On Wednesday 19 December 2012 at 09:41:13AM, Henrik Rydberg has written:
> Hi Christophe,
>
> > Fix horizontal two-finger scrolling on Sentelic touchpads. Currently
> > horizontal two-fingers scrolling does not work with these touchpads.
> > The patch also makes vertical two-finger scrolling smoother in some
> > applications e.g. Firefox.
> >
> > Signed-off-by: Christophe TORDEUX <christophe@xxxxxxxxxxx>
> > ----
> >
> > This patch was inspired by the reading of
> > drivers/input/mouse/synaptics.c. It is known to work (tested) on
> > Sentelic touchpads version STL3888_C0. Very probably works on all later
> > versions, and possibly works on some earlier versions of the hardware in
> > question.
>
> Thanks for the patch. I would prefer it if this driver was converted
> to use input_mt_assign_slots() instead.
Thanks for you feedback.
As for converting the driver to use input_mt_assign_slot, I can try,
and I have the device, give me just a couple of weeks.
I guess a measure of complete success would be to have the device able
to function in userspace purely based on multitouch events, with
non-multitouch absolute events being kept simply as a legacy feature,
in case one needs to use input userspace drivers like evdev which are
not specifically touchpad/touchscreen drivers.
However I'm not sure the synaptics xorg driver is fit to achieve this.
With this userspace driver as it stands today, a few more or less
functional multitouch features are achieved because the synaptics xorg
driver interprets a mixture of MT and non-MT absolute events. The
sequence in which MT and non-MT events arrive matters.
> Why not save the variables instead of the array?
> Relying on the box (x1, y1, max(x1, x2), max(y1, y2)) seems to assume
> a lot about the behavior of the device.
Well, as for your other objections, I guess I would agree today we
don't have to report touches which are (potentially) not located where
the physical fingers were detected, like I was doing in this early
attempt. Besides, saving data from the device's previous packet is not
an absolute necessity.
I'm now going to post a much simpler patch, and probably cleaner. The
way I see it this patch will have a double purpose:
1) fix horizontal and improve vertical scrolling, though maybe in a
coincidental way and 2) work towards making the non-MT absolute events
a purely legacy single finger fallback mode, without breaking anything.
I'll post my new patch in something like an hour.
> >
> > diff -uprN -X vanilla/linux-3.7-rc8/Documentation/dontdiff
> > vanilla/linux-3.7-rc8/drivers/input/mouse/sentelic.c
> > linux-3.7-rc8/drivers/input/mouse/sentelic.c
> > --- vanilla/linux-3.7-rc8/drivers/input/mouse/sentelic.c 2012-12-11 04:32:44.476930522 +0100
> > +++ linux-3.7-rc8/drivers/input/mouse/sentelic.c 2012-12-12 01:16:41.766018017 +0100
> > @@ -706,8 +706,10 @@ static psmouse_ret_t fsp_process_byte(st
> > struct input_dev *dev = psmouse->dev;
> > struct fsp_data *ad = psmouse->private;
> > unsigned char *packet = psmouse->packet;
> > + unsigned char *last_packet = ad->last_pkt;
> > unsigned char button_status = 0, lscroll = 0, rscroll = 0;
> > unsigned short abs_x, abs_y, fgrs = 0;
> > + unsigned short last_abs_x, last_abs_y;
> > int rel_x, rel_y;
> >
> > if (psmouse->pktcnt < 4)
> > @@ -734,6 +736,9 @@ static psmouse_ret_t fsp_process_byte(st
> >
> > abs_x = GET_ABS_X(packet);
> > abs_y = GET_ABS_Y(packet);
> > + last_abs_x = GET_ABS_X(last_packet);
> > + last_abs_y = GET_ABS_Y(last_packet);
> > + memcpy(ad->last_pkt, packet, psmouse->pktcnt);
>
> >
> > if (packet[0] & FSP_PB0_MFMC) {
> > /*
> > @@ -753,10 +758,19 @@ static psmouse_ret_t fsp_process_byte(st
> > */
> > fgrs = 1;
> > fsp_set_slot(dev, 0, false, 0, 0);
> > +
> > + /* We don't need to re-order
> > + * coordinates if last finger
> > + * was same finger
> > + */
> > + last_abs_x = abs_x;
> > + last_abs_y = abs_y;
> > }
> > ad->last_mt_fgr = 2;
> >
> > - fsp_set_slot(dev, 1, fgrs == 2, abs_x, abs_y);
> > + fsp_set_slot(dev, 1, fgrs == 2,
> > + max(abs_x, last_abs_x),
> > + max(abs_y, last_abs_y));
>
>
> > } else {
> > /* 1st finger */
> > if (ad->last_mt_fgr == 1) {
> > @@ -767,9 +781,18 @@ static psmouse_ret_t fsp_process_byte(st
> > */
> > fgrs = 1;
> > fsp_set_slot(dev, 1, false, 0, 0);
> > +
> > + /* We don't need to re-order
> > + * coordinates if last finger
> > + * was same finger
> > + */
> > + last_abs_x = abs_x;
> > + last_abs_y = abs_y;
> > }
> > ad->last_mt_fgr = 1;
> > - fsp_set_slot(dev, 0, fgrs != 0, abs_x, abs_y);
> > + fsp_set_slot(dev, 0, fgrs != 0,
> > + min(abs_x, last_abs_x),
> > + min(abs_y, last_abs_y));
> > }
> > } else {
> > /* SFAC packet */
> > @@ -788,12 +811,18 @@ static psmouse_ret_t fsp_process_byte(st
> > if (abs_x != 0 && abs_y != 0)
> > fgrs = 1;
> >
> > + /* We don't need to re-order
> > + * coordinates of SFAC packets
> > + */
> > + last_abs_x = abs_x;
> > + last_abs_y = abs_y;
> > +
> > fsp_set_slot(dev, 0, fgrs > 0, abs_x, abs_y);
> > fsp_set_slot(dev, 1, false, 0, 0);
> > }
> > if (fgrs > 0) {
> > - input_report_abs(dev, ABS_X, abs_x);
> > - input_report_abs(dev, ABS_Y, abs_y);
> > + input_report_abs(dev, ABS_X, min(abs_x, last_abs_x));
> > + input_report_abs(dev, ABS_Y, min(abs_y, last_abs_y));
>
> Looks like this point could differ from the reported MT positions.
>
> > }
> > input_report_key(dev, BTN_LEFT, packet[0] & 0x01);
> > input_report_key(dev, BTN_RIGHT, packet[0] & 0x02);
> > diff -uprN -X vanilla/linux-3.7-rc8/Documentation/dontdiff vanilla/linux-3.7-rc8/drivers/input/mouse/sentelic.h linux-3.7-rc8/drivers/input/mouse/sentelic.h
> > --- vanilla/linux-3.7-rc8/drivers/input/mouse/sentelic.h 2012-12-11 04:32:44.476930522 +0100
> > +++ linux-3.7-rc8/drivers/input/mouse/sentelic.h 2012-12-12 01:16:41.954018011 +0100
> > @@ -117,6 +117,7 @@ struct fsp_data {
> > unsigned char last_reg; /* Last register we requested read from */
> > unsigned char last_val;
> > unsigned int last_mt_fgr; /* Last seen finger(multitouch) */
> > + unsigned char last_pkt[8]; /* Last packet we processed from fsp */
> > };
> >
> > #ifdef CONFIG_MOUSE_PS2_SENTELIC
>
> Thanks.
> Henrik
>
>
--
Christophe TORDEUX
113 rue de la Barrière
19000 Tulle
FRANCE
mob: +33-6-86-50-83-65
tel: +33-5-55-93-64-35
http://christophe.tordeux.net
Attachment:
signature.asc
Description: Digital signature
