linux-omarchy 7.2.5-3 black screen, emergency shell, dead TPM, touchscreen or Wi-Fi on some Intel machines (Intel IOMMU on by default)
The mechanism and the fix hold. On this workstation linux-omarchy 7.2.5-3 has CONFIG_INTEL_IOMMU_DEFAULT_ON=y in /proc/config.gz, linux 7.2.3.arch1-3 is still installed, and /usr/lib/limine/limine-common-functions loads /etc/default/limine last ('4. Load /etc/default/limine (highest priority)'). limine-mkinitcpio with no argument pipes 'rebuild' to limine-mkinitcpio-install, and limine-update is limine-install plus limine-mkinitcpio, so both steps rebuild the UKIs. The pkgs.omarchy.org stable database fetched on 2026-10-05 still lists linux-omarchy-7.2.5-3, and no quattro migration after 1789444024.sh touches Limine or the IOMMU. omarchy-pkgs#677 confirms the XPS 13 9343 PTT fault, the 90 s wait and the 7.2.7rc1 fix. #12097 has a commenter using exactly the e, append, F10 one-boot path and the same drop-in. The cause and symptom contained fabricated precision. The cause called the Apple AHCI SSD controller 'Toshiba 1179:010b' for every Mac, but #14160's MacBookPro11,5 has a Samsung AHCI controller and #12119's MacBookPro13,3 has a Samsung NVMe SSD. The Toshiba controller is only the MacBookAir6,1. It also named only Southern Islands amdgpu chips while the 13,3 carries a Polaris Radeon Pro 450. The cited #13030 (AX210 Wi-Fi lost on 2 of 7 cold boots, no control run) and the #12550 radeon-bound 11,5 with no faults were missing. The title said 'older Intel machines', which the RTX 4070 Ti desktop and the 2019 ASUS laptop do not fit. Cause, symptom and title were rewritten. The fix is unchanged. Earlier audit's findings kept: the danger text about Btrfs corruption on affected Macs stays. Not exercised: no boot, no rebuild.
On Intel Macs whose internal SSD is affected (MacBookPro11,x on #14160, MacBookAir6,1 on #12097), DMA writes to the SSD fault while the IOMMU is on, and #14160 shows Btrfs corruption errors climbing during such a boot. Do not keep booting linux-omarchy to experiment on these machines. Boot linux, or add intel_iommu=off, before doing anything else, and run sudo btrfs scrub start -B / afterwards. On a MacBook Air the stock linux kernel may have no working Wi-Fi driver, so plan for a USB network adapter or phone tethering if you need to download anything from it.
Symptom
After the Omarchy 4.0.4 update installed the linux-omarchy kernel, the machine no longer boots properly or loses a device, but choosing the plain linux entry in the Limine menu still works. Reported forms:
- MacBookPro13,3 (2016): black screen, last line around
amdgpu 0000:01:00.0: [drm] VCE enabled in VM mode. Withamdgpublacklisted it gets further and drops to an emergency shell because the NVMe SSD never appears:
ERROR: device '/dev/mapper/omarchy_root' not found. Skipping fsck.
mount: /new_root: special device /dev/mapper/omarchy_root does not exist
You are now being dropped into an emergency shell.
- MacBookPro11,5 (2015): the session starts, then the Radeon R9 M370X wedges with
ring gfx timeout, and on one machine Btrfs corruption counters climb. - MacBookAir6,1: the internal SSD fails to identify, the encrypted root never appears, and the kernel panics with
Attempted to kill init!. - Dell XPS 13 9343: every boot stalls 90 seconds and the TPM disappears:
systemd[1]: Timed out waiting for device /dev/tpmrm0.
- Surface Pro 4 touchscreen dead, one desktop RTX 4070 Ti whose GSP firmware fails to boot, and one ASUS laptop whose Intel AX210 Wi-Fi drops off the PCIe bus on some cold boots.
journalctl -k on the failing kernel shows lines like DMAR: [DMA Read NO_PASID] Request device [0000:00:16.7] fault addr ....
Cause
linux-omarchy 7.2.5-3 is built with CONFIG_INTEL_IOMMU_DEFAULT_ON=y. Arch's linux 7.2.3 leaves it unset. With VT-d DMA translation on by default, a device whose buffers the firmware DMAR table does not cover, or that needs a PCI DMA alias quirk the kernel lacks, faults on DMA.
The reported devices differ by machine. On the MacBookAir6,1 it is the Toshiba AHCI controller 1179:010b, which needs a quirk_dma_func1_alias entry that a reporter sent to the Linux PCI maintainers on 2026-09-29. On the MacBookPro11,5 it is the internal Samsung AHCI SSD controller and the Southern Islands R9 M370X under amdgpu. On the MacBookPro13,3 it is the NVMe SSD, the Apple SPI keyboard and the Polaris Radeon Pro 450. Elsewhere it is the Intel ME PTT function at 00:16.7 that backs a firmware TPM, the Surface IPTS touch controller at 00:16.4, and an Intel AX210 on one ASUS laptop, where 2 of 7 cold boots failed and no intel_iommu=off control was run.
On omacom/omarchy-pkgs#677 the reporter confirmed that a 7.2.7rc1 build with the option unset removed the faults and the 90 s wait. As of 2026-10-05 the stable repository still ships linux-omarchy 7.2.5-3.
Not every report in these threads is the IOMMU. The iMac18,3 reporter on #12119 retracted their data point after testing the variables separately, and on #12550 a MacBookPro11,5 with its dGPU bound to radeon and the panel on Intel showed no faults with the IOMMU on.
Fix
## Get booted now
At the Limine menu pick the linux entry. The migration deliberately leaves the stock Arch kernel installed. Or highlight linux-omarchy, press e, append intel_iommu=off to the kernel command line and boot with F10.
## Make it persistent (Omarchy 4)
The command line is embedded in the UKI, so add a limine-entry-tool drop-in and rebuild. Never edit /boot/limine.conf.
echo 'KERNEL_CMDLINE[default]+=" intel_iommu=off"' | sudo tee /etc/limine-entry-tool.d/intel-iommu-off.conf
sudo limine-mkinitcpio
[default] applies to both kernels. That is harmless for the stock linux kernel, whose IOMMU is already off by default.
## Or keep booting the stock kernel by default
/etc/default/limine loads last and wins over every drop-in. The migration wrote a BOOT_ORDER line there. Put linux first:
sudo sed -i 's/^BOOT_ORDER=.*/BOOT_ORDER="linux, linux-omarchy, linux-omarchy-*, *, *fallback, Snapshots"/' /etc/default/limine
sudo limine-update
## Undo once a fixed kernel ships
When pacman -Q linux-omarchy shows a build with the option unset, remove the workaround:
zgrep INTEL_IOMMU_DEFAULT_ON /proc/config.gz # run while booted on linux-omarchy
sudo rm /etc/limine-entry-tool.d/intel-iommu-off.conf
sudo limine-mkinitcpio
Verify
After a reboot on the linux-omarchy entry: uname -r ends in -omarchy, grep -o intel_iommu=off /proc/cmdline prints the parameter, journalctl -k -b | grep -c 'DMAR: \[DMA' prints 0, and on the TPM case ls /dev/tpmrm0 exists and systemd-analyze no longer shows a 90 s userspace phase.
Sources
- https://github.com/omacom/omarchy/issues/12119
- https://github.com/omacom/omarchy/issues/12097
- https://github.com/omacom/omarchy/issues/12550
- https://github.com/omacom/omarchy/issues/14160
- https://github.com/omacom/omarchy-pkgs/issues/677
- https://github.com/omacom/omarchy/issues/13030
- https://gitlab.com/Zesko/limine-entry-tool
- https://wiki.archlinux.org/title/Kernel_parameters