Age | Commit message (Collapse) | Author |
|
showNeighborRib() executes net.ParseCIDR() for vrf prefix like
10.100:100:10.0.0.0/24 and then crashes.
Probably ListRib() API should support an option to return sorted
destinations.
Signed-off-by: FUJITA Tomonori <fujita.tomonori@gmail.com>
|
|
|
|
https://github.com/osrg/gobgp/commit/318a91e31b8baed86bb66a3767525da68ae49a5d
The above commit added a bug to close a closed channel; after deleting
a dynamic neighbor, handleFSMMessage() must return. Otherwise
cleanInfiniteChannel() is called twice.
https://travis-ci.org/osrg/gobgp/jobs/483896487
Signed-off-by: FUJITA Tomonori <fujita.tomonori@gmail.com>
|
|
|
|
|
|
also avoid unittest race.
Signed-off-by: FUJITA Tomonori <fujita.tomonori@gmail.com>
|
|
megacheck is deprecated.
Signed-off-by: FUJITA Tomonori <fujita.tomonori@gmail.com>
|
|
Signed-off-by: FUJITA Tomonori <fujita.tomonori@gmail.com>
|
|
Signed-off-by: FUJITA Tomonori <fujita.tomonori@gmail.com>
|
|
Signed-off-by: FUJITA Tomonori <fujita.tomonori@gmail.com>
|
|
Signed-off-by: FUJITA Tomonori <fujita.tomonori@gmail.com>
|
|
channel when sequential retry to connect zebra.
- Introducing MIN_ZAPIVER and MAX_ZAPIVER for avoiding editing files except for zapi.go when new ZAPI version will be supported in future.
- Fix a bug avoiding panic by double close for a channel.
- Changing algorithm for sequential retry to connect zebra.
|
|
|
|
Use ListPeer() instead.
Signed-off-by: FUJITA Tomonori <fujita.tomonori@gmail.com>
|
|
fix the regression of the adj counter due to
301b48532d4a5510c9d4ffdc44eb2754fdd1a3d1
Signed-off-by: FUJITA Tomonori <fujita.tomonori@gmail.com>
|
|
Signed-off-by: FUJITA Tomonori <fujita.tomonori@gmail.com>
|
|
Signed-off-by: FUJITA Tomonori <fujita.tomonori@gmail.com>
|
|
Signed-off-by: FUJITA Tomonori <fujita.tomonori@gmail.com>
|
|
Signed-off-by: FUJITA Tomonori <fujita.tomonori@gmail.com>
|
|
Signed-off-by: FUJITA Tomonori <fujita.tomonori@gmail.com>
|
|
Let's simply use 'validation' because there is no other validation
structure.
We could add validation results from other than rpki so drop rpki name.
Signed-off-by: FUJITA Tomonori <fujita.tomonori@gmail.com>
|
|
Signed-off-by: FUJITA Tomonori <fujita.tomonori@gmail.com>
|
|
- AddPath() with an api.Path object then DeletePath() works with the same api.Path object.
- ListPath() returns an api.Path object then DeletePath() works with the same api.Path object.
The above is guaranteed with and without PeerInfo (SourceAsn and SourceId).
Signed-off-by: FUJITA Tomonori <fujita.tomonori@gmail.com>
|
|
Signed-off-by: FUJITA Tomonori <fujita.tomonori@gmail.com>
|
|
All what we need is cancelling. Let's use the standard way for it,
context.
Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
|
|
before dialing
Use the feature of go 1.11 to remove the hack.
Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
|
|
|
|
this error can be reproduced by:
gobgp p statement s1 add action as-prepend 65001 1
gobgp p add p1 s1
gobgp n 192.168.0.5 p import add p1
gobgp n 192.168.0.5 softresetin
then:
panic: runtime error: invalid memory address or nil pointer dereference
|
|
Seems that bluemix.net is unstable nowadays.
Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
|
|
Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
|
|
AddPaths in AfiSafi should be used; it can handle configuration per
family.
Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
|
|
It's not configuration.
Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
|
|
|
|
|
|
Both should be in PeerState like OpenConfig.
Also remove unused supported_capabilities.
Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
|
|
both assume that getBestFromLocal() returns filtered paths by policies
but it doesn't. Paths filtered by policies are returned as withdraw
paths.
Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
|
|
Also rename PathAttrsAny to PathAttrs. Both are the first choice
rather than the binary format members.
Support SortType member to ListPathRequest to add an option to return
unordered paths.
Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
|
|
Can't modify a Path object in a table.
Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
|
|
use google/protobuf/timestamp.proto instead of our own way to
represent time.
Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
|
|
fixes #1893
Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
|
|
PathAttributeNextHop struct can handle ipv6 address but
NewPathAttributeNextHop() can't.
Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
|
|
The error message of 'go vet' goes to stderr so the length of the
stdout is zero. Always the test passes successfully.
'go vet' exits with non zero when something wrong happens.
Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
|
|
This broke vet for me. Not sure why it didn't break in CI.
|
|
|
|
https://github.com/osrg/gobgp/issues/1763#issuecomment-437594975
Follow Chris's proposal; more consistent with gRPC streaming API.
Also supports context properly.
Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
|
|
fix 'receiver name ss should be consistent with previous receiver name s for BgpServer'.
Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
|
|
Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
|
|
|
|
|
|
timeout context is for only connecting. We should refactor the code to
avoid global context variant.
Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
|