Public bug reported:
Problem description
===================
On Ubuntu 26.04.1, PipeWire's real-time scheduling behavior appears
inconsistent after login. During startup, PipeWire's module-rt debug log says
that RTKit granted realtime priority 20 to the audio data-loop threads. A
later check of those same thread IDs with chrt shows SCHED_OTHER with priority
0. Restarting the user audio services then results in SCHED_RR priority 20,
confirmed by both chrt and RTKit's journal. The available logs do not show
when or why the threads from the original startup changed back.
This affects the data-loop threads of pipewire, pipewire-pulse, and
wireplumber. This difference was reproduced in one session: the threads were
SCHED_OTHER before a manual service restart and SCHED_RR after it. The
realtime policy was still active 20 seconds after the restart.
System information
==================
- Ubuntu 26.04.1
- Kernel: 7.0.0-43
- PipeWire: 1.6.2-1ubuntu1.2
- WirePlumber: 0.5.13-1ubuntu1.1
- RTKit: 0.14-1
- Bluetooth adapter: Intel AX200
Startup log evidence (18 November 2026)
=======================================
PipeWire and PipeWire-Pulse started at approximately 14:06:58.086. RTKit was
reported started at 14:06:58.147. PipeWire's debug log then reported:
14:06:58.151925 pipewire[2211]: acquired realtime priority 20 for thread
2230 using RTKit
14:06:58.152533 pipewire-pulse[2213]: acquired realtime priority 20 for
thread 2228 using RTKit
14:06:58.223953 wireplumber[2217]: acquired realtime priority 20 for thread
2262 using RTKit
The same module logs show that each thread initially used
SCHED_OTHER|SCHED_RESET_ON_FORK before the RTKit request. A later check with
`chrt -p` showed TIDs 2230, 2228, and 2262 as SCHED_OTHER, priority 0. The
debug logs do not show a transition back to SCHED_OTHER. RTKit's journal for
this boot did not list promotions for these three TIDs, although it did list
successful promotions for other processes owned by UID 1000.
Manual service restart comparison (18 November 2026, 14:54)
===========================================================
Immediately before restart, the original data-loop TIDs 2230, 2228, and 2262
were all SCHED_OTHER. After restarting pipewire, pipewire-pulse, and
wireplumber, the new data-loop TIDs were SCHED_RR priority 20:
- pipewire PID 12679, data-loop TID 12689
- pipewire-pulse PID 12681, data-loop TID 12691
- wireplumber PID 12682, data-loop TID 12701
`chrt -p` confirmed SCHED_RR|SCHED_RESET_ON_FORK priority 20 for all three,
both immediately and 20 seconds after restart. RTKit's system journal
explicitly recorded successful grants for those exact TIDs at 14:54:50.110,
.113, and .123. The services remained active. This makes the behavior
reproducible across a boot-started instance and a manually restarted one.
Expected behavior
=================
If the module reports that RTKit granted realtime scheduling, the audio data
loop should remain at the granted realtime policy unless a documented limit,
revocation, or later policy change explains otherwise. Restarting the same
services in the same user session should not be necessary to obtain the
expected policy.
Actual behavior
===============
After login, PipeWire reports acquiring realtime priority, but a later
scheduler query shows the same threads running as SCHED_OTHER. Restarting the
user services produces stable SCHED_RR threads and corresponding RTKit grant
messages. It is not yet known whether the initial policy was applied and then
reverted, or whether another mechanism explains the discrepancy.
Relevant configuration
======================
The stock PipeWire configuration loads `libpipewire-module-rt` with
`rt.prio = 88`; PipeWire-Pulse and WirePlumber also load the realtime module.
The user services have `LimitRTPRIO=0`, so they rely on RTKit rather than
direct realtime privileges. No persistent service or privilege changes were
made for this investigation. Temporary systemd user drop-ins enabled debug
logging for this boot and were removed after collecting the logs.
Logs to attach
==============
We may send the complete journals log if deemed necessary.
Question
========
Can PipeWire, RTKit, or another system component explain why module-rt logs a
successful realtime grant while the same data-loop threads are later observed
as SCHED_OTHER? Is there a supported way to log a later policy change or
revocation for these threads?
ProblemType: Bug
DistroRelease: Ubuntu 26.04
Package: pipewire 1.6.2-1ubuntu1.2
ProcVersionSignature: Ubuntu 7.0.0-43.43-generic 7.0.14
Uname: Linux 7.0.0-43-generic x86_64
ApportVersion: 2.34.1-0ubuntu0.1
Architecture: amd64
AudioDevicesInUse:
USER PID ACCESS COMMAND
/dev/snd/controlC0: andrea 2217 F.... wireplumber
/dev/snd/seq: andrea 2211 F.... pipewire
CasperMD5CheckResult: unknown
CurrentDesktop: MATE
Date: Wed Nov 18 14:52:37 2026
MachineType: Dell Inc. XPS 13 7390
PulseList: Error: command ['pacmd', 'list'] failed with exit code 1: No
PulseAudio daemon running, or not running as session daemon.
SourcePackage: pipewire
UpgradeStatus: Upgraded to resolute on 2026-09-30 (49 days ago)
dmi.bios.date: 12/17/2024
dmi.bios.release: 1.29
dmi.bios.vendor: Dell Inc.
dmi.bios.version: 1.29.0
dmi.board.name: 0G2D0W
dmi.board.vendor: Dell Inc.
dmi.board.version: A00
dmi.chassis.type: 10
dmi.chassis.vendor: Dell Inc.
dmi.modalias:
dmi:bvnDellInc.:bvr1.29.0:bd12/17/2024:br1.29:svnDellInc.:pnXPS137390:pvr:rvnDellInc.:rn0G2D0W:rvrA00:cvnDellInc.:ct10:cvr:sku0962:pfaXPS:
dmi.product.family: XPS
dmi.product.name: XPS 13 7390
dmi.product.sku: 0962
dmi.sys.vendor: Dell Inc.
** Affects: pipewire (Ubuntu)
Importance: Undecided
Status: New
** Tags: amd64 apport-bug resolute
--
You received this bug notification because you are a member of Ubuntu
Bugs, which is subscribed to Ubuntu.
https://bugs.launchpad.net/bugs/2174841
Title:
PipeWire audio threads do not remain in realtime scheduling after
login
To manage notifications about this bug go to:
https://bugs.launchpad.net/ubuntu/+source/pipewire/+bug/2174841/+subscriptions
--
ubuntu-bugs mailing list
[email protected]
https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs