AIKIDO-2026-955281

TL;DR

tinytag reads audio metadata by iterating over the atoms inside M4A and MP4 containers. When a freeform ---- atom contains a sub-atom whose declared size is zero, the MP4 parser fails to advance past that sub-atom and spins in an endless loop. Reading such a crafted file therefore never returns and keeps the process or worker performing metadata extraction busy until it is killed. The fix reuses the primary atom parsing code path, which enforces proper size checks so zero-sized sub-atoms can no longer stall parsing.