2021-01-31 13:48:10

by Alexander Popov

Subject: [PATCH 1/1] vsock: fix the race conditions in multi-transport support

There are multiple similar bugs implicitly introduced by the
commit c0cfa2d8a788fcf4 ("vsock: add multi-transports support") and
commit 6a2c0962105ae8ce ("vsock: prevent transport modules unloading").

The bug pattern:
[1] vsock_sock.transport pointer is copied to a local variable,
[2] lock_sock() is called,
[3] the local variable is used.
VSOCK multi-transport support introduced the race condition:
vsock_sock.transport value may change between [1] and [2].

Let's copy vsock_sock.transport pointer to local variables after
the lock_sock() call.

Signed-off-by: Alexander Popov <[email protected]>
---
net/vmw_vsock/af_vsock.c | 17 ++++++++++++-----
1 file changed, 12 insertions(+), 5 deletions(-)

diff --git a/net/vmw_vsock/af_vsock.c b/net/vmw_vsock/af_vsock.c
index d10916ab4526..28edac1f9aa6 100644
--- a/net/vmw_vsock/af_vsock.c
+++ b/net/vmw_vsock/af_vsock.c
@@ -997,9 +997,12 @@ static __poll_t vsock_poll(struct file *file, struct socket *sock,
mask |= EPOLLOUT | EPOLLWRNORM | EPOLLWRBAND;

} else if (sock->type == SOCK_STREAM) {
- const struct vsock_transport *transport = vsk->transport;
+ const struct vsock_transport *transport = NULL;
+
lock_sock(sk);

+ transport = vsk->transport;
+
/* Listening sockets that have connections in their accept
* queue can be read.
*/
@@ -1082,10 +1085,11 @@ static int vsock_dgram_sendmsg(struct socket *sock, struct msghdr *msg,
err = 0;
sk = sock->sk;
vsk = vsock_sk(sk);
- transport = vsk->transport;

lock_sock(sk);

+ transport = vsk->transport;
+
err = vsock_auto_bind(vsk);
if (err)
goto out;
@@ -1544,10 +1548,11 @@ static int vsock_stream_setsockopt(struct socket *sock,
err = 0;
sk = sock->sk;
vsk = vsock_sk(sk);
- transport = vsk->transport;

lock_sock(sk);

+ transport = vsk->transport;
+
switch (optname) {
case SO_VM_SOCKETS_BUFFER_SIZE:
COPY_IN(val);
@@ -1680,7 +1685,6 @@ static int vsock_stream_sendmsg(struct socket *sock, struct msghdr *msg,

sk = sock->sk;
vsk = vsock_sk(sk);
- transport = vsk->transport;
total_written = 0;
err = 0;

@@ -1689,6 +1693,8 @@ static int vsock_stream_sendmsg(struct socket *sock, struct msghdr *msg,

lock_sock(sk);

+ transport = vsk->transport;
+
/* Callers should not provide a destination with stream sockets. */
if (msg->msg_namelen) {
err = sk->sk_state == TCP_ESTABLISHED ? -EISCONN : -EOPNOTSUPP;
@@ -1823,11 +1829,12 @@ vsock_stream_recvmsg(struct socket *sock, struct msghdr *msg, size_t len,

sk = sock->sk;
vsk = vsock_sk(sk);
- transport = vsk->transport;
err = 0;

lock_sock(sk);

+ transport = vsk->transport;
+
if (!transport || sk->sk_state != TCP_ESTABLISHED) {
/* Recvmsg is supposed to return 0 if a peer performs an
* orderly shutdown. Differentiate between that case and when a
--
2.26.2

2021-01-31 20:16:27

by Linus Torvalds

Subject: Re: [PATCH 1/1] vsock: fix the race conditions in multi-transport support

[ I'm checking lkml for at least some of the emails that I'm cc'd on ]

On Sun, Jan 31, 2021 at 2:59 AM Alexander Popov <[email protected]> wrote:
>
> There are multiple similar bugs implicitly introduced by the
> commit [...]

Note: this got eaten or delayed by the mailing list issues that seem
to be plaguing lkml - I'm not seeing it on lore, although google does
find it on mail-archive.com.

The maintainers are cc'd, but it means - for example - that if
maintainers rely on patchwork, I thin kthat will be missing this email
too.

Linus

2021-02-01 08:31:23

by Stefano Garzarella

Subject: Re: [PATCH 1/1] vsock: fix the race conditions in multi-transport support

On Sun, Jan 31, 2021 at 01:59:14PM +0300, Alexander Popov wrote:
>There are multiple similar bugs implicitly introduced by the
>commit c0cfa2d8a788fcf4 ("vsock: add multi-transports support") and
>commit 6a2c0962105ae8ce ("vsock: prevent transport modules unloading").
>
>The bug pattern:
> [1] vsock_sock.transport pointer is copied to a local variable,
> [2] lock_sock() is called,
> [3] the local variable is used.
>VSOCK multi-transport support introduced the race condition:
>vsock_sock.transport value may change between [1] and [2].
>
>Let's copy vsock_sock.transport pointer to local variables after
>the lock_sock() call.

We can add:

Fixes: c0cfa2d8a788 ("vsock: add multi-transports support")

>
>Signed-off-by: Alexander Popov <[email protected]>
>---
> net/vmw_vsock/af_vsock.c | 17 ++++++++++++-----
> 1 file changed, 12 insertions(+), 5 deletions(-)
>
>diff --git a/net/vmw_vsock/af_vsock.c b/net/vmw_vsock/af_vsock.c
>index d10916ab4526..28edac1f9aa6 100644
>--- a/net/vmw_vsock/af_vsock.c
>+++ b/net/vmw_vsock/af_vsock.c
>@@ -997,9 +997,12 @@ static __poll_t vsock_poll(struct file *file, struct socket *sock,
> mask |= EPOLLOUT | EPOLLWRNORM | EPOLLWRBAND;
>
> } else if (sock->type == SOCK_STREAM) {
>- const struct vsock_transport *transport = vsk->transport;
>+ const struct vsock_transport *transport = NULL;

I think we can avoid initializing to NULL since we assign it shortly
after.

>+
> lock_sock(sk);
>
>+ transport = vsk->transport;
>+
> /* Listening sockets that have connections in their accept
> * queue can be read.
> */
>@@ -1082,10 +1085,11 @@ static int vsock_dgram_sendmsg(struct socket *sock, struct msghdr *msg,
> err = 0;
> sk = sock->sk;
> vsk = vsock_sk(sk);
>- transport = vsk->transport;
>
> lock_sock(sk);
>
>+ transport = vsk->transport;
>+
> err = vsock_auto_bind(vsk);
> if (err)
> goto out;
>@@ -1544,10 +1548,11 @@ static int vsock_stream_setsockopt(struct
>socket *sock,
> err = 0;
> sk = sock->sk;
> vsk = vsock_sk(sk);
>- transport = vsk->transport;
>
> lock_sock(sk);
>
>+ transport = vsk->transport;
>+
> switch (optname) {
> case SO_VM_SOCKETS_BUFFER_SIZE:
> COPY_IN(val);
>@@ -1680,7 +1685,6 @@ static int vsock_stream_sendmsg(struct socket *sock, struct msghdr *msg,
>
> sk = sock->sk;
> vsk = vsock_sk(sk);
>- transport = vsk->transport;
> total_written = 0;
> err = 0;
>
>@@ -1689,6 +1693,8 @@ static int vsock_stream_sendmsg(struct socket *sock, struct msghdr *msg,
>
> lock_sock(sk);
>
>+ transport = vsk->transport;
>+
> /* Callers should not provide a destination with stream sockets. */
> if (msg->msg_namelen) {
> err = sk->sk_state == TCP_ESTABLISHED ? -EISCONN : -EOPNOTSUPP;
>@@ -1823,11 +1829,12 @@ vsock_stream_recvmsg(struct socket *sock, struct msghdr *msg, size_t len,
>
> sk = sock->sk;
> vsk = vsock_sk(sk);
>- transport = vsk->transport;
> err = 0;
>
> lock_sock(sk);
>
>+ transport = vsk->transport;
>+
> if (!transport || sk->sk_state != TCP_ESTABLISHED) {
> /* Recvmsg is supposed to return 0 if a peer performs an
> * orderly shutdown. Differentiate between that case and when a
>--
>2.26.2
>

Thanks for fixing this issues. With the small changes applied:

Reviewed-by: Stefano Garzarella <[email protected]>

Thanks,
Stefano

2021-02-01 08:56:10

by Alexander Popov

Subject: Re: [PATCH 1/1] vsock: fix the race conditions in multi-transport support

On 01.02.2021 11:26, Stefano Garzarella wrote:
> On Sun, Jan 31, 2021 at 01:59:14PM +0300, Alexander Popov wrote:
>> There are multiple similar bugs implicitly introduced by the
>> commit c0cfa2d8a788fcf4 ("vsock: add multi-transports support") and
>> commit 6a2c0962105ae8ce ("vsock: prevent transport modules unloading").
>>
>> The bug pattern:
>> [1] vsock_sock.transport pointer is copied to a local variable,
>> [2] lock_sock() is called,
>> [3] the local variable is used.
>> VSOCK multi-transport support introduced the race condition:
>> vsock_sock.transport value may change between [1] and [2].
>>
>> Let's copy vsock_sock.transport pointer to local variables after
>> the lock_sock() call.
>
> We can add:
>
> Fixes: c0cfa2d8a788 ("vsock: add multi-transports support")
>
>>
>> Signed-off-by: Alexander Popov <[email protected]>
>> ---
>> net/vmw_vsock/af_vsock.c | 17 ++++++++++++-----
>> 1 file changed, 12 insertions(+), 5 deletions(-)
>>
>> diff --git a/net/vmw_vsock/af_vsock.c b/net/vmw_vsock/af_vsock.c
>> index d10916ab4526..28edac1f9aa6 100644
>> --- a/net/vmw_vsock/af_vsock.c
>> +++ b/net/vmw_vsock/af_vsock.c
>> @@ -997,9 +997,12 @@ static __poll_t vsock_poll(struct file *file, struct socket *sock,
>> mask |= EPOLLOUT | EPOLLWRNORM | EPOLLWRBAND;
>>
>> } else if (sock->type == SOCK_STREAM) {
>> - const struct vsock_transport *transport = vsk->transport;
>> + const struct vsock_transport *transport = NULL;
>
> I think we can avoid initializing to NULL since we assign it shortly
> after.
>
>> +
>> lock_sock(sk);
>>
>> + transport = vsk->transport;
>> +
>> /* Listening sockets that have connections in their accept
>> * queue can be read.
>> */
>> @@ -1082,10 +1085,11 @@ static int vsock_dgram_sendmsg(struct socket *sock, struct msghdr *msg,
>> err = 0;
>> sk = sock->sk;
>> vsk = vsock_sk(sk);
>> - transport = vsk->transport;
>>
>> lock_sock(sk);
>>
>> + transport = vsk->transport;
>> +
>> err = vsock_auto_bind(vsk);
>> if (err)
>> goto out;
>> @@ -1544,10 +1548,11 @@ static int vsock_stream_setsockopt(struct
>> socket *sock,
>> err = 0;
>> sk = sock->sk;
>> vsk = vsock_sk(sk);
>> - transport = vsk->transport;
>>
>> lock_sock(sk);
>>
>> + transport = vsk->transport;
>> +
>> switch (optname) {
>> case SO_VM_SOCKETS_BUFFER_SIZE:
>> COPY_IN(val);
>> @@ -1680,7 +1685,6 @@ static int vsock_stream_sendmsg(struct socket *sock, struct msghdr *msg,
>>
>> sk = sock->sk;
>> vsk = vsock_sk(sk);
>> - transport = vsk->transport;
>> total_written = 0;
>> err = 0;
>>
>> @@ -1689,6 +1693,8 @@ static int vsock_stream_sendmsg(struct socket *sock, struct msghdr *msg,
>>
>> lock_sock(sk);
>>
>> + transport = vsk->transport;
>> +
>> /* Callers should not provide a destination with stream sockets. */
>> if (msg->msg_namelen) {
>> err = sk->sk_state == TCP_ESTABLISHED ? -EISCONN : -EOPNOTSUPP;
>> @@ -1823,11 +1829,12 @@ vsock_stream_recvmsg(struct socket *sock, struct msghdr *msg, size_t len,
>>
>> sk = sock->sk;
>> vsk = vsock_sk(sk);
>> - transport = vsk->transport;
>> err = 0;
>>
>> lock_sock(sk);
>>
>> + transport = vsk->transport;
>> +
>> if (!transport || sk->sk_state != TCP_ESTABLISHED) {
>> /* Recvmsg is supposed to return 0 if a peer performs an
>> * orderly shutdown. Differentiate between that case and when a
>> --
>> 2.26.2
>>
>
> Thanks for fixing this issues. With the small changes applied:
>
> Reviewed-by: Stefano Garzarella <[email protected]>

Hello Stefano,

Thanks for the review.

I've just sent the v2.

Best regards,
Alexander