ndhc/ChangeLog
Nicholas J. Kain 7a09ba2cf7 Initial commit.
2010-11-12 04:02:18 -05:00

65 lines
2.0 KiB
Plaintext

20040906
accept_conns() trivial correctness fix: EINTR return from accept should
force continue rather than break in handling loop.
fail_on_fdne() is now file_exists() with a slightly different API.
20040818
Use inet_aton rather than custom function; daemon.[ch] removed.
20040817
ifchd no longer logs a "FATAL - select returned an error!" when it
is -KILL'ed.
20040626
pending_exit is now static volatile sig_atomic_t rather than
static volatile int for pedantic correctness issues.
20040614
Added --interface (-i) option that allows one to restrict the
interfaces that ifchd clients are allowed to modify. I
reccomend that this flag be used to further mitigate the possible
effects of a compromised client. By default, all interfaces
may be modified by clients.
Minor cleanups.
Clear corresponding namesvrs and domains on socket disconnection.
20040613
Factor out Linux-specific code into linux.c.
Nameservers and search domains now tracked per-connection to
prevent races where a client could force a writeout of data
provided by another client.
Interface name now cleared on connection close.
Make all headers idempotent.
Minor improvements to error messages.
20040612
Removed iffd[] array; this change makes the state machine action
functions (perform_*) depend on less external state.
HOSTNAME command now supported; it is not enabled by default.
If you wish for remote daemons to be able to change the hostname
of the local machine, use the --hostname (-o) option.
20040610
Added MTU support.
20040609
Robustify so that suicide() isn't called at every possible failure.
Harmless failures will now simply print an error to the log. Risky
or severe errors still suicide().
20040608
Added support for resolv.conf, namely "nameserver" and "search" entries.
Make strlist more robust and paranoid.
20040607
Improved accept() code to be much more robust; it can now properly error
recover. I hope this will fix the elusive random-exit problem that
plagues machines running the 2.6 kernel.
Refactored dispatch_work() to be cleaner.