summaryrefslogtreecommitdiff
path: root/block
diff options
context:
space:
mode:
authorTejun Heo <tj@kernel.org>2019-11-07 22:18:03 +0300
committerJens Axboe <axboe@kernel.dk>2019-11-07 22:28:13 +0300
commitf73316482977ac401ac37245c9df48079d4e11f3 (patch)
tree1177e1189b87661f46900c6404047da82364dd93 /block
parent8a80d5d6638b7d58480a83aef49d587de63d4cbb (diff)
downloadlinux-f73316482977ac401ac37245c9df48079d4e11f3.tar.xz
blk-cgroup: reimplement basic IO stats using cgroup rstat
blk-cgroup has been using blkg_rwstat to track basic IO stats. Unfortunately, reading recursive stats scales badly as itinvolves walking all descendants. On systems with a huge number of cgroups (dead or alive), this can lead to substantial CPU cost when reading IO stats. This patch reimplements basic IO stats using cgroup rstat which uses more memory but makes recursive stat reading O(# descendants which have been active since last reading) instead of O(# descendants). * blk-cgroup core no longer uses sync/async stats. Introduce new stat enums - BLKG_IOSTAT_{READ|WRITE|DISCARD}. * Add blkg_iostat[_set] which encapsulates byte and io stats, last values for propagation delta calculation and u64_stats_sync for correctness on 32bit archs. * Update the new percpu stat counters directly and implement blkcg_rstat_flush() to implement propagation. * blkg_print_stat() can now bring the stats up to date by calling cgroup_rstat_flush() and print them instead of directly summing up all descendants. * It now allocates 96 bytes per cpu. It used to be 40 bytes. Signed-off-by: Tejun Heo <tj@kernel.org> Cc: Dan Schatzberg <dschatzberg@fb.com> Cc: Daniel Xu <dlxu@fb.com> Signed-off-by: Jens Axboe <axboe@kernel.dk>
Diffstat (limited to 'block')
-rw-r--r--block/blk-cgroup.c124
1 files changed, 99 insertions, 25 deletions
diff --git a/block/blk-cgroup.c b/block/blk-cgroup.c
index e7e93377e320..b3429be62057 100644
--- a/block/blk-cgroup.c
+++ b/block/blk-cgroup.c
@@ -80,8 +80,7 @@ static void blkg_free(struct blkcg_gq *blkg)
if (blkg->pd[i])
blkcg_policy[i]->pd_free_fn(blkg->pd[i]);
- blkg_rwstat_exit(&blkg->stat_ios);
- blkg_rwstat_exit(&blkg->stat_bytes);
+ free_percpu(blkg->iostat_cpu);
percpu_ref_exit(&blkg->refcnt);
kfree(blkg);
}
@@ -146,7 +145,7 @@ static struct blkcg_gq *blkg_alloc(struct blkcg *blkcg, struct request_queue *q,
gfp_t gfp_mask)
{
struct blkcg_gq *blkg;
- int i;
+ int i, cpu;
/* alloc and init base part */
blkg = kzalloc_node(sizeof(*blkg), gfp_mask, q->node);
@@ -156,8 +155,8 @@ static struct blkcg_gq *blkg_alloc(struct blkcg *blkcg, struct request_queue *q,
if (percpu_ref_init(&blkg->refcnt, blkg_release, 0, gfp_mask))
goto err_free;
- if (blkg_rwstat_init(&blkg->stat_bytes, gfp_mask) ||
- blkg_rwstat_init(&blkg->stat_ios, gfp_mask))
+ blkg->iostat_cpu = alloc_percpu_gfp(struct blkg_iostat_set, gfp_mask);
+ if (!blkg->iostat_cpu)
goto err_free;
blkg->q = q;
@@ -167,6 +166,10 @@ static struct blkcg_gq *blkg_alloc(struct blkcg *blkcg, struct request_queue *q,
INIT_WORK(&blkg->async_bio_work, blkg_async_bio_workfn);
blkg->blkcg = blkcg;
+ u64_stats_init(&blkg->iostat.sync);
+ for_each_possible_cpu(cpu)
+ u64_stats_init(&per_cpu_ptr(blkg->iostat_cpu, cpu)->sync);
+
for (i = 0; i < BLKCG_MAX_POLS; i++) {
struct blkcg_policy *pol = blkcg_policy[i];
struct blkg_policy_data *pd;
@@ -393,7 +396,6 @@ struct blkcg_gq *blkg_lookup_create(struct blkcg *blkcg,
static void blkg_destroy(struct blkcg_gq *blkg)
{
struct blkcg *blkcg = blkg->blkcg;
- struct blkcg_gq *parent = blkg->parent;
int i;
lockdep_assert_held(&blkg->q->queue_lock);
@@ -410,11 +412,6 @@ static void blkg_destroy(struct blkcg_gq *blkg)
pol->pd_offline_fn(blkg->pd[i]);
}
- if (parent) {
- blkg_rwstat_add_aux(&parent->stat_bytes, &blkg->stat_bytes);
- blkg_rwstat_add_aux(&parent->stat_ios, &blkg->stat_ios);
- }
-
blkg->online = false;
radix_tree_delete(&blkcg->blkg_tree, blkg->q->id);
@@ -464,7 +461,7 @@ static int blkcg_reset_stats(struct cgroup_subsys_state *css,
{
struct blkcg *blkcg = css_to_blkcg(css);
struct blkcg_gq *blkg;
- int i;
+ int i, cpu;
mutex_lock(&blkcg_pol_mutex);
spin_lock_irq(&blkcg->lock);
@@ -475,8 +472,12 @@ static int blkcg_reset_stats(struct cgroup_subsys_state *css,
* anyway. If you get hit by a race, retry.
*/
hlist_for_each_entry(blkg, &blkcg->blkg_list, blkcg_node) {
- blkg_rwstat_reset(&blkg->stat_bytes);
- blkg_rwstat_reset(&blkg->stat_ios);
+ for_each_possible_cpu(cpu) {
+ struct blkg_iostat_set *bis =
+ per_cpu_ptr(blkg->iostat_cpu, cpu);
+ memset(bis, 0, sizeof(*bis));
+ }
+ memset(&blkg->iostat, 0, sizeof(blkg->iostat));
for (i = 0; i < BLKCG_MAX_POLS; i++) {
struct blkcg_policy *pol = blkcg_policy[i];
@@ -840,16 +841,18 @@ static int blkcg_print_stat(struct seq_file *sf, void *v)
struct blkcg *blkcg = css_to_blkcg(seq_css(sf));
struct blkcg_gq *blkg;
+ cgroup_rstat_flush(blkcg->css.cgroup);
rcu_read_lock();
hlist_for_each_entry_rcu(blkg, &blkcg->blkg_list, blkcg_node) {
+ struct blkg_iostat_set *bis = &blkg->iostat;
const char *dname;
char *buf;
- struct blkg_rwstat_sample rwstat;
u64 rbytes, wbytes, rios, wios, dbytes, dios;
size_t size = seq_get_buf(sf, &buf), off = 0;
int i;
bool has_stats = false;
+ unsigned seq;
spin_lock_irq(&blkg->q->queue_lock);
@@ -868,17 +871,16 @@ static int blkcg_print_stat(struct seq_file *sf, void *v)
*/
off += scnprintf(buf+off, size-off, "%s ", dname);
- blkg_rwstat_recursive_sum(blkg, NULL,
- offsetof(struct blkcg_gq, stat_bytes), &rwstat);
- rbytes = rwstat.cnt[BLKG_RWSTAT_READ];
- wbytes = rwstat.cnt[BLKG_RWSTAT_WRITE];
- dbytes = rwstat.cnt[BLKG_RWSTAT_DISCARD];
+ do {
+ seq = u64_stats_fetch_begin(&bis->sync);
- blkg_rwstat_recursive_sum(blkg, NULL,
- offsetof(struct blkcg_gq, stat_ios), &rwstat);
- rios = rwstat.cnt[BLKG_RWSTAT_READ];
- wios = rwstat.cnt[BLKG_RWSTAT_WRITE];
- dios = rwstat.cnt[BLKG_RWSTAT_DISCARD];
+ rbytes = bis->cur.bytes[BLKG_IOSTAT_READ];
+ wbytes = bis->cur.bytes[BLKG_IOSTAT_WRITE];
+ dbytes = bis->cur.bytes[BLKG_IOSTAT_DISCARD];
+ rios = bis->cur.ios[BLKG_IOSTAT_READ];
+ wios = bis->cur.ios[BLKG_IOSTAT_WRITE];
+ dios = bis->cur.ios[BLKG_IOSTAT_DISCARD];
+ } while (u64_stats_fetch_retry(&bis->sync, seq));
if (rbytes || wbytes || rios || wios) {
has_stats = true;
@@ -1214,6 +1216,77 @@ static int blkcg_can_attach(struct cgroup_taskset *tset)
return ret;
}
+static void blkg_iostat_set(struct blkg_iostat *dst, struct blkg_iostat *src)
+{
+ int i;
+
+ for (i = 0; i < BLKG_IOSTAT_NR; i++) {
+ dst->bytes[i] = src->bytes[i];
+ dst->ios[i] = src->ios[i];
+ }
+}
+
+static void blkg_iostat_add(struct blkg_iostat *dst, struct blkg_iostat *src)
+{
+ int i;
+
+ for (i = 0; i < BLKG_IOSTAT_NR; i++) {
+ dst->bytes[i] += src->bytes[i];
+ dst->ios[i] += src->ios[i];
+ }
+}
+
+static void blkg_iostat_sub(struct blkg_iostat *dst, struct blkg_iostat *src)
+{
+ int i;
+
+ for (i = 0; i < BLKG_IOSTAT_NR; i++) {
+ dst->bytes[i] -= src->bytes[i];
+ dst->ios[i] -= src->ios[i];
+ }
+}
+
+static void blkcg_rstat_flush(struct cgroup_subsys_state *css, int cpu)
+{
+ struct blkcg *blkcg = css_to_blkcg(css);
+ struct blkcg_gq *blkg;
+
+ rcu_read_lock();
+
+ hlist_for_each_entry_rcu(blkg, &blkcg->blkg_list, blkcg_node) {
+ struct blkcg_gq *parent = blkg->parent;
+ struct blkg_iostat_set *bisc = per_cpu_ptr(blkg->iostat_cpu, cpu);
+ struct blkg_iostat cur, delta;
+ unsigned seq;
+
+ /* fetch the current per-cpu values */
+ do {
+ seq = u64_stats_fetch_begin(&bisc->sync);
+ blkg_iostat_set(&cur, &bisc->cur);
+ } while (u64_stats_fetch_retry(&bisc->sync, seq));
+
+ /* propagate percpu delta to global */
+ u64_stats_update_begin(&blkg->iostat.sync);
+ blkg_iostat_set(&delta, &cur);
+ blkg_iostat_sub(&delta, &bisc->last);
+ blkg_iostat_add(&blkg->iostat.cur, &delta);
+ blkg_iostat_add(&bisc->last, &delta);
+ u64_stats_update_end(&blkg->iostat.sync);
+
+ /* propagate global delta to parent */
+ if (parent) {
+ u64_stats_update_begin(&parent->iostat.sync);
+ blkg_iostat_set(&delta, &blkg->iostat.cur);
+ blkg_iostat_sub(&delta, &blkg->iostat.last);
+ blkg_iostat_add(&parent->iostat.cur, &delta);
+ blkg_iostat_add(&blkg->iostat.last, &delta);
+ u64_stats_update_end(&parent->iostat.sync);
+ }
+ }
+
+ rcu_read_unlock();
+}
+
static void blkcg_bind(struct cgroup_subsys_state *root_css)
{
int i;
@@ -1246,6 +1319,7 @@ struct cgroup_subsys io_cgrp_subsys = {
.css_offline = blkcg_css_offline,
.css_free = blkcg_css_free,
.can_attach = blkcg_can_attach,
+ .css_rstat_flush = blkcg_rstat_flush,
.bind = blkcg_bind,
.dfl_cftypes = blkcg_files,
.legacy_cftypes = blkcg_legacy_files,