$ the-wire · showcase
regmap's raw-spinlock allocation fix, and Omarchy's LUKS layout regression
By RepoJournal · Filed · About Linux · Composed from the cited sources · methodology
Two fixes lead the day: MMIO regmaps with a raw spinlock and a cache no longer allocate under that lock, and Omarchy keeps a non-Latin vconsole.conf out of the initramfs so LUKS passphrases stay typeable.
Merge tag 'regmap-fix-v7.3-rc7' of git://git.kernel.org/pub/scm/linux/kernel/git/broonie/regmap torvalds/linux
Mark Brown's pull describes the root cause plainly: earlier lockdep fixes done for the maple tree on sleeping buses left MMIO regmaps with a cache allocating under a raw spinlock, and the merge corrects that. While fixing it, Michael Reeves also noticed the debugfs files were briefly exposed. If you run regmap over MMIO with a cached, raw-spinlock-backed map, this is the warning flood you are s...
Keep a non-Latin vconsole.conf out of the initramfs again (#14885) basecamp/omarchy
Since Arch's 26.134.222-3 plymouth hook copies /etc/vconsole.conf into the initramfs on its own, the earlier guard that only removed the file from FILES stopped working, and with a first XKB layout that types no Latin letters Plymouth asks for the LUKS passphrase in a layout that cannot produce it. omarchy_vconsole.conf now refuses the file. Anyone with a non-Latin keymap and an encrypted root ...
Have a moved gateway answer, running doctor once if it does not basecamp/omarchy
A gateway set up by an older OpenClaw keeps its agent database at that release's schema, so the seeded runtime starts and then exits 78 asking for `openclaw doctor --fix`, while systemctl had already reported active. The move now requires a gateway that was running to answer before the run continues, and runs doctor once if it does not. This hit exactly the 4.0.4 users the migration exists for.
Revert migration-core changes from #14212 (#15864) basecamp/omarchy
With 4.0.6 and the seed package landing on each channel together, the deferred migration is unnecessary: migration 1792334754 now moves OpenClaw when the package is the seed and otherwise finishes with nothing to do, and omarchy-migrate, its test and agents/skills/migrations.md revert to their previous state. Machines without the package behave as they already did.
Merge tag 'scsi-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/jejb/scsi torvalds/linux
James Bottomley's SCSI pull is all driver and tape ULD fixes plus one bsg core fix, where reply_len is now threaded through the generic pieces; he calls the race in auth updates for iscsi configfs the biggest item in the set. Small collections also landed: three i2c clock management fixes and a dropped HALT_ACK reset in qcom-cci, driver-specific spi fixes, and a firewire race in split transacti...
