1
0
Fork 0

Add support for apple m1

The apple m1 uses position independent executables (pie).
I have implemented this in tccmacho.c

Apple also uses the stack different for var_args.
Also characters are signed instead of unsigned.
This is implemented in arm64-gen.c/tccdefs.h

Add bounds checking lib to lib/Makefile.

Add underscore support in lib/atomic.S and lib/fetch_and_add.S

Disable __clear_cache in lib/lib-arm64.c (Use system version).
I will try to fix this in future push.

Disable test_asm_call in tests/tcctest.c. Clang does not support @plt.
Also disable weak symbols test.
I will try to fix weak support in future push.

Disable tests/tests2/124_atomic_counter.c for 64BITS.
This is a bug in the atomic code and will be fixed in future push.

You have to use --dwarf configure option. stabs only works with -run.

tested on apple x86_64(10.5) and arm64(12.3).
This commit is contained in:
herman ten brugge 2022-11-16 12:52:51 -06:00
parent c6b722f3f8
commit c8ef84c854
10 changed files with 736 additions and 129 deletions

View File

@ -38,7 +38,9 @@
#define MAX_ALIGN 16
#ifndef TCC_TARGET_MACHO
#define CHAR_IS_UNSIGNED
#endif
/* define if return values need to be extended explicitely
at caller side (for interfacing with non-TCC compilers) */
@ -810,7 +812,7 @@ static int arm64_hfa(CType *type, unsigned *fsize)
return 0;
}
static unsigned long arm64_pcs_aux(int n, CType **type, unsigned long *a)
static unsigned long arm64_pcs_aux(int variadic, int n, CType **type, unsigned long *a)
{
int nx = 0; // next integer register
int nv = 0; // next vector register
@ -827,6 +829,12 @@ static unsigned long arm64_pcs_aux(int n, CType **type, unsigned long *a)
else
size = type_size(type[i], &align);
#if defined(__APPLE__)
if (variadic && i == variadic) {
nx = 8;
nv = 8;
}
#endif
if (hfa)
// B.2
;
@ -931,7 +939,7 @@ static unsigned long arm64_pcs_aux(int n, CType **type, unsigned long *a)
return ns - 32;
}
static unsigned long arm64_pcs(int n, CType **type, unsigned long *a)
static unsigned long arm64_pcs(int variadic, int n, CType **type, unsigned long *a)
{
unsigned long stack;
@ -939,12 +947,12 @@ static unsigned long arm64_pcs(int n, CType **type, unsigned long *a)
if ((type[0]->t & VT_BTYPE) == VT_VOID)
a[0] = -1;
else {
arm64_pcs_aux(1, type, a);
arm64_pcs_aux(0, 1, type, a);
assert(a[0] == 0 || a[0] == 1 || a[0] == 16);
}
// Argument types:
stack = arm64_pcs_aux(n, type + 1, a + 1);
stack = arm64_pcs_aux(variadic, n, type + 1, a + 1);
if (0) {
int i;
@ -970,6 +978,16 @@ static unsigned long arm64_pcs(int n, CType **type, unsigned long *a)
return stack;
}
static int n_func_args(CType *type)
{
int n_args = 0;
Sym *arg;
for (arg = type->ref->next; arg; arg = arg->next)
n_args++;
return n_args;
}
ST_FUNC void gfunc_call(int nb_args)
{
CType *return_type;
@ -977,6 +995,8 @@ ST_FUNC void gfunc_call(int nb_args)
unsigned long *a, *a1;
unsigned long stack;
int i;
int variadic = (vtop[-nb_args].type.ref->f.func_type == FUNC_ELLIPSIS);
int var_nb_arg = n_func_args(&vtop[-nb_args].type);
#ifdef CONFIG_TCC_BCHECK
if (tcc_state->do_bounds_check)
@ -995,7 +1015,7 @@ ST_FUNC void gfunc_call(int nb_args)
for (i = 0; i < nb_args; i++)
t[nb_args - i] = &vtop[-i].type;
stack = arm64_pcs(nb_args, t, a);
stack = arm64_pcs(variadic ? var_nb_arg : 0, nb_args, t, a);
// Allocate space for structs replaced by pointer:
for (i = nb_args; i; i--)
@ -1055,7 +1075,7 @@ ST_FUNC void gfunc_call(int nb_args)
}
else {
gv(RC_INT);
arm64_strx(arm64_type_size(vtop[0].type.t),
arm64_strx(3, // arm64_type_size(vtop[0].type.t),
intr(vtop[0].r), 31, a[i] - 32);
}
}
@ -1167,6 +1187,8 @@ ST_FUNC void gfunc_prolog(Sym *func_sym)
int use_x8 = 0;
int last_int = 0;
int last_float = 0;
int variadic = func_sym->type.ref->f.func_type == FUNC_ELLIPSIS;
int var_nb_arg = n_func_args(&func_sym->type);
func_vc = 144; // offset of where x8 is stored
@ -1178,9 +1200,9 @@ ST_FUNC void gfunc_prolog(Sym *func_sym)
for (sym = func_type->ref; sym; sym = sym->next)
t[i++] = &sym->type;
arm64_func_va_list_stack = arm64_pcs(n - 1, t, a);
arm64_func_va_list_stack = arm64_pcs(variadic ? var_nb_arg : 0, n - 1, t, a);
if (func_sym->type.ref->f.func_type == FUNC_ELLIPSIS) {
if (variadic) {
use_x8 = 1;
last_int = 4;
last_float = 4;
@ -1282,6 +1304,7 @@ ST_FUNC void gen_va_start(void)
o(0x910383be); // add x30,x29,#224
o(0xf900001e | r << 5); // str x30,[x(r)]
#if !defined(__APPLE__)
if (arm64_func_va_list_gr_offs) {
if (arm64_func_va_list_stack)
o(0x910383be); // add x30,x29,#224
@ -1298,6 +1321,7 @@ ST_FUNC void gen_va_start(void)
arm64_movimm(30, arm64_func_va_list_vr_offs);
o(0xb9001c1e | r << 5); // str w30,[x(r),#28]
#endif
--vtop;
}
@ -1321,6 +1345,7 @@ ST_FUNC void gen_va_arg(CType *t)
if (!hfa) {
uint32_t n = size > 16 ? 8 : (size + 7) & -8;
#if !defined(__APPLE__)
o(0xb940181e | r0 << 5); // ldr w30,[x(r0),#24] // __gr_offs
if (align == 16) {
assert(0); // this path untested but needed for __uint128_t
@ -1329,23 +1354,28 @@ ST_FUNC void gen_va_arg(CType *t)
}
o(0x310003c0 | r1 | n << 10); // adds w(r1),w30,#(n)
o(0x540000ad); // b.le .+20
#endif
o(0xf9400000 | r1 | r0 << 5); // ldr x(r1),[x(r0)] // __stack
o(0x9100001e | r1 << 5 | n << 10); // add x30,x(r1),#(n)
o(0xf900001e | r0 << 5); // str x30,[x(r0)] // __stack
#if !defined(__APPLE__)
o(0x14000004); // b .+16
o(0xb9001800 | r1 | r0 << 5); // str w(r1),[x(r0),#24] // __gr_offs
o(0xf9400400 | r1 | r0 << 5); // ldr x(r1),[x(r0),#8] // __gr_top
o(0x8b3ec000 | r1 | r1 << 5); // add x(r1),x(r1),w30,sxtw
#endif
if (size > 16)
o(0xf9400000 | r1 | r1 << 5); // ldr x(r1),[x(r1)]
}
else {
uint32_t rsz = hfa << 4;
uint32_t ssz = (size + 7) & -(uint32_t)8;
#if !defined(__APPLE__)
uint32_t rsz = hfa << 4;
uint32_t b1, b2;
o(0xb9401c1e | r0 << 5); // ldr w30,[x(r0),#28] // __vr_offs
o(0x310003c0 | r1 | rsz << 10); // adds w(r1),w30,#(rsz)
b1 = ind; o(0x5400000d); // b.le lab1
#endif
o(0xf9400000 | r1 | r0 << 5); // ldr x(r1),[x(r0)] // __stack
if (fsize == 16) {
o(0x91003c00 | r1 | r1 << 5); // add x(r1),x(r1),#15
@ -1353,6 +1383,7 @@ ST_FUNC void gen_va_arg(CType *t)
}
o(0x9100001e | r1 << 5 | ssz << 10); // add x30,x(r1),#(ssz)
o(0xf900001e | r0 << 5); // str x30,[x(r0)] // __stack
#if !defined(__APPLE__)
b2 = ind; o(0x14000000); // b lab2
// lab1:
write32le(cur_text_section->data + b1, 0x5400000d | (ind - b1) << 3);
@ -1376,6 +1407,7 @@ ST_FUNC void gen_va_arg(CType *t)
}
// lab2:
write32le(cur_text_section->data + b2, 0x14000000 | (ind - b2) >> 2);
#endif
}
}
@ -1390,7 +1422,7 @@ ST_FUNC void gfunc_return(CType *func_type)
CType *t = func_type;
unsigned long a;
arm64_pcs(0, &t, &a);
arm64_pcs(0, 0, &t, &a);
switch (a) {
case -1:
break;

View File

@ -207,11 +207,18 @@
&~3), *(type *)(ap - ((sizeof(type)+3)&~3)))
#elif defined __aarch64__
#if defined __APPLE__
typedef struct {
void *__stack;
} __builtin_va_list;
#else
typedef struct {
void *__stack, *__gr_top, *__vr_top;
int __gr_offs, __vr_offs;
} __builtin_va_list;
#endif
#elif defined __riscv
typedef char *__builtin_va_list;
#define __va_reg_size (__riscv_xlen >> 3)

View File

@ -55,7 +55,7 @@ OBJ-x86_64-osx = $(X86_64_O) va_list.o $(BCHECK_O)
OBJ-i386-win32 = $(I386_O) chkstk.o $(B_O) $(WIN_O)
OBJ-x86_64-win32 = $(X86_64_O) chkstk.o $(B_O) $(WIN_O)
OBJ-arm64 = $(ARM64_O) $(BCHECK_O) $(DSO_O)
OBJ-arm64-osx = $(ARM64_O) va_list.o
OBJ-arm64-osx = $(ARM64_O) $(BCHECK_O)
OBJ-arm = $(ARM_O) $(BCHECK_O) $(DSO_O)
OBJ-arm-fpa = $(ARM_O) $(DSO_O)
OBJ-arm-fpa-ld = $(ARM_O) $(DSO_O)

View File

@ -5,6 +5,13 @@
* __atomic_compare_exchange_4
* __atomic_compare_exchange_8
*/
#ifdef __leading_underscore
# define _(s) _##s
#else
# define _(s) s
#endif
#if defined __arm__
#ifndef __TINYC__
@ -14,9 +21,9 @@
.text
.align 2
.global __atomic_compare_exchange_1
.type __atomic_compare_exchange_1, %function
__atomic_compare_exchange_1:
.global _(__atomic_compare_exchange_1)
.type _(__atomic_compare_exchange_1), %function
_(__atomic_compare_exchange_1):
#ifdef __TINYC__
.int 0xe52de004
.int 0xe5d13000
@ -50,11 +57,11 @@ __atomic_compare_exchange_1:
strbne ip, [r1]
ldr pc, [sp], #4
#endif
.size __atomic_compare_exchange_1, .-__atomic_compare_exchange_1
.size _(__atomic_compare_exchange_1), .-_(__atomic_compare_exchange_1)
.global __atomic_compare_exchange_2
.type __atomic_compare_exchange_2, %function
__atomic_compare_exchange_2:
.global _(__atomic_compare_exchange_2)
.type _(__atomic_compare_exchange_2), %function
_(__atomic_compare_exchange_2):
#ifdef __TINYC__
.int 0xe52de004
.int 0xe1d130b0
@ -88,11 +95,11 @@ __atomic_compare_exchange_2:
strhne ip, [r1]
ldr pc, [sp], #4
#endif
.size __atomic_compare_exchange_2, .-__atomic_compare_exchange_2
.size _(__atomic_compare_exchange_2), .-_(__atomic_compare_exchange_2)
.global __atomic_compare_exchange_4
.type __atomic_compare_exchange_4, %function
__atomic_compare_exchange_4:
.global _(__atomic_compare_exchange_4)
.type _(__atomic_compare_exchange_4), %function
_(__atomic_compare_exchange_4):
#ifdef __TINYC__
.int 0xe52de004
.int 0xe5913000
@ -126,7 +133,7 @@ __atomic_compare_exchange_4:
strne ip, [r1]
ldr pc, [sp], #4
#endif
.size __atomic_compare_exchange_4, .-__atomic_compare_exchange_4
.size _(__atomic_compare_exchange_4), .-_(__atomic_compare_exchange_4)
/* ---------------------------------------------- */
#elif defined __aarch64__
@ -134,9 +141,9 @@ __atomic_compare_exchange_4:
.text
.align 2
.global __atomic_compare_exchange_1
.type __atomic_compare_exchange_1, %function
__atomic_compare_exchange_1:
.global _(__atomic_compare_exchange_1)
.type _(__atomic_compare_exchange_1), %function
_(__atomic_compare_exchange_1):
#ifdef __TINYC__
.int 0xa9be7bfd
.int 0x910003fd
@ -187,11 +194,11 @@ __atomic_compare_exchange_1:
ldp x29, x30, [sp], 32
ret
#endif
.size __atomic_compare_exchange_1, .-__atomic_compare_exchange_1
.size _(__atomic_compare_exchange_1), .-_(__atomic_compare_exchange_1)
.global __atomic_compare_exchange_2
.type __atomic_compare_exchange_2, %function
__atomic_compare_exchange_2:
.global _(__atomic_compare_exchange_2)
.type _(__atomic_compare_exchange_2), %function
_(__atomic_compare_exchange_2):
#ifdef __TINYC__
.int 0xa9be7bfd
.int 0x910003fd
@ -242,11 +249,11 @@ __atomic_compare_exchange_2:
ldp x29, x30, [sp], 32
ret
#endif
.size __atomic_compare_exchange_2, .-__atomic_compare_exchange_2
.size _(__atomic_compare_exchange_2), .-_(__atomic_compare_exchange_2)
.global __atomic_compare_exchange_4
.type __atomic_compare_exchange_4, %function
__atomic_compare_exchange_4:
.global _(__atomic_compare_exchange_4)
.type _(__atomic_compare_exchange_4), %function
_(__atomic_compare_exchange_4):
#ifdef __TINYC__
.int 0xa9be7bfd
.int 0x910003fd
@ -297,11 +304,11 @@ __atomic_compare_exchange_4:
ldp x29, x30, [sp], 32
ret
#endif
.size __atomic_compare_exchange_4, .-__atomic_compare_exchange_4
.size _(__atomic_compare_exchange_4), .-_(__atomic_compare_exchange_4)
.global __atomic_compare_exchange_8
.type __atomic_compare_exchange_8, %function
__atomic_compare_exchange_8:
.global _(__atomic_compare_exchange_8)
.type _(__atomic_compare_exchange_8), %function
_(__atomic_compare_exchange_8):
#ifdef __TINYC__
.int 0xa9be7bfd
.int 0x910003fd
@ -352,7 +359,7 @@ __atomic_compare_exchange_8:
ldp x29, x30, [sp], 32
ret
#endif
.size __atomic_compare_exchange_8, .-__atomic_compare_exchange_8
.size _(__atomic_compare_exchange_8), .-_(__atomic_compare_exchange_8)
/* ---------------------------------------------- */
#elif defined __riscv
@ -360,9 +367,9 @@ __atomic_compare_exchange_8:
.text
.align 2
.global __atomic_compare_exchange_1
.type __atomic_compare_exchange_1, %function
__atomic_compare_exchange_1:
.global _(__atomic_compare_exchange_1)
.type _(__atomic_compare_exchange_1), %function
_(__atomic_compare_exchange_1):
#ifdef __TINYC__
.short 0x1141
.short 0x86ba
@ -443,11 +450,11 @@ __atomic_compare_exchange_1:
addi sp,sp,16
jr ra
#endif
.size __atomic_compare_exchange_1, .-__atomic_compare_exchange_1
.size _(__atomic_compare_exchange_1), .-_(__atomic_compare_exchange_1)
.global __atomic_compare_exchange_2
.type __atomic_compare_exchange_2, %function
__atomic_compare_exchange_2:
.global _(__atomic_compare_exchange_2)
.type _(__atomic_compare_exchange_2), %function
_(__atomic_compare_exchange_2):
#ifdef __TINYC__
.short 0x1141
.short 0x86ba
@ -530,11 +537,11 @@ __atomic_compare_exchange_2:
addi sp,sp,16
jr ra
#endif
.size __atomic_compare_exchange_2, .-__atomic_compare_exchange_2
.size _(__atomic_compare_exchange_2), .-_(__atomic_compare_exchange_2)
.global __atomic_compare_exchange_4
.type __atomic_compare_exchange_4, %function
__atomic_compare_exchange_4:
.global _(__atomic_compare_exchange_4)
.type _(__atomic_compare_exchange_4), %function
_(__atomic_compare_exchange_4):
#ifdef __TINYC__
.short 0x419c
.int 0x0f50000f
@ -565,11 +572,11 @@ __atomic_compare_exchange_4:
andi a0,a0,1
ret
#endif
.size __atomic_compare_exchange_4, .-__atomic_compare_exchange_4
.size _(__atomic_compare_exchange_4), .-_(__atomic_compare_exchange_4)
.global __atomic_compare_exchange_8
.type __atomic_compare_exchange_8, %function
__atomic_compare_exchange_8:
.global _(__atomic_compare_exchange_8)
.type _(__atomic_compare_exchange_8), %function
_(__atomic_compare_exchange_8):
#ifdef __TINYC__
.short 0x619c
.int 0x0f50000f
@ -600,7 +607,7 @@ __atomic_compare_exchange_8:
andi a0,a0,1
ret
#endif
.size __atomic_compare_exchange_8, .-__atomic_compare_exchange_8
.size _(__atomic_compare_exchange_8), .-_(__atomic_compare_exchange_8)
/* ---------------------------------------------- */
#endif

View File

@ -1,11 +1,20 @@
/* ---------------------------------------------- */
#ifdef __leading_underscore
# define _(s) _##s
#else
# define _(s) s
#endif
.globl _(__bound_alloca)
_(__bound_alloca):
#if defined __arm__
.text
.align 2
.global fetch_and_add_arm
.type fetch_and_add_arm, %function
fetch_and_add_arm:
.global _(fetch_and_add_arm)
.type _(fetch_and_add_arm), %function
_(fetch_and_add_arm):
mcr p15, #0, r0, c7, c10, #5
.L0:
ldrex r3, [r0]
@ -15,16 +24,16 @@ fetch_and_add_arm:
bne .L0
mcr p15, #0, r0, c7, c10, #5
bx lr
.size fetch_and_add_arm, .-fetch_and_add_arm
.size _(fetch_and_add_arm), .-_(fetch_and_add_arm)
/* ---------------------------------------------- */
#elif defined __aarch64__
.text
.align 2
.global fetch_and_add_arm64
.type fetch_and_add_arm64, %function
fetch_and_add_arm64:
.global _(fetch_and_add_arm64)
.type _(fetch_and_add_arm64), %function
_(fetch_and_add_arm64):
#ifdef __TINYC__
.int 0x885f7c02
.int 0x0b010042
@ -36,21 +45,21 @@ fetch_and_add_arm64:
ldxr w2, [x0]
add w2, w2, w1
stlxr w3, w2, [x0]
cbnz w3, fetch_and_add_arm64
cbnz w3, _(fetch_and_add_arm64)
dmb ish
ret
#endif
.size fetch_and_add_arm64, .-fetch_and_add_arm64
.size _(fetch_and_add_arm64), .-_(fetch_and_add_arm64)
/* ---------------------------------------------- */
#elif defined __riscv
.text
.align 2
.global fetch_and_add_riscv64
.type fetch_and_add_riscv64, %function
fetch_and_add_riscv64:
.global _(fetch_and_add_riscv64)
.type _(fetch_and_add_riscv64), %function
_(fetch_and_add_riscv64):
#ifdef __TINYC__
.int 0x0f50000f
.int 0x004b5202f
@ -61,7 +70,7 @@ fetch_and_add_riscv64:
ret
#endif
.size fetch_and_add_riscv64, .-fetch_and_add_riscv64
.size _(fetch_and_add_riscv64), .-_(fetch_and_add_riscv64)
/* ---------------------------------------------- */
#endif

View File

@ -24,7 +24,7 @@ void *memcpy(void*,void*,__SIZE_TYPE__);
#include <string.h>
#endif
#ifndef __riscv
#if !defined __riscv && !defined __APPLE__
void __clear_cache(void *beg, void *end)
{
__arm64_clear_cache(beg, end);

View File

@ -1487,7 +1487,7 @@ static void add_init_array_defines(TCCState *s1, const char *section_name)
addr_t end_offset;
char buf[1024];
s = find_section_create(s1, section_name, 0);
if (!s) {
if (!s || !(s->sh_flags & SHF_ALLOC)) {
end_offset = 0;
s = data_section;
} else {

View File

@ -31,9 +31,28 @@
to setup our own crt code. We're not using lazy linking, so even function
calls are resolved at startup. */
#if !defined TCC_TARGET_X86_64 && !defined TCC_TARGET_ARM64
#error Platform not supported
#endif
#define DEBUG_MACHO 0
#define dprintf if (DEBUG_MACHO) printf
#define MH_EXECUTE (0x2)
#define MH_DYLDLINK (0x4)
#define MH_PIE (0x200000)
#define CPU_SUBTYPE_LIB64 (0x80000000)
#define CPU_SUBTYPE_X86_ALL (3)
#define CPU_SUBTYPE_ARM64_ALL (0)
#define CPU_ARCH_ABI64 (0x01000000)
#define CPU_TYPE_X86 (7)
#define CPU_TYPE_X86_64 (CPU_TYPE_X86 | CPU_ARCH_ABI64)
#define CPU_TYPE_ARM (12)
#define CPU_TYPE_ARM64 (CPU_TYPE_ARM | CPU_ARCH_ABI64)
struct fat_header {
uint32_t magic; /* FAT_MAGIC or FAT_MAGIC_64 */
uint32_t nfat_arch; /* number of structs that follow */
@ -86,6 +105,7 @@ struct load_command {
#define LC_LOAD_DYLINKER 0xe
#define LC_SEGMENT_64 0x19
#define LC_REEXPORT_DYLIB (0x1f | LC_REQ_DYLD)
#define LC_DYLD_INFO_ONLY (0x22|LC_REQ_DYLD)
#define LC_MAIN (0x28|LC_REQ_DYLD)
typedef int vm_prot_t;
@ -122,6 +142,8 @@ struct section_64 { /* for 64-bit architectures */
#define S_REGULAR 0x0
#define S_ZEROFILL 0x1
#define S_NON_LAZY_SYMBOL_POINTERS 0x6
#define S_LAZY_SYMBOL_POINTERS 0x7
#define S_SYMBOL_STUBS 0x8
#define S_MOD_INIT_FUNC_POINTERS 0x9
#define S_MOD_TERM_FUNC_POINTERS 0xa
@ -187,6 +209,38 @@ struct dysymtab_command {
uint32_t nlocrel; /* number of local relocation entries */
};
#define BIND_OPCODE_DONE 0x00
#define BIND_OPCODE_SET_DYLIB_SPECIAL_IMM 0x30
#define BIND_OPCODE_SET_SYMBOL_TRAILING_FLAGS_IMM 0x40
#define BIND_OPCODE_SET_TYPE_IMM 0x50
#define BIND_OPCODE_SET_SEGMENT_AND_OFFSET_ULEB 0x70
#define BIND_OPCODE_DO_BIND 0x90
#define BIND_TYPE_POINTER 1
#define BIND_SPECIAL_DYLIB_FLAT_LOOKUP -2
#define REBASE_OPCODE_DONE 0x00
#define REBASE_OPCODE_SET_TYPE_IMM 0x10
#define REBASE_OPCODE_SET_SEGMENT_AND_OFFSET_ULEB 0x20
#define REBASE_OPCODE_DO_REBASE_IMM_TIMES 0x50
#define REBASE_TYPE_POINTER 1
struct dyld_info_command {
uint32_t cmd; /* LC_DYLD_INFO or LC_DYLD_INFO_ONLY */
uint32_t cmdsize; /* sizeof(struct dyld_info_command) */
uint32_t rebase_off; /* file offset to rebase info */
uint32_t rebase_size; /* size of rebase info */
uint32_t bind_off; /* file offset to binding info */
uint32_t bind_size; /* size of binding info */
uint32_t weak_bind_off; /* file offset to weak binding info */
uint32_t weak_bind_size; /* size of weak binding info */
uint32_t lazy_bind_off; /* file offset to lazy binding info */
uint32_t lazy_bind_size; /* size of lazy binding infs */
uint32_t export_off; /* file offset to lazy binding info */
uint32_t export_size; /* size of lazy binding infs */
};
#define INDIRECT_SYMBOL_LOCAL 0x80000000
struct entry_point_command {
@ -201,6 +255,7 @@ enum skind {
sk_discard,
sk_text,
sk_stubs,
sk_stub_helper,
sk_ro_data,
sk_uw_info,
sk_nl_ptr, // non-lazy pointers, aka GOT
@ -208,6 +263,14 @@ enum skind {
sk_init,
sk_fini,
sk_rw_data,
sk_stab,
sk_stab_str,
sk_debug_info,
sk_debug_abbrev,
sk_debug_line,
sk_debug_aranges,
sk_debug_str,
sk_debug_line_str,
sk_bss,
sk_linkedit,
sk_last
@ -241,9 +304,26 @@ struct macho {
} sk_to_sect[sk_last];
int *elfsectomacho;
int *e2msym;
Section *symtab, *strtab, *wdata, *indirsyms, *stubs;
int stubsym;
Section *rebase, *binding, *lazy_binding, *exports;
Section *symtab, *strtab, *wdata, *indirsyms;
Section *stubs, *stub_helper, *la_symbol_ptr;
int nr_plt, n_got;
struct dyld_info_command *dyldinfo;
int stubsym, helpsym, lasym, dyld_private, dyld_stub_binder;
uint32_t ilocal, iextdef, iundef;
int n_lazy_bind_rebase;
struct lazy_bind_rebase {
int section;
int bind;
int bind_offset;
int la_symbol_offset;
ElfW_Rel rel;
} *lazy_bind_rebase;
int n_bind;
struct bind {
int section;
ElfW_Rel rel;
} *bind;
};
#define SHT_LINKEDIT (SHT_LOOS + 42)
@ -306,20 +386,165 @@ static void * add_dylib(struct macho *mo, char *name)
return lc;
}
static void write_uleb128(Section *section, uint64_t value)
{
do {
unsigned char byte = value & 0x7f;
uint8_t *ptr = section_ptr_add(section, 1);
value >>= 7;
*ptr = byte | (value ? 0x80 : 0);
} while (value != 0);
}
static void tcc_macho_add_destructor(TCCState *s1)
{
int init_sym, mh_execute_header, at_exit_sym;
Section *s;
ElfW_Rel *rel;
uint8_t *ptr;
s = find_section(s1, ".fini_array");
if (s->data_offset == 0)
return;
init_sym = put_elf_sym(s1->symtab, text_section->data_offset, 0,
ELFW(ST_INFO)(STB_LOCAL, STT_FUNC), 0,
text_section->sh_num, "___GLOBAL_init_65535");
mh_execute_header = put_elf_sym(s1->symtab, 0x100000000ll, 0,
ELFW(ST_INFO)(STB_LOCAL, STT_OBJECT), 0,
SHN_ABS, "__mh_execute_header");
at_exit_sym = put_elf_sym(s1->symtab, 0, 0,
ELFW(ST_INFO)(STB_GLOBAL, STT_FUNC), 0,
SHN_UNDEF, "___cxa_atexit");
#ifdef TCC_TARGET_X86_64
ptr = section_ptr_add(text_section, 4);
ptr[0] = 0x55; // pushq %rbp
ptr[1] = 0x48; // movq %rsp, %rbp
ptr[2] = 0x89;
ptr[3] = 0xe5;
for_each_elem(s->reloc, 0, rel, ElfW_Rel) {
int sym_index = ELFW(R_SYM)(rel->r_info);
ptr = section_ptr_add(text_section, 26);
ptr[0] = 0x48; // lea destructor(%rip),%rax
ptr[1] = 0x8d;
ptr[2] = 0x05;
put_elf_reloca(s1->symtab, text_section,
text_section->data_offset - 23,
R_X86_64_PC32, sym_index, -4);
ptr[7] = 0x48; // mov %rax,%rdi
ptr[8] = 0x89;
ptr[9] = 0xc7;
ptr[10] = 0x31; // xorl %ecx, %ecx
ptr[11] = 0xc9;
ptr[12] = 0x89; // movl %ecx, %esi
ptr[13] = 0xce;
ptr[14] = 0x48; // lea mh_execute_header(%rip),%rdx
ptr[15] = 0x8d;
ptr[16] = 0x15;
put_elf_reloca(s1->symtab, text_section,
text_section->data_offset - 9,
R_X86_64_PC32, mh_execute_header, -4);
ptr[21] = 0xe8; // call __cxa_atexit
put_elf_reloca(s1->symtab, text_section,
text_section->data_offset - 4,
R_X86_64_PLT32, at_exit_sym, -4);
}
ptr = section_ptr_add(text_section, 2);
ptr[0] = 0x5d; // pop %rbp
ptr[1] = 0xc3; // ret
#elif defined TCC_TARGET_ARM64
ptr = section_ptr_add(text_section, 8);
write32le(ptr, 0xa9bf7bfd); // stp x29, x30, [sp, #-16]!
write32le(ptr + 4, 0x910003fd); // mov x29, sp
for_each_elem(s->reloc, 0, rel, ElfW_Rel) {
int sym_index = ELFW(R_SYM)(rel->r_info);
ptr = section_ptr_add(text_section, 24);
put_elf_reloc(s1->symtab, text_section,
text_section->data_offset - 24,
R_AARCH64_ADR_PREL_PG_HI21, sym_index);
write32le(ptr, 0x90000000); // adrp x0, destructor@page
put_elf_reloc(s1->symtab, text_section,
text_section->data_offset - 20,
R_AARCH64_LDST8_ABS_LO12_NC, sym_index);
write32le(ptr + 4, 0x91000000); // add x0,x0,destructor@pageoff
write32le(ptr + 8, 0xd2800001); // mov x1, #0
put_elf_reloc(s1->symtab, text_section,
text_section->data_offset - 12,
R_AARCH64_ADR_PREL_PG_HI21, mh_execute_header);
write32le(ptr + 12, 0x90000002); // adrp x2, mh_execute_header@page
put_elf_reloc(s1->symtab, text_section,
text_section->data_offset - 8,
R_AARCH64_LDST8_ABS_LO12_NC, mh_execute_header);
write32le(ptr + 16, 0x91000042); // add x2,x2,mh_execute_header@pageoff
put_elf_reloc(s1->symtab, text_section,
text_section->data_offset - 4,
R_AARCH64_CALL26, at_exit_sym);
write32le(ptr + 20, 0x94000000); // bl __cxa_atexit
}
ptr = section_ptr_add(text_section, 8);
write32le(ptr, 0xa8c17bfd); // ldp x29, x30, [sp], #16
write32le(ptr + 4, 0xd65f03c0); // ret
#endif
s->reloc->data_offset = s->data_offset = 0;
s->sh_flags &= ~SHF_ALLOC;
add_array (s1, ".init_array", init_sym);
}
static void check_relocs(TCCState *s1, struct macho *mo)
{
uint8_t *jmp;
Section *s;
ElfW_Rel *rel;
ElfW(Sym) *sym;
int i, type, gotplt_entry, sym_index, for_code;
int sh_num, debug, bind_offset, la_symbol_offset;
uint32_t *pi, *goti;
struct sym_attr *attr;
s1->got = new_section(s1, ".got", SHT_PROGBITS, SHF_ALLOC | SHF_WRITE);
mo->indirsyms = new_section(s1, "LEINDIR", SHT_LINKEDIT, SHF_ALLOC | SHF_WRITE);
#ifdef TCC_TARGET_X86_64
jmp = section_ptr_add(mo->stub_helper, 16);
jmp[0] = 0x4c; /* leaq _dyld_private(%rip), %r11 */
jmp[1] = 0x8d;
jmp[2] = 0x1d;
put_elf_reloca(s1->symtab, mo->stub_helper, 3,
R_X86_64_PC32, mo->dyld_private, -4);
jmp[7] = 0x41; /* pushq %r11 */
jmp[8] = 0x53;
jmp[9] = 0xff; /* jmpq *dyld_stub_binder@GOT(%rip) */
jmp[10] = 0x25;
put_elf_reloca(s1->symtab, mo->stub_helper, 11,
R_X86_64_GOTPCREL, mo->dyld_stub_binder, -4);
jmp[15] = 0x90; /* nop */
#elif defined TCC_TARGET_ARM64
jmp = section_ptr_add(mo->stub_helper, 24);
put_elf_reloc(s1->symtab, mo->stub_helper, 0,
R_AARCH64_ADR_PREL_PG_HI21, mo->dyld_private);
write32le(jmp, 0x90000011); // adrp x17, _dyld_private@page
put_elf_reloc(s1->symtab, mo->stub_helper, 4,
R_AARCH64_LDST64_ABS_LO12_NC, mo->dyld_private);
write32le(jmp + 4, 0x91000231); // add x17,x17,_dyld_private@pageoff
write32le(jmp + 8, 0xa9bf47f0); // stp x16/x17, [sp, #-16]!
put_elf_reloc(s1->symtab, mo->stub_helper, 12,
R_AARCH64_ADR_GOT_PAGE, mo->dyld_stub_binder);
write32le(jmp + 12, 0x90000010); // adrp x16, dyld_stub_binder@page
put_elf_reloc(s1->symtab, mo->stub_helper, 16,
R_AARCH64_LD64_GOT_LO12_NC, mo->dyld_stub_binder);
write32le(jmp + 16, 0xf9400210); // ldr x16,[x16,dyld_stub_binder@pageoff]
write32le(jmp + 20, 0xd61f0200); // br x16
#endif
goti = NULL;
mo->nr_plt = mo->n_got = 0;
for (i = 1; i < s1->nb_sections; i++) {
s = s1->sections[i];
if (s->sh_type != SHT_RELX)
continue;
sh_num = s1->sections[s->sh_info]->sh_num;
debug = sh_num >= s1->dwlo && sh_num < s1->dwhi;
for_each_elem(s, 0, rel, ElfW_Rel) {
type = ELFW(R_TYPE)(rel->r_info);
gotplt_entry = gotplt_entry_type(type);
@ -329,44 +554,150 @@ static void check_relocs(TCCState *s1, struct macho *mo)
address due to codegen (i.e. a reloc requiring a got slot). */
sym_index = ELFW(R_SYM)(rel->r_info);
sym = &((ElfW(Sym) *)symtab_section->data)[sym_index];
if (sym->st_shndx == SHN_UNDEF
|| gotplt_entry == ALWAYS_GOTPLT_ENTRY) {
if (!debug &&
(sym->st_shndx == SHN_UNDEF
|| gotplt_entry == ALWAYS_GOTPLT_ENTRY)) {
attr = get_sym_attr(s1, sym_index, 1);
if (!attr->dyn_index) {
uint32_t *pi = section_ptr_add(mo->indirsyms, sizeof(*pi));
attr->got_offset = s1->got->data_offset;
attr->plt_offset = -1;
attr->dyn_index = 1; /* used as flag */
section_ptr_add(s1->got, PTR_SIZE);
section_ptr_add(s1->got, PTR_SIZE);
put_elf_reloc(s1->symtab, s1->got, attr->got_offset,
R_DATA_PTR, sym_index);
goti = tcc_realloc(goti, (mo->n_got + 1) * sizeof(*goti));
if (ELFW(ST_BIND)(sym->st_info) == STB_LOCAL) {
if (sym->st_shndx == SHN_UNDEF)
tcc_error("undefined local symbol???");
*pi = INDIRECT_SYMBOL_LOCAL;
/* The pointer slot we generated must point to the
symbol, whose address is only known after layout,
so register a simple relocation for that. */
put_elf_reloc(s1->symtab, s1->got, attr->got_offset,
R_DATA_PTR, sym_index);
} else
*pi = mo->e2msym[sym_index];
tcc_error("undefined local symbo: '%s'",
(char *) symtab_section->link->data + sym->st_name);
goti[mo->n_got++] = INDIRECT_SYMBOL_LOCAL;
} else {
goti[mo->n_got++] = mo->e2msym[sym_index];
if (sym->st_shndx == SHN_UNDEF
#ifdef TCC_TARGET_X86_64
&& type == R_X86_64_GOTPCREL
#elif defined TCC_TARGET_ARM64
&& type == R_AARCH64_ADR_GOT_PAGE
#endif
) {
mo->bind =
tcc_realloc(mo->bind,
(mo->n_bind + 1) *
sizeof(struct bind));
mo->bind[mo->n_bind].section = s1->got->reloc->sh_info;
mo->bind[mo->n_bind].rel = *rel;
mo->bind[mo->n_bind].rel.r_offset = attr->got_offset;
mo->n_bind++;
s1->got->reloc->data_offset -= sizeof (ElfW_Rel);
}
}
}
if (for_code) {
if (for_code && sym->st_shndx == SHN_UNDEF) {
if (attr->plt_offset == -1) {
uint8_t *jmp;
pi = section_ptr_add(mo->indirsyms, sizeof(*pi));
*pi = mo->e2msym[sym_index];
mo->nr_plt++;
attr->plt_offset = mo->stubs->data_offset;
#ifdef TCC_TARGET_X86_64
if (type != R_X86_64_PLT32)
continue;
/* __stubs */
jmp = section_ptr_add(mo->stubs, 6);
jmp[0] = 0xff; /* jmpq *ofs(%rip) */
jmp[0] = 0xff; /* jmpq *__la_symbol_ptr(%rip) */
jmp[1] = 0x25;
put_elf_reloc(s1->symtab, mo->stubs,
attr->plt_offset + 2,
R_X86_64_GOTPCREL, sym_index);
put_elf_reloca(s1->symtab, mo->stubs,
mo->stubs->data_offset - 4,
R_X86_64_PC32, mo->lasym,
mo->la_symbol_ptr->data_offset - 4);
/* __stub_helper */
bind_offset = mo->stub_helper->data_offset + 1;
jmp = section_ptr_add(mo->stub_helper, 10);
jmp[0] = 0x68; /* pushq $bind_offset */
jmp[5] = 0xe9; /* jmpq __stub_helper */
write32le(jmp + 6, -mo->stub_helper->data_offset);
/* __la_symbol_ptr */
la_symbol_offset = mo->la_symbol_ptr->data_offset;
put_elf_reloca(s1->symtab, mo->la_symbol_ptr,
mo->la_symbol_ptr->data_offset,
R_DATA_PTR, mo->helpsym,
mo->stub_helper->data_offset - 10);
jmp = section_ptr_add(mo->la_symbol_ptr, PTR_SIZE);
#elif defined TCC_TARGET_ARM64
if (type != R_AARCH64_CALL26)
continue;
/* __stubs */
jmp = section_ptr_add(mo->stubs, 12);
put_elf_reloca(s1->symtab, mo->stubs,
mo->stubs->data_offset - 12,
R_AARCH64_ADR_PREL_PG_HI21, mo->lasym,
mo->la_symbol_ptr->data_offset);
write32le(jmp, // adrp x16, __la_symbol_ptr@page
0x90000010);
put_elf_reloca(s1->symtab, mo->stubs,
mo->stubs->data_offset - 8,
R_AARCH64_LDST64_ABS_LO12_NC, mo->lasym,
mo->la_symbol_ptr->data_offset);
write32le(jmp + 4, // ldr x16,[x16, __la_symbol_ptr@pageoff]
0xf9400210);
write32le(jmp + 8, // br x16
0xd61f0200);
/* __stub_helper */
bind_offset = mo->stub_helper->data_offset + 8;
jmp = section_ptr_add(mo->stub_helper, 12);
write32le(jmp + 0, // ldr w16, l0
0x18000050);
write32le(jmp + 4, // b stubHelperHeader
0x14000000 +
((-(mo->stub_helper->data_offset - 8) / 4) &
0x3ffffff));
write32le(jmp + 8, 0); // l0: .long bind_offset
/* __la_symbol_ptr */
la_symbol_offset = mo->la_symbol_ptr->data_offset;
put_elf_reloca(s1->symtab, mo->la_symbol_ptr,
mo->la_symbol_ptr->data_offset,
R_DATA_PTR, mo->helpsym,
mo->stub_helper->data_offset - 12);
jmp = section_ptr_add(mo->la_symbol_ptr, PTR_SIZE);
#endif
mo->lazy_bind_rebase =
tcc_realloc(mo->lazy_bind_rebase,
(mo->n_lazy_bind_rebase + 1) *
sizeof(struct lazy_bind_rebase));
mo->lazy_bind_rebase[mo->n_lazy_bind_rebase].section =
mo->stub_helper->reloc->sh_info;
mo->lazy_bind_rebase[mo->n_lazy_bind_rebase].bind = 1;
mo->lazy_bind_rebase[mo->n_lazy_bind_rebase].bind_offset = bind_offset;
mo->lazy_bind_rebase[mo->n_lazy_bind_rebase].la_symbol_offset = la_symbol_offset;
mo->lazy_bind_rebase[mo->n_lazy_bind_rebase].rel = *rel;
mo->lazy_bind_rebase[mo->n_lazy_bind_rebase].rel.r_offset =
attr->plt_offset;
mo->n_lazy_bind_rebase++;
}
rel->r_info = ELFW(R_INFO)(mo->stubsym, type);
rel->r_addend += attr->plt_offset;
}
}
if (type == R_DATA_PTR) {
mo->lazy_bind_rebase =
tcc_realloc(mo->lazy_bind_rebase,
(mo->n_lazy_bind_rebase + 1) *
sizeof(struct lazy_bind_rebase));
mo->lazy_bind_rebase[mo->n_lazy_bind_rebase].section = s->sh_info;
mo->lazy_bind_rebase[mo->n_lazy_bind_rebase].bind = 0;
mo->lazy_bind_rebase[mo->n_lazy_bind_rebase].rel = *rel;
mo->n_lazy_bind_rebase++;
}
}
}
pi = section_ptr_add(mo->indirsyms, mo->n_got * sizeof(*pi));
memcpy(pi, goti, mo->n_got * sizeof(*pi));
pi = section_ptr_add(mo->indirsyms, mo->nr_plt * sizeof(*pi));
memcpy(pi, mo->indirsyms->data, mo->nr_plt * sizeof(*pi));
tcc_free(goti);
}
static int check_symbols(TCCState *s1, struct macho *mo)
@ -444,9 +775,12 @@ static void convert_symbol(TCCState *s1, struct macho *mo, struct nlist_64 *pn)
n.n_type = N_ABS, n.n_sect = 0;
else if (sym->st_shndx >= SHN_LORESERVE)
tcc_error("unhandled ELF symbol section %d %s", sym->st_shndx, name);
else if (!mo->elfsectomacho[sym->st_shndx])
tcc_error("ELF section %d not mapped into Mach-O for symbol %s",
sym->st_shndx, name);
else if (!mo->elfsectomacho[sym->st_shndx]) {
int sh_num = s1->sections[sym->st_shndx]->sh_num;
if (sh_num < s1->dwlo || sh_num >= s1->dwhi)
tcc_error("ELF section %d(%s) not mapped into Mach-O for symbol %s",
sym->st_shndx, s1->sections[sym->st_shndx]->name, name);
}
else
n.n_sect = mo->elfsectomacho[sym->st_shndx];
if (ELFW(ST_BIND)(sym->st_info) == STB_GLOBAL)
@ -465,9 +799,9 @@ static void convert_symbols(TCCState *s1, struct macho *mo)
convert_symbol(s1, mo, pn);
}
static int machosymcmp(const void *_a, const void *_b)
static int machosymcmp(const void *_a, const void *_b, void *arg)
{
TCCState *s1 = tcc_state;
TCCState *s1 = arg;
int ea = ((struct nlist_64 *)_a)->n_value;
int eb = ((struct nlist_64 *)_b)->n_value;
ElfSym *sa = (ElfSym *)symtab_section->data + ea;
@ -492,6 +826,37 @@ static int machosymcmp(const void *_a, const void *_b)
return ea - eb;
}
/* cannot use qsort because code has to be reentrant */
static void tcc_qsort (void *base, size_t nel, size_t width,
int (*comp)(const void *, const void *, void *), void *arg)
{
size_t wnel, gap, wgap, i, j, k;
char *a, *b, tmp;
wnel = width * nel;
for (gap = 0; ++gap < nel;)
gap *= 3;
while ( gap /= 3 ) {
wgap = width * gap;
for (i = wgap; i < wnel; i += width) {
for (j = i - wgap; ;j -= wgap) {
a = j + (char *)base;
b = a + wgap;
if ( (*comp)(a, b, arg) <= 0 )
break;
k = width;
do {
tmp = *a;
*a++ = *b;
*b++ = tmp;
} while ( --k );
if (j < wgap)
break;
}
}
}
}
static void create_symtab(TCCState *s1, struct macho *mo)
{
int sym_index, sym_end;
@ -500,9 +865,30 @@ static void create_symtab(TCCState *s1, struct macho *mo)
/* Stub creation belongs to check_relocs, but we need to create
the symbol now, so its included in the sorting. */
mo->stubs = new_section(s1, "__stubs", SHT_PROGBITS, SHF_ALLOC | SHF_EXECINSTR);
mo->stub_helper = new_section(s1, "__stub_helper", SHT_PROGBITS, SHF_ALLOC | SHF_EXECINSTR);
s1->got = new_section(s1, ".got", SHT_PROGBITS, SHF_ALLOC | SHF_WRITE);
mo->la_symbol_ptr = new_section(s1, "__la_symbol_ptr", SHT_PROGBITS, SHF_ALLOC | SHF_WRITE);
mo->stubsym = put_elf_sym(s1->symtab, 0, 0,
ELFW(ST_INFO)(STB_LOCAL, STT_SECTION), 0,
mo->stubs->sh_num, ".__stubs");
mo->helpsym = put_elf_sym(s1->symtab, 0, 0,
ELFW(ST_INFO)(STB_LOCAL, STT_SECTION), 0,
mo->stub_helper->sh_num, ".__stub_helper");
mo->lasym = put_elf_sym(s1->symtab, 0, 0,
ELFW(ST_INFO)(STB_LOCAL, STT_SECTION), 0,
mo->la_symbol_ptr->sh_num, ".__la_symbol_ptr");
section_ptr_add(data_section, -data_section->data_offset & (PTR_SIZE - 1));
mo->dyld_private = put_elf_sym(s1->symtab, data_section->data_offset, PTR_SIZE,
ELFW(ST_INFO)(STB_LOCAL, STT_OBJECT), 0,
data_section->sh_num, ".__dyld_private");
section_ptr_add(data_section, PTR_SIZE);
mo->dyld_stub_binder = put_elf_sym(s1->symtab, 0, 0,
ELFW(ST_INFO)(STB_GLOBAL, STT_OBJECT), 0,
SHN_UNDEF, "dyld_stub_binder");
mo->rebase = new_section(s1, "REBASE", SHT_LINKEDIT, SHF_ALLOC | SHF_WRITE);
mo->binding = new_section(s1, "BINDING", SHT_LINKEDIT, SHF_ALLOC | SHF_WRITE);
mo->lazy_binding = new_section(s1, "LAZY_BINDING", SHT_LINKEDIT, SHF_ALLOC | SHF_WRITE);
mo->exports = new_section(s1, "EXPORT", SHT_LINKEDIT, SHF_ALLOC | SHF_WRITE);
mo->symtab = new_section(s1, "LESYMTAB", SHT_LINKEDIT, SHF_ALLOC | SHF_WRITE);
mo->strtab = new_section(s1, "LESTRTAB", SHT_LINKEDIT, SHF_ALLOC | SHF_WRITE);
@ -515,9 +901,7 @@ static void create_symtab(TCCState *s1, struct macho *mo)
pn[sym_index - 1].n_strx = put_elf_str(mo->strtab, name);
pn[sym_index - 1].n_value = sym_index;
}
tcc_enter_state(s1); /* qsort needs global state */
qsort(pn, sym_end - 1, sizeof(*pn), machosymcmp);
tcc_exit_state(s1);
tcc_qsort(pn, sym_end - 1, sizeof(*pn), machosymcmp, s1);
mo->e2msym = tcc_malloc(sym_end * sizeof(*mo->e2msym));
mo->e2msym[0] = -1;
for (sym_index = 1; sym_index < sym_end; ++sym_index) {
@ -530,22 +914,121 @@ const struct {
uint32_t flags;
const char *name;
} skinfo[sk_last] = {
/*[sk_unknown] =*/ { 0 },
/*[sk_discard] =*/ { 0 },
/*[sk_text] =*/ { 1, S_REGULAR | S_ATTR_PURE_INSTRUCTIONS
| S_ATTR_SOME_INSTRUCTIONS, "__text" },
/*[sk_stubs] =*/ { 0 },
/*[sk_ro_data] =*/ { 1, S_REGULAR, "__rodata" },
/*[sk_uw_info] =*/ { 0 },
/*[sk_nl_ptr] =*/ { 2, S_NON_LAZY_SYMBOL_POINTERS, "__got" },
/*[sk_la_ptr] =*/ { 0 },
/*[sk_init] =*/ { 2, S_MOD_INIT_FUNC_POINTERS, "__mod_init_func" },
/*[sk_fini] =*/ { 2, S_MOD_TERM_FUNC_POINTERS, "__mod_term_func" },
/*[sk_rw_data] =*/ { 2, S_REGULAR, "__data" },
/*[sk_bss] =*/ { 2, S_ZEROFILL, "__bss" },
/*[sk_linkedit] =*/ { 3, S_REGULAR, NULL },
/*[sk_unknown] =*/ { 0 },
/*[sk_discard] =*/ { 0 },
/*[sk_text] =*/ { 1, S_REGULAR | S_ATTR_PURE_INSTRUCTIONS
| S_ATTR_SOME_INSTRUCTIONS, "__text" },
/*[sk_stubs] =*/ { 1, S_REGULAR | S_ATTR_PURE_INSTRUCTIONS | S_SYMBOL_STUBS
| S_ATTR_SOME_INSTRUCTIONS , "__stubs" },
/*[sk_stub_helper] =*/ { 1, S_REGULAR | S_ATTR_PURE_INSTRUCTIONS
| S_ATTR_SOME_INSTRUCTIONS , "__stub_helper" },
/*[sk_ro_data] =*/ { 1, S_REGULAR, "__rodata" },
/*[sk_uw_info] =*/ { 0 },
/*[sk_nl_ptr] =*/ { 2, S_NON_LAZY_SYMBOL_POINTERS, "__got" },
/*[sk_la_ptr] =*/ { 2, S_LAZY_SYMBOL_POINTERS, "__la_symbol_ptr" },
/*[sk_init] =*/ { 2, S_MOD_INIT_FUNC_POINTERS, "__mod_init_func" },
/*[sk_fini] =*/ { 2, S_MOD_TERM_FUNC_POINTERS, "__mod_term_func" },
/*[sk_rw_data] =*/ { 2, S_REGULAR, "__data" },
/*[sk_stab] =*/ { 2, S_REGULAR, "__stab" },
/*[sk_stab_str] =*/ { 2, S_REGULAR, "__stab_str" },
/*[sk_debug_info] =*/ { 2, S_REGULAR, "__debug_info" },
/*[sk_debug_abbrev] =*/ { 2, S_REGULAR, "__debug_abbrev" },
/*[sk_debug_line] =*/ { 2, S_REGULAR, "__debug_line" },
/*[sk_debug_aranges] =*/ { 2, S_REGULAR, "__debug_aranges" },
/*[sk_debug_str] =*/ { 2, S_REGULAR, "__debug_str" },
/*[sk_debug_line_str] =*/ { 2, S_REGULAR, "__debug_line_str" },
/*[sk_bss] =*/ { 2, S_ZEROFILL, "__bss" },
/*[sk_linkedit] =*/ { 3, S_REGULAR, NULL },
};
static void set_segment_and_offset(struct macho *mo, addr_t addr,
uint8_t *ptr, int opcode,
Section *sec, addr_t offset)
{
int i;
struct segment_command_64 *seg = NULL;
for (i = 1; i < mo->nseg - 1; i++) {
seg = get_segment(mo, i);
if (addr >= seg->vmaddr && addr < (seg->vmaddr + seg->vmsize))
break;
}
*ptr = opcode | i;
write_uleb128(sec, offset - seg->vmaddr);
}
static void do_bind_rebase(TCCState *s1, struct macho *mo)
{
int i;
uint8_t *ptr;
ElfW(Sym) *sym;
char *name;
for (i = 0; i < mo->n_lazy_bind_rebase; i++) {
int sym_index = ELFW(R_SYM)(mo->lazy_bind_rebase[i].rel.r_info);
Section *s = s1->sections[mo->lazy_bind_rebase[i].section];
sym = &((ElfW(Sym) *)symtab_section->data)[sym_index];
name = (char *) symtab_section->link->data + sym->st_name;
if (mo->lazy_bind_rebase[i].bind) {
write32le(mo->stub_helper->data +
mo->lazy_bind_rebase[i].bind_offset,
mo->lazy_binding->data_offset);
ptr = section_ptr_add(mo->lazy_binding, 1);
set_segment_and_offset(mo, mo->la_symbol_ptr->sh_addr, ptr,
BIND_OPCODE_SET_SEGMENT_AND_OFFSET_ULEB,
mo->lazy_binding,
mo->la_symbol_ptr->sh_addr +
mo->lazy_bind_rebase[i].la_symbol_offset);
ptr = section_ptr_add(mo->lazy_binding, 5 + strlen(name));
*ptr++ = BIND_OPCODE_SET_DYLIB_SPECIAL_IMM |
(BIND_SPECIAL_DYLIB_FLAT_LOOKUP & 0xf);
*ptr++ = BIND_OPCODE_SET_SYMBOL_TRAILING_FLAGS_IMM | 0;
strcpy(ptr, name);
ptr += strlen(name) + 1;
*ptr++ = BIND_OPCODE_DO_BIND;
*ptr = BIND_OPCODE_DONE;
}
else {
ptr = section_ptr_add(mo->rebase, 2);
*ptr++ = REBASE_OPCODE_SET_TYPE_IMM | REBASE_TYPE_POINTER;
set_segment_and_offset(mo, s->sh_addr, ptr,
REBASE_OPCODE_SET_SEGMENT_AND_OFFSET_ULEB,
mo->rebase,
mo->lazy_bind_rebase[i].rel.r_offset +
s->sh_addr);
ptr = section_ptr_add(mo->rebase, 1);
*ptr = REBASE_OPCODE_DO_REBASE_IMM_TIMES | 1;
}
}
for (i = 0; i < mo->n_bind; i++) {
int sym_index = ELFW(R_SYM)(mo->bind[i].rel.r_info);
Section *s = s1->sections[mo->bind[i].section];
sym = &((ElfW(Sym) *)symtab_section->data)[sym_index];
name = (char *) symtab_section->link->data + sym->st_name;
ptr = section_ptr_add(mo->binding, 5 + strlen(name));
*ptr++ = BIND_OPCODE_SET_DYLIB_SPECIAL_IMM |
(BIND_SPECIAL_DYLIB_FLAT_LOOKUP & 0xf);
*ptr++ = BIND_OPCODE_SET_SYMBOL_TRAILING_FLAGS_IMM | 0;
strcpy(ptr, name);
ptr += strlen(name) + 1;
*ptr++ = BIND_OPCODE_SET_TYPE_IMM | BIND_TYPE_POINTER;
set_segment_and_offset(mo, s->sh_addr, ptr,
BIND_OPCODE_SET_SEGMENT_AND_OFFSET_ULEB,
mo->binding,
mo->bind[i].rel.r_offset + s->sh_addr);
ptr = section_ptr_add(mo->binding, 1);
*ptr++ = BIND_OPCODE_DO_BIND;
}
if (mo->rebase->data_offset) {
ptr = section_ptr_add(mo->rebase, 1);
*ptr = REBASE_OPCODE_DONE;
}
tcc_free(mo->lazy_bind_rebase);
tcc_free(mo->bind);
}
static void collect_sections(TCCState *s1, struct macho *mo)
{
int i, sk, numsec;
@ -584,6 +1067,7 @@ static void collect_sections(TCCState *s1, struct macho *mo)
str = (char*)dyldlc + dyldlc->name;
strcpy(str, "/usr/lib/dyld");
mo->dyldinfo = add_lc(mo, LC_DYLD_INFO_ONLY, sizeof(*mo->dyldinfo));
symlc = add_lc(mo, LC_SYMTAB, sizeof(*symlc));
dysymlc = add_lc(mo, LC_DYSYMTAB, sizeof(*dysymlc));
@ -610,12 +1094,37 @@ static void collect_sections(TCCState *s1, struct macho *mo)
case SHT_FINI_ARRAY: sk = sk_fini; break;
case SHT_NOBITS: sk = sk_bss; break;
case SHT_SYMTAB: sk = sk_discard; break;
case SHT_STRTAB: sk = s == stabstr_section ? sk_ro_data : sk_discard; break;
case SHT_STRTAB:
if (s == stabstr_section)
sk = sk_stab_str;
else
sk = sk_discard;
break;
case SHT_RELX: sk = sk_discard; break;
case SHT_LINKEDIT: sk = sk_linkedit; break;
case SHT_PROGBITS:
if (s == s1->got)
if (s == mo->stubs)
sk = sk_stubs;
else if (s == mo->stub_helper)
sk = sk_stub_helper;
else if (s == s1->got)
sk = sk_nl_ptr;
else if (s == mo->la_symbol_ptr)
sk = sk_la_ptr;
else if (s == stab_section)
sk = sk_stab;
else if (s == dwarf_info_section)
sk = sk_debug_info;
else if (s == dwarf_abbrev_section)
sk = sk_debug_abbrev;
else if (s == dwarf_line_section)
sk = sk_debug_line;
else if (s == dwarf_aranges_section)
sk = sk_debug_aranges;
else if (s == dwarf_str_section)
sk = sk_debug_str;
else if (s == dwarf_line_str_section)
sk = sk_debug_line_str;
else if (flags & SHF_EXECINSTR)
sk = sk_text;
else if (flags & SHF_WRITE)
@ -637,9 +1146,14 @@ static void collect_sections(TCCState *s1, struct macho *mo)
mo->elfsectomacho = tcc_mallocz(sizeof(*mo->elfsectomacho) * s1->nb_sections);
for (sk = sk_unknown; sk < sk_last; sk++) {
struct section_64 *sec = NULL;
#define SEG_PAGE_SIZE 16384
if (sk == sk_linkedit)
do_bind_rebase(s1, mo);
if (seg) {
seg->vmsize = curaddr - seg->vmaddr;
seg->filesize = fileofs - seg->fileoff;
seg->vmsize = (curaddr - seg->vmaddr + SEG_PAGE_SIZE - 1) & -SEG_PAGE_SIZE;
seg->filesize = (fileofs - seg->fileoff + SEG_PAGE_SIZE - 1) & -SEG_PAGE_SIZE;
curaddr = seg->vmaddr + seg->vmsize;
fileofs = seg->fileoff + seg->filesize;
}
if (skinfo[sk].seg && mo->sk_to_sect[sk].s) {
uint64_t al = 0;
@ -652,6 +1166,16 @@ static void collect_sections(TCCState *s1, struct macho *mo)
mo->sk_to_sect[sk].machosect = si;
sec = get_section(seg, si);
sec->flags = skinfo[sk].flags;
if (sk == sk_stubs)
#ifdef TCC_TARGET_X86_64
sec->reserved2 = 6;
#elif defined TCC_TARGET_ARM64
sec->reserved2 = 12;
#endif
if (sk == sk_nl_ptr)
sec->reserved1 = mo->nr_plt;
if (sk == sk_la_ptr)
sec->reserved1 = mo->nr_plt + mo->n_got;
}
if (seg->vmaddr == -1) {
curaddr = (curaddr + 4095) & -4096;
@ -737,6 +1261,23 @@ static void collect_sections(TCCState *s1, struct macho *mo)
dysymlc->nundefsym = symlc->nsyms - dysymlc->iundefsym;
dysymlc->indirectsymoff = mo->indirsyms->sh_offset;
dysymlc->nindirectsyms = mo->indirsyms->data_offset / sizeof(uint32_t);
if (mo->rebase->data_offset) {
mo->dyldinfo->rebase_off = mo->rebase->sh_offset;
mo->dyldinfo->rebase_size = mo->rebase->data_offset;
}
if (mo->binding->data_offset) {
mo->dyldinfo->bind_off = mo->binding->sh_offset;
mo->dyldinfo->bind_size = mo->binding->data_offset;
}
if (mo->lazy_binding->data_offset) {
mo->dyldinfo->lazy_bind_off = mo->lazy_binding->sh_offset;
mo->dyldinfo->lazy_bind_size = mo->lazy_binding->data_offset;
}
if (mo->exports->data_offset) {
mo->dyldinfo->export_off = mo->exports->sh_offset;
mo->dyldinfo->export_size = mo->exports->data_offset;
}
}
static void macho_write(TCCState *s1, struct macho *mo, FILE *fp)
@ -745,10 +1286,15 @@ static void macho_write(TCCState *s1, struct macho *mo, FILE *fp)
uint64_t fileofs = 0;
Section *s;
mo->mh.mh.magic = MH_MAGIC_64;
mo->mh.mh.cputype = 0x1000007; // x86_64
mo->mh.mh.cpusubtype = 0x80000003;// all | CPU_SUBTYPE_LIB64
mo->mh.mh.filetype = 2; // MH_EXECUTE
mo->mh.mh.flags = 4; // DYLDLINK
#ifdef TCC_TARGET_X86_64
mo->mh.mh.cputype = CPU_TYPE_X86_64;
mo->mh.mh.cpusubtype = CPU_SUBTYPE_LIB64 | CPU_SUBTYPE_X86_ALL;
#elif defined TCC_TARGET_ARM64
mo->mh.mh.cputype = CPU_TYPE_ARM64;
mo->mh.mh.cpusubtype = CPU_SUBTYPE_ARM64_ALL;
#endif
mo->mh.mh.filetype = MH_EXECUTE;
mo->mh.mh.flags = MH_DYLDLINK | MH_PIE;
mo->mh.mh.ncmds = mo->nlc;
mo->mh.mh.sizeofcmds = 0;
for (i = 0; i < mo->nlc; i++)
@ -803,6 +1349,7 @@ ST_FUNC int macho_output_file(TCCState *s1, const char *filename)
printf("<- %s\n", filename);
tcc_add_runtime(s1);
tcc_macho_add_destructor(s1);
resolve_common_syms(s1);
create_symtab(s1, &mo);
check_relocs(s1, &mo);
@ -959,8 +1506,13 @@ ST_FUNC int macho_load_dll(TCCState * s1, int fd, const char* filename, int lev)
fh.nfat_arch * sizeof(*fa));
swap = fh.magic == FAT_CIGAM;
for (i = 0; i < SWAP(fh.nfat_arch); i++)
if (SWAP(fa[i].cputype) == 0x01000007 /* CPU_TYPE_X86_64 */
&& SWAP(fa[i].cpusubtype) == 3) /* CPU_SUBTYPE_X86_ALL */
#ifdef TCC_TARGET_X86_64
if (SWAP(fa[i].cputype) == CPU_TYPE_X86_64
&& SWAP(fa[i].cpusubtype) == CPU_SUBTYPE_X86_ALL)
#elif defined TCC_TARGET_ARM64
if (SWAP(fa[i].cputype) == CPU_TYPE_ARM64
&& SWAP(fa[i].cpusubtype) == CPU_SUBTYPE_ARM64_ALL)
#endif
break;
if (i == SWAP(fh.nfat_arch)) {
tcc_free(fa);

View File

@ -3648,7 +3648,7 @@ void test_asm_dead_code(void)
void test_asm_call(void)
{
#if defined __x86_64__ && !defined _WIN64
#if defined __x86_64__ && !defined _WIN64 && !defined(__APPLE__)
static char str[] = "PATH";
char *s;
/* This tests if a reference to an undefined symbol from an asm
@ -3660,10 +3660,8 @@ void test_asm_call(void)
tested here). */
/* two pushes so stack remains aligned */
asm volatile ("push %%rdi; push %%rdi; mov %0, %%rdi;"
#if 1 && !defined(__TINYC__) && (defined(__PIC__) || defined(__PIE__)) && !defined(__APPLE__)
#if 1 && !defined(__TINYC__) && (defined(__PIC__) || defined(__PIE__))
"call getenv@plt;"
#elif defined(__APPLE__)
"call _getenv;"
#else
"call getenv;"
#endif
@ -3892,7 +3890,7 @@ void builtin_test(void)
//printf("bera: %p\n", __builtin_extract_return_addr((void*)43));
}
#ifdef _WIN32
#if defined _WIN32 || defined __APPLE__
void weak_test(void) {}
#else
extern int __attribute__((weak)) weak_f1(void);

View File

@ -14,9 +14,11 @@
abort(); \
} while (0)
#ifndef __APPLE__
#if defined __x86_64__ || defined __aarch64__ || defined __riscv
#define HAS_64BITS
#endif
#endif
typedef struct {
atomic_flag flag;