Validate Module Artifacts (CLI + WASIX ModuleCache)

lib/api/src/js/module.rs

@@ -238,6 +238,18 @@ impl Module {
         return Err(DeserializeError::Generic("You need to enable the `js-serializable-module` feature flag to deserialize a `Module`".to_string()));
     }
 
+    pub fn deserialize_checked(
+        _engine: &impl AsEngineRef,
+        _bytes: impl IntoBytes,
+    ) -> Result<Self, DeserializeError> {
+        #[cfg(feature = "js-serializable-module")]
+        return Self::from_binary(_engine, &_bytes.into_bytes())
+            .map_err(|e| DeserializeError::Compiler(e));
+
+        #[cfg(not(feature = "js-serializable-module"))]
+        return Err(DeserializeError::Generic("You need to enable the `js-serializable-module` feature flag to deserialize a `Module`".to_string()));
+    }
+
     pub unsafe fn deserialize_from_file(
         engine: &impl AsEngineRef,
         path: impl AsRef<Path>,
@@ -246,6 +258,14 @@ impl Module {
         Self::deserialize(engine, bytes)
     }
 
+    pub fn deserialize_from_file_checked(
+        engine: &impl AsEngineRef,
+        path: impl AsRef<Path>,
+    ) -> Result<Self, DeserializeError> {
+        let bytes = std::fs::read(path.as_ref())?;
+        Self::deserialize_checked(engine, bytes)
+    }
+
     pub fn set_name(&mut self, name: &str) -> bool {
         self.name = Some(name.to_string());
         true

lib/api/src/module.rs

@@ -219,6 +219,8 @@ impl Module {
 
     /// Deserializes a serialized Module binary into a `Module`.
     ///
+    /// Note: You should usually prefer the safe [`Module::deserialize_checked`].
+    ///
     /// # Important
     ///
     /// This function only accepts a custom binary format, which will be different
@@ -253,6 +255,56 @@ impl Module {
         Ok(Self(module_imp::Module::deserialize(engine, bytes)?))
     }
 
+    /// Deserializes a serialized Module binary into a `Module`.
+    ///
+    /// # Important
+    ///
+    /// This function only accepts a custom binary format, which will be different
+    /// than the `wasm` binary format and may change among Wasmer versions.
+    /// (it should be the result of the serialization of a Module via the
+    /// `Module::serialize` method.).
+    ///
+    /// # Usage
+    ///
+    /// ```ignore
+    /// # use wasmer::*;
+    /// # fn main() -> anyhow::Result<()> {
+    /// # let mut store = Store::default();
+    /// let module = Module::deserialize_checked(&store, serialized_data)?;
+    /// # Ok(())
+    /// # }
+    /// ```
+    pub fn deserialize_checked(
+        engine: &impl AsEngineRef,
+        bytes: impl IntoBytes,
+    ) -> Result<Self, DeserializeError> {
+        Ok(Self(module_imp::Module::deserialize_checked(
+            engine, bytes,
+        )?))
+    }
+
+    /// Deserializes a a serialized Module located in a `Path` into a `Module`.
+    /// > Note: the module has to be serialized before with the `serialize` method.
+    ///
+    /// # Usage
+    ///
+    /// ```ignore
+    /// # use wasmer::*;
+    /// # let mut store = Store::default();
+    /// # fn main() -> anyhow::Result<()> {
+    /// let module = Module::deserialize_from_file(&store, path)?;
+    /// # Ok(())
+    /// # }
+    /// ```
+    pub fn deserialize_from_file_checked(
+        engine: &impl AsEngineRef,
+        path: impl AsRef<Path>,
+    ) -> Result<Self, DeserializeError> {
+        Ok(Self(module_imp::Module::deserialize_from_file_checked(
+            engine, path,
+        )?))
+    }
+
     /// Deserializes a a serialized Module located in a `Path` into a `Module`.
     /// > Note: the module has to be serialized before with the `serialize` method.
     ///

lib/api/src/sys/module.rs

@@ -78,6 +78,19 @@ impl Module {
         Ok(Self::from_artifact(artifact))
     }
 
+    pub fn deserialize_checked(
+        engine: &impl AsEngineRef,
+        bytes: impl IntoBytes,
+    ) -> Result<Self, DeserializeError> {
+        let bytes = bytes.into_bytes();
+        let artifact = engine
+            .as_engine_ref()
+            .engine()
+            .0
+            .deserialize_checked(&bytes)?;
+        Ok(Self::from_artifact(artifact))
+    }
+
     pub unsafe fn deserialize_from_file(
         engine: &impl AsEngineRef,
         path: impl AsRef<Path>,
@@ -90,6 +103,18 @@ impl Module {
         Ok(Self::from_artifact(artifact))
     }
 
+    pub fn deserialize_from_file_checked(
+        engine: &impl AsEngineRef,
+        path: impl AsRef<Path>,
+    ) -> Result<Self, DeserializeError> {
+        let artifact = engine
+            .as_engine_ref()
+            .engine()
+            .0
+            .deserialize_from_file_checked(path.as_ref())?;
+        Ok(Self::from_artifact(artifact))
+    }
+
     fn from_artifact(artifact: Arc<Artifact>) -> Self {
         Self { artifact }
     }

lib/cli/src/commands/run.rs

@@ -427,7 +427,7 @@ impl RunWithPathBuf {
         if wasmer_compiler::Artifact::is_deserializable(&contents) {
             let engine = wasmer_compiler::EngineBuilder::headless();
             let store = Store::new(engine);
-            let module = unsafe { Module::deserialize_from_file(&store, &self.path)? };
+            let module = Module::deserialize_from_file_checked(&store, &self.path)?;
             return Ok((store, module));
         }
         let (store, compiler_type) = self.store.get_store()?;

lib/compiler/src/engine/artifact.rs

@@ -148,6 +148,30 @@ impl Artifact {
     /// # Safety
     /// This function is unsafe because rkyv reads directly without validating
     /// the data.
+    pub fn deserialize_checked(engine: &Engine, bytes: &[u8]) -> Result<Self, DeserializeError> {
+        if !ArtifactBuild::is_deserializable(bytes) {
+            return Err(DeserializeError::Incompatible(
+                "Magic header not found".to_string(),
+            ));
+        }
+
+        let bytes = Self::get_byte_slice(bytes, ArtifactBuild::MAGIC_HEADER.len(), bytes.len())?;
+
+        let metadata_len = MetadataHeader::parse(bytes)?;
+        let metadata_slice = Self::get_byte_slice(bytes, MetadataHeader::LEN, bytes.len())?;
+        let metadata_slice = Self::get_byte_slice(metadata_slice, 0, metadata_len)?;
+
+        let serializable = SerializableModule::deserialize_checked(metadata_slice)?;
+        let artifact = ArtifactBuild::from_serializable(serializable);
+        let mut inner_engine = engine.inner_mut();
+        Self::from_parts(&mut inner_engine, artifact, engine.target())
+            .map_err(DeserializeError::Compiler)
+    }
+
+    /// Deserialize a ArtifactBuild
+    ///
+    /// # Safety
+    /// This function is unsafe because rkyv reads directly without validating the data.
     pub unsafe fn deserialize(engine: &Engine, bytes: &[u8]) -> Result<Self, DeserializeError> {
         if !ArtifactBuild::is_deserializable(bytes) {
             let static_artifact = Self::deserialize_object(engine, bytes);

lib/compiler/src/engine/inner.rs

@@ -200,12 +200,28 @@ impl Engine {
         Ok(Arc::new(Artifact::deserialize(self, bytes)?))
     }
 
+    /// Deserializes a WebAssembly module
     #[cfg(not(target_arch = "wasm32"))]
+    pub fn deserialize_checked(&self, bytes: &[u8]) -> Result<Arc<Artifact>, DeserializeError> {
+        Ok(Arc::new(Artifact::deserialize_checked(self, bytes)?))
+    }
+
     /// Deserializes a WebAssembly module from a path
+    #[cfg(not(target_arch = "wasm32"))]
+    pub fn deserialize_from_file_checked(
+        &self,
+        file_ref: &Path,
+    ) -> Result<Arc<Artifact>, DeserializeError> {
+        let contents = std::fs::read(file_ref)?;
+        self.deserialize_checked(&contents)
+    }
+
+    /// Deserialize from a file path.
     ///
     /// # Safety
     ///
-    /// The file's content must represent a serialized WebAssembly module.
+    /// See [`crate::Module::deserialize_from_file`].
+    #[cfg(not(target_arch = "wasm32"))]
     pub unsafe fn deserialize_from_file(
         &self,
         file_ref: &Path,

lib/types/Cargo.toml

@@ -16,10 +16,11 @@ serde_bytes = { version = "0.11", optional = true }
 thiserror = "1.0"
 more-asserts = "0.2"
 indexmap = { version = "1.6" }
-rkyv = { version = "0.7.40", features = ["indexmap"] }
+rkyv = { version = "0.7.40", features = ["indexmap", "validation", "strict"] }
 enum-iterator = "0.7.0"
 target-lexicon = { version = "0.12.2", default-features = false }
 enumset = "1.0"
+bytecheck = "0.6.8"
 
 [dev-dependencies]
 memoffset = "0.6"

lib/types/src/compilation/address_map.rs

@@ -10,6 +10,7 @@ use serde::{Deserialize, Serialize};
 /// Single source location to generated address mapping.
 #[cfg_attr(feature = "enable-serde", derive(Serialize, Deserialize))]
 #[derive(RkyvSerialize, RkyvDeserialize, Archive, Debug, Clone, PartialEq, Eq)]
+#[archive_attr(derive(rkyv::CheckBytes))]
 pub struct InstructionAddressMap {
     /// Original source location.
     pub srcloc: SourceLoc,
@@ -24,6 +25,7 @@ pub struct InstructionAddressMap {
 /// Function and its instructions addresses mappings.
 #[cfg_attr(feature = "enable-serde", derive(Serialize, Deserialize))]
 #[derive(RkyvSerialize, RkyvDeserialize, Archive, Debug, Clone, PartialEq, Eq, Default)]
+#[archive_attr(derive(rkyv::CheckBytes))]
 pub struct FunctionAddressMap {
     /// Instructions maps.
     /// The array is sorted by the InstructionAddressMap::code_offset field.

lib/types/src/compilation/function.rs

@@ -21,6 +21,7 @@ use serde::{Deserialize, Serialize};
 /// the frame information after a `Trap`.
 #[cfg_attr(feature = "enable-serde", derive(Deserialize, Serialize))]
 #[derive(RkyvSerialize, RkyvDeserialize, Archive, Debug, Clone, PartialEq, Eq, Default)]
+#[archive_attr(derive(rkyv::CheckBytes))]
 pub struct CompiledFunctionFrameInfo {
     /// The traps (in the function body).
     ///
@@ -34,6 +35,7 @@ pub struct CompiledFunctionFrameInfo {
 /// The function body.
 #[cfg_attr(feature = "enable-serde", derive(Deserialize, Serialize))]
 #[derive(RkyvSerialize, RkyvDeserialize, Archive, Debug, Clone, PartialEq, Eq)]
+#[archive_attr(derive(rkyv::CheckBytes))]
 pub struct FunctionBody {
     /// The function body bytes.
     #[cfg_attr(feature = "enable-serde", serde(with = "serde_bytes"))]
@@ -50,6 +52,7 @@ pub struct FunctionBody {
 /// and unwind information).
 #[cfg_attr(feature = "enable-serde", derive(Deserialize, Serialize))]
 #[derive(RkyvSerialize, RkyvDeserialize, Archive, Debug, Clone, PartialEq, Eq)]
+#[archive_attr(derive(rkyv::CheckBytes))]
 pub struct CompiledFunction {
     /// The function body.
     pub body: FunctionBody,
@@ -74,7 +77,9 @@ pub type CustomSections = PrimaryMap<SectionIndex, CustomSection>;
 /// In the future this structure may also hold other information useful
 /// for debugging.
 #[cfg_attr(feature = "enable-serde", derive(Deserialize, Serialize))]
-#[derive(RkyvSerialize, RkyvDeserialize, Archive, Debug, PartialEq, Eq, Clone)]
+#[derive(
+    RkyvSerialize, RkyvDeserialize, Archive, rkyv::CheckBytes, Debug, PartialEq, Eq, Clone,
+)]
 #[archive(as = "Self")]
 pub struct Dwarf {
     /// The section index in the [`Compilation`] that corresponds to the exception frames.

lib/types/src/compilation/module.rs

@@ -13,6 +13,7 @@ use std::sync::Arc;
 /// or the `MemoryStyle` and `TableStyle`).
 #[cfg_attr(feature = "enable-serde", derive(Deserialize, Serialize))]
 #[derive(Debug, Clone, PartialEq, Eq, RkyvSerialize, RkyvDeserialize, Archive)]
+#[archive_attr(derive(rkyv::CheckBytes))]
 pub struct CompileModuleInfo {
     /// The features used for compiling the module
     pub features: Features,

lib/types/src/compilation/relocation.rs

@@ -21,8 +21,11 @@ use serde::{Deserialize, Serialize};
 
 /// Relocation kinds for every ISA.
 #[cfg_attr(feature = "enable-serde", derive(Serialize, Deserialize))]
-#[derive(RkyvSerialize, RkyvDeserialize, Archive, Copy, Clone, Debug, PartialEq, Eq)]
+#[derive(
+    RkyvSerialize, RkyvDeserialize, Archive, rkyv::CheckBytes, Copy, Clone, Debug, PartialEq, Eq,
+)]
 #[archive(as = "Self")]
+#[repr(u8)]
 pub enum RelocationKind {
     /// absolute 4-byte
     Abs4,
@@ -90,6 +93,7 @@ impl fmt::Display for RelocationKind {
 /// A record of a relocation to perform.
 #[cfg_attr(feature = "enable-serde", derive(Serialize, Deserialize))]
 #[derive(RkyvSerialize, RkyvDeserialize, Archive, Debug, Clone, PartialEq, Eq)]
+#[archive_attr(derive(rkyv::CheckBytes))]
 pub struct Relocation {
     /// The relocation kind.
     pub kind: RelocationKind,
@@ -103,8 +107,11 @@ pub struct Relocation {
 
 /// Destination function. Can be either user function or some special one, like `memory.grow`.
 #[cfg_attr(feature = "enable-serde", derive(Serialize, Deserialize))]
-#[derive(RkyvSerialize, RkyvDeserialize, Archive, Debug, Copy, Clone, PartialEq, Eq)]
+#[derive(
+    RkyvSerialize, RkyvDeserialize, Archive, rkyv::CheckBytes, Debug, Copy, Clone, PartialEq, Eq,
+)]
 #[archive(as = "Self")]
+#[repr(u8)]
 pub enum RelocationTarget {
     /// A relocation to a function defined locally in the wasm (not an imported one).
     LocalFunc(LocalFunctionIndex),

lib/types/src/compilation/section.rs

@@ -17,6 +17,7 @@ use serde::{Deserialize, Serialize};
     RkyvSerialize,
     RkyvDeserialize,
     Archive,
+    rkyv::CheckBytes,
     Copy,
     Clone,
     PartialEq,
@@ -37,8 +38,11 @@ entity_impl!(SectionIndex);
 ///
 /// Determines how a custom section may be used.
 #[cfg_attr(feature = "enable-serde", derive(Serialize, Deserialize))]
-#[derive(RkyvSerialize, RkyvDeserialize, Archive, Debug, Clone, PartialEq, Eq)]
+#[derive(
+    RkyvSerialize, RkyvDeserialize, Archive, rkyv::CheckBytes, Debug, Clone, PartialEq, Eq,
+)]
 #[archive(as = "Self")]
+#[repr(u8)]
 pub enum CustomSectionProtection {
     /// A custom section with read permission.
     Read,
@@ -53,6 +57,7 @@ pub enum CustomSectionProtection {
 /// in the emitted module.
 #[cfg_attr(feature = "enable-serde", derive(Serialize, Deserialize))]
 #[derive(RkyvSerialize, RkyvDeserialize, Archive, Debug, Clone, PartialEq, Eq)]
+#[archive_attr(derive(rkyv::CheckBytes))]
 pub struct CustomSection {
     /// Memory protection that applies to this section.
     pub protection: CustomSectionProtection,
@@ -72,6 +77,7 @@ pub struct CustomSection {
 /// The bytes in the section.
 #[cfg_attr(feature = "enable-serde", derive(Serialize, Deserialize))]
 #[derive(RkyvSerialize, RkyvDeserialize, Archive, Debug, Clone, PartialEq, Eq, Default)]
+#[archive_attr(derive(rkyv::CheckBytes))]
 pub struct SectionBody(#[cfg_attr(feature = "enable-serde", serde(with = "serde_bytes"))] Vec<u8>);
 
 impl SectionBody {

lib/types/src/compilation/sourceloc.rs

@@ -23,7 +23,7 @@ use serde::{Deserialize, Serialize};
 )]
 #[derive(RkyvSerialize, RkyvDeserialize, Archive)]
 #[repr(transparent)]
-#[derive(Debug, Clone, Copy, PartialEq, Eq)]
+#[derive(Debug, Clone, Copy, PartialEq, Eq, rkyv::CheckBytes)]
 #[archive(as = "Self")]
 pub struct SourceLoc(u32);