Auto merge of rust-lang#109035 - scottmcm:ptr-read-should-know-undef,…

… r=WaffleLapkin,JakobDegen

Ensure `ptr::read` gets all the same LLVM `load` metadata that dereferencing does

I was looking into `array::IntoIter` optimization, and noticed that it wasn't annotating the loads with `noundef` for simple things like `array::IntoIter<i32, N>`.  Trying to narrow it down, it seems that was because `MaybeUninit::assume_init_read` isn't marking the load as initialized (<https://rust.godbolt.org/z/Mxd8TPTnv>), which is unfortunate since that's basically its reason to exist.

The root cause is that `ptr::read` is currently implemented via the *untyped* `copy_nonoverlapping`, and thus the `load` doesn't get any type-aware metadata: no `noundef`, no `!range`.  This PR solves that by lowering `ptr::read(p)` to `copy *p` in MIR, for which the backends already do the right thing.

Fortuitiously, this also improves the IR we give to LLVM for things like `mem::replace`, and fixes a couple of long-standing bugs where `ptr::read` on `Copy` types was worse than `*`ing them.

Zulip conversation: <https://rust-lang.zulipchat.com/#narrow/stream/219381-t-libs/topic/Move.20array.3A.3AIntoIter.20to.20ManuallyDrop/near/341189936>

cc `@erikdesjardins` `@JakobDegen` `@workingjubilee` `@the8472`

Fixes rust-lang#106369
Fixes rust-lang#73258

compiler/rustc_hir_analysis/src/check/intrinsic.rs

@@ -363,6 +363,8 @@ pub fn check_intrinsic_type(tcx: TyCtxt<'_>, it: &hir::ForeignItem<'_>) {
             sym::likely => (0, vec![tcx.types.bool], tcx.types.bool),
             sym::unlikely => (0, vec![tcx.types.bool], tcx.types.bool),
 
+            sym::read_via_copy => (1, vec![tcx.mk_imm_ptr(param(0))], param(0)),
+
             sym::discriminant_value => {
                 let assoc_items = tcx.associated_item_def_ids(
                     tcx.require_lang_item(hir::LangItem::DiscriminantKind, None),

compiler/rustc_lint_defs/src/builtin.rs

@@ -1026,12 +1026,13 @@ declare_lint! {
     /// ### Example
     ///
     /// ```rust,compile_fail
-    /// #![feature(const_ptr_read)]
+    /// #![feature(const_mut_refs)]
     /// const FOO: () = unsafe {
     ///     let x = &[0_u8; 4];
     ///     let y = x.as_ptr().cast::<u32>();
-    ///     y.read(); // the address of a `u8` array is unknown and thus we don't know if
-    ///     // it is aligned enough for reading a `u32`.
+    ///     let mut z = 123;
+    ///     y.copy_to_nonoverlapping(&mut z, 1); // the address of a `u8` array is unknown
+    ///     // and thus we don't know if it is aligned enough for copying a `u32`.
     /// };
     /// ```
     ///

compiler/rustc_mir_transform/src/lower_intrinsics.rs

@@ -149,6 +149,35 @@ impl<'tcx> MirPass<'tcx> for LowerIntrinsics {
                             terminator.kind = TerminatorKind::Goto { target };
                         }
                     }
+                    sym::read_via_copy => {
+                        let [arg] = args.as_slice() else {
+                            span_bug!(terminator.source_info.span, "Wrong number of arguments");
+                        };
+                        let derefed_place =
+                            if let Some(place) = arg.place() && let Some(local) = place.as_local() {
+                                tcx.mk_place_deref(local.into())
+                            } else {
+                                span_bug!(terminator.source_info.span, "Only passing a local is supported");
+                            };
+                        terminator.kind = match *target {
+                            None => {
+                                // No target means this read something uninhabited,
+                                // so it must be unreachable, and we don't need to
+                                // preserve the assignment either.
+                                TerminatorKind::Unreachable
+                            }
+                            Some(target) => {
+                                block.statements.push(Statement {
+                                    source_info: terminator.source_info,
+                                    kind: StatementKind::Assign(Box::new((
+                                        *destination,
+                                        Rvalue::Use(Operand::Copy(derefed_place)),
+                                    ))),
+                                });
+                                TerminatorKind::Goto { target }
+                            }
+                        }
+                    }
                     sym::discriminant_value => {
                         if let (Some(target), Some(arg)) = (*target, args[0].place()) {
                             let arg = tcx.mk_place_deref(arg);

compiler/rustc_span/src/symbol.rs

@@ -1153,6 +1153,7 @@ symbols! {
         read_enum_variant_arg,
         read_struct,
         read_struct_field,
+        read_via_copy,
         readonly,
         realloc,
         reason,

library/core/src/intrinsics.rs

@@ -2020,6 +2020,16 @@ extern "rust-intrinsic" {
     #[rustc_safe_intrinsic]
     pub fn saturating_sub<T: Copy>(a: T, b: T) -> T;
 
+    /// This is an implementation detail of [`crate::ptr::read`] and should
+    /// not be used anywhere else.  See its comments for why this exists.
+    ///
+    /// This intrinsic can *only* be called where the argument is a local without
+    /// projections (`read_via_copy(p)`, not `read_via_copy(*p)`) so that it
+    /// trivially obeys runtime-MIR rules about derefs in operands.
+    #[cfg(not(bootstrap))]
+    #[rustc_const_unstable(feature = "const_ptr_read", issue = "80377")]
+    pub fn read_via_copy<T>(p: *const T) -> T;
+
     /// Returns the value of the discriminant for the variant in 'v';
     /// if `T` has no discriminant, returns `0`.
     ///

library/core/src/ptr/mod.rs

@@ -1135,27 +1135,58 @@ pub const unsafe fn replace<T>(dst: *mut T, mut src: T) -> T {
 #[rustc_const_unstable(feature = "const_ptr_read", issue = "80377")]
 #[cfg_attr(miri, track_caller)] // even without panics, this helps for Miri backtraces
 pub const unsafe fn read<T>(src: *const T) -> T {
-    // We are calling the intrinsics directly to avoid function calls in the generated code
-    // as `intrinsics::copy_nonoverlapping` is a wrapper function.
-    extern "rust-intrinsic" {
-        #[rustc_const_stable(feature = "const_intrinsic_copy", since = "1.63.0")]
-        fn copy_nonoverlapping<T>(src: *const T, dst: *mut T, count: usize);
-    }
+    // It would be semantically correct to implement this via `copy_nonoverlapping`
+    // and `MaybeUninit`, as was done before PR #109035. Calling `assume_init`
+    // provides enough information to know that this is a typed operation.
 
-    let mut tmp = MaybeUninit::<T>::uninit();
-    // SAFETY: the caller must guarantee that `src` is valid for reads.
-    // `src` cannot overlap `tmp` because `tmp` was just allocated on
-    // the stack as a separate allocated object.
+    // However, as of March 2023 the compiler was not capable of taking advantage
+    // of that information.  Thus the implementation here switched to an intrinsic,
+    // which lowers to `_0 = *src` in MIR, to address a few issues:
     //
-    // Also, since we just wrote a valid value into `tmp`, it is guaranteed
-    // to be properly initialized.
+    // - Using `MaybeUninit::assume_init` after a `copy_nonoverlapping` was not
+    //   turning the untyped copy into a typed load. As such, the generated
+    //   `load` in LLVM didn't get various metadata, such as `!range` (#73258),
+    //   `!nonnull`, and `!noundef`, resulting in poorer optimization.
+    // - Going through the extra local resulted in multiple extra copies, even
+    //   in optimized MIR.  (Ignoring StorageLive/Dead, the intrinsic is one
+    //   MIR statement, while the previous implementation was eight.)  LLVM
+    //   could sometimes optimize them away, but because `read` is at the core
+    //   of so many things, not having them in the first place improves what we
+    //   hand off to the backend.  For example, `mem::replace::<Big>` previously
+    //   emitted 4 `alloca` and 6 `memcpy`s, but is now 1 `alloc` and 3 `memcpy`s.
+    // - In general, this approach keeps us from getting any more bugs (like
+    //   #106369) that boil down to "`read(p)` is worse than `*p`", as this
+    //   makes them look identical to the backend (or other MIR consumers).
+    //
+    // Future enhancements to MIR optimizations might well allow this to return
+    // to the previous implementation, rather than using an intrinsic.
+
+    // SAFETY: the caller must guarantee that `src` is valid for reads.
     unsafe {
         assert_unsafe_precondition!(
             "ptr::read requires that the pointer argument is aligned and non-null",
             [T](src: *const T) => is_aligned_and_not_null(src)
         );
-        copy_nonoverlapping(src, tmp.as_mut_ptr(), 1);
-        tmp.assume_init()
+
+        #[cfg(bootstrap)]
+        {
+            // We are calling the intrinsics directly to avoid function calls in the
+            // generated code as `intrinsics::copy_nonoverlapping` is a wrapper function.
+            extern "rust-intrinsic" {
+                #[rustc_const_stable(feature = "const_intrinsic_copy", since = "1.63.0")]
+                fn copy_nonoverlapping<T>(src: *const T, dst: *mut T, count: usize);
+            }
+
+            // `src` cannot overlap `tmp` because `tmp` was just allocated on
+            // the stack as a separate allocated object.
+            let mut tmp = MaybeUninit::<T>::uninit();
+            copy_nonoverlapping(src, tmp.as_mut_ptr(), 1);
+            tmp.assume_init()
+        }
+        #[cfg(not(bootstrap))]
+        {
+            crate::intrinsics::read_via_copy(src)
+        }
     }
 }
 

tests/codegen/issues/issue-106369.rs

@@ -0,0 +1,15 @@
+// compile-flags: -O
+// ignore-debug (the extra assertions get in the way)
+
+#![crate_type = "lib"]
+
+// From <https://github.com/rust-lang/rust/issues/106369#issuecomment-1369095304>
+
+// CHECK-LABEL: @issue_106369(
+#[no_mangle]
+pub unsafe fn issue_106369(ptr: *const &i32) -> bool {
+    // CHECK-NOT: icmp
+    // CHECK: ret i1 true
+    // CHECK-NOT: icmp
+    Some(std::ptr::read(ptr)).is_some()
+}

tests/codegen/issues/issue-73258.rs

@@ -0,0 +1,38 @@
+// compile-flags: -O
+// ignore-debug (the extra assertions get in the way)
+
+#![crate_type = "lib"]
+
+// Adapted from <https://github.com/rust-lang/rust/issues/73258#issue-637346014>
+
+#[derive(Clone, Copy)]
+#[repr(u8)]
+pub enum Foo {
+    A, B, C, D,
+}
+
+// CHECK-LABEL: @issue_73258(
+#[no_mangle]
+pub unsafe fn issue_73258(ptr: *const Foo) -> Foo {
+    // CHECK-NOT: icmp
+    // CHECK-NOT: call
+    // CHECK-NOT: br
+    // CHECK-NOT: select
+
+    // CHECK: %[[R:.+]] = load i8
+    // CHECK-SAME: !range !
+
+    // CHECK-NOT: icmp
+    // CHECK-NOT: call
+    // CHECK-NOT: br
+    // CHECK-NOT: select
+
+    // CHECK: ret i8 %[[R]]
+
+    // CHECK-NOT: icmp
+    // CHECK-NOT: call
+    // CHECK-NOT: br
+    // CHECK-NOT: select
+    let k: Option<Foo> = Some(ptr.read());
+    return k.unwrap();
+}

tests/codegen/mem-replace-big-type.rs

@@ -0,0 +1,36 @@
+// This test ensures that `mem::replace::<T>` only ever calls `@llvm.memcpy`
+// with `size_of::<T>()` as the size, and never goes through any wrapper that
+// may e.g. multiply `size_of::<T>()` with a variable "count" (which is only
+// known to be `1` after inlining).
+
+// compile-flags: -C no-prepopulate-passes -Zinline-mir=no
+// ignore-debug: the debug assertions get in the way
+
+#![crate_type = "lib"]
+
+#[repr(C, align(8))]
+pub struct Big([u64; 7]);
+pub fn replace_big(dst: &mut Big, src: Big) -> Big {
+    // Before the `read_via_copy` intrinsic, this emitted six `memcpy`s.
+    std::mem::replace(dst, src)
+}
+
+// NOTE(eddyb) the `CHECK-NOT`s ensure that the only calls of `@llvm.memcpy` in
+// the entire output, are the direct calls we want, from `ptr::replace`.
+
+// CHECK-NOT: call void @llvm.memcpy
+
+// For a large type, we expect exactly three `memcpy`s
+// CHECK-LABEL: define internal void @{{.+}}mem{{.+}}replace{{.+}}sret(%Big)
+    // CHECK-NOT: alloca
+    // CHECK: alloca %Big
+    // CHECK-NOT: alloca
+    // CHECK-NOT: call void @llvm.memcpy
+    // CHECK: call void @llvm.memcpy.{{.+}}({{i8\*|ptr}} align 8 %{{.*}}, {{i8\*|ptr}} align 8 %{{.*}}, i{{.*}} 56, i1 false)
+    // CHECK-NOT: call void @llvm.memcpy
+    // CHECK: call void @llvm.memcpy.{{.+}}({{i8\*|ptr}} align 8 %{{.*}}, {{i8\*|ptr}} align 8 %{{.*}}, i{{.*}} 56, i1 false)
+    // CHECK-NOT: call void @llvm.memcpy
+    // CHECK: call void @llvm.memcpy.{{.+}}({{i8\*|ptr}} align 8 %{{.*}}, {{i8\*|ptr}} align 8 %{{.*}}, i{{.*}} 56, i1 false)
+    // CHECK-NOT: call void @llvm.memcpy
+
+// CHECK-NOT: call void @llvm.memcpy

tests/codegen/mem-replace-direct-memcpy.rs

@@ -13,12 +13,21 @@ pub fn replace_byte(dst: &mut u8, src: u8) -> u8 {
 }
 
 // NOTE(eddyb) the `CHECK-NOT`s ensure that the only calls of `@llvm.memcpy` in
-// the entire output, are the two direct calls we want, from `ptr::replace`.
+// the entire output, are the direct calls we want, from `ptr::replace`.
 
 // CHECK-NOT: call void @llvm.memcpy
-// CHECK: ; core::mem::replace
-// CHECK-NOT: call void @llvm.memcpy
-// CHECK: call void @llvm.memcpy.{{.+}}({{i8\*|ptr}} align 1 %{{.*}}, {{i8\*|ptr}} align 1 %{{.*}}, i{{.*}} 1, i1 false)
-// CHECK-NOT: call void @llvm.memcpy
-// CHECK: call void @llvm.memcpy.{{.+}}({{i8\*|ptr}} align 1 %{{.*}}, {{i8\*|ptr}} align 1 %{{.*}}, i{{.*}} 1, i1 false)
+
+// For a small type, we expect one each of `load`/`store`/`memcpy` instead
+// CHECK-LABEL: define internal noundef i8 @{{.+}}mem{{.+}}replace
+    // CHECK-NOT: alloca
+    // CHECK: alloca i8
+    // CHECK-NOT: alloca
+    // CHECK-NOT: call void @llvm.memcpy
+    // CHECK: load i8
+    // CHECK-NOT: call void @llvm.memcpy
+    // CHECK: store i8
+    // CHECK-NOT: call void @llvm.memcpy
+    // CHECK: call void @llvm.memcpy.{{.+}}({{i8\*|ptr}} align 1 %{{.*}}, {{i8\*|ptr}} align 1 %{{.*}}, i{{.*}} 1, i1 false)
+    // CHECK-NOT: call void @llvm.memcpy
+
 // CHECK-NOT: call void @llvm.memcpy

tests/codegen/ptr-read-metadata.rs

@@ -0,0 +1,96 @@
+// compile-flags: -O -Z merge-functions=disabled
+// no-system-llvm
+// ignore-debug (the extra assertions get in the way)
+
+#![crate_type = "lib"]
+
+// Ensure that various forms of reading pointers correctly annotate the `load`s
+// with `!noundef` and `!range` metadata to enable extra optimization.
+
+use std::mem::MaybeUninit;
+
+// CHECK-LABEL: define noundef i8 @copy_byte(
+#[no_mangle]
+pub unsafe fn copy_byte(p: *const u8) -> u8 {
+    // CHECK-NOT: load
+    // CHECK: load i8, ptr %p, align 1
+    // CHECK-SAME: !noundef !
+    // CHECK-NOT: load
+    *p
+}
+
+// CHECK-LABEL: define noundef i8 @read_byte(
+#[no_mangle]
+pub unsafe fn read_byte(p: *const u8) -> u8 {
+    // CHECK-NOT: load
+    // CHECK: load i8, ptr %p, align 1
+    // CHECK-SAME: !noundef !
+    // CHECK-NOT: load
+    p.read()
+}
+
+// CHECK-LABEL: define i8 @read_byte_maybe_uninit(
+#[no_mangle]
+pub unsafe fn read_byte_maybe_uninit(p: *const MaybeUninit<u8>) -> MaybeUninit<u8> {
+    // CHECK-NOT: load
+    // CHECK: load i8, ptr %p, align 1
+    // CHECK-NOT: noundef
+    // CHECK-NOT: load
+    p.read()
+}
+
+// CHECK-LABEL: define noundef i8 @read_byte_assume_init(
+#[no_mangle]
+pub unsafe fn read_byte_assume_init(p: &MaybeUninit<u8>) -> u8 {
+    // CHECK-NOT: load
+    // CHECK: load i8, ptr %p, align 1
+    // CHECK-SAME: !noundef !
+    // CHECK-NOT: load
+    p.assume_init_read()
+}
+
+// CHECK-LABEL: define noundef i32 @copy_char(
+#[no_mangle]
+pub unsafe fn copy_char(p: *const char) -> char {
+    // CHECK-NOT: load
+    // CHECK: load i32, ptr %p
+    // CHECK-SAME: !range ![[RANGE:[0-9]+]]
+    // CHECK-SAME: !noundef !
+    // CHECK-NOT: load
+    *p
+}
+
+// CHECK-LABEL: define noundef i32 @read_char(
+#[no_mangle]
+pub unsafe fn read_char(p: *const char) -> char {
+    // CHECK-NOT: load
+    // CHECK: load i32, ptr %p
+    // CHECK-SAME: !range ![[RANGE]]
+    // CHECK-SAME: !noundef !
+    // CHECK-NOT: load
+    p.read()
+}
+
+// CHECK-LABEL: define i32 @read_char_maybe_uninit(
+#[no_mangle]
+pub unsafe fn read_char_maybe_uninit(p: *const MaybeUninit<char>) -> MaybeUninit<char> {
+    // CHECK-NOT: load
+    // CHECK: load i32, ptr %p
+    // CHECK-NOT: range
+    // CHECK-NOT: noundef
+    // CHECK-NOT: load
+    p.read()
+}
+
+// CHECK-LABEL: define noundef i32 @read_char_assume_init(
+#[no_mangle]
+pub unsafe fn read_char_assume_init(p: &MaybeUninit<char>) -> char {
+    // CHECK-NOT: load
+    // CHECK: load i32, ptr %p
+    // CHECK-SAME: !range ![[RANGE]]
+    // CHECK-SAME: !noundef !
+    // CHECK-NOT: load
+    p.assume_init_read()
+}
+
+// CHECK: ![[RANGE]] = !{i32 0, i32 1114112}