Commit Graph

705 Commits

Author SHA1 Message Date
Nicholas J. Kain
6047f04a12 Guard against carrier being spuriously set down.
Corrects a possible regression introduced by the previous patch.
2022-01-11 22:16:44 -05:00
Nicholas J. Kain
f3766990f9 Speed up interface carrier checking.
This is done by performing one synchronous query for carrier state at
the start of the program; after that, we just monitor the nlsocket for
carrier state changes and update the cached state accordingly.

The benefit is that ifchd needs to do a lot less work and this should
reduce the CPU cycle consumption; prior to this commit, the CPU time
ends up being a few CPU-minutes per month.
2021-04-25 05:55:01 -04:00
Nicholas J. Kain
ad1546ecb0 ifchd: Minor corrections in informative prints 2021-04-25 05:45:21 -04:00
Nicholas J. Kain
d61902cad4 ifchd: Better handle incomplete command buffers.
Before the handling would constantly acculmulate a prefix of previous
incomplete commands.  Now it still has a latent defect where the entire
buffer will be discarded given a spurious command, but ndhc shouldn't
generate such commands so it shouldn't matter.
2021-04-25 05:45:21 -04:00
Nicholas J. Kain
eb463ab086 poll() revents bits are reset by poll() 2021-04-25 05:45:21 -04:00
Nicholas J. Kain
c8eb238f31 ifchd-parse: Simplify leftover-buffer copying. 2021-04-24 20:50:08 -04:00
Nicholas J. Kain
b8bffa16c5 Build without -fno-strict-overflow.
I only built with this flag to mitigate accidental UB.  Now that
UBSan exists, there's no point as UBSan does better and actually
allows offending code to be located and fixed easily.
2020-11-25 04:44:26 -05:00
Nicholas J. Kain
28e1b34bac Simplify logging and fix some format specifiers. 2020-11-24 21:02:51 -05:00
Nicholas J. Kain
2e3c504cc5 Add and use safe_ftruncate() wrapper. 2020-11-06 18:47:51 -05:00
Nicholas J. Kain
a91b1d34ac Check and clear events from poll() even if interrupted by signal. 2020-11-06 18:32:38 -05:00
Nicholas J. Kain
1e60b4b8e7 Optimize internet checksum to use 32-bit fetches.
It's actually not very hard to do everything with 32-bit fetches
and 32-bit registers, aside from the necessary bit fetches for
tails.
2020-11-01 01:06:58 -04:00
Nicholas J. Kain
ff8b910222 Rename net_checksum161c to net_checksum16.
How many characters are summed at a time is an irrelevant
implementation detail.
2020-10-27 14:45:15 -04:00
Nicholas J. Kain
918555219b Make fingerprinting more robust with rapid carrier state changes.
The existing code tracked the fingerprinting completion state with
a bool, which is insufficient; what is required is a tristate
(none|inprogress|done) where the inprogress state reverts to
none on carrier down, but done state stays done on carrier down.
2020-10-27 14:14:18 -04:00
Nicholas J. Kain
d07469a5fa Optimize internet checksum to use 16-bit fetches.
We could use 32-bit fetches with the same technique on 64-bit
architectures, or SIMD could be used to do very fast 128-bit
fetches, but this isn't a performance bottleneck and this
method is very simple and relatively fast.
2020-10-24 17:14:20 -04:00
Nicholas J. Kain
5fdf3bd83e Remove unused definitions from CMakeLists.txt. 2020-10-24 15:45:57 -04:00
Nicholas J. Kain
b86735c478 Minor signal handling fixes. 2020-10-21 09:49:22 -04:00
Nicholas J. Kain
d58338b70b serverid checks for RENEW and REBIND DHCP(ACK|NAK) were too strict.
Use the same test that is already used for the DHCPACK sent by
the server in the REQUESTING state.
2020-10-21 06:47:11 -04:00
Nicholas J. Kain
b07ebb60a2 Revert bb7841f3ee.
That trivial warning fix causes warnings with -Wcast-qual, which
is a more valid warning than whatever was causing the warnings
under -Wall -pedantic with the normal Makefile.
2020-10-20 07:55:47 -04:00
Nicholas J. Kain
48bbb0b805 Add a missing newline to an error message.
I'm almost certain I remember adding this before, but I guess
it somehow got lost.
2020-10-20 07:53:23 -04:00
Nicholas J. Kain
bb7841f3ee Trivial warning fix. 2020-10-20 07:18:49 -04:00
Nicholas J. Kain
e6b5247d54 Update the plain Makefile. 2020-10-20 07:18:19 -04:00
Nicholas J. Kain
1ba7c9a2ba Documentation update -- signalfd and epoll aren't used anymore. 2020-10-20 07:06:19 -04:00
Nicholas J. Kain
f1e20305ba Don't depend on external ncmlib. 2020-10-20 07:00:20 -04:00
Nicholas J. Kain
4575f74164 Remove legacy support for exiting after obtaining a DHCP lease. 2020-10-20 06:55:04 -04:00
Nicholas J. Kain
ade4e988af Remove legacy support for forking to background. 2020-10-20 06:55:04 -04:00
Nicholas J. Kain
58067200d6 Remove legacy support for writing a pidfile. 2020-10-20 06:55:04 -04:00
Nicholas J. Kain
4d33c00e04 Use poll() instead of epoll() for ndhc-master. 2020-10-20 05:58:29 -04:00
Nicholas J. Kain
af9d45719f Use poll() instead of epoll() for ifchd and sockd.
poll() is simpler, portable, and the scalability advantages of
epoll are inapplicable here.
2020-10-20 05:41:09 -04:00
Nicholas J. Kain
06a541261e Stop using signalfd and audit signal handling code.
There's really no advantage to using signalfd in ndhc, particularly
since the normal POSIX signal API is now used for handling SIGCHLD in
ndhc-master.  So just use the tried and true volatile sig_atomic_t set
and check approach.

The only intended behavior change is in the dhcp RELEASE state --
before there would be a spurious attempt at renewing a nonexistent
lease when the RENEW signal was received.
2020-10-20 04:42:58 -04:00
Nicholas J. Kain
8d89ca9f19 Reliably force restart when a subprocess has a fatal error.
Suppose a system call such as bind() fails in the sockd subprocess in
request_sockd_fd().  sockd will suicide().  This will send a SIGCHLD to
the master process, which the master process should respond to by
calling suicide(), forcing a process supervisor to respawn the entire
ndhc program.

But, this doesn't reliably happen prior to this commit because of the
interaction between request_sock_fd() and signalfd() [or equivalently
self-pipe-trick] signal handling.

request_sock_fd() makes ndhc-master synchronously wait for a response
from sockd via safe_recvmsg().  The normal goto-like signal handling
path is suppressed when using signalfd() , so when SIGCHLD is received,
it will not be handled until io is dispatched for the signalfd or pipe.
But such code will never be reached because ndhc-master is waiting in
safe_recvmsg() and thus never polls signal fd status.

So, revert to using traditional POSIX sigaction() for SIGCHLD, which
provides exactly the required behavior for proper functioning.
2020-10-20 04:41:51 -04:00
Nicholas J. Kain
f0340b1475 Correct ba046c02c7.
Apparently I had forgotten the counter-intuitive semantics of
signalfd(): it's necessary to BLOCK the signals that will be
handled exclusively by signalfd() so that the default POSIX
signal handling mechanism won't intercept the signals first.

The lack of response to ctrl+c is a legitimate bug that is
now properly fixed; ba046c02c7 fixed that issue, but
regressed the handling of other signals.
2020-10-19 13:03:35 -04:00
Nicholas J. Kain
32bc422d0e Add a heuristic to detect when server ignores dhcp renews.
If we get no response to three renews (unicast), switch to sending
rebinds (broadcast).  Servers are supposed to always reply with
a DHCPACK or DHCPNAK even if the server doesn't update its internal
lease duration database, so this behavior should be RFC compliant.
2020-10-19 07:03:03 -04:00
Nicholas J. Kain
f4365897bc Make renew and rebinding directly track whether DHCPREQUEST was sent.
Before it was inferred by examining timeouts.  Also, simplify
the associated timeout code so that there are no longer effectively
two redundant paths.
2020-10-19 06:36:26 -04:00
Nicholas J. Kain
4df035ced3 Make sure xids in packets sent conform to RFC2131 pg36 table5. 2020-10-19 05:48:52 -04:00
Nicholas J. Kain
5dc35eca6d Merge send_renew() and send_rebind() into send_renew_or_rebind(). 2020-10-19 05:26:47 -04:00
Nicholas J. Kain
23d23c108a DHCPREQUEST in REBINDING state shouldn't have reqip option.
See RFC2131 pg31 paragraph 6.
2020-10-19 05:13:47 -04:00
Nicholas J. Kain
9b8c63d998 Give up on fingerprinting router/gateway if it doesn't reply. 2020-10-19 04:28:06 -04:00
Nicholas J. Kain
87ac82fa45 Add back project() for CMake. Corrects regression in c5a1edd5f6. 2020-10-19 04:28:06 -04:00
Nicholas J. Kain
b2af308647 Trivial refactoring. 2020-10-19 04:28:06 -04:00
Nicholas J. Kain
7bd551d564 Give up on fingerprinting relay agent/server if it doesn't reply.
Try to send/wait three times; then if there's no response, then
assume that the relay agent is ignoring or firewalled from
receiving ARP requests.
2020-10-19 04:28:06 -04:00
Nicholas J. Kain
ba046c02c7 Make custom SIGUSR signals work again.
These were broken long ago when converting from signal()
to sigprocmask().  This change also makes ctrl+c work again.
2020-09-02 21:04:59 -04:00
Nicholas J. Kain
fb143995d2 Silence unused variable warnings in Ragel-generated code. 2019-01-01 01:10:34 -05:00
Nicholas J. Kain
253a97662d sockd: Preserve a const qualifier. 2018-10-26 13:07:37 -04:00
Nicholas J. Kain
3f5285b7ce options.c: Remove an unnecessary cast. 2018-10-26 13:07:19 -04:00
Nicholas J. Kain
c5a1edd5f6 Use more modern CMake syntax. 2018-10-26 13:07:05 -04:00
Nicholas J. Kain
56b6ae2cd3 Quit using NULL macro. 2018-10-26 07:17:39 -04:00
Nicholas J. Kain
05a075aeb2 Replace '(c)' with 'Copyright'.
'(c)' may not be a valid substitute for 'Copyright' in some legal
domains/interpretations.  So be safe, since I obviously am asserting
copyright on my legal work.
2018-10-26 07:11:16 -04:00
Nicholas J. Kain
8983df3c86 Update copyright dates. 2018-02-18 08:25:10 -05:00
Nicholas J. Kain
a66f007931 Trivial documentation updates. 2018-02-18 08:25:10 -05:00
Nicholas J. Kain
e08d3b15b5 Remove seccomp support.
It breaks with the existing whitelists on the latest glibc and is
just too much maintenance burden.  It also causes the most questions
for new users.

Something like openbsd's pledge() would be fine, but I have no
intention of maintaining such a thing.

Most of the value-gain would come from disallowing high-risk
syscalls like ptrace() and the perf syscalls, anyway.

ndhc already uses extensive defense-in-depth and wasn't using
seccomp on non-(x86|x86-64) platforms, so it's not a huge loss.
2018-02-09 03:33:04 -05:00