aboutsummaryrefslogtreecommitdiff
path: root/sys/net
diff options
context:
space:
mode:
Diffstat (limited to 'sys/net')
-rw-r--r--sys/net/if.c64
-rw-r--r--sys/net/if_bridge.c56
-rw-r--r--sys/net/if_bridgevar.h2
-rw-r--r--sys/net/if_clone.h2
-rw-r--r--sys/net/if_epair.c62
-rw-r--r--sys/net/if_ethersubr.c2
-rw-r--r--sys/net/if_ovpn.c6
-rw-r--r--sys/net/if_tap.h2
-rw-r--r--sys/net/if_tun.h2
-rw-r--r--sys/net/if_tuntap.c88
-rw-r--r--sys/net/if_var.h1
-rw-r--r--sys/net/iflib.c120
-rw-r--r--sys/net/pfvar.h24
13 files changed, 340 insertions, 91 deletions
diff --git a/sys/net/if.c b/sys/net/if.c
index 79c883fd4a0a..b6a798aa0fab 100644
--- a/sys/net/if.c
+++ b/sys/net/if.c
@@ -74,7 +74,6 @@
#include <vm/uma.h>
#include <net/bpf.h>
-#include <net/ethernet.h>
#include <net/if.h>
#include <net/if_arp.h>
#include <net/if_clone.h>
@@ -1102,6 +1101,7 @@ if_detach_internal(struct ifnet *ifp, bool vmove)
struct ifaddr *ifa;
int i;
struct domain *dp;
+ void *if_afdata[AF_MAX];
#ifdef VIMAGE
bool shutdown;
@@ -1225,15 +1225,30 @@ finish_vnet_shutdown:
IF_AFDATA_LOCK(ifp);
i = ifp->if_afdata_initialized;
ifp->if_afdata_initialized = 0;
+ if (i != 0) {
+ /*
+ * Defer the dom_ifdetach call.
+ */
+ _Static_assert(sizeof(if_afdata) == sizeof(ifp->if_afdata),
+ "array size mismatch");
+ memcpy(if_afdata, ifp->if_afdata, sizeof(if_afdata));
+ memset(ifp->if_afdata, 0, sizeof(ifp->if_afdata));
+ }
IF_AFDATA_UNLOCK(ifp);
if (i == 0)
return;
+ /*
+ * XXXZL: This net epoch wait is not necessary if we have done right.
+ * But if we do not, at least we can make a guarantee that threads those
+ * enter net epoch will see NULL address family dependent data,
+ * e.g. if_afdata[AF_INET6]. A clear NULL pointer derefence is much
+ * better than writing to freed memory.
+ */
+ NET_EPOCH_WAIT();
SLIST_FOREACH(dp, &domains, dom_next) {
- if (dp->dom_ifdetach && ifp->if_afdata[dp->dom_family]) {
- (*dp->dom_ifdetach)(ifp,
- ifp->if_afdata[dp->dom_family]);
- ifp->if_afdata[dp->dom_family] = NULL;
- }
+ if (dp->dom_ifdetach != NULL &&
+ if_afdata[dp->dom_family] != NULL)
+ (*dp->dom_ifdetach)(ifp, if_afdata[dp->dom_family]);
}
}
@@ -2589,16 +2604,7 @@ ifhwioctl(u_long cmd, struct ifnet *ifp, caddr_t data, struct thread *td)
* flip. They require special handling because in-kernel
* consumers may indepdently toggle them.
*/
- if ((ifp->if_flags ^ new_flags) & IFF_PPROMISC) {
- if (new_flags & IFF_PPROMISC)
- ifp->if_flags |= IFF_PROMISC;
- else if (ifp->if_pcount == 0)
- ifp->if_flags &= ~IFF_PROMISC;
- if (log_promisc_mode_change)
- if_printf(ifp, "permanently promiscuous mode %s\n",
- ((new_flags & IFF_PPROMISC) ?
- "enabled" : "disabled"));
- }
+ if_setppromisc(ifp, new_flags & IFF_PPROMISC);
if ((ifp->if_flags ^ new_flags) & IFF_PALLMULTI) {
if (new_flags & IFF_PALLMULTI)
ifp->if_flags |= IFF_ALLMULTI;
@@ -4456,6 +4462,32 @@ if_getmtu_family(const if_t ifp, int family)
return (ifp->if_mtu);
}
+void
+if_setppromisc(if_t ifp, bool ppromisc)
+{
+ int new_flags;
+
+ if (ppromisc)
+ new_flags = ifp->if_flags | IFF_PPROMISC;
+ else
+ new_flags = ifp->if_flags & ~IFF_PPROMISC;
+ if ((ifp->if_flags ^ new_flags) & IFF_PPROMISC) {
+ if (new_flags & IFF_PPROMISC)
+ new_flags |= IFF_PROMISC;
+ /*
+ * Only unset IFF_PROMISC if there are no more consumers of
+ * promiscuity, i.e. the ifp->if_pcount refcount is 0.
+ */
+ else if (ifp->if_pcount == 0)
+ new_flags &= ~IFF_PROMISC;
+ if (log_promisc_mode_change)
+ if_printf(ifp, "permanently promiscuous mode %s\n",
+ ((new_flags & IFF_PPROMISC) ?
+ "enabled" : "disabled"));
+ }
+ ifp->if_flags = new_flags;
+}
+
/*
* Methods for drivers to access interface unicast and multicast
* link level addresses. Driver shall not know 'struct ifaddr' neither
diff --git a/sys/net/if_bridge.c b/sys/net/if_bridge.c
index 66555fd1feb5..d7911a348d87 100644
--- a/sys/net/if_bridge.c
+++ b/sys/net/if_bridge.c
@@ -522,11 +522,11 @@ SYSCTL_BOOL(_net_link_bridge, OID_AUTO, log_mac_flap,
"Log MAC address port flapping");
/* allow IP addresses on bridge members */
-VNET_DEFINE_STATIC(bool, member_ifaddrs) = false;
+VNET_DEFINE_STATIC(bool, member_ifaddrs) = true;
#define V_member_ifaddrs VNET(member_ifaddrs)
SYSCTL_BOOL(_net_link_bridge, OID_AUTO, member_ifaddrs,
CTLFLAG_RW | CTLFLAG_VNET, &VNET_NAME(member_ifaddrs), false,
- "Allow layer 3 addresses on bridge members");
+ "Allow layer 3 addresses on bridge members (deprecated)");
static bool
bridge_member_ifaddrs(void)
@@ -1448,24 +1448,30 @@ bridge_ioctl_add(struct bridge_softc *sc, void *arg)
/*
* If member_ifaddrs is disabled, do not allow an interface with
- * assigned IP addresses to be added to a bridge.
+ * assigned IP addresses to be added to a bridge. Skip this check
+ * for gif interfaces, because the IP address assigned to a gif
+ * interface is separate from the bridge's Ethernet segment.
*/
- if (!V_member_ifaddrs) {
+ if (ifs->if_type != IFT_GIF) {
struct ifaddr *ifa;
CK_STAILQ_FOREACH(ifa, &ifs->if_addrhead, ifa_link) {
-#ifdef INET
- if (ifa->ifa_addr->sa_family == AF_INET)
- return (EXTERROR(EINVAL,
- "Member interface may not have "
- "an IPv4 address configured"));
-#endif
-#ifdef INET6
- if (ifa->ifa_addr->sa_family == AF_INET6)
+ if (ifa->ifa_addr->sa_family != AF_INET &&
+ ifa->ifa_addr->sa_family != AF_INET6)
+ continue;
+
+ if (V_member_ifaddrs) {
+ if_printf(sc->sc_ifp,
+ "WARNING: Adding member interface %s which "
+ "has an IP address assigned is deprecated "
+ "and will be unsupported in a future "
+ "release.\n", ifs->if_xname);
+ break;
+ } else {
return (EXTERROR(EINVAL,
"Member interface may not have "
- "an IPv6 address configured"));
-#endif
+ "an IP address assigned"));
+ }
}
}
@@ -2398,6 +2404,12 @@ bridge_enqueue(struct bridge_softc *sc, struct ifnet *dst_ifp, struct mbuf *m,
return (EINVAL);
}
+ /* Do VLAN filtering. */
+ if (!bridge_vfilter_out(bif, m)) {
+ m_freem(m);
+ return (0);
+ }
+
/* We may be sending a fragment so traverse the mbuf */
for (; m; m = m0) {
m0 = m->m_nextpkt;
@@ -2817,10 +2829,6 @@ bridge_forward(struct bridge_softc *sc, struct bridge_iflist *sbif,
if (sbif->bif_flags & dbif->bif_flags & IFBIF_PRIVATE)
goto drop;
- /* Do VLAN filtering. */
- if (!bridge_vfilter_out(dbif, m))
- goto drop;
-
if ((dbif->bif_flags & IFBIF_STP) &&
dbif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING)
goto drop;
@@ -3189,10 +3197,6 @@ bridge_broadcast(struct bridge_softc *sc, struct ifnet *src_if,
if (sbif && (sbif->bif_flags & dbif->bif_flags & IFBIF_PRIVATE))
continue;
- /* Do VLAN filtering. */
- if (!bridge_vfilter_out(dbif, m))
- continue;
-
if ((dbif->bif_flags & IFBIF_STP) &&
dbif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING)
continue;
@@ -3358,6 +3362,14 @@ bridge_vfilter_out(const struct bridge_iflist *dbif, const struct mbuf *m)
NET_EPOCH_ASSERT();
+ /*
+ * If the interface is in span mode, then bif_sc will be NULL.
+ * Since the purpose of span interfaces is to receive all frames,
+ * pass everything.
+ */
+ if (dbif->bif_sc == NULL)
+ return (true);
+
/* If VLAN filtering isn't enabled, pass everything. */
if ((dbif->bif_sc->sc_flags & IFBRF_VLANFILTER) == 0)
return (true);
diff --git a/sys/net/if_bridgevar.h b/sys/net/if_bridgevar.h
index b0f579f688ac..5ed8c19f3128 100644
--- a/sys/net/if_bridgevar.h
+++ b/sys/net/if_bridgevar.h
@@ -159,7 +159,7 @@ struct ifbreq {
uint32_t ifbr_addrexceeded; /* member if addr violations */
ether_vlanid_t ifbr_pvid; /* member if PVID */
uint16_t ifbr_vlanproto; /* member if VLAN protocol */
- uint8_t pad[32];
+ uint8_t pad[28];
};
/* BRDGGIFFLAGS, BRDGSIFFLAGS */
diff --git a/sys/net/if_clone.h b/sys/net/if_clone.h
index 5a74ffa1cc2f..d780e49af25f 100644
--- a/sys/net/if_clone.h
+++ b/sys/net/if_clone.h
@@ -153,7 +153,7 @@ int if_clone_destroy(const char *);
int if_clone_list(struct if_clonereq *);
void if_clone_restoregroup(struct ifnet *);
-/* The below interfaces are used only by epair(4). */
+/* The below interfaces are used only by epair(4) and tun(4)/tap(4). */
void if_clone_addif(struct if_clone *, struct ifnet *);
int if_clone_destroyif(struct if_clone *, struct ifnet *);
diff --git a/sys/net/if_epair.c b/sys/net/if_epair.c
index a213a84e17db..fbffa8f359a0 100644
--- a/sys/net/if_epair.c
+++ b/sys/net/if_epair.c
@@ -67,9 +67,9 @@
#include <net/if_var.h>
#include <net/if_clone.h>
#include <net/if_media.h>
-#include <net/if_var.h>
#include <net/if_private.h>
#include <net/if_types.h>
+#include <net/if_vlan_var.h>
#include <net/netisr.h>
#ifdef RSS
#include <net/rss_config.h>
@@ -435,6 +435,21 @@ epair_media_status(struct ifnet *ifp __unused, struct ifmediareq *imr)
imr->ifm_active = IFM_ETHER | IFM_10G_T | IFM_FDX;
}
+/*
+ * Update ifp->if_hwassist according to the current value of ifp->if_capenable.
+ */
+static void
+epair_caps_changed(struct ifnet *ifp)
+{
+ uint64_t hwassist = 0;
+
+ if (ifp->if_capenable & IFCAP_TXCSUM)
+ hwassist |= CSUM_IP_TCP | CSUM_IP_UDP;
+ if (ifp->if_capenable & IFCAP_TXCSUM_IPV6)
+ hwassist |= CSUM_IP6_TCP | CSUM_IP6_UDP;
+ ifp->if_hwassist = hwassist;
+}
+
static int
epair_ioctl(struct ifnet *ifp, u_long cmd, caddr_t data)
{
@@ -462,6 +477,44 @@ epair_ioctl(struct ifnet *ifp, u_long cmd, caddr_t data)
error = 0;
break;
+ case SIOCGIFCAP:
+ ifr->ifr_reqcap = ifp->if_capabilities;
+ ifr->ifr_curcap = ifp->if_capenable;
+ error = 0;
+ break;
+ case SIOCSIFCAP:
+ /*
+ * Enable/disable capabilities as requested, besides
+ * IFCAP_RXCSUM(_IPV6), which always remain enabled.
+ * Incoming packets may have the mbuf flag CSUM_DATA_VALID set.
+ * Without IFCAP_RXCSUM(_IPV6), this flag would have to be
+ * removed, which does not seem helpful.
+ */
+ ifp->if_capenable = ifr->ifr_reqcap | IFCAP_RXCSUM |
+ IFCAP_RXCSUM_IPV6;
+ epair_caps_changed(ifp);
+ /*
+ * If IFCAP_TXCSUM(_IPV6) has been changed, change it on the
+ * other epair interface as well.
+ * A bridge disables IFCAP_TXCSUM(_IPV6) when adding one epair
+ * interface if another interface in the bridge has it disabled.
+ * In that case this capability needs to be disabled on the
+ * other epair interface to avoid sending packets in the bridge
+ * that rely on this capability.
+ */
+ sc = ifp->if_softc;
+ if ((ifp->if_capenable ^ sc->oifp->if_capenable) &
+ (IFCAP_TXCSUM | IFCAP_TXCSUM_IPV6)) {
+ sc->oifp->if_capenable &=
+ ~(IFCAP_TXCSUM | IFCAP_TXCSUM_IPV6);
+ sc->oifp->if_capenable |= ifp->if_capenable &
+ (IFCAP_TXCSUM | IFCAP_TXCSUM_IPV6);
+ epair_caps_changed(sc->oifp);
+ }
+ VLAN_CAPABILITIES(ifp);
+ error = 0;
+ break;
+
default:
/* Let the common ethernet handler process this. */
error = ether_ioctl(ifp, cmd, data);
@@ -573,8 +626,11 @@ epair_setup_ifp(struct epair_softc *sc, char *name, int unit)
ifp->if_dname = epairname;
ifp->if_dunit = unit;
ifp->if_flags = IFF_BROADCAST | IFF_SIMPLEX | IFF_MULTICAST;
- ifp->if_capabilities = IFCAP_VLAN_MTU;
- ifp->if_capenable = IFCAP_VLAN_MTU;
+ ifp->if_capabilities = IFCAP_VLAN_MTU | IFCAP_TXCSUM |
+ IFCAP_TXCSUM_IPV6 | IFCAP_RXCSUM | IFCAP_RXCSUM_IPV6;
+ ifp->if_capenable = IFCAP_VLAN_MTU | IFCAP_TXCSUM |
+ IFCAP_TXCSUM_IPV6 | IFCAP_RXCSUM | IFCAP_RXCSUM_IPV6;
+ epair_caps_changed(ifp);
ifp->if_transmit = epair_transmit;
ifp->if_qflush = epair_qflush;
ifp->if_start = epair_start;
diff --git a/sys/net/if_ethersubr.c b/sys/net/if_ethersubr.c
index 3ae0c01c0efc..9c157bf3d3c2 100644
--- a/sys/net/if_ethersubr.c
+++ b/sys/net/if_ethersubr.c
@@ -695,7 +695,7 @@ ether_input_internal(struct ifnet *ifp, struct mbuf *m)
* seen by upper protocol layers.
*/
if (!ETHER_IS_MULTICAST(eh->ether_dhost) &&
- bcmp(IF_LLADDR(ifp), eh->ether_dhost, ETHER_ADDR_LEN) != 0)
+ memcmp(IF_LLADDR(ifp), eh->ether_dhost, ETHER_ADDR_LEN) != 0)
m->m_flags |= M_PROMISC;
}
diff --git a/sys/net/if_ovpn.c b/sys/net/if_ovpn.c
index fe015632f33e..1c18baac3417 100644
--- a/sys/net/if_ovpn.c
+++ b/sys/net/if_ovpn.c
@@ -904,9 +904,11 @@ ovpn_create_kkey_dir(struct ovpn_kkey_dir **kdirp,
kdir->cipher = cipher;
kdir->keylen = keylen;
kdir->tx_seq = 1;
- memcpy(kdir->key, key, keylen);
+ if (keylen != 0)
+ memcpy(kdir->key, key, keylen);
kdir->noncelen = ivlen;
- memcpy(kdir->nonce, iv, ivlen);
+ if (ivlen != 0)
+ memcpy(kdir->nonce, iv, ivlen);
if (kdir->cipher != OVPN_CIPHER_ALG_NONE) {
/* Crypto init */
diff --git a/sys/net/if_tap.h b/sys/net/if_tap.h
index d84cd2eba6f3..8297b8d9e3d2 100644
--- a/sys/net/if_tap.h
+++ b/sys/net/if_tap.h
@@ -57,6 +57,8 @@
#define TAPGIFNAME TUNGIFNAME
#define TAPSVNETHDR _IOW('t', 91, int)
#define TAPGVNETHDR _IOR('t', 94, int)
+#define TAPSTRANSIENT TUNSTRANSIENT
+#define TAPGTRANSIENT TUNGTRANSIENT
/* VMware ioctl's */
#define VMIO_SIOCSIFFLAGS _IOWINT('V', 0)
diff --git a/sys/net/if_tun.h b/sys/net/if_tun.h
index a8fb61db45a2..ccdc25944823 100644
--- a/sys/net/if_tun.h
+++ b/sys/net/if_tun.h
@@ -43,5 +43,7 @@ struct tuninfo {
#define TUNSIFPID _IO('t', 95)
#define TUNSIFHEAD _IOW('t', 96, int)
#define TUNGIFHEAD _IOR('t', 97, int)
+#define TUNSTRANSIENT _IOW('t', 98, int)
+#define TUNGTRANSIENT _IOR('t', 99, int)
#endif /* !_NET_IF_TUN_H_ */
diff --git a/sys/net/if_tuntap.c b/sys/net/if_tuntap.c
index 5e6f65c04b2f..c8dbb6aa8893 100644
--- a/sys/net/if_tuntap.c
+++ b/sys/net/if_tuntap.c
@@ -132,6 +132,7 @@ struct tuntap_softc {
#define TUN_DYING 0x0200
#define TUN_L2 0x0400
#define TUN_VMNET 0x0800
+#define TUN_TRANSIENT 0x1000
#define TUN_DRIVER_IDENT_MASK (TUN_L2 | TUN_VMNET)
#define TUN_READY (TUN_OPEN | TUN_INITED)
@@ -443,6 +444,18 @@ tuntap_name2info(const char *name, int *outunit, int *outflags)
return (0);
}
+static struct if_clone *
+tuntap_cloner_from_flags(int tun_flags)
+{
+
+ for (u_int i = 0; i < NDRV; i++)
+ if ((tun_flags & TUN_DRIVER_IDENT_MASK) ==
+ tuntap_drivers[i].ident_flags)
+ return (V_tuntap_driver_cloners[i]);
+
+ return (NULL);
+}
+
/*
* Get driver information from a set of flags specified. Masks the identifying
* part of the flags and compares it against all of the available
@@ -615,19 +628,39 @@ out:
CURVNET_RESTORE();
}
-static void
-tun_destroy(struct tuntap_softc *tp)
+static int
+tun_destroy(struct tuntap_softc *tp, bool may_intr)
{
+ int error;
TUN_LOCK(tp);
+
+ /*
+ * Transient tunnels may have set TUN_DYING if we're being destroyed as
+ * a result of the last close, which we'll allow.
+ */
+ MPASS((tp->tun_flags & (TUN_DYING | TUN_TRANSIENT)) != TUN_DYING);
tp->tun_flags |= TUN_DYING;
- if (tp->tun_busy != 0)
- cv_wait_unlock(&tp->tun_cv, &tp->tun_mtx);
- else
- TUN_UNLOCK(tp);
+ error = 0;
+ while (tp->tun_busy != 0) {
+ if (may_intr)
+ error = cv_wait_sig(&tp->tun_cv, &tp->tun_mtx);
+ else
+ cv_wait(&tp->tun_cv, &tp->tun_mtx);
+ if (error != 0) {
+ tp->tun_flags &= ~TUN_DYING;
+ TUN_UNLOCK(tp);
+ return (error);
+ }
+ }
+ TUN_UNLOCK(tp);
CURVNET_SET(TUN2IFP(tp)->if_vnet);
+ mtx_lock(&tunmtx);
+ TAILQ_REMOVE(&tunhead, tp, tun_list);
+ mtx_unlock(&tunmtx);
+
/* destroy_dev will take care of any alias. */
destroy_dev(tp->tun_dev);
seldrain(&tp->tun_rsel);
@@ -648,6 +681,8 @@ tun_destroy(struct tuntap_softc *tp)
cv_destroy(&tp->tun_cv);
free(tp, M_TUN);
CURVNET_RESTORE();
+
+ return (0);
}
static int
@@ -655,12 +690,7 @@ tun_clone_destroy(struct if_clone *ifc __unused, struct ifnet *ifp, uint32_t fla
{
struct tuntap_softc *tp = ifp->if_softc;
- mtx_lock(&tunmtx);
- TAILQ_REMOVE(&tunhead, tp, tun_list);
- mtx_unlock(&tunmtx);
- tun_destroy(tp);
-
- return (0);
+ return (tun_destroy(tp, true));
}
static void
@@ -702,9 +732,9 @@ tun_uninit(const void *unused __unused)
mtx_lock(&tunmtx);
while ((tp = TAILQ_FIRST(&tunhead)) != NULL) {
- TAILQ_REMOVE(&tunhead, tp, tun_list);
mtx_unlock(&tunmtx);
- tun_destroy(tp);
+ /* tun_destroy() will remove it from the tailq. */
+ tun_destroy(tp, false);
mtx_lock(&tunmtx);
}
mtx_unlock(&tunmtx);
@@ -1217,6 +1247,23 @@ out:
tun_vnethdr_set(ifp, 0);
tun_unbusy_locked(tp);
+ if ((tp->tun_flags & TUN_TRANSIENT) != 0) {
+ struct if_clone *cloner;
+ int error __diagused;
+
+ /* Mark it busy so that nothing can re-open it. */
+ tp->tun_flags |= TUN_DYING;
+ TUN_UNLOCK(tp);
+
+ CURVNET_SET_QUIET(ifp->if_home_vnet);
+ cloner = tuntap_cloner_from_flags(tp->tun_flags);
+ CURVNET_RESTORE();
+
+ error = if_clone_destroyif(cloner, ifp);
+ MPASS(error == 0 || error == EINTR || error == ERESTART);
+ return;
+ }
+
TUN_UNLOCK(tp);
}
@@ -1668,6 +1715,19 @@ tunioctl(struct cdev *dev, u_long cmd, caddr_t data, int flag,
case TUNGDEBUG:
*(int *)data = tundebug;
break;
+ case TUNSTRANSIENT:
+ TUN_LOCK(tp);
+ if (*(int *)data)
+ tp->tun_flags |= TUN_TRANSIENT;
+ else
+ tp->tun_flags &= ~TUN_TRANSIENT;
+ TUN_UNLOCK(tp);
+ break;
+ case TUNGTRANSIENT:
+ TUN_LOCK(tp);
+ *(int *)data = (tp->tun_flags & TUN_TRANSIENT) != 0;
+ TUN_UNLOCK(tp);
+ break;
case FIONBIO:
break;
case FIOASYNC:
diff --git a/sys/net/if_var.h b/sys/net/if_var.h
index 08435e7bd5f6..f2df612b19c1 100644
--- a/sys/net/if_var.h
+++ b/sys/net/if_var.h
@@ -622,6 +622,7 @@ int if_setmtu(if_t ifp, int mtu);
int if_getmtu(const if_t ifp);
int if_getmtu_family(const if_t ifp, int family);
void if_notifymtu(if_t ifp);
+void if_setppromisc(const if_t ifp, bool ppromisc);
int if_setflagbits(if_t ifp, int set, int clear);
int if_setflags(if_t ifp, int flags);
int if_getflags(const if_t ifp);
diff --git a/sys/net/iflib.c b/sys/net/iflib.c
index 2b43f6f19051..1e6d98291c04 100644
--- a/sys/net/iflib.c
+++ b/sys/net/iflib.c
@@ -142,6 +142,7 @@ struct iflib_ctx;
static void iru_init(if_rxd_update_t iru, iflib_rxq_t rxq, uint8_t flid);
static void iflib_timer(void *arg);
static void iflib_tqg_detach(if_ctx_t ctx);
+static int iflib_simple_transmit(if_t ifp, struct mbuf *m);
typedef struct iflib_filter_info {
driver_filter_t *ifi_filter;
@@ -198,6 +199,7 @@ struct iflib_ctx {
uint8_t ifc_sysctl_use_logical_cores;
uint16_t ifc_sysctl_extra_msix_vectors;
bool ifc_cpus_are_physical_cores;
+ bool ifc_sysctl_simple_tx;
qidx_t ifc_sysctl_ntxds[8];
qidx_t ifc_sysctl_nrxds[8];
@@ -710,7 +712,7 @@ static uint32_t iflib_txq_can_drain(struct ifmp_ring *);
static void iflib_altq_if_start(if_t ifp);
static int iflib_altq_if_transmit(if_t ifp, struct mbuf *m);
#endif
-static int iflib_register(if_ctx_t);
+static void iflib_register(if_ctx_t);
static void iflib_deregister(if_ctx_t);
static void iflib_unregister_vlan_handlers(if_ctx_t ctx);
static uint16_t iflib_get_mbuf_size_for(unsigned int size);
@@ -725,6 +727,7 @@ static void iflib_free_intr_mem(if_ctx_t ctx);
#ifndef __NO_STRICT_ALIGNMENT
static struct mbuf *iflib_fixup_rx(struct mbuf *m);
#endif
+static __inline int iflib_completed_tx_reclaim(iflib_txq_t txq, int thresh);
static SLIST_HEAD(cpu_offset_list, cpu_offset) cpu_offsets =
SLIST_HEAD_INITIALIZER(cpu_offsets);
@@ -2624,8 +2627,10 @@ iflib_stop(if_ctx_t ctx)
#endif /* DEV_NETMAP */
CALLOUT_UNLOCK(txq);
- /* clean any enqueued buffers */
- iflib_ifmp_purge(txq);
+ if (!ctx->ifc_sysctl_simple_tx) {
+ /* clean any enqueued buffers */
+ iflib_ifmp_purge(txq);
+ }
/* Free any existing tx buffers. */
for (j = 0; j < txq->ift_size; j++) {
iflib_txsd_free(ctx, txq, j);
@@ -3635,13 +3640,22 @@ defrag:
* cxgb
*/
if (__predict_false(nsegs + 2 > TXQ_AVAIL(txq))) {
- txq->ift_no_desc_avail++;
- bus_dmamap_unload(buf_tag, map);
- DBG_COUNTER_INC(encap_txq_avail_fail);
- DBG_COUNTER_INC(encap_txd_encap_fail);
- if ((txq->ift_task.gt_task.ta_flags & TASK_ENQUEUED) == 0)
- GROUPTASK_ENQUEUE(&txq->ift_task);
- return (ENOBUFS);
+ (void)iflib_completed_tx_reclaim(txq, RECLAIM_THRESH(ctx));
+ if (__predict_false(nsegs + 2 > TXQ_AVAIL(txq))) {
+ txq->ift_no_desc_avail++;
+ bus_dmamap_unload(buf_tag, map);
+ DBG_COUNTER_INC(encap_txq_avail_fail);
+ DBG_COUNTER_INC(encap_txd_encap_fail);
+ if (ctx->ifc_sysctl_simple_tx) {
+ *m_headp = m_head = iflib_remove_mbuf(txq);
+ m_freem(*m_headp);
+ DBG_COUNTER_INC(tx_frees);
+ *m_headp = NULL;
+ }
+ if ((txq->ift_task.gt_task.ta_flags & TASK_ENQUEUED) == 0)
+ GROUPTASK_ENQUEUE(&txq->ift_task);
+ return (ENOBUFS);
+ }
}
/*
* On Intel cards we can greatly reduce the number of TX interrupts
@@ -4014,6 +4028,12 @@ _task_fn_tx(void *context)
netmap_tx_irq(ifp, txq->ift_id))
goto skip_ifmp;
#endif
+ if (ctx->ifc_sysctl_simple_tx) {
+ mtx_lock(&txq->ift_mtx);
+ (void)iflib_completed_tx_reclaim(txq, RECLAIM_THRESH(ctx));
+ mtx_unlock(&txq->ift_mtx);
+ goto skip_ifmp;
+ }
#ifdef ALTQ
if (if_altq_is_enabled(ifp))
iflib_altq_if_start(ifp);
@@ -4027,9 +4047,8 @@ _task_fn_tx(void *context)
*/
if (abdicate)
ifmp_ring_check_drainage(txq->ift_br, TX_BATCH_SIZE);
-#ifdef DEV_NETMAP
+
skip_ifmp:
-#endif
if (ctx->ifc_flags & IFC_LEGACY)
IFDI_INTR_ENABLE(ctx);
else
@@ -4285,6 +4304,10 @@ iflib_if_transmit(if_t ifp, struct mbuf *m)
ifmp_ring_check_drainage(txq->ift_br, TX_BATCH_SIZE);
m_freem(m);
DBG_COUNTER_INC(tx_frees);
+ if (err == ENOBUFS)
+ if_inc_counter(ifp, IFCOUNTER_OQDROPS, 1);
+ else
+ if_inc_counter(ifp, IFCOUNTER_OERRORS, 1);
}
return (err);
@@ -5123,15 +5146,19 @@ iflib_device_register(device_t dev, void *sc, if_shared_ctx_t sctx, if_ctx_t *ct
ctx->ifc_dev = dev;
ctx->ifc_softc = sc;
- if ((err = iflib_register(ctx)) != 0) {
- device_printf(dev, "iflib_register failed %d\n", err);
- goto fail_ctx_free;
- }
+ iflib_register(ctx);
iflib_add_device_sysctl_pre(ctx);
scctx = &ctx->ifc_softc_ctx;
ifp = ctx->ifc_ifp;
-
+ if (ctx->ifc_sysctl_simple_tx) {
+#ifndef ALTQ
+ if_settransmitfn(ifp, iflib_simple_transmit);
+ device_printf(dev, "using simple if_transmit\n");
+#else
+ device_printf(dev, "ALTQ prevents using simple if_transmit\n");
+#endif
+ }
iflib_reset_qvalues(ctx);
IFNET_WLOCK();
CTX_LOCK(ctx);
@@ -5343,7 +5370,6 @@ iflib_device_register(device_t dev, void *sc, if_shared_ctx_t sctx, if_ctx_t *ct
DEBUGNET_SET(ctx->ifc_ifp, iflib);
- if_setgetcounterfn(ctx->ifc_ifp, iflib_if_get_counter);
iflib_add_device_sysctl_post(ctx);
iflib_add_pfil(ctx);
ctx->ifc_flags |= IFC_INIT_DONE;
@@ -5367,7 +5393,6 @@ fail_unlock:
CTX_UNLOCK(ctx);
IFNET_WUNLOCK();
iflib_deregister(ctx);
-fail_ctx_free:
device_set_softc(ctx->ifc_dev, NULL);
if (ctx->ifc_flags & IFC_SC_ALLOCATED)
free(ctx->ifc_softc, M_IFLIB);
@@ -5665,7 +5690,7 @@ _iflib_pre_assert(if_softc_ctx_t scctx)
MPASS(scctx->isc_txrx->ift_rxd_flush);
}
-static int
+static void
iflib_register(if_ctx_t ctx)
{
if_shared_ctx_t sctx = ctx->ifc_sctx;
@@ -5698,6 +5723,7 @@ iflib_register(if_ctx_t ctx)
if_settransmitfn(ifp, iflib_if_transmit);
#endif
if_setqflushfn(ifp, iflib_if_qflush);
+ if_setgetcounterfn(ifp, iflib_if_get_counter);
if_setflags(ifp, IFF_BROADCAST | IFF_SIMPLEX | IFF_MULTICAST);
ctx->ifc_vlan_attach_event =
EVENTHANDLER_REGISTER(vlan_config, iflib_vlan_register, ctx,
@@ -5711,7 +5737,6 @@ iflib_register(if_ctx_t ctx)
ifmedia_init(ctx->ifc_mediap, IFM_IMASK,
iflib_media_change, iflib_media_status);
}
- return (0);
}
static void
@@ -6766,6 +6791,9 @@ iflib_add_device_sysctl_pre(if_ctx_t ctx)
SYSCTL_ADD_CONST_STRING(ctx_list, oid_list, OID_AUTO, "driver_version",
CTLFLAG_RD, ctx->ifc_sctx->isc_driver_version, "driver version");
+ SYSCTL_ADD_BOOL(ctx_list, oid_list, OID_AUTO, "simple_tx",
+ CTLFLAG_RDTUN, &ctx->ifc_sysctl_simple_tx, 0,
+ "use simple tx ring");
SYSCTL_ADD_U16(ctx_list, oid_list, OID_AUTO, "override_ntxqs",
CTLFLAG_RWTUN, &ctx->ifc_sysctl_ntxqs, 0,
"# of txqs to use, 0 => use default #");
@@ -7088,3 +7116,53 @@ iflib_debugnet_poll(if_t ifp, int count)
return (0);
}
#endif /* DEBUGNET */
+
+
+static inline iflib_txq_t
+iflib_simple_select_queue(if_ctx_t ctx, struct mbuf *m)
+{
+ int qidx;
+
+ if ((NTXQSETS(ctx) > 1) && M_HASHTYPE_GET(m))
+ qidx = QIDX(ctx, m);
+ else
+ qidx = NTXQSETS(ctx) + FIRST_QSET(ctx) - 1;
+ return (&ctx->ifc_txqs[qidx]);
+}
+
+static int
+iflib_simple_transmit(if_t ifp, struct mbuf *m)
+{
+ if_ctx_t ctx;
+ iflib_txq_t txq;
+ int error;
+ int bytes_sent = 0, pkt_sent = 0, mcast_sent = 0;
+
+
+ ctx = if_getsoftc(ifp);
+ if ((if_getdrvflags(ifp) & (IFF_DRV_RUNNING | IFF_DRV_OACTIVE)) !=
+ IFF_DRV_RUNNING)
+ return (EBUSY);
+ txq = iflib_simple_select_queue(ctx, m);
+ mtx_lock(&txq->ift_mtx);
+ error = iflib_encap(txq, &m);
+ if (error == 0) {
+ pkt_sent++;
+ bytes_sent += m->m_pkthdr.len;
+ mcast_sent += !!(m->m_flags & M_MCAST);
+ (void)iflib_txd_db_check(txq, true);
+ } else {
+ if (error == ENOBUFS)
+ if_inc_counter(ifp, IFCOUNTER_OQDROPS, 1);
+ else
+ if_inc_counter(ifp, IFCOUNTER_OERRORS, 1);
+ }
+ (void)iflib_completed_tx_reclaim(txq, RECLAIM_THRESH(ctx));
+ mtx_unlock(&txq->ift_mtx);
+ if_inc_counter(ifp, IFCOUNTER_OBYTES, bytes_sent);
+ if_inc_counter(ifp, IFCOUNTER_OPACKETS, pkt_sent);
+ if (mcast_sent)
+ if_inc_counter(ifp, IFCOUNTER_OMCASTS, mcast_sent);
+
+ return (error);
+}
diff --git a/sys/net/pfvar.h b/sys/net/pfvar.h
index d6c13470f2eb..8b102f198de8 100644
--- a/sys/net/pfvar.h
+++ b/sys/net/pfvar.h
@@ -326,6 +326,7 @@ pf_counter_u64_zero(struct pf_counter_u64 *pfcu64)
_Static_assert(sizeof(time_t) == 4 || sizeof(time_t) == 8, "unexpected time_t size");
SYSCTL_DECL(_net_pf);
+MALLOC_DECLARE(M_PF);
MALLOC_DECLARE(M_PFHASH);
MALLOC_DECLARE(M_PF_RULE_ITEM);
@@ -645,6 +646,7 @@ struct pf_kpool {
int tblidx;
u_int16_t proxy_port[2];
u_int8_t opts;
+ sa_family_t ipv6_nexthop_af;
};
struct pf_rule_actions {
@@ -859,8 +861,8 @@ struct pf_krule {
u_int8_t keep_state;
sa_family_t af;
u_int8_t proto;
- u_int8_t type;
- u_int8_t code;
+ uint16_t type;
+ uint16_t code;
u_int8_t flags;
u_int8_t flagset;
u_int8_t min_ttl;
@@ -1749,6 +1751,7 @@ struct pf_kstatus {
counter_u64_t lcounters[KLCNT_MAX]; /* limit counters */
struct pf_counter_u64 fcounters[FCNT_MAX]; /* state operation counters */
counter_u64_t scounters[SCNT_MAX]; /* src_node operation counters */
+ counter_u64_t ncounters[NCNT_MAX];
uint32_t states;
uint32_t src_nodes;
uint32_t running;
@@ -2389,8 +2392,6 @@ extern u_int16_t pf_cksum_fixup(u_int16_t, u_int16_t, u_int16_t,
extern u_int16_t pf_proto_cksum_fixup(struct mbuf *, u_int16_t,
u_int16_t, u_int16_t, u_int8_t);
-VNET_DECLARE(struct ifnet *, sync_ifp);
-#define V_sync_ifp VNET(sync_ifp);
VNET_DECLARE(struct pf_krule, pf_default_rule);
#define V_pf_default_rule VNET(pf_default_rule)
extern void pf_addrcpy(struct pf_addr *, const struct pf_addr *,
@@ -2421,7 +2422,7 @@ int pf_multihome_scan_init(int, int, struct pf_pdesc *);
int pf_multihome_scan_asconf(int, int, struct pf_pdesc *);
u_int32_t pf_new_isn(struct pf_kstate *);
-void *pf_pull_hdr(const struct mbuf *, int, void *, int, u_short *, u_short *,
+void *pf_pull_hdr(const struct mbuf *, int, void *, int, u_short *,
sa_family_t);
void pf_change_a(void *, u_int16_t *, u_int32_t, u_int8_t);
void pf_change_proto_a(struct mbuf *, void *, u_int16_t *, u_int32_t,
@@ -2438,6 +2439,7 @@ int pf_match_port(u_int8_t, u_int16_t, u_int16_t, u_int16_t);
void pf_normalize_init(void);
void pf_normalize_cleanup(void);
+uint64_t pf_normalize_get_frag_count(void);
int pf_normalize_tcp(struct pf_pdesc *);
void pf_normalize_tcp_cleanup(struct pf_kstate *);
int pf_normalize_tcp_init(struct pf_pdesc *,
@@ -2541,22 +2543,23 @@ struct mbuf *pf_build_tcp(const struct pf_krule *, sa_family_t,
const struct pf_addr *, const struct pf_addr *,
u_int16_t, u_int16_t, u_int32_t, u_int32_t,
u_int8_t, u_int16_t, u_int16_t, u_int8_t, int,
- u_int16_t, u_int16_t, u_int, int);
+ u_int16_t, u_int16_t, u_int, int, u_short *);
void pf_send_tcp(const struct pf_krule *, sa_family_t,
const struct pf_addr *, const struct pf_addr *,
u_int16_t, u_int16_t, u_int32_t, u_int32_t,
u_int8_t, u_int16_t, u_int16_t, u_int8_t, int,
- u_int16_t, u_int16_t, int);
+ u_int16_t, u_int16_t, int, u_short *);
void pf_syncookies_init(void);
void pf_syncookies_cleanup(void);
int pf_get_syncookies(struct pfioc_nv *);
int pf_set_syncookies(struct pfioc_nv *);
int pf_synflood_check(struct pf_pdesc *);
-void pf_syncookie_send(struct pf_pdesc *);
+void pf_syncookie_send(struct pf_pdesc *, u_short *);
bool pf_syncookie_check(struct pf_pdesc *);
u_int8_t pf_syncookie_validate(struct pf_pdesc *);
-struct mbuf * pf_syncookie_recreate_syn(struct pf_pdesc *);
+struct mbuf * pf_syncookie_recreate_syn(struct pf_pdesc *,
+ u_short *);
VNET_DECLARE(struct pf_kstatus, pf_status);
#define V_pf_status VNET(pf_status)
@@ -2612,6 +2615,7 @@ struct pf_kruleset *pf_find_kruleset(const char *);
struct pf_kruleset *pf_get_leaf_kruleset(char *, char **);
struct pf_kruleset *pf_find_or_create_kruleset(const char *);
void pf_rs_initialize(void);
+void pf_rule_tree_free(struct pf_krule_global *);
struct pf_krule *pf_krule_alloc(void);
@@ -2680,7 +2684,7 @@ u_short pf_map_addr(sa_family_t, struct pf_krule *,
struct pf_addr *, struct pf_kpool *);
u_short pf_map_addr_sn(u_int8_t, struct pf_krule *,
struct pf_addr *, struct pf_addr *,
- sa_family_t *, struct pfi_kkif **nkif,
+ sa_family_t *, struct pfi_kkif **,
struct pf_addr *, struct pf_kpool *,
pf_sn_types_t);
int pf_get_transaddr_af(struct pf_krule *,
generated by cgit v1.2.3 (git 2.25.1) at 2025-09-20 13:37:47 +0000