Date: Sun, 13 Mar 2011 00:08:30 +0100 From: Jilles Tjoelker <jilles@stack.nl> To: Doug Barton <dougb@FreeBSD.org> Cc: svn-src-head@freebsd.org, svn-src-all@freebsd.org, src-committers@freebsd.org Subject: Re: svn commit: r219578 - head/etc Message-ID: <20110312230830.GA269@stack.nl> In-Reply-To: <201103122113.p2CLD8LO030205@svn.freebsd.org> References: <201103122113.p2CLD8LO030205@svn.freebsd.org>
next in thread | previous in thread | raw e-mail | index | archive | help
On Sat, Mar 12, 2011 at 09:13:08PM +0000, Doug Barton wrote: > Author: dougb > Date: Sat Mar 12 21:13:08 2011 > New Revision: 219578 > URL: http://svn.freebsd.org/changeset/base/219578 > Log: > Use the allexport option in load_rc_config() in order to avoid having > to repeatedly read the conf files. Depending on what is enabled the > files are being read anywhere from 15, 30, or more times currently. > By loading the values in the environment this is reduced to 1, with > perhaps a couple more, again depending on what is enabled. I wonder if it is a good idea to pollute the environment of many daemons with this. Although sshd and cron clean it up, there is at least one daemon that passes the environment on. One of those is devd. While that saves four /etc/rc.conf reads on my minimalistic 9-current VM, it implies that devd must be restarted to pick up changes to /etc/rc.conf. That seems a POLA violation. With the old code, my VM reads rc.conf nine times at boot. One of them is the initial load, four of them are via devd and the other four are because rc.d scripts call each other via execve(2) instead of via the run_rc_script function. The latter four can be avoided but the other five seem unavoidable. I found this by adding the following line to /etc/rc.conf: echo "rc.conf is being read at $(date) by $(ps -p $$ -o pid= -o ppid= -o args=)" >/dev/console (The new code reads rc.conf once.) > The speed-up for boot and shutdown is negligible when rc.conf is > on local disk, noticable when accessing files over NFS, and dramatic > when pulling rc.conf values from a database. > This change also includes a minor optimization to the conditional > for $_rc_conf_loaded. I don't know if we ever supported 'set -u' in rc.d scripts, but this definitely breaks it. To fix this, test "${_rc_conf_loaded-}" instead of "$_rc_conf_loaded" > Modified: > head/etc/rc.subr > > Modified: head/etc/rc.subr > ============================================================================== > --- head/etc/rc.subr Sat Mar 12 20:36:52 2011 (r219577) > +++ head/etc/rc.subr Sat Mar 12 21:13:08 2011 (r219578) > @@ -998,9 +998,8 @@ load_rc_config() > err 3 'USAGE: load_rc_config name' > fi > > - if ${_rc_conf_loaded:-false}; then > - : > - else > + if [ -z "$_rc_conf_loaded" ]; then > + set -o allexport > if [ -r /etc/defaults/rc.conf ]; then > debug "Sourcing /etc/defaults/rc.conf" > . /etc/defaults/rc.conf > @@ -1010,6 +1009,7 @@ load_rc_config() > . /etc/rc.conf > fi > _rc_conf_loaded=true > + set +o allexport > fi > if [ -f /etc/rc.conf.d/"$_name" ]; then > debug "Sourcing /etc/rc.conf.d/${_name}" -- Jilles Tjoelker
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20110312230830.GA269>