block/mq: Cure cpu hotplug lock inversion

By poking at /debug/sched_features I triggered the following splat:

 [] ======================================================
 [] WARNING: possible circular locking dependency detected
 [] 4.11.0-00873-g964c8b7-dirty raspberrypi#694 Not tainted
 [] ------------------------------------------------------
 [] bash/2109 is trying to acquire lock:
 []  (cpu_hotplug_lock.rw_sem){++++++}, at: [<ffffffff8120cb8b>] static_key_slow_dec+0x1b/0x50
 []
 [] but task is already holding lock:
 []  (&sb->s_type->i_mutex_key#4){+++++.}, at: [<ffffffff81140216>] sched_feat_write+0x86/0x170
 []
 [] which lock already depends on the new lock.
 []
 []
 [] the existing dependency chain (in reverse order) is:
 []
 [] -> #2 (&sb->s_type->i_mutex_key#4){+++++.}:
 []        lock_acquire+0x100/0x210
 []        down_write+0x28/0x60
 []        start_creating+0x5e/0xf0
 []        debugfs_create_dir+0x13/0x110
 []        blk_mq_debugfs_register+0x21/0x70
 []        blk_mq_register_dev+0x64/0xd0
 []        blk_register_queue+0x6a/0x170
 []        device_add_disk+0x22d/0x440
 []        loop_add+0x1f3/0x280
 []        loop_init+0x104/0x142
 []        do_one_initcall+0x43/0x180
 []        kernel_init_freeable+0x1de/0x266
 []        kernel_init+0xe/0x100
 []        ret_from_fork+0x31/0x40
 []
 [] -> #1 (all_q_mutex){+.+.+.}:
 []        lock_acquire+0x100/0x210
 []        __mutex_lock+0x6c/0x960
 []        mutex_lock_nested+0x1b/0x20
 []        blk_mq_init_allocated_queue+0x37c/0x4e0
 []        blk_mq_init_queue+0x3a/0x60
 []        loop_add+0xe5/0x280
 []        loop_init+0x104/0x142
 []        do_one_initcall+0x43/0x180
 []        kernel_init_freeable+0x1de/0x266
 []        kernel_init+0xe/0x100
 []        ret_from_fork+0x31/0x40

 []  *** DEADLOCK ***
 []
 [] 3 locks held by bash/2109:
 []  #0:  (sb_writers#11){.+.+.+}, at: [<ffffffff81292bcd>] vfs_write+0x17d/0x1a0
 []  #1:  (debugfs_srcu){......}, at: [<ffffffff8155a90d>] full_proxy_write+0x5d/0xd0
 []  #2:  (&sb->s_type->i_mutex_key#4){+++++.}, at: [<ffffffff81140216>] sched_feat_write+0x86/0x170
 []
 [] stack backtrace:
 [] CPU: 9 PID: 2109 Comm: bash Not tainted 4.11.0-00873-g964c8b7-dirty raspberrypi#694
 [] Hardware name: Intel Corporation S2600GZ/S2600GZ, BIOS SE5C600.86B.02.02.0002.122320131210 12/23/2013
 [] Call Trace:

 []  lock_acquire+0x100/0x210
 []  get_online_cpus+0x2a/0x90
 []  static_key_slow_dec+0x1b/0x50
 []  static_key_disable+0x20/0x30
 []  sched_feat_write+0x131/0x170
 []  full_proxy_write+0x97/0xd0
 []  __vfs_write+0x28/0x120
 []  vfs_write+0xb5/0x1a0
 []  SyS_write+0x49/0xa0
 []  entry_SYSCALL_64_fastpath+0x23/0xc2

This is because of the cpu hotplug lock rework. Break the chain at #1
by reversing the lock acquisition order. This way i_mutex_key#4 no
longer depends on cpu_hotplug_lock and things are good.

Cc: Jens Axboe <axboe@kernel.dk>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Signed-off-by: Jens Axboe <axboe@fb.com>

block/blk-mq.c

@@ -2336,15 +2336,15 @@ struct request_queue *blk_mq_init_allocated_queue(struct blk_mq_tag_set *set,
 
 	blk_mq_init_cpu_queues(q, set->nr_hw_queues);
 
-	get_online_cpus();
 	mutex_lock(&all_q_mutex);
+	get_online_cpus();
 
 	list_add_tail(&q->all_q_node, &all_q_list);
 	blk_mq_add_queue_tag_set(set, q);
 	blk_mq_map_swqueue(q, cpu_online_mask);
 
-	mutex_unlock(&all_q_mutex);
 	put_online_cpus();
+	mutex_unlock(&all_q_mutex);
 
 	if (!(set->flags & BLK_MQ_F_NO_SCHED)) {
 		int ret;