added thumb mode emulation

This commit is contained in:
bunnei 2014-04-03 19:03:55 -04:00
parent c9b5b89e21
commit 7bf24c066b
6 changed files with 522 additions and 4 deletions

View file

@ -145,6 +145,7 @@
<ClCompile Include="src\arm\armvirt.cpp" /> <ClCompile Include="src\arm\armvirt.cpp" />
<ClCompile Include="src\arm\disassembler\arm_disasm.cpp" /> <ClCompile Include="src\arm\disassembler\arm_disasm.cpp" />
<ClCompile Include="src\arm\mmu\arm1176jzf_s_mmu.cpp" /> <ClCompile Include="src\arm\mmu\arm1176jzf_s_mmu.cpp" />
<ClCompile Include="src\arm\thumbemu.cpp" />
<ClCompile Include="src\core.cpp" /> <ClCompile Include="src\core.cpp" />
<ClCompile Include="src\core_timing.cpp" /> <ClCompile Include="src\core_timing.cpp" />
<ClCompile Include="src\elf\elf_reader.cpp" /> <ClCompile Include="src\elf\elf_reader.cpp" />

View file

@ -40,6 +40,9 @@
<ClCompile Include="src\arm\mmu\arm1176jzf_s_mmu.cpp"> <ClCompile Include="src\arm\mmu\arm1176jzf_s_mmu.cpp">
<Filter>arm\mmu</Filter> <Filter>arm\mmu</Filter>
</ClCompile> </ClCompile>
<ClCompile Include="src\arm\thumbemu.cpp">
<Filter>arm</Filter>
</ClCompile>
</ItemGroup> </ItemGroup>
<ItemGroup> <ItemGroup>
<Filter Include="arm"> <Filter Include="arm">

View file

@ -936,10 +936,10 @@ ARMul_Emulate26 (ARMul_State * state)
pipelined PC value is used when executing Thumb code, and also for pipelined PC value is used when executing Thumb code, and also for
dealing with the BL instruction. */ dealing with the BL instruction. */
if (TFLAG) { if (TFLAG) {
ARMword new; ARMword new_instr;
/* Check if in Thumb mode. */ /* Check if in Thumb mode. */
switch (ARMul_ThumbDecode (state, pc, instr, &new)) { switch (ARMul_ThumbDecode(state, pc, instr, &new_instr)) {
case t_undefined: case t_undefined:
/* This is a Thumb instruction. */ /* This is a Thumb instruction. */
ARMul_UndefInstr (state, instr); ARMul_UndefInstr (state, instr);
@ -952,7 +952,7 @@ ARMul_Emulate26 (ARMul_State * state)
case t_decoded: case t_decoded:
/* ARM instruction available. */ /* ARM instruction available. */
//printf("t decode %04lx -> %08lx\n", instr & 0xffff, new); //printf("t decode %04lx -> %08lx\n", instr & 0xffff, new);
instr = new; instr = new_instr;
/* So continue instruction decoding. */ /* So continue instruction decoding. */
break; break;
default: default:

View file

@ -4,6 +4,7 @@
#include "common.h" #include "common.h"
#define MODE32 #define MODE32
#define MODET
typedef struct typedef struct
{ {

View file

@ -0,0 +1,513 @@
/* thumbemu.c -- Thumb instruction emulation.
Copyright (C) 1996, Cygnus Software Technologies Ltd.
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. */
/* We can provide simple Thumb simulation by decoding the Thumb
instruction into its corresponding ARM instruction, and using the
existing ARM simulator. */
#include "skyeye_defs.h"
#ifndef MODET /* required for the Thumb instruction support */
#if 1
#error "MODET needs to be defined for the Thumb world to work"
#else
#define MODET (1)
#endif
#endif
#include "armdefs.h"
#include "armemu.h"
#include "armos.h"
/* Decode a 16bit Thumb instruction. The instruction is in the low
16-bits of the tinstr field, with the following Thumb instruction
held in the high 16-bits. Passing in two Thumb instructions allows
easier simulation of the special dual BL instruction. */
tdstate
ARMul_ThumbDecode (
ARMul_State *state,
ARMword pc,
ARMword tinstr,
ARMword *ainstr)
{
tdstate valid = t_decoded; /* default assumes a valid instruction */
ARMword next_instr;
if (state->bigendSig) {
next_instr = tinstr & 0xFFFF;
tinstr >>= 16;
}
else {
next_instr = tinstr >> 16;
tinstr &= 0xFFFF;
}
#if 1 /* debugging to catch non updates */
*ainstr = 0xDEADC0DE;
#endif
switch ((tinstr & 0xF800) >> 11) {
case 0: /* LSL */
case 1: /* LSR */
case 2: /* ASR */
/* Format 1 */
*ainstr = 0xE1B00000 /* base opcode */
| ((tinstr & 0x1800) >> (11 - 5)) /* shift type */
|((tinstr & 0x07C0) << (7 - 6)) /* imm5 */
|((tinstr & 0x0038) >> 3) /* Rs */
|((tinstr & 0x0007) << 12); /* Rd */
break;
case 3: /* ADD/SUB */
/* Format 2 */
{
ARMword subset[4] = {
0xE0900000, /* ADDS Rd,Rs,Rn */
0xE0500000, /* SUBS Rd,Rs,Rn */
0xE2900000, /* ADDS Rd,Rs,#imm3 */
0xE2500000 /* SUBS Rd,Rs,#imm3 */
};
/* It is quicker indexing into a table, than performing switch
or conditionals: */
*ainstr = subset[(tinstr & 0x0600) >> 9] /* base opcode */
|((tinstr & 0x01C0) >> 6) /* Rn or imm3 */
|((tinstr & 0x0038) << (16 - 3)) /* Rs */
|((tinstr & 0x0007) << (12 - 0)); /* Rd */
}
break;
case 4: /* MOV */
case 5: /* CMP */
case 6: /* ADD */
case 7: /* SUB */
/* Format 3 */
{
ARMword subset[4] = {
0xE3B00000, /* MOVS Rd,#imm8 */
0xE3500000, /* CMP Rd,#imm8 */
0xE2900000, /* ADDS Rd,Rd,#imm8 */
0xE2500000, /* SUBS Rd,Rd,#imm8 */
};
*ainstr = subset[(tinstr & 0x1800) >> 11] /* base opcode */
|((tinstr & 0x00FF) >> 0) /* imm8 */
|((tinstr & 0x0700) << (16 - 8)) /* Rn */
|((tinstr & 0x0700) << (12 - 8)); /* Rd */
}
break;
case 8: /* Arithmetic and high register transfers */
/* TODO: Since the subsets for both Format 4 and Format 5
instructions are made up of different ARM encodings, we could
save the following conditional, and just have one large
subset. */
if ((tinstr & (1 << 10)) == 0) {
/* Format 4 */
enum OpcodeType { t_norm, t_shift, t_neg, t_mul };
struct ThumbOpcode {
ARMword opcode;
OpcodeType otype;
};
ThumbOpcode subset[16] = {
{
0xE0100000, t_norm}, /* ANDS Rd,Rd,Rs */
{
0xE0300000, t_norm}, /* EORS Rd,Rd,Rs */
{
0xE1B00010, t_shift}, /* MOVS Rd,Rd,LSL Rs */
{
0xE1B00030, t_shift}, /* MOVS Rd,Rd,LSR Rs */
{
0xE1B00050, t_shift}, /* MOVS Rd,Rd,ASR Rs */
{
0xE0B00000, t_norm}, /* ADCS Rd,Rd,Rs */
{
0xE0D00000, t_norm}, /* SBCS Rd,Rd,Rs */
{
0xE1B00070, t_shift}, /* MOVS Rd,Rd,ROR Rs */
{
0xE1100000, t_norm}, /* TST Rd,Rs */
{
0xE2700000, t_neg}, /* RSBS Rd,Rs,#0 */
{
0xE1500000, t_norm}, /* CMP Rd,Rs */
{
0xE1700000, t_norm}, /* CMN Rd,Rs */
{
0xE1900000, t_norm}, /* ORRS Rd,Rd,Rs */
{
0xE0100090, t_mul}, /* MULS Rd,Rd,Rs */
{
0xE1D00000, t_norm}, /* BICS Rd,Rd,Rs */
{
0xE1F00000, t_norm} /* MVNS Rd,Rs */
};
*ainstr = subset[(tinstr & 0x03C0) >> 6].opcode; /* base */
switch (subset[(tinstr & 0x03C0) >> 6].otype) {
case t_norm:
*ainstr |= ((tinstr & 0x0007) << 16) /* Rn */
|((tinstr & 0x0007) << 12) /* Rd */
|((tinstr & 0x0038) >> 3); /* Rs */
break;
case t_shift:
*ainstr |= ((tinstr & 0x0007) << 12) /* Rd */
|((tinstr & 0x0007) >> 0) /* Rm */
|((tinstr & 0x0038) << (8 - 3)); /* Rs */
break;
case t_neg:
*ainstr |= ((tinstr & 0x0007) << 12) /* Rd */
|((tinstr & 0x0038) << (16 - 3)); /* Rn */
break;
case t_mul:
*ainstr |= ((tinstr & 0x0007) << 16) /* Rd */
|((tinstr & 0x0007) << 8) /* Rs */
|((tinstr & 0x0038) >> 3); /* Rm */
break;
}
}
else {
/* Format 5 */
ARMword Rd = ((tinstr & 0x0007) >> 0);
ARMword Rs = ((tinstr & 0x0038) >> 3);
if (tinstr & (1 << 7))
Rd += 8;
if (tinstr & (1 << 6))
Rs += 8;
switch ((tinstr & 0x03C0) >> 6) {
case 0x1: /* ADD Rd,Rd,Hs */
case 0x2: /* ADD Hd,Hd,Rs */
case 0x3: /* ADD Hd,Hd,Hs */
*ainstr = 0xE0800000 /* base */
| (Rd << 16) /* Rn */
|(Rd << 12) /* Rd */
|(Rs << 0); /* Rm */
break;
case 0x5: /* CMP Rd,Hs */
case 0x6: /* CMP Hd,Rs */
case 0x7: /* CMP Hd,Hs */
*ainstr = 0xE1500000 /* base */
| (Rd << 16) /* Rn */
|(Rd << 12) /* Rd */
|(Rs << 0); /* Rm */
break;
case 0x9: /* MOV Rd,Hs */
case 0xA: /* MOV Hd,Rs */
case 0xB: /* MOV Hd,Hs */
*ainstr = 0xE1A00000 /* base */
| (Rd << 16) /* Rn */
|(Rd << 12) /* Rd */
|(Rs << 0); /* Rm */
break;
case 0xC: /* BX Rs */
case 0xD: /* BX Hs */
*ainstr = 0xE12FFF10 /* base */
| ((tinstr & 0x0078) >> 3); /* Rd */
break;
case 0x0: /* UNDEFINED */
case 0x4: /* UNDEFINED */
case 0x8: /* UNDEFINED */
valid = t_undefined;
break;
case 0xE: /* BLX */
case 0xF: /* BLX */
if (state->is_v5) {
*ainstr = 0xE1200030 /* base */
|(Rs << 0); /* Rm */
} else {
valid = t_undefined;
}
break;
}
}
break;
case 9: /* LDR Rd,[PC,#imm8] */
/* Format 6 */
*ainstr = 0xE59F0000 /* base */
| ((tinstr & 0x0700) << (12 - 8)) /* Rd */
|((tinstr & 0x00FF) << (2 - 0)); /* off8 */
break;
case 10:
case 11:
/* TODO: Format 7 and Format 8 perform the same ARM encoding, so
the following could be merged into a single subset, saving on
the following boolean: */
if ((tinstr & (1 << 9)) == 0) {
/* Format 7 */
ARMword subset[4] = {
0xE7800000, /* STR Rd,[Rb,Ro] */
0xE7C00000, /* STRB Rd,[Rb,Ro] */
0xE7900000, /* LDR Rd,[Rb,Ro] */
0xE7D00000 /* LDRB Rd,[Rb,Ro] */
};
*ainstr = subset[(tinstr & 0x0C00) >> 10] /* base */
|((tinstr & 0x0007) << (12 - 0)) /* Rd */
|((tinstr & 0x0038) << (16 - 3)) /* Rb */
|((tinstr & 0x01C0) >> 6); /* Ro */
}
else {
/* Format 8 */
ARMword subset[4] = {
0xE18000B0, /* STRH Rd,[Rb,Ro] */
0xE19000D0, /* LDRSB Rd,[Rb,Ro] */
0xE19000B0, /* LDRH Rd,[Rb,Ro] */
0xE19000F0 /* LDRSH Rd,[Rb,Ro] */
};
*ainstr = subset[(tinstr & 0x0C00) >> 10] /* base */
|((tinstr & 0x0007) << (12 - 0)) /* Rd */
|((tinstr & 0x0038) << (16 - 3)) /* Rb */
|((tinstr & 0x01C0) >> 6); /* Ro */
}
break;
case 12: /* STR Rd,[Rb,#imm5] */
case 13: /* LDR Rd,[Rb,#imm5] */
case 14: /* STRB Rd,[Rb,#imm5] */
case 15: /* LDRB Rd,[Rb,#imm5] */
/* Format 9 */
{
ARMword subset[4] = {
0xE5800000, /* STR Rd,[Rb,#imm5] */
0xE5900000, /* LDR Rd,[Rb,#imm5] */
0xE5C00000, /* STRB Rd,[Rb,#imm5] */
0xE5D00000 /* LDRB Rd,[Rb,#imm5] */
};
/* The offset range defends on whether we are transferring a
byte or word value: */
*ainstr = subset[(tinstr & 0x1800) >> 11] /* base */
|((tinstr & 0x0007) << (12 - 0)) /* Rd */
|((tinstr & 0x0038) << (16 - 3)) /* Rb */
|((tinstr & 0x07C0) >> (6 - ((tinstr & (1 << 12)) ? 0 : 2))); /* off5 */
}
break;
case 16: /* STRH Rd,[Rb,#imm5] */
case 17: /* LDRH Rd,[Rb,#imm5] */
/* Format 10 */
*ainstr = ((tinstr & (1 << 11)) /* base */
? 0xE1D000B0 /* LDRH */
: 0xE1C000B0) /* STRH */
|((tinstr & 0x0007) << (12 - 0)) /* Rd */
|((tinstr & 0x0038) << (16 - 3)) /* Rb */
|((tinstr & 0x01C0) >> (6 - 1)) /* off5, low nibble */
|((tinstr & 0x0600) >> (9 - 8)); /* off5, high nibble */
break;
case 18: /* STR Rd,[SP,#imm8] */
case 19: /* LDR Rd,[SP,#imm8] */
/* Format 11 */
*ainstr = ((tinstr & (1 << 11)) /* base */
? 0xE59D0000 /* LDR */
: 0xE58D0000) /* STR */
|((tinstr & 0x0700) << (12 - 8)) /* Rd */
|((tinstr & 0x00FF) << 2); /* off8 */
break;
case 20: /* ADD Rd,PC,#imm8 */
case 21: /* ADD Rd,SP,#imm8 */
/* Format 12 */
if ((tinstr & (1 << 11)) == 0) {
/* NOTE: The PC value used here should by word aligned */
/* We encode shift-left-by-2 in the rotate immediate field,
so no shift of off8 is needed. */
*ainstr = 0xE28F0F00 /* base */
| ((tinstr & 0x0700) << (12 - 8)) /* Rd */
|(tinstr & 0x00FF); /* off8 */
}
else {
/* We encode shift-left-by-2 in the rotate immediate field,
so no shift of off8 is needed. */
*ainstr = 0xE28D0F00 /* base */
| ((tinstr & 0x0700) << (12 - 8)) /* Rd */
|(tinstr & 0x00FF); /* off8 */
}
break;
case 22:
case 23:
if ((tinstr & 0x0F00) == 0x0000) {
/* Format 13 */
/* NOTE: The instruction contains a shift left of 2
equivalent (implemented as ROR #30): */
*ainstr = ((tinstr & (1 << 7)) /* base */
? 0xE24DDF00 /* SUB */
: 0xE28DDF00) /* ADD */
|(tinstr & 0x007F); /* off7 */
}
else if ((tinstr & 0x0F00) == 0x0e00)
*ainstr = 0xEF000000 | SWI_Breakpoint;
else {
/* Format 14 */
ARMword subset[4] = {
0xE92D0000, /* STMDB sp!,{rlist} */
0xE92D4000, /* STMDB sp!,{rlist,lr} */
0xE8BD0000, /* LDMIA sp!,{rlist} */
0xE8BD8000 /* LDMIA sp!,{rlist,pc} */
};
*ainstr = subset[((tinstr & (1 << 11)) >> 10) | ((tinstr & (1 << 8)) >> 8)] /* base */
|(tinstr & 0x00FF); /* mask8 */
}
break;
case 24: /* STMIA */
case 25: /* LDMIA */
/* Format 15 */
*ainstr = ((tinstr & (1 << 11)) /* base */
? 0xE8B00000 /* LDMIA */
: 0xE8A00000) /* STMIA */
|((tinstr & 0x0700) << (16 - 8)) /* Rb */
|(tinstr & 0x00FF); /* mask8 */
break;
case 26: /* Bcc */
case 27: /* Bcc/SWI */
if ((tinstr & 0x0F00) == 0x0F00) {
if (tinstr == (ARMul_ABORTWORD & 0xffff) &&
state->AbortAddr == pc) {
*ainstr = ARMul_ABORTWORD;
break;
}
/* Format 17 : SWI */
*ainstr = 0xEF000000;
/* Breakpoint must be handled specially. */
if ((tinstr & 0x00FF) == 0x18)
*ainstr |= ((tinstr & 0x00FF) << 16);
/* New breakpoint value. See gdb/arm-tdep.c */
else if ((tinstr & 0x00FF) == 0xFE)
*ainstr |= SWI_Breakpoint;
else
*ainstr |= (tinstr & 0x00FF);
}
else if ((tinstr & 0x0F00) != 0x0E00) {
/* Format 16 */
int doit = FALSE;
/* TODO: Since we are doing a switch here, we could just add
the SWI and undefined instruction checks into this
switch to same on a couple of conditionals: */
switch ((tinstr & 0x0F00) >> 8) {
case EQ:
doit = ZFLAG;
break;
case NE:
doit = !ZFLAG;
break;
case VS:
doit = VFLAG;
break;
case VC:
doit = !VFLAG;
break;
case MI:
doit = NFLAG;
break;
case PL:
doit = !NFLAG;
break;
case CS:
doit = CFLAG;
break;
case CC:
doit = !CFLAG;
break;
case HI:
doit = (CFLAG && !ZFLAG);
break;
case LS:
doit = (!CFLAG || ZFLAG);
break;
case GE:
doit = ((!NFLAG && !VFLAG)
|| (NFLAG && VFLAG));
break;
case LT:
doit = ((NFLAG && !VFLAG)
|| (!NFLAG && VFLAG));
break;
case GT:
doit = ((!NFLAG && !VFLAG && !ZFLAG)
|| (NFLAG && VFLAG && !ZFLAG));
break;
case LE:
doit = ((NFLAG && !VFLAG)
|| (!NFLAG && VFLAG)) || ZFLAG;
break;
}
if (doit) {
state->Reg[15] = (pc + 4
+ (((tinstr & 0x7F) << 1)
| ((tinstr & (1 << 7)) ?
0xFFFFFF00 : 0)));
FLUSHPIPE;
}
valid = t_branch;
}
else /* UNDEFINED : cc=1110(AL) uses different format */
valid = t_undefined;
break;
case 28: /* B */
/* Format 18 */
state->Reg[15] = (pc + 4 + (((tinstr & 0x3FF) << 1)
| ((tinstr & (1 << 10)) ?
0xFFFFF800 : 0)));
FLUSHPIPE;
valid = t_branch;
break;
case 29:
if(tinstr & 0x1)
valid = t_undefined;
else{
/* BLX 1 for armv5t and above */
ARMword tmp = (pc + 2);
state->Reg[15] =
(state->Reg[14] + ((tinstr & 0x07FF) << 1)) & 0xFFFFFFFC;
state->Reg[14] = (tmp | 1);
CLEART;
DEBUG_LOG(ARM11, "In %s, After BLX(1),LR=0x%x,PC=0x%x, offset=0x%x\n", __FUNCTION__, state->Reg[14], state->Reg[15], (tinstr &0x7FF) << 1);
valid = t_branch;
FLUSHPIPE;
}
break;
case 30: /* BL instruction 1 */
/* Format 19 */
/* There is no single ARM instruction equivalent for this Thumb
instruction. To keep the simulation simple (from the user
perspective) we check if the following instruction is the
second half of this BL, and if it is we simulate it
immediately. */
state->Reg[14] = state->Reg[15]
+ (((tinstr & 0x07FF) << 12)
| ((tinstr & (1 << 10)) ? 0xFF800000 : 0));
valid = t_branch; /* in-case we don't have the 2nd half */
//tinstr = next_instr; /* move the instruction down */
//if (((tinstr & 0xF800) >> 11) != 31)
// break; /* exit, since not correct instruction */
/* else we fall through to process the second half of the BL */
//pc += 2; /* point the pc at the 2nd half */
state->Reg[15] = pc + 2;
FLUSHPIPE;
break;
case 31: /* BL instruction 2 */
/* Format 19 */
/* There is no single ARM instruction equivalent for this
instruction. Also, it should only ever be matched with the
fmt19 "BL instruction 1" instruction. However, we do allow
the simulation of it on its own, with undefined results if
r14 is not suitably initialised. */
{
ARMword tmp = (pc + 2);
state->Reg[15] =
(state->Reg[14] + ((tinstr & 0x07FF) << 1));
state->Reg[14] = (tmp | 1);
valid = t_branch;
FLUSHPIPE;
}
break;
}
return valid;
}