Rollup merge of #118029 - saethlin:allocid-gc, r=RalfJung

Expand Miri's BorTag GC to a Provenance GC

As suggested in #3080 (comment)

We previously solved memory growth issues associated with the Stacked Borrows and Tree Borrows runtimes with a GC. But of course we also have state accumulation associated with whole allocations elsewhere in the interpreter, and this PR starts tackling those.

To do this, we expand the visitor for the GC so that it can visit a BorTag or an AllocId. Instead of collecting all live AllocIds into a single HashSet, we just collect from the Machine itself then go through an accessor `InterpCx::is_alloc_live` which checks a number of allocation data structures in the core interpreter. This avoids the overhead of all the inserts that collecting their keys would require.

r? ``@RalfJung``

Author: Noratrieb

.github/workflows/ci.yml

@@ -37,7 +37,7 @@ jobs:
 
       - name: Set the tag GC interval to 1 on linux
         if: runner.os == 'Linux'
-        run: echo "MIRIFLAGS=-Zmiri-tag-gc=1" >> $GITHUB_ENV
+        run: echo "MIRIFLAGS=-Zmiri-provenance-gc=1" >> $GITHUB_ENV
 
       # Cache the global cargo directory, but NOT the local `target` directory which
       # we cannot reuse anyway when the nightly changes (and it grows quite large

README.md

@@ -411,10 +411,10 @@ to Miri failing to detect cases of undefined behavior in a program.
   without an explicit value), `none` means it never recurses, `scalar` means it only recurses for
   types where we would also emit `noalias` annotations in the generated LLVM IR (types passed as
   individual scalars or pairs of scalars). Setting this to `none` is **unsound**.
-* `-Zmiri-tag-gc=<blocks>` configures how often the pointer tag garbage collector runs. The default
-  is to search for and remove unreachable tags once every `10000` basic blocks. Setting this to
-  `0` disables the garbage collector, which causes some programs to have explosive memory usage
-  and/or super-linear runtime.
+* `-Zmiri-provenance-gc=<blocks>` configures how often the pointer provenance garbage collector runs.
+  The default is to search for and remove unreachable provenance once every `10000` basic blocks. Setting
+  this to `0` disables the garbage collector, which causes some programs to have explosive memory
+  usage and/or super-linear runtime.
 * `-Zmiri-track-alloc-id=<id1>,<id2>,...` shows a backtrace when the given allocations are
   being allocated or freed.  This helps in debugging memory leaks and
   use after free bugs. Specifying this argument multiple times does not overwrite the previous

src/bin/miri.rs

@@ -531,10 +531,10 @@ fn main() {
                 Err(err) => show_error!("-Zmiri-report-progress requires a `u32`: {}", err),
             };
             miri_config.report_progress = Some(interval);
-        } else if let Some(param) = arg.strip_prefix("-Zmiri-tag-gc=") {
+        } else if let Some(param) = arg.strip_prefix("-Zmiri-provenance-gc=") {
             let interval = match param.parse::<u32>() {
                 Ok(i) => i,
-                Err(err) => show_error!("-Zmiri-tag-gc requires a `u32`: {}", err),
+                Err(err) => show_error!("-Zmiri-provenance-gc requires a `u32`: {}", err),
             };
             miri_config.gc_interval = interval;
         } else if let Some(param) = arg.strip_prefix("-Zmiri-measureme=") {

src/borrow_tracker/mod.rs

@@ -75,8 +75,8 @@ pub struct FrameState {
     protected_tags: SmallVec<[(AllocId, BorTag); 2]>,
 }
 
-impl VisitTags for FrameState {
-    fn visit_tags(&self, _visit: &mut dyn FnMut(BorTag)) {
+impl VisitProvenance for FrameState {
+    fn visit_provenance(&self, _visit: &mut VisitWith<'_>) {
         // `protected_tags` are already recorded by `GlobalStateInner`.
     }
 }
@@ -110,10 +110,10 @@ pub struct GlobalStateInner {
     unique_is_unique: bool,
 }
 
-impl VisitTags for GlobalStateInner {
-    fn visit_tags(&self, visit: &mut dyn FnMut(BorTag)) {
+impl VisitProvenance for GlobalStateInner {
+    fn visit_provenance(&self, visit: &mut VisitWith<'_>) {
         for &tag in self.protected_tags.keys() {
-            visit(tag);
+            visit(None, Some(tag));
         }
         // The only other candidate is base_ptr_tags, and that does not need visiting since we don't ever
         // GC the bottommost/root tag.
@@ -236,6 +236,10 @@ impl GlobalStateInner {
             tag
         })
     }
+
+    pub fn remove_unreachable_allocs(&mut self, allocs: &LiveAllocs<'_, '_, '_>) {
+        self.base_ptr_tags.retain(|id, _| allocs.is_live(*id));
+    }
 }
 
 /// Which borrow tracking method to use
@@ -503,11 +507,11 @@ impl AllocState {
     }
 }
 
-impl VisitTags for AllocState {
-    fn visit_tags(&self, visit: &mut dyn FnMut(BorTag)) {
+impl VisitProvenance for AllocState {
+    fn visit_provenance(&self, visit: &mut VisitWith<'_>) {
         match self {
-            AllocState::StackedBorrows(sb) => sb.visit_tags(visit),
-            AllocState::TreeBorrows(tb) => tb.visit_tags(visit),
+            AllocState::StackedBorrows(sb) => sb.visit_provenance(visit),
+            AllocState::TreeBorrows(tb) => tb.visit_provenance(visit),
         }
     }
 }

src/borrow_tracker/stacked_borrows/mod.rs

@@ -462,10 +462,10 @@ impl Stacks {
     }
 }
 
-impl VisitTags for Stacks {
-    fn visit_tags(&self, visit: &mut dyn FnMut(BorTag)) {
+impl VisitProvenance for Stacks {
+    fn visit_provenance(&self, visit: &mut VisitWith<'_>) {
         for tag in self.exposed_tags.iter().copied() {
-            visit(tag);
+            visit(None, Some(tag));
         }
     }
 }

src/borrow_tracker/tree_borrows/diagnostics.rs

@@ -200,7 +200,7 @@ impl<'tcx> Tree {
     /// Climb the tree to get the tag of a distant ancestor.
     /// Allows operations on tags that are unreachable by the program
     /// but still exist in the tree. Not guaranteed to perform consistently
-    /// if `tag-gc=1`.
+    /// if `provenance-gc=1`.
     fn nth_parent(&self, tag: BorTag, nth_parent: u8) -> Option<BorTag> {
         let mut idx = self.tag_mapping.get(&tag).unwrap();
         for _ in 0..nth_parent {

src/borrow_tracker/tree_borrows/tree.rs

@@ -742,11 +742,11 @@ impl Tree {
     }
 }
 
-impl VisitTags for Tree {
-    fn visit_tags(&self, visit: &mut dyn FnMut(BorTag)) {
+impl VisitProvenance for Tree {
+    fn visit_provenance(&self, visit: &mut VisitWith<'_>) {
         // To ensure that the root never gets removed, we visit it
         // (the `root` node of `Tree` is not an `Option<_>`)
-        visit(self.nodes.get(self.root).unwrap().tag)
+        visit(None, Some(self.nodes.get(self.root).unwrap().tag))
     }
 }
 

src/concurrency/data_race.rs

@@ -790,9 +790,9 @@ pub struct VClockAlloc {
     alloc_ranges: RefCell<RangeMap<MemoryCellClocks>>,
 }
 
-impl VisitTags for VClockAlloc {
-    fn visit_tags(&self, _visit: &mut dyn FnMut(BorTag)) {
-        // No tags here.
+impl VisitProvenance for VClockAlloc {
+    fn visit_provenance(&self, _visit: &mut VisitWith<'_>) {
+        // No tags or allocIds here.
     }
 }
 
@@ -1404,8 +1404,8 @@ pub struct GlobalState {
     pub track_outdated_loads: bool,
 }
 
-impl VisitTags for GlobalState {
-    fn visit_tags(&self, _visit: &mut dyn FnMut(BorTag)) {
+impl VisitProvenance for GlobalState {
+    fn visit_provenance(&self, _visit: &mut VisitWith<'_>) {
         // We don't have any tags.
     }
 }

src/concurrency/init_once.rs

@@ -45,10 +45,10 @@ pub(super) struct InitOnce<'mir, 'tcx> {
     data_race: VClock,
 }
 
-impl<'mir, 'tcx> VisitTags for InitOnce<'mir, 'tcx> {
-    fn visit_tags(&self, visit: &mut dyn FnMut(BorTag)) {
+impl<'mir, 'tcx> VisitProvenance for InitOnce<'mir, 'tcx> {
+    fn visit_provenance(&self, visit: &mut VisitWith<'_>) {
         for waiter in self.waiters.iter() {
-            waiter.callback.visit_tags(visit);
+            waiter.callback.visit_provenance(visit);
         }
     }
 }

src/concurrency/sync.rs

@@ -181,10 +181,10 @@ pub(crate) struct SynchronizationState<'mir, 'tcx> {
     pub(super) init_onces: IndexVec<InitOnceId, InitOnce<'mir, 'tcx>>,
 }
 
-impl<'mir, 'tcx> VisitTags for SynchronizationState<'mir, 'tcx> {
-    fn visit_tags(&self, visit: &mut dyn FnMut(BorTag)) {
+impl<'mir, 'tcx> VisitProvenance for SynchronizationState<'mir, 'tcx> {
+    fn visit_provenance(&self, visit: &mut VisitWith<'_>) {
         for init_once in self.init_onces.iter() {
-            init_once.visit_tags(visit);
+            init_once.visit_provenance(visit);
         }
     }
 }