[dm-devel] [PATCH for-dm-3.14-fixes 2/8] dm thin: set flag when over the metadata low watermark threshold

[Joe Thornber]

NACK

On Thu, Feb 20, 2014 at 09:55:59PM -0500, Mike Snitzer wrote:
> The threshold boundary code in persistent-data/dm-space-map-metadata.c
> was too racey and resulted in a flood of warnings and events.

This code never runs concurrently, so I don't see how it can be racey.
Scanning the code it seems to work as advertised; it is an edge
triggered threshold, activating every time the free space crosses the
threshold from high to low.

- You haven't explained what your changes within
  dm-space-map-metadata.c do.  Is the threshold no longer edge
  triggered?

- With your changes, how is the threshold reset?  For instance:

  i) we cross the lwm and issue an event

  ii) the admin deletes a ton of old snapshots to free up some
      metadata space, taking us well above the lwm.

  iii) 3 months later we cross the lwm again.  Will the admin be notified?

I suspect the real issue is when the free space is near the threshold
it 'bobbles', crossing back and forth and triggering often (not a
race).  I suggest the correct way to fix this is to change the
threshold code to introduce some hysteresis.  For example by saying it
can't trigger unless free space has crossed another higher threshold
first (lwm + FUDGE_FACTOR).  Fixes for this should remain within
dm-space-map-metadata.c.

- Joe





> 
> Check the 'metadata_low_water_triggered' flag in metadata_low_callback()
> before logging a warning and sending an event.
> 
> Signed-off-by: Mike Snitzer <snitzer at redhat.com>
> ---
>  drivers/md/dm-thin.c                               | 11 ++++++++++-
>  drivers/md/persistent-data/dm-space-map-metadata.c | 19 +------------------
>  2 files changed, 11 insertions(+), 19 deletions(-)
> 
> diff --git a/drivers/md/dm-thin.c b/drivers/md/dm-thin.c
> index 9facc6f..42d08eb 100644
> --- a/drivers/md/dm-thin.c
> +++ b/drivers/md/dm-thin.c
> @@ -165,6 +165,7 @@ struct pool {
>  
>  	struct pool_features pf;
>  	bool low_water_triggered:1;	/* A dm event has been sent */
> +	bool metadata_low_water_triggered:1;
>  
>  	struct dm_bio_prison *prison;
>  	struct dm_kcopyd_client *copier;
> @@ -1824,6 +1825,7 @@ static struct pool *pool_create(struct mapped_device *pool_md,
>  	INIT_LIST_HEAD(&pool->prepared_mappings);
>  	INIT_LIST_HEAD(&pool->prepared_discards);
>  	pool->low_water_triggered = false;
> +	pool->metadata_low_water_triggered = false;
>  	bio_list_init(&pool->retry_on_resume_list);
>  
>  	pool->shared_read_ds = dm_deferred_set_create();
> @@ -1993,10 +1995,16 @@ static int parse_pool_features(struct dm_arg_set *as, struct pool_features *pf,
>  static void metadata_low_callback(void *context)
>  {
>  	struct pool *pool = context;
> +	unsigned long flags;
> +
> +	if (pool->metadata_low_water_triggered)
> +		return;
>  
>  	DMWARN("%s: reached low water mark for metadata device: sending event.",
>  	       dm_device_name(pool->pool_md));
> -
> +	spin_lock_irqsave(&pool->lock, flags);
> +	pool->metadata_low_water_triggered = true;
> +	spin_unlock_irqrestore(&pool->lock, flags);
>  	dm_table_event(pool->ti->table);
>  }
>  
> @@ -2350,6 +2358,7 @@ static void pool_resume(struct dm_target *ti)
>  
>  	spin_lock_irqsave(&pool->lock, flags);
>  	pool->low_water_triggered = false;
> +	pool->metadata_low_water_triggered = false;
>  	__requeue_bios(pool);
>  	spin_unlock_irqrestore(&pool->lock, flags);
>  
> diff --git a/drivers/md/persistent-data/dm-space-map-metadata.c b/drivers/md/persistent-data/dm-space-map-metadata.c
> index 536782e..a0231d3 100644
> --- a/drivers/md/persistent-data/dm-space-map-metadata.c
> +++ b/drivers/md/persistent-data/dm-space-map-metadata.c
> @@ -21,9 +21,7 @@
>   */
>  struct threshold {
>  	bool threshold_set;
> -	bool value_set;
>  	dm_block_t threshold;
> -	dm_block_t current_value;
>  	dm_sm_threshold_fn fn;
>  	void *context;
>  };
> @@ -31,7 +29,6 @@ struct threshold {
>  static void threshold_init(struct threshold *t)
>  {
>  	t->threshold_set = false;
> -	t->value_set = false;
>  }
>  
>  static void set_threshold(struct threshold *t, dm_block_t value,
> @@ -43,24 +40,10 @@ static void set_threshold(struct threshold *t, dm_block_t value,
>  	t->context = context;
>  }
>  
> -static bool below_threshold(struct threshold *t, dm_block_t value)
> -{
> -	return t->threshold_set && value <= t->threshold;
> -}
> -
> -static bool threshold_already_triggered(struct threshold *t)
> -{
> -	return t->value_set && below_threshold(t, t->current_value);
> -}
> -
>  static void check_threshold(struct threshold *t, dm_block_t value)
>  {
> -	if (below_threshold(t, value) &&
> -	    !threshold_already_triggered(t))
> +	if (t->threshold_set && value <= t->threshold)
>  		t->fn(t->context);
> -
> -	t->value_set = true;
> -	t->current_value = value;
>  }
>  
>  /*----------------------------------------------------------------*/
> -- 
> 1.8.3.1
>