cycle_kernel_lock() got pushed down to ipath_open(). I tried hard to
understand what it might protect, but finally gave up.
Roland noted that qlogic seems to have abandoned the ipath driver and
came to the following wise conclusion: "So I guess if the BKL stuff is
blocking you in any way, we can just drop it from ipath and leave it
as yet another race condition in a rotting old driver."
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
LKML-Reference: <adad44tj090.fsf@cisco.com>
Cc: Roland Dreier <rdreier@cisco.com>
static int ipath_open(struct inode *in, struct file *fp)
{
/* The real work is performed later in ipath_assign_port() */
- cycle_kernel_lock();
fp->private_data = kzalloc(sizeof(struct ipath_filedata), GFP_KERNEL);
return fp->private_data ? 0 : -ENOMEM;
}