Commit e568dc3e authored by Lizhi Hou's avatar Lizhi Hou
Browse files

accel/amdxdna: Add IOCTL parameter for telemetry data 



Extend DRM_IOCTL_AMDXDNA_GET_INFO to include additional parameters
that allow collection of telemetry data.

Reviewed-by: default avatarMario Limonciello (AMD) <superm1@kernel.org>
Signed-off-by: default avatarLizhi Hou <lizhi.hou@amd.com>
Link: https://patch.msgid.link/20251104062546.833771-3-lizhi.hou@amd.com
parent 1556c170

drivers/accel/amdxdna/aie2_message.c

+50 −6
Original line number Diff line number Diff line
@@ -47,7 +47,7 @@ static int aie2_send_mgmt_msg_wait(struct amdxdna_dev_hdl *ndev, 
		ndev->mgmt_chann = NULL;

	}



	if (!ret && *hdl->data != AIE2_STATUS_SUCCESS) {

	if (!ret && *hdl->status != AIE2_STATUS_SUCCESS) {

		XDNA_ERR(xdna, "command opcode 0x%x failed, status 0x%x",

			 msg->opcode, *hdl->data);

		ret = -EINVAL;
@@ -336,11 +336,6 @@ int aie2_query_status(struct amdxdna_dev_hdl *ndev, char __user *buf, 
		goto fail;

	}



	if (resp.status != AIE2_STATUS_SUCCESS) {

		XDNA_ERR(xdna, "Query NPU status failed, status 0x%x", resp.status);

		ret = -EINVAL;

		goto fail;

	}

	XDNA_DBG(xdna, "Query NPU status completed");



	if (size < resp.size) {
@@ -362,6 +357,55 @@ int aie2_query_status(struct amdxdna_dev_hdl *ndev, char __user *buf, 
	return ret;

}



int aie2_query_telemetry(struct amdxdna_dev_hdl *ndev,

			 char __user *buf, u32 size,

			 struct amdxdna_drm_query_telemetry_header *header)

{

	DECLARE_AIE2_MSG(get_telemetry, MSG_OP_GET_TELEMETRY);

	struct amdxdna_dev *xdna = ndev->xdna;

	dma_addr_t dma_addr;

	u8 *addr;

	int ret;



	if (header->type >= MAX_TELEMETRY_TYPE)

		return -EINVAL;



	addr = dma_alloc_noncoherent(xdna->ddev.dev, size, &dma_addr,

				     DMA_FROM_DEVICE, GFP_KERNEL);

	if (!addr)

		return -ENOMEM;



	req.buf_addr = dma_addr;

	req.buf_size = size;

	req.type = header->type;



	drm_clflush_virt_range(addr, size); /* device can access */

	ret = aie2_send_mgmt_msg_wait(ndev, &msg);

	if (ret) {

		XDNA_ERR(xdna, "Query telemetry failed, status %d", ret);

		goto free_buf;

	}



	if (size < resp.size) {

		ret = -EINVAL;

		XDNA_ERR(xdna, "Bad buffer size. Available: %u. Needs: %u", size, resp.size);

		goto free_buf;

	}



	if (copy_to_user(buf, addr, resp.size)) {

		ret = -EFAULT;

		XDNA_ERR(xdna, "Failed to copy telemetry to user space");

		goto free_buf;

	}



	header->major = resp.major;

	header->minor = resp.minor;



free_buf:

	dma_free_noncoherent(xdna->ddev.dev, size, addr, dma_addr, DMA_FROM_DEVICE);

	return ret;

}



int aie2_register_asyn_event_msg(struct amdxdna_dev_hdl *ndev, dma_addr_t addr, u32 size,

				 void *handle, int (*cb)(void*, void __iomem *, size_t))

{

drivers/accel/amdxdna/aie2_msg_priv.h

+24 −1
Original line number Diff line number Diff line
@@ -9,6 +9,7 @@ 
enum aie2_msg_opcode {

	MSG_OP_CREATE_CONTEXT              = 0x2,

	MSG_OP_DESTROY_CONTEXT             = 0x3,

	MSG_OP_GET_TELEMETRY               = 0x4,

	MSG_OP_SYNC_BO                     = 0x7,

	MSG_OP_EXECUTE_BUFFER_CF           = 0xC,

	MSG_OP_QUERY_COL_STATUS            = 0xD,
@@ -137,6 +138,28 @@ struct destroy_ctx_resp { 
	enum aie2_msg_status	status;

} __packed;



enum telemetry_type {

	TELEMETRY_TYPE_DISABLED,

	TELEMETRY_TYPE_HEALTH,

	TELEMETRY_TYPE_ERROR_INFO,

	TELEMETRY_TYPE_PROFILING,

	TELEMETRY_TYPE_DEBUG,

	MAX_TELEMETRY_TYPE

};



struct get_telemetry_req {

	enum telemetry_type	type;

	__u64	buf_addr;

	__u32	buf_size;

} __packed;



struct get_telemetry_resp {

	__u32	major;

	__u32	minor;

	__u32	size;

	enum aie2_msg_status	status;

} __packed;



struct execute_buffer_req {

	__u32	cu_idx;

	__u32	payload[19];

drivers/accel/amdxdna/aie2_pci.c

+73 −0
Original line number Diff line number Diff line
@@ -862,6 +862,76 @@ static int aie2_query_resource_info(struct amdxdna_client *client, 
	return 0;

}



static int aie2_fill_hwctx_map(struct amdxdna_hwctx *hwctx, void *arg)

{

	struct amdxdna_dev *xdna = hwctx->client->xdna;

	u32 *map = arg;



	if (hwctx->fw_ctx_id >= xdna->dev_handle->priv->hwctx_limit) {

		XDNA_ERR(xdna, "Invalid fw ctx id %d/%d ", hwctx->fw_ctx_id,

			 xdna->dev_handle->priv->hwctx_limit);

		return -EINVAL;

	}



	map[hwctx->fw_ctx_id] = hwctx->id;

	return 0;

}



static int aie2_get_telemetry(struct amdxdna_client *client,

			      struct amdxdna_drm_get_info *args)

{

	struct amdxdna_drm_query_telemetry_header *header __free(kfree) = NULL;

	u32 telemetry_data_sz, header_sz, elem_num;

	struct amdxdna_dev *xdna = client->xdna;

	struct amdxdna_client *tmp_client;

	int ret;



	elem_num = xdna->dev_handle->priv->hwctx_limit;

	header_sz = struct_size(header, map, elem_num);

	if (args->buffer_size <= header_sz) {

		XDNA_ERR(xdna, "Invalid buffer size");

		return -EINVAL;

	}



	telemetry_data_sz = args->buffer_size - header_sz;

	if (telemetry_data_sz > SZ_4M) {

		XDNA_ERR(xdna, "Buffer size is too big, %d", telemetry_data_sz);

		return -EINVAL;

	}



	header = kzalloc(header_sz, GFP_KERNEL);

	if (!header)

		return -ENOMEM;



	if (copy_from_user(header, u64_to_user_ptr(args->buffer), sizeof(*header))) {

		XDNA_ERR(xdna, "Failed to copy telemetry header from user");

		return -EFAULT;

	}



	header->map_num_elements = elem_num;

	list_for_each_entry(tmp_client, &xdna->client_list, node) {

		ret = amdxdna_hwctx_walk(tmp_client, &header->map,

					 aie2_fill_hwctx_map);

		if (ret)

			return ret;

	}



	ret = aie2_query_telemetry(xdna->dev_handle,

				   u64_to_user_ptr(args->buffer + header_sz),

				   telemetry_data_sz, header);

	if (ret) {

		XDNA_ERR(xdna, "Query telemetry failed ret %d", ret);

		return ret;

	}



	if (copy_to_user(u64_to_user_ptr(args->buffer), header, header_sz)) {

		XDNA_ERR(xdna, "Copy header failed");

		return -EFAULT;

	}



	return 0;

}



static int aie2_get_info(struct amdxdna_client *client, struct amdxdna_drm_get_info *args)

{

	struct amdxdna_dev *xdna = client->xdna;
@@ -896,6 +966,9 @@ static int aie2_get_info(struct amdxdna_client *client, struct amdxdna_drm_get_i 
	case DRM_AMDXDNA_GET_POWER_MODE:

		ret = aie2_get_power_mode(client, args);

		break;

	case DRM_AMDXDNA_QUERY_TELEMETRY:

		ret = aie2_get_telemetry(client, args);

		break;

	case DRM_AMDXDNA_QUERY_RESOURCE_INFO:

		ret = aie2_query_resource_info(client, args);

		break;

drivers/accel/amdxdna/aie2_pci.h

+3 −0
Original line number Diff line number Diff line
@@ -305,6 +305,9 @@ int aie2_create_context(struct amdxdna_dev_hdl *ndev, struct amdxdna_hwctx *hwct 
int aie2_destroy_context(struct amdxdna_dev_hdl *ndev, struct amdxdna_hwctx *hwctx);

int aie2_map_host_buf(struct amdxdna_dev_hdl *ndev, u32 context_id, u64 addr, u64 size);

int aie2_query_status(struct amdxdna_dev_hdl *ndev, char __user *buf, u32 size, u32 *cols_filled);

int aie2_query_telemetry(struct amdxdna_dev_hdl *ndev,

			 char __user *buf, u32 size,

			 struct amdxdna_drm_query_telemetry_header *header);

int aie2_register_asyn_event_msg(struct amdxdna_dev_hdl *ndev, dma_addr_t addr, u32 size,

				 void *handle, int (*cb)(void*, void __iomem *, size_t));

int aie2_config_cu(struct amdxdna_hwctx *hwctx,

drivers/accel/amdxdna/amdxdna_mailbox_helper.h

+4 −2
Original line number Diff line number Diff line
@@ -16,16 +16,18 @@ struct xdna_notify { 
	u32			*data;

	size_t			size;

	int			error;

	u32			*status;

};



#define DECLARE_XDNA_MSG_COMMON(name, op, status)			\

#define DECLARE_XDNA_MSG_COMMON(name, op, s)				\

	struct name##_req	req = { 0 };				\

	struct name##_resp	resp = { status	};			\

	struct name##_resp	resp = { .status = s };			\

	struct xdna_notify	hdl = {					\

		.error = 0,						\

		.data = (u32 *)&resp,					\

		.size = sizeof(resp),					\

		.comp = COMPLETION_INITIALIZER_ONSTACK(hdl.comp),	\

		.status = (u32 *)&resp.status,				\

	};								\

	struct xdna_mailbox_msg msg = {					\

		.send_data = (u8 *)&req,				\