From 8c5727bd9b490283f4cdee3b44dae1911f36e5be Mon Sep 17 00:00:00 2001
From: David Gibson <david@gibson.dropbear.id.au>
Date: Wed, 19 Mar 2025 16:14:22 +1100
Subject: [PATCH] migrate, tcp: Migrate RFC 7323 timestamp
Currently our migration of the state of TCP sockets omits the RFC 7323
timestamp. In some circumstances that can result in data sent from the
target machine not being received, because it is discarded on the peer due
to PAWS checking.
Add code to dump and restore the timestamp across migration.
Link: https://bugs.passt.top/show_bug.cgi?id=115
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
[sbrivio: Minor style fixes]
Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
(cherry picked from commit cfb3740568ab291d7be00e457658c45ce9367ed5)
---
tcp.c | 59 ++++++++++++++++++++++++++++++++++++++++++++++++++++++
tcp_conn.h | 2 ++
2 files changed, 61 insertions(+)
diff --git a/tcp.c b/tcp.c
index 2c226ae..ac2b7d3 100644
--- a/tcp.c
+++ b/tcp.c
@@ -2857,6 +2857,57 @@ static int tcp_flow_dump_mss(int s, struct tcp_tap_transfer_ext *t)
return 0;
}
+
+/**
+ * tcp_flow_dump_timestamp() - Dump RFC 7323 timestamp via TCP_TIMESTAMP
+ * @conn: Pointer to the TCP connection structure
+ * @t: Extended migration data (tcpi_options must be populated)
+ *
+ * Return: 0 on success, negative error code on failure
+ */
+static int tcp_flow_dump_timestamp(const struct tcp_tap_conn *conn,
+ struct tcp_tap_transfer_ext *t)
+{
+ int val = 0;
+
+ if (t->tcpi_options & TCPI_OPT_TIMESTAMPS) {
+ socklen_t sl = sizeof(val);
+
+ if (getsockopt(conn->sock, SOL_TCP, TCP_TIMESTAMP, &val, &sl)) {
+ int rc = -errno;
+ flow_perror(conn, "Getting RFC 7323 timestamp");
+ return rc;
+ }
+ }
+
+ t->timestamp = (uint32_t)val;
+ return 0;
+}
+
+/**
+ * tcp_flow_repair_timestamp() - Restore RFC 7323 timestamp via TCP_TIMESTAMP
+ * @conn: Pointer to the TCP connection structure
+ * @t: Extended migration data
+ *
+ * Return: 0 on success, negative error code on failure
+ */
+static int tcp_flow_repair_timestamp(const struct tcp_tap_conn *conn,
+ const struct tcp_tap_transfer_ext *t)
+{
+ int val = (int)t->timestamp;
+
+ if (t->tcpi_options & TCPI_OPT_TIMESTAMPS) {
+ if (setsockopt(conn->sock, SOL_TCP, TCP_TIMESTAMP,
+ &val, sizeof(val))) {
+ int rc = -errno;
+ flow_perror(conn, "Setting RFC 7323 timestamp");
+ return rc;
+ }
+ }
+
+ return 0;
+}
+
/**
* tcp_flow_dump_wnd() - Dump current tcp_repair_window parameters
* @c: Execution context
@@ -3244,6 +3295,9 @@ int tcp_flow_migrate_source_ext(int fd, int fidx,
if ((rc = tcp_flow_dump_mss(s, t)))
goto fail;
+ if ((rc = tcp_flow_dump_timestamp(conn, t)))
+ goto fail;
+
if ((rc = tcp_flow_dump_wnd(s, t)))
goto fail;
@@ -3289,6 +3343,7 @@ int tcp_flow_migrate_source_ext(int fd, int fidx,
t->notsent = htonl(t->notsent);
t->rcvq = htonl(t->rcvq);
t->mss = htonl(t->mss);
+ t->timestamp = htonl(t->timestamp);
t->snd_wl1 = htonl(t->snd_wl1);
t->snd_wnd = htonl(t->snd_wnd);
@@ -3503,6 +3558,7 @@ int tcp_flow_migrate_target_ext(struct ctx *c, union flow *flow, int fd)
t.notsent = ntohl(t.notsent);
t.rcvq = ntohl(t.rcvq);
t.mss = ntohl(t.mss);
+ t.timestamp = ntohl(t.timestamp);
t.snd_wl1 = ntohl(t.snd_wl1);
t.snd_wnd = ntohl(t.snd_wnd);
@@ -3542,6 +3598,9 @@ int tcp_flow_migrate_target_ext(struct ctx *c, union flow *flow, int fd)
/* We weren't able to create the socket, discard flow */
goto fail;
+ if (tcp_flow_repair_timestamp(conn, &t))
+ goto fail;
+
if (tcp_flow_select_queue(s, TCP_SEND_QUEUE))
goto fail;
diff --git a/tcp_conn.h b/tcp_conn.h
index 42dff48..c6501f8 100644
--- a/tcp_conn.h
+++ b/tcp_conn.h
@@ -152,6 +152,7 @@ struct tcp_tap_transfer {
* @notsent: Part of pending send queue that wasn't sent out yet
* @rcvq: Length of pending receive queue
* @mss: Socket-side MSS clamp
+ * @timestamp: RFC 7323 timestamp
* @snd_wl1: Next sequence used in window probe (next sequence - 1)
* @snd_wnd: Socket-side sending window
* @max_window: Window clamp
@@ -171,6 +172,7 @@ struct tcp_tap_transfer_ext {
uint32_t rcvq;
uint32_t mss;
+ uint32_t timestamp;
/* We can't just use struct tcp_repair_window: we need network order */
uint32_t snd_wl1;