blk-cgroup: show global disk stats in root cgroup io.stat
Subsystems (4)
PATCH 1/1 blk-cgroup: show global disk stats in root cgroup io.stat
In order to improve consistency and usability in cgroup stat accounting, we would like to support the root cgroup's io.stat.
Since the root cgroup has processes doing io even if the system has no explicitly created cgroups, we need to be careful to avoid overhead in that case. For that reason, the rstat algorithms don't handle the root cgroup, so just turning the file on wouldn't give correct statistics.
To get around this, we simulate flushing the iostat struct by filling it out directly from global disk stats. The result is a root cgroup io.stat file consistent with both /proc/diskstats and io.stat.
--- Documentation/admin-guide/cgroup-v2.rst | 3 +- block/blk-cgroup.c | 199 +++++++++++++++--------- block/genhd.c | 4 +- include/linux/genhd.h | 1 + 4 files changed, 129 insertions(+), 78 deletions(-)
Documentation/admin-guide/cgroup-v2.rst +1 -2
@@ -1465,8 +1465,7 @@ IO Interface Files ~~~~~~~~~~~~~~~~~~ io.stat - A read-only nested-keyed file which exists on non-root - cgroups. + A read-only nested-keyed file. Lines are keyed by $MAJ:$MIN device numbers and not ordered. The following nested keys are defined.
block/blk-cgroup.c +126 -73
222 lines not on this page · raw message
block/genhd.c +1 -3
@@ -38,8 +38,6 @@ static struct kobject *block_depr;static DEFINE_SPINLOCK (ext_devt_lock);static DEFINE_IDR (ext_devt_idr); -static const struct device_type disk_type; -static void disk_check_events (struct disk_events *ev,unsigned int *clearing_ptr);static void disk_alloc_events (struct gendisk *disk);
@@ -1566,7 +1564,7 @@ static char *block_devnode(struct device *dev, umode_t *mode,return NULL ; } -static const struct device_type disk_type = { +const struct device_type disk_type = { .name ="disk" , .groups = disk_attr_groups, .release = disk_release,
include/linux/genhd.h +1 -0
@@ -26,6 +26,7 @@#define disk_to_dev (disk) (&(disk)->part0.__dev)#define part_to_dev (part) (&((part)->__dev)) +extern const struct device_type disk_type;extern struct device_type part_type;extern struct class block_class;
Signature · 1 line
2.24.1
Tejun Heo
maintainerHello, Boris.
Quoted 14 lines · Boris Burkov
In order to improve consistency and usability in cgroup stat accounting, we would like to support the root cgroup's io.stat.
Since the root cgroup has processes doing io even if the system has no explicitly created cgroups, we need to be careful to avoid overhead in that case. For that reason, the rstat algorithms don't handle the root cgroup, so just turning the file on wouldn't give correct statistics.
To get around this, we simulate flushing the iostat struct by filling it out directly from global disk stats. The result is a root cgroup io.stat file consistent with both
/proc/diskstatsand io.stat.Signed-off-by Boris Burkov <[email protected]>Suggested-by Tejun Heo <[email protected]>
...
Review
Quoted 2 lines
(quoted diff) +2
-0+static void blkg_iostat_set(struct blkg_iostat *dst, struct blkg_iostat *src) +{
Can you please separate out code reorganization to a separate patch so that the actual change can be reviewed clearly?
Quoted 11 lines
+/*
+ * The rstat algorithms intentionally don't handle the root cgroup to avoid
+ * incurring overhead when no cgroups are defined. For that reason,
+ *cgroup_rstat_flushinblkcg_print_statdoes not actually fill out the
+ * iostat in the root cgroup'sblkcg_gq.
+ *
+ * However, we would like to re-use the printing code between the root and
+ * non-root cgroups to the extent possible. For that reason, we simulate
+ * flushing the root cgroup's stats by explicitly filling in the iostat
+ * with disk level statistics.
+ */
This is clever and neat.
Quoted 38 lines
block/blk-cgroup.c +38
-0+static void blkcg_fill_root_iostats (void ) +{ +struct class_dev_iter iter; +struct device *dev; + +class_dev_iter_init (&iter, &block_class,NULL , &disk_type); +while ((dev =class_dev_iter_next (&iter))) { +struct gendisk *disk =dev_to_disk (dev); +struct hd_struct *part =disk_get_part (disk,0 ); +struct blkcg_gq *blkg =blk_queue_root_blkg (disk->queue); +struct blkg_iostat tmp; +int cpu; + +memset (&tmp,0 ,sizeof (tmp)); +for_each_possible_cpu (cpu) { +struct disk_stats *cpu_dkstats; + + cpu_dkstats =per_cpu_ptr (part->dkstats, cpu); + tmp.ios[BLKG_IOSTAT_READ ] += + cpu_dkstats->ios[STAT_READ ]; + tmp.ios[BLKG_IOSTAT_WRITE ] += + cpu_dkstats->ios[STAT_WRITE ]; + tmp.ios[BLKG_IOSTAT_DISCARD ] += + cpu_dkstats->ios[STAT_DISCARD ]; +// convert sectors to bytes + tmp.bytes[BLKG_IOSTAT_READ ] += + cpu_dkstats->sectors[STAT_READ ] <<9 ; + tmp.bytes[BLKG_IOSTAT_WRITE ] += + cpu_dkstats->sectors[STAT_WRITE ] <<9 ; + tmp.bytes[BLKG_IOSTAT_DISCARD ] += + cpu_dkstats->sectors[STAT_DISCARD ] <<9 ; + +u64_stats_update_begin (&blkg->iostat.sync); +blkg_iostat_set (&blkg->iostat.cur, &tmp); +u64_stats_update_end (&blkg->iostat.sync); + } + } +}
...
Quoted 27 lines
block/genhd.c +1
-3@@ -38,8 +38,6 @@ static struct kobject *block_depr;static DEFINE_SPINLOCK (ext_devt_lock);static DEFINE_IDR (ext_devt_idr); -static const struct device_type disk_type; -static void disk_check_events (struct disk_events *ev,unsigned int *clearing_ptr);static void disk_alloc_events (struct gendisk *disk);@@ -1566,7 +1564,7 @@ static char *block_devnode(struct device *dev, umode_t *mode,return NULL ; } -static const struct device_type disk_type = { +const struct device_type disk_type = { .name ="disk" , .groups = disk_attr_groups, .release = disk_release,include/linux/genhd.h +0
-0
include/linux/genhd.h line 29
Quoted 4 lines
include/linux/genhd.h +1
-0@@ -26,6 +26,7 @@#define disk_to_dev (disk) (&(disk)->part0.__dev)#define part_to_dev (part) (&((part)->__dev)) +extern const struct device_type disk_type;
So, this is fine but I'd explicitly mention it in the patch description.
Thanks.
Signature · 1 line
tejun
PATCH 1/2 blk-cgroup: make iostat functions visible to stat printing
Previously, the code which printed io.stat only needed access to the generic rstat flushing code, but since we plan to write some more specific code for preparing root cgroup stats, we need to manipulate iostat structs directly. Since declaring static functions ahead does not seem like common practice in this file, simply move the iostat functions up. We only plan to use blkg_iostat_set, but it seems better to keep them all together.
--- block/blk-cgroup.c | 142 ++++++++++++++++++++++----------------------- 1 file changed, 71 insertions(+), 71 deletions(-)
block/blk-cgroup.c +71 -71
@@ -739,6 +739,77 @@ void blkg_conf_finish(struct blkg_conf_ctx *ctx) }EXPORT_SYMBOL_GPL (blkg_conf_finish); +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 int 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 int blkcg_print_stat (struct seq_file *sf,void *v) {struct blkcg *blkcg =css_to_blkcg (seq_css (sf));
@@ -1114,77 +1185,6 @@ 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;
Signature · 1 line
2.24.1
Re: [PATCH 1/2 blk-cgroup/for-5.8] blk-cgroup: make iostat functions visible to stat printing
Quoted 9 lines · Boris Burkov
Previously, the code which printed io.stat only needed access to the generic rstat flushing code, but since we plan to write some more specific code for preparing root cgroup stats, we need to manipulate iostat structs directly. Since declaring static functions ahead does not seem like common practice in this file, simply move the iostat functions up. We only plan to use
blkg_iostat_set, but it seems better to keep them all together.Signed-off-by Boris Burkov <[email protected]>
Thanks.
Signature · 1 line
tejun
PATCH 2/2 blk-cgroup: show global disk stats in root cgroup io.stat
In reply to Tejun HeoIn order to improve consistency and usability in cgroup stat accounting, we would like to support the root cgroup's io.stat.
Since the root cgroup has processes doing io even if the system has no explicitly created cgroups, we need to be careful to avoid overhead in that case. For that reason, the rstat algorithms don't handle the root cgroup, so just turning the file on wouldn't give correct statistics.
To get around this, we simulate flushing the iostat struct by filling it out directly from global disk stats. The result is a root cgroup io.stat file consistent with both /proc/diskstats and io.stat.
Note that in order to collect the disk stats, we needed to iterate over devices. To facilitate that, we had to change the linkage of a disk_type to external so that it can be used from blk-cgroup.c to iterate over disks.
--- Documentation/admin-guide/cgroup-v2.rst | 3 +- block/blk-cgroup.c | 57 ++++++++++++++++++++++++- block/genhd.c | 4 +- include/linux/genhd.h | 1 + 4 files changed, 58 insertions(+), 7 deletions(-)
Documentation/admin-guide/cgroup-v2.rst +1 -2
@@ -1465,8 +1465,7 @@ IO Interface Files ~~~~~~~~~~~~~~~~~~ io.stat - A read-only nested-keyed file which exists on non-root - cgroups. + A read-only nested-keyed file. Lines are keyed by $MAJ:$MIN device numbers and not ordered. The following nested keys are defined.
block/blk-cgroup.c +55 -2
74 lines not on this page · raw message
block/genhd.c +1 -3
@@ -38,8 +38,6 @@ static struct kobject *block_depr;static DEFINE_SPINLOCK (ext_devt_lock);static DEFINE_IDR (ext_devt_idr); -static const struct device_type disk_type; -static void disk_check_events (struct disk_events *ev,unsigned int *clearing_ptr);static void disk_alloc_events (struct gendisk *disk);
@@ -1566,7 +1564,7 @@ static char *block_devnode(struct device *dev, umode_t *mode,return NULL ; } -static const struct device_type disk_type = { +const struct device_type disk_type = { .name ="disk" , .groups = disk_attr_groups, .release = disk_release,
include/linux/genhd.h +1 -0
@@ -26,6 +26,7 @@#define disk_to_dev (disk) (&(disk)->part0.__dev)#define part_to_dev (part) (&((part)->__dev)) +extern const struct device_type disk_type;extern struct device_type part_type;extern struct class block_class;
Signature · 1 line
2.24.1
Quoted 12 lines · Boris Burkov
In order to improve consistency and usability in cgroup stat accounting, we would like to support the root cgroup's io.stat.
Since the root cgroup has processes doing io even if the system has no explicitly created cgroups, we need to be careful to avoid overhead in that case. For that reason, the rstat algorithms don't handle the root cgroup, so just turning the file on wouldn't give correct statistics.
To get around this, we simulate flushing the iostat struct by filling it out directly from global disk stats. The result is a root cgroup io.stat file consistent with both
/proc/diskstatsand io.stat.
Quoted 7 lines
Note that in order to collect the disk stats, we needed to iterate over devices. To facilitate that, we had to change the linkage of a
disk_typeto external so that it can be used fromblk-cgroup.cto iterate over disks.Signed-off-by Boris Burkov <[email protected]>Suggested-by Tejun Heo <[email protected]>
Thanks.
Signature · 1 line
tejun