svn commit: r290197 - in head: etc/defaults etc/rc.d sys/kern

[Edward Tomasz Napierala]

On 1221T1807, Jilles Tjoelker wrote:
> On Fri, Oct 30, 2015 at 03:52:10PM +0000, Edward Tomasz Napierala wrote:
> > Author: trasz
> > Date: Fri Oct 30 15:52:10 2015
> > New Revision: 290197
> > URL: https://svnweb.freebsd.org/changeset/base/290197
> 
> > Log:
> >   After r290196, the kernel won't wait for stuff like gmirror nodes
> >   if they are not required for mounting rootfs.  However, it's possible
> >   that some setups try to mount them in mountcritlocal (ie from fstab).
> 
> >   Export the list of current root mount holds using a new sysctl,
> >   vfs.root_mount_hold, and make mountcritlocal retry if "mount -a" fails
> >   and the list is not empty.
> 
> >   MFC after:	1 month
> >   Sponsored by:	The FreeBSD Foundation
> >   Differential Revision:	https://reviews.freebsd.org/D3709
> 
> I like the faster startup, but the rc.d script clearly will not wait as
> intended. See below.
> 
> > Modified:
> >   head/etc/defaults/rc.conf
> >   head/etc/rc.d/mountcritlocal
> >   head/sys/kern/vfs_mountroot.c
> 
> > [snip]
> > Modified: head/etc/rc.d/mountcritlocal
> > ==============================================================================
> > --- head/etc/rc.d/mountcritlocal	Fri Oct 30 15:35:04 2015	(r290196)
> > +++ head/etc/rc.d/mountcritlocal	Fri Oct 30 15:52:10 2015	(r290197)
> > @@ -15,7 +15,7 @@ stop_cmd=sync
> >  
> >  mountcritlocal_start()
> >  {
> > -	local err
> > +	local err holders waited
> >  
> >  	# Set up the list of network filesystem types for which mounting
> >  	# should be delayed until after network initialization.
> > @@ -35,8 +35,42 @@ mountcritlocal_start()
> >  		mount_excludes="${mount_excludes}${fstype},"
> >  	done
> >  	mount_excludes=${mount_excludes%,}
> > +
> > +	# Originally, root mount hold had to be released before mounting the root
> > +	# filesystem.  This delayed the boot, so it was changed to only wait if
> > +	# the root device isn't readily available.  This can result in this script
> > +	# executing before all the devices - such as graid(8) - are available.
> > +	# Thus, should the mount fail, we will wait for the root mount hold release
> > +	# and retry.
> 
> These lines are a bit too long.
> 
> >  	mount -a -t ${mount_excludes}
> >  	err=$?
> > +	if [ $? -ne 0 ]; then
> 
> The assignment will set $? to 0, so the new code will never be executed.
> "$err" should be tested instead of $?.

Thanks, fixed.  Previously, I've tested it by changing "-ne" to "-eq",
and didn't even realize I've missed a step.