Message ID | 20161112191545.GA19531@dtor-ws (mailing list archive) |
---|---|
State | New, archived |
Headers | show |
On Saturday 12 November 2016 20:15:45 Dmitry Torokhov wrote: > BYD automatic protocol detection is extremely unreliable and is often > triggers false positives on regular mice, Sentelic touchpads, and > other devices. BYD has several documents that have recommended > detection sequence, but they conflict with each other and, as far as > I can see, still would not produce unique enough output to reliably > differentiate BYD from other PS/2 devices. > > OEMs sourcing BYD devices also do not do us any favors by not > supplying any reasonable DMI data and instead leaving turds like "To > Be Filled By O.E.M." in place of vendor data, or "System Serial > Number" as serial number. ... > echo -n "byd" > /sys/bus/serio/devices/serio1/drvctl I agree, if we cannot detect BYD devices correctly without breaking other non-BYD devices then detection must be done by user... So you can add my Reviewed-by: Pali Rohár to this patch. Anyway, BYD show Linux as supported OS for their devices. It is really rude from BYD that they show Linux as supported OS, but Linux driver is reverse-engineered by community and they do not want to cooperate and at least provide some useful documentation for detection...
On Sat, Nov 12, 2016 at 11:39 AM, Pali Rohár <pali.rohar@gmail.com> wrote: > On Saturday 12 November 2016 20:15:45 Dmitry Torokhov wrote: >> BYD automatic protocol detection is extremely unreliable and is often >> triggers false positives on regular mice, Sentelic touchpads, and >> other devices. BYD has several documents that have recommended >> detection sequence, but they conflict with each other and, as far as >> I can see, still would not produce unique enough output to reliably >> differentiate BYD from other PS/2 devices. >> >> OEMs sourcing BYD devices also do not do us any favors by not >> supplying any reasonable DMI data and instead leaving turds like "To >> Be Filled By O.E.M." in place of vendor data, or "System Serial >> Number" as serial number. > ... >> echo -n "byd" > /sys/bus/serio/devices/serio1/drvctl > > I agree, if we cannot detect BYD devices correctly without breaking > other non-BYD devices then detection must be done by user... > > So you can add my Reviewed-by: Pali Rohár to this patch. > > Anyway, BYD show Linux as supported OS for their devices. It is really > rude from BYD that they show Linux as supported OS, but Linux driver is > reverse-engineered by community and they do not want to cooperate and at > least provide some useful documentation for detection... I am not even sure they have a good way of detecting their devices. The Windows model is that vendor supplies OEM with their driver and it just works (I do not think these devices have external PS/2 ports so there is no concern with mis-detection). Thanks.
diff --git a/drivers/input/mouse/psmouse-base.c b/drivers/input/mouse/psmouse-base.c index fb4b185..bee2674 100644 --- a/drivers/input/mouse/psmouse-base.c +++ b/drivers/input/mouse/psmouse-base.c @@ -1115,10 +1115,6 @@ static int psmouse_extensions(struct psmouse *psmouse, if (psmouse_try_protocol(psmouse, PSMOUSE_TOUCHKIT_PS2, &max_proto, set_properties, true)) return PSMOUSE_TOUCHKIT_PS2; - - if (psmouse_try_protocol(psmouse, PSMOUSE_BYD, - &max_proto, set_properties, true)) - return PSMOUSE_BYD; } /*
BYD automatic protocol detection is extremely unreliable and is often triggers false positives on regular mice, Sentelic touchpads, and other devices. BYD has several documents that have recommended detection sequence, but they conflict with each other and, as far as I can see, still would not produce unique enough output to reliably differentiate BYD from other PS/2 devices. OEMs sourcing BYD devices also do not do us any favors by not supplying any reasonable DMI data and instead leaving turds like "To Be Filled By O.E.M." in place of vendor data, or "System Serial Number" as serial number. On top of that BYD is not truly modern multitouch controller, but rather a single-touch transitional device that only reports absolute coordinates at the beginning of finger contact and then reverts to reporting displacements, and thus not very precise; the only benefit from using BYD mode vs the legacy PS/2 mode is possibility of edge scrolling. Given the above, and the fact that BYD devices are somewhat uncommon, let's disable automatic detection of BYD devices. Users who know they have BYD trackpads or want to experiment can attempt to activate BYD protocol via sysfs: echo -n "byd" > /sys/bus/serio/devices/serio1/drvctl Fixes: https://bugzilla.kernel.org/show_bug.cgi?id=151691 Fixes: https://bugzilla.kernel.org/show_bug.cgi?id=175421 Fixes: https://bugzilla.kernel.org/show_bug.cgi?id=120781 Fixes: https://bugzilla.kernel.org/show_bug.cgi?id=121281 Cc: stable@vger.kernel.org Signed-off-by: Dmitry Torokhov <dmitry.torokhov@gmail.com> --- drivers/input/mouse/psmouse-base.c | 4 ---- 1 file changed, 4 deletions(-)