macemu/SheepShaver/src/kpx_cpu/sheepshaver_glue.cpp

1023 lines
26 KiB
C++
Raw Normal View History

/*
* sheepshaver_glue.cpp - Glue Kheperix CPU to SheepShaver CPU engine interface
*
2004-01-12 15:37:24 +00:00
* SheepShaver (C) 1997-2004 Christian Bauer and Marc Hellwig
*
* This program is free software; you can redistribute it and/or modify
* it under the terms of the GNU General Public License as published by
* the Free Software Foundation; either version 2 of the License, or
* (at your option) any later version.
*
* This program is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
* GNU General Public License for more details.
*
* You should have received a copy of the GNU General Public License
* along with this program; if not, write to the Free Software
* Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
*/
#include "sysdeps.h"
#include "cpu_emulation.h"
#include "main.h"
#include "prefs.h"
#include "xlowmem.h"
#include "emul_op.h"
#include "rom_patches.h"
#include "macos_util.h"
#include "block-alloc.hpp"
#include "sigsegv.h"
#include "cpu/ppc/ppc-cpu.hpp"
#include "cpu/ppc/ppc-operations.hpp"
#include "cpu/ppc/ppc-instructions.hpp"
#include "thunks.h"
// Used for NativeOp trampolines
#include "video.h"
#include "name_registry.h"
#include "serial.h"
#include "ether.h"
#include <stdio.h>
#if ENABLE_MON
#include "mon.h"
#include "mon_disass.h"
#endif
#define DEBUG 0
#include "debug.h"
// Emulation time statistics
#define EMUL_TIME_STATS 1
#if EMUL_TIME_STATS
static clock_t emul_start_time;
static uint32 interrupt_count = 0;
static clock_t interrupt_time = 0;
static uint32 exec68k_count = 0;
static clock_t exec68k_time = 0;
static uint32 native_exec_count = 0;
static clock_t native_exec_time = 0;
static uint32 macos_exec_count = 0;
static clock_t macos_exec_time = 0;
#endif
static void enter_mon(void)
{
// Start up mon in real-mode
#if ENABLE_MON
char *arg[4] = {"mon", "-m", "-r", NULL};
mon(3, arg);
#endif
}
// From main_*.cpp
extern uintptr SignalStackBase();
// PowerPC EmulOp to exit from emulation looop
const uint32 POWERPC_EXEC_RETURN = POWERPC_EMUL_OP | 1;
// Enable multicore (main/interrupts) cpu emulation?
#define MULTICORE_CPU (ASYNC_IRQ ? 1 : 0)
// Enable Execute68k() safety checks?
#define SAFE_EXEC_68K 1
// Save FP state in Execute68k()?
#define SAVE_FP_EXEC_68K 1
// Interrupts in EMUL_OP mode?
#define INTERRUPTS_IN_EMUL_OP_MODE 1
// Interrupts in native mode?
#define INTERRUPTS_IN_NATIVE_MODE 1
// Pointer to Kernel Data
static KernelData * const kernel_data = (KernelData *)KERNEL_DATA_BASE;
// SIGSEGV handler
static sigsegv_return_t sigsegv_handler(sigsegv_address_t, sigsegv_address_t);
// JIT Compiler enabled?
static inline bool enable_jit_p()
{
return PrefsFindBool("jit");
}
/**
* PowerPC emulator glue with special 'sheep' opcodes
**/
enum {
PPC_I(SHEEP) = PPC_I(MAX),
PPC_I(SHEEP_MAX)
};
class sheepshaver_cpu
: public powerpc_cpu
{
void init_decoder();
void execute_sheep(uint32 opcode);
public:
// Constructor
sheepshaver_cpu();
// CR & XER accessors
uint32 get_cr() const { return cr().get(); }
void set_cr(uint32 v) { cr().set(v); }
uint32 get_xer() const { return xer().get(); }
void set_xer(uint32 v) { xer().set(v); }
// Execute 68k routine
void execute_68k(uint32 entry, M68kRegisters *r);
// Execute ppc routine
void execute_ppc(uint32 entry);
// Execute MacOS/PPC code
uint32 execute_macos_code(uint32 tvect, int nargs, uint32 const *args);
// Resource manager thunk
void get_resource(uint32 old_get_resource);
// Handle MacOS interrupt
void interrupt(uint32 entry);
void handle_interrupt();
// Lazy memory allocator (one item at a time)
void *operator new(size_t size)
{ return allocator_helper< sheepshaver_cpu, lazy_allocator >::allocate(); }
void operator delete(void *p)
{ allocator_helper< sheepshaver_cpu, lazy_allocator >::deallocate(p); }
// FIXME: really make surre array allocation fail at link time?
void *operator new[](size_t);
void operator delete[](void *p);
// Make sure the SIGSEGV handler can access CPU registers
friend sigsegv_return_t sigsegv_handler(sigsegv_address_t, sigsegv_address_t);
};
lazy_allocator< sheepshaver_cpu > allocator_helper< sheepshaver_cpu, lazy_allocator >::allocator;
sheepshaver_cpu::sheepshaver_cpu()
: powerpc_cpu(enable_jit_p())
{
init_decoder();
}
void sheepshaver_cpu::init_decoder()
{
#ifndef PPC_NO_STATIC_II_INDEX_TABLE
static bool initialized = false;
if (initialized)
return;
initialized = true;
#endif
static const instr_info_t sheep_ii_table[] = {
{ "sheep",
(execute_pmf)&sheepshaver_cpu::execute_sheep,
NULL,
PPC_I(SHEEP),
D_form, 6, 0, CFLOW_JUMP | CFLOW_TRAP
}
};
const int ii_count = sizeof(sheep_ii_table)/sizeof(sheep_ii_table[0]);
D(bug("SheepShaver extra decode table has %d entries\n", ii_count));
for (int i = 0; i < ii_count; i++) {
const instr_info_t * ii = &sheep_ii_table[i];
init_decoder_entry(ii);
}
}
// Forward declaration for native opcode handler
static void NativeOp(int selector);
/* NativeOp instruction format:
+------------+--------------------------+--+----------+------------+
| 6 | |FN| OP | 2 |
+------------+--------------------------+--+----------+------------+
0 5 |6 19 20 21 25 26 31
*/
typedef bit_field< 20, 20 > FN_field;
typedef bit_field< 21, 25 > NATIVE_OP_field;
typedef bit_field< 26, 31 > EMUL_OP_field;
// Execute SheepShaver instruction
void sheepshaver_cpu::execute_sheep(uint32 opcode)
{
// D(bug("Extended opcode %08x at %08x (68k pc %08x)\n", opcode, pc(), gpr(24)));
assert((((opcode >> 26) & 0x3f) == 6) && OP_MAX <= 64 + 3);
switch (opcode & 0x3f) {
case 0: // EMUL_RETURN
QuitEmulator();
break;
case 1: // EXEC_RETURN
spcflags().set(SPCFLAG_CPU_EXEC_RETURN);
break;
case 2: // EXEC_NATIVE
NativeOp(NATIVE_OP_field::extract(opcode));
if (FN_field::test(opcode))
pc() = lr();
else
pc() += 4;
break;
default: { // EMUL_OP
M68kRegisters r68;
WriteMacInt32(XLM_68K_R25, gpr(25));
WriteMacInt32(XLM_RUN_MODE, MODE_EMUL_OP);
for (int i = 0; i < 8; i++)
r68.d[i] = gpr(8 + i);
for (int i = 0; i < 7; i++)
r68.a[i] = gpr(16 + i);
r68.a[7] = gpr(1);
uint32 saved_cr = get_cr() & CR_field<2>::mask();
uint32 saved_xer = get_xer();
EmulOp(&r68, gpr(24), EMUL_OP_field::extract(opcode) - 3);
set_cr(saved_cr);
set_xer(saved_xer);
for (int i = 0; i < 8; i++)
gpr(8 + i) = r68.d[i];
for (int i = 0; i < 7; i++)
gpr(16 + i) = r68.a[i];
gpr(1) = r68.a[7];
WriteMacInt32(XLM_RUN_MODE, MODE_68K);
pc() += 4;
break;
}
}
}
// Handle MacOS interrupt
void sheepshaver_cpu::interrupt(uint32 entry)
{
#if EMUL_TIME_STATS
interrupt_count++;
const clock_t interrupt_start = clock();
#endif
#if !MULTICORE_CPU
// Save program counters and branch registers
uint32 saved_pc = pc();
uint32 saved_lr = lr();
uint32 saved_ctr= ctr();
uint32 saved_sp = gpr(1);
#endif
// Initialize stack pointer to SheepShaver alternate stack base
gpr(1) = SignalStackBase() - 64;
// Build trampoline to return from interrupt
SheepVar32 trampoline = POWERPC_EXEC_RETURN;
// Prepare registers for nanokernel interrupt routine
kernel_data->v[0x004 >> 2] = htonl(gpr(1));
kernel_data->v[0x018 >> 2] = htonl(gpr(6));
gpr(6) = ntohl(kernel_data->v[0x65c >> 2]);
assert(gpr(6) != 0);
WriteMacInt32(gpr(6) + 0x13c, gpr(7));
WriteMacInt32(gpr(6) + 0x144, gpr(8));
WriteMacInt32(gpr(6) + 0x14c, gpr(9));
WriteMacInt32(gpr(6) + 0x154, gpr(10));
WriteMacInt32(gpr(6) + 0x15c, gpr(11));
WriteMacInt32(gpr(6) + 0x164, gpr(12));
WriteMacInt32(gpr(6) + 0x16c, gpr(13));
gpr(1) = KernelDataAddr;
gpr(7) = ntohl(kernel_data->v[0x660 >> 2]);
gpr(8) = 0;
gpr(10) = trampoline.addr();
gpr(12) = trampoline.addr();
gpr(13) = get_cr();
// rlwimi. r7,r7,8,0,0
uint32 result = op_ppc_rlwimi::apply(gpr(7), 8, 0x80000000, gpr(7));
record_cr0(result);
gpr(7) = result;
gpr(11) = 0xf072; // MSR (SRR1)
cr().set((gpr(11) & 0x0fff0000) | (get_cr() & ~0x0fff0000));
// Enter nanokernel
execute(entry);
#if !MULTICORE_CPU
// Restore program counters and branch registers
pc() = saved_pc;
lr() = saved_lr;
ctr()= saved_ctr;
gpr(1) = saved_sp;
#endif
#if EMUL_TIME_STATS
interrupt_time += (clock() - interrupt_start);
#endif
}
// Execute 68k routine
void sheepshaver_cpu::execute_68k(uint32 entry, M68kRegisters *r)
{
#if EMUL_TIME_STATS
exec68k_count++;
const clock_t exec68k_start = clock();
#endif
#if SAFE_EXEC_68K
if (ReadMacInt32(XLM_RUN_MODE) != MODE_EMUL_OP)
printf("FATAL: Execute68k() not called from EMUL_OP mode\n");
#endif
// Save program counters and branch registers
uint32 saved_pc = pc();
uint32 saved_lr = lr();
uint32 saved_ctr= ctr();
uint32 saved_cr = get_cr();
// Create MacOS stack frame
// FIXME: make sure MacOS doesn't expect PPC registers to live on top
uint32 sp = gpr(1);
gpr(1) -= 56;
WriteMacInt32(gpr(1), sp);
// Save PowerPC registers
uint32 saved_GPRs[19];
memcpy(&saved_GPRs[0], &gpr(13), sizeof(uint32)*(32-13));
#if SAVE_FP_EXEC_68K
double saved_FPRs[18];
memcpy(&saved_FPRs[0], &fpr(14), sizeof(double)*(32-14));
#endif
// Setup registers for 68k emulator
cr().set(CR_SO_field<2>::mask()); // Supervisor mode
for (int i = 0; i < 8; i++) // d[0]..d[7]
gpr(8 + i) = r->d[i];
for (int i = 0; i < 7; i++) // a[0]..a[6]
gpr(16 + i) = r->a[i];
gpr(23) = 0;
gpr(24) = entry;
gpr(25) = ReadMacInt32(XLM_68K_R25); // MSB of SR
gpr(26) = 0;
gpr(28) = 0; // VBR
gpr(29) = ntohl(kernel_data->ed.v[0x74 >> 2]); // Pointer to opcode table
gpr(30) = ntohl(kernel_data->ed.v[0x78 >> 2]); // Address of emulator
gpr(31) = KernelDataAddr + 0x1000;
// Push return address (points to EXEC_RETURN opcode) on stack
gpr(1) -= 4;
WriteMacInt32(gpr(1), XLM_EXEC_RETURN_OPCODE);
// Rentering 68k emulator
WriteMacInt32(XLM_RUN_MODE, MODE_68K);
// Set r0 to 0 for 68k emulator
gpr(0) = 0;
// Execute 68k opcode
uint32 opcode = ReadMacInt16(gpr(24));
gpr(27) = (int32)(int16)ReadMacInt16(gpr(24) += 2);
gpr(29) += opcode * 8;
execute(gpr(29));
// Save r25 (contains current 68k interrupt level)
WriteMacInt32(XLM_68K_R25, gpr(25));
// Reentering EMUL_OP mode
WriteMacInt32(XLM_RUN_MODE, MODE_EMUL_OP);
// Save 68k registers
for (int i = 0; i < 8; i++) // d[0]..d[7]
r->d[i] = gpr(8 + i);
for (int i = 0; i < 7; i++) // a[0]..a[6]
r->a[i] = gpr(16 + i);
// Restore PowerPC registers
memcpy(&gpr(13), &saved_GPRs[0], sizeof(uint32)*(32-13));
#if SAVE_FP_EXEC_68K
memcpy(&fpr(14), &saved_FPRs[0], sizeof(double)*(32-14));
#endif
// Cleanup stack
gpr(1) += 56;
// Restore program counters and branch registers
pc() = saved_pc;
lr() = saved_lr;
ctr()= saved_ctr;
set_cr(saved_cr);
#if EMUL_TIME_STATS
exec68k_time += (clock() - exec68k_start);
#endif
}
// Call MacOS PPC code
uint32 sheepshaver_cpu::execute_macos_code(uint32 tvect, int nargs, uint32 const *args)
{
#if EMUL_TIME_STATS
macos_exec_count++;
const clock_t macos_exec_start = clock();
#endif
// Save program counters and branch registers
uint32 saved_pc = pc();
uint32 saved_lr = lr();
uint32 saved_ctr= ctr();
// Build trampoline with EXEC_RETURN
SheepVar32 trampoline = POWERPC_EXEC_RETURN;
lr() = trampoline.addr();
gpr(1) -= 64; // Create stack frame
uint32 proc = ReadMacInt32(tvect); // Get routine address
uint32 toc = ReadMacInt32(tvect + 4); // Get TOC pointer
// Save PowerPC registers
uint32 regs[8];
regs[0] = gpr(2);
for (int i = 0; i < nargs; i++)
regs[i + 1] = gpr(i + 3);
// Prepare and call MacOS routine
gpr(2) = toc;
for (int i = 0; i < nargs; i++)
gpr(i + 3) = args[i];
execute(proc);
uint32 retval = gpr(3);
// Restore PowerPC registers
for (int i = 0; i <= nargs; i++)
gpr(i + 2) = regs[i];
// Cleanup stack
gpr(1) += 64;
// Restore program counters and branch registers
pc() = saved_pc;
lr() = saved_lr;
ctr()= saved_ctr;
#if EMUL_TIME_STATS
macos_exec_time += (clock() - macos_exec_start);
#endif
return retval;
}
// Execute ppc routine
inline void sheepshaver_cpu::execute_ppc(uint32 entry)
{
// Save branch registers
uint32 saved_lr = lr();
SheepVar32 trampoline = POWERPC_EXEC_RETURN;
WriteMacInt32(trampoline.addr(), POWERPC_EXEC_RETURN);
lr() = trampoline.addr();
execute(entry);
// Restore branch registers
lr() = saved_lr;
}
// Resource Manager thunk
extern "C" void check_load_invoc(uint32 type, int16 id, uint32 h);
inline void sheepshaver_cpu::get_resource(uint32 old_get_resource)
{
uint32 type = gpr(3);
int16 id = gpr(4);
// Create stack frame
gpr(1) -= 56;
// Call old routine
execute_ppc(old_get_resource);
// Call CheckLoad()
uint32 handle = gpr(3);
check_load_invoc(type, id, handle);
gpr(3) = handle;
// Cleanup stack
gpr(1) += 56;
}
/**
* SheepShaver CPU engine interface
**/
static sheepshaver_cpu *main_cpu = NULL; // CPU emulator to handle usual control flow
static sheepshaver_cpu *interrupt_cpu = NULL; // CPU emulator to handle interrupts
static sheepshaver_cpu *current_cpu = NULL; // Current CPU emulator context
void FlushCodeCache(uintptr start, uintptr end)
{
D(bug("FlushCodeCache(%08x, %08x)\n", start, end));
main_cpu->invalidate_cache_range(start, end);
#if MULTICORE_CPU
interrupt_cpu->invalidate_cache_range(start, end);
#endif
}
static inline void cpu_push(sheepshaver_cpu *new_cpu)
{
#if MULTICORE_CPU
current_cpu = new_cpu;
#endif
}
static inline void cpu_pop()
{
#if MULTICORE_CPU
current_cpu = main_cpu;
#endif
}
// Dump PPC registers
static void dump_registers(void)
{
current_cpu->dump_registers();
}
// Dump log
static void dump_log(void)
{
current_cpu->dump_log();
}
/*
* Initialize CPU emulation
*/
static sigsegv_return_t sigsegv_handler(sigsegv_address_t fault_address, sigsegv_address_t fault_instruction)
{
#if ENABLE_VOSF
// Handle screen fault
extern bool Screen_fault_handler(sigsegv_address_t, sigsegv_address_t);
if (Screen_fault_handler(fault_address, fault_instruction))
return SIGSEGV_RETURN_SUCCESS;
#endif
const uintptr addr = (uintptr)fault_address;
#if HAVE_SIGSEGV_SKIP_INSTRUCTION
// Ignore writes to ROM
if ((addr - ROM_BASE) < ROM_SIZE)
return SIGSEGV_RETURN_SKIP_INSTRUCTION;
// Get program counter of target CPU
sheepshaver_cpu * const cpu = current_cpu;
const uint32 pc = cpu->pc();
// Fault in Mac ROM or RAM?
bool mac_fault = (pc >= ROM_BASE) && (pc < (ROM_BASE + ROM_AREA_SIZE)) || (pc >= RAMBase) && (pc < (RAMBase + RAMSize));
if (mac_fault) {
// "VM settings" during MacOS 8 installation
if (pc == ROM_BASE + 0x488160 && cpu->gpr(20) == 0xf8000000)
return SIGSEGV_RETURN_SKIP_INSTRUCTION;
// MacOS 8.5 installation
else if (pc == ROM_BASE + 0x488140 && cpu->gpr(16) == 0xf8000000)
return SIGSEGV_RETURN_SKIP_INSTRUCTION;
// MacOS 8 serial drivers on startup
else if (pc == ROM_BASE + 0x48e080 && (cpu->gpr(8) == 0xf3012002 || cpu->gpr(8) == 0xf3012000))
return SIGSEGV_RETURN_SKIP_INSTRUCTION;
// MacOS 8.1 serial drivers on startup
else if (pc == ROM_BASE + 0x48c5e0 && (cpu->gpr(20) == 0xf3012002 || cpu->gpr(20) == 0xf3012000))
return SIGSEGV_RETURN_SKIP_INSTRUCTION;
else if (pc == ROM_BASE + 0x4a10a0 && (cpu->gpr(20) == 0xf3012002 || cpu->gpr(20) == 0xf3012000))
return SIGSEGV_RETURN_SKIP_INSTRUCTION;
// Ignore all other faults, if requested
if (PrefsFindBool("ignoresegv"))
return SIGSEGV_RETURN_SKIP_INSTRUCTION;
}
#else
#error "FIXME: You don't have the capability to skip instruction within signal handlers"
#endif
printf("SIGSEGV\n");
printf(" pc %p\n", fault_instruction);
printf(" ea %p\n", fault_address);
printf(" cpu %s\n", current_cpu == main_cpu ? "main" : "interrupts");
dump_registers();
current_cpu->dump_log();
enter_mon();
QuitEmulator();
return SIGSEGV_RETURN_FAILURE;
}
void init_emul_ppc(void)
{
// Initialize main CPU emulator
main_cpu = new sheepshaver_cpu();
main_cpu->set_register(powerpc_registers::GPR(3), any_register((uint32)ROM_BASE + 0x30d000));
main_cpu->set_register(powerpc_registers::GPR(4), any_register(KernelDataAddr + 0x1000));
WriteMacInt32(XLM_RUN_MODE, MODE_68K);
#if MULTICORE_CPU
// Initialize alternate CPU emulator to handle interrupts
interrupt_cpu = new sheepshaver_cpu();
#endif
// Install the handler for SIGSEGV
sigsegv_install_handler(sigsegv_handler);
#if ENABLE_MON
// Install "regs" command in cxmon
mon_add_command("regs", dump_registers, "regs Dump PowerPC registers\n");
mon_add_command("log", dump_log, "log Dump PowerPC emulation log\n");
#endif
#if EMUL_TIME_STATS
emul_start_time = clock();
#endif
}
/*
* Deinitialize emulation
*/
void exit_emul_ppc(void)
{
#if EMUL_TIME_STATS
clock_t emul_end_time = clock();
printf("### Statistics for SheepShaver emulation parts\n");
const clock_t emul_time = emul_end_time - emul_start_time;
printf("Total emulation time : %.1f sec\n", double(emul_time) / double(CLOCKS_PER_SEC));
printf("Total interrupt count: %d (%2.1f Hz)\n", interrupt_count,
(double(interrupt_count) * CLOCKS_PER_SEC) / double(emul_time));
#define PRINT_STATS(LABEL, VAR_PREFIX) do { \
printf("Total " LABEL " count : %d\n", VAR_PREFIX##_count); \
printf("Total " LABEL " time : %.1f sec (%.1f%%)\n", \
double(VAR_PREFIX##_time) / double(CLOCKS_PER_SEC), \
100.0 * double(VAR_PREFIX##_time) / double(emul_time)); \
} while (0)
PRINT_STATS("Execute68k[Trap] execution", exec68k);
PRINT_STATS("NativeOp execution", native_exec);
PRINT_STATS("MacOS routine execution", macos_exec);
#undef PRINT_STATS
printf("\n");
#endif
delete main_cpu;
#if MULTICORE_CPU
delete interrupt_cpu;
#endif
}
/*
* Emulation loop
*/
void emul_ppc(uint32 entry)
{
current_cpu = main_cpu;
#if 0
current_cpu->start_log();
#endif
// start emulation loop and enable code translation or caching
current_cpu->execute(entry);
}
/*
* Handle PowerPC interrupt
*/
#if ASYNC_IRQ
void HandleInterrupt(void)
{
main_cpu->handle_interrupt();
}
#else
void TriggerInterrupt(void)
{
#if 0
WriteMacInt32(0x16a, ReadMacInt32(0x16a) + 1);
#else
// Trigger interrupt to main cpu only
if (main_cpu)
main_cpu->trigger_interrupt();
#endif
}
#endif
void sheepshaver_cpu::handle_interrupt(void)
{
// Do nothing if interrupts are disabled
if (*(int32 *)XLM_IRQ_NEST > 0)
return;
// Do nothing if there is no interrupt pending
if (InterruptFlags == 0)
return;
// Disable MacOS stack sniffer
WriteMacInt32(0x110, 0);
// Interrupt action depends on current run mode
switch (ReadMacInt32(XLM_RUN_MODE)) {
case MODE_68K:
// 68k emulator active, trigger 68k interrupt level 1
assert(current_cpu == main_cpu);
WriteMacInt16(tswap32(kernel_data->v[0x67c >> 2]), 1);
set_cr(get_cr() | tswap32(kernel_data->v[0x674 >> 2]));
break;
#if INTERRUPTS_IN_NATIVE_MODE
case MODE_NATIVE:
// 68k emulator inactive, in nanokernel?
assert(current_cpu == main_cpu);
if (gpr(1) != KernelDataAddr) {
// Prepare for 68k interrupt level 1
WriteMacInt16(tswap32(kernel_data->v[0x67c >> 2]), 1);
WriteMacInt32(tswap32(kernel_data->v[0x658 >> 2]) + 0xdc,
ReadMacInt32(tswap32(kernel_data->v[0x658 >> 2]) + 0xdc)
| tswap32(kernel_data->v[0x674 >> 2]));
// Execute nanokernel interrupt routine (this will activate the 68k emulator)
DisableInterrupt();
cpu_push(interrupt_cpu);
if (ROMType == ROMTYPE_NEWWORLD)
current_cpu->interrupt(ROM_BASE + 0x312b1c);
else
current_cpu->interrupt(ROM_BASE + 0x312a3c);
cpu_pop();
}
break;
#endif
#if INTERRUPTS_IN_EMUL_OP_MODE
case MODE_EMUL_OP:
// 68k emulator active, within EMUL_OP routine, execute 68k interrupt routine directly when interrupt level is 0
if ((ReadMacInt32(XLM_68K_R25) & 7) == 0) {
#if 1
// Execute full 68k interrupt routine
M68kRegisters r;
uint32 old_r25 = ReadMacInt32(XLM_68K_R25); // Save interrupt level
WriteMacInt32(XLM_68K_R25, 0x21); // Execute with interrupt level 1
static const uint8 proc[] = {
0x3f, 0x3c, 0x00, 0x00, // move.w #$0000,-(sp) (fake format word)
0x48, 0x7a, 0x00, 0x0a, // pea @1(pc) (return address)
0x40, 0xe7, // move sr,-(sp) (saved SR)
0x20, 0x78, 0x00, 0x064, // move.l $64,a0
0x4e, 0xd0, // jmp (a0)
M68K_RTS >> 8, M68K_RTS & 0xff // @1
};
Execute68k((uint32)proc, &r);
WriteMacInt32(XLM_68K_R25, old_r25); // Restore interrupt level
#else
// Only update cursor
if (HasMacStarted()) {
if (InterruptFlags & INTFLAG_VIA) {
ClearInterruptFlag(INTFLAG_VIA);
ADBInterrupt();
ExecuteNative(NATIVE_VIDEO_VBL);
}
}
#endif
}
break;
#endif
}
}
static void get_resource(void);
static void get_1_resource(void);
static void get_ind_resource(void);
static void get_1_ind_resource(void);
static void r_get_resource(void);
#define GPR(REG) current_cpu->gpr(REG)
static void NativeOp(int selector)
{
#if EMUL_TIME_STATS
native_exec_count++;
const clock_t native_exec_start = clock();
#endif
switch (selector) {
case NATIVE_PATCH_NAME_REGISTRY:
DoPatchNameRegistry();
break;
case NATIVE_VIDEO_INSTALL_ACCEL:
VideoInstallAccel();
break;
case NATIVE_VIDEO_VBL:
VideoVBL();
break;
case NATIVE_VIDEO_DO_DRIVER_IO:
GPR(3) = (int32)(int16)VideoDoDriverIO((void *)GPR(3), (void *)GPR(4),
(void *)GPR(5), GPR(6), GPR(7));
break;
#ifdef WORDS_BIGENDIAN
case NATIVE_ETHER_IRQ:
EtherIRQ();
break;
case NATIVE_ETHER_INIT:
GPR(3) = InitStreamModule((void *)GPR(3));
break;
case NATIVE_ETHER_TERM:
TerminateStreamModule();
break;
case NATIVE_ETHER_OPEN:
GPR(3) = ether_open((queue_t *)GPR(3), (void *)GPR(4), GPR(5), GPR(6), (void*)GPR(7));
break;
case NATIVE_ETHER_CLOSE:
GPR(3) = ether_close((queue_t *)GPR(3), GPR(4), (void *)GPR(5));
break;
case NATIVE_ETHER_WPUT:
GPR(3) = ether_wput((queue_t *)GPR(3), (mblk_t *)GPR(4));
break;
case NATIVE_ETHER_RSRV:
GPR(3) = ether_rsrv((queue_t *)GPR(3));
break;
#else
case NATIVE_ETHER_INIT:
// FIXME: needs more complicated thunks
GPR(3) = false;
break;
#endif
case NATIVE_SERIAL_NOTHING:
case NATIVE_SERIAL_OPEN:
case NATIVE_SERIAL_PRIME_IN:
case NATIVE_SERIAL_PRIME_OUT:
case NATIVE_SERIAL_CONTROL:
case NATIVE_SERIAL_STATUS:
case NATIVE_SERIAL_CLOSE: {
typedef int16 (*SerialCallback)(uint32, uint32);
static const SerialCallback serial_callbacks[] = {
SerialNothing,
SerialOpen,
SerialPrimeIn,
SerialPrimeOut,
SerialControl,
SerialStatus,
SerialClose
};
GPR(3) = serial_callbacks[selector - NATIVE_SERIAL_NOTHING](GPR(3), GPR(4));
break;
}
case NATIVE_GET_RESOURCE:
case NATIVE_GET_1_RESOURCE:
case NATIVE_GET_IND_RESOURCE:
case NATIVE_GET_1_IND_RESOURCE:
case NATIVE_R_GET_RESOURCE: {
typedef void (*GetResourceCallback)(void);
static const GetResourceCallback get_resource_callbacks[] = {
get_resource,
get_1_resource,
get_ind_resource,
get_1_ind_resource,
r_get_resource
};
get_resource_callbacks[selector - NATIVE_GET_RESOURCE]();
break;
}
case NATIVE_DISABLE_INTERRUPT:
DisableInterrupt();
break;
case NATIVE_ENABLE_INTERRUPT:
EnableInterrupt();
break;
case NATIVE_MAKE_EXECUTABLE:
MakeExecutable(0, (void *)GPR(4), GPR(5));
break;
default:
printf("FATAL: NATIVE_OP called with bogus selector %d\n", selector);
QuitEmulator();
break;
}
#if EMUL_TIME_STATS
native_exec_time += (clock() - native_exec_start);
#endif
}
/*
* Execute 68k subroutine (must be ended with EXEC_RETURN)
* This must only be called by the emul_thread when in EMUL_OP mode
* r->a[7] is unused, the routine runs on the caller's stack
*/
void Execute68k(uint32 pc, M68kRegisters *r)
{
current_cpu->execute_68k(pc, r);
}
/*
* Execute 68k A-Trap from EMUL_OP routine
* r->a[7] is unused, the routine runs on the caller's stack
*/
void Execute68kTrap(uint16 trap, M68kRegisters *r)
{
SheepVar proc_var(4);
uint32 proc = proc_var.addr();
WriteMacInt16(proc, trap);
WriteMacInt16(proc + 2, M68K_RTS);
Execute68k(proc, r);
}
/*
* Call MacOS PPC code
*/
uint32 call_macos(uint32 tvect)
{
return current_cpu->execute_macos_code(tvect, 0, NULL);
}
uint32 call_macos1(uint32 tvect, uint32 arg1)
{
const uint32 args[] = { arg1 };
return current_cpu->execute_macos_code(tvect, sizeof(args)/sizeof(args[0]), args);
}
uint32 call_macos2(uint32 tvect, uint32 arg1, uint32 arg2)
{
const uint32 args[] = { arg1, arg2 };
return current_cpu->execute_macos_code(tvect, sizeof(args)/sizeof(args[0]), args);
}
uint32 call_macos3(uint32 tvect, uint32 arg1, uint32 arg2, uint32 arg3)
{
const uint32 args[] = { arg1, arg2, arg3 };
return current_cpu->execute_macos_code(tvect, sizeof(args)/sizeof(args[0]), args);
}
uint32 call_macos4(uint32 tvect, uint32 arg1, uint32 arg2, uint32 arg3, uint32 arg4)
{
const uint32 args[] = { arg1, arg2, arg3, arg4 };
return current_cpu->execute_macos_code(tvect, sizeof(args)/sizeof(args[0]), args);
}
uint32 call_macos5(uint32 tvect, uint32 arg1, uint32 arg2, uint32 arg3, uint32 arg4, uint32 arg5)
{
const uint32 args[] = { arg1, arg2, arg3, arg4, arg5 };
return current_cpu->execute_macos_code(tvect, sizeof(args)/sizeof(args[0]), args);
}
uint32 call_macos6(uint32 tvect, uint32 arg1, uint32 arg2, uint32 arg3, uint32 arg4, uint32 arg5, uint32 arg6)
{
const uint32 args[] = { arg1, arg2, arg3, arg4, arg5, arg6 };
return current_cpu->execute_macos_code(tvect, sizeof(args)/sizeof(args[0]), args);
}
uint32 call_macos7(uint32 tvect, uint32 arg1, uint32 arg2, uint32 arg3, uint32 arg4, uint32 arg5, uint32 arg6, uint32 arg7)
{
const uint32 args[] = { arg1, arg2, arg3, arg4, arg5, arg6, arg7 };
return current_cpu->execute_macos_code(tvect, sizeof(args)/sizeof(args[0]), args);
}
/*
* Resource Manager thunks
*/
void get_resource(void)
{
current_cpu->get_resource(ReadMacInt32(XLM_GET_RESOURCE));
}
void get_1_resource(void)
{
current_cpu->get_resource(ReadMacInt32(XLM_GET_1_RESOURCE));
}
void get_ind_resource(void)
{
current_cpu->get_resource(ReadMacInt32(XLM_GET_IND_RESOURCE));
}
void get_1_ind_resource(void)
{
current_cpu->get_resource(ReadMacInt32(XLM_GET_1_IND_RESOURCE));
}
void r_get_resource(void)
{
current_cpu->get_resource(ReadMacInt32(XLM_R_GET_RESOURCE));
}