| Message ID | 20210521171418.393871-3-hdegoede@redhat.com (mailing list archive) |
|---|---|
| State | Superseded |
| Headers | show |
| Series | iio: accel: bmc150: Add support for yoga's with dual accelerometers with an ACPI HID of DUAL250E | expand |
On Fri, May 21, 2021 at 11:23 PM Hans de Goede <hdegoede@redhat.com> wrote: > On machines with dual accelerometers described in a single ACPI fwnode, > the bmc150_accel_probe() instantiates a second i2c-client for the second > accelerometer. > > A pointer to this manually instantiated second i2c-client is stored > inside the iio_dev's private-data through bmc150_set_second_device(), > so that the i2c-client can be unregistered from bmc150_accel_remove(). > > Before this commit bmc150_set_second_device() took only 1 argument so it > would store the pointer in private-data of the iio_dev belonging to the > manually instantiated i2c-client, leading to the bmc150_accel_remove() > call for the second_dev trying to unregister *itself* while it was > being removed, leading to a deadlock and rmmod hanging. > > Change bmc150_set_second_device() to take 2 arguments: 1. The i2c-client > which is instantiating the second i2c-client for the 2nd accelerometer and > 2. The second-device pointer itself (which also is an i2c-client). > > This will store the second_device pointer in the private data of the > iio_dev belonging to the (ACPI instantiated) i2c-client for the first > accelerometer and will make bmc150_accel_remove() unregister the > second_device i2c-client when called for the first client, > avoiding the deadlock. I would rather call it aux device (at least in the code). The terminology maybe needs more clarification (like the main one in the block with the display panel and aux in the keyboard). If you disagree, ignore this comment. I have no strong opinion here since I don't know the difference between them (accelerometers).
Hi, On 5/22/21 8:06 PM, Andy Shevchenko wrote: > On Fri, May 21, 2021 at 11:23 PM Hans de Goede <hdegoede@redhat.com> wrote: >> On machines with dual accelerometers described in a single ACPI fwnode, >> the bmc150_accel_probe() instantiates a second i2c-client for the second >> accelerometer. >> >> A pointer to this manually instantiated second i2c-client is stored >> inside the iio_dev's private-data through bmc150_set_second_device(), >> so that the i2c-client can be unregistered from bmc150_accel_remove(). >> >> Before this commit bmc150_set_second_device() took only 1 argument so it >> would store the pointer in private-data of the iio_dev belonging to the >> manually instantiated i2c-client, leading to the bmc150_accel_remove() >> call for the second_dev trying to unregister *itself* while it was >> being removed, leading to a deadlock and rmmod hanging. >> >> Change bmc150_set_second_device() to take 2 arguments: 1. The i2c-client >> which is instantiating the second i2c-client for the 2nd accelerometer and >> 2. The second-device pointer itself (which also is an i2c-client). >> >> This will store the second_device pointer in the private data of the >> iio_dev belonging to the (ACPI instantiated) i2c-client for the first >> accelerometer and will make bmc150_accel_remove() unregister the >> second_device i2c-client when called for the first client, >> avoiding the deadlock. > > I would rather call it aux device (at least in the code). The > terminology maybe needs more clarification (like the main one in the > block with the display panel and aux in the keyboard). > > If you disagree, ignore this comment. I have no strong opinion here > since I don't know the difference between them (accelerometers). Thank you for your input, but both sensors are identical, so calling one aux sounds of to me, so lets keep this as is. Regards, Hans
diff --git a/drivers/iio/accel/bmc150-accel-core.c b/drivers/iio/accel/bmc150-accel-core.c index 3a3f67930165..8ff358c37a81 100644 --- a/drivers/iio/accel/bmc150-accel-core.c +++ b/drivers/iio/accel/bmc150-accel-core.c @@ -1815,11 +1815,11 @@ struct i2c_client *bmc150_get_second_device(struct i2c_client *client) } EXPORT_SYMBOL_GPL(bmc150_get_second_device); -void bmc150_set_second_device(struct i2c_client *client) +void bmc150_set_second_device(struct i2c_client *client, struct i2c_client *second_dev) { struct bmc150_accel_data *data = iio_priv(i2c_get_clientdata(client)); - data->second_device = client; + data->second_device = second_dev; } EXPORT_SYMBOL_GPL(bmc150_set_second_device); diff --git a/drivers/iio/accel/bmc150-accel-i2c.c b/drivers/iio/accel/bmc150-accel-i2c.c index 69f709319484..2afaae0294ee 100644 --- a/drivers/iio/accel/bmc150-accel-i2c.c +++ b/drivers/iio/accel/bmc150-accel-i2c.c @@ -70,7 +70,7 @@ static int bmc150_accel_probe(struct i2c_client *client, second_dev = i2c_acpi_new_device(&client->dev, 1, &board_info); if (!IS_ERR(second_dev)) - bmc150_set_second_device(second_dev); + bmc150_set_second_device(client, second_dev); } #endif diff --git a/drivers/iio/accel/bmc150-accel.h b/drivers/iio/accel/bmc150-accel.h index 6024f15b9700..e30c1698f6fb 100644 --- a/drivers/iio/accel/bmc150-accel.h +++ b/drivers/iio/accel/bmc150-accel.h @@ -18,7 +18,7 @@ int bmc150_accel_core_probe(struct device *dev, struct regmap *regmap, int irq, const char *name, bool block_supported); int bmc150_accel_core_remove(struct device *dev); struct i2c_client *bmc150_get_second_device(struct i2c_client *second_device); -void bmc150_set_second_device(struct i2c_client *second_device); +void bmc150_set_second_device(struct i2c_client *client, struct i2c_client *second_dev); extern const struct dev_pm_ops bmc150_accel_pm_ops; extern const struct regmap_config bmc150_regmap_conf;
On machines with dual accelerometers described in a single ACPI fwnode, the bmc150_accel_probe() instantiates a second i2c-client for the second accelerometer. A pointer to this manually instantiated second i2c-client is stored inside the iio_dev's private-data through bmc150_set_second_device(), so that the i2c-client can be unregistered from bmc150_accel_remove(). Before this commit bmc150_set_second_device() took only 1 argument so it would store the pointer in private-data of the iio_dev belonging to the manually instantiated i2c-client, leading to the bmc150_accel_remove() call for the second_dev trying to unregister *itself* while it was being removed, leading to a deadlock and rmmod hanging. Change bmc150_set_second_device() to take 2 arguments: 1. The i2c-client which is instantiating the second i2c-client for the 2nd accelerometer and 2. The second-device pointer itself (which also is an i2c-client). This will store the second_device pointer in the private data of the iio_dev belonging to the (ACPI instantiated) i2c-client for the first accelerometer and will make bmc150_accel_remove() unregister the second_device i2c-client when called for the first client, avoiding the deadlock. Fixes: 5bfb3a4bd8f6 ("iio: accel: bmc150: Check for a second ACPI device for BOSC0200") Cc: Jeremy Cline <jeremy@jcline.org> Signed-off-by: Hans de Goede <hdegoede@redhat.com> --- drivers/iio/accel/bmc150-accel-core.c | 4 ++-- drivers/iio/accel/bmc150-accel-i2c.c | 2 +- drivers/iio/accel/bmc150-accel.h | 2 +- 3 files changed, 4 insertions(+), 4 deletions(-)