Rollup merge of rust-lang#127275 - RalfJung:offset-from-isize-min, r=Amanieu

offset_from, offset: clearly separate safety requirements the user needs to prove from corollaries that automatically follow

By landing rust-lang#116675 we decided that objects larger than `isize::MAX` cannot exist in the address space of a Rust program, which lets us simplify these rules.

For `offset_from`, we can even state that the *absolute* distance fits into an `isize`, and therefore exclude `isize::MIN`. This PR also changes Miri to treat an `isize::MIN` difference like the other isize-overflowing cases.

Author: matthiaskrgr

core/src/ptr/const_ptr.rs

@@ -390,37 +390,26 @@ impl<T: ?Sized> *const T {
         if self.is_null() { None } else { Some(unsafe { &*(self as *const MaybeUninit<T>) }) }
     }
 
-    /// Calculates the offset from a pointer.
+    /// Adds an offset to a pointer.
     ///
     /// `count` is in units of T; e.g., a `count` of 3 represents a pointer
     /// offset of `3 * size_of::<T>()` bytes.
     ///
     /// # Safety
     ///
-    /// If any of the following conditions are violated, the result is Undefined
-    /// Behavior:
+    /// If any of the following conditions are violated, the result is Undefined Behavior:
     ///
-    /// * If the computed offset, **in bytes**, is non-zero, then both the starting and resulting
-    ///   pointer must be either in bounds or at the end of the same [allocated object].
-    ///   (If it is zero, then the function is always well-defined.)
+    /// * The computed offset, `count * size_of::<T>()` bytes, must not overflow `isize`.
     ///
-    /// * The computed offset, **in bytes**, cannot overflow an `isize`.
+    /// * If the computed offset is non-zero, then `self` must be derived from a pointer to some
+    ///   [allocated object], and the entire memory range between `self` and the result must be in
+    ///   bounds of that allocated object. In particular, this range must not "wrap around" the edge
+    ///   of the address space.
     ///
-    /// * The offset being in bounds cannot rely on "wrapping around" the address
-    ///   space. That is, the infinite-precision sum, **in bytes** must fit in a usize.
-    ///
-    /// The compiler and standard library generally tries to ensure allocations
-    /// never reach a size where an offset is a concern. For instance, `Vec`
-    /// and `Box` ensure they never allocate more than `isize::MAX` bytes, so
-    /// `vec.as_ptr().add(vec.len())` is always safe.
-    ///
-    /// Most platforms fundamentally can't even construct such an allocation.
-    /// For instance, no known 64-bit platform can ever serve a request
-    /// for 2<sup>63</sup> bytes due to page-table limitations or splitting the address space.
-    /// However, some 32-bit and 16-bit platforms may successfully serve a request for
-    /// more than `isize::MAX` bytes with things like Physical Address
-    /// Extension. As such, memory acquired directly from allocators or memory
-    /// mapped files *may* be too large to handle with this function.
+    /// Allocated objects can never be larger than `isize::MAX` bytes, so if the computed offset
+    /// stays in bounds of the allocated object, it is guaranteed to satisfy the first requirement.
+    /// This implies, for instance, that `vec.as_ptr().add(vec.len())` (for `vec: Vec<T>`) is always
+    /// safe.
     ///
     /// Consider using [`wrapping_offset`] instead if these constraints are
     /// difficult to satisfy. The only advantage of this method is that it
@@ -611,8 +600,7 @@ impl<T: ?Sized> *const T {
     ///
     /// # Safety
     ///
-    /// If any of the following conditions are violated, the result is Undefined
-    /// Behavior:
+    /// If any of the following conditions are violated, the result is Undefined Behavior:
     ///
     /// * `self` and `origin` must either
     ///
@@ -623,26 +611,10 @@ impl<T: ?Sized> *const T {
     /// * The distance between the pointers, in bytes, must be an exact multiple
     ///   of the size of `T`.
     ///
-    /// * The distance between the pointers, **in bytes**, cannot overflow an `isize`.
-    ///
-    /// * The distance being in bounds cannot rely on "wrapping around" the address space.
-    ///
-    /// Rust types are never larger than `isize::MAX` and Rust allocations never wrap around the
-    /// address space, so two pointers within some value of any Rust type `T` will always satisfy
-    /// the last two conditions. The standard library also generally ensures that allocations
-    /// never reach a size where an offset is a concern. For instance, `Vec` and `Box` ensure they
-    /// never allocate more than `isize::MAX` bytes, so `ptr_into_vec.offset_from(vec.as_ptr())`
-    /// always satisfies the last two conditions.
-    ///
-    /// Most platforms fundamentally can't even construct such a large allocation.
-    /// For instance, no known 64-bit platform can ever serve a request
-    /// for 2<sup>63</sup> bytes due to page-table limitations or splitting the address space.
-    /// However, some 32-bit and 16-bit platforms may successfully serve a request for
-    /// more than `isize::MAX` bytes with things like Physical Address
-    /// Extension. As such, memory acquired directly from allocators or memory
-    /// mapped files *may* be too large to handle with this function.
-    /// (Note that [`offset`] and [`add`] also have a similar limitation and hence cannot be used on
-    /// such large allocations either.)
+    /// As a consequence, the absolute distance between the pointers, in bytes, computed on
+    /// mathematical integers (without "wrapping around"), cannot overflow an `isize`. This is
+    /// implied by the in-bounds requirement, and the fact that no allocated object can be larger
+    /// than `isize::MAX` bytes.
     ///
     /// The requirement for pointers to be derived from the same allocated object is primarily
     /// needed for `const`-compatibility: the distance between pointers into *different* allocated
@@ -879,37 +851,26 @@ impl<T: ?Sized> *const T {
         }
     }
 
-    /// Calculates the offset from a pointer (convenience for `.offset(count as isize)`).
+    /// Adds an offset to a pointer (convenience for `.offset(count as isize)`).
     ///
     /// `count` is in units of T; e.g., a `count` of 3 represents a pointer
     /// offset of `3 * size_of::<T>()` bytes.
     ///
     /// # Safety
     ///
-    /// If any of the following conditions are violated, the result is Undefined
-    /// Behavior:
+    /// If any of the following conditions are violated, the result is Undefined Behavior:
     ///
-    /// * If the computed offset, **in bytes**, is non-zero, then both the starting and resulting
-    ///   pointer must be either in bounds or at the end of the same [allocated object].
-    ///   (If it is zero, then the function is always well-defined.)
+    /// * The computed offset, `count * size_of::<T>()` bytes, must not overflow `isize`.
     ///
-    /// * The computed offset, **in bytes**, cannot overflow an `isize`.
+    /// * If the computed offset is non-zero, then `self` must be derived from a pointer to some
+    ///   [allocated object], and the entire memory range between `self` and the result must be in
+    ///   bounds of that allocated object. In particular, this range must not "wrap around" the edge
+    ///   of the address space.
     ///
-    /// * The offset being in bounds cannot rely on "wrapping around" the address
-    ///   space. That is, the infinite-precision sum must fit in a `usize`.
-    ///
-    /// The compiler and standard library generally tries to ensure allocations
-    /// never reach a size where an offset is a concern. For instance, `Vec`
-    /// and `Box` ensure they never allocate more than `isize::MAX` bytes, so
-    /// `vec.as_ptr().add(vec.len())` is always safe.
-    ///
-    /// Most platforms fundamentally can't even construct such an allocation.
-    /// For instance, no known 64-bit platform can ever serve a request
-    /// for 2<sup>63</sup> bytes due to page-table limitations or splitting the address space.
-    /// However, some 32-bit and 16-bit platforms may successfully serve a request for
-    /// more than `isize::MAX` bytes with things like Physical Address
-    /// Extension. As such, memory acquired directly from allocators or memory
-    /// mapped files *may* be too large to handle with this function.
+    /// Allocated objects can never be larger than `isize::MAX` bytes, so if the computed offset
+    /// stays in bounds of the allocated object, it is guaranteed to satisfy the first requirement.
+    /// This implies, for instance, that `vec.as_ptr().add(vec.len())` (for `vec: Vec<T>`) is always
+    /// safe.
     ///
     /// Consider using [`wrapping_add`] instead if these constraints are
     /// difficult to satisfy. The only advantage of this method is that it
@@ -963,38 +924,27 @@ impl<T: ?Sized> *const T {
         unsafe { self.cast::<u8>().add(count).with_metadata_of(self) }
     }
 
-    /// Calculates the offset from a pointer (convenience for
+    /// Subtracts an offset from a pointer (convenience for
     /// `.offset((count as isize).wrapping_neg())`).
     ///
     /// `count` is in units of T; e.g., a `count` of 3 represents a pointer
     /// offset of `3 * size_of::<T>()` bytes.
     ///
     /// # Safety
     ///
-    /// If any of the following conditions are violated, the result is Undefined
-    /// Behavior:
-    ///
-    /// * If the computed offset, **in bytes**, is non-zero, then both the starting and resulting
-    ///   pointer must be either in bounds or at the end of the same [allocated object].
-    ///   (If it is zero, then the function is always well-defined.)
-    ///
-    /// * The computed offset cannot exceed `isize::MAX` **bytes**.
+    /// If any of the following conditions are violated, the result is Undefined Behavior:
     ///
-    /// * The offset being in bounds cannot rely on "wrapping around" the address
-    ///   space. That is, the infinite-precision sum must fit in a usize.
+    /// * The computed offset, `count * size_of::<T>()` bytes, must not overflow `isize`.
     ///
-    /// The compiler and standard library generally tries to ensure allocations
-    /// never reach a size where an offset is a concern. For instance, `Vec`
-    /// and `Box` ensure they never allocate more than `isize::MAX` bytes, so
-    /// `vec.as_ptr().add(vec.len()).sub(vec.len())` is always safe.
+    /// * If the computed offset is non-zero, then `self` must be derived from a pointer to some
+    ///   [allocated object], and the entire memory range between `self` and the result must be in
+    ///   bounds of that allocated object. In particular, this range must not "wrap around" the edge
+    ///   of the address space.
     ///
-    /// Most platforms fundamentally can't even construct such an allocation.
-    /// For instance, no known 64-bit platform can ever serve a request
-    /// for 2<sup>63</sup> bytes due to page-table limitations or splitting the address space.
-    /// However, some 32-bit and 16-bit platforms may successfully serve a request for
-    /// more than `isize::MAX` bytes with things like Physical Address
-    /// Extension. As such, memory acquired directly from allocators or memory
-    /// mapped files *may* be too large to handle with this function.
+    /// Allocated objects can never be larger than `isize::MAX` bytes, so if the computed offset
+    /// stays in bounds of the allocated object, it is guaranteed to satisfy the first requirement.
+    /// This implies, for instance, that `vec.as_ptr().add(vec.len())` (for `vec: Vec<T>`) is always
+    /// safe.
     ///
     /// Consider using [`wrapping_sub`] instead if these constraints are
     /// difficult to satisfy. The only advantage of this method is that it