Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

Bump module serialization format to v2 #1223

Merged
merged 2 commits into from Feb 8, 2022
Merged

Bump module serialization format to v2 #1223

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
c2d807b
Create test: file_system_cache_store_uses_expected_path
webmaster128 Feb 8, 2022
4ffa110
Bump module serialization format to v2
webmaster128 Feb 8, 2022
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Jump to
Jump to file
Failed to load files.
Diff view
Diff view
2 changes: 2 additions & 0 deletions CHANGELOG.md
View file
Open in desktop
Expand Up @@ -21,6 +21,8 @@ and this project adheres to
### Fixed

- cosmwasm-vm: Fix `AddAssign` implementation of `GasInfo`.
- cosmwasm-vm: Bump `MODULE_SERIALIZATION_VERSION` to "v2" because the module
serialization format changed between Wasmer 2.0.0 and 2.1.x.

## [1.0.0-beta4] - 2021-12-23

Expand Down
55 changes: 44 additions & 11 deletions packages/vm/src/modules/file_system_cache.rs
View file
Open in desktop
Expand Up @@ -14,7 +14,23 @@ use crate::errors::{VmError, VmResult};
/// The string is used as a folder and should be named in a way that is
/// easy to interprete for system admins. It should allow easy clearing
/// of old versions.
const MODULE_SERIALIZATION_VERSION: &str = "v1";
///
/// See https://github.com/wasmerio/wasmer/issues/2781 for more information
/// on Wasmer's module stability concept.
///
/// ## Version history:
/// - **v1**:<br>
/// cosmwasm_vm < 1.0.0-beta5. This is working well up to Wasmer 2.0.0 as
/// [in wasmvm 1.0.0-beta2](https://github.com/CosmWasm/wasmvm/blob/v1.0.0-beta2/libwasmvm/Cargo.lock#L1412-L1413)
/// and [wasmvm 0.16.3](https://github.com/CosmWasm/wasmvm/blob/v0.16.3/libwasmvm/Cargo.lock#L1408-L1409).
/// Versions that ship with Wasmer 2.1.x such [as wasmvm 1.0.0-beta3](https://github.com/CosmWasm/wasmvm/blob/v1.0.0-beta3/libwasmvm/Cargo.lock#L1534-L1535)
/// to [wasmvm 1.0.0-beta5](https://github.com/CosmWasm/wasmvm/blob/v1.0.0-beta5/libwasmvm/Cargo.lock#L1530-L1531)
/// are broken, i.e. they will crash when reading older v1 modules.
/// - **v2**:<br>
/// Version for cosmwasm_vm 1.0.0-beta5 / wasmvm 1.0.0-beta6 that ships with Wasmer 2.1.1.
/// - **v3**:<br>
/// Version for Wasmer 2.2.0 which contains a [module breaking change to 2.1.x](https://github.com/wasmerio/wasmer/pull/2747).
const MODULE_SERIALIZATION_VERSION: &str = "v2";
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Can't this be derived or imported from wasmer somehow? That will make this code support all versions transparently (new test below would have to me modified).

Copy link
Member Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

No, the version is not exposed yet. This might change in the future. They added a versioning schema in wasmerio/wasmer#2747. However, it is not exported. So currently they just say it can change at any time.

If we get the version, we can check much better that we don't miss an update. However, we might want to keep control over the version number since it can include our own format changes as well. E.g. we could rename the file names from <checksum> to <checksum>.module or compress the modules before writing them to disk.

Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

If we get the version, we can check much better that we don't miss an update. However, we might want to keep control over the version number since it can include our own format changes as well. E.g. we could rename the file names from <checksum> to <checksum>.module or compress the modules before writing them to disk.

When wasmer module format version is published, we can consider that plus our own version number, either in a subdir or with a minor / patch format, i.e. {}/{} or {}.{} / {}-{}.

Copy link
Member Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

In 9f5716c you see an approach that hacks into the private header information. This way we ensure a module version change is noticed.

We could hash the full Wasmer module header into the file path, like format!("v3-wasmer{}", &header_hash[0..8]). Then the bumping happens automatically.

The only issue I see with this approach is that it is not obvious for admins which cache directory is used and which one is obsolete.

Copy link
Contributor

@maurolacy maurolacy Feb 8, 2022 •
edited

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

We could hash the full Wasmer module header into the file path, like format!("v3-wasmer{}", &header_hash[0..8]). Then the bumping happens automatically.

I like it. At least, until there's a public wasmer module format version number.

The only issue I see with this approach is that it is not obvious for admins which cache directory is used and which one is obsolete.

Not a big issue. The one being used will always contain more recent files, and those can be easily detected

Copy link
Member Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Good stuff. Let's continue the conversation in #1224 as Wasmer 2.2.0 will be the first version in which the relevant information is stored in the module headers.


/// Representation of a directory that contains compiled Wasm artifacts.
pub struct FileSystemCache {
Expand Down Expand Up @@ -116,22 +132,21 @@ mod tests {
const TESTING_MEMORY_LIMIT: Option<Size> = Some(Size::mebi(16));
const TESTING_GAS_LIMIT: u64 = 500_000_000;

const SOME_WAT: &str = r#"(module
(type $t0 (func (param i32) (result i32)))
(func $add_one (export "add_one") (type $t0) (param $p0 i32) (result i32)
get_local $p0
i32.const 1
i32.add))
"#;

#[test]
fn file_system_cache_run() {
let tmp_dir = TempDir::new().unwrap();
let mut cache = unsafe { FileSystemCache::new(tmp_dir.path()).unwrap() };

// Create module
let wasm = wat::parse_str(
r#"(module
(type $t0 (func (param i32) (result i32)))
(func $add_one (export "add_one") (type $t0) (param $p0 i32) (result i32)
get_local $p0
i32.const 1
i32.add))
"#,
)
.unwrap();
let wasm = wat::parse_str(SOME_WAT).unwrap();
let checksum = Checksum::generate(&wasm);

// Module does not exist
Expand Down Expand Up @@ -160,4 +175,22 @@ mod tests {
assert_eq!(result[0].unwrap_i32(), 43);
}
}

#[test]
fn file_system_cache_store_uses_expected_path() {
let tmp_dir = TempDir::new().unwrap();
let mut cache = unsafe { FileSystemCache::new(tmp_dir.path()).unwrap() };

// Create module
let wasm = wat::parse_str(SOME_WAT).unwrap();
let checksum = Checksum::generate(&wasm);

// Store module
let module = compile(&wasm, None, &[]).unwrap();
cache.store(&checksum, &module).unwrap();

let file_path = format!("{}/v2/{}", tmp_dir.path().to_string_lossy(), checksum);
let serialized_module = fs::read(file_path).unwrap();
assert_eq!(serialized_module.len(), 1040);
}
}