[linux-block.git] / net / ipv4 / bpf_tcp_ca.c

// SPDX-License-Identifier: GPL-2.0
/* Copyright (c) 2019 Facebook  */

#include <linux/init.h>
#include <linux/types.h>
#include <linux/bpf_verifier.h>
#include <linux/bpf.h>
#include <linux/btf.h>
#include <linux/btf_ids.h>
#include <linux/filter.h>
#include <net/tcp.h>
#include <net/bpf_sk_storage.h>

/* "extern" is to avoid sparse warning.  It is only used in bpf_struct_ops.c. */
extern struct bpf_struct_ops bpf_tcp_congestion_ops;

static u32 unsupported_ops[] = {
	offsetof(struct tcp_congestion_ops, get_info),
};

static const struct btf_type *tcp_sock_type;
static u32 tcp_sock_id, sock_id;

static int bpf_tcp_ca_init(struct btf *btf)
{
	s32 type_id;

	type_id = btf_find_by_name_kind(btf, "sock", BTF_KIND_STRUCT);
	if (type_id < 0)
		return -EINVAL;
	sock_id = type_id;

	type_id = btf_find_by_name_kind(btf, "tcp_sock", BTF_KIND_STRUCT);
	if (type_id < 0)
		return -EINVAL;
	tcp_sock_id = type_id;
	tcp_sock_type = btf_type_by_id(btf, tcp_sock_id);

	return 0;
}

static bool is_unsupported(u32 member_offset)
{
	unsigned int i;

	for (i = 0; i < ARRAY_SIZE(unsupported_ops); i++) {
		if (member_offset == unsupported_ops[i])
			return true;
	}

	return false;
}

extern struct btf *btf_vmlinux;

static bool bpf_tcp_ca_is_valid_access(int off, int size,
				       enum bpf_access_type type,
				       const struct bpf_prog *prog,
				       struct bpf_insn_access_aux *info)
{
	if (!bpf_tracing_btf_ctx_access(off, size, type, prog, info))
		return false;

	if (info->reg_type == PTR_TO_BTF_ID && info->btf_id == sock_id)
		/* promote it to tcp_sock */
		info->btf_id = tcp_sock_id;

	return true;
}

static int bpf_tcp_ca_btf_struct_access(struct bpf_verifier_log *log,
					const struct btf *btf,
					const struct btf_type *t, int off,
					int size, enum bpf_access_type atype,
					u32 *next_btf_id,
					enum bpf_type_flag *flag)
{
	size_t end;

	if (atype == BPF_READ)
		return btf_struct_access(log, btf, t, off, size, atype, next_btf_id,
					 flag);

	if (t != tcp_sock_type) {
		bpf_log(log, "only read is supported\n");
		return -EACCES;
	}

	switch (off) {
	case offsetof(struct sock, sk_pacing_rate):
		end = offsetofend(struct sock, sk_pacing_rate);
		break;
	case offsetof(struct sock, sk_pacing_status):
		end = offsetofend(struct sock, sk_pacing_status);
		break;
	case bpf_ctx_range(struct inet_connection_sock, icsk_ca_priv):
		end = offsetofend(struct inet_connection_sock, icsk_ca_priv);
		break;
	case offsetof(struct inet_connection_sock, icsk_ack.pending):
		end = offsetofend(struct inet_connection_sock,
				  icsk_ack.pending);
		break;
	case offsetof(struct tcp_sock, snd_cwnd):
		end = offsetofend(struct tcp_sock, snd_cwnd);
		break;
	case offsetof(struct tcp_sock, snd_cwnd_cnt):
		end = offsetofend(struct tcp_sock, snd_cwnd_cnt);
		break;
	case offsetof(struct tcp_sock, snd_ssthresh):
		end = offsetofend(struct tcp_sock, snd_ssthresh);
		break;
	case offsetof(struct tcp_sock, ecn_flags):
		end = offsetofend(struct tcp_sock, ecn_flags);
		break;
	default:
		bpf_log(log, "no write support to tcp_sock at off %d\n", off);
		return -EACCES;
	}

	if (off + size > end) {
		bpf_log(log,
			"write access at off %d with size %d beyond the member of tcp_sock ended at %zu\n",
			off, size, end);
		return -EACCES;
	}

	return NOT_INIT;
}

BPF_CALL_2(bpf_tcp_send_ack, struct tcp_sock *, tp, u32, rcv_nxt)
{
	/* bpf_tcp_ca prog cannot have NULL tp */
	__tcp_send_ack((struct sock *)tp, rcv_nxt);
	return 0;
}

static const struct bpf_func_proto bpf_tcp_send_ack_proto = {
	.func		= bpf_tcp_send_ack,
	.gpl_only	= false,
	/* In case we want to report error later */
	.ret_type	= RET_INTEGER,
	.arg1_type	= ARG_PTR_TO_BTF_ID,
	.arg1_btf_id	= &tcp_sock_id,
	.arg2_type	= ARG_ANYTHING,
};

static u32 prog_ops_moff(const struct bpf_prog *prog)
{
	const struct btf_member *m;
	const struct btf_type *t;
	u32 midx;

	midx = prog->expected_attach_type;
	t = bpf_tcp_congestion_ops.type;
	m = &btf_type_member(t)[midx];

	return __btf_member_bit_offset(t, m) / 8;
}

static const struct bpf_func_proto *
bpf_tcp_ca_get_func_proto(enum bpf_func_id func_id,
			  const struct bpf_prog *prog)
{
	switch (func_id) {
	case BPF_FUNC_tcp_send_ack:
		return &bpf_tcp_send_ack_proto;
	case BPF_FUNC_sk_storage_get:
		return &bpf_sk_storage_get_proto;
	case BPF_FUNC_sk_storage_delete:
		return &bpf_sk_storage_delete_proto;
	case BPF_FUNC_setsockopt:
		/* Does not allow release() to call setsockopt.
		 * release() is called when the current bpf-tcp-cc
		 * is retiring.  It is not allowed to call
		 * setsockopt() to make further changes which
		 * may potentially allocate new resources.
		 */
		if (prog_ops_moff(prog) !=
		    offsetof(struct tcp_congestion_ops, release))
			return &bpf_sk_setsockopt_proto;
		return NULL;
	case BPF_FUNC_getsockopt:
		/* Since get/setsockopt is usually expected to
		 * be available together, disable getsockopt for
		 * release also to avoid usage surprise.
		 * The bpf-tcp-cc already has a more powerful way
		 * to read tcp_sock from the PTR_TO_BTF_ID.
		 */
		if (prog_ops_moff(prog) !=
		    offsetof(struct tcp_congestion_ops, release))
			return &bpf_sk_getsockopt_proto;
		return NULL;
	case BPF_FUNC_ktime_get_coarse_ns:
		return &bpf_ktime_get_coarse_ns_proto;
	default:
		return bpf_base_func_proto(func_id);
	}
}

BTF_SET8_START(bpf_tcp_ca_check_kfunc_ids)
BTF_ID_FLAGS(func, tcp_reno_ssthresh)
BTF_ID_FLAGS(func, tcp_reno_cong_avoid)
BTF_ID_FLAGS(func, tcp_reno_undo_cwnd)
BTF_ID_FLAGS(func, tcp_slow_start)
BTF_ID_FLAGS(func, tcp_cong_avoid_ai)
BTF_SET8_END(bpf_tcp_ca_check_kfunc_ids)

static const struct btf_kfunc_id_set bpf_tcp_ca_kfunc_set = {
	.owner = THIS_MODULE,
	.set   = &bpf_tcp_ca_check_kfunc_ids,
};

static const struct bpf_verifier_ops bpf_tcp_ca_verifier_ops = {
	.get_func_proto		= bpf_tcp_ca_get_func_proto,
	.is_valid_access	= bpf_tcp_ca_is_valid_access,
	.btf_struct_access	= bpf_tcp_ca_btf_struct_access,
};

static int bpf_tcp_ca_init_member(const struct btf_type *t,
				  const struct btf_member *member,
				  void *kdata, const void *udata)
{
	const struct tcp_congestion_ops *utcp_ca;
	struct tcp_congestion_ops *tcp_ca;
	u32 moff;

	utcp_ca = (const struct tcp_congestion_ops *)udata;
	tcp_ca = (struct tcp_congestion_ops *)kdata;

	moff = __btf_member_bit_offset(t, member) / 8;
	switch (moff) {
	case offsetof(struct tcp_congestion_ops, flags):
		if (utcp_ca->flags & ~TCP_CONG_MASK)
			return -EINVAL;
		tcp_ca->flags = utcp_ca->flags;
		return 1;
	case offsetof(struct tcp_congestion_ops, name):
		if (bpf_obj_name_cpy(tcp_ca->name, utcp_ca->name,
				     sizeof(tcp_ca->name)) <= 0)
			return -EINVAL;
		if (tcp_ca_find(utcp_ca->name))
			return -EEXIST;
		return 1;
	}

	return 0;
}

static int bpf_tcp_ca_check_member(const struct btf_type *t,
				   const struct btf_member *member)
{
	if (is_unsupported(__btf_member_bit_offset(t, member) / 8))
		return -ENOTSUPP;
	return 0;
}

static int bpf_tcp_ca_reg(void *kdata)
{
	return tcp_register_congestion_control(kdata);
}

static void bpf_tcp_ca_unreg(void *kdata)
{
	tcp_unregister_congestion_control(kdata);
}

struct bpf_struct_ops bpf_tcp_congestion_ops = {
	.verifier_ops = &bpf_tcp_ca_verifier_ops,
	.reg = bpf_tcp_ca_reg,
	.unreg = bpf_tcp_ca_unreg,
	.check_member = bpf_tcp_ca_check_member,
	.init_member = bpf_tcp_ca_init_member,
	.init = bpf_tcp_ca_init,
	.name = "tcp_congestion_ops",
};

static int __init bpf_tcp_ca_kfunc_init(void)
{
	return register_btf_kfunc_id_set(BPF_PROG_TYPE_STRUCT_OPS, &bpf_tcp_ca_kfunc_set);
}
late_initcall(bpf_tcp_ca_kfunc_init);
Commit	Line	Data
0baf26b0 MKL	1	// SPDX-License-Identifier: GPL-2.0
	2	/* Copyright (c) 2019 Facebook */
	3
b202d844	4	#include <linux/init.h>
0baf26b0 MKL	5	#include <linux/types.h>
	6	#include <linux/bpf_verifier.h>
	7	#include <linux/bpf.h>
	8	#include <linux/btf.h>
e78aea8b	9	#include <linux/btf_ids.h>
0baf26b0 MKL	10	#include <linux/filter.h>
0baf26b0 MKL	11	#include <net/tcp.h>
ab14fd4e	12	#include <net/bpf_sk_storage.h>
0baf26b0	13
eb18b49e MKL	14	/* "extern" is to avoid sparse warning. It is only used in bpf_struct_ops.c. */
	15	extern struct bpf_struct_ops bpf_tcp_congestion_ops;
	16
0baf26b0 MKL	17	static u32 unsupported_ops[] = {
	18	offsetof(struct tcp_congestion_ops, get_info),
	19	};
	20
	21	static const struct btf_type *tcp_sock_type;
	22	static u32 tcp_sock_id, sock_id;
	23
	24	static int bpf_tcp_ca_init(struct btf *btf)
	25	{
	26	s32 type_id;
	27
	28	type_id = btf_find_by_name_kind(btf, "sock", BTF_KIND_STRUCT);
	29	if (type_id < 0)
	30	return -EINVAL;
	31	sock_id = type_id;
	32
	33	type_id = btf_find_by_name_kind(btf, "tcp_sock", BTF_KIND_STRUCT);
	34	if (type_id < 0)
	35	return -EINVAL;
	36	tcp_sock_id = type_id;
	37	tcp_sock_type = btf_type_by_id(btf, tcp_sock_id);
	38
	39	return 0;
	40	}
	41
0baf26b0 MKL	42	static bool is_unsupported(u32 member_offset)
	43	{
	44	unsigned int i;
	45
	46	for (i = 0; i < ARRAY_SIZE(unsupported_ops); i++) {
	47	if (member_offset == unsupported_ops[i])
	48	return true;
	49	}
	50
	51	return false;
	52	}
	53
	54	extern struct btf *btf_vmlinux;
	55
	56	static bool bpf_tcp_ca_is_valid_access(int off, int size,
	57	enum bpf_access_type type,
	58	const struct bpf_prog *prog,
	59	struct bpf_insn_access_aux *info)
	60	{
35346ab6	61	if (!bpf_tracing_btf_ctx_access(off, size, type, prog, info))
0baf26b0 MKL	62	return false;
	63
	64	if (info->reg_type == PTR_TO_BTF_ID && info->btf_id == sock_id)
	65	/* promote it to tcp_sock */
	66	info->btf_id = tcp_sock_id;
	67
	68	return true;
	69	}
	70
	71	static int bpf_tcp_ca_btf_struct_access(struct bpf_verifier_log *log,
22dc4a0f	72	const struct btf *btf,
0baf26b0 MKL	73	const struct btf_type *t, int off,
0baf26b0 MKL	74	int size, enum bpf_access_type atype,
c6f1bfe8 YS	75	u32 *next_btf_id,
c6f1bfe8 YS	76	enum bpf_type_flag *flag)
0baf26b0 MKL	77	{
	78	size_t end;
	79
	80	if (atype == BPF_READ)
c6f1bfe8 YS	81	return btf_struct_access(log, btf, t, off, size, atype, next_btf_id,
c6f1bfe8 YS	82	flag);
0baf26b0 MKL	83
	84	if (t != tcp_sock_type) {
	85	bpf_log(log, "only read is supported\n");
	86	return -EACCES;
	87	}
	88
	89	switch (off) {
41c95dd6 JTH	90	case offsetof(struct sock, sk_pacing_rate):
	91	end = offsetofend(struct sock, sk_pacing_rate);
	92	break;
	93	case offsetof(struct sock, sk_pacing_status):
	94	end = offsetofend(struct sock, sk_pacing_status);
	95	break;
0baf26b0 MKL	96	case bpf_ctx_range(struct inet_connection_sock, icsk_ca_priv):
	97	end = offsetofend(struct inet_connection_sock, icsk_ca_priv);
	98	break;
	99	case offsetof(struct inet_connection_sock, icsk_ack.pending):
	100	end = offsetofend(struct inet_connection_sock,
	101	icsk_ack.pending);
	102	break;
	103	case offsetof(struct tcp_sock, snd_cwnd):
	104	end = offsetofend(struct tcp_sock, snd_cwnd);
	105	break;
	106	case offsetof(struct tcp_sock, snd_cwnd_cnt):
	107	end = offsetofend(struct tcp_sock, snd_cwnd_cnt);
	108	break;
	109	case offsetof(struct tcp_sock, snd_ssthresh):
	110	end = offsetofend(struct tcp_sock, snd_ssthresh);
	111	break;
	112	case offsetof(struct tcp_sock, ecn_flags):
	113	end = offsetofend(struct tcp_sock, ecn_flags);
	114	break;
	115	default:
	116	bpf_log(log, "no write support to tcp_sock at off %d\n", off);
	117	return -EACCES;
	118	}
	119
	120	if (off + size > end) {
	121	bpf_log(log,
	122	"write access at off %d with size %d beyond the member of tcp_sock ended at %zu\n",
	123	off, size, end);
	124	return -EACCES;
	125	}
	126
	127	return NOT_INIT;
	128	}
	129
206057fe MKL	130	BPF_CALL_2(bpf_tcp_send_ack, struct tcp_sock *, tp, u32, rcv_nxt)
	131	{
	132	/* bpf_tcp_ca prog cannot have NULL tp */
	133	__tcp_send_ack((struct sock *)tp, rcv_nxt);
	134	return 0;
	135	}
	136
	137	static const struct bpf_func_proto bpf_tcp_send_ack_proto = {
	138	.func = bpf_tcp_send_ack,
	139	.gpl_only = false,
	140	/* In case we want to report error later */
	141	.ret_type = RET_INTEGER,
	142	.arg1_type = ARG_PTR_TO_BTF_ID,
9436ef6e	143	.arg1_btf_id = &tcp_sock_id,
206057fe	144	.arg2_type = ARG_ANYTHING,
206057fe MKL	145	};
206057fe MKL	146
eb18b49e MKL	147	static u32 prog_ops_moff(const struct bpf_prog *prog)
	148	{
	149	const struct btf_member *m;
	150	const struct btf_type *t;
	151	u32 midx;
	152
	153	midx = prog->expected_attach_type;
	154	t = bpf_tcp_congestion_ops.type;
	155	m = &btf_type_member(t)[midx];
	156
8293eb99	157	return __btf_member_bit_offset(t, m) / 8;
eb18b49e MKL	158	}
eb18b49e MKL	159
0baf26b0 MKL	160	static const struct bpf_func_proto *
	161	bpf_tcp_ca_get_func_proto(enum bpf_func_id func_id,
	162	const struct bpf_prog *prog)
	163	{
206057fe MKL	164	switch (func_id) {
	165	case BPF_FUNC_tcp_send_ack:
	166	return &bpf_tcp_send_ack_proto;
ab14fd4e	167	case BPF_FUNC_sk_storage_get:
592a3498	168	return &bpf_sk_storage_get_proto;
ab14fd4e	169	case BPF_FUNC_sk_storage_delete:
592a3498	170	return &bpf_sk_storage_delete_proto;
eb18b49e MKL	171	case BPF_FUNC_setsockopt:
	172	/* Does not allow release() to call setsockopt.
	173	* release() is called when the current bpf-tcp-cc
	174	* is retiring. It is not allowed to call
	175	* setsockopt() to make further changes which
	176	* may potentially allocate new resources.
	177	*/
	178	if (prog_ops_moff(prog) !=
	179	offsetof(struct tcp_congestion_ops, release))
	180	return &bpf_sk_setsockopt_proto;
	181	return NULL;
	182	case BPF_FUNC_getsockopt:
	183	/* Since get/setsockopt is usually expected to
	184	* be available together, disable getsockopt for
	185	* release also to avoid usage surprise.
	186	* The bpf-tcp-cc already has a more powerful way
	187	* to read tcp_sock from the PTR_TO_BTF_ID.
	188	*/
	189	if (prog_ops_moff(prog) !=
	190	offsetof(struct tcp_congestion_ops, release))
	191	return &bpf_sk_getsockopt_proto;
	192	return NULL;
5e0bc308 DB	193	case BPF_FUNC_ktime_get_coarse_ns:
5e0bc308 DB	194	return &bpf_ktime_get_coarse_ns_proto;
206057fe MKL	195	default:
	196	return bpf_base_func_proto(func_id);
	197	}
0baf26b0 MKL	198	}
0baf26b0 MKL	199
a4703e31 KKD	200	BTF_SET8_START(bpf_tcp_ca_check_kfunc_ids)
	201	BTF_ID_FLAGS(func, tcp_reno_ssthresh)
	202	BTF_ID_FLAGS(func, tcp_reno_cong_avoid)
	203	BTF_ID_FLAGS(func, tcp_reno_undo_cwnd)
	204	BTF_ID_FLAGS(func, tcp_slow_start)
	205	BTF_ID_FLAGS(func, tcp_cong_avoid_ai)
	206	BTF_SET8_END(bpf_tcp_ca_check_kfunc_ids)
e78aea8b	207
b202d844	208	static const struct btf_kfunc_id_set bpf_tcp_ca_kfunc_set = {
a4703e31 KKD	209	.owner = THIS_MODULE,
a4703e31 KKD	210	.set = &bpf_tcp_ca_check_kfunc_ids,
b202d844	211	};
e78aea8b	212
0baf26b0 MKL	213	static const struct bpf_verifier_ops bpf_tcp_ca_verifier_ops = {
	214	.get_func_proto = bpf_tcp_ca_get_func_proto,
	215	.is_valid_access = bpf_tcp_ca_is_valid_access,
	216	.btf_struct_access = bpf_tcp_ca_btf_struct_access,
	217	};
	218
	219	static int bpf_tcp_ca_init_member(const struct btf_type *t,
	220	const struct btf_member *member,
	221	void kdata, const void udata)
	222	{
	223	const struct tcp_congestion_ops *utcp_ca;
	224	struct tcp_congestion_ops *tcp_ca;
0baf26b0 MKL	225	u32 moff;
	226
	227	utcp_ca = (const struct tcp_congestion_ops *)udata;
	228	tcp_ca = (struct tcp_congestion_ops *)kdata;
	229
8293eb99	230	moff = __btf_member_bit_offset(t, member) / 8;
0baf26b0 MKL	231	switch (moff) {
	232	case offsetof(struct tcp_congestion_ops, flags):
	233	if (utcp_ca->flags & ~TCP_CONG_MASK)
	234	return -EINVAL;
	235	tcp_ca->flags = utcp_ca->flags;
	236	return 1;
	237	case offsetof(struct tcp_congestion_ops, name):
8e7ae251 MKL	238	if (bpf_obj_name_cpy(tcp_ca->name, utcp_ca->name,
8e7ae251 MKL	239	sizeof(tcp_ca->name)) <= 0)
0baf26b0 MKL	240	return -EINVAL;
	241	if (tcp_ca_find(utcp_ca->name))
	242	return -EEXIST;
0baf26b0 MKL	243	return 1;
	244	}
	245
0baf26b0 MKL	246	return 0;
	247	}
	248
	249	static int bpf_tcp_ca_check_member(const struct btf_type *t,
	250	const struct btf_member *member)
	251	{
8293eb99	252	if (is_unsupported(__btf_member_bit_offset(t, member) / 8))
0baf26b0 MKL	253	return -ENOTSUPP;
	254	return 0;
	255	}
	256
	257	static int bpf_tcp_ca_reg(void *kdata)
	258	{
	259	return tcp_register_congestion_control(kdata);
	260	}
	261
	262	static void bpf_tcp_ca_unreg(void *kdata)
	263	{
	264	tcp_unregister_congestion_control(kdata);
	265	}
	266
0baf26b0 MKL	267	struct bpf_struct_ops bpf_tcp_congestion_ops = {
	268	.verifier_ops = &bpf_tcp_ca_verifier_ops,
	269	.reg = bpf_tcp_ca_reg,
	270	.unreg = bpf_tcp_ca_unreg,
	271	.check_member = bpf_tcp_ca_check_member,
	272	.init_member = bpf_tcp_ca_init_member,
	273	.init = bpf_tcp_ca_init,
	274	.name = "tcp_congestion_ops",
	275	};
b202d844 KKD	276
	277	static int __init bpf_tcp_ca_kfunc_init(void)
	278	{
	279	return register_btf_kfunc_id_set(BPF_PROG_TYPE_STRUCT_OPS, &bpf_tcp_ca_kfunc_set);
	280	}
	281	late_initcall(bpf_tcp_ca_kfunc_init);