Merge branch 'for-3.5/core' of git://git.kernel.dk/linux-block

Merge block/IO core bits from Jens Axboe:
 "This is a bit bigger on the core side than usual, but that is purely
  because we decided to hold off on parts of Tejun's submission on 3.4
  to give it a bit more time to simmer.  As a consequence, it's seen a
  long cycle in for-next.

  It contains:

   - Bug fix from Dan, wrong locking type.
   - Relax splice gifting restriction from Eric.
   - A ton of updates from Tejun, primarily for blkcg.  This improves
     the code a lot, making the API nicer and cleaner, and also includes
     fixes for how we handle and tie policies and re-activate on
     switches.  The changes also include generic bug fixes.
   - A simple fix from Vivek, along with a fix for doing proper delayed
     allocation of the blkcg stats."

Fix up annoying conflict just due to different merge resolution in
Documentation/feature-removal-schedule.txt

* 'for-3.5/core' of git://git.kernel.dk/linux-block: (92 commits)
  blkcg: tg_stats_alloc_lock is an irq lock
  vmsplice: relax alignement requirements for SPLICE_F_GIFT
  blkcg: use radix tree to index blkgs from blkcg
  blkcg: fix blkcg->css ref leak in __blkg_lookup_create()
  block: fix elvpriv allocation failure handling
  block: collapse blk_alloc_request() into get_request()
  blkcg: collapse blkcg_policy_ops into blkcg_policy
  blkcg: embed struct blkg_policy_data in policy specific data
  blkcg: mass rename of blkcg API
  blkcg: style cleanups for blk-cgroup.h
  blkcg: remove blkio_group->path[]
  blkcg: blkg_rwstat_read() was missing inline
  blkcg: shoot down blkgs if all policies are deactivated
  blkcg: drop stuff unused after per-queue policy activation update
  blkcg: implement per-queue policy activation
  blkcg: add request_queue->root_blkg
  blkcg: make request_queue bypassing on allocation
  blkcg: make sure blkg_lookup() returns %NULL if @q is bypassing
  blkcg: make blkg_conf_prep() take @pol and return with queue lock held
  blkcg: remove static policy ID enums
  ...

block/Kconfig.iosched

@@ -23,8 +23,6 @@ config IOSCHED_DEADLINE
 
 config IOSCHED_CFQ
 	tristate "CFQ I/O scheduler"
-	# If BLK_CGROUP is a module, CFQ has to be built as module.
-	depends on (BLK_CGROUP=m && m) || !BLK_CGROUP || BLK_CGROUP=y
 	default y
 	---help---
 	  The CFQ I/O scheduler tries to distribute bandwidth equally
@@ -34,8 +32,6 @@ config IOSCHED_CFQ
 
 	  This is the default I/O scheduler.
 
-	  Note: If BLK_CGROUP=m, then CFQ can be built only as module.
-
 config CFQ_GROUP_IOSCHED
 	bool "CFQ Group Scheduling support"
 	depends on IOSCHED_CFQ && BLK_CGROUP

block/blk-cgroup.h

@@ -15,350 +15,371 @@
 
 #include <linux/cgroup.h>
 #include <linux/u64_stats_sync.h>
-
-enum blkio_policy_id {
-	BLKIO_POLICY_PROP = 0,		/* Proportional Bandwidth division */
-	BLKIO_POLICY_THROTL,		/* Throttling */
-};
+#include <linux/seq_file.h>
+#include <linux/radix-tree.h>
 
 /* Max limits for throttle policy */
 #define THROTL_IOPS_MAX		UINT_MAX
 
-#if defined(CONFIG_BLK_CGROUP) || defined(CONFIG_BLK_CGROUP_MODULE)
-
-#ifndef CONFIG_BLK_CGROUP
-/* When blk-cgroup is a module, its subsys_id isn't a compile-time constant */
-extern struct cgroup_subsys blkio_subsys;
-#define blkio_subsys_id blkio_subsys.subsys_id
-#endif
-
-enum stat_type {
-	/* Total time spent (in ns) between request dispatch to the driver and
-	 * request completion for IOs doen by this cgroup. This may not be
-	 * accurate when NCQ is turned on. */
-	BLKIO_STAT_SERVICE_TIME = 0,
-	/* Total time spent waiting in scheduler queue in ns */
-	BLKIO_STAT_WAIT_TIME,
-	/* Number of IOs queued up */
-	BLKIO_STAT_QUEUED,
-	/* All the single valued stats go below this */
-	BLKIO_STAT_TIME,
-#ifdef CONFIG_DEBUG_BLK_CGROUP
-	/* Time not charged to this cgroup */
-	BLKIO_STAT_UNACCOUNTED_TIME,
-	BLKIO_STAT_AVG_QUEUE_SIZE,
-	BLKIO_STAT_IDLE_TIME,
-	BLKIO_STAT_EMPTY_TIME,
-	BLKIO_STAT_GROUP_WAIT_TIME,
-	BLKIO_STAT_DEQUEUE
-#endif
-};
+/* CFQ specific, out here for blkcg->cfq_weight */
+#define CFQ_WEIGHT_MIN		10
+#define CFQ_WEIGHT_MAX		1000
+#define CFQ_WEIGHT_DEFAULT	500
 
-/* Per cpu stats */
-enum stat_type_cpu {
-	BLKIO_STAT_CPU_SECTORS,
-	/* Total bytes transferred */
-	BLKIO_STAT_CPU_SERVICE_BYTES,
-	/* Total IOs serviced, post merge */
-	BLKIO_STAT_CPU_SERVICED,
-	/* Number of IOs merged */
-	BLKIO_STAT_CPU_MERGED,
-	BLKIO_STAT_CPU_NR
-};
+#ifdef CONFIG_BLK_CGROUP
 
-enum stat_sub_type {
-	BLKIO_STAT_READ = 0,
-	BLKIO_STAT_WRITE,
-	BLKIO_STAT_SYNC,
-	BLKIO_STAT_ASYNC,
-	BLKIO_STAT_TOTAL
-};
+enum blkg_rwstat_type {
+	BLKG_RWSTAT_READ,
+	BLKG_RWSTAT_WRITE,
+	BLKG_RWSTAT_SYNC,
+	BLKG_RWSTAT_ASYNC,
 
-/* blkg state flags */
-enum blkg_state_flags {
-	BLKG_waiting = 0,
-	BLKG_idling,
-	BLKG_empty,
+	BLKG_RWSTAT_NR,
+	BLKG_RWSTAT_TOTAL = BLKG_RWSTAT_NR,
 };
 
-/* cgroup files owned by proportional weight policy */
-enum blkcg_file_name_prop {
-	BLKIO_PROP_weight = 1,
-	BLKIO_PROP_weight_device,
-	BLKIO_PROP_io_service_bytes,
-	BLKIO_PROP_io_serviced,
-	BLKIO_PROP_time,
-	BLKIO_PROP_sectors,
-	BLKIO_PROP_unaccounted_time,
-	BLKIO_PROP_io_service_time,
-	BLKIO_PROP_io_wait_time,
-	BLKIO_PROP_io_merged,
-	BLKIO_PROP_io_queued,
-	BLKIO_PROP_avg_queue_size,
-	BLKIO_PROP_group_wait_time,
-	BLKIO_PROP_idle_time,
-	BLKIO_PROP_empty_time,
-	BLKIO_PROP_dequeue,
-};
+struct blkcg_gq;
 
-/* cgroup files owned by throttle policy */
-enum blkcg_file_name_throtl {
-	BLKIO_THROTL_read_bps_device,
-	BLKIO_THROTL_write_bps_device,
-	BLKIO_THROTL_read_iops_device,
-	BLKIO_THROTL_write_iops_device,
-	BLKIO_THROTL_io_service_bytes,
-	BLKIO_THROTL_io_serviced,
-};
+struct blkcg {
+	struct cgroup_subsys_state	css;
+	spinlock_t			lock;
 
-struct blkio_cgroup {
-	struct cgroup_subsys_state css;
-	unsigned int weight;
-	spinlock_t lock;
-	struct hlist_head blkg_list;
-	struct list_head policy_list; /* list of blkio_policy_node */
-};
+	struct radix_tree_root		blkg_tree;
+	struct blkcg_gq			*blkg_hint;
+	struct hlist_head		blkg_list;
+
+	/* for policies to test whether associated blkcg has changed */
+	uint64_t			id;
 
-struct blkio_group_stats {
-	/* total disk time and nr sectors dispatched by this group */
-	uint64_t time;
-	uint64_t stat_arr[BLKIO_STAT_QUEUED + 1][BLKIO_STAT_TOTAL];
-#ifdef CONFIG_DEBUG_BLK_CGROUP
-	/* Time not charged to this cgroup */
-	uint64_t unaccounted_time;
-
-	/* Sum of number of IOs queued across all samples */
-	uint64_t avg_queue_size_sum;
-	/* Count of samples taken for average */
-	uint64_t avg_queue_size_samples;
-	/* How many times this group has been removed from service tree */
-	unsigned long dequeue;
-
-	/* Total time spent waiting for it to be assigned a timeslice. */
-	uint64_t group_wait_time;
-	uint64_t start_group_wait_time;
-
-	/* Time spent idling for this blkio_group */
-	uint64_t idle_time;
-	uint64_t start_idle_time;
-	/*
-	 * Total time when we have requests queued and do not contain the
-	 * current active queue.
-	 */
-	uint64_t empty_time;
-	uint64_t start_empty_time;
-	uint16_t flags;
-#endif
+	/* TODO: per-policy storage in blkcg */
+	unsigned int			cfq_weight;	/* belongs to cfq */
 };
 
-/* Per cpu blkio group stats */
-struct blkio_group_stats_cpu {
-	uint64_t sectors;
-	uint64_t stat_arr_cpu[BLKIO_STAT_CPU_NR][BLKIO_STAT_TOTAL];
-	struct u64_stats_sync syncp;
+struct blkg_stat {
+	struct u64_stats_sync		syncp;
+	uint64_t			cnt;
 };
 
-struct blkio_group {
-	/* An rcu protected unique identifier for the group */
-	void *key;
-	struct hlist_node blkcg_node;
-	unsigned short blkcg_id;
-	/* Store cgroup path */
-	char path[128];
-	/* The device MKDEV(major, minor), this group has been created for */
-	dev_t dev;
-	/* policy which owns this blk group */
-	enum blkio_policy_id plid;
-
-	/* Need to serialize the stats in the case of reset/update */
-	spinlock_t stats_lock;
-	struct blkio_group_stats stats;
-	/* Per cpu stats pointer */
-	struct blkio_group_stats_cpu __percpu *stats_cpu;
+struct blkg_rwstat {
+	struct u64_stats_sync		syncp;
+	uint64_t			cnt[BLKG_RWSTAT_NR];
 };
 
-struct blkio_policy_node {
-	struct list_head node;
-	dev_t dev;
-	/* This node belongs to max bw policy or porportional weight policy */
-	enum blkio_policy_id plid;
-	/* cgroup file to which this rule belongs to */
-	int fileid;
-
-	union {
-		unsigned int weight;
-		/*
-		 * Rate read/write in terms of bytes per second
-		 * Whether this rate represents read or write is determined
-		 * by file type "fileid".
-		 */
-		u64 bps;
-		unsigned int iops;
-	} val;
+/*
+ * A blkcg_gq (blkg) is association between a block cgroup (blkcg) and a
+ * request_queue (q).  This is used by blkcg policies which need to track
+ * information per blkcg - q pair.
+ *
+ * There can be multiple active blkcg policies and each has its private
+ * data on each blkg, the size of which is determined by
+ * blkcg_policy->pd_size.  blkcg core allocates and frees such areas
+ * together with blkg and invokes pd_init/exit_fn() methods.
+ *
+ * Such private data must embed struct blkg_policy_data (pd) at the
+ * beginning and pd_size can't be smaller than pd.
+ */
+struct blkg_policy_data {
+	/* the blkg this per-policy data belongs to */
+	struct blkcg_gq			*blkg;
+
+	/* used during policy activation */
+	struct list_head		alloc_node;
 };
 
-extern unsigned int blkcg_get_weight(struct blkio_cgroup *blkcg,
-				     dev_t dev);
-extern uint64_t blkcg_get_read_bps(struct blkio_cgroup *blkcg,
-				     dev_t dev);
-extern uint64_t blkcg_get_write_bps(struct blkio_cgroup *blkcg,
-				     dev_t dev);
-extern unsigned int blkcg_get_read_iops(struct blkio_cgroup *blkcg,
-				     dev_t dev);
-extern unsigned int blkcg_get_write_iops(struct blkio_cgroup *blkcg,
-				     dev_t dev);
-
-typedef void (blkio_unlink_group_fn) (void *key, struct blkio_group *blkg);
-
-typedef void (blkio_update_group_weight_fn) (void *key,
-			struct blkio_group *blkg, unsigned int weight);
-typedef void (blkio_update_group_read_bps_fn) (void * key,
-			struct blkio_group *blkg, u64 read_bps);
-typedef void (blkio_update_group_write_bps_fn) (void *key,
-			struct blkio_group *blkg, u64 write_bps);
-typedef void (blkio_update_group_read_iops_fn) (void *key,
-			struct blkio_group *blkg, unsigned int read_iops);
-typedef void (blkio_update_group_write_iops_fn) (void *key,
-			struct blkio_group *blkg, unsigned int write_iops);
-
-struct blkio_policy_ops {
-	blkio_unlink_group_fn *blkio_unlink_group_fn;
-	blkio_update_group_weight_fn *blkio_update_group_weight_fn;
-	blkio_update_group_read_bps_fn *blkio_update_group_read_bps_fn;
-	blkio_update_group_write_bps_fn *blkio_update_group_write_bps_fn;
-	blkio_update_group_read_iops_fn *blkio_update_group_read_iops_fn;
-	blkio_update_group_write_iops_fn *blkio_update_group_write_iops_fn;
+/* association between a blk cgroup and a request queue */
+struct blkcg_gq {
+	/* Pointer to the associated request_queue */
+	struct request_queue		*q;
+	struct list_head		q_node;
+	struct hlist_node		blkcg_node;
+	struct blkcg			*blkcg;
+	/* reference count */
+	int				refcnt;
+
+	struct blkg_policy_data		*pd[BLKCG_MAX_POLS];
+
+	struct rcu_head			rcu_head;
 };
 
-struct blkio_policy_type {
-	struct list_head list;
-	struct blkio_policy_ops ops;
-	enum blkio_policy_id plid;
+typedef void (blkcg_pol_init_pd_fn)(struct blkcg_gq *blkg);
+typedef void (blkcg_pol_exit_pd_fn)(struct blkcg_gq *blkg);
+typedef void (blkcg_pol_reset_pd_stats_fn)(struct blkcg_gq *blkg);
+
+struct blkcg_policy {
+	int				plid;
+	/* policy specific private data size */
+	size_t				pd_size;
+	/* cgroup files for the policy */
+	struct cftype			*cftypes;
+
+	/* operations */
+	blkcg_pol_init_pd_fn		*pd_init_fn;
+	blkcg_pol_exit_pd_fn		*pd_exit_fn;
+	blkcg_pol_reset_pd_stats_fn	*pd_reset_stats_fn;
 };
 
+extern struct blkcg blkcg_root;
+
+struct blkcg *cgroup_to_blkcg(struct cgroup *cgroup);
+struct blkcg *bio_blkcg(struct bio *bio);
+struct blkcg_gq *blkg_lookup(struct blkcg *blkcg, struct request_queue *q);
+struct blkcg_gq *blkg_lookup_create(struct blkcg *blkcg,
+				    struct request_queue *q);
+int blkcg_init_queue(struct request_queue *q);
+void blkcg_drain_queue(struct request_queue *q);
+void blkcg_exit_queue(struct request_queue *q);
+
 /* Blkio controller policy registration */
-extern void blkio_policy_register(struct blkio_policy_type *);
-extern void blkio_policy_unregister(struct blkio_policy_type *);
+int blkcg_policy_register(struct blkcg_policy *pol);
+void blkcg_policy_unregister(struct blkcg_policy *pol);
+int blkcg_activate_policy(struct request_queue *q,
+			  const struct blkcg_policy *pol);
+void blkcg_deactivate_policy(struct request_queue *q,
+			     const struct blkcg_policy *pol);
+
+void blkcg_print_blkgs(struct seq_file *sf, struct blkcg *blkcg,
+		       u64 (*prfill)(struct seq_file *,
+				     struct blkg_policy_data *, int),
+		       const struct blkcg_policy *pol, int data,
+		       bool show_total);
+u64 __blkg_prfill_u64(struct seq_file *sf, struct blkg_policy_data *pd, u64 v);
+u64 __blkg_prfill_rwstat(struct seq_file *sf, struct blkg_policy_data *pd,
+			 const struct blkg_rwstat *rwstat);
+u64 blkg_prfill_stat(struct seq_file *sf, struct blkg_policy_data *pd, int off);
+u64 blkg_prfill_rwstat(struct seq_file *sf, struct blkg_policy_data *pd,
+		       int off);
+
+struct blkg_conf_ctx {
+	struct gendisk			*disk;
+	struct blkcg_gq			*blkg;
+	u64				v;
+};
+
+int blkg_conf_prep(struct blkcg *blkcg, const struct blkcg_policy *pol,
+		   const char *input, struct blkg_conf_ctx *ctx);
+void blkg_conf_finish(struct blkg_conf_ctx *ctx);
+
+
+/**
+ * blkg_to_pdata - get policy private data
+ * @blkg: blkg of interest
+ * @pol: policy of interest
+ *
+ * Return pointer to private data associated with the @blkg-@pol pair.
+ */
+static inline struct blkg_policy_data *blkg_to_pd(struct blkcg_gq *blkg,
+						  struct blkcg_policy *pol)
+{
+	return blkg ? blkg->pd[pol->plid] : NULL;
+}
+
+/**
+ * pdata_to_blkg - get blkg associated with policy private data
+ * @pd: policy private data of interest
+ *
+ * @pd is policy private data.  Determine the blkg it's associated with.
+ */
+static inline struct blkcg_gq *pd_to_blkg(struct blkg_policy_data *pd)
+{
+	return pd ? pd->blkg : NULL;
+}
+
+/**
+ * blkg_path - format cgroup path of blkg
+ * @blkg: blkg of interest
+ * @buf: target buffer
+ * @buflen: target buffer length
+ *
+ * Format the path of the cgroup of @blkg into @buf.
+ */
+static inline int blkg_path(struct blkcg_gq *blkg, char *buf, int buflen)
+{
+	int ret;
+
+	rcu_read_lock();
+	ret = cgroup_path(blkg->blkcg->css.cgroup, buf, buflen);
+	rcu_read_unlock();
+	if (ret)
+		strncpy(buf, "<unavailable>", buflen);
+	return ret;
+}
 
-static inline char *blkg_path(struct blkio_group *blkg)
+/**
+ * blkg_get - get a blkg reference
+ * @blkg: blkg to get
+ *
+ * The caller should be holding queue_lock and an existing reference.
+ */
+static inline void blkg_get(struct blkcg_gq *blkg)
 {
-	return blkg->path;
+	lockdep_assert_held(blkg->q->queue_lock);
+	WARN_ON_ONCE(!blkg->refcnt);
+	blkg->refcnt++;
 }
 
-#else
+void __blkg_release(struct blkcg_gq *blkg);
 
-struct blkio_group {
+/**
+ * blkg_put - put a blkg reference
+ * @blkg: blkg to put
+ *
+ * The caller should be holding queue_lock.
+ */
+static inline void blkg_put(struct blkcg_gq *blkg)
+{
+	lockdep_assert_held(blkg->q->queue_lock);
+	WARN_ON_ONCE(blkg->refcnt <= 0);
+	if (!--blkg->refcnt)
+		__blkg_release(blkg);
+}
+
+/**
+ * blkg_stat_add - add a value to a blkg_stat
+ * @stat: target blkg_stat
+ * @val: value to add
+ *
+ * Add @val to @stat.  The caller is responsible for synchronizing calls to
+ * this function.
+ */
+static inline void blkg_stat_add(struct blkg_stat *stat, uint64_t val)
+{
+	u64_stats_update_begin(&stat->syncp);
+	stat->cnt += val;
+	u64_stats_update_end(&stat->syncp);
+}
+
+/**
+ * blkg_stat_read - read the current value of a blkg_stat
+ * @stat: blkg_stat to read
+ *
+ * Read the current value of @stat.  This function can be called without
+ * synchroniztion and takes care of u64 atomicity.
+ */
+static inline uint64_t blkg_stat_read(struct blkg_stat *stat)
+{
+	unsigned int start;
+	uint64_t v;
+
+	do {
+		start = u64_stats_fetch_begin(&stat->syncp);
+		v = stat->cnt;
+	} while (u64_stats_fetch_retry(&stat->syncp, start));
+
+	return v;
+}
+
+/**
+ * blkg_stat_reset - reset a blkg_stat
+ * @stat: blkg_stat to reset
+ */
+static inline void blkg_stat_reset(struct blkg_stat *stat)
+{
+	stat->cnt = 0;
+}
+
+/**
+ * blkg_rwstat_add - add a value to a blkg_rwstat
+ * @rwstat: target blkg_rwstat
+ * @rw: mask of REQ_{WRITE|SYNC}
+ * @val: value to add
+ *
+ * Add @val to @rwstat.  The counters are chosen according to @rw.  The
+ * caller is responsible for synchronizing calls to this function.
+ */
+static inline void blkg_rwstat_add(struct blkg_rwstat *rwstat,
+				   int rw, uint64_t val)
+{
+	u64_stats_update_begin(&rwstat->syncp);
+
+	if (rw & REQ_WRITE)
+		rwstat->cnt[BLKG_RWSTAT_WRITE] += val;
+	else
+		rwstat->cnt[BLKG_RWSTAT_READ] += val;
+	if (rw & REQ_SYNC)
+		rwstat->cnt[BLKG_RWSTAT_SYNC] += val;
+	else
+		rwstat->cnt[BLKG_RWSTAT_ASYNC] += val;
+
+	u64_stats_update_end(&rwstat->syncp);
+}
+
+/**
+ * blkg_rwstat_read - read the current values of a blkg_rwstat
+ * @rwstat: blkg_rwstat to read
+ *
+ * Read the current snapshot of @rwstat and return it as the return value.
+ * This function can be called without synchronization and takes care of
+ * u64 atomicity.
+ */
+static inline struct blkg_rwstat blkg_rwstat_read(struct blkg_rwstat *rwstat)
+{
+	unsigned int start;
+	struct blkg_rwstat tmp;
+
+	do {
+		start = u64_stats_fetch_begin(&rwstat->syncp);
+		tmp = *rwstat;
+	} while (u64_stats_fetch_retry(&rwstat->syncp, start));
+
+	return tmp;
+}
+
+/**
+ * blkg_rwstat_sum - read the total count of a blkg_rwstat
+ * @rwstat: blkg_rwstat to read
+ *
+ * Return the total count of @rwstat regardless of the IO direction.  This
+ * function can be called without synchronization and takes care of u64
+ * atomicity.
+ */
+static inline uint64_t blkg_rwstat_sum(struct blkg_rwstat *rwstat)
+{
+	struct blkg_rwstat tmp = blkg_rwstat_read(rwstat);
+
+	return tmp.cnt[BLKG_RWSTAT_READ] + tmp.cnt[BLKG_RWSTAT_WRITE];
+}
+
+/**
+ * blkg_rwstat_reset - reset a blkg_rwstat
+ * @rwstat: blkg_rwstat to reset
+ */
+static inline void blkg_rwstat_reset(struct blkg_rwstat *rwstat)
+{
+	memset(rwstat->cnt, 0, sizeof(rwstat->cnt));
+}
+
+#else	/* CONFIG_BLK_CGROUP */
+
+struct cgroup;
+
+struct blkg_policy_data {
 };
 
-struct blkio_policy_type {
+struct blkcg_gq {
 };
 
-static inline void blkio_policy_register(struct blkio_policy_type *blkiop) { }
-static inline void blkio_policy_unregister(struct blkio_policy_type *blkiop) { }
-
-static inline char *blkg_path(struct blkio_group *blkg) { return NULL; }
-
-#endif
-
-#define BLKIO_WEIGHT_MIN	10
-#define BLKIO_WEIGHT_MAX	1000
-#define BLKIO_WEIGHT_DEFAULT	500
-
-#ifdef CONFIG_DEBUG_BLK_CGROUP
-void blkiocg_update_avg_queue_size_stats(struct blkio_group *blkg);
-void blkiocg_update_dequeue_stats(struct blkio_group *blkg,
-				unsigned long dequeue);
-void blkiocg_update_set_idle_time_stats(struct blkio_group *blkg);
-void blkiocg_update_idle_time_stats(struct blkio_group *blkg);
-void blkiocg_set_start_empty_time(struct blkio_group *blkg);
-
-#define BLKG_FLAG_FNS(name)						\
-static inline void blkio_mark_blkg_##name(				\
-		struct blkio_group_stats *stats)			\
-{									\
-	stats->flags |= (1 << BLKG_##name);				\
-}									\
-static inline void blkio_clear_blkg_##name(				\
-		struct blkio_group_stats *stats)			\
-{									\
-	stats->flags &= ~(1 << BLKG_##name);				\
-}									\
-static inline int blkio_blkg_##name(struct blkio_group_stats *stats)	\
-{									\
-	return (stats->flags & (1 << BLKG_##name)) != 0;		\
-}									\
-
-BLKG_FLAG_FNS(waiting)
-BLKG_FLAG_FNS(idling)
-BLKG_FLAG_FNS(empty)
-#undef BLKG_FLAG_FNS
-#else
-static inline void blkiocg_update_avg_queue_size_stats(
-						struct blkio_group *blkg) {}
-static inline void blkiocg_update_dequeue_stats(struct blkio_group *blkg,
-						unsigned long dequeue) {}
-static inline void blkiocg_update_set_idle_time_stats(struct blkio_group *blkg)
-{}
-static inline void blkiocg_update_idle_time_stats(struct blkio_group *blkg) {}
-static inline void blkiocg_set_start_empty_time(struct blkio_group *blkg) {}
-#endif
-
-#if defined(CONFIG_BLK_CGROUP) || defined(CONFIG_BLK_CGROUP_MODULE)
-extern struct blkio_cgroup blkio_root_cgroup;
-extern struct blkio_cgroup *cgroup_to_blkio_cgroup(struct cgroup *cgroup);
-extern struct blkio_cgroup *task_blkio_cgroup(struct task_struct *tsk);
-extern void blkiocg_add_blkio_group(struct blkio_cgroup *blkcg,
-	struct blkio_group *blkg, void *key, dev_t dev,
-	enum blkio_policy_id plid);
-extern int blkio_alloc_blkg_stats(struct blkio_group *blkg);
-extern int blkiocg_del_blkio_group(struct blkio_group *blkg);
-extern struct blkio_group *blkiocg_lookup_group(struct blkio_cgroup *blkcg,
-						void *key);
-void blkiocg_update_timeslice_used(struct blkio_group *blkg,
-					unsigned long time,
-					unsigned long unaccounted_time);
-void blkiocg_update_dispatch_stats(struct blkio_group *blkg, uint64_t bytes,
-						bool direction, bool sync);
-void blkiocg_update_completion_stats(struct blkio_group *blkg,
-	uint64_t start_time, uint64_t io_start_time, bool direction, bool sync);
-void blkiocg_update_io_merged_stats(struct blkio_group *blkg, bool direction,
-					bool sync);
-void blkiocg_update_io_add_stats(struct blkio_group *blkg,
-		struct blkio_group *curr_blkg, bool direction, bool sync);
-void blkiocg_update_io_remove_stats(struct blkio_group *blkg,
-					bool direction, bool sync);
-#else
-struct cgroup;
-static inline struct blkio_cgroup *
-cgroup_to_blkio_cgroup(struct cgroup *cgroup) { return NULL; }
-static inline struct blkio_cgroup *
-task_blkio_cgroup(struct task_struct *tsk) { return NULL; }
-
-static inline void blkiocg_add_blkio_group(struct blkio_cgroup *blkcg,
-		struct blkio_group *blkg, void *key, dev_t dev,
-		enum blkio_policy_id plid) {}
-
-static inline int blkio_alloc_blkg_stats(struct blkio_group *blkg) { return 0; }
-
-static inline int
-blkiocg_del_blkio_group(struct blkio_group *blkg) { return 0; }
-
-static inline struct blkio_group *
-blkiocg_lookup_group(struct blkio_cgroup *blkcg, void *key) { return NULL; }
-static inline void blkiocg_update_timeslice_used(struct blkio_group *blkg,
-						unsigned long time,
-						unsigned long unaccounted_time)
-{}
-static inline void blkiocg_update_dispatch_stats(struct blkio_group *blkg,
-				uint64_t bytes, bool direction, bool sync) {}
-static inline void blkiocg_update_completion_stats(struct blkio_group *blkg,
-		uint64_t start_time, uint64_t io_start_time, bool direction,
-		bool sync) {}
-static inline void blkiocg_update_io_merged_stats(struct blkio_group *blkg,
-						bool direction, bool sync) {}
-static inline void blkiocg_update_io_add_stats(struct blkio_group *blkg,
-		struct blkio_group *curr_blkg, bool direction, bool sync) {}
-static inline void blkiocg_update_io_remove_stats(struct blkio_group *blkg,
-						bool direction, bool sync) {}
-#endif
-#endif /* _BLK_CGROUP_H */
+struct blkcg_policy {
+};
+
+static inline struct blkcg *cgroup_to_blkcg(struct cgroup *cgroup) { return NULL; }
+static inline struct blkcg *bio_blkcg(struct bio *bio) { return NULL; }
+static inline struct blkcg_gq *blkg_lookup(struct blkcg *blkcg, void *key) { return NULL; }
+static inline int blkcg_init_queue(struct request_queue *q) { return 0; }
+static inline void blkcg_drain_queue(struct request_queue *q) { }
+static inline void blkcg_exit_queue(struct request_queue *q) { }
+static inline int blkcg_policy_register(struct blkcg_policy *pol) { return 0; }
+static inline void blkcg_policy_unregister(struct blkcg_policy *pol) { }
+static inline int blkcg_activate_policy(struct request_queue *q,
+					const struct blkcg_policy *pol) { return 0; }
+static inline void blkcg_deactivate_policy(struct request_queue *q,
+					   const struct blkcg_policy *pol) { }
+
+static inline struct blkg_policy_data *blkg_to_pd(struct blkcg_gq *blkg,
+						  struct blkcg_policy *pol) { return NULL; }
+static inline struct blkcg_gq *pd_to_blkg(struct blkg_policy_data *pd) { return NULL; }
+static inline char *blkg_path(struct blkcg_gq *blkg) { return NULL; }
+static inline void blkg_get(struct blkcg_gq *blkg) { }
+static inline void blkg_put(struct blkcg_gq *blkg) { }
+
+#endif	/* CONFIG_BLK_CGROUP */
+#endif	/* _BLK_CGROUP_H */

block/blk-core.c

@@ -29,11 +29,13 @@
 #include <linux/fault-inject.h>
 #include <linux/list_sort.h>
 #include <linux/delay.h>
+#include <linux/ratelimit.h>
 
 #define CREATE_TRACE_POINTS
 #include <trace/events/block.h>
 
 #include "blk.h"
+#include "blk-cgroup.h"
 
 EXPORT_TRACEPOINT_SYMBOL_GPL(block_bio_remap);
 EXPORT_TRACEPOINT_SYMBOL_GPL(block_rq_remap);
@@ -280,7 +282,7 @@ EXPORT_SYMBOL(blk_stop_queue);
  *
  *     This function does not cancel any asynchronous activity arising
  *     out of elevator or throttling code. That would require elevaotor_exit()
- *     and blk_throtl_exit() to be called with queue lock initialized.
+ *     and blkcg_exit_queue() to be called with queue lock initialized.
  *
  */
 void blk_sync_queue(struct request_queue *q)
@@ -365,17 +367,23 @@ void blk_drain_queue(struct request_queue *q, bool drain_all)
 
 		spin_lock_irq(q->queue_lock);
 
-		elv_drain_elevator(q);
-		if (drain_all)
-			blk_throtl_drain(q);
+		/*
+		 * The caller might be trying to drain @q before its
+		 * elevator is initialized.
+		 */
+		if (q->elevator)
+			elv_drain_elevator(q);
+
+		blkcg_drain_queue(q);
 
 		/*
 		 * This function might be called on a queue which failed
-		 * driver init after queue creation.  Some drivers
-		 * (e.g. fd) get unhappy in such cases.  Kick queue iff
-		 * dispatch queue has something on it.
+		 * driver init after queue creation or is not yet fully
+		 * active yet.  Some drivers (e.g. fd and loop) get unhappy
+		 * in such cases.  Kick queue iff dispatch queue has
+		 * something on it and @q has request_fn set.
 		 */
-		if (!list_empty(&q->queue_head))
+		if (!list_empty(&q->queue_head) && q->request_fn)
 			__blk_run_queue(q);
 
 		drain |= q->rq.elvpriv;
@@ -402,6 +410,49 @@ void blk_drain_queue(struct request_queue *q, bool drain_all)
 	}
 }
 
+/**
+ * blk_queue_bypass_start - enter queue bypass mode
+ * @q: queue of interest
+ *
+ * In bypass mode, only the dispatch FIFO queue of @q is used.  This
+ * function makes @q enter bypass mode and drains all requests which were
+ * throttled or issued before.  On return, it's guaranteed that no request
+ * is being throttled or has ELVPRIV set and blk_queue_bypass() %true
+ * inside queue or RCU read lock.
+ */
+void blk_queue_bypass_start(struct request_queue *q)
+{
+	bool drain;
+
+	spin_lock_irq(q->queue_lock);
+	drain = !q->bypass_depth++;
+	queue_flag_set(QUEUE_FLAG_BYPASS, q);
+	spin_unlock_irq(q->queue_lock);
+
+	if (drain) {
+		blk_drain_queue(q, false);
+		/* ensure blk_queue_bypass() is %true inside RCU read lock */
+		synchronize_rcu();
+	}
+}
+EXPORT_SYMBOL_GPL(blk_queue_bypass_start);
+
+/**
+ * blk_queue_bypass_end - leave queue bypass mode
+ * @q: queue of interest
+ *
+ * Leave bypass mode and restore the normal queueing behavior.
+ */
+void blk_queue_bypass_end(struct request_queue *q)
+{
+	spin_lock_irq(q->queue_lock);
+	if (!--q->bypass_depth)
+		queue_flag_clear(QUEUE_FLAG_BYPASS, q);
+	WARN_ON_ONCE(q->bypass_depth < 0);
+	spin_unlock_irq(q->queue_lock);
+}
+EXPORT_SYMBOL_GPL(blk_queue_bypass_end);
+
 /**
  * blk_cleanup_queue - shutdown a request queue
  * @q: request queue to shutdown
@@ -418,6 +469,19 @@ void blk_cleanup_queue(struct request_queue *q)
 	queue_flag_set_unlocked(QUEUE_FLAG_DEAD, q);
 
 	spin_lock_irq(lock);
+
+	/*
+	 * Dead queue is permanently in bypass mode till released.  Note
+	 * that, unlike blk_queue_bypass_start(), we aren't performing
+	 * synchronize_rcu() after entering bypass mode to avoid the delay
+	 * as some drivers create and destroy a lot of queues while
+	 * probing.  This is still safe because blk_release_queue() will be
+	 * called only after the queue refcnt drops to zero and nothing,
+	 * RCU or not, would be traversing the queue by then.
+	 */
+	q->bypass_depth++;
+	queue_flag_set(QUEUE_FLAG_BYPASS, q);
+
 	queue_flag_set(QUEUE_FLAG_NOMERGES, q);
 	queue_flag_set(QUEUE_FLAG_NOXMERGES, q);
 	queue_flag_set(QUEUE_FLAG_DEAD, q);
@@ -428,13 +492,8 @@ void blk_cleanup_queue(struct request_queue *q)
 	spin_unlock_irq(lock);
 	mutex_unlock(&q->sysfs_lock);
 
-	/*
-	 * Drain all requests queued before DEAD marking.  The caller might
-	 * be trying to tear down @q before its elevator is initialized, in
-	 * which case we don't want to call into draining.
-	 */
-	if (q->elevator)
-		blk_drain_queue(q, true);
+	/* drain all requests queued before DEAD marking */
+	blk_drain_queue(q, true);
 
 	/* @q won't process any more request, flush async actions */
 	del_timer_sync(&q->backing_dev_info.laptop_mode_wb_timer);
@@ -498,14 +557,15 @@ struct request_queue *blk_alloc_queue_node(gfp_t gfp_mask, int node_id)
 	if (err)
 		goto fail_id;
 
-	if (blk_throtl_init(q))
-		goto fail_id;
-
 	setup_timer(&q->backing_dev_info.laptop_mode_wb_timer,
 		    laptop_mode_timer_fn, (unsigned long) q);
 	setup_timer(&q->timeout, blk_rq_timed_out_timer, (unsigned long) q);
+	INIT_LIST_HEAD(&q->queue_head);
 	INIT_LIST_HEAD(&q->timeout_list);
 	INIT_LIST_HEAD(&q->icq_list);
+#ifdef CONFIG_BLK_CGROUP
+	INIT_LIST_HEAD(&q->blkg_list);
+#endif
 	INIT_LIST_HEAD(&q->flush_queue[0]);
 	INIT_LIST_HEAD(&q->flush_queue[1]);
 	INIT_LIST_HEAD(&q->flush_data_in_flight);
@@ -522,6 +582,18 @@ struct request_queue *blk_alloc_queue_node(gfp_t gfp_mask, int node_id)
 	 */
 	q->queue_lock = &q->__queue_lock;
 
+	/*
+	 * A queue starts its life with bypass turned on to avoid
+	 * unnecessary bypass on/off overhead and nasty surprises during
+	 * init.  The initial bypass will be finished at the end of
+	 * blk_init_allocated_queue().
+	 */
+	q->bypass_depth = 1;
+	__set_bit(QUEUE_FLAG_BYPASS, &q->queue_flags);
+
+	if (blkcg_init_queue(q))
+		goto fail_id;
+
 	return q;
 
 fail_id:
@@ -614,15 +686,15 @@ blk_init_allocated_queue(struct request_queue *q, request_fn_proc *rfn,
 
 	q->sg_reserved_size = INT_MAX;
 
-	/*
-	 * all done
-	 */
-	if (!elevator_init(q, NULL)) {
-		blk_queue_congestion_threshold(q);
-		return q;
-	}
+	/* init elevator */
+	if (elevator_init(q, NULL))
+		return NULL;
 
-	return NULL;
+	blk_queue_congestion_threshold(q);
+
+	/* all done, end the initial bypass */
+	blk_queue_bypass_end(q);
+	return q;
 }
 EXPORT_SYMBOL(blk_init_allocated_queue);
 
@@ -648,33 +720,6 @@ static inline void blk_free_request(struct request_queue *q, struct request *rq)
 	mempool_free(rq, q->rq.rq_pool);
 }
 
-static struct request *
-blk_alloc_request(struct request_queue *q, struct io_cq *icq,
-		  unsigned int flags, gfp_t gfp_mask)
-{
-	struct request *rq = mempool_alloc(q->rq.rq_pool, gfp_mask);
-
-	if (!rq)
-		return NULL;
-
-	blk_rq_init(q, rq);
-
-	rq->cmd_flags = flags | REQ_ALLOCED;
-
-	if (flags & REQ_ELVPRIV) {
-		rq->elv.icq = icq;
-		if (unlikely(elv_set_request(q, rq, gfp_mask))) {
-			mempool_free(rq, q->rq.rq_pool);
-			return NULL;
-		}
-		/* @rq->elv.icq holds on to io_context until @rq is freed */
-		if (icq)
-			get_io_context(icq->ioc);
-	}
-
-	return rq;
-}
-
 /*
  * ioc_batching returns true if the ioc is a valid batching request and
  * should be given priority access to a request.
@@ -762,6 +807,22 @@ static bool blk_rq_should_init_elevator(struct bio *bio)
 	return true;
 }
 
+/**
+ * rq_ioc - determine io_context for request allocation
+ * @bio: request being allocated is for this bio (can be %NULL)
+ *
+ * Determine io_context to use for request allocation for @bio.  May return
+ * %NULL if %current->io_context doesn't exist.
+ */
+static struct io_context *rq_ioc(struct bio *bio)
+{
+#ifdef CONFIG_BLK_CGROUP
+	if (bio && bio->bi_ioc)
+		return bio->bi_ioc;
+#endif
+	return current->io_context;
+}
+
 /**
  * get_request - get a free request
  * @q: request_queue to allocate request from
@@ -779,7 +840,7 @@ static bool blk_rq_should_init_elevator(struct bio *bio)
 static struct request *get_request(struct request_queue *q, int rw_flags,
 				   struct bio *bio, gfp_t gfp_mask)
 {
-	struct request *rq = NULL;
+	struct request *rq;
 	struct request_list *rl = &q->rq;
 	struct elevator_type *et;
 	struct io_context *ioc;
@@ -789,7 +850,7 @@ static struct request *get_request(struct request_queue *q, int rw_flags,
 	int may_queue;
 retry:
 	et = q->elevator->type;
-	ioc = current->io_context;
+	ioc = rq_ioc(bio);
 
 	if (unlikely(blk_queue_dead(q)))
 		return NULL;
@@ -808,7 +869,7 @@ retry:
 			 */
 			if (!ioc && !retried) {
 				spin_unlock_irq(q->queue_lock);
-				create_io_context(current, gfp_mask, q->node);
+				create_io_context(gfp_mask, q->node);
 				spin_lock_irq(q->queue_lock);
 				retried = true;
 				goto retry;
@@ -831,7 +892,7 @@ retry:
 					 * process is not a "batcher", and not
 					 * exempted by the IO scheduler
 					 */
-					goto out;
+					return NULL;
 				}
 			}
 		}
@@ -844,7 +905,7 @@ retry:
 	 * allocated with any setting of ->nr_requests
 	 */
 	if (rl->count[is_sync] >= (3 * q->nr_requests / 2))
-		goto out;
+		return NULL;
 
 	rl->count[is_sync]++;
 	rl->starved[is_sync] = 0;
@@ -859,8 +920,7 @@ retry:
 	 * Also, lookup icq while holding queue_lock.  If it doesn't exist,
 	 * it will be created after releasing queue_lock.
 	 */
-	if (blk_rq_should_init_elevator(bio) &&
-	    !test_bit(QUEUE_FLAG_ELVSWITCH, &q->queue_flags)) {
+	if (blk_rq_should_init_elevator(bio) && !blk_queue_bypass(q)) {
 		rw_flags |= REQ_ELVPRIV;
 		rl->elvpriv++;
 		if (et->icq_cache && ioc)
@@ -871,41 +931,36 @@ retry:
 		rw_flags |= REQ_IO_STAT;
 	spin_unlock_irq(q->queue_lock);
 
-	/* create icq if missing */
-	if ((rw_flags & REQ_ELVPRIV) && unlikely(et->icq_cache && !icq)) {
-		icq = ioc_create_icq(q, gfp_mask);
-		if (!icq)
-			goto fail_icq;
-	}
-
-	rq = blk_alloc_request(q, icq, rw_flags, gfp_mask);
+	/* allocate and init request */
+	rq = mempool_alloc(q->rq.rq_pool, gfp_mask);
+	if (!rq)
+		goto fail_alloc;
 
-fail_icq:
-	if (unlikely(!rq)) {
-		/*
-		 * Allocation failed presumably due to memory. Undo anything
-		 * we might have messed up.
-		 *
-		 * Allocating task should really be put onto the front of the
-		 * wait queue, but this is pretty rare.
-		 */
-		spin_lock_irq(q->queue_lock);
-		freed_request(q, rw_flags);
+	blk_rq_init(q, rq);
+	rq->cmd_flags = rw_flags | REQ_ALLOCED;
+
+	/* init elvpriv */
+	if (rw_flags & REQ_ELVPRIV) {
+		if (unlikely(et->icq_cache && !icq)) {
+			create_io_context(gfp_mask, q->node);
+			ioc = rq_ioc(bio);
+			if (!ioc)
+				goto fail_elvpriv;
+
+			icq = ioc_create_icq(ioc, q, gfp_mask);
+			if (!icq)
+				goto fail_elvpriv;
+		}
 
-		/*
-		 * in the very unlikely event that allocation failed and no
-		 * requests for this direction was pending, mark us starved
-		 * so that freeing of a request in the other direction will
-		 * notice us. another possible fix would be to split the
-		 * rq mempool into READ and WRITE
-		 */
-rq_starved:
-		if (unlikely(rl->count[is_sync] == 0))
-			rl->starved[is_sync] = 1;
+		rq->elv.icq = icq;
+		if (unlikely(elv_set_request(q, rq, bio, gfp_mask)))
+			goto fail_elvpriv;
 
-		goto out;
+		/* @rq->elv.icq holds io_context until @rq is freed */
+		if (icq)
+			get_io_context(icq->ioc);
 	}
-
+out:
 	/*
 	 * ioc may be NULL here, and ioc_batching will be false. That's
 	 * OK, if the queue is under the request limit then requests need
@@ -916,8 +971,48 @@ rq_starved:
 		ioc->nr_batch_requests--;
 
 	trace_block_getrq(q, bio, rw_flags & 1);
-out:
 	return rq;
+
+fail_elvpriv:
+	/*
+	 * elvpriv init failed.  ioc, icq and elvpriv aren't mempool backed
+	 * and may fail indefinitely under memory pressure and thus
+	 * shouldn't stall IO.  Treat this request as !elvpriv.  This will
+	 * disturb iosched and blkcg but weird is bettern than dead.
+	 */
+	printk_ratelimited(KERN_WARNING "%s: request aux data allocation failed, iosched may be disturbed\n",
+			   dev_name(q->backing_dev_info.dev));
+
+	rq->cmd_flags &= ~REQ_ELVPRIV;
+	rq->elv.icq = NULL;
+
+	spin_lock_irq(q->queue_lock);
+	rl->elvpriv--;
+	spin_unlock_irq(q->queue_lock);
+	goto out;
+
+fail_alloc:
+	/*
+	 * Allocation failed presumably due to memory. Undo anything we
+	 * might have messed up.
+	 *
+	 * Allocating task should really be put onto the front of the wait
+	 * queue, but this is pretty rare.
+	 */
+	spin_lock_irq(q->queue_lock);
+	freed_request(q, rw_flags);
+
+	/*
+	 * in the very unlikely event that allocation failed and no
+	 * requests for this direction was pending, mark us starved so that
+	 * freeing of a request in the other direction will notice
+	 * us. another possible fix would be to split the rq mempool into
+	 * READ and WRITE
+	 */
+rq_starved:
+	if (unlikely(rl->count[is_sync] == 0))
+		rl->starved[is_sync] = 1;
+	return NULL;
 }
 
 /**
@@ -961,7 +1056,7 @@ static struct request *get_request_wait(struct request_queue *q, int rw_flags,
 		 * up to a big batch of them for a small period time.
 		 * See ioc_batching, ioc_set_batching
 		 */
-		create_io_context(current, GFP_NOIO, q->node);
+		create_io_context(GFP_NOIO, q->node);
 		ioc_set_batching(q, current->io_context);
 
 		spin_lock_irq(q->queue_lock);

block/blk-ioc.c

@@ -155,20 +155,20 @@ void put_io_context(struct io_context *ioc)
 }
 EXPORT_SYMBOL(put_io_context);
 
-/* Called by the exiting task */
-void exit_io_context(struct task_struct *task)
+/**
+ * put_io_context_active - put active reference on ioc
+ * @ioc: ioc of interest
+ *
+ * Undo get_io_context_active().  If active reference reaches zero after
+ * put, @ioc can never issue further IOs and ioscheds are notified.
+ */
+void put_io_context_active(struct io_context *ioc)
 {
-	struct io_context *ioc;
-	struct io_cq *icq;
 	struct hlist_node *n;
 	unsigned long flags;
+	struct io_cq *icq;
 
-	task_lock(task);
-	ioc = task->io_context;
-	task->io_context = NULL;
-	task_unlock(task);
-
-	if (!atomic_dec_and_test(&ioc->nr_tasks)) {
+	if (!atomic_dec_and_test(&ioc->active_ref)) {
 		put_io_context(ioc);
 		return;
 	}
@@ -197,6 +197,20 @@ retry:
 	put_io_context(ioc);
 }
 
+/* Called by the exiting task */
+void exit_io_context(struct task_struct *task)
+{
+	struct io_context *ioc;
+
+	task_lock(task);
+	ioc = task->io_context;
+	task->io_context = NULL;
+	task_unlock(task);
+
+	atomic_dec(&ioc->nr_tasks);
+	put_io_context_active(ioc);
+}
+
 /**
  * ioc_clear_queue - break any ioc association with the specified queue
  * @q: request_queue being cleared
@@ -218,19 +232,18 @@ void ioc_clear_queue(struct request_queue *q)
 	}
 }
 
-void create_io_context_slowpath(struct task_struct *task, gfp_t gfp_flags,
-				int node)
+int create_task_io_context(struct task_struct *task, gfp_t gfp_flags, int node)
 {
 	struct io_context *ioc;
 
 	ioc = kmem_cache_alloc_node(iocontext_cachep, gfp_flags | __GFP_ZERO,
 				    node);
 	if (unlikely(!ioc))
-		return;
+		return -ENOMEM;
 
 	/* initialize */
 	atomic_long_set(&ioc->refcount, 1);
-	atomic_set(&ioc->nr_tasks, 1);
+	atomic_set(&ioc->active_ref, 1);
 	spin_lock_init(&ioc->lock);
 	INIT_RADIX_TREE(&ioc->icq_tree, GFP_ATOMIC | __GFP_HIGH);
 	INIT_HLIST_HEAD(&ioc->icq_list);
@@ -250,6 +263,8 @@ void create_io_context_slowpath(struct task_struct *task, gfp_t gfp_flags,
 	else
 		kmem_cache_free(iocontext_cachep, ioc);
 	task_unlock(task);
+
+	return 0;
 }
 
 /**
@@ -281,7 +296,7 @@ struct io_context *get_task_io_context(struct task_struct *task,
 			return ioc;
 		}
 		task_unlock(task);
-	} while (create_io_context(task, gfp_flags, node));
+	} while (!create_task_io_context(task, gfp_flags, node));
 
 	return NULL;
 }
@@ -325,26 +340,23 @@ EXPORT_SYMBOL(ioc_lookup_icq);
 
 /**
  * ioc_create_icq - create and link io_cq
+ * @ioc: io_context of interest
  * @q: request_queue of interest
  * @gfp_mask: allocation mask
  *
- * Make sure io_cq linking %current->io_context and @q exists.  If either
- * io_context and/or icq don't exist, they will be created using @gfp_mask.
+ * Make sure io_cq linking @ioc and @q exists.  If icq doesn't exist, they
+ * will be created using @gfp_mask.
  *
  * The caller is responsible for ensuring @ioc won't go away and @q is
  * alive and will stay alive until this function returns.
  */
-struct io_cq *ioc_create_icq(struct request_queue *q, gfp_t gfp_mask)
+struct io_cq *ioc_create_icq(struct io_context *ioc, struct request_queue *q,
+			     gfp_t gfp_mask)
 {
 	struct elevator_type *et = q->elevator->type;
-	struct io_context *ioc;
 	struct io_cq *icq;
 
 	/* allocate stuff */
-	ioc = create_io_context(current, gfp_mask, q->node);
-	if (!ioc)
-		return NULL;
-
 	icq = kmem_cache_alloc_node(et->icq_cache, gfp_mask | __GFP_ZERO,
 				    q->node);
 	if (!icq)
@@ -382,74 +394,6 @@ struct io_cq *ioc_create_icq(struct request_queue *q, gfp_t gfp_mask)
 	return icq;
 }
 
-void ioc_set_icq_flags(struct io_context *ioc, unsigned int flags)
-{
-	struct io_cq *icq;
-	struct hlist_node *n;
-
-	hlist_for_each_entry(icq, n, &ioc->icq_list, ioc_node)
-		icq->flags |= flags;
-}
-
-/**
- * ioc_ioprio_changed - notify ioprio change
- * @ioc: io_context of interest
- * @ioprio: new ioprio
- *
- * @ioc's ioprio has changed to @ioprio.  Set %ICQ_IOPRIO_CHANGED for all
- * icq's.  iosched is responsible for checking the bit and applying it on
- * request issue path.
- */
-void ioc_ioprio_changed(struct io_context *ioc, int ioprio)
-{
-	unsigned long flags;
-
-	spin_lock_irqsave(&ioc->lock, flags);
-	ioc->ioprio = ioprio;
-	ioc_set_icq_flags(ioc, ICQ_IOPRIO_CHANGED);
-	spin_unlock_irqrestore(&ioc->lock, flags);
-}
-
-/**
- * ioc_cgroup_changed - notify cgroup change
- * @ioc: io_context of interest
- *
- * @ioc's cgroup has changed.  Set %ICQ_CGROUP_CHANGED for all icq's.
- * iosched is responsible for checking the bit and applying it on request
- * issue path.
- */
-void ioc_cgroup_changed(struct io_context *ioc)
-{
-	unsigned long flags;
-
-	spin_lock_irqsave(&ioc->lock, flags);
-	ioc_set_icq_flags(ioc, ICQ_CGROUP_CHANGED);
-	spin_unlock_irqrestore(&ioc->lock, flags);
-}
-EXPORT_SYMBOL(ioc_cgroup_changed);
-
-/**
- * icq_get_changed - fetch and clear icq changed mask
- * @icq: icq of interest
- *
- * Fetch and clear ICQ_*_CHANGED bits from @icq.  Grabs and releases
- * @icq->ioc->lock.
- */
-unsigned icq_get_changed(struct io_cq *icq)
-{
-	unsigned int changed = 0;
-	unsigned long flags;
-
-	if (unlikely(icq->flags & ICQ_CHANGED_MASK)) {
-		spin_lock_irqsave(&icq->ioc->lock, flags);
-		changed = icq->flags & ICQ_CHANGED_MASK;
-		icq->flags &= ~ICQ_CHANGED_MASK;
-		spin_unlock_irqrestore(&icq->ioc->lock, flags);
-	}
-	return changed;
-}
-EXPORT_SYMBOL(icq_get_changed);
-
 static int __init blk_ioc_init(void)
 {
 	iocontext_cachep = kmem_cache_create("blkdev_ioc",

block/blk-sysfs.c

@@ -9,6 +9,7 @@
 #include <linux/blktrace_api.h>
 
 #include "blk.h"
+#include "blk-cgroup.h"
 
 struct queue_sysfs_entry {
 	struct attribute attr;
@@ -479,6 +480,8 @@ static void blk_release_queue(struct kobject *kobj)
 
 	blk_sync_queue(q);
 
+	blkcg_exit_queue(q);
+
 	if (q->elevator) {
 		spin_lock_irq(q->queue_lock);
 		ioc_clear_queue(q);
@@ -486,15 +489,12 @@ static void blk_release_queue(struct kobject *kobj)
 		elevator_exit(q->elevator);
 	}
 
-	blk_throtl_exit(q);
-
 	if (rl->rq_pool)
 		mempool_destroy(rl->rq_pool);
 
 	if (q->queue_tags)
 		__blk_queue_free_tags(q);
 
-	blk_throtl_release(q);
 	blk_trace_shutdown(q);
 
 	bdi_destroy(&q->backing_dev_info);

block/blk-throttle.c

@@ -21,6 +21,8 @@ static int throtl_quantum = 32;
 /* Throttling is performed over 100ms slice and after that slice is renewed */
 static unsigned long throtl_slice = HZ/10;	/* 100 ms */
 
+static struct blkcg_policy blkcg_policy_throtl;
+
 /* A workqueue to queue throttle related work */
 static struct workqueue_struct *kthrotld_workqueue;
 static void throtl_schedule_delayed_work(struct throtl_data *td,
@@ -38,9 +40,17 @@ struct throtl_rb_root {
 
 #define rb_entry_tg(node)	rb_entry((node), struct throtl_grp, rb_node)
 
+/* Per-cpu group stats */
+struct tg_stats_cpu {
+	/* total bytes transferred */
+	struct blkg_rwstat		service_bytes;
+	/* total IOs serviced, post merge */
+	struct blkg_rwstat		serviced;
+};
+
 struct throtl_grp {
-	/* List of throtl groups on the request queue*/
-	struct hlist_node tg_node;
+	/* must be the first member */
+	struct blkg_policy_data pd;
 
 	/* active throtl group service_tree member */
 	struct rb_node rb_node;
@@ -52,8 +62,6 @@ struct throtl_grp {
 	 */
 	unsigned long disptime;
 
-	struct blkio_group blkg;
-	atomic_t ref;
 	unsigned int flags;
 
 	/* Two lists for READ and WRITE */
@@ -80,18 +88,18 @@ struct throtl_grp {
 	/* Some throttle limits got updated for the group */
 	int limits_changed;
 
-	struct rcu_head rcu_head;
+	/* Per cpu stats pointer */
+	struct tg_stats_cpu __percpu *stats_cpu;
+
+	/* List of tgs waiting for per cpu stats memory to be allocated */
+	struct list_head stats_alloc_node;
 };
 
 struct throtl_data
 {
-	/* List of throtl groups */
-	struct hlist_head tg_list;
-
 	/* service tree for active throtl groups */
 	struct throtl_rb_root tg_service_tree;
 
-	struct throtl_grp *root_tg;
 	struct request_queue *queue;
 
 	/* Total Number of queued bios on READ and WRITE lists */
@@ -108,6 +116,33 @@ struct throtl_data
 	int limits_changed;
 };
 
+/* list and work item to allocate percpu group stats */
+static DEFINE_SPINLOCK(tg_stats_alloc_lock);
+static LIST_HEAD(tg_stats_alloc_list);
+
+static void tg_stats_alloc_fn(struct work_struct *);
+static DECLARE_DELAYED_WORK(tg_stats_alloc_work, tg_stats_alloc_fn);
+
+static inline struct throtl_grp *pd_to_tg(struct blkg_policy_data *pd)
+{
+	return pd ? container_of(pd, struct throtl_grp, pd) : NULL;
+}
+
+static inline struct throtl_grp *blkg_to_tg(struct blkcg_gq *blkg)
+{
+	return pd_to_tg(blkg_to_pd(blkg, &blkcg_policy_throtl));
+}
+
+static inline struct blkcg_gq *tg_to_blkg(struct throtl_grp *tg)
+{
+	return pd_to_blkg(&tg->pd);
+}
+
+static inline struct throtl_grp *td_root_tg(struct throtl_data *td)
+{
+	return blkg_to_tg(td->queue->root_blkg);
+}
+
 enum tg_state_flags {
 	THROTL_TG_FLAG_on_rr = 0,	/* on round-robin busy list */
 };
@@ -128,244 +163,150 @@ static inline int throtl_tg_##name(const struct throtl_grp *tg)		\
 
 THROTL_TG_FNS(on_rr);
 
-#define throtl_log_tg(td, tg, fmt, args...)				\
-	blk_add_trace_msg((td)->queue, "throtl %s " fmt,		\
-				blkg_path(&(tg)->blkg), ##args);      	\
+#define throtl_log_tg(td, tg, fmt, args...)	do {			\
+	char __pbuf[128];						\
+									\
+	blkg_path(tg_to_blkg(tg), __pbuf, sizeof(__pbuf));		\
+	blk_add_trace_msg((td)->queue, "throtl %s " fmt, __pbuf, ##args); \
+} while (0)
 
 #define throtl_log(td, fmt, args...)	\
 	blk_add_trace_msg((td)->queue, "throtl " fmt, ##args)
 
-static inline struct throtl_grp *tg_of_blkg(struct blkio_group *blkg)
-{
-	if (blkg)
-		return container_of(blkg, struct throtl_grp, blkg);
-
-	return NULL;
-}
-
 static inline unsigned int total_nr_queued(struct throtl_data *td)
 {
 	return td->nr_queued[0] + td->nr_queued[1];
 }
 
-static inline struct throtl_grp *throtl_ref_get_tg(struct throtl_grp *tg)
-{
-	atomic_inc(&tg->ref);
-	return tg;
-}
-
-static void throtl_free_tg(struct rcu_head *head)
+/*
+ * Worker for allocating per cpu stat for tgs. This is scheduled on the
+ * system_nrt_wq once there are some groups on the alloc_list waiting for
+ * allocation.
+ */
+static void tg_stats_alloc_fn(struct work_struct *work)
 {
-	struct throtl_grp *tg;
+	static struct tg_stats_cpu *stats_cpu;	/* this fn is non-reentrant */
+	struct delayed_work *dwork = to_delayed_work(work);
+	bool empty = false;
+
+alloc_stats:
+	if (!stats_cpu) {
+		stats_cpu = alloc_percpu(struct tg_stats_cpu);
+		if (!stats_cpu) {
+			/* allocation failed, try again after some time */
+			queue_delayed_work(system_nrt_wq, dwork,
+					   msecs_to_jiffies(10));
+			return;
+		}
+	}
 
-	tg = container_of(head, struct throtl_grp, rcu_head);
-	free_percpu(tg->blkg.stats_cpu);
-	kfree(tg);
-}
+	spin_lock_irq(&tg_stats_alloc_lock);
 
-static void throtl_put_tg(struct throtl_grp *tg)
-{
-	BUG_ON(atomic_read(&tg->ref) <= 0);
-	if (!atomic_dec_and_test(&tg->ref))
-		return;
+	if (!list_empty(&tg_stats_alloc_list)) {
+		struct throtl_grp *tg = list_first_entry(&tg_stats_alloc_list,
+							 struct throtl_grp,
+							 stats_alloc_node);
+		swap(tg->stats_cpu, stats_cpu);
+		list_del_init(&tg->stats_alloc_node);
+	}
 
-	/*
-	 * A group is freed in rcu manner. But having an rcu lock does not
-	 * mean that one can access all the fields of blkg and assume these
-	 * are valid. For example, don't try to follow throtl_data and
-	 * request queue links.
-	 *
-	 * Having a reference to blkg under an rcu allows acess to only
-	 * values local to groups like group stats and group rate limits
-	 */
-	call_rcu(&tg->rcu_head, throtl_free_tg);
+	empty = list_empty(&tg_stats_alloc_list);
+	spin_unlock_irq(&tg_stats_alloc_lock);
+	if (!empty)
+		goto alloc_stats;
 }
 
-static void throtl_init_group(struct throtl_grp *tg)
+static void throtl_pd_init(struct blkcg_gq *blkg)
 {
-	INIT_HLIST_NODE(&tg->tg_node);
+	struct throtl_grp *tg = blkg_to_tg(blkg);
+	unsigned long flags;
+
 	RB_CLEAR_NODE(&tg->rb_node);
 	bio_list_init(&tg->bio_lists[0]);
 	bio_list_init(&tg->bio_lists[1]);
 	tg->limits_changed = false;
 
-	/* Practically unlimited BW */
-	tg->bps[0] = tg->bps[1] = -1;
-	tg->iops[0] = tg->iops[1] = -1;
+	tg->bps[READ] = -1;
+	tg->bps[WRITE] = -1;
+	tg->iops[READ] = -1;
+	tg->iops[WRITE] = -1;
 
 	/*
-	 * Take the initial reference that will be released on destroy
-	 * This can be thought of a joint reference by cgroup and
-	 * request queue which will be dropped by either request queue
-	 * exit or cgroup deletion path depending on who is exiting first.
+	 * Ugh... We need to perform per-cpu allocation for tg->stats_cpu
+	 * but percpu allocator can't be called from IO path.  Queue tg on
+	 * tg_stats_alloc_list and allocate from work item.
 	 */
-	atomic_set(&tg->ref, 1);
+	spin_lock_irqsave(&tg_stats_alloc_lock, flags);
+	list_add(&tg->stats_alloc_node, &tg_stats_alloc_list);
+	queue_delayed_work(system_nrt_wq, &tg_stats_alloc_work, 0);
+	spin_unlock_irqrestore(&tg_stats_alloc_lock, flags);
 }
 
-/* Should be called with rcu read lock held (needed for blkcg) */
-static void
-throtl_add_group_to_td_list(struct throtl_data *td, struct throtl_grp *tg)
+static void throtl_pd_exit(struct blkcg_gq *blkg)
 {
-	hlist_add_head(&tg->tg_node, &td->tg_list);
-	td->nr_undestroyed_grps++;
-}
-
-static void
-__throtl_tg_fill_dev_details(struct throtl_data *td, struct throtl_grp *tg)
-{
-	struct backing_dev_info *bdi = &td->queue->backing_dev_info;
-	unsigned int major, minor;
-
-	if (!tg || tg->blkg.dev)
-		return;
-
-	/*
-	 * Fill in device details for a group which might not have been
-	 * filled at group creation time as queue was being instantiated
-	 * and driver had not attached a device yet
-	 */
-	if (bdi->dev && dev_name(bdi->dev)) {
-		sscanf(dev_name(bdi->dev), "%u:%u", &major, &minor);
-		tg->blkg.dev = MKDEV(major, minor);
-	}
-}
-
-/*
- * Should be called with without queue lock held. Here queue lock will be
- * taken rarely. It will be taken only once during life time of a group
- * if need be
- */
-static void
-throtl_tg_fill_dev_details(struct throtl_data *td, struct throtl_grp *tg)
-{
-	if (!tg || tg->blkg.dev)
-		return;
-
-	spin_lock_irq(td->queue->queue_lock);
-	__throtl_tg_fill_dev_details(td, tg);
-	spin_unlock_irq(td->queue->queue_lock);
-}
-
-static void throtl_init_add_tg_lists(struct throtl_data *td,
-			struct throtl_grp *tg, struct blkio_cgroup *blkcg)
-{
-	__throtl_tg_fill_dev_details(td, tg);
-
-	/* Add group onto cgroup list */
-	blkiocg_add_blkio_group(blkcg, &tg->blkg, (void *)td,
-				tg->blkg.dev, BLKIO_POLICY_THROTL);
+	struct throtl_grp *tg = blkg_to_tg(blkg);
+	unsigned long flags;
 
-	tg->bps[READ] = blkcg_get_read_bps(blkcg, tg->blkg.dev);
-	tg->bps[WRITE] = blkcg_get_write_bps(blkcg, tg->blkg.dev);
-	tg->iops[READ] = blkcg_get_read_iops(blkcg, tg->blkg.dev);
-	tg->iops[WRITE] = blkcg_get_write_iops(blkcg, tg->blkg.dev);
+	spin_lock_irqsave(&tg_stats_alloc_lock, flags);
+	list_del_init(&tg->stats_alloc_node);
+	spin_unlock_irqrestore(&tg_stats_alloc_lock, flags);
 
-	throtl_add_group_to_td_list(td, tg);
+	free_percpu(tg->stats_cpu);
 }
 
-/* Should be called without queue lock and outside of rcu period */
-static struct throtl_grp *throtl_alloc_tg(struct throtl_data *td)
+static void throtl_pd_reset_stats(struct blkcg_gq *blkg)
 {
-	struct throtl_grp *tg = NULL;
-	int ret;
+	struct throtl_grp *tg = blkg_to_tg(blkg);
+	int cpu;
 
-	tg = kzalloc_node(sizeof(*tg), GFP_ATOMIC, td->queue->node);
-	if (!tg)
-		return NULL;
+	if (tg->stats_cpu == NULL)
+		return;
 
-	ret = blkio_alloc_blkg_stats(&tg->blkg);
+	for_each_possible_cpu(cpu) {
+		struct tg_stats_cpu *sc = per_cpu_ptr(tg->stats_cpu, cpu);
 
-	if (ret) {
-		kfree(tg);
-		return NULL;
+		blkg_rwstat_reset(&sc->service_bytes);
+		blkg_rwstat_reset(&sc->serviced);
 	}
-
-	throtl_init_group(tg);
-	return tg;
 }
 
-static struct
-throtl_grp *throtl_find_tg(struct throtl_data *td, struct blkio_cgroup *blkcg)
+static struct throtl_grp *throtl_lookup_tg(struct throtl_data *td,
+					   struct blkcg *blkcg)
 {
-	struct throtl_grp *tg = NULL;
-	void *key = td;
-
 	/*
-	 * This is the common case when there are no blkio cgroups.
- 	 * Avoid lookup in this case
- 	 */
-	if (blkcg == &blkio_root_cgroup)
-		tg = td->root_tg;
-	else
-		tg = tg_of_blkg(blkiocg_lookup_group(blkcg, key));
+	 * This is the common case when there are no blkcgs.  Avoid lookup
+	 * in this case
+	 */
+	if (blkcg == &blkcg_root)
+		return td_root_tg(td);
 
-	__throtl_tg_fill_dev_details(td, tg);
-	return tg;
+	return blkg_to_tg(blkg_lookup(blkcg, td->queue));
 }
 
-static struct throtl_grp * throtl_get_tg(struct throtl_data *td)
+static struct throtl_grp *throtl_lookup_create_tg(struct throtl_data *td,
+						  struct blkcg *blkcg)
 {
-	struct throtl_grp *tg = NULL, *__tg = NULL;
-	struct blkio_cgroup *blkcg;
 	struct request_queue *q = td->queue;
-
-	/* no throttling for dead queue */
-	if (unlikely(blk_queue_dead(q)))
-		return NULL;
-
-	rcu_read_lock();
-	blkcg = task_blkio_cgroup(current);
-	tg = throtl_find_tg(td, blkcg);
-	if (tg) {
-		rcu_read_unlock();
-		return tg;
-	}
-
-	/*
-	 * Need to allocate a group. Allocation of group also needs allocation
-	 * of per cpu stats which in-turn takes a mutex() and can block. Hence
-	 * we need to drop rcu lock and queue_lock before we call alloc.
-	 */
-	rcu_read_unlock();
-	spin_unlock_irq(q->queue_lock);
-
-	tg = throtl_alloc_tg(td);
-
-	/* Group allocated and queue is still alive. take the lock */
-	spin_lock_irq(q->queue_lock);
-
-	/* Make sure @q is still alive */
-	if (unlikely(blk_queue_dead(q))) {
-		kfree(tg);
-		return NULL;
-	}
-
-	/*
-	 * Initialize the new group. After sleeping, read the blkcg again.
-	 */
-	rcu_read_lock();
-	blkcg = task_blkio_cgroup(current);
+	struct throtl_grp *tg = NULL;
 
 	/*
-	 * If some other thread already allocated the group while we were
-	 * not holding queue lock, free up the group
+	 * This is the common case when there are no blkcgs.  Avoid lookup
+	 * in this case
 	 */
-	__tg = throtl_find_tg(td, blkcg);
-
-	if (__tg) {
-		kfree(tg);
-		rcu_read_unlock();
-		return __tg;
-	}
-
-	/* Group allocation failed. Account the IO to root group */
-	if (!tg) {
-		tg = td->root_tg;
-		return tg;
+	if (blkcg == &blkcg_root) {
+		tg = td_root_tg(td);
+	} else {
+		struct blkcg_gq *blkg;
+
+		blkg = blkg_lookup_create(blkcg, q);
+
+		/* if %NULL and @q is alive, fall back to root_tg */
+		if (!IS_ERR(blkg))
+			tg = blkg_to_tg(blkg);
+		else if (!blk_queue_dead(q))
+			tg = td_root_tg(td);
 	}
 
-	throtl_init_add_tg_lists(td, tg, blkcg);
-	rcu_read_unlock();
 	return tg;
 }
 
@@ -734,16 +675,41 @@ static bool tg_may_dispatch(struct throtl_data *td, struct throtl_grp *tg,
 	return 0;
 }
 
+static void throtl_update_dispatch_stats(struct blkcg_gq *blkg, u64 bytes,
+					 int rw)
+{
+	struct throtl_grp *tg = blkg_to_tg(blkg);
+	struct tg_stats_cpu *stats_cpu;
+	unsigned long flags;
+
+	/* If per cpu stats are not allocated yet, don't do any accounting. */
+	if (tg->stats_cpu == NULL)
+		return;
+
+	/*
+	 * Disabling interrupts to provide mutual exclusion between two
+	 * writes on same cpu. It probably is not needed for 64bit. Not
+	 * optimizing that case yet.
+	 */
+	local_irq_save(flags);
+
+	stats_cpu = this_cpu_ptr(tg->stats_cpu);
+
+	blkg_rwstat_add(&stats_cpu->serviced, rw, 1);
+	blkg_rwstat_add(&stats_cpu->service_bytes, rw, bytes);
+
+	local_irq_restore(flags);
+}
+
 static void throtl_charge_bio(struct throtl_grp *tg, struct bio *bio)
 {
 	bool rw = bio_data_dir(bio);
-	bool sync = rw_is_sync(bio->bi_rw);
 
 	/* Charge the bio to the group */
 	tg->bytes_disp[rw] += bio->bi_size;
 	tg->io_disp[rw]++;
 
-	blkiocg_update_dispatch_stats(&tg->blkg, bio->bi_size, rw, sync);
+	throtl_update_dispatch_stats(tg_to_blkg(tg), bio->bi_size, bio->bi_rw);
 }
 
 static void throtl_add_bio_tg(struct throtl_data *td, struct throtl_grp *tg,
@@ -753,7 +719,7 @@ static void throtl_add_bio_tg(struct throtl_data *td, struct throtl_grp *tg,
 
 	bio_list_add(&tg->bio_lists[rw], bio);
 	/* Take a bio reference on tg */
-	throtl_ref_get_tg(tg);
+	blkg_get(tg_to_blkg(tg));
 	tg->nr_queued[rw]++;
 	td->nr_queued[rw]++;
 	throtl_enqueue_tg(td, tg);
@@ -786,8 +752,8 @@ static void tg_dispatch_one_bio(struct throtl_data *td, struct throtl_grp *tg,
 
 	bio = bio_list_pop(&tg->bio_lists[rw]);
 	tg->nr_queued[rw]--;
-	/* Drop bio reference on tg */
-	throtl_put_tg(tg);
+	/* Drop bio reference on blkg */
+	blkg_put(tg_to_blkg(tg));
 
 	BUG_ON(td->nr_queued[rw] <= 0);
 	td->nr_queued[rw]--;
@@ -865,8 +831,8 @@ static int throtl_select_dispatch(struct throtl_data *td, struct bio_list *bl)
 
 static void throtl_process_limit_change(struct throtl_data *td)
 {
-	struct throtl_grp *tg;
-	struct hlist_node *pos, *n;
+	struct request_queue *q = td->queue;
+	struct blkcg_gq *blkg, *n;
 
 	if (!td->limits_changed)
 		return;
@@ -875,7 +841,9 @@ static void throtl_process_limit_change(struct throtl_data *td)
 
 	throtl_log(td, "limits changed");
 
-	hlist_for_each_entry_safe(tg, pos, n, &td->tg_list, tg_node) {
+	list_for_each_entry_safe(blkg, n, &q->blkg_list, q_node) {
+		struct throtl_grp *tg = blkg_to_tg(blkg);
+
 		if (!tg->limits_changed)
 			continue;
 
@@ -973,120 +941,159 @@ throtl_schedule_delayed_work(struct throtl_data *td, unsigned long delay)
 	}
 }
 
-static void
-throtl_destroy_tg(struct throtl_data *td, struct throtl_grp *tg)
+static u64 tg_prfill_cpu_rwstat(struct seq_file *sf,
+				struct blkg_policy_data *pd, int off)
 {
-	/* Something wrong if we are trying to remove same group twice */
-	BUG_ON(hlist_unhashed(&tg->tg_node));
+	struct throtl_grp *tg = pd_to_tg(pd);
+	struct blkg_rwstat rwstat = { }, tmp;
+	int i, cpu;
 
-	hlist_del_init(&tg->tg_node);
+	for_each_possible_cpu(cpu) {
+		struct tg_stats_cpu *sc = per_cpu_ptr(tg->stats_cpu, cpu);
 
-	/*
-	 * Put the reference taken at the time of creation so that when all
-	 * queues are gone, group can be destroyed.
-	 */
-	throtl_put_tg(tg);
-	td->nr_undestroyed_grps--;
+		tmp = blkg_rwstat_read((void *)sc + off);
+		for (i = 0; i < BLKG_RWSTAT_NR; i++)
+			rwstat.cnt[i] += tmp.cnt[i];
+	}
+
+	return __blkg_prfill_rwstat(sf, pd, &rwstat);
 }
 
-static void throtl_release_tgs(struct throtl_data *td)
+static int tg_print_cpu_rwstat(struct cgroup *cgrp, struct cftype *cft,
+			       struct seq_file *sf)
 {
-	struct hlist_node *pos, *n;
-	struct throtl_grp *tg;
+	struct blkcg *blkcg = cgroup_to_blkcg(cgrp);
 
-	hlist_for_each_entry_safe(tg, pos, n, &td->tg_list, tg_node) {
-		/*
-		 * If cgroup removal path got to blk_group first and removed
-		 * it from cgroup list, then it will take care of destroying
-		 * cfqg also.
-		 */
-		if (!blkiocg_del_blkio_group(&tg->blkg))
-			throtl_destroy_tg(td, tg);
-	}
+	blkcg_print_blkgs(sf, blkcg, tg_prfill_cpu_rwstat, &blkcg_policy_throtl,
+			  cft->private, true);
+	return 0;
 }
 
-/*
- * Blk cgroup controller notification saying that blkio_group object is being
- * delinked as associated cgroup object is going away. That also means that
- * no new IO will come in this group. So get rid of this group as soon as
- * any pending IO in the group is finished.
- *
- * This function is called under rcu_read_lock(). key is the rcu protected
- * pointer. That means "key" is a valid throtl_data pointer as long as we are
- * rcu read lock.
- *
- * "key" was fetched from blkio_group under blkio_cgroup->lock. That means
- * it should not be NULL as even if queue was going away, cgroup deltion
- * path got to it first.
- */
-void throtl_unlink_blkio_group(void *key, struct blkio_group *blkg)
+static u64 tg_prfill_conf_u64(struct seq_file *sf, struct blkg_policy_data *pd,
+			      int off)
 {
-	unsigned long flags;
-	struct throtl_data *td = key;
+	struct throtl_grp *tg = pd_to_tg(pd);
+	u64 v = *(u64 *)((void *)tg + off);
 
-	spin_lock_irqsave(td->queue->queue_lock, flags);
-	throtl_destroy_tg(td, tg_of_blkg(blkg));
-	spin_unlock_irqrestore(td->queue->queue_lock, flags);
+	if (v == -1)
+		return 0;
+	return __blkg_prfill_u64(sf, pd, v);
 }
 
-static void throtl_update_blkio_group_common(struct throtl_data *td,
-				struct throtl_grp *tg)
+static u64 tg_prfill_conf_uint(struct seq_file *sf, struct blkg_policy_data *pd,
+			       int off)
 {
-	xchg(&tg->limits_changed, true);
-	xchg(&td->limits_changed, true);
-	/* Schedule a work now to process the limit change */
-	throtl_schedule_delayed_work(td, 0);
+	struct throtl_grp *tg = pd_to_tg(pd);
+	unsigned int v = *(unsigned int *)((void *)tg + off);
+
+	if (v == -1)
+		return 0;
+	return __blkg_prfill_u64(sf, pd, v);
 }
 
-/*
- * For all update functions, key should be a valid pointer because these
- * update functions are called under blkcg_lock, that means, blkg is
- * valid and in turn key is valid. queue exit path can not race because
- * of blkcg_lock
- *
- * Can not take queue lock in update functions as queue lock under blkcg_lock
- * is not allowed. Under other paths we take blkcg_lock under queue_lock.
- */
-static void throtl_update_blkio_group_read_bps(void *key,
-				struct blkio_group *blkg, u64 read_bps)
+static int tg_print_conf_u64(struct cgroup *cgrp, struct cftype *cft,
+			     struct seq_file *sf)
 {
-	struct throtl_data *td = key;
-	struct throtl_grp *tg = tg_of_blkg(blkg);
-
-	tg->bps[READ] = read_bps;
-	throtl_update_blkio_group_common(td, tg);
+	blkcg_print_blkgs(sf, cgroup_to_blkcg(cgrp), tg_prfill_conf_u64,
+			  &blkcg_policy_throtl, cft->private, false);
+	return 0;
 }
 
-static void throtl_update_blkio_group_write_bps(void *key,
-				struct blkio_group *blkg, u64 write_bps)
+static int tg_print_conf_uint(struct cgroup *cgrp, struct cftype *cft,
+			      struct seq_file *sf)
 {
-	struct throtl_data *td = key;
-	struct throtl_grp *tg = tg_of_blkg(blkg);
-
-	tg->bps[WRITE] = write_bps;
-	throtl_update_blkio_group_common(td, tg);
+	blkcg_print_blkgs(sf, cgroup_to_blkcg(cgrp), tg_prfill_conf_uint,
+			  &blkcg_policy_throtl, cft->private, false);
+	return 0;
 }
 
-static void throtl_update_blkio_group_read_iops(void *key,
-			struct blkio_group *blkg, unsigned int read_iops)
+static int tg_set_conf(struct cgroup *cgrp, struct cftype *cft, const char *buf,
+		       bool is_u64)
 {
-	struct throtl_data *td = key;
-	struct throtl_grp *tg = tg_of_blkg(blkg);
+	struct blkcg *blkcg = cgroup_to_blkcg(cgrp);
+	struct blkg_conf_ctx ctx;
+	struct throtl_grp *tg;
+	struct throtl_data *td;
+	int ret;
+
+	ret = blkg_conf_prep(blkcg, &blkcg_policy_throtl, buf, &ctx);
+	if (ret)
+		return ret;
+
+	tg = blkg_to_tg(ctx.blkg);
+	td = ctx.blkg->q->td;
+
+	if (!ctx.v)
+		ctx.v = -1;
+
+	if (is_u64)
+		*(u64 *)((void *)tg + cft->private) = ctx.v;
+	else
+		*(unsigned int *)((void *)tg + cft->private) = ctx.v;
+
+	/* XXX: we don't need the following deferred processing */
+	xchg(&tg->limits_changed, true);
+	xchg(&td->limits_changed, true);
+	throtl_schedule_delayed_work(td, 0);
 
-	tg->iops[READ] = read_iops;
-	throtl_update_blkio_group_common(td, tg);
+	blkg_conf_finish(&ctx);
+	return 0;
 }
 
-static void throtl_update_blkio_group_write_iops(void *key,
-			struct blkio_group *blkg, unsigned int write_iops)
+static int tg_set_conf_u64(struct cgroup *cgrp, struct cftype *cft,
+			   const char *buf)
 {
-	struct throtl_data *td = key;
-	struct throtl_grp *tg = tg_of_blkg(blkg);
+	return tg_set_conf(cgrp, cft, buf, true);
+}
 
-	tg->iops[WRITE] = write_iops;
-	throtl_update_blkio_group_common(td, tg);
+static int tg_set_conf_uint(struct cgroup *cgrp, struct cftype *cft,
+			    const char *buf)
+{
+	return tg_set_conf(cgrp, cft, buf, false);
 }
 
+static struct cftype throtl_files[] = {
+	{
+		.name = "throttle.read_bps_device",
+		.private = offsetof(struct throtl_grp, bps[READ]),
+		.read_seq_string = tg_print_conf_u64,
+		.write_string = tg_set_conf_u64,
+		.max_write_len = 256,
+	},
+	{
+		.name = "throttle.write_bps_device",
+		.private = offsetof(struct throtl_grp, bps[WRITE]),
+		.read_seq_string = tg_print_conf_u64,
+		.write_string = tg_set_conf_u64,
+		.max_write_len = 256,
+	},
+	{
+		.name = "throttle.read_iops_device",
+		.private = offsetof(struct throtl_grp, iops[READ]),
+		.read_seq_string = tg_print_conf_uint,
+		.write_string = tg_set_conf_uint,
+		.max_write_len = 256,
+	},
+	{
+		.name = "throttle.write_iops_device",
+		.private = offsetof(struct throtl_grp, iops[WRITE]),
+		.read_seq_string = tg_print_conf_uint,
+		.write_string = tg_set_conf_uint,
+		.max_write_len = 256,
+	},
+	{
+		.name = "throttle.io_service_bytes",
+		.private = offsetof(struct tg_stats_cpu, service_bytes),
+		.read_seq_string = tg_print_cpu_rwstat,
+	},
+	{
+		.name = "throttle.io_serviced",
+		.private = offsetof(struct tg_stats_cpu, serviced),
+		.read_seq_string = tg_print_cpu_rwstat,
+	},
+	{ }	/* terminate */
+};
+
 static void throtl_shutdown_wq(struct request_queue *q)
 {
 	struct throtl_data *td = q->td;
@@ -1094,19 +1101,13 @@ static void throtl_shutdown_wq(struct request_queue *q)
 	cancel_delayed_work_sync(&td->throtl_work);
 }
 
-static struct blkio_policy_type blkio_policy_throtl = {
-	.ops = {
-		.blkio_unlink_group_fn = throtl_unlink_blkio_group,
-		.blkio_update_group_read_bps_fn =
-					throtl_update_blkio_group_read_bps,
-		.blkio_update_group_write_bps_fn =
-					throtl_update_blkio_group_write_bps,
-		.blkio_update_group_read_iops_fn =
-					throtl_update_blkio_group_read_iops,
-		.blkio_update_group_write_iops_fn =
-					throtl_update_blkio_group_write_iops,
-	},
-	.plid = BLKIO_POLICY_THROTL,
+static struct blkcg_policy blkcg_policy_throtl = {
+	.pd_size		= sizeof(struct throtl_grp),
+	.cftypes		= throtl_files,
+
+	.pd_init_fn		= throtl_pd_init,
+	.pd_exit_fn		= throtl_pd_exit,
+	.pd_reset_stats_fn	= throtl_pd_reset_stats,
 };
 
 bool blk_throtl_bio(struct request_queue *q, struct bio *bio)
@@ -1114,7 +1115,7 @@ bool blk_throtl_bio(struct request_queue *q, struct bio *bio)
 	struct throtl_data *td = q->td;
 	struct throtl_grp *tg;
 	bool rw = bio_data_dir(bio), update_disptime = true;
-	struct blkio_cgroup *blkcg;
+	struct blkcg *blkcg;
 	bool throttled = false;
 
 	if (bio->bi_rw & REQ_THROTTLED) {
@@ -1122,33 +1123,31 @@ bool blk_throtl_bio(struct request_queue *q, struct bio *bio)
 		goto out;
 	}
 
+	/* bio_associate_current() needs ioc, try creating */
+	create_io_context(GFP_ATOMIC, q->node);
+
 	/*
 	 * A throtl_grp pointer retrieved under rcu can be used to access
 	 * basic fields like stats and io rates. If a group has no rules,
 	 * just update the dispatch stats in lockless manner and return.
 	 */
-
 	rcu_read_lock();
-	blkcg = task_blkio_cgroup(current);
-	tg = throtl_find_tg(td, blkcg);
+	blkcg = bio_blkcg(bio);
+	tg = throtl_lookup_tg(td, blkcg);
 	if (tg) {
-		throtl_tg_fill_dev_details(td, tg);
-
 		if (tg_no_rule_group(tg, rw)) {
-			blkiocg_update_dispatch_stats(&tg->blkg, bio->bi_size,
-					rw, rw_is_sync(bio->bi_rw));
-			rcu_read_unlock();
-			goto out;
+			throtl_update_dispatch_stats(tg_to_blkg(tg),
+						     bio->bi_size, bio->bi_rw);
+			goto out_unlock_rcu;
 		}
 	}
-	rcu_read_unlock();
 
 	/*
 	 * Either group has not been allocated yet or it is not an unlimited
 	 * IO group
 	 */
 	spin_lock_irq(q->queue_lock);
-	tg = throtl_get_tg(td);
+	tg = throtl_lookup_create_tg(td, blkcg);
 	if (unlikely(!tg))
 		goto out_unlock;
 
@@ -1189,6 +1188,7 @@ queue_bio:
 			tg->io_disp[rw], tg->iops[rw],
 			tg->nr_queued[READ], tg->nr_queued[WRITE]);
 
+	bio_associate_current(bio);
 	throtl_add_bio_tg(q->td, tg, bio);
 	throttled = true;
 
@@ -1199,6 +1199,8 @@ queue_bio:
 
 out_unlock:
 	spin_unlock_irq(q->queue_lock);
+out_unlock_rcu:
+	rcu_read_unlock();
 out:
 	return throttled;
 }
@@ -1241,79 +1243,31 @@ void blk_throtl_drain(struct request_queue *q)
 int blk_throtl_init(struct request_queue *q)
 {
 	struct throtl_data *td;
-	struct throtl_grp *tg;
+	int ret;
 
 	td = kzalloc_node(sizeof(*td), GFP_KERNEL, q->node);
 	if (!td)
 		return -ENOMEM;
 
-	INIT_HLIST_HEAD(&td->tg_list);
 	td->tg_service_tree = THROTL_RB_ROOT;
 	td->limits_changed = false;
 	INIT_DELAYED_WORK(&td->throtl_work, blk_throtl_work);
 
-	/* alloc and Init root group. */
+	q->td = td;
 	td->queue = q;
-	tg = throtl_alloc_tg(td);
 
-	if (!tg) {
+	/* activate policy */
+	ret = blkcg_activate_policy(q, &blkcg_policy_throtl);
+	if (ret)
 		kfree(td);
-		return -ENOMEM;
-	}
-
-	td->root_tg = tg;
-
-	rcu_read_lock();
-	throtl_init_add_tg_lists(td, tg, &blkio_root_cgroup);
-	rcu_read_unlock();
-
-	/* Attach throtl data to request queue */
-	q->td = td;
-	return 0;
+	return ret;
 }
 
 void blk_throtl_exit(struct request_queue *q)
 {
-	struct throtl_data *td = q->td;
-	bool wait = false;
-
-	BUG_ON(!td);
-
-	throtl_shutdown_wq(q);
-
-	spin_lock_irq(q->queue_lock);
-	throtl_release_tgs(td);
-
-	/* If there are other groups */
-	if (td->nr_undestroyed_grps > 0)
-		wait = true;
-
-	spin_unlock_irq(q->queue_lock);
-
-	/*
-	 * Wait for tg->blkg->key accessors to exit their grace periods.
-	 * Do this wait only if there are other undestroyed groups out
-	 * there (other than root group). This can happen if cgroup deletion
-	 * path claimed the responsibility of cleaning up a group before
-	 * queue cleanup code get to the group.
-	 *
-	 * Do not call synchronize_rcu() unconditionally as there are drivers
-	 * which create/delete request queue hundreds of times during scan/boot
-	 * and synchronize_rcu() can take significant time and slow down boot.
-	 */
-	if (wait)
-		synchronize_rcu();
-
-	/*
-	 * Just being safe to make sure after previous flush if some body did
-	 * update limits through cgroup and another work got queued, cancel
-	 * it.
-	 */
+	BUG_ON(!q->td);
 	throtl_shutdown_wq(q);
-}
-
-void blk_throtl_release(struct request_queue *q)
-{
+	blkcg_deactivate_policy(q, &blkcg_policy_throtl);
 	kfree(q->td);
 }
 
@@ -1323,8 +1277,7 @@ static int __init throtl_init(void)
 	if (!kthrotld_workqueue)
 		panic("Failed to create kthrotld\n");
 
-	blkio_policy_register(&blkio_policy_throtl);
-	return 0;
+	return blkcg_policy_register(&blkcg_policy_throtl);
 }
 
 module_init(throtl_init);

block/blk.h

@@ -23,7 +23,8 @@ void blk_rq_bio_prep(struct request_queue *q, struct request *rq,
 			struct bio *bio);
 int blk_rq_append_bio(struct request_queue *q, struct request *rq,
 		      struct bio *bio);
-void blk_drain_queue(struct request_queue *q, bool drain_all);
+void blk_queue_bypass_start(struct request_queue *q);
+void blk_queue_bypass_end(struct request_queue *q);
 void blk_dequeue_request(struct request *rq);
 void __blk_queue_free_tags(struct request_queue *q);
 bool __blk_end_bidi_request(struct request *rq, int error,
@@ -144,9 +145,6 @@ void blk_queue_congestion_threshold(struct request_queue *q);
 
 int blk_dev_init(void);
 
-void elv_quiesce_start(struct request_queue *q);
-void elv_quiesce_end(struct request_queue *q);
-
 
 /*
  * Return the threshold (number of used requests) at which the queue is
@@ -186,32 +184,30 @@ static inline int blk_do_io_stat(struct request *rq)
  */
 void get_io_context(struct io_context *ioc);
 struct io_cq *ioc_lookup_icq(struct io_context *ioc, struct request_queue *q);
-struct io_cq *ioc_create_icq(struct request_queue *q, gfp_t gfp_mask);
+struct io_cq *ioc_create_icq(struct io_context *ioc, struct request_queue *q,
+			     gfp_t gfp_mask);
 void ioc_clear_queue(struct request_queue *q);
 
-void create_io_context_slowpath(struct task_struct *task, gfp_t gfp_mask,
-				int node);
+int create_task_io_context(struct task_struct *task, gfp_t gfp_mask, int node);
 
 /**
  * create_io_context - try to create task->io_context
- * @task: target task
  * @gfp_mask: allocation mask
  * @node: allocation node
  *
- * If @task->io_context is %NULL, allocate a new io_context and install it.
- * Returns the current @task->io_context which may be %NULL if allocation
- * failed.
+ * If %current->io_context is %NULL, allocate a new io_context and install
+ * it.  Returns the current %current->io_context which may be %NULL if
+ * allocation failed.
  *
  * Note that this function can't be called with IRQ disabled because
- * task_lock which protects @task->io_context is IRQ-unsafe.
+ * task_lock which protects %current->io_context is IRQ-unsafe.
  */
-static inline struct io_context *create_io_context(struct task_struct *task,
-						   gfp_t gfp_mask, int node)
+static inline struct io_context *create_io_context(gfp_t gfp_mask, int node)
 {
 	WARN_ON_ONCE(irqs_disabled());
-	if (unlikely(!task->io_context))
-		create_io_context_slowpath(task, gfp_mask, node);
-	return task->io_context;
+	if (unlikely(!current->io_context))
+		create_task_io_context(current, gfp_mask, node);
+	return current->io_context;
 }
 
 /*
@@ -222,7 +218,6 @@ extern bool blk_throtl_bio(struct request_queue *q, struct bio *bio);
 extern void blk_throtl_drain(struct request_queue *q);
 extern int blk_throtl_init(struct request_queue *q);
 extern void blk_throtl_exit(struct request_queue *q);
-extern void blk_throtl_release(struct request_queue *q);
 #else /* CONFIG_BLK_DEV_THROTTLING */
 static inline bool blk_throtl_bio(struct request_queue *q, struct bio *bio)
 {
@@ -231,7 +226,6 @@ static inline bool blk_throtl_bio(struct request_queue *q, struct bio *bio)
 static inline void blk_throtl_drain(struct request_queue *q) { }
 static inline int blk_throtl_init(struct request_queue *q) { return 0; }
 static inline void blk_throtl_exit(struct request_queue *q) { }
-static inline void blk_throtl_release(struct request_queue *q) { }
 #endif /* CONFIG_BLK_DEV_THROTTLING */
 
 #endif /* BLK_INTERNAL_H */

block/cfq.h

@@ -1,115 +0,0 @@
-#ifndef _CFQ_H
-#define _CFQ_H
-#include "blk-cgroup.h"
-
-#ifdef CONFIG_CFQ_GROUP_IOSCHED
-static inline void cfq_blkiocg_update_io_add_stats(struct blkio_group *blkg,
-	struct blkio_group *curr_blkg, bool direction, bool sync)
-{
-	blkiocg_update_io_add_stats(blkg, curr_blkg, direction, sync);
-}
-
-static inline void cfq_blkiocg_update_dequeue_stats(struct blkio_group *blkg,
-			unsigned long dequeue)
-{
-	blkiocg_update_dequeue_stats(blkg, dequeue);
-}
-
-static inline void cfq_blkiocg_update_timeslice_used(struct blkio_group *blkg,
-			unsigned long time, unsigned long unaccounted_time)
-{
-	blkiocg_update_timeslice_used(blkg, time, unaccounted_time);
-}
-
-static inline void cfq_blkiocg_set_start_empty_time(struct blkio_group *blkg)
-{
-	blkiocg_set_start_empty_time(blkg);
-}
-
-static inline void cfq_blkiocg_update_io_remove_stats(struct blkio_group *blkg,
-				bool direction, bool sync)
-{
-	blkiocg_update_io_remove_stats(blkg, direction, sync);
-}
-
-static inline void cfq_blkiocg_update_io_merged_stats(struct blkio_group *blkg,
-		bool direction, bool sync)
-{
-	blkiocg_update_io_merged_stats(blkg, direction, sync);
-}
-
-static inline void cfq_blkiocg_update_idle_time_stats(struct blkio_group *blkg)
-{
-	blkiocg_update_idle_time_stats(blkg);
-}
-
-static inline void
-cfq_blkiocg_update_avg_queue_size_stats(struct blkio_group *blkg)
-{
-	blkiocg_update_avg_queue_size_stats(blkg);
-}
-
-static inline void
-cfq_blkiocg_update_set_idle_time_stats(struct blkio_group *blkg)
-{
-	blkiocg_update_set_idle_time_stats(blkg);
-}
-
-static inline void cfq_blkiocg_update_dispatch_stats(struct blkio_group *blkg,
-				uint64_t bytes, bool direction, bool sync)
-{
-	blkiocg_update_dispatch_stats(blkg, bytes, direction, sync);
-}
-
-static inline void cfq_blkiocg_update_completion_stats(struct blkio_group *blkg, uint64_t start_time, uint64_t io_start_time, bool direction, bool sync)
-{
-	blkiocg_update_completion_stats(blkg, start_time, io_start_time,
-				direction, sync);
-}
-
-static inline void cfq_blkiocg_add_blkio_group(struct blkio_cgroup *blkcg,
-			struct blkio_group *blkg, void *key, dev_t dev) {
-	blkiocg_add_blkio_group(blkcg, blkg, key, dev, BLKIO_POLICY_PROP);
-}
-
-static inline int cfq_blkiocg_del_blkio_group(struct blkio_group *blkg)
-{
-	return blkiocg_del_blkio_group(blkg);
-}
-
-#else /* CFQ_GROUP_IOSCHED */
-static inline void cfq_blkiocg_update_io_add_stats(struct blkio_group *blkg,
-	struct blkio_group *curr_blkg, bool direction, bool sync) {}
-
-static inline void cfq_blkiocg_update_dequeue_stats(struct blkio_group *blkg,
-			unsigned long dequeue) {}
-
-static inline void cfq_blkiocg_update_timeslice_used(struct blkio_group *blkg,
-			unsigned long time, unsigned long unaccounted_time) {}
-static inline void cfq_blkiocg_set_start_empty_time(struct blkio_group *blkg) {}
-static inline void cfq_blkiocg_update_io_remove_stats(struct blkio_group *blkg,
-				bool direction, bool sync) {}
-static inline void cfq_blkiocg_update_io_merged_stats(struct blkio_group *blkg,
-		bool direction, bool sync) {}
-static inline void cfq_blkiocg_update_idle_time_stats(struct blkio_group *blkg)
-{
-}
-static inline void
-cfq_blkiocg_update_avg_queue_size_stats(struct blkio_group *blkg) {}
-
-static inline void
-cfq_blkiocg_update_set_idle_time_stats(struct blkio_group *blkg) {}
-
-static inline void cfq_blkiocg_update_dispatch_stats(struct blkio_group *blkg,
-				uint64_t bytes, bool direction, bool sync) {}
-static inline void cfq_blkiocg_update_completion_stats(struct blkio_group *blkg, uint64_t start_time, uint64_t io_start_time, bool direction, bool sync) {}
-
-static inline void cfq_blkiocg_add_blkio_group(struct blkio_cgroup *blkcg,
-			struct blkio_group *blkg, void *key, dev_t dev) {}
-static inline int cfq_blkiocg_del_blkio_group(struct blkio_group *blkg)
-{
-	return 0;
-}
-
-#endif /* CFQ_GROUP_IOSCHED */
-#endif

block/deadline-iosched.c

@@ -337,13 +337,13 @@ static void deadline_exit_queue(struct elevator_queue *e)
 /*
  * initialize elevator private data (deadline_data).
  */
-static void *deadline_init_queue(struct request_queue *q)
+static int deadline_init_queue(struct request_queue *q)
 {
 	struct deadline_data *dd;
 
 	dd = kmalloc_node(sizeof(*dd), GFP_KERNEL | __GFP_ZERO, q->node);
 	if (!dd)
-		return NULL;
+		return -ENOMEM;
 
 	INIT_LIST_HEAD(&dd->fifo_list[READ]);
 	INIT_LIST_HEAD(&dd->fifo_list[WRITE]);
@@ -354,7 +354,9 @@ static void *deadline_init_queue(struct request_queue *q)
 	dd->writes_starved = writes_starved;
 	dd->front_merges = 1;
 	dd->fifo_batch = fifo_batch;
-	return dd;
+
+	q->elevator->elevator_data = dd;
+	return 0;
 }
 
 /*

block/elevator.c

@@ -38,6 +38,7 @@
 #include <trace/events/block.h>
 
 #include "blk.h"
+#include "blk-cgroup.h"
 
 static DEFINE_SPINLOCK(elv_list_lock);
 static LIST_HEAD(elv_list);
@@ -121,15 +122,6 @@ static struct elevator_type *elevator_get(const char *name)
 	return e;
 }
 
-static int elevator_init_queue(struct request_queue *q,
-			       struct elevator_queue *eq)
-{
-	eq->elevator_data = eq->type->ops.elevator_init_fn(q);
-	if (eq->elevator_data)
-		return 0;
-	return -ENOMEM;
-}
-
 static char chosen_elevator[ELV_NAME_MAX];
 
 static int __init elevator_setup(char *str)
@@ -188,7 +180,6 @@ static void elevator_release(struct kobject *kobj)
 int elevator_init(struct request_queue *q, char *name)
 {
 	struct elevator_type *e = NULL;
-	struct elevator_queue *eq;
 	int err;
 
 	if (unlikely(q->elevator))
@@ -222,17 +213,16 @@ int elevator_init(struct request_queue *q, char *name)
 		}
 	}
 
-	eq = elevator_alloc(q, e);
-	if (!eq)
+	q->elevator = elevator_alloc(q, e);
+	if (!q->elevator)
 		return -ENOMEM;
 
-	err = elevator_init_queue(q, eq);
+	err = e->ops.elevator_init_fn(q);
 	if (err) {
-		kobject_put(&eq->kobj);
+		kobject_put(&q->elevator->kobj);
 		return err;
 	}
 
-	q->elevator = eq;
 	return 0;
 }
 EXPORT_SYMBOL(elevator_init);
@@ -564,25 +554,6 @@ void elv_drain_elevator(struct request_queue *q)
 	}
 }
 
-void elv_quiesce_start(struct request_queue *q)
-{
-	if (!q->elevator)
-		return;
-
-	spin_lock_irq(q->queue_lock);
-	queue_flag_set(QUEUE_FLAG_ELVSWITCH, q);
-	spin_unlock_irq(q->queue_lock);
-
-	blk_drain_queue(q, false);
-}
-
-void elv_quiesce_end(struct request_queue *q)
-{
-	spin_lock_irq(q->queue_lock);
-	queue_flag_clear(QUEUE_FLAG_ELVSWITCH, q);
-	spin_unlock_irq(q->queue_lock);
-}
-
 void __elv_add_request(struct request_queue *q, struct request *rq, int where)
 {
 	trace_block_rq_insert(q, rq);
@@ -692,12 +663,13 @@ struct request *elv_former_request(struct request_queue *q, struct request *rq)
 	return NULL;
 }
 
-int elv_set_request(struct request_queue *q, struct request *rq, gfp_t gfp_mask)
+int elv_set_request(struct request_queue *q, struct request *rq,
+		    struct bio *bio, gfp_t gfp_mask)
 {
 	struct elevator_queue *e = q->elevator;
 
 	if (e->type->ops.elevator_set_req_fn)
-		return e->type->ops.elevator_set_req_fn(q, rq, gfp_mask);
+		return e->type->ops.elevator_set_req_fn(q, rq, bio, gfp_mask);
 	return 0;
 }
 
@@ -801,8 +773,9 @@ static struct kobj_type elv_ktype = {
 	.release	= elevator_release,
 };
 
-int __elv_register_queue(struct request_queue *q, struct elevator_queue *e)
+int elv_register_queue(struct request_queue *q)
 {
+	struct elevator_queue *e = q->elevator;
 	int error;
 
 	error = kobject_add(&e->kobj, &q->kobj, "%s", "iosched");
@@ -820,11 +793,6 @@ int __elv_register_queue(struct request_queue *q, struct elevator_queue *e)
 	}
 	return error;
 }
-
-int elv_register_queue(struct request_queue *q)
-{
-	return __elv_register_queue(q, q->elevator);
-}
 EXPORT_SYMBOL(elv_register_queue);
 
 void elv_unregister_queue(struct request_queue *q)
@@ -907,53 +875,60 @@ EXPORT_SYMBOL_GPL(elv_unregister);
  */
 static int elevator_switch(struct request_queue *q, struct elevator_type *new_e)
 {
-	struct elevator_queue *old_elevator, *e;
+	struct elevator_queue *old = q->elevator;
+	bool registered = old->registered;
 	int err;
 
-	/* allocate new elevator */
-	e = elevator_alloc(q, new_e);
-	if (!e)
-		return -ENOMEM;
+	/*
+	 * Turn on BYPASS and drain all requests w/ elevator private data.
+	 * Block layer doesn't call into a quiesced elevator - all requests
+	 * are directly put on the dispatch list without elevator data
+	 * using INSERT_BACK.  All requests have SOFTBARRIER set and no
+	 * merge happens either.
+	 */
+	blk_queue_bypass_start(q);
+
+	/* unregister and clear all auxiliary data of the old elevator */
+	if (registered)
+		elv_unregister_queue(q);
+
+	spin_lock_irq(q->queue_lock);
+	ioc_clear_queue(q);
+	spin_unlock_irq(q->queue_lock);
 
-	err = elevator_init_queue(q, e);
+	/* allocate, init and register new elevator */
+	err = -ENOMEM;
+	q->elevator = elevator_alloc(q, new_e);
+	if (!q->elevator)
+		goto fail_init;
+
+	err = new_e->ops.elevator_init_fn(q);
 	if (err) {
-		kobject_put(&e->kobj);
-		return err;
+		kobject_put(&q->elevator->kobj);
+		goto fail_init;
 	}
 
-	/* turn on BYPASS and drain all requests w/ elevator private data */
-	elv_quiesce_start(q);
-
-	/* unregister old queue, register new one and kill old elevator */
-	if (q->elevator->registered) {
-		elv_unregister_queue(q);
-		err = __elv_register_queue(q, e);
+	if (registered) {
+		err = elv_register_queue(q);
 		if (err)
 			goto fail_register;
 	}
 
-	/* done, clear io_cq's, switch elevators and turn off BYPASS */
-	spin_lock_irq(q->queue_lock);
-	ioc_clear_queue(q);
-	old_elevator = q->elevator;
-	q->elevator = e;
-	spin_unlock_irq(q->queue_lock);
-
-	elevator_exit(old_elevator);
-	elv_quiesce_end(q);
+	/* done, kill the old one and finish */
+	elevator_exit(old);
+	blk_queue_bypass_end(q);
 
-	blk_add_trace_msg(q, "elv switch: %s", e->type->elevator_name);
+	blk_add_trace_msg(q, "elv switch: %s", new_e->elevator_name);
 
 	return 0;
 
 fail_register:
-	/*
-	 * switch failed, exit the new io scheduler and reattach the old
-	 * one again (along with re-adding the sysfs dir)
-	 */
-	elevator_exit(e);
+	elevator_exit(q->elevator);
+fail_init:
+	/* switch failed, restore and re-register old elevator */
+	q->elevator = old;
 	elv_register_queue(q);
-	elv_quiesce_end(q);
+	blk_queue_bypass_end(q);
 
 	return err;
 }

block/noop-iosched.c

@@ -59,15 +59,17 @@ noop_latter_request(struct request_queue *q, struct request *rq)
 	return list_entry(rq->queuelist.next, struct request, queuelist);
 }
 
-static void *noop_init_queue(struct request_queue *q)
+static int noop_init_queue(struct request_queue *q)
 {
 	struct noop_data *nd;
 
 	nd = kmalloc_node(sizeof(*nd), GFP_KERNEL, q->node);
 	if (!nd)
-		return NULL;
+		return -ENOMEM;
+
 	INIT_LIST_HEAD(&nd->queue);
-	return nd;
+	q->elevator->elevator_data = nd;
+	return 0;
 }
 
 static void noop_exit_queue(struct elevator_queue *e)

fs/bio.c

@@ -19,12 +19,14 @@
 #include <linux/swap.h>
 #include <linux/bio.h>
 #include <linux/blkdev.h>
+#include <linux/iocontext.h>
 #include <linux/slab.h>
 #include <linux/init.h>
 #include <linux/kernel.h>
 #include <linux/export.h>
 #include <linux/mempool.h>
 #include <linux/workqueue.h>
+#include <linux/cgroup.h>
 #include <scsi/sg.h>		/* for struct sg_iovec */
 
 #include <trace/events/block.h>
@@ -418,6 +420,7 @@ void bio_put(struct bio *bio)
 	 * last put frees it
 	 */
 	if (atomic_dec_and_test(&bio->bi_cnt)) {
+		bio_disassociate_task(bio);
 		bio->bi_next = NULL;
 		bio->bi_destructor(bio);
 	}
@@ -1646,6 +1649,64 @@ bad:
 }
 EXPORT_SYMBOL(bioset_create);
 
+#ifdef CONFIG_BLK_CGROUP
+/**
+ * bio_associate_current - associate a bio with %current
+ * @bio: target bio
+ *
+ * Associate @bio with %current if it hasn't been associated yet.  Block
+ * layer will treat @bio as if it were issued by %current no matter which
+ * task actually issues it.
+ *
+ * This function takes an extra reference of @task's io_context and blkcg
+ * which will be put when @bio is released.  The caller must own @bio,
+ * ensure %current->io_context exists, and is responsible for synchronizing
+ * calls to this function.
+ */
+int bio_associate_current(struct bio *bio)
+{
+	struct io_context *ioc;
+	struct cgroup_subsys_state *css;
+
+	if (bio->bi_ioc)
+		return -EBUSY;
+
+	ioc = current->io_context;
+	if (!ioc)
+		return -ENOENT;
+
+	/* acquire active ref on @ioc and associate */
+	get_io_context_active(ioc);
+	bio->bi_ioc = ioc;
+
+	/* associate blkcg if exists */
+	rcu_read_lock();
+	css = task_subsys_state(current, blkio_subsys_id);
+	if (css && css_tryget(css))
+		bio->bi_css = css;
+	rcu_read_unlock();
+
+	return 0;
+}
+
+/**
+ * bio_disassociate_task - undo bio_associate_current()
+ * @bio: target bio
+ */
+void bio_disassociate_task(struct bio *bio)
+{
+	if (bio->bi_ioc) {
+		put_io_context(bio->bi_ioc);
+		bio->bi_ioc = NULL;
+	}
+	if (bio->bi_css) {
+		css_put(bio->bi_css);
+		bio->bi_css = NULL;
+	}
+}
+
+#endif /* CONFIG_BLK_CGROUP */
+
 static void __init biovec_init_slabs(void)
 {
 	int i;

fs/ioprio.c

@@ -50,7 +50,7 @@ int set_task_ioprio(struct task_struct *task, int ioprio)
 
 	ioc = get_task_io_context(task, GFP_ATOMIC, NUMA_NO_NODE);
 	if (ioc) {
-		ioc_ioprio_changed(ioc, ioprio);
+		ioc->ioprio = ioprio;
 		put_io_context(ioc);
 	}
 

fs/splice.c

@@ -1388,7 +1388,7 @@ static long do_splice(struct file *in, loff_t __user *off_in,
  */
 static int get_iovec_page_array(const struct iovec __user *iov,
 				unsigned int nr_vecs, struct page **pages,
-				struct partial_page *partial, int aligned,
+				struct partial_page *partial, bool aligned,
 				unsigned int pipe_buffers)
 {
 	int buffers = 0, error = 0;
@@ -1626,7 +1626,7 @@ static long vmsplice_to_pipe(struct file *file, const struct iovec __user *iov,
 		return -ENOMEM;
 
 	spd.nr_pages = get_iovec_page_array(iov, nr_segs, spd.pages,
-					    spd.partial, flags & SPLICE_F_GIFT,
+					    spd.partial, false,
 					    pipe->buffers);
 	if (spd.nr_pages <= 0)
 		ret = spd.nr_pages;

include/linux/bio.h

@@ -269,6 +269,14 @@ extern struct bio_vec *bvec_alloc_bs(gfp_t, int, unsigned long *, struct bio_set
 extern void bvec_free_bs(struct bio_set *, struct bio_vec *, unsigned int);
 extern unsigned int bvec_nr_vecs(unsigned short idx);
 
+#ifdef CONFIG_BLK_CGROUP
+int bio_associate_current(struct bio *bio);
+void bio_disassociate_task(struct bio *bio);
+#else	/* CONFIG_BLK_CGROUP */
+static inline int bio_associate_current(struct bio *bio) { return -ENOENT; }
+static inline void bio_disassociate_task(struct bio *bio) { }
+#endif	/* CONFIG_BLK_CGROUP */
+
 /*
  * bio_set is used to allow other portions of the IO system to
  * allocate their own private memory pools for bio and iovec structures.

include/linux/blk_types.h

@@ -14,6 +14,8 @@ struct bio;
 struct bio_integrity_payload;
 struct page;
 struct block_device;
+struct io_context;
+struct cgroup_subsys_state;
 typedef void (bio_end_io_t) (struct bio *, int);
 typedef void (bio_destructor_t) (struct bio *);
 
@@ -66,6 +68,14 @@ struct bio {
 	bio_end_io_t		*bi_end_io;
 
 	void			*bi_private;
+#ifdef CONFIG_BLK_CGROUP
+	/*
+	 * Optional ioc and css associated with this bio.  Put on bio
+	 * release.  Read comment on top of bio_associate_current().
+	 */
+	struct io_context	*bi_ioc;
+	struct cgroup_subsys_state *bi_css;
+#endif
 #if defined(CONFIG_BLK_DEV_INTEGRITY)
 	struct bio_integrity_payload *bi_integrity;  /* data integrity */
 #endif

include/linux/blkdev.h

@@ -32,10 +32,17 @@ struct blk_trace;
 struct request;
 struct sg_io_hdr;
 struct bsg_job;
+struct blkcg_gq;
 
 #define BLKDEV_MIN_RQ	4
 #define BLKDEV_MAX_RQ	128	/* Default maximum */
 
+/*
+ * Maximum number of blkcg policies allowed to be registered concurrently.
+ * Defined here to simplify include dependency.
+ */
+#define BLKCG_MAX_POLS		2
+
 struct request;
 typedef void (rq_end_io_fn)(struct request *, int);
 
@@ -363,6 +370,11 @@ struct request_queue {
 	struct list_head	timeout_list;
 
 	struct list_head	icq_list;
+#ifdef CONFIG_BLK_CGROUP
+	DECLARE_BITMAP		(blkcg_pols, BLKCG_MAX_POLS);
+	struct blkcg_gq		*root_blkg;
+	struct list_head	blkg_list;
+#endif
 
 	struct queue_limits	limits;
 
@@ -390,12 +402,17 @@ struct request_queue {
 
 	struct mutex		sysfs_lock;
 
+	int			bypass_depth;
+
 #if defined(CONFIG_BLK_DEV_BSG)
 	bsg_job_fn		*bsg_job_fn;
 	int			bsg_job_size;
 	struct bsg_class_device bsg_dev;
 #endif
 
+#ifdef CONFIG_BLK_CGROUP
+	struct list_head	all_q_node;
+#endif
 #ifdef CONFIG_BLK_DEV_THROTTLING
 	/* Throttle data */
 	struct throtl_data *td;
@@ -407,7 +424,7 @@ struct request_queue {
 #define	QUEUE_FLAG_SYNCFULL	3	/* read queue has been filled */
 #define QUEUE_FLAG_ASYNCFULL	4	/* write queue has been filled */
 #define QUEUE_FLAG_DEAD		5	/* queue being torn down */
-#define QUEUE_FLAG_ELVSWITCH	6	/* don't use elevator, just do FIFO */
+#define QUEUE_FLAG_BYPASS	6	/* act as dumb FIFO queue */
 #define QUEUE_FLAG_BIDI		7	/* queue supports bidi requests */
 #define QUEUE_FLAG_NOMERGES     8	/* disable merge attempts */
 #define QUEUE_FLAG_SAME_COMP	9	/* complete on same CPU-group */
@@ -491,6 +508,7 @@ static inline void queue_flag_clear(unsigned int flag, struct request_queue *q)
 #define blk_queue_tagged(q)	test_bit(QUEUE_FLAG_QUEUED, &(q)->queue_flags)
 #define blk_queue_stopped(q)	test_bit(QUEUE_FLAG_STOPPED, &(q)->queue_flags)
 #define blk_queue_dead(q)	test_bit(QUEUE_FLAG_DEAD, &(q)->queue_flags)
+#define blk_queue_bypass(q)	test_bit(QUEUE_FLAG_BYPASS, &(q)->queue_flags)
 #define blk_queue_nomerges(q)	test_bit(QUEUE_FLAG_NOMERGES, &(q)->queue_flags)
 #define blk_queue_noxmerges(q)	\
 	test_bit(QUEUE_FLAG_NOXMERGES, &(q)->queue_flags)

include/linux/elevator.h

@@ -28,12 +28,13 @@ typedef int (elevator_may_queue_fn) (struct request_queue *, int);
 
 typedef void (elevator_init_icq_fn) (struct io_cq *);
 typedef void (elevator_exit_icq_fn) (struct io_cq *);
-typedef int (elevator_set_req_fn) (struct request_queue *, struct request *, gfp_t);
+typedef int (elevator_set_req_fn) (struct request_queue *, struct request *,
+				   struct bio *, gfp_t);
 typedef void (elevator_put_req_fn) (struct request *);
 typedef void (elevator_activate_req_fn) (struct request_queue *, struct request *);
 typedef void (elevator_deactivate_req_fn) (struct request_queue *, struct request *);
 
-typedef void *(elevator_init_fn) (struct request_queue *);
+typedef int (elevator_init_fn) (struct request_queue *);
 typedef void (elevator_exit_fn) (struct elevator_queue *);
 
 struct elevator_ops
@@ -129,7 +130,8 @@ extern void elv_unregister_queue(struct request_queue *q);
 extern int elv_may_queue(struct request_queue *, int);
 extern void elv_abort_queue(struct request_queue *);
 extern void elv_completed_request(struct request_queue *, struct request *);
-extern int elv_set_request(struct request_queue *, struct request *, gfp_t);
+extern int elv_set_request(struct request_queue *q, struct request *rq,
+			   struct bio *bio, gfp_t gfp_mask);
 extern void elv_put_request(struct request_queue *, struct request *);
 extern void elv_drain_elevator(struct request_queue *);
 

include/linux/iocontext.h

@@ -6,11 +6,7 @@
 #include <linux/workqueue.h>
 
 enum {
-	ICQ_IOPRIO_CHANGED	= 1 << 0,
-	ICQ_CGROUP_CHANGED	= 1 << 1,
 	ICQ_EXITED		= 1 << 2,
-
-	ICQ_CHANGED_MASK	= ICQ_IOPRIO_CHANGED | ICQ_CGROUP_CHANGED,
 };
 
 /*
@@ -100,6 +96,7 @@ struct io_cq {
  */
 struct io_context {
 	atomic_long_t refcount;
+	atomic_t active_ref;
 	atomic_t nr_tasks;
 
 	/* all the fields below are protected by this lock */
@@ -120,29 +117,37 @@ struct io_context {
 	struct work_struct release_work;
 };
 
-static inline struct io_context *ioc_task_link(struct io_context *ioc)
+/**
+ * get_io_context_active - get active reference on ioc
+ * @ioc: ioc of interest
+ *
+ * Only iocs with active reference can issue new IOs.  This function
+ * acquires an active reference on @ioc.  The caller must already have an
+ * active reference on @ioc.
+ */
+static inline void get_io_context_active(struct io_context *ioc)
 {
-	/*
-	 * if ref count is zero, don't allow sharing (ioc is going away, it's
-	 * a race).
-	 */
-	if (ioc && atomic_long_inc_not_zero(&ioc->refcount)) {
-		atomic_inc(&ioc->nr_tasks);
-		return ioc;
-	}
+	WARN_ON_ONCE(atomic_long_read(&ioc->refcount) <= 0);
+	WARN_ON_ONCE(atomic_read(&ioc->active_ref) <= 0);
+	atomic_long_inc(&ioc->refcount);
+	atomic_inc(&ioc->active_ref);
+}
+
+static inline void ioc_task_link(struct io_context *ioc)
+{
+	get_io_context_active(ioc);
 
-	return NULL;
+	WARN_ON_ONCE(atomic_read(&ioc->nr_tasks) <= 0);
+	atomic_inc(&ioc->nr_tasks);
 }
 
 struct task_struct;
 #ifdef CONFIG_BLOCK
 void put_io_context(struct io_context *ioc);
+void put_io_context_active(struct io_context *ioc);
 void exit_io_context(struct task_struct *task);
 struct io_context *get_task_io_context(struct task_struct *task,
 				       gfp_t gfp_flags, int node);
-void ioc_ioprio_changed(struct io_context *ioc, int ioprio);
-void ioc_cgroup_changed(struct io_context *ioc);
-unsigned int icq_get_changed(struct io_cq *icq);
 #else
 struct io_context;
 static inline void put_io_context(struct io_context *ioc) { }

include/linux/ioprio.h

@@ -42,26 +42,14 @@ enum {
 };
 
 /*
- * if process has set io priority explicitly, use that. if not, convert
- * the cpu scheduler nice value to an io priority
+ * Fallback BE priority
  */
 #define IOPRIO_NORM	(4)
-static inline int task_ioprio(struct io_context *ioc)
-{
-	if (ioprio_valid(ioc->ioprio))
-		return IOPRIO_PRIO_DATA(ioc->ioprio);
-
-	return IOPRIO_NORM;
-}
-
-static inline int task_ioprio_class(struct io_context *ioc)
-{
-	if (ioprio_valid(ioc->ioprio))
-		return IOPRIO_PRIO_CLASS(ioc->ioprio);
-
-	return IOPRIO_CLASS_BE;
-}
 
+/*
+ * if process has set io priority explicitly, use that. if not, convert
+ * the cpu scheduler nice value to an io priority
+ */
 static inline int task_nice_ioprio(struct task_struct *task)
 {
 	return (task_nice(task) + 20) / 5;

init/Kconfig

@@ -803,7 +803,7 @@ config RT_GROUP_SCHED
 endif #CGROUP_SCHED
 
 config BLK_CGROUP
-	tristate "Block IO controller"
+	bool "Block IO controller"
 	depends on BLOCK
 	default n
 	---help---

kernel/fork.c

@@ -976,9 +976,8 @@ static int copy_io(unsigned long clone_flags, struct task_struct *tsk)
 	 * Share io context with parent, if CLONE_IO is set
 	 */
 	if (clone_flags & CLONE_IO) {
-		tsk->io_context = ioc_task_link(ioc);
-		if (unlikely(!tsk->io_context))
-			return -ENOMEM;
+		ioc_task_link(ioc);
+		tsk->io_context = ioc;
 	} else if (ioprio_valid(ioc->ioprio)) {
 		new_ioc = get_task_io_context(tsk, GFP_KERNEL, NUMA_NO_NODE);
 		if (unlikely(!new_ioc))