[BUG] Scheduling while atomic in apple_bce (aaudio_handle_stream_timestamp → mutex_lock) causing random reboots on Ubuntu 24.04 (MacBookPro16,1) #179
Description
[BUG] Scheduling while atomic in apple_bce (aaudio_handle_stream_timestamp → mutex_lock) causing random reboots on Ubuntu 24.04 (MacBookPro16,1)
Bug Report: Kernel BUG "scheduling while atomic" in apple_bce aaudio DMA IRQ handler causing random hard reboots
Hardware:
• Model: MacBook Pro 16" 2019 (MacBookPro16,1) with Intel Core i9
• BIOS/iBridge: 2094.40.1.0.0.0 (iBridge: 23.16.12048.0.0,0) dated 11/08/2025
OS / Distribution / Kernel:
• Ubuntu 24.04 LTS (noble)
• Kernel: 6.19.5-1-t2-noble (from https://github.com/AdityaGarg8/t2-ubuntu-repo)
• Also reproduced on 6.18.2-1-t2-noble
• Tainted: G WC (WARN + out-of-tree module apple_bce)
Symptom:
• Random hard reboots / power cycles after long uptime (e.g., 7+ days stable, then sudden reboot with no panic screen or obvious trigger).
• Occurs during idle/light use; no heavy load, specific apps, or audio playback required.
• Kernel logs repeatedly show "BUG: scheduling while atomic" in the BCE DMA IRQ thread (irq/88-bce_dma or irq/98-bce_dma), likely leading to instability/watchdog/panic.
Key log excerpts (from journalctl -k):
First trace (with mutex details):
BUG: scheduling while atomic: irq/88-bce_dma/529/0x00000002
...
Call Trace:
...
__mutex_lock_slowpath+0x13/0x20
mutex_lock+0x3b/0x50
_snd_pcm_stream_lock_irqsave+0x36/0x60 [snd_pcm]
aaudio_handle_stream_timestamp+0x4f/0xf0 [apple_bce]
aaudio_handle_timestamp+0x2d/0x59 [apple_bce]
aaudio_handle_cmd_timestamp+0xd3/0x190 [apple_bce]
aaudio_handle_command+0x83/0x90 [apple_bce]
aaudio_bce_in_queue_completion+0x22f/0x2e0 [apple_bce]
bce_handle_cq_completions+0x19a/0x230 [apple_bce]
bce_handle_dma_irq+0x7e/0xe0 [apple_bce]
...
Followed by:
aaudio_pcm_pointer while not started
BUG: scheduling while atomic: irq/88-bce_dma/529/0x00000000
Mitigation / Workarounds Tried:
• Kernel upgrade: 6.18.2-t2 → 6.19.5-t2-noble → same issue persists.
• GRUB cmdline (per wiki): quiet splash intel_iommu=on iommu=pt pcie_ports=compat
• Also tested with added threadirqs → no improvement.
• Blacklisting apple_bce (via /etc/modprobe.d/blacklist.conf + update-initramfs -u -k all) → completely eliminates the BUG messages and reboots (but disables internal audio and Touch Bar).
• Audio stack: PipeWire (default on Ubuntu 24.04). Happens even with audio muted/idle; likely background timestamp/polling trigger.
• No hardware issues seen in macOS (if previously dual-booted); SSD health fine via smartctl.
Additional Info:
• No other oops/panics in logs.
• Seems tied to aaudio timestamp handling attempting mutex in atomic/IRQ context (e.g., _snd_pcm_stream_lock_irqsave from aaudio_handle_stream_timestamp when stream not started).
• Not widely reported (no matching issues found in apple-bce-drv or related repos).
Request / Questions:
• Is this a known issue in apple_bce / aaudio subsystem?
• Any quick workarounds besides blacklisting the module (e.g., config to disable timestamp queries, or defer locking)?
• Would a code change help (e.g., use spinlock, move lock to non-atomic path, skip if !started, or workqueue deferral)?
• Happy to test patches, debug branches, or provide more info (full dmesg, lsmod, pactl list, etc.).
Thanks for maintaining t2linux — this setup is otherwise very stable!