The motivation for this patch is to address a concerning behavior of
tx-addpath-bestpath-per-AS. Prior to this patch, all paths' TX ID was
pre-determined as the path was received from a peer. However, this meant
that any time the path selected as best from an AS changed, bgpd had no
choice but to withdraw the previous best path, and advertise the new
best-path under a new TX ID. This could cause significant network
disruption, especially for the subset of prefixes coming from only one
AS that were also communicated over a bestpath-per-AS session.
The patch's general approach is best illustrated by
txaddpath_update_ids. After a bestpath run (required for best-per-AS to
know what will and will not be sent as addpaths) ID numbers will be
stripped from paths that no longer need to be sent, and held in a pool.
Then, paths that will be sent as addpaths and do not already have ID
numbers will allocate new ID numbers, pulling first from that pool.
Finally, anything left in the pool will be returned to the allocator.
In order for this to work, ID numbers had to be split by strategy. The
tx-addpath-All strategy would keep every ID number "in use" constantly,
preventing IDs from being transferred to different paths. Rather than
create two variables for ID, this patch create a more generic array that
will easily enable more addpath strategies to be implemented. The
previously described ID manipulations will happen per addpath strategy,
and will only be run for strategies that are enabled on at least one
peer.
Finally, the ID numbers are allocated from an allocator that tracks per
AFI/SAFI/Addpath Strategy which IDs are in use. Though it would be very
improbable, there was the possibility with the free-running counter
approach for rollover to cause two paths on the same prefix to get
assigned the same TX ID. As remote as the possibility is, we prefer to
not leave it to chance.
This ID re-use method is not perfect. In some cases you could still get
withdraw-then-add behaviors where not strictly necessary. In the case of
bestpath-per-AS this requires one AS to advertise a prefix for the first
time, then a second AS withdraws that prefix, all within the space of an
already pending MRAI timer. In those situations a withdraw-then-add is
more forgivable, and fixing it would probably require a much more
significant effort, as IDs would need to be moved to ADVs instead of
paths.
Signed-off-by Mitchell Skiba <mskiba@amazon.com>
This commit introduces unit tests for BGP peer attributes and checks all
three involved components, which are:
- CLI Configuration Input: The appropriate commands to configure the
attribute on either a peer or peer-group are being executed the same way
an end user would do it.
- CLI Configuration Output: The output of 'show running-config' is being
checked for presence/absence of expected configuration strings.
- Internal Data Structures: The internal data structures for maintaining
flag/filter states (value + override + invert) are being checked after
each operation to ensure the override has been implemented properly.
All attributes to be tested must be defined within the 'peer_attrs'
structure, which contains all peer attributes as of today and checks
them with both IPv4 Unicast and IPv6 Unicast. More address families are
supposed to be introduced at a later point in time.
Each attribute is being checked in its own 'clean' BGP environment, so
everything gets reset after each attribute to avoid any weird edge
cases. The 'correct' BGP startup and shutdown routine was taken from
'bgp_main.c' to ensure that we are not leaking any memory or acting
different than the real 'bgpd' would do.
Signed-off-by: Pascal Mathis <mail@pascalmathis.com>
* Add general-purpose DFS traversal code
* Add ability to dump any graph to DOT language
* Add tests for graph datastructure
Signed-off-by: Quentin Young <qlyoung@cumulusnetworks.com>
Needed these while rewriting LSDB iteration.
NB: this commit fails because of a bug in ospf_lsdb_get_next, which will
SEGV when the LSDB is actually empty. Whooo... (this is fixed in the
following commits.)
Signed-off-by: David Lamparter <equinox@opensourcerouting.org>
This adds some common CLI testtool code as well as a tool that has a
bunch of commands to be poked for their correct processing.
The tool doesn't work correctly from a script at stdin at this point
because the vty code will throw away all buffered when it sees EOF, so
the tail end of the input file is lost.
Signed-off-by: David Lamparter <equinox@opensourcerouting.org>
(cherry picked from commit db93eec18d8f1e840b32ba2cdf8baf2510f6e1a5)
837d16c ("*: use array_size() helper macro") accidentally changed one of
the expressions in the backtrace code, which afterwards read:
zlog_backtrace_sigsafe():
if (((size = backtrace(array,array_size(array)) <= 0) ||
which boils down to: (size = backtrace(...) <= 0). The braces were
intended to go: (size = backtrace(...)) <= 0.
All in all, this makes a nice textbook example of the original author
being too clever (trying to save a single line by pulling the assignment
into the condition) and the next person touching the code tripping over
it...
This code occurs another time in zlog_backtrace() where it is actually
correct. Pulling out the assignment nonetheless. Also, new test
program.
Cc: Andrew J. Schorr <ajschorr@alumni.princeton.edu>
Cc: Balaji.G <balajig81@gmail.com>
Cc: Scott Feldman <sfeldma@cumulusnetworks.com>
Signed-off-by: David Lamparter <equinox@opensourcerouting.org>
Change the datastructure for recursive routes. This brings the following
benefits:
By using struct nexthop also to store nexthops obtained by recursive
resolution, we can get rid of quite a bit of code duplication in the fib
management. (rt_netlink, rt_socket, ...)
With the new datastructure we can make use of all available paths when
recursive routes are resolved with multipath routes.
Signed-off-by: Christian Franke <chris@opensourcerouting.org>
Signed-off-by: David Lamparter <equinox@opensourcerouting.org>
DejaGNU seems to be the 'standard' GNU test framework (which by itself
doesn't say much), but it seems relatively usable and the "remote
system" capabilities might come in handy for virtualisation-based tests
for kernel interactions or something.
Signed-off-by: David Lamparter <equinox@opensourcerouting.org>
* lib/table.[ch]
- Add a function (route_table_get_next()) to get the route_node in
a tree that succeeds a given prefix in iteration order.
This allows one to reliably walk nodes in a tree while allowing
modifications, and is useful for achieving scale and
performance. Other approaches are also possible -- the main plus
point of this one is that it does not require any state about
the walk to be maintained in the table data structures.
- Add an iterator for walking the nodes in a tree. This introduces
a new structure (route_table_iter_t) and the following main
functions.
route_table_iter_init()
route_table_iter_pause()
route_table_iter_next()
route_table_iter_cleanup()
The iterator normally uses node pointers and the existing
route_next() function to walk nodes efficiently. When an
iteration is 'paused' with route_table_iter_pause(), it stores
the last prefix processed. The next call to
route_table_iter_next() transparently invokes
route_table_get_next() with the prefix to resume iteration.
* bgpd/bgp_table.[ch]
Add wrappers for the new table features described above.
* tests/table_test.c
Add tests for the new table code.
Signed-off-by: David Lamparter <equinox@opensourcerouting.org>
