Split large unsafe function slide (#2406)

The old slice was doing several things at the same time: demonstrating
both external functions as well as unsafe Rust functions.

We now treat those two topics separately. In addition, the “Calling
Unsafe Functions” heading has become its own slide with a non-crashing
example that shows what can go wrong if an argument is misunderstood
in a call to an unsafe function. The old example didn’t actually
illustrate the danger clearly: it would produce mangled UTF-8 output,
which the Playground server refuses to print.

Part of #2445.

---------

Co-authored-by: Dustin J. Mitchell <[email protected]>
Co-authored-by: Andrew Walbran <[email protected]>

Author: 3 people

src/SUMMARY.md

@@ -222,6 +222,9 @@
   - [Mutable Static Variables](unsafe-rust/mutable-static.md)
   - [Unions](unsafe-rust/unions.md)
   - [Unsafe Functions](unsafe-rust/unsafe-functions.md)
+    - [Unsafe Rust Functions](unsafe-rust/unsafe-functions/rust.md)
+    - [Unsafe External Functions](unsafe-rust/unsafe-functions/extern-c.md)
+    - [Calling Unsafe Functions](unsafe-rust/unsafe-functions/calling.md)
   - [Unsafe Traits](unsafe-rust/unsafe-traits.md)
   - [Exercise: FFI Wrapper](unsafe-rust/exercise.md)
     - [Solution](unsafe-rust/solution.md)

src/unsafe-rust/unsafe-functions.md

@@ -1,100 +1,19 @@
 ---
-minutes: 5
+minutes: 15
 ---
 
 # Unsafe Functions
 
-## Calling Unsafe Functions
-
 A function or method can be marked `unsafe` if it has extra preconditions you
-must uphold to avoid undefined behaviour:
-
-```rust,editable
-extern "C" {
-    fn abs(input: i32) -> i32;
-}
-
-fn main() {
-    let emojis = "🗻∈🌏";
-
-    // SAFETY: The indices are in the correct order, within the bounds of the
-    // string slice, and lie on UTF-8 sequence boundaries.
-    unsafe {
-        println!("emoji: {}", emojis.get_unchecked(0..4));
-        println!("emoji: {}", emojis.get_unchecked(4..7));
-        println!("emoji: {}", emojis.get_unchecked(7..11));
-    }
-
-    println!("char count: {}", count_chars(unsafe { emojis.get_unchecked(0..7) }));
-
-    // SAFETY: `abs` doesn't deal with pointers and doesn't have any safety
-    // requirements.
-    unsafe {
-        println!("Absolute value of -3 according to C: {}", abs(-3));
-    }
-
-    // Not upholding the UTF-8 encoding requirement breaks memory safety!
-    // println!("emoji: {}", unsafe { emojis.get_unchecked(0..3) });
-    // println!("char count: {}", count_chars(unsafe {
-    // emojis.get_unchecked(0..3) }));
-}
-
-fn count_chars(s: &str) -> usize {
-    s.chars().count()
-}
-```
-
-## Writing Unsafe Functions
+must uphold to avoid undefined behaviour.
 
-You can mark your own functions as `unsafe` if they require particular
-conditions to avoid undefined behaviour.
+There are two main categories:
 
-```rust,editable
-/// Swaps the values pointed to by the given pointers.
-///
-/// # Safety
-///
-/// The pointers must be valid and properly aligned.
-unsafe fn swap(a: *mut u8, b: *mut u8) {
-    let temp = *a;
-    *a = *b;
-    *b = temp;
-}
-
-fn main() {
-    let mut a = 42;
-    let mut b = 66;
-
-    // SAFETY: ...
-    unsafe {
-        swap(&mut a, &mut b);
-    }
-
-    println!("a = {}, b = {}", a, b);
-}
-```
+- Rust functions declared unsafe with `unsafe fn`.
+- Foreign functions in `extern "C"` blocks.
 
 <details>
 
-## Calling Unsafe Functions
-
-`get_unchecked`, like most `_unchecked` functions, is unsafe, because it can
-create UB if the range is incorrect. `abs` is unsafe for a different reason: it
-is an external function (FFI). Calling external functions is usually only a
-problem when those functions do things with pointers which might violate Rust's
-memory model, but in general any C function might have undefined behaviour under
-any arbitrary circumstances.
-
-The `"C"` in this example is the ABI;
-[other ABIs are available too](https://doc.rust-lang.org/reference/items/external-blocks.html).
-
-## Writing Unsafe Functions
-
-We wouldn't actually use pointers for a `swap` function - it can be done safely
-with references.
-
-Note that unsafe code is allowed within an unsafe function without an `unsafe`
-block. We can prohibit this with `#[deny(unsafe_op_in_unsafe_fn)]`. Try adding
-it and see what happens. This will likely change in a future Rust edition.
+We will look at the two kinds of unsafe functions next.
 
 </details>

src/unsafe-rust/unsafe-functions/calling.md

@@ -0,0 +1,48 @@
+# Calling Unsafe Functions
+
+Failing to uphold the safety requirements breaks memory safety!
+
+```rust,editable
+#[derive(Debug)]
+#[repr(C)]
+struct KeyPair {
+    pk: [u16; 4], // 8 bytes
+    sk: [u16; 4], // 8 bytes
+}
+
+const PK_BYTE_LEN: usize = 8;
+
+fn log_public_key(pk_ptr: *const u16) {
+    let pk: &[u16] = unsafe { std::slice::from_raw_parts(pk_ptr, PK_BYTE_LEN) };
+    println!("{pk:?}");
+}
+
+fn main() {
+    let key_pair = KeyPair { pk: [1, 2, 3, 4], sk: [0, 0, 42, 0] };
+    log_public_key(key_pair.pk.as_ptr());
+}
+```
+
+Always include a safety comment for each `unsafe` block. It must explain why the
+code is actually safe. This example is missing a safety comment and is unsound.
+
+<details>
+
+Key points:
+
+- The second argument to `slice::from_raw_parts` is the number of _elements_,
+  not bytes! This example demonstrates unexpected behavior by reading past the
+  end of one array and into another.
+- This is not actually undefined behaviour, as `KeyPair` has a defined
+  representation (due to `repr(C)`) and no padding, so the contents of the
+  second array is also valid to read through the same pointer.
+- `log_public_key` should be unsafe, because `pk_ptr` must meet certain
+  prerequisites to avoid undefined behaviour. A safe function which can cause
+  undefined behaviour is said to be `unsound`. What should its safety
+  documentation say?
+- The standard library contains many low-level unsafe functions. Prefer the safe
+  alternatives when possible!
+- If you use an unsafe function as an optimization, make sure to add a benchmark
+  to demonstrate the gain.
+
+</details>

src/unsafe-rust/unsafe-functions/extern-c.md

@@ -0,0 +1,44 @@
+# Unsafe External Functions
+
+Functions in a foreign language may also be unsafe:
+
+```rust,editable
+use std::ffi::c_char;
+
+unsafe extern "C" {
+    // `abs` doesn't deal with pointers and doesn't have any safety requirements.
+    safe fn abs(input: i32) -> i32;
+
+    /// # Safety
+    ///
+    /// `s` must be a pointer to a NUL-terminated C string which is valid and
+    /// not modified for the duration of this function call.
+    unsafe fn strlen(s: *const c_char) -> usize;
+}
+
+fn main() {
+    println!("Absolute value of -3 according to C: {}", abs(-3));
+
+    unsafe {
+        // SAFETY: We pass a pointer to a C string literal which is valid for
+        // the duration of the program.
+        println!("String length: {}", strlen(c"String".as_ptr()));
+    }
+}
+```
+
+<details>
+
+- Rust used to consider all extern functions unsafe, but this changed in Rust
+  1.82 with `unsafe extern` blocks.
+- `abs` must be explicitly marked as `safe` because it is an external function
+  (FFI). Calling external functions is usually only a problem when those
+  functions do things with pointers which which might violate Rust's memory
+  model, but in general any C function might have undefined behaviour under any
+  arbitrary circumstances.
+- The `"C"` in this example is the ABI;
+  [other ABIs are available too](https://doc.rust-lang.org/reference/items/external-blocks.html).
+- Note that there is no verification that the Rust function signature matches
+  that of the function definition -- that's up to you!
+
+</details>

src/unsafe-rust/unsafe-functions/rust.md

@@ -0,0 +1,43 @@
+# Unsafe Rust Functions
+
+You can mark your own functions as `unsafe` if they require particular
+preconditions to avoid undefined behaviour.
+
+```rust,editable
+/// Swaps the values pointed to by the given pointers.
+///
+/// # Safety
+///
+/// The pointers must be valid, properly aligned, and not otherwise accessed for
+/// the duration of the function call.
+unsafe fn swap(a: *mut u8, b: *mut u8) {
+    let temp = *a;
+    *a = *b;
+    *b = temp;
+}
+
+fn main() {
+    let mut a = 42;
+    let mut b = 66;
+
+    // SAFETY: The pointers must be valid, aligned and unique because they came
+    // from references.
+    unsafe {
+        swap(&mut a, &mut b);
+    }
+
+    println!("a = {}, b = {}", a, b);
+}
+```
+
+<details>
+
+We wouldn't actually use pointers for a `swap` function --- it can be done
+safely with references.
+
+Note that unsafe code is allowed within an unsafe function without an `unsafe`
+block. We can prohibit this with `#[deny(unsafe_op_in_unsafe_fn)]`. Try adding
+it and see what happens. This will
+[change in the 2024 Rust edition](https://github.com/rust-lang/rust/issues/120535).
+
+</details>