On Fri, 30 May 2025 13:42:11 -0400 "Ted Unangst" <[email protected]> wrote:
> On 2025-05-29, izzy Meyer wrote: > > > I was able to get it to attach as a synaptics device with the attached > > diff though. I took it from chris's tips (CC'd) on that linked thread in > > the initial email here. I dunno if this is notable, but chris mentioned > > an xfer error that would occasionally show up. I've been running this > > diff with the #675 userland (i sysupgrade'd last night) for all of > > today so far and the error has yet to pop up. Is there a possibility of > > supporting active multiplexing in the future so i don't need to patch > > this #define? > > Oh, this is puzzling. Why did it work before? But definitely helpful > to have at least a workaround solution. Thanks. > I have a Panasonic CF-SZ6 which is very similar to the CF-NX4 in specification and behaviour. On both 7.6 and 7.7-release, the touchpad was detected as a default PS/2 mouse (with no vertical scrolling), not as a Synaptics device, because it fails to respond to any capability probing commands. It's my understanding from this thread that this is still the situation on the latest snapshots after Ted's patch which disables ISA bus probing, at least after a reboot. Having determined in debugging that the machine was using Active Muxing on the PS/2 Aux devices and ignoring the commands in the driver to switch back to legacy mode, I was able to make the Synaptics device attach using a rather crude patch changing KBC_AUXWRITE in sys/dev/ic/i8042reg.h to 0x91 (the routing command for Active Muxing AUX1) instead of 0xd4 (Legacy AUX). This all relates to the pms driver after attachment. (See https://marc.info/?l=openbsd-bugs&m=173019316731591&w=2, and http://www.isdaman.com/alsos/hardware/mouse/mouse2.pdf) There are occasional "pms0: not in sync yet, discard input (state = 3, 90 90 cf 3a | 77 ea)" or "pms0: not in sync yet, discard input (state = 0, 00 | 00 00 c0 00 80)" (and even more rarely some other byte patterns) but no apparent impact on operation. Most probably this could be done more elegantly, perhaps by using a FLAGS value that could be applied in UKC, but this is sadly beyond my knowledge to implement. I'll try to test with a new snapshot in the next few days, to check if this method of getting the touchpad to respond to capability probing still works with an ACPI attachment -- Chris
