udev patterns are shell globs, so nvme[0-9]*n[0-9]* also matched every
partition. Partitions have no queue/scheduler, and udev logged a "Could not
chase sysfs attribute" for each one at boot. Restrict the match to
SUBSYSTEM block with DEVTYPE disk, which also keeps mmcblk boot areas and
NVMe multipath nodes out.
Reword the comment: kyber targets a read latency rather than bounding it,
and the kernel's default choice depends on the device rather than being a
fixed NVMe versus everything-else split.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
The kernel leaves NVMe on none and everything else on mq-deadline. Neither
bounds latency once the device queue fills, so a large build, copy, or
package upgrade can make the desktop sluggish while reads wait behind a
wall of writes.
Kyber keeps separate read and sync-write queues and throttles the depth it
submits to hit a 2ms read target, which keeps interactive reads flowing
under heavy writes at negligible CPU cost. The trade is a small ceiling on
peak throughput on very fast devices, which matters for a storage server
chasing IOPS but not for a desktop.
Ships as a package-owned udev rule in /etc so it applies at boot and on
hot-plug. Existing installs pick it up at the next boot. zram is left alone
since it has nothing to schedule.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
