update to Rust 1.78.0 / bump OPTE to 0.29.250

[iliana]

The most severe changes in this PR are dead code removal; I am not exactly sure why, but it seems like dead code detection became a little more robust between 1.77 and 1.78.

The 1.78.0 toolchain includes two new Clippy lints that are fixed in this PR:

• assigning_clones identifies cases where some value is set to the clone of another value where that type has a custom Clone::clone_from implementation. In this particular case we see this with Vec<_>, which has a more efficient implementation.
• unnecessary_get_then_check identifies uses of .get().is_some() / .get().is_none() on map types, suggesting .contains() / .contains_key() instead. (In one case this identified a HashSet contains-then-insert, which is unnecessary due to how HashSet::insert works.)

[iliana]

Interesting:

thread 'main' panicked at /home/build/.cargo/git/checkouts/opte-c3062ea19cee8fd4/7ee353a/lib/opte-ioctl/src/lib.rs:294:9:
assertion failed: rioctl.resp_len_actual != 0
note: run with `RUST_BACKTRACE=1` environment variable to display a backtrace

This appears to go away when I change the rust-toolchain.toml back.

[iliana]

OPTE fix at oxidecomputer/opte#527

[jgallagher]

Thanks - all the code changes LGTM. I'm a little nervous about the helios/deploy failure, but I don't see any smoking guns there - just NTP failing to sync, which I think is a known flake (#5598).

[iliana]

It has now happened twice in a row, which is kind of concerning (I hit retry once already).

[FelixMcFelix]

I had a quick peek at the logs there:

2024-05-15T16:15:46.446Z	+ pfexec /opt/oxide/opte/bin/opteadm dump-v2b
2024-05-15T16:15:46.446Z	Virtual to Boundary Mappings
2024-05-15T16:15:46.446Z	======================================================================
2024-05-15T16:15:46.446Z	
2024-05-15T16:15:46.446Z	IPv4 mappings
2024-05-15T16:15:46.446Z	----------------------------------------------------------------------
2024-05-15T16:15:46.446Z	TUNNELED PREFIX  BOUNDARY IP  VNI
2024-05-15T16:15:46.446Z	
2024-05-15T16:15:46.446Z	IPv6 mappings
2024-05-15T16:15:46.447Z	----------------------------------------------------------------------
2024-05-15T16:15:46.447Z	TUNNELED PREFIX  BOUNDARY IP  VNI
2024-05-15T16:15:46.447Z

I don't think this is a flake. Part of an OPTE API bump now is that you need to update maghemite and its binaries to a compatible version, as mgd also inserts V2B mappings using the opte-ioctl crate. The most obvious consequence of this mismatch is that if the API version changes, you lose external connectivity since OPTE ports are never told of upstream routes. Levon's RPW work handles the upgrade here, among other spots.

[iliana]

Aha. I will hold off on this until that can get merged then.