Extract addr-policy parsing code.

6 files changed, 254 insertions, 203 deletions

diff --git a/src/feature/dirparse/parsecommon.c b/src/feature/dirparse/parsecommon.c
index ab815f585d..c12f199e4e 100644
--- a/src/feature/dirparse/parsecommon.c
+++ b/src/feature/dirparse/parsecommon.c
@@ -51,7 +51,7 @@ token_clear(directory_token_t *tok)
 int
 tokenize_string(memarea_t *area,
                 const char *start, const char *end, smartlist_t *out,
-                token_rule_t *table, int flags)
+                const token_rule_t *table, int flags)
 {
   const char **s;
   directory_token_t *tok = NULL;
@@ -257,7 +257,7 @@ token_check_object(memarea_t *area, const char *kwd,
  */
 directory_token_t *
 get_next_token(memarea_t *area,
-               const char **s, const char *eos, token_rule_t *table)
+               const char **s, const char *eos, const token_rule_t *table)
 {
   /** Reject any object at least this big; it is probably an overflow, an
    * attack, a bug, or some other nonsense. */
diff --git a/src/feature/dirparse/parsecommon.h b/src/feature/dirparse/parsecommon.h
index d0f3810c0b..f14862f04a 100644
--- a/src/feature/dirparse/parsecommon.h
+++ b/src/feature/dirparse/parsecommon.h
@@ -302,12 +302,12 @@ void token_clear(directory_token_t *tok);
 int tokenize_string(struct memarea_t *area,
                     const char *start, const char *end,
                     struct smartlist_t *out,
-                    token_rule_t *table,
+                    const token_rule_t *table,
                     int flags);
 directory_token_t *get_next_token(struct memarea_t *area,
                                   const char **s,
                                   const char *eos,
-                                  token_rule_t *table);
+                                  const token_rule_t *table);
 
 directory_token_t *find_by_keyword_(struct smartlist_t *s,
                                     directory_keyword keyword,
diff --git a/src/feature/dirparse/policy_parse.c b/src/feature/dirparse/policy_parse.c
new file mode 100644
index 0000000000..e102a62282
--- /dev/null
+++ b/src/feature/dirparse/policy_parse.c
@@ -0,0 +1,218 @@
+/* Copyright (c) 2001 Matej Pfajfar.
+ * Copyright (c) 2001-2004, Roger Dingledine.
+ * Copyright (c) 2004-2006, Roger Dingledine, Nick Mathewson.
+ * Copyright (c) 2007-2018, The Tor Project, Inc. */
+/* See LICENSE for licensing information */
+
+/**
+ * \file policy_parse.c
+ * \brief Code to parse address policies.
+ **/
+
+#define  EXPOSE_ROUTERDESC_TOKEN_TABLE
+
+#include "core/or/or.h"
+
+#include "core/or/policies.h"
+#include "feature/dirparse/parsecommon.h"
+#include "feature/dirparse/policy_parse.h"
+#include "feature/dirparse/routerparse.h"
+#include "feature/dirparse/unparseable.h"
+#include "lib/memarea/memarea.h"
+
+#include "core/or/addr_policy_st.h"
+
+static addr_policy_t *router_parse_addr_policy_private(directory_token_t *tok);
+
+/** Parse the addr policy in the string <b>s</b> and return it.  If
+ * assume_action is nonnegative, then insert its action (ADDR_POLICY_ACCEPT or
+ * ADDR_POLICY_REJECT) for items that specify no action.
+ *
+ * Returns NULL on policy errors.
+ *
+ * Set *<b>malformed_list</b> to true if the entire policy list should be
+ * discarded. Otherwise, set it to false, and only this item should be ignored
+ * on error - the rest of the policy list can continue to be processed and
+ * used.
+ *
+ * The addr_policy_t returned by this function can have its address set to
+ * AF_UNSPEC for '*'.  Use policy_expand_unspec() to turn this into a pair
+ * of AF_INET and AF_INET6 items.
+ */
+MOCK_IMPL(addr_policy_t *,
+router_parse_addr_policy_item_from_string,(const char *s, int assume_action,
+                                           int *malformed_list))
+{
+  directory_token_t *tok = NULL;
+  const char *cp, *eos;
+  /* Longest possible policy is
+   * "accept6 [ffff:ffff:..255]/128:10000-65535",
+   * which contains a max-length IPv6 address, plus 26 characters.
+   * But note that there can be an arbitrary amount of space between the
+   * accept and the address:mask/port element.
+   * We don't need to multiply TOR_ADDR_BUF_LEN by 2, as there is only one
+   * IPv6 address. But making the buffer shorter might cause valid long lines,
+   * which parsed in previous versions, to fail to parse in new versions.
+   * (These lines would have to have excessive amounts of whitespace.) */
+  char line[TOR_ADDR_BUF_LEN*2 + 32];
+  addr_policy_t *r;
+  memarea_t *area = NULL;
+
+  tor_assert(malformed_list);
+  *malformed_list = 0;
+
+  s = eat_whitespace(s);
+  /* We can only do assume_action on []-quoted IPv6, as "a" (accept)
+   * and ":" (port separator) are ambiguous */
+  if ((*s == '*' || *s == '[' || TOR_ISDIGIT(*s)) && assume_action >= 0) {
+    if (tor_snprintf(line, sizeof(line), "%s %s",
+               assume_action == ADDR_POLICY_ACCEPT?"accept":"reject", s)<0) {
+      log_warn(LD_DIR, "Policy %s is too long.", escaped(s));
+      return NULL;
+    }
+    cp = line;
+    tor_strlower(line);
+  } else { /* assume an already well-formed address policy line */
+    cp = s;
+  }
+
+  eos = cp + strlen(cp);
+  area = memarea_new();
+  tok = get_next_token(area, &cp, eos, routerdesc_token_table);
+  if (tok->tp == ERR_) {
+    log_warn(LD_DIR, "Error reading address policy: %s", tok->error);
+    goto err;
+  }
+  if (tok->tp != K_ACCEPT && tok->tp != K_ACCEPT6 &&
+      tok->tp != K_REJECT && tok->tp != K_REJECT6) {
+    log_warn(LD_DIR, "Expected 'accept' or 'reject'.");
+    goto err;
+  }
+
+  /* Use the extended interpretation of accept/reject *,
+   * expanding it into an IPv4 wildcard and an IPv6 wildcard.
+   * Also permit *4 and *6 for IPv4 and IPv6 only wildcards. */
+  r = router_parse_addr_policy(tok, TAPMP_EXTENDED_STAR);
+  if (!r) {
+    goto err;
+  }
+
+  /* Ensure that accept6/reject6 fields are followed by IPv6 addresses.
+   * AF_UNSPEC addresses are only permitted on the accept/reject field type.
+   * Unlike descriptors, torrcs exit policy accept/reject can be followed by
+   * either an IPv4 or IPv6 address. */
+  if ((tok->tp == K_ACCEPT6 || tok->tp == K_REJECT6) &&
+       tor_addr_family(&r->addr) != AF_INET6) {
+    /* This is a non-fatal error, just ignore this one entry. */
+    *malformed_list = 0;
+    log_warn(LD_DIR, "IPv4 address '%s' with accept6/reject6 field type in "
+             "exit policy. Ignoring, but continuing to parse rules. (Use "
+             "accept/reject with IPv4 addresses.)",
+             tok->n_args == 1 ? tok->args[0] : "");
+    addr_policy_free(r);
+    r = NULL;
+    goto done;
+  }
+
+  goto done;
+ err:
+  *malformed_list = 1;
+  r = NULL;
+ done:
+  token_clear(tok);
+  if (area) {
+    DUMP_AREA(area, "policy item");
+    memarea_drop_all(area);
+  }
+  return r;
+}
+
+/** Given a K_ACCEPT[6] or K_REJECT[6] token and a router, create and return
+ * a new exit_policy_t corresponding to the token. If TAPMP_EXTENDED_STAR
+ * is set in fmt_flags, K_ACCEPT6 and K_REJECT6 tokens followed by *
+ * expand to IPv6-only policies, otherwise they expand to IPv4 and IPv6
+ * policies */
+addr_policy_t *
+router_parse_addr_policy(directory_token_t *tok, unsigned fmt_flags)
+{
+  addr_policy_t newe;
+  char *arg;
+
+  tor_assert(tok->tp == K_REJECT || tok->tp == K_REJECT6 ||
+             tok->tp == K_ACCEPT || tok->tp == K_ACCEPT6);
+
+  if (tok->n_args != 1)
+    return NULL;
+  arg = tok->args[0];
+
+  if (!strcmpstart(arg,"private"))
+    return router_parse_addr_policy_private(tok);
+
+  memset(&newe, 0, sizeof(newe));
+
+  if (tok->tp == K_REJECT || tok->tp == K_REJECT6)
+    newe.policy_type = ADDR_POLICY_REJECT;
+  else
+    newe.policy_type = ADDR_POLICY_ACCEPT;
+
+  /* accept6/reject6 * produces an IPv6 wildcard address only.
+   * (accept/reject * produces rules for IPv4 and IPv6 wildcard addresses.) */
+  if ((fmt_flags & TAPMP_EXTENDED_STAR)
+      && (tok->tp == K_ACCEPT6 || tok->tp == K_REJECT6)) {
+    fmt_flags |= TAPMP_STAR_IPV6_ONLY;
+  }
+
+  if (tor_addr_parse_mask_ports(arg, fmt_flags, &newe.addr, &newe.maskbits,
+                                &newe.prt_min, &newe.prt_max) < 0) {
+    log_warn(LD_DIR,"Couldn't parse line %s. Dropping", escaped(arg));
+    return NULL;
+  }
+
+  return addr_policy_get_canonical_entry(&newe);
+}
+
+/** Parse an exit policy line of the format "accept[6]/reject[6] private:...".
+ * This didn't exist until Tor 0.1.1.15, so nobody should generate it in
+ * router descriptors until earlier versions are obsolete.
+ *
+ * accept/reject and accept6/reject6 private all produce rules for both
+ * IPv4 and IPv6 addresses.
+ */
+static addr_policy_t *
+router_parse_addr_policy_private(directory_token_t *tok)
+{
+  const char *arg;
+  uint16_t port_min, port_max;
+  addr_policy_t result;
+
+  arg = tok->args[0];
+  if (strcmpstart(arg, "private"))
+    return NULL;
+
+  arg += strlen("private");
+  arg = (char*) eat_whitespace(arg);
+  if (!arg || *arg != ':')
+    return NULL;
+
+  if (parse_port_range(arg+1, &port_min, &port_max)<0)
+    return NULL;
+
+  memset(&result, 0, sizeof(result));
+  if (tok->tp == K_REJECT || tok->tp == K_REJECT6)
+    result.policy_type = ADDR_POLICY_REJECT;
+  else
+    result.policy_type = ADDR_POLICY_ACCEPT;
+  result.is_private = 1;
+  result.prt_min = port_min;
+  result.prt_max = port_max;
+
+  if (tok->tp == K_ACCEPT6 || tok->tp == K_REJECT6) {
+    log_warn(LD_GENERAL,
+             "'%s' expands into rules which apply to all private IPv4 and "
+             "IPv6 addresses. (Use accept/reject private:* for IPv4 and "
+             "IPv6.)", tok->n_args == 1 ? tok->args[0] : "");
+  }
+
+  return addr_policy_get_canonical_entry(&result);
+}
+
diff --git a/src/feature/dirparse/policy_parse.h b/src/feature/dirparse/policy_parse.h
new file mode 100644
index 0000000000..887aa9261b
--- /dev/null
+++ b/src/feature/dirparse/policy_parse.h
@@ -0,0 +1,25 @@
+/* Copyright (c) 2001 Matej Pfajfar.
+ * Copyright (c) 2001-2004, Roger Dingledine.
+ * Copyright (c) 2004-2006, Roger Dingledine, Nick Mathewson.
+ * Copyright (c) 2007-2018, The Tor Project, Inc. */
+/* See LICENSE for licensing information */
+
+/**
+ * \file policy_parse.h
+ * \brief Header file for policy_parse.c.
+ **/
+
+#ifndef TOR_POLICY_PARSE_H
+#define TOR_POLICY_PARSE_H
+
+#include "lib/testsupport/testsupport.h"
+
+struct directory_token_t;
+
+MOCK_DECL(addr_policy_t *, router_parse_addr_policy_item_from_string,
+         (const char *s, int assume_action, int *malformed_list));
+
+addr_policy_t *router_parse_addr_policy(struct directory_token_t *tok,
+                                        unsigned fmt_flags);
+
+#endif /* !defined(TOR_POLICY_PARSE_H) */
diff --git a/src/feature/dirparse/routerparse.c b/src/feature/dirparse/routerparse.c
index 0e97735643..54c0e89480 100644
--- a/src/feature/dirparse/routerparse.c
+++ b/src/feature/dirparse/routerparse.c
@@ -51,6 +51,7 @@
  **/
 
 #define ROUTERPARSE_PRIVATE
+#define EXPOSE_ROUTERDESC_TOKEN_TABLE
 
 #include "core/or/or.h"
 #include "app/config/config.h"
@@ -61,6 +62,7 @@
 #include "feature/dirauth/shared_random.h"
 #include "feature/dircommon/voting_schedule.h"
 #include "feature/dirparse/parsecommon.h"
+#include "feature/dirparse/policy_parse.h"
 #include "feature/dirparse/routerparse.h"
 #include "feature/hs_common/shared_random_client.h"
 #include "feature/nodelist/authcert.h"
@@ -111,7 +113,7 @@
 /****************************************************************************/
 
 /** List of tokens recognized in router descriptors */
-static token_rule_t routerdesc_token_table[] = {
+const token_rule_t routerdesc_token_table[] = {
   T0N("reject",              K_REJECT,              ARGS,    NO_OBJ ),
   T0N("accept",              K_ACCEPT,              ARGS,    NO_OBJ ),
   T0N("reject6",             K_REJECT6,             ARGS,    NO_OBJ ),
@@ -197,10 +199,6 @@ static token_rule_t extrainfo_token_table[] = {
 
 /* static function prototypes */
 static int router_add_exit_policy(routerinfo_t *router,directory_token_t *tok);
-static addr_policy_t *router_parse_addr_policy(directory_token_t *tok,
-                                               unsigned fmt_flags);
-static addr_policy_t *router_parse_addr_policy_private(directory_token_t *tok);
-
 static smartlist_t *find_all_exitpolicy(smartlist_t *s);
 
 /** Set <b>digest</b> to the SHA-1 digest of the hash of the first router in
@@ -1201,109 +1199,6 @@ extrainfo_parse_entry_from_string(const char *s, const char *end,
   return extrainfo;
 }
 
-/** Parse the addr policy in the string <b>s</b> and return it.  If
- * assume_action is nonnegative, then insert its action (ADDR_POLICY_ACCEPT or
- * ADDR_POLICY_REJECT) for items that specify no action.
- *
- * Returns NULL on policy errors.
- *
- * Set *<b>malformed_list</b> to true if the entire policy list should be
- * discarded. Otherwise, set it to false, and only this item should be ignored
- * on error - the rest of the policy list can continue to be processed and
- * used.
- *
- * The addr_policy_t returned by this function can have its address set to
- * AF_UNSPEC for '*'.  Use policy_expand_unspec() to turn this into a pair
- * of AF_INET and AF_INET6 items.
- */
-MOCK_IMPL(addr_policy_t *,
-router_parse_addr_policy_item_from_string,(const char *s, int assume_action,
-                                           int *malformed_list))
-{
-  directory_token_t *tok = NULL;
-  const char *cp, *eos;
-  /* Longest possible policy is
-   * "accept6 [ffff:ffff:..255]/128:10000-65535",
-   * which contains a max-length IPv6 address, plus 26 characters.
-   * But note that there can be an arbitrary amount of space between the
-   * accept and the address:mask/port element.
-   * We don't need to multiply TOR_ADDR_BUF_LEN by 2, as there is only one
-   * IPv6 address. But making the buffer shorter might cause valid long lines,
-   * which parsed in previous versions, to fail to parse in new versions.
-   * (These lines would have to have excessive amounts of whitespace.) */
-  char line[TOR_ADDR_BUF_LEN*2 + 32];
-  addr_policy_t *r;
-  memarea_t *area = NULL;
-
-  tor_assert(malformed_list);
-  *malformed_list = 0;
-
-  s = eat_whitespace(s);
-  /* We can only do assume_action on []-quoted IPv6, as "a" (accept)
-   * and ":" (port separator) are ambiguous */
-  if ((*s == '*' || *s == '[' || TOR_ISDIGIT(*s)) && assume_action >= 0) {
-    if (tor_snprintf(line, sizeof(line), "%s %s",
-               assume_action == ADDR_POLICY_ACCEPT?"accept":"reject", s)<0) {
-      log_warn(LD_DIR, "Policy %s is too long.", escaped(s));
-      return NULL;
-    }
-    cp = line;
-    tor_strlower(line);
-  } else { /* assume an already well-formed address policy line */
-    cp = s;
-  }
-
-  eos = cp + strlen(cp);
-  area = memarea_new();
-  tok = get_next_token(area, &cp, eos, routerdesc_token_table);
-  if (tok->tp == ERR_) {
-    log_warn(LD_DIR, "Error reading address policy: %s", tok->error);
-    goto err;
-  }
-  if (tok->tp != K_ACCEPT && tok->tp != K_ACCEPT6 &&
-      tok->tp != K_REJECT && tok->tp != K_REJECT6) {
-    log_warn(LD_DIR, "Expected 'accept' or 'reject'.");
-    goto err;
-  }
-
-  /* Use the extended interpretation of accept/reject *,
-   * expanding it into an IPv4 wildcard and an IPv6 wildcard.
-   * Also permit *4 and *6 for IPv4 and IPv6 only wildcards. */
-  r = router_parse_addr_policy(tok, TAPMP_EXTENDED_STAR);
-  if (!r) {
-    goto err;
-  }
-
-  /* Ensure that accept6/reject6 fields are followed by IPv6 addresses.
-   * AF_UNSPEC addresses are only permitted on the accept/reject field type.
-   * Unlike descriptors, torrcs exit policy accept/reject can be followed by
-   * either an IPv4 or IPv6 address. */
-  if ((tok->tp == K_ACCEPT6 || tok->tp == K_REJECT6) &&
-       tor_addr_family(&r->addr) != AF_INET6) {
-    /* This is a non-fatal error, just ignore this one entry. */
-    *malformed_list = 0;
-    log_warn(LD_DIR, "IPv4 address '%s' with accept6/reject6 field type in "
-             "exit policy. Ignoring, but continuing to parse rules. (Use "
-             "accept/reject with IPv4 addresses.)",
-             tok->n_args == 1 ? tok->args[0] : "");
-    addr_policy_free(r);
-    r = NULL;
-    goto done;
-  }
-
-  goto done;
- err:
-  *malformed_list = 1;
-  r = NULL;
- done:
-  token_clear(tok);
-  if (area) {
-    DUMP_AREA(area, "policy item");
-    memarea_drop_all(area);
-  }
-  return r;
-}
-
 /** Add an exit policy stored in the token <b>tok</b> to the router info in
  * <b>router</b>.  Return 0 on success, -1 on failure. */
 static int
@@ -1340,95 +1235,6 @@ router_add_exit_policy(routerinfo_t *router, directory_token_t *tok)
   return 0;
 }
 
-/** Given a K_ACCEPT[6] or K_REJECT[6] token and a router, create and return
- * a new exit_policy_t corresponding to the token. If TAPMP_EXTENDED_STAR
- * is set in fmt_flags, K_ACCEPT6 and K_REJECT6 tokens followed by *
- * expand to IPv6-only policies, otherwise they expand to IPv4 and IPv6
- * policies */
-static addr_policy_t *
-router_parse_addr_policy(directory_token_t *tok, unsigned fmt_flags)
-{
-  addr_policy_t newe;
-  char *arg;
-
-  tor_assert(tok->tp == K_REJECT || tok->tp == K_REJECT6 ||
-             tok->tp == K_ACCEPT || tok->tp == K_ACCEPT6);
-
-  if (tok->n_args != 1)
-    return NULL;
-  arg = tok->args[0];
-
-  if (!strcmpstart(arg,"private"))
-    return router_parse_addr_policy_private(tok);
-
-  memset(&newe, 0, sizeof(newe));
-
-  if (tok->tp == K_REJECT || tok->tp == K_REJECT6)
-    newe.policy_type = ADDR_POLICY_REJECT;
-  else
-    newe.policy_type = ADDR_POLICY_ACCEPT;
-
-  /* accept6/reject6 * produces an IPv6 wildcard address only.
-   * (accept/reject * produces rules for IPv4 and IPv6 wildcard addresses.) */
-  if ((fmt_flags & TAPMP_EXTENDED_STAR)
-      && (tok->tp == K_ACCEPT6 || tok->tp == K_REJECT6)) {
-    fmt_flags |= TAPMP_STAR_IPV6_ONLY;
-  }
-
-  if (tor_addr_parse_mask_ports(arg, fmt_flags, &newe.addr, &newe.maskbits,
-                                &newe.prt_min, &newe.prt_max) < 0) {
-    log_warn(LD_DIR,"Couldn't parse line %s. Dropping", escaped(arg));
-    return NULL;
-  }
-
-  return addr_policy_get_canonical_entry(&newe);
-}
-
-/** Parse an exit policy line of the format "accept[6]/reject[6] private:...".
- * This didn't exist until Tor 0.1.1.15, so nobody should generate it in
- * router descriptors until earlier versions are obsolete.
- *
- * accept/reject and accept6/reject6 private all produce rules for both
- * IPv4 and IPv6 addresses.
- */
-static addr_policy_t *
-router_parse_addr_policy_private(directory_token_t *tok)
-{
-  const char *arg;
-  uint16_t port_min, port_max;
-  addr_policy_t result;
-
-  arg = tok->args[0];
-  if (strcmpstart(arg, "private"))
-    return NULL;
-
-  arg += strlen("private");
-  arg = (char*) eat_whitespace(arg);
-  if (!arg || *arg != ':')
-    return NULL;
-
-  if (parse_port_range(arg+1, &port_min, &port_max)<0)
-    return NULL;
-
-  memset(&result, 0, sizeof(result));
-  if (tok->tp == K_REJECT || tok->tp == K_REJECT6)
-    result.policy_type = ADDR_POLICY_REJECT;
-  else
-    result.policy_type = ADDR_POLICY_ACCEPT;
-  result.is_private = 1;
-  result.prt_min = port_min;
-  result.prt_max = port_max;
-
-  if (tok->tp == K_ACCEPT6 || tok->tp == K_REJECT6) {
-    log_warn(LD_GENERAL,
-             "'%s' expands into rules which apply to all private IPv4 and "
-             "IPv6 addresses. (Use accept/reject private:* for IPv4 and "
-             "IPv6.)", tok->n_args == 1 ? tok->args[0] : "");
-  }
-
-  return addr_policy_get_canonical_entry(&result);
-}
-
 /** Return a newly allocated smartlist of all accept or reject tokens in
  * <b>s</b>.
  */
diff --git a/src/feature/dirparse/routerparse.h b/src/feature/dirparse/routerparse.h
index 72f812ee4a..3022d60d36 100644
--- a/src/feature/dirparse/routerparse.h
+++ b/src/feature/dirparse/routerparse.h
@@ -34,8 +34,6 @@ struct digest_ri_map_t;
 extrainfo_t *extrainfo_parse_entry_from_string(const char *s, const char *end,
                              int cache_copy, struct digest_ri_map_t *routermap,
                              int *can_dl_again_out);
-MOCK_DECL(addr_policy_t *, router_parse_addr_policy_item_from_string,
-         (const char *s, int assume_action, int *malformed_list));
 
 int find_single_ipv6_orport(const smartlist_t *list,
                             tor_addr_t *addr_out,
@@ -44,6 +42,10 @@ int find_single_ipv6_orport(const smartlist_t *list,
 void routerparse_init(void);
 void routerparse_free_all(void);
 
+#ifdef EXPOSE_ROUTERDESC_TOKEN_TABLE
+extern const struct token_rule_t routerdesc_token_table[];
+#endif
+
 #define ED_DESC_SIGNATURE_PREFIX "Tor router descriptor signature v1"
 
 #endif /* !defined(TOR_ROUTERPARSE_H) */