From 9cd3e072b0be17446e37d7414eac8a3499e0601e Mon Sep 17 00:00:00 2001 From: Eric Dumazet Date: Sun, 29 Nov 2015 20:03:10 -0800 Subject: net: rename SOCK_ASYNC_NOSPACE and SOCK_ASYNC_WAITDATA This patch is a cleanup to make following patch easier to review. Goal is to move SOCK_ASYNC_NOSPACE and SOCK_ASYNC_WAITDATA from (struct socket)->flags to a (struct socket_wq)->flags to benefit from RCU protection in sock_wake_async() To ease backports, we rename both constants. Two new helpers, sk_set_bit(int nr, struct sock *sk) and sk_clear_bit(int net, struct sock *sk) are added so that following patch can change their implementation. Signed-off-by: Eric Dumazet Signed-off-by: David S. Miller --- net/sctp/socket.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'net/sctp/socket.c') diff --git a/net/sctp/socket.c b/net/sctp/socket.c index 897c01c029ca..2353985d689c 100644 --- a/net/sctp/socket.c +++ b/net/sctp/socket.c @@ -6458,7 +6458,7 @@ unsigned int sctp_poll(struct file *file, struct socket *sock, poll_table *wait) if (sctp_writeable(sk)) { mask |= POLLOUT | POLLWRNORM; } else { - set_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags); + sk_set_bit(SOCKWQ_ASYNC_NOSPACE, sk); /* * Since the socket is not locked, the buffer * might be made available after the writeable check and -- cgit v1.2.3 From ceb5d58b217098a657f3850b7a2640f995032e62 Mon Sep 17 00:00:00 2001 From: Eric Dumazet Date: Sun, 29 Nov 2015 20:03:11 -0800 Subject: net: fix sock_wake_async() rcu protection Dmitry provided a syzkaller (http://github.com/google/syzkaller) triggering a fault in sock_wake_async() when async IO is requested. Said program stressed af_unix sockets, but the issue is generic and should be addressed in core networking stack. The problem is that by the time sock_wake_async() is called, we should not access the @flags field of 'struct socket', as the inode containing this socket might be freed without further notice, and without RCU grace period. We already maintain an RCU protected structure, "struct socket_wq" so moving SOCKWQ_ASYNC_NOSPACE & SOCKWQ_ASYNC_WAITDATA into it is the safe route. It also reduces number of cache lines needing dirtying, so might provide a performance improvement anyway. In followup patches, we might move remaining flags (SOCK_NOSPACE, SOCK_PASSCRED, SOCK_PASSSEC) to save 8 bytes and let 'struct socket' being mostly read and let it being shared between cpus. Reported-by: Dmitry Vyukov Signed-off-by: Eric Dumazet Signed-off-by: David S. Miller --- net/sctp/socket.c | 24 ++++++++++++++---------- 1 file changed, 14 insertions(+), 10 deletions(-) (limited to 'net/sctp/socket.c') diff --git a/net/sctp/socket.c b/net/sctp/socket.c index 2353985d689c..5e35ef34008b 100644 --- a/net/sctp/socket.c +++ b/net/sctp/socket.c @@ -6801,26 +6801,30 @@ no_packet: static void __sctp_write_space(struct sctp_association *asoc) { struct sock *sk = asoc->base.sk; - struct socket *sock = sk->sk_socket; - if ((sctp_wspace(asoc) > 0) && sock) { - if (waitqueue_active(&asoc->wait)) - wake_up_interruptible(&asoc->wait); + if (sctp_wspace(asoc) <= 0) + return; + + if (waitqueue_active(&asoc->wait)) + wake_up_interruptible(&asoc->wait); - if (sctp_writeable(sk)) { - wait_queue_head_t *wq = sk_sleep(sk); + if (sctp_writeable(sk)) { + struct socket_wq *wq; - if (wq && waitqueue_active(wq)) - wake_up_interruptible(wq); + rcu_read_lock(); + wq = rcu_dereference(sk->sk_wq); + if (wq) { + if (waitqueue_active(&wq->wait)) + wake_up_interruptible(&wq->wait); /* Note that we try to include the Async I/O support * here by modeling from the current TCP/UDP code. * We have not tested with it yet. */ if (!(sk->sk_shutdown & SEND_SHUTDOWN)) - sock_wake_async(sock, - SOCK_WAKE_SPACE, POLL_OUT); + sock_wake_async(wq, SOCK_WAKE_SPACE, POLL_OUT); } + rcu_read_unlock(); } } -- cgit v1.2.3 From cacc06215271104b40773c99547c506095db6ad4 Mon Sep 17 00:00:00 2001 From: Marcelo Ricardo Leitner Date: Mon, 30 Nov 2015 14:32:54 -0200 Subject: sctp: use GFP_USER for user-controlled kmalloc Dmitry Vyukov reported that the user could trigger a kernel warning by using a large len value for getsockopt SCTP_GET_LOCAL_ADDRS, as that value directly affects the value used as a kmalloc() parameter. This patch thus switches the allocation flags from all user-controllable kmalloc size to GFP_USER to put some more restrictions on it and also disables the warn, as they are not necessary. Signed-off-by: Marcelo Ricardo Leitner Acked-by: Daniel Borkmann Signed-off-by: David S. Miller --- net/sctp/socket.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'net/sctp/socket.c') diff --git a/net/sctp/socket.c b/net/sctp/socket.c index 5e35ef34008b..f6161e356734 100644 --- a/net/sctp/socket.c +++ b/net/sctp/socket.c @@ -972,7 +972,7 @@ static int sctp_setsockopt_bindx(struct sock *sk, return -EFAULT; /* Alloc space for the address array in kernel memory. */ - kaddrs = kmalloc(addrs_size, GFP_KERNEL); + kaddrs = kmalloc(addrs_size, GFP_USER | __GFP_NOWARN); if (unlikely(!kaddrs)) return -ENOMEM; @@ -4928,7 +4928,7 @@ static int sctp_getsockopt_local_addrs(struct sock *sk, int len, to = optval + offsetof(struct sctp_getaddrs, addrs); space_left = len - offsetof(struct sctp_getaddrs, addrs); - addrs = kmalloc(space_left, GFP_KERNEL); + addrs = kmalloc(space_left, GFP_USER | __GFP_NOWARN); if (!addrs) return -ENOMEM; -- cgit v1.2.3 From 602dd62dfbda3e63a2d6a3cbde953ebe82bf5087 Mon Sep 17 00:00:00 2001 From: Eric Dumazet Date: Tue, 1 Dec 2015 07:20:07 -0800 Subject: ipv6: sctp: implement sctp_v6_destroy_sock() Dmitry Vyukov reported a memory leak using IPV6 SCTP sockets. We need to call inet6_destroy_sock() to properly release inet6 specific fields. Reported-by: Dmitry Vyukov Signed-off-by: Eric Dumazet Acked-by: Daniel Borkmann Signed-off-by: David S. Miller --- net/sctp/socket.c | 9 ++++++++- 1 file changed, 8 insertions(+), 1 deletion(-) (limited to 'net/sctp/socket.c') diff --git a/net/sctp/socket.c b/net/sctp/socket.c index f6161e356734..03c8256063ec 100644 --- a/net/sctp/socket.c +++ b/net/sctp/socket.c @@ -7379,6 +7379,13 @@ struct proto sctp_prot = { #if IS_ENABLED(CONFIG_IPV6) +#include +static void sctp_v6_destroy_sock(struct sock *sk) +{ + sctp_destroy_sock(sk); + inet6_destroy_sock(sk); +} + struct proto sctpv6_prot = { .name = "SCTPv6", .owner = THIS_MODULE, @@ -7388,7 +7395,7 @@ struct proto sctpv6_prot = { .accept = sctp_accept, .ioctl = sctp_ioctl, .init = sctp_init_sock, - .destroy = sctp_destroy_sock, + .destroy = sctp_v6_destroy_sock, .shutdown = sctp_shutdown, .setsockopt = sctp_setsockopt, .getsockopt = sctp_getsockopt, -- cgit v1.2.3 From 01ce63c90170283a9855d1db4fe81934dddce648 Mon Sep 17 00:00:00 2001 From: Marcelo Ricardo Leitner Date: Fri, 4 Dec 2015 15:14:04 -0200 Subject: sctp: update the netstamp_needed counter when copying sockets Dmitry Vyukov reported that SCTP was triggering a WARN on socket destroy related to disabling sock timestamp. When SCTP accepts an association or peel one off, it copies sock flags but forgot to call net_enable_timestamp() if a packet timestamping flag was copied, leading to extra calls to net_disable_timestamp() whenever such clones were closed. The fix is to call net_enable_timestamp() whenever we copy a sock with that flag on, like tcp does. Reported-by: Dmitry Vyukov Signed-off-by: Marcelo Ricardo Leitner Acked-by: Vlad Yasevich Signed-off-by: David S. Miller --- net/sctp/socket.c | 3 +++ 1 file changed, 3 insertions(+) (limited to 'net/sctp/socket.c') diff --git a/net/sctp/socket.c b/net/sctp/socket.c index 03c8256063ec..4c9282bdd067 100644 --- a/net/sctp/socket.c +++ b/net/sctp/socket.c @@ -7199,6 +7199,9 @@ void sctp_copy_sock(struct sock *newsk, struct sock *sk, newinet->mc_ttl = 1; newinet->mc_index = 0; newinet->mc_list = NULL; + + if (newsk->sk_flags & SK_FLAGS_TIMESTAMP) + net_enable_timestamp(); } static inline void sctp_copy_descendant(struct sock *sk_to, -- cgit v1.2.3 From 50a5ffb1ef535e3c6989711c51b5d61b543a3b45 Mon Sep 17 00:00:00 2001 From: Marcelo Ricardo Leitner Date: Fri, 4 Dec 2015 15:14:05 -0200 Subject: sctp: also copy sk_tsflags when copying the socket As we are keeping timestamps on when copying the socket, we also have to copy sk_tsflags. This is needed since b9f40e21ef42 ("net-timestamp: move timestamp flags out of sk_flags"). Signed-off-by: Marcelo Ricardo Leitner Acked-by: Vlad Yasevich Signed-off-by: David S. Miller --- net/sctp/socket.c | 1 + 1 file changed, 1 insertion(+) (limited to 'net/sctp/socket.c') diff --git a/net/sctp/socket.c b/net/sctp/socket.c index 4c9282bdd067..1a32ecdb8bae 100644 --- a/net/sctp/socket.c +++ b/net/sctp/socket.c @@ -7167,6 +7167,7 @@ void sctp_copy_sock(struct sock *newsk, struct sock *sk, newsk->sk_type = sk->sk_type; newsk->sk_bound_dev_if = sk->sk_bound_dev_if; newsk->sk_flags = sk->sk_flags; + newsk->sk_tsflags = sk->sk_tsflags; newsk->sk_no_check_tx = sk->sk_no_check_tx; newsk->sk_no_check_rx = sk->sk_no_check_rx; newsk->sk_reuse = sk->sk_reuse; -- cgit v1.2.3 From 69b5777f2e5779bb987d4a25a33401d5ac257c14 Mon Sep 17 00:00:00 2001 From: lucien Date: Sat, 5 Dec 2015 15:15:17 +0800 Subject: sctp: hold the chunks only after the chunk is enqueued in outq When a msg is sent, sctp will hold the chunks of this msg and then try to enqueue them. But if the chunks are not enqueued in sctp_outq_tail() because of the invalid state, sctp_cmd_interpreter() may still return success to sctp_sendmsg() after calling sctp_outq_flush(), these chunks will become orphans and will leak. So we fix them by moving sctp_chunk_hold() to sctp_outq_tail(), where we are sure that the chunk is going to get queued. Signed-off-by: Xin Long Signed-off-by: Marcelo Ricardo Leitner Signed-off-by: David S. Miller --- net/sctp/socket.c | 2 -- 1 file changed, 2 deletions(-) (limited to 'net/sctp/socket.c') diff --git a/net/sctp/socket.c b/net/sctp/socket.c index 1a32ecdb8bae..bd57300c8e91 100644 --- a/net/sctp/socket.c +++ b/net/sctp/socket.c @@ -1952,8 +1952,6 @@ static int sctp_sendmsg(struct sock *sk, struct msghdr *msg, size_t msg_len) /* Now send the (possibly) fragmented message. */ list_for_each_entry(chunk, &datamsg->chunks, frag_list) { - sctp_chunk_hold(chunk); - /* Do accounting for the write space. */ sctp_set_owner_w(chunk); -- cgit v1.2.3 From 8b570dc9f7b634e853866ce40097c0342ac5bb81 Mon Sep 17 00:00:00 2001 From: lucien Date: Sat, 5 Dec 2015 15:19:27 +0800 Subject: sctp: only drop the reference on the datamsg after sending a msg If the chunks are enqueued successfully but sctp_cmd_interpreter() return err to sctp_sendmsg() (mainly because of no mem), the chunks will get re-queued, but we are dropping the reference and freeing them. The fix is to just drop the reference on the datamsg just as it had succeeded, as: - if the chunks weren't queued, this is enough to get them freed. - if they were queued, they will get freed when they finally get out or discarded. Signed-off-by: Xin Long Marcelo Ricardo Leitner Signed-off-by: Marcelo Ricardo Leitner Signed-off-by: David S. Miller --- net/sctp/socket.c | 6 ++---- 1 file changed, 2 insertions(+), 4 deletions(-) (limited to 'net/sctp/socket.c') diff --git a/net/sctp/socket.c b/net/sctp/socket.c index bd57300c8e91..9b6cc6de80d8 100644 --- a/net/sctp/socket.c +++ b/net/sctp/socket.c @@ -1964,15 +1964,13 @@ static int sctp_sendmsg(struct sock *sk, struct msghdr *msg, size_t msg_len) * breaks. */ err = sctp_primitive_SEND(net, asoc, datamsg); + sctp_datamsg_put(datamsg); /* Did the lower layer accept the chunk? */ - if (err) { - sctp_datamsg_free(datamsg); + if (err) goto out_free; - } pr_debug("%s: we sent primitively\n", __func__); - sctp_datamsg_put(datamsg); err = msg_len; if (unlikely(wait_connect)) { -- cgit v1.2.3