/* Capstone Disassembly Engine */ /* By Nguyen Anh Quynh , 2013-2014 */ #if defined (WIN32) || defined (WIN64) || defined (_WIN32) || defined (_WIN64) #pragma warning(disable:4996) #endif #include #include #include #include #include #include "utils.h" #include "MCRegisterInfo.h" #ifdef CAPSTONE_USE_SYS_DYN_MEM #define INSN_CACHE_SIZE 32 #else // reduce stack variable size for kernel/firmware #define INSN_CACHE_SIZE 8 #endif // default SKIPDATA mnemonic #define SKIPDATA_MNEM ".byte" cs_err (*arch_init[MAX_ARCH])(cs_struct *) = { NULL }; cs_err (*arch_option[MAX_ARCH]) (cs_struct *, cs_opt_type, size_t value) = { NULL }; void (*arch_destroy[MAX_ARCH]) (cs_struct *) = { NULL }; extern void ARM_enable(void); extern void AArch64_enable(void); extern void Mips_enable(void); extern void X86_enable(void); extern void PPC_enable(void); extern void Sparc_enable(void); extern void SystemZ_enable(void); extern void XCore_enable(void); static void archs_enable(void) { static bool initialized = false; if (initialized) return; #ifdef CAPSTONE_HAS_ARM ARM_enable(); #endif #ifdef CAPSTONE_HAS_ARM64 AArch64_enable(); #endif #ifdef CAPSTONE_HAS_MIPS Mips_enable(); #endif #ifdef CAPSTONE_HAS_POWERPC PPC_enable(); #endif #ifdef CAPSTONE_HAS_SPARC Sparc_enable(); #endif #ifdef CAPSTONE_HAS_SYSZ SystemZ_enable(); #endif #ifdef CAPSTONE_HAS_X86 X86_enable(); #endif #ifdef CAPSTONE_HAS_XCORE XCore_enable(); #endif initialized = true; } unsigned int all_arch = 0; #ifdef CAPSTONE_USE_SYS_DYN_MEM cs_malloc_t cs_mem_malloc = malloc; cs_calloc_t cs_mem_calloc = calloc; cs_realloc_t cs_mem_realloc = realloc; cs_free_t cs_mem_free = free; cs_vsnprintf_t cs_vsnprintf = vsnprintf; #else cs_malloc_t cs_mem_malloc = NULL; cs_calloc_t cs_mem_calloc = NULL; cs_realloc_t cs_mem_realloc = NULL; cs_free_t cs_mem_free = NULL; cs_vsnprintf_t cs_vsnprintf = NULL; #endif CAPSTONE_EXPORT unsigned int cs_version(int *major, int *minor) { archs_enable(); if (major != NULL && minor != NULL) { *major = CS_API_MAJOR; *minor = CS_API_MINOR; } return (CS_API_MAJOR << 8) + CS_API_MINOR; } CAPSTONE_EXPORT bool cs_support(int query) { archs_enable(); if (query == CS_ARCH_ALL) return all_arch == ((1 << CS_ARCH_ARM) | (1 << CS_ARCH_ARM64) | (1 << CS_ARCH_MIPS) | (1 << CS_ARCH_X86) | (1 << CS_ARCH_PPC) | (1 << CS_ARCH_SPARC) | (1 << CS_ARCH_SYSZ) | (1 << CS_ARCH_XCORE)); if ((unsigned int)query < CS_ARCH_MAX) return all_arch & (1 << query); if (query == CS_SUPPORT_DIET) { #ifdef CAPSTONE_DIET return true; #else return false; #endif } if (query == CS_SUPPORT_X86_REDUCE) { #if defined(CAPSTONE_HAS_X86) && defined(CAPSTONE_X86_REDUCE) return true; #else return false; #endif } // unsupported query return false; } CAPSTONE_EXPORT cs_err cs_errno(csh handle) { struct cs_struct *ud; if (!handle) return CS_ERR_CSH; ud = (struct cs_struct *)(uintptr_t)handle; return ud->errnum; } CAPSTONE_EXPORT const char *cs_strerror(cs_err code) { switch(code) { default: return "Unknown error code"; case CS_ERR_OK: return "OK (CS_ERR_OK)"; case CS_ERR_MEM: return "Out of memory (CS_ERR_MEM)"; case CS_ERR_ARCH: return "Invalid architecture (CS_ERR_ARCH)"; case CS_ERR_HANDLE: return "Invalid handle (CS_ERR_HANDLE)"; case CS_ERR_CSH: return "Invalid csh (CS_ERR_CSH)"; case CS_ERR_MODE: return "Invalid mode (CS_ERR_MODE)"; case CS_ERR_OPTION: return "Invalid option (CS_ERR_OPTION)"; case CS_ERR_DETAIL: return "Details are unavailable (CS_ERR_DETAIL)"; case CS_ERR_MEMSETUP: return "Dynamic memory management uninitialized (CS_ERR_MEMSETUP)"; case CS_ERR_VERSION: return "Different API version between core & binding (CS_ERR_VERSION)"; case CS_ERR_DIET: return "Information irrelevant in diet engine (CS_ERR_DIET)"; case CS_ERR_SKIPDATA: return "Information irrelevant for 'data' instruction in SKIPDATA mode (CS_ERR_SKIPDATA)"; } } CAPSTONE_EXPORT cs_err cs_open(cs_arch arch, cs_mode mode, csh *handle) { cs_err err; struct cs_struct *ud; if (!cs_mem_malloc || !cs_mem_calloc || !cs_mem_realloc || !cs_mem_free || !cs_vsnprintf) // Error: before cs_open(), dynamic memory management must be initialized // with cs_option(CS_OPT_MEM) return CS_ERR_MEMSETUP; archs_enable(); if (arch < CS_ARCH_MAX && arch_init[arch]) { ud = cs_mem_calloc(1, sizeof(*ud)); if (!ud) { // memory insufficient return CS_ERR_MEM; } ud->errnum = CS_ERR_OK; ud->arch = arch; ud->mode = mode; ud->big_endian = mode & CS_MODE_BIG_ENDIAN; // by default, do not break instruction into details ud->detail = CS_OPT_OFF; // default skipdata setup ud->skipdata_setup.mnemonic = SKIPDATA_MNEM; err = arch_init[ud->arch](ud); if (err) { cs_mem_free(ud); *handle = 0; return err; } *handle = (uintptr_t)ud; return CS_ERR_OK; } else { *handle = 0; return CS_ERR_ARCH; } } CAPSTONE_EXPORT cs_err cs_close(csh *handle) { struct cs_struct *ud; if (*handle == 0) // invalid handle return CS_ERR_CSH; ud = (struct cs_struct *)(*handle); if (ud->printer_info) cs_mem_free(ud->printer_info); // arch_destroy[ud->arch](ud); cs_mem_free(ud->insn_cache); memset(ud, 0, sizeof(*ud)); cs_mem_free(ud); // invalidate this handle by ZERO out its value. // this is to make sure it is unusable after cs_close() *handle = 0; return CS_ERR_OK; } // fill insn with mnemonic & operands info static void fill_insn(struct cs_struct *handle, cs_insn *insn, char *buffer, MCInst *mci, PostPrinter_t postprinter, const uint8_t *code) { #ifndef CAPSTONE_DIET char *sp, *mnem; #endif // fill the instruction bytes memcpy(insn->bytes, code, MIN(sizeof(insn->bytes), insn->size)); // map internal instruction opcode to public insn ID handle->insn_id(handle, insn, MCInst_getOpcode(mci)); // alias instruction might have ID saved in OpcodePub if (MCInst_getOpcodePub(mci)) insn->id = MCInst_getOpcodePub(mci); // post printer handles some corner cases (hacky) if (postprinter) postprinter((csh)handle, insn, buffer, mci); #ifndef CAPSTONE_DIET // fill in mnemonic & operands // find first space or tab sp = buffer; mnem = insn->mnemonic; for (sp = buffer; *sp; sp++) { if (*sp == ' '|| *sp == '\t') break; if (*sp == '|') // lock|rep prefix for x86 *sp = ' '; // copy to @mnemonic *mnem = *sp; mnem++; } *mnem = '\0'; // copy @op_str if (*sp) { *sp = '\0'; // find the next non-space char sp++; for (; ((*sp == ' ') || (*sp == '\t')); sp++); strncpy(insn->op_str, sp, sizeof(insn->op_str) - 1); insn->op_str[sizeof(insn->op_str) - 1] = '\0'; } else insn->op_str[0] = '\0'; #endif } // how many bytes will we skip when encountering data (CS_OPT_SKIPDATA)? // this very much depends on instruction alignment requirement of each arch. static uint8_t skipdata_size(cs_struct *handle) { switch(handle->arch) { default: // should never reach return -1; case CS_ARCH_ARM: // skip 2 bytes on Thumb mode. if (handle->mode & CS_MODE_THUMB) return 2; // otherwise, skip 4 bytes return 4; case CS_ARCH_ARM64: case CS_ARCH_MIPS: case CS_ARCH_PPC: case CS_ARCH_SPARC: // skip 4 bytes return 4; case CS_ARCH_SYSZ: // SystemZ instruction's length can be 2, 4 or 6 bytes, // so we just skip 2 bytes return 2; case CS_ARCH_X86: // X86 has no restriction on instruction alignment return 1; case CS_ARCH_XCORE: // XCore instruction's length can be 2 or 4 bytes, // so we just skip 2 bytes return 2; } } CAPSTONE_EXPORT cs_err cs_option(csh ud, cs_opt_type type, size_t value) { struct cs_struct *handle; archs_enable(); // cs_option() can be called with NULL handle just for CS_OPT_MEM // This is supposed to be executed before all other APIs (even cs_open()) if (type == CS_OPT_MEM) { cs_opt_mem *mem = (cs_opt_mem *)value; cs_mem_malloc = mem->malloc; cs_mem_calloc = mem->calloc; cs_mem_realloc = mem->realloc; cs_mem_free = mem->free; cs_vsnprintf = mem->vsnprintf; return CS_ERR_OK; } handle = (struct cs_struct *)(uintptr_t)ud; if (!handle) return CS_ERR_CSH; switch(type) { default: break; case CS_OPT_DETAIL: handle->detail = value; return CS_ERR_OK; case CS_OPT_SKIPDATA: handle->skipdata = (value == CS_OPT_ON); if (handle->skipdata) { if (handle->skipdata_size == 0) { // set the default skipdata size handle->skipdata_size = skipdata_size(handle); } } return CS_ERR_OK; case CS_OPT_SKIPDATA_SETUP: if (value) handle->skipdata_setup = *((cs_opt_skipdata *)value); return CS_ERR_OK; } return arch_option[handle->arch](handle, type, value); } // generate @op_str for data instruction of SKIPDATA static void skipdata_opstr(char *opstr, const uint8_t *buffer, size_t size) { char *p = opstr; int len; size_t i; if (!size) { opstr[0] = '\0'; return; } len = sprintf(p, "0x%02x", buffer[0]); p+= len; for(i = 1; i < size; i++) { len = sprintf(p, ", 0x%02x", buffer[i]); p+= len; } } // dynamicly allocate memory to contain disasm insn // NOTE: caller must free() the allocated memory itself to avoid memory leaking CAPSTONE_EXPORT size_t cs_disasm_ex(csh ud, const uint8_t *buffer, size_t size, uint64_t offset, size_t count, cs_insn **insn) { struct cs_struct *handle = (struct cs_struct *)(uintptr_t)ud; MCInst mci; uint16_t insn_size; size_t c = 0, i; unsigned int f = 0; cs_insn *insn_cache; void *total = NULL; size_t total_size = 0; bool r; void *tmp; size_t skipdata_bytes; // save all the original info of the buffer uint64_t offset_org; size_t size_org; const uint8_t *buffer_org; if (!handle) { // FIXME: how to handle this case: // handle->errnum = CS_ERR_HANDLE; return 0; } handle->errnum = CS_ERR_OK; // save the original offset for SKIPDATA buffer_org = buffer; offset_org = offset; size_org = size; total_size = (sizeof(cs_insn) * INSN_CACHE_SIZE); total = cs_mem_malloc(total_size); insn_cache = total; while (size > 0) { MCInst_Init(&mci); mci.csh = handle; // relative branches need to know the address & size of current insn mci.address = offset; if (handle->detail) { // allocate memory for @detail pointer insn_cache->detail = cs_mem_malloc(sizeof(cs_detail)); } else { insn_cache->detail = NULL; } // save all the information for non-detailed mode mci.flat_insn = insn_cache; mci.flat_insn->address = offset; r = handle->disasm(ud, buffer, size, &mci, &insn_size, offset, handle->getinsn_info); if (r) { SStream ss; SStream_Init(&ss); mci.flat_insn->size = insn_size; handle->printer(&mci, &ss, handle->printer_info); fill_insn(handle, insn_cache, ss.buffer, &mci, handle->post_printer, buffer); f++; if (f == INSN_CACHE_SIZE) { // resize total to contain newly disasm insns total_size += (sizeof(cs_insn) * INSN_CACHE_SIZE); tmp = cs_mem_realloc(total, total_size); if (tmp == NULL) { // insufficient memory if (handle->detail) { insn_cache = (cs_insn *)total; for (i = 0; i < c; i++, insn_cache++) cs_mem_free(insn_cache->detail); } cs_mem_free(total); *insn = NULL; handle->errnum = CS_ERR_MEM; return 0; } total = tmp; insn_cache = (cs_insn *)((char *)total + total_size - (sizeof(cs_insn) * INSN_CACHE_SIZE)); // reset f back to 0 f = 0; } else insn_cache++; c++; if (count > 0 && c == count) break; buffer += insn_size; size -= insn_size; offset += insn_size; } else { if (handle->detail) { // free memory of @detail pointer cs_mem_free(insn_cache->detail); } // encounter a broken instruction // if there is no request to skip data, or remaining data is too small, // then bail out if (!handle->skipdata || handle->skipdata_size > size) break; if (handle->skipdata_setup.callback) { skipdata_bytes = handle->skipdata_setup.callback(buffer_org, size_org, offset - offset_org, handle->skipdata_setup.user_data); if (skipdata_bytes > size) // remaining data is not enough break; if (!skipdata_bytes) // user requested not to skip data, so bail out break; } else skipdata_bytes = handle->skipdata_size; // we have to skip some amount of data, depending on arch & mode insn_cache->id = 0; // invalid ID for this "data" instruction insn_cache->address = offset; insn_cache->size = (uint16_t) skipdata_bytes; memcpy(insn_cache->bytes, buffer, skipdata_bytes); strncpy(insn_cache->mnemonic, handle->skipdata_setup.mnemonic, sizeof(insn_cache->mnemonic) - 1); skipdata_opstr(insn_cache->op_str, buffer, skipdata_bytes); insn_cache->detail = NULL; f++; if (f == INSN_CACHE_SIZE) { // resize total to contain newly disasm insns total_size += (sizeof(cs_insn) * INSN_CACHE_SIZE); tmp = cs_mem_realloc(total, total_size); if (tmp == NULL) { // insufficient memory if (handle->detail) { insn_cache = (cs_insn *)total; for (i = 0; i < c; i++, insn_cache++) cs_mem_free(insn_cache->detail); } cs_mem_free(total); *insn = NULL; handle->errnum = CS_ERR_MEM; return 0; } total = tmp; insn_cache = (cs_insn *)((char *)total + total_size - (sizeof(cs_insn) * INSN_CACHE_SIZE)); // reset f back to 0 f = 0; } else insn_cache++; buffer += skipdata_bytes; size -= skipdata_bytes; offset += skipdata_bytes; c++; } } if (f) { // resize total to contain newly disasm insns void *tmp = cs_mem_realloc(total, total_size - (INSN_CACHE_SIZE - f) * sizeof(*insn_cache)); if (tmp == NULL) { // insufficient memory // free all detail pointers if (handle->detail) { insn_cache = (cs_insn *)total; for (i = 0; i < c; i++, insn_cache++) cs_mem_free(insn_cache->detail); } cs_mem_free(total); *insn = NULL; handle->errnum = CS_ERR_MEM; return 0; } total = tmp; } else if (!c) { cs_mem_free(total); total = NULL; } *insn = total; return c; } CAPSTONE_EXPORT void cs_free(cs_insn *insn, size_t count) { size_t i; // free all detail pointers for (i = 0; i < count; i++) cs_mem_free(insn[i].detail); // then free pointer to cs_insn array cs_mem_free(insn); } // return friendly name of regiser in a string CAPSTONE_EXPORT const char *cs_reg_name(csh ud, unsigned int reg) { struct cs_struct *handle = (struct cs_struct *)(uintptr_t)ud; if (!handle || handle->reg_name == NULL) { return NULL; } return handle->reg_name(ud, reg); } CAPSTONE_EXPORT const char *cs_insn_name(csh ud, unsigned int insn) { struct cs_struct *handle = (struct cs_struct *)(uintptr_t)ud; if (!handle || handle->insn_name == NULL) { return NULL; } return handle->insn_name(ud, insn); } CAPSTONE_EXPORT const char *cs_group_name(csh ud, unsigned int group) { struct cs_struct *handle = (struct cs_struct *)(uintptr_t)ud; if (!handle || handle->group_name == NULL) { return NULL; } return handle->group_name(ud, group); } static bool arr_exist(unsigned char *arr, unsigned char max, unsigned int id) { int i; for (i = 0; i < max; i++) { if (arr[i] == id) return true; } return false; } CAPSTONE_EXPORT bool cs_insn_group(csh ud, cs_insn *insn, unsigned int group_id) { struct cs_struct *handle; if (!ud) return false; handle = (struct cs_struct *)(uintptr_t)ud; if (!handle->detail) { handle->errnum = CS_ERR_DETAIL; return false; } if(!insn->id) { handle->errnum = CS_ERR_SKIPDATA; return false; } if(!insn->detail) { handle->errnum = CS_ERR_DETAIL; return false; } return arr_exist(insn->detail->groups, insn->detail->groups_count, group_id); } CAPSTONE_EXPORT bool cs_reg_read(csh ud, cs_insn *insn, unsigned int reg_id) { struct cs_struct *handle; if (!ud) return false; handle = (struct cs_struct *)(uintptr_t)ud; if (!handle->detail) { handle->errnum = CS_ERR_DETAIL; return false; } if(!insn->id) { handle->errnum = CS_ERR_SKIPDATA; return false; } if(!insn->detail) { handle->errnum = CS_ERR_DETAIL; return false; } return arr_exist(insn->detail->regs_read, insn->detail->regs_read_count, reg_id); } CAPSTONE_EXPORT bool cs_reg_write(csh ud, cs_insn *insn, unsigned int reg_id) { struct cs_struct *handle; if (!ud) return false; handle = (struct cs_struct *)(uintptr_t)ud; if (!handle->detail) { handle->errnum = CS_ERR_DETAIL; return false; } if(!insn->id) { handle->errnum = CS_ERR_SKIPDATA; return false; } if(!insn->detail) { handle->errnum = CS_ERR_DETAIL; return false; } return arr_exist(insn->detail->regs_write, insn->detail->regs_write_count, reg_id); } CAPSTONE_EXPORT int cs_op_count(csh ud, cs_insn *insn, unsigned int op_type) { struct cs_struct *handle; unsigned int count = 0, i; if (!ud) return -1; handle = (struct cs_struct *)(uintptr_t)ud; if (!handle->detail) { handle->errnum = CS_ERR_DETAIL; return -1; } if(!insn->id) { handle->errnum = CS_ERR_SKIPDATA; return -1; } if(!insn->detail) { handle->errnum = CS_ERR_DETAIL; return -1; } handle->errnum = CS_ERR_OK; switch (handle->arch) { default: handle->errnum = CS_ERR_HANDLE; return -1; case CS_ARCH_ARM: for (i = 0; i < insn->detail->arm.op_count; i++) if (insn->detail->arm.operands[i].type == (arm_op_type)op_type) count++; break; case CS_ARCH_ARM64: for (i = 0; i < insn->detail->arm64.op_count; i++) if (insn->detail->arm64.operands[i].type == (arm64_op_type)op_type) count++; break; case CS_ARCH_X86: for (i = 0; i < insn->detail->x86.op_count; i++) if (insn->detail->x86.operands[i].type == (x86_op_type)op_type) count++; break; case CS_ARCH_MIPS: for (i = 0; i < insn->detail->mips.op_count; i++) if (insn->detail->mips.operands[i].type == (mips_op_type)op_type) count++; break; case CS_ARCH_PPC: for (i = 0; i < insn->detail->ppc.op_count; i++) if (insn->detail->ppc.operands[i].type == (ppc_op_type)op_type) count++; break; case CS_ARCH_SPARC: for (i = 0; i < insn->detail->sparc.op_count; i++) if (insn->detail->sparc.operands[i].type == (sparc_op_type)op_type) count++; break; case CS_ARCH_SYSZ: for (i = 0; i < insn->detail->sysz.op_count; i++) if (insn->detail->sysz.operands[i].type == (sysz_op_type)op_type) count++; break; case CS_ARCH_XCORE: for (i = 0; i < insn->detail->xcore.op_count; i++) if (insn->detail->xcore.operands[i].type == (xcore_op_type)op_type) count++; break; } return count; } CAPSTONE_EXPORT int cs_op_index(csh ud, cs_insn *insn, unsigned int op_type, unsigned int post) { struct cs_struct *handle; unsigned int count = 0, i; if (!ud) return -1; handle = (struct cs_struct *)(uintptr_t)ud; if (!handle->detail) { handle->errnum = CS_ERR_DETAIL; return -1; } if(!insn->id) { handle->errnum = CS_ERR_SKIPDATA; return -1; } if(!insn->detail) { handle->errnum = CS_ERR_DETAIL; return -1; } handle->errnum = CS_ERR_OK; switch (handle->arch) { default: handle->errnum = CS_ERR_HANDLE; return -1; case CS_ARCH_ARM: for (i = 0; i < insn->detail->arm.op_count; i++) { if (insn->detail->arm.operands[i].type == (arm_op_type)op_type) count++; if (count == post) return i; } break; case CS_ARCH_ARM64: for (i = 0; i < insn->detail->arm64.op_count; i++) { if (insn->detail->arm64.operands[i].type == (arm64_op_type)op_type) count++; if (count == post) return i; } break; case CS_ARCH_X86: for (i = 0; i < insn->detail->x86.op_count; i++) { if (insn->detail->x86.operands[i].type == (x86_op_type)op_type) count++; if (count == post) return i; } break; case CS_ARCH_MIPS: for (i = 0; i < insn->detail->mips.op_count; i++) { if (insn->detail->mips.operands[i].type == (mips_op_type)op_type) count++; if (count == post) return i; } break; case CS_ARCH_PPC: for (i = 0; i < insn->detail->ppc.op_count; i++) { if (insn->detail->ppc.operands[i].type == (ppc_op_type)op_type) count++; if (count == post) return i; } break; case CS_ARCH_SPARC: for (i = 0; i < insn->detail->sparc.op_count; i++) { if (insn->detail->sparc.operands[i].type == (sparc_op_type)op_type) count++; if (count == post) return i; } break; case CS_ARCH_SYSZ: for (i = 0; i < insn->detail->sysz.op_count; i++) { if (insn->detail->sysz.operands[i].type == (sysz_op_type)op_type) count++; if (count == post) return i; } break; case CS_ARCH_XCORE: for (i = 0; i < insn->detail->xcore.op_count; i++) { if (insn->detail->xcore.operands[i].type == (xcore_op_type)op_type) count++; if (count == post) return i; } break; } return -1; }