input: zinitix: multitouch broken by e941dc1

[timon37]

On my samsung a3 2015 multitouch doesn't work e.g. making pinch zoom-out in maps impossible (in pure maps this is a pretty serious blocker).

I've debuged it to be a regression caused by commit e941dc1

In my case finger_mask is most definitely not a mask, it's a finger_count. Which is basically irrelevant for the driver, as the fingers could be reported further down in the list (e.g. if you press one finger, then another, then release the first, the remaining finger remains in slot 1). Additionally I've never seen SUB_BIT_UP set. Only EXIST, DOWN, MOVE. Of which only EXIST matters, basically if it's set in a slot then it should be reported. Which is exactly how the driver worked before that commit.

I can of course provide a minimal patch that only changes the relevant bits. But I suspect that this will regress the issue Linus Walleij [email protected] ran into.

So at the moment the main questions are:

1. who else can test/verify (preferably we'd gather as many different devices as possible)
2. what's the difference between the devices, e.g. chip revision, firmware version, etc
3. whether there are any better workarounds for whatever issue Linus had. Given the severity of his issue I imagine he didn't even consider multitouch.

Mine's reported as:

[ 2242.801735] input: Zinitix Capacitive TouchScreen as /devices/platform/soc@0/78b9000.i2c/i2c-6/6-0020/input/input9
[ 2242.875593] zinitix_init_touch:251: Zinitix-TS 6-0020: chip revision 0001 firmware version 0004 regdata version 0009

[TravMurav]

Hi! I'd not be surprised if different firmware versions of the touch handle stuff slightly differently. You're probably best off duplicating this as a response to the patch you've mentioned and coordinating with Linus W. on the best way to resolve it (i.e. by mapping the quirk to a specific compatible value or firmware version)