aboutgitcodebugslistschat
diff options
context:
space:
mode:
authorDavid Gibson <david@gibson.dropbear.id.au>2025-12-02 15:02:11 +1100
committerStefano Brivio <sbrivio@redhat.com>2025-12-02 23:07:30 +0100
commite77cbf06b99ef3795344b239dc28e6a67bbb1eea (patch)
treeff9412e2a650ab012adcc46cc56ecf4b9237edb4
parente6f6eb24b1945f3c271f0b49fd60b9944e85439d (diff)
downloadpasst-e77cbf06b99ef3795344b239dc28e6a67bbb1eea.tar
passt-e77cbf06b99ef3795344b239dc28e6a67bbb1eea.tar.gz
passt-e77cbf06b99ef3795344b239dc28e6a67bbb1eea.tar.bz2
passt-e77cbf06b99ef3795344b239dc28e6a67bbb1eea.tar.lz
passt-e77cbf06b99ef3795344b239dc28e6a67bbb1eea.tar.xz
passt-e77cbf06b99ef3795344b239dc28e6a67bbb1eea.tar.zst
passt-e77cbf06b99ef3795344b239dc28e6a67bbb1eea.zip
tcp, udp: Remove fallback if creating dual stack socket fails
To save kernel memory we try to use "dual stack" sockets which can listen for both IPv4 and IPv6 connections where possible. To support kernels which don't allow dual stack sockets, we fall back to creating individual sockets for IPv4 and IPv6. This fallback causes some mild ugliness now, and will cause more difficulty with upcoming improvements to the forwarding logic. I don't think we need the fallback on the following grounds: 1) The fallback was broken since inception: The fallback was triggered if pif_sock_l4() failed attempting to create the dual stack socket. But even if the kernel didn't support them, pif_sock_l4() would not report a failure. - Dual stack sockets are distinguished by having the IPV6_V6ONLY sockopt set to 0. However, until the last patch, we only called setsockopt() if we wanted to set this to 1, so there was no kernel operation which could fail for dual stack sockets - we'd silently create a IPv6 only socket instead. - Even if we did call the setsockopt(), we only printed a debug() message for failures, we didn't report it to the caller 2) Dual stack sockets are not just a Linux extension The dual stack socket interface is described in RFC3493, specifically section 3.7 and section 5.3. It is supported on BSD: https://man.freebsd.org/cgi/man.cgi?query=ip6 and on Windows: https://learn.microsoft.com/en-us/windows/win32/winsock/ipproto-ipv6-socket-options 3) Linux has supported dual stack sockets for over 20 years According to ipv6(7) the IPV6_V6ONLY socket option was introduced in Linux 2.6 and Linux 2.4.21 both from 2003. Signed-off-by: David Gibson <david@gibson.dropbear.id.au> Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
-rw-r--r--tcp.c43
-rw-r--r--udp.c56
2 files changed, 51 insertions, 48 deletions
diff --git a/tcp.c b/tcp.c
index ae87e12..06b7a1b 100644
--- a/tcp.c
+++ b/tcp.c
@@ -2601,42 +2601,49 @@ static int tcp_sock_init_one(const struct ctx *c, uint8_t pif,
}
/**
- * tcp_sock_init() - Create listening sockets for a given host ("inbound") port
+ * tcp_sock_init() - Create listening socket for a given host ("inbound") port
* @c: Execution context
* @pif: Interface to open the socket for (PIF_HOST or PIF_SPLICE)
* @addr: Pointer to address for binding, NULL if not configured
* @ifname: Name of interface to bind to, NULL if not configured
* @port: Port, host order
*
- * Return: 0 on (partial) success, negative error code on (complete) failure
+ * Return: 0 on success, negative error code on failure
*/
int tcp_sock_init(const struct ctx *c, uint8_t pif,
const union inany_addr *addr, const char *ifname,
in_port_t port)
{
- int r4 = FD_REF_MAX + 1, r6 = FD_REF_MAX + 1;
+ int s;
ASSERT(!c->no_tcp);
- if (!addr && c->ifi4 && c->ifi6)
- /* Attempt to get a dual stack socket */
- if (tcp_sock_init_one(c, pif, NULL, ifname, port) >= 0)
+ if (!c->ifi4) {
+ if (!addr)
+ /* Restrict to v6 only */
+ addr = &inany_any6;
+ else if (inany_v4(addr))
+ /* Nothing to do */
return 0;
+ }
+ if (!c->ifi6) {
+ if (!addr)
+ /* Restrict to v4 only */
+ addr = &inany_any4;
+ else if (!inany_v4(addr))
+ /* Nothing to do */
+ return 0;
+ }
- /* Otherwise create a socket per IP version */
- if ((!addr || inany_v4(addr)) && c->ifi4)
- r4 = tcp_sock_init_one(c, pif,
- addr ? addr : &inany_any4, ifname, port);
-
- if ((!addr || !inany_v4(addr)) && c->ifi6)
- r6 = tcp_sock_init_one(c, pif,
- addr ? addr : &inany_any6, ifname, port);
-
- if (IN_INTERVAL(0, FD_REF_MAX, r4) || IN_INTERVAL(0, FD_REF_MAX, r6))
- return 0;
+ s = tcp_sock_init_one(c, pif, addr, ifname, port);
+ if (s < 0)
+ return s;
+ if (s > FD_REF_MAX)
+ return -EIO;
- return r4 < 0 ? r4 : r6;
+ return 0;
}
+
/**
* tcp_ns_sock_init() - Init socket to listen for spliced outbound connections
* @c: Execution context
diff --git a/udp.c b/udp.c
index b3ce9c7..3d097fb 100644
--- a/udp.c
+++ b/udp.c
@@ -1102,14 +1102,14 @@ int udp_tap_handler(const struct ctx *c, uint8_t pif,
}
/**
- * udp_sock_init() - Initialise listening sockets for a given port
+ * udp_sock_init() - Initialise listening socket for a given port
* @c: Execution context
* @pif: Interface to open the socket for (PIF_HOST or PIF_SPLICE)
* @addr: Pointer to address for binding, NULL if not configured
* @ifname: Name of interface to bind to, NULL if not configured
* @port: Port, host order
*
- * Return: 0 on (partial) success, negative error code on (complete) failure
+ * Return: 0 on success, negative error code on failure
*/
int udp_sock_init(const struct ctx *c, uint8_t pif,
const union inany_addr *addr, const char *ifname,
@@ -1119,8 +1119,8 @@ int udp_sock_init(const struct ctx *c, uint8_t pif,
.pif = pif,
.port = port,
};
- int r4 = FD_REF_MAX + 1, r6 = FD_REF_MAX + 1;
int (*socks)[NUM_PORTS];
+ int s;
ASSERT(!c->no_udp);
ASSERT(pif_is_socket(pif));
@@ -1130,40 +1130,36 @@ int udp_sock_init(const struct ctx *c, uint8_t pif,
else
socks = udp_splice_ns;
- if (!addr && c->ifi4 && c->ifi6) {
- int s;
-
- /* Attempt to get a dual stack socket */
- s = pif_sock_l4(c, EPOLL_TYPE_UDP_LISTEN, PIF_HOST,
- NULL, ifname, port, uref.u32);
- socks[V4][port] = s < 0 ? -1 : s;
- socks[V6][port] = s < 0 ? -1 : s;
- if (IN_INTERVAL(0, FD_REF_MAX, s))
+ if (!c->ifi4) {
+ if (!addr)
+ /* Restrict to v6 only */
+ addr = &inany_any6;
+ else if (inany_v4(addr))
+ /* Nothing to do */
return 0;
}
-
- if ((!addr || inany_v4(addr)) && c->ifi4) {
- const union inany_addr *a = addr ? addr : &inany_any4;
-
- r4 = pif_sock_l4(c, EPOLL_TYPE_UDP_LISTEN, pif, a, ifname,
- port, uref.u32);
-
- socks[V4][port] = r4 < 0 ? -1 : r4;
+ if (!c->ifi6) {
+ if (!addr)
+ /* Restrict to v4 only */
+ addr = &inany_any4;
+ else if (!inany_v4(addr))
+ /* Nothing to do */
+ return 0;
}
- if ((!addr || !inany_v4(addr)) && c->ifi6) {
- const union inany_addr *a = addr ? addr : &inany_any6;
-
- r6 = pif_sock_l4(c, EPOLL_TYPE_UDP_LISTEN, pif, a, ifname,
- port, uref.u32);
-
- socks[V6][port] = r6 < 0 ? -1 : r6;
+ s = pif_sock_l4(c, EPOLL_TYPE_UDP_LISTEN, pif,
+ addr, ifname, port, uref.u32);
+ if (s > FD_REF_MAX) {
+ close(s);
+ s = -EIO;
}
- if (IN_INTERVAL(0, FD_REF_MAX, r4) || IN_INTERVAL(0, FD_REF_MAX, r6))
- return 0;
+ if (!addr || inany_v4(addr))
+ socks[V4][port] = s < 0 ? -1 : s;
+ if (!addr || !inany_v4(addr))
+ socks[V6][port] = s < 0 ? -1 : s;
- return r4 < 0 ? r4 : r6;
+ return s < 0 ? s : 0;
}
/**