From dadff137931c27aa90aafc6783b6e9eb535b1117 Mon Sep 17 00:00:00 2001 From: Ivan Malov Date: Tue, 20 Oct 2020 10:13:41 +0100 Subject: [PATCH] net/sfc: support encap flow items in transfer rules Add support for flow items VXLAN, Geneve and NVGRE to MAE-specific RTE flow implementation. Having support for these items implies the ability to insert so-called outer MAE rules and refer to them in MAE action rules. The patch takes care of all necessary facilities to do that. Signed-off-by: Ivan Malov Signed-off-by: Andrew Rybchenko Reviewed-by: Andy Moreton --- doc/guides/nics/sfc_efx.rst | 6 + drivers/net/sfc/sfc_flow.c | 18 +- drivers/net/sfc/sfc_flow.h | 2 + drivers/net/sfc/sfc_mae.c | 603 ++++++++++++++++++++++++++++++++++-- drivers/net/sfc/sfc_mae.h | 34 ++ 5 files changed, 630 insertions(+), 33 deletions(-) diff --git a/doc/guides/nics/sfc_efx.rst b/doc/guides/nics/sfc_efx.rst index 1f6058e57a..cc5b9f120c 100644 --- a/doc/guides/nics/sfc_efx.rst +++ b/doc/guides/nics/sfc_efx.rst @@ -212,6 +212,12 @@ Supported pattern items (***transfer*** rules): - UDP (source/destination ports) +- VXLAN (exact match of VXLAN network identifier) + +- GENEVE (exact match of virtual network identifier) + +- NVGRE (exact match of virtual subnet ID) + Supported actions (***transfer*** rules): - OF_POP_VLAN diff --git a/drivers/net/sfc/sfc_flow.c b/drivers/net/sfc/sfc_flow.c index 6ccefef477..4321045d1a 100644 --- a/drivers/net/sfc/sfc_flow.c +++ b/drivers/net/sfc/sfc_flow.c @@ -1288,7 +1288,8 @@ sfc_flow_parse_pattern(const struct sfc_flow_item *flow_items, break; default: - if (is_ifrm) { + if (parse_ctx->type == SFC_FLOW_PARSE_CTX_FILTER && + is_ifrm) { rte_flow_error_set(error, EINVAL, RTE_FLOW_ERROR_TYPE_ITEM, pattern, @@ -2565,13 +2566,8 @@ sfc_flow_verify(struct sfc_adapter *sa, struct rte_flow *flow, } if (ops->verify != NULL) { - /* - * Use locking since verify method may need to - * access the list of already created rules. - */ - sfc_adapter_lock(sa); + SFC_ASSERT(sfc_adapter_is_locked(sa)); rc = ops->verify(sa, flow); - sfc_adapter_unlock(sa); } if (rc != 0) { @@ -2599,12 +2595,16 @@ sfc_flow_validate(struct rte_eth_dev *dev, if (flow == NULL) return -rte_errno; + sfc_adapter_lock(sa); + rc = sfc_flow_parse(dev, attr, pattern, actions, flow, error); if (rc == 0) rc = sfc_flow_verify(sa, flow, error); sfc_flow_free(sa, flow); + sfc_adapter_unlock(sa); + return rc; } @@ -2623,12 +2623,12 @@ sfc_flow_create(struct rte_eth_dev *dev, if (flow == NULL) goto fail_no_mem; + sfc_adapter_lock(sa); + rc = sfc_flow_parse(dev, attr, pattern, actions, flow, error); if (rc != 0) goto fail_bad_value; - sfc_adapter_lock(sa); - TAILQ_INSERT_TAIL(&sa->flow_list, flow, entries); if (sa->state == SFC_ADAPTER_STARTED) { diff --git a/drivers/net/sfc/sfc_flow.h b/drivers/net/sfc/sfc_flow.h index e991ae132c..09105953ff 100644 --- a/drivers/net/sfc/sfc_flow.h +++ b/drivers/net/sfc/sfc_flow.h @@ -67,6 +67,8 @@ struct sfc_flow_spec_filter { struct sfc_flow_spec_mae { /* Desired priority level */ unsigned int priority; + /* Outer rule registry entry */ + struct sfc_mae_outer_rule *outer_rule; /* EFX match specification */ efx_mae_match_spec_t *match_spec; /* Action set registry entry */ diff --git a/drivers/net/sfc/sfc_mae.c b/drivers/net/sfc/sfc_mae.c index 3dd233c6dc..c78a376a88 100644 --- a/drivers/net/sfc/sfc_mae.c +++ b/drivers/net/sfc/sfc_mae.c @@ -81,7 +81,10 @@ sfc_mae_attach(struct sfc_adapter *sa) goto fail_mae_assign_switch_port; mae->status = SFC_MAE_STATUS_SUPPORTED; + mae->nb_outer_rule_prios_max = limits.eml_max_n_outer_prios; mae->nb_action_rule_prios_max = limits.eml_max_n_action_prios; + mae->encap_types_supported = limits.eml_encap_types_supported; + TAILQ_INIT(&mae->outer_rules); TAILQ_INIT(&mae->action_sets); sfc_log_init(sa, "done"); @@ -119,6 +122,138 @@ sfc_mae_detach(struct sfc_adapter *sa) sfc_log_init(sa, "done"); } +static struct sfc_mae_outer_rule * +sfc_mae_outer_rule_attach(struct sfc_adapter *sa, + const efx_mae_match_spec_t *match_spec, + efx_tunnel_protocol_t encap_type) +{ + struct sfc_mae_outer_rule *rule; + struct sfc_mae *mae = &sa->mae; + + SFC_ASSERT(sfc_adapter_is_locked(sa)); + + TAILQ_FOREACH(rule, &mae->outer_rules, entries) { + if (efx_mae_match_specs_equal(rule->match_spec, match_spec) && + rule->encap_type == encap_type) { + ++(rule->refcnt); + return rule; + } + } + + return NULL; +} + +static int +sfc_mae_outer_rule_add(struct sfc_adapter *sa, + efx_mae_match_spec_t *match_spec, + efx_tunnel_protocol_t encap_type, + struct sfc_mae_outer_rule **rulep) +{ + struct sfc_mae_outer_rule *rule; + struct sfc_mae *mae = &sa->mae; + + SFC_ASSERT(sfc_adapter_is_locked(sa)); + + rule = rte_zmalloc("sfc_mae_outer_rule", sizeof(*rule), 0); + if (rule == NULL) + return ENOMEM; + + rule->refcnt = 1; + rule->match_spec = match_spec; + rule->encap_type = encap_type; + + rule->fw_rsrc.rule_id.id = EFX_MAE_RSRC_ID_INVALID; + + TAILQ_INSERT_TAIL(&mae->outer_rules, rule, entries); + + *rulep = rule; + + return 0; +} + +static void +sfc_mae_outer_rule_del(struct sfc_adapter *sa, + struct sfc_mae_outer_rule *rule) +{ + struct sfc_mae *mae = &sa->mae; + + SFC_ASSERT(sfc_adapter_is_locked(sa)); + SFC_ASSERT(rule->refcnt != 0); + + --(rule->refcnt); + + if (rule->refcnt != 0) + return; + + SFC_ASSERT(rule->fw_rsrc.rule_id.id == EFX_MAE_RSRC_ID_INVALID); + SFC_ASSERT(rule->fw_rsrc.refcnt == 0); + + efx_mae_match_spec_fini(sa->nic, rule->match_spec); + + TAILQ_REMOVE(&mae->outer_rules, rule, entries); + rte_free(rule); +} + +static int +sfc_mae_outer_rule_enable(struct sfc_adapter *sa, + struct sfc_mae_outer_rule *rule, + efx_mae_match_spec_t *match_spec_action) +{ + struct sfc_mae_fw_rsrc *fw_rsrc = &rule->fw_rsrc; + int rc; + + SFC_ASSERT(sfc_adapter_is_locked(sa)); + + if (fw_rsrc->refcnt == 0) { + SFC_ASSERT(fw_rsrc->rule_id.id == EFX_MAE_RSRC_ID_INVALID); + SFC_ASSERT(rule->match_spec != NULL); + + rc = efx_mae_outer_rule_insert(sa->nic, rule->match_spec, + rule->encap_type, + &fw_rsrc->rule_id); + if (rc != 0) + return rc; + } + + rc = efx_mae_match_spec_outer_rule_id_set(match_spec_action, + &fw_rsrc->rule_id); + if (rc != 0) { + if (fw_rsrc->refcnt == 0) { + (void)efx_mae_outer_rule_remove(sa->nic, + &fw_rsrc->rule_id); + } + return rc; + } + + ++(fw_rsrc->refcnt); + + return 0; +} + +static int +sfc_mae_outer_rule_disable(struct sfc_adapter *sa, + struct sfc_mae_outer_rule *rule) +{ + struct sfc_mae_fw_rsrc *fw_rsrc = &rule->fw_rsrc; + int rc; + + SFC_ASSERT(sfc_adapter_is_locked(sa)); + SFC_ASSERT(fw_rsrc->rule_id.id != EFX_MAE_RSRC_ID_INVALID); + SFC_ASSERT(fw_rsrc->refcnt != 0); + + if (fw_rsrc->refcnt == 1) { + rc = efx_mae_outer_rule_remove(sa->nic, &fw_rsrc->rule_id); + if (rc != 0) + return rc; + + fw_rsrc->rule_id.id = EFX_MAE_RSRC_ID_INVALID; + } + + --(fw_rsrc->refcnt); + + return 0; +} + static struct sfc_mae_action_set * sfc_mae_action_set_attach(struct sfc_adapter *sa, const efx_mae_actions_t *spec) @@ -253,6 +388,9 @@ sfc_mae_flow_cleanup(struct sfc_adapter *sa, SFC_ASSERT(spec_mae->rule_id.id == EFX_MAE_RSRC_ID_INVALID); + if (spec_mae->outer_rule != NULL) + sfc_mae_outer_rule_del(sa, spec_mae->outer_rule); + if (spec_mae->action_set != NULL) sfc_mae_action_set_del(sa, spec_mae->action_set); @@ -263,8 +401,8 @@ sfc_mae_flow_cleanup(struct sfc_adapter *sa, static int sfc_mae_set_ethertypes(struct sfc_mae_parse_ctx *ctx) { - efx_mae_match_spec_t *efx_spec = ctx->match_spec_action; struct sfc_mae_pattern_data *pdata = &ctx->pattern_data; + const efx_mae_field_id_t *fremap = ctx->field_ids_remap; const efx_mae_field_id_t field_ids[] = { EFX_MAE_FIELD_VLAN0_PROTO_BE, EFX_MAE_FIELD_VLAN1_PROTO_BE, @@ -279,7 +417,8 @@ sfc_mae_set_ethertypes(struct sfc_mae_parse_ctx *ctx) * no L3 item, it's 0x0000/0x0000. */ et = &pdata->ethertypes[pdata->nb_vlan_tags]; - rc = efx_mae_match_spec_field_set(efx_spec, EFX_MAE_FIELD_ETHER_TYPE_BE, + rc = efx_mae_match_spec_field_set(ctx->match_spec, + fremap[EFX_MAE_FIELD_ETHER_TYPE_BE], sizeof(et->value), (const uint8_t *)&et->value, sizeof(et->mask), @@ -296,7 +435,8 @@ sfc_mae_set_ethertypes(struct sfc_mae_parse_ctx *ctx) for (i = 0; i < pdata->nb_vlan_tags; ++i) { et = &pdata->ethertypes[i]; - rc = efx_mae_match_spec_field_set(efx_spec, field_ids[i], + rc = efx_mae_match_spec_field_set(ctx->match_spec, + fremap[field_ids[i]], sizeof(et->value), (const uint8_t *)&et->value, sizeof(et->mask), @@ -312,7 +452,7 @@ static int sfc_mae_rule_process_pattern_data(struct sfc_mae_parse_ctx *ctx, struct rte_flow_error *error) { - efx_mae_match_spec_t *efx_spec = ctx->match_spec_action; + const efx_mae_field_id_t *fremap = ctx->field_ids_remap; struct sfc_mae_pattern_data *pdata = &ctx->pattern_data; struct sfc_mae_ethertype *ethertypes = pdata->ethertypes; const rte_be16_t supported_tpids[] = { @@ -411,7 +551,8 @@ sfc_mae_rule_process_pattern_data(struct sfc_mae_parse_ctx *ctx, valuep = (const uint8_t *)&pdata->l3_next_proto_value; maskp = (const uint8_t *)&pdata->l3_next_proto_mask; - rc = efx_mae_match_spec_field_set(efx_spec, EFX_MAE_FIELD_IP_PROTO, + rc = efx_mae_match_spec_field_set(ctx->match_spec, + fremap[EFX_MAE_FIELD_IP_PROTO], sizeof(pdata->l3_next_proto_value), valuep, sizeof(pdata->l3_next_proto_mask), @@ -478,7 +619,7 @@ sfc_mae_rule_parse_item_port_id(const struct rte_flow_item *item, "Can't find RTE ethdev by the port ID"); } - rc = efx_mae_match_spec_mport_set(ctx_mae->match_spec_action, + rc = efx_mae_match_spec_mport_set(ctx_mae->match_spec, &mport_sel, NULL); if (rc != 0) { return rte_flow_error_set(error, rc, @@ -536,8 +677,7 @@ sfc_mae_rule_parse_item_phy_port(const struct rte_flow_item *item, "Failed to convert the PHY_PORT index"); } - rc = efx_mae_match_spec_mport_set(ctx_mae->match_spec_action, - &mport_v, NULL); + rc = efx_mae_match_spec_mport_set(ctx_mae->match_spec, &mport_v, NULL); if (rc != 0) { return rte_flow_error_set(error, rc, RTE_FLOW_ERROR_TYPE_ITEM, item, @@ -573,8 +713,7 @@ sfc_mae_rule_parse_item_pf(const struct rte_flow_item *item, "Failed to convert the PF ID"); } - rc = efx_mae_match_spec_mport_set(ctx_mae->match_spec_action, - &mport_v, NULL); + rc = efx_mae_match_spec_mport_set(ctx_mae->match_spec, &mport_v, NULL); if (rc != 0) { return rte_flow_error_set(error, rc, RTE_FLOW_ERROR_TYPE_ITEM, item, @@ -639,8 +778,7 @@ sfc_mae_rule_parse_item_vf(const struct rte_flow_item *item, "Failed to convert the PF + VF IDs"); } - rc = efx_mae_match_spec_mport_set(ctx_mae->match_spec_action, - &mport_v, NULL); + rc = efx_mae_match_spec_mport_set(ctx_mae->match_spec, &mport_v, NULL); if (rc != 0) { return rte_flow_error_set(error, rc, RTE_FLOW_ERROR_TYPE_ITEM, item, @@ -689,9 +827,10 @@ sfc_mae_item_build_supp_mask(const struct sfc_mae_field_locator *field_locators, static int sfc_mae_parse_item(const struct sfc_mae_field_locator *field_locators, unsigned int nb_field_locators, const uint8_t *spec, - const uint8_t *mask, efx_mae_match_spec_t *efx_spec, + const uint8_t *mask, struct sfc_mae_parse_ctx *ctx, struct rte_flow_error *error) { + const efx_mae_field_id_t *fremap = ctx->field_ids_remap; unsigned int i; int rc = 0; @@ -701,7 +840,8 @@ sfc_mae_parse_item(const struct sfc_mae_field_locator *field_locators, if (fl->field_id == SFC_MAE_FIELD_HANDLING_DEFERRED) continue; - rc = efx_mae_match_spec_field_set(efx_spec, fl->field_id, + rc = efx_mae_match_spec_field_set(ctx->match_spec, + fremap[fl->field_id], fl->size, spec + fl->ofst, fl->size, mask + fl->ofst); if (rc != 0) @@ -782,7 +922,7 @@ sfc_mae_rule_parse_item_eth(const struct rte_flow_item *item, } return sfc_mae_parse_item(flocs_eth, RTE_DIM(flocs_eth), spec, mask, - ctx_mae->match_spec_action, error); + ctx_mae, error); } static const struct sfc_mae_field_locator flocs_vlan[] = { @@ -878,8 +1018,7 @@ sfc_mae_rule_parse_item_vlan(const struct rte_flow_item *item, return 0; } - return sfc_mae_parse_item(flocs, nb_flocs, spec, mask, - ctx_mae->match_spec_action, error); + return sfc_mae_parse_item(flocs, nb_flocs, spec, mask, ctx_mae, error); } static const struct sfc_mae_field_locator flocs_ipv4[] = { @@ -956,7 +1095,7 @@ sfc_mae_rule_parse_item_ipv4(const struct rte_flow_item *item, } return sfc_mae_parse_item(flocs_ipv4, RTE_DIM(flocs_ipv4), spec, mask, - ctx_mae->match_spec_action, error); + ctx_mae, error); } static const struct sfc_mae_field_locator flocs_ipv6[] = { @@ -993,6 +1132,7 @@ sfc_mae_rule_parse_item_ipv6(const struct rte_flow_item *item, { rte_be16_t ethertype_ipv6_be = RTE_BE16(RTE_ETHER_TYPE_IPV6); struct sfc_mae_parse_ctx *ctx_mae = ctx->mae; + const efx_mae_field_id_t *fremap = ctx_mae->field_ids_remap; struct sfc_mae_pattern_data *pdata = &ctx_mae->pattern_data; struct rte_flow_item_ipv6 supp_mask; const uint8_t *spec = NULL; @@ -1034,7 +1174,7 @@ sfc_mae_rule_parse_item_ipv6(const struct rte_flow_item *item, } rc = sfc_mae_parse_item(flocs_ipv6, RTE_DIM(flocs_ipv6), spec, mask, - ctx_mae->match_spec_action, error); + ctx_mae, error); if (rc != 0) return rc; @@ -1046,8 +1186,8 @@ sfc_mae_rule_parse_item_ipv6(const struct rte_flow_item *item, vtc_flow = rte_be_to_cpu_32(vtc_flow_be); tc_mask = (vtc_flow & RTE_IPV6_HDR_TC_MASK) >> RTE_IPV6_HDR_TC_SHIFT; - rc = efx_mae_match_spec_field_set(ctx_mae->match_spec_action, - EFX_MAE_FIELD_IP_TOS, + rc = efx_mae_match_spec_field_set(ctx_mae->match_spec, + fremap[EFX_MAE_FIELD_IP_TOS], sizeof(tc_value), &tc_value, sizeof(tc_mask), &tc_mask); if (rc != 0) { @@ -1094,6 +1234,16 @@ sfc_mae_rule_parse_item_tcp(const struct rte_flow_item *item, const uint8_t *mask = NULL; int rc; + /* + * When encountered among outermost items, item TCP is invalid. + * Check which match specification is being constructed now. + */ + if (ctx_mae->match_spec != ctx_mae->match_spec_action) { + return rte_flow_error_set(error, EINVAL, + RTE_FLOW_ERROR_TYPE_ITEM, item, + "TCP in outer frame is invalid"); + } + sfc_mae_item_build_supp_mask(flocs_tcp, RTE_DIM(flocs_tcp), &supp_mask, sizeof(supp_mask)); @@ -1112,7 +1262,7 @@ sfc_mae_rule_parse_item_tcp(const struct rte_flow_item *item, return 0; return sfc_mae_parse_item(flocs_tcp, RTE_DIM(flocs_tcp), spec, mask, - ctx_mae->match_spec_action, error); + ctx_mae, error); } static const struct sfc_mae_field_locator flocs_udp[] = { @@ -1158,7 +1308,157 @@ sfc_mae_rule_parse_item_udp(const struct rte_flow_item *item, return 0; return sfc_mae_parse_item(flocs_udp, RTE_DIM(flocs_udp), spec, mask, - ctx_mae->match_spec_action, error); + ctx_mae, error); +} + +static const struct sfc_mae_field_locator flocs_tunnel[] = { + { + /* + * The size and offset values are relevant + * for Geneve and NVGRE, too. + */ + .size = RTE_SIZEOF_FIELD(struct rte_flow_item_vxlan, vni), + .ofst = offsetof(struct rte_flow_item_vxlan, vni), + }, +}; + +/* + * An auxiliary registry which allows using non-encap. field IDs + * directly when building a match specification of type ACTION. + * + * See sfc_mae_rule_parse_pattern() and sfc_mae_rule_parse_item_tunnel(). + */ +static const efx_mae_field_id_t field_ids_no_remap[] = { +#define FIELD_ID_NO_REMAP(_field) \ + [EFX_MAE_FIELD_##_field] = EFX_MAE_FIELD_##_field + + FIELD_ID_NO_REMAP(ETHER_TYPE_BE), + FIELD_ID_NO_REMAP(ETH_SADDR_BE), + FIELD_ID_NO_REMAP(ETH_DADDR_BE), + FIELD_ID_NO_REMAP(VLAN0_TCI_BE), + FIELD_ID_NO_REMAP(VLAN0_PROTO_BE), + FIELD_ID_NO_REMAP(VLAN1_TCI_BE), + FIELD_ID_NO_REMAP(VLAN1_PROTO_BE), + FIELD_ID_NO_REMAP(SRC_IP4_BE), + FIELD_ID_NO_REMAP(DST_IP4_BE), + FIELD_ID_NO_REMAP(IP_PROTO), + FIELD_ID_NO_REMAP(IP_TOS), + FIELD_ID_NO_REMAP(IP_TTL), + FIELD_ID_NO_REMAP(SRC_IP6_BE), + FIELD_ID_NO_REMAP(DST_IP6_BE), + FIELD_ID_NO_REMAP(L4_SPORT_BE), + FIELD_ID_NO_REMAP(L4_DPORT_BE), + FIELD_ID_NO_REMAP(TCP_FLAGS_BE), + +#undef FIELD_ID_NO_REMAP +}; + +/* + * An auxiliary registry which allows using "ENC" field IDs + * when building a match specification of type OUTER. + * + * See sfc_mae_rule_encap_parse_init(). + */ +static const efx_mae_field_id_t field_ids_remap_to_encap[] = { +#define FIELD_ID_REMAP_TO_ENCAP(_field) \ + [EFX_MAE_FIELD_##_field] = EFX_MAE_FIELD_ENC_##_field + + FIELD_ID_REMAP_TO_ENCAP(ETHER_TYPE_BE), + FIELD_ID_REMAP_TO_ENCAP(ETH_SADDR_BE), + FIELD_ID_REMAP_TO_ENCAP(ETH_DADDR_BE), + FIELD_ID_REMAP_TO_ENCAP(VLAN0_TCI_BE), + FIELD_ID_REMAP_TO_ENCAP(VLAN0_PROTO_BE), + FIELD_ID_REMAP_TO_ENCAP(VLAN1_TCI_BE), + FIELD_ID_REMAP_TO_ENCAP(VLAN1_PROTO_BE), + FIELD_ID_REMAP_TO_ENCAP(SRC_IP4_BE), + FIELD_ID_REMAP_TO_ENCAP(DST_IP4_BE), + FIELD_ID_REMAP_TO_ENCAP(IP_PROTO), + FIELD_ID_REMAP_TO_ENCAP(IP_TOS), + FIELD_ID_REMAP_TO_ENCAP(IP_TTL), + FIELD_ID_REMAP_TO_ENCAP(SRC_IP6_BE), + FIELD_ID_REMAP_TO_ENCAP(DST_IP6_BE), + FIELD_ID_REMAP_TO_ENCAP(L4_SPORT_BE), + FIELD_ID_REMAP_TO_ENCAP(L4_DPORT_BE), + +#undef FIELD_ID_REMAP_TO_ENCAP +}; + +static int +sfc_mae_rule_parse_item_tunnel(const struct rte_flow_item *item, + struct sfc_flow_parse_ctx *ctx, + struct rte_flow_error *error) +{ + struct sfc_mae_parse_ctx *ctx_mae = ctx->mae; + uint8_t vnet_id_v[sizeof(uint32_t)] = {0}; + uint8_t vnet_id_m[sizeof(uint32_t)] = {0}; + const struct rte_flow_item_vxlan *vxp; + uint8_t supp_mask[sizeof(uint64_t)]; + const uint8_t *spec = NULL; + const uint8_t *mask = NULL; + const void *def_mask; + int rc; + + /* + * We're about to start processing inner frame items. + * Process pattern data that has been deferred so far + * and reset pattern data storage. + */ + rc = sfc_mae_rule_process_pattern_data(ctx_mae, error); + if (rc != 0) + return rc; + + memset(&ctx_mae->pattern_data, 0, sizeof(ctx_mae->pattern_data)); + + sfc_mae_item_build_supp_mask(flocs_tunnel, RTE_DIM(flocs_tunnel), + &supp_mask, sizeof(supp_mask)); + + /* + * This tunnel item was preliminarily detected by + * sfc_mae_rule_encap_parse_init(). Default mask + * was also picked by that helper. Use it here. + */ + def_mask = ctx_mae->tunnel_def_mask; + + rc = sfc_flow_parse_init(item, + (const void **)&spec, (const void **)&mask, + (const void *)&supp_mask, def_mask, + sizeof(def_mask), error); + if (rc != 0) + return rc; + + /* + * This item and later ones comprise a + * match specification of type ACTION. + */ + ctx_mae->match_spec = ctx_mae->match_spec_action; + + /* This item and later ones use non-encap. EFX MAE field IDs. */ + ctx_mae->field_ids_remap = field_ids_no_remap; + + if (spec == NULL) + return 0; + + /* + * Field EFX_MAE_FIELD_ENC_VNET_ID_BE is a 32-bit one. + * Copy 24-bit VNI, which is BE, at offset 1 in it. + * The extra byte is 0 both in the mask and in the value. + */ + vxp = (const struct rte_flow_item_vxlan *)spec; + memcpy(vnet_id_v + 1, &vxp->vni, sizeof(vxp->vni)); + + vxp = (const struct rte_flow_item_vxlan *)mask; + memcpy(vnet_id_m + 1, &vxp->vni, sizeof(vxp->vni)); + + rc = efx_mae_match_spec_field_set(ctx_mae->match_spec, + EFX_MAE_FIELD_ENC_VNET_ID_BE, + sizeof(vnet_id_v), vnet_id_v, + sizeof(vnet_id_m), vnet_id_m); + if (rc != 0) { + rc = rte_flow_error_set(error, rc, RTE_FLOW_ERROR_TYPE_ITEM, + item, "Failed to set VXLAN VNI"); + } + + return rc; } static const struct sfc_flow_item sfc_flow_items[] = { @@ -1248,8 +1548,178 @@ static const struct sfc_flow_item sfc_flow_items[] = { .ctx_type = SFC_FLOW_PARSE_CTX_MAE, .parse = sfc_mae_rule_parse_item_udp, }, + { + .type = RTE_FLOW_ITEM_TYPE_VXLAN, + .prev_layer = SFC_FLOW_ITEM_L4, + .layer = SFC_FLOW_ITEM_START_LAYER, + .ctx_type = SFC_FLOW_PARSE_CTX_MAE, + .parse = sfc_mae_rule_parse_item_tunnel, + }, + { + .type = RTE_FLOW_ITEM_TYPE_GENEVE, + .prev_layer = SFC_FLOW_ITEM_L4, + .layer = SFC_FLOW_ITEM_START_LAYER, + .ctx_type = SFC_FLOW_PARSE_CTX_MAE, + .parse = sfc_mae_rule_parse_item_tunnel, + }, + { + .type = RTE_FLOW_ITEM_TYPE_NVGRE, + .prev_layer = SFC_FLOW_ITEM_L3, + .layer = SFC_FLOW_ITEM_START_LAYER, + .ctx_type = SFC_FLOW_PARSE_CTX_MAE, + .parse = sfc_mae_rule_parse_item_tunnel, + }, }; +static int +sfc_mae_rule_process_outer(struct sfc_adapter *sa, + struct sfc_mae_parse_ctx *ctx, + struct sfc_mae_outer_rule **rulep, + struct rte_flow_error *error) +{ + struct sfc_mae_outer_rule *rule; + int rc; + + if (ctx->encap_type == EFX_TUNNEL_PROTOCOL_NONE) { + *rulep = NULL; + return 0; + } + + SFC_ASSERT(ctx->match_spec_outer != NULL); + + if (!efx_mae_match_spec_is_valid(sa->nic, ctx->match_spec_outer)) { + return rte_flow_error_set(error, ENOTSUP, + RTE_FLOW_ERROR_TYPE_ITEM, NULL, + "Inconsistent pattern (outer)"); + } + + *rulep = sfc_mae_outer_rule_attach(sa, ctx->match_spec_outer, + ctx->encap_type); + if (*rulep != NULL) { + efx_mae_match_spec_fini(sa->nic, ctx->match_spec_outer); + } else { + rc = sfc_mae_outer_rule_add(sa, ctx->match_spec_outer, + ctx->encap_type, rulep); + if (rc != 0) { + return rte_flow_error_set(error, rc, + RTE_FLOW_ERROR_TYPE_ITEM, NULL, + "Failed to process the pattern"); + } + } + + /* + * Depending on whether we reuse an existing outer rule or create a + * new one (see above), outer rule ID is either a valid value or + * EFX_MAE_RSRC_ID_INVALID. Set it in the action rule match + * specification (and the full mask, too) in order to have correct + * class comparisons of the new rule with existing ones. + * Also, action rule match specification will be validated shortly, + * and having the full mask set for outer rule ID indicates that we + * will use this field, and support for this field has to be checked. + */ + rule = *rulep; + rc = efx_mae_match_spec_outer_rule_id_set(ctx->match_spec_action, + &rule->fw_rsrc.rule_id); + if (rc != 0) { + sfc_mae_outer_rule_del(sa, *rulep); + *rulep = NULL; + + return rte_flow_error_set(error, rc, + RTE_FLOW_ERROR_TYPE_ITEM, NULL, + "Failed to process the pattern"); + } + + return 0; +} + +static int +sfc_mae_rule_encap_parse_init(struct sfc_adapter *sa, + const struct rte_flow_item pattern[], + struct sfc_mae_parse_ctx *ctx, + struct rte_flow_error *error) +{ + struct sfc_mae *mae = &sa->mae; + int rc; + + if (pattern == NULL) { + rte_flow_error_set(error, EINVAL, + RTE_FLOW_ERROR_TYPE_ITEM_NUM, NULL, + "NULL pattern"); + return -rte_errno; + } + + for (;;) { + switch (pattern->type) { + case RTE_FLOW_ITEM_TYPE_VXLAN: + ctx->encap_type = EFX_TUNNEL_PROTOCOL_VXLAN; + ctx->tunnel_def_mask = &rte_flow_item_vxlan_mask; + RTE_BUILD_BUG_ON(sizeof(ctx->tunnel_def_mask) != + sizeof(rte_flow_item_vxlan_mask)); + break; + case RTE_FLOW_ITEM_TYPE_GENEVE: + ctx->encap_type = EFX_TUNNEL_PROTOCOL_GENEVE; + ctx->tunnel_def_mask = &rte_flow_item_geneve_mask; + RTE_BUILD_BUG_ON(sizeof(ctx->tunnel_def_mask) != + sizeof(rte_flow_item_geneve_mask)); + break; + case RTE_FLOW_ITEM_TYPE_NVGRE: + ctx->encap_type = EFX_TUNNEL_PROTOCOL_NVGRE; + ctx->tunnel_def_mask = &rte_flow_item_nvgre_mask; + RTE_BUILD_BUG_ON(sizeof(ctx->tunnel_def_mask) != + sizeof(rte_flow_item_nvgre_mask)); + break; + case RTE_FLOW_ITEM_TYPE_END: + break; + default: + ++pattern; + continue; + }; + + break; + } + + if (pattern->type == RTE_FLOW_ITEM_TYPE_END) + return 0; + + if ((mae->encap_types_supported & (1U << ctx->encap_type)) == 0) { + return rte_flow_error_set(error, ENOTSUP, + RTE_FLOW_ERROR_TYPE_ITEM, + pattern, "Unsupported tunnel item"); + } + + if (ctx->priority >= mae->nb_outer_rule_prios_max) { + return rte_flow_error_set(error, ENOTSUP, + RTE_FLOW_ERROR_TYPE_ATTR_PRIORITY, + NULL, "Unsupported priority level"); + } + + rc = efx_mae_match_spec_init(sa->nic, EFX_MAE_RULE_OUTER, ctx->priority, + &ctx->match_spec_outer); + if (rc != 0) { + return rte_flow_error_set(error, rc, + RTE_FLOW_ERROR_TYPE_ITEM, pattern, + "Failed to initialise outer rule match specification"); + } + + /* Outermost items comprise a match specification of type OUTER. */ + ctx->match_spec = ctx->match_spec_outer; + + /* Outermost items use "ENC" EFX MAE field IDs. */ + ctx->field_ids_remap = field_ids_remap_to_encap; + + return 0; +} + +static void +sfc_mae_rule_encap_parse_fini(struct sfc_adapter *sa, + struct sfc_mae_parse_ctx *ctx) +{ + if (ctx->encap_type == EFX_TUNNEL_PROTOCOL_NONE) + return; + + efx_mae_match_spec_fini(sa->nic, ctx->match_spec_outer); +} + int sfc_mae_rule_parse_pattern(struct sfc_adapter *sa, const struct rte_flow_item pattern[], @@ -1261,6 +1731,7 @@ sfc_mae_rule_parse_pattern(struct sfc_adapter *sa, int rc; memset(&ctx_mae, 0, sizeof(ctx_mae)); + ctx_mae.priority = spec->priority; ctx_mae.sa = sa; rc = efx_mae_match_spec_init(sa->nic, EFX_MAE_RULE_ACTION, @@ -1273,9 +1744,24 @@ sfc_mae_rule_parse_pattern(struct sfc_adapter *sa, goto fail_init_match_spec_action; } + /* + * As a preliminary setting, assume that there is no encapsulation + * in the pattern. That is, pattern items are about to comprise a + * match specification of type ACTION and use non-encap. field IDs. + * + * sfc_mae_rule_encap_parse_init() below may override this. + */ + ctx_mae.encap_type = EFX_TUNNEL_PROTOCOL_NONE; + ctx_mae.match_spec = ctx_mae.match_spec_action; + ctx_mae.field_ids_remap = field_ids_no_remap; + ctx.type = SFC_FLOW_PARSE_CTX_MAE; ctx.mae = &ctx_mae; + rc = sfc_mae_rule_encap_parse_init(sa, pattern, &ctx_mae, error); + if (rc != 0) + goto fail_encap_parse_init; + rc = sfc_flow_parse_pattern(sfc_flow_items, RTE_DIM(sfc_flow_items), pattern, &ctx, error); if (rc != 0) @@ -1285,6 +1771,10 @@ sfc_mae_rule_parse_pattern(struct sfc_adapter *sa, if (rc != 0) goto fail_process_pattern_data; + rc = sfc_mae_rule_process_outer(sa, &ctx_mae, &spec->outer_rule, error); + if (rc != 0) + goto fail_process_outer; + if (!efx_mae_match_spec_is_valid(sa->nic, ctx_mae.match_spec_action)) { rc = rte_flow_error_set(error, ENOTSUP, RTE_FLOW_ERROR_TYPE_ITEM, NULL, @@ -1297,8 +1787,12 @@ sfc_mae_rule_parse_pattern(struct sfc_adapter *sa, return 0; fail_validate_match_spec_action: +fail_process_outer: fail_process_pattern_data: fail_parse_pattern: + sfc_mae_rule_encap_parse_fini(sa, &ctx_mae); + +fail_encap_parse_init: efx_mae_match_spec_fini(sa->nic, ctx_mae.match_spec_action); fail_init_match_spec_action: @@ -1670,6 +2164,37 @@ sfc_mae_rules_class_cmp(struct sfc_adapter *sa, return (rc == 0) ? have_same_class : false; } +static int +sfc_mae_outer_rule_class_verify(struct sfc_adapter *sa, + struct sfc_mae_outer_rule *rule) +{ + struct sfc_mae_fw_rsrc *fw_rsrc = &rule->fw_rsrc; + struct sfc_mae_outer_rule *entry; + struct sfc_mae *mae = &sa->mae; + + if (fw_rsrc->rule_id.id != EFX_MAE_RSRC_ID_INVALID) { + /* An active rule is reused. It's class is wittingly valid. */ + return 0; + } + + TAILQ_FOREACH_REVERSE(entry, &mae->outer_rules, + sfc_mae_outer_rules, entries) { + const efx_mae_match_spec_t *left = entry->match_spec; + const efx_mae_match_spec_t *right = rule->match_spec; + + if (entry == rule) + continue; + + if (sfc_mae_rules_class_cmp(sa, left, right)) + return 0; + } + + sfc_info(sa, "for now, the HW doesn't support rule validation, and HW " + "support for outer frame pattern items is not guaranteed; " + "other than that, the items are valid from SW standpoint"); + return 0; +} + static int sfc_mae_action_rule_class_verify(struct sfc_adapter *sa, struct sfc_flow_spec_mae *spec) @@ -1722,12 +2247,20 @@ sfc_mae_flow_verify(struct sfc_adapter *sa, { struct sfc_flow_spec *spec = &flow->spec; struct sfc_flow_spec_mae *spec_mae = &spec->mae; + struct sfc_mae_outer_rule *outer_rule = spec_mae->outer_rule; + int rc; SFC_ASSERT(sfc_adapter_is_locked(sa)); if (sa->state != SFC_ADAPTER_STARTED) return EAGAIN; + if (outer_rule != NULL) { + rc = sfc_mae_outer_rule_class_verify(sa, outer_rule); + if (rc != 0) + return rc; + } + return sfc_mae_action_rule_class_verify(sa, spec_mae); } @@ -1737,6 +2270,7 @@ sfc_mae_flow_insert(struct sfc_adapter *sa, { struct sfc_flow_spec *spec = &flow->spec; struct sfc_flow_spec_mae *spec_mae = &spec->mae; + struct sfc_mae_outer_rule *outer_rule = spec_mae->outer_rule; struct sfc_mae_action_set *action_set = spec_mae->action_set; struct sfc_mae_fw_rsrc *fw_rsrc = &action_set->fw_rsrc; int rc; @@ -1744,6 +2278,13 @@ sfc_mae_flow_insert(struct sfc_adapter *sa, SFC_ASSERT(spec_mae->rule_id.id == EFX_MAE_RSRC_ID_INVALID); SFC_ASSERT(action_set != NULL); + if (outer_rule != NULL) { + rc = sfc_mae_outer_rule_enable(sa, outer_rule, + spec_mae->match_spec); + if (rc != 0) + goto fail_outer_rule_enable; + } + rc = sfc_mae_action_set_enable(sa, action_set); if (rc != 0) goto fail_action_set_enable; @@ -1760,6 +2301,10 @@ fail_action_rule_insert: (void)sfc_mae_action_set_disable(sa, action_set); fail_action_set_enable: + if (outer_rule != NULL) + (void)sfc_mae_outer_rule_disable(sa, outer_rule); + +fail_outer_rule_enable: return rc; } @@ -1770,6 +2315,7 @@ sfc_mae_flow_remove(struct sfc_adapter *sa, struct sfc_flow_spec *spec = &flow->spec; struct sfc_flow_spec_mae *spec_mae = &spec->mae; struct sfc_mae_action_set *action_set = spec_mae->action_set; + struct sfc_mae_outer_rule *outer_rule = spec_mae->outer_rule; int rc; SFC_ASSERT(spec_mae->rule_id.id != EFX_MAE_RSRC_ID_INVALID); @@ -1781,5 +2327,14 @@ sfc_mae_flow_remove(struct sfc_adapter *sa, spec_mae->rule_id.id = EFX_MAE_RSRC_ID_INVALID; - return sfc_mae_action_set_disable(sa, action_set); + rc = sfc_mae_action_set_disable(sa, action_set); + if (rc != 0) { + sfc_err(sa, "failed to disable the action set (rc = %d)", rc); + /* Despite the error, proceed with outer rule removal. */ + } + + if (outer_rule != NULL) + return sfc_mae_outer_rule_disable(sa, outer_rule); + + return 0; } diff --git a/drivers/net/sfc/sfc_mae.h b/drivers/net/sfc/sfc_mae.h index 8d9b4039f3..53ddead979 100644 --- a/drivers/net/sfc/sfc_mae.h +++ b/drivers/net/sfc/sfc_mae.h @@ -26,9 +26,21 @@ struct sfc_mae_fw_rsrc { RTE_STD_C11 union { efx_mae_aset_id_t aset_id; + efx_mae_rule_id_t rule_id; }; }; +/** Outer rule registry entry */ +struct sfc_mae_outer_rule { + TAILQ_ENTRY(sfc_mae_outer_rule) entries; + unsigned int refcnt; + efx_mae_match_spec_t *match_spec; + efx_tunnel_protocol_t encap_type; + struct sfc_mae_fw_rsrc fw_rsrc; +}; + +TAILQ_HEAD(sfc_mae_outer_rules, sfc_mae_outer_rule); + /** Action set registry entry */ struct sfc_mae_action_set { TAILQ_ENTRY(sfc_mae_action_set) entries; @@ -53,8 +65,14 @@ struct sfc_mae { uint16_t switch_port_id; /** NIC support for MAE status */ enum sfc_mae_status status; + /** Priority level limit for MAE outer rules */ + unsigned int nb_outer_rule_prios_max; /** Priority level limit for MAE action rules */ unsigned int nb_action_rule_prios_max; + /** Encapsulation support status */ + uint32_t encap_types_supported; + /** Outer rule registry */ + struct sfc_mae_outer_rules outer_rules; /** Action set registry */ struct sfc_mae_action_sets action_sets; }; @@ -149,8 +167,24 @@ struct sfc_mae_pattern_data { struct sfc_mae_parse_ctx { struct sfc_adapter *sa; efx_mae_match_spec_t *match_spec_action; + efx_mae_match_spec_t *match_spec_outer; + /* + * This points to either of the above two specifications depending + * on which part of the pattern is being parsed (outer / inner). + */ + efx_mae_match_spec_t *match_spec; + /* + * This points to either "field_ids_remap_to_encap" + * or "field_ids_no_remap" (see sfc_mae.c) depending on + * which part of the pattern is being parsed. + */ + const efx_mae_field_id_t *field_ids_remap; + /* This points to a tunnel-specific default mask. */ + const void *tunnel_def_mask; bool match_mport_set; struct sfc_mae_pattern_data pattern_data; + efx_tunnel_protocol_t encap_type; + unsigned int priority; }; int sfc_mae_attach(struct sfc_adapter *sa); -- 2.20.1