$ the-wire · showcase
Omarchy hardens lock and sleep hook installs, restrains Kitty remote control
By RepoJournal · Filed · About Linux
Omarchy ships hardening for lock authentication and sleep-hook ownership, while Kitty's remote-control defaults move to the system config to close an untrusted-terminal attack channel.
Omarchy hardened two privilege boundaries. The lock authentication command lookup no longer trusts the environment it runs in [1], and installed sleep hooks (keyboard-backlight and hybrid-GPU, plus the supergfxd drop-in) are now published through hidden root-owned staging inodes with atomic replacement; unsafe existing copies are repaired without trusting current contents [2][3]. The related merge [3] carries the same ownership hardening for sleep hooks, and the "root-owned sleep hooks" description says the change applies "the same ownership boundary to the supergfxd service drop-in installed by the hybrid-GPU toggle" [2], so a complete install should close both paths.
Kitty's defaults now ship in `/etc/xdg/kitty/kitty.conf` with `allow_remote_control` set to `socket-only`, rejecting the remote-control channel from terminal output while keeping the local Unix socket used for active-tab directory lookup. The user config only contains the theme include and commented examples, so removing the include disconnects Kitty from Omarchy's theming system [4]. Separately, the locate configuration script `install/config/locate.sh` is gone: the two needed `updatedb` options (index Btrfs subvolume mounts, exclude `/.snapshots`) are fixed arguments in a package-owned drop-in for `plocate-updatedb.service`, passed identically in the installer and explicit AUR refresh, removing a privileged script and its migration [5].
On the kernel side, Linus pulled the tracing merge with a fix for tracefs files that did not take the `trace_array` reference, closing a lifetime gap around instances created and destroyed via the `instances` directory [6]. The locking merge fixes a softirq processing delay bug in `local_interrupt_disable()` that "should mostly only affect the Rust runtime", removes the now-unnecessary `hardirq_disable_count()`, invalidates stale lockdep `class_cache` entries, and fixes an `rt_mutex` issue [7]. The IRQ merge reverts an mbigen driver change that regressed two-port Hi1616 chips, fixes a too-long-preemption-off in the stm32mp-exti driver caused by a time unit ambiguity, and removes the unused `irq_domain_add_linear()` [8].
Action items
- → Update Omarchy to include the lock authentication and sleep-hook hardening fixes basecamp/omarchy [immediate]
- → Review your Kitty config for overrides that re-enable allow_remote_control basecamp/omarchy [immediate]
- → On affected Hi1616 two-port systems, hold off until the mbigen revert lands in your kernel torvalds/linux [monitor]
References
- [1] Merge pull request #10225 from mdisec/security/harden-quattro-fprintd-path ↗ basecamp/omarchy
- [2] Harden ownership of installed sleep hooks ↗ basecamp/omarchy
- [3] Merge pull request #10425 from acrogenesis/security/root-owned-sleep-hooks ↗ basecamp/omarchy
- [4] Move Kitty defaults to the system config and restrict remote control ↗ basecamp/omarchy
- [5] Replace locate configuration script with packaged service defaults ↗ basecamp/omarchy
- [6] Merge tag 'trace-v7.3-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/trace/linux-trace ↗ torvalds/linux
- [7] Merge tag 'locking-urgent-2026-09-06' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip ↗ torvalds/linux
- [8] Merge tag 'irq-urgent-2026-09-06' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip ↗ torvalds/linux