Ensure determinism

[damip]

To make sure execution is deterministic we need to:

• Disable wasm threads support (wasmer frontpage states that it supports wasm threads that are non deterministic )
• Canonicalize NaNs ( https://docs.rs/wasmer-compiler-singlepass-near/latest/wasmer_compiler_singlepass/struct.Singlepass.html#method.canonicalize_nans )
• Limit stack depth to a config defined value (I found this but dunno if it is related: https://docs.rs/wasmer-compiler-singlepass/latest/wasmer_compiler_singlepass/struct.Singlepass.html#method.enable_stack_check )
• Limit memory usage to a config defined value ( see https://github.com/wasmerio/wasmer/blob/c33b89c/examples/tunables_limit_memory.rs )
• uninitialized memory
• look for other sources of nondeterminism (?)

[yvan-sraka]

Could we write stress tests to any of the potential issues described here to see if we're able to produce examples of a bytecode with two runs in identical conditions that end up in different states?

It would be nice to explore the space of all possible SC to find edge cases!

[damip]

Yes that's something we should do. Of course we can't explore the whole SC space but we could try a few corner cases. Let's also look at how others set up the parameters of their wasmer interpreter (eg. near). This is a security-critical topic.

[sydhds]

Near disable all experimental wasm features (https://github.com/near/nearcore/blob/ea96faffff96010ef0a1faa58fe504e26f5e7bea/runtime/near-vm-runner/src/prepare.rs).

Not sure if all experimental features imply non determinism. Should we follow what Near do or investigate every features?

[damip]

The safest would be to have them all always disabled by config, and pin to a given wasmer version

[adrien-zinger]

PR for the NaN: #105

[damip]

Another question: what are those things fed with (time, RNG seed etc...) ?

• https://www.assemblyscript.org/stdlib/crypto.html
• https://www.assemblyscript.org/stdlib/process.html
• https://www.assemblyscript.org/stdlib/console.html
• https://www.assemblyscript.org/stdlib/date.html

Are we sure all those are deterministic ?

[damip]

@adrien-zinger do you mind taking a look ?