sugar/linux
1
0
Fork 0
Code Activity

accel/amdxdna: Add IOCTL parameter for telemetry data

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

Reviewed-by: Mario Limonciello (AMD) <superm1@kernel.org>
Signed-off-by: Lizhi Hou <lizhi.hou@amd.com>
Link: https://patch.msgid.link/20251104062546.833771-3-lizhi.hou@amd.com
This commit is contained in:
Lizhi Hou
2025-11-03 22:25:45 -08:00
parent 1556c170d2
commit e568dc3e62
7 changed files with 173 additions and 10 deletions
Download Patch File Download Diff File Expand all files Collapse all files

56
drivers/accel/amdxdna/aie2_message.c
View File

@@ -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 @@ fail:
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))
{

25
drivers/accel/amdxdna/aie2_msg_priv.h
View File

@@ -9,7 +9,8 @@
enum aie2_msg_opcode {
MSG_OP_CREATE_CONTEXT = 0x2,
MSG_OP_DESTROY_CONTEXT = 0x3,
MSG_OP_SYNC_BO = 0x7,
MSG_OP_GET_TELEMETRY = 0x4,
MSG_OP_SYNC_BO = 0x7,
MSG_OP_EXECUTE_BUFFER_CF = 0xC,
MSG_OP_QUERY_COL_STATUS = 0xD,
MSG_OP_QUERY_AIE_TILE_INFO = 0xE,
@@ -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];

73
drivers/accel/amdxdna/aie2_pci.c
View File

@@ -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;

3
drivers/accel/amdxdna/aie2_pci.h
View File

@@ -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,

6
drivers/accel/amdxdna/amdxdna_mailbox_helper.h
View File

@@ -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, \

3
drivers/accel/amdxdna/amdxdna_pci_drv.c
View File

@@ -30,9 +30,10 @@ MODULE_FIRMWARE("amdnpu/17f0_20/npu.sbin");
* 0.2: Support getting last error hardware error
* 0.3: Support firmware debug buffer
* 0.4: Support getting resource information
* 0.5: Support getting telemetry data
*/
#define AMDXDNA_DRIVER_MAJOR 0
#define AMDXDNA_DRIVER_MINOR 4
#define AMDXDNA_DRIVER_MINOR 5
/*
* Bind the driver base on (vendor_id, device_id) pair and later use the

17
include/uapi/drm/amdxdna_accel.h
View File

@@ -442,6 +442,7 @@ enum amdxdna_drm_get_param {
DRM_AMDXDNA_QUERY_HW_CONTEXTS,
DRM_AMDXDNA_QUERY_FIRMWARE_VERSION = 8,
DRM_AMDXDNA_GET_POWER_MODE,
DRM_AMDXDNA_QUERY_TELEMETRY,
DRM_AMDXDNA_QUERY_RESOURCE_INFO = 12,
};
@@ -461,6 +462,22 @@ struct amdxdna_drm_get_resource_info {
__u64 npu_task_curr;
};
/**
* struct amdxdna_drm_query_telemetry_header - Telemetry data header
*/
struct amdxdna_drm_query_telemetry_header {
/** @major: Firmware telemetry interface major version number */
__u32 major;
/** @minor: Firmware telemetry interface minor version number */
__u32 minor;
/** @type: Telemetry query type */
__u32 type;
/** @map_num_elements: Total number of elements in the map table */
__u32 map_num_elements;
/** @map: Element map */
__u32 map[];
};
/**
* struct amdxdna_drm_get_info - Get some information from the AIE hardware.
* @param: Value in enum amdxdna_drm_get_param. Specifies the structure passed in the buffer.