runtime: fix unsync_load on atomic types

tokio/src/loom/std/atomic_u16.rs

@@ -23,7 +23,7 @@ impl AtomicU16 {
     /// All mutations must have happened before the unsynchronized load.
     /// Additionally, there must be no concurrent mutations.
     pub(crate) unsafe fn unsync_load(&self) -> u16 {
-        *(*self.inner.get()).get_mut()
+        core::ptr::read(self.inner.get() as *const u16)
     }
 }
 

tokio/src/loom/std/atomic_u32.rs

@@ -23,7 +23,7 @@ impl AtomicU32 {
     /// All mutations must have happened before the unsynchronized load.
     /// Additionally, there must be no concurrent mutations.
     pub(crate) unsafe fn unsync_load(&self) -> u32 {
-        *(*self.inner.get()).get_mut()
+        core::ptr::read(self.inner.get() as *const u32)
     }
 }
 

tokio/src/loom/std/atomic_usize.rs

@@ -23,7 +23,7 @@ impl AtomicUsize {
     /// All mutations must have happened before the unsynchronized load.
     /// Additionally, there must be no concurrent mutations.
     pub(crate) unsafe fn unsync_load(&self) -> usize {
-        *(*self.inner.get()).get_mut()
+        core::ptr::read(self.inner.get() as *const usize)
     }
 
     pub(crate) fn with_mut<R>(&mut self, f: impl FnOnce(&mut usize) -> R) -> R {

tokio/src/runtime/tests/queue.rs

@@ -111,7 +111,7 @@ const fn normal_or_miri(normal: usize, miri: usize) -> usize {
 
 #[test]
 fn stress1() {
-    const NUM_ITER: usize = 1;
+    const NUM_ITER: usize = 5;
     const NUM_STEAL: usize = normal_or_miri(1_000, 10);
     const NUM_LOCAL: usize = normal_or_miri(1_000, 10);
     const NUM_PUSH: usize = normal_or_miri(500, 10);