On Wed, Dec 10, 2014 at 6:01 PM, Jason Gerecke <killertofu@xxxxxxxxx> wrote:
> On Wed, Dec 10, 2014 at 12:25 PM, Benjamin Tissoires
> <benjamin.tissoires@xxxxxxxxx> wrote:
>> On Fri, Dec 5, 2014 at 4:37 PM, Jason Gerecke <killertofu@xxxxxxxxx> wrote:
>>> The existing generic touch code only reports events after reading an
>>> entire HID report, which practically means that only data about the last
>>> contact in a report will ever be provided to userspace. This patch uses
>>> a trick from hid-multitouch.c to discover what type of field is at the
>>> end of each contact; when such a field is encountered all the stored
>>> contact data will be reported.
>>>
>>> Signed-off-by: Jason Gerecke <killertofu@xxxxxxxxx>
>>> ---
>>
>> Thanks for the v2 Jason.
>>
>> I think we are closes to it. I have some comments on the style, but
>> now that I have read it more carefully, I have a better idea of it,
>> and the patch is not so scary.
>>
>> Jiri. If you reviewed it and took it, that's fine. If not, then I
>> think there is still room for improvements.
>>
>
> I'm really hoping to get this in for 3.19 if its possible, so I'm
> going to try to get this turned-around today...
>
>> Cheers,
>> Benjamin
>>
>>> Changes from v1:
>>> * Patch renamed from "HID: wacom: Report input events immediately upon receipt"
>>> * Instead of reporting on-reciept, events are bundled and sent at the end of each contact
>>> * Adds code to discover 'last_slot_field' based on hid-multitouch.c implementation
>>>
>>> drivers/hid/wacom_wac.c | 94 ++++++++++++++++++++++++++++++++-----------------
>>> drivers/hid/wacom_wac.h | 1 +
>>> 2 files changed, 63 insertions(+), 32 deletions(-)
>>>
>>> diff --git a/drivers/hid/wacom_wac.c b/drivers/hid/wacom_wac.c
>>> index c33f889..75fc16e 100644
>>> --- a/drivers/hid/wacom_wac.c
>>> +++ b/drivers/hid/wacom_wac.c
>>> @@ -1381,35 +1381,68 @@ static void wacom_wac_finger_usage_mapping(struct hid_device *hdev,
>>> {
>>> struct wacom *wacom = hid_get_drvdata(hdev);
>>> struct wacom_wac *wacom_wac = &wacom->wacom_wac;
>>> - unsigned touch_max = wacom_wac->features.touch_max;
>>
>> not sure we really need to remove the temporary variable touch_max here.
>>
>>> + struct wacom_features *features = &wacom_wac->features;
>>>
>>> switch (usage->hid) {
>>> case HID_GD_X:
>>> - if (touch_max == 1)
>>> + features->last_slot_field = usage->hid;
>>> + if (features->touch_max == 1)
>>
>> these changes (touchmax => features->touch_max) hinder a little bit
>> the actual change, that is worrisome :)
>>
> ACK.
>
>>> wacom_map_usage(wacom, usage, field, EV_ABS, ABS_X, 4);
>>> else
>>> wacom_map_usage(wacom, usage, field, EV_ABS,
>>> ABS_MT_POSITION_X, 4);
>>> break;
>>> case HID_GD_Y:
>>> - if (touch_max == 1)
>>> + features->last_slot_field = usage->hid;
>>> + if (features->touch_max == 1)
>>> wacom_map_usage(wacom, usage, field, EV_ABS, ABS_Y, 4);
>>> else
>>> wacom_map_usage(wacom, usage, field, EV_ABS,
>>> ABS_MT_POSITION_Y, 4);
>>> break;
>>> case HID_DG_CONTACTID:
>>> + features->last_slot_field = usage->hid;
>>> break;
>>> case HID_DG_INRANGE:
>>> + features->last_slot_field = usage->hid;
>>> break;
>>> case HID_DG_INVERT:
>>> + features->last_slot_field = usage->hid;
>>> break;
>>> case HID_DG_TIPSWITCH:
>>> + features->last_slot_field = usage->hid;
>>> wacom_map_usage(wacom, usage, field, EV_KEY, BTN_TOUCH, 0);
>>> break;
>>> }
>>> }
>>>
>>> +static void wacom_wac_finger_slot(struct wacom_wac *wacom_wac,
>>> + struct input_dev *input)
>>> +{
>>> + struct hid_data *hid_data = &wacom_wac->hid_data;
>>> + bool mt = wacom_wac->features.touch_max > 1;
>>> + bool prox = hid_data->tipswitch &&
>>> + !wacom_wac->shared->stylus_in_proximity;
>>> +
>>> + if (mt) {
>>
>> Personal taste. I don't like the "if (mt)" approach in this patch. I
>> preferred the old approach:
>> if (mt)
>> wacom_wac_finger_mt_slot()
>> else
>> wacom_wac_finger_touch()
>>
>> i.e. one function per case.
>>
> I don't really like the idea of breaking this into two functions. The
> differences between how a each MT contact is reported and how the sole
> ST contact is reported are trivial, and I feel that breaking them
> apart leads to unnecessary duplication. I'll split it out if you feel
> that is warranted though.
No, do whatever you like. It was more a comment I added before reading
the entire patch, so after having the global view, it's fine this way
too.
>
>>> + int slot;
>>> +
>>> + slot = input_mt_get_slot_by_key(input, hid_data->id);
>>> + input_mt_slot(input, slot);
>>> + input_mt_report_slot_state(input, MT_TOOL_FINGER, prox);
>>> + }
>>> + else {
>>> + input_report_key(input, BTN_TOUCH, prox);
>>> + }
>>> +
>>> + if (prox) {
>>> + input_report_abs(input, mt ? ABS_MT_POSITION_X : ABS_X,
>>> + hid_data->x);
>>> + input_report_abs(input, mt ? ABS_MT_POSITION_Y : ABS_Y,
>>> + hid_data->y);
>>> + }
>>> +}
>>> +
>>> static int wacom_wac_finger_event(struct hid_device *hdev,
>>> struct hid_field *field, struct hid_usage *usage, __s32 value)
>>> {
>>> @@ -1432,36 +1465,35 @@ static int wacom_wac_finger_event(struct hid_device *hdev,
>>> }
>>>
>>>
>>> + if (usage->usage_index + 1 == field->report_count) {
>>> + if (usage->hid == wacom_wac->features.last_slot_field)
>>
>> OK, I think it would be fair to assume that the FW guys will send N
>> full touches by report. IIRC, I had the case where sometimes they add
>> some random axis in the end (like X or Y), just to make me cry :)
>> should be good here though (and we can have access to the FW guys if
>> they do something too much funny :-P )
>>
> That explains why the hid-multitouch code seemed so much more
> convoluted than necessary :) Our FW team hasn't made any descriptors
> which are that pathological, and I certainly hope they don't start.
Actually, I may not be fair with the FW guys. I think the problem we
had was that while we were parsing the multitouch report and then the
mouse emulation report, we ended up in having ABS_Y as the last slot
field. Either we (I) were (was) not smart enough to check on the type
of the report, or either it was difficult to detect in input_mapping
which collection was the mouse, and which was the touch (and this is
when the FW guys go nuts and do anything in the report descriptor :-P
)
>
>>> + wacom_wac_finger_slot(wacom_wac, wacom_wac->input);
>>> + }
>>> +
>>> return 0;
>>> }
>>>
>>> -static void wacom_wac_finger_mt_report(struct wacom_wac *wacom_wac,
>>> - struct input_dev *input, bool touch)
>>> +static int wacom_wac_finger_touches(struct hid_device *hdev)
>>
>> It took me a while to understand what this function was. It simply
>> returns the number of touches that has been recorded in the report (I
>> was mislead by the git diff I think).
> Almost correct -- it returns the number of touches that are currently
> active (which may be larger than the number of touches recorded in the
> report).
Yeah, that's what I meant, you understood me :)
>
>> We should rename this in something else
>> (wacom_wac_finger_get_touches_count or something better - I am bad at
>> naming things).
>>
> How about 'wacom_wac_finger_count_touches'?
works for me.
>
>> Second question, why the need to return the touch count?
>>
> This function is used only to update the (boolean) value of
> 'wacom_wac->shared->touch_down'. Strictly speaking it could save a
> small amount of processing time and return a boolean directly, but it
> seemed trivial enough to just return the count in case it was useful
> in the future.
>
> If the question was more of "why do we need to bother going through
> the slot state?" then the answer is that we can't rely on the
> information in any single report. Many of our MT devices use what
> Microsoft calls "hybrid mode", so the complete sensor state may span
> multiple reports. All the fingers in one report may go up even while
> other fingers remain on the tablet.
I like when you are able to decrypt my mind :)
ok, fine.
We could still abort earlier in the function to report a boolean, but
the finger count is fine too.
>
>>> {
>>> - int slot;
>>> - struct hid_data *hid_data = &wacom_wac->hid_data;
>>> + struct wacom *wacom = hid_get_drvdata(hdev);
>>> + struct wacom_wac *wacom_wac = &wacom->wacom_wac;
>>> + struct input_dev *input = wacom_wac->input;
>>> + unsigned touch_max = wacom_wac->features.touch_max;
>>> + int count = 0;
>>> + int i;
>>>
>>> - slot = input_mt_get_slot_by_key(input, hid_data->id);
>>> + if (touch_max == 1)
>>> + return wacom_wac->hid_data.tipswitch &&
>>> + !wacom_wac->shared->stylus_in_proximity;
>>>
>>> - input_mt_slot(input, slot);
>>> - input_mt_report_slot_state(input, MT_TOOL_FINGER, touch);
>>> - if (touch) {
>>> - input_report_abs(input, ABS_MT_POSITION_X, hid_data->x);
>>> - input_report_abs(input, ABS_MT_POSITION_Y, hid_data->y);
>>> + for (i = 0; i < input->mt->num_slots; i++) {
>>> + struct input_mt_slot *ps = &input->mt->slots[i];
>>> + int id = input_mt_get_value(ps, ABS_MT_TRACKING_ID);
>>> + if (id >= 0)
>>> + count++;
>>> }
>>> - input_mt_sync_frame(input);
>>> -}
>>> -
>>> -static void wacom_wac_finger_single_touch_report(struct wacom_wac *wacom_wac,
>>> - struct input_dev *input, bool touch)
>>> -{
>>> - struct hid_data *hid_data = &wacom_wac->hid_data;
>>>
>>> - if (touch) {
>>> - input_report_abs(input, ABS_X, hid_data->x);
>>> - input_report_abs(input, ABS_Y, hid_data->y);
>>> - }
>>> - input_report_key(input, BTN_TOUCH, touch);
>>> + return count;
>>> }
>>>
>>> static void wacom_wac_finger_report(struct hid_device *hdev,
>>> @@ -1470,18 +1502,16 @@ static void wacom_wac_finger_report(struct hid_device *hdev,
>>> struct wacom *wacom = hid_get_drvdata(hdev);
>>> struct wacom_wac *wacom_wac = &wacom->wacom_wac;
>>> struct input_dev *input = wacom_wac->input;
>>> - bool touch = wacom_wac->hid_data.tipswitch &&
>>> - !wacom_wac->shared->stylus_in_proximity;
>>> - unsigned touch_max = wacom_wac->features.touch_max;
>>>
>>> - if (touch_max > 1)
>>> - wacom_wac_finger_mt_report(wacom_wac, input, touch);
>>> + if (wacom_wac->features.touch_max > 1)
>>
>> I think removing the touch_max variable hides a little bit the changes too.
>>
>>> + input_mt_sync_frame(input);
>>> else
>>> - wacom_wac_finger_single_touch_report(wacom_wac, input, touch);
>>> + wacom_wac_finger_slot(wacom_wac, input);
>>
>> Not sure there is a need to call wacom_wac_finger_slot() for single
>> touch devices. Your current implementation in wacom_finger_event
>> should match also single finger tablets.
>>
> I'll double-check to make sure. I hadn't considered that even the ST
> case might match in 'wacom_finger_event'.
thanks
>
>>> +
>>> input_sync(input);
>>>
>>> /* keep touch state for pen event */
>>> - wacom_wac->shared->touch_down = touch;
>>> + wacom_wac->shared->touch_down = wacom_wac_finger_touches(hdev);
>>> }
>>>
>>> #define WACOM_PEN_FIELD(f) (((f)->logical == HID_DG_STYLUS) || \
>>> diff --git a/drivers/hid/wacom_wac.h b/drivers/hid/wacom_wac.h
>>> index 128cbb3..6e256206 100644
>>> --- a/drivers/hid/wacom_wac.h
>>> +++ b/drivers/hid/wacom_wac.h
>>> @@ -144,6 +144,7 @@ struct wacom_features {
>>> int pktlen;
>>> bool check_for_hid_type;
>>> int hid_type;
>>> + int last_slot_field;
>>> };
>>>
>>> struct wacom_shared {
>>> --
>>> 2.1.3
>>>
Cheers,
Benjamin
--
To unsubscribe from this list: send the line "unsubscribe linux-input" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
