diff options
Diffstat (limited to 'src/platform/nm-platform.h')
| -rw-r--r-- | src/platform/nm-platform.h | 330 |
1 files changed, 165 insertions, 165 deletions
diff --git a/src/platform/nm-platform.h b/src/platform/nm-platform.h index 441c4828e8..c01008018a 100644 --- a/src/platform/nm-platform.h +++ b/src/platform/nm-platform.h @@ -73,7 +73,7 @@ typedef enum { NMP_NLM_FLAG_F_ECHO = 0x08, /* NLM_F_ECHO, Echo this request */ /* use our own platform enum for the nlmsg-flags. Otherwise, we'd have - * to include <linux/netlink.h> */ + * to include <linux/netlink.h> */ NMP_NLM_FLAG_F_REPLACE = 0x100, /* NLM_F_REPLACE, Override existing */ NMP_NLM_FLAG_F_EXCL = 0x200, /* NLM_F_EXCL, Do not touch, if it exists */ NMP_NLM_FLAG_F_CREATE = 0x400, /* NLM_F_CREATE, Create, if it does not exist */ @@ -82,15 +82,15 @@ typedef enum { NMP_NLM_FLAG_FMASK = 0xFFFF, /* a mask for all NMP_NLM_FLAG_F_* flags */ /* instructs NM to suppress logging an error message for any failures - * received from kernel. - * - * It will still log with debug-level, and it will still log - * other failures aside the kernel response. */ + * received from kernel. + * + * It will still log with debug-level, and it will still log + * other failures aside the kernel response. */ NMP_NLM_FLAG_SUPPRESS_NETLINK_FAILURE = 0x10000, /* the following aliases correspond to iproute2's `ip route CMD` for - * RTM_NEWROUTE, with CMD being one of add, change, replace, prepend, - * append and test. */ + * RTM_NEWROUTE, with CMD being one of add, change, replace, prepend, + * append and test. */ NMP_NLM_FLAG_ADD = NMP_NLM_FLAG_F_CREATE | NMP_NLM_FLAG_F_EXCL, NMP_NLM_FLAG_CHANGE = NMP_NLM_FLAG_F_REPLACE, NMP_NLM_FLAG_REPLACE = NMP_NLM_FLAG_F_CREATE | NMP_NLM_FLAG_F_REPLACE, @@ -101,40 +101,40 @@ typedef enum { typedef enum { /* compare fields which kernel considers as similar routes. - * It is a looser comparisong then NM_PLATFORM_IP_ROUTE_CMP_TYPE_ID - * and means that `ip route add` would fail to add two routes - * that have the same NM_PLATFORM_IP_ROUTE_CMP_TYPE_WEAK_ID. - * On the other hand, `ip route append` would allow that, as - * long as NM_PLATFORM_IP_ROUTE_CMP_TYPE_ID differs. */ + * It is a looser comparisong then NM_PLATFORM_IP_ROUTE_CMP_TYPE_ID + * and means that `ip route add` would fail to add two routes + * that have the same NM_PLATFORM_IP_ROUTE_CMP_TYPE_WEAK_ID. + * On the other hand, `ip route append` would allow that, as + * long as NM_PLATFORM_IP_ROUTE_CMP_TYPE_ID differs. */ NM_PLATFORM_IP_ROUTE_CMP_TYPE_WEAK_ID, /* compare two routes as kernel would allow to add them with - * `ip route append`. In other words, kernel does not allow you to - * add two routes (at the same time) which compare equal according - * to NM_PLATFORM_IP_ROUTE_CMP_TYPE_ID. - * - * For the ID we can only recognize route fields that we actually implement. - * However, kernel supports more routing options, some of them also part of - * the ID. NetworkManager is oblivious to these options and will wrongly think - * that two routes are identical, while they are not. That can lead to an - * inconsistent platform cache. Not much what we can do about that, except - * implementing all options that kernel supports *sigh*. See rh#1337860. - */ + * `ip route append`. In other words, kernel does not allow you to + * add two routes (at the same time) which compare equal according + * to NM_PLATFORM_IP_ROUTE_CMP_TYPE_ID. + * + * For the ID we can only recognize route fields that we actually implement. + * However, kernel supports more routing options, some of them also part of + * the ID. NetworkManager is oblivious to these options and will wrongly think + * that two routes are identical, while they are not. That can lead to an + * inconsistent platform cache. Not much what we can do about that, except + * implementing all options that kernel supports *sigh*. See rh#1337860. + */ NM_PLATFORM_IP_ROUTE_CMP_TYPE_ID, /* compare all fields as they make sense for kernel. For example, - * a route destination 192.168.1.5/24 is not accepted by kernel and - * we treat it identical to 192.168.1.0/24. Semantically these - * routes are identical, but NM_PLATFORM_IP_ROUTE_CMP_TYPE_FULL will - * report them as different. - * - * The result shall be identical to call first nm_platform_ip_route_normalize() - * on both routes and then doing a full comparison. */ + * a route destination 192.168.1.5/24 is not accepted by kernel and + * we treat it identical to 192.168.1.0/24. Semantically these + * routes are identical, but NM_PLATFORM_IP_ROUTE_CMP_TYPE_FULL will + * report them as different. + * + * The result shall be identical to call first nm_platform_ip_route_normalize() + * on both routes and then doing a full comparison. */ NM_PLATFORM_IP_ROUTE_CMP_TYPE_SEMANTICALLY, /* compare all fields. This should have the same effect as memcmp(), - * except allowing for undefined data in holes between field alignment. - */ + * except allowing for undefined data in holes between field alignment. + */ NM_PLATFORM_IP_ROUTE_CMP_TYPE_FULL, } NMPlatformIPRouteCmpType; @@ -158,16 +158,16 @@ GBytes * nmp_link_address_get_as_bytes(const NMPLinkAddress *addr); typedef enum { /* match-flags are strictly inclusive. That means, - * by default nothing is matched, but if you enable a particular - * flag, a candidate that matches passes the check. - * - * In other words: adding more flags can only extend the result - * set of matching objects. - * - * Also, the flags form partitions. Like, an address can be either of - * ADDRTYPE_NORMAL or ADDRTYPE_LINKLOCAL, but never both. Same for - * the ADDRSTATE match types. - */ + * by default nothing is matched, but if you enable a particular + * flag, a candidate that matches passes the check. + * + * In other words: adding more flags can only extend the result + * set of matching objects. + * + * Also, the flags form partitions. Like, an address can be either of + * ADDRTYPE_NORMAL or ADDRTYPE_LINKLOCAL, but never both. Same for + * the ADDRSTATE match types. + */ NM_PLATFORM_MATCH_WITH_NONE = 0, NM_PLATFORM_MATCH_WITH_ADDRTYPE_NORMAL = (1LL << 0), @@ -187,7 +187,7 @@ typedef enum { struct _NMPlatformObject { /* the object type has no fields of its own, it is only used to having - * a special pointer type that can be used to indicate "any" type. */ + * a special pointer type that can be used to indicate "any" type. */ char _dummy_don_t_use_me; }; @@ -214,8 +214,8 @@ struct _NMPlatformLink { int master; /* rtnl_link_get_link(), IFLA_LINK. - * If IFLA_LINK_NETNSID indicates that the parent is in another namespace, - * this field be set to (negative) NM_PLATFORM_LINK_OTHER_NETNS. */ + * If IFLA_LINK_NETNSID indicates that the parent is in another namespace, + * this field be set to (negative) NM_PLATFORM_LINK_OTHER_NETNS. */ int parent; /* IFF_* flags. Note that the flags in 'struct ifinfomsg' are declared as 'unsigned'. */ @@ -236,8 +236,8 @@ struct _NMPlatformLink { NMUtilsIPv6IfaceId inet6_token; /* The bitwise inverse of rtnl_link_inet6_get_addr_gen_mode(). It is inverse - * to have a default of 0 -- meaning: unspecified. That way, a struct - * initialized with memset(0) has and unset value.*/ + * to have a default of 0 -- meaning: unspecified. That way, a struct + * initialized with memset(0) has and unset value.*/ guint8 inet6_addr_gen_mode_inv; /* Statistics */ @@ -247,7 +247,7 @@ struct _NMPlatformLink { guint64 tx_bytes; /* @connected is mostly identical to (@n_ifi_flags & IFF_UP). Except for bridge/bond masters, - * where we coerce the link as disconnect if it has no slaves. */ + * where we coerce the link as disconnect if it has no slaves. */ bool connected : 1; bool initialized : 1; @@ -287,35 +287,35 @@ typedef enum { NMIPConfigSource addr_source; \ \ /* Timestamp in seconds in the reference system of nm_utils_get_monotonic_timestamp_*(). - * - * The rules are: - * 1 @lifetime==0: @timestamp and @preferred is irrelevant (but mostly set to 0 too). Such addresses - * are permanent. This rule is so that unset addresses (calloc) are permanent by default. - * 2 @lifetime==@preferred==NM_PLATFORM_LIFETIME_PERMANENT: @timestamp is irrelevant (but mostly - * set to 0). Such addresses are permanent. - * 3 Non permanent addresses should (almost) always have @timestamp > 0. 0 is not a valid timestamp - * and never returned by nm_utils_get_monotonic_timestamp_sec(). In this case @valid/@preferred - * is anchored at @timestamp. - * 4 Non permanent addresses with @timestamp == 0 are implicitly anchored at *now*, thus the time - * moves as time goes by. This is usually not useful, except e.g. nm_platform_ip[46]_address_add(). - * - * Non permanent addresses from DHCP/RA might have the @timestamp set to the moment of when the - * lease was received. Addresses from kernel might have the @timestamp based on the last modification - * time of the addresses. But don't rely on this behaviour, the @timestamp is only defined for anchoring - * @lifetime and @preferred. - */ \ + * + * The rules are: + * 1 @lifetime==0: @timestamp and @preferred is irrelevant (but mostly set to 0 too). Such addresses + * are permanent. This rule is so that unset addresses (calloc) are permanent by default. + * 2 @lifetime==@preferred==NM_PLATFORM_LIFETIME_PERMANENT: @timestamp is irrelevant (but mostly + * set to 0). Such addresses are permanent. + * 3 Non permanent addresses should (almost) always have @timestamp > 0. 0 is not a valid timestamp + * and never returned by nm_utils_get_monotonic_timestamp_sec(). In this case @valid/@preferred + * is anchored at @timestamp. + * 4 Non permanent addresses with @timestamp == 0 are implicitly anchored at *now*, thus the time + * moves as time goes by. This is usually not useful, except e.g. nm_platform_ip[46]_address_add(). + * + * Non permanent addresses from DHCP/RA might have the @timestamp set to the moment of when the + * lease was received. Addresses from kernel might have the @timestamp based on the last modification + * time of the addresses. But don't rely on this behaviour, the @timestamp is only defined for anchoring + * @lifetime and @preferred. + */ \ guint32 timestamp; \ guint32 lifetime; /* seconds since timestamp */ \ guint32 preferred; /* seconds since timestamp */ \ \ /* ifa_flags in 'struct ifaddrmsg' from <linux/if_addr.h>, extended to 32 bit by - * IFA_FLAGS attribute. */ \ + * IFA_FLAGS attribute. */ \ guint32 n_ifa_flags; \ \ guint8 plen; \ \ /* FIXME(l3cfg): the external marker won't be necessary anymore, because we only - * merge addresses we care about, and ignore (don't remove) external addresses. */ \ + * merge addresses we care about, and ignore (don't remove) external addresses. */ \ bool external : 1; \ \ bool use_ip4_broadcast_address : 1; \ @@ -345,20 +345,20 @@ struct _NMPlatformIP4Address { in_addr_t address; /* The IFA_ADDRESS PTP peer address. This field is rather important, because - * it constitutes the identifier for the IPv4 address (e.g. you can add two - * addresses that only differ by their peer's network-part. - * - * Beware that for most cases, NetworkManager doesn't want to set an explicit - * peer-address. However, that corresponds to setting the peer address to @address - * itself. Leaving peer-address unset/zero, means explicitly setting the peer - * address to 0.0.0.0, which you probably don't want. - * */ + * it constitutes the identifier for the IPv4 address (e.g. you can add two + * addresses that only differ by their peer's network-part. + * + * Beware that for most cases, NetworkManager doesn't want to set an explicit + * peer-address. However, that corresponds to setting the peer address to @address + * itself. Leaving peer-address unset/zero, means explicitly setting the peer + * address to 0.0.0.0, which you probably don't want. + * */ in_addr_t peer_address; /* PTP peer address */ /* IFA_BROADCAST. - * - * This parameter is ignored unless use_ip4_broadcast_address is TRUE. - * See nm_platform_ip4_broadcast_address_from_addr(). */ + * + * This parameter is ignored unless use_ip4_broadcast_address is TRUE. + * See nm_platform_ip4_broadcast_address_from_addr(). */ in_addr_t broadcast_address; char label[NMP_IFNAMSIZ]; @@ -407,32 +407,32 @@ typedef union { __NMPlatformObjWithIfindex_COMMON; \ \ /* The NMIPConfigSource. For routes that we receive from cache this corresponds - * to the rtm_protocol field (and is one of the NM_IP_CONFIG_SOURCE_RTPROT_* values). - * When adding a route, the source will be coerced to the protocol using - * nmp_utils_ip_config_source_coerce_to_rtprot(). - * - * rtm_protocol is part of the primary key of an IPv4 route (meaning, you can add - * two IPv4 routes that only differ in their rtm_protocol. For IPv6, that is not - * the case. - * - * When deleting an IPv4/IPv6 route, the rtm_protocol field must match (even - * if it is not part of the primary key for IPv6) -- unless rtm_protocol is set - * to zero, in which case the first matching route (with proto ignored) is deleted. */ \ + * to the rtm_protocol field (and is one of the NM_IP_CONFIG_SOURCE_RTPROT_* values). + * When adding a route, the source will be coerced to the protocol using + * nmp_utils_ip_config_source_coerce_to_rtprot(). + * + * rtm_protocol is part of the primary key of an IPv4 route (meaning, you can add + * two IPv4 routes that only differ in their rtm_protocol. For IPv6, that is not + * the case. + * + * When deleting an IPv4/IPv6 route, the rtm_protocol field must match (even + * if it is not part of the primary key for IPv6) -- unless rtm_protocol is set + * to zero, in which case the first matching route (with proto ignored) is deleted. */ \ NMIPConfigSource rt_source; \ \ guint8 plen; \ \ /* RTA_METRICS: - * - * For IPv4 routes, these properties are part of their - * ID (meaning: you can add otherwise identical IPv4 routes that - * only differ by the metric property). - * On the other hand, for IPv6 you cannot add two IPv6 routes that only differ - * by an RTA_METRICS property. - * - * When deleting a route, kernel seems to ignore the RTA_METRICS properties. - * That is a problem/bug for IPv4 because you cannot explicitly select which - * route to delete. Kernel just picks the first. See rh#1475642. */ \ + * + * For IPv4 routes, these properties are part of their + * ID (meaning: you can add otherwise identical IPv4 routes that + * only differ by the metric property). + * On the other hand, for IPv6 you cannot add two IPv6 routes that only differ + * by an RTA_METRICS property. + * + * When deleting a route, kernel seems to ignore the RTA_METRICS properties. + * That is a problem/bug for IPv4 because you cannot explicitly select which + * route to delete. Kernel just picks the first. See rh#1475642. */ \ \ /* RTA_METRICS.RTAX_LOCK (iproute2: "lock" arguments) */ \ bool lock_window : 1; \ @@ -442,15 +442,15 @@ typedef union { bool lock_mtu : 1; \ \ /* rtnh_flags - * - * Routes with rtm_flags RTM_F_CLONED are hidden by platform and - * do not exist from the point-of-view of platform users. - * Such a route is not alive, according to nmp_object_is_alive(). - * - * NOTE: currently we ignore all flags except RTM_F_CLONED - * and RTNH_F_ONLINK. - * We also may not properly consider the flags as part of the ID - * in route-cmp. */ \ + * + * Routes with rtm_flags RTM_F_CLONED are hidden by platform and + * do not exist from the point-of-view of platform users. + * Such a route is not alive, according to nmp_object_is_alive(). + * + * NOTE: currently we ignore all flags except RTM_F_CLONED + * and RTNH_F_ONLINK. + * We also may not properly consider the flags as part of the ID + * in route-cmp. */ \ unsigned r_rtm_flags; \ \ /* RTA_METRICS.RTAX_ADVMSS (iproute2: advmss) */ \ @@ -475,17 +475,17 @@ typedef union { guint32 metric; \ \ /* rtm_table, RTA_TABLE. - * - * This is not the original table ID. Instead, 254 (RT_TABLE_MAIN) and - * zero (RT_TABLE_UNSPEC) are swapped, so that the default is the main - * table. Use nm_platform_route_table_coerce()/nm_platform_route_table_uncoerce(). */ \ + * + * This is not the original table ID. Instead, 254 (RT_TABLE_MAIN) and + * zero (RT_TABLE_UNSPEC) are swapped, so that the default is the main + * table. Use nm_platform_route_table_coerce()/nm_platform_route_table_uncoerce(). */ \ guint32 table_coerced; \ \ /* rtm_type. - * - * This is not the original type, if type_coerced is 0 then - * it means RTN_UNSPEC otherwise the type value is preserved. - * */ \ + * + * This is not the original type, if type_coerced is 0 then + * it means RTN_UNSPEC otherwise the type value is preserved. + * */ \ guint8 type_coerced; \ \ /*end*/ @@ -515,29 +515,29 @@ struct _NMPlatformIP4Route { in_addr_t gateway; /* RTA_PREFSRC (called "src" by iproute2). - * - * pref_src is part of the ID of an IPv4 route. When deleting a route, - * pref_src must match, unless set to 0.0.0.0 to match any. */ + * + * pref_src is part of the ID of an IPv4 route. When deleting a route, + * pref_src must match, unless set to 0.0.0.0 to match any. */ in_addr_t pref_src; /* rtm_tos (iproute2: tos) - * - * For IPv4, tos is part of the weak-id (like metric). - * - * For IPv6, tos is ignored by kernel. */ + * + * For IPv4, tos is part of the weak-id (like metric). + * + * For IPv6, tos is ignored by kernel. */ guint8 tos; /* The bitwise inverse of the route scope rtm_scope. It is inverted so that the - * default value (RT_SCOPE_NOWHERE) is zero. Use nm_platform_route_scope_inv() - * to convert back and forth between the inverse representation and the - * real value. - * - * rtm_scope is part of the primary key for IPv4 routes. When deleting a route, - * the scope must match, unless it is left at RT_SCOPE_NOWHERE, in which case the first - * matching route is deleted. - * - * For IPv6 routes, the scope is ignored and kernel always assumes global scope. - * Hence, this field is only in NMPlatformIP4Route. */ + * default value (RT_SCOPE_NOWHERE) is zero. Use nm_platform_route_scope_inv() + * to convert back and forth between the inverse representation and the + * real value. + * + * rtm_scope is part of the primary key for IPv4 routes. When deleting a route, + * the scope must match, unless it is left at RT_SCOPE_NOWHERE, in which case the first + * matching route is deleted. + * + * For IPv6 routes, the scope is ignored and kernel always assumes global scope. + * Hence, this field is only in NMPlatformIP4Route. */ guint8 scope_inv; }; @@ -549,28 +549,28 @@ struct _NMPlatformIP6Route { struct in6_addr gateway; /* RTA_PREFSRC (called "src" by iproute2). - * - * pref_src is not part of the ID for an IPv6 route. You cannot add two - * routes that only differ by pref_src. - * - * When deleting a route, pref_src is ignored by kernel. */ + * + * pref_src is not part of the ID for an IPv6 route. You cannot add two + * routes that only differ by pref_src. + * + * When deleting a route, pref_src is ignored by kernel. */ struct in6_addr pref_src; /* RTA_SRC and rtm_src_len (called "from" by iproute2). - * - * Kernel clears the host part of src/src_plen. - * - * src/src_plen is part of the ID of a route just like network/plen. That is, - * Not only `ip route append`, but also `ip route add` allows to add routes that only - * differ in their src/src_plen. - */ + * + * Kernel clears the host part of src/src_plen. + * + * src/src_plen is part of the ID of a route just like network/plen. That is, + * Not only `ip route append`, but also `ip route add` allows to add routes that only + * differ in their src/src_plen. + */ struct in6_addr src; guint8 src_plen; /* RTA_PREF router preference. - * - * The type is guint8 to keep the struct size small. But the values are compatible with - * the NMIcmpv6RouterPref enum. */ + * + * The type is guint8 to keep the struct size small. But the values are compatible with + * the NMIcmpv6RouterPref enum. */ guint8 rt_pref; }; @@ -674,7 +674,7 @@ typedef struct { __NMPlatformObjWithIfindex_COMMON; /* beware, kind is embedded in an NMPObject, hence you must - * take care of the lifetime of the string. */ + * take care of the lifetime of the string. */ const char *kind; int addr_family; @@ -702,7 +702,7 @@ typedef struct { typedef struct { /* beware, kind is embedded in an NMPObject, hence you must - * take care of the lifetime of the string. */ + * take care of the lifetime of the string. */ const char *kind; union { @@ -718,7 +718,7 @@ typedef struct { __NMPlatformObjWithIfindex_COMMON; /* beware, kind is embedded in an NMPObject, hence you must - * take care of the lifetime of the string. */ + * take care of the lifetime of the string. */ const char *kind; int addr_family; @@ -988,7 +988,7 @@ typedef enum { NM_PLATFORM_KERNEL_SUPPORT_TYPE_IFLA_BR_VLAN_STATS_ENABLED, /* this also includes FRA_SPORT_RANGE and FRA_DPORT_RANGE which - * were added at the same time. */ + * were added at the same time. */ NM_PLATFORM_KERNEL_SUPPORT_TYPE_FRA_IP_PROTO, _NM_PLATFORM_KERNEL_SUPPORT_NUM, @@ -1270,7 +1270,7 @@ nm_platform_ip4_broadcast_address_from_addr(const NMPlatformIP4Address *addr) return addr->broadcast_address; /* the set broadcast-address gets ignored, and we determine a default brd base - * on the peer IFA_ADDRESS. */ + * on the peer IFA_ADDRESS. */ if (addr->peer_address != 0u && addr->plen < 31 /* RFC3021 */) return nm_platform_ip4_broadcast_address_create(addr->peer_address, addr->plen); return 0u; @@ -1290,11 +1290,11 @@ static inline guint32 nm_platform_route_table_coerce(guint32 table) { /* For kernel, the default table is RT_TABLE_MAIN (254). - * We want that in NMPlatformIPRoute.table_coerced a numeric - * zero is the default. Hence, @table_coerced swaps the - * value 0 and 254. Use nm_platform_route_table_coerce() - * and nm_platform_route_table_uncoerce() to convert between - * the two domains. */ + * We want that in NMPlatformIPRoute.table_coerced a numeric + * zero is the default. Hence, @table_coerced swaps the + * value 0 and 254. Use nm_platform_route_table_coerce() + * and nm_platform_route_table_uncoerce() to convert between + * the two domains. */ switch (table) { case 0 /* RT_TABLE_UNSPEC */: return 254; @@ -1334,13 +1334,13 @@ static inline gboolean nm_platform_route_table_is_main(guint32 table) { /* same as - * nm_platform_route_table_uncoerce (table, TRUE) == RT_TABLE_MAIN - * and - * nm_platform_route_table_uncoerce (nm_platform_route_table_coerce (table), TRUE) == RT_TABLE_MAIN - * - * That is, the function operates the same on @table and its coerced - * form. - */ + * nm_platform_route_table_uncoerce (table, TRUE) == RT_TABLE_MAIN + * and + * nm_platform_route_table_uncoerce (nm_platform_route_table_coerce (table), TRUE) == RT_TABLE_MAIN + * + * That is, the function operates the same on @table and its coerced + * form. + */ return table == 0 || table == 254; } |