goldberg_emulator/overlay_experimental/Linux_Detour.cpp

638 lines
23 KiB
C++

#include "Linux_Detour.h"
#include <algorithm>
#include <list>
#include <cstdint>
#include <unistd.h>
#include <malloc.h>
#include <sys/mman.h>
#include <errno.h>
//------------------------------------------------------------------------------//
// Helper funcs
//------------------------------------------------------------------------------//
constexpr static auto relative_jump_size = 5;
constexpr static auto relative_addr_jump_size = sizeof(int32_t);
constexpr static auto absolute_jump_size = 6;
struct
{
bool has_r_m;
uint8_t base_size;
} s_opcodes[256] =
{
{false, 0}, {false, 0}, {false, 0}, {false, 0}, {false, 0}, {false, 0}, {false, 0}, {false, 0}, // 0 - 7
{false, 0}, {false, 0}, {false, 0}, {false, 0}, {false, 0}, {false, 0}, {false, 0}, {false, 0}, // 8 - 15
{false, 0}, {false, 0}, {false, 0}, {false, 0}, {false, 0}, {false, 0}, {false, 0}, {false, 0}, // 16 - 23
{false, 0}, {false, 0}, {false, 0}, {false, 0}, {false, 0}, {false, 0}, {false, 0}, {false, 0}, // 24 - 31
{false, 0}, {false, 0}, {false, 0}, {false, 0}, {false, 0}, {false, 0}, {false, 0}, {false, 0}, // 32 - 39
{false, 0}, {false, 0}, {false, 0}, {false, 0}, {false, 0}, {false, 0}, {false, 0}, {false, 0}, // 40 - 47
{false, 0}, {false, 0}, {false, 0}, {false, 0}, {false, 0}, {false, 0}, {false, 0}, {false, 0}, // 48 - 55
{false, 0}, {false, 0}, {false, 0}, {false, 0}, {false, 0}, {false, 0}, {false, 0}, {false, 0}, // 56 - 63
{false, 0}, {false, 0}, {false, 0}, {false, 0}, {false, 0}, {false, 0}, {false, 0}, {false, 0}, // 64 - 71
{false, 0}, {false, 0}, {false, 0}, {false, 0}, {false, 0}, {false, 0}, {false, 0}, {false, 0}, // 72 - 79
// PUSH ...
{false, 1}, {false, 1}, {false, 1}, {false, 1}, {false, 1}, {false, 1}, {false, 1}, {false, 1}, // 80 - 87
// POP ...
{false, 1}, {false, 1}, {false, 1}, {false, 1}, {false, 1}, {false, 1}, {false, 1}, {false, 1}, // 88 - 95
{false, 0}, {false, 0}, {false, 0}, {false, 0}, {false, 0}, {false, 0}, {false, 0}, {false, 0}, // 96 - 103
{false, 0}, {false, 0}, {false, 0}, {false, 0}, {false, 0}, {false, 0}, {false, 0}, {false, 0}, // 104 - 111
{false, 0}, {false, 0}, {false, 0}, {false, 0}, {false, 0}, {false, 0}, {false, 0}, {false, 0}, // 112 - 129
{false, 0}, {false, 0}, {false, 0}, {false, 0}, {false, 0}, {false, 0}, {false, 0}, {false, 0}, // 120 - 127
// MOV, ADD, ... R8 <- IMM8 TEST_8 TEST XCHG_8 XCHG
{true , 3}, {true , 6}, {true , 3}, {true , 3}, {true , 2}, {true , 2}, {true , 2}, {true , 2}, // 128 - 135
// MOV_8 MOV MOV_R8_B MOV_R32_D MOV_32_ES LEA MOV_ES_32 POP
{true , 2}, {true , 2}, {true , 2}, {true , 2}, {true , 2}, {true , 2}, {true , 2}, {false, 2}, // 136 - 143
// NOP
{false, 1}, {false, 0}, {false, 0}, {false, 0}, {false, 0}, {false, 0}, {false, 0}, {false, 0}, // 144 - 151
{false, 0}, {false, 0}, {false, 0}, {false, 0}, {false, 0}, {false, 0}, {false, 0}, {false, 0}, // 152 - 159
{false, 0}, {false, 0}, {false, 0}, {false, 0}, {false, 0}, {false, 0}, {false, 0}, {false, 0}, // 160 - 167
{false, 0}, {false, 0}, {false, 0}, {false, 0}, {false, 0}, {false, 0}, {false, 0}, {false, 0}, // 168 - 175
// MOV_AL MOV_CL MOV_DL MOV_BL MOV_AH MOV_CH MOV_DH MOV_BH
{false, 2}, {false, 2}, {false, 2}, {false, 2}, {false, 2}, {false, 2}, {false, 2}, {false, 2}, // 176 - 183
// MOV_EAX MOV_ECX MOV_EDX MOV_EBX MOV_ESP MOV_EBP MOV_ESI MOV_EDI,
{false, 5}, {false, 5}, {false, 5}, {false, 5}, {false, 5}, {false, 5}, {false, 5}, {false, 5}, // 184 - 191
// RETN_IMM16 RETN
{false, 0}, {false, 0}, {false, 3}, {false, 1}, {false, 0}, {false, 0}, {false, 0}, {false, 0}, // 192 - 199
// LEAVE RETF_IMM16 RETF INT INT_IMM8 INTO
{false, 0}, {false, 1}, {false, 3}, {false, 1}, {false, 1}, {false, 1}, {false, 1}, {false, 0}, // 200 - 207
{false, 0}, {false, 0}, {false, 0}, {false, 0}, {false, 0}, {false, 0}, {false, 0}, {false, 0}, // 208 - 215
{false, 0}, {false, 0}, {false, 0}, {false, 0}, {false, 0}, {false, 0}, {false, 0}, {false, 0}, // 216 - 223
{false, 0}, {false, 0}, {false, 0}, {false, 0}, {false, 0}, {false, 0}, {false, 0}, {false, 0}, // 224 - 231
// CALL JMP LJMP SHORT_JMP
{false, 0}, {false, 0}, {false, 0}, {false, 0}, {false, 0}, {false, 0}, {false, 0}, {false, 0}, // 232 - 239
{false, 0}, {false, 0}, {false, 0}, {false, 0}, {false, 0}, {false, 0}, {false, 0}, {false, 0}, // 240 - 247
// EXTENDED
{false, 0}, {false, 0}, {false, 0}, {false, 0}, {false, 0}, {false, 0}, {false, 0}, {false, 6}, // 248 - 255
};
static constexpr auto mod_mask = 0xC0;
static constexpr auto register_addressing_mode = 0xC0;
static constexpr auto four_bytes_signed_displacement = 0x80;
static constexpr auto one_byte_signed_displacement = 0x40;
static constexpr auto rm_mask = 0x05;
static constexpr auto displacement_only_addressing = 0x05;
static constexpr auto sib_with_no_displacement = 0x04;
static constexpr auto register_indirect_addressing_mode = 0x00;
enum opcodes_e // Commonly used opcode in the beginning of functions
{
PUSH_EAX = 0x50, PUSH_ECX, PUSH_EDX, PUSH_EBX, PUSH_ESP, PUSH_EBP, PUSH_ESI, PUSH_EDI,
POP_EAX , POP_ECX , POP_EDX , POP_EBX , POP_ESP , POP_EBP , POP_ESI , POP_EDI ,
R8_IMM8 = 0x80, R32_IMM32, R8_IMM8_2, R32_IMM8,
TEST_8, TEST,
XCHG_8, XCHG,
MOV_8 , MOV,
MOV_8_B, MOV_32_D, MOV_32_ES,
LEA,
MOV_ES_32, POP,
NOP,
MOV_AL = 0xB0, MOV_CL, MOV_DL, MOV_BL, MOV_AH, MOV_CH, MOV_DH, MOV_BH,
MOV_EAX, MOV_ECX, MOV_EDX, MOV_EBX, MOV_ESP, MOV_EBP, MOV_ESI, MOV_EDI,
RETN_IMM16 = 0xC2, RETN,
LEAVE = 0xC9, RETF_IMM16, RETF, INT, INT_IMM8, INTO,
CALL = 0xE8, // 5 bytes don't process
JMP, // 5 bytes don't process
LJMP, // 7 bytes, don't process
SHORT_JMP, // 2 bytes don't process
EXTENDED = 0xFF,
};
const char* opcode_name(uint8_t opcode)
{
#define NAME(opcode_name) case opcode_name: return #opcode_name
switch( opcode )
{
NAME(PUSH_EAX); NAME(PUSH_ECX); NAME(PUSH_EDX); NAME(PUSH_EBX);
NAME(PUSH_ESP); NAME(PUSH_EBP); NAME(PUSH_ESI); NAME(PUSH_EDI);
NAME(POP_EAX); NAME(POP_ECX); NAME(POP_EDX); NAME(POP_EBX);
NAME(POP_ESP); NAME(POP_EBP); NAME(POP_ESI); NAME(POP_EDI);
NAME(R8_IMM8); NAME(R32_IMM32); NAME(R8_IMM8_2); NAME(R32_IMM8);
NAME(TEST_8); NAME(TEST); NAME(XCHG_8); NAME(XCHG); NAME(MOV_8); NAME(MOV);
NAME(MOV_8_B); NAME(MOV_32_D); NAME(MOV_32_ES);
NAME(LEA);
NAME(MOV_ES_32); NAME(POP);
NAME(NOP);
NAME(MOV_AL) ; NAME(MOV_CL) ; NAME(MOV_DL) ; NAME(MOV_BL) ; NAME(MOV_AH) ; NAME(MOV_CH) ; NAME(MOV_DH) ; NAME(MOV_BH) ;
NAME(MOV_EAX); NAME(MOV_ECX); NAME(MOV_EDX); NAME(MOV_EBX); NAME(MOV_ESP); NAME(MOV_EBP); NAME(MOV_ESI); NAME(MOV_EDI);
NAME(RETN_IMM16); NAME(RETN); NAME(LEAVE); NAME(RETF_IMM16); NAME(RETF);
NAME(INT); NAME(INT_IMM8); NAME(INTO); NAME(CALL); NAME(JMP); NAME(LJMP); NAME(SHORT_JMP);
NAME(EXTENDED);
}
#undef NAME
return "no registered";
}
#pragma pack(push, 1)
struct trampoline_x86_t
{
uint8_t trampolineBytes[16+relative_jump_size]; // trampoline + original function opcodes
uint8_t hookJump[relative_jump_size]; // jump to hook addr, needed because of relative jump overflow
uint8_t nOriginalBytes; // number of original function bytes bkp
uint8_t originalBytes[16]; // original function bytes
};
typedef trampoline_x86_t trampoline_t;
struct trampoline_region_t
{
uint32_t header;
uint8_t numTrampolines; // current trampolines allocated
trampoline_t *trampolines_start; // start pointer of current region trampolines
trampoline_t *next_free_trampoline; // next free trampoline in region
};
#pragma pack(pop)
struct transaction_t
{
bool detach;
void** ppOriginalFunc;
trampoline_t *trampoline;
};
static std::list<trampoline_region_t> trampoline_regions;
static bool transaction_started = false;
static std::list<transaction_t> cur_transaction;
inline size_t page_size()
{
static size_t _page_size = sysconf(_SC_PAGESIZE);
return _page_size;
}
inline size_t region_size()
{
return page_size();
}
static uint8_t max_trampolines_in_region = region_size() / sizeof(trampoline_t);
inline void* library_address_by_handle(void *library)
{
return (library == nullptr ? nullptr : *reinterpret_cast<void**>(library));
}
inline size_t page_align(size_t size, size_t page_size)
{
return (size+(page_size-1)) & (((size_t)-1)^(page_size-1));
}
inline void* page_addr(void* addr, size_t page_size)
{
return reinterpret_cast<void*>(reinterpret_cast<size_t>(addr) & (((size_t)-1)^(page_size-1)));
}
////////////////////////////////////////////////////
/// Tiny disasm
bool is_opcode_terminating_function(uint8_t* pCode)
{
switch( *pCode )
{
case LEAVE:
case RETN: case RETN_IMM16:
case RETF: case RETF_IMM16:
case INT: case INT_IMM8: case INTO:
return true;
}
return false;
}
int find_space_for_trampoline(uint8_t* func, int bytes_needed)
{
if( func == nullptr )
return -1;
int code_len = -1;
bool search = true;
uint8_t *startCode = reinterpret_cast<uint8_t*>(func);
uint8_t *pCode = startCode;
while( search ) // Find opcodes size and try to find at least 5 bytes for our JMP
{
if( is_opcode_terminating_function(pCode) )
break;
if( s_opcodes[*pCode].has_r_m )
{ // MOD-REG-R/M Byte
// 7 6 5 4 3 2 1 0 - bits
//[ MOD ][ REG ][ R/M ]
switch( pCode[1] & mod_mask ) // Check MOD to know how many bytes we have after this opcode
{
case register_addressing_mode : pCode += s_opcodes[*pCode].base_size ; break;// register addressing mode [opcode] [R/M] [XX]
case four_bytes_signed_displacement: pCode += s_opcodes[*pCode].base_size+5; break;// address mode byte + 4 bytes displacement
case one_byte_signed_displacement : pCode += s_opcodes[*pCode].base_size+2; break;// address mode byte + 1 byte displacement
case 0x00:
switch( pCode[1] & rm_mask )
{
case sib_with_no_displacement : pCode += s_opcodes[*pCode].base_size+1; break;// SIB with no displacement
case displacement_only_addressing : pCode += s_opcodes[*pCode].base_size+4; break;// 4 bytes Displacement only addressing mode
case register_indirect_addressing_mode: pCode += s_opcodes[*pCode].base_size; // Register indirect addressing mode
}
}
}
else if( s_opcodes[*pCode].base_size )
{
pCode += s_opcodes[*pCode].base_size;
}
else
{
switch( *pCode )
{
case CALL: case JMP: case LJMP: case SHORT_JMP:
//std::cerr << "CALL and JMP are not supported for trampolines." << std::endl;
search = false;
break;
case EXTENDED:
//std::cerr << "IMPORT_JUMP is not handled" << std::endl;
if (pCode[1] == 0x25) // This is an imported function
{ // Get the true function call
//pCode = (uint8_t*)*(pCode+2);
//startCode = pCode;
// For now disable this case
search = false;
}
else
{
search = false;
}
break;
default:
//std::cerr << "opcode " << std::hex << (uint32_t)*pCode << " no registered" << std::endl;
search = false;
}
}
if( (pCode - startCode) >= bytes_needed && search )
{
search = false;
code_len = pCode-startCode;
}
}
return code_len;
}
///////////////////////////////////////////
// Tiny asm
inline uint8_t* gen_immediate_addr(uint8_t* opcode_addr, uint8_t* dest)
{
*reinterpret_cast<int32_t*>(opcode_addr) = (dest - (opcode_addr + relative_addr_jump_size));
return opcode_addr + relative_addr_jump_size;
}
inline uint8_t* gen_immediate_jump(uint8_t* opcode_addr, uint8_t* dest)
{
*opcode_addr++ = JMP;
return gen_immediate_addr(opcode_addr, dest);
}
inline uint8_t* gen_immediate_call(uint8_t* opcode_addr, uint8_t* dest)
{
*opcode_addr++ = CALL;
return gen_immediate_addr(opcode_addr, dest);
}
uint8_t* relative_addr_to_absolute(int32_t rel_addr, uint8_t *code_addr)
{
return code_addr + rel_addr + relative_jump_size;
}
void alloc_new_trampoline_region()
{
trampoline_region_t region;
region.numTrampolines = 0;
// allocate new trampoline right in the middle of memory so relative jump can access any function
region.trampolines_start = reinterpret_cast<trampoline_t*>(mmap((void*)std::numeric_limits<int32_t>::max(), // allocate the page near the half of memory addressing
region_size(), // size
PROT_EXEC|PROT_WRITE|PROT_READ, // protection
MAP_PRIVATE|MAP_32BIT|MAP_ANONYMOUS, // don't map a file but memory
-1, // fd = -1
0) // offset
);
// Fill the region with 0
std::fill(reinterpret_cast<uint8_t*>(region.trampolines_start), reinterpret_cast<uint8_t*>(region.trampolines_start)+region_size(), 0);
region.next_free_trampoline = region.trampolines_start;
// Protect trampoline region memory
mprotect((void*)region.trampolines_start, region_size(), PROT_READ|PROT_EXEC);
trampoline_regions.push_back(region);
}
trampoline_t* get_free_trampoline()
{
if (!transaction_started)
return nullptr;
trampoline_t *res = nullptr;
auto it = std::find_if(trampoline_regions.begin(), trampoline_regions.end(), [&res](trampoline_region_t &region){
if( region.numTrampolines == max_trampolines_in_region )
return false;
return true;
});
if( it == trampoline_regions.end() )
{
alloc_new_trampoline_region();
it = --trampoline_regions.end();
}
res = it->next_free_trampoline;
trampoline_t *next_new_trampoline = res+1;
if( it->numTrampolines != max_trampolines_in_region )
{
while( next_new_trampoline->nOriginalBytes != 0 )
{
++next_new_trampoline;
}
}
else
{
next_new_trampoline = nullptr;
}
it->next_free_trampoline = next_new_trampoline;
++it->numTrampolines;
return res;
}
void clear_trampoline(trampoline_region_t& region, trampoline_t *trampoline)
{
--region.numTrampolines;
std::fill(reinterpret_cast<uint8_t*>(trampoline), reinterpret_cast<uint8_t*>(trampoline+1), 0);
if( region.next_free_trampoline == nullptr || region.next_free_trampoline > trampoline )
region.next_free_trampoline = trampoline;
}
inline bool is_page_inside_region(void *page, trampoline_region_t& region)
{
if( page >= region.trampolines_start && page <= (region.trampolines_start+region_size()) )
return true;
return false;
}
//------------------------------------------------------------------------------//
/*
#include <signal.h>
#include <dirent.h>
static pid_t tid = 0;
struct dism_pthread_t
{
uint32_t d0;
uint32_t d4;
uint32_t d8;
uint32_t dc;
uint32_t d10;
uint32_t d14;
uint32_t d18;
uint32_t d1c;
uint32_t d20;
uint32_t d24;
uint32_t d28;
uint32_t d2c;
uint32_t d30;
uint32_t d34;
uint32_t d38;
uint32_t d3c;
uint32_t d40;
uint32_t d44;
uint32_t d48;
uint32_t d4c;
uint32_t d50;
uint32_t d54;
uint32_t d58;
uint32_t d5c;
uint32_t d60;
uint32_t d64;
pid_t task_id;
};
*/
int Linux_Detour::update_thread(pthread_t thread_id)
{
//dism_pthread_t *dt = (dism_pthread_t*)thread_id;
// dt->task_id == syscall(SYS_gettid);
return 0;
}
int Linux_Detour::transaction_begin()
{
if( transaction_started )
return -1;
transaction_started = true;
return 0;
}
int Linux_Detour::transaction_abort()
{
if(!transaction_started)
return -1;
for( auto &i : cur_transaction )
{
trampoline_t *trampoline = i.trampoline;
void *page_start = page_addr(reinterpret_cast<void*>(trampoline), page_size());
auto it = std::find_if(trampoline_regions.begin(), trampoline_regions.end(), [page_start](trampoline_region_t &region){
if( is_page_inside_region(page_start, region) )
return true;
return false;
});
if( it != trampoline_regions.end() )
{
clear_trampoline(*it, trampoline);
}
}
cur_transaction.clear();
transaction_started = false;
return 0;
}
int Linux_Detour::transaction_commit()
{
if (!transaction_started )
return -1;
for( auto &i : cur_transaction)
{
trampoline_t *trampoline = i.trampoline;
void **ppOriginalFunc = i.ppOriginalFunc;
int res;
if( i.detach )
{
void* trampoline_page = page_addr(reinterpret_cast<void*>(trampoline), page_size());
*ppOriginalFunc = (void*)(relative_addr_to_absolute(*reinterpret_cast<int32_t*>(trampoline->trampolineBytes+trampoline->nOriginalBytes+1),
trampoline->trampolineBytes));
void* originalFunctionPage = page_addr(*ppOriginalFunc, page_size());
// Allow write on the original func
res = mprotect(originalFunctionPage, page_size()*2, PROT_READ|PROT_WRITE|PROT_EXEC);
// Write the original opcodes
std::copy(trampoline->originalBytes, trampoline->originalBytes+trampoline->nOriginalBytes,
reinterpret_cast<uint8_t*>(*ppOriginalFunc));
// Remove write permission
res = mprotect(originalFunctionPage, page_size()*2, PROT_READ|PROT_EXEC);
// Allow write on trampoline page
res = mprotect(trampoline_page, page_size()*2, PROT_READ|PROT_WRITE|PROT_EXEC);
clear_trampoline(*reinterpret_cast<trampoline_region_t*>(trampoline_page), trampoline);
// Remove write permission
res = mprotect(trampoline_page, page_size()*2, PROT_READ|PROT_EXEC);
}
else
{
void* originalFunctionPage = page_addr(*ppOriginalFunc, page_size());
// Allow write on the original func
res = mprotect(originalFunctionPage, page_size()*2, PROT_READ|PROT_WRITE|PROT_EXEC);
// Write the jump to trampoline
gen_immediate_jump(reinterpret_cast<uint8_t*>(*ppOriginalFunc), trampoline->hookJump);
// Remove write permission
res = mprotect(originalFunctionPage, page_size()*2, PROT_READ|PROT_EXEC);
*ppOriginalFunc = (void*)(trampoline->trampolineBytes);
}
}
cur_transaction.clear();
transaction_started = false;
return 0;
}
int Linux_Detour::unhook_func(void** ppOriginalFunc, void* _hook)
{
if( !transaction_started )
return -EPERM;
if( ppOriginalFunc == nullptr || _hook == nullptr || *ppOriginalFunc == nullptr )
return -EINVAL;
trampoline_t *trampoline = reinterpret_cast<trampoline_t*>(*ppOriginalFunc);
void *page_start = page_addr(reinterpret_cast<void*>(trampoline), page_size());
auto it = std::find_if(trampoline_regions.begin(), trampoline_regions.end(), [page_start](trampoline_region_t &region){
if( is_page_inside_region(page_start, region) )
return true;
return false;
});
if( it != trampoline_regions.end() )
{
cur_transaction.push_back({true, ppOriginalFunc, trampoline});
}
return -EINVAL;
}
int Linux_Detour::hook_func(void** ppOriginalFunc, void* _hook)
{
if( !transaction_started )
return -EPERM;
if( ppOriginalFunc == nullptr || _hook == nullptr || *ppOriginalFunc == nullptr )
return -EINVAL;
uint8_t* hook = reinterpret_cast<uint8_t*>(_hook);
uint8_t* pOriginalFunc = reinterpret_cast<uint8_t*>(*ppOriginalFunc);
int code_len = find_space_for_trampoline(pOriginalFunc, relative_jump_size);
if( code_len < relative_jump_size )
return -ENOSPC;
// Allocate the trampoline, try to put it right in the middle of the mem, so a relative jump can access any function in the app (+/-2GB)
//
// Our hook is a 5 bytes JMP (1 bytes for opcode, 4 for RELATIVE jump)
// /!\ TODO: Add checks on JMP overflow
trampoline_t *trampoline = get_free_trampoline();
uint8_t *pTrampolineCode = trampoline->trampolineBytes;
void* trampoline_page = page_addr(reinterpret_cast<void*>(trampoline), page_size());
// Enable write to the trampoline region
mprotect(trampoline_page, page_size()*2, PROT_READ|PROT_WRITE|PROT_EXEC);
// Create relative jmp to hook
gen_immediate_jump(trampoline->hookJump, hook);
// Copy original opcodes
trampoline->nOriginalBytes = code_len;
std::copy((uint8_t*)pOriginalFunc, ((uint8_t*)pOriginalFunc)+code_len, trampoline->originalBytes);
std::copy((uint8_t*)pOriginalFunc, ((uint8_t*)pOriginalFunc)+code_len, pTrampolineCode);
pTrampolineCode += code_len;
// Create the relative jmp to original (function + backed up opcodes)
pTrampolineCode = gen_immediate_jump(pTrampolineCode, pOriginalFunc+code_len);
pTrampolineCode += relative_addr_jump_size;
// Disable trampoline region write
mprotect(trampoline_page, page_size()*2, PROT_READ|PROT_EXEC);
cur_transaction.push_back({false, ppOriginalFunc, trampoline});
return 0;
}
/* ------ DOCUMENTATION ------
http://www.c-jump.com/CIS77/CPU/x86/lecture.html <- some help to understand [MOD][REG][R/M] (see paragraph #6)
http://shell-storm.org/online/Online-Assembler-and-Disassembler <- online assembler
http://ref.x86asm.net/coder32.html <- opcodes reference
X86
push ebx : 0x53
sub esp ?? : 0x83 0xEC 0x??
call ????????: 0xE8 0x?? 0x?? 0x?? 0x??
// relative jmp: ???????? = dst_addr - curr_addr - 5
jmp ???????? : 0xe9 0x?? 0x?? 0x?? 0x??
destination = 0x8dba8
jmp location: 0x91995 - opcodes: e9 0e c2 ff ff
0e c2 ff ff = 0x8dba8 - 0x91995 - 5
// short jmp: ?? = dst_addr - curr_addr - 2
jmp short ??: 0xeb 0x??
destination = 0x91964
jmp location: 0x9198f - opcodes: 0xeb 0xd3
d3 = 0x91964 - 0x9198f - 2
X64
TODO:
Hint: make relative jump to near (+/-2Gb) code : 5 bytes
load absolute addr into R11 and call a jmp on register R11 : 13 bytes
Or
Reuse x86 relative jmp method
Example:
mov r11, 0x0123456789abcdef -> 0x49 0xbb 0xef 0xcd 0xab 0x89 0x67 0x45 0x23 0x01
jmp r11 -> 0x41 0xff 0xe3
*/