fix: leaking amd card fd

Signed-off-by: thxCode <thxcode0824@gmail.com>

Author: thxCode

gpustack_runtime/detector/amd.py

@@ -142,12 +142,11 @@ def detect(self) -> Devices | None:
                 if (
                     not dev_cores or not dev_asic_family_id
                 ) and dev_card_id is not None:
-                    with contextlib.suppress(pyamdgpu.AMDGPUError):
-                        _, _, dev_gpudev = pyamdgpu.amdgpu_device_initialize(
-                            dev_card_id,
-                        )
+                    with (
+                        contextlib.suppress(pyamdgpu.AMDGPUError),
+                        pyamdgpu.amdgpu_device(dev_card_id) as dev_gpudev,
+                    ):
                         dev_gpudev_info = pyamdgpu.amdgpu_query_gpu_info(dev_gpudev)
-                        pyamdgpu.amdgpu_device_deinitialize(dev_gpudev)
                         if not dev_cores:
                             dev_cores = dev_gpudev_info.cu_active_number
                         if not dev_asic_family_id:

gpustack_runtime/detector/hygon.py

@@ -132,12 +132,11 @@ def detect(self) -> Devices | None:
 
                 dev_cores = dev_hsa_agent.compute_units
                 if not dev_cores and dev_card_id is not None:
-                    with contextlib.suppress(pyamdgpu.AMDGPUError):
-                        _, _, dev_gpudev = pyamdgpu.amdgpu_device_initialize(
-                            dev_card_id,
-                        )
+                    with (
+                        contextlib.suppress(pyamdgpu.AMDGPUError),
+                        pyamdgpu.amdgpu_device(dev_card_id) as dev_gpudev,
+                    ):
                         dev_gpudev_info = pyamdgpu.amdgpu_query_gpu_info(dev_gpudev)
-                        pyamdgpu.amdgpu_device_deinitialize(dev_gpudev)
                         dev_cores = dev_gpudev_info.cu_active_number
 
                 dev_cores_util = pyrocmsmi.rsmi_dev_busy_percent_get(dev_idx)

gpustack_runtime/detector/pyamdgpu/__init__.py

@@ -7,6 +7,7 @@
 import os
 import sys
 import threading
+from contextlib import contextmanager
 from ctypes import *
 from typing import ClassVar
 
@@ -247,31 +248,32 @@ def _LoadAMDGPULibrary():
 
 
 ## C function wrappers ##
-def amdgpu_device_initialize(card=1):
+@contextmanager
+def amdgpu_device(card=1):
     _LoadAMDGPULibrary()
 
     try:
         fd = os.open(f"/dev/dri/card{card}", os.O_RDONLY)
     except Exception:
         raise AMDGPUError(AMDGPU_ERROR_CARD_NOTFOUND)
 
-    c_major = c_uint32()
-    c_minor = c_uint32()
     device = c_amdgpu_device_t()
-    fn = _amdgpuGetFunctionPointer("amdgpu_device_initialize")
-    # If receive an error print here, try
-    # sudo vim /etc/default/grub
-    # and add "amdgpu.dc=0" to GRUB_CMDLINE_LINUX_DEFAULT
-    # then run "sudo update-grub" and reboot.
-    ret = fn(fd, byref(c_major), byref(c_minor), byref(device))
-    _amdgpuCheckReturn(ret)
-    return c_major.value, c_minor.value, device
 
+    try:
+        c_major = c_uint32()
+        c_minor = c_uint32()
+        fn = _amdgpuGetFunctionPointer("amdgpu_device_initialize")
+        ret = fn(fd, byref(c_major), byref(c_minor), byref(device))
+        _amdgpuCheckReturn(ret)
+
+        yield device
 
-def amdgpu_device_deinitialize(device):
-    fn = _amdgpuGetFunctionPointer("amdgpu_device_deinitialize")
-    ret = fn(device)
-    _amdgpuCheckReturn(ret)
+    finally:
+        try:
+            fn = _amdgpuGetFunctionPointer("amdgpu_device_deinitialize")
+            fn(device)
+        finally:
+            os.close(fd)
 
 
 def amdgpu_query_gpu_info(device):