Source: https://code.google.com/p/google-security-research/issues/detail?id=385&can=1
We have encountered a number of Windows kernel crashes in the ATMFD.DLL OpenType driver while processing corrupted OTF font files, such as:
---
PAGE_FAULT_IN_NONPAGED_AREA (50)
Invalid system memory was referenced. This cannot be protected by try-except,
it must be protected by a Probe. Typically the address is just plain bad or it
is pointing at freed memory.
Arguments:
Arg1: a3a3a3db, memory referenced.
Arg2: 00000001, value 0 = read operation, 1 = write operation.
Arg3: 91f445c9, If non-zero, the instruction address which referenced the bad memory
address.
Arg4: 00000002, (reserved)
Debugging Details:
------------------
*** ERROR: Module load completed but symbols could not be loaded for ATMFD.DLL
WRITE_ADDRESS: a3a3a3db
FAULTING_IP:
ATMFD+345c9
91f445c9 83483810 or dword ptr [eax+38h],10h
MM_INTERNAL_CODE: 2
IMAGE_NAME: ATMFD.DLL
DEBUG_FLR_IMAGE_TIMESTAMP: 54e6a55a
MODULE_NAME: ATMFD
FAULTING_MODULE: 91f10000 ATMFD
DEFAULT_BUCKET_ID: WIN7_DRIVER_FAULT
BUGCHECK_STR: 0x50
PROCESS_NAME: csrss.exe
CURRENT_IRQL: 2
ANALYSIS_VERSION: 6.3.9600.17237 (debuggers(dbg).140716-0327) x86fre
TRAP_FRAME: 879cfd8c -- (.trap 0xffffffff879cfd8c)
ErrCode = 00000002
eax=a3a3a3a3 ebx=00000008 ecx=00000004 edx=fb964900 esi=fb80e380 edi=fb80e3a0
eip=91f445c9 esp=879cfe00 ebp=879cfe10 iopl=0 nv up ei ng nz ac pe cy
cs=0008 ss=0010 ds=0023 es=0023 fs=0030 gs=0000 efl=00010297
ATMFD+0x345c9:
91f445c9 83483810 or dword ptr [eax+38h],10h ds:0023:a3a3a3db=????????
Resetting default scope
LAST_CONTROL_TRANSFER: from 826f4ce7 to 826902d8
STACK_TEXT:
879cf8dc 826f4ce7 00000003 5e64199c 00000065 nt!RtlpBreakWithStatusInstruction
879cf92c 826f57e5 00000003 c06028e8 a3a3a3db nt!KiBugCheckDebugBreak+0x1c
879cfcf0 826a3391 00000050 a3a3a3db 00000001 nt!KeBugCheck2+0x68b
879cfd74 82655c48 00000001 a3a3a3db 00000000 nt!MmAccessFault+0x104
879cfd74 91f445c9 00000001 a3a3a3db 00000000 nt!KiTrap0E+0xdc
WARNING: Stack unwind information not available. Following frames may be wrong.
879cfe10 91f39eec fb964900 ff657038 91f563ec ATMFD+0x345c9
879cfe34 91f3e987 fb9bec70 91f563ec 00000f5c ATMFD+0x29eec
879d0544 91f3f6e0 fb9bec70 91f4f028 879d0790 ATMFD+0x2e987
879d0600 91f327ae fb9bec70 91f4f028 879d0790 ATMFD+0x2f6e0
879d06ec 91f32858 fb9bec70 879d0790 879d0814 ATMFD+0x227ae
879d0718 91f232b2 fb9bec70 91f4f028 879d0790 ATMFD+0x22858
879d087c 91f23689 ffffffff 879d099c fb874f58 ATMFD+0x132b2
879d08d0 91f1406d ffffffff 879d099c 00000000 ATMFD+0x13689
879d0924 91c7dcf2 ff7a5010 fbeeccf0 00000001 ATMFD+0x406d
879d096c 91c667cb ff7a5010 fbeeccf0 00000001 win32k!PDEVOBJ::QueryFontData+0x3e
879d09e0 91c91513 ffa6a130 fb81e8d0 0000004f win32k!xInsertMetricsRFONTOBJ+0x9c
879d0a14 91c935f5 00000020 879d0b2c 879d0c92 win32k!RFONTOBJ::bGetGlyphMetrics+0x131
879d0cb8 91ca6684 040101b7 00000060 00000040 win32k!GreGetCharABCWidthsW+0x147
879d0d14 82652a66 040101b7 00000040 00000040 win32k!NtGdiGetCharABCWidthsW+0xf8
879d0d14 771870f4 040101b7 00000040 00000040 nt!KiSystemServicePostCall
0012f1d4 00000000 00000000 00000000 00000000 ntdll!KiFastSystemCallRet
---
The crash occurs while trying to modify memory under an invalid address. More specifically, the "a3a3a3a3" bytes found in the EAX register at the time of the crash are a repetition of a random byte filled by Driver Verifier in each pool allocation before returning it to the caller. This means that the dereferenced pointer is in fact an uninitialized value from the kernel pool.
The offending code (containing the crashing instruction) is as follows:
---
.text:000445C6 loc_445C6:
.text:000445C6 mov eax, [esi+ecx*4]
.text:000445C9 or dword ptr [eax+38h], 10h
.text:000445CD inc ecx
.text:000445CE cmp ecx, ebx
.text:000445D0 jl short loc_445C6
---
As shown above, there is a loop iterating EBX times over an array of addresses stored in [ESI]. At the time of the bugcheck, EBX=8, so the code expects 8 valid pointers in the array; however, only 4 pointers are properly initialized:
---
kd> dd fb80e380
fb80e380 fb964980 fb9649c0 fb964900 fb964940
fb80e390 a3a3a3a3 a3a3a3a3 a3a3a3a3 a3a3a3a3
fb80e3a0 a3a3a3a3 a3a3a3a3 a3a3a3a3 a3a3a3a3
---
The issue reproduces on Windows 7. It is easiest to reproduce with Special Pools enabled for ATMFD.DLL (leading to an immediate crash when the bug is triggered), but it should also be possible to observe a crash on a default Windows installation in ATMFD.DLL.
Attached is an archive with two proof of concept font files together with corresponding kernel crash logs.
Proof of Concept:
https://gitlab.com/exploit-database/exploitdb-bin-sploits/-/raw/main/bin-sploits/37920.zip
