[RFC 0/2] ACPI: Adding new acpi_driver type drivers ?

[Hans de Goede <hdegoede@redhat.com>]

Hi Rafael,

I recently learned that some Dell AIOs (1) use a backlight controller board
connected to an UART. Canonical even submitted a driver for this in 2017:
https://lkml.org/lkml/2017/10/26/78

This UART has a DELL0501 HID with CID set to PNP0501 so that the UART is
still handled by 8250_pnp.c. Unfortunately there is no separate ACPI device
with an UartSerialBusV2() resource to model the backlight-controller.

The RFC patch 2/2 in this series uses acpi_quirk_skip_serdev_enumeration()
to still create a serdev for this for a backlight driver to bind to
instead of creating a /dev/ttyS0.

Like other cases where the UartSerialBusV2() resource is missing or broken
this will only create the serdev-controller device and the serdev-device
itself will need to be instantiated by the consumer (the backlight driver).

Unlike existing other cases which use DMI modaliases to load on a specific
board to work around brokeness of that board's specific ACPI tables, the
intend here is to have a single driver for all Dell AIOs using the DELL0501
HID for their UART, without needing to maintain a list of DMI matches.

This means that the dell-uart-backlight driver will need something to bind
to. The original driver from 2017 used an acpi_driver for this matching on
and binding to the DELL0501 acpi_device.

AFAIK you are trying to get rid of having drivers bind directly to
acpi_device-s so I assume that you don't want me to introduce a new one.
So to get a device to bind to without introducing a new acpi_driver
patch 2/2 if this series creates a platform_device for this.

The creation of this platform_device is why this is marked as RFC,
if you are ok with this solution I guess you can merge this series
already as is. With the caveat that the matching dell-uart-backlight
driver is still under development (its progressing nicely and the
serdev-device instantation + binding a serdev driver to it already
works).

If you have a different idea how to handle this I'm certainly open
to suggestions.

Regards,

Hans

1) All In One a monitor with a PC builtin


p.s.

I also tried this approach, but that did not work:

This was an attempt to create both a pdev from acpi_default_enumeration()
by making the PNP scan handler attach() method return 0 rather then 1;
and get a pnp_device created for the UART driver as well by
making acpi_is_pnp_device() return true.

This approach does not work due to the following code in pnpacpi_add_device():

	/* Skip devices that are already bound */
	if (device->physical_node_count)
		return 0;

diff --git a/drivers/acpi/acpi_pnp.c b/drivers/acpi/acpi_pnp.c
index 01abf26764b0..847c08deea7b 100644
--- a/drivers/acpi/acpi_pnp.c
+++ b/drivers/acpi/acpi_pnp.c
@@ -353,10 +353,17 @@ static bool acpi_pnp_match(const char *idstr, const struct acpi_device_id **matc
  * given ACPI device object, the PNP scan handler will not attach to that
  * object, because there is a proper non-PNP driver in the kernel for the
  * device represented by it.
+ *
+ * The DELL0501 ACPI HID represents an UART (CID is set to PNP0501) with
+ * a backlight-controller attached. There is no separate ACPI device with
+ * an UartSerialBusV2() resource to model the backlight-controller.
+ * This setup requires instantiating both a pnp_device for the UART as well
+ * as a platform_device for the backlight-controller driver to bind too.
  */
 static const struct acpi_device_id acpi_nonpnp_device_ids[] = {
 	{"INTC1080"},
 	{"INTC1081"},
+	{"DELL0501"},
 	{""},
 };
 
@@ -376,13 +383,16 @@ static struct acpi_scan_handler acpi_pnp_handler = {
  * For CMOS RTC devices, the PNP ACPI scan handler does not work, because
  * there is a CMOS RTC ACPI scan handler installed already, so we need to
  * check those devices and enumerate them to the PNP bus directly.
+ * For DELL0501 devices the PNP ACPI scan handler is skipped to create
+ * a platform_device, see the acpi_nonpnp_device_ids[] comment.
  */
-static int is_cmos_rtc_device(struct acpi_device *adev)
+static int is_special_pnp_device(struct acpi_device *adev)
 {
 	static const struct acpi_device_id ids[] = {
 		{ "PNP0B00" },
 		{ "PNP0B01" },
 		{ "PNP0B02" },
+		{ "DELL0501" },
 		{""},
 	};
 	return !acpi_match_device_ids(adev, ids);
@@ -390,7 +400,7 @@ static int is_cmos_rtc_device(struct acpi_device *adev)
 
 bool acpi_is_pnp_device(struct acpi_device *adev)
 {
-	return adev->handler == &acpi_pnp_handler || is_cmos_rtc_device(adev);
+	return adev->handler == &acpi_pnp_handler || is_special_pnp_device(adev);
 }
 EXPORT_SYMBOL_GPL(acpi_is_pnp_device);

 
Hans de Goede (2):
  ACPI: x86: Move acpi_quirk_skip_serdev_enumeration() out of
    CONFIG_X86_ANDROID_TABLETS
  ACPI: x86: Add DELL0501 handling to
    acpi_quirk_skip_serdev_enumeration()

 drivers/acpi/x86/utils.c | 38 ++++++++++++++++++++++++++++++++++----
 include/acpi/acpi_bus.h  | 22 +++++++++++-----------
 2 files changed, 45 insertions(+), 15 deletions(-)

-- 
2.43.0