Added the ve2 driver tracing support

[saifuddin-xilinx]

No description provided.

[Copilot]

Pull request overview

This PR adds VE2 driver tracing hooks and introduces a per-hardware-context ID to correlate trace events across hwctx lifecycle, command submission, scheduling, and IRQ completion paths.

Changes:

• Added a new generic trace event (__amdxdna_trace_point) and wrapper macro to record profiling points with function name + 3 integer args.
• Introduced hwctx ID allocation via an xarray in the VE2 device handle, and stored an ID in amdxdna_ctx_priv.
• Instrumented VE2 management/scheduler, IRQ handler, and HSA queue submission/commit/wait paths with new tracepoints.

Reviewed changes

Copilot reviewed 5 out of 5 changed files in this pull request and generated 4 comments.

Show a summary per file

File	Description
src/driver/amdxdna/ve2_of.h	Adds hwctx ID field and device-level XArray state for ID allocation.
src/driver/amdxdna/ve2_of.c	Initializes/destroys the hwctx ID XArray during device init/fini.
src/driver/amdxdna/ve2_mgmt.c	Adds tracepoints around partition init, scheduling, scheduler work, and IRQ completion.
src/driver/amdxdna/ve2_hwctx.c	Allocates/releases hwctx IDs; adds tracepoints around queue write/read index updates, submit, wait, init/fini.
src/driver/amdxdna/amdxdna_trace.h	Adds new trace event + macro wrapper capturing __func__.

Comments suppressed due to low confidence (3)

src/driver/amdxdna/ve2_of.h:90

• struct amdxdna_ctx_priv::id is declared as u64, but IDs are allocated via xa_alloc_cyclic() using a (u32 *)&priv->id cast. This relies on type-punning and only initializes the low 32 bits; it’s safer and clearer to store the ID as u32 (or allocate into a temporary u32 and then assign to the u64).

struct amdxdna_ctx_priv {
	u64				id; /* Unique incrementing hwctx ID */
	u32				start_col;
	u32				num_col;
	u32				mem_bitmap;
	u32				state;

src/driver/amdxdna/ve2_hwctx.c:1428

• After successfully allocating an ID with xa_alloc_cyclic(), later failures (e.g., ve2_xrs_request() or host queue creation) jump to cleanup_priv and free priv without removing the entry from hwctx_ids. This leaks the ID and leaves a stale pointer in the XArray; add a cleanup path that calls xa_erase() (or xa_release()) before freeing priv on all post-allocation error paths.

	/* Allocate unique ID using XArray (thread-safe, supports ID recycling) */
	ret = xa_alloc_cyclic(&xdna->dev_handle->hwctx_ids, (u32 *)&priv->id, priv,
			      XA_LIMIT(1, U32_MAX), &xdna->dev_handle->next_hwctx_id, GFP_KERNEL);
	if (ret < 0) {
		XDNA_ERR(xdna, "Failed to allocate hwctx ID, ret=%d", ret);
		goto cleanup_priv;
	}

	trace_amdxdna_trace_point("XRT_PROFILING_TRACE_ENTER",
				  client->pid, 0, priv->id, 0);
	XDNA_DBG(xdna, "hwctx init: enter hwctx=%p pid=%d", hwctx, client->pid);
	init_waitqueue_head(&priv->waitq);

	ret = ve2_xrs_request(xdna, hwctx);
	if (ret) {
		XDNA_ERR(xdna, "XRS resource request failed, ret=%d", ret);
		goto cleanup_priv;
	}

src/driver/amdxdna/ve2_hwctx.c:1000

• check_read_index() introduces static u64 read_index_value, which is shared across all callers/contexts and can be concurrently modified, producing incorrect comparisons/logs. This value should be a plain local variable (stack) instead of static.

static inline bool check_read_index(struct amdxdna_ctx *hwctx,
				    u64 seq, u32 print_interval)
{
	struct amdxdna_ctx_priv *priv_ctx = hwctx->priv;
	static u64 counter;
	static u64 read_index_value;
	u64 *read_index;

	if (!hwctx || !priv_ctx || !priv_ctx->hwctx_hsa_queue.hsa_queue_p)
		return false;

---

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

[Copilot]

Pull request overview

Copilot reviewed 5 out of 5 changed files in this pull request and generated 9 comments.

Comments suppressed due to low confidence (1)

src/driver/amdxdna/ve2_hwctx.c:1466

• xa_alloc_cyclic() inserts priv into xdna->dev_handle->hwctx_ids before the rest of hwctx initialization. If any later step fails (ve2_xrs_request(), ve2_create_host_queue(), etc.), the cleanup path frees priv but never removes the XArray entry, leaving a stale pointer and leaking the ID. Track whether allocation succeeded and xa_erase() the ID on all failure paths (or delay allocation until after the last failing step).

	/* Allocate unique ID using XArray (thread-safe, supports ID recycling) */
	ret = xa_alloc_cyclic(&xdna->dev_handle->hwctx_ids, (u32 *)&priv->id, priv,
			      XA_LIMIT(1, U32_MAX), &xdna->dev_handle->next_hwctx_id, GFP_KERNEL);
	if (ret < 0) {
		XDNA_ERR(xdna, "Failed to allocate hwctx ID, ret=%d", ret);
		goto cleanup_priv;
	}

	trace_amdxdna_trace_point("XRT_PROFILING_TRACE_ENTER",
				  client->pid, 0, priv->id, 0);
	XDNA_DBG(xdna, "hwctx init: enter hwctx=%p pid=%d", hwctx, client->pid);
	init_waitqueue_head(&priv->waitq);

	ret = ve2_xrs_request(xdna, hwctx);
	if (ret) {
		XDNA_ERR(xdna, "XRS resource request failed, ret=%d", ret);
		goto cleanup_priv;
	}

	/* Auto-select memory bitmap based on start_col */
	ve2_auto_select_mem_bitmap(xdna, hwctx);

	/* One host_queue entry per hwctx */
	ret = ve2_create_host_queue(xdna, hwctx, &priv->hwctx_hsa_queue);
	if (ret) {
		XDNA_ERR(xdna, "Failed to create host queue, ret=%d", ret);
		goto cleanup_xrs;
	}

	if (enable_polling) {
		XDNA_DBG(xdna, "Running in timer mode");
		timer_setup(&priv->event_timer, timeout_cb, 0);
		mod_timer(&priv->event_timer, jiffies + CTX_TIMER);
	} else {
		XDNA_DBG(xdna, "Running in interrupt mode");
	}

	if (verbosity >= VERBOSITY_LEVEL_DBG)
		ve2_clear_firmware_status(xdna, hwctx);

	mutex_init(&priv->privctx_lock);
	priv->state = AMDXDNA_HWCTX_STATE_IDLE;

	XDNA_DBG(xdna, "hwctx init: ready hwctx=%p start_col=%u pid=%d",
		 hwctx, priv->start_col, hwctx->client->pid);

	trace_amdxdna_trace_point("XRT_PROFILING_TRACE_EXIT",
				  hwctx->client->pid, priv->start_col, priv->id, 0);
	return 0;

cleanup_xrs:
	/* Releases XRS and partition (ve2_mgmt_destroy_partition calls ve2_xrs_release). */
	ve2_mgmt_destroy_partition(hwctx);
cleanup_priv:
	kfree(hwctx->priv);
	hwctx->priv = NULL;

---

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

[Copilot]

Pull request overview

Copilot reviewed 5 out of 5 changed files in this pull request and generated 6 comments.

Comments suppressed due to low confidence (1)

src/driver/amdxdna/ve2_hwctx.c:1471

• On error paths after a successful xa_alloc_cyclic() (e.g., failures in ve2_xrs_request() / ve2_create_host_queue()), cleanup falls through to cleanup_priv and frees hwctx->priv without removing the entry from xdna->dev_handle->hwctx_ids. This leaves a dangling pointer in the XArray and leaks the allocated ID. Ensure the allocated ID is xa_erase()d on all failure paths before freeing priv.

cleanup_xrs:
	/* Releases XRS and partition (ve2_mgmt_destroy_partition calls ve2_xrs_release). */
	ve2_mgmt_destroy_partition(hwctx);
cleanup_priv:
	kfree(hwctx->priv);

---

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