On Tue, 2017-07-04 at 14:46 +0200, Hans de Goede wrote:
> ACPI video devices get tagged by the kernel with the custom LNXVIDEO
> HID so that normal pnp-id matching can be used and are handled by the
> acpi-video driver.
>
> Sometimes the ACPI nodes describing these contain a SERIAL_TYPE_I2C
> ACPI
> resource. Before this commit the presence of this resource would cause
> the
> i2c-core to create a /sys/bus/i2c/devices/i2c-LNXVIDEO:00 device for
> this
> with a modalias of: "i2c:LNXVIDEO:00".
>
> There is no i2c driver for this custom HID, the acpi-video driver
> binds
> directly to the ACPI device /sys/bus/acpi/devices/LNXVIDEO\:00 which
> has
> a modalias of "acpi:LNXVIDEO:" .
>
> Not only is the creation of an i2c-client for this undesirable, it is
> actually causing problems. This weird pseudo-resource claims an i2c
> speed of 100KHz and typically points to the i2c bus which is used by
> the
> touchscreen controller. Some touchscreen controllers only work
> properly at
> 400KHz, at 100KHz they cause errors like these:
>
> i2c_designware 80860F41:03: i2c_dw_handle_tx_abort: lost arbitration
> i2c_designware 80860F41:03: i2c_dw_handle_tx_abort: lost arbitration
> i2c_designware 80860F41:03: i2c_dw_handle_tx_abort: lost arbitration
> i2c_designware 80860F41:03: i2c_dw_handle_tx_abort: lost arbitration
> silead_ts i2c-MSSL1680:00: Registers clear error -11
>
> This commit makes the i2c-core ignore LNXVIDEO compatible ACPI devices
> which has 2 positive results:
>
> 1) The bogus i2c-client for these is no longer created.
> 2) i2c_acpi_lookup_speed now ignores the 100KHz speed from the pseudo
> i2c-resouce and properly returns 400KHz as speed for the touchscreen
> i2c bus, fixing the touchscreen not working on various devies.
>
Thanks for fixing this.
Reviewed-by: Andy Shevchenko <andriy.shevchenko@xxxxxxxxxxxxxxx>
One comment below.
> Signed-off-by: Hans de Goede <hdegoede@xxxxxxxxxx>
> ---
> Changes in v2:
> -Rebase on top of linux-i2c/for-next
> ---
> drivers/i2c/i2c-core-acpi.c | 11 +++++++++++
> 1 file changed, 11 insertions(+)
>
> diff --git a/drivers/i2c/i2c-core-acpi.c b/drivers/i2c/i2c-core-acpi.c
> index 052005579ed6..bda281293d28 100644
> --- a/drivers/i2c/i2c-core-acpi.c
> +++ b/drivers/i2c/i2c-core-acpi.c
> @@ -82,11 +82,22 @@ static int i2c_acpi_do_lookup(struct acpi_device
> *adev,
> struct i2c_board_info *info = lookup->info;
> struct list_head resource_list;
> int ret;
> + static const struct acpi_device_id video_device_ids[] = {
> + { ACPI_VIDEO_HID, 0 },
> + {}
> + };
Can we move this out of the function body?
>
> if (acpi_bus_get_status(adev) || !adev->status.present ||
> acpi_device_enumerated(adev))
> return -EINVAL;
>
> + /*
> + * ACPI video acpi_devices, which are handled by the acpi-
> video driver
> + * sometimes contain a SERIAL_TYPE_I2C ACPI resource, ignore
> these.
> + */
> + if (acpi_match_device_ids(adev, video_device_ids) == 0)
> + return -ENODEV;
> +
> memset(info, 0, sizeof(*info));
> lookup->device_handle = acpi_device_handle(adev);
>
--
Andy Shevchenko <andriy.shevchenko@xxxxxxxxxxxxxxx>
Intel Finland Oy
