FreeBSD user home directory

[Mark Millard]

On 2016-Jul-20, at 11:59 PM, Mark Millard <markmi at dsl-only.net> wrote:

> On 2016-Jul-20, at 8:03 PM, Glen Barber <gjb at FreeBSD.org> wrote:
> 
>> On Wed, Jul 20, 2016 at 07:54:27PM -0700, Mark Millard wrote:
>>> Looking at my armv6 and amd64 11.0's (long in use, originally
>>> -CURRENT, now -STABLE, maintained via source updates):
>>> 
>>> amd64 and armv6 (rpi2) both have real /usr/home directories.
>>> 
>>> armv6 (and rpi2) has no /home path established at all, not even
>>> as a symbolic link to elsewhere.
>>> 
>>> amd64 has /home -> usr/home via a symbolic link.
>>> 
>>> (I do not have access to check my memory and will not for weeks
>>> but if I remember right my powerpc64 and powerpc 11.0's were like
>>> amd64 above. They dated back to somewhat before 2016-June-04 when
>>> last updated.)
>>> 
>>> If I remember right my old powerpc and powerpc 10.x-STABLE's and
>>> 10.x-RELEASES also agreed with amd64 above. (At the time I only was
>>> experimenting with powerpc64 and powerpc FreeBSD.)
>>> 
>>> In comparison today's -r303119 says:
>>> 
>>>> Log:
>>>> Create a /usr/home -> /home symlink for the arm images to
>>>> avoid /usr/home confusingly being created as a directory.
>>> 
>>> 
>>> May be which path is to directly be the actual directory by default
>>> has changed --since all of my contexts started long ago.
>>> 
>>> But what all my confirmable examples suggest is that /usr/home
>>> is normally the directory.
>>> 
>>> I did not manually control or create /usr/home for any of the
>>> contexts as far as I can remember. It was automatic as a side effect
>>> of some activity.
>>> 
>> 
>> Right, but as we do not provide binary upgrade paths for tier-2
>> architectures, nothing should be affected for source-based upgrades.
>> Especially in this case.
>> 
>>> If there is variability up to now or across architectures it might
>>> be appropriate to have an UPDATING entry to indicate the new uniform
>>> answer or whatever describes how things now are.
>>> 
>>> Are there alternative standard FreeBSD installation techniques
>>> that may be should all be made to match for such properties? (POLA
>>> for such defaults: lack of variability across [the major or official]
>>> techniques?)
>>> 
>> 
>> This is discussion that is not applicable for the commit to which you
>> reference.  It creates a symlink on an image that is "installed" by
>> writing a raw filesystem onto an SD card via dd(1).  This does not
>> affect source-based upgrades.
>> 
>> Glen
> 
> Just an FYI:
> 
> I think I found were my amd64 /usr/home came from: I happened to have
> experimented with zfs for that and /usr/libexec/bsdinstall/zfsboot uses
> the following in defining ZFSBOOT_DATASETS
> 
>        # Home directories separated so they are common to all BEs
>        /usr/home       # NB: /home is a symlink to /usr/home
> 
> (BE in "BEs" is "Boot Environment".) My only amd64 environment is the
> only context that I've experimented with zfs in.
> 
> It is true that this is not likely to have much in common with how my
> armv6, powerpc64, and powerpc environments got to be /usr/home style
> (all tier 2).

Another FYI: I think I've tracked down where my armv6 /usr/home came from:
I happened to experiment with crochet for my very first armv6 creation and
I've never had to start from scratch again for armv6 after that.

# grep -i home ~/crochet/option/User/setup.sh
HOME_DIR=/usr/home
    mkdir -p ${BOARD_FREEBSD_MOUNTPOINT}${HOME_DIR}/$1
    $PW -V ${BOARD_FREEBSD_MOUNTPOINT}/etc/ useradd -n $1 -s /bin/csh -g wheel -w yes -d ${HOME_DIR}/$1
    $CHOWN $UGID ${BOARD_FREEBSD_MOUNTPOINT}${HOME_DIR}/$1

So crochet uses /usr/home as the directly existing directory (or did at
that time anyway).

(I'm not saying that crochet is a standard technique. I had forgotten that
I had happened to decide to experiment with it just the one time when I
first created a armv6 context.)