imsg into the buffer. since imsg_get by definition only returns one imsg we
missed the next imsg(s) until the next poll event on the socket in question,
building up a queue on that socket. didn't show up as a problem yet...
factor out imsg_read, which reads into the buffer. imsg_get now entirely
operates on the buffers and does not read(2) itself.
make all callers cope by calling imsg_read on poll events and calling
imsg_get in a loop until all imsgs are processed.
internal view to the kernel routing table respectively removing them all from
the kernel routing table
kroute_shutdown is now a simple wrapper to kroute_fib_decouple
the error number as parameter instead of accessing errno, because in one
place the error number was not in errno but fetched from a socket.
now, of course it makes much more sense to just set errno to the error number
just fecthed in this one place instead of having hundreds of fatal() calls
all transfer the errno round and round and round...
fix this, and also provide a fatalx, which does not care for errno and doesn't
invoke strerror.
oh, btw, in the place where we fetch the err # from the socket, we don't
call fatal anymore anyway...
of void they can now report errors upstream and do not need to panic any
more. so do that and handle the errors in bgpd.c in the vein that we at least
can clean up before exit.
there are no direct fatal() call in kroute.c now any more, nor any in bgpd.c
after forking.
periodic scans:
-keep a tree of nexthops with valid/invalid flags
-provide kroute_match, which takes an IP address and gives the kernel route
for that
-find the kernel route for a given nexthop with that
-keep a marker on the kernel route that a nexthop depends on it
-on removal of the kernel route, re-evaluate the affected nexthops for
validity.
ok claudio@
them when shutting down.
the rde needs the main process to clean up the routing table on exit so the
parent process needs to be in service until the RDE is done.
ok claudio@
o define new imsg types for this
o process these imsgs in the parent process
now "only" debugging and the rde sending these messages is missing.
ok claudio@
finally gives us read buffers per pipe instead of per process, eleminating
a possible race.
also gets us a real imsg_init() that does all the boring init work
introduce msgbuf API and bundle all info needed for the write buffers in a
struct msgbuf.
also switch to a write queue per handled connection (each bgp session, each
pipe) instead of one big one.
fixes some subtle problems and is overall nicer.
ok claudio@
aligned memory (u_int16_t) therefor crashed the RDE on my sparc64. All
buffer specific functions use now void * instead of u_char * so most cast
are now history. Tested on sparc64 and i386. OK henning@
condition happened. fatal()s from subsystems used by all 3 processes like
the imsg subsystem were hard to track down without knowing in which process
the condition happened.
started by me some time ago with moral support from theo, the proceeded up to
the point where the session engine worked correctly. claudio jeker joined
then and did a lot of work in the RDE.
it is not particulary usefull as application right now as parts are still
missing but is imported to enable more people to work on it.
status:
BGP sessions get established fine, OPEN messages and then KEEPALIVEs
exchanged etc. session FSM works fine; NOTIFICATIONs are handled fine, and
all connection drops etc I provoked get handled fine.
Incoming UPDATE messgages are parsed well and the data entered to the RIB,
the decision process is not yet there, neither is outgoing UPDATEs or sync
to the kernel routing table.
not connected to the builds yet.
