mirror of
https://github.com/outbackdingo/firezone.git
synced 2026-01-27 18:18:55 +00:00
d1d0874699
Within Firezone's Rust codebase, we use the `etherparse` crate extensively to parse network packets. To provide a more ergonomic API, this is all encapsulated in our `ip-packet` crate. For #7518, we need to write an eBPF kernel that parses and manipulates network packets. Etherparse itself doesn't provide any facilities to manipulate network packets. That is an open feature request: https://github.com/JulianSchmid/etherparse/issues/9. For the packet manipulation that we are doing in `connlib`, we already wrote certain extensions to the `etherparse` crate but today, those are all within the `ip-packet` crate. In order to reuse that within the eBPF kernel, we cannot just depend on `ip-packet` directly because eBPF is a no-std and no-alloc environment, thus no crate in the dependency tree is allowed to depend on Rust's std-lib. `etherparse` itself actually has an `std` feature flag that we can turn off. Introducing the same in `ip-packet` would require a lot of conditional-compilation gates using `#[cfg]`. it is much easier to just introduce a new crate that houses all our in-house extensions to `etherparse`. Eventually, we can hopefully upstream those which is another motivator to separate this out.
18 lines
495 B
Rust
18 lines
495 B
Rust
/// Writes the given byte-array to the slice at the specified index.
|
|
///
|
|
/// # Safety
|
|
///
|
|
/// `offset` + `bytes.len()` must be within the slice.
|
|
pub unsafe fn write_to_offset_unchecked<const N: usize>(
|
|
slice: &mut [u8],
|
|
offset: usize,
|
|
bytes: [u8; N],
|
|
) {
|
|
debug_assert!(offset + N <= slice.len());
|
|
|
|
let (_front, rest) = unsafe { slice.split_at_mut_unchecked(offset) };
|
|
let (target, _rest) = unsafe { rest.split_at_mut_unchecked(N) };
|
|
|
|
target.copy_from_slice(&bytes)
|
|
}
|