From: Roman Gushchin guro@fb.com
maillist inclusion category: feature bugzilla: https://gitee.com/openeuler/kernel/issues/I5F6X6 CVE: NA
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 --- 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 ++ init/Kconfig | 10 ++++++ kernel/bpf/btf.c | 1 + kernel/bpf/syscall.c | 9 +++++ kernel/bpf/trampoline.c | 1 + kernel/bpf/verifier.c | 11 +++++- kernel/sched/Makefile | 1 + kernel/sched/bpf_sched.c | 62 +++++++++++++++++++++++++++++++++ tools/include/uapi/linux/bpf.h | 2 ++ 12 files changed, 130 insertions(+), 1 deletion(-) 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 a8137bb6dd3c..5732b485c539 100644 --- a/include/linux/bpf_types.h +++ b/include/linux/bpf_types.h @@ -77,6 +77,10 @@ BPF_PROG_TYPE(BPF_PROG_TYPE_LSM, lsm, void *, void *) #endif /* CONFIG_BPF_LSM */ #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 8fae845d80e2..4f8f3f2113a7 100644 --- a/include/uapi/linux/bpf.h +++ b/include/uapi/linux/bpf.h @@ -199,6 +199,7 @@ enum bpf_prog_type { BPF_PROG_TYPE_EXT, BPF_PROG_TYPE_LSM, BPF_PROG_TYPE_SK_LOOKUP, + BPF_PROG_TYPE_SCHED, };
enum bpf_attach_type { @@ -240,6 +241,7 @@ enum bpf_attach_type { BPF_XDP_CPUMAP, BPF_SK_LOOKUP, BPF_XDP, + BPF_SCHED, __MAX_BPF_ATTACH_TYPE };
diff --git a/init/Kconfig b/init/Kconfig index 0ac3206686db..0935a7d84a26 100644 --- a/init/Kconfig +++ b/init/Kconfig @@ -1758,6 +1758,16 @@ config BPF_LSM
If you are unsure how to answer this question, answer N.
+config BPF_SCHED + bool "SCHED Instrumentation with BPF" + depends on BPF_EVENTS + depends on BPF_SYSCALL + help + Enables instrumentation of the sched hooks with eBPF programs for + implementing dynamic scheduling policies. + + If you are unsure how to answer this question, answer N. + config BPF_SYSCALL bool "Enable bpf() system call" select BPF diff --git a/kernel/bpf/btf.c b/kernel/bpf/btf.c index fba28f17e61a..9a0a9895ec62 100644 --- a/kernel/bpf/btf.c +++ b/kernel/bpf/btf.c @@ -4479,6 +4479,7 @@ bool btf_ctx_access(int off, int size, enum bpf_access_type type, return true; t = btf_type_by_id(btf, t->type); break; + case BPF_SCHED: 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 419dbc3d060e..ff65862ae5ce 100644 --- a/kernel/bpf/syscall.c +++ b/kernel/bpf/syscall.c @@ -1997,6 +1997,7 @@ 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: + case BPF_PROG_TYPE_SCHED: break; default: return -EINVAL; @@ -2108,6 +2109,7 @@ 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 */ + case BPF_PROG_TYPE_SCHED: return true; default: return false; @@ -2608,6 +2610,12 @@ static int bpf_tracing_prog_attach(struct bpf_prog *prog, goto out_put_prog; } break; + case BPF_PROG_TYPE_SCHED: + if (prog->expected_attach_type != BPF_SCHED) { + err = -EINVAL; + goto out_put_prog; + } + break; default: err = -EINVAL; goto out_put_prog; @@ -2838,6 +2846,7 @@ static int bpf_raw_tracepoint_open(const union bpf_attr *attr) case BPF_PROG_TYPE_TRACING: case BPF_PROG_TYPE_EXT: case BPF_PROG_TYPE_LSM: + case BPF_PROG_TYPE_SCHED: if (attr->raw_tracepoint.name) { /* The attach point for this category of programs * should be specified via btf_id during program load. diff --git a/kernel/bpf/trampoline.c b/kernel/bpf/trampoline.c index 986dabc3d11f..cc6ba35a1d14 100644 --- a/kernel/bpf/trampoline.c +++ b/kernel/bpf/trampoline.c @@ -357,6 +357,7 @@ 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; + case BPF_SCHED: 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 6423f1714a2f..d26104b258ba 100644 --- a/kernel/bpf/verifier.c +++ b/kernel/bpf/verifier.c @@ -22,6 +22,7 @@ #include <linux/error-injection.h> #include <linux/bpf_lsm.h> #include <linux/btf_ids.h> +#include <linux/bpf_sched.h>
#include "disasm.h"
@@ -12178,6 +12179,7 @@ int bpf_check_attach_target(struct bpf_verifier_log *log, case BPF_LSM_MAC: case BPF_TRACE_FENTRY: case BPF_TRACE_FEXIT: + case BPF_SCHED: if (!btf_type_is_func(t)) { bpf_log(log, "attach_btf_id %u is not a function\n", btf_id); @@ -12283,7 +12285,8 @@ static int check_attach_btf_id(struct bpf_verifier_env *env)
if (prog->type != BPF_PROG_TYPE_TRACING && prog->type != BPF_PROG_TYPE_LSM && - prog->type != BPF_PROG_TYPE_EXT) + prog->type != BPF_PROG_TYPE_EXT && + prog->type != BPF_PROG_TYPE_SCHED) return 0;
ret = bpf_check_attach_target(&env->log, prog, tgt_prog, btf_id, &tgt_info); @@ -12323,6 +12326,12 @@ static int check_attach_btf_id(struct bpf_verifier_env *env) return ret; }
+ if (prog->type == BPF_PROG_TYPE_SCHED) { + ret = bpf_sched_verify_prog(&env->log, prog); + if (ret < 0) + return ret; + } + key = bpf_trampoline_compute_key(tgt_prog, btf_id); tr = bpf_trampoline_get(key, &tgt_info); if (!tr) diff --git a/kernel/sched/Makefile b/kernel/sched/Makefile index 5fc9c9b70862..8ae9e39eb83a 100644 --- a/kernel/sched/Makefile +++ b/kernel/sched/Makefile @@ -36,3 +36,4 @@ obj-$(CONFIG_CPU_FREQ_GOV_SCHEDUTIL) += cpufreq_schedutil.o obj-$(CONFIG_MEMBARRIER) += membarrier.o obj-$(CONFIG_CPU_ISOLATION) += isolation.o obj-$(CONFIG_PSI) += psi.o +obj-$(CONFIG_BPF_SCHED) += bpf_sched.o \ No newline at end of file diff --git a/kernel/sched/bpf_sched.c b/kernel/sched/bpf_sched.c new file mode 100644 index 000000000000..2f05c186cfd0 --- /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 NULL; + } +} + +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/tools/include/uapi/linux/bpf.h b/tools/include/uapi/linux/bpf.h index a44cb5155825..7d7c04f698ec 100644 --- a/tools/include/uapi/linux/bpf.h +++ b/tools/include/uapi/linux/bpf.h @@ -199,6 +199,7 @@ enum bpf_prog_type { BPF_PROG_TYPE_EXT, BPF_PROG_TYPE_LSM, BPF_PROG_TYPE_SK_LOOKUP, + BPF_PROG_TYPE_SCHED, };
enum bpf_attach_type { @@ -240,6 +241,7 @@ enum bpf_attach_type { BPF_XDP_CPUMAP, BPF_SK_LOOKUP, BPF_XDP, + BPF_SCHED, __MAX_BPF_ATTACH_TYPE };