[DRAFT] userspace LL/audio test PR (latest ver: V27)#10558
Draft
kv2019i wants to merge 52 commits into
Draft
Conversation
Collaborator
Author
|
Running the pipeline_two_components_user test added in this PR, on a Intel PTL, looks like this: |
kv2019i
force-pushed
the
feat/userspace-ll-wip
branch
from
9cdb1be to
37b0412
Compare
Collaborator
Author
|
V2 snapshot pushed:
|
kv2019i
force-pushed
the
feat/userspace-ll-wip
branch
from
37b0412 to
7317b18
Compare
Collaborator
Author
|
V3 snapshot pushed:
|
Collaborator
Author
|
Test output for the new test added in V3: |
kv2019i
commented
Mar 17, 2026
| /* works? yes */ | ||
| //return 0; | ||
|
|
||
| printk("ipc %p\n", ipc); |
Collaborator
Author
There was a problem hiding this comment.
Oops, these shouldn't be here. :)
kv2019i
commented
Mar 17, 2026
|
|
||
| /* create the pipeline */ | ||
| pipe = pipeline_new(NULL, pipe_desc->primary.r.instance_id, | ||
| pipe = pipeline_new(heap, pipe_desc->primary.r.instance_id, |
Collaborator
Author
There was a problem hiding this comment.
@lyakh @jsarha @lgirdwood This is where you'd plug in the vregions stuff to pass a separate heap to each pipe (based on topology description of its needs). In this series, as a placeholder, I use the zephyr_ll_user_heap() instead.
Collaborator
There was a problem hiding this comment.
right, maybe put a comment there for now to make re-discovery easier
lyakh
reviewed
Mar 18, 2026
lyakh
left a comment
lyakh
left a comment
Collaborator
There was a problem hiding this comment.
last reviewed commit so far "schedule: zephyr_ll: implement thread_init/free domain ops"
| #ifdef CONFIG_SOF_USERSPACE_LL | ||
| void comp_grant_access_to_thread(const struct comp_dev *dev, struct k_thread *th) | ||
| { | ||
| assert(dev->list_mutex); |
Collaborator
There was a problem hiding this comment.
description a bit confusing - this is only granting access to a mutex. Also list_mutex is only added to comp_dev in the next commit.
| } | ||
|
|
||
| stream_addr = rballoc_align(flags, size, align); | ||
| stream_addr = sof_heap_alloc(heap, flags, size, align); |
Collaborator
There was a problem hiding this comment.
the commit, that is mentioned in the commit message, only moved buffer context objects to particular heaps. This commit moves actual data buffers to them too, which is different and (arguably) more risky
| #define HDA_DMA_BUFFER_PERIOD_COUNT 4 | ||
|
|
||
| SHARED_DATA struct sof_dma dma[] = { | ||
| APP_TASK_DATA SHARED_DATA struct sof_dma dma[] = { |
Collaborator
There was a problem hiding this comment.
do I understand correctly, that this kind of userspace access makes that data writable to userspace?
| comp_err(dev, "requested channel %d is busy", hda_chan); | ||
| return -ENODEV; | ||
| } | ||
| hd->chan = &hd->dma->chan[channel]; |
Collaborator
There was a problem hiding this comment.
is this also not needed for the legacy mode? Also below
| uint64_t next_sync; | ||
| uint64_t period_in_cycles; | ||
| #endif | ||
| struct k_heap *heap; |
Collaborator
There was a problem hiding this comment.
wasn't this already referenced in the previous commit?
|
|
||
| k_spinlock_init(&dd->dai->lock); | ||
| #ifdef CONFIG_SOF_USERSPACE_LL | ||
| dd->dai->lock = k_object_alloc(K_OBJ_MUTEX); |
Collaborator
There was a problem hiding this comment.
check for NULL? Possibly in other locations too
| k_mutex_lock(dai->lock, K_FOREVER); | ||
| props = dai_get_properties(dai->dev, direction, stream_id); | ||
| hs_id = props->dma_hs_id; | ||
| ret = dai_get_properties_copy(dai->dev, direction, stream_id, &props); |
Collaborator
There was a problem hiding this comment.
I'm guessing this is made a syscall in one of the commits
| mod_heap = &mod_heap_user->heap; | ||
| } else { | ||
| #ifdef CONFIG_SOF_USERSPACE_LL | ||
| mod_heap = zephyr_ll_user_heap(); |
Collaborator
There was a problem hiding this comment.
looks good, but this else is entered under multiple conditions, might need to double-check
| .schedule_task_before = zephyr_ll_task_schedule_before, | ||
| .schedule_task_after = zephyr_ll_task_schedule_after, | ||
| .schedule_task_free = zephyr_ll_task_free, | ||
| .schedule_task_free = zephyr_ll_task_sched_free, |
Collaborator
There was a problem hiding this comment.
let's "spend" 3 more characters and make it ..._schedule_free()
| return -ENOMEM; | ||
| } | ||
| tr_err(&ll_tr, "Failed to allocate thread object for core %d", core); | ||
| dt->handler = NULL; |
lyakh
reviewed
Mar 18, 2026
| { | ||
| const struct sof_man_fw_desc *desc = basefw_vendor_get_manifest(); | ||
| const struct sof_man_module *mod; | ||
| uint32_t i; |
| uint32_t i; | ||
|
|
||
| if (!desc) | ||
| return -1; |
| return (int)i; | ||
| } | ||
|
|
||
| return -1; |
| union ipc4_connector_node_id node_id; | ||
| uint32_t dma_buffer_size; | ||
| uint32_t config_length; | ||
| } __packed __aligned(4); |
Collaborator
There was a problem hiding this comment.
does __aligned actually make sense in a type definition?
| pipe_msg.extension.dat = ipc_user->ipc_msg_ext; | ||
|
|
||
| /* Execute pipeline creation in user context */ | ||
| ipc_user->result = ipc_pipeline_new(ipc_user->ipc, (ipc_pipe_new *)&pipe_msg); |
Collaborator
There was a problem hiding this comment.
that's brave! ;-) I'd put a huge "TODO" here to make sure not to ship this by chance :-)
|
|
||
| /* create the pipeline */ | ||
| pipe = pipeline_new(NULL, pipe_desc->primary.r.instance_id, | ||
| pipe = pipeline_new(heap, pipe_desc->primary.r.instance_id, |
Collaborator
There was a problem hiding this comment.
right, maybe put a comment there for now to make re-discovery easier
kv2019i
force-pushed
the
feat/userspace-ll-wip
branch
from
7317b18 to
c029d49
Compare
Collaborator
Author
|
V4 snapshot pushed:
|
Collaborator
Author
|
Example output with V4 patchset: |
kv2019i
force-pushed
the
feat/userspace-ll-wip
branch
from
c029d49 to
9c778e4
Compare
Collaborator
Author
|
V5 pushed:
|
kv2019i
force-pushed
the
feat/userspace-ll-wip
branch
from
9c778e4 to
c4b127f
Compare
Collaborator
Author
|
V6 pushed:
|
kv2019i
force-pushed
the
feat/userspace-ll-wip
branch
from
c4b127f to
ad359d6
Compare
Collaborator
Author
|
V7 submitted:
|
kv2019i
force-pushed
the
feat/userspace-ll-wip
branch
from
ad359d6 to
c7e9a37
Compare
Collaborator
Author
|
V8 submitted:
|
kv2019i
force-pushed
the
feat/userspace-ll-wip
branch
from
c7e9a37 to
8af1be4
Compare
Collaborator
Author
|
V9 submitted:
|
Collaborator
Author
|
V10:
|
kv2019i
force-pushed
the
feat/userspace-ll-wip
branch
from
aa5efe0 to
35a61b6
Compare
Collaborator
Author
|
V11 pushed:
|
kv2019i
force-pushed
the
feat/userspace-ll-wip
branch
from
35a61b6 to
214c051
Compare
Collaborator
Author
|
V12 pushed:
|
When LL scheduler is run in user-space, use a different Zephyr thread name. Signed-off-by: Kai Vehmanen <kai.vehmanen@linux.intel.com>
The COHERENT_CHECK_NONSHARED_CORES debug macros call cpu_get_id() which invokes arch_proc_id() - a privileged hardware register read that faults in user-space context. Disable the entire debug block at compile time when CONFIG_SOF_USERSPACE_LL is enabled. This also fixes the same latent issue in CORE_CHECK_STRUCT and CORE_CHECK_STRUCT_INIT. Signed-off-by: Kai Vehmanen <kai.vehmanen@linux.intel.com>
Modify the checks in zephyr_ll_assert_core() to make them safe to call from user-space LL threads. Signed-off-by: Kai Vehmanen <kai.vehmanen@linux.intel.com>
Needs more review, but makes the tests pass again. Signed-off-by: Kai Vehmanen <kai.vehmanen@linux.intel.com>
Build fails when building with CONFIG_THREAD_NAME disabled. Fix the issue by conditional compilation of code using CONFIG_THREAD_MAX_NAME_LEN. Signed-off-by: Kai Vehmanen <kai.vehmanen@linux.intel.com>
In user-space LL builds the low-latency scheduler runs its work in a dedicated privileged domain thread, created together with its timer and access grants by scheduler_init_context() (zephyr_ll_init_context() -> domain_thread_init()). This context is per-core and must exist on every core that runs LL tasks. So far it was only established for the primary core, so LL tasks could not be scheduled on secondary cores when CONFIG_SOF_USERSPACE_LL is enabled. Allocate an LL task in secondary_core_init() and run scheduler_init_context() on it, giving each secondary core its own LL domain thread. A dedicated sec_core_init UUID is registered for the task. The whole block is compiled in only for CONFIG_SOF_USERSPACE_LL. Signed-off-by: Kai Vehmanen <kai.vehmanen@linux.intel.com>
Port chain DMA to SOF's sof_dma_* syscall wrappers (as host and dai already do) so its DMA operations can run unprivileged. Channel handles are stored as integer indices instead of kernel-only struct dma_chan_data pointers, matching the sof_dma_* API which takes indices. Both channel indices are initialised to -EINVAL so an incomplete initialisation is detected consistently for host and link. No functional change for existing (privileged) builds. Signed-off-by: Kai Vehmanen <kai.vehmanen@linux.intel.com>
When the LL pipeline runs in user-space (CONFIG_SOF_USERSPACE_LL) the component and its private data must reside on the user heap so the unprivileged user LL thread can access them. Introduce chain_dev_alloc()/ chain_cd_alloc() and their free counterparts to keep the config-specific allocation out of chain_task_create()/chain_task_free() instead of sprinkling #ifdefs through the control flow. The non-user-space path is unchanged (comp_alloc()/rzalloc()). Signed-off-by: Kai Vehmanen <kai.vehmanen@linux.intel.com>
For CONFIG_SOF_USERSPACE_LL the DMA buffer must be allocated from the user LL heap so it is reachable by the unprivileged user LL thread that runs chain_task_run(). Pass the LL alloc context to buffer_alloc() instead of NULL; for non-user-space builds alloc_ctx stays NULL and the default heap is used as before. Signed-off-by: Kai Vehmanen <kai.vehmanen@linux.intel.com>
Currently building with userspace LL and the debug overlay generates non-functional images. This is due to two incompatibilities: (1) cpu_get_id() called in userspace in zephyr_ll_task_init() and (2) the heap_alloc boot-test currently causing an exception. Fix the former and disable the latter to re-enabld debug builds with the userspace enabled. Signed-off-by: Guennadi Liakhovetski <guennadi.liakhovetski@linux.intel.com>
Make ipc_msg_send() a Zephyr system call so audio processing modules running in user-space LL threads can queue IPC messages (e.g. position updates, notifications) back to the host. The change takes effect only if CONFIG_SOF_USERSPACE_LL=y. Follows the same pattern used for ipc_msg_reply(): a dedicated header with __syscall declaration, z_impl/z_vrfy split, and syscall header registration in CMakeLists.txt. The verifier validates that the msg struct is writable (the implementation touches the list linkage) and that the data buffer, when provided, is readable up to msg->tx_size bytes. Signed-off-by: Jyri Sarha <jyri.sarha@linux.intel.com> (cherry picked from commit b25c15c)
Add an optional heap parameter to ipc_msg_w_ext_init() and ipc_msg_init() so callers can direct allocations to a specific heap. When the heap argument is NULL the existing rzalloc() path is used; when non-NULL, sof_heap_alloc()/sof_heap_free() are used instead. This allows IPC messages to be allocated from userspace-accessible heaps. For audio module contexts, introduce mod_ipc_msg_w_ext_init() and mod_ipc_msg_init() in generic.h. These use mod_zalloc()/ mod_free() for allocations that are automatically tracked and freed with the module lifecycle. ipc_msg_w_ext_init() is moved from a static inline in msg.h to a non-inline function in ipc-common.c due to the additional sof_heap_alloc dependency. Update all existing callers: - Module context callers (cadence, tdfb, sound_dose) use the new mod_ipc_msg_*() variants. - host-zephyr.c uses hd->heap, pipeline-graph.c uses the heap parameter from pipeline_new(). - Remaining kernel-context callers pass NULL for the default heap. Signed-off-by: Jyri Sarha <jyri.sarha@linux.intel.com> (cherry picked from commit 11a17d5)
Move user-facing notification functions (send_copier_gateway_xrun_notif_msg, send_gateway_xrun_notif_msg, send_mixer_underrun_notif_msg, send_process_data_error_notif_msg) to a new notification-user.c file so they can run in userspace. The send_resource_notif() function, which depends on the kernel-side notification pool and IPC message infrastructure, is converted to a Zephyr syscall. The implementation is renamed to z_impl_send_resource_notif() and remains in notification.c alongside is_notif_filtered_out() and ipc4_update_notification_mask(). The send_resource_notif() is converted to a system call only if CONFIG_SOF_USERSPACE_LL=y, without it the behaviour is same as befofe. A z_vrfy_send_resource_notif() handler is added to validate the user-provided data buffer and other parameters before forwarding to the kernel implementation. Signed-off-by: Jyri Sarha <jyri.sarha@linux.intel.com> (cherry picked from commit e385f10)
Add a new test to userspace_ll set that tests more of the functionality needed to run full audio pipelines in user-space. The test creates a pipeline with two components (IPC4 host and DAI copiers), does pipeline prepare, one copy cycle in prepared state and tears down the pipeline. One user-space thread is created to manage the pipelines. This would be equivalent to user IPC handler thread. Another user-space thread is created for the LL scheduler. Signed-off-by: Kai Vehmanen <kai.vehmanen@linux.intel.com>
Place the pipeline position lookup table in the sysuser memory partition and replace k_spinlock with a dynamically allocated k_mutex when CONFIG_SOF_USERSPACE_LL is enabled. Spinlocks disable interrupts which is a privileged operation unavailable from user-mode threads. The mutex pointer is stored in a separate APP_SYSUSER_BSS variable outside the SHARED_DATA struct so Zephyr's kernel object tracking can recognize it for syscall verification. Move pipeline_posn_init() from task_main_start() to primary_core_init() before platform_init(), so the mutex is allocated before ipc_user_init() grants thread access to it. In pipeline_posn_get(), bypass the sof_get() kernel singleton and access the shared structure directly when running in user-space. Grant the ipc_user_init thread access to the pipeline position mutex via new pipeline_posn_grant_access() helper. Signed-off-by: Kai Vehmanen <kai.vehmanen@linux.intel.com>
…pace task zephyr_ll_task_sched_free() frees an active (RUNNING/RESCHEDULE) task by setting pdata->freeing and waiting on pdata->sem for the scheduler thread to remove the task from its run list before the memory is released. Under CONFIG_SOF_USERSPACE_LL this function runs in kernel context while pdata->sem is a sys_sem allocated on the user heap. sys_sem_take() returns -EINVAL immediately when called from kernel context, so the wait is a no-op: pdata is freed (and the struct task is subsequently freed by pipeline_free()) while the task is still linked in sch->tasks with n_tasks != 0 and the scheduling domain handler still set. Because n_tasks is non-zero, schedule_free() does not stop the LL timer, and the next timer tick runs zephyr_ll_run() over the dangling task, dereferencing freed memory and taking a load/store-privilege exception (EXCCAUSE 26) in the user-space LL thread. Stop relying on the cross-privilege semaphore handshake in this path. When the task must be waited for, mark it cancelled so that, should it actually be mid-execution on the scheduler's temporary list, it is removed via the cancel path without re-running task->run() on resources the caller may already have freed. If the task is still linked on the run list, the scheduler thread is provably not executing it (a running task is moved off sch->tasks with the lock dropped), so remove it directly and skip the wait. This guarantees the task is delisted (n_tasks -> 0, handler -> NULL) before pdata is freed, eliminating both the dangling list entry and the stray timer wakeups. Verified on PTL with the standalone user-space LL boot tests: the userspace_ll suite, including pipeline_two_components_user, now passes without the fatal exception at teardown. Signed-off-by: Kai Vehmanen <kai.vehmanen@linux.intel.com>
If SOF is built with CONFIG_SOF_USERSPACE_LL, the IPC user handled will require access to coldrodata sections to initialize audio modules. This logic is not required for LLEXT modules, which have existing code to add access to coldrodata (and other sections). This commit is needed for builds where LLEXT is not used. Signed-off-by: Kai Vehmanen <kai.vehmanen@linux.intel.com>
This is a set of temporary changes to audio code to remove calls to privileged interfaces that are not mandatory to run simple audio tests. These need proper solutions to be able to run all use-cases in user LL version. Signed-off-by: Kai Vehmanen <kai.vehmanen@linux.intel.com>
This reverts commit 768faad.
This reverts commit 7bb70f5.
kv2019i
force-pushed
the
feat/userspace-ll-wip
branch
from
70358c5 to
537a39b
Compare
Collaborator
Author
|
V27 pushed:
|
kv2019i
changed the title
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
SOF has recently gained ability to run DP (=preemptable audio tasks) in Zephyr user-space.
This PR is an early stage pull-request for changes to extend this capability to all of the audio pipeline code, and specifically the LL (low-latency) tasks.
This early stage as the design is not set in stone and the PR uses a number of short cuts in order to move (and tests) incrementally larger sets of audio functionality.