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];
47 struct sock
*sk_redir
;
49 struct list_head list
;
52 struct sk_psock_progs
{
53 struct bpf_prog
*msg_parser
;
54 struct bpf_prog
*skb_parser
;
55 struct bpf_prog
*skb_verdict
;
58 enum sk_psock_state_bits
{
62 struct sk_psock_link
{
63 struct list_head list
;
68 struct sk_psock_parser
{
69 struct strparser strp
;
71 void (*saved_data_ready
)(struct sock
*sk
);
74 struct sk_psock_work_state
{
82 struct sock
*sk_redir
;
87 struct sk_psock_progs progs
;
88 struct sk_psock_parser parser
;
89 struct sk_buff_head ingress_skb
;
90 struct list_head ingress_msg
;
92 struct list_head link
;
95 void (*saved_unhash
)(struct sock
*sk
);
96 void (*saved_close
)(struct sock
*sk
, long timeout
);
97 void (*saved_write_space
)(struct sock
*sk
);
98 struct proto
*sk_proto
;
99 struct sk_psock_work_state work_state
;
100 struct work_struct work
;
103 struct work_struct gc
;
107 int sk_msg_alloc(struct sock
*sk
, struct sk_msg
*msg
, int len
,
108 int elem_first_coalesce
);
109 int sk_msg_clone(struct sock
*sk
, struct sk_msg
*dst
, struct sk_msg
*src
,
111 void sk_msg_trim(struct sock
*sk
, struct sk_msg
*msg
, int len
);
112 int sk_msg_free(struct sock
*sk
, struct sk_msg
*msg
);
113 int sk_msg_free_nocharge(struct sock
*sk
, struct sk_msg
*msg
);
114 void sk_msg_free_partial(struct sock
*sk
, struct sk_msg
*msg
, u32 bytes
);
115 void sk_msg_free_partial_nocharge(struct sock
*sk
, struct sk_msg
*msg
,
118 void sk_msg_return(struct sock
*sk
, struct sk_msg
*msg
, int bytes
);
119 void sk_msg_return_zero(struct sock
*sk
, struct sk_msg
*msg
, int bytes
);
121 int sk_msg_zerocopy_from_iter(struct sock
*sk
, struct iov_iter
*from
,
122 struct sk_msg
*msg
, u32 bytes
);
123 int sk_msg_memcopy_from_iter(struct sock
*sk
, struct iov_iter
*from
,
124 struct sk_msg
*msg
, u32 bytes
);
126 static inline void sk_msg_check_to_free(struct sk_msg
*msg
, u32 i
, u32 bytes
)
128 WARN_ON(i
== msg
->sg
.end
&& bytes
);
131 static inline void sk_msg_apply_bytes(struct sk_psock
*psock
, u32 bytes
)
133 if (psock
->apply_bytes
) {
134 if (psock
->apply_bytes
< bytes
)
135 psock
->apply_bytes
= 0;
137 psock
->apply_bytes
-= bytes
;
141 #define sk_msg_iter_var_prev(var) \
144 var = MAX_MSG_FRAGS - 1; \
149 #define sk_msg_iter_var_next(var) \
152 if (var == MAX_MSG_FRAGS) \
156 #define sk_msg_iter_prev(msg, which) \
157 sk_msg_iter_var_prev(msg->sg.which)
159 #define sk_msg_iter_next(msg, which) \
160 sk_msg_iter_var_next(msg->sg.which)
162 static inline void sk_msg_clear_meta(struct sk_msg
*msg
)
164 memset(&msg
->sg
, 0, offsetofend(struct sk_msg_sg
, copy
));
167 static inline void sk_msg_init(struct sk_msg
*msg
)
169 BUILD_BUG_ON(ARRAY_SIZE(msg
->sg
.data
) - 1 != MAX_MSG_FRAGS
);
170 memset(msg
, 0, sizeof(*msg
));
171 sg_init_marker(msg
->sg
.data
, MAX_MSG_FRAGS
);
174 static inline void sk_msg_xfer(struct sk_msg
*dst
, struct sk_msg
*src
,
177 dst
->sg
.data
[which
] = src
->sg
.data
[which
];
178 dst
->sg
.data
[which
].length
= size
;
179 dst
->sg
.size
+= size
;
180 src
->sg
.data
[which
].length
-= size
;
181 src
->sg
.data
[which
].offset
+= size
;
184 static inline void sk_msg_xfer_full(struct sk_msg
*dst
, struct sk_msg
*src
)
186 memcpy(dst
, src
, sizeof(*src
));
190 static inline bool sk_msg_full(const struct sk_msg
*msg
)
192 return (msg
->sg
.end
== msg
->sg
.start
) && msg
->sg
.size
;
195 static inline u32
sk_msg_elem_used(const struct sk_msg
*msg
)
197 if (sk_msg_full(msg
))
198 return MAX_MSG_FRAGS
;
200 return msg
->sg
.end
>= msg
->sg
.start
?
201 msg
->sg
.end
- msg
->sg
.start
:
202 msg
->sg
.end
+ (MAX_MSG_FRAGS
- msg
->sg
.start
);
205 static inline struct scatterlist
*sk_msg_elem(struct sk_msg
*msg
, int which
)
207 return &msg
->sg
.data
[which
];
210 static inline struct scatterlist
sk_msg_elem_cpy(struct sk_msg
*msg
, int which
)
212 return msg
->sg
.data
[which
];
215 static inline struct page
*sk_msg_page(struct sk_msg
*msg
, int which
)
217 return sg_page(sk_msg_elem(msg
, which
));
220 static inline bool sk_msg_to_ingress(const struct sk_msg
*msg
)
222 return msg
->flags
& BPF_F_INGRESS
;
225 static inline void sk_msg_compute_data_pointers(struct sk_msg
*msg
)
227 struct scatterlist
*sge
= sk_msg_elem(msg
, msg
->sg
.start
);
229 if (msg
->sg
.copy
[msg
->sg
.start
]) {
231 msg
->data_end
= NULL
;
233 msg
->data
= sg_virt(sge
);
234 msg
->data_end
= msg
->data
+ sge
->length
;
238 static inline void sk_msg_page_add(struct sk_msg
*msg
, struct page
*page
,
241 struct scatterlist
*sge
;
244 sge
= sk_msg_elem(msg
, msg
->sg
.end
);
245 sg_set_page(sge
, page
, len
, offset
);
248 msg
->sg
.copy
[msg
->sg
.end
] = true;
250 sk_msg_iter_next(msg
, end
);
253 static inline void sk_msg_sg_copy(struct sk_msg
*msg
, u32 i
, bool copy_state
)
256 msg
->sg
.copy
[i
] = copy_state
;
257 sk_msg_iter_var_next(i
);
258 if (i
== msg
->sg
.end
)
263 static inline void sk_msg_sg_copy_set(struct sk_msg
*msg
, u32 start
)
265 sk_msg_sg_copy(msg
, start
, true);
268 static inline void sk_msg_sg_copy_clear(struct sk_msg
*msg
, u32 start
)
270 sk_msg_sg_copy(msg
, start
, false);
273 static inline struct sk_psock
*sk_psock(const struct sock
*sk
)
275 return rcu_dereference_sk_user_data(sk
);
278 static inline void sk_psock_queue_msg(struct sk_psock
*psock
,
281 list_add_tail(&msg
->list
, &psock
->ingress_msg
);
284 static inline bool sk_psock_queue_empty(const struct sk_psock
*psock
)
286 return psock
? list_empty(&psock
->ingress_msg
) : true;
289 static inline void sk_psock_report_error(struct sk_psock
*psock
, int err
)
291 struct sock
*sk
= psock
->sk
;
294 sk
->sk_error_report(sk
);
297 struct sk_psock
*sk_psock_init(struct sock
*sk
, int node
);
299 int sk_psock_init_strp(struct sock
*sk
, struct sk_psock
*psock
);
300 void sk_psock_start_strp(struct sock
*sk
, struct sk_psock
*psock
);
301 void sk_psock_stop_strp(struct sock
*sk
, struct sk_psock
*psock
);
303 int sk_psock_msg_verdict(struct sock
*sk
, struct sk_psock
*psock
,
306 static inline struct sk_psock_link
*sk_psock_init_link(void)
308 return kzalloc(sizeof(struct sk_psock_link
),
309 GFP_ATOMIC
| __GFP_NOWARN
);
312 static inline void sk_psock_free_link(struct sk_psock_link
*link
)
317 struct sk_psock_link
*sk_psock_link_pop(struct sk_psock
*psock
);
318 #if defined(CONFIG_BPF_STREAM_PARSER)
319 void sk_psock_unlink(struct sock
*sk
, struct sk_psock_link
*link
);
321 static inline void sk_psock_unlink(struct sock
*sk
,
322 struct sk_psock_link
*link
)
327 void __sk_psock_purge_ingress_msg(struct sk_psock
*psock
);
329 static inline void sk_psock_cork_free(struct sk_psock
*psock
)
332 sk_msg_free(psock
->sk
, psock
->cork
);
338 static inline void sk_psock_update_proto(struct sock
*sk
,
339 struct sk_psock
*psock
,
342 psock
->saved_unhash
= sk
->sk_prot
->unhash
;
343 psock
->saved_close
= sk
->sk_prot
->close
;
344 psock
->saved_write_space
= sk
->sk_write_space
;
346 psock
->sk_proto
= sk
->sk_prot
;
350 static inline void sk_psock_restore_proto(struct sock
*sk
,
351 struct sk_psock
*psock
)
353 if (psock
->sk_proto
) {
354 sk
->sk_prot
= psock
->sk_proto
;
355 psock
->sk_proto
= NULL
;
359 static inline void sk_psock_set_state(struct sk_psock
*psock
,
360 enum sk_psock_state_bits bit
)
362 set_bit(bit
, &psock
->state
);
365 static inline void sk_psock_clear_state(struct sk_psock
*psock
,
366 enum sk_psock_state_bits bit
)
368 clear_bit(bit
, &psock
->state
);
371 static inline bool sk_psock_test_state(const struct sk_psock
*psock
,
372 enum sk_psock_state_bits bit
)
374 return test_bit(bit
, &psock
->state
);
377 static inline struct sk_psock
*sk_psock_get_checked(struct sock
*sk
)
379 struct sk_psock
*psock
;
382 psock
= sk_psock(sk
);
384 if (sk
->sk_prot
->recvmsg
!= tcp_bpf_recvmsg
) {
385 psock
= ERR_PTR(-EBUSY
);
389 if (!refcount_inc_not_zero(&psock
->refcnt
))
390 psock
= ERR_PTR(-EBUSY
);
397 static inline struct sk_psock
*sk_psock_get(struct sock
*sk
)
399 struct sk_psock
*psock
;
402 psock
= sk_psock(sk
);
403 if (psock
&& !refcount_inc_not_zero(&psock
->refcnt
))
409 void sk_psock_stop(struct sock
*sk
, struct sk_psock
*psock
);
410 void sk_psock_destroy(struct rcu_head
*rcu
);
411 void sk_psock_drop(struct sock
*sk
, struct sk_psock
*psock
);
413 static inline void sk_psock_put(struct sock
*sk
, struct sk_psock
*psock
)
415 if (refcount_dec_and_test(&psock
->refcnt
))
416 sk_psock_drop(sk
, psock
);
419 static inline void psock_set_prog(struct bpf_prog
**pprog
,
420 struct bpf_prog
*prog
)
422 prog
= xchg(pprog
, prog
);
427 static inline void psock_progs_drop(struct sk_psock_progs
*progs
)
429 psock_set_prog(&progs
->msg_parser
, NULL
);
430 psock_set_prog(&progs
->skb_parser
, NULL
);
431 psock_set_prog(&progs
->skb_verdict
, NULL
);
434 #endif /* _LINUX_SKMSG_H */