bpf: support BPF_PROG_QUERY for progs attached to sockmap
Quoted 176 lines · Jakub Sitnicki
Quoted 9 lines · Di Zhu
Quoted 7 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.Quoted 6 lines
Quoted 6 lines
Signed-off-by Di Zhu <[email protected]>Acked-by Yonghong Song <[email protected]>--- include/linux/bpf.h | 9 +++++ kernel/bpf/syscall.c | 5 +++ net/core/sock_map.c | 78++++++++++++++++++++++++++++++++++++++++----
Quoted 7 lines
include/linux/bpf.h int bpf_prog_test_run_syscall(struct bpf_prog
Quoted 7 lines
3 files changed, 85 insertions(+), 7 deletions(-)include/linux/bpf.h +0
-0@@ -2071,6 +2071,9 @@ int bpf_prog_test_run_syscall(struct bpf_prog*prog,
Quoted 1 line
include/linux/bpf.h int bpf_prog_test_run_syscall(struct bpf_prog · line 2071
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 int bpf_prog_test_run_syscall(struct bpf_prog · line 2072
Quoted 1 line
include/linux/bpf.h +0
-0int sock_map_prog_detach(const union bpf_attr *attr, enum bpf_prog_typeptype);
Quoted 1 line
include/linux/bpf.h int bpf_prog_test_run_syscall(struct bpf_prog · line 2073
Quoted 1 line
include/linux/bpf.h +0
-0int sock_map_update_elem_sys(struct bpf_map *map, void *key, void*value, u64 flags);
Quoted 7 lines
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 @@ -2124,6 +2127,12 @@ static inline intsock_map_update_elem_sys(struct
bpf_map*map, void *key, voidQuoted 6 lines
include/linux/bpf.h static inline int · line 2132
Quoted 6 lines
include/linux/bpf.h +3
-0{return -EOPNOTSUPP ; } + +static inline int sock_map_bpf_prog_query (const union bpf_attr *attr, + union bpf_attr __user*uattr)
Quoted 11 lines
kernel/bpf/syscall.c static int bpf_prog_query(const union bpf_attr
Quoted 11 lines
include/linux/bpf.h +3
-0+{ +return -EINVAL ; +}#endif /* CONFIG_BPF_SYSCALL */ #endif /* CONFIG_NET && CONFIG_BPF_SYSCALL */ kernel/bpf/syscall.c +0
-0@@ -3318,6 +3318,11 @@ static int bpf_prog_query(const union bpf_attr*attr,
Quoted 8 lines
Quoted 8 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);Quoted 8 lines
net/core/sock_map.c static struct sk_psock_progs
Quoted 8 lines
kernel/bpf/syscall.c +0
-0default :return -EINVAL ; }net/core/sock_map.c +0
-0@@ -1416,38 +1416,50 @@ static struct sk_psock_progs*sock_map_progs(struct
bpf_map*map)Quoted 4 lines
net/core/sock_map.c static struct sk_psock_progs · line 1419
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,
Quoted 2 lines
net/core/sock_map.c static struct sk_psock_progs · line 1419
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 35 lines
Quoted 35 lines
net/core/sock_map.c +5
-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; + *pprog = &progs->skb_verdict;break ;default :return -EOPNOTSUPP ; }Quoted 4 lines
net/core/sock_map.c static struct sk_psock_progs · line 1453
Quoted 4 lines
net/core/sock_map.c +4
-0+return 0 ; +} + +static int sock_map_prog_update(struct bpf_map *map, struct bpf_prog*prog,
Quoted 12 lines
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
net/core/sock_map.c static int sock_map_prog_update(struct bpf_map
Quoted 1 line
net/core/sock_map.c +0
-0@@ -1455,6 +1467,58 @@ static int sock_map_prog_update(struct bpf_map*map, struct
bpf_prog*prog,Quoted 30 lines
Quoted 30 lines
net/core/sock_map.c +27
-0return 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 ; +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 (); + + ret =sock_map_prog_lookup (map, &pprog, attr->query.attach_type); +if (ret) + goto end; + + prog = *pprog;Quoted 8 lines
net/core/sock_map.c static int sock_map_prog_update(struct bpf_map · line 1502
Quoted 6 lines
net/core/sock_map.c +6
-0+ prog_cnt = !prog ?0 :1 ; + +if (!attr->query.prog_cnt || !prog_ids || !prog_cnt) + goto end; + + id = prog->aux->id;^ This looks like a concurrent read/write.
Quoted 4 lines
You mean that
bpf_prog_load()might be setting it in a different thread? I think ID is allocated and fixed before prog FD is available to the user-space, soprog->aux->idis set in stone and immutable in that regard.
What we're talking about here is that bpf_prog_free_id() will write the id identifier synchronously.
Quoted 32 lines
Quoted 6 lines
Would wrap with
READ_ONCE()and correspondingWRITE_ONCE()inbpf_prog_free_id(). See [1] for rationale.[1]
https://github.com/google/kernel-sanitizers/blob/master/other/READ_WRITE_O
NCE.mdQuoted 14 lines
Quoted 4 lines
net/core/sock_map.c +3
-0+ +/* we do not hold the refcnt, the bpf prog may be released +* asynchronously and the id would be set to 0. net/core/sock_map.c static int sock_map_prog_update(struct bpf_map · line 1515
Quoted 10 lines
net/core/sock_map.c +10
-0+*/ +if (id ==0 ) + prog_cnt =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 9 lines
Quoted 9 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) {
Quoted 177 lines · zhudi (E)
Quoted 177 lines · Jakub Sitnicki
Quoted 9 lines · Di Zhu
Quoted 7 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.Quoted 6 lines
Quoted 6 lines
Signed-off-by Di Zhu <[email protected]>Acked-by Yonghong Song <[email protected]>--- include/linux/bpf.h | 9 +++++ kernel/bpf/syscall.c | 5 +++ net/core/sock_map.c | 78++++++++++++++++++++++++++++++++++++++++----
Quoted 7 lines
include/linux/bpf.h int bpf_prog_test_run_syscall(struct bpf_prog
Quoted 7 lines
3 files changed, 85 insertions(+), 7 deletions(-)include/linux/bpf.h +0
-0@@ -2071,6 +2071,9 @@ int bpf_prog_test_run_syscall(struct bpf_prog*prog,
Quoted 1 line
include/linux/bpf.h int bpf_prog_test_run_syscall(struct bpf_prog · line 2071
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 int bpf_prog_test_run_syscall(struct bpf_prog · line 2072
Quoted 1 line
include/linux/bpf.h +0
-0int sock_map_prog_detach(const union bpf_attr *attr, enum bpf_prog_typeptype);
Quoted 1 line
include/linux/bpf.h int bpf_prog_test_run_syscall(struct bpf_prog · line 2073
Quoted 1 line
include/linux/bpf.h +0
-0int sock_map_update_elem_sys(struct bpf_map *map, void *key, void*value, u64 flags);
Quoted 7 lines
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 @@ -2124,6 +2127,12 @@ static inline intsock_map_update_elem_sys(struct
bpf_map*map, void *key, voidQuoted 6 lines
include/linux/bpf.h static inline int · line 2132
Quoted 6 lines
include/linux/bpf.h +3
-0{return -EOPNOTSUPP ; } + +static inline int sock_map_bpf_prog_query (const union bpf_attr *attr, + union bpf_attr __user*uattr)
Quoted 11 lines
kernel/bpf/syscall.c static int bpf_prog_query(const union bpf_attr
Quoted 11 lines
include/linux/bpf.h +3
-0+{ +return -EINVAL ; +}#endif /* CONFIG_BPF_SYSCALL */ #endif /* CONFIG_NET && CONFIG_BPF_SYSCALL */ kernel/bpf/syscall.c +0
-0@@ -3318,6 +3318,11 @@ static int bpf_prog_query(const union bpf_attr*attr,
Quoted 8 lines
Quoted 8 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);Quoted 8 lines
net/core/sock_map.c static struct sk_psock_progs
Quoted 8 lines
kernel/bpf/syscall.c +0
-0default :return -EINVAL ; }net/core/sock_map.c +0
-0@@ -1416,38 +1416,50 @@ static struct sk_psock_progs*sock_map_progs(struct
bpf_map*map)Quoted 4 lines
net/core/sock_map.c static struct sk_psock_progs · line 1419
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,
Quoted 2 lines
net/core/sock_map.c static struct sk_psock_progs · line 1419
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 35 lines
Quoted 35 lines
net/core/sock_map.c +5
-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; + *pprog = &progs->skb_verdict;break ;default :return -EOPNOTSUPP ; }Quoted 4 lines
net/core/sock_map.c static struct sk_psock_progs · line 1453
Quoted 4 lines
net/core/sock_map.c +4
-0+return 0 ; +} + +static int sock_map_prog_update(struct bpf_map *map, struct bpf_prog*prog,
Quoted 12 lines
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
net/core/sock_map.c static int sock_map_prog_update(struct bpf_map
Quoted 1 line
net/core/sock_map.c +0
-0@@ -1455,6 +1467,58 @@ static int sock_map_prog_update(struct bpf_map*map, struct
bpf_prog*prog,Quoted 38 lines
Quoted 26 lines
net/core/sock_map.c +23
-0return 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 ; +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 (); + + ret =sock_map_prog_lookup (map, &pprog, attr->query.attach_type);net/core/sock_map.c static int sock_map_prog_update(struct bpf_map · line 1502
Quoted 10 lines
net/core/sock_map.c +10
-0+if (ret) + goto end; + + prog = *pprog; + prog_cnt = !prog ?0 :1 ; + +if (!attr->query.prog_cnt || !prog_ids || !prog_cnt) + goto end; + + id = prog->aux->id;^ This looks like a concurrent read/write.
Quoted 7 lines
Quoted 4 lines
You mean that
bpf_prog_load()might be setting it in a different thread? I think ID is allocated and fixed before prog FD is available to the user-space, soprog->aux->idis set in stone and immutable in that regard.What we're talking about here is that
bpf_prog_free_id()will write the id identifier synchronously.
Hm.. let's say bpf_prog_free_id() happens right after we read id 123. It's impossible to distinguish that from reading valid ID (that's not yet freed), returning it to user-space and before user-space can do anything about that this program and it's ID are freed. User-space either way will get an ID that's not valid anymore. I don't see any use of READ_ONCE/WRITE_ONCE with prog->aux->id, which is why I was asking what changed.
Quoted 33 lines
Quoted 33 lines
Quoted 7 lines
Would wrap with
READ_ONCE()and correspondingWRITE_ONCE()inbpf_prog_free_id(). See [1] for rationale.[1]
https://github.com/google/kernel-sanitizers/blob/master/other/READ_WRITE_O
NCE.mdQuoted 14 lines
Quoted 4 lines
net/core/sock_map.c +3
-0+ +/* we do not hold the refcnt, the bpf prog may be released +* asynchronously and the id would be set to 0. net/core/sock_map.c static int sock_map_prog_update(struct bpf_map · line 1515
Quoted 10 lines
net/core/sock_map.c +10
-0+*/ +if (id ==0 ) + prog_cnt =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 9 lines
Quoted 9 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) {
Quoted 5 lines · Andrii Nakryiko
Quoted 4 lines · zhudi (E)
Quoted 4 lines · Jakub Sitnicki
Quoted 2 lines · Di Zhu
[...]
Quoted 44 lines
Quoted 36 lines
Quoted 36 lines
Quoted 35 lines
Quoted 23 lines
(quoted diff) +23
-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; + 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(); + + ret = sock_map_prog_lookup(map, &pprog, attr->query.attach_type);Review
Quoted 10 lines
(quoted diff) +10
-0+ if (ret) + goto end; + + prog = *pprog; + prog_cnt = !prog ? 0 : 1; + + if (!attr->query.prog_cnt || !prog_ids || !prog_cnt) + goto end; + + id = prog->aux->id;^ This looks like a concurrent read/write.
Quoted 7 lines
Quoted 4 lines
You mean that
bpf_prog_load()might be setting it in a different thread? I think ID is allocated and fixed before prog FD is available to the user-space, soprog->aux->idis set in stone and immutable in that regard.What we're talking about here is that
bpf_prog_free_id()will write the id identifier synchronously.
Quoted 8 lines
Hm.. let's say
bpf_prog_free_id()happens right after we read id 123. It's impossible to distinguish that from reading valid ID (that's not yet freed), returning it to user-space and before user-space can do anything about that this program and it's ID are freed. User-space either way will get an ID that's not valid anymore. I don't see any use of READ_ONCE/WRITE_ONCE withprog->aux->id, which is why I was asking what changed.
You're right, READ_ONCE/WRITE_ONCE is not improving anything here.
I've suggested it not to make the query op more reliable, but rather to mark the shared access.
But in this case annotating it with data_race() [1] would be a better fit, I think, because we don't care if we get the old or the new value.
Quoted 10 lines
Quoted 10 lines
Quoted 10 lines
Quoted 7 lines
Would wrap with
READ_ONCE()and correspondingWRITE_ONCE()inbpf_prog_free_id(). See [1] for rationale.[1]
https://github.com/google/kernel-sanitizers/blob/master/other/READ_WRITE_O
NCE.mdQuoted 1 line
[...]