userspace: prepare DP task to handle all the module methods (part 3 of #10287) #10379

lyakh · 2025-11-19T15:27:26Z

Prepare further parts for the userspace conversion: the DP thread, fast-get and the ring-buffer

Copilot

Pull Request Overview

This PR prepares the userspace conversion by refactoring DP thread initialization, fast-get/fast-put APIs, and ring buffer allocation to support user-mode execution with proper heap management.

Moved DP thread creation from scheduler_dp_task_shedule to scheduler_dp_task_init to initialize threads earlier
Converted fast_get and fast_put to syscalls that accept a k_heap parameter for userspace compatibility
Updated ring buffer allocation to use heap-based allocation instead of global memory functions

Reviewed Changes

Copilot reviewed 15 out of 15 changed files in this pull request and generated 4 comments.

Show a summary per file

File	Description
zephyr/lib/fast-get.c	Converted fast_get/fast_put to syscalls with heap parameter and added verification wrappers
zephyr/CMakeLists.txt	Added syscall header generation for fast-get.h
src/schedule/zephyr_dp_schedule.c	Moved thread creation from schedule to init, introduced undeclared variable bug
src/library_manager/llext_manager.c	Added memory domain partition management for loadable extensions
src/include/sof/llext_manager.h	Added llext_manager_add_domain function declaration
src/include/sof/lib/fast-get.h	Updated fast_get/fast_put signatures to include heap parameter as syscalls
src/include/sof/audio/component_ext.h	Removed schedule_task_cancel from comp_trigger_local (moved to module_adapter_free)
src/include/sof/audio/audio_buffer.h	Added heap pointer to sof_audio_buffer struct
src/audio/src/src_lite.c	Removed unused set_configuration and get_configuration interface assignments
src/audio/src/src_common.h	Removed declarations for unused src_set_config and src_get_config
src/audio/src/src_common.c	Removed unused src_set_config and src_get_config implementations
src/audio/src/src.c	Removed unused set_configuration and get_configuration interface assignments
src/audio/module_adapter/module_adapter.c	Moved DP task initialization earlier and added schedule_task_cancel to free function
src/audio/module_adapter/module/generic.c	Updated fast_get/fast_put calls to pass heap parameter
src/audio/buffers/ring_buffer.c	Converted to heap-based allocation with proper memory management

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

Copilot · 2025-11-19T15:33:40Z

src/schedule/zephyr_dp_schedule.c

+	pdata->thread_id = k_thread_create(pdata->thread, (__sparse_force void *)p_stack,
+					   stack_size, dp_thread_fn, *task, NULL, NULL,
+					   CONFIG_DP_THREAD_PRIORITY, (*task)->flags, K_FOREVER);
+
+	k_thread_access_grant(pdata->thread_id, pdata->sem);
+	scheduler_dp_grant(pdata->thread_id, cpu_get_id());
+
+	/* pin the thread to specific core */
+	ret = k_thread_cpu_pin(pdata->thread_id, core);


The variable pdata is used but never declared in this function. It should either be declared as struct task_dp_pdata *pdata = &task_memory->pdata; before line 621, or all occurrences of pdata should be changed to task_memory->pdata.

Copilot · 2025-11-19T15:33:40Z

src/schedule/zephyr_dp_schedule.c

+
+#ifdef CONFIG_USERSPACE
+	if ((*task)->flags & K_USER) {
+		ret = user_memory_init_shared(pdata->thread_id, pdata->mod);


The variable pdata is undeclared. Should use task_memory->pdata instead.

Copilot · 2025-11-19T15:33:41Z

src/schedule/zephyr_dp_schedule.c

+	k_sem_init(pdata->sem, 0, 1);
+	k_thread_start(pdata->thread_id);


The variable pdata is undeclared. Should use task_memory->pdata instead.

Copilot · 2025-11-19T15:33:41Z

src/schedule/zephyr_dp_schedule.c

 	return 0;
+
+e_thread:
+	k_thread_abort(pdata->thread_id);


The variable pdata is undeclared. Should use task_memory->pdata instead.

We need to move IPC processing for DP scheduled components into their thread context. For that the thread has to be started early. Create it immediately when creating DP task context. Note, that k_thread_create() never fails, so no need to check for error. Signed-off-by: Guennadi Liakhovetski <guennadi.liakhovetski@linux.intel.com>

We need to run all module callbacks in DP thread context, for this the thread has to be started early - before the first module callback is called. Signed-off-by: Guennadi Liakhovetski <guennadi.liakhovetski@linux.intel.com>

DP processing threads should have as long as life time as possible to process all the relevant IPCs in the thread context. Move thread termination to be called immediately before freeing module data. Signed-off-by: Guennadi Liakhovetski <guennadi.liakhovetski@linux.intel.com>

src_set_config() and src_get_config() aren't used, they would return an error if ever called. It's easier to just remove them. Signed-off-by: Guennadi Liakhovetski <guennadi.liakhovetski@linux.intel.com>

Add functions for adding LLEXT partitions to a memory domain for user-space modules. Signed-off-by: Guennadi Liakhovetski <guennadi.liakhovetski@linux.intel.com>

Allocate the ring-buffer object on module heap too for DP access. Signed-off-by: Guennadi Liakhovetski <guennadi.liakhovetski@linux.intel.com>

The ring-buffer data buffer has to be accessible to user-space DP modules, allocate it on module heap. Signed-off-by: Guennadi Liakhovetski <guennadi.liakhovetski@linux.intel.com>

kv2019i

Series looks good, question on last patch adding new syscalls.

kv2019i · 2025-11-19T18:17:46Z

zephyr/lib/fast-get.c

+#ifdef CONFIG_USERSPACE
+void z_vrfy_fast_put(struct k_heap *heap, const void *sram_ptr)
+{
+	z_impl_fast_put(heap, sram_ptr);


Shouldn't we have checks the user thread has access to 'heap' and 'sram_ptr'?

@kv2019i we should, yes, unfortunately it isn't very easy to check that perfectly... I've added some minimal checking, we need something better eventually

Thanks for adding the checks. We should add TODO/FIXMEs if there are known gaps.

Pass a "heap" argument to fast_get() and fast_put() for user-space DP allocations. Signed-off-by: Guennadi Liakhovetski <guennadi.liakhovetski@linux.intel.com>

Convert fast_get() and fast_put() to syscalls. Signed-off-by: Guennadi Liakhovetski <guennadi.liakhovetski@linux.intel.com>

kv2019i · 2025-11-20T15:54:24Z

zephyr/lib/fast-get.c

+#ifdef CONFIG_USERSPACE
+void z_vrfy_fast_put(struct k_heap *heap, const void *sram_ptr)
+{
+	K_OOPS(K_SYSCALL_MEMORY_WRITE(heap, sizeof(*heap)));


This is a bit surprising, I didn't realize k_heap object is fully visible to kernel space. But at least this is safe, it thread has write access, then this is ok.

kv2019i · 2025-11-20T15:58:22Z

zephyr/lib/fast-get.c

+#ifdef CONFIG_USERSPACE
+void z_vrfy_fast_put(struct k_heap *heap, const void *sram_ptr)
+{
+	z_impl_fast_put(heap, sram_ptr);


Thanks for adding the checks. We should add TODO/FIXMEs if there are known gaps.

softwarecki · 2025-11-21T15:50:12Z

Wow, that was merged really fast without giving a chance for review. Great job.

softwarecki · 2025-11-21T16:34:11Z

Additionally, this breaks building with CONFIG_USERSPACE enabled. Brilliant work!

sof/zephyr/lib/fast-get.c:176:2: warning: implicit declaration of function 'K_OOPS' is invalid in C99 [-Wimplicit-function-declaration]
        K_OOPS(K_SYSCALL_MEMORY_WRITE(heap, sizeof(*heap)));
        ^
sof/zephyr/lib/fast-get.c:176:9: warning: implicit declaration of function 'K_SYSCALL_MEMORY_WRITE' is invalid in C99 [-Wimplicit-function-declaration]
        K_OOPS(K_SYSCALL_MEMORY_WRITE(heap, sizeof(*heap)));
               ^
sof/zephyr/lib/fast-get.c:177:55: error: invalid application of 'sizeof' to an incomplete type 'struct z_heap'
        K_OOPS(K_SYSCALL_MEMORY_WRITE(heap->heap.heap, sizeof(*heap->heap.heap)));
                                                             ^~~~~~~~~~~~~~~~~~
zephyr/include/zephyr/sys/sys_heap.h:58:9: note: forward declaration of 'struct z_heap'
        struct z_heap *heap;
               ^
sof/zephyr/lib/fast-get.c:185:10: fatal error: 'zephyr/syscalls/mod_fast_put_mrsh.c' file not found
#include <zephyr/syscalls/mod_fast_put_mrsh.c>
         ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2 warnings and 2 errors generated.

Copilot AI review requested due to automatic review settings November 19, 2025 15:27

lyakh requested review from LaurentiuM1234, abonislawski, dbaluta, iuliana-prodan, jxstelter, kv2019i, lbetlej, lgirdwood, marcinszkudlinski, mmaka1, pblaszko, plbossart, ranj063 and singalsu as code owners November 19, 2025 15:27

Copilot started reviewing on behalf of lyakh November 19, 2025 15:30 View session

Copilot finished reviewing on behalf of lyakh November 19, 2025 15:32

Copilot AI reviewed Nov 19, 2025

View reviewed changes

lyakh added 7 commits November 19, 2025 16:54

schedule: dp: start the task early

acdb39b

We need to run all module callbacks in DP thread context, for this the thread has to be started early - before the first module callback is called. Signed-off-by: Guennadi Liakhovetski <guennadi.liakhovetski@linux.intel.com>

audio: src: remove unused functions

80dcb94

src_set_config() and src_get_config() aren't used, they would return an error if ever called. It's easier to just remove them. Signed-off-by: Guennadi Liakhovetski <guennadi.liakhovetski@linux.intel.com>

llext: add support for adding LLEXT memory to a memory domain

3ecabe3

Add functions for adding LLEXT partitions to a memory domain for user-space modules. Signed-off-by: Guennadi Liakhovetski <guennadi.liakhovetski@linux.intel.com>

ring-buffer: allocate on module heap

984ed50

Allocate the ring-buffer object on module heap too for DP access. Signed-off-by: Guennadi Liakhovetski <guennadi.liakhovetski@linux.intel.com>

ring-buffer: allocate the data buffer on module heap

93b8fd9

The ring-buffer data buffer has to be accessible to user-space DP modules, allocate it on module heap. Signed-off-by: Guennadi Liakhovetski <guennadi.liakhovetski@linux.intel.com>

lyakh force-pushed the user-p2 branch 5 times, most recently from b38c2f7 to 0bb8ad5 Compare November 19, 2025 16:19

lyakh force-pushed the user-p2 branch from 0bb8ad5 to 60029f3 Compare November 19, 2025 17:04

kv2019i reviewed Nov 19, 2025

View reviewed changes

lyakh added 2 commits November 20, 2025 09:59

fast-get: allocate on specific heap

298a9db

Pass a "heap" argument to fast_get() and fast_put() for user-space DP allocations. Signed-off-by: Guennadi Liakhovetski <guennadi.liakhovetski@linux.intel.com>

fast-get: convert to syscalls

879ba45

Convert fast_get() and fast_put() to syscalls. Signed-off-by: Guennadi Liakhovetski <guennadi.liakhovetski@linux.intel.com>

lyakh force-pushed the user-p2 branch from 60029f3 to 879ba45 Compare November 20, 2025 10:57

lyakh requested a review from softwarecki November 20, 2025 11:19

lgirdwood approved these changes Nov 20, 2025

View reviewed changes

kv2019i approved these changes Nov 20, 2025

View reviewed changes

lgirdwood merged commit 4c4ca6a into thesofproject:main Nov 20, 2025
34 of 42 checks passed

lyakh deleted the user-p2 branch November 20, 2025 18:23

lyakh mentioned this pull request Nov 24, 2025

fast-get: fix userspace #10388

Merged

This was referenced Nov 29, 2025

alloc: sof_heap: Fix ring buffer memory allocation #10402

Merged

Enable user-space for DP (part 4 of #10287) #10386

Merged

		k_sem_init(pdata->sem, 0, 1);
		k_thread_start(pdata->thread_id);

userspace: prepare DP task to handle all the module methods (part 3 of #10287) #10379

userspace: prepare DP task to handle all the module methods (part 3 of #10287) #10379

Uh oh!

Conversation

lyakh commented Nov 19, 2025

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull Request Overview

Reviewed Changes

Uh oh!

Copilot AI Nov 19, 2025

Choose a reason for hiding this comment

Uh oh!

Copilot AI Nov 19, 2025

Choose a reason for hiding this comment

Uh oh!

Copilot AI Nov 19, 2025

Choose a reason for hiding this comment

Uh oh!

Copilot AI Nov 19, 2025

Choose a reason for hiding this comment

Uh oh!

kv2019i left a comment

Choose a reason for hiding this comment

Uh oh!

kv2019i Nov 19, 2025

Choose a reason for hiding this comment

Uh oh!

lyakh Nov 20, 2025

Choose a reason for hiding this comment

Uh oh!

kv2019i Nov 20, 2025

Choose a reason for hiding this comment

Uh oh!

kv2019i Nov 20, 2025

Choose a reason for hiding this comment

Uh oh!

kv2019i Nov 20, 2025

Choose a reason for hiding this comment

Uh oh!

Uh oh!

softwarecki commented Nov 21, 2025

Uh oh!

softwarecki commented Nov 21, 2025

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants