hulk inclusion category: feature bugzilla: https://gitee.com/openeuler/kernel/issues/ICA1GK -------------------------------- Add a kfunc to get iostat from a blkcg, which treats cgroup v1, v2 and different block policy differently. Signed-off-by: GONG Ruiqi <gongruiqi1@huawei.com> --- block/Kconfig.iosched | 3 + block/blk-cgroup.c | 159 +++++++++++++++++++++++++++++++++++++++++ kernel/bpf-rvi/Kconfig | 1 + 3 files changed, 163 insertions(+) diff --git a/block/Kconfig.iosched b/block/Kconfig.iosched index 27f11320b8d1..392b03dc3350 100644 --- a/block/Kconfig.iosched +++ b/block/Kconfig.iosched @@ -44,4 +44,7 @@ config BFQ_CGROUP_DEBUG Enable some debugging help. Currently it exports additional stat files in a cgroup which can be useful for debugging. +config BPF_RVI_BLK_BFQ + bool + endmenu diff --git a/block/blk-cgroup.c b/block/blk-cgroup.c index 37e0b92e2e87..85e7e3307837 100644 --- a/block/blk-cgroup.c +++ b/block/blk-cgroup.c @@ -29,10 +29,19 @@ #include <linux/resume_user_mode.h> #include <linux/psi.h> #include <linux/part_stat.h> +#ifdef CONFIG_BPF_RVI +#include <linux/bpf.h> +#include <linux/btf.h> +#include <linux/btf_ids.h> +#endif + #include "blk.h" #include "blk-cgroup.h" #include "blk-ioprio.h" #include "blk-throttle.h" +#ifdef CONFIG_BPF_RVI +#include "bfq-iosched.h" +#endif static void __blkcg_rstat_flush(struct blkcg *blkcg, int cpu); @@ -2208,5 +2217,155 @@ bool blk_cgroup_congested(void) return ret; } +#ifdef CONFIG_BPF_RVI +struct blkg_rw_iostat { + struct blkg_rwstat_sample throttle_bytes; + struct blkg_rwstat_sample throttle_ios; + struct blkg_rwstat_sample bfq_bytes; + struct blkg_rwstat_sample bfq_ios; + struct blkg_rwstat_sample bfq_service_time; + struct blkg_rwstat_sample bfq_wait_time; + struct blkg_rwstat_sample bfq_merged; + struct blkg_iostat v2_iostat; +}; + +/* + * Getting: + * + * - "throttle.io_{service_bytes,serviced}_recursive" + * via offsetof(struct throtl_grp, stat_{bytes,ios}) + * - "bfq.io_{merged,{wait,service}_time,service_bytes,serviced}_recursive" + * via offsetof(struct bfq_group, stats.{merged,{wait,service}_time,bytes,ios}) + */ +static void blkcg_get_one_stat_v1(struct blkcg_gq *blkg, struct blkg_rw_iostat *iostat) +{ + struct blkcg_policy *pol; + +#ifdef CONFIG_BLK_DEV_THROTTLING // what blkcg_policy_throtl depends on + pol = &blkcg_policy_throtl; + if (blkcg_policy_enabled(blkg->q, pol)) { + // throttle.io_service_bytes_recursive + blkg_rwstat_recursive_sum(blkg, pol, + offsetof(struct throtl_grp, stat_bytes), + &iostat->throttle_bytes); + // throttle.io_serviced_recursive + blkg_rwstat_recursive_sum(blkg, pol, + offsetof(struct throtl_grp, stat_ios), + &iostat->throttle_ios); + } +#endif + + /* + * CONFIG_BPF_RVI_BLK_BFQ: blkcg_policy_bfq is in block/bfq-cgroup.c, which could be + * built as a module if CONFIG_IOSCHED_BFQ=m + * CONFIG_BFQ_GROUP_IOSCHED: what struct bfq_group.stats depends on + */ +#if defined(CONFIG_BPF_RVI_BLK_BFQ) && defined(CONFIG_BFQ_GROUP_IOSCHED) + pol = &blkcg_policy_bfq; + if (blkcg_policy_enabled(blkg->q, pol)) { + // bfq.io_service_bytes_recursive + blkg_rwstat_recursive_sum(blkg, pol, + offsetof(struct bfq_group, stats.bytes), + &iostat->bfq_bytes); + // bfq.io_serviced_recursive + blkg_rwstat_recursive_sum(blkg, pol, + offsetof(struct bfq_group, stats.ios), + &iostat->bfq_ios); +#ifdef CONFIG_BFQ_CGROUP_DEBUG + // bfq.io_service_time_recursive + blkg_rwstat_recursive_sum(blkg, pol, + offsetof(struct bfq_group, stats.service_time), + &iostat->bfq_service_time); + // bfq.io_wait_time_recursive + blkg_rwstat_recursive_sum(blkg, pol, + offsetof(struct bfq_group, stats.wait_time), + &iostat->bfq_wait_time); + // bfq.io_merged_recursive + blkg_rwstat_recursive_sum(blkg, pol, + offsetof(struct bfq_group, stats.merged), + &iostat->bfq_merged); +#endif + } +#endif +} + +/* Reference: blkcg_print_one_stat() */ +static void blkcg_get_one_stat_v2(struct blkcg_gq *blkg, struct blkg_rw_iostat *iostat) +{ + struct blkg_iostat_set *bis = &blkg->iostat; + unsigned int seq; + + if (!blkg->online) + return; + + do { + seq = u64_stats_fetch_begin(&bis->sync); + iostat->v2_iostat = bis->cur; + } while (u64_stats_fetch_retry(&bis->sync, seq)); +} + +/* + * Basically inmitating: + * + * - v1: + * - tg_print_rwstat_recursive() in block/blk-throttle.c + * - bfqg_print_rwstat_recursive() in block/bfq-cgroup.c + * - v2: + * - blkcg_print_stat() + * + * without the final printing (e.g. the __blkg_prfill_rwstat() part). + * + * Note that a subsystem can only exist in either cgroup v1 or v2 at the same time. + */ +__bpf_kfunc void bpf_blkcg_get_dev_iostat(struct blkcg *blkcg, int major, int minor, + struct blkg_rw_iostat *iostat, bool is_v2) +{ + struct blkcg_gq *blkg; + char dev_name[64]; + + if (!blkcg || !iostat) + return; + + if (is_v2) { + if (blkcg == &blkcg_root) + blkcg_fill_root_iostats(); + else + cgroup_rstat_flush_atomic(blkcg->css.cgroup); + } + + // memset(iostat, 0, sizeof(*iostat)); + snprintf(dev_name, sizeof(dev_name), "%d:%d", major, minor); + rcu_read_lock(); + hlist_for_each_entry_rcu(blkg, &blkcg->blkg_list, blkcg_node) { + if (strcmp(dev_name, blkg_dev_name(blkg))) + continue; + spin_lock_irq(&blkg->q->queue_lock); + if (is_v2) + blkcg_get_one_stat_v2(blkg, iostat); + else + blkcg_get_one_stat_v1(blkg, iostat); + spin_unlock_irq(&blkg->q->queue_lock); + break; + } + rcu_read_unlock(); +} + +BTF_SET8_START(bpf_blkcg_kfunc_ids) +BTF_ID_FLAGS(func, bpf_blkcg_get_dev_iostat) +BTF_SET8_END(bpf_blkcg_kfunc_ids) + +static const struct btf_kfunc_id_set bpf_blkcg_kfunc_set = { + .owner = THIS_MODULE, + .set = &bpf_blkcg_kfunc_ids, +}; + +static int __init bpf_blkcg_kfunc_init(void) +{ + return register_btf_kfunc_id_set(BPF_PROG_TYPE_TRACING, + &bpf_blkcg_kfunc_set); +} +late_initcall(bpf_blkcg_kfunc_init); +#endif /* CONFIG_BPF_RVI */ + module_param(blkcg_debug_stats, bool, 0644); MODULE_PARM_DESC(blkcg_debug_stats, "True if you want debug stats, false if not"); diff --git a/kernel/bpf-rvi/Kconfig b/kernel/bpf-rvi/Kconfig index 0e356ae6fc85..8a9cbac36a0c 100644 --- a/kernel/bpf-rvi/Kconfig +++ b/kernel/bpf-rvi/Kconfig @@ -7,6 +7,7 @@ config BPF_RVI depends on BPF_SYSCALL depends on BPF_JIT depends on CPUSETS + select BPF_RVI_BLK_BFQ if IOSCHED_BFQ = y # built-in required default n help A resource view is a bundle of interfaces under /proc and /sys -- 2.25.1