gofer operations accumulate dentries touched in a slice to call
checkCachingLocked on them when the operation is over. In case
the same dentry is touched multiple times during the operation,
checkCachingLocked, and consequently destroyLocked, may be called
more than once for the same dentry.
Updates #1198
PiperOrigin-RevId: 305276819
Software GSO implementation currently has a complicated code path with
implicit assumptions that all packets to WritePackets carry same Data
and it does this to avoid allocations on the path etc. But this makes it
hard to reuse the WritePackets API.
This change breaks all such assumptions by introducing a new Vectorised
View API ReadToVV which can be used to cleanly split a VV into multiple
independent VVs. Further this change also makes packet buffers linkable
to form an intrusive list. This allows us to get rid of the array of
packet buffers that are passed in the WritePackets API call and replace
it with a list of packet buffers.
While this code does introduce some more allocations in the benchmarks
it doesn't cause any degradation.
Updates #231
PiperOrigin-RevId: 304731742
This change involves several steps:
- Refactor the VFS1 unix socket implementation to share methods between VFS1
and VFS2 where possible. Re-implement the rest.
- Override the default PRead, Read, PWrite, Write, Ioctl, Release methods in
FileDescriptionDefaultImpl.
- Add functions to create and initialize a new Dentry/Inode and FileDescription
for a Unix socket file.
Updates #1476
PiperOrigin-RevId: 304689796
NAME_MAX should be enforced per filesystem implementation
because other file systems may not have the same restriction.
Gofer filesystem now keeps a reference to the kernel clock to
avoid lookup in the Context on file access to update atime.
Update access, modification, and status change times in tmpfs.
Updates #1197, #1198.
PiperOrigin-RevId: 304527148
As per RFC 6980 section 5, nodes MUST silently ignore NDP messages if
the packet carrying them include an IPv6 Fragmentation Header.
Test: ipv6_test.TestNDPValidation
PiperOrigin-RevId: 304519379
This is mostly required for PipeTest_OffsetCalls.
The options are DenyPRead/PWrite rather than AllowPRead/PWrite since, in Linux
terms, fs/open.c:do_dentry_open sets FMODE_PREAD|FMODE_PWRITE unconditionally
(although it allows filesystem implementations of open to unset these flags),
so they're set for most FDs; it's usually FDs created outside of open(2) that
don't get them, e.g.:
- Syscall-created pipes (fs/pipe.c:create_pipe_files =>
fs/file_table.c:alloc_file_pseudo)
- Epoll instances (fs/eventpoll.c:do_epoll_create =>
fs/anon_inodes.c:anon_inode_getfile => alloc_file_pseudo)
- Sockets (net/socket.c:sock_alloc_file => alloc_file_pseudo)
This CL adds the flags to epoll instances; a subsequent CL reworks the VFS2
implementation of pipe FDs to be filesystem-independent and adds the flags
there, and sockets aren't implemented yet.
Updates #1035
PiperOrigin-RevId: 304506434
The test is flaky in cooperative S/R mode because TCP timers are not restored
across a S/R. This can cause the TCPLinger2 timer to not fire. This change
disables S/R before setting the TCP_LINGER2 timeout.
PiperOrigin-RevId: 304430536
Previously, go-marshal only allowed literals for array
lengths. However, it's very common for ABI structs to have a fix-sized
array whose length is defined by a constant; for example PATH_MAX.
Having to convert all such arrays to have literal lengths is too
awkward.
PiperOrigin-RevId: 304289345
Timeouts were increased to deflake pkg/tcpip/stack:stack_x_test tests
that depend on timers. Some timeouts used previously were intended for
tests that do not depend on timers, so this change updates those
timeouts to give more time for a timer-based event to occur. This
change also de-parallelizes non-subtests to reduce the number of active
timers.
Test: bazel test //pkg/tcpip/stack:stack_x_test --runs_per_test=500
PiperOrigin-RevId: 304287622
Refactor the existing socket interface to share methods between VFS1 and VFS2.
The method signatures do not contain anything filesystem-related, so they don't
need to be re-defined for VFS2.
Updates #1476, #1478, #1484, #1485.
PiperOrigin-RevId: 304184545
Previously, the import statement collision detection mechanism aborted
go-marshal whenever it detected two imports in any package that has
the same local name. Consider this trivial package, defined by the the
following two source files:
file1.go:
package example
import (
path/a/to/foo
)
...
file2.go:
package example
import (
another/package/with/final/component/foo
)
...
Go-marshal previously couldn't handle generating code for the the
above package, even if none of the types marked for marshalling used
either of the imported foo packages. This turns out to be too
restrictive as we run into this a lot in practice. Examples include
"encoding/binary" vs "gvisor/pkg/binary/binary", and "sync" vs
"gvisor/pkg/sync/sync".
This change allows go-marshal to proceed with marshalling, and only
abort if the code generated by go-marshal references any such
ambiguous import names.
PiperOrigin-RevId: 304131190
A socket mount where anonymous sockets will reside is added to the
VirtualFilesystem. Socketfs is built on top of kernfs.
Updates #1476, #1478, #1484, #1485.
PiperOrigin-RevId: 304095251
When copyUp fails, we attempt to clean up the upper filesystem by removing any
files that have already been copied-up. If the cleanup fails, we panic because
the "overlay filesystem is in an inconsistent state".
This CL adds the original copy-up error to the panic information, to hopefully
make it easier to track down how the overlay filesystem got into the
inconsistent state.
PiperOrigin-RevId: 304053370
In the case of other signals (preemption), inject a normal bounce and
defer the signal until the vCPU has been returned from guest mode.
PiperOrigin-RevId: 303799678
Both have analogues in Linux:
* struct file_system_type has a char *name field.
* struct super_block keeps a pointer to the file_system_type.
These fields are necessary to support the `filesystem type` field in
/proc/[pid]/mountinfo.
PiperOrigin-RevId: 303434063
Enables handling the Hop by Hop and Destination Options extension
headers, but options are not yet supported. All options will be
treated as unknown and their respective action will be followed.
Note, the stack does not yet support sending ICMPv6 error messages in
response to options that cannot be handled/parsed. That will come
in a later change (Issue #2211).
Tests:
- header_test.TestIPv6UnknownExtHdrOption
- header_test.TestIPv6OptionsExtHdrIterErr
- header_test.TestIPv6OptionsExtHdrIter
- ipv6_test.TestReceiveIPv6ExtHdrs
PiperOrigin-RevId: 303433085
BoundEndpointAt() is needed to support Unix sockets bound at a
file path, corresponding to BoundEndpoint() in VFS1.
Updates #1476.
PiperOrigin-RevId: 303258251
Using the host-defined file owner matches VFS1. It is more correct to use the
host-defined mode, since the cached value may become out of date. However,
kernfs.Inode.Mode() does not return an error--other filesystems on kernfs are
in-memory so retrieving mode should not fail. Therefore, if the host syscall
fails, we rely on a cached value instead.
Updates #1672.
PiperOrigin-RevId: 303220864
Fabricio Voznika
Dean Deng
Eyal Soha
Jamie Liu
Nicolas Lacasse
Bhasker Hariharan
Adin Scannell
Andrei Vagin
Tamir Duberstein
gVisor bot
Uros Prestor
Ghanan Gowripalan
Zach Koopmans
Rahat Mahmood
Michael Pratt