Merge tag 'for-4.17/block-20180402' of git://git.kernel.dk/linux-block

Pull block layer updates from Jens Axboe:
 "It's a pretty quiet round this time, which is nice. This contains:

   - series from Bart, cleaning up the way we set/test/clear atomic
     queue flags.

   - series from Bart, fixing races between gendisk and queue
     registration and removal.

   - set of bcache fixes and improvements from various folks, by way of
     Michael Lyle.

   - set of lightnvm updates from Matias, most of it being the 1.2 to
     2.0 transition.

   - removal of unused DIO flags from Nikolay.

   - blk-mq/sbitmap memory ordering fixes from Omar.

   - divide-by-zero fix for BFQ from Paolo.

   - minor documentation patches from Randy.

   - timeout fix from Tejun.

   - Alpha "can't write a char atomically" fix from Mikulas.

   - set of NVMe fixes by way of Keith.

   - bsg and bsg-lib improvements from Christoph.

   - a few sed-opal fixes from Jonas.

   - cdrom check-disk-change deadlock fix from Maurizio.

   - various little fixes, comment fixes, etc from various folks"

* tag 'for-4.17/block-20180402' of git://git.kernel.dk/linux-block: (139 commits)
  blk-mq: Directly schedule q->timeout_work when aborting a request
  blktrace: fix comment in blktrace_api.h
  lightnvm: remove function name in strings
  lightnvm: pblk: remove some unnecessary NULL checks
  lightnvm: pblk: don't recover unwritten lines
  lightnvm: pblk: implement 2.0 support
  lightnvm: pblk: implement get log report chunk
  lightnvm: pblk: rename ppaf* to addrf*
  lightnvm: pblk: check for supported version
  lightnvm: implement get log report chunk helpers
  lightnvm: make address conversions depend on generic device
  lightnvm: add support for 2.0 address format
  lightnvm: normalize geometry nomenclature
  lightnvm: complete geo structure with maxoc*
  lightnvm: add shorten OCSSD version in geo
  lightnvm: add minor version to generic geometry
  lightnvm: simplify geometry structure
  lightnvm: pblk: refactor init/exit sequences
  lightnvm: Avoid validation of default op value
  lightnvm: centralize permission check for lightnvm ioctl
  ...

1 files changed, 62 insertions, 16 deletions

diff --git a/block/blk-cgroup.c b/block/blk-cgroup.c
index c2033a232a44..1c16694ae145 100644
--- a/block/blk-cgroup.c
+++ b/block/blk-cgroup.c
@@ -307,11 +307,28 @@ struct blkcg_gq *blkg_lookup_create(struct blkcg *blkcg,
 	}
 }
 
+static void blkg_pd_offline(struct blkcg_gq *blkg)
+{
+	int i;
+
+	lockdep_assert_held(blkg->q->queue_lock);
+	lockdep_assert_held(&blkg->blkcg->lock);
+
+	for (i = 0; i < BLKCG_MAX_POLS; i++) {
+		struct blkcg_policy *pol = blkcg_policy[i];
+
+		if (blkg->pd[i] && !blkg->pd[i]->offline &&
+		    pol->pd_offline_fn) {
+			pol->pd_offline_fn(blkg->pd[i]);
+			blkg->pd[i]->offline = true;
+		}
+	}
+}
+
 static void blkg_destroy(struct blkcg_gq *blkg)
 {
 	struct blkcg *blkcg = blkg->blkcg;
 	struct blkcg_gq *parent = blkg->parent;
-	int i;
 
 	lockdep_assert_held(blkg->q->queue_lock);
 	lockdep_assert_held(&blkcg->lock);
@@ -320,13 +337,6 @@ static void blkg_destroy(struct blkcg_gq *blkg)
 	WARN_ON_ONCE(list_empty(&blkg->q_node));
 	WARN_ON_ONCE(hlist_unhashed(&blkg->blkcg_node));
 
-	for (i = 0; i < BLKCG_MAX_POLS; i++) {
-		struct blkcg_policy *pol = blkcg_policy[i];
-
-		if (blkg->pd[i] && pol->pd_offline_fn)
-			pol->pd_offline_fn(blkg->pd[i]);
-	}
-
 	if (parent) {
 		blkg_rwstat_add_aux(&parent->stat_bytes, &blkg->stat_bytes);
 		blkg_rwstat_add_aux(&parent->stat_ios, &blkg->stat_ios);
@@ -369,6 +379,7 @@ static void blkg_destroy_all(struct request_queue *q)
 		struct blkcg *blkcg = blkg->blkcg;
 
 		spin_lock(&blkcg->lock);
+		blkg_pd_offline(blkg);
 		blkg_destroy(blkg);
 		spin_unlock(&blkcg->lock);
 	}
@@ -995,25 +1006,25 @@ static struct cftype blkcg_legacy_files[] = {
  * @css: css of interest
  *
  * This function is called when @css is about to go away and responsible
- * for shooting down all blkgs associated with @css.  blkgs should be
- * removed while holding both q and blkcg locks.  As blkcg lock is nested
- * inside q lock, this function performs reverse double lock dancing.
+ * for offlining all blkgs pd and killing all wbs associated with @css.
+ * blkgs pd offline should be done while holding both q and blkcg locks.
+ * As blkcg lock is nested inside q lock, this function performs reverse
+ * double lock dancing.
  *
  * This is the blkcg counterpart of ioc_release_fn().
  */
 static void blkcg_css_offline(struct cgroup_subsys_state *css)
 {
 	struct blkcg *blkcg = css_to_blkcg(css);
+	struct blkcg_gq *blkg;
 
 	spin_lock_irq(&blkcg->lock);
 
-	while (!hlist_empty(&blkcg->blkg_list)) {
-		struct blkcg_gq *blkg = hlist_entry(blkcg->blkg_list.first,
-						struct blkcg_gq, blkcg_node);
+	hlist_for_each_entry(blkg, &blkcg->blkg_list, blkcg_node) {
 		struct request_queue *q = blkg->q;
 
 		if (spin_trylock(q->queue_lock)) {
-			blkg_destroy(blkg);
+			blkg_pd_offline(blkg);
 			spin_unlock(q->queue_lock);
 		} else {
 			spin_unlock_irq(&blkcg->lock);
@@ -1027,11 +1038,43 @@ static void blkcg_css_offline(struct cgroup_subsys_state *css)
 	wb_blkcg_offline(blkcg);
 }
 
+/**
+ * blkcg_destroy_all_blkgs - destroy all blkgs associated with a blkcg
+ * @blkcg: blkcg of interest
+ *
+ * This function is called when blkcg css is about to free and responsible for
+ * destroying all blkgs associated with @blkcg.
+ * blkgs should be removed while holding both q and blkcg locks. As blkcg lock
+ * is nested inside q lock, this function performs reverse double lock dancing.
+ */
+static void blkcg_destroy_all_blkgs(struct blkcg *blkcg)
+{
+	spin_lock_irq(&blkcg->lock);
+	while (!hlist_empty(&blkcg->blkg_list)) {
+		struct blkcg_gq *blkg = hlist_entry(blkcg->blkg_list.first,
+						    struct blkcg_gq,
+						    blkcg_node);
+		struct request_queue *q = blkg->q;
+
+		if (spin_trylock(q->queue_lock)) {
+			blkg_destroy(blkg);
+			spin_unlock(q->queue_lock);
+		} else {
+			spin_unlock_irq(&blkcg->lock);
+			cpu_relax();
+			spin_lock_irq(&blkcg->lock);
+		}
+	}
+	spin_unlock_irq(&blkcg->lock);
+}
+
 static void blkcg_css_free(struct cgroup_subsys_state *css)
 {
 	struct blkcg *blkcg = css_to_blkcg(css);
 	int i;
 
+	blkcg_destroy_all_blkgs(blkcg);
+
 	mutex_lock(&blkcg_pol_mutex);
 
 	list_del(&blkcg->all_blkcgs_node);
@@ -1371,8 +1414,11 @@ void blkcg_deactivate_policy(struct request_queue *q,
 		spin_lock(&blkg->blkcg->lock);
 
 		if (blkg->pd[pol->plid]) {
-			if (pol->pd_offline_fn)
+			if (!blkg->pd[pol->plid]->offline &&
+			    pol->pd_offline_fn) {
 				pol->pd_offline_fn(blkg->pd[pol->plid]);
+				blkg->pd[pol->plid]->offline = true;
+			}
 			pol->pd_free_fn(blkg->pd[pol->plid]);
 			blkg->pd[pol->plid] = NULL;
 		}