From: Eric Dumazet edumazet@google.com
commit 086d49058cd8471046ae9927524708820f5fd1c7 upstream.
IPv6 has this hack changing sk->sk_prot when an IPv6 socket is 'converted' to an IPv4 one with IPV6_ADDRFORM option.
This operation is only performed for TCP and UDP, knowing their 'struct proto' for the two network families are populated in the same way, and can not disappear while a reader might use and dereference sk->sk_prot.
If we think about it all reads of sk->sk_prot while either socket lock or RTNL is not acquired should be using READ_ONCE().
Also note that other layers like MPTCP, XFRM, CHELSIO_TLS also write over sk->sk_prot.
BUG: KCSAN: data-race in inet6_recvmsg / ipv6_setsockopt
write to 0xffff8881386f7aa8 of 8 bytes by task 26932 on cpu 0: do_ipv6_setsockopt net/ipv6/ipv6_sockglue.c:492 [inline] ipv6_setsockopt+0x3758/0x3910 net/ipv6/ipv6_sockglue.c:1019 udpv6_setsockopt+0x85/0x90 net/ipv6/udp.c:1649 sock_common_setsockopt+0x5d/0x70 net/core/sock.c:3489 __sys_setsockopt+0x209/0x2a0 net/socket.c:2180 __do_sys_setsockopt net/socket.c:2191 [inline] __se_sys_setsockopt net/socket.c:2188 [inline] __x64_sys_setsockopt+0x62/0x70 net/socket.c:2188 do_syscall_x64 arch/x86/entry/common.c:50 [inline] do_syscall_64+0x44/0xd0 arch/x86/entry/common.c:80 entry_SYSCALL_64_after_hwframe+0x44/0xae
read to 0xffff8881386f7aa8 of 8 bytes by task 26911 on cpu 1: inet6_recvmsg+0x7a/0x210 net/ipv6/af_inet6.c:659 ____sys_recvmsg+0x16c/0x320 ___sys_recvmsg net/socket.c:2674 [inline] do_recvmmsg+0x3f5/0xae0 net/socket.c:2768 __sys_recvmmsg net/socket.c:2847 [inline] __do_sys_recvmmsg net/socket.c:2870 [inline] __se_sys_recvmmsg net/socket.c:2863 [inline] __x64_sys_recvmmsg+0xde/0x160 net/socket.c:2863 do_syscall_x64 arch/x86/entry/common.c:50 [inline] do_syscall_64+0x44/0xd0 arch/x86/entry/common.c:80 entry_SYSCALL_64_after_hwframe+0x44/0xae
value changed: 0xffffffff85e0e980 -> 0xffffffff85e01580
Reported by Kernel Concurrency Sanitizer on: CPU: 1 PID: 26911 Comm: syz-executor.3 Not tainted 5.17.0-rc2-syzkaller-00316-g0457e5153e0e-dirty #0 Hardware name: Google Google Compute Engine/Google Compute Engine, BIOS Google 01/01/2011
Reported-by: syzbot syzkaller@googlegroups.com Signed-off-by: Eric Dumazet edumazet@google.com Signed-off-by: David S. Miller davem@davemloft.net Signed-off-by: Kazunori Kobayashi kazunori.kobayashi@miraclelinux.com --- net/ipv6/af_inet6.c | 24 ++++++++++++++++++------ net/ipv6/ipv6_sockglue.c | 6 ++++-- 2 files changed, 22 insertions(+), 8 deletions(-)
diff --git a/net/ipv6/af_inet6.c b/net/ipv6/af_inet6.c index 8e0c33b683010..bbfa5d3a16f01 100644 --- a/net/ipv6/af_inet6.c +++ b/net/ipv6/af_inet6.c @@ -452,11 +452,14 @@ int inet6_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len) { struct sock *sk = sock->sk; u32 flags = BIND_WITH_LOCK; + const struct proto *prot; int err = 0;
+ /* IPV6_ADDRFORM can change sk->sk_prot under us. */ + prot = READ_ONCE(sk->sk_prot); /* If the socket has its own bind function then use it. */ - if (sk->sk_prot->bind) - return sk->sk_prot->bind(sk, uaddr, addr_len); + if (prot->bind) + return prot->bind(sk, uaddr, addr_len);
if (addr_len < SIN6_LEN_RFC2133) return -EINVAL; @@ -572,6 +575,7 @@ int inet6_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg) void __user *argp = (void __user *)arg; struct sock *sk = sock->sk; struct net *net = sock_net(sk); + const struct proto *prot;
switch (cmd) { case SIOCADDRT: @@ -589,9 +593,11 @@ int inet6_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg) case SIOCSIFDSTADDR: return addrconf_set_dstaddr(net, argp); default: - if (!sk->sk_prot->ioctl) + /* IPV6_ADDRFORM can change sk->sk_prot under us. */ + prot = READ_ONCE(sk->sk_prot); + if (!prot->ioctl) return -ENOIOCTLCMD; - return sk->sk_prot->ioctl(sk, cmd, arg); + return prot->ioctl(sk, cmd, arg); } /*NOTREACHED*/ return 0; @@ -653,11 +659,14 @@ INDIRECT_CALLABLE_DECLARE(int udpv6_sendmsg(struct sock *, struct msghdr *, int inet6_sendmsg(struct socket *sock, struct msghdr *msg, size_t size) { struct sock *sk = sock->sk; + const struct proto *prot;
if (unlikely(inet_send_prepare(sk))) return -EAGAIN;
- return INDIRECT_CALL_2(sk->sk_prot->sendmsg, tcp_sendmsg, udpv6_sendmsg, + /* IPV6_ADDRFORM can change sk->sk_prot under us. */ + prot = READ_ONCE(sk->sk_prot); + return INDIRECT_CALL_2(prot->sendmsg, tcp_sendmsg, udpv6_sendmsg, sk, msg, size); }
@@ -667,13 +676,16 @@ int inet6_recvmsg(struct socket *sock, struct msghdr *msg, size_t size, int flags) { struct sock *sk = sock->sk; + const struct proto *prot; int addr_len = 0; int err;
if (likely(!(flags & MSG_ERRQUEUE))) sock_rps_record_flow(sk);
- err = INDIRECT_CALL_2(sk->sk_prot->recvmsg, tcp_recvmsg, udpv6_recvmsg, + /* IPV6_ADDRFORM can change sk->sk_prot under us. */ + prot = READ_ONCE(sk->sk_prot); + err = INDIRECT_CALL_2(prot->recvmsg, tcp_recvmsg, udpv6_recvmsg, sk, msg, size, flags & MSG_DONTWAIT, flags & ~MSG_DONTWAIT, &addr_len); if (err >= 0) diff --git a/net/ipv6/ipv6_sockglue.c b/net/ipv6/ipv6_sockglue.c index 197e12d5607f1..485d381aa4475 100644 --- a/net/ipv6/ipv6_sockglue.c +++ b/net/ipv6/ipv6_sockglue.c @@ -475,7 +475,8 @@ static int do_ipv6_setsockopt(struct sock *sk, int level, int optname, sock_prot_inuse_add(net, sk->sk_prot, -1); sock_prot_inuse_add(net, &tcp_prot, 1); local_bh_enable(); - sk->sk_prot = &tcp_prot; + /* Paired with READ_ONCE(sk->sk_prot) in net/ipv6/af_inet6.c */ + WRITE_ONCE(sk->sk_prot, &tcp_prot); icsk->icsk_af_ops = &ipv4_specific; sk->sk_socket->ops = &inet_stream_ops; sk->sk_family = PF_INET; @@ -489,7 +490,8 @@ static int do_ipv6_setsockopt(struct sock *sk, int level, int optname, sock_prot_inuse_add(net, sk->sk_prot, -1); sock_prot_inuse_add(net, prot, 1); local_bh_enable(); - sk->sk_prot = prot; + /* Paired with READ_ONCE(sk->sk_prot) in net/ipv6/af_inet6.c */ + WRITE_ONCE(sk->sk_prot, prot); sk->sk_socket->ops = &inet_dgram_ops; sk->sk_family = PF_INET; }