linux-stable/drivers/i2c
Hans de Goede cd86d1403b i2c: i2c-designware-platdrv: Always use a dynamic adapter number
Before this commit the i2c-designware-platdrv assumes that if the pdev
has an apci-companion it should use a dynamic adapter-nr and it sets
adapter->nr to -1, otherwise it will use pdev->id as adapter->nr.

There are 3 ways how platform_device-s to which i2c-designware-platdrv
will bind can be instantiated:

1) Through of / devicetree
2) Through ACPI enumeration
3) Explicitly instantiated through platform_device_create + add

1) In case of devicetree-instantiation the drivers/of code always sets
pdev->id to PLATFORM_DEVID_NONE, which is -1 so in this case both paths
to set adapter->nr end up doing the same thing.

2) In case of ACPI instantiation the device will always have an
ACPI-companion, so we are already using dynamic adapter-nrs.

3) There are 2 places manually instantiating a designware_i2c platform_dev:
drivers/mfd/intel_quark_i2c_gpio.c
drivers/mfd/intel-lpss.c

In the intel_quark_i2c_gpio.c case pdev->id is always 0, so switching to
dynamic adapter-nrs here could lead to the bus-number no longer being
stable, but the quark X1000 only has 1 i2c-controller, which will also
be assigned bus-number 0 when using dynamic adapter-nrs.

In the intel-lpss.c case intel_lpss_probe() is called from either
intel-lpss-acpi.c in which case there always is an ACPI-companion, or
from intel-lpss-pci.c. In most cases devices handled by intel-lpss-pci.c
also have an ACPI-companion, so we use a dynamic adapter-nr. But in some
cases the ACPI-companion is missing and we would use pdev->id (allocated
from intel_lpss_devid_ida). Devices which use the intel-lpss-pci.c code
typically have many i2c busses, so using pdev->id in this case may lead
to a bus-number conflict, triggering a WARN(id < 0, "couldn't get idr")
in i2c-core-base.c causing an oops an the adapter registration to fail.
So in this case using non dynamic adapter-nrs is actually undesirable.

One machine on which this oops was triggering is the Apollo Lake based
Acer TravelMate Spin B118.

TL;DR: Switching to always using dynamic adapter-numbers does not make
any difference in most cases and in the one case where it does make a
difference the behavior change is desirable because the old behavior
caused an oops.

BugLink: https://bugzilla.redhat.com/show_bug.cgi?id=1687065
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Acked-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Acked-by: Jarkko Nikula <jarkko.nikula@linux.intel.com>
Signed-off-by: Wolfram Sang <wsa@the-dreams.de>
2019-03-13 18:07:10 +01:00
..
algos i2c: algo-bit: convert to SPDX header 2019-02-05 14:05:03 +01:00
busses i2c: i2c-designware-platdrv: Always use a dynamic adapter number 2019-03-13 18:07:10 +01:00
muxes Merge branch 'i2c/for-4.20' of git://git.kernel.org/pub/scm/linux/kernel/git/wsa/linux 2018-10-29 14:44:03 -07:00
i2c-boardinfo.c i2c: i2c-boardinfo: fix memory leaks on devinfo 2017-11-27 19:14:29 +01:00
i2c-core-acpi.c i2c: acpi: Introduce i2c_acpi_get_i2c_resource() helper 2018-12-03 21:40:02 +02:00
i2c-core-base.c i2c: add extra check to safe DMA buffer helper 2019-03-13 18:03:26 +01:00
i2c-core-of.c i2c: of: Try to find an I2C adapter matching the parent 2019-02-05 13:41:49 +01:00
i2c-core-slave.c i2c: remove i2c_lock_adapter and use i2c_lock_bus directly 2018-07-13 00:09:37 +02:00
i2c-core-smbus.c i2c: core-smbus: don't trace smbus_reply data on errors 2019-02-14 18:01:33 +01:00
i2c-core.h i2c: make i2c_check_addr_validity() static 2018-03-24 14:07:23 +01:00
i2c-dev.c i2c: expand minor range when registering chrdev region 2019-02-15 09:53:34 +01:00
i2c-mux.c Merge branch 'i2c/for-4.19' of git://git.kernel.org/pub/scm/linux/kernel/git/wsa/linux 2018-08-21 17:40:46 -07:00
i2c-slave-eeprom.c
i2c-smbus.c i2c: i2c-smbus: add of_i2c_setup_smbus_alert 2017-10-28 23:42:47 +02:00
i2c-stub.c treewide: kzalloc() -> kcalloc() 2018-06-12 16:19:22 -07:00
Kconfig treewide: surround Kconfig file paths with double quotes 2018-12-22 00:25:54 +09:00
Makefile i2c: don't use any __deprecated handling anymore 2018-08-24 17:26:43 +02:00