Re: [PATCH v6 1/8] tcp: extract buffer management from tcp_send_flag()

[Stefano Brivio <sbrivio@redhat.com>]

I couldn't find out why this patch breaks the the pasta_podman/bats
test, yet, that is:

not ok 19 [505] Single TCP port forwarding, IPv4, tap
# (from function `bail-now' in file test/podman/test/system/helpers.bash, line 235,
#  from function `assert' in file test/podman/test/system/helpers.bash, line 929,
#  from function `pasta_test_do' in file test/podman/test/system/505-networking-pasta.bats, line 239,
#  in test file test/podman/test/system/505-networking-pasta.bats, line 472)
#   `pasta_test_do' failed
#
# [22:54:18.306131353] $ test/podman/bin/podman rm -t 0 --all --force --ignore
#
# [22:54:18.367462243] $ test/podman/bin/podman ps --all --external --format {{.ID}} {{.Names}}
#
# [22:54:18.394935392] $ test/podman/bin/podman images --all --format {{.Repository}}:{{.Tag}} {{.ID}}
# [22:54:18.419773379] quay.io/libpod/testimage:20240123 1f6acd4c4a1d
#
# [22:54:19.246631856] $ test/podman/bin/podman info --format {{.Host.Pasta.Executable}}
# [22:54:20.084392405] /home/sbrivio/passt/pasta
#
# [22:54:20.167980222] $ test/podman/bin/podman run --net=pasta -p [88.198.0.164]:5727:5727/tcp quay.io/libpod/testimage:20240123 sh -c for port in $(seq 5727 5727); do                              socat -u TCP4-LISTEN:${port},bind=[88.198.0.164] STDOUT &                          done; wait
# [22:54:37.256040883] x2024/06/12 20:54:37 socat[3] E read(6, 0x7fe675cd6000, 8192): Connection reset by peer
# #/vvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvv
# #|     FAIL: Mismatch between data sent and received
# #| expected: = x
# #|   actual:   x2024/06/12 20:54:37 socat\[3\] E read\(6\, 0x7fe675cd6000\, 8192\): Connection reset by peer
# #\^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

meaning that the data transfer is actually fine, but we reset the
connection instead of an orderly shutdown.

I found a few issues while looking for that:

On Wed, 12 Jun 2024 17:47:27 +0200
Laurent Vivier <lvivier@redhat.com> wrote:

> This commit isolates the internal data structure management used for storing
> data (e.g., tcp4_l2_flags_iov[], tcp6_l2_flags_iov[], tcp4_flags_ip[],
> tcp4_flags[], ...) from the tcp_send_flag() function. The extracted
> functionality is relocated to a new function named tcp_fill_flag_header().
> 
> tcp_fill_flag_header() is now a generic function that accepts parameters such
> as struct tcphdr and a data pointer. tcp_send_flag() utilizes this parameter to
> pass memory pointers from tcp4_l2_flags_iov[] and tcp6_l2_flags_iov[].
> 
> This separation sets the stage for utilizing tcp_fill_flag_header() to
> set the memory provided by the guest via vhost-user in future developments.
> 
> Signed-off-by: Laurent Vivier <lvivier@redhat.com>
> ---
> 
> Notes:
>     v6:
>       - rename tcp_fill_flag_header() to tcp_prepare_flags()
>       - set optlen to 0 in tcp_prepare_flags()
>     
>     v5:
>       - use tcp_fill_flag_header() rather than tcp_fill_headers4() and
>         tcp_fill_headers6().
> 
>  tcp.c | 72 +++++++++++++++++++++++++++++++++++++++--------------------
>  1 file changed, 48 insertions(+), 24 deletions(-)
> 
> diff --git a/tcp.c b/tcp.c
> index dd8d46e08628..6800209d4122 100644
> --- a/tcp.c
> +++ b/tcp.c
> @@ -1567,24 +1567,25 @@ static void tcp_update_seqack_from_tap(const struct ctx *c,
>  }
>  
>  /**
> - * tcp_send_flag() - Send segment with flags to tap (no payload)
> + * tcp_prepare_flags() - Prepare header for flags-only segment (no payload)
>   * @c:		Execution context
>   * @conn:	Connection pointer
>   * @flags:	TCP flags: if not set, send segment only if ACK is due
> + * @th:		TCP header to update
> + * @data:	buffer to store TCP option
> + * @optlen:	size of the TCP option buffer (output parameter)
>   *
> - * Return: negative error code on connection reset, 0 otherwise
> + * Return: < 0 error code on connection reset,
> + *	     0 if there is no flag to send
> + *	     1 otherwise

This is often called with if (tcp_send_flag(...)) or
if (!tcp_send_flag(...)). Those need to be replaced with 
if (tcp_send_flag(...) < 0) or if (tcp_send_flag(...) >= 0) in the
callers.

>   */
> -static int tcp_send_flag(struct ctx *c, struct tcp_tap_conn *conn, int flags)
> +static int tcp_prepare_flags(struct ctx *c, struct tcp_tap_conn *conn,
> +			     int flags, struct tcphdr *th, char *data,
> +			     size_t *optlen)
>  {
> -	struct tcp_flags_t *payload;
>  	struct tcp_info tinfo = { 0 };
>  	socklen_t sl = sizeof(tinfo);
>  	int s = conn->sock;
> -	size_t optlen = 0;
> -	struct tcphdr *th;
> -	struct iovec *iov;
> -	size_t l4len;
> -	char *data;
>  
>  	if (SEQ_GE(conn->seq_ack_to_tap, conn->seq_from_tap) &&
>  	    !flags && conn->wnd_to_tap)
> @@ -1606,20 +1607,11 @@ static int tcp_send_flag(struct ctx *c, struct tcp_tap_conn *conn, int flags)
>  	if (!tcp_update_seqack_wnd(c, conn, flags, &tinfo) && !flags)
>  		return 0;
>  
> -	if (CONN_V4(conn))
> -		iov = tcp4_l2_flags_iov[tcp4_flags_used++];
> -	else
> -		iov = tcp6_l2_flags_iov[tcp6_flags_used++];
> -
> -	payload = iov[TCP_IOV_PAYLOAD].iov_base;
> -	th = &payload->th;
> -	data = payload->opts;
> -
>  	if (flags & SYN) {
>  		int mss;
>  
>  		/* Options: MSS, NOP and window scale (8 bytes) */
> -		optlen = OPT_MSS_LEN + 1 + OPT_WS_LEN;
> +		*optlen = OPT_MSS_LEN + 1 + OPT_WS_LEN;
>  
>  		*data++ = OPT_MSS;
>  		*data++ = OPT_MSS_LEN;
> @@ -1651,19 +1643,16 @@ static int tcp_send_flag(struct ctx *c, struct tcp_tap_conn *conn, int flags)
>  		*data++ = conn->ws_to_tap;
>  	} else if (!(flags & RST)) {
>  		flags |= ACK;
> +		*optlen = 0;

*optlen also needs to be set to 0 if (flags & RST), say:

	} else {
		*optlen = 0;

		if (!(flags & RST))
			flags |= ACK;
	}

>  	}
>  
> -	th->doff = (sizeof(*th) + optlen) / 4;
> +	th->doff = (sizeof(*th) + *optlen) / 4;
>  
>  	th->ack = !!(flags & ACK);
>  	th->rst = !!(flags & RST);
>  	th->syn = !!(flags & SYN);
>  	th->fin = !!(flags & FIN);
>  
> -	l4len = tcp_l2_buf_fill_headers(c, conn, iov, optlen, NULL,
> -					conn->seq_to_tap);
> -	iov[TCP_IOV_PAYLOAD].iov_len = l4len;
> -
>  	if (th->ack) {
>  		if (SEQ_GE(conn->seq_ack_to_tap, conn->seq_from_tap))
>  			conn_flag(c, conn, ~ACK_TO_TAP_DUE);
> @@ -1678,6 +1667,41 @@ static int tcp_send_flag(struct ctx *c, struct tcp_tap_conn *conn, int flags)
>  	if (th->fin || th->syn)
>  		conn->seq_to_tap++;
>  
> +	return 1;
> +}
> +
> +/**
> + * tcp_send_flag() - Send segment with flags to tap (no payload)
> + * @c:         Execution context
> + * @conn:      Connection pointer
> + * @flags:     TCP flags: if not set, send segment only if ACK is due
> + *
> + * Return: negative error code on connection reset, 0 otherwise
> + */
> +static int tcp_send_flag(struct ctx *c, struct tcp_tap_conn *conn, int flags)
> +{
> +	struct tcp_flags_t *payload;
> +	struct iovec *iov;
> +	size_t optlen;
> +	size_t l4len;
> +	int ret;
> +
> +	if (CONN_V4(conn))
> +		iov = tcp4_l2_flags_iov[tcp4_flags_used++];
> +	else
> +		iov = tcp6_l2_flags_iov[tcp6_flags_used++];

We increase the counters here, but we don't decrease them back if we
hit if (ret <= 0) return ret; later.

> +
> +	payload = iov[TCP_IOV_PAYLOAD].iov_base;
> +
> +	ret = tcp_prepare_flags(c, conn, flags, &payload->th,
> +				payload->opts, &optlen);
> +	if (ret <= 0)
> +		return ret;
> +
> +	l4len = tcp_l2_buf_fill_headers(c, conn, iov, optlen, NULL,
> +					conn->seq_to_tap);
> +	iov[TCP_IOV_PAYLOAD].iov_len = l4len;
> +
>  	if (flags & DUP_ACK) {
>  		struct iovec *dup_iov;
>  		int i;

Here's the diff I have so far, I'm not necessarily recommending any of
that, it was just a quick try:

diff --git a/tcp.c b/tcp.c
index 6800209..6bff6bc 100644
--- a/tcp.c
+++ b/tcp.c
@@ -1644,6 +1644,8 @@ static int tcp_prepare_flags(struct ctx *c, struct tcp_tap_conn *conn,
 	} else if (!(flags & RST)) {
 		flags |= ACK;
 		*optlen = 0;
+	} else {
+		*optlen = 0;
 	}
 
 	th->doff = (sizeof(*th) + *optlen) / 4;
@@ -1695,8 +1697,14 @@ static int tcp_send_flag(struct ctx *c, struct tcp_tap_conn *conn, int flags)
 
 	ret = tcp_prepare_flags(c, conn, flags, &payload->th,
 				payload->opts, &optlen);
-	if (ret <= 0)
+	if (ret <= 0) {
+		if (CONN_V4(conn))
+			tcp4_flags_used--;
+		else
+			tcp6_flags_used--;
+
 		return ret;
+	}
 
 	l4len = tcp_l2_buf_fill_headers(c, conn, iov, optlen, NULL,
 					conn->seq_to_tap);
@@ -1738,7 +1746,7 @@ static void tcp_rst_do(struct ctx *c, struct tcp_tap_conn *conn)
 	if (conn->events == CLOSED)
 		return;
 
-	if (!tcp_send_flag(c, conn, RST))
+	if (tcp_send_flag(c, conn, RST) >= 0)
 		conn_event(c, conn, CLOSED);
 }
 
@@ -2111,7 +2119,7 @@ static void tcp_conn_from_tap(struct ctx *c, sa_family_t af,
 	} else {
 		tcp_get_sndbuf(conn);
 
-		if (tcp_send_flag(c, conn, SYN | ACK))
+		if (tcp_send_flag(c, conn, SYN | ACK) < 0)
 			goto cancel;
 
 		conn_event(c, conn, TAP_SYN_ACK_SENT);
@@ -2282,9 +2290,11 @@ static int tcp_data_from_sock(struct ctx *c, struct tcp_tap_conn *conn)
 
 	if (!len) {
 		if ((conn->events & (SOCK_FIN_RCVD | TAP_FIN_SENT)) == SOCK_FIN_RCVD) {
-			if ((ret = tcp_send_flag(c, conn, FIN | ACK))) {
+			int rc = tcp_send_flag(c, conn, FIN | ACK);
+
+			if (rc < 0) {
 				tcp_rst(c, conn);
-				return ret;
+				return rc;
 			}
 
 			conn_event(c, conn, TAP_FIN_SENT);
@@ -2716,7 +2726,7 @@ static void tcp_connect_finish(struct ctx *c, struct tcp_tap_conn *conn)
 		return;
 	}
 
-	if (tcp_send_flag(c, conn, SYN | ACK))
+	if (tcp_send_flag(c, conn, SYN | ACK) < 0)
 		return;
 
 	conn_event(c, conn, TAP_SYN_ACK_SENT);

-- 
@@ -1644,6 +1644,8 @@ static int tcp_prepare_flags(struct ctx *c, struct tcp_tap_conn *conn,
 	} else if (!(flags & RST)) {
 		flags |= ACK;
 		*optlen = 0;
+	} else {
+		*optlen = 0;
 	}
 
 	th->doff = (sizeof(*th) + *optlen) / 4;
@@ -1695,8 +1697,14 @@ static int tcp_send_flag(struct ctx *c, struct tcp_tap_conn *conn, int flags)
 
 	ret = tcp_prepare_flags(c, conn, flags, &payload->th,
 				payload->opts, &optlen);
-	if (ret <= 0)
+	if (ret <= 0) {
+		if (CONN_V4(conn))
+			tcp4_flags_used--;
+		else
+			tcp6_flags_used--;
+
 		return ret;
+	}
 
 	l4len = tcp_l2_buf_fill_headers(c, conn, iov, optlen, NULL,
 					conn->seq_to_tap);
@@ -1738,7 +1746,7 @@ static void tcp_rst_do(struct ctx *c, struct tcp_tap_conn *conn)
 	if (conn->events == CLOSED)
 		return;
 
-	if (!tcp_send_flag(c, conn, RST))
+	if (tcp_send_flag(c, conn, RST) >= 0)
 		conn_event(c, conn, CLOSED);
 }
 
@@ -2111,7 +2119,7 @@ static void tcp_conn_from_tap(struct ctx *c, sa_family_t af,
 	} else {
 		tcp_get_sndbuf(conn);
 
-		if (tcp_send_flag(c, conn, SYN | ACK))
+		if (tcp_send_flag(c, conn, SYN | ACK) < 0)
 			goto cancel;
 
 		conn_event(c, conn, TAP_SYN_ACK_SENT);
@@ -2282,9 +2290,11 @@ static int tcp_data_from_sock(struct ctx *c, struct tcp_tap_conn *conn)
 
 	if (!len) {
 		if ((conn->events & (SOCK_FIN_RCVD | TAP_FIN_SENT)) == SOCK_FIN_RCVD) {
-			if ((ret = tcp_send_flag(c, conn, FIN | ACK))) {
+			int rc = tcp_send_flag(c, conn, FIN | ACK);
+
+			if (rc < 0) {
 				tcp_rst(c, conn);
-				return ret;
+				return rc;
 			}
 
 			conn_event(c, conn, TAP_FIN_SENT);
@@ -2716,7 +2726,7 @@ static void tcp_connect_finish(struct ctx *c, struct tcp_tap_conn *conn)
 		return;
 	}
 
-	if (tcp_send_flag(c, conn, SYN | ACK))
+	if (tcp_send_flag(c, conn, SYN | ACK) < 0)
 		return;
 
 	conn_event(c, conn, TAP_SYN_ACK_SENT);

-- 
Stefano