ipc4: ace: Add dynamic log level control via IPC with runtime filtering #10536

tmleman · 2026-02-10T20:40:33Z

This PR adds support for dynamic firmware log level control through the IPC4_ENABLE_LOGS command, allowing the host to adjust logging verbosity at runtime without firmware recompilation.

Dependencies:

This PR depends on Zephyr upstream PR that adds CONFIG_LOG_RUNTIME_DEFAULT_LEVEL:

logging: Add CONFIG_LOG_RUNTIME_DEFAULT_LEVEL for startup filter control zephyrproject-rtos/zephyr#103820

Changes Overview:

IPC4 logging handler enhancement - Parse logs_mask field from IPC4_ENABLE_LOGS to determine requested log level and pass it to the backend, enabling dynamic control (ERROR/WARNING/INFO/DEBUG) based on host request
Platform configuration updates - Enable runtime filtering with ERROR default level for all Intel ADSP ACE platforms (MTL/LNL/PTL/WCL/NVL/NVLS), providing quiet boot while maintaining full debug capability

This commit will be replaced with actual zephyr update after necessary changes are merged. Signed-off-by: Tomasz Leman <tomasz.m.leman@intel.com>

Copilot

Pull request overview

Adds IPC4-driven runtime log level control so the host can adjust firmware logging verbosity dynamically, and enables Zephyr runtime log filtering on Intel ADSP ACE boards with a quiet default.

Changes:

Parse logs_mask from IPC4_ENABLE_LOGS and derive a Zephyr log level used when enabling the backend.
Enable CONFIG_LOG_RUNTIME_FILTERING and set CONFIG_LOG_RUNTIME_DEFAULT_LEVEL=1 (ERROR) across multiple Intel ADSP ACE board configs.
Point the Zephyr west manifest to an upstream PR branch providing CONFIG_LOG_RUNTIME_DEFAULT_LEVEL.

Reviewed changes

Copilot reviewed 9 out of 9 changed files in this pull request and generated 4 comments.

Show a summary per file

File	Description
west.yml	Switch Zephyr revision to the upstream PR providing `CONFIG_LOG_RUNTIME_DEFAULT_LEVEL`.
src/ipc/ipc4/logging.c	Compute log level from IPC4 `logs_mask` and pass it to `log_backend_enable()`.
app/debug_overlay.conf	Sets runtime filtering to `n` in the debug overlay config.
app/boards/intel_adsp_ace40_nvls.conf	Enables runtime filtering and sets default runtime level to ERROR.
app/boards/intel_adsp_ace40_nvl.conf	Enables runtime filtering and sets default runtime level to ERROR.
app/boards/intel_adsp_ace30_wcl.conf	Enables runtime filtering and sets default runtime level to ERROR.
app/boards/intel_adsp_ace30_ptl.conf	Enables runtime filtering and sets default runtime level to ERROR.
app/boards/intel_adsp_ace20_lnl.conf	Enables runtime filtering and sets default runtime level to ERROR.
app/boards/intel_adsp_ace15_mtpm.conf	Enables runtime filtering and sets default runtime level to ERROR.

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

west.yml

src/ipc/ipc4/logging.c

app/debug_overlay.conf

wjablon1 · 2026-02-11T08:38:51Z

src/ipc/ipc4/logging.c

-		/* Enable backend if not already active */
-		if (!log_backend_is_active(log_backend))
-			log_backend_enable(log_backend, mtrace_log_hook, CONFIG_SOF_LOG_LEVEL);
+		log_backend_enable(log_backend, mtrace_log_hook, log_level);


Does it make sense to enable backend for LOG_LEVEL_NONE?

First of all, this doesn't seem to be illegal (pointless, but not illegal).

This IPC properly implemented using what Zephyr logger offers will allow, for example, to silence all logs for a specified backend and then set them to a higher level only for a specified source. And in such a scenario, enabling with level set to NONE makes perfect sense.

abonislawski · 2026-02-11T08:49:49Z

@kv2019i what does Linux support for this IPC look like? Presumably it needs to be implemented? Is it fine to immediately switch the default log filtering state?

lyakh · 2026-02-11T11:36:03Z

app/boards/intel_adsp_ace30_ptl.conf

 CONFIG_LOG_FLUSH_SLEEP_US=5000
 CONFIG_WINSTREAM_CONSOLE=n
+CONFIG_LOG_RUNTIME_FILTERING=y
+CONFIG_LOG_RUNTIME_DEFAULT_LEVEL=1


Either I don't understand something or something isn't working. This means, that only ERR level logs should appear in mtrace, right? But if you check this PR logs it's still the usual INFO. If indeed this is trying to set log level to ERR, is this a good idea? We need INF level logs in CI

At this moment, according to my knowledge, everything works as I intended. The default log filtering level is set to ERROR at startup. This doesn't change the log level that gets compiled in, it still remains INFO. The fact that CI assumes that logs will appear in the memory window without sending IPC which is supposed to enable them, I've known for some time (I once wanted to disable logger backend autostart). The FW used in CI is built with the app/debug_overlay.conf and there runtime filtering is disabled to maintain the existing behavior expected by CI.

Implement dynamic log level selection based on the logs_mask field from the IPC4_ENABLE_LOGS command, allowing the host to control firmware logging verbosity at runtime. Previously, logging was enabled with a fixed compile-time level (CONFIG_SOF_LOG_LEVEL), preventing dynamic control. This change parses the logs_mask[0] bits 0-4 to determine the requested log level and passes it to the backend. Mask bit mapping to Zephyr log levels: - bit 0 (critical/error) -> LOG_LEVEL_ERR (1) - bit 1 (high/warning) -> LOG_LEVEL_WRN (2) - bit 2 (medium) -> LOG_LEVEL_INF (3) - bit 3 (low/info) -> LOG_LEVEL_INF (3) - bit 4 (verbose/debug) -> LOG_LEVEL_DBG (4) The highest set bit determines the maximum log level. This allows the host to request ERROR-only logging for quiet operation, or DEBUG for verbose troubleshooting, without firmware recompilation. Additionally, log_backend_enable() is now always called (it is idempotent and safe for reconfiguration), removing the previous check that prevented level changes on already-active backends. A TODO comment documents that the mask handling should be improved with proper macros or structs in the future for better code maintainability, but the current implementation is sufficient. Signed-off-by: Tomasz Leman <tomasz.m.leman@intel.com>

kv2019i

Good new feature, but we need to figure out how to introduce without breaking existing systems. Some comments inline.

kv2019i · 2026-02-11T11:54:30Z

src/ipc/ipc4/logging.c

+		else if (mask & BIT(1))
+			log_level = LOG_LEVEL_WRN;
+		else if (mask & BIT(0))
+			log_level = LOG_LEVEL_ERR;


We need to think about how to introduce this. Current shipping Linux kernels do not set log_mask and we cannot retroactively change shipped kernels (update of FW must not break user systems with older kernel), so this would effectively break mtrace for all existing systems as log level is forced to LOG_LEVEL_NONE. I think this needs to be a a build option (so we can keep the old behaviour in Linux overlay) and/or the default needs to be LOG_LEVEL_INF (like it is now). Looping in our kernel folks @bardliao @ujfalusi @ranj063

Why does FW that doesn't unnecessarily spam with logs mean breaking the user's system?

Enable CONFIG_LOG_RUNTIME_FILTERING and set CONFIG_LOG_RUNTIME_DEFAULT_LEVEL to 1 (ERROR) for all Intel ADSP ACE platforms. This provides quiet firmware boot with minimal logging overhead while maintaining full DEBUG compile-time capability for dynamic log level control via IPC. Platforms configured: - intel_adsp_ace15_mtpm - intel_adsp_ace20_lnl - intel_adsp_ace30_ptl - intel_adsp_ace30_wcl - intel_adsp_ace40_nvl - intel_adsp_ace40_nvls With these settings: - Firmware boots showing only ERROR level logs (quiet startup) - Host can dynamically increase verbosity to INFO or DEBUG via IPC4_ENABLE_LOGS without recompilation - All log macros (LOG_ERR, LOG_INF, etc.) remain compiled in Runtime filtering is explicitly disabled in debug_overlay.conf to maintain current debug build behavior where logs are not filtered at runtime. This works in conjunction with the IPC4 logging handler changes that parse logs_mask to set the appropriate runtime filter level. Signed-off-by: Tomasz Leman <tomasz.m.leman@intel.com>

Export log_dynamic_drc symbol when CONFIG_LOG_RUNTIME_FILTERING is enabled to fix LLEXT module linking failure. When runtime filtering is enabled, LOG_MODULE_REGISTER() creates both log_const_<module> and log_dynamic_<module> structures. LLEXT modules need to link against these symbols at load time. Previously, only log_const_drc was exported, causing the LLEXT linker to fail with: <wrn> llext: llext_link_plt: PLT: cannot find idx 46 name log_dynamic_drc <err> llext: do_llext_load: Failed to link, ret -2 <err> lib_manager: llext_manager_link_single: linking failed: -2 This prevented the DRC module from loading when built as LLEXT with runtime log filtering enabled. The fix conditionally exports log_dynamic_drc when runtime filtering is configured, allowing the LLEXT module to resolve all required logging symbols. Signed-off-by: Tomasz Leman <tomasz.m.leman@intel.com>

west: Changes from pull request in review

c6e666a

This commit will be replaced with actual zephyr update after necessary changes are merged. Signed-off-by: Tomasz Leman <tomasz.m.leman@intel.com>

Copilot AI review requested due to automatic review settings February 10, 2026 20:40

tmleman requested review from lbetlej, lgirdwood, mmaka1 and plbossart as code owners February 10, 2026 20:40

tmleman added the DNM Do Not Merge tag label Feb 10, 2026

tmleman requested review from dbaluta and kv2019i as code owners February 10, 2026 20:40

Copilot AI reviewed Feb 10, 2026

View reviewed changes

west.yml Show resolved Hide resolved

src/ipc/ipc4/logging.c Show resolved Hide resolved

src/ipc/ipc4/logging.c Outdated Show resolved Hide resolved

app/debug_overlay.conf Outdated Show resolved Hide resolved

tmleman requested review from abonislawski, serhiy-katsyuba-intel, softwarecki and wjablon1 February 10, 2026 20:43

wjablon1 reviewed Feb 11, 2026

View reviewed changes

lyakh reviewed Feb 11, 2026

View reviewed changes

tmleman force-pushed the topic/upstream/pr/intel/ace/logging/runtime_filtering branch from 21ebebb to e55ea21 Compare February 11, 2026 11:44

tmleman requested a review from a team as a code owner February 11, 2026 11:44

kv2019i requested changes Feb 11, 2026

View reviewed changes

tmleman added 2 commits February 11, 2026 13:43

tmleman force-pushed the topic/upstream/pr/intel/ace/logging/runtime_filtering branch from e55ea21 to 4ac3119 Compare February 11, 2026 12:43

tmleman requested review from kv2019i, lyakh and wjablon1 February 11, 2026 13:20

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

ipc4: ace: Add dynamic log level control via IPC with runtime filtering #10536

ipc4: ace: Add dynamic log level control via IPC with runtime filtering #10536

Uh oh!

tmleman commented Feb 10, 2026

Uh oh!

Copilot AI left a comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

wjablon1 Feb 11, 2026

Uh oh!

tmleman Feb 11, 2026

Uh oh!

abonislawski commented Feb 11, 2026

Uh oh!

lyakh Feb 11, 2026

Uh oh!

tmleman Feb 11, 2026

Uh oh!

kv2019i left a comment

Uh oh!

kv2019i Feb 11, 2026

Uh oh!

tmleman Feb 11, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

5 participants

ipc4: ace: Add dynamic log level control via IPC with runtime filtering #10536

Are you sure you want to change the base?

ipc4: ace: Add dynamic log level control via IPC with runtime filtering #10536

Uh oh!

Conversation

tmleman commented Feb 10, 2026

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Reviewed changes

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

wjablon1 Feb 11, 2026

Choose a reason for hiding this comment

Uh oh!

tmleman Feb 11, 2026

Choose a reason for hiding this comment

Uh oh!

abonislawski commented Feb 11, 2026

Uh oh!

lyakh Feb 11, 2026

Choose a reason for hiding this comment

Uh oh!

tmleman Feb 11, 2026

Choose a reason for hiding this comment

Uh oh!

kv2019i left a comment

Choose a reason for hiding this comment

Uh oh!

kv2019i Feb 11, 2026

Choose a reason for hiding this comment

Uh oh!

tmleman Feb 11, 2026

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

5 participants