Merge branch 'for-4.14/block-postmerge' of git://git.kernel.dk/linux-block

Pull followup block layer updates from Jens Axboe:
 "I ended up splitting the main pull request for this series into two,
  mainly because of clashes between NVMe fixes that went into 4.13 after
  the for-4.14 branches were split off. This pull request is mostly
  NVMe, but not exclusively. In detail, it contains:

   - Two pull request for NVMe changes from Christoph. Nothing new on
     the feature front, basically just fixes all over the map for the
     core bits, transport, rdma, etc.

   - Series from Bart, cleaning up various bits in the BFQ scheduler.

   - Series of bcache fixes, which has been lingering for a release or
     two. Coly sent this in, but patches from various people in this
     area.

   - Set of patches for BFQ from Paolo himself, updating both
     documentation and fixing some corner cases in performance.

   - Series from Omar, attempting to now get the 4k loop support
     correct. Our confidence level is higher this time.

   - Series from Shaohua for loop as well, improving O_DIRECT
     performance and fixing a use-after-free"

* 'for-4.14/block-postmerge' of git://git.kernel.dk/linux-block: (74 commits)
  bcache: initialize dirty stripes in flash_dev_run()
  loop: set physical block size to logical block size
  bcache: fix bch_hprint crash and improve output
  bcache: Update continue_at() documentation
  bcache: silence static checker warning
  bcache: fix for gc and write-back race
  bcache: increase the number of open buckets
  bcache: Correct return value for sysfs attach errors
  bcache: correct cache_dirty_target in __update_writeback_rate()
  bcache: gc does not work when triggering by manual command
  bcache: Don't reinvent the wheel but use existing llist API
  bcache: do not subtract sectors_to_gc for bypassed IO
  bcache: fix sequential large write IO bypass
  bcache: Fix leak of bdev reference
  block/loop: remove unused field
  block/loop: fix use after free
  bfq: Use icq_to_bic() consistently
  bfq: Suppress compiler warnings about comparisons
  bfq: Check kstrtoul() return value
  bfq: Declare local functions static
  ...

10 files changed, 99 insertions, 57 deletions

diff --git a/drivers/md/bcache/alloc.c b/drivers/md/bcache/alloc.c
index ca4abe1ccd8d..cacbe2dbd5c3 100644
--- a/drivers/md/bcache/alloc.c
+++ b/drivers/md/bcache/alloc.c
@@ -68,6 +68,8 @@
 #include <linux/random.h>
 #include <trace/events/bcache.h>
 
+#define MAX_OPEN_BUCKETS 128
+
 /* Bucket heap / gen */
 
 uint8_t bch_inc_gen(struct cache *ca, struct bucket *b)
@@ -671,7 +673,7 @@ int bch_open_buckets_alloc(struct cache_set *c)
 
 	spin_lock_init(&c->data_bucket_lock);
 
-	for (i = 0; i < 6; i++) {
+	for (i = 0; i < MAX_OPEN_BUCKETS; i++) {
 		struct open_bucket *b = kzalloc(sizeof(*b), GFP_KERNEL);
 		if (!b)
 			return -ENOMEM;
diff --git a/drivers/md/bcache/bcache.h b/drivers/md/bcache/bcache.h
index dee542fff68e..2ed9bd231d84 100644
--- a/drivers/md/bcache/bcache.h
+++ b/drivers/md/bcache/bcache.h
@@ -333,6 +333,7 @@ struct cached_dev {
 	/* Limit number of writeback bios in flight */
 	struct semaphore	in_flight;
 	struct task_struct	*writeback_thread;
+	struct workqueue_struct	*writeback_write_wq;
 
 	struct keybuf		writeback_keys;
 
diff --git a/drivers/md/bcache/closure.c b/drivers/md/bcache/closure.c
index 864e673aec39..7d5286b05036 100644
--- a/drivers/md/bcache/closure.c
+++ b/drivers/md/bcache/closure.c
@@ -70,21 +70,10 @@ void __closure_wake_up(struct closure_waitlist *wait_list)
 	list = llist_del_all(&wait_list->list);
 
 	/* We first reverse the list to preserve FIFO ordering and fairness */
-
-	while (list) {
-		struct llist_node *t = list;
-		list = llist_next(list);
-
-		t->next = reverse;
-		reverse = t;
-	}
+	reverse = llist_reverse_order(list);
 
 	/* Then do the wakeups */
-
-	while (reverse) {
-		cl = container_of(reverse, struct closure, list);
-		reverse = llist_next(reverse);
-
+	llist_for_each_entry(cl, reverse, list) {
 		closure_set_waiting(cl, 0);
 		closure_sub(cl, CLOSURE_WAITING + 1);
 	}
diff --git a/drivers/md/bcache/closure.h b/drivers/md/bcache/closure.h
index 1ec84ca81146..295b7e43f92c 100644
--- a/drivers/md/bcache/closure.h
+++ b/drivers/md/bcache/closure.h
@@ -312,8 +312,6 @@ static inline void closure_wake_up(struct closure_waitlist *list)
  * been dropped with closure_put()), it will resume execution at @fn running out
  * of @wq (or, if @wq is NULL, @fn will be called by closure_put() directly).
  *
- * NOTE: This macro expands to a return in the calling function!
- *
  * This is because after calling continue_at() you no longer have a ref on @cl,
  * and whatever @cl owns may be freed out from under you - a running closure fn
  * has a ref on its own closure which continue_at() drops.
@@ -340,8 +338,6 @@ do {									\
  * Causes @fn to be executed out of @cl, in @wq context (or called directly if
  * @wq is NULL).
  *
- * NOTE: like continue_at(), this macro expands to a return in the caller!
- *
  * The ref the caller of continue_at_nobarrier() had on @cl is now owned by @fn,
  * thus it's not safe to touch anything protected by @cl after a
  * continue_at_nobarrier().
diff --git a/drivers/md/bcache/request.c b/drivers/md/bcache/request.c
index 0e1463d0c334..681b4f12b05a 100644
--- a/drivers/md/bcache/request.c
+++ b/drivers/md/bcache/request.c
@@ -196,12 +196,12 @@ static void bch_data_insert_start(struct closure *cl)
 	struct data_insert_op *op = container_of(cl, struct data_insert_op, cl);
 	struct bio *bio = op->bio, *n;
 
-	if (atomic_sub_return(bio_sectors(bio), &op->c->sectors_to_gc) < 0)
-		wake_up_gc(op->c);
-
 	if (op->bypass)
 		return bch_data_invalidate(cl);
 
+	if (atomic_sub_return(bio_sectors(bio), &op->c->sectors_to_gc) < 0)
+		wake_up_gc(op->c);
+
 	/*
 	 * Journal writes are marked REQ_PREFLUSH; if the original write was a
 	 * flush, it'll wait on the journal write.
@@ -400,12 +400,6 @@ static bool check_should_bypass(struct cached_dev *dc, struct bio *bio)
 	if (!congested && !dc->sequential_cutoff)
 		goto rescale;
 
-	if (!congested &&
-	    mode == CACHE_MODE_WRITEBACK &&
-	    op_is_write(bio->bi_opf) &&
-	    op_is_sync(bio->bi_opf))
-		goto rescale;
-
 	spin_lock(&dc->io_lock);
 
 	hlist_for_each_entry(i, iohash(dc, bio->bi_iter.bi_sector), hash)
diff --git a/drivers/md/bcache/super.c b/drivers/md/bcache/super.c
index 974d832e54a6..fc0a31b13ac4 100644
--- a/drivers/md/bcache/super.c
+++ b/drivers/md/bcache/super.c
@@ -1026,7 +1026,7 @@ int bch_cached_dev_attach(struct cached_dev *dc, struct cache_set *c)
 	}
 
 	if (BDEV_STATE(&dc->sb) == BDEV_STATE_DIRTY) {
-		bch_sectors_dirty_init(dc);
+		bch_sectors_dirty_init(&dc->disk);
 		atomic_set(&dc->has_dirty, 1);
 		atomic_inc(&dc->count);
 		bch_writeback_queue(dc);
@@ -1059,6 +1059,8 @@ static void cached_dev_free(struct closure *cl)
 	cancel_delayed_work_sync(&dc->writeback_rate_update);
 	if (!IS_ERR_OR_NULL(dc->writeback_thread))
 		kthread_stop(dc->writeback_thread);
+	if (dc->writeback_write_wq)
+		destroy_workqueue(dc->writeback_write_wq);
 
 	mutex_lock(&bch_register_lock);
 
@@ -1228,6 +1230,7 @@ static int flash_dev_run(struct cache_set *c, struct uuid_entry *u)
 		goto err;
 
 	bcache_device_attach(d, c, u - c->uuids);
+	bch_sectors_dirty_init(d);
 	bch_flash_dev_request_init(d);
 	add_disk(d->disk);
 
@@ -1374,9 +1377,6 @@ static void cache_set_flush(struct closure *cl)
 	struct btree *b;
 	unsigned i;
 
-	if (!c)
-		closure_return(cl);
-
 	bch_cache_accounting_destroy(&c->accounting);
 
 	kobject_put(&c->internal);
@@ -1964,6 +1964,8 @@ static ssize_t register_bcache(struct kobject *k, struct kobj_attribute *attr,
 			else
 				err = "device busy";
 			mutex_unlock(&bch_register_lock);
+			if (!IS_ERR(bdev))
+				bdput(bdev);
 			if (attr == &ksysfs_register_quiet)
 				goto out;
 		}
diff --git a/drivers/md/bcache/sysfs.c b/drivers/md/bcache/sysfs.c
index f90f13616980..104c57cd666c 100644
--- a/drivers/md/bcache/sysfs.c
+++ b/drivers/md/bcache/sysfs.c
@@ -192,7 +192,7 @@ STORE(__cached_dev)
 {
 	struct cached_dev *dc = container_of(kobj, struct cached_dev,
 					     disk.kobj);
-	unsigned v = size;
+	ssize_t v = size;
 	struct cache_set *c;
 	struct kobj_uevent_env *env;
 
@@ -227,7 +227,7 @@ STORE(__cached_dev)
 		bch_cached_dev_run(dc);
 
 	if (attr == &sysfs_cache_mode) {
-		ssize_t v = bch_read_string_list(buf, bch_cache_modes + 1);
+		v = bch_read_string_list(buf, bch_cache_modes + 1);
 
 		if (v < 0)
 			return v;
@@ -615,8 +615,21 @@ STORE(__bch_cache_set)
 		bch_cache_accounting_clear(&c->accounting);
 	}
 
-	if (attr == &sysfs_trigger_gc)
+	if (attr == &sysfs_trigger_gc) {
+		/*
+		 * Garbage collection thread only works when sectors_to_gc < 0,
+		 * when users write to sysfs entry trigger_gc, most of time
+		 * they want to forcibly triger gargage collection. Here -1 is
+		 * set to c->sectors_to_gc, to make gc_should_run() give a
+		 * chance to permit gc thread to run. "give a chance" means
+		 * before going into gc_should_run(), there is still chance
+		 * that c->sectors_to_gc being set to other positive value. So
+		 * writing sysfs entry trigger_gc won't always make sure gc
+		 * thread takes effect.
+		 */
+		atomic_set(&c->sectors_to_gc, -1);
 		wake_up_gc(c);
+	}
 
 	if (attr == &sysfs_prune_cache) {
 		struct shrink_control sc;
diff --git a/drivers/md/bcache/util.c b/drivers/md/bcache/util.c
index 8c3a938f4bf0..176d3c2ef5f5 100644
--- a/drivers/md/bcache/util.c
+++ b/drivers/md/bcache/util.c
@@ -74,24 +74,44 @@ STRTO_H(strtouint, unsigned int)
 STRTO_H(strtoll, long long)
 STRTO_H(strtoull, unsigned long long)
 
+/**
+ * bch_hprint() - formats @v to human readable string for sysfs.
+ *
+ * @v - signed 64 bit integer
+ * @buf - the (at least 8 byte) buffer to format the result into.
+ *
+ * Returns the number of bytes used by format.
+ */
 ssize_t bch_hprint(char *buf, int64_t v)
 {
 	static const char units[] = "?kMGTPEZY";
-	char dec[4] = "";
-	int u, t = 0;
-
-	for (u = 0; v >= 1024 || v <= -1024; u++) {
-		t = v & ~(~0 << 10);
-		v >>= 10;
-	}
-
-	if (!u)
-		return sprintf(buf, "%llu", v);
-
-	if (v < 100 && v > -100)
-		snprintf(dec, sizeof(dec), ".%i", t / 100);
-
-	return sprintf(buf, "%lli%s%c", v, dec, units[u]);
+	int u = 0, t;
+
+	uint64_t q;
+
+	if (v < 0)
+		q = -v;
+	else
+		q = v;
+
+	/* For as long as the number is more than 3 digits, but at least
+	 * once, shift right / divide by 1024.  Keep the remainder for
+	 * a digit after the decimal point.
+	 */
+	do {
+		u++;
+
+		t = q & ~(~0 << 10);
+		q >>= 10;
+	} while (q >= 1000);
+
+	if (v < 0)
+		/* '-', up to 3 digits, '.', 1 digit, 1 character, null;
+		 * yields 8 bytes.
+		 */
+		return sprintf(buf, "-%llu.%i%c", q, t * 10 / 1024, units[u]);
+	else
+		return sprintf(buf, "%llu.%i%c", q, t * 10 / 1024, units[u]);
 }
 
 ssize_t bch_snprint_string_list(char *buf, size_t size, const char * const list[],
diff --git a/drivers/md/bcache/writeback.c b/drivers/md/bcache/writeback.c
index c49022a8dc9d..e663ca082183 100644
--- a/drivers/md/bcache/writeback.c
+++ b/drivers/md/bcache/writeback.c
@@ -21,7 +21,8 @@
 static void __update_writeback_rate(struct cached_dev *dc)
 {
 	struct cache_set *c = dc->disk.c;
-	uint64_t cache_sectors = c->nbuckets * c->sb.bucket_size;
+	uint64_t cache_sectors = c->nbuckets * c->sb.bucket_size -
+				bcache_flash_devs_sectors_dirty(c);
 	uint64_t cache_dirty_target =
 		div_u64(cache_sectors * dc->writeback_percent, 100);
 
@@ -186,7 +187,7 @@ static void write_dirty(struct closure *cl)
 
 	closure_bio_submit(&io->bio, cl);
 
-	continue_at(cl, write_dirty_finish, system_wq);
+	continue_at(cl, write_dirty_finish, io->dc->writeback_write_wq);
 }
 
 static void read_dirty_endio(struct bio *bio)
@@ -206,7 +207,7 @@ static void read_dirty_submit(struct closure *cl)
 
 	closure_bio_submit(&io->bio, cl);
 
-	continue_at(cl, write_dirty, system_wq);
+	continue_at(cl, write_dirty, io->dc->writeback_write_wq);
 }
 
 static void read_dirty(struct cached_dev *dc)
@@ -481,17 +482,17 @@ static int sectors_dirty_init_fn(struct btree_op *_op, struct btree *b,
 	return MAP_CONTINUE;
 }
 
-void bch_sectors_dirty_init(struct cached_dev *dc)
+void bch_sectors_dirty_init(struct bcache_device *d)
 {
 	struct sectors_dirty_init op;
 
 	bch_btree_op_init(&op.op, -1);
-	op.inode = dc->disk.id;
+	op.inode = d->id;
 
-	bch_btree_map_keys(&op.op, dc->disk.c, &KEY(op.inode, 0, 0),
+	bch_btree_map_keys(&op.op, d->c, &KEY(op.inode, 0, 0),
 			   sectors_dirty_init_fn, 0);
 
-	dc->disk.sectors_dirty_last = bcache_dev_sectors_dirty(&dc->disk);
+	d->sectors_dirty_last = bcache_dev_sectors_dirty(d);
 }
 
 void bch_cached_dev_writeback_init(struct cached_dev *dc)
@@ -515,6 +516,11 @@ void bch_cached_dev_writeback_init(struct cached_dev *dc)
 
 int bch_cached_dev_writeback_start(struct cached_dev *dc)
 {
+	dc->writeback_write_wq = alloc_workqueue("bcache_writeback_wq",
+						WQ_MEM_RECLAIM, 0);
+	if (!dc->writeback_write_wq)
+		return -ENOMEM;
+
 	dc->writeback_thread = kthread_create(bch_writeback_thread, dc,
 					      "bcache_writeback");
 	if (IS_ERR(dc->writeback_thread))
diff --git a/drivers/md/bcache/writeback.h b/drivers/md/bcache/writeback.h
index 629bd1a502fd..e35421d20d2e 100644
--- a/drivers/md/bcache/writeback.h
+++ b/drivers/md/bcache/writeback.h
@@ -14,6 +14,25 @@ static inline uint64_t bcache_dev_sectors_dirty(struct bcache_device *d)
 	return ret;
 }
 
+static inline uint64_t  bcache_flash_devs_sectors_dirty(struct cache_set *c)
+{
+	uint64_t i, ret = 0;
+
+	mutex_lock(&bch_register_lock);
+
+	for (i = 0; i < c->nr_uuids; i++) {
+		struct bcache_device *d = c->devices[i];
+
+		if (!d || !UUID_FLASH_ONLY(&c->uuids[i]))
+			continue;
+	   ret += bcache_dev_sectors_dirty(d);
+	}
+
+	mutex_unlock(&bch_register_lock);
+
+	return ret;
+}
+
 static inline unsigned offset_to_stripe(struct bcache_device *d,
 					uint64_t offset)
 {
@@ -84,7 +103,7 @@ static inline void bch_writeback_add(struct cached_dev *dc)
 
 void bcache_dev_sectors_dirty_add(struct cache_set *, unsigned, uint64_t, int);
 
-void bch_sectors_dirty_init(struct cached_dev *dc);
+void bch_sectors_dirty_init(struct bcache_device *);
 void bch_cached_dev_writeback_init(struct cached_dev *);
 int bch_cached_dev_writeback_start(struct cached_dev *);