From: Stefano Brivio <sbrivio@redhat.com>
To: Laurent Vivier <lvivier@redhat.com>
Cc: passt-dev@passt.top, David Gibson <david@gibson.dropbear.id.au>
Subject: Re: [PATCH v6 1/8] tcp: extract buffer management from tcp_send_flag()
Date: Wed, 12 Jun 2024 23:22:10 +0200 [thread overview]
Message-ID: <20240612232210.3b1f975f@elisabeth> (raw)
In-Reply-To: <20240612154734.1044883-2-lvivier@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
next prev parent reply other threads:[~2024-06-12 21:23 UTC|newest]
Thread overview: 25+ messages / expand[flat|nested] mbox.gz Atom feed top
2024-06-12 15:47 [PATCH v6 0/8] Add vhost-user support to passt (part 2) Laurent Vivier
2024-06-12 15:47 ` [PATCH v6 1/8] tcp: extract buffer management from tcp_send_flag() Laurent Vivier
2024-06-12 21:22 ` Stefano Brivio [this message]
2024-06-13 6:07 ` Stefano Brivio
2024-06-13 8:24 ` Laurent Vivier
2024-06-13 10:14 ` Stefano Brivio
2024-06-13 10:22 ` Laurent Vivier
2024-06-13 10:49 ` Stefano Brivio
2024-06-13 10:58 ` Laurent Vivier
2024-06-13 7:31 ` Laurent Vivier
2024-06-13 9:35 ` Stefano Brivio
2024-06-13 14:36 ` David Gibson
2024-06-12 15:47 ` [PATCH v6 2/8] tcp: move buffers management functions to their own file Laurent Vivier
2024-06-12 15:54 ` Stefano Brivio
2024-06-12 15:47 ` [PATCH v6 3/8] tap: refactor packets handling functions Laurent Vivier
2024-06-12 15:52 ` Stefano Brivio
2024-06-12 16:00 ` Laurent Vivier
2024-06-12 15:47 ` [PATCH v6 4/8] udp: refactor UDP header update functions Laurent Vivier
2024-06-12 15:47 ` [PATCH v6 5/8] udp: rename udp_sock_handler() to udp_buf_sock_handler() Laurent Vivier
2024-06-12 15:47 ` [PATCH v6 6/8] vhost-user: compare mode MODE_PASTA and not MODE_PASST Laurent Vivier
2024-06-12 15:47 ` [PATCH v6 7/8] iov: remove iov_copy() Laurent Vivier
2024-06-12 15:47 ` [PATCH v6 8/8] tap: use in->buf_size rather than sizeof(pkt_buf) Laurent Vivier
2024-06-12 17:16 ` [PATCH v6 0/8] Add vhost-user support to passt (part 2) Stefano Brivio
2024-06-12 17:37 ` Stefano Brivio
2024-06-12 21:23 ` Stefano Brivio
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20240612232210.3b1f975f@elisabeth \
--to=sbrivio@redhat.com \
--cc=david@gibson.dropbear.id.au \
--cc=lvivier@redhat.com \
--cc=passt-dev@passt.top \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
Code repositories for project(s) associated with this public inbox
https://passt.top/passt
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for IMAP folder(s).