Arch Linux

d.ALT

Member

Registered: 2019-05-10

Posts: 956

BTRFS: not using the toplevel subvolume (ID=5) as "/" (from Wiki)

Hey guys!

Please I need some clarifications about what Wiki is suggesting about "/" when mounting subvolumes using BTRFS.
I cannot fully understand what to do, I really appreciate your help.

Should I create a "parent" subvolume with path /main and mount / inside /main?
And then use other subvolumes, eg.: /home = /main/home, etc...?
But... Then... What should I declare inside /etc/fstab?

Here's my current configuration:

$ mount
/dev/sda3 on / type btrfs (rw,relatime,compress=zstd:3,ssd,space_cache=v2,autodefrag,subvolid=5,subvol=/)


$ cat /etc/fstab 
# Static information about the filesystems.
# See fstab(5) for details.

# <file system> <dir> <type> <options> <dump> <pass>
# /dev/sda3 LABEL=ARCH
UUID=3cdec25f-04e2-4bec-b602-8ae0059411de       /               btrfs           rw,relatime,autodefrag,compress=zstd:3,ssd,space_cache=v2,subvolid=5,subvol=/   0 0

# /dev/sda1 LABEL=ESP
UUID=4542-BF6D          /efi            vfat            rw,relatime,fmask=0022,dmask=0022,codepage=437,iocharset=ascii,shortname=mixed,utf8,errors=remount-ro   0 2

# /dev/sda2 LABEL=SWAP
UUID=85e7c388-fc29-46b1-a8aa-9b5354acc8ef       none            swap            defaults        0 0


$ sudo blkid
/dev/sda2: LABEL="SWAP" UUID="85e7c388-fc29-46b1-a8aa-9b5354acc8ef" TYPE="swap" PARTLABEL="SWAP" PARTUUID="628cf2cb-2850-48c5-8fe3-f88157efd40a"
/dev/sda3: LABEL="ARCH" UUID="3cdec25f-04e2-4bec-b602-8ae0059411de" UUID_SUB="fff3faf7-824f-4c71-9ffe-cd944e2b432f" BLOCK_SIZE="4096" TYPE="btrfs" PARTLABEL="ARCH" PARTUUID="5ed06b59-0420-4469-a24c-0ffa0d7d5ff7"
/dev/sda1: LABEL_FATBOOT="ESP" LABEL="ESP" UUID="4542-BF6D" BLOCK_SIZE="512" TYPE="vfat" PARTLABEL="ESP" PARTUUID="b7d43ba1-76e0-4cb9-83b9-f67c8d51a523"


$ sudo btrfs subvolume list --sort=path -t -a -p -u -q /
ID      gen     parent  top level       parent_uuid     uuid    path
--      ---     ------  ---------       -----------     ----    ----
257     727     5       5               -                                       b478f0f0-5442-8f4e-935c-73a6f3788549    var/lib/machines
256     727     5       5               -                                       de7c3ab7-980d-2543-9f11-3d35c372db06    var/lib/portables


$ sudo lsblk -f
NAME   FSTYPE FSVER LABEL UUID                                 FSAVAIL FSUSE% MOUNTPOINTS
sda                                                                           
├─sda1 vfat   FAT32 ESP   4542-BF6D                             260,2M     0% /efi
├─sda2 swap   1     SWAP  85e7c388-fc29-46b1-a8aa-9b5354acc8ef                [SWAP]
└─sda3 btrfs        ARCH  3cdec25f-04e2-4bec-b602-8ae0059411de   57,9G    45% /

Thanks

---

<49,17,III,I>    Fama di loro il mondo esser non lassa;
<50,17,III,I>    misericordia e giustizia li sdegna:
<51,17,III,I>    non ragioniam di lor, ma guarda e passa.

dawnofman

Member

Registered: 2019-07-26

Posts: 140

Re: BTRFS: not using the toplevel subvolume (ID=5) as "/" (from Wiki)

Hey guys!

Please I need some clarifications about what Wiki is suggesting about "/" when mounting subvolumes using BTRFS.
I cannot fully understand what to do, I really appreciate your help.

Should I create a "parent" subvolume with path /main and mount / inside /main?
And then use other subvolumes, eg.: /home = /main/home, etc...?
But... Then... What should I declare inside /etc/fstab?

Last month I re-installed and by the way started using BTRFS for the first time: lots of doubts at the time, indeed. Please, see my recent post: is it good practice to create subvolumes while installing to BTRFS ?

It seems there are two main approaches to BTRFS volume creation:

- creating whatever you want on the root mount-point (like we always did for ext4 etc); eg: (for a data drive) /data (EXT4) ... period.
- creating a skeleton root (default mount) with only any other needed mount-points for any sub-volumes we want.

Either approach has its pros and cons, but, the second one, was advised/recommended/whatever on many sites while I did my research on the matter; better content separation will ease future snapshot management (if ever needed) etc etc.

Implementing the second approach took me a while, needed to piece the bits together and many sites/pages were outdated, were short on details, and/or lacked something very important: to make the parent root volume the default one after creating all the other sub-volumes, otherwise, no successful boot at all.

In short for any given drive:

parent (root) sub-volume (default)
sub-volume # 1
sub-volume # 2
sub-volume # 3
...

From my fstab (check relevant volume ...86a935552eb0); eg:

PARTUUID=eca82c38-b61e-4800-a15b-3eadc552d8b3 /boot vfat defaults,discard,utf8,dmask=0022,fmask=0022,umask=0022,noatime,rw 0 2

PARTUUID=c195aefd-3769-4fc7-8fec-86a935552eb0 / btrfs defaults,ssd,space_cache=v2,compress=zstd,commit=30,noatime,subvolid=256,subvol=/,rw 0 0
PARTUUID=c195aefd-3769-4fc7-8fec-86a935552eb0 /var btrfs defaults,ssd,space_cache=v2,compress=zstd,commit=30,noatime,subvolid=257,subvol=/sub-volume-data,rw 0 0
PARTUUID=c195aefd-3769-4fc7-8fec-86a935552eb0 /tmp btrfs defaults,ssd,space_cache=v2,compress=zstd,commit=30,noatime,subvolid=258,subvol=/sub-volume-temp,rw 0 0
PARTUUID=c195aefd-3769-4fc7-8fec-86a935552eb0 /root btrfs defaults,ssd,space_cache=v2,compress=zstd,commit=30,noatime,subvolid=259,subvol=/sub-volume-homeroot,rw 0 0
PARTUUID=c195aefd-3769-4fc7-8fec-86a935552eb0 /home btrfs defaults,ssd,space_cache=v2,compress=zstd,commit=30,noatime,subvolid=260,subvol=/sub-volume-home,rw 0 0

PARTUUID=e588a3c0-7664-41c3-b8f1-44ecafb76051 /data ext4 defaults,noatime,rw 0 2

My system drive:

- has a sub-volume ID=256 which is the root one for /
- has a sub-volume ID=257 which is labeled sub-volume-data for /var
- has a sub-volume ID=258 which is labeled sub-volume-temp for /tmp
- has a sub-volume ID=259 which is labeled sub-volume-homeroot for /root
- has a sub-volume ID=260 which is labeled sub-volume-home for /home

Following is an excerpt of my installation script that I think could prove helpful for your understanding on the subject:

### assuming target system drive is already manually-formatted; eg: for HGST SSD 100 GB:

### eg: sgdisk --new='1:4096:+1G ' --typecode='1:EF00' --change-name='1:system-EFI' '/dev/sd?'; ###  1 GB system-EFI for FAT32
### eg: sgdisk --new='2:    :+90G' --typecode='2:8304' --change-name='2:system-NIX' '/dev/sd?'; ### 90 GB system-NIX for BTRFS including sub‐volumes for var, tmp, root, home
### eg: sgdisk --print                                                              '/dev/sd?';

### eg: mkfs.fat           -F '32'      -n 'system-EFI' '/dev/sd?1'; ### FAT32 … do not use any outdated file-system: FAT → VFAT → FAT32 (this one)
### eg: mkfs.btrfs --csum='crc32c' --label='system-NIX' '/dev/sd?2'; ### BTRFS

### ∙ notes for BTRFS configuration: [/etc/mkinitcpio.conf] with modules='(crc32c)' to enable faster compression times
### ∙ notes for BTRFS configuration [/etc/fstab] with compress=zstd … when automatically mounted
### ∙ notes for BTRFS mount -o compress='zstd' /dev/sd?# /mnt/whatever/; … when manually mounted

typeset strBTRFSoptions='defaults,ssd,space_cache=v2,compress=zstd,commit=30,noatime,rw';    ### my default options to mount BTRFS file-systems
typeset strFAT32options='defaults,discard,utf8,dmask=0022,fmask=0022,umask=0022,noatime,rw'; ### my default options to mount FAT32 file-systems

### CAUTION … with the following IDs !

### volume IDs as drive WWN-based IDs … FSTAB using these may be possible (see man mount) but udevd will be required in HOOKS in [/etc/mkinitcpio.conf] … AND, there's no guarantee the EFI entries can also handle these:

### volume ID from [/dev/disk/by-id/] wwn-0x 5000cca0132a327c -part1 → ../../sdd1 … on system drive
### volume ID from [/dev/disk/by-id/] wwn-0x 5000cca0132a327c -part2 → ../../sdd2 … on system drive

### volume ID from [/dev/disk/by-id/] wwn-0x 55cd2e404c63af53 -part1 → ../../sdc1 … on data-buffer-2 drive … currently mounted as data

### volume IDs as UUIDs … will be using this set instead:

### volume ID from [/dev/disk/by-partuuid/] c3c15aa8-9577-4103-b1e2-e4f554d7fd15 → /dev/sdd1/ … on system drive
### volume ID from [/dev/disk/by-partuuid/] f47191f1-cd5c-4a21-8c8c-39c40412e77a → /dev/sdd2/ … on system drive

### volume ID from [/dev/disk/by-partuuid/] e588a3c0-7664-41c3-b8f1-44ecafb76051 → /dev/sdc1/ … on data-buffer-2 drive … currently mounted as data

typeset strIDPATHdrives='/dev/disk/by-id/wwn-0x';  ### ie: will be identifying drives by WWN
typeset strIDPATHvolumes='/dev/disk/by-partuuid/'; ### ie: will be identifying volumes by UUID
typeset strIDPATHcurrent='./configure-boot-storage-ID-current-';

typeset strDriveSystemWWN='5000cca0132a327c';      ### ie: HGST  SSD 100 GB  SAS
typeset strDriveDataBuffer2WWN='55cd2e404c63af53'; ### ie: Intel SSD 160 GB SATA

typeset strVolumeSystemEFIUUID='eca82c38-b61e-4800-a15b-3eadc552d8b3';   ### ie: should be [/dev/sd?1/]
typeset strVolumeSystemNIXUUID='c195aefd-3769-4fc7-8fec-86a935552eb0';   ### ie: should be [/dev/sd?2/]
typeset strVolumeDataBuffer2UUID='e588a3c0-7664-41c3-b8f1-44ecafb76051'; ### ie: should be [/dev/sd?1/]

typeset strDriveSystem="${strIDPATHdrives}${strDriveSystemWWN}";
typeset strDriveDataBuffer2="${strIDPATHdrives}${strDriveDataBuffer2WWN}";

typeset strVolumeSystemEFI="${strIDPATHvolumes}${strVolumeSystemEFIUUID}";
typeset strVolumeSystemNIX="${strIDPATHvolumes}${strVolumeSystemNIXUUID}";
typeset strVolumeSystemNIXID='sub-volume-';
typeset strVolumeSystemNIXIDroot="${strVolumeSystemNIXID}root";
typeset strVolumeSystemNIXIDdata="${strVolumeSystemNIXID}data";
typeset strVolumeSystemNIXIDtemp="${strVolumeSystemNIXID}temp";
typeset strVolumeSystemNIXIDhome="${strVolumeSystemNIXID}home";
typeset strVolumeSystemNIXIDhomeroot="${strVolumeSystemNIXID}homeroot";
typeset strVolumeDataBuffer2="${strIDPATHvolumes}${strVolumeDataBuffer2UUID}";

typeset strSystemTMP='/mnt/system/';

if /usr/bin/test 1 -eq 0; then ### safe-check to avoid running this accidentally while in production

   ### CAUTION … DO NOT ALLOW this to run before checking the above sgdisk output

   if ! /usr/bin/test -d "${strSystemTMP}"; then

      /usr/bin/mkdir "${strSystemTMP}"; ### ie: on the liveIMG file-system

   fi;

   if /usr/bin/test -d "${strSystemTMP}"; then ### we are on the live image environment: it's a go:

      if /usr/bin/test ! '['$(/usr/bin/findmnt --noheadings --output='TARGET' --mountpoint="${strSystemTMP}")'/]' = "${strSystemTMP}"; then

         /usr/bin/printf "mounting [${strSystemTMP}] …\n";

         /usr/bin/mount -t 'btrfs' --options="${strBTRFSoptions}" "${strVolumeSystemNIX}" "${strSystemTMP}"; ### ie: on a volatile mount-point on the live image file-system

      fi;

      if /usr/bin/test '['$(/usr/bin/findmnt --noheadings --output='TARGET' --mountpoint="${strSystemTMP}")'/]' = "[${strSystemTMP}]" \
         && /usr/bin/test -z $(/usr/bin/find "${strSystemTMP}" -mindepth '1' -maxdepth '1' -print -quit) \
         ; then ### proceed if there's no leftovers (possibly from a previous attempt):

         /usr/bin/printf "creating sub‐volumes on [${strSystemTMP}] …\n";

         ### the rationale here is to create sub‐volumes for anything holding variable user data; ie: RW data
         ### this could prove useful for system backups/snapshots as advised on some forums … although the pros and the cons are not very clear to me right now

         /usr/bin/btrfs subvolume create "${strSystemTMP}${strVolumeSystemNIXIDroot}"; ### this will be the root file‐system holding all other sub‐volumes
         /usr/bin/btrfs subvolume create "${strSystemTMP}${strVolumeSystemNIXIDdata}";
         /usr/bin/btrfs subvolume create "${strSystemTMP}${strVolumeSystemNIXIDtemp}";
         /usr/bin/btrfs subvolume create "${strSystemTMP}${strVolumeSystemNIXIDhomeroot}";
         /usr/bin/btrfs subvolume create "${strSystemTMP}${strVolumeSystemNIXIDhome}";

         ### from BTRFS btrfs-subvolume man page: a freshly‐created file‐system is also a sub‐volume that cannot be removed or replaced by another sub‐volume
         ### from BTRFS btrfs-subvolume man page: this is also the sub‐volume that will be mounted by default (unless changed with set-default)

         /usr/bin/btrfs subvolume list -p "${strSystemTMP}";

         /usr/bin/printf "current default sub‐volume is: "; /usr/bin/btrfs subvolume get-default "${strSystemTMP}";

         typeset strVolumeSystemNIXIDrootNO=$(/usr/bin/btrfs subvolume list "${strSystemTMP}" \
            | /usr/bin/grep --regexp="${strVolumeSystemNIXIDroot}" \
            | /usr/bin/grep --only-matching --perl-regexp '^ID [0-9]+' \
            | /usr/bin/sed 's/ID //'\
            ); ### needing BTRFS sub‐volume number ID

         /usr/bin/printf "setting default sub‐volume to [${strVolumeSystemNIXIDrootNO}]=[${strVolumeSystemNIXIDroot}] …\n";

         /usr/bin/btrfs subvolume set-default "${strVolumeSystemNIXIDrootNO}" "${strSystemTMP}"; ### needs an unmount/mount to take action

         /usr/bin/btrfs subvolume list -p "${strSystemTMP}";

         /usr/bin/printf "current default sub‐volume is: "; /usr/bin/btrfs subvolume get-default "${strSystemTMP}";

         unset -v strVolumeSystemNIXIDrootNO;

         ### so we currently have; eg: /dev/sd?2 on /mnt/system type btrfs (rw,noatime,compress=zstd:3,ssd,space_cache=v2,commit=120,subvolid=5,subvol=/) … as following:

         ### /mnt/system/sub-volume-root
         ### /mnt/system/sub-volume-data
         ### /mnt/system/sub-volume-temp
         ### /mnt/system/sub-volume-homeroot
         ### /mnt/system/sub-volume-home

         ### there are no more directories and/or files on [/mnt/system/] at this point; only the sub‐volumes shown as normal directories

         ### we won't need this default sub‐volume anymore; we'll be using the new default one:

         /usr/bin/printf "re-mounting [${strSystemTMP}] to get the new default sub‐volume up‐and‐running …";

         /usr/bin/sync && /usr/bin/umount "${strSystemTMP}";
         /usr/bin/mount -t 'btrfs' --options="${strBTRFSoptions}" "${strVolumeSystemNIX}" "${strSystemTMP}"; ### ie: on a volatile mount-point on the live image file-system

         /usr/bin/printf "creating mount-points on [${strSystemTMP}] …\n";

         /usr/bin/mkdir --mode='700' \
            "${strSystemTMP}boot/" \
            "${strSystemTMP}data/" \
            "${strSystemTMP}home/" \
            "${strSystemTMP}root/" \
            "${strSystemTMP}tmp/" \
            "${strSystemTMP}tmp-workspace/" \
            "${strSystemTMP}var/" \
            "${strSystemTMP}usb/" \
            ;

         /usr/bin/printf "mounting [${strSystemTMP}boot/] …\n";            /usr/bin/mount -t  'vfat' --options="${strFAT32options}"                                        "${strVolumeSystemEFI}" "${strSystemTMP}boot/";
         /usr/bin/printf "mounting [${strVolumeSystemNIXIDdata}] …\n";     /usr/bin/mount -t 'btrfs' --options="${strBTRFSoptions},subvol=${strVolumeSystemNIXIDdata}"     "${strVolumeSystemNIX}" "${strSystemTMP}var/";
         /usr/bin/printf "mounting [${strVolumeSystemNIXIDtemp}] …\n";     /usr/bin/mount -t 'btrfs' --options="${strBTRFSoptions},subvol=${strVolumeSystemNIXIDtemp}"     "${strVolumeSystemNIX}" "${strSystemTMP}tmp/";
         /usr/bin/printf "mounting [${strVolumeSystemNIXIDhomeroot}] …\n"; /usr/bin/mount -t 'btrfs' --options="${strBTRFSoptions},subvol=${strVolumeSystemNIXIDhomeroot}" "${strVolumeSystemNIX}" "${strSystemTMP}root/";
         /usr/bin/printf "mounting [${strVolumeSystemNIXIDhome}] …\n";     /usr/bin/mount -t 'btrfs' --options="${strBTRFSoptions},subvol=${strVolumeSystemNIXIDhome}"     "${strVolumeSystemNIX}" "${strSystemTMP}home/";

         if    /usr/bin/test '['$(/usr/bin/findmnt --noheadings --output='TARGET' --mountpoint="${strSystemTMP}boot")']' = "[${strSystemTMP}boot]" \
            && /usr/bin/test '['$(/usr/bin/findmnt --noheadings --output='TARGET' --mountpoint="${strSystemTMP}var" )']' = "[${strSystemTMP}var]" \
            && /usr/bin/test '['$(/usr/bin/findmnt --noheadings --output='TARGET' --mountpoint="${strSystemTMP}tmp" )']' = "[${strSystemTMP}tmp]" \
            && /usr/bin/test '['$(/usr/bin/findmnt --noheadings --output='TARGET' --mountpoint="${strSystemTMP}root")']' = "[${strSystemTMP}root]" \
            && /usr/bin/test '['$(/usr/bin/findmnt --noheadings --output='TARGET' --mountpoint="${strSystemTMP}home")']' = "[${strSystemTMP}home]" \
            ; then

            /usr/bin/printf "everything already‐mounted on [${strSystemTMP}]: changing permissions accordingly …\n";

            /usr/bin/chmod '700' "${strSystemTMP}home/";
            /usr/bin/chmod '700' "${strSystemTMP}root/";

            /usr/bin/printf "everything already‐mounted on [${strSystemTMP}]: updating drive and volume IDs on FSTAB and EFI entries …\n";

            /usr/bin/genfstab -U "${strSystemTMP}" > '/etc/fstab';

            ...

         fi;

      fi;

   fi;

fi;

edit: fixed a wrong permission in the script that I forgot to change after my last installation:

from:

/usr/bin/chmod '700' "${strSystemTMP}home/";

to:

/usr/bin/chmod '755' "${strSystemTMP}home/";

Last edited by dawnofman (2022-06-01 05:23:34)

ratcheer

Member

Registered: 2011-10-09

Posts: 924

Re: BTRFS: not using the toplevel subvolume (ID=5) as "/" (from Wiki)

I always put my root  (/) on a subvolume of the btrfs root subvolume (subvolid=5). I also make my /home a subvolume of the btrfs root subvolume. / and /home are, of course, mounted during the course of normal operations.

The only time I mount the btrfs root subvolume is to make snapshots of / and /home. After my snapshots have been migtared to external storage, I umount the btrfs root subvolume.