Are you sure this is safe? For fileio I think we just do a vec![0; bufsize]

The length is set by next line unsafe set_len and with_capacity just help it a bit. If read_exact has any problem, regardless of zero-filled vector or garbage filled vector(with set_len), it causes the same problem. (But I believe read_exact will work corrent)

Perhaps, though I'd prefer a vector with zeroes rather than UB, but if you think this is fine than feel free to use it.

This is neither UB nor unsafe. Please tell me if I am missing something.

Vec<T>::set_len is a vector version of MaybeUninit, so it may cause UB by the element type T and capaticy. In this case, the type T is u8. so it does not cause any type related unsafe case. About the capacity, we set it in this line. ("new_len must be less than or equal to capacity()." from https://doc.rust-lang.org/std/vec/struct.Vec.html#method.set_len) So this is successfully avoiding both of the unsafe case.
By meeting both of the criteria of set_len, even in the worst case, the result is just garbage bytes in vector. Any wrong u8 value in safe vector is not related to UB.

The next steps are also guaranteed to be safe.
In document, the Read trait requires (though it is not forced, I believe std followed the requirements) the buffer is expected to be read-only. ("It is recommended that implementations only write data to buf instead of reading its contents." from https://doc.rust-lang.org/std/io/trait.Read.html#method.read_exact)
So, we don't need to set it to any value because read_exact never use it. And after calling read_exact successfully, the vector will be full of new read data; otherwise the result will be Err.

Because set_len itself is not safe, I will put with_capacity into unsafe block and add a comment they must be coupled forever to help other people who will read this code next time.