recipes/rc2014/ps2: drive a shell with ps/2 kbd!

kernel/kbd.asm

@@ -0,0 +1,81 @@
+; kbd - implement GetC for PS/2 keyboard
+;
+; Status: Work in progress. See recipes/rc2014/ps2
+;
+; *** Defines ***
+; The port of the device where we read scan codes. See recipe rc2014/ps2.
+; KBD_PORT
+
+; *** Variables ***
+.equ	KBD_SKIP_NEXT	KBD_RAMSTART
+.equ	KBD_RAMEND	KBD_SKIP_NEXT+1
+
+kbdInit:
+	xor	a
+	ld	(KBD_SKIP_NEXT), a
+	ret
+
+kbdGetC:
+	in	a, (KBD_PORT)
+	or	a		; cp 0
+	ret	z
+	; scan code not zero, maybe we have something.
+	; Do we need to skip it?
+	push	af		;  <|
+	ld	a, (KBD_SKIP_NEXT) ;|
+	or	a		;   |
+	jr	nz, .skip	;   |
+	pop	af		;  <|
+	cp	0x80
+	jr	nc, .outOfBounds
+	; No need to skip, code within bounds, we have something! Let's see if
+	; there's a ASCII code associated to it.
+	push	hl		; <|
+	ld	hl, kbdScanCodes ; |
+	call	addHL		;  |
+	ld	a, (hl)		;  |
+	pop	hl		; <|
+	or	a		; cp 0
+	jp	z, unsetZ	; no code. Keep A at 0, but unset Z
+	; We have something!
+	cp	a		; ensure Z
+	ret
+.outOfBounds:
+	; A scan code over 0x80 is out of bounds. Ignore.
+	; If F0 (break code) or E0 (extended code), we also skip the next code
+	cp	0xf0
+	jr	z, .skipNext
+	cp	0xe0
+	jr	z, .skipNext
+	xor	a
+	jp	unsetZ
+.skipNext:
+	ld	(KBD_SKIP_NEXT), a
+	xor	a
+	jp	unsetZ
+.skip:
+	pop	af		; equilibrate stack
+	xor	a
+	ld	(KBD_SKIP_NEXT), a
+	jp	unsetZ
+
+; A list of the value associated with the 0x80 possible scan codes of the set
+; 2 of the PS/2 keyboard specs. 0 means no value. That value is a character than
+; can be read in a GetC routine. No make code in the PS/2 set 2 reaches 0x80.
+kbdScanCodes:
+; 0x00    1   2   3   4   5   6   7   8   9   a   b   c   d   e   f
+.db   0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  9,'`',  0
+; 0x10 9 = TAB
+.db   0,  0,  0,  0,  0,'Q','1',  0,  0,  0,'Z','S','A','W','2',  0
+; 0x20 32 = SPACE
+.db   0,'C','X','D','E','4','3',  0,  0, 32,'V','F','T','R','5',  0
+; 0x30
+.db   0,'N','B','H','G','Y','6',  0,  0,  0,'M','J','U','7','8',  0
+; 0x40 59 = ;
+.db   0,',','K','I','O','0','9',  0,  0,'.','/','L', 59,'P','-',  0
+; 0x50 13 = RETURN 39 = '
+.db   0,  0, 39,  0,'[','=',  0,  0,  0,  0,  13,']',  0,'\',  0,  0
+; 0x60 8 = BKSP
+.db   0,  0,  0,  0,  0,  0,  8,  0,  0,  0,  0,  0,  0,  0,  0,  0
+; 0x70 27 = ESC
+.db   0,  0,  0,  0,  0,  0, 27,  0,  0,  0,  0,  0,  0,  0,  0,  0

recipes/.gitignore

@@ -1,2 +1,4 @@
 *.bin
 *.cfs
+*.hex
+*.obj

recipes/rc2014/ps2/Makefile

@@ -1,7 +1,9 @@
 PROGNAME = ps2ctl
 AVRDUDEMCU ?= t45
 AVRDUDEARGS ?= -c usbtiny -P usb 
-TARGETS = $(PROGNAME).hex
+TARGETS = $(PROGNAME).hex os.bin
+ZASM = ../../../tools/zasm.sh
+KERNEL = ../../../kernel
 
 # Rules
 
@@ -14,9 +16,10 @@ send: $(PROGNAME).hex
 	avrdude $(AVRDUDEARGS) -p $(AVRDUDEMCU) -U flash:w:$<
 
 $(PROGNAME).hex: $(PROGNAME).asm
-$(TARGETS):
 	avra -o $@ $<
 
-clean:
+os.bin: glue.asm
-	rm -f $(TARGETS) *.eep.hex *.obj
+	$(ZASM) $(KERNEL) < $< > $@
 
+clean:
+	rm -f $(TARGETS) *.eep.hex *.obj os.bin

recipes/rc2014/ps2/README.md

@@ -36,16 +36,15 @@ address selection + `IORQ` + `RO`
 
 ## Using the PS/2 interface
 
-As of now, the interface is incomplete and can only be queried through the
+After having built and flashed the `glue.asm` supplied with this recipe, you end
-shell's `iord`. I've set my device up for addr `8` (that is, I wired `A3`
+up with a shell driven by the PS/2 keyboard (but it still outputs to ACIA).
-through the inverter, the rest through diodes, and hooked this pudding to `OE`).
 
-When doing `iord 8` in the shell, I get the scan code of the last key I pressed,
+You will see, by typing on the keyboard, that it kinda works, but in a very
-unless the 595 was "busy" with another code. For example, if I press `A`, my
+basic and glitchy way. You will get double letters sometimes, and at some point,
-next `iord 8` will yield `1C` (the "make" code for "A" in the PS/2 protocol).
+communications are likely to become "corrupted" (you reliably get the wrong
+letters). That's because parity checks, timeouts and reset procedures aren't
+implemented yet.
 
-Doing a second `iord 8` right after a first will yield `0`, indicating that the
+But still, it kinda works!
-device properly detect the first reading attempt and properly flushes the value
-from the 595.
 
 [avra]: https://github.com/hsoft/avra

recipes/rc2014/ps2/glue.asm

@@ -0,0 +1,42 @@
+.equ	RAMSTART	0x8000
+.equ	RAMEND		0xffff
+.equ	ACIA_CTL	0x80	; Control and status. RS off.
+.equ	ACIA_IO		0x81	; Transmit. RS on.
+.equ	KBD_PORT	0x08
+
+jp	init
+
+; interrupt hook
+.fill	0x38-$
+jp	aciaInt
+
+#include "err.h"
+#include "core.asm"
+#include "parse.asm"
+.equ	ACIA_RAMSTART	RAMSTART
+#include "acia.asm"
+
+.equ	KBD_RAMSTART	ACIA_RAMEND
+#include "kbd.asm"
+
+.equ	STDIO_RAMSTART	KBD_RAMEND
+#include "stdio.asm"
+
+.equ	SHELL_RAMSTART	STDIO_RAMEND
+.equ	SHELL_EXTRA_CMD_COUNT 0
+#include "shell.asm"
+
+init:
+	di
+	; setup stack
+	ld	hl, RAMEND
+	ld	sp, hl
+	im 1
+
+	call	aciaInit
+	ld	hl, kbdGetC
+	ld	de, aciaPutC
+	call	stdioInit
+	call	shellInit
+	ei
+	jp	shellLoop

recipes/rc2014/ps2/ps2ctl.asm

@@ -233,16 +233,27 @@ sendTo595Loop:
 	dec	r16
 	brne	sendTo595Loop	; not zero yet? loop
 
-	; toggle RCLK
+	; We're finished sending our data to the 595 and we're ready to go back
-	sbi	PORTB, RCLK
+	; to business as usual. However, timing is important here. The z80 is
-	cbi	PORTB, RCLK
+	; very fast and constantly hammers our 595 with polls. While this
+	; routine was running, it was getting zeroes, which is fine, but as soon
+	; as we trigger RCLK, the z80 is going to fetch that value. What we want
+	; to do is to enable back the interrupts as soon as RCLK is triggered
+	; so that the z80 doesn't have enough time to poll twice. If it did, we
+	; would return a double character. This is why RCLK triggering is the
+	; last operation.
 
 	; release PS/2
 	cbi	DDRB, DATA
 
 	; Set R2 to "595 is busy"
 	inc	r2
+
+	; toggle RCLK
+	sbi	PORTB, RCLK
+	cbi	PORTB, RCLK
 	sei
+
 	rjmp	loop
 
 ; Check that Y is within bounds, reset to SRAM_START if not.