about summary refs log tree commit diff
path: root/tvix/store/src/fs/mod.rs
blob: c1ae6c9c19011935c588df7d9ac7dae980a2b5e9 (plain) (blame)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
mod file_attr;
mod inode_tracker;
mod inodes;

#[cfg(feature = "fuse")]
pub mod fuse;

#[cfg(feature = "virtiofs")]
pub mod virtiofs;

#[cfg(test)]
mod tests;

use crate::pathinfoservice::PathInfoService;

use fuse_backend_rs::abi::fuse_abi::stat64;
use fuse_backend_rs::api::filesystem::{Context, FileSystem, FsOptions, ROOT_ID};
use futures::StreamExt;
use nix_compat::store_path::StorePath;
use parking_lot::RwLock;
use std::{
    collections::HashMap,
    io,
    sync::atomic::AtomicU64,
    sync::{atomic::Ordering, Arc},
    time::Duration,
};
use tokio::{
    io::{AsyncReadExt, AsyncSeekExt},
    sync::mpsc,
};
use tracing::{debug, info_span, instrument, warn};
use tvix_castore::proto as castorepb;
use tvix_castore::{
    blobservice::{BlobReader, BlobService},
    directoryservice::DirectoryService,
    proto::{node::Node, NamedNode},
    B3Digest,
};

use self::{
    file_attr::{gen_file_attr, ROOT_FILE_ATTR},
    inode_tracker::InodeTracker,
    inodes::{DirectoryInodeData, InodeData},
};

/// This implements a read-only FUSE filesystem for a tvix-store
/// with the passed [BlobService], [DirectoryService] and [PathInfoService].
///
/// We don't allow listing on the root mountpoint (inode 0).
/// In the future, this might be made configurable once a listing method is
/// added to [self.path_info_service], and then show all store paths in that
/// store.
///
/// Linux uses inodes in filesystems. When implementing FUSE, most calls are
/// *for* a given inode.
///
/// This means, we need to have a stable mapping of inode numbers to the
/// corresponding store nodes.
///
/// We internally delegate all inode allocation and state keeping to the
/// inode tracker, and store the currently "explored" store paths together with
/// root inode of the root.
///
/// There's some places where inodes are allocated / data inserted into
/// the inode tracker, if not allocated before already:
///  - Processing a `lookup` request, either in the mount root, or somewhere
///    deeper
///  - Processing a `readdir` request
///
///  Things pointing to the same contents get the same inodes, irrespective of
///  their own location.
///  This means:
///  - Symlinks with the same target will get the same inode.
///  - Regular/executable files with the same contents will get the same inode
///  - Directories with the same contents will get the same inode.
///
/// Due to the above being valid across the whole store, and considering the
/// merkle structure is a DAG, not a tree, this also means we can't do "bucketed
/// allocation", aka reserve Directory.size inodes for each PathInfo.
pub struct TvixStoreFs {
    blob_service: Arc<dyn BlobService>,
    directory_service: Arc<dyn DirectoryService>,
    path_info_service: Arc<dyn PathInfoService>,

    /// Whether to (try) listing elements in the root.
    list_root: bool,

    /// This maps a given StorePath to the inode we allocated for the root inode.
    store_paths: RwLock<HashMap<StorePath, u64>>,

    /// This keeps track of inodes and data alongside them.
    inode_tracker: RwLock<InodeTracker>,

    /// This holds all open file handles
    #[allow(clippy::type_complexity)]
    file_handles: RwLock<HashMap<u64, Arc<tokio::sync::Mutex<Box<dyn BlobReader>>>>>,

    next_file_handle: AtomicU64,

    tokio_handle: tokio::runtime::Handle,
}

impl TvixStoreFs {
    pub fn new(
        blob_service: Arc<dyn BlobService>,
        directory_service: Arc<dyn DirectoryService>,
        path_info_service: Arc<dyn PathInfoService>,
        list_root: bool,
    ) -> Self {
        Self {
            blob_service,
            directory_service,
            path_info_service,

            list_root,

            store_paths: RwLock::new(HashMap::default()),
            inode_tracker: RwLock::new(Default::default()),

            file_handles: RwLock::new(Default::default()),
            next_file_handle: AtomicU64::new(1),
            tokio_handle: tokio::runtime::Handle::current(),
        }
    }

    /// Retrieves the inode for a given StorePath, if present.
    /// This obtains a read lock on self.store_paths.
    fn get_inode_for_store_path(&self, store_path: &StorePath) -> Option<u64> {
        let store_paths = self.store_paths.read();
        store_paths.get(store_path).cloned()
    }

    /// For a given inode, look up the given directory behind it (from
    /// self.inode_tracker), and return its children.
    /// The inode_tracker MUST know about this inode already, and it MUST point
    /// to a [InodeData::Directory].
    /// It is ok if it's a [DirectoryInodeData::Sparse] - in that case, a lookup
    /// in self.directory_service is performed, and self.inode_tracker is updated with the
    /// [DirectoryInodeData::Populated].
    #[instrument(skip(self), err)]
    fn get_directory_children(&self, ino: u64) -> io::Result<(B3Digest, Vec<(u64, Node)>)> {
        let data = self.inode_tracker.read().get(ino).unwrap();
        match *data {
            // if it's populated already, return children.
            InodeData::Directory(DirectoryInodeData::Populated(
                ref parent_digest,
                ref children,
            )) => Ok((parent_digest.clone(), children.clone())),
            // if it's sparse, fetch data using directory_service, populate child nodes
            // and update it in [self.inode_tracker].
            InodeData::Directory(DirectoryInodeData::Sparse(ref parent_digest, _)) => {
                let directory = self
                    .tokio_handle
                    .block_on(self.tokio_handle.spawn({
                        let directory_service = self.directory_service.clone();
                        let parent_digest = parent_digest.to_owned();
                        async move { directory_service.get(&parent_digest).await }
                    }))
                    .unwrap()?
                    .ok_or_else(|| {
                        warn!(directory.digest=%parent_digest, "directory not found");
                        // If the Directory can't be found, this is a hole, bail out.
                        io::Error::from_raw_os_error(libc::EIO)
                    })?;

                // Turn the retrieved directory into a InodeData::Directory(DirectoryInodeData::Populated(..)),
                // allocating inodes for the children on the way.
                let children = {
                    let mut inode_tracker = self.inode_tracker.write();

                    let children: Vec<(u64, castorepb::node::Node)> = directory
                        .nodes()
                        .map(|child_node| {
                            let child_ino = inode_tracker.put((&child_node).into());
                            (child_ino, child_node)
                        })
                        .collect();

                    // replace.
                    inode_tracker.replace(
                        ino,
                        Arc::new(InodeData::Directory(DirectoryInodeData::Populated(
                            parent_digest.clone(),
                            children.clone(),
                        ))),
                    );

                    children
                };

                Ok((parent_digest.clone(), children))
            }
            // if the parent inode was not a directory, this doesn't make sense
            InodeData::Regular(..) | InodeData::Symlink(_) => {
                Err(io::Error::from_raw_os_error(libc::ENOTDIR))
            }
        }
    }

    /// This will turn a lookup request for a name in the root to a ino and
    /// [InodeData].
    /// It will peek in [self.store_paths], and then either look it up from
    /// [self.inode_tracker],
    /// or otherwise fetch from [self.path_info_service], and then insert into
    /// [self.inode_tracker].
    /// In the case the name can't be found, a libc::ENOENT is returned.
    fn name_in_root_to_ino_and_data(
        &self,
        name: &std::ffi::CStr,
    ) -> io::Result<(u64, Arc<InodeData>)> {
        // parse the name into a [StorePath].
        let store_path = StorePath::from_bytes(name.to_bytes()).map_err(|e| {
            debug!(e=?e, "unable to parse as store path");
            // This is not an error, but a "ENOENT", as someone can stat
            // a file inside the root that's no valid store path
            io::Error::from_raw_os_error(libc::ENOENT)
        })?;

        // Look up the inode for that store path.
        // If there's one, [self.inode_tracker] MUST also contain the data,
        // which we can then return.
        if let Some(inode) = self.get_inode_for_store_path(&store_path) {
            return Ok((
                inode,
                self.inode_tracker
                    .read()
                    .get(inode)
                    .expect("must exist")
                    .to_owned(),
            ));
        }

        // We don't have it yet, look it up in [self.path_info_service].
        match self.tokio_handle.block_on({
            let path_info_service = self.path_info_service.clone();
            let digest = *store_path.digest();
            async move { path_info_service.get(digest).await }
        }) {
            // if there was an error looking up the path_info, propagate up an IO error.
            Err(_e) => Err(io::Error::from_raw_os_error(libc::EIO)),
            // the pathinfo doesn't exist, so the file doesn't exist.
            Ok(None) => Err(io::Error::from_raw_os_error(libc::ENOENT)),
            // The pathinfo does exist
            Ok(Some(path_info)) => {
                // There must be a root node (ensured by the validation happening inside clients)
                let root_node = path_info.node.unwrap().node.unwrap();

                // The name must match what's passed in the lookup, otherwise this is also a ENOENT.
                if root_node.get_name() != store_path.to_string().as_bytes() {
                    debug!(root_node.name=?root_node.get_name(), store_path.name=%store_path.to_string(), "store path mismatch");
                    return Err(io::Error::from_raw_os_error(libc::ENOENT));
                }

                // Let's check if someone else beat us to updating the inode tracker and
                // store_paths map. This avoids locking inode_tracker for writing.
                if let Some(ino) = self.store_paths.read().get(&store_path) {
                    return Ok((
                        *ino,
                        self.inode_tracker.read().get(*ino).expect("must exist"),
                    ));
                }

                // Only in case it doesn't, lock [self.store_paths] and
                // [self.inode_tracker] for writing.
                let mut store_paths = self.store_paths.write();
                let mut inode_tracker = self.inode_tracker.write();

                // insert the (sparse) inode data and register in
                // self.store_paths.
                let inode_data: InodeData = (&root_node).into();
                let ino = inode_tracker.put(inode_data.clone());
                store_paths.insert(store_path, ino);

                Ok((ino, Arc::new(inode_data)))
            }
        }
    }
}

impl FileSystem for TvixStoreFs {
    type Handle = u64;
    type Inode = u64;

    fn init(&self, _capable: FsOptions) -> io::Result<FsOptions> {
        Ok(FsOptions::empty())
    }

    #[tracing::instrument(skip_all, fields(rq.inode = inode))]
    fn getattr(
        &self,
        _ctx: &Context,
        inode: Self::Inode,
        _handle: Option<Self::Handle>,
    ) -> io::Result<(stat64, Duration)> {
        if inode == ROOT_ID {
            return Ok((ROOT_FILE_ATTR.into(), Duration::MAX));
        }

        match self.inode_tracker.read().get(inode) {
            None => Err(io::Error::from_raw_os_error(libc::ENOENT)),
            Some(node) => {
                debug!(node = ?node, "found node");
                Ok((gen_file_attr(&node, inode).into(), Duration::MAX))
            }
        }
    }

    #[tracing::instrument(skip_all, fields(rq.parent_inode = parent, rq.name = ?name))]
    fn lookup(
        &self,
        _ctx: &Context,
        parent: Self::Inode,
        name: &std::ffi::CStr,
    ) -> io::Result<fuse_backend_rs::api::filesystem::Entry> {
        debug!("lookup");

        // This goes from a parent inode to a node.
        // - If the parent is [ROOT_ID], we need to check
        //   [self.store_paths] (fetching from PathInfoService if needed)
        // - Otherwise, lookup the parent in [self.inode_tracker] (which must be
        //   a [InodeData::Directory]), and find the child with that name.
        if parent == ROOT_ID {
            let (ino, inode_data) = self.name_in_root_to_ino_and_data(name)?;

            debug!(inode_data=?&inode_data, ino=ino, "Some");
            return Ok(fuse_backend_rs::api::filesystem::Entry {
                inode: ino,
                attr: gen_file_attr(&inode_data, ino).into(),
                attr_timeout: Duration::MAX,
                entry_timeout: Duration::MAX,
                ..Default::default()
            });
        }
        // This is the "lookup for "a" inside inode 42.
        // We already know that inode 42 must be a directory.
        let (parent_digest, children) = self.get_directory_children(parent)?;

        let span = info_span!("lookup", directory.digest = %parent_digest);
        let _enter = span.enter();

        // Search for that name in the list of children and return the FileAttrs.

        // in the children, find the one with the desired name.
        if let Some((child_ino, _)) = children.iter().find(|e| e.1.get_name() == name.to_bytes()) {
            // lookup the child [InodeData] in [self.inode_tracker].
            // We know the inodes for children have already been allocated.
            let child_inode_data = self.inode_tracker.read().get(*child_ino).unwrap();

            // Reply with the file attributes for the child.
            // For child directories, we still have all data we need to reply.
            Ok(fuse_backend_rs::api::filesystem::Entry {
                inode: *child_ino,
                attr: gen_file_attr(&child_inode_data, *child_ino).into(),
                attr_timeout: Duration::MAX,
                entry_timeout: Duration::MAX,
                ..Default::default()
            })
        } else {
            // Child not found, return ENOENT.
            Err(io::Error::from_raw_os_error(libc::ENOENT))
        }
    }

    // TODO: readdirplus?

    #[tracing::instrument(skip_all, fields(rq.inode = inode, rq.offset = offset))]
    fn readdir(
        &self,
        _ctx: &Context,
        inode: Self::Inode,
        _handle: Self::Handle,
        _size: u32,
        offset: u64,
        add_entry: &mut dyn FnMut(fuse_backend_rs::api::filesystem::DirEntry) -> io::Result<usize>,
    ) -> io::Result<()> {
        debug!("readdir");

        if inode == ROOT_ID {
            if !self.list_root {
                return Err(io::Error::from_raw_os_error(libc::EPERM)); // same error code as ipfs/kubo
            } else {
                let path_info_service = self.path_info_service.clone();
                let (tx, mut rx) = mpsc::channel(16);

                // This task will run in the background immediately and will exit
                // after the stream ends or if we no longer want any more entries.
                self.tokio_handle.spawn(async move {
                    let mut stream = path_info_service.list().skip(offset as usize).enumerate();
                    while let Some(path_info) = stream.next().await {
                        if tx.send(path_info).await.is_err() {
                            // If we get a send error, it means the sync code
                            // doesn't want any more entries.
                            break;
                        }
                    }
                });

                while let Some((i, path_info)) = rx.blocking_recv() {
                    let path_info = match path_info {
                        Err(e) => {
                            warn!("failed to retrieve pathinfo: {}", e);
                            return Err(io::Error::from_raw_os_error(libc::EPERM));
                        }
                        Ok(path_info) => path_info,
                    };

                    // We know the root node exists and the store_path can be parsed because clients MUST validate.
                    let root_node = path_info.node.unwrap().node.unwrap();
                    let store_path = StorePath::from_bytes(root_node.get_name()).unwrap();

                    // obtain the inode, or allocate a new one.
                    let ino = self
                        .get_inode_for_store_path(&store_path)
                        .unwrap_or_else(|| {
                            // insert the (sparse) inode data and register in
                            // self.store_paths.
                            let ino = self.inode_tracker.write().put((&root_node).into());
                            self.store_paths.write().insert(store_path.clone(), ino);
                            ino
                        });

                    let ty = match root_node {
                        Node::Directory(_) => libc::S_IFDIR,
                        Node::File(_) => libc::S_IFREG,
                        Node::Symlink(_) => libc::S_IFLNK,
                    };

                    let written = add_entry(fuse_backend_rs::api::filesystem::DirEntry {
                        ino,
                        offset: offset + i as u64 + 1,
                        type_: ty,
                        name: store_path.to_string().as_bytes(),
                    })?;
                    // If the buffer is full, add_entry will return `Ok(0)`.
                    if written == 0 {
                        break;
                    }
                }

                return Ok(());
            }
        }

        // lookup the children, or return an error if it's not a directory.
        let (parent_digest, children) = self.get_directory_children(inode)?;

        let span = info_span!("lookup", directory.digest = %parent_digest);
        let _enter = span.enter();

        for (i, (ino, child_node)) in children.iter().skip(offset as usize).enumerate() {
            // the second parameter will become the "offset" parameter on the next call.
            let written = add_entry(fuse_backend_rs::api::filesystem::DirEntry {
                ino: *ino,
                offset: offset + i as u64 + 1,
                type_: match child_node {
                    #[allow(clippy::unnecessary_cast)]
                    // libc::S_IFDIR is u32 on Linux and u16 on MacOS
                    Node::Directory(_) => libc::S_IFDIR as u32,
                    #[allow(clippy::unnecessary_cast)]
                    // libc::S_IFDIR is u32 on Linux and u16 on MacOS
                    Node::File(_) => libc::S_IFREG as u32,
                    #[allow(clippy::unnecessary_cast)]
                    // libc::S_IFDIR is u32 on Linux and u16 on MacOS
                    Node::Symlink(_) => libc::S_IFLNK as u32,
                },
                name: child_node.get_name(),
            })?;
            // If the buffer is full, add_entry will return `Ok(0)`.
            if written == 0 {
                break;
            }
        }

        Ok(())
    }

    #[tracing::instrument(skip_all, fields(rq.inode = inode))]
    fn open(
        &self,
        _ctx: &Context,
        inode: Self::Inode,
        _flags: u32,
        _fuse_flags: u32,
    ) -> io::Result<(
        Option<Self::Handle>,
        fuse_backend_rs::api::filesystem::OpenOptions,
    )> {
        if inode == ROOT_ID {
            return Err(io::Error::from_raw_os_error(libc::ENOSYS));
        }

        // lookup the inode
        match *self.inode_tracker.read().get(inode).unwrap() {
            // read is invalid on non-files.
            InodeData::Directory(..) | InodeData::Symlink(_) => {
                warn!("is directory");
                Err(io::Error::from_raw_os_error(libc::EISDIR))
            }
            InodeData::Regular(ref blob_digest, _blob_size, _) => {
                let span = info_span!("read", blob.digest = %blob_digest);
                let _enter = span.enter();

                let blob_service = self.blob_service.clone();
                let blob_digest = blob_digest.clone();

                let task = self
                    .tokio_handle
                    .spawn(async move { blob_service.open_read(&blob_digest).await });

                let blob_reader = self.tokio_handle.block_on(task).unwrap();

                match blob_reader {
                    Ok(None) => {
                        warn!("blob not found");
                        Err(io::Error::from_raw_os_error(libc::EIO))
                    }
                    Err(e) => {
                        warn!(e=?e, "error opening blob");
                        Err(io::Error::from_raw_os_error(libc::EIO))
                    }
                    Ok(Some(blob_reader)) => {
                        // get a new file handle
                        // TODO: this will overflow after 2**64 operations,
                        // which is fine for now.
                        // See https://cl.tvl.fyi/c/depot/+/8834/comment/a6684ce0_d72469d1
                        // for the discussion on alternatives.
                        let fh = self.next_file_handle.fetch_add(1, Ordering::SeqCst);

                        debug!("add file handle {}", fh);
                        self.file_handles
                            .write()
                            .insert(fh, Arc::new(tokio::sync::Mutex::new(blob_reader)));

                        Ok((
                            Some(fh),
                            fuse_backend_rs::api::filesystem::OpenOptions::empty(),
                        ))
                    }
                }
            }
        }
    }

    #[tracing::instrument(skip_all, fields(rq.inode = inode, fh = handle))]
    fn release(
        &self,
        _ctx: &Context,
        inode: Self::Inode,
        _flags: u32,
        handle: Self::Handle,
        _flush: bool,
        _flock_release: bool,
        _lock_owner: Option<u64>,
    ) -> io::Result<()> {
        // remove and get ownership on the blob reader
        match self.file_handles.write().remove(&handle) {
            // drop it, which will close it.
            Some(blob_reader) => drop(blob_reader),
            None => {
                // These might already be dropped if a read error occured.
                debug!("file_handle {} not found", handle);
            }
        }

        Ok(())
    }

    #[tracing::instrument(skip_all, fields(rq.inode = inode, rq.offset = offset, rq.size = size))]
    fn read(
        &self,
        _ctx: &Context,
        inode: Self::Inode,
        handle: Self::Handle,
        w: &mut dyn fuse_backend_rs::api::filesystem::ZeroCopyWriter,
        size: u32,
        offset: u64,
        _lock_owner: Option<u64>,
        _flags: u32,
    ) -> io::Result<usize> {
        debug!("read");

        // We need to take out the blob reader from self.file_handles, so we can
        // interact with it in the separate task.
        // On success, we pass it back out of the task, so we can put it back in self.file_handles.
        let blob_reader = match self.file_handles.read().get(&handle) {
            Some(blob_reader) => blob_reader.clone(),
            None => {
                warn!("file handle {} unknown", handle);
                return Err(io::Error::from_raw_os_error(libc::EIO));
            }
        };

        let task = self.tokio_handle.spawn(async move {
            let mut blob_reader = blob_reader.lock().await;

            // seek to the offset specified, which is relative to the start of the file.
            let resp = blob_reader.seek(io::SeekFrom::Start(offset)).await;

            match resp {
                Ok(pos) => {
                    debug_assert_eq!(offset, pos);
                }
                Err(e) => {
                    warn!("failed to seek to offset {}: {}", offset, e);
                    return Err(io::Error::from_raw_os_error(libc::EIO));
                }
            }

            // As written in the fuse docs, read should send exactly the number
            // of bytes requested except on EOF or error.

            let mut buf: Vec<u8> = Vec::with_capacity(size as usize);

            // copy things from the internal buffer into buf to fill it till up until size
            tokio::io::copy(&mut blob_reader.as_mut().take(size as u64), &mut buf).await?;

            Ok(buf)
        });

        let buf = self.tokio_handle.block_on(task).unwrap()?;

        w.write(&buf)
    }

    #[tracing::instrument(skip_all, fields(rq.inode = inode))]
    fn readlink(&self, _ctx: &Context, inode: Self::Inode) -> io::Result<Vec<u8>> {
        if inode == ROOT_ID {
            return Err(io::Error::from_raw_os_error(libc::ENOSYS));
        }

        // lookup the inode
        match *self.inode_tracker.read().get(inode).unwrap() {
            InodeData::Directory(..) | InodeData::Regular(..) => {
                Err(io::Error::from_raw_os_error(libc::EINVAL))
            }
            InodeData::Symlink(ref target) => Ok(target.to_vec()),
        }
    }
}