scsi, bnx2fc: Fix CPU hotplug callback registration
commit7229b6d0b2a9e62d5eb4886dcd0137664c735718
authorSrivatsa S. Bhat <srivatsa.bhat@linux.vnet.ibm.com>
Mon, 10 Mar 2014 20:39:45 +0000 (11 02:09 +0530)
committerRafael J. Wysocki <rafael.j.wysocki@intel.com>
Thu, 20 Mar 2014 12:43:44 +0000 (20 13:43 +0100)
tree9c20d95dd9b4ab0171b83baa6e564d77d4a6f1e1
parentbc0003c93f0e68b5383fb36579f7efa267b7dc16
scsi, bnx2fc: Fix CPU hotplug callback registration

Subsystems that want to register CPU hotplug callbacks, as well as perform
initialization for the CPUs that are already online, often do it as shown
below:

get_online_cpus();

for_each_online_cpu(cpu)
init_cpu(cpu);

register_cpu_notifier(&foobar_cpu_notifier);

put_online_cpus();

This is wrong, since it is prone to ABBA deadlocks involving the
cpu_add_remove_lock and the cpu_hotplug.lock (when running concurrently
with CPU hotplug operations).

Instead, the correct and race-free way of performing the callback
registration is:

cpu_notifier_register_begin();

for_each_online_cpu(cpu)
init_cpu(cpu);

/* Note the use of the double underscored version of the API */
__register_cpu_notifier(&foobar_cpu_notifier);

cpu_notifier_register_done();

Fix the bnx2fc code in scsi by using this latter form of callback
registration.

Cc: Eddie Wai <eddie.wai@broadcom.com>
Cc: "James E.J. Bottomley" <JBottomley@parallels.com>
Cc: Ingo Molnar <mingo@kernel.org>
Signed-off-by: Srivatsa S. Bhat <srivatsa.bhat@linux.vnet.ibm.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
drivers/scsi/bnx2fc/bnx2fc_fcoe.c