bpf: support BPF_PROG_QUERY for progs attached to sockmap
Quoted 234 lines
-----邮件原件-----
发件人: Yonghong Song [mailto:[email protected]]
发送时间: 2021年11月4日 14:31
收件人: zhudi (E) <[email protected]>; [email protected]; [email protected]; [email protected]; [email protected]; [email protected]; [email protected]; [email protected]; [email protected]; [email protected]
抄送: [email protected]; [email protected]
主题: Re: [PATCH bpf-next v5 1/2] bpf: supportBPF_PROG_QUERYfor progs attached to sockmapQuoted 37 lines · zhudi (E)
Quoted 37 lines · Di Zhu
Quoted 20 lines
Right now there is no way to query whether BPF programs are attached to a sockmap or not.
we can use the standard interface in libbpf to query, such as:
bpf_prog_query(mapFd, BPF_SK_SKB_STREAM_PARSER, 0, NULL, ...); the mapFd is the fd of sockmap.Signed-off-by Di Zhu <[email protected]>---
/* v2 */
- John Fastabend <[email protected]>
- add selftest code/* v3 */
- avoid sleeping caused bycopy_to_user()in rcu critical zone/* v4 */
- Alexei Starovoitov <[email protected]> -split into two patches, one for core code and one for selftest.Quoted 7 lines
/* v5 */
- Yonghong Song <[email protected]>
-Some naming and formatting changes--- include/linux/bpf.h | 9 ++++++ kernel/bpf/syscall.c | 5 +++ net/core/sock_map.c | 74+++++++++++++++++++++++++++++++++++++++-----
include/linux/bpf.h int bpf_prog_test_run_syscall(struct bpf_prog
Quoted 7 lines
3 files changed, 81 insertions(+), 7 deletions(-)include/linux/bpf.h +0
-0@@ -1961,6 +1961,9 @@ int bpf_prog_test_run_syscall(struct bpf_prog*prog,
Quoted 3 lines
Quoted 3 lines
include/linux/bpf.h int bpf_prog_test_run_syscall(struct bpf_prog · line 1961
Quoted 1 line
include/linux/bpf.h +0
-0int sock_map_get_from_fd(const union bpf_attr *attr, struct bpf_prog*prog);
Quoted 1 line
include/linux/bpf.h +0
-0int sock_map_prog_detach(const union bpf_attr *attr, enum
bpf_prog_typeQuoted 10 lines
Quoted 10 lines
ptype);
include/linux/bpf.h int bpf_prog_test_run_syscall(struct bpf_prog · line 1963
Quoted 1 line
include/linux/bpf.h +0
-0int sock_map_update_elem_sys(struct bpf_map *map, void *key, void*value, u64 flags);
include/linux/bpf.h static inline int
Quoted 7 lines
include/linux/bpf.h +3
-0+int sock_map_bpf_prog_query (const union bpf_attr *attr, +union bpf_attr __user *uattr); +void sock_map_unhash (struct sock *sk);void sock_map_close (struct sock *sk,long timeout);#else @@ -2014,6 +2017,12 @@ static inline intsock_map_update_elem_sys(struct
Quoted 96 lines
Quoted 96 lines
bpf_map*map, void *key, voidkernel/bpf/syscall.c static int bpf_prog_query(const union bpf_attr
Quoted 17 lines
include/linux/bpf.h +6
-0{return -EOPNOTSUPP ; } + +static inline int sock_map_bpf_prog_query (const union bpf_attr *attr, +union bpf_attr __user *uattr) +{ +return -EINVAL ; +}#endif /* CONFIG_BPF_SYSCALL */ #endif /* CONFIG_NET && CONFIG_BPF_SYSCALL */ kernel/bpf/syscall.c +0
-0@@ -3275,6 +3275,11 @@ static int bpf_prog_query(const union bpf_attr*attr,
net/core/sock_map.c static struct sk_psock_progs
Quoted 16 lines
kernel/bpf/syscall.c +5
-0case BPF_FLOW_DISSECTOR :case BPF_SK_LOOKUP :return netns_bpf_prog_query (attr, uattr); +case BPF_SK_SKB_STREAM_PARSER : +case BPF_SK_SKB_STREAM_VERDICT : +case BPF_SK_MSG_VERDICT : +case BPF_SK_SKB_VERDICT : +return sock_map_bpf_prog_query (attr, uattr);default :return -EINVAL ; }net/core/sock_map.c +0
-0@@ -1412,38 +1412,50 @@ static struct sk_psock_progs*sock_map_progs(struct
bpf_map*map)net/core/sock_map.c static struct sk_psock_progs · line 1415
Quoted 4 lines
net/core/sock_map.c +0
-1return NULL ; } -static int sock_map_prog_update(struct bpf_map *map, struct bpf_prog*prog,
net/core/sock_map.c static struct sk_psock_progs · line 1415
Quoted 2 lines
net/core/sock_map.c +1
-1- struct bpf_prog *old, u32 which) +static int sock_map_prog_lookup(struct bpf_map *map, struct bpf_prog***pprog,
Quoted 29 lines
net/core/sock_map.c +4
-5+ u32 which) {struct sk_psock_progs *progs =sock_map_progs (map); -struct bpf_prog **pprog;if (!progs)return -EOPNOTSUPP ;switch (which) {case BPF_SK_MSG_VERDICT : - pprog = &progs->msg_parser; + *pprog = &progs->msg_parser;break ;#if IS_ENABLED (CONFIG_BPF_STREAM_PARSER )case BPF_SK_SKB_STREAM_PARSER : - pprog = &progs->stream_parser; + *pprog = &progs->stream_parser;break ;#endif case BPF_SK_SKB_STREAM_VERDICT :if (progs->skb_verdict)return -EBUSY ; - pprog = &progs->stream_verdict; + *pprog = &progs->stream_verdict;break ;case BPF_SK_SKB_VERDICT :if (progs->stream_verdict)return -EBUSY ; - pprog = &progs->skb_verdict;net/core/sock_map.c static struct sk_psock_progs · line 1449
Quoted 10 lines
net/core/sock_map.c +5
-0+ *pprog = &progs->skb_verdict;break ;default :return -EOPNOTSUPP ; } +return 0 ; +} + +static int sock_map_prog_update(struct bpf_map *map, struct bpf_prog*prog,
Quoted 12 lines
net/core/sock_map.c +9
-0+ struct bpf_prog *old, u32 which) +{ +struct bpf_prog **pprog; +int ret; + + ret =sock_map_prog_lookup (map, &pprog, which); +if (ret) +return ret; +if (old)return psock_replace_prog (pprog, prog, old);Quoted 1 line
Quoted 1 line
net/core/sock_map.c static int sock_map_prog_update(struct
Quoted 1 line
net/core/sock_map.c +0
-0@@ -1451,6 +1463,54 @@ static int sock_map_prog_update(struct
bpf_mapQuoted 18 lines
Quoted 4 lines
*map, struct
bpf_prog*prog,Quoted 3 lines
net/core/sock_map.c +0
-0return 0 ; }Quoted 12 lines
net/core/sock_map.c static int sock_map_prog_update(struct · line 1475
Quoted 10 lines
net/core/sock_map.c +10
-0+int sock_map_bpf_prog_query (const union bpf_attr *attr, +union bpf_attr __user *uattr) +{ +__u32 __user *prog_ids =u64_to_user_ptr (attr->query.prog_ids); +u32 prog_cnt =0 , flags =0 , ufd = attr->target_fd; +struct bpf_prog **pprog; +struct bpf_prog *prog; +struct bpf_map *map; +struct fd f; +u32 id =0 ;There is no need to initialize 'id = 0'. id will be assigned later.
Quoted 1 line
if (!attr->query.prog_cnt || !prog_ids || !prog_cnt) is met, the id will not be
assigned later.
Quoted 1 line
At the end of the function, we judge whether to copy the program ID by the
value of the id.
Ya, that is true. id = 0 is indeed needed.
Quoted 40 lines
Quoted 15 lines
Quoted 15 lines
net/core/sock_map.c +12
-0+int ret; + +if (attr->query.query_flags) +return -EINVAL ; + + f =fdget (ufd); + map =__bpf_map_get (f); +if (IS_ERR (map)) +return PTR_ERR (map); + +rcu_read_lock (); +Quoted 8 lines
net/core/sock_map.c static int sock_map_prog_update(struct · line 1493
Quoted 6 lines
net/core/sock_map.c +6
-0+ ret =sock_map_prog_lookup (map, &pprog, attr->query.attach_type); +if (ret) + goto end; + + prog = *pprog; + prog_cnt = (!prog) ?0 :1 ;(!prog) => !prog ?
Yes, It's just my habit
Quoted 12 lines
net/core/sock_map.c static int sock_map_prog_update(struct · line 1500
Quoted 8 lines
net/core/sock_map.c +7
-0+ +if (!attr->query.prog_cnt || !prog_ids || !prog_cnt) + goto end; + + id = prog->aux->id; +if (id ==0 ) + prog_cnt =0 ;id will never be 0, see function
bpf_prog_alloc_id()issyscall.c. So 'if (id == 0)' check is not needed.Quoted 1 line
Because we do not hold the reference count, the program may be released
synchronously
Quoted 8 lines
synchronously => asynchronously
Quoted 3 lines
and its ID will be set to 0 in
bpf_prog_free_id().Is that right?
Just checked the code again. You are right. Maybe add a comment here to make it clear why we check id == 0 here?
Yes, I'll add a comment
Quoted 21 lines
Quoted 10 lines
Quoted 10 lines
net/core/sock_map.c static int sock_map_prog_update(struct · line 1507
Quoted 10 lines
net/core/sock_map.c +7
-0+ +end: +rcu_read_unlock (); + +if (copy_to_user (&uattr->query.attach_flags, &flags,sizeof (flags)) || + (id !=0 &©_to_user (prog_ids, &id,sizeof (u32))) || + copy_to_user(&uattr->query.prog_cnt, &prog_cnt,
sizeof(prog_cnt)))Quoted 10 lines
Quoted 10 lines
Quoted 10 lines
net/core/sock_map.c +6
-0+ ret = -EFAULT ; + +fdput (f); +return ret; +} +static void sock_map_unlink (struct sock *sk,struct sk_psock_link *link) {switch (link->map->map_type) {