April 2013 - tor-commits - lists.torproject.org

[tor/release-0.2.4] Add support for days of the week to intervals
by arma＠torproject.org 11 Apr '13

11 Apr '13

commit cf734a08f60f141ce2c21c703a403aeb74017b1f Author: Nick Mathewson <nickm(a)torproject.org> Date: Mon Mar 11 21:35:50 2013 -0400 Add support for days of the week to intervals --- src/or/config.c | 2 ++ 1 files changed, 2 insertions(+), 0 deletions(-) diff --git a/src/or/config.c b/src/or/config.c index 6ccd65a..adbd367 100644 --- a/src/or/config.c +++ b/src/or/config.c @@ -6534,6 +6534,8 @@ static struct unit_table_t time_units[] = { { "days", 24*60*60 }, { "week", 7*24*60*60 }, { "weeks", 7*24*60*60 }, + { "months", 2629728, }, /* about 30.437 days */ + { "months", 2629728, }, { NULL, 0 }, };

1 0

[tor/release-0.2.4] Merge remote-tracking branch 'public/bug7164_diagnostic' into maint-0.2.4
by arma＠torproject.org 11 Apr '13

11 Apr '13

commit 323cb655be7802a3f527bfa4278478d9fbd23944 Merge: 6f20a74 adfc3de Author: Nick Mathewson <nickm(a)torproject.org> Date: Tue Mar 19 16:16:48 2013 -0400 Merge remote-tracking branch 'public/bug7164_diagnostic' into maint-0.2.4 changes/bug7164_diagnostic | 4 ++++ src/or/microdesc.c | 20 +++++++++++--------- src/or/microdesc.h | 4 +++- 3 files changed, 18 insertions(+), 10 deletions(-)

1 0

[tor/release-0.2.4] Don't assert when writing a cell to a CLOSING connection.
by arma＠torproject.org 11 Apr '13

11 Apr '13

commit ca50fb4f81a1f48c82d75825687085cda4627424 Author: Nick Mathewson <nickm(a)torproject.org> Date: Tue Mar 19 16:35:40 2013 -0400 Don't assert when writing a cell to a CLOSING connection. Instead, drop the cell. Fixes another case of bug 7350; bugfix on 0.2.4.4-alpha --- changes/bug7350 | 4 ++++ src/or/channel.c | 24 ++++++++++++++++++++++++ 2 files changed, 28 insertions(+), 0 deletions(-) diff --git a/changes/bug7350 b/changes/bug7350 new file mode 100644 index 0000000..b0ee9d0 --- /dev/null +++ b/changes/bug7350 @@ -0,0 +1,4 @@ + o Major bugfixes: + - Avoid an assertion when we discover that we'd like to write a cell + onto a closing connection: just discard the cell. Fixes another + case of bug 7350; bugfix on 0.2.4.4-alpha. diff --git a/src/or/channel.c b/src/or/channel.c index 82db061..4e9086f 100644 --- a/src/or/channel.c +++ b/src/or/channel.c @@ -1751,6 +1751,14 @@ channel_write_cell(channel_t *chan, cell_t *cell) tor_assert(chan); tor_assert(cell); + if (chan->state == CHANNEL_STATE_CLOSING) { + log_debug(LD_CHANNEL, "Discarding cell_t %p on closing channel %p with " + "global ID "U64_FORMAT, cell, chan, + U64_PRINTF_ARG(chan->global_identifier)); + tor_free(cell); + return; + } + log_debug(LD_CHANNEL, "Writing cell_t %p to channel %p with global ID " U64_FORMAT, @@ -1777,6 +1785,14 @@ channel_write_packed_cell(channel_t *chan, packed_cell_t *packed_cell) tor_assert(chan); tor_assert(packed_cell); + if (chan->state == CHANNEL_STATE_CLOSING) { + log_debug(LD_CHANNEL, "Discarding packed_cell_t %p on closing channel %p " + "with global ID "U64_FORMAT, packed_cell, chan, + U64_PRINTF_ARG(chan->global_identifier)); + packed_cell_free(packed_cell); + return; + } + log_debug(LD_CHANNEL, "Writing packed_cell_t %p to channel %p with global ID " U64_FORMAT, @@ -1805,6 +1821,14 @@ channel_write_var_cell(channel_t *chan, var_cell_t *var_cell) tor_assert(chan); tor_assert(var_cell); + if (chan->state == CHANNEL_STATE_CLOSING) { + log_debug(LD_CHANNEL, "Discarding var_cell_t %p on closing channel %p " + "with global ID "U64_FORMAT, var_cell, chan, + U64_PRINTF_ARG(chan->global_identifier)); + var_cell_free(var_cell); + return; + } + log_debug(LD_CHANNEL, "Writing var_cell_t %p to channel %p with global ID " U64_FORMAT,

1 0

[tor/release-0.2.4] Merge branch 'bug7350_redux' into maint-0.2.4
by arma＠torproject.org 11 Apr '13

11 Apr '13

commit f1caa2f2146ac750962e6506fe672b03749c50f9 Merge: 323cb65 ca50fb4 Author: Nick Mathewson <nickm(a)torproject.org> Date: Tue Mar 19 16:43:00 2013 -0400 Merge branch 'bug7350_redux' into maint-0.2.4 changes/bug7350 | 4 ++++ src/or/channel.c | 24 ++++++++++++++++++++++++ 2 files changed, 28 insertions(+), 0 deletions(-)

1 0

[tor/release-0.2.4] Merge branch 'bug8240_v2_squashed' into maint-0.2.4
by arma＠torproject.org 11 Apr '13

11 Apr '13

commit 6f20a74d52741cce521cf03b8afee570e3cb367b Merge: a7b4633 18752bc Author: Nick Mathewson <nickm(a)torproject.org> Date: Tue Mar 19 16:15:27 2013 -0400 Merge branch 'bug8240_v2_squashed' into maint-0.2.4 Conflicts: doc/tor.1.txt src/or/circuitbuild.c src/or/config.c src/or/or.h changes/ticket8240 | 4 ++++ doc/tor.1.txt | 6 ++++++ src/common/util.h | 11 +++++++++++ src/or/config.c | 1 + src/or/confparse.c | 2 ++ src/or/entrynodes.c | 35 +++++++++++++++++++++++++++++++++-- src/or/or.h | 2 ++ 7 files changed, 59 insertions(+), 2 deletions(-) diff --cc doc/tor.1.txt index b73d4a0,0c13a5c..3be90be --- a/doc/tor.1.txt +++ b/doc/tor.1.txt @@@ -1045,10 -959,12 +1045,16 @@@ The following options are useful only f If UseEntryGuards is set to 1, we will try to pick a total of NUM routers as long-term entries for our circuits. (Default: 3) +**NumDirectoryGuards** __NUM__:: + If UseEntryGuardsAsDirectoryGuards is enabled, we try to make sure we + have at least NUM routers to use as directory guards. (Default: 3) + + **GuardLifetime** __N__ **days**|**weeks**|**months**:: + If nonzero, and UseEntryGuards is set, minimum time to keep a guard before + picking a new one. If zero, we use the GuardLifetime parameter from the + consensus directory. No value here may be less than 1 month or greater + than 5 years; out-of-range values are clamped. (Default: 0) + **SafeSocks** **0**|**1**:: When this option is enabled, Tor will reject application connections that use unsafe variants of the socks protocol -- ones that only provide an IP diff --cc src/or/confparse.c index 98fde98,0000000..8863d92 mode 100644,000000..100644 --- a/src/or/confparse.c +++ b/src/or/confparse.c @@@ -1,1229 -1,0 +1,1231 @@@ +/* Copyright (c) 2001 Matej Pfajfar. + * Copyright (c) 2001-2004, Roger Dingledine. + * Copyright (c) 2004-2006, Roger Dingledine, Nick Mathewson. + * Copyright (c) 2007-2013, The Tor Project, Inc. */ +/* See LICENSE for licensing information */ + +#include "or.h" +#include "confparse.h" +#include "routerset.h" + +static uint64_t config_parse_memunit(const char *s, int *ok); +static int config_parse_msec_interval(const char *s, int *ok); +static int config_parse_interval(const char *s, int *ok); +static void config_reset(const config_format_t *fmt, void *options, + const config_var_t *var, int use_defaults); + +/** Allocate an empty configuration object of a given format type. */ +void * +config_new(const config_format_t *fmt) +{ + void *opts = tor_malloc_zero(fmt->size); + *(uint32_t*)STRUCT_VAR_P(opts, fmt->magic_offset) = fmt->magic; + CONFIG_CHECK(fmt, opts); + return opts; +} + +/* + * Functions to parse config options + */ + +/** If option is an official abbreviation for a longer option, + * return the longer option. Otherwise return option. + * If command_line is set, apply all abbreviations. Otherwise, only + * apply abbreviations that work for the config file and the command line. + * If warn_obsolete is set, warn about deprecated names. */ +const char * +config_expand_abbrev(const config_format_t *fmt, const char *option, + int command_line, int warn_obsolete) +{ + int i; + if (! fmt->abbrevs) + return option; + for (i=0; fmt->abbrevs[i].abbreviated; ++i) { + /* Abbreviations are case insensitive. */ + if (!strcasecmp(option,fmt->abbrevs[i].abbreviated) && + (command_line || !fmt->abbrevs[i].commandline_only)) { + if (warn_obsolete && fmt->abbrevs[i].warn) { + log_warn(LD_CONFIG, + "The configuration option '%s' is deprecated; " + "use '%s' instead.", + fmt->abbrevs[i].abbreviated, + fmt->abbrevs[i].full); + } + /* Keep going through the list in case we want to rewrite it more. + * (We could imagine recursing here, but I don't want to get the + * user into an infinite loop if we craft our list wrong.) */ + option = fmt->abbrevs[i].full; + } + } + return option; +} + +/** Helper: allocate a new configuration option mapping 'key' to 'val', + * append it to *lst. */ +void +config_line_append(config_line_t **lst, + const char *key, + const char *val) +{ + config_line_t *newline; + + newline = tor_malloc_zero(sizeof(config_line_t)); + newline->key = tor_strdup(key); + newline->value = tor_strdup(val); + newline->next = NULL; + while (*lst) + lst = &((*lst)->next); + + (*lst) = newline; +} + +/** Helper: parse the config string and strdup into key/value + * strings. Set *result to the list, or NULL if parsing the string + * failed. Return 0 on success, -1 on failure. Warn and ignore any + * misformatted lines. + * + * If extended is set, then treat keys beginning with / and with + as + * indicating "clear" and "append" respectively. */ +int +config_get_lines(const char *string, config_line_t **result, int extended) +{ + config_line_t *list = NULL, **next; + char *k, *v; + const char *parse_err; + + next = &list; + do { + k = v = NULL; + string = parse_config_line_from_str_verbose(string, &k, &v, &parse_err); + if (!string) { + log_warn(LD_CONFIG, "Error while parsing configuration: %s", + parse_err?parse_err:"<unknown>"); + config_free_lines(list); + tor_free(k); + tor_free(v); + return -1; + } + if (k && v) { + unsigned command = CONFIG_LINE_NORMAL; + if (extended) { + if (k[0] == '+') { + char *k_new = tor_strdup(k+1); + tor_free(k); + k = k_new; + command = CONFIG_LINE_APPEND; + } else if (k[0] == '/') { + char *k_new = tor_strdup(k+1); + tor_free(k); + k = k_new; + tor_free(v); + v = tor_strdup(""); + command = CONFIG_LINE_CLEAR; + } + } + /* This list can get long, so we keep a pointer to the end of it + * rather than using config_line_append over and over and getting + * n^2 performance. */ + *next = tor_malloc_zero(sizeof(config_line_t)); + (*next)->key = k; + (*next)->value = v; + (*next)->next = NULL; + (*next)->command = command; + next = &((*next)->next); + } else { + tor_free(k); + tor_free(v); + } + } while (*string); + + *result = list; + return 0; +} + +/** + * Free all the configuration lines on the linked list front. + */ +void +config_free_lines(config_line_t *front) +{ + config_line_t *tmp; + + while (front) { + tmp = front; + front = tmp->next; + + tor_free(tmp->key); + tor_free(tmp->value); + tor_free(tmp); + } +} + +/** As config_find_option, but return a non-const pointer. */ +config_var_t * +config_find_option_mutable(config_format_t *fmt, const char *key) +{ + int i; + size_t keylen = strlen(key); + if (!keylen) + return NULL; /* if they say "--" on the command line, it's not an option */ + /* First, check for an exact (case-insensitive) match */ + for (i=0; fmt->vars[i].name; ++i) { + if (!strcasecmp(key, fmt->vars[i].name)) { + return &fmt->vars[i]; + } + } + /* If none, check for an abbreviated match */ + for (i=0; fmt->vars[i].name; ++i) { + if (!strncasecmp(key, fmt->vars[i].name, keylen)) { + log_warn(LD_CONFIG, "The abbreviation '%s' is deprecated. " + "Please use '%s' instead", + key, fmt->vars[i].name); + return &fmt->vars[i]; + } + } + /* Okay, unrecognized option */ + return NULL; +} + +/** If key is a configuration option, return the corresponding const + * config_var_t. Otherwise, if key is a non-standard abbreviation, + * warn, and return the corresponding const config_var_t. Otherwise return + * NULL. + */ +const config_var_t * +config_find_option(const config_format_t *fmt, const char *key) +{ + return config_find_option_mutable((config_format_t*)fmt, key); +} + +/** Return the number of option entries in fmt. */ +static int +config_count_options(const config_format_t *fmt) +{ + int i; + for (i=0; fmt->vars[i].name; ++i) + ; + return i; +} + +/* + * Functions to assign config options. + */ + +/** c-\>key is known to be a real key. Update options + * with c-\>value and return 0, or return -1 if bad value. + * + * Called from config_assign_line() and option_reset(). + */ +static int +config_assign_value(const config_format_t *fmt, void *options, + config_line_t *c, char **msg) +{ + int i, ok; + const config_var_t *var; + void *lvalue; + + CONFIG_CHECK(fmt, options); + + var = config_find_option(fmt, c->key); + tor_assert(var); + + lvalue = STRUCT_VAR_P(options, var->var_offset); + + switch (var->type) { + + case CONFIG_TYPE_PORT: + if (!strcasecmp(c->value, "auto")) { + *(int *)lvalue = CFG_AUTO_PORT; + break; + } + /* fall through */ + case CONFIG_TYPE_INT: + case CONFIG_TYPE_UINT: + i = (int)tor_parse_long(c->value, 10, + var->type==CONFIG_TYPE_INT ? INT_MIN : 0, + var->type==CONFIG_TYPE_PORT ? 65535 : INT_MAX, + &ok, NULL); + if (!ok) { + tor_asprintf(msg, + "Int keyword '%s %s' is malformed or out of bounds.", + c->key, c->value); + return -1; + } + *(int *)lvalue = i; + break; + + case CONFIG_TYPE_INTERVAL: { + i = config_parse_interval(c->value, &ok); + if (!ok) { + tor_asprintf(msg, + "Interval '%s %s' is malformed or out of bounds.", + c->key, c->value); + return -1; + } + *(int *)lvalue = i; + break; + } + + case CONFIG_TYPE_MSEC_INTERVAL: { + i = config_parse_msec_interval(c->value, &ok); + if (!ok) { + tor_asprintf(msg, + "Msec interval '%s %s' is malformed or out of bounds.", + c->key, c->value); + return -1; + } + *(int *)lvalue = i; + break; + } + + case CONFIG_TYPE_MEMUNIT: { + uint64_t u64 = config_parse_memunit(c->value, &ok); + if (!ok) { + tor_asprintf(msg, + "Value '%s %s' is malformed or out of bounds.", + c->key, c->value); + return -1; + } + *(uint64_t *)lvalue = u64; + break; + } + + case CONFIG_TYPE_BOOL: + i = (int)tor_parse_long(c->value, 10, 0, 1, &ok, NULL); + if (!ok) { + tor_asprintf(msg, + "Boolean '%s %s' expects 0 or 1.", + c->key, c->value); + return -1; + } + *(int *)lvalue = i; + break; + + case CONFIG_TYPE_AUTOBOOL: + if (!strcmp(c->value, "auto")) + *(int *)lvalue = -1; + else if (!strcmp(c->value, "0")) + *(int *)lvalue = 0; + else if (!strcmp(c->value, "1")) + *(int *)lvalue = 1; + else { + tor_asprintf(msg, "Boolean '%s %s' expects 0, 1, or 'auto'.", + c->key, c->value); + return -1; + } + break; + + case CONFIG_TYPE_STRING: + case CONFIG_TYPE_FILENAME: + tor_free(*(char **)lvalue); + *(char **)lvalue = tor_strdup(c->value); + break; + + case CONFIG_TYPE_DOUBLE: + *(double *)lvalue = atof(c->value); + break; + + case CONFIG_TYPE_ISOTIME: + if (parse_iso_time(c->value, (time_t *)lvalue)) { + tor_asprintf(msg, + "Invalid time '%s' for keyword '%s'", c->value, c->key); + return -1; + } + break; + + case CONFIG_TYPE_ROUTERSET: + if (*(routerset_t**)lvalue) { + routerset_free(*(routerset_t**)lvalue); + } + *(routerset_t**)lvalue = routerset_new(); + if (routerset_parse(*(routerset_t**)lvalue, c->value, c->key)<0) { + tor_asprintf(msg, "Invalid exit list '%s' for option '%s'", + c->value, c->key); + return -1; + } + break; + + case CONFIG_TYPE_CSV: + if (*(smartlist_t**)lvalue) { + SMARTLIST_FOREACH(*(smartlist_t**)lvalue, char *, cp, tor_free(cp)); + smartlist_clear(*(smartlist_t**)lvalue); + } else { + *(smartlist_t**)lvalue = smartlist_new(); + } + + smartlist_split_string(*(smartlist_t**)lvalue, c->value, ",", + SPLIT_SKIP_SPACE|SPLIT_IGNORE_BLANK, 0); + break; + + case CONFIG_TYPE_LINELIST: + case CONFIG_TYPE_LINELIST_S: + { + config_line_t *lastval = *(config_line_t**)lvalue; + if (lastval && lastval->fragile) { + if (c->command != CONFIG_LINE_APPEND) { + config_free_lines(lastval); + *(config_line_t**)lvalue = NULL; + } else { + lastval->fragile = 0; + } + } + + config_line_append((config_line_t**)lvalue, c->key, c->value); + } + break; + case CONFIG_TYPE_OBSOLETE: + log_warn(LD_CONFIG, "Skipping obsolete configuration option '%s'", c->key); + break; + case CONFIG_TYPE_LINELIST_V: + tor_asprintf(msg, + "You may not provide a value for virtual option '%s'", c->key); + return -1; + default: + tor_assert(0); + break; + } + return 0; +} + +/** Mark every linelist in options "fragile", so that fresh assignments + * to it will replace old ones. */ +static void +config_mark_lists_fragile(const config_format_t *fmt, void *options) +{ + int i; + tor_assert(fmt); + tor_assert(options); + + for (i = 0; fmt->vars[i].name; ++i) { + const config_var_t *var = &fmt->vars[i]; + config_line_t *list; + if (var->type != CONFIG_TYPE_LINELIST && + var->type != CONFIG_TYPE_LINELIST_V) + continue; + + list = *(config_line_t **)STRUCT_VAR_P(options, var->var_offset); + if (list) + list->fragile = 1; + } +} + +/** If c is a syntactically valid configuration line, update + * options with its value and return 0. Otherwise return -1 for bad + * key, -2 for bad value. + * + * If clear_first is set, clear the value first. Then if + * use_defaults is set, set the value to the default. + * + * Called from config_assign(). + */ +static int +config_assign_line(const config_format_t *fmt, void *options, + config_line_t *c, int use_defaults, + int clear_first, bitarray_t *options_seen, char **msg) +{ + const config_var_t *var; + + CONFIG_CHECK(fmt, options); + + var = config_find_option(fmt, c->key); + if (!var) { + if (fmt->extra) { + void *lvalue = STRUCT_VAR_P(options, fmt->extra->var_offset); + log_info(LD_CONFIG, + "Found unrecognized option '%s'; saving it.", c->key); + config_line_append((config_line_t**)lvalue, c->key, c->value); + return 0; + } else { + tor_asprintf(msg, + "Unknown option '%s'. Failing.", c->key); + return -1; + } + } + + /* Put keyword into canonical case. */ + if (strcmp(var->name, c->key)) { + tor_free(c->key); + c->key = tor_strdup(var->name); + } + + if (!strlen(c->value)) { + /* reset or clear it, then return */ + if (!clear_first) { + if ((var->type == CONFIG_TYPE_LINELIST || + var->type == CONFIG_TYPE_LINELIST_S) && + c->command != CONFIG_LINE_CLEAR) { + /* We got an empty linelist from the torrc or command line. + As a special case, call this an error. Warn and ignore. */ + log_warn(LD_CONFIG, + "Linelist option '%s' has no value. Skipping.", c->key); + } else { /* not already cleared */ + config_reset(fmt, options, var, use_defaults); + } + } + return 0; + } else if (c->command == CONFIG_LINE_CLEAR && !clear_first) { + config_reset(fmt, options, var, use_defaults); + } + + if (options_seen && (var->type != CONFIG_TYPE_LINELIST && + var->type != CONFIG_TYPE_LINELIST_S)) { + /* We're tracking which options we've seen, and this option is not + * supposed to occur more than once. */ + int var_index = (int)(var - fmt->vars); + if (bitarray_is_set(options_seen, var_index)) { + log_warn(LD_CONFIG, "Option '%s' used more than once; all but the last " + "value will be ignored.", var->name); + } + bitarray_set(options_seen, var_index); + } + + if (config_assign_value(fmt, options, c, msg) < 0) + return -2; + return 0; +} + +/** Restore the option named key in options to its default value. + * Called from config_assign(). */ +static void +config_reset_line(const config_format_t *fmt, void *options, + const char *key, int use_defaults) +{ + const config_var_t *var; + + CONFIG_CHECK(fmt, options); + + var = config_find_option(fmt, key); + if (!var) + return; /* give error on next pass. */ + + config_reset(fmt, options, var, use_defaults); +} + +/** Return true iff value needs to be quoted and escaped to be used in + * a configuration file. */ +static int +config_value_needs_escape(const char *value) +{ + if (*value == '\"') + return 1; + while (*value) { + switch (*value) + { + case '\r': + case '\n': + case '#': + /* Note: quotes and backspaces need special handling when we are using + * quotes, not otherwise, so they don't trigger escaping on their + * own. */ + return 1; + default: + if (!TOR_ISPRINT(*value)) + return 1; + } + ++value; + } + return 0; +} + +/** Return a newly allocated deep copy of the lines in inp. */ +config_line_t * +config_lines_dup(const config_line_t *inp) +{ + config_line_t *result = NULL; + config_line_t **next_out = &result; + while (inp) { + *next_out = tor_malloc_zero(sizeof(config_line_t)); + (*next_out)->key = tor_strdup(inp->key); + (*next_out)->value = tor_strdup(inp->value); + inp = inp->next; + next_out = &((*next_out)->next); + } + (*next_out) = NULL; + return result; +} + +/** Return newly allocated line or lines corresponding to key in the + * configuration options. If escape_val is true and a + * value needs to be quoted before it's put in a config file, quote and + * escape that value. Return NULL if no such key exists. */ +config_line_t * +config_get_assigned_option(const config_format_t *fmt, const void *options, + const char *key, int escape_val) +{ + const config_var_t *var; + const void *value; + config_line_t *result; + tor_assert(options && key); + + CONFIG_CHECK(fmt, options); + + var = config_find_option(fmt, key); + if (!var) { + log_warn(LD_CONFIG, "Unknown option '%s'. Failing.", key); + return NULL; + } + value = STRUCT_VAR_P(options, var->var_offset); + + result = tor_malloc_zero(sizeof(config_line_t)); + result->key = tor_strdup(var->name); + switch (var->type) + { + case CONFIG_TYPE_STRING: + case CONFIG_TYPE_FILENAME: + if (*(char**)value) { + result->value = tor_strdup(*(char**)value); + } else { + tor_free(result->key); + tor_free(result); + return NULL; + } + break; + case CONFIG_TYPE_ISOTIME: + if (*(time_t*)value) { + result->value = tor_malloc(ISO_TIME_LEN+1); + format_iso_time(result->value, *(time_t*)value); + } else { + tor_free(result->key); + tor_free(result); + } + escape_val = 0; /* Can't need escape. */ + break; + case CONFIG_TYPE_PORT: + if (*(int*)value == CFG_AUTO_PORT) { + result->value = tor_strdup("auto"); + escape_val = 0; + break; + } + /* fall through */ + case CONFIG_TYPE_INTERVAL: + case CONFIG_TYPE_MSEC_INTERVAL: + case CONFIG_TYPE_UINT: + case CONFIG_TYPE_INT: + /* This means every or_options_t uint or bool element + * needs to be an int. Not, say, a uint16_t or char. */ + tor_asprintf(&result->value, "%d", *(int*)value); + escape_val = 0; /* Can't need escape. */ + break; + case CONFIG_TYPE_MEMUNIT: + tor_asprintf(&result->value, U64_FORMAT, + U64_PRINTF_ARG(*(uint64_t*)value)); + escape_val = 0; /* Can't need escape. */ + break; + case CONFIG_TYPE_DOUBLE: + tor_asprintf(&result->value, "%f", *(double*)value); + escape_val = 0; /* Can't need escape. */ + break; + + case CONFIG_TYPE_AUTOBOOL: + if (*(int*)value == -1) { + result->value = tor_strdup("auto"); + escape_val = 0; + break; + } + /* fall through */ + case CONFIG_TYPE_BOOL: + result->value = tor_strdup(*(int*)value ? "1" : "0"); + escape_val = 0; /* Can't need escape. */ + break; + case CONFIG_TYPE_ROUTERSET: + result->value = routerset_to_string(*(routerset_t**)value); + break; + case CONFIG_TYPE_CSV: + if (*(smartlist_t**)value) + result->value = + smartlist_join_strings(*(smartlist_t**)value, ",", 0, NULL); + else + result->value = tor_strdup(""); + break; + case CONFIG_TYPE_OBSOLETE: + log_fn(LOG_INFO, LD_CONFIG, + "You asked me for the value of an obsolete config option '%s'.", + key); + tor_free(result->key); + tor_free(result); + return NULL; + case CONFIG_TYPE_LINELIST_S: + log_warn(LD_CONFIG, + "Can't return context-sensitive '%s' on its own", key); + tor_free(result->key); + tor_free(result); + return NULL; + case CONFIG_TYPE_LINELIST: + case CONFIG_TYPE_LINELIST_V: + tor_free(result->key); + tor_free(result); + result = config_lines_dup(*(const config_line_t**)value); + break; + default: + tor_free(result->key); + tor_free(result); + log_warn(LD_BUG,"Unknown type %d for known key '%s'", + var->type, key); + return NULL; + } + + if (escape_val) { + config_line_t *line; + for (line = result; line; line = line->next) { + if (line->value && config_value_needs_escape(line->value)) { + char *newval = esc_for_log(line->value); + tor_free(line->value); + line->value = newval; + } + } + } + + return result; +} +/** Iterate through the linked list of requested options list. + * For each item, convert as appropriate and assign to options. + * If an item is unrecognized, set *msg and return -1 immediately, + * else return 0 for success. + * + * If clear_first, interpret config options as replacing (not + * extending) their previous values. If clear_first is set, + * then use_defaults to decide if you set to defaults after + * clearing, or make the value 0 or NULL. + * + * Here are the use cases: + * 1. A non-empty AllowInvalid line in your torrc. Appends to current + * if linelist, replaces current if csv. + * 2. An empty AllowInvalid line in your torrc. Should clear it. + * 3. "RESETCONF AllowInvalid" sets it to default. + * 4. "SETCONF AllowInvalid" makes it NULL. + * 5. "SETCONF AllowInvalid=foo" clears it and sets it to "foo". + * + * Use_defaults Clear_first + * 0 0 "append" + * 1 0 undefined, don't use + * 0 1 "set to null first" + * 1 1 "set to defaults first" + * Return 0 on success, -1 on bad key, -2 on bad value. + * + * As an additional special case, if a LINELIST config option has + * no value and clear_first is 0, then warn and ignore it. + */ + +/* +There are three call cases for config_assign() currently. + +Case one: Torrc entry +options_init_from_torrc() calls config_assign(0, 0) + calls config_assign_line(0, 0). + if value is empty, calls config_reset(0) and returns. + calls config_assign_value(), appends. + +Case two: setconf +options_trial_assign() calls config_assign(0, 1) + calls config_reset_line(0) + calls config_reset(0) + calls option_clear(). + calls config_assign_line(0, 1). + if value is empty, returns. + calls config_assign_value(), appends. + +Case three: resetconf +options_trial_assign() calls config_assign(1, 1) + calls config_reset_line(1) + calls config_reset(1) + calls option_clear(). + calls config_assign_value(default) + calls config_assign_line(1, 1). + returns. +*/ +int +config_assign(const config_format_t *fmt, void *options, config_line_t *list, + int use_defaults, int clear_first, char **msg) +{ + config_line_t *p; + bitarray_t *options_seen; + const int n_options = config_count_options(fmt); + + CONFIG_CHECK(fmt, options); + + /* pass 1: normalize keys */ + for (p = list; p; p = p->next) { + const char *full = config_expand_abbrev(fmt, p->key, 0, 1); + if (strcmp(full,p->key)) { + tor_free(p->key); + p->key = tor_strdup(full); + } + } + + /* pass 2: if we're reading from a resetting source, clear all + * mentioned config options, and maybe set to their defaults. */ + if (clear_first) { + for (p = list; p; p = p->next) + config_reset_line(fmt, options, p->key, use_defaults); + } + + options_seen = bitarray_init_zero(n_options); + /* pass 3: assign. */ + while (list) { + int r; + if ((r=config_assign_line(fmt, options, list, use_defaults, + clear_first, options_seen, msg))) { + bitarray_free(options_seen); + return r; + } + list = list->next; + } + bitarray_free(options_seen); + + /** Now we're done assigning a group of options to the configuration. + * Subsequent group assignments should _replace_ linelists, not extend + * them. */ + config_mark_lists_fragile(fmt, options); + + return 0; +} + +/** Reset config option var to 0, 0.0, NULL, or the equivalent. + * Called from config_reset() and config_free(). */ +static void +config_clear(const config_format_t *fmt, void *options, + const config_var_t *var) +{ + void *lvalue = STRUCT_VAR_P(options, var->var_offset); + (void)fmt; /* unused */ + switch (var->type) { + case CONFIG_TYPE_STRING: + case CONFIG_TYPE_FILENAME: + tor_free(*(char**)lvalue); + break; + case CONFIG_TYPE_DOUBLE: + *(double*)lvalue = 0.0; + break; + case CONFIG_TYPE_ISOTIME: + *(time_t*)lvalue = 0; + break; + case CONFIG_TYPE_INTERVAL: + case CONFIG_TYPE_MSEC_INTERVAL: + case CONFIG_TYPE_UINT: + case CONFIG_TYPE_INT: + case CONFIG_TYPE_PORT: + case CONFIG_TYPE_BOOL: + *(int*)lvalue = 0; + break; + case CONFIG_TYPE_AUTOBOOL: + *(int*)lvalue = -1; + break; + case CONFIG_TYPE_MEMUNIT: + *(uint64_t*)lvalue = 0; + break; + case CONFIG_TYPE_ROUTERSET: + if (*(routerset_t**)lvalue) { + routerset_free(*(routerset_t**)lvalue); + *(routerset_t**)lvalue = NULL; + } + break; + case CONFIG_TYPE_CSV: + if (*(smartlist_t**)lvalue) { + SMARTLIST_FOREACH(*(smartlist_t **)lvalue, char *, cp, tor_free(cp)); + smartlist_free(*(smartlist_t **)lvalue); + *(smartlist_t **)lvalue = NULL; + } + break; + case CONFIG_TYPE_LINELIST: + case CONFIG_TYPE_LINELIST_S: + config_free_lines(*(config_line_t **)lvalue); + *(config_line_t **)lvalue = NULL; + break; + case CONFIG_TYPE_LINELIST_V: + /* handled by linelist_s. */ + break; + case CONFIG_TYPE_OBSOLETE: + break; + } +} + +/** Clear the option indexed by var in options. Then if + * use_defaults, set it to its default value. + * Called by config_init() and option_reset_line() and option_assign_line(). */ +static void +config_reset(const config_format_t *fmt, void *options, + const config_var_t *var, int use_defaults) +{ + config_line_t *c; + char *msg = NULL; + CONFIG_CHECK(fmt, options); + config_clear(fmt, options, var); /* clear it first */ + if (!use_defaults) + return; /* all done */ + if (var->initvalue) { + c = tor_malloc_zero(sizeof(config_line_t)); + c->key = tor_strdup(var->name); + c->value = tor_strdup(var->initvalue); + if (config_assign_value(fmt, options, c, &msg) < 0) { + log_warn(LD_BUG, "Failed to assign default: %s", msg); + tor_free(msg); /* if this happens it's a bug */ + } + config_free_lines(c); + } +} + +/** Release storage held by options. */ +void +config_free(const config_format_t *fmt, void *options) +{ + int i; + + if (!options) + return; + + tor_assert(fmt); + + for (i=0; fmt->vars[i].name; ++i) + config_clear(fmt, options, &(fmt->vars[i])); + if (fmt->extra) { + config_line_t **linep = STRUCT_VAR_P(options, fmt->extra->var_offset); + config_free_lines(*linep); + *linep = NULL; + } + tor_free(options); +} + +/** Return true iff a and b contain identical keys and values in identical + * order. */ +int +config_lines_eq(config_line_t *a, config_line_t *b) +{ + while (a && b) { + if (strcasecmp(a->key, b->key) || strcmp(a->value, b->value)) + return 0; + a = a->next; + b = b->next; + } + if (a || b) + return 0; + return 1; +} + +/** Return the number of lines in a whose key is key. */ +int +config_count_key(const config_line_t *a, const char *key) +{ + int n = 0; + while (a) { + if (!strcasecmp(a->key, key)) { + ++n; + } + a = a->next; + } + return n; +} + +/** Return true iff the option name has the same value in o1 + * and o2. Must not be called for LINELIST_S or OBSOLETE options. + */ +int +config_is_same(const config_format_t *fmt, + const void *o1, const void *o2, + const char *name) +{ + config_line_t *c1, *c2; + int r = 1; + CONFIG_CHECK(fmt, o1); + CONFIG_CHECK(fmt, o2); + + c1 = config_get_assigned_option(fmt, o1, name, 0); + c2 = config_get_assigned_option(fmt, o2, name, 0); + r = config_lines_eq(c1, c2); + config_free_lines(c1); + config_free_lines(c2); + return r; +} + +/** Copy storage held by old into a new or_options_t and return it. */ +void * +config_dup(const config_format_t *fmt, const void *old) +{ + void *newopts; + int i; + config_line_t *line; + + newopts = config_new(fmt); + for (i=0; fmt->vars[i].name; ++i) { + if (fmt->vars[i].type == CONFIG_TYPE_LINELIST_S) + continue; + if (fmt->vars[i].type == CONFIG_TYPE_OBSOLETE) + continue; + line = config_get_assigned_option(fmt, old, fmt->vars[i].name, 0); + if (line) { + char *msg = NULL; + if (config_assign(fmt, newopts, line, 0, 0, &msg) < 0) { + log_err(LD_BUG, "config_get_assigned_option() generated " + "something we couldn't config_assign(): %s", msg); + tor_free(msg); + tor_assert(0); + } + } + config_free_lines(line); + } + return newopts; +} +/** Set all vars in the configuration object options to their default + * values. */ +void +config_init(const config_format_t *fmt, void *options) +{ + int i; + const config_var_t *var; + CONFIG_CHECK(fmt, options); + + for (i=0; fmt->vars[i].name; ++i) { + var = &fmt->vars[i]; + if (!var->initvalue) + continue; /* defaults to NULL or 0 */ + config_reset(fmt, options, var, 1); + } +} + +/** Allocate and return a new string holding the written-out values of the vars + * in 'options'. If 'minimal', do not write out any default-valued vars. + * Else, if comment_defaults, write default values as comments. + */ +char * +config_dump(const config_format_t *fmt, const void *default_options, + const void *options, int minimal, + int comment_defaults) +{ + smartlist_t *elements; + const void *defaults = default_options; + void *defaults_tmp = NULL; + config_line_t *line, *assigned; + char *result; + int i; + char *msg = NULL; + + if (defaults == NULL) { + defaults = defaults_tmp = config_new(fmt); + config_init(fmt, defaults_tmp); + } + + /* XXX use a 1 here so we don't add a new log line while dumping */ + if (default_options == NULL) { + if (fmt->validate_fn(NULL, defaults_tmp, 1, &msg) < 0) { + log_err(LD_BUG, "Failed to validate default config."); + tor_free(msg); + tor_assert(0); + } + } + + elements = smartlist_new(); + for (i=0; fmt->vars[i].name; ++i) { + int comment_option = 0; + if (fmt->vars[i].type == CONFIG_TYPE_OBSOLETE || + fmt->vars[i].type == CONFIG_TYPE_LINELIST_S) + continue; + /* Don't save 'hidden' control variables. */ + if (!strcmpstart(fmt->vars[i].name, "__")) + continue; + if (minimal && config_is_same(fmt, options, defaults, fmt->vars[i].name)) + continue; + else if (comment_defaults && + config_is_same(fmt, options, defaults, fmt->vars[i].name)) + comment_option = 1; + + line = assigned = + config_get_assigned_option(fmt, options, fmt->vars[i].name, 1); + + for (; line; line = line->next) { + smartlist_add_asprintf(elements, "%s%s %s\n", + comment_option ? "# " : "", + line->key, line->value); + } + config_free_lines(assigned); + } + + if (fmt->extra) { + line = *(config_line_t**)STRUCT_VAR_P(options, fmt->extra->var_offset); + for (; line; line = line->next) { + smartlist_add_asprintf(elements, "%s %s\n", line->key, line->value); + } + } + + result = smartlist_join_strings(elements, "", 0, NULL); + SMARTLIST_FOREACH(elements, char *, cp, tor_free(cp)); + smartlist_free(elements); + if (defaults_tmp) + config_free(fmt, defaults_tmp); + return result; +} + +/** Mapping from a unit name to a multiplier for converting that unit into a + * base unit. Used by config_parse_unit. */ +struct unit_table_t { + const char *unit; /**< The name of the unit */ + uint64_t multiplier; /**< How many of the base unit appear in this unit */ +}; + +/** Table to map the names of memory units to the number of bytes they + * contain. */ +static struct unit_table_t memory_units[] = { + { "", 1 }, + { "b", 1<< 0 }, + { "byte", 1<< 0 }, + { "bytes", 1<< 0 }, + { "kb", 1<<10 }, + { "kbyte", 1<<10 }, + { "kbytes", 1<<10 }, + { "kilobyte", 1<<10 }, + { "kilobytes", 1<<10 }, + { "m", 1<<20 }, + { "mb", 1<<20 }, + { "mbyte", 1<<20 }, + { "mbytes", 1<<20 }, + { "megabyte", 1<<20 }, + { "megabytes", 1<<20 }, + { "gb", 1<<30 }, + { "gbyte", 1<<30 }, + { "gbytes", 1<<30 }, + { "gigabyte", 1<<30 }, + { "gigabytes", 1<<30 }, + { "tb", U64_LITERAL(1)<<40 }, + { "terabyte", U64_LITERAL(1)<<40 }, + { "terabytes", U64_LITERAL(1)<<40 }, + { NULL, 0 }, +}; + +/** Table to map the names of time units to the number of seconds they + * contain. */ +static struct unit_table_t time_units[] = { + { "", 1 }, + { "second", 1 }, + { "seconds", 1 }, + { "minute", 60 }, + { "minutes", 60 }, + { "hour", 60*60 }, + { "hours", 60*60 }, + { "day", 24*60*60 }, + { "days", 24*60*60 }, + { "week", 7*24*60*60 }, + { "weeks", 7*24*60*60 }, ++ { "month", 2629728, }, /* about 30.437 days */ ++ { "months", 2629728, }, + { NULL, 0 }, +}; + +/** Table to map the names of time units to the number of milliseconds + * they contain. */ +static struct unit_table_t time_msec_units[] = { + { "", 1 }, + { "msec", 1 }, + { "millisecond", 1 }, + { "milliseconds", 1 }, + { "second", 1000 }, + { "seconds", 1000 }, + { "minute", 60*1000 }, + { "minutes", 60*1000 }, + { "hour", 60*60*1000 }, + { "hours", 60*60*1000 }, + { "day", 24*60*60*1000 }, + { "days", 24*60*60*1000 }, + { "week", 7*24*60*60*1000 }, + { "weeks", 7*24*60*60*1000 }, + { NULL, 0 }, +}; + +/** Parse a string val containing a number, zero or more + * spaces, and an optional unit string. If the unit appears in the + * table u, then multiply the number by the unit multiplier. + * On success, set *ok to 1 and return this product. + * Otherwise, set *ok to 0. + */ +static uint64_t +config_parse_units(const char *val, struct unit_table_t *u, int *ok) +{ + uint64_t v = 0; + double d = 0; + int use_float = 0; + char *cp; + + tor_assert(ok); + + v = tor_parse_uint64(val, 10, 0, UINT64_MAX, ok, &cp); + if (!*ok || (cp && *cp == '.')) { + d = tor_parse_double(val, 0, UINT64_MAX, ok, &cp); + if (!*ok) + goto done; + use_float = 1; + } + + if (!cp) { + *ok = 1; + v = use_float ? DBL_TO_U64(d) : v; + goto done; + } + + cp = (char*) eat_whitespace(cp); + + for ( ;u->unit;++u) { + if (!strcasecmp(u->unit, cp)) { + if (use_float) + v = u->multiplier * d; + else + v *= u->multiplier; + *ok = 1; + goto done; + } + } + log_warn(LD_CONFIG, "Unknown unit '%s'.", cp); + *ok = 0; + done: + + if (*ok) + return v; + else + return 0; +} + +/** Parse a string in the format "number unit", where unit is a unit of + * information (byte, KB, M, etc). On success, set *ok to true + * and return the number of bytes specified. Otherwise, set + * *ok to false and return 0. */ +static uint64_t +config_parse_memunit(const char *s, int *ok) +{ + uint64_t u = config_parse_units(s, memory_units, ok); + return u; +} + +/** Parse a string in the format "number unit", where unit is a unit of + * time in milliseconds. On success, set *ok to true and return + * the number of milliseconds in the provided interval. Otherwise, set + * *ok to 0 and return -1. */ +static int +config_parse_msec_interval(const char *s, int *ok) +{ + uint64_t r; + r = config_parse_units(s, time_msec_units, ok); + if (!ok) + return -1; + if (r > INT_MAX) { + log_warn(LD_CONFIG, "Msec interval '%s' is too long", s); + *ok = 0; + return -1; + } + return (int)r; +} + +/** Parse a string in the format "number unit", where unit is a unit of time. + * On success, set *ok to true and return the number of seconds in + * the provided interval. Otherwise, set *ok to 0 and return -1. + */ +static int +config_parse_interval(const char *s, int *ok) +{ + uint64_t r; + r = config_parse_units(s, time_units, ok); + if (!ok) + return -1; + if (r > INT_MAX) { + log_warn(LD_CONFIG, "Interval '%s' is too long", s); + *ok = 0; + return -1; + } + return (int)r; +} + diff --cc src/or/entrynodes.c index e92c0c1,0000000..5bb0a73 mode 100644,000000..100644 --- a/src/or/entrynodes.c +++ b/src/or/entrynodes.c @@@ -1,2173 -1,0 +1,2204 @@@ +/* Copyright (c) 2001 Matej Pfajfar. + * Copyright (c) 2001-2004, Roger Dingledine. + * Copyright (c) 2004-2006, Roger Dingledine, Nick Mathewson. + * Copyright (c) 2007-2013, The Tor Project, Inc. */ +/* See LICENSE for licensing information */ + +/** + * \file entrynodes.c + * \brief Code to manage our fixed first nodes for various functions. + * + * Entry nodes can be guards (for general use) or bridges (for censorship + * circumvention). + **/ + +#include "or.h" +#include "circuitbuild.h" +#include "circuitstats.h" +#include "config.h" +#include "confparse.h" +#include "connection.h" +#include "connection_or.h" +#include "control.h" +#include "directory.h" +#include "entrynodes.h" +#include "main.h" +#include "microdesc.h" ++#include "networkstatus.h" +#include "nodelist.h" +#include "policies.h" +#include "router.h" +#include "routerlist.h" +#include "routerparse.h" +#include "routerset.h" +#include "transports.h" +#include "statefile.h" + +/** Information about a configured bridge. Currently this just matches the + * ones in the torrc file, but one day we may be able to learn about new + * bridges on our own, and remember them in the state file. */ +typedef struct { + /** Address of the bridge. */ + tor_addr_t addr; + /** TLS port for the bridge. */ + uint16_t port; + /** Boolean: We are re-parsing our bridge list, and we are going to remove + * this one if we don't find it in the list of configured bridges. */ + unsigned marked_for_removal : 1; + /** Expected identity digest, or all zero bytes if we don't know what the + * digest should be. */ + char identity[DIGEST_LEN]; + + /** Name of pluggable transport protocol taken from its config line. */ + char *transport_name; + + /** When should we next try to fetch a descriptor for this bridge? */ + download_status_t fetch_status; +} bridge_info_t; + +/** A list of our chosen entry guards. */ +static smartlist_t *entry_guards = NULL; +/** A value of 1 means that the entry_guards list has changed + * and those changes need to be flushed to disk. */ +static int entry_guards_dirty = 0; + +static void bridge_free(bridge_info_t *bridge); +static const node_t *choose_random_entry_impl(cpath_build_state_t *state, + int for_directory, + dirinfo_type_t dirtype); + +/** Return the list of entry guards, creating it if necessary. */ +const smartlist_t * +get_entry_guards(void) +{ + if (! entry_guards) + entry_guards = smartlist_new(); + return entry_guards; +} + +/** Check whether the entry guard e is usable, given the directory + * authorities' opinion about the router (stored in ri) and the user's + * configuration (in options). Set e->bad_since + * accordingly. Return true iff the entry guard's status changes. + * + * If it's not usable, set *reason to a static string explaining why. + */ +static int +entry_guard_set_status(entry_guard_t *e, const node_t *node, + time_t now, const or_options_t *options, + const char **reason) +{ + char buf[HEX_DIGEST_LEN+1]; + int changed = 0; + + *reason = NULL; + + /* Do we want to mark this guard as bad? */ + if (!node) + *reason = "unlisted"; + else if (!node->is_running) + *reason = "down"; + else if (options->UseBridges && (!node->ri || + node->ri->purpose != ROUTER_PURPOSE_BRIDGE)) + *reason = "not a bridge"; + else if (options->UseBridges && !node_is_a_configured_bridge(node)) + *reason = "not a configured bridge"; + else if (!options->UseBridges && !node->is_possible_guard && + !routerset_contains_node(options->EntryNodes,node)) + *reason = "not recommended as a guard"; + else if (routerset_contains_node(options->ExcludeNodes, node)) + *reason = "excluded"; + else if (e->path_bias_disabled) + *reason = "path-biased"; + + if (*reason && ! e->bad_since) { + /* Router is newly bad. */ + base16_encode(buf, sizeof(buf), e->identity, DIGEST_LEN); + log_info(LD_CIRC, "Entry guard %s (%s) is %s: marking as unusable.", + e->nickname, buf, *reason); + + e->bad_since = now; + control_event_guard(e->nickname, e->identity, "BAD"); + changed = 1; + } else if (!*reason && e->bad_since) { + /* There's nothing wrong with the router any more. */ + base16_encode(buf, sizeof(buf), e->identity, DIGEST_LEN); + log_info(LD_CIRC, "Entry guard %s (%s) is no longer unusable: " + "marking as ok.", e->nickname, buf); + + e->bad_since = 0; + control_event_guard(e->nickname, e->identity, "GOOD"); + changed = 1; + } + + if (node) { + int is_dir = node_is_dir(node) && node->rs && + node->rs->version_supports_microdesc_cache; + if (options->UseBridges && node_is_a_configured_bridge(node)) + is_dir = 1; + if (e->is_dir_cache != is_dir) { + e->is_dir_cache = is_dir; + changed = 1; + } + } + + return changed; +} + +/** Return true iff enough time has passed since we last tried to connect + * to the unreachable guard e that we're willing to try again. */ +static int +entry_is_time_to_retry(entry_guard_t *e, time_t now) +{ + long diff; + if (e->last_attempted < e->unreachable_since) + return 1; + diff = now - e->unreachable_since; + if (diff < 6*60*60) + return now > (e->last_attempted + 60*60); + else if (diff < 3*24*60*60) + return now > (e->last_attempted + 4*60*60); + else if (diff < 7*24*60*60) + return now > (e->last_attempted + 18*60*60); + else + return now > (e->last_attempted + 36*60*60); +} + +/** Return the node corresponding to e, if e is + * working well enough that we are willing to use it as an entry + * right now. (Else return NULL.) In particular, it must be + * - Listed as either up or never yet contacted; + * - Present in the routerlist; + * - Listed as 'stable' or 'fast' by the current dirserver consensus, + * if demanded by need_uptime or need_capacity + * (unless it's a configured EntryNode); + * - Allowed by our current ReachableORAddresses config option; and + * - Currently thought to be reachable by us (unless assume_reachable + * is true). + * + * If the answer is no, set *msg to an explanation of why. + * + * If need_descriptor is true, only return the node if we currently have + * a descriptor (routerinfo or microdesc) for it. + */ +static INLINE const node_t * +entry_is_live(entry_guard_t *e, int need_uptime, int need_capacity, + int assume_reachable, int need_descriptor, const char **msg) +{ + const node_t *node; + const or_options_t *options = get_options(); + tor_assert(msg); + + if (e->path_bias_disabled) { + *msg = "path-biased"; + return NULL; + } + if (e->bad_since) { + *msg = "bad"; + return NULL; + } + /* no good if it's unreachable, unless assume_unreachable or can_retry. */ + if (!assume_reachable && !e->can_retry && + e->unreachable_since && !entry_is_time_to_retry(e, time(NULL))) { + *msg = "unreachable"; + return NULL; + } + node = node_get_by_id(e->identity); + if (!node) { + *msg = "no node info"; + return NULL; + } + if (need_descriptor && !node_has_descriptor(node)) { + *msg = "no descriptor"; + return NULL; + } + if (get_options()->UseBridges) { + if (node_get_purpose(node) != ROUTER_PURPOSE_BRIDGE) { + *msg = "not a bridge"; + return NULL; + } + if (!node_is_a_configured_bridge(node)) { + *msg = "not a configured bridge"; + return NULL; + } + } else { /* !get_options()->UseBridges */ + if (node_get_purpose(node) != ROUTER_PURPOSE_GENERAL) { + *msg = "not general-purpose"; + return NULL; + } + } + if (routerset_contains_node(options->EntryNodes, node)) { + /* they asked for it, they get it */ + need_uptime = need_capacity = 0; + } + if (node_is_unreliable(node, need_uptime, need_capacity, 0)) { + *msg = "not fast/stable"; + return NULL; + } + if (!fascist_firewall_allows_node(node)) { + *msg = "unreachable by config"; + return NULL; + } + return node; +} + +/** Return the number of entry guards that we think are usable. */ +int +num_live_entry_guards(int for_directory) +{ + int n = 0; + const char *msg; + if (! entry_guards) + return 0; + SMARTLIST_FOREACH_BEGIN(entry_guards, entry_guard_t *, entry) { + if (for_directory && !entry->is_dir_cache) + continue; + if (entry_is_live(entry, 0, 1, 0, !for_directory, &msg)) + ++n; + } SMARTLIST_FOREACH_END(entry); + return n; +} + +/** If digest matches the identity of any node in the + * entry_guards list, return that node. Else return NULL. */ +entry_guard_t * +entry_guard_get_by_id_digest(const char *digest) +{ + SMARTLIST_FOREACH(entry_guards, entry_guard_t *, entry, + if (tor_memeq(digest, entry->identity, DIGEST_LEN)) + return entry; + ); + return NULL; +} + +/** Dump a description of our list of entry guards to the log at level + * severity. */ +static void +log_entry_guards(int severity) +{ + smartlist_t *elements = smartlist_new(); + char *s; + + SMARTLIST_FOREACH_BEGIN(entry_guards, entry_guard_t *, e) + { + const char *msg = NULL; + if (entry_is_live(e, 0, 1, 0, 0, &msg)) + smartlist_add_asprintf(elements, "%s [%s] (up %s)", + e->nickname, + hex_str(e->identity, DIGEST_LEN), + e->made_contact ? "made-contact" : "never-contacted"); + else + smartlist_add_asprintf(elements, "%s [%s] (%s, %s)", + e->nickname, + hex_str(e->identity, DIGEST_LEN), + msg, + e->made_contact ? "made-contact" : "never-contacted"); + } + SMARTLIST_FOREACH_END(e); + + s = smartlist_join_strings(elements, ",", 0, NULL); + SMARTLIST_FOREACH(elements, char*, cp, tor_free(cp)); + smartlist_free(elements); + log_fn(severity,LD_CIRC,"%s",s); + tor_free(s); +} + +/** Called when one or more guards that we would previously have used for some + * purpose are no longer in use because a higher-priority guard has become + * usable again. */ +static void +control_event_guard_deferred(void) +{ + /* XXXX We don't actually have a good way to figure out _how many_ entries + * are live for some purpose. We need an entry_is_even_slightly_live() + * function for this to work right. NumEntryGuards isn't reliable: if we + * need guards with weird properties, we can have more than that number + * live. + **/ +#if 0 + int n = 0; + const char *msg; + const or_options_t *options = get_options(); + if (!entry_guards) + return; + SMARTLIST_FOREACH(entry_guards, entry_guard_t *, entry, + { + if (entry_is_live(entry, 0, 1, 0, &msg)) { + if (n++ == options->NumEntryGuards) { + control_event_guard(entry->nickname, entry->identity, "DEFERRED"); + return; + } + } + }); +#endif +} + ++/** Largest amount that we'll backdate chosen_on_date */ ++#define CHOSEN_ON_DATE_SLOP (30*86400) ++ +/** Add a new (preferably stable and fast) router to our + * entry_guards list. Return a pointer to the router if we succeed, + * or NULL if we can't find any more suitable entries. + * + * If chosen is defined, use that one, and if it's not + * already in our entry_guards list, put it at the *beginning*. + * Else, put the one we pick at the end of the list. */ +static const node_t * +add_an_entry_guard(const node_t *chosen, int reset_status, int prepend, + int for_directory) +{ + const node_t *node; + entry_guard_t *entry; + + if (chosen) { + node = chosen; + entry = entry_guard_get_by_id_digest(node->identity); + if (entry) { + if (reset_status) { + entry->bad_since = 0; + entry->can_retry = 1; + } + entry->is_dir_cache = node->rs && + node->rs->version_supports_microdesc_cache; + if (get_options()->UseBridges && node_is_a_configured_bridge(node)) + entry->is_dir_cache = 1; + return NULL; + } + } else if (!for_directory) { + node = choose_good_entry_server(CIRCUIT_PURPOSE_C_GENERAL, NULL); + if (!node) + return NULL; + } else { + const routerstatus_t *rs; + rs = router_pick_directory_server(MICRODESC_DIRINFO|V3_DIRINFO, + PDS_PREFER_TUNNELED_DIR_CONNS_|PDS_FOR_GUARD); + if (!rs) + return NULL; + node = node_get_by_id(rs->identity_digest); + if (!node) + return NULL; + } + if (node->using_as_guard) + return NULL; + if (entry_guard_get_by_id_digest(node->identity) != NULL) { + log_info(LD_CIRC, "I was about to add a duplicate entry guard."); + /* This can happen if we choose a guard, then the node goes away, then + * comes back. */ + ((node_t*) node)->using_as_guard = 1; + return NULL; + } + entry = tor_malloc_zero(sizeof(entry_guard_t)); + log_info(LD_CIRC, "Chose %s as new entry guard.", + node_describe(node)); + strlcpy(entry->nickname, node_get_nickname(node), sizeof(entry->nickname)); + memcpy(entry->identity, node->identity, DIGEST_LEN); + entry->is_dir_cache = node_is_dir(node) && + node->rs && node->rs->version_supports_microdesc_cache; + if (get_options()->UseBridges && node_is_a_configured_bridge(node)) + entry->is_dir_cache = 1; + + /* Choose expiry time smudged over the past month. The goal here + * is to a) spread out when Tor clients rotate their guards, so they + * don't all select them on the same day, and b) avoid leaving a + * precise timestamp in the state file about when we first picked + * this guard. For details, see the Jan 2010 or-dev thread. */ + entry->chosen_on_date = time(NULL) - crypto_rand_int(3600*24*30); + entry->chosen_by_version = tor_strdup(VERSION); + ((node_t*)node)->using_as_guard = 1; + if (prepend) + smartlist_insert(entry_guards, 0, entry); + else + smartlist_add(entry_guards, entry); + control_event_guard(entry->nickname, entry->identity, "NEW"); + control_event_guard_deferred(); + log_entry_guards(LOG_INFO); + return node; +} + +/** If the use of entry guards is configured, choose more entry guards + * until we have enough in the list. */ +static void +pick_entry_guards(const or_options_t *options, int for_directory) +{ + int changed = 0; + const int num_needed = for_directory ? options->NumDirectoryGuards : + options->NumEntryGuards; + + tor_assert(entry_guards); + + while (num_live_entry_guards(for_directory) < num_needed) { + if (!add_an_entry_guard(NULL, 0, 0, for_directory)) + break; + changed = 1; + } + if (changed) + entry_guards_changed(); +} + +/** How long (in seconds) do we allow an entry guard to be nonfunctional, + * unlisted, excluded, or otherwise nonusable before we give up on it? */ +#define ENTRY_GUARD_REMOVE_AFTER (30*24*60*60) + +/** Release all storage held by e. */ +static void +entry_guard_free(entry_guard_t *e) +{ + if (!e) + return; + tor_free(e->chosen_by_version); + tor_free(e); +} + ++/** ++ * Return the minimum lifetime of working entry guard, in seconds, ++ * as given in the consensus networkstatus. (Plus CHOSEN_ON_DATE_SLOP, ++ * so that we can do the chosen_on_date randomization while achieving the ++ * desired minimum lifetime.) ++ */ ++static int32_t ++guards_get_lifetime(void) ++{ ++ const or_options_t *options = get_options(); ++#define DFLT_GUARD_LIFETIME (86400 * 30) /* One month. */ ++#define MIN_GUARD_LIFETIME (86400 * 60) /* Two months. */ ++#define MAX_GUARD_LIFETIME (86400 * 1826) /* Five years. */ ++ ++ if (options->GuardLifetime >= 1) { ++ return CLAMP(MIN_GUARD_LIFETIME, ++ options->GuardLifetime, ++ MAX_GUARD_LIFETIME) + CHOSEN_ON_DATE_SLOP; ++ } ++ ++ return networkstatus_get_param(NULL, "GuardLifetime", ++ DFLT_GUARD_LIFETIME, ++ MIN_GUARD_LIFETIME, ++ MAX_GUARD_LIFETIME) + CHOSEN_ON_DATE_SLOP; ++} ++ +/** Remove any entry guard which was selected by an unknown version of Tor, + * or which was selected by a version of Tor that's known to select + * entry guards badly, or which was selected more 2 months ago. */ +/* XXXX The "obsolete guards" and "chosen long ago guards" things should + * probably be different functions. */ +static int +remove_obsolete_entry_guards(time_t now) +{ + int changed = 0, i; ++ int32_t guard_lifetime = guards_get_lifetime(); + + for (i = 0; i < smartlist_len(entry_guards); ++i) { + entry_guard_t *entry = smartlist_get(entry_guards, i); + const char *ver = entry->chosen_by_version; + const char *msg = NULL; + tor_version_t v; + int version_is_bad = 0, date_is_bad = 0; + if (!ver) { + msg = "does not say what version of Tor it was selected by"; + version_is_bad = 1; + } else if (tor_version_parse(ver, &v)) { + msg = "does not seem to be from any recognized version of Tor"; + version_is_bad = 1; + } else { + char *tor_ver = NULL; + tor_asprintf(&tor_ver, "Tor %s", ver); + if ((tor_version_as_new_as(tor_ver, "0.1.0.10-alpha") && + !tor_version_as_new_as(tor_ver, "0.1.2.16-dev")) || + (tor_version_as_new_as(tor_ver, "0.2.0.0-alpha") && + !tor_version_as_new_as(tor_ver, "0.2.0.6-alpha")) || + /* above are bug 440; below are bug 1217 */ + (tor_version_as_new_as(tor_ver, "0.2.1.3-alpha") && + !tor_version_as_new_as(tor_ver, "0.2.1.23")) || + (tor_version_as_new_as(tor_ver, "0.2.2.0-alpha") && + !tor_version_as_new_as(tor_ver, "0.2.2.7-alpha"))) { + msg = "was selected without regard for guard bandwidth"; + version_is_bad = 1; + } + tor_free(tor_ver); + } - if (!version_is_bad && entry->chosen_on_date + 3600*24*60 < now) { - /* It's been 2 months since the date listed in our state file. */ ++ if (!version_is_bad && entry->chosen_on_date + guard_lifetime < now) { ++ /* It's been too long since the date listed in our state file. */ + msg = "was selected several months ago"; + date_is_bad = 1; + } + + if (version_is_bad || date_is_bad) { /* we need to drop it */ + char dbuf[HEX_DIGEST_LEN+1]; + tor_assert(msg); + base16_encode(dbuf, sizeof(dbuf), entry->identity, DIGEST_LEN); + log_fn(version_is_bad ? LOG_NOTICE : LOG_INFO, LD_CIRC, + "Entry guard '%s' (%s) %s. (Version=%s.) Replacing it.", + entry->nickname, dbuf, msg, ver?escaped(ver):"none"); + control_event_guard(entry->nickname, entry->identity, "DROPPED"); + entry_guard_free(entry); + smartlist_del_keeporder(entry_guards, i--); + log_entry_guards(LOG_INFO); + changed = 1; + } + } + + return changed ? 1 : 0; +} + +/** Remove all entry guards that have been down or unlisted for so + * long that we don't think they'll come up again. Return 1 if we + * removed any, or 0 if we did nothing. */ +static int +remove_dead_entry_guards(time_t now) +{ + char dbuf[HEX_DIGEST_LEN+1]; + char tbuf[ISO_TIME_LEN+1]; + int i; + int changed = 0; + + for (i = 0; i < smartlist_len(entry_guards); ) { + entry_guard_t *entry = smartlist_get(entry_guards, i); + if (entry->bad_since && + ! entry->path_bias_disabled && + entry->bad_since + ENTRY_GUARD_REMOVE_AFTER < now) { + + base16_encode(dbuf, sizeof(dbuf), entry->identity, DIGEST_LEN); + format_local_iso_time(tbuf, entry->bad_since); + log_info(LD_CIRC, "Entry guard '%s' (%s) has been down or unlisted " + "since %s local time; removing.", + entry->nickname, dbuf, tbuf); + control_event_guard(entry->nickname, entry->identity, "DROPPED"); + entry_guard_free(entry); + smartlist_del_keeporder(entry_guards, i); + log_entry_guards(LOG_INFO); + changed = 1; + } else + ++i; + } + return changed ? 1 : 0; +} + +/** A new directory or router-status has arrived; update the down/listed + * status of the entry guards. + * + * An entry is 'down' if the directory lists it as nonrunning. + * An entry is 'unlisted' if the directory doesn't include it. + * + * Don't call this on startup; only on a fresh download. Otherwise we'll + * think that things are unlisted. + */ +void +entry_guards_compute_status(const or_options_t *options, time_t now) +{ + int changed = 0; + digestmap_t *reasons; + + if (! entry_guards) + return; + + if (options->EntryNodes) /* reshuffle the entry guard list if needed */ + entry_nodes_should_be_added(); + + reasons = digestmap_new(); + SMARTLIST_FOREACH_BEGIN(entry_guards, entry_guard_t *, entry) + { + const node_t *r = node_get_by_id(entry->identity); + const char *reason = NULL; + if (entry_guard_set_status(entry, r, now, options, &reason)) + changed = 1; + + if (entry->bad_since) + tor_assert(reason); + if (reason) + digestmap_set(reasons, entry->identity, (char*)reason); + } + SMARTLIST_FOREACH_END(entry); + + if (remove_dead_entry_guards(now)) + changed = 1; + if (remove_obsolete_entry_guards(now)) + changed = 1; + + if (changed) { + SMARTLIST_FOREACH_BEGIN(entry_guards, entry_guard_t *, entry) { + const char *reason = digestmap_get(reasons, entry->identity); + const char *live_msg = ""; + const node_t *r = entry_is_live(entry, 0, 1, 0, 0, &live_msg); + log_info(LD_CIRC, "Summary: Entry %s [%s] is %s, %s%s%s, and %s%s.", + entry->nickname, + hex_str(entry->identity, DIGEST_LEN), + entry->unreachable_since ? "unreachable" : "reachable", + entry->bad_since ? "unusable" : "usable", + reason ? ", ": "", + reason ? reason : "", + r ? "live" : "not live / ", + r ? "" : live_msg); + } SMARTLIST_FOREACH_END(entry); + log_info(LD_CIRC, " (%d/%d entry guards are usable/new)", + num_live_entry_guards(0), smartlist_len(entry_guards)); + log_entry_guards(LOG_INFO); + entry_guards_changed(); + } + + digestmap_free(reasons, NULL); +} + +/** Called when a connection to an OR with the identity digest digest + * is established (succeeded==1) or has failed (succeeded==0). + * If the OR is an entry, change that entry's up/down status. + * Return 0 normally, or -1 if we want to tear down the new connection. + * + * If mark_relay_status, also call router_set_status() on this + * relay. + * + * XXX024 change succeeded and mark_relay_status into 'int flags'. + */ +int +entry_guard_register_connect_status(const char *digest, int succeeded, + int mark_relay_status, time_t now) +{ + int changed = 0; + int refuse_conn = 0; + int first_contact = 0; + entry_guard_t *entry = NULL; + int idx = -1; + char buf[HEX_DIGEST_LEN+1]; + + if (! entry_guards) + return 0; + + SMARTLIST_FOREACH_BEGIN(entry_guards, entry_guard_t *, e) { + tor_assert(e); + if (tor_memeq(e->identity, digest, DIGEST_LEN)) { + entry = e; + idx = e_sl_idx; + break; + } + } SMARTLIST_FOREACH_END(e); + + if (!entry) + return 0; + + base16_encode(buf, sizeof(buf), entry->identity, DIGEST_LEN); + + if (succeeded) { + if (entry->unreachable_since) { + log_info(LD_CIRC, "Entry guard '%s' (%s) is now reachable again. Good.", + entry->nickname, buf); + entry->can_retry = 0; + entry->unreachable_since = 0; + entry->last_attempted = now; + control_event_guard(entry->nickname, entry->identity, "UP"); + changed = 1; + } + if (!entry->made_contact) { + entry->made_contact = 1; + first_contact = changed = 1; + } + } else { /* ! succeeded */ + if (!entry->made_contact) { + /* We've never connected to this one. */ + log_info(LD_CIRC, + "Connection to never-contacted entry guard '%s' (%s) failed. " + "Removing from the list. %d/%d entry guards usable/new.", + entry->nickname, buf, + num_live_entry_guards(0)-1, smartlist_len(entry_guards)-1); + control_event_guard(entry->nickname, entry->identity, "DROPPED"); + entry_guard_free(entry); + smartlist_del_keeporder(entry_guards, idx); + log_entry_guards(LOG_INFO); + changed = 1; + } else if (!entry->unreachable_since) { + log_info(LD_CIRC, "Unable to connect to entry guard '%s' (%s). " + "Marking as unreachable.", entry->nickname, buf); + entry->unreachable_since = entry->last_attempted = now; + control_event_guard(entry->nickname, entry->identity, "DOWN"); + changed = 1; + entry->can_retry = 0; /* We gave it an early chance; no good. */ + } else { + char tbuf[ISO_TIME_LEN+1]; + format_iso_time(tbuf, entry->unreachable_since); + log_debug(LD_CIRC, "Failed to connect to unreachable entry guard " + "'%s' (%s). It has been unreachable since %s.", + entry->nickname, buf, tbuf); + entry->last_attempted = now; + entry->can_retry = 0; /* We gave it an early chance; no good. */ + } + } + + /* if the caller asked us to, also update the is_running flags for this + * relay */ + if (mark_relay_status) + router_set_status(digest, succeeded); + + if (first_contact) { + /* We've just added a new long-term entry guard. Perhaps the network just + * came back? We should give our earlier entries another try too, + * and close this connection so we don't use it before we've given + * the others a shot. */ + SMARTLIST_FOREACH_BEGIN(entry_guards, entry_guard_t *, e) { + if (e == entry) + break; + if (e->made_contact) { + const char *msg; + const node_t *r = entry_is_live(e, 0, 1, 1, 0, &msg); + if (r && e->unreachable_since) { + refuse_conn = 1; + e->can_retry = 1; + } + } + } SMARTLIST_FOREACH_END(e); + if (refuse_conn) { + log_info(LD_CIRC, + "Connected to new entry guard '%s' (%s). Marking earlier " + "entry guards up. %d/%d entry guards usable/new.", + entry->nickname, buf, + num_live_entry_guards(0), smartlist_len(entry_guards)); + log_entry_guards(LOG_INFO); + changed = 1; + } + } + + if (changed) + entry_guards_changed(); + return refuse_conn ? -1 : 0; +} + +/** When we try to choose an entry guard, should we parse and add + * config's EntryNodes first? */ +static int should_add_entry_nodes = 0; + +/** Called when the value of EntryNodes changes in our configuration. */ +void +entry_nodes_should_be_added(void) +{ + log_info(LD_CIRC, "EntryNodes config option set. Putting configured " + "relays at the front of the entry guard list."); + should_add_entry_nodes = 1; +} + +/** Update the using_as_guard fields of all the nodes. We do this after we + * remove entry guards from the list: This is the only function that clears + * the using_as_guard field. */ +static void +update_node_guard_status(void) +{ + smartlist_t *nodes = nodelist_get_list(); + SMARTLIST_FOREACH(nodes, node_t *, node, node->using_as_guard = 0); + SMARTLIST_FOREACH_BEGIN(entry_guards, entry_guard_t *, entry) { + node_t *node = node_get_mutable_by_id(entry->identity); + if (node) + node->using_as_guard = 1; + } SMARTLIST_FOREACH_END(entry); +} + +/** Adjust the entry guards list so that it only contains entries from + * EntryNodes, adding new entries from EntryNodes to the list as needed. */ +static void +entry_guards_set_from_config(const or_options_t *options) +{ + smartlist_t *entry_nodes, *worse_entry_nodes, *entry_fps; + smartlist_t *old_entry_guards_on_list, *old_entry_guards_not_on_list; + tor_assert(entry_guards); + + should_add_entry_nodes = 0; + + if (!options->EntryNodes) { + /* It's possible that a controller set EntryNodes, thus making + * should_add_entry_nodes set, then cleared it again, all before the + * call to choose_random_entry() that triggered us. If so, just return. + */ + return; + } + + { + char *string = routerset_to_string(options->EntryNodes); + log_info(LD_CIRC,"Adding configured EntryNodes '%s'.", string); + tor_free(string); + } + + entry_nodes = smartlist_new(); + worse_entry_nodes = smartlist_new(); + entry_fps = smartlist_new(); + old_entry_guards_on_list = smartlist_new(); + old_entry_guards_not_on_list = smartlist_new(); + + /* Split entry guards into those on the list and those not. */ + + routerset_get_all_nodes(entry_nodes, options->EntryNodes, + options->ExcludeNodes, 0); + SMARTLIST_FOREACH(entry_nodes, const node_t *,node, + smartlist_add(entry_fps, (void*)node->identity)); + + SMARTLIST_FOREACH(entry_guards, entry_guard_t *, e, { + if (smartlist_contains_digest(entry_fps, e->identity)) + smartlist_add(old_entry_guards_on_list, e); + else + smartlist_add(old_entry_guards_not_on_list, e); + }); + + /* Remove all currently configured guard nodes, excluded nodes, unreachable + * nodes, or non-Guard nodes from entry_nodes. */ + SMARTLIST_FOREACH_BEGIN(entry_nodes, const node_t *, node) { + if (entry_guard_get_by_id_digest(node->identity)) { + SMARTLIST_DEL_CURRENT(entry_nodes, node); + continue; + } else if (routerset_contains_node(options->ExcludeNodes, node)) { + SMARTLIST_DEL_CURRENT(entry_nodes, node); + continue; + } else if (!fascist_firewall_allows_node(node)) { + SMARTLIST_DEL_CURRENT(entry_nodes, node); + continue; + } else if (! node->is_possible_guard) { + smartlist_add(worse_entry_nodes, (node_t*)node); + SMARTLIST_DEL_CURRENT(entry_nodes, node); + } + } SMARTLIST_FOREACH_END(node); + + /* Now build the new entry_guards list. */ + smartlist_clear(entry_guards); + /* First, the previously configured guards that are in EntryNodes. */ + smartlist_add_all(entry_guards, old_entry_guards_on_list); + /* Next, scramble the rest of EntryNodes, putting the guards first. */ + smartlist_shuffle(entry_nodes); + smartlist_shuffle(worse_entry_nodes); + smartlist_add_all(entry_nodes, worse_entry_nodes); + + /* Next, the rest of EntryNodes */ + SMARTLIST_FOREACH_BEGIN(entry_nodes, const node_t *, node) { + add_an_entry_guard(node, 0, 0, 0); + if (smartlist_len(entry_guards) > options->NumEntryGuards * 10) + break; + } SMARTLIST_FOREACH_END(node); + log_notice(LD_GENERAL, "%d entries in guards", smartlist_len(entry_guards)); + /* Finally, free the remaining previously configured guards that are not in + * EntryNodes. */ + SMARTLIST_FOREACH(old_entry_guards_not_on_list, entry_guard_t *, e, + entry_guard_free(e)); + + update_node_guard_status(); + + smartlist_free(entry_nodes); + smartlist_free(worse_entry_nodes); + smartlist_free(entry_fps); + smartlist_free(old_entry_guards_on_list); + smartlist_free(old_entry_guards_not_on_list); + entry_guards_changed(); +} + +/** Return 0 if we're fine adding arbitrary routers out of the + * directory to our entry guard list, or return 1 if we have a + * list already and we must stick to it. + */ +int +entry_list_is_constrained(const or_options_t *options) +{ + if (options->EntryNodes) + return 1; + if (options->UseBridges) + return 1; + return 0; +} + +/** Return true iff this node can answer directory questions about + * microdescriptors. */ +static int +node_understands_microdescriptors(const node_t *node) +{ + tor_assert(node); + if (node->rs && node->rs->version_supports_microdesc_cache) + return 1; + if (node->ri && tor_version_supports_microdescriptors(node->ri->platform)) + return 1; + return 0; +} + +/** Return true iff node is able to answer directory questions + * of type dirinfo. */ +static int +node_can_handle_dirinfo(const node_t *node, dirinfo_type_t dirinfo) +{ + /* Checking dirinfo for any type other than microdescriptors isn't required + yet, since we only choose directory guards that can support microdescs, + routerinfos, and networkstatuses, AND we don't use directory guards if + we're configured to do direct downloads of anything else. The only case + where we might have a guard that doesn't know about a type of directory + information is when we're retrieving directory information from a + bridge. */ + + if ((dirinfo & MICRODESC_DIRINFO) && + !node_understands_microdescriptors(node)) + return 0; + return 1; +} + +/** Pick a live (up and listed) entry guard from entry_guards. If + * state is non-NULL, this is for a specific circuit -- + * make sure not to pick this circuit's exit or any node in the + * exit's family. If state is NULL, we're looking for a random + * guard (likely a bridge). If dirinfo is not NO_DIRINFO, then + * only select from nodes that know how to answer directory questions + * of that type. */ +const node_t * +choose_random_entry(cpath_build_state_t *state) +{ + return choose_random_entry_impl(state, 0, 0); +} + +/** Pick a live (up and listed) directory guard from entry_guards for + * downloading information of type type. */ +const node_t * +choose_random_dirguard(dirinfo_type_t type) +{ + return choose_random_entry_impl(NULL, 1, type); +} + +/** Helper for choose_random{entry,dirguard}. */ +static const node_t * +choose_random_entry_impl(cpath_build_state_t *state, int for_directory, + dirinfo_type_t dirinfo_type) +{ + const or_options_t *options = get_options(); + smartlist_t *live_entry_guards = smartlist_new(); + smartlist_t *exit_family = smartlist_new(); + const node_t *chosen_exit = + state?build_state_get_exit_node(state) : NULL; + const node_t *node = NULL; + int need_uptime = state ? state->need_uptime : 0; + int need_capacity = state ? state->need_capacity : 0; + int preferred_min, consider_exit_family = 0; + int need_descriptor = !for_directory; + const int num_needed = for_directory ? options->NumDirectoryGuards : + options->NumEntryGuards; + + if (chosen_exit) { + nodelist_add_node_and_family(exit_family, chosen_exit); + consider_exit_family = 1; + } + + if (!entry_guards) + entry_guards = smartlist_new(); + + if (should_add_entry_nodes) + entry_guards_set_from_config(options); + + if (!entry_list_is_constrained(options) && + smartlist_len(entry_guards) < num_needed) + pick_entry_guards(options, for_directory); + + retry: + smartlist_clear(live_entry_guards); + SMARTLIST_FOREACH_BEGIN(entry_guards, entry_guard_t *, entry) { + const char *msg; + node = entry_is_live(entry, need_uptime, need_capacity, 0, + need_descriptor, &msg); + if (!node) + continue; /* down, no point */ + if (for_directory) { + if (!entry->is_dir_cache) + continue; /* We need a directory and didn't get one. */ + } + if (node == chosen_exit) + continue; /* don't pick the same node for entry and exit */ + if (consider_exit_family && smartlist_contains(exit_family, node)) + continue; /* avoid relays that are family members of our exit */ + if (dirinfo_type != NO_DIRINFO && + !node_can_handle_dirinfo(node, dirinfo_type)) + continue; /* this node won't be able to answer our dir questions */ +#if 0 /* since EntryNodes is always strict now, this clause is moot */ + if (options->EntryNodes && + !routerset_contains_node(options->EntryNodes, node)) { + /* We've come to the end of our preferred entry nodes. */ + if (smartlist_len(live_entry_guards)) + goto choose_and_finish; /* only choose from the ones we like */ + if (options->StrictNodes) { + /* in theory this case should never happen, since + * entry_guards_set_from_config() drops unwanted relays */ + tor_fragile_assert(); + } else { + log_info(LD_CIRC, + "No relays from EntryNodes available. Using others."); + } + } +#endif + smartlist_add(live_entry_guards, (void*)node); + if (!entry->made_contact) { + /* Always start with the first not-yet-contacted entry + * guard. Otherwise we might add several new ones, pick + * the second new one, and now we've expanded our entry + * guard list without needing to. */ + goto choose_and_finish; + } + if (smartlist_len(live_entry_guards) >= num_needed) + goto choose_and_finish; /* we have enough */ + } SMARTLIST_FOREACH_END(entry); + + if (entry_list_is_constrained(options)) { + /* If we prefer the entry nodes we've got, and we have at least + * one choice, that's great. Use it. */ + preferred_min = 1; + } else { + /* Try to have at least 2 choices available. This way we don't + * get stuck with a single live-but-crummy entry and just keep + * using him. + * (We might get 2 live-but-crummy entry guards, but so be it.) */ + preferred_min = 2; + } + + if (smartlist_len(live_entry_guards) < preferred_min) { + if (!entry_list_is_constrained(options)) { + /* still no? try adding a new entry then */ + /* XXX if guard doesn't imply fast and stable, then we need + * to tell add_an_entry_guard below what we want, or it might + * be a long time til we get it. -RD */ + node = add_an_entry_guard(NULL, 0, 0, for_directory); + if (node) { + entry_guards_changed(); + /* XXX we start over here in case the new node we added shares + * a family with our exit node. There's a chance that we'll just + * load up on entry guards here, if the network we're using is + * one big family. Perhaps we should teach add_an_entry_guard() + * to understand nodes-to-avoid-if-possible? -RD */ + goto retry; + } + } + if (!node && need_uptime) { + need_uptime = 0; /* try without that requirement */ + goto retry; + } + if (!node && need_capacity) { + /* still no? last attempt, try without requiring capacity */ + need_capacity = 0; + goto retry; + } +#if 0 + /* Removing this retry logic: if we only allow one exit, and it is in the + same family as all our entries, then we are just plain not going to win + here. */ + if (!node && entry_list_is_constrained(options) && consider_exit_family) { + /* still no? if we're using bridges or have strictentrynodes + * set, and our chosen exit is in the same family as all our + * bridges/entry guards, then be flexible about families. */ + consider_exit_family = 0; + goto retry; + } +#endif + /* live_entry_guards may be empty below. Oh well, we tried. */ + } + + choose_and_finish: + if (entry_list_is_constrained(options)) { + /* We need to weight by bandwidth, because our bridges or entryguards + * were not already selected proportional to their bandwidth. */ + node = node_sl_choose_by_bandwidth(live_entry_guards, WEIGHT_FOR_GUARD); + } else { + /* We choose uniformly at random here, because choose_good_entry_server() + * already weights its choices by bandwidth, so we don't want to + * *double*-weight our guard selection. */ + node = smartlist_choose(live_entry_guards); + } + smartlist_free(live_entry_guards); + smartlist_free(exit_family); + return node; +} + +/** Parse state and learn about the entry guards it describes. + * If set is true, and there are no errors, replace the global + * entry_list with what we find. + * On success, return 0. On failure, alloc into *msg a string + * describing the error, and return -1. + */ +int +entry_guards_parse_state(or_state_t *state, int set, char **msg) +{ + entry_guard_t *node = NULL; + smartlist_t *new_entry_guards = smartlist_new(); + config_line_t *line; + time_t now = time(NULL); + const char *state_version = state->TorVersion; + digestmap_t *added_by = digestmap_new(); + + *msg = NULL; + for (line = state->EntryGuards; line; line = line->next) { + if (!strcasecmp(line->key, "EntryGuard")) { + smartlist_t *args = smartlist_new(); + node = tor_malloc_zero(sizeof(entry_guard_t)); + /* all entry guards on disk have been contacted */ + node->made_contact = 1; + smartlist_add(new_entry_guards, node); + smartlist_split_string(args, line->value, " ", + SPLIT_SKIP_SPACE|SPLIT_IGNORE_BLANK, 0); + if (smartlist_len(args)<2) { + *msg = tor_strdup("Unable to parse entry nodes: " + "Too few arguments to EntryGuard"); + } else if (!is_legal_nickname(smartlist_get(args,0))) { + *msg = tor_strdup("Unable to parse entry nodes: " + "Bad nickname for EntryGuard"); + } else { + strlcpy(node->nickname, smartlist_get(args,0), MAX_NICKNAME_LEN+1); + if (base16_decode(node->identity, DIGEST_LEN, smartlist_get(args,1), + strlen(smartlist_get(args,1)))<0) { + *msg = tor_strdup("Unable to parse entry nodes: " + "Bad hex digest for EntryGuard"); + } + } + if (smartlist_len(args) >= 3) { + const char *is_cache = smartlist_get(args, 2); + if (!strcasecmp(is_cache, "DirCache")) { + node->is_dir_cache = 1; + } else if (!strcasecmp(is_cache, "NoDirCache")) { + node->is_dir_cache = 0; + } else { + log_warn(LD_CONFIG, "Bogus third argument to EntryGuard line: %s", + escaped(is_cache)); + } + } + SMARTLIST_FOREACH(args, char*, cp, tor_free(cp)); + smartlist_free(args); + if (*msg) + break; + } else if (!strcasecmp(line->key, "EntryGuardDownSince") || + !strcasecmp(line->key, "EntryGuardUnlistedSince")) { + time_t when; + time_t last_try = 0; + if (!node) { + *msg = tor_strdup("Unable to parse entry nodes: " + "EntryGuardDownSince/UnlistedSince without EntryGuard"); + break; + } + if (parse_iso_time(line->value, &when)<0) { + *msg = tor_strdup("Unable to parse entry nodes: " + "Bad time in EntryGuardDownSince/UnlistedSince"); + break; + } + if (when > now) { + /* It's a bad idea to believe info in the future: you can wind + * up with timeouts that aren't allowed to happen for years. */ + continue; + } + if (strlen(line->value) >= ISO_TIME_LEN+ISO_TIME_LEN+1) { + /* ignore failure */ + (void) parse_iso_time(line->value+ISO_TIME_LEN+1, &last_try); + } + if (!strcasecmp(line->key, "EntryGuardDownSince")) { + node->unreachable_since = when; + node->last_attempted = last_try; + } else { + node->bad_since = when; + } + } else if (!strcasecmp(line->key, "EntryGuardAddedBy")) { + char d[DIGEST_LEN]; + /* format is digest version date */ + if (strlen(line->value) < HEX_DIGEST_LEN+1+1+1+ISO_TIME_LEN) { + log_warn(LD_BUG, "EntryGuardAddedBy line is not long enough."); + continue; + } + if (base16_decode(d, sizeof(d), line->value, HEX_DIGEST_LEN)<0 || + line->value[HEX_DIGEST_LEN] != ' ') { + log_warn(LD_BUG, "EntryGuardAddedBy line %s does not begin with " + "hex digest", escaped(line->value)); + continue; + } + digestmap_set(added_by, d, tor_strdup(line->value+HEX_DIGEST_LEN+1)); + } else if (!strcasecmp(line->key, "EntryGuardPathUseBias")) { + const or_options_t *options = get_options(); + double use_cnt, success_cnt; + + if (!node) { + *msg = tor_strdup("Unable to parse entry nodes: " + "EntryGuardPathUseBias without EntryGuard"); + break; + } + + if (tor_sscanf(line->value, "%lf %lf", + &use_cnt, &success_cnt) != 2) { + log_info(LD_GENERAL, "Malformed path use bias line for node %s", + node->nickname); + continue; + } + + node->use_attempts = use_cnt; + node->use_successes = success_cnt; + + log_info(LD_GENERAL, "Read %f/%f path use bias for node %s", + node->use_successes, node->use_attempts, node->nickname); + + /* Note: We rely on the < comparison here to allow us to set a 0 + * rate and disable the feature entirely. If refactoring, don't + * change to <= */ + if (pathbias_get_use_success_count(node)/node->use_attempts + < pathbias_get_extreme_use_rate(options) && + pathbias_get_dropguards(options)) { + node->path_bias_disabled = 1; + log_info(LD_GENERAL, + "Path use bias is too high (%f/%f); disabling node %s", + node->circ_successes, node->circ_attempts, node->nickname); + } + } else if (!strcasecmp(line->key, "EntryGuardPathBias")) { + const or_options_t *options = get_options(); + double hop_cnt, success_cnt, timeouts, collapsed, successful_closed, + unusable; + + if (!node) { + *msg = tor_strdup("Unable to parse entry nodes: " + "EntryGuardPathBias without EntryGuard"); + break; + } + + /* First try 3 params, then 2. */ + /* In the long run: circuit_success ~= successful_circuit_close + + * collapsed_circuits + + * unusable_circuits */ + if (tor_sscanf(line->value, "%lf %lf %lf %lf %lf %lf", + &hop_cnt, &success_cnt, &successful_closed, + &collapsed, &unusable, &timeouts) != 6) { + int old_success, old_hops; + if (tor_sscanf(line->value, "%u %u", &old_success, &old_hops) != 2) { + continue; + } + log_info(LD_GENERAL, "Reading old-style EntryGuardPathBias %s", + escaped(line->value)); + + success_cnt = old_success; + successful_closed = old_success; + hop_cnt = old_hops; + timeouts = 0; + collapsed = 0; + unusable = 0; + } + + node->circ_attempts = hop_cnt; + node->circ_successes = success_cnt; + + node->successful_circuits_closed = successful_closed; + node->timeouts = timeouts; + node->collapsed_circuits = collapsed; + node->unusable_circuits = unusable; + + log_info(LD_GENERAL, "Read %f/%f path bias for node %s", + node->circ_successes, node->circ_attempts, node->nickname); + /* Note: We rely on the < comparison here to allow us to set a 0 + * rate and disable the feature entirely. If refactoring, don't + * change to <= */ + if (pathbias_get_close_success_count(node)/node->circ_attempts + < pathbias_get_extreme_rate(options) && + pathbias_get_dropguards(options)) { + node->path_bias_disabled = 1; + log_info(LD_GENERAL, + "Path bias is too high (%f/%f); disabling node %s", + node->circ_successes, node->circ_attempts, node->nickname); + } + + } else { + log_warn(LD_BUG, "Unexpected key %s", line->key); + } + } + + SMARTLIST_FOREACH_BEGIN(new_entry_guards, entry_guard_t *, e) { + char *sp; + char *val = digestmap_get(added_by, e->identity); + if (val && (sp = strchr(val, ' '))) { + time_t when; + *sp++ = '\0'; + if (parse_iso_time(sp, &when)<0) { + log_warn(LD_BUG, "Can't read time %s in EntryGuardAddedBy", sp); + } else { + e->chosen_by_version = tor_strdup(val); + e->chosen_on_date = when; + } + } else { + if (state_version) { + e->chosen_by_version = tor_strdup(state_version); + e->chosen_on_date = time(NULL) - crypto_rand_int(3600*24*30); + } + } + if (e->path_bias_disabled && !e->bad_since) + e->bad_since = time(NULL); + } + SMARTLIST_FOREACH_END(e); + + if (*msg || !set) { + SMARTLIST_FOREACH(new_entry_guards, entry_guard_t *, e, + entry_guard_free(e)); + smartlist_free(new_entry_guards); + } else { /* !err && set */ + if (entry_guards) { + SMARTLIST_FOREACH(entry_guards, entry_guard_t *, e, + entry_guard_free(e)); + smartlist_free(entry_guards); + } + entry_guards = new_entry_guards; + entry_guards_dirty = 0; + /* XXX024 hand new_entry_guards to this func, and move it up a + * few lines, so we don't have to re-dirty it */ + if (remove_obsolete_entry_guards(now)) + entry_guards_dirty = 1; + + update_node_guard_status(); + } + digestmap_free(added_by, tor_free_); + return *msg ? -1 : 0; +} + +/** Our list of entry guards has changed, or some element of one + * of our entry guards has changed. Write the changes to disk within + * the next few minutes. + */ +void +entry_guards_changed(void) +{ + time_t when; + entry_guards_dirty = 1; + + /* or_state_save() will call entry_guards_update_state(). */ + when = get_options()->AvoidDiskWrites ? time(NULL) + 3600 : time(NULL)+600; + or_state_mark_dirty(get_or_state(), when); +} + +/** If the entry guard info has not changed, do nothing and return. + * Otherwise, free the EntryGuards piece of state and create + * a new one out of the global entry_guards list, and then mark + * state dirty so it will get saved to disk. + */ +void +entry_guards_update_state(or_state_t *state) +{ + config_line_t **next, *line; + if (! entry_guards_dirty) + return; + + config_free_lines(state->EntryGuards); + next = &state->EntryGuards; + *next = NULL; + if (!entry_guards) + entry_guards = smartlist_new(); + SMARTLIST_FOREACH_BEGIN(entry_guards, entry_guard_t *, e) { + char dbuf[HEX_DIGEST_LEN+1]; + if (!e->made_contact) + continue; /* don't write this one to disk */ + *next = line = tor_malloc_zero(sizeof(config_line_t)); + line->key = tor_strdup("EntryGuard"); + base16_encode(dbuf, sizeof(dbuf), e->identity, DIGEST_LEN); + tor_asprintf(&line->value, "%s %s %sDirCache", e->nickname, dbuf, + e->is_dir_cache ? "" : "No"); + next = &(line->next); + if (e->unreachable_since) { + *next = line = tor_malloc_zero(sizeof(config_line_t)); + line->key = tor_strdup("EntryGuardDownSince"); + line->value = tor_malloc(ISO_TIME_LEN+1+ISO_TIME_LEN+1); + format_iso_time(line->value, e->unreachable_since); + if (e->last_attempted) { + line->value[ISO_TIME_LEN] = ' '; + format_iso_time(line->value+ISO_TIME_LEN+1, e->last_attempted); + } + next = &(line->next); + } + if (e->bad_since) { + *next = line = tor_malloc_zero(sizeof(config_line_t)); + line->key = tor_strdup("EntryGuardUnlistedSince"); + line->value = tor_malloc(ISO_TIME_LEN+1); + format_iso_time(line->value, e->bad_since); + next = &(line->next); + } + if (e->chosen_on_date && e->chosen_by_version && + !strchr(e->chosen_by_version, ' ')) { + char d[HEX_DIGEST_LEN+1]; + char t[ISO_TIME_LEN+1]; + *next = line = tor_malloc_zero(sizeof(config_line_t)); + line->key = tor_strdup("EntryGuardAddedBy"); + base16_encode(d, sizeof(d), e->identity, DIGEST_LEN); + format_iso_time(t, e->chosen_on_date); + tor_asprintf(&line->value, "%s %s %s", + d, e->chosen_by_version, t); + next = &(line->next); + } + if (e->circ_attempts > 0) { + *next = line = tor_malloc_zero(sizeof(config_line_t)); + line->key = tor_strdup("EntryGuardPathBias"); + /* In the long run: circuit_success ~= successful_circuit_close + + * collapsed_circuits + + * unusable_circuits */ + tor_asprintf(&line->value, "%f %f %f %f %f %f", + e->circ_attempts, e->circ_successes, + pathbias_get_close_success_count(e), + e->collapsed_circuits, + e->unusable_circuits, e->timeouts); + next = &(line->next); + } + if (e->use_attempts > 0) { + *next = line = tor_malloc_zero(sizeof(config_line_t)); + line->key = tor_strdup("EntryGuardPathUseBias"); + + tor_asprintf(&line->value, "%f %f", + e->use_attempts, + pathbias_get_use_success_count(e)); + next = &(line->next); + } + + } SMARTLIST_FOREACH_END(e); + if (!get_options()->AvoidDiskWrites) + or_state_mark_dirty(get_or_state(), 0); + entry_guards_dirty = 0; +} + +/** If question is the string "entry-guards", then dump + * to *answer a newly allocated string describing all of + * the nodes in the global entry_guards list. See control-spec.txt + * for details. + * For backward compatibility, we also handle the string "helper-nodes". + * */ +int +getinfo_helper_entry_guards(control_connection_t *conn, + const char *question, char **answer, + const char **errmsg) +{ + (void) conn; + (void) errmsg; + + if (!strcmp(question,"entry-guards") || + !strcmp(question,"helper-nodes")) { + smartlist_t *sl = smartlist_new(); + char tbuf[ISO_TIME_LEN+1]; + char nbuf[MAX_VERBOSE_NICKNAME_LEN+1]; + if (!entry_guards) + entry_guards = smartlist_new(); + SMARTLIST_FOREACH_BEGIN(entry_guards, entry_guard_t *, e) { + const char *status = NULL; + time_t when = 0; + const node_t *node; + + if (!e->made_contact) { + status = "never-connected"; + } else if (e->bad_since) { + when = e->bad_since; + status = "unusable"; + } else { + status = "up"; + } + + node = node_get_by_id(e->identity); + if (node) { + node_get_verbose_nickname(node, nbuf); + } else { + nbuf[0] = '$'; + base16_encode(nbuf+1, sizeof(nbuf)-1, e->identity, DIGEST_LEN); + /* e->nickname field is not very reliable if we don't know about + * this router any longer; don't include it. */ + } + + if (when) { + format_iso_time(tbuf, when); + smartlist_add_asprintf(sl, "%s %s %s\n", nbuf, status, tbuf); + } else { + smartlist_add_asprintf(sl, "%s %s\n", nbuf, status); + } + } SMARTLIST_FOREACH_END(e); + *answer = smartlist_join_strings(sl, "", 0, NULL); + SMARTLIST_FOREACH(sl, char *, c, tor_free(c)); + smartlist_free(sl); + } + return 0; +} + +/** A list of configured bridges. Whenever we actually get a descriptor + * for one, we add it as an entry guard. Note that the order of bridges + * in this list does not necessarily correspond to the order of bridges + * in the torrc. */ +static smartlist_t *bridge_list = NULL; + +/** Mark every entry of the bridge list to be removed on our next call to + * sweep_bridge_list unless it has first been un-marked. */ +void +mark_bridge_list(void) +{ + if (!bridge_list) + bridge_list = smartlist_new(); + SMARTLIST_FOREACH(bridge_list, bridge_info_t *, b, + b->marked_for_removal = 1); +} + +/** Remove every entry of the bridge list that was marked with + * mark_bridge_list if it has not subsequently been un-marked. */ +void +sweep_bridge_list(void) +{ + if (!bridge_list) + bridge_list = smartlist_new(); + SMARTLIST_FOREACH_BEGIN(bridge_list, bridge_info_t *, b) { + if (b->marked_for_removal) { + SMARTLIST_DEL_CURRENT(bridge_list, b); + bridge_free(b); + } + } SMARTLIST_FOREACH_END(b); +} + +/** Initialize the bridge list to empty, creating it if needed. */ +static void +clear_bridge_list(void) +{ + if (!bridge_list) + bridge_list = smartlist_new(); + SMARTLIST_FOREACH(bridge_list, bridge_info_t *, b, bridge_free(b)); + smartlist_clear(bridge_list); +} + +/** Free the bridge bridge. */ +static void +bridge_free(bridge_info_t *bridge) +{ + if (!bridge) + return; + + tor_free(bridge->transport_name); + tor_free(bridge); +} + +/** If we have a bridge configured whose digest matches digest, or a + * bridge with no known digest whose address matches any of the + * tor_addr_port_t's in orports, return that bridge. Else return + * NULL. */ +static bridge_info_t * +get_configured_bridge_by_orports_digest(const char *digest, + const smartlist_t *orports) +{ + if (!bridge_list) + return NULL; + SMARTLIST_FOREACH_BEGIN(bridge_list, bridge_info_t *, bridge) + { + if (tor_digest_is_zero(bridge->identity)) { + SMARTLIST_FOREACH_BEGIN(orports, tor_addr_port_t *, ap) + { + if (tor_addr_compare(&bridge->addr, &ap->addr, CMP_EXACT) == 0 && + bridge->port == ap->port) + return bridge; + } + SMARTLIST_FOREACH_END(ap); + } + if (digest && tor_memeq(bridge->identity, digest, DIGEST_LEN)) + return bridge; + } + SMARTLIST_FOREACH_END(bridge); + return NULL; +} + +/** If we have a bridge configured whose digest matches digest, or a + * bridge with no known digest whose address matches addr:/port, + * return that bridge. Else return NULL. */ +static bridge_info_t * +get_configured_bridge_by_addr_port_digest(const tor_addr_t *addr, + uint16_t port, + const char *digest) +{ + if (!bridge_list) + return NULL; + SMARTLIST_FOREACH_BEGIN(bridge_list, bridge_info_t *, bridge) + { + if (tor_digest_is_zero(bridge->identity) && + !tor_addr_compare(&bridge->addr, addr, CMP_EXACT) && + bridge->port == port) + return bridge; + if (digest && tor_memeq(bridge->identity, digest, DIGEST_LEN)) + return bridge; + } + SMARTLIST_FOREACH_END(bridge); + return NULL; +} + +/** Wrapper around get_configured_bridge_by_addr_port_digest() to look + * it up via router descriptor ri. */ +static bridge_info_t * +get_configured_bridge_by_routerinfo(const routerinfo_t *ri) +{ + bridge_info_t *bi = NULL; + smartlist_t *orports = router_get_all_orports(ri); + bi = get_configured_bridge_by_orports_digest(ri->cache_info.identity_digest, + orports); + SMARTLIST_FOREACH(orports, tor_addr_port_t *, p, tor_free(p)); + smartlist_free(orports); + return bi; +} + +/** Return 1 if ri is one of our known bridges, else 0. */ +int +routerinfo_is_a_configured_bridge(const routerinfo_t *ri) +{ + return get_configured_bridge_by_routerinfo(ri) ? 1 : 0; +} + +/** Return 1 if node is one of our configured bridges, else 0. */ +int +node_is_a_configured_bridge(const node_t *node) +{ + int retval = 0; + smartlist_t *orports = node_get_all_orports(node); + retval = get_configured_bridge_by_orports_digest(node->identity, + orports) != NULL; + SMARTLIST_FOREACH(orports, tor_addr_port_t *, p, tor_free(p)); + smartlist_free(orports); + return retval; +} + +/** We made a connection to a router at addr:port + * without knowing its digest. Its digest turned out to be digest. + * If it was a bridge, and we still don't know its digest, record it. + */ +void +learned_router_identity(const tor_addr_t *addr, uint16_t port, + const char *digest) +{ + bridge_info_t *bridge = + get_configured_bridge_by_addr_port_digest(addr, port, digest); + if (bridge && tor_digest_is_zero(bridge->identity)) { + char *transport_info = NULL; + const char *transport_name = + find_transport_name_by_bridge_addrport(addr, port); + if (transport_name) + tor_asprintf(&transport_info, " (with transport '%s')", transport_name); + + memcpy(bridge->identity, digest, DIGEST_LEN); + log_notice(LD_DIR, "Learned fingerprint %s for bridge %s%s.", + hex_str(digest, DIGEST_LEN), fmt_addrport(addr, port), + transport_info ? transport_info : ""); + tor_free(transport_info); + } +} + +/** Return true if bridge has the same identity digest as + * digest. If digest is NULL, it matches + * bridges with unspecified identity digests. */ +static int +bridge_has_digest(const bridge_info_t *bridge, const char *digest) +{ + if (digest) + return tor_memeq(digest, bridge->identity, DIGEST_LEN); + else + return tor_digest_is_zero(bridge->identity); +} + +/** We are about to add a new bridge at addr:port, with optional + * digest and transport_name. Mark for removal any previously + * existing bridge with the same address and port, and warn the user as + * appropriate. + */ +static void +bridge_resolve_conflicts(const tor_addr_t *addr, uint16_t port, + const char *digest, const char *transport_name) +{ + /* Iterate the already-registered bridge list: + + If you find a bridge with the same adress and port, mark it for + removal. It doesn't make sense to have two active bridges with + the same IP:PORT. If the bridge in question has a different + digest or transport than digest/transport_name, + it's probably a misconfiguration and we should warn the user. + */ + SMARTLIST_FOREACH_BEGIN(bridge_list, bridge_info_t *, bridge) { + if (bridge->marked_for_removal) + continue; + + if (tor_addr_eq(&bridge->addr, addr) && (bridge->port == port)) { + + bridge->marked_for_removal = 1; + + if (!bridge_has_digest(bridge, digest) || + strcmp_opt(bridge->transport_name, transport_name)) { + /* warn the user */ + char *bridge_description_new, *bridge_description_old; + tor_asprintf(&bridge_description_new, "%s:%s:%s", + fmt_addrport(addr, port), + digest ? hex_str(digest, DIGEST_LEN) : "", + transport_name ? transport_name : ""); + tor_asprintf(&bridge_description_old, "%s:%s:%s", + fmt_addrport(&bridge->addr, bridge->port), + tor_digest_is_zero(bridge->identity) ? + "" : hex_str(bridge->identity,DIGEST_LEN), + bridge->transport_name ? bridge->transport_name : ""); + + log_warn(LD_GENERAL,"Tried to add bridge '%s', but we found a conflict" + " with the already registered bridge '%s'. We will discard" + " the old bridge and keep '%s'. If this is not what you" + " wanted, please change your configuration file accordingly.", + bridge_description_new, bridge_description_old, + bridge_description_new); + + tor_free(bridge_description_new); + tor_free(bridge_description_old); + } + } + } SMARTLIST_FOREACH_END(bridge); +} + +/** Remember a new bridge at addr:port. If digest + * is set, it tells us the identity key too. If we already had the + * bridge in our list, unmark it, and don't actually add anything new. + * If transport_name is non-NULL - the bridge is associated with a + * pluggable transport - we assign the transport to the bridge. */ +void +bridge_add_from_config(const tor_addr_t *addr, uint16_t port, + const char *digest, const char *transport_name) +{ + bridge_info_t *b; + + bridge_resolve_conflicts(addr, port, digest, transport_name); + + b = tor_malloc_zero(sizeof(bridge_info_t)); + tor_addr_copy(&b->addr, addr); + b->port = port; + if (digest) + memcpy(b->identity, digest, DIGEST_LEN); + if (transport_name) + b->transport_name = tor_strdup(transport_name); + b->fetch_status.schedule = DL_SCHED_BRIDGE; + if (!bridge_list) + bridge_list = smartlist_new(); + + smartlist_add(bridge_list, b); +} + +/** Return true iff routerset contains the bridge bridge. */ +static int +routerset_contains_bridge(const routerset_t *routerset, + const bridge_info_t *bridge) +{ + int result; + extend_info_t *extinfo; + tor_assert(bridge); + if (!routerset) + return 0; + + extinfo = extend_info_new( + NULL, bridge->identity, NULL, NULL, &bridge->addr, bridge->port); + result = routerset_contains_extendinfo(routerset, extinfo); + extend_info_free(extinfo); + return result; +} + +/** If digest is one of our known bridges, return it. */ +static bridge_info_t * +find_bridge_by_digest(const char *digest) +{ + SMARTLIST_FOREACH(bridge_list, bridge_info_t *, bridge, + { + if (tor_memeq(bridge->identity, digest, DIGEST_LEN)) + return bridge; + }); + return NULL; +} + +/** Given the addr and port of a bridge, if that bridge + * supports a pluggable transport, return its name. Otherwise, return + * NULL. */ +const char * +find_transport_name_by_bridge_addrport(const tor_addr_t *addr, uint16_t port) +{ + if (!bridge_list) + return NULL; + + SMARTLIST_FOREACH_BEGIN(bridge_list, const bridge_info_t *, bridge) { + if (tor_addr_eq(&bridge->addr, addr) && + (bridge->port == port)) + return bridge->transport_name; + } SMARTLIST_FOREACH_END(bridge); + + return NULL; +} + +/** If addr and port match the address and port of a + * bridge of ours that uses pluggable transports, place its transport + * in transport. + * + * Return 0 on success (found a transport, or found a bridge with no + * transport, or found no bridge); return -1 if we should be using a + * transport, but the transport could not be found. + */ +int +find_transport_by_bridge_addrport(const tor_addr_t *addr, uint16_t port, + const transport_t **transport) +{ + *transport = NULL; + if (!bridge_list) + return 0; + + SMARTLIST_FOREACH_BEGIN(bridge_list, const bridge_info_t *, bridge) { + if (tor_addr_eq(&bridge->addr, addr) && + (bridge->port == port)) { /* bridge matched */ + if (bridge->transport_name) { /* it also uses pluggable transports */ + *transport = transport_get_by_name(bridge->transport_name); + if (*transport == NULL) { /* it uses pluggable transports, but + the transport could not be found! */ + return -1; + } + return 0; + } else { /* bridge matched, but it doesn't use transports. */ + break; + } + } + } SMARTLIST_FOREACH_END(bridge); + + *transport = NULL; + return 0; +} + +/** We need to ask bridge for its server descriptor. */ +static void +launch_direct_bridge_descriptor_fetch(bridge_info_t *bridge) +{ + char *address; + const or_options_t *options = get_options(); + + if (connection_get_by_type_addr_port_purpose( + CONN_TYPE_DIR, &bridge->addr, bridge->port, + DIR_PURPOSE_FETCH_SERVERDESC)) + return; /* it's already on the way */ + + if (routerset_contains_bridge(options->ExcludeNodes, bridge)) { + download_status_mark_impossible(&bridge->fetch_status); + log_warn(LD_APP, "Not using bridge at %s: it is in ExcludeNodes.", + safe_str_client(fmt_and_decorate_addr(&bridge->addr))); + return; + } + + address = tor_dup_addr(&bridge->addr); + + directory_initiate_command(address, &bridge->addr, + bridge->port, 0/*no dirport*/, + bridge->identity, + DIR_PURPOSE_FETCH_SERVERDESC, + ROUTER_PURPOSE_BRIDGE, + DIRIND_ONEHOP, "authority.z", NULL, 0, 0); + tor_free(address); +} + +/** Fetching the bridge descriptor from the bridge authority returned a + * "not found". Fall back to trying a direct fetch. */ +void +retry_bridge_descriptor_fetch_directly(const char *digest) +{ + bridge_info_t *bridge = find_bridge_by_digest(digest); + if (!bridge) + return; /* not found? oh well. */ + + launch_direct_bridge_descriptor_fetch(bridge); +} + +/** For each bridge in our list for which we don't currently have a + * descriptor, fetch a new copy of its descriptor -- either directly + * from the bridge or via a bridge authority. */ +void +fetch_bridge_descriptors(const or_options_t *options, time_t now) +{ + int num_bridge_auths = get_n_authorities(BRIDGE_DIRINFO); + int ask_bridge_directly; + int can_use_bridge_authority; + + if (!bridge_list) + return; + + /* If we still have unconfigured managed proxies, don't go and + connect to a bridge. */ + if (pt_proxies_configuration_pending()) + return; + + SMARTLIST_FOREACH_BEGIN(bridge_list, bridge_info_t *, bridge) + { + if (!download_status_is_ready(&bridge->fetch_status, now, + IMPOSSIBLE_TO_DOWNLOAD)) + continue; /* don't bother, no need to retry yet */ + if (routerset_contains_bridge(options->ExcludeNodes, bridge)) { + download_status_mark_impossible(&bridge->fetch_status); + log_warn(LD_APP, "Not using bridge at %s: it is in ExcludeNodes.", + safe_str_client(fmt_and_decorate_addr(&bridge->addr))); + continue; + } + + /* schedule another fetch as if this one will fail, in case it does */ + download_status_failed(&bridge->fetch_status, 0); + + can_use_bridge_authority = !tor_digest_is_zero(bridge->identity) && + num_bridge_auths; + ask_bridge_directly = !can_use_bridge_authority || + !options->UpdateBridgesFromAuthority; + log_debug(LD_DIR, "ask_bridge_directly=%d (%d, %d, %d)", + ask_bridge_directly, tor_digest_is_zero(bridge->identity), + !options->UpdateBridgesFromAuthority, !num_bridge_auths); + + if (ask_bridge_directly && + !fascist_firewall_allows_address_or(&bridge->addr, bridge->port)) { + log_notice(LD_DIR, "Bridge at '%s' isn't reachable by our " + "firewall policy. %s.", + fmt_addrport(&bridge->addr, bridge->port), + can_use_bridge_authority ? + "Asking bridge authority instead" : "Skipping"); + if (can_use_bridge_authority) + ask_bridge_directly = 0; + else + continue; + } + + if (ask_bridge_directly) { + /* we need to ask the bridge itself for its descriptor. */ + launch_direct_bridge_descriptor_fetch(bridge); + } else { + /* We have a digest and we want to ask an authority. We could + * combine all the requests into one, but that may give more + * hints to the bridge authority than we want to give. */ + char resource[10 + HEX_DIGEST_LEN]; + memcpy(resource, "fp/", 3); + base16_encode(resource+3, HEX_DIGEST_LEN+1, + bridge->identity, DIGEST_LEN); + memcpy(resource+3+HEX_DIGEST_LEN, ".z", 3); + log_info(LD_DIR, "Fetching bridge info '%s' from bridge authority.", + resource); + directory_get_from_dirserver(DIR_PURPOSE_FETCH_SERVERDESC, + ROUTER_PURPOSE_BRIDGE, resource, 0); + } + } + SMARTLIST_FOREACH_END(bridge); +} + +/** If our bridge is configured to be a different address than + * the bridge gives in node, rewrite the routerinfo + * we received to use the address we meant to use. Now we handle + * multihomed bridges better. + */ +static void +rewrite_node_address_for_bridge(const bridge_info_t *bridge, node_t *node) +{ + /* XXXX move this function. */ + /* XXXX overridden addresses should really live in the node_t, so that the + * routerinfo_t and the microdesc_t can be immutable. But we can only + * do that safely if we know that no function that connects to an OR + * does so through an address from any source other than node_get_addr(). + */ + tor_addr_t addr; + + if (node->ri) { + routerinfo_t *ri = node->ri; + tor_addr_from_ipv4h(&addr, ri->addr); + + if ((!tor_addr_compare(&bridge->addr, &addr, CMP_EXACT) && + bridge->port == ri->or_port) || + (!tor_addr_compare(&bridge->addr, &ri->ipv6_addr, CMP_EXACT) && + bridge->port == ri->ipv6_orport)) { + /* they match, so no need to do anything */ + } else { + if (tor_addr_family(&bridge->addr) == AF_INET) { + ri->addr = tor_addr_to_ipv4h(&bridge->addr); + tor_free(ri->address); + ri->address = tor_dup_ip(ri->addr); + ri->or_port = bridge->port; + log_info(LD_DIR, + "Adjusted bridge routerinfo for '%s' to match configured " + "address %s:%d.", + ri->nickname, ri->address, ri->or_port); + } else if (tor_addr_family(&bridge->addr) == AF_INET6) { + tor_addr_copy(&ri->ipv6_addr, &bridge->addr); + ri->ipv6_orport = bridge->port; + log_info(LD_DIR, + "Adjusted bridge routerinfo for '%s' to match configured " + "address %s.", + ri->nickname, fmt_addrport(&ri->ipv6_addr, ri->ipv6_orport)); + } else { + log_err(LD_BUG, "Address family not supported: %d.", + tor_addr_family(&bridge->addr)); + return; + } + } + + /* Mark which address to use based on which bridge_t we got. */ + node->ipv6_preferred = (tor_addr_family(&bridge->addr) == AF_INET6 && + !tor_addr_is_null(&node->ri->ipv6_addr)); + + /* XXXipv6 we lack support for falling back to another address for + the same relay, warn the user */ + if (!tor_addr_is_null(&ri->ipv6_addr)) { + tor_addr_port_t ap; + node_get_pref_orport(node, &ap); + log_notice(LD_CONFIG, + "Bridge '%s' has both an IPv4 and an IPv6 address. " + "Will prefer using its %s address (%s).", + ri->nickname, + tor_addr_family(&ap.addr) == AF_INET6 ? "IPv6" : "IPv4", + fmt_addrport(&ap.addr, ap.port)); + } + } + if (node->rs) { + routerstatus_t *rs = node->rs; + tor_addr_from_ipv4h(&addr, rs->addr); + + if (!tor_addr_compare(&bridge->addr, &addr, CMP_EXACT) && + bridge->port == rs->or_port) { + /* they match, so no need to do anything */ + } else { + rs->addr = tor_addr_to_ipv4h(&bridge->addr); + rs->or_port = bridge->port; + log_info(LD_DIR, + "Adjusted bridge routerstatus for '%s' to match " + "configured address %s.", + rs->nickname, fmt_addrport(&bridge->addr, rs->or_port)); + } + } +} + +/** We just learned a descriptor for a bridge. See if that + * digest is in our entry guard list, and add it if not. */ +void +learned_bridge_descriptor(routerinfo_t *ri, int from_cache) +{ + tor_assert(ri); + tor_assert(ri->purpose == ROUTER_PURPOSE_BRIDGE); + if (get_options()->UseBridges) { + int first = !any_bridge_descriptors_known(); + bridge_info_t *bridge = get_configured_bridge_by_routerinfo(ri); + time_t now = time(NULL); + router_set_status(ri->cache_info.identity_digest, 1); + + if (bridge) { /* if we actually want to use this one */ + node_t *node; + /* it's here; schedule its re-fetch for a long time from now. */ + if (!from_cache) + download_status_reset(&bridge->fetch_status); + + node = node_get_mutable_by_id(ri->cache_info.identity_digest); + tor_assert(node); + rewrite_node_address_for_bridge(bridge, node); + add_an_entry_guard(node, 1, 1, 0); + + log_notice(LD_DIR, "new bridge descriptor '%s' (%s): %s", ri->nickname, + from_cache ? "cached" : "fresh", router_describe(ri)); + /* set entry->made_contact so if it goes down we don't drop it from + * our entry node list */ + entry_guard_register_connect_status(ri->cache_info.identity_digest, + 1, 0, now); + if (first) + routerlist_retry_directory_downloads(now); + } + } +} + +/** Return 1 if any of our entry guards have descriptors that + * are marked with purpose 'bridge' and are running. Else return 0. + * + * We use this function to decide if we're ready to start building + * circuits through our bridges, or if we need to wait until the + * directory "server/authority" requests finish. */ +int +any_bridge_descriptors_known(void) +{ + tor_assert(get_options()->UseBridges); + return choose_random_entry(NULL) != NULL; +} + +/** Return 1 if there are any directory conns fetching bridge descriptors + * that aren't marked for close. We use this to guess if we should tell + * the controller that we have a problem. */ +int +any_pending_bridge_descriptor_fetches(void) +{ + smartlist_t *conns = get_connection_array(); + SMARTLIST_FOREACH_BEGIN(conns, connection_t *, conn) { + if (conn->type == CONN_TYPE_DIR && + conn->purpose == DIR_PURPOSE_FETCH_SERVERDESC && + TO_DIR_CONN(conn)->router_purpose == ROUTER_PURPOSE_BRIDGE && + !conn->marked_for_close && + conn->linked && + conn->linked_conn && !conn->linked_conn->marked_for_close) { + log_debug(LD_DIR, "found one: %s", conn->address); + return 1; + } + } SMARTLIST_FOREACH_END(conn); + return 0; +} + +/** Return 1 if we have at least one descriptor for an entry guard + * (bridge or member of EntryNodes) and all descriptors we know are + * down. Else return 0. If act is 1, then mark the down guards + * up; else just observe and report. */ +static int +entries_retry_helper(const or_options_t *options, int act) +{ + const node_t *node; + int any_known = 0; + int any_running = 0; + int need_bridges = options->UseBridges != 0; + if (!entry_guards) + entry_guards = smartlist_new(); + SMARTLIST_FOREACH_BEGIN(entry_guards, entry_guard_t *, e) { + node = node_get_by_id(e->identity); + if (node && node_has_descriptor(node) && + node_is_bridge(node) == need_bridges) { + any_known = 1; + if (node->is_running) + any_running = 1; /* some entry is both known and running */ + else if (act) { + /* Mark all current connections to this OR as unhealthy, since + * otherwise there could be one that started 30 seconds + * ago, and in 30 seconds it will time out, causing us to mark + * the node down and undermine the retry attempt. We mark even + * the established conns, since if the network just came back + * we'll want to attach circuits to fresh conns. */ + connection_or_set_bad_connections(node->identity, 1); + + /* mark this entry node for retry */ + router_set_status(node->identity, 1); + e->can_retry = 1; + e->bad_since = 0; + } + } + } SMARTLIST_FOREACH_END(e); + log_debug(LD_DIR, "%d: any_known %d, any_running %d", + act, any_known, any_running); + return any_known && !any_running; +} + +/** Do we know any descriptors for our bridges / entrynodes, and are + * all the ones we have descriptors for down? */ +int +entries_known_but_down(const or_options_t *options) +{ + tor_assert(entry_list_is_constrained(options)); + return entries_retry_helper(options, 0); +} + +/** Mark all down known bridges / entrynodes up. */ +void +entries_retry_all(const or_options_t *options) +{ + tor_assert(entry_list_is_constrained(options)); + entries_retry_helper(options, 1); +} + +/** Return true if at least one of our bridges runs a Tor version that can + * provide microdescriptors to us. If not, we'll fall back to asking for + * full descriptors. */ +int +any_bridge_supports_microdescriptors(void) +{ + const node_t *node; + if (!get_options()->UseBridges || !entry_guards) + return 0; + SMARTLIST_FOREACH_BEGIN(entry_guards, entry_guard_t *, e) { + node = node_get_by_id(e->identity); + if (node && node->is_running && + node_is_bridge(node) && node_is_a_configured_bridge(node) && + node_understands_microdescriptors(node)) { + /* This is one of our current bridges, and we know enough about + * it to know that it will be able to answer our microdescriptor + * questions. */ + return 1; + } + } SMARTLIST_FOREACH_END(e); + return 0; +} + +/** Release all storage held by the list of entry guards and related + * memory structs. */ +void +entry_guards_free_all(void) +{ + if (entry_guards) { + SMARTLIST_FOREACH(entry_guards, entry_guard_t *, e, + entry_guard_free(e)); + smartlist_free(entry_guards); + entry_guards = NULL; + } + clear_bridge_list(); + smartlist_free(bridge_list); + bridge_list = NULL; + circuit_build_times_free_timeouts(&circ_times); +} + diff --cc src/or/or.h index 00f72ad,b54834d..6731552 --- a/src/or/or.h +++ b/src/or/or.h @@@ -3987,46 -3600,14 +3987,48 @@@ typedef struct */ int PathBiasCircThreshold; double PathBiasNoticeRate; - double PathBiasDisableRate; + double PathBiasWarnRate; + double PathBiasExtremeRate; + int PathBiasDropGuards; int PathBiasScaleThreshold; - int PathBiasScaleFactor; /** @} */ + /** + * Parameters for path-bias use detection + * @{ + */ + int PathBiasUseThreshold; + double PathBiasNoticeUseRate; + double PathBiasExtremeUseRate; + int PathBiasScaleUseThreshold; + /** @} */ + + int IPv6Exit; /**< Do we support exiting to IPv6 addresses? */ + + char *TLSECGroup; /**< One of "P256", "P224", or nil for auto */ + + /** Autobool: should we use the ntor handshake if we can? */ + int UseNTorHandshake; + + /** Fraction: */ + double PathsNeededToBuildCircuits; + + /** Do we serve v2 directory info at all? This is a temporary option, since + * we'd like to disable v2 directory serving entirely, but we need a way to + * make it temporarily disableable, in order to do fast testing and be + * able to turn it back on if it turns out to be non-workable. + * + * XXXX025 Make this always-on, or always-off. Right now, it's only + * enableable for authorities. + */ + int DisableV2DirectoryInfo_; + + /** What expiry time shall we place on our SSL certs? "0" means we + * should guess a suitable value. */ + int SSLKeyLifetime; + + /** How long (seconds) do we keep a guard before picking a new one? */ + int GuardLifetime; - } or_options_t; /** Persistent state for an onion router, as saved to disk. */

1 0

[tor/release-0.2.4] Merge commit '2f98bf5c9fac4dfd1bb07564ce08b13d1e330252' into maint-0.2.4
by arma＠torproject.org 11 Apr '13

11 Apr '13

commit 90d2162f32f511f23449211ad900a880fd384afb Merge: 5b0b51c 2f98bf5 Author: Nick Mathewson <nickm(a)torproject.org> Date: Wed Mar 20 10:52:26 2013 -0400 Merge commit '2f98bf5c9fac4dfd1bb07564ce08b13d1e330252' into maint-0.2.4 (This is the part of the Bug 8042 patch that warns about unsigned time_t) changes/warn-unsigned-time_t | 5 +++++ configure.ac | 4 ++++ 2 files changed, 9 insertions(+), 0 deletions(-)

1 0

[tor/release-0.2.4] Merge remote-tracking branch 'public/bug6304_v3' into maint-0.2.4
by arma＠torproject.org 11 Apr '13

11 Apr '13

commit 5b0b51ca3ff55d6e82081adf4cb1dbfa24ac4872 Merge: f1caa2f b9a8f8c Author: Nick Mathewson <nickm(a)torproject.org> Date: Wed Mar 20 10:31:28 2013 -0400 Merge remote-tracking branch 'public/bug6304_v3' into maint-0.2.4 Conflicts: src/or/circuitbuild.c src/or/config.c changes/bug6304 | 4 ++++ src/or/circuitstats.c | 19 ++++++++----------- src/or/config.c | 12 ++++++++---- 3 files changed, 20 insertions(+), 15 deletions(-) diff --cc src/or/circuitstats.c index 73e34d9,0000000..fe9c80d mode 100644,000000..100644 --- a/src/or/circuitstats.c +++ b/src/or/circuitstats.c @@@ -1,1559 -1,0 +1,1556 @@@ +/* Copyright (c) 2001 Matej Pfajfar. + * Copyright (c) 2001-2004, Roger Dingledine. + * Copyright (c) 2004-2006, Roger Dingledine, Nick Mathewson. + * Copyright (c) 2007-2013, The Tor Project, Inc. */ +/* See LICENSE for licensing information */ + +#define CIRCUITSTATS_PRIVATE + +#include "or.h" +#include "circuitbuild.h" +#include "circuitstats.h" +#include "config.h" +#include "confparse.h" +#include "control.h" +#include "networkstatus.h" +#include "statefile.h" + +#undef log +#include <math.h> + +#define CBT_BIN_TO_MS(bin) ((bin)*CBT_BIN_WIDTH + (CBT_BIN_WIDTH/2)) + +/** Global list of circuit build times */ +// XXXX: Add this as a member for entry_guard_t instead of global? +// Then we could do per-guard statistics, as guards are likely to +// vary in their own latency. The downside of this is that guards +// can change frequently, so we'd be building a lot more circuits +// most likely. +/* XXXX024 Make this static; add accessor functions. */ +circuit_build_times_t circ_times; + +/** If set, we're running the unit tests: we should avoid clobbering + * our state file or accessing get_options() or get_or_state() */ +static int unit_tests = 0; + +/** + * This function decides if CBT learning should be disabled. It returns + * true if one or more of the following four conditions are met: + * + * 1. If the cbtdisabled consensus parameter is set. + * 2. If the torrc option LearnCircuitBuildTimeout is false. + * 3. If we are a directory authority + * 4. If we fail to write circuit build time history to our state file. + */ +int +circuit_build_times_disabled(void) +{ + if (unit_tests) { + return 0; + } else { + int consensus_disabled = networkstatus_get_param(NULL, "cbtdisabled", + 0, 0, 1); + int config_disabled = !get_options()->LearnCircuitBuildTimeout; + int dirauth_disabled = get_options()->AuthoritativeDir; + int state_disabled = did_last_state_file_write_fail() ? 1 : 0; + + if (consensus_disabled || config_disabled || dirauth_disabled || + state_disabled) { + log_debug(LD_CIRC, + "CircuitBuildTime learning is disabled. " + "Consensus=%d, Config=%d, AuthDir=%d, StateFile=%d", + consensus_disabled, config_disabled, dirauth_disabled, + state_disabled); + return 1; + } else { + log_debug(LD_CIRC, + "CircuitBuildTime learning is not disabled. " + "Consensus=%d, Config=%d, AuthDir=%d, StateFile=%d", + consensus_disabled, config_disabled, dirauth_disabled, + state_disabled); + return 0; + } + } +} + +/** + * Retrieve and bounds-check the cbtmaxtimeouts consensus paramter. + * + * Effect: When this many timeouts happen in the last 'cbtrecentcount' + * circuit attempts, the client should discard all of its history and + * begin learning a fresh timeout value. + */ +static int32_t +circuit_build_times_max_timeouts(void) +{ + int32_t cbt_maxtimeouts; + + cbt_maxtimeouts = networkstatus_get_param(NULL, "cbtmaxtimeouts", + CBT_DEFAULT_MAX_RECENT_TIMEOUT_COUNT, + CBT_MIN_MAX_RECENT_TIMEOUT_COUNT, + CBT_MAX_MAX_RECENT_TIMEOUT_COUNT); + + if (!(get_options()->LearnCircuitBuildTimeout)) { + log_debug(LD_BUG, + "circuit_build_times_max_timeouts() called, cbtmaxtimeouts is" + " %d", + cbt_maxtimeouts); + } + + return cbt_maxtimeouts; +} + +/** + * Retrieve and bounds-check the cbtnummodes consensus paramter. + * + * Effect: This value governs how many modes to use in the weighted + * average calculation of Pareto parameter Xm. A value of 3 introduces + * some bias (2-5% of CDF) under ideal conditions, but allows for better + * performance in the event that a client chooses guard nodes of radically + * different performance characteristics. + */ +static int32_t +circuit_build_times_default_num_xm_modes(void) +{ + int32_t num = networkstatus_get_param(NULL, "cbtnummodes", + CBT_DEFAULT_NUM_XM_MODES, + CBT_MIN_NUM_XM_MODES, + CBT_MAX_NUM_XM_MODES); + + if (!(get_options()->LearnCircuitBuildTimeout)) { + log_debug(LD_BUG, + "circuit_build_times_default_num_xm_modes() called, cbtnummodes" + " is %d", + num); + } + + return num; +} + +/** + * Retrieve and bounds-check the cbtmincircs consensus paramter. + * + * Effect: This is the minimum number of circuits to build before + * computing a timeout. + */ +static int32_t +circuit_build_times_min_circs_to_observe(void) +{ + int32_t num = networkstatus_get_param(NULL, "cbtmincircs", + CBT_DEFAULT_MIN_CIRCUITS_TO_OBSERVE, + CBT_MIN_MIN_CIRCUITS_TO_OBSERVE, + CBT_MAX_MIN_CIRCUITS_TO_OBSERVE); + + if (!(get_options()->LearnCircuitBuildTimeout)) { + log_debug(LD_BUG, + "circuit_build_times_min_circs_to_observe() called, cbtmincircs" + " is %d", + num); + } + + return num; +} + +/** Return true iff cbt has recorded enough build times that we + * want to start acting on the timeout it implies. */ +int +circuit_build_times_enough_to_compute(circuit_build_times_t *cbt) +{ + return cbt->total_build_times >= circuit_build_times_min_circs_to_observe(); +} + +/** + * Retrieve and bounds-check the cbtquantile consensus paramter. + * + * Effect: This is the position on the quantile curve to use to set the + * timeout value. It is a percent (10-99). + */ +double +circuit_build_times_quantile_cutoff(void) +{ + int32_t num = networkstatus_get_param(NULL, "cbtquantile", + CBT_DEFAULT_QUANTILE_CUTOFF, + CBT_MIN_QUANTILE_CUTOFF, + CBT_MAX_QUANTILE_CUTOFF); + + if (!(get_options()->LearnCircuitBuildTimeout)) { + log_debug(LD_BUG, + "circuit_build_times_quantile_cutoff() called, cbtquantile" + " is %d", + num); + } + + return num/100.0; +} + +/** + * Retrieve and bounds-check the cbtclosequantile consensus paramter. + * + * Effect: This is the position on the quantile curve to use to set the + * timeout value to use to actually close circuits. It is a percent + * (0-99). + */ +static double +circuit_build_times_close_quantile(void) +{ + int32_t param; + /* Cast is safe - circuit_build_times_quantile_cutoff() is capped */ + int32_t min = (int)tor_lround(100*circuit_build_times_quantile_cutoff()); + param = networkstatus_get_param(NULL, "cbtclosequantile", + CBT_DEFAULT_CLOSE_QUANTILE, + CBT_MIN_CLOSE_QUANTILE, + CBT_MAX_CLOSE_QUANTILE); + + if (!(get_options()->LearnCircuitBuildTimeout)) { + log_debug(LD_BUG, + "circuit_build_times_close_quantile() called, cbtclosequantile" + " is %d", param); + } + + if (param < min) { + log_warn(LD_DIR, "Consensus parameter cbtclosequantile is " + "too small, raising to %d", min); + param = min; + } + return param / 100.0; +} + +/** + * Retrieve and bounds-check the cbttestfreq consensus paramter. + * + * Effect: Describes how often in seconds to build a test circuit to + * gather timeout values. Only applies if less than 'cbtmincircs' + * have been recorded. + */ +static int32_t +circuit_build_times_test_frequency(void) +{ + int32_t num = networkstatus_get_param(NULL, "cbttestfreq", + CBT_DEFAULT_TEST_FREQUENCY, + CBT_MIN_TEST_FREQUENCY, + CBT_MAX_TEST_FREQUENCY); + + if (!(get_options()->LearnCircuitBuildTimeout)) { + log_debug(LD_BUG, + "circuit_build_times_test_frequency() called, cbttestfreq is %d", + num); + } + + return num; +} + +/** + * Retrieve and bounds-check the cbtmintimeout consensus parameter. + * + * Effect: This is the minimum allowed timeout value in milliseconds. + * The minimum is to prevent rounding to 0 (we only check once + * per second). + */ +static int32_t +circuit_build_times_min_timeout(void) +{ + int32_t num = networkstatus_get_param(NULL, "cbtmintimeout", + CBT_DEFAULT_TIMEOUT_MIN_VALUE, + CBT_MIN_TIMEOUT_MIN_VALUE, + CBT_MAX_TIMEOUT_MIN_VALUE); + + if (!(get_options()->LearnCircuitBuildTimeout)) { + log_debug(LD_BUG, + "circuit_build_times_min_timeout() called, cbtmintimeout is %d", + num); + } + + return num; +} + +/** + * Retrieve and bounds-check the cbtinitialtimeout consensus paramter. + * + * Effect: This is the timeout value to use before computing a timeout, + * in milliseconds. + */ +int32_t +circuit_build_times_initial_timeout(void) +{ + int32_t min = circuit_build_times_min_timeout(); + int32_t param = networkstatus_get_param(NULL, "cbtinitialtimeout", + CBT_DEFAULT_TIMEOUT_INITIAL_VALUE, + CBT_MIN_TIMEOUT_INITIAL_VALUE, + CBT_MAX_TIMEOUT_INITIAL_VALUE); + + if (!(get_options()->LearnCircuitBuildTimeout)) { + log_debug(LD_BUG, + "circuit_build_times_initial_timeout() called, " + "cbtinitialtimeout is %d", + param); + } + + if (param < min) { + log_warn(LD_DIR, "Consensus parameter cbtinitialtimeout is too small, " + "raising to %d", min); + param = min; + } + return param; +} + +/** + * Retrieve and bounds-check the cbtrecentcount consensus paramter. + * + * Effect: This is the number of circuit build times to keep track of + * for deciding if we hit cbtmaxtimeouts and need to reset our state + * and learn a new timeout. + */ +static int32_t +circuit_build_times_recent_circuit_count(networkstatus_t *ns) +{ + int32_t num; + num = networkstatus_get_param(ns, "cbtrecentcount", + CBT_DEFAULT_RECENT_CIRCUITS, + CBT_MIN_RECENT_CIRCUITS, + CBT_MAX_RECENT_CIRCUITS); + + if (!(get_options()->LearnCircuitBuildTimeout)) { + log_debug(LD_BUG, + "circuit_build_times_recent_circuit_count() called, " + "cbtrecentcount is %d", + num); + } + + return num; +} + +/** + * This function is called when we get a consensus update. + * + * It checks to see if we have changed any consensus parameters + * that require reallocation or discard of previous stats. + */ +void +circuit_build_times_new_consensus_params(circuit_build_times_t *cbt, + networkstatus_t *ns) +{ + int32_t num; + + /* + * First check if we're doing adaptive timeouts at all; nothing to + * update if we aren't. + */ + + if (!circuit_build_times_disabled()) { + num = circuit_build_times_recent_circuit_count(ns); + + if (num > 0) { + if (num != cbt->liveness.num_recent_circs) { + int8_t *recent_circs; + log_notice(LD_CIRC, "The Tor Directory Consensus has changed how many " + "circuits we must track to detect network failures from %d " + "to %d.", cbt->liveness.num_recent_circs, num); + + tor_assert(cbt->liveness.timeouts_after_firsthop || + cbt->liveness.num_recent_circs == 0); + + /* + * Technically this is a circular array that we are reallocating + * and memcopying. However, since it only consists of either 1s + * or 0s, and is only used in a statistical test to determine when + * we should discard our history after a sufficient number of 1's + * have been reached, it is fine if order is not preserved or + * elements are lost. + * + * cbtrecentcount should only be changing in cases of severe network + * distress anyway, so memory correctness here is paramount over + * doing acrobatics to preserve the array. + */ + recent_circs = tor_malloc_zero(sizeof(int8_t)*num); + if (cbt->liveness.timeouts_after_firsthop && + cbt->liveness.num_recent_circs > 0) { + memcpy(recent_circs, cbt->liveness.timeouts_after_firsthop, + sizeof(int8_t)*MIN(num, cbt->liveness.num_recent_circs)); + } + + // Adjust the index if it needs it. + if (num < cbt->liveness.num_recent_circs) { + cbt->liveness.after_firsthop_idx = MIN(num-1, + cbt->liveness.after_firsthop_idx); + } + + tor_free(cbt->liveness.timeouts_after_firsthop); + cbt->liveness.timeouts_after_firsthop = recent_circs; + cbt->liveness.num_recent_circs = num; + } + /* else no change, nothing to do */ + } else { /* num == 0 */ + /* + * Weird. This probably shouldn't happen, so log a warning, but try + * to do something sensible anyway. + */ + + log_warn(LD_CIRC, + "The cbtrecentcircs consensus parameter came back zero! " + "This disables adaptive timeouts since we can't keep track of " + "any recent circuits."); + + circuit_build_times_free_timeouts(cbt); + } + } else { + /* + * Adaptive timeouts are disabled; this might be because of the + * LearnCircuitBuildTimes config parameter, and hence permanent, or + * the cbtdisabled consensus parameter, so it may be a new condition. + * Treat it like getting num == 0 above and free the circuit history + * if we have any. + */ + + circuit_build_times_free_timeouts(cbt); + } +} + +/** + * Return the initial default or configured timeout in milliseconds + */ +static double +circuit_build_times_get_initial_timeout(void) +{ + double timeout; + + /* + * Check if we have LearnCircuitBuildTimeout, and if we don't, + * always use CircuitBuildTimeout, no questions asked. + */ - if (get_options()->LearnCircuitBuildTimeout) { - if (!unit_tests && get_options()->CircuitBuildTimeout) { - timeout = get_options()->CircuitBuildTimeout*1000; - if (timeout < circuit_build_times_min_timeout()) { - log_warn(LD_CIRC, "Config CircuitBuildTimeout too low. Setting to %ds", - circuit_build_times_min_timeout()/1000); - timeout = circuit_build_times_min_timeout(); - } - } else { - timeout = circuit_build_times_initial_timeout(); ++ if (!unit_tests && get_options()->CircuitBuildTimeout) { ++ timeout = get_options()->CircuitBuildTimeout*1000; ++ if (get_options()->LearnCircuitBuildTimeout && ++ timeout < circuit_build_times_min_timeout()) { ++ log_warn(LD_CIRC, "Config CircuitBuildTimeout too low. Setting to %ds", ++ circuit_build_times_min_timeout()/1000); ++ timeout = circuit_build_times_min_timeout(); + } + } else { - timeout = get_options()->CircuitBuildTimeout*1000; ++ timeout = circuit_build_times_initial_timeout(); + } + + return timeout; +} + +/** + * Reset the build time state. + * + * Leave estimated parameters, timeout and network liveness intact + * for future use. + */ +void +circuit_build_times_reset(circuit_build_times_t *cbt) +{ + memset(cbt->circuit_build_times, 0, sizeof(cbt->circuit_build_times)); + cbt->total_build_times = 0; + cbt->build_times_idx = 0; + cbt->have_computed_timeout = 0; +} + +/** + * Initialize the buildtimes structure for first use. + * + * Sets the initial timeout values based on either the config setting, + * the consensus param, or the default (CBT_DEFAULT_TIMEOUT_INITIAL_VALUE). + */ +void +circuit_build_times_init(circuit_build_times_t *cbt) +{ + memset(cbt, 0, sizeof(*cbt)); + /* + * Check if we really are using adaptive timeouts, and don't keep + * track of this stuff if not. + */ + if (!circuit_build_times_disabled()) { + cbt->liveness.num_recent_circs = + circuit_build_times_recent_circuit_count(NULL); + cbt->liveness.timeouts_after_firsthop = + tor_malloc_zero(sizeof(int8_t)*cbt->liveness.num_recent_circs); + } else { + cbt->liveness.num_recent_circs = 0; + cbt->liveness.timeouts_after_firsthop = NULL; + } + cbt->close_ms = cbt->timeout_ms = circuit_build_times_get_initial_timeout(); + control_event_buildtimeout_set(cbt, BUILDTIMEOUT_SET_EVENT_RESET); +} + +/** + * Free the saved timeouts, if the cbtdisabled consensus parameter got turned + * on or something. + */ + +void +circuit_build_times_free_timeouts(circuit_build_times_t *cbt) +{ + if (!cbt) return; + + if (cbt->liveness.timeouts_after_firsthop) { + tor_free(cbt->liveness.timeouts_after_firsthop); + } + + cbt->liveness.num_recent_circs = 0; +} + +#if 0 +/** + * Rewind our build time history by n positions. + */ +static void +circuit_build_times_rewind_history(circuit_build_times_t *cbt, int n) +{ + int i = 0; + + cbt->build_times_idx -= n; + cbt->build_times_idx %= CBT_NCIRCUITS_TO_OBSERVE; + + for (i = 0; i < n; i++) { + cbt->circuit_build_times[(i+cbt->build_times_idx) + %CBT_NCIRCUITS_TO_OBSERVE]=0; + } + + if (cbt->total_build_times > n) { + cbt->total_build_times -= n; + } else { + cbt->total_build_times = 0; + } + + log_info(LD_CIRC, + "Rewound history by %d places. Current index: %d. " + "Total: %d", n, cbt->build_times_idx, cbt->total_build_times); +} +#endif + +/** + * Add a new build time value time to the set of build times. Time + * units are milliseconds. + * + * circuit_build_times cbt is a circular array, so loop around when + * array is full. + */ +int +circuit_build_times_add_time(circuit_build_times_t *cbt, build_time_t time) +{ + if (time <= 0 || time > CBT_BUILD_TIME_MAX) { + log_warn(LD_BUG, "Circuit build time is too large (%u)." + "This is probably a bug.", time); + tor_fragile_assert(); + return -1; + } + + log_debug(LD_CIRC, "Adding circuit build time %u", time); + + cbt->circuit_build_times[cbt->build_times_idx] = time; + cbt->build_times_idx = (cbt->build_times_idx + 1) % CBT_NCIRCUITS_TO_OBSERVE; + if (cbt->total_build_times < CBT_NCIRCUITS_TO_OBSERVE) + cbt->total_build_times++; + + if ((cbt->total_build_times % CBT_SAVE_STATE_EVERY) == 0) { + /* Save state every n circuit builds */ + if (!unit_tests && !get_options()->AvoidDiskWrites) + or_state_mark_dirty(get_or_state(), 0); + } + + return 0; +} + +/** + * Return maximum circuit build time + */ +static build_time_t +circuit_build_times_max(circuit_build_times_t *cbt) +{ + int i = 0; + build_time_t max_build_time = 0; + for (i = 0; i < CBT_NCIRCUITS_TO_OBSERVE; i++) { + if (cbt->circuit_build_times[i] > max_build_time + && cbt->circuit_build_times[i] != CBT_BUILD_ABANDONED) + max_build_time = cbt->circuit_build_times[i]; + } + return max_build_time; +} + +#if 0 +/** Return minimum circuit build time */ +build_time_t +circuit_build_times_min(circuit_build_times_t *cbt) +{ + int i = 0; + build_time_t min_build_time = CBT_BUILD_TIME_MAX; + for (i = 0; i < CBT_NCIRCUITS_TO_OBSERVE; i++) { + if (cbt->circuit_build_times[i] && /* 0 <-> uninitialized */ + cbt->circuit_build_times[i] < min_build_time) + min_build_time = cbt->circuit_build_times[i]; + } + if (min_build_time == CBT_BUILD_TIME_MAX) { + log_warn(LD_CIRC, "No build times less than CBT_BUILD_TIME_MAX!"); + } + return min_build_time; +} +#endif + +/** + * Calculate and return a histogram for the set of build times. + * + * Returns an allocated array of histrogram bins representing + * the frequency of index*CBT_BIN_WIDTH millisecond + * build times. Also outputs the number of bins in nbins. + * + * The return value must be freed by the caller. + */ +static uint32_t * +circuit_build_times_create_histogram(circuit_build_times_t *cbt, + build_time_t *nbins) +{ + uint32_t *histogram; + build_time_t max_build_time = circuit_build_times_max(cbt); + int i, c; + + *nbins = 1 + (max_build_time / CBT_BIN_WIDTH); + histogram = tor_malloc_zero(*nbins * sizeof(build_time_t)); + + // calculate histogram + for (i = 0; i < CBT_NCIRCUITS_TO_OBSERVE; i++) { + if (cbt->circuit_build_times[i] == 0 + || cbt->circuit_build_times[i] == CBT_BUILD_ABANDONED) + continue; /* 0 <-> uninitialized */ + + c = (cbt->circuit_build_times[i] / CBT_BIN_WIDTH); + histogram[c]++; + } + + return histogram; +} + +/** + * Return the Pareto start-of-curve parameter Xm. + * + * Because we are not a true Pareto curve, we compute this as the + * weighted average of the N most frequent build time bins. N is either + * 1 if we don't have enough circuit build time data collected, or + * determined by the consensus parameter cbtnummodes (default 3). + */ +static build_time_t +circuit_build_times_get_xm(circuit_build_times_t *cbt) +{ + build_time_t i, nbins; + build_time_t *nth_max_bin; + int32_t bin_counts=0; + build_time_t ret = 0; + uint32_t *histogram = circuit_build_times_create_histogram(cbt, &nbins); + int n=0; + int num_modes = circuit_build_times_default_num_xm_modes(); + + tor_assert(nbins > 0); + tor_assert(num_modes > 0); + + // Only use one mode if < 1000 buildtimes. Not enough data + // for multiple. + if (cbt->total_build_times < CBT_NCIRCUITS_TO_OBSERVE) + num_modes = 1; + + nth_max_bin = (build_time_t*)tor_malloc_zero(num_modes*sizeof(build_time_t)); + + /* Determine the N most common build times */ + for (i = 0; i < nbins; i++) { + if (histogram[i] >= histogram[nth_max_bin[0]]) { + nth_max_bin[0] = i; + } + + for (n = 1; n < num_modes; n++) { + if (histogram[i] >= histogram[nth_max_bin[n]] && + (!histogram[nth_max_bin[n-1]] + || histogram[i] < histogram[nth_max_bin[n-1]])) { + nth_max_bin[n] = i; + } + } + } + + for (n = 0; n < num_modes; n++) { + bin_counts += histogram[nth_max_bin[n]]; + ret += CBT_BIN_TO_MS(nth_max_bin[n])*histogram[nth_max_bin[n]]; + log_info(LD_CIRC, "Xm mode #%d: %u %u", n, CBT_BIN_TO_MS(nth_max_bin[n]), + histogram[nth_max_bin[n]]); + } + + /* The following assert is safe, because we don't get called when we + * haven't observed at least CBT_MIN_MIN_CIRCUITS_TO_OBSERVE circuits. */ + tor_assert(bin_counts > 0); + + ret /= bin_counts; + tor_free(histogram); + tor_free(nth_max_bin); + + return ret; +} + +/** + * Output a histogram of current circuit build times to + * the or_state_t state structure. + */ +void +circuit_build_times_update_state(circuit_build_times_t *cbt, + or_state_t *state) +{ + uint32_t *histogram; + build_time_t i = 0; + build_time_t nbins = 0; + config_line_t **next, *line; + + histogram = circuit_build_times_create_histogram(cbt, &nbins); + // write to state + config_free_lines(state->BuildtimeHistogram); + next = &state->BuildtimeHistogram; + *next = NULL; + + state->TotalBuildTimes = cbt->total_build_times; + state->CircuitBuildAbandonedCount = 0; + + for (i = 0; i < CBT_NCIRCUITS_TO_OBSERVE; i++) { + if (cbt->circuit_build_times[i] == CBT_BUILD_ABANDONED) + state->CircuitBuildAbandonedCount++; + } + + for (i = 0; i < nbins; i++) { + // compress the histogram by skipping the blanks + if (histogram[i] == 0) continue; + *next = line = tor_malloc_zero(sizeof(config_line_t)); + line->key = tor_strdup("CircuitBuildTimeBin"); + tor_asprintf(&line->value, "%d %d", + CBT_BIN_TO_MS(i), histogram[i]); + next = &(line->next); + } + + if (!unit_tests) { + if (!get_options()->AvoidDiskWrites) + or_state_mark_dirty(get_or_state(), 0); + } + + tor_free(histogram); +} + +/** + * Shuffle the build times array. + * + * Adapted from http://en.wikipedia.org/wiki/Fisher-Yates_shuffle + */ +static void +circuit_build_times_shuffle_and_store_array(circuit_build_times_t *cbt, + build_time_t *raw_times, + uint32_t num_times) +{ + uint32_t n = num_times; + if (num_times > CBT_NCIRCUITS_TO_OBSERVE) { + log_notice(LD_CIRC, "The number of circuit times that this Tor version " + "uses to calculate build times is less than the number stored " + "in your state file. Decreasing the circuit time history from " + "%lu to %d.", (unsigned long)num_times, + CBT_NCIRCUITS_TO_OBSERVE); + } + + if (n > INT_MAX-1) { + log_warn(LD_CIRC, "For some insane reasons, you had %lu circuit build " + "observations in your state file. That's far too many; probably " + "there's a bug here.", (unsigned long)n); + n = INT_MAX-1; + } + + /* This code can only be run on a compact array */ + while (n-- > 1) { + int k = crypto_rand_int(n + 1); /* 0 <= k <= n. */ + build_time_t tmp = raw_times[k]; + raw_times[k] = raw_times[n]; + raw_times[n] = tmp; + } + + /* Since the times are now shuffled, take a random CBT_NCIRCUITS_TO_OBSERVE + * subset (ie the first CBT_NCIRCUITS_TO_OBSERVE values) */ + for (n = 0; n < MIN(num_times, CBT_NCIRCUITS_TO_OBSERVE); n++) { + circuit_build_times_add_time(cbt, raw_times[n]); + } +} + +/** + * Filter old synthetic timeouts that were created before the + * new right-censored Pareto calculation was deployed. + * + * Once all clients before 0.2.1.13-alpha are gone, this code + * will be unused. + */ +static int +circuit_build_times_filter_timeouts(circuit_build_times_t *cbt) +{ + int num_filtered=0, i=0; + double timeout_rate = 0; + build_time_t max_timeout = 0; + + timeout_rate = circuit_build_times_timeout_rate(cbt); + max_timeout = (build_time_t)cbt->close_ms; + + for (i = 0; i < CBT_NCIRCUITS_TO_OBSERVE; i++) { + if (cbt->circuit_build_times[i] > max_timeout) { + build_time_t replaced = cbt->circuit_build_times[i]; + num_filtered++; + cbt->circuit_build_times[i] = CBT_BUILD_ABANDONED; + + log_debug(LD_CIRC, "Replaced timeout %d with %d", replaced, + cbt->circuit_build_times[i]); + } + } + + log_info(LD_CIRC, + "We had %d timeouts out of %d build times, " + "and filtered %d above the max of %u", + (int)(cbt->total_build_times*timeout_rate), + cbt->total_build_times, num_filtered, max_timeout); + + return num_filtered; +} + +/** + * Load histogram from state, shuffling the resulting array + * after we do so. Use this result to estimate parameters and + * calculate the timeout. + * + * Return -1 on error. + */ +int +circuit_build_times_parse_state(circuit_build_times_t *cbt, + or_state_t *state) +{ + int tot_values = 0; + uint32_t loaded_cnt = 0, N = 0; + config_line_t *line; + unsigned int i; + build_time_t *loaded_times; + int err = 0; + circuit_build_times_init(cbt); + + if (circuit_build_times_disabled()) { + return 0; + } + + /* build_time_t 0 means uninitialized */ + loaded_times = tor_malloc_zero(sizeof(build_time_t)*state->TotalBuildTimes); + + for (line = state->BuildtimeHistogram; line; line = line->next) { + smartlist_t *args = smartlist_new(); + smartlist_split_string(args, line->value, " ", + SPLIT_SKIP_SPACE|SPLIT_IGNORE_BLANK, 0); + if (smartlist_len(args) < 2) { + log_warn(LD_GENERAL, "Unable to parse circuit build times: " + "Too few arguments to CircuitBuildTime"); + err = 1; + SMARTLIST_FOREACH(args, char*, cp, tor_free(cp)); + smartlist_free(args); + break; + } else { + const char *ms_str = smartlist_get(args,0); + const char *count_str = smartlist_get(args,1); + uint32_t count, k; + build_time_t ms; + int ok; + ms = (build_time_t)tor_parse_ulong(ms_str, 0, 0, + CBT_BUILD_TIME_MAX, &ok, NULL); + if (!ok) { + log_warn(LD_GENERAL, "Unable to parse circuit build times: " + "Unparsable bin number"); + err = 1; + SMARTLIST_FOREACH(args, char*, cp, tor_free(cp)); + smartlist_free(args); + break; + } + count = (uint32_t)tor_parse_ulong(count_str, 0, 0, + UINT32_MAX, &ok, NULL); + if (!ok) { + log_warn(LD_GENERAL, "Unable to parse circuit build times: " + "Unparsable bin count"); + err = 1; + SMARTLIST_FOREACH(args, char*, cp, tor_free(cp)); + smartlist_free(args); + break; + } + + if (loaded_cnt+count+state->CircuitBuildAbandonedCount + > state->TotalBuildTimes) { + log_warn(LD_CIRC, + "Too many build times in state file. " + "Stopping short before %d", + loaded_cnt+count); + SMARTLIST_FOREACH(args, char*, cp, tor_free(cp)); + smartlist_free(args); + break; + } + + for (k = 0; k < count; k++) { + loaded_times[loaded_cnt++] = ms; + } + N++; + SMARTLIST_FOREACH(args, char*, cp, tor_free(cp)); + smartlist_free(args); + } + } + + log_info(LD_CIRC, + "Adding %d timeouts.", state->CircuitBuildAbandonedCount); + for (i=0; i < state->CircuitBuildAbandonedCount; i++) { + loaded_times[loaded_cnt++] = CBT_BUILD_ABANDONED; + } + + if (loaded_cnt != state->TotalBuildTimes) { + log_warn(LD_CIRC, + "Corrupt state file? Build times count mismatch. " + "Read %d times, but file says %d", loaded_cnt, + state->TotalBuildTimes); + err = 1; + circuit_build_times_reset(cbt); + goto done; + } + + circuit_build_times_shuffle_and_store_array(cbt, loaded_times, loaded_cnt); + + /* Verify that we didn't overwrite any indexes */ + for (i=0; i < CBT_NCIRCUITS_TO_OBSERVE; i++) { + if (!cbt->circuit_build_times[i]) + break; + tot_values++; + } + log_info(LD_CIRC, + "Loaded %d/%d values from %d lines in circuit time histogram", + tot_values, cbt->total_build_times, N); + + if (cbt->total_build_times != tot_values + || cbt->total_build_times > CBT_NCIRCUITS_TO_OBSERVE) { + log_warn(LD_CIRC, + "Corrupt state file? Shuffled build times mismatch. " + "Read %d times, but file says %d", tot_values, + state->TotalBuildTimes); + err = 1; + circuit_build_times_reset(cbt); + goto done; + } + + circuit_build_times_set_timeout(cbt); + + if (!state->CircuitBuildAbandonedCount && cbt->total_build_times) { + circuit_build_times_filter_timeouts(cbt); + } + + done: + tor_free(loaded_times); + return err ? -1 : 0; +} + +/** + * Estimates the Xm and Alpha parameters using + * http://en.wikipedia.org/wiki/Pareto_distribution#Parameter_estimation + * + * The notable difference is that we use mode instead of min to estimate Xm. + * This is because our distribution is frechet-like. We claim this is + * an acceptable approximation because we are only concerned with the + * accuracy of the CDF of the tail. + */ +int +circuit_build_times_update_alpha(circuit_build_times_t *cbt) +{ + build_time_t *x=cbt->circuit_build_times; + double a = 0; + int n=0,i=0,abandoned_count=0; + build_time_t max_time=0; + + /* http://en.wikipedia.org/wiki/Pareto_distribution#Parameter_estimation */ + /* We sort of cheat here and make our samples slightly more pareto-like + * and less frechet-like. */ + cbt->Xm = circuit_build_times_get_xm(cbt); + + tor_assert(cbt->Xm > 0); + + for (i=0; i< CBT_NCIRCUITS_TO_OBSERVE; i++) { + if (!x[i]) { + continue; + } + + if (x[i] < cbt->Xm) { + a += tor_mathlog(cbt->Xm); + } else if (x[i] == CBT_BUILD_ABANDONED) { + abandoned_count++; + } else { + a += tor_mathlog(x[i]); + if (x[i] > max_time) + max_time = x[i]; + } + n++; + } + + /* + * We are erring and asserting here because this can only happen + * in codepaths other than startup. The startup state parsing code + * performs this same check, and resets state if it hits it. If we + * hit it at runtime, something serious has gone wrong. + */ + if (n!=cbt->total_build_times) { + log_err(LD_CIRC, "Discrepancy in build times count: %d vs %d", n, + cbt->total_build_times); + } + tor_assert(n==cbt->total_build_times); + + if (max_time <= 0) { + /* This can happen if Xm is actually the *maximum* value in the set. + * It can also happen if we've abandoned every single circuit somehow. + * In either case, tell the caller not to compute a new build timeout. */ + log_warn(LD_BUG, + "Could not determine largest build time (%d). " + "Xm is %dms and we've abandoned %d out of %d circuits.", max_time, + cbt->Xm, abandoned_count, n); + return 0; + } + + a += abandoned_count*tor_mathlog(max_time); + + a -= n*tor_mathlog(cbt->Xm); + // Estimator comes from Eq #4 in: + // "Bayesian estimation based on trimmed samples from Pareto populations" + // by Arturo J. Fernández. We are right-censored only. + a = (n-abandoned_count)/a; + + cbt->alpha = a; + + return 1; +} + +/** + * This is the Pareto Quantile Function. It calculates the point x + * in the distribution such that F(x) = quantile (ie quantile*100% + * of the mass of the density function is below x on the curve). + * + * We use it to calculate the timeout and also to generate synthetic + * values of time for circuits that timeout before completion. + * + * See http://en.wikipedia.org/wiki/Quantile_function, + * http://en.wikipedia.org/wiki/Inverse_transform_sampling and + * http://en.wikipedia.org/wiki/Pareto_distribution#Generating_a_ + * random_sample_from_Pareto_distribution + * That's right. I'll cite wikipedia all day long. + * + * Return value is in milliseconds. + */ +double +circuit_build_times_calculate_timeout(circuit_build_times_t *cbt, + double quantile) +{ + double ret; + tor_assert(quantile >= 0); + tor_assert(1.0-quantile > 0); + tor_assert(cbt->Xm > 0); + + ret = cbt->Xm/pow(1.0-quantile,1.0/cbt->alpha); + if (ret > INT32_MAX) { + ret = INT32_MAX; + } + tor_assert(ret > 0); + return ret; +} + +/** Pareto CDF */ +double +circuit_build_times_cdf(circuit_build_times_t *cbt, double x) +{ + double ret; + tor_assert(cbt->Xm > 0); + ret = 1.0-pow(cbt->Xm/x,cbt->alpha); + tor_assert(0 <= ret && ret <= 1.0); + return ret; +} + +/** + * Generate a synthetic time using our distribution parameters. + * + * The return value will be within the [q_lo, q_hi) quantile points + * on the CDF. + */ +build_time_t +circuit_build_times_generate_sample(circuit_build_times_t *cbt, + double q_lo, double q_hi) +{ + double randval = crypto_rand_double(); + build_time_t ret; + double u; + + /* Generate between [q_lo, q_hi) */ + /*XXXX This is what nextafter is supposed to be for; we should use it on the + * platforms that support it. */ + q_hi -= 1.0/(INT32_MAX); + + tor_assert(q_lo >= 0); + tor_assert(q_hi < 1); + tor_assert(q_lo < q_hi); + + u = q_lo + (q_hi-q_lo)*randval; + + tor_assert(0 <= u && u < 1.0); + /* circuit_build_times_calculate_timeout returns <= INT32_MAX */ + ret = (build_time_t) + tor_lround(circuit_build_times_calculate_timeout(cbt, u)); + tor_assert(ret > 0); + return ret; +} + +/** + * Estimate an initial alpha parameter by solving the quantile + * function with a quantile point and a specific timeout value. + */ +void +circuit_build_times_initial_alpha(circuit_build_times_t *cbt, + double quantile, double timeout_ms) +{ + // Q(u) = Xm/((1-u)^(1/a)) + // Q(0.8) = Xm/((1-0.8))^(1/a)) = CircBuildTimeout + // CircBuildTimeout = Xm/((1-0.8))^(1/a)) + // CircBuildTimeout = Xm*((1-0.8))^(-1/a)) + // ln(CircBuildTimeout) = ln(Xm)+ln(((1-0.8)))*(-1/a) + // -ln(1-0.8)/(ln(CircBuildTimeout)-ln(Xm))=a + tor_assert(quantile >= 0); + tor_assert(cbt->Xm > 0); + cbt->alpha = tor_mathlog(1.0-quantile)/ + (tor_mathlog(cbt->Xm)-tor_mathlog(timeout_ms)); + tor_assert(cbt->alpha > 0); +} + +/** + * Returns true if we need circuits to be built + */ +int +circuit_build_times_needs_circuits(circuit_build_times_t *cbt) +{ + /* Return true if < MIN_CIRCUITS_TO_OBSERVE */ + return !circuit_build_times_enough_to_compute(cbt); +} + +/** + * Returns true if we should build a timeout test circuit + * right now. + */ +int +circuit_build_times_needs_circuits_now(circuit_build_times_t *cbt) +{ + return circuit_build_times_needs_circuits(cbt) && + approx_time()-cbt->last_circ_at > circuit_build_times_test_frequency(); +} + +/** + * Called to indicate that the network showed some signs of liveness, + * i.e. we received a cell. + * + * This is used by circuit_build_times_network_check_live() to decide + * if we should record the circuit build timeout or not. + * + * This function is called every time we receive a cell. Avoid + * syscalls, events, and other high-intensity work. + */ +void +circuit_build_times_network_is_live(circuit_build_times_t *cbt) +{ + time_t now = approx_time(); + if (cbt->liveness.nonlive_timeouts > 0) { + log_notice(LD_CIRC, + "Tor now sees network activity. Restoring circuit build " + "timeout recording. Network was down for %d seconds " + "during %d circuit attempts.", + (int)(now - cbt->liveness.network_last_live), + cbt->liveness.nonlive_timeouts); + } + cbt->liveness.network_last_live = now; + cbt->liveness.nonlive_timeouts = 0; +} + +/** + * Called to indicate that we completed a circuit. Because this circuit + * succeeded, it doesn't count as a timeout-after-the-first-hop. + * + * This is used by circuit_build_times_network_check_changed() to determine + * if we had too many recent timeouts and need to reset our learned timeout + * to something higher. + */ +void +circuit_build_times_network_circ_success(circuit_build_times_t *cbt) +{ + /* Check for NULLness because we might not be using adaptive timeouts */ + if (cbt->liveness.timeouts_after_firsthop && + cbt->liveness.num_recent_circs > 0) { + cbt->liveness.timeouts_after_firsthop[cbt->liveness.after_firsthop_idx] + = 0; + cbt->liveness.after_firsthop_idx++; + cbt->liveness.after_firsthop_idx %= cbt->liveness.num_recent_circs; + } +} + +/** + * A circuit just timed out. If it failed after the first hop, record it + * in our history for later deciding if the network speed has changed. + * + * This is used by circuit_build_times_network_check_changed() to determine + * if we had too many recent timeouts and need to reset our learned timeout + * to something higher. + */ +static void +circuit_build_times_network_timeout(circuit_build_times_t *cbt, + int did_onehop) +{ + /* Check for NULLness because we might not be using adaptive timeouts */ + if (cbt->liveness.timeouts_after_firsthop && + cbt->liveness.num_recent_circs > 0) { + if (did_onehop) { + cbt->liveness.timeouts_after_firsthop[cbt->liveness.after_firsthop_idx] + = 1; + cbt->liveness.after_firsthop_idx++; + cbt->liveness.after_firsthop_idx %= cbt->liveness.num_recent_circs; + } + } +} + +/** + * A circuit was just forcibly closed. If there has been no recent network + * activity at all, but this circuit was launched back when we thought the + * network was live, increment the number of "nonlive" circuit timeouts. + * + * This is used by circuit_build_times_network_check_live() to decide + * if we should record the circuit build timeout or not. + */ +static void +circuit_build_times_network_close(circuit_build_times_t *cbt, + int did_onehop, time_t start_time) +{ + time_t now = time(NULL); + /* + * Check if this is a timeout that was for a circuit that spent its + * entire existence during a time where we have had no network activity. + */ + if (cbt->liveness.network_last_live < start_time) { + if (did_onehop) { + char last_live_buf[ISO_TIME_LEN+1]; + char start_time_buf[ISO_TIME_LEN+1]; + char now_buf[ISO_TIME_LEN+1]; + format_local_iso_time(last_live_buf, cbt->liveness.network_last_live); + format_local_iso_time(start_time_buf, start_time); + format_local_iso_time(now_buf, now); + log_warn(LD_BUG, + "Circuit somehow completed a hop while the network was " + "not live. Network was last live at %s, but circuit launched " + "at %s. It's now %s.", last_live_buf, start_time_buf, + now_buf); + } + cbt->liveness.nonlive_timeouts++; + if (cbt->liveness.nonlive_timeouts == 1) { + log_notice(LD_CIRC, + "Tor has not observed any network activity for the past %d " + "seconds. Disabling circuit build timeout recording.", + (int)(now - cbt->liveness.network_last_live)); + } else { + log_info(LD_CIRC, + "Got non-live timeout. Current count is: %d", + cbt->liveness.nonlive_timeouts); + } + } +} + +/** + * When the network is not live, we do not record circuit build times. + * + * The network is considered not live if there has been at least one + * circuit build that began and ended (had its close_ms measurement + * period expire) since we last received a cell. + * + * Also has the side effect of rewinding the circuit time history + * in the case of recent liveness changes. + */ +int +circuit_build_times_network_check_live(circuit_build_times_t *cbt) +{ + if (cbt->liveness.nonlive_timeouts > 0) { + return 0; + } + + return 1; +} + +/** + * Returns true if we have seen more than MAX_RECENT_TIMEOUT_COUNT of + * the past RECENT_CIRCUITS time out after the first hop. Used to detect + * if the network connection has changed significantly, and if so, + * resets our circuit build timeout to the default. + * + * Also resets the entire timeout history in this case and causes us + * to restart the process of building test circuits and estimating a + * new timeout. + */ +int +circuit_build_times_network_check_changed(circuit_build_times_t *cbt) +{ + int total_build_times = cbt->total_build_times; + int timeout_count=0; + int i; + + if (cbt->liveness.timeouts_after_firsthop && + cbt->liveness.num_recent_circs > 0) { + /* how many of our recent circuits made it to the first hop but then + * timed out? */ + for (i = 0; i < cbt->liveness.num_recent_circs; i++) { + timeout_count += cbt->liveness.timeouts_after_firsthop[i]; + } + } + + /* If 80% of our recent circuits are timing out after the first hop, + * we need to re-estimate a new initial alpha and timeout. */ + if (timeout_count < circuit_build_times_max_timeouts()) { + return 0; + } + + circuit_build_times_reset(cbt); + if (cbt->liveness.timeouts_after_firsthop && + cbt->liveness.num_recent_circs > 0) { + memset(cbt->liveness.timeouts_after_firsthop, 0, + sizeof(*cbt->liveness.timeouts_after_firsthop)* + cbt->liveness.num_recent_circs); + } + cbt->liveness.after_firsthop_idx = 0; + + /* Check to see if this has happened before. If so, double the timeout + * to give people on abysmally bad network connections a shot at access */ + if (cbt->timeout_ms >= circuit_build_times_get_initial_timeout()) { + if (cbt->timeout_ms > INT32_MAX/2 || cbt->close_ms > INT32_MAX/2) { + log_warn(LD_CIRC, "Insanely large circuit build timeout value. " + "(timeout = %fmsec, close = %fmsec)", + cbt->timeout_ms, cbt->close_ms); + } else { + cbt->timeout_ms *= 2; + cbt->close_ms *= 2; + } + } else { + cbt->close_ms = cbt->timeout_ms + = circuit_build_times_get_initial_timeout(); + } + + control_event_buildtimeout_set(cbt, BUILDTIMEOUT_SET_EVENT_RESET); + + log_notice(LD_CIRC, + "Your network connection speed appears to have changed. Resetting " + "timeout to %lds after %d timeouts and %d buildtimes.", + tor_lround(cbt->timeout_ms/1000), timeout_count, + total_build_times); + + return 1; +} + +/** + * Count the number of timeouts in a set of cbt data. + */ +double +circuit_build_times_timeout_rate(const circuit_build_times_t *cbt) +{ + int i=0,timeouts=0; + for (i = 0; i < CBT_NCIRCUITS_TO_OBSERVE; i++) { + if (cbt->circuit_build_times[i] >= cbt->timeout_ms) { + timeouts++; + } + } + + if (!cbt->total_build_times) + return 0; + + return ((double)timeouts)/cbt->total_build_times; +} + +/** + * Count the number of closed circuits in a set of cbt data. + */ +double +circuit_build_times_close_rate(const circuit_build_times_t *cbt) +{ + int i=0,closed=0; + for (i = 0; i < CBT_NCIRCUITS_TO_OBSERVE; i++) { + if (cbt->circuit_build_times[i] == CBT_BUILD_ABANDONED) { + closed++; + } + } + + if (!cbt->total_build_times) + return 0; + + return ((double)closed)/cbt->total_build_times; +} + +/** + * Store a timeout as a synthetic value. + * + * Returns true if the store was successful and we should possibly + * update our timeout estimate. + */ +int +circuit_build_times_count_close(circuit_build_times_t *cbt, + int did_onehop, + time_t start_time) +{ + if (circuit_build_times_disabled()) { + cbt->close_ms = cbt->timeout_ms + = circuit_build_times_get_initial_timeout(); + return 0; + } + + /* Record this force-close to help determine if the network is dead */ + circuit_build_times_network_close(cbt, did_onehop, start_time); + + /* Only count timeouts if network is live.. */ + if (!circuit_build_times_network_check_live(cbt)) { + return 0; + } + + circuit_build_times_add_time(cbt, CBT_BUILD_ABANDONED); + return 1; +} + +/** + * Update timeout counts to determine if we need to expire + * our build time history due to excessive timeouts. + * + * We do not record any actual time values at this stage; + * we are only interested in recording the fact that a timeout + * happened. We record the time values via + * circuit_build_times_count_close() and circuit_build_times_add_time(). + */ +void +circuit_build_times_count_timeout(circuit_build_times_t *cbt, + int did_onehop) +{ + if (circuit_build_times_disabled()) { + cbt->close_ms = cbt->timeout_ms + = circuit_build_times_get_initial_timeout(); + return; + } + + /* Register the fact that a timeout just occurred. */ + circuit_build_times_network_timeout(cbt, did_onehop); + + /* If there are a ton of timeouts, we should reset + * the circuit build timeout. */ + circuit_build_times_network_check_changed(cbt); +} + +/** + * Estimate a new timeout based on history and set our timeout + * variable accordingly. + */ +static int +circuit_build_times_set_timeout_worker(circuit_build_times_t *cbt) +{ + build_time_t max_time; + if (!circuit_build_times_enough_to_compute(cbt)) + return 0; + + if (!circuit_build_times_update_alpha(cbt)) + return 0; + + cbt->timeout_ms = circuit_build_times_calculate_timeout(cbt, + circuit_build_times_quantile_cutoff()); + + cbt->close_ms = circuit_build_times_calculate_timeout(cbt, + circuit_build_times_close_quantile()); + + max_time = circuit_build_times_max(cbt); + + if (cbt->timeout_ms > max_time) { + log_info(LD_CIRC, + "Circuit build timeout of %dms is beyond the maximum build " + "time we have ever observed. Capping it to %dms.", + (int)cbt->timeout_ms, max_time); + cbt->timeout_ms = max_time; + } + + if (max_time < INT32_MAX/2 && cbt->close_ms > 2*max_time) { + log_info(LD_CIRC, + "Circuit build measurement period of %dms is more than twice " + "the maximum build time we have ever observed. Capping it to " + "%dms.", (int)cbt->close_ms, 2*max_time); + cbt->close_ms = 2*max_time; + } + + /* Sometimes really fast guard nodes give us such a steep curve + * that this ends up being not that much greater than timeout_ms. + * Make it be at least 1 min to handle this case. */ + cbt->close_ms = MAX(cbt->close_ms, circuit_build_times_initial_timeout()); + + cbt->have_computed_timeout = 1; + return 1; +} + +/** + * Exposed function to compute a new timeout. Dispatches events and + * also filters out extremely high timeout values. + */ +void +circuit_build_times_set_timeout(circuit_build_times_t *cbt) +{ + long prev_timeout = tor_lround(cbt->timeout_ms/1000); + double timeout_rate; + + /* + * Just return if we aren't using adaptive timeouts + */ + if (circuit_build_times_disabled()) + return; + + if (!circuit_build_times_set_timeout_worker(cbt)) + return; + + if (cbt->timeout_ms < circuit_build_times_min_timeout()) { + log_info(LD_CIRC, "Set buildtimeout to low value %fms. Setting to %dms", + cbt->timeout_ms, circuit_build_times_min_timeout()); + cbt->timeout_ms = circuit_build_times_min_timeout(); + if (cbt->close_ms < cbt->timeout_ms) { + /* This shouldn't happen because of MAX() in timeout_worker above, + * but doing it just in case */ + cbt->close_ms = circuit_build_times_initial_timeout(); + } + } + + control_event_buildtimeout_set(cbt, BUILDTIMEOUT_SET_EVENT_COMPUTED); + + timeout_rate = circuit_build_times_timeout_rate(cbt); + + if (prev_timeout > tor_lround(cbt->timeout_ms/1000)) { + log_info(LD_CIRC, + "Based on %d circuit times, it looks like we don't need to " + "wait so long for circuits to finish. We will now assume a " + "circuit is too slow to use after waiting %ld seconds.", + cbt->total_build_times, + tor_lround(cbt->timeout_ms/1000)); + log_info(LD_CIRC, + "Circuit timeout data: %fms, %fms, Xm: %d, a: %f, r: %f", + cbt->timeout_ms, cbt->close_ms, cbt->Xm, cbt->alpha, + timeout_rate); + } else if (prev_timeout < tor_lround(cbt->timeout_ms/1000)) { + log_info(LD_CIRC, + "Based on %d circuit times, it looks like we need to wait " + "longer for circuits to finish. We will now assume a " + "circuit is too slow to use after waiting %ld seconds.", + cbt->total_build_times, + tor_lround(cbt->timeout_ms/1000)); + log_info(LD_CIRC, + "Circuit timeout data: %fms, %fms, Xm: %d, a: %f, r: %f", + cbt->timeout_ms, cbt->close_ms, cbt->Xm, cbt->alpha, + timeout_rate); + } else { + log_info(LD_CIRC, + "Set circuit build timeout to %lds (%fms, %fms, Xm: %d, a: %f," + " r: %f) based on %d circuit times", + tor_lround(cbt->timeout_ms/1000), + cbt->timeout_ms, cbt->close_ms, cbt->Xm, cbt->alpha, timeout_rate, + cbt->total_build_times); + } +} +/** Make a note that we're running unit tests (rather than running Tor + * itself), so we avoid clobbering our state file. */ +void +circuitbuild_running_unit_tests(void) +{ + unit_tests = 1; +} + diff --cc src/or/config.c index 68c0cca,b04a874..2b8eab7 --- a/src/or/config.c +++ b/src/or/config.c @@@ -2700,71 -3761,21 +2700,75 @@@ options_validate(or_options_t *old_opti options->LearnCircuitBuildTimeout = 0; } + if (options->Tor2webMode && options->UseEntryGuards) { + /* tor2web mode clients do not (and should not) use entry guards + * in any meaningful way. Further, tor2web mode causes the hidden + * service client code to do things which break the path bias + * detector, and it's far easier to turn off entry guards (and + * thus the path bias detector with it) than to figure out how to + * make a piece of code which cannot possibly help tor2web mode + * users compatible with tor2web mode. + */ + log_notice(LD_CONFIG, + "Tor2WebMode is enabled; disabling UseEntryGuards."); + options->UseEntryGuards = 0; + } + + if (!(options->UseEntryGuards) && + (options->RendConfigLines != NULL)) { + log_warn(LD_CONFIG, + "UseEntryGuards is disabled, but you have configured one or more " + "hidden services on this Tor instance. Your hidden services " + "will be very easy to locate using a well-known attack -- see " + "http://freehaven.net/anonbib/#hs-attack06 for details."); + } + - if (!(options->LearnCircuitBuildTimeout) && - options->CircuitBuildTimeout < RECOMMENDED_MIN_CIRCUIT_BUILD_TIMEOUT) { + if (!options->LearnCircuitBuildTimeout && options->CircuitBuildTimeout && + options->CircuitBuildTimeout < RECOMMENDED_MIN_CIRCUIT_BUILD_TIMEOUT) { log_warn(LD_CONFIG, - "CircuitBuildTimeout is shorter (%d seconds) than recommended " - "(%d seconds), and LearnCircuitBuildTimeout is disabled. " + "CircuitBuildTimeout is shorter (%d seconds) than the recommended " + "minimum (%d seconds), and LearnCircuitBuildTimeout is disabled. " "If tor isn't working, raise this value or enable " "LearnCircuitBuildTimeout.", options->CircuitBuildTimeout, RECOMMENDED_MIN_CIRCUIT_BUILD_TIMEOUT ); + } else if (!options->LearnCircuitBuildTimeout && + !options->CircuitBuildTimeout) { + log_notice(LD_CONFIG, "You disabled LearnCircuitBuildTimeout, but didn't " + "a CircuitBuildTimeout. I'll pick a plausible default."); } + if (options->PathBiasNoticeRate > 1.0) { + tor_asprintf(msg, + "PathBiasNoticeRate is too high. " + "It must be between 0 and 1.0"); + return -1; + } + if (options->PathBiasWarnRate > 1.0) { + tor_asprintf(msg, + "PathBiasWarnRate is too high. " + "It must be between 0 and 1.0"); + return -1; + } + if (options->PathBiasExtremeRate > 1.0) { + tor_asprintf(msg, + "PathBiasExtremeRate is too high. " + "It must be between 0 and 1.0"); + return -1; + } + if (options->PathBiasNoticeUseRate > 1.0) { + tor_asprintf(msg, + "PathBiasNoticeUseRate is too high. " + "It must be between 0 and 1.0"); + return -1; + } + if (options->PathBiasExtremeUseRate > 1.0) { + tor_asprintf(msg, + "PathBiasExtremeUseRate is too high. " + "It must be between 0 and 1.0"); + return -1; + } + if (options->MaxCircuitDirtiness < MIN_MAX_CIRCUIT_DIRTINESS) { log_warn(LD_CONFIG, "MaxCircuitDirtiness option is too short; " "raising to %d seconds.", MIN_MAX_CIRCUIT_DIRTINESS);

1 0

[tor/release-0.2.4] Let testing networks override ABSOLUTE_MIN_VALUE_FOR_FAST_FLAG
by arma＠torproject.org 11 Apr '13

11 Apr '13

commit a660fe6fd51cd511cdc610e4a50f06d59cbf74aa Author: Nick Mathewson <nickm(a)torproject.org> Date: Wed Mar 20 13:34:57 2013 -0400 Let testing networks override ABSOLUTE_MIN_VALUE_FOR_FAST_FLAG This adds a new option to fix bug 8508 which broke chutney networks. The bug was introduced by 317d16de. --- changes/bug8408 | 4 ++++ doc/tor.1.txt | 5 +++++ src/or/config.c | 1 + src/or/dirserv.c | 3 +++ src/or/or.h | 3 +++ 5 files changed, 16 insertions(+), 0 deletions(-) diff --git a/changes/bug8408 b/changes/bug8408 new file mode 100644 index 0000000..ae9cf17 --- /dev/null +++ b/changes/bug8408 @@ -0,0 +1,4 @@ + o Minor bugfixes: + - Allow TestingTorNetworks to override the 4096-byte minimum for the Fast + threshold. Otherwise they can't bootstrap until they've observed more + traffic. Fixes bug 8508; bugfix on 0.2.4.10-alpha. diff --git a/doc/tor.1.txt b/doc/tor.1.txt index 3be90be..5727ec3 100644 --- a/doc/tor.1.txt +++ b/doc/tor.1.txt @@ -2019,6 +2019,11 @@ The following options are used for running a testing Tor network. time. Changing this requires that **TestingTorNetwork** is set. (Default: 10 minutes) +**TestingMinFastFlagThreshold** __N__ **bytes**|**KBytes**|**MBytes**|**GBytes**:: + Minimum value for the Fast flag. Overrides the ordinary minimum taken + from the consensus when TestingTorNetwork is set. (Default: 0.) + + SIGNALS ------- diff --git a/src/or/config.c b/src/or/config.c index 2b8eab7..9e2230e 100644 --- a/src/or/config.c +++ b/src/or/config.c @@ -341,6 +341,7 @@ static config_var_t option_vars_[] = { V(PerConnBWRate, MEMUNIT, "0"), V(PidFile, STRING, NULL), V(TestingTorNetwork, BOOL, "0"), + V(TestingMinFastFlagThreshold, MEMUNIT, "0"), V(OptimisticData, AUTOBOOL, "auto"), V(PortForwarding, BOOL, "0"), V(PortForwardingHelper, FILENAME, "tor-fw-helper"), diff --git a/src/or/dirserv.c b/src/or/dirserv.c index 8e8f79a..81f328a 100644 --- a/src/or/dirserv.c +++ b/src/or/dirserv.c @@ -2005,6 +2005,9 @@ dirserv_compute_performance_thresholds(routerlist_t *rl, ABSOLUTE_MIN_VALUE_FOR_FAST_FLAG, ABSOLUTE_MIN_VALUE_FOR_FAST_FLAG, INT32_MAX); + if (options->TestingTorNetwork) { + min_fast = (int32_t)options->TestingMinFastFlagThreshold; + } max_fast = networkstatus_get_param(NULL, "FastFlagMaxThreshold", INT32_MAX, min_fast, INT32_MAX); if (fast_bandwidth < (uint32_t)min_fast) diff --git a/src/or/or.h b/src/or/or.h index 6731552..2b0102d 100644 --- a/src/or/or.h +++ b/src/or/or.h @@ -3910,6 +3910,9 @@ typedef struct { * of certain configuration options. */ int TestingTorNetwork; + /** Minimum value for the Fast flag threshold on testing networks. */ + uint64_t TestingMinFastFlagThreshold; + /** If true, and we have GeoIP data, and we're a bridge, keep a per-country * count of how many client addresses have contacted us so that we can help * the bridge authority guess which countries have blocked access to us. */

1 0

[tor/release-0.2.4] Set default minimum bandwidth for exit flag to zero for TestingTorNetwork
by arma＠torproject.org 11 Apr '13

11 Apr '13

commit 5c5198e713365fecf44741ae6f287c2ebdef18f6 Author: Andrea Shepard <andrea(a)torproject.org> Date: Wed Mar 20 11:16:41 2013 -0700 Set default minimum bandwidth for exit flag to zero for TestingTorNetwork --- src/or/config.c | 1 + src/or/dirserv.c | 10 ++++++++-- src/or/or.h | 3 +++ 3 files changed, 12 insertions(+), 2 deletions(-) diff --git a/src/or/config.c b/src/or/config.c index 9e2230e..a238a1a 100644 --- a/src/or/config.c +++ b/src/or/config.c @@ -341,6 +341,7 @@ static config_var_t option_vars_[] = { V(PerConnBWRate, MEMUNIT, "0"), V(PidFile, STRING, NULL), V(TestingTorNetwork, BOOL, "0"), + V(TestingMinExitFlagThreshold, MEMUNIT, "0"), V(TestingMinFastFlagThreshold, MEMUNIT, "0"), V(OptimisticData, AUTOBOOL, "auto"), V(PortForwarding, BOOL, "0"), diff --git a/src/or/dirserv.c b/src/or/dirserv.c index 81f328a..e837e4b 100644 --- a/src/or/dirserv.c +++ b/src/or/dirserv.c @@ -1894,11 +1894,17 @@ router_counts_toward_thresholds(const node_t *node, time_t now, /* Have measured bw? */ int have_mbw = dirserv_has_measured_bw(node->ri->cache_info.identity_digest); + uint64_t min_bw = ABSOLUTE_MIN_BW_VALUE_TO_CONSIDER; + const or_options_t *options = get_options(); + + if (options->TestingTorNetwork) { + min_bw = (int64_t)options->TestingMinExitFlagThreshold; + } return node->ri && router_is_active(node->ri, node, now) && !digestmap_get(omit_as_sybil, node->ri->cache_info.identity_digest) && - (dirserv_get_credible_bandwidth(node->ri) >= - ABSOLUTE_MIN_BW_VALUE_TO_CONSIDER) && (have_mbw || !require_mbw); + (dirserv_get_credible_bandwidth(node->ri) >= min_bw) && + (have_mbw || !require_mbw); } /** Look through the routerlist, the Mean Time Between Failure history, and diff --git a/src/or/or.h b/src/or/or.h index 2b0102d..4e19140 100644 --- a/src/or/or.h +++ b/src/or/or.h @@ -3910,6 +3910,9 @@ typedef struct { * of certain configuration options. */ int TestingTorNetwork; + /** Minimum value for the Exit flag threshold on testing networks. */ + uint64_t TestingMinExitFlagThreshold; + /** Minimum value for the Fast flag threshold on testing networks. */ uint64_t TestingMinFastFlagThreshold;

1 0

[tor/release-0.2.4] Fix a small memory leak in the unit tests
by arma＠torproject.org 11 Apr '13

11 Apr '13

commit 42fb61d172b172687cd57c86dd72b117f03f2136 Author: Nick Mathewson <nickm(a)torproject.org> Date: Thu Mar 21 07:52:36 2013 -0400 Fix a small memory leak in the unit tests Found by coverity; this is CID 992692. --- src/test/test_util.c | 26 +++++++++++++------------- 1 files changed, 13 insertions(+), 13 deletions(-) diff --git a/src/test/test_util.c b/src/test/test_util.c index 7ab54e1..2c4f804 100644 --- a/src/test/test_util.c +++ b/src/test/test_util.c @@ -802,7 +802,7 @@ test_util_strmisc(void) { char buf[1024]; int i; - char *cp; + char *cp, *cp_tmp; /* Test strl operations */ test_eq(5, strlcpy(buf, "Hello", 0)); @@ -1005,20 +1005,20 @@ test_util_strmisc(void) /* Test strndup and memdup */ { const char *s = "abcdefghijklmnopqrstuvwxyz"; - cp = tor_strndup(s, 30); - test_streq(cp, s); /* same string, */ - test_neq_ptr(cp, s); /* but different pointers. */ - tor_free(cp); + cp_tmp = tor_strndup(s, 30); + test_streq(cp_tmp, s); /* same string, */ + test_neq_ptr(cp_tmp, s); /* but different pointers. */ + tor_free(cp_tmp); - cp = tor_strndup(s, 5); - test_streq(cp, "abcde"); - tor_free(cp); + cp_tmp = tor_strndup(s, 5); + test_streq(cp_tmp, "abcde"); + tor_free(cp_tmp); s = "a\0b\0c\0d\0e\0"; - cp = tor_memdup(s,10); - test_memeq(cp, s, 10); /* same ram, */ - test_neq_ptr(cp, s); /* but different pointers. */ - tor_free(cp); + cp_tmp = tor_memdup(s,10); + test_memeq(cp_tmp, s, 10); /* same ram, */ + test_neq_ptr(cp_tmp, s); /* but different pointers. */ + tor_free(cp_tmp); } /* Test str-foo functions */ @@ -1097,7 +1097,7 @@ test_util_strmisc(void) tt_int_op(strcmp_len("blah", "", 0), ==, 0); done: - ; + tor_free(cp_tmp); } static void

1 0