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)
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];
41 static_assert(BITS_PER_LONG >= NR_MSG_FRAG_IDS);
43 /* UAPI in filter.c depends on struct sk_msg_sg being first element. */
52 struct sock *sk_redir;
54 struct list_head list;
57 struct sk_psock_progs {
58 struct bpf_prog *msg_parser;
59 struct bpf_prog *stream_parser;
60 struct bpf_prog *stream_verdict;
61 struct bpf_prog *skb_verdict;
64 enum sk_psock_state_bits {
68 struct sk_psock_link {
69 struct list_head list;
74 struct sk_psock_work_state {
82 struct sock *sk_redir;
87 struct sk_psock_progs progs;
88 #if IS_ENABLED(CONFIG_BPF_STREAM_PARSER)
89 struct strparser strp;
91 struct sk_buff_head ingress_skb;
92 struct list_head ingress_msg;
93 spinlock_t ingress_lock;
95 struct list_head link;
98 void (*saved_unhash)(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_wait_data(struct sock *sk, struct sk_psock *psock, int flags,
130 long timeo, int *err);
131 int sk_msg_recvmsg(struct sock *sk, struct sk_psock *psock, struct msghdr *msg,
134 static inline void sk_msg_check_to_free(struct sk_msg *msg, u32 i, u32 bytes)
136 WARN_ON(i == msg->sg.end && bytes);
139 static inline void sk_msg_apply_bytes(struct sk_psock *psock, u32 bytes)
141 if (psock->apply_bytes) {
142 if (psock->apply_bytes < bytes)
143 psock->apply_bytes = 0;
145 psock->apply_bytes -= bytes;
149 static inline u32 sk_msg_iter_dist(u32 start, u32 end)
151 return end >= start ? end - start : end + (NR_MSG_FRAG_IDS - start);
154 #define sk_msg_iter_var_prev(var) \
157 var = NR_MSG_FRAG_IDS - 1; \
162 #define sk_msg_iter_var_next(var) \
165 if (var == NR_MSG_FRAG_IDS) \
169 #define sk_msg_iter_prev(msg, which) \
170 sk_msg_iter_var_prev(msg->sg.which)
172 #define sk_msg_iter_next(msg, which) \
173 sk_msg_iter_var_next(msg->sg.which)
175 static inline void sk_msg_clear_meta(struct sk_msg *msg)
177 memset(&msg->sg, 0, offsetofend(struct sk_msg_sg, copy));
180 static inline void sk_msg_init(struct sk_msg *msg)
182 BUILD_BUG_ON(ARRAY_SIZE(msg->sg.data) - 1 != NR_MSG_FRAG_IDS);
183 memset(msg, 0, sizeof(*msg));
184 sg_init_marker(msg->sg.data, NR_MSG_FRAG_IDS);
187 static inline void sk_msg_xfer(struct sk_msg *dst, struct sk_msg *src,
190 dst->sg.data[which] = src->sg.data[which];
191 dst->sg.data[which].length = size;
192 dst->sg.size += size;
193 src->sg.size -= size;
194 src->sg.data[which].length -= size;
195 src->sg.data[which].offset += size;
198 static inline void sk_msg_xfer_full(struct sk_msg *dst, struct sk_msg *src)
200 memcpy(dst, src, sizeof(*src));
204 static inline bool sk_msg_full(const struct sk_msg *msg)
206 return sk_msg_iter_dist(msg->sg.start, msg->sg.end) == MAX_MSG_FRAGS;
209 static inline u32 sk_msg_elem_used(const struct sk_msg *msg)
211 return sk_msg_iter_dist(msg->sg.start, msg->sg.end);
214 static inline struct scatterlist *sk_msg_elem(struct sk_msg *msg, int which)
216 return &msg->sg.data[which];
219 static inline struct scatterlist sk_msg_elem_cpy(struct sk_msg *msg, int which)
221 return msg->sg.data[which];
224 static inline struct page *sk_msg_page(struct sk_msg *msg, int which)
226 return sg_page(sk_msg_elem(msg, which));
229 static inline bool sk_msg_to_ingress(const struct sk_msg *msg)
231 return msg->flags & BPF_F_INGRESS;
234 static inline void sk_msg_compute_data_pointers(struct sk_msg *msg)
236 struct scatterlist *sge = sk_msg_elem(msg, msg->sg.start);
238 if (test_bit(msg->sg.start, &msg->sg.copy)) {
240 msg->data_end = NULL;
242 msg->data = sg_virt(sge);
243 msg->data_end = msg->data + sge->length;
247 static inline void sk_msg_page_add(struct sk_msg *msg, struct page *page,
250 struct scatterlist *sge;
253 sge = sk_msg_elem(msg, msg->sg.end);
254 sg_set_page(sge, page, len, offset);
257 __set_bit(msg->sg.end, &msg->sg.copy);
259 sk_msg_iter_next(msg, end);
262 static inline void sk_msg_sg_copy(struct sk_msg *msg, u32 i, bool copy_state)
266 __set_bit(i, &msg->sg.copy);
268 __clear_bit(i, &msg->sg.copy);
269 sk_msg_iter_var_next(i);
270 if (i == msg->sg.end)
275 static inline void sk_msg_sg_copy_set(struct sk_msg *msg, u32 start)
277 sk_msg_sg_copy(msg, start, true);
280 static inline void sk_msg_sg_copy_clear(struct sk_msg *msg, u32 start)
282 sk_msg_sg_copy(msg, start, false);
285 static inline struct sk_psock *sk_psock(const struct sock *sk)
287 return rcu_dereference_sk_user_data(sk);
290 static inline void sk_psock_queue_msg(struct sk_psock *psock,
293 spin_lock_bh(&psock->ingress_lock);
294 list_add_tail(&msg->list, &psock->ingress_msg);
295 spin_unlock_bh(&psock->ingress_lock);
298 static inline struct sk_msg *sk_psock_dequeue_msg(struct sk_psock *psock)
302 spin_lock_bh(&psock->ingress_lock);
303 msg = list_first_entry_or_null(&psock->ingress_msg, struct sk_msg, list);
305 list_del(&msg->list);
306 spin_unlock_bh(&psock->ingress_lock);
310 static inline struct sk_msg *sk_psock_peek_msg(struct sk_psock *psock)
314 spin_lock_bh(&psock->ingress_lock);
315 msg = list_first_entry_or_null(&psock->ingress_msg, struct sk_msg, list);
316 spin_unlock_bh(&psock->ingress_lock);
320 static inline struct sk_msg *sk_psock_next_msg(struct sk_psock *psock,
325 spin_lock_bh(&psock->ingress_lock);
326 if (list_is_last(&msg->list, &psock->ingress_msg))
329 ret = list_next_entry(msg, list);
330 spin_unlock_bh(&psock->ingress_lock);
334 static inline bool sk_psock_queue_empty(const struct sk_psock *psock)
336 return psock ? list_empty(&psock->ingress_msg) : true;
339 static inline void kfree_sk_msg(struct sk_msg *msg)
342 consume_skb(msg->skb);
346 static inline void sk_psock_report_error(struct sk_psock *psock, int err)
348 struct sock *sk = psock->sk;
351 sk->sk_error_report(sk);
354 struct sk_psock *sk_psock_init(struct sock *sk, int node);
355 void sk_psock_stop(struct sk_psock *psock, bool wait);
357 #if IS_ENABLED(CONFIG_BPF_STREAM_PARSER)
358 int sk_psock_init_strp(struct sock *sk, struct sk_psock *psock);
359 void sk_psock_start_strp(struct sock *sk, struct sk_psock *psock);
360 void sk_psock_stop_strp(struct sock *sk, struct sk_psock *psock);
362 static inline int sk_psock_init_strp(struct sock *sk, struct sk_psock *psock)
367 static inline void sk_psock_start_strp(struct sock *sk, struct sk_psock *psock)
371 static inline void sk_psock_stop_strp(struct sock *sk, struct sk_psock *psock)
376 void sk_psock_start_verdict(struct sock *sk, struct sk_psock *psock);
377 void sk_psock_stop_verdict(struct sock *sk, struct sk_psock *psock);
379 int sk_psock_msg_verdict(struct sock *sk, struct sk_psock *psock,
382 static inline struct sk_psock_link *sk_psock_init_link(void)
384 return kzalloc(sizeof(struct sk_psock_link),
385 GFP_ATOMIC | __GFP_NOWARN);
388 static inline void sk_psock_free_link(struct sk_psock_link *link)
393 struct sk_psock_link *sk_psock_link_pop(struct sk_psock *psock);
395 static inline void sk_psock_cork_free(struct sk_psock *psock)
398 sk_msg_free(psock->sk, psock->cork);
404 static inline void sk_psock_restore_proto(struct sock *sk,
405 struct sk_psock *psock)
407 if (psock->psock_update_sk_prot)
408 psock->psock_update_sk_prot(sk, psock, true);
411 static inline void sk_psock_set_state(struct sk_psock *psock,
412 enum sk_psock_state_bits bit)
414 set_bit(bit, &psock->state);
417 static inline void sk_psock_clear_state(struct sk_psock *psock,
418 enum sk_psock_state_bits bit)
420 clear_bit(bit, &psock->state);
423 static inline bool sk_psock_test_state(const struct sk_psock *psock,
424 enum sk_psock_state_bits bit)
426 return test_bit(bit, &psock->state);
429 static inline struct sk_psock *sk_psock_get(struct sock *sk)
431 struct sk_psock *psock;
434 psock = sk_psock(sk);
435 if (psock && !refcount_inc_not_zero(&psock->refcnt))
441 void sk_psock_drop(struct sock *sk, struct sk_psock *psock);
443 static inline void sk_psock_put(struct sock *sk, struct sk_psock *psock)
445 if (refcount_dec_and_test(&psock->refcnt))
446 sk_psock_drop(sk, psock);
449 static inline void sk_psock_data_ready(struct sock *sk, struct sk_psock *psock)
451 if (psock->saved_data_ready)
452 psock->saved_data_ready(sk);
454 sk->sk_data_ready(sk);
457 static inline void psock_set_prog(struct bpf_prog **pprog,
458 struct bpf_prog *prog)
460 prog = xchg(pprog, prog);
465 static inline int psock_replace_prog(struct bpf_prog **pprog,
466 struct bpf_prog *prog,
467 struct bpf_prog *old)
469 if (cmpxchg(pprog, old, prog) != old)
478 static inline void psock_progs_drop(struct sk_psock_progs *progs)
480 psock_set_prog(&progs->msg_parser, NULL);
481 psock_set_prog(&progs->stream_parser, NULL);
482 psock_set_prog(&progs->stream_verdict, NULL);
483 psock_set_prog(&progs->skb_verdict, NULL);
486 int sk_psock_tls_strp_read(struct sk_psock *psock, struct sk_buff *skb);
488 static inline bool sk_psock_strp_enabled(struct sk_psock *psock)
492 return !!psock->saved_data_ready;
495 #if IS_ENABLED(CONFIG_NET_SOCK_MSG)
497 /* We only have one bit so far. */
498 #define BPF_F_PTR_MASK ~(BPF_F_INGRESS)
500 static inline bool skb_bpf_ingress(const struct sk_buff *skb)
502 unsigned long sk_redir = skb->_sk_redir;
504 return sk_redir & BPF_F_INGRESS;
507 static inline void skb_bpf_set_ingress(struct sk_buff *skb)
509 skb->_sk_redir |= BPF_F_INGRESS;
512 static inline void skb_bpf_set_redir(struct sk_buff *skb, struct sock *sk_redir,
515 skb->_sk_redir = (unsigned long)sk_redir;
517 skb->_sk_redir |= BPF_F_INGRESS;
520 static inline struct sock *skb_bpf_redirect_fetch(const struct sk_buff *skb)
522 unsigned long sk_redir = skb->_sk_redir;
524 return (struct sock *)(sk_redir & BPF_F_PTR_MASK);
527 static inline void skb_bpf_redirect_clear(struct sk_buff *skb)
531 #endif /* CONFIG_NET_SOCK_MSG */
532 #endif /* _LINUX_SKMSG_H */