Merge tag 'io_uring-bio-cache.5-2021-08-30' of git://git.kernel.dk/linux-block

Pull support for struct bio recycling from Jens Axboe:
 "This adds bio recycling support for polled IO, allowing quick reuse of
  a bio for high IOPS scenarios via a percpu bio_set list.

  It's good for almost a 10% improvement in performance, bumping our
  per-core IO limit from ~3.2M IOPS to ~3.5M IOPS"

* tag 'io_uring-bio-cache.5-2021-08-30' of git://git.kernel.dk/linux-block:
  bio: improve kerneldoc documentation for bio_alloc_kiocb()
  block: provide bio_clear_hipri() helper
  block: use the percpu bio cache in __blkdev_direct_IO
  io_uring: enable use of bio alloc cache
  block: clear BIO_PERCPU_CACHE flag if polling isn't supported
  bio: add allocation cache abstraction
  fs: add kiocb alloc cache flag
  bio: optimize initialization of a bio

4 files changed, 162 insertions, 18 deletions

diff --git a/block/bio.c b/block/bio.c
index 265bff6b549a..1319dd2aa0b8 100644
--- a/block/bio.c
+++ b/block/bio.c
@@ -25,6 +25,11 @@
 #include "blk.h"
 #include "blk-rq-qos.h"
 
+struct bio_alloc_cache {
+	struct bio_list		free_list;
+	unsigned int		nr;
+};
+
 static struct biovec_slab {
 	int nr_vecs;
 	char *name;
@@ -246,12 +251,40 @@ static void bio_free(struct bio *bio)
 void bio_init(struct bio *bio, struct bio_vec *table,
 	      unsigned short max_vecs)
 {
-	memset(bio, 0, sizeof(*bio));
+	bio->bi_next = NULL;
+	bio->bi_bdev = NULL;
+	bio->bi_opf = 0;
+	bio->bi_flags = 0;
+	bio->bi_ioprio = 0;
+	bio->bi_write_hint = 0;
+	bio->bi_status = 0;
+	bio->bi_iter.bi_sector = 0;
+	bio->bi_iter.bi_size = 0;
+	bio->bi_iter.bi_idx = 0;
+	bio->bi_iter.bi_bvec_done = 0;
+	bio->bi_end_io = NULL;
+	bio->bi_private = NULL;
+#ifdef CONFIG_BLK_CGROUP
+	bio->bi_blkg = NULL;
+	bio->bi_issue.value = 0;
+#ifdef CONFIG_BLK_CGROUP_IOCOST
+	bio->bi_iocost_cost = 0;
+#endif
+#endif
+#ifdef CONFIG_BLK_INLINE_ENCRYPTION
+	bio->bi_crypt_context = NULL;
+#endif
+#ifdef CONFIG_BLK_DEV_INTEGRITY
+	bio->bi_integrity = NULL;
+#endif
+	bio->bi_vcnt = 0;
+
 	atomic_set(&bio->__bi_remaining, 1);
 	atomic_set(&bio->__bi_cnt, 1);
 
-	bio->bi_io_vec = table;
 	bio->bi_max_vecs = max_vecs;
+	bio->bi_io_vec = table;
+	bio->bi_pool = NULL;
 }
 EXPORT_SYMBOL(bio_init);
 
@@ -586,6 +619,53 @@ void guard_bio_eod(struct bio *bio)
 	bio_truncate(bio, maxsector << 9);
 }
 
+#define ALLOC_CACHE_MAX		512
+#define ALLOC_CACHE_SLACK	 64
+
+static void bio_alloc_cache_prune(struct bio_alloc_cache *cache,
+				  unsigned int nr)
+{
+	unsigned int i = 0;
+	struct bio *bio;
+
+	while ((bio = bio_list_pop(&cache->free_list)) != NULL) {
+		cache->nr--;
+		bio_free(bio);
+		if (++i == nr)
+			break;
+	}
+}
+
+static int bio_cpu_dead(unsigned int cpu, struct hlist_node *node)
+{
+	struct bio_set *bs;
+
+	bs = hlist_entry_safe(node, struct bio_set, cpuhp_dead);
+	if (bs->cache) {
+		struct bio_alloc_cache *cache = per_cpu_ptr(bs->cache, cpu);
+
+		bio_alloc_cache_prune(cache, -1U);
+	}
+	return 0;
+}
+
+static void bio_alloc_cache_destroy(struct bio_set *bs)
+{
+	int cpu;
+
+	if (!bs->cache)
+		return;
+
+	cpuhp_state_remove_instance_nocalls(CPUHP_BIO_DEAD, &bs->cpuhp_dead);
+	for_each_possible_cpu(cpu) {
+		struct bio_alloc_cache *cache;
+
+		cache = per_cpu_ptr(bs->cache, cpu);
+		bio_alloc_cache_prune(cache, -1U);
+	}
+	free_percpu(bs->cache);
+}
+
 /**
  * bio_put - release a reference to a bio
  * @bio:   bio to release reference to
@@ -596,16 +676,23 @@ void guard_bio_eod(struct bio *bio)
  **/
 void bio_put(struct bio *bio)
 {
-	if (!bio_flagged(bio, BIO_REFFED))
-		bio_free(bio);
-	else {
+	if (unlikely(bio_flagged(bio, BIO_REFFED))) {
 		BIO_BUG_ON(!atomic_read(&bio->__bi_cnt));
+		if (!atomic_dec_and_test(&bio->__bi_cnt))
+			return;
+	}
 
-		/*
-		 * last put frees it
-		 */
-		if (atomic_dec_and_test(&bio->__bi_cnt))
-			bio_free(bio);
+	if (bio_flagged(bio, BIO_PERCPU_CACHE)) {
+		struct bio_alloc_cache *cache;
+
+		bio_uninit(bio);
+		cache = per_cpu_ptr(bio->bi_pool->cache, get_cpu());
+		bio_list_add_head(&cache->free_list, bio);
+		if (++cache->nr > ALLOC_CACHE_MAX + ALLOC_CACHE_SLACK)
+			bio_alloc_cache_prune(cache, ALLOC_CACHE_SLACK);
+		put_cpu();
+	} else {
+		bio_free(bio);
 	}
 }
 EXPORT_SYMBOL(bio_put);
@@ -1496,6 +1583,7 @@ int biovec_init_pool(mempool_t *pool, int pool_entries)
  */
 void bioset_exit(struct bio_set *bs)
 {
+	bio_alloc_cache_destroy(bs);
 	if (bs->rescue_workqueue)
 		destroy_workqueue(bs->rescue_workqueue);
 	bs->rescue_workqueue = NULL;
@@ -1557,12 +1645,18 @@ int bioset_init(struct bio_set *bs,
 	    biovec_init_pool(&bs->bvec_pool, pool_size))
 		goto bad;
 
-	if (!(flags & BIOSET_NEED_RESCUER))
-		return 0;
-
-	bs->rescue_workqueue = alloc_workqueue("bioset", WQ_MEM_RECLAIM, 0);
-	if (!bs->rescue_workqueue)
-		goto bad;
+	if (flags & BIOSET_NEED_RESCUER) {
+		bs->rescue_workqueue = alloc_workqueue("bioset",
+							WQ_MEM_RECLAIM, 0);
+		if (!bs->rescue_workqueue)
+			goto bad;
+	}
+	if (flags & BIOSET_PERCPU_CACHE) {
+		bs->cache = alloc_percpu(struct bio_alloc_cache);
+		if (!bs->cache)
+			goto bad;
+		cpuhp_state_add_instance_nocalls(CPUHP_BIO_DEAD, &bs->cpuhp_dead);
+	}
 
 	return 0;
 bad:
@@ -1589,6 +1683,46 @@ int bioset_init_from_src(struct bio_set *bs, struct bio_set *src)
 }
 EXPORT_SYMBOL(bioset_init_from_src);
 
+/**
+ * bio_alloc_kiocb - Allocate a bio from bio_set based on kiocb
+ * @kiocb:	kiocb describing the IO
+ * @nr_iovecs:	number of iovecs to pre-allocate
+ * @bs:		bio_set to allocate from
+ *
+ * Description:
+ *    Like @bio_alloc_bioset, but pass in the kiocb. The kiocb is only
+ *    used to check if we should dip into the per-cpu bio_set allocation
+ *    cache. The allocation uses GFP_KERNEL internally. On return, the
+ *    bio is marked BIO_PERCPU_CACHEABLE, and the final put of the bio
+ *    MUST be done from process context, not hard/soft IRQ.
+ *
+ */
+struct bio *bio_alloc_kiocb(struct kiocb *kiocb, unsigned short nr_vecs,
+			    struct bio_set *bs)
+{
+	struct bio_alloc_cache *cache;
+	struct bio *bio;
+
+	if (!(kiocb->ki_flags & IOCB_ALLOC_CACHE) || nr_vecs > BIO_INLINE_VECS)
+		return bio_alloc_bioset(GFP_KERNEL, nr_vecs, bs);
+
+	cache = per_cpu_ptr(bs->cache, get_cpu());
+	bio = bio_list_pop(&cache->free_list);
+	if (bio) {
+		cache->nr--;
+		put_cpu();
+		bio_init(bio, nr_vecs ? bio->bi_inline_vecs : NULL, nr_vecs);
+		bio->bi_pool = bs;
+		bio_set_flag(bio, BIO_PERCPU_CACHE);
+		return bio;
+	}
+	put_cpu();
+	bio = bio_alloc_bioset(GFP_KERNEL, nr_vecs, bs);
+	bio_set_flag(bio, BIO_PERCPU_CACHE);
+	return bio;
+}
+EXPORT_SYMBOL_GPL(bio_alloc_kiocb);
+
 static int __init init_bio(void)
 {
 	int i;
@@ -1603,6 +1737,9 @@ static int __init init_bio(void)
 				SLAB_HWCACHE_ALIGN | SLAB_PANIC, NULL);
 	}
 
+	cpuhp_setup_state_multi(CPUHP_BIO_DEAD, "block/bio:dead", NULL,
+					bio_cpu_dead);
+
 	if (bioset_init(&fs_bio_set, BIO_POOL_SIZE, 0, BIOSET_NEED_BVECS))
 		panic("bio: can't allocate bios\n");
 
diff --git a/block/blk-core.c b/block/blk-core.c
index b5098739f72a..5454db2fa263 100644
--- a/block/blk-core.c
+++ b/block/blk-core.c
@@ -821,7 +821,7 @@ static noinline_for_stack bool submit_bio_checks(struct bio *bio)
 	}
 
 	if (!test_bit(QUEUE_FLAG_POLL, &q->queue_flags))
-		bio->bi_opf &= ~REQ_HIPRI;
+		bio_clear_hipri(bio);
 
 	switch (bio_op(bio)) {
 	case REQ_OP_DISCARD:
diff --git a/block/blk-merge.c b/block/blk-merge.c
index eeba8422ae82..7a5c81c02c80 100644
--- a/block/blk-merge.c
+++ b/block/blk-merge.c
@@ -285,7 +285,7 @@ split:
 	 * iopoll in direct IO routine. Given performance gain of iopoll for
 	 * big IO can be trival, disable iopoll when split needed.
 	 */
-	bio->bi_opf &= ~REQ_HIPRI;
+	bio_clear_hipri(bio);
 
 	return bio_split(bio, sectors, GFP_NOIO, bs);
 }
diff --git a/block/blk.h b/block/blk.h
index 346d184c82b0..8c96b0c90c48 100644
--- a/block/blk.h
+++ b/block/blk.h
@@ -366,4 +366,11 @@ extern struct device_attribute dev_attr_events;
 extern struct device_attribute dev_attr_events_async;
 extern struct device_attribute dev_attr_events_poll_msecs;
 
+static inline void bio_clear_hipri(struct bio *bio)
+{
+	/* can't support alloc cache if we turn off polling */
+	bio_clear_flag(bio, BIO_PERCPU_CACHE);
+	bio->bi_opf &= ~REQ_HIPRI;
+}
+
 #endif /* BLK_INTERNAL_H */