When an i2c-hid device is a panel follower, drm_atomic_helper_shutdown()
unprepares the panel before device_shutdown() gets around to the i2c
client. The panel_unpreparing() callback already calls
i2c_hid_core_suspend() with force_poweroff set, which sends
SET_POWER(SLEEP) and then drops the device's regulators.

So by the time i2c_hid_core_shutdown() runs, the controller is unpowered
and can no longer complete a transfer. On MT8196 the SET_POWER(SLEEP)
issued there burns the full i2c-mt65xx bus timeout before failing,
adding two seconds to every reboot.

Skip the redundant SET_POWER for panel followers, matching how
i2c_hid_core_pm_suspend() and friends already bail out for them.
free_irq() and the shutdown_tail() hook are unaffected.

Fixes: 96a37bfd232a ("HID: i2c-hid: Support being a panel follower")
Signed-off-by: Justin Yeh <justin.yeh@mediatek.com>
---
 drivers/hid/i2c-hid/i2c-hid-core.c | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)

diff --git a/drivers/hid/i2c-hid/i2c-hid-core.c b/drivers/hid/i2c-hid/i2c-hid-core.c
index 0ff07fdab442..2e787c17256d 100644
--- a/drivers/hid/i2c-hid/i2c-hid-core.c
+++ b/drivers/hid/i2c-hid/i2c-hid-core.c
@@ -1355,7 +1355,9 @@ void i2c_hid_core_shutdown(struct i2c_client *client)
 {
 	struct i2c_hid *ihid = i2c_get_clientdata(client);
 
-	i2c_hid_set_power(ihid, I2C_HID_PWR_SLEEP);
+	if (!ihid->is_panel_follower)
+		i2c_hid_set_power(ihid, I2C_HID_PWR_SLEEP);
+
 	free_irq(client->irq, ihid);
 
 	i2c_hid_core_shutdown_tail(ihid);
-- 
2.45.2

Hi,

On Thu, Sep 3, 2026 at 8:50 PM Justin Yeh <justin.yeh@mediatek.com> wrote:
>
> When an i2c-hid device is a panel follower, drm_atomic_helper_shutdown()
> unprepares the panel before device_shutdown() gets around to the i2c
> client. The panel_unpreparing() callback already calls
> i2c_hid_core_suspend() with force_poweroff set, which sends
> SET_POWER(SLEEP) and then drops the device's regulators.
>
> So by the time i2c_hid_core_shutdown() runs, the controller is unpowered
> and can no longer complete a transfer. On MT8196 the SET_POWER(SLEEP)
> issued there burns the full i2c-mt65xx bus timeout before failing,
> adding two seconds to every reboot.

I don't think there is any guarantee here though, is there? In your
case the panel might have been powered off before
i2c_hid_core_shutdown(), but is there something that ensures that? I'm
not aware of it.

Would it be better to follow in the example of i2c_hid_core_remove()
and just remove ourselves as a panel follower? That would
automatically call the power-off functions if needed, right?

-Doug