CVS log for src/etc/rc

Up to [cvs.NetBSD.org] / src / etc

Request diff between arbitrary revisions

Keyword substitution: kv
Default branch: MAIN

Ooops, overlooked one MOUNTCRITLOCAL reference in the rename to
CRITLOCALMOUNTED.

Split the local disk availability step into two phases to allow scripts
that pre-populate parts of the system (e.g. a tmpfs based /var) an
easy place to plug in like:

# REQUIRE: mountcritlocal
# BEFORE: MOUNTCRITLOCAL

This also cleans up the existing special handling a bit by separating it
into new scripts. All later scripts now depend on MOUNTCRITLOCAL.
Discussed on tech-userlevel some time ago.

Sync with HEAD

Sync with HEAD

set cloexec to the rc postprocessor fd (rudolf in tech-userlevel)

Sync with HEAD

add +

set 7 and 8 to close on exec.

Rebase to HEAD as of a few days ago.

Rebase.

In the background subshell that flushes partial output every few
seconds, detach from the controlling terminal.  Without this, pressing
control-T (to send SIGINFO from the terminal) would almost always result
in the sleep process printing "about N seconds left of the original 3",
which is useless information.

In rc.subr, add _have_rc_postprocessor function and use it instead of
inline tests like [ -n "${_rc_postprocessor_fd}" ]. The new function
performs a few new tests, including verifying that /etc/rc is still
running (using a new _rc_pid variable set by /etc/rc).

This is intended to deal with the case that a script run from /etc/rc
spawns a background process, then /etc/rc exits, but the background
process still has environment variables inherited from /etc/rc.

Fixes PR 46546.

sync with head.

for a reference, the tree before this commit was tagged
as yamt-pagecache-tag8.

this commit was splitted into small chunks to avoid
a limitation of cvs.  ("Protocol error: too many arguments")

Send a "nop" metadata message to the postprocessor every few seconds,
to flush partial output lines.  This should help with PR 48714.

Add comments about the intended use of each type of metadata message.

Allow "echo -n" to work as expected in rc.d scripts that are executed
by /etc/rc.  Similarly for printf with a format that does not end with
"\n".  Previously, the partial line would not be visible on the console
until a newline was printed, possibly after an annoying delay.

This is done by adding echo() and printf() shell functions to rc.subr,
so that naive use of the echo and printf commands in rc.d scripts will
call these functions instead of the underlying commands.  These shell
functions send a new "nop" metadata message after the partial line, and
the rc_postprocess function in /etc/rc disentangles the partial line of
plain output from the metadata "nop".

Also add a "-n" option to the print_rc_normal function in rc.subr,
and make some cosmetic changes.

print human readable exit code.

Add a postprocessor to /etc/rc, which logs messages to /var/run/rc.log,
and which can suppress output in silent mode.  Silent mode is enabled
via the new rc_silent variable, which defaults to a value that depends
on the kern.boothowto sysctl.

Part of the /etc/rc silent changes requested in PR 41946
and proposed in tech-userlevel.

Sync with HEAD.

Third (and last) commit. See http://mail-index.netbsd.org/source-changes/2009/05/13/msg221222.html

Add rc_directories to specify where to look for rc scripts.
For the moment all scripts must be in /root and non-existent directories
are skipped.

Change how internal boolean variables are used to:
	if ! ${_somevar:-false}; then
		_somevar=true
	fi
(Consisent, slightly quicker, and slightly cleaner)

Add two new rc.conf(5) variables:
    rc_rcorder_flags		extra flags to rcorder(8) in /etc/rc
    rcshutdown_rcorder_flags	extra flags to rcorder(8) in /etc/rc.shutdown

This can be used to specify extra directories to search for rc.d scripts in.
For example, adding the following to rc.conf(5):
	rc_rcorder_flags="/usr/pkg/etc/rc.d/*"
	rcshutdown_rcorder_flags="/usr/pkg/etc/rc.d/*"
will add the files in /usr/pkg/etc/rc.d to the list of files that rcorder(8)
uses to build the list of scripts to start or stop.

I proposed this functionality on tech-userlevel@ over one month ago.
Closes the recent [misc/16888], which asked for a similar feature.

consistently set rc_(fast|force)=yes

- Rename a bunch of "internal" run_rc_command() variables and officially
  make them "externally" available:
	Previous	Current		Purpose
	--------	-------		-------
	_arg		rc_arg		Argument to command, after fast/force
					processing performed (and prefix
					removed)

	_flags		rc_flags	Flags to start the default command
					with.  Defaults to ${name}_flags,
					unless overridden by $flags from the
					environment.  This variable may be
					changed by the precmd method.

	_pid		rc_pid		PID of command (if appropriate).

	_rc_run_fast	rc_fast		Not empty is "fast" was provided.

	_rc_run_force	rc_force	Not empty is "force" was provided.

- Use rc_flags instead of _flags or ${name}_flags in various rc.d scripts,
  so that $flags from the environment overrides ${name}_flags from rc.conf(5).
  Fixes [bin/15800].

use symbolic signal names instead of numbers

explicitly set the umask to 022 to be safe

move date before rcorder

various speed-ups (and related work):
- only load rc.conf if $_rc_conf_loaded is not set
- use case instead of if for various string comparisons
- print the date at the start of the boot as well as at the end
- if $rc_fast_and_loose is set, always run the rc.d scripts in the
  current shell rather than in a subshell. this is not on by default
  because it's potentially dangerous (a rogue command could terminate
  the boot), but it is provided as an optional speedup for people
  with slow machines that have an expensive fork

- describe what rc.shutdown does, and what invokes it
- remove `.sh' suffix on script name in description

pull up rev 1.153 (approved by thorpej):
	don't start rc.d scripts that have a `nostart' keyword

don't start rc.d scripts with a `nostart' keyword

revert previous; the fast-start functionality already existed, and
works at multi-user boot. (starting /etc/rc after single user boot
doesn't exhibit this fast start behaviour).

Add an "rc_fastboot" variable check in rc.subr, that if set adds
"fast" to the start of the run_rc_command arg.

/etc/rc calls each rc.d script with rc_fastboot=yes so that rc.subr
aware scripts to use "faststart" to skip the process running check.

Check rc_configured also here, and bail out if it isn't set to yes.

run scripts in /etc/rc.d/* in the order returned by rcorder

make clearing of /tmp after reboot configurable (default on), as per recent
discussion on current-users

ip6forwarding -> ip6mode change (somehow i forgot to commit...)
PR: 9375

change IPv6 configuration syntax to avoid ambiguity.
ip6mode=host: IPv6 host
ip6mode=autohost: IPv6 host, with autoconfig
ip6mode=router: IPv6 router

backward compatibility to "ip6forwading" is provided.

Allow NFS servers to start up.

Start automount daemon even if NFS client is not activated. This makes
sense for system where "amd" is used to mount local devices like e.g.
CD-ROM drives. Fixes PR misc/9126 by Bjoern Labitzke.

Make some warning message consistient with others.

Since the logger(1) puts space between its arguments, there is no need
to pass it explicitly.

Make sure that warning message starts from the beginning of line.  Some
warning recently added didn't care about it.

Cosmetic changes; folding long lines, using tab instead of spaces where
appropriate and fixing some indentation.

don't need to background xfs now that it auto-backgrounds itself...

Pull up to last week's -current.

Add some boot time warnings:
 - if ypserv=yes but portmap=no
 - if yppasswdd=yes but ypserv=no
 - if nfs_server=yes but portmap=no

Per discussion on tech-userlevel@netbsd.org.

accounting should be controllable via $accounting=YES in rc.conf...

Remove update, it's no longer used.

Use checkyesno to check the state of $rc_configured like for any other
variable set in /etc/rc.conf. Pointed out by Erik Bertelsen
<erik@mediator.uni-c.dk> in pr 8522.

move old dev_mkdb and "kvm_mkdb /netbsd" up into the section which builds
kvm_mkdb based on machdep.boot_kernel. also failover to "kvm_mkdb /netbsd".

send sysctl machdep errors to /dev/null
merge in patches to implement multiple bootup configurations (look for
/etc/etc.current, /etc/etc.default and friends)
only start up NFS services if either nfs_client or nfs_server is yes.

Add support for $defcorename

If machdep.booted_kernel can be found in sysctl output, feed that name into
kvm_mkdb so that when we boot we always have an upto-date kvm db, allowing
dmesg and friends to run without error when we do "boot netbsd.fun" to test
kernels, etc.

add sanity check on rtadvd invocation (allowed on IPv6 routers only).
add rc.conf entry for rtsold.

Pull up revision 1.129:
  Fix denial-of-service attack involving vi preserve.  (christos)

check if vi recovery files are regular files before trying to e-mail them.
Stops a denial of service attack where the user can put a named pipe or
any other device that blocks i/o in /var/tmp/vi.recover/recover.*
[from dynamo@ime.net]

Create temp _motd file in /etc

add IPv6 configuration (ip6forwarding, rtsol, route6d and rtadvd).
Right now netsetart does not bother you even if you do not have
"options INET6" in the kernel.

usbd is dead.

Pull up revision 1.124:
Configure wscons much earlier so that a virtual console can e.g. be
used for "syslogd" output.

Configure wscons much earlier so that a virtual console can e.g. be
used for "syslogd" output.

Change mount_critical_filesystems() so that if given an argument
of "local" it will not mount filesystem types it identifies as
network file systems (currently just nfs). Call the function twice,
before networking starts with arg "local", after networking starts
with arg "network". Remove critfs_require_network as is now redundant.

start xdm after wscons configuration - the X server might want an own
virtual screen

Add dhcrelay startup (allow only if dhcpd isn't started)

Nuke the line 'raidctl -r raid$dev'.  a) The '-r' option is now '-i', and
b) with component labels and clean-bits working, this shouldn't be
required at all on each boot!  Thanks to Andrew Gillham for pointing this out.

- Move critical filesystem mounting code into a shell function and put
  that in rc.subr.
- Support critfs_require_network rc.conf variable.

change to new syslog socket. make symlink from old to new location, too.

rc.wscons can take arguments, so we shouldn't simply source it but exec in
a subshell. Being here, add a prototype wscons_flags variable.

introduce a "wscons" rc.conf flag and run /etc/rc.wscons from /etc/rc
if it is set

No need to involve awk in the securelevel part, since sysctl has the -n switch.
Pointed out by NAKAJIMA Yoshihiro.

Use peers too when setting the time via ntpdate.

Add missing '' for awk in the securelevel part.

oops, that's -n not -z.  pointed out by hubert...

increase the securelevel (using the same rules as before) much earlier in the
boot process.  before cron, sendmail, inetd, etc, are run, but after lkms are
loaded.  this avoids the chances of `@reboot', mail .forwards, remote logins,
etc., happening before the securelevel has been raised.

support dmesg=NO and dmesg_flags, and do not cut the old boot messages..they might be useful.  fix a typo (rbootd is not rarpd)

Enable cron to be turned on and off from rc.conf, like other daemons.
From Soren S. Jorvang <soren@t.dk>, closes his PR bin/5250.

Add $portmap_flags variable, now that portmap has optional flags, and
by default, pass '-l' flag (libwrap logging) by analogy with inetd

save a copy of the boot messages. Closes 6398 from msanders@confusion.net

Add support for RAIDframe configuration as suggested by Eric S. Hvozda
in PR install/6676.

allow /var/crash to be a symlink to a directory

Add USB support.  Supported so far:
* UHCI and OHCI host controllers on PCI
* Hubs
* HID devices withe special drivers for mouse and keyboard
* Printers

Strip comments from "${amd_master}".

add "screenblank" support.

send ntpdate output to syslog rather than stdout (with -s).
suggested by "Soren S.  Jorvang" <soren@t.dk>

exclude 127.127 servers when automagically determining ntpdate hosts

include rc.subr and use appropriately

Move ipnat from /etc/netstart to /etc/rc after /usr is mounted.

Fix pr 4587 (from Matthias Scheler) -- pause slightly after xfs starts.

Pull rev 1.97 up from trunk (mrg)

print a warning if no swap devices are configured and "no_swap" is not YES.

Pull rev 1.96 up from trunk (jonathan)

Add options to rc.conf and rc to start xfs (x font server) and xdm.
Add entries to rc.conf.5.
While we're here, note that the domestic kerberos damons are still
Kerberos IV, not kerberos version 5.

Replace the (slightly broken) check to avoid double mounting of
filesystems by a correct one. (too bad that there was no agreement
about a mount option doing the same...)
Introduce a new rc.conf variable to mount user-defined filesystems
early in startup (eg for syslog, crashdumps, ...).

add $ipmon and $ipmon_flags, to control startup of ipmon(8)

fix an rc.conf flag check which was broken with the "checkyesno" changes

avoid to mount /usr twice if LKM load already did it

no more DEFAULT for $ntpdate_hosts; from Geoff Wing in PR bin/4073

Curt, what is in rc.func ?
I only see rc.subr so I renamed rc.func to rc.subr.

Split rc.conf variables for starting programs into two: an on/off
switch and a set of flags. Get rid of DEFAULT flags entirely. Print
warnings if on/off switches are not set, or are set incorrectly.
Add a shell function to simplify this on/off switch testing.

ldconfig processes /etc/ld.so.conf internally; run it without cat'ing
the contents into an argument vector.

Move ntpdate before `starting network daemons' to get cleaner output
during boot. Besides, it's not a daemon anyway.

add apmd to /etc/rc and /etc/rc.conf per Lennart Augstsson in pr 3313

Fix boot time vi recovery as reported by NAKAJIMA Yoshihiro in pr-3430
Really weird bug. On inspection, the shell code in rc looked fine, but
when tested, sure enough, the semantics of the shell were not what I
thought they were. Very unexpected to me, and obviously to the
hundreds of people who must have read those lines without seeing that
they were broken.

Move savecore up before named to get a dump on low memory systems and
move 'mount -a' between yp and nfs daemons.

Take 2 to get the mount order "right", taking care of possible LKM
loads before netstart this time.
LKMs can be loaded at 3 strategic places in /etc/rc - before netstart,
before "mount -a" and after all. "before netstart" works only if
the /usr filesystem is local (or on NFS, if root is too).
An additional field ("when") in lkm.conf controls when the LKM is
loaded.

Back out the last two commits for now till we find a better way to fix
the mount problems.

rc.lkm needs /usr/bin/ld, so try to mount it with -t nonfs before rc.lkm.

Mount only "/" at the beginning (instead of all non-NFS filesystems).
This makes that the ordering of fstab is used for mounts which is
a) more intuitive and b) needed if {mfs,union,...} on top of NFS
is wanted.

Before disk checks are performed, do a "swapctl -A -t blk" so that
block-type devices are available during disk checks, which may consume
large amounts of memory if large file systems are present.  Once "critical"
file systems (e.g. /usr and /var) are mounted, perform a "swapctl -A -t noblk"
to enable swapping on any swap files that may be listed in /etc/fstab.

use swapctl instead of swapon

Make xntpd put its pid in /var/run/xntpd.pid.

Spawn a separate shell for netstart, making sure to check the return value
and exit with error if appropriate.

Add support for an ntpdate_hosts variable.
settings:
	NO:	don't run ntpdate on boot
	DEFAULT:extract a list of hosts from /etc/ntp.conf
	otherwise, a list of hosts to poll with ntpdate

make motd update conditional on an rc.conf variable (update_motd), and
move it before interactive logins are possible (just in case).

move last items from rc.local to rc; rc.local is now actually for local stuff.

Added mopd.

When displaying the names for statd and lockd, include the `rpc.'.

Add .  GC , which isn't used.

Do the DEFAULT thing with mountd_flags.  Don't output an extra
newline for statd and lockd.

Minor cleanup.

If foo_flags is DEFAULT, run the program with a default set of flags.  Most
users will use this, and therefore will never have to worry about flags
changing between releases.  It's also a bit less error-prone.

Add configuration flags for lockd/statd, and start them if an
NFS server or client is configured and the flags are not NO.

cleanup Lite-1 merge

import 4.4BSD-Lite

oops;  use $savecore_flags.  from kilbi@rad.rwth-aachen.de

move start-time configuration from netstart, rc, rc.local into
the new /etc/rc.conf.  this also includes the following changes:

	netstart:  ifaliases now includes a netmask, to ensure
		   aliases work.

	rc/rc.local:  make _everything_ based on configuration
	in rc.conf.  also move things around such that DNS and
	YP are started early enough that other daemons needing
	their services also exist.  move ldconfig from rc.local
	here, and also make it use /etc/ld.so.conf if it exists.
	run /etc/rc.lkm if it exists, a boot time.

add inetd_flags, mostly from <jbernard@tater.Mines.EDU> in pr#2987

add dhcpd startup;  disabled by default

move rpc daemon startup to after network startup so that if the machine is a DNS server and wants to resolve via itself, one doesn't have to wait for all the DNS queries to time out as each rpc daemon starts. This was particularly ugly with mountd.

Quote the YP daemon flags variables.

Change how the YP daemons are started.  Previously, ypbind was always
started if the directory /var/yp exists.  Now, ypbind, ypserv, and
rpc.yppasswdd are started like other daemons; there are flags variables
for these programs.  To disable them, set the variables to "NO", otherwise,
their contents are passed as flags.

add note about why we use /usr/sbin/sendmail.

now that sendmail supports SIGHUP, call sendmail with full path.

Add $mrouted_flags to netstart and use it in rc to start mrouted in a manner
similar (well, identical, really) to how $routed_flags is used to start
routed.

New-style RCS ids.

fix typo in comment

Conditionally start rbootd like rarpd and rpc.bootparamd, and disable
it by default.

Fix typos in two comments.

Configure ccd devices before the filesystem checks, as suggested by
Thomas Eberhardt in PR #1408.

Actually pass all the arguments to amd that we meant to.  Reported by
Dave Leonard (D.Leonard@pleiades.newcastle.edu.au).

Create an empty mountdtab after removing it, to avoid getting a warning
from mountd on startup.  Change $name_server to $named_flags.

always check quotas (there is no reason not to try)

quotas file(s) changed name in 4.4-lite

if install will do the trick, why not use it\?

Update nfsd and nfsiod invocations.

Move building of link editor cache from rc to rc.local, like in SunOS,
because the end user may need to add local directories to be searched.
And rc.local is a much more appropriate place for local changes.

I emailed pk on this matter, and he said that the overhead of not
having linker hints for the daemons started by rc would not be
significant.

This change also closes bug #217.

Attempt to make sure /var is mounted before futzing with it.

Move code that cleans /var/run so that it is before the automount deamon
startup puts its pid file there (This addressess most of bug #217).

add ttyflags -a, early on in rc.

keep amd pid in /var/run/amd.pid

punctuation

kill uucp lock files.  from Mike Long <mike.long@analog.com>

slight change to the way AMD is invoked, to make it sensible for most uses

make booting comments a bit cleaner

automatically enable accounting and rotate logs

don't delete preserve files!

Do the right thing for nvi.

The cron daemon now lives in /usr/sbin

Add bootparamd gunk.

Add rarpd gunk.

The cron daemon has been renamed from crond to cron.  This name change
is from Vixie Cron 3.0, to make it like the 4.3BSD cron.

start amd if requested
do nfs mounting after mountd has been started (so two machines can cross-mount
nfs from each other, else they hang if booted simultaneously)

initialize rpc services in proper order

Use `.' to run rc.local, rather than starting another shell, so that variables
from netstart are passed on.

fix comment, add check

Put ldconfig later so it doesn't get wiped when reaping /var/run.

Build ld.so.hints on each boot.

more flags

quote sendmail_flags test, from Patrick Bridges <patrick@buffy.cs.msstate.edu>

(1) clean up lots (rename fooflags -> foo_flags, etc.)
(2) implement suggestion made by Garrett Wollman of "sendmail flags"
	to set flags passed to sendmail and/or turn it off
(3) document origins of variables

never rotate accounting logs, and turn on accounting the way done in net/2

Use `-e fastboot', not `-r fastboot'.

it's clear that mountd started from inetd doesn't work

mountd is now started by inetd.

Don't start rwalld - now started by inetd.

prune dirs rm'd when cleaning tmp

turn on accounting if /var/account exists

bsd->netbsd

new kernel is bsd

fixes for nfs_client, nfs_server division.

always run portmap

mountd's log of who has what filesystem mounted should not persist across
reboots.

rwalld should be run *after* the portmapper.

screwed it up again...  = vs ==.

screwed up comparison for rwalld and rwhod...

updated for addition of rwalld

got rid of damned highlight characters, and changed
method to turn on rwhod.

initial import of 386bsd-0.1 sources

Initial revision