AI/rocm-systems
2
0
Форкнуть 0
Код Задачи Запросы на слияние Действия Пакеты Проекты Релизы Вики Активность

Python_Cli_Examples (#696)

Просмотр исходного кода 
* Adjusted Python CLI examples

Signed-off-by: Maisam Arif <Maisam.Arif@amd.com>

[ROCm/amdsmi commit: ea225b459b]
Этот коммит содержится в:
Saeed, Oosman
2025-09-16 18:53:07 -05:00
коммит произвёл GitHub
родитель 10bfc7c056
Коммит d3fbbb4d36
1 изменённых файлов: 84 добавлений и 137 удалений
Скачать Patch файл Скачать Diff файл Показать все файлы Свернуть все файлы
projects/amdsmi/docs/how-to/amdsmi-cli-tool.md
+84 -137
Просмотреть файл
@@ -1081,153 +1081,100 @@ GPU: 0
### Listing CPER entries using amd-smi
This example code shows how to list CPER entries for all GPUs into files
```bash
~$ sudo amd-smi ras --cper --severity all --folder /tmp/cper_dump/
timestamp gpu_id severity file_name list of afids
2000/06/27 10:45:13 0 FATAL fatal-1.cper 30
2000/06/27 10:45:13 1 FATAL fatal-2.cper 30
2000/06/27 10:45:13 2 FATAL fatal-3.cper 30
2000/06/27 10:45:13 3 FATAL fatal-4.cper 30
2000/06/27 10:45:13 4 FATAL fatal-5.cper 30
2000/06/27 10:45:13 5 FATAL fatal-6.cper 30
2000/06/27 10:45:13 6 FATAL fatal-7.cper 30
2000/06/27 10:45:13 7 FATAL fatal-8.cper 30
```
This example code shows how to list CPER entries for a given GPU into files
```python
from amdsmi import *
import os
amdsmi_init()
def get_severity_mask(severity):
severity_mask = 0
if severity == "all":
# Set bits for NON_FATAL_UNCORRECTED (0), FATAL (1), and NON_FATAL_CORRECTED (2)
severity_mask |= ((1 << 0) | (1 << 1) | (1 << 2))
elif severity == "fatal":
# Set bit corresponding to AMDSMI_CPER_SEV_FATAL (which is 1)
severity_mask |= (1 << 1)
elif severity in ("nonfatal", "nonfatal-uncorrected"):
# Set bit corresponding to AMDSMI_CPER_SEV_NON_FATAL_UNCORRECTED (which is 0)
severity_mask |= (1 << 0)
elif severity in ("nonfatal-corrected", "corrected"):
# Set bit corresponding to AMDSMI_CPER_SEV_NON_FATAL_CORRECTED (which is 2)
severity_mask |= (1 << 2)
return severity_mask
def gpuid(device):
for gpu_index, device_handle in enumerate(amdsmi_interface.amdsmi_get_processor_handles()):
if device.value == device_handle.value:
return gpu_index
def dump_cper_entry(entry, cper_data, key):
try:
os.mkdir("/tmp/cper_dump", mode=0o777, dir_fd=None)
except FileExistsError:
pass
cper_file = f"/tmp/cper_dump/cper_entry_{key}.bin"
with open(cper_file, "wb") as file:
size = cper_data[key]["size"]
data = cper_data[key]["bytes"]
data = bytes(x % 256 for x in data[:size])
file.write(data)
print(f" Wrote cper data to file: {cper_file}")
json_file = f"/tmp/cper_dump/cper_entry_{key}.json"
with open(json_file, "wt") as file:
file.write(str(entry))
def get_gpu_cper_entries():
try:
devices = amdsmi_interface.amdsmi_get_processor_handles()
buffer_size = 1024*100
initial_cursor = 0
severity = "all"
for device in devices:
while True:
entries, new_cursor, cper_data, status_code = amdsmi_get_gpu_cper_entries(
device, get_severity_mask(severity), buffer_size, initial_cursor)
gpu_id = gpuid(device)
print("#############################################################################")
print(f"cper entries for severity: '{severity}', gpu #{gpu_id}, cursor: {initial_cursor}-{new_cursor - 1}")
for key, entry in entries.items():
print("----------------")
print("Entry", initial_cursor + key)
print(" Error Severity:", entry.get("error_severity", "Unknown"))
print(" Notify Type:", entry.get("notify_type", "Unknown"))
print(" Timestamp:", entry.get("timestamp", ""))
print(f" Cper entry metadata: {entry}")
dump_cper_entry(entry, cper_data, key)
if initial_cursor == new_cursor:
break
initial_cursor=new_cursor
break
except AmdSmiException as e:
print(e)
get_gpu_cper_entries()
```bash
~$ sudo amd-smi ras --cper --severity all --folder /tmp/cper_dump/ --gpu 1
timestamp gpu_id severity file_name list of afids
2000/06/27 10:45:13 1 FATAL fatal-1.cper 30
```
Output:
This example code shows how to list CPER entries and their JSON data for a given GPU into files
```shell
cper entries for severity: 'all', gpu #0, cursor: 0-3
----------------
Entry 0
Error Severity: non_fatal_corrected
Notify Type: CMC
Timestamp: 2025/09/07 00:14:22
Cper entry metadata: {'error_severity': 'non_fatal_corrected', 'notify_type': 'CMC', 'timestamp': '2025/09/07 00:14:22', 'signature': b'CPER', 'revision': 256, 'signature_end': '0xffffffff', 'sec_cnt': 1, 'record_length': 472, 'platform_id': b'0x1002:0x74A2', 'creator_id': b'amdgpu', 'record_id': b'5:1', 'flags': 0, 'persistence_info': 0}
Wrote cper data to file: /tmp/cper_dump/cper_entry_0.bin
----------------
Entry 1
Error Severity: non_fatal_corrected
Notify Type: CMC
Timestamp: 2025/09/07 00:14:26
Cper entry metadata: {'error_severity': 'non_fatal_corrected', 'notify_type': 'CMC', 'timestamp': '2025/09/07 00:14:26', 'signature': b'CPER', 'revision': 256, 'signature_end': '0xffffffff', 'sec_cnt': 1, 'record_length': 472, 'platform_id': b'0x1002:0x74A2', 'creator_id': b'amdgpu', 'record_id': b'5:2', 'flags': 0, 'persistence_info': 0}
Wrote cper data to file: /tmp/cper_dump/cper_entry_1.bin
----------------
Entry 2
Error Severity: non_fatal_corrected
Notify Type: CMC
Timestamp: 2025/09/08 06:12:11
Cper entry metadata: {'error_severity': 'non_fatal_corrected', 'notify_type': 'CMC', 'timestamp': '2025/09/08 06:12:11', 'signature': b'CPER', 'revision': 256, 'signature_end': '0xffffffff', 'sec_cnt': 1, 'record_length': 472, 'platform_id': b'0x1002:0x74A2', 'creator_id': b'amdgpu', 'record_id': b'5:3', 'flags': 0, 'persistence_info': 0}
Wrote cper data to file: /tmp/cper_dump/cper_entry_2.bin
----------------
Entry 3
Error Severity: non_fatal_corrected
Notify Type: CMC
Timestamp: 2025/09/08 06:13:59
Cper entry metadata: {'error_severity': 'non_fatal_corrected', 'notify_type': 'CMC', 'timestamp': '2025/09/08 06:13:59', 'signature': b'CPER', 'revision': 256, 'signature_end': '0xffffffff', 'sec_cnt': 1, 'record_length': 472, 'platform_id': b'0x1002:0x74A2', 'creator_id': b'amdgpu', 'record_id': b'5:4', 'flags': 0, 'persistence_info': 0}
Wrote cper data to file: /tmp/cper_dump/cper_entry_3.bin
#############################################################################
cper entries for severity: 'all', gpu #0, cursor: 4-3
```bash
~$ sudo amd-smi ras --cper --severity all --folder /tmp/cper_dump/ --gpu 1 --json
timestamp gpu_id severity file_name list of afids
2000/06/27 10:45:13 1 FATAL fatal-1.cper 30
~$ ls -alh /tmp/cper_dump/
total 12K
drwxr-xr-x 2 root root 46 Sep 16 21:12 .
drwxrwxrwt 1 root root 4.0K Sep 16 18:03 ..
-rw-r--r-- 1 root root 376 Sep 16 21:12 fatal-1.cper
-rw-r--r-- 1 root root 347 Sep 16 21:12 fatal-1.json
~$ cat /tmp/cper_dump/fatal-1.json
{
"error_severity": "fatal",
"notify_type": "MCE",
"timestamp": "2000/06/27 10:45:13",
"signature": "CPER",
"revision": 256,
"signature_end": "0xffffffff",
"sec_cnt": 1,
"record_length": 376,
"platform_id": "111102-G40307-0C",
"creator_id": "136c692517001839",
"record_id": "f0000031",
"flags": 0,
"persistence_info": 0
}
```
### Listing AFID numbers from CPER files
This example code shows how to continuously list CPER entries without exiting
This example code shows how to retrieve the AFID numbers from CPER files
```python
from amdsmi import *
import os
amdsmi_init()
def amdsmi_get_afids_from_cper():
directory_path = "/tmp/cper_dump/"
print(f"Searching for cper file in {directory_path}")
with os.scandir(directory_path) as cper_files:
for cper_file in cper_files:
if cper_file.is_file(): # Check if the entry is a file (not a subdirectory)
if ".bin" in cper_file.path:
print(f"Found {cper_file.path}")
with open(cper_file.path, "rb") as file:
raw = file.read()
afids, num_afids = amdsmi_interface.amdsmi_get_afids_from_cper(raw)
print(f"afids: {afids}")
amdsmi_get_afids_from_cper()
```bash
~$ sudo amd-smi ras --cper --follow --severity all --folder /tmp/cper_dump
Press CTRL + C to stop.
timestamp gpu_id severity file_name list of afids
2000/06/27 10:45:13 0 FATAL fatal-1.cper 30
2000/06/27 10:45:13 1 FATAL fatal-2.cper 30
2000/06/27 10:45:13 2 FATAL fatal-3.cper 30
2000/06/27 10:45:13 3 FATAL fatal-4.cper 30
2000/06/27 10:45:13 4 FATAL fatal-5.cper 30
2000/06/27 10:45:13 5 FATAL fatal-6.cper 30
2000/06/27 10:45:13 6 FATAL fatal-7.cper 30
2000/06/27 10:45:13 7 FATAL fatal-8.cper 30
...
```
Output:
This example code shows how to list CPER entries with a limited number of entries
```shell
sudo python3 afid.py
Searching for cper file in /tmp/cper_dump/
Found /tmp/cper_dump/cper_entry_0.bin
afids: [17]
Found /tmp/cper_dump/cper_entry_1.bin
afids: [17]
```bash
~$ sudo amd-smi ras --cper --severity all --folder /tmp/cper_dump --file-limit 5
timestamp gpu_id severity file_name list of afids
2000/06/27 10:45:13 0 FATAL fatal-1.cper 30
2000/06/27 10:45:13 1 FATAL fatal-2.cper 30
2000/06/27 10:45:13 2 FATAL fatal-3.cper 30
2000/06/27 10:45:13 3 FATAL fatal-4.cper 30
2000/06/27 10:45:13 4 FATAL fatal-5.cper 30
```
This example code shows how to list a specific severity of CPER entries only
```bash
~$ sudo amd-smi ras --cper --severity fatal --folder /tmp/cper_dump/
timestamp gpu_id severity file_name list of afids
2000/06/27 10:45:13 0 FATAL fatal-1.cper 30
```
This example code shows how to dump AFID errors in a CPER file
```bash
~$ sudo amd-smi ras --afid --cper-file /tmp/cper_dump/fatal-1.cper
```
Refer to amd_smi_cper_example.py & amd_smi_afid_example.py for API examples