md/raid5-cache: Clear conf->log after finishing work

A NULL pointer dereferlence on conf->log is seen randomly with
the mdadm test 21raid5cache. Kasan reporst:

BUG: KASAN: null-ptr-deref in r5l_reclaimable_space+0xf5/0x140
Read of size 8 at addr 0000000000000860 by task md0_reclaim/3086

Call Trace:
  dump_stack_lvl+0x5a/0x74
  kasan_report.cold+0x5f/0x1a9
  __asan_load8+0x69/0x90
  r5l_reclaimable_space+0xf5/0x140
  r5l_do_reclaim+0xf4/0x5e0
  r5l_reclaim_thread+0x69/0x3b0
  md_thread+0x1a2/0x2c0
  kthread+0x177/0x1b0
  ret_from_fork+0x22/0x30

This is caused by conf->log being cleared in r5l_exit_log() before
stopping the reclaim thread.

To fix this, clear conf->log after the reclaim_thread is unregistered
and after flushing disable_writeback_work.

Signed-off-by: Logan Gunthorpe <logang@deltatee.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Song Liu <song@kernel.org>
Signed-off-by: Jens Axboe <axboe@kernel.dk>

drivers/md/raid5-cache.c

@@ -3165,12 +3165,13 @@ void r5l_exit_log(struct r5conf *conf)
 {
 	struct r5l_log *log = conf->log;
 
-	conf->log = NULL;
-
 	/* Ensure disable_writeback_work wakes up and exits */
 	wake_up(&conf->mddev->sb_wait);
 	flush_work(&log->disable_writeback_work);
 	md_unregister_thread(&log->reclaim_thread);
+
+	conf->log = NULL;
+
 	mempool_exit(&log->meta_pool);
 	bioset_exit(&log->bs);
 	mempool_exit(&log->io_pool);