[dm-devel] [PATCH 1/2] dm-snapshot: fix crash with the realtime kernel

[Mike Snitzer]

On Mon, Nov 11 2019 at  8:59am -0500,
Mikulas Patocka <mpatocka at redhat.com> wrote:

> Snapshot doesn't work with realtime kernels since the commit f79ae415b64c.
> hlist_bl is implemented as a raw spinlock and the code takes two non-raw
> spinlocks while holding hlist_bl (non-raw spinlocks are blocking mutexes
> in the realtime kernel, so they couldn't be taken inside a raw spinlock).
> 
> This patch fixes the problem by using non-raw spinlock
> exception_table_lock instead of the hlist_bl lock.
> 
> Signed-off-by: Mikulas Patocka <mpatocka at redhat.com>
> Fixes: f79ae415b64c ("dm snapshot: Make exception tables scalable")
> 
> ---
>  drivers/md/dm-snap.c |   65 ++++++++++++++++++++++++++++++++-------------------
>  1 file changed, 42 insertions(+), 23 deletions(-)
> 
> Index: linux-2.6/drivers/md/dm-snap.c
> ===================================================================
> --- linux-2.6.orig/drivers/md/dm-snap.c	2019-11-08 15:51:42.000000000 +0100
> +++ linux-2.6/drivers/md/dm-snap.c	2019-11-08 15:54:58.000000000 +0100
> @@ -141,6 +141,10 @@ struct dm_snapshot {
>  	 * for them to be committed.
>  	 */
>  	struct bio_list bios_queued_during_merge;
> +
> +#ifdef CONFIG_PREEMPT_RT_BASE
> +	spinlock_t exception_table_lock;
> +#endif
>  };
>  
>  /*
> @@ -625,30 +629,42 @@ static uint32_t exception_hash(struct dm
>  
>  /* Lock to protect access to the completed and pending exception hash tables. */
>  struct dm_exception_table_lock {
> +#ifndef CONFIG_PREEMPT_RT_BASE
>  	struct hlist_bl_head *complete_slot;
>  	struct hlist_bl_head *pending_slot;
> +#endif
>  };

Why not put the spinlock_t in 'struct dm_exception_table_lock' with the
member name 'lock'?
  
>  static void dm_exception_table_lock_init(struct dm_snapshot *s, chunk_t chunk,
>  					 struct dm_exception_table_lock *lock)
>  {
> +#ifndef CONFIG_PREEMPT_RT_BASE
>  	struct dm_exception_table *complete = &s->complete;
>  	struct dm_exception_table *pending = &s->pending;
>  
>  	lock->complete_slot = &complete->table[exception_hash(complete, chunk)];
>  	lock->pending_slot = &pending->table[exception_hash(pending, chunk)];
> +#endif
>  }
>  
> -static void dm_exception_table_lock(struct dm_exception_table_lock *lock)
> +static void dm_exception_table_lock(struct dm_snapshot *s, struct dm_exception_table_lock *lock)
>  {
> +#ifdef CONFIG_PREEMPT_RT_BASE
> +	spin_lock(&s->exception_table_lock);
> +#else
>  	hlist_bl_lock(lock->complete_slot);
>  	hlist_bl_lock(lock->pending_slot);
> +#endif
>  }
>  
> -static void dm_exception_table_unlock(struct dm_exception_table_lock *lock)
> +static void dm_exception_table_unlock(struct dm_snapshot *s, struct dm_exception_table_lock *lock)
>  {
> +#ifdef CONFIG_PREEMPT_RT_BASE
> +	spin_unlock(&s->exception_table_lock);
> +#else
>  	hlist_bl_unlock(lock->pending_slot);
>  	hlist_bl_unlock(lock->complete_slot);
> +#endif
>  }
>  
>  static int dm_exception_table_init(struct dm_exception_table *et,
> @@ -835,9 +851,9 @@ static int dm_add_exception(void *contex
>  	 */
>  	dm_exception_table_lock_init(s, old, &lock);
>  
> -	dm_exception_table_lock(&lock);
> +	dm_exception_table_lock(s, &lock);
>  	dm_insert_exception(&s->complete, e);
> -	dm_exception_table_unlock(&lock);
> +	dm_exception_table_unlock(s, &lock);
>  
>  	return 0;
>  }

That way you don't need the extra 'struct dm_snapshot' arg to all the
various dm_exception_table_{lock,unlock} calls.

> @@ -1318,6 +1334,9 @@ static int snapshot_ctr(struct dm_target
>  	s->first_merging_chunk = 0;
>  	s->num_merging_chunks = 0;
>  	bio_list_init(&s->bios_queued_during_merge);
> +#ifdef CONFIG_PREEMPT_RT_BASE
> +	spin_lock_init(&s->exception_table_lock);
> +#endif
>  
>  	/* Allocate hash table for COW data */
>  	if (init_hash_tables(s)) {

And this spin_lock_init() would go in dm_exception_table_lock_init()
in appropriate #ifdef with spin_lock_init(&lock->lock)

Doing it that way would seriously reduce the size of this patch.

Unless I'm missing something, please submit a v2 and cc linux-rt-user
mailing list and the other direct CCs suggested by others in reply to
patch 2/2.

Thanks,
Mike