Re: [PATCH V2] accel/amdxdna: Add debug prints for command submission

[Mario Limonciello]

On 3/16/26 12:56, Lizhi Hou wrote:
> Add debug prints to help diagnose issues with incoming command submissions.
>
> Signed-off-by: Lizhi Hou <lizhi.hou@xxxxxxx>

Reviewed-by: Mario Limonciello <mario.limonciello@xxxxxxx>

> ---
>   drivers/accel/amdxdna/aie2_message.c | 34 ++++++++++++++++++++++++----
>   1 file changed, 29 insertions(+), 5 deletions(-)
>
> diff --git a/drivers/accel/amdxdna/aie2_message.c b/drivers/accel/amdxdna/aie2_message.c
> index b764c7e8816a..4ec591306854 100644
> --- a/drivers/accel/amdxdna/aie2_message.c
> +++ b/drivers/accel/amdxdna/aie2_message.c
> @@ -1002,7 +1002,7 @@ int aie2_cmdlist_multi_execbuf(struct amdxdna_hwctx *hwctx,
>   	struct amdxdna_cmd_chain *payload;
>   	struct xdna_mailbox_msg msg;
>   	union exec_chain_req req;
> -	u32 payload_len;
> +	u32 payload_len, ccnt;
>   	u32 offset = 0;
>   	size_t size;
>   	int ret;
> @@ -1011,12 +1011,24 @@ int aie2_cmdlist_multi_execbuf(struct amdxdna_hwctx *hwctx,
>   
>   	op = amdxdna_cmd_get_op(cmd_abo);
>   	payload = amdxdna_cmd_get_payload(cmd_abo, &payload_len);
> -	if (op != ERT_CMD_CHAIN || !payload ||
> -	    payload_len < struct_size(payload, data, payload->command_count))
> +	if (op != ERT_CMD_CHAIN) {
> +		XDNA_DBG(xdna, "Invalid op code %d", op);
>   		return -EINVAL;
> +	}
> +
> +	if (!payload) {
> +		XDNA_DBG(xdna, "Failed to get command payload");
> +		return -EINVAL;
> +	}
> +
> +	ccnt = payload->command_count;
> +	if (payload_len < struct_size(payload, data, ccnt)) {
> +		XDNA_DBG(xdna, "Invalid command count %d", ccnt);
> +		return -EINVAL;
> +	}
>   
>   	op = ERT_INVALID_CMD;
> -	for (i = 0; i < payload->command_count; i++) {
> +	for (i = 0; i < ccnt; i++) {
>   		u32 boh = (u32)(payload->data[i]);
>   		struct amdxdna_gem_obj *abo;
>   
> @@ -1035,19 +1047,26 @@ int aie2_cmdlist_multi_execbuf(struct amdxdna_hwctx *hwctx,
>   
>   		offset += size;
>   	}
> +
> +	XDNA_DBG(xdna, "Total %d commands:", ccnt);
> +	print_hex_dump_debug("cmdbufs: ", DUMP_PREFIX_OFFSET, 16, 4,
> +			     cmdbuf_abo->mem.kva, offset, false);
> +
>   	msg.opcode = EXEC_MSG_OPS(xdna)->get_chain_msg_op(op);
>   	if (msg.opcode == MSG_OP_MAX_OPCODE)
>   		return -EOPNOTSUPP;
>   
>   	/* The offset is the accumulated total size of the cmd buffer */
>   	EXEC_MSG_OPS(xdna)->init_chain_req(&req, cmdbuf_abo->mem.dev_addr,
> -					   offset, payload->command_count);
> +					   offset, ccnt);
>   	drm_clflush_virt_range(cmdbuf_abo->mem.kva, offset);
>   
>   	msg.handle = job;
>   	msg.notify_cb = notify_cb;
>   	msg.send_data = (u8 *)&req;
>   	msg.send_size = sizeof(req);
> +	print_hex_dump_debug("cmdlist msg: ", DUMP_PREFIX_OFFSET, 16, 4,
> +			     &req, msg.send_size, false);
>   	ret = xdna_mailbox_send_msg(chann, &msg, TX_TIMEOUT);
>   	if (ret) {
>   		XDNA_ERR(xdna, "Send message failed");
> @@ -1076,6 +1095,9 @@ int aie2_cmdlist_single_execbuf(struct amdxdna_hwctx *hwctx,
>   	if (ret)
>   		return ret;
>   
> +	print_hex_dump_debug("cmdbuf: ", DUMP_PREFIX_OFFSET, 16, 4,
> +			     cmdbuf_abo->mem.kva, size, false);
> +
>   	msg.opcode = EXEC_MSG_OPS(xdna)->get_chain_msg_op(op);
>   	if (msg.opcode == MSG_OP_MAX_OPCODE)
>   		return -EOPNOTSUPP;
> @@ -1088,6 +1110,8 @@ int aie2_cmdlist_single_execbuf(struct amdxdna_hwctx *hwctx,
>   	msg.notify_cb = notify_cb;
>   	msg.send_data = (u8 *)&req;
>   	msg.send_size = sizeof(req);
> +	print_hex_dump_debug("cmdlist msg: ", DUMP_PREFIX_OFFSET, 16, 4,
> +			     &req, msg.send_size, false);
>   	ret = xdna_mailbox_send_msg(chann, &msg, TX_TIMEOUT);
>   	if (ret) {
>   		XDNA_ERR(hwctx->client->xdna, "Send message failed");