From 45f6fad84cc305103b28d73482b344d7f5b76f39 Mon Sep 17 00:00:00 2001 From: Eric Dumazet Date: Sun, 29 Nov 2015 19:37:57 -0800 Subject: ipv6: add complete rcu protection around np->opt This patch addresses multiple problems : UDP/RAW sendmsg() need to get a stable struct ipv6_txoptions while socket is not locked : Other threads can change np->opt concurrently. Dmitry posted a syzkaller (http://github.com/google/syzkaller) program desmonstrating use-after-free. Starting with TCP/DCCP lockless listeners, tcp_v6_syn_recv_sock() and dccp_v6_request_recv_sock() also need to use RCU protection to dereference np->opt once (before calling ipv6_dup_options()) This patch adds full RCU protection to np->opt Reported-by: Dmitry Vyukov Signed-off-by: Eric Dumazet Acked-by: Hannes Frederic Sowa Signed-off-by: David S. Miller --- net/ipv6/tcp_ipv6.c | 28 +++++++++++++++++----------- 1 file changed, 17 insertions(+), 11 deletions(-) (limited to 'net/ipv6/tcp_ipv6.c') diff --git a/net/ipv6/tcp_ipv6.c b/net/ipv6/tcp_ipv6.c index c5429a636f1a..6a50bb4a0dae 100644 --- a/net/ipv6/tcp_ipv6.c +++ b/net/ipv6/tcp_ipv6.c @@ -120,6 +120,7 @@ static int tcp_v6_connect(struct sock *sk, struct sockaddr *uaddr, struct ipv6_pinfo *np = inet6_sk(sk); struct tcp_sock *tp = tcp_sk(sk); struct in6_addr *saddr = NULL, *final_p, final; + struct ipv6_txoptions *opt; struct flowi6 fl6; struct dst_entry *dst; int addr_type; @@ -235,7 +236,8 @@ static int tcp_v6_connect(struct sock *sk, struct sockaddr *uaddr, fl6.fl6_dport = usin->sin6_port; fl6.fl6_sport = inet->inet_sport; - final_p = fl6_update_dst(&fl6, np->opt, &final); + opt = rcu_dereference_protected(np->opt, sock_owned_by_user(sk)); + final_p = fl6_update_dst(&fl6, opt, &final); security_sk_classify_flow(sk, flowi6_to_flowi(&fl6)); @@ -263,9 +265,9 @@ static int tcp_v6_connect(struct sock *sk, struct sockaddr *uaddr, tcp_fetch_timewait_stamp(sk, dst); icsk->icsk_ext_hdr_len = 0; - if (np->opt) - icsk->icsk_ext_hdr_len = (np->opt->opt_flen + - np->opt->opt_nflen); + if (opt) + icsk->icsk_ext_hdr_len = opt->opt_flen + + opt->opt_nflen; tp->rx_opt.mss_clamp = IPV6_MIN_MTU - sizeof(struct tcphdr) - sizeof(struct ipv6hdr); @@ -461,7 +463,8 @@ static int tcp_v6_send_synack(const struct sock *sk, struct dst_entry *dst, if (np->repflow && ireq->pktopts) fl6->flowlabel = ip6_flowlabel(ipv6_hdr(ireq->pktopts)); - err = ip6_xmit(sk, skb, fl6, np->opt, np->tclass); + err = ip6_xmit(sk, skb, fl6, rcu_dereference(np->opt), + np->tclass); err = net_xmit_eval(err); } @@ -972,6 +975,7 @@ static struct sock *tcp_v6_syn_recv_sock(const struct sock *sk, struct sk_buff * struct inet_request_sock *ireq; struct ipv6_pinfo *newnp; const struct ipv6_pinfo *np = inet6_sk(sk); + struct ipv6_txoptions *opt; struct tcp6_sock *newtcp6sk; struct inet_sock *newinet; struct tcp_sock *newtp; @@ -1098,13 +1102,15 @@ static struct sock *tcp_v6_syn_recv_sock(const struct sock *sk, struct sk_buff * but we make one more one thing there: reattach optmem to newsk. */ - if (np->opt) - newnp->opt = ipv6_dup_options(newsk, np->opt); - + opt = rcu_dereference(np->opt); + if (opt) { + opt = ipv6_dup_options(newsk, opt); + RCU_INIT_POINTER(newnp->opt, opt); + } inet_csk(newsk)->icsk_ext_hdr_len = 0; - if (newnp->opt) - inet_csk(newsk)->icsk_ext_hdr_len = (newnp->opt->opt_nflen + - newnp->opt->opt_flen); + if (opt) + inet_csk(newsk)->icsk_ext_hdr_len = opt->opt_nflen + + opt->opt_flen; tcp_ca_openreq_child(newsk, dst); -- cgit v1.2.3 From 6bd4f355df2eae80b8a5c7b097371cd1e05f20d5 Mon Sep 17 00:00:00 2001 From: Eric Dumazet Date: Wed, 2 Dec 2015 21:53:57 -0800 Subject: ipv6: kill sk_dst_lock While testing the np->opt RCU conversion, I found that UDP/IPv6 was using a mixture of xchg() and sk_dst_lock to protect concurrent changes to sk->sk_dst_cache, leading to possible corruptions and crashes. ip6_sk_dst_lookup_flow() uses sk_dst_check() anyway, so the simplest way to fix the mess is to remove sk_dst_lock completely, as we did for IPv4. __ip6_dst_store() and ip6_dst_store() share same implementation. sk_setup_caps() being called with socket lock being held or not, we have to use sk_dst_set() instead of __sk_dst_set() Note that I had to move the "np->dst_cookie = rt6_get_cookie(rt);" in ip6_dst_store() before the sk_setup_caps(sk, dst) call. This is because ip6_dst_store() can be called from process context, without any lock held. As soon as the dst is installed in sk->sk_dst_cache, dst can be freed from another cpu doing a concurrent ip6_dst_store() Doing the dst dereference before doing the install is needed to make sure no use after free would trigger. Signed-off-by: Eric Dumazet Reported-by: Dmitry Vyukov Signed-off-by: David S. Miller --- include/net/ip6_route.h | 17 ++++------------- include/net/sock.h | 3 +-- net/core/sock.c | 4 +--- net/dccp/ipv6.c | 4 ++-- net/ipv6/af_inet6.c | 2 +- net/ipv6/icmp.c | 14 -------------- net/ipv6/inet6_connection_sock.c | 10 +--------- net/ipv6/tcp_ipv6.c | 4 ++-- 8 files changed, 12 insertions(+), 46 deletions(-) (limited to 'net/ipv6/tcp_ipv6.c') diff --git a/include/net/ip6_route.h b/include/net/ip6_route.h index 2bfb2ad2fab1..877f682989b8 100644 --- a/include/net/ip6_route.h +++ b/include/net/ip6_route.h @@ -133,27 +133,18 @@ void rt6_clean_tohost(struct net *net, struct in6_addr *gateway); /* * Store a destination cache entry in a socket */ -static inline void __ip6_dst_store(struct sock *sk, struct dst_entry *dst, - const struct in6_addr *daddr, - const struct in6_addr *saddr) +static inline void ip6_dst_store(struct sock *sk, struct dst_entry *dst, + const struct in6_addr *daddr, + const struct in6_addr *saddr) { struct ipv6_pinfo *np = inet6_sk(sk); - struct rt6_info *rt = (struct rt6_info *) dst; + np->dst_cookie = rt6_get_cookie((struct rt6_info *)dst); sk_setup_caps(sk, dst); np->daddr_cache = daddr; #ifdef CONFIG_IPV6_SUBTREES np->saddr_cache = saddr; #endif - np->dst_cookie = rt6_get_cookie(rt); -} - -static inline void ip6_dst_store(struct sock *sk, struct dst_entry *dst, - struct in6_addr *daddr, struct in6_addr *saddr) -{ - spin_lock(&sk->sk_dst_lock); - __ip6_dst_store(sk, dst, daddr, saddr); - spin_unlock(&sk->sk_dst_lock); } static inline bool ipv6_unicast_destination(const struct sk_buff *skb) diff --git a/include/net/sock.h b/include/net/sock.h index 0434138c5f95..52d27ee924f4 100644 --- a/include/net/sock.h +++ b/include/net/sock.h @@ -254,7 +254,6 @@ struct cg_proto; * @sk_wq: sock wait queue and async head * @sk_rx_dst: receive input route used by early demux * @sk_dst_cache: destination cache - * @sk_dst_lock: destination cache lock * @sk_policy: flow policy * @sk_receive_queue: incoming packets * @sk_wmem_alloc: transmit queue bytes committed @@ -393,7 +392,7 @@ struct sock { #endif struct dst_entry *sk_rx_dst; struct dst_entry __rcu *sk_dst_cache; - spinlock_t sk_dst_lock; + /* Note: 32bit hole on 64bit arches */ atomic_t sk_wmem_alloc; atomic_t sk_omem_alloc; int sk_sndbuf; diff --git a/net/core/sock.c b/net/core/sock.c index 9d79569935a3..e31dfcee1729 100644 --- a/net/core/sock.c +++ b/net/core/sock.c @@ -1530,7 +1530,6 @@ struct sock *sk_clone_lock(const struct sock *sk, const gfp_t priority) skb_queue_head_init(&newsk->sk_receive_queue); skb_queue_head_init(&newsk->sk_write_queue); - spin_lock_init(&newsk->sk_dst_lock); rwlock_init(&newsk->sk_callback_lock); lockdep_set_class_and_name(&newsk->sk_callback_lock, af_callback_keys + newsk->sk_family, @@ -1607,7 +1606,7 @@ void sk_setup_caps(struct sock *sk, struct dst_entry *dst) { u32 max_segs = 1; - __sk_dst_set(sk, dst); + sk_dst_set(sk, dst); sk->sk_route_caps = dst->dev->features; if (sk->sk_route_caps & NETIF_F_GSO) sk->sk_route_caps |= NETIF_F_GSO_SOFTWARE; @@ -2388,7 +2387,6 @@ void sock_init_data(struct socket *sock, struct sock *sk) } else sk->sk_wq = NULL; - spin_lock_init(&sk->sk_dst_lock); rwlock_init(&sk->sk_callback_lock); lockdep_set_class_and_name(&sk->sk_callback_lock, af_callback_keys + sk->sk_family, diff --git a/net/dccp/ipv6.c b/net/dccp/ipv6.c index e7e0b9bc2a43..9c6d0508e63a 100644 --- a/net/dccp/ipv6.c +++ b/net/dccp/ipv6.c @@ -459,7 +459,7 @@ static struct sock *dccp_v6_request_recv_sock(const struct sock *sk, * comment in that function for the gory details. -acme */ - __ip6_dst_store(newsk, dst, NULL, NULL); + ip6_dst_store(newsk, dst, NULL, NULL); newsk->sk_route_caps = dst->dev->features & ~(NETIF_F_IP_CSUM | NETIF_F_TSO); newdp6 = (struct dccp6_sock *)newsk; @@ -883,7 +883,7 @@ static int dccp_v6_connect(struct sock *sk, struct sockaddr *uaddr, np->saddr = *saddr; inet->inet_rcv_saddr = LOOPBACK4_IPV6; - __ip6_dst_store(sk, dst, NULL, NULL); + ip6_dst_store(sk, dst, NULL, NULL); icsk->icsk_ext_hdr_len = 0; if (opt) diff --git a/net/ipv6/af_inet6.c b/net/ipv6/af_inet6.c index 38d66ddfb937..8ec0df75f1c4 100644 --- a/net/ipv6/af_inet6.c +++ b/net/ipv6/af_inet6.c @@ -673,7 +673,7 @@ int inet6_sk_rebuild_header(struct sock *sk) return PTR_ERR(dst); } - __ip6_dst_store(sk, dst, NULL, NULL); + ip6_dst_store(sk, dst, NULL, NULL); } return 0; diff --git a/net/ipv6/icmp.c b/net/ipv6/icmp.c index 36c5a98b0472..0a37ddc7af51 100644 --- a/net/ipv6/icmp.c +++ b/net/ipv6/icmp.c @@ -834,11 +834,6 @@ void icmpv6_flow_init(struct sock *sk, struct flowi6 *fl6, security_sk_classify_flow(sk, flowi6_to_flowi(fl6)); } -/* - * Special lock-class for __icmpv6_sk: - */ -static struct lock_class_key icmpv6_socket_sk_dst_lock_key; - static int __net_init icmpv6_sk_init(struct net *net) { struct sock *sk; @@ -860,15 +855,6 @@ static int __net_init icmpv6_sk_init(struct net *net) net->ipv6.icmp_sk[i] = sk; - /* - * Split off their lock-class, because sk->sk_dst_lock - * gets used from softirqs, which is safe for - * __icmpv6_sk (because those never get directly used - * via userspace syscalls), but unsafe for normal sockets. - */ - lockdep_set_class(&sk->sk_dst_lock, - &icmpv6_socket_sk_dst_lock_key); - /* Enough space for 2 64K ICMP packets, including * sk_buff struct overhead. */ diff --git a/net/ipv6/inet6_connection_sock.c b/net/ipv6/inet6_connection_sock.c index 3ff5208772bb..a7ca2cde2ecb 100644 --- a/net/ipv6/inet6_connection_sock.c +++ b/net/ipv6/inet6_connection_sock.c @@ -110,14 +110,6 @@ void inet6_csk_addr2sockaddr(struct sock *sk, struct sockaddr *uaddr) } EXPORT_SYMBOL_GPL(inet6_csk_addr2sockaddr); -static inline -void __inet6_csk_dst_store(struct sock *sk, struct dst_entry *dst, - const struct in6_addr *daddr, - const struct in6_addr *saddr) -{ - __ip6_dst_store(sk, dst, daddr, saddr); -} - static inline struct dst_entry *__inet6_csk_dst_check(struct sock *sk, u32 cookie) { @@ -153,7 +145,7 @@ static struct dst_entry *inet6_csk_route_socket(struct sock *sk, dst = ip6_dst_lookup_flow(sk, fl6, final_p); if (!IS_ERR(dst)) - __inet6_csk_dst_store(sk, dst, NULL, NULL); + ip6_dst_store(sk, dst, NULL, NULL); } return dst; } diff --git a/net/ipv6/tcp_ipv6.c b/net/ipv6/tcp_ipv6.c index 6a50bb4a0dae..e7aab561b7b4 100644 --- a/net/ipv6/tcp_ipv6.c +++ b/net/ipv6/tcp_ipv6.c @@ -257,7 +257,7 @@ static int tcp_v6_connect(struct sock *sk, struct sockaddr *uaddr, inet->inet_rcv_saddr = LOOPBACK4_IPV6; sk->sk_gso_type = SKB_GSO_TCPV6; - __ip6_dst_store(sk, dst, NULL, NULL); + ip6_dst_store(sk, dst, NULL, NULL); if (tcp_death_row.sysctl_tw_recycle && !tp->rx_opt.ts_recent_stamp && @@ -1060,7 +1060,7 @@ static struct sock *tcp_v6_syn_recv_sock(const struct sock *sk, struct sk_buff * */ newsk->sk_gso_type = SKB_GSO_TCPV6; - __ip6_dst_store(newsk, dst, NULL, NULL); + ip6_dst_store(newsk, dst, NULL, NULL); inet6_sk_rx_dst_set(newsk, skb); newtcp6sk = (struct tcp6_sock *)newsk; -- cgit v1.2.3 From 5037e9ef9454917b047f9f3a19b4dd179fbf7cd4 Mon Sep 17 00:00:00 2001 From: Eric Dumazet Date: Mon, 14 Dec 2015 14:08:53 -0800 Subject: net: fix IP early demux races David Wilder reported crashes caused by dst reuse. I am seeing a crash on a distro V4.2.3 kernel caused by a double release of a dst_entry. In ipv4_dst_destroy() the call to list_empty() finds a poisoned next pointer, indicating the dst_entry has already been removed from the list and freed. The crash occurs 18 to 24 hours into a run of a network stress exerciser. Thanks to his detailed report and analysis, we were able to understand the core issue. IP early demux can associate a dst to skb, after a lookup in TCP/UDP sockets. When socket cache is not properly set, we want to store into sk->sk_dst_cache the dst for future IP early demux lookups, by acquiring a stable refcount on the dst. Problem is this acquisition is simply using an atomic_inc(), which works well, unless the dst was queued for destruction from dst_release() noticing dst refcount went to zero, if DST_NOCACHE was set on dst. We need to make sure current refcount is not zero before incrementing it, or risk double free as David reported. This patch, being a stable candidate, adds two new helpers, and use them only from IP early demux problematic paths. It might be possible to merge in net-next skb_dst_force() and skb_dst_force_safe(), but I prefer having the smallest patch for stable kernels : Maybe some skb_dst_force() callers do not expect skb->dst can suddenly be cleared. Can probably be backported back to linux-3.6 kernels Reported-by: David J. Wilder Tested-by: David J. Wilder Signed-off-by: Eric Dumazet Signed-off-by: David S. Miller --- include/net/dst.h | 33 +++++++++++++++++++++++++++++++++ include/net/sock.h | 2 +- net/ipv4/tcp_ipv4.c | 5 ++--- net/ipv6/tcp_ipv6.c | 3 +-- 4 files changed, 37 insertions(+), 6 deletions(-) (limited to 'net/ipv6/tcp_ipv6.c') diff --git a/include/net/dst.h b/include/net/dst.h index 1279f9b09791..c7329dcd90cc 100644 --- a/include/net/dst.h +++ b/include/net/dst.h @@ -322,6 +322,39 @@ static inline void skb_dst_force(struct sk_buff *skb) } } +/** + * dst_hold_safe - Take a reference on a dst if possible + * @dst: pointer to dst entry + * + * This helper returns false if it could not safely + * take a reference on a dst. + */ +static inline bool dst_hold_safe(struct dst_entry *dst) +{ + if (dst->flags & DST_NOCACHE) + return atomic_inc_not_zero(&dst->__refcnt); + dst_hold(dst); + return true; +} + +/** + * skb_dst_force_safe - makes sure skb dst is refcounted + * @skb: buffer + * + * If dst is not yet refcounted and not destroyed, grab a ref on it. + */ +static inline void skb_dst_force_safe(struct sk_buff *skb) +{ + if (skb_dst_is_noref(skb)) { + struct dst_entry *dst = skb_dst(skb); + + if (!dst_hold_safe(dst)) + dst = NULL; + + skb->_skb_refdst = (unsigned long)dst; + } +} + /** * __skb_tunnel_rx - prepare skb for rx reinsert diff --git a/include/net/sock.h b/include/net/sock.h index c4205e0a3a2d..28790fe18206 100644 --- a/include/net/sock.h +++ b/include/net/sock.h @@ -817,7 +817,7 @@ void sk_stream_write_space(struct sock *sk); static inline void __sk_add_backlog(struct sock *sk, struct sk_buff *skb) { /* dont let skb dst not refcounted, we are going to leave rcu lock */ - skb_dst_force(skb); + skb_dst_force_safe(skb); if (!sk->sk_backlog.tail) sk->sk_backlog.head = skb; diff --git a/net/ipv4/tcp_ipv4.c b/net/ipv4/tcp_ipv4.c index db003438aaf5..d8841a2f1569 100644 --- a/net/ipv4/tcp_ipv4.c +++ b/net/ipv4/tcp_ipv4.c @@ -1493,7 +1493,7 @@ bool tcp_prequeue(struct sock *sk, struct sk_buff *skb) if (likely(sk->sk_rx_dst)) skb_dst_drop(skb); else - skb_dst_force(skb); + skb_dst_force_safe(skb); __skb_queue_tail(&tp->ucopy.prequeue, skb); tp->ucopy.memory += skb->truesize; @@ -1721,8 +1721,7 @@ void inet_sk_rx_dst_set(struct sock *sk, const struct sk_buff *skb) { struct dst_entry *dst = skb_dst(skb); - if (dst) { - dst_hold(dst); + if (dst && dst_hold_safe(dst)) { sk->sk_rx_dst = dst; inet_sk(sk)->rx_dst_ifindex = skb->skb_iif; } diff --git a/net/ipv6/tcp_ipv6.c b/net/ipv6/tcp_ipv6.c index e7aab561b7b4..6b8a8a9091fa 100644 --- a/net/ipv6/tcp_ipv6.c +++ b/net/ipv6/tcp_ipv6.c @@ -93,10 +93,9 @@ static void inet6_sk_rx_dst_set(struct sock *sk, const struct sk_buff *skb) { struct dst_entry *dst = skb_dst(skb); - if (dst) { + if (dst && dst_hold_safe(dst)) { const struct rt6_info *rt = (const struct rt6_info *)dst; - dst_hold(dst); sk->sk_rx_dst = dst; inet_sk(sk)->rx_dst_ifindex = skb->skb_iif; inet6_sk(sk)->rx_dst_cookie = rt6_get_cookie(rt); -- cgit v1.2.3