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_close)(struct sock *sk, long timeout);
99 void (*saved_write_space)(struct sock *sk);
100 void (*saved_data_ready)(struct sock *sk);
101 int (*psock_update_sk_prot)(struct sock *sk, struct sk_psock *psock,
103 struct proto *sk_proto;
104 struct mutex work_mutex;
105 struct sk_psock_work_state work_state;
106 struct work_struct work;
107 struct rcu_work rwork;
110 int sk_msg_alloc(struct sock *sk, struct sk_msg *msg, int len,
111 int elem_first_coalesce);
112 int sk_msg_clone(struct sock *sk, struct sk_msg *dst, struct sk_msg *src,
114 void sk_msg_trim(struct sock *sk, struct sk_msg *msg, int len);
115 int sk_msg_free(struct sock *sk, struct sk_msg *msg);
116 int sk_msg_free_nocharge(struct sock *sk, struct sk_msg *msg);
117 void sk_msg_free_partial(struct sock *sk, struct sk_msg *msg, u32 bytes);
118 void sk_msg_free_partial_nocharge(struct sock *sk, struct sk_msg *msg,
121 void sk_msg_return(struct sock *sk, struct sk_msg *msg, int bytes);
122 void sk_msg_return_zero(struct sock *sk, struct sk_msg *msg, int bytes);
124 int sk_msg_zerocopy_from_iter(struct sock *sk, struct iov_iter *from,
125 struct sk_msg *msg, u32 bytes);
126 int sk_msg_memcopy_from_iter(struct sock *sk, struct iov_iter *from,
127 struct sk_msg *msg, u32 bytes);
128 int sk_msg_recvmsg(struct sock *sk, struct sk_psock *psock, struct msghdr *msg,
130 bool sk_msg_is_readable(struct sock *sk);
132 static inline void sk_msg_check_to_free(struct sk_msg *msg, u32 i, u32 bytes)
134 WARN_ON(i == msg->sg.end && bytes);
137 static inline void sk_msg_apply_bytes(struct sk_psock *psock, u32 bytes)
139 if (psock->apply_bytes) {
140 if (psock->apply_bytes < bytes)
141 psock->apply_bytes = 0;
143 psock->apply_bytes -= bytes;
147 static inline u32 sk_msg_iter_dist(u32 start, u32 end)
149 return end >= start ? end - start : end + (NR_MSG_FRAG_IDS - start);
152 #define sk_msg_iter_var_prev(var) \
155 var = NR_MSG_FRAG_IDS - 1; \
160 #define sk_msg_iter_var_next(var) \
163 if (var == NR_MSG_FRAG_IDS) \
167 #define sk_msg_iter_prev(msg, which) \
168 sk_msg_iter_var_prev(msg->sg.which)
170 #define sk_msg_iter_next(msg, which) \
171 sk_msg_iter_var_next(msg->sg.which)
173 static inline void sk_msg_init(struct sk_msg *msg)
175 BUILD_BUG_ON(ARRAY_SIZE(msg->sg.data) - 1 != NR_MSG_FRAG_IDS);
176 memset(msg, 0, sizeof(*msg));
177 sg_init_marker(msg->sg.data, NR_MSG_FRAG_IDS);
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.size -= size;
187 src->sg.data[which].length -= size;
188 src->sg.data[which].offset += size;
191 static inline void sk_msg_xfer_full(struct sk_msg *dst, struct sk_msg *src)
193 memcpy(dst, src, sizeof(*src));
197 static inline bool sk_msg_full(const struct sk_msg *msg)
199 return sk_msg_iter_dist(msg->sg.start, msg->sg.end) == MAX_MSG_FRAGS;
202 static inline u32 sk_msg_elem_used(const struct sk_msg *msg)
204 return sk_msg_iter_dist(msg->sg.start, msg->sg.end);
207 static inline struct scatterlist *sk_msg_elem(struct sk_msg *msg, int which)
209 return &msg->sg.data[which];
212 static inline struct scatterlist sk_msg_elem_cpy(struct sk_msg *msg, int which)
214 return msg->sg.data[which];
217 static inline struct page *sk_msg_page(struct sk_msg *msg, int which)
219 return sg_page(sk_msg_elem(msg, which));
222 static inline bool sk_msg_to_ingress(const struct sk_msg *msg)
224 return msg->flags & BPF_F_INGRESS;
227 static inline void sk_msg_compute_data_pointers(struct sk_msg *msg)
229 struct scatterlist *sge = sk_msg_elem(msg, msg->sg.start);
231 if (test_bit(msg->sg.start, msg->sg.copy)) {
233 msg->data_end = NULL;
235 msg->data = sg_virt(sge);
236 msg->data_end = msg->data + sge->length;
240 static inline void sk_msg_page_add(struct sk_msg *msg, struct page *page,
243 struct scatterlist *sge;
246 sge = sk_msg_elem(msg, msg->sg.end);
247 sg_set_page(sge, page, len, offset);
250 __set_bit(msg->sg.end, msg->sg.copy);
252 sk_msg_iter_next(msg, end);
255 static inline void sk_msg_sg_copy(struct sk_msg *msg, u32 i, bool copy_state)
259 __set_bit(i, msg->sg.copy);
261 __clear_bit(i, msg->sg.copy);
262 sk_msg_iter_var_next(i);
263 if (i == msg->sg.end)
268 static inline void sk_msg_sg_copy_set(struct sk_msg *msg, u32 start)
270 sk_msg_sg_copy(msg, start, true);
273 static inline void sk_msg_sg_copy_clear(struct sk_msg *msg, u32 start)
275 sk_msg_sg_copy(msg, start, false);
278 static inline struct sk_psock *sk_psock(const struct sock *sk)
280 return rcu_dereference_sk_user_data(sk);
283 static inline void sk_psock_set_state(struct sk_psock *psock,
284 enum sk_psock_state_bits bit)
286 set_bit(bit, &psock->state);
289 static inline void sk_psock_clear_state(struct sk_psock *psock,
290 enum sk_psock_state_bits bit)
292 clear_bit(bit, &psock->state);
295 static inline bool sk_psock_test_state(const struct sk_psock *psock,
296 enum sk_psock_state_bits bit)
298 return test_bit(bit, &psock->state);
301 static inline void sock_drop(struct sock *sk, struct sk_buff *skb)
303 sk_drops_add(sk, skb);
307 static inline void sk_psock_queue_msg(struct sk_psock *psock,
310 spin_lock_bh(&psock->ingress_lock);
311 if (sk_psock_test_state(psock, SK_PSOCK_TX_ENABLED))
312 list_add_tail(&msg->list, &psock->ingress_msg);
314 sk_msg_free(psock->sk, msg);
317 spin_unlock_bh(&psock->ingress_lock);
320 static inline struct sk_msg *sk_psock_dequeue_msg(struct sk_psock *psock)
324 spin_lock_bh(&psock->ingress_lock);
325 msg = list_first_entry_or_null(&psock->ingress_msg, struct sk_msg, list);
327 list_del(&msg->list);
328 spin_unlock_bh(&psock->ingress_lock);
332 static inline struct sk_msg *sk_psock_peek_msg(struct sk_psock *psock)
336 spin_lock_bh(&psock->ingress_lock);
337 msg = list_first_entry_or_null(&psock->ingress_msg, struct sk_msg, list);
338 spin_unlock_bh(&psock->ingress_lock);
342 static inline struct sk_msg *sk_psock_next_msg(struct sk_psock *psock,
347 spin_lock_bh(&psock->ingress_lock);
348 if (list_is_last(&msg->list, &psock->ingress_msg))
351 ret = list_next_entry(msg, list);
352 spin_unlock_bh(&psock->ingress_lock);
356 static inline bool sk_psock_queue_empty(const struct sk_psock *psock)
358 return psock ? list_empty(&psock->ingress_msg) : true;
361 static inline void kfree_sk_msg(struct sk_msg *msg)
364 consume_skb(msg->skb);
368 static inline void sk_psock_report_error(struct sk_psock *psock, int err)
370 struct sock *sk = psock->sk;
376 struct sk_psock *sk_psock_init(struct sock *sk, int node);
377 void sk_psock_stop(struct sk_psock *psock, bool wait);
379 #if IS_ENABLED(CONFIG_BPF_STREAM_PARSER)
380 int sk_psock_init_strp(struct sock *sk, struct sk_psock *psock);
381 void sk_psock_start_strp(struct sock *sk, struct sk_psock *psock);
382 void sk_psock_stop_strp(struct sock *sk, struct sk_psock *psock);
384 static inline int sk_psock_init_strp(struct sock *sk, struct sk_psock *psock)
389 static inline void sk_psock_start_strp(struct sock *sk, struct sk_psock *psock)
393 static inline void sk_psock_stop_strp(struct sock *sk, struct sk_psock *psock)
398 void sk_psock_start_verdict(struct sock *sk, struct sk_psock *psock);
399 void sk_psock_stop_verdict(struct sock *sk, struct sk_psock *psock);
401 int sk_psock_msg_verdict(struct sock *sk, struct sk_psock *psock,
404 static inline struct sk_psock_link *sk_psock_init_link(void)
406 return kzalloc(sizeof(struct sk_psock_link),
407 GFP_ATOMIC | __GFP_NOWARN);
410 static inline void sk_psock_free_link(struct sk_psock_link *link)
415 struct sk_psock_link *sk_psock_link_pop(struct sk_psock *psock);
417 static inline void sk_psock_cork_free(struct sk_psock *psock)
420 sk_msg_free(psock->sk, psock->cork);
426 static inline void sk_psock_restore_proto(struct sock *sk,
427 struct sk_psock *psock)
429 if (psock->psock_update_sk_prot)
430 psock->psock_update_sk_prot(sk, psock, true);
433 static inline struct sk_psock *sk_psock_get(struct sock *sk)
435 struct sk_psock *psock;
438 psock = sk_psock(sk);
439 if (psock && !refcount_inc_not_zero(&psock->refcnt))
445 void sk_psock_drop(struct sock *sk, struct sk_psock *psock);
447 static inline void sk_psock_put(struct sock *sk, struct sk_psock *psock)
449 if (refcount_dec_and_test(&psock->refcnt))
450 sk_psock_drop(sk, psock);
453 static inline void sk_psock_data_ready(struct sock *sk, struct sk_psock *psock)
455 if (psock->saved_data_ready)
456 psock->saved_data_ready(sk);
458 sk->sk_data_ready(sk);
461 static inline void psock_set_prog(struct bpf_prog **pprog,
462 struct bpf_prog *prog)
464 prog = xchg(pprog, prog);
469 static inline int psock_replace_prog(struct bpf_prog **pprog,
470 struct bpf_prog *prog,
471 struct bpf_prog *old)
473 if (cmpxchg(pprog, old, prog) != old)
482 static inline void psock_progs_drop(struct sk_psock_progs *progs)
484 psock_set_prog(&progs->msg_parser, NULL);
485 psock_set_prog(&progs->stream_parser, NULL);
486 psock_set_prog(&progs->stream_verdict, NULL);
487 psock_set_prog(&progs->skb_verdict, NULL);
490 int sk_psock_tls_strp_read(struct sk_psock *psock, struct sk_buff *skb);
492 static inline bool sk_psock_strp_enabled(struct sk_psock *psock)
496 return !!psock->saved_data_ready;
499 static inline bool sk_is_udp(const struct sock *sk)
501 return sk->sk_type == SOCK_DGRAM &&
502 sk->sk_protocol == IPPROTO_UDP;
505 #if IS_ENABLED(CONFIG_NET_SOCK_MSG)
507 #define BPF_F_STRPARSER (1UL << 1)
509 /* We only have two bits so far. */
510 #define BPF_F_PTR_MASK ~(BPF_F_INGRESS | BPF_F_STRPARSER)
512 static inline bool skb_bpf_strparser(const struct sk_buff *skb)
514 unsigned long sk_redir = skb->_sk_redir;
516 return sk_redir & BPF_F_STRPARSER;
519 static inline void skb_bpf_set_strparser(struct sk_buff *skb)
521 skb->_sk_redir |= BPF_F_STRPARSER;
524 static inline bool skb_bpf_ingress(const struct sk_buff *skb)
526 unsigned long sk_redir = skb->_sk_redir;
528 return sk_redir & BPF_F_INGRESS;
531 static inline void skb_bpf_set_ingress(struct sk_buff *skb)
533 skb->_sk_redir |= BPF_F_INGRESS;
536 static inline void skb_bpf_set_redir(struct sk_buff *skb, struct sock *sk_redir,
539 skb->_sk_redir = (unsigned long)sk_redir;
541 skb->_sk_redir |= BPF_F_INGRESS;
544 static inline struct sock *skb_bpf_redirect_fetch(const struct sk_buff *skb)
546 unsigned long sk_redir = skb->_sk_redir;
548 return (struct sock *)(sk_redir & BPF_F_PTR_MASK);
551 static inline void skb_bpf_redirect_clear(struct sk_buff *skb)
555 #endif /* CONFIG_NET_SOCK_MSG */
556 #endif /* _LINUX_SKMSG_H */