|
| 1 | +package daemon |
| 2 | + |
| 3 | +import ( |
| 4 | + "os/exec" |
| 5 | + "strconv" |
| 6 | + |
| 7 | + "github.com/containerd/containerd/contrib/nvidia" |
| 8 | + "github.com/docker/docker/pkg/capabilities" |
| 9 | + "github.com/opencontainers/runtime-spec/specs-go" |
| 10 | + "github.com/pkg/errors" |
| 11 | +) |
| 12 | + |
| 13 | +// TODO: nvidia should not be hard-coded, and should be a device plugin instead on the daemon object. |
| 14 | +// TODO: add list of device capabilities in daemon/node info |
| 15 | + |
| 16 | +var errConflictCountDeviceIDs = errors.New("cannot set both Count and DeviceIDs on device request") |
| 17 | + |
| 18 | +// stolen from github.com/containerd/containerd/contrib/nvidia |
| 19 | +const nvidiaCLI = "nvidia-container-cli" |
| 20 | + |
| 21 | +// These are NVIDIA-specific capabilities stolen from github.com/containerd/containerd/contrib/nvidia.allCaps |
| 22 | +var allNvidiaCaps = map[nvidia.Capability]struct{}{ |
| 23 | + nvidia.Compute: {}, |
| 24 | + nvidia.Compat32: {}, |
| 25 | + nvidia.Graphics: {}, |
| 26 | + nvidia.Utility: {}, |
| 27 | + nvidia.Video: {}, |
| 28 | + nvidia.Display: {}, |
| 29 | +} |
| 30 | + |
| 31 | +func init() { |
| 32 | + if _, err := exec.LookPath(nvidiaCLI); err != nil { |
| 33 | + // do not register Nvidia driver if helper binary is not present. |
| 34 | + return |
| 35 | + } |
| 36 | + capset := capabilities.Set{"gpu": struct{}{}, "nvidia": struct{}{}} |
| 37 | + nvidiaDriver := &deviceDriver{ |
| 38 | + capset: capset, |
| 39 | + updateSpec: setNvidiaGPUs, |
| 40 | + } |
| 41 | + for c := range capset { |
| 42 | + nvidiaDriver.capset[c] = struct{}{} |
| 43 | + } |
| 44 | + registerDeviceDriver("nvidia", nvidiaDriver) |
| 45 | +} |
| 46 | + |
| 47 | +func setNvidiaGPUs(s *specs.Spec, dev *deviceInstance) error { |
| 48 | + var opts []nvidia.Opts |
| 49 | + |
| 50 | + req := dev.req |
| 51 | + if req.Count != 0 && len(req.DeviceIDs) > 0 { |
| 52 | + return errConflictCountDeviceIDs |
| 53 | + } |
| 54 | + |
| 55 | + if len(req.DeviceIDs) > 0 { |
| 56 | + var ids []int |
| 57 | + var uuids []string |
| 58 | + for _, devID := range req.DeviceIDs { |
| 59 | + id, err := strconv.Atoi(devID) |
| 60 | + if err == nil { |
| 61 | + ids = append(ids, id) |
| 62 | + continue |
| 63 | + } |
| 64 | + // if not an integer, then assume UUID. |
| 65 | + uuids = append(uuids, devID) |
| 66 | + } |
| 67 | + if len(ids) > 0 { |
| 68 | + opts = append(opts, nvidia.WithDevices(ids...)) |
| 69 | + } |
| 70 | + if len(uuids) > 0 { |
| 71 | + opts = append(opts, nvidia.WithDeviceUUIDs(uuids...)) |
| 72 | + } |
| 73 | + } |
| 74 | + |
| 75 | + if req.Count < 0 { |
| 76 | + opts = append(opts, nvidia.WithAllDevices) |
| 77 | + } else if req.Count > 0 { |
| 78 | + opts = append(opts, nvidia.WithDevices(countToDevices(req.Count)...)) |
| 79 | + } |
| 80 | + |
| 81 | + var nvidiaCaps []nvidia.Capability |
| 82 | + // req.Capabilities contains device capabilities, some but not all are NVIDIA driver capabilities. |
| 83 | + for _, c := range dev.selectedCaps { |
| 84 | + nvcap := nvidia.Capability(c) |
| 85 | + if _, isNvidiaCap := allNvidiaCaps[nvcap]; isNvidiaCap { |
| 86 | + nvidiaCaps = append(nvidiaCaps, nvcap) |
| 87 | + continue |
| 88 | + } |
| 89 | + // TODO: nvidia.WithRequiredCUDAVersion |
| 90 | + // for now we let the prestart hook verify cuda versions but errors are not pretty. |
| 91 | + } |
| 92 | + |
| 93 | + if nvidiaCaps != nil { |
| 94 | + opts = append(opts, nvidia.WithCapabilities(nvidiaCaps...)) |
| 95 | + } |
| 96 | + |
| 97 | + return nvidia.WithGPUs(opts...)(nil, nil, nil, s) |
| 98 | +} |
| 99 | + |
| 100 | +// countToDevices returns the list 0, 1, ... count-1 of deviceIDs. |
| 101 | +func countToDevices(count int) []int { |
| 102 | + devices := make([]int, count) |
| 103 | + for i := range devices { |
| 104 | + devices[i] = i |
| 105 | + } |
| 106 | + return devices |
| 107 | +} |
0 commit comments