Linux kernel mirror (for testing)
git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git
tjh.dev
kernel
os
linux
1/* SPDX-License-Identifier: GPL-2.0 */
2#ifndef _BPF_CGROUP_H
3#define _BPF_CGROUP_H
4
5#include <linux/bpf.h>
6#include <linux/errno.h>
7#include <linux/jump_label.h>
8#include <linux/percpu.h>
9#include <linux/percpu-refcount.h>
10#include <linux/rbtree.h>
11#include <uapi/linux/bpf.h>
12
13struct sock;
14struct sockaddr;
15struct cgroup;
16struct sk_buff;
17struct bpf_map;
18struct bpf_prog;
19struct bpf_sock_ops_kern;
20struct bpf_cgroup_storage;
21struct ctl_table;
22struct ctl_table_header;
23
24#ifdef CONFIG_CGROUP_BPF
25
26extern struct static_key_false cgroup_bpf_enabled_key;
27#define cgroup_bpf_enabled static_branch_unlikely(&cgroup_bpf_enabled_key)
28
29DECLARE_PER_CPU(struct bpf_cgroup_storage*,
30 bpf_cgroup_storage[MAX_BPF_CGROUP_STORAGE_TYPE]);
31
32#define for_each_cgroup_storage_type(stype) \
33 for (stype = 0; stype < MAX_BPF_CGROUP_STORAGE_TYPE; stype++)
34
35struct bpf_cgroup_storage_map;
36
37struct bpf_storage_buffer {
38 struct rcu_head rcu;
39 char data[];
40};
41
42struct bpf_cgroup_storage {
43 union {
44 struct bpf_storage_buffer *buf;
45 void __percpu *percpu_buf;
46 };
47 struct bpf_cgroup_storage_map *map;
48 struct bpf_cgroup_storage_key key;
49 struct list_head list_map;
50 struct list_head list_cg;
51 struct rb_node node;
52 struct rcu_head rcu;
53};
54
55struct bpf_cgroup_link {
56 struct bpf_link link;
57 struct cgroup *cgroup;
58 enum bpf_attach_type type;
59};
60
61struct bpf_prog_list {
62 struct list_head node;
63 struct bpf_prog *prog;
64 struct bpf_cgroup_link *link;
65 struct bpf_cgroup_storage *storage[MAX_BPF_CGROUP_STORAGE_TYPE];
66};
67
68struct bpf_prog_array;
69
70struct cgroup_bpf {
71 /* array of effective progs in this cgroup */
72 struct bpf_prog_array __rcu *effective[MAX_BPF_ATTACH_TYPE];
73
74 /* attached progs to this cgroup and attach flags
75 * when flags == 0 or BPF_F_ALLOW_OVERRIDE the progs list will
76 * have either zero or one element
77 * when BPF_F_ALLOW_MULTI the list can have up to BPF_CGROUP_MAX_PROGS
78 */
79 struct list_head progs[MAX_BPF_ATTACH_TYPE];
80 u32 flags[MAX_BPF_ATTACH_TYPE];
81
82 /* list of cgroup shared storages */
83 struct list_head storages;
84
85 /* temp storage for effective prog array used by prog_attach/detach */
86 struct bpf_prog_array *inactive;
87
88 /* reference counter used to detach bpf programs after cgroup removal */
89 struct percpu_ref refcnt;
90
91 /* cgroup_bpf is released using a work queue */
92 struct work_struct release_work;
93};
94
95int cgroup_bpf_inherit(struct cgroup *cgrp);
96void cgroup_bpf_offline(struct cgroup *cgrp);
97
98int __cgroup_bpf_attach(struct cgroup *cgrp,
99 struct bpf_prog *prog, struct bpf_prog *replace_prog,
100 struct bpf_cgroup_link *link,
101 enum bpf_attach_type type, u32 flags);
102int __cgroup_bpf_detach(struct cgroup *cgrp, struct bpf_prog *prog,
103 struct bpf_cgroup_link *link,
104 enum bpf_attach_type type);
105int __cgroup_bpf_query(struct cgroup *cgrp, const union bpf_attr *attr,
106 union bpf_attr __user *uattr);
107
108/* Wrapper for __cgroup_bpf_*() protected by cgroup_mutex */
109int cgroup_bpf_attach(struct cgroup *cgrp,
110 struct bpf_prog *prog, struct bpf_prog *replace_prog,
111 struct bpf_cgroup_link *link, enum bpf_attach_type type,
112 u32 flags);
113int cgroup_bpf_detach(struct cgroup *cgrp, struct bpf_prog *prog,
114 enum bpf_attach_type type);
115int cgroup_bpf_query(struct cgroup *cgrp, const union bpf_attr *attr,
116 union bpf_attr __user *uattr);
117
118int __cgroup_bpf_run_filter_skb(struct sock *sk,
119 struct sk_buff *skb,
120 enum bpf_attach_type type);
121
122int __cgroup_bpf_run_filter_sk(struct sock *sk,
123 enum bpf_attach_type type);
124
125int __cgroup_bpf_run_filter_sock_addr(struct sock *sk,
126 struct sockaddr *uaddr,
127 enum bpf_attach_type type,
128 void *t_ctx);
129
130int __cgroup_bpf_run_filter_sock_ops(struct sock *sk,
131 struct bpf_sock_ops_kern *sock_ops,
132 enum bpf_attach_type type);
133
134int __cgroup_bpf_check_dev_permission(short dev_type, u32 major, u32 minor,
135 short access, enum bpf_attach_type type);
136
137int __cgroup_bpf_run_filter_sysctl(struct ctl_table_header *head,
138 struct ctl_table *table, int write,
139 char **buf, size_t *pcount, loff_t *ppos,
140 enum bpf_attach_type type);
141
142int __cgroup_bpf_run_filter_setsockopt(struct sock *sock, int *level,
143 int *optname, char __user *optval,
144 int *optlen, char **kernel_optval);
145int __cgroup_bpf_run_filter_getsockopt(struct sock *sk, int level,
146 int optname, char __user *optval,
147 int __user *optlen, int max_optlen,
148 int retval);
149
150static inline enum bpf_cgroup_storage_type cgroup_storage_type(
151 struct bpf_map *map)
152{
153 if (map->map_type == BPF_MAP_TYPE_PERCPU_CGROUP_STORAGE)
154 return BPF_CGROUP_STORAGE_PERCPU;
155
156 return BPF_CGROUP_STORAGE_SHARED;
157}
158
159static inline void bpf_cgroup_storage_set(struct bpf_cgroup_storage
160 *storage[MAX_BPF_CGROUP_STORAGE_TYPE])
161{
162 enum bpf_cgroup_storage_type stype;
163
164 for_each_cgroup_storage_type(stype)
165 this_cpu_write(bpf_cgroup_storage[stype], storage[stype]);
166}
167
168struct bpf_cgroup_storage *
169cgroup_storage_lookup(struct bpf_cgroup_storage_map *map,
170 void *key, bool locked);
171struct bpf_cgroup_storage *bpf_cgroup_storage_alloc(struct bpf_prog *prog,
172 enum bpf_cgroup_storage_type stype);
173void bpf_cgroup_storage_free(struct bpf_cgroup_storage *storage);
174void bpf_cgroup_storage_link(struct bpf_cgroup_storage *storage,
175 struct cgroup *cgroup,
176 enum bpf_attach_type type);
177void bpf_cgroup_storage_unlink(struct bpf_cgroup_storage *storage);
178int bpf_cgroup_storage_assign(struct bpf_prog_aux *aux, struct bpf_map *map);
179
180int bpf_percpu_cgroup_storage_copy(struct bpf_map *map, void *key, void *value);
181int bpf_percpu_cgroup_storage_update(struct bpf_map *map, void *key,
182 void *value, u64 flags);
183
184/* Wrappers for __cgroup_bpf_run_filter_skb() guarded by cgroup_bpf_enabled. */
185#define BPF_CGROUP_RUN_PROG_INET_INGRESS(sk, skb) \
186({ \
187 int __ret = 0; \
188 if (cgroup_bpf_enabled) \
189 __ret = __cgroup_bpf_run_filter_skb(sk, skb, \
190 BPF_CGROUP_INET_INGRESS); \
191 \
192 __ret; \
193})
194
195#define BPF_CGROUP_RUN_PROG_INET_EGRESS(sk, skb) \
196({ \
197 int __ret = 0; \
198 if (cgroup_bpf_enabled && sk && sk == skb->sk) { \
199 typeof(sk) __sk = sk_to_full_sk(sk); \
200 if (sk_fullsock(__sk)) \
201 __ret = __cgroup_bpf_run_filter_skb(__sk, skb, \
202 BPF_CGROUP_INET_EGRESS); \
203 } \
204 __ret; \
205})
206
207#define BPF_CGROUP_RUN_SK_PROG(sk, type) \
208({ \
209 int __ret = 0; \
210 if (cgroup_bpf_enabled) { \
211 __ret = __cgroup_bpf_run_filter_sk(sk, type); \
212 } \
213 __ret; \
214})
215
216#define BPF_CGROUP_RUN_PROG_INET_SOCK(sk) \
217 BPF_CGROUP_RUN_SK_PROG(sk, BPF_CGROUP_INET_SOCK_CREATE)
218
219#define BPF_CGROUP_RUN_PROG_INET_SOCK_RELEASE(sk) \
220 BPF_CGROUP_RUN_SK_PROG(sk, BPF_CGROUP_INET_SOCK_RELEASE)
221
222#define BPF_CGROUP_RUN_PROG_INET4_POST_BIND(sk) \
223 BPF_CGROUP_RUN_SK_PROG(sk, BPF_CGROUP_INET4_POST_BIND)
224
225#define BPF_CGROUP_RUN_PROG_INET6_POST_BIND(sk) \
226 BPF_CGROUP_RUN_SK_PROG(sk, BPF_CGROUP_INET6_POST_BIND)
227
228#define BPF_CGROUP_RUN_SA_PROG(sk, uaddr, type) \
229({ \
230 int __ret = 0; \
231 if (cgroup_bpf_enabled) \
232 __ret = __cgroup_bpf_run_filter_sock_addr(sk, uaddr, type, \
233 NULL); \
234 __ret; \
235})
236
237#define BPF_CGROUP_RUN_SA_PROG_LOCK(sk, uaddr, type, t_ctx) \
238({ \
239 int __ret = 0; \
240 if (cgroup_bpf_enabled) { \
241 lock_sock(sk); \
242 __ret = __cgroup_bpf_run_filter_sock_addr(sk, uaddr, type, \
243 t_ctx); \
244 release_sock(sk); \
245 } \
246 __ret; \
247})
248
249#define BPF_CGROUP_RUN_PROG_INET4_BIND_LOCK(sk, uaddr) \
250 BPF_CGROUP_RUN_SA_PROG_LOCK(sk, uaddr, BPF_CGROUP_INET4_BIND, NULL)
251
252#define BPF_CGROUP_RUN_PROG_INET6_BIND_LOCK(sk, uaddr) \
253 BPF_CGROUP_RUN_SA_PROG_LOCK(sk, uaddr, BPF_CGROUP_INET6_BIND, NULL)
254
255#define BPF_CGROUP_PRE_CONNECT_ENABLED(sk) (cgroup_bpf_enabled && \
256 sk->sk_prot->pre_connect)
257
258#define BPF_CGROUP_RUN_PROG_INET4_CONNECT(sk, uaddr) \
259 BPF_CGROUP_RUN_SA_PROG(sk, uaddr, BPF_CGROUP_INET4_CONNECT)
260
261#define BPF_CGROUP_RUN_PROG_INET6_CONNECT(sk, uaddr) \
262 BPF_CGROUP_RUN_SA_PROG(sk, uaddr, BPF_CGROUP_INET6_CONNECT)
263
264#define BPF_CGROUP_RUN_PROG_INET4_CONNECT_LOCK(sk, uaddr) \
265 BPF_CGROUP_RUN_SA_PROG_LOCK(sk, uaddr, BPF_CGROUP_INET4_CONNECT, NULL)
266
267#define BPF_CGROUP_RUN_PROG_INET6_CONNECT_LOCK(sk, uaddr) \
268 BPF_CGROUP_RUN_SA_PROG_LOCK(sk, uaddr, BPF_CGROUP_INET6_CONNECT, NULL)
269
270#define BPF_CGROUP_RUN_PROG_UDP4_SENDMSG_LOCK(sk, uaddr, t_ctx) \
271 BPF_CGROUP_RUN_SA_PROG_LOCK(sk, uaddr, BPF_CGROUP_UDP4_SENDMSG, t_ctx)
272
273#define BPF_CGROUP_RUN_PROG_UDP6_SENDMSG_LOCK(sk, uaddr, t_ctx) \
274 BPF_CGROUP_RUN_SA_PROG_LOCK(sk, uaddr, BPF_CGROUP_UDP6_SENDMSG, t_ctx)
275
276#define BPF_CGROUP_RUN_PROG_UDP4_RECVMSG_LOCK(sk, uaddr) \
277 BPF_CGROUP_RUN_SA_PROG_LOCK(sk, uaddr, BPF_CGROUP_UDP4_RECVMSG, NULL)
278
279#define BPF_CGROUP_RUN_PROG_UDP6_RECVMSG_LOCK(sk, uaddr) \
280 BPF_CGROUP_RUN_SA_PROG_LOCK(sk, uaddr, BPF_CGROUP_UDP6_RECVMSG, NULL)
281
282/* The SOCK_OPS"_SK" macro should be used when sock_ops->sk is not a
283 * fullsock and its parent fullsock cannot be traced by
284 * sk_to_full_sk().
285 *
286 * e.g. sock_ops->sk is a request_sock and it is under syncookie mode.
287 * Its listener-sk is not attached to the rsk_listener.
288 * In this case, the caller holds the listener-sk (unlocked),
289 * set its sock_ops->sk to req_sk, and call this SOCK_OPS"_SK" with
290 * the listener-sk such that the cgroup-bpf-progs of the
291 * listener-sk will be run.
292 *
293 * Regardless of syncookie mode or not,
294 * calling bpf_setsockopt on listener-sk will not make sense anyway,
295 * so passing 'sock_ops->sk == req_sk' to the bpf prog is appropriate here.
296 */
297#define BPF_CGROUP_RUN_PROG_SOCK_OPS_SK(sock_ops, sk) \
298({ \
299 int __ret = 0; \
300 if (cgroup_bpf_enabled) \
301 __ret = __cgroup_bpf_run_filter_sock_ops(sk, \
302 sock_ops, \
303 BPF_CGROUP_SOCK_OPS); \
304 __ret; \
305})
306
307#define BPF_CGROUP_RUN_PROG_SOCK_OPS(sock_ops) \
308({ \
309 int __ret = 0; \
310 if (cgroup_bpf_enabled && (sock_ops)->sk) { \
311 typeof(sk) __sk = sk_to_full_sk((sock_ops)->sk); \
312 if (__sk && sk_fullsock(__sk)) \
313 __ret = __cgroup_bpf_run_filter_sock_ops(__sk, \
314 sock_ops, \
315 BPF_CGROUP_SOCK_OPS); \
316 } \
317 __ret; \
318})
319
320#define BPF_CGROUP_RUN_PROG_DEVICE_CGROUP(type, major, minor, access) \
321({ \
322 int __ret = 0; \
323 if (cgroup_bpf_enabled) \
324 __ret = __cgroup_bpf_check_dev_permission(type, major, minor, \
325 access, \
326 BPF_CGROUP_DEVICE); \
327 \
328 __ret; \
329})
330
331
332#define BPF_CGROUP_RUN_PROG_SYSCTL(head, table, write, buf, count, pos) \
333({ \
334 int __ret = 0; \
335 if (cgroup_bpf_enabled) \
336 __ret = __cgroup_bpf_run_filter_sysctl(head, table, write, \
337 buf, count, pos, \
338 BPF_CGROUP_SYSCTL); \
339 __ret; \
340})
341
342#define BPF_CGROUP_RUN_PROG_SETSOCKOPT(sock, level, optname, optval, optlen, \
343 kernel_optval) \
344({ \
345 int __ret = 0; \
346 if (cgroup_bpf_enabled) \
347 __ret = __cgroup_bpf_run_filter_setsockopt(sock, level, \
348 optname, optval, \
349 optlen, \
350 kernel_optval); \
351 __ret; \
352})
353
354#define BPF_CGROUP_GETSOCKOPT_MAX_OPTLEN(optlen) \
355({ \
356 int __ret = 0; \
357 if (cgroup_bpf_enabled) \
358 get_user(__ret, optlen); \
359 __ret; \
360})
361
362#define BPF_CGROUP_RUN_PROG_GETSOCKOPT(sock, level, optname, optval, optlen, \
363 max_optlen, retval) \
364({ \
365 int __ret = retval; \
366 if (cgroup_bpf_enabled) \
367 __ret = __cgroup_bpf_run_filter_getsockopt(sock, level, \
368 optname, optval, \
369 optlen, max_optlen, \
370 retval); \
371 __ret; \
372})
373
374int cgroup_bpf_prog_attach(const union bpf_attr *attr,
375 enum bpf_prog_type ptype, struct bpf_prog *prog);
376int cgroup_bpf_prog_detach(const union bpf_attr *attr,
377 enum bpf_prog_type ptype);
378int cgroup_bpf_link_attach(const union bpf_attr *attr, struct bpf_prog *prog);
379int cgroup_bpf_prog_query(const union bpf_attr *attr,
380 union bpf_attr __user *uattr);
381#else
382
383struct bpf_prog;
384struct cgroup_bpf {};
385static inline int cgroup_bpf_inherit(struct cgroup *cgrp) { return 0; }
386static inline void cgroup_bpf_offline(struct cgroup *cgrp) {}
387
388static inline int cgroup_bpf_prog_attach(const union bpf_attr *attr,
389 enum bpf_prog_type ptype,
390 struct bpf_prog *prog)
391{
392 return -EINVAL;
393}
394
395static inline int cgroup_bpf_prog_detach(const union bpf_attr *attr,
396 enum bpf_prog_type ptype)
397{
398 return -EINVAL;
399}
400
401static inline int cgroup_bpf_link_attach(const union bpf_attr *attr,
402 struct bpf_prog *prog)
403{
404 return -EINVAL;
405}
406
407static inline int cgroup_bpf_prog_query(const union bpf_attr *attr,
408 union bpf_attr __user *uattr)
409{
410 return -EINVAL;
411}
412
413static inline void bpf_cgroup_storage_set(
414 struct bpf_cgroup_storage *storage[MAX_BPF_CGROUP_STORAGE_TYPE]) {}
415static inline int bpf_cgroup_storage_assign(struct bpf_prog_aux *aux,
416 struct bpf_map *map) { return 0; }
417static inline struct bpf_cgroup_storage *bpf_cgroup_storage_alloc(
418 struct bpf_prog *prog, enum bpf_cgroup_storage_type stype) { return NULL; }
419static inline void bpf_cgroup_storage_free(
420 struct bpf_cgroup_storage *storage) {}
421static inline int bpf_percpu_cgroup_storage_copy(struct bpf_map *map, void *key,
422 void *value) {
423 return 0;
424}
425static inline int bpf_percpu_cgroup_storage_update(struct bpf_map *map,
426 void *key, void *value, u64 flags) {
427 return 0;
428}
429
430#define cgroup_bpf_enabled (0)
431#define BPF_CGROUP_RUN_SA_PROG_LOCK(sk, uaddr, type, t_ctx) ({ 0; })
432#define BPF_CGROUP_PRE_CONNECT_ENABLED(sk) (0)
433#define BPF_CGROUP_RUN_PROG_INET_INGRESS(sk,skb) ({ 0; })
434#define BPF_CGROUP_RUN_PROG_INET_EGRESS(sk,skb) ({ 0; })
435#define BPF_CGROUP_RUN_PROG_INET_SOCK(sk) ({ 0; })
436#define BPF_CGROUP_RUN_PROG_INET_SOCK_RELEASE(sk) ({ 0; })
437#define BPF_CGROUP_RUN_PROG_INET4_BIND_LOCK(sk, uaddr) ({ 0; })
438#define BPF_CGROUP_RUN_PROG_INET6_BIND_LOCK(sk, uaddr) ({ 0; })
439#define BPF_CGROUP_RUN_PROG_INET4_POST_BIND(sk) ({ 0; })
440#define BPF_CGROUP_RUN_PROG_INET6_POST_BIND(sk) ({ 0; })
441#define BPF_CGROUP_RUN_PROG_INET4_CONNECT(sk, uaddr) ({ 0; })
442#define BPF_CGROUP_RUN_PROG_INET4_CONNECT_LOCK(sk, uaddr) ({ 0; })
443#define BPF_CGROUP_RUN_PROG_INET6_CONNECT(sk, uaddr) ({ 0; })
444#define BPF_CGROUP_RUN_PROG_INET6_CONNECT_LOCK(sk, uaddr) ({ 0; })
445#define BPF_CGROUP_RUN_PROG_UDP4_SENDMSG_LOCK(sk, uaddr, t_ctx) ({ 0; })
446#define BPF_CGROUP_RUN_PROG_UDP6_SENDMSG_LOCK(sk, uaddr, t_ctx) ({ 0; })
447#define BPF_CGROUP_RUN_PROG_UDP4_RECVMSG_LOCK(sk, uaddr) ({ 0; })
448#define BPF_CGROUP_RUN_PROG_UDP6_RECVMSG_LOCK(sk, uaddr) ({ 0; })
449#define BPF_CGROUP_RUN_PROG_SOCK_OPS(sock_ops) ({ 0; })
450#define BPF_CGROUP_RUN_PROG_DEVICE_CGROUP(type,major,minor,access) ({ 0; })
451#define BPF_CGROUP_RUN_PROG_SYSCTL(head,table,write,buf,count,pos) ({ 0; })
452#define BPF_CGROUP_GETSOCKOPT_MAX_OPTLEN(optlen) ({ 0; })
453#define BPF_CGROUP_RUN_PROG_GETSOCKOPT(sock, level, optname, optval, \
454 optlen, max_optlen, retval) ({ retval; })
455#define BPF_CGROUP_RUN_PROG_SETSOCKOPT(sock, level, optname, optval, optlen, \
456 kernel_optval) ({ 0; })
457
458#define for_each_cgroup_storage_type(stype) for (; false; )
459
460#endif /* CONFIG_CGROUP_BPF */
461
462#endif /* _BPF_CGROUP_H */