kernfs

package
v0.0.0-...-f1656df Latest Latest
Warning

This package is not in the latest version of its module.

 Go to latest Published: Dec 19, 2024 License: Apache-2.0, MIT Imports: 19 Imported by: 31

Documentation

Overview

Package kernfs provides the tools to implement inode-based filesystems. Kernfs has two main features:

  1. The Inode interface, which maps VFS's path-based filesystem operations to specific filesystem nodes. Kernfs uses the Inode interface to provide a blanket implementation for the vfs.FilesystemImpl. Kernfs also serves as the synchronization mechanism for all filesystem operations by holding a filesystem-wide lock across all operations.

  2. Various utility types which provide generic implementations for various parts of the Inode and vfs.FileDescription interfaces. Client filesystems based on kernfs can embed the appropriate set of these to avoid having to reimplement common filesystem operations. See inode_impl_util.go and fd_impl_util.go.

Reference Model:

Kernfs dentries represents named pointers to inodes. Kernfs is solely responsible for maintaining and modifying its dentry tree; inode implementations can not access the tree. Dentries and inodes have independent lifetimes and reference counts. A child dentry unconditionally holds a reference on its parent directory's dentry. A dentry also holds a reference on the inode it points to (although that might not be the only reference on the inode). Due to this inodes can outlive the dentries that point to them. Multiple dentries can point to the same inode (for example, in the case of hardlinks). File descriptors hold a reference to the dentry they're opened on.

Dentries are guaranteed to exist while holding Filesystem.mu for reading. Dropping dentries require holding Filesystem.mu for writing. To queue dentries for destruction from a read critical section, see Filesystem.deferDecRef.

Lock ordering: 

kernfs.Filesystem.mu
	kernel.TaskSet.mu
  	kernel.Task.mu
	kernfs.Dentry.dirMu
  	vfs.VirtualFilesystem.mountMu
    	vfs.Dentry.mu
	(inode implementation locks, if any)

kernfs.Filesystem.deferredDecRefsMu

Index

Constants

View Source 
const (
	// Consider the end of the file to be after the final static entry. This is
	// the default option.
	SeekEndStaticEntries = iota
	// Consider the end of the file to be at offset 0.
	SeekEndZero
)

Constants related to SEEK_END behaviour for FDs.

Variables

This section is empty.

Functions

This section is empty.

Types

type CachedMappable 

type CachedMappable struct {
	// contains filtered or unexported fields
}

CachedMappable implements memmap.Mappable. This utility can be embedded in a kernfs.Inode that represents a host file to make the inode mappable. CachedMappable caches the mappings of the host file. CachedMappable must be initialized (via Init) with a hostFD before use.

+stateify savable

func (*CachedMappable) AddMapping 

func (i *CachedMappable) AddMapping(ctx context.Context, ms memmap.MappingSpace, ar hostarch.AddrRange, offset uint64, writable bool) error

AddMapping implements memmap.Mappable.AddMapping.

func (*CachedMappable) CopyMapping 

func (i *CachedMappable) CopyMapping(ctx context.Context, ms memmap.MappingSpace, srcAR, dstAR hostarch.AddrRange, offset uint64, writable bool) error

CopyMapping implements memmap.Mappable.CopyMapping.

func (*CachedMappable) Init 

func (i *CachedMappable) Init(hostFD int)

Init initializes i.pf. This must be called before using CachedMappable.

func (*CachedMappable) InitFileMapperOnce 

func (i *CachedMappable) InitFileMapperOnce()

InitFileMapperOnce initializes the host file mapper. It ensures that the file mapper is initialized just once.

func (*CachedMappable) InvalidateRange 

func (i *CachedMappable) InvalidateRange(r memmap.MappableRange)

InvalidateRange invalidates the passed range on i.mappings.

func (*CachedMappable) InvalidateUnsavable 

func (i *CachedMappable) InvalidateUnsavable(ctx context.Context) error

InvalidateUnsavable implements memmap.Mappable.InvalidateUnsavable.

func (*CachedMappable) RemoveMapping 

func (i *CachedMappable) RemoveMapping(ctx context.Context, ms memmap.MappingSpace, ar hostarch.AddrRange, offset uint64, writable bool)

RemoveMapping implements memmap.Mappable.RemoveMapping.

func (*CachedMappable) Translate 

func (i *CachedMappable) Translate(ctx context.Context, required, optional memmap.MappableRange, at hostarch.AccessType) ([]memmap.Translation, error)

Translate implements memmap.Mappable.Translate.

type Dentry 

type Dentry struct {
	// contains filtered or unexported fields
}

Dentry implements vfs.DentryImpl.

A kernfs dentry is similar to a dentry in a traditional filesystem: it's a named reference to an inode. A dentry generally lives as long as it's part of a mounted filesystem tree. Kernfs drops dentries once all references to them are dropped. Dentries hold a single reference to the inode they point to, and child dentries hold a reference on their parent.

Must be initialized by Init prior to first use.

+stateify savable

func (*Dentry) DecRef 

func (d *Dentry) DecRef(ctx context.Context)

DecRef implements vfs.DentryImpl.DecRef.

func (*Dentry) FSLocalPath 

func (d *Dentry) FSLocalPath() string

FSLocalPath returns an absolute path to d, relative to the root of its filesystem.

func (*Dentry) IncRef 

func (d *Dentry) IncRef()

IncRef implements vfs.DentryImpl.IncRef.

func (*Dentry) Init 

func (d *Dentry) Init(fs *Filesystem, inode Inode)

Init initializes this dentry.

Precondition: Caller must hold a reference on inode.

Postcondition: Caller's reference on inode is transferred to the dentry.

func (*Dentry) InitRoot 

func (d *Dentry) InitRoot(fs *Filesystem, inode Inode)

InitRoot initializes this dentry as the root of the filesystem.

Precondition: Caller must hold a reference on inode.

Postcondition: Caller's reference on inode is transferred to the dentry.

func (*Dentry) Inode 

func (d *Dentry) Inode() Inode

Inode returns the dentry's inode.

func (*Dentry) InotifyWithParent 

func (d *Dentry) InotifyWithParent(ctx context.Context, events, cookie uint32, et vfs.EventType)

InotifyWithParent implements vfs.DentryImpl.InotifyWithParent.

func (*Dentry) Invalidate 

func (d *Dentry) Invalidate(ctx context.Context)

Invalidate invalidates the dentry and its children.

func (*Dentry) LeakMessage 

func (d *Dentry) LeakMessage() string

LeakMessage implements refs.CheckedObject.LeakMessage.

func (*Dentry) LogRefs 

func (d *Dentry) LogRefs() bool

LogRefs implements refs.CheckedObject.LogRefs.

This should only be set to true for debugging purposes, as it can generate an extremely large amount of output and drastically degrade performance.

func (*Dentry) OnZeroWatches 

func (d *Dentry) OnZeroWatches(context.Context)

OnZeroWatches implements vfs.Dentry.OnZeroWatches.

func (*Dentry) Parent 

func (d *Dentry) Parent() *Dentry

Parent returns the parent of this Dentry. This is not safe in general, the filesystem may concurrently move d elsewhere. The caller is responsible for ensuring the returned result remains valid while it is used.

func (*Dentry) RefType 

func (d *Dentry) RefType() string

RefType implements refs.CheckedObject.Type.

func (*Dentry) TryIncRef 

func (d *Dentry) TryIncRef() bool

TryIncRef implements vfs.DentryImpl.TryIncRef.

func (*Dentry) VFSDentry 

func (d *Dentry) VFSDentry() *vfs.Dentry

VFSDentry returns the generic vfs dentry for this kernfs dentry.

func (*Dentry) WalkDentryTree 

func (d *Dentry) WalkDentryTree(ctx context.Context, vfsObj *vfs.VirtualFilesystem, p fspath.Path) (*Dentry, error)

WalkDentryTree traverses p in the dentry tree for this filesystem. Note that this only traverses the dentry tree and is not a general path traversal. No symlinks and dynamic children are resolved, and no permission checks are performed. The caller is responsible for ensuring the returned Dentry exists for an appropriate lifetime.

p is interpreted starting at d, and may be absolute or relative (absolute vs relative paths both refer to the same target here, since p is absolute from d). p may contain "." and "..", but will not allow traversal above d (similar to ".." at the root dentry).

This is useful for filesystem internals, where the filesystem may not be mounted yet. For a mounted filesystem, use GetDentryAt.

func (*Dentry) Watches 

func (d *Dentry) Watches() *vfs.Watches

Watches implements vfs.DentryImpl.Watches.

type DynamicBytesFD 

type DynamicBytesFD struct {
	vfs.FileDescriptionDefaultImpl
	vfs.DynamicBytesFileDescriptionImpl
	vfs.LockFD
	// contains filtered or unexported fields
}

DynamicBytesFD implements vfs.FileDescriptionImpl for an FD backed by a DynamicBytesFile.

Must be initialized with Init before first use.

+stateify savable

func (*DynamicBytesFD) Init 

func (fd *DynamicBytesFD) Init(m *vfs.Mount, d *Dentry, data vfs.DynamicBytesSource, locks *vfs.FileLocks, flags uint32) error

Init initializes a DynamicBytesFD.

func (*DynamicBytesFD) PRead 

func (fd *DynamicBytesFD) PRead(ctx context.Context, dst usermem.IOSequence, offset int64, opts vfs.ReadOptions) (int64, error)

PRead implements vfs.FileDescriptionImpl.PRead.

func (*DynamicBytesFD) PWrite 

func (fd *DynamicBytesFD) PWrite(ctx context.Context, src usermem.IOSequence, offset int64, opts vfs.WriteOptions) (int64, error)

PWrite implements vfs.FileDescriptionImpl.PWrite.

func (*DynamicBytesFD) Read 

func (fd *DynamicBytesFD) Read(ctx context.Context, dst usermem.IOSequence, opts vfs.ReadOptions) (int64, error)

Read implements vfs.FileDescriptionImpl.Read.

func (*DynamicBytesFD) Release 

func (fd *DynamicBytesFD) Release(context.Context)

Release implements vfs.FileDescriptionImpl.Release.

func (*DynamicBytesFD) Seek 

func (fd *DynamicBytesFD) Seek(ctx context.Context, offset int64, whence int32) (int64, error)

Seek implements vfs.FileDescriptionImpl.Seek.

func (*DynamicBytesFD) SetStat 

func (fd *DynamicBytesFD) SetStat(context.Context, vfs.SetStatOptions) error

SetStat implements vfs.FileDescriptionImpl.SetStat.

func (*DynamicBytesFD) Stat 

func (fd *DynamicBytesFD) Stat(ctx context.Context, opts vfs.StatOptions) (linux.Statx, error)

Stat implements vfs.FileDescriptionImpl.Stat.

func (*DynamicBytesFD) Write 

func (fd *DynamicBytesFD) Write(ctx context.Context, src usermem.IOSequence, opts vfs.WriteOptions) (int64, error)

Write implements vfs.FileDescriptionImpl.Write.

type DynamicBytesFile 

type DynamicBytesFile struct {
	InodeAttrs
	InodeNoStatFS
	InodeNoopRefCount
	InodeNotAnonymous
	InodeNotDirectory
	InodeNotSymlink
	InodeWatches
	InodeFSOwned
	// contains filtered or unexported fields
}

DynamicBytesFile implements kernfs.Inode and represents a read-only file whose contents are backed by a vfs.DynamicBytesSource. If data additionally implements vfs.WritableDynamicBytesSource, the file also supports dispatching writes to the implementer, but note that this will not update the source data.

Must be instantiated with NewDynamicBytesFile or initialized with Init before first use.

+stateify savable

func (*DynamicBytesFile) Data 

func (f *DynamicBytesFile) Data() vfs.DynamicBytesSource

Data returns the underlying data source.

func (*DynamicBytesFile) Init 

func (f *DynamicBytesFile) Init(ctx context.Context, creds *auth.Credentials, devMajor, devMinor uint32, ino uint64, data vfs.DynamicBytesSource, perm linux.FileMode)

Init initializes a dynamic bytes file.

func (*DynamicBytesFile) Locks 

func (f *DynamicBytesFile) Locks() *vfs.FileLocks

Locks returns the file locks for this file.

func (*DynamicBytesFile) Open 

func (f *DynamicBytesFile) Open(ctx context.Context, rp *vfs.ResolvingPath, d *Dentry, opts vfs.OpenOptions) (*vfs.FileDescription, error)

Open implements Inode.Open.

func (*DynamicBytesFile) SetStat 

func (*DynamicBytesFile) SetStat(context.Context, *vfs.Filesystem, *auth.Credentials, vfs.SetStatOptions) error

SetStat implements Inode.SetStat. By default DynamicBytesFile doesn't allow inode attributes to be changed. Override SetStat() making it call f.InodeAttrs to allow it.

type Filesystem 

type Filesystem struct {

	// MaxCachedDentries is the maximum size of cachedDentries. If not set,
	// defaults to 0 and kernfs does not cache any dentries. This is immutable.
	MaxCachedDentries uint64
	// contains filtered or unexported fields
}

Filesystem mostly implements vfs.FilesystemImpl for a generic in-memory filesystem. Concrete implementations are expected to embed this in their own Filesystem type.

+stateify savable

func (*Filesystem) AccessAt 

func (fs *Filesystem) AccessAt(ctx context.Context, rp *vfs.ResolvingPath, creds *auth.Credentials, ats vfs.AccessTypes) error

AccessAt implements vfs.Filesystem.Impl.AccessAt.

func (*Filesystem) BoundEndpointAt 

func (fs *Filesystem) BoundEndpointAt(ctx context.Context, rp *vfs.ResolvingPath, opts vfs.BoundEndpointOptions) (transport.BoundEndpoint, error)

BoundEndpointAt implements vfs.FilesystemImpl.BoundEndpointAt.

func (*Filesystem) GetDentryAt 

func (fs *Filesystem) GetDentryAt(ctx context.Context, rp *vfs.ResolvingPath, opts vfs.GetDentryOptions) (*vfs.Dentry, error)

GetDentryAt implements vfs.FilesystemImpl.GetDentryAt.

func (*Filesystem) GetParentDentryAt 

func (fs *Filesystem) GetParentDentryAt(ctx context.Context, rp *vfs.ResolvingPath) (*vfs.Dentry, error)

GetParentDentryAt implements vfs.FilesystemImpl.GetParentDentryAt.

func (*Filesystem) GetXattrAt 

func (fs *Filesystem) GetXattrAt(ctx context.Context, rp *vfs.ResolvingPath, opts vfs.GetXattrOptions) (string, error)

GetXattrAt implements vfs.FilesystemImpl.GetXattrAt.

func (*Filesystem) IsDescendant 

func (fs *Filesystem) IsDescendant(vfsroot, vd vfs.VirtualDentry) bool

IsDescendant implements vfs.FilesystemImpl.IsDescendant.

func (*Filesystem) LinkAt 

func (fs *Filesystem) LinkAt(ctx context.Context, rp *vfs.ResolvingPath, vd vfs.VirtualDentry) error

LinkAt implements vfs.FilesystemImpl.LinkAt.

func (*Filesystem) ListXattrAt 

func (fs *Filesystem) ListXattrAt(ctx context.Context, rp *vfs.ResolvingPath, size uint64) ([]string, error)

ListXattrAt implements vfs.FilesystemImpl.ListXattrAt.

func (*Filesystem) MkdirAt 

func (fs *Filesystem) MkdirAt(ctx context.Context, rp *vfs.ResolvingPath, opts vfs.MkdirOptions) error

MkdirAt implements vfs.FilesystemImpl.MkdirAt.

func (*Filesystem) MknodAt 

func (fs *Filesystem) MknodAt(ctx context.Context, rp *vfs.ResolvingPath, opts vfs.MknodOptions) error

MknodAt implements vfs.FilesystemImpl.MknodAt.

func (*Filesystem) NextIno 

func (fs *Filesystem) NextIno() uint64

NextIno allocates a new inode number on this filesystem.

func (*Filesystem) OpenAt 

func (fs *Filesystem) OpenAt(ctx context.Context, rp *vfs.ResolvingPath, opts vfs.OpenOptions) (*vfs.FileDescription, error)

OpenAt implements vfs.FilesystemImpl.OpenAt.

func (*Filesystem) PrependPath 

func (fs *Filesystem) PrependPath(ctx context.Context, vfsroot, vd vfs.VirtualDentry, b *fspath.Builder) error

PrependPath implements vfs.FilesystemImpl.PrependPath.

func (*Filesystem) ReadlinkAt 

func (fs *Filesystem) ReadlinkAt(ctx context.Context, rp *vfs.ResolvingPath) (string, error)

ReadlinkAt implements vfs.FilesystemImpl.ReadlinkAt.

func (*Filesystem) Release 

func (fs *Filesystem) Release(ctx context.Context)

Release implements vfs.FilesystemImpl.Release.

func (*Filesystem) RemoveXattrAt 

func (fs *Filesystem) RemoveXattrAt(ctx context.Context, rp *vfs.ResolvingPath, name string) error

RemoveXattrAt implements vfs.FilesystemImpl.RemoveXattrAt.

func (*Filesystem) RenameAt 

func (fs *Filesystem) RenameAt(ctx context.Context, rp *vfs.ResolvingPath, oldParentVD vfs.VirtualDentry, oldName string, opts vfs.RenameOptions) error

RenameAt implements vfs.FilesystemImpl.RenameAt.

func (*Filesystem) RmdirAt 

func (fs *Filesystem) RmdirAt(ctx context.Context, rp *vfs.ResolvingPath) error

RmdirAt implements vfs.FilesystemImpl.RmdirAt.

func (*Filesystem) SafeDecRef 

func (fs *Filesystem) SafeDecRef(ctx context.Context, vd vfs.VirtualDentry)

SafeDecRef safely DecRef the virtual dentry making sure DecRef is deferred in case Filesystem.mu is held. See comment on Filesystem.mu.

func (*Filesystem) SafeDecRefFD 

func (fs *Filesystem) SafeDecRefFD(ctx context.Context, fd *vfs.FileDescription)

SafeDecRefFD safely DecRef the FileDescription making sure DecRef is deferred in case Filesystem.mu is held. See comment on Filesystem.mu.

func (*Filesystem) SetStatAt 

func (fs *Filesystem) SetStatAt(ctx context.Context, rp *vfs.ResolvingPath, opts vfs.SetStatOptions) error

SetStatAt implements vfs.FilesystemImpl.SetStatAt.

func (*Filesystem) SetXattrAt 

func (fs *Filesystem) SetXattrAt(ctx context.Context, rp *vfs.ResolvingPath, opts vfs.SetXattrOptions) error

SetXattrAt implements vfs.FilesystemImpl.SetXattrAt.

func (*Filesystem) StatAt 

func (fs *Filesystem) StatAt(ctx context.Context, rp *vfs.ResolvingPath, opts vfs.StatOptions) (linux.Statx, error)

StatAt implements vfs.FilesystemImpl.StatAt.

func (*Filesystem) StatFSAt 

func (fs *Filesystem) StatFSAt(ctx context.Context, rp *vfs.ResolvingPath) (linux.Statfs, error)

StatFSAt implements vfs.FilesystemImpl.StatFSAt.

func (*Filesystem) SymlinkAt 

func (fs *Filesystem) SymlinkAt(ctx context.Context, rp *vfs.ResolvingPath, target string) error

SymlinkAt implements vfs.FilesystemImpl.SymlinkAt.

func (*Filesystem) Sync 

func (fs *Filesystem) Sync(ctx context.Context) error

Sync implements vfs.FilesystemImpl.Sync.

func (*Filesystem) UnlinkAt 

func (fs *Filesystem) UnlinkAt(ctx context.Context, rp *vfs.ResolvingPath) error

UnlinkAt implements vfs.FilesystemImpl.UnlinkAt.

func (*Filesystem) VFSFilesystem 

func (fs *Filesystem) VFSFilesystem() *vfs.Filesystem

VFSFilesystem returns the generic vfs filesystem object.

type GenericDirectoryFD 

type GenericDirectoryFD struct {
	vfs.FileDescriptionDefaultImpl
	vfs.DirectoryFileDescriptionDefaultImpl
	vfs.LockFD
	// contains filtered or unexported fields
}

GenericDirectoryFD implements vfs.FileDescriptionImpl for a generic directory inode that uses OrderChildren to track child nodes.

Note that GenericDirectoryFD holds a lock over OrderedChildren while calling IterDirents callback. The IterDirents callback therefore cannot hash or unhash children, or recursively call IterDirents on the same underlying inode.

Must be initialize with Init before first use.

Lock ordering: mu => children.mu.

+stateify savable

func NewGenericDirectoryFD 

func NewGenericDirectoryFD(m *vfs.Mount, d *Dentry, children *OrderedChildren, locks *vfs.FileLocks, opts *vfs.OpenOptions, fdOpts GenericDirectoryFDOptions) (*GenericDirectoryFD, error)

NewGenericDirectoryFD creates a new GenericDirectoryFD and returns its dentry.

func (*GenericDirectoryFD) Allocate 

func (fd *GenericDirectoryFD) Allocate(ctx context.Context, mode, offset, length uint64) error

Allocate implements vfs.FileDescriptionImpl.Allocate.

func (*GenericDirectoryFD) ConfigureMMap 

func (fd *GenericDirectoryFD) ConfigureMMap(ctx context.Context, opts *memmap.MMapOpts) error

ConfigureMMap implements vfs.FileDescriptionImpl.ConfigureMMap.

func (*GenericDirectoryFD) Init 

func (fd *GenericDirectoryFD) Init(children *OrderedChildren, locks *vfs.FileLocks, opts *vfs.OpenOptions, fdOpts GenericDirectoryFDOptions) error

Init initializes a GenericDirectoryFD. Use it when overriding GenericDirectoryFD. Caller must call fd.VFSFileDescription.Init() with the correct implementation.

func (*GenericDirectoryFD) IterDirents 

func (fd *GenericDirectoryFD) IterDirents(ctx context.Context, cb vfs.IterDirentsCallback) error

IterDirents implements vfs.FileDescriptionImpl.IterDirents. IterDirents holds o.mu when calling cb.

func (*GenericDirectoryFD) PRead 

func (fd *GenericDirectoryFD) PRead(ctx context.Context, dst usermem.IOSequence, offset int64, opts vfs.ReadOptions) (int64, error)

PRead implements vfs.FileDescriptionImpl.PRead.

func (*GenericDirectoryFD) PWrite 

func (fd *GenericDirectoryFD) PWrite(ctx context.Context, src usermem.IOSequence, offset int64, opts vfs.WriteOptions) (int64, error)

PWrite implements vfs.FileDescriptionImpl.PWrite.

func (*GenericDirectoryFD) Read 

func (fd *GenericDirectoryFD) Read(ctx context.Context, dst usermem.IOSequence, opts vfs.ReadOptions) (int64, error)

Read implements vfs.FileDescriptionImpl.Read.

func (*GenericDirectoryFD) Release 

func (fd *GenericDirectoryFD) Release(context.Context)

Release implements vfs.FileDescriptionImpl.Release.

func (*GenericDirectoryFD) Seek 

func (fd *GenericDirectoryFD) Seek(ctx context.Context, offset int64, whence int32) (int64, error)

Seek implements vfs.FileDescriptionImpl.Seek.

func (*GenericDirectoryFD) SetStat 

func (fd *GenericDirectoryFD) SetStat(ctx context.Context, opts vfs.SetStatOptions) error

SetStat implements vfs.FileDescriptionImpl.SetStat.

func (*GenericDirectoryFD) Stat 

func (fd *GenericDirectoryFD) Stat(ctx context.Context, opts vfs.StatOptions) (linux.Statx, error)

Stat implements vfs.FileDescriptionImpl.Stat.

func (*GenericDirectoryFD) VFSFileDescription 

func (fd *GenericDirectoryFD) VFSFileDescription() *vfs.FileDescription

VFSFileDescription returns a pointer to the vfs.FileDescription representing this object.

func (*GenericDirectoryFD) Write 

func (fd *GenericDirectoryFD) Write(ctx context.Context, src usermem.IOSequence, opts vfs.WriteOptions) (int64, error)

Write implements vfs.FileDescriptionImpl.Write.

type GenericDirectoryFDOptions 

type GenericDirectoryFDOptions struct {
	SeekEnd SeekEndConfig
}

GenericDirectoryFDOptions contains configuration for a GenericDirectoryFD.

+stateify savable

type Inode 

type Inode interface {

	// Open creates a file description for the filesystem object represented by
	// this inode. The returned file description should hold a reference on the
	// dentry for its lifetime.
	//
	// Precondition: rp.Done(). vfsd.Impl() must be the kernfs Dentry containing
	// the inode on which Open() is being called.
	Open(ctx context.Context, rp *vfs.ResolvingPath, d *Dentry, opts vfs.OpenOptions) (*vfs.FileDescription, error)

	// StatFS returns filesystem statistics for the client filesystem. This
	// corresponds to vfs.FilesystemImpl.StatFSAt. If the client filesystem
	// doesn't support statfs(2), this should return ENOSYS.
	StatFS(ctx context.Context, fs *vfs.Filesystem) (linux.Statfs, error)

	// Keep indicates whether the dentry created after Inode.Lookup should be
	// kept in the kernfs dentry tree.
	Keep() bool

	// Valid should return true if this inode is still valid, or needs to
	// be resolved again by a call to Lookup.
	Valid(ctx context.Context, parent *Dentry, name string) bool

	// Watches returns the set of inotify watches associated with this inode.
	Watches() *vfs.Watches

	// Anonymous indicates that the Inode is anonymous. It will never have
	// a name or parent.
	Anonymous() bool

	// RegisterDentry is called when a new dentry representing the inode is
	// created.
	RegisterDentry(d *Dentry)

	// UnregisterDentry is called when the specified dentry is destroyed.
	UnregisterDentry(d *Dentry)
	// contains filtered or unexported methods
}

The Inode interface maps filesystem-level operations that operate on paths to equivalent operations on specific filesystem nodes.

The interface methods are groups into logical categories as sub interfaces below. Generally, an implementation for each sub interface can be provided by embedding an appropriate type from inode_impl_utils.go. The sub interfaces are purely organizational. Methods declared directly in the main interface have no generic implementations, and should be explicitly provided by the client filesystem.

Generally, implementations are not responsible for tasks that are common to all filesystems. These include:

  • Checking that dentries passed to methods are of the appropriate file type.
  • Checking permissions.

Inode functions may be called holding filesystem wide locks and are not allowed to call vfs functions that may reenter, unless otherwise noted.

Specific responsibilities of implementations are documented below.

func NewStaticDir 

func NewStaticDir(ctx context.Context, creds *auth.Credentials, devMajor, devMinor uint32, ino uint64, perm linux.FileMode, children map[string]Inode, fdOpts GenericDirectoryFDOptions) Inode

NewStaticDir creates a new static directory and returns its dentry. 

func NewStaticSymlink(ctx context.Context, creds *auth.Credentials, devMajor, devMinor uint32, ino uint64, target string) Inode

NewStaticSymlink creates a new symlink file pointing to 'target'.

type InodeAlwaysValid 

type InodeAlwaysValid struct{}

InodeAlwaysValid partially implements Inode.

+stateify savable

func (*InodeAlwaysValid) Valid 

func (*InodeAlwaysValid) Valid(context.Context, *Dentry, string) bool

Valid implements Inode.Valid.

type InodeAnonymous 

type InodeAnonymous struct{}

InodeAnonymous partially implements Inode.

+stateify savable

func (*InodeAnonymous) Anonymous 

func (*InodeAnonymous) Anonymous() bool

Anonymous implements Inode.Anonymous

type InodeAttrs 

type InodeAttrs struct {
	// contains filtered or unexported fields
}

InodeAttrs partially implements the Inode interface, specifically the inodeMetadata sub interface. InodeAttrs provides functionality related to inode attributes.

Must be initialized by Init prior to first use.

+stateify savable

func (*InodeAttrs) CheckPermissions 

func (a *InodeAttrs) CheckPermissions(_ context.Context, creds *auth.Credentials, ats vfs.AccessTypes) error

CheckPermissions implements Inode.CheckPermissions. 

func (a *InodeAttrs) DecLinks()

DecLinks implements Inode.DecLinks.

func (*InodeAttrs) DevMajor 

func (a *InodeAttrs) DevMajor() uint32

DevMajor returns the device major number.

func (*InodeAttrs) DevMinor 

func (a *InodeAttrs) DevMinor() uint32

DevMinor returns the device minor number.

func (*InodeAttrs) GID 

func (a *InodeAttrs) GID() auth.KGID

GID implements Inode.GID. 

func (a *InodeAttrs) IncLinks(n uint32)

IncLinks implements Inode.IncLinks.

func (*InodeAttrs) Init 

func (a *InodeAttrs) Init(ctx context.Context, creds *auth.Credentials, devMajor, devMinor uint32, ino uint64, mode linux.FileMode)

Init initializes this InodeAttrs.

func (*InodeAttrs) InitWithIDs 

func (a *InodeAttrs) InitWithIDs(ctx context.Context, uid auth.KUID, gid auth.KGID, devMajor, devMinor uint32, ino uint64, mode linux.FileMode)

InitWithIDs initializes this InodeAttrs.

func (*InodeAttrs) Ino 

func (a *InodeAttrs) Ino() uint64

Ino returns the inode id. 

func (a *InodeAttrs) Links() uint32

Links returns the link count.

func (*InodeAttrs) Mode 

func (a *InodeAttrs) Mode() linux.FileMode

Mode implements Inode.Mode.

func (*InodeAttrs) SetStat 

func (a *InodeAttrs) SetStat(ctx context.Context, fs *vfs.Filesystem, creds *auth.Credentials, opts vfs.SetStatOptions) error

SetStat implements Inode.SetStat.

func (*InodeAttrs) Stat 

func (a *InodeAttrs) Stat(context.Context, *vfs.Filesystem, vfs.StatOptions) (linux.Statx, error)

Stat partially implements Inode.Stat. Note that this function doesn't provide all the stat fields, and the embedder should consider extending the result with filesystem-specific fields.

func (*InodeAttrs) TouchAtime 

func (a *InodeAttrs) TouchAtime(ctx context.Context, mnt *vfs.Mount)

TouchAtime updates a.atime to the current time.

func (*InodeAttrs) TouchCMtime 

func (a *InodeAttrs) TouchCMtime(ctx context.Context)

TouchCMtime updates a.{c/m}time to the current time. The caller should synchronize calls to this so that ctime and mtime are updated to the same value.

func (*InodeAttrs) UID 

func (a *InodeAttrs) UID() auth.KUID

UID implements Inode.UID.

type InodeDirectoryNoNewChildren 

type InodeDirectoryNoNewChildren struct{}

InodeDirectoryNoNewChildren partially implements the Inode interface. InodeDirectoryNoNewChildren represents a directory inode which does not support creation of new children.

+stateify savable

func (InodeDirectoryNoNewChildren) NewDir 

func (InodeDirectoryNoNewChildren) NewDir(context.Context, string, vfs.MkdirOptions) (Inode, error)

NewDir implements Inode.NewDir.

func (InodeDirectoryNoNewChildren) NewFile 

func (InodeDirectoryNoNewChildren) NewFile(context.Context, string, vfs.OpenOptions) (Inode, error)

NewFile implements Inode.NewFile. 

func (InodeDirectoryNoNewChildren) NewLink(context.Context, string, Inode) (Inode, error)

NewLink implements Inode.NewLink.

func (InodeDirectoryNoNewChildren) NewNode 

func (InodeDirectoryNoNewChildren) NewNode(context.Context, string, vfs.MknodOptions) (Inode, error)

NewNode implements Inode.NewNode. 

func (InodeDirectoryNoNewChildren) NewSymlink(context.Context, string, string) (Inode, error)

NewSymlink implements Inode.NewSymlink.

type InodeFSOwned 

type InodeFSOwned struct{}

InodeFSOwned represents inodes whose lifecycle is entirely managed by the filesystem.

+stateify savable

func (*InodeFSOwned) RegisterDentry 

func (*InodeFSOwned) RegisterDentry(d *Dentry)

RegisterDentry implements Inode.RegisterDentry.

func (*InodeFSOwned) UnregisterDentry 

func (*InodeFSOwned) UnregisterDentry(d *Dentry)

UnregisterDentry implements Remove.UnregisterDentry.

type InodeNoStatFS 

type InodeNoStatFS struct{}

InodeNoStatFS partially implements the Inode interface, where the client filesystem doesn't support statfs(2).

+stateify savable

func (*InodeNoStatFS) StatFS 

func (*InodeNoStatFS) StatFS(context.Context, *vfs.Filesystem) (linux.Statfs, error)

StatFS implements Inode.StatFS.

type InodeNoopRefCount 

type InodeNoopRefCount struct {
	InodeTemporary
}

InodeNoopRefCount partially implements the Inode interface, specifically the inodeRefs sub interface. InodeNoopRefCount implements a simple reference count for inodes, performing no extra actions when references are obtained or released. This is suitable for simple file inodes that don't reference any resources.

+stateify savable

func (InodeNoopRefCount) DecRef 

func (InodeNoopRefCount) DecRef(context.Context)

DecRef implements Inode.DecRef.

func (InodeNoopRefCount) IncRef 

func (InodeNoopRefCount) IncRef()

IncRef implements Inode.IncRef.

func (InodeNoopRefCount) TryIncRef 

func (InodeNoopRefCount) TryIncRef() bool

TryIncRef implements Inode.TryIncRef.

type InodeNotAnonymous 

type InodeNotAnonymous struct{}

InodeNotAnonymous partially implements Inode.

+stateify savable

func (*InodeNotAnonymous) Anonymous 

func (*InodeNotAnonymous) Anonymous() bool

Anonymous implements Inode.Anonymous

type InodeNotDirectory 

type InodeNotDirectory struct {
	InodeAlwaysValid
}

InodeNotDirectory partially implements the Inode interface, specifically the inodeDirectory and inodeDynamicDirectory sub interfaces. Inodes that do not represent directories can embed this to provide no-op implementations for directory-related functions.

+stateify savable

func (InodeNotDirectory) HasChildren 

func (InodeNotDirectory) HasChildren() bool

HasChildren implements Inode.HasChildren.

func (InodeNotDirectory) IterDirents 

func (InodeNotDirectory) IterDirents(ctx context.Context, mnt *vfs.Mount, callback vfs.IterDirentsCallback, offset, relOffset int64) (newOffset int64, err error)

IterDirents implements Inode.IterDirents.

func (InodeNotDirectory) Lookup 

func (InodeNotDirectory) Lookup(ctx context.Context, name string) (Inode, error)

Lookup implements Inode.Lookup.

func (InodeNotDirectory) NewDir 

func (InodeNotDirectory) NewDir(context.Context, string, vfs.MkdirOptions) (Inode, error)

NewDir implements Inode.NewDir.

func (InodeNotDirectory) NewFile 

func (InodeNotDirectory) NewFile(context.Context, string, vfs.OpenOptions) (Inode, error)

NewFile implements Inode.NewFile. 

func (InodeNotDirectory) NewLink(context.Context, string, Inode) (Inode, error)

NewLink implements Inode.NewLinkink.

func (InodeNotDirectory) NewNode 

func (InodeNotDirectory) NewNode(context.Context, string, vfs.MknodOptions) (Inode, error)

NewNode implements Inode.NewNode. 

func (InodeNotDirectory) NewSymlink(context.Context, string, string) (Inode, error)

NewSymlink implements Inode.NewSymlink.

func (InodeNotDirectory) Rename 

func (InodeNotDirectory) Rename(context.Context, string, string, Inode, Inode) error

Rename implements Inode.Rename.

func (InodeNotDirectory) RmDir 

func (InodeNotDirectory) RmDir(context.Context, string, Inode) error

RmDir implements Inode.RmDir. 

func (InodeNotDirectory) Unlink(context.Context, string, Inode) error

Unlink implements Inode.Unlink. 

type InodeNotSymlink struct{}

InodeNotSymlink partially implements the Inode interface, specifically the inodeSymlink sub interface. All inodes that are not symlinks may embed this to return the appropriate errors from symlink-related functions.

+stateify savable 

func (InodeNotSymlink) Getlink(context.Context, *vfs.Mount) (vfs.VirtualDentry, string, error)

Getlink implements Inode.Getlink. 

func (InodeNotSymlink) Readlink(context.Context, *vfs.Mount) (string, error)

Readlink implements Inode.Readlink. 

type InodeSymlink struct {
	InodeNotDirectory
}

InodeSymlink partially implements Inode interface for symlinks.

+stateify savable

func (InodeSymlink) Open 

func (InodeSymlink) Open(ctx context.Context, rp *vfs.ResolvingPath, d *Dentry, opts vfs.OpenOptions) (*vfs.FileDescription, error)

Open implements Inode.Open.

type InodeTemporary 

type InodeTemporary struct{}

InodeTemporary partially implements Inode.

+stateify savable

func (*InodeTemporary) Keep 

func (*InodeTemporary) Keep() bool

Keep implements Inode.Keep.

type InodeWatches 

type InodeWatches struct {
	// contains filtered or unexported fields
}

InodeWatches partially implements Inode.

+stateify savable

func (*InodeWatches) Watches 

func (i *InodeWatches) Watches() *vfs.Watches

Watches implements Inode.Watches.

type OrderedChildren 

type OrderedChildren struct {
	// contains filtered or unexported fields
}

OrderedChildren partially implements the Inode interface. OrderedChildren can be embedded in directory inodes to keep track of children in the directory, and can then be used to implement a generic directory FD -- see GenericDirectoryFD.

OrderedChildren can represent a node in an Inode tree. The children inodes might be directories themselves using OrderedChildren; hence extending the tree. The parent inode (OrderedChildren user) holds a ref on all its static children. This lets the static inodes outlive their associated dentry. While the dentry might have to be regenerated via a Lookup() call, we can keep reusing the same static inode. These static children inodes are finally DecRef'd when this directory inode is being destroyed. This makes OrderedChildren suitable for static directory entries as well.

Must be initialize with Init before first use.

+stateify savable

func (*OrderedChildren) Destroy 

func (o *OrderedChildren) Destroy(ctx context.Context)

Destroy clears the children stored in o. It should be called by structs embedding OrderedChildren upon destruction, i.e. when their reference count reaches zero.

func (*OrderedChildren) ForEachChild 

func (o *OrderedChildren) ForEachChild(fn func(string, Inode))

ForEachChild calls fn on all children tracked by this ordered children.

func (*OrderedChildren) HasChildren 

func (o *OrderedChildren) HasChildren() bool

HasChildren implements Inode.HasChildren.

func (*OrderedChildren) Init 

func (o *OrderedChildren) Init(opts OrderedChildrenOptions)

Init initializes an OrderedChildren.

func (*OrderedChildren) Insert 

func (o *OrderedChildren) Insert(name string, child Inode) error

Insert inserts a dynamic child into o. This ignores the writability of o, as this is not part of the vfs.FilesystemImpl interface, and is a lower-level operation.

func (*OrderedChildren) Inserter 

func (o *OrderedChildren) Inserter(name string, makeChild func() Inode) (Inode, error)

Inserter is like Insert, but obtains the child to insert by calling makeChild. makeChild is only called if the insert will succeed. This allows the caller to atomically check and insert a child without having to clean up the child on failure.

func (*OrderedChildren) IterDirents 

func (o *OrderedChildren) IterDirents(ctx context.Context, mnt *vfs.Mount, cb vfs.IterDirentsCallback, offset, relOffset int64) (newOffset int64, err error)

IterDirents implements Inode.IterDirents.

func (*OrderedChildren) Lookup 

func (o *OrderedChildren) Lookup(ctx context.Context, name string) (Inode, error)

Lookup implements Inode.Lookup.

func (*OrderedChildren) Populate 

func (o *OrderedChildren) Populate(children map[string]Inode) uint32

Populate inserts static children into this OrderedChildren. Populate returns the number of directories inserted, which the caller may use to update the link count for the parent directory.

Precondition:

  • d must represent a directory inode.
  • children must not contain any conflicting entries already in o.
  • Caller must hold a reference on all inodes passed.

Postcondition: Caller's references on inodes are transferred to o.

func (*OrderedChildren) Rename 

func (o *OrderedChildren) Rename(ctx context.Context, oldname, newname string, child, dstDir Inode) error

Rename implements Inode.Rename.

Precondition: Rename may only be called across two directory inodes with identical implementations of Rename. Practically, this means filesystems that implement Rename by embedding OrderedChildren for any directory implementation must use OrderedChildren for all directory implementations that will support Rename.

Postcondition: reference on any replaced dentry transferred to caller.

func (*OrderedChildren) RmDir 

func (o *OrderedChildren) RmDir(ctx context.Context, name string, child Inode) error

RmDir implements Inode.RmDir. 

func (o *OrderedChildren) Unlink(ctx context.Context, name string, child Inode) error

Unlink implements Inode.Unlink.

type OrderedChildrenOptions 

type OrderedChildrenOptions struct {
	// Writable indicates whether vfs.FilesystemImpl methods implemented by
	// OrderedChildren may modify the tracked children. This applies to
	// operations related to rename, unlink and rmdir. If an OrderedChildren is
	// not writable, these operations all fail with EPERM.
	//
	// Note that writable users must implement the sticky bit (I_SVTX).
	Writable bool
}

OrderedChildrenOptions contains initialization options for OrderedChildren.

+stateify savable

type SeekEndConfig 

type SeekEndConfig int

SeekEndConfig describes the SEEK_END behaviour for FDs.

+stateify savable

type StaticDirectory 

type StaticDirectory struct {
	InodeAlwaysValid
	InodeAttrs
	InodeDirectoryNoNewChildren
	InodeNoStatFS
	InodeNotAnonymous
	InodeNotSymlink
	InodeTemporary
	InodeWatches
	OrderedChildren
	StaticDirectoryRefs
	InodeFSOwned
	// contains filtered or unexported fields
}

StaticDirectory is a standard implementation of a directory with static contents.

+stateify savable

func (*StaticDirectory) DecRef 

func (s *StaticDirectory) DecRef(ctx context.Context)

DecRef implements Inode.DecRef.

func (*StaticDirectory) Init 

func (s *StaticDirectory) Init(ctx context.Context, creds *auth.Credentials, devMajor, devMinor uint32, ino uint64, perm linux.FileMode, fdOpts GenericDirectoryFDOptions)

Init initializes StaticDirectory.

func (*StaticDirectory) Open 

func (s *StaticDirectory) Open(ctx context.Context, rp *vfs.ResolvingPath, d *Dentry, opts vfs.OpenOptions) (*vfs.FileDescription, error)

Open implements Inode.Open.

func (*StaticDirectory) SetStat 

func (*StaticDirectory) SetStat(context.Context, *vfs.Filesystem, *auth.Credentials, vfs.SetStatOptions) error

SetStat implements Inode.SetStat not allowing inode attributes to be changed. 

type StaticSymlink struct {
	InodeAttrs
	InodeNoopRefCount
	InodeNotAnonymous
	InodeSymlink
	InodeNoStatFS
	InodeWatches
	InodeFSOwned
	// contains filtered or unexported fields
}

StaticSymlink provides an Inode implementation for symlinks that point to a immutable target.

+stateify savable 

func (s *StaticSymlink) Getlink(context.Context, *vfs.Mount) (vfs.VirtualDentry, string, error)

Getlink implements Inode.Getlink.

func (*StaticSymlink) Init 

func (s *StaticSymlink) Init(ctx context.Context, creds *auth.Credentials, devMajor uint32, devMinor uint32, ino uint64, target string)

Init initializes the instance. 

func (s *StaticSymlink) Readlink(_ context.Context, _ *vfs.Mount) (string, error)

Readlink implements Inode.Readlink.

func (*StaticSymlink) SetStat 

func (*StaticSymlink) SetStat(context.Context, *vfs.Filesystem, *auth.Credentials, vfs.SetStatOptions) error

SetStat implements Inode.SetStat not allowing inode attributes to be changed.

Source Files

View all Source files

Jump to

Keyboard shortcuts

? : This menu
/ : Search site
f or F : Jump to
y or Y : Canonical URL
go.dev uses cookies from Google to deliver and enhance the quality of its services and to analyze traffic. Learn more.