[BUG] [CML] Infinite ipc tx error loop after suspend/resume floods journal #10442

Describe the bug
After suspend/resume, the SOF driver enters an infinite error loop when PipeWire attempts to use the audio device. The driver repeatedly fails with -22 (EINVAL) on sof_pcm_prepare, flooding the journal with thousands of errors per second. The system does not recover automatically.

I diagnosed this by checking journalctl -r which showed the error spam. The errors occur at a rate high enough that systemd-journald reports "Missed N kernel messages".

To Reproduce

1. Have PipeWire running with an active audio client (e.g., browser, Discord)
2. Suspend the system (S3)
3. Resume
4. Observe error flood in journalctl -f

Reproduction Rate
Unable to determine exact rate as suspend/resume normally works. This occurred once so far, but when it does occur, the error loop is 100% consistent until PipeWire is restarted.

Expected behavior
The driver should either:

• Successfully reinitialize after resume, or
• Fail gracefully with rate-limited errors, allowing userspace to recover without manual intervention

Impact
Annoyance, but significant when it occurs:

• High CPU usage from systemd-journald processing the flood
• Journal grew by ~4GB in minutes before I noticed
• Requires manual restart of PipeWire to recover

Environment

1. Branch name and commit hash:
   • Kernel: 6.17.9-arch1-1 (Arch Linux mainline kernel)
   • SOF: sof-firmware 2025.05.1-1 (Arch Linux package)
2. Name of the topology file
   • Topology: (default, using skl_hda_dsp_generic as shown in logs)
3. Name of the platform(s)
   • Platform: Comet Lake PCH-LP cAVS (device 0000:00:1f.3)
   • Driver: snd_sof_pci_intel_cnl
   • Audio server: PipeWire 1.4.9

Screenshots or console output
Error log from journalctl (repeats hundreds of times per second):

sof-audio-pci-intel-cnl 0000:00:1f.3: ipc tx error for 0x60010000 (msg/reply size: 108/20): -22
sof-audio-pci-intel-cnl 0000:00:1f.3: sof_ipc3_pcm_hw_params: pcm0 (HDA Analog), dir 1: STREAM_PCM_PARAMS ipc failed for stream_tag 2
sof-audio-pci-intel-cnl 0000:00:1f.3: sof_pcm_prepare: pcm0 (HDA Analog), dir 1: failed to set hw_params after resume
sof-audio-pci-intel-cnl 0000:00:1f.3: ASoC error (-22): at snd_soc_pcm_component_prepare() on 0000:00:1f.3
systemd-journald[284]: Missed 2 kernel messages

Workaround - restarting PipeWire clears the error state:

systemctl --user restart pipewire pipewire-pulse wireplumber

Additional context
The bug appears to be in the driver's resume path failing to restore DSP state properly. PipeWire repeatedly retries the broken device, triggering the kernel error on each attempt. There is no rate limiting on the kernel side, causing the journal flood.

I do not have sof-logger output as the issue has been worked around and I cannot reliably reproduce it.