[Bug 292604] netwait always warns of lack of network functionality

https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=292604

Jeremy Chadwick <[email protected]> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |[email protected]

--- Comment #6 from Jeremy Chadwick <[email protected]> ---
I am the original author of the netwait script.  dougb vetted it.  It's a very
sensitive shell script that has unforeseen effects depending on various network
configurations (especially FreeBSD users relying on VLANs).  Both Doug and I
spent a long time getting it right for everyone.  Yes it's just a shell script,
but how it works / the edge cases are not immediately obvious.

I upgraded my single remaining FreeBSD system from 14.3 to 14.4 then 15.0 and
15.1 and saw the /etc/rc warning.  Needless to say I am quite irritated.

The commits that broke this is
https://github.com/freebsd/freebsd-src/commit/5ead817c3b7af6d6b5fea222ab144db2c3167b22
dated October 2025.  Author is des.

Please test things thoroughly before pushing to src.  It only takes 3 lines to
verify proper behaviour:

netwait_enable="yes"
netwait_ip="8.8.8.8 4.2.2.1"
netwait_if="IFACE_NAME"

Followed by /etc/rc.d/netwait start.  That's it.  Not hard.

While I'm here:

Commit
https://github.com/freebsd/freebsd-src/commit/630c9dba7a2479dbc60858eb94125fb315128b88
-- 11 years ago -- added comments to the top of the script -- specifically
lines 9 through 17 -- that tell users who use DHCP to use
synchronous_dhclient="yes" instead of netwait, and that this is mainly for
static IP systems.

THIS IS WRONG.  I repeat, clearly: THIS IS WRONG.  The author of these comments
does not understand what this script is for and what the purpose is.

The purpose is to verify ACTUAL REAL-WORLD NETWORK CONNECTIVITY is functional
_prior_ to the rest of the rc.d framework/dependencies continuing on.  There
are tons of daemons that require actual functional networking layer before they
should start.  Just because DHCP/dhclient can sit and wait until it exits
cleanly DOES NOT mean the rest of the network (esp. uplink) is actually
functional at that point.

It checks network link state, but it also uses ping -- which is the critical
part.  The entire point of the ping is to tests real-world packet flow
(preferably to an endpoint that is off-network, e.g. Internet, but everyone's
use-case varies).

Furthermore, static IP vs. DHCP has absolutely no relevancy to the purpose of
this script. These comments are completely wrong and misleading, and thus
should be removed.

I have no complaints about the REQUIRE change, or the use of iterating over
multiple interfaces.  It's the comments that are wrong.

-- 
You are receiving this mail because:
You are the assignee for the bug.