[Author Prev][Author Next][Thread Prev][Thread Next][Author Index][Thread Index]
[tor-commits] [tor/master] Merge branch 'bug23512-v4-033' into bug23512-v4-master
commit 8a83c4b61397a5923257746f5b06418f8aef9959
Merge: aebc98d58 ad10cafd9
Author: Mike Perry <mikeperry-git@xxxxxxxxxxxxxx>
Date: Tue Sep 18 00:17:14 2018 +0000
Merge branch 'bug23512-v4-033' into bug23512-v4-master
changes/bug23512 | 6 +++
src/core/or/channeltls.h | 2 +
src/core/or/circuitlist.c | 59 +++++++++++++++++++++++
src/core/or/circuitlist.h | 2 +
src/core/or/or.h | 12 +++++
src/core/or/relay.c | 1 +
src/core/or/scheduler_kist.c | 2 -
src/feature/stats/rephist.c | 21 ++++----
src/feature/stats/rephist.h | 10 ++--
src/test/test_relay.c | 111 +++++++++++++++++++++++++++++++++++++++++--
10 files changed, 207 insertions(+), 19 deletions(-)
diff --cc src/core/or/channeltls.h
index 4a10d51d9,000000000..12715450b
mode 100644,000000..100644
--- a/src/core/or/channeltls.h
+++ b/src/core/or/channeltls.h
@@@ -1,77 -1,0 +1,79 @@@
+/* * Copyright (c) 2012-2018, The Tor Project, Inc. */
+/* See LICENSE for licensing information */
+
+/**
+ * \file channeltls.h
+ * \brief Header file for channeltls.c
+ **/
+
+#ifndef TOR_CHANNELTLS_H
+#define TOR_CHANNELTLS_H
+
+#include "core/or/or.h"
+#include "core/or/channel.h"
+
+struct ed25519_public_key_t;
+struct curve25519_public_key_t;
+
++#define TLS_PER_CELL_OVERHEAD 29
++
+#define BASE_CHAN_TO_TLS(c) (channel_tls_from_base((c)))
+#define TLS_CHAN_TO_BASE(c) (channel_tls_to_base((c)))
+
+#define TLS_CHAN_MAGIC 0x8a192427U
+
+#ifdef TOR_CHANNEL_INTERNAL_
+
+struct channel_tls_s {
+ /* Base channel_t struct */
+ channel_t base_;
+ /* or_connection_t pointer */
+ or_connection_t *conn;
+};
+
+#endif /* defined(TOR_CHANNEL_INTERNAL_) */
+
+channel_t * channel_tls_connect(const tor_addr_t *addr, uint16_t port,
+ const char *id_digest,
+ const struct ed25519_public_key_t *ed_id);
+channel_listener_t * channel_tls_get_listener(void);
+channel_listener_t * channel_tls_start_listener(void);
+channel_t * channel_tls_handle_incoming(or_connection_t *orconn);
+
+/* Casts */
+
+channel_t * channel_tls_to_base(channel_tls_t *tlschan);
+channel_tls_t * channel_tls_from_base(channel_t *chan);
+
+/* Things for connection_or.c to call back into */
+void channel_tls_handle_cell(cell_t *cell, or_connection_t *conn);
+void channel_tls_handle_state_change_on_orconn(channel_tls_t *chan,
+ or_connection_t *conn,
+ uint8_t old_state,
+ uint8_t state);
+void channel_tls_handle_var_cell(var_cell_t *var_cell,
+ or_connection_t *conn);
+void channel_tls_update_marks(or_connection_t *conn);
+
+/* Cleanup at shutdown */
+void channel_tls_free_all(void);
+
+extern uint64_t stats_n_authorize_cells_processed;
+extern uint64_t stats_n_authenticate_cells_processed;
+extern uint64_t stats_n_versions_cells_processed;
+extern uint64_t stats_n_netinfo_cells_processed;
+extern uint64_t stats_n_vpadding_cells_processed;
+extern uint64_t stats_n_certs_cells_processed;
+extern uint64_t stats_n_auth_challenge_cells_processed;
+
+#ifdef CHANNELTLS_PRIVATE
+STATIC void channel_tls_process_certs_cell(var_cell_t *cell,
+ channel_tls_t *tlschan);
+STATIC void channel_tls_process_auth_challenge_cell(var_cell_t *cell,
+ channel_tls_t *tlschan);
+STATIC void channel_tls_common_init(channel_tls_t *tlschan);
+STATIC void channel_tls_process_authenticate_cell(var_cell_t *cell,
+ channel_tls_t *tlschan);
+#endif /* defined(CHANNELTLS_PRIVATE) */
+
+#endif /* !defined(TOR_CHANNELTLS_H) */
diff --cc src/core/or/circuitlist.c
index 78ecadab7,000000000..7cf5e3445
mode 100644,000000..100644
--- a/src/core/or/circuitlist.c
+++ b/src/core/or/circuitlist.c
@@@ -1,2751 -1,0 +1,2810 @@@
+/* Copyright 2001 Matej Pfajfar.
+ * Copyright (c) 2001-2004, Roger Dingledine.
+ * Copyright (c) 2004-2006, Roger Dingledine, Nick Mathewson.
+ * Copyright (c) 2007-2018, The Tor Project, Inc. */
+/* See LICENSE for licensing information */
+
+/**
+ * \file circuitlist.c
+ *
+ * \brief Manage global structures that list and index circuits, and
+ * look up circuits within them.
+ *
+ * One of the most frequent operations in Tor occurs every time that
+ * a relay cell arrives on a channel. When that happens, we need to
+ * find which circuit it is associated with, based on the channel and the
+ * circuit ID in the relay cell.
+ *
+ * To handle that, we maintain a global list of circuits, and a hashtable
+ * mapping [channel,circID] pairs to circuits. Circuits are added to and
+ * removed from this mapping using circuit_set_p_circid_chan() and
+ * circuit_set_n_circid_chan(). To look up a circuit from this map, most
+ * callers should use circuit_get_by_circid_channel(), though
+ * circuit_get_by_circid_channel_even_if_marked() is appropriate under some
+ * circumstances.
+ *
+ * We also need to allow for the possibility that we have blocked use of a
+ * circuit ID (because we are waiting to send a DESTROY cell), but the
+ * circuit is not there any more. For that case, we allow placeholder
+ * entries in the table, using channel_mark_circid_unusable().
+ *
+ * To efficiently handle a channel that has just opened, we also maintain a
+ * list of the circuits waiting for channels, so we can attach them as
+ * needed without iterating through the whole list of circuits, using
+ * circuit_get_all_pending_on_channel().
+ *
+ * In this module, we also handle the list of circuits that have been
+ * marked for close elsewhere, and close them as needed. (We use this
+ * "mark now, close later" pattern here and elsewhere to avoid
+ * unpredictable recursion if we closed every circuit immediately upon
+ * realizing it needed to close.) See circuit_mark_for_close() for the
+ * mark function, and circuit_close_all_marked() for the close function.
+ *
+ * For hidden services, we need to be able to look up introduction point
+ * circuits and rendezvous circuits by cookie, key, etc. These are
+ * currently handled with linear searches in
+ * circuit_get_ready_rend_circuit_by_rend_data(),
+ * circuit_get_next_by_pk_and_purpose(), and with hash lookups in
+ * circuit_get_rendezvous() and circuit_get_intro_point().
+ *
+ * This module is also the entry point for our out-of-memory handler
+ * logic, which was originally circuit-focused.
+ **/
+#define CIRCUITLIST_PRIVATE
+#include "lib/cc/torint.h" /* TOR_PRIuSZ */
+
+#include "core/or/or.h"
+#include "core/or/channel.h"
++#include "core/or/channeltls.h"
+#include "feature/client/circpathbias.h"
+#include "core/or/circuitbuild.h"
+#include "core/or/circuitlist.h"
+#include "core/or/circuituse.h"
+#include "core/or/circuitstats.h"
+#include "core/mainloop/connection.h"
+#include "app/config/config.h"
+#include "core/or/connection_edge.h"
+#include "core/or/connection_or.h"
+#include "feature/control/control.h"
+#include "lib/crypt_ops/crypto_rand.h"
+#include "lib/crypt_ops/crypto_util.h"
+#include "lib/crypt_ops/crypto_dh.h"
+#include "feature/dircache/directory.h"
+#include "feature/client/entrynodes.h"
+#include "core/mainloop/main.h"
+#include "feature/hs/hs_circuit.h"
+#include "feature/hs/hs_circuitmap.h"
+#include "feature/hs/hs_ident.h"
+#include "feature/nodelist/networkstatus.h"
+#include "feature/nodelist/nodelist.h"
+#include "core/crypto/onion.h"
+#include "core/crypto/onion_fast.h"
+#include "core/or/policies.h"
+#include "core/or/relay.h"
+#include "core/crypto/relay_crypto.h"
+#include "feature/rend/rendclient.h"
+#include "feature/rend/rendcommon.h"
+#include "feature/stats/rephist.h"
+#include "feature/nodelist/routerlist.h"
+#include "feature/nodelist/routerset.h"
+#include "core/or/channelpadding.h"
+#include "lib/compress/compress.h"
+#include "lib/compress/compress_lzma.h"
+#include "lib/compress/compress_zlib.h"
+#include "lib/compress/compress_zstd.h"
+#include "lib/container/buffers.h"
+
+#include "ht.h"
+
+#include "core/or/cpath_build_state_st.h"
+#include "core/or/crypt_path_reference_st.h"
+#include "feature/dircommon/dir_connection_st.h"
+#include "core/or/edge_connection_st.h"
+#include "core/or/half_edge_st.h"
+#include "core/or/extend_info_st.h"
+#include "core/or/or_circuit_st.h"
+#include "core/or/origin_circuit_st.h"
+
+/********* START VARIABLES **********/
+
+/** A global list of all circuits at this hop. */
+static smartlist_t *global_circuitlist = NULL;
+
+/** A global list of all origin circuits. Every element of this is also
+ * an element of global_circuitlist. */
+static smartlist_t *global_origin_circuit_list = NULL;
+
+/** A list of all the circuits in CIRCUIT_STATE_CHAN_WAIT. */
+static smartlist_t *circuits_pending_chans = NULL;
+
+/** List of all the (origin) circuits whose state is
+ * CIRCUIT_STATE_GUARD_WAIT. */
+static smartlist_t *circuits_pending_other_guards = NULL;
+
+/** A list of all the circuits that have been marked with
+ * circuit_mark_for_close and which are waiting for circuit_about_to_free. */
+static smartlist_t *circuits_pending_close = NULL;
+
+static void circuit_free_cpath_node(crypt_path_t *victim);
+static void cpath_ref_decref(crypt_path_reference_t *cpath_ref);
+static void circuit_about_to_free_atexit(circuit_t *circ);
+static void circuit_about_to_free(circuit_t *circ);
+
+/**
+ * A cached value of the current state of the origin circuit list. Has the
+ * value 1 if we saw any opened circuits recently (since the last call to
+ * circuit_any_opened_circuits(), which gets called around once a second by
+ * circuit_expire_building). 0 otherwise.
+ */
+static int any_opened_circs_cached_val = 0;
+
+/********* END VARIABLES ************/
+
+or_circuit_t *
+TO_OR_CIRCUIT(circuit_t *x)
+{
+ tor_assert(x->magic == OR_CIRCUIT_MAGIC);
+ return DOWNCAST(or_circuit_t, x);
+}
+const or_circuit_t *
+CONST_TO_OR_CIRCUIT(const circuit_t *x)
+{
+ tor_assert(x->magic == OR_CIRCUIT_MAGIC);
+ return DOWNCAST(or_circuit_t, x);
+}
+origin_circuit_t *
+TO_ORIGIN_CIRCUIT(circuit_t *x)
+{
+ tor_assert(x->magic == ORIGIN_CIRCUIT_MAGIC);
+ return DOWNCAST(origin_circuit_t, x);
+}
+const origin_circuit_t *
+CONST_TO_ORIGIN_CIRCUIT(const circuit_t *x)
+{
+ tor_assert(x->magic == ORIGIN_CIRCUIT_MAGIC);
+ return DOWNCAST(origin_circuit_t, x);
+}
+
+/** A map from channel and circuit ID to circuit. (Lookup performance is
+ * very important here, since we need to do it every time a cell arrives.) */
+typedef struct chan_circid_circuit_map_t {
+ HT_ENTRY(chan_circid_circuit_map_t) node;
+ channel_t *chan;
+ circid_t circ_id;
+ circuit_t *circuit;
+ /* For debugging 12184: when was this placeholder item added? */
+ time_t made_placeholder_at;
+} chan_circid_circuit_map_t;
+
+/** Helper for hash tables: compare the channel and circuit ID for a and
+ * b, and return less than, equal to, or greater than zero appropriately.
+ */
+static inline int
+chan_circid_entries_eq_(chan_circid_circuit_map_t *a,
+ chan_circid_circuit_map_t *b)
+{
+ return a->chan == b->chan && a->circ_id == b->circ_id;
+}
+
+/** Helper: return a hash based on circuit ID and the pointer value of
+ * chan in <b>a</b>. */
+static inline unsigned int
+chan_circid_entry_hash_(chan_circid_circuit_map_t *a)
+{
+ /* Try to squeze the siphash input into 8 bytes to save any extra siphash
+ * rounds. This hash function is in the critical path. */
+ uintptr_t chan = (uintptr_t) (void*) a->chan;
+ uint32_t array[2];
+ array[0] = a->circ_id;
+ /* The low bits of the channel pointer are uninteresting, since the channel
+ * is a pretty big structure. */
+ array[1] = (uint32_t) (chan >> 6);
+ return (unsigned) siphash24g(array, sizeof(array));
+}
+
+/** Map from [chan,circid] to circuit. */
+static HT_HEAD(chan_circid_map, chan_circid_circuit_map_t)
+ chan_circid_map = HT_INITIALIZER();
+HT_PROTOTYPE(chan_circid_map, chan_circid_circuit_map_t, node,
+ chan_circid_entry_hash_, chan_circid_entries_eq_)
+HT_GENERATE2(chan_circid_map, chan_circid_circuit_map_t, node,
+ chan_circid_entry_hash_, chan_circid_entries_eq_, 0.6,
+ tor_reallocarray_, tor_free_)
+
+/** The most recently returned entry from circuit_get_by_circid_chan;
+ * used to improve performance when many cells arrive in a row from the
+ * same circuit.
+ */
+static chan_circid_circuit_map_t *_last_circid_chan_ent = NULL;
+
+/** Implementation helper for circuit_set_{p,n}_circid_channel: A circuit ID
+ * and/or channel for circ has just changed from <b>old_chan, old_id</b>
+ * to <b>chan, id</b>. Adjust the chan,circid map as appropriate, removing
+ * the old entry (if any) and adding a new one. */
+static void
+circuit_set_circid_chan_helper(circuit_t *circ, int direction,
+ circid_t id,
+ channel_t *chan)
+{
+ chan_circid_circuit_map_t search;
+ chan_circid_circuit_map_t *found;
+ channel_t *old_chan, **chan_ptr;
+ circid_t old_id, *circid_ptr;
+ int make_active, attached = 0;
+
+ if (direction == CELL_DIRECTION_OUT) {
+ chan_ptr = &circ->n_chan;
+ circid_ptr = &circ->n_circ_id;
+ make_active = circ->n_chan_cells.n > 0;
+ } else {
+ or_circuit_t *c = TO_OR_CIRCUIT(circ);
+ chan_ptr = &c->p_chan;
+ circid_ptr = &c->p_circ_id;
+ make_active = c->p_chan_cells.n > 0;
+ }
+ old_chan = *chan_ptr;
+ old_id = *circid_ptr;
+
+ if (id == old_id && chan == old_chan)
+ return;
+
+ if (_last_circid_chan_ent &&
+ ((old_id == _last_circid_chan_ent->circ_id &&
+ old_chan == _last_circid_chan_ent->chan) ||
+ (id == _last_circid_chan_ent->circ_id &&
+ chan == _last_circid_chan_ent->chan))) {
+ _last_circid_chan_ent = NULL;
+ }
+
+ if (old_chan) {
+ /*
+ * If we're changing channels or ID and had an old channel and a non
+ * zero old ID and weren't marked for close (i.e., we should have been
+ * attached), detach the circuit. ID changes require this because
+ * circuitmux hashes on (channel_id, circuit_id).
+ */
+ if (old_id != 0 && (old_chan != chan || old_id != id) &&
+ !(circ->marked_for_close)) {
+ tor_assert(old_chan->cmux);
+ circuitmux_detach_circuit(old_chan->cmux, circ);
+ }
+
+ /* we may need to remove it from the conn-circid map */
+ search.circ_id = old_id;
+ search.chan = old_chan;
+ found = HT_REMOVE(chan_circid_map, &chan_circid_map, &search);
+ if (found) {
+ tor_free(found);
+ if (direction == CELL_DIRECTION_OUT) {
+ /* One fewer circuits use old_chan as n_chan */
+ --(old_chan->num_n_circuits);
+ } else {
+ /* One fewer circuits use old_chan as p_chan */
+ --(old_chan->num_p_circuits);
+ }
+ }
+ }
+
+ /* Change the values only after we have possibly made the circuit inactive
+ * on the previous chan. */
+ *chan_ptr = chan;
+ *circid_ptr = id;
+
+ if (chan == NULL)
+ return;
+
+ /* now add the new one to the conn-circid map */
+ search.circ_id = id;
+ search.chan = chan;
+ found = HT_FIND(chan_circid_map, &chan_circid_map, &search);
+ if (found) {
+ found->circuit = circ;
+ found->made_placeholder_at = 0;
+ } else {
+ found = tor_malloc_zero(sizeof(chan_circid_circuit_map_t));
+ found->circ_id = id;
+ found->chan = chan;
+ found->circuit = circ;
+ HT_INSERT(chan_circid_map, &chan_circid_map, found);
+ }
+
+ /*
+ * Attach to the circuitmux if we're changing channels or IDs and
+ * have a new channel and ID to use and the circuit is not marked for
+ * close.
+ */
+ if (chan && id != 0 && (old_chan != chan || old_id != id) &&
+ !(circ->marked_for_close)) {
+ tor_assert(chan->cmux);
+ circuitmux_attach_circuit(chan->cmux, circ, direction);
+ attached = 1;
+ }
+
+ /*
+ * This is a no-op if we have no cells, but if we do it marks us active to
+ * the circuitmux
+ */
+ if (make_active && attached)
+ update_circuit_on_cmux(circ, direction);
+
+ /* Adjust circuit counts on new channel */
+ if (direction == CELL_DIRECTION_OUT) {
+ ++chan->num_n_circuits;
+ } else {
+ ++chan->num_p_circuits;
+ }
+}
+
+/** Mark that circuit id <b>id</b> shouldn't be used on channel <b>chan</b>,
+ * even if there is no circuit on the channel. We use this to keep the
+ * circuit id from getting re-used while we have queued but not yet sent
+ * a destroy cell. */
+void
+channel_mark_circid_unusable(channel_t *chan, circid_t id)
+{
+ chan_circid_circuit_map_t search;
+ chan_circid_circuit_map_t *ent;
+
+ /* See if there's an entry there. That wouldn't be good. */
+ memset(&search, 0, sizeof(search));
+ search.chan = chan;
+ search.circ_id = id;
+ ent = HT_FIND(chan_circid_map, &chan_circid_map, &search);
+
+ if (ent && ent->circuit) {
+ /* we have a problem. */
+ log_warn(LD_BUG, "Tried to mark %u unusable on %p, but there was already "
+ "a circuit there.", (unsigned)id, chan);
+ } else if (ent) {
+ /* It's already marked. */
+ if (!ent->made_placeholder_at)
+ ent->made_placeholder_at = approx_time();
+ } else {
+ ent = tor_malloc_zero(sizeof(chan_circid_circuit_map_t));
+ ent->chan = chan;
+ ent->circ_id = id;
+ /* leave circuit at NULL. */
+ ent->made_placeholder_at = approx_time();
+ HT_INSERT(chan_circid_map, &chan_circid_map, ent);
+ }
+}
+
+/** Mark that a circuit id <b>id</b> can be used again on <b>chan</b>.
+ * We use this to re-enable the circuit ID after we've sent a destroy cell.
+ */
+void
+channel_mark_circid_usable(channel_t *chan, circid_t id)
+{
+ chan_circid_circuit_map_t search;
+ chan_circid_circuit_map_t *ent;
+
+ /* See if there's an entry there. That wouldn't be good. */
+ memset(&search, 0, sizeof(search));
+ search.chan = chan;
+ search.circ_id = id;
+ ent = HT_REMOVE(chan_circid_map, &chan_circid_map, &search);
+ if (ent && ent->circuit) {
+ log_warn(LD_BUG, "Tried to mark %u usable on %p, but there was already "
+ "a circuit there.", (unsigned)id, chan);
+ return;
+ }
+ if (_last_circid_chan_ent == ent)
+ _last_circid_chan_ent = NULL;
+ tor_free(ent);
+}
+
+/** Called to indicate that a DESTROY is pending on <b>chan</b> with
+ * circuit ID <b>id</b>, but hasn't been sent yet. */
+void
+channel_note_destroy_pending(channel_t *chan, circid_t id)
+{
+ circuit_t *circ = circuit_get_by_circid_channel_even_if_marked(id,chan);
+ if (circ) {
+ if (circ->n_chan == chan && circ->n_circ_id == id) {
+ circ->n_delete_pending = 1;
+ } else {
+ or_circuit_t *orcirc = TO_OR_CIRCUIT(circ);
+ if (orcirc->p_chan == chan && orcirc->p_circ_id == id) {
+ circ->p_delete_pending = 1;
+ }
+ }
+ return;
+ }
+ channel_mark_circid_unusable(chan, id);
+}
+
+/** Called to indicate that a DESTROY is no longer pending on <b>chan</b> with
+ * circuit ID <b>id</b> -- typically, because it has been sent. */
+MOCK_IMPL(void,
+channel_note_destroy_not_pending,(channel_t *chan, circid_t id))
+{
+ circuit_t *circ = circuit_get_by_circid_channel_even_if_marked(id,chan);
+ if (circ) {
+ if (circ->n_chan == chan && circ->n_circ_id == id) {
+ circ->n_delete_pending = 0;
+ } else {
+ or_circuit_t *orcirc = TO_OR_CIRCUIT(circ);
+ if (orcirc->p_chan == chan && orcirc->p_circ_id == id) {
+ circ->p_delete_pending = 0;
+ }
+ }
+ /* XXXX this shouldn't happen; log a bug here. */
+ return;
+ }
+ channel_mark_circid_usable(chan, id);
+}
+
+/** Set the p_conn field of a circuit <b>circ</b>, along
+ * with the corresponding circuit ID, and add the circuit as appropriate
+ * to the (chan,id)-\>circuit map. */
+void
+circuit_set_p_circid_chan(or_circuit_t *or_circ, circid_t id,
+ channel_t *chan)
+{
+ circuit_t *circ = TO_CIRCUIT(or_circ);
+ channel_t *old_chan = or_circ->p_chan;
+ circid_t old_id = or_circ->p_circ_id;
+
+ circuit_set_circid_chan_helper(circ, CELL_DIRECTION_IN, id, chan);
+
+ if (chan) {
+ chan->timestamp_last_had_circuits = approx_time();
+ }
+
+ if (circ->p_delete_pending && old_chan) {
+ channel_mark_circid_unusable(old_chan, old_id);
+ circ->p_delete_pending = 0;
+ }
+}
+
+/** Set the n_conn field of a circuit <b>circ</b>, along
+ * with the corresponding circuit ID, and add the circuit as appropriate
+ * to the (chan,id)-\>circuit map. */
+void
+circuit_set_n_circid_chan(circuit_t *circ, circid_t id,
+ channel_t *chan)
+{
+ channel_t *old_chan = circ->n_chan;
+ circid_t old_id = circ->n_circ_id;
+
+ circuit_set_circid_chan_helper(circ, CELL_DIRECTION_OUT, id, chan);
+
+ if (chan) {
+ chan->timestamp_last_had_circuits = approx_time();
+ }
+
+ if (circ->n_delete_pending && old_chan) {
+ channel_mark_circid_unusable(old_chan, old_id);
+ circ->n_delete_pending = 0;
+ }
+}
+
+/** Change the state of <b>circ</b> to <b>state</b>, adding it to or removing
+ * it from lists as appropriate. */
+void
+circuit_set_state(circuit_t *circ, uint8_t state)
+{
+ tor_assert(circ);
+ if (state == circ->state)
+ return;
+ if (PREDICT_UNLIKELY(!circuits_pending_chans))
+ circuits_pending_chans = smartlist_new();
+ if (PREDICT_UNLIKELY(!circuits_pending_other_guards))
+ circuits_pending_other_guards = smartlist_new();
+ if (circ->state == CIRCUIT_STATE_CHAN_WAIT) {
+ /* remove from waiting-circuit list. */
+ smartlist_remove(circuits_pending_chans, circ);
+ }
+ if (state == CIRCUIT_STATE_CHAN_WAIT) {
+ /* add to waiting-circuit list. */
+ smartlist_add(circuits_pending_chans, circ);
+ }
+ if (circ->state == CIRCUIT_STATE_GUARD_WAIT) {
+ smartlist_remove(circuits_pending_other_guards, circ);
+ }
+ if (state == CIRCUIT_STATE_GUARD_WAIT) {
+ smartlist_add(circuits_pending_other_guards, circ);
+ }
+ if (state == CIRCUIT_STATE_GUARD_WAIT || state == CIRCUIT_STATE_OPEN)
+ tor_assert(!circ->n_chan_create_cell);
+ circ->state = state;
+}
+
+/** Append to <b>out</b> all circuits in state CHAN_WAIT waiting for
+ * the given connection. */
+void
+circuit_get_all_pending_on_channel(smartlist_t *out, channel_t *chan)
+{
+ tor_assert(out);
+ tor_assert(chan);
+
+ if (!circuits_pending_chans)
+ return;
+
+ SMARTLIST_FOREACH_BEGIN(circuits_pending_chans, circuit_t *, circ) {
+ if (circ->marked_for_close)
+ continue;
+ if (!circ->n_hop)
+ continue;
+ tor_assert(circ->state == CIRCUIT_STATE_CHAN_WAIT);
+ if (tor_digest_is_zero(circ->n_hop->identity_digest)) {
+ /* Look at addr/port. This is an unkeyed connection. */
+ if (!channel_matches_extend_info(chan, circ->n_hop))
+ continue;
+ } else {
+ /* We expected a key. See if it's the right one. */
+ if (tor_memneq(chan->identity_digest,
+ circ->n_hop->identity_digest, DIGEST_LEN))
+ continue;
+ }
+ smartlist_add(out, circ);
+ } SMARTLIST_FOREACH_END(circ);
+}
+
+/** Return the number of circuits in state CHAN_WAIT, waiting for the given
+ * channel. */
+int
+circuit_count_pending_on_channel(channel_t *chan)
+{
+ int cnt;
+ smartlist_t *sl = smartlist_new();
+
+ tor_assert(chan);
+
+ circuit_get_all_pending_on_channel(sl, chan);
+ cnt = smartlist_len(sl);
+ smartlist_free(sl);
+ log_debug(LD_CIRC,"or_conn to %s, %d pending circs",
+ channel_get_canonical_remote_descr(chan),
+ cnt);
+ return cnt;
+}
+
+/** Remove <b>origin_circ</b> from the global list of origin circuits.
+ * Called when we are freeing a circuit.
+ */
+static void
+circuit_remove_from_origin_circuit_list(origin_circuit_t *origin_circ)
+{
+ int origin_idx = origin_circ->global_origin_circuit_list_idx;
+ if (origin_idx < 0)
+ return;
+ origin_circuit_t *c2;
+ tor_assert(origin_idx <= smartlist_len(global_origin_circuit_list));
+ c2 = smartlist_get(global_origin_circuit_list, origin_idx);
+ tor_assert(origin_circ == c2);
+ smartlist_del(global_origin_circuit_list, origin_idx);
+ if (origin_idx < smartlist_len(global_origin_circuit_list)) {
+ origin_circuit_t *replacement =
+ smartlist_get(global_origin_circuit_list, origin_idx);
+ replacement->global_origin_circuit_list_idx = origin_idx;
+ }
+ origin_circ->global_origin_circuit_list_idx = -1;
+}
+
+/** Add <b>origin_circ</b> to the global list of origin circuits. Called
+ * when creating the circuit. */
+static void
+circuit_add_to_origin_circuit_list(origin_circuit_t *origin_circ)
+{
+ tor_assert(origin_circ->global_origin_circuit_list_idx == -1);
+ smartlist_t *lst = circuit_get_global_origin_circuit_list();
+ smartlist_add(lst, origin_circ);
+ origin_circ->global_origin_circuit_list_idx = smartlist_len(lst) - 1;
+}
+
+/** Detach from the global circuit list, and deallocate, all
+ * circuits that have been marked for close.
+ */
+void
+circuit_close_all_marked(void)
+{
+ if (circuits_pending_close == NULL)
+ return;
+
+ smartlist_t *lst = circuit_get_global_list();
+ SMARTLIST_FOREACH_BEGIN(circuits_pending_close, circuit_t *, circ) {
+ tor_assert(circ->marked_for_close);
+
+ /* Remove it from the circuit list. */
+ int idx = circ->global_circuitlist_idx;
+ smartlist_del(lst, idx);
+ if (idx < smartlist_len(lst)) {
+ circuit_t *replacement = smartlist_get(lst, idx);
+ replacement->global_circuitlist_idx = idx;
+ }
+ circ->global_circuitlist_idx = -1;
+
+ /* Remove it from the origin circuit list, if appropriate. */
+ if (CIRCUIT_IS_ORIGIN(circ)) {
+ circuit_remove_from_origin_circuit_list(TO_ORIGIN_CIRCUIT(circ));
+ }
+
+ circuit_about_to_free(circ);
+ circuit_free(circ);
+ } SMARTLIST_FOREACH_END(circ);
+
+ smartlist_clear(circuits_pending_close);
+}
+
+/** Return a pointer to the global list of circuits. */
+MOCK_IMPL(smartlist_t *,
+circuit_get_global_list,(void))
+{
+ if (NULL == global_circuitlist)
+ global_circuitlist = smartlist_new();
+ return global_circuitlist;
+}
+
+/** Return a pointer to the global list of origin circuits. */
+smartlist_t *
+circuit_get_global_origin_circuit_list(void)
+{
+ if (NULL == global_origin_circuit_list)
+ global_origin_circuit_list = smartlist_new();
+ return global_origin_circuit_list;
+}
+
+/**
+ * Return true if we have any opened general-purpose 3 hop
+ * origin circuits.
+ *
+ * The result from this function is cached for use by
+ * circuit_any_opened_circuits_cached().
+ */
+int
+circuit_any_opened_circuits(void)
+{
+ SMARTLIST_FOREACH_BEGIN(circuit_get_global_origin_circuit_list(),
+ const origin_circuit_t *, next_circ) {
+ if (!TO_CIRCUIT(next_circ)->marked_for_close &&
+ next_circ->has_opened &&
+ TO_CIRCUIT(next_circ)->state == CIRCUIT_STATE_OPEN &&
+ TO_CIRCUIT(next_circ)->purpose != CIRCUIT_PURPOSE_C_MEASURE_TIMEOUT &&
+ next_circ->build_state &&
+ next_circ->build_state->desired_path_len == DEFAULT_ROUTE_LEN) {
+ circuit_cache_opened_circuit_state(1);
+ return 1;
+ }
+ } SMARTLIST_FOREACH_END(next_circ);
+
+ circuit_cache_opened_circuit_state(0);
+ return 0;
+}
+
+/**
+ * Cache the "any circuits opened" state, as specified in param
+ * circuits_are_opened. This is a helper function to update
+ * the circuit opened status whenever we happen to look at the
+ * circuit list.
+ */
+void
+circuit_cache_opened_circuit_state(int circuits_are_opened)
+{
+ any_opened_circs_cached_val = circuits_are_opened;
+}
+
+/**
+ * Return true if there were any opened circuits since the last call to
+ * circuit_any_opened_circuits(), or since circuit_expire_building() last
+ * ran (it runs roughly once per second).
+ */
+int
+circuit_any_opened_circuits_cached(void)
+{
+ return any_opened_circs_cached_val;
+}
+
+/** Function to make circ-\>state human-readable */
+const char *
+circuit_state_to_string(int state)
+{
+ static char buf[64];
+ switch (state) {
+ case CIRCUIT_STATE_BUILDING: return "doing handshakes";
+ case CIRCUIT_STATE_ONIONSKIN_PENDING: return "processing the onion";
+ case CIRCUIT_STATE_CHAN_WAIT: return "connecting to server";
+ case CIRCUIT_STATE_GUARD_WAIT: return "waiting to see how other "
+ "guards perform";
+ case CIRCUIT_STATE_OPEN: return "open";
+ default:
+ log_warn(LD_BUG, "Unknown circuit state %d", state);
+ tor_snprintf(buf, sizeof(buf), "unknown state [%d]", state);
+ return buf;
+ }
+}
+
+/** Map a circuit purpose to a string suitable to be displayed to a
+ * controller. */
+const char *
+circuit_purpose_to_controller_string(uint8_t purpose)
+{
+ static char buf[32];
+ switch (purpose) {
+ case CIRCUIT_PURPOSE_OR:
+ case CIRCUIT_PURPOSE_INTRO_POINT:
+ case CIRCUIT_PURPOSE_REND_POINT_WAITING:
+ case CIRCUIT_PURPOSE_REND_ESTABLISHED:
+ return "SERVER"; /* A controller should never see these, actually. */
+
+ case CIRCUIT_PURPOSE_C_GENERAL:
+ return "GENERAL";
+
+ case CIRCUIT_PURPOSE_C_HSDIR_GET:
+ return "HS_CLIENT_HSDIR";
+
+ case CIRCUIT_PURPOSE_C_INTRODUCING:
+ case CIRCUIT_PURPOSE_C_INTRODUCE_ACK_WAIT:
+ case CIRCUIT_PURPOSE_C_INTRODUCE_ACKED:
+ return "HS_CLIENT_INTRO";
+
+ case CIRCUIT_PURPOSE_C_ESTABLISH_REND:
+ case CIRCUIT_PURPOSE_C_REND_READY:
+ case CIRCUIT_PURPOSE_C_REND_READY_INTRO_ACKED:
+ case CIRCUIT_PURPOSE_C_REND_JOINED:
+ return "HS_CLIENT_REND";
+
+ case CIRCUIT_PURPOSE_S_HSDIR_POST:
+ return "HS_SERVICE_HSDIR";
+
+ case CIRCUIT_PURPOSE_S_ESTABLISH_INTRO:
+ case CIRCUIT_PURPOSE_S_INTRO:
+ return "HS_SERVICE_INTRO";
+
+ case CIRCUIT_PURPOSE_S_CONNECT_REND:
+ case CIRCUIT_PURPOSE_S_REND_JOINED:
+ return "HS_SERVICE_REND";
+
+ case CIRCUIT_PURPOSE_TESTING:
+ return "TESTING";
+ case CIRCUIT_PURPOSE_C_MEASURE_TIMEOUT:
+ return "MEASURE_TIMEOUT";
+ case CIRCUIT_PURPOSE_CONTROLLER:
+ return "CONTROLLER";
+ case CIRCUIT_PURPOSE_PATH_BIAS_TESTING:
+ return "PATH_BIAS_TESTING";
+ case CIRCUIT_PURPOSE_HS_VANGUARDS:
+ return "HS_VANGUARDS";
+
+ default:
+ tor_snprintf(buf, sizeof(buf), "UNKNOWN_%d", (int)purpose);
+ return buf;
+ }
+}
+
+/** Return a string specifying the state of the hidden-service circuit
+ * purpose <b>purpose</b>, or NULL if <b>purpose</b> is not a
+ * hidden-service-related circuit purpose. */
+const char *
+circuit_purpose_to_controller_hs_state_string(uint8_t purpose)
+{
+ switch (purpose)
+ {
+ default:
+ log_fn(LOG_WARN, LD_BUG,
+ "Unrecognized circuit purpose: %d",
+ (int)purpose);
+ tor_fragile_assert();
+ /* fall through */
+
+ case CIRCUIT_PURPOSE_OR:
+ case CIRCUIT_PURPOSE_C_GENERAL:
+ case CIRCUIT_PURPOSE_C_MEASURE_TIMEOUT:
+ case CIRCUIT_PURPOSE_TESTING:
+ case CIRCUIT_PURPOSE_CONTROLLER:
+ case CIRCUIT_PURPOSE_PATH_BIAS_TESTING:
+ case CIRCUIT_PURPOSE_HS_VANGUARDS:
+ return NULL;
+
+ case CIRCUIT_PURPOSE_INTRO_POINT:
+ return "OR_HSSI_ESTABLISHED";
+ case CIRCUIT_PURPOSE_REND_POINT_WAITING:
+ return "OR_HSCR_ESTABLISHED";
+ case CIRCUIT_PURPOSE_REND_ESTABLISHED:
+ return "OR_HS_R_JOINED";
+
+ case CIRCUIT_PURPOSE_C_HSDIR_GET:
+ case CIRCUIT_PURPOSE_C_INTRODUCING:
+ return "HSCI_CONNECTING";
+ case CIRCUIT_PURPOSE_C_INTRODUCE_ACK_WAIT:
+ return "HSCI_INTRO_SENT";
+ case CIRCUIT_PURPOSE_C_INTRODUCE_ACKED:
+ return "HSCI_DONE";
+
+ case CIRCUIT_PURPOSE_C_ESTABLISH_REND:
+ return "HSCR_CONNECTING";
+ case CIRCUIT_PURPOSE_C_REND_READY:
+ return "HSCR_ESTABLISHED_IDLE";
+ case CIRCUIT_PURPOSE_C_REND_READY_INTRO_ACKED:
+ return "HSCR_ESTABLISHED_WAITING";
+ case CIRCUIT_PURPOSE_C_REND_JOINED:
+ return "HSCR_JOINED";
+
+ case CIRCUIT_PURPOSE_S_HSDIR_POST:
+ case CIRCUIT_PURPOSE_S_ESTABLISH_INTRO:
+ return "HSSI_CONNECTING";
+ case CIRCUIT_PURPOSE_S_INTRO:
+ return "HSSI_ESTABLISHED";
+
+ case CIRCUIT_PURPOSE_S_CONNECT_REND:
+ return "HSSR_CONNECTING";
+ case CIRCUIT_PURPOSE_S_REND_JOINED:
+ return "HSSR_JOINED";
+ }
+}
+
+/** Return a human-readable string for the circuit purpose <b>purpose</b>. */
+const char *
+circuit_purpose_to_string(uint8_t purpose)
+{
+ static char buf[32];
+
+ switch (purpose)
+ {
+ case CIRCUIT_PURPOSE_OR:
+ return "Circuit at relay";
+ case CIRCUIT_PURPOSE_INTRO_POINT:
+ return "Acting as intro point";
+ case CIRCUIT_PURPOSE_REND_POINT_WAITING:
+ return "Acting as rendezvous (pending)";
+ case CIRCUIT_PURPOSE_REND_ESTABLISHED:
+ return "Acting as rendezvous (established)";
+ case CIRCUIT_PURPOSE_C_GENERAL:
+ return "General-purpose client";
+ case CIRCUIT_PURPOSE_C_INTRODUCING:
+ return "Hidden service client: Connecting to intro point";
+ case CIRCUIT_PURPOSE_C_INTRODUCE_ACK_WAIT:
+ return "Hidden service client: Waiting for ack from intro point";
+ case CIRCUIT_PURPOSE_C_INTRODUCE_ACKED:
+ return "Hidden service client: Received ack from intro point";
+ case CIRCUIT_PURPOSE_C_ESTABLISH_REND:
+ return "Hidden service client: Establishing rendezvous point";
+ case CIRCUIT_PURPOSE_C_REND_READY:
+ return "Hidden service client: Pending rendezvous point";
+ case CIRCUIT_PURPOSE_C_REND_READY_INTRO_ACKED:
+ return "Hidden service client: Pending rendezvous point (ack received)";
+ case CIRCUIT_PURPOSE_C_REND_JOINED:
+ return "Hidden service client: Active rendezvous point";
+ case CIRCUIT_PURPOSE_C_HSDIR_GET:
+ return "Hidden service client: Fetching HS descriptor";
+
+ case CIRCUIT_PURPOSE_C_MEASURE_TIMEOUT:
+ return "Measuring circuit timeout";
+
+ case CIRCUIT_PURPOSE_S_ESTABLISH_INTRO:
+ return "Hidden service: Establishing introduction point";
+ case CIRCUIT_PURPOSE_S_INTRO:
+ return "Hidden service: Introduction point";
+ case CIRCUIT_PURPOSE_S_CONNECT_REND:
+ return "Hidden service: Connecting to rendezvous point";
+ case CIRCUIT_PURPOSE_S_REND_JOINED:
+ return "Hidden service: Active rendezvous point";
+ case CIRCUIT_PURPOSE_S_HSDIR_POST:
+ return "Hidden service: Uploading HS descriptor";
+
+ case CIRCUIT_PURPOSE_TESTING:
+ return "Testing circuit";
+
+ case CIRCUIT_PURPOSE_CONTROLLER:
+ return "Circuit made by controller";
+
+ case CIRCUIT_PURPOSE_PATH_BIAS_TESTING:
+ return "Path-bias testing circuit";
+
+ case CIRCUIT_PURPOSE_HS_VANGUARDS:
+ return "Hidden service: Pre-built vanguard circuit";
+
+ default:
+ tor_snprintf(buf, sizeof(buf), "UNKNOWN_%d", (int)purpose);
+ return buf;
+ }
+}
+
+/** Pick a reasonable package_window to start out for our circuits.
+ * Originally this was hard-coded at 1000, but now the consensus votes
+ * on the answer. See proposal 168. */
+int32_t
+circuit_initial_package_window(void)
+{
+ int32_t num = networkstatus_get_param(NULL, "circwindow", CIRCWINDOW_START,
+ CIRCWINDOW_START_MIN,
+ CIRCWINDOW_START_MAX);
+ /* If the consensus tells us a negative number, we'd assert. */
+ if (num < 0)
+ num = CIRCWINDOW_START;
+ return num;
+}
+
+/** Initialize the common elements in a circuit_t, and add it to the global
+ * list. */
+static void
+init_circuit_base(circuit_t *circ)
+{
+ tor_gettimeofday(&circ->timestamp_created);
+
+ // Gets reset when we send CREATE_FAST.
+ // circuit_expire_building() expects these to be equal
+ // until the orconn is built.
+ circ->timestamp_began = circ->timestamp_created;
+
+ circ->package_window = circuit_initial_package_window();
+ circ->deliver_window = CIRCWINDOW_START;
+ cell_queue_init(&circ->n_chan_cells);
+
+ smartlist_add(circuit_get_global_list(), circ);
+ circ->global_circuitlist_idx = smartlist_len(circuit_get_global_list()) - 1;
+}
+
+/** If we haven't yet decided on a good timeout value for circuit
+ * building, we close idle circuits aggressively so we can get more
+ * data points. These are the default, min, and max consensus values */
+#define DFLT_IDLE_TIMEOUT_WHILE_LEARNING (3*60)
+#define MIN_IDLE_TIMEOUT_WHILE_LEARNING (10)
+#define MAX_IDLE_TIMEOUT_WHILE_LEARNING (1000*60)
+
+/** Allocate space for a new circuit, initializing with <b>p_circ_id</b>
+ * and <b>p_conn</b>. Add it to the global circuit list.
+ */
+origin_circuit_t *
+origin_circuit_new(void)
+{
+ origin_circuit_t *circ;
+ /* never zero, since a global ID of 0 is treated specially by the
+ * controller */
+ static uint32_t n_circuits_allocated = 1;
+
+ circ = tor_malloc_zero(sizeof(origin_circuit_t));
+ circ->base_.magic = ORIGIN_CIRCUIT_MAGIC;
+
+ circ->next_stream_id = crypto_rand_int(1<<16);
+ circ->global_identifier = n_circuits_allocated++;
+ circ->remaining_relay_early_cells = MAX_RELAY_EARLY_CELLS_PER_CIRCUIT;
+ circ->remaining_relay_early_cells -= crypto_rand_int(2);
+
+ init_circuit_base(TO_CIRCUIT(circ));
+
+ /* Add to origin-list. */
+ circ->global_origin_circuit_list_idx = -1;
+ circuit_add_to_origin_circuit_list(circ);
+
+ circuit_build_times_update_last_circ(get_circuit_build_times_mutable());
+
+ if (! circuit_build_times_disabled(get_options()) &&
+ circuit_build_times_needs_circuits(get_circuit_build_times())) {
+ /* Circuits should be shorter lived if we need more of them
+ * for learning a good build timeout */
+ circ->circuit_idle_timeout =
+ networkstatus_get_param(NULL, "cbtlearntimeout",
+ DFLT_IDLE_TIMEOUT_WHILE_LEARNING,
+ MIN_IDLE_TIMEOUT_WHILE_LEARNING,
+ MAX_IDLE_TIMEOUT_WHILE_LEARNING);
+ } else {
+ // This should always be larger than the current port prediction time
+ // remaining, or else we'll end up with the case where a circuit times out
+ // and another one is built, effectively doubling the timeout window.
+ //
+ // We also randomize it by up to 5% more (ie 5% of 0 to 3600 seconds,
+ // depending on how much circuit prediction time is remaining) so that
+ // we don't close a bunch of unused circuits all at the same time.
+ int prediction_time_remaining =
+ predicted_ports_prediction_time_remaining(time(NULL));
+ circ->circuit_idle_timeout = prediction_time_remaining+1+
+ crypto_rand_int(1+prediction_time_remaining/20);
+
+ if (circ->circuit_idle_timeout <= 0) {
+ log_warn(LD_BUG,
+ "Circuit chose a negative idle timeout of %d based on "
+ "%d seconds of predictive building remaining.",
+ circ->circuit_idle_timeout,
+ prediction_time_remaining);
+ circ->circuit_idle_timeout =
+ networkstatus_get_param(NULL, "cbtlearntimeout",
+ DFLT_IDLE_TIMEOUT_WHILE_LEARNING,
+ MIN_IDLE_TIMEOUT_WHILE_LEARNING,
+ MAX_IDLE_TIMEOUT_WHILE_LEARNING);
+ }
+
+ log_info(LD_CIRC,
+ "Circuit %"PRIu32" chose an idle timeout of %d based on "
+ "%d seconds of predictive building remaining.",
+ (circ->global_identifier),
+ circ->circuit_idle_timeout,
+ prediction_time_remaining);
+ }
+
+ return circ;
+}
+
+/** Allocate a new or_circuit_t, connected to <b>p_chan</b> as
+ * <b>p_circ_id</b>. If <b>p_chan</b> is NULL, the circuit is unattached. */
+or_circuit_t *
+or_circuit_new(circid_t p_circ_id, channel_t *p_chan)
+{
+ /* CircIDs */
+ or_circuit_t *circ;
+
+ circ = tor_malloc_zero(sizeof(or_circuit_t));
+ circ->base_.magic = OR_CIRCUIT_MAGIC;
+
+ if (p_chan)
+ circuit_set_p_circid_chan(circ, p_circ_id, p_chan);
+
+ circ->remaining_relay_early_cells = MAX_RELAY_EARLY_CELLS_PER_CIRCUIT;
+ cell_queue_init(&circ->p_chan_cells);
+
+ init_circuit_base(TO_CIRCUIT(circ));
+
+ return circ;
+}
+
+/** Free all storage held in circ->testing_cell_stats */
+void
+circuit_clear_testing_cell_stats(circuit_t *circ)
+{
+ if (!circ || !circ->testing_cell_stats)
+ return;
+ SMARTLIST_FOREACH(circ->testing_cell_stats, testing_cell_stats_entry_t *,
+ ent, tor_free(ent));
+ smartlist_free(circ->testing_cell_stats);
+ circ->testing_cell_stats = NULL;
+}
+
+/** Deallocate space associated with circ.
+ */
+STATIC void
+circuit_free_(circuit_t *circ)
+{
+ circid_t n_circ_id = 0;
+ void *mem;
+ size_t memlen;
+ int should_free = 1;
+ if (!circ)
+ return;
+
+ /* We keep a copy of this so we can log its value before it gets unset. */
+ n_circ_id = circ->n_circ_id;
+
+ circuit_clear_testing_cell_stats(circ);
+
+ /* Cleanup circuit from anything HS v3 related. We also do this when the
+ * circuit is closed. This is to avoid any code path that free registered
+ * circuits without closing them before. This needs to be done before the
+ * hs identifier is freed. */
+ hs_circ_cleanup(circ);
+
+ if (CIRCUIT_IS_ORIGIN(circ)) {
+ origin_circuit_t *ocirc = TO_ORIGIN_CIRCUIT(circ);
+ mem = ocirc;
+ memlen = sizeof(origin_circuit_t);
+ tor_assert(circ->magic == ORIGIN_CIRCUIT_MAGIC);
+
+ circuit_remove_from_origin_circuit_list(ocirc);
+
+ if (ocirc->half_streams) {
+ SMARTLIST_FOREACH_BEGIN(ocirc->half_streams, half_edge_t*,
+ half_conn) {
+ tor_free(half_conn);
+ } SMARTLIST_FOREACH_END(half_conn);
+ smartlist_free(ocirc->half_streams);
+ }
+
+ if (ocirc->build_state) {
+ extend_info_free(ocirc->build_state->chosen_exit);
+ circuit_free_cpath_node(ocirc->build_state->pending_final_cpath);
+ cpath_ref_decref(ocirc->build_state->service_pending_final_cpath_ref);
+ }
+ tor_free(ocirc->build_state);
+
+ /* Cancel before freeing, if we haven't already succeeded or failed. */
+ if (ocirc->guard_state) {
+ entry_guard_cancel(ô->guard_state);
+ }
+ circuit_guard_state_free(ocirc->guard_state);
+
+ circuit_clear_cpath(ocirc);
+
+ crypto_pk_free(ocirc->intro_key);
+ rend_data_free(ocirc->rend_data);
+
+ /* Finally, free the identifier of the circuit and nullify it so multiple
+ * cleanup will work. */
+ hs_ident_circuit_free(ocirc->hs_ident);
+ ocirc->hs_ident = NULL;
+
+ tor_free(ocirc->dest_address);
+ if (ocirc->socks_username) {
+ memwipe(ocirc->socks_username, 0x12, ocirc->socks_username_len);
+ tor_free(ocirc->socks_username);
+ }
+ if (ocirc->socks_password) {
+ memwipe(ocirc->socks_password, 0x06, ocirc->socks_password_len);
+ tor_free(ocirc->socks_password);
+ }
+ addr_policy_list_free(ocirc->prepend_policy);
+ } else {
+ or_circuit_t *ocirc = TO_OR_CIRCUIT(circ);
+ /* Remember cell statistics for this circuit before deallocating. */
+ if (get_options()->CellStatistics)
+ rep_hist_buffer_stats_add_circ(circ, time(NULL));
+ mem = ocirc;
+ memlen = sizeof(or_circuit_t);
+ tor_assert(circ->magic == OR_CIRCUIT_MAGIC);
+
+ should_free = (ocirc->workqueue_entry == NULL);
+
+ relay_crypto_clear(ô->crypto);
+
+ if (ocirc->rend_splice) {
+ or_circuit_t *other = ocirc->rend_splice;
+ tor_assert(other->base_.magic == OR_CIRCUIT_MAGIC);
+ other->rend_splice = NULL;
+ }
+
+ /* remove from map. */
+ circuit_set_p_circid_chan(ocirc, 0, NULL);
+
+ /* Clear cell queue _after_ removing it from the map. Otherwise our
+ * "active" checks will be violated. */
+ cell_queue_clear(ô->p_chan_cells);
+ }
+
+ extend_info_free(circ->n_hop);
+ tor_free(circ->n_chan_create_cell);
+
+ if (circ->global_circuitlist_idx != -1) {
+ int idx = circ->global_circuitlist_idx;
+ circuit_t *c2 = smartlist_get(global_circuitlist, idx);
+ tor_assert(c2 == circ);
+ smartlist_del(global_circuitlist, idx);
+ if (idx < smartlist_len(global_circuitlist)) {
+ c2 = smartlist_get(global_circuitlist, idx);
+ c2->global_circuitlist_idx = idx;
+ }
+ }
+
+ /* Remove from map. */
+ circuit_set_n_circid_chan(circ, 0, NULL);
+
+ /* Clear cell queue _after_ removing it from the map. Otherwise our
+ * "active" checks will be violated. */
+ cell_queue_clear(&circ->n_chan_cells);
+
+ log_info(LD_CIRC, "Circuit %u (id: %" PRIu32 ") has been freed.",
+ n_circ_id,
+ CIRCUIT_IS_ORIGIN(circ) ?
+ TO_ORIGIN_CIRCUIT(circ)->global_identifier : 0);
+
+ if (should_free) {
+ memwipe(mem, 0xAA, memlen); /* poison memory */
+ tor_free(mem);
+ } else {
+ /* If we made it here, this is an or_circuit_t that still has a pending
+ * cpuworker request which we weren't able to cancel. Instead, set up
+ * the magic value so that when the reply comes back, we'll know to discard
+ * the reply and free this structure.
+ */
+ memwipe(mem, 0xAA, memlen);
+ circ->magic = DEAD_CIRCUIT_MAGIC;
+ }
+}
+
+/** Deallocate the linked list circ-><b>cpath</b>, and remove the cpath from
+ * <b>circ</b>. */
+void
+circuit_clear_cpath(origin_circuit_t *circ)
+{
+ crypt_path_t *victim, *head, *cpath;
+
+ head = cpath = circ->cpath;
+
+ if (!cpath)
+ return;
+
+ /* it's a circular list, so we have to notice when we've
+ * gone through it once. */
+ while (cpath->next && cpath->next != head) {
+ victim = cpath;
+ cpath = victim->next;
+ circuit_free_cpath_node(victim);
+ }
+
+ circuit_free_cpath_node(cpath);
+
+ circ->cpath = NULL;
+}
+
+/** Release all storage held by circuits. */
+void
+circuit_free_all(void)
+{
+ smartlist_t *lst = circuit_get_global_list();
+
+ SMARTLIST_FOREACH_BEGIN(lst, circuit_t *, tmp) {
+ if (! CIRCUIT_IS_ORIGIN(tmp)) {
+ or_circuit_t *or_circ = TO_OR_CIRCUIT(tmp);
+ while (or_circ->resolving_streams) {
+ edge_connection_t *next_conn;
+ next_conn = or_circ->resolving_streams->next_stream;
+ connection_free_(TO_CONN(or_circ->resolving_streams));
+ or_circ->resolving_streams = next_conn;
+ }
+ }
+ tmp->global_circuitlist_idx = -1;
+ circuit_about_to_free_atexit(tmp);
+ circuit_free(tmp);
+ SMARTLIST_DEL_CURRENT(lst, tmp);
+ } SMARTLIST_FOREACH_END(tmp);
+
+ smartlist_free(lst);
+ global_circuitlist = NULL;
+
+ smartlist_free(global_origin_circuit_list);
+ global_origin_circuit_list = NULL;
+
+ smartlist_free(circuits_pending_chans);
+ circuits_pending_chans = NULL;
+
+ smartlist_free(circuits_pending_close);
+ circuits_pending_close = NULL;
+
+ smartlist_free(circuits_pending_other_guards);
+ circuits_pending_other_guards = NULL;
+
+ {
+ chan_circid_circuit_map_t **elt, **next, *c;
+ for (elt = HT_START(chan_circid_map, &chan_circid_map);
+ elt;
+ elt = next) {
+ c = *elt;
+ next = HT_NEXT_RMV(chan_circid_map, &chan_circid_map, elt);
+
+ tor_assert(c->circuit == NULL);
+ tor_free(c);
+ }
+ }
+ HT_CLEAR(chan_circid_map, &chan_circid_map);
+}
+
+/** Deallocate space associated with the cpath node <b>victim</b>. */
+static void
+circuit_free_cpath_node(crypt_path_t *victim)
+{
+ if (!victim)
+ return;
+
+ relay_crypto_clear(&victim->crypto);
+ onion_handshake_state_release(&victim->handshake_state);
+ crypto_dh_free(victim->rend_dh_handshake_state);
+ extend_info_free(victim->extend_info);
+
+ memwipe(victim, 0xBB, sizeof(crypt_path_t)); /* poison memory */
+ tor_free(victim);
+}
+
+/** Release a crypt_path_reference_t*, which may be NULL. */
+static void
+cpath_ref_decref(crypt_path_reference_t *cpath_ref)
+{
+ if (cpath_ref != NULL) {
+ if (--(cpath_ref->refcount) == 0) {
+ circuit_free_cpath_node(cpath_ref->cpath);
+ tor_free(cpath_ref);
+ }
+ }
+}
+
+/** A helper function for circuit_dump_by_conn() below. Log a bunch
+ * of information about circuit <b>circ</b>.
+ */
+static void
+circuit_dump_conn_details(int severity,
+ circuit_t *circ,
+ int conn_array_index,
+ const char *type,
+ circid_t this_circid,
+ circid_t other_circid)
+{
+ tor_log(severity, LD_CIRC, "Conn %d has %s circuit: circID %u "
+ "(other side %u), state %d (%s), born %ld:",
+ conn_array_index, type, (unsigned)this_circid, (unsigned)other_circid,
+ circ->state, circuit_state_to_string(circ->state),
+ (long)circ->timestamp_began.tv_sec);
+ if (CIRCUIT_IS_ORIGIN(circ)) { /* circ starts at this node */
+ circuit_log_path(severity, LD_CIRC, TO_ORIGIN_CIRCUIT(circ));
+ }
+}
+
+/** Log, at severity <b>severity</b>, information about each circuit
+ * that is connected to <b>conn</b>.
+ */
+void
+circuit_dump_by_conn(connection_t *conn, int severity)
+{
+ edge_connection_t *tmpconn;
+
+ SMARTLIST_FOREACH_BEGIN(circuit_get_global_list(), circuit_t *, circ) {
+ circid_t n_circ_id = circ->n_circ_id, p_circ_id = 0;
+
+ if (circ->marked_for_close) {
+ continue;
+ }
+
+ if (!CIRCUIT_IS_ORIGIN(circ)) {
+ p_circ_id = TO_OR_CIRCUIT(circ)->p_circ_id;
+ }
+
+ if (CIRCUIT_IS_ORIGIN(circ)) {
+ for (tmpconn=TO_ORIGIN_CIRCUIT(circ)->p_streams; tmpconn;
+ tmpconn=tmpconn->next_stream) {
+ if (TO_CONN(tmpconn) == conn) {
+ circuit_dump_conn_details(severity, circ, conn->conn_array_index,
+ "App-ward", p_circ_id, n_circ_id);
+ }
+ }
+ }
+
+ if (! CIRCUIT_IS_ORIGIN(circ)) {
+ for (tmpconn=TO_OR_CIRCUIT(circ)->n_streams; tmpconn;
+ tmpconn=tmpconn->next_stream) {
+ if (TO_CONN(tmpconn) == conn) {
+ circuit_dump_conn_details(severity, circ, conn->conn_array_index,
+ "Exit-ward", n_circ_id, p_circ_id);
+ }
+ }
+ }
+ }
+ SMARTLIST_FOREACH_END(circ);
+}
+
+/** Return the circuit whose global ID is <b>id</b>, or NULL if no
+ * such circuit exists. */
+origin_circuit_t *
+circuit_get_by_global_id(uint32_t id)
+{
+ SMARTLIST_FOREACH_BEGIN(circuit_get_global_list(), circuit_t *, circ) {
+ if (CIRCUIT_IS_ORIGIN(circ) &&
+ TO_ORIGIN_CIRCUIT(circ)->global_identifier == id) {
+ if (circ->marked_for_close)
+ return NULL;
+ else
+ return TO_ORIGIN_CIRCUIT(circ);
+ }
+ }
+ SMARTLIST_FOREACH_END(circ);
+ return NULL;
+}
+
+/** Return a circ such that:
+ * - circ-\>n_circ_id or circ-\>p_circ_id is equal to <b>circ_id</b>, and
+ * - circ is attached to <b>chan</b>, either as p_chan or n_chan.
+ * Return NULL if no such circuit exists.
+ *
+ * If <b>found_entry_out</b> is provided, set it to true if we have a
+ * placeholder entry for circid/chan, and leave it unset otherwise.
+ */
+static inline circuit_t *
+circuit_get_by_circid_channel_impl(circid_t circ_id, channel_t *chan,
+ int *found_entry_out)
+{
+ chan_circid_circuit_map_t search;
+ chan_circid_circuit_map_t *found;
+
+ if (_last_circid_chan_ent &&
+ circ_id == _last_circid_chan_ent->circ_id &&
+ chan == _last_circid_chan_ent->chan) {
+ found = _last_circid_chan_ent;
+ } else {
+ search.circ_id = circ_id;
+ search.chan = chan;
+ found = HT_FIND(chan_circid_map, &chan_circid_map, &search);
+ _last_circid_chan_ent = found;
+ }
+ if (found && found->circuit) {
+ log_debug(LD_CIRC,
+ "circuit_get_by_circid_channel_impl() returning circuit %p for"
+ " circ_id %u, channel ID %"PRIu64 " (%p)",
+ found->circuit, (unsigned)circ_id,
+ (chan->global_identifier), chan);
+ if (found_entry_out)
+ *found_entry_out = 1;
+ return found->circuit;
+ }
+
+ log_debug(LD_CIRC,
+ "circuit_get_by_circid_channel_impl() found %s for"
+ " circ_id %u, channel ID %"PRIu64 " (%p)",
+ found ? "placeholder" : "nothing",
+ (unsigned)circ_id,
+ (chan->global_identifier), chan);
+
+ if (found_entry_out)
+ *found_entry_out = found ? 1 : 0;
+
+ return NULL;
+ /* The rest of this checks for bugs. Disabled by default. */
+ /* We comment it out because coverity complains otherwise.
+ {
+ circuit_t *circ;
+ TOR_LIST_FOREACH(circ, &global_circuitlist, head) {
+ if (! CIRCUIT_IS_ORIGIN(circ)) {
+ or_circuit_t *or_circ = TO_OR_CIRCUIT(circ);
+ if (or_circ->p_chan == chan && or_circ->p_circ_id == circ_id) {
+ log_warn(LD_BUG,
+ "circuit matches p_chan, but not in hash table (Bug!)");
+ return circ;
+ }
+ }
+ if (circ->n_chan == chan && circ->n_circ_id == circ_id) {
+ log_warn(LD_BUG,
+ "circuit matches n_chan, but not in hash table (Bug!)");
+ return circ;
+ }
+ }
+ return NULL;
+ } */
+}
+
+/** Return a circ such that:
+ * - circ-\>n_circ_id or circ-\>p_circ_id is equal to <b>circ_id</b>, and
+ * - circ is attached to <b>chan</b>, either as p_chan or n_chan.
+ * - circ is not marked for close.
+ * Return NULL if no such circuit exists.
+ */
+circuit_t *
+circuit_get_by_circid_channel(circid_t circ_id, channel_t *chan)
+{
+ circuit_t *circ = circuit_get_by_circid_channel_impl(circ_id, chan, NULL);
+ if (!circ || circ->marked_for_close)
+ return NULL;
+ else
+ return circ;
+}
+
+/** Return a circ such that:
+ * - circ-\>n_circ_id or circ-\>p_circ_id is equal to <b>circ_id</b>, and
+ * - circ is attached to <b>chan</b>, either as p_chan or n_chan.
+ * Return NULL if no such circuit exists.
+ */
+circuit_t *
+circuit_get_by_circid_channel_even_if_marked(circid_t circ_id,
+ channel_t *chan)
+{
+ return circuit_get_by_circid_channel_impl(circ_id, chan, NULL);
+}
+
+/** Return true iff the circuit ID <b>circ_id</b> is currently used by a
+ * circuit, marked or not, on <b>chan</b>, or if the circ ID is reserved until
+ * a queued destroy cell can be sent.
+ *
+ * (Return 1 if the circuit is present, marked or not; Return 2
+ * if the circuit ID is pending a destroy.)
+ **/
+int
+circuit_id_in_use_on_channel(circid_t circ_id, channel_t *chan)
+{
+ int found = 0;
+ if (circuit_get_by_circid_channel_impl(circ_id, chan, &found) != NULL)
+ return 1;
+ if (found)
+ return 2;
+ return 0;
+}
+
+/** Helper for debugging 12184. Returns the time since which 'circ_id' has
+ * been marked unusable on 'chan'. */
+time_t
+circuit_id_when_marked_unusable_on_channel(circid_t circ_id, channel_t *chan)
+{
+ chan_circid_circuit_map_t search;
+ chan_circid_circuit_map_t *found;
+
+ memset(&search, 0, sizeof(search));
+ search.circ_id = circ_id;
+ search.chan = chan;
+
+ found = HT_FIND(chan_circid_map, &chan_circid_map, &search);
+
+ if (! found || found->circuit)
+ return 0;
+
+ return found->made_placeholder_at;
+}
+
+/** Return the circuit that a given edge connection is using. */
+circuit_t *
+circuit_get_by_edge_conn(edge_connection_t *conn)
+{
+ circuit_t *circ;
+
+ circ = conn->on_circuit;
+ tor_assert(!circ ||
+ (CIRCUIT_IS_ORIGIN(circ) ? circ->magic == ORIGIN_CIRCUIT_MAGIC
+ : circ->magic == OR_CIRCUIT_MAGIC));
+
+ return circ;
+}
+
+/** For each circuit that has <b>chan</b> as n_chan or p_chan, unlink the
+ * circuit from the chan,circid map, and mark it for close if it hasn't
+ * been marked already.
+ */
+void
+circuit_unlink_all_from_channel(channel_t *chan, int reason)
+{
+ smartlist_t *detached = smartlist_new();
+
+/* #define DEBUG_CIRCUIT_UNLINK_ALL */
+
+ channel_unlink_all_circuits(chan, detached);
+
+#ifdef DEBUG_CIRCUIT_UNLINK_ALL
+ {
+ smartlist_t *detached_2 = smartlist_new();
+ int mismatch = 0, badlen = 0;
+
+ SMARTLIST_FOREACH_BEGIN(circuit_get_global_list(), circuit_t *, circ) {
+ if (circ->n_chan == chan ||
+ (!CIRCUIT_IS_ORIGIN(circ) &&
+ TO_OR_CIRCUIT(circ)->p_chan == chan)) {
+ smartlist_add(detached_2, circ);
+ }
+ }
+ SMARTLIST_FOREACH_END(circ);
+
+ if (smartlist_len(detached) != smartlist_len(detached_2)) {
+ log_warn(LD_BUG, "List of detached circuits had the wrong length! "
+ "(got %d, should have gotten %d)",
+ (int)smartlist_len(detached),
+ (int)smartlist_len(detached_2));
+ badlen = 1;
+ }
+ smartlist_sort_pointers(detached);
+ smartlist_sort_pointers(detached_2);
+
+ SMARTLIST_FOREACH(detached, circuit_t *, c,
+ if (c != smartlist_get(detached_2, c_sl_idx))
+ mismatch = 1;
+ );
+
+ if (mismatch)
+ log_warn(LD_BUG, "Mismatch in list of detached circuits.");
+
+ if (badlen || mismatch) {
+ smartlist_free(detached);
+ detached = detached_2;
+ } else {
+ log_notice(LD_CIRC, "List of %d circuits was as expected.",
+ (int)smartlist_len(detached));
+ smartlist_free(detached_2);
+ }
+ }
+#endif /* defined(DEBUG_CIRCUIT_UNLINK_ALL) */
+
+ SMARTLIST_FOREACH_BEGIN(detached, circuit_t *, circ) {
+ int mark = 0;
+ if (circ->n_chan == chan) {
+
+ circuit_set_n_circid_chan(circ, 0, NULL);
+ mark = 1;
+
+ /* If we didn't request this closure, pass the remote
+ * bit to mark_for_close. */
+ if (chan->reason_for_closing != CHANNEL_CLOSE_REQUESTED)
+ reason |= END_CIRC_REASON_FLAG_REMOTE;
+ }
+ if (! CIRCUIT_IS_ORIGIN(circ)) {
+ or_circuit_t *or_circ = TO_OR_CIRCUIT(circ);
+ if (or_circ->p_chan == chan) {
+ circuit_set_p_circid_chan(or_circ, 0, NULL);
+ mark = 1;
+ }
+ }
+ if (!mark) {
+ log_warn(LD_BUG, "Circuit on detached list which I had no reason "
+ "to mark");
+ continue;
+ }
+ if (!circ->marked_for_close)
+ circuit_mark_for_close(circ, reason);
+ } SMARTLIST_FOREACH_END(circ);
+
+ smartlist_free(detached);
+}
+
+/** Return a circ such that
+ * - circ-\>rend_data-\>onion_address is equal to
+ * <b>rend_data</b>-\>onion_address,
+ * - circ-\>rend_data-\>rend_cookie is equal to
+ * <b>rend_data</b>-\>rend_cookie, and
+ * - circ-\>purpose is equal to CIRCUIT_PURPOSE_C_REND_READY.
+ *
+ * Return NULL if no such circuit exists.
+ */
+origin_circuit_t *
+circuit_get_ready_rend_circ_by_rend_data(const rend_data_t *rend_data)
+{
+ SMARTLIST_FOREACH_BEGIN(circuit_get_global_list(), circuit_t *, circ) {
+ if (!circ->marked_for_close &&
+ circ->purpose == CIRCUIT_PURPOSE_C_REND_READY) {
+ origin_circuit_t *ocirc = TO_ORIGIN_CIRCUIT(circ);
+ if (ocirc->rend_data == NULL) {
+ continue;
+ }
+ if (!rend_cmp_service_ids(rend_data_get_address(rend_data),
+ rend_data_get_address(ocirc->rend_data)) &&
+ tor_memeq(ocirc->rend_data->rend_cookie,
+ rend_data->rend_cookie,
+ REND_COOKIE_LEN))
+ return ocirc;
+ }
+ }
+ SMARTLIST_FOREACH_END(circ);
+ return NULL;
+}
+
+/** Return the first service introduction circuit originating from the global
+ * circuit list after <b>start</b> or at the start of the list if <b>start</b>
+ * is NULL. Return NULL if no circuit is found.
+ *
+ * A service introduction point circuit has a purpose of either
+ * CIRCUIT_PURPOSE_S_ESTABLISH_INTRO or CIRCUIT_PURPOSE_S_INTRO. This does not
+ * return a circuit marked for close and its state must be open. */
+origin_circuit_t *
+circuit_get_next_service_intro_circ(origin_circuit_t *start)
+{
+ int idx = 0;
+ smartlist_t *lst = circuit_get_global_list();
+
+ if (start) {
+ idx = TO_CIRCUIT(start)->global_circuitlist_idx + 1;
+ }
+
+ for ( ; idx < smartlist_len(lst); ++idx) {
+ circuit_t *circ = smartlist_get(lst, idx);
+
+ /* Ignore a marked for close circuit or purpose not matching a service
+ * intro point or if the state is not open. */
+ if (circ->marked_for_close || circ->state != CIRCUIT_STATE_OPEN ||
+ (circ->purpose != CIRCUIT_PURPOSE_S_ESTABLISH_INTRO &&
+ circ->purpose != CIRCUIT_PURPOSE_S_INTRO)) {
+ continue;
+ }
+ /* The purposes we are looking for are only for origin circuits so the
+ * following is valid. */
+ return TO_ORIGIN_CIRCUIT(circ);
+ }
+ /* Not found. */
+ return NULL;
+}
+
+/** Return the first service rendezvous circuit originating from the global
+ * circuit list after <b>start</b> or at the start of the list if <b>start</b>
+ * is NULL. Return NULL if no circuit is found.
+ *
+ * A service rendezvous point circuit has a purpose of either
+ * CIRCUIT_PURPOSE_S_CONNECT_REND or CIRCUIT_PURPOSE_S_REND_JOINED. This does
+ * not return a circuit marked for close and its state must be open. */
+origin_circuit_t *
+circuit_get_next_service_rp_circ(origin_circuit_t *start)
+{
+ int idx = 0;
+ smartlist_t *lst = circuit_get_global_list();
+
+ if (start) {
+ idx = TO_CIRCUIT(start)->global_circuitlist_idx + 1;
+ }
+
+ for ( ; idx < smartlist_len(lst); ++idx) {
+ circuit_t *circ = smartlist_get(lst, idx);
+
+ /* Ignore a marked for close circuit or purpose not matching a service
+ * intro point or if the state is not open. */
+ if (circ->marked_for_close || circ->state != CIRCUIT_STATE_OPEN ||
+ (circ->purpose != CIRCUIT_PURPOSE_S_CONNECT_REND &&
+ circ->purpose != CIRCUIT_PURPOSE_S_REND_JOINED)) {
+ continue;
+ }
+ /* The purposes we are looking for are only for origin circuits so the
+ * following is valid. */
+ return TO_ORIGIN_CIRCUIT(circ);
+ }
+ /* Not found. */
+ return NULL;
+}
+
+/** Return the first circuit originating here in global_circuitlist after
+ * <b>start</b> whose purpose is <b>purpose</b>, and where <b>digest</b> (if
+ * set) matches the private key digest of the rend data associated with the
+ * circuit. Return NULL if no circuit is found. If <b>start</b> is NULL,
+ * begin at the start of the list.
+ */
+origin_circuit_t *
+circuit_get_next_by_pk_and_purpose(origin_circuit_t *start,
+ const uint8_t *digest, uint8_t purpose)
+{
+ int idx;
+ smartlist_t *lst = circuit_get_global_list();
+ tor_assert(CIRCUIT_PURPOSE_IS_ORIGIN(purpose));
+ if (start == NULL)
+ idx = 0;
+ else
+ idx = TO_CIRCUIT(start)->global_circuitlist_idx + 1;
+
+ for ( ; idx < smartlist_len(lst); ++idx) {
+ circuit_t *circ = smartlist_get(lst, idx);
+ origin_circuit_t *ocirc;
+
+ if (circ->marked_for_close)
+ continue;
+ if (circ->purpose != purpose)
+ continue;
+ /* At this point we should be able to get a valid origin circuit because
+ * the origin purpose we are looking for matches this circuit. */
+ if (BUG(!CIRCUIT_PURPOSE_IS_ORIGIN(circ->purpose))) {
+ break;
+ }
+ ocirc = TO_ORIGIN_CIRCUIT(circ);
+ if (!digest)
+ return ocirc;
+ if (rend_circuit_pk_digest_eq(ocirc, digest)) {
+ return ocirc;
+ }
+ }
+ return NULL;
+}
+
+/** We might cannibalize this circuit: Return true if its last hop can be used
+ * as a v3 rendezvous point. */
+static int
+circuit_can_be_cannibalized_for_v3_rp(const origin_circuit_t *circ)
+{
+ if (!circ->build_state) {
+ return 0;
+ }
+
+ extend_info_t *chosen_exit = circ->build_state->chosen_exit;
+ if (BUG(!chosen_exit)) {
+ return 0;
+ }
+
+ const node_t *rp_node = node_get_by_id(chosen_exit->identity_digest);
+ if (rp_node) {
+ if (node_supports_v3_rendezvous_point(rp_node)) {
+ return 1;
+ }
+ }
+
+ return 0;
+}
+
+/** We are trying to create a circuit of purpose <b>purpose</b> and we are
+ * looking for cannibalizable circuits. Return the circuit purpose we would be
+ * willing to cannibalize. */
+static uint8_t
+get_circuit_purpose_needed_to_cannibalize(uint8_t purpose)
+{
+ if (circuit_should_use_vanguards(purpose)) {
+ /* If we are using vanguards, then we should only cannibalize vanguard
+ * circuits so that we get the same path construction logic. */
+ return CIRCUIT_PURPOSE_HS_VANGUARDS;
+ } else {
+ /* If no vanguards are used just get a general circuit! */
+ return CIRCUIT_PURPOSE_C_GENERAL;
+ }
+}
+
+/** Return a circuit that is open, is CIRCUIT_PURPOSE_C_GENERAL,
+ * has a timestamp_dirty value of 0, has flags matching the CIRCLAUNCH_*
+ * flags in <b>flags</b>, and if info is defined, does not already use info
+ * as any of its hops; or NULL if no circuit fits this description.
+ *
+ * The <b>purpose</b> argument refers to the purpose of the circuit we want to
+ * create, not the purpose of the circuit we want to cannibalize.
+ *
+ * If !CIRCLAUNCH_NEED_UPTIME, prefer returning non-uptime circuits.
+ *
+ * To "cannibalize" a circuit means to extend it an extra hop, and use it
+ * for some other purpose than we had originally intended. We do this when
+ * we want to perform some low-bandwidth task at a specific relay, and we
+ * would like the circuit to complete as soon as possible. (If we were going
+ * to use a lot of bandwidth, we wouldn't want a circuit with an extra hop.
+ * If we didn't care about circuit completion latency, we would just build
+ * a new circuit.)
+ */
+origin_circuit_t *
+circuit_find_to_cannibalize(uint8_t purpose_to_produce, extend_info_t *info,
+ int flags)
+{
+ origin_circuit_t *best=NULL;
+ int need_uptime = (flags & CIRCLAUNCH_NEED_UPTIME) != 0;
+ int need_capacity = (flags & CIRCLAUNCH_NEED_CAPACITY) != 0;
+ int internal = (flags & CIRCLAUNCH_IS_INTERNAL) != 0;
+ const or_options_t *options = get_options();
+ /* We want the circuit we are trying to cannibalize to have this purpose */
+ int purpose_to_search_for;
+
+ /* Make sure we're not trying to create a onehop circ by
+ * cannibalization. */
+ tor_assert(!(flags & CIRCLAUNCH_ONEHOP_TUNNEL));
+
+ purpose_to_search_for = get_circuit_purpose_needed_to_cannibalize(
+ purpose_to_produce);
+
+ tor_assert_nonfatal(purpose_to_search_for == CIRCUIT_PURPOSE_C_GENERAL ||
+ purpose_to_search_for == CIRCUIT_PURPOSE_HS_VANGUARDS);
+
+ log_debug(LD_CIRC,
+ "Hunting for a circ to cannibalize: purpose %d, uptime %d, "
+ "capacity %d, internal %d",
+ purpose_to_produce, need_uptime, need_capacity, internal);
+
+ SMARTLIST_FOREACH_BEGIN(circuit_get_global_list(), circuit_t *, circ_) {
+ if (CIRCUIT_IS_ORIGIN(circ_) &&
+ circ_->state == CIRCUIT_STATE_OPEN &&
+ !circ_->marked_for_close &&
+ circ_->purpose == purpose_to_search_for &&
+ !circ_->timestamp_dirty) {
+ origin_circuit_t *circ = TO_ORIGIN_CIRCUIT(circ_);
+
+ /* Only cannibalize from reasonable length circuits. If we
+ * want C_GENERAL, then only choose 3 hop circs. If we want
+ * HS_VANGUARDS, only choose 4 hop circs.
+ */
+ if (circ->build_state->desired_path_len !=
+ route_len_for_purpose(purpose_to_search_for, NULL)) {
+ goto next;
+ }
+
+ /* Ignore any circuits for which we can't use the Guard. It is possible
+ * that the Guard was removed from the samepled set after the circuit
+ * was created so avoid using it. */
+ if (!entry_guard_could_succeed(circ->guard_state)) {
+ goto next;
+ }
+
+ if ((!need_uptime || circ->build_state->need_uptime) &&
+ (!need_capacity || circ->build_state->need_capacity) &&
+ (internal == circ->build_state->is_internal) &&
+ !circ->unusable_for_new_conns &&
+ circ->remaining_relay_early_cells &&
+ !circ->build_state->onehop_tunnel &&
+ !circ->isolation_values_set) {
+ if (info) {
+ /* need to make sure we don't duplicate hops */
+ crypt_path_t *hop = circ->cpath;
+ const node_t *ri1 = node_get_by_id(info->identity_digest);
+ do {
+ const node_t *ri2;
+ if (tor_memeq(hop->extend_info->identity_digest,
+ info->identity_digest, DIGEST_LEN))
+ goto next;
+ if (ri1 &&
+ (ri2 = node_get_by_id(hop->extend_info->identity_digest))
+ && nodes_in_same_family(ri1, ri2))
+ goto next;
+ hop=hop->next;
+ } while (hop!=circ->cpath);
+ }
+ if (options->ExcludeNodes) {
+ /* Make sure no existing nodes in the circuit are excluded for
+ * general use. (This may be possible if StrictNodes is 0, and we
+ * thought we needed to use an otherwise excluded node for, say, a
+ * directory operation.) */
+ crypt_path_t *hop = circ->cpath;
+ do {
+ if (routerset_contains_extendinfo(options->ExcludeNodes,
+ hop->extend_info))
+ goto next;
+ hop = hop->next;
+ } while (hop != circ->cpath);
+ }
+
+ if ((flags & CIRCLAUNCH_IS_V3_RP) &&
+ !circuit_can_be_cannibalized_for_v3_rp(circ)) {
+ log_debug(LD_GENERAL, "Skipping uncannibalizable circuit for v3 "
+ "rendezvous point.");
+ goto next;
+ }
+
+ if (!best || (best->build_state->need_uptime && !need_uptime))
+ best = circ;
+ next: ;
+ }
+ }
+ }
+ SMARTLIST_FOREACH_END(circ_);
+ return best;
+}
+
+/**
+ * Check whether any of the origin circuits that are waiting to see if
+ * their guard is good enough to use can be upgraded to "ready". If so,
+ * return a new smartlist containing them. Otherwise return NULL.
+ */
+smartlist_t *
+circuit_find_circuits_to_upgrade_from_guard_wait(void)
+{
+ /* Only if some circuit is actually waiting on an upgrade should we
+ * run the algorithm. */
+ if (! circuits_pending_other_guards ||
+ smartlist_len(circuits_pending_other_guards)==0)
+ return NULL;
+ /* Only if we have some origin circuits should we run the algorithm. */
+ if (!global_origin_circuit_list)
+ return NULL;
+
+ /* Okay; we can pass our circuit list to entrynodes.c.*/
+ smartlist_t *result = smartlist_new();
+ int circuits_upgraded = entry_guards_upgrade_waiting_circuits(
+ get_guard_selection_info(),
+ global_origin_circuit_list,
+ result);
+ if (circuits_upgraded && smartlist_len(result)) {
+ return result;
+ } else {
+ smartlist_free(result);
+ return NULL;
+ }
+}
+
+/** Return the number of hops in circuit's path. If circ has no entries,
+ * or is NULL, returns 0. */
+int
+circuit_get_cpath_len(origin_circuit_t *circ)
+{
+ int n = 0;
+ if (circ && circ->cpath) {
+ crypt_path_t *cpath, *cpath_next = NULL;
+ for (cpath = circ->cpath; cpath_next != circ->cpath; cpath = cpath_next) {
+ cpath_next = cpath->next;
+ ++n;
+ }
+ }
+ return n;
+}
+
+/** Return the number of opened hops in circuit's path.
+ * If circ has no entries, or is NULL, returns 0. */
+int
+circuit_get_cpath_opened_len(const origin_circuit_t *circ)
+{
+ int n = 0;
+ if (circ && circ->cpath) {
+ crypt_path_t *cpath, *cpath_next = NULL;
+ for (cpath = circ->cpath;
+ cpath->state == CPATH_STATE_OPEN
+ && cpath_next != circ->cpath;
+ cpath = cpath_next) {
+ cpath_next = cpath->next;
+ ++n;
+ }
+ }
+ return n;
+}
+
+/** Return the <b>hopnum</b>th hop in <b>circ</b>->cpath, or NULL if there
+ * aren't that many hops in the list. <b>hopnum</b> starts at 1.
+ * Returns NULL if <b>hopnum</b> is 0 or negative. */
+crypt_path_t *
+circuit_get_cpath_hop(origin_circuit_t *circ, int hopnum)
+{
+ if (circ && circ->cpath && hopnum > 0) {
+ crypt_path_t *cpath, *cpath_next = NULL;
+ for (cpath = circ->cpath; cpath_next != circ->cpath; cpath = cpath_next) {
+ cpath_next = cpath->next;
+ if (--hopnum <= 0)
+ return cpath;
+ }
+ }
+ return NULL;
+}
+
+/** Go through the circuitlist; mark-for-close each circuit that starts
+ * at us but has not yet been used. */
+void
+circuit_mark_all_unused_circs(void)
+{
+ SMARTLIST_FOREACH_BEGIN(circuit_get_global_list(), circuit_t *, circ) {
+ if (CIRCUIT_IS_ORIGIN(circ) &&
+ !circ->marked_for_close &&
+ !circ->timestamp_dirty)
+ circuit_mark_for_close(circ, END_CIRC_REASON_FINISHED);
+ }
+ SMARTLIST_FOREACH_END(circ);
+}
+
+/** Go through the circuitlist; for each circuit that starts at us
+ * and is dirty, frob its timestamp_dirty so we won't use it for any
+ * new streams.
+ *
+ * This is useful for letting the user change pseudonyms, so new
+ * streams will not be linkable to old streams.
+ */
+void
+circuit_mark_all_dirty_circs_as_unusable(void)
+{
+ SMARTLIST_FOREACH_BEGIN(circuit_get_global_list(), circuit_t *, circ) {
+ if (CIRCUIT_IS_ORIGIN(circ) &&
+ !circ->marked_for_close &&
+ circ->timestamp_dirty) {
+ mark_circuit_unusable_for_new_conns(TO_ORIGIN_CIRCUIT(circ));
+ }
+ }
+ SMARTLIST_FOREACH_END(circ);
+}
+
++/**
++ * Report any queued cells on or_circuits as written in our bandwidth
++ * totals, for the specified channel direction.
++ *
++ * When we close a circuit or clear its cell queues, we've read
++ * data and recorded those bytes in our read statistics, but we're
++ * not going to write it. This discrepancy can be used by an adversary
++ * to infer information from our public relay statistics and perform
++ * attacks such as guard discovery.
++ *
++ * This function is in the critical path of circuit_mark_for_close().
++ * It must be (and is) O(1)!
++ *
++ * See https://trac.torproject.org/projects/tor/ticket/23512.
++ */
++void
++circuit_synchronize_written_or_bandwidth(const circuit_t *c,
++ circuit_channel_direction_t dir)
++{
++ uint64_t cells;
++ uint64_t cell_size;
++ uint64_t written_sync;
++ const channel_t *chan = NULL;
++ const or_circuit_t *or_circ;
++
++ if (!CIRCUIT_IS_ORCIRC(c))
++ return;
++
++ or_circ = CONST_TO_OR_CIRCUIT(c);
++
++ if (dir == CIRCUIT_N_CHAN) {
++ chan = c->n_chan;
++ cells = c->n_chan_cells.n;
++ } else {
++ chan = or_circ->p_chan;
++ cells = or_circ->p_chan_cells.n;
++ }
++
++ /* If we still know the chan, determine real cell size. Otherwise,
++ * assume it's a wide circid channel */
++ if (chan)
++ cell_size = get_cell_network_size(chan->wide_circ_ids);
++ else
++ cell_size = CELL_MAX_NETWORK_SIZE;
++
++ /* The missing written bytes are the cell counts times their cell
++ * size plus TLS per cell overhead */
++ written_sync = cells*(cell_size+TLS_PER_CELL_OVERHEAD);
++
++ /* Report the missing bytes as written, to avoid asymmetry.
++ * We must use time() for consistency with rephist, even though on
++ * some very old rare platforms, approx_time() may be faster. */
++ rep_hist_note_bytes_written(written_sync, time(NULL));
++}
++
+/** Mark <b>circ</b> to be closed next time we call
+ * circuit_close_all_marked(). Do any cleanup needed:
+ * - If state is onionskin_pending, remove circ from the onion_pending
+ * list.
+ * - If circ isn't open yet: call circuit_build_failed() if we're
+ * the origin.
+ * - If purpose is C_INTRODUCE_ACK_WAIT, report the intro point
+ * failure we just had to the hidden service client module.
+ * - If purpose is C_INTRODUCING and <b>reason</b> isn't TIMEOUT,
+ * report to the hidden service client module that the intro point
+ * we just tried may be unreachable.
+ * - Send appropriate destroys and edge_destroys for conns and
+ * streams attached to circ.
+ * - If circ->rend_splice is set (we are the midpoint of a joined
+ * rendezvous stream), then mark the other circuit to close as well.
+ */
+MOCK_IMPL(void,
+circuit_mark_for_close_, (circuit_t *circ, int reason, int line,
+ const char *file))
+{
+ int orig_reason = reason; /* Passed to the controller */
+ assert_circuit_ok(circ);
+ tor_assert(line);
+ tor_assert(file);
+
+ if (circ->marked_for_close) {
+ log_warn(LD_BUG,
+ "Duplicate call to circuit_mark_for_close at %s:%d"
+ " (first at %s:%d)", file, line,
+ circ->marked_for_close_file, circ->marked_for_close);
+ return;
+ }
+ if (reason == END_CIRC_AT_ORIGIN) {
+ if (!CIRCUIT_IS_ORIGIN(circ)) {
+ log_warn(LD_BUG, "Specified 'at-origin' non-reason for ending circuit, "
+ "but circuit was not at origin. (called %s:%d, purpose=%d)",
+ file, line, circ->purpose);
+ }
+ reason = END_CIRC_REASON_NONE;
+ }
+
+ if (CIRCUIT_IS_ORIGIN(circ)) {
+ if (pathbias_check_close(TO_ORIGIN_CIRCUIT(circ), reason) == -1) {
+ /* Don't close it yet, we need to test it first */
+ return;
+ }
+
+ /* We don't send reasons when closing circuits at the origin. */
+ reason = END_CIRC_REASON_NONE;
+ }
+
++ circuit_synchronize_written_or_bandwidth(circ, CIRCUIT_N_CHAN);
++ circuit_synchronize_written_or_bandwidth(circ, CIRCUIT_P_CHAN);
++
+ if (reason & END_CIRC_REASON_FLAG_REMOTE)
+ reason &= ~END_CIRC_REASON_FLAG_REMOTE;
+
+ if (reason < END_CIRC_REASON_MIN_ || reason > END_CIRC_REASON_MAX_) {
+ if (!(orig_reason & END_CIRC_REASON_FLAG_REMOTE))
+ log_warn(LD_BUG, "Reason %d out of range at %s:%d", reason, file, line);
+ reason = END_CIRC_REASON_NONE;
+ }
+
+ circ->marked_for_close = line;
+ circ->marked_for_close_file = file;
+ circ->marked_for_close_reason = reason;
+ circ->marked_for_close_orig_reason = orig_reason;
+
+ if (!CIRCUIT_IS_ORIGIN(circ)) {
+ or_circuit_t *or_circ = TO_OR_CIRCUIT(circ);
+ if (or_circ->rend_splice) {
+ if (!or_circ->rend_splice->base_.marked_for_close) {
+ /* do this after marking this circuit, to avoid infinite recursion. */
+ circuit_mark_for_close(TO_CIRCUIT(or_circ->rend_splice), reason);
+ }
+ or_circ->rend_splice = NULL;
+ }
+ }
+
+ /* Notify the HS subsystem that this circuit is closing. */
+ hs_circ_cleanup(circ);
+
+ if (circuits_pending_close == NULL)
+ circuits_pending_close = smartlist_new();
+
+ smartlist_add(circuits_pending_close, circ);
+ mainloop_schedule_postloop_cleanup();
+
+ log_info(LD_GENERAL, "Circuit %u (id: %" PRIu32 ") marked for close at "
+ "%s:%d (orig reason: %d, new reason: %d)",
+ circ->n_circ_id,
+ CIRCUIT_IS_ORIGIN(circ) ?
+ TO_ORIGIN_CIRCUIT(circ)->global_identifier : 0,
+ file, line, orig_reason, reason);
+}
+
+/** Called immediately before freeing a marked circuit <b>circ</b> from
+ * circuit_free_all() while shutting down Tor; this is a safe-at-shutdown
+ * version of circuit_about_to_free(). It's important that it at least
+ * do circuitmux_detach_circuit() when appropriate.
+ */
+static void
+circuit_about_to_free_atexit(circuit_t *circ)
+{
+
+ if (circ->n_chan) {
+ circuit_clear_cell_queue(circ, circ->n_chan);
+ circuitmux_detach_circuit(circ->n_chan->cmux, circ);
+ circuit_set_n_circid_chan(circ, 0, NULL);
+ }
+
+ if (! CIRCUIT_IS_ORIGIN(circ)) {
+ or_circuit_t *or_circ = TO_OR_CIRCUIT(circ);
+
+ if (or_circ->p_chan) {
+ circuit_clear_cell_queue(circ, or_circ->p_chan);
+ circuitmux_detach_circuit(or_circ->p_chan->cmux, circ);
+ circuit_set_p_circid_chan(or_circ, 0, NULL);
+ }
+ }
+}
+
+/** Called immediately before freeing a marked circuit <b>circ</b>.
+ * Disconnects the circuit from other data structures, launches events
+ * as appropriate, and performs other housekeeping.
+ */
+static void
+circuit_about_to_free(circuit_t *circ)
+{
+
+ int reason = circ->marked_for_close_reason;
+ int orig_reason = circ->marked_for_close_orig_reason;
+
+ if (circ->state == CIRCUIT_STATE_ONIONSKIN_PENDING) {
+ onion_pending_remove(TO_OR_CIRCUIT(circ));
+ }
+ /* If the circuit ever became OPEN, we sent it to the reputation history
+ * module then. If it isn't OPEN, we send it there now to remember which
+ * links worked and which didn't.
+ */
+ if (circ->state != CIRCUIT_STATE_OPEN &&
+ circ->state != CIRCUIT_STATE_GUARD_WAIT) {
+ if (CIRCUIT_IS_ORIGIN(circ)) {
+ origin_circuit_t *ocirc = TO_ORIGIN_CIRCUIT(circ);
+ circuit_build_failed(ocirc); /* take actions if necessary */
+ }
+ }
+ if (circ->state == CIRCUIT_STATE_CHAN_WAIT) {
+ if (circuits_pending_chans)
+ smartlist_remove(circuits_pending_chans, circ);
+ }
+ if (circuits_pending_other_guards) {
+ smartlist_remove(circuits_pending_other_guards, circ);
+ }
+ if (CIRCUIT_IS_ORIGIN(circ)) {
+ control_event_circuit_status(TO_ORIGIN_CIRCUIT(circ),
+ (circ->state == CIRCUIT_STATE_OPEN ||
+ circ->state == CIRCUIT_STATE_GUARD_WAIT) ?
+ CIRC_EVENT_CLOSED:CIRC_EVENT_FAILED,
+ orig_reason);
+ }
+
+ if (circ->purpose == CIRCUIT_PURPOSE_C_INTRODUCE_ACK_WAIT) {
+ origin_circuit_t *ocirc = TO_ORIGIN_CIRCUIT(circ);
+ int timed_out = (reason == END_CIRC_REASON_TIMEOUT);
+ tor_assert(circ->state == CIRCUIT_STATE_OPEN);
+ tor_assert(ocirc->build_state->chosen_exit);
+ if (orig_reason != END_CIRC_REASON_IP_NOW_REDUNDANT &&
+ ocirc->rend_data) {
+ /* treat this like getting a nack from it */
+ log_info(LD_REND, "Failed intro circ %s to %s (awaiting ack). %s",
+ safe_str_client(rend_data_get_address(ocirc->rend_data)),
+ safe_str_client(build_state_get_exit_nickname(ocirc->build_state)),
+ timed_out ? "Recording timeout." : "Removing from descriptor.");
+ rend_client_report_intro_point_failure(ocirc->build_state->chosen_exit,
+ ocirc->rend_data,
+ timed_out ?
+ INTRO_POINT_FAILURE_TIMEOUT :
+ INTRO_POINT_FAILURE_GENERIC);
+ }
+ } else if (circ->purpose == CIRCUIT_PURPOSE_C_INTRODUCING &&
+ reason != END_CIRC_REASON_TIMEOUT) {
+ origin_circuit_t *ocirc = TO_ORIGIN_CIRCUIT(circ);
+ if (ocirc->build_state->chosen_exit && ocirc->rend_data) {
+ if (orig_reason != END_CIRC_REASON_IP_NOW_REDUNDANT &&
+ ocirc->rend_data) {
+ log_info(LD_REND, "Failed intro circ %s to %s "
+ "(building circuit to intro point). "
+ "Marking intro point as possibly unreachable.",
+ safe_str_client(rend_data_get_address(ocirc->rend_data)),
+ safe_str_client(build_state_get_exit_nickname(
+ ocirc->build_state)));
+ rend_client_report_intro_point_failure(ocirc->build_state->chosen_exit,
+ ocirc->rend_data,
+ INTRO_POINT_FAILURE_UNREACHABLE);
+ }
+ }
+ }
+
+ if (circ->n_chan) {
+ circuit_clear_cell_queue(circ, circ->n_chan);
+ /* Only send destroy if the channel isn't closing anyway */
+ if (!CHANNEL_CONDEMNED(circ->n_chan)) {
+ channel_send_destroy(circ->n_circ_id, circ->n_chan, reason);
+ }
+ circuitmux_detach_circuit(circ->n_chan->cmux, circ);
+ circuit_set_n_circid_chan(circ, 0, NULL);
+ }
+
+ if (! CIRCUIT_IS_ORIGIN(circ)) {
+ or_circuit_t *or_circ = TO_OR_CIRCUIT(circ);
+ edge_connection_t *conn;
+ for (conn=or_circ->n_streams; conn; conn=conn->next_stream)
+ connection_edge_destroy(or_circ->p_circ_id, conn);
+ or_circ->n_streams = NULL;
+
+ while (or_circ->resolving_streams) {
+ conn = or_circ->resolving_streams;
+ or_circ->resolving_streams = conn->next_stream;
+ if (!conn->base_.marked_for_close) {
+ /* The client will see a DESTROY, and infer that the connections
+ * are closing because the circuit is getting torn down. No need
+ * to send an end cell. */
+ conn->edge_has_sent_end = 1;
+ conn->end_reason = END_STREAM_REASON_DESTROY;
+ conn->end_reason |= END_STREAM_REASON_FLAG_ALREADY_SENT_CLOSED;
+ connection_mark_for_close(TO_CONN(conn));
+ }
+ conn->on_circuit = NULL;
+ }
+
+ if (or_circ->p_chan) {
+ circuit_clear_cell_queue(circ, or_circ->p_chan);
+ /* Only send destroy if the channel isn't closing anyway */
+ if (!CHANNEL_CONDEMNED(or_circ->p_chan)) {
+ channel_send_destroy(or_circ->p_circ_id, or_circ->p_chan, reason);
+ }
+ circuitmux_detach_circuit(or_circ->p_chan->cmux, circ);
+ circuit_set_p_circid_chan(or_circ, 0, NULL);
+ }
+ } else {
+ origin_circuit_t *ocirc = TO_ORIGIN_CIRCUIT(circ);
+ edge_connection_t *conn;
+ for (conn=ocirc->p_streams; conn; conn=conn->next_stream)
+ connection_edge_destroy(circ->n_circ_id, conn);
+ ocirc->p_streams = NULL;
+ }
+}
+
+/** Given a marked circuit <b>circ</b>, aggressively free its cell queues to
+ * recover memory. */
+static void
+marked_circuit_free_cells(circuit_t *circ)
+{
+ if (!circ->marked_for_close) {
+ log_warn(LD_BUG, "Called on non-marked circuit");
+ return;
+ }
+ cell_queue_clear(&circ->n_chan_cells);
+ if (circ->n_mux)
+ circuitmux_clear_num_cells(circ->n_mux, circ);
+ if (! CIRCUIT_IS_ORIGIN(circ)) {
+ or_circuit_t *orcirc = TO_OR_CIRCUIT(circ);
+ cell_queue_clear(&orcirc->p_chan_cells);
+ if (orcirc->p_mux)
+ circuitmux_clear_num_cells(orcirc->p_mux, circ);
+ }
+}
+
+static size_t
+single_conn_free_bytes(connection_t *conn)
+{
+ size_t result = 0;
+ if (conn->inbuf) {
+ result += buf_allocation(conn->inbuf);
+ buf_clear(conn->inbuf);
+ }
+ if (conn->outbuf) {
+ result += buf_allocation(conn->outbuf);
+ buf_clear(conn->outbuf);
+ conn->outbuf_flushlen = 0;
+ }
+ if (conn->type == CONN_TYPE_DIR) {
+ dir_connection_t *dir_conn = TO_DIR_CONN(conn);
+ if (dir_conn->compress_state) {
+ result += tor_compress_state_size(dir_conn->compress_state);
+ tor_compress_free(dir_conn->compress_state);
+ dir_conn->compress_state = NULL;
+ }
+ }
+ return result;
+}
+
+/** Aggressively free buffer contents on all the buffers of all streams in the
+ * list starting at <b>stream</b>. Return the number of bytes recovered. */
+static size_t
+marked_circuit_streams_free_bytes(edge_connection_t *stream)
+{
+ size_t result = 0;
+ for ( ; stream; stream = stream->next_stream) {
+ connection_t *conn = TO_CONN(stream);
+ result += single_conn_free_bytes(conn);
+ if (conn->linked_conn) {
+ result += single_conn_free_bytes(conn->linked_conn);
+ }
+ }
+ return result;
+}
+
+/** Aggressively free buffer contents on all the buffers of all streams on
+ * circuit <b>c</b>. Return the number of bytes recovered. */
+static size_t
+marked_circuit_free_stream_bytes(circuit_t *c)
+{
+ if (CIRCUIT_IS_ORIGIN(c)) {
+ return marked_circuit_streams_free_bytes(TO_ORIGIN_CIRCUIT(c)->p_streams);
+ } else {
+ return marked_circuit_streams_free_bytes(TO_OR_CIRCUIT(c)->n_streams);
+ }
+}
+
+/** Return the number of cells used by the circuit <b>c</b>'s cell queues. */
+STATIC size_t
+n_cells_in_circ_queues(const circuit_t *c)
+{
+ size_t n = c->n_chan_cells.n;
+ if (! CIRCUIT_IS_ORIGIN(c)) {
+ circuit_t *cc = (circuit_t *) c;
+ n += TO_OR_CIRCUIT(cc)->p_chan_cells.n;
+ }
+ return n;
+}
+
+/**
+ * Return the age of the oldest cell queued on <b>c</b>, in timestamp units.
+ * Return 0 if there are no cells queued on c. Requires that <b>now</b> be
+ * the current coarse timestamp.
+ *
+ * This function will return incorrect results if the oldest cell queued on
+ * the circuit is older than about 2**32 msec (about 49 days) old.
+ */
+STATIC uint32_t
+circuit_max_queued_cell_age(const circuit_t *c, uint32_t now)
+{
+ uint32_t age = 0;
+ packed_cell_t *cell;
+
+ if (NULL != (cell = TOR_SIMPLEQ_FIRST(&c->n_chan_cells.head)))
+ age = now - cell->inserted_timestamp;
+
+ if (! CIRCUIT_IS_ORIGIN(c)) {
+ const or_circuit_t *orcirc = CONST_TO_OR_CIRCUIT(c);
+ if (NULL != (cell = TOR_SIMPLEQ_FIRST(&orcirc->p_chan_cells.head))) {
+ uint32_t age2 = now - cell->inserted_timestamp;
+ if (age2 > age)
+ return age2;
+ }
+ }
+ return age;
+}
+
+/** Return the age of the oldest buffer chunk on <b>conn</b>, where age is
+ * taken in timestamp units before the time <b>now</b>. If the connection has
+ * no data, treat it as having age zero.
+ **/
+static uint32_t
+conn_get_buffer_age(const connection_t *conn, uint32_t now_ts)
+{
+ uint32_t age = 0, age2;
+ if (conn->outbuf) {
+ age2 = buf_get_oldest_chunk_timestamp(conn->outbuf, now_ts);
+ if (age2 > age)
+ age = age2;
+ }
+ if (conn->inbuf) {
+ age2 = buf_get_oldest_chunk_timestamp(conn->inbuf, now_ts);
+ if (age2 > age)
+ age = age2;
+ }
+ return age;
+}
+
+/** Return the age in timestamp units of the oldest buffer chunk on any stream
+ * in the linked list <b>stream</b>, where age is taken in timestamp units
+ * before the timestamp <b>now</b>. */
+static uint32_t
+circuit_get_streams_max_data_age(const edge_connection_t *stream, uint32_t now)
+{
+ uint32_t age = 0, age2;
+ for (; stream; stream = stream->next_stream) {
+ const connection_t *conn = TO_CONN(stream);
+ age2 = conn_get_buffer_age(conn, now);
+ if (age2 > age)
+ age = age2;
+ if (conn->linked_conn) {
+ age2 = conn_get_buffer_age(conn->linked_conn, now);
+ if (age2 > age)
+ age = age2;
+ }
+ }
+ return age;
+}
+
+/** Return the age in timestamp units of the oldest buffer chunk on any stream
+ * attached to the circuit <b>c</b>, where age is taken before the timestamp
+ * <b>now</b>. */
+STATIC uint32_t
+circuit_max_queued_data_age(const circuit_t *c, uint32_t now)
+{
+ if (CIRCUIT_IS_ORIGIN(c)) {
+ return circuit_get_streams_max_data_age(
+ CONST_TO_ORIGIN_CIRCUIT(c)->p_streams, now);
+ } else {
+ return circuit_get_streams_max_data_age(
+ CONST_TO_OR_CIRCUIT(c)->n_streams, now);
+ }
+}
+
+/** Return the age of the oldest cell or stream buffer chunk on the circuit
+ * <b>c</b>, where age is taken in timestamp units before the timestamp
+ * <b>now</b> */
+STATIC uint32_t
+circuit_max_queued_item_age(const circuit_t *c, uint32_t now)
+{
+ uint32_t cell_age = circuit_max_queued_cell_age(c, now);
+ uint32_t data_age = circuit_max_queued_data_age(c, now);
+ if (cell_age > data_age)
+ return cell_age;
+ else
+ return data_age;
+}
+
+/** Helper to sort a list of circuit_t by age of oldest item, in descending
+ * order. */
+static int
+circuits_compare_by_oldest_queued_item_(const void **a_, const void **b_)
+{
+ const circuit_t *a = *a_;
+ const circuit_t *b = *b_;
+ uint32_t age_a = a->age_tmp;
+ uint32_t age_b = b->age_tmp;
+
+ if (age_a < age_b)
+ return 1;
+ else if (age_a == age_b)
+ return 0;
+ else
+ return -1;
+}
+
+static uint32_t now_ts_for_buf_cmp;
+
+/** Helper to sort a list of circuit_t by age of oldest item, in descending
+ * order. */
+static int
+conns_compare_by_buffer_age_(const void **a_, const void **b_)
+{
+ const connection_t *a = *a_;
+ const connection_t *b = *b_;
+ time_t age_a = conn_get_buffer_age(a, now_ts_for_buf_cmp);
+ time_t age_b = conn_get_buffer_age(b, now_ts_for_buf_cmp);
+
+ if (age_a < age_b)
+ return 1;
+ else if (age_a == age_b)
+ return 0;
+ else
+ return -1;
+}
+
+#define FRACTION_OF_DATA_TO_RETAIN_ON_OOM 0.90
+
+/** We're out of memory for cells, having allocated <b>current_allocation</b>
+ * bytes' worth. Kill the 'worst' circuits until we're under
+ * FRACTION_OF_DATA_TO_RETAIN_ON_OOM of our maximum usage. */
+void
+circuits_handle_oom(size_t current_allocation)
+{
+ smartlist_t *circlist;
+ smartlist_t *connection_array = get_connection_array();
+ int conn_idx;
+ size_t mem_to_recover;
+ size_t mem_recovered=0;
+ int n_circuits_killed=0;
+ int n_dirconns_killed=0;
+ uint32_t now_ts;
+ log_notice(LD_GENERAL, "We're low on memory (cell queues total alloc:"
+ " %"TOR_PRIuSZ" buffer total alloc: %" TOR_PRIuSZ ","
+ " tor compress total alloc: %" TOR_PRIuSZ
+ " (zlib: %" TOR_PRIuSZ ", zstd: %" TOR_PRIuSZ ","
+ " lzma: %" TOR_PRIuSZ "),"
+ " rendezvous cache total alloc: %" TOR_PRIuSZ "). Killing"
+ " circuits withover-long queues. (This behavior is controlled by"
+ " MaxMemInQueues.)",
+ cell_queues_get_total_allocation(),
+ buf_get_total_allocation(),
+ tor_compress_get_total_allocation(),
+ tor_zlib_get_total_allocation(),
+ tor_zstd_get_total_allocation(),
+ tor_lzma_get_total_allocation(),
+ rend_cache_get_total_allocation());
+
+ {
+ size_t mem_target = (size_t)(get_options()->MaxMemInQueues *
+ FRACTION_OF_DATA_TO_RETAIN_ON_OOM);
+ if (current_allocation <= mem_target)
+ return;
+ mem_to_recover = current_allocation - mem_target;
+ }
+
+ now_ts = monotime_coarse_get_stamp();
+
+ circlist = circuit_get_global_list();
+ SMARTLIST_FOREACH_BEGIN(circlist, circuit_t *, circ) {
+ circ->age_tmp = circuit_max_queued_item_age(circ, now_ts);
+ } SMARTLIST_FOREACH_END(circ);
+
+ /* This is O(n log n); there are faster algorithms we could use instead.
+ * Let's hope this doesn't happen enough to be in the critical path. */
+ smartlist_sort(circlist, circuits_compare_by_oldest_queued_item_);
+
+ /* Fix up the indices before we run into trouble */
+ SMARTLIST_FOREACH_BEGIN(circlist, circuit_t *, circ) {
+ circ->global_circuitlist_idx = circ_sl_idx;
+ } SMARTLIST_FOREACH_END(circ);
+
+ /* Now sort the connection array ... */
+ now_ts_for_buf_cmp = now_ts;
+ smartlist_sort(connection_array, conns_compare_by_buffer_age_);
+ now_ts_for_buf_cmp = 0;
+
+ /* Fix up the connection array to its new order. */
+ SMARTLIST_FOREACH_BEGIN(connection_array, connection_t *, conn) {
+ conn->conn_array_index = conn_sl_idx;
+ } SMARTLIST_FOREACH_END(conn);
+
+ /* Okay, now the worst circuits and connections are at the front of their
+ * respective lists. Let's mark them, and reclaim their storage
+ * aggressively. */
+ conn_idx = 0;
+ SMARTLIST_FOREACH_BEGIN(circlist, circuit_t *, circ) {
+ size_t n;
+ size_t freed;
+
+ /* Free storage in any non-linked directory connections that have buffered
+ * data older than this circuit. */
+ while (conn_idx < smartlist_len(connection_array)) {
+ connection_t *conn = smartlist_get(connection_array, conn_idx);
+ uint32_t conn_age = conn_get_buffer_age(conn, now_ts);
+ if (conn_age < circ->age_tmp) {
+ break;
+ }
+ if (conn->type == CONN_TYPE_DIR && conn->linked_conn == NULL) {
+ if (!conn->marked_for_close)
+ connection_mark_for_close(conn);
+ mem_recovered += single_conn_free_bytes(conn);
+
+ ++n_dirconns_killed;
+
+ if (mem_recovered >= mem_to_recover)
+ goto done_recovering_mem;
+ }
+ ++conn_idx;
+ }
+
+ /* Now, kill the circuit. */
+ n = n_cells_in_circ_queues(circ);
+ if (! circ->marked_for_close) {
+ circuit_mark_for_close(circ, END_CIRC_REASON_RESOURCELIMIT);
+ }
+ marked_circuit_free_cells(circ);
+ freed = marked_circuit_free_stream_bytes(circ);
+
+ ++n_circuits_killed;
+
+ mem_recovered += n * packed_cell_mem_cost();
+ mem_recovered += freed;
+
+ if (mem_recovered >= mem_to_recover)
+ goto done_recovering_mem;
+ } SMARTLIST_FOREACH_END(circ);
+
+ done_recovering_mem:
+
+ log_notice(LD_GENERAL, "Removed %"TOR_PRIuSZ" bytes by killing %d circuits; "
+ "%d circuits remain alive. Also killed %d non-linked directory "
+ "connections.",
+ mem_recovered,
+ n_circuits_killed,
+ smartlist_len(circlist) - n_circuits_killed,
+ n_dirconns_killed);
+}
+
+/** Verify that cpath layer <b>cp</b> has all of its invariants
+ * correct. Trigger an assert if anything is invalid.
+ */
+void
+assert_cpath_layer_ok(const crypt_path_t *cp)
+{
+// tor_assert(cp->addr); /* these are zero for rendezvous extra-hops */
+// tor_assert(cp->port);
+ tor_assert(cp);
+ tor_assert(cp->magic == CRYPT_PATH_MAGIC);
+ switch (cp->state)
+ {
+ case CPATH_STATE_OPEN:
+ relay_crypto_assert_ok(&cp->crypto);
+ /* fall through */
+ case CPATH_STATE_CLOSED:
+ /*XXXX Assert that there's no handshake_state either. */
+ tor_assert(!cp->rend_dh_handshake_state);
+ break;
+ case CPATH_STATE_AWAITING_KEYS:
+ /* tor_assert(cp->dh_handshake_state); */
+ break;
+ default:
+ log_fn(LOG_ERR, LD_BUG, "Unexpected state %d", cp->state);
+ tor_assert(0);
+ }
+ tor_assert(cp->package_window >= 0);
+ tor_assert(cp->deliver_window >= 0);
+}
+
+/** Verify that cpath <b>cp</b> has all of its invariants
+ * correct. Trigger an assert if anything is invalid.
+ */
+static void
+assert_cpath_ok(const crypt_path_t *cp)
+{
+ const crypt_path_t *start = cp;
+
+ do {
+ assert_cpath_layer_ok(cp);
+ /* layers must be in sequence of: "open* awaiting? closed*" */
+ if (cp != start) {
+ if (cp->state == CPATH_STATE_AWAITING_KEYS) {
+ tor_assert(cp->prev->state == CPATH_STATE_OPEN);
+ } else if (cp->state == CPATH_STATE_OPEN) {
+ tor_assert(cp->prev->state == CPATH_STATE_OPEN);
+ }
+ }
+ cp = cp->next;
+ tor_assert(cp);
+ } while (cp != start);
+}
+
+/** Verify that circuit <b>c</b> has all of its invariants
+ * correct. Trigger an assert if anything is invalid.
+ */
+MOCK_IMPL(void,
+assert_circuit_ok,(const circuit_t *c))
+{
+ edge_connection_t *conn;
+ const or_circuit_t *or_circ = NULL;
+ const origin_circuit_t *origin_circ = NULL;
+
+ tor_assert(c);
+ tor_assert(c->magic == ORIGIN_CIRCUIT_MAGIC || c->magic == OR_CIRCUIT_MAGIC);
+ tor_assert(c->purpose >= CIRCUIT_PURPOSE_MIN_ &&
+ c->purpose <= CIRCUIT_PURPOSE_MAX_);
+
+ if (CIRCUIT_IS_ORIGIN(c))
+ origin_circ = CONST_TO_ORIGIN_CIRCUIT(c);
+ else
+ or_circ = CONST_TO_OR_CIRCUIT(c);
+
+ if (c->n_chan) {
+ tor_assert(!c->n_hop);
+
+ if (c->n_circ_id) {
+ /* We use the _impl variant here to make sure we don't fail on marked
+ * circuits, which would not be returned by the regular function. */
+ circuit_t *c2 = circuit_get_by_circid_channel_impl(c->n_circ_id,
+ c->n_chan, NULL);
+ tor_assert(c == c2);
+ }
+ }
+ if (or_circ && or_circ->p_chan) {
+ if (or_circ->p_circ_id) {
+ /* ibid */
+ circuit_t *c2 =
+ circuit_get_by_circid_channel_impl(or_circ->p_circ_id,
+ or_circ->p_chan, NULL);
+ tor_assert(c == c2);
+ }
+ }
+ if (or_circ)
+ for (conn = or_circ->n_streams; conn; conn = conn->next_stream)
+ tor_assert(conn->base_.type == CONN_TYPE_EXIT);
+
+ tor_assert(c->deliver_window >= 0);
+ tor_assert(c->package_window >= 0);
+ if (c->state == CIRCUIT_STATE_OPEN ||
+ c->state == CIRCUIT_STATE_GUARD_WAIT) {
+ tor_assert(!c->n_chan_create_cell);
+ if (or_circ) {
+ relay_crypto_assert_ok(&or_circ->crypto);
+ }
+ }
+ if (c->state == CIRCUIT_STATE_CHAN_WAIT && !c->marked_for_close) {
+ tor_assert(circuits_pending_chans &&
+ smartlist_contains(circuits_pending_chans, c));
+ } else {
+ tor_assert(!circuits_pending_chans ||
+ !smartlist_contains(circuits_pending_chans, c));
+ }
+ if (origin_circ && origin_circ->cpath) {
+ assert_cpath_ok(origin_circ->cpath);
+ }
+ if (c->purpose == CIRCUIT_PURPOSE_REND_ESTABLISHED) {
+ tor_assert(or_circ);
+ if (!c->marked_for_close) {
+ tor_assert(or_circ->rend_splice);
+ tor_assert(or_circ->rend_splice->rend_splice == or_circ);
+ }
+ tor_assert(or_circ->rend_splice != or_circ);
+ } else {
+ tor_assert(!or_circ || !or_circ->rend_splice);
+ }
+}
diff --cc src/core/or/circuitlist.h
index b069604a1,000000000..dac11431c
mode 100644,000000..100644
--- a/src/core/or/circuitlist.h
+++ b/src/core/or/circuitlist.h
@@@ -1,247 -1,0 +1,249 @@@
+/* Copyright (c) 2001 Matej Pfajfar.
+ * Copyright (c) 2001-2004, Roger Dingledine.
+ * Copyright (c) 2004-2006, Roger Dingledine, Nick Mathewson.
+ * Copyright (c) 2007-2018, The Tor Project, Inc. */
+/* See LICENSE for licensing information */
+
+/**
+ * \file circuitlist.h
+ * \brief Header file for circuitlist.c.
+ **/
+
+#ifndef TOR_CIRCUITLIST_H
+#define TOR_CIRCUITLIST_H
+
+#include "lib/testsupport/testsupport.h"
+#include "feature/hs/hs_ident.h"
+
+/** Circuit state: I'm the origin, still haven't done all my handshakes. */
+#define CIRCUIT_STATE_BUILDING 0
+/** Circuit state: Waiting to process the onionskin. */
+#define CIRCUIT_STATE_ONIONSKIN_PENDING 1
+/** Circuit state: I'd like to deliver a create, but my n_chan is still
+ * connecting. */
+#define CIRCUIT_STATE_CHAN_WAIT 2
+/** Circuit state: the circuit is open but we don't want to actually use it
+ * until we find out if a better guard will be available.
+ */
+#define CIRCUIT_STATE_GUARD_WAIT 3
+/** Circuit state: onionskin(s) processed, ready to send/receive cells. */
+#define CIRCUIT_STATE_OPEN 4
+
+#define CIRCUIT_PURPOSE_MIN_ 1
+
+/* these circuits were initiated elsewhere */
+#define CIRCUIT_PURPOSE_OR_MIN_ 1
+/** OR-side circuit purpose: normal circuit, at OR. */
+#define CIRCUIT_PURPOSE_OR 1
+/** OR-side circuit purpose: At OR, from the service, waiting for intro from
+ * clients. */
+#define CIRCUIT_PURPOSE_INTRO_POINT 2
+/** OR-side circuit purpose: At OR, from the client, waiting for the service.
+ */
+#define CIRCUIT_PURPOSE_REND_POINT_WAITING 3
+/** OR-side circuit purpose: At OR, both circuits have this purpose. */
+#define CIRCUIT_PURPOSE_REND_ESTABLISHED 4
+#define CIRCUIT_PURPOSE_OR_MAX_ 4
+
+/* these circuits originate at this node */
+
+/* here's how circ client-side purposes work:
+ * normal circuits are C_GENERAL.
+ * circuits that are c_introducing are either on their way to
+ * becoming open, or they are open and waiting for a
+ * suitable rendcirc before they send the intro.
+ * circuits that are c_introduce_ack_wait have sent the intro,
+ * but haven't gotten a response yet.
+ * circuits that are c_establish_rend are either on their way
+ * to becoming open, or they are open and have sent the
+ * establish_rendezvous cell but haven't received an ack.
+ * circuits that are c_rend_ready are open and have received a
+ * rend ack, but haven't heard from the service yet. if they have a
+ * buildstate->pending_final_cpath then they're expecting a
+ * cell from the service, else they're not.
+ * circuits that are c_rend_ready_intro_acked are open, and
+ * some intro circ has sent its intro and received an ack.
+ * circuits that are c_rend_joined are open, have heard from
+ * the service, and are talking to it.
+ */
+/** Client-side circuit purpose: Normal circuit, with cpath. */
+#define CIRCUIT_PURPOSE_C_GENERAL 5
+#define CIRCUIT_PURPOSE_C_HS_MIN_ 6
+/** Client-side circuit purpose: at the client, connecting to intro point. */
+#define CIRCUIT_PURPOSE_C_INTRODUCING 6
+/** Client-side circuit purpose: at the client, sent INTRODUCE1 to intro point,
+ * waiting for ACK/NAK. */
+#define CIRCUIT_PURPOSE_C_INTRODUCE_ACK_WAIT 7
+/** Client-side circuit purpose: at the client, introduced and acked, closing.
+ */
+#define CIRCUIT_PURPOSE_C_INTRODUCE_ACKED 8
+/** Client-side circuit purpose: at the client, waiting for ack. */
+#define CIRCUIT_PURPOSE_C_ESTABLISH_REND 9
+/** Client-side circuit purpose: at the client, waiting for the service. */
+#define CIRCUIT_PURPOSE_C_REND_READY 10
+/** Client-side circuit purpose: at the client, waiting for the service,
+ * INTRODUCE has been acknowledged. */
+#define CIRCUIT_PURPOSE_C_REND_READY_INTRO_ACKED 11
+/** Client-side circuit purpose: at the client, rendezvous established. */
+#define CIRCUIT_PURPOSE_C_REND_JOINED 12
+/** This circuit is used for getting hsdirs */
+#define CIRCUIT_PURPOSE_C_HSDIR_GET 13
+#define CIRCUIT_PURPOSE_C_HS_MAX_ 13
+/** This circuit is used for build time measurement only */
+#define CIRCUIT_PURPOSE_C_MEASURE_TIMEOUT 14
+#define CIRCUIT_PURPOSE_C_MAX_ 14
+
+#define CIRCUIT_PURPOSE_S_HS_MIN_ 15
+/** Hidden-service-side circuit purpose: at the service, waiting for
+ * introductions. */
+#define CIRCUIT_PURPOSE_S_ESTABLISH_INTRO 15
+/** Hidden-service-side circuit purpose: at the service, successfully
+ * established intro. */
+#define CIRCUIT_PURPOSE_S_INTRO 16
+/** Hidden-service-side circuit purpose: at the service, connecting to rend
+ * point. */
+#define CIRCUIT_PURPOSE_S_CONNECT_REND 17
+/** Hidden-service-side circuit purpose: at the service, rendezvous
+ * established. */
+#define CIRCUIT_PURPOSE_S_REND_JOINED 18
+/** This circuit is used for uploading hsdirs */
+#define CIRCUIT_PURPOSE_S_HSDIR_POST 19
+#define CIRCUIT_PURPOSE_S_HS_MAX_ 19
+
+/** A testing circuit; not meant to be used for actual traffic. */
+#define CIRCUIT_PURPOSE_TESTING 20
+/** A controller made this circuit and Tor should not use it. */
+#define CIRCUIT_PURPOSE_CONTROLLER 21
+/** This circuit is used for path bias probing only */
+#define CIRCUIT_PURPOSE_PATH_BIAS_TESTING 22
+
+/** This circuit is used for vanguards/restricted paths.
+ *
+ * This type of circuit is *only* created preemptively and never
+ * on-demand. When an HS operation needs to take place (e.g. connect to an
+ * intro point), these circuits are then cannibalized and repurposed to the
+ * actual needed HS purpose. */
+#define CIRCUIT_PURPOSE_HS_VANGUARDS 23
+
+#define CIRCUIT_PURPOSE_MAX_ 23
+/** A catch-all for unrecognized purposes. Currently we don't expect
+ * to make or see any circuits with this purpose. */
+#define CIRCUIT_PURPOSE_UNKNOWN 255
+
+/** True iff the circuit purpose <b>p</b> is for a circuit that
+ * originated at this node. */
+#define CIRCUIT_PURPOSE_IS_ORIGIN(p) ((p)>CIRCUIT_PURPOSE_OR_MAX_)
+/** True iff the circuit purpose <b>p</b> is for a circuit that originated
+ * here to serve as a client. (Hidden services don't count here.) */
+#define CIRCUIT_PURPOSE_IS_CLIENT(p) \
+ ((p)> CIRCUIT_PURPOSE_OR_MAX_ && \
+ (p)<=CIRCUIT_PURPOSE_C_MAX_)
+/** True iff the circuit_t <b>c</b> is actually an origin_circuit_t. */
+#define CIRCUIT_IS_ORIGIN(c) (CIRCUIT_PURPOSE_IS_ORIGIN((c)->purpose))
+/** True iff the circuit purpose <b>p</b> is for an established rendezvous
+ * circuit. */
+#define CIRCUIT_PURPOSE_IS_ESTABLISHED_REND(p) \
+ ((p) == CIRCUIT_PURPOSE_C_REND_JOINED || \
+ (p) == CIRCUIT_PURPOSE_S_REND_JOINED)
+/** True iff the circuit_t c is actually an or_circuit_t */
+#define CIRCUIT_IS_ORCIRC(c) (((circuit_t *)(c))->magic == OR_CIRCUIT_MAGIC)
+
+/** True iff this circuit purpose should count towards the global
+ * pending rate limit (set by MaxClientCircuitsPending). We count all
+ * general purpose circuits, as well as the first step of client onion
+ * service connections (HSDir gets). */
+#define CIRCUIT_PURPOSE_COUNTS_TOWARDS_MAXPENDING(p) \
+ ((p) == CIRCUIT_PURPOSE_C_GENERAL || \
+ (p) == CIRCUIT_PURPOSE_C_HSDIR_GET)
+
+/** Convert a circuit_t* to a pointer to the enclosing or_circuit_t. Assert
+ * if the cast is impossible. */
+or_circuit_t *TO_OR_CIRCUIT(circuit_t *);
+const or_circuit_t *CONST_TO_OR_CIRCUIT(const circuit_t *);
+/** Convert a circuit_t* to a pointer to the enclosing origin_circuit_t.
+ * Assert if the cast is impossible. */
+origin_circuit_t *TO_ORIGIN_CIRCUIT(circuit_t *);
+const origin_circuit_t *CONST_TO_ORIGIN_CIRCUIT(const circuit_t *);
+
+MOCK_DECL(smartlist_t *, circuit_get_global_list, (void));
+smartlist_t *circuit_get_global_origin_circuit_list(void);
+int circuit_any_opened_circuits(void);
+int circuit_any_opened_circuits_cached(void);
+void circuit_cache_opened_circuit_state(int circuits_are_opened);
+
+const char *circuit_state_to_string(int state);
+const char *circuit_purpose_to_controller_string(uint8_t purpose);
+const char *circuit_purpose_to_controller_hs_state_string(uint8_t purpose);
+const char *circuit_purpose_to_string(uint8_t purpose);
+void circuit_dump_by_conn(connection_t *conn, int severity);
+void circuit_set_p_circid_chan(or_circuit_t *circ, circid_t id,
+ channel_t *chan);
+void circuit_set_n_circid_chan(circuit_t *circ, circid_t id,
+ channel_t *chan);
+void channel_mark_circid_unusable(channel_t *chan, circid_t id);
+void channel_mark_circid_usable(channel_t *chan, circid_t id);
+time_t circuit_id_when_marked_unusable_on_channel(circid_t circ_id,
+ channel_t *chan);
+void circuit_set_state(circuit_t *circ, uint8_t state);
+void circuit_close_all_marked(void);
+int32_t circuit_initial_package_window(void);
+origin_circuit_t *origin_circuit_new(void);
+or_circuit_t *or_circuit_new(circid_t p_circ_id, channel_t *p_chan);
+circuit_t *circuit_get_by_circid_channel(circid_t circ_id,
+ channel_t *chan);
+circuit_t *
+circuit_get_by_circid_channel_even_if_marked(circid_t circ_id,
+ channel_t *chan);
+int circuit_id_in_use_on_channel(circid_t circ_id, channel_t *chan);
+circuit_t *circuit_get_by_edge_conn(edge_connection_t *conn);
+void circuit_unlink_all_from_channel(channel_t *chan, int reason);
+origin_circuit_t *circuit_get_by_global_id(uint32_t id);
+origin_circuit_t *circuit_get_ready_rend_circ_by_rend_data(
+ const rend_data_t *rend_data);
+origin_circuit_t *circuit_get_next_by_pk_and_purpose(origin_circuit_t *start,
+ const uint8_t *digest, uint8_t purpose);
+origin_circuit_t *circuit_get_next_service_intro_circ(origin_circuit_t *start);
+origin_circuit_t *circuit_get_next_service_rp_circ(origin_circuit_t *start);
+origin_circuit_t *circuit_get_next_service_hsdir_circ(origin_circuit_t *start);
+origin_circuit_t *circuit_find_to_cannibalize(uint8_t purpose,
+ extend_info_t *info, int flags);
+void circuit_mark_all_unused_circs(void);
+void circuit_mark_all_dirty_circs_as_unusable(void);
++void circuit_synchronize_written_or_bandwidth(const circuit_t *c,
++ circuit_channel_direction_t dir);
+MOCK_DECL(void, circuit_mark_for_close_, (circuit_t *circ, int reason,
+ int line, const char *file));
+int circuit_get_cpath_len(origin_circuit_t *circ);
+int circuit_get_cpath_opened_len(const origin_circuit_t *);
+void circuit_clear_cpath(origin_circuit_t *circ);
+crypt_path_t *circuit_get_cpath_hop(origin_circuit_t *circ, int hopnum);
+void circuit_get_all_pending_on_channel(smartlist_t *out,
+ channel_t *chan);
+int circuit_count_pending_on_channel(channel_t *chan);
+
+#define circuit_mark_for_close(c, reason) \
+ circuit_mark_for_close_((c), (reason), __LINE__, SHORT_FILE__)
+
+void assert_cpath_layer_ok(const crypt_path_t *cp);
+MOCK_DECL(void, assert_circuit_ok,(const circuit_t *c));
+void circuit_free_all(void);
+void circuits_handle_oom(size_t current_allocation);
+
+void circuit_clear_testing_cell_stats(circuit_t *circ);
+
+void channel_note_destroy_pending(channel_t *chan, circid_t id);
+MOCK_DECL(void, channel_note_destroy_not_pending,
+ (channel_t *chan, circid_t id));
+
+smartlist_t *circuit_find_circuits_to_upgrade_from_guard_wait(void);
+
+#ifdef CIRCUITLIST_PRIVATE
+STATIC void circuit_free_(circuit_t *circ);
+#define circuit_free(circ) FREE_AND_NULL(circuit_t, circuit_free_, (circ))
+STATIC size_t n_cells_in_circ_queues(const circuit_t *c);
+STATIC uint32_t circuit_max_queued_data_age(const circuit_t *c, uint32_t now);
+STATIC uint32_t circuit_max_queued_cell_age(const circuit_t *c, uint32_t now);
+STATIC uint32_t circuit_max_queued_item_age(const circuit_t *c, uint32_t now);
+#endif /* defined(CIRCUITLIST_PRIVATE) */
+
+#endif /* !defined(TOR_CIRCUITLIST_H) */
diff --cc src/core/or/or.h
index eae027012,000000000..4d8b6d787
mode 100644,000000..100644
--- a/src/core/or/or.h
+++ b/src/core/or/or.h
@@@ -1,1083 -1,0 +1,1095 @@@
+/* Copyright (c) 2001 Matej Pfajfar.
+ * Copyright (c) 2001-2004, Roger Dingledine.
+ * Copyright (c) 2004-2006, Roger Dingledine, Nick Mathewson.
+ * Copyright (c) 2007-2018, The Tor Project, Inc. */
+/* See LICENSE for licensing information */
+
+/**
+ * \file or.h
+ * \brief Master header file for Tor-specific functionality.
+ **/
+
+#ifndef TOR_OR_H
+#define TOR_OR_H
+
+#include "orconfig.h"
+#include "lib/cc/torint.h"
+
+#ifdef HAVE_SIGNAL_H
+#include <signal.h>
+#endif
+#ifdef HAVE_TIME_H
+#include <time.h>
+#endif
+
+#include "lib/arch/bytes.h"
+#include "lib/cc/compat_compiler.h"
+#include "lib/cc/torint.h"
+#include "lib/container/map.h"
+#include "lib/container/smartlist.h"
+#include "lib/crypt_ops/crypto_cipher.h"
+#include "lib/crypt_ops/crypto_rsa.h"
+#include "lib/ctime/di_ops.h"
+#include "lib/defs/dh_sizes.h"
+#include "lib/encoding/binascii.h"
+#include "lib/encoding/cstring.h"
+#include "lib/encoding/time_fmt.h"
+#include "lib/err/torerr.h"
+#include "lib/fs/dir.h"
+#include "lib/fs/files.h"
+#include "lib/fs/mmap.h"
+#include "lib/fs/path.h"
+#include "lib/fs/userdb.h"
+#include "lib/intmath/addsub.h"
+#include "lib/intmath/bits.h"
+#include "lib/intmath/cmp.h"
+#include "lib/intmath/logic.h"
+#include "lib/intmath/muldiv.h"
+#include "lib/log/escape.h"
+#include "lib/log/ratelim.h"
+#include "lib/log/util_bug.h"
+#include "lib/malloc/malloc.h"
+#include "lib/net/address.h"
+#include "lib/net/inaddr.h"
+#include "lib/net/socket.h"
+#include "lib/string/compat_ctype.h"
+#include "lib/string/compat_string.h"
+#include "lib/string/parse_int.h"
+#include "lib/string/printf.h"
+#include "lib/string/scanf.h"
+#include "lib/string/util_string.h"
+#include "lib/testsupport/testsupport.h"
+#include "lib/thread/threads.h"
+#include "lib/time/compat_time.h"
+#include "lib/wallclock/approx_time.h"
+#include "lib/wallclock/timeval.h"
+
+#include "ht.h"
+
+// These, more than other includes, are for keeping the other struct
+// definitions working. We should remove them when we minimize our includes.
+#include "core/or/entry_port_cfg_st.h"
+
+struct ed25519_public_key_t;
+struct curve25519_public_key_t;
+
+/* These signals are defined to help handle_control_signal work.
+ */
+#ifndef SIGHUP
+#define SIGHUP 1
+#endif
+#ifndef SIGINT
+#define SIGINT 2
+#endif
+#ifndef SIGUSR1
+#define SIGUSR1 10
+#endif
+#ifndef SIGUSR2
+#define SIGUSR2 12
+#endif
+#ifndef SIGTERM
+#define SIGTERM 15
+#endif
+/* Controller signals start at a high number so we don't
+ * conflict with system-defined signals. */
+#define SIGNEWNYM 129
+#define SIGCLEARDNSCACHE 130
+#define SIGHEARTBEAT 131
+
+#if (SIZEOF_CELL_T != 0)
+/* On Irix, stdlib.h defines a cell_t type, so we need to make sure
+ * that our stuff always calls cell_t something different. */
+#define cell_t tor_cell_t
+#endif
+
+/** Helper macro: Given a pointer to to.base_, of type from*, return &to. */
+#define DOWNCAST(to, ptr) ((to*)SUBTYPE_P(ptr, to, base_))
+
+/** Length of longest allowable configured nickname. */
+#define MAX_NICKNAME_LEN 19
+/** Length of a router identity encoded as a hexadecimal digest, plus
+ * possible dollar sign. */
+#define MAX_HEX_NICKNAME_LEN (HEX_DIGEST_LEN+1)
+/** Maximum length of verbose router identifier: dollar sign, hex ID digest,
+ * equal sign or tilde, nickname. */
+#define MAX_VERBOSE_NICKNAME_LEN (1+HEX_DIGEST_LEN+1+MAX_NICKNAME_LEN)
+
+/** For HTTP parsing: Maximum number of bytes we'll accept in the headers
+ * of an HTTP request or response. */
+#define MAX_HEADERS_SIZE 50000
+
+/** Maximum size, in bytes, of a single router descriptor uploaded to us
+ * as a directory authority. Caches and clients fetch whatever descriptors
+ * the authorities tell them to fetch, and don't care about size. */
+#define MAX_DESCRIPTOR_UPLOAD_SIZE 20000
+
+/** Maximum size of a single extrainfo document, as above. */
+#define MAX_EXTRAINFO_UPLOAD_SIZE 50000
+
+/** Minimum lifetime for an onion key in days. */
+#define MIN_ONION_KEY_LIFETIME_DAYS (1)
+
+/** Maximum lifetime for an onion key in days. */
+#define MAX_ONION_KEY_LIFETIME_DAYS (90)
+
+/** Default lifetime for an onion key in days. */
+#define DEFAULT_ONION_KEY_LIFETIME_DAYS (28)
+
+/** Minimum grace period for acceptance of an onion key in days.
+ * The maximum value is defined in proposal #274 as being the current network
+ * consensus parameter for "onion-key-rotation-days". */
+#define MIN_ONION_KEY_GRACE_PERIOD_DAYS (1)
+
+/** Default grace period for acceptance of an onion key in days. */
+#define DEFAULT_ONION_KEY_GRACE_PERIOD_DAYS (7)
+
+/** How often we should check the network consensus if it is time to rotate or
+ * expire onion keys. */
+#define ONION_KEY_CONSENSUS_CHECK_INTERVAL (60*60)
+
+/** How often do we rotate TLS contexts? */
+#define MAX_SSL_KEY_LIFETIME_INTERNAL (2*60*60)
+
+/** How old do we allow a router to get before removing it
+ * from the router list? In seconds. */
+#define ROUTER_MAX_AGE (60*60*48)
+/** How old can a router get before we (as a server) will no longer
+ * consider it live? In seconds. */
+#define ROUTER_MAX_AGE_TO_PUBLISH (60*60*24)
+/** How old do we let a saved descriptor get before force-removing it? */
+#define OLD_ROUTER_DESC_MAX_AGE (60*60*24*5)
+
+/* Proxy client types */
+#define PROXY_NONE 0
+#define PROXY_CONNECT 1
+#define PROXY_SOCKS4 2
+#define PROXY_SOCKS5 3
+/* !!!! If there is ever a PROXY_* type over 3, we must grow the proxy_type
+ * field in or_connection_t */
+
+/* Pluggable transport proxy type. Don't use this in or_connection_t,
+ * instead use the actual underlying proxy type (see above). */
+#define PROXY_PLUGGABLE 4
+
+/** How many circuits do we want simultaneously in-progress to handle
+ * a given stream? */
+#define MIN_CIRCUITS_HANDLING_STREAM 2
+
+/* These RELAY_COMMAND constants define values for relay cell commands, and
+* must match those defined in tor-spec.txt. */
+#define RELAY_COMMAND_BEGIN 1
+#define RELAY_COMMAND_DATA 2
+#define RELAY_COMMAND_END 3
+#define RELAY_COMMAND_CONNECTED 4
+#define RELAY_COMMAND_SENDME 5
+#define RELAY_COMMAND_EXTEND 6
+#define RELAY_COMMAND_EXTENDED 7
+#define RELAY_COMMAND_TRUNCATE 8
+#define RELAY_COMMAND_TRUNCATED 9
+#define RELAY_COMMAND_DROP 10
+#define RELAY_COMMAND_RESOLVE 11
+#define RELAY_COMMAND_RESOLVED 12
+#define RELAY_COMMAND_BEGIN_DIR 13
+#define RELAY_COMMAND_EXTEND2 14
+#define RELAY_COMMAND_EXTENDED2 15
+
+#define RELAY_COMMAND_ESTABLISH_INTRO 32
+#define RELAY_COMMAND_ESTABLISH_RENDEZVOUS 33
+#define RELAY_COMMAND_INTRODUCE1 34
+#define RELAY_COMMAND_INTRODUCE2 35
+#define RELAY_COMMAND_RENDEZVOUS1 36
+#define RELAY_COMMAND_RENDEZVOUS2 37
+#define RELAY_COMMAND_INTRO_ESTABLISHED 38
+#define RELAY_COMMAND_RENDEZVOUS_ESTABLISHED 39
+#define RELAY_COMMAND_INTRODUCE_ACK 40
+
+/* Reasons why an OR connection is closed. */
+#define END_OR_CONN_REASON_DONE 1
+#define END_OR_CONN_REASON_REFUSED 2 /* connection refused */
+#define END_OR_CONN_REASON_OR_IDENTITY 3
+#define END_OR_CONN_REASON_CONNRESET 4 /* connection reset by peer */
+#define END_OR_CONN_REASON_TIMEOUT 5
+#define END_OR_CONN_REASON_NO_ROUTE 6 /* no route to host/net */
+#define END_OR_CONN_REASON_IO_ERROR 7 /* read/write error */
+#define END_OR_CONN_REASON_RESOURCE_LIMIT 8 /* sockets, buffers, etc */
+#define END_OR_CONN_REASON_PT_MISSING 9 /* PT failed or not available */
+#define END_OR_CONN_REASON_MISC 10
+
+/* Reasons why we (or a remote OR) might close a stream. See tor-spec.txt for
+ * documentation of these. The values must match. */
+#define END_STREAM_REASON_MISC 1
+#define END_STREAM_REASON_RESOLVEFAILED 2
+#define END_STREAM_REASON_CONNECTREFUSED 3
+#define END_STREAM_REASON_EXITPOLICY 4
+#define END_STREAM_REASON_DESTROY 5
+#define END_STREAM_REASON_DONE 6
+#define END_STREAM_REASON_TIMEOUT 7
+#define END_STREAM_REASON_NOROUTE 8
+#define END_STREAM_REASON_HIBERNATING 9
+#define END_STREAM_REASON_INTERNAL 10
+#define END_STREAM_REASON_RESOURCELIMIT 11
+#define END_STREAM_REASON_CONNRESET 12
+#define END_STREAM_REASON_TORPROTOCOL 13
+#define END_STREAM_REASON_NOTDIRECTORY 14
+#define END_STREAM_REASON_ENTRYPOLICY 15
+
+/* These high-numbered end reasons are not part of the official spec,
+ * and are not intended to be put in relay end cells. They are here
+ * to be more informative when sending back socks replies to the
+ * application. */
+/* XXXX 256 is no longer used; feel free to reuse it. */
+/** We were unable to attach the connection to any circuit at all. */
+/* XXXX the ways we use this one don't make a lot of sense. */
+#define END_STREAM_REASON_CANT_ATTACH 257
+/** We can't connect to any directories at all, so we killed our streams
+ * before they can time out. */
+#define END_STREAM_REASON_NET_UNREACHABLE 258
+/** This is a SOCKS connection, and the client used (or misused) the SOCKS
+ * protocol in a way we couldn't handle. */
+#define END_STREAM_REASON_SOCKSPROTOCOL 259
+/** This is a transparent proxy connection, but we can't extract the original
+ * target address:port. */
+#define END_STREAM_REASON_CANT_FETCH_ORIG_DEST 260
+/** This is a connection on the NATD port, and the destination IP:Port was
+ * either ill-formed or out-of-range. */
+#define END_STREAM_REASON_INVALID_NATD_DEST 261
+/** The target address is in a private network (like 127.0.0.1 or 10.0.0.1);
+ * you don't want to do that over a randomly chosen exit */
+#define END_STREAM_REASON_PRIVATE_ADDR 262
+/** This is an HTTP tunnel connection and the client used or misused HTTP in a
+ * way we can't handle.
+ */
+#define END_STREAM_REASON_HTTPPROTOCOL 263
+
+/** Bitwise-and this value with endreason to mask out all flags. */
+#define END_STREAM_REASON_MASK 511
+
+/** Bitwise-or this with the argument to control_event_stream_status
+ * to indicate that the reason came from an END cell. */
+#define END_STREAM_REASON_FLAG_REMOTE 512
+/** Bitwise-or this with the argument to control_event_stream_status
+ * to indicate that we already sent a CLOSED stream event. */
+#define END_STREAM_REASON_FLAG_ALREADY_SENT_CLOSED 1024
+/** Bitwise-or this with endreason to indicate that we already sent
+ * a socks reply, and no further reply needs to be sent from
+ * connection_mark_unattached_ap(). */
+#define END_STREAM_REASON_FLAG_ALREADY_SOCKS_REPLIED 2048
+
+/* 'type' values to use in RESOLVED cells. Specified in tor-spec.txt. */
+#define RESOLVED_TYPE_HOSTNAME 0
+#define RESOLVED_TYPE_IPV4 4
+#define RESOLVED_TYPE_IPV6 6
+#define RESOLVED_TYPE_ERROR_TRANSIENT 0xF0
+#define RESOLVED_TYPE_ERROR 0xF1
+
+/* Negative reasons are internal: we never send them in a DESTROY or TRUNCATE
+ * call; they only go to the controller for tracking */
+
+/* Closing introduction point that were opened in parallel. */
+#define END_CIRC_REASON_IP_NOW_REDUNDANT -4
+
+/** Our post-timeout circuit time measurement period expired.
+ * We must give up now */
+#define END_CIRC_REASON_MEASUREMENT_EXPIRED -3
+
+/** We couldn't build a path for this circuit. */
+#define END_CIRC_REASON_NOPATH -2
+/** Catch-all "other" reason for closing origin circuits. */
+#define END_CIRC_AT_ORIGIN -1
+
+/* Reasons why we (or a remote OR) might close a circuit. See tor-spec.txt
+ * section 5.4 for documentation of these. */
+#define END_CIRC_REASON_MIN_ 0
+#define END_CIRC_REASON_NONE 0
+#define END_CIRC_REASON_TORPROTOCOL 1
+#define END_CIRC_REASON_INTERNAL 2
+#define END_CIRC_REASON_REQUESTED 3
+#define END_CIRC_REASON_HIBERNATING 4
+#define END_CIRC_REASON_RESOURCELIMIT 5
+#define END_CIRC_REASON_CONNECTFAILED 6
+#define END_CIRC_REASON_OR_IDENTITY 7
+#define END_CIRC_REASON_CHANNEL_CLOSED 8
+#define END_CIRC_REASON_FINISHED 9
+#define END_CIRC_REASON_TIMEOUT 10
+#define END_CIRC_REASON_DESTROYED 11
+#define END_CIRC_REASON_NOSUCHSERVICE 12
+#define END_CIRC_REASON_MAX_ 12
+
+/** Bitwise-OR this with the argument to circuit_mark_for_close() or
+ * control_event_circuit_status() to indicate that the reason was
+ * passed through from a destroy or truncate cell. */
+#define END_CIRC_REASON_FLAG_REMOTE 512
+
+/** Length of 'y' portion of 'y.onion' URL. */
+#define REND_SERVICE_ID_LEN_BASE32 16
+
+/** Length of 'y.onion' including '.onion' URL. */
+#define REND_SERVICE_ADDRESS_LEN (16+1+5)
+
+/** Length of a binary-encoded rendezvous service ID. */
+#define REND_SERVICE_ID_LEN 10
+
+/** Time period for which a v2 descriptor will be valid. */
+#define REND_TIME_PERIOD_V2_DESC_VALIDITY (24*60*60)
+
+/** Time period within which two sets of v2 descriptors will be uploaded in
+ * parallel. */
+#define REND_TIME_PERIOD_OVERLAPPING_V2_DESCS (60*60)
+
+/** Number of non-consecutive replicas (i.e. distributed somewhere
+ * in the ring) for a descriptor. */
+#define REND_NUMBER_OF_NON_CONSECUTIVE_REPLICAS 2
+
+/** Number of consecutive replicas for a descriptor. */
+#define REND_NUMBER_OF_CONSECUTIVE_REPLICAS 3
+
+/** Length of v2 descriptor ID (32 base32 chars = 160 bits). */
+#define REND_DESC_ID_V2_LEN_BASE32 BASE32_DIGEST_LEN
+
+/** Length of the base32-encoded secret ID part of versioned hidden service
+ * descriptors. */
+#define REND_SECRET_ID_PART_LEN_BASE32 BASE32_DIGEST_LEN
+
+/** Length of the base32-encoded hash of an introduction point's
+ * identity key. */
+#define REND_INTRO_POINT_ID_LEN_BASE32 BASE32_DIGEST_LEN
+
+/** Length of the descriptor cookie that is used for client authorization
+ * to hidden services. */
+#define REND_DESC_COOKIE_LEN 16
+
+/** Length of the base64-encoded descriptor cookie that is used for
+ * exchanging client authorization between hidden service and client. */
+#define REND_DESC_COOKIE_LEN_BASE64 22
+
+/** Length of client identifier in encrypted introduction points for hidden
+ * service authorization type 'basic'. */
+#define REND_BASIC_AUTH_CLIENT_ID_LEN 4
+
+/** Multiple of the number of clients to which the real number of clients
+ * is padded with fake clients for hidden service authorization type
+ * 'basic'. */
+#define REND_BASIC_AUTH_CLIENT_MULTIPLE 16
+
+/** Length of client entry consisting of client identifier and encrypted
+ * session key for hidden service authorization type 'basic'. */
+#define REND_BASIC_AUTH_CLIENT_ENTRY_LEN (REND_BASIC_AUTH_CLIENT_ID_LEN \
+ + CIPHER_KEY_LEN)
+
+/** Maximum size of v2 hidden service descriptors. */
+#define REND_DESC_MAX_SIZE (20 * 1024)
+
+/** Legal characters for use in authorized client names for a hidden
+ * service. */
+#define REND_LEGAL_CLIENTNAME_CHARACTERS \
+ "abcdefghijklmnopqrstuvwxyzABCDEFGHIJKLMNOPQRSTUVWXYZ0123456789+-_"
+
+/** Maximum length of authorized client names for a hidden service. */
+#define REND_CLIENTNAME_MAX_LEN 16
+
+/** Length of the rendezvous cookie that is used to connect circuits at the
+ * rendezvous point. */
+#define REND_COOKIE_LEN DIGEST_LEN
+
+/** Client authorization type that a hidden service performs. */
+typedef enum rend_auth_type_t {
+ REND_NO_AUTH = 0,
+ REND_BASIC_AUTH = 1,
+ REND_STEALTH_AUTH = 2,
+} rend_auth_type_t;
+
+/** Client-side configuration of authorization for a hidden service. */
+typedef struct rend_service_authorization_t {
+ uint8_t descriptor_cookie[REND_DESC_COOKIE_LEN];
+ char onion_address[REND_SERVICE_ADDRESS_LEN+1];
+ rend_auth_type_t auth_type;
+} rend_service_authorization_t;
+
+/** Client- and server-side data that is used for hidden service connection
+ * establishment. Not all fields contain data depending on where this struct
+ * is used. */
+typedef struct rend_data_t {
+ /* Hidden service protocol version of this base object. */
+ uint32_t version;
+
+ /** List of HSDir fingerprints on which this request has been sent to. This
+ * contains binary identity digest of the directory of size DIGEST_LEN. */
+ smartlist_t *hsdirs_fp;
+
+ /** Rendezvous cookie used by both, client and service. */
+ char rend_cookie[REND_COOKIE_LEN];
+
+ /** Number of streams associated with this rendezvous circuit. */
+ int nr_streams;
+} rend_data_t;
+
+typedef struct rend_data_v2_t {
+ /* Rendezvous base data. */
+ rend_data_t base_;
+
+ /** Onion address (without the .onion part) that a client requests. */
+ char onion_address[REND_SERVICE_ID_LEN_BASE32+1];
+
+ /** Descriptor ID for each replicas computed from the onion address. If
+ * the onion address is empty, this array MUST be empty. We keep them so
+ * we know when to purge our entry in the last hsdir request table. */
+ char descriptor_id[REND_NUMBER_OF_NON_CONSECUTIVE_REPLICAS][DIGEST_LEN];
+
+ /** (Optional) descriptor cookie that is used by a client. */
+ char descriptor_cookie[REND_DESC_COOKIE_LEN];
+
+ /** Authorization type for accessing a service used by a client. */
+ rend_auth_type_t auth_type;
+
+ /** Descriptor ID for a client request. The control port command HSFETCH
+ * uses this. It's set if the descriptor query should only use this
+ * descriptor ID. */
+ char desc_id_fetch[DIGEST_LEN];
+
+ /** Hash of the hidden service's PK used by a service. */
+ char rend_pk_digest[DIGEST_LEN];
+} rend_data_v2_t;
+
+/* From a base rend_data_t object <b>d</d>, return the v2 object. */
+static inline
+rend_data_v2_t *TO_REND_DATA_V2(const rend_data_t *d)
+{
+ tor_assert(d);
+ tor_assert(d->version == 2);
+ return DOWNCAST(rend_data_v2_t, d);
+}
+
+/* Stub because we can't include hs_ident.h. */
+struct hs_ident_edge_conn_t;
+struct hs_ident_dir_conn_t;
+struct hs_ident_circuit_t;
+
+typedef struct hsdir_index_t hsdir_index_t;
+
+/** Time interval for tracking replays of DH public keys received in
+ * INTRODUCE2 cells. Used only to avoid launching multiple
+ * simultaneous attempts to connect to the same rendezvous point. */
+#define REND_REPLAY_TIME_INTERVAL (5 * 60)
+
+/** Used to indicate which way a cell is going on a circuit. */
+typedef enum {
+ CELL_DIRECTION_IN=1, /**< The cell is moving towards the origin. */
+ CELL_DIRECTION_OUT=2, /**< The cell is moving away from the origin. */
+} cell_direction_t;
+
++/**
++ * An enum to allow us to specify which channel in a circuit
++ * we're interested in.
++ *
++ * This is needed because our data structures and other fields
++ * for channel delivery are disassociated from the channel.
++ */
++typedef enum {
++ CIRCUIT_N_CHAN = 0,
++ CIRCUIT_P_CHAN = 1
++} circuit_channel_direction_t;
++
+/** Initial value for both sides of a circuit transmission window when the
+ * circuit is initialized. Measured in cells. */
+#define CIRCWINDOW_START 1000
+#define CIRCWINDOW_START_MIN 100
+#define CIRCWINDOW_START_MAX 1000
+/** Amount to increment a circuit window when we get a circuit SENDME. */
+#define CIRCWINDOW_INCREMENT 100
+/** Initial value on both sides of a stream transmission window when the
+ * stream is initialized. Measured in cells. */
+#define STREAMWINDOW_START 500
+#define STREAMWINDOW_START_MAX 500
+/** Amount to increment a stream window when we get a stream SENDME. */
+#define STREAMWINDOW_INCREMENT 50
+
+/** Maximum number of queued cells on a circuit for which we are the
+ * midpoint before we give up and kill it. This must be >= circwindow
+ * to avoid killing innocent circuits, and >= circwindow*2 to give
+ * leaky-pipe a chance of working someday. The ORCIRC_MAX_MIDDLE_KILL_THRESH
+ * ratio controls the margin of error between emitting a warning and
+ * killing the circuit.
+ */
+#define ORCIRC_MAX_MIDDLE_CELLS (CIRCWINDOW_START_MAX*2)
+/** Ratio of hard (circuit kill) to soft (warning) thresholds for the
+ * ORCIRC_MAX_MIDDLE_CELLS tests.
+ */
+#define ORCIRC_MAX_MIDDLE_KILL_THRESH (1.1f)
+
+/* Cell commands. These values are defined in tor-spec.txt. */
+#define CELL_PADDING 0
+#define CELL_CREATE 1
+#define CELL_CREATED 2
+#define CELL_RELAY 3
+#define CELL_DESTROY 4
+#define CELL_CREATE_FAST 5
+#define CELL_CREATED_FAST 6
+#define CELL_VERSIONS 7
+#define CELL_NETINFO 8
+#define CELL_RELAY_EARLY 9
+#define CELL_CREATE2 10
+#define CELL_CREATED2 11
+#define CELL_PADDING_NEGOTIATE 12
+
+#define CELL_VPADDING 128
+#define CELL_CERTS 129
+#define CELL_AUTH_CHALLENGE 130
+#define CELL_AUTHENTICATE 131
+#define CELL_AUTHORIZE 132
+#define CELL_COMMAND_MAX_ 132
+
+/** How long to test reachability before complaining to the user. */
+#define TIMEOUT_UNTIL_UNREACHABILITY_COMPLAINT (20*60)
+
+/** Legal characters in a nickname. */
+#define LEGAL_NICKNAME_CHARACTERS \
+ "abcdefghijklmnopqrstuvwxyzABCDEFGHIJKLMNOPQRSTUVWXYZ0123456789"
+
+/** Name to use in client TLS certificates if no nickname is given. Once
+ * Tor 0.1.2.x is obsolete, we can remove this. */
+#define DEFAULT_CLIENT_NICKNAME "client"
+
+/** Name chosen by routers that don't configure nicknames */
+#define UNNAMED_ROUTER_NICKNAME "Unnamed"
+
+/** Number of bytes in a SOCKS4 header. */
+#define SOCKS4_NETWORK_LEN 8
+
+/*
+ * Relay payload:
+ * Relay command [1 byte]
+ * Recognized [2 bytes]
+ * Stream ID [2 bytes]
+ * Partial SHA-1 [4 bytes]
+ * Length [2 bytes]
+ * Relay payload [498 bytes]
+ */
+
+/** Number of bytes in a cell, minus cell header. */
+#define CELL_PAYLOAD_SIZE 509
+/** Number of bytes in a cell transmitted over the network, in the longest
+ * form */
+#define CELL_MAX_NETWORK_SIZE 514
+
+/** Maximum length of a header on a variable-length cell. */
+#define VAR_CELL_MAX_HEADER_SIZE 7
+
+static int get_cell_network_size(int wide_circ_ids);
+static inline int get_cell_network_size(int wide_circ_ids)
+{
+ return wide_circ_ids ? CELL_MAX_NETWORK_SIZE : CELL_MAX_NETWORK_SIZE - 2;
+}
+static int get_var_cell_header_size(int wide_circ_ids);
+static inline int get_var_cell_header_size(int wide_circ_ids)
+{
+ return wide_circ_ids ? VAR_CELL_MAX_HEADER_SIZE :
+ VAR_CELL_MAX_HEADER_SIZE - 2;
+}
+static int get_circ_id_size(int wide_circ_ids);
+static inline int get_circ_id_size(int wide_circ_ids)
+{
+ return wide_circ_ids ? 4 : 2;
+}
+
+/** Number of bytes in a relay cell's header (not including general cell
+ * header). */
+#define RELAY_HEADER_SIZE (1+2+2+4+2)
+/** Largest number of bytes that can fit in a relay cell payload. */
+#define RELAY_PAYLOAD_SIZE (CELL_PAYLOAD_SIZE-RELAY_HEADER_SIZE)
+
+/** Identifies a circuit on an or_connection */
+typedef uint32_t circid_t;
+/** Identifies a stream on a circuit */
+typedef uint16_t streamid_t;
+
+/* channel_t typedef; struct channel_s is in channel.h */
+
+typedef struct channel_s channel_t;
+
+/* channel_listener_t typedef; struct channel_listener_s is in channel.h */
+
+typedef struct channel_listener_s channel_listener_t;
+
+/* TLS channel stuff */
+
+typedef struct channel_tls_s channel_tls_t;
+
+/* circuitmux_t typedef; struct circuitmux_s is in circuitmux.h */
+
+typedef struct circuitmux_s circuitmux_t;
+
+typedef struct cell_t cell_t;
+typedef struct var_cell_t var_cell_t;
+typedef struct packed_cell_t packed_cell_t;
+typedef struct cell_queue_t cell_queue_t;
+typedef struct destroy_cell_t destroy_cell_t;
+typedef struct destroy_cell_queue_t destroy_cell_queue_t;
+typedef struct ext_or_cmd_t ext_or_cmd_t;
+
+/** Beginning of a RELAY cell payload. */
+typedef struct {
+ uint8_t command; /**< The end-to-end relay command. */
+ uint16_t recognized; /**< Used to tell whether cell is for us. */
+ streamid_t stream_id; /**< Which stream is this cell associated with? */
+ char integrity[4]; /**< Used to tell whether cell is corrupted. */
+ uint16_t length; /**< How long is the payload body? */
+} relay_header_t;
+
+typedef struct socks_request_t socks_request_t;
+typedef struct entry_port_cfg_t entry_port_cfg_t;
+typedef struct server_port_cfg_t server_port_cfg_t;
+
+/** Minimum length of the random part of an AUTH_CHALLENGE cell. */
+#define OR_AUTH_CHALLENGE_LEN 32
+
+/**
+ * @name Certificate types for CERTS cells.
+ *
+ * These values are defined by the protocol, and affect how an X509
+ * certificate in a CERTS cell is interpreted and used.
+ *
+ * @{ */
+/** A certificate that authenticates a TLS link key. The subject key
+ * must match the key used in the TLS handshake; it must be signed by
+ * the identity key. */
+#define OR_CERT_TYPE_TLS_LINK 1
+/** A self-signed identity certificate. The subject key must be a
+ * 1024-bit RSA key. */
+#define OR_CERT_TYPE_ID_1024 2
+/** A certificate that authenticates a key used in an AUTHENTICATE cell
+ * in the v3 handshake. The subject key must be a 1024-bit RSA key; it
+ * must be signed by the identity key */
+#define OR_CERT_TYPE_AUTH_1024 3
+/* DOCDOC */
+#define OR_CERT_TYPE_RSA_ED_CROSSCERT 7
+/**@}*/
+
+/** The first supported type of AUTHENTICATE cell. It contains
+ * a bunch of structures signed with an RSA1024 key. The signed
+ * structures include a HMAC using negotiated TLS secrets, and a digest
+ * of all cells sent or received before the AUTHENTICATE cell (including
+ * the random server-generated AUTH_CHALLENGE cell).
+ */
+#define AUTHTYPE_RSA_SHA256_TLSSECRET 1
+/** As AUTHTYPE_RSA_SHA256_TLSSECRET, but instead of using the
+ * negotiated TLS secrets, uses exported keying material from the TLS
+ * session as described in RFC 5705.
+ *
+ * Not used by today's tors, since everything that supports this
+ * also supports ED25519_SHA256_5705, which is better.
+ **/
+#define AUTHTYPE_RSA_SHA256_RFC5705 2
+/** As AUTHTYPE_RSA_SHA256_RFC5705, but uses an Ed25519 identity key to
+ * authenticate. */
+#define AUTHTYPE_ED25519_SHA256_RFC5705 3
+/*
+ * NOTE: authchallenge_type_is_better() relies on these AUTHTYPE codes
+ * being sorted in order of preference. If we someday add one with
+ * a higher numerical value that we don't like as much, we should revise
+ * authchallenge_type_is_better().
+ */
+
+/** The length of the part of the AUTHENTICATE cell body that the client and
+ * server can generate independently (when using RSA_SHA256_TLSSECRET). It
+ * contains everything except the client's timestamp, the client's randomly
+ * generated nonce, and the signature. */
+#define V3_AUTH_FIXED_PART_LEN (8+(32*6))
+/** The length of the part of the AUTHENTICATE cell body that the client
+ * signs. */
+#define V3_AUTH_BODY_LEN (V3_AUTH_FIXED_PART_LEN + 8 + 16)
+
+typedef struct or_handshake_certs_t or_handshake_certs_t;
+typedef struct or_handshake_state_t or_handshake_state_t;
+
+/** Length of Extended ORPort connection identifier. */
+#define EXT_OR_CONN_ID_LEN DIGEST_LEN /* 20 */
+/*
+ * OR_CONN_HIGHWATER and OR_CONN_LOWWATER moved from connection_or.c so
+ * channeltls.c can see them too.
+ */
+
+/** When adding cells to an OR connection's outbuf, keep adding until the
+ * outbuf is at least this long, or we run out of cells. */
+#define OR_CONN_HIGHWATER (32*1024)
+
+/** Add cells to an OR connection's outbuf whenever the outbuf's data length
+ * drops below this size. */
+#define OR_CONN_LOWWATER (16*1024)
+
+typedef struct connection_t connection_t;
+typedef struct control_connection_t control_connection_t;
+typedef struct dir_connection_t dir_connection_t;
+typedef struct edge_connection_t edge_connection_t;
+typedef struct entry_connection_t entry_connection_t;
+typedef struct listener_connection_t listener_connection_t;
+typedef struct or_connection_t or_connection_t;
+
+/** Cast a connection_t subtype pointer to a connection_t **/
+#define TO_CONN(c) (&(((c)->base_)))
+
+/** Cast a entry_connection_t subtype pointer to a connection_t **/
+#define ENTRY_TO_CONN(c) (TO_CONN(ENTRY_TO_EDGE_CONN(c)))
+
+typedef struct addr_policy_t addr_policy_t;
+
+typedef struct cached_dir_t cached_dir_t;
+
+/** Enum used to remember where a signed_descriptor_t is stored and how to
+ * manage the memory for signed_descriptor_body. */
+typedef enum {
+ /** The descriptor isn't stored on disk at all: the copy in memory is
+ * canonical; the saved_offset field is meaningless. */
+ SAVED_NOWHERE=0,
+ /** The descriptor is stored in the cached_routers file: the
+ * signed_descriptor_body is meaningless; the signed_descriptor_len and
+ * saved_offset are used to index into the mmaped cache file. */
+ SAVED_IN_CACHE,
+ /** The descriptor is stored in the cached_routers.new file: the
+ * signed_descriptor_body and saved_offset fields are both set. */
+ /* FFFF (We could also mmap the file and grow the mmap as needed, or
+ * lazy-load the descriptor text by using seek and read. We don't, for
+ * now.)
+ */
+ SAVED_IN_JOURNAL
+} saved_location_t;
+#define saved_location_bitfield_t ENUM_BF(saved_location_t)
+
+/** Enumeration: what directory object is being downloaded?
+ * This determines which schedule is selected to perform the download. */
+typedef enum {
+ DL_SCHED_GENERIC = 0,
+ DL_SCHED_CONSENSUS = 1,
+ DL_SCHED_BRIDGE = 2,
+} download_schedule_t;
+#define download_schedule_bitfield_t ENUM_BF(download_schedule_t)
+
+/** Enumeration: is the download schedule for downloading from an authority,
+ * or from any available directory mirror?
+ * During bootstrap, "any" means a fallback (or an authority, if there
+ * are no fallbacks).
+ * When we have a valid consensus, "any" means any directory server. */
+typedef enum {
+ DL_WANT_ANY_DIRSERVER = 0,
+ DL_WANT_AUTHORITY = 1,
+} download_want_authority_t;
+#define download_want_authority_bitfield_t \
+ ENUM_BF(download_want_authority_t)
+
+/** Enumeration: do we want to increment the schedule position each time a
+ * connection is attempted (these attempts can be concurrent), or do we want
+ * to increment the schedule position after a connection fails? */
+typedef enum {
+ DL_SCHED_INCREMENT_FAILURE = 0,
+ DL_SCHED_INCREMENT_ATTEMPT = 1,
+} download_schedule_increment_t;
+#define download_schedule_increment_bitfield_t \
+ ENUM_BF(download_schedule_increment_t)
+
+typedef struct download_status_t download_status_t;
+
+/** If n_download_failures is this high, the download can never happen. */
+#define IMPOSSIBLE_TO_DOWNLOAD 255
+
+/** The max size we expect router descriptor annotations we create to
+ * be. We'll accept larger ones if we see them on disk, but we won't
+ * create any that are larger than this. */
+#define ROUTER_ANNOTATION_BUF_LEN 256
+
+typedef struct signed_descriptor_t signed_descriptor_t;
+
+/** A signed integer representing a country code. */
+typedef int16_t country_t;
+
+/** Flags used to summarize the declared protocol versions of a relay,
+ * so we don't need to parse them again and again. */
+typedef struct protover_summary_flags_t {
+ /** True iff we have a proto line for this router, or a versions line
+ * from which we could infer the protocols. */
+ unsigned int protocols_known:1;
+
+ /** True iff this router has a version or protocol list that allows it to
+ * accept EXTEND2 cells. This requires Relay=2. */
+ unsigned int supports_extend2_cells:1;
+
+ /** True iff this router has a protocol list that allows it to negotiate
+ * ed25519 identity keys on a link handshake with us. This
+ * requires LinkAuth=3. */
+ unsigned int supports_ed25519_link_handshake_compat:1;
+
+ /** True iff this router has a protocol list that allows it to negotiate
+ * ed25519 identity keys on a link handshake, at all. This requires some
+ * LinkAuth=X for X >= 3. */
+ unsigned int supports_ed25519_link_handshake_any:1;
+
+ /** True iff this router has a protocol list that allows it to be an
+ * introduction point supporting ed25519 authentication key which is part of
+ * the v3 protocol detailed in proposal 224. This requires HSIntro=4. */
+ unsigned int supports_ed25519_hs_intro : 1;
+
+ /** True iff this router has a protocol list that allows it to be an hidden
+ * service directory supporting version 3 as seen in proposal 224. This
+ * requires HSDir=2. */
+ unsigned int supports_v3_hsdir : 1;
+
+ /** True iff this router has a protocol list that allows it to be an hidden
+ * service rendezvous point supporting version 3 as seen in proposal 224.
+ * This requires HSRend=2. */
+ unsigned int supports_v3_rendezvous_point: 1;
+} protover_summary_flags_t;
+
+typedef struct routerinfo_t routerinfo_t;
+typedef struct extrainfo_t extrainfo_t;
+typedef struct routerstatus_t routerstatus_t;
+
+typedef struct microdesc_t microdesc_t;
+typedef struct node_t node_t;
+typedef struct vote_microdesc_hash_t vote_microdesc_hash_t;
+typedef struct vote_routerstatus_t vote_routerstatus_t;
+typedef struct document_signature_t document_signature_t;
+typedef struct networkstatus_voter_info_t networkstatus_voter_info_t;
+typedef struct networkstatus_sr_info_t networkstatus_sr_info_t;
+
+/** Enumerates recognized flavors of a consensus networkstatus document. All
+ * flavors of a consensus are generated from the same set of votes, but they
+ * present different types information to different versions of Tor. */
+typedef enum {
+ FLAV_NS = 0,
+ FLAV_MICRODESC = 1,
+} consensus_flavor_t;
+
+/** How many different consensus flavors are there? */
+#define N_CONSENSUS_FLAVORS ((int)(FLAV_MICRODESC)+1)
+
+typedef struct networkstatus_t networkstatus_t;
+typedef struct ns_detached_signatures_t ns_detached_signatures_t;
+typedef struct desc_store_t desc_store_t;
+typedef struct routerlist_t routerlist_t;
+typedef struct extend_info_t extend_info_t;
+typedef struct authority_cert_t authority_cert_t;
+
+/** Bitfield enum type listing types of information that directory authorities
+ * can be authoritative about, and that directory caches may or may not cache.
+ *
+ * Note that the granularity here is based on authority granularity and on
+ * cache capabilities. Thus, one particular bit may correspond in practice to
+ * a few types of directory info, so long as every authority that pronounces
+ * officially about one of the types prounounces officially about all of them,
+ * and so long as every cache that caches one of them caches all of them.
+ */
+typedef enum {
+ NO_DIRINFO = 0,
+ /** Serves/signs v3 directory information: votes, consensuses, certs */
+ V3_DIRINFO = 1 << 2,
+ /** Serves bridge descriptors. */
+ BRIDGE_DIRINFO = 1 << 4,
+ /** Serves extrainfo documents. */
+ EXTRAINFO_DIRINFO=1 << 5,
+ /** Serves microdescriptors. */
+ MICRODESC_DIRINFO=1 << 6,
+} dirinfo_type_t;
+
+#define ALL_DIRINFO ((dirinfo_type_t)((1<<7)-1))
+
+#define ONION_HANDSHAKE_TYPE_TAP 0x0000
+#define ONION_HANDSHAKE_TYPE_FAST 0x0001
+#define ONION_HANDSHAKE_TYPE_NTOR 0x0002
+#define MAX_ONION_HANDSHAKE_TYPE 0x0002
+
+typedef struct onion_handshake_state_t onion_handshake_state_t;
+typedef struct relay_crypto_t relay_crypto_t;
+typedef struct crypt_path_t crypt_path_t;
+typedef struct crypt_path_reference_t crypt_path_reference_t;
+
+#define CPATH_KEY_MATERIAL_LEN (20*2+16*2)
+
+typedef struct cpath_build_state_t cpath_build_state_t;
+
+struct create_cell_t;
+
+/** Entry in the cell stats list of a circuit; used only if CELL_STATS
+ * events are enabled. */
+typedef struct testing_cell_stats_entry_t {
+ uint8_t command; /**< cell command number. */
+ /** Waiting time in centiseconds if this event is for a removed cell,
+ * or 0 if this event is for adding a cell to the queue. 22 bits can
+ * store more than 11 hours, enough to assume that a circuit with this
+ * delay would long have been closed. */
+ unsigned int waiting_time:22;
+ unsigned int removed:1; /**< 0 for added to, 1 for removed from queue. */
+ unsigned int exitward:1; /**< 0 for app-ward, 1 for exit-ward. */
+} testing_cell_stats_entry_t;
+
+typedef struct circuit_t circuit_t;
+typedef struct origin_circuit_t origin_circuit_t;
+typedef struct or_circuit_t or_circuit_t;
+
+/** Largest number of relay_early cells that we can send on a given
+ * circuit. */
+#define MAX_RELAY_EARLY_CELLS_PER_CIRCUIT 8
+
+typedef enum path_state_t path_state_t;
+#define path_state_bitfield_t ENUM_BF(path_state_t)
+
+#if REND_COOKIE_LEN != DIGEST_LEN
+#error "The REND_TOKEN_LEN macro assumes REND_COOKIE_LEN == DIGEST_LEN"
+#endif
+#define REND_TOKEN_LEN DIGEST_LEN
+
+/** Convert a circuit subtype to a circuit_t. */
+#define TO_CIRCUIT(x) (&((x)->base_))
+
+/** @name Isolation flags
+
+ Ways to isolate client streams
+
+ @{
+*/
+/** Isolate based on destination port */
+#define ISO_DESTPORT (1u<<0)
+/** Isolate based on destination address */
+#define ISO_DESTADDR (1u<<1)
+/** Isolate based on SOCKS authentication */
+#define ISO_SOCKSAUTH (1u<<2)
+/** Isolate based on client protocol choice */
+#define ISO_CLIENTPROTO (1u<<3)
+/** Isolate based on client address */
+#define ISO_CLIENTADDR (1u<<4)
+/** Isolate based on session group (always on). */
+#define ISO_SESSIONGRP (1u<<5)
+/** Isolate based on newnym epoch (always on). */
+#define ISO_NYM_EPOCH (1u<<6)
+/** Isolate all streams (Internal only). */
+#define ISO_STREAM (1u<<7)
+/**@}*/
+
+/** Default isolation level for ports. */
+#define ISO_DEFAULT (ISO_CLIENTADDR|ISO_SOCKSAUTH|ISO_SESSIONGRP|ISO_NYM_EPOCH)
+
+/** Indicates that we haven't yet set a session group on a port_cfg_t. */
+#define SESSION_GROUP_UNSET -1
+/** Session group reserved for directory connections */
+#define SESSION_GROUP_DIRCONN -2
+/** Session group reserved for resolve requests launched by a controller */
+#define SESSION_GROUP_CONTROL_RESOLVE -3
+/** First automatically allocated session group number */
+#define SESSION_GROUP_FIRST_AUTO -4
+
+typedef struct port_cfg_t port_cfg_t;
+typedef struct routerset_t routerset_t;
+
+/** A magic value for the (Socks|OR|...)Port options below, telling Tor
+ * to pick its own port. */
+#define CFG_AUTO_PORT 0xc4005e
+
+typedef struct or_options_t or_options_t;
+
+#define LOG_PROTOCOL_WARN (get_protocol_warning_severity_level())
+
+typedef struct or_state_t or_state_t;
+
+#define MAX_SOCKS_ADDR_LEN 256
+
+/********************************* circuitbuild.c **********************/
+
+/** How many hops does a general-purpose circuit have by default? */
+#define DEFAULT_ROUTE_LEN 3
+
+/* Circuit Build Timeout "public" structures. */
+
+/** Precision multiplier for the Bw weights */
+#define BW_WEIGHT_SCALE 10000
+#define BW_MIN_WEIGHT_SCALE 1
+#define BW_MAX_WEIGHT_SCALE INT32_MAX
+
+typedef struct circuit_build_times_s circuit_build_times_t;
+
+/********************************* config.c ***************************/
+
+/********************************* connection_edge.c *************************/
+
+/** Enumerates possible origins of a client-side address mapping. */
+typedef enum {
+ /** We're remapping this address because the controller told us to. */
+ ADDRMAPSRC_CONTROLLER,
+ /** We're remapping this address because of an AutomapHostsOnResolve
+ * configuration. */
+ ADDRMAPSRC_AUTOMAP,
+ /** We're remapping this address because our configuration (via torrc, the
+ * command line, or a SETCONF command) told us to. */
+ ADDRMAPSRC_TORRC,
+ /** We're remapping this address because we have TrackHostExit configured,
+ * and we want to remember to use the same exit next time. */
+ ADDRMAPSRC_TRACKEXIT,
+ /** We're remapping this address because we got a DNS resolution from a
+ * Tor server that told us what its value was. */
+ ADDRMAPSRC_DNS,
+
+ /** No remapping has occurred. This isn't a possible value for an
+ * addrmap_entry_t; it's used as a null value when we need to answer "Why
+ * did this remapping happen." */
+ ADDRMAPSRC_NONE
+} addressmap_entry_source_t;
+#define addressmap_entry_source_bitfield_t ENUM_BF(addressmap_entry_source_t)
+
+#define WRITE_STATS_INTERVAL (24*60*60)
+
+/********************************* dirvote.c ************************/
+
+typedef struct vote_timing_t vote_timing_t;
+
+/********************************* microdesc.c *************************/
+
+typedef struct microdesc_cache_t microdesc_cache_t;
+
+/********************************* rendcommon.c ***************************/
+
+typedef struct rend_authorized_client_t rend_authorized_client_t;
+typedef struct rend_encoded_v2_service_descriptor_t
+ rend_encoded_v2_service_descriptor_t;
+
+/** The maximum number of non-circuit-build-timeout failures a hidden
+ * service client will tolerate while trying to build a circuit to an
+ * introduction point. See also rend_intro_point_t.unreachable_count. */
+#define MAX_INTRO_POINT_REACHABILITY_FAILURES 5
+
+/** The minimum and maximum number of distinct INTRODUCE2 cells which a
+ * hidden service's introduction point will receive before it begins to
+ * expire. */
+#define INTRO_POINT_MIN_LIFETIME_INTRODUCTIONS 16384
+/* Double the minimum value so the interval is [min, min * 2]. */
+#define INTRO_POINT_MAX_LIFETIME_INTRODUCTIONS \
+ (INTRO_POINT_MIN_LIFETIME_INTRODUCTIONS * 2)
+
+/** The minimum number of seconds that an introduction point will last
+ * before expiring due to old age. (If it receives
+ * INTRO_POINT_LIFETIME_INTRODUCTIONS INTRODUCE2 cells, it may expire
+ * sooner.)
+ *
+ * XXX Should this be configurable? */
+#define INTRO_POINT_LIFETIME_MIN_SECONDS (18*60*60)
+/** The maximum number of seconds that an introduction point will last
+ * before expiring due to old age.
+ *
+ * XXX Should this be configurable? */
+#define INTRO_POINT_LIFETIME_MAX_SECONDS (24*60*60)
+
+/** The maximum number of circuit creation retry we do to an intro point
+ * before giving up. We try to reuse intro point that fails during their
+ * lifetime so this is a hard limit on the amount of time we do that. */
+#define MAX_INTRO_POINT_CIRCUIT_RETRIES 3
+
+typedef struct rend_intro_point_t rend_intro_point_t;
+typedef struct rend_service_descriptor_t rend_service_descriptor_t;
+
+/********************************* routerlist.c ***************************/
+
+typedef struct dir_server_t dir_server_t;
+
+#define RELAY_REQUIRED_MIN_BANDWIDTH (75*1024)
+#define BRIDGE_REQUIRED_MIN_BANDWIDTH (50*1024)
+
+#define ROUTER_MAX_DECLARED_BANDWIDTH INT32_MAX
+
+typedef struct tor_version_t tor_version_t;
+
+#endif /* !defined(TOR_OR_H) */
diff --cc src/core/or/relay.c
index 407873e2a,000000000..260d84797
mode 100644,000000..100644
--- a/src/core/or/relay.c
+++ b/src/core/or/relay.c
@@@ -1,3168 -1,0 +1,3169 @@@
+/* Copyright (c) 2001 Matej Pfajfar.
+ * Copyright (c) 2001-2004, Roger Dingledine.
+ * Copyright (c) 2004-2006, Roger Dingledine, Nick Mathewson.
+ * Copyright (c) 2007-2018, The Tor Project, Inc. */
+/* See LICENSE for licensing information */
+
+/**
+ * \file relay.c
+ * \brief Handle relay cell encryption/decryption, plus packaging and
+ * receiving from circuits, plus queuing on circuits.
+ *
+ * This is a core modules that makes Tor work. It's responsible for
+ * dealing with RELAY cells (the ones that travel more than one hop along a
+ * circuit), by:
+ * <ul>
+ * <li>constructing relays cells,
+ * <li>encrypting relay cells,
+ * <li>decrypting relay cells,
+ * <li>demultiplexing relay cells as they arrive on a connection,
+ * <li>queueing relay cells for retransmission,
+ * <li>or handling relay cells that are for us to receive (as an exit or a
+ * client).
+ * </ul>
+ *
+ * RELAY cells are generated throughout the code at the client or relay side,
+ * using relay_send_command_from_edge() or one of the functions like
+ * connection_edge_send_command() that calls it. Of particular interest is
+ * connection_edge_package_raw_inbuf(), which takes information that has
+ * arrived on an edge connection socket, and packages it as a RELAY_DATA cell
+ * -- this is how information is actually sent across the Tor network. The
+ * cryptography for these functions is handled deep in
+ * circuit_package_relay_cell(), which either adds a single layer of
+ * encryption (if we're an exit), or multiple layers (if we're the origin of
+ * the circuit). After construction and encryption, the RELAY cells are
+ * passed to append_cell_to_circuit_queue(), which queues them for
+ * transmission and tells the circuitmux (see circuitmux.c) that the circuit
+ * is waiting to send something.
+ *
+ * Incoming RELAY cells arrive at circuit_receive_relay_cell(), called from
+ * command.c. There they are decrypted and, if they are for us, are passed to
+ * connection_edge_process_relay_cell(). If they're not for us, they're
+ * re-queued for retransmission again with append_cell_to_circuit_queue().
+ *
+ * The connection_edge_process_relay_cell() function handles all the different
+ * types of relay cells, launching requests or transmitting data as needed.
+ **/
+
+#define RELAY_PRIVATE
+#include "core/or/or.h"
+#include "feature/client/addressmap.h"
+#include "lib/err/backtrace.h"
+#include "lib/container/buffers.h"
+#include "core/or/channel.h"
+#include "feature/client/circpathbias.h"
+#include "core/or/circuitbuild.h"
+#include "core/or/circuitlist.h"
+#include "core/or/circuituse.h"
+#include "lib/compress/compress.h"
+#include "app/config/config.h"
+#include "core/mainloop/connection.h"
+#include "core/or/connection_edge.h"
+#include "core/or/connection_or.h"
+#include "feature/control/control.h"
+#include "lib/crypt_ops/crypto_rand.h"
+#include "lib/crypt_ops/crypto_util.h"
+#include "feature/dircache/directory.h"
+#include "feature/relay/dns.h"
+#include "feature/stats/geoip.h"
+#include "feature/hs/hs_cache.h"
+#include "core/mainloop/main.h"
+#include "feature/nodelist/networkstatus.h"
+#include "feature/nodelist/nodelist.h"
+#include "core/crypto/onion.h"
+#include "core/or/policies.h"
+#include "core/or/reasons.h"
+#include "core/or/relay.h"
+#include "core/crypto/relay_crypto.h"
+#include "feature/rend/rendcache.h"
+#include "feature/rend/rendcommon.h"
+#include "feature/relay/router.h"
+#include "feature/nodelist/routerlist.h"
+#include "feature/nodelist/routerparse.h"
+#include "core/or/scheduler.h"
+#include "feature/stats/rephist.h"
+
+#include "core/or/cell_st.h"
+#include "core/or/cell_queue_st.h"
+#include "core/or/cpath_build_state_st.h"
+#include "feature/dircommon/dir_connection_st.h"
+#include "core/or/destroy_cell_queue_st.h"
+#include "core/or/entry_connection_st.h"
+#include "core/or/extend_info_st.h"
+#include "core/or/or_circuit_st.h"
+#include "core/or/origin_circuit_st.h"
+#include "feature/nodelist/routerinfo_st.h"
+#include "core/or/socks_request_st.h"
+
+#include "lib/intmath/weakrng.h"
+
+static edge_connection_t *relay_lookup_conn(circuit_t *circ, cell_t *cell,
+ cell_direction_t cell_direction,
+ crypt_path_t *layer_hint);
+
+static void circuit_consider_sending_sendme(circuit_t *circ,
+ crypt_path_t *layer_hint);
+static void circuit_resume_edge_reading(circuit_t *circ,
+ crypt_path_t *layer_hint);
+static int circuit_resume_edge_reading_helper(edge_connection_t *conn,
+ circuit_t *circ,
+ crypt_path_t *layer_hint);
+static int circuit_consider_stop_edge_reading(circuit_t *circ,
+ crypt_path_t *layer_hint);
+static int circuit_queue_streams_are_blocked(circuit_t *circ);
+static void adjust_exit_policy_from_exitpolicy_failure(origin_circuit_t *circ,
+ entry_connection_t *conn,
+ node_t *node,
+ const tor_addr_t *addr);
+
+/** Stop reading on edge connections when we have this many cells
+ * waiting on the appropriate queue. */
+#define CELL_QUEUE_HIGHWATER_SIZE 256
+/** Start reading from edge connections again when we get down to this many
+ * cells. */
+#define CELL_QUEUE_LOWWATER_SIZE 64
+
+/** Stats: how many relay cells have originated at this hop, or have
+ * been relayed onward (not recognized at this hop)?
+ */
+uint64_t stats_n_relay_cells_relayed = 0;
+/** Stats: how many relay cells have been delivered to streams at this
+ * hop?
+ */
+uint64_t stats_n_relay_cells_delivered = 0;
+/** Stats: how many circuits have we closed due to the cell queue limit being
+ * reached (see append_cell_to_circuit_queue()) */
+uint64_t stats_n_circ_max_cell_reached = 0;
+
+/** Used to tell which stream to read from first on a circuit. */
+static tor_weak_rng_t stream_choice_rng = TOR_WEAK_RNG_INIT;
+
+/**
+ * Update channel usage state based on the type of relay cell and
+ * circuit properties.
+ *
+ * This is needed to determine if a client channel is being
+ * used for application traffic, and if a relay channel is being
+ * used for multihop circuits and application traffic. The decision
+ * to pad in channelpadding.c depends upon this info (as well as
+ * consensus parameters) to decide what channels to pad.
+ */
+static void
+circuit_update_channel_usage(circuit_t *circ, cell_t *cell)
+{
+ if (CIRCUIT_IS_ORIGIN(circ)) {
+ /*
+ * The client state was first set much earlier in
+ * circuit_send_next_onion_skin(), so we can start padding as early as
+ * possible.
+ *
+ * However, if padding turns out to be expensive, we may want to not do
+ * it until actual application traffic starts flowing (which is controlled
+ * via consensus param nf_pad_before_usage).
+ *
+ * So: If we're an origin circuit and we've created a full length circuit,
+ * then any CELL_RELAY cell means application data. Increase the usage
+ * state of the channel to indicate this.
+ *
+ * We want to wait for CELL_RELAY specifically here, so we know that
+ * the channel was definitely being used for data and not for extends.
+ * By default, we pad as soon as a channel has been used for *any*
+ * circuits, so this state is irrelevant to the padding decision in
+ * the default case. However, if padding turns out to be expensive,
+ * we would like the ability to avoid padding until we're absolutely
+ * sure that a channel is used for enough application data to be worth
+ * padding.
+ *
+ * (So it does not matter that CELL_RELAY_EARLY can actually contain
+ * application data. This is only a load reducing option and that edge
+ * case does not matter if we're desperately trying to reduce overhead
+ * anyway. See also consensus parameter nf_pad_before_usage).
+ */
+ if (BUG(!circ->n_chan))
+ return;
+
+ if (circ->n_chan->channel_usage == CHANNEL_USED_FOR_FULL_CIRCS &&
+ cell->command == CELL_RELAY) {
+ circ->n_chan->channel_usage = CHANNEL_USED_FOR_USER_TRAFFIC;
+ }
+ } else {
+ /* If we're a relay circuit, the question is more complicated. Basically:
+ * we only want to pad connections that carry multihop (anonymous)
+ * circuits.
+ *
+ * We assume we're more than one hop if either the previous hop
+ * is not a client, or if the previous hop is a client and there's
+ * a next hop. Then, circuit traffic starts at RELAY_EARLY, and
+ * user application traffic starts when we see RELAY cells.
+ */
+ or_circuit_t *or_circ = TO_OR_CIRCUIT(circ);
+
+ if (BUG(!or_circ->p_chan))
+ return;
+
+ if (!channel_is_client(or_circ->p_chan) ||
+ (channel_is_client(or_circ->p_chan) && circ->n_chan)) {
+ if (cell->command == CELL_RELAY_EARLY) {
+ if (or_circ->p_chan->channel_usage < CHANNEL_USED_FOR_FULL_CIRCS) {
+ or_circ->p_chan->channel_usage = CHANNEL_USED_FOR_FULL_CIRCS;
+ }
+ } else if (cell->command == CELL_RELAY) {
+ or_circ->p_chan->channel_usage = CHANNEL_USED_FOR_USER_TRAFFIC;
+ }
+ }
+ }
+}
+
+/** Receive a relay cell:
+ * - Crypt it (encrypt if headed toward the origin or if we <b>are</b> the
+ * origin; decrypt if we're headed toward the exit).
+ * - Check if recognized (if exitward).
+ * - If recognized and the digest checks out, then find if there's a stream
+ * that the cell is intended for, and deliver it to the right
+ * connection_edge.
+ * - If not recognized, then we need to relay it: append it to the appropriate
+ * cell_queue on <b>circ</b>.
+ *
+ * Return -<b>reason</b> on failure.
+ */
+int
+circuit_receive_relay_cell(cell_t *cell, circuit_t *circ,
+ cell_direction_t cell_direction)
+{
+ channel_t *chan = NULL;
+ crypt_path_t *layer_hint=NULL;
+ char recognized=0;
+ int reason;
+
+ tor_assert(cell);
+ tor_assert(circ);
+ tor_assert(cell_direction == CELL_DIRECTION_OUT ||
+ cell_direction == CELL_DIRECTION_IN);
+ if (circ->marked_for_close)
+ return 0;
+
+ if (relay_decrypt_cell(circ, cell, cell_direction, &layer_hint, &recognized)
+ < 0) {
+ log_fn(LOG_PROTOCOL_WARN, LD_PROTOCOL,
+ "relay crypt failed. Dropping connection.");
+ return -END_CIRC_REASON_INTERNAL;
+ }
+
+ circuit_update_channel_usage(circ, cell);
+
+ if (recognized) {
+ edge_connection_t *conn = NULL;
+
+ if (circ->purpose == CIRCUIT_PURPOSE_PATH_BIAS_TESTING) {
+ if (pathbias_check_probe_response(circ, cell) == -1) {
+ pathbias_count_valid_cells(circ, cell);
+ }
+
+ /* We need to drop this cell no matter what to avoid code that expects
+ * a certain purpose (such as the hidserv code). */
+ return 0;
+ }
+
+ conn = relay_lookup_conn(circ, cell, cell_direction, layer_hint);
+ if (cell_direction == CELL_DIRECTION_OUT) {
+ ++stats_n_relay_cells_delivered;
+ log_debug(LD_OR,"Sending away from origin.");
+ if ((reason=connection_edge_process_relay_cell(cell, circ, conn, NULL))
+ < 0) {
+ log_fn(LOG_PROTOCOL_WARN, LD_PROTOCOL,
+ "connection_edge_process_relay_cell (away from origin) "
+ "failed.");
+ return reason;
+ }
+ }
+ if (cell_direction == CELL_DIRECTION_IN) {
+ ++stats_n_relay_cells_delivered;
+ log_debug(LD_OR,"Sending to origin.");
+ if ((reason = connection_edge_process_relay_cell(cell, circ, conn,
+ layer_hint)) < 0) {
+ /* If a client is trying to connect to unknown hidden service port,
+ * END_CIRC_AT_ORIGIN is sent back so we can then close the circuit.
+ * Do not log warn as this is an expected behavior for a service. */
+ if (reason != END_CIRC_AT_ORIGIN) {
+ log_warn(LD_OR,
+ "connection_edge_process_relay_cell (at origin) failed.");
+ }
+ return reason;
+ }
+ }
+ return 0;
+ }
+
+ /* not recognized. pass it on. */
+ if (cell_direction == CELL_DIRECTION_OUT) {
+ cell->circ_id = circ->n_circ_id; /* switch it */
+ chan = circ->n_chan;
+ } else if (! CIRCUIT_IS_ORIGIN(circ)) {
+ cell->circ_id = TO_OR_CIRCUIT(circ)->p_circ_id; /* switch it */
+ chan = TO_OR_CIRCUIT(circ)->p_chan;
+ } else {
+ log_fn(LOG_PROTOCOL_WARN, LD_OR,
+ "Dropping unrecognized inbound cell on origin circuit.");
+ /* If we see unrecognized cells on path bias testing circs,
+ * it's bad mojo. Those circuits need to die.
+ * XXX: Shouldn't they always die? */
+ if (circ->purpose == CIRCUIT_PURPOSE_PATH_BIAS_TESTING) {
+ TO_ORIGIN_CIRCUIT(circ)->path_state = PATH_STATE_USE_FAILED;
+ return -END_CIRC_REASON_TORPROTOCOL;
+ } else {
+ return 0;
+ }
+ }
+
+ if (!chan) {
+ // XXXX Can this splice stuff be done more cleanly?
+ if (! CIRCUIT_IS_ORIGIN(circ) &&
+ TO_OR_CIRCUIT(circ)->rend_splice &&
+ cell_direction == CELL_DIRECTION_OUT) {
+ or_circuit_t *splice_ = TO_OR_CIRCUIT(circ)->rend_splice;
+ tor_assert(circ->purpose == CIRCUIT_PURPOSE_REND_ESTABLISHED);
+ tor_assert(splice_->base_.purpose == CIRCUIT_PURPOSE_REND_ESTABLISHED);
+ cell->circ_id = splice_->p_circ_id;
+ cell->command = CELL_RELAY; /* can't be relay_early anyway */
+ if ((reason = circuit_receive_relay_cell(cell, TO_CIRCUIT(splice_),
+ CELL_DIRECTION_IN)) < 0) {
+ log_warn(LD_REND, "Error relaying cell across rendezvous; closing "
+ "circuits");
+ /* XXXX Do this here, or just return -1? */
+ circuit_mark_for_close(circ, -reason);
+ return reason;
+ }
+ return 0;
+ }
+ log_fn(LOG_PROTOCOL_WARN, LD_PROTOCOL,
+ "Didn't recognize cell, but circ stops here! Closing circ.");
+ return -END_CIRC_REASON_TORPROTOCOL;
+ }
+
+ log_debug(LD_OR,"Passing on unrecognized cell.");
+
+ ++stats_n_relay_cells_relayed; /* XXXX no longer quite accurate {cells}
+ * we might kill the circ before we relay
+ * the cells. */
+
+ append_cell_to_circuit_queue(circ, chan, cell, cell_direction, 0);
+ return 0;
+}
+
+/** Package a relay cell from an edge:
+ * - Encrypt it to the right layer
+ * - Append it to the appropriate cell_queue on <b>circ</b>.
+ */
+static int
+circuit_package_relay_cell(cell_t *cell, circuit_t *circ,
+ cell_direction_t cell_direction,
+ crypt_path_t *layer_hint, streamid_t on_stream,
+ const char *filename, int lineno)
+{
+ channel_t *chan; /* where to send the cell */
+
+ if (circ->marked_for_close) {
+ /* Circuit is marked; send nothing. */
+ return 0;
+ }
+
+ if (cell_direction == CELL_DIRECTION_OUT) {
+ chan = circ->n_chan;
+ if (!chan) {
+ log_warn(LD_BUG,"outgoing relay cell sent from %s:%d has n_chan==NULL."
+ " Dropping. Circuit is in state %s (%d), and is "
+ "%smarked for close. (%s:%d, %d)", filename, lineno,
+ circuit_state_to_string(circ->state), circ->state,
+ circ->marked_for_close ? "" : "not ",
+ circ->marked_for_close_file?circ->marked_for_close_file:"",
+ circ->marked_for_close, circ->marked_for_close_reason);
+ if (CIRCUIT_IS_ORIGIN(circ)) {
+ circuit_log_path(LOG_WARN, LD_BUG, TO_ORIGIN_CIRCUIT(circ));
+ }
+ log_backtrace(LOG_WARN,LD_BUG,"");
+ return 0; /* just drop it */
+ }
+ if (!CIRCUIT_IS_ORIGIN(circ)) {
+ log_warn(LD_BUG,"outgoing relay cell sent from %s:%d on non-origin "
+ "circ. Dropping.", filename, lineno);
+ log_backtrace(LOG_WARN,LD_BUG,"");
+ return 0; /* just drop it */
+ }
+
+ relay_encrypt_cell_outbound(cell, TO_ORIGIN_CIRCUIT(circ), layer_hint);
+
+ /* Update circ written totals for control port */
+ origin_circuit_t *ocirc = TO_ORIGIN_CIRCUIT(circ);
+ ocirc->n_written_circ_bw = tor_add_u32_nowrap(ocirc->n_written_circ_bw,
+ CELL_PAYLOAD_SIZE);
+
+ } else { /* incoming cell */
+ if (CIRCUIT_IS_ORIGIN(circ)) {
+ /* We should never package an _incoming_ cell from the circuit
+ * origin; that means we messed up somewhere. */
+ log_warn(LD_BUG,"incoming relay cell at origin circuit. Dropping.");
+ assert_circuit_ok(circ);
+ return 0; /* just drop it */
+ }
+ or_circuit_t *or_circ = TO_OR_CIRCUIT(circ);
+ relay_encrypt_cell_inbound(cell, or_circ);
+ chan = or_circ->p_chan;
+ }
+ ++stats_n_relay_cells_relayed;
+
+ append_cell_to_circuit_queue(circ, chan, cell, cell_direction, on_stream);
+ return 0;
+}
+
+/** If cell's stream_id matches the stream_id of any conn that's
+ * attached to circ, return that conn, else return NULL.
+ */
+static edge_connection_t *
+relay_lookup_conn(circuit_t *circ, cell_t *cell,
+ cell_direction_t cell_direction, crypt_path_t *layer_hint)
+{
+ edge_connection_t *tmpconn;
+ relay_header_t rh;
+
+ relay_header_unpack(&rh, cell->payload);
+
+ if (!rh.stream_id)
+ return NULL;
+
+ /* IN or OUT cells could have come from either direction, now
+ * that we allow rendezvous *to* an OP.
+ */
+
+ if (CIRCUIT_IS_ORIGIN(circ)) {
+ for (tmpconn = TO_ORIGIN_CIRCUIT(circ)->p_streams; tmpconn;
+ tmpconn=tmpconn->next_stream) {
+ if (rh.stream_id == tmpconn->stream_id &&
+ !tmpconn->base_.marked_for_close &&
+ tmpconn->cpath_layer == layer_hint) {
+ log_debug(LD_APP,"found conn for stream %d.", rh.stream_id);
+ return tmpconn;
+ }
+ }
+ } else {
+ for (tmpconn = TO_OR_CIRCUIT(circ)->n_streams; tmpconn;
+ tmpconn=tmpconn->next_stream) {
+ if (rh.stream_id == tmpconn->stream_id &&
+ !tmpconn->base_.marked_for_close) {
+ log_debug(LD_EXIT,"found conn for stream %d.", rh.stream_id);
+ if (cell_direction == CELL_DIRECTION_OUT ||
+ connection_edge_is_rendezvous_stream(tmpconn))
+ return tmpconn;
+ }
+ }
+ for (tmpconn = TO_OR_CIRCUIT(circ)->resolving_streams; tmpconn;
+ tmpconn=tmpconn->next_stream) {
+ if (rh.stream_id == tmpconn->stream_id &&
+ !tmpconn->base_.marked_for_close) {
+ log_debug(LD_EXIT,"found conn for stream %d.", rh.stream_id);
+ return tmpconn;
+ }
+ }
+ }
+ return NULL; /* probably a begin relay cell */
+}
+
+/** Pack the relay_header_t host-order structure <b>src</b> into
+ * network-order in the buffer <b>dest</b>. See tor-spec.txt for details
+ * about the wire format.
+ */
+void
+relay_header_pack(uint8_t *dest, const relay_header_t *src)
+{
+ set_uint8(dest, src->command);
+ set_uint16(dest+1, htons(src->recognized));
+ set_uint16(dest+3, htons(src->stream_id));
+ memcpy(dest+5, src->integrity, 4);
+ set_uint16(dest+9, htons(src->length));
+}
+
+/** Unpack the network-order buffer <b>src</b> into a host-order
+ * relay_header_t structure <b>dest</b>.
+ */
+void
+relay_header_unpack(relay_header_t *dest, const uint8_t *src)
+{
+ dest->command = get_uint8(src);
+ dest->recognized = ntohs(get_uint16(src+1));
+ dest->stream_id = ntohs(get_uint16(src+3));
+ memcpy(dest->integrity, src+5, 4);
+ dest->length = ntohs(get_uint16(src+9));
+}
+
+/** Convert the relay <b>command</b> into a human-readable string. */
+static const char *
+relay_command_to_string(uint8_t command)
+{
+ static char buf[64];
+ switch (command) {
+ case RELAY_COMMAND_BEGIN: return "BEGIN";
+ case RELAY_COMMAND_DATA: return "DATA";
+ case RELAY_COMMAND_END: return "END";
+ case RELAY_COMMAND_CONNECTED: return "CONNECTED";
+ case RELAY_COMMAND_SENDME: return "SENDME";
+ case RELAY_COMMAND_EXTEND: return "EXTEND";
+ case RELAY_COMMAND_EXTENDED: return "EXTENDED";
+ case RELAY_COMMAND_TRUNCATE: return "TRUNCATE";
+ case RELAY_COMMAND_TRUNCATED: return "TRUNCATED";
+ case RELAY_COMMAND_DROP: return "DROP";
+ case RELAY_COMMAND_RESOLVE: return "RESOLVE";
+ case RELAY_COMMAND_RESOLVED: return "RESOLVED";
+ case RELAY_COMMAND_BEGIN_DIR: return "BEGIN_DIR";
+ case RELAY_COMMAND_ESTABLISH_INTRO: return "ESTABLISH_INTRO";
+ case RELAY_COMMAND_ESTABLISH_RENDEZVOUS: return "ESTABLISH_RENDEZVOUS";
+ case RELAY_COMMAND_INTRODUCE1: return "INTRODUCE1";
+ case RELAY_COMMAND_INTRODUCE2: return "INTRODUCE2";
+ case RELAY_COMMAND_RENDEZVOUS1: return "RENDEZVOUS1";
+ case RELAY_COMMAND_RENDEZVOUS2: return "RENDEZVOUS2";
+ case RELAY_COMMAND_INTRO_ESTABLISHED: return "INTRO_ESTABLISHED";
+ case RELAY_COMMAND_RENDEZVOUS_ESTABLISHED:
+ return "RENDEZVOUS_ESTABLISHED";
+ case RELAY_COMMAND_INTRODUCE_ACK: return "INTRODUCE_ACK";
+ case RELAY_COMMAND_EXTEND2: return "EXTEND2";
+ case RELAY_COMMAND_EXTENDED2: return "EXTENDED2";
+ default:
+ tor_snprintf(buf, sizeof(buf), "Unrecognized relay command %u",
+ (unsigned)command);
+ return buf;
+ }
+}
+
+/** Make a relay cell out of <b>relay_command</b> and <b>payload</b>, and send
+ * it onto the open circuit <b>circ</b>. <b>stream_id</b> is the ID on
+ * <b>circ</b> for the stream that's sending the relay cell, or 0 if it's a
+ * control cell. <b>cpath_layer</b> is NULL for OR->OP cells, or the
+ * destination hop for OP->OR cells.
+ *
+ * If you can't send the cell, mark the circuit for close and return -1. Else
+ * return 0.
+ */
+MOCK_IMPL(int,
+relay_send_command_from_edge_,(streamid_t stream_id, circuit_t *circ,
+ uint8_t relay_command, const char *payload,
+ size_t payload_len, crypt_path_t *cpath_layer,
+ const char *filename, int lineno))
+{
+ cell_t cell;
+ relay_header_t rh;
+ cell_direction_t cell_direction;
+ /* XXXX NM Split this function into a separate versions per circuit type? */
+
+ tor_assert(circ);
+ tor_assert(payload_len <= RELAY_PAYLOAD_SIZE);
+
+ memset(&cell, 0, sizeof(cell_t));
+ cell.command = CELL_RELAY;
+ if (CIRCUIT_IS_ORIGIN(circ)) {
+ tor_assert(cpath_layer);
+ cell.circ_id = circ->n_circ_id;
+ cell_direction = CELL_DIRECTION_OUT;
+ } else {
+ tor_assert(! cpath_layer);
+ cell.circ_id = TO_OR_CIRCUIT(circ)->p_circ_id;
+ cell_direction = CELL_DIRECTION_IN;
+ }
+
+ memset(&rh, 0, sizeof(rh));
+ rh.command = relay_command;
+ rh.stream_id = stream_id;
+ rh.length = payload_len;
+ relay_header_pack(cell.payload, &rh);
+ if (payload_len)
+ memcpy(cell.payload+RELAY_HEADER_SIZE, payload, payload_len);
+
+ log_debug(LD_OR,"delivering %d cell %s.", relay_command,
+ cell_direction == CELL_DIRECTION_OUT ? "forward" : "backward");
+
+ if (relay_command == RELAY_COMMAND_DROP)
+ rep_hist_padding_count_write(PADDING_TYPE_DROP);
+
+ /* If we are sending an END cell and this circuit is used for a tunneled
+ * directory request, advance its state. */
+ if (relay_command == RELAY_COMMAND_END && circ->dirreq_id)
+ geoip_change_dirreq_state(circ->dirreq_id, DIRREQ_TUNNELED,
+ DIRREQ_END_CELL_SENT);
+
+ if (cell_direction == CELL_DIRECTION_OUT && circ->n_chan) {
+ /* if we're using relaybandwidthrate, this conn wants priority */
+ channel_timestamp_client(circ->n_chan);
+ }
+
+ if (cell_direction == CELL_DIRECTION_OUT) {
+ origin_circuit_t *origin_circ = TO_ORIGIN_CIRCUIT(circ);
+ if (origin_circ->remaining_relay_early_cells > 0 &&
+ (relay_command == RELAY_COMMAND_EXTEND ||
+ relay_command == RELAY_COMMAND_EXTEND2 ||
+ cpath_layer != origin_circ->cpath)) {
+ /* If we've got any relay_early cells left and (we're sending
+ * an extend cell or we're not talking to the first hop), use
+ * one of them. Don't worry about the conn protocol version:
+ * append_cell_to_circuit_queue will fix it up. */
+ cell.command = CELL_RELAY_EARLY;
+ --origin_circ->remaining_relay_early_cells;
+ log_debug(LD_OR, "Sending a RELAY_EARLY cell; %d remaining.",
+ (int)origin_circ->remaining_relay_early_cells);
+ /* Memorize the command that is sent as RELAY_EARLY cell; helps debug
+ * task 878. */
+ origin_circ->relay_early_commands[
+ origin_circ->relay_early_cells_sent++] = relay_command;
+ } else if (relay_command == RELAY_COMMAND_EXTEND ||
+ relay_command == RELAY_COMMAND_EXTEND2) {
+ /* If no RELAY_EARLY cells can be sent over this circuit, log which
+ * commands have been sent as RELAY_EARLY cells before; helps debug
+ * task 878. */
+ smartlist_t *commands_list = smartlist_new();
+ int i = 0;
+ char *commands = NULL;
+ for (; i < origin_circ->relay_early_cells_sent; i++)
+ smartlist_add(commands_list, (char *)
+ relay_command_to_string(origin_circ->relay_early_commands[i]));
+ commands = smartlist_join_strings(commands_list, ",", 0, NULL);
+ log_warn(LD_BUG, "Uh-oh. We're sending a RELAY_COMMAND_EXTEND cell, "
+ "but we have run out of RELAY_EARLY cells on that circuit. "
+ "Commands sent before: %s", commands);
+ tor_free(commands);
+ smartlist_free(commands_list);
+ }
+
+ /* Let's assume we're well-behaved: Anything that we decide to send is
+ * valid, delivered data. */
+ circuit_sent_valid_data(origin_circ, rh.length);
+ }
+
+ if (circuit_package_relay_cell(&cell, circ, cell_direction, cpath_layer,
+ stream_id, filename, lineno) < 0) {
+ log_warn(LD_BUG,"circuit_package_relay_cell failed. Closing.");
+ circuit_mark_for_close(circ, END_CIRC_REASON_INTERNAL);
+ return -1;
+ }
+ return 0;
+}
+
+/** Make a relay cell out of <b>relay_command</b> and <b>payload</b>, and
+ * send it onto the open circuit <b>circ</b>. <b>fromconn</b> is the stream
+ * that's sending the relay cell, or NULL if it's a control cell.
+ * <b>cpath_layer</b> is NULL for OR->OP cells, or the destination hop
+ * for OP->OR cells.
+ *
+ * If you can't send the cell, mark the circuit for close and
+ * return -1. Else return 0.
+ */
+int
+connection_edge_send_command(edge_connection_t *fromconn,
+ uint8_t relay_command, const char *payload,
+ size_t payload_len)
+{
+ /* XXXX NM Split this function into a separate versions per circuit type? */
+ circuit_t *circ;
+ crypt_path_t *cpath_layer = fromconn->cpath_layer;
+ tor_assert(fromconn);
+ circ = fromconn->on_circuit;
+
+ if (fromconn->base_.marked_for_close) {
+ log_warn(LD_BUG,
+ "called on conn that's already marked for close at %s:%d.",
+ fromconn->base_.marked_for_close_file,
+ fromconn->base_.marked_for_close);
+ return 0;
+ }
+
+ if (!circ) {
+ if (fromconn->base_.type == CONN_TYPE_AP) {
+ log_info(LD_APP,"no circ. Closing conn.");
+ connection_mark_unattached_ap(EDGE_TO_ENTRY_CONN(fromconn),
+ END_STREAM_REASON_INTERNAL);
+ } else {
+ log_info(LD_EXIT,"no circ. Closing conn.");
+ fromconn->edge_has_sent_end = 1; /* no circ to send to */
+ fromconn->end_reason = END_STREAM_REASON_INTERNAL;
+ connection_mark_for_close(TO_CONN(fromconn));
+ }
+ return -1;
+ }
+
+ if (circ->marked_for_close) {
+ /* The circuit has been marked, but not freed yet. When it's freed, it
+ * will mark this connection for close. */
+ return -1;
+ }
+
+#ifdef MEASUREMENTS_21206
+ /* Keep track of the number of RELAY_DATA cells sent for directory
+ * connections. */
+ connection_t *linked_conn = TO_CONN(fromconn)->linked_conn;
+
+ if (linked_conn && linked_conn->type == CONN_TYPE_DIR) {
+ ++(TO_DIR_CONN(linked_conn)->data_cells_sent);
+ }
+#endif /* defined(MEASUREMENTS_21206) */
+
+ return relay_send_command_from_edge(fromconn->stream_id, circ,
+ relay_command, payload,
+ payload_len, cpath_layer);
+}
+
+/** How many times will I retry a stream that fails due to DNS
+ * resolve failure or misc error?
+ */
+#define MAX_RESOLVE_FAILURES 3
+
+/** Return 1 if reason is something that you should retry if you
+ * get the end cell before you've connected; else return 0. */
+static int
+edge_reason_is_retriable(int reason)
+{
+ return reason == END_STREAM_REASON_HIBERNATING ||
+ reason == END_STREAM_REASON_RESOURCELIMIT ||
+ reason == END_STREAM_REASON_EXITPOLICY ||
+ reason == END_STREAM_REASON_RESOLVEFAILED ||
+ reason == END_STREAM_REASON_MISC ||
+ reason == END_STREAM_REASON_NOROUTE;
+}
+
+/** Called when we receive an END cell on a stream that isn't open yet,
+ * from the client side.
+ * Arguments are as for connection_edge_process_relay_cell().
+ */
+static int
+connection_ap_process_end_not_open(
+ relay_header_t *rh, cell_t *cell, origin_circuit_t *circ,
+ entry_connection_t *conn, crypt_path_t *layer_hint)
+{
+ node_t *exitrouter;
+ int reason = *(cell->payload+RELAY_HEADER_SIZE);
+ int control_reason;
+ edge_connection_t *edge_conn = ENTRY_TO_EDGE_CONN(conn);
+ (void) layer_hint; /* unused */
+
+ if (rh->length > 0) {
+ if (reason == END_STREAM_REASON_TORPROTOCOL ||
+ reason == END_STREAM_REASON_DESTROY) {
+ /* Both of these reasons could mean a failed tag
+ * hit the exit and it complained. Do not probe.
+ * Fail the circuit. */
+ circ->path_state = PATH_STATE_USE_FAILED;
+ return -END_CIRC_REASON_TORPROTOCOL;
+ } else if (reason == END_STREAM_REASON_INTERNAL) {
+ /* We can't infer success or failure, since older Tors report
+ * ENETUNREACH as END_STREAM_REASON_INTERNAL. */
+ } else {
+ /* Path bias: If we get a valid reason code from the exit,
+ * it wasn't due to tagging.
+ *
+ * We rely on recognized+digest being strong enough to make
+ * tags unlikely to allow us to get tagged, yet 'recognized'
+ * reason codes here. */
+ pathbias_mark_use_success(circ);
+ }
+ }
+
+ /* This end cell is now valid. */
+ circuit_read_valid_data(circ, rh->length);
+
+ if (rh->length == 0) {
+ reason = END_STREAM_REASON_MISC;
+ }
+
+ control_reason = reason | END_STREAM_REASON_FLAG_REMOTE;
+
+ if (edge_reason_is_retriable(reason) &&
+ /* avoid retry if rend */
+ !connection_edge_is_rendezvous_stream(edge_conn)) {
+ const char *chosen_exit_digest =
+ circ->build_state->chosen_exit->identity_digest;
+ log_info(LD_APP,"Address '%s' refused due to '%s'. Considering retrying.",
+ safe_str(conn->socks_request->address),
+ stream_end_reason_to_string(reason));
+ exitrouter = node_get_mutable_by_id(chosen_exit_digest);
+ switch (reason) {
+ case END_STREAM_REASON_EXITPOLICY: {
+ tor_addr_t addr;
+ tor_addr_make_unspec(&addr);
+ if (rh->length >= 5) {
+ int ttl = -1;
+ tor_addr_make_unspec(&addr);
+ if (rh->length == 5 || rh->length == 9) {
+ tor_addr_from_ipv4n(&addr,
+ get_uint32(cell->payload+RELAY_HEADER_SIZE+1));
+ if (rh->length == 9)
+ ttl = (int)ntohl(get_uint32(cell->payload+RELAY_HEADER_SIZE+5));
+ } else if (rh->length == 17 || rh->length == 21) {
+ tor_addr_from_ipv6_bytes(&addr,
+ (char*)(cell->payload+RELAY_HEADER_SIZE+1));
+ if (rh->length == 21)
+ ttl = (int)ntohl(get_uint32(cell->payload+RELAY_HEADER_SIZE+17));
+ }
+ if (tor_addr_is_null(&addr)) {
+ log_info(LD_APP,"Address '%s' resolved to 0.0.0.0. Closing,",
+ safe_str(conn->socks_request->address));
+ connection_mark_unattached_ap(conn, END_STREAM_REASON_TORPROTOCOL);
+ return 0;
+ }
+
+ if ((tor_addr_family(&addr) == AF_INET &&
+ !conn->entry_cfg.ipv4_traffic) ||
+ (tor_addr_family(&addr) == AF_INET6 &&
+ !conn->entry_cfg.ipv6_traffic)) {
+ log_fn(LOG_PROTOCOL_WARN, LD_APP,
+ "Got an EXITPOLICY failure on a connection with a "
+ "mismatched family. Closing.");
+ connection_mark_unattached_ap(conn, END_STREAM_REASON_TORPROTOCOL);
+ return 0;
+ }
+ if (get_options()->ClientDNSRejectInternalAddresses &&
+ tor_addr_is_internal(&addr, 0)) {
+ log_info(LD_APP,"Address '%s' resolved to internal. Closing,",
+ safe_str(conn->socks_request->address));
+ connection_mark_unattached_ap(conn, END_STREAM_REASON_TORPROTOCOL);
+ return 0;
+ }
+
+ client_dns_set_addressmap(conn,
+ conn->socks_request->address, &addr,
+ conn->chosen_exit_name, ttl);
+
+ {
+ char new_addr[TOR_ADDR_BUF_LEN];
+ tor_addr_to_str(new_addr, &addr, sizeof(new_addr), 1);
+ if (strcmp(conn->socks_request->address, new_addr)) {
+ strlcpy(conn->socks_request->address, new_addr,
+ sizeof(conn->socks_request->address));
+ control_event_stream_status(conn, STREAM_EVENT_REMAP, 0);
+ }
+ }
+ }
+ /* check if the exit *ought* to have allowed it */
+
+ adjust_exit_policy_from_exitpolicy_failure(circ,
+ conn,
+ exitrouter,
+ &addr);
+
+ if (conn->chosen_exit_optional ||
+ conn->chosen_exit_retries) {
+ /* stop wanting a specific exit */
+ conn->chosen_exit_optional = 0;
+ /* A non-zero chosen_exit_retries can happen if we set a
+ * TrackHostExits for this address under a port that the exit
+ * relay allows, but then try the same address with a different
+ * port that it doesn't allow to exit. We shouldn't unregister
+ * the mapping, since it is probably still wanted on the
+ * original port. But now we give away to the exit relay that
+ * we probably have a TrackHostExits on it. So be it. */
+ conn->chosen_exit_retries = 0;
+ tor_free(conn->chosen_exit_name); /* clears it */
+ }
+ if (connection_ap_detach_retriable(conn, circ, control_reason) >= 0)
+ return 0;
+ /* else, conn will get closed below */
+ break;
+ }
+ case END_STREAM_REASON_CONNECTREFUSED:
+ if (!conn->chosen_exit_optional)
+ break; /* break means it'll close, below */
+ /* Else fall through: expire this circuit, clear the
+ * chosen_exit_name field, and try again. */
+ /* Falls through. */
+ case END_STREAM_REASON_RESOLVEFAILED:
+ case END_STREAM_REASON_TIMEOUT:
+ case END_STREAM_REASON_MISC:
+ case END_STREAM_REASON_NOROUTE:
+ if (client_dns_incr_failures(conn->socks_request->address)
+ < MAX_RESOLVE_FAILURES) {
+ /* We haven't retried too many times; reattach the connection. */
+ circuit_log_path(LOG_INFO,LD_APP,circ);
+ /* Mark this circuit "unusable for new streams". */
+ mark_circuit_unusable_for_new_conns(circ);
+
+ if (conn->chosen_exit_optional) {
+ /* stop wanting a specific exit */
+ conn->chosen_exit_optional = 0;
+ tor_free(conn->chosen_exit_name); /* clears it */
+ }
+ if (connection_ap_detach_retriable(conn, circ, control_reason) >= 0)
+ return 0;
+ /* else, conn will get closed below */
+ } else {
+ log_notice(LD_APP,
+ "Have tried resolving or connecting to address '%s' "
+ "at %d different places. Giving up.",
+ safe_str(conn->socks_request->address),
+ MAX_RESOLVE_FAILURES);
+ /* clear the failures, so it will have a full try next time */
+ client_dns_clear_failures(conn->socks_request->address);
+ }
+ break;
+ case END_STREAM_REASON_HIBERNATING:
+ case END_STREAM_REASON_RESOURCELIMIT:
+ if (exitrouter) {
+ policies_set_node_exitpolicy_to_reject_all(exitrouter);
+ }
+ if (conn->chosen_exit_optional) {
+ /* stop wanting a specific exit */
+ conn->chosen_exit_optional = 0;
+ tor_free(conn->chosen_exit_name); /* clears it */
+ }
+ if (connection_ap_detach_retriable(conn, circ, control_reason) >= 0)
+ return 0;
+ /* else, will close below */
+ break;
+ } /* end switch */
+ log_info(LD_APP,"Giving up on retrying; conn can't be handled.");
+ }
+
+ log_info(LD_APP,
+ "Edge got end (%s) before we're connected. Marking for close.",
+ stream_end_reason_to_string(rh->length > 0 ? reason : -1));
+ circuit_log_path(LOG_INFO,LD_APP,circ);
+ /* need to test because of detach_retriable */
+ if (!ENTRY_TO_CONN(conn)->marked_for_close)
+ connection_mark_unattached_ap(conn, control_reason);
+ return 0;
+}
+
+/** Called when we have gotten an END_REASON_EXITPOLICY failure on <b>circ</b>
+ * for <b>conn</b>, while attempting to connect via <b>node</b>. If the node
+ * told us which address it rejected, then <b>addr</b> is that address;
+ * otherwise it is AF_UNSPEC.
+ *
+ * If we are sure the node should have allowed this address, mark the node as
+ * having a reject *:* exit policy. Otherwise, mark the circuit as unusable
+ * for this particular address.
+ **/
+static void
+adjust_exit_policy_from_exitpolicy_failure(origin_circuit_t *circ,
+ entry_connection_t *conn,
+ node_t *node,
+ const tor_addr_t *addr)
+{
+ int make_reject_all = 0;
+ const sa_family_t family = tor_addr_family(addr);
+
+ if (node) {
+ tor_addr_t tmp;
+ int asked_for_family = tor_addr_parse(&tmp, conn->socks_request->address);
+ if (family == AF_UNSPEC) {
+ make_reject_all = 1;
+ } else if (node_exit_policy_is_exact(node, family) &&
+ asked_for_family != -1 && !conn->chosen_exit_name) {
+ make_reject_all = 1;
+ }
+
+ if (make_reject_all) {
+ log_info(LD_APP,
+ "Exitrouter %s seems to be more restrictive than its exit "
+ "policy. Not using this router as exit for now.",
+ node_describe(node));
+ policies_set_node_exitpolicy_to_reject_all(node);
+ }
+ }
+
+ if (family != AF_UNSPEC)
+ addr_policy_append_reject_addr(&circ->prepend_policy, addr);
+}
+
+/** Helper: change the socks_request->address field on conn to the
+ * dotted-quad representation of <b>new_addr</b>,
+ * and send an appropriate REMAP event. */
+static void
+remap_event_helper(entry_connection_t *conn, const tor_addr_t *new_addr)
+{
+ tor_addr_to_str(conn->socks_request->address, new_addr,
+ sizeof(conn->socks_request->address),
+ 1);
+ control_event_stream_status(conn, STREAM_EVENT_REMAP,
+ REMAP_STREAM_SOURCE_EXIT);
+}
+
+/** Extract the contents of a connected cell in <b>cell</b>, whose relay
+ * header has already been parsed into <b>rh</b>. On success, set
+ * <b>addr_out</b> to the address we're connected to, and <b>ttl_out</b> to
+ * the ttl of that address, in seconds, and return 0. On failure, return
+ * -1.
+ *
+ * Note that the resulting address can be UNSPEC if the connected cell had no
+ * address (as for a stream to an union service or a tunneled directory
+ * connection), and that the ttl can be absent (in which case <b>ttl_out</b>
+ * is set to -1). */
+STATIC int
+connected_cell_parse(const relay_header_t *rh, const cell_t *cell,
+ tor_addr_t *addr_out, int *ttl_out)
+{
+ uint32_t bytes;
+ const uint8_t *payload = cell->payload + RELAY_HEADER_SIZE;
+
+ tor_addr_make_unspec(addr_out);
+ *ttl_out = -1;
+ if (rh->length == 0)
+ return 0;
+ if (rh->length < 4)
+ return -1;
+ bytes = ntohl(get_uint32(payload));
+
+ /* If bytes is 0, this is maybe a v6 address. Otherwise it's a v4 address */
+ if (bytes != 0) {
+ /* v4 address */
+ tor_addr_from_ipv4h(addr_out, bytes);
+ if (rh->length >= 8) {
+ bytes = ntohl(get_uint32(payload + 4));
+ if (bytes <= INT32_MAX)
+ *ttl_out = bytes;
+ }
+ } else {
+ if (rh->length < 25) /* 4 bytes of 0s, 1 addr, 16 ipv4, 4 ttl. */
+ return -1;
+ if (get_uint8(payload + 4) != 6)
+ return -1;
+ tor_addr_from_ipv6_bytes(addr_out, (char*)(payload + 5));
+ bytes = ntohl(get_uint32(payload + 21));
+ if (bytes <= INT32_MAX)
+ *ttl_out = (int) bytes;
+ }
+ return 0;
+}
+
+/** Drop all storage held by <b>addr</b>. */
+STATIC void
+address_ttl_free_(address_ttl_t *addr)
+{
+ if (!addr)
+ return;
+ tor_free(addr->hostname);
+ tor_free(addr);
+}
+
+/** Parse a resolved cell in <b>cell</b>, with parsed header in <b>rh</b>.
+ * Return -1 on parse error. On success, add one or more newly allocated
+ * address_ttl_t to <b>addresses_out</b>; set *<b>errcode_out</b> to
+ * one of 0, RESOLVED_TYPE_ERROR, or RESOLVED_TYPE_ERROR_TRANSIENT, and
+ * return 0. */
+STATIC int
+resolved_cell_parse(const cell_t *cell, const relay_header_t *rh,
+ smartlist_t *addresses_out, int *errcode_out)
+{
+ const uint8_t *cp;
+ uint8_t answer_type;
+ size_t answer_len;
+ address_ttl_t *addr;
+ size_t remaining;
+ int errcode = 0;
+ smartlist_t *addrs;
+
+ tor_assert(cell);
+ tor_assert(rh);
+ tor_assert(addresses_out);
+ tor_assert(errcode_out);
+
+ *errcode_out = 0;
+
+ if (rh->length > RELAY_PAYLOAD_SIZE)
+ return -1;
+
+ addrs = smartlist_new();
+
+ cp = cell->payload + RELAY_HEADER_SIZE;
+
+ remaining = rh->length;
+ while (remaining) {
+ const uint8_t *cp_orig = cp;
+ if (remaining < 2)
+ goto err;
+ answer_type = *cp++;
+ answer_len = *cp++;
+ if (remaining < 2 + answer_len + 4) {
+ goto err;
+ }
+ if (answer_type == RESOLVED_TYPE_IPV4) {
+ if (answer_len != 4) {
+ goto err;
+ }
+ addr = tor_malloc_zero(sizeof(*addr));
+ tor_addr_from_ipv4n(&addr->addr, get_uint32(cp));
+ cp += 4;
+ addr->ttl = ntohl(get_uint32(cp));
+ cp += 4;
+ smartlist_add(addrs, addr);
+ } else if (answer_type == RESOLVED_TYPE_IPV6) {
+ if (answer_len != 16)
+ goto err;
+ addr = tor_malloc_zero(sizeof(*addr));
+ tor_addr_from_ipv6_bytes(&addr->addr, (const char*) cp);
+ cp += 16;
+ addr->ttl = ntohl(get_uint32(cp));
+ cp += 4;
+ smartlist_add(addrs, addr);
+ } else if (answer_type == RESOLVED_TYPE_HOSTNAME) {
+ if (answer_len == 0) {
+ goto err;
+ }
+ addr = tor_malloc_zero(sizeof(*addr));
+ addr->hostname = tor_memdup_nulterm(cp, answer_len);
+ cp += answer_len;
+ addr->ttl = ntohl(get_uint32(cp));
+ cp += 4;
+ smartlist_add(addrs, addr);
+ } else if (answer_type == RESOLVED_TYPE_ERROR_TRANSIENT ||
+ answer_type == RESOLVED_TYPE_ERROR) {
+ errcode = answer_type;
+ /* Ignore the error contents */
+ cp += answer_len + 4;
+ } else {
+ cp += answer_len + 4;
+ }
+ tor_assert(((ssize_t)remaining) >= (cp - cp_orig));
+ remaining -= (cp - cp_orig);
+ }
+
+ if (errcode && smartlist_len(addrs) == 0) {
+ /* Report an error only if there were no results. */
+ *errcode_out = errcode;
+ }
+
+ smartlist_add_all(addresses_out, addrs);
+ smartlist_free(addrs);
+
+ return 0;
+
+ err:
+ /* On parse error, don't report any results */
+ SMARTLIST_FOREACH(addrs, address_ttl_t *, a, address_ttl_free(a));
+ smartlist_free(addrs);
+ return -1;
+}
+
+/** Helper for connection_edge_process_resolved_cell: given an error code,
+ * an entry_connection, and a list of address_ttl_t *, report the best answer
+ * to the entry_connection. */
+static void
+connection_ap_handshake_socks_got_resolved_cell(entry_connection_t *conn,
+ int error_code,
+ smartlist_t *results)
+{
+ address_ttl_t *addr_ipv4 = NULL;
+ address_ttl_t *addr_ipv6 = NULL;
+ address_ttl_t *addr_hostname = NULL;
+ address_ttl_t *addr_best = NULL;
+
+ /* If it's an error code, that's easy. */
+ if (error_code) {
+ tor_assert(error_code == RESOLVED_TYPE_ERROR ||
+ error_code == RESOLVED_TYPE_ERROR_TRANSIENT);
+ connection_ap_handshake_socks_resolved(conn,
+ error_code,0,NULL,-1,-1);
+ return;
+ }
+
+ /* Get the first answer of each type. */
+ SMARTLIST_FOREACH_BEGIN(results, address_ttl_t *, addr) {
+ if (addr->hostname) {
+ if (!addr_hostname) {
+ addr_hostname = addr;
+ }
+ } else if (tor_addr_family(&addr->addr) == AF_INET) {
+ if (!addr_ipv4 && conn->entry_cfg.ipv4_traffic) {
+ addr_ipv4 = addr;
+ }
+ } else if (tor_addr_family(&addr->addr) == AF_INET6) {
+ if (!addr_ipv6 && conn->entry_cfg.ipv6_traffic) {
+ addr_ipv6 = addr;
+ }
+ }
+ } SMARTLIST_FOREACH_END(addr);
+
+ /* Now figure out which type we wanted to deliver. */
+ if (conn->socks_request->command == SOCKS_COMMAND_RESOLVE_PTR) {
+ if (addr_hostname) {
+ connection_ap_handshake_socks_resolved(conn,
+ RESOLVED_TYPE_HOSTNAME,
+ strlen(addr_hostname->hostname),
+ (uint8_t*)addr_hostname->hostname,
+ addr_hostname->ttl,-1);
+ } else {
+ connection_ap_handshake_socks_resolved(conn,
+ RESOLVED_TYPE_ERROR,0,NULL,-1,-1);
+ }
+ return;
+ }
+
+ if (conn->entry_cfg.prefer_ipv6) {
+ addr_best = addr_ipv6 ? addr_ipv6 : addr_ipv4;
+ } else {
+ addr_best = addr_ipv4 ? addr_ipv4 : addr_ipv6;
+ }
+
+ /* Now convert it to the ugly old interface */
+ if (! addr_best) {
+ connection_ap_handshake_socks_resolved(conn,
+ RESOLVED_TYPE_ERROR,0,NULL,-1,-1);
+ return;
+ }
+
+ connection_ap_handshake_socks_resolved_addr(conn,
+ &addr_best->addr,
+ addr_best->ttl,
+ -1);
+
+ remap_event_helper(conn, &addr_best->addr);
+}
+
+/** Handle a RELAY_COMMAND_RESOLVED cell that we received on a non-open AP
+ * stream. */
+STATIC int
+connection_edge_process_resolved_cell(edge_connection_t *conn,
+ const cell_t *cell,
+ const relay_header_t *rh)
+{
+ entry_connection_t *entry_conn = EDGE_TO_ENTRY_CONN(conn);
+ smartlist_t *resolved_addresses = NULL;
+ int errcode = 0;
+
+ if (conn->base_.state != AP_CONN_STATE_RESOLVE_WAIT) {
+ log_fn(LOG_PROTOCOL_WARN, LD_APP, "Got a 'resolved' cell while "
+ "not in state resolve_wait. Dropping.");
+ return 0;
+ }
+ tor_assert(SOCKS_COMMAND_IS_RESOLVE(entry_conn->socks_request->command));
+
+ resolved_addresses = smartlist_new();
+ if (resolved_cell_parse(cell, rh, resolved_addresses, &errcode)) {
+ log_fn(LOG_PROTOCOL_WARN, LD_PROTOCOL,
+ "Dropping malformed 'resolved' cell");
+ connection_mark_unattached_ap(entry_conn, END_STREAM_REASON_TORPROTOCOL);
+ goto done;
+ }
+
+ if (get_options()->ClientDNSRejectInternalAddresses) {
+ int orig_len = smartlist_len(resolved_addresses);
+ SMARTLIST_FOREACH_BEGIN(resolved_addresses, address_ttl_t *, addr) {
+ if (addr->hostname == NULL && tor_addr_is_internal(&addr->addr, 0)) {
+ log_info(LD_APP, "Got a resolved cell with answer %s; dropping that "
+ "answer.",
+ safe_str_client(fmt_addr(&addr->addr)));
+ address_ttl_free(addr);
+ SMARTLIST_DEL_CURRENT(resolved_addresses, addr);
+ }
+ } SMARTLIST_FOREACH_END(addr);
+ if (orig_len && smartlist_len(resolved_addresses) == 0) {
+ log_info(LD_APP, "Got a resolved cell with only private addresses; "
+ "dropping it.");
+ connection_ap_handshake_socks_resolved(entry_conn,
+ RESOLVED_TYPE_ERROR_TRANSIENT,
+ 0, NULL, 0, TIME_MAX);
+ connection_mark_unattached_ap(entry_conn,
+ END_STREAM_REASON_TORPROTOCOL);
+ goto done;
+ }
+ }
+
+ /* This is valid data at this point. Count it */
+ if (conn->on_circuit && CIRCUIT_IS_ORIGIN(conn->on_circuit)) {
+ circuit_read_valid_data(TO_ORIGIN_CIRCUIT(conn->on_circuit),
+ rh->length);
+ }
+
+ connection_ap_handshake_socks_got_resolved_cell(entry_conn,
+ errcode,
+ resolved_addresses);
+
+ connection_mark_unattached_ap(entry_conn,
+ END_STREAM_REASON_DONE |
+ END_STREAM_REASON_FLAG_ALREADY_SOCKS_REPLIED);
+
+ done:
+ SMARTLIST_FOREACH(resolved_addresses, address_ttl_t *, addr,
+ address_ttl_free(addr));
+ smartlist_free(resolved_addresses);
+ return 0;
+}
+
+/** An incoming relay cell has arrived from circuit <b>circ</b> to
+ * stream <b>conn</b>.
+ *
+ * The arguments here are the same as in
+ * connection_edge_process_relay_cell() below; this function is called
+ * from there when <b>conn</b> is defined and not in an open state.
+ */
+static int
+connection_edge_process_relay_cell_not_open(
+ relay_header_t *rh, cell_t *cell, circuit_t *circ,
+ edge_connection_t *conn, crypt_path_t *layer_hint)
+{
+ if (rh->command == RELAY_COMMAND_END) {
+ if (CIRCUIT_IS_ORIGIN(circ) && conn->base_.type == CONN_TYPE_AP) {
+ return connection_ap_process_end_not_open(rh, cell,
+ TO_ORIGIN_CIRCUIT(circ),
+ EDGE_TO_ENTRY_CONN(conn),
+ layer_hint);
+ } else {
+ /* we just got an 'end', don't need to send one */
+ conn->edge_has_sent_end = 1;
+ conn->end_reason = *(cell->payload+RELAY_HEADER_SIZE) |
+ END_STREAM_REASON_FLAG_REMOTE;
+ connection_mark_for_close(TO_CONN(conn));
+ return 0;
+ }
+ }
+
+ if (conn->base_.type == CONN_TYPE_AP &&
+ rh->command == RELAY_COMMAND_CONNECTED) {
+ tor_addr_t addr;
+ int ttl;
+ entry_connection_t *entry_conn = EDGE_TO_ENTRY_CONN(conn);
+ tor_assert(CIRCUIT_IS_ORIGIN(circ));
+ if (conn->base_.state != AP_CONN_STATE_CONNECT_WAIT) {
+ log_fn(LOG_PROTOCOL_WARN, LD_APP,
+ "Got 'connected' while not in state connect_wait. Dropping.");
+ return 0;
+ }
+ CONNECTION_AP_EXPECT_NONPENDING(entry_conn);
+ conn->base_.state = AP_CONN_STATE_OPEN;
+ log_info(LD_APP,"'connected' received for circid %u streamid %d "
+ "after %d seconds.",
+ (unsigned)circ->n_circ_id,
+ rh->stream_id,
+ (int)(time(NULL) - conn->base_.timestamp_last_read_allowed));
+ if (connected_cell_parse(rh, cell, &addr, &ttl) < 0) {
+ log_fn(LOG_PROTOCOL_WARN, LD_APP,
+ "Got a badly formatted connected cell. Closing.");
+ connection_edge_end(conn, END_STREAM_REASON_TORPROTOCOL);
+ connection_mark_unattached_ap(entry_conn, END_STREAM_REASON_TORPROTOCOL);
+ return 0;
+ }
+ if (tor_addr_family(&addr) != AF_UNSPEC) {
+ /* The family is not UNSPEC: so we were given an address in the
+ * connected cell. (This is normal, except for BEGINDIR and onion
+ * service streams.) */
+ const sa_family_t family = tor_addr_family(&addr);
+ if (tor_addr_is_null(&addr) ||
+ (get_options()->ClientDNSRejectInternalAddresses &&
+ tor_addr_is_internal(&addr, 0))) {
+ log_info(LD_APP, "...but it claims the IP address was %s. Closing.",
+ fmt_addr(&addr));
+ connection_edge_end(conn, END_STREAM_REASON_TORPROTOCOL);
+ connection_mark_unattached_ap(entry_conn,
+ END_STREAM_REASON_TORPROTOCOL);
+ return 0;
+ }
+
+ if ((family == AF_INET && ! entry_conn->entry_cfg.ipv4_traffic) ||
+ (family == AF_INET6 && ! entry_conn->entry_cfg.ipv6_traffic)) {
+ log_fn(LOG_PROTOCOL_WARN, LD_APP,
+ "Got a connected cell to %s with unsupported address family."
+ " Closing.", fmt_addr(&addr));
+ connection_edge_end(conn, END_STREAM_REASON_TORPROTOCOL);
+ connection_mark_unattached_ap(entry_conn,
+ END_STREAM_REASON_TORPROTOCOL);
+ return 0;
+ }
+
+ client_dns_set_addressmap(entry_conn,
+ entry_conn->socks_request->address, &addr,
+ entry_conn->chosen_exit_name, ttl);
+
+ remap_event_helper(entry_conn, &addr);
+ }
+ circuit_log_path(LOG_INFO,LD_APP,TO_ORIGIN_CIRCUIT(circ));
+ /* don't send a socks reply to transparent conns */
+ tor_assert(entry_conn->socks_request != NULL);
+ if (!entry_conn->socks_request->has_finished) {
+ connection_ap_handshake_socks_reply(entry_conn, NULL, 0, 0);
+ }
+
+ /* Was it a linked dir conn? If so, a dir request just started to
+ * fetch something; this could be a bootstrap status milestone. */
+ log_debug(LD_APP, "considering");
+ if (TO_CONN(conn)->linked_conn &&
+ TO_CONN(conn)->linked_conn->type == CONN_TYPE_DIR) {
+ connection_t *dirconn = TO_CONN(conn)->linked_conn;
+ log_debug(LD_APP, "it is! %d", dirconn->purpose);
+ switch (dirconn->purpose) {
+ case DIR_PURPOSE_FETCH_CERTIFICATE:
+ if (consensus_is_waiting_for_certs())
+ control_event_bootstrap(BOOTSTRAP_STATUS_LOADING_KEYS, 0);
+ break;
+ case DIR_PURPOSE_FETCH_CONSENSUS:
+ control_event_bootstrap(BOOTSTRAP_STATUS_LOADING_STATUS, 0);
+ break;
+ case DIR_PURPOSE_FETCH_SERVERDESC:
+ case DIR_PURPOSE_FETCH_MICRODESC:
+ if (TO_DIR_CONN(dirconn)->router_purpose == ROUTER_PURPOSE_GENERAL)
+ control_event_boot_dir(BOOTSTRAP_STATUS_LOADING_DESCRIPTORS,
+ count_loading_descriptors_progress());
+ break;
+ }
+ }
+ /* This is definitely a success, so forget about any pending data we
+ * had sent. */
+ if (entry_conn->pending_optimistic_data) {
+ buf_free(entry_conn->pending_optimistic_data);
+ entry_conn->pending_optimistic_data = NULL;
+ }
+
+ /* This is valid data at this point. Count it */
+ circuit_read_valid_data(TO_ORIGIN_CIRCUIT(circ), rh->length);
+
+ /* handle anything that might have queued */
+ if (connection_edge_package_raw_inbuf(conn, 1, NULL) < 0) {
+ /* (We already sent an end cell if possible) */
+ connection_mark_for_close(TO_CONN(conn));
+ return 0;
+ }
+ return 0;
+ }
+ if (conn->base_.type == CONN_TYPE_AP &&
+ rh->command == RELAY_COMMAND_RESOLVED) {
+ return connection_edge_process_resolved_cell(conn, cell, rh);
+ }
+
+ log_fn(LOG_PROTOCOL_WARN, LD_PROTOCOL,
+ "Got an unexpected relay command %d, in state %d (%s). Dropping.",
+ rh->command, conn->base_.state,
+ conn_state_to_string(conn->base_.type, conn->base_.state));
+ return 0; /* for forward compatibility, don't kill the circuit */
+// connection_edge_end(conn, END_STREAM_REASON_TORPROTOCOL);
+// connection_mark_for_close(conn);
+// return -1;
+}
+
+/** An incoming relay cell has arrived on circuit <b>circ</b>. If
+ * <b>conn</b> is NULL this is a control cell, else <b>cell</b> is
+ * destined for <b>conn</b>.
+ *
+ * If <b>layer_hint</b> is defined, then we're the origin of the
+ * circuit, and it specifies the hop that packaged <b>cell</b>.
+ *
+ * Return -reason if you want to warn and tear down the circuit, else 0.
+ */
+STATIC int
+connection_edge_process_relay_cell(cell_t *cell, circuit_t *circ,
+ edge_connection_t *conn,
+ crypt_path_t *layer_hint)
+{
+ static int num_seen=0;
+ relay_header_t rh;
+ unsigned domain = layer_hint?LD_APP:LD_EXIT;
+ int reason;
+ int optimistic_data = 0; /* Set to 1 if we receive data on a stream
+ * that's in the EXIT_CONN_STATE_RESOLVING
+ * or EXIT_CONN_STATE_CONNECTING states. */
+
+ tor_assert(cell);
+ tor_assert(circ);
+
+ relay_header_unpack(&rh, cell->payload);
+// log_fn(LOG_DEBUG,"command %d stream %d", rh.command, rh.stream_id);
+ num_seen++;
+ log_debug(domain, "Now seen %d relay cells here (command %d, stream %d).",
+ num_seen, rh.command, rh.stream_id);
+
+ if (rh.length > RELAY_PAYLOAD_SIZE) {
+ log_fn(LOG_PROTOCOL_WARN, LD_PROTOCOL,
+ "Relay cell length field too long. Closing circuit.");
+ return - END_CIRC_REASON_TORPROTOCOL;
+ }
+
+ if (rh.stream_id == 0) {
+ switch (rh.command) {
+ case RELAY_COMMAND_BEGIN:
+ case RELAY_COMMAND_CONNECTED:
+ case RELAY_COMMAND_END:
+ case RELAY_COMMAND_RESOLVE:
+ case RELAY_COMMAND_RESOLVED:
+ case RELAY_COMMAND_BEGIN_DIR:
+ log_fn(LOG_PROTOCOL_WARN, LD_PROTOCOL, "Relay command %d with zero "
+ "stream_id. Dropping.", (int)rh.command);
+ return 0;
+ default:
+ ;
+ }
+ }
+
+ /* either conn is NULL, in which case we've got a control cell, or else
+ * conn points to the recognized stream. */
+
+ if (conn && !connection_state_is_open(TO_CONN(conn))) {
+ if (conn->base_.type == CONN_TYPE_EXIT &&
+ (conn->base_.state == EXIT_CONN_STATE_CONNECTING ||
+ conn->base_.state == EXIT_CONN_STATE_RESOLVING) &&
+ rh.command == RELAY_COMMAND_DATA) {
+ /* Allow DATA cells to be delivered to an exit node in state
+ * EXIT_CONN_STATE_CONNECTING or EXIT_CONN_STATE_RESOLVING.
+ * This speeds up HTTP, for example. */
+ optimistic_data = 1;
+ } else if (rh.stream_id == 0 && rh.command == RELAY_COMMAND_DATA) {
+ log_warn(LD_BUG, "Somehow I had a connection that matched a "
+ "data cell with stream ID 0.");
+ } else {
+ return connection_edge_process_relay_cell_not_open(
+ &rh, cell, circ, conn, layer_hint);
+ }
+ }
+
+ switch (rh.command) {
+ case RELAY_COMMAND_DROP:
+ rep_hist_padding_count_read(PADDING_TYPE_DROP);
+// log_info(domain,"Got a relay-level padding cell. Dropping.");
+ return 0;
+ case RELAY_COMMAND_BEGIN:
+ case RELAY_COMMAND_BEGIN_DIR:
+ if (layer_hint &&
+ circ->purpose != CIRCUIT_PURPOSE_S_REND_JOINED) {
+ log_fn(LOG_PROTOCOL_WARN, LD_APP,
+ "Relay begin request unsupported at AP. Dropping.");
+ return 0;
+ }
+ if (circ->purpose == CIRCUIT_PURPOSE_S_REND_JOINED &&
+ layer_hint != TO_ORIGIN_CIRCUIT(circ)->cpath->prev) {
+ log_fn(LOG_PROTOCOL_WARN, LD_APP,
+ "Relay begin request to Hidden Service "
+ "from intermediary node. Dropping.");
+ return 0;
+ }
+ if (conn) {
+ log_fn(LOG_PROTOCOL_WARN, domain,
+ "Begin cell for known stream. Dropping.");
+ return 0;
+ }
+ if (rh.command == RELAY_COMMAND_BEGIN_DIR &&
+ circ->purpose != CIRCUIT_PURPOSE_S_REND_JOINED) {
+ /* Assign this circuit and its app-ward OR connection a unique ID,
+ * so that we can measure download times. The local edge and dir
+ * connection will be assigned the same ID when they are created
+ * and linked. */
+ static uint64_t next_id = 0;
+ circ->dirreq_id = ++next_id;
+ TO_OR_CIRCUIT(circ)->p_chan->dirreq_id = circ->dirreq_id;
+ }
+ return connection_exit_begin_conn(cell, circ);
+ case RELAY_COMMAND_DATA:
+ ++stats_n_data_cells_received;
+ if (( layer_hint && --layer_hint->deliver_window < 0) ||
+ (!layer_hint && --circ->deliver_window < 0)) {
+ log_fn(LOG_PROTOCOL_WARN, LD_PROTOCOL,
+ "(relay data) circ deliver_window below 0. Killing.");
+ if (conn) {
+ /* XXXX Do we actually need to do this? Will killing the circuit
+ * not send an END and mark the stream for close as appropriate? */
+ connection_edge_end(conn, END_STREAM_REASON_TORPROTOCOL);
+ connection_mark_for_close(TO_CONN(conn));
+ }
+ return -END_CIRC_REASON_TORPROTOCOL;
+ }
+ log_debug(domain,"circ deliver_window now %d.", layer_hint ?
+ layer_hint->deliver_window : circ->deliver_window);
+
+ circuit_consider_sending_sendme(circ, layer_hint);
+
+ if (rh.stream_id == 0) {
+ log_fn(LOG_PROTOCOL_WARN, LD_PROTOCOL, "Relay data cell with zero "
+ "stream_id. Dropping.");
+ return 0;
+ } else if (!conn) {
+ if (CIRCUIT_IS_ORIGIN(circ)) {
+ origin_circuit_t *ocirc = TO_ORIGIN_CIRCUIT(circ);
+ if (connection_half_edge_is_valid_data(ocirc->half_streams,
+ rh.stream_id)) {
+ circuit_read_valid_data(ocirc, rh.length);
+ log_info(domain,
+ "data cell on circ %u valid on half-closed "
+ "stream id %d", ocirc->global_identifier, rh.stream_id);
+ }
+ }
+
+ log_info(domain,"data cell dropped, unknown stream (streamid %d).",
+ rh.stream_id);
+ return 0;
+ }
+
+ if (--conn->deliver_window < 0) { /* is it below 0 after decrement? */
+ log_fn(LOG_PROTOCOL_WARN, LD_PROTOCOL,
+ "(relay data) conn deliver_window below 0. Killing.");
+ return -END_CIRC_REASON_TORPROTOCOL;
+ }
+ /* Total all valid application bytes delivered */
+ if (CIRCUIT_IS_ORIGIN(circ) && rh.length > 0) {
+ circuit_read_valid_data(TO_ORIGIN_CIRCUIT(circ), rh.length);
+ }
+
+ stats_n_data_bytes_received += rh.length;
+ connection_buf_add((char*)(cell->payload + RELAY_HEADER_SIZE),
+ rh.length, TO_CONN(conn));
+
+#ifdef MEASUREMENTS_21206
+ /* Count number of RELAY_DATA cells received on a linked directory
+ * connection. */
+ connection_t *linked_conn = TO_CONN(conn)->linked_conn;
+
+ if (linked_conn && linked_conn->type == CONN_TYPE_DIR) {
+ ++(TO_DIR_CONN(linked_conn)->data_cells_received);
+ }
+#endif /* defined(MEASUREMENTS_21206) */
+
+ if (!optimistic_data) {
+ /* Only send a SENDME if we're not getting optimistic data; otherwise
+ * a SENDME could arrive before the CONNECTED.
+ */
+ connection_edge_consider_sending_sendme(conn);
+ }
+
+ return 0;
+ case RELAY_COMMAND_END:
+ reason = rh.length > 0 ?
+ get_uint8(cell->payload+RELAY_HEADER_SIZE) : END_STREAM_REASON_MISC;
+ if (!conn) {
+ if (CIRCUIT_IS_ORIGIN(circ)) {
+ origin_circuit_t *ocirc = TO_ORIGIN_CIRCUIT(circ);
+ if (connection_half_edge_is_valid_end(ocirc->half_streams,
+ rh.stream_id)) {
+
+ circuit_read_valid_data(ocirc, rh.length);
+ log_info(domain,
+ "end cell (%s) on circ %u valid on half-closed "
+ "stream id %d",
+ stream_end_reason_to_string(reason),
+ ocirc->global_identifier, rh.stream_id);
+ return 0;
+ }
+ }
+ log_info(domain,"end cell (%s) dropped, unknown stream.",
+ stream_end_reason_to_string(reason));
+ return 0;
+ }
+/* XXX add to this log_fn the exit node's nickname? */
+ log_info(domain,TOR_SOCKET_T_FORMAT": end cell (%s) for stream %d. "
+ "Removing stream.",
+ conn->base_.s,
+ stream_end_reason_to_string(reason),
+ conn->stream_id);
+ if (conn->base_.type == CONN_TYPE_AP) {
+ entry_connection_t *entry_conn = EDGE_TO_ENTRY_CONN(conn);
+ if (entry_conn->socks_request &&
+ !entry_conn->socks_request->has_finished)
+ log_warn(LD_BUG,
+ "open stream hasn't sent socks answer yet? Closing.");
+ }
+ /* We just *got* an end; no reason to send one. */
+ conn->edge_has_sent_end = 1;
+ if (!conn->end_reason)
+ conn->end_reason = reason | END_STREAM_REASON_FLAG_REMOTE;
+ if (!conn->base_.marked_for_close) {
+ /* only mark it if not already marked. it's possible to
+ * get the 'end' right around when the client hangs up on us. */
+ connection_mark_and_flush(TO_CONN(conn));
+
+ /* Total all valid application bytes delivered */
+ if (CIRCUIT_IS_ORIGIN(circ)) {
+ circuit_read_valid_data(TO_ORIGIN_CIRCUIT(circ), rh.length);
+ }
+ }
+ return 0;
+ case RELAY_COMMAND_EXTEND:
+ case RELAY_COMMAND_EXTEND2: {
+ static uint64_t total_n_extend=0, total_nonearly=0;
+ total_n_extend++;
+ if (rh.stream_id) {
+ log_fn(LOG_PROTOCOL_WARN, domain,
+ "'extend' cell received for non-zero stream. Dropping.");
+ return 0;
+ }
+ if (cell->command != CELL_RELAY_EARLY &&
+ !networkstatus_get_param(NULL,"AllowNonearlyExtend",0,0,1)) {
+#define EARLY_WARNING_INTERVAL 3600
+ static ratelim_t early_warning_limit =
+ RATELIM_INIT(EARLY_WARNING_INTERVAL);
+ char *m;
+ if (cell->command == CELL_RELAY) {
+ ++total_nonearly;
+ if ((m = rate_limit_log(&early_warning_limit, approx_time()))) {
+ double percentage = ((double)total_nonearly)/total_n_extend;
+ percentage *= 100;
+ log_fn(LOG_PROTOCOL_WARN, domain, "EXTEND cell received, "
+ "but not via RELAY_EARLY. Dropping.%s", m);
+ log_fn(LOG_PROTOCOL_WARN, domain, " (We have dropped %.02f%% of "
+ "all EXTEND cells for this reason)", percentage);
+ tor_free(m);
+ }
+ } else {
+ log_fn(LOG_WARN, domain,
+ "EXTEND cell received, in a cell with type %d! Dropping.",
+ cell->command);
+ }
+ return 0;
+ }
+ return circuit_extend(cell, circ);
+ }
+ case RELAY_COMMAND_EXTENDED:
+ case RELAY_COMMAND_EXTENDED2:
+ if (!layer_hint) {
+ log_fn(LOG_PROTOCOL_WARN, LD_PROTOCOL,
+ "'extended' unsupported at non-origin. Dropping.");
+ return 0;
+ }
+ log_debug(domain,"Got an extended cell! Yay.");
+ {
+ extended_cell_t extended_cell;
+ if (extended_cell_parse(&extended_cell, rh.command,
+ (const uint8_t*)cell->payload+RELAY_HEADER_SIZE,
+ rh.length)<0) {
+ log_warn(LD_PROTOCOL,
+ "Can't parse EXTENDED cell; killing circuit.");
+ return -END_CIRC_REASON_TORPROTOCOL;
+ }
+ if ((reason = circuit_finish_handshake(TO_ORIGIN_CIRCUIT(circ),
+ &extended_cell.created_cell)) < 0) {
+ circuit_mark_for_close(circ, -reason);
+ return 0; /* We don't want to cause a warning, so we mark the circuit
+ * here. */
+ }
+ }
+ if ((reason=circuit_send_next_onion_skin(TO_ORIGIN_CIRCUIT(circ)))<0) {
+ log_info(domain,"circuit_send_next_onion_skin() failed.");
+ return reason;
+ }
+ /* Total all valid bytes delivered. */
+ if (CIRCUIT_IS_ORIGIN(circ)) {
+ circuit_read_valid_data(TO_ORIGIN_CIRCUIT(circ), rh.length);
+ }
+ return 0;
+ case RELAY_COMMAND_TRUNCATE:
+ if (layer_hint) {
+ log_fn(LOG_PROTOCOL_WARN, LD_APP,
+ "'truncate' unsupported at origin. Dropping.");
+ return 0;
+ }
+ if (circ->n_hop) {
+ if (circ->n_chan)
+ log_warn(LD_BUG, "n_chan and n_hop set on the same circuit!");
+ extend_info_free(circ->n_hop);
+ circ->n_hop = NULL;
+ tor_free(circ->n_chan_create_cell);
+ circuit_set_state(circ, CIRCUIT_STATE_OPEN);
+ }
+ if (circ->n_chan) {
+ uint8_t trunc_reason = get_uint8(cell->payload + RELAY_HEADER_SIZE);
++ circuit_synchronize_written_or_bandwidth(circ, CIRCUIT_N_CHAN);
+ circuit_clear_cell_queue(circ, circ->n_chan);
+ channel_send_destroy(circ->n_circ_id, circ->n_chan,
+ trunc_reason);
+ circuit_set_n_circid_chan(circ, 0, NULL);
+ }
+ log_debug(LD_EXIT, "Processed 'truncate', replying.");
+ {
+ char payload[1];
+ payload[0] = (char)END_CIRC_REASON_REQUESTED;
+ relay_send_command_from_edge(0, circ, RELAY_COMMAND_TRUNCATED,
+ payload, sizeof(payload), NULL);
+ }
+ return 0;
+ case RELAY_COMMAND_TRUNCATED:
+ if (!layer_hint) {
+ log_fn(LOG_PROTOCOL_WARN, LD_EXIT,
+ "'truncated' unsupported at non-origin. Dropping.");
+ return 0;
+ }
+
+ /* Count the truncated as valid, for completeness. The
+ * circuit is being torn down anyway, though. */
+ if (CIRCUIT_IS_ORIGIN(circ)) {
+ circuit_read_valid_data(TO_ORIGIN_CIRCUIT(circ),
+ rh.length);
+ }
+ circuit_truncated(TO_ORIGIN_CIRCUIT(circ),
+ get_uint8(cell->payload + RELAY_HEADER_SIZE));
+ return 0;
+ case RELAY_COMMAND_CONNECTED:
+ if (conn) {
+ log_fn(LOG_PROTOCOL_WARN, LD_PROTOCOL,
+ "'connected' unsupported while open. Closing circ.");
+ return -END_CIRC_REASON_TORPROTOCOL;
+ }
+
+ if (CIRCUIT_IS_ORIGIN(circ)) {
+ origin_circuit_t *ocirc = TO_ORIGIN_CIRCUIT(circ);
+ if (connection_half_edge_is_valid_connected(ocirc->half_streams,
+ rh.stream_id)) {
+ circuit_read_valid_data(ocirc, rh.length);
+ log_info(domain,
+ "connected cell on circ %u valid on half-closed "
+ "stream id %d", ocirc->global_identifier, rh.stream_id);
+ return 0;
+ }
+ }
+
+ log_info(domain,
+ "'connected' received on circid %u for streamid %d, "
+ "no conn attached anymore. Ignoring.",
+ (unsigned)circ->n_circ_id, rh.stream_id);
+ return 0;
+ case RELAY_COMMAND_SENDME:
+ if (!rh.stream_id) {
+ if (layer_hint) {
+ if (layer_hint->package_window + CIRCWINDOW_INCREMENT >
+ CIRCWINDOW_START_MAX) {
+ static struct ratelim_t exit_warn_ratelim = RATELIM_INIT(600);
+ log_fn_ratelim(&exit_warn_ratelim, LOG_WARN, LD_PROTOCOL,
+ "Unexpected sendme cell from exit relay. "
+ "Closing circ.");
+ return -END_CIRC_REASON_TORPROTOCOL;
+ }
+ layer_hint->package_window += CIRCWINDOW_INCREMENT;
+ log_debug(LD_APP,"circ-level sendme at origin, packagewindow %d.",
+ layer_hint->package_window);
+ circuit_resume_edge_reading(circ, layer_hint);
+
+ /* We count circuit-level sendme's as valid delivered data because
+ * they are rate limited.
+ */
+ if (CIRCUIT_IS_ORIGIN(circ)) {
+ circuit_read_valid_data(TO_ORIGIN_CIRCUIT(circ),
+ rh.length);
+ }
+
+ } else {
+ if (circ->package_window + CIRCWINDOW_INCREMENT >
+ CIRCWINDOW_START_MAX) {
+ static struct ratelim_t client_warn_ratelim = RATELIM_INIT(600);
+ log_fn_ratelim(&client_warn_ratelim,LOG_PROTOCOL_WARN, LD_PROTOCOL,
+ "Unexpected sendme cell from client. "
+ "Closing circ (window %d).",
+ circ->package_window);
+ return -END_CIRC_REASON_TORPROTOCOL;
+ }
+ circ->package_window += CIRCWINDOW_INCREMENT;
+ log_debug(LD_APP,
+ "circ-level sendme at non-origin, packagewindow %d.",
+ circ->package_window);
+ circuit_resume_edge_reading(circ, layer_hint);
+ }
+ return 0;
+ }
+ if (!conn) {
+ if (CIRCUIT_IS_ORIGIN(circ)) {
+ origin_circuit_t *ocirc = TO_ORIGIN_CIRCUIT(circ);
+ if (connection_half_edge_is_valid_sendme(ocirc->half_streams,
+ rh.stream_id)) {
+ circuit_read_valid_data(ocirc, rh.length);
+ log_info(domain,
+ "sendme cell on circ %u valid on half-closed "
+ "stream id %d", ocirc->global_identifier, rh.stream_id);
+ }
+ }
+
+ log_info(domain,"sendme cell dropped, unknown stream (streamid %d).",
+ rh.stream_id);
+ return 0;
+ }
+
+ /* Don't allow the other endpoint to request more than our maximum
+ * (i.e. initial) stream SENDME window worth of data. Well-behaved
+ * stock clients will not request more than this max (as per the check
+ * in the while loop of connection_edge_consider_sending_sendme()).
+ */
+ if (conn->package_window + STREAMWINDOW_INCREMENT >
+ STREAMWINDOW_START_MAX) {
+ static struct ratelim_t stream_warn_ratelim = RATELIM_INIT(600);
+ log_fn_ratelim(&stream_warn_ratelim, LOG_PROTOCOL_WARN, LD_PROTOCOL,
+ "Unexpected stream sendme cell. Closing circ (window %d).",
+ conn->package_window);
+ return -END_CIRC_REASON_TORPROTOCOL;
+ }
+
+ /* At this point, the stream sendme is valid */
+ if (CIRCUIT_IS_ORIGIN(circ)) {
+ circuit_read_valid_data(TO_ORIGIN_CIRCUIT(circ),
+ rh.length);
+ }
+
+ conn->package_window += STREAMWINDOW_INCREMENT;
+ log_debug(domain,"stream-level sendme, packagewindow now %d.",
+ conn->package_window);
+ if (circuit_queue_streams_are_blocked(circ)) {
+ /* Still waiting for queue to flush; don't touch conn */
+ return 0;
+ }
+ connection_start_reading(TO_CONN(conn));
+ /* handle whatever might still be on the inbuf */
+ if (connection_edge_package_raw_inbuf(conn, 1, NULL) < 0) {
+ /* (We already sent an end cell if possible) */
+ connection_mark_for_close(TO_CONN(conn));
+ return 0;
+ }
+ return 0;
+ case RELAY_COMMAND_RESOLVE:
+ if (layer_hint) {
+ log_fn(LOG_PROTOCOL_WARN, LD_APP,
+ "resolve request unsupported at AP; dropping.");
+ return 0;
+ } else if (conn) {
+ log_fn(LOG_PROTOCOL_WARN, domain,
+ "resolve request for known stream; dropping.");
+ return 0;
+ } else if (circ->purpose != CIRCUIT_PURPOSE_OR) {
+ log_fn(LOG_PROTOCOL_WARN, domain,
+ "resolve request on circ with purpose %d; dropping",
+ circ->purpose);
+ return 0;
+ }
+ connection_exit_begin_resolve(cell, TO_OR_CIRCUIT(circ));
+ return 0;
+ case RELAY_COMMAND_RESOLVED:
+ if (conn) {
+ log_fn(LOG_PROTOCOL_WARN, domain,
+ "'resolved' unsupported while open. Closing circ.");
+ return -END_CIRC_REASON_TORPROTOCOL;
+ }
+
+ if (CIRCUIT_IS_ORIGIN(circ)) {
+ origin_circuit_t *ocirc = TO_ORIGIN_CIRCUIT(circ);
+ if (connection_half_edge_is_valid_resolved(ocirc->half_streams,
+ rh.stream_id)) {
+ circuit_read_valid_data(ocirc, rh.length);
+ log_info(domain,
+ "resolved cell on circ %u valid on half-closed "
+ "stream id %d", ocirc->global_identifier, rh.stream_id);
+ return 0;
+ }
+ }
+
+ log_info(domain,
+ "'resolved' received, no conn attached anymore. Ignoring.");
+ return 0;
+ case RELAY_COMMAND_ESTABLISH_INTRO:
+ case RELAY_COMMAND_ESTABLISH_RENDEZVOUS:
+ case RELAY_COMMAND_INTRODUCE1:
+ case RELAY_COMMAND_INTRODUCE2:
+ case RELAY_COMMAND_INTRODUCE_ACK:
+ case RELAY_COMMAND_RENDEZVOUS1:
+ case RELAY_COMMAND_RENDEZVOUS2:
+ case RELAY_COMMAND_INTRO_ESTABLISHED:
+ case RELAY_COMMAND_RENDEZVOUS_ESTABLISHED:
+ rend_process_relay_cell(circ, layer_hint,
+ rh.command, rh.length,
+ cell->payload+RELAY_HEADER_SIZE);
+ return 0;
+ }
+ log_fn(LOG_PROTOCOL_WARN, LD_PROTOCOL,
+ "Received unknown relay command %d. Perhaps the other side is using "
+ "a newer version of Tor? Dropping.",
+ rh.command);
+ return 0; /* for forward compatibility, don't kill the circuit */
+}
+
+/** How many relay_data cells have we built, ever? */
+uint64_t stats_n_data_cells_packaged = 0;
+/** How many bytes of data have we put in relay_data cells have we built,
+ * ever? This would be RELAY_PAYLOAD_SIZE*stats_n_data_cells_packaged if
+ * every relay cell we ever sent were completely full of data. */
+uint64_t stats_n_data_bytes_packaged = 0;
+/** How many relay_data cells have we received, ever? */
+uint64_t stats_n_data_cells_received = 0;
+/** How many bytes of data have we received relay_data cells, ever? This would
+ * be RELAY_PAYLOAD_SIZE*stats_n_data_cells_packaged if every relay cell we
+ * ever received were completely full of data. */
+uint64_t stats_n_data_bytes_received = 0;
+
+/** If <b>conn</b> has an entire relay payload of bytes on its inbuf (or
+ * <b>package_partial</b> is true), and the appropriate package windows aren't
+ * empty, grab a cell and send it down the circuit.
+ *
+ * If *<b>max_cells</b> is given, package no more than max_cells. Decrement
+ * *<b>max_cells</b> by the number of cells packaged.
+ *
+ * Return -1 (and send a RELAY_COMMAND_END cell if necessary) if conn should
+ * be marked for close, else return 0.
+ */
+int
+connection_edge_package_raw_inbuf(edge_connection_t *conn, int package_partial,
+ int *max_cells)
+{
+ size_t bytes_to_process, length;
+ char payload[CELL_PAYLOAD_SIZE];
+ circuit_t *circ;
+ const unsigned domain = conn->base_.type == CONN_TYPE_AP ? LD_APP : LD_EXIT;
+ int sending_from_optimistic = 0;
+ entry_connection_t *entry_conn =
+ conn->base_.type == CONN_TYPE_AP ? EDGE_TO_ENTRY_CONN(conn) : NULL;
+ const int sending_optimistically =
+ entry_conn &&
+ conn->base_.type == CONN_TYPE_AP &&
+ conn->base_.state != AP_CONN_STATE_OPEN;
+ crypt_path_t *cpath_layer = conn->cpath_layer;
+
+ tor_assert(conn);
+
+ if (conn->base_.marked_for_close) {
+ log_warn(LD_BUG,
+ "called on conn that's already marked for close at %s:%d.",
+ conn->base_.marked_for_close_file, conn->base_.marked_for_close);
+ return 0;
+ }
+
+ if (max_cells && *max_cells <= 0)
+ return 0;
+
+ repeat_connection_edge_package_raw_inbuf:
+
+ circ = circuit_get_by_edge_conn(conn);
+ if (!circ) {
+ log_info(domain,"conn has no circuit! Closing.");
+ conn->end_reason = END_STREAM_REASON_CANT_ATTACH;
+ return -1;
+ }
+
+ if (circuit_consider_stop_edge_reading(circ, cpath_layer))
+ return 0;
+
+ if (conn->package_window <= 0) {
+ log_info(domain,"called with package_window %d. Skipping.",
+ conn->package_window);
+ connection_stop_reading(TO_CONN(conn));
+ return 0;
+ }
+
+ sending_from_optimistic = entry_conn &&
+ entry_conn->sending_optimistic_data != NULL;
+
+ if (PREDICT_UNLIKELY(sending_from_optimistic)) {
+ bytes_to_process = buf_datalen(entry_conn->sending_optimistic_data);
+ if (PREDICT_UNLIKELY(!bytes_to_process)) {
+ log_warn(LD_BUG, "sending_optimistic_data was non-NULL but empty");
+ bytes_to_process = connection_get_inbuf_len(TO_CONN(conn));
+ sending_from_optimistic = 0;
+ }
+ } else {
+ bytes_to_process = connection_get_inbuf_len(TO_CONN(conn));
+ }
+
+ if (!bytes_to_process)
+ return 0;
+
+ if (!package_partial && bytes_to_process < RELAY_PAYLOAD_SIZE)
+ return 0;
+
+ if (bytes_to_process > RELAY_PAYLOAD_SIZE) {
+ length = RELAY_PAYLOAD_SIZE;
+ } else {
+ length = bytes_to_process;
+ }
+ stats_n_data_bytes_packaged += length;
+ stats_n_data_cells_packaged += 1;
+
+ if (PREDICT_UNLIKELY(sending_from_optimistic)) {
+ /* XXXX We could be more efficient here by sometimes packing
+ * previously-sent optimistic data in the same cell with data
+ * from the inbuf. */
+ buf_get_bytes(entry_conn->sending_optimistic_data, payload, length);
+ if (!buf_datalen(entry_conn->sending_optimistic_data)) {
+ buf_free(entry_conn->sending_optimistic_data);
+ entry_conn->sending_optimistic_data = NULL;
+ }
+ } else {
+ connection_buf_get_bytes(payload, length, TO_CONN(conn));
+ }
+
+ log_debug(domain,TOR_SOCKET_T_FORMAT": Packaging %d bytes (%d waiting).",
+ conn->base_.s,
+ (int)length, (int)connection_get_inbuf_len(TO_CONN(conn)));
+
+ if (sending_optimistically && !sending_from_optimistic) {
+ /* This is new optimistic data; remember it in case we need to detach and
+ retry */
+ if (!entry_conn->pending_optimistic_data)
+ entry_conn->pending_optimistic_data = buf_new();
+ buf_add(entry_conn->pending_optimistic_data, payload, length);
+ }
+
+ if (connection_edge_send_command(conn, RELAY_COMMAND_DATA,
+ payload, length) < 0 ) {
+ /* circuit got marked for close, don't continue, don't need to mark conn */
+ return 0;
+ }
+
+ if (!cpath_layer) { /* non-rendezvous exit */
+ tor_assert(circ->package_window > 0);
+ circ->package_window--;
+ } else { /* we're an AP, or an exit on a rendezvous circ */
+ tor_assert(cpath_layer->package_window > 0);
+ cpath_layer->package_window--;
+ }
+
+ if (--conn->package_window <= 0) { /* is it 0 after decrement? */
+ connection_stop_reading(TO_CONN(conn));
+ log_debug(domain,"conn->package_window reached 0.");
+ circuit_consider_stop_edge_reading(circ, cpath_layer);
+ return 0; /* don't process the inbuf any more */
+ }
+ log_debug(domain,"conn->package_window is now %d",conn->package_window);
+
+ if (max_cells) {
+ *max_cells -= 1;
+ if (*max_cells <= 0)
+ return 0;
+ }
+
+ /* handle more if there's more, or return 0 if there isn't */
+ goto repeat_connection_edge_package_raw_inbuf;
+}
+
+/** Called when we've just received a relay data cell, when
+ * we've just finished flushing all bytes to stream <b>conn</b>,
+ * or when we've flushed *some* bytes to the stream <b>conn</b>.
+ *
+ * If conn->outbuf is not too full, and our deliver window is
+ * low, send back a suitable number of stream-level sendme cells.
+ */
+void
+connection_edge_consider_sending_sendme(edge_connection_t *conn)
+{
+ circuit_t *circ;
+
+ if (connection_outbuf_too_full(TO_CONN(conn)))
+ return;
+
+ circ = circuit_get_by_edge_conn(conn);
+ if (!circ) {
+ /* this can legitimately happen if the destroy has already
+ * arrived and torn down the circuit */
+ log_info(LD_APP,"No circuit associated with conn. Skipping.");
+ return;
+ }
+
+ while (conn->deliver_window <= STREAMWINDOW_START - STREAMWINDOW_INCREMENT) {
+ log_debug(conn->base_.type == CONN_TYPE_AP ?LD_APP:LD_EXIT,
+ "Outbuf %d, Queuing stream sendme.",
+ (int)conn->base_.outbuf_flushlen);
+ conn->deliver_window += STREAMWINDOW_INCREMENT;
+ if (connection_edge_send_command(conn, RELAY_COMMAND_SENDME,
+ NULL, 0) < 0) {
+ log_warn(LD_APP,"connection_edge_send_command failed. Skipping.");
+ return; /* the circuit's closed, don't continue */
+ }
+ }
+}
+
+/** The circuit <b>circ</b> has received a circuit-level sendme
+ * (on hop <b>layer_hint</b>, if we're the OP). Go through all the
+ * attached streams and let them resume reading and packaging, if
+ * their stream windows allow it.
+ */
+static void
+circuit_resume_edge_reading(circuit_t *circ, crypt_path_t *layer_hint)
+{
+ if (circuit_queue_streams_are_blocked(circ)) {
+ log_debug(layer_hint?LD_APP:LD_EXIT,"Too big queue, no resuming");
+ return;
+ }
+ log_debug(layer_hint?LD_APP:LD_EXIT,"resuming");
+
+ if (CIRCUIT_IS_ORIGIN(circ))
+ circuit_resume_edge_reading_helper(TO_ORIGIN_CIRCUIT(circ)->p_streams,
+ circ, layer_hint);
+ else
+ circuit_resume_edge_reading_helper(TO_OR_CIRCUIT(circ)->n_streams,
+ circ, layer_hint);
+}
+
+void
+stream_choice_seed_weak_rng(void)
+{
+ crypto_seed_weak_rng(&stream_choice_rng);
+}
+
+/** A helper function for circuit_resume_edge_reading() above.
+ * The arguments are the same, except that <b>conn</b> is the head
+ * of a linked list of edge streams that should each be considered.
+ */
+static int
+circuit_resume_edge_reading_helper(edge_connection_t *first_conn,
+ circuit_t *circ,
+ crypt_path_t *layer_hint)
+{
+ edge_connection_t *conn;
+ int n_packaging_streams, n_streams_left;
+ int packaged_this_round;
+ int cells_on_queue;
+ int cells_per_conn;
+ edge_connection_t *chosen_stream = NULL;
+ int max_to_package;
+
+ if (first_conn == NULL) {
+ /* Don't bother to try to do the rest of this if there are no connections
+ * to resume. */
+ return 0;
+ }
+
+ /* How many cells do we have space for? It will be the minimum of
+ * the number needed to exhaust the package window, and the minimum
+ * needed to fill the cell queue. */
+ max_to_package = circ->package_window;
+ if (CIRCUIT_IS_ORIGIN(circ)) {
+ cells_on_queue = circ->n_chan_cells.n;
+ } else {
+ or_circuit_t *or_circ = TO_OR_CIRCUIT(circ);
+ cells_on_queue = or_circ->p_chan_cells.n;
+ }
+ if (CELL_QUEUE_HIGHWATER_SIZE - cells_on_queue < max_to_package)
+ max_to_package = CELL_QUEUE_HIGHWATER_SIZE - cells_on_queue;
+
+ /* Once we used to start listening on the streams in the order they
+ * appeared in the linked list. That leads to starvation on the
+ * streams that appeared later on the list, since the first streams
+ * would always get to read first. Instead, we just pick a random
+ * stream on the list, and enable reading for streams starting at that
+ * point (and wrapping around as if the list were circular). It would
+ * probably be better to actually remember which streams we've
+ * serviced in the past, but this is simple and effective. */
+
+ /* Select a stream uniformly at random from the linked list. We
+ * don't need cryptographic randomness here. */
+ {
+ int num_streams = 0;
+ for (conn = first_conn; conn; conn = conn->next_stream) {
+ num_streams++;
+ if (tor_weak_random_one_in_n(&stream_choice_rng, num_streams)) {
+ chosen_stream = conn;
+ }
+ /* Invariant: chosen_stream has been chosen uniformly at random from
+ * among the first num_streams streams on first_conn.
+ *
+ * (Note that we iterate over every stream on the circuit, so that after
+ * we've considered the first stream, we've chosen it with P=1; and
+ * after we consider the second stream, we've switched to it with P=1/2
+ * and stayed with the first stream with P=1/2; and after we've
+ * considered the third stream, we've switched to it with P=1/3 and
+ * remained with one of the first two streams with P=(2/3), giving each
+ * one P=(1/2)(2/3) )=(1/3).) */
+ }
+ }
+
+ /* Count how many non-marked streams there are that have anything on
+ * their inbuf, and enable reading on all of the connections. */
+ n_packaging_streams = 0;
+ /* Activate reading starting from the chosen stream */
+ for (conn=chosen_stream; conn; conn = conn->next_stream) {
+ /* Start reading for the streams starting from here */
+ if (conn->base_.marked_for_close || conn->package_window <= 0)
+ continue;
+ if (!layer_hint || conn->cpath_layer == layer_hint) {
+ connection_start_reading(TO_CONN(conn));
+
+ if (connection_get_inbuf_len(TO_CONN(conn)) > 0)
+ ++n_packaging_streams;
+ }
+ }
+ /* Go back and do the ones we skipped, circular-style */
+ for (conn = first_conn; conn != chosen_stream; conn = conn->next_stream) {
+ if (conn->base_.marked_for_close || conn->package_window <= 0)
+ continue;
+ if (!layer_hint || conn->cpath_layer == layer_hint) {
+ connection_start_reading(TO_CONN(conn));
+
+ if (connection_get_inbuf_len(TO_CONN(conn)) > 0)
+ ++n_packaging_streams;
+ }
+ }
+
+ if (n_packaging_streams == 0) /* avoid divide-by-zero */
+ return 0;
+
+ again:
+
+ cells_per_conn = CEIL_DIV(max_to_package, n_packaging_streams);
+
+ packaged_this_round = 0;
+ n_streams_left = 0;
+
+ /* Iterate over all connections. Package up to cells_per_conn cells on
+ * each. Update packaged_this_round with the total number of cells
+ * packaged, and n_streams_left with the number that still have data to
+ * package.
+ */
+ for (conn=first_conn; conn; conn=conn->next_stream) {
+ if (conn->base_.marked_for_close || conn->package_window <= 0)
+ continue;
+ if (!layer_hint || conn->cpath_layer == layer_hint) {
+ int n = cells_per_conn, r;
+ /* handle whatever might still be on the inbuf */
+ r = connection_edge_package_raw_inbuf(conn, 1, &n);
+
+ /* Note how many we packaged */
+ packaged_this_round += (cells_per_conn-n);
+
+ if (r<0) {
+ /* Problem while packaging. (We already sent an end cell if
+ * possible) */
+ connection_mark_for_close(TO_CONN(conn));
+ continue;
+ }
+
+ /* If there's still data to read, we'll be coming back to this stream. */
+ if (connection_get_inbuf_len(TO_CONN(conn)))
+ ++n_streams_left;
+
+ /* If the circuit won't accept any more data, return without looking
+ * at any more of the streams. Any connections that should be stopped
+ * have already been stopped by connection_edge_package_raw_inbuf. */
+ if (circuit_consider_stop_edge_reading(circ, layer_hint))
+ return -1;
+ /* XXXX should we also stop immediately if we fill up the cell queue?
+ * Probably. */
+ }
+ }
+
+ /* If we made progress, and we are willing to package more, and there are
+ * any streams left that want to package stuff... try again!
+ */
+ if (packaged_this_round && packaged_this_round < max_to_package &&
+ n_streams_left) {
+ max_to_package -= packaged_this_round;
+ n_packaging_streams = n_streams_left;
+ goto again;
+ }
+
+ return 0;
+}
+
+/** Check if the package window for <b>circ</b> is empty (at
+ * hop <b>layer_hint</b> if it's defined).
+ *
+ * If yes, tell edge streams to stop reading and return 1.
+ * Else return 0.
+ */
+static int
+circuit_consider_stop_edge_reading(circuit_t *circ, crypt_path_t *layer_hint)
+{
+ edge_connection_t *conn = NULL;
+ unsigned domain = layer_hint ? LD_APP : LD_EXIT;
+
+ if (!layer_hint) {
+ or_circuit_t *or_circ = TO_OR_CIRCUIT(circ);
+ log_debug(domain,"considering circ->package_window %d",
+ circ->package_window);
+ if (circ->package_window <= 0) {
+ log_debug(domain,"yes, not-at-origin. stopped.");
+ for (conn = or_circ->n_streams; conn; conn=conn->next_stream)
+ connection_stop_reading(TO_CONN(conn));
+ return 1;
+ }
+ return 0;
+ }
+ /* else, layer hint is defined, use it */
+ log_debug(domain,"considering layer_hint->package_window %d",
+ layer_hint->package_window);
+ if (layer_hint->package_window <= 0) {
+ log_debug(domain,"yes, at-origin. stopped.");
+ for (conn = TO_ORIGIN_CIRCUIT(circ)->p_streams; conn;
+ conn=conn->next_stream) {
+ if (conn->cpath_layer == layer_hint)
+ connection_stop_reading(TO_CONN(conn));
+ }
+ return 1;
+ }
+ return 0;
+}
+
+/** Check if the deliver_window for circuit <b>circ</b> (at hop
+ * <b>layer_hint</b> if it's defined) is low enough that we should
+ * send a circuit-level sendme back down the circuit. If so, send
+ * enough sendmes that the window would be overfull if we sent any
+ * more.
+ */
+static void
+circuit_consider_sending_sendme(circuit_t *circ, crypt_path_t *layer_hint)
+{
+// log_fn(LOG_INFO,"Considering: layer_hint is %s",
+// layer_hint ? "defined" : "null");
+ while ((layer_hint ? layer_hint->deliver_window : circ->deliver_window) <=
+ CIRCWINDOW_START - CIRCWINDOW_INCREMENT) {
+ log_debug(LD_CIRC,"Queuing circuit sendme.");
+ if (layer_hint)
+ layer_hint->deliver_window += CIRCWINDOW_INCREMENT;
+ else
+ circ->deliver_window += CIRCWINDOW_INCREMENT;
+ if (relay_send_command_from_edge(0, circ, RELAY_COMMAND_SENDME,
+ NULL, 0, layer_hint) < 0) {
+ log_warn(LD_CIRC,
+ "relay_send_command_from_edge failed. Circuit's closed.");
+ return; /* the circuit's closed, don't continue */
+ }
+ }
+}
+
+/** The total number of cells we have allocated. */
+static size_t total_cells_allocated = 0;
+
+/** Release storage held by <b>cell</b>. */
+static inline void
+packed_cell_free_unchecked(packed_cell_t *cell)
+{
+ --total_cells_allocated;
+ tor_free(cell);
+}
+
+/** Allocate and return a new packed_cell_t. */
+STATIC packed_cell_t *
+packed_cell_new(void)
+{
+ ++total_cells_allocated;
+ return tor_malloc_zero(sizeof(packed_cell_t));
+}
+
+/** Return a packed cell used outside by channel_t lower layer */
+void
+packed_cell_free_(packed_cell_t *cell)
+{
+ if (!cell)
+ return;
+ packed_cell_free_unchecked(cell);
+}
+
+/** Log current statistics for cell pool allocation at log level
+ * <b>severity</b>. */
+void
+dump_cell_pool_usage(int severity)
+{
+ int n_circs = 0;
+ int n_cells = 0;
+ SMARTLIST_FOREACH_BEGIN(circuit_get_global_list(), circuit_t *, c) {
+ n_cells += c->n_chan_cells.n;
+ if (!CIRCUIT_IS_ORIGIN(c))
+ n_cells += TO_OR_CIRCUIT(c)->p_chan_cells.n;
+ ++n_circs;
+ }
+ SMARTLIST_FOREACH_END(c);
+ tor_log(severity, LD_MM,
+ "%d cells allocated on %d circuits. %d cells leaked.",
+ n_cells, n_circs, (int)total_cells_allocated - n_cells);
+}
+
+/** Allocate a new copy of packed <b>cell</b>. */
+static inline packed_cell_t *
+packed_cell_copy(const cell_t *cell, int wide_circ_ids)
+{
+ packed_cell_t *c = packed_cell_new();
+ cell_pack(c, cell, wide_circ_ids);
+ return c;
+}
+
+/** Append <b>cell</b> to the end of <b>queue</b>. */
+void
+cell_queue_append(cell_queue_t *queue, packed_cell_t *cell)
+{
+ TOR_SIMPLEQ_INSERT_TAIL(&queue->head, cell, next);
+ ++queue->n;
+}
+
+/** Append a newly allocated copy of <b>cell</b> to the end of the
+ * <b>exitward</b> (or app-ward) <b>queue</b> of <b>circ</b>. If
+ * <b>use_stats</b> is true, record statistics about the cell.
+ */
+void
+cell_queue_append_packed_copy(circuit_t *circ, cell_queue_t *queue,
+ int exitward, const cell_t *cell,
+ int wide_circ_ids, int use_stats)
+{
+ packed_cell_t *copy = packed_cell_copy(cell, wide_circ_ids);
+ (void)circ;
+ (void)exitward;
+ (void)use_stats;
+
+ copy->inserted_timestamp = monotime_coarse_get_stamp();
+
+ cell_queue_append(queue, copy);
+}
+
+/** Initialize <b>queue</b> as an empty cell queue. */
+void
+cell_queue_init(cell_queue_t *queue)
+{
+ memset(queue, 0, sizeof(cell_queue_t));
+ TOR_SIMPLEQ_INIT(&queue->head);
+}
+
+/** Remove and free every cell in <b>queue</b>. */
+void
+cell_queue_clear(cell_queue_t *queue)
+{
+ packed_cell_t *cell;
+ while ((cell = TOR_SIMPLEQ_FIRST(&queue->head))) {
+ TOR_SIMPLEQ_REMOVE_HEAD(&queue->head, next);
+ packed_cell_free_unchecked(cell);
+ }
+ TOR_SIMPLEQ_INIT(&queue->head);
+ queue->n = 0;
+}
+
+/** Extract and return the cell at the head of <b>queue</b>; return NULL if
+ * <b>queue</b> is empty. */
+STATIC packed_cell_t *
+cell_queue_pop(cell_queue_t *queue)
+{
+ packed_cell_t *cell = TOR_SIMPLEQ_FIRST(&queue->head);
+ if (!cell)
+ return NULL;
+ TOR_SIMPLEQ_REMOVE_HEAD(&queue->head, next);
+ --queue->n;
+ return cell;
+}
+
+/** Initialize <b>queue</b> as an empty cell queue. */
+void
+destroy_cell_queue_init(destroy_cell_queue_t *queue)
+{
+ memset(queue, 0, sizeof(destroy_cell_queue_t));
+ TOR_SIMPLEQ_INIT(&queue->head);
+}
+
+/** Remove and free every cell in <b>queue</b>. */
+void
+destroy_cell_queue_clear(destroy_cell_queue_t *queue)
+{
+ destroy_cell_t *cell;
+ while ((cell = TOR_SIMPLEQ_FIRST(&queue->head))) {
+ TOR_SIMPLEQ_REMOVE_HEAD(&queue->head, next);
+ tor_free(cell);
+ }
+ TOR_SIMPLEQ_INIT(&queue->head);
+ queue->n = 0;
+}
+
+/** Extract and return the cell at the head of <b>queue</b>; return NULL if
+ * <b>queue</b> is empty. */
+STATIC destroy_cell_t *
+destroy_cell_queue_pop(destroy_cell_queue_t *queue)
+{
+ destroy_cell_t *cell = TOR_SIMPLEQ_FIRST(&queue->head);
+ if (!cell)
+ return NULL;
+ TOR_SIMPLEQ_REMOVE_HEAD(&queue->head, next);
+ --queue->n;
+ return cell;
+}
+
+/** Append a destroy cell for <b>circid</b> to <b>queue</b>. */
+void
+destroy_cell_queue_append(destroy_cell_queue_t *queue,
+ circid_t circid,
+ uint8_t reason)
+{
+ destroy_cell_t *cell = tor_malloc_zero(sizeof(destroy_cell_t));
+ cell->circid = circid;
+ cell->reason = reason;
+ /* Not yet used, but will be required for OOM handling. */
+ cell->inserted_timestamp = monotime_coarse_get_stamp();
+
+ TOR_SIMPLEQ_INSERT_TAIL(&queue->head, cell, next);
+ ++queue->n;
+}
+
+/** Convert a destroy_cell_t to a newly allocated cell_t. Frees its input. */
+static packed_cell_t *
+destroy_cell_to_packed_cell(destroy_cell_t *inp, int wide_circ_ids)
+{
+ packed_cell_t *packed = packed_cell_new();
+ cell_t cell;
+ memset(&cell, 0, sizeof(cell));
+ cell.circ_id = inp->circid;
+ cell.command = CELL_DESTROY;
+ cell.payload[0] = inp->reason;
+ cell_pack(packed, &cell, wide_circ_ids);
+
+ tor_free(inp);
+ return packed;
+}
+
+/** Return the total number of bytes used for each packed_cell in a queue.
+ * Approximate. */
+size_t
+packed_cell_mem_cost(void)
+{
+ return sizeof(packed_cell_t);
+}
+
+/* DOCDOC */
+size_t
+cell_queues_get_total_allocation(void)
+{
+ return total_cells_allocated * packed_cell_mem_cost();
+}
+
+/** How long after we've been low on memory should we try to conserve it? */
+#define MEMORY_PRESSURE_INTERVAL (30*60)
+
+/** The time at which we were last low on memory. */
+static time_t last_time_under_memory_pressure = 0;
+
+/** Check whether we've got too much space used for cells. If so,
+ * call the OOM handler and return 1. Otherwise, return 0. */
+STATIC int
+cell_queues_check_size(void)
+{
+ time_t now = time(NULL);
+ size_t alloc = cell_queues_get_total_allocation();
+ alloc += buf_get_total_allocation();
+ alloc += tor_compress_get_total_allocation();
+ const size_t rend_cache_total = rend_cache_get_total_allocation();
+ alloc += rend_cache_total;
+ const size_t geoip_client_cache_total =
+ geoip_client_cache_total_allocation();
+ alloc += geoip_client_cache_total;
+ const size_t dns_cache_total = dns_cache_total_allocation();
+ alloc += dns_cache_total;
+ if (alloc >= get_options()->MaxMemInQueues_low_threshold) {
+ last_time_under_memory_pressure = approx_time();
+ if (alloc >= get_options()->MaxMemInQueues) {
+ /* If we're spending over 20% of the memory limit on hidden service
+ * descriptors, free them until we're down to 10%. Do the same for geoip
+ * client cache. */
+ if (rend_cache_total > get_options()->MaxMemInQueues / 5) {
+ const size_t bytes_to_remove =
+ rend_cache_total - (size_t)(get_options()->MaxMemInQueues / 10);
+ alloc -= hs_cache_handle_oom(now, bytes_to_remove);
+ }
+ if (geoip_client_cache_total > get_options()->MaxMemInQueues / 5) {
+ const size_t bytes_to_remove =
+ geoip_client_cache_total -
+ (size_t)(get_options()->MaxMemInQueues / 10);
+ alloc -= geoip_client_cache_handle_oom(now, bytes_to_remove);
+ }
+ if (dns_cache_total > get_options()->MaxMemInQueues / 5) {
+ const size_t bytes_to_remove =
+ dns_cache_total - (size_t)(get_options()->MaxMemInQueues / 10);
+ alloc -= dns_cache_handle_oom(now, bytes_to_remove);
+ }
+ circuits_handle_oom(alloc);
+ return 1;
+ }
+ }
+ return 0;
+}
+
+/** Return true if we've been under memory pressure in the last
+ * MEMORY_PRESSURE_INTERVAL seconds. */
+int
+have_been_under_memory_pressure(void)
+{
+ return last_time_under_memory_pressure + MEMORY_PRESSURE_INTERVAL
+ < approx_time();
+}
+
+/**
+ * Update the number of cells available on the circuit's n_chan or p_chan's
+ * circuit mux.
+ */
+void
+update_circuit_on_cmux_(circuit_t *circ, cell_direction_t direction,
+ const char *file, int lineno)
+{
+ channel_t *chan = NULL;
+ or_circuit_t *or_circ = NULL;
+ circuitmux_t *cmux = NULL;
+
+ tor_assert(circ);
+
+ /* Okay, get the channel */
+ if (direction == CELL_DIRECTION_OUT) {
+ chan = circ->n_chan;
+ } else {
+ or_circ = TO_OR_CIRCUIT(circ);
+ chan = or_circ->p_chan;
+ }
+
+ tor_assert(chan);
+ tor_assert(chan->cmux);
+
+ /* Now get the cmux */
+ cmux = chan->cmux;
+
+ /* Cmux sanity check */
+ if (! circuitmux_is_circuit_attached(cmux, circ)) {
+ log_warn(LD_BUG, "called on non-attached circuit from %s:%d",
+ file, lineno);
+ return;
+ }
+ tor_assert(circuitmux_attached_circuit_direction(cmux, circ) == direction);
+
+ /* Update the number of cells we have for the circuit mux */
+ if (direction == CELL_DIRECTION_OUT) {
+ circuitmux_set_num_cells(cmux, circ, circ->n_chan_cells.n);
+ } else {
+ circuitmux_set_num_cells(cmux, circ, or_circ->p_chan_cells.n);
+ }
+}
+
+/** Remove all circuits from the cmux on <b>chan</b>.
+ *
+ * If <b>circuits_out</b> is non-NULL, add all detached circuits to
+ * <b>circuits_out</b>.
+ **/
+void
+channel_unlink_all_circuits(channel_t *chan, smartlist_t *circuits_out)
+{
+ tor_assert(chan);
+ tor_assert(chan->cmux);
+
+ circuitmux_detach_all_circuits(chan->cmux, circuits_out);
+ chan->num_n_circuits = 0;
+ chan->num_p_circuits = 0;
+}
+
+/** Block (if <b>block</b> is true) or unblock (if <b>block</b> is false)
+ * every edge connection that is using <b>circ</b> to write to <b>chan</b>,
+ * and start or stop reading as appropriate.
+ *
+ * If <b>stream_id</b> is nonzero, block only the edge connection whose
+ * stream_id matches it.
+ *
+ * Returns the number of streams whose status we changed.
+ */
+static int
+set_streams_blocked_on_circ(circuit_t *circ, channel_t *chan,
+ int block, streamid_t stream_id)
+{
+ edge_connection_t *edge = NULL;
+ int n = 0;
+ if (circ->n_chan == chan) {
+ circ->streams_blocked_on_n_chan = block;
+ if (CIRCUIT_IS_ORIGIN(circ))
+ edge = TO_ORIGIN_CIRCUIT(circ)->p_streams;
+ } else {
+ circ->streams_blocked_on_p_chan = block;
+ tor_assert(!CIRCUIT_IS_ORIGIN(circ));
+ edge = TO_OR_CIRCUIT(circ)->n_streams;
+ }
+
+ for (; edge; edge = edge->next_stream) {
+ connection_t *conn = TO_CONN(edge);
+ if (stream_id && edge->stream_id != stream_id)
+ continue;
+
+ if (edge->edge_blocked_on_circ != block) {
+ ++n;
+ edge->edge_blocked_on_circ = block;
+ }
+
+ if (!conn->read_event) {
+ /* This connection is a placeholder for something; probably a DNS
+ * request. It can't actually stop or start reading.*/
+ continue;
+ }
+
+ if (block) {
+ if (connection_is_reading(conn))
+ connection_stop_reading(conn);
+ } else {
+ /* Is this right? */
+ if (!connection_is_reading(conn))
+ connection_start_reading(conn);
+ }
+ }
+
+ return n;
+}
+
+/** Extract the command from a packed cell. */
+static uint8_t
+packed_cell_get_command(const packed_cell_t *cell, int wide_circ_ids)
+{
+ if (wide_circ_ids) {
+ return get_uint8(cell->body+4);
+ } else {
+ return get_uint8(cell->body+2);
+ }
+}
+
+/** Extract the circuit ID from a packed cell. */
+circid_t
+packed_cell_get_circid(const packed_cell_t *cell, int wide_circ_ids)
+{
+ if (wide_circ_ids) {
+ return ntohl(get_uint32(cell->body));
+ } else {
+ return ntohs(get_uint16(cell->body));
+ }
+}
+
+/** Pull as many cells as possible (but no more than <b>max</b>) from the
+ * queue of the first active circuit on <b>chan</b>, and write them to
+ * <b>chan</b>->outbuf. Return the number of cells written. Advance
+ * the active circuit pointer to the next active circuit in the ring. */
+MOCK_IMPL(int,
+channel_flush_from_first_active_circuit, (channel_t *chan, int max))
+{
+ circuitmux_t *cmux = NULL;
+ int n_flushed = 0;
+ cell_queue_t *queue;
+ destroy_cell_queue_t *destroy_queue=NULL;
+ circuit_t *circ;
+ or_circuit_t *or_circ;
+ int streams_blocked;
+ packed_cell_t *cell;
+
+ /* Get the cmux */
+ tor_assert(chan);
+ tor_assert(chan->cmux);
+ cmux = chan->cmux;
+
+ /* Main loop: pick a circuit, send a cell, update the cmux */
+ while (n_flushed < max) {
+ circ = circuitmux_get_first_active_circuit(cmux, &destroy_queue);
+ if (destroy_queue) {
+ destroy_cell_t *dcell;
+ /* this code is duplicated from some of the logic below. Ugly! XXXX */
+ /* If we are given a destroy_queue here, then it is required to be
+ * nonempty... */
+ tor_assert(destroy_queue->n > 0);
+ dcell = destroy_cell_queue_pop(destroy_queue);
+ /* ...and pop() will always yield a cell from a nonempty queue. */
+ tor_assert(dcell);
+ /* frees dcell */
+ cell = destroy_cell_to_packed_cell(dcell, chan->wide_circ_ids);
+ /* Send the DESTROY cell. It is very unlikely that this fails but just
+ * in case, get rid of the channel. */
+ if (channel_write_packed_cell(chan, cell) < 0) {
+ /* The cell has been freed. */
+ channel_mark_for_close(chan);
+ continue;
+ }
+ /* Update the cmux destroy counter */
+ circuitmux_notify_xmit_destroy(cmux);
+ cell = NULL;
+ ++n_flushed;
+ continue;
+ }
+ /* If it returns NULL, no cells left to send */
+ if (!circ) break;
+
+ if (circ->n_chan == chan) {
+ queue = &circ->n_chan_cells;
+ streams_blocked = circ->streams_blocked_on_n_chan;
+ } else {
+ or_circ = TO_OR_CIRCUIT(circ);
+ tor_assert(or_circ->p_chan == chan);
+ queue = &TO_OR_CIRCUIT(circ)->p_chan_cells;
+ streams_blocked = circ->streams_blocked_on_p_chan;
+ }
+
+ /* Circuitmux told us this was active, so it should have cells */
+ if (/*BUG(*/ queue->n == 0 /*)*/) {
+ log_warn(LD_BUG, "Found a supposedly active circuit with no cells "
+ "to send. Trying to recover.");
+ circuitmux_set_num_cells(cmux, circ, 0);
+ if (! circ->marked_for_close)
+ circuit_mark_for_close(circ, END_CIRC_REASON_INTERNAL);
+ continue;
+ }
+
+ tor_assert(queue->n > 0);
+
+ /*
+ * Get just one cell here; once we've sent it, that can change the circuit
+ * selection, so we have to loop around for another even if this circuit
+ * has more than one.
+ */
+ cell = cell_queue_pop(queue);
+
+ /* Calculate the exact time that this cell has spent in the queue. */
+ if (get_options()->CellStatistics ||
+ get_options()->TestingEnableCellStatsEvent) {
+ uint32_t timestamp_now = monotime_coarse_get_stamp();
+ uint32_t msec_waiting =
+ (uint32_t) monotime_coarse_stamp_units_to_approx_msec(
+ timestamp_now - cell->inserted_timestamp);
+
+ if (get_options()->CellStatistics && !CIRCUIT_IS_ORIGIN(circ)) {
+ or_circ = TO_OR_CIRCUIT(circ);
+ or_circ->total_cell_waiting_time += msec_waiting;
+ or_circ->processed_cells++;
+ }
+
+ if (get_options()->TestingEnableCellStatsEvent) {
+ uint8_t command = packed_cell_get_command(cell, chan->wide_circ_ids);
+
+ testing_cell_stats_entry_t *ent =
+ tor_malloc_zero(sizeof(testing_cell_stats_entry_t));
+ ent->command = command;
+ ent->waiting_time = msec_waiting / 10;
+ ent->removed = 1;
+ if (circ->n_chan == chan)
+ ent->exitward = 1;
+ if (!circ->testing_cell_stats)
+ circ->testing_cell_stats = smartlist_new();
+ smartlist_add(circ->testing_cell_stats, ent);
+ }
+ }
+
+ /* If we just flushed our queue and this circuit is used for a
+ * tunneled directory request, possibly advance its state. */
+ if (queue->n == 0 && chan->dirreq_id)
+ geoip_change_dirreq_state(chan->dirreq_id,
+ DIRREQ_TUNNELED,
+ DIRREQ_CIRC_QUEUE_FLUSHED);
+
+ /* Now send the cell. It is very unlikely that this fails but just in
+ * case, get rid of the channel. */
+ if (channel_write_packed_cell(chan, cell) < 0) {
+ /* The cell has been freed at this point. */
+ channel_mark_for_close(chan);
+ continue;
+ }
+ cell = NULL;
+
+ /*
+ * Don't packed_cell_free_unchecked(cell) here because the channel will
+ * do so when it gets out of the channel queue (probably already did, in
+ * which case that was an immediate double-free bug).
+ */
+
+ /* Update the counter */
+ ++n_flushed;
+
+ /*
+ * Now update the cmux; tell it we've just sent a cell, and how many
+ * we have left.
+ */
+ circuitmux_notify_xmit_cells(cmux, circ, 1);
+ circuitmux_set_num_cells(cmux, circ, queue->n);
+ if (queue->n == 0)
+ log_debug(LD_GENERAL, "Made a circuit inactive.");
+
+ /* Is the cell queue low enough to unblock all the streams that are waiting
+ * to write to this circuit? */
+ if (streams_blocked && queue->n <= CELL_QUEUE_LOWWATER_SIZE)
+ set_streams_blocked_on_circ(circ, chan, 0, 0); /* unblock streams */
+
+ /* If n_flushed < max still, loop around and pick another circuit */
+ }
+
+ /* Okay, we're done sending now */
+ return n_flushed;
+}
+
+/* Minimum value is the maximum circuit window size.
+ *
+ * SENDME cells makes it that we can control how many cells can be inflight on
+ * a circuit from end to end. This logic makes it that on any circuit cell
+ * queue, we have a maximum of cells possible.
+ *
+ * Because the Tor protocol allows for a client to exit at any hop in a
+ * circuit and a circuit can be of a maximum of 8 hops, so in theory the
+ * normal worst case will be the circuit window start value times the maximum
+ * number of hops (8). Having more cells then that means something is wrong.
+ *
+ * However, because padding cells aren't counted in the package window, we set
+ * the maximum size to a reasonably large size for which we expect that we'll
+ * never reach in theory. And if we ever do because of future changes, we'll
+ * be able to control it with a consensus parameter.
+ *
+ * XXX: Unfortunately, END cells aren't accounted for in the circuit window
+ * which means that for instance if a client opens 8001 streams, the 8001
+ * following END cells will queue up in the circuit which will get closed if
+ * the max limit is 8000. Which is sad because it is allowed by the Tor
+ * protocol. But, we need an upper bound on circuit queue in order to avoid
+ * DoS memory pressure so the default size is a middle ground between not
+ * having any limit and having a very restricted one. This is why we can also
+ * control it through a consensus parameter. */
+#define RELAY_CIRC_CELL_QUEUE_SIZE_MIN CIRCWINDOW_START_MAX
+/* We can't have a consensus parameter above this value. */
+#define RELAY_CIRC_CELL_QUEUE_SIZE_MAX INT32_MAX
+/* Default value is set to a large value so we can handle padding cells
+ * properly which aren't accounted for in the SENDME window. Default is 50000
+ * allowed cells in the queue resulting in ~25MB. */
+#define RELAY_CIRC_CELL_QUEUE_SIZE_DEFAULT \
+ (50 * RELAY_CIRC_CELL_QUEUE_SIZE_MIN)
+
+/* The maximum number of cell a circuit queue can contain. This is updated at
+ * every new consensus and controlled by a parameter. */
+static int32_t max_circuit_cell_queue_size =
+ RELAY_CIRC_CELL_QUEUE_SIZE_DEFAULT;
+
+/* Called when the consensus has changed. At this stage, the global consensus
+ * object has NOT been updated. It is called from
+ * notify_before_networkstatus_changes(). */
+void
+relay_consensus_has_changed(const networkstatus_t *ns)
+{
+ tor_assert(ns);
+
+ /* Update the circuit max cell queue size from the consensus. */
+ max_circuit_cell_queue_size =
+ networkstatus_get_param(ns, "circ_max_cell_queue_size",
+ RELAY_CIRC_CELL_QUEUE_SIZE_DEFAULT,
+ RELAY_CIRC_CELL_QUEUE_SIZE_MIN,
+ RELAY_CIRC_CELL_QUEUE_SIZE_MAX);
+}
+
+/** Add <b>cell</b> to the queue of <b>circ</b> writing to <b>chan</b>
+ * transmitting in <b>direction</b>.
+ *
+ * The given <b>cell</b> is copied onto the circuit queue so the caller must
+ * cleanup the memory.
+ *
+ * This function is part of the fast path. */
+void
+append_cell_to_circuit_queue(circuit_t *circ, channel_t *chan,
+ cell_t *cell, cell_direction_t direction,
+ streamid_t fromstream)
+{
+ or_circuit_t *orcirc = NULL;
+ cell_queue_t *queue;
+ int streams_blocked;
+ int exitward;
+ if (circ->marked_for_close)
+ return;
+
+ exitward = (direction == CELL_DIRECTION_OUT);
+ if (exitward) {
+ queue = &circ->n_chan_cells;
+ streams_blocked = circ->streams_blocked_on_n_chan;
+ } else {
+ orcirc = TO_OR_CIRCUIT(circ);
+ queue = &orcirc->p_chan_cells;
+ streams_blocked = circ->streams_blocked_on_p_chan;
+ }
+
+ if (PREDICT_UNLIKELY(queue->n >= max_circuit_cell_queue_size)) {
+ log_fn(LOG_PROTOCOL_WARN, LD_PROTOCOL,
+ "%s circuit has %d cells in its queue, maximum allowed is %d. "
+ "Closing circuit for safety reasons.",
+ (exitward) ? "Outbound" : "Inbound", queue->n,
+ max_circuit_cell_queue_size);
+ circuit_mark_for_close(circ, END_CIRC_REASON_RESOURCELIMIT);
+ stats_n_circ_max_cell_reached++;
+ return;
+ }
+
+ /* Very important that we copy to the circuit queue because all calls to
+ * this function use the stack for the cell memory. */
+ cell_queue_append_packed_copy(circ, queue, exitward, cell,
+ chan->wide_circ_ids, 1);
+
+ /* Check and run the OOM if needed. */
+ if (PREDICT_UNLIKELY(cell_queues_check_size())) {
+ /* We ran the OOM handler which might have closed this circuit. */
+ if (circ->marked_for_close)
+ return;
+ }
+
+ /* If we have too many cells on the circuit, we should stop reading from
+ * the edge streams for a while. */
+ if (!streams_blocked && queue->n >= CELL_QUEUE_HIGHWATER_SIZE)
+ set_streams_blocked_on_circ(circ, chan, 1, 0); /* block streams */
+
+ if (streams_blocked && fromstream) {
+ /* This edge connection is apparently not blocked; block it. */
+ set_streams_blocked_on_circ(circ, chan, 1, fromstream);
+ }
+
+ update_circuit_on_cmux(circ, direction);
+ if (queue->n == 1) {
+ /* This was the first cell added to the queue. We just made this
+ * circuit active. */
+ log_debug(LD_GENERAL, "Made a circuit active.");
+ }
+
+ /* New way: mark this as having waiting cells for the scheduler */
+ scheduler_channel_has_waiting_cells(chan);
+}
+
+/** Append an encoded value of <b>addr</b> to <b>payload_out</b>, which must
+ * have at least 18 bytes of free space. The encoding is, as specified in
+ * tor-spec.txt:
+ * RESOLVED_TYPE_IPV4 or RESOLVED_TYPE_IPV6 [1 byte]
+ * LENGTH [1 byte]
+ * ADDRESS [length bytes]
+ * Return the number of bytes added, or -1 on error */
+int
+append_address_to_payload(uint8_t *payload_out, const tor_addr_t *addr)
+{
+ uint32_t a;
+ switch (tor_addr_family(addr)) {
+ case AF_INET:
+ payload_out[0] = RESOLVED_TYPE_IPV4;
+ payload_out[1] = 4;
+ a = tor_addr_to_ipv4n(addr);
+ memcpy(payload_out+2, &a, 4);
+ return 6;
+ case AF_INET6:
+ payload_out[0] = RESOLVED_TYPE_IPV6;
+ payload_out[1] = 16;
+ memcpy(payload_out+2, tor_addr_to_in6_addr8(addr), 16);
+ return 18;
+ case AF_UNSPEC:
+ default:
+ return -1;
+ }
+}
+
+/** Given <b>payload_len</b> bytes at <b>payload</b>, starting with an address
+ * encoded as by append_address_to_payload(), try to decode the address into
+ * *<b>addr_out</b>. Return the next byte in the payload after the address on
+ * success, or NULL on failure. */
+const uint8_t *
+decode_address_from_payload(tor_addr_t *addr_out, const uint8_t *payload,
+ int payload_len)
+{
+ if (payload_len < 2)
+ return NULL;
+ if (payload_len < 2+payload[1])
+ return NULL;
+
+ switch (payload[0]) {
+ case RESOLVED_TYPE_IPV4:
+ if (payload[1] != 4)
+ return NULL;
+ tor_addr_from_ipv4n(addr_out, get_uint32(payload+2));
+ break;
+ case RESOLVED_TYPE_IPV6:
+ if (payload[1] != 16)
+ return NULL;
+ tor_addr_from_ipv6_bytes(addr_out, (char*)(payload+2));
+ break;
+ default:
+ tor_addr_make_unspec(addr_out);
+ break;
+ }
+ return payload + 2 + payload[1];
+}
+
+/** Remove all the cells queued on <b>circ</b> for <b>chan</b>. */
+void
+circuit_clear_cell_queue(circuit_t *circ, channel_t *chan)
+{
+ cell_queue_t *queue;
+ cell_direction_t direction;
+
+ if (circ->n_chan == chan) {
+ queue = &circ->n_chan_cells;
+ direction = CELL_DIRECTION_OUT;
+ } else {
+ or_circuit_t *orcirc = TO_OR_CIRCUIT(circ);
+ tor_assert(orcirc->p_chan == chan);
+ queue = &orcirc->p_chan_cells;
+ direction = CELL_DIRECTION_IN;
+ }
+
+ /* Clear the queue */
+ cell_queue_clear(queue);
+
+ /* Update the cell counter in the cmux */
+ if (chan->cmux && circuitmux_is_circuit_attached(chan->cmux, circ))
+ update_circuit_on_cmux(circ, direction);
+}
+
+/** Return 1 if we shouldn't restart reading on this circuit, even if
+ * we get a SENDME. Else return 0.
+*/
+static int
+circuit_queue_streams_are_blocked(circuit_t *circ)
+{
+ if (CIRCUIT_IS_ORIGIN(circ)) {
+ return circ->streams_blocked_on_n_chan;
+ } else {
+ return circ->streams_blocked_on_p_chan;
+ }
+}
diff --cc src/core/or/scheduler_kist.c
index 449478df7,000000000..41c346ac7
mode 100644,000000..100644
--- a/src/core/or/scheduler_kist.c
+++ b/src/core/or/scheduler_kist.c
@@@ -1,842 -1,0 +1,840 @@@
+/* Copyright (c) 2017-2018, The Tor Project, Inc. */
+/* See LICENSE for licensing information */
+
+#define SCHEDULER_KIST_PRIVATE
+
+#include "core/or/or.h"
+#include "lib/container/buffers.h"
+#include "app/config/config.h"
+#include "core/mainloop/connection.h"
+#include "feature/nodelist/networkstatus.h"
+#define TOR_CHANNEL_INTERNAL_
+#include "core/or/channel.h"
+#include "core/or/channeltls.h"
+#define SCHEDULER_PRIVATE_
+#include "core/or/scheduler.h"
+#include "lib/math/fp.h"
+
+#include "core/or/or_connection_st.h"
+
- #define TLS_PER_CELL_OVERHEAD 29
-
+#ifdef HAVE_SYS_IOCTL_H
+#include <sys/ioctl.h>
+#endif
+
+#ifdef HAVE_KIST_SUPPORT
+/* Kernel interface needed for KIST. */
+#include <netinet/tcp.h>
+#include <linux/sockios.h>
+#endif /* HAVE_KIST_SUPPORT */
+
+/*****************************************************************************
+ * Data structures and supporting functions
+ *****************************************************************************/
+
+/* Socket_table hash table stuff. The socket_table keeps track of per-socket
+ * limit information imposed by kist and used by kist. */
+
+static uint32_t
+socket_table_ent_hash(const socket_table_ent_t *ent)
+{
+ return (uint32_t)ent->chan->global_identifier;
+}
+
+static unsigned
+socket_table_ent_eq(const socket_table_ent_t *a, const socket_table_ent_t *b)
+{
+ return a->chan == b->chan;
+}
+
+typedef HT_HEAD(socket_table_s, socket_table_ent_s) socket_table_t;
+
+static socket_table_t socket_table = HT_INITIALIZER();
+
+HT_PROTOTYPE(socket_table_s, socket_table_ent_s, node, socket_table_ent_hash,
+ socket_table_ent_eq)
+HT_GENERATE2(socket_table_s, socket_table_ent_s, node, socket_table_ent_hash,
+ socket_table_ent_eq, 0.6, tor_reallocarray, tor_free_)
+
+/* outbuf_table hash table stuff. The outbuf_table keeps track of which
+ * channels have data sitting in their outbuf so the kist scheduler can force
+ * a write from outbuf to kernel periodically during a run and at the end of a
+ * run. */
+
+typedef struct outbuf_table_ent_s {
+ HT_ENTRY(outbuf_table_ent_s) node;
+ channel_t *chan;
+} outbuf_table_ent_t;
+
+static uint32_t
+outbuf_table_ent_hash(const outbuf_table_ent_t *ent)
+{
+ return (uint32_t)ent->chan->global_identifier;
+}
+
+static unsigned
+outbuf_table_ent_eq(const outbuf_table_ent_t *a, const outbuf_table_ent_t *b)
+{
+ return a->chan->global_identifier == b->chan->global_identifier;
+}
+
+HT_PROTOTYPE(outbuf_table_s, outbuf_table_ent_s, node, outbuf_table_ent_hash,
+ outbuf_table_ent_eq)
+HT_GENERATE2(outbuf_table_s, outbuf_table_ent_s, node, outbuf_table_ent_hash,
+ outbuf_table_ent_eq, 0.6, tor_reallocarray, tor_free_)
+
+/*****************************************************************************
+ * Other internal data
+ *****************************************************************************/
+
+/* Store the last time the scheduler was run so we can decide when to next run
+ * the scheduler based on it. */
+static monotime_t scheduler_last_run;
+/* This is a factor for the extra_space calculation in kist per-socket limits.
+ * It is the number of extra congestion windows we want to write to the kernel.
+ */
+static double sock_buf_size_factor = 1.0;
+/* How often the scheduler runs. */
+STATIC int sched_run_interval = KIST_SCHED_RUN_INTERVAL_DEFAULT;
+
+#ifdef HAVE_KIST_SUPPORT
+/* Indicate if KIST lite mode is on or off. We can disable it at runtime.
+ * Important to have because of the KISTLite -> KIST possible transition. */
+static unsigned int kist_lite_mode = 0;
+/* Indicate if we don't have the kernel support. This can happen if the kernel
+ * changed and it doesn't recognized the values passed to the syscalls needed
+ * by KIST. In that case, fallback to the naive approach. */
+static unsigned int kist_no_kernel_support = 0;
+#else /* !(defined(HAVE_KIST_SUPPORT)) */
+static unsigned int kist_lite_mode = 1;
+#endif /* defined(HAVE_KIST_SUPPORT) */
+
+/*****************************************************************************
+ * Internally called function implementations
+ *****************************************************************************/
+
+/* Little helper function to get the length of a channel's output buffer */
+static inline size_t
+channel_outbuf_length(channel_t *chan)
+{
+ /* In theory, this can not happen because we can not scheduler a channel
+ * without a connection that has its outbuf initialized. Just in case, bug
+ * on this so we can understand a bit more why it happened. */
+ if (SCHED_BUG(BASE_CHAN_TO_TLS(chan)->conn == NULL, chan)) {
+ return 0;
+ }
+ return buf_datalen(TO_CONN(BASE_CHAN_TO_TLS(chan)->conn)->outbuf);
+}
+
+/* Little helper function for HT_FOREACH_FN. */
+static int
+each_channel_write_to_kernel(outbuf_table_ent_t *ent, void *data)
+{
+ (void) data; /* Make compiler happy. */
+ channel_write_to_kernel(ent->chan);
+ return 0; /* Returning non-zero removes the element from the table. */
+}
+
+/* Free the given outbuf table entry ent. */
+static int
+free_outbuf_info_by_ent(outbuf_table_ent_t *ent, void *data)
+{
+ (void) data; /* Make compiler happy. */
+ log_debug(LD_SCHED, "Freeing outbuf table entry from chan=%" PRIu64,
+ ent->chan->global_identifier);
+ tor_free(ent);
+ return 1; /* So HT_FOREACH_FN will remove the element */
+}
+
+/* Free the given socket table entry ent. */
+static int
+free_socket_info_by_ent(socket_table_ent_t *ent, void *data)
+{
+ (void) data; /* Make compiler happy. */
+ log_debug(LD_SCHED, "Freeing socket table entry from chan=%" PRIu64,
+ ent->chan->global_identifier);
+ tor_free(ent);
+ return 1; /* So HT_FOREACH_FN will remove the element */
+}
+
+/* Clean up socket_table. Probably because the KIST sched impl is going away */
+static void
+free_all_socket_info(void)
+{
+ HT_FOREACH_FN(socket_table_s, &socket_table, free_socket_info_by_ent, NULL);
+ HT_CLEAR(socket_table_s, &socket_table);
+}
+
+static socket_table_ent_t *
+socket_table_search(socket_table_t *table, const channel_t *chan)
+{
+ socket_table_ent_t search, *ent = NULL;
+ search.chan = chan;
+ ent = HT_FIND(socket_table_s, table, &search);
+ return ent;
+}
+
+/* Free a socket entry in table for the given chan. */
+static void
+free_socket_info_by_chan(socket_table_t *table, const channel_t *chan)
+{
+ socket_table_ent_t *ent = NULL;
+ ent = socket_table_search(table, chan);
+ if (!ent)
+ return;
+ log_debug(LD_SCHED, "scheduler free socket info for chan=%" PRIu64,
+ chan->global_identifier);
+ HT_REMOVE(socket_table_s, table, ent);
+ free_socket_info_by_ent(ent, NULL);
+}
+
+/* Perform system calls for the given socket in order to calculate kist's
+ * per-socket limit as documented in the function body. */
+MOCK_IMPL(void,
+update_socket_info_impl, (socket_table_ent_t *ent))
+{
+#ifdef HAVE_KIST_SUPPORT
+ int64_t tcp_space, extra_space;
+ const tor_socket_t sock =
+ TO_CONN(BASE_CHAN_TO_TLS((channel_t *) ent->chan)->conn)->s;
+ struct tcp_info tcp;
+ socklen_t tcp_info_len = sizeof(tcp);
+
+ if (kist_no_kernel_support || kist_lite_mode) {
+ goto fallback;
+ }
+
+ /* Gather information */
+ if (getsockopt(sock, SOL_TCP, TCP_INFO, (void *)&(tcp), &tcp_info_len) < 0) {
+ if (errno == EINVAL) {
+ /* Oops, this option is not provided by the kernel, we'll have to
+ * disable KIST entirely. This can happen if tor was built on a machine
+ * with the support previously or if the kernel was updated and lost the
+ * support. */
+ log_notice(LD_SCHED, "Looks like our kernel doesn't have the support "
+ "for KIST anymore. We will fallback to the naive "
+ "approach. Remove KIST from the Schedulers list "
+ "to disable.");
+ kist_no_kernel_support = 1;
+ }
+ goto fallback;
+ }
+ if (ioctl(sock, SIOCOUTQNSD, &(ent->notsent)) < 0) {
+ if (errno == EINVAL) {
+ log_notice(LD_SCHED, "Looks like our kernel doesn't have the support "
+ "for KIST anymore. We will fallback to the naive "
+ "approach. Remove KIST from the Schedulers list "
+ "to disable.");
+ /* Same reason as the above. */
+ kist_no_kernel_support = 1;
+ }
+ goto fallback;
+ }
+ ent->cwnd = tcp.tcpi_snd_cwnd;
+ ent->unacked = tcp.tcpi_unacked;
+ ent->mss = tcp.tcpi_snd_mss;
+
+ /* In order to reduce outbound kernel queuing delays and thus improve Tor's
+ * ability to prioritize circuits, KIST wants to set a socket write limit
+ * that is near the amount that the socket would be able to immediately send
+ * into the Internet.
+ *
+ * We first calculate how much the socket could send immediately (assuming
+ * completely full packets) according to the congestion window and the number
+ * of unacked packets.
+ *
+ * Then we add a little extra space in a controlled way. We do this so any
+ * when the kernel gets ACKs back for data currently sitting in the "TCP
+ * space", it will already have some more data to send immediately. It will
+ * not have to wait for the scheduler to run again. The amount of extra space
+ * is a factor of the current congestion window. With the suggested
+ * sock_buf_size_factor value of 1.0, we allow at most 2*cwnd bytes to sit in
+ * the kernel: 1 cwnd on the wire waiting for ACKs and 1 cwnd ready and
+ * waiting to be sent when those ACKs finally come.
+ *
+ * In the below diagram, we see some bytes in the TCP-space (denoted by '*')
+ * that have be sent onto the wire and are waiting for ACKs. We have a little
+ * more room in "TCP space" that we can fill with data that will be
+ * immediately sent. We also see the "extra space" KIST calculates. The sum
+ * of the empty "TCP space" and the "extra space" is the kist-imposed write
+ * limit for this socket.
+ *
+ * <----------------kernel-outbound-socket-queue----------------|
+ * <*********---------------------------------------------------|
+ * |----TCP-space-----|----extra-space-----|
+ * |------------------|
+ * ^ ((cwnd - unacked) * mss) bytes
+ * |--------------------|
+ * ^ ((cwnd * mss) * factor) bytes
+ */
+
+ /* These values from the kernel are uint32_t, they will always fit into a
+ * int64_t tcp_space variable but if the congestion window cwnd is smaller
+ * than the unacked packets, the remaining TCP space is set to 0. */
+ if (ent->cwnd >= ent->unacked) {
+ tcp_space = (ent->cwnd - ent->unacked) * (int64_t)(ent->mss);
+ } else {
+ tcp_space = 0;
+ }
+
+ /* The clamp_double_to_int64 makes sure the first part fits into an int64_t.
+ * In fact, if sock_buf_size_factor is still forced to be >= 0 in config.c,
+ * then it will be positive for sure. Then we subtract a uint32_t. Getting a
+ * negative value is OK, see after how it is being handled. */
+ extra_space =
+ clamp_double_to_int64(
+ (ent->cwnd * (int64_t)ent->mss) * sock_buf_size_factor) -
+ ent->notsent;
+ if ((tcp_space + extra_space) < 0) {
+ /* This means that the "notsent" queue is just too big so we shouldn't put
+ * more in the kernel for now. */
+ ent->limit = 0;
+ } else {
+ /* The positive sum of two int64_t will always fit into an uint64_t.
+ * And we know this will always be positive, since we checked above. */
+ ent->limit = (uint64_t)tcp_space + (uint64_t)extra_space;
+ }
+ return;
+
+#else /* !(defined(HAVE_KIST_SUPPORT)) */
+ goto fallback;
+#endif /* defined(HAVE_KIST_SUPPORT) */
+
+ fallback:
+ /* If all of a sudden we don't have kist support, we just zero out all the
+ * variables for this socket since we don't know what they should be. We
+ * also allow the socket to write as much as it can from the estimated
+ * number of cells the lower layer can accept, effectively returning it to
+ * Vanilla scheduler behavior. */
+ ent->cwnd = ent->unacked = ent->mss = ent->notsent = 0;
+ /* This function calls the specialized channel object (currently channeltls)
+ * and ask how many cells it can write on the outbuf which we then multiply
+ * by the size of the cells for this channel. The cast is because this
+ * function requires a non-const channel object, meh. */
+ ent->limit = channel_num_cells_writeable((channel_t *) ent->chan) *
+ (get_cell_network_size(ent->chan->wide_circ_ids) +
+ TLS_PER_CELL_OVERHEAD);
+}
+
+/* Given a socket that isn't in the table, add it.
+ * Given a socket that is in the table, re-init values that need init-ing
+ * every scheduling run
+ */
+static void
+init_socket_info(socket_table_t *table, const channel_t *chan)
+{
+ socket_table_ent_t *ent = NULL;
+ ent = socket_table_search(table, chan);
+ if (!ent) {
+ log_debug(LD_SCHED, "scheduler init socket info for chan=%" PRIu64,
+ chan->global_identifier);
+ ent = tor_malloc_zero(sizeof(*ent));
+ ent->chan = chan;
+ HT_INSERT(socket_table_s, table, ent);
+ }
+ ent->written = 0;
+}
+
+/* Add chan to the outbuf table if it isn't already in it. If it is, then don't
+ * do anything */
+static void
+outbuf_table_add(outbuf_table_t *table, channel_t *chan)
+{
+ outbuf_table_ent_t search, *ent;
+ search.chan = chan;
+ ent = HT_FIND(outbuf_table_s, table, &search);
+ if (!ent) {
+ log_debug(LD_SCHED, "scheduler init outbuf info for chan=%" PRIu64,
+ chan->global_identifier);
+ ent = tor_malloc_zero(sizeof(*ent));
+ ent->chan = chan;
+ HT_INSERT(outbuf_table_s, table, ent);
+ }
+}
+
+static void
+outbuf_table_remove(outbuf_table_t *table, channel_t *chan)
+{
+ outbuf_table_ent_t search, *ent;
+ search.chan = chan;
+ ent = HT_FIND(outbuf_table_s, table, &search);
+ if (ent) {
+ HT_REMOVE(outbuf_table_s, table, ent);
+ free_outbuf_info_by_ent(ent, NULL);
+ }
+}
+
+/* Set the scheduler running interval. */
+static void
+set_scheduler_run_interval(void)
+{
+ int old_sched_run_interval = sched_run_interval;
+ sched_run_interval = kist_scheduler_run_interval();
+ if (old_sched_run_interval != sched_run_interval) {
+ log_info(LD_SCHED, "Scheduler KIST changing its running interval "
+ "from %" PRId32 " to %" PRId32,
+ old_sched_run_interval, sched_run_interval);
+ }
+}
+
+/* Return true iff the channel hasn't hit its kist-imposed write limit yet */
+static int
+socket_can_write(socket_table_t *table, const channel_t *chan)
+{
+ socket_table_ent_t *ent = NULL;
+ ent = socket_table_search(table, chan);
+ if (SCHED_BUG(!ent, chan)) {
+ return 1; // Just return true, saying that kist wouldn't limit the socket
+ }
+
+ /* We previously calculated a write limit for this socket. In the below
+ * calculation, first determine how much room is left in bytes. Then divide
+ * that by the amount of space a cell takes. If there's room for at least 1
+ * cell, then KIST will allow the socket to write. */
+ int64_t kist_limit_space =
+ (int64_t) (ent->limit - ent->written) /
+ (CELL_MAX_NETWORK_SIZE + TLS_PER_CELL_OVERHEAD);
+ return kist_limit_space > 0;
+}
+
+/* Update the channel's socket kernel information. */
+static void
+update_socket_info(socket_table_t *table, const channel_t *chan)
+{
+ socket_table_ent_t *ent = NULL;
+ ent = socket_table_search(table, chan);
+ if (SCHED_BUG(!ent, chan)) {
+ return; // Whelp. Entry didn't exist for some reason so nothing to do.
+ }
+ update_socket_info_impl(ent);
+ log_debug(LD_SCHED, "chan=%" PRIu64 " updated socket info, limit: %" PRIu64
+ ", cwnd: %" PRIu32 ", unacked: %" PRIu32
+ ", notsent: %" PRIu32 ", mss: %" PRIu32,
+ ent->chan->global_identifier, ent->limit, ent->cwnd, ent->unacked,
+ ent->notsent, ent->mss);
+}
+
+/* Increment the channel's socket written value by the number of bytes. */
+static void
+update_socket_written(socket_table_t *table, channel_t *chan, size_t bytes)
+{
+ socket_table_ent_t *ent = NULL;
+ ent = socket_table_search(table, chan);
+ if (SCHED_BUG(!ent, chan)) {
+ return; // Whelp. Entry didn't exist so nothing to do.
+ }
+
+ log_debug(LD_SCHED, "chan=%" PRIu64 " wrote %lu bytes, old was %" PRIi64,
+ chan->global_identifier, (unsigned long) bytes, ent->written);
+
+ ent->written += bytes;
+}
+
+/*
+ * A naive KIST impl would write every single cell all the way to the kernel.
+ * That would take a lot of system calls. A less bad KIST impl would write a
+ * channel's outbuf to the kernel only when we are switching to a different
+ * channel. But if we have two channels with equal priority, we end up writing
+ * one cell for each and bouncing back and forth. This KIST impl avoids that
+ * by only writing a channel's outbuf to the kernel if it has 8 cells or more
+ * in it.
+ */
+MOCK_IMPL(int, channel_should_write_to_kernel,
+ (outbuf_table_t *table, channel_t *chan))
+{
+ outbuf_table_add(table, chan);
+ /* CELL_MAX_NETWORK_SIZE * 8 because we only want to write the outbuf to the
+ * kernel if there's 8 or more cells waiting */
+ return channel_outbuf_length(chan) > (CELL_MAX_NETWORK_SIZE * 8);
+}
+
+/* Little helper function to write a channel's outbuf all the way to the
+ * kernel */
+MOCK_IMPL(void, channel_write_to_kernel, (channel_t *chan))
+{
+ log_debug(LD_SCHED, "Writing %lu bytes to kernel for chan %" PRIu64,
+ (unsigned long)channel_outbuf_length(chan),
+ chan->global_identifier);
+ connection_handle_write(TO_CONN(BASE_CHAN_TO_TLS(chan)->conn), 0);
+}
+
+/* Return true iff the scheduler has work to perform. */
+static int
+have_work(void)
+{
+ smartlist_t *cp = get_channels_pending();
+ IF_BUG_ONCE(!cp) {
+ return 0; // channels_pending doesn't exist so... no work?
+ }
+ return smartlist_len(cp) > 0;
+}
+
+/* Function of the scheduler interface: free_all() */
+static void
+kist_free_all(void)
+{
+ free_all_socket_info();
+}
+
+/* Function of the scheduler interface: on_channel_free() */
+static void
+kist_on_channel_free_fn(const channel_t *chan)
+{
+ free_socket_info_by_chan(&socket_table, chan);
+}
+
+/* Function of the scheduler interface: on_new_consensus() */
+static void
+kist_scheduler_on_new_consensus(void)
+{
+ set_scheduler_run_interval();
+}
+
+/* Function of the scheduler interface: on_new_options() */
+static void
+kist_scheduler_on_new_options(void)
+{
+ sock_buf_size_factor = get_options()->KISTSockBufSizeFactor;
+
+ /* Calls kist_scheduler_run_interval which calls get_options(). */
+ set_scheduler_run_interval();
+}
+
+/* Function of the scheduler interface: init() */
+static void
+kist_scheduler_init(void)
+{
+ /* When initializing the scheduler, the last run could be 0 because it is
+ * declared static or a value in the past that was set when it was last
+ * used. In both cases, we want to initialize it to now so we don't risk
+ * using the value 0 which doesn't play well with our monotonic time
+ * interface.
+ *
+ * One side effect is that the first scheduler run will be at the next tick
+ * that is in now + 10 msec (KIST_SCHED_RUN_INTERVAL_DEFAULT) by default. */
+ monotime_get(&scheduler_last_run);
+
+ kist_scheduler_on_new_options();
+ IF_BUG_ONCE(sched_run_interval == 0) {
+ log_warn(LD_SCHED, "We are initing the KIST scheduler and noticed the "
+ "KISTSchedRunInterval is telling us to not use KIST. That's "
+ "weird! We'll continue using KIST, but at %" PRId32 "ms.",
+ KIST_SCHED_RUN_INTERVAL_DEFAULT);
+ sched_run_interval = KIST_SCHED_RUN_INTERVAL_DEFAULT;
+ }
+}
+
+/* Function of the scheduler interface: schedule() */
+static void
+kist_scheduler_schedule(void)
+{
+ struct monotime_t now;
+ struct timeval next_run;
+ int64_t diff;
+
+ if (!have_work()) {
+ return;
+ }
+ monotime_get(&now);
+
+ /* If time is really monotonic, we can never have now being smaller than the
+ * last scheduler run. The scheduler_last_run at first is set to 0.
+ * Unfortunately, not all platforms guarantee monotonic time so we log at
+ * info level but don't make it more noisy. */
+ diff = monotime_diff_msec(&scheduler_last_run, &now);
+ if (diff < 0) {
+ log_info(LD_SCHED, "Monotonic time between now and last run of scheduler "
+ "is negative: %" PRId64 ". Setting diff to 0.", diff);
+ diff = 0;
+ }
+ if (diff < sched_run_interval) {
+ next_run.tv_sec = 0;
+ /* Takes 1000 ms -> us. This will always be valid because diff can NOT be
+ * negative and can NOT be bigger than sched_run_interval so values can
+ * only go from 1000 usec (diff set to interval - 1) to 100000 usec (diff
+ * set to 0) for the maximum allowed run interval (100ms). */
+ next_run.tv_usec = (int) ((sched_run_interval - diff) * 1000);
+ /* Re-adding an event reschedules it. It does not duplicate it. */
+ scheduler_ev_add(&next_run);
+ } else {
+ scheduler_ev_active();
+ }
+}
+
+/* Function of the scheduler interface: run() */
+static void
+kist_scheduler_run(void)
+{
+ /* Define variables */
+ channel_t *chan = NULL; // current working channel
+ /* The last distinct chan served in a sched loop. */
+ channel_t *prev_chan = NULL;
+ int flush_result; // temporarily store results from flush calls
+ /* Channels to be re-adding to pending at the end */
+ smartlist_t *to_readd = NULL;
+ smartlist_t *cp = get_channels_pending();
+
+ outbuf_table_t outbuf_table = HT_INITIALIZER();
+
+ /* For each pending channel, collect new kernel information */
+ SMARTLIST_FOREACH_BEGIN(cp, const channel_t *, pchan) {
+ init_socket_info(&socket_table, pchan);
+ update_socket_info(&socket_table, pchan);
+ } SMARTLIST_FOREACH_END(pchan);
+
+ log_debug(LD_SCHED, "Running the scheduler. %d channels pending",
+ smartlist_len(cp));
+
+ /* The main scheduling loop. Loop until there are no more pending channels */
+ while (smartlist_len(cp) > 0) {
+ /* get best channel */
+ chan = smartlist_pqueue_pop(cp, scheduler_compare_channels,
+ offsetof(channel_t, sched_heap_idx));
+ if (SCHED_BUG(!chan, NULL)) {
+ /* Some-freaking-how a NULL got into the channels_pending. That should
+ * never happen, but it should be harmless to ignore it and keep looping.
+ */
+ continue;
+ }
+ outbuf_table_add(&outbuf_table, chan);
+
+ /* if we have switched to a new channel, consider writing the previous
+ * channel's outbuf to the kernel. */
+ if (!prev_chan) {
+ prev_chan = chan;
+ }
+ if (prev_chan != chan) {
+ if (channel_should_write_to_kernel(&outbuf_table, prev_chan)) {
+ channel_write_to_kernel(prev_chan);
+ outbuf_table_remove(&outbuf_table, prev_chan);
+ }
+ prev_chan = chan;
+ }
+
+ /* Only flush and write if the per-socket limit hasn't been hit */
+ if (socket_can_write(&socket_table, chan)) {
+ /* flush to channel queue/outbuf */
+ flush_result = (int)channel_flush_some_cells(chan, 1); // 1 for num cells
+ /* XXX: While flushing cells, it is possible that the connection write
+ * fails leading to the channel to be closed which triggers a release
+ * and free its entry in the socket table. And because of a engineering
+ * design issue, the error is not propagated back so we don't get an
+ * error at this point. So before we continue, make sure the channel is
+ * open and if not just ignore it. See #23751. */
+ if (!CHANNEL_IS_OPEN(chan)) {
+ /* Channel isn't open so we put it back in IDLE mode. It is either
+ * renegotiating its TLS session or about to be released. */
+ scheduler_set_channel_state(chan, SCHED_CHAN_IDLE);
+ continue;
+ }
+ /* flush_result has the # cells flushed */
+ if (flush_result > 0) {
+ update_socket_written(&socket_table, chan, flush_result *
+ (CELL_MAX_NETWORK_SIZE + TLS_PER_CELL_OVERHEAD));
+ } else {
+ /* XXX: This can happen because tor sometimes does flush in an
+ * opportunistic way cells from the circuit to the outbuf so the
+ * channel can end up here without having anything to flush nor needed
+ * to write to the kernel. Hopefully we'll fix that soon but for now
+ * we have to handle this case which happens kind of often. */
+ log_debug(LD_SCHED,
+ "We didn't flush anything on a chan that we think "
+ "can write and wants to write. The channel's state is '%s' "
+ "and in scheduler state '%s'. We're going to mark it as "
+ "waiting_for_cells (as that's most likely the issue) and "
+ "stop scheduling it this round.",
+ channel_state_to_string(chan->state),
+ get_scheduler_state_string(chan->scheduler_state));
+ scheduler_set_channel_state(chan, SCHED_CHAN_WAITING_FOR_CELLS);
+ continue;
+ }
+ }
+
+ /* Decide what to do with the channel now */
+
+ if (!channel_more_to_flush(chan) &&
+ !socket_can_write(&socket_table, chan)) {
+
+ /* Case 1: no more cells to send, and cannot write */
+
+ /*
+ * You might think we should put the channel in SCHED_CHAN_IDLE. And
+ * you're probably correct. While implementing KIST, we found that the
+ * scheduling system would sometimes lose track of channels when we did
+ * that. We suspect it has to do with the difference between "can't
+ * write because socket/outbuf is full" and KIST's "can't write because
+ * we've arbitrarily decided that that's enough for now." Sometimes
+ * channels run out of cells at the same time they hit their
+ * kist-imposed write limit and maybe the rest of Tor doesn't put the
+ * channel back in pending when it is supposed to.
+ *
+ * This should be investigated again. It is as simple as changing
+ * SCHED_CHAN_WAITING_FOR_CELLS to SCHED_CHAN_IDLE and seeing if Tor
+ * starts having serious throughput issues. Best done in shadow/chutney.
+ */
+ scheduler_set_channel_state(chan, SCHED_CHAN_WAITING_FOR_CELLS);
+ } else if (!channel_more_to_flush(chan)) {
+
+ /* Case 2: no more cells to send, but still open for writes */
+
+ scheduler_set_channel_state(chan, SCHED_CHAN_WAITING_FOR_CELLS);
+ } else if (!socket_can_write(&socket_table, chan)) {
+
+ /* Case 3: cells to send, but cannot write */
+
+ /*
+ * We want to write, but can't. If we left the channel in
+ * channels_pending, we would never exit the scheduling loop. We need to
+ * add it to a temporary list of channels to be added to channels_pending
+ * after the scheduling loop is over. They can hopefully be taken care of
+ * in the next scheduling round.
+ */
+ if (!to_readd) {
+ to_readd = smartlist_new();
+ }
+ smartlist_add(to_readd, chan);
+ } else {
+
+ /* Case 4: cells to send, and still open for writes */
+
+ scheduler_set_channel_state(chan, SCHED_CHAN_PENDING);
+ if (!SCHED_BUG(chan->sched_heap_idx != -1, chan)) {
+ smartlist_pqueue_add(cp, scheduler_compare_channels,
+ offsetof(channel_t, sched_heap_idx), chan);
+ }
+ }
+ } /* End of main scheduling loop */
+
+ /* Write the outbuf of any channels that still have data */
+ HT_FOREACH_FN(outbuf_table_s, &outbuf_table, each_channel_write_to_kernel,
+ NULL);
+ /* We are done with it. */
+ HT_FOREACH_FN(outbuf_table_s, &outbuf_table, free_outbuf_info_by_ent, NULL);
+ HT_CLEAR(outbuf_table_s, &outbuf_table);
+
+ log_debug(LD_SCHED, "len pending=%d, len to_readd=%d",
+ smartlist_len(cp),
+ (to_readd ? smartlist_len(to_readd) : -1));
+
+ /* Re-add any channels we need to */
+ if (to_readd) {
+ SMARTLIST_FOREACH_BEGIN(to_readd, channel_t *, readd_chan) {
+ scheduler_set_channel_state(readd_chan, SCHED_CHAN_PENDING);
+ if (!smartlist_contains(cp, readd_chan)) {
+ if (!SCHED_BUG(chan->sched_heap_idx != -1, chan)) {
+ /* XXXX Note that the check above is in theory redundant with
+ * the smartlist_contains check. But let's make sure we're
+ * not messing anything up, and leave them both for now. */
+ smartlist_pqueue_add(cp, scheduler_compare_channels,
+ offsetof(channel_t, sched_heap_idx), readd_chan);
+ }
+ }
+ } SMARTLIST_FOREACH_END(readd_chan);
+ smartlist_free(to_readd);
+ }
+
+ monotime_get(&scheduler_last_run);
+}
+
+/*****************************************************************************
+ * Externally called function implementations not called through scheduler_t
+ *****************************************************************************/
+
+/* Stores the kist scheduler function pointers. */
+static scheduler_t kist_scheduler = {
+ .type = SCHEDULER_KIST,
+ .free_all = kist_free_all,
+ .on_channel_free = kist_on_channel_free_fn,
+ .init = kist_scheduler_init,
+ .on_new_consensus = kist_scheduler_on_new_consensus,
+ .schedule = kist_scheduler_schedule,
+ .run = kist_scheduler_run,
+ .on_new_options = kist_scheduler_on_new_options,
+};
+
+/* Return the KIST scheduler object. If it didn't exists, return a newly
+ * allocated one but init() is not called. */
+scheduler_t *
+get_kist_scheduler(void)
+{
+ return &kist_scheduler;
+}
+
+/* Check the torrc (and maybe consensus) for the configured KIST scheduler run
+ * interval.
+ * - If torrc > 0, then return the positive torrc value (should use KIST, and
+ * should use the set value)
+ * - If torrc == 0, then look in the consensus for what the value should be.
+ * - If == 0, then return 0 (don't use KIST)
+ * - If > 0, then return the positive consensus value
+ * - If consensus doesn't say anything, return 10 milliseconds, default.
+ */
+int
+kist_scheduler_run_interval(void)
+{
+ int run_interval = get_options()->KISTSchedRunInterval;
+
+ if (run_interval != 0) {
+ log_debug(LD_SCHED, "Found KISTSchedRunInterval=%" PRId32 " in torrc. "
+ "Using that.", run_interval);
+ return run_interval;
+ }
+
+ log_debug(LD_SCHED, "KISTSchedRunInterval=0, turning to the consensus.");
+
+ /* Will either be the consensus value or the default. Note that 0 can be
+ * returned which means the consensus wants us to NOT use KIST. */
+ return networkstatus_get_param(NULL, "KISTSchedRunInterval",
+ KIST_SCHED_RUN_INTERVAL_DEFAULT,
+ KIST_SCHED_RUN_INTERVAL_MIN,
+ KIST_SCHED_RUN_INTERVAL_MAX);
+}
+
+/* Set KISTLite mode that is KIST without kernel support. */
+void
+scheduler_kist_set_lite_mode(void)
+{
+ kist_lite_mode = 1;
+ kist_scheduler.type = SCHEDULER_KIST_LITE;
+ log_info(LD_SCHED,
+ "Setting KIST scheduler without kernel support (KISTLite mode)");
+}
+
+/* Set KIST mode that is KIST with kernel support. */
+void
+scheduler_kist_set_full_mode(void)
+{
+ kist_lite_mode = 0;
+ kist_scheduler.type = SCHEDULER_KIST;
+ log_info(LD_SCHED,
+ "Setting KIST scheduler with kernel support (KIST mode)");
+}
+
+#ifdef HAVE_KIST_SUPPORT
+
+/* Return true iff the scheduler subsystem should use KIST. */
+int
+scheduler_can_use_kist(void)
+{
+ if (kist_no_kernel_support) {
+ /* We have no kernel support so we can't use KIST. */
+ return 0;
+ }
+
+ /* We do have the support, time to check if we can get the interval that the
+ * consensus can be disabling. */
+ int run_interval = kist_scheduler_run_interval();
+ log_debug(LD_SCHED, "Determined KIST sched_run_interval should be "
+ "%" PRId32 ". Can%s use KIST.",
+ run_interval, (run_interval > 0 ? "" : " not"));
+ return run_interval > 0;
+}
+
+#else /* !(defined(HAVE_KIST_SUPPORT)) */
+
+int
+scheduler_can_use_kist(void)
+{
+ return 0;
+}
+
+#endif /* defined(HAVE_KIST_SUPPORT) */
diff --cc src/feature/stats/rephist.c
index 405efc26e,000000000..edb079ef2
mode 100644,000000..100644
--- a/src/feature/stats/rephist.c
+++ b/src/feature/stats/rephist.c
@@@ -1,3222 -1,0 +1,3227 @@@
+/* Copyright (c) 2004-2006, Roger Dingledine, Nick Mathewson.
+ * Copyright (c) 2007-2018, The Tor Project, Inc. */
+/* See LICENSE for licensing information */
+
+/**
+ * \file rephist.c
+ * \brief Basic history and performance-tracking functionality.
+ *
+ * Basic history and performance-tracking functionality to remember
+ * which servers have worked in the past, how much bandwidth we've
+ * been using, which ports we tend to want, and so on; further,
+ * exit port statistics, cell statistics, and connection statistics.
+ *
+ * The history and information tracked in this module could sensibly be
+ * divided into several categories:
+ *
+ * <ul><li>Statistics used by authorities to remember the uptime and
+ * stability information about various relays, including "uptime",
+ * "weighted fractional uptime" and "mean time between failures".
+ *
+ * <li>Bandwidth usage history, used by relays to self-report how much
+ * bandwidth they've used for different purposes over last day or so,
+ * in order to generate the {dirreq-,}{read,write}-history lines in
+ * that they publish.
+ *
+ * <li>Predicted ports, used by clients to remember how long it's been
+ * since they opened an exit connection to each given target
+ * port. Clients use this information in order to try to keep circuits
+ * open to exit nodes that can connect to the ports that they care
+ * about. (The predicted ports mechanism also handles predicted circuit
+ * usage that _isn't_ port-specific, such as resolves, internal circuits,
+ * and so on.)
+ *
+ * <li>Public key operation counters, for tracking how many times we've
+ * done each public key operation. (This is unmaintained and we should
+ * remove it.)
+ *
+ * <li>Exit statistics by port, used by exits to keep track of the
+ * number of streams and bytes they've served at each exit port, so they
+ * can generate their exit-kibibytes-{read,written} and
+ * exit-streams-opened statistics.
+ *
+ * <li>Circuit stats, used by relays instances to tract circuit
+ * queue fullness and delay over time, and generate cell-processed-cells,
+ * cell-queued-cells, cell-time-in-queue, and cell-circuits-per-decile
+ * statistics.
+ *
+ * <li>Descriptor serving statistics, used by directory caches to track
+ * how many descriptors they've served.
+ *
+ * <li>Connection statistics, used by relays to track one-way and
+ * bidirectional connections.
+ *
+ * <li>Onion handshake statistics, used by relays to count how many
+ * TAP and ntor handshakes they've handled.
+ *
+ * <li>Hidden service statistics, used by relays to count rendezvous
+ * traffic and HSDir-stored descriptors.
+ *
+ * <li>Link protocol statistics, used by relays to count how many times
+ * each link protocol has been used.
+ *
+ * </ul>
+ *
+ * The entry points for this module are scattered throughout the
+ * codebase. Sending data, receiving data, connecting to a relay,
+ * losing a connection to a relay, and so on can all trigger a change in
+ * our current stats. Relays also invoke this module in order to
+ * extract their statistics when building routerinfo and extrainfo
+ * objects in router.c.
+ *
+ * TODO: This module should be broken up.
+ *
+ * (The "rephist" name originally stood for "reputation and history". )
+ **/
+
+#include "core/or/or.h"
+#include "core/or/circuitlist.h"
+#include "core/or/circuituse.h"
+#include "app/config/config.h"
+#include "lib/crypt_ops/crypto_rand.h"
+#include "feature/nodelist/networkstatus.h"
+#include "feature/nodelist/nodelist.h"
+#include "feature/stats/rephist.h"
+#include "feature/relay/router.h"
+#include "feature/nodelist/routerlist.h"
+#include "ht.h"
+#include "core/or/channelpadding.h"
+#include "core/or/connection_or.h"
+#include "app/config/statefile.h"
+
+#include "feature/nodelist/networkstatus_st.h"
+#include "core/or/or_circuit_st.h"
+#include "app/config/or_state_st.h"
+
+#include "lib/container/bloomfilt.h"
+#include "lib/container/order.h"
+#include "lib/math/fp.h"
+#include "lib/math/laplace.h"
+#include "lib/time/tvdiff.h"
+
+#ifdef HAVE_FCNTL_H
+#include <fcntl.h>
+#endif
+
+static void bw_arrays_init(void);
+static void predicted_ports_alloc(void);
+
++typedef struct bw_array_t bw_array_t;
++STATIC uint64_t find_largest_max(bw_array_t *b);
++STATIC void commit_max(bw_array_t *b);
++STATIC void advance_obs(bw_array_t *b);
++
+/** Total number of bytes currently allocated in fields used by rephist.c. */
+uint64_t rephist_total_alloc=0;
+/** Number of or_history_t objects currently allocated. */
+uint32_t rephist_total_num=0;
+
+/** If the total weighted run count of all runs for a router ever falls
+ * below this amount, the router can be treated as having 0 MTBF. */
+#define STABILITY_EPSILON 0.0001
+/** Value by which to discount all old intervals for MTBF purposes. This
+ * is compounded every STABILITY_INTERVAL. */
+#define STABILITY_ALPHA 0.95
+/** Interval at which to discount all old intervals for MTBF purposes. */
+#define STABILITY_INTERVAL (12*60*60)
+/* (This combination of ALPHA, INTERVAL, and EPSILON makes it so that an
+ * interval that just ended counts twice as much as one that ended a week ago,
+ * 20X as much as one that ended a month ago, and routers that have had no
+ * uptime data for about half a year will get forgotten.) */
+
+/** History of an OR. */
+typedef struct or_history_t {
+ /** When did we start tracking this OR? */
+ time_t since;
+ /** When did we most recently note a change to this OR? */
+ time_t changed;
+
+ /** The address at which we most recently connected to this OR
+ * successfully. */
+ tor_addr_t last_reached_addr;
+
+ /** The port at which we most recently connected to this OR successfully */
+ uint16_t last_reached_port;
+
+ /* === For MTBF tracking: */
+ /** Weighted sum total of all times that this router has been online.
+ */
+ unsigned long weighted_run_length;
+ /** If the router is now online (according to stability-checking rules),
+ * when did it come online? */
+ time_t start_of_run;
+ /** Sum of weights for runs in weighted_run_length. */
+ double total_run_weights;
+ /* === For fractional uptime tracking: */
+ time_t start_of_downtime;
+ unsigned long weighted_uptime;
+ unsigned long total_weighted_time;
+} or_history_t;
+
+/**
+ * This structure holds accounting needed to calculate the padding overhead.
+ */
+typedef struct padding_counts_t {
+ /** Total number of cells we have received, including padding */
+ uint64_t read_cell_count;
+ /** Total number of cells we have sent, including padding */
+ uint64_t write_cell_count;
+ /** Total number of CELL_PADDING cells we have received */
+ uint64_t read_pad_cell_count;
+ /** Total number of CELL_PADDING cells we have sent */
+ uint64_t write_pad_cell_count;
+ /** Total number of read cells on padding-enabled conns */
+ uint64_t enabled_read_cell_count;
+ /** Total number of sent cells on padding-enabled conns */
+ uint64_t enabled_write_cell_count;
+ /** Total number of read CELL_PADDING cells on padding-enabled cons */
+ uint64_t enabled_read_pad_cell_count;
+ /** Total number of sent CELL_PADDING cells on padding-enabled cons */
+ uint64_t enabled_write_pad_cell_count;
+ /** Total number of RELAY_DROP cells we have received */
+ uint64_t read_drop_cell_count;
+ /** Total number of RELAY_DROP cells we have sent */
+ uint64_t write_drop_cell_count;
+ /** The maximum number of padding timers we've seen in 24 hours */
+ uint64_t maximum_chanpad_timers;
+ /** When did we first copy padding_current into padding_published? */
+ char first_published_at[ISO_TIME_LEN+1];
+} padding_counts_t;
+
+/** Holds the current values of our padding statistics.
+ * It is not published until it is transferred to padding_published. */
+static padding_counts_t padding_current;
+
+/** Remains fixed for a 24 hour period, and then is replaced
+ * by a redacted copy of padding_current */
+static padding_counts_t padding_published;
+
+/** When did we last multiply all routers' weighted_run_length and
+ * total_run_weights by STABILITY_ALPHA? */
+static time_t stability_last_downrated = 0;
+
+/** */
+static time_t started_tracking_stability = 0;
+
+/** Map from hex OR identity digest to or_history_t. */
+static digestmap_t *history_map = NULL;
+
+/** Return the or_history_t for the OR with identity digest <b>id</b>,
+ * creating it if necessary. */
+static or_history_t *
+get_or_history(const char* id)
+{
+ or_history_t *hist;
+
+ if (tor_digest_is_zero(id))
+ return NULL;
+
+ hist = digestmap_get(history_map, id);
+ if (!hist) {
+ hist = tor_malloc_zero(sizeof(or_history_t));
+ rephist_total_alloc += sizeof(or_history_t);
+ rephist_total_num++;
+ hist->since = hist->changed = time(NULL);
+ tor_addr_make_unspec(&hist->last_reached_addr);
+ digestmap_set(history_map, id, hist);
+ }
+ return hist;
+}
+
+/** Helper: free storage held by a single OR history entry. */
+static void
+free_or_history(void *_hist)
+{
+ or_history_t *hist = _hist;
+ rephist_total_alloc -= sizeof(or_history_t);
+ rephist_total_num--;
+ tor_free(hist);
+}
+
+/** Initialize the static data structures for tracking history. */
+void
+rep_hist_init(void)
+{
+ history_map = digestmap_new();
+ bw_arrays_init();
+ predicted_ports_alloc();
+}
+
+/** We have just decided that this router with identity digest <b>id</b> is
+ * reachable, meaning we will give it a "Running" flag for the next while. */
+void
+rep_hist_note_router_reachable(const char *id, const tor_addr_t *at_addr,
+ const uint16_t at_port, time_t when)
+{
+ or_history_t *hist = get_or_history(id);
+ int was_in_run = 1;
+ char tbuf[ISO_TIME_LEN+1];
+ int addr_changed, port_changed;
+
+ tor_assert(hist);
+ tor_assert((!at_addr && !at_port) || (at_addr && at_port));
+
+ addr_changed = at_addr && !tor_addr_is_null(&hist->last_reached_addr) &&
+ tor_addr_compare(at_addr, &hist->last_reached_addr, CMP_EXACT) != 0;
+ port_changed = at_port && hist->last_reached_port &&
+ at_port != hist->last_reached_port;
+
+ if (!started_tracking_stability)
+ started_tracking_stability = time(NULL);
+ if (!hist->start_of_run) {
+ hist->start_of_run = when;
+ was_in_run = 0;
+ }
+ if (hist->start_of_downtime) {
+ long down_length;
+
+ format_local_iso_time(tbuf, hist->start_of_downtime);
+ log_info(LD_HIST, "Router %s is now Running; it had been down since %s.",
+ hex_str(id, DIGEST_LEN), tbuf);
+ if (was_in_run)
+ log_info(LD_HIST, " (Paradoxically, it was already Running too.)");
+
+ down_length = when - hist->start_of_downtime;
+ hist->total_weighted_time += down_length;
+ hist->start_of_downtime = 0;
+ } else if (addr_changed || port_changed) {
+ /* If we're reachable, but the address changed, treat this as some
+ * downtime. */
+ int penalty = get_options()->TestingTorNetwork ? 240 : 3600;
+ networkstatus_t *ns;
+
+ if ((ns = networkstatus_get_latest_consensus())) {
+ int fresh_interval = (int)(ns->fresh_until - ns->valid_after);
+ int live_interval = (int)(ns->valid_until - ns->valid_after);
+ /* on average, a descriptor addr change takes .5 intervals to make it
+ * into a consensus, and half a liveness period to make it to
+ * clients. */
+ penalty = (int)(fresh_interval + live_interval) / 2;
+ }
+ format_local_iso_time(tbuf, hist->start_of_run);
+ log_info(LD_HIST,"Router %s still seems Running, but its address appears "
+ "to have changed since the last time it was reachable. I'm "
+ "going to treat it as having been down for %d seconds",
+ hex_str(id, DIGEST_LEN), penalty);
+ rep_hist_note_router_unreachable(id, when-penalty);
+ rep_hist_note_router_reachable(id, NULL, 0, when);
+ } else {
+ format_local_iso_time(tbuf, hist->start_of_run);
+ if (was_in_run)
+ log_debug(LD_HIST, "Router %s is still Running; it has been Running "
+ "since %s", hex_str(id, DIGEST_LEN), tbuf);
+ else
+ log_info(LD_HIST,"Router %s is now Running; it was previously untracked",
+ hex_str(id, DIGEST_LEN));
+ }
+ if (at_addr)
+ tor_addr_copy(&hist->last_reached_addr, at_addr);
+ if (at_port)
+ hist->last_reached_port = at_port;
+}
+
+/** We have just decided that this router is unreachable, meaning
+ * we are taking away its "Running" flag. */
+void
+rep_hist_note_router_unreachable(const char *id, time_t when)
+{
+ or_history_t *hist = get_or_history(id);
+ char tbuf[ISO_TIME_LEN+1];
+ int was_running = 0;
+ if (!started_tracking_stability)
+ started_tracking_stability = time(NULL);
+
+ tor_assert(hist);
+ if (hist->start_of_run) {
+ /*XXXX We could treat failed connections differently from failed
+ * connect attempts. */
+ long run_length = when - hist->start_of_run;
+ format_local_iso_time(tbuf, hist->start_of_run);
+
+ hist->total_run_weights += 1.0;
+ hist->start_of_run = 0;
+ if (run_length < 0) {
+ unsigned long penalty = -run_length;
+#define SUBTRACT_CLAMPED(var, penalty) \
+ do { (var) = (var) < (penalty) ? 0 : (var) - (penalty); } while (0)
+
+ SUBTRACT_CLAMPED(hist->weighted_run_length, penalty);
+ SUBTRACT_CLAMPED(hist->weighted_uptime, penalty);
+ } else {
+ hist->weighted_run_length += run_length;
+ hist->weighted_uptime += run_length;
+ hist->total_weighted_time += run_length;
+ }
+ was_running = 1;
+ log_info(LD_HIST, "Router %s is now non-Running: it had previously been "
+ "Running since %s. Its total weighted uptime is %lu/%lu.",
+ hex_str(id, DIGEST_LEN), tbuf, hist->weighted_uptime,
+ hist->total_weighted_time);
+ }
+ if (!hist->start_of_downtime) {
+ hist->start_of_downtime = when;
+
+ if (!was_running)
+ log_info(LD_HIST, "Router %s is now non-Running; it was previously "
+ "untracked.", hex_str(id, DIGEST_LEN));
+ } else {
+ if (!was_running) {
+ format_local_iso_time(tbuf, hist->start_of_downtime);
+
+ log_info(LD_HIST, "Router %s is still non-Running; it has been "
+ "non-Running since %s.", hex_str(id, DIGEST_LEN), tbuf);
+ }
+ }
+}
+
+/** Mark a router with ID <b>id</b> as non-Running, and retroactively declare
+ * that it has never been running: give it no stability and no WFU. */
+void
+rep_hist_make_router_pessimal(const char *id, time_t when)
+{
+ or_history_t *hist = get_or_history(id);
+ tor_assert(hist);
+
+ rep_hist_note_router_unreachable(id, when);
+
+ hist->weighted_run_length = 0;
+ hist->weighted_uptime = 0;
+}
+
+/** Helper: Discount all old MTBF data, if it is time to do so. Return
+ * the time at which we should next discount MTBF data. */
+time_t
+rep_hist_downrate_old_runs(time_t now)
+{
+ digestmap_iter_t *orhist_it;
+ const char *digest1;
+ or_history_t *hist;
+ void *hist_p;
+ double alpha = 1.0;
+
+ if (!history_map)
+ history_map = digestmap_new();
+ if (!stability_last_downrated)
+ stability_last_downrated = now;
+ if (stability_last_downrated + STABILITY_INTERVAL > now)
+ return stability_last_downrated + STABILITY_INTERVAL;
+
+ /* Okay, we should downrate the data. By how much? */
+ while (stability_last_downrated + STABILITY_INTERVAL < now) {
+ stability_last_downrated += STABILITY_INTERVAL;
+ alpha *= STABILITY_ALPHA;
+ }
+
+ log_info(LD_HIST, "Discounting all old stability info by a factor of %f",
+ alpha);
+
+ /* Multiply every w_r_l, t_r_w pair by alpha. */
+ for (orhist_it = digestmap_iter_init(history_map);
+ !digestmap_iter_done(orhist_it);
+ orhist_it = digestmap_iter_next(history_map,orhist_it)) {
+ digestmap_iter_get(orhist_it, &digest1, &hist_p);
+ hist = hist_p;
+
+ hist->weighted_run_length =
+ (unsigned long)(hist->weighted_run_length * alpha);
+ hist->total_run_weights *= alpha;
+
+ hist->weighted_uptime = (unsigned long)(hist->weighted_uptime * alpha);
+ hist->total_weighted_time = (unsigned long)
+ (hist->total_weighted_time * alpha);
+ }
+
+ return stability_last_downrated + STABILITY_INTERVAL;
+}
+
+/** Helper: Return the weighted MTBF of the router with history <b>hist</b>. */
+static double
+get_stability(or_history_t *hist, time_t when)
+{
+ long total = hist->weighted_run_length;
+ double total_weights = hist->total_run_weights;
+
+ if (hist->start_of_run) {
+ /* We're currently in a run. Let total and total_weights hold the values
+ * they would hold if the current run were to end now. */
+ total += (when-hist->start_of_run);
+ total_weights += 1.0;
+ }
+ if (total_weights < STABILITY_EPSILON) {
+ /* Round down to zero, and avoid divide-by-zero. */
+ return 0.0;
+ }
+
+ return total / total_weights;
+}
+
+/** Return the total amount of time we've been observing, with each run of
+ * time downrated by the appropriate factor. */
+static long
+get_total_weighted_time(or_history_t *hist, time_t when)
+{
+ long total = hist->total_weighted_time;
+ if (hist->start_of_run) {
+ total += (when - hist->start_of_run);
+ } else if (hist->start_of_downtime) {
+ total += (when - hist->start_of_downtime);
+ }
+ return total;
+}
+
+/** Helper: Return the weighted percent-of-time-online of the router with
+ * history <b>hist</b>. */
+static double
+get_weighted_fractional_uptime(or_history_t *hist, time_t when)
+{
+ long total = hist->total_weighted_time;
+ long up = hist->weighted_uptime;
+
+ if (hist->start_of_run) {
+ long run_length = (when - hist->start_of_run);
+ up += run_length;
+ total += run_length;
+ } else if (hist->start_of_downtime) {
+ total += (when - hist->start_of_downtime);
+ }
+
+ if (!total) {
+ /* Avoid calling anybody's uptime infinity (which should be impossible if
+ * the code is working), or NaN (which can happen for any router we haven't
+ * observed up or down yet). */
+ return 0.0;
+ }
+
+ return ((double) up) / total;
+}
+
+/** Return how long the router whose identity digest is <b>id</b> has
+ * been reachable. Return 0 if the router is unknown or currently deemed
+ * unreachable. */
+long
+rep_hist_get_uptime(const char *id, time_t when)
+{
+ or_history_t *hist = get_or_history(id);
+ if (!hist)
+ return 0;
+ if (!hist->start_of_run || when < hist->start_of_run)
+ return 0;
+ return when - hist->start_of_run;
+}
+
+/** Return an estimated MTBF for the router whose identity digest is
+ * <b>id</b>. Return 0 if the router is unknown. */
+double
+rep_hist_get_stability(const char *id, time_t when)
+{
+ or_history_t *hist = get_or_history(id);
+ if (!hist)
+ return 0.0;
+
+ return get_stability(hist, when);
+}
+
+/** Return an estimated percent-of-time-online for the router whose identity
+ * digest is <b>id</b>. Return 0 if the router is unknown. */
+double
+rep_hist_get_weighted_fractional_uptime(const char *id, time_t when)
+{
+ or_history_t *hist = get_or_history(id);
+ if (!hist)
+ return 0.0;
+
+ return get_weighted_fractional_uptime(hist, when);
+}
+
+/** Return a number representing how long we've known about the router whose
+ * digest is <b>id</b>. Return 0 if the router is unknown.
+ *
+ * Be careful: this measure increases monotonically as we know the router for
+ * longer and longer, but it doesn't increase linearly.
+ */
+long
+rep_hist_get_weighted_time_known(const char *id, time_t when)
+{
+ or_history_t *hist = get_or_history(id);
+ if (!hist)
+ return 0;
+
+ return get_total_weighted_time(hist, when);
+}
+
+/** Return true if we've been measuring MTBFs for long enough to
+ * pronounce on Stability. */
+int
+rep_hist_have_measured_enough_stability(void)
+{
+ /* XXXX++ This doesn't do so well when we change our opinion
+ * as to whether we're tracking router stability. */
+ return started_tracking_stability < time(NULL) - 4*60*60;
+}
+
+/** Log all the reliability data we have remembered, with the chosen
+ * severity.
+ */
+void
+rep_hist_dump_stats(time_t now, int severity)
+{
+ digestmap_iter_t *orhist_it;
+ const char *name1, *digest1;
+ char hexdigest1[HEX_DIGEST_LEN+1];
+ or_history_t *or_history;
+ void *or_history_p;
+ const node_t *node;
+
+ rep_history_clean(now - get_options()->RephistTrackTime);
+
+ tor_log(severity, LD_HIST, "--------------- Dumping history information:");
+
+ for (orhist_it = digestmap_iter_init(history_map);
+ !digestmap_iter_done(orhist_it);
+ orhist_it = digestmap_iter_next(history_map,orhist_it)) {
+ double s;
+ long stability;
+ digestmap_iter_get(orhist_it, &digest1, &or_history_p);
+ or_history = (or_history_t*) or_history_p;
+
+ if ((node = node_get_by_id(digest1)) && node_get_nickname(node))
+ name1 = node_get_nickname(node);
+ else
+ name1 = "(unknown)";
+ base16_encode(hexdigest1, sizeof(hexdigest1), digest1, DIGEST_LEN);
+ s = get_stability(or_history, now);
+ stability = (long)s;
+ tor_log(severity, LD_HIST,
+ "OR %s [%s]: wmtbf %lu:%02lu:%02lu",
+ name1, hexdigest1,
+ stability/3600, (stability/60)%60, stability%60);
+ }
+}
+
+/** Remove history info for routers/links that haven't changed since
+ * <b>before</b>.
+ */
+void
+rep_history_clean(time_t before)
+{
+ int authority = authdir_mode(get_options());
+ or_history_t *or_history;
+ void *or_history_p;
+ digestmap_iter_t *orhist_it;
+ const char *d1;
+
+ orhist_it = digestmap_iter_init(history_map);
+ while (!digestmap_iter_done(orhist_it)) {
+ int should_remove;
+ digestmap_iter_get(orhist_it, &d1, &or_history_p);
+ or_history = or_history_p;
+
+ should_remove = authority ?
+ (or_history->total_run_weights < STABILITY_EPSILON &&
+ !or_history->start_of_run)
+ : (or_history->changed < before);
+ if (should_remove) {
+ orhist_it = digestmap_iter_next_rmv(history_map, orhist_it);
+ free_or_history(or_history);
+ continue;
+ }
+ orhist_it = digestmap_iter_next(history_map, orhist_it);
+ }
+}
+
+/** Write MTBF data to disk. Return 0 on success, negative on failure.
+ *
+ * If <b>missing_means_down</b>, then if we're about to write an entry
+ * that is still considered up but isn't in our routerlist, consider it
+ * to be down. */
+int
+rep_hist_record_mtbf_data(time_t now, int missing_means_down)
+{
+ char time_buf[ISO_TIME_LEN+1];
+
+ digestmap_iter_t *orhist_it;
+ const char *digest;
+ void *or_history_p;
+ or_history_t *hist;
+ open_file_t *open_file = NULL;
+ FILE *f;
+
+ {
+ char *filename = get_datadir_fname("router-stability");
+ f = start_writing_to_stdio_file(filename, OPEN_FLAGS_REPLACE|O_TEXT, 0600,
+ &open_file);
+ tor_free(filename);
+ if (!f)
+ return -1;
+ }
+
+ /* File format is:
+ * FormatLine *KeywordLine Data
+ *
+ * FormatLine = "format 1" NL
+ * KeywordLine = Keyword SP Arguments NL
+ * Data = "data" NL *RouterMTBFLine "." NL
+ * RouterMTBFLine = Fingerprint SP WeightedRunLen SP
+ * TotalRunWeights [SP S=StartRunTime] NL
+ */
+#define PUT(s) STMT_BEGIN if (fputs((s),f)<0) goto err; STMT_END
+#define PRINTF(args) STMT_BEGIN if (fprintf args <0) goto err; STMT_END
+
+ PUT("format 2\n");
+
+ format_iso_time(time_buf, time(NULL));
+ PRINTF((f, "stored-at %s\n", time_buf));
+
+ if (started_tracking_stability) {
+ format_iso_time(time_buf, started_tracking_stability);
+ PRINTF((f, "tracked-since %s\n", time_buf));
+ }
+ if (stability_last_downrated) {
+ format_iso_time(time_buf, stability_last_downrated);
+ PRINTF((f, "last-downrated %s\n", time_buf));
+ }
+
+ PUT("data\n");
+
+ /* XXX Nick: now bridge auths record this for all routers too.
+ * Should we make them record it only for bridge routers? -RD
+ * Not for 0.2.0. -NM */
+ for (orhist_it = digestmap_iter_init(history_map);
+ !digestmap_iter_done(orhist_it);
+ orhist_it = digestmap_iter_next(history_map,orhist_it)) {
+ char dbuf[HEX_DIGEST_LEN+1];
+ const char *t = NULL;
+ digestmap_iter_get(orhist_it, &digest, &or_history_p);
+ hist = (or_history_t*) or_history_p;
+
+ base16_encode(dbuf, sizeof(dbuf), digest, DIGEST_LEN);
+
+ if (missing_means_down && hist->start_of_run &&
+ !connection_or_digest_is_known_relay(digest)) {
+ /* We think this relay is running, but it's not listed in our
+ * consensus. Somehow it fell out without telling us it went
+ * down. Complain and also correct it. */
+ log_info(LD_HIST,
+ "Relay '%s' is listed as up in rephist, but it's not in "
+ "our routerlist. Correcting.", dbuf);
+ rep_hist_note_router_unreachable(digest, now);
+ }
+
+ PRINTF((f, "R %s\n", dbuf));
+ if (hist->start_of_run > 0) {
+ format_iso_time(time_buf, hist->start_of_run);
+ t = time_buf;
+ }
+ PRINTF((f, "+MTBF %lu %.5f%s%s\n",
+ hist->weighted_run_length, hist->total_run_weights,
+ t ? " S=" : "", t ? t : ""));
+ t = NULL;
+ if (hist->start_of_downtime > 0) {
+ format_iso_time(time_buf, hist->start_of_downtime);
+ t = time_buf;
+ }
+ PRINTF((f, "+WFU %lu %lu%s%s\n",
+ hist->weighted_uptime, hist->total_weighted_time,
+ t ? " S=" : "", t ? t : ""));
+ }
+
+ PUT(".\n");
+
+#undef PUT
+#undef PRINTF
+
+ return finish_writing_to_file(open_file);
+ err:
+ abort_writing_to_file(open_file);
+ return -1;
+}
+
+/** Helper: return the first j >= i such that !strcmpstart(sl[j], prefix) and
+ * such that no line sl[k] with i <= k < j starts with "R ". Return -1 if no
+ * such line exists. */
+static int
+find_next_with(smartlist_t *sl, int i, const char *prefix)
+{
+ for ( ; i < smartlist_len(sl); ++i) {
+ const char *line = smartlist_get(sl, i);
+ if (!strcmpstart(line, prefix))
+ return i;
+ if (!strcmpstart(line, "R "))
+ return -1;
+ }
+ return -1;
+}
+
+/** How many bad times has parse_possibly_bad_iso_time() parsed? */
+static int n_bogus_times = 0;
+/** Parse the ISO-formatted time in <b>s</b> into *<b>time_out</b>, but
+ * round any pre-1970 date to Jan 1, 1970. */
+static int
+parse_possibly_bad_iso_time(const char *s, time_t *time_out)
+{
+ int year;
+ char b[5];
+ strlcpy(b, s, sizeof(b));
+ b[4] = '\0';
+ year = (int)tor_parse_long(b, 10, 0, INT_MAX, NULL, NULL);
+ if (year < 1970) {
+ *time_out = 0;
+ ++n_bogus_times;
+ return 0;
+ } else
+ return parse_iso_time(s, time_out);
+}
+
+/** We've read a time <b>t</b> from a file stored at <b>stored_at</b>, which
+ * says we started measuring at <b>started_measuring</b>. Return a new number
+ * that's about as much before <b>now</b> as <b>t</b> was before
+ * <b>stored_at</b>.
+ */
+static inline time_t
+correct_time(time_t t, time_t now, time_t stored_at, time_t started_measuring)
+{
+ if (t < started_measuring - 24*60*60*365)
+ return 0;
+ else if (t < started_measuring)
+ return started_measuring;
+ else if (t > stored_at)
+ return 0;
+ else {
+ long run_length = stored_at - t;
+ t = (time_t)(now - run_length);
+ if (t < started_measuring)
+ t = started_measuring;
+ return t;
+ }
+}
+
+/** Load MTBF data from disk. Returns 0 on success or recoverable error, -1
+ * on failure. */
+int
+rep_hist_load_mtbf_data(time_t now)
+{
+ /* XXXX won't handle being called while history is already populated. */
+ smartlist_t *lines;
+ const char *line = NULL;
+ int r=0, i;
+ time_t last_downrated = 0, stored_at = 0, tracked_since = 0;
+ time_t latest_possible_start = now;
+ long format = -1;
+
+ {
+ char *filename = get_datadir_fname("router-stability");
+ char *d = read_file_to_str(filename, RFTS_IGNORE_MISSING, NULL);
+ tor_free(filename);
+ if (!d)
+ return -1;
+ lines = smartlist_new();
+ smartlist_split_string(lines, d, "\n", SPLIT_SKIP_SPACE, 0);
+ tor_free(d);
+ }
+
+ {
+ const char *firstline;
+ if (smartlist_len(lines)>4) {
+ firstline = smartlist_get(lines, 0);
+ if (!strcmpstart(firstline, "format "))
+ format = tor_parse_long(firstline+strlen("format "),
+ 10, -1, LONG_MAX, NULL, NULL);
+ }
+ }
+ if (format != 1 && format != 2) {
+ log_warn(LD_HIST,
+ "Unrecognized format in mtbf history file. Skipping.");
+ goto err;
+ }
+ for (i = 1; i < smartlist_len(lines); ++i) {
+ line = smartlist_get(lines, i);
+ if (!strcmp(line, "data"))
+ break;
+ if (!strcmpstart(line, "last-downrated ")) {
+ if (parse_iso_time(line+strlen("last-downrated "), &last_downrated)<0)
+ log_warn(LD_HIST,"Couldn't parse downrate time in mtbf "
+ "history file.");
+ }
+ if (!strcmpstart(line, "stored-at ")) {
+ if (parse_iso_time(line+strlen("stored-at "), &stored_at)<0)
+ log_warn(LD_HIST,"Couldn't parse stored time in mtbf "
+ "history file.");
+ }
+ if (!strcmpstart(line, "tracked-since ")) {
+ if (parse_iso_time(line+strlen("tracked-since "), &tracked_since)<0)
+ log_warn(LD_HIST,"Couldn't parse started-tracking time in mtbf "
+ "history file.");
+ }
+ }
+ if (last_downrated > now)
+ last_downrated = now;
+ if (tracked_since > now)
+ tracked_since = now;
+
+ if (!stored_at) {
+ log_warn(LD_HIST, "No stored time recorded.");
+ goto err;
+ }
+
+ if (line && !strcmp(line, "data"))
+ ++i;
+
+ n_bogus_times = 0;
+
+ for (; i < smartlist_len(lines); ++i) {
+ char digest[DIGEST_LEN];
+ char hexbuf[HEX_DIGEST_LEN+1];
+ char mtbf_timebuf[ISO_TIME_LEN+1];
+ char wfu_timebuf[ISO_TIME_LEN+1];
+ time_t start_of_run = 0;
+ time_t start_of_downtime = 0;
+ int have_mtbf = 0, have_wfu = 0;
+ long wrl = 0;
+ double trw = 0;
+ long wt_uptime = 0, total_wt_time = 0;
+ int n;
+ or_history_t *hist;
+ line = smartlist_get(lines, i);
+ if (!strcmp(line, "."))
+ break;
+
+ mtbf_timebuf[0] = '\0';
+ wfu_timebuf[0] = '\0';
+
+ if (format == 1) {
+ n = tor_sscanf(line, "%40s %ld %lf S=%10s %8s",
+ hexbuf, &wrl, &trw, mtbf_timebuf, mtbf_timebuf+11);
+ if (n != 3 && n != 5) {
+ log_warn(LD_HIST, "Couldn't scan line %s", escaped(line));
+ continue;
+ }
+ have_mtbf = 1;
+ } else {
+ // format == 2.
+ int mtbf_idx, wfu_idx;
+ if (strcmpstart(line, "R ") || strlen(line) < 2+HEX_DIGEST_LEN)
+ continue;
+ strlcpy(hexbuf, line+2, sizeof(hexbuf));
+ mtbf_idx = find_next_with(lines, i+1, "+MTBF ");
+ wfu_idx = find_next_with(lines, i+1, "+WFU ");
+ if (mtbf_idx >= 0) {
+ const char *mtbfline = smartlist_get(lines, mtbf_idx);
+ n = tor_sscanf(mtbfline, "+MTBF %lu %lf S=%10s %8s",
+ &wrl, &trw, mtbf_timebuf, mtbf_timebuf+11);
+ if (n == 2 || n == 4) {
+ have_mtbf = 1;
+ } else {
+ log_warn(LD_HIST, "Couldn't scan +MTBF line %s",
+ escaped(mtbfline));
+ }
+ }
+ if (wfu_idx >= 0) {
+ const char *wfuline = smartlist_get(lines, wfu_idx);
+ n = tor_sscanf(wfuline, "+WFU %lu %lu S=%10s %8s",
+ &wt_uptime, &total_wt_time,
+ wfu_timebuf, wfu_timebuf+11);
+ if (n == 2 || n == 4) {
+ have_wfu = 1;
+ } else {
+ log_warn(LD_HIST, "Couldn't scan +WFU line %s", escaped(wfuline));
+ }
+ }
+ if (wfu_idx > i)
+ i = wfu_idx;
+ if (mtbf_idx > i)
+ i = mtbf_idx;
+ }
+ if (base16_decode(digest, DIGEST_LEN,
+ hexbuf, HEX_DIGEST_LEN) != DIGEST_LEN) {
+ log_warn(LD_HIST, "Couldn't hex string %s", escaped(hexbuf));
+ continue;
+ }
+ hist = get_or_history(digest);
+ if (!hist)
+ continue;
+
+ if (have_mtbf) {
+ if (mtbf_timebuf[0]) {
+ mtbf_timebuf[10] = ' ';
+ if (parse_possibly_bad_iso_time(mtbf_timebuf, &start_of_run)<0)
+ log_warn(LD_HIST, "Couldn't parse time %s",
+ escaped(mtbf_timebuf));
+ }
+ hist->start_of_run = correct_time(start_of_run, now, stored_at,
+ tracked_since);
+ if (hist->start_of_run < latest_possible_start + wrl)
+ latest_possible_start = (time_t)(hist->start_of_run - wrl);
+
+ hist->weighted_run_length = wrl;
+ hist->total_run_weights = trw;
+ }
+ if (have_wfu) {
+ if (wfu_timebuf[0]) {
+ wfu_timebuf[10] = ' ';
+ if (parse_possibly_bad_iso_time(wfu_timebuf, &start_of_downtime)<0)
+ log_warn(LD_HIST, "Couldn't parse time %s", escaped(wfu_timebuf));
+ }
+ }
+ hist->start_of_downtime = correct_time(start_of_downtime, now, stored_at,
+ tracked_since);
+ hist->weighted_uptime = wt_uptime;
+ hist->total_weighted_time = total_wt_time;
+ }
+ if (strcmp(line, "."))
+ log_warn(LD_HIST, "Truncated MTBF file.");
+
+ if (tracked_since < 86400*365) /* Recover from insanely early value. */
+ tracked_since = latest_possible_start;
+
+ stability_last_downrated = last_downrated;
+ started_tracking_stability = tracked_since;
+
+ goto done;
+ err:
+ r = -1;
+ done:
+ SMARTLIST_FOREACH(lines, char *, cp, tor_free(cp));
+ smartlist_free(lines);
+ return r;
+}
+
+/** For how many seconds do we keep track of individual per-second bandwidth
+ * totals? */
+#define NUM_SECS_ROLLING_MEASURE 10
+/** How large are the intervals for which we track and report bandwidth use? */
+#define NUM_SECS_BW_SUM_INTERVAL (24*60*60)
+/** How far in the past do we remember and publish bandwidth use? */
+#define NUM_SECS_BW_SUM_IS_VALID (5*24*60*60)
+/** How many bandwidth usage intervals do we remember? (derived) */
+#define NUM_TOTALS (NUM_SECS_BW_SUM_IS_VALID/NUM_SECS_BW_SUM_INTERVAL)
+
+/** Structure to track bandwidth use, and remember the maxima for a given
+ * time period.
+ */
+typedef struct bw_array_t {
+ /** Observation array: Total number of bytes transferred in each of the last
+ * NUM_SECS_ROLLING_MEASURE seconds. This is used as a circular array. */
+ uint64_t obs[NUM_SECS_ROLLING_MEASURE];
+ int cur_obs_idx; /**< Current position in obs. */
+ time_t cur_obs_time; /**< Time represented in obs[cur_obs_idx] */
+ uint64_t total_obs; /**< Total for all members of obs except
+ * obs[cur_obs_idx] */
+ uint64_t max_total; /**< Largest value that total_obs has taken on in the
+ * current period. */
+ uint64_t total_in_period; /**< Total bytes transferred in the current
+ * period. */
+
+ /** When does the next period begin? */
+ time_t next_period;
+ /** Where in 'maxima' should the maximum bandwidth usage for the current
+ * period be stored? */
+ int next_max_idx;
+ /** How many values in maxima/totals have been set ever? */
+ int num_maxes_set;
+ /** Circular array of the maximum
+ * bandwidth-per-NUM_SECS_ROLLING_MEASURE usage for the last
+ * NUM_TOTALS periods */
+ uint64_t maxima[NUM_TOTALS];
+ /** Circular array of the total bandwidth usage for the last NUM_TOTALS
+ * periods */
+ uint64_t totals[NUM_TOTALS];
+} bw_array_t;
+
+/** Shift the current period of b forward by one. */
- static void
++STATIC void
+commit_max(bw_array_t *b)
+{
+ /* Store total from current period. */
+ b->totals[b->next_max_idx] = b->total_in_period;
+ /* Store maximum from current period. */
+ b->maxima[b->next_max_idx++] = b->max_total;
+ /* Advance next_period and next_max_idx */
+ b->next_period += NUM_SECS_BW_SUM_INTERVAL;
+ if (b->next_max_idx == NUM_TOTALS)
+ b->next_max_idx = 0;
+ if (b->num_maxes_set < NUM_TOTALS)
+ ++b->num_maxes_set;
+ /* Reset max_total. */
+ b->max_total = 0;
+ /* Reset total_in_period. */
+ b->total_in_period = 0;
+}
+
+/** Shift the current observation time of <b>b</b> forward by one second. */
- static inline void
++STATIC void
+advance_obs(bw_array_t *b)
+{
+ int nextidx;
+ uint64_t total;
+
+ /* Calculate the total bandwidth for the last NUM_SECS_ROLLING_MEASURE
+ * seconds; adjust max_total as needed.*/
+ total = b->total_obs + b->obs[b->cur_obs_idx];
+ if (total > b->max_total)
+ b->max_total = total;
+
+ nextidx = b->cur_obs_idx+1;
+ if (nextidx == NUM_SECS_ROLLING_MEASURE)
+ nextidx = 0;
+
+ b->total_obs = total - b->obs[nextidx];
+ b->obs[nextidx]=0;
+ b->cur_obs_idx = nextidx;
+
+ if (++b->cur_obs_time >= b->next_period)
+ commit_max(b);
+}
+
+/** Add <b>n</b> bytes to the number of bytes in <b>b</b> for second
+ * <b>when</b>. */
+static inline void
+add_obs(bw_array_t *b, time_t when, uint64_t n)
+{
+ if (when < b->cur_obs_time)
+ return; /* Don't record data in the past. */
+
+ /* If we're currently adding observations for an earlier second than
+ * 'when', advance b->cur_obs_time and b->cur_obs_idx by an
+ * appropriate number of seconds, and do all the other housekeeping. */
+ while (when > b->cur_obs_time) {
+ /* Doing this one second at a time is potentially inefficient, if we start
+ with a state file that is very old. Fortunately, it doesn't seem to
+ show up in profiles, so we can just ignore it for now. */
+ advance_obs(b);
+ }
+
+ b->obs[b->cur_obs_idx] += n;
+ b->total_in_period += n;
+}
+
+/** Allocate, initialize, and return a new bw_array. */
+static bw_array_t *
+bw_array_new(void)
+{
+ bw_array_t *b;
+ time_t start;
+ b = tor_malloc_zero(sizeof(bw_array_t));
+ rephist_total_alloc += sizeof(bw_array_t);
+ start = time(NULL);
+ b->cur_obs_time = start;
+ b->next_period = start + NUM_SECS_BW_SUM_INTERVAL;
+ return b;
+}
+
+#define bw_array_free(val) \
+ FREE_AND_NULL(bw_array_t, bw_array_free_, (val))
+
+/** Free storage held by bandwidth array <b>b</b>. */
+static void
+bw_array_free_(bw_array_t *b)
+{
+ if (!b) {
+ return;
+ }
+
+ rephist_total_alloc -= sizeof(bw_array_t);
+ tor_free(b);
+}
+
+/** Recent history of bandwidth observations for read operations. */
+static bw_array_t *read_array = NULL;
+/** Recent history of bandwidth observations for write operations. */
- static bw_array_t *write_array = NULL;
++STATIC bw_array_t *write_array = NULL;
+/** Recent history of bandwidth observations for read operations for the
+ directory protocol. */
+static bw_array_t *dir_read_array = NULL;
+/** Recent history of bandwidth observations for write operations for the
+ directory protocol. */
+static bw_array_t *dir_write_array = NULL;
+
+/** Set up [dir-]read_array and [dir-]write_array, freeing them if they
+ * already exist. */
+static void
+bw_arrays_init(void)
+{
+ bw_array_free(read_array);
+ bw_array_free(write_array);
+ bw_array_free(dir_read_array);
+ bw_array_free(dir_write_array);
+
+ read_array = bw_array_new();
+ write_array = bw_array_new();
+ dir_read_array = bw_array_new();
+ dir_write_array = bw_array_new();
+}
+
+/** Remember that we read <b>num_bytes</b> bytes in second <b>when</b>.
+ *
+ * Add num_bytes to the current running total for <b>when</b>.
+ *
+ * <b>when</b> can go back to time, but it's safe to ignore calls
+ * earlier than the latest <b>when</b> you've heard of.
+ */
+void
- rep_hist_note_bytes_written(size_t num_bytes, time_t when)
++rep_hist_note_bytes_written(uint64_t num_bytes, time_t when)
+{
+/* Maybe a circular array for recent seconds, and step to a new point
+ * every time a new second shows up. Or simpler is to just to have
+ * a normal array and push down each item every second; it's short.
+ */
+/* When a new second has rolled over, compute the sum of the bytes we've
+ * seen over when-1 to when-1-NUM_SECS_ROLLING_MEASURE, and stick it
+ * somewhere. See rep_hist_bandwidth_assess() below.
+ */
+ add_obs(write_array, when, num_bytes);
+}
+
+/** Remember that we wrote <b>num_bytes</b> bytes in second <b>when</b>.
+ * (like rep_hist_note_bytes_written() above)
+ */
+void
- rep_hist_note_bytes_read(size_t num_bytes, time_t when)
++rep_hist_note_bytes_read(uint64_t num_bytes, time_t when)
+{
+/* if we're smart, we can make this func and the one above share code */
+ add_obs(read_array, when, num_bytes);
+}
+
+/** Remember that we wrote <b>num_bytes</b> directory bytes in second
+ * <b>when</b>. (like rep_hist_note_bytes_written() above)
+ */
+void
- rep_hist_note_dir_bytes_written(size_t num_bytes, time_t when)
++rep_hist_note_dir_bytes_written(uint64_t num_bytes, time_t when)
+{
+ add_obs(dir_write_array, when, num_bytes);
+}
+
+/** Remember that we read <b>num_bytes</b> directory bytes in second
+ * <b>when</b>. (like rep_hist_note_bytes_written() above)
+ */
+void
- rep_hist_note_dir_bytes_read(size_t num_bytes, time_t when)
++rep_hist_note_dir_bytes_read(uint64_t num_bytes, time_t when)
+{
+ add_obs(dir_read_array, when, num_bytes);
+}
+
+/** Helper: Return the largest value in b->maxima. (This is equal to the
+ * most bandwidth used in any NUM_SECS_ROLLING_MEASURE period for the last
+ * NUM_SECS_BW_SUM_IS_VALID seconds.)
+ */
- static uint64_t
++STATIC uint64_t
+find_largest_max(bw_array_t *b)
+{
+ int i;
+ uint64_t max;
+ max=0;
+ for (i=0; i<NUM_TOTALS; ++i) {
+ if (b->maxima[i]>max)
+ max = b->maxima[i];
+ }
+ return max;
+}
+
+/** Find the largest sums in the past NUM_SECS_BW_SUM_IS_VALID (roughly)
+ * seconds. Find one sum for reading and one for writing. They don't have
+ * to be at the same time.
+ *
+ * Return the smaller of these sums, divided by NUM_SECS_ROLLING_MEASURE.
+ */
+MOCK_IMPL(int,
+rep_hist_bandwidth_assess,(void))
+{
+ uint64_t w,r;
+ r = find_largest_max(read_array);
+ w = find_largest_max(write_array);
+ if (r>w)
+ return (int)(((double)w)/NUM_SECS_ROLLING_MEASURE);
+ else
+ return (int)(((double)r)/NUM_SECS_ROLLING_MEASURE);
+}
+
+/** Print the bandwidth history of b (either [dir-]read_array or
+ * [dir-]write_array) into the buffer pointed to by buf. The format is
+ * simply comma separated numbers, from oldest to newest.
+ *
+ * It returns the number of bytes written.
+ */
+static size_t
+rep_hist_fill_bandwidth_history(char *buf, size_t len, const bw_array_t *b)
+{
+ char *cp = buf;
+ int i, n;
+ const or_options_t *options = get_options();
+ uint64_t cutoff;
+
+ if (b->num_maxes_set <= b->next_max_idx) {
+ /* We haven't been through the circular array yet; time starts at i=0.*/
+ i = 0;
+ } else {
+ /* We've been around the array at least once. The next i to be
+ overwritten is the oldest. */
+ i = b->next_max_idx;
+ }
+
+ if (options->RelayBandwidthRate) {
+ /* We don't want to report that we used more bandwidth than the max we're
+ * willing to relay; otherwise everybody will know how much traffic
+ * we used ourself. */
+ cutoff = options->RelayBandwidthRate * NUM_SECS_BW_SUM_INTERVAL;
+ } else {
+ cutoff = UINT64_MAX;
+ }
+
+ for (n=0; n<b->num_maxes_set; ++n,++i) {
+ uint64_t total;
+ if (i >= NUM_TOTALS)
+ i -= NUM_TOTALS;
+ tor_assert(i < NUM_TOTALS);
+ /* Round the bandwidth used down to the nearest 1k. */
+ total = b->totals[i] & ~0x3ff;
+ if (total > cutoff)
+ total = cutoff;
+
+ if (n==(b->num_maxes_set-1))
+ tor_snprintf(cp, len-(cp-buf), "%"PRIu64, (total));
+ else
+ tor_snprintf(cp, len-(cp-buf), "%"PRIu64",", (total));
+ cp += strlen(cp);
+ }
+ return cp-buf;
+}
+
+/** Allocate and return lines for representing this server's bandwidth
+ * history in its descriptor. We publish these lines in our extra-info
+ * descriptor.
+ */
+char *
+rep_hist_get_bandwidth_lines(void)
+{
+ char *buf, *cp;
+ char t[ISO_TIME_LEN+1];
+ int r;
+ bw_array_t *b = NULL;
+ const char *desc = NULL;
+ size_t len;
+
+ /* [dirreq-](read|write)-history yyyy-mm-dd HH:MM:SS (n s) n,n,n... */
+/* The n,n,n part above. Largest representation of a uint64_t is 20 chars
+ * long, plus the comma. */
+#define MAX_HIST_VALUE_LEN (21*NUM_TOTALS)
+ len = (67+MAX_HIST_VALUE_LEN)*4;
+ buf = tor_malloc_zero(len);
+ cp = buf;
+ for (r=0;r<4;++r) {
+ char tmp[MAX_HIST_VALUE_LEN];
+ size_t slen;
+ switch (r) {
+ case 0:
+ b = write_array;
+ desc = "write-history";
+ break;
+ case 1:
+ b = read_array;
+ desc = "read-history";
+ break;
+ case 2:
+ b = dir_write_array;
+ desc = "dirreq-write-history";
+ break;
+ case 3:
+ b = dir_read_array;
+ desc = "dirreq-read-history";
+ break;
+ }
+ tor_assert(b);
+ slen = rep_hist_fill_bandwidth_history(tmp, MAX_HIST_VALUE_LEN, b);
+ /* If we don't have anything to write, skip to the next entry. */
+ if (slen == 0)
+ continue;
+ format_iso_time(t, b->next_period-NUM_SECS_BW_SUM_INTERVAL);
+ tor_snprintf(cp, len-(cp-buf), "%s %s (%d s) ",
+ desc, t, NUM_SECS_BW_SUM_INTERVAL);
+ cp += strlen(cp);
+ strlcat(cp, tmp, len-(cp-buf));
+ cp += slen;
+ strlcat(cp, "\n", len-(cp-buf));
+ ++cp;
+ }
+ return buf;
+}
+
+/** Write a single bw_array_t into the Values, Ends, Interval, and Maximum
+ * entries of an or_state_t. Done before writing out a new state file. */
+static void
+rep_hist_update_bwhist_state_section(or_state_t *state,
+ const bw_array_t *b,
+ smartlist_t **s_values,
+ smartlist_t **s_maxima,
+ time_t *s_begins,
+ int *s_interval)
+{
+ int i,j;
+ uint64_t maxval;
+
+ if (*s_values) {
+ SMARTLIST_FOREACH(*s_values, char *, val, tor_free(val));
+ smartlist_free(*s_values);
+ }
+ if (*s_maxima) {
+ SMARTLIST_FOREACH(*s_maxima, char *, val, tor_free(val));
+ smartlist_free(*s_maxima);
+ }
+ if (! server_mode(get_options())) {
+ /* Clients don't need to store bandwidth history persistently;
+ * force these values to the defaults. */
+ /* FFFF we should pull the default out of config.c's state table,
+ * so we don't have two defaults. */
+ if (*s_begins != 0 || *s_interval != 900) {
+ time_t now = time(NULL);
+ time_t save_at = get_options()->AvoidDiskWrites ? now+3600 : now+600;
+ or_state_mark_dirty(state, save_at);
+ }
+ *s_begins = 0;
+ *s_interval = 900;
+ *s_values = smartlist_new();
+ *s_maxima = smartlist_new();
+ return;
+ }
+ *s_begins = b->next_period;
+ *s_interval = NUM_SECS_BW_SUM_INTERVAL;
+
+ *s_values = smartlist_new();
+ *s_maxima = smartlist_new();
+ /* Set i to first position in circular array */
+ i = (b->num_maxes_set <= b->next_max_idx) ? 0 : b->next_max_idx;
+ for (j=0; j < b->num_maxes_set; ++j,++i) {
+ if (i >= NUM_TOTALS)
+ i = 0;
+ smartlist_add_asprintf(*s_values, "%"PRIu64,
+ (b->totals[i] & ~0x3ff));
+ maxval = b->maxima[i] / NUM_SECS_ROLLING_MEASURE;
+ smartlist_add_asprintf(*s_maxima, "%"PRIu64,
+ (maxval & ~0x3ff));
+ }
+ smartlist_add_asprintf(*s_values, "%"PRIu64,
+ (b->total_in_period & ~0x3ff));
+ maxval = b->max_total / NUM_SECS_ROLLING_MEASURE;
+ smartlist_add_asprintf(*s_maxima, "%"PRIu64,
+ (maxval & ~0x3ff));
+}
+
+/** Update <b>state</b> with the newest bandwidth history. Done before
+ * writing out a new state file. */
+void
+rep_hist_update_state(or_state_t *state)
+{
+#define UPDATE(arrname,st) \
+ rep_hist_update_bwhist_state_section(state,\
+ (arrname),\
+ &state->BWHistory ## st ## Values, \
+ &state->BWHistory ## st ## Maxima, \
+ &state->BWHistory ## st ## Ends, \
+ &state->BWHistory ## st ## Interval)
+
+ UPDATE(write_array, Write);
+ UPDATE(read_array, Read);
+ UPDATE(dir_write_array, DirWrite);
+ UPDATE(dir_read_array, DirRead);
+
+ if (server_mode(get_options())) {
+ or_state_mark_dirty(state, time(NULL)+(2*3600));
+ }
+#undef UPDATE
+}
+
+/** Load a single bw_array_t from its Values, Ends, Maxima, and Interval
+ * entries in an or_state_t. Done while reading the state file. */
+static int
+rep_hist_load_bwhist_state_section(bw_array_t *b,
+ const smartlist_t *s_values,
+ const smartlist_t *s_maxima,
+ const time_t s_begins,
+ const int s_interval)
+{
+ time_t now = time(NULL);
+ int retval = 0;
+ time_t start;
+
+ uint64_t v, mv;
+ int i,ok,ok_m = 0;
+ int have_maxima = s_maxima && s_values &&
+ (smartlist_len(s_values) == smartlist_len(s_maxima));
+
+ if (s_values && s_begins >= now - NUM_SECS_BW_SUM_INTERVAL*NUM_TOTALS) {
+ start = s_begins - s_interval*(smartlist_len(s_values));
+ if (start > now)
+ return 0;
+ b->cur_obs_time = start;
+ b->next_period = start + NUM_SECS_BW_SUM_INTERVAL;
+ SMARTLIST_FOREACH_BEGIN(s_values, const char *, cp) {
+ const char *maxstr = NULL;
+ v = tor_parse_uint64(cp, 10, 0, UINT64_MAX, &ok, NULL);
+ if (have_maxima) {
+ maxstr = smartlist_get(s_maxima, cp_sl_idx);
+ mv = tor_parse_uint64(maxstr, 10, 0, UINT64_MAX, &ok_m, NULL);
+ mv *= NUM_SECS_ROLLING_MEASURE;
+ } else {
+ /* No maxima known; guess average rate to be conservative. */
+ mv = (v / s_interval) * NUM_SECS_ROLLING_MEASURE;
+ }
+ if (!ok) {
+ retval = -1;
+ log_notice(LD_HIST, "Could not parse value '%s' into a number.'",cp);
+ }
+ if (maxstr && !ok_m) {
+ retval = -1;
+ log_notice(LD_HIST, "Could not parse maximum '%s' into a number.'",
+ maxstr);
+ }
+
+ if (start < now) {
+ time_t cur_start = start;
+ time_t actual_interval_len = s_interval;
+ uint64_t cur_val = 0;
+ /* Calculate the average per second. This is the best we can do
+ * because our state file doesn't have per-second resolution. */
+ if (start + s_interval > now)
+ actual_interval_len = now - start;
+ cur_val = v / actual_interval_len;
+ /* This is potentially inefficient, but since we don't do it very
+ * often it should be ok. */
+ while (cur_start < start + actual_interval_len) {
+ add_obs(b, cur_start, cur_val);
+ ++cur_start;
+ }
+ b->max_total = mv;
+ /* This will result in some fairly choppy history if s_interval
+ * is not the same as NUM_SECS_BW_SUM_INTERVAL. XXXX */
+ start += actual_interval_len;
+ }
+ } SMARTLIST_FOREACH_END(cp);
+ }
+
+ /* Clean up maxima and observed */
+ for (i=0; i<NUM_SECS_ROLLING_MEASURE; ++i) {
+ b->obs[i] = 0;
+ }
+ b->total_obs = 0;
+
+ return retval;
+}
+
+/** Set bandwidth history from the state file we just loaded. */
+int
+rep_hist_load_state(or_state_t *state, char **err)
+{
+ int all_ok = 1;
+
+ /* Assert they already have been malloced */
+ tor_assert(read_array && write_array);
+ tor_assert(dir_read_array && dir_write_array);
+
+#define LOAD(arrname,st) \
+ if (rep_hist_load_bwhist_state_section( \
+ (arrname), \
+ state->BWHistory ## st ## Values, \
+ state->BWHistory ## st ## Maxima, \
+ state->BWHistory ## st ## Ends, \
+ state->BWHistory ## st ## Interval)<0) \
+ all_ok = 0
+
+ LOAD(write_array, Write);
+ LOAD(read_array, Read);
+ LOAD(dir_write_array, DirWrite);
+ LOAD(dir_read_array, DirRead);
+
+#undef LOAD
+ if (!all_ok) {
+ *err = tor_strdup("Parsing of bandwidth history values failed");
+ /* and create fresh arrays */
+ bw_arrays_init();
+ return -1;
+ }
+ return 0;
+}
+
+/*********************************************************************/
+
+/** A single predicted port: used to remember which ports we've made
+ * connections to, so that we can try to keep making circuits that can handle
+ * those ports. */
+typedef struct predicted_port_t {
+ /** The port we connected to */
+ uint16_t port;
+ /** The time at which we last used it */
+ time_t time;
+} predicted_port_t;
+
+/** A list of port numbers that have been used recently. */
+static smartlist_t *predicted_ports_list=NULL;
+/** How long do we keep predicting circuits? */
+static time_t prediction_timeout=0;
+/** When was the last time we added a prediction entry (HS or port) */
+static time_t last_prediction_add_time=0;
+
+/**
+ * How much time left until we stop predicting circuits?
+ */
+int
+predicted_ports_prediction_time_remaining(time_t now)
+{
+ time_t seconds_waited;
+ time_t seconds_left;
+
+ /* Protect against overflow of return value. This can happen if the clock
+ * jumps backwards in time. Update the last prediction time (aka last
+ * active time) to prevent it. This update is preferable to using monotonic
+ * time because it prevents clock jumps into the past from simply causing
+ * very long idle timeouts while the monotonic time stands still. */
+ seconds_waited = time_diff(last_prediction_add_time, now);
+ if (seconds_waited == TIME_MAX) {
+ last_prediction_add_time = now;
+ seconds_waited = 0;
+ }
+
+ /* Protect against underflow of the return value. This can happen for very
+ * large periods of inactivity/system sleep. */
+ if (seconds_waited > prediction_timeout)
+ return 0;
+
+ seconds_left = time_diff(seconds_waited, prediction_timeout);
+ if (BUG(seconds_left == TIME_MAX))
+ return INT_MAX;
+
+ return (int)(seconds_left);
+}
+
+/** We just got an application request for a connection with
+ * port <b>port</b>. Remember it for the future, so we can keep
+ * some circuits open that will exit to this port.
+ */
+static void
+add_predicted_port(time_t now, uint16_t port)
+{
+ predicted_port_t *pp = tor_malloc(sizeof(predicted_port_t));
+
+ // If the list is empty, re-randomize predicted ports lifetime
+ if (!any_predicted_circuits(now)) {
+ prediction_timeout =
+ (time_t)channelpadding_get_circuits_available_timeout();
+ }
+
+ last_prediction_add_time = now;
+
+ log_info(LD_CIRC,
+ "New port prediction added. Will continue predictive circ building "
+ "for %d more seconds.",
+ predicted_ports_prediction_time_remaining(now));
+
+ pp->port = port;
+ pp->time = now;
+ rephist_total_alloc += sizeof(*pp);
+ smartlist_add(predicted_ports_list, pp);
+}
+
+/**
+ * Allocate whatever memory and structs are needed for predicting
+ * which ports will be used. Also seed it with port 80, so we'll build
+ * circuits on start-up.
+ */
+static void
+predicted_ports_alloc(void)
+{
+ predicted_ports_list = smartlist_new();
+}
+
+void
+predicted_ports_init(void)
+{
+ add_predicted_port(time(NULL), 443); // Add a port to get us started
+}
+
+/** Free whatever memory is needed for predicting which ports will
+ * be used.
+ */
+static void
+predicted_ports_free_all(void)
+{
+ rephist_total_alloc -=
+ smartlist_len(predicted_ports_list)*sizeof(predicted_port_t);
+ SMARTLIST_FOREACH(predicted_ports_list, predicted_port_t *,
+ pp, tor_free(pp));
+ smartlist_free(predicted_ports_list);
+}
+
+/** Remember that <b>port</b> has been asked for as of time <b>now</b>.
+ * This is used for predicting what sorts of streams we'll make in the
+ * future and making exit circuits to anticipate that.
+ */
+void
+rep_hist_note_used_port(time_t now, uint16_t port)
+{
+ tor_assert(predicted_ports_list);
+
+ if (!port) /* record nothing */
+ return;
+
+ SMARTLIST_FOREACH_BEGIN(predicted_ports_list, predicted_port_t *, pp) {
+ if (pp->port == port) {
+ pp->time = now;
+
+ last_prediction_add_time = now;
+ log_info(LD_CIRC,
+ "New port prediction added. Will continue predictive circ "
+ "building for %d more seconds.",
+ predicted_ports_prediction_time_remaining(now));
+ return;
+ }
+ } SMARTLIST_FOREACH_END(pp);
+ /* it's not there yet; we need to add it */
+ add_predicted_port(now, port);
+}
+
+/** Return a newly allocated pointer to a list of uint16_t * for ports that
+ * are likely to be asked for in the near future.
+ */
+smartlist_t *
+rep_hist_get_predicted_ports(time_t now)
+{
+ int predicted_circs_relevance_time;
+ smartlist_t *out = smartlist_new();
+ tor_assert(predicted_ports_list);
+
+ predicted_circs_relevance_time = (int)prediction_timeout;
+
+ /* clean out obsolete entries */
+ SMARTLIST_FOREACH_BEGIN(predicted_ports_list, predicted_port_t *, pp) {
+ if (pp->time + predicted_circs_relevance_time < now) {
+ log_debug(LD_CIRC, "Expiring predicted port %d", pp->port);
+
+ rephist_total_alloc -= sizeof(predicted_port_t);
+ tor_free(pp);
+ SMARTLIST_DEL_CURRENT(predicted_ports_list, pp);
+ } else {
+ smartlist_add(out, tor_memdup(&pp->port, sizeof(uint16_t)));
+ }
+ } SMARTLIST_FOREACH_END(pp);
+ return out;
+}
+
+/**
+ * Take a list of uint16_t *, and remove every port in the list from the
+ * current list of predicted ports.
+ */
+void
+rep_hist_remove_predicted_ports(const smartlist_t *rmv_ports)
+{
+ /* Let's do this on O(N), not O(N^2). */
+ bitarray_t *remove_ports = bitarray_init_zero(UINT16_MAX);
+ SMARTLIST_FOREACH(rmv_ports, const uint16_t *, p,
+ bitarray_set(remove_ports, *p));
+ SMARTLIST_FOREACH_BEGIN(predicted_ports_list, predicted_port_t *, pp) {
+ if (bitarray_is_set(remove_ports, pp->port)) {
+ tor_free(pp);
+ rephist_total_alloc -= sizeof(*pp);
+ SMARTLIST_DEL_CURRENT(predicted_ports_list, pp);
+ }
+ } SMARTLIST_FOREACH_END(pp);
+ bitarray_free(remove_ports);
+}
+
+/** The user asked us to do a resolve. Rather than keeping track of
+ * timings and such of resolves, we fake it for now by treating
+ * it the same way as a connection to port 80. This way we will continue
+ * to have circuits lying around if the user only uses Tor for resolves.
+ */
+void
+rep_hist_note_used_resolve(time_t now)
+{
+ rep_hist_note_used_port(now, 80);
+}
+
+/** The last time at which we needed an internal circ. */
+static time_t predicted_internal_time = 0;
+/** The last time we needed an internal circ with good uptime. */
+static time_t predicted_internal_uptime_time = 0;
+/** The last time we needed an internal circ with good capacity. */
+static time_t predicted_internal_capacity_time = 0;
+
+/** Remember that we used an internal circ at time <b>now</b>. */
+void
+rep_hist_note_used_internal(time_t now, int need_uptime, int need_capacity)
+{
+ // If the list is empty, re-randomize predicted ports lifetime
+ if (!any_predicted_circuits(now)) {
+ prediction_timeout = channelpadding_get_circuits_available_timeout();
+ }
+
+ last_prediction_add_time = now;
+
+ log_info(LD_CIRC,
+ "New port prediction added. Will continue predictive circ building "
+ "for %d more seconds.",
+ predicted_ports_prediction_time_remaining(now));
+
+ predicted_internal_time = now;
+ if (need_uptime)
+ predicted_internal_uptime_time = now;
+ if (need_capacity)
+ predicted_internal_capacity_time = now;
+}
+
+/** Return 1 if we've used an internal circ recently; else return 0. */
+int
+rep_hist_get_predicted_internal(time_t now, int *need_uptime,
+ int *need_capacity)
+{
+ int predicted_circs_relevance_time;
+
+ predicted_circs_relevance_time = (int)prediction_timeout;
+
+ if (!predicted_internal_time) { /* initialize it */
+ predicted_internal_time = now;
+ predicted_internal_uptime_time = now;
+ predicted_internal_capacity_time = now;
+ }
+ if (predicted_internal_time + predicted_circs_relevance_time < now)
+ return 0; /* too long ago */
+ if (predicted_internal_uptime_time + predicted_circs_relevance_time >= now)
+ *need_uptime = 1;
+ // Always predict that we need capacity.
+ *need_capacity = 1;
+ return 1;
+}
+
+/** Any ports used lately? These are pre-seeded if we just started
+ * up or if we're running a hidden service. */
+int
+any_predicted_circuits(time_t now)
+{
+ int predicted_circs_relevance_time;
+ predicted_circs_relevance_time = (int)prediction_timeout;
+
+ return smartlist_len(predicted_ports_list) ||
+ predicted_internal_time + predicted_circs_relevance_time >= now;
+}
+
+/** Return 1 if we have no need for circuits currently, else return 0. */
+int
+rep_hist_circbuilding_dormant(time_t now)
+{
+ const or_options_t *options = get_options();
+
+ if (any_predicted_circuits(now))
+ return 0;
+
+ /* see if we'll still need to build testing circuits */
+ if (server_mode(options) &&
+ (!check_whether_orport_reachable(options) ||
+ !circuit_enough_testing_circs()))
+ return 0;
+ if (!check_whether_dirport_reachable(options))
+ return 0;
+
+ return 1;
+}
+
+/*** Exit port statistics ***/
+
+/* Some constants */
+/** To what multiple should byte numbers be rounded up? */
+#define EXIT_STATS_ROUND_UP_BYTES 1024
+/** To what multiple should stream counts be rounded up? */
+#define EXIT_STATS_ROUND_UP_STREAMS 4
+/** Number of TCP ports */
+#define EXIT_STATS_NUM_PORTS 65536
+/** Top n ports that will be included in exit stats. */
+#define EXIT_STATS_TOP_N_PORTS 10
+
+/* The following data structures are arrays and no fancy smartlists or maps,
+ * so that all write operations can be done in constant time. This comes at
+ * the price of some memory (1.25 MB) and linear complexity when writing
+ * stats for measuring relays. */
+/** Number of bytes read in current period by exit port */
+static uint64_t *exit_bytes_read = NULL;
+/** Number of bytes written in current period by exit port */
+static uint64_t *exit_bytes_written = NULL;
+/** Number of streams opened in current period by exit port */
+static uint32_t *exit_streams = NULL;
+
+/** Start time of exit stats or 0 if we're not collecting exit stats. */
+static time_t start_of_exit_stats_interval;
+
+/** Initialize exit port stats. */
+void
+rep_hist_exit_stats_init(time_t now)
+{
+ start_of_exit_stats_interval = now;
+ exit_bytes_read = tor_calloc(EXIT_STATS_NUM_PORTS, sizeof(uint64_t));
+ exit_bytes_written = tor_calloc(EXIT_STATS_NUM_PORTS, sizeof(uint64_t));
+ exit_streams = tor_calloc(EXIT_STATS_NUM_PORTS, sizeof(uint32_t));
+}
+
+/** Reset counters for exit port statistics. */
+void
+rep_hist_reset_exit_stats(time_t now)
+{
+ start_of_exit_stats_interval = now;
+ memset(exit_bytes_read, 0, EXIT_STATS_NUM_PORTS * sizeof(uint64_t));
+ memset(exit_bytes_written, 0, EXIT_STATS_NUM_PORTS * sizeof(uint64_t));
+ memset(exit_streams, 0, EXIT_STATS_NUM_PORTS * sizeof(uint32_t));
+}
+
+/** Stop collecting exit port stats in a way that we can re-start doing
+ * so in rep_hist_exit_stats_init(). */
+void
+rep_hist_exit_stats_term(void)
+{
+ start_of_exit_stats_interval = 0;
+ tor_free(exit_bytes_read);
+ tor_free(exit_bytes_written);
+ tor_free(exit_streams);
+}
+
+/** Helper for qsort: compare two ints. Does not handle overflow properly,
+ * but works fine for sorting an array of port numbers, which is what we use
+ * it for. */
+static int
+compare_int_(const void *x, const void *y)
+{
+ return (*(int*)x - *(int*)y);
+}
+
+/** Return a newly allocated string containing the exit port statistics
+ * until <b>now</b>, or NULL if we're not collecting exit stats. Caller
+ * must ensure start_of_exit_stats_interval is in the past. */
+char *
+rep_hist_format_exit_stats(time_t now)
+{
+ int i, j, top_elements = 0, cur_min_idx = 0, cur_port;
+ uint64_t top_bytes[EXIT_STATS_TOP_N_PORTS];
+ int top_ports[EXIT_STATS_TOP_N_PORTS];
+ uint64_t cur_bytes = 0, other_read = 0, other_written = 0,
+ total_read = 0, total_written = 0;
+ uint32_t total_streams = 0, other_streams = 0;
+ smartlist_t *written_strings, *read_strings, *streams_strings;
+ char *written_string, *read_string, *streams_string;
+ char t[ISO_TIME_LEN+1];
+ char *result;
+
+ if (!start_of_exit_stats_interval)
+ return NULL; /* Not initialized. */
+
+ tor_assert(now >= start_of_exit_stats_interval);
+
+ /* Go through all ports to find the n ports that saw most written and
+ * read bytes.
+ *
+ * Invariant: at the end of the loop for iteration i,
+ * total_read is the sum of all exit_bytes_read[0..i]
+ * total_written is the sum of all exit_bytes_written[0..i]
+ * total_stream is the sum of all exit_streams[0..i]
+ *
+ * top_elements = MAX(EXIT_STATS_TOP_N_PORTS,
+ * #{j | 0 <= j <= i && volume(i) > 0})
+ *
+ * For all 0 <= j < top_elements,
+ * top_bytes[j] > 0
+ * 0 <= top_ports[j] <= 65535
+ * top_bytes[j] = volume(top_ports[j])
+ *
+ * There is no j in 0..i and k in 0..top_elements such that:
+ * volume(j) > top_bytes[k] AND j is not in top_ports[0..top_elements]
+ *
+ * There is no j!=cur_min_idx in 0..top_elements such that:
+ * top_bytes[j] < top_bytes[cur_min_idx]
+ *
+ * where volume(x) == exit_bytes_read[x]+exit_bytes_written[x]
+ *
+ * Worst case: O(EXIT_STATS_NUM_PORTS * EXIT_STATS_TOP_N_PORTS)
+ */
+ for (i = 1; i < EXIT_STATS_NUM_PORTS; i++) {
+ total_read += exit_bytes_read[i];
+ total_written += exit_bytes_written[i];
+ total_streams += exit_streams[i];
+ cur_bytes = exit_bytes_read[i] + exit_bytes_written[i];
+ if (cur_bytes == 0) {
+ continue;
+ }
+ if (top_elements < EXIT_STATS_TOP_N_PORTS) {
+ top_bytes[top_elements] = cur_bytes;
+ top_ports[top_elements++] = i;
+ } else if (cur_bytes > top_bytes[cur_min_idx]) {
+ top_bytes[cur_min_idx] = cur_bytes;
+ top_ports[cur_min_idx] = i;
+ } else {
+ continue;
+ }
+ cur_min_idx = 0;
+ for (j = 1; j < top_elements; j++) {
+ if (top_bytes[j] < top_bytes[cur_min_idx]) {
+ cur_min_idx = j;
+ }
+ }
+ }
+
+ /* Add observations of top ports to smartlists. */
+ written_strings = smartlist_new();
+ read_strings = smartlist_new();
+ streams_strings = smartlist_new();
+ other_read = total_read;
+ other_written = total_written;
+ other_streams = total_streams;
+ /* Sort the ports; this puts them out of sync with top_bytes, but we
+ * won't be using top_bytes again anyway */
+ qsort(top_ports, top_elements, sizeof(int), compare_int_);
+ for (j = 0; j < top_elements; j++) {
+ cur_port = top_ports[j];
+ if (exit_bytes_written[cur_port] > 0) {
+ uint64_t num = round_uint64_to_next_multiple_of(
+ exit_bytes_written[cur_port],
+ EXIT_STATS_ROUND_UP_BYTES);
+ num /= 1024;
+ smartlist_add_asprintf(written_strings, "%d=%"PRIu64,
+ cur_port, (num));
+ other_written -= exit_bytes_written[cur_port];
+ }
+ if (exit_bytes_read[cur_port] > 0) {
+ uint64_t num = round_uint64_to_next_multiple_of(
+ exit_bytes_read[cur_port],
+ EXIT_STATS_ROUND_UP_BYTES);
+ num /= 1024;
+ smartlist_add_asprintf(read_strings, "%d=%"PRIu64,
+ cur_port, (num));
+ other_read -= exit_bytes_read[cur_port];
+ }
+ if (exit_streams[cur_port] > 0) {
+ uint32_t num = round_uint32_to_next_multiple_of(
+ exit_streams[cur_port],
+ EXIT_STATS_ROUND_UP_STREAMS);
+ smartlist_add_asprintf(streams_strings, "%d=%u", cur_port, num);
+ other_streams -= exit_streams[cur_port];
+ }
+ }
+
+ /* Add observations of other ports in a single element. */
+ other_written = round_uint64_to_next_multiple_of(other_written,
+ EXIT_STATS_ROUND_UP_BYTES);
+ other_written /= 1024;
+ smartlist_add_asprintf(written_strings, "other=%"PRIu64,
+ (other_written));
+ other_read = round_uint64_to_next_multiple_of(other_read,
+ EXIT_STATS_ROUND_UP_BYTES);
+ other_read /= 1024;
+ smartlist_add_asprintf(read_strings, "other=%"PRIu64,
+ (other_read));
+ other_streams = round_uint32_to_next_multiple_of(other_streams,
+ EXIT_STATS_ROUND_UP_STREAMS);
+ smartlist_add_asprintf(streams_strings, "other=%u", other_streams);
+
+ /* Join all observations in single strings. */
+ written_string = smartlist_join_strings(written_strings, ",", 0, NULL);
+ read_string = smartlist_join_strings(read_strings, ",", 0, NULL);
+ streams_string = smartlist_join_strings(streams_strings, ",", 0, NULL);
+ SMARTLIST_FOREACH(written_strings, char *, cp, tor_free(cp));
+ SMARTLIST_FOREACH(read_strings, char *, cp, tor_free(cp));
+ SMARTLIST_FOREACH(streams_strings, char *, cp, tor_free(cp));
+ smartlist_free(written_strings);
+ smartlist_free(read_strings);
+ smartlist_free(streams_strings);
+
+ /* Put everything together. */
+ format_iso_time(t, now);
+ tor_asprintf(&result, "exit-stats-end %s (%d s)\n"
+ "exit-kibibytes-written %s\n"
+ "exit-kibibytes-read %s\n"
+ "exit-streams-opened %s\n",
+ t, (unsigned) (now - start_of_exit_stats_interval),
+ written_string,
+ read_string,
+ streams_string);
+ tor_free(written_string);
+ tor_free(read_string);
+ tor_free(streams_string);
+ return result;
+}
+
+/** If 24 hours have passed since the beginning of the current exit port
+ * stats period, write exit stats to $DATADIR/stats/exit-stats (possibly
+ * overwriting an existing file) and reset counters. Return when we would
+ * next want to write exit stats or 0 if we never want to write. */
+time_t
+rep_hist_exit_stats_write(time_t now)
+{
+ char *str = NULL;
+
+ if (!start_of_exit_stats_interval)
+ return 0; /* Not initialized. */
+ if (start_of_exit_stats_interval + WRITE_STATS_INTERVAL > now)
+ goto done; /* Not ready to write. */
+
+ log_info(LD_HIST, "Writing exit port statistics to disk.");
+
+ /* Generate history string. */
+ str = rep_hist_format_exit_stats(now);
+
+ /* Reset counters. */
+ rep_hist_reset_exit_stats(now);
+
+ /* Try to write to disk. */
+ if (!check_or_create_data_subdir("stats")) {
+ write_to_data_subdir("stats", "exit-stats", str, "exit port statistics");
+ }
+
+ done:
+ tor_free(str);
+ return start_of_exit_stats_interval + WRITE_STATS_INTERVAL;
+}
+
+/** Note that we wrote <b>num_written</b> bytes and read <b>num_read</b>
+ * bytes to/from an exit connection to <b>port</b>. */
+void
+rep_hist_note_exit_bytes(uint16_t port, size_t num_written,
+ size_t num_read)
+{
+ if (!start_of_exit_stats_interval)
+ return; /* Not initialized. */
+ exit_bytes_written[port] += num_written;
+ exit_bytes_read[port] += num_read;
+ log_debug(LD_HIST, "Written %lu bytes and read %lu bytes to/from an "
+ "exit connection to port %d.",
+ (unsigned long)num_written, (unsigned long)num_read, port);
+}
+
+/** Note that we opened an exit stream to <b>port</b>. */
+void
+rep_hist_note_exit_stream_opened(uint16_t port)
+{
+ if (!start_of_exit_stats_interval)
+ return; /* Not initialized. */
+ exit_streams[port]++;
+ log_debug(LD_HIST, "Opened exit stream to port %d", port);
+}
+
+/*** cell statistics ***/
+
+/** Start of the current buffer stats interval or 0 if we're not
+ * collecting buffer statistics. */
+static time_t start_of_buffer_stats_interval;
+
+/** Initialize buffer stats. */
+void
+rep_hist_buffer_stats_init(time_t now)
+{
+ start_of_buffer_stats_interval = now;
+}
+
+/** Statistics from a single circuit. Collected when the circuit closes, or
+ * when we flush statistics to disk. */
+typedef struct circ_buffer_stats_t {
+ /** Average number of cells in the circuit's queue */
+ double mean_num_cells_in_queue;
+ /** Average time a cell waits in the queue. */
+ double mean_time_cells_in_queue;
+ /** Total number of cells sent over this circuit */
+ uint32_t processed_cells;
+} circ_buffer_stats_t;
+
+/** List of circ_buffer_stats_t. */
+static smartlist_t *circuits_for_buffer_stats = NULL;
+
+/** Remember cell statistics <b>mean_num_cells_in_queue</b>,
+ * <b>mean_time_cells_in_queue</b>, and <b>processed_cells</b> of a
+ * circuit. */
+void
+rep_hist_add_buffer_stats(double mean_num_cells_in_queue,
+ double mean_time_cells_in_queue, uint32_t processed_cells)
+{
+ circ_buffer_stats_t *stats;
+ if (!start_of_buffer_stats_interval)
+ return; /* Not initialized. */
+ stats = tor_malloc_zero(sizeof(circ_buffer_stats_t));
+ stats->mean_num_cells_in_queue = mean_num_cells_in_queue;
+ stats->mean_time_cells_in_queue = mean_time_cells_in_queue;
+ stats->processed_cells = processed_cells;
+ if (!circuits_for_buffer_stats)
+ circuits_for_buffer_stats = smartlist_new();
+ smartlist_add(circuits_for_buffer_stats, stats);
+}
+
+/** Remember cell statistics for circuit <b>circ</b> at time
+ * <b>end_of_interval</b> and reset cell counters in case the circuit
+ * remains open in the next measurement interval. */
+void
+rep_hist_buffer_stats_add_circ(circuit_t *circ, time_t end_of_interval)
+{
+ time_t start_of_interval;
+ int interval_length;
+ or_circuit_t *orcirc;
+ double mean_num_cells_in_queue, mean_time_cells_in_queue;
+ uint32_t processed_cells;
+ if (CIRCUIT_IS_ORIGIN(circ))
+ return;
+ orcirc = TO_OR_CIRCUIT(circ);
+ if (!orcirc->processed_cells)
+ return;
+ start_of_interval = (circ->timestamp_created.tv_sec >
+ start_of_buffer_stats_interval) ?
+ (time_t)circ->timestamp_created.tv_sec :
+ start_of_buffer_stats_interval;
+ interval_length = (int) (end_of_interval - start_of_interval);
+ if (interval_length <= 0)
+ return;
+ processed_cells = orcirc->processed_cells;
+ /* 1000.0 for s -> ms; 2.0 because of app-ward and exit-ward queues */
+ mean_num_cells_in_queue = (double) orcirc->total_cell_waiting_time /
+ (double) interval_length / 1000.0 / 2.0;
+ mean_time_cells_in_queue =
+ (double) orcirc->total_cell_waiting_time /
+ (double) orcirc->processed_cells;
+ orcirc->total_cell_waiting_time = 0;
+ orcirc->processed_cells = 0;
+ rep_hist_add_buffer_stats(mean_num_cells_in_queue,
+ mean_time_cells_in_queue,
+ processed_cells);
+}
+
+/** Sorting helper: return -1, 1, or 0 based on comparison of two
+ * circ_buffer_stats_t */
+static int
+buffer_stats_compare_entries_(const void **_a, const void **_b)
+{
+ const circ_buffer_stats_t *a = *_a, *b = *_b;
+ if (a->processed_cells < b->processed_cells)
+ return 1;
+ else if (a->processed_cells > b->processed_cells)
+ return -1;
+ else
+ return 0;
+}
+
+/** Stop collecting cell stats in a way that we can re-start doing so in
+ * rep_hist_buffer_stats_init(). */
+void
+rep_hist_buffer_stats_term(void)
+{
+ rep_hist_reset_buffer_stats(0);
+}
+
+/** Clear history of circuit statistics and set the measurement interval
+ * start to <b>now</b>. */
+void
+rep_hist_reset_buffer_stats(time_t now)
+{
+ if (!circuits_for_buffer_stats)
+ circuits_for_buffer_stats = smartlist_new();
+ SMARTLIST_FOREACH(circuits_for_buffer_stats, circ_buffer_stats_t *,
+ stats, tor_free(stats));
+ smartlist_clear(circuits_for_buffer_stats);
+ start_of_buffer_stats_interval = now;
+}
+
+/** Return a newly allocated string containing the buffer statistics until
+ * <b>now</b>, or NULL if we're not collecting buffer stats. Caller must
+ * ensure start_of_buffer_stats_interval is in the past. */
+char *
+rep_hist_format_buffer_stats(time_t now)
+{
+#define SHARES 10
+ uint64_t processed_cells[SHARES];
+ uint32_t circs_in_share[SHARES];
+ int number_of_circuits, i;
+ double queued_cells[SHARES], time_in_queue[SHARES];
+ smartlist_t *processed_cells_strings, *queued_cells_strings,
+ *time_in_queue_strings;
+ char *processed_cells_string, *queued_cells_string,
+ *time_in_queue_string;
+ char t[ISO_TIME_LEN+1];
+ char *result;
+
+ if (!start_of_buffer_stats_interval)
+ return NULL; /* Not initialized. */
+
+ tor_assert(now >= start_of_buffer_stats_interval);
+
+ /* Calculate deciles if we saw at least one circuit. */
+ memset(processed_cells, 0, SHARES * sizeof(uint64_t));
+ memset(circs_in_share, 0, SHARES * sizeof(uint32_t));
+ memset(queued_cells, 0, SHARES * sizeof(double));
+ memset(time_in_queue, 0, SHARES * sizeof(double));
+ if (!circuits_for_buffer_stats)
+ circuits_for_buffer_stats = smartlist_new();
+ number_of_circuits = smartlist_len(circuits_for_buffer_stats);
+ if (number_of_circuits > 0) {
+ smartlist_sort(circuits_for_buffer_stats,
+ buffer_stats_compare_entries_);
+ i = 0;
+ SMARTLIST_FOREACH_BEGIN(circuits_for_buffer_stats,
+ circ_buffer_stats_t *, stats)
+ {
+ int share = i++ * SHARES / number_of_circuits;
+ processed_cells[share] += stats->processed_cells;
+ queued_cells[share] += stats->mean_num_cells_in_queue;
+ time_in_queue[share] += stats->mean_time_cells_in_queue;
+ circs_in_share[share]++;
+ }
+ SMARTLIST_FOREACH_END(stats);
+ }
+
+ /* Write deciles to strings. */
+ processed_cells_strings = smartlist_new();
+ queued_cells_strings = smartlist_new();
+ time_in_queue_strings = smartlist_new();
+ for (i = 0; i < SHARES; i++) {
+ smartlist_add_asprintf(processed_cells_strings,
+ "%"PRIu64, !circs_in_share[i] ? 0 :
+ (processed_cells[i] /
+ circs_in_share[i]));
+ }
+ for (i = 0; i < SHARES; i++) {
+ smartlist_add_asprintf(queued_cells_strings, "%.2f",
+ circs_in_share[i] == 0 ? 0.0 :
+ queued_cells[i] / (double) circs_in_share[i]);
+ }
+ for (i = 0; i < SHARES; i++) {
+ smartlist_add_asprintf(time_in_queue_strings, "%.0f",
+ circs_in_share[i] == 0 ? 0.0 :
+ time_in_queue[i] / (double) circs_in_share[i]);
+ }
+
+ /* Join all observations in single strings. */
+ processed_cells_string = smartlist_join_strings(processed_cells_strings,
+ ",", 0, NULL);
+ queued_cells_string = smartlist_join_strings(queued_cells_strings,
+ ",", 0, NULL);
+ time_in_queue_string = smartlist_join_strings(time_in_queue_strings,
+ ",", 0, NULL);
+ SMARTLIST_FOREACH(processed_cells_strings, char *, cp, tor_free(cp));
+ SMARTLIST_FOREACH(queued_cells_strings, char *, cp, tor_free(cp));
+ SMARTLIST_FOREACH(time_in_queue_strings, char *, cp, tor_free(cp));
+ smartlist_free(processed_cells_strings);
+ smartlist_free(queued_cells_strings);
+ smartlist_free(time_in_queue_strings);
+
+ /* Put everything together. */
+ format_iso_time(t, now);
+ tor_asprintf(&result, "cell-stats-end %s (%d s)\n"
+ "cell-processed-cells %s\n"
+ "cell-queued-cells %s\n"
+ "cell-time-in-queue %s\n"
+ "cell-circuits-per-decile %d\n",
+ t, (unsigned) (now - start_of_buffer_stats_interval),
+ processed_cells_string,
+ queued_cells_string,
+ time_in_queue_string,
+ CEIL_DIV(number_of_circuits, SHARES));
+ tor_free(processed_cells_string);
+ tor_free(queued_cells_string);
+ tor_free(time_in_queue_string);
+ return result;
+#undef SHARES
+}
+
+/** If 24 hours have passed since the beginning of the current buffer
+ * stats period, write buffer stats to $DATADIR/stats/buffer-stats
+ * (possibly overwriting an existing file) and reset counters. Return
+ * when we would next want to write buffer stats or 0 if we never want to
+ * write. */
+time_t
+rep_hist_buffer_stats_write(time_t now)
+{
+ char *str = NULL;
+
+ if (!start_of_buffer_stats_interval)
+ return 0; /* Not initialized. */
+ if (start_of_buffer_stats_interval + WRITE_STATS_INTERVAL > now)
+ goto done; /* Not ready to write */
+
+ /* Add open circuits to the history. */
+ SMARTLIST_FOREACH_BEGIN(circuit_get_global_list(), circuit_t *, circ) {
+ rep_hist_buffer_stats_add_circ(circ, now);
+ }
+ SMARTLIST_FOREACH_END(circ);
+
+ /* Generate history string. */
+ str = rep_hist_format_buffer_stats(now);
+
+ /* Reset both buffer history and counters of open circuits. */
+ rep_hist_reset_buffer_stats(now);
+
+ /* Try to write to disk. */
+ if (!check_or_create_data_subdir("stats")) {
+ write_to_data_subdir("stats", "buffer-stats", str, "buffer statistics");
+ }
+
+ done:
+ tor_free(str);
+ return start_of_buffer_stats_interval + WRITE_STATS_INTERVAL;
+}
+
+/*** Descriptor serving statistics ***/
+
+/** Digestmap to track which descriptors were downloaded this stats
+ * collection interval. It maps descriptor digest to pointers to 1,
+ * effectively turning this into a list. */
+static digestmap_t *served_descs = NULL;
+
+/** Number of how many descriptors were downloaded in total during this
+ * interval. */
+static unsigned long total_descriptor_downloads;
+
+/** Start time of served descs stats or 0 if we're not collecting those. */
+static time_t start_of_served_descs_stats_interval;
+
+/** Initialize descriptor stats. */
+void
+rep_hist_desc_stats_init(time_t now)
+{
+ if (served_descs) {
+ log_warn(LD_BUG, "Called rep_hist_desc_stats_init() when desc stats were "
+ "already initialized. This is probably harmless.");
+ return; // Already initialized
+ }
+ served_descs = digestmap_new();
+ total_descriptor_downloads = 0;
+ start_of_served_descs_stats_interval = now;
+}
+
+/** Reset served descs stats to empty, starting a new interval <b>now</b>. */
+static void
+rep_hist_reset_desc_stats(time_t now)
+{
+ rep_hist_desc_stats_term();
+ rep_hist_desc_stats_init(now);
+}
+
+/** Stop collecting served descs stats, so that rep_hist_desc_stats_init() is
+ * safe to be called again. */
+void
+rep_hist_desc_stats_term(void)
+{
+ digestmap_free(served_descs, NULL);
+ served_descs = NULL;
+ start_of_served_descs_stats_interval = 0;
+ total_descriptor_downloads = 0;
+}
+
+/** Helper for rep_hist_desc_stats_write(). Return a newly allocated string
+ * containing the served desc statistics until now, or NULL if we're not
+ * collecting served desc stats. Caller must ensure that now is not before
+ * start_of_served_descs_stats_interval. */
+static char *
+rep_hist_format_desc_stats(time_t now)
+{
+ char t[ISO_TIME_LEN+1];
+ char *result;
+
+ digestmap_iter_t *iter;
+ const char *key;
+ void *val;
+ unsigned size;
+ int *vals, max = 0, q3 = 0, md = 0, q1 = 0, min = 0;
+ int n = 0;
+
+ if (!start_of_served_descs_stats_interval)
+ return NULL;
+
+ size = digestmap_size(served_descs);
+ if (size > 0) {
+ vals = tor_calloc(size, sizeof(int));
+ for (iter = digestmap_iter_init(served_descs);
+ !digestmap_iter_done(iter);
+ iter = digestmap_iter_next(served_descs, iter)) {
+ uintptr_t count;
+ digestmap_iter_get(iter, &key, &val);
+ count = (uintptr_t)val;
+ vals[n++] = (int)count;
+ (void)key;
+ }
+ max = find_nth_int(vals, size, size-1);
+ q3 = find_nth_int(vals, size, (3*size-1)/4);
+ md = find_nth_int(vals, size, (size-1)/2);
+ q1 = find_nth_int(vals, size, (size-1)/4);
+ min = find_nth_int(vals, size, 0);
+ tor_free(vals);
+ }
+
+ format_iso_time(t, now);
+
+ tor_asprintf(&result,
+ "served-descs-stats-end %s (%d s) total=%lu unique=%u "
+ "max=%d q3=%d md=%d q1=%d min=%d\n",
+ t,
+ (unsigned) (now - start_of_served_descs_stats_interval),
+ total_descriptor_downloads,
+ size, max, q3, md, q1, min);
+
+ return result;
+}
+
+/** If WRITE_STATS_INTERVAL seconds have passed since the beginning of
+ * the current served desc stats interval, write the stats to
+ * $DATADIR/stats/served-desc-stats (possibly appending to an existing file)
+ * and reset the state for the next interval. Return when we would next want
+ * to write served desc stats or 0 if we won't want to write. */
+time_t
+rep_hist_desc_stats_write(time_t now)
+{
+ char *filename = NULL, *str = NULL;
+
+ if (!start_of_served_descs_stats_interval)
+ return 0; /* We're not collecting stats. */
+ if (start_of_served_descs_stats_interval + WRITE_STATS_INTERVAL > now)
+ return start_of_served_descs_stats_interval + WRITE_STATS_INTERVAL;
+
+ str = rep_hist_format_desc_stats(now);
+ tor_assert(str != NULL);
+
+ if (check_or_create_data_subdir("stats") < 0) {
+ goto done;
+ }
+ filename = get_datadir_fname2("stats", "served-desc-stats");
+ if (append_bytes_to_file(filename, str, strlen(str), 0) < 0)
+ log_warn(LD_HIST, "Unable to write served descs statistics to disk!");
+
+ rep_hist_reset_desc_stats(now);
+
+ done:
+ tor_free(filename);
+ tor_free(str);
+ return start_of_served_descs_stats_interval + WRITE_STATS_INTERVAL;
+}
+
+/** Called to note that we've served a given descriptor (by
+ * digest). Increments the count of descriptors served, and the number
+ * of times we've served this descriptor. */
+void
+rep_hist_note_desc_served(const char * desc)
+{
+ void *val;
+ uintptr_t count;
+ if (!served_descs)
+ return; // We're not collecting stats
+ val = digestmap_get(served_descs, desc);
+ count = (uintptr_t)val;
+ if (count != INT_MAX)
+ ++count;
+ digestmap_set(served_descs, desc, (void*)count);
+ total_descriptor_downloads++;
+}
+
+/*** Connection statistics ***/
+
+/** Start of the current connection stats interval or 0 if we're not
+ * collecting connection statistics. */
+static time_t start_of_conn_stats_interval;
+
+/** Initialize connection stats. */
+void
+rep_hist_conn_stats_init(time_t now)
+{
+ start_of_conn_stats_interval = now;
+}
+
+/* Count connections that we read and wrote less than these many bytes
+ * from/to as below threshold. */
+#define BIDI_THRESHOLD 20480
+
+/* Count connections that we read or wrote at least this factor as many
+ * bytes from/to than we wrote or read to/from as mostly reading or
+ * writing. */
+#define BIDI_FACTOR 10
+
+/* Interval length in seconds for considering read and written bytes for
+ * connection stats. */
+#define BIDI_INTERVAL 10
+
+/** Start of next BIDI_INTERVAL second interval. */
+static time_t bidi_next_interval = 0;
+
+/** Number of connections that we read and wrote less than BIDI_THRESHOLD
+ * bytes from/to in BIDI_INTERVAL seconds. */
+static uint32_t below_threshold = 0;
+
+/** Number of connections that we read at least BIDI_FACTOR times more
+ * bytes from than we wrote to in BIDI_INTERVAL seconds. */
+static uint32_t mostly_read = 0;
+
+/** Number of connections that we wrote at least BIDI_FACTOR times more
+ * bytes to than we read from in BIDI_INTERVAL seconds. */
+static uint32_t mostly_written = 0;
+
+/** Number of connections that we read and wrote at least BIDI_THRESHOLD
+ * bytes from/to, but not BIDI_FACTOR times more in either direction in
+ * BIDI_INTERVAL seconds. */
+static uint32_t both_read_and_written = 0;
+
+/** Entry in a map from connection ID to the number of read and written
+ * bytes on this connection in a BIDI_INTERVAL second interval. */
+typedef struct bidi_map_entry_t {
+ HT_ENTRY(bidi_map_entry_t) node;
+ uint64_t conn_id; /**< Connection ID */
+ size_t read; /**< Number of read bytes */
+ size_t written; /**< Number of written bytes */
+} bidi_map_entry_t;
+
+/** Map of OR connections together with the number of read and written
+ * bytes in the current BIDI_INTERVAL second interval. */
+static HT_HEAD(bidimap, bidi_map_entry_t) bidi_map =
+ HT_INITIALIZER();
+
+static int
+bidi_map_ent_eq(const bidi_map_entry_t *a, const bidi_map_entry_t *b)
+{
+ return a->conn_id == b->conn_id;
+}
+
+/* DOCDOC bidi_map_ent_hash */
+static unsigned
+bidi_map_ent_hash(const bidi_map_entry_t *entry)
+{
+ return (unsigned) entry->conn_id;
+}
+
+HT_PROTOTYPE(bidimap, bidi_map_entry_t, node, bidi_map_ent_hash,
+ bidi_map_ent_eq)
+HT_GENERATE2(bidimap, bidi_map_entry_t, node, bidi_map_ent_hash,
+ bidi_map_ent_eq, 0.6, tor_reallocarray_, tor_free_)
+
+/* DOCDOC bidi_map_free */
+static void
+bidi_map_free_all(void)
+{
+ bidi_map_entry_t **ptr, **next, *ent;
+ for (ptr = HT_START(bidimap, &bidi_map); ptr; ptr = next) {
+ ent = *ptr;
+ next = HT_NEXT_RMV(bidimap, &bidi_map, ptr);
+ tor_free(ent);
+ }
+ HT_CLEAR(bidimap, &bidi_map);
+}
+
+/** Reset counters for conn statistics. */
+void
+rep_hist_reset_conn_stats(time_t now)
+{
+ start_of_conn_stats_interval = now;
+ below_threshold = 0;
+ mostly_read = 0;
+ mostly_written = 0;
+ both_read_and_written = 0;
+ bidi_map_free_all();
+}
+
+/** Stop collecting connection stats in a way that we can re-start doing
+ * so in rep_hist_conn_stats_init(). */
+void
+rep_hist_conn_stats_term(void)
+{
+ rep_hist_reset_conn_stats(0);
+}
+
+/** We read <b>num_read</b> bytes and wrote <b>num_written</b> from/to OR
+ * connection <b>conn_id</b> in second <b>when</b>. If this is the first
+ * observation in a new interval, sum up the last observations. Add bytes
+ * for this connection. */
+void
+rep_hist_note_or_conn_bytes(uint64_t conn_id, size_t num_read,
+ size_t num_written, time_t when)
+{
+ if (!start_of_conn_stats_interval)
+ return;
+ /* Initialize */
+ if (bidi_next_interval == 0)
+ bidi_next_interval = when + BIDI_INTERVAL;
+ /* Sum up last period's statistics */
+ if (when >= bidi_next_interval) {
+ bidi_map_entry_t **ptr, **next, *ent;
+ for (ptr = HT_START(bidimap, &bidi_map); ptr; ptr = next) {
+ ent = *ptr;
+ if (ent->read + ent->written < BIDI_THRESHOLD)
+ below_threshold++;
+ else if (ent->read >= ent->written * BIDI_FACTOR)
+ mostly_read++;
+ else if (ent->written >= ent->read * BIDI_FACTOR)
+ mostly_written++;
+ else
+ both_read_and_written++;
+ next = HT_NEXT_RMV(bidimap, &bidi_map, ptr);
+ tor_free(ent);
+ }
+ while (when >= bidi_next_interval)
+ bidi_next_interval += BIDI_INTERVAL;
+ log_info(LD_GENERAL, "%d below threshold, %d mostly read, "
+ "%d mostly written, %d both read and written.",
+ below_threshold, mostly_read, mostly_written,
+ both_read_and_written);
+ }
+ /* Add this connection's bytes. */
+ if (num_read > 0 || num_written > 0) {
+ bidi_map_entry_t *entry, lookup;
+ lookup.conn_id = conn_id;
+ entry = HT_FIND(bidimap, &bidi_map, &lookup);
+ if (entry) {
+ entry->written += num_written;
+ entry->read += num_read;
+ } else {
+ entry = tor_malloc_zero(sizeof(bidi_map_entry_t));
+ entry->conn_id = conn_id;
+ entry->written = num_written;
+ entry->read = num_read;
+ HT_INSERT(bidimap, &bidi_map, entry);
+ }
+ }
+}
+
+/** Return a newly allocated string containing the connection statistics
+ * until <b>now</b>, or NULL if we're not collecting conn stats. Caller must
+ * ensure start_of_conn_stats_interval is in the past. */
+char *
+rep_hist_format_conn_stats(time_t now)
+{
+ char *result, written[ISO_TIME_LEN+1];
+
+ if (!start_of_conn_stats_interval)
+ return NULL; /* Not initialized. */
+
+ tor_assert(now >= start_of_conn_stats_interval);
+
+ format_iso_time(written, now);
+ tor_asprintf(&result, "conn-bi-direct %s (%d s) %d,%d,%d,%d\n",
+ written,
+ (unsigned) (now - start_of_conn_stats_interval),
+ below_threshold,
+ mostly_read,
+ mostly_written,
+ both_read_and_written);
+ return result;
+}
+
+/** If 24 hours have passed since the beginning of the current conn stats
+ * period, write conn stats to $DATADIR/stats/conn-stats (possibly
+ * overwriting an existing file) and reset counters. Return when we would
+ * next want to write conn stats or 0 if we never want to write. */
+time_t
+rep_hist_conn_stats_write(time_t now)
+{
+ char *str = NULL;
+
+ if (!start_of_conn_stats_interval)
+ return 0; /* Not initialized. */
+ if (start_of_conn_stats_interval + WRITE_STATS_INTERVAL > now)
+ goto done; /* Not ready to write */
+
+ /* Generate history string. */
+ str = rep_hist_format_conn_stats(now);
+
+ /* Reset counters. */
+ rep_hist_reset_conn_stats(now);
+
+ /* Try to write to disk. */
+ if (!check_or_create_data_subdir("stats")) {
+ write_to_data_subdir("stats", "conn-stats", str, "connection statistics");
+ }
+
+ done:
+ tor_free(str);
+ return start_of_conn_stats_interval + WRITE_STATS_INTERVAL;
+}
+
+/** 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.
+ * @{ */
+STATIC int onion_handshakes_requested[MAX_ONION_HANDSHAKE_TYPE+1] = {0};
+STATIC int onion_handshakes_assigned[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)
+{
+ if (type <= MAX_ONION_HANDSHAKE_TYPE)
+ onion_handshakes_requested[type]++;
+}
+
+/** We've sent an onionskin (using the <b>type</b> handshake) to a
+ * cpuworker. */
+void
+rep_hist_note_circuit_handshake_assigned(uint16_t type)
+{
+ if (type <= MAX_ONION_HANDSHAKE_TYPE)
+ onion_handshakes_assigned[type]++;
+}
+
+/** Log our onionskin statistics since the last time we were called. */
+void
+rep_hist_log_circuit_handshake_stats(time_t now)
+{
+ (void)now;
+ log_notice(LD_HEARTBEAT, "Circuit handshake stats since last time: "
+ "%d/%d TAP, %d/%d NTor.",
+ onion_handshakes_assigned[ONION_HANDSHAKE_TYPE_TAP],
+ onion_handshakes_requested[ONION_HANDSHAKE_TYPE_TAP],
+ onion_handshakes_assigned[ONION_HANDSHAKE_TYPE_NTOR],
+ onion_handshakes_requested[ONION_HANDSHAKE_TYPE_NTOR]);
+ memset(onion_handshakes_assigned, 0, sizeof(onion_handshakes_assigned));
+ memset(onion_handshakes_requested, 0, sizeof(onion_handshakes_requested));
+}
+
+/* Hidden service statistics section */
+
+/** Start of the current hidden service stats interval or 0 if we're
+ * not collecting hidden service statistics. */
+static time_t start_of_hs_stats_interval;
+
+/** Carries the various hidden service statistics, and any other
+ * information needed. */
+typedef struct hs_stats_t {
+ /** How many relay cells have we seen as rendezvous points? */
+ uint64_t rp_relay_cells_seen;
+
+ /** Set of unique public key digests we've seen this stat period
+ * (could also be implemented as sorted smartlist). */
+ digestmap_t *onions_seen_this_period;
+} hs_stats_t;
+
+/** Our statistics structure singleton. */
+static hs_stats_t *hs_stats = NULL;
+
+/** Allocate, initialize and return an hs_stats_t structure. */
+static hs_stats_t *
+hs_stats_new(void)
+{
+ hs_stats_t *new_hs_stats = tor_malloc_zero(sizeof(hs_stats_t));
+ new_hs_stats->onions_seen_this_period = digestmap_new();
+
+ return new_hs_stats;
+}
+
+#define hs_stats_free(val) \
+ FREE_AND_NULL(hs_stats_t, hs_stats_free_, (val))
+
+/** Free an hs_stats_t structure. */
+static void
+hs_stats_free_(hs_stats_t *victim_hs_stats)
+{
+ if (!victim_hs_stats) {
+ return;
+ }
+
+ digestmap_free(victim_hs_stats->onions_seen_this_period, NULL);
+ tor_free(victim_hs_stats);
+}
+
+/** Initialize hidden service statistics. */
+void
+rep_hist_hs_stats_init(time_t now)
+{
+ if (!hs_stats) {
+ hs_stats = hs_stats_new();
+ }
+
+ start_of_hs_stats_interval = now;
+}
+
+/** Clear history of hidden service statistics and set the measurement
+ * interval start to <b>now</b>. */
+static void
+rep_hist_reset_hs_stats(time_t now)
+{
+ if (!hs_stats) {
+ hs_stats = hs_stats_new();
+ }
+
+ hs_stats->rp_relay_cells_seen = 0;
+
+ digestmap_free(hs_stats->onions_seen_this_period, NULL);
+ hs_stats->onions_seen_this_period = digestmap_new();
+
+ start_of_hs_stats_interval = now;
+}
+
+/** Stop collecting hidden service stats in a way that we can re-start
+ * doing so in rep_hist_buffer_stats_init(). */
+void
+rep_hist_hs_stats_term(void)
+{
+ rep_hist_reset_hs_stats(0);
+}
+
+/** We saw a new HS relay cell, Count it! */
+void
+rep_hist_seen_new_rp_cell(void)
+{
+ if (!hs_stats) {
+ return; // We're not collecting stats
+ }
+
+ hs_stats->rp_relay_cells_seen++;
+}
+
+/** As HSDirs, we saw another hidden service with public key
+ * <b>pubkey</b>. Check whether we have counted it before, if not
+ * count it now! */
+void
+rep_hist_stored_maybe_new_hs(const crypto_pk_t *pubkey)
+{
+ char pubkey_hash[DIGEST_LEN];
+
+ if (!hs_stats) {
+ return; // We're not collecting stats
+ }
+
+ /* Get the digest of the pubkey which will be used to detect whether
+ we've seen this hidden service before or not. */
+ if (crypto_pk_get_digest(pubkey, pubkey_hash) < 0) {
+ /* This fail should not happen; key has been validated by
+ descriptor parsing code first. */
+ return;
+ }
+
+ /* Check if this is the first time we've seen this hidden
+ service. If it is, count it as new. */
+ if (!digestmap_get(hs_stats->onions_seen_this_period,
+ pubkey_hash)) {
+ digestmap_set(hs_stats->onions_seen_this_period,
+ pubkey_hash, (void*)(uintptr_t)1);
+ }
+}
+
+/* The number of cells that are supposed to be hidden from the adversary
+ * by adding noise from the Laplace distribution. This value, divided by
+ * EPSILON, is Laplace parameter b. It must be greather than 0. */
+#define REND_CELLS_DELTA_F 2048
+/* Security parameter for obfuscating number of cells with a value between
+ * ]0.0, 1.0]. Smaller values obfuscate observations more, but at the same
+ * time make statistics less usable. */
+#define REND_CELLS_EPSILON 0.3
+/* The number of cells that are supposed to be hidden from the adversary
+ * by rounding up to the next multiple of this number. */
+#define REND_CELLS_BIN_SIZE 1024
+/* The number of service identities that are supposed to be hidden from the
+ * adversary by adding noise from the Laplace distribution. This value,
+ * divided by EPSILON, is Laplace parameter b. It must be greater than 0. */
+#define ONIONS_SEEN_DELTA_F 8
+/* Security parameter for obfuscating number of service identities with a
+ * value between ]0.0, 1.0]. Smaller values obfuscate observations more, but
+ * at the same time make statistics less usable. */
+#define ONIONS_SEEN_EPSILON 0.3
+/* The number of service identities that are supposed to be hidden from
+ * the adversary by rounding up to the next multiple of this number. */
+#define ONIONS_SEEN_BIN_SIZE 8
+
+/** Allocate and return a string containing hidden service stats that
+ * are meant to be placed in the extra-info descriptor. */
+static char *
+rep_hist_format_hs_stats(time_t now)
+{
+ char t[ISO_TIME_LEN+1];
+ char *hs_stats_string;
+ int64_t obfuscated_cells_seen;
+ int64_t obfuscated_onions_seen;
+
+ uint64_t rounded_cells_seen
+ = round_uint64_to_next_multiple_of(hs_stats->rp_relay_cells_seen,
+ REND_CELLS_BIN_SIZE);
+ rounded_cells_seen = MIN(rounded_cells_seen, INT64_MAX);
+ obfuscated_cells_seen = add_laplace_noise((int64_t)rounded_cells_seen,
+ crypto_rand_double(),
+ REND_CELLS_DELTA_F, REND_CELLS_EPSILON);
+
+ uint64_t rounded_onions_seen =
+ round_uint64_to_next_multiple_of((size_t)digestmap_size(
+ hs_stats->onions_seen_this_period),
+ ONIONS_SEEN_BIN_SIZE);
+ rounded_onions_seen = MIN(rounded_onions_seen, INT64_MAX);
+ obfuscated_onions_seen = add_laplace_noise((int64_t)rounded_onions_seen,
+ crypto_rand_double(), ONIONS_SEEN_DELTA_F,
+ ONIONS_SEEN_EPSILON);
+
+ format_iso_time(t, now);
+ tor_asprintf(&hs_stats_string, "hidserv-stats-end %s (%d s)\n"
+ "hidserv-rend-relayed-cells %"PRId64" delta_f=%d "
+ "epsilon=%.2f bin_size=%d\n"
+ "hidserv-dir-onions-seen %"PRId64" delta_f=%d "
+ "epsilon=%.2f bin_size=%d\n",
+ t, (unsigned) (now - start_of_hs_stats_interval),
+ (obfuscated_cells_seen), REND_CELLS_DELTA_F,
+ REND_CELLS_EPSILON, REND_CELLS_BIN_SIZE,
+ (obfuscated_onions_seen),
+ ONIONS_SEEN_DELTA_F,
+ ONIONS_SEEN_EPSILON, ONIONS_SEEN_BIN_SIZE);
+
+ return hs_stats_string;
+}
+
+/** If 24 hours have passed since the beginning of the current HS
+ * stats period, write buffer stats to $DATADIR/stats/hidserv-stats
+ * (possibly overwriting an existing file) and reset counters. Return
+ * when we would next want to write buffer stats or 0 if we never want to
+ * write. */
+time_t
+rep_hist_hs_stats_write(time_t now)
+{
+ char *str = NULL;
+
+ if (!start_of_hs_stats_interval) {
+ return 0; /* Not initialized. */
+ }
+
+ if (start_of_hs_stats_interval + WRITE_STATS_INTERVAL > now) {
+ goto done; /* Not ready to write */
+ }
+
+ /* Generate history string. */
+ str = rep_hist_format_hs_stats(now);
+
+ /* Reset HS history. */
+ rep_hist_reset_hs_stats(now);
+
+ /* Try to write to disk. */
+ if (!check_or_create_data_subdir("stats")) {
+ write_to_data_subdir("stats", "hidserv-stats", str,
+ "hidden service stats");
+ }
+
+ done:
+ tor_free(str);
+ return start_of_hs_stats_interval + WRITE_STATS_INTERVAL;
+}
+
+static uint64_t link_proto_count[MAX_LINK_PROTO+1][2];
+
+/** Note that we negotiated link protocol version <b>link_proto</b>, on
+ * a connection that started here iff <b>started_here</b> is true.
+ */
+void
+rep_hist_note_negotiated_link_proto(unsigned link_proto, int started_here)
+{
+ started_here = !!started_here; /* force to 0 or 1 */
+ if (link_proto > MAX_LINK_PROTO) {
+ log_warn(LD_BUG, "Can't log link protocol %u", link_proto);
+ return;
+ }
+
+ link_proto_count[link_proto][started_here]++;
+}
+
+/**
+ * Update the maximum count of total pending channel padding timers
+ * in this period.
+ */
+void
+rep_hist_padding_count_timers(uint64_t num_timers)
+{
+ if (num_timers > padding_current.maximum_chanpad_timers) {
+ padding_current.maximum_chanpad_timers = num_timers;
+ }
+}
+
+/**
+ * Count a cell that we sent for padding overhead statistics.
+ *
+ * RELAY_COMMAND_DROP and CELL_PADDING are accounted separately. Both should be
+ * counted for PADDING_TYPE_TOTAL.
+ */
+void
+rep_hist_padding_count_write(padding_type_t type)
+{
+ switch (type) {
+ case PADDING_TYPE_DROP:
+ padding_current.write_drop_cell_count++;
+ break;
+ case PADDING_TYPE_CELL:
+ padding_current.write_pad_cell_count++;
+ break;
+ case PADDING_TYPE_TOTAL:
+ padding_current.write_cell_count++;
+ break;
+ case PADDING_TYPE_ENABLED_TOTAL:
+ padding_current.enabled_write_cell_count++;
+ break;
+ case PADDING_TYPE_ENABLED_CELL:
+ padding_current.enabled_write_pad_cell_count++;
+ break;
+ }
+}
+
+/**
+ * Count a cell that we've received for padding overhead statistics.
+ *
+ * RELAY_COMMAND_DROP and CELL_PADDING are accounted separately. Both should be
+ * counted for PADDING_TYPE_TOTAL.
+ */
+void
+rep_hist_padding_count_read(padding_type_t type)
+{
+ switch (type) {
+ case PADDING_TYPE_DROP:
+ padding_current.read_drop_cell_count++;
+ break;
+ case PADDING_TYPE_CELL:
+ padding_current.read_pad_cell_count++;
+ break;
+ case PADDING_TYPE_TOTAL:
+ padding_current.read_cell_count++;
+ break;
+ case PADDING_TYPE_ENABLED_TOTAL:
+ padding_current.enabled_read_cell_count++;
+ break;
+ case PADDING_TYPE_ENABLED_CELL:
+ padding_current.enabled_read_pad_cell_count++;
+ break;
+ }
+}
+
+/**
+ * Reset our current padding statistics. Called once every 24 hours.
+ */
+void
+rep_hist_reset_padding_counts(void)
+{
+ memset(&padding_current, 0, sizeof(padding_current));
+}
+
+/**
+ * Copy our current cell counts into a structure for listing in our
+ * extra-info descriptor. Also perform appropriate rounding and redaction.
+ *
+ * This function is called once every 24 hours.
+ */
+#define MIN_CELL_COUNTS_TO_PUBLISH 1
+#define ROUND_CELL_COUNTS_TO 10000
+void
+rep_hist_prep_published_padding_counts(time_t now)
+{
+ memcpy(&padding_published, &padding_current, sizeof(padding_published));
+
+ if (padding_published.read_cell_count < MIN_CELL_COUNTS_TO_PUBLISH ||
+ padding_published.write_cell_count < MIN_CELL_COUNTS_TO_PUBLISH) {
+ memset(&padding_published, 0, sizeof(padding_published));
+ return;
+ }
+
+ format_iso_time(padding_published.first_published_at, now);
+#define ROUND_AND_SET_COUNT(x) (x) = round_uint64_to_next_multiple_of((x), \
+ ROUND_CELL_COUNTS_TO)
+ ROUND_AND_SET_COUNT(padding_published.read_pad_cell_count);
+ ROUND_AND_SET_COUNT(padding_published.write_pad_cell_count);
+ ROUND_AND_SET_COUNT(padding_published.read_drop_cell_count);
+ ROUND_AND_SET_COUNT(padding_published.write_drop_cell_count);
+ ROUND_AND_SET_COUNT(padding_published.write_cell_count);
+ ROUND_AND_SET_COUNT(padding_published.read_cell_count);
+ ROUND_AND_SET_COUNT(padding_published.enabled_read_cell_count);
+ ROUND_AND_SET_COUNT(padding_published.enabled_read_pad_cell_count);
+ ROUND_AND_SET_COUNT(padding_published.enabled_write_cell_count);
+ ROUND_AND_SET_COUNT(padding_published.enabled_write_pad_cell_count);
+#undef ROUND_AND_SET_COUNT
+}
+
+/**
+ * Returns an allocated string for extra-info documents for publishing
+ * padding statistics from the last 24 hour interval.
+ */
+char *
+rep_hist_get_padding_count_lines(void)
+{
+ char *result = NULL;
+
+ if (!padding_published.read_cell_count ||
+ !padding_published.write_cell_count) {
+ return NULL;
+ }
+
+ tor_asprintf(&result, "padding-counts %s (%d s)"
+ " bin-size=%"PRIu64
+ " write-drop=%"PRIu64
+ " write-pad=%"PRIu64
+ " write-total=%"PRIu64
+ " read-drop=%"PRIu64
+ " read-pad=%"PRIu64
+ " read-total=%"PRIu64
+ " enabled-read-pad=%"PRIu64
+ " enabled-read-total=%"PRIu64
+ " enabled-write-pad=%"PRIu64
+ " enabled-write-total=%"PRIu64
+ " max-chanpad-timers=%"PRIu64
+ "\n",
+ padding_published.first_published_at,
+ REPHIST_CELL_PADDING_COUNTS_INTERVAL,
+ (uint64_t)ROUND_CELL_COUNTS_TO,
+ (padding_published.write_drop_cell_count),
+ (padding_published.write_pad_cell_count),
+ (padding_published.write_cell_count),
+ (padding_published.read_drop_cell_count),
+ (padding_published.read_pad_cell_count),
+ (padding_published.read_cell_count),
+ (padding_published.enabled_read_pad_cell_count),
+ (padding_published.enabled_read_cell_count),
+ (padding_published.enabled_write_pad_cell_count),
+ (padding_published.enabled_write_cell_count),
+ (padding_published.maximum_chanpad_timers)
+ );
+
+ return result;
+}
+
+/** Log a heartbeat message explaining how many connections of each link
+ * protocol version we have used.
+ */
+void
+rep_hist_log_link_protocol_counts(void)
+{
+ log_notice(LD_HEARTBEAT,
+ "Since startup, we have initiated "
+ "%"PRIu64" v1 connections, "
+ "%"PRIu64" v2 connections, "
+ "%"PRIu64" v3 connections, and "
+ "%"PRIu64" v4 connections; and received "
+ "%"PRIu64" v1 connections, "
+ "%"PRIu64" v2 connections, "
+ "%"PRIu64" v3 connections, and "
+ "%"PRIu64" v4 connections.",
+ (link_proto_count[1][1]),
+ (link_proto_count[2][1]),
+ (link_proto_count[3][1]),
+ (link_proto_count[4][1]),
+ (link_proto_count[1][0]),
+ (link_proto_count[2][0]),
+ (link_proto_count[3][0]),
+ (link_proto_count[4][0]));
+}
+
+/** Free all storage held by the OR/link history caches, by the
+ * bandwidth history arrays, by the port history, or by statistics . */
+void
+rep_hist_free_all(void)
+{
+ hs_stats_free(hs_stats);
+ digestmap_free(history_map, free_or_history);
+
+ bw_array_free(read_array);
+ read_array = NULL;
+
+ bw_array_free(write_array);
+ write_array = NULL;
+
+ bw_array_free(dir_read_array);
+ dir_read_array = NULL;
+
+ bw_array_free(dir_write_array);
+ dir_write_array = NULL;
+
+ tor_free(exit_bytes_read);
+ tor_free(exit_bytes_written);
+ tor_free(exit_streams);
+ predicted_ports_free_all();
+ bidi_map_free_all();
+
+ if (circuits_for_buffer_stats) {
+ SMARTLIST_FOREACH(circuits_for_buffer_stats, circ_buffer_stats_t *, s,
+ tor_free(s));
+ smartlist_free(circuits_for_buffer_stats);
+ circuits_for_buffer_stats = NULL;
+ }
+ rep_hist_desc_stats_term();
+ total_descriptor_downloads = 0;
+
+ tor_assert_nonfatal(rephist_total_alloc == 0);
+ tor_assert_nonfatal_once(rephist_total_num == 0);
+}
diff --cc src/feature/stats/rephist.h
index 67a015a4c,000000000..e17a72248
mode 100644,000000..100644
--- a/src/feature/stats/rephist.h
+++ b/src/feature/stats/rephist.h
@@@ -1,140 -1,0 +1,142 @@@
+/* Copyright (c) 2001 Matej Pfajfar.
+ * Copyright (c) 2001-2004, Roger Dingledine.
+ * Copyright (c) 2004-2006, Roger Dingledine, Nick Mathewson.
+ * Copyright (c) 2007-2018, The Tor Project, Inc. */
+/* See LICENSE for licensing information */
+
+/**
+ * \file rephist.h
+ * \brief Header file for rephist.c.
+ **/
+
+#ifndef TOR_REPHIST_H
+#define TOR_REPHIST_H
+
+void rep_hist_init(void);
+void rep_hist_dump_stats(time_t now, int severity);
- void rep_hist_note_bytes_read(size_t num_bytes, time_t when);
- void rep_hist_note_bytes_written(size_t num_bytes, time_t when);
++void rep_hist_note_bytes_read(uint64_t num_bytes, time_t when);
++void rep_hist_note_bytes_written(uint64_t num_bytes, time_t when);
+
+void rep_hist_make_router_pessimal(const char *id, time_t when);
+
- void rep_hist_note_dir_bytes_read(size_t num_bytes, time_t when);
- void rep_hist_note_dir_bytes_written(size_t num_bytes, time_t when);
++void rep_hist_note_dir_bytes_read(uint64_t num_bytes, time_t when);
++void rep_hist_note_dir_bytes_written(uint64_t num_bytes, time_t when);
+
+MOCK_DECL(int, rep_hist_bandwidth_assess, (void));
+char *rep_hist_get_bandwidth_lines(void);
+void rep_hist_update_state(or_state_t *state);
+int rep_hist_load_state(or_state_t *state, char **err);
+void rep_history_clean(time_t before);
+
+void rep_hist_note_router_reachable(const char *id, const tor_addr_t *at_addr,
+ const uint16_t at_port, time_t when);
+void rep_hist_note_router_unreachable(const char *id, time_t when);
+int rep_hist_record_mtbf_data(time_t now, int missing_means_down);
+int rep_hist_load_mtbf_data(time_t now);
+
+time_t rep_hist_downrate_old_runs(time_t now);
+long rep_hist_get_uptime(const char *id, time_t when);
+double rep_hist_get_stability(const char *id, time_t when);
+double rep_hist_get_weighted_fractional_uptime(const char *id, time_t when);
+long rep_hist_get_weighted_time_known(const char *id, time_t when);
+int rep_hist_have_measured_enough_stability(void);
+
+void predicted_ports_init(void);
+void rep_hist_note_used_port(time_t now, uint16_t port);
+smartlist_t *rep_hist_get_predicted_ports(time_t now);
+void rep_hist_remove_predicted_ports(const smartlist_t *rmv_ports);
+void rep_hist_note_used_resolve(time_t now);
+void rep_hist_note_used_internal(time_t now, int need_uptime,
+ int need_capacity);
+int rep_hist_get_predicted_internal(time_t now, int *need_uptime,
+ int *need_capacity);
+
+int any_predicted_circuits(time_t now);
+int rep_hist_circbuilding_dormant(time_t now);
+int predicted_ports_prediction_time_remaining(time_t now);
+
+void rep_hist_exit_stats_init(time_t now);
+void rep_hist_reset_exit_stats(time_t now);
+void rep_hist_exit_stats_term(void);
+char *rep_hist_format_exit_stats(time_t now);
+time_t rep_hist_exit_stats_write(time_t now);
+void rep_hist_note_exit_bytes(uint16_t port, size_t num_written,
+ size_t num_read);
+void rep_hist_note_exit_stream_opened(uint16_t port);
+
+void rep_hist_buffer_stats_init(time_t now);
+void rep_hist_buffer_stats_add_circ(circuit_t *circ,
+ time_t end_of_interval);
+time_t rep_hist_buffer_stats_write(time_t now);
+void rep_hist_buffer_stats_term(void);
+void rep_hist_add_buffer_stats(double mean_num_cells_in_queue,
+ double mean_time_cells_in_queue, uint32_t processed_cells);
+char *rep_hist_format_buffer_stats(time_t now);
+void rep_hist_reset_buffer_stats(time_t now);
+
+void rep_hist_desc_stats_init(time_t now);
+void rep_hist_note_desc_served(const char * desc);
+void rep_hist_desc_stats_term(void);
+time_t rep_hist_desc_stats_write(time_t now);
+
+void rep_hist_conn_stats_init(time_t now);
+void rep_hist_note_or_conn_bytes(uint64_t conn_id, size_t num_read,
+ size_t num_written, time_t when);
+void rep_hist_reset_conn_stats(time_t now);
+char *rep_hist_format_conn_stats(time_t now);
+time_t rep_hist_conn_stats_write(time_t now);
+void rep_hist_conn_stats_term(void);
+
+void rep_hist_note_circuit_handshake_requested(uint16_t type);
+void rep_hist_note_circuit_handshake_assigned(uint16_t type);
+void rep_hist_log_circuit_handshake_stats(time_t now);
+
+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);
+char *rep_hist_get_hs_stats_string(void);
+void rep_hist_seen_new_rp_cell(void);
+void rep_hist_stored_maybe_new_hs(const crypto_pk_t *pubkey);
+
+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);
+
+extern uint64_t rephist_total_alloc;
+extern uint32_t rephist_total_num;
+#ifdef TOR_UNIT_TESTS
+extern int onion_handshakes_requested[MAX_ONION_HANDSHAKE_TYPE+1];
+extern int onion_handshakes_assigned[MAX_ONION_HANDSHAKE_TYPE+1];
++typedef struct bw_array_t bw_array_t;
++extern bw_array_t *write_array;
+#endif
+
+/**
+ * Represents the type of a cell for padding accounting
+ */
+typedef enum padding_type_t {
+ /** A RELAY_DROP cell */
+ PADDING_TYPE_DROP,
+ /** A CELL_PADDING cell */
+ PADDING_TYPE_CELL,
+ /** Total counts of padding and non-padding together */
+ PADDING_TYPE_TOTAL,
+ /** Total cell counts for all padding-enabled channels */
+ PADDING_TYPE_ENABLED_TOTAL,
+ /** CELL_PADDING counts for all padding-enabled channels */
+ PADDING_TYPE_ENABLED_CELL
+} padding_type_t;
+
+/** The amount of time over which the padding cell counts were counted */
+#define REPHIST_CELL_PADDING_COUNTS_INTERVAL (24*60*60)
+void rep_hist_padding_count_read(padding_type_t type);
+void rep_hist_padding_count_write(padding_type_t type);
+char *rep_hist_get_padding_count_lines(void);
+void rep_hist_reset_padding_counts(void);
+void rep_hist_prep_published_padding_counts(time_t now);
+void rep_hist_padding_count_timers(uint64_t num_timers);
+
+#endif /* !defined(TOR_REPHIST_H) */
+
diff --cc src/test/test_relay.c
index 25084fab3,ffb013442..65b9a2f94
--- a/src/test/test_relay.c
+++ b/src/test/test_relay.c
@@@ -1,21 -1,20 +1,23 @@@
-/* Copyright (c) 2014-2017, The Tor Project, Inc. */
+/* Copyright (c) 2014-2018, The Tor Project, Inc. */
/* See LICENSE for licensing information */
-#include "or.h"
+#include "core/or/or.h"
#define CIRCUITBUILD_PRIVATE
-#include "circuitbuild.h"
-#include "circuitlist.h"
-#include "rephist.h"
-#include "channeltls.h"
+#include "core/or/circuitbuild.h"
+#include "core/or/circuitlist.h"
++#include "core/or/channeltls.h"
++#include "feature/stats/rephist.h"
#define RELAY_PRIVATE
-#include "relay.h"
+#include "core/or/relay.h"
/* For init/free stuff */
-#include "scheduler.h"
+#include "core/or/scheduler.h"
+
+#include "core/or/cell_st.h"
+#include "core/or/or_circuit_st.h"
/* Test suite stuff */
-#include "test.h"
-#include "fakechans.h"
+#include "test/test.h"
+#include "test/fakechans.h"
static or_circuit_t * new_fake_orcirc(channel_t *nchan, channel_t *pchan);
@@@ -129,5 -225,8 +228,7 @@@ test_relay_append_cell_to_circuit_queue
struct testcase_t relay_tests[] = {
{ "append_cell_to_circuit_queue", test_relay_append_cell_to_circuit_queue,
TT_FORK, NULL, NULL },
+ { "close_circ_rephist", test_relay_close_circuit,
+ TT_FORK, NULL, NULL },
END_OF_TESTCASES
};
-
_______________________________________________
tor-commits mailing list
tor-commits@xxxxxxxxxxxxxxxxxxxx
https://lists.torproject.org/cgi-bin/mailman/listinfo/tor-commits