openexr
Affected
- introduced
3.3.2, fixed3.3.3(ecosystem)
Fixed versions
3.3.3
Fix
- Fixed version
3.3.3pip install --upgrade openexr==3.3.3
Advisory
Also known as: CVE-2025-48073, GHSA-qhpm-86v7-phmm
OpenEXR ScanLineProcess::run_fill NULL Pointer Write In "reduceMemory" Mode
Summary
When reading a deep scanline image with a large sample count in reduceMemory mode, it is possible to crash a target application with a NULL pointer dereference in a write operation.
Details
In the ScanLineProcess::run_fill function, implemented in src/lib/OpenEXR/ImfDeepScanLineInputFile.cpp, the following code is used to write the fillValue in the sample buffer:
switch (fills.type)
{
case OPENEXR_IMF_INTERNAL_NAMESPACE::UINT:
{
unsigned int fillVal = (unsigned int) (fills.fillValue);
unsigned int* fillptr = static_cast<unsigned int*> (dest);
for ( int32_t s = 0; s < samps; ++s )
fillptr[s] = fillVal; // <--- POTENTIAL CRASH HERE
break;
}
However, when reduceMemory mode is enabled in the readDeepScanLine function in src/lib/OpenEXRUtil/ImfCheckFile.cpp, with large sample counts, the sample data will not be read, as shown below:
// limit total number of samples read in reduceMemory mode
//
if (!reduceMemory ||
fileBufferSize + bufferSize < gMaxBytesPerDeepScanline) // <--- CHECK ON LARGE SAMPLE COUNTS AND reduceMemory
{
// SNIP...
try
{
in.readPixels (y);
}
Therefore, in those cases, the sample buffer would not be allocated, resulting in a potential write operation on a NULL pointer.
PoC
NOTE: please download the runfill_crash.exr file from the following link:
https://github.com/ShielderSec/poc/tree/main/CVE-2025-48073
exrcheck binary in a macOS or GNU/Linux machine with ASAN.runfill_crash.exr file with the following command:exrcheck -m runfill_crash.exr
exrcheck crashes with ASAN stack-trace.Impact
An attacker may cause a denial of service by crashing the application.
Not rated by the source.
3.3.2, fixed 3.3.3 (ecosystem)3.3.3
3.3.3pip install --upgrade openexr==3.3.3Fix commands are the export's own generic commands. Ojuju never runs them. Check them against your setup first.
Check your own computer: guard packages check
Source: PyPA Advisory Database · CC BY 4.0