Under specific circumstances there might be two mixed-up
netlink sessions (one for scan, the other for route change
request). This patch separates netlink scans and requests
to two fds (and seq counters).
This should fix http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=428865
Old AS path maching supposes thath AS number appears
only once in AS path, but that is not true. It also
contains some bugs related to AS path sets.
New code does not use any assumptions about semantic
structure of AS path. It is asymptotically slower than
the old code, but on real paths it is not significant.
It also allows '?' for matching one arbitrary AS number.
Cryptographic authentication in OSPF is defective by
design - there might be several packets independently
sent to the network (for example HELLO, LSUPD and LSACK)
where they might be reordered and that causes crypt.
sequence number error.
That can be workarounded by not incresing sequence number
too often. Now we update it only when last packet was sent
before at least one second. This can constitute a risk of
replay attacks, but RFC supposes something similar (like time
in seconds used as CSN).
If a DBDES packet from a master to a slave is lost, then the old code
does not retransmit it and instead send a next one with the same
sequence number. That leads to silent desynchronization of LSA
databases.
Prefix sets were broken beyond any repair and have to be reimplemented.
They are reimplemented using a trie with bitmasks in nodes.
There is also change in the interpretation of minus prefix pattern,
but the old interpretation was already inconsistent with
the documentation and broken.
There is also some bugfixes in filter code related to set variables.
Routes comming through pipe from primary to secondary table were
filtered by both EXPORT and IMPORT filters, but they should be
only filtered by EXPORT filters.
KRF_INSTALLED flag was not cleared during reconfiguration
that lead to not removing routes during reconfigure when
export rules changed.
We also should not try to remove routes we didi not installed,
on Linux this leads to warnings (as kernel checks route source
field and do not allow to remove non-bird routes) but we should
not rely on it.
AS4 optional attribute errors were handled by session
drop (according to BGP RFC). This patch implements
error handling according to new BGP AS4 draft (*)
- ignoring invalid AS4 optional attributes.
(*) http://www.ietf.org/internet-drafts/draft-chen-rfc4893bis-02.txt
This patch extends the length for attributes from 1024 to 2048
(because both AS_PATH and AS4_PATH attributes take 2+4 B per AS).
If there is not enough space for attributes, Bird skips that
route group. Old behavior (skipping remaining attributes)
leads to skipping required attributes and session drop.
When OSPF neighbor state drops down to EXSTART,
clear LSA request and retransmit lists, as specified
by RFC. I hope that this will prevent oscillations
between EXSTART and LOADING states, which sometimes
happened.
It also contains related fix from Yury Shevchuk that
properly resets DB summary list iterator.
Filter code used 'aux' integer field of 'symbol' struct to store ptr
to next symbol and both 'aux2' and 'def' fields for value.
Changed to just 'def' for value and 'aux2' for ptr to next symbol.
Also another minor bugfix.
When capability related error is received, next connect will be
without capabilities. Also cease error subcodes descriptions
(according to [RFC4486]) are added.
BGP keeps its copy of configuration ptr and didn't update it during
reconfiguration. But old configuration is freed during reconfiguration.
That leads to unnecessary reset of BGP connection during reconfiguration
(old conf is corrupted and therefore different) and possibly other strange
behavior.
Fixes two race conditions causing crash of Bird, several unhandled
cases during BGP initialization, and some other bugs. Also changes
handling of startup delay to be more useful and implement
reporting of last error in 'show protocols' command.
WALK_LIST_DELSAFE (in ev_run_list) is not safe with regard
to deletion of next node. When some events are rescheduled
during event execution, it may lead to deletion of next
node and some events are skipped. Such skipped nodes remain
in temporary list on stack and the last of them contains
'next' pointer to stack area. When this event is later
scheduled, it damages stack area trying to remove it from
the list, which leads to random crashes with funny
backtraces :-) .
The core state machine was broken - it didn't free resources
in START -> DOWN transition and might freed resources after
UP -> STOP transition before protocol turned down. It leads
to deadlock on olock acquisition when lock was not freed
during previous stop.
The current behavior is that resources, allocated during
DOWN -> * transition, are freed in * -> DOWN transition,
and flushing (scheduled in UP -> *) just counteract
feeding (scheduled in * -> UP). Protocol fell down
when both flushing is done (if needed) and protocol
reports DOWN.
BTW, is thera a reason why neighbour cache item acquired
by protocol is not tracked by resource mechanism?
When protocol started, feeding was scheduled. If protocol
got down before feeding was executed, then function
responsible for connecting protocol to kernel routing
tables was called after the function responsible for
disconnecting, then resource pool of protocol was freed,
but freed linked list structures remains in the list.
