An attacker with high physical access can exploit this vulnerability to read the contents of stack memory or global memory. This information could help with explotation of other vulnerabilities in DXE to elevate privileges from ring 3 or ring 0 (depends on the operating system) to a DXE driver and execute arbitrary code. Malicious code installed as a result of this exploitation could survive operating system (OS) boot process and runtime, or modify NVRAM area on the SPI flash storage (to gain persistence). Additionally, threat actors could use this vulnerability to bypass OS security mechanisms (modify privileged memory or runtime variables), influence OS boot process, and in some cases allow an attacker to hook or modify EFI Runtime services.
Binarly REsearch Team has discovered a memory contents leak / information disclosure vulnerability that allows a potential attacker to dump stack memory or global memory into an NVRAM variable. This in turn could help building a successful attack vector based on exploiting a memory corruption vulnerability.
An attacker with high physical access can exploit this vulnerability to read the contents of stack memory or global memory. This information could help with explotation of other vulnerabilities in DXE to elevate privileges from ring 3 or ring 0 (depends on the operating system) to a DXE driver and execute arbitrary code. Malicious code installed as a result of this exploitation could survive operating system (OS) boot process and runtime, or modify NVRAM area on the SPI flash storage (to gain persistence). Additionally, threat actors could use this vulnerability to bypass OS security mechanisms (modify privileged memory or runtime variables), influence OS boot process, and in some cases allow an attacker to hook or modify EFI Runtime services.
Let's take XPS 17 9710, Precision 5760's firmware (version: 0.1.14.0, module sha256: 4b32444ad888c2597de681707449065e75e4b620aa50c2f2e875eca97b875d73) as an example.
The following code in the module actually allows leaking memory:
gRT->GetVariable()
offset: 0x103d
gRT->SetVariable()
offset: 0x1074
__int64 __fastcall sub_E74()
{
char v0; // bl
__int64 result; // rax
unsigned __int8 v2; // al
__int64 (__fastcall **v3)(__int64); // rax
bool v4; // zf
int v5; // eax
UINTN v6; // [rsp+30h] [rbp-D0h] BYREF
UINTN DataSize; // [rsp+38h] [rbp-C8h] BYREF
char Data[91]; // [rsp+40h] [rbp-C0h] BYREF
int v9; // [rsp+9Bh] [rbp-65h]
int v10; // [rsp+9Fh] [rbp-61h]
int v11; // [rsp+A9h] [rbp-57h]
char v12[2048]; // [rsp+2D0h] [rbp+1D0h] BYREF
UINT32 Attributes; // [rsp+AE0h] [rbp+9E0h] BYREF
UINT32 v14; // [rsp+AE8h] [rbp+9E8h] BYREF
void *Interface; // [rsp+AF0h] [rbp+9F0h] BYREF
void *v16; // [rsp+AF8h] [rbp+9F8h] BYREF
Interface = 0i64;
v16 = 0i64;
v0 = 0;
result = gBS->LocateProtocol(&PLATFORM_NVS_AREA_PROTOCOL_GUID, 0i64, &Interface);
if ( result >= 0 )
{
if ( (gBS->LocateProtocol(&UNKNOWN_PROTOCOL_GUID_0, 0i64, &v16) & 0x8000000000000000ui64) == 0i64 )
{
while ( 1 )
{
(*((void (__fastcall **)(void *, __int64, __int64))v16 + 3))(v16, 166i64, 1i64);
__outbyte(0x910u, 0x11u);
v2 = __inbyte(0x911u);
if ( ((v2 + 1) & 0xFE) != 0 )
break;
if ( (unsigned __int8)++v0 >= 3u )
goto LABEL_7;
}
switch ( v2 )
{
case 1u:
LABEL_7:
*(_BYTE *)(*(_QWORD *)Interface + 2079i64) = 0;
break;
case 2u:
*(_BYTE *)(*(_QWORD *)Interface + 2079i64) = 1;
break;
case 4u:
*(_BYTE *)(*(_QWORD *)Interface + 2079i64) = 2;
break;
case 8u:
*(_BYTE *)(*(_QWORD *)Interface + 2079i64) = 3;
break;
}
}
DataSize = 654i64;
result = gRT->GetVariable((CHAR16 *)L"CpuSetup", &CPU_SETUP_VARIABLE_GUID, &Attributes, &DataSize, Data);
if ( result >= 0 )
{
v3 = (__int64 (__fastcall **)(__int64))sub_1824();
v4 = (v3[1](306i64) & 1) == 0;
v5 = 100000;
if ( v4 )
v5 = 109000;
v11 = v5;
v10 = v5;
v9 = v5;
gRT->SetVariable((CHAR16 *)L"CpuSetup", &CPU_SETUP_VARIABLE_GUID, Attributes, 0x28Eui64, Data);
v6 = 2028i64;
gRT->GetVariable( // <= first call (we can rewrite DataSize here)
(CHAR16 *)L"PchSetup",
&PCH_SETUP_VARIABLE_GUID,
&v14,
&v6,
v12);
v12[1712] = 0;
return gRT->SetVariable( // <= second call
(CHAR16 *)L"PchSetup",
&PCH_SETUP_VARIABLE_GUID,
v14,
v6,
v12);
}
}
return result;
}
The gRT->SetVariable()
service is called with the DataSize
as an argument, which will be overwritten inside the gRT->GetVariable()
service if the length of PchSetup
NVRAM variable is greater than 2028
.
Thus, a potential attacker can dump X - 2028
bytes from the stack (or global memory) into PchSetup NVRAM variable by setting PchSetup
NVRAM variable's size to X > 2028
.
To fix this vulnerability the DataSize
must be re-initialized with the size of PchSetup
before calling gRT->SetVariable()
.
This bug is subject to a 90 day disclosure deadline. After 90 days elapsed or a patch has been made broadly available (whichever is earlier), the bug report will become visible to the public.
Binarly REsearch Team