Re: Touchpad stickiness on AMD laptops (was Dell Inspiron/XPS)

[Hans de Goede <hdegoede@redhat.com>]

Hi,

On 1/17/22 09:39, Miroslav Bendík wrote:
>> [AMD Official Use Only]
>>
>>> Now i am trying to change ASF registers instead of SMBus registers.
>>> I have tried to enable interrupts and set listen address, but it don't
>>> work or
>>> i can't recognize the difference between interrupts generated by
>>> transfers and
>>> interrupts generated from slave.
>> Try reading the value of SFx0A ASFStatus bit 5 (it's write to clear if it's an interrupt).
>>
>>> outb_p(0x02, 0x15 + piix4_smba); // SlaveIntrListenEn
>>> outb_p(0x2c << 1 | 0x01, 0x09 + piix4_smba); // ListenAdr | ListenAdrEn
>> ASFx04 SlaveAddress instead of  ASFx09 ListenAdr
>> ?
>>
>>
> Little bit more informations:
> 
> Interrupts are generated only if ASFx09 ListenAdr is:
> 
> (0x08 << 1) | 0x01
> (0x10 << 1) | 0x01
> 
> and touchpad is initialized with synaptics_intertouch=1
> 
> There is maybe small correlation between frequency and touch, but i am
> not 100% sure.

I know very litlle about this, but I believe that when using
host-notify that after receiving the host-notify you are supposed to
do an I2C read from the SMBus Alert Response Address (ARA, 0x0c) to find
out the source of the notify (since multiple devices on the bus may
be notify capable). I guess that the controller may not do that itself
and that as long as you have not done it the touchpad may keep repeating
the notify.

But as said I know very little about this, so take this with a big
grain of salt :)  I guess you may want to read up a bit on how this
is supposed to work at the bus level. I believe that the SMBUS spec
is public.

Regards,

Hans




> 
> There are no register changed in interrupt handler except of
> ASFx13 DataBankSel. I can't determine if interrupt is generated from
> transfer, or from external event.
> 
> ASF should be system for remote management. It should have access to
> SMBus and data / command registers are identical, this means, that SMBus
> should work (except block transfers).
> 
> If ASF just mirrors SMBus, then question is, why i can't access to
> touchpad using SMBus? One strange thing is, that i2cdetect on standard
> SMbus (0xb00), port 0 returns:
> 
>      0  1  2  3  4  5  6  7  8  9  a  b  c  d  e  f
> 00:                         -- -- -- -- -- -- -- --
> 10: -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- --
> 20: -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- --
> 30: -- -- -- -- -- -- 36 37 -- -- -- -- -- -- -- --
> 40: -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- --
> 50: 50 -- -- -- -- -- -- -- 58 -- -- -- -- -- -- --
> 60: -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- --
> 70: -- -- -- -- -- -- -- --
> 
> Address 0x58 is exactly 0x2c (synaptics) moved 1 bit left, but i2c-piix4
> correctly moves address.
>