Auto merge of #106790 - the8472:rawvec-niche, r=scottmcm

add more niches to rawvec

Previously RawVec only had a single niche in its `NonNull` pointer. With this change it now has `isize::MAX` niches since half the value-space of the capacity field is never needed, we can't have a capacity larger than isize::MAX.

Author: bors

library/alloc/src/raw_vec.rs

@@ -25,6 +25,16 @@ enum AllocInit {
     Zeroed,
 }
 
+#[repr(transparent)]
+#[cfg_attr(target_pointer_width = "16", rustc_layout_scalar_valid_range_end(0x7fff))]
+#[cfg_attr(target_pointer_width = "32", rustc_layout_scalar_valid_range_end(0x7fff_ffff))]
+#[cfg_attr(target_pointer_width = "64", rustc_layout_scalar_valid_range_end(0x7fff_ffff_ffff_ffff))]
+struct Cap(usize);
+
+impl Cap {
+    const ZERO: Cap = unsafe { Cap(0) };
+}
+
 /// A low-level utility for more ergonomically allocating, reallocating, and deallocating
 /// a buffer of memory on the heap without having to worry about all the corner cases
 /// involved. This type is excellent for building your own data structures like Vec and VecDeque.
@@ -50,7 +60,12 @@ enum AllocInit {
 #[allow(missing_debug_implementations)]
 pub(crate) struct RawVec<T, A: Allocator = Global> {
     ptr: Unique<T>,
-    cap: usize,
+    /// Never used for ZSTs; it's `capacity()`'s responsibility to return usize::MAX in that case.
+    ///
+    /// # Safety
+    ///
+    /// `cap` must be in the `0..=isize::MAX` range.
+    cap: Cap,
     alloc: A,
 }
 
@@ -119,7 +134,7 @@ impl<T, A: Allocator> RawVec<T, A> {
     /// the returned `RawVec`.
     pub const fn new_in(alloc: A) -> Self {
         // `cap: 0` means "unallocated". zero-sized types are ignored.
-        Self { ptr: Unique::dangling(), cap: 0, alloc }
+        Self { ptr: Unique::dangling(), cap: Cap::ZERO, alloc }
     }
 
     /// Like `with_capacity`, but parameterized over the choice of
@@ -194,7 +209,7 @@ impl<T, A: Allocator> RawVec<T, A> {
             // here should change to `ptr.len() / mem::size_of::<T>()`.
             Self {
                 ptr: unsafe { Unique::new_unchecked(ptr.cast().as_ptr()) },
-                cap: capacity,
+                cap: unsafe { Cap(capacity) },
                 alloc,
             }
         }
@@ -207,12 +222,13 @@ impl<T, A: Allocator> RawVec<T, A> {
     /// The `ptr` must be allocated (via the given allocator `alloc`), and with the given
     /// `capacity`.
     /// The `capacity` cannot exceed `isize::MAX` for sized types. (only a concern on 32-bit
-    /// systems). ZST vectors may have a capacity up to `usize::MAX`.
+    /// systems). For ZSTs capacity is ignored.
     /// If the `ptr` and `capacity` come from a `RawVec` created via `alloc`, then this is
     /// guaranteed.
     #[inline]
     pub unsafe fn from_raw_parts_in(ptr: *mut T, capacity: usize, alloc: A) -> Self {
-        Self { ptr: unsafe { Unique::new_unchecked(ptr) }, cap: capacity, alloc }
+        let cap = if T::IS_ZST { Cap::ZERO } else { unsafe { Cap(capacity) } };
+        Self { ptr: unsafe { Unique::new_unchecked(ptr) }, cap, alloc }
     }
 
     /// Gets a raw pointer to the start of the allocation. Note that this is
@@ -228,7 +244,7 @@ impl<T, A: Allocator> RawVec<T, A> {
     /// This will always be `usize::MAX` if `T` is zero-sized.
     #[inline(always)]
     pub fn capacity(&self) -> usize {
-        if T::IS_ZST { usize::MAX } else { self.cap }
+        if T::IS_ZST { usize::MAX } else { self.cap.0 }
     }
 
     /// Returns a shared reference to the allocator backing this `RawVec`.
@@ -237,7 +253,7 @@ impl<T, A: Allocator> RawVec<T, A> {
     }
 
     fn current_memory(&self) -> Option<(NonNull<u8>, Layout)> {
-        if T::IS_ZST || self.cap == 0 {
+        if T::IS_ZST || self.cap.0 == 0 {
             None
         } else {
             // We could use Layout::array here which ensures the absence of isize and usize overflows
@@ -247,7 +263,7 @@ impl<T, A: Allocator> RawVec<T, A> {
             let _: () = const { assert!(mem::size_of::<T>() % mem::align_of::<T>() == 0) };
             unsafe {
                 let align = mem::align_of::<T>();
-                let size = mem::size_of::<T>().unchecked_mul(self.cap);
+                let size = mem::size_of::<T>().unchecked_mul(self.cap.0);
                 let layout = Layout::from_size_align_unchecked(size, align);
                 Some((self.ptr.cast().into(), layout))
             }
@@ -375,12 +391,15 @@ impl<T, A: Allocator> RawVec<T, A> {
         additional > self.capacity().wrapping_sub(len)
     }
 
-    fn set_ptr_and_cap(&mut self, ptr: NonNull<[u8]>, cap: usize) {
+    /// # Safety:
+    ///
+    /// `cap` must not exceed `isize::MAX`.
+    unsafe fn set_ptr_and_cap(&mut self, ptr: NonNull<[u8]>, cap: usize) {
         // Allocators currently return a `NonNull<[u8]>` whose length matches
         // the size requested. If that ever changes, the capacity here should
         // change to `ptr.len() / mem::size_of::<T>()`.
         self.ptr = unsafe { Unique::new_unchecked(ptr.cast().as_ptr()) };
-        self.cap = cap;
+        self.cap = unsafe { Cap(cap) };
     }
 
     // This method is usually instantiated many times. So we want it to be as
@@ -405,14 +424,15 @@ impl<T, A: Allocator> RawVec<T, A> {
 
         // This guarantees exponential growth. The doubling cannot overflow
         // because `cap <= isize::MAX` and the type of `cap` is `usize`.
-        let cap = cmp::max(self.cap * 2, required_cap);
+        let cap = cmp::max(self.cap.0 * 2, required_cap);
         let cap = cmp::max(Self::MIN_NON_ZERO_CAP, cap);
 
         let new_layout = Layout::array::<T>(cap);
 
         // `finish_grow` is non-generic over `T`.
         let ptr = finish_grow(new_layout, self.current_memory(), &mut self.alloc)?;
-        self.set_ptr_and_cap(ptr, cap);
+        // SAFETY: finish_grow would have resulted in a capacity overflow if we tried to allocate more than isize::MAX items
+        unsafe { self.set_ptr_and_cap(ptr, cap) };
         Ok(())
     }
 
@@ -431,7 +451,10 @@ impl<T, A: Allocator> RawVec<T, A> {
 
         // `finish_grow` is non-generic over `T`.
         let ptr = finish_grow(new_layout, self.current_memory(), &mut self.alloc)?;
-        self.set_ptr_and_cap(ptr, cap);
+        // SAFETY: finish_grow would have resulted in a capacity overflow if we tried to allocate more than isize::MAX items
+        unsafe {
+            self.set_ptr_and_cap(ptr, cap);
+        }
         Ok(())
     }
 
@@ -449,7 +472,7 @@ impl<T, A: Allocator> RawVec<T, A> {
         if cap == 0 {
             unsafe { self.alloc.deallocate(ptr, layout) };
             self.ptr = Unique::dangling();
-            self.cap = 0;
+            self.cap = Cap::ZERO;
         } else {
             let ptr = unsafe {
                 // `Layout::array` cannot overflow here because it would have
@@ -460,7 +483,10 @@ impl<T, A: Allocator> RawVec<T, A> {
                     .shrink(ptr, layout, new_layout)
                     .map_err(|_| AllocError { layout: new_layout, non_exhaustive: () })?
             };
-            self.set_ptr_and_cap(ptr, cap);
+            // SAFETY: if the allocation is valid, then the capacity is too
+            unsafe {
+                self.set_ptr_and_cap(ptr, cap);
+            }
         }
         Ok(())
     }

library/alloc/src/raw_vec/tests.rs

@@ -1,4 +1,5 @@
 use super::*;
+use core::mem::size_of;
 use std::cell::Cell;
 
 #[test]
@@ -161,3 +162,11 @@ fn zst_reserve_exact_panic() {
 
     v.reserve_exact(101, usize::MAX - 100);
 }
+
+#[test]
+fn niches() {
+    let baseline = size_of::<RawVec<u8>>();
+    assert_eq!(size_of::<Option<RawVec<u8>>>(), baseline);
+    assert_eq!(size_of::<Option<Option<RawVec<u8>>>>(), baseline);
+    assert_eq!(size_of::<Option<Option<Option<RawVec<u8>>>>>(), baseline);
+}

src/etc/gdb_providers.py

@@ -154,7 +154,11 @@ def __init__(self, valobj):
         self._valobj = valobj
         self._head = int(valobj["head"])
         self._size = int(valobj["len"])
-        self._cap = int(valobj["buf"]["cap"])
+        # BACKCOMPAT: rust 1.75
+        cap = valobj["buf"]["cap"]
+        if cap.type.code != gdb.TYPE_CODE_INT:
+            cap = cap[ZERO_FIELD]
+        self._cap = int(cap)
         self._data_ptr = unwrap_unique_or_non_null(valobj["buf"]["ptr"])
 
     def to_string(self):

src/etc/lldb_providers.py

@@ -267,7 +267,8 @@ class StdVecSyntheticProvider:
     """Pretty-printer for alloc::vec::Vec<T>
 
     struct Vec<T> { buf: RawVec<T>, len: usize }
-    struct RawVec<T> { ptr: Unique<T>, cap: usize, ... }
+    rust 1.75: struct RawVec<T> { ptr: Unique<T>, cap: usize, ... }
+    rust 1.76: struct RawVec<T> { ptr: Unique<T>, cap: Cap(usize), ... }
     rust 1.31.1: struct Unique<T: ?Sized> { pointer: NonZero<*const T>, ... }
     rust 1.33.0: struct Unique<T: ?Sized> { pointer: *const T, ... }
     rust 1.62.0: struct Unique<T: ?Sized> { pointer: NonNull<T>, ... }
@@ -390,7 +391,10 @@ def update(self):
         self.head = self.valobj.GetChildMemberWithName("head").GetValueAsUnsigned()
         self.size = self.valobj.GetChildMemberWithName("len").GetValueAsUnsigned()
         self.buf = self.valobj.GetChildMemberWithName("buf")
-        self.cap = self.buf.GetChildMemberWithName("cap").GetValueAsUnsigned()
+        cap = self.buf.GetChildMemberWithName("cap")
+        if cap.GetType().num_fields == 1:
+            cap = cap.GetChildAtIndex(0)
+        self.cap = cap.GetValueAsUnsigned()
 
         self.data_ptr = unwrap_unique_or_non_null(self.buf.GetChildMemberWithName("ptr"))
 

src/etc/natvis/liballoc.natvis

@@ -4,7 +4,7 @@
     <DisplayString>{{ len={len} }}</DisplayString>
     <Expand>
       <Item Name="[len]" ExcludeView="simple">len</Item>
-      <Item Name="[capacity]" ExcludeView="simple">buf.cap</Item>
+      <Item Name="[capacity]" ExcludeView="simple">buf.cap.__0</Item>
       <ArrayItems>
         <Size>len</Size>
         <ValuePointer>buf.ptr.pointer.pointer</ValuePointer>
@@ -15,15 +15,15 @@
     <DisplayString>{{ len={len} }}</DisplayString>
     <Expand>
       <Item Name="[len]" ExcludeView="simple">len</Item>
-      <Item Name="[capacity]" ExcludeView="simple">buf.cap</Item>
+      <Item Name="[capacity]" ExcludeView="simple">buf.cap.__0</Item>
       <CustomListItems>
         <Variable Name="i" InitialValue="0" />
         <Size>len</Size>
         <Loop>
           <If Condition="i == len">
             <Break/>
           </If>
-          <Item>buf.ptr.pointer.pointer[(i + head) % buf.cap]</Item>
+          <Item>buf.ptr.pointer.pointer[(i + head) % buf.cap.__0]</Item>
           <Exec>i = i + 1</Exec>
         </Loop>
       </CustomListItems>
@@ -45,7 +45,7 @@
     <StringView>(char*)vec.buf.ptr.pointer.pointer,[vec.len]s8</StringView>
     <Expand>
       <Item Name="[len]" ExcludeView="simple">vec.len</Item>
-      <Item Name="[capacity]" ExcludeView="simple">vec.buf.cap</Item>
+      <Item Name="[capacity]" ExcludeView="simple">vec.buf.cap.__0</Item>
       <Synthetic Name="[chars]">
         <DisplayString>{(char*)vec.buf.ptr.pointer.pointer,[vec.len]s8}</DisplayString>
         <Expand>

tests/codegen/issues/issue-86106.rs

@@ -9,19 +9,25 @@
 // CHECK-LABEL: define {{(dso_local )?}}void @string_new
 #[no_mangle]
 pub fn string_new() -> String {
-    // CHECK: store ptr inttoptr
+    // CHECK-NOT: load i8
+    // CHECK: store i{{32|64}}
     // CHECK-NEXT: getelementptr
-    // CHECK-NEXT: call void @llvm.memset
+    // CHECK-NEXT: store ptr
+    // CHECK-NEXT: getelementptr
+    // CHECK-NEXT: store i{{32|64}}
     // CHECK-NEXT: ret void
     String::new()
 }
 
 // CHECK-LABEL: define {{(dso_local )?}}void @empty_to_string
 #[no_mangle]
 pub fn empty_to_string() -> String {
-    // CHECK: store ptr inttoptr
+    // CHECK-NOT: load i8
+    // CHECK: store i{{32|64}}
+    // CHECK-NEXT: getelementptr
+    // CHECK-NEXT: store ptr
     // CHECK-NEXT: getelementptr
-    // CHECK-NEXT: call void @llvm.memset
+    // CHECK-NEXT: store i{{32|64}}
     // CHECK-NEXT: ret void
     "".to_string()
 }
@@ -32,19 +38,25 @@ pub fn empty_to_string() -> String {
 // CHECK-LABEL: @empty_vec
 #[no_mangle]
 pub fn empty_vec() -> Vec<u8> {
-    // CHECK: store ptr inttoptr
+    // CHECK: store i{{32|64}}
+    // CHECK-NOT: load i8
     // CHECK-NEXT: getelementptr
-    // CHECK-NEXT: call void @llvm.memset
+    // CHECK-NEXT: store ptr
+    // CHECK-NEXT: getelementptr
+    // CHECK-NEXT: store i{{32|64}}
     // CHECK-NEXT: ret void
     vec![]
 }
 
 // CHECK-LABEL: @empty_vec_clone
 #[no_mangle]
 pub fn empty_vec_clone() -> Vec<u8> {
-    // CHECK: store ptr inttoptr
+    // CHECK: store i{{32|64}}
+    // CHECK-NOT: load i8
+    // CHECK-NEXT: getelementptr
+    // CHECK-NEXT: store ptr
     // CHECK-NEXT: getelementptr
-    // CHECK-NEXT: call void @llvm.memset
+    // CHECK-NEXT: store i{{32|64}}
     // CHECK-NEXT: ret void
     vec![].clone()
 }

tests/ui/hygiene/panic-location.run.stderr

@@ -1,3 +1,3 @@
-thread 'main' panicked at library/alloc/src/raw_vec.rs:545:5:
+thread 'main' panicked at library/alloc/src/raw_vec.rs:571:5:
 capacity overflow
 note: run with `RUST_BACKTRACE=1` environment variable to display a backtrace