On Fri, 2005-08-26 at 14:52 -0700, Andrew Morton wrote:
> Is there not some nasty little hack we can do to make the crash go away in
> the interim?
Well, it's nasty. The problem is that klist_next() refers to the
previous element to get the next one. However, this previous element
has just been removed and freed. We could try to open code the
necessary gets and puts in next_device().
Could someone check the attached, I think it's the correct open coding?
James
diff --git a/drivers/base/core.c b/drivers/base/core.c
--- a/drivers/base/core.c
+++ b/drivers/base/core.c
@@ -369,8 +369,20 @@ void device_unregister(struct device * d
static struct device * next_device(struct klist_iter * i)
{
+ struct device *dev;
+ struct device *odev = i->i_cur ?
+ container_of(i->i_cur, struct device, knode_parent) : NULL;
struct klist_node * n = klist_next(i);
- return n ? container_of(n, struct device, knode_parent) : NULL;
+
+ if (odev)
+ put_device(odev);
+
+ dev = n ? container_of(n, struct device, knode_parent) : NULL;
+
+ if (dev)
+ get_device(dev);
+
+ return dev;
}
/**
-
: send the line "unsubscribe linux-scsi" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
