-		return -EINVAL;
-
+	/* Prevent concurrent online/offline processing and ungrouping. */
+	if (atomic_cmpxchg(&gdev->onoff, 0, 1) != 0)
+		return -EAGAIN;
+	if (gdev->state != CCWGROUP_OFFLINE) {
+		rc = -EINVAL;
+		goto out;
+	}
 	/* Note that we cannot unregister the device from one of its
 	 * attribute methods, so we have to use this roundabout approach.
 	 */
 	rc = device_schedule_callback(dev, ccwgroup_ungroup_callback);
-	if (rc)
-		count = rc;
+out:
+	if (rc) {
+		/* Release onoff "lock" when ungrouping failed. */
+		atomic_set(&gdev->onoff, 0);
+		return rc;
+	}
 	return count;
 }