On Tue, Nov 10, 2015 at 12:13 AM, Dmitry Torokhov
[off-list ref] wrote:

On Thu, Nov 05, 2015 at 03:36:56PM -0800, Andrew Duggan wrote:

quoted

From: Benjamin Tissoires <redacted>

When .unified_input is set to true in the platform data, the
functions should rely on the common input node created by rmi_driver
to forward events instead of having their own input node.

This node is named "Synaptics PRODUCT_ID" to be able to
differentiate the various models.

Signed-off-by: Benjamin Tissoires <redacted>
Tested-by: Andrew Duggan <redacted>
---
 drivers/input/rmi4/rmi_driver.c | 43 +++++++++++++++++++++++++++++++++++++++++
 drivers/input/rmi4/rmi_driver.h |  6 ++++++
 drivers/input/rmi4/rmi_f01.c    |  7 +++++++
 include/linux/rmi.h             |  2 ++
 4 files changed, 58 insertions(+)

diff --git a/drivers/input/rmi4/rmi_driver.c b/drivers/input/rmi4/rmi_driver.c
index b9db709..95f9386 100644
--- a/drivers/input/rmi4/rmi_driver.c
+++ b/drivers/input/rmi4/rmi_driver.c

@@ -310,6 +310,9 @@ static int process_interrupt_requests(struct rmi_device *rmi_dev)
              if (entry->irq_mask)
                      process_one_interrupt(data, entry);

+     if (data->input)
+             input_sync(data->input);
+
      return 0;
 }

@@ -330,6 +333,25 @@ static int rmi_driver_set_input_params(struct rmi_device *rmi_dev,
      return 0;
 }

+static void rmi_driver_set_input_name(struct rmi_device *rmi_dev,
+                             struct input_dev *input)
+{
+     struct rmi_driver_data *data = dev_get_drvdata(&rmi_dev->dev);
+     char *device_name = rmi_f01_get_product_ID(data->f01_container);
+     char *name;
+
+     if (!device_name)
+             return;

Should we still give device some name?

Actually, I think the test is need needed. rmi_f01_get_product_ID()
returns a pointer to an array embedded in struct f01_data.
rmi_driver_set_input_name() is called after f01 has been initialized
and I think if there is an error while processing f01 we bail out. So
basically device_name can not be null. (If I read the code correctly).

quoted

+
+     name = devm_kasprintf(&rmi_dev->dev, GFP_KERNEL,
+                           "Synaptics %s", device_name);
+     if (!name)
+             return;

Are we guaranteed that devm_kasprintf only called in probe() path?

Right now, yes. This function is called only in probe.

What about errors?

Indeed, I should have tackled these first :(

quoted

+
+     input->name = name;
+}
+
+
 static int rmi_driver_set_irq_bits(struct rmi_device *rmi_dev,
                                 unsigned long *mask)
 {

@@ -720,6 +742,8 @@ static int rmi_driver_remove(struct device *dev)
      const struct rmi_device_platform_data *pdata =
                                      rmi_get_platform_data(rmi_dev);

+     if (data->input)
+             input_unregister_device(data->input);

Isn't this too early? Can sensor still be sending data here? (Looking...
yeah, it looks like it can. Bad.)

oops, my bad.

quoted

      disable_sensor(rmi_dev);
      rmi_free_function_list(rmi_dev);

@@ -832,6 +856,15 @@ static int rmi_driver_probe(struct device *dev)
      data->current_irq_mask  = irq_memory + size * 2;
      data->new_irq_mask      = irq_memory + size * 3;

+     if (pdata->unified_input) {
+             data->input = input_allocate_device();
+             if (data->input) {
+                     rmi_driver_set_input_params(rmi_dev, data->input);
+                     sprintf(data->input_phys, "%s/input0", dev_name(dev));
+                     data->input->phys = data->input_phys;
+             }

Bail if error?

yes, we need to change it

quoted

+     }
+
      irq_count = 0;
      dev_dbg(dev, "Creating functions.");
      retval = rmi_scan_pdt(rmi_dev, &irq_count, rmi_create_function);

@@ -866,6 +899,15 @@ static int rmi_driver_probe(struct device *dev)
              mutex_init(&data->suspend_mutex);
      }

+     if (data->input) {
+             rmi_driver_set_input_name(rmi_dev, data->input);
+             if (input_register_device(data->input)) {
+                     dev_err(dev, "%s: Failed to register input device.\n",
+                             __func__);
+                     goto err_destroy_functions;
+             }
+     }
+
      if (gpio_is_valid(pdata->attn_gpio)) {
              static const char GPIO_LABEL[] = "attn";
              unsigned long gpio_flags = GPIOF_DIR_IN;

@@ -921,6 +963,7 @@ static int rmi_driver_probe(struct device *dev)
      return 0;

 err_destroy_functions:
+     input_free_device(data->input);
      rmi_free_function_list(rmi_dev);
      kfree(irq_memory);
 err_free_mem:

diff --git a/drivers/input/rmi4/rmi_driver.h b/drivers/input/rmi4/rmi_driver.h
index dda564f..36ca34b 100644
--- a/drivers/input/rmi4/rmi_driver.h
+++ b/drivers/input/rmi4/rmi_driver.h

@@ -13,6 +13,7 @@
 #include <linux/ctype.h>
 #include <linux/hrtimer.h>
 #include <linux/ktime.h>
+#include <linux/input.h>
 #include "rmi_bus.h"

 #define RMI_DRIVER_VERSION "1.6"

@@ -29,6 +30,8 @@

 #define RMI_PDT_PROPS_HAS_BSR 0x02

+#define NAME_BUFFER_SIZE 256
+
 struct rmi_driver_data {
      struct list_head function_list;

@@ -49,6 +52,8 @@ struct rmi_driver_data {
      unsigned long *current_irq_mask;
      unsigned long *new_irq_mask;
      struct mutex irq_mutex;
+     struct input_dev *input;
+     char input_phys[NAME_BUFFER_SIZE];

      /* Following are used when polling. */
      struct hrtimer poll_timer;

@@ -112,6 +117,7 @@ void rmi_unregister_physical_driver(void);

 int rmi_register_f01_handler(void);
 void rmi_unregister_f01_handler(void);
+char *rmi_f01_get_product_ID(struct rmi_function *fn);

 #ifdef CONFIG_RMI4_F11
 int rmi_register_f11_handler(void);

diff --git a/drivers/input/rmi4/rmi_f01.c b/drivers/input/rmi4/rmi_f01.c
index ee5f4a1..2d72dc8 100644
--- a/drivers/input/rmi4/rmi_f01.c
+++ b/drivers/input/rmi4/rmi_f01.c

@@ -176,6 +176,13 @@ static int rmi_f01_read_properties(struct rmi_device *rmi_dev,
      return 0;
 }

+char *rmi_f01_get_product_ID(struct rmi_function *fn)
+{
+     struct f01_data *f01 = dev_get_drvdata(&fn->dev);
+
+     return f01->properties.product_id;
+}

Should we have something like rmi_to_input_id() that is similar to
usb_to_input_id()?

Works for me.

quoted

+
 static int rmi_f01_probe(struct rmi_function *fn)
 {
      struct rmi_device *rmi_dev = fn->rmi_dev;

diff --git a/include/linux/rmi.h b/include/linux/rmi.h
index ca35b2f..1d22985 100644
--- a/include/linux/rmi.h
+++ b/include/linux/rmi.h

@@ -277,6 +277,8 @@ struct rmi_device_platform_data {
      struct rmi_f30_gpioled_map *gpioled_map;
      struct rmi_button_map *f41_button_map;

+     bool unified_input;
+
 #ifdef CONFIG_RMI4_FWLIB
      char *firmware_name;
 #endif
--

2.1.4

Thanks.

Thanks for the review!

Cheers,
Benjamin