# Correction + update: this is an Ubuntu resolute (26.04) kernel *build* regression, not an upstream -31 regression
When I filed this bug (2026-09-08, via apport) I attributed it to an upstream regression between 7.0.0-28 and 7.0.0-31. Further bisection shows that attribution was wrong. The failure tracks the **Ubuntu build**, not the upstream version number, and it is still present in the latest resolute kernels. `uname -r` cannot distinguish the builds — the dpkg version suffix (`~24.04.1` = noble-hwe build) is what matters. ## Evidence matrix (version.build = dpkg version) | Kernel build | Userspace | Result | |---|---|---| | 7.0.0-30.30~24.04.1 (noble-hwe) | 24.04 | GOOD (jack + dmic verified) | | 7.0.0-30.30 (resolute, SAME upstream version) | 26.04 GA | BAD: headphone jack silent, dmic full-scale clipping | | 7.0.0-31.31 (resolute) | 26.04 GA | BAD: jack silent, speaker OK | | 7.0.0-31.31 (resolute) | 26.04.1 | GOOD: same kernel binary that failed on 26.04 GA worked after 26.04.1 updates — implicates 26.04 GA *userspace* for that phase | | 7.0.0-38.38 (resolute) | 26.04.1 | BAD: dmic captures one initial pop (peak ~28006) then pure digital silence — a NEW signature, distinct from the earlier saturated-buzz/clipping pattern | | 7.0.0-42.42 (resolute) | 26.04.1 | BAD: dmic captures pure digital silence (peak = 0 in every 100 ms window), verified both via PipeWire and direct `arecord -D hw:CARD=sofsoundwire,DEV=4 -f S16_LE -r 48000 -c 2` | The genuine upstream -28 → -30 fix (likely upstream 7.0.10's `soundwire: cadence: Clear message complete before signaling waiting thread`) is a separate, already-resolved issue. ## Hardware / stack (unchanged since filing) Dell Precision 5690, Meteor Lake, sof-soundwire machine driver, rt713 (jack + PDM dmic, ALSA capture device 4) + rt1316 + rt1713, topology `sof-mtl-rt713-l0-rt1316-l12-rt1713-l3.tplg`, SOF firmware 2.14.1.1. ## What has been ruled out (on the bad builds) - SOF firmware and topology are byte-identical between good and bad states. - alsa-ucm-conf 1.2.10 vs 1.2.15.3 makes no difference; all DMIC kcontrols are correct (FU1E on, ADC muxes DMIC1/DMIC2), PipeWire source unmuted. - dmesg shows ZERO SoundWire/codec errors on the bad builds: firmware boots, topology loads, all codecs enumerate. The stream runs and delivers zeros. - Full DSP/codec re-init (`modprobe -r snd_sof_pci_intel_mtl && modprobe snd_sof_pci_intel_mtl`) does not help. - Hardware is proven good (works in Windows, and under the noble-hwe build on this same machine). ## Current state - 7.0.0-38.38 and 7.0.0-42.42 (both resolute) both have the dead-dmic problem. - The held noble-hwe 7.0.0-30.30~24.04.1 build remains my known-good fallback (at the cost of losing the IPU6 camera, since linux-modules-ipu6 was never published for -30). - Happy to test patches or provide traces (SoundWire debugfs, SOF IPC logs, `alsabat`-style loopback) on request. -- You received this bug notification because you are a member of Ubuntu Bugs, which is subscribed to Ubuntu. https://bugs.launchpad.net/bugs/2166754 Title: [regression] sof-soundwire rt713: headphone jack silent on 7.0.0-31 (worked on 7.0.0-30, broken on 7.0.0-28) — Dell Precision 5690 / Meteor Lake To manage notifications about this bug go to: https://bugs.launchpad.net/ubuntu/+source/linux/+bug/2166754/+subscriptions -- ubuntu-bugs mailing list [email protected] https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs
