Re: [RFC v9 1/1] leds: asus-aura-scsi: Add ASUS Aura RGB LED driver for ROG NVMe enclosures

[Liang Haowen <nbg2974@gmail.com>]

On Thu, 15 Oct 2026, Lee Jones wrote:

> A driver source file without 'Kconfig' or 'Makefile' integration is dead code in
> the tree.  Either wire it up now or hold off on submitting until the rest of the
> pieces are ready.

v10 wires it up: the drivers/leds/rgb/Kconfig and Makefile entries
are in, and the series is posted as a new thread under the same
subject.

Your 25 Sep note never reached me: neither it nor the copy
kernel.org relayed on 14 Oct is in any folder of my mailbox. I
first saw its content today, quoted in this mail. The three week
gap was delivery, not silence; this is my first chance to answer
it.

> Thank you for your contribution! Sashiko AI review found 3 potential issue(s) to consider:
> - [Low] leds: asus-aura-scsi: undefined function kzalloc_obj() causes build failure
> - [Low] leds: asus-aura-scsi: blk_rq_map_kern() missing required request_queue argument
> - [Low] leds: asus-aura-scsi: strscpy() called with missing size argument
>
> Was this even tested?

Yes. The driver builds with clang (LLVM=1) against 7.2.4 with zero
warnings, loads on the enclosure, and was exercised today: the four
LED zones driven to independent colours and checked by eye, 60
sequential and 100 concurrent updates, the sdev deleted mid-write
(zero splats, clean detach), rmmod, and a fresh attach through a
host scan.

All three findings assume a tree older than v7.0:

- kzalloc_obj() has existed since v7.0 (Kees Cook's series,
  include/linux/slab.h).
- blk_rq_map_kern() takes four arguments on current kernels; the
  queue comes from the request itself, and scsi_execute_cmd()
  calls it the same way.
- strscpy() with two arguments is backed by __strscpy0 in
  include/linux/string.h, which infers sizeof(dst) and rejects
  non-array destinations. v10 passes the size explicitly anyway,
  since it costs nothing.

kzalloc_obj() has now been reported by the bot in all nine of
its reviews and blk_rq_map_kern() in eight, and the answer has
not changed since the first time. If sashiko could review
against a current tree, or drop those two checks, the noise
would stop.

Best regards,
Liang