github.com/nicocha30/gvisor-ligolo@v0.0.0-20230726075806-989fa2c0a413/pkg/sentry/devices/nvproxy/nvproxy.go (about)

     1  // Copyright 2023 The gVisor Authors.
     2  //
     3  // Licensed under the Apache License, Version 2.0 (the "License");
     4  // you may not use this file except in compliance with the License.
     5  // You may obtain a copy of the License at
     6  //
     7  //     http://www.apache.org/licenses/LICENSE-2.0
     8  //
     9  // Unless required by applicable law or agreed to in writing, software
    10  // distributed under the License is distributed on an "AS IS" BASIS,
    11  // WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
    12  // See the License for the specific language governing permissions and
    13  // limitations under the License.
    14  
    15  // Package nvproxy implements proxying for the Nvidia GPU Linux kernel driver:
    16  // https://github.com/NVIDIA/open-gpu-kernel-modules.
    17  //
    18  // Supported Nvidia GPUs: T4, L4, A100, A10G, V100 and H100.
    19  package nvproxy
    20  
    21  import (
    22  	"fmt"
    23  
    24  	"github.com/nicocha30/gvisor-ligolo/pkg/abi/nvgpu"
    25  	"github.com/nicocha30/gvisor-ligolo/pkg/context"
    26  	"github.com/nicocha30/gvisor-ligolo/pkg/hostarch"
    27  	"github.com/nicocha30/gvisor-ligolo/pkg/log"
    28  	"github.com/nicocha30/gvisor-ligolo/pkg/marshal"
    29  	"github.com/nicocha30/gvisor-ligolo/pkg/sentry/fsimpl/devtmpfs"
    30  	"github.com/nicocha30/gvisor-ligolo/pkg/sentry/mm"
    31  	"github.com/nicocha30/gvisor-ligolo/pkg/sentry/vfs"
    32  )
    33  
    34  // Register registers all devices implemented by this package in vfsObj.
    35  func Register(vfsObj *vfs.VirtualFilesystem, uvmDevMajor uint32) error {
    36  	// The kernel driver's interface is unstable, so only allow versions of the
    37  	// driver that are known to be supported.
    38  	version, err := hostDriverVersion()
    39  	if err != nil {
    40  		return fmt.Errorf("failed to get Nvidia driver version: %w", err)
    41  	}
    42  	switch version {
    43  	case
    44  		"525.60.13",
    45  		"525.105.17":
    46  		log.Infof("Nvidia driver version: %s", version)
    47  	default:
    48  		return fmt.Errorf("unsupported Nvidia driver version: %s", version)
    49  	}
    50  
    51  	nvp := &nvproxy{
    52  		objsLive: make(map[nvgpu.Handle]*object),
    53  	}
    54  	for minor := uint32(0); minor <= nvgpu.NV_CONTROL_DEVICE_MINOR; minor++ {
    55  		if err := vfsObj.RegisterDevice(vfs.CharDevice, nvgpu.NV_MAJOR_DEVICE_NUMBER, minor, &frontendDevice{
    56  			nvp:   nvp,
    57  			minor: minor,
    58  		}, &vfs.RegisterDeviceOptions{
    59  			GroupName: "nvidia-frontend",
    60  		}); err != nil {
    61  			return err
    62  		}
    63  	}
    64  	if err := vfsObj.RegisterDevice(vfs.CharDevice, uvmDevMajor, nvgpu.NVIDIA_UVM_PRIMARY_MINOR_NUMBER, &uvmDevice{
    65  		nvp: nvp,
    66  	}, &vfs.RegisterDeviceOptions{
    67  		GroupName: "nvidia-uvm",
    68  	}); err != nil {
    69  		return err
    70  	}
    71  	return nil
    72  }
    73  
    74  // CreateDriverDevtmpfsFiles creates device special files in dev that should
    75  // always exist when this package is enabled. It does not create per-device
    76  // files in dev; see CreateIndexDevtmpfsFile.
    77  func CreateDriverDevtmpfsFiles(ctx context.Context, dev *devtmpfs.Accessor, uvmDevMajor uint32) error {
    78  	if err := dev.CreateDeviceFile(ctx, "nvidiactl", vfs.CharDevice, nvgpu.NV_MAJOR_DEVICE_NUMBER, nvgpu.NV_CONTROL_DEVICE_MINOR, 0666); err != nil {
    79  		return err
    80  	}
    81  	if err := dev.CreateDeviceFile(ctx, "nvidia-uvm", vfs.CharDevice, uvmDevMajor, nvgpu.NVIDIA_UVM_PRIMARY_MINOR_NUMBER, 0666); err != nil {
    82  		return err
    83  	}
    84  	return nil
    85  }
    86  
    87  // CreateIndexDevtmpfsFile creates the device special file in dev for the
    88  // device with the given index.
    89  func CreateIndexDevtmpfsFile(ctx context.Context, dev *devtmpfs.Accessor, index uint32) error {
    90  	return dev.CreateDeviceFile(ctx, fmt.Sprintf("nvidia%d", index), vfs.CharDevice, nvgpu.NV_MAJOR_DEVICE_NUMBER, index, 0666)
    91  }
    92  
    93  // +stateify savable
    94  type nvproxy struct {
    95  	objsMu   objsMutex
    96  	objsLive map[nvgpu.Handle]*object
    97  }
    98  
    99  // object tracks an object allocated through the driver.
   100  //
   101  // +stateify savable
   102  type object struct {
   103  	impl objectImpl
   104  }
   105  
   106  func (o *object) init(impl objectImpl) {
   107  	o.impl = impl
   108  }
   109  
   110  // Release is called after the represented object is freed.
   111  func (o *object) Release(ctx context.Context) {
   112  	o.impl.Release(ctx)
   113  }
   114  
   115  type objectImpl interface {
   116  	Release(ctx context.Context)
   117  }
   118  
   119  // osDescMem is an objectImpl tracking an OS descriptor.
   120  //
   121  // +stateify savable
   122  type osDescMem struct {
   123  	object
   124  	pinnedRanges []mm.PinnedRange
   125  }
   126  
   127  // Release implements objectImpl.Release.
   128  func (o *osDescMem) Release(ctx context.Context) {
   129  	ctx.Infof("nvproxy: unpinning pages for released OS descriptor")
   130  	mm.Unpin(o.pinnedRanges)
   131  }
   132  
   133  type marshalPtr[T any] interface {
   134  	*T
   135  	marshal.Marshallable
   136  }
   137  
   138  func addrFromP64(p nvgpu.P64) hostarch.Addr {
   139  	return hostarch.Addr(uintptr(uint64(p)))
   140  }