| Commit message (Collapse) | Author | Age | Files | Lines |
... | |
| |\ \ \ \ |
|
| | | | | | |
|
| | | | | | |
|
| |/ / / /
| | | | |
| | | | |
| | | | |
| | | | |
| | | | | |
Add environment variable (VTYSH_LOG) for logging.
If a command fails, exit with non-zero exit code and don't
continue multipart commands.
|
| |\ \ \ \
| | | |_|/
| | |/| | |
|
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | | |
Earlier changes to route management caused any route change to cause
the whole RIB to marked as changed. This then causes route flapping.
Fix by being more selective about what changes get flagged.
Bugfix 3488
|
| | | |/
| | |/|
| | | |
| | | | |
This makes code work like comments and probable original design.
|
| |\| | |
|
| | | |
| | | |
| | | |
| | | |
| | | | |
Mostly from incorrect usage of z option. Need to use:
printf(" size = %zd", x->size);
|
| | | |
| | | |
| | | |
| | | |
| | | | |
Make the trivial boolean functions inline to help performance
and code size.
|
| | | |
| | | |
| | | |
| | | |
| | | |
| | | | |
This adds reference counting to avoid cases where bgp instance
is refered to by peers after bgp has been deleted. The style and
format matchs other reference counting in Quagga
|
| | | |
| | | |
| | | |
| | | | |
These trivial functions can just be inlined.
|
| | | |
| | | |
| | | |
| | | |
| | | | |
These trivial functions take more code to call than to just
do inline.
|
| | | |
| | | |
| | | |
| | | | |
The peer_unlock() in node_cleanup could be freeing the peer.
|
|\ \ \ \
| | |/ /
| |/| | |
|
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | | |
Earlier commit for Bug 3434 caused removed routes not to be marked
as SELECTED. This caused them not to be removed. Fix is to always
check both flags.
Bugfix 3479
|
| | | |
| | | |
| | | |
| | | |
| | | | |
When multipath is being used, the state of one nexthop maybe down,
but another one may be up.
|
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | | |
This includes two fixes. First entries that are marked for removal
should not be used (clear SELECTED flag). Also, when recursive route
changes, the CHANGED flag needs to be set to force update.
Bugfix 3434
|
| |\ \ \ |
|
| | | | |
| | | | |
| | | | |
| | | | | |
The command to turn off zebra rib debugging was wrong.
|
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | | |
Revert "Revert "Only announce connected routes if link is detected""
This reverts commit 38701029fe978ca18645408771414a91588413d3.
|
| |/ / /
| | | |
| | | |
| | | |
| | | | |
Adding a route that already exists, or deleting a route that is
already gone, should not be an error.
|
|\| | | |
|
| | | |
| | | |
| | | |
| | | |
| | | |
| | | | |
Want to remember connected routes even if they aren't active at the
current time.
This reverts commit 84cc940a9b74d5a867a5598f4b0f6c657eb61c9c.
|
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | | |
If a link state change event causes a different nexthop, then
the RIB (and FIB) need to be updated. Solved by forcing the CHANGED
flag. This code could be rearranged to be cleaner in future.
Bugfix 3434
|
|\| | | |
|
| | | |
| | | |
| | | |
| | | |
| | | | |
Make the trivial boolean functions inline to help performance
and code size.
|
| |/ /
| | |
| | |
| | |
| | | |
Mostly from incorrect usage of z option. Need to use:
printf(" size = %zd", x->size);
|
|\| | |
|
| | |
| | |
| | |
| | |
| | | |
- added code in vyatta-quagga.postinst to write smux values into
repesctive routing daemon config files if not written till then
|
|\| | |
|
| |\ \ |
|
| | |\ \ |
|
| | | | | |
|
| | |/ /
| |/| |
| | | |
| | | |
| | | | |
Bugfix 3433, 3434
Need to rescan for recursive routes after any change to static route.
|
| | | |
| | | |
| | | |
| | | |
| | | | |
Bugfix 2847
Set TOS to internet control for BGP traffic
|
| | | |
| | | |
| | | |
| | | |
| | | | |
This is the IPV6 analog to earlier changes to IPV4 static route
management.
|
| | | |
| | | |
| | | |
| | | |
| | | | |
The shell here document only strips tabs not spaces,
so the config file ended up with messy entry
|
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | | |
Need to handle all the different flavors of connected routes
in linkstate management
Bugfix: 2904, 3416
|
| | | |
| | | |
| | | |
| | | |
| | | | |
The route_info table can be static const, and the vrf_table can be
static.
|
| |/ /
| | |
| | |
| | |
| | | |
Need to check both IFF_UP and IFF_RUNNING if link-detect is enabled,
before announcing routes.
|
|\| | |
|
| | |
| | |
| | |
| | |
| | | |
This reverts commit a182091782d22218942466b062c5ef2406853491.
More changes that should NOT go into hollywood
|
| | |
| | |
| | |
| | |
| | | |
This reverts commit 17cf20dabb137205497d81f7c55efa45a4e50748.
Stuff for next release snuck into hollywood branch
|
| | | |
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
Having kernel cleanup FIB was not ideal for a couple reasons:
1. It still generates lots of netlink events if there are lots
of static routes.
2. Kernel changes were not accepted in mainline kernel "do it in
userspace"
|
| |\ \ |
|
| | |\ \ |
|
| | | | | |
|
| | |/ /
| | | |
| | | |
| | | |
| | | | |
If route server client is being used, then peer->rib entry
would get freed twice.
|