Amd device plugin

[mgsharm]

Description of changes:

This PR adds support for AMD GPU detection and device plugin functionality to Bottlerocket:

• libdrm package: Added libdrm 2.4.129 as one of the device plugin dependencies. Reference:
  https://github.com/ROCm/k8s-device-plugin/blob/763445e18f3838fa72b22e31a04ec25987334bff/Dockerfile#L15
• rocm-k8s-device-plugin package: Added AMD K8s device plugin v1.31.0.8 from ROCm for GPU resource management

Dependencies

• link-to-kernel-kit-pr - Adds AMD GPU DKMS driver and detection services

Testing done

AMD GPU Instance (MI300X with 8 GPUs)

PCI device detection:

bash-5.1# lspci -d 1002:75a3
51:00.0 Processing accelerators: Advanced Micro Devices, Inc. [AMD/ATI] Device 75a3
52:00.0 Processing accelerators: Advanced Micro Devices, Inc. [AMD/ATI] Device 75a3
62:00.0 Processing accelerators: Advanced Micro Devices, Inc. [AMD/ATI] Device 75a3
63:00.0 Processing accelerators: Advanced Micro Devices, Inc. [AMD/ATI] Device 75a3
73:00.0 Processing accelerators: Advanced Micro Devices, Inc. [AMD/ATI] Device 75a3
74:00.0 Processing accelerators: Advanced Micro Devices, Inc. [AMD/ATI] Device 75a3
84:00.0 Processing accelerators: Advanced Micro Devices, Inc. [AMD/ATI] Device 75a3
85:00.0 Processing accelerators: Advanced Micro Devices, Inc. [AMD/ATI] Device 75a3

ROCm device plugin running:

bash-5.1# systemctl list-units | grep -i rocm
  rocm-k8s-device-plugin.service                                                                                                                loaded active     
running   Start ROCm kubernetes device plugin

bash-5.1# systemctl status rocm-k8s-device-plugin.service
● rocm-k8s-device-plugin.service - Start ROCm kubernetes device plugin
     Loaded: loaded (/x86_64-bottlerocket-linux-gnu/sys-root/usr/lib/systemd/system/rocm-k8s-device-plugin.service; enabled; preset: enabled)
    Drop-In: /x86_64-bottlerocket-linux-gnu/sys-root/usr/lib/systemd/system/service.d
             └─00-aws-config.conf
     Active: active (running) since Wed 2025-11-26 06:22:34 UTC; 2min 47s ago
 Invocation: ae2b73174a7245b5976c1b9eb7835695
    Process: 47801 ExecStartPre=/usr/bin/sleep 0.1 (code=exited, status=0/SUCCESS)
    Process: 47823 ExecStartPre=/usr/bin/test -S /var/lib/kubelet/device-plugins/kubelet.sock (code=exited, status=0/SUCCESS)
   Main PID: 47825 (rocm-device-plu)
      Tasks: 12 (limit: 629145)
     Memory: 26.9M (peak: 28M)
        CPU: 63ms
     CGroup: /system.slice/rocm-k8s-device-plugin.service
             └─47825 /usr/bin/rocm-device-plugin -logtostderr=true -stderrthreshold=INFO -v=5

Nov 26 06:22:34 ip-172-31-60-25.us-west-2.compute.internal rocm-device-plugin[47825]: I1126 06:22:34.799514   47825 amdgpu.go:261] Devices map: map[0000:51:00.0:map[card:0 computePartitionType:spx devID:10833827524210284120 memoryPartitionType:nps1 nodeId:2 numaNode:0 renderD:128] 0000:52:00.0:map[card:8 computePartitionType:spx devID:5789222631865278233 memoryPartitionType:nps1 nodeId:3 numaNode:0 renderD:136] 0000:62:00.0:map[card:16 computePartitionType:spx devID:1537282919108789315 memoryPartitionType:nps1 nodeId:4 numaNode:0 renderD:144] 0000:63:00.0:map[card:24 computePartitionType:spx devID:17650589009955793669 memoryPartitionType:nps1 nodeId:5 numaNode:0 renderD:152] 0000:73:00.0:map[card:32 computePartitionType:spx devID:17599352850648317852 memoryPartitionType:nps1 nodeId:6 numaNode:1 renderD:160] 0000:74:00.0:map[card:40 computePartitionType:spx devID:16824009056103876010 memoryPartitionType:nps1 nodeId:7 numaNode:1 renderD:168] 0000:84:00.0:map[card:48 computePartitionType:spx devID:7459436697213601231 memoryPartitionType:nps1 nodeId:8 numaNode:1 renderD:176] 0000:85:00.0:map[card:56 computePartitionType:spx devID:10369896732957786260 memoryPartitionType:nps1 nodeId:9 numaNode:1 renderD:184]]
Nov 26 06:22:34 ip-172-31-60-25.us-west-2.compute.internal rocm-device-plugin[47825]: I1126 06:22:34.799540   47825 amdgpu.go:278] Partition counts: map[spx_nps1:8]
Nov 26 06:22:34 ip-172-31-60-25.us-west-2.compute.internal rocm-device-plugin[47825]: I1126 06:22:34.799548   47825 plugin.go:254] Watching GPU with bus ID: 0000:52:00.0 NUMA Node: [0]
Nov 26 06:22:34 ip-172-31-60-25.us-west-2.compute.internal rocm-device-plugin[47825]: I1126 06:22:34.799558   47825 plugin.go:254] Watching GPU with bus ID: 0000:62:00.0 NUMA Node: [0]
Nov 26 06:22:34 ip-172-31-60-25.us-west-2.compute.internal rocm-device-plugin[47825]: I1126 06:22:34.799561   47825 plugin.go:254] Watching GPU with bus ID: 0000:63:00.0 NUMA Node: [0]
Nov 26 06:22:34 ip-172-31-60-25.us-west-2.compute.internal rocm-device-plugin[47825]: I1126 06:22:34.799564   47825 plugin.go:254] Watching GPU with bus ID: 0000:73:00.0 NUMA Node: [1]
Nov 26 06:22:34 ip-172-31-60-25.us-west-2.compute.internal rocm-device-plugin[47825]: I1126 06:22:34.799567   47825 plugin.go:254] Watching GPU with bus ID: 0000:74:00.0 NUMA Node: [1]
Nov 26 06:22:34 ip-172-31-60-25.us-west-2.compute.internal rocm-device-plugin[47825]: I1126 06:22:34.799570   47825 plugin.go:254] Watching GPU with bus ID: 0000:84:00.0 NUMA Node: [1]
Nov 26 06:22:34 ip-172-31-60-25.us-west-2.compute.internal rocm-device-plugin[47825]: I1126 06:22:34.799572   47825 plugin.go:254] Watching GPU with bus ID: 0000:85:00.0 NUMA Node: [1]
Nov 26 06:22:34 ip-172-31-60-25.us-west-2.compute.internal rocm-device-plugin[47825]: I1126 06:22:34.799575   47825 plugin.go:254] Watching GPU with bus ID: 0000:51:00.0 NUMA Node: [0]

GPU capacity advertised to Kubernetes:

  ➜ kubectl get nodes -o json | jq '.items[].status.capacity'
  {
    "amd.com/gpu": "8",
    "cpu": "192",
    "ephemeral-storage": "15471392Ki",
    "hugepages-1Gi": "0",
    "hugepages-2Mi": "0",
    "memory": "4206782968Ki",
    "pods": "110"
  }

All services and targets that rocm-k8s-device-plugin.service depends on:

bash-5.1#   systemctl list-dependencies rocm-k8s-device-plugin.service
rocm-k8s-device-plugin.service
● ├─kubelet.service
● ├─[email protected]
● ├─system.slice
● └─sysinit.target
●   ├─dev-hugepages.mount
●   ├─dev-mqueue.mount
●   ├─kmod-static-nodes.service
●   ├─ldconfig.service
●   ├─proc-sys-fs-binfmt_misc.mount
●   ├─sys-fs-fuse-connections.mount
●   ├─sys-kernel-config.mount
●   ├─sys-kernel-debug.mount
●   ├─sys-kernel-tracing.mount
●   ├─systemd-journal-catalog-update.service
●   ├─systemd-journal-flush.service
●   ├─systemd-journald.service
●   ├─systemd-machine-id-commit.service
●   ├─systemd-modules-load.service
●   ├─systemd-network-generator.service
○   ├─systemd-pstore.service
●   ├─systemd-random-seed.service
○   ├─systemd-repart.service
●   ├─systemd-resolved.service
●   ├─systemd-sysctl.service
●   ├─systemd-sysusers.service
●   ├─systemd-tmpfiles-setup-dev-early.service
●   ├─systemd-tmpfiles-setup-dev.service
●   ├─systemd-tmpfiles-setup.service
●   ├─systemd-udev-load-credentials.service
●   ├─systemd-udev-trigger.service
●   ├─systemd-udevd.service
●   ├─local-fs.target
●   │ ├─\x2ebottlerocket.mount
●   │ ├─has-boot-ever-succeeded.service
●   │ ├─local.mount
●   │ ├─mnt.mount
●   │ ├─opt.mount
●   │ ├─prepare-local-fs.service
●   │ ├─prepare-opt.service
●   │ ├─prepare-var-lib-containerd.service
●   │ ├─prepare-var-lib-kubelet.service
●   │ ├─prepare-var.service
○   │ ├─repart-data-fallback.service
○   │ ├─repart-data-preferred.service
●   │ ├─repart-local.service
●   │ ├─selinux-policy-files.service
●   │ ├─systemd-remount-fs.service
●   │ ├─tmp.mount
●   │ ├─var.mount
●   │ └─x86_64\x2dbottlerocket\x2dlinux\x2dgnu-sys\x2droot-usr-lib-modules.mount
●   └─swap.target

All services that depend on [email protected] :

bash-5.1#   systemctl list-dependencies --reverse [email protected]
[email protected]
● └─rocm-k8s-device-plugin.service

Terms of contribution:

By submitting this pull request, I agree that this contribution is dual-licensed under the terms of both the Apache License, version 2.0, and the MIT license.

[mgsharm]

Make rocm-k8s-device-plugin depend on [email protected] to ensure AMD GPU driver loads first.