diff options
Diffstat (limited to 'src/platform/nmp-object.c')
| -rw-r--r-- | src/platform/nmp-object.c | 180 |
1 files changed, 90 insertions, 90 deletions
diff --git a/src/platform/nmp-object.c b/src/platform/nmp-object.c index b47bf86087..77e6808ea3 100644 --- a/src/platform/nmp-object.c +++ b/src/platform/nmp-object.c @@ -54,25 +54,25 @@ typedef struct { struct _NMPCache { /* the cache contains only one hash table for all object types, and similarly - * it contains only one NMMultiIndex. - * This works, because different object types don't ever compare equal and - * because their index ids also don't overlap. - * - * For routes and addresses, the cache contains an address if (and only if) the - * object was reported via netlink. - * For links, the cache contain a link if it was reported by either netlink - * or udev. That means, a link object can be alive, even if it was already - * removed via netlink. - * - * This effectively merges the udev-device cache into the NMPCache. - */ + * it contains only one NMMultiIndex. + * This works, because different object types don't ever compare equal and + * because their index ids also don't overlap. + * + * For routes and addresses, the cache contains an address if (and only if) the + * object was reported via netlink. + * For links, the cache contain a link if it was reported by either netlink + * or udev. That means, a link object can be alive, even if it was already + * removed via netlink. + * + * This effectively merges the udev-device cache into the NMPCache. + */ NMDedupMultiIndex *multi_idx; /* an idx_type entry for each NMP_CACHE_ID_TYPE. Note that NONE (zero) - * is skipped, so the index is shifted by one: idx_type[cache_id_type - 1]. - * - * Don't bother, use _idx_type_get() instead! */ + * is skipped, so the index is shifted by one: idx_type[cache_id_type - 1]. + * + * Don't bother, use _idx_type_get() instead! */ DedupMultiIdxType idx_types[NMP_CACHE_ID_TYPE_MAX]; gboolean use_udev; @@ -226,9 +226,9 @@ nm_sock_addr_union_to_string(const NMSockAddrUnion *sa, char *buf, gsize len) return buf; /* maybe we should use getnameinfo(), but here implement it ourself. - * - * We want to see the actual bytes for debugging (as we understand them), - * and now what getnameinfo() makes of it. Also, it's simpler this way. */ + * + * We want to see the actual bytes for debugging (as we understand them), + * and now what getnameinfo() makes of it. Also, it's simpler this way. */ switch (sa->sa.sa_family) { case AF_INET: @@ -300,7 +300,7 @@ _idx_obj_part(const DedupMultiIdxType *idx_type, NMPObjectType obj_type; /* the hash/equals functions are strongly related. So, keep them - * side-by-side and do it all in _idx_obj_part(). */ + * side-by-side and do it all in _idx_obj_part(). */ nm_assert(idx_type); nm_assert(idx_type->parent.klass == &_dedup_multi_idx_type_class); @@ -321,8 +321,8 @@ _idx_obj_part(const DedupMultiIdxType *idx_type, case NMP_CACHE_ID_TYPE_LINK_BY_IFNAME: if (NMP_OBJECT_GET_TYPE(obj_a) != NMP_OBJECT_TYPE_LINK) { /* first check, whether obj_a is suitable for this idx_type. - * If not, return 0 (which is correct for partitionable(), hash() and equal() - * functions. */ + * If not, return 0 (which is correct for partitionable(), hash() and equal() + * functions. */ if (h) nm_hash_update_val(h, obj_a); return 0; @@ -652,7 +652,7 @@ _nmp_object_fixup_link_udev_fields(NMPObject **obj_new, NMPObject *obj_orig, gbo obj = *obj_new ?: obj_orig; /* The link contains internal fields that are combined by - * properties from netlink and udev. Update those properties */ + * properties from netlink and udev. Update those properties */ /* When a link is not in netlink, its udev fields don't matter. */ if (obj->_link.netlink.is_in_netlink) { @@ -661,12 +661,12 @@ _nmp_object_fixup_link_udev_fields(NMPObject **obj_new, NMPObject *obj_orig, gbo initialized = TRUE; else if (!use_udev) { /* If we don't use udev, we immediately mark the link as initialized. - * - * For that, we consult @use_udev argument, that is cached via - * nmp_cache_use_udev_get(). It is on purpose not to test - * for a writable /sys on every call. A minor reason for that is - * performance, but the real reason is reproducibility. - * */ + * + * For that, we consult @use_udev argument, that is cached via + * nmp_cache_use_udev_get(). It is on purpose not to test + * for a writable /sys on every call. A minor reason for that is + * performance, but the real reason is reproducibility. + * */ initialized = TRUE; } } @@ -1217,9 +1217,9 @@ _vt_cmd_obj_cmp_link(const NMPObject *obj1, const NMPObject *obj2) return 1; /* Only compare based on pointer values. That is ugly because it's not a - * stable sort order. - * - * Have this check as very last. */ + * stable sort order. + * + * Have this check as very last. */ return (obj1->_link.udev.device < obj2->_link.udev.device) ? -1 : 1; } @@ -1451,7 +1451,7 @@ nmp_object_id_cmp(const NMPObject *obj1, const NMPObject *obj2) if (!klass->cmd_plobj_id_cmp) { /* the klass doesn't implement ID cmp(). That means, different objects - * never compare equal, but the cmp() according to their pointer value. */ + * never compare equal, but the cmp() according to their pointer value. */ NM_CMP_DIRECT_PTR(obj1, obj2); return 0; } @@ -1525,7 +1525,7 @@ nmp_object_id_hash_update(const NMPObject *obj, NMHashState *h) if (!klass->cmd_plobj_id_hash_update) { /* The klass doesn't implement ID compare. It means, to use pointer - * equality. */ + * equality. */ nm_hash_update_val(h, obj); return; } @@ -1679,20 +1679,20 @@ static gboolean _vt_cmd_obj_is_alive_ipx_route(const NMPObject *obj) { /* We want to ignore routes that are RTM_F_CLONED but we still - * let nmp_object_from_nl() create such route objects, instead of - * returning NULL right away. - * - * The idea is, that if we have the same route (according to its id) - * in the cache with !RTM_F_CLONED, an update that changes the route - * to be RTM_F_CLONED must remove the instance. - * - * If nmp_object_from_nl() would just return NULL, we couldn't look - * into the cache to see if it contains a route that now disappears - * (because it changed to be cloned). - * - * Instead we create a dead object, and nmp_cache_update_netlink() - * will remove the old version of the update. - **/ + * let nmp_object_from_nl() create such route objects, instead of + * returning NULL right away. + * + * The idea is, that if we have the same route (according to its id) + * in the cache with !RTM_F_CLONED, an update that changes the route + * to be RTM_F_CLONED must remove the instance. + * + * If nmp_object_from_nl() would just return NULL, we couldn't look + * into the cache to see if it contains a route that now disappears + * (because it changed to be cloned). + * + * Instead we create a dead object, and nmp_cache_update_netlink() + * will remove the old version of the update. + **/ return NMP_OBJECT_CAST_IP_ROUTE(obj)->ifindex > 0 && !NM_FLAGS_HAS(obj->ip_route.r_rtm_flags, RTM_F_CLONED); } @@ -1873,7 +1873,7 @@ nmp_cache_link_connected_needs_toggle(const NMPCache * cache, return FALSE; /* if native IFF_LOWER_UP is down, link.connected must also be down - * regardless of the slaves. */ + * regardless of the slaves. */ if (!NM_FLAGS_HAS(master->link.n_ifi_flags, IFF_LOWER_UP)) return !!master->link.connected; @@ -2305,13 +2305,13 @@ static NMDedupMultiIdxMode _obj_get_add_mode(const NMPObject *obj) { /* new objects are usually appended to the list. Except for - * addresses, which are prepended during `ip address add`. - * - * Actually, for routes it is more complicated, because depending on - * `ip route append`, `ip route replace`, `ip route prepend`, the object - * will be added at the tail, at the front, or even replace an element - * in the list. However, that is handled separately by nmp_cache_update_netlink_route() - * and of no concern here. */ + * addresses, which are prepended during `ip address add`. + * + * Actually, for routes it is more complicated, because depending on + * `ip route append`, `ip route replace`, `ip route prepend`, the object + * will be added at the tail, at the front, or even replace an element + * in the list. However, that is handled separately by nmp_cache_update_netlink_route() + * and of no concern here. */ if (NM_IN_SET(NMP_OBJECT_GET_TYPE(obj), NMP_OBJECT_TYPE_IP4_ADDRESS, NMP_OBJECT_TYPE_IP6_ADDRESS)) @@ -2385,7 +2385,7 @@ _idxcache_update_other_cache_ids(NMPCache * cache, if (obj_old && nm_dedup_multi_idx_type_id_equal(idx_type, obj_old, obj_new) && nm_dedup_multi_idx_type_partition_equal(idx_type, obj_old, obj_new)) { /* optimize. We just looked up the @obj_old entry and @obj_new compares equal - * according to idx_obj_id_equal(). entry_new is the same as entry_old. */ + * according to idx_obj_id_equal(). entry_new is the same as entry_old. */ entry_new = entry_old; } else { entry_new = nm_dedup_multi_index_lookup_obj(cache->multi_idx, idx_type, obj_new); @@ -2440,9 +2440,9 @@ _idxcache_update(NMPCache * cache, nm_auto_nmpobj const NMPObject *obj_old = NULL; /* we update an object in the cache. - * - * Note that @entry_old MUST be what is currently tracked in multi_idx, and it must - * have the same ID as @obj_new. */ + * + * Note that @entry_old MUST be what is currently tracked in multi_idx, and it must + * have the same ID as @obj_new. */ nm_assert(cache); nm_assert(entry_old || obj_new); @@ -2469,11 +2469,11 @@ _idxcache_update(NMPCache * cache, obj_old = nmp_object_ref(entry_old->obj); /* first update the main index NMP_CACHE_ID_TYPE_OBJECT_TYPE. - * We already know the pre-existing @entry old, so all that - * nm_dedup_multi_index_add_full() effectively does, is update the - * obj reference. - * - * We also get the new boxed object, which we need below. */ + * We already know the pre-existing @entry old, so all that + * nm_dedup_multi_index_add_full() effectively does, is update the + * obj reference. + * + * We also get the new boxed object, which we need below. */ if (obj_new) { nm_auto_nmpobj NMPObject *obj_old2 = NULL; @@ -2494,7 +2494,7 @@ _idxcache_update(NMPCache * cache, nm_dedup_multi_index_remove_entry(cache->multi_idx, entry_old); /* now update all other indexes. We know the previously boxed entry, and the - * newly boxed one. */ + * newly boxed one. */ klass = NMP_OBJECT_GET_CLASS(entry_new ? entry_new->obj : obj_old); for (i_idx_type = klass->supported_cache_ids; *i_idx_type; i_idx_type++) { NMPCacheIdType id_type = *i_idx_type; @@ -2534,7 +2534,7 @@ nmp_cache_remove(NMPCache * cache, if (equals_by_ptr && obj_old != obj_needle) { /* We found an identical object, but we only delete it if it's the same pointer as - * @obj_needle. */ + * @obj_needle. */ return NMP_CACHE_OPS_UNCHANGED; } if (only_dirty && !entry_old->dirty) { @@ -2568,9 +2568,9 @@ nmp_cache_remove_netlink(NMPCache * cache, if (NMP_OBJECT_GET_TYPE(obj_needle) == NMP_OBJECT_TYPE_LINK) { /* For nmp_cache_remove_netlink() we have an incomplete @obj_needle instance to be - * removed from netlink. Link objects are alive without being in netlink when they - * have a udev-device. All we want to do in this case is clear the netlink.is_in_netlink - * flag. */ + * removed from netlink. Link objects are alive without being in netlink when they + * have a udev-device. All we want to do in this case is clear the netlink.is_in_netlink + * flag. */ NM_SET_OUT(out_obj_old, nmp_object_ref(obj_old)); @@ -2647,8 +2647,8 @@ nmp_cache_update_netlink(NMPCache * cache, nm_assert(NMP_OBJECT_IS_VALID(obj_hand_over)); nm_assert(!NMP_OBJECT_IS_STACKINIT(obj_hand_over)); /* A link object from netlink must have the udev related fields unset. - * We could implement to handle that, but there is no need to support such - * a use-case */ + * We could implement to handle that, but there is no need to support such + * a use-case */ nm_assert(NMP_OBJECT_GET_TYPE(obj_hand_over) != NMP_OBJECT_TYPE_LINK || (!obj_hand_over->_link.udev.device && !obj_hand_over->link.driver)); nm_assert(nm_dedup_multi_index_obj_find(cache->multi_idx, obj_hand_over) != obj_hand_over); @@ -2685,14 +2685,14 @@ nmp_cache_update_netlink(NMPCache * cache, } if (obj_old->_link.udev.device) { /* @obj_hand_over is not in netlink. - * - * This is similar to nmp_cache_remove_netlink(), but there we preserve the - * preexisting netlink properties. The use case of that is when kernel_get_object() - * cannot load an object (based on the id of a needle). - * - * Here we keep the data provided from @obj_hand_over. The usecase is when receiving - * a valid @obj_hand_over instance from netlink with RTM_DELROUTE. - */ + * + * This is similar to nmp_cache_remove_netlink(), but there we preserve the + * preexisting netlink properties. The use case of that is when kernel_get_object() + * cannot load an object (based on the id of a needle). + * + * Here we keep the data provided from @obj_hand_over. The usecase is when receiving + * a valid @obj_hand_over instance from netlink with RTM_DELROUTE. + */ is_alive = TRUE; } else is_alive = FALSE; @@ -2765,8 +2765,8 @@ nmp_cache_update_netlink_route(NMPCache * cache, nm_assert(NMP_OBJECT_IS_VALID(obj_hand_over)); nm_assert(!NMP_OBJECT_IS_STACKINIT(obj_hand_over)); /* A link object from netlink must have the udev related fields unset. - * We could implement to handle that, but there is no need to support such - * a use-case */ + * We could implement to handle that, but there is no need to support such + * a use-case */ nm_assert(NM_IN_SET(NMP_OBJECT_GET_TYPE(obj_hand_over), NMP_OBJECT_TYPE_IP4_ROUTE, NMP_OBJECT_TYPE_IP6_ROUTE)); @@ -2809,12 +2809,12 @@ update_done: NM_SET_OUT(out_obj_new, nmp_object_ref(nm_dedup_multi_entry_get_obj(entry_new))); /* a RTM_GETROUTE event may signal that another object was replaced. - * Find out whether that is the case and return it as @obj_replaced. - * - * Also, fixup the order of @entry_new within NMP_CACHE_ID_TYPE_ROUTES_BY_WEAK_ID - * index. For most parts, we don't care about the order of objects (including routes). - * But NMP_CACHE_ID_TYPE_ROUTES_BY_WEAK_ID we must keep in the correct order, to - * properly find @obj_replaced. */ + * Find out whether that is the case and return it as @obj_replaced. + * + * Also, fixup the order of @entry_new within NMP_CACHE_ID_TYPE_ROUTES_BY_WEAK_ID + * index. For most parts, we don't care about the order of objects (including routes). + * But NMP_CACHE_ID_TYPE_ROUTES_BY_WEAK_ID we must keep in the correct order, to + * properly find @obj_replaced. */ resync_required = FALSE; entry_replace = NULL; if (is_dump) @@ -2824,15 +2824,15 @@ update_done: if (NM_FLAGS_HAS(nlmsgflags, NLM_F_REPLACE) && nmp_cache_lookup_all(cache, NMP_CACHE_ID_TYPE_ROUTES_BY_WEAK_ID, obj_hand_over)) { /* hm. @obj_hand_over was not added, meaning it was not alive. - * However, we track some other objects with the same weak-id. - * It's unclear what that means. To be sure, resync. */ + * However, we track some other objects with the same weak-id. + * It's unclear what that means. To be sure, resync. */ resync_required = TRUE; } goto out; } /* FIXME: for routes, we only maintain the order correctly for the BY_WEAK_ID - * index. For all other indexes their order becomes messed up. */ + * index. For all other indexes their order becomes messed up. */ entry_cur = _lookup_entry_with_idx_type(cache, NMP_CACHE_ID_TYPE_ROUTES_BY_WEAK_ID, entry_new->obj); if (!entry_cur) { |