| Message ID | 20211230141722.512395-1-hdegoede@redhat.com |
|---|---|
| Headers | show |
| Series | ACPI / pdx86: Add support for x86 Android tablets with broken DSDTs | expand |
On Thu, Dec 30, 2021 at 3:17 PM Hans de Goede <hdegoede@redhat.com> wrote: > > x86 ACPI boards which ship with only Android as their factory image usually > declare a whole bunch of bogus I2C devs in their ACPI tables and sometimes > there are issues with serdev devices on these boards too, e.g. the resource > points to the wrong serdev_controller. > > Instantiating I2C / serdev devs for these bogus devs causes various issues, > e.g. GPIO/IRQ resource conflicts because sometimes drivers do bind to them. > The Android x86 kernel fork shipped on these devices has some special code > to remove the bogus I2C clients (and serdevs are ignored completely). > > Introduce acpi_quirk_skip_i2c_client_enumeration() and > acpi_quirk_skip_serdev_enumeration() helpers. Which can be used by the I2C/ > serdev code to skip instantiating any I2C or serdev devs on broken boards. > > These 2 helpers are added to drivers/acpi/x86/utils.c so that the DMI table > can be shared between the I2C and serdev code. > > Note these boards typically do actually have I2C and serdev devices, just > different ones then the ones described in their DSDT. The devices which > are actually present are manually instantiated by the > drivers/platform/x86/x86-android-tablets.c kernel module. > > The new helpers are only build if CONFIG_X86_ANDROID_TABLETS is enabled, > otherwise they are empty stubs to not unnecessarily grow the kernel size. > > Signed-off-by: Hans de Goede <hdegoede@redhat.com> > --- > Changes in v2: > - Move the i2c_acpi_known_good_ids[] list and checking into this patch / > into the new acpi_quirk_skip_i2c_client_enumeration() function Acked-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> or please let me know if you want me to take this series. Thanks! > --- > drivers/acpi/x86/utils.c | 111 +++++++++++++++++++++++++++++++++++++++ > include/acpi/acpi_bus.h | 16 ++++++ > 2 files changed, 127 insertions(+) > > diff --git a/drivers/acpi/x86/utils.c b/drivers/acpi/x86/utils.c > index a2ae1ac41319..375a0911f06d 100644 > --- a/drivers/acpi/x86/utils.c > +++ b/drivers/acpi/x86/utils.c > @@ -10,6 +10,7 @@ > > #include <linux/acpi.h> > #include <linux/dmi.h> > +#include <linux/platform_device.h> > #include <asm/cpu_device_id.h> > #include <asm/intel-family.h> > #include "../internal.h" > @@ -208,3 +209,113 @@ bool force_storage_d3(void) > { > return x86_match_cpu(storage_d3_cpu_ids); > } > + > +#if IS_ENABLED(CONFIG_X86_ANDROID_TABLETS) > +/* > + * x86 ACPI boards which ship with only Android as their factory image usually > + * declare a whole bunch of bogus I2C devices in their ACPI tables and sometimes > + * there are issues with serdev devices on these boards too, e.g. the resource > + * points to the wrong serdev_controller. > + * > + * Instantiating I2C / serdev devs for these bogus devs causes various issues, > + * e.g. GPIO/IRQ resource conflicts because sometimes drivers do bind to them. > + * The Android x86 kernel fork shipped on these devices has some special code > + * to remove the bogus I2C clients (and AFAICT serdevs are ignored completely). > + * > + * The acpi_quirk_skip_*_enumeration() functions below are used by the I2C or > + * serdev code to skip instantiating any I2C or serdev devs on broken boards. > + * > + * In case of I2C an exception is made for HIDs on the i2c_acpi_known_good_ids > + * list. These are known to always be correct (and in case of the audio-codecs > + * the drivers heavily rely on the codec being enumerated through ACPI). > + * > + * Note these boards typically do actually have I2C and serdev devices, > + * just different ones then the ones described in their DSDT. The devices > + * which are actually present are manually instantiated by the > + * drivers/platform/x86/x86-android-tablets.c kernel module. > + */ > +#define ACPI_QUIRK_SKIP_I2C_CLIENTS BIT(0) > +#define ACPI_QUIRK_UART1_TTY_UART2_SKIP BIT(1) > + > +static const struct dmi_system_id acpi_skip_serial_bus_enumeration_ids[] = { > + { > + .matches = { > + DMI_EXACT_MATCH(DMI_SYS_VENDOR, "ASUSTeK COMPUTER INC."), > + DMI_EXACT_MATCH(DMI_PRODUCT_NAME, "ME176C"), > + }, > + .driver_data = (void *)(ACPI_QUIRK_SKIP_I2C_CLIENTS | > + ACPI_QUIRK_UART1_TTY_UART2_SKIP), > + }, > + { > + .matches = { > + DMI_MATCH(DMI_SYS_VENDOR, "ASUSTeK COMPUTER INC."), > + DMI_MATCH(DMI_PRODUCT_NAME, "TF103C"), > + }, > + .driver_data = (void *)ACPI_QUIRK_SKIP_I2C_CLIENTS, > + }, > + { > + /* Whitelabel (sold as various brands) TM800A550L */ > + .matches = { > + DMI_MATCH(DMI_BOARD_VENDOR, "AMI Corporation"), > + DMI_MATCH(DMI_BOARD_NAME, "Aptio CRB"), > + /* Above strings are too generic, also match on BIOS version */ > + DMI_MATCH(DMI_BIOS_VERSION, "ZY-8-BI-PX4S70VTR400-X423B-005-D"), > + }, > + .driver_data = (void *)ACPI_QUIRK_SKIP_I2C_CLIENTS, > + }, > + {} > +}; > + > +static const struct acpi_device_id i2c_acpi_known_good_ids[] = { > + { "10EC5640", 0 }, /* RealTek ALC5640 audio codec */ > + { "INT33F4", 0 }, /* X-Powers AXP288 PMIC */ > + { "INT33FD", 0 }, /* Intel Crystal Cove PMIC */ > + { "NPCE69A", 0 }, /* Asus Transformer keyboard dock */ > + {} > +}; > + > +bool acpi_quirk_skip_i2c_client_enumeration(struct acpi_device *adev) > +{ > + const struct dmi_system_id *dmi_id; > + long quirks; > + > + dmi_id = dmi_first_match(acpi_skip_serial_bus_enumeration_ids); > + if (!dmi_id) > + return false; > + > + quirks = (unsigned long)dmi_id->driver_data; > + if (!(quirks & ACPI_QUIRK_SKIP_I2C_CLIENTS)) > + return false; > + > + return acpi_match_device_ids(adev, i2c_acpi_known_good_ids); > +} > +EXPORT_SYMBOL_GPL(acpi_quirk_skip_i2c_client_enumeration); > + > +int acpi_quirk_skip_serdev_enumeration(struct device *controller_parent, bool *skip) > +{ > + struct acpi_device *adev = ACPI_COMPANION(controller_parent); > + const struct dmi_system_id *dmi_id; > + long quirks = 0; > + > + *skip = false; > + > + /* !dev_is_platform() to not match on PNP enumerated debug UARTs */ > + if (!adev || !adev->pnp.unique_id || !dev_is_platform(controller_parent)) > + return 0; > + > + dmi_id = dmi_first_match(acpi_skip_serial_bus_enumeration_ids); > + if (dmi_id) > + quirks = (unsigned long)dmi_id->driver_data; > + > + if (quirks & ACPI_QUIRK_UART1_TTY_UART2_SKIP) { > + if (!strcmp(adev->pnp.unique_id, "1")) > + return -ENODEV; /* Create tty cdev instead of serdev */ > + > + if (!strcmp(adev->pnp.unique_id, "2")) > + *skip = true; > + } > + > + return 0; > +} > +EXPORT_SYMBOL_GPL(acpi_quirk_skip_serdev_enumeration); > +#endif > diff --git a/include/acpi/acpi_bus.h b/include/acpi/acpi_bus.h > index 5895f6c7f6db..102b1cf433c7 100644 > --- a/include/acpi/acpi_bus.h > +++ b/include/acpi/acpi_bus.h > @@ -624,6 +624,22 @@ static inline bool acpi_device_override_status(struct acpi_device *adev, > } > #endif > > +#if IS_ENABLED(CONFIG_X86_ANDROID_TABLETS) > +bool acpi_quirk_skip_i2c_client_enumeration(struct acpi_device *adev); > +int acpi_quirk_skip_serdev_enumeration(struct device *controller_parent, bool *skip); > +#else > +static inline bool acpi_quirk_skip_i2c_client_enumeration(struct acpi_device *adev) > +{ > + return false; > +} > +static inline int > +acpi_quirk_skip_serdev_enumeration(struct device *controller_parent, bool *skip) > +{ > + *skip = false; > + return 0; > +} > +#endif > + > #ifdef CONFIG_PM > void acpi_pm_wakeup_event(struct device *dev); > acpi_status acpi_add_pm_notifier(struct acpi_device *adev, struct device *dev, > -- > 2.33.1 >
Hi Rafael,
> or please let me know if you want me to take this series.
Yes, see cover-letter :)
All the best,
Wolfram
Hi, On 12/30/21 19:00, Rafael J. Wysocki wrote: > On Thu, Dec 30, 2021 at 6:56 PM Wolfram Sang <wsa@kernel.org> wrote: >> >> Hi Rafael, >> >>> or please let me know if you want me to take this series. >> >> Yes, see cover-letter :) > > Oh well. > > I'll take it now, thanks! Thank you. I see this in your bleeding-edge branch now. I was wondering what this means wrt this making it into 5.17, or if you consider this 5.18 material? Note 5.18 is fine, just wondering (mostly for the timing of merging the pdx86 side of this, which probably also is more 5.18 material). Regards, Hans
Hi, On 12/30/21 19:19, Rafael J. Wysocki wrote: > On Thu, Dec 30, 2021 at 7:18 PM Hans de Goede <hdegoede@redhat.com> wrote: >> >> Hi, >> >> On 12/30/21 19:00, Rafael J. Wysocki wrote: >>> On Thu, Dec 30, 2021 at 6:56 PM Wolfram Sang <wsa@kernel.org> wrote: >>>> >>>> Hi Rafael, >>>> >>>>> or please let me know if you want me to take this series. >>>> >>>> Yes, see cover-letter :) >>> >>> Oh well. >>> >>> I'll take it now, thanks! >> >> Thank you. >> >> I see this in your bleeding-edge branch now. I was wondering what this >> means wrt this making it into 5.17, or if you consider this 5.18 >> material? > > 5.17 Nice, then it probably is best to also get the matching x86-android-tablets changes merged for 5.17. I'll give it a couple of days for people to review those and then merge those into pdx86/for-next. Regards, Hans
Hi All, Here is v2 of my patch-series to deal with x86 Android tablets with broken DSDTs. The only change in v2 is moving the i2c_acpi_known_good_ids[] list and check to drivers/acpi/x86/utils.c. Since the pdx86 patches are unchanged I'm only sending out the ACPI core changes this time. Since the i2c and serdev ACPI enumeration changes are very small and depend on patch 1, I believe it would be best for all 3 ACPI patches to be merged through Rafael's ACPI tree. Rafael, acks have already been given for merging all patches through the ACPI tree by Mika, Wolfram and Greg. For some more background info here is the v1 cover-letter again: As a small(ish) hoppy project over the holidays I've been looking into getting some (somewhat older) x86 tablets which ship with Android as the only OS on their factory image working with the mainline kernel. These typically have pretty broken DSDTs since the Android image kernel just has everything hardcoded. This patch-series makes most things on 3 of these tablets work with the mainline kernel and lays the groundwork for adding support for similar tablets. Since the ACPI tables on these devices clearly are buggy this series is written so as to add minimal changes to the ACPI core code, leaving all of the heavy lifting to the recently introduced (in linux-next) drivers/platform/x86/x86-android-tablets.c module, which when built as a module only autoloads on affected devices based on DMI matching. And when this module is disabled the added acpi_quirk_skip_*_enumeration() helpers are replaced by inline stubs and even the minimally added core code will be optimized away. Regards, Hans Hans de Goede (3): ACPI / x86: Add acpi_quirk_skip_[i2c_client|serdev]_enumeration() helpers i2c: acpi: Do not instantiate I2C-clients on boards with known bogus DSDT entries serdev: Do not instantiate serdevs on boards with known bogus DSDT entries drivers/acpi/x86/utils.c | 111 ++++++++++++++++++++++++++++++++++++ drivers/i2c/i2c-core-acpi.c | 7 +++ drivers/tty/serdev/core.c | 14 +++++ include/acpi/acpi_bus.h | 16 ++++++ 4 files changed, 148 insertions(+)