Re: [PATCH net v2 1/2] vsock: Fix memory leak in vsock_connect()

[Stefano Garzarella]

On Sun, Aug 07, 2022 at 02:00:11AM -0700, Peilin Ye wrote:
> From: Peilin Ye <peilin.ye@xxxxxxxxxxxxx>
>
> An O_NONBLOCK vsock_connect() request may try to reschedule
> @connect_work.  Imagine the following sequence of vsock_connect()
> requests:
>
>  1. The 1st, non-blocking request schedules @connect_work, which will
>     expire after 200 jiffies.  Socket state is now SS_CONNECTING;
>
>  2. Later, the 2nd, blocking request gets interrupted by a signal after
>     a few jiffies while waiting for the connection to be established.
>     Socket state is back to SS_UNCONNECTED, but @connect_work is still
>     pending, and will expire after 100 jiffies.
>
>  3. Now, the 3rd, non-blocking request tries to schedule @connect_work
>     again.  Since @connect_work is already scheduled,
>     schedule_delayed_work() silently returns.  sock_hold() is called
>     twice, but sock_put() will only be called once in
>     vsock_connect_timeout(), causing a memory leak reported by syzbot:
>
>  BUG: memory leak
>  unreferenced object 0xffff88810ea56a40 (size 1232):
>    comm "syz-executor756", pid 3604, jiffies 4294947681 (age 12.350s)
>    hex dump (first 32 bytes):
>      00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00  ................
>      28 00 07 40 00 00 00 00 00 00 00 00 00 00 00 00  (..@............
>    backtrace:
>      [<ffffffff837c830e>] sk_prot_alloc+0x3e/0x1b0 net/core/sock.c:1930
>      [<ffffffff837cbe22>] sk_alloc+0x32/0x2e0 net/core/sock.c:1989
>      [<ffffffff842ccf68>] __vsock_create.constprop.0+0x38/0x320 net/vmw_vsock/af_vsock.c:734
>      [<ffffffff842ce8f1>] vsock_create+0xc1/0x2d0 net/vmw_vsock/af_vsock.c:2203
>      [<ffffffff837c0cbb>] __sock_create+0x1ab/0x2b0 net/socket.c:1468
>      [<ffffffff837c3acf>] sock_create net/socket.c:1519 [inline]
>      [<ffffffff837c3acf>] __sys_socket+0x6f/0x140 net/socket.c:1561
>      [<ffffffff837c3bba>] __do_sys_socket net/socket.c:1570 [inline]
>      [<ffffffff837c3bba>] __se_sys_socket net/socket.c:1568 [inline]
>      [<ffffffff837c3bba>] __x64_sys_socket+0x1a/0x20 net/socket.c:1568
>      [<ffffffff84512815>] do_syscall_x64 arch/x86/entry/common.c:50 [inline]
>      [<ffffffff84512815>] do_syscall_64+0x35/0x80 arch/x86/entry/common.c:80
>      [<ffffffff84600068>] entry_SYSCALL_64_after_hwframe+0x44/0xae
>  <...>
>
> Use mod_delayed_work() instead: if @connect_work is already scheduled,
> reschedule it, and undo sock_hold() to keep the reference count
> balanced.
>
> Reported-and-tested-by: syzbot+b03f55bf128f9a38f064@xxxxxxxxxxxxxxxxxxxxxxxxx
> Fixes: d021c344051a ("VSOCK: Introduce VM Sockets")
> Co-developed-by: Stefano Garzarella <sgarzare@xxxxxxxxxx>
> Signed-off-by: Stefano Garzarella <sgarzare@xxxxxxxxxx>
> Signed-off-by: Peilin Ye <peilin.ye@xxxxxxxxxxxxx>
> ---
> change since v1:
>  - merged with Stefano's patch [1]
>
> [1] https://gitlab.com/sgarzarella/linux/-/commit/2d0f0b9cbbb30d58fdcbca7c1a857fd8f3110d61
>
> Hi Stefano,
>
> About the Fixes: tag, [2] introduced @connect_work, but all it did was
> breaking @dwork into two and moving some INIT_DELAYED_WORK()'s, so I don't
> think [2] introduced this memory leak?
>
> Since [2] has already been backported to 4.9 and 4.14, I think we can
> Fixes: commit d021c344051a ("VSOCK: Introduce VM Sockets"), too, to make
> backporting easier?

Yep, I think it should be fine!

> [2] commit 455f05ecd2b2 ("vsock: split dwork to avoid reinitializations")
>
> Thanks,
> Peilin Ye
>
> net/vmw_vsock/af_vsock.c | 8 +++++++-
> 1 file changed, 7 insertions(+), 1 deletion(-)
>
> diff --git a/net/vmw_vsock/af_vsock.c b/net/vmw_vsock/af_vsock.c
> index f04abf662ec6..fe14f6cbca22 100644
> --- a/net/vmw_vsock/af_vsock.c
> +++ b/net/vmw_vsock/af_vsock.c
> @@ -1391,7 +1391,13 @@ static int vsock_connect(struct socket *sock, struct sockaddr *addr,
> 			 * timeout fires.
> 			 */
> 			sock_hold(sk);
> -			schedule_delayed_work(&vsk->connect_work, timeout);
> +
> +			/* If the timeout function is already scheduled,
> +			 * reschedule it, then ungrab the socket refcount to
> +			 * keep it balanced.
> +			 */
> +			if (mod_delayed_work(system_wq, &vsk->connect_work, timeout))
                            ^
Checkpatch warns here about line lenght.
If you have to re-send, please split it.

Anyway, the patch LGTM:

Reviewed-by: Stefano Garzarella <sgarzare@xxxxxxxxxx>

Thanks,
Stefano