2024-03-18 19:50:16

by Khazhy Kumykov

Subject: [PATCH 1/2] iscsi_tcp: do not bind sockets that already have extra callbacks

This attempts to avoid a situation where a misbehaving iscsi daemon
passes a socket for a different iSCSI connection to BIND_CONN - which
would result in infinite recursion and stack overflow. This will
also prevent passing *other* sockets which had sk_user_data overridden,
but that wouldn't have been safe anyways - since we throw away that
pointer anyways. This does not cover all hypothetical scenarios where we
pass bad sockets to BIND_CONN.

This also papers over a different bug - we allow a daemon to call
BIND_CONN twice for the same connection - which would result in, at the
least, failing to uninitialize/teardown the previous socket, which will
be addressed separately.

Signed-off-by: Khazhismel Kumykov <[email protected]>
---
drivers/scsi/iscsi_tcp.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/scsi/iscsi_tcp.c b/drivers/scsi/iscsi_tcp.c
index 8e14cea15f98..e8ed60b777c6 100644
--- a/drivers/scsi/iscsi_tcp.c
+++ b/drivers/scsi/iscsi_tcp.c
@@ -725,7 +725,7 @@ iscsi_sw_tcp_conn_bind(struct iscsi_cls_session *cls_session,
}

err = -EINVAL;
- if (!sk_is_tcp(sock->sk))
+ if (!sk_is_tcp(sock->sk) || sock->sk->sk_user_data)
goto free_socket;

err = iscsi_conn_bind(cls_session, cls_conn, is_leading);
--
2.44.0.291.gc1ea87d7ee-goog

2024-03-18 19:50:29

by Khazhy Kumykov

Subject: [PATCH 2/2] iscsi_tcp: disallow binding the same connection twice

iscsi_sw_tcp_conn_bind does not check or cleanup previously bound
sockets, nor should we allow binding the same connection twice.

Signed-off-by: Khazhismel Kumykov <[email protected]>
---
drivers/scsi/iscsi_tcp.c | 3 +++
1 file changed, 3 insertions(+)

diff --git a/drivers/scsi/iscsi_tcp.c b/drivers/scsi/iscsi_tcp.c
index e8ed60b777c6..8cf5dc203a82 100644
--- a/drivers/scsi/iscsi_tcp.c
+++ b/drivers/scsi/iscsi_tcp.c
@@ -716,6 +716,9 @@ iscsi_sw_tcp_conn_bind(struct iscsi_cls_session *cls_session,
struct socket *sock;
int err;

+ if (tcp_sw_conn->sock)
+ return -EINVAL;
+
/* lookup for existing socket */
sock = sockfd_lookup((int)transport_eph, &err);
if (!sock) {
--
2.44.0.291.gc1ea87d7ee-goog

2024-03-19 19:29:26

by Mike Christie

Subject: Re: [PATCH 1/2] iscsi_tcp: do not bind sockets that already have extra callbacks

On 3/18/24 2:49 PM, Khazhismel Kumykov wrote:
> This attempts to avoid a situation where a misbehaving iscsi daemon
> passes a socket for a different iSCSI connection to BIND_CONN - which
> would result in infinite recursion and stack overflow. This will
> also prevent passing *other* sockets which had sk_user_data overridden,
> but that wouldn't have been safe anyways - since we throw away that
> pointer anyways. This does not cover all hypothetical scenarios where we
> pass bad sockets to BIND_CONN.
>
> This also papers over a different bug - we allow a daemon to call
> BIND_CONN twice for the same connection - which would result in, at the
> least, failing to uninitialize/teardown the previous socket, which will
> be addressed separately.
>
> Signed-off-by: Khazhismel Kumykov <[email protected]>
> ---
> drivers/scsi/iscsi_tcp.c | 2 +-
> 1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/drivers/scsi/iscsi_tcp.c b/drivers/scsi/iscsi_tcp.c
> index 8e14cea15f98..e8ed60b777c6 100644
> --- a/drivers/scsi/iscsi_tcp.c
> +++ b/drivers/scsi/iscsi_tcp.c
> @@ -725,7 +725,7 @@ iscsi_sw_tcp_conn_bind(struct iscsi_cls_session *cls_session,
> }
>
> err = -EINVAL;
> - if (!sk_is_tcp(sock->sk))
> + if (!sk_is_tcp(sock->sk) || sock->sk->sk_user_data)
> goto free_socket;
>
> err = iscsi_conn_bind(cls_session, cls_conn, is_leading);


Reviewed-by: Mike Christie <[email protected]>

2024-03-19 19:34:28

by Mike Christie

Subject: Re: [PATCH 2/2] iscsi_tcp: disallow binding the same connection twice

On 3/18/24 2:49 PM, Khazhismel Kumykov wrote:
> iscsi_sw_tcp_conn_bind does not check or cleanup previously bound
> sockets, nor should we allow binding the same connection twice.
>

This looks like a problem for all the iscsi drivers.

I think you could:

1. Add a check for ISCSI_CONN_FLAG_BOUND in iscsi_conn_bind.
2. Have iscsi_sw_tcp_conn_stop do:

/* stop xmit side */
- iscsi_suspend_tx(conn);
+ iscsi_conn_unbind(cls_conn, true);

to clear the flag when we clean up the conn for relogin.

3. Fix up the other iscsi drivers so they call:

iscsi_conn_unbind(cls_conn, true);

in their failure paths so when they fail they clear ISCSI_CONN_FLAG_BOUND and
iscsi_conn_bind can be called on the retry.



> Signed-off-by: Khazhismel Kumykov <[email protected]>
> ---
> drivers/scsi/iscsi_tcp.c | 3 +++
> 1 file changed, 3 insertions(+)
>
> diff --git a/drivers/scsi/iscsi_tcp.c b/drivers/scsi/iscsi_tcp.c
> index e8ed60b777c6..8cf5dc203a82 100644
> --- a/drivers/scsi/iscsi_tcp.c
> +++ b/drivers/scsi/iscsi_tcp.c
> @@ -716,6 +716,9 @@ iscsi_sw_tcp_conn_bind(struct iscsi_cls_session *cls_session,
> struct socket *sock;
> int err;
>
> + if (tcp_sw_conn->sock)
> + return -EINVAL;
> +
> /* lookup for existing socket */
> sock = sockfd_lookup((int)transport_eph, &err);
> if (!sock) {