Age | Commit message (Collapse) | Author |
|
Travis has been super-slow lately (> 6 hours per build). Cirrus is much
faster: about 20 minutes. Cirrus also has slightly better test
coverage, mainly because it doesn't use SECCOMP.
Also,
* Fix the Redox CI build. The old Travis configuration didn't actually
build for Redox, so we never noticed that Redox can't be built with a
stable compiler. Thanks to @coolreader18 for finding this.
* Disable the udp_offload tests on cross-tested platforms. These tests
are failing with ENOPROTOOPT in Cirrus-CI. I suspect it's due to a
lack of support in QEMU. These tests were skipped on Travis because
its kernel was too old.
* Fix require_kernel_version on Cirrus-CI. Cirrus reports the Linux
kernel version as 4.19.112+, which the semver crate can't handle.
* Fix test_setfsuid on Cirrus. When run on Cirrus, it seems like the
file in /tmp gets deleted as soon as it's closed. Probably an
overzealous temporary file cleaner. Use /var/tmp, because no
temporary file cleaner should run in there.
* Skip mount tests on Cirrus. They fail for an unknown reason.
Issue #1351
* Skip the AF_ALG tests on Cirrus-CI
Issue #1352
|
|
There were two problems discovered with the `recvmmsg(2)` implementation
that this changeset attempts to fix:
1. As mentioned in nix-rust/issues/1325, `recvmmsg(2)` can return fewer
messages than requested, and
2. Passing the return value of `recvmmsg(2)` as the number of bytes in
the messages received is incorrect.
This changeset incorporates the proposed fix from nix-rust/issues/1325,
as well as passing the correct value (`mmsghdr.msg_len`) for the number
of bytes in a given message.
|
|
Fix deprecation warning from libc update.
|
|
This adds Ipv4PacketInfo and Ipv6PacketInfo to ControlMessage,
allowing these to be used with sendmsg/sendmmsg.
This change contains the following squashed commits:
Add Ipv{4,6}PacketInfo to ControlMessage.
Add documentation links to Ipv{4,6}PacketInfo
Add changelog entry for Ipv{4,6}PacketInfo
Add link to PR in the Changelog.
Add extra build environments.
Add tests for Ipv{4,6}PacketInfo.
Swap #[test] and #[cfg]
The CI appears to be running the test, even though it's not cfg'd for
that platform. I _think_ this might be due to these being in the wrong
order. So lets try swapping them.
s/freebsd/netbsd/ for Ipv4PacketInfo
netbsd supports in_pktinfo, not freebsd.
Fix the cfg for Ipv{4,6}PacketInfo usage.
Ah, I see what I did wrong. I had fixed the definitions, but I had the
wrong cfg() in the usage. This has the usage match the definitions.
Change SOL_IPV6 to IPPROTO_IPV6.
FreeBSD doesn't have SOL_IPV6, but does have IPPROTO_IPV6, and the two
constants are defined as being equal. So change to use IPPROTO_IPV6.
Skip Ipv6PacketInfo test if v6 is not available.
If IPv6 is not available, then when we try and bind to ip6-localhost,
we'll get a EADDRNOTAVAIL, so skip the test.
This should mean that the test will run on any machine that has a v6
loopback address.
More architecture cfg() fixes.
These all need to be the same, and they were not. Make them them all
the same. Attempt III.
Fix up mismatched cfg's again.
Take IV. Make sure the cfg's that use a enum variant match the enum
definition.
|
|
sendmmsg() / recvmmsg() in #1208.
In #1208, sendmmsg() / recvmmsg() were added, but OpenBSD(who doesn't support these)
was included on the list of allowed operating systems for sendmmsg() related things.
This broke the build on OpenBSD.
For more Rust-world examples, see: https://github.com/rust-lang/libc/commit/6f6297301d49ff67b8ca044d651f36a56950298c
|
|
|
|
|
|
|
|
|
|
|
|
recvfrom(2) only returns the sender's address for protocols that provide
it. Usually, that means it returns the sender's address for datagram
sockets but not for stream sockets.
Fixes #1144
|
|
Travis is now using Seccomp, and Docker's default Seccomp policy
disables execveat (though, weirdly, not fexecve). It also prohibits any
operations on AF_ALG sockets.
While I'm here, replace close/dup with dup2, which is more reliable.
Also, drop the fork mutex earlier. This way all of the exeve tests will
run, even if one fails.
https://docs.docker.com/engine/security/seccomp/
|
|
NULL bytes have no special significance in an abstrace address, and the
length of the address is solely decided by the length member. If the
length is set to sun_path.len(), all the NULL bytes will be considered
part of the address.
Tests are updated accordingly.
Closes #1119
Signed-off-by: Yuxuan Shui <yshuiv7@gmail.com>
|
|
1095: Fix warnings on Rust 1.37.0 r=asomers a=asomers
* Replace obsolete range syntax "..." with inclusive range "..="
* Use dyn Trait syntax instead of Box<Trait>
* Raise MSRV to 1.27.0 (for dyn Trait syntax)
* Test with nightly again
Co-authored-by: Alan Somers <asomers@gmail.com>
|
|
* Replace obsolete range syntax "..." with inclusive range "..="
* Use dyn Trait syntax instead of Box<Trait>
* Raise MSRV to 1.27.0 (for dyn Trait syntax)
* Raise MSRV to 1.31.0 (because of rand)
tempfile pulls in rand, and rand pulls in fuchsia-cprng, which requires
1.31.0. Why rand pulls in fuchsia-cprng I don't know. It's specified
as a target-specific dependency, but Cargo tries to build it anyway
(only on Linux, not on FreeBSD or OSX). A bug in Cargo 1.27.0?
|
|
The current VSOCK implementation does not support loopback devices,
so, for now, we expect a failure in the spawned thread when it
tries to connect.
Signed-off-by: Stefano Garzarella <sgarzare@redhat.com>
|
|
Abstract paths should always be N-1 in length where N is the length of
the `sun_path` field (first byte is \0). Given that,
`UnixAddr::new_abstract()` should always return this N-1 length, not
just the length of the string provided (the rest of the array will be
\0s).
|
|
Derive Clone, Copy, Eq, Hash, and PartialEq for all types. Not all
traits are supported by all types, which is why many are missing
some.
|
|
received using recvmsg.
|
|
|
|
On some platforms the alignment of cmsg_data could be less than the
alignment of the messages that it contains. That led to unaligned
reads
on those platforms. This change fixes the issue by always copying the
message contents into aligned objects. The change is not 100%
backwards
compatible when using recvmsg. Users may have to replace code like
this:
```rust
if let ControlMessage::ScmRights(&fds) = cmsg {
```
with this:
```rust
if let ControlMessageOwned::ScmRights(fds) = cmsg {
```
Fixes #999
|
|
CmsgSpace had three problems:
1) It would oversize buffers that expect multiple control messages
2) It didn't use the libc CMSG_SPACE(3) macro, so it might actually
undersize a buffer for a single control message.
3) It could do bad things on drop, if you instantiate it with a type
that implements Drop (which none of the currently supported
ControlMessage types do).
Fixes #994
|
|
There were two problems:
1) It would always return Ok, even on error
2) It could panic if there was an error, because
sockaddr_storage_to_addr would be called on uninitialized memory.
|
|
Include IP_PKTINFO and IP6_PKTINFO on netbsd/openbsd.
|
|
993: Don't reference packed structs. r=asomers a=pusateri
Fixes #992.
Don't merge this yet. I have more testing to do. I just am pushing it up for others.
Co-authored-by: Tom Pusateri <pusateri@bangj.com>
|
|
Some tests have been disabled and will need further review.
|
|
ignore pktinfo tests on qemu mips,mips64,powerpc64
Original work by @mcginty.
|
|
Signed-off-by: Igor Gnatenko <i.gnatenko.brain@gmail.com>
|
|
|
|
|
|
|
|
Rather than using the native implementation of these constants
on supported platforms, the native implementation was instead
emulated. This was also hidden from the user even though this
could result in data races and the functionality being broken.
Native functionality is, however, not support on macos/ios.
Rather than enable this emulation solely for this platform, it
should be removed as this is a dangerous abstraction.
|
|
|
|
cc #664 (unsure if this is everything needed)
|
|
The libc_bitflags! macro was replaced with a non-recursive one supporting
only public structs. I could not figure out how to make the old macro work
with the upgrade, so I reworked part of the bitflags! macro directly to suit
our needs, much as the original recursive macro was made. There are no uses
of this macro for non-public structs, so this is not a problem for internal code.
|
|
This introduces an `as_abstract()` getter to `UnixAddr` in order to
retrieve the name of an abstract unix socket.
This also adds tests around abstract addresses and clarify docs,
adding explicit semantics.
|
|
|
|
Most could be replaced by simple raw pointer casts (or even perfectly
safe coercions!).
cc #373
|
|
|
|
|
|
|
|
test_getsockname used an IPv4 socket and assumed that localhost was
"127.0.0.1". But that assumption doesn't hold on IPv6-only hosts or on
shared-IP FreeBSD jails. Unfortunately, the Rust standard library
doesn't provide a good way to resolve localhost. So change the test to
use a unix-domain socket instead.
|
|
There were multiple errors regarding Unix domain sockets:
* UnixAddr::path assumed that gethostbyname and similar functions would
include the terminating null as part of len. That is not universally
true. In fact, POSIX only guarantees that len will be at least large
enough to store the non-null-terminated path. So it could be larger or
smaller than nix was assuming. Since abstract sockets' paths are not
strings, we can't modify gethostbyname. Instead, I implemented the fix in
UnixAddr::path and UnixAddr::new. I clarified the documentation too.
* SockAddr::as_ffi_pair contained a Linuxism.
* sockaddr_storage_to_addr forgot to adjust sun_len when creating a UnixAddr
|
|
Fixes #329
|
|
|
|
|
|
The returned length of AF_UNIX sockaddrs is significant, and generally
does not match the length of the entire structure. For filesystem
sockets, this is ignorable because the path is also NUL-terminated, but
for unbound sockets (e.g., a socketpair) or abstract-namespace
sockets (a Linux extension where the address is an arbitrary
bytestring), we need to keep track of the length.
Fixes #177. Also add a UnixAddr::new_abstract function and some better
handling of abstract-namespace socket addresses to fix #169.
|
|
The best specification for control message layout appears to be
[RFC 2292, section 4](https://tools.ietf.org/html/rfc2292#section-4),
despite this not being a wire protocol. These definitions have also been
checked against glibc 2.19 <bits/socket.h> and Linux 4.0
<linux/socket.h>, and tested on Debian 8.1 and FreeBSD 10.2 x86_64.
The API differs a bit from the cmsg(3) API for type-safety reasons (and
also because the cmsg(3) API is terrible). See test/sys/test_socket.rs
for an example.
Only supports SCM_RIGHTS at the moment.
Fixes #88.
|
|
This reverts commit 046af7d1ba82506f9bc48e62ac0584361025fc02.
|
|
The best specification for control message layout appears to be
[RFC 2292, section 4](https://tools.ietf.org/html/rfc2292#section-4),
despite this not being a wire protocol. These definitions have also been
checked against glibc 2.19 <bits/socket.h> and Linux 4.0
<linux/socket.h>, and tested on Debian 8.1 and FreeBSD 10.2 x86_64.
The API differs a bit from the cmsg(4) API for type-safety reasons (and
also because the cmsg(4) API is terrible). See test/sys/test_socket.rs
for an example.
Only supports SCM_RIGHTS at the moment.
Fixes #88.
|