unicorn/uc.c

1336 lines
36 KiB
C

/* Unicorn Emulator Engine */
/* By Nguyen Anh Quynh <aquynh@gmail.com>, 2015 */
#if defined(UNICORN_HAS_OSXKERNEL)
#include <libkern/libkern.h>
#else
#include <stddef.h>
#include <stdio.h>
#include <stdlib.h>
#endif
#include <time.h> // nanosleep
#include <string.h>
#include "uc_priv.h"
// target specific headers
#include "qemu/target/arm/unicorn.h"
#include "qemu/target/i386/unicorn.h"
#include "qemu/target/m68k/unicorn.h"
#include "qemu/target/mips/unicorn.h"
#include "qemu/target/riscv/unicorn.h"
#include "qemu/target/sparc/unicorn.h"
#include "qemu/include/hw/boards.h"
#include "qemu/include/qemu/queue.h"
static void free_table(gpointer key, gpointer value, gpointer data)
{
TypeInfo *ti = (TypeInfo*) value;
g_hash_table_destroy(ti->class->properties);
g_free((void *) ti->class);
g_free((void *) ti->name);
g_free((void *) ti->parent);
g_free((void *) ti);
}
UNICORN_EXPORT
unsigned int uc_version(unsigned int *major, unsigned int *minor)
{
if (major != NULL && minor != NULL) {
*major = UC_API_MAJOR;
*minor = UC_API_MINOR;
}
return (UC_API_MAJOR << 8) + UC_API_MINOR;
}
UNICORN_EXPORT
uc_err uc_errno(uc_engine *uc)
{
return uc->errnum;
}
UNICORN_EXPORT
const char *uc_strerror(uc_err code)
{
switch(code) {
default:
return "Unknown error code";
case UC_ERR_OK:
return "OK (UC_ERR_OK)";
case UC_ERR_NOMEM:
return "No memory available or memory not present (UC_ERR_NOMEM)";
case UC_ERR_ARCH:
return "Invalid/unsupported architecture (UC_ERR_ARCH)";
case UC_ERR_HANDLE:
return "Invalid handle (UC_ERR_HANDLE)";
case UC_ERR_MODE:
return "Invalid mode (UC_ERR_MODE)";
case UC_ERR_VERSION:
return "Different API version between core & binding (UC_ERR_VERSION)";
case UC_ERR_READ_UNMAPPED:
return "Invalid memory read (UC_ERR_READ_UNMAPPED)";
case UC_ERR_WRITE_UNMAPPED:
return "Invalid memory write (UC_ERR_WRITE_UNMAPPED)";
case UC_ERR_FETCH_UNMAPPED:
return "Invalid memory fetch (UC_ERR_FETCH_UNMAPPED)";
case UC_ERR_HOOK:
return "Invalid hook type (UC_ERR_HOOK)";
case UC_ERR_INSN_INVALID:
return "Invalid instruction (UC_ERR_INSN_INVALID)";
case UC_ERR_MAP:
return "Invalid memory mapping (UC_ERR_MAP)";
case UC_ERR_WRITE_PROT:
return "Write to write-protected memory (UC_ERR_WRITE_PROT)";
case UC_ERR_READ_PROT:
return "Read from non-readable memory (UC_ERR_READ_PROT)";
case UC_ERR_FETCH_PROT:
return "Fetch from non-executable memory (UC_ERR_FETCH_PROT)";
case UC_ERR_ARG:
return "Invalid argument (UC_ERR_ARG)";
case UC_ERR_READ_UNALIGNED:
return "Read from unaligned memory (UC_ERR_READ_UNALIGNED)";
case UC_ERR_WRITE_UNALIGNED:
return "Write to unaligned memory (UC_ERR_WRITE_UNALIGNED)";
case UC_ERR_FETCH_UNALIGNED:
return "Fetch from unaligned memory (UC_ERR_FETCH_UNALIGNED)";
case UC_ERR_RESOURCE:
return "Insufficient resource (UC_ERR_RESOURCE)";
case UC_ERR_EXCEPTION:
return "Unhandled CPU exception (UC_ERR_EXCEPTION)";
case UC_ERR_TIMEOUT:
return "Emulation timed out (UC_ERR_TIMEOUT)";
}
}
UNICORN_EXPORT
bool uc_arch_supported(uc_arch arch)
{
switch (arch) {
#ifdef UNICORN_HAS_ARM
case UC_ARCH_ARM: return true;
#endif
#ifdef UNICORN_HAS_ARM64
case UC_ARCH_ARM64: return true;
#endif
#ifdef UNICORN_HAS_M68K
case UC_ARCH_M68K: return true;
#endif
#ifdef UNICORN_HAS_MIPS
case UC_ARCH_MIPS: return true;
#endif
#ifdef UNICORN_HAS_PPC
case UC_ARCH_PPC: return true;
#endif
#ifdef UNICORN_HAS_SPARC
case UC_ARCH_SPARC: return true;
#endif
#ifdef UNICORN_HAS_RISCV
case UC_ARCH_RISCV: return true;
#endif
#ifdef UNICORN_HAS_X86
case UC_ARCH_X86: return true;
#endif
/* Invalid or disabled arch */
default: return false;
}
}
UNICORN_EXPORT
uc_err uc_open(uc_arch arch, uc_mode mode, uc_engine **result)
{
struct uc_struct *uc;
if (arch < UC_ARCH_MAX) {
uc = calloc(1, sizeof(*uc));
if (!uc) {
// memory insufficient
return UC_ERR_NOMEM;
}
uc->errnum = UC_ERR_OK;
uc->arch = arch;
uc->mode = mode;
// uc->ram_list = { .blocks = QLIST_HEAD_INITIALIZER(ram_list.blocks) };
uc->ram_list.blocks.lh_first = NULL;
uc->memory_listeners.tqh_first = NULL;
uc->memory_listeners.tqh_last = &uc->memory_listeners.tqh_first;
uc->address_spaces.tqh_first = NULL;
uc->address_spaces.tqh_last = &uc->address_spaces.tqh_first;
uc->phys_map_node_alloc_hint = 16;
switch(arch) {
default:
break;
#ifdef UNICORN_HAS_M68K
case UC_ARCH_M68K:
if ((mode & ~UC_MODE_M68K_MASK) ||
!(mode & UC_MODE_BIG_ENDIAN)) {
free(uc);
return UC_ERR_MODE;
}
uc->init_arch = m68k_uc_init;
break;
#endif
#ifdef UNICORN_HAS_X86
case UC_ARCH_X86:
if ((mode & ~UC_MODE_X86_MASK) ||
(mode & UC_MODE_BIG_ENDIAN) ||
!(mode & (UC_MODE_16|UC_MODE_32|UC_MODE_64))) {
free(uc);
return UC_ERR_MODE;
}
uc->init_arch = x86_uc_init;
break;
#endif
#ifdef UNICORN_HAS_ARM
case UC_ARCH_ARM:
if ((mode & ~UC_MODE_ARM_MASK)) {
free(uc);
return UC_ERR_MODE;
}
if (mode & UC_MODE_BIG_ENDIAN) {
uc->init_arch = armeb_uc_init;
} else {
uc->init_arch = arm_uc_init;
}
if (mode & UC_MODE_THUMB)
uc->thumb = 1;
break;
#endif
#ifdef UNICORN_HAS_ARM64
case UC_ARCH_ARM64:
if (mode & ~UC_MODE_ARM_MASK) {
free(uc);
return UC_ERR_MODE;
}
if (mode & UC_MODE_BIG_ENDIAN) {
uc->init_arch = arm64eb_uc_init;
} else {
uc->init_arch = arm64_uc_init;
}
break;
#endif
#if defined(UNICORN_HAS_MIPS) || defined(UNICORN_HAS_MIPSEL) || defined(UNICORN_HAS_MIPS64) || defined(UNICORN_HAS_MIPS64EL)
case UC_ARCH_MIPS:
if ((mode & ~UC_MODE_MIPS_MASK) ||
!(mode & (UC_MODE_MIPS32|UC_MODE_MIPS64))) {
free(uc);
return UC_ERR_MODE;
}
if (mode & UC_MODE_BIG_ENDIAN) {
#ifdef UNICORN_HAS_MIPS
if (mode & UC_MODE_MIPS32)
uc->init_arch = mips_uc_init;
#endif
#ifdef UNICORN_HAS_MIPS64
if (mode & UC_MODE_MIPS64)
uc->init_arch = mips64_uc_init;
#endif
} else { // little endian
#ifdef UNICORN_HAS_MIPSEL
if (mode & UC_MODE_MIPS32)
uc->init_arch = mipsel_uc_init;
#endif
#ifdef UNICORN_HAS_MIPS64EL
if (mode & UC_MODE_MIPS64)
uc->init_arch = mips64el_uc_init;
#endif
}
break;
#endif
#ifdef UNICORN_HAS_RISCV
case UC_ARCH_RISCV:
if (mode & ~UC_MODE_RISCV_MASK) {
free(uc);
return UC_ERR_MODE;
}
if (mode & UC_MODE_RISCV64) {
#ifdef UNICORN_HAS_RISCV64
uc->init_arch = riscv64_uc_init;
#endif
} else {
#ifdef UNICORN_HAS_RISCV32
uc->init_arch = riscv32_uc_init;
#endif
}
break;
#endif /* UNICORN_HAS_RISCV */
#ifdef UNICORN_HAS_SPARC
case UC_ARCH_SPARC:
if ((mode & ~UC_MODE_SPARC_MASK) ||
!(mode & UC_MODE_BIG_ENDIAN) ||
!(mode & (UC_MODE_SPARC32|UC_MODE_SPARC64))) {
free(uc);
return UC_ERR_MODE;
}
if (mode & UC_MODE_SPARC64)
uc->init_arch = sparc64_uc_init;
else
uc->init_arch = sparc_uc_init;
break;
#endif
}
if (uc->init_arch == NULL) {
return UC_ERR_ARCH;
}
if (machine_initialize(uc)) {
return UC_ERR_RESOURCE;
}
*result = uc;
if (uc->reg_reset)
uc->reg_reset(uc);
return UC_ERR_OK;
} else {
return UC_ERR_ARCH;
}
}
static void free_hooks(uc_engine *uc)
{
struct list_item *cur;
struct hook *hook;
int i;
// free hooks and hook lists
for (i = 0; i < UC_HOOK_MAX; i++) {
cur = uc->hook[i].head;
// hook can be in more than one list
// so we refcount to know when to free
while (cur) {
hook = (struct hook *)cur->data;
if (--hook->refs == 0) {
free(hook);
}
cur = cur->next;
}
list_clear(&uc->hook[i]);
}
}
UNICORN_EXPORT
uc_err uc_close(uc_engine *uc)
{
// Cleanup internally.
if (uc->release)
uc->release(uc->tcg_ctx);
g_free(uc->tcg_ctx);
// Cleanup CPU.
g_free(uc->cpu->cpu_ases);
g_free(uc->cpu->thread);
// Cleanup all objects.
OBJECT(uc->machine_state->accelerator)->ref = 1;
OBJECT(uc->machine_state)->ref = 1;
OBJECT(uc->owner)->ref = 1;
OBJECT(uc->root)->ref = 1;
object_unref(uc, OBJECT(uc->machine_state->accelerator));
object_unref(uc, OBJECT(uc->machine_state));
object_unref(uc, OBJECT(uc->cpu));
object_unref(uc, OBJECT(&uc->io_mem_notdirty));
object_unref(uc, OBJECT(&uc->io_mem_unassigned));
object_unref(uc, OBJECT(&uc->io_mem_rom));
object_unref(uc, OBJECT(uc->root));
// System memory.
g_free(uc->system_memory);
// Thread relateds.
if (uc->qemu_thread_data)
g_free(uc->qemu_thread_data);
// Other auxilaries.
free(uc->l1_map);
if (uc->bounce.buffer) {
free(uc->bounce.buffer);
}
g_hash_table_foreach(uc->type_table, free_table, uc);
g_hash_table_destroy(uc->type_table);
free_hooks(uc);
free(uc->mapped_blocks);
// finally, free uc itself.
memset(uc, 0, sizeof(*uc));
free(uc);
return UC_ERR_OK;
}
UNICORN_EXPORT
uc_err uc_reg_read_batch(uc_engine *uc, int *ids, void **vals, int count)
{
if (uc->reg_read)
uc->reg_read(uc, (unsigned int *)ids, vals, count);
else
return -1; // FIXME: need a proper uc_err
return UC_ERR_OK;
}
UNICORN_EXPORT
uc_err uc_reg_write_batch(uc_engine *uc, int *ids, void *const *vals, int count)
{
int ret = UC_ERR_OK;
if (uc->reg_write) {
ret = uc->reg_write(uc, (unsigned int *)ids, vals, count);
} else {
return UC_ERR_EXCEPTION; // FIXME: need a proper uc_err
}
return ret;
}
UNICORN_EXPORT
uc_err uc_reg_read(uc_engine *uc, int regid, void *value)
{
return uc_reg_read_batch(uc, &regid, &value, 1);
}
UNICORN_EXPORT
uc_err uc_reg_write(uc_engine *uc, int regid, const void *value)
{
return uc_reg_write_batch(uc, &regid, (void *const *)&value, 1);
}
// check if a memory area is mapped
// this is complicated because an area can overlap adjacent blocks
static bool check_mem_area(uc_engine *uc, uint64_t address, size_t size)
{
size_t count = 0, len;
while(count < size) {
MemoryRegion *mr = memory_mapping(uc, address);
if (mr) {
len = (size_t)MIN(size - count, mr->end - address);
count += len;
address += len;
} else // this address is not mapped in yet
break;
}
return (count == size);
}
UNICORN_EXPORT
uc_err uc_mem_read(uc_engine *uc, uint64_t address, void *_bytes, size_t size)
{
size_t count = 0, len;
uint8_t *bytes = _bytes;
if (uc->mem_redirect) {
address = uc->mem_redirect(address);
}
if (!check_mem_area(uc, address, size))
return UC_ERR_READ_UNMAPPED;
// memory area can overlap adjacent memory blocks
while(count < size) {
MemoryRegion *mr = memory_mapping(uc, address);
if (mr) {
len = (size_t)MIN(size - count, mr->end - address);
if (uc->read_mem(uc->cpu->as, address, bytes, len) == false)
break;
count += len;
address += len;
bytes += len;
} else // this address is not mapped in yet
break;
}
if (count == size)
return UC_ERR_OK;
else
return UC_ERR_READ_UNMAPPED;
}
UNICORN_EXPORT
uc_err uc_mem_write(uc_engine *uc, uint64_t address, const void *_bytes, size_t size)
{
size_t count = 0, len;
const uint8_t *bytes = _bytes;
if (uc->mem_redirect) {
address = uc->mem_redirect(address);
}
if (!check_mem_area(uc, address, size))
return UC_ERR_WRITE_UNMAPPED;
// memory area can overlap adjacent memory blocks
while(count < size) {
MemoryRegion *mr = memory_mapping(uc, address);
if (mr) {
uint32_t operms = mr->perms;
if (!(operms & UC_PROT_WRITE)) // write protected
// but this is not the program accessing memory, so temporarily mark writable
uc->readonly_mem(mr, false);
len = (size_t)MIN(size - count, mr->end - address);
if (uc->write_mem(uc->cpu->as, address, bytes, len) == false)
break;
if (!(operms & UC_PROT_WRITE)) // write protected
// now write protect it again
uc->readonly_mem(mr, true);
count += len;
address += len;
bytes += len;
} else // this address is not mapped in yet
break;
}
if (count == size)
return UC_ERR_OK;
else
return UC_ERR_WRITE_UNMAPPED;
}
#define TIMEOUT_STEP 2 // microseconds
static void *_timeout_fn(void *arg)
{
struct uc_struct *uc = arg;
int64_t current_time = get_clock();
do {
usleep(TIMEOUT_STEP);
// perhaps emulation is even done before timeout?
if (uc->emulation_done)
break;
} while((uint64_t)(get_clock() - current_time) < uc->timeout);
// timeout before emulation is done?
if (!uc->emulation_done) {
uc->timed_out = true;
// force emulation to stop
uc_emu_stop(uc);
}
return NULL;
}
static void enable_emu_timer(uc_engine *uc, uint64_t timeout)
{
uc->timeout = timeout;
qemu_thread_create(uc, &uc->timer, "timeout", _timeout_fn,
uc, QEMU_THREAD_JOINABLE);
}
static void hook_count_cb(struct uc_struct *uc, uint64_t address, uint32_t size, void *user_data)
{
// count this instruction. ah ah ah.
uc->emu_counter++;
if (uc->emu_counter > uc->emu_count)
uc_emu_stop(uc);
}
UNICORN_EXPORT
uc_err uc_emu_start(uc_engine* uc, uint64_t begin, uint64_t until, uint64_t timeout, size_t count)
{
// reset the counter
uc->emu_counter = 0;
uc->invalid_error = UC_ERR_OK;
uc->block_full = false;
uc->emulation_done = false;
uc->timed_out = false;
switch(uc->arch) {
default:
break;
#ifdef UNICORN_HAS_M68K
case UC_ARCH_M68K:
uc_reg_write(uc, UC_M68K_REG_PC, &begin);
break;
#endif
#ifdef UNICORN_HAS_X86
case UC_ARCH_X86:
switch(uc->mode) {
default:
break;
case UC_MODE_16: {
uint64_t ip;
uint16_t cs;
uc_reg_read(uc, UC_X86_REG_CS, &cs);
// compensate for later adding up IP & CS
ip = begin - cs*16;
uc_reg_write(uc, UC_X86_REG_IP, &ip);
break;
}
case UC_MODE_32:
uc_reg_write(uc, UC_X86_REG_EIP, &begin);
break;
case UC_MODE_64:
uc_reg_write(uc, UC_X86_REG_RIP, &begin);
break;
}
break;
#endif
#ifdef UNICORN_HAS_ARM
case UC_ARCH_ARM:
uc_reg_write(uc, UC_ARM_REG_R15, &begin);
break;
#endif
#ifdef UNICORN_HAS_ARM64
case UC_ARCH_ARM64:
uc_reg_write(uc, UC_ARM64_REG_PC, &begin);
break;
#endif
#ifdef UNICORN_HAS_MIPS
case UC_ARCH_MIPS:
// TODO: MIPS32/MIPS64/BIGENDIAN etc
uc_reg_write(uc, UC_MIPS_REG_PC, &begin);
break;
#endif
#ifdef UNICORN_HAS_RISCV
case UC_ARCH_RISCV:
uc_reg_write(uc, UC_RISCV_REG_PC, &begin);
break;
#endif
#ifdef UNICORN_HAS_SPARC
case UC_ARCH_SPARC:
// TODO: Sparc/Sparc64
uc_reg_write(uc, UC_SPARC_REG_PC, &begin);
break;
#endif
}
uc->stop_request = false;
uc->emu_count = count;
// remove count hook if counting isn't necessary
if (count <= 0 && uc->count_hook != 0) {
uc_hook_del(uc, uc->count_hook);
uc->count_hook = 0;
}
// set up count hook to count instructions.
if (count > 0 && uc->count_hook == 0) {
uc_err err;
// callback to count instructions must be run before everything else,
// so instead of appending, we must insert the hook at the begin
// of the hook list
uc->hook_insert = 1;
err = uc_hook_add(uc, &uc->count_hook, UC_HOOK_CODE, hook_count_cb, NULL, 1, 0);
// restore to append mode for uc_hook_add()
uc->hook_insert = 0;
if (err != UC_ERR_OK) {
return err;
}
}
uc->addr_end = until;
if (timeout)
enable_emu_timer(uc, timeout * 1000); // microseconds -> nanoseconds
if (uc->vm_start(uc)) {
return UC_ERR_RESOURCE;
}
// emulation is done
uc->emulation_done = true;
if (timeout) {
// wait for the timer to finish
qemu_thread_join(&uc->timer);
}
if (uc->timed_out) {
return UC_ERR_TIMEOUT;
}
return uc->invalid_error;
}
UNICORN_EXPORT
uc_err uc_emu_stop(uc_engine *uc)
{
if (uc->emulation_done)
return UC_ERR_OK;
uc->stop_request = true;
// TODO: make this atomic somehow?
if (uc->current_cpu) {
// exit the current TB
cpu_exit(uc->current_cpu);
}
return UC_ERR_OK;
}
// find if a memory range overlaps with existing mapped regions
static bool memory_overlap(struct uc_struct *uc, uint64_t begin, size_t size)
{
unsigned int i;
uint64_t end = begin + size - 1;
for(i = 0; i < uc->mapped_block_count; i++) {
// begin address falls inside this region?
if (begin >= uc->mapped_blocks[i]->addr && begin <= uc->mapped_blocks[i]->end - 1)
return true;
// end address falls inside this region?
if (end >= uc->mapped_blocks[i]->addr && end <= uc->mapped_blocks[i]->end - 1)
return true;
// this region falls totally inside this range?
if (begin < uc->mapped_blocks[i]->addr && end > uc->mapped_blocks[i]->end - 1)
return true;
}
// not found
return false;
}
// common setup/error checking shared between uc_mem_map and uc_mem_map_ptr
static uc_err mem_map(uc_engine *uc, uint64_t address, size_t size, uint32_t perms, MemoryRegion *block)
{
MemoryRegion **regions;
if (block == NULL)
return UC_ERR_NOMEM;
if ((uc->mapped_block_count & (MEM_BLOCK_INCR - 1)) == 0) { //time to grow
regions = (MemoryRegion**)g_realloc(uc->mapped_blocks,
sizeof(MemoryRegion*) * (uc->mapped_block_count + MEM_BLOCK_INCR));
if (regions == NULL) {
return UC_ERR_NOMEM;
}
uc->mapped_blocks = regions;
}
uc->mapped_blocks[uc->mapped_block_count] = block;
uc->mapped_block_count++;
return UC_ERR_OK;
}
static uc_err mem_map_check(uc_engine *uc, uint64_t address, size_t size, uint32_t perms)
{
if (size == 0)
// invalid memory mapping
return UC_ERR_ARG;
// address cannot wrapp around
if (address + size - 1 < address)
return UC_ERR_ARG;
// address must be aligned to uc->target_page_size
if ((address & uc->target_page_align) != 0)
return UC_ERR_ARG;
// size must be multiple of uc->target_page_size
if ((size & uc->target_page_align) != 0)
return UC_ERR_ARG;
// check for only valid permissions
if ((perms & ~UC_PROT_ALL) != 0)
return UC_ERR_ARG;
// this area overlaps existing mapped regions?
if (memory_overlap(uc, address, size)) {
return UC_ERR_MAP;
}
return UC_ERR_OK;
}
UNICORN_EXPORT
uc_err uc_mem_map(uc_engine *uc, uint64_t address, size_t size, uint32_t perms)
{
uc_err res;
if (uc->mem_redirect) {
address = uc->mem_redirect(address);
}
res = mem_map_check(uc, address, size, perms);
if (res)
return res;
return mem_map(uc, address, size, perms, uc->memory_map(uc, address, size, perms));
}
UNICORN_EXPORT
uc_err uc_mem_map_ptr(uc_engine *uc, uint64_t address, size_t size, uint32_t perms, void *ptr)
{
uc_err res;
if (ptr == NULL)
return UC_ERR_ARG;
if (uc->mem_redirect) {
address = uc->mem_redirect(address);
}
res = mem_map_check(uc, address, size, perms);
if (res)
return res;
return mem_map(uc, address, size, UC_PROT_ALL, uc->memory_map_ptr(uc, address, size, perms, ptr));
}
// Create a backup copy of the indicated MemoryRegion.
// Generally used in prepartion for splitting a MemoryRegion.
static uint8_t *copy_region(struct uc_struct *uc, MemoryRegion *mr)
{
uint8_t *block = (uint8_t *)g_malloc0((size_t)int128_get64(mr->size));
if (block != NULL) {
uc_err err = uc_mem_read(uc, mr->addr, block, (size_t)int128_get64(mr->size));
if (err != UC_ERR_OK) {
free(block);
block = NULL;
}
}
return block;
}
/*
Split the given MemoryRegion at the indicated address for the indicated size
this may result in the create of up to 3 spanning sections. If the delete
parameter is true, the no new section will be created to replace the indicate
range. This functions exists to support uc_mem_protect and uc_mem_unmap.
This is a static function and callers have already done some preliminary
parameter validation.
The do_delete argument indicates that we are being called to support
uc_mem_unmap. In this case we save some time by choosing NOT to remap
the areas that are intended to get unmapped
*/
// TODO: investigate whether qemu region manipulation functions already offered
// this capability
static bool split_region(struct uc_struct *uc, MemoryRegion *mr, uint64_t address,
size_t size, bool do_delete)
{
uint8_t *backup;
uint32_t perms;
uint64_t begin, end, chunk_end;
size_t l_size, m_size, r_size;
chunk_end = address + size;
// if this region belongs to area [address, address+size],
// then there is no work to do.
if (address <= mr->addr && chunk_end >= mr->end)
return true;
if (size == 0)
// trivial case
return true;
if (address >= mr->end || chunk_end <= mr->addr)
// impossible case
return false;
backup = copy_region(uc, mr);
if (backup == NULL)
return false;
// save the essential information required for the split before mr gets deleted
perms = mr->perms;
begin = mr->addr;
end = mr->end;
// unmap this region first, then do split it later
if (uc_mem_unmap(uc, mr->addr, (size_t)int128_get64(mr->size)) != UC_ERR_OK)
goto error;
/* overlapping cases
* |------mr------|
* case 1 |---size--|
* case 2 |--size--|
* case 3 |---size--|
*/
// adjust some things
if (address < begin)
address = begin;
if (chunk_end > end)
chunk_end = end;
// compute sub region sizes
l_size = (size_t)(address - begin);
r_size = (size_t)(end - chunk_end);
m_size = (size_t)(chunk_end - address);
// If there are error in any of the below operations, things are too far gone
// at that point to recover. Could try to remap orignal region, but these smaller
// allocation just failed so no guarantee that we can recover the original
// allocation at this point
if (l_size > 0) {
if (uc_mem_map(uc, begin, l_size, perms) != UC_ERR_OK)
goto error;
if (uc_mem_write(uc, begin, backup, l_size) != UC_ERR_OK)
goto error;
}
if (m_size > 0 && !do_delete) {
if (uc_mem_map(uc, address, m_size, perms) != UC_ERR_OK)
goto error;
if (uc_mem_write(uc, address, backup + l_size, m_size) != UC_ERR_OK)
goto error;
}
if (r_size > 0) {
if (uc_mem_map(uc, chunk_end, r_size, perms) != UC_ERR_OK)
goto error;
if (uc_mem_write(uc, chunk_end, backup + l_size + m_size, r_size) != UC_ERR_OK)
goto error;
}
free(backup);
return true;
error:
free(backup);
return false;
}
UNICORN_EXPORT
uc_err uc_mem_protect(struct uc_struct *uc, uint64_t address, size_t size, uint32_t perms)
{
MemoryRegion *mr;
uint64_t addr = address;
size_t count, len;
bool remove_exec = false;
if (size == 0)
// trivial case, no change
return UC_ERR_OK;
// address must be aligned to uc->target_page_size
if ((address & uc->target_page_align) != 0)
return UC_ERR_ARG;
// size must be multiple of uc->target_page_size
if ((size & uc->target_page_align) != 0)
return UC_ERR_ARG;
// check for only valid permissions
if ((perms & ~UC_PROT_ALL) != 0)
return UC_ERR_ARG;
if (uc->mem_redirect) {
address = uc->mem_redirect(address);
}
// check that user's entire requested block is mapped
if (!check_mem_area(uc, address, size))
return UC_ERR_NOMEM;
// Now we know entire region is mapped, so change permissions
// We may need to split regions if this area spans adjacent regions
addr = address;
count = 0;
while(count < size) {
mr = memory_mapping(uc, addr);
len = (size_t)MIN(size - count, mr->end - addr);
if (!split_region(uc, mr, addr, len, false))
return UC_ERR_NOMEM;
mr = memory_mapping(uc, addr);
// will this remove EXEC permission?
if (((mr->perms & UC_PROT_EXEC) != 0) && ((perms & UC_PROT_EXEC) == 0))
remove_exec = true;
mr->perms = perms;
uc->readonly_mem(mr, (perms & UC_PROT_WRITE) == 0);
count += len;
addr += len;
}
// if EXEC permission is removed, then quit TB and continue at the same place
if (remove_exec) {
uc->quit_request = true;
uc_emu_stop(uc);
}
return UC_ERR_OK;
}
UNICORN_EXPORT
uc_err uc_mem_unmap(struct uc_struct *uc, uint64_t address, size_t size)
{
MemoryRegion *mr;
uint64_t addr;
size_t count, len;
if (size == 0)
// nothing to unmap
return UC_ERR_OK;
// address must be aligned to uc->target_page_size
if ((address & uc->target_page_align) != 0)
return UC_ERR_ARG;
// size must be multiple of uc->target_page_size
if ((size & uc->target_page_align) != 0)
return UC_ERR_ARG;
if (uc->mem_redirect) {
address = uc->mem_redirect(address);
}
// check that user's entire requested block is mapped
if (!check_mem_area(uc, address, size))
return UC_ERR_NOMEM;
// Now we know entire region is mapped, so do the unmap
// We may need to split regions if this area spans adjacent regions
addr = address;
count = 0;
while(count < size) {
mr = memory_mapping(uc, addr);
len = (size_t)MIN(size - count, mr->end - addr);
if (!split_region(uc, mr, addr, len, true))
return UC_ERR_NOMEM;
// if we can retrieve the mapping, then no splitting took place
// so unmap here
mr = memory_mapping(uc, addr);
if (mr != NULL)
uc->memory_unmap(uc, mr);
count += len;
addr += len;
}
return UC_ERR_OK;
}
// find the memory region of this address
MemoryRegion *memory_mapping(struct uc_struct* uc, uint64_t address)
{
unsigned int i;
if (uc->mapped_block_count == 0)
return NULL;
if (uc->mem_redirect) {
address = uc->mem_redirect(address);
}
// try with the cache index first
i = uc->mapped_block_cache_index;
if (i < uc->mapped_block_count && address >= uc->mapped_blocks[i]->addr && address < uc->mapped_blocks[i]->end)
return uc->mapped_blocks[i];
for(i = 0; i < uc->mapped_block_count; i++) {
if (address >= uc->mapped_blocks[i]->addr && address <= uc->mapped_blocks[i]->end - 1) {
// cache this index for the next query
uc->mapped_block_cache_index = i;
return uc->mapped_blocks[i];
}
}
// not found
return NULL;
}
UNICORN_EXPORT
uc_err uc_hook_add(uc_engine *uc, uc_hook *hh, int type, void *callback,
void *user_data, uint64_t begin, uint64_t end, ...)
{
int ret = UC_ERR_OK;
int i = 0;
struct hook *hook = calloc(1, sizeof(struct hook));
if (hook == NULL) {
return UC_ERR_NOMEM;
}
hook->begin = begin;
hook->end = end;
hook->type = type;
hook->callback = callback;
hook->user_data = user_data;
hook->refs = 0;
*hh = (uc_hook)hook;
// UC_HOOK_INSN has an extra argument for instruction ID
if (type & UC_HOOK_INSN) {
va_list valist;
va_start(valist, end);
hook->insn = va_arg(valist, int);
va_end(valist);
if (uc->insn_hook_validate) {
if (! uc->insn_hook_validate(hook->insn)) {
free(hook);
return UC_ERR_HOOK;
}
}
if (uc->hook_insert) {
if (list_insert(&uc->hook[UC_HOOK_INSN_IDX], hook) == NULL) {
free(hook);
return UC_ERR_NOMEM;
}
} else {
if (list_append(&uc->hook[UC_HOOK_INSN_IDX], hook) == NULL) {
free(hook);
return UC_ERR_NOMEM;
}
}
hook->refs++;
return UC_ERR_OK;
}
while ((type >> i) > 0) {
if ((type >> i) & 1) {
// TODO: invalid hook error?
if (i < UC_HOOK_MAX) {
if (uc->hook_insert) {
if (list_insert(&uc->hook[i], hook) == NULL) {
if (hook->refs == 0) {
free(hook);
}
return UC_ERR_NOMEM;
}
} else {
if (list_append(&uc->hook[i], hook) == NULL) {
if (hook->refs == 0) {
free(hook);
}
return UC_ERR_NOMEM;
}
}
hook->refs++;
}
}
i++;
}
// we didn't use the hook
// TODO: return an error?
if (hook->refs == 0) {
free(hook);
}
return ret;
}
UNICORN_EXPORT
uc_err uc_hook_del(uc_engine *uc, uc_hook hh)
{
int i;
struct hook *hook = (struct hook *)hh;
// we can't dereference hook->type if hook is invalid
// so for now we need to iterate over all possible types to remove the hook
// which is less efficient
// an optimization would be to align the hook pointer
// and store the type mask in the hook pointer.
for (i = 0; i < UC_HOOK_MAX; i++) {
if (list_remove(&uc->hook[i], (void *)hook)) {
if (--hook->refs == 0) {
free(hook);
break;
}
}
}
return UC_ERR_OK;
}
// TCG helper
void helper_uc_tracecode(int32_t size, uc_hook_type type, void *handle, int64_t address);
void helper_uc_tracecode(int32_t size, uc_hook_type type, void *handle, int64_t address)
{
struct uc_struct *uc = handle;
struct list_item *cur = uc->hook[type].head;
struct hook *hook;
// sync PC in CPUArchState with address
if (uc->set_pc) {
uc->set_pc(uc, address);
}
while (cur != NULL && !uc->stop_request) {
hook = (struct hook *)cur->data;
if (HOOK_BOUND_CHECK(hook, (uint64_t)address)) {
((uc_cb_hookcode_t)hook->callback)(uc, address, size, hook->user_data);
}
cur = cur->next;
}
}
UNICORN_EXPORT
uint32_t uc_mem_regions(uc_engine *uc, uc_mem_region **regions, uint32_t *count)
{
uint32_t i;
uc_mem_region *r = NULL;
*count = uc->mapped_block_count;
if (*count) {
r = g_malloc0(*count * sizeof(uc_mem_region));
if (r == NULL) {
// out of memory
return UC_ERR_NOMEM;
}
}
for (i = 0; i < *count; i++) {
r[i].begin = uc->mapped_blocks[i]->addr;
r[i].end = uc->mapped_blocks[i]->end - 1;
r[i].perms = uc->mapped_blocks[i]->perms;
}
*regions = r;
return UC_ERR_OK;
}
UNICORN_EXPORT
uc_err uc_query(uc_engine *uc, uc_query_type type, size_t *result)
{
if (type == UC_QUERY_PAGE_SIZE) {
*result = uc->target_page_size;
return UC_ERR_OK;
}
if (type == UC_QUERY_ARCH) {
*result = uc->arch;
return UC_ERR_OK;
}
switch(uc->arch) {
#ifdef UNICORN_HAS_ARM
case UC_ARCH_ARM:
return uc->query(uc, type, result);
#endif
default:
return UC_ERR_ARG;
}
return UC_ERR_OK;
}
static size_t cpu_context_size(uc_arch arch, uc_mode mode)
{
// each of these constants is defined by offsetof(CPUXYZState, tlb_table)
// tbl_table is the first entry in the CPU_COMMON macro, so it marks the end
// of the interesting CPU registers
switch (arch) {
#ifdef UNICORN_HAS_M68K
case UC_ARCH_M68K: return M68K_REGS_STORAGE_SIZE;
#endif
#ifdef UNICORN_HAS_X86
case UC_ARCH_X86: return X86_REGS_STORAGE_SIZE;
#endif
#ifdef UNICORN_HAS_ARM
case UC_ARCH_ARM: return mode & UC_MODE_BIG_ENDIAN ? ARM_REGS_STORAGE_SIZE_armeb : ARM_REGS_STORAGE_SIZE_arm;
#endif
#ifdef UNICORN_HAS_ARM64
case UC_ARCH_ARM64: return mode & UC_MODE_BIG_ENDIAN ? ARM64_REGS_STORAGE_SIZE_aarch64eb : ARM64_REGS_STORAGE_SIZE_aarch64;
#endif
#ifdef UNICORN_HAS_MIPS
case UC_ARCH_MIPS:
if (mode & UC_MODE_MIPS64) {
if (mode & UC_MODE_BIG_ENDIAN) {
return MIPS64_REGS_STORAGE_SIZE_mips64;
} else {
return MIPS64_REGS_STORAGE_SIZE_mips64el;
}
} else {
if (mode & UC_MODE_BIG_ENDIAN) {
return MIPS_REGS_STORAGE_SIZE_mips;
} else {
return MIPS_REGS_STORAGE_SIZE_mipsel;
}
}
#endif
#ifdef UNICORN_HAS_RISCV
case UC_ARCH_RISCV:
if (mode & UC_MODE_RISCV64) {
return RISCV64_REGS_STORAGE_SIZE_riscv64;
}
return RISCV32_REGS_STORAGE_SIZE_riscv32;
#endif
#ifdef UNICORN_HAS_SPARC
case UC_ARCH_SPARC: return mode & UC_MODE_SPARC64 ? SPARC64_REGS_STORAGE_SIZE : SPARC_REGS_STORAGE_SIZE;
#endif
default: return 0;
}
}
UNICORN_EXPORT
uc_err uc_context_alloc(uc_engine *uc, uc_context **context)
{
struct uc_context **_context = context;
size_t size = cpu_context_size(uc->arch, uc->mode);
*_context = malloc(size + sizeof(uc_context));
if (*_context) {
(*_context)->size = size;
return UC_ERR_OK;
} else {
return UC_ERR_NOMEM;
}
}
UNICORN_EXPORT
uc_err uc_free(void *mem)
{
g_free(mem);
return UC_ERR_OK;
}
UNICORN_EXPORT
size_t uc_context_size(uc_engine *uc)
{
return cpu_context_size(uc->arch, uc->mode);
}
UNICORN_EXPORT
uc_err uc_context_save(uc_engine *uc, uc_context *context)
{
struct uc_context *_context = context;
memcpy(_context->data, uc->cpu->env_ptr, _context->size);
return UC_ERR_OK;
}
UNICORN_EXPORT
uc_err uc_context_restore(uc_engine *uc, uc_context *context)
{
struct uc_context *_context = context;
memcpy(uc->cpu->env_ptr, _context->data, _context->size);
return UC_ERR_OK;
}