Add IP_RECVIF & IP_RECVDSTADDR. #1002

pusateri · 2018-12-28T11:14:11Z

Add IP_RECVIF & IP_RECVDSTADDR on freebsd, ios, macos, netbsd, openbsd
Include IP_PKTINFO on netbsd
Include IP6_PKTINFO on netbsd, openbsd.

FreeBSD/OpenBSD do not support IP_PKTINFO for IPv4 but use IP_RECVIF for interface index and use IP_RECVDSTADDR for destination address.
NetBSD and macOS also support IP_RECVIF and IP_RECVDSTADDR in addition to IP_PKTINFO for IPv4.
(For IPv6, all use IPV6_PKTINFO)

pusateri · 2018-12-30T00:43:02Z

Missing a constant in libc. Will be ready when libc pull request is merged rust-lang/libc#1189

pusateri · 2019-01-02T18:33:33Z

Ok, libc change has been merged. Not sure how to re-run the CI tests but no code changes should be needed.

Susurrus · 2019-01-02T19:27:16Z

I restarted Travis manually. Should've only done the one target, but I didn't realize it was only 1 until too late.

asomers · 2019-01-04T18:59:09Z

I'm afraid you'll need to rebase, due to CI changes.

pusateri · 2019-01-05T01:52:53Z

Just tried this myself and I can't get the doc test to fail on FreeBSD 11.2 amd64. I notice there's a cfg option to turn this doc test off for i386. What's that all about?

% freebsd-version
11.2-RELEASE-p5

% rustup show
Default host: x86_64-unknown-freebsd

installed toolchains
--------------------

stable-x86_64-unknown-freebsd
1.24.1-x86_64-unknown-freebsd (default)

active toolchain
----------------

1.24.1-x86_64-unknown-freebsd (default)
rustc 1.24.1 (d3ae9a9e0 2018-02-27)

% cargo test --doc
    Finished dev [unoptimized + debuginfo] target(s) in 0.0 secs
   Doc-tests nix

running 75 tests
test src/pty.rs - pty::posix_openpt (line 92) ... ok
test src/lib.rs - Error::as_errno (line 117) ... ok
test src/ifaddrs.rs - ifaddrs::getifaddrs (line 112) ... ok
test src/sys/aio.rs - sys::aio::AioCb<'a>::cancel (line 728) ... ok
test src/sys/aio.rs - sys::aio::AioCb<'a>::error (line 783) ... ok
test src/sys/aio.rs - sys::aio::AioCb<'a>::from_boxed_slice (line 365) ... ok
test src/sys/aio.rs - sys::aio::AioCb<'a>::from_boxed_mut_slice (line 482) ... ok
test src/sys/aio.rs - sys::aio::AioCb<'a>::from_boxed_slice (line 397) ... ok
test src/sys/aio.rs - sys::aio::AioCb<'a>::from_boxed_slice (line 423) ... ok
test src/sys/aio.rs - sys::aio::AioCb<'a>::from_fd (line 228) ... ok
test src/sys/aio.rs - sys::aio::AioCb<'a>::from_mut_slice (line 289) ... ok
test src/sys/aio.rs - sys::aio::LioCb<'a>::listio_resubmit (line 1156) ... ok
test src/sys/aio.rs - sys::aio::AioCb<'a>::from_slice (line 644) ... ok
test src/sys/aio.rs - sys::aio::aio_suspend (line 981) ... ok
test src/sys/ioctl/mod.rs - ioctl_none (line 269) ... ignored
test src/sys/ioctl/mod.rs - ioctl_none (line 287) ... ok
test src/sys/ioctl/mod.rs - ioctl_none_bad (line 310) ... ignored
test src/sys/ioctl/mod.rs - ioctl_none_bad (line 319) ... ok
test src/sys/ioctl/mod.rs - ioctl_read (line 353) ... ignored
test src/sys/ioctl/bsd.rs - request_code_none (line 41) ... ok
test src/sys/ioctl/mod.rs - ioctl_read_bad (line 390) ... ignored
test src/sys/aio.rs - sys::aio::LioCb<'a>::listio (line 1099) ... ok
test src/sys/ioctl/mod.rs - ioctl_read_buf (line 694) ... ignored
test src/sys/ioctl/mod.rs - ioctl_readwrite (line 629) ... ignored
test src/sys/ioctl/mod.rs - ioctl_read (line 362) ... ok
test src/sys/ioctl/mod.rs - ioctl_readwrite_bad (line 665) ... ignored
test src/sys/ioctl/mod.rs - ioctl_readwrite_buf (line 762) ... ignored
test src/sys/ioctl/mod.rs - ioctl_write_buf (line 723) ... ignored
test src/sys/aio.rs - sys::aio::aio_cancel_all (line 927) ... ok
test src/sys/ioctl/mod.rs - ioctl_write_int (line 503) ... ignored
test src/sys/ioctl/mod.rs - ioctl_read_bad (line 399) ... ok
test src/sys/ioctl/mod.rs - ioctl_write_int_bad (line 584) ... ignored
test src/sys/ioctl/mod.rs - ioctl_readwrite (line 638) ... ok
test src/sys/ioctl/mod.rs - ioctl_write_int_bad (line 593) ... ok
test src/sys/ioctl/mod.rs - ioctl_write_ptr (line 428) ... ignored
test src/sys/ioctl/mod.rs - ioctl_write_int (line 516) ... ok
test src/sys/ioctl/mod.rs - ioctl_write_ptr_bad (line 464) ... ignored
test src/sys/ioctl/mod.rs - ioctl_write_buf (line 732) ... ok
test src/sys/ioctl/mod.rs - ioctl_write_int_bad (line 602) ... ok
test src/sys/ioctl/mod.rs - ioctl_write_ptr (line 437) ... ok
test src/sys/ioctl/mod.rs - sys::ioctl (line 125) ... ok
test src/sys/ioctl/mod.rs - ioctl_write_ptr_bad (line 473) ... ok
test src/sys/ioctl/mod.rs - sys::ioctl (line 167) ... ok
test src/sys/ioctl/mod.rs - sys::ioctl (line 179) ... ok
test src/sys/ioctl/mod.rs - sys::ioctl (line 217) ... ok
test src/sys/ioctl/mod.rs - sys::ioctl (line 56) ... ok
test src/sys/ioctl/mod.rs - sys::ioctl (line 67) ... ok
test src/sys/ioctl/mod.rs - sys::ioctl (line 92) ... ok
test src/sys/ptrace/bsd.rs - sys::ptrace::bsd::step (line 123) ... ok
test src/sys/select.rs - sys::select::FdSet::highest (line 49) ... ok
test src/sys/socket/mod.rs - sys::socket::CmsgSpace (line 356) ... ok
test src/sys/socket/mod.rs - sys::socket::setsockopt (line 1214) ... ok
test src/sys/termios.rs - sys::termios (line 108) ... ok
test src/sys/termios.rs - sys::termios (line 122) ... ok
test src/sys/termios.rs - sys::termios (line 25) ... ok
test src/sys/termios.rs - sys::termios (line 137) ... ok
test src/sys/socket/mod.rs - sys::socket::ControlMessage::ScmTimestamp (line 473) ... ok
test src/sys/termios.rs - sys::termios (line 94) ... ignored
test src/sys/termios.rs - sys::termios (line 40) ... ok
test src/sys/termios.rs - sys::termios::Termios::get_libc_termios (line 200) ... ok
test src/sys/termios.rs - sys::termios (line 65) ... ok
test src/sys/termios.rs - sys::termios (line 79) ... ok
test src/sys/wait.rs - sys::wait::WaitStatus::from_raw (line 173) ... ok
test src/unistd.rs - unistd::close (line 902) ... ok
test src/unistd.rs - unistd::fork (line 192) ... ok
test src/unistd.rs - unistd::alarm (line 1499) ... ok
test src/unistd.rs - unistd::gethostname (line 873) ... ok
test src/unistd.rs - unistd::initgroups (line 1443) ... ok
test src/unistd.rs - unistd::close (line 916) ... ok
test src/unistd.rs - unistd::getcwd (line 544) ... ok
test src/unistd.rs - unistd::mkstemp (line 1611) ... ok
test src/unistd.rs - unistd::mkdir (line 438) ... ok
test src/unistd.rs - unistd::setgroups (line 1310) ... ok
test src/unistd.rs - unistd::mkfifo (line 481) ... ok
test src/unistd.rs - unistd::alarm (line 1513) ... ok

test result: ok. 61 passed; 0 failed; 14 ignored; 0 measured; 0 filtered out

pusateri · 2019-01-06T02:50:49Z

There may be a problem with the test environment.

asomers · 2019-01-06T04:12:19Z

The test is not really valid. It's basically asserting that the sleep plus the 2nd sendmsg call take no more than 255ms, when it reality they took 275ms. But there's nothing that guarantees that. In a non-real time operating system, you could be delayed at any time. The test is also wildly overcomplicated. I'll fix it.

asomers · 2019-01-07T16:13:53Z

test/sys/test_socket.rs

+                .intersects(MsgFlags::MSG_TRUNC | MsgFlags::MSG_CTRUNC)
+        );
+
+        for cmsg in msg.cmsgs() {


There's nothing to ensure that cmsgs were actually received. Shouldn't you have some kind of assertion to that effect?

Your latest updated didn't address this issue. Do you plan to?

Do you just want something like this:

assert_eq!(msg.cmsgs().count(), 2, "expected 2 cmsgs");

or did you have something different in mind?

That statement wouldn't check that you got the right types of cmsgs. I think you need to check that the cmsgs were of the right types, too.

the match condition '_' in the for loop would assert if it was any other type. So this would ensure there were two messages and the types were either Ipv4RecvIf or Ipv4RecvDstAddr. Since the kernel can't send two of the same type, I think this is sufficient, no?

I think the test should verify that we got one of each. Otherwise it wouldn't catch a regression that decodes the types incorrectly.

Ok, just added some code that I think covers this. So now, I check to make sure they're are only two and I check to make sure each type is received and it fails if something other than the expected two message types are received.

asomers · 2019-01-07T16:15:02Z

test/sys/test_socket.rs

+    ).expect("send socket failed");
+    sendmsg(send, &iov, &[], MsgFlags::empty(), Some(&sa)).expect("sendmsg failed");
+
+    thread.join().unwrap();


I don't think you need a separate thread. sendmsg(2) should return before recvmsg(2) gets called.

asomers · 2019-01-07T16:15:54Z

test/sys/test_socket.rs

+    use std::io::Write;
+    use std::thread;
+
+    fn loopback_v4addr() -> Option<InterfaceAddress> {


I'm getting deja vu here. Didn't a recent PR of yours contain something similar or identical to this function? If so, could you make it common code?

asomers

Ok, I think this is looking good now! It just needs a CHANGELOG entry and a squash.

Include IP_PKTINFO and IP6_PKTINFO on netbsd/openbsd.

pusateri · 2019-01-25T23:05:13Z

Anything else I need to do here?

asomers · 2019-01-25T23:07:04Z

I've restarted the Cirrus test. We'll see if it actually runs this time.

pusateri · 2019-01-30T21:19:29Z

Still is still ready to merge and the CI is still failing for no good reason. I don't think the FreeBSD CI change is working well.

asomers

Ok, the third time was the charm. I've opened an issue upstream with Cirrus; we'll see what they say.

bors r+

1002: Add IP_RECVIF & IP_RECVDSTADDR. r=asomers a=pusateri Add IP_RECVIF & IP_RECVDSTADDR on freebsd, ios, macos, netbsd, openbsd Include IP_PKTINFO on netbsd Include IP6_PKTINFO on netbsd, openbsd. FreeBSD/OpenBSD do not support IP_PKTINFO for IPv4 but use IP_RECVIF for interface index and use IP_RECVDSTADDR for destination address. NetBSD and macOS also support IP_RECVIF and IP_RECVDSTADDR in addition to IP_PKTINFO for IPv4. (For IPv6, all use IPV6_PKTINFO) Co-authored-by: Tom Pusateri <pusateri@bangj.com>

bors · 2019-01-30T22:27:37Z

Build failed

FreeBSD 11.2

asomers · 2019-01-30T22:36:35Z

bors retry

1002: Add IP_RECVIF & IP_RECVDSTADDR. r=asomers a=pusateri Add IP_RECVIF & IP_RECVDSTADDR on freebsd, ios, macos, netbsd, openbsd Include IP_PKTINFO on netbsd Include IP6_PKTINFO on netbsd, openbsd. FreeBSD/OpenBSD do not support IP_PKTINFO for IPv4 but use IP_RECVIF for interface index and use IP_RECVDSTADDR for destination address. NetBSD and macOS also support IP_RECVIF and IP_RECVDSTADDR in addition to IP_PKTINFO for IPv4. (For IPv6, all use IPV6_PKTINFO) Co-authored-by: Tom Pusateri <pusateri@bangj.com>

bors · 2019-01-30T23:28:07Z

Build succeeded

continuous-integration/travis-ci/push
FreeBSD 11.2

asomers · 2019-01-31T04:13:13Z

FWIW The Cirrus people say they've fixed the bug now.

pusateri · 2019-01-31T04:28:39Z

Great. Thanks! False negatives waste a lot of time.

This was an oversight from PR nix-rust#1002

pusateri mentioned this pull request Jan 4, 2019

Consider supporting advanced networking APIs tokio-rs/tokio#799

Closed

asomers reviewed Jan 7, 2019

View reviewed changes

asomers requested changes Jan 23, 2019

View reviewed changes

Add IP_RECVIF & IP_RECVDSTADDR.

147f791

Include IP_PKTINFO and IP6_PKTINFO on netbsd/openbsd.

asomers approved these changes Jan 30, 2019

View reviewed changes

bors bot merged commit 147f791 into nix-rust:master Jan 30, 2019

pusateri deleted the recvif branch January 31, 2019 04:28

asomers added a commit to asomers/nix that referenced this pull request Jan 31, 2019

Enable Ipv6PacketInfo on more OSes.

8220a19

This was an oversight from PR nix-rust#1002

asomers added a commit to asomers/nix that referenced this pull request Jan 31, 2019

Enable Ipv6PacketInfo on more OSes.

8eb3c9b

This was an oversight from PR nix-rust#1002

asomers added a commit to asomers/nix that referenced this pull request Feb 12, 2019

Enable IPv4PacketInfo and Ipv6PacketInfo on more OSes.

2125d63

This was an oversight from PR nix-rust#1002

asomers added a commit to asomers/nix that referenced this pull request Feb 12, 2019

Enable IPv4PacketInfo and Ipv6PacketInfo on more OSes.

f11fe7f

This was an oversight from PR nix-rust#1002

asomers added a commit to asomers/nix that referenced this pull request Feb 14, 2019

Enable IPv4PacketInfo and Ipv6PacketInfo on more OSes.

b5c4c7a

This was an oversight from PR nix-rust#1002

vdagonneau pushed a commit to vdagonneau/nix that referenced this pull request Feb 20, 2019

Enable IPv4PacketInfo and Ipv6PacketInfo on more OSes.

abbd9b0

This was an oversight from PR nix-rust#1002

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Add IP_RECVIF & IP_RECVDSTADDR. #1002

Add IP_RECVIF & IP_RECVDSTADDR. #1002

pusateri commented Dec 28, 2018

pusateri commented Dec 30, 2018 •

edited

Loading

pusateri commented Jan 2, 2019

Susurrus commented Jan 2, 2019

asomers commented Jan 4, 2019

pusateri commented Jan 5, 2019

pusateri commented Jan 6, 2019

asomers commented Jan 6, 2019 •

edited

Loading

asomers Jan 7, 2019

asomers Jan 13, 2019

pusateri Jan 14, 2019

asomers Jan 14, 2019

pusateri Jan 20, 2019 •

edited

Loading

asomers Jan 20, 2019

pusateri Jan 22, 2019

asomers Jan 7, 2019

asomers Jan 7, 2019

asomers left a comment

pusateri commented Jan 25, 2019

asomers commented Jan 25, 2019

pusateri commented Jan 30, 2019

asomers left a comment

bors bot commented Jan 30, 2019

asomers commented Jan 30, 2019

bors bot commented Jan 30, 2019

asomers commented Jan 31, 2019

pusateri commented Jan 31, 2019

Add IP_RECVIF & IP_RECVDSTADDR. #1002

Add IP_RECVIF & IP_RECVDSTADDR. #1002

Conversation

pusateri commented Dec 28, 2018

pusateri commented Dec 30, 2018 • edited Loading

pusateri commented Jan 2, 2019

Susurrus commented Jan 2, 2019

asomers commented Jan 4, 2019

pusateri commented Jan 5, 2019

pusateri commented Jan 6, 2019

asomers commented Jan 6, 2019 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

pusateri Jan 20, 2019 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

asomers left a comment

Choose a reason for hiding this comment

pusateri commented Jan 25, 2019

asomers commented Jan 25, 2019

pusateri commented Jan 30, 2019

asomers left a comment

Choose a reason for hiding this comment

bors bot commented Jan 30, 2019

Build failed

asomers commented Jan 30, 2019

bors bot commented Jan 30, 2019

Build succeeded

asomers commented Jan 31, 2019

pusateri commented Jan 31, 2019

pusateri commented Dec 30, 2018 •

edited

Loading

asomers commented Jan 6, 2019 •

edited

Loading

pusateri Jan 20, 2019 •

edited

Loading