Package: src:linux Version: 6.12.115-2 Severity: important Tags: regression
Dear Maintainer, after upgrading the kernel from 6.12.113+deb13-amd64 (6.12.113-1) to 6.12.115+deb13.1-amd64 (6.12.115-2) on Debian 13 (trixie), a MOTU M4 USB audio interface (lsusb 07fd:000b) produces no sound at all. Booting the previous kernel 6.12.113 restores audio; the regression is fully reproducible across both kernels. Symptom: any playback routed to the MOTU M4 hangs. Application frames are accepted but the device never clocks. The ALSA PCM status shows state RUNNING, hw_ptr 0, tstamp 0, while appl_ptr advances, i.e. the hardware pointer never advances. Confirmed down to raw ALSA (aplay -D plughw with PipeWire stopped and the device free), so this is not a PipeWire/userspace issue. The internal Intel HDA output plays fine on the same 6.12.115 kernel. PipeWire restart and USB re-enumeration do not help. Regression range: works on 6.12.113+deb13-amd64, broken on 6.12.115+deb13.1-amd64. The only sound/usb change between these two Debian point releases is the backport of the CVE-2026-23208 fix "ALSA: usb-audio: Prevent excessive number of frames" (sound/usb/pcm.c), which clamps the number of frames per data URB. This clamp appears to over-restrict the MOTU M4 playback path (implicit feedback), so no valid audio URBs are submitted and the stream never starts clocking. Hardware: Intel i7-12700KF, MOTU M4 on a SuperSpeed USB port. Workaround in place: booting 6.12.113 (audio works). Could the frame-count check from the CVE-2026-23208 fix be revisited so it no longer breaks legitimate implicit-feedback USB audio devices like the MOTU M4? I can provide dmesg, alsa-info output, or test patched kernels. Thank you.

