rx_hook may be cleaned in some callback so we check it beforeexecuting sk_read().
Fixes segfault with multiple routing tables.
This is bug, isn't it?
Fix problem with local time changes.
Fixes bugs in IPv6 code caused by recent commits.
Ignore messages related to wireless extensions.
Multihop BGP was completely broken, because listening socket has alwaysttl 1.
Merge branch 'dev' into out
Check of socket name length
Do not repeat 'Invalid broadcast address' error message.
'Invalid broadcast address' error is reported only once foran interface and not during every interface scan.
Ignore unknown netlink events.
Bird sometimes reported 'bird: nl_parse_link: Malformed message received'.The cause is that bird asynchronously received netlink packet fromwireless driver about some wireless event on its link layer. In thatcase bird shouldn't complain.
Bugfix for routing table breaking bug.
Here is a patch fixing a bug that causes breakage of a local routingtable during shutdown of Bird. The problem was caused by shutdownof 'device' protocol before shutdown of 'kernel' protocol. When'device' protocol went down, the route (with local network prefix)...
Implementation of MD5 authentication of BGP sessions.
Previous patch reverted. :-(
Buffer overflow fix.
Fix behavior of ipa_opposite().
It was giving wrong results on /30 networks.
Close fd of config file after reconfiguration.
Detach from the TTY properly.
Yes, we will go for 1.0.12
Let's go for 1.0.11.
Real write is only in sk_maybe_write. Previous change partially reverted.Thank you MJ.
Added s->err_hook wrapper that empties socket.
Small changed to reduce the number of warnings.
Yes, I'd like to go to 1.0.10
Time added in password management.
Everything is tested and works. I thins it's right time to go to 1.0.9.
sk_write bugfix for BSD.
Used parameter can be marked as unused. :-) Thanx MJ.
Some new warnings eliminated.
Marked unused parameters as unused.
Deleted some unused code.
Test old instance of BIRD.
Staticized lots of local functions.
Marked unused parameters in sysdep code as such.
Renamed log() to log_msg(), but still keeping the old name as a macro.This is done to avoid clashes with gcc-3.3 which has built-in logarithms :)
Cleaned up sk_reallocate() and friends.Also, removed the `if (s)' test, because I believe that as the wholesocket interface doesn't accent NULL pointers, sk_reallocate() shouldn'tbe the only exception.
OSPF is ready for changing MTU.
typo in README
CONFIG_SKIP_MC_BIND added.BSD hates it, Linux needs it.
Deleted useless rfree. (Socked was freed by cli_free())
Code clean up.
Reverting last patch.
Caching loopback interface.
Minor changes caused by MJ's comment.
#ifdef CONFIG_UNIX_DONTROUTE added.
Useles log()s deleted.
bzero has 2 arguments.
memset -> bzero
Move CLI socket to the newly created CLI's pool.
(thanks to Andreas for the original idea)
Rewritten the I/O loop. All socket operations are now safe, meaning thatyou can delete the socket from anywhere in the hooks and nothing should break.Also, the receive/transmit buffers are now regular xmalloc()'ed buffers,not separate resources which would need shuffling around between pools....
Fix handling on full pipe to client in bird. Prevent packet overflowsfor even only medium sized route table output. Fix a strange garbledoutput problem in the client. The latter seems to be caused by somelibrary doing tcflush while there is still command output pending. So...
Previous change was mistake.
*BSD port added. (Tested on FreeBSD and NetBSD)
Prepared for release 1.0.8.
This probably IS the memory leak we're looking for. Alien routes weren'tcorrectly disposed of.
Releasing version 1.0.6.
Current Linux kernels don't remember rtm_protocol for IPv6 routes and supply RTPROT_BOOT instead.Work around that.
Report link-scope addresses as well.
Added missing includes.
Added missing #include and wondering how could it ever compile.
Don't loop forever when trying to skip an out-sequence netlink reply.
Yes, 1.0.5 released.
Fixed <time.h> vs. <sys/time.h> problems.
Serious bug in ext lsa origination found. Going for 1.0.4.
Version 1.0.3 released.
Released as version 1.0.2.
Released as version 1.0.1.
Check broadcast address sanity before believing it.
If compiled with --enable-debug, don't even try to log to syslog unlessthe user forces it in the config file.
Fix numbering of routing tables in IPv6 version.
Use our own SUN_LEN if libc doesn't provide it.
During initialization, log to both syslog and stderr. When a configurationfile has been read and it doesn't specify any logging, log to syslog only(if syslog is not available, then stderr).
Use SUN_LEN() for length of UNIX domain addresses. This should fix problemswith connection to clients on libc5 machines.
Released as 1.0.0, but marked it as a BETA version.
Fixes for the programmer's manual.
... as well there.
Updated CLI helps.
Fixes to the progdoc.
Spelling fixes to progdoc.
Documented all the sysdeps (only briefly, I admit).
Except for Filters, RIP and OSPF, the progdocs are complete.
Documented sockets, events and timers.
Split random number functions off io.c, so that they can be documentedseparately.
Removed lots of trailing newlines in log messages.
Please note that the only calls which don't add newlines automatically(i.e., don't print a full line of output) are debug() and DBG.
tm_format_reltime() works with both past and future timestamps.
Updated for new scope handling.
Also, provide proper address scopes in struct ifa.
Before configuration file is read, log to both syslog and stderr.
Updated all the Doc files to new format.
Tracing of CLI connections/commands can be now controlledby `debug commands <level>' in the configuration. Level 0 meansno tracing, 1 means connections only, 2 includes all commands.
Support --version and --help.
When in persist mode, don't delete routes from kernel tables even ifthey cease to exist in our routing tables due to protocols having shut down.
Adapted to new rt_notify semantics.
Multicast problems should be gone, although the fix is Linux only and we'llneed to figure out something better when working on new ports.
If a broadcast address is missing, go fake one.
In non-debug mode, set default logging to syslog only, not stderr.
Fixed type in daemonization code.
When not debugging, daemonize automatically.
Forgot to add a spiky comment :)
Autoconf is evil. The sysconfdir and similar variables are unusable inC includes as they contain substitutions specific to make.
Worked around by creating sysconf/paths.h which is created fromthe Makefile instead of by the configure script.
When reporting a bug(), call abort() instead of exit(), so that wecan analyse the core.