Impact
As a Rust library the wasmtime crate clearly marks which functions are safe and which are unsafe, guaranteeing that if consumers never use unsafe then it should not be possible to have memory unsafety issues in their embeddings of Wasmtime. An issue was discovered in the safe API of Linker::func APIs. These APIs were previously not sound when one Engine was used to create the Linker and then a different Engine was used to create a Store and then the Linker was used to instantiate a module into that Store. Cross-Engine usage of functions is not supported in Wasmtime and this can result in type confusion of function pointers, resulting in being able to safely call a function with the wrong type.
Triggering this bug requires using at least two Engine values in an embedding and then additionally using two different values with a Linker (one at the creation time of the Linker and another when instantiating a module with the Linker).
It's expected that usage of more-than-one Engine in an embedding is relatively rare since an Engine is intended to be a globally shared resource, so the expectation is that the impact of this issue is relatively small.
The fix implemented is to change this behavior to panic!() in Rust instead of silently allowing it. Using different Engine instances with a Linker is a programmer bug that wasmtime catches at runtime.
Patches
This bug has been patched and users should upgrade to Wasmtime version 0.30.0.
Workarounds
If you cannot upgrade Wasmtime and are using more than one Engine in your embedding it's recommended to instead use only one Engine for the entire program if possible. An Engine is designed to be a globally shared resource that is suitable to have only one for the lifetime of an entire process. If using multiple Engines is required then code should be audited to ensure that Linker is only used with one Engine.
For more information
If you have any questions or comments about this advisory:
Reach out to us on the Bytecode Alliance Zulip chat Open an issue in the bytecodealliance/wasmtime repository
Impact
There was a use-after-free bug when passing externrefs from the host to guest Wasm content.
To trigger the bug, you have to explicitly pass multiple externrefs from the host to a Wasm instance at the same time, either by
passing multiple externrefs as arguments from host code to a Wasm function, or returning multiple externrefs to Wasm from a multi-value return function defined in the host. If you do not have host code that matches one of these shapes, then you are not impacted.
If Wasmtime's VMExternRefActivationsTable became filled to capacity after passing the first externref in, then passing in the second externref could trigger a garbage collection. However the first externref is not rooted until we pass control to Wasm, and therefore could be reclaimed by the collector if nothing else was holding a reference to it or otherwise keeping it alive. Then, when control was passed to Wasm after the garbage collection, Wasm could use the first externref, which at this point has already been freed.
We have reason to believe that the effective impact of this bug is relatively small because usage of externref is currently quite rare.
Patches
The bug has been fixed, and users should upgrade to Wasmtime 0.30.0.
Additionally, we have updated our primary externref fuzz target such that it better exercises these code paths and we can have greater confidence in their correctness going forward.
Workarounds
If you cannot upgrade Wasmtime yet, you can avoid the bug by disabling reference types support in Wasmtime by passing false to wasmtime::Config::wasmreferencetypes.
References
The reference types Wasm proposal, which introduces externref
For more information
If you have any questions or comments about this advisory:
Reach out to us on the Bytecode Alliance Zulip chat Open an issue in the bytecodealliance/wasmtime repository