fuse/nodefs/fsconnector.go

// Copyright 2016 the Go-FUSE Authors. All rights reserved.
// Use of this source code is governed by a BSD-style
// license that can be found in the LICENSE file.

package nodefs

// This file contains the internal logic of the
// FileSystemConnector. The functions for satisfying the raw interface
// are in fsops.go

import (
	"log"
	"path/filepath"
	"strings"
	"sync"
	"time"
	"unsafe"

	"github.com/hanwen/go-fuse/v2/fuse"
)

// Tests should set to true.
var paranoia = false

// FileSystemConnector translates the raw FUSE protocol (serialized
// structs of uint32/uint64) to operations on Go objects representing
// files and directories.
type FileSystemConnector struct {
	debug bool

	// Callbacks for talking back to the kernel.
	server *fuse.Server

	// Translate between uint64 handles and *Inode.
	inodeMap handleMap

	// The root of the FUSE file system.
	rootNode *Inode

	// This lock prevents Lookup() and Forget() from running concurrently.
	// Locking at this level is a big hammer, but makes sure we don't return
	// forgotten nodes to the kernel. Problems solved by this lock:
	// https://github.com/hanwen/go-fuse/issues/168
	// https://github.com/rfjakob/gocryptfs/issues/322
	//
	// The lock is shared: several concurrent Lookups are allowed to be
	// run simultaneously, while Forget is exclusive.
	lookupLock sync.RWMutex
}

// NewOptions generates FUSE options that correspond to libfuse's
// defaults.
func NewOptions() *Options {
	return &Options{
		NegativeTimeout: 0,
		AttrTimeout:     time.Second,
		EntryTimeout:    time.Second,
		Owner:           fuse.CurrentOwner(),
	}
}

// NewFileSystemConnector creates a FileSystemConnector with the given
// options.
func NewFileSystemConnector(root Node, opts *Options) (c *FileSystemConnector) {
	c = new(FileSystemConnector)
	if opts == nil {
		opts = NewOptions()
	}
	c.inodeMap = newPortableHandleMap()
	c.rootNode = newInode(true, root)

	c.verify()
	c.mountRoot(opts)

	// FUSE does not issue a LOOKUP for 1 (obviously), but it does
	// issue a forget.  This lookupUpdate is to make the counts match.
	c.lookupUpdate(c.rootNode)
	c.debug = opts.Debug

	return c
}

// Server returns the fuse.Server that talking to the kernel.
func (c *FileSystemConnector) Server() *fuse.Server {
	return c.server
}

// SetDebug toggles printing of debug information. This function is
// deprecated. Set the Debug option in the Options struct instead.
func (c *FileSystemConnector) SetDebug(debug bool) {
	c.debug = debug
}

// This verifies invariants of the data structure.  This routine
// acquires tree locks as it walks the inode tree.
func (c *FileSystemConnector) verify() {
	if !paranoia {
		return
	}
	root := c.rootNode
	root.verify(c.rootNode.mountPoint)
}

// childLookup fills entry information for a newly created child inode
func (c *rawBridge) childLookup(out *fuse.EntryOut, n *Inode, context *fuse.Context) {
	n.Node().GetAttr(&out.Attr, nil, context)
	n.mount.fillEntry(out)
	out.NodeId, out.Generation = c.fsConn().lookupUpdate(n)
	if out.Ino == 0 {
		out.Ino = out.NodeId
	}
	if out.Nlink == 0 {
		// With Nlink == 0, newer kernels will refuse link
		// operations.
		out.Nlink = 1
	}
}

func (c *rawBridge) toInode(nodeid uint64) *Inode {
	if nodeid == fuse.FUSE_ROOT_ID {
		return c.rootNode
	}
	i := (*Inode)(unsafe.Pointer(c.inodeMap.Decode(nodeid)))
	return i
}

// Must run outside treeLock.  Returns the nodeId and generation.
func (c *FileSystemConnector) lookupUpdate(node *Inode) (id, generation uint64) {
	id, generation = c.inodeMap.Register(&node.handled)
	c.verify()
	return
}

// forgetUpdate decrements the reference counter for "nodeID" by "forgetCount".
// Must run outside treeLock.
func (c *FileSystemConnector) forgetUpdate(nodeID uint64, forgetCount int) {
	if nodeID == fuse.FUSE_ROOT_ID {
		c.rootNode.Node().OnUnmount()

		// We never got a lookup for root, so don't try to
		// forget root.
		return
	}

	// Prevent concurrent modification of the tree while we are processing
	// the FORGET
	node := (*Inode)(unsafe.Pointer(c.inodeMap.Decode(nodeID)))
	node.mount.treeLock.Lock()
	defer node.mount.treeLock.Unlock()

	if forgotten, _ := c.inodeMap.Forget(nodeID, forgetCount); forgotten {
		if len(node.children) > 0 || !node.Node().Deletable() ||
			node == c.rootNode || node.mountPoint != nil {
			// We cannot forget a directory that still has children as these
			// would become unreachable.
			return
		}
		// We have to remove ourself from all parents.
		// Create a copy of node.parents so we can safely iterate over it
		// while modifying the original.
		parents := make(map[parentData]struct{}, len(node.parents))
		for k, v := range node.parents {
			parents[k] = v
		}

		for p := range parents {
			// This also modifies node.parents
			p.parent.rmChild(p.name)
		}

		node.fsInode.OnForget()
	}
	// TODO - try to drop children even forget was not successful.
	c.verify()
}

// InodeCount returns the number of inodes registered with the kernel.
func (c *FileSystemConnector) InodeHandleCount() int {
	return c.inodeMap.Count()
}

// Finds a node within the currently known inodes, returns the last
// known node and the remaining unknown path components.  If parent is
// nil, start from FUSE mountpoint.
func (c *FileSystemConnector) Node(parent *Inode, fullPath string) (*Inode, []string) {
	if parent == nil {
		parent = c.rootNode
	}
	if fullPath == "" {
		return parent, nil
	}

	sep := string(filepath.Separator)
	fullPath = strings.TrimLeft(filepath.Clean(fullPath), sep)
	comps := strings.Split(fullPath, sep)

	node := parent
	if node.mountPoint == nil {
		node.mount.treeLock.RLock()
		defer node.mount.treeLock.RUnlock()
	}

	for i, component := range comps {
		if len(component) == 0 {
			continue
		}

		if node.mountPoint != nil {
			node.mount.treeLock.RLock()
			defer node.mount.treeLock.RUnlock()
		}

		next := node.children[component]
		if next == nil {
			return node, comps[i:]
		}
		node = next
	}

	return node, nil
}

// Follows the path from the given parent, doing lookups as
// necessary. The path should be '/' separated without leading slash.
func (c *FileSystemConnector) LookupNode(parent *Inode, path string) *Inode {
	if path == "" {
		return parent
	}

	components := strings.Split(path, "/")
	for _, r := range components {
		var a fuse.Attr
		// This will not affect inode ID lookup counts, which
		// are only update in response to kernel requests.
		var dummy fuse.InHeader
		child, _ := c.internalLookup(nil, &a, parent, r, &dummy)
		if child == nil {
			return nil
		}

		parent = child
	}

	return parent
}

func (c *FileSystemConnector) mountRoot(opts *Options) {
	c.rootNode.mountFs(opts)
	c.rootNode.mount.connector = c
	c.verify()
}

// Mount() generates a synthetic directory node, and mounts the file
// system there.  If opts is nil, the mount options of the root file
// system are inherited.  The encompassing filesystem should pretend
// the mount point does not exist.
//
// It returns ENOENT if the directory containing the mount point does
// not exist, and EBUSY if the intended mount point already exists.
func (c *FileSystemConnector) Mount(parent *Inode, name string, root Node, opts *Options) fuse.Status {
	node, code := c.lockMount(parent, name, root, opts)
	if !code.Ok() {
		return code
	}

	node.Node().OnMount(c)
	return code
}

func (c *FileSystemConnector) lockMount(parent *Inode, name string, root Node, opts *Options) (*Inode, fuse.Status) {
	defer c.verify()
	parent.mount.treeLock.Lock()
	defer parent.mount.treeLock.Unlock()
	node := parent.children[name]
	if node != nil {
		return nil, fuse.EBUSY
	}

	node = newInode(true, root)
	if opts == nil {
		opts = c.rootNode.mountPoint.options
	}

	node.mountFs(opts)
	node.mount.connector = c
	parent.addChild(name, node)

	node.mountPoint.parentInode = parent
	if c.debug {
		log.Printf("Mount %T on subdir %s, parent i%d", node,
			name, c.inodeMap.Handle(&parent.handled))
	}
	return node, fuse.OK
}

// Unmount() tries to unmount the given inode.  It returns EINVAL if the
// path does not exist, or is not a mount point, and EBUSY if there
// are open files or submounts below this node.
func (c *FileSystemConnector) Unmount(node *Inode) fuse.Status {
	// TODO - racy.
	if node.mountPoint == nil {
		log.Println("not a mountpoint:", c.inodeMap.Handle(&node.handled))
		return fuse.EINVAL
	}

	nodeID := c.inodeMap.Handle(&node.handled)

	// Must lock parent to update tree structure.
	parentNode := node.mountPoint.parentInode
	parentNode.mount.treeLock.Lock()
	defer parentNode.mount.treeLock.Unlock()

	mount := node.mountPoint
	name := node.mountPoint.mountName()
	if mount.openFiles.Count() > 0 {
		return fuse.EBUSY
	}

	node.mount.treeLock.Lock()
	defer node.mount.treeLock.Unlock()

	if mount.mountInode != node {
		log.Panicf("got two different mount inodes %v vs %v",
			c.inodeMap.Handle(&mount.mountInode.handled),
			c.inodeMap.Handle(&node.handled))
	}

	if !node.canUnmount() {
		return fuse.EBUSY
	}

	delete(parentNode.children, name)
	node.Node().OnUnmount()

	parentId := c.inodeMap.Handle(&parentNode.handled)
	if parentNode == c.rootNode {
		// TODO - test coverage. Currently covered by zipfs/multizip_test.go
		parentId = fuse.FUSE_ROOT_ID
	}

	// We have to wait until the kernel has forgotten the
	// mountpoint, so the write to node.mountPoint is no longer
	// racy.
	mount.treeLock.Unlock()
	parentNode.mount.treeLock.Unlock()
	code := c.server.DeleteNotify(parentId, nodeID, name)

	if code.Ok() {
		delay := 100 * time.Microsecond

		for {
			// This operation is rare, so we kludge it to avoid
			// contention.
			time.Sleep(delay)
			delay = delay * 2
			if !c.inodeMap.Has(nodeID) {
				break
			}

			if delay >= time.Second {
				// We limit the wait at one second. If
				// it takes longer, something else is
				// amiss, and we would be waiting forever.
				log.Println("kernel did not issue FORGET for node on Unmount.")
				break
			}
		}

	}

	parentNode.mount.treeLock.Lock()
	mount.treeLock.Lock()
	mount.mountInode = nil
	node.mountPoint = nil

	return fuse.OK
}

// FileNotify notifies the kernel that data and metadata of this inode
// has changed.  After this call completes, the kernel will issue a
// new GetAttr requests for metadata and new Read calls for content.
// Use negative offset for metadata-only invalidation, and zero-length
// for invalidating all content.
func (c *FileSystemConnector) FileNotify(node *Inode, off int64, length int64) fuse.Status {
	var nID uint64
	if node == c.rootNode {
		nID = fuse.FUSE_ROOT_ID
	} else {
		nID = c.inodeMap.Handle(&node.handled)
	}

	if nID == 0 {
		return fuse.OK
	}
	return c.server.InodeNotify(nID, off, length)
}

// FileNotifyStoreCache notifies the kernel about changed data of the inode.
//
// This call is similar to FileNotify, but instead of only invalidating a data
// region, it puts updated data directly to the kernel cache:
//
// After this call completes, the kernel has put updated data into the inode's cache,
// and will use data from that cache for non direct-IO reads from the inode
// in corresponding data region. After kernel's cache data is evicted, the kernel
// will have to issue new Read calls on user request to get data content.
//
// ENOENT is returned if the kernel does not currently have entry for this
// inode in its dentry cache.
func (c *FileSystemConnector) FileNotifyStoreCache(node *Inode, off int64, data []byte) fuse.Status {
	var nID uint64
	if node == c.rootNode {
		nID = fuse.FUSE_ROOT_ID
	} else {
		nID = c.inodeMap.Handle(&node.handled)
	}

	if nID == 0 {
		// the kernel does not currently know about this inode.
		return fuse.ENOENT
	}
	return c.server.InodeNotifyStoreCache(nID, off, data)
}

// FileRetrieveCache retrieves data from kernel's inode cache.
//
// This call retrieves data from kernel's inode cache @ offset and up to
// len(dest) bytes. If kernel cache has fewer consecutive data starting at
// offset, that fewer amount is returned. In particular if inode data at offset
// is not cached (0, OK) is returned.
//
// If the kernel does not currently have entry for this inode in its dentry
// cache (0, OK) is still returned, pretending that the inode could be known to
// the kernel, but kernel's inode cache is empty.
func (c *FileSystemConnector) FileRetrieveCache(node *Inode, off int64, dest []byte) (n int, st fuse.Status) {
	var nID uint64
	if node == c.rootNode {
		nID = fuse.FUSE_ROOT_ID
	} else {
		nID = c.inodeMap.Handle(&node.handled)
	}

	if nID == 0 {
		// the kernel does not currently know about this inode.
		// -> we can pretend that its cache for the inode is empty.
		return 0, fuse.OK
	}
	return c.server.InodeRetrieveCache(nID, off, dest)
}

// EntryNotify makes the kernel forget the entry data from the given
// name from a directory.  After this call, the kernel will issue a
// new lookup request for the given name when necessary. No filesystem
// related locks should be held when calling this.
func (c *FileSystemConnector) EntryNotify(node *Inode, name string) fuse.Status {
	var nID uint64
	if node == c.rootNode {
		nID = fuse.FUSE_ROOT_ID
	} else {
		nID = c.inodeMap.Handle(&node.handled)
	}

	if nID == 0 {
		return fuse.OK
	}
	return c.server.EntryNotify(nID, name)
}

// DeleteNotify signals to the kernel that the named entry in dir for
// the child disappeared. No filesystem related locks should be held
// when calling this.
func (c *FileSystemConnector) DeleteNotify(dir *Inode, child *Inode, name string) fuse.Status {
	var nID uint64

	if dir == c.rootNode {
		nID = fuse.FUSE_ROOT_ID
	} else {
		nID = c.inodeMap.Handle(&dir.handled)
	}

	if nID == 0 {
		return fuse.OK
	}

	chId := c.inodeMap.Handle(&child.handled)

	return c.server.DeleteNotify(nID, chId, name)
}