diff options
author | Andrew Yourtchenko <ayourtch@gmail.com> | 2017-11-17 14:38:18 +0100 |
---|---|---|
committer | Damjan Marion <dmarion.lists@gmail.com> | 2018-04-17 18:25:05 +0000 |
commit | 669d07dc016757b856e1014a415996cf9f0ebc58 (patch) | |
tree | bd86de6e168fd66563f3f81aa971403c0409bbe9 /src/plugins/abf/abf_itf_attach.c | |
parent | 2926eca95138577be8d88eb8d6a442d93f182309 (diff) |
ACL based forwarding
A poor man's flow switching or policy based rounting.
An ACL is used to match packets and is associated with a [set of] forwarding paths
that determine how to forward matched packets - collectively this association is a
'policy'.
Policies are then 'attached', in a priority order, to an interface when thaey are
encountered as an input feature. If a packet matches no policies it is forwarded
normally in the IP FIB.
This commit is used to test the "ACL-as-a-service" functionality,
which currently compiles, and the existing traffic ACL tests pass in both hash and linear modes.
Change-Id: I0b274ec9f2e645352fa898b43eb54c457e195964
Signed-off-by: Neale Ranns <nranns@cisco.com>
Signed-off-by: Andrew Yourtchenko <ayourtch@gmail.com>
Signed-off-by: Ole Troan <ot@cisco.com>
Diffstat (limited to 'src/plugins/abf/abf_itf_attach.c')
-rw-r--r-- | src/plugins/abf/abf_itf_attach.c | 758 |
1 files changed, 758 insertions, 0 deletions
diff --git a/src/plugins/abf/abf_itf_attach.c b/src/plugins/abf/abf_itf_attach.c new file mode 100644 index 00000000000..2e30db9d12f --- /dev/null +++ b/src/plugins/abf/abf_itf_attach.c @@ -0,0 +1,758 @@ +/* + * Copyright (c) 2017 Cisco and/or its affiliates. + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at: + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +#include <plugins/abf/abf_itf_attach.h> +#include <vnet/fib/fib_path_list.h> +#include <plugins/acl/exports.h> + +/** + * Forward declarations; + */ +extern vlib_node_registration_t abf_ip4_node; +extern vlib_node_registration_t abf_ip6_node; + +/** + * FIB node registered type for the bonds + */ +static fib_node_type_t abf_itf_attach_fib_node_type; + +/** + * Pool of ABF interface attachment objects + */ +abf_itf_attach_t *abf_itf_attach_pool; + +/** + * A per interface vector of attached policies. used in the data-plane + */ +static u32 **abf_per_itf[FIB_PROTOCOL_MAX]; + +/** + * Per interface values of ACL lookup context IDs. used in the data-plane + */ +static u32 *abf_alctx_per_itf[FIB_PROTOCOL_MAX]; + +/** + * ABF ACL module user id returned during the initialization + */ +static u32 abf_acl_user_id; + +/** + * A DB of attachments; key={abf_index,sw_if_index} + */ +static uword *abf_itf_attach_db; + +static u64 +abf_itf_attach_mk_key (u32 abf_index, u32 sw_if_index) +{ + u64 key; + + key = abf_index; + key = key << 32; + key |= sw_if_index; + + return (key); +} + +static abf_itf_attach_t * +abf_itf_attach_db_find (u32 abf_index, u32 sw_if_index) +{ + uword *p; + u64 key; + + key = abf_itf_attach_mk_key (abf_index, sw_if_index); + + p = hash_get (abf_itf_attach_db, key); + + if (NULL != p) + return (pool_elt_at_index (abf_itf_attach_pool, p[0])); + + return (NULL); +} + +static void +abf_itf_attach_db_add (u32 abf_index, u32 sw_if_index, abf_itf_attach_t * aia) +{ + u64 key; + + key = abf_itf_attach_mk_key (abf_index, sw_if_index); + + hash_set (abf_itf_attach_db, key, aia - abf_itf_attach_pool); +} + +static void +abf_itf_attach_db_del (u32 abf_index, u32 sw_if_index) +{ + u64 key; + + key = abf_itf_attach_mk_key (abf_index, sw_if_index); + + hash_unset (abf_itf_attach_db, key); +} + +static void +abf_itf_attach_stack (abf_itf_attach_t * aia) +{ + /* + * stack the DPO on the forwarding contributed by the path-list + */ + dpo_id_t via_dpo = DPO_INVALID; + abf_policy_t *ap; + + ap = abf_policy_get (aia->aia_abf); + + fib_path_list_contribute_forwarding (ap->ap_pl, + (FIB_PROTOCOL_IP4 == aia->aia_proto ? + FIB_FORW_CHAIN_TYPE_UNICAST_IP4 : + FIB_FORW_CHAIN_TYPE_UNICAST_IP6), + FIB_PATH_LIST_FWD_FLAG_COLLAPSE, + &via_dpo); + + dpo_stack_from_node ((FIB_PROTOCOL_IP4 == aia->aia_proto ? + abf_ip4_node.index : + abf_ip6_node.index), &aia->aia_dpo, &via_dpo); + dpo_reset (&via_dpo); +} + +static int +abf_cmp_attach_for_sort (void *v1, void *v2) +{ + const abf_itf_attach_t *aia1; + const abf_itf_attach_t *aia2; + + aia1 = abf_itf_attach_get (*(u32 *) v1); + aia2 = abf_itf_attach_get (*(u32 *) v2); + + return (aia1->aia_prio - aia2->aia_prio); +} + +void +abf_setup_acl_lc (fib_protocol_t fproto, u32 sw_if_index) +{ + u32 *acl_vec = 0; + u32 *aiai; + abf_itf_attach_t *aia; + + if (~0 == abf_alctx_per_itf[fproto][sw_if_index]) + return; + + vec_foreach (aiai, abf_per_itf[fproto][sw_if_index]) + { + aia = abf_itf_attach_get (*aiai); + vec_add1 (acl_vec, aia->aia_acl); + } + acl_plugin_set_acl_vec_for_context (abf_alctx_per_itf[fproto][sw_if_index], + acl_vec); + vec_free (acl_vec); +} + +int +abf_itf_attach (fib_protocol_t fproto, + u32 policy_id, u32 priority, u32 sw_if_index) +{ + abf_itf_attach_t *aia; + abf_policy_t *ap; + u32 api, aiai; + + api = abf_policy_find (policy_id); + + ASSERT (INDEX_INVALID != api); + ap = abf_policy_get (api); + + /* + * check this is not a duplicate + */ + aia = abf_itf_attach_db_find (policy_id, sw_if_index); + + if (NULL != aia) + return (VNET_API_ERROR_ENTRY_ALREADY_EXISTS); + + /* + * construt a new attachemnt object + */ + pool_get (abf_itf_attach_pool, aia); + + fib_node_init (&aia->aia_node, abf_itf_attach_fib_node_type); + aia->aia_prio = priority; + aia->aia_proto = fproto; + aia->aia_acl = ap->ap_acl; + aia->aia_abf = api; + aia->aia_sw_if_index = sw_if_index; + aiai = aia - abf_itf_attach_pool; + abf_itf_attach_db_add (policy_id, sw_if_index, aia); + + /* + * stack the DPO on the forwarding contributed by the path-list + */ + abf_itf_attach_stack (aia); + + /* + * Insert the policy on the interfaces list. + */ + vec_validate_init_empty (abf_per_itf[fproto], sw_if_index, NULL); + vec_add1 (abf_per_itf[fproto][sw_if_index], aia - abf_itf_attach_pool); + if (1 == vec_len (abf_per_itf[fproto][sw_if_index])) + { + /* + * when enabling the first ABF polciy on the interface + * we need to enable the interface input feature + */ + vnet_feature_enable_disable ((FIB_PROTOCOL_IP4 == fproto ? + "ip4-unicast" : + "ip6-unicast"), + (FIB_PROTOCOL_IP4 == fproto ? + "abf-input-ip4" : + "abf-input-ip6"), + sw_if_index, 1, NULL, 0); + + /* if this is the first ABF policy, we need to acquire an ACL lookup context */ + vec_validate_init_empty (abf_alctx_per_itf[fproto], sw_if_index, ~0); + abf_alctx_per_itf[fproto][sw_if_index] = + acl_plugin_get_lookup_context_index (abf_acl_user_id, sw_if_index, 0); + } + else + { + vec_sort_with_function (abf_per_itf[fproto][sw_if_index], + abf_cmp_attach_for_sort); + } + + /* Prepare and set the list of ACLs for lookup within the context */ + abf_setup_acl_lc (fproto, sw_if_index); + + /* + * become a child of the ABF poilcy so we are notified when + * its forwarding changes. + */ + aia->aia_sibling = fib_node_child_add (abf_policy_fib_node_type, + api, + abf_itf_attach_fib_node_type, aiai); + + return (0); +} + +int +abf_itf_detach (fib_protocol_t fproto, u32 policy_id, u32 sw_if_index) +{ + abf_itf_attach_t *aia; + u32 index; + + /* + * check this is a valid attahment + */ + aia = abf_itf_attach_db_find (policy_id, sw_if_index); + + if (NULL == aia) + return (VNET_API_ERROR_ENTRY_ALREADY_EXISTS); + + /* + * first remove from the interface's vecotr + */ + ASSERT (abf_per_itf[fproto]); + ASSERT (abf_per_itf[fproto][sw_if_index]); + + index = vec_search (abf_per_itf[fproto][sw_if_index], + aia - abf_itf_attach_pool); + + ASSERT (index != ~0); + vec_del1 (abf_per_itf[fproto][sw_if_index], index); + + if (0 == vec_len (abf_per_itf[fproto][sw_if_index])) + { + /* + * when deleting the last ABF polciy on the interface + * we need to disable the interface input feature + */ + vnet_feature_enable_disable ((FIB_PROTOCOL_IP4 == fproto ? + "ip4-unicast" : + "ip6-unicast"), + (FIB_PROTOCOL_IP4 == fproto ? + "abf-input-ip4" : + "abf-input-ip6"), + sw_if_index, 0, NULL, 0); + + /* Return the lookup context, invalidate its id in our records */ + acl_plugin_put_lookup_context_index (abf_alctx_per_itf[fproto] + [sw_if_index]); + abf_alctx_per_itf[fproto][sw_if_index] = ~0; + } + else + { + vec_sort_with_function (abf_per_itf[fproto][sw_if_index], + abf_cmp_attach_for_sort); + } + + /* Prepare and set the list of ACLs for lookup within the context */ + abf_setup_acl_lc (fproto, sw_if_index); + + /* + * remove the dependency on the policy + */ + fib_node_child_remove (abf_policy_fib_node_type, + aia->aia_abf, aia->aia_sibling); + + /* + * remove the attahcment from the DB + */ + abf_itf_attach_db_del (policy_id, sw_if_index); + + /* + * release our locks on FIB forwarding data + */ + dpo_reset (&aia->aia_dpo); + + /* + * return the object + */ + pool_put (abf_itf_attach_pool, aia); + + return (0); +} + +static u8 * +format_abf_intf_attach (u8 * s, va_list * args) +{ + abf_itf_attach_t *aia = va_arg (*args, abf_itf_attach_t *); + abf_policy_t *ap; + + ap = abf_policy_get (aia->aia_abf); + s = format (s, "abf-interface-attach: policy:%d prioity:%d", + ap->ap_id, aia->aia_prio); + s = format (s, "\n %U", format_dpo_id, &aia->aia_dpo, 2); + + return (s); +} + +static clib_error_t * +abf_itf_attach_cmd (vlib_main_t * vm, + unformat_input_t * input, vlib_cli_command_t * cmd) +{ + u32 policy_id, sw_if_index; + fib_protocol_t fproto; + u32 is_del, priority; + vnet_main_t *vnm; + + is_del = 0; + sw_if_index = policy_id = ~0; + vnm = vnet_get_main (); + fproto = FIB_PROTOCOL_MAX; + priority = 0; + + while (unformat_check_input (input) != UNFORMAT_END_OF_INPUT) + { + if (unformat (input, "del")) + is_del = 1; + else if (unformat (input, "add")) + is_del = 0; + else if (unformat (input, "ip4")) + fproto = FIB_PROTOCOL_IP4; + else if (unformat (input, "ip6")) + fproto = FIB_PROTOCOL_IP6; + else if (unformat (input, "policy %d", &policy_id)) + ; + else if (unformat (input, "priority %d", &priority)) + ; + else if (unformat (input, "%U", + unformat_vnet_sw_interface, vnm, &sw_if_index)) + ; + else + return (clib_error_return (0, "unknown input '%U'", + format_unformat_error, input)); + } + + if (~0 == policy_id) + { + return (clib_error_return (0, "invalid policy ID:%d", policy_id)); + } + if (~0 == sw_if_index) + { + return (clib_error_return (0, "invalid interface name")); + } + if (FIB_PROTOCOL_MAX == fproto) + { + return (clib_error_return (0, "Specify either ip4 or ip6")); + } + + if (~0 == abf_policy_find (policy_id)) + return (clib_error_return (0, "invalid policy ID:%d", policy_id)); + + if (is_del) + abf_itf_detach (fproto, policy_id, sw_if_index); + else + abf_itf_attach (fproto, policy_id, priority, sw_if_index); + + return (NULL); +} + +/* *INDENT-OFF* */ +/** + * Attach an ABF policy to an interface. + */ +VLIB_CLI_COMMAND (abf_itf_attach_cmd_node, static) = { + .path = "abf attach", + .function = abf_itf_attach_cmd, + .short_help = "abf attach <ip4|ip6> [del] policy <value> <interface>", + // this is not MP safe +}; +/* *INDENT-ON* */ + +static clib_error_t * +abf_show_attach_cmd (vlib_main_t * vm, + unformat_input_t * input, vlib_cli_command_t * cmd) +{ + const abf_itf_attach_t *aia; + u32 sw_if_index, *aiai; + fib_protocol_t fproto; + vnet_main_t *vnm; + + sw_if_index = ~0; + vnm = vnet_get_main (); + + while (unformat_check_input (input) != UNFORMAT_END_OF_INPUT) + { + if (unformat (input, "%U", + unformat_vnet_sw_interface, vnm, &sw_if_index)) + ; + else + return (clib_error_return (0, "unknown input '%U'", + format_unformat_error, input)); + } + + if (~0 == sw_if_index) + { + vlib_cli_output (vm, "specify an interface"); + } + + /* *INDENT-OFF* */ + FOR_EACH_FIB_IP_PROTOCOL(fproto) + { + if (sw_if_index < vec_len(abf_per_itf[fproto])) + { + if (vec_len(abf_per_itf[fproto][sw_if_index])) + vlib_cli_output(vm, "%U:", format_fib_protocol, fproto); + + vec_foreach(aiai, abf_per_itf[fproto][sw_if_index]) + { + aia = pool_elt_at_index(abf_itf_attach_pool, *aiai); + vlib_cli_output(vm, " %U", format_abf_intf_attach, aia); + } + } + } + /* *INDENT-ON* */ + return (NULL); +} + +/* *INDENT-OFF* */ +VLIB_CLI_COMMAND (abf_show_attach_cmd_node, static) = { + .path = "show abf attach", + .function = abf_show_attach_cmd, + .short_help = "show abf attach <interface>", + .is_mp_safe = 1, +}; +/* *INDENT-ON* */ + +void +abf_itf_attach_walk (abf_itf_attach_walk_cb_t cb, void *ctx) +{ + u32 aii; + + /* *INDENT-OFF* */ + pool_foreach_index(aii, abf_itf_attach_pool, + ({ + if (!cb(aii, ctx)) + break; + })); + /* *INDENT-ON* */ +} + +typedef enum abf_next_t_ +{ + ABF_NEXT_DROP, + ABF_N_NEXT, +} abf_next_t; + +typedef struct abf_input_trace_t_ +{ + abf_next_t next; + index_t index; +} abf_input_trace_t; + +typedef enum +{ +#define abf_error(n,s) ABF_ERROR_##n, +#include "abf_error.def" +#undef abf_error + ABF_N_ERROR, +} abf_error_t; + +always_inline uword +abf_input_inline (vlib_main_t * vm, + vlib_node_runtime_t * node, + vlib_frame_t * frame, fib_protocol_t fproto) +{ + u32 n_left_from, *from, *to_next, next_index, matches; + + from = vlib_frame_vector_args (frame); + n_left_from = frame->n_vectors; + next_index = node->cached_next_index; + matches = 0; + + while (n_left_from > 0) + { + u32 n_left_to_next; + + vlib_get_next_frame (vm, node, next_index, to_next, n_left_to_next); + + while (n_left_from > 0 && n_left_to_next > 0) + { + const u32 *attachments0; + const abf_itf_attach_t *aia0; + abf_next_t next0 = ABF_NEXT_DROP; + vlib_buffer_t *b0; + u32 bi0, sw_if_index0; + fa_5tuple_opaque_t fa_5tuple0; + u32 match_acl_index = ~0; + u32 match_acl_pos = ~0; + u32 match_rule_index = ~0; + u32 trace_bitmap = 0; + u8 action; + + bi0 = from[0]; + to_next[0] = bi0; + from += 1; + to_next += 1; + n_left_from -= 1; + n_left_to_next -= 1; + + b0 = vlib_get_buffer (vm, bi0); + sw_if_index0 = vnet_buffer (b0)->sw_if_index[VLIB_RX]; + + ASSERT (vec_len (abf_per_itf[fproto]) > sw_if_index0); + attachments0 = abf_per_itf[fproto][sw_if_index0]; + + ASSERT (vec_len (abf_alctx_per_itf[fproto]) > sw_if_index0); + /* + * check if any of the policies attached to this interface matches. + */ + u32 lc_index = abf_alctx_per_itf[fproto][sw_if_index0]; + + acl_plugin_fill_5tuple (lc_index, b0, (FIB_PROTOCOL_IP6 == fproto), + 1, 0, &fa_5tuple0); + + if (acl_plugin_match_5tuple + (lc_index, &fa_5tuple0, (FIB_PROTOCOL_IP6 == fproto), &action, + &match_acl_pos, &match_acl_index, &match_rule_index, + &trace_bitmap)) + { + /* + * match: + * follow the DPO chain + */ + aia0 = abf_itf_attach_get (attachments0[match_acl_pos]); + + next0 = aia0->aia_dpo.dpoi_next_node; + vnet_buffer (b0)->ip.adj_index[VLIB_TX] = + aia0->aia_dpo.dpoi_index; + matches++; + } + else + { + /* + * miss: + * move on down the feature arc + */ + vnet_feature_next (sw_if_index0, &next0, b0); + } + + if (PREDICT_FALSE (b0->flags & VLIB_BUFFER_IS_TRACED)) + { + abf_input_trace_t *tr; + + tr = vlib_add_trace (vm, node, b0, sizeof (*tr)); + tr->next = next0; + tr->index = vnet_buffer (b0)->ip.adj_index[VLIB_TX]; + } + + /* verify speculative enqueue, maybe switch current next frame */ + vlib_validate_buffer_enqueue_x1 (vm, node, next_index, + to_next, n_left_to_next, bi0, + next0); + } + + vlib_put_next_frame (vm, node, next_index, n_left_to_next); + } + + vlib_node_increment_counter (vm, + (fproto = FIB_PROTOCOL_IP6 ? + abf_ip4_node.index : + abf_ip6_node.index), + ABF_ERROR_MATCHED, matches); + + return frame->n_vectors; +} + +static uword +abf_input_ip4 (vlib_main_t * vm, + vlib_node_runtime_t * node, vlib_frame_t * frame) +{ + return abf_input_inline (vm, node, frame, FIB_PROTOCOL_IP4); +} + +static uword +abf_input_ip6 (vlib_main_t * vm, + vlib_node_runtime_t * node, vlib_frame_t * frame) +{ + return abf_input_inline (vm, node, frame, FIB_PROTOCOL_IP6); +} + +static u8 * +format_abf_input_trace (u8 * s, va_list * args) +{ + CLIB_UNUSED (vlib_main_t * vm) = va_arg (*args, vlib_main_t *); + CLIB_UNUSED (vlib_node_t * node) = va_arg (*args, vlib_node_t *); + abf_input_trace_t *t = va_arg (*args, abf_input_trace_t *); + + s = format (s, " next %d index %d", t->next, t->index); + return s; +} + +static char *abf_error_strings[] = { +#define abf_error(n,s) s, +#include "abf_error.def" +#undef abf_error +}; + +/* *INDENT-OFF* */ +VLIB_REGISTER_NODE (abf_ip4_node) = +{ + .function = abf_input_ip4, + .name = "abf-input-ip4", + .vector_size = sizeof (u32), + .format_trace = format_abf_input_trace, + .type = VLIB_NODE_TYPE_INTERNAL, + .n_errors = ABF_N_ERROR, + .error_strings = abf_error_strings, + .n_next_nodes = ABF_N_NEXT, + .next_nodes = + { + [ABF_NEXT_DROP] = "error-drop", + } +}; + +VLIB_REGISTER_NODE (abf_ip6_node) = +{ + .function = abf_input_ip6, + .name = "abf-input-ip6", + .vector_size = sizeof (u32), + .format_trace = format_abf_input_trace, + .type = VLIB_NODE_TYPE_INTERNAL, + .n_errors = 0, + .n_next_nodes = ABF_N_NEXT, + + .next_nodes = + { + [ABF_NEXT_DROP] = "error-drop", + } +}; + +VNET_FEATURE_INIT (abf_ip4_feat, static) = +{ + .arc_name = "ip4-unicast", + .node_name = "abf-input-ip4", + .runs_after = VNET_FEATURES ("acl-plugin-in-ip4-fa"), +}; + +VNET_FEATURE_INIT (abf_ip6_feat, static) = +{ + .arc_name = "ip6-unicast", + .node_name = "abf-input-ip6", + .runs_after = VNET_FEATURES ("acl-plugin-in-ip6-fa"), +}; +/* *INDENT-ON* */ + +static fib_node_t * +abf_itf_attach_get_node (fib_node_index_t index) +{ + abf_itf_attach_t *aia = abf_itf_attach_get (index); + return (&(aia->aia_node)); +} + +static abf_itf_attach_t * +abf_itf_attach_get_from_node (fib_node_t * node) +{ + return ((abf_itf_attach_t *) (((char *) node) - + STRUCT_OFFSET_OF (abf_itf_attach_t, + aia_node))); +} + +static void +abf_itf_attach_last_lock_gone (fib_node_t * node) +{ + /* + * ABF interface attachments are leaves on the graph. + * we do not manage locks from children. + */ +} + +/* + * abf_itf_attach_back_walk_notify + * + * A back walk has reached this BIER fmask + */ +static fib_node_back_walk_rc_t +abf_itf_attach_back_walk_notify (fib_node_t * node, + fib_node_back_walk_ctx_t * ctx) +{ + /* + * re-stack the fmask on the n-eos of the via + */ + abf_itf_attach_t *aia = abf_itf_attach_get_from_node (node); + + abf_itf_attach_stack (aia); + + return (FIB_NODE_BACK_WALK_CONTINUE); +} + +/* + * The BIER fmask's graph node virtual function table + */ +static const fib_node_vft_t abf_itf_attach_vft = { + .fnv_get = abf_itf_attach_get_node, + .fnv_last_lock = abf_itf_attach_last_lock_gone, + .fnv_back_walk = abf_itf_attach_back_walk_notify, +}; + +static clib_error_t * +abf_itf_bond_init (vlib_main_t * vm) +{ + abf_itf_attach_fib_node_type = + fib_node_register_new_type (&abf_itf_attach_vft); + clib_error_t *acl_init_res = acl_plugin_exports_init (); + if (acl_init_res) + return (acl_init_res); + + abf_acl_user_id = + acl_plugin_register_user_module ("abp plugin", "sw_if_index", NULL); + + return (NULL); +} + +VLIB_INIT_FUNCTION (abf_itf_bond_init); + +/* + * fd.io coding-style-patch-verification: ON + * + * Local Variables: + * eval: (c-set-style "gnu") + * End: + */ |