`WaveFormat::parse` reads past the end of a `&WAVEFORMATEX`
Historical advisory saved 24 Oct 2026 13:19:54 UTC. The publisher may have updated it since then. Latest advisory.
osv · RUSTSEC-2026-0332
Published`WaveFormat::parse` is a safe function that takes a `&WAVEFORMATEX`, which is only valid for reads of `size_of::<WAVEFORMATEX>()` (18) bytes. When the header has `wFormatTag == WAVE_FORMAT_EXTENSIBLE` and `cbSize >= 22`, the function reinterprets the reference as a `WAVEFORMATEXTENSIBLE` and reads 40 bytes. Nothing in the signature guarantees that the header sits at the start of a larger buffer, so safe code can trigger an out-of-bounds read: ```rust use wasapi::WaveFormat; use windows::Win32::Media::Audio::WAVEFORMATEX; use windows::Win32::Media::KernelStreaming::WAVE_FORMAT_EXTENSIBLE; let header = WAVEFORMATEX { wFormatTag: WAVE_FORMAT_EXTENSIBLE as u16, nChannels: 2, nSamplesPerSec: 48000, nAvgBytesPerSec: 384000, nBlockAlign: 8, wBitsPerSample: 32, cbSize: 22, }; // Reads 22 bytes past the end of `header`. let _ = WaveFormat::parse(&header); ``` A more likely way to hit this is to copy the header out of a format pointer returned by WASAPI (`let fmt = unsafe { *ptr };`) and pass `&fmt`. The copy keeps `cbSize`, but not the 22 bytes that follow it. The bytes read past the end are returned as the channel mask and subformat of the parsed format. The flaw was corrected in commit `2562db7`, released in 0.25.0. `parse` is now an `unsafe fn` that takes a `*const WAVEFORMATEX`, and the caller must guarantee that the pointer is valid for reads of `size_of::<WAVEFORMATEX>() + cbSize` bytes. `WaveFormat::parse_from_blob_bytes` (available since 0.23.0) is the safe alternative and checks the slice length against `cbSize`.
- Source published
- 8 Sep 2026 12:00:00 UTC
- Source modified
- 7 Oct 2026 14:45:03 UTC
- Saved here
- 24 Oct 2026 13:19:54 UTC
- Source terms
- OSV · upstream source license applies
Original source record · Cite this exact observation · Complete original publisher JSON
Source: osv · OSV · upstream source license applies. The publisher's wording is preserved. ChangeIntel groups the information and labels changes. The JSON record includes additional source fields.