| lists.openwall.net | lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC | |
|
Open Source and information security mailing list archives
| ||
|
Message-ID: <s5hzkuk1t22.wl%tiwai@suse.de>
Date: Mon, 11 Oct 2010 20:29:57 +0200
From: Takashi Iwai <tiwai@...e.de>
To: Chris Bagwell <chris@...bagwell.com>
Cc: Chase Douglas <chase.douglas@...onical.com>,
linux-input@...r.kernel.org, xorg-devel@...ts.x.org,
Dmitry Torokhov <dmitry.torokhov@...il.com>,
Andy Whitcroft <apw@...onical.com>,
Henrik Rydberg <rydberg@...omail.se>,
linux-kernel@...r.kernel.org,
Peter Hutterer <peter.hutterer@...-t.net>,
Duncan McGreggor <duncan.mcgreggor@...onical.com>
Subject: Re: [PATCH 3/3] Input: synaptics - remove touches over button click area
At Mon, 11 Oct 2010 12:46:40 -0500,
Chris Bagwell wrote:
>
> On Mon, Oct 11, 2010 at 12:10 PM, Takashi Iwai <tiwai@...e.de> wrote:
> > At Mon, 11 Oct 2010 11:24:04 -0500,
> > Chris Bagwell wrote:
> >>
> >> On Fri, Oct 8, 2010 at 9:58 AM, Chase Douglas
> >> <chase.douglas@...onical.com> wrote:
> >> > Now that we have proper multitouch support, we can handle integrated
> >> > buttons better. If we know the top of the buttons on the touchpad, we
> >> > can ignore any touches that occur within the touchpad area while a
> >> > button is clicked. It may be possible to get the button area by querying
> >> > the device, but for now allow the user to manually set it.
> >> >
> >> > A note on why this works: the Synaptics touchpads have pseudo touch
> >> > tracking. When two touches are on the touchpad, an MT touch packet with
> >> > just the X, Y, and pressure values is sent before a normal Synaptics
> >> > touch packet. When one touch is obviously in motion and the other is
> >> > stationary, the touchpad controller sends the touch in motion in the
> >> > normal packet and the stationary touch in the MT packet. Single touch
> >> > emulation is provided by the normal packet, so an action like clicking
> >> > a button and dragging with another finger still works as expected.
> >> >
> >> > Tested on a Dell Mini 1012 with synaptics_multitouch=1 and
> >> > synaptics_button_thresh=4100.
> >> >
> >> > Signed-off-by: Chase Douglas <chase.douglas@...onical.com>
> >> > ---
> >> > drivers/input/mouse/synaptics.c | 16 +++++++++++++++-
> >> > 1 files changed, 15 insertions(+), 1 deletions(-)
> >> >
> >> > diff --git a/drivers/input/mouse/synaptics.c b/drivers/input/mouse/synaptics.c
> >> > index 7289d88..e67778d 100644
> >> > --- a/drivers/input/mouse/synaptics.c
> >> > +++ b/drivers/input/mouse/synaptics.c
> >> > @@ -49,6 +49,12 @@ module_param(synaptics_multitouch, bool, 0644);
> >> > MODULE_PARM_DESC(synaptics_multitouch,
> >> > "Enable multitouch mode on Synaptics devices");
> >> >
> >> > +static int synaptics_button_thresh = YMIN_NOMINAL + YMAX_NOMINAL;
> >> > +module_param(synaptics_button_thresh, int, 0644);
> >> > +MODULE_PARM_DESC(synaptics_button_thres,
> >> > + "Y coordinate threshold of integrated buttons on Synaptics "
> >> > + "devices");
> >> > +
> >> > /*****************************************************************************
> >> > * Stuff we need even when we do not want native Synaptics support
> >> > ****************************************************************************/
> >> > @@ -463,6 +469,10 @@ static void synaptics_parse_hw_state(unsigned char buf[], struct synaptics_data
> >> > }
> >> > }
> >> >
> >> > +#define TOUCH_OVER_BUTTON(hw) (((hw).left || (hw).middle || (hw).right) && \
> >> > + (YMAX_NOMINAL + YMIN_NOMINAL - (hw).y > \
> >> > + synaptics_button_thresh))
> >> > +
> >> > /*
> >> > * called for each full received packet from the touchpad
> >> > */
> >> > @@ -477,7 +487,7 @@ static void synaptics_process_packet(struct psmouse *psmouse)
> >> > synaptics_parse_hw_state(psmouse->packet, priv, &hw);
> >> >
> >> > if (SYN_MULTITOUCH(priv, &hw)) {
> >> > - if (hw.z > 0) {
> >> > + if (hw.z > 0 && !TOUCH_OVER_BUTTON(hw)) {
> >> > input_report_abs(dev, ABS_MT_POSITION_X, hw.x);
> >> > input_report_abs(dev, ABS_MT_POSITION_Y,
> >> > YMAX_NOMINAL + YMIN_NOMINAL - hw.y);
> >> > @@ -509,6 +519,10 @@ static void synaptics_process_packet(struct psmouse *psmouse)
> >> > return;
> >> > }
> >> >
> >> > + /* If touch occurs over depressed button, ignore it */
> >> > + if (TOUCH_OVER_BUTTON(hw))
> >> > + hw.z = 0;
> >> > +
> >> > if (hw.z > 0) {
> >> > priv->num_fingers++;
> >> > finger_width = 5;
> >> > --
> >> > 1.7.1
> >> >
> >> >
> >>
> >> I'm convinced now that clickpad style touchpads can't work without
> >> multi-touch and something like logic in xf86-input-multitouch.
> >
> > Actually Clickpad works without multi-touch patch. With my patches to
> > synaptics, it worked in some level. There are many restrictions (e.g.
> > pushing the button first then drag), though.
> >
>
> True, but if I understand synaptic hardware MT behavior (sends
> actively moving finger in higher resolution packet regardless of
> original finger touch) then your patch will result in jumpy cursor on
> X side and that side would need patches to attempt to guess invalid
> data and discard. I've worked on a few similar patches to various
> xf86-input-* and generally they've failed to detect difference between
> invalid packets vs. fast user movements.
Right. I also tackled to this once.
> The main point of my 3 options was to address jumpy cursor in
> xf86-input-* that are not MT aware. I think ABS_X/ABS_Y should only
> allow its meaning to change at detectable time periods so user can
> account for it and, specifically, that time period is best at
> transition of BTN_TOOL_DOUBLETAP.
>
> Assuming its easy enough to support exact rules for ABS_X/ABS_Y
> changing meanings on kernel side (which I think it probably is pretty
> easy), I think we should do it so that applications don't have to
> become MT-aware as the official solution for jumpy cursors.
Hm, so are you suggesting that we need to transform MT events to
non-MT events in the kernel side with some clever filtering?
I don't know whether this is really needed... If the patch is minimal
amount, it might be worth to get in, though.
However, I feel that we should forget about clickpad support in
kernel; as you mentioned, Clickpad works properly only with the
multi-touch, after all...
Rather I'm concerned how to decide the kernel driver's behavior,
whether behaving in MT or non-MT mode. As Henrik suggested, in MT
mode, all events should be ABS_MT_* type. But non-MT-aware user-space
apps won't understand all these events. So, we'll need some fallback
mechanism for non-MT apps, anyway.
I've implemented multi_touch module option for that purpose, but it's
hacikish. Dmitry, do you have any suggestion?
thanks,
Takashi
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
Powered by blists - more mailing lists