diff options
Diffstat (limited to 'src/feature')
45 files changed, 1545 insertions, 171 deletions
diff --git a/src/feature/api/tor_api.c b/src/feature/api/tor_api.c index 051be50b3a..88e91ebfd5 100644 --- a/src/feature/api/tor_api.c +++ b/src/feature/api/tor_api.c @@ -18,9 +18,9 @@ // Include this after the above headers, to insure that they don't // depend on anything else. #include "orconfig.h" +#include "lib/cc/compat_compiler.h" #include "lib/cc/torint.h" #include "feature/api/tor_api_internal.h" -#include "lib/cc/compat_compiler.h" #include <stdio.h> #include <stdlib.h> @@ -39,7 +39,9 @@ #include "lib/net/socketpair.h" #define raw_socketpair tor_ersatz_socketpair #define raw_closesocket closesocket +#if !defined(HAVE_SNPRINTF) #define snprintf _snprintf +#endif #else /* !defined(_WIN32) */ #define raw_socketpair socketpair #define raw_closesocket close diff --git a/src/feature/client/bridges.c b/src/feature/client/bridges.c index d40bcc6c8e..9e36d26929 100644 --- a/src/feature/client/bridges.c +++ b/src/feature/client/bridges.c @@ -943,9 +943,17 @@ rewrite_node_address_for_bridge(const bridge_info_t *bridge, node_t *node) } /** We just learned a descriptor for a bridge. See if that - * digest is in our entry guard list, and add it if not. */ + * digest is in our entry guard list, and add it if not. Schedule the + * next fetch for a long time from now, and initiate any follow-up + * activities like continuing to bootstrap. + * + * <b>from_cache</b> * tells us whether we fetched it from disk (else + * the network) + * + * <b>desc_is_new</b> tells us if we preferred it to the old version we + * had, if any. */ void -learned_bridge_descriptor(routerinfo_t *ri, int from_cache) +learned_bridge_descriptor(routerinfo_t *ri, int from_cache, int desc_is_new) { tor_assert(ri); tor_assert(ri->purpose == ROUTER_PURPOSE_BRIDGE); @@ -961,12 +969,14 @@ learned_bridge_descriptor(routerinfo_t *ri, int from_cache) if (bridge) { /* if we actually want to use this one */ node_t *node; - /* it's here; schedule its re-fetch for a long time from now. */ if (!from_cache) { /* This schedules the re-fetch at a constant interval, which produces * a pattern of bridge traffic. But it's better than trying all * configured bridges several times in the first few minutes. */ download_status_reset(&bridge->fetch_status); + /* it's here; schedule its re-fetch for a long time from now. */ + bridge->fetch_status.next_attempt_at += + get_options()->TestingBridgeDownloadInitialDelay; } node = node_get_mutable_by_id(ri->cache_info.identity_digest); @@ -982,8 +992,10 @@ learned_bridge_descriptor(routerinfo_t *ri, int from_cache) entry_guard_learned_bridge_identity(&bridge->addrport_configured, (const uint8_t*)ri->cache_info.identity_digest); - log_notice(LD_DIR, "new bridge descriptor '%s' (%s): %s", ri->nickname, - from_cache ? "cached" : "fresh", router_describe(ri)); + if (desc_is_new) + log_notice(LD_DIR, "new bridge descriptor '%s' (%s): %s", + ri->nickname, + from_cache ? "cached" : "fresh", router_describe(ri)); /* If we didn't have a reachable bridge before this one, try directory * documents again. */ if (first) { diff --git a/src/feature/client/bridges.h b/src/feature/client/bridges.h index a42363f683..dd3e498a0a 100644 --- a/src/feature/client/bridges.h +++ b/src/feature/client/bridges.h @@ -46,7 +46,8 @@ void learned_router_identity(const tor_addr_t *addr, uint16_t port, void bridge_add_from_config(struct bridge_line_t *bridge_line); void retry_bridge_descriptor_fetch_directly(const char *digest); void fetch_bridge_descriptors(const or_options_t *options, time_t now); -void learned_bridge_descriptor(routerinfo_t *ri, int from_cache); +void learned_bridge_descriptor(routerinfo_t *ri, + int from_cache, int desc_is_new); const smartlist_t *get_socks_args_by_bridge_addrport(const tor_addr_t *addr, uint16_t port); diff --git a/src/feature/client/circpathbias.c b/src/feature/client/circpathbias.c index 9c0ecc56ad..ff9e05a645 100644 --- a/src/feature/client/circpathbias.c +++ b/src/feature/client/circpathbias.c @@ -363,6 +363,17 @@ pathbias_should_count(origin_circuit_t *circ) return 0; } + /* Ignore circuits where the controller helped choose the path. When + * this happens, we can't be sure whether the path was chosen randomly + * or not. */ + if (circ->any_hop_from_controller) { + /* (In this case, we _don't_ check to see if shouldcount is changing, + * since it's possible that an already-created circuit later gets extended + * by the controller. */ + circ->pathbias_shouldcount = PATHBIAS_SHOULDCOUNT_IGNORED; + return 0; + } + /* Completely ignore one hop circuits */ if (circ->build_state->onehop_tunnel || circ->build_state->desired_path_len == 1) { diff --git a/src/feature/client/entrynodes.c b/src/feature/client/entrynodes.c index 502cb99690..32ecb4f705 100644 --- a/src/feature/client/entrynodes.c +++ b/src/feature/client/entrynodes.c @@ -132,6 +132,7 @@ #include "feature/client/entrynodes.h" #include "feature/client/transports.h" #include "feature/control/control_events.h" +#include "feature/dirclient/dlstatus.h" #include "feature/dircommon/directory.h" #include "feature/nodelist/describe.h" #include "feature/nodelist/microdesc.h" @@ -559,7 +560,7 @@ get_extreme_restriction_threshold(void) int32_t pct = networkstatus_get_param(NULL, "guard-extreme-restriction-percent", DFLT_EXTREME_RESTRICTION_PERCENT, - 1, INT32_MAX); + 1, 100); return pct / 100.0; } @@ -576,6 +577,18 @@ mark_guard_maybe_reachable(entry_guard_t *guard) guard->is_reachable = GUARD_REACHABLE_MAYBE; if (guard->is_filtered_guard) guard->is_usable_filtered_guard = 1; + + /* Check if it is a bridge and we don't have its descriptor yet */ + if (guard->bridge_addr && !guard_has_descriptor(guard)) { + /* Reset the descriptor fetch retry schedule, so it gives it another + * go soon. It's important to keep any "REACHABLE_MAYBE" bridges in + * sync with the descriptor fetch schedule, since we will refuse to + * use the network until our first primary bridges are either + * known-usable or known-unusable. See bug 40396. */ + download_status_t *dl = get_bridge_dl_status_by_id(guard->identity); + if (dl) + download_status_reset(dl); + } } /** @@ -2046,6 +2059,14 @@ entry_guard_consider_retry(entry_guard_t *guard) get_retry_schedule(guard->failing_since, now, guard->is_primary); const time_t last_attempt = guard->last_tried_to_connect; + /* Check if it is a bridge and we don't have its descriptor yet */ + if (guard->bridge_addr && !guard_has_descriptor(guard)) { + /* We want to leave the retry schedule to fetch_bridge_descriptors(), + * so we don't have two retry schedules clobbering each other. See + * bugs 40396 and 40497 for details of why we need this exception. */ + return; + } + if (BUG(last_attempt == 0) || now >= last_attempt + delay) { /* We should mark this retriable. */ @@ -2271,6 +2292,13 @@ entry_guards_note_guard_failure(guard_selection_t *gs, guard->is_primary?"primary ":"", guard->confirmed_idx>=0?"confirmed ":"", entry_guard_describe(guard)); + + /* Schedule a re-assessment of whether we have enough dir info to + * use the network. Counterintuitively, *losing* a bridge might actually + * be just what we need to *resume* using the network, if we had it in + * state GUARD_REACHABLE_MAYBE and we were stalling to learn this + * outcome. See bug 40396 for more details. */ + router_dir_info_changed(); } /** @@ -2295,6 +2323,12 @@ entry_guards_note_guard_success(guard_selection_t *gs, /* If guard was not already marked as reachable, send a GUARD UP signal */ if (guard->is_reachable != GUARD_REACHABLE_YES) { control_event_guard(guard->nickname, guard->identity, "UP"); + + /* Schedule a re-assessment of whether we have enough dir info to + * use the network. One of our guards has just moved to + * GUARD_REACHABLE_YES, so maybe we can resume using the network + * now. */ + router_dir_info_changed(); } guard->is_reachable = GUARD_REACHABLE_YES; @@ -3538,6 +3572,11 @@ entry_guards_changed_for_guard_selection(guard_selection_t *gs) entry_guards_update_guards_in_state() */ or_state_mark_dirty(get_or_state(), when); + + /* Schedule a re-assessment of whether we have enough dir info to + * use the network. When we add or remove or disable or enable a + * guard, the decision could shift. */ + router_dir_info_changed(); } /** Our list of entry guards has changed for the default guard selection @@ -3930,6 +3969,253 @@ guard_selection_free_(guard_selection_t *gs) tor_free(gs); } +/**********************************************************************/ + +/** Layer2 guard subsystem (vanguards-lite) used for onion service circuits */ + +/** A simple representation of a layer2 guard. We just need its identity so + * that we feed it into a routerset, and a sampled timestamp to do expiration + * checks. */ +typedef struct layer2_guard_t { + /** Identity of the guard */ + char identity[DIGEST_LEN]; + /** When does this guard expire? (randomized timestamp) */ + time_t expire_on_date; +} layer2_guard_t; + +#define layer2_guard_free(val) \ + FREE_AND_NULL(layer2_guard_t, layer2_guard_free_, (val)) + +/** Return true if the vanguards-lite subsystem is enabled */ +bool +vanguards_lite_is_enabled(void) +{ + /* First check torrc option and then maybe also the consensus parameter. */ + const or_options_t *options = get_options(); + + /* If the option is explicitly disabled, that's the final word here */ + if (options->VanguardsLiteEnabled == 0) { + return false; + } + + /* If the option is set to auto, then check the consensus parameter */ + if (options->VanguardsLiteEnabled == -1) { + return networkstatus_get_param(NULL, "vanguards-lite-enabled", + 1, /* default to "on" */ + 0, 1); + } + + /* else it's enabled */ + tor_assert_nonfatal(options->VanguardsLiteEnabled == 1); + return options->VanguardsLiteEnabled; +} + +static void +layer2_guard_free_(layer2_guard_t *l2) +{ + if (!l2) { + return; + } + + tor_free(l2); +} + +/** Global list and routerset of L2 guards. They are both synced and they get + * updated periodically. We need both the list and the routerset: we use the + * smartlist to keep track of expiration times and the routerset is what we + * return to the users of this subsystem. */ +static smartlist_t *layer2_guards = NULL; +static routerset_t *layer2_routerset = NULL; + +/** Number of L2 guards */ +#define NUMBER_SECOND_GUARDS 4 +/** Make sure that the number of L2 guards is less than the number of + * MAX_SANE_RESTRICTED_NODES */ +CTASSERT(NUMBER_SECOND_GUARDS < 20); + +/** Lifetime of L2 guards: + * 1 to 12 days, for an average of a week using the max(x,x) distribution */ +#define MIN_SECOND_GUARD_LIFETIME (3600*24) +#define MAX_SECOND_GUARD_LIFETIME (3600*24*12) + +/** Return the number of guards our L2 guardset should have */ +static int +get_number_of_layer2_hs_guards(void) +{ + return (int) networkstatus_get_param(NULL, + "guard-hs-l2-number", + NUMBER_SECOND_GUARDS, + 1, 19); +} + +/** Return the minimum lifetime of L2 guards */ +static int +get_min_lifetime_of_layer2_hs_guards(void) +{ + return (int) networkstatus_get_param(NULL, + "guard-hs-l2-lifetime-min", + MIN_SECOND_GUARD_LIFETIME, + 1, INT32_MAX); +} + +/** Return the maximum lifetime of L2 guards */ +static int +get_max_lifetime_of_layer2_hs_guards(void) +{ + return (int) networkstatus_get_param(NULL, + "guard-hs-l2-lifetime-max", + MAX_SECOND_GUARD_LIFETIME, + 1, INT32_MAX); +} + +/** + * Sample and return a lifetime for an L2 guard. + * + * Lifetime randomized uniformly between min and max consensus params. + */ +static int +get_layer2_hs_guard_lifetime(void) +{ + int min = get_min_lifetime_of_layer2_hs_guards(); + int max = get_max_lifetime_of_layer2_hs_guards(); + + if (BUG(min >= max)) { + return min; + } + + return crypto_rand_int_range(min, max); +} + +/** Maintain the L2 guard list. Make sure the list contains enough guards, do + * expirations as necessary, and keep all the data structures of this + * subsystem synchronized */ +void +maintain_layer2_guards(void) +{ + if (!router_have_minimum_dir_info()) { + return; + } + + /* Create the list if it doesn't exist */ + if (!layer2_guards) { + layer2_guards = smartlist_new(); + } + + /* Go through the list and perform any needed expirations */ + SMARTLIST_FOREACH_BEGIN(layer2_guards, layer2_guard_t *, g) { + /* Expire based on expiration date */ + if (g->expire_on_date <= approx_time()) { + log_info(LD_GENERAL, "Removing expired Layer2 guard %s", + safe_str_client(hex_str(g->identity, DIGEST_LEN))); + // Nickname may be gone from consensus and doesn't matter anyway + control_event_guard("None", g->identity, "BAD_L2"); + layer2_guard_free(g); + SMARTLIST_DEL_CURRENT_KEEPORDER(layer2_guards, g); + continue; + } + + /* Expire if relay has left consensus */ + if (router_get_consensus_status_by_id(g->identity) == NULL) { + log_info(LD_GENERAL, "Removing missing Layer2 guard %s", + safe_str_client(hex_str(g->identity, DIGEST_LEN))); + // Nickname may be gone from consensus and doesn't matter anyway + control_event_guard("None", g->identity, "BAD_L2"); + layer2_guard_free(g); + SMARTLIST_DEL_CURRENT_KEEPORDER(layer2_guards, g); + continue; + } + } SMARTLIST_FOREACH_END(g); + + /* Find out how many guards we need to add */ + int new_guards_needed_n = + get_number_of_layer2_hs_guards() - smartlist_len(layer2_guards); + if (new_guards_needed_n <= 0) { + return; + } + + log_info(LD_GENERAL, "Adding %d guards to Layer2 routerset", + new_guards_needed_n); + + /* Add required guards to the list */ + smartlist_t *excluded = smartlist_new(); + for (int i = 0; i < new_guards_needed_n; i++) { + const node_t *choice = NULL; + const or_options_t *options = get_options(); + /* Pick Stable nodes */ + router_crn_flags_t flags = CRN_NEED_DESC|CRN_NEED_UPTIME; + choice = router_choose_random_node(excluded, options->ExcludeNodes, flags); + if (!choice) { + break; + } + + /* We found our node: create an L2 guard out of it */ + layer2_guard_t *layer2_guard = tor_malloc_zero(sizeof(layer2_guard_t)); + memcpy(layer2_guard->identity, choice->identity, DIGEST_LEN); + layer2_guard->expire_on_date = approx_time() + + get_layer2_hs_guard_lifetime(); + smartlist_add(layer2_guards, layer2_guard); + log_info(LD_GENERAL, "Adding Layer2 guard %s", + safe_str_client(hex_str(layer2_guard->identity, DIGEST_LEN))); + // Nickname can also be None here because it is looked up later + control_event_guard("None", layer2_guard->identity, + "GOOD_L2"); + /* Exclude this node and its family so that we don't double-pick. */ + nodelist_add_node_and_family(excluded, choice); + } + + /* Some cleanup */ + smartlist_free(excluded); + + /* Now that the list is up to date, synchronize the routerset */ + routerset_free(layer2_routerset); + layer2_routerset = routerset_new(); + + SMARTLIST_FOREACH_BEGIN (layer2_guards, layer2_guard_t *, g) { + routerset_parse(layer2_routerset, + hex_str(g->identity, DIGEST_LEN), + "l2 guards"); + } SMARTLIST_FOREACH_END(g); +} + +/** + * Reset vanguards-lite list(s). + * + * Used for SIGNAL NEWNYM. + */ +void +purge_vanguards_lite(void) +{ + if (!layer2_guards) + return; + + /* Go through the list and perform any needed expirations */ + SMARTLIST_FOREACH_BEGIN(layer2_guards, layer2_guard_t *, g) { + layer2_guard_free(g); + } SMARTLIST_FOREACH_END(g); + + smartlist_clear(layer2_guards); + + /* Pick new l2 guards */ + maintain_layer2_guards(); +} + +/** Return a routerset containing the L2 guards or NULL if it's not yet + * initialized. Callers must not free the routerset. Designed for use in + * pick_vanguard_middle_node() and should not be used anywhere else. Do not + * store this pointer -- any future calls to maintain_layer2_guards() and + * purge_vanguards_lite() can invalidate it. */ +const routerset_t * +get_layer2_guards(void) +{ + if (!layer2_guards) { + maintain_layer2_guards(); + } + + return layer2_routerset; +} + +/*****************************************************************************/ + /** Release all storage held by the list of entry guards and related * memory structs. */ void @@ -3946,4 +4232,15 @@ entry_guards_free_all(void) guard_contexts = NULL; } circuit_build_times_free_timeouts(get_circuit_build_times_mutable()); + + if (!layer2_guards) { + return; + } + + SMARTLIST_FOREACH_BEGIN(layer2_guards, layer2_guard_t *, g) { + layer2_guard_free(g); + } SMARTLIST_FOREACH_END(g); + + smartlist_free(layer2_guards); + routerset_free(layer2_routerset); } diff --git a/src/feature/client/entrynodes.h b/src/feature/client/entrynodes.h index 88ed8f649e..08fd7cf745 100644 --- a/src/feature/client/entrynodes.h +++ b/src/feature/client/entrynodes.h @@ -651,4 +651,9 @@ guard_get_guardfraction_bandwidth(guardfraction_bandwidth_t *guardfraction_bw, int orig_bandwidth, uint32_t guardfraction_percentage); +bool vanguards_lite_is_enabled(void); +const routerset_t *get_layer2_guards(void); +void maintain_layer2_guards(void); +void purge_vanguards_lite(void); + #endif /* !defined(TOR_ENTRYNODES_H) */ diff --git a/src/feature/client/transports.c b/src/feature/client/transports.c index 167beb96c6..80903ac9e5 100644 --- a/src/feature/client/transports.c +++ b/src/feature/client/transports.c @@ -843,7 +843,7 @@ handle_methods_done(const managed_proxy_t *mp) tor_assert(mp->transports); if (smartlist_len(mp->transports) == 0) - log_notice(LD_GENERAL, "Managed proxy '%s' was spawned successfully, " + log_warn(LD_GENERAL, "Managed proxy '%s' was spawned successfully, " "but it didn't launch any pluggable transport listeners!", mp->argv[0]); @@ -903,14 +903,22 @@ handle_proxy_line(const char *line, managed_proxy_t *mp) if (mp->conf_state != PT_PROTO_ACCEPTING_METHODS) goto err; + /* Log the error but do not kill the managed proxy. + * A proxy may contain several transports and if one + * of them is misconfigured, we still want to use + * the other transports. A managed proxy with no usable + * transports will log a warning. + * See https://gitlab.torproject.org/tpo/core/tor/-/issues/7362 + * */ parse_client_method_error(line); - goto err; + return; } else if (!strcmpstart(line, PROTO_SMETHOD_ERROR)) { if (mp->conf_state != PT_PROTO_ACCEPTING_METHODS) goto err; + /* Log the error but do not kill the managed proxy */ parse_server_method_error(line); - goto err; + return; } else if (!strcmpstart(line, PROTO_CMETHOD)) { if (mp->conf_state != PT_PROTO_ACCEPTING_METHODS) goto err; diff --git a/src/feature/control/control_cmd.c b/src/feature/control/control_cmd.c index bd0d41d29e..b19935e456 100644 --- a/src/feature/control/control_cmd.c +++ b/src/feature/control/control_cmd.c @@ -817,6 +817,8 @@ handle_control_extendcircuit(control_connection_t *conn, circ->first_hop_from_controller = 1; } + circ->any_hop_from_controller = 1; + /* now circ refers to something that is ready to be extended */ first_node = zero_circ; SMARTLIST_FOREACH(nodes, const node_t *, node, @@ -1075,7 +1077,7 @@ static const control_cmd_syntax_t redirectstream_syntax = { .max_args = UINT_MAX, // XXX should be 3. }; -/** Called when we receive a REDIRECTSTERAM command. Try to change the target +/** Called when we receive a REDIRECTSTREAM command. Try to change the target * address of the named AP stream, and report success or failure. */ static int handle_control_redirectstream(control_connection_t *conn, diff --git a/src/feature/dirauth/dirauth_options.inc b/src/feature/dirauth/dirauth_options.inc index 05726b8c2f..4fd07a8859 100644 --- a/src/feature/dirauth/dirauth_options.inc +++ b/src/feature/dirauth/dirauth_options.inc @@ -27,6 +27,10 @@ CONF_VAR(AuthDirHasIPv6Connectivity, BOOL, 0, "0") * good. */ CONF_VAR(AuthDirListBadExits, BOOL, 0, "0") +/** True iff we should list middle-only relays, and vote for all other + * relays as possibly suitable for other positions. */ +CONF_VAR(AuthDirListMiddleOnly, BOOL, 0, "0") + /** Do not permit more than this number of servers per IP address. */ CONF_VAR(AuthDirMaxServersPerAddr, POSINT, 0, "2") @@ -109,4 +113,7 @@ CONF_VAR(VersioningAuthoritativeDirectory, BOOL, 0, "0") * pressure or not. */ CONF_VAR(AuthDirRejectRequestsUnderLoad, BOOL, 0, "1") +/** Boolean: Should we not give bandwidth weight measurements to dirauths? */ +CONF_VAR(AuthDirDontVoteOnDirAuthBandwidth, BOOL, 0, "1") + END_CONF_STRUCT(dirauth_options_t) diff --git a/src/feature/dirauth/dirvote.c b/src/feature/dirauth/dirvote.c index fa906c0c3c..cdd2c132ef 100644 --- a/src/feature/dirauth/dirvote.c +++ b/src/feature/dirauth/dirvote.c @@ -1479,6 +1479,21 @@ compute_nth_protocol_set(int n, int n_voters, const smartlist_t *votes) return result; } +/** Helper: Takes a smartlist of `const char *` flags, and a flag to remove. + * + * Removes that flag if it is present in the list. Doesn't free it. + */ +static void +remove_flag(smartlist_t *sl, const char *flag) +{ + /* We can't use smartlist_string_remove() here, since that doesn't preserve + * order, and since it frees elements from the string. */ + + int idx = smartlist_string_pos(sl, flag); + if (idx >= 0) + smartlist_del_keeporder(sl, idx); +} + /** Given a list of vote networkstatus_t in <b>votes</b>, our public * authority <b>identity_key</b>, our private authority <b>signing_key</b>, * and the number of <b>total_authorities</b> that we believe exist in our @@ -1633,6 +1648,9 @@ networkstatus_compute_consensus(smartlist_t *votes, tor_free(votesec_list); tor_free(distsec_list); } + // True if anybody is voting on the BadExit flag. + const bool badexit_flag_is_listed = + smartlist_contains_string(flags, "BadExit"); chunks = smartlist_new(); @@ -1924,7 +1942,7 @@ networkstatus_compute_consensus(smartlist_t *votes, const char *chosen_name = NULL; int exitsummary_disagreement = 0; int is_named = 0, is_unnamed = 0, is_running = 0, is_valid = 0; - int is_guard = 0, is_exit = 0, is_bad_exit = 0; + int is_guard = 0, is_exit = 0, is_bad_exit = 0, is_middle_only = 0; int naming_conflict = 0; int n_listing = 0; char microdesc_digest[DIGEST256_LEN]; @@ -2055,7 +2073,6 @@ networkstatus_compute_consensus(smartlist_t *votes, } /* Set the flags. */ - smartlist_add(chosen_flags, (char*)"s"); /* for the start of the line. */ SMARTLIST_FOREACH_BEGIN(flags, const char *, fl) { if (!strcmp(fl, "Named")) { if (is_named) @@ -2077,6 +2094,8 @@ networkstatus_compute_consensus(smartlist_t *votes, is_running = 1; else if (!strcmp(fl, "BadExit")) is_bad_exit = 1; + else if (!strcmp(fl, "MiddleOnly")) + is_middle_only = 1; else if (!strcmp(fl, "Valid")) is_valid = 1; } @@ -2093,6 +2112,22 @@ networkstatus_compute_consensus(smartlist_t *votes, if (!is_valid) continue; + /* Starting with consensus method 32, we handle the middle-only + * flag specially: when it is present, we clear some flags, and + * set others. */ + if (is_middle_only && consensus_method >= MIN_METHOD_FOR_MIDDLEONLY) { + remove_flag(chosen_flags, "Exit"); + remove_flag(chosen_flags, "V2Dir"); + remove_flag(chosen_flags, "Guard"); + remove_flag(chosen_flags, "HSDir"); + is_exit = is_guard = 0; + if (! is_bad_exit && badexit_flag_is_listed) { + is_bad_exit = 1; + smartlist_add(chosen_flags, (char *)"BadExit"); + smartlist_sort_strings(chosen_flags); // restore order. + } + } + /* Pick the version. */ if (smartlist_len(versions)) { sort_version_list(versions, 0); @@ -2253,6 +2288,8 @@ networkstatus_compute_consensus(smartlist_t *votes, smartlist_add_asprintf(chunks, "m %s\n", m); } /* Next line is all flags. The "\n" is missing. */ + smartlist_add_asprintf(chunks, "s%s", + smartlist_len(chosen_flags)?" ":""); smartlist_add(chunks, smartlist_join_strings(chosen_flags, " ", 0, NULL)); /* Now the version line. */ @@ -2265,7 +2302,8 @@ networkstatus_compute_consensus(smartlist_t *votes, smartlist_add_asprintf(chunks, "pr %s\n", chosen_protocol_list); } /* Now the weight line. */ - if (rs_out.has_bandwidth) { + if (rs_out.has_bandwidth && (!rs_out.is_authority || + !dirauth_get_options()->AuthDirDontVoteOnDirAuthBandwidth)) { char *guardfraction_str = NULL; int unmeasured = rs_out.bw_is_unmeasured; @@ -4581,6 +4619,7 @@ const char DIRVOTE_UNIVERSAL_FLAGS[] = * depending on our configuration. */ const char DIRVOTE_OPTIONAL_FLAGS[] = "BadExit " + "MiddleOnly " "Running"; /** Return a new networkstatus_t* containing our current opinion. (For v3 @@ -4598,7 +4637,8 @@ dirserv_generate_networkstatus_vote_obj(crypto_pk_t *private_key, smartlist_t *routers, *routerstatuses; char identity_digest[DIGEST_LEN]; char signing_key_digest[DIGEST_LEN]; - const int listbadexits = d_options->AuthDirListBadExits; + const int list_bad_exits = d_options->AuthDirListBadExits; + const int list_middle_only = d_options->AuthDirListMiddleOnly; routerlist_t *rl = router_get_routerlist(); time_t now = time(NULL); time_t cutoff = now - ROUTER_MAX_AGE_TO_PUBLISH; @@ -4703,7 +4743,8 @@ dirserv_generate_networkstatus_vote_obj(crypto_pk_t *private_key, vrs = tor_malloc_zero(sizeof(vote_routerstatus_t)); rs = &vrs->status; dirauth_set_routerstatus_from_routerinfo(rs, node, ri, now, - listbadexits); + list_bad_exits, + list_middle_only); if (ri->cache_info.signing_key_cert) { memcpy(vrs->ed25519_id, @@ -4825,8 +4866,10 @@ dirserv_generate_networkstatus_vote_obj(crypto_pk_t *private_key, 0, SPLIT_SKIP_SPACE|SPLIT_IGNORE_BLANK, 0); if (vote_on_reachability) smartlist_add_strdup(v3_out->known_flags, "Running"); - if (listbadexits) + if (list_bad_exits) smartlist_add_strdup(v3_out->known_flags, "BadExit"); + if (list_middle_only) + smartlist_add_strdup(v3_out->known_flags, "MiddleOnly"); smartlist_sort_strings(v3_out->known_flags); if (d_options->ConsensusParams) { diff --git a/src/feature/dirauth/dirvote.h b/src/feature/dirauth/dirvote.h index 3420098315..64aaec116e 100644 --- a/src/feature/dirauth/dirvote.h +++ b/src/feature/dirauth/dirvote.h @@ -53,7 +53,7 @@ #define MIN_SUPPORTED_CONSENSUS_METHOD 28 /** The highest consensus method that we currently support. */ -#define MAX_SUPPORTED_CONSENSUS_METHOD 31 +#define MAX_SUPPORTED_CONSENSUS_METHOD 32 /** * Lowest consensus method where microdescriptor lines are put in canonical @@ -70,6 +70,10 @@ */ #define MIN_METHOD_FOR_CORRECT_BWWEIGHTSCALE 31 +/** Lowest consensus method for which we handle the MiddleOnly flag specially. + */ +#define MIN_METHOD_FOR_MIDDLEONLY 32 + /** Default bandwidth to clip unmeasured bandwidths to using method >= * MIN_METHOD_TO_CLIP_UNMEASURED_BW. (This is not a consensus method; do not * get confused with the above macros.) */ diff --git a/src/feature/dirauth/process_descs.c b/src/feature/dirauth/process_descs.c index eca987b8b5..a75f516dca 100644 --- a/src/feature/dirauth/process_descs.c +++ b/src/feature/dirauth/process_descs.c @@ -226,6 +226,8 @@ dirserv_load_fingerprint_file(void) add_status = RTR_BADEXIT; } else if (!strcasecmp(nickname, "!invalid")) { add_status = RTR_INVALID; + } else if (!strcasecmp(nickname, "!middleonly")) { + add_status = RTR_MIDDLEONLY; } /* Check if fingerprint is RSA or ed25519 by verifying it. */ @@ -412,11 +414,11 @@ dirserv_rejects_tor_version(const char *platform, return true; } - /* Series between Tor 0.3.6 and 0.4.1 inclusive are unsupported. Reject - * them. 0.3.6.0-alpha-dev only existed for a short time, before it was - * renamed to 0.4.0.0-alpha-dev. */ + /* Series between Tor 0.3.6.x and 0.4.5.5-rc inclusive are unsupported. + * Reject them. 0.3.6.0-alpha-dev only existed for a short time, before it + * was renamed to 0.4.0.0-alpha-dev. */ if (tor_version_as_new_as(platform,"0.3.6.0-alpha-dev") && - !tor_version_as_new_as(platform,"0.4.2.1-alpha")) { + !tor_version_as_new_as(platform,"0.4.5.6")) { if (msg) { *msg = please_upgrade_string; } @@ -496,6 +498,13 @@ dirserv_get_status_impl(const char *id_digest, result |= RTR_BADEXIT; } + if (authdir_policy_middleonly_address(ipv4_addr, ipv4_orport)) { + log_fn(severity, LD_DIRSERV, + "Marking '%s' as middle-only because of address '%s'", + nickname, fmt_addr(ipv4_addr)); + result |= RTR_MIDDLEONLY; + } + if (!authdir_policy_permits_address(ipv4_addr, ipv4_orport)) { log_fn(severity, LD_DIRSERV, "Rejecting '%s' because of address '%s'", nickname, fmt_addr(ipv4_addr)); @@ -630,6 +639,7 @@ dirserv_set_node_flags_from_authoritative_status(node_t *node, { node->is_valid = (authstatus & RTR_INVALID) ? 0 : 1; node->is_bad_exit = (authstatus & RTR_BADEXIT) ? 1 : 0; + node->is_middle_only = (authstatus & RTR_MIDDLEONLY) ? 1 : 0; } /** True iff <b>a</b> is more severe than <b>b</b>. */ @@ -963,6 +973,11 @@ directory_remove_invalid(void) (r & RTR_BADEXIT) ? "bad" : "good"); node->is_bad_exit = (r&RTR_BADEXIT) ? 1: 0; } + if (bool_neq((r & RTR_MIDDLEONLY), node->is_middle_only)) { + log_info(LD_DIRSERV, "Router '%s' is now %smiddle-only", description, + (r & RTR_MIDDLEONLY) ? "" : "not"); + node->is_middle_only = (r&RTR_MIDDLEONLY) ? 1: 0; + } } SMARTLIST_FOREACH_END(node); routerlist_assert_ok(rl); diff --git a/src/feature/dirauth/process_descs.h b/src/feature/dirauth/process_descs.h index 6c056d11dd..a509eb1fbe 100644 --- a/src/feature/dirauth/process_descs.h +++ b/src/feature/dirauth/process_descs.h @@ -45,7 +45,8 @@ typedef struct authdir_config_t { #define RTR_REJECT 4 /**< We will not publish this router. */ /* 8 Historically used to avoid using this as a dir. */ #define RTR_BADEXIT 16 /**< We'll tell clients not to use this as an exit. */ -/* 32 Historically used to indicade Unnamed */ +/** We'll vote to only use this router as a midpoint. */ +#define RTR_MIDDLEONLY 32 #endif /* defined(PROCESS_DESCS_PRIVATE) || defined(TOR_UNIT_TESTS) */ diff --git a/src/feature/dirauth/voteflags.c b/src/feature/dirauth/voteflags.c index d755a270be..05c19ff501 100644 --- a/src/feature/dirauth/voteflags.c +++ b/src/feature/dirauth/voteflags.c @@ -565,7 +565,8 @@ dirauth_set_routerstatus_from_routerinfo(routerstatus_t *rs, node_t *node, const routerinfo_t *ri, time_t now, - int listbadexits) + int listbadexits, + int listmiddleonly) { const or_options_t *options = get_options(); uint32_t routerbw_kb = dirserv_get_credible_bandwidth_kb(ri); @@ -597,6 +598,14 @@ dirauth_set_routerstatus_from_routerinfo(routerstatus_t *rs, /* Override rs->is_bad_exit */ rs->is_bad_exit = listbadexits && node->is_bad_exit; + /* Override rs->is_middle_only and related flags. */ + rs->is_middle_only = listmiddleonly && node->is_middle_only; + if (rs->is_middle_only) { + if (listbadexits) + rs->is_bad_exit = 1; + rs->is_exit = rs->is_possible_guard = rs->is_hs_dir = rs->is_v2_dir = 0; + } + /* Set rs->is_staledesc. */ rs->is_staledesc = (ri->cache_info.published_on + DESC_IS_STALE_INTERVAL) < now; diff --git a/src/feature/dirauth/voteflags.h b/src/feature/dirauth/voteflags.h index 818a0bafd2..8371f1c315 100644 --- a/src/feature/dirauth/voteflags.h +++ b/src/feature/dirauth/voteflags.h @@ -22,7 +22,8 @@ void dirauth_set_routerstatus_from_routerinfo(routerstatus_t *rs, node_t *node, const routerinfo_t *ri, time_t now, - int listbadexits); + int listbadexits, + int listmiddleonly); void dirserv_compute_performance_thresholds(digestmap_t *omit_as_sybil); #endif /* defined(HAVE_MODULE_DIRAUTH) */ diff --git a/src/feature/dircache/dircache.c b/src/feature/dircache/dircache.c index 7fdb1bc70f..7319b96caf 100644 --- a/src/feature/dircache/dircache.c +++ b/src/feature/dircache/dircache.c @@ -1569,6 +1569,8 @@ directory_handle_command_post,(dir_connection_t *conn, const char *headers, char *url = NULL; const or_options_t *options = get_options(); + (void) body_len; + log_debug(LD_DIRSERV,"Received POST command."); conn->base_.state = DIR_CONN_STATE_SERVER_WRITING; diff --git a/src/feature/dirclient/dir_server_st.h b/src/feature/dirclient/dir_server_st.h index ed6b00647e..ac45f3787b 100644 --- a/src/feature/dirclient/dir_server_st.h +++ b/src/feature/dirclient/dir_server_st.h @@ -16,6 +16,8 @@ #include "core/or/or.h" #include "feature/nodelist/routerstatus_st.h" +struct smartlist_t; + /** Represents information about a single trusted or fallback directory * server. */ struct dir_server_t { @@ -48,6 +50,10 @@ struct dir_server_t { time_t addr_current_at; /**< When was the document that we derived the * address information from published? */ + /** Authority only. Can be null. If present, a list of auth_dirport_t + * representing HTTP dirports for this authority. */ + struct smartlist_t *auth_dirports; + routerstatus_t fake_status; /**< Used when we need to pass this trusted * dir_server_t to * directory_request_set_routerstatus. diff --git a/src/feature/dirclient/dirclient.c b/src/feature/dirclient/dirclient.c index 0b6a8101a5..4e9c8e2f45 100644 --- a/src/feature/dirclient/dirclient.c +++ b/src/feature/dirclient/dirclient.c @@ -1134,6 +1134,7 @@ directory_request_set_routerstatus(directory_request_t *req, { req->routerstatus = status; } + /** * Helper: update the addresses, ports, and identities in <b>req</b> * from the routerstatus object in <b>req</b>. Return 0 on success. @@ -1176,7 +1177,7 @@ directory_request_set_dir_from_routerstatus(directory_request_t *req) return -1; } - /* At this point, if we are a client making a direct connection to a + /* At this point, if we are a client making a direct connection to a * directory server, we have selected a server that has at least one address * allowed by ClientUseIPv4/6 and Reachable{"",OR,Dir}Addresses. This * selection uses the preference in ClientPreferIPv6{OR,Dir}Port, if @@ -1191,6 +1192,37 @@ directory_request_set_dir_from_routerstatus(directory_request_t *req) return -1; } + /* One last thing: If we're talking to an authority, we might want to use + * a special HTTP port for it based on our purpose. + */ + if (req->indirection == DIRIND_DIRECT_CONN && status->is_authority) { + const dir_server_t *ds = router_get_trusteddirserver_by_digest( + status->identity_digest); + if (ds) { + const tor_addr_port_t *v4 = NULL; + if (authdir_mode_v3(get_options())) { + // An authority connecting to another authority should always + // prefer the VOTING usage, if one is specifically configured. + v4 = trusted_dir_server_get_dirport_exact( + ds, AUTH_USAGE_VOTING, AF_INET); + } + if (! v4) { + // Everybody else should prefer a usage dependent on their + // the dir_purpose. + auth_dirport_usage_t usage = + auth_dirport_usage_for_purpose(req->dir_purpose); + v4 = trusted_dir_server_get_dirport(ds, usage, AF_INET); + } + tor_assert_nonfatal(v4); + if (v4) { + // XXXX We could, if we wanted, also select a v6 address. But a v4 + // address must exist here, and we as a relay are required to support + // ipv4. So we just that. + tor_addr_port_copy(&use_dir_ap, v4); + } + } + } + directory_request_set_or_addr_port(req, &use_or_ap); directory_request_set_dir_addr_port(req, &use_dir_ap); directory_request_set_directory_id_digest(req, status->identity_digest); @@ -1209,7 +1241,7 @@ directory_initiate_request,(directory_request_t *request)) tor_assert_nonfatal( ! directory_request_dir_contact_info_specified(request)); if (directory_request_set_dir_from_routerstatus(request) < 0) { - return; + return; // or here XXXX } } @@ -1324,6 +1356,8 @@ directory_initiate_request,(directory_request_t *request)) entry_guard_cancel(&guard_state); } + // XXXX This is the case where we replace. + switch (connection_connect(TO_CONN(conn), conn->base_.address, &addr, port, &socket_error)) { case -1: diff --git a/src/feature/dirclient/dlstatus.c b/src/feature/dirclient/dlstatus.c index 8be2983a5d..c21dd113b4 100644 --- a/src/feature/dirclient/dlstatus.c +++ b/src/feature/dirclient/dlstatus.c @@ -73,15 +73,14 @@ find_dl_min_delay(const download_status_t *dls, const or_options_t *options) } } case DL_SCHED_BRIDGE: - if (options->UseBridges && num_bridges_usable(0) > 0) { - /* A bridge client that is sure that one or more of its bridges are - * running can afford to wait longer to update bridge descriptors. */ - return options->TestingBridgeDownloadInitialDelay; - } else { - /* A bridge client which might have no running bridges, must try to - * get bridge descriptors straight away. */ - return options->TestingBridgeBootstrapDownloadInitialDelay; - } + /* Be conservative here: always return the 'during bootstrap' delay + * value, so we never delay while trying to fetch descriptors + * for new bridges. Once we do succeed at fetching a descriptor + * for our bridge, we will adjust its next_attempt_at based on + * the longer "TestingBridgeDownloadInitialDelay" value. See + * learned_bridge_descriptor() for details. + */ + return options->TestingBridgeBootstrapDownloadInitialDelay; default: tor_assert(0); } diff --git a/src/feature/dircommon/consdiff.c b/src/feature/dircommon/consdiff.c index c877227adc..323f2bd576 100644 --- a/src/feature/dircommon/consdiff.c +++ b/src/feature/dircommon/consdiff.c @@ -1128,7 +1128,7 @@ consdiff_get_digests(const smartlist_t *diff, { const cdline_t *line2 = smartlist_get(diff, 1); char *h = tor_memdup_nulterm(line2->s, line2->len); - smartlist_split_string(hash_words, h, " ", 0, 0); + smartlist_split_string(hash_words, h, " ", 0, 4); tor_free(h); } diff --git a/src/feature/dirparse/ns_parse.c b/src/feature/dirparse/ns_parse.c index 947b3810a4..cd3e2731be 100644 --- a/src/feature/dirparse/ns_parse.c +++ b/src/feature/dirparse/ns_parse.c @@ -434,6 +434,8 @@ routerstatus_parse_entry_from_string(memarea_t *area, rs->is_possible_guard = 1; else if (!strcmp(tok->args[i], "BadExit")) rs->is_bad_exit = 1; + else if (!strcmp(tok->args[i], "MiddleOnly")) + rs->is_middle_only = 1; else if (!strcmp(tok->args[i], "Authority")) rs->is_authority = 1; else if (!strcmp(tok->args[i], "Unnamed") && diff --git a/src/feature/hs/hs_config.c b/src/feature/hs/hs_config.c index 73f9176186..a76893fe1a 100644 --- a/src/feature/hs/hs_config.c +++ b/src/feature/hs/hs_config.c @@ -548,15 +548,19 @@ config_service(config_line_t *line, const or_options_t *options, tor_assert(service->config.version <= HS_VERSION_MAX); - /* Check permission on service directory that was just parsed. And this must - * be done regardless of the service version. Do not ask for the directory - * to be created, this is done when the keys are loaded because we could be - * in validation mode right now. */ - if (hs_check_service_private_dir(options->User, - service->config.directory_path, - service->config.dir_group_readable, - 0) < 0) { - goto err; + /* If we're running with TestingTorNetwork enabled, we relax the permissions + * check on the hs directory. */ + if (!options->TestingTorNetwork) { + /* Check permission on service directory that was just parsed. And this + * must be done regardless of the service version. Do not ask for the + * directory to be created, this is done when the keys are loaded because + * we could be in validation mode right now. */ + if (hs_check_service_private_dir(options->User, + service->config.directory_path, + service->config.dir_group_readable, + 0) < 0) { + goto err; + } } /* We'll try to learn the service version here by loading the key(s) if @@ -640,6 +644,7 @@ hs_config_service_all(const or_options_t *options, int validate_only) int rv = config_service(section, options, new_service_list); config_free_lines(section); if (rv < 0) { + config_free_lines(remaining); goto err; } } diff --git a/src/feature/hs/hs_descriptor.c b/src/feature/hs/hs_descriptor.c index 70ff4e9690..a37eab5b5d 100644 --- a/src/feature/hs/hs_descriptor.c +++ b/src/feature/hs/hs_descriptor.c @@ -1607,8 +1607,8 @@ decrypt_desc_layer,(const hs_descriptor_t *desc, * put in decrypted_out which contains the superencrypted layer of the * descriptor. Return the length of decrypted_out on success else 0 is * returned and decrypted_out is set to NULL. */ -static size_t -desc_decrypt_superencrypted(const hs_descriptor_t *desc, char **decrypted_out) +MOCK_IMPL(STATIC size_t, +desc_decrypt_superencrypted,(const hs_descriptor_t *desc,char **decrypted_out)) { size_t superencrypted_len = 0; char *superencrypted_plaintext = NULL; @@ -1639,10 +1639,10 @@ desc_decrypt_superencrypted(const hs_descriptor_t *desc, char **decrypted_out) * decrypted_out which contains the encrypted layer of the descriptor. * Return the length of decrypted_out on success else 0 is returned and * decrypted_out is set to NULL. */ -static size_t -desc_decrypt_encrypted(const hs_descriptor_t *desc, - const curve25519_secret_key_t *client_auth_sk, - char **decrypted_out) +MOCK_IMPL(STATIC size_t, +desc_decrypt_encrypted,(const hs_descriptor_t *desc, + const curve25519_secret_key_t *client_auth_sk, + char **decrypted_out)) { size_t encrypted_len = 0; char *encrypted_plaintext = NULL; @@ -2145,7 +2145,7 @@ desc_decode_plaintext_v3(smartlist_t *tokens, /** Decode the version 3 superencrypted section of the given descriptor desc. * The desc_superencrypted_out will be populated with the decoded data. */ -static hs_desc_decode_status_t +STATIC hs_desc_decode_status_t desc_decode_superencrypted_v3(const hs_descriptor_t *desc, hs_desc_superencrypted_data_t * desc_superencrypted_out) @@ -2259,7 +2259,7 @@ desc_decode_superencrypted_v3(const hs_descriptor_t *desc, /** Decode the version 3 encrypted section of the given descriptor desc. The * desc_encrypted_out will be populated with the decoded data. */ -static hs_desc_decode_status_t +STATIC hs_desc_decode_status_t desc_decode_encrypted_v3(const hs_descriptor_t *desc, const curve25519_secret_key_t *client_auth_sk, hs_desc_encrypted_data_t *desc_encrypted_out) diff --git a/src/feature/hs/hs_descriptor.h b/src/feature/hs/hs_descriptor.h index 7e437faeb8..d959431369 100644 --- a/src/feature/hs/hs_descriptor.h +++ b/src/feature/hs/hs_descriptor.h @@ -339,6 +339,25 @@ MOCK_DECL(STATIC size_t, decrypt_desc_layer,(const hs_descriptor_t *desc, bool is_superencrypted_layer, char **decrypted_out)); +STATIC hs_desc_decode_status_t desc_decode_encrypted_v3( + const hs_descriptor_t *desc, + const curve25519_secret_key_t *client_auth_sk, + hs_desc_encrypted_data_t *desc_encrypted_out); + +STATIC hs_desc_decode_status_t +desc_decode_superencrypted_v3(const hs_descriptor_t *desc, + hs_desc_superencrypted_data_t * + desc_superencrypted_out); + +MOCK_DECL(STATIC size_t, desc_decrypt_encrypted,( + const hs_descriptor_t *desc, + const curve25519_secret_key_t *client_auth_sk, + char **decrypted_out)); + +MOCK_DECL(STATIC size_t, desc_decrypt_superencrypted,( + const hs_descriptor_t *desc, + char **decrypted_out)); + #endif /* defined(HS_DESCRIPTOR_PRIVATE) */ #endif /* !defined(TOR_HS_DESCRIPTOR_H) */ diff --git a/src/feature/hs/hs_metrics.c b/src/feature/hs/hs_metrics.c index e023eab90c..0f1824c51c 100644 --- a/src/feature/hs/hs_metrics.c +++ b/src/feature/hs/hs_metrics.c @@ -29,18 +29,6 @@ port_to_str(const uint16_t port) return buf; } -/** Return a static buffer pointer that contains a formatted label on the form - * of key=value. - * - * Subsequent call to this function invalidates the previous buffer. */ -static const char * -format_label(const char *key, const char *value) -{ - static char buf[128]; - tor_snprintf(buf, sizeof(buf), "%s=%s", key, value); - return buf; -} - /** Initialize a metrics store for the given service. * * Essentially, this goes over the base_metrics array and adds them all to the @@ -61,12 +49,12 @@ init_store(hs_service_t *service) /* Add labels to the entry. */ metrics_store_entry_add_label(entry, - format_label("onion", service->onion_address)); + metrics_format_label("onion", service->onion_address)); if (base_metrics[i].port_as_label && service->config.ports) { SMARTLIST_FOREACH_BEGIN(service->config.ports, const hs_port_config_t *, p) { metrics_store_entry_add_label(entry, - format_label("port", port_to_str(p->virtual_port))); + metrics_format_label("port", port_to_str(p->virtual_port))); } SMARTLIST_FOREACH_END(p); } } @@ -96,7 +84,7 @@ hs_metrics_update_by_service(const hs_metrics_key_t key, SMARTLIST_FOREACH_BEGIN(entries, metrics_store_entry_t *, entry) { if (port == 0 || metrics_store_entry_has_label(entry, - format_label("port", port_to_str(port)))) { + metrics_format_label("port", port_to_str(port)))) { metrics_store_entry_update(entry, n); break; } diff --git a/src/feature/nodelist/dirlist.c b/src/feature/nodelist/dirlist.c index 1f18bd71a2..1f1ac4d106 100644 --- a/src/feature/nodelist/dirlist.c +++ b/src/feature/nodelist/dirlist.c @@ -43,6 +43,14 @@ #include "feature/dirclient/dir_server_st.h" #include "feature/nodelist/node_st.h" +/** Information about an (HTTP) dirport for a directory authority. */ +struct auth_dirport_t { + /** What is the intended usage for this dirport? One of AUTH_USAGE_* */ + auth_dirport_usage_t usage; + /** What is the correct address/port ? */ + tor_addr_port_t dirport; +}; + /** Global list of a dir_server_t object for each directory * authority. */ static smartlist_t *trusted_dir_servers = NULL; @@ -66,6 +74,11 @@ add_trusted_dir_to_nodelist_addr_set(const dir_server_t *dir) /* IPv6 DirPort is not a thing yet for authorities. */ nodelist_add_addr_to_address_set(&dir->ipv6_addr, dir->ipv6_orport, 0); } + if (dir->auth_dirports) { + SMARTLIST_FOREACH_BEGIN(dir->auth_dirports, const auth_dirport_t *, p) { + nodelist_add_addr_to_address_set(&p->dirport.addr, 0, p->dirport.port); + } SMARTLIST_FOREACH_END(p); + } } /** Go over the trusted directory server list and add their address(es) to the @@ -256,7 +269,10 @@ MOCK_IMPL(int, router_digest_is_trusted_dir_type, /** Return true iff the given address matches a trusted directory that matches * at least one bit of type. * - * If type is NO_DIRINFO or ALL_DIRINFO, any authority is matched. */ + * If type is NO_DIRINFO or ALL_DIRINFO, any authority is matched. + * + * Only ORPorts' addresses are considered. + */ bool router_addr_is_trusted_dir_type(const tor_addr_t *addr, dirinfo_type_t type) { @@ -281,6 +297,39 @@ router_addr_is_trusted_dir_type(const tor_addr_t *addr, dirinfo_type_t type) return false; } +/** Return an appropriate usage value describing which authdir port to use + * for a given directory connection purpose. + */ +auth_dirport_usage_t +auth_dirport_usage_for_purpose(int purpose) +{ + switch (purpose) { + case DIR_PURPOSE_FETCH_SERVERDESC: + case DIR_PURPOSE_FETCH_EXTRAINFO: + case DIR_PURPOSE_FETCH_CONSENSUS: + case DIR_PURPOSE_FETCH_CERTIFICATE: + case DIR_PURPOSE_FETCH_MICRODESC: + return AUTH_USAGE_DOWNLOAD; + + case DIR_PURPOSE_UPLOAD_DIR: + return AUTH_USAGE_UPLOAD; + + case DIR_PURPOSE_UPLOAD_VOTE: + case DIR_PURPOSE_UPLOAD_SIGNATURES: + case DIR_PURPOSE_FETCH_DETACHED_SIGNATURES: + case DIR_PURPOSE_FETCH_STATUS_VOTE: + return AUTH_USAGE_VOTING; + + case DIR_PURPOSE_SERVER: + case DIR_PURPOSE_UPLOAD_HSDESC: + case DIR_PURPOSE_FETCH_HSDESC: + case DIR_PURPOSE_HAS_FETCHED_HSDESC: + default: + tor_assert_nonfatal_unreached(); + return AUTH_USAGE_LEGACY; + } +} + /** Create a directory server at <b>address</b>:<b>port</b>, with OR identity * key <b>digest</b> which has DIGEST_LEN bytes. If <b>address</b> is NULL, * add ourself. If <b>is_authority</b>, this is a directory authority. Return @@ -357,6 +406,7 @@ dir_server_new(int is_authority, ent->fake_status.ipv4_dirport = ent->ipv4_dirport; ent->fake_status.ipv4_orport = ent->ipv4_orport; ent->fake_status.ipv6_orport = ent->ipv6_orport; + ent->fake_status.is_authority = !! is_authority; return ent; } @@ -404,10 +454,98 @@ trusted_dir_server_new(const char *nickname, const char *address, ipv6_addrport, digest, v3_auth_digest, type, weight); + + if (ipv4_dirport) { + tor_addr_port_t p; + memset(&p, 0, sizeof(p)); + tor_addr_copy(&p.addr, &ipv4_addr); + p.port = ipv4_dirport; + trusted_dir_server_add_dirport(result, AUTH_USAGE_LEGACY, &p); + } tor_free(hostname); return result; } +/** + * Add @a dirport as an HTTP DirPort contact point for the directory authority + * @a ds, for use when contacting that authority for the given @a usage. + * + * Multiple ports of the same usage are allowed; if present, then only + * the first one of each address family is currently used. + */ +void +trusted_dir_server_add_dirport(dir_server_t *ds, + auth_dirport_usage_t usage, + const tor_addr_port_t *dirport) +{ + tor_assert(ds); + tor_assert(dirport); + + if (BUG(! ds->is_authority)) { + return; + } + + if (ds->auth_dirports == NULL) { + ds->auth_dirports = smartlist_new(); + } + + auth_dirport_t *port = tor_malloc_zero(sizeof(auth_dirport_t)); + port->usage = usage; + tor_addr_port_copy(&port->dirport, dirport); + smartlist_add(ds->auth_dirports, port); +} + +/** + * Helper for trusted_dir_server_get_dirport: only return the exact requested + * usage type. + */ +const tor_addr_port_t * +trusted_dir_server_get_dirport_exact(const dir_server_t *ds, + auth_dirport_usage_t usage, + int addr_family) +{ + tor_assert(ds); + tor_assert_nonfatal(addr_family == AF_INET || addr_family == AF_INET6); + if (ds->auth_dirports == NULL) + return NULL; + + SMARTLIST_FOREACH_BEGIN(ds->auth_dirports, const auth_dirport_t *, port) { + if (port->usage == usage && + tor_addr_family(&port->dirport.addr) == addr_family) { + return &port->dirport; + } + } SMARTLIST_FOREACH_END(port); + + return NULL; +} + +/** + * Return the DirPort of the authority @a ds for with the usage type + * @a usage and address family @a addr_family. If none is found, try + * again with an AUTH_USAGE_LEGACY dirport, if there is one. Return NULL + * if no port can be found. + */ +const tor_addr_port_t * +trusted_dir_server_get_dirport(const dir_server_t *ds, + auth_dirport_usage_t usage, + int addr_family) +{ + const tor_addr_port_t *port; + + while (1) { + port = trusted_dir_server_get_dirport_exact(ds, usage, addr_family); + if (port) + return port; + + // If we tried LEGACY, there is no fallback from this point. + if (usage == AUTH_USAGE_LEGACY) + return NULL; + + // Try again with LEGACY. + usage = AUTH_USAGE_LEGACY; + } +} + /** Return a new dir_server_t for a fallback directory server at * <b>addr</b>:<b>or_port</b>/<b>dir_port</b>, with identity key digest * <b>id_digest</b> */ @@ -447,6 +585,10 @@ dir_server_free_(dir_server_t *ds) if (!ds) return; + if (ds->auth_dirports) { + SMARTLIST_FOREACH(ds->auth_dirports, auth_dirport_t *, p, tor_free(p)); + smartlist_free(ds->auth_dirports); + } tor_free(ds->nickname); tor_free(ds->description); tor_free(ds->address); diff --git a/src/feature/nodelist/dirlist.h b/src/feature/nodelist/dirlist.h index f744fecf92..3b4faf07af 100644 --- a/src/feature/nodelist/dirlist.h +++ b/src/feature/nodelist/dirlist.h @@ -11,6 +11,28 @@ #ifndef TOR_DIRLIST_H #define TOR_DIRLIST_H +typedef struct auth_dirport_t auth_dirport_t; +/** + * Different usages for an authority's HTTP directory port. + * + * Historically, only legacy ports existed; proposal 330 added multiple types + * of dirport to better enable authorities to offload work and resist DoS + * attacks. + **/ +typedef enum auth_dirport_usage_t { + /** Flag for an authority's dirport that is intended for misc/legacy + * usage. May be used when no other dirport is available. */ + AUTH_USAGE_LEGACY, + /** Flag for an authority's dirport that is intended for descriptor uploads + * only. */ + AUTH_USAGE_UPLOAD, + /** Flag for an authority's dirport that is intended for voting only */ + AUTH_USAGE_VOTING, + /** Flag for an authority's dirport that is intended for relay downloads + * only. */ + AUTH_USAGE_DOWNLOAD, +} auth_dirport_usage_t; + int get_n_authorities(dirinfo_type_t type); const smartlist_t *router_get_trusted_dir_servers(void); const smartlist_t *router_get_fallback_dir_servers(void); @@ -18,6 +40,8 @@ smartlist_t *router_get_trusted_dir_servers_mutable(void); smartlist_t *router_get_fallback_dir_servers_mutable(void); void mark_all_dirservers_up(smartlist_t *server_list); +auth_dirport_usage_t auth_dirport_usage_for_purpose(int purpose); + dir_server_t *router_get_trusteddirserver_by_digest(const char *d); dir_server_t *router_get_fallback_dirserver_by_digest( const char *digest); @@ -28,6 +52,14 @@ MOCK_DECL(dir_server_t *, trusteddirserver_get_by_v3_auth_digest, MOCK_DECL(int, router_digest_is_trusted_dir_type, (const char *digest, dirinfo_type_t type)); +const tor_addr_port_t *trusted_dir_server_get_dirport(const dir_server_t *ds, + auth_dirport_usage_t usage, + int addr_family); +const tor_addr_port_t *trusted_dir_server_get_dirport_exact( + const dir_server_t *ds, + auth_dirport_usage_t usage, + int addr_family); + bool router_addr_is_trusted_dir_type(const tor_addr_t *addr, dirinfo_type_t type); #define router_addr_is_trusted_dir(d) \ @@ -41,6 +73,9 @@ dir_server_t *trusted_dir_server_new(const char *nickname, const char *address, const tor_addr_port_t *addrport_ipv6, const char *digest, const char *v3_auth_digest, dirinfo_type_t type, double weight); +void trusted_dir_server_add_dirport(dir_server_t *ds, + auth_dirport_usage_t usage, + const tor_addr_port_t *dirport); dir_server_t *fallback_dir_server_new(const tor_addr_t *addr, uint16_t dir_port, uint16_t or_port, const tor_addr_port_t *addrport_ipv6, diff --git a/src/feature/nodelist/fmt_routerstatus.c b/src/feature/nodelist/fmt_routerstatus.c index 6db40c0b68..95379a7721 100644 --- a/src/feature/nodelist/fmt_routerstatus.c +++ b/src/feature/nodelist/fmt_routerstatus.c @@ -87,7 +87,7 @@ routerstatus_format_entry(const routerstatus_t *rs, const char *version, goto done; smartlist_add_asprintf(chunks, - "s%s%s%s%s%s%s%s%s%s%s%s%s\n", + "s%s%s%s%s%s%s%s%s%s%s%s%s%s\n", /* These must stay in alphabetical order. */ rs->is_authority?" Authority":"", rs->is_bad_exit?" BadExit":"", @@ -95,6 +95,7 @@ routerstatus_format_entry(const routerstatus_t *rs, const char *version, rs->is_fast?" Fast":"", rs->is_possible_guard?" Guard":"", rs->is_hs_dir?" HSDir":"", + rs->is_middle_only?" MiddleOnly":"", rs->is_flagged_running?" Running":"", rs->is_stable?" Stable":"", rs->is_staledesc?" StaleDesc":"", diff --git a/src/feature/nodelist/networkstatus.c b/src/feature/nodelist/networkstatus.c index af808a6ba7..77e2b547f5 100644 --- a/src/feature/nodelist/networkstatus.c +++ b/src/feature/nodelist/networkstatus.c @@ -45,6 +45,8 @@ #include "core/or/channel.h" #include "core/or/channelpadding.h" #include "core/or/circuitpadding.h" +#include "core/or/congestion_control_common.h" +#include "core/or/congestion_control_flow.h" #include "core/or/circuitmux.h" #include "core/or/circuitmux_ewma.h" #include "core/or/circuitstats.h" @@ -1664,7 +1666,6 @@ notify_before_networkstatus_changes(const networkstatus_t *old_c, dos_consensus_has_changed(new_c); relay_consensus_has_changed(new_c); hs_dos_consensus_has_changed(new_c); - rep_hist_consensus_has_changed(new_c); } /* Called after a new consensus has been put in the global state. It is safe @@ -1701,6 +1702,11 @@ notify_after_networkstatus_changes(void) channelpadding_new_consensus_params(c); circpad_new_consensus_params(c); router_new_consensus_params(c); + congestion_control_new_consensus_params(c); + flow_control_new_consensus_params(c); + + /* Maintenance of our L2 guard list */ + maintain_layer2_guards(); } /** Copy all the ancillary information (like router download status and so on) diff --git a/src/feature/nodelist/node_st.h b/src/feature/nodelist/node_st.h index b15e7154c4..df67a47ada 100644 --- a/src/feature/nodelist/node_st.h +++ b/src/feature/nodelist/node_st.h @@ -70,6 +70,8 @@ struct node_t { unsigned int is_exit:1; /**< Do we think this is an OK exit? */ unsigned int is_bad_exit:1; /**< Do we think this exit is censored, borked, * or otherwise nasty? */ + /** Is this unsuitable for use as anything besides a middle relay? */ + unsigned int is_middle_only:1; unsigned int is_hs_dir:1; /**< True iff this router is a hidden service * directory according to the authorities. */ diff --git a/src/feature/nodelist/nodelist.c b/src/feature/nodelist/nodelist.c index 121dc8823a..c676e8dfb4 100644 --- a/src/feature/nodelist/nodelist.c +++ b/src/feature/nodelist/nodelist.c @@ -2820,6 +2820,7 @@ update_router_have_minimum_dir_info(void) const networkstatus_t *consensus = networkstatus_get_reasonably_live_consensus(now,usable_consensus_flavor()); int using_md; + static int be_loud_when_things_work_again = 0; if (!consensus) { if (!networkstatus_get_latest_consensus()) @@ -2875,8 +2876,9 @@ update_router_have_minimum_dir_info(void) if (res && !have_min_dir_info) { control_event_client_status(LOG_NOTICE, "ENOUGH_DIR_INFO"); control_event_boot_dir(BOOTSTRAP_STATUS_ENOUGH_DIRINFO, 0); - log_info(LD_DIR, - "We now have enough directory information to build circuits."); + tor_log(be_loud_when_things_work_again ? LOG_NOTICE : LOG_INFO, LD_DIR, + "We now have enough directory information to build circuits."); + be_loud_when_things_work_again = 0; } /* If paths have just become unavailable in this update. */ @@ -2885,6 +2887,10 @@ update_router_have_minimum_dir_info(void) tor_log(quiet ? LOG_INFO : LOG_NOTICE, LD_DIR, "Our directory information is no longer up-to-date " "enough to build circuits: %s", dir_info_status); + if (!quiet) { + /* remember to do a notice-level log when things come back */ + be_loud_when_things_work_again = 1; + } /* a) make us log when we next complete a circuit, so we know when Tor * is back up and usable, and b) disable some activities that Tor diff --git a/src/feature/nodelist/routerlist.c b/src/feature/nodelist/routerlist.c index 565d4596d4..c00f7ffb26 100644 --- a/src/feature/nodelist/routerlist.c +++ b/src/feature/nodelist/routerlist.c @@ -1617,6 +1617,13 @@ router_add_to_routerlist(routerinfo_t *router, const char **msg, "descriptor for router %s", router_describe(router)); } else { + if (router->purpose == ROUTER_PURPOSE_BRIDGE) { + /* Even if we're not going to keep this descriptor, we need to + * let the bridge descriptor fetch subsystem know that we + * succeeded at getting it -- so we can adjust the retry schedule + * to stop trying for a while. */ + learned_bridge_descriptor(router, from_cache, 0); + } log_info(LD_DIR, "Dropping descriptor that we already have for router %s", router_describe(router)); @@ -2012,6 +2019,30 @@ routerlist_remove_old_routers(void) router_rebuild_store(RRS_DONT_REMOVE_OLD,&routerlist->extrainfo_store); } +/* Drop every bridge descriptor in our routerlist. Used by the external + * 'bridgestrap' tool to discard bridge descriptors so that it can then + * do a clean reachability test. */ +void +routerlist_drop_bridge_descriptors(void) +{ + routerinfo_t *router; + int i; + + if (!routerlist) + return; + + for (i = 0; i < smartlist_len(routerlist->routers); ++i) { + router = smartlist_get(routerlist->routers, i); + if (router->purpose == ROUTER_PURPOSE_BRIDGE) { + log_notice(LD_DIR, + "Dropping existing bridge descriptor for %s", + router_describe(router)); + routerlist_remove(routerlist, router, 0, time(NULL)); + i--; + } + } +} + /** We just added a new set of descriptors. Take whatever extra steps * we need. */ void @@ -2023,7 +2054,7 @@ routerlist_descriptors_added(smartlist_t *sl, int from_cache) control_event_descriptors_changed(sl); SMARTLIST_FOREACH_BEGIN(sl, routerinfo_t *, ri) { if (ri->purpose == ROUTER_PURPOSE_BRIDGE) - learned_bridge_descriptor(ri, from_cache); + learned_bridge_descriptor(ri, from_cache, 1); if (ri->needs_retest_if_added) { ri->needs_retest_if_added = 0; dirserv_single_reachability_test(approx_time(), ri); diff --git a/src/feature/nodelist/routerlist.h b/src/feature/nodelist/routerlist.h index 7dc748c94b..7ba305baf6 100644 --- a/src/feature/nodelist/routerlist.h +++ b/src/feature/nodelist/routerlist.h @@ -145,6 +145,7 @@ was_router_added_t router_add_extrainfo_to_routerlist( int from_cache, int from_fetch); void routerlist_descriptors_added(smartlist_t *sl, int from_cache); void routerlist_remove_old_routers(void); +void routerlist_drop_bridge_descriptors(void); int router_load_single_router(const char *s, uint8_t purpose, int cache, const char **msg); int router_load_routers_from_string(const char *s, const char *eos, diff --git a/src/feature/nodelist/routerstatus_st.h b/src/feature/nodelist/routerstatus_st.h index 46ff0bdeac..55b76de581 100644 --- a/src/feature/nodelist/routerstatus_st.h +++ b/src/feature/nodelist/routerstatus_st.h @@ -51,6 +51,8 @@ struct routerstatus_t { * choice as an entry guard. */ unsigned int is_bad_exit:1; /**< True iff this node is a bad choice for * an exit node. */ + unsigned int is_middle_only:1; /**< True iff this node is marked as bad + * for anything besides middle positions. */ unsigned int is_hs_dir:1; /**< True iff this router is a v2-or-later hidden * service directory. */ unsigned int is_v2_dir:1; /** True iff this router publishes an open DirPort diff --git a/src/feature/relay/dns.c b/src/feature/relay/dns.c index 6a703f2ab3..c6e0439338 100644 --- a/src/feature/relay/dns.c +++ b/src/feature/relay/dns.c @@ -1539,6 +1539,16 @@ evdns_callback(int result, char type, int count, int ttl, void *addresses, tor_addr_make_unspec(&addr); + /* Note down any DNS errors to the statistics module */ + if (result == DNS_ERR_TIMEOUT) { + /* libevent timed out while resolving a name. However, because libevent + * handles retries and timeouts internally, this means that all attempts of + * libevent timed out. If we wanted to get more granular information about + * individual libevent attempts, we would have to implement our own DNS + * timeout/retry logic */ + rep_hist_note_overload(OVERLOAD_GENERAL); + } + /* Keep track of whether IPv6 is working */ if (type == DNS_IPv6_AAAA) { if (result == DNS_ERR_TIMEOUT) { @@ -1642,7 +1652,7 @@ evdns_callback(int result, char type, int count, int ttl, void *addresses, /* The result can be changed within this function thus why we note the result * at the end. */ - rep_hist_note_dns_query(type, result); + rep_hist_note_dns_error(type, result); tor_free(arg_); } @@ -1662,6 +1672,9 @@ launch_one_resolve(const char *address, uint8_t query_type, addr[0] = (char) query_type; memcpy(addr+1, address, addr_len + 1); + /* Note the query for our statistics. */ + rep_hist_note_dns_request(query_type); + switch (query_type) { case DNS_IPv4_A: req = evdns_base_resolve_ipv4(the_evdns_base, diff --git a/src/feature/relay/include.am b/src/feature/relay/include.am index 84bb1ff35e..8a121cef01 100644 --- a/src/feature/relay/include.am +++ b/src/feature/relay/include.am @@ -15,6 +15,7 @@ MODULE_RELAY_SOURCES = \ src/feature/relay/routermode.c \ src/feature/relay/relay_config.c \ src/feature/relay/relay_handshake.c \ + src/feature/relay/relay_metrics.c \ src/feature/relay/relay_periodic.c \ src/feature/relay/relay_sys.c \ src/feature/relay/routerkeys.c \ @@ -30,6 +31,7 @@ noinst_HEADERS += \ src/feature/relay/onion_queue.h \ src/feature/relay/relay_config.h \ src/feature/relay/relay_handshake.h \ + src/feature/relay/relay_metrics.h \ src/feature/relay/relay_periodic.h \ src/feature/relay/relay_sys.h \ src/feature/relay/relay_find_addr.h \ diff --git a/src/feature/relay/onion_queue.c b/src/feature/relay/onion_queue.c index 85ec0dc74a..c09f4d5b9b 100644 --- a/src/feature/relay/onion_queue.c +++ b/src/feature/relay/onion_queue.c @@ -164,6 +164,7 @@ onion_pending_add(or_circuit_t *circ, create_cell_t *onionskin) #define WARN_TOO_MANY_CIRC_CREATIONS_INTERVAL (60) static ratelim_t last_warned = RATELIM_INIT(WARN_TOO_MANY_CIRC_CREATIONS_INTERVAL); + rep_hist_note_circuit_handshake_dropped(onionskin->handshake_type); if (onionskin->handshake_type == ONION_HANDSHAKE_TYPE_NTOR) { char *m; /* Note this ntor onionskin drop as an overload */ diff --git a/src/feature/relay/relay_metrics.c b/src/feature/relay/relay_metrics.c new file mode 100644 index 0000000000..fc8eb10d1b --- /dev/null +++ b/src/feature/relay/relay_metrics.c @@ -0,0 +1,401 @@ +/* Copyright (c) 2021, The Tor Project, Inc. */ +/* See LICENSE for licensing information */ + +/** + * @file relay_metrics.c + * @brief Relay metrics exposed through the MetricsPort + **/ + +#define RELAY_METRICS_ENTRY_PRIVATE + +#include "orconfig.h" + +#include "core/or/or.h" +#include "core/or/relay.h" + +#include "lib/malloc/malloc.h" +#include "lib/container/smartlist.h" +#include "lib/metrics/metrics_store.h" +#include "lib/log/util_bug.h" + +#include "feature/relay/relay_metrics.h" +#include "feature/stats/rephist.h" + +#include <event2/dns.h> + +/** Declarations of each fill function for metrics defined in base_metrics. */ +static void fill_dns_error_values(void); +static void fill_dns_query_values(void); +static void fill_global_bw_limit_values(void); +static void fill_socket_values(void); +static void fill_onionskins_values(void); +static void fill_oom_values(void); +static void fill_tcp_exhaustion_values(void); + +/** The base metrics that is a static array of metrics added to the metrics + * store. + * + * The key member MUST be also the index of the entry in the array. */ +static const relay_metrics_entry_t base_metrics[] = +{ + { + .key = RELAY_METRICS_NUM_OOM_BYTES, + .type = METRICS_TYPE_COUNTER, + .name = METRICS_NAME(relay_load_oom_bytes_total), + .help = "Total number of bytes the OOM has freed by subsystem", + .fill_fn = fill_oom_values, + }, + { + .key = RELAY_METRICS_NUM_ONIONSKINS, + .type = METRICS_TYPE_COUNTER, + .name = METRICS_NAME(relay_load_onionskins_total), + .help = "Total number of onionskins handled", + .fill_fn = fill_onionskins_values, + }, + { + .key = RELAY_METRICS_NUM_SOCKETS, + .type = METRICS_TYPE_GAUGE, + .name = METRICS_NAME(relay_load_socket_total), + .help = "Total number of sockets", + .fill_fn = fill_socket_values, + }, + { + .key = RELAY_METRICS_NUM_GLOBAL_RW_LIMIT, + .type = METRICS_TYPE_COUNTER, + .name = METRICS_NAME(relay_load_global_rate_limit_reached_total), + .help = "Total number of global connection bucket limit reached", + .fill_fn = fill_global_bw_limit_values, + }, + { + .key = RELAY_METRICS_NUM_DNS, + .type = METRICS_TYPE_COUNTER, + .name = METRICS_NAME(relay_exit_dns_query_total), + .help = "Total number of DNS queries done by this relay", + .fill_fn = fill_dns_query_values, + }, + { + .key = RELAY_METRICS_NUM_DNS_ERRORS, + .type = METRICS_TYPE_COUNTER, + .name = METRICS_NAME(relay_exit_dns_error_total), + .help = "Total number of DNS errors encountered by this relay", + .fill_fn = fill_dns_error_values, + }, + { + .key = RELAY_METRICS_NUM_TCP_EXHAUSTION, + .type = METRICS_TYPE_COUNTER, + .name = METRICS_NAME(relay_load_tcp_exhaustion_total), + .help = "Total number of times we ran out of TCP ports", + .fill_fn = fill_tcp_exhaustion_values, + }, +}; +static const size_t num_base_metrics = ARRAY_LENGTH(base_metrics); + +/** The only and single store of all the relay metrics. */ +static metrics_store_t *the_store; + +/** Helper function to convert an handshake type into a string. */ +static inline const char * +handshake_type_to_str(const uint16_t type) +{ + switch (type) { + case ONION_HANDSHAKE_TYPE_TAP: + return "tap"; + case ONION_HANDSHAKE_TYPE_FAST: + return "fast"; + case ONION_HANDSHAKE_TYPE_NTOR: + return "ntor"; + default: + // LCOV_EXCL_START + tor_assert_unreached(); + // LCOV_EXCL_STOP + } +} + +/** Fill function for the RELAY_METRICS_NUM_DNS metrics. */ +static void +fill_tcp_exhaustion_values(void) +{ + metrics_store_entry_t *sentry; + const relay_metrics_entry_t *rentry = + &base_metrics[RELAY_METRICS_NUM_TCP_EXHAUSTION]; + + sentry = metrics_store_add(the_store, rentry->type, rentry->name, + rentry->help); + metrics_store_entry_update(sentry, rep_hist_get_n_tcp_exhaustion()); +} + +/* NOTE: Disable the record type label until libevent is fixed. */ +#if 0 +/** Helper array containing mapping for the name of the different DNS records + * and their corresponding libevent values. */ +static struct dns_type { + const char *name; + uint8_t type; +} dns_types[] = { + { .name = "A", .type = DNS_IPv4_A }, + { .name = "PTR", .type = DNS_PTR }, + { .name = "AAAA", .type = DNS_IPv6_AAAA }, +}; +static const size_t num_dns_types = ARRAY_LENGTH(dns_types); +#endif + +/** Fill function for the RELAY_METRICS_NUM_DNS_ERRORS metrics. */ +static void +fill_dns_error_values(void) +{ + metrics_store_entry_t *sentry; + const relay_metrics_entry_t *rentry = + &base_metrics[RELAY_METRICS_NUM_DNS_ERRORS]; + + /* Helper array to map libeven DNS errors to their names and so we can + * iterate over this array to add all metrics. */ + static struct dns_error { + const char *name; + uint8_t key; + } errors[] = { + { .name = "success", .key = DNS_ERR_NONE }, + { .name = "format", .key = DNS_ERR_FORMAT }, + { .name = "serverfailed", .key = DNS_ERR_SERVERFAILED }, + { .name = "notexist", .key = DNS_ERR_NOTEXIST }, + { .name = "notimpl", .key = DNS_ERR_NOTIMPL }, + { .name = "refused", .key = DNS_ERR_REFUSED }, + { .name = "truncated", .key = DNS_ERR_TRUNCATED }, + { .name = "unknown", .key = DNS_ERR_UNKNOWN }, + { .name = "tor_timeout", .key = DNS_ERR_TIMEOUT }, + { .name = "shutdown", .key = DNS_ERR_SHUTDOWN }, + { .name = "cancel", .key = DNS_ERR_CANCEL }, + { .name = "nodata", .key = DNS_ERR_NODATA }, + }; + static const size_t num_errors = ARRAY_LENGTH(errors); + + /* NOTE: Disable the record type label until libevent is fixed. */ +#if 0 + for (size_t i = 0; i < num_dns_types; i++) { + /* Dup the label because metrics_format_label() returns a pointer to a + * string on the stack and we need that label for all metrics. */ + char *record_label = + tor_strdup(metrics_format_label("record", dns_types[i].name)); + + for (size_t j = 0; j < num_errors; j++) { + sentry = metrics_store_add(the_store, rentry->type, rentry->name, + rentry->help); + metrics_store_entry_add_label(sentry, record_label); + metrics_store_entry_add_label(sentry, + metrics_format_label("reason", errors[j].name)); + metrics_store_entry_update(sentry, + rep_hist_get_n_dns_error(dns_types[i].type, errors[j].key)); + } + tor_free(record_label); + } +#endif + + /* Put in the DNS errors, unfortunately not per-type for now. */ + for (size_t j = 0; j < num_errors; j++) { + sentry = metrics_store_add(the_store, rentry->type, rentry->name, + rentry->help); + metrics_store_entry_add_label(sentry, + metrics_format_label("reason", errors[j].name)); + metrics_store_entry_update(sentry, + rep_hist_get_n_dns_error(0, errors[j].key)); + } +} + +/** Fill function for the RELAY_METRICS_NUM_DNS metrics. */ +static void +fill_dns_query_values(void) +{ + metrics_store_entry_t *sentry; + const relay_metrics_entry_t *rentry = + &base_metrics[RELAY_METRICS_NUM_DNS]; + + /* NOTE: Disable the record type label until libevent is fixed (#40490). */ +#if 0 + for (size_t i = 0; i < num_dns_types; i++) { + /* Dup the label because metrics_format_label() returns a pointer to a + * string on the stack and we need that label for all metrics. */ + char *record_label = + tor_strdup(metrics_format_label("record", dns_types[i].name)); + sentry = metrics_store_add(the_store, rentry->type, rentry->name, + rentry->help); + metrics_store_entry_add_label(sentry, record_label); + metrics_store_entry_update(sentry, + rep_hist_get_n_dns_request(dns_types[i].type)); + tor_free(record_label); + } +#endif + + sentry = metrics_store_add(the_store, rentry->type, rentry->name, + rentry->help); + metrics_store_entry_update(sentry, rep_hist_get_n_dns_request(0)); +} + +/** Fill function for the RELAY_METRICS_NUM_GLOBAL_RW_LIMIT metrics. */ +static void +fill_global_bw_limit_values(void) +{ + metrics_store_entry_t *sentry; + const relay_metrics_entry_t *rentry = + &base_metrics[RELAY_METRICS_NUM_GLOBAL_RW_LIMIT]; + + sentry = metrics_store_add(the_store, rentry->type, rentry->name, + rentry->help); + metrics_store_entry_add_label(sentry, + metrics_format_label("side", "read")); + metrics_store_entry_update(sentry, rep_hist_get_n_read_limit_reached()); + + sentry = metrics_store_add(the_store, rentry->type, rentry->name, + rentry->help); + metrics_store_entry_add_label(sentry, + metrics_format_label("side", "write")); + metrics_store_entry_update(sentry, rep_hist_get_n_write_limit_reached()); +} + +/** Fill function for the RELAY_METRICS_NUM_SOCKETS metrics. */ +static void +fill_socket_values(void) +{ + metrics_store_entry_t *sentry; + const relay_metrics_entry_t *rentry = + &base_metrics[RELAY_METRICS_NUM_SOCKETS]; + + sentry = metrics_store_add(the_store, rentry->type, rentry->name, + rentry->help); + metrics_store_entry_add_label(sentry, + metrics_format_label("state", "opened")); + metrics_store_entry_update(sentry, get_n_open_sockets()); + + sentry = metrics_store_add(the_store, rentry->type, rentry->name, + rentry->help); + metrics_store_entry_update(sentry, get_max_sockets()); +} + +/** Fill function for the RELAY_METRICS_NUM_ONIONSKINS metrics. */ +static void +fill_onionskins_values(void) +{ + metrics_store_entry_t *sentry; + const relay_metrics_entry_t *rentry = + &base_metrics[RELAY_METRICS_NUM_ONIONSKINS]; + + for (uint16_t t = 0; t <= MAX_ONION_HANDSHAKE_TYPE; t++) { + /* Dup the label because metrics_format_label() returns a pointer to a + * string on the stack and we need that label for all metrics. */ + char *type_label = + tor_strdup(metrics_format_label("type", handshake_type_to_str(t))); + sentry = metrics_store_add(the_store, rentry->type, rentry->name, + rentry->help); + metrics_store_entry_add_label(sentry, type_label); + metrics_store_entry_add_label(sentry, + metrics_format_label("action", "processed")); + metrics_store_entry_update(sentry, + rep_hist_get_circuit_n_handshake_assigned(t)); + + sentry = metrics_store_add(the_store, rentry->type, rentry->name, + rentry->help); + metrics_store_entry_add_label(sentry, type_label); + metrics_store_entry_add_label(sentry, + metrics_format_label("action", "dropped")); + metrics_store_entry_update(sentry, + rep_hist_get_circuit_n_handshake_dropped(t)); + tor_free(type_label); + } +} + +/** Fill function for the RELAY_METRICS_NUM_OOM_BYTES metrics. */ +static void +fill_oom_values(void) +{ + metrics_store_entry_t *sentry; + const relay_metrics_entry_t *rentry = + &base_metrics[RELAY_METRICS_NUM_OOM_BYTES]; + + sentry = metrics_store_add(the_store, rentry->type, rentry->name, + rentry->help); + metrics_store_entry_add_label(sentry, + metrics_format_label("subsys", "cell")); + metrics_store_entry_update(sentry, oom_stats_n_bytes_removed_cell); + + sentry = metrics_store_add(the_store, rentry->type, rentry->name, + rentry->help); + metrics_store_entry_add_label(sentry, + metrics_format_label("subsys", "dns")); + metrics_store_entry_update(sentry, oom_stats_n_bytes_removed_dns); + + sentry = metrics_store_add(the_store, rentry->type, rentry->name, + rentry->help); + metrics_store_entry_add_label(sentry, + metrics_format_label("subsys", "geoip")); + metrics_store_entry_update(sentry, oom_stats_n_bytes_removed_geoip); + + sentry = metrics_store_add(the_store, rentry->type, rentry->name, + rentry->help); + metrics_store_entry_add_label(sentry, + metrics_format_label("subsys", "hsdir")); + metrics_store_entry_update(sentry, oom_stats_n_bytes_removed_hsdir); +} + +/** Reset the global store and fill it with all the metrics from base_metrics + * and their associated values. + * + * To pull this off, every metrics has a "fill" function that is called and in + * charge of adding the metrics to the store, appropriate labels and finally + * updating the value to report. */ +static void +fill_store(void) +{ + /* Reset the current store, we are about to fill it with all the things. */ + metrics_store_reset(the_store); + + /* Call the fill function for each metrics. */ + for (size_t i = 0; i < num_base_metrics; i++) { + if (BUG(!base_metrics[i].fill_fn)) { + continue; + } + base_metrics[i].fill_fn(); + } +} + +/** Return a list of all the relay metrics stores. This is the + * function attached to the .get_metrics() member of the subsys_t. */ +const smartlist_t * +relay_metrics_get_stores(void) +{ + /* We can't have the caller to free the returned list so keep it static, + * simply update it. */ + static smartlist_t *stores_list = NULL; + + /* We dynamically fill the store with all the metrics upon a request. The + * reason for this is because the exposed metrics of a relay are often + * internal counters in the fast path and thus we fetch the value when a + * metrics port request arrives instead of keeping a local metrics store of + * those values. */ + fill_store(); + + if (!stores_list) { + stores_list = smartlist_new(); + smartlist_add(stores_list, the_store); + } + + return stores_list; +} + +/** Initialize the relay metrics. */ +void +relay_metrics_init(void) +{ + if (BUG(the_store)) { + return; + } + the_store = metrics_store_new(); +} + +/** Free the relay metrics. */ +void +relay_metrics_free(void) +{ + if (!the_store) { + return; + } + /* NULL is set with this call. */ + metrics_store_free(the_store); +} diff --git a/src/feature/relay/relay_metrics.h b/src/feature/relay/relay_metrics.h new file mode 100644 index 0000000000..00dfeaa624 --- /dev/null +++ b/src/feature/relay/relay_metrics.h @@ -0,0 +1,55 @@ +/* Copyright (c) 2021, The Tor Project, Inc. */ +/* See LICENSE for licensing information */ + +/** + * @file relay_metrics.h + * @brief Header for feature/relay/relay_metrics.c + **/ + +#ifndef TOR_FEATURE_RELAY_RELAY_METRICS_H +#define TOR_FEATURE_RELAY_RELAY_METRICS_H + +#include "lib/container/smartlist.h" +#include "lib/metrics/metrics_common.h" + +/** Metrics key for each reported metrics. This key is also used as an index in + * the base_metrics array. */ +typedef enum { + /** Number of OOM invocation. */ + RELAY_METRICS_NUM_OOM_BYTES = 0, + /** Number of onionskines handled. */ + RELAY_METRICS_NUM_ONIONSKINS = 1, + /** Number of sockets. */ + RELAY_METRICS_NUM_SOCKETS = 2, + /** Number of global connection rate limit. */ + RELAY_METRICS_NUM_GLOBAL_RW_LIMIT = 3, + /** Number of DNS queries. */ + RELAY_METRICS_NUM_DNS = 4, + /** Number of DNS query errors. */ + RELAY_METRICS_NUM_DNS_ERRORS = 5, + /** Number of TCP exhaustion reached. */ + RELAY_METRICS_NUM_TCP_EXHAUSTION = 6, +} relay_metrics_key_t; + +/** The metadata of a relay metric. */ +typedef struct relay_metrics_entry_t { + /* Metric key used as a static array index. */ + relay_metrics_key_t key; + /* Metric type. */ + metrics_type_t type; + /* Metrics output name. */ + const char *name; + /* Metrics output help comment. */ + const char *help; + /* Update value function. */ + void (*fill_fn)(void); +} relay_metrics_entry_t; + +/* Init. */ +void relay_metrics_init(void); +void relay_metrics_free(void); + +/* Accessors. */ +const smartlist_t *relay_metrics_get_stores(void); + +#endif /* !defined(TOR_FEATURE_RELAY_RELAY_METRICS_H) */ diff --git a/src/feature/relay/relay_periodic.c b/src/feature/relay/relay_periodic.c index ee94590e01..dd9be4e36f 100644 --- a/src/feature/relay/relay_periodic.c +++ b/src/feature/relay/relay_periodic.c @@ -219,7 +219,7 @@ reachability_warnings_callback(time_t now, const or_options_t *options) tor_asprintf(&where4, "%s:%d", address4, me->ipv4_orport); if (!v6_ok) tor_asprintf(&where6, "[%s]:%d", address6, me->ipv6_orport); - const char *opt_and = (!v4_ok && !v6_ok) ? "and" : ""; + const char *opt_and = (!v4_ok && !v6_ok) ? " and " : ""; /* IPv4 reachability test worked but not the IPv6. We will _not_ * publish the descriptor if our IPv6 was configured. We will if it diff --git a/src/feature/relay/relay_sys.c b/src/feature/relay/relay_sys.c index 25fc0bbd32..9c43734b84 100644 --- a/src/feature/relay/relay_sys.c +++ b/src/feature/relay/relay_sys.c @@ -14,6 +14,7 @@ #include "feature/relay/dns.h" #include "feature/relay/ext_orport.h" +#include "feature/relay/relay_metrics.h" #include "feature/relay/onion_queue.h" #include "feature/relay/relay_periodic.h" #include "feature/relay/relay_sys.h" @@ -25,6 +26,7 @@ static int subsys_relay_initialize(void) { + relay_metrics_init(); relay_register_periodic_events(); return 0; } @@ -37,6 +39,7 @@ subsys_relay_shutdown(void) clear_pending_onions(); routerkeys_free_all(); router_free_all(); + relay_metrics_free(); } const struct subsys_fns_t sys_relay = { @@ -46,4 +49,6 @@ const struct subsys_fns_t sys_relay = { .level = RELAY_SUBSYS_LEVEL, .initialize = subsys_relay_initialize, .shutdown = subsys_relay_shutdown, + + .get_metrics = relay_metrics_get_stores, }; diff --git a/src/feature/rend/rendmid.c b/src/feature/rend/rendmid.c index df838aa527..8f6a45dfef 100644 --- a/src/feature/rend/rendmid.c +++ b/src/feature/rend/rendmid.c @@ -131,7 +131,11 @@ rend_mid_rendezvous(or_circuit_t *circ, const uint8_t *request, rend_circ = hs_circuitmap_get_rend_circ_relay_side(request); if (!rend_circ) { - log_fn(LOG_PROTOCOL_WARN, LD_PROTOCOL, + /* Once this was a LOG_PROTOCOL_WARN, but it can happen naturally if a + * client gives up on a rendezvous circuit after sending INTRODUCE1, but + * before the onion service sends the RENDEZVOUS1 cell. + */ + log_fn(LOG_DEBUG, LD_PROTOCOL, "Rejecting RENDEZVOUS1 cell with unrecognized rendezvous cookie %s.", hexid); reason = END_CIRC_REASON_TORPROTOCOL; diff --git a/src/feature/stats/geoip_stats.c b/src/feature/stats/geoip_stats.c index b4b107c3f7..a0fe8597c1 100644 --- a/src/feature/stats/geoip_stats.c +++ b/src/feature/stats/geoip_stats.c @@ -1206,11 +1206,11 @@ format_bridge_stats_controller(time_t now) char * format_client_stats_heartbeat(time_t now) { - const int n_hours = 6; + const int n_seconds = get_options()->HeartbeatPeriod; char *out = NULL; int n_clients = 0; clientmap_entry_t **ent; - unsigned cutoff = (unsigned)( (now-n_hours*3600)/60 ); + unsigned cutoff = (unsigned)( (now-n_seconds)/60 ); if (!start_of_bridge_stats_interval) return NULL; /* Not initialized. */ @@ -1226,8 +1226,7 @@ format_client_stats_heartbeat(time_t now) } tor_asprintf(&out, "Heartbeat: " - "In the last %d hours, I have seen %d unique clients.", - n_hours, + "Since last heartbeat message, I have seen %d unique clients.", n_clients); return out; diff --git a/src/feature/stats/rephist.c b/src/feature/stats/rephist.c index 2bfa14d326..5ff4ef1d2e 100644 --- a/src/feature/stats/rephist.c +++ b/src/feature/stats/rephist.c @@ -206,18 +206,33 @@ typedef struct { uint64_t overload_fd_exhausted; } overload_stats_t; +/** Current state of overload stats */ +static overload_stats_t overload_stats; + +/** Counters to count the number of times we've reached an overload for the + * global connection read/write limit. Reported on the MetricsPort. */ +static uint64_t stats_n_read_limit_reached = 0; +static uint64_t stats_n_write_limit_reached = 0; + +/** Total number of times we've reached TCP port exhaustion. */ +static uint64_t stats_n_tcp_exhaustion = 0; + /***** DNS statistics *****/ -/** Represents the statistics of DNS queries seen if it is an Exit. */ +/** Overload DNS statistics. The information in this object is used to assess + * if, due to DNS errors, we should emit a general overload signal or not. + * + * NOTE: This structure is _not_ per DNS query type like the statistics below + * because of a libevent bug + * (https://github.com/libevent/libevent/issues/1219), on error, the type is + * not propagated up back to the user and so we need to keep our own stats for + * the overload signal. */ typedef struct { /** Total number of DNS request seen at an Exit. They might not all end * successfully or might even be lost by tor. This counter is incremented * right before the DNS request is initiated. */ uint64_t stats_n_request; - /** Total number of DNS timeout errors. */ - uint64_t stats_n_error_timeout; - /** When is the next assessment time of the general overload for DNS errors. * Once this time is reached, all stats are reset and this time is set to the * next assessment time. */ @@ -227,121 +242,230 @@ typedef struct { /** Keep track of the DNS requests for the general overload state. */ static overload_dns_stats_t overload_dns_stats; -/* We use a scale here so we can represent percentages with decimal points by - * scaling the value by this factor and so 0.5% becomes a value of 500. - * Default is 1% and thus min and max range is 0 to 100%. */ -#define OVERLOAD_DNS_TIMEOUT_PERCENT_SCALE 1000.0 -#define OVERLOAD_DNS_TIMEOUT_PERCENT_DEFAULT 1000 -#define OVERLOAD_DNS_TIMEOUT_PERCENT_MIN 0 -#define OVERLOAD_DNS_TIMEOUT_PERCENT_MAX 100000 - -/** Consensus parameter: indicate what fraction of DNS timeout errors over the - * total number of DNS requests must be reached before we trigger a general - * overload signal .*/ -static double overload_dns_timeout_fraction = - OVERLOAD_DNS_TIMEOUT_PERCENT_DEFAULT / - OVERLOAD_DNS_TIMEOUT_PERCENT_SCALE / 100.0; - -/* Number of seconds for the assessment period. Default is 10 minutes (600) and - * the min max range is within a 32bit value. */ -#define OVERLOAD_DNS_TIMEOUT_PERIOD_SECS_DEFAULT (10 * 60) -#define OVERLOAD_DNS_TIMEOUT_PERIOD_SECS_MIN 0 -#define OVERLOAD_DNS_TIMEOUT_PERIOD_SECS_MAX INT32_MAX - -/** Consensus parameter: Period, in seconds, over which we count the number of - * DNS requests and timeout errors. After that period, we assess if we trigger - * an overload or not. */ -static int32_t overload_dns_timeout_period_secs = - OVERLOAD_DNS_TIMEOUT_PERIOD_SECS_DEFAULT; +/** Represents the statistics of DNS queries seen if it is an Exit. */ +typedef struct { + /* Total number of DNS errors found in RFC 1035 (from 0 to 5 code). */ + uint64_t stats_n_error_none; /* 0 */ + uint64_t stats_n_error_format; /* 1 */ + uint64_t stats_n_error_serverfailed; /* 2 */ + uint64_t stats_n_error_notexist; /* 3 */ + uint64_t stats_n_error_notimpl; /* 4 */ + uint64_t stats_n_error_refused; /* 5 */ + + /* Total number of DNS errors specific to libevent. */ + uint64_t stats_n_error_truncated; /* 65 */ + uint64_t stats_n_error_unknown; /* 66 */ + uint64_t stats_n_error_tor_timeout; /* 67 */ + uint64_t stats_n_error_shutdown; /* 68 */ + uint64_t stats_n_error_cancel; /* 69 */ + uint64_t stats_n_error_nodata; /* 70 */ + + /* Total number of DNS request seen at an Exit. They might not all end + * successfully or might even be lost by tor. This counter is incremented + * right before the DNS request is initiated. */ + uint64_t stats_n_request; +} dns_stats_t; + +/* This is disabled because of the libevent bug where on error we don't get the + * DNS query type back. Once it is fixed, we can re-enable this. */ +#if 0 +/** DNS statistics store for each DNS record type for which tor supports only + * three at the moment: A, PTR and AAAA. */ +static dns_stats_t dns_A_stats; +static dns_stats_t dns_PTR_stats; +static dns_stats_t dns_AAAA_stats; +#endif -/** Current state of overload stats */ -static overload_stats_t overload_stats; +/** DNS query statistics store. It covers all type of queries. */ +static dns_stats_t dns_all_stats; -/** Return true if this overload happened within the last `n_hours`. */ -static bool -overload_happened_recently(time_t overload_time, int n_hours) +/** Return the point to the DNS statistics store. Ignore the type for now + * because of a libevent problem. */ +static inline dns_stats_t * +get_dns_stats_by_type(const int type) { - /* An overload is relevant if it happened in the last 72 hours */ - if (overload_time > approx_time() - 3600 * n_hours) { - return true; - } - return false; + (void) type; + return &dns_all_stats; } -/** Assess the DNS timeout errors and if we have enough to trigger a general - * overload. */ -static void -overload_general_dns_assessment(void) +#if 0 +/** From a libevent record type, return a pointer to the corresponding DNS + * statistics store. NULL is returned if the type is unhandled. */ +static inline dns_stats_t * +get_dns_stats_by_type(const int type) { - /* Initialize the time. Should be done once. */ - if (overload_dns_stats.next_assessment_time == 0) { - goto reset; + switch (type) { + case DNS_IPv4_A: + return &dns_A_stats; + case DNS_PTR: + return &dns_PTR_stats; + case DNS_IPv6_AAAA: + return &dns_AAAA_stats; + default: + return NULL; } +} +#endif - /* Not the time yet. */ - if (overload_dns_stats.next_assessment_time > approx_time()) { - return; +/** Return the DNS error count for the given libevent DNS type and error code. + * The possible types are: DNS_IPv4_A, DNS_PTR, DNS_IPv6_AAAA. */ +uint64_t +rep_hist_get_n_dns_error(int type, uint8_t error) +{ + dns_stats_t *dns_stats = get_dns_stats_by_type(type); + if (BUG(!dns_stats)) { + return 0; } - reset: - /* Reset counters for the next period. */ - overload_dns_stats.stats_n_error_timeout = 0; - overload_dns_stats.stats_n_request = 0; - overload_dns_stats.next_assessment_time = - approx_time() + overload_dns_timeout_period_secs; + switch (error) { + case DNS_ERR_NONE: + return dns_stats->stats_n_error_none; + case DNS_ERR_FORMAT: + return dns_stats->stats_n_error_format; + case DNS_ERR_SERVERFAILED: + return dns_stats->stats_n_error_serverfailed; + case DNS_ERR_NOTEXIST: + return dns_stats->stats_n_error_notexist; + case DNS_ERR_NOTIMPL: + return dns_stats->stats_n_error_notimpl; + case DNS_ERR_REFUSED: + return dns_stats->stats_n_error_refused; + case DNS_ERR_TRUNCATED: + return dns_stats->stats_n_error_truncated; + case DNS_ERR_UNKNOWN: + return dns_stats->stats_n_error_unknown; + case DNS_ERR_TIMEOUT: + return dns_stats->stats_n_error_tor_timeout; + case DNS_ERR_SHUTDOWN: + return dns_stats->stats_n_error_shutdown; + case DNS_ERR_CANCEL: + return dns_stats->stats_n_error_cancel; + case DNS_ERR_NODATA: + return dns_stats->stats_n_error_nodata; + default: + /* Unhandled code sent back by libevent. */ + return 0; + } } -/** Called just before the consensus will be replaced. Update the consensus - * parameters in case they changed. */ -void -rep_hist_consensus_has_changed(const networkstatus_t *ns) +/** Return the total number of DNS request seen for the given libevent DNS + * record type. Possible types are: DNS_IPv4_A, DNS_PTR, DNS_IPv6_AAAA. */ +uint64_t +rep_hist_get_n_dns_request(int type) { - overload_dns_timeout_fraction = - networkstatus_get_param(ns, "overload_dns_timeout_scale_percent", - OVERLOAD_DNS_TIMEOUT_PERCENT_DEFAULT, - OVERLOAD_DNS_TIMEOUT_PERCENT_MIN, - OVERLOAD_DNS_TIMEOUT_PERCENT_MAX) / - OVERLOAD_DNS_TIMEOUT_PERCENT_SCALE / 100.0; - - overload_dns_timeout_period_secs = - networkstatus_get_param(ns, "overload_dns_timeout_period_secs", - OVERLOAD_DNS_TIMEOUT_PERIOD_SECS_DEFAULT, - OVERLOAD_DNS_TIMEOUT_PERIOD_SECS_MIN, - OVERLOAD_DNS_TIMEOUT_PERIOD_SECS_MAX); + dns_stats_t *dns_stats = get_dns_stats_by_type(type); + if (BUG(!dns_stats)) { + return 0; + } + return dns_stats->stats_n_request; } /** Note a DNS error for the given given libevent DNS record type and error * code. Possible types are: DNS_IPv4_A, DNS_PTR, DNS_IPv6_AAAA. * - * IMPORTANT: Libevent is _not_ returning the type in case of an error and so - * if error is anything but DNS_ERR_NONE, the type is not usable and set to 0. + * NOTE: Libevent is _not_ returning the type in case of an error and so if + * error is anything but DNS_ERR_NONE, the type is not usable and set to 0. * * See: https://gitlab.torproject.org/tpo/core/tor/-/issues/40490 */ void -rep_hist_note_dns_query(int type, uint8_t error) +rep_hist_note_dns_error(int type, uint8_t error) { - (void) type; + overload_dns_stats.stats_n_request++; - /* Assess if we need to trigger a general overload with regards to the DNS - * errors or not. */ - overload_general_dns_assessment(); + /* Again, the libevent bug (see function comment), for an error that is + * anything but DNS_ERR_NONE, the type is always 0 which means that we don't + * have a DNS stat object for it so this code will do nothing until libevent + * is fixed. */ + dns_stats_t *dns_stats = get_dns_stats_by_type(type); + /* Unsupported DNS query type. */ + if (!dns_stats) { + return; + } - /* We only care about timeouts for the moment. */ switch (error) { + case DNS_ERR_NONE: + dns_stats->stats_n_error_none++; + break; + case DNS_ERR_FORMAT: + dns_stats->stats_n_error_format++; + break; + case DNS_ERR_SERVERFAILED: + dns_stats->stats_n_error_serverfailed++; + break; + case DNS_ERR_NOTEXIST: + dns_stats->stats_n_error_notexist++; + break; + case DNS_ERR_NOTIMPL: + dns_stats->stats_n_error_notimpl++; + break; + case DNS_ERR_REFUSED: + dns_stats->stats_n_error_refused++; + break; + case DNS_ERR_TRUNCATED: + dns_stats->stats_n_error_truncated++; + break; + case DNS_ERR_UNKNOWN: + dns_stats->stats_n_error_unknown++; + break; case DNS_ERR_TIMEOUT: - overload_dns_stats.stats_n_error_timeout++; + dns_stats->stats_n_error_tor_timeout++; + break; + case DNS_ERR_SHUTDOWN: + dns_stats->stats_n_error_shutdown++; + break; + case DNS_ERR_CANCEL: + dns_stats->stats_n_error_cancel++; + break; + case DNS_ERR_NODATA: + dns_stats->stats_n_error_nodata++; break; default: + /* Unhandled code sent back by libevent. */ break; } +} - /* Increment total number of requests. */ - overload_dns_stats.stats_n_request++; +/** Note a DNS request for the given given libevent DNS record type. */ +void +rep_hist_note_dns_request(int type) +{ + dns_stats_t *dns_stats = get_dns_stats_by_type(type); + if (BUG(!dns_stats)) { + return; + } + dns_stats->stats_n_request++; +} + +/***** END of DNS statistics *****/ + +/** Return true if this overload happened within the last `n_hours`. */ +static bool +overload_happened_recently(time_t overload_time, int n_hours) +{ + /* An overload is relevant if it happened in the last 72 hours */ + if (overload_time > approx_time() - 3600 * n_hours) { + return true; + } + return false; } /* The current version of the overload stats version */ #define OVERLOAD_STATS_VERSION 1 +/** Return the stats_n_read_limit_reached counter. */ +uint64_t +rep_hist_get_n_read_limit_reached(void) +{ + return stats_n_read_limit_reached; +} + +/** Return the stats_n_write_limit_reached counter. */ +uint64_t +rep_hist_get_n_write_limit_reached(void) +{ + return stats_n_write_limit_reached; +} + /** Returns an allocated string for server descriptor for publising information * on whether we are overloaded or not. */ char * @@ -420,6 +544,7 @@ rep_hist_note_overload(overload_type_t overload) SET_TO_START_OF_HOUR(overload_stats.overload_general_time); break; case OVERLOAD_READ: { + stats_n_read_limit_reached++; SET_TO_START_OF_HOUR(overload_stats.overload_ratelimits_time); if (approx_time() >= last_read_counted + 60) { /* Count once a minute */ overload_stats.overload_read_count++; @@ -428,6 +553,7 @@ rep_hist_note_overload(overload_type_t overload) break; } case OVERLOAD_WRITE: { + stats_n_write_limit_reached++; SET_TO_START_OF_HOUR(overload_stats.overload_ratelimits_time); if (approx_time() >= last_write_counted + 60) { /* Count once a minute */ overload_stats.overload_write_count++; @@ -442,6 +568,22 @@ rep_hist_note_overload(overload_type_t overload) } } +/** Note down that we've reached a TCP port exhaustion. This triggers an + * overload general event. */ +void +rep_hist_note_tcp_exhaustion(void) +{ + stats_n_tcp_exhaustion++; + rep_hist_note_overload(OVERLOAD_GENERAL); +} + +/** Return the total number of TCP exhaustion times we've reached. */ +uint64_t +rep_hist_get_n_tcp_exhaustion(void) +{ + return stats_n_tcp_exhaustion; +} + /** Return the or_history_t for the OR with identity digest <b>id</b>, * creating it if necessary. */ static or_history_t * @@ -641,7 +783,7 @@ rep_hist_downrate_old_runs(time_t now) return stability_last_downrated + STABILITY_INTERVAL; /* Okay, we should downrate the data. By how much? */ - while (stability_last_downrated + STABILITY_INTERVAL < now) { + while (stability_last_downrated + STABILITY_INTERVAL <= now) { stability_last_downrated += STABILITY_INTERVAL; alpha *= STABILITY_ALPHA; } @@ -1908,11 +2050,18 @@ rep_hist_note_desc_served(const char * desc) /** Internal statistics to track how many requests of each type of * handshake we've received, and how many we've assigned to cpuworkers. * Useful for seeing trends in cpu load. + * + * They are reset at every heartbeat. * @{ */ STATIC int onion_handshakes_requested[MAX_ONION_HANDSHAKE_TYPE+1] = {0}; STATIC int onion_handshakes_assigned[MAX_ONION_HANDSHAKE_TYPE+1] = {0}; /**@}*/ +/** Counters keeping the same stats as above but for the entire duration of the + * process (not reset). */ +static uint64_t stats_n_onionskin_assigned[MAX_ONION_HANDSHAKE_TYPE+1] = {0}; +static uint64_t stats_n_onionskin_dropped[MAX_ONION_HANDSHAKE_TYPE+1] = {0}; + /** A new onionskin (using the <b>type</b> handshake) has arrived. */ void rep_hist_note_circuit_handshake_requested(uint16_t type) @@ -1926,8 +2075,20 @@ rep_hist_note_circuit_handshake_requested(uint16_t type) void rep_hist_note_circuit_handshake_assigned(uint16_t type) { - if (type <= MAX_ONION_HANDSHAKE_TYPE) + if (type <= MAX_ONION_HANDSHAKE_TYPE) { onion_handshakes_assigned[type]++; + stats_n_onionskin_assigned[type]++; + } +} + +/** We've just drop an onionskin (using the <b>type</b> handshake) due to being + * overloaded. */ +void +rep_hist_note_circuit_handshake_dropped(uint16_t type) +{ + if (type <= MAX_ONION_HANDSHAKE_TYPE) { + stats_n_onionskin_dropped[type]++; + } } /** Get the circuit handshake value that is requested. */ @@ -1950,6 +2111,26 @@ rep_hist_get_circuit_handshake_assigned, (uint16_t type)) return onion_handshakes_assigned[type]; } +/** Get the total number of circuit handshake value that is assigned. */ +MOCK_IMPL(uint64_t, +rep_hist_get_circuit_n_handshake_assigned, (uint16_t type)) +{ + if (BUG(type > MAX_ONION_HANDSHAKE_TYPE)) { + return 0; + } + return stats_n_onionskin_assigned[type]; +} + +/** Get the total number of circuit handshake value that is dropped. */ +MOCK_IMPL(uint64_t, +rep_hist_get_circuit_n_handshake_dropped, (uint16_t type)) +{ + if (BUG(type > MAX_ONION_HANDSHAKE_TYPE)) { + return 0; + } + return stats_n_onionskin_dropped[type]; +} + /** Log our onionskin statistics since the last time we were called. */ void rep_hist_log_circuit_handshake_stats(time_t now) diff --git a/src/feature/stats/rephist.h b/src/feature/stats/rephist.h index 749b4996a8..7f414de4c8 100644 --- a/src/feature/stats/rephist.h +++ b/src/feature/stats/rephist.h @@ -58,11 +58,17 @@ time_t rep_hist_desc_stats_write(time_t now); void rep_hist_note_circuit_handshake_requested(uint16_t type); void rep_hist_note_circuit_handshake_assigned(uint16_t type); +void rep_hist_note_circuit_handshake_dropped(uint16_t type); void rep_hist_log_circuit_handshake_stats(time_t now); MOCK_DECL(int, rep_hist_get_circuit_handshake_requested, (uint16_t type)); MOCK_DECL(int, rep_hist_get_circuit_handshake_assigned, (uint16_t type)); +MOCK_DECL(uint64_t, rep_hist_get_circuit_n_handshake_assigned, + (uint16_t type)); +MOCK_DECL(uint64_t, rep_hist_get_circuit_n_handshake_dropped, + (uint16_t type)); + void rep_hist_hs_stats_init(time_t now); void rep_hist_hs_stats_term(void); time_t rep_hist_hs_stats_write(time_t now, bool is_v3); @@ -72,14 +78,16 @@ void rep_hist_seen_new_rp_cell(bool is_v2); char *rep_hist_get_hs_v3_stats_string(void); void rep_hist_hsdir_stored_maybe_new_v3_onion(const uint8_t *blinded_key); -void rep_hist_note_dns_query(int type, uint8_t error); - void rep_hist_free_all(void); void rep_hist_note_negotiated_link_proto(unsigned link_proto, int started_here); void rep_hist_log_link_protocol_counts(void); -void rep_hist_consensus_has_changed(const networkstatus_t *ns); + +uint64_t rep_hist_get_n_dns_error(int type, uint8_t error); +uint64_t rep_hist_get_n_dns_request(int type); +void rep_hist_note_dns_request(int type); +void rep_hist_note_dns_error(int type, uint8_t error); extern uint64_t rephist_total_alloc; extern uint32_t rephist_total_num; @@ -162,6 +170,12 @@ void rep_hist_note_overload(overload_type_t overload); char *rep_hist_get_overload_general_line(void); char *rep_hist_get_overload_stats_lines(void); +void rep_hist_note_tcp_exhaustion(void); +uint64_t rep_hist_get_n_tcp_exhaustion(void); + +uint64_t rep_hist_get_n_read_limit_reached(void); +uint64_t rep_hist_get_n_write_limit_reached(void); + #ifdef TOR_UNIT_TESTS struct hs_v2_stats_t; const struct hs_v2_stats_t *rep_hist_get_hs_v2_stats(void); |