[sf-lug] Chromebooks' partitioning optimizing

[Michael Paoli]

Well, there are additional means besides symbolic link(s).

Most notably:
With Linux, one can not only mount a filesystem atop a
directory (the mount point), and whether or not it's empty
(nominally it would be, if it's not, that's commonly referred to as
an overmount), but Linux also allows the same filesystem to be
mounted multiple places at the same time!
I'll give brief example that happens to cover both:
# mount | grep -F -e ' /var ' -e ' /mnt '
/dev/mapper/tigger-var on /var type ext3 (rw,nosuid,nodev,relatime)
# mount -t ext3 -o rw,nosuid,nodev,relatime /dev/mapper/tigger-var /mnt
# mount | grep -F -e ' /var ' -e ' /mnt '
/dev/mapper/tigger-var on /var type ext3 (rw,nosuid,nodev,relatime)
/dev/mapper/tigger-var on /mnt type ext3 (rw,nosuid,nodev,relatime)
# ls -1d /{var,mnt}/cache/apt/archives/.fi*
ls: cannot access '/var/cache/apt/archives/.fi*': No such file or directory
/mnt/cache/apt/archives/.find_deb_in_snapshot.debian.org.out
# mount | grep -F -e ' /var/cache/apt/archives ' | sort -u
/dev/mapper/tigger-apt.archives on /var/cache/apt/archives type ext3
(rw,nosuid,nodev,relatime)
#
So, with that additional mount command, I then also mounted that
/var filesystem also on /mnt directory (mountpoint).
Also under that filesystem, relative to its root (as seen under its
mount point when mounted), there is file:
cache/apt/archives/.find_deb_in_snapshot.debian.org.out
however, can't access that at present under the mount of that
filesystem on /var, because overmount - another filesystem is
mounted atop
(/var/)cache/apt/archives
One typically wouldn't mount atop a non-empty directory
but in this particular case I'm doing it quite intentionally
(I want to reduce the probability of the
.find_deb_in_snapshot.debian.org.out
file getting accidentally deleted), and with that overmount,
can't access that file without either unmounting the overmount,
or mounting that filesystem elsewhere (including the possibility of
concurrently).
Note mounting same filesystem elsewhere at same time,
one may be restricted to using same mount options for each such
additional mount.

Another means is bind mount (other *nix flavors may have same or highly
similar, but may use other terminology and syntax, e.g. Solaris calls it
loopback mount, but that's not to be confused with Linux's loopback
devices, which are another thing entirely).
Bind mount allows one to mount a directory atop another directory.
Continuing from our example above:
# cd "$(mktemp -d)"
# mkdir mnt var
# (for n in mnt var; do mount -o bind /"$n" "$n"; done)
# ls -1d {var,mnt}/cache/apt/archives{,/.fi*}; pwd -P
mnt/cache/apt/archives
mnt/cache/apt/archives/.find_deb_in_snapshot.debian.org.out
var/cache/apt/archives
var/cache/apt/archives/.find_deb_in_snapshot.debian.org.out
/tmp/tmp.XuNMlkIZ5v
#
Note that the bind mount doesn't carry along the overmounts - or
any other mounts from the source directory to under the target.
E.g.:
# (for p in {,/}var/local{,/pub}; do echo "$(df -h "$p" | awk
'{if(NR==2) print $1,$6;}') #$p"; done)
/dev/mapper/tigger-var /tmp/tmp.XuNMlkIZ5v/var #var/local
/dev/mapper/tigger-var /tmp/tmp.XuNMlkIZ5v/var #var/local/pub
/dev/mapper/tigger-var /var #/var/local
/dev/mapper/tigger-pub /var/local/pub #/var/local/pub
# umount var && umount mnt && rmdir * && (d="$(pwd -P)" && cd / &&
rmdir "$d") && cd && umount /mnt
#
Linux also allows bind mounting a non-directory atop a non-directory, e.g.:
# (t="$(mktemp)" && mount -o bind /etc/motd "$t" && { echo cmp
/etc/motd "$t" && ls -ldi /etc/motd "$t" && cmp /etc/motd "$t" && echo
IDENTICAL; umount "$t" && rm "$t"; })
cmp /etc/motd /tmp/tmp.1cyXUErh2m
88332 -rw-r--r-- 1 root root 286 Aug 25 17:48 /etc/motd
88332 -rw-r--r-- 1 root root 286 Aug 25 17:48 /tmp/tmp.1cyXUErh2m
IDENTICAL
#
And one can specify bind mounts in /etc/fstab, e.g. (here Ss just shrinks
the whitespace):
$ grep -F -e bind /etc/fstab | Ss
/var/cache/apt/archives/archives /var/cache/apt/archives none bind 0 0
/dev/null /var/lib/named/dev/null none bind 0 0
/dev/random /var/lib/named/dev/random none bind 0 0
/run/named /var/lib/named/run/named none bind 0 0
/usr/share/dns /var/lib/named/usr/share/dns none bind 0 0
/var/cache/bind /var/lib/named/var/cache/bind none bind 0 0
/var/lib/bind /var/lib/named/var/lib/bind none bind 0 0
/run/systemd/notify /var/lib/named/run/systemd/notify none bind,nofail 0 0
/home/r/root/log/log /var/local/log/log none bind 0 0
/var/local/boinc/boinc/boinc-client /var/lib/boinc-client none bind 0 0
$
https://www.mpaoli.net/~michael/bin/Ss

And ln(1) syntax is similar-ish to cp(1).
Notably the last non-option argument is what's created
or possibly replaced, but that if that last non-option argument is
a directory, then link(s) are generally created within that directory.
ln [OPTION]... [-T] TARGET LINK_NAME
ln [OPTION]... TARGET... DIRECTORY
E.g.:
$ cd "$(mktemp -d)"
$ (for f in f0 f1; do echo "$f" > "$f"; done); grep . f?
f0:f0
f1:f1
$ mkdir d
$ ln -s f0 L0; ln -s f1 L1; ln -s f? d
$ (for l in $(find * -type l -print); do echo -n "$l -> "; readlink "$l"; done)
L0 -> f0
L1 -> f1
d/f1 -> f1
d/f0 -> f0
$ mkdir D && ln -s D d && { echo -n 'd/D -> ' && readlink d/D; }
d/D -> D
$
But note many of those links aren't very useful, as they loop:
$ cat d/*
cat: d/D: Too many levels of symbolic links
cat: d/f0: Too many levels of symbolic links
cat: d/f1: Too many levels of symbolic links
$
More (potentially) useful, e.g.:
$ ln -s /tmp . && ln -s /nope . && ln -s /tmp d && (for l in tmp nope
d/tmp; do echo -n "$l -> $(readlink "$l") " && (cd "$l" && { echo "$l:
"$(pwd; pwd -P); }); done)
tmp -> /tmp tmp: /tmp/tmp.9c0Du9Upqx/tmp /tmp
nope -> /nope -bash: cd: nope: No such file or directory
d/tmp -> /tmp d/tmp: /tmp/tmp.9c0Du9Upqx/d/tmp /tmp
$ rm -rf * && mkdir -p d/d/d/d/d/d/d/d/d && ln -s "$(pwd
-P)"/d/d/d/d/d/d/d/d/d D
$ (cd D && { pwd; pwd -P; })
/tmp/tmp.9c0Du9Upqx/D
/tmp/tmp.9c0Du9Upqx/d/d/d/d/d/d/d/d/d
$ (cd "$(pwd -P)"/D && { pwd; pwd -P; })
/tmp/tmp.9c0Du9Upqx/D
/tmp/tmp.9c0Du9Upqx/d/d/d/d/d/d/d/d/d
$ readlink D
/tmp/tmp.9c0Du9Upqx/d/d/d/d/d/d/d/d/d
$

On Sun, Feb 1, 2026 at 1:01 PM <aaronco36 at sdf.org> wrote:
> > --> After creating and partitioning /dev/mmcblk1p3 as /usr/local ....
> > then create /usr/local/opt and symlink /opt to /usr/local/opt
> > ( 'sudo cp -v /opt /usr/local/opt ; sudo ln -sf /opt /usr/local/opt' )
> The symlink 'ln' syntax from its man page (e.g.,
> https://www.man7.org/linux/man-pages/man1/ln.1.html ) is "ln [OPTION]...
> TARGET... DIRECTORY"
> Therefore, as /usr/local/opt is the [new but actual!] TARGET, 2nd half of
> above should be corrected to ...
> 'sudo ln -sf /usr/local/opt /opt'
> The directionality of the 'ln' command is different from the 'cp' command,
> and that very fact sometimes messes me up... am sorry! :-(

On Sun, Feb 1, 2026 at 11:46 AM <aaronco36 at sdf.org> wrote:
>
> A good way to deal w/ the Chromebooks' low HW specs is to use SD cards
> in their SD slots to increase their available internal diskspace.
> Already wrote to Bobbie abt this the last time.
> E.g. w/ 32 GB microSD card /dev/mmcblk1p
> 1) 4 GB swap-part
> 2) 4 GB /var
> 3) 4 GB /usr/local
> 4) Remaining space as Ext'd for MBR BIOS
> 5) 1 GB /tmp
> 6) 16 GB+ /home
>
> --> After creating and partitioning /dev/mmcblk1p3 as /usr/local ....
> then create /usr/local/opt and symlink /opt to /usr/local/opt
> ( 'sudo cp -v /opt /usr/local/opt ; sudo ln -sf /opt /usr/local/opt' )
>
> The Chromebooks' default internal 16.0 GB /dev/mmcblk0p as
> 1) ~300 MB EFI boot (efs)
> 2) 14 GB+ /<root>
> ( swap on (2) as 2 GB swap _file_ )