Receiving ACK with a valid SYN cookie, cookie_v4_check() allocates struct
request_sock and then can allocate inet_rsk(req)->ireq_opt. After that,
tcp_v4_syn_recv_sock() allocates struct sock and copies ireq_opt to
inet_sk(sk)->inet_opt. Normally, tcp_v4_syn_recv_sock() inserts the full
socket into ehash and sets NULL to ireq_opt. Otherwise,
tcp_v4_syn_recv_sock() has to reset inet_opt by NULL and free the full
socket.
The commit 01770a1661657 ("tcp: fix race condition when creating child
sockets from syncookies") added a new path, in which more than one cores
create full sockets for the same SYN cookie. Currently, the core which
loses the race frees the full socket without resetting inet_opt, resulting
in that both sock_put() and reqsk_put() call kfree() for the same memory:
sock_put
sk_free
__sk_free
sk_destruct
__sk_destruct
sk->sk_destruct/inet_sock_destruct
kfree(rcu_dereference_protected(inet->inet_opt, 1));
reqsk_put
reqsk_free
__reqsk_free
req->rsk_ops->destructor/tcp_v4_reqsk_destructor
kfree(rcu_dereference_protected(inet_rsk(req)->ireq_opt, 1));
Calling kmalloc() between the double kfree() can lead to use-after-free, so
this patch fixes it by setting NULL to inet_opt before sock_put().
As a side note, this kind of issue does not happen for IPv6. This is
because tcp_v6_syn_recv_sock() clones both ipv6_opt and pktopts which
correspond to ireq_opt in IPv4.
Fixes: 01770a166165 ("tcp: fix race condition when creating child sockets from syncookies")
CC: Ricardo Dias <[email protected]>
Signed-off-by: Kuniyuki Iwashima <[email protected]>
Reviewed-by: Benjamin Herrenschmidt <[email protected]>
---
net/ipv4/tcp_ipv4.c | 4 ++--
1 file changed, 2 insertions(+), 2 deletions(-)
diff --git a/net/ipv4/tcp_ipv4.c b/net/ipv4/tcp_ipv4.c
index 58207c7769d0..87eb614dab27 100644
--- a/net/ipv4/tcp_ipv4.c
+++ b/net/ipv4/tcp_ipv4.c
@@ -1595,6 +1595,8 @@ struct sock *tcp_v4_syn_recv_sock(const struct sock *sk, struct sk_buff *skb,
tcp_move_syn(newtp, req);
ireq->ireq_opt = NULL;
} else {
+ newinet->inet_opt = NULL;
+
if (!req_unhash && found_dup_sk) {
/* This code path should only be executed in the
* syncookie case only
@@ -1602,8 +1604,6 @@ struct sock *tcp_v4_syn_recv_sock(const struct sock *sk, struct sk_buff *skb,
bh_unlock_sock(newsk);
sock_put(newsk);
newsk = NULL;
- } else {
- newinet->inet_opt = NULL;
}
}
return newsk;
--
2.17.2 (Apple Git-113)
On Mon, 18 Jan 2021 14:59:20 +0900 Kuniyuki Iwashima wrote:
> Receiving ACK with a valid SYN cookie, cookie_v4_check() allocates struct
> request_sock and then can allocate inet_rsk(req)->ireq_opt. After that,
> tcp_v4_syn_recv_sock() allocates struct sock and copies ireq_opt to
> inet_sk(sk)->inet_opt. Normally, tcp_v4_syn_recv_sock() inserts the full
> socket into ehash and sets NULL to ireq_opt. Otherwise,
> tcp_v4_syn_recv_sock() has to reset inet_opt by NULL and free the full
> socket.
>
> The commit 01770a1661657 ("tcp: fix race condition when creating child
> sockets from syncookies") added a new path, in which more than one cores
> create full sockets for the same SYN cookie. Currently, the core which
> loses the race frees the full socket without resetting inet_opt, resulting
> in that both sock_put() and reqsk_put() call kfree() for the same memory:
>
> sock_put
> sk_free
> __sk_free
> sk_destruct
> __sk_destruct
> sk->sk_destruct/inet_sock_destruct
> kfree(rcu_dereference_protected(inet->inet_opt, 1));
>
> reqsk_put
> reqsk_free
> __reqsk_free
> req->rsk_ops->destructor/tcp_v4_reqsk_destructor
> kfree(rcu_dereference_protected(inet_rsk(req)->ireq_opt, 1));
>
> Calling kmalloc() between the double kfree() can lead to use-after-free, so
> this patch fixes it by setting NULL to inet_opt before sock_put().
>
> As a side note, this kind of issue does not happen for IPv6. This is
> because tcp_v6_syn_recv_sock() clones both ipv6_opt and pktopts which
> correspond to ireq_opt in IPv4.
>
> Fixes: 01770a166165 ("tcp: fix race condition when creating child sockets from syncookies")
> CC: Ricardo Dias <[email protected]>
> Signed-off-by: Kuniyuki Iwashima <[email protected]>
> Reviewed-by: Benjamin Herrenschmidt <[email protected]>
Ricardo, Eric, any reason this was written this way?
> diff --git a/net/ipv4/tcp_ipv4.c b/net/ipv4/tcp_ipv4.c
> index 58207c7769d0..87eb614dab27 100644
> --- a/net/ipv4/tcp_ipv4.c
> +++ b/net/ipv4/tcp_ipv4.c
> @@ -1595,6 +1595,8 @@ struct sock *tcp_v4_syn_recv_sock(const struct sock *sk, struct sk_buff *skb,
> tcp_move_syn(newtp, req);
> ireq->ireq_opt = NULL;
> } else {
> + newinet->inet_opt = NULL;
> +
> if (!req_unhash && found_dup_sk) {
> /* This code path should only be executed in the
> * syncookie case only
> @@ -1602,8 +1604,6 @@ struct sock *tcp_v4_syn_recv_sock(const struct sock *sk, struct sk_buff *skb,
> bh_unlock_sock(newsk);
> sock_put(newsk);
> newsk = NULL;
> - } else {
> - newinet->inet_opt = NULL;
> }
> }
> return newsk;
On Wed, Jan 20, 2021 at 2:17 AM Jakub Kicinski <[email protected]> wrote:
>
> On Mon, 18 Jan 2021 14:59:20 +0900 Kuniyuki Iwashima wrote:
> > Receiving ACK with a valid SYN cookie, cookie_v4_check() allocates struct
> > request_sock and then can allocate inet_rsk(req)->ireq_opt. After that,
> > tcp_v4_syn_recv_sock() allocates struct sock and copies ireq_opt to
> > inet_sk(sk)->inet_opt. Normally, tcp_v4_syn_recv_sock() inserts the full
> > socket into ehash and sets NULL to ireq_opt. Otherwise,
> > tcp_v4_syn_recv_sock() has to reset inet_opt by NULL and free the full
> > socket.
> >
> > The commit 01770a1661657 ("tcp: fix race condition when creating child
> > sockets from syncookies") added a new path, in which more than one cores
> > create full sockets for the same SYN cookie. Currently, the core which
> > loses the race frees the full socket without resetting inet_opt, resulting
> > in that both sock_put() and reqsk_put() call kfree() for the same memory:
> >
> > sock_put
> > sk_free
> > __sk_free
> > sk_destruct
> > __sk_destruct
> > sk->sk_destruct/inet_sock_destruct
> > kfree(rcu_dereference_protected(inet->inet_opt, 1));
> >
> > reqsk_put
> > reqsk_free
> > __reqsk_free
> > req->rsk_ops->destructor/tcp_v4_reqsk_destructor
> > kfree(rcu_dereference_protected(inet_rsk(req)->ireq_opt, 1));
> >
> > Calling kmalloc() between the double kfree() can lead to use-after-free, so
> > this patch fixes it by setting NULL to inet_opt before sock_put().
> >
> > As a side note, this kind of issue does not happen for IPv6. This is
> > because tcp_v6_syn_recv_sock() clones both ipv6_opt and pktopts which
> > correspond to ireq_opt in IPv4.
> >
> > Fixes: 01770a166165 ("tcp: fix race condition when creating child sockets from syncookies")
> > CC: Ricardo Dias <[email protected]>
> > Signed-off-by: Kuniyuki Iwashima <[email protected]>
> > Reviewed-by: Benjamin Herrenschmidt <[email protected]>
>
> Ricardo, Eric, any reason this was written this way?
Well, I guess that was a plain bug.
IPv4 options are not used often I think.
Reviewed-by: Eric Dumazet <[email protected]>
On Wed, 20 Jan 2021 14:07:35 +0100 Eric Dumazet wrote:
> On Wed, Jan 20, 2021 at 2:17 AM Jakub Kicinski <[email protected]> wrote:
> > On Mon, 18 Jan 2021 14:59:20 +0900 Kuniyuki Iwashima wrote:
> > > Receiving ACK with a valid SYN cookie, cookie_v4_check() allocates struct
> > > request_sock and then can allocate inet_rsk(req)->ireq_opt. After that,
> > > tcp_v4_syn_recv_sock() allocates struct sock and copies ireq_opt to
> > > inet_sk(sk)->inet_opt. Normally, tcp_v4_syn_recv_sock() inserts the full
> > > socket into ehash and sets NULL to ireq_opt. Otherwise,
> > > tcp_v4_syn_recv_sock() has to reset inet_opt by NULL and free the full
> > > socket.
> > >
> > > The commit 01770a1661657 ("tcp: fix race condition when creating child
> > > sockets from syncookies") added a new path, in which more than one cores
> > > create full sockets for the same SYN cookie. Currently, the core which
> > > loses the race frees the full socket without resetting inet_opt, resulting
> > > in that both sock_put() and reqsk_put() call kfree() for the same memory:
> > >
> > > sock_put
> > > sk_free
> > > __sk_free
> > > sk_destruct
> > > __sk_destruct
> > > sk->sk_destruct/inet_sock_destruct
> > > kfree(rcu_dereference_protected(inet->inet_opt, 1));
> > >
> > > reqsk_put
> > > reqsk_free
> > > __reqsk_free
> > > req->rsk_ops->destructor/tcp_v4_reqsk_destructor
> > > kfree(rcu_dereference_protected(inet_rsk(req)->ireq_opt, 1));
> > >
> > > Calling kmalloc() between the double kfree() can lead to use-after-free, so
> > > this patch fixes it by setting NULL to inet_opt before sock_put().
> > >
> > > As a side note, this kind of issue does not happen for IPv6. This is
> > > because tcp_v6_syn_recv_sock() clones both ipv6_opt and pktopts which
> > > correspond to ireq_opt in IPv4.
> > >
> > > Fixes: 01770a166165 ("tcp: fix race condition when creating child sockets from syncookies")
> > > CC: Ricardo Dias <[email protected]>
> > > Signed-off-by: Kuniyuki Iwashima <[email protected]>
> > > Reviewed-by: Benjamin Herrenschmidt <[email protected]>
> >
> > Ricardo, Eric, any reason this was written this way?
>
> Well, I guess that was a plain bug.
>
> IPv4 options are not used often I think.
I see.
> Reviewed-by: Eric Dumazet <[email protected]>
Applied, thank you!