Kernel/user partition: remove the SOF DMA trace - #11421

Part of #11306 (kernel / user-space partition): remove the SOF DMA trace (src/trace, CONFIG_TRACE). It was the last kernel-only directory outside src/kernel, and nothing builds it any more except the host builds, which only used it to print traces. Follows #11420 (merged).

Patches

1. Move the SOF probe log backend options to src/kernel/logging. The CONFIG_LOG_BACKEND_SOF_PROBE* options were defined in src/trace/Kconfig. They move to a Kconfig file next to the backend; the values are unchanged.

2. Print host build traces without CONFIG_TRACE.

   • The testbench and host library printed traces through the CONFIG_LIBRARY part of trace-soflogger.h, which needed CONFIG_TRACE.
   • That printing moves to a header of its own, trace-host.h, with the same levels and output as before.
   • The host library now builds with CONFIG_TRACE=n. Its code is unchanged; only the trace contexts move out of .trace_ctx, a section only the firmware dictionary logger uses.

3. Remove the SOF DMA trace. Zephyr builds can't enable it: zephyr/include/sof/trace/preproc.h stops the build if CONFIG_TRACE is set. Removed:

   • src/trace and its Kconfig options (TRACE, TRACEV, TRACEE, TRACEM and the filtering options), trace-soflogger.h and dma-trace.h;
   • src/ipc/dma-copy.c and CONFIG_TRACE_CHANNEL, used only by the DMA trace;
   • the CONFIG_TRACE code in:
     • the IPC3 handler and PM gate (the stub that accepts trace commands stays);
     • the i.MX and AMD platform init;
     • system init and the panic handler;
     • the cmocka mocks;
   • ipc_build_trace_posn(), ipc_trigger_trace_xfer(), the trace fields of struct sof, and dump_hex() with its two Waves CONFIG_TRACEV users;
   • the CONFIG_TRACE=n lines in the board and app configs.

   Kept:

   • in trace.h, the Zephyr logging, host and no-op trace macros;
   • the host ABI in include/ipc/trace.h;
   • the dma_copy_*() declarations and DMA_TRACE_CHANNEL, which the legacy (non-native driver) probe and AMD DMA code still reference. They go with that code (Phase 0b).

4. Docs: AGENTS.md, the user-space README and the init README no longer list src/trace.

Functional change

• Zephyr firmware: no build compiled the DMA trace itself. Two changes:
  • the unused ipc_build_trace_posn() and ipc_trigger_trace_xfer() are no longer built;
  • struct sof is two pointers smaller, which changes field offsets in the code that uses it. On i.MX that shows up in the IPC3 helpers and pipeline-graph.c.
• Host builds: they print the same traces as before.

Testing

• Per commit:
  • Builds: tgl, tgl-h, mtl, lnl, ptl, wcl, nvl, imx8, imx8x, imx8m, imx8m_cm7, imx8ulp, imx95 and qemu_xtensa, and the extra CI configurations:
    • native_sim IPC3/IPC4 stub builds;
    • imx93 64-bit;
    • PTL with each userspace overlay;
    • PTL with CONFIG_SOF_USERSPACE_PROXY=n;
    • PTL and MTL with os_linux_overlay.conf;llext_relocatable.conf.
  • Comparisons: objects, LLEXT modules and firmware images. Only the third patch changes objects, as described above. sof_init() changes size on Cortex-M7 because it embeds the git hash.
  • Signing: rimage runs for every Intel platform, and the generated configs are unchanged.
  • Cadence xt-clang (RI-2022.10): MTL, PTL and NVL compile.
  • ZTest: the unit tests pass, run as in CI (18 of 18 configurations).
  • Host: the host library and the cmocka unit tests build and pass for all nine defconfigs and the library config.
• At the tip:
  • the testbench and the tools pass;
  • the platforms that already fail on main (acp_6_0, acp_7_0, acp_7_x, mt8186, mt8188, mt8195, mt8196, mt8365) have unchanged error sets.
• Hardware: TGL (UP Xtreme i11), sof-hda-generic-cavs25-2ch.tplg, comparing main (a206532) and this branch. Both boot, play 5 s, and play again after runtime suspend, with no DSP errors.

All of the above was run after rebasing onto a206532 (current main, including #11420).

🤖 Generated with Claude Code