Re: [PATCH 2/2] HID: hid-sensor-hub: Add collection device

On 19/02/15 23:54, Srinivas Pandruvada wrote:

On Wed, 2015-01-21 at 21:53 +0000, Jonathan Cameron wrote:

quoted

On 07/01/15 18:24, Srinivas Pandruvada wrote:

quoted

HID sensor hub exports several sensor which are fusion sensors, where
data is interpreted from one or more sensors. Some of them can't be
exported via IIO like sysfs as the user space download some firmware,
which defines what sensor will look like. They can be a part of a
collection.

This functionality is beginning to turn up on a number of devices.
I'd argue that where at all possible, we should try for standard
interfaces on the resulting 'fused sensors'.  Still you do indeed
always need a means of uploading the firmware!

I will submitted RFCs, once I see in production releases.

Cool

Currently planning to use misc driver for FW update.
What do you think about using dev/iio:deviceX for such case to download
FW for each sensor? We also need mechanism to initiate and reset after
FW download.

I think using the main iio device is pushing things a little.  We could add
the ability to get another anonymous cdev like we do for events I suppose.
That way, at least it will be tied to the normal iio device.

Sort of feels like there should be a generic / standard way of doing
this sort of firmware update, irrespective of what the device does.
I guess, maybe it's always non generic enough that this isn't
possible.

Thanks,
Srinivas

quoted

quoted

Creating a MFD cell for a collection to write a standalone driver
to manage collections. Most of the time they will be propritery drivers
may not be even upstreamed. This patch allows framework to have
capability to write such drivers.

Signed-off-by: Srinivas Pandruvada <srinivas.pandruvada-VuQAYsv1563Yd54FQh9/CA@public.gmane.org>

A few comments inline...

How does userspace download a firmware?

 
 ---

quoted

 drivers/hid/hid-sensor-hub.c   | 50 ++++++++++++++++++++++++++++++++++++++----
 include/linux/hid-sensor-ids.h |  2 ++
 2 files changed, 48 insertions(+), 4 deletions(-)

diff --git a/drivers/hid/hid-sensor-hub.c b/drivers/hid/hid-sensor-hub.c
index 865cd56..83b6e15 100644
--- a/drivers/hid/hid-sensor-hub.c
+++ b/drivers/hid/hid-sensor-hub.c

@@ -119,11 +119,12 @@ static struct hid_sensor_hub_callbacks *sensor_hub_get_callback(
 
 	spin_lock_irqsave(&pdata->dyn_callback_lock, flags);
 	list_for_each_entry(callback, &pdata->dyn_callback_list, list)
-		if (callback->usage_id == usage_id &&
+		if (callback->usage_id == HID_USAGE_SENSOR_TYPE_COLLECTION ||
+			(callback->usage_id == usage_id &&
 			(collection_index >=
 				callback->hsdev->start_collection_index) &&
 			(collection_index <
-				callback->hsdev->end_collection_index)) {
+				callback->hsdev->end_collection_index))) {
 			*priv = callback->priv;
 			*hsdev = callback->hsdev;
 			spin_unlock_irqrestore(&pdata->dyn_callback_lock,

@@ -159,7 +160,12 @@ int sensor_hub_register_callback(struct hid_sensor_hub_device *hsdev,
 	callback->usage_callback = usage_callback;
 	callback->usage_id = usage_id;
 	callback->priv = NULL;
-	list_add_tail(&callback->list, &pdata->dyn_callback_list);
+	/* Give higher priority to collection device, so add to front */

Comment should probably state why we want to give them higher priorities.

quoted

+	if (usage_id == HID_USAGE_SENSOR_TYPE_COLLECTION)
+		list_add(&callback->list, &pdata->dyn_callback_list);
+	else
+		list_add_tail(&callback->list, &pdata->dyn_callback_list);
+
 	spin_unlock_irqrestore(&pdata->dyn_callback_lock, flags);
 
 	return 0;

@@ -547,6 +553,37 @@ static __u8 *sensor_hub_report_fixup(struct hid_device *hdev, __u8 *rdesc,
 	return rdesc;
 }
 
+static int sensor_hub_add_collection_device(struct hid_device *hdev,
+					    struct sensor_hub_data *sd)
+{
+	struct hid_sensor_hub_device *hsdev;
+	char *name;
+
+	hsdev = devm_kzalloc(&hdev->dev, sizeof(*hsdev), GFP_KERNEL);
+	if (!hsdev)
+		return -ENOMEM;
+
+	hsdev->hdev = hdev;
+	hsdev->vendor_id = hdev->vendor;
+	hsdev->product_id = hdev->product;
+	hsdev->usage = HID_USAGE_SENSOR_TYPE_COLLECTION;
+	mutex_init(&hsdev->mutex);
+	name = devm_kasprintf(&hdev->dev, GFP_KERNEL, "HID-SENSOR-%x",
+			      HID_USAGE_SENSOR_TYPE_COLLECTION);
+	if (name == NULL) {
+		hid_err(hdev, "Failed MFD device name\n");
+		return -ENOMEM;
+	}
+	sd->hid_sensor_hub_client_devs[sd->hid_sensor_client_cnt].name = name;
+	sd->hid_sensor_hub_client_devs[
+			sd->hid_sensor_client_cnt].platform_data = hsdev;

Odd breaking up of the line...  Perhaps break before .platform_data instead?

quoted

+	sd->hid_sensor_hub_client_devs[
+			sd->hid_sensor_client_cnt].pdata_size = sizeof(*hsdev);
+	sd->hid_sensor_client_cnt++;
+
+	return 0;
+}
+
 static int sensor_hub_probe(struct hid_device *hdev,
 				const struct hid_device_id *id)
 {

@@ -591,7 +628,8 @@ static int sensor_hub_probe(struct hid_device *hdev,
 		ret = -EINVAL;
 		goto err_stop_hw;
 	}
-	sd->hid_sensor_hub_client_devs = devm_kzalloc(&hdev->dev, dev_cnt *
+	sd->hid_sensor_hub_client_devs = devm_kzalloc(&hdev->dev,
+						      (dev_cnt + 1) *
 						      sizeof(struct mfd_cell),
 						      GFP_KERNEL);
 	if (sd->hid_sensor_hub_client_devs == NULL) {

@@ -645,6 +683,10 @@ static int sensor_hub_probe(struct hid_device *hdev,
 	if (last_hsdev)
 		last_hsdev->end_collection_index = i;
 
+	ret = sensor_hub_add_collection_device(hdev, sd);
+	if (ret)
+		goto err_stop_hw;
+
 	ret = mfd_add_hotplug_devices(&hdev->dev,
 			sd->hid_sensor_hub_client_devs,
 			sd->hid_sensor_client_cnt);

diff --git a/include/linux/hid-sensor-ids.h b/include/linux/hid-sensor-ids.h
index 109f0e6..300ffea 100644
--- a/include/linux/hid-sensor-ids.h
+++ b/include/linux/hid-sensor-ids.h

@@ -21,6 +21,8 @@
 
 #define HID_MAX_PHY_DEVICES					0xFF
 
+#define HID_USAGE_SENSOR_TYPE_COLLECTION			0x200001
+
 /* Accel 3D (200073) */
 #define HID_USAGE_SENSOR_ACCEL_3D				0x200073
 #define HID_USAGE_SENSOR_DATA_ACCELERATION			0x200452