introduce multicore load initial corpus

[R9295]

implements and closes #1746
I ended up implementing a multicore version unlike what I wrote in the issue. It traverses the corpus entries twice (at most), which is inefficient but is probably the simplest implementation I could think of.

to test:

cd fuzzers/libfuzzer_libpng_norestart
RUST_LOG="debug" LIBAFL_DEBUG_OUTPUT=1 ./fuzzer_libpng --cores 1,2 --input seeds > debug
cat debug

[domenukk]

Thank you! Looks like very good already! However, we need to consider restarts during testcase evaluation time. You probably want to store all inputs to be processed in a Vec in the state for the current core, and if the vec already exists (after restart) skip over this initial loading, and continue processing the vec (?)

[R9295]

@domenukk Thanks, I will update it.
A question though: why is it built this way? If it crashes or has a timeout, wouldn't that be achieving a "solution"

[domenukk]

It follows AFL++'s (optional) concept of ignoring timeouts and crashes in the initial seed corpus, since that stuff is already stored.
We could consider optionally adding it to the solutions of course, it might be useful for someuse cases, who knows?

[domenukk]

Run cd libafl && cargo test --no-default-features
   Compiling libafl_bolts v0.11.2 (/home/runner/work/LibAFL/LibAFL/libafl_bolts)
   Compiling libafl v0.11.2 (/home/runner/work/LibAFL/LibAFL/libafl)
error[E0560]: struct `StdState<I, C, R, SC>` has no field named `multicore_inputs_processed`
    --> libafl/src/state/mod.rs:10[8](https://github.com/AFLplusplus/LibAFL/actions/runs/8157117940/job/22296046890?pr=1905#step:14:9)3:13
     |
[10](https://github.com/AFLplusplus/LibAFL/actions/runs/8157117940/job/22296046890?pr=1905#step:14:11)83 |             multicore_inputs_processed: None,
     |             ^^^^^^^^^^^^^^^^^^^^^^^^^^ `StdState<I, C, R, SC>` does not have this field
     |
     = note: all struct fields are already assigned

Looks like no_std is broken now

[R9295]

@domenukk should be good now

[domenukk]

error: item in documentation is missing backticks
   --> libafl/src/state/mod.rs:332:27
    |
332 |     /// relevant only for load_multicore_inputs

Last tiny things in docs, else looks good to go! :)

[domenukk]

One last cargo +nightly fmt in the example fuzzer you changed 😬

[R9295]

@domenukk not sure why one of the runs is failing..

[tokatoka]

the error is due to my previous changes..

204 |         let mut executor = InProcessExecutor::batched_timeouts(
    |                                               ^^^^^^^^^^^^^^^^
    |                                               |
    |                                               function or associated item not found in `GenericInProcessExecutor<_, &mut _, (), _, _>`
    |                                               help: there is an associated function with a similar name: `batched_timeout`

can you change this to batched_timeout as suggested?

[tokatoka]

i'll fix it

[domenukk]

Thanks!