Re: [PATCH v4 07/28] block: Introduce zone write plugging

[Christoph Hellwig <hch@lst.de>]

> +	/* We can remove zone write plugs for zones that are empty or full. */
> +	return !zwplug->wp_offset ||
> +		zwplug->wp_offset >= disk->zone_capacity;

Nit: this condition easily fits onto a single line.

> +static inline struct blk_zone_wplug *
> +disk_lookup_zone_wplug(struct gendisk *disk, sector_t sector)
> +{
> +	unsigned int zno = disk_zone_no(disk, sector);
> +	unsigned int idx = hash_32(zno, disk->zone_wplugs_hash_bits);
> +	struct blk_zone_wplug *zwplug;
> +
> +	rcu_read_lock();
> +	hlist_for_each_entry_rcu(zwplug, &disk->zone_wplugs_hash[idx], node) {
> +		if (zwplug->zone_no == zno)
> +			goto unlock;
> +	}
> +	zwplug = NULL;
> +
> +unlock:
> +	rcu_read_unlock();
> +	return zwplug;
> +}

Did we lose an atomic_inc_unless_zero here?  This now just does a lookup
under RCU, but nothing to prevent the zwplug from beeing freed?

> +	/* Resize the zone write plug memory pool if needed. */
> +	if (disk->zone_wplugs_pool->min_nr != pool_size)
> +		return mempool_resize(disk->zone_wplugs_pool, pool_size);

Note that a mempool_resize to the current size work just fine.  It takes
a pointless lock, but given that this is something that doesn't happen
frequently that probably doesn't matter.

> +#include <linux/mempool.h>

> +	mempool_t		*zone_wplugs_pool;

Please use struct mempool_s here so that you only need a forward
declaration instead of pulling in another header.