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
17 #define NR_MSG_FRAG_IDS (MAX_MSG_FRAGS + 1)
32 DECLARE_BITMAP(copy, MAX_MSG_FRAGS + 2);
33 /* The extra two elements:
34 * 1) used for chaining the front and sections when the list becomes
35 * partitioned (e.g. end < start). The crypto APIs require the
37 * 2) to chain tailer SG entries after the message.
39 struct scatterlist data[MAX_MSG_FRAGS + 2];
42 /* UAPI in filter.c depends on struct sk_msg_sg being first element. */
51 struct sock *sk_redir;
53 struct list_head list;
56 struct sk_psock_progs {
57 struct bpf_prog *msg_parser;
58 struct bpf_prog *stream_parser;
59 struct bpf_prog *stream_verdict;
60 struct bpf_prog *skb_verdict;
63 enum sk_psock_state_bits {
67 struct sk_psock_link {
68 struct list_head list;
73 struct sk_psock_work_state {
81 struct sock *sk_redir;
86 struct sk_psock_progs progs;
87 #if IS_ENABLED(CONFIG_BPF_STREAM_PARSER)
88 struct strparser strp;
90 struct sk_buff_head ingress_skb;
91 struct list_head ingress_msg;
92 spinlock_t ingress_lock;
94 struct list_head link;
97 void (*saved_unhash)(struct sock *sk);
98 void (*saved_destroy)(struct sock *sk);
99 void (*saved_close)(struct sock *sk, long timeout);
100 void (*saved_write_space)(struct sock *sk);
101 void (*saved_data_ready)(struct sock *sk);
102 int (*psock_update_sk_prot)(struct sock *sk, struct sk_psock *psock,
104 struct proto *sk_proto;
105 struct mutex work_mutex;
106 struct sk_psock_work_state work_state;
107 struct work_struct work;
108 struct rcu_work rwork;
111 int sk_msg_alloc(struct sock *sk, struct sk_msg *msg, int len,
112 int elem_first_coalesce);
113 int sk_msg_clone(struct sock *sk, struct sk_msg *dst, struct sk_msg *src,
115 void sk_msg_trim(struct sock *sk, struct sk_msg *msg, int len);
116 int sk_msg_free(struct sock *sk, struct sk_msg *msg);
117 int sk_msg_free_nocharge(struct sock *sk, struct sk_msg *msg);
118 void sk_msg_free_partial(struct sock *sk, struct sk_msg *msg, u32 bytes);
119 void sk_msg_free_partial_nocharge(struct sock *sk, struct sk_msg *msg,
122 void sk_msg_return(struct sock *sk, struct sk_msg *msg, int bytes);
123 void sk_msg_return_zero(struct sock *sk, struct sk_msg *msg, int bytes);
125 int sk_msg_zerocopy_from_iter(struct sock *sk, struct iov_iter *from,
126 struct sk_msg *msg, u32 bytes);
127 int sk_msg_memcopy_from_iter(struct sock *sk, struct iov_iter *from,
128 struct sk_msg *msg, u32 bytes);
129 int sk_msg_recvmsg(struct sock *sk, struct sk_psock *psock, struct msghdr *msg,
131 bool sk_msg_is_readable(struct sock *sk);
133 static inline void sk_msg_check_to_free(struct sk_msg *msg, u32 i, u32 bytes)
135 WARN_ON(i == msg->sg.end && bytes);
138 static inline void sk_msg_apply_bytes(struct sk_psock *psock, u32 bytes)
140 if (psock->apply_bytes) {
141 if (psock->apply_bytes < bytes)
142 psock->apply_bytes = 0;
144 psock->apply_bytes -= bytes;
148 static inline u32 sk_msg_iter_dist(u32 start, u32 end)
150 return end >= start ? end - start : end + (NR_MSG_FRAG_IDS - start);
153 #define sk_msg_iter_var_prev(var) \
156 var = NR_MSG_FRAG_IDS - 1; \
161 #define sk_msg_iter_var_next(var) \
164 if (var == NR_MSG_FRAG_IDS) \
168 #define sk_msg_iter_prev(msg, which) \
169 sk_msg_iter_var_prev(msg->sg.which)
171 #define sk_msg_iter_next(msg, which) \
172 sk_msg_iter_var_next(msg->sg.which)
174 static inline void sk_msg_init(struct sk_msg *msg)
176 BUILD_BUG_ON(ARRAY_SIZE(msg->sg.data) - 1 != NR_MSG_FRAG_IDS);
177 memset(msg, 0, sizeof(*msg));
178 sg_init_marker(msg->sg.data, NR_MSG_FRAG_IDS);
181 static inline void sk_msg_xfer(struct sk_msg *dst, struct sk_msg *src,
184 dst->sg.data[which] = src->sg.data[which];
185 dst->sg.data[which].length = size;
186 dst->sg.size += size;
187 src->sg.size -= size;
188 src->sg.data[which].length -= size;
189 src->sg.data[which].offset += size;
192 static inline void sk_msg_xfer_full(struct sk_msg *dst, struct sk_msg *src)
194 memcpy(dst, src, sizeof(*src));
198 static inline bool sk_msg_full(const struct sk_msg *msg)
200 return sk_msg_iter_dist(msg->sg.start, msg->sg.end) == MAX_MSG_FRAGS;
203 static inline u32 sk_msg_elem_used(const struct sk_msg *msg)
205 return sk_msg_iter_dist(msg->sg.start, msg->sg.end);
208 static inline struct scatterlist *sk_msg_elem(struct sk_msg *msg, int which)
210 return &msg->sg.data[which];
213 static inline struct scatterlist sk_msg_elem_cpy(struct sk_msg *msg, int which)
215 return msg->sg.data[which];
218 static inline struct page *sk_msg_page(struct sk_msg *msg, int which)
220 return sg_page(sk_msg_elem(msg, which));
223 static inline bool sk_msg_to_ingress(const struct sk_msg *msg)
225 return msg->flags & BPF_F_INGRESS;
228 static inline void sk_msg_compute_data_pointers(struct sk_msg *msg)
230 struct scatterlist *sge = sk_msg_elem(msg, msg->sg.start);
232 if (test_bit(msg->sg.start, msg->sg.copy)) {
234 msg->data_end = NULL;
236 msg->data = sg_virt(sge);
237 msg->data_end = msg->data + sge->length;
241 static inline void sk_msg_page_add(struct sk_msg *msg, struct page *page,
244 struct scatterlist *sge;
247 sge = sk_msg_elem(msg, msg->sg.end);
248 sg_set_page(sge, page, len, offset);
251 __set_bit(msg->sg.end, msg->sg.copy);
253 sk_msg_iter_next(msg, end);
256 static inline void sk_msg_sg_copy(struct sk_msg *msg, u32 i, bool copy_state)
260 __set_bit(i, msg->sg.copy);
262 __clear_bit(i, msg->sg.copy);
263 sk_msg_iter_var_next(i);
264 if (i == msg->sg.end)
269 static inline void sk_msg_sg_copy_set(struct sk_msg *msg, u32 start)
271 sk_msg_sg_copy(msg, start, true);
274 static inline void sk_msg_sg_copy_clear(struct sk_msg *msg, u32 start)
276 sk_msg_sg_copy(msg, start, false);
279 static inline struct sk_psock *sk_psock(const struct sock *sk)
281 return __rcu_dereference_sk_user_data_with_flags(sk,
285 static inline void sk_psock_set_state(struct sk_psock *psock,
286 enum sk_psock_state_bits bit)
288 set_bit(bit, &psock->state);
291 static inline void sk_psock_clear_state(struct sk_psock *psock,
292 enum sk_psock_state_bits bit)
294 clear_bit(bit, &psock->state);
297 static inline bool sk_psock_test_state(const struct sk_psock *psock,
298 enum sk_psock_state_bits bit)
300 return test_bit(bit, &psock->state);
303 static inline void sock_drop(struct sock *sk, struct sk_buff *skb)
305 sk_drops_add(sk, skb);
309 static inline void sk_psock_queue_msg(struct sk_psock *psock,
312 spin_lock_bh(&psock->ingress_lock);
313 if (sk_psock_test_state(psock, SK_PSOCK_TX_ENABLED))
314 list_add_tail(&msg->list, &psock->ingress_msg);
316 sk_msg_free(psock->sk, msg);
319 spin_unlock_bh(&psock->ingress_lock);
322 static inline struct sk_msg *sk_psock_dequeue_msg(struct sk_psock *psock)
326 spin_lock_bh(&psock->ingress_lock);
327 msg = list_first_entry_or_null(&psock->ingress_msg, struct sk_msg, list);
329 list_del(&msg->list);
330 spin_unlock_bh(&psock->ingress_lock);
334 static inline struct sk_msg *sk_psock_peek_msg(struct sk_psock *psock)
338 spin_lock_bh(&psock->ingress_lock);
339 msg = list_first_entry_or_null(&psock->ingress_msg, struct sk_msg, list);
340 spin_unlock_bh(&psock->ingress_lock);
344 static inline struct sk_msg *sk_psock_next_msg(struct sk_psock *psock,
349 spin_lock_bh(&psock->ingress_lock);
350 if (list_is_last(&msg->list, &psock->ingress_msg))
353 ret = list_next_entry(msg, list);
354 spin_unlock_bh(&psock->ingress_lock);
358 static inline bool sk_psock_queue_empty(const struct sk_psock *psock)
360 return psock ? list_empty(&psock->ingress_msg) : true;
363 static inline void kfree_sk_msg(struct sk_msg *msg)
366 consume_skb(msg->skb);
370 static inline void sk_psock_report_error(struct sk_psock *psock, int err)
372 struct sock *sk = psock->sk;
378 struct sk_psock *sk_psock_init(struct sock *sk, int node);
379 void sk_psock_stop(struct sk_psock *psock, bool wait);
381 #if IS_ENABLED(CONFIG_BPF_STREAM_PARSER)
382 int sk_psock_init_strp(struct sock *sk, struct sk_psock *psock);
383 void sk_psock_start_strp(struct sock *sk, struct sk_psock *psock);
384 void sk_psock_stop_strp(struct sock *sk, struct sk_psock *psock);
386 static inline int sk_psock_init_strp(struct sock *sk, struct sk_psock *psock)
391 static inline void sk_psock_start_strp(struct sock *sk, struct sk_psock *psock)
395 static inline void sk_psock_stop_strp(struct sock *sk, struct sk_psock *psock)
400 void sk_psock_start_verdict(struct sock *sk, struct sk_psock *psock);
401 void sk_psock_stop_verdict(struct sock *sk, struct sk_psock *psock);
403 int sk_psock_msg_verdict(struct sock *sk, struct sk_psock *psock,
406 static inline struct sk_psock_link *sk_psock_init_link(void)
408 return kzalloc(sizeof(struct sk_psock_link),
409 GFP_ATOMIC | __GFP_NOWARN);
412 static inline void sk_psock_free_link(struct sk_psock_link *link)
417 struct sk_psock_link *sk_psock_link_pop(struct sk_psock *psock);
419 static inline void sk_psock_cork_free(struct sk_psock *psock)
422 sk_msg_free(psock->sk, psock->cork);
428 static inline void sk_psock_restore_proto(struct sock *sk,
429 struct sk_psock *psock)
431 if (psock->psock_update_sk_prot)
432 psock->psock_update_sk_prot(sk, psock, true);
435 static inline struct sk_psock *sk_psock_get(struct sock *sk)
437 struct sk_psock *psock;
440 psock = sk_psock(sk);
441 if (psock && !refcount_inc_not_zero(&psock->refcnt))
447 void sk_psock_drop(struct sock *sk, struct sk_psock *psock);
449 static inline void sk_psock_put(struct sock *sk, struct sk_psock *psock)
451 if (refcount_dec_and_test(&psock->refcnt))
452 sk_psock_drop(sk, psock);
455 static inline void sk_psock_data_ready(struct sock *sk, struct sk_psock *psock)
457 if (psock->saved_data_ready)
458 psock->saved_data_ready(sk);
460 sk->sk_data_ready(sk);
463 static inline void psock_set_prog(struct bpf_prog **pprog,
464 struct bpf_prog *prog)
466 prog = xchg(pprog, prog);
471 static inline int psock_replace_prog(struct bpf_prog **pprog,
472 struct bpf_prog *prog,
473 struct bpf_prog *old)
475 if (cmpxchg(pprog, old, prog) != old)
484 static inline void psock_progs_drop(struct sk_psock_progs *progs)
486 psock_set_prog(&progs->msg_parser, NULL);
487 psock_set_prog(&progs->stream_parser, NULL);
488 psock_set_prog(&progs->stream_verdict, NULL);
489 psock_set_prog(&progs->skb_verdict, NULL);
492 int sk_psock_tls_strp_read(struct sk_psock *psock, struct sk_buff *skb);
494 static inline bool sk_psock_strp_enabled(struct sk_psock *psock)
498 return !!psock->saved_data_ready;
501 static inline bool sk_is_udp(const struct sock *sk)
503 return sk->sk_type == SOCK_DGRAM &&
504 sk->sk_protocol == IPPROTO_UDP;
507 #if IS_ENABLED(CONFIG_NET_SOCK_MSG)
509 #define BPF_F_STRPARSER (1UL << 1)
511 /* We only have two bits so far. */
512 #define BPF_F_PTR_MASK ~(BPF_F_INGRESS | BPF_F_STRPARSER)
514 static inline bool skb_bpf_strparser(const struct sk_buff *skb)
516 unsigned long sk_redir = skb->_sk_redir;
518 return sk_redir & BPF_F_STRPARSER;
521 static inline void skb_bpf_set_strparser(struct sk_buff *skb)
523 skb->_sk_redir |= BPF_F_STRPARSER;
526 static inline bool skb_bpf_ingress(const struct sk_buff *skb)
528 unsigned long sk_redir = skb->_sk_redir;
530 return sk_redir & BPF_F_INGRESS;
533 static inline void skb_bpf_set_ingress(struct sk_buff *skb)
535 skb->_sk_redir |= BPF_F_INGRESS;
538 static inline void skb_bpf_set_redir(struct sk_buff *skb, struct sock *sk_redir,
541 skb->_sk_redir = (unsigned long)sk_redir;
543 skb->_sk_redir |= BPF_F_INGRESS;
546 static inline struct sock *skb_bpf_redirect_fetch(const struct sk_buff *skb)
548 unsigned long sk_redir = skb->_sk_redir;
550 return (struct sock *)(sk_redir & BPF_F_PTR_MASK);
553 static inline void skb_bpf_redirect_clear(struct sk_buff *skb)
557 #endif /* CONFIG_NET_SOCK_MSG */
558 #endif /* _LINUX_SKMSG_H */