1
0
mirror of https://github.com/hsoft/collapseos.git synced 2024-11-01 22:20:56 +11:00
collapseos/doc/hw/z80/code/ps2ctl.asm
Virgil Dupras b12ac4b672 doc: fix PS/2 recipe
During the recipe overhaul, I forgot to carry PS/2 schematics over.
Also, the recipe should go in hw/z80 because it's specific to Z80.
2020-12-05 19:48:38 -05:00

432 lines
11 KiB
NASM

.include "tn45def.inc"
; Receives keystrokes from PS/2 keyboard and send them to the 595. As long as
; that number is not collected, we buffer the scan code received from ps/2. As
; soon as that number is collected we put the next number in the buffer. If the
; buffer is empty, we do nothing (the 595 already had its SRCLR pin triggered
; and shows 0).
;
; PS/2 is a bidirectional protocol, but in this program, we only care about
; receiving keystrokes. We don't send anything to the keyboard.
;
; The PS/2 keyboard has two data wires: Clock and Data. It is the keyboard that
; drives the clock with about 30-50 us between each clock.
;
; We wire the Clock to INT0 (PB2) and make it trigger an interrupt on the
; falling edge (the edge, in the PS/2 protocol, when data is set).
;
; Data is sent by the keyboard in 11-bit frames. 1 start bit (0), 8 data bits,
; one parity bit, one stop bit (1).
;
; Parity bit is set if number of bits in data bits is even. Unset otherwise.
;
; *** Receiving a data frame ***
;
; In idle mode, R18 is zero. When INT0 is triggered, it is increased and R17 is
; loaded with 0x80. We do this because we're going to right shift our data in
; (byte is sent LSB first). When the carry flag is set, we'll know we're
; finished. When that happens, we increase R18 again. We're waiting for parity
; bit. When we get it, we check parity and increase R18 again. We're waiting
; for stop bit. After we receive stop bit, we reset R18 to 0.
;
; On error, we ignore and reset our counters.
; *** Buffering scan codes ***
;
; The buffer starts at SRAM and stops at 0x100. It leaves space for the stack
; and makes overflow check easy. Also, we don't need a very big buffer. In this
; address space, Z chasing Y. When Y == Z, the buffer is empty. When 0x100 is
; reached, we go back to SRAM_START.
;
; Whenever a new scan code is received, we place it in Y and increase it.
; Whenever we send a scan code to the 595 (which can't be done when Z == Y
; because Z points to an invalid value), we send the value of Z and increase.
; *** Sending to the 595 ***
;
; Whenever a scan code is read from the 595, CE goes low and triggers a PCINT
; on PB4. When we get it, we clear the GPIOR0/1 flag to indicate that we're
; ready to send a new scan code to the 595.
;
; Because that CE flip/flop is real fast (375ns), it requires us to run at 8MHz.
;
; During the PCINT, we also trigger RCLK once because CE is also wired to SRCLR
; and we want the z80 to be able to know that the device has nothing to give
; (has a value of zero) rather than having to second guess (is this value, which
; is the same as the one that was read before, a new value or not?). With that
; "quick zero-in" scheme, there's no ambiguity: no scan code can be ready twice
; because it's replaced by a 0 as soon as it's read, until it can be filled with
; the next char in the buffer.
; *** Register Usage ***
;
; GPIOR0 flags:
; 0 - when set, indicates that the DATA pin was high when we received a
; bit through INT0. When we receive a bit, we set flag T to indicate
; it.
; 1 - When set, indicate that the 595 holds a value that hasn't been read
; by the z80 yet.
;
; R16: tmp stuff
; R17: recv buffer. Whenever we receive a bit, we push it in there.
; R18: recv step:
; - 0: idle
; - 1: receiving data
; - 2: awaiting parity bit
; - 3: awaiting stop bit
; R19: Register used for parity computations and tmp value in some other places
; R20: data being sent to the 595
; Y: pointer to the memory location where the next scan code from ps/2 will be
; written.
; Z: pointer to the next scan code to push to the 595
;
; *** Constants ***
;
.equ CLK = PINB2
.equ DATA = PINB1
.equ SRCLK = PINB3
.equ CE = PINB4
.equ RCLK = PINB0
; init value for TCNT0 so that overflow occurs in 100us
.equ TIMER_INITVAL = 0x100-100
rjmp main
rjmp hdlINT0
rjmp hdlPCINT
; Read DATA and set GPIOR0/0 if high. Then, set flag T.
; no SREG fiddling because no SREG-modifying instruction
hdlINT0:
sbic PINB, DATA ; DATA clear? skip next
sbi GPIOR0, 0
set
reti
; Only PB4 is hooked to PCINT and we don't bother checking the value of the PB4
; pin: things go too fast for this.
; no SREG fiddling because no SREG-modifying instruction
hdlPCINT:
; SRCLR has been triggered. Let's trigger RCLK too.
sbi PORTB, RCLK
cbi PORTB, RCLK
cbi GPIOR0, 1 ; 595 is now free
reti
main:
ldi r16, low(RAMEND)
out SPL, r16
ldi r16, high(RAMEND)
out SPH, r16
; Set clock prescaler to 1 (8MHz)
ldi r16, (1<<CLKPCE)
out CLKPR, r16
clr r16
out CLKPR, r16
; init variables
clr r18
out GPIOR0, r18
; Setup int0/PCINT
; INT0, falling edge
ldi r16, (1<<ISC01)
out MCUCR, r16
; Enable both INT0 and PCINT
ldi r16, (1<<INT0)|(1<<PCIE)
out GIMSK, r16
; For PCINT, enable only PB4
ldi r16, (1<<PCINT4)
out PCMSK, r16
; Setup buffer
clr YH
ldi YL, low(SRAM_START)
clr ZH
ldi ZL, low(SRAM_START)
; Setup timer. We use the timer to clear up "processbit" registers after
; 100us without a clock. This allows us to start the next frame in a
; fresh state. at 8MHZ, setting the counter's prescaler to 8 gives us
; a nice 1us for each TCNT0.
ldi r16, (1<<CS01) ; clk/8 prescaler
out TCCR0B, r16
; init DDRB
sbi DDRB, SRCLK
cbi PORTB, RCLK ; RCLK is generally kept low
sbi DDRB, RCLK
sei
loop:
brts processbit ; flag T set? we have a bit to process
cp YL, ZL ; if YL == ZL, buffer is empty
brne sendTo595 ; YL != ZL? our buffer has data
; nothing to do. Before looping, let's check if our communication timer
; overflowed.
in r16, TIFR
sbrc r16, TOV0
rjmp processbitReset ; Timer0 overflow? reset processbit
; Nothing to do for real.
rjmp loop
; Process the data bit received in INT0 handler.
processbit:
in r19, GPIOR0 ; backup GPIOR0 before we reset T
andi r19, 0x1 ; only keep the first flag
cbi GPIOR0, 0
clt ; ready to receive another bit
; We've received a bit. reset timer
rcall resetTimer
; Which step are we at?
tst r18
breq processbits0
cpi r18, 1
breq processbits1
cpi r18, 2
breq processbits2
; step 3: stop bit
clr r18 ; happens in all cases
; DATA has to be set
tst r19 ; Was DATA set?
breq loop ; not set? error, don't push to buffer
; push r17 to the buffer
st Y+, r17
rcall checkBoundsY
rjmp loop
processbits0:
; step 0 - start bit
; DATA has to be cleared
tst r19 ; Was DATA set?
brne loop ; Set? error. no need to do anything. keep r18
; as-is.
; DATA is cleared. prepare r17 and r18 for step 1
inc r18
ldi r17, 0x80
rjmp loop
processbits1:
; step 1 - receive bit
; We're about to rotate the carry flag into r17. Let's set it first
; depending on whether DATA is set.
clc
sbrc r19, 0 ; skip if DATA cleared.
sec
; Carry flag is set
ror r17
; Good. now, are we finished rotating? If carry flag is set, it means
; that we've rotated in 8 bits.
brcc loop ; we haven't finished yet
; We're finished, go to step 2
inc r18
rjmp loop
processbits2:
; step 2 - parity bit
mov r1, r19
mov r19, r17
rcall checkParity ; --> r16
cp r1, r16
brne processbitError ; r1 != r16? wrong parity
inc r18
rjmp loop
processbitError:
clr r18
ldi r19, 0xfe
rcall sendToPS2
rjmp loop
processbitReset:
clr r18
rcall resetTimer
rjmp loop
; send next scan code in buffer to 595, MSB.
sendTo595:
sbic GPIOR0, 1
rjmp loop ; flag 1 set? 595 is "busy". Don't send.
; We disable any interrupt handling during this routine. Whatever it
; is, it has no meaning to us at this point in time and processing it
; might mess things up.
cli
sbi DDRB, DATA
ld r20, Z+
rcall checkBoundsZ
ldi r16, 8
sendTo595Loop:
cbi PORTB, DATA
sbrc r20, 7 ; if leftmost bit isn't cleared, set DATA high
sbi PORTB, DATA
; toggle SRCLK
cbi PORTB, SRCLK
lsl r20
sbi PORTB, SRCLK
dec r16
brne sendTo595Loop ; not zero yet? loop
; We're finished sending our data to the 595 and we're ready to go back
; to business as usual. However, timing is important here. The z80 is
; 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 GPIOR0/1 to "595 is busy"
sbi GPIOR0, 1
; toggle RCLK
sbi PORTB, RCLK
cbi PORTB, RCLK
sei
rjmp loop
resetTimer:
ldi r16, TIMER_INITVAL
out TCNT0, r16
ldi r16, (1<<TOV0)
out TIFR, r16
ret
; Send the value of r19 to the PS/2 keyboard
sendToPS2:
; We don't use the general INT0 mechanism here. However, we still want
; to listen to PCINT, so we don't disable interrupts entirely, just
; INT0.
ldi r16, (1<<PCIE)
out GIMSK, r16
; First, indicate our request to send by holding both Clock low for
; 100us, then pull Data low
; lines low for 100us.
cbi PORTB, CLK
sbi DDRB, CLK
rcall resetTimer
; Wait until the timer overflows
in r16, TIFR
sbrs r16, TOV0
rjmp PC-2
; Good, 100us passed.
; Pull Data low, that's our start bit.
cbi PORTB, DATA
sbi DDRB, DATA
; Now, let's release the clock. At the next raising edge, we'll be
; expected to have set up our first bit (LSB). We set up when CLK is
; low.
cbi DDRB, CLK ; Should be starting high now.
; We will do the next loop 8 times
ldi r16, 8
; Let's remember initial r19 for parity
mov r1, r19
sendToPS2Loop:
; Wait for CLK to go low
sbic PINB, CLK
rjmp PC-1
; set up DATA
cbi PORTB, DATA
sbrc r19, 0 ; skip if LSB is clear
sbi PORTB, DATA
lsr r19
; Wait for CLK to go high
sbis PINB, CLK
rjmp PC-1
dec r16
brne sendToPS2Loop ; not zero? loop
; Data was sent, CLK is high. Let's send parity
mov r19, r1 ; recall saved value
rcall checkParity ; --> r16
; Wait for CLK to go low
sbic PINB, CLK
rjmp PC-1
; set parity bit
cbi PORTB, DATA
sbrc r16, 0 ; parity bit in r16
sbi PORTB, DATA
; Wait for CLK to go high
sbis PINB, CLK
rjmp PC-1
; Wait for CLK to go low
sbic PINB, CLK
rjmp PC-1
; We can now release the DATA line
cbi DDRB, DATA
; Wait for DATA to go low. That's our ACK
sbic PINB, DATA
rjmp PC-1
; Wait for CLK to go low
sbic PINB, CLK
rjmp PC-1
; We're finished! Enable INT0, reset timer, everything back to normal!
rcall resetTimer
clt ; also, make sure T isn't mistakely set.
ldi r16, (1<<INT0)|(1<<PCIE)
out GIMSK, r16
ret
; Check that Y is within bounds, reset to SRAM_START if not.
checkBoundsY:
tst YL
breq PC+2
ret ; not zero, nothing to do
; YL is zero. Reset Y
clr YH
ldi YL, low(SRAM_START)
ret
; Check that Z is within bounds, reset to SRAM_START if not.
checkBoundsZ:
tst ZL
breq PC+2
ret ; not zero, nothing to do
; ZL is zero. Reset Z
clr ZH
ldi ZL, low(SRAM_START)
ret
; Counts the number of 1s in r19 and set r16 to 1 if there's an even number of
; 1s, 0 if they're odd.
checkParity:
ldi r16, 1
lsr r19
brcc PC+2 ; Carry unset? skip next
inc r16 ; Carry set? We had a 1
tst r19 ; is r19 zero yet?
brne checkParity+1 ; no? loop and skip first LDI
andi r16, 0x1 ; Sets Z accordingly
ret