[evdev]: Minor KPI changes

[wulf]

Summary

This combines 2 mostly no-op changes affecting ft5406, ums, ukbd and some other drivers

1. Remove evdev.ko linkage dependency on kbd driver
   
   Move evdev_ev_kbd_event() helper from evdev to kbd.c as otherwise evdev unconditionally requires all keyboard and console stuff to be compiled in to kernel. This dependency happens as evdev_ev_kbd_event() helper references kbdsw global variable defined in kbd.c through use of kbdd_ioctl() macro.
   
   While here make all keyboard drivers respect evdev_rcpt_mask while setting typematic rate and LEDs with evdev interface.
   
   Requested by: Milan Obuch <[e-mail]>

2. remove soft context from evdev methods parameter list.
   
   Now softc should be retrieved from struct edvev * pointer with evdev_get_softc() helper.
   
   wmt(4) is a sample of driver that support both KPI.

[• hselasky]

sys/dev/usb/input/wmt.c

233 ↗

(On Diff #46375)

Why is there an #if 1 here ?

[wulf]

sys/dev/usb/input/wmt.c

233 ↗

(On Diff #46375)

I am going to replace it with somethig like

#if (__FreeBSD_version >= 12XXXXXX)

when XXXXXX will be known

[wulf]

#if 1 replaced with #if __FreeBSD_version >= 1200000

[andrew]

sys/arm/broadcom/bcm2835/bcm2835_ft5406.c

180 ↗

(On Diff #46376)

Why are you casting this? A void * shouldn't need to be cast to another pointer.

sys/dev/usb/input/wmt.c

233 ↗

(On Diff #46375)

Do we need the check?

[wulf]

sys/arm/broadcom/bcm2835/bcm2835_ft5406.c

180 ↗

(On Diff #46376)

I think I took it from some code snippets originated in caddr_t times. I will remove these useless casts.

sys/dev/usb/input/wmt.c

233 ↗

(On Diff #46375)

This check is not needed for already MFC-ed to 11 drivers but can be useful for not yet committed ones, if any exist, as it makes such a MFC simpler.
It is placed here only for reference as this file is a sample of driver that supports both versions of KPI. It should be removed after 11-EOL.

[wulf]

Useless softc casts removed

[gonzo]

LGTM