summaryrefslogtreecommitdiffstats
path: root/bgpd/bgp_fsm.c
Commit message (Collapse)AuthorAgeFilesLines
* Fix bug in NOTIFICATION handling before FSM reaches Established state.ex07Chris Hall2010-09-101-3/+4
| | | | | | | | | | | | | | | | | | | Change removes the setting of session->active to false which was being done before a NOTIFICATION message is sent. This flag should remain true when the session is not being stopped -- which is the case if a NOTIFICATION is sent before Established state is reached. Effect of this bug was to trip up the accept() side of the session, bringing bgpd down on an assert(). This bug may be triggered by a peer that accepts a connection and then remains silent, for whatever reason -- causing bgpd to issue an HoldTimer Expired NOTIFICATION. Version advanced to 0.99.15ex07. Other changes purely cosmetic -- eg changing some 'int' to 'bool', and a few small documentation edits.
* Fix various capability flags so that sessions are restarted.Chris Hall2010-08-041-2/+18
| | | | | | | | | | | | | | | | | | For some reason (unknown) changing the following did NOT restart any active session: neighbor dont-capability-negotiate neighbor override-capability neighbor strict-capability-match since these can all change the nature of a session, this seems to be a (long time) bug, and has been fixed. Also, per RFC 5492: if (during OpenSent and OpenConfirm) get a NOTIFICATION: "Open/Unsupported Optional Parameter", treat that as "don't understand capabilities" and suppress sending of capabilities. Advanced to "ex05".
* Reworking of peer state handling.Chris Hall2010-07-191-0/+16
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | This fixes a reported assert() in 'no neighbor xxx'. It also fixes other issues found when reviewing and testing that. Also fixed is a reported segfault caused by prefix lists without explicit sequence numbers. Peer State ========== Tightened up the peer state handling, including: * shades of pIdle -- depending on some peer flags and the session, if any * state transitions -- particularly pEstablished -> pIdle or pClearing * handling if deleting peer and associated session * handling of PEER_DOWN_XXX -- why peer was last downed * handling of last NOTIFICATION sent/received RS Client RIBs ============== Cleared up places where RS Client RIBs were not disposed of properly when a peer's afi/safi state changed, in particular: -- when deactivating an afi/safi -- when unsetting the rs client state for an afi/safi -- when binding a peer to a group for an afi/safi In the past these issues were probably invisible, except for a slight leak of memory. With the newer code these issues triggered some asserts when closing down a peer or the entire program. Program Termination =================== Now terminates by deleting all peers -- essentially 'no neighbor' for all peers. Each time a peer is set to be deleted bm->peer_linger_count is incremented, when the peer finally is deleted, the count is decremented. So while in program_terminating state, all nexuses continue to run until the Routing Engine spots that there are no peers left to delete. Then all nexuses are shut down and the program finally terminates. To do this, when termination starts a new Routing Engine foreground task is added, which calls program_terminate_if_all_peers_deleted(). Accept() Status for Session =========================== The accept() code needs to find the session associated with the incoming IP address. Then it needs to see if can accept the incoming connection. It looks up the IP address in the Peer Index (under its Mutex). Previously: the Peer Index entry contained a pointer to the accepting connection (if any), and the session had a pointer to the Peer Index entry so that it could set/clear the accept field in that entry. Now: have removed the accept field in the Peer Index entry, and the pointer from the session to that entry. There is now an "accept" flag in the session structure -- so setting/clearing accept does not have to fiddle with any state to do with the peer. Which seems cleaner. To find the session, the accept() code goes via the Peer Index and then via the peer to find the session. This is done under the Peer Index Mutex. To support that, the Routing Engine only changes the peer->session pointer under the same Mutex. General Changes =============== 1. name changes: peer_lock -> bgp_peer_lock etc. 2. bgp_notify_send -> bgp_peer_down_error bgp_notify_send_with_data -> bgp_peer_down_error_with_data 3. name changes: bgp_peer_sIdle -> bgp_peer_pIdle etc. 4. changing int to bool 5. added "received" flag to bgp_notify structure Files Affected ============== configure.ac -- set version to 0.99.15ex02 bgpd/bgp.h -- format changes only bgpd/bgp_advertise.c -- (1) bgpd/bgp_attr.c -- (2) bgpd/bgp_common.h -- (3) bgpd/bgp_connection.c -- (4) for session->active - adding session->accept flag - removing peer index accept entry - adding bgp_connection_query_accept() bgpd/bgp_connection.h -- adding bgp_connection_query_accept() bgpd/bgp_debug.c -- (3) (5) - changed bgp_notify_print() to remove "sending" parameter and use (5) bgpd/bgp_debug.h -- changed bgp_notify_print() bgpd/bgp_fsm.c -- added TODO for NSF and for CollisionDetectEstablishedState bgpd/bgp_main.c -- (4) for various flags - added static bool program_terminating - used "mqb_priority" name instead of "1" - ignore SIGHUP and SIGTERM messages once is "program_terminating" - added program_terminate_if_all_peers_deleted() - in SIGTERM: set program_terminating and add the foreground hook. bgpd/bgp_mplsvpn.c -- (3) bgpd/bgp_msg_read.c -- (5) set the "received" flag on incoming notifications - update call of bgp_notify_print() bgpd/bgp_msg_write.c -- update call of bgp_notify_print() bgpd/bgp_network.c -- (4) bgpd/bgp_nexthop.c -- (3) bgpd/bgp_notification.c -- add "received" flag to notification structure, which is false by default. bgpd/bgp_notification.h -- add "received" flag to notification structure - add bgp_notify_set_received() - add bgp_notify_get_received() bgpd/bgp_open.c -- (2) bgpd/bgp_open_state.c -- in bgp_peer_open_state_receive(): - copy the session->hold_timer_interval and session->keepalive_timer_interval values (as negotiated by the BGP Engine) into the peer - set PEER_CAP_RESTART_RCV if have - fix typo, use: open_recv->can_preserve not: open_recv->can_g_restart bgpd/bgp_packet.c -- delete bgp_notify_send() and bgp_notify_send_with_data() - (1) (2) (3) - bgp_clear_route_normal() -> bgp_clear_routes() and now returns "completed" state. bgpd/bgp_packet.h -- delete: bgp_keepalive_send() bgp_open_send() bgp_notify_send() bgp_notify_send_with_data() bgpd/bgp_peer.c -- changed: bgp_session_has_established() bgp_session session() bgp_session_has_disabled() to void and to take session, not peer. - removed lock of session structure in bgp_session_do_event() -- was holding the lock for far longer than necessary, particularly when clearing routes ! - in bgp_session_has_established(): - lock session structure where and only where required - tidied up timer handling - in bgp_session_has_stopped(): - extend IdleHoldTime - examine notification etc to see why session came down. - invoke bgp_peer_down_notify(), which will start the process of downing the peer. - in bgp_session_has_disabled(): - removed defer_enable - removed calls to bgp_peer_stop() etc. That is done in bgp_peer_down() et all. - now if session is marked delete_me, then do that; otherwise, can enable again. - removed program_terminate_if_all_disabled(), replaced by new mechanism - in bgp_peer_stop(): - changed to void function and added nsf parameter. - sets pClearing state. - MUST now only be called when pEstablished. - removed some code to bgp_peer_reset_idle(). So... bgp_peer_stop() brings pEstablished peer to halt, while bgp_peer_reset_idle() prepares it for new session. - now passes nsf to bgp_clear_all_routes(), which returns flag to say whether task is complete or whether it continues in background. - clearing of NSF_MODE and flags moved to bgp_peer_reset_idle(). - sets pIdle state if route clearing completed - renamed peer_nsf_stop() -> bgp_peer_nsf_stop() - if is pIdle or pClearing and have NSF routes, then stops timers and clears out all the NSF routes. - added bgp_peer_clear_all_stale_routes() - added bgp_peer_shutdown() -- used when peer is downed for PEER_DOWN_USER_SHUTDOWN ! - added bgp_peer_reset_idle() -- used when peer goes pIdle or is about to enable session. - deleted bgp_peer_timers_stop() -- see bgp_peer_change_status(). - replaced bgp_peer_clearing_completed() - if pClearing, sets pIdle and enables if can - if pDeleting, unlocks peer - replaced bgp_timer_set() by bgp_peer_timers_set() -- deals in new peer states only. - renamed peer_new() -> bgp_peer_new() - renamed peer_create() -> bgp_peer_create() - added setting of PEER_STATUS_REAL_PEER - changed auto activation to reflect what actually happens. - changed bgp_session_init_new() call because it now sets peer->session. - sets timers suitable for pIdle. before any auto enable. - renamed peer_delete() -> bgp_peer_delete() - removed call of peer_nsf_stop() - added bgp_peer_down() with PEER_DOWN_NEIGHBOR_DELETE -- which does all the work of flattening an active peer, and returns it pIdle or pClearing. - if pClearing, lock the peer so that when clearing completes, it can unlock it. - sets pDeleting state, and increments bm->peer_linger_count. - tightened procedure for dealing with various references to peer -- including use of the PEER_STATUS_REAL_PEER flag. - tidied up dealing with rsclient RIBs and shared pointers to group versions of same. - removed call of bgp_timer_set(), now done in bgp_peer_change_status(). - now unregisters the peer immediately, so can register a new one before this one is completely deleted. - deletes session if it can. - moved peer_lock() & peer_unlock() from bgpd.c and renamed bgp_peer_lock() & bgp_peer_unlock() - renamed peer_free() bgp_peer_free() and made static. - peer must be pDeleting -- so have been through bgp_peer_delete() - peer->session must be NULL - decrements bm->peer_linger_count - deleted peer->clear_node_queue handling - deleted bgp_session_free() -- that's done in bgp_peer_delete() or elsewhere. - unlocked bgp at end - assert peer->session == NULL, to be sure - set peer->lock == -54321 - in bgp_peer_enable(): - recast as switch() on peer state - added bgp_peer_reset)idle() before enabling the session. - renamed bgp_peer_disable() -> bgp_peer_down() - takes PEER_DOWN_XXX argument, which drives what notification is sent, and sets the peer->last_reset status. A small number of PEER_DOWN_XXX are special. - removed the IdleHoldTimer stuff. - copies outbound notification to session. - for PEER_DOWN_NSF_CLOSE_SESSION, keep non- stale routes. - for PEER_DOWN_USER_SHUTDOWN, do bgp_peer_shutdown() - after disabling any session and doing any shutdown, proceed as per peer->status: pIdle -- flush stale routes bgp_peer_enable() pEstablished -- bgp_peer_stop() pClearing -- flush stale routes - added bgp_peer_down_notify(). - added bgp_peer_down_error(), which replaces bgp_notify_send(). - added bgp_peer_down_error_with_data(), which replaces bgp_notify_send_with_data() The "down_error" functions calculate the appropriate PEER_DOWN_XXX value, and call bgp_peer_down_notify(). - added bgp_peer_map_peer_down(), to map PEER_DOWN_XXX to a notification message. - added bgp_peer_map_notification, to map notification message to a PEER_DOWN_XXX. - renamed peer_change_status() -> bgp_peer_change_status() - do most things only if state changes. - add call to bgp_peer_reset_idle() as enter pIdle state. - at all times do bgp_peer_timer_set() - renamed peer_timers_set() -> bgp_peer_timers_set() - commoned up code for Graceful Restart and Graceful Restart Stale timers and stale routes. - changed Graceful Restart Stale time to cope if it should expire before Graceful Restart ! bgpd/bgp_peer.h -- added PEER_DOWN_XXX values and tidied up + PEER_DOWN_NULL + PEER_DOWN_UNSPECIFIED + PEER_DOWN_CONFIG_CHANGE + PEER_DOWN_AF_DEACTIVATE + PEER_DOWN_PASSWORD_CHANGE + PEER_DOWN_ALLOWAS_IN_CHANGE + PEER_DOWN_INTERFACE_DOWN + PEER_DOWN_MAX_PREFIX + PEER_DOWN_HEADER_ERROR + PEER_DOWN_OPEN_ERROR + PEER_DOWN_UPDATE_ERROR + PEER_DOWN_HOLD_TIMER + PEER_DOWN_FSM_ERROR + PEER_DOWN_DYN_CAP_ERROR - PEER_DOWN_NOTIFY_SEND (deleted) - added typedef peer_down_t - struct peer: - deleted redundant clear_node_queue - removed PEER_STATUS_ACCEPT_PEER flag - added PEER_STATUS_REAL_PEER flag - (3) - deleted bgp_peer_reenable() -- redundant - deleted bgp_peer_stop() -- now static - replaced bgp_peer_disable() by bgp_peer_down() - added bgp_peer_down_error() - added bgp_peer_down_error_with_data() - deleted peer_change_status() -- now static - renamed peer_new() -> bgp_peer_new() - renamed peer_create() -> bgp_peer_create() - renamed peer_delete() -> bgp_peer_delete() - added bgp_peer_lock() - added bgp_peer_unlock() - deleted peer_free() - deleted peer_nsf_stop() bgpd/bgp_peer_index.c -- removed accept entry from bgp_peer_index_entry structure - added explicit next_free entry to the structure - sets next_free to point at self in entries which are in use -- and checks this. - change bgp_peer_index_seek_accept() to link to session via the peer data structure, and to call bgp_connection_query_accept() under the Peer Index Mutex. - (4) for bgp_peer_index_seek_accept() bgpd/bgp_peer_index.h -- removed accept entry from bgp_peer_index_entry structure - added explicit next_free entry to the structure - (4) for bgp_peer_index_seek_accept() bgpd/bgp_route.c -- (1) (2) (3) - in bgp_process_rsclient(), bgp_process_main(), and bgp_processq_del(): - extra dasserts() - clear rn->wq_next - unlock table *after* unlock node (bug fix) - in bgp_process(), lock bgp before table. - in bgp_maximum_prefix_restart_timer(), replace call of peer_clear() by unset of flag and bgp_peer_enable() -- peer is already down. - added bgp_maximum_prefix_cancel_timer() - deleted bgp_clear_this_route() -- code now inline in only caller. - renamed bgp_clear_route_normal() -> bgp_clear_routes() - takes an "nsf" argument to invoke NSF "clearing", iff nsf set for afi/safi. Sets PEER_STATUS_NSF_WAIT if so. - returns bool "completed" if clearing has completed immediately -- ie no background work left to be done. - renamed bgp_clear_route_all() -> bgp_clear_all_routes() - takes "nsf" argument and returns "completed" as for bgp_clear_routes(). - removed call: bgp_peer_clearing_completed() - renamed bgp_clear_route_rsclient() -> bgp_clear_rsclient_rib() - deleted bgp_cleanup_routes() -- was used during termination, no longer required because termination deletes all peers. bgpd/bgp_route.h -- deleted bgp_cleanup_routes() -- program termination now deletes all peers, which implicitly cleans up all routes. - renamed: bgp_clear_route_normal() -> bgp_clear_routes() - renamed: bgp_clear_route_rsclient() -> bgp_clear_rsclient_rib() - renamed: bgp_clear_route_all() -> bgp_clear_all_routes() - added: bgp_maximum_prefix_cancel_timer() bgpd/bgp_session.c -- (3) - deleted bgp_session_defer_if_limping() - in bgp_session_init_new() - changed to void and removed session argument -- always creates a new session - peer MUST not have a session - removed Peer Index pointer stuff as Peer Index no longer has accept field - sets session->peer and locks peer - sets peer->session under Peer Index Mutex - sets session->delete_me false - sets session->accept flag false - replaced bgp_session_free() by bgp_session_delete() - changed to void function - if session is active, set the delete_me flag so session will be deleted when goes sDisabled. - make sure that session Mutex has been released by the BGP Engine before destroying it... otherwise: tears. - unhook session from peer under Peer Index mutex -- for accept() stuff. - unhook peer from session. - unlock peer. - in bgp_session_enable() - assert that peer is pIdle. - clear delete_me for completeness - clear additional fields - in bgp_session_disable() - clear session->accept - in bgp_session_is_active() - no longer interested in Peer Index stuff - deleted bgp_session_defer_if_limping() bgpd/bgp_session.h -- in bgp_session structure: - removed index_entry pointer to Peer Index - added delete_me flag - removed defer_enable flag - added accept flag - removed session parameter from bgp_session_init_new() - deleted bgp_session_free() - added bgp_session_delete() - bgp_session_is_active() now returns bool bgpd/bgp_table.c -- bgp_node_free() sets lock count = -54321 - in bgp_table_free(): - assert that route node is empty - set lock count = -54321 - bgp_node_delete() asserts that is not on_wq - (1) bgpd/bgp_vty.c -- (1) (3) (4) - change peer_af_flag_modify_vty() to call peer_af_flag_modify() not set or unset. - change name: bgp_clear_route_rsclient() -> bgp_clear_rsclient_rib() - in peer_rsclient_set_vty(): - add peer to bgp->rsclient list after all validation is complete - in peer_rsclient_unset_vty(): - removed code for deleting the rsclient RIB etc to peer_rsclient_unset(). - added peer_rsclient_unset() bgpd/bgp_zebra.c -- bgp_peer_disable() -> bgp_peer_down() and now takes PEER_DOWN_INTERFACE_DOWN argument. bgpd/bgpd.c -- (1) (3) - replaced setting peer->last_reset and call of bgp_notify_send() by call of the new bgp_peer_down(). - bgp_router_id_set() ... PEER_DOWN_RID_CHANGE - bgp_cluster_id_set() ... PEER_DOWN_CLID_CHANGE - bgp_cluster_id_unset() ... PEER_DOWN_CLID_CHANGE - bgp_confederation_id_set() ... PEER_DOWN_CONFED_ID_CHANGE ... PEER_DOWN_CONFED_ID_CHANGE - bgp_confederation_id_unset() ... PEER_DOWN_CONFED_ID_CHANGE - bgp_confederation_peers_add() ... PEER_DOWN_CONFED_PEER_CHANGE - bgp_confederation_peers_remove() ... PEER_DOWN_CONFED_PEER_CHANGE - peer_as_change() ... PEER_DOWN_REMOTE_AS_CHANGE - peer_activate() ... PEER_DOWN_AF_ACTIVATE - peer_deactivate() ... PEER_DOWN_AF_DEACTIVATE - peer_group_bind() ... PEER_DOWN_RMAP_BIND - peer_group_unbind() ... PEER_DOWN_RMAP_UNBIND - peer_change_action() ... why_changed ... why_changed - peer_flag_modify_action() ... action->peer_down - peer_update_source_if_set() ... PEER_DOWN_UPDATE_SOURCE_CHANGE ... PEER_DOWN_UPDATE_SOURCE_CHANGE - peer_update_source_addr_set() ... PEER_DOWN_UPDATE_SOURCE_CHANGE ... PEER_DOWN_UPDATE_SOURCE_CHANGE - peer_update_source_unset() ... PEER_DOWN_UPDATE_SOURCE_CHANGE ... PEER_DOWN_UPDATE_SOURCE_CHANGE - peer_local_as_set() ... PEER_DOWN_LOCAL_AS_CHANGE ... PEER_DOWN_LOCAL_AS_CHANGE - peer_local_as_unset() ... PEER_DOWN_LOCAL_AS_CHANGE ... PEER_DOWN_LOCAL_AS_CHANGE - peer_password_set() ... PEER_DOWN_PASSWORD_CHANGE ... PEER_DOWN_PASSWORD_CHANGE - peer_password_unset() ... PEER_DOWN_PASSWORD_CHANGE ... PEER_DOWN_PASSWORD_CHANGE - peer_clear() ... PEER_DOWN_USER_RESET - bgp_terminate() ... PEER_DOWN_USER_RESET - deleted peer_lock() & peer_unlock(). See bgp_peer_lock() & bgp_peer_unlock() in bgp_peer - in peer_as_change(), move downing of peer to after all config changes have been made. - in peer_remote_as() implicitly activate iff !BGP_FLAG_NO_DEFAULT and is IPv4/Unicast. (but only ever called with IPv4/Unicast or nothing at all.) - in peer_deactivate() - if cannot dynamically reconfigure, then will down the peer PEER_DOWN_AF_DEACTIVATE. - uses new peer_rsclient_unset() to tidy away any rsclient RIB etc. - in peer_change_action(): - added 'why_down' argument - replace if's by switch() - in struct peer_flag_action, updated entry types - in peer_flag_action_list[], added the appropriate PEER_DOWN_XXX values. - in peer_af_flag_action_list[] - added the appropriate PEER_DOWN_XXX values - added multiple flag entries - in peer_flag_action_set(): - changed to return const address of peer_flag_action structure -- or NULL. - table may now contain entries which the given flag must be a subset of. - in peer_flag_modify_action(): - now takes peer_flag_action* and whether flag has been set or not. - allow *only* peer_change_none or peer_change_reset - deal with clearing PEER_FLAG_SHUTDOWN, otherwise bgp_peer_down(). - in peer_group_bind(): - uses new peer_rsclient_unset() to tidy away any rsclient RIB etc. - in peer_flag_modify(): - takes bool set flag - changed to suit peer_flag_action_set() and peer_flag_modify_action() - in peer_flag_set() and peer_flag_unset() changed to bool flag - added peer_af_flag_modify_action(), common code for use in peer_af_flag_modify(). - in peer_af_flag_modify(): - takes bool set flag - changed to suit peer_flag_action_set() and peer_flag_modify_action() - use peer_af_flag_modify_action() - in peer_af_flag_set() and peer_af_flag_unset() changed to bool flag - in peer_clear(): adjust for new bgp_peer_down() mechanics. - in bgp_master_init(): account for peer_linger_count (starting at 0) - in bgp_terminate(): - removed program_terminating -- see flag now in bgp_main.c - implement "retain_mode" by using BGP_OPT_NO_FIB flag to turn off changing the FIB as routes are deleted. - either bgp_peer_delete() if terminating or bgp_peer_down() all peers. - flush process queues. - deleted program_terminate_if_all_disabled() - in peer_lookup(), removed handling of PEER_STATUS_ACCEPT_PEER(). - deleted peer_lookup_with_open(). - in bgp_config_write_family(), removed handling of PEER_STATUS_ACCEPT_PEER(). - in bgp_config_write(), removed handling of PEER_STATUS_ACCEPT_PEER(). bgpd/bgpd.h -- add peer_linger_count entry to the bgp_master structure. - remove: peer_lock(), peer_unlock() and peer_delete() - (4) for peer_af_flag_modify() - added peer_rsclient_unset() lib/plist.c -- fixed handling of prefix lists with no explicit sequence numbers. lib/qpnexus.c -- (4) for main_thread & terminate flags - change qpn_terminate() so does nothing if terminate flag is already set. lib/qpnexus.h -- (4) for main_thread & terminate flags tests/bgp_capability_test.c -- (3)
* New functions for error numbers and addresses in messages.Chris Hall2010-04-081-9/+7
| | | | | | | | Implemented less onerous ways of including descriptions of errors and IP addresses in logging and other messages. Implemented mapping of error numbers to error names, which is generally more meaningful.
* First beta releaseChris Hall2010-04-061-30/+104
| | | | | | | Various bug fixes and improvements. Running with a fair amount of debug/assert code, which must be removed at some date.
* Major updateChris Hall2010-03-161-3/+3
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | bgpd/bgp_advertise.c bgpd/bgp_advertise.h The adj_in and adj_out objects are now put on a list based on the peer to whom the route belongs. The adj_in and adj_out objects also now point to the bgp_node which they are routes for. This substantially reduces the work needed to shut down a peer. bgpd/bgp_damp.c Changes to adj_in and adj_out forced small change to macros used in bgp_damp.c to manage its lists. bgpd/bgp_debug.c Replaced direct access to vty->node by the required vty_get_node(). bgpd/bgp_dump.c Changes to the names of fields in bgp_info structures. bgpd/bgp_engine.h Modified the debug and trace functions. bgpd/bgp_fsm.c Make use of sockunion2str() consistent with common usage. Improved some documentation. bgpd/bgp_main.c Use the newly extended qpn_add_hook_function() facility. bgpd/bgp_mplsvpn.c Changes to the names of fields in bgp_info structures. bgpd/bgp_msg_read.c Bug fix: correct handling of capability code length. Improvement: better casting in calculation of message length. bgpd/bgp_msg_write.c Bug fix: correct byte ordering of bgp_id in open message. bgpd/bgp_network.c Bug fix: correct handling of incoming connections. Takes advantage of improvements in sockunion.c. bgpd/bgp_nexthop.c Changes to the names of fields in bgp_info structures. bgpd/bgp_open_state.c Remove mistaken #include of memtypes.h bgpd/bgp_packet.c Improvements to handling of withdrawing routes for peers. bgpd/bgp_peer.c Tidying up the state of peers as they are enabled and disabled. Improvements to handling of withdrawing routes for peers. bgpd/bgp_peer.h Adding list bases for lists of routes originated by the peer. bgpd/bgp_peer_index.c Bug fix: correct freeing of peer indexes. bgpd/bgp_route.c Implement lists of bgp_info based in the owning peer. Adjust for name changes to bgp_info fields. Reimplemented all the clearing functions to use the lists of items that belong to the peer -- rather than searching route tables for stuff to withdraw. Changed work queue handling for added/changed routes, so that queues run through existing items, rather than having queues of auxiliary items -- lower memory overhead. bgpd/bgp_route.h Added fields to bgp_info to allow all bgp_info originated by each peer to live on lists based in the peer. And changed the name of existing fields to avoid confusion. bgpd/bgp_routemap.c Removing redundant code and fixing a memory leak. bgpd/bgp_table.h Based work queue for added/changed routes directly in the table, rather than having auxiliary structures. bgpd/bgp_vty.c Use vty_get_node() and vty_set_node() rather than direct access to the vty field. bgpd/bgpd.c Implement changes to route clearing. bgpd/bgpd.h Changes to work queue handling. lib/buffer.c Changes to allow embedded buffer structures. lib/buffer.h Moved struct buffer here so that could have embedded buffer structurs. lib/command.c Substantial tidy up and document exercise. Restructured the top level command processing and finding of descriptions and command completion. Removal of unpleasant messing around with the insides of vector structures. Movement of some command actions to vty.c. Uses uty.h to pick up the "private" functions from vty.c et al. lib/command.h Moved the "node" values to node_type.h, so that can use an enum node_type in places where cannot include command.h. lib/command_queue.c Updated to cope with the called command changing the node value. Improved handling of revoked commands, so the the command line handler does not get stuck waiting for a command to complete which has been revoked ! lib/command_queue.h Improved message format. lib/if.c Use vty_set_node(). lib/keychain.c Use vty_set_node(). new lib/keystroke.c new lib/keystroke.h New code to implement a keystroke FIFO. This moves some complexity out of the command handler. The handling of mixtures of escapes and Telnet IACs is tightened up. It would be possible to extend this to, say, UTF-8. Regularises the "stealing" of keystrokes for the "--more--" output handling... which was a bit hit and miss. new lib/list_util.c new lib/list_util.h New code to implement various forms of linked list, where the list pointers are embedded in structures. lib/log.c Changed the handling of log messages, so that all types of log output (except syslog) use the same message buffer scheme, and the message is constructed once and once only. Changes to the handling of VTY_LOCK() etc. Uses uty.h to pick up the "private" functions from vty.c et al. lib/log.h Changes to the buffering of log messages. new lib/mem_tracker.c New code to track memory allocation/deallocation, for debug purposes. lib/memory.c lib/memory.h Updated to allow the use of the mem_tracker. lib/memtypes.awk Made the memtypes into a named enum MTYPE. lib/memtypes.c Various new memory types. lib/mqueue.c lib/mqueue.h Add mqueue_finish function for close-down. lib/network.c lib/network.h Added non-blocking read_nb() and write_nb(). new lib/node_type.h As above. lib/plist.c Remove vty_puts() which wasn't a good idea. lib/qlib_init.c Added qps_init() to first stage and mqueue_finish to finish. lib/qpnexus.c lib/qpnexus.h More flexible hooks for in_thread_init and in_thread_final. lib/qpselect.c lib/qpselect.h Added qps_start_up() to build the required maps once and for all. Added qdebug to control the debug checks and validation. Improved validation and test functions. new lib/qstring.c new lib/qstring.h New code for limited flexible string handling. lib/qtimers.c Added qdebug to control the debug checks and validation. lib/routemap.c Use vty_set_node(). lib/sockunion.c lib/sockunion.h Tidied up and regularised the handling of sin_len and sin6_len. Created common function for setting port into socket. Created common function for initialisation/allocation of new sockunion. Reduced various functions by using common sub-functions. Rationalised some code. Added sockunion_listen() and sockunion_new_sockaddr(). Renamed sockunion_new() to sockunion_new_prefix(). Improved some logging messages. Added documentation. new lib/uty.h Functions etc. used only by vty/command/log/vty_io and vty_cli. lib/vector.c lib/vector.h Added vector_t type. Removed VECTOR_INDEX, vector_only_wrapper_free() and vector_only_index_free() -- following improvement of code in command.c. Added vector_set_min_length(), vector_set_new_min_length() and vector_length() functions. new lib/vio_fifo.c new lib/vio_fifo.h New code to manage simple FIFO of indefinite length. lib/vty.c lib/vty.h Reworked. Broken into vty.c, vty_io.c and vty_cli.c. new lib/vty_cli.c new lib/vty_cli.h CLI handling parts of the vty family. new lib/vty_io.c new lib/vty_io.h I/O parts of the vty family. lib/workqueue.h Introduced tyedefs for the various call-back entries. new tests/test-list_util.c Tests for the list-util stuff. vtysh/vtysh.c Small change to interface for cmd_execute_command()
* Revised thread/timer handling, work queue and scheduling.Chris Hall2010-02-161-62/+43
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | Updated quagga thread handling to use qtimers when using the new qpnexus -- so all timers are qtimers in the new scheme. Updated work queue handling so that each work queue item is a single malloced structure, not three. (Only bgpd and zebra use the work queue system.) When using qpnexus the background thread queue is no longer a timer queue, but simply a list of pending background threads. When a background thread is waiting on a timer, it is in the qtimer pile, same like any other thread. When using qpnexus, the only remaining quagga thread queues are the event and ready queues. Revised the qpnexus loop so that only when there is nothing else to do will it consider the background threads. Revised write I/O in the BGP Engine so that all writing is via the connection's write buffer. Revised the write I/O in the Routeing Engine, so that it passes groups of updates in a single mqueue message. This all reduces the number of TCP packets sent (because BGP messages are collected together in the connection's write buffer) and reduces the number of mqueue messages involved. (No need for TCP_CORK.) Code and comments review for the new code. modified: bgpd/bgp_advertise.c modified: bgpd/bgp_common.h modified: bgpd/bgp_connection.c modified: bgpd/bgp_connection.h modified: bgpd/bgp_engine.h modified: bgpd/bgp_fsm.c modified: bgpd/bgp_main.c modified: bgpd/bgp_msg_read.c modified: bgpd/bgp_msg_write.c modified: bgpd/bgp_network.c modified: bgpd/bgp_packet.c modified: bgpd/bgp_packet.h modified: bgpd/bgp_peer.c modified: bgpd/bgp_peer_index.h modified: bgpd/bgp_route.c modified: bgpd/bgp_route_refresh.h modified: bgpd/bgp_session.c modified: bgpd/bgp_session.h modified: bgpd/bgpd.c new file: bgpd/bgpd.cx modified: lib/mqueue.h modified: lib/qpnexus.c modified: lib/qpnexus.h modified: lib/qpselect.c modified: lib/qtimers.c modified: lib/qtimers.h modified: lib/sigevent.c modified: lib/stream.c modified: lib/stream.h modified: lib/thread.c modified: lib/thread.h modified: lib/workqueue.c modified: lib/workqueue.h modified: tests/heavy-wq.c modified: zebra/zebra_rib.c
* Fix retry for connect() connections + debug for mqueue.Chris Hall2010-02-051-7/+6
| | | | | | | | | modified: bgpd/bgp_connection.c modified: bgpd/bgp_connection.h modified: bgpd/bgp_engine.h modified: bgpd/bgp_fsm.c modified: bgpd/bgp_session.c modified: bgpd/bgpd.c
* Review and tidy new code and comments.Chris Hall2010-02-041-116/+162
| | | | | | | | | | | | | | | | | | | Follows David Lamparter's code for update-source handling. modified: bgpd/bgp_connection.c modified: bgpd/bgp_connection.h modified: bgpd/bgp_fsm.c modified: bgpd/bgp_fsm.h modified: bgpd/bgp_msg_read.c modified: bgpd/bgp_msg_read.h modified: bgpd/bgp_notification.c modified: bgpd/bgp_notification.h modified: bgpd/bgp_peer.c modified: bgpd/bgp_peer.h modified: bgpd/bgp_session.c modified: bgpd/bgp_session.h modified: lib/prefix.c modified: lib/prefix.h
* Tidying up the new FSM and exception handling.Chris Hall2010-02-041-623/+361
| | | | | | | | | | | modified: bgpd/bgp_common.h modified: bgpd/bgp_fsm.c modified: bgpd/bgp_fsm.h modified: bgpd/bgp_msg_read.c modified: bgpd/bgp_msg_write.c modified: bgpd/bgp_notification.c modified: bgpd/bgp_notification.h modified: bgpd/bgp_packet.c
* Clearing up the reaping of connections, and handling of notifications.Chris Hall2010-02-011-144/+174
| | | | | | | | | modified: bgpd/bgp_connection.c modified: bgpd/bgp_fsm.c modified: bgpd/bgp_fsm.h modified: bgpd/bgp_msg_read.c modified: bgpd/bgp_peer.c modified: bgpd/bgp_session.c
* Binding to interfaces and counting of messages.Chris Hall2010-01-271-20/+18
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | Wired up message counters in bgp_session structure. Added fields to session for neighbor interface and neighbor update-source -- so that these can be set when connect() is done. Peering Engine resolves any interface name to an address, so that BGP Engine doesn't have to. Reinstated as much code as necessary in bgp_network to bind to specific interfaces, as set in the session. Moved setting of bgp_nexthop_set() back into Routeing Engine. Result is that only Peering Engine talks to Zebra or uses the iflist. Wired up setting of TTL. Reworked connections locking of the session mutex so more robust if/when connections are cut loose from the session. Made peer_index entry point at connection, not session. Works better in bgp_network that way. modified: bgpd/bgp_connection.c modified: bgpd/bgp_connection.h modified: bgpd/bgp_fsm.c modified: bgpd/bgp_msg_read.c modified: bgpd/bgp_msg_write.c modified: bgpd/bgp_network.c modified: bgpd/bgp_network.h modified: bgpd/bgp_peer.c modified: bgpd/bgp_peer.h modified: bgpd/bgp_peer_index.c modified: bgpd/bgp_peer_index.h modified: bgpd/bgp_session.c modified: bgpd/bgp_session.h modified: lib/prefix.h modified: lib/sockunion.c modified: lib/sockunion.h
* Wiring up ROUTE-REFRESH and ORF handling. Work-in-progress.Chris Hall2010-01-271-4/+30
| | | | | | | | | | | | | | | | | | | | modified: bgpd/bgp.h modified: bgpd/bgp_common.c modified: bgpd/bgp_common.h modified: bgpd/bgp_connection.c modified: bgpd/bgp_connection.h modified: bgpd/bgp_fsm.c modified: bgpd/bgp_fsm.h modified: bgpd/bgp_msg_read.c modified: bgpd/bgp_msg_read.h modified: bgpd/bgp_msg_write.c modified: bgpd/bgp_notification.c modified: bgpd/bgp_notification.h modified: bgpd/bgp_open_state.c modified: bgpd/bgp_open_state.h modified: bgpd/bgp_route_refresh.c modified: bgpd/bgp_route_refresh.h modified: lib/qafi_safi.h
* Fix bgp_session_do_disable() to return exactly one eDisabled event.Chris Hall2010-01-251-2/+8
| | | | | | | modified: bgpd/bgp_fsm.c modified: bgpd/bgp_notification.c modified: bgpd/bgp_notification.h modified: bgpd/bgp_session.c
* Getting BGP Engine to start and removing warnings.Chris Hall2010-01-241-209/+192
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | Finish the wiring required to get bgp_msg_read to process OPEN messages into the connection's open_state, and be able to check for correct peer AS, etc. Removed bugs preventing messages from being written. Added BGP Id check to collision detection logic. Removed as many warnings from comilation as possible. Replaced horrible FIFO kludge in the process. (Introduced the even more horrible miyagi kludge.) modified: bgpd/bgp_advertise.c modified: bgpd/bgp_advertise.h modified: bgpd/bgp_connection.c modified: bgpd/bgp_connection.h modified: bgpd/bgp_debug.c modified: bgpd/bgp_fsm.c modified: bgpd/bgp_msg_read.c modified: bgpd/bgp_msg_write.c modified: bgpd/bgp_network.c modified: bgpd/bgp_nexthop.c modified: bgpd/bgp_notification.c modified: bgpd/bgp_open.c modified: bgpd/bgp_packet.c modified: bgpd/bgp_session.c modified: bgpd/bgp_session.h modified: bgpd/bgpd.c modified: lib/Makefile.am modified: lib/distribute.c modified: lib/if_rmap.c new file: lib/miyagi.h modified: lib/prefix.h modified: lib/sockopt.c modified: lib/stream.c modified: lib/thread.c modified: lib/vty.c modified: lib/zebra.h modified: tests/bgp_capability_test.c modified: tests/bgp_mp_attr_test.c modified: tests/ecommunity_test.c modified: tests/heavy-thread.c modified: tests/heavy-wq.c modified: tests/heavy.c modified: tests/main.c modified: tests/test-checksum.c modified: tests/test-sig.c modified: watchquagga/watchquagga.c modified: zebra/if_netlink.c modified: zebra/ioctl.c modified: zebra/rt_netlink.c modified: zebra/rtread_netlink.c
* Blitz on bgp_msg_read, particularly OPEN message handlingChris Hall2010-01-231-1/+2
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | In the BGP Engine the OPEN message needs to be processed into the open_recv structure in the *connection*. The OPEN that arrives must be checked for acceptability before it is acknowledged. Later the connection may be discarded in collision resolution, or the connection may become the Established connection, and the open_recv structure is passed to the session and hence to the Peering Engine. modified: bgpd/bgp.h modified: bgpd/bgp_common.c modified: bgpd/bgp_common.h modified: bgpd/bgp_connection.c modified: bgpd/bgp_connection.h modified: bgpd/bgp_fsm.c modified: bgpd/bgp_msg_read.c modified: bgpd/bgp_msg_read.h modified: bgpd/bgp_msg_write.c modified: bgpd/bgp_notification.c modified: bgpd/bgp_notification.h modified: bgpd/bgp_open.h modified: bgpd/bgp_open_state.c modified: bgpd/bgp_open_state.h modified: bgpd/bgp_packet.c modified: bgpd/bgp_session.h modified: lib/distribute.c modified: lib/if_rmap.c modified: lib/qafi_safi.h modified: lib/stream.c modified: lib/stream.h
* Weed out last vestiges of FSM setting session->state.Chris Hall2010-01-211-4/+1
|
* Keep timer running past IO errors in Connect/Active states.Chris Hall2010-01-211-17/+13
|
* Adding verify debug code for qtimers and tidied up.Chris Hall2010-01-201-8/+17
| | | | | | | | | | | | | | | | modified: bgpd/bgp_connection.c -- using _unset functions -- added bgp_connection_close_file modified: bgpd/bgp_connection.h modified: bgpd/bgp_fsm.c -- checked notification handling -- uses bgp_connection_close_file modified: bgpd/bgp_notification.c -- added _unset function modified: bgpd/bgp_notification.h modified: bgpd/bgp_open_state.c -- added _unset function modified: bgpd/bgp_open_state.h modified: lib/qtimers.c -- added debug _verify function modified: lib/qtimers.h modified: lib/sockunion.c -- added _unset function modified: lib/sockunion.h
* Fix bug in bgp_connection not clearing pointers after free. Plantedpaulo2010-01-201-0/+5
| | | | asserts to try to track timer crash.
* Use sigprocmask when non-threaded. Fixed assert.paulo2010-01-191-1/+1
|
* Refining the session state handling.Chris Hall2010-01-181-3/+6
| | | | | | | | | | | | | | | | | | | | | | | modified: bgpd/bgp_common.h Renaming of states: sStopping -> sLimping sStopped -> sDisabled modified: bgpd/bgp_fsm.c Removed last vestige of session->state stuff. modified: bgpd/bgp_peer.c Upgraded event processing so that disable session has full hand shake with BGP Engine. modified: bgpd/bgp_session.c Changed to implement sLimping and sDisabled states and the disable session hand-shake. modified: bgpd/bgp_session.h
* Wired up receiving updates, with update packet sent from BGP to Routingpaulo2010-01-151-0/+3
| | | | | | | | | | | engine. Wired up Routing engine side of sending update packets to BGP engine with flow control. Eliminated a lot of compilation errors and warnings although a few remain. Have skipped out now unused code, particularly in bgp_packet.c that will be deleted eventually.
* Make session state a Peering Engine item.Chris Hall2010-01-141-19/+21
| | | | | | | | | | | | | | | | | | | | | | | | | | Along with the introduction of the sStopping session state, move responsibility for the session->state field entirely back to Peering Engine, and get it to change same when session events arrive. Stop FSM from writing the session->state, session->event, session->notification, session->err and session->ordinal fields. (Which may now be used by Peering Engine if it wants.) Handle sStopping as an active state where required. Not issue a further session disable once in sStopping state. Wire up bgp_connection and bgp_msg_read. modified: bgpd/bgp_connection.c modified: bgpd/bgp_fsm.c modified: bgpd/bgp_msg_read.c modified: bgpd/bgp_msg_read.h modified: bgpd/bgp_peer.c modified: bgpd/bgp_peer_index.c modified: bgpd/bgp_session.c modified: bgpd/bgp_session.h
* Further work-in-progressChris Hall2010-01-101-10/+69
| | | | | | | | | | | | | | | | | | | | | | | | | | | modified: .gitignore modified: bgpd/Makefile.am modified: bgpd/bgp.h modified: bgpd/bgp_common.h modified: bgpd/bgp_connection.c modified: bgpd/bgp_connection.h modified: bgpd/bgp_fsm.c modified: bgpd/bgp_msg_write.c modified: bgpd/bgp_msg_write.h modified: bgpd/bgp_notification.c modified: bgpd/bgp_notification.h modified: bgpd/bgp_open_state.c modified: bgpd/bgp_open_state.h modified: bgpd/bgp_packet.h new file: bgpd/bgp_route_refresh.c new file: bgpd/bgp_route_refresh.h modified: bgpd/bgp_session.c modified: bgpd/bgp_session.h modified: lib/Makefile.am new file: lib/confirm.h modified: lib/memtypes.c modified: lib/stream.c modified: lib/stream.h modified: lib/zassert.h
* Further work-in-progress.Chris Hall2010-01-081-176/+258
| | | | | | | | | | | | | | | | | modified: bgpd/Makefile.am modified: bgpd/bgp_common.h modified: bgpd/bgp_connection.c modified: bgpd/bgp_connection.h modified: bgpd/bgp_engine.c modified: bgpd/bgp_fsm.c modified: bgpd/bgp_fsm.h new file: bgpd/bgp_msg_write.c new file: bgpd/bgp_msg_write.h modified: bgpd/bgp_notification.c modified: bgpd/bgp_notification.h modified: bgpd/bgp_session.c modified: bgpd/bgp_session.h modified: lib/sockunion.h
* Continuing work-in-progressChris Hall2010-01-081-149/+204
| | | | | | | | | | | | | | | | | | modified: bgpd/Makefile.am modified: bgpd/bgp.h modified: bgpd/bgp_common.h modified: bgpd/bgp_connection.c modified: bgpd/bgp_connection.h modified: bgpd/bgp_engine.h modified: bgpd/bgp_fsm.c modified: bgpd/bgp_network.c modified: bgpd/bgp_network.h new file: bgpd/bgp_peer_index.c new file: bgpd/bgp_peer_index.h modified: bgpd/bgp_session.c modified: bgpd/bgp_session.h modified: lib/memtypes.c modified: lib/mqueue.h
* Work in progress on BGP EngineChris Hall2010-01-061-380/+789
| | | | | | | | | | | | modified: bgpd/bgp_common.h modified: bgpd/bgp_connection.c modified: bgpd/bgp_connection.h modified: bgpd/bgp_fsm.c modified: bgpd/bgp_fsm.h modified: bgpd/bgp_network.c modified: bgpd/bgp_open_state.h modified: bgpd/bgp_session.c modified: bgpd/bgp_session.h
* Initial commit for bgp_engine branch -- seeding new files...Chris Hall2010-01-041-993/+1877
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | On branch bgp_engine modified: .gitignore modified: bgpd/Makefile.am new file: bgpd/bgp.h new file: bgpd/bgp_common.c new file: bgpd/bgp_common.h new file: bgpd/bgp_connection.c new file: bgpd/bgp_connection.h modified: bgpd/bgp_debug.c new file: bgpd/bgp_engine.c new file: bgpd/bgp_engine.h modified: bgpd/bgp_fsm.c modified: bgpd/bgp_fsm.h modified: bgpd/bgp_network.c modified: bgpd/bgp_network.h new file: bgpd/bgp_notification.c new file: bgpd/bgp_notification.h modified: bgpd/bgp_open.c new file: bgpd/bgp_open_state.c new file: bgpd/bgp_open_state.h modified: bgpd/bgp_packet.c new file: bgpd/bgp_peer.c new file: bgpd/bgp_peer.h modified: bgpd/bgp_route.c new file: bgpd/bgp_session.c new file: bgpd/bgp_session.h modified: bgpd/bgp_vty.c modified: bgpd/bgp_zebra.c modified: bgpd/bgpd.c modified: bgpd/bgpd.h modified: lib/Makefile.am modified: lib/memtypes.c modified: lib/mqueue.c modified: lib/mqueue.h new file: lib/qafi_safi.h modified: lib/qpselect.c modified: lib/qpselect.h modified: lib/qpthreads.c modified: lib/qpthreads.h modified: lib/sockopt.c modified: lib/sockunion.c modified: lib/sockunion.h modified: lib/stream.c modified: lib/stream.h modified: lib/symtab.h modified: lib/zebra.h
* bgpd: compile warnings cleanupStephen Hemminger2009-12-101-1/+1
| | | | | | | | | | | | * bgpd/bgp_fsm.c * bgp_clearing_completed(): only used in one file, can be static * bgpd/bgp_packet.c * afi2str(): sayonara * safi2str(): sayonara * bgpd/bgp_route.c * bgp_distance_reset(): sayonara * bgpd/bgp_zebra.c * bgp_ifindex_by_nexthop(): sayonara
* bgpd/trivial: Fix indentation in previousPaul Jakma2009-07-281-4/+4
|
* bgpd: fd leak in bgpdSteve Hill2009-07-281-15/+33
| | | | | | | | | | | | | | | | | * bgp_fsm.c: I have found an fd leak in bgpd that is caused by the 'new' Clearing state. I've been seeing it from hold timer failures, but it can also be triggered by other things. When Hold_Timer_expired fires in Established state, a notify is sent and BGP_Stop event queued. The fsm then transitions into Clearing state. That is the problem; When the BGP_Stop event is serviced, the state table says to ignore it while in Clearing. Thus bgp_stop is not called and the fd leaks. Previously the peer would be in Idle state, which correctly handles the BGP_Stop event. Fix by making bgp_stop safe to call from Clearing state, without losing ClearingCompleted events, and then ensuring it is called prior to transition from Clearing->Idle.
* [bgpd/cleanup] Make BGP FSM table read-only staticStephen Hemminger2009-06-181-1/+1
| | | | The finite state machine table is immutable.
* [cleanup] functions taking no args should be declared with void argsStephen Hemminger2009-06-121-1/+1
| | | | | Use Ansi-C prototypes rather than old K&R method of declaring function without arguments
* Revert "[bgpd] Add 'bgp open-accept' option, to send OPEN immediately on ↵Paul Jakma2008-08-261-2/+1
| | | | | | | | accepted conns" Revert commit d664ae1182c29b74b409bc8594b7bd0575e91ce9. An experimental patch which violates RFC4271 quite badly, but managed to accidently sneak its way in.
* [bgpd] Add 'bgp open-accept' option, to send OPEN immediately on accepted connsPaul Jakma2008-08-221-1/+2
| | | | | | | | | | | 2007-08-31 Paul Jakma <paul.jakma@sun.com> * (general) Add 'bgp open-accept' option, to allow bgpd to send OPEN on accepted connections, i.e. to not wait till after collision-detect to send OPEN, which appears to be allowed in RFC4271. This may help speed up establishing sessions, or help avoid FSM problems with sessions to certain peers. Not enabled by default though.
* [bgpd] bug #368: Fix possible loop between peers going Idle<->OpenSentPaul Jakma2007-06-221-3/+3
| | | | | | | | | 2007-06-22 Paul Jakma <paul.jakma@sun.com> * bgp_fsm.c: (struct FSM) Bug #368. TCP Errors during OpenSent should cycle to Active, not to Idle or else peer bringup can race and cycle Idle<->Active. Reported and fix tested by Mukesh Agrawal.
* [bgpd] Peer delete can race with reconfig leading to crashPaul Jakma2007-02-221-8/+8
| | | | | | | | | | | | | | | | | | | 2007-02-22 Paul Jakma <paul.jakma@sun.com> * bgp_fsm.c: (bgp_fsm_change_status) Handle state change into clearing or greater here. Simpler. (bgp_event) Clearing state change work moved to previous * bgp_route.c: (bgp_clear_route_node) Clearing adj-in here is too late, as it leaves a race between a peer being deleted and an identical peer being configured before clearing completes, leading to a crash. Simplest fix is to clean peers Adj-in up-front, rather than queueing such work. (bgp_clear_route_table) Clear peer's Adj-In and Adj-Out up-front here, rather than queueing such work. Extensive comment added on the various bits of indexed data that exist and how they need to be dealt with. (bgp_clear_route) Update comment.
* [bgpd] Bug #302, bgpd can get stuck in state ClearingPaul Jakma2006-12-081-5/+8
| | | | | | | | | | | | 2006-12-07 Paul Jakma <paul.jakma@sun.com> * bgp_fsm.c: Bug #302 fix, diagnosis, suggestions and testing by Juergen Kammer <j.kammer@eurodata.de>. Fix follows from his suggested fix, just made in a slightly different way. (bgp_event) Transitions into Clearing always must call bgp_clear_route_all(). (bgp_stop) No need to clear routes here, BGP FSM should do it.
* [bgpd] Bug #302 fixes. ClearingCompleted event gets flushed, leaving peers ↵Paul Jakma2006-10-151-3/+3
| | | | | | | | | | | | | | stuck in Clearing. 2006-10-14 Paul Jakma <paul.jakma@sun.com> * bgp_fsm.h: Remove BGP_EVENT_FLUSH_ADD, dangerous and not needed. * bgp_fsm.c: (bgp_stop) Move BGP_EVENT_FLUSH to the top of the of the function, otherwise it could flush a ClearingCompleted event, bug #302. * bgp_packet.c: Replace all BGP_EVENT_FLUSH_ADD with BGP_EVENT_ADD, fixing bug #302.
* [bgpd] simplify peer refcounts, squash slow peer leakPaul Jakma2006-09-141-6/+2
| | | | | | | | | | | | | | | 2006-09-14 Paul Jakma <paul.jakma@sun.com> * (general) fix the peer refcount issue exposed by previous, by just removing refcounting of peer threads, which is mostly senseless as they're references leading from struct peer, which peer_free cancels anyway. No need to muck around.. * bgp_fsm.h: Just remove the refcounting from the various TIMER/READ/WRITE/EVENT ON/OFF/ADD macros. * bgp_fsm.c: (bgp_stop) use BGP_EVENT_FLUSH, no refcounts attached to events anymore. (bgp_event) remove peer_unlock, events not refcounted. * bgpd.c: (peer_free) flush events before free.
* [bgpd] Fix 0.99 shutdown regression, introduce Clearing and Deleted statesPaul Jakma2006-09-141-21/+83
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | 2006-09-14 Paul Jakma <paul.jakma@sun.com> * (general) Fix some niggly issues around 'shutdown' and clearing by adding a Clearing FSM wait-state and a hidden 'Deleted' FSM state, to allow deleted peers to 'cool off' and hit 0 references. This introduces a slow memory leak of struct peer, however that's more a testament to the fragility of the reference counting than a bug in this patch, cleanup of reference counting to fix this is to follow. * bgpd.h: Add Clearing, Deleted states and Clearing_Completed and event. * bgp_debug.c: (bgp_status_msg[]) Add strings for Clearing and Deleted. * bgp_fsm.h: Don't allow timer/event threads to set anything for Deleted peers. * bgp_fsm.c: (bgp_timer_set) Add Clearing and Deleted. Deleted needs to stop everything. (bgp_stop) Remove explicit fsm_change_status call, the general framework handles the transition. (bgp_start) Log a warning if a start is attempted on a peer that should stay down, trying to start a peer. (struct .. FSM) Add Clearing_Completed events, has little influence except when in state Clearing to signal wait-state can end. Add Clearing and Deleted states, former is a wait-state, latter is a placeholder state to allow peers to disappear quietly once refcounts settle. (bgp_event) Try reduce verbosity of FSM state-change debug, changes to same state are not interesting (Established->Established) Allow NULL action functions in FSM. * bgp_packet.c: (bgp_write) Use FSM events, rather than trying to twiddle directly with FSM state behind the back of FSM. (bgp_write_notify) ditto. (bgp_read) Remove the vague ACCEPT_PEER peer_unlock, or else this patch crashes, now it leaks instead. * bgp_route.c: (bgp_clear_node_complete) Clearing_Completed event, to end clearing. (bgp_clear_route) See extensive comments. * bgpd.c: (peer_free) should only be called while in Deleted, peer refcounting controls when peer_free is called. bgp_sync_delete should be here, not in peer_delete. (peer_delete) Initiate delete. Transition to Deleted state manually. When removing peer from indices that provide visibility of it, take great care to be idempotent wrt the reference counting of struct peer through those indices. Use bgp_timer_set, rather than replicating. Call to bgp_sync_delete isn't appropriate here, sync can be referenced while shutting down and finishing deletion. (peer_group_bind) Take care to be idempotent wrt list references indexing peers.
* [bgpd] Fix crash on shutdown of peerPaul Jakma2006-07-021-19/+22
| | | | | | | | 2006-07-02 Paul Jakma <paul.jakma@sun.com> * bgp_fsm.c: (bgp_{stop,start}) Move clear/free of certain bits of state from stop to start, as they may be used via peer references on clearing queues..
* [bgpd] Record afi/safi in bgp_table. Serialise peer clear with FSM.Paul Jakma2006-02-211-0/+1
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | 2006-02-21 Paul Jakma <paul.jakma@sun.com> * bgpd.h: move the clear_node_queue to be peer specific. Add a new peer status flag, PEER_STATUS_CLEARING. * bgp_table.h: (struct bgp_table) Add fields to record afi, safi of the table. (bgp_table_init) Take afi and safi to create table for. * bgp_table.c: (bgp_table_init) record the afi and safi. * bgp_nexthop.c: Update all calls to bgp_table_init. * bgp_vty.c: ditto. * bgpd.c: ditto. * bgp_fsm.c: (bgp_timer_set) dont bring up a session which is clearing. * bgp_route.c: (general) Update all bgp_table_init calls. (bgp_process_{rsclient,main}) clear_node is serialised via PEER_STATUS_CLEARING and fsm now. (struct bgp_clear_node_queue) can be removed. struct bgp_node can be the queue item data directly, as struct peer can be kept in the new wq global user data and afi/safi can be retrieved via bgp_node -> bgp_table. (bgp_clear_route_node) fix to get peer via wq->spec.data, afi/safi via bgp_node->bgp_table. (bgp_clear_node_queue_del) no more item data to delete, only unlock the bgp_node. (bgp_clear_node_complete) only need to unset CLEARING flag and unlock struct peer. (bgp_clear_node_queue_init) queue attaches to struct peer now. record peer name as queue name. (bgp_clear_route_table) If queue transitions to active, serialise clearing by setting PEER_STATUS_CLEARING rather than plugging process queue, and lock peer while queue active. Update to pass only bgp_node as per-queue-item specific data.
* 2005-06-28 Paul Jakma <paul.jakma@sun.com>paul2005-06-281-15/+15
| | | | | | | | | | * (global) The great bgpd extern and static'ification. * bgp_routemap.c: remove unused ROUTE_MATCH_ASPATH_OLD code (route_set_metric_compile) fix u_int32_t to ULONG_MAX comparison warnings. * bgp_route.h: (bgp_process, bgp_withdraw, bgp_update) export these used by various files which had their own private declarations, in the case of mplsvpn - incorrect.
* 2005-06-01 Paul Jakma <paul.jakma@sun.com>post_bgp_workqueuspaul2005-06-011-36/+41
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | * bgpd/(general) refcount struct peer and bgp_info, hence allowing us add work_queues for bgp_process. * bgpd/bgp_route.h: (struct bgp_info) Add 'lock' field for refcount. Add bgp_info_{lock,unlock} helper functions. Add bgp_info_{add,delete} helpers, to remove need for users managing locking/freeing of bgp_info and bgp_node's. * bgpd/bgp_table.h: (struct bgp_node) Add a flags field, and BGP_NODE_PROCESS_SCHEDULED to merge redundant processing of nodes. * bgpd/bgp_fsm.h: Make the ON/OFF/ADD/REMOVE macros lock and unlock peer reference as appropriate. * bgpd/bgp_damp.c: Remove its internal prototypes for bgp_info_delete/free. Just use bgp_info_delete. * bgpd/bgpd.h: (struct bgp_master) Add work_queue pointers. (struct peer) Add reference count 'lock' (peer_lock,peer_unlock) New helpers to take/release reference on struct peer. * bgpd/bgp_advertise.c: (general) Add peer and bgp_info refcounting and balance how references are taken and released. (bgp_advertise_free) release bgp_info reference, if appropriate (bgp_adj_out_free) unlock peer (bgp_advertise_clean) leave the adv references alone, or else call bgp_advertise_free cant unlock them. (bgp_adj_out_set) lock the peer on new adj's, leave the reference alone otherwise. lock the new bgp_info reference. (bgp_adj_in_set) lock the peer reference (bgp_adj_in_remove) and unlock it here (bgp_sync_delete) make hash_free on peer conditional, just in case. * bgpd/bgp_fsm.c: (general) document that the timers depend on bgp_event to release a peer reference. (bgp_fsm_change_status) moved up the file, unchanged. (bgp_stop) Decrement peer lock as many times as cancel_event canceled - shouldnt be needed but just in case. stream_fifo_clean of obuf made conditional, just in case. (bgp_event) always unlock the peer, regardless of return value of bgp_fsm_change_status. * bgpd/bgp_packet.c: (general) change several bgp_stop's to BGP_EVENT's. (bgp_read) Add a mysterious extra peer_unlock for ACCEPT_PEERs along with a comment on it. * bgpd/bgp_route.c: (general) Add refcounting of bgp_info, cleanup some of the resource management around bgp_info. Refcount peer. Add workqueues for bgp_process and clear_table. (bgp_info_new) make static (bgp_info_free) Ditto, and unlock the peer reference. (bgp_info_lock,bgp_info_unlock) new exported functions (bgp_info_add) Add a bgp_info to a bgp_node in correct fashion, taking care of reference counts. (bgp_info_delete) do the opposite of bgp_info_add. (bgp_process_rsclient) Converted into a work_queue work function. (bgp_process_main) ditto. (bgp_processq_del) process work queue item deconstructor (bgp_process_queue_init) process work queue init (bgp_process) call init function if required, set up queue item and add to queue, rather than calling process functions directly. (bgp_rib_remove) let bgp_info_delete manage bgp_info refcounts (bgp_rib_withdraw) ditto (bgp_update_rsclient) let bgp_info_add manage refcounts (bgp_update_main) ditto (bgp_clear_route_node) clear_node_queue work function, does per-node aspects of what bgp_clear_route_table did previously (bgp_clear_node_queue_del) clear_node_queue item delete function (bgp_clear_node_complete) clear_node_queue completion function, it unplugs the process queues, which have to be blocked while clear_node_queue is being processed to prevent a race. (bgp_clear_node_queue_init) init function for clear_node_queue work queues (bgp_clear_route_table) Sets up items onto a workqueue now, rather than clearing each node directly. Plugs both process queues to avoid potential race. (bgp_static_withdraw_rsclient) let bgp_info_{add,delete} manage bgp_info refcounts. (bgp_static_update_rsclient) ditto (bgp_static_update_main) ditto (bgp_static_update_vpnv4) ditto, remove unneeded cast. (bgp_static_withdraw) see bgp_static_withdraw_rsclient (bgp_static_withdraw_vpnv4) ditto (bgp_aggregate_{route,add,delete}) ditto (bgp_redistribute_{add,delete,withdraw}) ditto * bgpd/bgp_vty.c: (peer_rsclient_set_vty) lock rsclient list peer reference (peer_rsclient_unset_vty) ditto, but unlock same reference * bgpd/bgpd.c: (peer_free) handle frees of info to be kept for lifetime of struct peer. (peer_lock,peer_unlock) peer refcount helpers (peer_new) add initial refcounts (peer_create,peer_create_accept) lock peer as appropriate (peer_delete) unlock as appropriate, move out some free's to peer_free. (peer_group_bind,peer_group_unbind) peer refcounting as appropriate. (bgp_create) check CALLOC return value. (bgp_terminate) free workqueues too. * lib/memtypes.c: Add MTYPE_BGP_PROCESS_QUEUE and MTYPE_BGP_CLEAR_NODE_QUEUE
* 2005-05-19 Paul Jakma <paul@dishone.st>paul2005-05-191-2/+2
| | | | | | | | * bgp_fsm.c: (bgp_stop) use sockunion_free, not XFREE.. * bgp_network.c: (bgp_getsockname) ditto * bgp_routemap.c: (route_match_peer) ditto, als use a ret value and remove one sockunion_free. * bgpd.c: (peer_delete) ditto
* * bgp_fsm.c, bgp_open.c, bgp_packet.c, bgp_route.[ch], bgp_vty.c,hasso2005-02-021-2/+131
| | | | | | | | | bgpd.[ch]: Add BGP_INFO_STALE flag and end-of-rib support. "bgp graceful-restart" commands added. Show numbers of individual messages in "show ip bgp neighbor" command. Final pieces of graceful restart. [merge from GNU Zebra]
* * bgp_nexthop.c: Improve debug.hasso2005-02-011-7/+22
| | | | | | | | | | | | | * bgpd.[ch], bgp_nexthop.c, bgp_snmp.c: Remove useless bgp_get_master() function. * bgp_packet.c: MP AFI_IP update and withdraw parsing. * bgp_fsm.c: Reset peer synctime in bgp_stop(). bgp_fsm_change_status() is better place to log about peer status change than bgp_event(). Log in bgp_connect_success(). * bgp_vty.c: Fix typo in comment. * bgp_attr.c: Better log about unknown attribute. [merge from GNU Zebra]
* 2004-12-08 Andrew J. Schorr <ajschorr@alumni.princeton.edu>ajs2004-12-081-5/+5
| | | | * *.c: Change level of debug messages to LOG_DEBUG.
generated by cgit v1.2.3 (git 2.25.1) at 2025-08-10 13:48:33 +0000