In VFS1, both fs/host and fs/gofer used the same utils for host file mappings.
Refactor parts of fsimpl/gofer to create similar utils to share with
fsimpl/host (memory accounting code moved to fsutil, page rounding arithmetic
moved to usermem).
Updates #1476.
PiperOrigin-RevId: 312345090
Fixes#1812. (The more direct cause of the deadlock is panic unsafety because
the historically high cost of defer means that we avoid it in hot paths,
including much of MM; defer is much cheaper as of Go 1.14, but still a
measurable overhead.)
PiperOrigin-RevId: 294560316
Updates #1198
Opening host pipes (by spinning in fdpipe) and host sockets is not yet
complete, and will be done in a future CL.
Major differences from VFS1 gofer client (sentry/fs/gofer), with varying levels
of backportability:
- "Cache policies" are replaced by InteropMode, which control the behavior of
timestamps in addition to caching. Under InteropModeExclusive (analogous to
cacheAll) and InteropModeWritethrough (analogous to cacheAllWritethrough),
client timestamps are *not* written back to the server (it is not possible in
9P or Linux for clients to set ctime, so writing back client-authoritative
timestamps results in incoherence between atime/mtime and ctime). Under
InteropModeShared (analogous to cacheRemoteRevalidating), client timestamps
are not used at all (remote filesystem clocks are authoritative). cacheNone
is translated to InteropModeShared + new option
filesystemOptions.specialRegularFiles.
- Under InteropModeShared, "unstable attribute" reloading for permission
checks, lookup, and revalidation are fused, which is feasible in VFS2 since
gofer.filesystem controls path resolution. This results in a ~33% reduction
in RPCs for filesystem operations compared to cacheRemoteRevalidating. For
example, consider stat("/foo/bar/baz") where "/foo/bar/baz" fails
revalidation, resulting in the instantiation of a new dentry:
VFS1 RPCs:
getattr("/") // fs.MountNamespace.FindLink() => fs.Inode.CheckPermission() => gofer.inodeOperations.check() => gofer.inodeOperations.UnstableAttr()
walkgetattr("/", "foo") = fid1 // fs.Dirent.walk() => gofer.session.Revalidate() => gofer.cachePolicy.Revalidate()
clunk(fid1)
getattr("/foo") // CheckPermission
walkgetattr("/foo", "bar") = fid2 // Revalidate
clunk(fid2)
getattr("/foo/bar") // CheckPermission
walkgetattr("/foo/bar", "baz") = fid3 // Revalidate
clunk(fid3)
walkgetattr("/foo/bar", "baz") = fid4 // fs.Dirent.walk() => gofer.inodeOperations.Lookup
getattr("/foo/bar/baz") // linux.stat() => gofer.inodeOperations.UnstableAttr()
VFS2 RPCs:
getattr("/") // gofer.filesystem.walkExistingLocked()
walkgetattr("/", "foo") = fid1 // gofer.filesystem.stepExistingLocked()
clunk(fid1)
// No getattr: walkgetattr already updated metadata for permission check
walkgetattr("/foo", "bar") = fid2
clunk(fid2)
walkgetattr("/foo/bar", "baz") = fid3
// No clunk: fid3 used for new gofer.dentry
// No getattr: walkgetattr already updated metadata for stat()
- gofer.filesystem.unlinkAt() does not require instantiation of a dentry that
represents the file to be deleted. Updates #898.
- gofer.regularFileFD.OnClose() skips Tflushf for regular files under
InteropModeExclusive, as it's nonsensical to request a remote file flush
without flushing locally-buffered writes to that remote file first.
- Symlink targets are cached when InteropModeShared is not in effect.
- p9.QID.Path (which is already required to be unique for each file within a
server, and is accordingly already synthesized from device/inode numbers in
all known gofers) is used as-is for inode numbers, rather than being mapped
along with attr.RDev in the client to yet another synthetic inode number.
- Relevant parts of fsutil.CachingInodeOperations are inlined directly into
gofer package code. This avoids having to duplicate part of its functionality
in fsutil.HostMappable.
PiperOrigin-RevId: 293190213
Because the abi will depend on the core types for marshalling (usermem,
context, safemem, safecopy), these need to be flattened from the sentry
directory. These packages contain no sentry-specific details.
PiperOrigin-RevId: 291811289
There was a very bare get/setxattr in the InodeOperations interface. Add
context.Context to both, size to getxattr, and flags to setxattr.
Note that extended attributes are passed around as strings in this
implementation, so size is automatically encoded into the value. Size is
added in getxattr so that implementations can return ERANGE if a value is larger
than can fit in the user-allocated buffer. This prevents us from unnecessarily
passing around an arbitrarily large xattr when the user buffer is actually too
small.
Don't use the existing xattrwalk and xattrcreate messages and define our
own, mainly for the sake of simplicity.
Extended attributes will be implemented in future commits.
PiperOrigin-RevId: 290121300
* Rename syncutil to sync.
* Add aliases to sync types.
* Replace existing usage of standard library sync package.
This will make it easier to swap out synchronization primitives. For example,
this will allow us to use primitives from github.com/sasha-s/go-deadlock to
check for lock ordering violations.
Updates #1472
PiperOrigin-RevId: 289033387
- Renamed memfs to tmpfs.
- Copied fileRangeSet bits from fs/fsutil/ to fsimpl/tmpfs/
- Changed tmpfs to be backed by filemem instead of byte slice.
- regularFileReadWriter uses a sync.Pool, similar to gofer client.
PiperOrigin-RevId: 288356380
Linux kernel before 4.19 doesn't implement a feature that updates
open FD after a file is open for write (and is copied to the upper
layer). Already open FD will continue to read the old file content
until they are reopened. This is especially problematic for gVisor
because it caches open files.
Flag was added to force readonly files to be reopenned when the
same file is open for write. This is only needed if using kernels
prior to 4.19.
Closes#1006
It's difficult to really test this because we never run on tests
on older kernels. I'm adding a test in GKE which uses kernels
with the overlayfs problem for 1.14 and lower.
PiperOrigin-RevId: 275115289
The gofer's CachingInodeOperations implementation contains an optimization for
the common open-read-close pattern when we have a host FD. In this case, the
host kernel will update the timestamp for us to a reasonably close time, so we
don't need an extra RPC to the gofer.
However, when the app explicitly sets the timestamps (via futimes or similar)
then we actually DO need to update the timestamps, because the host kernel
won't do it for us.
To fix this, a new boolean `forceSetTimestamps` was added to
CachineInodeOperations.SetMaskedAttributes. It is only set by
gofer.InodeOperations.SetTimestamps.
PiperOrigin-RevId: 272048146
This also allows the tee(2) implementation to be enabled, since dup can now be
properly supported via WriteTo.
Note that this change necessitated some minor restructoring with the
fs.FileOperations splice methods. If the *fs.File is passed through directly,
then only public API methods are accessible, which will deadlock immediately
since the locking is already done by fs.Splice. Instead, we pass through an
abstract io.Reader or io.Writer, which elide locks and use the underlying
fs.FileOperations directly.
PiperOrigin-RevId: 268805207
Get/Set pipe size and ioctl support were missing from
overlayfs. It required moving the pipe.Sizer interface
to fs so that overlay could get access.
Fixes#318
PiperOrigin-RevId: 255511125
Currently, the overlay dirCache is only used for a single logical use of
getdents. i.e., it is discard when the FD is closed or seeked back to
the beginning.
But the initial work of getting the directory contents can be quite
expensive (particularly sorting large directories), so we should keep it
as long as possible.
This is very similar to the readdirCache in fs/gofer.
Since the upper filesystem does not have to allow caching readdir
entries, the new CacheReaddir MountSourceOperations method controls this
behavior.
This caching should be trivially movable to all Inodes if desired,
though that adds an additional copy step for non-overlay Inodes.
(Overlay Inodes already do the extra copy).
PiperOrigin-RevId: 255477592
This does not actually implement an efficient splice or sendfile. Rather, it
adds a generic plumbing to the file internals so that this can be added. All
file implementations use the stub fileutil.NoSplice implementation, which
causes sendfile and splice to fall back to an internal copy.
A basic splice system call interface is added, along with a test.
PiperOrigin-RevId: 249335960
Change-Id: Ic5568be2af0a505c19e7aec66d5af2480ab0939b
* Creation of files, directories (and other fs objects) in a directory
should always update ctime.
* Same for removal.
* atime should not be updated on lookup, only readdir.
I've also renamed some misleading functions that update mtime and ctime.
PiperOrigin-RevId: 249115063
Change-Id: I30fa275fa7db96d01aa759ed64628c18bb3a7dc7
There is a lot of redundancy that we can simplify in the stat_times
test. This will make it easier to add new tests. However, the
simplification reveals that cached uattrs on goferfs don't properly
update ctime on rename.
PiperOrigin-RevId: 248773425
Change-Id: I52662728e1e9920981555881f9a85f9ce04041cf
This feature allows MemoryFile to delay eviction of "optional"
allocations, such as unused cached file pages.
Note that this incidentally makes CachingInodeOperations writeback
asynchronous, in the sense that it doesn't occur until eviction; this is
necessary because between when a cached page becomes evictable and when
it's evicted, file writes (via CachingInodeOperations.Write) may dirty
the page.
As currently implemented, this feature won't meaningfully impact
steady-state memory usage or caching; the reclaimer goroutine will
schedule eviction as soon as it runs out of other work to do. Future CLs
increase caching by adding constraints on when eviction is scheduled.
PiperOrigin-RevId: 246014822
Change-Id: Ia85feb25a2de92a48359eb84434b6ec6f9bea2cb
Based on the guidelines at
https://opensource.google.com/docs/releasing/authors/.
1. $ rg -l "Google LLC" | xargs sed -i 's/Google LLC.*/The gVisor Authors./'
2. Manual fixup of "Google Inc" references.
3. Add AUTHORS file. Authors may request to be added to this file.
4. Point netstack AUTHORS to gVisor AUTHORS. Drop CONTRIBUTORS.
Fixes#209
PiperOrigin-RevId: 245823212
Change-Id: I64530b24ad021a7d683137459cafc510f5ee1de9
RootFromContext can return a dirent with reference taken, or nil. We must call
DecRef if (and only if) a real dirent is returned.
PiperOrigin-RevId: 242965515
Change-Id: Ie2b7b4cb19ee09b6ccf788b71f3fd7efcdf35a11
Strings are a better fit for this usage because they are immutable in Go, and
can contain arbitrary bytes. It also allows us to avoid casting bytes to string
(and the associated allocation) in the hot path when checking for overlay
whiteouts.
PiperOrigin-RevId: 242208856
Change-Id: I7699ae6302492eca71787dd0b72e0a5a217a3db2
- Document fsutil.CachedFileObject.FD() requirements on access
permissions, and change gofer.inodeFileState.FD() to honor them.
Fixes#147.
- Combine gofer.inodeFileState.readonly and
gofer.inodeFileState.readthrough, and simplify handle caching logic.
- Inline gofer.cachePolicy.cacheHandles into
gofer.inodeFileState.setSharedHandles, because users with access to
gofer.inodeFileState don't necessarily have access to the fs.Inode
(predictably, this is a save/restore problem).
Before this CL:
$ docker run --runtime=runsc-d -v $(pwd)/gvisor/repro:/root/repro -it ubuntu bash
root@34d51017ed67:/# /root/repro/runsc-b147
mmap: 0x7f3c01e45000
Segmentation fault
After this CL:
$ docker run --runtime=runsc-d -v $(pwd)/gvisor/repro:/root/repro -it ubuntu bash
root@d3c3cb56bbf9:/# /root/repro/runsc-b147
mmap: 0x7f78987ec000
o
PiperOrigin-RevId: 240818413
Change-Id: I49e1d4a81a0cb9177832b0a9f31a10da722a896b
MM.insertPMAsLocked() passes vma.maxPerms to memmap.Mappable.Translate
(although it unsets AccessType.Write if the vma is private). This
somewhat simplifies handling of pmas, since it means only COW-break
needs to replace existing pmas. However, it also means that a MAP_SHARED
mapping of a file opened O_RDWR dirties the file, regardless of the
mapping's permissions and whether or not the mapping is ever actually
written to with I/O that ignores permissions (e.g.
ptrace(PTRACE_POKEDATA)).
To fix this:
- Change the pma-getting path to request only the permissions that are
required for the calling access.
- Change memmap.Mappable.Translate to take requested permissions, and
return allowed permissions. This preserves the existing behavior in the
common cases where the memmap.Mappable isn't
fsutil.CachingInodeOperations and doesn't care if the translated
platform.File pages are written to.
- Change the MM.getPMAsLocked path to support permission upgrading of
pmas outside of copy-on-write.
PiperOrigin-RevId: 240196979
Change-Id: Ie0147c62c1fbc409467a6fa16269a413f3d7d571
In the case of a rename replacing an existing destination inode, ramfs
Rename failed to first remove the replaced inode. This caused:
1. A leak of a reference to the inode (making it live indefinitely).
2. For directories, a leak of the replaced directory's .. link to the
parent. This would cause the parent's link count to incorrectly
increase.
(2) is much simpler to test than (1), so that's what I've done.
agentfs has a similar bug with link count only, so the Dirent layer
informs the Inode if this is a replacing rename.
Fixes#133
PiperOrigin-RevId: 239105698
Change-Id: I4450af2462d8ae3339def812287213d2cbeebde0
This is in preparation for improved page cache reclaim, which requires
greater integration between the page cache and page allocator.
PiperOrigin-RevId: 238444706
Change-Id: Id24141b3678d96c7d7dc24baddd9be555bffafe4
- Redefine some memmap.Mappable, platform.File, and platform.Memory
semantics in terms of File reference counts (no functional change).
- Make AddressSpace.MapFile take a platform.File instead of a raw FD,
and replace platform.File.MapInto with platform.File.FD. This allows
kvm.AddressSpace.MapFile to always use platform.File.MapInternal instead
of maintaining its own (redundant) cache of file mappings in the sentry
address space.
PiperOrigin-RevId: 238044504
Change-Id: Ib73a11e4275c0da0126d0194aa6c6017a9cef64f
Dirty should be set only when the attribute is changed in the cache
only. Instances where the change was also sent to the backing file
doesn't need to dirty the attribute.
Also remove size update during WriteOut as writing dirty page would
naturaly grow the file if needed.
RELNOTES: relnotes is needed for the parent CL.
PiperOrigin-RevId: 232068978
Change-Id: I00ba54693a2c7adc06efa9e030faf8f2e8e7f188
This changed required making fsutil.HostMappable use
a backing file to ensure the correct FD would be used
for read/write operations.
RELNOTES: relnotes is needed for the parent CL.
PiperOrigin-RevId: 231836164
Change-Id: I8ae9639715529874ea7d80a65e2c711a5b4ce254
Nothing reads them and they can simply get stale.
Generated with:
$ sed -i "s/licenses(\(.*\)).*/licenses(\1)/" **/BUILD
PiperOrigin-RevId: 231818945
Change-Id: Ibc3f9838546b7e94f13f217060d31f4ada9d4bf0
We were modifying InodeSimpleAttributes.Unstable.AccessTime without holding
the necessary lock. Luckily for us, InodeSimpleAttributes already has a
NotifyAccess method that will do the update while holding the lock.
In addition, we were holding dfo.dir.mu.Lock while setting AccessTime, which
is unnecessary, so that lock has been removed.
PiperOrigin-RevId: 231278447
Change-Id: I81ed6d3dbc0b18e3f90c1df5e5a9c06132761769
When file size changes outside the sandbox, page cache was not
refreshing file size which is required for cacheRemoteRevalidating.
In fact, cacheRemoteRevalidating should be skipping the cache
completely since it's not really benefiting from it. The cache is
cache is already bypassed for unstable attributes (see
cachePolicy.cacheUAttrs). And althought the cache is called to
map pages, they will always miss the cache and map directly from
the host.
Created a HostMappable struct that maps directly to the host and
use it for files with cacheRemoteRevalidating.
Closes#124
PiperOrigin-RevId: 230998440
Change-Id: Ic5f632eabe33b47241e05e98c95e9b2090ae08fc
More helper structs have been added to the fsutil package to make it easier to
implement fs.InodeOperations and fs.FileOperations.
PiperOrigin-RevId: 229305982
Change-Id: Ib6f8d3862f4216745116857913dbfa351530223b
Andrei Vagin
Dean Deng
Michael Pratt
gVisor bot
Jamie Liu
Adin Scannell
Ian Gudger
Nicolas Lacasse
Kevin Krakauer
Fabricio Voznika
Shiva Prasanth
Rahat Mahmood