Hi All, So here is my third attempt at dealing with multiple i2c devices being described in a single ACPI fwnode. Like the second attemp This is still based on having a special driver for these corner cases which can be a module and only loaded when on hardware which needs this. New this time around is that this now is a platform driver, avoiding the need for the I2C_CLIENT_IGNORE_BUSY hack and having 2 i2c-clients at the address of the first i2c ACPI resource. While working on the ACPI changes for enumerating ACPI nodes which are handled by the new i2c-multi-instantiate driver as platform devices I found a bug caused by a combination of my: "ACPI / scan: Use acpi_bus_get_status() to initialize ACPI_TYPE_DEVICE devs" and the somewhat (but not much) later added: "ACPI / scan: Do not enumerate Indirect IO host children" commits, specifically the second one does not work because of the first one, I guess it was tested against a kernel without the first commit. My changes for enumerating relevant ACPI nodes as platform devices hit the same problem. Thus the first patch of this series is an ACPI bugfix for that problem. The second patch changes acpi/scan.c to enumerate relevant ACPI nodes as platform devices. The third patch is an ACPI cleanup patch made possible by the first patch. All 3 these patches are new in this version of this series. And finally the fourth patch is the actual i2c-multi-instantiate driver, now as platform driver living under drivers/platform/x86 . Regards, Hans
