Add AMD GPU collector

CMakeLists.txt

@@ -671,6 +671,7 @@ set(PROC_PLUGIN_FILES
         collectors/proc.plugin/sys_fs_btrfs.c
         collectors/proc.plugin/sys_class_power_supply.c
         collectors/proc.plugin/sys_devices_pci_aer.c
+        collectors/proc.plugin/sys_class_drm.c
         )
 
 set(TC_PLUGIN_FILES

Makefile.am

@@ -423,6 +423,7 @@ PROC_PLUGIN_FILES = \
     collectors/proc.plugin/sys_fs_btrfs.c \
     collectors/proc.plugin/sys_class_power_supply.c \
     collectors/proc.plugin/sys_class_infiniband.c \
+    collectors/proc.plugin/sys_class_drm.c \
     $(NULL)
 
 PROFILE_PLUGIN_FILES = \

collectors/all.h

@@ -70,6 +70,10 @@
 
 #define NETDATA_CHART_PRIO_INTERRUPTS_PER_CORE        1100 // +1 per core
 
+// AMD GPUs
+
+#define NETDATA_CHART_PRIO_DRM_AMDGPU                 1005
+
 // Memory Section - 1xxx
 
 #define NETDATA_CHART_PRIO_MEM_SYSTEM_AVAILABLE       1010

collectors/proc.plugin/README.md

@@ -31,6 +31,7 @@ In detail, it collects metrics from:
 -   `/proc/spl/kstat/zfs/pool/state` (state of ZFS pools)
 -   `/sys/class/power_supply` (power supply properties)
 -   `/sys/class/infiniband` (infiniband interconnect)
+-   `/sys/class/drm` (AMD GPUs)
 -   `ipc` (IPC semaphores and message queues)
 -   `ksm` Kernel Same-Page Merging performance (several files under `/sys/kernel/mm/ksm`).
 -   `netdata` (internal Netdata resources utilization)
@@ -579,6 +580,36 @@ Default configuration will monitor only enabled infiniband ports, and refresh ne
   # refresh ports state every seconds = 30
 ```
 
+## AMD GPUs
+
+This module monitors every AMD GPU card discovered at agent startup.
+
+### Monitored GPU metrics
+
+The following charts will be provided:
+
+-   **GPU utilization**
+-   **GPU memory utilization**
+-   **GPU clock frequency**
+-   **GPU memory clock frequency**
+-   **VRAM memory usage percentage**
+-   **VRAM memory usage**
+-   **visible VRAM memory usage percentage**
+-   **visible VRAM memory usage**
+-   **GTT memory usage percentage**
+-   **GTT memory usage**
+
+### configuration
+
+The `drm` path can be configured if it differs from the default:
+
+```
+[plugin:proc:/sys/class/drm]
+  # directory to monitor = /sys/class/drm
+```
+
+> [!NOTE]  
+> Temperature, fan speed, voltage and power metrics for AMD GPUs can be monitored using the [Sensors](https://github.com/netdata/netdata/blob/master/collectors/charts.d.plugin/sensors/README.md) plugin.
 
 ## IPC
 

collectors/proc.plugin/metadata.yaml

@@ -5180,3 +5180,138 @@ modules:
                 - name: now
                 - name: max
                 - name: max_design
+  - meta:
+      plugin_name: proc.plugin
+      module_name: /sys/class/drm
+      monitored_instance:
+        name: AMD GPU
+        link: "https://www.amd.com"
+        categories:
+          - data-collection.hardware-devices-and-sensors
+        icon_filename: "amd.png"
+      related_resources:
+        integrations:
+          list: []
+      info_provided_to_referring_integrations:
+        description: ""
+      keywords:
+        - amd
+        - gpu
+        - hardware
+      most_popular: false
+    overview:
+      data_collection:
+        metrics_description: "This integration monitors AMD GPU metrics, such as utilization, clock frequency and memory usage."
+        method_description: ""
+      supported_platforms:
+        include:
+          - Linux
+        exclude: []
+      multi_instance: true
+      additional_permissions:
+        description: ""
+      default_behavior:
+        auto_detection:
+          description: ""
+        limits:
+          description: ""
+        performance_impact:
+          description: ""
+    setup:
+      prerequisites:
+        list: []
+      configuration:
+        file:
+          name: ""
+          description: ""
+        options:
+          description: ""
+          folding:
+            title: ""
+            enabled: true
+          list: []
+        examples:
+          folding:
+            enabled: true
+            title: ""
+          list: []
+    troubleshooting:
+      problems:
+        list: []
+    alerts: []
+    metrics:
+      folding:
+        title: Metrics
+        enabled: false
+      description: ""
+      availability: []
+      scopes:
+        - name: gpu
+          description: "These metrics refer to the GPU."
+          labels:
+            - name: product_name
+              description: GPU product name (e.g. AMD RX 6600)
+          metrics:
+            - name: amdgpu.gpu_utilization
+              description: GPU utilization
+              unit: "percentage"
+              chart_type: line
+              dimensions:
+                - name: utilization
+            - name: amdgpu.gpu_mem_utilization
+              description: GPU memory utilization
+              unit: "percentage"
+              chart_type: line
+              dimensions:
+                - name: utilization
+            - name: amdgpu.gpu_clk_frequency
+              description: GPU clock frequency
+              unit: "MHz"
+              chart_type: line
+              dimensions:
+                - name: frequency
+            - name: amdgpu.gpu_mem_clk_frequency
+              description: GPU memory clock frequency
+              unit: "MHz"
+              chart_type: line
+              dimensions:
+                - name: frequency
+            - name: amdgpu.gpu_mem_vram_usage_perc
+              description: VRAM memory usage percentage
+              unit: "percentage"
+              chart_type: line
+              dimensions:
+                - name: used
+            - name: amdgpu.gpu_mem_vram_usage
+              description: VRAM memory usage
+              unit: "bytes"
+              chart_type: area
+              dimensions:
+                - name: used
+                - name: free
+            - name: amdgpu.gpu_mem_vis_vram_usage_perc
+              description: visible VRAM memory usage percentage
+              unit: "percentage"
+              chart_type: line
+              dimensions:
+                - name: used
+            - name: amdgpu.gpu_mem_vis_vram_usage
+              description: visible VRAM memory usage
+              unit: "bytes"
+              chart_type: area
+              dimensions:
+                - name: used
+                - name: free
+            - name: amdgpu.gpu_mem_gtt_usage_perc
+              description: GTT memory usage percentage
+              unit: "percentage"
+              chart_type: line
+              dimensions:
+                - name: used
+            - name: amdgpu.gpu_mem_gtt_usage
+              description: GTT memory usage
+              unit: "bytes"
+              chart_type: area
+              dimensions:
+                - name: used
+                - name: free

collectors/proc.plugin/plugin_proc.c

@@ -70,8 +70,11 @@ static struct proc_module {
     // IPC metrics
     {.name = "ipc",                          .dim = "ipc",          .func = do_ipc},
 
-    {.name = "/sys/class/power_supply",      .dim = "power_supply", .func = do_sys_class_power_supply},
     // linux power supply metrics
+    {.name = "/sys/class/power_supply",      .dim = "power_supply", .func = do_sys_class_power_supply},
+
+    // GPU metrics
+    {.name = "/sys/class/drm",               .dim = "drm",          .func = do_sys_class_drm},
 
     // the terminator of this array
     {.name = NULL, .dim = NULL, .func = NULL}

collectors/proc.plugin/plugin_proc.h

@@ -46,6 +46,7 @@ int do_ipc(int update_every, usec_t dt);
 int do_sys_class_power_supply(int update_every, usec_t dt);
 int do_proc_pagetypeinfo(int update_every, usec_t dt);
 int do_sys_class_infiniband(int update_every, usec_t dt);
+int do_sys_class_drm(int update_every, usec_t dt);
 int get_numa_node_count(void);
 
 // metrics that need to be shared among data collectors