A bad /etc/fstab entry drops the system to emergency mode or hangs 90 seconds
The systemd mechanics held against the installed systemd 261.2-1 man page (man -P cat systemd.mount, same text as https://man.archlinux.org/man/systemd.mount.5): fstab lines become mount units at boot and on daemon-reload, a block-backed mount gains Requires= and After= on its device unit, nofail turns the Requires= from local-fs.target into Wants= and drops the Before= ordering, and x-systemd.device-timeout= caps the device wait and is honoured only in fstab. systemctl show reports DefaultDeviceTimeoutUSec=1min 30s here, and /usr/lib/systemd/system/local-fs.target carries OnFailure=emergency.target, so the 90 second stall and the drop to emergency mode are confirmed on this machine. findmnt --verify is in findmnt(8) of util-linux 2.42.2, and mount -a is the correct dry run of the file. The root password prompt is reachable on Omarchy 4: the ISO configurator (omacom/omarchy-iso, quattro, configs/airootfs/root/configurator) writes root_enc_password with the same SHA-512 hash as the user's password into user_credentials.json, and orchestrator/archinstall_adapter.py's root_user() creates root from it, so root is not locked and the emergency prompt takes the login password. The wizard text says so too: "Used for user + root, and disk encryption when enabled". I could not confirm it against /etc/shadow here (root only), so it is from source, not this machine. What was wrong for Omarchy 4: the symptom names /boot/efi and the example fstab shows an ext4 root, while /etc/fstab here mounts the ESP at /boot (vfat, fmask=0077,dmask=0077, pass 2) and root is btrfs subvolumes (@, @home, @log, @pkg) on /dev/mapper/root, all pass 0. Omarchy's initramfs uses the busybox encrypt hook (HOOKS in /etc/mkinitcpio.conf.d/omarchy_hooks.conf) with cryptdevice=PARTUUID=...:root and rw on the kernel command line, so by the time emergency.target starts the mapper is open and / is mounted read-write, and lsblk -f and blkid show the mapper rather than the raw partition. The swap advice was the other miss: Omarchy has no swap partition, it hibernates to /swap/swapfile with resume=/dev/mapper/root resume_offset=... coming from /etc/limine-entry-tool.d/resume.conf, which is embedded in the UKI, so dropping resume= means removing that drop-in and running limine-mkinitcpio, not editing a GRUB line. omarchy-hibernation-remove deletes the swapfile, the fstab line and the resume hook but leaves resume.conf in place. On plain Arch with the mkinitcpio resume hook, hooks/resume calls resolve_device, which waits rootdelay (default 10 s per init_functions poll_device) and then continues, so the wait is 10 s rather than indefinite. The GRUB paths in the plain Arch branch are from https://wiki.archlinux.org/title/GRUB. The three EndeavourOS forum threads the record cites still resolve but were not re-read for this pass. The emergency shell itself was not exercised, and nothing was rebooted.
Never add nofail to the root filesystem line, and on Omarchy 4 never add it to the /home, /var/log or /var/cache/pacman/pkg subvolume lines either: they sit on the same device as root and Omarchy's update path expects them mounted. Editing fstab with a broken syntax (missing field, wrong pass number) can itself cause emergency mode: always run sudo findmnt --verify and sudo mount -a before rebooting. On Omarchy the ESP is /boot and holds the Limine UKI, so a wrong UUID on that line means the next kernel or firmware update writes the UKI into the root filesystem instead of the ESP and the machine boots the old kernel.
Symptom
Boot stalls with:
A start job is running for /dev/disk/by-uuid/1a2b-3c4d (1min 30s / no limit)
and then:
[FAILED] Failed to mount /boot.
[DEPEND] Dependency failed for Local File Systems.
You are in emergency mode. After logging in, type "journalctl -xb" to view
system logs, "systemctl reboot" to reboot ...
Give root password for maintenance (or press Control-D to continue):
The mount point named is whatever line failed: /boot on Omarchy 4, where the ESP is mounted there, /boot/efi on EndeavourOS and many hand-built Arch installs. Usually after swapping a disk, removing an external drive, reformatting the ESP, or turning off a swap partition. On Omarchy the prompt appears after Plymouth quits, so it can follow a few seconds of blank splash.
Cause
systemd generates a mount unit for every /etc/fstab line. If the device never appears, the generated .device unit waits (default 90 s) and then the mount fails. Because local-fs.target *requires* it, the whole boot fails into emergency mode. A reformatted ESP or a re-created swap partition has a new UUID, so the old fstab line can never be satisfied.
Fix
At the emergency prompt log in as root. On Omarchy 4 the root password is the password you chose at install, because the installer sets root and your user to the same one. If you cannot log in, boot a live USB and chroot instead. On an encrypted Omarchy install the LUKS container is already open (/dev/mapper/root) and / is mounted read-write by the time you reach this prompt. Compare fstab against reality:
lsblk -f
blkid
cat /etc/fstab
Fix the UUIDs, or delete lines for devices that no longer exist. Root filesystem must stay mandatory, but make optional mounts non-fatal.
**Plain Arch (ext4 root, ESP at /boot/efi or /boot):**
# <device> <dir> <type> <options> <dump> <pass>
UUID=1a2b-3c4d /boot vfat defaults,noatime 0 2
UUID=aaaa-bbbb-cccc / ext4 rw,relatime 0 1
UUID=dddd-eeee-ffff none swap defaults 0 0
# external / removable: never block boot
UUID=1111-2222 /mnt/data ext4 defaults,nofail,x-systemd.device-timeout=10s 0 2
**Omarchy 4 (btrfs subvolumes on /dev/mapper/root, ESP at /boot):** the installer writes one line per subvolume, all with the same UUID and pass 0, and the ESP with pass 2. Keep those as they are and only add nofail to drives you bolted on later:
# /dev/mapper/root
UUID=<btrfs-uuid> / btrfs rw,relatime,compress=zstd:3,ssd,space_cache=v2,subvol=/@ 0 0
UUID=<btrfs-uuid> /home btrfs rw,relatime,compress=zstd:3,ssd,space_cache=v2,subvol=/@home 0 0
UUID=<btrfs-uuid> /var/cache/pacman/pkg btrfs rw,relatime,compress=zstd:3,ssd,space_cache=v2,subvol=/@pkg 0 0
UUID=<btrfs-uuid> /var/log btrfs rw,relatime,compress=zstd:3,ssd,space_cache=v2,subvol=/@log 0 0
# /dev/nvme0n1p1
UUID=<esp-uuid> /boot vfat rw,relatime,fmask=0077,dmask=0077,codepage=437,iocharset=ascii,shortname=mixed,utf8,errors=remount-ro 0 2
# Btrfs swapfile for system hibernation
/swap/swapfile none swap defaults,pri=0 0 0
# external / removable: never block boot
UUID=1111-2222 /mnt/data ext4 defaults,nofail,x-systemd.device-timeout=10s 0 2
nofail makes the mount *wanted* rather than *required* by local-fs.target and removes the ordering before it, so boot continues whether or not the device turns up. x-systemd.device-timeout= caps how long systemd waits for the device instead of the 90-second default. It only works in /etc/fstab, not in a unit file.
Apply and test without rebooting blind:
sudo findmnt --verify # parse check of /etc/fstab
sudo systemctl daemon-reload
sudo mount -a # must return with no errors
sudo systemctl reboot
daemon-reload regenerates the mount units from the edited file. mount -a mounts every line not already mounted and fails loudly on a bad one, but it does not exercise the device timeout, so a nofail line for a missing drive still passes.
If the failing entry is swap you removed, also drop resume= from the kernel command line so the initramfs stops looking for it:
- **Plain Arch:** remove
resume=from your bootloader's kernel line (for GRUB,GRUB_CMDLINE_LINUX_DEFAULTin/etc/default/grubthengrub-mkconfig -o /boot/grub/grub.cfg). With the mkinitcpioresumehook the wait is about 10 seconds (rootdelay), after which boot continues with an error in the log. - **Omarchy 4:** the swap is a swapfile, and
resume=/dev/mapper/root resume_offset=...lives in/etc/limine-entry-tool.d/resume.conf, which is baked into the UKI. Preferomarchy-hibernation-remove, which disables the swapfile, drops its fstab line and removes theresumehook, then delete the drop-in and rebuild the UKI:
omarchy-hibernation-remove
sudo rm /etc/limine-entry-tool.d/resume.conf
sudo limine-mkinitcpio
Verify
sudo mount -a produces no output, systemctl --failed is empty after reboot, and systemd-analyze blame | head no longer shows a ~90 s device job.
Sources
- https://man.archlinux.org/man/systemd.mount.5
- https://forum.endeavouros.com/t/emergency-mode-failed-to-mount-boot-efi-vfat-error-tried-base-reinstall-issue-persists/75747
- https://forum.endeavouros.com/t/a-start-job-is-running-for-dev-disk-uui-running-for-5min/80783
- https://forum.endeavouros.com/t/whenever-kernel-updates-efi-mount-fails-on-reboot/77575
- https://man.archlinux.org/man/sulogin.8
- https://man.archlinux.org/man/findmnt.8
- https://wiki.archlinux.org/title/Fstab
- https://wiki.archlinux.org/title/GRUB
- https://gitlab.archlinux.org/archlinux/mkinitcpio/mkinitcpio/-/raw/master/hooks/resume
- https://gitlab.archlinux.org/archlinux/mkinitcpio/mkinitcpio/-/raw/master/init_functions
- https://github.com/omacom/omarchy-iso/blob/quattro/configs/airootfs/root/configurator
- https://github.com/omacom/omarchy-iso/blob/quattro/configs/airootfs/usr/share/omarchy-iso/orchestrator/archinstall_adapter.py
- https://github.com/omacom/omarchy/blob/quattro/bin/omarchy-hibernation-remove