Merge pull request #162 from influxdata/crepererum/evil-payload

feat: infrastructure for "evil" test payloads

Author: crepererum

Cargo.lock

@@ -3,6 +3,7 @@ resolver = "3"
 members = [
   "arrow2bytes",
   "guests/bundle",
+  "guests/evil",
   "guests/python",
   "guests/rust",
   "host",
@@ -24,6 +25,7 @@ datafusion-expr = { version = "49.0.1", default-features = false }
 datafusion-sql = { version = "49.0.1", default-features = false }
 datafusion-udf-wasm-arrow2bytes = { path = "arrow2bytes", version = "0.1.0" }
 datafusion-udf-wasm-bundle = { path = "guests/bundle", version = "0.1.0" }
+datafusion-udf-wasm-evil = { path = "guests/evil", version = "0.1.0" }
 datafusion-udf-wasm-guest = { path = "guests/rust", version = "0.1.0" }
 datafusion-udf-wasm-host = { path = "host", version = "0.1.0" }
 datafusion-udf-wasm-python = { path = "guests/python", version = "0.1.0" }
@@ -33,6 +35,7 @@ hyper = { version = "1.8", default-features = false }
 insta = { version = "1.43.2", "default-features" = false }
 log = { version = "0.4.28", default-features = false }
 pyo3 = { version = "0.27.1", default-features = false, features = ["macros"] }
+regex = { version = "1", default-features = false }
 sqlparser = { version = "0.55.0", default-features = false, features = [
   "std",
   "visitor"

Cargo.toml

@@ -1,2 +1,3 @@
+mod evil
 mod python
 mod rust

guests/Justfile

@@ -6,12 +6,14 @@ license.workspace = true
 
 [build-dependencies]
 # these need to be marked as build dependencies so the build script reruns whenever they change
+datafusion-udf-wasm-evil = { workspace = true, optional = true }
 datafusion-udf-wasm-guest = { workspace = true, optional = true }
 datafusion-udf-wasm-python = { workspace = true, optional = true }
 # the actual build-time dependencies
 serde_json = "1.0.145"
 
 [features]
+evil = ["dep:datafusion-udf-wasm-evil"]
 example = ["dep:datafusion-udf-wasm-guest"]
 python = ["dep:datafusion-udf-wasm-python"]
 

guests/bundle/Cargo.toml

@@ -287,6 +287,15 @@ fn just_build(cwd: &Path, just_cmd: &str, cargo_target_dir: &Path) {
 ///
 /// This must be in-sync with the feature list in `Cargo.toml`.
 const FEATURES: &[Feature] = &[
+    Feature {
+        name: "evil",
+        package: "datafusion-udf-wasm-evil",
+        just_cmds: &[JustCmd {
+            artifact_type: ArtifactType::Lib,
+            const_name: "EVIL",
+            doc: "Evil payloads.",
+        }],
+    },
     Feature {
         name: "example",
         package: "datafusion-udf-wasm-guest",

guests/bundle/build.rs

@@ -0,0 +1,18 @@
+[package]
+name = "datafusion-udf-wasm-evil"
+version.workspace = true
+edition.workspace = true
+license.workspace = true
+
+[lib]
+crate-type = ["cdylib"]
+
+[dependencies]
+arrow.workspace = true
+datafusion-common.workspace = true
+datafusion-expr.workspace = true
+datafusion-udf-wasm-guest.workspace = true
+tar.workspace = true
+
+[lints]
+workspace = true

guests/evil/Cargo.toml

@@ -0,0 +1,14 @@
+[private]
+build profile:
+    @echo ::group::guests::evil::build-{{profile}}
+    cargo build --target=wasm32-wasip2 --profile={{replace(profile, "debug", "dev")}}
+    @echo ::endgroup::
+
+# build library in debug mode
+build-debug: (build "debug")
+
+# build library in release mode
+build-release: (build "release")
+
+# checks build
+check-build: build-debug

guests/evil/Justfile

@@ -0,0 +1,5 @@
+# Evil Test Payloads
+
+These are payloads that try to stress the WASM sandbox. They are NOT meant to be used in production.
+
+For efficiency we generate a single WASM binary that is multiplexed using the `EVIL` environment variable.

guests/evil/README.md

@@ -0,0 +1,58 @@
+//! Library that multiplexes different evil payloads.
+//!
+//! We need that because evil payloads may act long before the actual UDFs are available.
+use std::sync::Arc;
+
+use datafusion_common::Result as DataFusionResult;
+use datafusion_expr::ScalarUDFImpl;
+use datafusion_udf_wasm_guest::export;
+
+mod root;
+mod runtime;
+
+/// Method that returns the root filesystem.
+type RootFn = Box<dyn Fn() -> Option<Vec<u8>>>;
+
+/// Method that enumerates UDFs.
+type UdfsFn = Box<dyn Fn(String) -> DataFusionResult<Vec<Arc<dyn ScalarUDFImpl>>>>;
+
+/// An evil.
+struct Evil {
+    /// Root file system.
+    root: RootFn,
+
+    /// Returns  UDFs.
+    udfs: UdfsFn,
+}
+
+impl Evil {
+    /// Get evil, multiplexed by env.
+    fn get() -> Self {
+        match std::env::var("EVIL").expect("evil specified").as_str() {
+            "root::many_files" => Self {
+                root: Box::new(root::many_files::root),
+                udfs: Box::new(root::many_files::udfs),
+            },
+            "runtime" => Self {
+                root: Box::new(runtime::root),
+                udfs: Box::new(runtime::udfs),
+            },
+            other => panic!("unknown evil: {other}"),
+        }
+    }
+}
+
+/// Return root file system.
+fn root() -> Option<Vec<u8>> {
+    (Evil::get().root)()
+}
+
+/// Returns our evil UDFs.
+fn udfs(source: String) -> DataFusionResult<Vec<Arc<dyn ScalarUDFImpl>>> {
+    (Evil::get().udfs)(source)
+}
+
+export! {
+    root_fs_tar: root,
+    scalar_udfs: udfs,
+}

guests/evil/src/lib.rs

@@ -0,0 +1,31 @@
+//! Evil payloads that creates A LOT of files.
+use std::sync::Arc;
+
+use datafusion_common::Result as DataFusionResult;
+use datafusion_expr::ScalarUDFImpl;
+
+/// Return root file system.
+#[expect(clippy::unnecessary_wraps, reason = "public API through export! macro")]
+pub(crate) fn root() -> Option<Vec<u8>> {
+    let mut ar = tar::Builder::new(Vec::new());
+
+    const LIMIT: u64 = 10_000;
+    for i in 0..=LIMIT {
+        let mut header = tar::Header::new_gnu();
+        header.set_path(i.to_string()).unwrap();
+        header.set_size(0);
+        header.set_cksum();
+
+        ar.append(&header, b"".as_slice()).unwrap();
+    }
+
+    Some(ar.into_inner().unwrap())
+}
+
+/// Returns UDFs.
+///
+/// The passed `source` is ignored.
+#[expect(clippy::unnecessary_wraps, reason = "public API through export! macro")]
+pub(crate) fn udfs(_source: String) -> DataFusionResult<Vec<Arc<dyn ScalarUDFImpl>>> {
+    Ok(vec![])
+}