diff options
| -rw-r--r-- | ospfd/ChangeLog.opaque.txt | 221 |
1 files changed, 221 insertions, 0 deletions
diff --git a/ospfd/ChangeLog.opaque.txt b/ospfd/ChangeLog.opaque.txt new file mode 100644 index 00000000..782e332a --- /dev/null +++ b/ospfd/ChangeLog.opaque.txt @@ -0,0 +1,221 @@ +----- * ----- * ----- * ----- * ----- * ----- * ----- * ----- * ----- * ----- +Changes 2002.12.20 + +1. Bug fixes + + 1.1 When an opaque LSA is being removed from (or added to) the LSDB, + it does not mean a change in network topology. Therefore, SPF + recalculation should not be triggered in that case. + There was an assertion failure problem "assert (rn && rn->info)" + inside the function "ospf_ase_incremental_update()", because + the upper function "ospf_lsa_maxage_walker_remover()" called it + when a type-11 opaque LSA is removed due to MaxAge. + + 1.2 Type-9 LSA is defined to have "link-local" flooding scope. + In the Database exchange procedure with a new neighbor, a type-9 + LSA was added in the database summary of a DD message, even if + the link is different from the one that have bound to. + +2. Feature enhancements + + 2.1 Though a "wildcard" concept to handle type-9/10/11 LSAs altogether + has introduced about a year ago, it was only a symbol definition + and actual handling mechanism was not implemented. Now it works. + +----- * ----- * ----- * ----- * ----- * ----- * ----- * ----- * ----- * ----- +Changes 2002.7.8 + +1. Bug fixes + + 1.1 When "ospf_delete_opaque_functab()" is called, internal structure + "oipt" remain unfreed. If register/delete functab is repeated, + illegal memory access happens due to this "oipt". + + 1.2 In "free_opaque_info_per_id()", there was a crucial typo which + ignores a condition test. + + "if (oipi->lsa != NULL);" <-- semicolon! + +2. Feature enhancements + + None. + +----- * ----- * ----- * ----- * ----- * ----- * ----- * ----- * ----- * ----- +Changes 2001.12.03 + +1. Bug fixes + + 1.1 Though a new member "oi" has added to "struct ospf_lsa" to control + flooding scope of type-9 Opaque-LSAs, the value was always NULL + because no one set it. + + 1.2 In the function "show_ip_ospf_database_summary()" and "show_lsa_ + detail_adv_router()", VTY output for type-11 Opaque-LSAs did not + work properly. + + 1.3 URL for the opaque-type assignment reference has changed. + + 1.4 In the file "ospf_mpls_te.c", printf formats have changed to + avoid compiler warning messages; "%lu" -> "%u", "%lx" -> "%x". + Note that this hack depends on OS, compiler and their versions. + + 1.5 One of attached documentation "opaque_lsa.txt" has changed to + reflect the latest coding. + +2. Feature enhancements + + 2.1 Knowing that it is an ugly hack, an "officially unallocated" + opaque-type value 0 has newly introduced as a "wildcard", + which matches to all opaque-type. + This value must not be flooded to the network, of course. + + 2.2 The Opaque-core module makes use of newly introduced hooks to + dispatch every LSDB change (LSA installation and deletion) to + preregistered opaque users. + Therefore, by providing appropriate callback functions as new + parameters of "ospf_register_opaque_functab()", an opaque user + can refer to every LSA instance to be installed into, or to be + deleted from, the LSDB. + +----- * ----- * ----- * ----- * ----- * ----- * ----- * ----- * ----- * ----- +Changes 2001.10.31 + +1. Bug fixes + + 1.1 Since each LSA has their own lifetime, they will remain in a + routing domain (being stored in LSDB of each router), until their + age naturally reach to MaxAge or explicitly being flushed by the + originated router. Therefore, if a router restarted with a short + downtime, it is possible that previously flooded self-originated + LSAs might received if the NSM status is not less than Exchange. + + There were some problems in the way of handling self-originated + Opaque-LSAs if they are contained in a received LSUpd message, + but not installed to the local LSDB yet. + Regardless of some conditions to start originating Opaque-LSAs + (there should be at least one opaque-capable full-state neighbor), + the function "ospf_flood()" will be called to flood and install + this brand-new looking LSA. + As the result, when the NSM of an opaque-capable neighbor gets + full, internal state inconsistency happens; a user of Opaque-LSA + such as MPLS-TE can refer to self-originated LSAs in the local + LSDB, but cannot modify their contents... + + Above problems have fixed with a policy "flush it from the whole + routing domain and keep silent until the flushing completed". + By using this sweeping technique, we can be free from confusion + caused by self-originated LSAs received via network. + + 1.2 The function "ospf_opaque_type_name()" contained massive ifdefs + corresponding to each "opaque-type". + These unnecessary ifdefs are removed completely. + + 1.3 In the function "ospf_delete_opaque_functab()", there was an + improper loop control that causes illegal memory access. + Original coding was "next = nextnode (node)". + + 1.4 The function "ospf_mpls_te_ism_change()" could not handle the + case when the ISM changes from Waiting to DR/BDR/Other. + So, there was a case that even if one of an ISM become + operational and MPLS-TE module has started, the corresponding + Opaque-LSA cannot be originated. + + 1.5 The function "ospf_opaque_lsa_reoriginate_schedule()" did not + allow to be called multiple times, simply because handling + module for the given "lsa-type & opaque-type" already exists. + But this assumption seems to be wrong. + Change the policy to allow this function to be called multiple + times and let the caller to decide what should do when the + corresponding callback function "(* functab->lsa_originator)()" + is called. + +2. Feature enhancements + + 2.1 The global bitmap "opaque" has introduced instead of former flag + "OpaqueCapable", to store complex conditions to handle Opaque-LSAs. + + 2.2 The MPLS-TE module now referes to "draft-katz-yeung-ospf-traffic + -06.txt", no significant changes with 05 version, though. + +----- * ----- * ----- * ----- * ----- * ----- * ----- * ----- * ----- * ----- +Changes 2001.08.03 + +1. Bug fixes + + 1.1 Even if the ospfd started with opaque capability enabled, when + the ospfd receives an unknown opaque-type (unregistered by the + function "ospf_register_opaque_functab()" beforehand), the LSA + was discarded. As the result, only the opaque-LSAs that have + commonly registered by opaque-capable ospf routers can be + flooded in a routing domain. + + This behavior has fixed so that arbitrary opaque-type LSAs can + be flooded among opaque-capable ospf routers. + If the ospfd has opaque-LSA capability but disabled at runtime, + received opaque-LSAs can be accepted and registered to LSDB as + is, but not be flooded to the network; those opaque LSAs will + remain in LSDB until explicitly flushed by incoming LSUpd + messages with MaxAge, or their age naturally reaches to MaxAge. + + 1.2 The function "ospf_register_opaque_functab()" did not check + if the entry corresponding to the given "lsa-type, opaque-type" + combination already exists or not. + This problem has fixed not to allow multiple registration. + + 1.3 Since type-11 (AS external) LSAs will be flooded beyond areas, + there is little relationship between "struct lsa" and "struct + area". More specifically, the pointer address "lsa->area" can + be NULL if the lsa-type is 11, thus an illegal memory access + will happen. This problem has fixed. + + 1.4 When self-originated opaque-LSAs are received via network and + if the corresponding opaque-type functions are not available + (they have already deleted) at that time, those LSAs were + dropped due to "unknown opaque-type" error. + After the problem 1.1 has fixed, those "self-originated" LSAs + were registered to LSDB and then flooded to the network, even + if the processing functions did not exist... + + After all, this problem has fixed so that those LSAs should + explicitly be flushed from the routing domain immediately, if + the processing functions cannot find at that time. + + 1.5 Some typo have fixed. + + --- EXAMPLE --- + static int + opaque_lsa_originate_callback (list funclist, void *lsa_type_dependent) + ^^^^^ + --- EXAMPLE --- + +2. Feature enhancements + + 2.1 According to the description of rfc2328 in section 10.8, any + change in the router's optional capabilities should trigger + the option re-negotiation procedures with neighbors. + + --- EXCERPT --- + If for some reason the router's optional + capabilities change, the Database Exchange procedure should be + restarted by reverting to neighbor state ExStart. + --- EXCERPT --- + + For the opaque-capability changes, this feature has implemented. + More specifically, if "ospf opaque-lsa" or "no ospf opaque-lsa" + VTY command is given at runtime, all self-originated LSAs will + be flushed immediately and then all neighbor status will be + forced to ExStart by generating SeqNumberMismatch events. + + 2.1 When we change opaque-capability dynamically (ON -> OFF -> ON), + there was no trigger at "OFF->ON" timing to reactivate opaque + LSA handling modules (such as MPLS-TE) that have once forcibly + stopped at "ON->OFF" timing. + Now this dynamic reactivation feature has added. + + 2.2 The MPLS-TE module now referes to "draft-katz-yeung-ospf-traffic + -05.txt", no significant changes with 04 version, though. + +----- * ----- * ----- * ----- * ----- * ----- * ----- * ----- * ----- * ----- +Changes 2001.03.28 + + Initial release of Opaque-LSA/MPLS-TE extensions for the zebra/ospfd. |