Re: [PATCH 1/2] Input: ilitek_ts: add stylus input support

Thank you for your contribution! Sashiko AI review found 4 potential issue(s) to consider:

New issues:
- [High] Lazy initialization of the pen input device from within the IRQ handler causes a devres LIFO ordering inversion, leading to a memory leak and Use-After-Free during device unbind.
- [High] The pen input device axis limits are not synced with Device Tree properties, leading to mismatched axis boundaries and massive coordinate underflow values.
- [Low] Direct bit manipulation is used instead of the preferred `input_set_capability()` helper.

Pre-existing issues:
- [High] I2C transfers are performed using buffers allocated on the stack, which is unsafe for DMA.
--

commit 0912736a7a818d72cc6868aba6ad50c6f06afd78
Author: Kristian Mide [off-list ref]

Input: ilitek_ts: add stylus input support

Add a separate stylus input device for report ID 0x0c packets,
with pressure, hover, and side-button support. The pen device is
created lazily on first pen report so touchscreen-only hardware does
not expose stylus capabilities up front.

[ ... ]

[Severity: High]
This is a pre-existing issue and was not introduced by the patch under
review, but since buf is passed here, is it safe to allocate buf as a
512-byte array on the stack in ilitek_process_and_report_v6() and pass it to
I2C transfer functions?

When CONFIG_VMAP_STACK is enabled, stack addresses cannot be safely mapped
for DMA. If the underlying I2C host controller uses DMA, could attempting
to map this stack buffer lead to memory corruption or kernel panics?

+
 	if (buf[0] != ILITEK_TP_I2C_REPORT_ID) {
 		dev_err(dev, "get touch info failed. Wrong id: 0x%02X\n", buf[0]);
 		return -EINVAL;

[ ... ]

[Severity: Low]
This isn't a bug, but could we use the input_set_capability() helper here
instead of direct bit manipulation?

+
+	input_set_abs_params(pen_input, ABS_X,
+			     ts->screen_min_x, ts->screen_max_x, 0, 0);
+	input_set_abs_params(pen_input, ABS_Y,
+			     ts->screen_min_y, ts->screen_max_y, 0, 0);

[Severity: High]
Should the touchscreen properties (DT properties) be applied to these
absolute parameters?

The touchscreen_report_pos() function called during event processing will
apply inversion and swapping based on Device Tree properties to the
coordinates.

If the limits advertised here are strictly from the raw hardware bounds and
are unaware of those DT properties, could this cause the device to report
out-of-bounds coordinates or result in unsigned underflows when an axis is
inverted?

+	input_set_abs_params(pen_input, ABS_PRESSURE, 0,
+			     ILITEK_PEN_PRESSURE_MAX, 0, 0);
+	input_set_abs_params(pen_input, ABS_DISTANCE, 0,
+			     ILITEK_PEN_DISTANCE_MAX, 0, 0);
+	input_abs_set_res(pen_input, ABS_X, ILITEK_PEN_RESOLUTION);
+	input_abs_set_res(pen_input, ABS_Y, ILITEK_PEN_RESOLUTION);
+
+	error = input_register_device(pen_input);
+	if (error)
+		goto err_free_pen_input;
+
+	error = devm_add_action_or_reset(dev, ilitek_unregister_pen_input, ts);

[Severity: High]
Since this function is called lazily from the threaded IRQ handler after
probe has completed, does adding a devm action here break the LIFO teardown
ordering?

During device unbind, devres will release resources in LIFO order. This
action would run first and set ts->pen_input_dev = NULL, but the IRQ
handler would still be active.

If a hardware interrupt fires during unbind after this has been
unregistered, would the IRQ handler observe the NULL pointer, allocate a
new pen device, and add a new devres action that devres_release_all()
will never process, resulting in a memory leak and a potential
use-after-free?

+	if (error)
+		return error;
+
+	return 0;
+
+err_free_pen_input:
+	ts->pen_input_dev = NULL;
+	input_free_device(pen_input);
+	return error;
+}

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20260626214248.5563-1-kristian@mide.dk?part=1