Re: [PATCH 2/4] drivers/misc: add Goodix GXFP5130 eSPI fingerprint sensor driver

[Metehan Günen]

On Sat, Jul 18, 2026 at 11:12:28AM +0200, Greg KH wrote:
> This is not how you use debugfs.
>
> Also, this implements a custom trace function, again, use the built-in
> kernel functions for this.

Understood on both counts. The IS_ERR_OR_NULL() guard around the
debugfs_create_dir() return value is wrong — debugfs functions handle a
NULL or error dir silently in subsequent calls, so the check is
unnecessary. The gxfp_trace module will be removed entirely in v2 and
replaced with dev_dbg() throughout.

> This implies that the driver came from an external company that does
> not have much kernel development experience, or a LLM wrote this.  As
> there is no copyright information on the code, which is it?

The original code was written by Void755 (github.com/Void755) for
out-of-tree use under GPL-2.0 on hardware not supported by any existing
driver. I ported it to mainline kernel style with assistance from an AI
tool, which is responsible for the debugfs misuse, the custom trace
system, and the other style issues you found — it applied patterns that
compile but do not follow kernel conventions. I should have caught these
before sending.

v2 will carry proper copyright headers:

  Copyright (C) 2024 Void755 <github.com/Void755>
  Copyright (C) 2026 Metehan Günen <metehangnen@xxxxxxxxx>

> That's an abuse of READ_ONCE() :(

Agreed — READ_ONCE() is for lockless concurrent access paths, not a
plain init flag. Will be replaced with a regular boolean check in v2.

Thanks,
Metehan Günen