raid5: Use sequence counter with associated spinlock
authorAhmed S. Darwish <a.darwish@linutronix.de>
Mon, 20 Jul 2020 15:55:25 +0000 (17:55 +0200)
committerPeter Zijlstra <peterz@infradead.org>
Wed, 29 Jul 2020 14:14:27 +0000 (16:14 +0200)
A sequence counter write side critical section must be protected by some
form of locking to serialize writers. A plain seqcount_t does not
contain the information of which lock must be held when entering a write
side critical section.

Use the new seqcount_spinlock_t data type, which allows to associate a
spinlock with the sequence counter. This enables lockdep to verify that
the spinlock used for writer serialization is held when the write side
critical section is entered.

If lockdep is disabled this lock association is compiled out and has
neither storage size nor runtime overhead.

Signed-off-by: Ahmed S. Darwish <a.darwish@linutronix.de>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Acked-by: Song Liu <song@kernel.org>
Link: https://lkml.kernel.org/r/20200720155530.1173732-20-a.darwish@linutronix.de
drivers/md/raid5.c
drivers/md/raid5.h

index ab8067f..892aefe 100644 (file)
@@ -6935,7 +6935,7 @@ static struct r5conf *setup_conf(struct mddev *mddev)
        } else
                goto abort;
        spin_lock_init(&conf->device_lock);
-       seqcount_init(&conf->gen_lock);
+       seqcount_spinlock_init(&conf->gen_lock, &conf->device_lock);
        mutex_init(&conf->cache_size_mutex);
        init_waitqueue_head(&conf->wait_for_quiescent);
        init_waitqueue_head(&conf->wait_for_stripe);
index f90e070..a2c9e9e 100644 (file)
@@ -589,7 +589,7 @@ struct r5conf {
        int                     prev_chunk_sectors;
        int                     prev_algo;
        short                   generation; /* increments with every reshape */
-       seqcount_t              gen_lock;       /* lock against generation changes */
+       seqcount_spinlock_t     gen_lock;       /* lock against generation changes */
        unsigned long           reshape_checkpoint; /* Time we last updated
                                                     * metadata */
        long long               min_offset_diff; /* minimum difference between