Skip to content
Browse files

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
  ...
  • Loading branch information...
2 parents 2f83766 + ff26eaa commit 0d167518e045cc8bb63f0a8a0a85ad4fa4e0044f @torvalds committed
View
4 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
View
2,100 block/blk-cgroup.c
669 additions, 1,431 deletions not shown because the diff is too large. Please use a local Git client to view these changes.
View
647 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 */
View
281 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;
@@ -403,6 +411,49 @@ 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.
@@ -763,6 +808,22 @@ static bool blk_rq_should_init_elevator(struct bio *bio)
}
/**
+ * 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
* @rw_flags: RW and SYNC flags
@@ -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 @@ static struct request *get_request(struct request_queue *q, int rw_flags,
*/
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 @@ static struct request *get_request(struct request_queue *q, int rw_flags,
* process is not a "batcher", and not
* exempted by the IO scheduler
*/
- goto out;
+ return NULL;
}
}
}
@@ -844,7 +905,7 @@ static struct request *get_request(struct request_queue *q, int rw_flags,
* 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 @@ static struct request *get_request(struct request_queue *q, int rw_flags,
* 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 @@ static struct request *get_request(struct request_queue *q, int rw_flags,
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 @@ static struct request *get_request(struct request_queue *q, int rw_flags,
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);
View
126 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 @@ void exit_io_context(struct task_struct *task)
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",
View
6 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);
View
697 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 @@ bool blk_throtl_bio(struct request_queue *q, struct bio *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 @@ bool blk_throtl_bio(struct request_queue *q, struct bio *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.
- */