1 /* SPDX-License-Identifier: GPL-2.0 */
2 /* Copyright (c) 2017 - 2018 Covalent IO, Inc. http://covalent.io */
8 #include <linux/filter.h>
9 #include <linux/scatterlist.h>
10 #include <linux/skbuff.h>
14 #include <net/strparser.h>
16 #define MAX_MSG_FRAGS MAX_SKB_FRAGS
31 bool copy[MAX_MSG_FRAGS];
32 /* The extra element is used for chaining the front and sections when
33 * the list becomes partitioned (e.g. end < start). The crypto APIs
34 * require the chaining.
36 struct scatterlist data[MAX_MSG_FRAGS + 1];
39 /* UAPI in filter.c depends on struct sk_msg_sg being first element. */
48 struct sock *sk_redir;
50 struct list_head list;
53 struct sk_psock_progs {
54 struct bpf_prog *msg_parser;
55 struct bpf_prog *skb_parser;
56 struct bpf_prog *skb_verdict;
59 enum sk_psock_state_bits {
63 struct sk_psock_link {
64 struct list_head list;
69 struct sk_psock_parser {
70 struct strparser strp;
72 void (*saved_data_ready)(struct sock *sk);
75 struct sk_psock_work_state {
83 struct sock *sk_redir;
88 struct sk_psock_progs progs;
89 struct sk_psock_parser parser;
90 struct sk_buff_head ingress_skb;
91 struct list_head ingress_msg;
93 struct list_head link;
96 void (*saved_unhash)(struct sock *sk);
97 void (*saved_close)(struct sock *sk, long timeout);
98 void (*saved_write_space)(struct sock *sk);
99 struct proto *sk_proto;
100 struct sk_psock_work_state work_state;
101 struct work_struct work;
104 struct work_struct gc;
108 int sk_msg_alloc(struct sock *sk, struct sk_msg *msg, int len,
109 int elem_first_coalesce);
110 int sk_msg_clone(struct sock *sk, struct sk_msg *dst, struct sk_msg *src,
112 void sk_msg_trim(struct sock *sk, struct sk_msg *msg, int len);
113 int sk_msg_free(struct sock *sk, struct sk_msg *msg);
114 int sk_msg_free_nocharge(struct sock *sk, struct sk_msg *msg);
115 void sk_msg_free_partial(struct sock *sk, struct sk_msg *msg, u32 bytes);
116 void sk_msg_free_partial_nocharge(struct sock *sk, struct sk_msg *msg,
119 void sk_msg_return(struct sock *sk, struct sk_msg *msg, int bytes);
120 void sk_msg_return_zero(struct sock *sk, struct sk_msg *msg, int bytes);
122 int sk_msg_zerocopy_from_iter(struct sock *sk, struct iov_iter *from,
123 struct sk_msg *msg, u32 bytes);
124 int sk_msg_memcopy_from_iter(struct sock *sk, struct iov_iter *from,
125 struct sk_msg *msg, u32 bytes);
127 static inline void sk_msg_check_to_free(struct sk_msg *msg, u32 i, u32 bytes)
129 WARN_ON(i == msg->sg.end && bytes);
132 static inline void sk_msg_apply_bytes(struct sk_psock *psock, u32 bytes)
134 if (psock->apply_bytes) {
135 if (psock->apply_bytes < bytes)
136 psock->apply_bytes = 0;
138 psock->apply_bytes -= bytes;
142 static inline u32 sk_msg_iter_dist(u32 start, u32 end)
144 return end >= start ? end - start : end + (MAX_MSG_FRAGS - start);
147 #define sk_msg_iter_var_prev(var) \
150 var = MAX_MSG_FRAGS - 1; \
155 #define sk_msg_iter_var_next(var) \
158 if (var == MAX_MSG_FRAGS) \
162 #define sk_msg_iter_prev(msg, which) \
163 sk_msg_iter_var_prev(msg->sg.which)
165 #define sk_msg_iter_next(msg, which) \
166 sk_msg_iter_var_next(msg->sg.which)
168 static inline void sk_msg_clear_meta(struct sk_msg *msg)
170 memset(&msg->sg, 0, offsetofend(struct sk_msg_sg, copy));
173 static inline void sk_msg_init(struct sk_msg *msg)
175 BUILD_BUG_ON(ARRAY_SIZE(msg->sg.data) - 1 != MAX_MSG_FRAGS);
176 memset(msg, 0, sizeof(*msg));
177 sg_init_marker(msg->sg.data, MAX_MSG_FRAGS);
180 static inline void sk_msg_xfer(struct sk_msg *dst, struct sk_msg *src,
183 dst->sg.data[which] = src->sg.data[which];
184 dst->sg.data[which].length = size;
185 dst->sg.size += size;
186 src->sg.data[which].length -= size;
187 src->sg.data[which].offset += size;
190 static inline void sk_msg_xfer_full(struct sk_msg *dst, struct sk_msg *src)
192 memcpy(dst, src, sizeof(*src));
196 static inline bool sk_msg_full(const struct sk_msg *msg)
198 return (msg->sg.end == msg->sg.start) && msg->sg.size;
201 static inline u32 sk_msg_elem_used(const struct sk_msg *msg)
203 if (sk_msg_full(msg))
204 return MAX_MSG_FRAGS;
206 return sk_msg_iter_dist(msg->sg.start, msg->sg.end);
209 static inline struct scatterlist *sk_msg_elem(struct sk_msg *msg, int which)
211 return &msg->sg.data[which];
214 static inline struct scatterlist sk_msg_elem_cpy(struct sk_msg *msg, int which)
216 return msg->sg.data[which];
219 static inline struct page *sk_msg_page(struct sk_msg *msg, int which)
221 return sg_page(sk_msg_elem(msg, which));
224 static inline bool sk_msg_to_ingress(const struct sk_msg *msg)
226 return msg->flags & BPF_F_INGRESS;
229 static inline void sk_msg_compute_data_pointers(struct sk_msg *msg)
231 struct scatterlist *sge = sk_msg_elem(msg, msg->sg.start);
233 if (msg->sg.copy[msg->sg.start]) {
235 msg->data_end = NULL;
237 msg->data = sg_virt(sge);
238 msg->data_end = msg->data + sge->length;
242 static inline void sk_msg_page_add(struct sk_msg *msg, struct page *page,
245 struct scatterlist *sge;
248 sge = sk_msg_elem(msg, msg->sg.end);
249 sg_set_page(sge, page, len, offset);
252 msg->sg.copy[msg->sg.end] = true;
254 sk_msg_iter_next(msg, end);
257 static inline void sk_msg_sg_copy(struct sk_msg *msg, u32 i, bool copy_state)
260 msg->sg.copy[i] = copy_state;
261 sk_msg_iter_var_next(i);
262 if (i == msg->sg.end)
267 static inline void sk_msg_sg_copy_set(struct sk_msg *msg, u32 start)
269 sk_msg_sg_copy(msg, start, true);
272 static inline void sk_msg_sg_copy_clear(struct sk_msg *msg, u32 start)
274 sk_msg_sg_copy(msg, start, false);
277 static inline struct sk_psock *sk_psock(const struct sock *sk)
279 return rcu_dereference_sk_user_data(sk);
282 static inline void sk_psock_queue_msg(struct sk_psock *psock,
285 list_add_tail(&msg->list, &psock->ingress_msg);
288 static inline bool sk_psock_queue_empty(const struct sk_psock *psock)
290 return psock ? list_empty(&psock->ingress_msg) : true;
293 static inline void sk_psock_report_error(struct sk_psock *psock, int err)
295 struct sock *sk = psock->sk;
298 sk->sk_error_report(sk);
301 struct sk_psock *sk_psock_init(struct sock *sk, int node);
303 int sk_psock_init_strp(struct sock *sk, struct sk_psock *psock);
304 void sk_psock_start_strp(struct sock *sk, struct sk_psock *psock);
305 void sk_psock_stop_strp(struct sock *sk, struct sk_psock *psock);
307 int sk_psock_msg_verdict(struct sock *sk, struct sk_psock *psock,
310 static inline struct sk_psock_link *sk_psock_init_link(void)
312 return kzalloc(sizeof(struct sk_psock_link),
313 GFP_ATOMIC | __GFP_NOWARN);
316 static inline void sk_psock_free_link(struct sk_psock_link *link)
321 struct sk_psock_link *sk_psock_link_pop(struct sk_psock *psock);
322 #if defined(CONFIG_BPF_STREAM_PARSER)
323 void sk_psock_unlink(struct sock *sk, struct sk_psock_link *link);
325 static inline void sk_psock_unlink(struct sock *sk,
326 struct sk_psock_link *link)
331 void __sk_psock_purge_ingress_msg(struct sk_psock *psock);
333 static inline void sk_psock_cork_free(struct sk_psock *psock)
336 sk_msg_free(psock->sk, psock->cork);
342 static inline void sk_psock_update_proto(struct sock *sk,
343 struct sk_psock *psock,
346 psock->saved_unhash = sk->sk_prot->unhash;
347 psock->saved_close = sk->sk_prot->close;
348 psock->saved_write_space = sk->sk_write_space;
350 psock->sk_proto = sk->sk_prot;
354 static inline void sk_psock_restore_proto(struct sock *sk,
355 struct sk_psock *psock)
357 sk->sk_write_space = psock->saved_write_space;
359 if (psock->sk_proto) {
360 struct inet_connection_sock *icsk = inet_csk(sk);
361 bool has_ulp = !!icsk->icsk_ulp_data;
364 tcp_update_ulp(sk, psock->sk_proto);
366 sk->sk_prot = psock->sk_proto;
367 psock->sk_proto = NULL;
371 static inline void sk_psock_set_state(struct sk_psock *psock,
372 enum sk_psock_state_bits bit)
374 set_bit(bit, &psock->state);
377 static inline void sk_psock_clear_state(struct sk_psock *psock,
378 enum sk_psock_state_bits bit)
380 clear_bit(bit, &psock->state);
383 static inline bool sk_psock_test_state(const struct sk_psock *psock,
384 enum sk_psock_state_bits bit)
386 return test_bit(bit, &psock->state);
389 static inline struct sk_psock *sk_psock_get_checked(struct sock *sk)
391 struct sk_psock *psock;
394 psock = sk_psock(sk);
396 if (sk->sk_prot->recvmsg != tcp_bpf_recvmsg) {
397 psock = ERR_PTR(-EBUSY);
401 if (!refcount_inc_not_zero(&psock->refcnt))
402 psock = ERR_PTR(-EBUSY);
409 static inline struct sk_psock *sk_psock_get(struct sock *sk)
411 struct sk_psock *psock;
414 psock = sk_psock(sk);
415 if (psock && !refcount_inc_not_zero(&psock->refcnt))
421 void sk_psock_stop(struct sock *sk, struct sk_psock *psock);
422 void sk_psock_destroy(struct rcu_head *rcu);
423 void sk_psock_drop(struct sock *sk, struct sk_psock *psock);
425 static inline void sk_psock_put(struct sock *sk, struct sk_psock *psock)
427 if (refcount_dec_and_test(&psock->refcnt))
428 sk_psock_drop(sk, psock);
431 static inline void sk_psock_data_ready(struct sock *sk, struct sk_psock *psock)
433 if (psock->parser.enabled)
434 psock->parser.saved_data_ready(sk);
436 sk->sk_data_ready(sk);
439 static inline void psock_set_prog(struct bpf_prog **pprog,
440 struct bpf_prog *prog)
442 prog = xchg(pprog, prog);
447 static inline void psock_progs_drop(struct sk_psock_progs *progs)
449 psock_set_prog(&progs->msg_parser, NULL);
450 psock_set_prog(&progs->skb_parser, NULL);
451 psock_set_prog(&progs->skb_verdict, NULL);
454 #endif /* _LINUX_SKMSG_H */