diff options
| author | Claudio Jeker <claudio@cvs.openbsd.org> | 2015-06-21 12:16:30 +0000 |
|---|---|---|
| committer | Claudio Jeker <claudio@cvs.openbsd.org> | 2015-06-21 12:16:30 +0000 |
| commit | fcbe0144a9d84344101b85e0e75c55c48c3aedec (patch) | |
| tree | d2f47138a35d8a7c372401e7e4167d2aea4e139a | |
| parent | 21b29d8136604b9cb75567326275572803aec0f2 (diff) | |
There is a race between sending notifications to the SE and getting a new
peer_up event in the RDE. This can be triggered by graceful restart. So
remove the panic and replace it with roughly what peer_down does.
OK phessler and henning
| -rw-r--r-- | usr.sbin/bgpd/rde.c | 17 |
1 files changed, 14 insertions, 3 deletions
diff --git a/usr.sbin/bgpd/rde.c b/usr.sbin/bgpd/rde.c index fea40aef92a..8bc792566d9 100644 --- a/usr.sbin/bgpd/rde.c +++ b/usr.sbin/bgpd/rde.c @@ -1,4 +1,4 @@ -/* $OpenBSD: rde.c,v 1.331 2015/06/19 14:54:12 phessler Exp $ */ +/* $OpenBSD: rde.c,v 1.332 2015/06/21 12:16:29 claudio Exp $ */ /* * Copyright (c) 2003, 2004 Henning Brauer <henning@openbsd.org> @@ -3190,8 +3190,19 @@ peer_up(u_int32_t id, struct session_up *sup) } if (peer->state != PEER_DOWN && peer->state != PEER_NONE && - peer->state != PEER_UP) - fatalx("peer_up: bad state"); + peer->state != PEER_UP) { + /* + * There is a race condition when doing PEER_ERR -> PEER_DOWN. + * So just do a full reset of the peer here. + */ + for (i = 0; i < AID_MAX; i++) { + peer->staletime[i] = 0; + peer_flush(peer, i); + } + up_down(peer); + peer->prefix_cnt = 0; + peer->state = PEER_DOWN; + } peer->remote_bgpid = ntohl(sup->remote_bgpid); peer->short_as = sup->short_as; memcpy(&peer->remote_addr, &sup->remote_addr, |