We have encountered a Windows kernel crash in the ATMFD.DLL OpenType driver while processing a corrupted OTF font file, see below:
N bytes of memory was allocated and more than N bytes are being referenced.
This cannot be protected by try-except.
When possible, the guilty driver's name (Unicode string) is printed on
the bugcheck screen and saved in KiBugCheckDriver.
Arg1: fb69b01e, memory referenced
Arg2: 00000000, value 0 = read operation, 1 = write operation
Arg3: 8f635862, if non-zero, the address which referenced memory.
Arg4: 00000000, (reserved)
8f635862 803802 cmp byte ptr [eax],2
CPU_MICROCODE: 6,3e,4,0 (F,M,S,R) SIG: 19'00000000 (cache) 19'00000000 (init)
ANALYSIS_SESSION_TIME: 03-21-2017 10:49:20.0375
ANALYSIS_VERSION: 10.0.10586.567 x86fre
LAST_CONTROL_TRANSFER: from 8f636088 to 8f635862
WARNING: Stack unwind information not available. Following frames may be wrong.
9625f538 8f636088 9625f790 05f70000 9625f790 ATMFD+0x35862
9625f630 8f6355b1 9625f790 05f70000 9625f64c ATMFD+0x36088
9625f734 8f635711 9625f790 05f70000 9625f868 ATMFD+0x355b1
9625f768 8f6051b0 00000000 9625f790 05f70000 ATMFD+0x35711
9625f808 8f607279 00000004 00000001 00000002 ATMFD+0x51b0
9625f888 8f603d14 00000000 00000000 94bb3200 ATMFD+0x7279
9625f96c 8f6e7b8d 00000004 fbad2fc0 fbadaff8 ATMFD+0x3d14
9625f9b4 8f6e7adf 00000001 fbad2fc0 fbadaff8 win32k!PDEVOBJ::LoadFontFile+0x3c
9625f9f4 8f6e74fc ffa6a130 0000002e fbad2fc0 win32k!vLoadFontFileView+0x291
9625fa80 8f6d6403 9625fb58 0000002e 00000001 win32k!PUBLIC_PFTOBJ::bLoadFonts+0x209
9625facc 8f6d73d8 9625fb58 0000002e 00000001 win32k!GreAddFontResourceWInternal+0xfb
9625fc14 8164ddb6 000d9b78 0000002e 00000001 win32k!NtGdiAddFontResourceW+0x142
9625fc14 77ad6c74 000d9b78 0000002e 00000001 nt!KiSystemServicePostCall
The bugcheck is caused by an attempt to read memory from an unmapped address. The specific expression being dereferenced by ATMFD.DLL is "base address of the Name INDEX data + NAME.offset[x] - 1", however no bounds checking is performed over the value of NAME.offset[x] before using it for pointer arithmetic. To our current knowledge, this condition can only lead to an out-of-bounds read, thus limiting the impact of the bug to remote denial of service, or potentially local kernel memory disclosure. However, we have not fully confirmed that the severity of the bug is not in fact more significant due to some further ATMFD logic we are not aware of.
Interestingly, the crash is almost identical to the one reported in Issue #386 (MSRC-30296) nearly two years ago, which was supposedly fixed as CVE-2015-2461 in the MS15-080 bulletin. The fact that the same bugcheck still reproduces can potentially mean that the patch was insufficient.
Only a single bitflip applied to a valid font file is sufficient to create an offending testcase (excluding SFNT table checksums). In our case, the byte at offset 0x375 in the original sample must be changed from 0x01 to 0x41. This corresponds to offset 0x71 of the "CFF " table. The PoC font can be found attached to this tracker entry.
The issue reproduces on Windows 7 (other platforms untested). It is easiest to reproduce with Special Pools enabled for ATMFD.DLL, leading to an immediate crash when the bug is triggered. The bugcheck occurs upon opening the font in any default utility such as the Windows Font Viewer -- no special tools are required.
Proof of Concept: