audio: comp_buffer: restore virtual heap routing for mem allocation

[wjablon1]

For SOF_USERSPACE_LL=n, CONFIG_USERSPACE=y configuration, the component buffers are allocated out of sof_heap instead of virtual_heap (alloc_ctx->heap == NULL + lack of SOF_MEM_FLAG_LARGE_BUFFER fall straight to sof_heap) which leads to out-of-memory errors for certain scenarios.

This commit fixes that by using sof_ctx_alloc and sof_ctx_free for all the configurations as it turns out these already have proper fallbacks in place when used with SOF_MEM_FLAG_LARGE_BUFFER flag.

[Copilot]

🟡 Changes recommended

buffer_set_size_range() can incorrectly reinitialize the stream to size 0 (and potentially return success) when all allocation attempts fail, which is a functional correctness issue in OOM scenarios.

Once you've addressed the issues Copilot identified, you can request another Copilot review.

Pull request overview

This PR fixes an out-of-memory scenario in Zephyr userspace builds by ensuring component buffer backing memory is allocated via sof_ctx_alloc() with the SOF_MEM_FLAG_LARGE_BUFFER flag, so the allocator can correctly route large allocations to the virtual heap (when applicable) even when alloc_ctx->heap == NULL.

Changes:

• Simplifies buffer data allocation/free paths to consistently use sof_ctx_alloc() / sof_ctx_free().
• Forces SOF_MEM_FLAG_LARGE_BUFFER on buffer stream allocations/resizes to restore virtual-heap routing in the affected configuration.
• Removes direct sof_heap_alloc()/sof_heap_free() fallbacks in comp_buffer.c for buffer stream memory.

File summaries

File	Description
src/audio/buffers/comp_buffer.c	Routes stream allocations/frees through sof_ctx_alloc/free with SOF_MEM_FLAG_LARGE_BUFFER to leverage allocator fallbacks (incl. virtual heap).

Review details

Suppressed comments (2)

src/audio/buffers/comp_buffer.c:317

• buffer_alloc_range() logs/returns errors using the original flags, but the allocation now always adds SOF_MEM_FLAG_LARGE_BUFFER. Logging the effective flags (and only logging "allocated" after success) avoids confusing traces when allocations fail.

	for (size = preferred_size; size >= minimum_size; size -= minimum_size) {
		stream_addr = sof_ctx_alloc(alloc, flags | SOF_MEM_FLAG_LARGE_BUFFER,
					    size, align);
		if (stream_addr)
			break;
	}

src/audio/buffers/comp_buffer.c:372

• On resize failure, the error log reports the current buffer size and the original flags, but the attempted allocation is for the requested size with SOF_MEM_FLAG_LARGE_BUFFER added. Logging the requested size and effective flags will make OOM debugging much clearer.

	new_ptr = sof_ctx_alloc(alloc, buffer->flags | SOF_MEM_FLAG_LARGE_BUFFER, size, alignment);

	/* we couldn't allocate bigger chunk */
	if (!new_ptr && size > audio_stream_get_size(&buffer->stream)) {
		buf_err(buffer, "resize can't alloc %u bytes of flags 0x%x",
			audio_stream_get_size(&buffer->stream), buffer->flags);
		return -ENOMEM;

• Files reviewed: 1/1 changed files
• Comments generated: 2
• Review effort level: Lite

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.

[intel-sofci]

PR 11182: test results

Run date: 2026-09-09 12:21 UTC

Tested commit: 168eb4f204eaf33003d31e3c116d702a03ee21d8

[lyakh]

what are actually cases when alloc == NULL? This appears to interfere with be10674 @serhiy-katsyuba-intel ?