Age | Commit message (Collapse) | Author |
|
Singleton filesystem like devpts and devtmpfs have a single filesystem shared
among all mounts, so they acquire a "self-reference" when initialized that
must be released when the entire virtual filesystem is released at sandbox
exit.
PiperOrigin-RevId: 336828852
|
|
This change aims to fix the memory leak issue reported inĀ #3933.
Background:
VFS2 kernfs kept accumulating invalid dentries if those dentries were not
walked on. After substantial consideration of the problem by our team, we
decided to have an LRU cache solution. This change is the first part to that
solution, where we don't cache anything. The LRU cache can be added on top of
this.
What has changed:
- Introduced the concept of an inode tree in kernfs.OrderedChildren.
This is helpful is cases where the lifecycle of an inode is different from
that of a dentry.
- OrderedChildren now deals with initialized inodes instead of initialized
dentries. It now implements Lookup() where it constructs a new dentry
using the inode.
- OrderedChildren holds a ref on all its children inodes. With this change,
now an inode can "outlive" a dentry pointing to it. See comments in
kernfs.OrderedChildren.
- The kernfs dentry tree is solely maintained by kernfs only. Inode
implementations can not modify the dentry tree.
- Dentries that reach ref count 0 are removed from the dentry tree.
- revalidateChildLocked now defer-DecRefs the newly created dentry from
Inode.Lookup(), limiting its life to the current filesystem operation. If
refs are picked on the dentry during the FS op (via an FD or something),
then it will stick around and will be removed when the FD is closed. So there
is essentially _no caching_ for Look()ed up dentries.
- kernfs.DecRef does not have the precondition that fs.mu must be locked.
Fixes #3933
PiperOrigin-RevId: 336768576
|
|
Update signatures for:
- All methods in inodeDirectory
- deferDecRef() and Filesystem.droppedDentries
- newSyntheticDirectory()
- `slot`s used in OrderedChildren and subsequent methods like
replaceChildLocked() and checkExistingLocked()
- stepExistingLocked(), walkParentDirLocked(), checkCreateLocked()
Updates #1193
PiperOrigin-RevId: 333558866
|
|
Updates #1663
PiperOrigin-RevId: 333539293
|
|
It is called from the kernfs code (OpenAt and revalidateChildLocked()).
For RemoveChildLocked, it is opposed. We need to call it from fuse.RmDir and
fuse.Unlink.
PiperOrigin-RevId: 333453218
|
|
There were some instances where we were not enabling leak checking.
PiperOrigin-RevId: 333418571
|
|
Update signatures for:
- walkExistingLocked
- checkDeleteLocked
- Inode.Open
Updates #1193
PiperOrigin-RevId: 333163381
|
|
Updates #1193.
PiperOrigin-RevId: 332939026
|
|
PiperOrigin-RevId: 332486111
|
|
This fixes a use-after-free in fuse.DeviceFD.Release.
PiperOrigin-RevId: 332394146
|
|
As noticed by @ayushr2, the "implements" comments are not
consistent, e.g.
// IterDirents implements kernfs.inodeDynamicLookup.
// Generate implements vfs.DynamicBytesSource.Generate.
This patch improves this by making the comments like this
consistently include the package name (when the interface
and struct are not in the same package) and method name.
Signed-off-by: Tiwei Bie <tiwei.btw@antgroup.com>
|
|
fix #3956
|
|
fix #3963
|
|
|
|
copylocks: directory.go:34:7: Allocate passes lock by value:
fuse/fuse.directoryFD contains fuse/fuse.fileDescription contains
pkg/sentry/vfs/vfs.FileDescription contains pkg/sync/sync.Mutex
|
|
|
|
|
|
|
|
|
|
Fixes #3696
|
|
|
|
This change implements Release for the FUSE filesystem
and expected behaviors of the FUSE devices.
It includes several checks for aborted connection
in the path for making a request and a function
to abort all the ongoing FUSE requests in order.
|
|
This commit fixes the potential unexpected errors
of original handling of FUSE_RELEASE responses while
keep the same behavior (ignoring any reply).
|
|
This change adds bookkeeping variables for the
FUSE request. With them, old insecure confusing
code we used to process async requests is replaced
by new clear compiling ones. Future code can take
advantage of them to have better control of each
requests.
|
|
This change decouples the code that is weakly
tied to the connection struct from connection.go,
rename variables and files with more meaningful choices,
adds detailed comments, explains lock orders,
and adds other minor improvement to make
the existing FUSE code more readable and
more organized.
Purpose is to avoid too much code in one file
and provide better structure for the
future commits.
|
|
This commit implements FUSE_SETATTR command. When a system call modifies
the metadata of a regular file or a folder by chown(2), chmod(2),
truncate(2), utime(2), or utimes(2), they should be translated to
corresponding FUSE_SETATTR command and sent to the FUSE server.
Fixes #3332
|
|
According to Linux 4.4's FUSE behavior, the flags and fh attributes in
FUSE_GETATTR are only used in read, write, and lseek. fstat(2) doesn't
use them either. Add tests to ensure the requests sent from FUSE module
are consistent with Linux's.
Updates #3655
|
|
This commit adds basic write(2) support for FUSE.
|
|
FUSE_CREATE is called when issuing creat(2) or open(2) with O_CREAT. It
creates a new file on the FUSE filesystem.
Fixes #3825
|
|
This change removes the unnecessary loop and avoids
the recursive call. It also fixes minor bugs in this
function.
|
|
|
|
This commit removes unused marshalling functions in linux abi package
and moves self-defined FUSEInitRes wrapper to fuse package.
Updates #3707
|
|
The fuse_init_out struct changes in different FUSE kernel versions. A
FUSE server may implement older versions of fuse_init_out, but they
share common attributes from the beginning. Implement variable-length
marshallable interface to support older versions of ABI.
Fixes #3707
|
|
kernfs uses inode.Getlink to resolve symlink when look up paths.
Updates #3452
|
|
According to readdir(3), the offset attribute in struct dirent is the
offset to the next dirent instead of the offset of itself. Send the
successive FUSE_READDIR requests with the offset retrieved from the last
entry.
Updates #3255
|
|
Until #3698 is resolved, this change is needed to ensure we're not
corrupting memory anywhere.
|
|
Fixes #3255.
This change adds support for IterDirents. You can now use `ls` in
the FUSE sandbox.
Co-authored-by: Craig Chi <craigchi@google.com>
|
|
Fixes #3587
Co-authored-by: Craig Chi <craigchi@google.com>
|
|
Fixes #3206
|
|
Fixes #3392
|
|
Fixes #3316
|
|
Fixes #3452
|
|
Fixes #3492
|
|
Fixes #3314
|
|
Fixes #3174
|
|
Fixes #3231
Co-authored-by: Boyuan He <heboyuan@google.com>
|
|
Original FUSE integration test has limited capabilities. To test more
situations, the new integration test framework introduces a protocol
to communicate between testing thread and the FUSE server. In summary,
this change includes:
1. Remove CompareResult() and break SetExpected() into
SetServerResponse() and GetServerActualRequest(). We no longer set
up an expected request because we want to retrieve the actual FUSE
request made to the FUSE server and check in the testing thread.
2. Declare a serial buffer data structure to save the received requests
and expected responses sequentially. The data structure contains a
cursor to indicate the progress of accessing. This change makes
sequential SetServerResponse() and GetServerActualRequest() possible.
3. Replace 2 single directional pipes with 1 bi-directional socketpair.
A protocol which starts with FuseTestCmd is used between the testing
thread and the FUSE server to provide various functionality.
Fixes #3405
|
|
CopyContext is a better name for the interface because from
go-marshal's perspective, the interface has nothing to do with a
task. A kernel.Task happens to implement the interface, but so can
other things like MemoryManager and IO sequences.
PiperOrigin-RevId: 331959678
|
|
PiperOrigin-RevId: 331256608
|
|
Updates #1487
PiperOrigin-RevId: 330580699
|