0
0
mirror of https://gitlab.nic.cz/labs/bird.git synced 2024-10-18 18:08:45 +00:00
Commit Graph

5026 Commits

Author SHA1 Message Date
Maria Matejka
1e7fb948a8 RCU: Rewritten to be more straightforward 2024-06-24 09:42:03 +02:00
Maria Matejka
78a7d16c9d Unit test for RCU 2024-06-24 09:42:03 +02:00
Maria Matejka
178ae9ba9e Netindex: consistency checks and deletion of whole hash 2024-06-24 09:42:03 +02:00
Maria Matejka
1371661156 If debugging, store a malloc circular log for easier debugging 2024-06-24 09:42:03 +02:00
Maria Matejka
102cea56f2 Hostentry: made src ea_list atomic to help with consistency 2024-06-24 09:42:03 +02:00
Maria Matejka
7ef2a042bc Slab: can be deleted as a single object 2024-06-24 09:42:03 +02:00
Maria Matejka
a9208e602d Export: Next feed step-up is now internal
There is no much reason to do anything else than the netindex technique.
2024-06-24 09:42:03 +02:00
Maria Matejka
dd841dd472 BGP: Simpler hashing in export table
We need a hashing simple enough to allow for feeding by netindex.
2024-06-24 09:42:03 +02:00
Maria Matejka
5ed9dcb1b6 BGP: Export uses common attribute cache
There is no real need for storing bucket attributes locally and we may
save some memory by caching the attributes in one central place.

If this becomes a contention problem, we should reduce the lock load
of the central attribute cache.
2024-06-24 09:42:03 +02:00
Maria Matejka
ea9f64e0ec BGP: pending TX prefixes link netindex instead of copying net_addr
This helps with memory consumption, allows for removal of multiple
slab/mblock ifs and prepares for easier feeds.
2024-06-24 09:42:03 +02:00
Maria Matejka
1eed19cadb BGP: refactored pending TX back into channel 2024-06-24 09:42:03 +02:00
Maria Matejka
5cbebfc940 Real almost-lockless feeds and more pull-like exports
Introducing a new omnipotent internal API to just pass route updates
from whatever point wherever we want.

From now on, all the exports should be processed by RT_WALK_EXPORTS
macro, and you can also issue a separate feed-only request to just get a
feed and finish.

The exporters can now also stop and the readers must expect that to
happen and recover. Main tables don't stop, though.
2024-06-24 09:42:03 +02:00
Maria Matejka
1d0e47e74b BGP: explicitly sending route refresh from CLI 2024-06-24 09:42:03 +02:00
Maria Matejka
402a574050 Protocol filter reload is now done by 'reload filters'
This prepares for the separate 'reload bgp' command triggering BGP Route Refresh
2024-06-24 09:42:03 +02:00
Maria Matejka
2357fc7d95 Static: Fixed undefined nexthop padding problems 2024-06-24 09:42:03 +02:00
Maria Matejka
b6e70f5ecf Netindex: fixed resolving net by index 2024-06-24 09:42:03 +02:00
Maria Matejka
2537e7f924 RIP partial reload never worked properly, running full reload always 2024-06-24 09:42:03 +02:00
Maria Matejka
b9db1d70fe Dropping obsolete protocol Perf
We have now better methods how to measure overall performance
and this obsolete protocol has basically rotten away. If anybody
needs its features, feel free to revive it in future.
2024-06-24 09:42:03 +02:00
Maria Matejka
23c861f95e OSPF partial reload never worked properly, running full reload always 2024-06-24 09:42:03 +02:00
Maria Matejka
89c92abb7a Simplified temporary resources
Also TMP_SAVED now uses the CLEANUP hook to allow for breaks and returns
2024-06-24 09:42:03 +02:00
Maria Matejka
39f69064ec SKIP_BACK_DECLARE: easier embedded-to-parent typecasting 2024-06-24 09:42:03 +02:00
Maria Matejka
66e98f69f8 Lockless feed of a single net 2024-06-24 09:42:03 +02:00
Maria Matejka
21c1df625a Using ea_lookup_tmp() for temporarily keeping attribute references
To avoid needs for keeping local temporary references for attributes,
now one can use ea_lookup_tmp() to ensure that the attributes are
valid and stored until the task ends. After that, the attributes are
automatically unref'd and also deallocated if needed.
2024-06-24 09:42:03 +02:00
Maria Matejka
bff3b7eb2c Refactored the deferring framework into a separate structure 2024-06-24 09:42:03 +02:00
Maria Matejka
e39a76d6f6 Avoiding RCU synchronization deadlock when locking in critical section
Explicitly marking domains eligible for RCU synchronization. It's then
forbidden to lock these domains in RCU critical section to avoid
possible deadlock.
2024-06-24 09:42:03 +02:00
Maria Matejka
2ca9e6fdeb RCU Unwinder refactored from route table to a separate structure 2024-06-24 09:42:03 +02:00
Maria Matejka
020c34136f Table feeds are now lockless
This commit makes the route chains in the tables atomic. This allows not
only standard exports but also feeds and bulk exports to be processed
without ever locking the table.

Design note: the overall data structures are quite brittle. We're using
RCU read-locks to keep track about readers, and we're indicating ongoing
work on the data structures by prepending a REF_OBSOLETE sentinel node
to make every reader go waiting.

All the operations are intended to stay inside nest/rt-table.c and it
may be even best to further refactor the code to hide the routing table
internal structure inside there. Nobody shall definitely write any
routines manipulating live routes in tables from outside.
2024-06-24 09:42:03 +02:00
Maria Matejka
629e29febf Fixed flush condition when stale cycle valid/set indicators wrap around 2024-06-24 09:42:03 +02:00
Maria Matejka
074384d6ba Kernel: replaced synchronous prune by a refeed 2024-06-24 09:42:03 +02:00
Maria Matejka
37e8cffc0a Lockless hostentry resolution
Now the hostentry doesn't need to lock table, instead it tracks the
hostentry version and retries if the hostentry changed while updating.
2024-06-24 09:42:03 +02:00
Maria Matejka
c872b8f826 Cached route attributes now have explicitly marked layers
Also the rta_* functions renamed to ea_* functions
2024-06-24 09:42:03 +02:00
Maria Matejka
ba3b8607fb Netindex: Dropping tmp handle mechanism in favor of deferred lfuc_unlock 2024-06-24 09:42:03 +02:00
Maria Matejka
a66be7641e Fixed all implicit seq_cst warnings caused by the previous commit 2024-06-24 09:42:03 +02:00
Maria Matejka
51795c3e05 Enabling warnings for implicit seq_cst in atomic access 2024-06-24 09:42:03 +02:00
Maria Matejka
72ecc1d99d Doc: short migration info for BIRD 2 -> 3 2024-06-24 09:42:03 +02:00
Maria Matejka
53e3bf9063 Fixed announcement inconsistency between feeds and regular exports 2024-06-24 09:42:03 +02:00
Maria Matejka
51ebbda95c Fixed best route announcements after NHU
When more routes in one net changed at once, the best route
announcements were inconsistent which confused exporters.
2024-06-24 09:42:03 +02:00
Maria Matejka
35ff820a3b Route flag REF_OBSOLETE
Marking routes obsolete when being removed from table, just to be sure.
2024-06-24 09:42:03 +02:00
Maria Matejka
929df85f1e Removed the obsolete birdloop flagger
This was useful when events were locking. As now sending events is lockless,
we can drop this obsolete routine for good.
2024-06-24 09:42:03 +02:00
Maria Matejka
feb16991ee Simplified table deletion (one less step) 2024-06-24 09:42:03 +02:00
Maria Matejka
7c4db11b9f Hostentry usecount converted to lfuc 2024-06-24 09:42:03 +02:00
Maria Matejka
a3b125e223 Lockfree usecount: deferring unlocks to the metaloop
This allows us for easy temporary locks without additional burden of explicit cleanup.
2024-06-24 09:42:03 +02:00
Maria Matejka
08f67979ce Route table export journal converted to the generic structure 2024-06-24 09:42:03 +02:00
Maria Matejka
3787c0f7a5 Lock free journal refactored into a separate data structure 2024-06-24 09:42:03 +02:00
Maria Matejka
cb6e7123c4 Replacing the NHU table loop flag with a proper event 2024-06-24 09:42:03 +02:00
Maria Matejka
90127cb2d2 Flowspec links don't peruse the all-hooks list for lookup 2024-06-24 09:42:03 +02:00
Maria Matejka
22f58ac316 Refactoring: back-merged export data structures
Table-specific structures inheriting commons are too crazy.
2024-06-24 09:42:03 +02:00
Maria Matejka
dea78a5c51 Export table dump temporarily switched off 2024-06-24 09:42:03 +02:00
Maria Matejka
561116f6fb Added and explicit target table to the ROA subscription object 2024-06-24 09:42:03 +02:00
Maria Matejka
35fe916155 Refactoring of struct rt_pending_export
Now it stores const rte * instead of struct rte_storage * to allow for
different storage backends.
2024-06-24 09:42:03 +02:00