/*
- * Copyright (c) 2010, 2011 Nicira Networks.
+ * Copyright (c) 2010, 2011, 2012 Nicira, Inc.
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
#include <config.h>
#include "cfm.h"
-#include <assert.h>
#include <stdint.h>
#include <stdlib.h>
#include <string.h>
#define CCM_RDI_MASK 0x80
#define CFM_HEALTH_INTERVAL 6
struct ccm {
- uint8_t mdlevel_version; /* MD Level and Version */
- uint8_t opcode;
- uint8_t flags;
- uint8_t tlv_offset;
+ uint8_t mdlevel_version; /* MD Level and Version */
+ uint8_t opcode;
+ uint8_t flags;
+ uint8_t tlv_offset;
ovs_be32 seq;
ovs_be16 mpid;
- uint8_t maid[CCM_MAID_LEN];
+ uint8_t maid[CCM_MAID_LEN];
/* Defined by ITU-T Y.1731 should be zero */
ovs_be16 interval_ms_x; /* Transmission interval in ms. */
ovs_be64 mpid64; /* MPID in extended mode. */
uint8_t opdown; /* Operationally down. */
- uint8_t zero[5];
+ uint8_t zero[5];
/* TLV space. */
uint8_t end_tlv;
struct hmap_node hmap_node; /* Node in all_cfms list. */
uint64_t mpid;
+ bool check_tnl_key; /* Verify the tunnel key of inbound packets? */
bool extended; /* Extended mode. */
- int fault; /* Connectivity fault status. */
- int recv_fault; /* Bit mask of faults occuring on receive. */
+ bool booted; /* A full fault interval has occured. */
+ enum cfm_fault_reason fault; /* Connectivity fault status. */
+ enum cfm_fault_reason recv_fault; /* Bit mask of faults occuring on
+ receive. */
bool opup; /* Operational State. */
bool remote_opup; /* Remote Operational State. */
received. */
int health_interval; /* Number of fault_intervals since health was
recomputed. */
-
+ long long int last_tx; /* Last CCM transmission time. */
};
/* Remote MPs represent foreign network entities that are configured to have
struct hmap_node node; /* Node in 'remote_mps' map. */
bool recv; /* CCM was received since last fault check. */
- bool rdi; /* Remote Defect Indicator. Indicates remote_mp isn't
- receiving CCMs that it's expecting to. */
bool opup; /* Operational State. */
uint32_t seq; /* Most recently received sequence number. */
uint8_t num_health_ccm; /* Number of received ccm frames every
CFM_HEALTH_INTERVAL * 'fault_interval'. */
+ long long int last_rx; /* Last CCM reception time. */
};
static void
ds_put_cfm_fault(struct ds *ds, int fault)
{
- size_t length = ds->length;
int i;
for (i = 0; i < CFM_FAULT_N_REASONS; i++) {
}
}
- if (ds->length > length) {
- ds_truncate(ds, ds->length - 1);
- }
+ ds_chomp(ds, ' ');
}
static void
md_len = strlen(ovs_md_name);
ma_len = strlen(ovs_ma_name);
- assert(md_len && ma_len && md_len + ma_len + 4 <= CCM_MAID_LEN);
+ ovs_assert(md_len && ma_len && md_len + ma_len + 4 <= CCM_MAID_LEN);
cfm->maid[0] = 4; /* MD name string format. */
cfm->maid[1] = md_len; /* MD name size. */
cfm->remote_opup = true;
cfm->fault_override = -1;
cfm->health = -1;
+ cfm->last_tx = 0;
return cfm;
}
cfm->health = (rmp->num_health_ccm * 100) / exp_ccm_recvd;
cfm->health = MIN(cfm->health, 100);
rmp->num_health_ccm = 0;
- assert(cfm->health >= 0 && cfm->health <= 100);
+ ovs_assert(cfm->health >= 0 && cfm->health <= 100);
}
cfm->health_interval = 0;
}
HMAP_FOR_EACH_SAFE (rmp, rmp_next, node, &cfm->remote_mps) {
if (!rmp->recv) {
- VLOG_DBG("%s: no CCM from RMP %"PRIu64" in the last %lldms",
- cfm->name, rmp->mpid, interval);
+ VLOG_INFO("%s: Received no CCM from RMP %"PRIu64" in the last"
+ " %lldms", cfm->name, rmp->mpid,
+ time_msec() - rmp->last_rx);
hmap_remove(&cfm->remote_mps, &rmp->node);
free(rmp);
} else {
rmp->recv = false;
- if (rmp->mpid == cfm->mpid) {
- VLOG_WARN_RL(&rl,"%s: received CCM with local MPID"
- " %"PRIu64, cfm->name, rmp->mpid);
- cfm->fault |= CFM_FAULT_LOOPBACK;
- }
-
- if (rmp->rdi) {
- VLOG_DBG("%s: RDI bit flagged from RMP %"PRIu64, cfm->name,
- rmp->mpid);
- cfm->fault |= CFM_FAULT_RDI;
- }
-
if (!rmp->opup) {
cfm->remote_opup = rmp->opup;
}
cfm->fault |= CFM_FAULT_RECV;
}
- if (old_cfm_fault != cfm->fault) {
+ if (old_cfm_fault != cfm->fault && !VLOG_DROP_INFO(&rl)) {
struct ds ds = DS_EMPTY_INITIALIZER;
+ ds_put_cstr(&ds, "from [");
+ ds_put_cfm_fault(&ds, old_cfm_fault);
+ ds_put_cstr(&ds, "] to [");
ds_put_cfm_fault(&ds, cfm->fault);
- VLOG_INFO_RL(&rl, "%s: CFM fault status changed: %s", cfm->name,
- ds_cstr_ro(&ds));
+ ds_put_char(&ds, ']');
+ VLOG_INFO("%s: CFM faults changed %s.", cfm->name, ds_cstr(&ds));
ds_destroy(&ds);
}
+ cfm->booted = true;
timer_set_duration(&cfm->fault_timer, interval);
+ VLOG_DBG("%s: new fault interval", cfm->name);
}
}
}
if (cfm->ccm_interval == 0) {
- assert(cfm->extended);
+ ovs_assert(cfm->extended);
ccm->interval_ms_x = htons(cfm->ccm_interval_ms);
+ } else {
+ ccm->interval_ms_x = htons(0);
}
- if (hmap_is_empty(&cfm->remote_mps)) {
+ if (cfm->booted && hmap_is_empty(&cfm->remote_mps)) {
ccm->flags |= CCM_RDI_MASK;
}
+
+ if (cfm->last_tx) {
+ long long int delay = time_msec() - cfm->last_tx;
+ if (delay > (cfm->ccm_interval_ms * 3 / 2)) {
+ VLOG_WARN("%s: long delay of %lldms (expected %dms) sending CCM"
+ " seq %"PRIu32, cfm->name, delay, cfm->ccm_interval_ms,
+ cfm->seq);
+ }
+ }
+ cfm->last_tx = time_msec();
}
void
}
cfm->mpid = s->mpid;
+ cfm->check_tnl_key = s->check_tnl_key;
cfm->extended = s->extended;
cfm->opup = s->opup;
interval = ms_to_ccm_interval(s->interval);
cfm_should_process_flow(const struct cfm *cfm, const struct flow *flow)
{
return (ntohs(flow->dl_type) == ETH_TYPE_CFM
- && eth_addr_equals(flow->dl_dst, cfm_ccm_addr(cfm)));
+ && eth_addr_equals(flow->dl_dst, cfm_ccm_addr(cfm))
+ && (!cfm->check_tnl_key || flow->tunnel.tun_id == htonll(0)));
}
/* Updates internal statistics relevant to packet 'p'. Should be called on
* expensive changes to the network topology. It seems prudent to trigger
* them judiciously, especially when CFM is used to check slave status of
* bonds. Furthermore, faults can be maliciously triggered by crafting
- * invalid CCMs. */
+ * unexpected CCMs. */
if (memcmp(ccm->maid, cfm->maid, sizeof ccm->maid)) {
cfm->recv_fault |= CFM_FAULT_MAID;
VLOG_WARN_RL(&rl, "%s: Received unexpected remote MAID from MAC "
uint64_t ccm_mpid;
uint32_t ccm_seq;
bool ccm_opdown;
- bool fault = false;
+ enum cfm_fault_reason cfm_fault = 0;
if (cfm->extended) {
ccm_mpid = ntohll(ccm->mpid64);
ccm_seq = ntohl(ccm->seq);
if (ccm_interval != cfm->ccm_interval) {
- VLOG_WARN_RL(&rl, "%s: received a CCM with an invalid interval"
+ cfm_fault |= CFM_FAULT_INTERVAL;
+ VLOG_WARN_RL(&rl, "%s: received a CCM with an unexpected interval"
" (%"PRIu8") from RMP %"PRIu64, cfm->name,
ccm_interval, ccm_mpid);
- fault = true;
}
if (cfm->extended && ccm_interval == 0
&& ccm_interval_ms_x != cfm->ccm_interval_ms) {
- VLOG_WARN_RL(&rl, "%s: received a CCM with an invalid extended"
+ cfm_fault |= CFM_FAULT_INTERVAL;
+ VLOG_WARN_RL(&rl, "%s: received a CCM with an unexpected extended"
" interval (%"PRIu16"ms) from RMP %"PRIu64, cfm->name,
ccm_interval_ms_x, ccm_mpid);
- fault = true;
}
rmp = lookup_remote_mp(cfm, ccm_mpid);
rmp = xzalloc(sizeof *rmp);
hmap_insert(&cfm->remote_mps, &rmp->node, hash_mpid(ccm_mpid));
} else {
- cfm->recv_fault |= CFM_FAULT_OVERFLOW;
+ cfm_fault |= CFM_FAULT_OVERFLOW;
VLOG_WARN_RL(&rl,
"%s: dropped CCM with MPID %"PRIu64" from MAC "
ETH_ADDR_FMT, cfm->name, ccm_mpid,
ETH_ADDR_ARGS(eth->eth_src));
- fault = true;
}
}
+ if (ccm_rdi) {
+ cfm_fault |= CFM_FAULT_RDI;
+ VLOG_DBG("%s: RDI bit flagged from RMP %"PRIu64, cfm->name,
+ ccm_mpid);
+ }
+
VLOG_DBG("%s: received CCM (seq %"PRIu32") (mpid %"PRIu64")"
" (interval %"PRIu8") (RDI %s)", cfm->name, ccm_seq,
ccm_mpid, ccm_interval, ccm_rdi ? "true" : "false");
- if (ccm_rdi) {
- fault = true;
- }
if (rmp) {
+ if (rmp->mpid == cfm->mpid) {
+ cfm_fault |= CFM_FAULT_LOOPBACK;
+ VLOG_WARN_RL(&rl,"%s: received CCM with local MPID"
+ " %"PRIu64, cfm->name, rmp->mpid);
+ }
+
if (rmp->seq && ccm_seq != (rmp->seq + 1)) {
VLOG_WARN_RL(&rl, "%s: (mpid %"PRIu64") detected sequence"
" numbers which indicate possible connectivity"
" problems (previous %"PRIu32") (current %"PRIu32
")", cfm->name, ccm_mpid, rmp->seq, ccm_seq);
- fault = true;
}
rmp->mpid = ccm_mpid;
- rmp->recv = true;
- if (!fault) {
+ if (!cfm_fault) {
rmp->num_health_ccm++;
}
+ rmp->recv = true;
+ cfm->recv_fault |= cfm_fault;
rmp->seq = ccm_seq;
- rmp->rdi = ccm_rdi;
rmp->opup = !ccm_opdown;
+ rmp->last_rx = time_msec();
}
}
}
/* Gets the operational state of 'cfm'. 'cfm' is considered operationally down
* if it has received a CCM with the operationally down bit set from any of its
- * remote maintenance points. Returns true if 'cfm' is operationally up. False
- * otherwise. */
-bool
+ * remote maintenance points. Returns 1 if 'cfm' is operationally up, 0 if
+ * 'cfm' is operationally down, or -1 if 'cfm' has no operational state
+ * (because it isn't in extended mode). */
+int
cfm_get_opup(const struct cfm *cfm)
{
- return cfm->remote_opup;
+ if (cfm->extended) {
+ return cfm->remote_opup;
+ } else {
+ return -1;
+ }
}
/* Populates 'rmps' with an array of remote maintenance points reachable by
cfm_print_details(struct ds *ds, const struct cfm *cfm)
{
struct remote_mp *rmp;
+ int fault;
ds_put_format(ds, "---- %s ----\n", cfm->name);
ds_put_format(ds, "MPID %"PRIu64":%s%s\n", cfm->mpid,
cfm->extended ? " extended" : "",
cfm->fault_override >= 0 ? " fault_override" : "");
-
- if (cfm_get_fault(cfm)) {
+ fault = cfm_get_fault(cfm);
+ if (fault) {
ds_put_cstr(ds, "\tfault: ");
- ds_put_cfm_fault(ds, cfm_get_fault(cfm));
+ ds_put_cfm_fault(ds, fault);
ds_put_cstr(ds, "\n");
}
timer_msecs_until_expired(&cfm->fault_timer));
HMAP_FOR_EACH (rmp, node, &cfm->remote_mps) {
- ds_put_format(ds, "Remote MPID %"PRIu64":%s\n",
- rmp->mpid,
- rmp->rdi ? " rdi" : "");
+ ds_put_format(ds, "Remote MPID %"PRIu64"\n", rmp->mpid);
ds_put_format(ds, "\trecv since check: %s\n",
rmp->recv ? "true" : "false");
ds_put_format(ds, "\topstate: %s\n", rmp->opup? "up" : "down");