X-Git-Url: http://git.onelab.eu/?a=blobdiff_plain;f=lib%2Fodp-util.c;h=030ad5e5a2f4ae5720180e717adc1af67d20b5d6;hb=ec9f40dce11c7e81bc41d42e3bbfaaf8287165ce;hp=9657595084de2bafd3da8d4df5eefc4f291b5686;hpb=d9065a90b6b955aa38586c952e8804ca7a22547e;p=sliver-openvswitch.git diff --git a/lib/odp-util.c b/lib/odp-util.c index 965759508..030ad5e5a 100644 --- a/lib/odp-util.c +++ b/lib/odp-util.c @@ -1,5 +1,5 @@ /* - * Copyright (c) 2009, 2010, 2011 Nicira Networks. + * 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. @@ -14,11 +14,13 @@ * limitations under the License. */ -#include #include +#include #include "odp-util.h" #include #include +#include +#include #include #include #include @@ -28,10 +30,13 @@ #include "flow.h" #include "netlink.h" #include "ofpbuf.h" -#include "openvswitch/tunnel.h" #include "packets.h" +#include "simap.h" #include "timeval.h" #include "util.h" +#include "vlog.h" + +VLOG_DEFINE_THIS_MODULE(odp_util); /* The interface between userspace and kernel uses an "OVS_*" prefix. * Since this is fairly non-specific for the OVS userspace components, @@ -39,28 +44,39 @@ * interactions with the datapath. */ -int +/* The set of characters that may separate one action or one key attribute + * from another. */ +static const char *delimiters = ", \t\r\n"; + +static int parse_odp_key_attr(const char *, const struct simap *port_names, + struct ofpbuf *); +static void format_odp_key_attr(const struct nlattr *a, struct ds *ds); + +/* Returns one the following for the action with the given OVS_ACTION_ATTR_* + * 'type': + * + * - For an action whose argument has a fixed length, returned that + * nonnegative length in bytes. + * + * - For an action with a variable-length argument, returns -2. + * + * - For an invalid 'type', returns -1. */ +static int odp_action_len(uint16_t type) { if (type > OVS_ACTION_ATTR_MAX) { return -1; } - switch ((enum ovs_action_type) type) { - case OVS_ACTION_ATTR_OUTPUT: return 4; - case OVS_ACTION_ATTR_USERSPACE: return 8; - case OVS_ACTION_ATTR_PUSH_VLAN: return 2; + switch ((enum ovs_action_attr) type) { + case OVS_ACTION_ATTR_OUTPUT: return sizeof(uint32_t); + 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_SET_DL_SRC: return ETH_ADDR_LEN; - case OVS_ACTION_ATTR_SET_DL_DST: return ETH_ADDR_LEN; - case OVS_ACTION_ATTR_SET_NW_SRC: return 4; - case OVS_ACTION_ATTR_SET_NW_DST: return 4; - case OVS_ACTION_ATTR_SET_NW_TOS: return 1; - case OVS_ACTION_ATTR_SET_TP_SRC: return 2; - case OVS_ACTION_ATTR_SET_TP_DST: return 2; - case OVS_ACTION_ATTR_SET_TUNNEL: return 8; - case OVS_ACTION_ATTR_SET_PRIORITY: return 4; - case OVS_ACTION_ATTR_POP_PRIORITY: 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; case OVS_ACTION_ATTR_UNSPEC: case __OVS_ACTION_ATTR_MAX: @@ -70,6 +86,39 @@ odp_action_len(uint16_t type) return -1; } +static const char * +ovs_key_attr_to_string(enum ovs_key_attr attr) +{ + static char unknown_attr[3 + INT_STRLEN(unsigned int) + 1]; + + switch (attr) { + case OVS_KEY_ATTR_UNSPEC: return "unspec"; + case OVS_KEY_ATTR_ENCAP: return "encap"; + 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"; + case OVS_KEY_ATTR_ETHERTYPE: return "eth_type"; + case OVS_KEY_ATTR_IPV4: return "ipv4"; + case OVS_KEY_ATTR_IPV6: return "ipv6"; + case OVS_KEY_ATTR_TCP: return "tcp"; + case OVS_KEY_ATTR_UDP: return "udp"; + case OVS_KEY_ATTR_ICMP: return "icmp"; + 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_MPLS: return "mpls"; + + case __OVS_KEY_ATTR_MAX: + default: + snprintf(unknown_attr, sizeof unknown_attr, "key%u", + (unsigned int) attr); + return unknown_attr; + } +} + static void format_generic_odp_action(struct ds *ds, const struct nlattr *a) { @@ -89,69 +138,277 @@ format_generic_odp_action(struct ds *ds, const struct nlattr *a) } } -void +static void +format_odp_sample_action(struct ds *ds, const struct nlattr *attr) +{ + static const struct nl_policy ovs_sample_policy[] = { + [OVS_SAMPLE_ATTR_PROBABILITY] = { .type = NL_A_U32 }, + [OVS_SAMPLE_ATTR_ACTIONS] = { .type = NL_A_NESTED } + }; + struct nlattr *a[ARRAY_SIZE(ovs_sample_policy)]; + double percentage; + const struct nlattr *nla_acts; + int len; + + ds_put_cstr(ds, "sample"); + + if (!nl_parse_nested(attr, ovs_sample_policy, a, ARRAY_SIZE(a))) { + ds_put_cstr(ds, "(error)"); + return; + } + + percentage = (100.0 * nl_attr_get_u32(a[OVS_SAMPLE_ATTR_PROBABILITY])) / + UINT32_MAX; + + ds_put_format(ds, "(sample=%.1f%%,", percentage); + + ds_put_cstr(ds, "actions("); + nla_acts = nl_attr_get(a[OVS_SAMPLE_ATTR_ACTIONS]); + len = nl_attr_get_size(a[OVS_SAMPLE_ATTR_ACTIONS]); + format_odp_actions(ds, nla_acts, len); + ds_put_format(ds, "))"); +} + +static const char * +slow_path_reason_to_string(enum slow_path_reason reason) +{ + switch (reason) { + case SLOW_CFM: + return "cfm"; + case SLOW_LACP: + return "lacp"; + case SLOW_STP: + return "stp"; + case SLOW_BFD: + return "bfd"; + case SLOW_CONTROLLER: + return "controller"; + case __SLOW_MAX: + default: + return NULL; + } +} + +static enum slow_path_reason +string_to_slow_path_reason(const char *string) +{ + enum slow_path_reason i; + + for (i = 1; i < __SLOW_MAX; i++) { + if (!strcmp(string, slow_path_reason_to_string(i))) { + return i; + } + } + + return 0; +} + +static int +parse_flags(const char *s, const char *(*bit_to_string)(uint32_t), + uint32_t *res) +{ + uint32_t result = 0; + int n = 0; + + if (s[n] != '(') { + return -EINVAL; + } + n++; + + 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; + } + + 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 (!bit) { + return -EINVAL; + } + } + n++; + + *res = result; + return n; +} + +static void +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_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)"); + return; + } + + ds_put_format(ds, "userspace(pid=%"PRIu32, + nl_attr_get_u32(a[OVS_USERSPACE_ATTR_PID])); + + userdata_attr = a[OVS_USERSPACE_ATTR_USERDATA]; + + if (userdata_attr) { + const uint8_t *userdata = nl_attr_get(userdata_attr); + size_t userdata_len = nl_attr_get_size(userdata_attr); + bool userdata_unspec = true; + union user_action_cookie cookie; + + if (userdata_len >= sizeof cookie.type + && userdata_len <= sizeof cookie) { + + memset(&cookie, 0, sizeof cookie); + memcpy(&cookie, userdata, userdata_len); + + userdata_unspec = false; + + if (userdata_len == sizeof cookie.sflow + && cookie.type == USER_ACTION_COOKIE_SFLOW) { + ds_put_format(ds, ",sFlow(" + "vid=%"PRIu16",pcp=%"PRIu8",output=%"PRIu32")", + vlan_tci_to_vid(cookie.sflow.vlan_tci), + vlan_tci_to_pcp(cookie.sflow.vlan_tci), + cookie.sflow.output); + } else if (userdata_len == sizeof cookie.slow_path + && cookie.type == USER_ACTION_COOKIE_SLOW_PATH) { + const char *reason; + reason = slow_path_reason_to_string(cookie.slow_path.reason); + reason = reason ? reason : ""; + ds_put_format(ds, ",slow_path(%s)", reason); + } else if (userdata_len == sizeof cookie.flow_sample + && cookie.type == USER_ACTION_COOKIE_FLOW_SAMPLE) { + ds_put_format(ds, ",flow_sample(probability=%"PRIu16 + ",collector_set_id=%"PRIu32 + ",obs_domain_id=%"PRIu32 + ",obs_point_id=%"PRIu32")", + cookie.flow_sample.probability, + cookie.flow_sample.collector_set_id, + cookie.flow_sample.obs_domain_id, + cookie.flow_sample.obs_point_id); + } else if (userdata_len == sizeof cookie.ipfix + && cookie.type == USER_ACTION_COOKIE_IPFIX) { + ds_put_format(ds, ",ipfix"); + } else { + userdata_unspec = true; + } + } + + if (userdata_unspec) { + size_t i; + ds_put_format(ds, ",userdata("); + for (i = 0; i < userdata_len; i++) { + ds_put_format(ds, "%02x", userdata[i]); + } + ds_put_char(ds, ')'); + } + } + + ds_put_char(ds, ')'); +} + +static void +format_vlan_tci(struct ds *ds, ovs_be16 vlan_tci) +{ + ds_put_format(ds, "vid=%"PRIu16",pcp=%d", + vlan_tci_to_vid(vlan_tci), + vlan_tci_to_pcp(vlan_tci)); + if (!(vlan_tci & htons(VLAN_CFI))) { + ds_put_cstr(ds, ",cfi=0"); + } +} + +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) { - const uint8_t *eth; - ovs_be32 ip; + int expected_len; + enum ovs_action_attr type = nl_attr_type(a); + const struct ovs_action_push_vlan *vlan; - if (nl_attr_get_size(a) != odp_action_len(nl_attr_type(a))) { + expected_len = odp_action_len(nl_attr_type(a)); + if (expected_len != -2 && nl_attr_get_size(a) != expected_len) { ds_put_format(ds, "bad length %zu, expected %d for: ", - nl_attr_get_size(a), odp_action_len(nl_attr_type(a))); + nl_attr_get_size(a), expected_len); format_generic_odp_action(ds, a); return; } - switch (nl_attr_type(a)) { + switch (type) { case OVS_ACTION_ATTR_OUTPUT: ds_put_format(ds, "%"PRIu16, nl_attr_get_u32(a)); break; case OVS_ACTION_ATTR_USERSPACE: - ds_put_format(ds, "userspace(%"PRIu64")", nl_attr_get_u64(a)); + format_odp_userspace_action(ds, a); break; - case OVS_ACTION_ATTR_SET_TUNNEL: - ds_put_format(ds, "set_tunnel(%#"PRIx64")", - ntohll(nl_attr_get_be64(a))); + case OVS_ACTION_ATTR_SET: + ds_put_cstr(ds, "set("); + format_odp_key_attr(nl_attr_get(a), ds); + ds_put_cstr(ds, ")"); break; case OVS_ACTION_ATTR_PUSH_VLAN: - ds_put_format(ds, "push_vlan(vid=%"PRIu16",pcp=%d)", - vlan_tci_to_vid(nl_attr_get_be16(a)), - vlan_tci_to_pcp(nl_attr_get_be16(a))); + vlan = nl_attr_get(a); + ds_put_cstr(ds, "push_vlan("); + if (vlan->vlan_tpid != htons(ETH_TYPE_VLAN)) { + ds_put_format(ds, "tpid=0x%04"PRIx16",", ntohs(vlan->vlan_tpid)); + } + format_vlan_tci(ds, vlan->vlan_tci); + ds_put_char(ds, ')'); break; case OVS_ACTION_ATTR_POP_VLAN: - ds_put_format(ds, "pop_vlan"); - break; - case OVS_ACTION_ATTR_SET_DL_SRC: - eth = nl_attr_get_unspec(a, ETH_ADDR_LEN); - ds_put_format(ds, "set_dl_src("ETH_ADDR_FMT")", ETH_ADDR_ARGS(eth)); + ds_put_cstr(ds, "pop_vlan"); break; - case OVS_ACTION_ATTR_SET_DL_DST: - eth = nl_attr_get_unspec(a, ETH_ADDR_LEN); - ds_put_format(ds, "set_dl_dst("ETH_ADDR_FMT")", ETH_ADDR_ARGS(eth)); + 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_SET_NW_SRC: - ip = nl_attr_get_be32(a); - ds_put_format(ds, "set_nw_src("IP_FMT")", IP_ARGS(&ip)); - break; - case OVS_ACTION_ATTR_SET_NW_DST: - ip = nl_attr_get_be32(a); - ds_put_format(ds, "set_nw_dst("IP_FMT")", IP_ARGS(&ip)); - break; - case OVS_ACTION_ATTR_SET_NW_TOS: - ds_put_format(ds, "set_nw_tos(%"PRIu8")", nl_attr_get_u8(a)); - break; - case OVS_ACTION_ATTR_SET_TP_SRC: - ds_put_format(ds, "set_tp_src(%"PRIu16")", ntohs(nl_attr_get_be16(a))); - break; - case OVS_ACTION_ATTR_SET_TP_DST: - ds_put_format(ds, "set_tp_dst(%"PRIu16")", ntohs(nl_attr_get_be16(a))); - break; - case OVS_ACTION_ATTR_SET_PRIORITY: - ds_put_format(ds, "set_priority(%#"PRIx32")", nl_attr_get_u32(a)); + } + 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_POP_PRIORITY: - ds_put_cstr(ds, "pop_priority"); + } + case OVS_ACTION_ATTR_SAMPLE: + format_odp_sample_action(ds, a); break; + case OVS_ACTION_ATTR_UNSPEC: + case __OVS_ACTION_ATTR_MAX: default: format_generic_odp_action(ds, a); break; @@ -173,18 +430,278 @@ format_odp_actions(struct ds *ds, const struct nlattr *actions, format_odp_action(ds, a); } if (left) { + int i; + if (left == actions_len) { ds_put_cstr(ds, ""); } - ds_put_format(ds, ",***%u leftover bytes***", left); + ds_put_format(ds, ",***%u leftover bytes*** (", left); + for (i = 0; i < left; i++) { + ds_put_format(ds, "%02x", ((const uint8_t *) a)[i]); + } + ds_put_char(ds, ')'); } } else { ds_put_cstr(ds, "drop"); } } + +static int +parse_odp_action(const char *s, const struct simap *port_names, + struct ofpbuf *actions) +{ + /* Many of the sscanf calls in this function use oversized destination + * fields because some sscanf() implementations truncate the range of %i + * directives, so that e.g. "%"SCNi16 interprets input of "0xfedc" as a + * value of 0x7fff. The other alternatives are to allow only a single + * radix (e.g. decimal or hexadecimal) or to write more sophisticated + * parsers. + * + * The tun_id parser has to use an alternative approach because there is no + * type larger than 64 bits. */ + + { + unsigned long long int port; + int n = -1; + + if (sscanf(s, "%lli%n", &port, &n) > 0 && n > 0) { + nl_msg_put_u32(actions, OVS_ACTION_ATTR_OUTPUT, port); + return n; + } + } + + if (port_names) { + int len = strcspn(s, delimiters); + struct simap_node *node; + + node = simap_find_len(port_names, s, len); + if (node) { + nl_msg_put_u32(actions, OVS_ACTION_ATTR_OUTPUT, node->data); + return len; + } + } + + { + unsigned long long int pid; + unsigned long long int output; + unsigned long long int probability; + unsigned long long int collector_set_id; + unsigned long long int obs_domain_id; + unsigned long long int obs_point_id; + int vid, pcp; + int n = -1; + + if (sscanf(s, "userspace(pid=%lli)%n", &pid, &n) > 0 && n > 0) { + 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", + &pid, &vid, &pcp, &output, &n) > 0 && n > 0) { + union user_action_cookie cookie; + uint16_t tci; + + tci = vid | (pcp << VLAN_PCP_SHIFT); + if (tci) { + tci |= VLAN_CFI; + } + + cookie.type = USER_ACTION_COOKIE_SFLOW; + cookie.sflow.vlan_tci = htons(tci); + cookie.sflow.output = output; + odp_put_userspace_action(pid, &cookie, sizeof cookie.sflow, + actions); + return n; + } else if (sscanf(s, "userspace(pid=%lli,slow_path(%n", &pid, &n) > 0 + && n > 0) { + union user_action_cookie cookie; + char reason[32]; + + if (s[n] == ')' && s[n + 1] == ')') { + reason[0] = '\0'; + n += 2; + } else if (sscanf(s + n, "%31[^)]))", reason) > 0) { + n += strlen(reason) + 2; + } else { + return -EINVAL; + } + + cookie.type = USER_ACTION_COOKIE_SLOW_PATH; + cookie.slow_path.unused = 0; + cookie.slow_path.reason = string_to_slow_path_reason(reason); + + if (reason[0] && !cookie.slow_path.reason) { + return -EINVAL; + } + + odp_put_userspace_action(pid, &cookie, sizeof cookie.slow_path, + actions); + return n; + } else if (sscanf(s, "userspace(pid=%lli,flow_sample(probability=%lli," + "collector_set_id=%lli,obs_domain_id=%lli," + "obs_point_id=%lli))%n", + &pid, &probability, &collector_set_id, + &obs_domain_id, &obs_point_id, &n) > 0 && n > 0) { + union user_action_cookie cookie; + + cookie.type = USER_ACTION_COOKIE_FLOW_SAMPLE; + cookie.flow_sample.probability = probability; + cookie.flow_sample.collector_set_id = collector_set_id; + cookie.flow_sample.obs_domain_id = obs_domain_id; + cookie.flow_sample.obs_point_id = obs_point_id; + odp_put_userspace_action(pid, &cookie, sizeof cookie.flow_sample, + actions); + return n; + } else if (sscanf(s, "userspace(pid=%lli,ipfix)%n", &pid, &n) > 0 + && n > 0) { + union user_action_cookie cookie; + + cookie.type = USER_ACTION_COOKIE_IPFIX; + odp_put_userspace_action(pid, &cookie, sizeof cookie.ipfix, + 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; + } + } + } + + if (!strncmp(s, "set(", 4)) { + size_t start_ofs; + int retval; + + start_ofs = nl_msg_start_nested(actions, OVS_ACTION_ATTR_SET); + retval = parse_odp_key_attr(s + 4, port_names, actions); + if (retval < 0) { + return retval; + } + if (s[retval + 4] != ')') { + return -EINVAL; + } + nl_msg_end_nested(actions, start_ofs); + return retval + 5; + } + + { + struct ovs_action_push_vlan push; + int tpid = ETH_TYPE_VLAN; + int vid, pcp; + int cfi = 1; + int n = -1; + + if ((sscanf(s, "push_vlan(vid=%i,pcp=%i)%n", &vid, &pcp, &n) > 0 + && n > 0) + || (sscanf(s, "push_vlan(vid=%i,pcp=%i,cfi=%i)%n", + &vid, &pcp, &cfi, &n) > 0 && n > 0) + || (sscanf(s, "push_vlan(tpid=%i,vid=%i,pcp=%i)%n", + &tpid, &vid, &pcp, &n) > 0 && n > 0) + || (sscanf(s, "push_vlan(tpid=%i,vid=%i,pcp=%i,cfi=%i)%n", + &tpid, &vid, &pcp, &cfi, &n) > 0 && n > 0)) { + push.vlan_tpid = htons(tpid); + push.vlan_tci = htons((vid << VLAN_VID_SHIFT) + | (pcp << VLAN_PCP_SHIFT) + | (cfi ? VLAN_CFI : 0)); + nl_msg_put_unspec(actions, OVS_ACTION_ATTR_PUSH_VLAN, + &push, sizeof push); + + return n; + } + } + + if (!strncmp(s, "pop_vlan", 8)) { + nl_msg_put_flag(actions, OVS_ACTION_ATTR_POP_VLAN); + return 8; + } + + { + double percentage; + int n = -1; + + if (sscanf(s, "sample(sample=%lf%%,actions(%n", &percentage, &n) > 0 + && percentage >= 0. && percentage <= 100.0 + && n > 0) { + size_t sample_ofs, actions_ofs; + double probability; + + probability = floor(UINT32_MAX * (percentage / 100.0) + .5); + sample_ofs = nl_msg_start_nested(actions, OVS_ACTION_ATTR_SAMPLE); + nl_msg_put_u32(actions, OVS_SAMPLE_ATTR_PROBABILITY, + (probability <= 0 ? 0 + : probability >= UINT32_MAX ? UINT32_MAX + : probability)); + + actions_ofs = nl_msg_start_nested(actions, + OVS_SAMPLE_ATTR_ACTIONS); + for (;;) { + int retval; + + n += strspn(s + n, delimiters); + if (s[n] == ')') { + break; + } + + retval = parse_odp_action(s + n, port_names, actions); + if (retval < 0) { + return retval; + } + n += retval; + } + nl_msg_end_nested(actions, actions_ofs); + nl_msg_end_nested(actions, sample_ofs); + + return s[n + 1] == ')' ? n + 2 : -EINVAL; + } + } + + return -EINVAL; +} + +/* Parses the string representation of datapath actions, in the format output + * by format_odp_action(). Returns 0 if successful, otherwise a positive errno + * value. On success, the ODP actions are appended to 'actions' as a series of + * Netlink attributes. On failure, no data is appended to 'actions'. Either + * way, 'actions''s data might be reallocated. */ +int +odp_actions_from_string(const char *s, const struct simap *port_names, + struct ofpbuf *actions) +{ + size_t old_size; + + if (!strcasecmp(s, "drop")) { + return 0; + } + + old_size = actions->size; + for (;;) { + int retval; + + s += strspn(s, delimiters); + if (!*s) { + return 0; + } + + retval = parse_odp_action(s, port_names, actions); + if (retval < 0 || !strchr(delimiters, s[retval])) { + actions->size = old_size; + return -retval; + } + s += retval; + } + + return 0; +} /* Returns the correct length of the payload for a flow key attribute of the - * specified 'type', or -1 if 'type' is unknown. */ + * specified 'type', -1 if 'type' is unknown, or -2 if the attribute's payload + * is variable length. */ static int odp_flow_key_attr_len(uint16_t type) { @@ -192,12 +709,16 @@ odp_flow_key_attr_len(uint16_t type) return -1; } - switch ((enum ovs_key_type) type) { - case OVS_KEY_ATTR_TUN_ID: return 8; + switch ((enum ovs_key_attr) type) { + case OVS_KEY_ATTR_ENCAP: return -2; + case OVS_KEY_ATTR_PRIORITY: return 4; + 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_8021Q: return sizeof(struct ovs_key_8021q); + 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); @@ -215,13 +736,10 @@ odp_flow_key_attr_len(uint16_t type) return -1; } - static void format_generic_odp_key(const struct nlattr *a, struct ds *ds) { size_t len = nl_attr_get_size(a); - - ds_put_format(ds, "key%"PRId16, nl_attr_type(a)); if (len) { const uint8_t *unspec; unsigned int i; @@ -235,11 +753,131 @@ format_generic_odp_key(const struct nlattr *a, struct ds *ds) } } +static const char * +ovs_frag_type_to_string(enum ovs_frag_type type) +{ + switch (type) { + case OVS_FRAG_TYPE_NONE: + return "no"; + case OVS_FRAG_TYPE_FIRST: + return "first"; + case OVS_FRAG_TYPE_LATER: + return "later"; + case __OVS_FRAG_TYPE_MAX: + default: + return ""; + } +} + +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) { const struct ovs_key_ethernet *eth_key; - const struct ovs_key_8021q *q_key; const struct ovs_key_ipv4 *ipv4_key; const struct ovs_key_ipv6 *ipv6_key; const struct ovs_key_tcp *tcp_key; @@ -248,54 +886,94 @@ 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; - - if (nl_attr_get_size(a) != odp_flow_key_attr_len(nl_attr_type(a))) { - ds_put_format(ds, "bad length %zu, expected %d for: ", + struct flow_tnl tun_key; + enum ovs_key_attr attr = nl_attr_type(a); + int expected_len; + + ds_put_cstr(ds, ovs_key_attr_to_string(attr)); + expected_len = odp_flow_key_attr_len(nl_attr_type(a)); + if (expected_len != -2 && nl_attr_get_size(a) != expected_len) { + ds_put_format(ds, "(bad length %zu, expected %d)", nl_attr_get_size(a), odp_flow_key_attr_len(nl_attr_type(a))); format_generic_odp_key(a, ds); return; } - switch (nl_attr_type(a)) { - case OVS_KEY_ATTR_TUN_ID: - ds_put_format(ds, "tun_id(%#"PRIx64")", ntohll(nl_attr_get_be64(a))); + switch (attr) { + case OVS_KEY_ATTR_ENCAP: + ds_put_cstr(ds, "("); + if (nl_attr_get_size(a)) { + odp_flow_key_format(nl_attr_get(a), nl_attr_get_size(a), ds); + } + ds_put_char(ds, ')'); + break; + + case OVS_KEY_ATTR_PRIORITY: + 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_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: - ds_put_format(ds, "in_port(%"PRIu32")", nl_attr_get_u32(a)); + ds_put_format(ds, "(%"PRIu32")", nl_attr_get_u32(a)); break; case OVS_KEY_ATTR_ETHERNET: eth_key = nl_attr_get(a); - ds_put_format(ds, "eth(src="ETH_ADDR_FMT",dst="ETH_ADDR_FMT")", + ds_put_format(ds, "(src="ETH_ADDR_FMT",dst="ETH_ADDR_FMT")", ETH_ADDR_ARGS(eth_key->eth_src), ETH_ADDR_ARGS(eth_key->eth_dst)); break; - case OVS_KEY_ATTR_8021Q: - q_key = nl_attr_get(a); - ds_put_cstr(ds, "vlan("); - if (q_key->q_tpid != htons(ETH_TYPE_VLAN)) { - ds_put_format(ds, "tpid=0x%04"PRIx16",", ntohs(q_key->q_tpid)); - } - ds_put_format(ds, "vid=%"PRIu16",pcp=%d)", - vlan_tci_to_vid(q_key->q_tci), - vlan_tci_to_pcp(q_key->q_tci)); + case OVS_KEY_ATTR_VLAN: + ds_put_char(ds, '('); + format_vlan_tci(ds, nl_attr_get_be16(a)); + ds_put_char(ds, ')'); break; - case OVS_KEY_ATTR_ETHERTYPE: - ds_put_format(ds, "eth_type(0x%04"PRIx16")", - ntohs(nl_attr_get_be16(a))); + 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_lse); + ds_put_char(ds, ')'); + break; + } + + case OVS_KEY_ATTR_ETHERTYPE: + ds_put_format(ds, "(0x%04"PRIx16")", + ntohs(nl_attr_get_be16(a))); break; case OVS_KEY_ATTR_IPV4: ipv4_key = nl_attr_get(a); - ds_put_format(ds, "ipv4(src="IP_FMT",dst="IP_FMT"," - "proto=%"PRId8",tos=%"PRIu8")", - IP_ARGS(&ipv4_key->ipv4_src), - IP_ARGS(&ipv4_key->ipv4_dst), - ipv4_key->ipv4_proto, ipv4_key->ipv4_tos); + 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), + ipv4_key->ipv4_proto, ipv4_key->ipv4_tos, + ipv4_key->ipv4_ttl, + ovs_frag_type_to_string(ipv4_key->ipv4_frag)); break; case OVS_KEY_ATTR_IPV6: { @@ -306,41 +984,44 @@ format_odp_key_attr(const struct nlattr *a, struct ds *ds) inet_ntop(AF_INET6, ipv6_key->ipv6_src, src_str, sizeof src_str); inet_ntop(AF_INET6, ipv6_key->ipv6_dst, dst_str, sizeof dst_str); - ds_put_format(ds, "ipv6(src=%s,dst=%s,proto=%"PRId8",tos=%"PRIu8")", - src_str, dst_str, ipv6_key->ipv6_proto, - ipv6_key->ipv6_tos); + ds_put_format(ds, "(src=%s,dst=%s,label=%#"PRIx32",proto=%"PRIu8 + ",tclass=%#"PRIx8",hlimit=%"PRIu8",frag=%s)", + src_str, dst_str, ntohl(ipv6_key->ipv6_label), + ipv6_key->ipv6_proto, ipv6_key->ipv6_tclass, + ipv6_key->ipv6_hlimit, + ovs_frag_type_to_string(ipv6_key->ipv6_frag)); break; } case OVS_KEY_ATTR_TCP: tcp_key = nl_attr_get(a); - ds_put_format(ds, "tcp(src=%"PRIu16",dst=%"PRIu16")", + ds_put_format(ds, "(src=%"PRIu16",dst=%"PRIu16")", ntohs(tcp_key->tcp_src), ntohs(tcp_key->tcp_dst)); break; case OVS_KEY_ATTR_UDP: udp_key = nl_attr_get(a); - ds_put_format(ds, "udp(src=%"PRIu16",dst=%"PRIu16")", + ds_put_format(ds, "(src=%"PRIu16",dst=%"PRIu16")", ntohs(udp_key->udp_src), ntohs(udp_key->udp_dst)); break; case OVS_KEY_ATTR_ICMP: icmp_key = nl_attr_get(a); - ds_put_format(ds, "icmp(type=%"PRIu8",code=%"PRIu8")", + ds_put_format(ds, "(type=%"PRIu8",code=%"PRIu8")", icmp_key->icmp_type, icmp_key->icmp_code); break; case OVS_KEY_ATTR_ICMPV6: icmpv6_key = nl_attr_get(a); - ds_put_format(ds, "icmpv6(type=%"PRIu8",code=%"PRIu8")", + ds_put_format(ds, "(type=%"PRIu8",code=%"PRIu8")", icmpv6_key->icmpv6_type, icmpv6_key->icmpv6_code); break; case OVS_KEY_ATTR_ARP: arp_key = nl_attr_get(a); - ds_put_format(ds, "arp(sip="IP_FMT",tip="IP_FMT",op=%"PRIu16"," + 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; @@ -351,7 +1032,7 @@ format_odp_key_attr(const struct nlattr *a, struct ds *ds) nd_key = nl_attr_get(a); inet_ntop(AF_INET6, nd_key->nd_target, target, sizeof target); - ds_put_format(ds, "nd(target=%s", target); + ds_put_format(ds, "(target=%s", target); if (!eth_addr_is_zero(nd_key->nd_sll)) { ds_put_format(ds, ",sll="ETH_ADDR_FMT, ETH_ADDR_ARGS(nd_key->nd_sll)); @@ -364,6 +1045,8 @@ format_odp_key_attr(const struct nlattr *a, struct ds *ds) break; } + case OVS_KEY_ATTR_UNSPEC: + case __OVS_KEY_ATTR_MAX: default: format_generic_odp_key(a, ds); break; @@ -386,10 +1069,16 @@ odp_flow_key_format(const struct nlattr *key, size_t key_len, struct ds *ds) format_odp_key_attr(a, ds); } if (left) { + int i; + if (left == key_len) { ds_put_cstr(ds, ""); } - ds_put_format(ds, ",***%u leftover bytes***", left); + ds_put_format(ds, ",***%u leftover bytes*** (", left); + for (i = 0; i < left; i++) { + ds_put_format(ds, "%02x", ((const uint8_t *) a)[i]); + } + ds_put_char(ds, ')'); } } else { ds_put_cstr(ds, ""); @@ -416,8 +1105,33 @@ put_nd_key(int n, const char *nd_target_s, return n; } +static bool +ovs_frag_type_from_string(const char *s, enum ovs_frag_type *type) +{ + if (!strcasecmp(s, "no")) { + *type = OVS_FRAG_TYPE_NONE; + } else if (!strcasecmp(s, "first")) { + *type = OVS_FRAG_TYPE_FIRST; + } else if (!strcasecmp(s, "later")) { + *type = OVS_FRAG_TYPE_LATER; + } else { + return false; + } + 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, struct ofpbuf *key) +parse_odp_key_attr(const char *s, const struct simap *port_names, + struct ofpbuf *key) { /* Many of the sscanf calls in this function use oversized destination * fields because some sscanf() implementations truncate the range of %i @@ -429,14 +1143,56 @@ parse_odp_key_attr(const char *s, struct ofpbuf *key) * The tun_id parser has to use an alternative approach because there is no * type larger than 64 bits. */ + { + unsigned long long int priority; + int n = -1; + + 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; } } @@ -451,6 +1207,20 @@ parse_odp_key_attr(const char *s, struct ofpbuf *key) } } + if (port_names && !strncmp(s, "in_port(", 8)) { + const char *name; + const struct simap_node *node; + int name_len; + + name = s + 8; + name_len = strcspn(s, ")"); + node = simap_find_len(port_names, name, name_len); + if (node) { + nl_msg_put_u32(key, OVS_KEY_ATTR_IN_PORT, node->data); + return 8 + name_len + 1; + } + } + { struct ovs_key_ethernet eth_key; int n = -1; @@ -466,21 +1236,25 @@ parse_odp_key_attr(const char *s, struct ofpbuf *key) } { - uint16_t tpid = ETH_TYPE_VLAN; uint16_t vid; int pcp; + int cfi; int n = -1; - if ((sscanf(s, "vlan(vid=%"SCNi16",pcp=%i)%n", - &vid, &pcp, &n) > 0 && n > 0) || - (sscanf(s, "vlan(tpid=%"SCNi16",vid=%"SCNi16",pcp=%i)%n", - &tpid, &vid, &pcp, &n) > 0 && n > 0)) { - struct ovs_key_8021q q_key; - - q_key.q_tpid = htons(tpid); - q_key.q_tci = htons((vid << VLAN_VID_SHIFT) | - (pcp << VLAN_PCP_SHIFT)); - nl_msg_put_unspec(key, OVS_KEY_ATTR_8021Q, &q_key, sizeof q_key); + if ((sscanf(s, "vlan(vid=%"SCNi16",pcp=%i)%n", &vid, &pcp, &n) > 0 + && n > 0)) { + nl_msg_put_be16(key, OVS_KEY_ATTR_VLAN, + htons((vid << VLAN_VID_SHIFT) | + (pcp << VLAN_PCP_SHIFT) | + VLAN_CFI)); + return n; + } else if ((sscanf(s, "vlan(vid=%"SCNi16",pcp=%i,cfi=%i)%n", + &vid, &pcp, &cfi, &n) > 0 + && n > 0)) { + nl_msg_put_be16(key, OVS_KEY_ATTR_VLAN, + htons((vid << VLAN_VID_SHIFT) | + (pcp << VLAN_PCP_SHIFT) | + (cfi ? VLAN_CFI : 0))); return n; } } @@ -495,25 +1269,46 @@ parse_odp_key_attr(const char *s, struct ofpbuf *key) } } + { + 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_lse = mpls_lse_from_components(label, tc, ttl, bos); + return n; + } + } + { ovs_be32 ipv4_src; ovs_be32 ipv4_dst; int ipv4_proto; int ipv4_tos; + int ipv4_ttl; + char frag[8]; + enum ovs_frag_type ipv4_frag; int n = -1; if (sscanf(s, "ipv4(src="IP_SCAN_FMT",dst="IP_SCAN_FMT"," - "proto=%i,tos=%i)%n", - IP_SCAN_ARGS(&ipv4_src), - IP_SCAN_ARGS(&ipv4_dst), &ipv4_proto, &ipv4_tos, &n) > 0 - && n > 0) { + "proto=%i,tos=%i,ttl=%i,frag=%7[a-z])%n", + IP_SCAN_ARGS(&ipv4_src), IP_SCAN_ARGS(&ipv4_dst), + &ipv4_proto, &ipv4_tos, &ipv4_ttl, frag, &n) > 0 + && n > 0 + && ovs_frag_type_from_string(frag, &ipv4_frag)) { struct ovs_key_ipv4 ipv4_key; - memset(&ipv4_key, 0, sizeof ipv4_key); ipv4_key.ipv4_src = ipv4_src; ipv4_key.ipv4_dst = ipv4_dst; ipv4_key.ipv4_proto = ipv4_proto; ipv4_key.ipv4_tos = ipv4_tos; + ipv4_key.ipv4_ttl = ipv4_ttl; + ipv4_key.ipv4_frag = ipv4_frag; nl_msg_put_unspec(key, OVS_KEY_ATTR_IPV4, &ipv4_key, sizeof ipv4_key); return n; @@ -523,23 +1318,31 @@ parse_odp_key_attr(const char *s, struct ofpbuf *key) { char ipv6_src_s[IPV6_SCAN_LEN + 1]; char ipv6_dst_s[IPV6_SCAN_LEN + 1]; + int ipv6_label; int ipv6_proto; - int ipv6_tos; + int ipv6_tclass; + int ipv6_hlimit; + char frag[8]; + enum ovs_frag_type ipv6_frag; int n = -1; if (sscanf(s, "ipv6(src="IPV6_SCAN_FMT",dst="IPV6_SCAN_FMT"," - "proto=%i,tos=%i)%n", - ipv6_src_s, ipv6_dst_s, - &ipv6_proto, &ipv6_tos, &n) > 0 && n > 0) { + "label=%i,proto=%i,tclass=%i,hlimit=%i,frag=%7[a-z])%n", + ipv6_src_s, ipv6_dst_s, &ipv6_label, + &ipv6_proto, &ipv6_tclass, &ipv6_hlimit, frag, &n) > 0 + && n > 0 + && ovs_frag_type_from_string(frag, &ipv6_frag)) { struct ovs_key_ipv6 ipv6_key; - memset(&ipv6_key, 0, sizeof ipv6_key); if (inet_pton(AF_INET6, ipv6_src_s, &ipv6_key.ipv6_src) != 1 || inet_pton(AF_INET6, ipv6_dst_s, &ipv6_key.ipv6_dst) != 1) { return -EINVAL; } + ipv6_key.ipv6_label = htonl(ipv6_label); ipv6_key.ipv6_proto = ipv6_proto; - ipv6_key.ipv6_tos = ipv6_tos; + ipv6_key.ipv6_tclass = ipv6_tclass; + ipv6_key.ipv6_hlimit = ipv6_hlimit; + ipv6_key.ipv6_frag = ipv6_frag; nl_msg_put_unspec(key, OVS_KEY_ATTR_IPV6, &ipv6_key, sizeof ipv6_key); return n; @@ -666,6 +1469,36 @@ parse_odp_key_attr(const char *s, struct ofpbuf *key) } } + if (!strncmp(s, "encap(", 6)) { + const char *start = s; + size_t encap; + + encap = nl_msg_start_nested(key, OVS_KEY_ATTR_ENCAP); + + s += 6; + for (;;) { + int retval; + + s += strspn(s, ", \t\r\n"); + if (!*s) { + return -EINVAL; + } else if (*s == ')') { + break; + } + + retval = parse_odp_key_attr(s, port_names, key); + if (retval < 0) { + return retval; + } + s += retval; + } + s++; + + nl_msg_end_nested(key, encap); + + return s - start; + } + return -EINVAL; } @@ -676,23 +1509,27 @@ parse_odp_key_attr(const char *s, struct ofpbuf *key) * data is appended to 'key'. Either way, 'key''s data might be * reallocated. * + * If 'port_names' is nonnull, it points to an simap that maps from a port name + * to a port number. (Port names may be used instead of port numbers in + * in_port.) + * * On success, the attributes appended to 'key' are individually syntactically * valid, but they may not be valid as a sequence. 'key' might, for example, - * be missing an "in_port" key, have duplicated keys, or have keys in the wrong - * order. odp_flow_key_to_flow() will detect those errors. */ + * have duplicated keys. odp_flow_key_to_flow() will detect those errors. */ int -odp_flow_key_from_string(const char *s, struct ofpbuf *key) +odp_flow_key_from_string(const char *s, const struct simap *port_names, + struct ofpbuf *key) { const size_t old_size = key->size; for (;;) { int retval; - s += strspn(s, ", \t\r\n"); + s += strspn(s, delimiters); if (!*s) { return 0; } - retval = parse_odp_key_attr(s, key); + retval = parse_odp_key_attr(s, port_names, key); if (retval < 0) { key->size = old_size; return -retval; @@ -703,19 +1540,43 @@ odp_flow_key_from_string(const char *s, struct ofpbuf *key) return 0; } -/* Appends a representation of 'flow' as OVS_KEY_ATTR_* attributes to 'buf'. */ +static uint8_t +ovs_to_odp_frag(uint8_t nw_frag) +{ + return (nw_frag == 0 ? OVS_FRAG_TYPE_NONE + : nw_frag == FLOW_NW_FRAG_ANY ? OVS_FRAG_TYPE_FIRST + : OVS_FRAG_TYPE_LATER); +} + +/* 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; + + if (flow->skb_priority) { + 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->in_port != OFPP_NONE) { - nl_msg_put_u32(buf, OVS_KEY_ATTR_IN_PORT, - ofp_port_to_odp_port(flow->in_port)); + if (flow->skb_mark) { + nl_msg_put_u32(buf, OVS_KEY_ATTR_SKB_MARK, flow->skb_mark); + } + + 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, @@ -723,17 +1584,19 @@ odp_flow_key_from_flow(struct ofpbuf *buf, const struct flow *flow) memcpy(eth_key->eth_src, flow->dl_src, ETH_ADDR_LEN); memcpy(eth_key->eth_dst, flow->dl_dst, ETH_ADDR_LEN); - if (flow->vlan_tci != htons(0)) { - struct ovs_key_8021q *q_key; - - q_key = nl_msg_put_unspec_uninit(buf, OVS_KEY_ATTR_8021Q, - sizeof *q_key); - q_key->q_tpid = htons(ETH_TYPE_VLAN); - q_key->q_tci = flow->vlan_tci & ~htons(VLAN_CFI); + if (flow->vlan_tci != htons(0) || flow->dl_type == htons(ETH_TYPE_VLAN)) { + nl_msg_put_be16(buf, OVS_KEY_ATTR_ETHERTYPE, htons(ETH_TYPE_VLAN)); + nl_msg_put_be16(buf, OVS_KEY_ATTR_VLAN, flow->vlan_tci); + encap = nl_msg_start_nested(buf, OVS_KEY_ATTR_ENCAP); + if (flow->vlan_tci == htons(0)) { + goto unencap; + } + } else { + encap = 0; } if (ntohs(flow->dl_type) < ETH_TYPE_MIN) { - return; + goto unencap; } nl_msg_put_be16(buf, OVS_KEY_ATTR_ETHERTYPE, flow->dl_type); @@ -743,22 +1606,26 @@ odp_flow_key_from_flow(struct ofpbuf *buf, const struct flow *flow) ipv4_key = nl_msg_put_unspec_uninit(buf, OVS_KEY_ATTR_IPV4, sizeof *ipv4_key); - memset(ipv4_key, 0, sizeof *ipv4_key); ipv4_key->ipv4_src = flow->nw_src; ipv4_key->ipv4_dst = flow->nw_dst; ipv4_key->ipv4_proto = flow->nw_proto; ipv4_key->ipv4_tos = flow->nw_tos; + ipv4_key->ipv4_ttl = flow->nw_ttl; + ipv4_key->ipv4_frag = ovs_to_odp_frag(flow->nw_frag); } else if (flow->dl_type == htons(ETH_TYPE_IPV6)) { struct ovs_key_ipv6 *ipv6_key; ipv6_key = nl_msg_put_unspec_uninit(buf, OVS_KEY_ATTR_IPV6, sizeof *ipv6_key); - memset(ipv6_key, 0, sizeof *ipv6_key); memcpy(ipv6_key->ipv6_src, &flow->ipv6_src, sizeof ipv6_key->ipv6_src); memcpy(ipv6_key->ipv6_dst, &flow->ipv6_dst, sizeof ipv6_key->ipv6_dst); + ipv6_key->ipv6_label = flow->ipv6_label; ipv6_key->ipv6_proto = flow->nw_proto; - ipv6_key->ipv6_tos = flow->nw_tos; - } else if (flow->dl_type == htons(ETH_TYPE_ARP)) { + 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) || + 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, @@ -771,9 +1638,15 @@ 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)) { + 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_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; @@ -818,242 +1691,769 @@ odp_flow_key_from_flow(struct ofpbuf *buf, const struct flow *flow) } } } + +unencap: + if (encap) { + nl_msg_end_nested(buf, encap); + } } -/* Converts the 'key_len' bytes of OVS_KEY_ATTR_* attributes in 'key' to a flow - * structure in 'flow'. Returns 0 if successful, otherwise EINVAL. */ -int -odp_flow_key_to_flow(const struct nlattr *key, size_t key_len, - struct flow *flow) +uint32_t +odp_flow_key_hash(const struct nlattr *key, size_t key_len) { + BUILD_ASSERT_DECL(!(NLA_ALIGNTO % sizeof(uint32_t))); + return hash_words((const uint32_t *) key, key_len / sizeof(uint32_t), 0); +} + +static void +log_odp_key_attributes(struct vlog_rate_limit *rl, const char *title, + uint64_t attrs, int out_of_range_attr, + const struct nlattr *key, size_t key_len) +{ + struct ds s; + int i; + + if (VLOG_DROP_DBG(rl)) { + return; + } + + ds_init(&s); + for (i = 0; i < 64; i++) { + if (attrs & (UINT64_C(1) << i)) { + ds_put_format(&s, " %s", ovs_key_attr_to_string(i)); + } + } + if (out_of_range_attr) { + ds_put_format(&s, " %d (and possibly others)", out_of_range_attr); + } + + ds_put_cstr(&s, ": "); + odp_flow_key_format(key, key_len, &s); + + VLOG_DBG("%s:%s", title, ds_cstr(&s)); + ds_destroy(&s); +} + +static bool +odp_to_ovs_frag(uint8_t odp_frag, struct flow *flow) +{ + static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5); + + if (odp_frag > OVS_FRAG_TYPE_LATER) { + VLOG_ERR_RL(&rl, "invalid frag %"PRIu8" in flow key", odp_frag); + return false; + } + + if (odp_frag != OVS_FRAG_TYPE_NONE) { + flow->nw_frag |= FLOW_NW_FRAG_ANY; + if (odp_frag == OVS_FRAG_TYPE_LATER) { + flow->nw_frag |= FLOW_NW_FRAG_LATER; + } + } + return true; +} + +static bool +parse_flow_nlattrs(const struct nlattr *key, size_t key_len, + const struct nlattr *attrs[], uint64_t *present_attrsp, + int *out_of_range_attrp) +{ + static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(10, 10); const struct nlattr *nla; - enum ovs_key_type prev_type; + uint64_t present_attrs; size_t left; - memset(flow, 0, sizeof *flow); - flow->dl_type = htons(FLOW_DL_TYPE_NONE); - flow->in_port = OFPP_NONE; - - prev_type = OVS_KEY_ATTR_UNSPEC; + 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) { - const struct ovs_key_ethernet *eth_key; - const struct ovs_key_8021q *q_key; - const struct ovs_key_ipv4 *ipv4_key; - const struct ovs_key_ipv6 *ipv6_key; - const struct ovs_key_tcp *tcp_key; - const struct ovs_key_udp *udp_key; - const struct ovs_key_icmp *icmp_key; - const struct ovs_key_icmpv6 *icmpv6_key; - const struct ovs_key_arp *arp_key; - const struct ovs_key_nd *nd_key; - uint16_t type = nl_attr_type(nla); - int len = odp_flow_key_attr_len(type); + size_t len = nl_attr_get_size(nla); + int expected_len = odp_flow_key_attr_len(type); + + if (len != expected_len && expected_len >= 0) { + VLOG_ERR_RL(&rl, "attribute %s has length %zu but should have " + "length %d", ovs_key_attr_to_string(type), + len, expected_len); + return false; + } + + if (type > OVS_KEY_ATTR_MAX) { + *out_of_range_attrp = type; + } else { + if (present_attrs & (UINT64_C(1) << type)) { + VLOG_ERR_RL(&rl, "duplicate %s attribute in flow key", + ovs_key_attr_to_string(type)); + return false; + } - if (nl_attr_get_size(nla) != len && len != -1) { - return EINVAL; + present_attrs |= UINT64_C(1) << type; + attrs[type] = nla; } + } + if (left) { + VLOG_ERR_RL(&rl, "trailing garbage in flow key"); + return false; + } -#define TRANSITION(PREV_TYPE, TYPE) (((PREV_TYPE) << 16) | (TYPE)) - switch (TRANSITION(prev_type, type)) { - case TRANSITION(OVS_KEY_ATTR_UNSPEC, OVS_KEY_ATTR_TUN_ID): - flow->tun_id = nl_attr_get_be64(nla); - break; + *present_attrsp = present_attrs; + return true; +} - case TRANSITION(OVS_KEY_ATTR_UNSPEC, OVS_KEY_ATTR_IN_PORT): - case TRANSITION(OVS_KEY_ATTR_TUN_ID, OVS_KEY_ATTR_IN_PORT): - if (nl_attr_get_u32(nla) >= UINT16_MAX) { - return EINVAL; - } - flow->in_port = odp_port_to_ofp_port(nl_attr_get_u32(nla)); - break; +static enum odp_key_fitness +check_expectations(uint64_t present_attrs, int out_of_range_attr, + uint64_t expected_attrs, + const struct nlattr *key, size_t key_len) +{ + uint64_t missing_attrs; + uint64_t extra_attrs; + + missing_attrs = expected_attrs & ~present_attrs; + if (missing_attrs) { + static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(10, 10); + log_odp_key_attributes(&rl, "expected but not present", + missing_attrs, 0, key, key_len); + return ODP_FIT_TOO_LITTLE; + } - case TRANSITION(OVS_KEY_ATTR_UNSPEC, OVS_KEY_ATTR_ETHERNET): - case TRANSITION(OVS_KEY_ATTR_TUN_ID, OVS_KEY_ATTR_ETHERNET): - case TRANSITION(OVS_KEY_ATTR_IN_PORT, OVS_KEY_ATTR_ETHERNET): - eth_key = nl_attr_get(nla); - memcpy(flow->dl_src, eth_key->eth_src, ETH_ADDR_LEN); - memcpy(flow->dl_dst, eth_key->eth_dst, ETH_ADDR_LEN); - break; + extra_attrs = present_attrs & ~expected_attrs; + if (extra_attrs || out_of_range_attr) { + static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(10, 10); + log_odp_key_attributes(&rl, "present but not expected", + extra_attrs, out_of_range_attr, key, key_len); + return ODP_FIT_TOO_MUCH; + } - case TRANSITION(OVS_KEY_ATTR_ETHERNET, OVS_KEY_ATTR_8021Q): - q_key = nl_attr_get(nla); - if (q_key->q_tpid != htons(ETH_TYPE_VLAN)) { - /* Only standard 0x8100 VLANs currently supported. */ - return EINVAL; - } - if (q_key->q_tci & htons(VLAN_CFI)) { - return EINVAL; - } - flow->vlan_tci = q_key->q_tci | htons(VLAN_CFI); - break; + return ODP_FIT_PERFECT; +} - case TRANSITION(OVS_KEY_ATTR_8021Q, OVS_KEY_ATTR_ETHERTYPE): - case TRANSITION(OVS_KEY_ATTR_ETHERNET, OVS_KEY_ATTR_ETHERTYPE): - flow->dl_type = nl_attr_get_be16(nla); - if (ntohs(flow->dl_type) < 1536) { - return EINVAL; - } - break; +static bool +parse_ethertype(const struct nlattr *attrs[OVS_KEY_ATTR_MAX + 1], + uint64_t present_attrs, uint64_t *expected_attrs, + struct flow *flow) +{ + static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5); + + if (present_attrs & (UINT64_C(1) << OVS_KEY_ATTR_ETHERTYPE)) { + flow->dl_type = nl_attr_get_be16(attrs[OVS_KEY_ATTR_ETHERTYPE]); + if (ntohs(flow->dl_type) < 1536) { + VLOG_ERR_RL(&rl, "invalid Ethertype %"PRIu16" in flow key", + ntohs(flow->dl_type)); + return false; + } + *expected_attrs |= UINT64_C(1) << OVS_KEY_ATTR_ETHERTYPE; + } else { + flow->dl_type = htons(FLOW_DL_TYPE_NONE); + } + return true; +} - case TRANSITION(OVS_KEY_ATTR_ETHERTYPE, OVS_KEY_ATTR_IPV4): - if (flow->dl_type != htons(ETH_TYPE_IP)) { - return EINVAL; - } - ipv4_key = nl_attr_get(nla); +static enum odp_key_fitness +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 (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; + + ipv4_key = nl_attr_get(attrs[OVS_KEY_ATTR_IPV4]); flow->nw_src = ipv4_key->ipv4_src; flow->nw_dst = ipv4_key->ipv4_dst; flow->nw_proto = ipv4_key->ipv4_proto; flow->nw_tos = ipv4_key->ipv4_tos; - if (flow->nw_tos & IP_ECN_MASK) { - return EINVAL; + flow->nw_ttl = ipv4_key->ipv4_ttl; + if (!odp_to_ovs_frag(ipv4_key->ipv4_frag, flow)) { + return ODP_FIT_ERROR; } - break; + } + } else if (flow->dl_type == htons(ETH_TYPE_IPV6)) { + expected_attrs |= UINT64_C(1) << OVS_KEY_ATTR_IPV6; + if (present_attrs & (UINT64_C(1) << OVS_KEY_ATTR_IPV6)) { + const struct ovs_key_ipv6 *ipv6_key; - case TRANSITION(OVS_KEY_ATTR_ETHERTYPE, OVS_KEY_ATTR_IPV6): - if (flow->dl_type != htons(ETH_TYPE_IPV6)) { - return EINVAL; - } - ipv6_key = nl_attr_get(nla); + ipv6_key = nl_attr_get(attrs[OVS_KEY_ATTR_IPV6]); memcpy(&flow->ipv6_src, ipv6_key->ipv6_src, sizeof flow->ipv6_src); memcpy(&flow->ipv6_dst, ipv6_key->ipv6_dst, sizeof flow->ipv6_dst); + flow->ipv6_label = ipv6_key->ipv6_label; flow->nw_proto = ipv6_key->ipv6_proto; - flow->nw_tos = ipv6_key->ipv6_tos; - if (flow->nw_tos & IP_ECN_MASK) { - return EINVAL; + flow->nw_tos = ipv6_key->ipv6_tclass; + flow->nw_ttl = ipv6_key->ipv6_hlimit; + if (!odp_to_ovs_frag(ipv6_key->ipv6_frag, flow)) { + return ODP_FIT_ERROR; } - break; + } + } 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; - case TRANSITION(OVS_KEY_ATTR_IPV4, OVS_KEY_ATTR_TCP): - case TRANSITION(OVS_KEY_ATTR_IPV6, OVS_KEY_ATTR_TCP): - if (flow->nw_proto != IPPROTO_TCP) { - return EINVAL; + arp_key = nl_attr_get(attrs[OVS_KEY_ATTR_ARP]); + flow->nw_src = arp_key->arp_sip; + flow->nw_dst = arp_key->arp_tip; + if (arp_key->arp_op & htons(0xff00)) { + VLOG_ERR_RL(&rl, "unsupported ARP opcode %"PRIu16" in flow " + "key", ntohs(arp_key->arp_op)); + return ODP_FIT_ERROR; } - tcp_key = nl_attr_get(nla); + flow->nw_proto = ntohs(arp_key->arp_op); + memcpy(flow->arp_sha, arp_key->arp_sha, ETH_ADDR_LEN); + memcpy(flow->arp_tha, arp_key->arp_tha, ETH_ADDR_LEN); + } + } + + if (flow->nw_proto == IPPROTO_TCP + && (flow->dl_type == htons(ETH_TYPE_IP) || + flow->dl_type == htons(ETH_TYPE_IPV6)) + && !(flow->nw_frag & FLOW_NW_FRAG_LATER)) { + expected_attrs |= UINT64_C(1) << OVS_KEY_ATTR_TCP; + if (present_attrs & (UINT64_C(1) << OVS_KEY_ATTR_TCP)) { + const struct ovs_key_tcp *tcp_key; + + tcp_key = nl_attr_get(attrs[OVS_KEY_ATTR_TCP]); flow->tp_src = tcp_key->tcp_src; flow->tp_dst = tcp_key->tcp_dst; - break; - - case TRANSITION(OVS_KEY_ATTR_IPV4, OVS_KEY_ATTR_UDP): - case TRANSITION(OVS_KEY_ATTR_IPV6, OVS_KEY_ATTR_UDP): - if (flow->nw_proto != IPPROTO_UDP) { - return EINVAL; - } - udp_key = nl_attr_get(nla); + } + } else if (flow->nw_proto == IPPROTO_UDP + && (flow->dl_type == htons(ETH_TYPE_IP) || + flow->dl_type == htons(ETH_TYPE_IPV6)) + && !(flow->nw_frag & FLOW_NW_FRAG_LATER)) { + expected_attrs |= UINT64_C(1) << OVS_KEY_ATTR_UDP; + if (present_attrs & (UINT64_C(1) << OVS_KEY_ATTR_UDP)) { + const struct ovs_key_udp *udp_key; + + udp_key = nl_attr_get(attrs[OVS_KEY_ATTR_UDP]); flow->tp_src = udp_key->udp_src; flow->tp_dst = udp_key->udp_dst; - break; - - case TRANSITION(OVS_KEY_ATTR_IPV4, OVS_KEY_ATTR_ICMP): - if (flow->nw_proto != IPPROTO_ICMP) { - return EINVAL; - } - icmp_key = nl_attr_get(nla); + } + } else if (flow->nw_proto == IPPROTO_ICMP + && flow->dl_type == htons(ETH_TYPE_IP) + && !(flow->nw_frag & FLOW_NW_FRAG_LATER)) { + expected_attrs |= UINT64_C(1) << OVS_KEY_ATTR_ICMP; + if (present_attrs & (UINT64_C(1) << OVS_KEY_ATTR_ICMP)) { + const struct ovs_key_icmp *icmp_key; + + icmp_key = nl_attr_get(attrs[OVS_KEY_ATTR_ICMP]); flow->tp_src = htons(icmp_key->icmp_type); flow->tp_dst = htons(icmp_key->icmp_code); - break; - - case TRANSITION(OVS_KEY_ATTR_IPV6, OVS_KEY_ATTR_ICMPV6): - if (flow->nw_proto != IPPROTO_ICMPV6) { - return EINVAL; - } - icmpv6_key = nl_attr_get(nla); + } + } else if (flow->nw_proto == IPPROTO_ICMPV6 + && flow->dl_type == htons(ETH_TYPE_IPV6) + && !(flow->nw_frag & FLOW_NW_FRAG_LATER)) { + expected_attrs |= UINT64_C(1) << OVS_KEY_ATTR_ICMPV6; + if (present_attrs & (UINT64_C(1) << OVS_KEY_ATTR_ICMPV6)) { + const struct ovs_key_icmpv6 *icmpv6_key; + + icmpv6_key = nl_attr_get(attrs[OVS_KEY_ATTR_ICMPV6]); flow->tp_src = htons(icmpv6_key->icmpv6_type); flow->tp_dst = htons(icmpv6_key->icmpv6_code); - break; - case TRANSITION(OVS_KEY_ATTR_ETHERTYPE, OVS_KEY_ATTR_ARP): - if (flow->dl_type != htons(ETH_TYPE_ARP)) { - return EINVAL; - } - arp_key = nl_attr_get(nla); - flow->nw_src = arp_key->arp_sip; - flow->nw_dst = arp_key->arp_tip; - if (arp_key->arp_op & htons(0xff00)) { - return EINVAL; + if (flow->tp_src == htons(ND_NEIGHBOR_SOLICIT) || + flow->tp_src == htons(ND_NEIGHBOR_ADVERT)) { + expected_attrs |= UINT64_C(1) << OVS_KEY_ATTR_ND; + if (present_attrs & (UINT64_C(1) << OVS_KEY_ATTR_ND)) { + const struct ovs_key_nd *nd_key; + + nd_key = nl_attr_get(attrs[OVS_KEY_ATTR_ND]); + memcpy(&flow->nd_target, nd_key->nd_target, + sizeof flow->nd_target); + memcpy(flow->arp_sha, nd_key->nd_sll, ETH_ADDR_LEN); + memcpy(flow->arp_tha, nd_key->nd_tll, ETH_ADDR_LEN); + } } - flow->nw_proto = ntohs(arp_key->arp_op); - memcpy(flow->arp_sha, arp_key->arp_sha, ETH_ADDR_LEN); - memcpy(flow->arp_tha, arp_key->arp_tha, ETH_ADDR_LEN); - break; + } + } - case TRANSITION(OVS_KEY_ATTR_ICMPV6, OVS_KEY_ATTR_ND): - if (flow->tp_src != htons(ND_NEIGHBOR_SOLICIT) - && flow->tp_src != htons(ND_NEIGHBOR_ADVERT)) { - return EINVAL; - } - nd_key = nl_attr_get(nla); - memcpy(&flow->nd_target, nd_key->nd_target, sizeof flow->nd_target); - memcpy(flow->arp_sha, nd_key->nd_sll, ETH_ADDR_LEN); - memcpy(flow->arp_tha, nd_key->nd_tll, ETH_ADDR_LEN); - break; + return check_expectations(present_attrs, out_of_range_attr, expected_attrs, + key, key_len); +} - default: - if (type == OVS_KEY_ATTR_UNSPEC - || prev_type == OVS_KEY_ATTR_UNSPEC) { - return EINVAL; - } - return EINVAL; +/* Parse 802.1Q header then encapsulated L3 attributes. */ +static enum odp_key_fitness +parse_8021q_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); + + const struct nlattr *encap + = (present_attrs & (UINT64_C(1) << OVS_KEY_ATTR_ENCAP) + ? attrs[OVS_KEY_ATTR_ENCAP] : NULL); + enum odp_key_fitness encap_fitness; + enum odp_key_fitness fitness; + ovs_be16 tci; + + /* Calculate fitness of outer attributes. */ + expected_attrs |= ((UINT64_C(1) << OVS_KEY_ATTR_VLAN) | + (UINT64_C(1) << OVS_KEY_ATTR_ENCAP)); + fitness = check_expectations(present_attrs, out_of_range_attr, + expected_attrs, key, key_len); + + /* Get the VLAN TCI value. */ + if (!(present_attrs & (UINT64_C(1) << OVS_KEY_ATTR_VLAN))) { + return ODP_FIT_TOO_LITTLE; + } + tci = nl_attr_get_be16(attrs[OVS_KEY_ATTR_VLAN]); + if (tci == htons(0)) { + /* Corner case for a truncated 802.1Q header. */ + if (fitness == ODP_FIT_PERFECT && nl_attr_get_size(encap)) { + return ODP_FIT_TOO_MUCH; } + return fitness; + } else if (!(tci & htons(VLAN_CFI))) { + VLOG_ERR_RL(&rl, "OVS_KEY_ATTR_VLAN 0x%04"PRIx16" is nonzero " + "but CFI bit is not set", ntohs(tci)); + return ODP_FIT_ERROR; + } + + /* Set vlan_tci. + * Remove the TPID from dl_type since it's not the real Ethertype. */ + flow->vlan_tci = tci; + flow->dl_type = htons(0); - prev_type = type; + /* Now parse the encapsulated attributes. */ + if (!parse_flow_nlattrs(nl_attr_get(encap), nl_attr_get_size(encap), + attrs, &present_attrs, &out_of_range_attr)) { + return ODP_FIT_ERROR; } - if (left) { - return EINVAL; + expected_attrs = 0; + + if (!parse_ethertype(attrs, present_attrs, &expected_attrs, flow)) { + return ODP_FIT_ERROR; } + encap_fitness = parse_l2_5_onward(attrs, present_attrs, out_of_range_attr, + expected_attrs, flow, key, key_len); - switch (prev_type) { - case OVS_KEY_ATTR_UNSPEC: - return EINVAL; + /* The overall fitness is the worse of the outer and inner attributes. */ + return MAX(fitness, encap_fitness); +} - case OVS_KEY_ATTR_TUN_ID: - case OVS_KEY_ATTR_IN_PORT: - return EINVAL; +/* Converts the 'key_len' bytes of OVS_KEY_ATTR_* attributes in 'key' to a flow + * 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 + * by looking at the attributes for lower-level protocols, e.g. if the network + * protocol in OVS_KEY_ATTR_IPV4 or OVS_KEY_ATTR_IPV6 is IPPROTO_TCP then we + * know that a OVS_KEY_ATTR_TCP attribute must appear and that otherwise it + * must be absent. */ +enum odp_key_fitness +odp_flow_key_to_flow(const struct nlattr *key, size_t key_len, + struct flow *flow) +{ + const struct nlattr *attrs[OVS_KEY_ATTR_MAX + 1]; + uint64_t expected_attrs; + uint64_t present_attrs; + int out_of_range_attr; - case OVS_KEY_ATTR_ETHERNET: - case OVS_KEY_ATTR_8021Q: - return 0; + memset(flow, 0, sizeof *flow); - case OVS_KEY_ATTR_ETHERTYPE: - if (flow->dl_type == htons(ETH_TYPE_IP) - || flow->dl_type == htons(ETH_TYPE_IPV6) - || flow->dl_type == htons(ETH_TYPE_ARP)) { - return EINVAL; + /* Parse attributes. */ + if (!parse_flow_nlattrs(key, key_len, attrs, &present_attrs, + &out_of_range_attr)) { + return ODP_FIT_ERROR; + } + expected_attrs = 0; + + /* Metadata. */ + if (present_attrs & (UINT64_C(1) << OVS_KEY_ATTR_PRIORITY)) { + flow->skb_priority = nl_attr_get_u32(attrs[OVS_KEY_ATTR_PRIORITY]); + expected_attrs |= UINT64_C(1) << OVS_KEY_ATTR_PRIORITY; + } + + 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_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; } - return 0; + } - case OVS_KEY_ATTR_IPV4: - if (flow->nw_proto == IPPROTO_TCP - || flow->nw_proto == IPPROTO_UDP - || flow->nw_proto == IPPROTO_ICMP) { - return EINVAL; + 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 = OVSP_NONE; + } + + /* Ethernet header. */ + if (present_attrs & (UINT64_C(1) << OVS_KEY_ATTR_ETHERNET)) { + const struct ovs_key_ethernet *eth_key; + + eth_key = nl_attr_get(attrs[OVS_KEY_ATTR_ETHERNET]); + memcpy(flow->dl_src, eth_key->eth_src, ETH_ADDR_LEN); + memcpy(flow->dl_dst, eth_key->eth_dst, ETH_ADDR_LEN); + } + expected_attrs |= UINT64_C(1) << OVS_KEY_ATTR_ETHERNET; + + /* Get Ethertype or 802.1Q TPID or FLOW_DL_TYPE_NONE. */ + if (!parse_ethertype(attrs, present_attrs, &expected_attrs, flow)) { + return ODP_FIT_ERROR; + } + + if (flow->dl_type == htons(ETH_TYPE_VLAN)) { + return parse_8021q_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. */ +const char * +odp_key_fitness_to_string(enum odp_key_fitness fitness) +{ + switch (fitness) { + case ODP_FIT_PERFECT: + return "OK"; + case ODP_FIT_TOO_MUCH: + return "too_much"; + case ODP_FIT_TOO_LITTLE: + return "too_little"; + case ODP_FIT_ERROR: + return "error"; + default: + return ""; + } +} + +/* Appends an OVS_ACTION_ATTR_USERSPACE action to 'odp_actions' that specifies + * 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 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 (userdata) { + userdata_ofs = odp_actions->size + NLA_HDRLEN; + nl_msg_put_unspec(odp_actions, OVS_USERSPACE_ATTR_USERDATA, + userdata, userdata_size); + } else { + userdata_ofs = 0; + } + nl_msg_end_nested(odp_actions, offset); + + 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. */ + +static void +commit_set_action(struct ofpbuf *odp_actions, enum ovs_key_attr key_type, + const void *key, size_t key_size) +{ + size_t offset = nl_msg_start_nested(odp_actions, OVS_ACTION_ATTR_SET); + nl_msg_put_unspec(odp_actions, key_type, key, key_size); + nl_msg_end_nested(odp_actions, offset); +} + +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) +{ + /* A valid IPV4_TUNNEL must have non-zero ip_dst. */ + if (flow->tunnel.ip_dst) { + if (!memcmp(&base->tunnel, &flow->tunnel, sizeof base->tunnel)) { + return; } - return 0; + memcpy(&base->tunnel, &flow->tunnel, sizeof base->tunnel); + odp_put_tunnel_action(&base->tunnel, odp_actions); + } +} - case OVS_KEY_ATTR_IPV6: - if (flow->nw_proto == IPPROTO_TCP - || flow->nw_proto == IPPROTO_UDP - || flow->nw_proto == IPPROTO_ICMPV6) { - return EINVAL; +static void +commit_set_ether_addr_action(const struct flow *flow, struct flow *base, + struct ofpbuf *odp_actions) +{ + struct ovs_key_ethernet eth_key; + + if (eth_addr_equals(base->dl_src, flow->dl_src) && + eth_addr_equals(base->dl_dst, flow->dl_dst)) { + return; + } + + memcpy(base->dl_src, flow->dl_src, ETH_ADDR_LEN); + memcpy(base->dl_dst, flow->dl_dst, ETH_ADDR_LEN); + + memcpy(eth_key.eth_src, base->dl_src, ETH_ADDR_LEN); + memcpy(eth_key.eth_dst, base->dl_dst, ETH_ADDR_LEN); + + commit_set_action(odp_actions, OVS_KEY_ATTR_ETHERNET, + ð_key, sizeof(eth_key)); +} + +static void +commit_vlan_action(const struct flow *flow, struct flow *base, + struct ofpbuf *odp_actions) +{ + if (base->vlan_tci == flow->vlan_tci) { + return; + } + + if (base->vlan_tci & htons(VLAN_CFI)) { + nl_msg_put_flag(odp_actions, OVS_ACTION_ATTR_POP_VLAN); + } + + if (flow->vlan_tci & htons(VLAN_CFI)) { + struct ovs_action_push_vlan vlan; + + vlan.vlan_tpid = htons(ETH_TYPE_VLAN); + vlan.vlan_tci = flow->vlan_tci; + nl_msg_put_unspec(odp_actions, OVS_ACTION_ATTR_PUSH_VLAN, + &vlan, sizeof vlan); + } + 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"); } - return 0; - case OVS_KEY_ATTR_ICMPV6: - if (flow->icmp_type == htons(ND_NEIGHBOR_SOLICIT) - || flow->icmp_type == htons(ND_NEIGHBOR_ADVERT)) { - return EINVAL; + 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"); } - return 0; - case OVS_KEY_ATTR_TCP: - case OVS_KEY_ATTR_UDP: - case OVS_KEY_ATTR_ICMP: - case OVS_KEY_ATTR_ARP: - case OVS_KEY_ATTR_ND: - return 0; + 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; - case __OVS_KEY_ATTR_MAX: - default: - NOT_REACHED(); + mpls_key.mpls_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) +{ + struct ovs_key_ipv4 ipv4_key; + + if (base->nw_src == flow->nw_src && + base->nw_dst == flow->nw_dst && + base->nw_tos == flow->nw_tos && + base->nw_ttl == flow->nw_ttl && + base->nw_frag == flow->nw_frag) { + return; } + + ipv4_key.ipv4_src = base->nw_src = flow->nw_src; + ipv4_key.ipv4_dst = base->nw_dst = flow->nw_dst; + ipv4_key.ipv4_tos = base->nw_tos = flow->nw_tos; + ipv4_key.ipv4_ttl = base->nw_ttl = flow->nw_ttl; + ipv4_key.ipv4_proto = base->nw_proto; + ipv4_key.ipv4_frag = ovs_to_odp_frag(base->nw_frag); + + commit_set_action(odp_actions, OVS_KEY_ATTR_IPV4, + &ipv4_key, sizeof(ipv4_key)); +} + +static void +commit_set_ipv6_action(const struct flow *flow, struct flow *base, + struct ofpbuf *odp_actions) +{ + struct ovs_key_ipv6 ipv6_key; + + if (ipv6_addr_equals(&base->ipv6_src, &flow->ipv6_src) && + ipv6_addr_equals(&base->ipv6_dst, &flow->ipv6_dst) && + base->ipv6_label == flow->ipv6_label && + base->nw_tos == flow->nw_tos && + base->nw_ttl == flow->nw_ttl && + base->nw_frag == flow->nw_frag) { + return; + } + + base->ipv6_src = flow->ipv6_src; + memcpy(&ipv6_key.ipv6_src, &base->ipv6_src, sizeof(ipv6_key.ipv6_src)); + base->ipv6_dst = flow->ipv6_dst; + memcpy(&ipv6_key.ipv6_dst, &base->ipv6_dst, sizeof(ipv6_key.ipv6_dst)); + + ipv6_key.ipv6_label = base->ipv6_label = flow->ipv6_label; + ipv6_key.ipv6_tclass = base->nw_tos = flow->nw_tos; + ipv6_key.ipv6_hlimit = base->nw_ttl = flow->nw_ttl; + ipv6_key.ipv6_proto = base->nw_proto; + ipv6_key.ipv6_frag = ovs_to_odp_frag(base->nw_frag); + + commit_set_action(odp_actions, OVS_KEY_ATTR_IPV6, + &ipv6_key, sizeof(ipv6_key)); +} + +static void +commit_set_nw_action(const struct flow *flow, struct flow *base, + struct ofpbuf *odp_actions) +{ + /* Check if flow really have an IP header. */ + if (!flow->nw_proto) { + return; + } + + if (base->dl_type == htons(ETH_TYPE_IP)) { + commit_set_ipv4_action(flow, base, odp_actions); + } else if (base->dl_type == htons(ETH_TYPE_IPV6)) { + commit_set_ipv6_action(flow, base, odp_actions); + } +} + +static void +commit_set_port_action(const struct flow *flow, struct flow *base, + struct ofpbuf *odp_actions) +{ + if (!is_ip_any(base) || (!base->tp_src && !base->tp_dst)) { + return; + } + + if (base->tp_src == flow->tp_src && + base->tp_dst == flow->tp_dst) { + return; + } + + if (flow->nw_proto == IPPROTO_TCP) { + struct ovs_key_tcp port_key; + + port_key.tcp_src = base->tp_src = flow->tp_src; + port_key.tcp_dst = base->tp_dst = flow->tp_dst; + + commit_set_action(odp_actions, OVS_KEY_ATTR_TCP, + &port_key, sizeof(port_key)); + + } else if (flow->nw_proto == IPPROTO_UDP) { + struct ovs_key_udp port_key; + + port_key.udp_src = base->tp_src = flow->tp_src; + port_key.udp_dst = base->tp_dst = flow->tp_dst; + + commit_set_action(odp_actions, OVS_KEY_ATTR_UDP, + &port_key, sizeof(port_key)); + } +} + +static void +commit_set_priority_action(const struct flow *flow, struct flow *base, + struct ofpbuf *odp_actions) +{ + if (base->skb_priority == flow->skb_priority) { + return; + } + base->skb_priority = flow->skb_priority; + + commit_set_action(odp_actions, OVS_KEY_ATTR_PRIORITY, + &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. 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_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); + /* Committing 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); }