From: David Gibson <david@gibson.dropbear.id.au>
To: passt-dev@passt.top, Stefano Brivio <sbrivio@redhat.com>
Cc: David Gibson <david@gibson.dropbear.id.au>
Subject: [PATCH v2 12/22] tcp, tcp_splice: Helpers for getting sockets from the pools
Date: Tue, 6 Feb 2024 12:17:24 +1100 [thread overview]
Message-ID: <20240206011734.884138-13-david@gibson.dropbear.id.au> (raw)
In-Reply-To: <20240206011734.884138-1-david@gibson.dropbear.id.au>
We maintain pools of ready-to-connect sockets in both the original and
(for pasta) guest namespace to reduce latency when starting new TCP
connections. If we exhaust those pools we have to take a higher
latency path to get a new socket.
Currently we open-code that fallback in the places we need it. To
improve clarity encapsulate that into helper functions.
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
---
tcp.c | 29 ++++++++++++++++++++++++-----
tcp_conn.h | 2 +-
tcp_splice.c | 46 +++++++++++++++++++++++++---------------------
3 files changed, 50 insertions(+), 27 deletions(-)
diff --git a/tcp.c b/tcp.c
index e15b932f..c06d1cc4 100644
--- a/tcp.c
+++ b/tcp.c
@@ -1792,7 +1792,7 @@ int tcp_conn_pool_sock(int pool[])
*
* Return: socket number on success, negative code if socket creation failed
*/
-int tcp_conn_new_sock(const struct ctx *c, sa_family_t af)
+static int tcp_conn_new_sock(const struct ctx *c, sa_family_t af)
{
int s;
@@ -1811,6 +1811,27 @@ int tcp_conn_new_sock(const struct ctx *c, sa_family_t af)
return s;
}
+/**
+ * tcp_conn_sock() - Obtain a connectable socket in the host/init namespace
+ * @c: Execution context
+ * @af: Address family (AF_INET or AF_INET6)
+ *
+ * Return: Socket fd on success, -errno on failure
+ */
+int tcp_conn_sock(const struct ctx *c, sa_family_t af)
+{
+ int *pool = af == AF_INET6 ? init_sock_pool6 : init_sock_pool4;
+ int s;
+
+ if ((s = tcp_conn_pool_sock(pool)) >= 0)
+ return s;
+
+ /* If the pool is empty we just open a new one without refilling the
+ * pool to keep latency down.
+ */
+ return tcp_conn_new_sock(c, af);
+}
+
/**
* tcp_conn_tap_mss() - Get MSS value advertised by tap/guest
* @conn: Connection pointer
@@ -1909,7 +1930,6 @@ static void tcp_conn_from_tap(struct ctx *c, sa_family_t af,
const struct tcphdr *th, const char *opts,
size_t optlen, const struct timespec *now)
{
- int *pool = af == AF_INET6 ? init_sock_pool6 : init_sock_pool4;
struct sockaddr_in addr4 = {
.sin_family = AF_INET,
.sin_port = th->dest,
@@ -1931,9 +1951,8 @@ static void tcp_conn_from_tap(struct ctx *c, sa_family_t af,
if (!(flow = flow_alloc()))
return;
- if ((s = tcp_conn_pool_sock(pool)) < 0)
- if ((s = tcp_conn_new_sock(c, af)) < 0)
- goto cancel;
+ if ((s = tcp_conn_sock(c, af)) < 0)
+ goto cancel;
if (!c->no_map_gw) {
if (af == AF_INET && IN4_ARE_ADDR_EQUAL(daddr, &c->ip4.gw))
diff --git a/tcp_conn.h b/tcp_conn.h
index 20c7cb8b..e55edafe 100644
--- a/tcp_conn.h
+++ b/tcp_conn.h
@@ -159,7 +159,7 @@ bool tcp_flow_defer(union flow *flow);
bool tcp_splice_flow_defer(union flow *flow);
void tcp_splice_timer(const struct ctx *c, union flow *flow);
int tcp_conn_pool_sock(int pool[]);
-int tcp_conn_new_sock(const struct ctx *c, sa_family_t af);
+int tcp_conn_sock(const struct ctx *c, sa_family_t af);
void tcp_sock_refill_pool(const struct ctx *c, int pool[], sa_family_t af);
void tcp_splice_refill(const struct ctx *c);
diff --git a/tcp_splice.c b/tcp_splice.c
index 576fe9be..609f3242 100644
--- a/tcp_splice.c
+++ b/tcp_splice.c
@@ -90,7 +90,7 @@ static const char *tcp_splice_flag_str[] __attribute((__unused__)) = {
};
/* Forward declaration */
-static int tcp_sock_refill_ns(void *arg);
+static int tcp_conn_sock_ns(const struct ctx *c, sa_family_t af);
/**
* tcp_splice_conn_epoll_events() - epoll events masks for given state
@@ -380,36 +380,19 @@ static int tcp_splice_connect(const struct ctx *c, struct tcp_splice_conn *conn,
static int tcp_splice_new(const struct ctx *c, struct tcp_splice_conn *conn,
in_port_t port, uint8_t pif)
{
+ sa_family_t af = CONN_V6(conn) ? AF_INET6 : AF_INET;
int s = -1;
- /* If the pool is empty we take slightly different approaches
- * for init or ns sockets. For init sockets we just open a
- * new one without refilling the pool to keep latency down.
- * For ns sockets, we're going to incur the latency of
- * entering the ns anyway, so we might as well refill the
- * pool.
- */
if (pif == PIF_SPLICE) {
- int *p = CONN_V6(conn) ? init_sock_pool6 : init_sock_pool4;
- sa_family_t af = CONN_V6(conn) ? AF_INET6 : AF_INET;
-
port += c->tcp.fwd_out.delta[port];
- s = tcp_conn_pool_sock(p);
- if (s < 0)
- s = tcp_conn_new_sock(c, af);
+ s = tcp_conn_sock(c, af);
} else {
- int *p = CONN_V6(conn) ? ns_sock_pool6 : ns_sock_pool4;
-
ASSERT(pif == PIF_HOST);
port += c->tcp.fwd_in.delta[port];
- /* If pool is empty, refill it first */
- if (p[TCP_SOCK_POOL_SIZE-1] < 0)
- NS_CALL(tcp_sock_refill_ns, c);
-
- s = tcp_conn_pool_sock(p);
+ s = tcp_conn_sock_ns(c, af);
}
if (s < 0) {
@@ -709,6 +692,27 @@ static int tcp_sock_refill_ns(void *arg)
return 0;
}
+/**
+ * tcp_conn_sock_ns() - Obtain a connectable socket in the namespace
+ * @c: Execution context
+ * @af: Address family (AF_INET or AF_INET6)
+ *
+ * Return: Socket fd in the namespace on success, -errno on failure
+ */
+static int tcp_conn_sock_ns(const struct ctx *c, sa_family_t af)
+{
+ int *p = af == AF_INET6 ? ns_sock_pool6 : ns_sock_pool4;
+
+ /* If the pool is empty we have to incur the latency of entering the ns.
+ * Therefore, we might as well refill the whole pool while we're at it,
+ * which differs from tcp_conn_sock().
+ */
+ if (p[TCP_SOCK_POOL_SIZE-1] < 0)
+ NS_CALL(tcp_sock_refill_ns, c);
+
+ return tcp_conn_pool_sock(p);
+}
+
/**
* tcp_splice_refill() - Refill pools of resources needed for splicing
* @c: Execution context
--
@@ -90,7 +90,7 @@ static const char *tcp_splice_flag_str[] __attribute((__unused__)) = {
};
/* Forward declaration */
-static int tcp_sock_refill_ns(void *arg);
+static int tcp_conn_sock_ns(const struct ctx *c, sa_family_t af);
/**
* tcp_splice_conn_epoll_events() - epoll events masks for given state
@@ -380,36 +380,19 @@ static int tcp_splice_connect(const struct ctx *c, struct tcp_splice_conn *conn,
static int tcp_splice_new(const struct ctx *c, struct tcp_splice_conn *conn,
in_port_t port, uint8_t pif)
{
+ sa_family_t af = CONN_V6(conn) ? AF_INET6 : AF_INET;
int s = -1;
- /* If the pool is empty we take slightly different approaches
- * for init or ns sockets. For init sockets we just open a
- * new one without refilling the pool to keep latency down.
- * For ns sockets, we're going to incur the latency of
- * entering the ns anyway, so we might as well refill the
- * pool.
- */
if (pif == PIF_SPLICE) {
- int *p = CONN_V6(conn) ? init_sock_pool6 : init_sock_pool4;
- sa_family_t af = CONN_V6(conn) ? AF_INET6 : AF_INET;
-
port += c->tcp.fwd_out.delta[port];
- s = tcp_conn_pool_sock(p);
- if (s < 0)
- s = tcp_conn_new_sock(c, af);
+ s = tcp_conn_sock(c, af);
} else {
- int *p = CONN_V6(conn) ? ns_sock_pool6 : ns_sock_pool4;
-
ASSERT(pif == PIF_HOST);
port += c->tcp.fwd_in.delta[port];
- /* If pool is empty, refill it first */
- if (p[TCP_SOCK_POOL_SIZE-1] < 0)
- NS_CALL(tcp_sock_refill_ns, c);
-
- s = tcp_conn_pool_sock(p);
+ s = tcp_conn_sock_ns(c, af);
}
if (s < 0) {
@@ -709,6 +692,27 @@ static int tcp_sock_refill_ns(void *arg)
return 0;
}
+/**
+ * tcp_conn_sock_ns() - Obtain a connectable socket in the namespace
+ * @c: Execution context
+ * @af: Address family (AF_INET or AF_INET6)
+ *
+ * Return: Socket fd in the namespace on success, -errno on failure
+ */
+static int tcp_conn_sock_ns(const struct ctx *c, sa_family_t af)
+{
+ int *p = af == AF_INET6 ? ns_sock_pool6 : ns_sock_pool4;
+
+ /* If the pool is empty we have to incur the latency of entering the ns.
+ * Therefore, we might as well refill the whole pool while we're at it,
+ * which differs from tcp_conn_sock().
+ */
+ if (p[TCP_SOCK_POOL_SIZE-1] < 0)
+ NS_CALL(tcp_sock_refill_ns, c);
+
+ return tcp_conn_pool_sock(p);
+}
+
/**
* tcp_splice_refill() - Refill pools of resources needed for splicing
* @c: Execution context
--
2.43.0
next prev parent reply other threads:[~2024-02-06 1:17 UTC|newest]
Thread overview: 38+ messages / expand[flat|nested] mbox.gz Atom feed top
2024-02-06 1:17 [PATCH v2 00/22] More flow table preliminaries: address handling improvements David Gibson
2024-02-06 1:17 ` [PATCH v2 01/22] treewide: Use sa_family_t for address family variables David Gibson
2024-02-06 1:17 ` [PATCH v2 02/22] inany: Helper to test for various address types David Gibson
2024-02-18 20:58 ` Stefano Brivio
2024-02-19 1:48 ` David Gibson
2024-02-06 1:17 ` [PATCH v2 03/22] inany: Add inany_ntop() helper David Gibson
2024-02-06 1:17 ` [PATCH v2 04/22] inany: Provide more conveniently typed constants for special addresses David Gibson
2024-02-06 1:17 ` [PATCH v2 05/22] inany: Introduce union sockaddr_inany David Gibson
2024-02-06 1:17 ` [PATCH v2 06/22] util: Allow IN4_IS_* macros to operate on untyped addresses David Gibson
2024-02-06 1:17 ` [PATCH v2 07/22] tcp, udp: Don't precompute port remappings in epoll references David Gibson
2024-02-06 1:17 ` [PATCH v2 08/22] flow: Add helper to determine a flow's protocol David Gibson
2024-02-06 1:17 ` [PATCH v2 09/22] tcp_splice: Simplify clean up logic David Gibson
2024-02-18 20:59 ` Stefano Brivio
2024-02-19 1:50 ` David Gibson
2024-02-06 1:17 ` [PATCH v2 10/22] tcp_splice: Don't use flow_trace() before setting flow type David Gibson
2024-02-06 1:17 ` [PATCH v2 11/22] flow: Clarify flow entry life cycle, introduce uniform logging David Gibson
2024-02-18 21:00 ` Stefano Brivio
2024-02-19 1:58 ` David Gibson
2024-02-06 1:17 ` David Gibson [this message]
2024-02-18 21:00 ` [PATCH v2 12/22] tcp, tcp_splice: Helpers for getting sockets from the pools Stefano Brivio
2024-02-19 1:51 ` David Gibson
2024-02-06 1:17 ` [PATCH v2 13/22] tcp_splice: More specific variable names in new splice path David Gibson
2024-02-18 21:00 ` Stefano Brivio
2024-02-19 1:53 ` David Gibson
2024-02-06 1:17 ` [PATCH v2 14/22] tcp_splice: Merge tcp_splice_new() into its caller David Gibson
2024-02-06 1:17 ` [PATCH v2 15/22] tcp_splice: Make tcp_splice_connect() create its own sockets David Gibson
2024-02-06 1:17 ` [PATCH v2 16/22] tcp_splice: Improve error reporting on connect path David Gibson
2024-02-18 21:01 ` Stefano Brivio
2024-02-19 3:23 ` David Gibson
2024-02-06 1:17 ` [PATCH v2 17/22] tcp_splice: Improve logic deciding when to splice David Gibson
2024-02-06 1:17 ` [PATCH v2 18/22] tcp, tcp_splice: Parse listening socket epoll ref in tcp_listen_handler() David Gibson
2024-02-06 1:17 ` [PATCH v2 19/22] tcp: Validate TCP endpoint addresses David Gibson
2024-02-22 12:45 ` Stefano Brivio
2024-02-23 3:56 ` David Gibson
2024-02-06 1:17 ` [PATCH v2 20/22] tap: Disallow loopback addresses on tap interface David Gibson
2024-02-06 1:17 ` [PATCH v2 21/22] port_fwd: Fix copypasta error in port_fwd_scan_udp() comments David Gibson
2024-02-06 1:17 ` [PATCH v2 22/22] fwd: Rename port_fwd.[ch] and their contents David Gibson
2024-02-27 14:22 ` [PATCH v2 00/22] More flow table preliminaries: address handling improvements 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=20240206011734.884138-13-david@gibson.dropbear.id.au \
--to=david@gibson.dropbear.id.au \
--cc=passt-dev@passt.top \
--cc=sbrivio@redhat.com \
/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).