X-Git-Url: http://git.onelab.eu/?a=blobdiff_plain;f=lib%2Fodp-util.c;h=751c1c92390dea32cdfd0b939137ac6b754e9dec;hb=a6ca7c672f6d5aaf24111d3414ea132ec36b40d6;hp=901dac3b4c80cba5649ff5a58af604632fde2127;hpb=6e9bea4da975729edf650c94430a02cc0518e383;p=sliver-openvswitch.git diff --git a/lib/odp-util.c b/lib/odp-util.c index 901dac3b4..751c1c923 100644 --- a/lib/odp-util.c +++ b/lib/odp-util.c @@ -1,5 +1,5 @@ /* - * Copyright (c) 2009, 2010, 2011, 2012 Nicira, Inc. + * Copyright (c) 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. @@ -73,6 +73,8 @@ odp_action_len(uint16_t type) case OVS_ACTION_ATTR_USERSPACE: return -2; case OVS_ACTION_ATTR_PUSH_VLAN: return sizeof(struct ovs_action_push_vlan); case OVS_ACTION_ATTR_POP_VLAN: return 0; + case OVS_ACTION_ATTR_PUSH_MPLS: return sizeof(struct ovs_action_push_mpls); + case OVS_ACTION_ATTR_POP_MPLS: return sizeof(ovs_be16); case OVS_ACTION_ATTR_SET: return -2; case OVS_ACTION_ATTR_SAMPLE: return -2; @@ -92,7 +94,9 @@ ovs_key_attr_to_string(enum ovs_key_attr attr) switch (attr) { case OVS_KEY_ATTR_UNSPEC: return "unspec"; case OVS_KEY_ATTR_ENCAP: return "encap"; - case OVS_KEY_ATTR_PRIORITY: return "priority"; + case OVS_KEY_ATTR_PRIORITY: return "skb_priority"; + case OVS_KEY_ATTR_SKB_MARK: return "skb_mark"; + case OVS_KEY_ATTR_TUNNEL: return "tunnel"; case OVS_KEY_ATTR_IN_PORT: return "in_port"; case OVS_KEY_ATTR_ETHERNET: return "eth"; case OVS_KEY_ATTR_VLAN: return "vlan"; @@ -105,7 +109,7 @@ ovs_key_attr_to_string(enum ovs_key_attr attr) case OVS_KEY_ATTR_ICMPV6: return "icmpv6"; case OVS_KEY_ATTR_ARP: return "arp"; case OVS_KEY_ATTR_ND: return "nd"; - case OVS_KEY_ATTR_TUN_ID: return "tun_id"; + case OVS_KEY_ATTR_MPLS: return "mpls"; case __OVS_KEY_ATTR_MAX: default: @@ -166,8 +170,10 @@ format_odp_sample_action(struct ds *ds, const struct nlattr *attr) } static const char * -slow_path_reason_to_string(enum slow_path_reason bit) +slow_path_reason_to_string(uint32_t data) { + enum slow_path_reason bit = (enum slow_path_reason) data; + switch (bit) { case SLOW_CFM: return "cfm"; @@ -186,29 +192,54 @@ slow_path_reason_to_string(enum slow_path_reason bit) } } -static void -format_slow_path_reason(struct ds *ds, uint32_t slow) +static int +parse_flags(const char *s, const char *(*bit_to_string)(uint32_t), + uint32_t *res) { - uint32_t bad = 0; + uint32_t result = 0; + int n = 0; - while (slow) { - uint32_t bit = rightmost_1bit(slow); - const char *s; + if (s[n] != '(') { + return -EINVAL; + } + n++; - s = slow_path_reason_to_string(bit); - if (s) { - ds_put_format(ds, "%s,", s); - } else { - bad |= bit; + while (s[n] != ')') { + unsigned long long int flags; + uint32_t bit; + int n0; + + if (sscanf(&s[n], "%lli%n", &flags, &n0) > 0 && n0 > 0) { + n += n0 + (s[n + n0] == ','); + result |= flags; + continue; } - slow &= ~bit; - } + for (bit = 1; bit; bit <<= 1) { + const char *name = bit_to_string(bit); + size_t len; + + if (!name) { + continue; + } + + len = strlen(name); + if (!strncmp(s + n, name, len) && + (s[n + len] == ',' || s[n + len] == ')')) { + result |= bit; + n += len + (s[n + len] == ','); + break; + } + } - if (bad) { - ds_put_format(ds, "0x%"PRIx32",", bad); + if (!bit) { + return -EINVAL; + } } - ds_chomp(ds, ','); + n++; + + *res = result; + return n; } static void @@ -216,9 +247,11 @@ format_odp_userspace_action(struct ds *ds, const struct nlattr *attr) { static const struct nl_policy ovs_userspace_policy[] = { [OVS_USERSPACE_ATTR_PID] = { .type = NL_A_U32 }, - [OVS_USERSPACE_ATTR_USERDATA] = { .type = NL_A_U64, .optional = true }, + [OVS_USERSPACE_ATTR_USERDATA] = { .type = NL_A_UNSPEC, + .optional = true }, }; struct nlattr *a[ARRAY_SIZE(ovs_userspace_policy)]; + const struct nlattr *userdata_attr; if (!nl_parse_nested(attr, ovs_userspace_policy, a, ARRAY_SIZE(a))) { ds_put_cstr(ds, "userspace(error)"); @@ -228,7 +261,8 @@ format_odp_userspace_action(struct ds *ds, const struct nlattr *attr) ds_put_format(ds, "userspace(pid=%"PRIu32, nl_attr_get_u32(a[OVS_USERSPACE_ATTR_PID])); - if (a[OVS_USERSPACE_ATTR_USERDATA]) { + userdata_attr = a[OVS_USERSPACE_ATTR_USERDATA]; + if (userdata_attr && nl_attr_get_size(userdata_attr) == sizeof(uint64_t)) { uint64_t userdata = nl_attr_get_u64(a[OVS_USERSPACE_ATTR_USERDATA]); union user_action_cookie cookie; @@ -245,10 +279,9 @@ format_odp_userspace_action(struct ds *ds, const struct nlattr *attr) case USER_ACTION_COOKIE_SLOW_PATH: ds_put_cstr(ds, ",slow_path("); - if (cookie.slow_path.reason) { - format_slow_path_reason(ds, cookie.slow_path.reason); - } - ds_put_char(ds, ')'); + format_flags(ds, slow_path_reason_to_string, + cookie.slow_path.reason, ','); + ds_put_format(ds, ")"); break; case USER_ACTION_COOKIE_UNSPEC: @@ -256,6 +289,16 @@ format_odp_userspace_action(struct ds *ds, const struct nlattr *attr) ds_put_format(ds, ",userdata=0x%"PRIx64, userdata); break; } + } else if (userdata_attr) { + const uint8_t *userdata = nl_attr_get(userdata_attr); + size_t len = nl_attr_get_size(userdata_attr); + size_t i; + + ds_put_format(ds, ",userdata("); + for (i = 0; i < len; i++) { + ds_put_format(ds, "%02x", userdata[i]); + } + ds_put_char(ds, ')'); } ds_put_char(ds, ')'); @@ -272,6 +315,16 @@ format_vlan_tci(struct ds *ds, ovs_be16 vlan_tci) } } +static void +format_mpls_lse(struct ds *ds, ovs_be32 mpls_lse) +{ + ds_put_format(ds, "label=%"PRIu32",tc=%d,ttl=%d,bos=%d", + mpls_lse_to_label(mpls_lse), + mpls_lse_to_tc(mpls_lse), + mpls_lse_to_ttl(mpls_lse), + mpls_lse_to_bos(mpls_lse)); +} + static void format_odp_action(struct ds *ds, const struct nlattr *a) { @@ -311,6 +364,18 @@ format_odp_action(struct ds *ds, const struct nlattr *a) case OVS_ACTION_ATTR_POP_VLAN: ds_put_cstr(ds, "pop_vlan"); break; + case OVS_ACTION_ATTR_PUSH_MPLS: { + const struct ovs_action_push_mpls *mpls = nl_attr_get(a); + ds_put_cstr(ds, "push_mpls("); + format_mpls_lse(ds, mpls->mpls_lse); + ds_put_format(ds, ",eth_type=0x%"PRIx16")", ntohs(mpls->mpls_ethertype)); + break; + } + case OVS_ACTION_ATTR_POP_MPLS: { + ovs_be16 ethertype = nl_attr_get_be16(a); + ds_put_format(ds, "pop_mpls(eth_type=0x%"PRIx16")", ntohs(ethertype)); + break; + } case OVS_ACTION_ATTR_SAMPLE: format_odp_sample_action(ds, a); break; @@ -396,7 +461,7 @@ parse_odp_action(const char *s, const struct simap *port_names, int n = -1; if (sscanf(s, "userspace(pid=%lli)%n", &pid, &n) > 0 && n > 0) { - odp_put_userspace_action(pid, NULL, actions); + odp_put_userspace_action(pid, NULL, 0, actions); return n; } else if (sscanf(s, "userspace(pid=%lli,sFlow(vid=%i," "pcp=%i,output=%lli))%n", @@ -412,54 +477,51 @@ parse_odp_action(const char *s, const struct simap *port_names, cookie.type = USER_ACTION_COOKIE_SFLOW; cookie.sflow.vlan_tci = htons(tci); cookie.sflow.output = output; - odp_put_userspace_action(pid, &cookie, actions); + odp_put_userspace_action(pid, &cookie, sizeof cookie, actions); return n; - } else if (sscanf(s, "userspace(pid=%lli,slow_path(%n", &pid, &n) > 0 + } else if (sscanf(s, "userspace(pid=%lli,slow_path%n", &pid, &n) > 0 && n > 0) { union user_action_cookie cookie; + int res; cookie.type = USER_ACTION_COOKIE_SLOW_PATH; cookie.slow_path.unused = 0; cookie.slow_path.reason = 0; - while (s[n] != ')') { - uint32_t bit; - - for (bit = 1; bit; bit <<= 1) { - const char *reason = slow_path_reason_to_string(bit); - size_t len = strlen(reason); - - if (reason - && !strncmp(s + n, reason, len) - && (s[n + len] == ',' || s[n + len] == ')')) - { - cookie.slow_path.reason |= bit; - n += len + (s[n + len] == ','); - break; - } - } - - if (!bit) { - return -EINVAL; - } + res = parse_flags(&s[n], slow_path_reason_to_string, + &cookie.slow_path.reason); + if (res < 0) { + return res; } - if (s[n + 1] != ')') { + n += res; + if (s[n] != ')') { return -EINVAL; } - n += 2; + n++; - odp_put_userspace_action(pid, &cookie, actions); + odp_put_userspace_action(pid, &cookie, sizeof cookie, actions); return n; } else if (sscanf(s, "userspace(pid=%lli,userdata=" "%31[x0123456789abcdefABCDEF])%n", &pid, userdata_s, &n) > 0 && n > 0) { - union user_action_cookie cookie; uint64_t userdata; userdata = strtoull(userdata_s, NULL, 0); - memcpy(&cookie, &userdata, sizeof cookie); - odp_put_userspace_action(pid, &cookie, actions); + odp_put_userspace_action(pid, &userdata, sizeof(userdata), + actions); return n; + } else if (sscanf(s, "userspace(pid=%lli,userdata(%n", &pid, &n) > 0 + && n > 0) { + struct ofpbuf buf; + char *end; + + ofpbuf_init(&buf, 16); + end = ofpbuf_put_hex(&buf, &s[n], NULL); + if (end[0] == ')' && end[1] == ')') { + odp_put_userspace_action(pid, buf.data, buf.size, actions); + ofpbuf_uninit(&buf); + return (end + 2) - s; + } } } @@ -601,11 +663,13 @@ odp_flow_key_attr_len(uint16_t type) switch ((enum ovs_key_attr) type) { case OVS_KEY_ATTR_ENCAP: return -2; case OVS_KEY_ATTR_PRIORITY: return 4; - case OVS_KEY_ATTR_TUN_ID: return 8; + case OVS_KEY_ATTR_SKB_MARK: return 4; + case OVS_KEY_ATTR_TUNNEL: return -2; case OVS_KEY_ATTR_IN_PORT: return 4; case OVS_KEY_ATTR_ETHERNET: return sizeof(struct ovs_key_ethernet); case OVS_KEY_ATTR_VLAN: return sizeof(ovs_be16); case OVS_KEY_ATTR_ETHERTYPE: return 2; + case OVS_KEY_ATTR_MPLS: return sizeof(struct ovs_key_mpls); case OVS_KEY_ATTR_IPV4: return sizeof(struct ovs_key_ipv4); case OVS_KEY_ATTR_IPV6: return sizeof(struct ovs_key_ipv6); case OVS_KEY_ATTR_TCP: return sizeof(struct ovs_key_tcp); @@ -656,6 +720,111 @@ ovs_frag_type_to_string(enum ovs_frag_type type) } } +static int +tunnel_key_attr_len(int type) +{ + switch (type) { + case OVS_TUNNEL_KEY_ATTR_ID: return 8; + case OVS_TUNNEL_KEY_ATTR_IPV4_SRC: return 4; + case OVS_TUNNEL_KEY_ATTR_IPV4_DST: return 4; + case OVS_TUNNEL_KEY_ATTR_TOS: return 1; + case OVS_TUNNEL_KEY_ATTR_TTL: return 1; + case OVS_TUNNEL_KEY_ATTR_DONT_FRAGMENT: return 0; + case OVS_TUNNEL_KEY_ATTR_CSUM: return 0; + case __OVS_TUNNEL_KEY_ATTR_MAX: + return -1; + } + return -1; +} + +static enum odp_key_fitness +tun_key_from_attr(const struct nlattr *attr, struct flow_tnl *tun) +{ + unsigned int left; + const struct nlattr *a; + bool ttl = false; + bool unknown = false; + + NL_NESTED_FOR_EACH(a, left, attr) { + uint16_t type = nl_attr_type(a); + size_t len = nl_attr_get_size(a); + int expected_len = tunnel_key_attr_len(type); + + if (len != expected_len && expected_len >= 0) { + return ODP_FIT_ERROR; + } + + switch (type) { + case OVS_TUNNEL_KEY_ATTR_ID: + tun->tun_id = nl_attr_get_be64(a); + tun->flags |= FLOW_TNL_F_KEY; + break; + case OVS_TUNNEL_KEY_ATTR_IPV4_SRC: + tun->ip_src = nl_attr_get_be32(a); + break; + case OVS_TUNNEL_KEY_ATTR_IPV4_DST: + tun->ip_dst = nl_attr_get_be32(a); + break; + case OVS_TUNNEL_KEY_ATTR_TOS: + tun->ip_tos = nl_attr_get_u8(a); + break; + case OVS_TUNNEL_KEY_ATTR_TTL: + tun->ip_ttl = nl_attr_get_u8(a); + ttl = true; + break; + case OVS_TUNNEL_KEY_ATTR_DONT_FRAGMENT: + tun->flags |= FLOW_TNL_F_DONT_FRAGMENT; + break; + case OVS_TUNNEL_KEY_ATTR_CSUM: + tun->flags |= FLOW_TNL_F_CSUM; + break; + default: + /* Allow this to show up as unexpected, if there are unknown + * tunnel attribute, eventually resulting in ODP_FIT_TOO_MUCH. */ + unknown = true; + break; + } + } + + if (!ttl) { + return ODP_FIT_ERROR; + } + if (unknown) { + return ODP_FIT_TOO_MUCH; + } + return ODP_FIT_PERFECT; +} + +static void +tun_key_to_attr(struct ofpbuf *a, const struct flow_tnl *tun_key) +{ + size_t tun_key_ofs; + + tun_key_ofs = nl_msg_start_nested(a, OVS_KEY_ATTR_TUNNEL); + + if (tun_key->flags & FLOW_TNL_F_KEY) { + nl_msg_put_be64(a, OVS_TUNNEL_KEY_ATTR_ID, tun_key->tun_id); + } + if (tun_key->ip_src) { + nl_msg_put_be32(a, OVS_TUNNEL_KEY_ATTR_IPV4_SRC, tun_key->ip_src); + } + if (tun_key->ip_dst) { + nl_msg_put_be32(a, OVS_TUNNEL_KEY_ATTR_IPV4_DST, tun_key->ip_dst); + } + if (tun_key->ip_tos) { + nl_msg_put_u8(a, OVS_TUNNEL_KEY_ATTR_TOS, tun_key->ip_tos); + } + nl_msg_put_u8(a, OVS_TUNNEL_KEY_ATTR_TTL, tun_key->ip_ttl); + if (tun_key->flags & FLOW_TNL_F_DONT_FRAGMENT) { + nl_msg_put_flag(a, OVS_TUNNEL_KEY_ATTR_DONT_FRAGMENT); + } + if (tun_key->flags & FLOW_TNL_F_CSUM) { + nl_msg_put_flag(a, OVS_TUNNEL_KEY_ATTR_CSUM); + } + + nl_msg_end_nested(a, tun_key_ofs); +} + static void format_odp_key_attr(const struct nlattr *a, struct ds *ds) { @@ -668,6 +837,7 @@ format_odp_key_attr(const struct nlattr *a, struct ds *ds) const struct ovs_key_icmpv6 *icmpv6_key; const struct ovs_key_arp *arp_key; const struct ovs_key_nd *nd_key; + struct flow_tnl tun_key; enum ovs_key_attr attr = nl_attr_type(a); int expected_len; @@ -691,11 +861,29 @@ format_odp_key_attr(const struct nlattr *a, struct ds *ds) break; case OVS_KEY_ATTR_PRIORITY: - ds_put_format(ds, "(%"PRIu32")", nl_attr_get_u32(a)); + ds_put_format(ds, "(%#"PRIx32")", nl_attr_get_u32(a)); + break; + + case OVS_KEY_ATTR_SKB_MARK: + ds_put_format(ds, "(%#"PRIx32")", nl_attr_get_u32(a)); break; - case OVS_KEY_ATTR_TUN_ID: - ds_put_format(ds, "(%#"PRIx64")", ntohll(nl_attr_get_be64(a))); + case OVS_KEY_ATTR_TUNNEL: + memset(&tun_key, 0, sizeof tun_key); + if (tun_key_from_attr(a, &tun_key) == ODP_FIT_ERROR) { + ds_put_format(ds, "(error)"); + } else { + ds_put_format(ds, "(tun_id=0x%"PRIx64",src="IP_FMT",dst="IP_FMT"," + "tos=0x%"PRIx8",ttl=%"PRIu8",flags(", + ntohll(tun_key.tun_id), + IP_ARGS(tun_key.ip_src), + IP_ARGS(tun_key.ip_dst), + tun_key.ip_tos, tun_key.ip_ttl); + + format_flags(ds, flow_tun_flag_to_string, + (uint32_t) tun_key.flags, ','); + ds_put_format(ds, "))"); + } break; case OVS_KEY_ATTR_IN_PORT: @@ -715,6 +903,14 @@ format_odp_key_attr(const struct nlattr *a, struct ds *ds) ds_put_char(ds, ')'); break; + case OVS_KEY_ATTR_MPLS: { + const struct ovs_key_mpls *mpls_key = nl_attr_get(a); + ds_put_char(ds, '('); + format_mpls_lse(ds, mpls_key->mpls_top_lse); + ds_put_char(ds, ')'); + break; + } + case OVS_KEY_ATTR_ETHERTYPE: ds_put_format(ds, "(0x%04"PRIx16")", ntohs(nl_attr_get_be16(a))); @@ -724,8 +920,8 @@ format_odp_key_attr(const struct nlattr *a, struct ds *ds) ipv4_key = nl_attr_get(a); ds_put_format(ds, "(src="IP_FMT",dst="IP_FMT",proto=%"PRIu8 ",tos=%#"PRIx8",ttl=%"PRIu8",frag=%s)", - IP_ARGS(&ipv4_key->ipv4_src), - IP_ARGS(&ipv4_key->ipv4_dst), + IP_ARGS(ipv4_key->ipv4_src), + IP_ARGS(ipv4_key->ipv4_dst), ipv4_key->ipv4_proto, ipv4_key->ipv4_tos, ipv4_key->ipv4_ttl, ovs_frag_type_to_string(ipv4_key->ipv4_frag)); @@ -776,7 +972,7 @@ format_odp_key_attr(const struct nlattr *a, struct ds *ds) arp_key = nl_attr_get(a); ds_put_format(ds, "(sip="IP_FMT",tip="IP_FMT",op=%"PRIu16"," "sha="ETH_ADDR_FMT",tha="ETH_ADDR_FMT")", - IP_ARGS(&arp_key->arp_sip), IP_ARGS(&arp_key->arp_tip), + IP_ARGS(arp_key->arp_sip), IP_ARGS(arp_key->arp_tip), ntohs(arp_key->arp_op), ETH_ADDR_ARGS(arp_key->arp_sha), ETH_ADDR_ARGS(arp_key->arp_tha)); break; @@ -875,6 +1071,15 @@ ovs_frag_type_from_string(const char *s, enum ovs_frag_type *type) return true; } +static ovs_be32 +mpls_lse_from_components(int mpls_label, int mpls_tc, int mpls_ttl, int mpls_bos) +{ + return (htonl((mpls_label << MPLS_LABEL_SHIFT) | + (mpls_tc << MPLS_TC_SHIFT) | + (mpls_ttl << MPLS_TTL_SHIFT) | + (mpls_bos << MPLS_BOS_SHIFT))); +} + static int parse_odp_key_attr(const char *s, const struct simap *port_names, struct ofpbuf *key) @@ -893,20 +1098,52 @@ parse_odp_key_attr(const char *s, const struct simap *port_names, unsigned long long int priority; int n = -1; - if (sscanf(s, "priority(%lli)%n", &priority, &n) > 0 && n > 0) { + if (sscanf(s, "skb_priority(%llx)%n", &priority, &n) > 0 && n > 0) { nl_msg_put_u32(key, OVS_KEY_ATTR_PRIORITY, priority); return n; } } + { + unsigned long long int mark; + int n = -1; + + if (sscanf(s, "skb_mark(%llx)%n", &mark, &n) > 0 && n > 0) { + nl_msg_put_u32(key, OVS_KEY_ATTR_SKB_MARK, mark); + return n; + } + } + { char tun_id_s[32]; + int tos, ttl; + struct flow_tnl tun_key; int n = -1; - if (sscanf(s, "tun_id(%31[x0123456789abcdefABCDEF])%n", - tun_id_s, &n) > 0 && n > 0) { - uint64_t tun_id = strtoull(tun_id_s, NULL, 0); - nl_msg_put_be64(key, OVS_KEY_ATTR_TUN_ID, htonll(tun_id)); + if (sscanf(s, "tunnel(tun_id=%31[x0123456789abcdefABCDEF]," + "src="IP_SCAN_FMT",dst="IP_SCAN_FMT + ",tos=%i,ttl=%i,flags%n", tun_id_s, + IP_SCAN_ARGS(&tun_key.ip_src), + IP_SCAN_ARGS(&tun_key.ip_dst), &tos, &ttl, + &n) > 0 && n > 0) { + int res; + uint32_t flags; + + tun_key.tun_id = htonll(strtoull(tun_id_s, NULL, 0)); + tun_key.ip_tos = tos; + tun_key.ip_ttl = ttl; + res = parse_flags(&s[n], flow_tun_flag_to_string, &flags); + tun_key.flags = (uint16_t) flags; + + if (res < 0) { + return res; + } + n += res; + if (s[n] != ')') { + return -EINVAL; + } + n++; + tun_key_to_attr(key, &tun_key); return n; } } @@ -983,6 +1220,22 @@ parse_odp_key_attr(const char *s, const struct simap *port_names, } } + { + int label, tc, ttl, bos; + int n = -1; + + if (sscanf(s, "mpls(label=%"SCNi32",tc=%i,ttl=%i,bos=%i)%n", + &label, &tc, &ttl, &bos, &n) > 0 && + n > 0) { + struct ovs_key_mpls *mpls; + + mpls = nl_msg_put_unspec_uninit(key, OVS_KEY_ATTR_MPLS, + sizeof *mpls); + mpls->mpls_top_lse = mpls_lse_from_components(label, tc, ttl, bos); + return n; + } + } + { ovs_be32 ipv4_src; ovs_be32 ipv4_dst; @@ -1247,11 +1500,16 @@ ovs_to_odp_frag(uint8_t nw_frag) } /* Appends a representation of 'flow' as OVS_KEY_ATTR_* attributes to 'buf'. + * 'flow->in_port' is ignored (since it is likely to be an OpenFlow port + * number rather than a datapath port number). Instead, if 'odp_in_port' + * is anything other than OVSP_NONE, it is included in 'buf' as the input + * port. * * 'buf' must have at least ODPUTIL_FLOW_KEY_BYTES bytes of space, or be * capable of being expanded to allow for that much space. */ void -odp_flow_key_from_flow(struct ofpbuf *buf, const struct flow *flow) +odp_flow_key_from_flow(struct ofpbuf *buf, const struct flow *flow, + uint32_t odp_in_port) { struct ovs_key_ethernet *eth_key; size_t encap; @@ -1260,13 +1518,16 @@ odp_flow_key_from_flow(struct ofpbuf *buf, const struct flow *flow) nl_msg_put_u32(buf, OVS_KEY_ATTR_PRIORITY, flow->skb_priority); } - if (flow->tun_id != htonll(0)) { - nl_msg_put_be64(buf, OVS_KEY_ATTR_TUN_ID, flow->tun_id); + if (flow->tunnel.ip_dst) { + tun_key_to_attr(buf, &flow->tunnel); + } + + if (flow->skb_mark) { + nl_msg_put_u32(buf, OVS_KEY_ATTR_SKB_MARK, flow->skb_mark); } - if (flow->in_port != OFPP_NONE && flow->in_port != OFPP_CONTROLLER) { - nl_msg_put_u32(buf, OVS_KEY_ATTR_IN_PORT, - ofp_port_to_odp_port(flow->in_port)); + if (odp_in_port != OVSP_NONE) { + nl_msg_put_u32(buf, OVS_KEY_ATTR_IN_PORT, odp_in_port); } eth_key = nl_msg_put_unspec_uninit(buf, OVS_KEY_ATTR_ETHERNET, @@ -1314,7 +1575,8 @@ odp_flow_key_from_flow(struct ofpbuf *buf, const struct flow *flow) ipv6_key->ipv6_tclass = flow->nw_tos; ipv6_key->ipv6_hlimit = flow->nw_ttl; ipv6_key->ipv6_frag = ovs_to_odp_frag(flow->nw_frag); - } else if (flow->dl_type == htons(ETH_TYPE_ARP)) { + } else if (flow->dl_type == htons(ETH_TYPE_ARP) || + flow->dl_type == htons(ETH_TYPE_RARP)) { struct ovs_key_arp *arp_key; arp_key = nl_msg_put_unspec_uninit(buf, OVS_KEY_ATTR_ARP, @@ -1327,10 +1589,15 @@ odp_flow_key_from_flow(struct ofpbuf *buf, const struct flow *flow) memcpy(arp_key->arp_tha, flow->arp_tha, ETH_ADDR_LEN); } - if ((flow->dl_type == htons(ETH_TYPE_IP) - || flow->dl_type == htons(ETH_TYPE_IPV6)) - && !(flow->nw_frag & FLOW_NW_FRAG_LATER)) { + if (flow->mpls_depth) { + struct ovs_key_mpls *mpls_key; + + mpls_key = nl_msg_put_unspec_uninit(buf, OVS_KEY_ATTR_MPLS, + sizeof *mpls_key); + mpls_key->mpls_top_lse = flow->mpls_lse; + } + if (is_ip_any(flow) && !(flow->nw_frag & FLOW_NW_FRAG_LATER)) { if (flow->nw_proto == IPPROTO_TCP) { struct ovs_key_tcp *tcp_key; @@ -1447,6 +1714,7 @@ parse_flow_nlattrs(const struct nlattr *key, size_t key_len, uint64_t present_attrs; size_t left; + BUILD_ASSERT(OVS_KEY_ATTR_MAX < CHAR_BIT * sizeof present_attrs); present_attrs = 0; *out_of_range_attrp = 0; NL_ATTR_FOR_EACH (nla, left, key, key_len) { @@ -1461,7 +1729,7 @@ parse_flow_nlattrs(const struct nlattr *key, size_t key_len, return false; } - if (type >= CHAR_BIT * sizeof present_attrs) { + if (type > OVS_KEY_ATTR_MAX) { *out_of_range_attrp = type; } else { if (present_attrs & (UINT64_C(1) << type)) { @@ -1532,14 +1800,22 @@ parse_ethertype(const struct nlattr *attrs[OVS_KEY_ATTR_MAX + 1], } static enum odp_key_fitness -parse_l3_onward(const struct nlattr *attrs[OVS_KEY_ATTR_MAX + 1], - uint64_t present_attrs, int out_of_range_attr, - uint64_t expected_attrs, struct flow *flow, - const struct nlattr *key, size_t key_len) +parse_l2_5_onward(const struct nlattr *attrs[OVS_KEY_ATTR_MAX + 1], + uint64_t present_attrs, int out_of_range_attr, + uint64_t expected_attrs, struct flow *flow, + const struct nlattr *key, size_t key_len) { static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5); - if (flow->dl_type == htons(ETH_TYPE_IP)) { + if (eth_type_mpls(flow->dl_type)) { + expected_attrs |= (UINT64_C(1) << OVS_KEY_ATTR_MPLS); + + if (!(present_attrs & (UINT64_C(1) << OVS_KEY_ATTR_MPLS))) { + return ODP_FIT_TOO_LITTLE; + } + flow->mpls_lse = nl_attr_get_be32(attrs[OVS_KEY_ATTR_MPLS]); + flow->mpls_depth++; + } else if (flow->dl_type == htons(ETH_TYPE_IP)) { expected_attrs |= UINT64_C(1) << OVS_KEY_ATTR_IPV4; if (present_attrs & (UINT64_C(1) << OVS_KEY_ATTR_IPV4)) { const struct ovs_key_ipv4 *ipv4_key; @@ -1570,7 +1846,8 @@ parse_l3_onward(const struct nlattr *attrs[OVS_KEY_ATTR_MAX + 1], return ODP_FIT_ERROR; } } - } else if (flow->dl_type == htons(ETH_TYPE_ARP)) { + } else if (flow->dl_type == htons(ETH_TYPE_ARP) || + flow->dl_type == htons(ETH_TYPE_RARP)) { expected_attrs |= UINT64_C(1) << OVS_KEY_ATTR_ARP; if (present_attrs & (UINT64_C(1) << OVS_KEY_ATTR_ARP)) { const struct ovs_key_arp *arp_key; @@ -1709,8 +1986,8 @@ parse_8021q_onward(const struct nlattr *attrs[OVS_KEY_ATTR_MAX + 1], if (!parse_ethertype(attrs, present_attrs, &expected_attrs, flow)) { return ODP_FIT_ERROR; } - encap_fitness = parse_l3_onward(attrs, present_attrs, out_of_range_attr, - expected_attrs, flow, key, key_len); + encap_fitness = parse_l2_5_onward(attrs, present_attrs, out_of_range_attr, + expected_attrs, flow, key, key_len); /* The overall fitness is the worse of the outer and inner attributes. */ return MAX(fitness, encap_fitness); @@ -1720,6 +1997,10 @@ parse_8021q_onward(const struct nlattr *attrs[OVS_KEY_ATTR_MAX + 1], * structure in 'flow'. Returns an ODP_FIT_* value that indicates how well * 'key' fits our expectations for what a flow key should contain. * + * The 'in_port' will be the datapath's understanding of the port. The + * caller will need to translate with odp_port_to_ofp_port() if the + * OpenFlow port is needed. + * * This function doesn't take the packet itself as an argument because none of * the currently understood OVS_KEY_ATTR_* attributes require it. Currently, * it is always possible to infer which additional attribute(s) should appear @@ -1731,7 +2012,6 @@ enum odp_key_fitness odp_flow_key_to_flow(const struct nlattr *key, size_t key_len, struct flow *flow) { - static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5); const struct nlattr *attrs[OVS_KEY_ATTR_MAX + 1]; uint64_t expected_attrs; uint64_t present_attrs; @@ -1752,22 +2032,27 @@ odp_flow_key_to_flow(const struct nlattr *key, size_t key_len, expected_attrs |= UINT64_C(1) << OVS_KEY_ATTR_PRIORITY; } - if (present_attrs & (UINT64_C(1) << OVS_KEY_ATTR_TUN_ID)) { - flow->tun_id = nl_attr_get_be64(attrs[OVS_KEY_ATTR_TUN_ID]); - expected_attrs |= UINT64_C(1) << OVS_KEY_ATTR_TUN_ID; + if (present_attrs & (UINT64_C(1) << OVS_KEY_ATTR_SKB_MARK)) { + flow->skb_mark = nl_attr_get_u32(attrs[OVS_KEY_ATTR_SKB_MARK]); + expected_attrs |= UINT64_C(1) << OVS_KEY_ATTR_SKB_MARK; } - if (present_attrs & (UINT64_C(1) << OVS_KEY_ATTR_IN_PORT)) { - uint32_t in_port = nl_attr_get_u32(attrs[OVS_KEY_ATTR_IN_PORT]); - if (in_port >= UINT16_MAX || in_port >= OFPP_MAX) { - VLOG_ERR_RL(&rl, "in_port %"PRIu32" out of supported range", - in_port); + if (present_attrs & (UINT64_C(1) << OVS_KEY_ATTR_TUNNEL)) { + enum odp_key_fitness res; + + res = tun_key_from_attr(attrs[OVS_KEY_ATTR_TUNNEL], &flow->tunnel); + if (res == ODP_FIT_ERROR) { return ODP_FIT_ERROR; + } else if (res == ODP_FIT_PERFECT) { + expected_attrs |= UINT64_C(1) << OVS_KEY_ATTR_TUNNEL; } - flow->in_port = odp_port_to_ofp_port(in_port); + } + + if (present_attrs & (UINT64_C(1) << OVS_KEY_ATTR_IN_PORT)) { + flow->in_port = nl_attr_get_u32(attrs[OVS_KEY_ATTR_IN_PORT]); expected_attrs |= UINT64_C(1) << OVS_KEY_ATTR_IN_PORT; } else { - flow->in_port = OFPP_NONE; + flow->in_port = OVSP_NONE; } /* Ethernet header. */ @@ -1789,8 +2074,8 @@ odp_flow_key_to_flow(const struct nlattr *key, size_t key_len, return parse_8021q_onward(attrs, present_attrs, out_of_range_attr, expected_attrs, flow, key, key_len); } - return parse_l3_onward(attrs, present_attrs, out_of_range_attr, - expected_attrs, flow, key, key_len); + return parse_l2_5_onward(attrs, present_attrs, out_of_range_attr, + expected_attrs, flow, key, key_len); } /* Returns 'fitness' as a string, for use in debug messages. */ @@ -1812,25 +2097,39 @@ odp_key_fitness_to_string(enum odp_key_fitness fitness) } /* Appends an OVS_ACTION_ATTR_USERSPACE action to 'odp_actions' that specifies - * Netlink PID 'pid'. If 'cookie' is nonnull, adds a userdata attribute whose - * contents contains 'cookie' and returns the offset within 'odp_actions' of - * the start of the cookie. (If 'cookie' is null, then the return value is not - * meaningful.) */ + * Netlink PID 'pid'. If 'userdata' is nonnull, adds a userdata attribute + * whose contents are the 'userdata_size' bytes at 'userdata' and returns the + * offset within 'odp_actions' of the start of the cookie. (If 'userdata' is + * null, then the return value is not meaningful.) */ size_t -odp_put_userspace_action(uint32_t pid, const union user_action_cookie *cookie, +odp_put_userspace_action(uint32_t pid, + const void *userdata, size_t userdata_size, struct ofpbuf *odp_actions) { + size_t userdata_ofs; size_t offset; offset = nl_msg_start_nested(odp_actions, OVS_ACTION_ATTR_USERSPACE); nl_msg_put_u32(odp_actions, OVS_USERSPACE_ATTR_PID, pid); - if (cookie) { + if (userdata) { + userdata_ofs = odp_actions->size + NLA_HDRLEN; nl_msg_put_unspec(odp_actions, OVS_USERSPACE_ATTR_USERDATA, - cookie, sizeof *cookie); + userdata, userdata_size); + } else { + userdata_ofs = 0; } nl_msg_end_nested(odp_actions, offset); - return cookie ? odp_actions->size - NLA_ALIGN(sizeof *cookie) : 0; + return userdata_ofs; +} + +void +odp_put_tunnel_action(const struct flow_tnl *tunnel, + struct ofpbuf *odp_actions) +{ + size_t offset = nl_msg_start_nested(odp_actions, OVS_ACTION_ATTR_SET); + tun_key_to_attr(odp_actions, tunnel); + nl_msg_end_nested(odp_actions, offset); } /* The commit_odp_actions() function and its helpers. */ @@ -1844,17 +2143,33 @@ commit_set_action(struct ofpbuf *odp_actions, enum ovs_key_attr key_type, nl_msg_end_nested(odp_actions, offset); } -static void -commit_set_tun_id_action(const struct flow *flow, struct flow *base, +void +odp_put_skb_mark_action(const uint32_t skb_mark, + struct ofpbuf *odp_actions) +{ + commit_set_action(odp_actions, OVS_KEY_ATTR_SKB_MARK, &skb_mark, + sizeof(skb_mark)); +} + +/* If any of the flow key data that ODP actions can modify are different in + * 'base->tunnel' and 'flow->tunnel', appends a set_tunnel ODP action to + * 'odp_actions' that change the flow tunneling information in key from + * 'base->tunnel' into 'flow->tunnel', and then changes 'base->tunnel' in the + * same way. In other words, operates the same as commit_odp_actions(), but + * only on tunneling information. */ +void +commit_odp_tunnel_action(const struct flow *flow, struct flow *base, struct ofpbuf *odp_actions) { - if (base->tun_id == flow->tun_id) { + if (!memcmp(&base->tunnel, &flow->tunnel, sizeof base->tunnel)) { return; } - base->tun_id = flow->tun_id; + memcpy(&base->tunnel, &flow->tunnel, sizeof base->tunnel); - commit_set_action(odp_actions, OVS_KEY_ATTR_TUN_ID, - &base->tun_id, sizeof(base->tun_id)); + /* A valid IPV4_TUNNEL must have non-zero ip_dst. */ + if (flow->tunnel.ip_dst) { + odp_put_tunnel_action(&base->tunnel, odp_actions); + } } static void @@ -1901,6 +2216,50 @@ commit_vlan_action(const struct flow *flow, struct flow *base, base->vlan_tci = flow->vlan_tci; } +static void +commit_mpls_action(const struct flow *flow, struct flow *base, + struct ofpbuf *odp_actions) +{ + if (flow->mpls_lse == base->mpls_lse && + flow->mpls_depth == base->mpls_depth) { + return; + } + + if (flow->mpls_depth < base->mpls_depth) { + if (base->mpls_depth - flow->mpls_depth > 1) { + static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(10, 10); + VLOG_WARN_RL(&rl, "Multiple mpls_pop actions reduced to " + " a single mpls_pop action"); + } + + nl_msg_put_be16(odp_actions, OVS_ACTION_ATTR_POP_MPLS, flow->dl_type); + } else if (flow->mpls_depth > base->mpls_depth) { + struct ovs_action_push_mpls *mpls; + + if (flow->mpls_depth - base->mpls_depth > 1) { + static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(10, 10); + VLOG_WARN_RL(&rl, "Multiple mpls_push actions reduced to " + " a single mpls_push action"); + } + + mpls = nl_msg_put_unspec_uninit(odp_actions, OVS_ACTION_ATTR_PUSH_MPLS, + sizeof *mpls); + memset(mpls, 0, sizeof *mpls); + mpls->mpls_ethertype = flow->dl_type; + mpls->mpls_lse = flow->mpls_lse; + } else { + struct ovs_key_mpls mpls_key; + + mpls_key.mpls_top_lse = flow->mpls_lse; + commit_set_action(odp_actions, OVS_KEY_ATTR_MPLS, + &mpls_key, sizeof(mpls_key)); + } + + base->dl_type = flow->dl_type; + base->mpls_lse = flow->mpls_lse; + base->mpls_depth = flow->mpls_depth; +} + static void commit_set_ipv4_action(const struct flow *flow, struct flow *base, struct ofpbuf *odp_actions) @@ -1976,7 +2335,7 @@ static void commit_set_port_action(const struct flow *flow, struct flow *base, struct ofpbuf *odp_actions) { - if (!base->tp_src && !base->tp_dst) { + if (!is_ip_any(base) || (!base->tp_src && !base->tp_dst)) { return; } @@ -2018,17 +2377,35 @@ commit_set_priority_action(const struct flow *flow, struct flow *base, &base->skb_priority, sizeof(base->skb_priority)); } +static void +commit_set_skb_mark_action(const struct flow *flow, struct flow *base, + struct ofpbuf *odp_actions) +{ + if (base->skb_mark == flow->skb_mark) { + return; + } + base->skb_mark = flow->skb_mark; + + odp_put_skb_mark_action(base->skb_mark, odp_actions); +} /* If any of the flow key data that ODP actions can modify are different in * 'base' and 'flow', appends ODP actions to 'odp_actions' that change the flow - * key from 'base' into 'flow', and then changes 'base' the same way. */ + * key from 'base' into 'flow', and then changes 'base' the same way. Does not + * commit set_tunnel actions. Users should call commit_odp_tunnel_action() + * in addition to this function if needed. */ void commit_odp_actions(const struct flow *flow, struct flow *base, struct ofpbuf *odp_actions) { - commit_set_tun_id_action(flow, base, odp_actions); commit_set_ether_addr_action(flow, base, odp_actions); commit_vlan_action(flow, base, odp_actions); commit_set_nw_action(flow, base, odp_actions); commit_set_port_action(flow, base, odp_actions); + /* Commiting MPLS actions should occur after committing nw and port + * actions. This is because committing MPLS actions may alter a packet so + * that it is no longer IP and thus nw and port actions are no longer valid. + */ + commit_mpls_action(flow, base, odp_actions); commit_set_priority_action(flow, base, odp_actions); + commit_set_skb_mark_action(flow, base, odp_actions); }