Re: [PATCH 02/15] HID: hid-oxp: validate input report lengths before decoding

[sashiko-bot@kernel.org]

Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:

Pre-existing issues:
- [High] Shared driver state variables are concurrently modified from atomic context and read from process context without protection, enabling a TOCTOU out-of-bounds read.
--

commit c83e5978640f43b909fcea30940f12727c190fcd
Author: Andrei Aldea <andrei1998@gmail.com>
Subject: HID: hid-oxp: validate input report lengths before decoding

This commit adds length checks to ensure short USB reports do not result in
out-of-bounds reads when decoding RGB status and printing debug information.

> diff --git a/drivers/hid/hid-oxp.c b/drivers/hid/hid-oxp.c
> index 3f2be8059ddb1..0c895bd778bbf 100644
> --- a/drivers/hid/hid-oxp.c
> +++ b/drivers/hid/hid-oxp.c
[ ... ]
> @@ -404,6 +411,8 @@ static int oxp_hid_raw_event_gen_2(struct hid_device *hdev,
>  
>  	if (data[3] != OXP_GET_PROPERTY)
>  		return 0;
> +	if (size < sizeof(*rgb_rep))
> +		return 0;
>  
>  	rgb_rep = (struct oxp_gen_2_rgb_report *)data;
>  	/* Ensure we save monocolor as the list value */

[Severity: High]
This isn't a bug introduced by this patch, but can the lockless assignment of
drvdata.rgb_effect here cause a Time-Of-Check to Time-Of-Use (TOCTOU)
out-of-bounds read in effect_show()?

In drivers/hid/hid-oxp.c, oxp_hid_raw_event_gen_2() updates the effect
locklessly from the raw event handler:

	drvdata.rgb_effect = rgb_rep->effect == OXP_EFFECT_MONO_TRUE ?
			     OXP_EFFECT_MONO_LIST :
			     rgb_rep->effect;

Concurrently, a userspace process reading the effect sysfs attribute calls
effect_show(), which performs two separate lockless reads of
drvdata.rgb_effect:

static ssize_t effect_show(...)
{
    ...
	if (drvdata.rgb_effect >= ARRAY_SIZE(oxp_rgb_effect_text))
		return -EINVAL;

	return sysfs_emit(buf, "%s\n", oxp_rgb_effect_text[drvdata.rgb_effect]);
}

If a malicious device sends continuous status event reports, can the value of
drvdata.rgb_effect change to an out-of-bounds value immediately after the
bounds check but before the array access? This would cause sysfs_emit() to
dereference a garbage pointer.

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20260910032115.28669-1-andrei1998@gmail.com?part=2