From 939a050ad96c08ea7793994fa42818ec29bccedd Mon Sep 17 00:00:00 2001 From: Hiroki Sato Date: Wed, 1 Oct 2014 21:37:32 +0000 Subject: [PATCH] Virtualize lagg(4) cloner. This change fixes a panic when tearing down if_lagg(4) interfaces which were cloned in a vnet jail. Sysctl nodes which are dynamically generated for each cloned interface (net.link.lagg.N.*) have been removed, and use_flowid and flowid_shift ifconfig(8) parameters have been added instead. Flags and per-interface statistics counters are displayed in "ifconfig -v". CR: D842 --- sbin/ifconfig/ifconfig.8 | 45 ++++++- sbin/ifconfig/iflagg.c | 80 +++++++++++- share/man/man4/lagg.4 | 8 +- sys/net/ieee8023ad_lacp.c | 69 ++--------- sys/net/if_lagg.c | 247 ++++++++++++++++++++++++++------------ sys/net/if_lagg.h | 17 ++- 6 files changed, 316 insertions(+), 150 deletions(-) diff --git a/sbin/ifconfig/ifconfig.8 b/sbin/ifconfig/ifconfig.8 index 61ccf3ce5681..02766a142fea 100644 --- a/sbin/ifconfig/ifconfig.8 +++ b/sbin/ifconfig/ifconfig.8 @@ -28,7 +28,7 @@ .\" From: @(#)ifconfig.8 8.3 (Berkeley) 1/5/94 .\" $FreeBSD$ .\" -.Dd September 9, 2014 +.Dd October 1, 2014 .Dt IFCONFIG 8 .Os .Sh NAME @@ -679,7 +679,7 @@ Set a flag to enable Neighbor Unreachability Detection. Clear a flag .Cm nud . .It Cm no_prefer_iface -Set a flag to not honor rule 5 of source address selection in RFC 3484. +Set a flag to not honor rule 5 of source address selection in RFC 3484. In practice this means the address on the outgoing interface will not be preferred, effectively yielding the decision to the address selection policy table, configurable with @@ -2331,9 +2331,16 @@ Remove the interface named by from the aggregation interface. .It Cm laggproto Ar proto Set the aggregation protocol. -The default is failover. -The available options are failover, lacp, loadbalance, roundrobin, broadcast -and none. +The default is +.Li failover . +The available options are +.Li failover , +.Li lacp , +.Li loadbalance , +.Li roundrobin , +.Li broadcast +and +.Li none . .It Cm lagghash Ar option Ns Oo , Ns Ar option Oc Set the packet layers to hash for aggregation protocols which load balance. The default is @@ -2348,6 +2355,34 @@ src/dst address for IPv4 or IPv6. .It Cm l4 src/dst port for TCP/UDP/SCTP. .El +.It Cm use_flowid +Enable local hash computation for RSS hash on the interface. +The +.Li loadbalance +and +.Li lacp +modes will use the RSS hash from the network card if available +to avoid computing one, this may give poor traffic distribution +if the hash is invalid or uses less of the protocol header information. +.Cm use_flowid +disables use of RSS hash from the network card. +The default value can be set via the +.Va net.link.lagg.default_use_flowid +.Xr sysctl 8 +variable. +.Li 0 +means +.Dq disabled +and +.Li 1 +means +.Dq enabled . +.It Cm -use_flowid +Disable local hash computation for RSS hash on the interface. +.It Cm flowid_shift Ar number +Set a shift parameter for RSS local hash computation. +Hash is calculated by using flowid bits in a packet header mbuf +which are shifted by the number of this parameter. .El .Pp The following parameters are specific to IP tunnel interfaces, diff --git a/sbin/ifconfig/iflagg.c b/sbin/ifconfig/iflagg.c index 29b8574f4b1a..edb6121634b1 100644 --- a/sbin/ifconfig/iflagg.c +++ b/sbin/ifconfig/iflagg.c @@ -68,7 +68,7 @@ setlaggproto(const char *val, int d, int s, const struct afswtch *afp) bzero(&ra, sizeof(ra)); ra.ra_proto = LAGG_PROTO_MAX; - for (i = 0; i < (sizeof(lpr) / sizeof(lpr[0])); i++) { + for (i = 0; i < nitems(lpr); i++) { if (strcmp(val, lpr[i].lpr_name) == 0) { ra.ra_proto = lpr[i].lpr_proto; break; @@ -82,6 +82,48 @@ setlaggproto(const char *val, int d, int s, const struct afswtch *afp) err(1, "SIOCSLAGG"); } +static void +setlaggflowidshift(const char *val, int d, int s, const struct afswtch *afp) +{ + struct lagg_reqall ra; + + bzero(&ra, sizeof(ra)); + ra.ra_opts = LAGG_OPT_FLOWIDSHIFT; + strlcpy(ra.ra_ifname, name, sizeof(ra.ra_ifname)); + ra.ra_flowid_shift = (int)strtol(val, NULL, 10); + if (ra.ra_flowid_shift & ~LAGG_OPT_FLOWIDSHIFT_MASK) + errx(1, "Invalid flowid_shift option: %s", val); + + if (ioctl(s, SIOCSLAGG, &ra) != 0) + err(1, "SIOCSLAGG"); +} + +static void +setlaggsetopt(const char *val, int d, int s, const struct afswtch *afp) +{ + struct lagg_reqall ra; + + bzero(&ra, sizeof(ra)); + ra.ra_opts = d; + switch (ra.ra_opts) { + case LAGG_OPT_USE_FLOWID: + case -LAGG_OPT_USE_FLOWID: + case LAGG_OPT_LACP_STRICT: + case -LAGG_OPT_LACP_STRICT: + case LAGG_OPT_LACP_TXTEST: + case -LAGG_OPT_LACP_TXTEST: + case LAGG_OPT_LACP_RXTEST: + case -LAGG_OPT_LACP_RXTEST: + break; + default: + err(1, "Invalid lagg option"); + } + strlcpy(ra.ra_ifname, name, sizeof(ra.ra_ifname)); + + if (ioctl(s, SIOCSLAGG, &ra) != 0) + err(1, "SIOCSLAGG"); +} + static void setlagghash(const char *val, int d, int s, const struct afswtch *afp) { @@ -169,7 +211,7 @@ lagg_status(int s) if (ioctl(s, SIOCGLAGG, &ra) == 0) { lp = (struct lacp_opreq *)&ra.ra_lacpreq; - for (i = 0; i < (sizeof(lpr) / sizeof(lpr[0])); i++) { + for (i = 0; i < nitems(lpr); i++) { if (ra.ra_proto == lpr[i].lpr_proto) { proto = lpr[i].lpr_name; break; @@ -197,9 +239,28 @@ lagg_status(int s) if (isport) printf(" laggdev %s", rp.rp_ifname); putchar('\n'); - if (verbose && ra.ra_proto == LAGG_PROTO_LACP) - printf("\tlag id: %s\n", - lacp_format_peer(lp, "\n\t\t ")); + if (verbose) { + printf("\tlagg options:\n"); + printf("\t\tuse_flowid: %d\n", + (ra.ra_opts & LAGG_OPT_USE_FLOWID) ? 1 : 0); + printf("\t\tflowid_shift: %d\n", ra.ra_flowid_shift); + switch (ra.ra_proto) { + case LAGG_PROTO_LACP: + printf("\t\tlacp_strict: %d\n", + (ra.ra_opts & LAGG_OPT_LACP_STRICT) ? 1 : 0); + printf("\t\tlacp_rxtest: %d\n", + (ra.ra_opts & LAGG_OPT_LACP_RXTEST) ? 1 : 0); + printf("\t\tlacp_txtest: %d\n", + (ra.ra_opts & LAGG_OPT_LACP_TXTEST) ? 1 : 0); + } + printf("\tlagg statistics:\n"); + printf("\t\tactive ports: %d\n", ra.ra_active); + printf("\t\tflapping: %u\n", ra.ra_flapping); + if (ra.ra_proto == LAGG_PROTO_LACP) { + printf("\tlag id: %s\n", + lacp_format_peer(lp, "\n\t\t ")); + } + } for (i = 0; i < ra.ra_ports; i++) { lp = (struct lacp_opreq *)&rpbuf[i].rp_lacpreq; @@ -226,6 +287,15 @@ static struct cmd lagg_cmds[] = { DEF_CMD_ARG("-laggport", unsetlaggport), DEF_CMD_ARG("laggproto", setlaggproto), DEF_CMD_ARG("lagghash", setlagghash), + DEF_CMD("use_flowid", LAGG_OPT_USE_FLOWID, setlaggsetopt), + DEF_CMD("-use_flowid", -LAGG_OPT_USE_FLOWID, setlaggsetopt), + DEF_CMD("lacp_strict", LAGG_OPT_LACP_STRICT, setlaggsetopt), + DEF_CMD("-lacp_strict", -LAGG_OPT_LACP_STRICT, setlaggsetopt), + DEF_CMD("lacp_txtest", LAGG_OPT_LACP_TXTEST, setlaggsetopt), + DEF_CMD("-lacp_txtest", -LAGG_OPT_LACP_TXTEST, setlaggsetopt), + DEF_CMD("lacp_rxtest", LAGG_OPT_LACP_RXTEST, setlaggsetopt), + DEF_CMD("-lacp_rxtest", -LAGG_OPT_LACP_RXTEST, setlaggsetopt), + DEF_CMD_ARG("flowid_shift", setlaggflowidshift), }; static struct afswtch af_lagg = { .af_name = "af_lagg", diff --git a/share/man/man4/lagg.4 b/share/man/man4/lagg.4 index 3ea9247c39f9..1b4031593155 100644 --- a/share/man/man4/lagg.4 +++ b/share/man/man4/lagg.4 @@ -16,7 +16,7 @@ .\" .\" $FreeBSD$ .\" -.Dd February 23, 2012 +.Dd October 1, 2014 .Dt LAGG 4 .Os .Sh NAME @@ -143,9 +143,9 @@ modes will use the RSS hash from the network card if available to avoid computing one, this may give poor traffic distribution if the hash is invalid or uses less of the protocol header information. Local hash computation can be forced per interface by setting the -.Va net.link.lagg.X.use_flowid -.Xr sysctl 8 -variable to zero where X is the interface number. +.Cm use_flowid +.Xr ifconfig 8 +flag. The default for new interfaces is set via the .Va net.link.lagg.default_use_flowid .Xr sysctl 8 . diff --git a/sys/net/ieee8023ad_lacp.c b/sys/net/ieee8023ad_lacp.c index 90ed6e0ed76e..548e81b8baa8 100644 --- a/sys/net/ieee8023ad_lacp.c +++ b/sys/net/ieee8023ad_lacp.c @@ -190,14 +190,15 @@ static const char *lacp_format_portid(const struct lacp_portid *, char *, static void lacp_dprintf(const struct lacp_port *, const char *, ...) __attribute__((__format__(__printf__, 2, 3))); -static int lacp_debug = 0; +static VNET_DEFINE(int, lacp_debug); +#define V_lacp_debug VNET(lacp_debug) SYSCTL_NODE(_net_link_lagg, OID_AUTO, lacp, CTLFLAG_RD, 0, "ieee802.3ad"); -SYSCTL_INT(_net_link_lagg_lacp, OID_AUTO, debug, CTLFLAG_RWTUN, - &lacp_debug, 0, "Enable LACP debug logging (1=debug, 2=trace)"); +SYSCTL_INT(_net_link_lagg_lacp, OID_AUTO, debug, CTLFLAG_RWTUN | CTLFLAG_VNET, + &VNET_NAME(lacp_debug), 0, "Enable LACP debug logging (1=debug, 2=trace)"); -#define LACP_DPRINTF(a) if (lacp_debug & 0x01) { lacp_dprintf a ; } -#define LACP_TRACE(a) if (lacp_debug & 0x02) { lacp_dprintf(a,"%s\n",__func__); } -#define LACP_TPRINTF(a) if (lacp_debug & 0x04) { lacp_dprintf a ; } +#define LACP_DPRINTF(a) if (V_lacp_debug & 0x01) { lacp_dprintf a ; } +#define LACP_TRACE(a) if (V_lacp_debug & 0x02) { lacp_dprintf(a,"%s\n",__func__); } +#define LACP_TPRINTF(a) if (V_lacp_debug & 0x04) { lacp_dprintf a ; } /* * partner administration variables. @@ -300,7 +301,7 @@ lacp_pdu_input(struct lacp_port *lp, struct mbuf *m) goto bad; } - if (lacp_debug > 0) { + if (V_lacp_debug > 0) { lacp_dprintf(lp, "lacpdu receive\n"); lacp_dump_lacpdu(du); } @@ -385,7 +386,7 @@ lacp_xmit_lacpdu(struct lacp_port *lp) sizeof(du->ldu_collector)); du->ldu_collector.lci_maxdelay = 0; - if (lacp_debug > 0) { + if (V_lacp_debug > 0) { lacp_dprintf(lp, "lacpdu transmit\n"); lacp_dump_lacpdu(du); } @@ -497,12 +498,14 @@ lacp_tick(void *arg) if ((lp->lp_state & LACP_STATE_AGGREGATION) == 0) continue; + CURVNET_SET(lp->lp_ifp->if_vnet); lacp_run_timers(lp); lacp_select(lp); lacp_sm_mux(lp); lacp_sm_tx(lp); lacp_sm_ptx_tx_schedule(lp); + CURVNET_RESTORE(); } callout_reset(&lsc->lsc_callout, hz, lacp_tick, lsc); } @@ -747,48 +750,10 @@ lacp_transit_expire(void *vp) lsc->lsc_suppress_distributing = FALSE; } -static void -lacp_attach_sysctl(struct lacp_softc *lsc, struct sysctl_oid *p_oid) -{ - struct lagg_softc *sc = lsc->lsc_softc; - - SYSCTL_ADD_UINT(&sc->ctx, SYSCTL_CHILDREN(p_oid), OID_AUTO, - "lacp_strict_mode", - CTLFLAG_RW, - &lsc->lsc_strict_mode, - lsc->lsc_strict_mode, - "Enable LACP strict mode"); -} - -static void -lacp_attach_sysctl_debug(struct lacp_softc *lsc, struct sysctl_oid *p_oid) -{ - struct lagg_softc *sc = lsc->lsc_softc; - struct sysctl_oid *oid; - - /* Create a child of the parent lagg interface */ - oid = SYSCTL_ADD_NODE(&sc->ctx, SYSCTL_CHILDREN(p_oid), - OID_AUTO, "debug", CTLFLAG_RD, NULL, "DEBUG"); - - SYSCTL_ADD_UINT(&sc->ctx, SYSCTL_CHILDREN(oid), OID_AUTO, - "rx_test", - CTLFLAG_RW, - &lsc->lsc_debug.lsc_rx_test, - lsc->lsc_debug.lsc_rx_test, - "Bitmap of if_dunit entries to drop RX frames for"); - SYSCTL_ADD_UINT(&sc->ctx, SYSCTL_CHILDREN(oid), OID_AUTO, - "tx_test", - CTLFLAG_RW, - &lsc->lsc_debug.lsc_tx_test, - lsc->lsc_debug.lsc_tx_test, - "Bitmap of if_dunit entries to drop TX frames for"); -} - void lacp_attach(struct lagg_softc *sc) { struct lacp_softc *lsc; - struct sysctl_oid *oid; lsc = malloc(sizeof(struct lacp_softc), M_DEVBUF, M_WAITOK | M_ZERO); @@ -802,14 +767,6 @@ lacp_attach(struct lagg_softc *sc) TAILQ_INIT(&lsc->lsc_aggregators); LIST_INIT(&lsc->lsc_ports); - /* Create a child of the parent lagg interface */ - oid = SYSCTL_ADD_NODE(&sc->ctx, SYSCTL_CHILDREN(sc->sc_oid), - OID_AUTO, "lacp", CTLFLAG_RD, NULL, "LACP"); - - /* Attach sysctl nodes */ - lacp_attach_sysctl(lsc, oid); - lacp_attach_sysctl_debug(lsc, oid); - callout_init_mtx(&lsc->lsc_transit_callout, &lsc->lsc_mtx, 0); callout_init_mtx(&lsc->lsc_callout, &lsc->lsc_mtx, 0); @@ -875,7 +832,7 @@ lacp_select_tx_port(struct lagg_softc *sc, struct mbuf *m) return (NULL); } - if (sc->use_flowid && (m->m_flags & M_FLOWID)) + if ((sc->sc_opts & LAGG_OPT_USE_FLOWID) && (m->m_flags & M_FLOWID)) hash = m->m_pkthdr.flowid >> sc->flowid_shift; else hash = lagg_hashmbuf(sc, m, lsc->lsc_hashkey); @@ -1374,7 +1331,7 @@ lacp_sm_mux(struct lacp_port *lp) enum lacp_selected selected = lp->lp_selected; struct lacp_aggregator *la; - if (lacp_debug > 1) + if (V_lacp_debug > 1) lacp_dprintf(lp, "%s: state= 0x%x, selected= 0x%x, " "p_sync= 0x%x, p_collecting= 0x%x\n", __func__, lp->lp_mux_state, selected, p_sync, p_collecting); diff --git a/sys/net/if_lagg.c b/sys/net/if_lagg.c index 634d64cde139..3de40edad046 100644 --- a/sys/net/if_lagg.c +++ b/sys/net/if_lagg.c @@ -51,6 +51,7 @@ __FBSDID("$FreeBSD$"); #include #include #include +#include #if defined(INET) || defined(INET6) #include @@ -81,13 +82,21 @@ static struct { {0, NULL} }; -SLIST_HEAD(__trhead, lagg_softc) lagg_list; /* list of laggs */ -static struct mtx lagg_list_mtx; +VNET_DEFINE(SLIST_HEAD(__trhead, lagg_softc), lagg_list); /* list of laggs */ +#define V_lagg_list VNET(lagg_list) +static VNET_DEFINE(struct mtx, lagg_list_mtx); +#define V_lagg_list_mtx VNET(lagg_list_mtx) +#define LAGG_LIST_LOCK_INIT(x) mtx_init(&V_lagg_list_mtx, \ + "if_lagg list", NULL, MTX_DEF) +#define LAGG_LIST_LOCK_DESTROY(x) mtx_destroy(&V_lagg_list_mtx) +#define LAGG_LIST_LOCK(x) mtx_lock(&V_lagg_list_mtx) +#define LAGG_LIST_UNLOCK(x) mtx_unlock(&V_lagg_list_mtx) eventhandler_tag lagg_detach_cookie = NULL; static int lagg_clone_create(struct if_clone *, int, caddr_t); static void lagg_clone_destroy(struct ifnet *); -static struct if_clone *lagg_cloner; +static VNET_DEFINE(struct if_clone *, lagg_cloner); +#define V_lagg_cloner VNET(lagg_cloner) static const char laggname[] = "lagg"; static void lagg_lladdr(struct lagg_softc *, uint8_t *); @@ -123,7 +132,6 @@ static void lagg_media_status(struct ifnet *, struct ifmediareq *); static struct lagg_port *lagg_link_active(struct lagg_softc *, struct lagg_port *); static const void *lagg_gethdr(struct mbuf *, u_int, u_int, void *); -static int lagg_sysctl_active(SYSCTL_HANDLER_ARGS); /* Simple round robin */ static void lagg_rr_attach(struct lagg_softc *); @@ -232,29 +240,55 @@ SYSCTL_DECL(_net_link); SYSCTL_NODE(_net_link, OID_AUTO, lagg, CTLFLAG_RW, 0, "Link Aggregation"); -static int lagg_failover_rx_all = 0; /* Allow input on any failover links */ -SYSCTL_INT(_net_link_lagg, OID_AUTO, failover_rx_all, CTLFLAG_RW, - &lagg_failover_rx_all, 0, +/* Allow input on any failover links */ +static VNET_DEFINE(int, lagg_failover_rx_all); +#define V_lagg_failover_rx_all VNET(lagg_failover_rx_all) +SYSCTL_INT(_net_link_lagg, OID_AUTO, failover_rx_all, CTLFLAG_RW | CTLFLAG_VNET, + &VNET_NAME(lagg_failover_rx_all), 0, "Accept input from any interface in a failover lagg"); -static int def_use_flowid = 1; /* Default value for using M_FLOWID */ + +/* Default value for using M_FLOWID */ +static VNET_DEFINE(int, def_use_flowid) = 1; +#define V_def_use_flowid VNET(def_use_flowid) SYSCTL_INT(_net_link_lagg, OID_AUTO, default_use_flowid, CTLFLAG_RWTUN, - &def_use_flowid, 0, + &VNET_NAME(def_use_flowid), 0, "Default setting for using flow id for load sharing"); -static int def_flowid_shift = 16; /* Default value for using M_FLOWID */ + +/* Default value for using M_FLOWID */ +static VNET_DEFINE(int, def_flowid_shift) = 16; +#define V_def_flowid_shift VNET(def_flowid_shift) SYSCTL_INT(_net_link_lagg, OID_AUTO, default_flowid_shift, CTLFLAG_RWTUN, - &def_flowid_shift, 0, + &VNET_NAME(def_flowid_shift), 0, "Default setting for flowid shift for load sharing"); +static void +vnet_lagg_init(const void *unused __unused) +{ + + LAGG_LIST_LOCK_INIT(); + SLIST_INIT(&V_lagg_list); + V_lagg_cloner = if_clone_simple(laggname, lagg_clone_create, + lagg_clone_destroy, 0); +} +VNET_SYSINIT(vnet_lagg_init, SI_SUB_PROTO_IFATTACHDOMAIN, SI_ORDER_ANY, + vnet_lagg_init, NULL); + +static void +vnet_lagg_uninit(const void *unused __unused) +{ + + if_clone_detach(V_lagg_cloner); + LAGG_LIST_LOCK_DESTROY(); +} +VNET_SYSUNINIT(vnet_lagg_uninit, SI_SUB_PROTO_IFATTACHDOMAIN, SI_ORDER_ANY, + vnet_lagg_uninit, NULL); + static int lagg_modevent(module_t mod, int type, void *data) { switch (type) { case MOD_LOAD: - mtx_init(&lagg_list_mtx, "if_lagg list", NULL, MTX_DEF); - SLIST_INIT(&lagg_list); - lagg_cloner = if_clone_simple(laggname, lagg_clone_create, - lagg_clone_destroy, 0); lagg_input_p = lagg_input; lagg_linkstate_p = lagg_port_state; lagg_detach_cookie = EVENTHANDLER_REGISTER( @@ -264,10 +298,8 @@ lagg_modevent(module_t mod, int type, void *data) case MOD_UNLOAD: EVENTHANDLER_DEREGISTER(ifnet_departure_event, lagg_detach_cookie); - if_clone_detach(lagg_cloner); lagg_input_p = NULL; lagg_linkstate_p = NULL; - mtx_destroy(&lagg_list_mtx); break; default: return (EOPNOTSUPP); @@ -445,8 +477,6 @@ lagg_clone_create(struct if_clone *ifc, int unit, caddr_t params) struct lagg_softc *sc; struct ifnet *ifp; static const u_char eaddr[6]; /* 00:00:00:00:00:00 */ - struct sysctl_oid *oid; - char num[14]; /* sufficient for 32 bits */ sc = malloc(sizeof(*sc), M_DEVBUF, M_WAITOK|M_ZERO); ifp = sc->sc_ifp = if_alloc(IFT_ETHER); @@ -455,29 +485,10 @@ lagg_clone_create(struct if_clone *ifc, int unit, caddr_t params) return (ENOSPC); } - sysctl_ctx_init(&sc->ctx); - snprintf(num, sizeof(num), "%u", unit); - sc->use_flowid = def_use_flowid; - sc->flowid_shift = def_flowid_shift; - sc->sc_oid = oid = SYSCTL_ADD_NODE(&sc->ctx, - &SYSCTL_NODE_CHILDREN(_net_link, lagg), - OID_AUTO, num, CTLFLAG_RD, NULL, ""); - SYSCTL_ADD_INT(&sc->ctx, SYSCTL_CHILDREN(oid), OID_AUTO, - "use_flowid", CTLTYPE_INT|CTLFLAG_RW, &sc->use_flowid, - sc->use_flowid, "Use flow id for load sharing"); - SYSCTL_ADD_INT(&sc->ctx, SYSCTL_CHILDREN(oid), OID_AUTO, - "flowid_shift", CTLTYPE_INT|CTLFLAG_RW, &sc->flowid_shift, - sc->flowid_shift, - "Shift flowid bits to prevent multiqueue collisions"); - SYSCTL_ADD_INT(&sc->ctx, SYSCTL_CHILDREN(oid), OID_AUTO, - "count", CTLTYPE_INT|CTLFLAG_RD, &sc->sc_count, sc->sc_count, - "Total number of ports"); - SYSCTL_ADD_PROC(&sc->ctx, SYSCTL_CHILDREN(oid), OID_AUTO, - "active", CTLTYPE_INT|CTLFLAG_RD, sc, 0, lagg_sysctl_active, - "I", "Total number of active ports"); - SYSCTL_ADD_INT(&sc->ctx, SYSCTL_CHILDREN(oid), OID_AUTO, - "flapping", CTLTYPE_INT|CTLFLAG_RD, &sc->sc_flapping, - sc->sc_flapping, "Total number of port change events"); + if (V_def_use_flowid) + sc->sc_opts |= LAGG_OPT_USE_FLOWID; + sc->flowid_shift = V_def_flowid_shift; + /* Hash all layers by default */ sc->sc_flags = LAGG_F_HASHL2|LAGG_F_HASHL3|LAGG_F_HASHL4; @@ -515,9 +526,9 @@ lagg_clone_create(struct if_clone *ifc, int unit, caddr_t params) lagg_unregister_vlan, sc, EVENTHANDLER_PRI_FIRST); /* Insert into the global list of laggs */ - mtx_lock(&lagg_list_mtx); - SLIST_INSERT_HEAD(&lagg_list, sc, sc_entries); - mtx_unlock(&lagg_list_mtx); + LAGG_LIST_LOCK(); + SLIST_INSERT_HEAD(&V_lagg_list, sc, sc_entries); + LAGG_LIST_UNLOCK(); return (0); } @@ -542,14 +553,13 @@ lagg_clone_destroy(struct ifnet *ifp) /* Unhook the aggregation protocol */ lagg_proto_detach(sc); - sysctl_ctx_free(&sc->ctx); ifmedia_removeall(&sc->sc_media); ether_ifdetach(ifp); if_free(ifp); - mtx_lock(&lagg_list_mtx); - SLIST_REMOVE(&lagg_list, sc, lagg_softc, sc_entries); - mtx_unlock(&lagg_list_mtx); + LAGG_LIST_LOCK(); + SLIST_REMOVE(&V_lagg_list, sc, lagg_softc, sc_entries); + LAGG_LIST_UNLOCK(); taskqueue_drain(taskqueue_swi, &sc->sc_lladdr_task); LAGG_LOCK_DESTROY(sc); @@ -752,10 +762,10 @@ lagg_port_create(struct lagg_softc *sc, struct ifnet *ifp) return (ENOMEM); /* Check if port is a stacked lagg */ - mtx_lock(&lagg_list_mtx); - SLIST_FOREACH(sc_ptr, &lagg_list, sc_entries) { + LAGG_LIST_LOCK(); + SLIST_FOREACH(sc_ptr, &V_lagg_list, sc_entries) { if (ifp == sc_ptr->sc_ifp) { - mtx_unlock(&lagg_list_mtx); + LAGG_LIST_UNLOCK(); free(lp, M_DEVBUF); return (EINVAL); /* XXX disable stacking for the moment, its untested */ @@ -763,14 +773,14 @@ lagg_port_create(struct lagg_softc *sc, struct ifnet *ifp) lp->lp_flags |= LAGG_PORT_STACK; if (lagg_port_checkstacking(sc_ptr) >= LAGG_MAX_STACKING) { - mtx_unlock(&lagg_list_mtx); + LAGG_LIST_UNLOCK(); free(lp, M_DEVBUF); return (E2BIG); } #endif } } - mtx_unlock(&lagg_list_mtx); + LAGG_LIST_UNLOCK(); /* Change the interface type */ lp->lp_iftype = ifp->if_type; @@ -1205,6 +1215,31 @@ lagg_ioctl(struct ifnet *ifp, u_long cmd, caddr_t data) LAGG_RLOCK(sc, &tracker); ra->ra_proto = sc->sc_proto; lagg_proto_request(sc, &ra->ra_psc); + ra->ra_opts = sc->sc_opts; + if (sc->sc_proto == LAGG_PROTO_LACP) { + struct lacp_softc *lsc; + + lsc = (struct lacp_softc *)sc->sc_psc; + if (lsc->lsc_debug.lsc_tx_test != 0) + ra->ra_opts |= LAGG_OPT_LACP_TXTEST; + if (lsc->lsc_debug.lsc_rx_test != 0) + ra->ra_opts |= LAGG_OPT_LACP_RXTEST; + if (lsc->lsc_strict_mode != 0) + ra->ra_opts |= LAGG_OPT_LACP_STRICT; + + ra->ra_active = sc->sc_active; + } else { + /* + * LACP tracks active links automatically, + * the others do not. + */ + ra->ra_active = 0; + SLIST_FOREACH(lp, &sc->sc_ports, lp_entries) + ra->ra_active += LAGG_PORTACTIVE(lp); + } + ra->ra_flapping = sc->sc_flapping; + ra->ra_flowid_shift = sc->flowid_shift; + count = 0; buf = outbuf; len = min(ra->ra_size, buflen); @@ -1225,9 +1260,88 @@ lagg_ioctl(struct ifnet *ifp, u_long cmd, caddr_t data) free(outbuf, M_TEMP); break; case SIOCSLAGG: + /* + * Set options or protocol depending on + * ra->ra_opts and ra->ra_proto. + */ error = priv_check(td, PRIV_NET_LAGG); if (error) break; + if (ra->ra_opts != 0) { + /* + * Set options. LACP options are stored in sc->sc_psc, + * not in sc_opts. + */ + int valid, lacp; + + switch (ra->ra_opts) { + case LAGG_OPT_USE_FLOWID: + case -LAGG_OPT_USE_FLOWID: + case LAGG_OPT_FLOWIDSHIFT: + valid = 1; + lacp = 0; + break; + case LAGG_OPT_LACP_TXTEST: + case -LAGG_OPT_LACP_TXTEST: + case LAGG_OPT_LACP_RXTEST: + case -LAGG_OPT_LACP_RXTEST: + case LAGG_OPT_LACP_STRICT: + case -LAGG_OPT_LACP_STRICT: + valid = lacp = 1; + break; + default: + valid = lacp = 0; + break; + } + + LAGG_WLOCK(sc); + if (valid == 0 || + (lacp == 1 && sc->sc_proto != LAGG_PROTO_LACP)) { + /* Invalid combination of options specified. */ + error = EINVAL; + LAGG_WUNLOCK(sc); + break; /* Return from SIOCSLAGG. */ + } + /* + * Store new options into sc->sc_opts except for + * FLOWIDSHIFT and LACP options. + */ + if (lacp == 0) { + if (ra->ra_opts == LAGG_OPT_FLOWIDSHIFT) + sc->flowid_shift = ra->ra_flowid_shift; + else if (ra->ra_opts > 0) + sc->sc_opts |= ra->ra_opts; + else + sc->sc_opts &= ~ra->ra_opts; + } else { + struct lacp_softc *lsc; + + lsc = (struct lacp_softc *)sc->sc_psc; + + switch (ra->ra_opts) { + case LAGG_OPT_LACP_TXTEST: + lsc->lsc_debug.lsc_tx_test = 1; + break; + case -LAGG_OPT_LACP_TXTEST: + lsc->lsc_debug.lsc_tx_test = 0; + break; + case LAGG_OPT_LACP_RXTEST: + lsc->lsc_debug.lsc_rx_test = 1; + break; + case -LAGG_OPT_LACP_RXTEST: + lsc->lsc_debug.lsc_rx_test = 0; + break; + case LAGG_OPT_LACP_STRICT: + lsc->lsc_strict_mode = 1; + break; + case -LAGG_OPT_LACP_STRICT: + lsc->lsc_strict_mode = 0; + break; + } + } + LAGG_WUNLOCK(sc); + break; /* Return from SIOCSLAGG. */ + } if (ra->ra_proto < 1 || ra->ra_proto >= LAGG_PROTO_MAX) { error = EPROTONOSUPPORT; break; @@ -1687,27 +1801,6 @@ lagg_gethdr(struct mbuf *m, u_int off, u_int len, void *buf) return (mtod(m, char *) + off); } -static int -lagg_sysctl_active(SYSCTL_HANDLER_ARGS) -{ - struct lagg_softc *sc = (struct lagg_softc *)arg1; - struct lagg_port *lp; - int error; - - /* LACP tracks active links automatically, the others do not */ - if (sc->sc_proto != LAGG_PROTO_LACP) { - sc->sc_active = 0; - SLIST_FOREACH(lp, &sc->sc_ports, lp_entries) - sc->sc_active += LAGG_PORTACTIVE(lp); - } - - error = sysctl_handle_int(oidp, &sc->sc_active, 0, req); - if ((error) || (req->newptr == NULL)) - return (error); - - return (0); -} - uint32_t lagg_hashmbuf(struct lagg_softc *sc, struct mbuf *m, uint32_t key) { @@ -1948,7 +2041,7 @@ lagg_fail_input(struct lagg_softc *sc, struct lagg_port *lp, struct mbuf *m) struct ifnet *ifp = sc->sc_ifp; struct lagg_port *tmp_tp; - if (lp == sc->sc_primary || lagg_failover_rx_all) { + if (lp == sc->sc_primary || V_lagg_failover_rx_all) { m->m_pkthdr.rcvif = ifp; return (m); } @@ -2043,7 +2136,7 @@ lagg_lb_start(struct lagg_softc *sc, struct mbuf *m) struct lagg_port *lp = NULL; uint32_t p = 0; - if (sc->use_flowid && (m->m_flags & M_FLOWID)) + if ((sc->sc_opts & LAGG_OPT_USE_FLOWID) && (m->m_flags & M_FLOWID)) p = m->m_pkthdr.flowid >> sc->flowid_shift; else p = lagg_hashmbuf(sc, m, lb->lb_key); diff --git a/sys/net/if_lagg.h b/sys/net/if_lagg.h index 5aeae571da77..6c39d1c451e9 100644 --- a/sys/net/if_lagg.h +++ b/sys/net/if_lagg.h @@ -125,6 +125,19 @@ struct lagg_reqall { struct lacp_opreq rpsc_lacp; } ra_psc; #define ra_lacpreq ra_psc.rpsc_lacp + int ra_opts; /* Option bitmap */ +#define LAGG_OPT_NONE 0x00 +#define LAGG_OPT_USE_FLOWID 0x01 /* use M_FLOWID */ +/* Pseudo flags which are used in ra_opts but not stored into sc_opts. */ +#define LAGG_OPT_FLOWIDSHIFT 0x02 /* Set flowid */ +#define LAGG_OPT_FLOWIDSHIFT_MASK 0x1f /* flowid is uint32_t */ +#define LAGG_OPT_LACP_STRICT 0x10 /* LACP strict mode */ +#define LAGG_OPT_LACP_TXTEST 0x20 /* LACP debug: txtest */ +#define LAGG_OPT_LACP_RXTEST 0x40 /* LACP debug: rxtest */ + u_int ra_count; /* number of ports */ + u_int ra_active; /* active port count */ + u_int ra_flapping; /* number of flapping */ + int ra_flowid_shift; /* shift the flowid */ }; #define SIOCGLAGG _IOWR('i', 143, struct lagg_reqall) @@ -212,9 +225,7 @@ struct lagg_softc { eventhandler_tag vlan_attach; eventhandler_tag vlan_detach; struct callout sc_callout; - struct sysctl_ctx_list ctx; /* sysctl variables */ - struct sysctl_oid *sc_oid; /* sysctl tree oid */ - int use_flowid; /* use M_FLOWID */ + u_int sc_opts; int flowid_shift; /* shift the flowid */ struct lagg_counters detached_counters; /* detached ports sum */ };