summaryrefslogtreecommitdiffhomepage
path: root/pkg/sentry/fsimpl
AgeCommit message (Collapse)Author
2020-12-17Set verityMu to be state nosaveChong Cai
PiperOrigin-RevId: 348092999
2020-12-17Fix seek on /proc/pid/cmdline when task is zombie.Nicolas Lacasse
PiperOrigin-RevId: 348056159
2020-12-17Set process group and session on host TTYFabricio Voznika
Closes #5128 PiperOrigin-RevId: 348052446
2020-12-15Change violation mode to an enumChong Cai
PiperOrigin-RevId: 347706953
2020-12-15Internal change.gVisor bot
PiperOrigin-RevId: 347671070
2020-12-11Internal change.gVisor bot
PiperOrigin-RevId: 347091372
2020-12-11Remove existing nogo exceptions.Adin Scannell
PiperOrigin-RevId: 347047550
2020-12-10Change merkle root file name to avoid collisionChong Cai
PiperOrigin-RevId: 346923826
2020-12-07Fix error handling on fusefs mount.Rahat Mahmood
Don't propagate arbitrary golang errors up from fusefs because errors that don't map to an errno result in a sentry panic. Reported-by: syzbot+697cb635346e456fddfc@syzkaller.appspotmail.com PiperOrigin-RevId: 346220306
2020-12-04Overlay runsc regular file mounts with regular files.Jamie Liu
Fixes #4991 PiperOrigin-RevId: 345800333
2020-12-03Implement `fcntl` options `F_GETSIG` and `F_SETSIG`.Etienne Perot
These options allow overriding the signal that gets sent to the process when I/O operations are available on the file descriptor, rather than the default `SIGIO` signal. Doing so also populates `siginfo` to contain extra information about which file descriptor caused the event (`si_fd`) and what events happened on it (`si_band`). The logic around which FD is populated within `si_fd` matches Linux's, which means it has some weird edge cases where that value may not actually refer to a file descriptor that is still valid. This CL also ports extra S/R logic regarding async handler in VFS2. Without this, async I/O handlers aren't properly re-registered after S/R. PiperOrigin-RevId: 345436598
2020-12-02Remove FileReadWriteSeeker from vfs.Jamie Liu
Previous experience has shown that these types of wrappers tends to create two kinds of problems: hidden allocations (e.g. each call to FileReadWriteSeeker.Read/Write allocates a usermem.BytesIO on the heap) and hidden lock ordering problems (e.g. VFS1 splice deadlocks). Since this is only needed by fsimpl/verity, move it there. PiperOrigin-RevId: 345377830
2020-12-02Clean up verity tests.Dean Deng
Refactor some utilities and rename some others for clarity. PiperOrigin-RevId: 345247836
2020-12-02Add /proc/sys/kernel/sem.Jing Chen
PiperOrigin-RevId: 345178956
2020-11-24Remove outdated TODO.Dean Deng
The bug has been fixed. PiperOrigin-RevId: 344088206
2020-11-23Don't evict gofer.dentries with inotify watches before saving.Jamie Liu
PiperOrigin-RevId: 343959348
2020-11-20Refactor verity test for readabilityChong Cai
1. Add getD/getDentry methods to avoid long casting line in each test 2. Factor all calls to vfs.OpenAt/UnlinkAt/RenameAt on lower filesystem to their own method (for both lower file and lower Merkle file) so the tests are more readable 3. Add descriptive test names for delete/remove tests PiperOrigin-RevId: 343540202
2020-11-19Remove racy stringification of socket fds from /proc/net/*.Rahat Mahmood
PiperOrigin-RevId: 343398191
2020-11-18[vfs] kernfs: Do not panic if destroyed dentry is cached.Ayush Ranjan
If a kernfs user does not cache dentries, then cacheLocked will destroy the dentry. The current DecRef implementation will be racy in this case as the following can happen: - Goroutine 1 calls DecRef and decreases ref count from 1 to 0. - Goroutine 2 acquires d.fs.mu for reading and calls IncRef and increasing the ref count from 0 to 1. - Goroutine 2 releases d.fs.mu and calls DecRef again decreasing ref count from 1 to 0. - Goroutine 1 now acquires d.fs.mu and calls cacheLocked which destroys the dentry. - Goroutine 2 now acquires d.fs.mu and calls cacheLocked to find that the dentry is already destroyed! Earlier we would panic in this case, we could instead just return instead of adding complexity to handle this race. This is similar to what the gofer client does. We do not want to lock d.fs.mu in the case that the filesystem caches dentries (common case as procfs and sysfs do this) to prevent congestion due to lock contention. PiperOrigin-RevId: 343229496
2020-11-18Port filesystem metrics to VFS2.Jamie Liu
PiperOrigin-RevId: 343196927
2020-11-17fs/fuse: don't dereference fuse.DeviceFD.fs if it is nilAndrei Vagin
PiperOrigin-RevId: 342992936
2020-11-17tmpfs: make sure that a dentry will not be destroyed before the open() callAndrei Vagin
If we don't hold a reference, the dentry can be destroyed by another thread. Reported-by: syzbot+f2132e50060c41f6d41f@syzkaller.appspotmail.com PiperOrigin-RevId: 342951940
2020-11-17Add consistent precondition formatting for verityChong Cai
Also add the lock order for verity fs, and add a lock to protect dentry hash. PiperOrigin-RevId: 342946537
2020-11-13Have fuse.DeviceFD hold reference on fuse.filesystem.Jamie Liu
This is actually just b/168751672 again; cl/332394146 was incorrectly reverted by cl/341411151. Document the reference holder to reduce the likelihood that this happens again. Also document a few other bugs observed in the process. PiperOrigin-RevId: 342339144
2020-11-13fs/tmpfs: change regularFile.size atomicallyAndrei Vagin
PiperOrigin-RevId: 342221309
2020-11-13fs/tmpfs: use atomic operations to access inode.modeAndrei Vagin
PiperOrigin-RevId: 342214859
2020-11-12Fix misuses of kernel.Task as context.Context.Jamie Liu
kernel.Task can only be used as context.Context by that Task's task goroutine. This is violated in at least two places: - In any case where one thread accesses the /proc/[tid] of any other thread, passing the kernel.Task for [tid] as the context.Context is incorrect. - Task.rebuildTraceContext() may be called by Kernel.RebuildTraceContexts() outside the scope of any task goroutine. Fix these (as well as a data race on Task.traceContext discovered during the course of finding the latter). PiperOrigin-RevId: 342174404
2020-11-12Add children names into verity hashChong Cai
children names map can be used to verify whether a child is expected during walking, so that we can detect unexpected modifications that deleted/renamed both the target file and the corresponding merkle tree file. PiperOrigin-RevId: 342170715
2020-11-12Rename kernel.TaskContext to kernel.TaskImage.Jamie Liu
This reduces confusion with context.Context (which is also relevant to kernel.Tasks) and is consistent with existing function kernel.LoadTaskImage(). PiperOrigin-RevId: 342167298
2020-11-12Filter dentries with non-zero refs in VFS2 gofer/overlay checks.Jamie Liu
PiperOrigin-RevId: 342161204
2020-11-11Read fsimpl/tmpfs timestamps atomically.Jamie Liu
PiperOrigin-RevId: 341982672
2020-11-09Print a debug message if /sys/kernel/debug/kcov is availableAndrei Vagin
This will help to debug: https://syzkaller.appspot.com/bug?id=0d717bd7028dceeb4b38f09aab2841c398b41d81 PiperOrigin-RevId: 341458715
2020-11-09Initialize references with a value of 1.Dean Deng
This lets us avoid treating a value of 0 as one reference. All references using the refsvfs2 template must call InitRefs() before the reference is incremented/decremented, or else a panic will occur. Therefore, it should be pretty easy to identify missing InitRef calls during testing. Updates #1486. PiperOrigin-RevId: 341411151
2020-11-06Allow VFS2 gofer.dentries to have separate read and write FDs.Jamie Liu
This is necessary to allow writes to files opened with O_WRONLY to go through host FDs. PiperOrigin-RevId: 341174509
2020-11-06[vfs] Return EEXIST when file already exists and rp.MustBeDir() is true.Ayush Ranjan
This is consistent with what Linux does. This was causing a PHP runtime test failure. Fixed it for VFS2. PiperOrigin-RevId: 341155209
2020-11-06[vfs] overlayfs: Do not unlink non-existing whiteout during file creation.Ayush Ranjan
We can reuse information about whether a whiteout exists on a given file path from stepLocked when creating a file at that path. This helps save an Unlink call to the upper filesystem if the whiteout does NOT exist (common case). Plumbs this information from lookupLocked() -> getChildLocked() -> stepLocked(). This also helped save a Lookup in RenameAt(). Fixes #1199 PiperOrigin-RevId: 341105351
2020-11-06Avoid extra DecRef on kernfs root for "kept" dentries.Dean Deng
The root dentry was not created through Inode.Lookup, so we should not release a reference even if inode.Keep() is true. PiperOrigin-RevId: 341103220
2020-11-03Fix more nogo testsTing-Yu Wang
PiperOrigin-RevId: 340536306
2020-11-03Make pipe min/max sizes match linux.Nicolas Lacasse
The default pipe size already matched linux, and is unchanged. Furthermore `atomicIOBytes` is made a proper constant (as it is in Linux). We were plumbing usermem.PageSize everywhere, so this is no functional change. PiperOrigin-RevId: 340497006
2020-11-02[vfs2] Refactor kernfs checkCreateLocked.Dean Deng
Don't return the filename, since it can already be determined by the caller. This was causing a panic in RenameAt, which relied on the name to be nonempty even if the error was EEXIST. Reported-by: syzbot+e9f117d000301e42361f@syzkaller.appspotmail.com PiperOrigin-RevId: 340381946
2020-11-02Pass hashing algorithms in verity fs optsChong Cai
PiperOrigin-RevId: 340275942
2020-10-30Adjust error handling in kernfs rename.Dean Deng
Read-only directories (e.g. under /sys, /proc) should return EPERM for rename. PiperOrigin-RevId: 339979022
2020-10-30Fix rename error handling for VFS2 kernfs.Dean Deng
The non-errno error was causing panics before. PiperOrigin-RevId: 339969348
2020-10-28Add leak checking for kernfs.Dentry.Dean Deng
Updates #1486. PiperOrigin-RevId: 339581879
2020-10-28Add logging option to leak checker.Dean Deng
Also refactor the template and CheckedObject interface to make this cleaner. Updates #1486. PiperOrigin-RevId: 339577120
2020-10-28Invalidate overlay.dentry.dirents during open() file creation.Jamie Liu
Updates #1199 PiperOrigin-RevId: 339528827
2020-10-28[vfs] Refactor hostfs mmap into kernfs util.Ayush Ranjan
PiperOrigin-RevId: 339505487
2020-10-27Add SHA512 to merkle tree libraryChong Cai
PiperOrigin-RevId: 339377254
2020-10-27Implement /proc/[pid]/memLennart
This PR implements /proc/[pid]/mem for `pkg/sentry/fs` (refer to #2716) and `pkg/sentry/fsimpl`. @majek COPYBARA_INTEGRATE_REVIEW=https://github.com/google/gvisor/pull/4060 from lnsp:proc-pid-mem 2caf9021254646f441be618a9bb5528610e44d43 PiperOrigin-RevId: 339369629
2020-10-27Assign VFS2 overlay device numbers based on layer device numbers.Jamie Liu
In VFS1's overlayfs, files use the device and inode number of the lower layer inode if one exists, and the upper layer inode otherwise. The former behavior is inefficient (requiring lower layer lookups even if the file exists and is otherwise wholly determined by the upper layer), and somewhat dangerous if the lower layer is also observable (since both the overlay and lower layer file will have the same device and inode numbers and thus appear to be the same file, despite being behaviorally different). VFS2 overlayfs imitates Linux overlayfs (in its default configuration) instead; it always uses the inode number from the originating layer, but synthesizes a unique device number for directories and another device number for non-directory files that have not been copied-up. As it turns out, the latter is insufficient (in VFS2, and possibly Linux as well), because a given layer may include files with different device numbers. If two distinct files on such a layer have device number X and Y respectively, but share inode number Z, then the overlay will map both files to some private device number X' and inode number Z, potentially confusing applications. Fix this by assigning synthetic device numbers based on the lower layer's device number, rather than the lower layer's vfs.Filesystem. PiperOrigin-RevId: 339300341
generated by cgit v1.2.3 (git 2.39.1) at 2024-12-21 09:55:52 +0000