From 026834db356ebb8d1211f5f94da348cd533c83c6 Mon Sep 17 00:00:00 2001
From: Alan Cox <alan@linux.intel.com>
Date: Sat, 25 Jul 2015 22:20:37 +0100
Subject: [PATCH] 68hc11test: begin fitting in firmware

---
 Kernel/platform-68hc11test/Makefile  |  10 ++-
 Kernel/platform-68hc11test/p68hc11.s | 106 ++++++++++++++++++++++++++-
 Kernel/platform-68hc11test/tricks.s  |   5 ++
 3 files changed, 115 insertions(+), 6 deletions(-)

diff --git a/Kernel/platform-68hc11test/Makefile b/Kernel/platform-68hc11test/Makefile
index d8d58e6b..7851ed52 100644
--- a/Kernel/platform-68hc11test/Makefile
+++ b/Kernel/platform-68hc11test/Makefile
@@ -3,7 +3,7 @@ CSRCS = devtty.c devrd.c
 CSRCS += devices.c main.c libc.c
 
 ASRCS = crt0.s p68hc11.s
-ASRCS += tricks.s commonmem.s
+ASRCS += tricks.s commonmem.s eeprom.s
 
 COBJS = $(CSRCS:.c=$(BINEXT))
 AOBJS = $(ASRCS:.s=$(BINEXT))
@@ -11,7 +11,7 @@ OBJS  = $(COBJS) $(AOBJS)
 
 JUNK = $(CSRCS:.c=.o) $(ASRCS:.s=.o)
 
-all:	$(OBJS)
+all:	$(OBJS) kernel.def eeprom.def
 
 $(COBJS): %$(BINEXT): %.c
 	$(CROSS_CC) $(CROSS_CCOPTS) -c $<
@@ -23,7 +23,7 @@ clean:
 	rm -f $(OBJS) $(JUNK)  core *~ 
 
 image:
-	$(CROSS_LD) -o ../fuzix.bin -M >../fuzix.map \
+	$(CROSS_LD) -o fuzix.out -M >../fuzix.map \
 	--script=memory.x \
 	crt0.o commonmem.o \
 	p68hc11.o ../start.o ../version.o ../lowlevel-68hc11.o \
@@ -32,3 +32,7 @@ image:
 	../syscall_proc.o ../syscall_other.o ../mm.o ../swap.o ../bankfixed.o \
 	../tty.o ../devsys.o ../usermem.o ../syscall_fs2.o ../syscall_exec16.o \
 	../usermem_std-68hc11.o devtty.o libc.o
+	$(CROSS_LD) -o eeprom.out -M >../eeprom.map \
+	--script=eeprom.x eeprom.o
+	m6811-elf-objcopy --output-target=binary fuzix.out ../fuzix.bin
+	m6811-elf-objcopy --output-target=binary eeprom.out ../eeprom.bin
diff --git a/Kernel/platform-68hc11test/p68hc11.s b/Kernel/platform-68hc11test/p68hc11.s
index e0bb1ca8..20b52e02 100644
--- a/Kernel/platform-68hc11test/p68hc11.s
+++ b/Kernel/platform-68hc11test/p68hc11.s
@@ -140,7 +140,7 @@ map_process:
 ;
 doexec:
 	xgdx		; function into X
-	ldaa usrbank	; bank
+	ldab usrbank	; bank
 	ldy U_DATA__U_ISP
 	jsr farcall
 ;
@@ -156,7 +156,7 @@ doexec:
 sigdispatch:
 	xgdx
 	ldy U_DATA__U_SYSCALL_SP
-	ldaa usrbank
+	ldab usrbank
 	; This may not return which is fine
 	jmp farcall
 	; Signal handler completed and returned back to the eeprom and thus
@@ -164,7 +164,8 @@ sigdispatch:
 
 ;
 ;	We have no common but instead route far accesses via the eeprom
-;	helpers in the firmware.
+;	helpers in the firmware. X points at the stackframe and B holds
+;	the previous bank.
 ;
 _ugetc:
 	xgdx
@@ -237,3 +238,102 @@ _uzero:
 	jmp farzero
 
 	rts
+
+;
+;	System call and interrupt vectors (via the eeprom firmware. It does
+;	the bank switches to 0 and return and the rti).
+;
+;	We only use SCI rx, timer and external irqs
+;	plus illegal opcodes and swi for syscall stuff
+;
+;	Y points to the far stack frame, B holds the bank we trapped from
+;
+;	FIXME: if we resched in the interrupt handler we break the eeprom
+;	irq vector handling assumptions. Need to rework this lot so we don't
+;	return via the eeprom irq path ?
+;
+
+
+serial_int:
+	ldaa #1
+do_int:
+	staa irqsrc
+	jsr interrupt_handler
+	clr irqsrc
+	tst U_DATA__U_CURSIG
+	bne int_signal
+	rts
+timer_int:
+	ldaa #2
+	bra do_int
+irq_int:
+	ldaa #3
+	bra do_int
+
+int_signal:
+	jsr dispatch_process_signal
+	rts
+;
+;	Synchronous events. This is ok to use tmp1 as we are not an
+;	irq affecting the kernel
+;
+;	On entry the Y register points to the far stack frame and A holds
+;	the bank we trapped from (and thus it's stack)
+;
+;	We are on a small trap handling stack with interrupts off. We must
+;	get off that stack stub
+;
+;	We need to copy the trap frame (9 bytes), above that is a return
+;	address (user space) and then arguments 2-4 (arg1 is in the trap
+;	frame X). Total 17 bytes
+;
+swi_int:
+	sty U_DATA__U_SYSCALL_SP ; userspace return stack
+	lds #kstack_top		; switch to kernel stack
+	clrb			; copy to kernel
+	ldx #17			; bytes to copy
+	stx tmp1
+	ldx #paramcopy
+	jsr farcopy
+	ldx #paramcopy
+	ldd 4,x			; stack frame X reg (arg1)
+	std U_DATA__U_ARGN
+	ldd 11,x		; arg 2
+	std U_DATA__U_ARGN1
+	ldd 13,x
+	std U_DATA__U_ARGN2
+	ldd 15,x
+	std U_DATA__U_ARGN3
+	ldd 4,x
+	std U_DATA__U_CALLNO
+	cli
+	jsr unix_syscall_entry
+	sei
+	ldx U_DATA__U_ERROR
+	ldd U_DATA__U_RETVAL
+	tst U_DATA__U_CURSIG
+	beq nosig
+	psha
+	pshb
+	pshx
+	;
+	;	FIXME: sort out the actual signal delivery logic
+	;
+	jsr dispatch_process_signal
+	pulx
+	pulb
+	pula
+nosig:
+	stx ret_x		; registers for farjmp return
+	std ret_d
+
+	ldy U_DATA__U_SYSCALL_SP
+	iny			; eat return address
+	iny
+	cli
+	rts
+
+	.sect .data
+irqsrc:	.byte 0
+
+.comm	paramcopy,17,1
diff --git a/Kernel/platform-68hc11test/tricks.s b/Kernel/platform-68hc11test/tricks.s
index b8c3b9b9..2f9be56b 100644
--- a/Kernel/platform-68hc11test/tricks.s
+++ b/Kernel/platform-68hc11test/tricks.s
@@ -116,6 +116,11 @@ badswitchmsg:
 	.byte 0
 
 ; new process pointer is in D
+;
+; FIXME: switchin needs to switch the interrupt stack descriptors or similar
+; on per task basis to handle the interrupt pre-emption case in a way that
+; keeps the firmware from getting grumpy.
+;
 switchin:
 	sei
 	
-- 
2.34.1