Auto merge of #1485 - RalfJung:miri-extern-fn, r=oli-obk

 Miri: use extern fn to expose interpreter operations to program; fix leak checker on Windows

This is the Miri side of rust-lang/rust#74681.

Fixes #1302
Fixes #1318

Author: bors

README.md

@@ -233,6 +233,29 @@ different Miri binaries, and as such worth documenting:
   interpret the code but compile it like rustc would. This is useful to be sure
   that the compiled `rlib`s are compatible with Miri.
 
+## Miri `extern` functions
+
+Miri provides some `extern` functions that programs can import to access
+Miri-specific functionality:
+
+```rust
+#[cfg(miri)]
+extern "Rust" {
+    /// Miri-provided extern function to mark the block `ptr` points to as a "root"
+    /// for some static memory. This memory and everything reachable by it is not
+    /// considered leaking even if it still exists when the program terminates.
+    ///
+    /// `ptr` has to point to the beginning of an allocated block.
+    fn miri_static_root(ptr: *const u8);
+
+    /// Miri-provided extern function to begin unwinding with the given payload.
+    ///
+    /// This is internal and unstable and should not be used; we give it here
+    /// just to be complete.
+    fn miri_start_panic(payload: *mut u8) -> !;
+}
+```
+
 ## Contributing and getting help
 
 If you want to contribute to Miri, great!  Please check out our

rust-version

@@ -1 +1 @@
-4825e12fc9c79954aa0fe18f5521efa6c19c7539
+0e11fc8053d32c44e7152865852acc5c3c54efb3

src/eval.rs

@@ -5,6 +5,7 @@ use std::ffi::OsStr;
 
 use rand::rngs::StdRng;
 use rand::SeedableRng;
+use log::info;
 
 use rustc_hir::def_id::DefId;
 use rustc_middle::ty::{self, layout::LayoutCx, TyCtxt};
@@ -195,8 +196,8 @@ pub fn create_ecx<'mir, 'tcx: 'mir>(
 /// Returns `Some(return_code)` if program executed completed.
 /// Returns `None` if an evaluation error occured.
 pub fn eval_main<'tcx>(tcx: TyCtxt<'tcx>, main_id: DefId, config: MiriConfig) -> Option<i64> {
-    // FIXME: on Windows, we ignore leaks (https://github.com/rust-lang/miri/issues/1302).
-    let ignore_leaks = config.ignore_leaks || tcx.sess.target.target.target_os == "windows";
+    // Copy setting before we move `config`.
+    let ignore_leaks = config.ignore_leaks;
 
     let (mut ecx, ret_place) = match create_ecx(tcx, main_id, config) {
         Ok(v) => v,
@@ -244,7 +245,8 @@ pub fn eval_main<'tcx>(tcx: TyCtxt<'tcx>, main_id: DefId, config: MiriConfig) ->
     match res {
         Ok(return_code) => {
             if !ignore_leaks {
-                let leaks = ecx.memory.leak_report();
+                info!("Additonal static roots: {:?}", ecx.machine.static_roots);
+                let leaks = ecx.memory.leak_report(&ecx.machine.static_roots);
                 if leaks != 0 {
                     tcx.sess.err("the evaluated program leaked memory");
                     // Ignore the provided return code - let the reported error

src/machine.rs

@@ -262,6 +262,9 @@ pub struct Evaluator<'mir, 'tcx> {
 
     /// Precomputed `TyLayout`s for primitive data types that are commonly used inside Miri.
     pub(crate) layouts: PrimitiveLayouts<'tcx>,
+
+    /// Allocations that are considered roots of static memory (that may leak).
+    pub(crate) static_roots: Vec<AllocId>,
 }
 
 impl<'mir, 'tcx> Evaluator<'mir, 'tcx> {
@@ -289,6 +292,7 @@ impl<'mir, 'tcx> Evaluator<'mir, 'tcx> {
             time_anchor: Instant::now(),
             layouts,
             threads: ThreadManager::default(),
+            static_roots: Vec::new(),
         }
     }
 }

src/shims/foreign_items.rs

@@ -111,7 +111,7 @@ pub trait EvalContextExt<'mir, 'tcx: 'mir>: crate::MiriEvalContextExt<'mir, 'tcx
         def_id: DefId,
         args: &[OpTy<'tcx, Tag>],
         ret: Option<(PlaceTy<'tcx, Tag>, mir::BasicBlock)>,
-        _unwind: Option<mir::BasicBlock>,
+        unwind: Option<mir::BasicBlock>,
     ) -> InterpResult<'tcx, Option<&'mir mir::Body<'tcx>>> {
         let this = self.eval_context_mut();
         let attrs = this.tcx.get_attrs(def_id);
@@ -126,6 +126,10 @@ pub trait EvalContextExt<'mir, 'tcx: 'mir>: crate::MiriEvalContextExt<'mir, 'tcx
         // First: functions that diverge.
         let (dest, ret) = match ret {
             None => match link_name {
+                "miri_start_panic" => {
+                    this.handle_miri_start_panic(args, unwind)?;
+                    return Ok(None);
+                }
                 // This matches calls to the foreign item `panic_impl`.
                 // The implementation is provided by the function with the `#[panic_handler]` attribute.
                 "panic_impl" => {
@@ -193,6 +197,17 @@ pub trait EvalContextExt<'mir, 'tcx: 'mir>: crate::MiriEvalContextExt<'mir, 'tcx
         // Here we dispatch all the shims for foreign functions. If you have a platform specific
         // shim, add it to the corresponding submodule.
         match link_name {
+            // Miri-specific extern functions
+            "miri_static_root" => {
+                let &[ptr] = check_arg_count(args)?;
+                let ptr = this.read_scalar(ptr)?.not_undef()?;
+                let ptr = this.force_ptr(ptr)?;
+                if ptr.offset != Size::ZERO {
+                    throw_unsup_format!("pointer passed to miri_static_root must point to beginning of an allocated block");
+                }
+                this.machine.static_roots.push(ptr.alloc_id);
+            }
+
             // Standard C allocation
             "malloc" => {
                 let &[size] = check_arg_count(args)?;

src/shims/intrinsics.rs

@@ -18,7 +18,7 @@ pub trait EvalContextExt<'mir, 'tcx: 'mir>: crate::MiriEvalContextExt<'mir, 'tcx
         instance: ty::Instance<'tcx>,
         args: &[OpTy<'tcx, Tag>],
         ret: Option<(PlaceTy<'tcx, Tag>, mir::BasicBlock)>,
-        unwind: Option<mir::BasicBlock>,
+        _unwind: Option<mir::BasicBlock>,
     ) -> InterpResult<'tcx> {
         let this = self.eval_context_mut();
         let intrinsic_name = this.tcx.item_name(instance.def_id());
@@ -32,13 +32,10 @@ pub trait EvalContextExt<'mir, 'tcx: 'mir>: crate::MiriEvalContextExt<'mir, 'tcx
             return Ok(());
         }
 
-        // First handle intrinsics without return place.
+        // All supported intrinsics have a return place.
         let intrinsic_name = &*intrinsic_name.as_str();
         let (dest, ret) = match ret {
-            None => match intrinsic_name {
-                "miri_start_panic" => return this.handle_miri_start_panic(args, unwind),
-                _ => throw_unsup_format!("unimplemented (diverging) intrinsic: {}", intrinsic_name),
-            },
+            None => throw_unsup_format!("unimplemented (diverging) intrinsic: {}", intrinsic_name),
             Some(p) => p,
         };
 

tests/compile-fail/memleak.rs

@@ -1,5 +1,3 @@
-// ignore-windows: We do not check leaks on Windows
-
 //error-pattern: the evaluated program leaked memory
 
 fn main() {

tests/compile-fail/memleak_rc.rs

@@ -1,5 +1,3 @@
-// ignore-windows: We do not check leaks on Windows
-
 //error-pattern: the evaluated program leaked memory
 
 use std::rc::Rc;