X-Git-Url: http://git.onelab.eu/?a=blobdiff_plain;f=lib%2Fdpif-linux.c;h=804a90ffd5abe2047ff1472d47e06c5d5e5bc1d7;hb=f80022d9df98d29ee41176a4bc6cb91025da84b8;hp=cde11672772e6e4c036ee053d9cd1cf6928d8659;hpb=cb22974d773942d66da42b700b8bca0db27a0920;p=sliver-openvswitch.git diff --git a/lib/dpif-linux.c b/lib/dpif-linux.c index cde116727..804a90ffd 100644 --- a/lib/dpif-linux.c +++ b/lib/dpif-linux.c @@ -1,5 +1,5 @@ /* - * Copyright (c) 2008, 2009, 2010, 2011, 2012 Nicira, Inc. + * Copyright (c) 2008, 2009, 2010, 2011, 2012, 2013 Nicira, Inc. * * Licensed under the Apache License, Version 2.0 (the "License"); * you may not use this file except in compliance with the License. @@ -47,7 +47,6 @@ #include "odp-util.h" #include "ofpbuf.h" #include "openvswitch/datapath-compat.h" -#include "openvswitch/tunnel.h" #include "packets.h" #include "poll-loop.h" #include "random.h" @@ -106,6 +105,8 @@ struct dpif_linux_flow { * the Netlink version of the command, even if actions_len is zero. */ const struct nlattr *key; /* OVS_FLOW_ATTR_KEY. */ size_t key_len; + const struct nlattr *mask; /* OVS_FLOW_ATTR_MASK. */ + size_t mask_len; const struct nlattr *actions; /* OVS_FLOW_ATTR_ACTIONS. */ size_t actions_len; const struct ovs_flow_stats *stats; /* OVS_FLOW_ATTR_STATS. */ @@ -168,7 +169,8 @@ static int dpif_linux_init(void); static void open_dpif(const struct dpif_linux_dp *, struct dpif **); static bool dpif_linux_nln_parse(struct ofpbuf *, void *); static void dpif_linux_port_changed(const void *vport, void *dpif); -static uint32_t dpif_linux_port_get_pid(const struct dpif *, uint32_t port_no); +static uint32_t dpif_linux_port_get_pid(const struct dpif *, + odp_port_t port_no); static void dpif_linux_vport_to_ofpbuf(const struct dpif_linux_vport *, struct ofpbuf *); @@ -260,7 +262,7 @@ open_dpif(const struct dpif_linux_dp *dp, struct dpif **dpifp) static void destroy_channels(struct dpif_linux *dpif) { - int i; + unsigned int i; if (dpif->epoll_fd < 0) { return; @@ -279,7 +281,7 @@ destroy_channels(struct dpif_linux *dpif) dpif_linux_vport_init(&vport_request); vport_request.cmd = OVS_VPORT_CMD_SET; vport_request.dp_ifindex = dpif->dp_ifindex; - vport_request.port_no = i; + vport_request.port_no = u32_to_odp(i); vport_request.upcall_pid = &upcall_pid; dpif_linux_vport_transact(&vport_request, NULL, NULL); @@ -299,9 +301,10 @@ destroy_channels(struct dpif_linux *dpif) } static int -add_channel(struct dpif_linux *dpif, uint32_t port_no, struct nl_sock *sock) +add_channel(struct dpif_linux *dpif, odp_port_t port_no, struct nl_sock *sock) { struct epoll_event event; + uint32_t port_idx = odp_to_u32(port_no); if (dpif->epoll_fd < 0) { return 0; @@ -309,11 +312,11 @@ add_channel(struct dpif_linux *dpif, uint32_t port_no, struct nl_sock *sock) /* We assume that the datapath densely chooses port numbers, which * can therefore be used as an index into an array of channels. */ - if (port_no >= dpif->uc_array_size) { - int new_size = port_no + 1; - int i; + if (port_idx >= dpif->uc_array_size) { + uint32_t new_size = port_idx + 1; + uint32_t i; - if (new_size > 65535) { + if (new_size > MAX_PORTS) { VLOG_WARN_RL(&error_rl, "%s: datapath port %"PRIu32" too big", dpif_name(&dpif->dpif), port_no); return EFBIG; @@ -332,33 +335,36 @@ add_channel(struct dpif_linux *dpif, uint32_t port_no, struct nl_sock *sock) memset(&event, 0, sizeof event); event.events = EPOLLIN; - event.data.u32 = port_no; + event.data.u32 = port_idx; if (epoll_ctl(dpif->epoll_fd, EPOLL_CTL_ADD, nl_sock_fd(sock), &event) < 0) { return errno; } - dpif->channels[port_no].sock = sock; - dpif->channels[port_no].last_poll = LLONG_MIN; + nl_sock_destroy(dpif->channels[port_idx].sock); + dpif->channels[port_idx].sock = sock; + dpif->channels[port_idx].last_poll = LLONG_MIN; return 0; } static void -del_channel(struct dpif_linux *dpif, uint32_t port_no) +del_channel(struct dpif_linux *dpif, odp_port_t port_no) { struct dpif_channel *ch; + uint32_t port_idx = odp_to_u32(port_no); - if (dpif->epoll_fd < 0 || port_no >= dpif->uc_array_size) { + if (dpif->epoll_fd < 0 || port_idx >= dpif->uc_array_size) { return; } - ch = &dpif->channels[port_no]; + ch = &dpif->channels[port_idx]; if (!ch->sock) { return; } epoll_ctl(dpif->epoll_fd, EPOLL_CTL_DEL, nl_sock_fd(ch->sock), NULL); + dpif->event_offset = dpif->n_events = 0; nl_sock_destroy(ch->sock); ch->sock = NULL; @@ -421,18 +427,78 @@ dpif_linux_get_stats(const struct dpif *dpif_, struct dpif_dp_stats *stats) return error; } +static const char * +get_vport_type(const struct dpif_linux_vport *vport) +{ + static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(5, 20); + + switch (vport->type) { + case OVS_VPORT_TYPE_NETDEV: + return "system"; + + case OVS_VPORT_TYPE_INTERNAL: + return "internal"; + + case OVS_VPORT_TYPE_GRE: + return "gre"; + + case OVS_VPORT_TYPE_GRE64: + return "gre64"; + + case OVS_VPORT_TYPE_VXLAN: + return "vxlan"; + + case OVS_VPORT_TYPE_LISP: + return "lisp"; + + case OVS_VPORT_TYPE_UNSPEC: + case __OVS_VPORT_TYPE_MAX: + break; + } + + VLOG_WARN_RL(&rl, "dp%d: port `%s' has unsupported type %u", + vport->dp_ifindex, vport->name, (unsigned int) vport->type); + return "unknown"; +} + +static enum ovs_vport_type +netdev_to_ovs_vport_type(const struct netdev *netdev) +{ + const char *type = netdev_get_type(netdev); + + if (!strcmp(type, "tap") || !strcmp(type, "system")) { + return OVS_VPORT_TYPE_NETDEV; + } else if (!strcmp(type, "internal")) { + return OVS_VPORT_TYPE_INTERNAL; + } else if (strstr(type, "gre64")) { + return OVS_VPORT_TYPE_GRE64; + } else if (strstr(type, "gre")) { + return OVS_VPORT_TYPE_GRE; + } else if (!strcmp(type, "vxlan")) { + return OVS_VPORT_TYPE_VXLAN; + } else if (!strcmp(type, "lisp")) { + return OVS_VPORT_TYPE_LISP; + } else { + return OVS_VPORT_TYPE_UNSPEC; + } +} + static int dpif_linux_port_add(struct dpif *dpif_, struct netdev *netdev, - uint32_t *port_nop) + odp_port_t *port_nop) { struct dpif_linux *dpif = dpif_linux_cast(dpif_); - const char *name = netdev_get_name(netdev); + const struct netdev_tunnel_config *tnl_cfg; + char namebuf[NETDEV_VPORT_NAME_BUFSIZE]; + const char *name = netdev_vport_get_dpif_port(netdev, + namebuf, sizeof namebuf); const char *type = netdev_get_type(netdev); struct dpif_linux_vport request, reply; - const struct ofpbuf *options; struct nl_sock *sock = NULL; uint32_t upcall_pid; struct ofpbuf *buf; + uint64_t options_stub[64 / 8]; + struct ofpbuf options; int error; if (dpif->epoll_fd >= 0) { @@ -445,7 +511,7 @@ dpif_linux_port_add(struct dpif *dpif_, struct netdev *netdev, dpif_linux_vport_init(&request); request.cmd = OVS_VPORT_CMD_NEW; request.dp_ifindex = dpif->dp_ifindex; - request.type = netdev_vport_get_vport_type(netdev); + request.type = netdev_to_ovs_vport_type(netdev); if (request.type == OVS_VPORT_TYPE_UNSPEC) { VLOG_WARN_RL(&error_rl, "%s: cannot create port `%s' because it has " "unsupported type `%s'", @@ -455,16 +521,19 @@ dpif_linux_port_add(struct dpif *dpif_, struct netdev *netdev, } request.name = name; - options = netdev_vport_get_options(netdev); - if (options && options->size) { - request.options = options->data; - request.options_len = options->size; - } - if (request.type == OVS_VPORT_TYPE_NETDEV) { netdev_linux_ethtool_set_flag(netdev, ETH_FLAG_LRO, "LRO", false); } + tnl_cfg = netdev_get_tunnel_config(netdev); + if (tnl_cfg && tnl_cfg->dst_port != 0) { + ofpbuf_use_stack(&options, options_stub, sizeof options_stub); + nl_msg_put_u16(&options, OVS_TUNNEL_ATTR_DST_PORT, + ntohs(tnl_cfg->dst_port)); + request.options = options.data; + request.options_len = options.size; + } + request.port_no = *port_nop; upcall_pid = sock ? nl_sock_pid(sock) : 0; request.upcall_pid = &upcall_pid; @@ -474,9 +543,11 @@ dpif_linux_port_add(struct dpif *dpif_, struct netdev *netdev, *port_nop = reply.port_no; VLOG_DBG("%s: assigning port %"PRIu32" to netlink pid %"PRIu32, dpif_name(dpif_), reply.port_no, upcall_pid); - } else if (error == EBUSY && *port_nop != UINT32_MAX) { - VLOG_INFO("%s: requested port %"PRIu32" is in use", - dpif_name(dpif_), *port_nop); + } else { + if (error == EBUSY && *port_nop != ODPP_NONE) { + VLOG_INFO("%s: requested port %"PRIu32" is in use", + dpif_name(dpif_), *port_nop); + } nl_sock_destroy(sock); ofpbuf_delete(buf); return error; @@ -505,7 +576,7 @@ dpif_linux_port_add(struct dpif *dpif_, struct netdev *netdev, } static int -dpif_linux_port_del(struct dpif *dpif_, uint32_t port_no) +dpif_linux_port_del(struct dpif *dpif_, odp_port_t port_no) { struct dpif_linux *dpif = dpif_linux_cast(dpif_); struct dpif_linux_vport vport; @@ -523,7 +594,7 @@ dpif_linux_port_del(struct dpif *dpif_, uint32_t port_no) } static int -dpif_linux_port_query__(const struct dpif *dpif, uint32_t port_no, +dpif_linux_port_query__(const struct dpif *dpif, odp_port_t port_no, const char *port_name, struct dpif_port *dpif_port) { struct dpif_linux_vport request; @@ -545,7 +616,7 @@ dpif_linux_port_query__(const struct dpif *dpif, uint32_t port_no, error = ENODEV; } else if (dpif_port) { dpif_port->name = xstrdup(reply.name); - dpif_port->type = xstrdup(netdev_vport_get_netdev_type(&reply)); + dpif_port->type = xstrdup(get_vport_type(&reply)); dpif_port->port_no = reply.port_no; } ofpbuf_delete(buf); @@ -554,7 +625,7 @@ dpif_linux_port_query__(const struct dpif *dpif, uint32_t port_no, } static int -dpif_linux_port_query_by_number(const struct dpif *dpif, uint32_t port_no, +dpif_linux_port_query_by_number(const struct dpif *dpif, odp_port_t port_no, struct dpif_port *dpif_port) { return dpif_linux_port_query__(dpif, port_no, NULL, dpif_port); @@ -567,23 +638,24 @@ dpif_linux_port_query_by_name(const struct dpif *dpif, const char *devname, return dpif_linux_port_query__(dpif, 0, devname, dpif_port); } -static int +static odp_port_t dpif_linux_get_max_ports(const struct dpif *dpif OVS_UNUSED) { - return MAX_PORTS; + return u32_to_odp(MAX_PORTS); } static uint32_t -dpif_linux_port_get_pid(const struct dpif *dpif_, uint32_t port_no) +dpif_linux_port_get_pid(const struct dpif *dpif_, odp_port_t port_no) { struct dpif_linux *dpif = dpif_linux_cast(dpif_); + uint32_t port_idx = odp_to_u32(port_no); if (dpif->epoll_fd < 0) { return 0; } else { - /* The UINT32_MAX "reserved" port number uses the "ovs-system"'s + /* The ODPP_NONE "reserved" port number uses the "ovs-system"'s * channel, since it is not heavily loaded. */ - int idx = (port_no >= dpif->uc_array_size) ? 0 : port_no; + uint32_t idx = port_idx >= dpif->uc_array_size ? 0 : port_idx; return nl_sock_pid(dpif->channels[idx].sock); } } @@ -645,7 +717,7 @@ dpif_linux_port_dump_next(const struct dpif *dpif OVS_UNUSED, void *state_, } dpif_port->name = CONST_CAST(char *, vport.name); - dpif_port->type = CONST_CAST(char *, netdev_vport_get_netdev_type(&vport)); + dpif_port->type = CONST_CAST(char *, get_vport_type(&vport)); dpif_port->port_no = vport.port_no; return 0; } @@ -731,7 +803,7 @@ static void dpif_linux_init_flow_put(struct dpif *dpif_, const struct dpif_flow_put *put, struct dpif_linux_flow *request) { - static struct nlattr dummy_action; + static const struct nlattr dummy_action; struct dpif_linux *dpif = dpif_linux_cast(dpif_); @@ -741,8 +813,12 @@ dpif_linux_init_flow_put(struct dpif *dpif_, const struct dpif_flow_put *put, request->dp_ifindex = dpif->dp_ifindex; request->key = put->key; request->key_len = put->key_len; + request->mask = put->mask; + request->mask_len = put->mask_len; /* Ensure that OVS_FLOW_ATTR_ACTIONS will always be included. */ - request->actions = put->actions ? put->actions : &dummy_action; + request->actions = (put->actions + ? put->actions + : CONST_CAST(struct nlattr *, &dummy_action)); request->actions_len = put->actions_len; if (put->flags & DPIF_FP_ZERO_STATS) { request->clear = true; @@ -833,6 +909,7 @@ dpif_linux_flow_dump_start(const struct dpif *dpif_, void **statep) static int dpif_linux_flow_dump_next(const struct dpif *dpif_ OVS_UNUSED, void *state_, const struct nlattr **key, size_t *key_len, + const struct nlattr **mask, size_t *mask_len, const struct nlattr **actions, size_t *actions_len, const struct dpif_flow_stats **stats) { @@ -860,7 +937,8 @@ dpif_linux_flow_dump_next(const struct dpif *dpif_ OVS_UNUSED, void *state_, if (error == ENOENT) { VLOG_DBG("dumped flow disappeared on get"); } else if (error) { - VLOG_WARN("error fetching dumped flow: %s", strerror(error)); + VLOG_WARN("error fetching dumped flow: %s", + ovs_strerror(error)); } } } while (error); @@ -873,6 +951,10 @@ dpif_linux_flow_dump_next(const struct dpif *dpif_ OVS_UNUSED, void *state_, *key = state->flow.key; *key_len = state->flow.key_len; } + if (mask) { + *mask = state->flow.mask; + *mask_len = state->flow.mask ? state->flow.mask_len : 0; + } if (stats) { dpif_linux_flow_get_stats(&state->flow, &state->stats); *stats = &state->stats; @@ -1127,7 +1209,7 @@ dpif_linux_recv_set(struct dpif *dpif_, bool enable) } else { VLOG_WARN_RL(&error_rl, "%s: failed to set upcall pid on port: %s", - dpif_name(&dpif->dpif), strerror(error)); + dpif_name(&dpif->dpif), ovs_strerror(error)); nl_sock_destroy(sock); if (error == ENODEV || error == ENOENT) { @@ -1174,7 +1256,7 @@ parse_odp_packet(struct ofpbuf *buf, struct dpif_upcall *upcall, [OVS_PACKET_ATTR_KEY] = { .type = NL_A_NESTED }, /* OVS_PACKET_CMD_ACTION only. */ - [OVS_PACKET_ATTR_USERDATA] = { .type = NL_A_U64, .optional = true }, + [OVS_PACKET_ATTR_USERDATA] = { .type = NL_A_UNSPEC, .optional = true }, }; struct ovs_header *ovs_header; @@ -1212,9 +1294,7 @@ parse_odp_packet(struct ofpbuf *buf, struct dpif_upcall *upcall, upcall->key = CONST_CAST(struct nlattr *, nl_attr_get(a[OVS_PACKET_ATTR_KEY])); upcall->key_len = nl_attr_get_size(a[OVS_PACKET_ATTR_KEY]); - upcall->userdata = (a[OVS_PACKET_ATTR_USERDATA] - ? nl_attr_get_u64(a[OVS_PACKET_ATTR_USERDATA]) - : 0); + upcall->userdata = a[OVS_PACKET_ATTR_USERDATA]; *dp_ifindex = ovs_header->dp_ifindex; return 0; @@ -1242,7 +1322,7 @@ dpif_linux_recv(struct dpif *dpif_, struct dpif_upcall *upcall, } while (retval < 0 && errno == EINTR); if (retval < 0) { static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1); - VLOG_WARN_RL(&rl, "epoll_wait failed (%s)", strerror(errno)); + VLOG_WARN_RL(&rl, "epoll_wait failed (%s)", ovs_strerror(errno)); } else if (retval > 0) { dpif->n_events = retval; } @@ -1413,7 +1493,7 @@ dpif_linux_is_internal_device(const char *name) ofpbuf_delete(buf); } else if (error != ENODEV && error != ENOENT) { VLOG_WARN_RL(&error_rl, "%s: vport query failed (%s)", - name, strerror(error)); + name, ovs_strerror(error)); } return reply.type == OVS_VPORT_TYPE_INTERNAL; @@ -1463,10 +1543,6 @@ dpif_linux_vport_from_ofpbuf(struct dpif_linux_vport *vport, [OVS_VPORT_ATTR_UPCALL_PID] = { .type = NL_A_U32 }, [OVS_VPORT_ATTR_STATS] = { NL_POLICY_FOR(struct ovs_vport_stats), .optional = true }, - [OVS_VPORT_ATTR_ADDRESS] = { .type = NL_A_UNSPEC, - .min_len = ETH_ADDR_LEN, - .max_len = ETH_ADDR_LEN, - .optional = true }, [OVS_VPORT_ATTR_OPTIONS] = { .type = NL_A_NESTED, .optional = true }, }; @@ -1491,7 +1567,7 @@ dpif_linux_vport_from_ofpbuf(struct dpif_linux_vport *vport, vport->cmd = genl->cmd; vport->dp_ifindex = ovs_header->dp_ifindex; - vport->port_no = nl_attr_get_u32(a[OVS_VPORT_ATTR_PORT_NO]); + vport->port_no = nl_attr_get_odp_port(a[OVS_VPORT_ATTR_PORT_NO]); vport->type = nl_attr_get_u32(a[OVS_VPORT_ATTR_TYPE]); vport->name = nl_attr_get_string(a[OVS_VPORT_ATTR_NAME]); if (a[OVS_VPORT_ATTR_UPCALL_PID]) { @@ -1500,9 +1576,6 @@ dpif_linux_vport_from_ofpbuf(struct dpif_linux_vport *vport, if (a[OVS_VPORT_ATTR_STATS]) { vport->stats = nl_attr_get(a[OVS_VPORT_ATTR_STATS]); } - if (a[OVS_VPORT_ATTR_ADDRESS]) { - vport->address = nl_attr_get(a[OVS_VPORT_ATTR_ADDRESS]); - } if (a[OVS_VPORT_ATTR_OPTIONS]) { vport->options = nl_attr_get(a[OVS_VPORT_ATTR_OPTIONS]); vport->options_len = nl_attr_get_size(a[OVS_VPORT_ATTR_OPTIONS]); @@ -1524,8 +1597,8 @@ dpif_linux_vport_to_ofpbuf(const struct dpif_linux_vport *vport, ovs_header = ofpbuf_put_uninit(buf, sizeof *ovs_header); ovs_header->dp_ifindex = vport->dp_ifindex; - if (vport->port_no != UINT32_MAX) { - nl_msg_put_u32(buf, OVS_VPORT_ATTR_PORT_NO, vport->port_no); + if (vport->port_no != ODPP_NONE) { + nl_msg_put_odp_port(buf, OVS_VPORT_ATTR_PORT_NO, vport->port_no); } if (vport->type != OVS_VPORT_TYPE_UNSPEC) { @@ -1545,11 +1618,6 @@ dpif_linux_vport_to_ofpbuf(const struct dpif_linux_vport *vport, vport->stats, sizeof *vport->stats); } - if (vport->address) { - nl_msg_put_unspec(buf, OVS_VPORT_ATTR_ADDRESS, - vport->address, ETH_ADDR_LEN); - } - if (vport->options) { nl_msg_put_nested(buf, OVS_VPORT_ATTR_OPTIONS, vport->options, vport->options_len); @@ -1561,7 +1629,7 @@ void dpif_linux_vport_init(struct dpif_linux_vport *vport) { memset(vport, 0, sizeof *vport); - vport->port_no = UINT32_MAX; + vport->port_no = ODPP_NONE; } /* Executes 'request' in the kernel datapath. If the command fails, returns a @@ -1778,6 +1846,7 @@ dpif_linux_flow_from_ofpbuf(struct dpif_linux_flow *flow, { static const struct nl_policy ovs_flow_policy[] = { [OVS_FLOW_ATTR_KEY] = { .type = NL_A_NESTED }, + [OVS_FLOW_ATTR_MASK] = { .type = NL_A_NESTED, .optional = true }, [OVS_FLOW_ATTR_ACTIONS] = { .type = NL_A_NESTED, .optional = true }, [OVS_FLOW_ATTR_STATS] = { NL_POLICY_FOR(struct ovs_flow_stats), .optional = true }, @@ -1809,6 +1878,11 @@ dpif_linux_flow_from_ofpbuf(struct dpif_linux_flow *flow, flow->dp_ifindex = ovs_header->dp_ifindex; flow->key = nl_attr_get(a[OVS_FLOW_ATTR_KEY]); flow->key_len = nl_attr_get_size(a[OVS_FLOW_ATTR_KEY]); + + if (a[OVS_FLOW_ATTR_MASK]) { + flow->mask = nl_attr_get(a[OVS_FLOW_ATTR_MASK]); + flow->mask_len = nl_attr_get_size(a[OVS_FLOW_ATTR_MASK]); + } if (a[OVS_FLOW_ATTR_ACTIONS]) { flow->actions = nl_attr_get(a[OVS_FLOW_ATTR_ACTIONS]); flow->actions_len = nl_attr_get_size(a[OVS_FLOW_ATTR_ACTIONS]); @@ -1844,6 +1918,10 @@ dpif_linux_flow_to_ofpbuf(const struct dpif_linux_flow *flow, nl_msg_put_unspec(buf, OVS_FLOW_ATTR_KEY, flow->key, flow->key_len); } + if (flow->mask_len) { + nl_msg_put_unspec(buf, OVS_FLOW_ATTR_MASK, flow->mask, flow->mask_len); + } + if (flow->actions || flow->actions_len) { nl_msg_put_unspec(buf, OVS_FLOW_ATTR_ACTIONS, flow->actions, flow->actions_len); @@ -1927,7 +2005,7 @@ report_loss(struct dpif *dpif_, struct dpif_channel *ch) static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(5, 5); struct ds s; - if (VLOG_DROP_ERR(&rl)) { + if (VLOG_DROP_WARN(&rl)) { return; }