From: Guan Jing guanjing6@huawei.com

hulk inclusion category: feature bugzilla: https://gitee.com/openeuler/kernel/issues/I8OIT1

--------------------------------

Add a tag for the task, useful to identify the special task. User can use the file system interface to mark different tags for specific workloads. The kernel subsystems can use the set_* helpers to mark it too. The bpf prog obtains the tags to detect different workloads.

Signed-off-by: Chen Hui judy.chenhui@huawei.com Signed-off-by: Ren Zhijie renzhijie2@huawei.com Signed-off-by: Hui Tang tanghui20@huawei.com Signed-off-by: Guan Jing guanjing6@huawei.com --- include/linux/sched.h | 5 +++++ init/init_task.c | 3 +++ kernel/sched/core.c | 3 +++ 3 files changed, 11 insertions(+)

diff --git a/include/linux/sched.h b/include/linux/sched.h index bd9031f5772c..c841cf1c150a 100644 --- a/include/linux/sched.h +++ b/include/linux/sched.h @@ -1547,6 +1547,10 @@ struct task_struct { struct user_event_mm *user_event_mm; #endif

+#ifdef CONFIG_BPF_SCHED + long tag; +#endif + #ifdef CONFIG_QOS_SCHED_DYNAMIC_AFFINITY cpumask_t *prefer_cpus; const cpumask_t *select_cpus; @@ -1555,6 +1559,7 @@ struct task_struct { #ifdef CONFIG_PSI_FINE_GRAINED int memstall_type; #endif + /* * New fields for task_struct should be added above here, so that * they are included in the randomized portion of task_struct. diff --git a/init/init_task.c b/init/init_task.c index ac0c5850f74b..2101c6e3432d 100644 --- a/init/init_task.c +++ b/init/init_task.c @@ -213,6 +213,9 @@ struct task_struct init_task #ifdef CONFIG_SECCOMP_FILTER .seccomp = { .filter_count = ATOMIC_INIT(0) }, #endif +#ifdef CONFIG_BPF_SCHED + .tag = 0, +#endif }; EXPORT_SYMBOL(init_task);

diff --git a/kernel/sched/core.c b/kernel/sched/core.c index bafb7b440263..427a1a67ed11 100644 --- a/kernel/sched/core.c +++ b/kernel/sched/core.c @@ -4542,6 +4542,9 @@ static void __sched_fork(unsigned long clone_flags, struct task_struct *p) p->migration_pending = NULL; #endif init_sched_mm_cid(p); +#ifdef CONFIG_BPF_SCHED + p->tag = 0; +#endif }

DEFINE_STATIC_KEY_FALSE(sched_numa_balancing);

From: Guan Jing guanjing6@huawei.com

hulk inclusion category: feature bugzilla: https://gitee.com/openeuler/kernel/issues/I8OIT1

--------------------------------

Add user interface of task group tag, bridges the information gap between user-mode and kernel-mode.

Signed-off-by: Chen Hui judy.chenhui@huawei.com Signed-off-by: Ren Zhijie renzhijie2@huawei.com Signed-off-by: Hui Tang tanghui20@huawei.com Signed-off-by: Guan Jing guanjing6@huawei.com --- include/linux/sched.h | 3 ++ kernel/sched/core.c | 81 +++++++++++++++++++++++++++++++++++++++++++ kernel/sched/sched.h | 3 ++ 3 files changed, 87 insertions(+)

diff --git a/include/linux/sched.h b/include/linux/sched.h index c841cf1c150a..857c0d8df803 100644 --- a/include/linux/sched.h +++ b/include/linux/sched.h @@ -2476,7 +2476,10 @@ void rseq_syscall(struct pt_regs *regs); static inline void rseq_syscall(struct pt_regs *regs) { } +#endif

+#ifdef CONFIG_BPF_SCHED +extern void sched_settag(struct task_struct *tsk, s64 tag); #endif

#ifdef CONFIG_SCHED_CORE diff --git a/kernel/sched/core.c b/kernel/sched/core.c index d6a27c003949..dc48b7dd9036 100644 --- a/kernel/sched/core.c +++ b/kernel/sched/core.c @@ -11422,6 +11422,80 @@ static inline s64 cpu_qos_read(struct cgroup_subsys_state *css, } #endif

+#ifdef CONFIG_BPF_SCHED +void sched_settag(struct task_struct *tsk, s64 tag) +{ + int queued, running, queue_flags = + DEQUEUE_SAVE | DEQUEUE_MOVE | DEQUEUE_NOCLOCK; + struct rq_flags rf; + struct rq *rq; + + if (tsk->tag == tag) + return; + + rq = task_rq_lock(tsk, &rf); + + running = task_current(rq, tsk); + queued = task_on_rq_queued(tsk); + + update_rq_clock(rq); + if (queued) + dequeue_task(rq, tsk, queue_flags); + if (running) + put_prev_task(rq, tsk); + + tsk->tag = tag; + + if (queued) + enqueue_task(rq, tsk, queue_flags); + if (running) + set_next_task(rq, tsk); + + task_rq_unlock(rq, tsk, &rf); +} + +int tg_change_tag(struct task_group *tg, void *data) +{ + struct css_task_iter it; + struct task_struct *tsk; + s64 tag = *(s64 *)data; + struct cgroup_subsys_state *css = &tg->css; + + tg->tag = tag; + + css_task_iter_start(css, 0, &it); + while ((tsk = css_task_iter_next(&it))) + sched_settag(tsk, tag); + css_task_iter_end(&it); + + return 0; +} + +static int cpu_tag_write(struct cgroup_subsys_state *css, + struct cftype *cftype, s64 tag) +{ + struct task_group *tg = css_tg(css); + + if (tg == &root_task_group) + return -EINVAL; + + if (tg->tag == tag) + return 0; + + rcu_read_lock(); + walk_tg_tree_from(tg, tg_change_tag, tg_nop, (void *)(&tag)); + rcu_read_unlock(); + + return 0; +} + +static inline s64 cpu_tag_read(struct cgroup_subsys_state *css, + struct cftype *cft) +{ + return css_tg(css)->tag; +} +#endif + static struct cftype cpu_legacy_files[] = { #ifdef CONFIG_FAIR_GROUP_SCHED { @@ -11492,6 +11566,13 @@ static struct cftype cpu_legacy_files[] = { .read_s64 = cpu_qos_read, .write_s64 = cpu_qos_write, }, +#endif +#ifdef CONFIG_BPF_SCHED + { + .name = "tag", + .read_s64 = cpu_tag_read, + .write_s64 = cpu_tag_write, + }, #endif { } /* Terminate */ }; diff --git a/kernel/sched/sched.h b/kernel/sched/sched.h index 29f9a049fa28..de81c7f2ad93 100644 --- a/kernel/sched/sched.h +++ b/kernel/sched/sched.h @@ -456,6 +456,9 @@ static inline int walk_tg_tree(tg_visitor down, tg_visitor up, void *data) }

extern int tg_nop(struct task_group *tg, void *data); +#ifdef CONFIG_BPF_SCHED +extern int tg_change_tag(struct task_group *tg, void *data); +#endif

extern void free_fair_sched_group(struct task_group *tg); extern int alloc_fair_sched_group(struct task_group *tg, struct task_group *parent);

From: Guan Jing guanjing6@huawei.com

maillist inclusion category: feature bugzilla: https://gitee.com/openeuler/kernel/issues/I8OIT1

Reference: https://lore.kernel.org/all/20210916162451.709260-1-guro@fb.com/

-------------------

This commit introduces basic definitions and infrastructure for scheduler bpf programs. It defines the BPF_PROG_TYPE_SCHED program type and the BPF_SCHED attachment type.

The implementation is inspired by lsm bpf programs and is based on kretprobes. This will allow to add new hooks with a minimal changes to the kernel code and without any changes to libbpf/bpftool. It's very convenient as I anticipate a large number of private patches being used for a long time before (or if at all) reaching upstream.

Sched programs are expected to return an int, which meaning will be context defined.

This patch doesn't add any real scheduler hooks (only a stub), it will be done by following patches in the series.

Scheduler bpf programs as now are very restricted in what they can do: only the bpf_printk() helper is available. The scheduler context can impose significant restrictions on what's safe and what's not. So let's extend their abilities on case by case basis when a need arise.

Signed-off-by: Roman Gushchin guro@fb.com Signed-off-by: Chen Hui judy.chenhui@huawei.com Signed-off-by: Ren Zhijie renzhijie2@huawei.com Signed-off-by: Hui Tang tanghui20@huawei.com Signed-off-by: Guan Jing guanjing6@huawei.com --- include/linux/bpf_sched.h | 26 ++++++++++++++ include/linux/bpf_types.h | 4 +++ include/linux/sched_hook_defs.h | 2 ++ include/uapi/linux/bpf.h | 2 ++ kernel/bpf/btf.c | 3 ++ kernel/bpf/syscall.c | 30 ++++++++++++++++ kernel/bpf/trampoline.c | 3 ++ kernel/bpf/verifier.c | 23 ++++++++++++ kernel/sched/bpf_sched.c | 62 +++++++++++++++++++++++++++++++++ kernel/sched/build_utility.c | 4 +++ tools/include/uapi/linux/bpf.h | 2 ++ tools/lib/bpf/bpf.c | 1 + 12 files changed, 162 insertions(+) create mode 100644 include/linux/bpf_sched.h create mode 100644 include/linux/sched_hook_defs.h create mode 100644 kernel/sched/bpf_sched.c

diff --git a/include/linux/bpf_sched.h b/include/linux/bpf_sched.h new file mode 100644 index 000000000000..874393e6a6aa --- /dev/null +++ b/include/linux/bpf_sched.h @@ -0,0 +1,26 @@ +/* SPDX-License-Identifier: GPL-2.0 */ +#ifndef _LINUX_BPF_SCHED_H +#define _LINUX_BPF_SCHED_H + +#include <linux/bpf.h> + +#ifdef CONFIG_BPF_SCHED + +#define BPF_SCHED_HOOK(RET, DEFAULT, NAME, ...) \ + RET bpf_sched_##NAME(__VA_ARGS__); +#include <linux/sched_hook_defs.h> +#undef BPF_SCHED_HOOK + +int bpf_sched_verify_prog(struct bpf_verifier_log *vlog, + const struct bpf_prog *prog); + +#else /* !CONFIG_BPF_SCHED */ + +static inline int bpf_sched_verify_prog(struct bpf_verifier_log *vlog, + const struct bpf_prog *prog) +{ + return -EOPNOTSUPP; +} + +#endif /* CONFIG_BPF_SCHED */ +#endif /* _LINUX_BPF_SCHED_H */ diff --git a/include/linux/bpf_types.h b/include/linux/bpf_types.h index fc0d6f32c687..dd79463eea4e 100644 --- a/include/linux/bpf_types.h +++ b/include/linux/bpf_types.h @@ -83,6 +83,10 @@ BPF_PROG_TYPE(BPF_PROG_TYPE_SYSCALL, bpf_syscall, BPF_PROG_TYPE(BPF_PROG_TYPE_NETFILTER, netfilter, struct bpf_nf_ctx, struct bpf_nf_ctx) #endif +#ifdef CONFIG_BPF_SCHED +BPF_PROG_TYPE(BPF_PROG_TYPE_SCHED, bpf_sched, + void *, void *) +#endif /* CONFIG_BPF_SCHED */

BPF_MAP_TYPE(BPF_MAP_TYPE_ARRAY, array_map_ops) BPF_MAP_TYPE(BPF_MAP_TYPE_PERCPU_ARRAY, percpu_array_map_ops) diff --git a/include/linux/sched_hook_defs.h b/include/linux/sched_hook_defs.h new file mode 100644 index 000000000000..14344004e335 --- /dev/null +++ b/include/linux/sched_hook_defs.h @@ -0,0 +1,2 @@ +/* SPDX-License-Identifier: GPL-2.0 */ +BPF_SCHED_HOOK(int, 0, dummy, void) diff --git a/include/uapi/linux/bpf.h b/include/uapi/linux/bpf.h index 4924f0cde1bc..9dd0b85549b6 100644 --- a/include/uapi/linux/bpf.h +++ b/include/uapi/linux/bpf.h @@ -988,6 +988,7 @@ enum bpf_prog_type { BPF_PROG_TYPE_SK_LOOKUP, BPF_PROG_TYPE_SYSCALL, /* a program that can execute syscalls */ BPF_PROG_TYPE_NETFILTER, + BPF_PROG_TYPE_SCHED, };

enum bpf_attach_type { @@ -1040,6 +1041,7 @@ enum bpf_attach_type { BPF_TCX_INGRESS, BPF_TCX_EGRESS, BPF_TRACE_UPROBE_MULTI, + BPF_SCHED, __MAX_BPF_ATTACH_TYPE };

diff --git a/kernel/bpf/btf.c b/kernel/bpf/btf.c index 8090d7fb11ef..b8aa9e66e356 100644 --- a/kernel/bpf/btf.c +++ b/kernel/bpf/btf.c @@ -5982,6 +5982,9 @@ bool btf_ctx_access(int off, int size, enum bpf_access_type type, return true; t = btf_type_by_id(btf, t->type); break; +#ifdef CONFIG_BPF_SCHED + case BPF_SCHED: +#endif case BPF_MODIFY_RETURN: /* For now the BPF_MODIFY_RETURN can only be attached to * functions that return an int. diff --git a/kernel/bpf/syscall.c b/kernel/bpf/syscall.c index d77b2f8b9364..793dfbced1ac 100644 --- a/kernel/bpf/syscall.c +++ b/kernel/bpf/syscall.c @@ -2412,6 +2412,9 @@ bpf_prog_load_check_attach(enum bpf_prog_type prog_type, case BPF_PROG_TYPE_LSM: case BPF_PROG_TYPE_STRUCT_OPS: case BPF_PROG_TYPE_EXT: +#ifdef CONFIG_BPF_SCHED + case BPF_PROG_TYPE_SCHED: +#endif break; default: return -EINVAL; @@ -2539,6 +2542,9 @@ static bool is_perfmon_prog_type(enum bpf_prog_type prog_type) case BPF_PROG_TYPE_LSM: case BPF_PROG_TYPE_STRUCT_OPS: /* has access to struct sock */ case BPF_PROG_TYPE_EXT: /* extends any prog */ +#ifdef CONFIG_BPF_SCHED + case BPF_PROG_TYPE_SCHED: +#endif return true; default: return false; @@ -3115,6 +3121,14 @@ static int bpf_tracing_prog_attach(struct bpf_prog *prog, goto out_put_prog; } break; +#ifdef CONFIG_BPF_SCHED + case BPF_PROG_TYPE_SCHED: + if (prog->expected_attach_type != BPF_SCHED) { + err = -EINVAL; + goto out_put_prog; + } + break; +#endif default: err = -EINVAL; goto out_put_prog; @@ -3582,6 +3596,9 @@ static int bpf_raw_tp_link_attach(struct bpf_prog *prog, case BPF_PROG_TYPE_TRACING: case BPF_PROG_TYPE_EXT: case BPF_PROG_TYPE_LSM: +#ifdef CONFIG_BPF_SCHED + case BPF_PROG_TYPE_SCHED: +#endif if (user_tp_name) /* The attach point for this category of programs * should be specified via btf_id during program load. @@ -3717,6 +3734,10 @@ attach_type_to_prog_type(enum bpf_attach_type attach_type) case BPF_TCX_INGRESS: case BPF_TCX_EGRESS: return BPF_PROG_TYPE_SCHED_CLS; +#ifdef CONFIG_BPF_SCHED + case BPF_SCHED: + return BPF_PROG_TYPE_SCHED; +#endif default: return BPF_PROG_TYPE_UNSPEC; } @@ -3744,6 +3765,12 @@ static int bpf_prog_attach_check_attach_type(const struct bpf_prog *prog, -EINVAL : 0; case BPF_PROG_TYPE_EXT: return 0; +#ifdef CONFIG_BPF_SCHED + case BPF_PROG_TYPE_SCHED: + if (!capable(CAP_SYS_ADMIN)) + return -EPERM; + return 0; +#endif case BPF_PROG_TYPE_NETFILTER: if (attach_type != BPF_NETFILTER) return -EINVAL; @@ -4922,6 +4949,9 @@ static int link_create(union bpf_attr *attr, bpfptr_t uattr) ret = cgroup_bpf_link_attach(attr, prog); break; case BPF_PROG_TYPE_EXT: +#ifdef CONFIG_BPF_SCHED + case BPF_PROG_TYPE_SCHED: +#endif ret = bpf_tracing_prog_attach(prog, attr->link_create.target_fd, attr->link_create.target_btf_id, diff --git a/kernel/bpf/trampoline.c b/kernel/bpf/trampoline.c index e97aeda3a86b..e5b97eb226e8 100644 --- a/kernel/bpf/trampoline.c +++ b/kernel/bpf/trampoline.c @@ -493,6 +493,9 @@ static enum bpf_tramp_prog_type bpf_attach_type_to_tramp(struct bpf_prog *prog) switch (prog->expected_attach_type) { case BPF_TRACE_FENTRY: return BPF_TRAMP_FENTRY; +#ifdef CONFIG_BPF_SCHED + case BPF_SCHED: +#endif case BPF_MODIFY_RETURN: return BPF_TRAMP_MODIFY_RETURN; case BPF_TRACE_FEXIT: diff --git a/kernel/bpf/verifier.c b/kernel/bpf/verifier.c index 824531d4c262..d3090e8778e1 100644 --- a/kernel/bpf/verifier.c +++ b/kernel/bpf/verifier.c @@ -28,6 +28,10 @@ #include <linux/cpumask.h> #include <net/xdp.h>

+#ifdef CONFIG_BPF_SCHED +#include <linux/bpf_sched.h> +#endif + #include "disasm.h"

static const struct bpf_verifier_ops * const bpf_verifier_ops[] = { @@ -19453,6 +19457,9 @@ int bpf_check_attach_target(struct bpf_verifier_log *log, case BPF_LSM_CGROUP: case BPF_TRACE_FENTRY: case BPF_TRACE_FEXIT: +#ifdef CONFIG_BPF_SCHED + case BPF_SCHED: +#endif if (!btf_type_is_func(t)) { bpf_log(log, "attach_btf_id %u is not a function\n", btf_id); @@ -19627,10 +19634,18 @@ static int check_attach_btf_id(struct bpf_verifier_env *env) if (prog->type == BPF_PROG_TYPE_STRUCT_OPS) return check_struct_ops_btf_id(env);

+#ifdef CONFIG_BPF_SCHED + if (prog->type != BPF_PROG_TYPE_TRACING && + prog->type != BPF_PROG_TYPE_LSM && + prog->type != BPF_PROG_TYPE_EXT && + prog->type != BPF_PROG_TYPE_SCHED) + return 0; +#else if (prog->type != BPF_PROG_TYPE_TRACING && prog->type != BPF_PROG_TYPE_LSM && prog->type != BPF_PROG_TYPE_EXT) return 0; +#endif

ret = bpf_check_attach_target(&env->log, prog, tgt_prog, btf_id, &tgt_info); if (ret) @@ -19673,6 +19688,14 @@ static int check_attach_btf_id(struct bpf_verifier_env *env) return -EINVAL; }

+#ifdef CONFIG_BPF_SCHED + if (prog->type == BPF_PROG_TYPE_SCHED) { + ret = bpf_sched_verify_prog(&env->log, prog); + if (ret < 0) + return ret; + } +#endif + key = bpf_trampoline_compute_key(tgt_prog, prog->aux->attach_btf, btf_id); tr = bpf_trampoline_get(key, &tgt_info); if (!tr) diff --git a/kernel/sched/bpf_sched.c b/kernel/sched/bpf_sched.c new file mode 100644 index 000000000000..2360404d4a07 --- /dev/null +++ b/kernel/sched/bpf_sched.c @@ -0,0 +1,62 @@ +// SPDX-License-Identifier: GPL-2.0 +#include <linux/bpf.h> +#include <linux/cgroup.h> +#include <linux/bpf_verifier.h> +#include <linux/bpf_sched.h> +#include <linux/btf_ids.h> +#include "sched.h" + +/* + * For every hook declare a nop function where a BPF program can be attached. + */ +#define BPF_SCHED_HOOK(RET, DEFAULT, NAME, ...) \ +noinline RET bpf_sched_##NAME(__VA_ARGS__) \ +{ \ + return DEFAULT; \ +} + +#include <linux/sched_hook_defs.h> +#undef BPF_SCHED_HOOK + +#define BPF_SCHED_HOOK(RET, DEFAULT, NAME, ...) BTF_ID(func, bpf_sched_##NAME) +BTF_SET_START(bpf_sched_hooks) +#include <linux/sched_hook_defs.h> +#undef BPF_SCHED_HOOK +BTF_SET_END(bpf_sched_hooks) + +int bpf_sched_verify_prog(struct bpf_verifier_log *vlog, + const struct bpf_prog *prog) +{ + if (!prog->gpl_compatible) { + bpf_log(vlog, + "sched programs must have a GPL compatible license\n"); + return -EINVAL; + } + + if (!btf_id_set_contains(&bpf_sched_hooks, prog->aux->attach_btf_id)) { + bpf_log(vlog, "attach_btf_id %u points to wrong type name %s\n", + prog->aux->attach_btf_id, prog->aux->attach_func_name); + return -EINVAL; + } + + return 0; +} + +static const struct bpf_func_proto * +bpf_sched_func_proto(enum bpf_func_id func_id, const struct bpf_prog *prog) +{ + switch (func_id) { + case BPF_FUNC_trace_printk: + return bpf_get_trace_printk_proto(); + default: + return bpf_base_func_proto(func_id); + } +} + +const struct bpf_prog_ops bpf_sched_prog_ops = { +}; + +const struct bpf_verifier_ops bpf_sched_verifier_ops = { + .get_func_proto = bpf_sched_func_proto, + .is_valid_access = btf_ctx_access, +}; diff --git a/kernel/sched/build_utility.c b/kernel/sched/build_utility.c index 99bdd96f454f..d44c584d9bc7 100644 --- a/kernel/sched/build_utility.c +++ b/kernel/sched/build_utility.c @@ -108,3 +108,7 @@ #ifdef CONFIG_SCHED_AUTOGROUP # include "autogroup.c" #endif + +#ifdef CONFIG_BPF_SCHED +# include "bpf_sched.c" +#endif diff --git a/tools/include/uapi/linux/bpf.h b/tools/include/uapi/linux/bpf.h index 4924f0cde1bc..9dd0b85549b6 100644 --- a/tools/include/uapi/linux/bpf.h +++ b/tools/include/uapi/linux/bpf.h @@ -988,6 +988,7 @@ enum bpf_prog_type { BPF_PROG_TYPE_SK_LOOKUP, BPF_PROG_TYPE_SYSCALL, /* a program that can execute syscalls */ BPF_PROG_TYPE_NETFILTER, + BPF_PROG_TYPE_SCHED, };

enum bpf_attach_type { @@ -1040,6 +1041,7 @@ enum bpf_attach_type { BPF_TCX_INGRESS, BPF_TCX_EGRESS, BPF_TRACE_UPROBE_MULTI, + BPF_SCHED, __MAX_BPF_ATTACH_TYPE };

diff --git a/tools/lib/bpf/bpf.c b/tools/lib/bpf/bpf.c index b0f1913763a3..ddbb16be651f 100644 --- a/tools/lib/bpf/bpf.c +++ b/tools/lib/bpf/bpf.c @@ -781,6 +781,7 @@ int bpf_link_create(int prog_fd, int target_fd, case BPF_TRACE_FENTRY: case BPF_TRACE_FEXIT: case BPF_MODIFY_RETURN: + case BPF_SCHED: case BPF_LSM_MAC: attr.link_create.tracing.cookie = OPTS_GET(opts, tracing.cookie, 0); if (!OPTS_ZEROED(opts, tracing))

From: Guan Jing guanjing6@huawei.com

maillist inclusion category: feature bugzilla: https://gitee.com/openeuler/kernel/issues/I8OIT1

Reference: https://lore.kernel.org/all/20210916162451.709260-1-guro@fb.com/

-------------------

This patch adds a support for loading and attaching scheduler bpf programs.

Signed-off-by: Roman Gushchin guro@fb.com Signed-off-by: Chen Hui judy.chenhui@huawei.com Signed-off-by: Ren Zhijie renzhijie2@huawei.com Signed-off-by: Hui Tang tanghui20@huawei.com Signed-off-by: Guan Jing guanjing6@huawei.com --- tools/lib/bpf/libbpf.c | 21 ++++++++++++++++++++- tools/lib/bpf/libbpf.h | 2 ++ tools/lib/bpf/libbpf.map | 1 + 3 files changed, 23 insertions(+), 1 deletion(-)

diff --git a/tools/lib/bpf/libbpf.c b/tools/lib/bpf/libbpf.c index 96ff1aa4bf6a..d683d1bcc0f4 100644 --- a/tools/lib/bpf/libbpf.c +++ b/tools/lib/bpf/libbpf.c @@ -3029,7 +3029,8 @@ static int bpf_object_fixup_btf(struct bpf_object *obj) static bool prog_needs_vmlinux_btf(struct bpf_program *prog) { if (prog->type == BPF_PROG_TYPE_STRUCT_OPS || - prog->type == BPF_PROG_TYPE_LSM) + prog->type == BPF_PROG_TYPE_LSM || + prog->type == BPF_PROG_TYPE_SCHED) return true;

/* BPF_PROG_TYPE_TRACING programs which do not attach to other programs @@ -8764,6 +8765,7 @@ static int attach_kprobe_multi(const struct bpf_program *prog, long cookie, stru static int attach_uprobe_multi(const struct bpf_program *prog, long cookie, struct bpf_link **link); static int attach_lsm(const struct bpf_program *prog, long cookie, struct bpf_link **link); static int attach_iter(const struct bpf_program *prog, long cookie, struct bpf_link **link); +static int attach_sched(const struct bpf_program *prog, long cookie, struct bpf_link **link);

static const struct bpf_sec_def section_defs[] = { SEC_DEF("socket", SOCKET_FILTER, 0, SEC_NONE), @@ -8858,6 +8860,7 @@ static const struct bpf_sec_def section_defs[] = { SEC_DEF("struct_ops.s+", STRUCT_OPS, 0, SEC_SLEEPABLE), SEC_DEF("sk_lookup", SK_LOOKUP, BPF_SK_LOOKUP, SEC_ATTACHABLE), SEC_DEF("netfilter", NETFILTER, BPF_NETFILTER, SEC_NONE), + SEC_DEF("sched/", SCHED, BPF_SCHED, SEC_ATTACH_BTF, attach_sched), };

int libbpf_register_prog_handler(const char *sec, @@ -9237,6 +9240,7 @@ static int bpf_object__collect_st_ops_relos(struct bpf_object *obj, #define BTF_TRACE_PREFIX "btf_trace_" #define BTF_LSM_PREFIX "bpf_lsm_" #define BTF_ITER_PREFIX "bpf_iter_" +#define BTF_SCHED_PREFIX "bpf_sched_" #define BTF_MAX_NAME_SIZE 128

void btf_get_kernel_prefix_kind(enum bpf_attach_type attach_type, @@ -9256,6 +9260,10 @@ void btf_get_kernel_prefix_kind(enum bpf_attach_type attach_type, *prefix = BTF_ITER_PREFIX; *kind = BTF_KIND_FUNC; break; + case BPF_SCHED: + *prefix = BTF_SCHED_PREFIX; + *kind = BTF_KIND_FUNC; + break; default: *prefix = ""; *kind = BTF_KIND_FUNC; @@ -12113,6 +12121,17 @@ struct bpf_link *bpf_program__attach_netfilter(const struct bpf_program *prog, return link; }

+struct bpf_link *bpf_program__attach_sched(const struct bpf_program *prog) +{ + return bpf_program__attach_btf_id(prog, NULL); +} + +static int attach_sched(const struct bpf_program *prog, long cookie, struct bpf_link **link) +{ + *link = bpf_program__attach_sched(prog); + return libbpf_get_error(*link); +} + struct bpf_link *bpf_program__attach(const struct bpf_program *prog) { struct bpf_link *link = NULL; diff --git a/tools/lib/bpf/libbpf.h b/tools/lib/bpf/libbpf.h index 0e52621cba43..aabdd973c1a5 100644 --- a/tools/lib/bpf/libbpf.h +++ b/tools/lib/bpf/libbpf.h @@ -769,6 +769,8 @@ bpf_program__attach_xdp(const struct bpf_program *prog, int ifindex); LIBBPF_API struct bpf_link * bpf_program__attach_freplace(const struct bpf_program *prog, int target_fd, const char *attach_func_name); +LIBBPF_API struct bpf_link * +bpf_program__attach_sched(const struct bpf_program *prog);

struct bpf_netfilter_opts { /* size of this struct, for forward/backward compatibility */ diff --git a/tools/lib/bpf/libbpf.map b/tools/lib/bpf/libbpf.map index 57712321490f..228ab00a5e69 100644 --- a/tools/lib/bpf/libbpf.map +++ b/tools/lib/bpf/libbpf.map @@ -236,6 +236,7 @@ LIBBPF_0.2.0 { perf_buffer__buffer_fd; perf_buffer__epoll_fd; perf_buffer__consume_buffer; + bpf_program__attach_sched; } LIBBPF_0.1.0;

LIBBPF_0.3.0 {

From: Guan Jing guanjing6@huawei.com

hulk inclusion category: feature bugzilla: https://gitee.com/openeuler/kernel/issues/I8OIT1

--------------------------------

Add hook of sched type in select_task_rq_fair(), as follows: 'cfs_select_rq' Replace the original core selection policy or implement dynamic CPU affinity.

Signed-off-by: Chen Hui judy.chenhui@huawei.com Signed-off-by: Hui Tang tanghui20@huawei.com Signed-off-by: Guan Jing guanjing6@huawei.com --- include/linux/sched.h | 12 ++++++++++++ include/linux/sched_hook_defs.h | 2 +- kernel/sched/core.c | 15 +++++++++++++++ kernel/sched/fair.c | 28 ++++++++++++++++++++++++++++ kernel/sched/sched.h | 4 ++++ scripts/bpf_doc.py | 2 ++ 6 files changed, 62 insertions(+), 1 deletion(-)

diff --git a/include/linux/sched.h b/include/linux/sched.h index 857c0d8df803..569d055518cc 100644 --- a/include/linux/sched.h +++ b/include/linux/sched.h @@ -2480,6 +2480,18 @@ static inline void rseq_syscall(struct pt_regs *regs)

#ifdef CONFIG_BPF_SCHED extern void sched_settag(struct task_struct *tsk, s64 tag); + +struct sched_migrate_ctx { + struct task_struct *task; + struct cpumask *select_idle_mask; + int prev_cpu; + int curr_cpu; + int is_sync; + int want_affine; + int wake_flags; + int sd_flag; + int new_cpu; +}; #endif

#ifdef CONFIG_SCHED_CORE diff --git a/include/linux/sched_hook_defs.h b/include/linux/sched_hook_defs.h index 14344004e335..0e91209826a1 100644 --- a/include/linux/sched_hook_defs.h +++ b/include/linux/sched_hook_defs.h @@ -1,2 +1,2 @@ /* SPDX-License-Identifier: GPL-2.0 */ -BPF_SCHED_HOOK(int, 0, dummy, void) +BPF_SCHED_HOOK(int, -1, cfs_select_rq, struct sched_migrate_ctx *ctx) diff --git a/kernel/sched/core.c b/kernel/sched/core.c index dc48b7dd9036..f49884275d02 100644 --- a/kernel/sched/core.c +++ b/kernel/sched/core.c @@ -2469,7 +2469,11 @@ static inline bool rq_has_pinned_tasks(struct rq *rq) * Per-CPU kthreads are allowed to run on !active && online CPUs, see * __set_cpus_allowed_ptr() and select_fallback_rq(). */ +#ifdef CONFIG_BPF_SCHED +inline bool is_cpu_allowed(struct task_struct *p, int cpu) +#else static inline bool is_cpu_allowed(struct task_struct *p, int cpu) +#endif { /* When not in the task's cpumask, no point in looking further. */ if (!cpumask_test_cpu(cpu, p->cpus_ptr)) @@ -9956,6 +9960,10 @@ LIST_HEAD(task_groups); static struct kmem_cache *task_group_cache __read_mostly; #endif

+#ifdef CONFIG_BPF_SCHED +DECLARE_PER_CPU(cpumask_var_t, select_idle_mask); +#endif + void __init sched_init(void) { unsigned long ptr = 0; @@ -10011,6 +10019,13 @@ void __init sched_init(void) global_rt_period(), global_rt_runtime()); #endif /* CONFIG_RT_GROUP_SCHED */

+#if defined(CONFIG_CPUMASK_OFFSTACK) && defined(CONFIG_BPF_SCHED) + for_each_possible_cpu(i) { + per_cpu(select_idle_mask, i) = (cpumask_var_t)kzalloc_node( + cpumask_size(), GFP_KERNEL, cpu_to_node(i)); + } +#endif + #ifdef CONFIG_CGROUP_SCHED task_group_cache = KMEM_CACHE(task_group, 0);

diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c index 9e32aae696ba..fc17a3af11b1 100644 --- a/kernel/sched/fair.c +++ b/kernel/sched/fair.c @@ -52,6 +52,7 @@ #include <asm/switch_to.h>

#include <linux/sched/cond_resched.h> +#include <linux/bpf_sched.h>

#include "sched.h" #include "stats.h" @@ -99,6 +100,10 @@ unsigned int sysctl_sched_child_runs_first __read_mostly;

const_debug unsigned int sysctl_sched_migration_cost = 500000UL;

+#ifdef CONFIG_BPF_SCHED +DEFINE_PER_CPU(cpumask_var_t, select_idle_mask); +#endif + int sched_thermal_decay_shift; static int __init setup_sched_thermal_decay_shift(char *str) { @@ -8468,6 +8473,10 @@ select_task_rq_fair(struct task_struct *p, int prev_cpu, int wake_flags) #ifdef CONFIG_QOS_SCHED_DYNAMIC_AFFINITY int idlest_cpu = -1; #endif +#ifdef CONFIG_BPF_SCHED + struct sched_migrate_ctx ctx; + int ret; +#endif

time = schedstat_start_time();

@@ -8502,6 +8511,25 @@ select_task_rq_fair(struct task_struct *p, int prev_cpu, int wake_flags) }

rcu_read_lock(); +#ifdef CONFIG_BPF_SCHED + if (bpf_sched_enabled()) { + ctx.task = p; + ctx.prev_cpu = prev_cpu; + ctx.curr_cpu = cpu; + ctx.is_sync = sync; + ctx.wake_flags = wake_flags; + ctx.want_affine = want_affine; + ctx.sd_flag = sd_flag; + ctx.select_idle_mask = this_cpu_cpumask_var_ptr(select_idle_mask); + + ret = bpf_sched_cfs_select_rq(&ctx); + if (ret >= 0 && is_cpu_allowed(p, ret)) { + rcu_read_unlock(); + return ret; + } + } +#endif + for_each_domain(cpu, tmp) { /* * If both 'cpu' and 'prev_cpu' are part of this domain, diff --git a/kernel/sched/sched.h b/kernel/sched/sched.h index de81c7f2ad93..b7ba69337453 100644 --- a/kernel/sched/sched.h +++ b/kernel/sched/sched.h @@ -3621,4 +3621,8 @@ static inline void init_sched_mm_cid(struct task_struct *t) { } extern u64 avg_vruntime(struct cfs_rq *cfs_rq); extern int entity_eligible(struct cfs_rq *cfs_rq, struct sched_entity *se);

+#ifdef CONFIG_BPF_SCHED +inline bool is_cpu_allowed(struct task_struct *p, int cpu); +#endif + #endif /* _KERNEL_SCHED_SCHED_H */ diff --git a/scripts/bpf_doc.py b/scripts/bpf_doc.py index 61b7dddedc46..a8a547fdb321 100755 --- a/scripts/bpf_doc.py +++ b/scripts/bpf_doc.py @@ -700,6 +700,7 @@ class PrinterHelpers(Printer): 'struct bpf_dynptr', 'struct iphdr', 'struct ipv6hdr', + 'struct sched_migrate_ctx', ] known_types = { '...', @@ -755,6 +756,7 @@ class PrinterHelpers(Printer): 'const struct bpf_dynptr', 'struct iphdr', 'struct ipv6hdr', + 'struct sched_migrate_ctx', } mapped_types = { 'u8': '__u8',

From: Guan Jing guanjing6@huawei.com

hulk inclusion category: feature bugzilla: https://gitee.com/openeuler/kernel/issues/I8OIT1

--------------------------------

Enable CONFIG_BPF_SCHED for x86, arm64 by default.

Signed-off-by: Guan Jing guanjing6@huawei.com --- arch/arm64/configs/openeuler_defconfig | 1 + arch/x86/configs/openeuler_defconfig | 1 + 2 files changed, 2 insertions(+)

diff --git a/arch/arm64/configs/openeuler_defconfig b/arch/arm64/configs/openeuler_defconfig index 703fe5bc2535..1fe486a62814 100644 --- a/arch/arm64/configs/openeuler_defconfig +++ b/arch/arm64/configs/openeuler_defconfig @@ -88,6 +88,7 @@ CONFIG_BPF_JIT_DEFAULT_ON=y # CONFIG_BPF_UNPRIV_DEFAULT_OFF is not set # CONFIG_BPF_PRELOAD is not set # CONFIG_BPF_LSM is not set +CONFIG_BPF_SCHED=y # end of BPF subsystem

CONFIG_PREEMPT_NONE_BUILD=y diff --git a/arch/x86/configs/openeuler_defconfig b/arch/x86/configs/openeuler_defconfig index c9e816e54003..a95f27894e05 100644 --- a/arch/x86/configs/openeuler_defconfig +++ b/arch/x86/configs/openeuler_defconfig @@ -106,6 +106,7 @@ CONFIG_BPF_JIT_DEFAULT_ON=y # CONFIG_BPF_UNPRIV_DEFAULT_OFF is not set # CONFIG_BPF_PRELOAD is not set # CONFIG_BPF_LSM is not set +CONFIG_BPF_SCHED=y # end of BPF subsystem

CONFIG_PREEMPT_BUILD=y