Merge branch 'bpf-trampoline'

Alexei Starovoitov says:

====================
Introduce BPF trampoline that works as a bridge between kernel functions, BPF
programs and other BPF programs.

The first use case is fentry/fexit BPF programs that are roughly equivalent to
kprobe/kretprobe. Unlike k[ret]probe there is practically zero overhead to call
a set of BPF programs before or after kernel function.

The second use case is heavily influenced by pain points in XDP development.
BPF trampoline allows attaching similar fentry/fexit BPF program to any
networking BPF program. It's now possible to see packets on input and output of
any XDP, TC, lwt, cgroup programs without disturbing them. This greatly helps
BPF-based network troubleshooting.

The third use case of BPF trampoline will be explored in the follow up patches.
The BPF trampoline will be used to dynamicly link BPF programs. It's more
generic mechanism than array and link list of programs used in tracing,
networking, cgroups. In many cases it can be used as a replacement for
bpf_tail_call-based program chaining. See [1] for long term design discussion.

v3 -> v4:
- Included Peter's
  "86/alternatives: Teach text_poke_bp() to emulate instructions" as a first patch.
  If it changes between now and merge window, I'll rebease to newer version.
  The patch is necessary to do s/text_poke/text_poke_bp/ in patch 3 to fix the race.
- In patch 4 fixed bpf_trampoline creation race spotted by Andrii.
- Added patch 15 that annotates prog->kern bpf context types. It made patches 16
  and 17 cleaner and more generic.
- Addressed Andrii's feedback in other patches.

v2 -> v3:
- Addressed Song's and Andrii's comments
- Fixed few minor bugs discovered while testing
- Added one more libbpf patch

v1 -> v2:
- Addressed Andrii's comments
- Added more test for fentry/fexit to kernel functions. Including stress test
  for maximum number of progs per trampoline.
- Fixed a race btf_resolve_helper_id()
- Added a patch to compare BTF types of functions arguments with actual types.
- Added support for attaching BPF program to another BPF program via trampoline
- Converted to use text_poke() API. That's the only viable mechanism to
  implement BPF-to-BPF attach. BPF-to-kernel attach can be refactored to use
  register_ftrace_direct() whenever it's available.

  [1] https://lore.kernel.org/bpf/20191112025112.bhzmrrh2pr76ssnh@ast-mbp.dhcp.thefacebook.com/
====================

Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>

6 files changed, 798 insertions, 36 deletions

diff --git a/kernel/bpf/Makefile b/kernel/bpf/Makefile
index e1d9adb212f9..3f671bf617e8 100644
--- a/kernel/bpf/Makefile
+++ b/kernel/bpf/Makefile
@@ -6,6 +6,7 @@ obj-$(CONFIG_BPF_SYSCALL) += syscall.o verifier.o inode.o helpers.o tnum.o
 obj-$(CONFIG_BPF_SYSCALL) += hashtab.o arraymap.o percpu_freelist.o bpf_lru_list.o lpm_trie.o map_in_map.o
 obj-$(CONFIG_BPF_SYSCALL) += local_storage.o queue_stack_maps.o
 obj-$(CONFIG_BPF_SYSCALL) += disasm.o
+obj-$(CONFIG_BPF_JIT) += trampoline.o
 obj-$(CONFIG_BPF_SYSCALL) += btf.o
 ifeq ($(CONFIG_NET),y)
 obj-$(CONFIG_BPF_SYSCALL) += devmap.o
diff --git a/kernel/bpf/btf.c b/kernel/bpf/btf.c
index 4639c4ba9a9b..40efde5eedcb 100644
--- a/kernel/bpf/btf.c
+++ b/kernel/bpf/btf.c
@@ -2,6 +2,8 @@
 /* Copyright (c) 2018 Facebook */
 
 #include <uapi/linux/btf.h>
+#include <uapi/linux/bpf.h>
+#include <uapi/linux/bpf_perf_event.h>
 #include <uapi/linux/types.h>
 #include <linux/seq_file.h>
 #include <linux/compiler.h>
@@ -16,6 +18,9 @@
 #include <linux/sort.h>
 #include <linux/bpf_verifier.h>
 #include <linux/btf.h>
+#include <linux/skmsg.h>
+#include <linux/perf_event.h>
+#include <net/sock.h>
 
 /* BTF (BPF Type Format) is the meta data format which describes
  * the data types of BPF program/map.  Hence, it basically focus
@@ -3439,13 +3444,112 @@ errout:
 
 extern char __weak _binary__btf_vmlinux_bin_start[];
 extern char __weak _binary__btf_vmlinux_bin_end[];
+extern struct btf *btf_vmlinux;
+
+#define BPF_MAP_TYPE(_id, _ops)
+static union {
+	struct bpf_ctx_convert {
+#define BPF_PROG_TYPE(_id, _name, prog_ctx_type, kern_ctx_type) \
+	prog_ctx_type _id##_prog; \
+	kern_ctx_type _id##_kern;
+#include <linux/bpf_types.h>
+#undef BPF_PROG_TYPE
+	} *__t;
+	/* 't' is written once under lock. Read many times. */
+	const struct btf_type *t;
+} bpf_ctx_convert;
+enum {
+#define BPF_PROG_TYPE(_id, _name, prog_ctx_type, kern_ctx_type) \
+	__ctx_convert##_id,
+#include <linux/bpf_types.h>
+#undef BPF_PROG_TYPE
+};
+static u8 bpf_ctx_convert_map[] = {
+#define BPF_PROG_TYPE(_id, _name, prog_ctx_type, kern_ctx_type) \
+	[_id] = __ctx_convert##_id,
+#include <linux/bpf_types.h>
+#undef BPF_PROG_TYPE
+};
+#undef BPF_MAP_TYPE
+
+static const struct btf_member *
+btf_get_prog_ctx_type(struct bpf_verifier_log *log, struct btf *btf,
+		      const struct btf_type *t, enum bpf_prog_type prog_type)
+{
+	const struct btf_type *conv_struct;
+	const struct btf_type *ctx_struct;
+	const struct btf_member *ctx_type;
+	const char *tname, *ctx_tname;
+
+	conv_struct = bpf_ctx_convert.t;
+	if (!conv_struct) {
+		bpf_log(log, "btf_vmlinux is malformed\n");
+		return NULL;
+	}
+	t = btf_type_by_id(btf, t->type);
+	while (btf_type_is_modifier(t))
+		t = btf_type_by_id(btf, t->type);
+	if (!btf_type_is_struct(t)) {
+		/* Only pointer to struct is supported for now.
+		 * That means that BPF_PROG_TYPE_TRACEPOINT with BTF
+		 * is not supported yet.
+		 * BPF_PROG_TYPE_RAW_TRACEPOINT is fine.
+		 */
+		bpf_log(log, "BPF program ctx type is not a struct\n");
+		return NULL;
+	}
+	tname = btf_name_by_offset(btf, t->name_off);
+	if (!tname) {
+		bpf_log(log, "BPF program ctx struct doesn't have a name\n");
+		return NULL;
+	}
+	/* prog_type is valid bpf program type. No need for bounds check. */
+	ctx_type = btf_type_member(conv_struct) + bpf_ctx_convert_map[prog_type] * 2;
+	/* ctx_struct is a pointer to prog_ctx_type in vmlinux.
+	 * Like 'struct __sk_buff'
+	 */
+	ctx_struct = btf_type_by_id(btf_vmlinux, ctx_type->type);
+	if (!ctx_struct)
+		/* should not happen */
+		return NULL;
+	ctx_tname = btf_name_by_offset(btf_vmlinux, ctx_struct->name_off);
+	if (!ctx_tname) {
+		/* should not happen */
+		bpf_log(log, "Please fix kernel include/linux/bpf_types.h\n");
+		return NULL;
+	}
+	/* only compare that prog's ctx type name is the same as
+	 * kernel expects. No need to compare field by field.
+	 * It's ok for bpf prog to do:
+	 * struct __sk_buff {};
+	 * int socket_filter_bpf_prog(struct __sk_buff *skb)
+	 * { // no fields of skb are ever used }
+	 */
+	if (strcmp(ctx_tname, tname))
+		return NULL;
+	return ctx_type;
+}
+
+static int btf_translate_to_vmlinux(struct bpf_verifier_log *log,
+				     struct btf *btf,
+				     const struct btf_type *t,
+				     enum bpf_prog_type prog_type)
+{
+	const struct btf_member *prog_ctx_type, *kern_ctx_type;
+
+	prog_ctx_type = btf_get_prog_ctx_type(log, btf, t, prog_type);
+	if (!prog_ctx_type)
+		return -ENOENT;
+	kern_ctx_type = prog_ctx_type + 1;
+	return kern_ctx_type->type;
+}
 
 struct btf *btf_parse_vmlinux(void)
 {
 	struct btf_verifier_env *env = NULL;
 	struct bpf_verifier_log *log;
 	struct btf *btf = NULL;
-	int err;
+	int err, i;
 
 	env = kzalloc(sizeof(*env), GFP_KERNEL | __GFP_NOWARN);
 	if (!env)
@@ -3479,6 +3583,26 @@ struct btf *btf_parse_vmlinux(void)
 	if (err)
 		goto errout;
 
+	/* find struct bpf_ctx_convert for type checking later */
+	for (i = 1; i <= btf->nr_types; i++) {
+		const struct btf_type *t;
+		const char *tname;
+
+		t = btf_type_by_id(btf, i);
+		if (!__btf_type_is_struct(t))
+			continue;
+		tname = __btf_name_by_offset(btf, t->name_off);
+		if (!strcmp(tname, "bpf_ctx_convert")) {
+			/* btf_parse_vmlinux() runs under bpf_verifier_lock */
+			bpf_ctx_convert.t = t;
+			break;
+		}
+	}
+	if (i > btf->nr_types) {
+		err = -ENOENT;
+		goto errout;
+	}
+
 	btf_verifier_env_free(env);
 	refcount_set(&btf->refcnt, 1);
 	return btf;
@@ -3492,17 +3616,29 @@ errout:
 	return ERR_PTR(err);
 }
 
-extern struct btf *btf_vmlinux;
+struct btf *bpf_prog_get_target_btf(const struct bpf_prog *prog)
+{
+	struct bpf_prog *tgt_prog = prog->aux->linked_prog;
+
+	if (tgt_prog) {
+		return tgt_prog->aux->btf;
+	} else {
+		return btf_vmlinux;
+	}
+}
 
 bool btf_ctx_access(int off, int size, enum bpf_access_type type,
 		    const struct bpf_prog *prog,
 		    struct bpf_insn_access_aux *info)
 {
 	const struct btf_type *t = prog->aux->attach_func_proto;
+	struct bpf_prog *tgt_prog = prog->aux->linked_prog;
+	struct btf *btf = bpf_prog_get_target_btf(prog);
 	const char *tname = prog->aux->attach_func_name;
 	struct bpf_verifier_log *log = info->log;
 	const struct btf_param *args;
 	u32 nr_args, arg;
+	int ret;
 
 	if (off % 8) {
 		bpf_log(log, "func '%s' offset %d is not multiple of 8\n",
@@ -3511,22 +3647,34 @@ bool btf_ctx_access(int off, int size, enum bpf_access_type type,
 	}
 	arg = off / 8;
 	args = (const struct btf_param *)(t + 1);
-	nr_args = btf_type_vlen(t);
+	/* if (t == NULL) Fall back to default BPF prog with 5 u64 arguments */
+	nr_args = t ? btf_type_vlen(t) : 5;
 	if (prog->aux->attach_btf_trace) {
 		/* skip first 'void *__data' argument in btf_trace_##name typedef */
 		args++;
 		nr_args--;
 	}
-	if (arg >= nr_args) {
+
+	if (prog->expected_attach_type == BPF_TRACE_FEXIT &&
+	    arg == nr_args) {
+		if (!t)
+			/* Default prog with 5 args. 6th arg is retval. */
+			return true;
+		/* function return type */
+		t = btf_type_by_id(btf, t->type);
+	} else if (arg >= nr_args) {
 		bpf_log(log, "func '%s' doesn't have %d-th argument\n",
-			tname, arg);
+			tname, arg + 1);
 		return false;
+	} else {
+		if (!t)
+			/* Default prog with 5 args */
+			return true;
+		t = btf_type_by_id(btf, args[arg].type);
 	}
-
-	t = btf_type_by_id(btf_vmlinux, args[arg].type);
 	/* skip modifiers */
 	while (btf_type_is_modifier(t))
-		t = btf_type_by_id(btf_vmlinux, t->type);
+		t = btf_type_by_id(btf, t->type);
 	if (btf_type_is_int(t))
 		/* accessing a scalar */
 		return true;
@@ -3534,7 +3682,7 @@ bool btf_ctx_access(int off, int size, enum bpf_access_type type,
 		bpf_log(log,
 			"func '%s' arg%d '%s' has type %s. Only pointer access is allowed\n",
 			tname, arg,
-			__btf_name_by_offset(btf_vmlinux, t->name_off),
+			__btf_name_by_offset(btf, t->name_off),
 			btf_kind_str[BTF_INFO_KIND(t->info)]);
 		return false;
 	}
@@ -3549,10 +3697,19 @@ bool btf_ctx_access(int off, int size, enum bpf_access_type type,
 	info->reg_type = PTR_TO_BTF_ID;
 	info->btf_id = t->type;
 
-	t = btf_type_by_id(btf_vmlinux, t->type);
+	if (tgt_prog) {
+		ret = btf_translate_to_vmlinux(log, btf, t, tgt_prog->type);
+		if (ret > 0) {
+			info->btf_id = ret;
+			return true;
+		} else {
+			return false;
+		}
+	}
+	t = btf_type_by_id(btf, t->type);
 	/* skip modifiers */
 	while (btf_type_is_modifier(t))
-		t = btf_type_by_id(btf_vmlinux, t->type);
+		t = btf_type_by_id(btf, t->type);
 	if (!btf_type_is_struct(t)) {
 		bpf_log(log,
 			"func '%s' arg%d type %s is not a struct\n",
@@ -3561,7 +3718,7 @@ bool btf_ctx_access(int off, int size, enum bpf_access_type type,
 	}
 	bpf_log(log, "func '%s' arg%d has btf_id %d type %s '%s'\n",
 		tname, arg, info->btf_id, btf_kind_str[BTF_INFO_KIND(t->info)],
-		__btf_name_by_offset(btf_vmlinux, t->name_off));
+		__btf_name_by_offset(btf, t->name_off));
 	return true;
 }
 
@@ -3716,7 +3873,8 @@ again:
 	return -EINVAL;
 }
 
-u32 btf_resolve_helper_id(struct bpf_verifier_log *log, void *fn, int arg)
+static int __btf_resolve_helper_id(struct bpf_verifier_log *log, void *fn,
+				   int arg)
 {
 	char fnname[KSYM_SYMBOL_LEN + 4] = "btf_";
 	const struct btf_param *args;
@@ -3784,6 +3942,185 @@ u32 btf_resolve_helper_id(struct bpf_verifier_log *log, void *fn, int arg)
 	return btf_id;
 }
 
+int btf_resolve_helper_id(struct bpf_verifier_log *log,
+			  const struct bpf_func_proto *fn, int arg)
+{
+	int *btf_id = &fn->btf_id[arg];
+	int ret;
+
+	if (fn->arg_type[arg] != ARG_PTR_TO_BTF_ID)
+		return -EINVAL;
+
+	ret = READ_ONCE(*btf_id);
+	if (ret)
+		return ret;
+	/* ok to race the search. The result is the same */
+	ret = __btf_resolve_helper_id(log, fn->func, arg);
+	if (!ret) {
+		/* Function argument cannot be type 'void' */
+		bpf_log(log, "BTF resolution bug\n");
+		return -EFAULT;
+	}
+	WRITE_ONCE(*btf_id, ret);
+	return ret;
+}
+
+static int __get_type_size(struct btf *btf, u32 btf_id,
+			   const struct btf_type **bad_type)
+{
+	const struct btf_type *t;
+
+	if (!btf_id)
+		/* void */
+		return 0;
+	t = btf_type_by_id(btf, btf_id);
+	while (t && btf_type_is_modifier(t))
+		t = btf_type_by_id(btf, t->type);
+	if (!t)
+		return -EINVAL;
+	if (btf_type_is_ptr(t))
+		/* kernel size of pointer. Not BPF's size of pointer*/
+		return sizeof(void *);
+	if (btf_type_is_int(t) || btf_type_is_enum(t))
+		return t->size;
+	*bad_type = t;
+	return -EINVAL;
+}
+
+int btf_distill_func_proto(struct bpf_verifier_log *log,
+			   struct btf *btf,
+			   const struct btf_type *func,
+			   const char *tname,
+			   struct btf_func_model *m)
+{
+	const struct btf_param *args;
+	const struct btf_type *t;
+	u32 i, nargs;
+	int ret;
+
+	if (!func) {
+		/* BTF function prototype doesn't match the verifier types.
+		 * Fall back to 5 u64 args.
+		 */
+		for (i = 0; i < 5; i++)
+			m->arg_size[i] = 8;
+		m->ret_size = 8;
+		m->nr_args = 5;
+		return 0;
+	}
+	args = (const struct btf_param *)(func + 1);
+	nargs = btf_type_vlen(func);
+	if (nargs >= MAX_BPF_FUNC_ARGS) {
+		bpf_log(log,
+			"The function %s has %d arguments. Too many.\n",
+			tname, nargs);
+		return -EINVAL;
+	}
+	ret = __get_type_size(btf, func->type, &t);
+	if (ret < 0) {
+		bpf_log(log,
+			"The function %s return type %s is unsupported.\n",
+			tname, btf_kind_str[BTF_INFO_KIND(t->info)]);
+		return -EINVAL;
+	}
+	m->ret_size = ret;
+
+	for (i = 0; i < nargs; i++) {
+		ret = __get_type_size(btf, args[i].type, &t);
+		if (ret < 0) {
+			bpf_log(log,
+				"The function %s arg%d type %s is unsupported.\n",
+				tname, i, btf_kind_str[BTF_INFO_KIND(t->info)]);
+			return -EINVAL;
+		}
+		m->arg_size[i] = ret;
+	}
+	m->nr_args = nargs;
+	return 0;
+}
+
+int btf_check_func_arg_match(struct bpf_verifier_env *env, int subprog)
+{
+	struct bpf_verifier_state *st = env->cur_state;
+	struct bpf_func_state *func = st->frame[st->curframe];
+	struct bpf_reg_state *reg = func->regs;
+	struct bpf_verifier_log *log = &env->log;
+	struct bpf_prog *prog = env->prog;
+	struct btf *btf = prog->aux->btf;
+	const struct btf_param *args;
+	const struct btf_type *t;
+	u32 i, nargs, btf_id;
+	const char *tname;
+
+	if (!prog->aux->func_info)
+		return 0;
+
+	btf_id = prog->aux->func_info[subprog].type_id;
+	if (!btf_id)
+		return 0;
+
+	if (prog->aux->func_info_aux[subprog].unreliable)
+		return 0;
+
+	t = btf_type_by_id(btf, btf_id);
+	if (!t || !btf_type_is_func(t)) {
+		bpf_log(log, "BTF of subprog %d doesn't point to KIND_FUNC\n",
+			subprog);
+		return -EINVAL;
+	}
+	tname = btf_name_by_offset(btf, t->name_off);
+
+	t = btf_type_by_id(btf, t->type);
+	if (!t || !btf_type_is_func_proto(t)) {
+		bpf_log(log, "Invalid type of func %s\n", tname);
+		return -EINVAL;
+	}
+	args = (const struct btf_param *)(t + 1);
+	nargs = btf_type_vlen(t);
+	if (nargs > 5) {
+		bpf_log(log, "Function %s has %d > 5 args\n", tname, nargs);
+		goto out;
+	}
+	/* check that BTF function arguments match actual types that the
+	 * verifier sees.
+	 */
+	for (i = 0; i < nargs; i++) {
+		t = btf_type_by_id(btf, args[i].type);
+		while (btf_type_is_modifier(t))
+			t = btf_type_by_id(btf, t->type);
+		if (btf_type_is_int(t) || btf_type_is_enum(t)) {
+			if (reg[i + 1].type == SCALAR_VALUE)
+				continue;
+			bpf_log(log, "R%d is not a scalar\n", i + 1);
+			goto out;
+		}
+		if (btf_type_is_ptr(t)) {
+			if (reg[i + 1].type == SCALAR_VALUE) {
+				bpf_log(log, "R%d is not a pointer\n", i + 1);
+				goto out;
+			}
+			/* If program is passing PTR_TO_CTX into subprogram
+			 * check that BTF type matches.
+			 */
+			if (reg[i + 1].type == PTR_TO_CTX &&
+			    !btf_get_prog_ctx_type(log, btf, t, prog->type))
+				goto out;
+			/* All other pointers are ok */
+			continue;
+		}
+		bpf_log(log, "Unrecognized argument type %s\n",
+			btf_kind_str[BTF_INFO_KIND(t->info)]);
+		goto out;
+	}
+	return 0;
+out:
+	/* LLVM optimizations can remove arguments from static functions. */
+	bpf_log(log,
+		"Type info disagrees with actual arguments due to compiler optimizations\n");
+	prog->aux->func_info_aux[subprog].unreliable = true;
+	return 0;
+}
+
 void btf_type_seq_show(const struct btf *btf, u32 type_id, void *obj,
 		       struct seq_file *m)
 {
diff --git a/kernel/bpf/core.c b/kernel/bpf/core.c
index 99693f3c4e99..b5945c3aaa8e 100644
--- a/kernel/bpf/core.c
+++ b/kernel/bpf/core.c
@@ -2015,6 +2015,7 @@ static void bpf_prog_free_deferred(struct work_struct *work)
 	if (aux->prog->has_callchain_buf)
 		put_callchain_buffers();
 #endif
+	bpf_trampoline_put(aux->trampoline);
 	for (i = 0; i < aux->func_cnt; i++)
 		bpf_jit_free(aux->func[i]);
 	if (aux->func_cnt) {
@@ -2030,6 +2031,8 @@ void bpf_prog_free(struct bpf_prog *fp)
 {
 	struct bpf_prog_aux *aux = fp->aux;
 
+	if (aux->linked_prog)
+		bpf_prog_put(aux->linked_prog);
 	INIT_WORK(&aux->work, bpf_prog_free_deferred);
 	schedule_work(&aux->work);
 }
@@ -2144,6 +2147,12 @@ int __weak skb_copy_bits(const struct sk_buff *skb, int offset, void *to,
 	return -EFAULT;
 }
 
+int __weak bpf_arch_text_poke(void *ip, enum bpf_text_poke_type t,
+			      void *addr1, void *addr2)
+{
+	return -ENOTSUPP;
+}
+
 DEFINE_STATIC_KEY_FALSE(bpf_stats_enabled_key);
 EXPORT_SYMBOL(bpf_stats_enabled_key);
 
diff --git a/kernel/bpf/syscall.c b/kernel/bpf/syscall.c
index 6d9ce95e5a8d..c88c815c2154 100644
--- a/kernel/bpf/syscall.c
+++ b/kernel/bpf/syscall.c
@@ -43,7 +43,7 @@ static DEFINE_SPINLOCK(map_idr_lock);
 int sysctl_unprivileged_bpf_disabled __read_mostly;
 
 static const struct bpf_map_ops * const bpf_map_types[] = {
-#define BPF_PROG_TYPE(_id, _ops)
+#define BPF_PROG_TYPE(_id, _name, prog_ctx_type, kern_ctx_type)
 #define BPF_MAP_TYPE(_id, _ops) \
 	[_id] = &_ops,
 #include <linux/bpf_types.h>
@@ -1189,7 +1189,7 @@ err_put:
 }
 
 static const struct bpf_prog_ops * const bpf_prog_types[] = {
-#define BPF_PROG_TYPE(_id, _name) \
+#define BPF_PROG_TYPE(_id, _name, prog_ctx_type, kern_ctx_type) \
 	[_id] = & _name ## _prog_ops,
 #define BPF_MAP_TYPE(_id, _ops)
 #include <linux/bpf_types.h>
@@ -1328,6 +1328,7 @@ static void __bpf_prog_put_rcu(struct rcu_head *rcu)
 	struct bpf_prog_aux *aux = container_of(rcu, struct bpf_prog_aux, rcu);
 
 	kvfree(aux->func_info);
+	kfree(aux->func_info_aux);
 	free_used_maps(aux);
 	bpf_prog_uncharge_memlock(aux->prog);
 	security_bpf_prog_free(aux);
@@ -1576,7 +1577,7 @@ static void bpf_prog_load_fixup_attach_type(union bpf_attr *attr)
 static int
 bpf_prog_load_check_attach(enum bpf_prog_type prog_type,
 			   enum bpf_attach_type expected_attach_type,
-			   u32 btf_id)
+			   u32 btf_id, u32 prog_fd)
 {
 	switch (prog_type) {
 	case BPF_PROG_TYPE_TRACING:
@@ -1584,7 +1585,7 @@ bpf_prog_load_check_attach(enum bpf_prog_type prog_type,
 			return -EINVAL;
 		break;
 	default:
-		if (btf_id)
+		if (btf_id || prog_fd)
 			return -EINVAL;
 		break;
 	}
@@ -1635,7 +1636,7 @@ bpf_prog_load_check_attach(enum bpf_prog_type prog_type,
 }
 
 /* last field in 'union bpf_attr' used by this command */
-#define	BPF_PROG_LOAD_LAST_FIELD attach_btf_id
+#define	BPF_PROG_LOAD_LAST_FIELD attach_prog_fd
 
 static int bpf_prog_load(union bpf_attr *attr, union bpf_attr __user *uattr)
 {
@@ -1678,7 +1679,8 @@ static int bpf_prog_load(union bpf_attr *attr, union bpf_attr __user *uattr)
 
 	bpf_prog_load_fixup_attach_type(attr);
 	if (bpf_prog_load_check_attach(type, attr->expected_attach_type,
-				       attr->attach_btf_id))
+				       attr->attach_btf_id,
+				       attr->attach_prog_fd))
 		return -EINVAL;
 
 	/* plain bpf_prog allocation */
@@ -1688,6 +1690,16 @@ static int bpf_prog_load(union bpf_attr *attr, union bpf_attr __user *uattr)
 
 	prog->expected_attach_type = attr->expected_attach_type;
 	prog->aux->attach_btf_id = attr->attach_btf_id;
+	if (attr->attach_prog_fd) {
+		struct bpf_prog *tgt_prog;
+
+		tgt_prog = bpf_prog_get(attr->attach_prog_fd);
+		if (IS_ERR(tgt_prog)) {
+			err = PTR_ERR(tgt_prog);
+			goto free_prog_nouncharge;
+		}
+		prog->aux->linked_prog = tgt_prog;
+	}
 
 	prog->aux->offload_requested = !!attr->prog_ifindex;
 
@@ -1799,6 +1811,49 @@ static int bpf_obj_get(const union bpf_attr *attr)
 				attr->file_flags);
 }
 
+static int bpf_tracing_prog_release(struct inode *inode, struct file *filp)
+{
+	struct bpf_prog *prog = filp->private_data;
+
+	WARN_ON_ONCE(bpf_trampoline_unlink_prog(prog));
+	bpf_prog_put(prog);
+	return 0;
+}
+
+static const struct file_operations bpf_tracing_prog_fops = {
+	.release	= bpf_tracing_prog_release,
+	.read		= bpf_dummy_read,
+	.write		= bpf_dummy_write,
+};
+
+static int bpf_tracing_prog_attach(struct bpf_prog *prog)
+{
+	int tr_fd, err;
+
+	if (prog->expected_attach_type != BPF_TRACE_FENTRY &&
+	    prog->expected_attach_type != BPF_TRACE_FEXIT) {
+		err = -EINVAL;
+		goto out_put_prog;
+	}
+
+	err = bpf_trampoline_link_prog(prog);
+	if (err)
+		goto out_put_prog;
+
+	tr_fd = anon_inode_getfd("bpf-tracing-prog", &bpf_tracing_prog_fops,
+				 prog, O_CLOEXEC);
+	if (tr_fd < 0) {
+		WARN_ON_ONCE(bpf_trampoline_unlink_prog(prog));
+		err = tr_fd;
+		goto out_put_prog;
+	}
+	return tr_fd;
+
+out_put_prog:
+	bpf_prog_put(prog);
+	return err;
+}
+
 struct bpf_raw_tracepoint {
 	struct bpf_raw_event_map *btp;
 	struct bpf_prog *prog;
@@ -1850,14 +1905,16 @@ static int bpf_raw_tracepoint_open(const union bpf_attr *attr)
 
 	if (prog->type == BPF_PROG_TYPE_TRACING) {
 		if (attr->raw_tracepoint.name) {
-			/* raw_tp name should not be specified in raw_tp
-			 * programs that were verified via in-kernel BTF info
+			/* The attach point for this category of programs
+			 * should be specified via btf_id during program load.
 			 */
 			err = -EINVAL;
 			goto out_put_prog;
 		}
-		/* raw_tp name is taken from type name instead */
-		tp_name = prog->aux->attach_func_name;
+		if (prog->expected_attach_type == BPF_TRACE_RAW_TP)
+			tp_name = prog->aux->attach_func_name;
+		else
+			return bpf_tracing_prog_attach(prog);
 	} else {
 		if (strncpy_from_user(buf,
 				      u64_to_user_ptr(attr->raw_tracepoint.name),
diff --git a/kernel/bpf/trampoline.c b/kernel/bpf/trampoline.c
new file mode 100644
index 000000000000..10ae59d65f13
--- /dev/null
+++ b/kernel/bpf/trampoline.c
@@ -0,0 +1,253 @@
+// SPDX-License-Identifier: GPL-2.0-only
+/* Copyright (c) 2019 Facebook */
+#include <linux/hash.h>
+#include <linux/bpf.h>
+#include <linux/filter.h>
+
+/* btf_vmlinux has ~22k attachable functions. 1k htab is enough. */
+#define TRAMPOLINE_HASH_BITS 10
+#define TRAMPOLINE_TABLE_SIZE (1 << TRAMPOLINE_HASH_BITS)
+
+static struct hlist_head trampoline_table[TRAMPOLINE_TABLE_SIZE];
+
+/* serializes access to trampoline_table */
+static DEFINE_MUTEX(trampoline_mutex);
+
+struct bpf_trampoline *bpf_trampoline_lookup(u64 key)
+{
+	struct bpf_trampoline *tr;
+	struct hlist_head *head;
+	void *image;
+	int i;
+
+	mutex_lock(&trampoline_mutex);
+	head = &trampoline_table[hash_64(key, TRAMPOLINE_HASH_BITS)];
+	hlist_for_each_entry(tr, head, hlist) {
+		if (tr->key == key) {
+			refcount_inc(&tr->refcnt);
+			goto out;
+		}
+	}
+	tr = kzalloc(sizeof(*tr), GFP_KERNEL);
+	if (!tr)
+		goto out;
+
+	/* is_root was checked earlier. No need for bpf_jit_charge_modmem() */
+	image = bpf_jit_alloc_exec(PAGE_SIZE);
+	if (!image) {
+		kfree(tr);
+		tr = NULL;
+		goto out;
+	}
+
+	tr->key = key;
+	INIT_HLIST_NODE(&tr->hlist);
+	hlist_add_head(&tr->hlist, head);
+	refcount_set(&tr->refcnt, 1);
+	mutex_init(&tr->mutex);
+	for (i = 0; i < BPF_TRAMP_MAX; i++)
+		INIT_HLIST_HEAD(&tr->progs_hlist[i]);
+
+	set_vm_flush_reset_perms(image);
+	/* Keep image as writeable. The alternative is to keep flipping ro/rw
+	 * everytime new program is attached or detached.
+	 */
+	set_memory_x((long)image, 1);
+	tr->image = image;
+out:
+	mutex_unlock(&trampoline_mutex);
+	return tr;
+}
+
+/* Each call __bpf_prog_enter + call bpf_func + call __bpf_prog_exit is ~50
+ * bytes on x86.  Pick a number to fit into PAGE_SIZE / 2
+ */
+#define BPF_MAX_TRAMP_PROGS 40
+
+static int bpf_trampoline_update(struct bpf_trampoline *tr)
+{
+	void *old_image = tr->image + ((tr->selector + 1) & 1) * PAGE_SIZE/2;
+	void *new_image = tr->image + (tr->selector & 1) * PAGE_SIZE/2;
+	struct bpf_prog *progs_to_run[BPF_MAX_TRAMP_PROGS];
+	int fentry_cnt = tr->progs_cnt[BPF_TRAMP_FENTRY];
+	int fexit_cnt = tr->progs_cnt[BPF_TRAMP_FEXIT];
+	struct bpf_prog **progs, **fentry, **fexit;
+	u32 flags = BPF_TRAMP_F_RESTORE_REGS;
+	struct bpf_prog_aux *aux;
+	int err;
+
+	if (fentry_cnt + fexit_cnt == 0) {
+		err = bpf_arch_text_poke(tr->func.addr, BPF_MOD_CALL_TO_NOP,
+					 old_image, NULL);
+		tr->selector = 0;
+		goto out;
+	}
+
+	/* populate fentry progs */
+	fentry = progs = progs_to_run;
+	hlist_for_each_entry(aux, &tr->progs_hlist[BPF_TRAMP_FENTRY], tramp_hlist)
+		*progs++ = aux->prog;
+
+	/* populate fexit progs */
+	fexit = progs;
+	hlist_for_each_entry(aux, &tr->progs_hlist[BPF_TRAMP_FEXIT], tramp_hlist)
+		*progs++ = aux->prog;
+
+	if (fexit_cnt)
+		flags = BPF_TRAMP_F_CALL_ORIG | BPF_TRAMP_F_SKIP_FRAME;
+
+	err = arch_prepare_bpf_trampoline(new_image, &tr->func.model, flags,
+					  fentry, fentry_cnt,
+					  fexit, fexit_cnt,
+					  tr->func.addr);
+	if (err)
+		goto out;
+
+	if (tr->selector)
+		/* progs already running at this address */
+		err = bpf_arch_text_poke(tr->func.addr, BPF_MOD_CALL_TO_CALL,
+					 old_image, new_image);
+	else
+		/* first time registering */
+		err = bpf_arch_text_poke(tr->func.addr, BPF_MOD_NOP_TO_CALL,
+					 NULL, new_image);
+	if (err)
+		goto out;
+	tr->selector++;
+out:
+	return err;
+}
+
+static enum bpf_tramp_prog_type bpf_attach_type_to_tramp(enum bpf_attach_type t)
+{
+	switch (t) {
+	case BPF_TRACE_FENTRY:
+		return BPF_TRAMP_FENTRY;
+	default:
+		return BPF_TRAMP_FEXIT;
+	}
+}
+
+int bpf_trampoline_link_prog(struct bpf_prog *prog)
+{
+	enum bpf_tramp_prog_type kind;
+	struct bpf_trampoline *tr;
+	int err = 0;
+
+	tr = prog->aux->trampoline;
+	kind = bpf_attach_type_to_tramp(prog->expected_attach_type);
+	mutex_lock(&tr->mutex);
+	if (tr->progs_cnt[BPF_TRAMP_FENTRY] + tr->progs_cnt[BPF_TRAMP_FEXIT]
+	    >= BPF_MAX_TRAMP_PROGS) {
+		err = -E2BIG;
+		goto out;
+	}
+	if (!hlist_unhashed(&prog->aux->tramp_hlist)) {
+		/* prog already linked */
+		err = -EBUSY;
+		goto out;
+	}
+	hlist_add_head(&prog->aux->tramp_hlist, &tr->progs_hlist[kind]);
+	tr->progs_cnt[kind]++;
+	err = bpf_trampoline_update(prog->aux->trampoline);
+	if (err) {
+		hlist_del(&prog->aux->tramp_hlist);
+		tr->progs_cnt[kind]--;
+	}
+out:
+	mutex_unlock(&tr->mutex);
+	return err;
+}
+
+/* bpf_trampoline_unlink_prog() should never fail. */
+int bpf_trampoline_unlink_prog(struct bpf_prog *prog)
+{
+	enum bpf_tramp_prog_type kind;
+	struct bpf_trampoline *tr;
+	int err;
+
+	tr = prog->aux->trampoline;
+	kind = bpf_attach_type_to_tramp(prog->expected_attach_type);
+	mutex_lock(&tr->mutex);
+	hlist_del(&prog->aux->tramp_hlist);
+	tr->progs_cnt[kind]--;
+	err = bpf_trampoline_update(prog->aux->trampoline);
+	mutex_unlock(&tr->mutex);
+	return err;
+}
+
+void bpf_trampoline_put(struct bpf_trampoline *tr)
+{
+	if (!tr)
+		return;
+	mutex_lock(&trampoline_mutex);
+	if (!refcount_dec_and_test(&tr->refcnt))
+		goto out;
+	WARN_ON_ONCE(mutex_is_locked(&tr->mutex));
+	if (WARN_ON_ONCE(!hlist_empty(&tr->progs_hlist[BPF_TRAMP_FENTRY])))
+		goto out;
+	if (WARN_ON_ONCE(!hlist_empty(&tr->progs_hlist[BPF_TRAMP_FEXIT])))
+		goto out;
+	bpf_jit_free_exec(tr->image);
+	hlist_del(&tr->hlist);
+	kfree(tr);
+out:
+	mutex_unlock(&trampoline_mutex);
+}
+
+/* The logic is similar to BPF_PROG_RUN, but with explicit rcu and preempt that
+ * are needed for trampoline. The macro is split into
+ * call _bpf_prog_enter
+ * call prog->bpf_func
+ * call __bpf_prog_exit
+ */
+u64 notrace __bpf_prog_enter(void)
+{
+	u64 start = 0;
+
+	rcu_read_lock();
+	preempt_disable();
+	if (static_branch_unlikely(&bpf_stats_enabled_key))
+		start = sched_clock();
+	return start;
+}
+
+void notrace __bpf_prog_exit(struct bpf_prog *prog, u64 start)
+{
+	struct bpf_prog_stats *stats;
+
+	if (static_branch_unlikely(&bpf_stats_enabled_key) &&
+	    /* static_key could be enabled in __bpf_prog_enter
+	     * and disabled in __bpf_prog_exit.
+	     * And vice versa.
+	     * Hence check that 'start' is not zero.
+	     */
+	    start) {
+		stats = this_cpu_ptr(prog->aux->stats);
+		u64_stats_update_begin(&stats->syncp);
+		stats->cnt++;
+		stats->nsecs += sched_clock() - start;
+		u64_stats_update_end(&stats->syncp);
+	}
+	preempt_enable();
+	rcu_read_unlock();
+}
+
+int __weak
+arch_prepare_bpf_trampoline(void *image, struct btf_func_model *m, u32 flags,
+			    struct bpf_prog **fentry_progs, int fentry_cnt,
+			    struct bpf_prog **fexit_progs, int fexit_cnt,
+			    void *orig_call)
+{
+	return -ENOTSUPP;
+}
+
+static int __init init_trampolines(void)
+{
+	int i;
+
+	for (i = 0; i < TRAMPOLINE_TABLE_SIZE; i++)
+		INIT_HLIST_HEAD(&trampoline_table[i]);
+	return 0;
+}
+late_initcall(init_trampolines);
diff --git a/kernel/bpf/verifier.c b/kernel/bpf/verifier.c
index 2f2374967b36..e9dc95a18d44 100644
--- a/kernel/bpf/verifier.c
+++ b/kernel/bpf/verifier.c
@@ -23,7 +23,7 @@
 #include "disasm.h"
 
 static const struct bpf_verifier_ops * const bpf_verifier_ops[] = {
-#define BPF_PROG_TYPE(_id, _name) \
+#define BPF_PROG_TYPE(_id, _name, prog_ctx_type, kern_ctx_type) \
 	[_id] = & _name ## _verifier_ops,
 #define BPF_MAP_TYPE(_id, _ops)
 #include <linux/bpf_types.h>
@@ -3970,6 +3970,9 @@ static int check_func_call(struct bpf_verifier_env *env, struct bpf_insn *insn,
 	/* only increment it after check_reg_arg() finished */
 	state->curframe++;
 
+	if (btf_check_func_arg_match(env, subprog))
+		return -EINVAL;
+
 	/* and go analyze first insn of the callee */
 	*insn_idx = target_insn;
 
@@ -4147,11 +4150,9 @@ static int check_helper_call(struct bpf_verifier_env *env, int func_id, int insn
 	meta.func_id = func_id;
 	/* check args */
 	for (i = 0; i < 5; i++) {
-		if (fn->arg_type[i] == ARG_PTR_TO_BTF_ID) {
-			if (!fn->btf_id[i])
-				fn->btf_id[i] = btf_resolve_helper_id(&env->log, fn->func, i);
-			meta.btf_id = fn->btf_id[i];
-		}
+		err = btf_resolve_helper_id(&env->log, fn, i);
+		if (err > 0)
+			meta.btf_id = err;
 		err = check_func_arg(env, BPF_REG_1 + i, fn->arg_type[i], &meta);
 		if (err)
 			return err;
@@ -6566,6 +6567,7 @@ static int check_btf_func(struct bpf_verifier_env *env,
 	u32 i, nfuncs, urec_size, min_size;
 	u32 krec_size = sizeof(struct bpf_func_info);
 	struct bpf_func_info *krecord;
+	struct bpf_func_info_aux *info_aux = NULL;
 	const struct btf_type *type;
 	struct bpf_prog *prog;
 	const struct btf *btf;
@@ -6599,6 +6601,9 @@ static int check_btf_func(struct bpf_verifier_env *env,
 	krecord = kvcalloc(nfuncs, krec_size, GFP_KERNEL | __GFP_NOWARN);
 	if (!krecord)
 		return -ENOMEM;
+	info_aux = kcalloc(nfuncs, sizeof(*info_aux), GFP_KERNEL | __GFP_NOWARN);
+	if (!info_aux)
+		goto err_free;
 
 	for (i = 0; i < nfuncs; i++) {
 		ret = bpf_check_uarg_tail_zero(urecord, krec_size, urec_size);
@@ -6650,29 +6655,31 @@ static int check_btf_func(struct bpf_verifier_env *env,
 			ret = -EINVAL;
 			goto err_free;
 		}
-
 		prev_offset = krecord[i].insn_off;
 		urecord += urec_size;
 	}
 
 	prog->aux->func_info = krecord;
 	prog->aux->func_info_cnt = nfuncs;
+	prog->aux->func_info_aux = info_aux;
 	return 0;
 
 err_free:
 	kvfree(krecord);
+	kfree(info_aux);
 	return ret;
 }
 
 static void adjust_btf_func(struct bpf_verifier_env *env)
 {
+	struct bpf_prog_aux *aux = env->prog->aux;
 	int i;
 
-	if (!env->prog->aux->func_info)
+	if (!aux->func_info)
 		return;
 
 	for (i = 0; i < env->subprog_cnt; i++)
-		env->prog->aux->func_info[i].insn_off = env->subprog_info[i].start;
+		aux->func_info[i].insn_off = env->subprog_info[i].start;
 }
 
 #define MIN_BPF_LINEINFO_SIZE	(offsetof(struct bpf_line_info, line_col) + \
@@ -7653,6 +7660,9 @@ static int do_check(struct bpf_verifier_env *env)
 			0 /* frameno */,
 			0 /* subprogno, zero == main subprog */);
 
+	if (btf_check_func_arg_match(env, 0))
+		return -EINVAL;
+
 	for (;;) {
 		struct bpf_insn *insn;
 		u8 class;
@@ -9380,10 +9390,17 @@ static void print_verification_stats(struct bpf_verifier_env *env)
 static int check_attach_btf_id(struct bpf_verifier_env *env)
 {
 	struct bpf_prog *prog = env->prog;
+	struct bpf_prog *tgt_prog = prog->aux->linked_prog;
 	u32 btf_id = prog->aux->attach_btf_id;
 	const char prefix[] = "btf_trace_";
+	int ret = 0, subprog = -1, i;
+	struct bpf_trampoline *tr;
 	const struct btf_type *t;
+	bool conservative = true;
 	const char *tname;
+	struct btf *btf;
+	long addr;
+	u64 key;
 
 	if (prog->type != BPF_PROG_TYPE_TRACING)
 		return 0;
@@ -9392,19 +9409,47 @@ static int check_attach_btf_id(struct bpf_verifier_env *env)
 		verbose(env, "Tracing programs must provide btf_id\n");
 		return -EINVAL;
 	}
-	t = btf_type_by_id(btf_vmlinux, btf_id);
+	btf = bpf_prog_get_target_btf(prog);
+	if (!btf) {
+		verbose(env,
+			"FENTRY/FEXIT program can only be attached to another program annotated with BTF\n");
+		return -EINVAL;
+	}
+	t = btf_type_by_id(btf, btf_id);
 	if (!t) {
 		verbose(env, "attach_btf_id %u is invalid\n", btf_id);
 		return -EINVAL;
 	}
-	tname = btf_name_by_offset(btf_vmlinux, t->name_off);
+	tname = btf_name_by_offset(btf, t->name_off);
 	if (!tname) {
 		verbose(env, "attach_btf_id %u doesn't have a name\n", btf_id);
 		return -EINVAL;
 	}
+	if (tgt_prog) {
+		struct bpf_prog_aux *aux = tgt_prog->aux;
+
+		for (i = 0; i < aux->func_info_cnt; i++)
+			if (aux->func_info[i].type_id == btf_id) {
+				subprog = i;
+				break;
+			}
+		if (subprog == -1) {
+			verbose(env, "Subprog %s doesn't exist\n", tname);
+			return -EINVAL;
+		}
+		conservative = aux->func_info_aux[subprog].unreliable;
+		key = ((u64)aux->id) << 32 | btf_id;
+	} else {
+		key = btf_id;
+	}
 
 	switch (prog->expected_attach_type) {
 	case BPF_TRACE_RAW_TP:
+		if (tgt_prog) {
+			verbose(env,
+				"Only FENTRY/FEXIT progs are attachable to another BPF prog\n");
+			return -EINVAL;
+		}
 		if (!btf_type_is_typedef(t)) {
 			verbose(env, "attach_btf_id %u is not a typedef\n",
 				btf_id);
@@ -9416,11 +9461,11 @@ static int check_attach_btf_id(struct bpf_verifier_env *env)
 			return -EINVAL;
 		}
 		tname += sizeof(prefix) - 1;
-		t = btf_type_by_id(btf_vmlinux, t->type);
+		t = btf_type_by_id(btf, t->type);
 		if (!btf_type_is_ptr(t))
 			/* should never happen in valid vmlinux build */
 			return -EINVAL;
-		t = btf_type_by_id(btf_vmlinux, t->type);
+		t = btf_type_by_id(btf, t->type);
 		if (!btf_type_is_func_proto(t))
 			/* should never happen in valid vmlinux build */
 			return -EINVAL;
@@ -9432,6 +9477,66 @@ static int check_attach_btf_id(struct bpf_verifier_env *env)
 		prog->aux->attach_func_proto = t;
 		prog->aux->attach_btf_trace = true;
 		return 0;
+	case BPF_TRACE_FENTRY:
+	case BPF_TRACE_FEXIT:
+		if (!btf_type_is_func(t)) {
+			verbose(env, "attach_btf_id %u is not a function\n",
+				btf_id);
+			return -EINVAL;
+		}
+		t = btf_type_by_id(btf, t->type);
+		if (!btf_type_is_func_proto(t))
+			return -EINVAL;
+		tr = bpf_trampoline_lookup(key);
+		if (!tr)
+			return -ENOMEM;
+		prog->aux->attach_func_name = tname;
+		/* t is either vmlinux type or another program's type */
+		prog->aux->attach_func_proto = t;
+		mutex_lock(&tr->mutex);
+		if (tr->func.addr) {
+			prog->aux->trampoline = tr;
+			goto out;
+		}
+		if (tgt_prog && conservative) {
+			prog->aux->attach_func_proto = NULL;
+			t = NULL;
+		}
+		ret = btf_distill_func_proto(&env->log, btf, t,
+					     tname, &tr->func.model);
+		if (ret < 0)
+			goto out;
+		if (tgt_prog) {
+			if (!tgt_prog->jited) {
+				/* for now */
+				verbose(env, "Can trace only JITed BPF progs\n");
+				ret = -EINVAL;
+				goto out;
+			}
+			if (tgt_prog->type == BPF_PROG_TYPE_TRACING) {
+				/* prevent cycles */
+				verbose(env, "Cannot recursively attach\n");
+				ret = -EINVAL;
+				goto out;
+			}
+			addr = (long) tgt_prog->aux->func[subprog]->bpf_func;
+		} else {
+			addr = kallsyms_lookup_name(tname);
+			if (!addr) {
+				verbose(env,
+					"The address of function %s cannot be found\n",
+					tname);
+				ret = -ENOENT;
+				goto out;
+			}
+		}
+		tr->func.addr = (void *)addr;
+		prog->aux->trampoline = tr;
+out:
+		mutex_unlock(&tr->mutex);
+		if (ret)
+			bpf_trampoline_put(tr);
+		return ret;
 	default:
 		return -EINVAL;
 	}