From c63829182c37c2d6d0608976d15fa61ebebe9e6b Mon Sep 17 00:00:00 2001 From: Cong Wang Date: Sun, 4 Jul 2021 12:02:47 -0700 Subject: af_unix: Implement ->psock_update_sk_prot() Now we can implement unix_bpf_update_proto() to update sk_prot, especially prot->close(). Signed-off-by: Cong Wang Signed-off-by: Alexei Starovoitov Link: https://lore.kernel.org/bpf/20210704190252.11866-7-xiyou.wangcong@gmail.com --- net/unix/unix_bpf.c | 47 +++++++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 47 insertions(+) create mode 100644 net/unix/unix_bpf.c (limited to 'net/unix/unix_bpf.c') diff --git a/net/unix/unix_bpf.c b/net/unix/unix_bpf.c new file mode 100644 index 000000000000..b1582a659427 --- /dev/null +++ b/net/unix/unix_bpf.c @@ -0,0 +1,47 @@ +// SPDX-License-Identifier: GPL-2.0 +/* Copyright (c) 2021 Cong Wang */ + +#include +#include +#include +#include + +static struct proto *unix_prot_saved __read_mostly; +static DEFINE_SPINLOCK(unix_prot_lock); +static struct proto unix_bpf_prot; + +static void unix_bpf_rebuild_protos(struct proto *prot, const struct proto *base) +{ + *prot = *base; + prot->close = sock_map_close; +} + +static void unix_bpf_check_needs_rebuild(struct proto *ops) +{ + if (unlikely(ops != smp_load_acquire(&unix_prot_saved))) { + spin_lock_bh(&unix_prot_lock); + if (likely(ops != unix_prot_saved)) { + unix_bpf_rebuild_protos(&unix_bpf_prot, ops); + smp_store_release(&unix_prot_saved, ops); + } + spin_unlock_bh(&unix_prot_lock); + } +} + +int unix_bpf_update_proto(struct sock *sk, struct sk_psock *psock, bool restore) +{ + if (restore) { + sk->sk_write_space = psock->saved_write_space; + WRITE_ONCE(sk->sk_prot, psock->sk_proto); + return 0; + } + + unix_bpf_check_needs_rebuild(psock->sk_proto); + WRITE_ONCE(sk->sk_prot, &unix_bpf_prot); + return 0; +} + +void __init unix_bpf_build_proto(void) +{ + unix_bpf_rebuild_protos(&unix_bpf_prot, &unix_proto); +} -- cgit From 9825d866ce0d11009513e06824885340062c166b Mon Sep 17 00:00:00 2001 From: Cong Wang Date: Sun, 4 Jul 2021 12:02:48 -0700 Subject: af_unix: Implement unix_dgram_bpf_recvmsg() We have to implement unix_dgram_bpf_recvmsg() to replace the original ->recvmsg() to retrieve skmsg from ingress_msg. AF_UNIX is again special here because the lack of sk_prot->recvmsg(). I simply add a special case inside unix_dgram_recvmsg() to call sk->sk_prot->recvmsg() directly. Signed-off-by: Cong Wang Signed-off-by: Alexei Starovoitov Link: https://lore.kernel.org/bpf/20210704190252.11866-8-xiyou.wangcong@gmail.com --- net/unix/unix_bpf.c | 75 +++++++++++++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 75 insertions(+) (limited to 'net/unix/unix_bpf.c') diff --git a/net/unix/unix_bpf.c b/net/unix/unix_bpf.c index b1582a659427..db0cda29fb2f 100644 --- a/net/unix/unix_bpf.c +++ b/net/unix/unix_bpf.c @@ -6,6 +6,80 @@ #include #include +#define unix_sk_has_data(__sk, __psock) \ + ({ !skb_queue_empty(&__sk->sk_receive_queue) || \ + !skb_queue_empty(&__psock->ingress_skb) || \ + !list_empty(&__psock->ingress_msg); \ + }) + +static int unix_msg_wait_data(struct sock *sk, struct sk_psock *psock, + long timeo) +{ + DEFINE_WAIT_FUNC(wait, woken_wake_function); + struct unix_sock *u = unix_sk(sk); + int ret = 0; + + if (sk->sk_shutdown & RCV_SHUTDOWN) + return 1; + + if (!timeo) + return ret; + + add_wait_queue(sk_sleep(sk), &wait); + sk_set_bit(SOCKWQ_ASYNC_WAITDATA, sk); + if (!unix_sk_has_data(sk, psock)) { + mutex_unlock(&u->iolock); + wait_woken(&wait, TASK_INTERRUPTIBLE, timeo); + mutex_lock(&u->iolock); + ret = unix_sk_has_data(sk, psock); + } + sk_clear_bit(SOCKWQ_ASYNC_WAITDATA, sk); + remove_wait_queue(sk_sleep(sk), &wait); + return ret; +} + +static int unix_dgram_bpf_recvmsg(struct sock *sk, struct msghdr *msg, + size_t len, int nonblock, int flags, + int *addr_len) +{ + struct unix_sock *u = unix_sk(sk); + struct sk_psock *psock; + int copied, ret; + + psock = sk_psock_get(sk); + if (unlikely(!psock)) + return __unix_dgram_recvmsg(sk, msg, len, flags); + + mutex_lock(&u->iolock); + if (!skb_queue_empty(&sk->sk_receive_queue) && + sk_psock_queue_empty(psock)) { + ret = __unix_dgram_recvmsg(sk, msg, len, flags); + goto out; + } + +msg_bytes_ready: + copied = sk_msg_recvmsg(sk, psock, msg, len, flags); + if (!copied) { + long timeo; + int data; + + timeo = sock_rcvtimeo(sk, nonblock); + data = unix_msg_wait_data(sk, psock, timeo); + if (data) { + if (!sk_psock_queue_empty(psock)) + goto msg_bytes_ready; + ret = __unix_dgram_recvmsg(sk, msg, len, flags); + goto out; + } + copied = -EAGAIN; + } + ret = copied; +out: + mutex_unlock(&u->iolock); + sk_psock_put(sk, psock); + return ret; +} + static struct proto *unix_prot_saved __read_mostly; static DEFINE_SPINLOCK(unix_prot_lock); static struct proto unix_bpf_prot; @@ -14,6 +88,7 @@ static void unix_bpf_rebuild_protos(struct proto *prot, const struct proto *base { *prot = *base; prot->close = sock_map_close; + prot->recvmsg = unix_dgram_bpf_recvmsg; } static void unix_bpf_check_needs_rebuild(struct proto *ops) -- cgit From 0b846445985895e75958ecd59061fd7bf77e0c3f Mon Sep 17 00:00:00 2001 From: Cong Wang Date: Fri, 23 Jul 2021 11:36:30 -0700 Subject: unix_bpf: Fix a potential deadlock in unix_dgram_bpf_recvmsg() As Eric noticed, __unix_dgram_recvmsg() may acquire u->iolock too, so we have to release it before calling this function. Fixes: 9825d866ce0d ("af_unix: Implement unix_dgram_bpf_recvmsg()") Reported-by: Eric Dumazet Signed-off-by: Cong Wang Signed-off-by: Andrii Nakryiko Acked-by: Jakub Sitnicki Acked-by: John Fastabend --- net/unix/unix_bpf.c | 16 ++++++++-------- 1 file changed, 8 insertions(+), 8 deletions(-) (limited to 'net/unix/unix_bpf.c') diff --git a/net/unix/unix_bpf.c b/net/unix/unix_bpf.c index db0cda29fb2f..177e883f451e 100644 --- a/net/unix/unix_bpf.c +++ b/net/unix/unix_bpf.c @@ -44,7 +44,7 @@ static int unix_dgram_bpf_recvmsg(struct sock *sk, struct msghdr *msg, { struct unix_sock *u = unix_sk(sk); struct sk_psock *psock; - int copied, ret; + int copied; psock = sk_psock_get(sk); if (unlikely(!psock)) @@ -53,8 +53,9 @@ static int unix_dgram_bpf_recvmsg(struct sock *sk, struct msghdr *msg, mutex_lock(&u->iolock); if (!skb_queue_empty(&sk->sk_receive_queue) && sk_psock_queue_empty(psock)) { - ret = __unix_dgram_recvmsg(sk, msg, len, flags); - goto out; + mutex_unlock(&u->iolock); + sk_psock_put(sk, psock); + return __unix_dgram_recvmsg(sk, msg, len, flags); } msg_bytes_ready: @@ -68,16 +69,15 @@ msg_bytes_ready: if (data) { if (!sk_psock_queue_empty(psock)) goto msg_bytes_ready; - ret = __unix_dgram_recvmsg(sk, msg, len, flags); - goto out; + mutex_unlock(&u->iolock); + sk_psock_put(sk, psock); + return __unix_dgram_recvmsg(sk, msg, len, flags); } copied = -EAGAIN; } - ret = copied; -out: mutex_unlock(&u->iolock); sk_psock_put(sk, psock); - return ret; + return copied; } static struct proto *unix_prot_saved __read_mostly; -- cgit From 83f31535565c63ac4f62c7b8592210929a630d3d Mon Sep 17 00:00:00 2001 From: Cong Wang Date: Sat, 31 Jul 2021 12:50:38 -0700 Subject: bpf, unix: Check socket type in unix_bpf_update_proto() As of now, only AF_UNIX datagram socket supports sockmap. But unix_proto is shared for all kinds of AF_UNIX sockets, so we have to check the socket type in unix_bpf_update_proto() to explicitly reject other types, otherwise they could be added into sockmap, too. Fixes: c63829182c37 ("af_unix: Implement ->psock_update_sk_prot()") Reported-by: Jakub Sitnicki Signed-off-by: Cong Wang Signed-off-by: Daniel Borkmann Acked-by: Jakub Sitnicki Link: https://lore.kernel.org/bpf/20210731195038.8084-1-xiyou.wangcong@gmail.com --- net/unix/unix_bpf.c | 3 +++ 1 file changed, 3 insertions(+) (limited to 'net/unix/unix_bpf.c') diff --git a/net/unix/unix_bpf.c b/net/unix/unix_bpf.c index 177e883f451e..20f53575b5c9 100644 --- a/net/unix/unix_bpf.c +++ b/net/unix/unix_bpf.c @@ -105,6 +105,9 @@ static void unix_bpf_check_needs_rebuild(struct proto *ops) int unix_bpf_update_proto(struct sock *sk, struct sk_psock *psock, bool restore) { + if (sk->sk_type != SOCK_DGRAM) + return -EOPNOTSUPP; + if (restore) { sk->sk_write_space = psock->saved_write_space; WRITE_ONCE(sk->sk_prot, psock->sk_proto); -- cgit From 94531cfcbe79c3598acf96806627b2137ca32eb9 Mon Sep 17 00:00:00 2001 From: Jiang Wang Date: Mon, 16 Aug 2021 19:03:21 +0000 Subject: af_unix: Add unix_stream_proto for sockmap Previously, sockmap for AF_UNIX protocol only supports dgram type. This patch add unix stream type support, which is similar to unix_dgram_proto. To support sockmap, dgram and stream cannot share the same unix_proto anymore, because they have different implementations, such as unhash for stream type (which will remove closed or disconnected sockets from the map), so rename unix_proto to unix_dgram_proto and add a new unix_stream_proto. Also implement stream related sockmap functions. And add dgram key words to those dgram specific functions. Signed-off-by: Jiang Wang Signed-off-by: Andrii Nakryiko Reviewed-by: Cong Wang Acked-by: Jakub Sitnicki Acked-by: John Fastabend Link: https://lore.kernel.org/bpf/20210816190327.2739291-3-jiang.wang@bytedance.com --- net/unix/unix_bpf.c | 93 ++++++++++++++++++++++++++++++++++++++++------------- 1 file changed, 71 insertions(+), 22 deletions(-) (limited to 'net/unix/unix_bpf.c') diff --git a/net/unix/unix_bpf.c b/net/unix/unix_bpf.c index 20f53575b5c9..b927e2baae50 100644 --- a/net/unix/unix_bpf.c +++ b/net/unix/unix_bpf.c @@ -38,9 +38,18 @@ static int unix_msg_wait_data(struct sock *sk, struct sk_psock *psock, return ret; } -static int unix_dgram_bpf_recvmsg(struct sock *sk, struct msghdr *msg, - size_t len, int nonblock, int flags, - int *addr_len) +static int __unix_recvmsg(struct sock *sk, struct msghdr *msg, + size_t len, int flags) +{ + if (sk->sk_type == SOCK_DGRAM) + return __unix_dgram_recvmsg(sk, msg, len, flags); + else + return __unix_stream_recvmsg(sk, msg, len, flags); +} + +static int unix_bpf_recvmsg(struct sock *sk, struct msghdr *msg, + size_t len, int nonblock, int flags, + int *addr_len) { struct unix_sock *u = unix_sk(sk); struct sk_psock *psock; @@ -48,14 +57,14 @@ static int unix_dgram_bpf_recvmsg(struct sock *sk, struct msghdr *msg, psock = sk_psock_get(sk); if (unlikely(!psock)) - return __unix_dgram_recvmsg(sk, msg, len, flags); + return __unix_recvmsg(sk, msg, len, flags); mutex_lock(&u->iolock); if (!skb_queue_empty(&sk->sk_receive_queue) && sk_psock_queue_empty(psock)) { mutex_unlock(&u->iolock); sk_psock_put(sk, psock); - return __unix_dgram_recvmsg(sk, msg, len, flags); + return __unix_recvmsg(sk, msg, len, flags); } msg_bytes_ready: @@ -71,7 +80,7 @@ msg_bytes_ready: goto msg_bytes_ready; mutex_unlock(&u->iolock); sk_psock_put(sk, psock); - return __unix_dgram_recvmsg(sk, msg, len, flags); + return __unix_recvmsg(sk, msg, len, flags); } copied = -EAGAIN; } @@ -80,30 +89,55 @@ msg_bytes_ready: return copied; } -static struct proto *unix_prot_saved __read_mostly; -static DEFINE_SPINLOCK(unix_prot_lock); -static struct proto unix_bpf_prot; +static struct proto *unix_dgram_prot_saved __read_mostly; +static DEFINE_SPINLOCK(unix_dgram_prot_lock); +static struct proto unix_dgram_bpf_prot; + +static struct proto *unix_stream_prot_saved __read_mostly; +static DEFINE_SPINLOCK(unix_stream_prot_lock); +static struct proto unix_stream_bpf_prot; -static void unix_bpf_rebuild_protos(struct proto *prot, const struct proto *base) +static void unix_dgram_bpf_rebuild_protos(struct proto *prot, const struct proto *base) { *prot = *base; prot->close = sock_map_close; - prot->recvmsg = unix_dgram_bpf_recvmsg; + prot->recvmsg = unix_bpf_recvmsg; +} + +static void unix_stream_bpf_rebuild_protos(struct proto *prot, + const struct proto *base) +{ + *prot = *base; + prot->close = sock_map_close; + prot->recvmsg = unix_bpf_recvmsg; + prot->unhash = sock_map_unhash; +} + +static void unix_dgram_bpf_check_needs_rebuild(struct proto *ops) +{ + if (unlikely(ops != smp_load_acquire(&unix_dgram_prot_saved))) { + spin_lock_bh(&unix_dgram_prot_lock); + if (likely(ops != unix_dgram_prot_saved)) { + unix_dgram_bpf_rebuild_protos(&unix_dgram_bpf_prot, ops); + smp_store_release(&unix_dgram_prot_saved, ops); + } + spin_unlock_bh(&unix_dgram_prot_lock); + } } -static void unix_bpf_check_needs_rebuild(struct proto *ops) +static void unix_stream_bpf_check_needs_rebuild(struct proto *ops) { - if (unlikely(ops != smp_load_acquire(&unix_prot_saved))) { - spin_lock_bh(&unix_prot_lock); - if (likely(ops != unix_prot_saved)) { - unix_bpf_rebuild_protos(&unix_bpf_prot, ops); - smp_store_release(&unix_prot_saved, ops); + if (unlikely(ops != smp_load_acquire(&unix_stream_prot_saved))) { + spin_lock_bh(&unix_stream_prot_lock); + if (likely(ops != unix_stream_prot_saved)) { + unix_stream_bpf_rebuild_protos(&unix_stream_bpf_prot, ops); + smp_store_release(&unix_stream_prot_saved, ops); } - spin_unlock_bh(&unix_prot_lock); + spin_unlock_bh(&unix_stream_prot_lock); } } -int unix_bpf_update_proto(struct sock *sk, struct sk_psock *psock, bool restore) +int unix_dgram_bpf_update_proto(struct sock *sk, struct sk_psock *psock, bool restore) { if (sk->sk_type != SOCK_DGRAM) return -EOPNOTSUPP; @@ -114,12 +148,27 @@ int unix_bpf_update_proto(struct sock *sk, struct sk_psock *psock, bool restore) return 0; } - unix_bpf_check_needs_rebuild(psock->sk_proto); - WRITE_ONCE(sk->sk_prot, &unix_bpf_prot); + unix_dgram_bpf_check_needs_rebuild(psock->sk_proto); + WRITE_ONCE(sk->sk_prot, &unix_dgram_bpf_prot); + return 0; +} + +int unix_stream_bpf_update_proto(struct sock *sk, struct sk_psock *psock, bool restore) +{ + if (restore) { + sk->sk_write_space = psock->saved_write_space; + WRITE_ONCE(sk->sk_prot, psock->sk_proto); + return 0; + } + + unix_stream_bpf_check_needs_rebuild(psock->sk_proto); + WRITE_ONCE(sk->sk_prot, &unix_stream_bpf_prot); return 0; } void __init unix_bpf_build_proto(void) { - unix_bpf_rebuild_protos(&unix_bpf_prot, &unix_proto); + unix_dgram_bpf_rebuild_protos(&unix_dgram_bpf_prot, &unix_dgram_proto); + unix_stream_bpf_rebuild_protos(&unix_stream_bpf_prot, &unix_stream_proto); + } -- cgit From af493388950b6ea3a86f860cfaffab137e024fc8 Mon Sep 17 00:00:00 2001 From: Cong Wang Date: Fri, 8 Oct 2021 13:33:05 -0700 Subject: net: Implement ->sock_is_readable() for UDP and AF_UNIX Yucong noticed we can't poll() sockets in sockmap even when they are the destination sockets of redirections. This is because we never poll any psock queues in ->poll(), except for TCP. With ->sock_is_readable() now we can overwrite >sock_is_readable(), invoke and implement it for both UDP and AF_UNIX sockets. Reported-by: Yucong Sun Signed-off-by: Cong Wang Signed-off-by: Alexei Starovoitov Link: https://lore.kernel.org/bpf/20211008203306.37525-4-xiyou.wangcong@gmail.com --- net/unix/unix_bpf.c | 2 ++ 1 file changed, 2 insertions(+) (limited to 'net/unix/unix_bpf.c') diff --git a/net/unix/unix_bpf.c b/net/unix/unix_bpf.c index b927e2baae50..452376c6f419 100644 --- a/net/unix/unix_bpf.c +++ b/net/unix/unix_bpf.c @@ -102,6 +102,7 @@ static void unix_dgram_bpf_rebuild_protos(struct proto *prot, const struct proto *prot = *base; prot->close = sock_map_close; prot->recvmsg = unix_bpf_recvmsg; + prot->sock_is_readable = sk_msg_is_readable; } static void unix_stream_bpf_rebuild_protos(struct proto *prot, @@ -110,6 +111,7 @@ static void unix_stream_bpf_rebuild_protos(struct proto *prot, *prot = *base; prot->close = sock_map_close; prot->recvmsg = unix_bpf_recvmsg; + prot->sock_is_readable = sk_msg_is_readable; prot->unhash = sock_map_unhash; } -- cgit