| Message ID | 20221006185349.74777-3-kuniyu@amazon.com (mailing list archive) |
|---|---|
| State | Accepted |
| Commit | 21985f43376cee092702d6cb963ff97a9d2ede68 |
| Delegated to: | Netdev Maintainers |
| Headers | show |
| Series | tcp/udp: Fix memory leaks and data races around IPV6_ADDRFORM. | expand |
On Thu, 2022-10-06 at 11:53 -0700, Kuniyuki Iwashima wrote: > Commit 4b340ae20d0e ("IPv6: Complete IPV6_DONTFRAG support") forgot > to add a change to free inet6_sk(sk)->rxpmtu while converting an IPv6 > socket into IPv4 with IPV6_ADDRFORM. After conversion, sk_prot is > changed to udp_prot and ->destroy() never cleans it up, resulting in > a memory leak. > > This is due to the discrepancy between inet6_destroy_sock() and > IPV6_ADDRFORM, so let's call inet6_destroy_sock() from IPV6_ADDRFORM > to remove the difference. > > However, this is not enough for now because rxpmtu can be changed > without lock_sock() after commit 03485f2adcde ("udpv6: Add lockless > sendmsg() support"). We will fix this case in the following patch. > > Note we will rename inet6_destroy_sock() to inet6_cleanup_sock() and > remove unnecessary inet6_destroy_sock() calls in sk_prot->destroy() > in the future. > > Fixes: 4b340ae20d0e ("IPv6: Complete IPV6_DONTFRAG support") > Signed-off-by: Kuniyuki Iwashima <kuniyu@amazon.com> > --- > Cc: Brian Haley <brian.haley@hp.com> > --- > include/net/ipv6.h | 1 + > net/ipv6/af_inet6.c | 6 ++++++ > net/ipv6/ipv6_sockglue.c | 20 ++++++++------------ > 3 files changed, 15 insertions(+), 12 deletions(-) > > diff --git a/include/net/ipv6.h b/include/net/ipv6.h > index d664ba5812d8..335a49ecd8a0 100644 > --- a/include/net/ipv6.h > +++ b/include/net/ipv6.h > @@ -1182,6 +1182,7 @@ void ipv6_icmp_error(struct sock *sk, struct sk_buff *skb, int err, __be16 port, > void ipv6_local_error(struct sock *sk, int err, struct flowi6 *fl6, u32 info); > void ipv6_local_rxpmtu(struct sock *sk, struct flowi6 *fl6, u32 mtu); > > +void inet6_cleanup_sock(struct sock *sk); > int inet6_release(struct socket *sock); > int inet6_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len); > int inet6_getname(struct socket *sock, struct sockaddr *uaddr, > diff --git a/net/ipv6/af_inet6.c b/net/ipv6/af_inet6.c > index d40b7d60e00e..ded827944fa6 100644 > --- a/net/ipv6/af_inet6.c > +++ b/net/ipv6/af_inet6.c > @@ -510,6 +510,12 @@ void inet6_destroy_sock(struct sock *sk) > } > EXPORT_SYMBOL_GPL(inet6_destroy_sock); > > +void inet6_cleanup_sock(struct sock *sk) > +{ > + inet6_destroy_sock(sk); > +} > +EXPORT_SYMBOL_GPL(inet6_cleanup_sock); > + > /* > * This does both peername and sockname. > */ > diff --git a/net/ipv6/ipv6_sockglue.c b/net/ipv6/ipv6_sockglue.c > index 408345fc4c5c..a20edae868fd 100644 > --- a/net/ipv6/ipv6_sockglue.c > +++ b/net/ipv6/ipv6_sockglue.c > @@ -431,9 +431,6 @@ int do_ipv6_setsockopt(struct sock *sk, int level, int optname, > if (optlen < sizeof(int)) > goto e_inval; > if (val == PF_INET) { > - struct ipv6_txoptions *opt; > - struct sk_buff *pktopt; > - > if (sk->sk_type == SOCK_RAW) > break; > > @@ -464,7 +461,6 @@ int do_ipv6_setsockopt(struct sock *sk, int level, int optname, > break; > } > > - fl6_free_socklist(sk); > __ipv6_sock_mc_close(sk); > __ipv6_sock_ac_close(sk); > > @@ -501,14 +497,14 @@ int do_ipv6_setsockopt(struct sock *sk, int level, int optname, > sk->sk_socket->ops = &inet_dgram_ops; > sk->sk_family = PF_INET; > } > - opt = xchg((__force struct ipv6_txoptions **)&np->opt, > - NULL); > - if (opt) { > - atomic_sub(opt->tot_len, &sk->sk_omem_alloc); > - txopt_put(opt); > - } > - pktopt = xchg(&np->pktoptions, NULL); > - kfree_skb(pktopt); > + > + /* Disable all options not to allocate memory anymore, > + * but there is still a race. See the lockless path > + * in udpv6_sendmsg() and ipv6_local_rxpmtu(). > + */ > + np->rxopt.all = 0; > + > + inet6_cleanup_sock(sk); I think there still a pending point raised from Eric here. """ Once the v6 socket has been transformed to IPv4 one, inet6_sock_destruct() is not going to be called. """ AFAICS the series is safe Kuniyuki noted: """ inet6_sock_destruct() is set to sk->sk_destruct(), which is not changed by the transformation and will be called from __sk_destruct(). """ [with the next patch] @Eric are you fine with the above? Thanks! Paolo > > /* > * ... and add it to the refcnt debug socks count
diff --git a/include/net/ipv6.h b/include/net/ipv6.h index d664ba5812d8..335a49ecd8a0 100644 --- a/include/net/ipv6.h +++ b/include/net/ipv6.h @@ -1182,6 +1182,7 @@ void ipv6_icmp_error(struct sock *sk, struct sk_buff *skb, int err, __be16 port, void ipv6_local_error(struct sock *sk, int err, struct flowi6 *fl6, u32 info); void ipv6_local_rxpmtu(struct sock *sk, struct flowi6 *fl6, u32 mtu); +void inet6_cleanup_sock(struct sock *sk); int inet6_release(struct socket *sock); int inet6_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len); int inet6_getname(struct socket *sock, struct sockaddr *uaddr, diff --git a/net/ipv6/af_inet6.c b/net/ipv6/af_inet6.c index d40b7d60e00e..ded827944fa6 100644 --- a/net/ipv6/af_inet6.c +++ b/net/ipv6/af_inet6.c @@ -510,6 +510,12 @@ void inet6_destroy_sock(struct sock *sk) } EXPORT_SYMBOL_GPL(inet6_destroy_sock); +void inet6_cleanup_sock(struct sock *sk) +{ + inet6_destroy_sock(sk); +} +EXPORT_SYMBOL_GPL(inet6_cleanup_sock); + /* * This does both peername and sockname. */ diff --git a/net/ipv6/ipv6_sockglue.c b/net/ipv6/ipv6_sockglue.c index 408345fc4c5c..a20edae868fd 100644 --- a/net/ipv6/ipv6_sockglue.c +++ b/net/ipv6/ipv6_sockglue.c @@ -431,9 +431,6 @@ int do_ipv6_setsockopt(struct sock *sk, int level, int optname, if (optlen < sizeof(int)) goto e_inval; if (val == PF_INET) { - struct ipv6_txoptions *opt; - struct sk_buff *pktopt; - if (sk->sk_type == SOCK_RAW) break; @@ -464,7 +461,6 @@ int do_ipv6_setsockopt(struct sock *sk, int level, int optname, break; } - fl6_free_socklist(sk); __ipv6_sock_mc_close(sk); __ipv6_sock_ac_close(sk); @@ -501,14 +497,14 @@ int do_ipv6_setsockopt(struct sock *sk, int level, int optname, sk->sk_socket->ops = &inet_dgram_ops; sk->sk_family = PF_INET; } - opt = xchg((__force struct ipv6_txoptions **)&np->opt, - NULL); - if (opt) { - atomic_sub(opt->tot_len, &sk->sk_omem_alloc); - txopt_put(opt); - } - pktopt = xchg(&np->pktoptions, NULL); - kfree_skb(pktopt); + + /* Disable all options not to allocate memory anymore, + * but there is still a race. See the lockless path + * in udpv6_sendmsg() and ipv6_local_rxpmtu(). + */ + np->rxopt.all = 0; + + inet6_cleanup_sock(sk); /* * ... and add it to the refcnt debug socks count
Commit 4b340ae20d0e ("IPv6: Complete IPV6_DONTFRAG support") forgot to add a change to free inet6_sk(sk)->rxpmtu while converting an IPv6 socket into IPv4 with IPV6_ADDRFORM. After conversion, sk_prot is changed to udp_prot and ->destroy() never cleans it up, resulting in a memory leak. This is due to the discrepancy between inet6_destroy_sock() and IPV6_ADDRFORM, so let's call inet6_destroy_sock() from IPV6_ADDRFORM to remove the difference. However, this is not enough for now because rxpmtu can be changed without lock_sock() after commit 03485f2adcde ("udpv6: Add lockless sendmsg() support"). We will fix this case in the following patch. Note we will rename inet6_destroy_sock() to inet6_cleanup_sock() and remove unnecessary inet6_destroy_sock() calls in sk_prot->destroy() in the future. Fixes: 4b340ae20d0e ("IPv6: Complete IPV6_DONTFRAG support") Signed-off-by: Kuniyuki Iwashima <kuniyu@amazon.com> --- Cc: Brian Haley <brian.haley@hp.com> --- include/net/ipv6.h | 1 + net/ipv6/af_inet6.c | 6 ++++++ net/ipv6/ipv6_sockglue.c | 20 ++++++++------------ 3 files changed, 15 insertions(+), 12 deletions(-)