IPv6: Satisfy unhinted assignments ordered by prefix length
[project/netifd.git] / interface-ip.c
index d18fa74..3b6be6b 100644 (file)
@@ -17,6 +17,7 @@
 #include <stdio.h>
 #include <unistd.h>
 
+#include <limits.h>
 #include <arpa/inet.h>
 
 #include "netifd.h"
@@ -50,7 +51,7 @@ static const struct blobmsg_policy route_attr[__ROUTE_MAX] = {
        [ROUTE_VALID] = { .name = "valid", .type = BLOBMSG_TYPE_INT32 },
 };
 
-const struct config_param_list route_attr_list = {
+const struct uci_blob_param_list route_attr_list = {
        .n_params = __ROUTE_MAX,
        .params = route_attr,
 };
@@ -90,6 +91,59 @@ match_if_addr(union if_addr *a1, union if_addr *a2, int mask)
        return !memcmp(p1, p2, sizeof(*p1));
 }
 
+static int set_ip_source_policy(bool add, bool v6, unsigned int priority,
+               const union if_addr *addr, uint8_t mask, struct interface *iface,
+               struct interface *in_iface, const char *action)
+{
+       struct iprule rule = {
+               .flags = IPRULE_PRIORITY,
+               .priority = priority
+       };
+
+       if (addr) {
+               rule.flags |= IPRULE_SRC;
+               rule.src_addr = *addr;
+               rule.src_mask = mask;
+       }
+
+       if (iface) {
+               rule.flags |= IPRULE_LOOKUP;
+               rule.lookup = (v6) ? iface->ip6table : iface->ip4table;
+
+               if (!rule.lookup)
+                       return 0;
+       } else if (action) {
+               rule.flags |= IPRULE_ACTION;
+               system_resolve_iprule_action(action, &rule.action);
+       }
+
+       if (in_iface && in_iface->l3_dev.dev) {
+               rule.flags |= IPRULE_IN;
+               strcpy(rule.in_dev, in_iface->l3_dev.dev->ifname);
+       }
+
+       rule.flags |= (v6) ? IPRULE_INET6 : IPRULE_INET4;
+
+       return (add) ? system_add_iprule(&rule) : system_del_iprule(&rule);
+}
+
+static int set_ip_lo_policy(bool add, bool v6, struct interface *iface)
+{
+       struct iprule rule = {
+               .flags = IPRULE_IN | IPRULE_LOOKUP | IPRULE_PRIORITY,
+               .priority = IPRULE_PRIORITY_NW + iface->l3_dev.dev->ifindex,
+               .lookup = (v6) ? iface->ip6table : iface->ip4table,
+               .in_dev = "lo"
+       };
+
+       if (!rule.lookup)
+               return 0;
+
+       rule.flags |= (v6) ? IPRULE_INET6 : IPRULE_INET4;
+
+       return (add) ? system_add_iprule(&rule) : system_del_iprule(&rule);
+}
+
 static bool
 __find_ip_addr_target(struct interface_ip_settings *ip, union if_addr *a, bool v6)
 {
@@ -102,7 +156,13 @@ __find_ip_addr_target(struct interface_ip_settings *ip, union if_addr *a, bool v
                if (v6 != ((addr->flags & DEVADDR_FAMILY) == DEVADDR_INET6))
                        continue;
 
-               if (!match_if_addr(&addr->addr, a, addr->mask))
+               // Handle offlink addresses correctly
+               unsigned int mask = addr->mask;
+               if ((addr->flags & DEVADDR_FAMILY) == DEVADDR_INET6 &&
+                               (addr->flags & DEVADDR_OFFLINK))
+                       mask = 128;
+
+               if (!match_if_addr(&addr->addr, a, mask))
                        continue;
 
                return true;
@@ -151,9 +211,8 @@ interface_ip_find_route_target(struct interface *iface, union if_addr *a,
 }
 
 struct interface *
-interface_ip_add_target_route(union if_addr *addr, bool v6)
+interface_ip_add_target_route(union if_addr *addr, bool v6, struct interface *iface)
 {
-       struct interface *iface;
        struct device_route *route, *r_next = NULL;
        bool defaultroute_target = false;
        int addrsize = v6 ? sizeof(addr->in6) : sizeof(addr->in);
@@ -169,7 +228,7 @@ interface_ip_add_target_route(union if_addr *addr, bool v6)
        else
                memcpy(&route->addr, addr, addrsize);
 
-       vlist_for_each_element(&interfaces, iface, node) {
+       if (iface) {
                /* look for locally addressable target first */
                if (interface_ip_find_addr_target(iface, addr, v6))
                        goto done;
@@ -177,6 +236,16 @@ interface_ip_add_target_route(union if_addr *addr, bool v6)
                /* do not stop at the first route, let the lookup compare
                 * masks to find the best match */
                interface_ip_find_route_target(iface, addr, v6, &r_next);
+       } else {
+               vlist_for_each_element(&interfaces, iface, node) {
+                       /* look for locally addressable target first */
+                       if (interface_ip_find_addr_target(iface, addr, v6))
+                               goto done;
+
+                       /* do not stop at the first route, let the lookup compare
+                        * masks to find the best match */
+                       interface_ip_find_route_target(iface, addr, v6, &r_next);
+               }
        }
 
        if (!r_next) {
@@ -188,6 +257,7 @@ interface_ip_add_target_route(union if_addr *addr, bool v6)
        memcpy(&route->nexthop, &r_next->nexthop, sizeof(route->nexthop));
        route->mtu = r_next->mtu;
        route->metric = r_next->metric;
+       route->table = r_next->table;
 
 done:
        route->iface = iface;
@@ -205,6 +275,7 @@ interface_ip_add_route(struct interface *iface, struct blob_attr *attr, bool v6)
        struct blob_attr *tb[__ROUTE_MAX], *cur;
        struct device_route *route;
        int af = v6 ? AF_INET6 : AF_INET;
+       bool is_proto_route = !!iface;
 
        blobmsg_parse(route_attr, __ROUTE_MAX, tb, blobmsg_data(attr), blobmsg_data_len(attr));
 
@@ -257,6 +328,12 @@ interface_ip_add_route(struct interface *iface, struct blob_attr *attr, bool v6)
                route->flags |= DEVROUTE_MTU;
        }
 
+       // Use source-based routing
+       if (is_proto_route) {
+               route->table = (v6) ? iface->ip6table : iface->ip4table;
+               route->flags |= DEVROUTE_SRCTABLE;
+       }
+
        if ((cur = tb[ROUTE_TABLE]) != NULL) {
                if (!system_resolve_rt_table(blobmsg_data(cur), &route->table)) {
                        DPRINTF("Failed to resolve routing table: %s\n", (char *) blobmsg_data(cur));
@@ -267,8 +344,12 @@ interface_ip_add_route(struct interface *iface, struct blob_attr *attr, bool v6)
                        route->flags |= DEVROUTE_TABLE;
        }
 
-       if ((cur = tb[ROUTE_VALID]) != NULL)
-               route->valid_until = system_get_rtime() + blobmsg_get_u32(cur);
+       if ((cur = tb[ROUTE_VALID]) != NULL) {
+               int64_t valid = blobmsg_get_u32(cur);
+               int64_t valid_until = valid + (int64_t)system_get_rtime();
+               if (valid_until <= LONG_MAX && valid != 0xffffffffLL) // Catch overflow
+                       route->valid_until = valid_until;
+       }
 
        vlist_add(&ip->route, &route->node, route);
        return;
@@ -304,7 +385,7 @@ route_cmp(const void *k1, const void *k2, void *ptr)
 static int
 prefix_cmp(const void *k1, const void *k2, void *ptr)
 {
-       return memcmp(k1, k2, sizeof(struct device_prefix) -
+       return memcmp(k1, k2, offsetof(struct device_prefix, pclass) -
                        offsetof(struct device_prefix, addr));
 }
 
@@ -346,6 +427,7 @@ interface_update_proto_addr(struct vlist_tree *tree,
        struct device *dev;
        struct device_addr *a_new = NULL, *a_old = NULL;
        bool keep = false;
+       bool v6 = false;
 
        ip = container_of(tree, struct interface_ip_settings, addr);
        iface = ip->iface;
@@ -384,6 +466,19 @@ interface_update_proto_addr(struct vlist_tree *tree,
        if (node_old) {
                if (!(a_old->flags & DEVADDR_EXTERNAL) && a_old->enabled && !keep) {
                        interface_handle_subnet_route(iface, a_old, false);
+
+                       if ((a_old->flags & DEVADDR_FAMILY) == DEVADDR_INET6)
+                               v6 = true;
+
+                       //This is needed for source routing to work correctly. If a device
+                       //has two connections to a network using the same subnet, adding
+                       //only the network-rule will cause packets to be routed through the
+                       //first matching network (source IP matches both masks).
+                       set_ip_source_policy(false, v6, IPRULE_PRIORITY_ADDR, &a_old->addr,
+                               (v6) ? 128 : 32, iface, NULL, NULL);
+                       set_ip_source_policy(false, v6, IPRULE_PRIORITY_NW, &a_old->addr,
+                               a_old->mask, iface, NULL, NULL);
+
                        system_del_address(dev, a_old);
                }
                free(a_old);
@@ -393,6 +488,15 @@ interface_update_proto_addr(struct vlist_tree *tree,
                a_new->enabled = true;
                if (!(a_new->flags & DEVADDR_EXTERNAL) && !keep) {
                        system_add_address(dev, a_new);
+
+                       if ((a_new->flags & DEVADDR_FAMILY) == DEVADDR_INET6)
+                               v6 = true;
+
+                       set_ip_source_policy(true, v6, IPRULE_PRIORITY_ADDR, &a_new->addr,
+                               (v6) ? 128 : 32, iface, NULL, NULL);
+                       set_ip_source_policy(true, v6, IPRULE_PRIORITY_NW, &a_new->addr,
+                               a_new->mask, iface, NULL, NULL);
+
                        if ((a_new->flags & DEVADDR_OFFLINK) || iface->metric)
                                interface_handle_subnet_route(iface, a_new, true);
                }
@@ -476,41 +580,6 @@ interface_update_host_route(struct vlist_tree *tree,
 
 static void
 interface_set_prefix_address(struct device_prefix_assignment *assignment,
-               const struct device_prefix *prefix, struct interface *iface, bool add);
-
-static void interface_trigger_ula_prefix(struct interface *iface,
-               const struct device_prefix *prefix, bool enable)
-{
-       if (prefix == ula_prefix || (prefix->addr.s6_addr[0] & 0xfe) != 0xfc)
-               return;
-
-       bool external_ula = false;
-       struct device_prefix_assignment *ula_assign = NULL;
-       struct device_prefix *c;
-       list_for_each_entry(c, &prefixes, head) {
-               if (c != ula_prefix && (c->addr.s6_addr[0] & 0xfe) != 0xfc)
-                       continue;
-
-               struct device_prefix_assignment *a;
-               list_for_each_entry(a, &c->assignments, head) {
-                       if (!strcmp(a->name, iface->name)) {
-                               if (c == ula_prefix)
-                                       ula_assign = a;
-                               else if (a->enabled)
-                                       external_ula = true;
-                       }
-               }
-
-       }
-
-       // Remove ULA assignment if there is an externally managed ULA and vice versa
-       if (ula_assign && ((enable && !external_ula) || (!enable && external_ula)))
-               interface_set_prefix_address(ula_assign, ula_prefix, iface, enable);
-}
-
-
-static void
-interface_set_prefix_address(struct device_prefix_assignment *assignment,
                const struct device_prefix *prefix, struct interface *iface, bool add)
 {
        const struct interface *uplink = prefix->iface;
@@ -535,11 +604,24 @@ interface_set_prefix_address(struct device_prefix_assignment *assignment,
                if (!addr.valid_until || addr.valid_until - now > 7200)
                        addr.valid_until = now + 7200;
                system_add_address(l3_downlink, &addr);
-               assignment->enabled = false;
+               if (prefix->iface) {
+                       set_ip_source_policy(false, true, IPRULE_PRIORITY_NW, &addr.addr,
+                                                       addr.mask, prefix->iface, iface, NULL);
+
+                       set_ip_source_policy(false, true, IPRULE_PRIORITY_REJECT, &addr.addr,
+                                                       addr.mask, NULL, iface, "unreachable");
+               }
 
-               interface_trigger_ula_prefix(iface, prefix, true);
+               assignment->enabled = false;
        } else if (add && (iface->state == IFS_UP || iface->state == IFS_SETUP)) {
                system_add_address(l3_downlink, &addr);
+               if (prefix->iface && !assignment->enabled) {
+                       set_ip_source_policy(true, true, IPRULE_PRIORITY_REJECT, &addr.addr,
+                                       addr.mask, NULL, iface, "unreachable");
+
+                       set_ip_source_policy(true, true, IPRULE_PRIORITY_NW, &addr.addr,
+                                       addr.mask, prefix->iface, iface, NULL);
+               }
                if (uplink && uplink->l3_dev.dev) {
                        int mtu = system_update_ipv6_mtu(
                                        uplink->l3_dev.dev, 0);
@@ -547,8 +629,6 @@ interface_set_prefix_address(struct device_prefix_assignment *assignment,
                                system_update_ipv6_mtu(l3_downlink, mtu);
                }
                assignment->enabled = true;
-
-               interface_trigger_ula_prefix(iface, prefix, false);
        }
 }
 
@@ -612,16 +692,33 @@ static void interface_update_prefix_assignments(struct device_prefix *prefix, bo
                list_add(&c->head, &prefix->assignments);
        }
 
+       bool assigned_any = false;
        struct list_head assign_later = LIST_HEAD_INIT(assign_later);
        vlist_for_each_element(&interfaces, iface, node) {
-               if (iface->config_ip.assignment_length < 48 ||
-                               iface->config_ip.assignment_length > 64)
+               if (iface->assignment_length < 48 ||
+                               iface->assignment_length > 64)
                        continue;
 
+               // Test whether there is a matching class
+               if (!list_empty(&iface->assignment_classes)) {
+                       bool found = false;
+
+                       struct interface_assignment_class *c;
+                       list_for_each_entry(c, &iface->assignment_classes, head) {
+                               if (!strcmp(c->name, prefix->pclass)) {
+                                       found = true;
+                                       break;
+                               }
+                       }
+
+                       if (!found)
+                               continue;
+               }
+
                size_t namelen = strlen(iface->name) + 1;
                c = malloc(sizeof(*c) + namelen);
-               c->length = iface->config_ip.assignment_length;
-               c->assigned = iface->config_ip.assignment_hint;
+               c->length = iface->assignment_length;
+               c->assigned = iface->assignment_hint;
                c->enabled = false;
                memcpy(c->name, iface->name, namelen);
 
@@ -632,24 +729,48 @@ static void interface_update_prefix_assignments(struct device_prefix *prefix, bo
                                netifd_log_message(L_WARNING, "Failed to assign requested subprefix "
                                                "of size %hhu for %s, trying other\n", c->length, c->name);
                        }
-                       list_add_tail(&c->head, &assign_later);
+
+                       struct list_head *next = &assign_later;
+                       struct device_prefix_assignment *n;
+                       list_for_each_entry(n, &assign_later, head) {
+                               if (n->length > c->length) {
+                                       next = &n->head;
+                                       break;
+                               }
+                       }
+                       list_add_tail(&c->head, next);
                }
+
+               if (c->assigned != -1)
+                       assigned_any = true;
        }
 
        // Then try to assign all other + failed custom assignments
        while (!list_empty(&assign_later)) {
                c = list_first_entry(&assign_later, struct device_prefix_assignment, head);
                list_del(&c->head);
-               if (!interface_prefix_assign(&prefix->assignments, c)) {
+
+               bool assigned = false;
+               do {
+                       assigned = interface_prefix_assign(&prefix->assignments, c);
+               } while (!assigned && ++c->length <= 64);
+
+               if (!assigned) {
                        netifd_log_message(L_WARNING, "Failed to assign subprefix "
                                        "of size %hhu for %s\n", c->length, c->name);
                        free(c);
+               } else {
+                       assigned_any = true;
                }
        }
 
        list_for_each_entry(c, &prefix->assignments, head)
                if ((iface = vlist_find(&interfaces, c->name, iface, node)))
                        interface_set_prefix_address(c, prefix, iface, true);
+
+       if (!assigned_any)
+               netifd_log_message(L_WARNING, "You have delegated IPv6-prefixes but haven't assigned them "
+                               "to any interface. Did you forget to set option ip6assign on your lan-interfaces?");
 }
 
 
@@ -681,6 +802,7 @@ interface_update_prefix(struct vlist_tree *tree,
        route.mask = (node_new) ? prefix_new->length : prefix_old->length;
        route.addr.in6 = (node_new) ? prefix_new->addr : prefix_old->addr;
 
+
        struct device_prefix_assignment *c;
        struct interface *iface;
 
@@ -694,12 +816,10 @@ interface_update_prefix(struct vlist_tree *tree,
        } else if (node_new) {
                // Set null-route to avoid routing loops
                system_add_route(NULL, &route);
-
                interface_update_prefix_assignments(prefix_new, true);
        } else if (node_old) {
-               interface_update_prefix_assignments(prefix_old, false);
-
                // Remove null-route
+               interface_update_prefix_assignments(prefix_old, false);
                system_del_route(NULL, &route);
        }
 
@@ -716,9 +836,12 @@ interface_update_prefix(struct vlist_tree *tree,
 struct device_prefix*
 interface_ip_add_device_prefix(struct interface *iface, struct in6_addr *addr,
                uint8_t length, time_t valid_until, time_t preferred_until,
-               struct in6_addr *excl_addr, uint8_t excl_length)
+               struct in6_addr *excl_addr, uint8_t excl_length, const char *pclass)
 {
-       struct device_prefix *prefix = calloc(1, sizeof(*prefix));
+       if (!pclass)
+               pclass = (iface) ? iface->name : "local";
+
+       struct device_prefix *prefix = calloc(1, sizeof(*prefix) + strlen(pclass) + 1);
        prefix->length = length;
        prefix->addr = *addr;
        prefix->preferred_until = preferred_until;
@@ -731,6 +854,8 @@ interface_ip_add_device_prefix(struct interface *iface, struct in6_addr *addr,
                prefix->excl_length = excl_length;
        }
 
+       strcpy(prefix->pclass, pclass);
+
        if (iface)
                vlist_add(&iface->proto_ip.prefix, &prefix->node, &prefix->addr);
        else
@@ -767,7 +892,7 @@ interface_ip_set_ula_prefix(const char *prefix)
                        interface_update_prefix(NULL, NULL, &ula_prefix->node);
 
                ula_prefix = interface_ip_add_device_prefix(NULL, &addr, length,
-                               0, 0, NULL, 0);
+                               0, 0, NULL, 0, NULL);
        }
 }
 
@@ -965,6 +1090,14 @@ void interface_ip_set_enabled(struct interface_ip_settings *ip, bool enabled)
                list_for_each_entry(a, &c->assignments, head)
                        if (!strcmp(a->name, ip->iface->name))
                                interface_set_prefix_address(a, c, ip->iface, enabled);
+
+       if (ip->iface && ip->iface->l3_dev.dev) {
+               set_ip_lo_policy(enabled, true, ip->iface);
+               set_ip_lo_policy(enabled, false, ip->iface);
+
+               set_ip_source_policy(enabled, true, IPRULE_PRIORITY_REJECT + ip->iface->l3_dev.dev->ifindex,
+                       NULL, 0, NULL, ip->iface, "failed_policy");
+       }
 }
 
 void