projects/linux·blob v6.4 · refs

591 lines · 18.0 KiB

1// SPDX-License-Identifier: GPL-2.0
2
3//! String representations.
4
5use alloc::vec::Vec;
6use core::fmt::{self, Write};
7use core::ops::{self, Deref, Index};
8
9use crate::{
10 bindings,
11 error::{code::*, Error},
12};
13
14/// Byte string without UTF-8 validity guarantee.
15///
16/// `BStr` is simply an alias to `[u8]`, but has a more evident semantical meaning.
17pub type BStr = [u8];
18
19/// Creates a new [`BStr`] from a string literal.
20///
21/// `b_str!` converts the supplied string literal to byte string, so non-ASCII
22/// characters can be included.
23///
24/// # Examples
25///
26/// ```
27/// # use kernel::b_str;
28/// # use kernel::str::BStr;
29/// const MY_BSTR: &BStr = b_str!("My awesome BStr!");
30/// ```
31#[macro_export]
32macro_rules! b_str {
33 ($str:literal) => {{
34 const S: &'static str = $str;
35 const C: &'static $crate::str::BStr = S.as_bytes();
36 C
37 }};
38}
39
40/// Possible errors when using conversion functions in [`CStr`].
41#[derive(Debug, Clone, Copy)]
42pub enum CStrConvertError {
43 /// Supplied bytes contain an interior `NUL`.
44 InteriorNul,
45
46 /// Supplied bytes are not terminated by `NUL`.
47 NotNulTerminated,
48}
49
50impl From<CStrConvertError> for Error {
51 #[inline]
52 fn from(_: CStrConvertError) -> Error {
53 EINVAL
54 }
55}
56
57/// A string that is guaranteed to have exactly one `NUL` byte, which is at the
58/// end.
59///
60/// Used for interoperability with kernel APIs that take C strings.
61#[repr(transparent)]
62pub struct CStr([u8]);
63
64impl CStr {
65 /// Returns the length of this string excluding `NUL`.
66 #[inline]
67 pub const fn len(&self) -> usize {
68 self.len_with_nul() - 1
69 }
70
71 /// Returns the length of this string with `NUL`.
72 #[inline]
73 pub const fn len_with_nul(&self) -> usize {
74 // SAFETY: This is one of the invariant of `CStr`.
75 // We add a `unreachable_unchecked` here to hint the optimizer that
76 // the value returned from this function is non-zero.
77 if self.0.is_empty() {
78 unsafe { core::hint::unreachable_unchecked() };
79 }
80 self.0.len()
81 }
82
83 /// Returns `true` if the string only includes `NUL`.
84 #[inline]
85 pub const fn is_empty(&self) -> bool {
86 self.len() == 0
87 }
88
89 /// Wraps a raw C string pointer.
90 ///
91 /// # Safety
92 ///
93 /// `ptr` must be a valid pointer to a `NUL`-terminated C string, and it must
94 /// last at least `'a`. When `CStr` is alive, the memory pointed by `ptr`
95 /// must not be mutated.
96 #[inline]
97 pub unsafe fn from_char_ptr<'a>(ptr: *const core::ffi::c_char) -> &'a Self {
98 // SAFETY: The safety precondition guarantees `ptr` is a valid pointer
99 // to a `NUL`-terminated C string.
100 let len = unsafe { bindings::strlen(ptr) } + 1;
101 // SAFETY: Lifetime guaranteed by the safety precondition.
102 let bytes = unsafe { core::slice::from_raw_parts(ptr as _, len as _) };
103 // SAFETY: As `len` is returned by `strlen`, `bytes` does not contain interior `NUL`.
104 // As we have added 1 to `len`, the last byte is known to be `NUL`.
105 unsafe { Self::from_bytes_with_nul_unchecked(bytes) }
106 }
107
108 /// Creates a [`CStr`] from a `[u8]`.
109 ///
110 /// The provided slice must be `NUL`-terminated, does not contain any
111 /// interior `NUL` bytes.
112 pub const fn from_bytes_with_nul(bytes: &[u8]) -> Result<&Self, CStrConvertError> {
113 if bytes.is_empty() {
114 return Err(CStrConvertError::NotNulTerminated);
115 }
116 if bytes[bytes.len() - 1] != 0 {
117 return Err(CStrConvertError::NotNulTerminated);
118 }
119 let mut i = 0;
120 // `i + 1 < bytes.len()` allows LLVM to optimize away bounds checking,
121 // while it couldn't optimize away bounds checks for `i < bytes.len() - 1`.
122 while i + 1 < bytes.len() {
123 if bytes[i] == 0 {
124 return Err(CStrConvertError::InteriorNul);
125 }
126 i += 1;
127 }
128 // SAFETY: We just checked that all properties hold.
129 Ok(unsafe { Self::from_bytes_with_nul_unchecked(bytes) })
130 }
131
132 /// Creates a [`CStr`] from a `[u8]` without performing any additional
133 /// checks.
134 ///
135 /// # Safety
136 ///
137 /// `bytes` *must* end with a `NUL` byte, and should only have a single
138 /// `NUL` byte (or the string will be truncated).
139 #[inline]
140 pub const unsafe fn from_bytes_with_nul_unchecked(bytes: &[u8]) -> &CStr {
141 // SAFETY: Properties of `bytes` guaranteed by the safety precondition.
142 unsafe { core::mem::transmute(bytes) }
143 }
144
145 /// Returns a C pointer to the string.
146 #[inline]
147 pub const fn as_char_ptr(&self) -> *const core::ffi::c_char {
148 self.0.as_ptr() as _
149 }
150
151 /// Convert the string to a byte slice without the trailing 0 byte.
152 #[inline]
153 pub fn as_bytes(&self) -> &[u8] {
154 &self.0[..self.len()]
155 }
156
157 /// Convert the string to a byte slice containing the trailing 0 byte.
158 #[inline]
159 pub const fn as_bytes_with_nul(&self) -> &[u8] {
160 &self.0
161 }
162
163 /// Yields a [`&str`] slice if the [`CStr`] contains valid UTF-8.
164 ///
165 /// If the contents of the [`CStr`] are valid UTF-8 data, this
166 /// function will return the corresponding [`&str`] slice. Otherwise,
167 /// it will return an error with details of where UTF-8 validation failed.
168 ///
169 /// # Examples
170 ///
171 /// ```
172 /// # use kernel::str::CStr;
173 /// let cstr = CStr::from_bytes_with_nul(b"foo\0").unwrap();
174 /// assert_eq!(cstr.to_str(), Ok("foo"));
175 /// ```
176 #[inline]
177 pub fn to_str(&self) -> Result<&str, core::str::Utf8Error> {
178 core::str::from_utf8(self.as_bytes())
179 }
180
181 /// Unsafely convert this [`CStr`] into a [`&str`], without checking for
182 /// valid UTF-8.
183 ///
184 /// # Safety
185 ///
186 /// The contents must be valid UTF-8.
187 ///
188 /// # Examples
189 ///
190 /// ```
191 /// # use kernel::c_str;
192 /// # use kernel::str::CStr;
193 /// // SAFETY: String literals are guaranteed to be valid UTF-8
194 /// // by the Rust compiler.
195 /// let bar = c_str!("ツ");
196 /// assert_eq!(unsafe { bar.as_str_unchecked() }, "ツ");
197 /// ```
198 #[inline]
199 pub unsafe fn as_str_unchecked(&self) -> &str {
200 unsafe { core::str::from_utf8_unchecked(self.as_bytes()) }
201 }
202}
203
204impl fmt::Display for CStr {
205 /// Formats printable ASCII characters, escaping the rest.
206 ///
207 /// ```
208 /// # use kernel::c_str;
209 /// # use kernel::str::CStr;
210 /// # use kernel::str::CString;
211 /// let penguin = c_str!("🐧");
212 /// let s = CString::try_from_fmt(fmt!("{}", penguin)).unwrap();
213 /// assert_eq!(s.as_bytes_with_nul(), "\\xf0\\x9f\\x90\\xa7\0".as_bytes());
214 ///
215 /// let ascii = c_str!("so \"cool\"");
216 /// let s = CString::try_from_fmt(fmt!("{}", ascii)).unwrap();
217 /// assert_eq!(s.as_bytes_with_nul(), "so \"cool\"\0".as_bytes());
218 /// ```
219 fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
220 for &c in self.as_bytes() {
221 if (0x20..0x7f).contains(&c) {
222 // Printable character.
223 f.write_char(c as char)?;
224 } else {
225 write!(f, "\\x{:02x}", c)?;
226 }
227 }
228 Ok(())
229 }
230}
231
232impl fmt::Debug for CStr {
233 /// Formats printable ASCII characters with a double quote on either end, escaping the rest.
234 ///
235 /// ```
236 /// # use kernel::c_str;
237 /// # use kernel::str::CStr;
238 /// # use kernel::str::CString;
239 /// let penguin = c_str!("🐧");
240 /// let s = CString::try_from_fmt(fmt!("{:?}", penguin)).unwrap();
241 /// assert_eq!(s.as_bytes_with_nul(), "\"\\xf0\\x9f\\x90\\xa7\"\0".as_bytes());
242 ///
243 /// // Embedded double quotes are escaped.
244 /// let ascii = c_str!("so \"cool\"");
245 /// let s = CString::try_from_fmt(fmt!("{:?}", ascii)).unwrap();
246 /// assert_eq!(s.as_bytes_with_nul(), "\"so \\\"cool\\\"\"\0".as_bytes());
247 /// ```
248 fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
249 f.write_str("\"")?;
250 for &c in self.as_bytes() {
251 match c {
252 // Printable characters.
253 b'\"' => f.write_str("\\\"")?,
254 0x20..=0x7e => f.write_char(c as char)?,
255 _ => write!(f, "\\x{:02x}", c)?,
256 }
257 }
258 f.write_str("\"")
259 }
260}
261
262impl AsRef<BStr> for CStr {
263 #[inline]
264 fn as_ref(&self) -> &BStr {
265 self.as_bytes()
266 }
267}
268
269impl Deref for CStr {
270 type Target = BStr;
271
272 #[inline]
273 fn deref(&self) -> &Self::Target {
274 self.as_bytes()
275 }
276}
277
278impl Index<ops::RangeFrom<usize>> for CStr {
279 type Output = CStr;
280
281 #[inline]
282 fn index(&self, index: ops::RangeFrom<usize>) -> &Self::Output {
283 // Delegate bounds checking to slice.
284 // Assign to _ to mute clippy's unnecessary operation warning.
285 let _ = &self.as_bytes()[index.start..];
286 // SAFETY: We just checked the bounds.
287 unsafe { Self::from_bytes_with_nul_unchecked(&self.0[index.start..]) }
288 }
289}
290
291impl Index<ops::RangeFull> for CStr {
292 type Output = CStr;
293
294 #[inline]
295 fn index(&self, _index: ops::RangeFull) -> &Self::Output {
296 self
297 }
298}
299
300mod private {
301 use core::ops;
302
303 // Marker trait for index types that can be forward to `BStr`.
304 pub trait CStrIndex {}
305
306 impl CStrIndex for usize {}
307 impl CStrIndex for ops::Range<usize> {}
308 impl CStrIndex for ops::RangeInclusive<usize> {}
309 impl CStrIndex for ops::RangeToInclusive<usize> {}
310}
311
312impl<Idx> Index<Idx> for CStr
313where
314 Idx: private::CStrIndex,
315 BStr: Index<Idx>,
316{
317 type Output = <BStr as Index<Idx>>::Output;
318
319 #[inline]
320 fn index(&self, index: Idx) -> &Self::Output {
321 &self.as_bytes()[index]
322 }
323}
324
325/// Creates a new [`CStr`] from a string literal.
326///
327/// The string literal should not contain any `NUL` bytes.
328///
329/// # Examples
330///
331/// ```
332/// # use kernel::c_str;
333/// # use kernel::str::CStr;
334/// const MY_CSTR: &CStr = c_str!("My awesome CStr!");
335/// ```
336#[macro_export]
337macro_rules! c_str {
338 ($str:expr) => {{
339 const S: &str = concat!($str, "\0");
340 const C: &$crate::str::CStr = match $crate::str::CStr::from_bytes_with_nul(S.as_bytes()) {
341 Ok(v) => v,
342 Err(_) => panic!("string contains interior NUL"),
343 };
344 C
345 }};
346}
347
348#[cfg(test)]
349mod tests {
350 use super::*;
351
352 #[test]
353 fn test_cstr_to_str() {
354 let good_bytes = b"\xf0\x9f\xa6\x80\0";
355 let checked_cstr = CStr::from_bytes_with_nul(good_bytes).unwrap();
356 let checked_str = checked_cstr.to_str().unwrap();
357 assert_eq!(checked_str, "🦀");
358 }
359
360 #[test]
361 #[should_panic]
362 fn test_cstr_to_str_panic() {
363 let bad_bytes = b"\xc3\x28\0";
364 let checked_cstr = CStr::from_bytes_with_nul(bad_bytes).unwrap();
365 checked_cstr.to_str().unwrap();
366 }
367
368 #[test]
369 fn test_cstr_as_str_unchecked() {
370 let good_bytes = b"\xf0\x9f\x90\xA7\0";
371 let checked_cstr = CStr::from_bytes_with_nul(good_bytes).unwrap();
372 let unchecked_str = unsafe { checked_cstr.as_str_unchecked() };
373 assert_eq!(unchecked_str, "🐧");
374 }
375}
376
377/// Allows formatting of [`fmt::Arguments`] into a raw buffer.
378///
379/// It does not fail if callers write past the end of the buffer so that they can calculate the
380/// size required to fit everything.
381///
382/// # Invariants
383///
384/// The memory region between `pos` (inclusive) and `end` (exclusive) is valid for writes if `pos`
385/// is less than `end`.
386pub(crate) struct RawFormatter {
387 // Use `usize` to use `saturating_*` functions.
388 beg: usize,
389 pos: usize,
390 end: usize,
391}
392
393impl RawFormatter {
394 /// Creates a new instance of [`RawFormatter`] with an empty buffer.
395 fn new() -> Self {
396 // INVARIANT: The buffer is empty, so the region that needs to be writable is empty.
397 Self {
398 beg: 0,
399 pos: 0,
400 end: 0,
401 }
402 }
403
404 /// Creates a new instance of [`RawFormatter`] with the given buffer pointers.
405 ///
406 /// # Safety
407 ///
408 /// If `pos` is less than `end`, then the region between `pos` (inclusive) and `end`
409 /// (exclusive) must be valid for writes for the lifetime of the returned [`RawFormatter`].
410 pub(crate) unsafe fn from_ptrs(pos: *mut u8, end: *mut u8) -> Self {
411 // INVARIANT: The safety requirements guarantee the type invariants.
412 Self {
413 beg: pos as _,
414 pos: pos as _,
415 end: end as _,
416 }
417 }
418
419 /// Creates a new instance of [`RawFormatter`] with the given buffer.
420 ///
421 /// # Safety
422 ///
423 /// The memory region starting at `buf` and extending for `len` bytes must be valid for writes
424 /// for the lifetime of the returned [`RawFormatter`].
425 pub(crate) unsafe fn from_buffer(buf: *mut u8, len: usize) -> Self {
426 let pos = buf as usize;
427 // INVARIANT: We ensure that `end` is never less then `buf`, and the safety requirements
428 // guarantees that the memory region is valid for writes.
429 Self {
430 pos,
431 beg: pos,
432 end: pos.saturating_add(len),
433 }
434 }
435
436 /// Returns the current insert position.
437 ///
438 /// N.B. It may point to invalid memory.
439 pub(crate) fn pos(&self) -> *mut u8 {
440 self.pos as _
441 }
442
443 /// Return the number of bytes written to the formatter.
444 pub(crate) fn bytes_written(&self) -> usize {
445 self.pos - self.beg
446 }
447}
448
449impl fmt::Write for RawFormatter {
450 fn write_str(&mut self, s: &str) -> fmt::Result {
451 // `pos` value after writing `len` bytes. This does not have to be bounded by `end`, but we
452 // don't want it to wrap around to 0.
453 let pos_new = self.pos.saturating_add(s.len());
454
455 // Amount that we can copy. `saturating_sub` ensures we get 0 if `pos` goes past `end`.
456 let len_to_copy = core::cmp::min(pos_new, self.end).saturating_sub(self.pos);
457
458 if len_to_copy > 0 {
459 // SAFETY: If `len_to_copy` is non-zero, then we know `pos` has not gone past `end`
460 // yet, so it is valid for write per the type invariants.
461 unsafe {
462 core::ptr::copy_nonoverlapping(
463 s.as_bytes().as_ptr(),
464 self.pos as *mut u8,
465 len_to_copy,
466 )
467 };
468 }
469
470 self.pos = pos_new;
471 Ok(())
472 }
473}
474
475/// Allows formatting of [`fmt::Arguments`] into a raw buffer.
476///
477/// Fails if callers attempt to write more than will fit in the buffer.
478pub(crate) struct Formatter(RawFormatter);
479
480impl Formatter {
481 /// Creates a new instance of [`Formatter`] with the given buffer.
482 ///
483 /// # Safety
484 ///
485 /// The memory region starting at `buf` and extending for `len` bytes must be valid for writes
486 /// for the lifetime of the returned [`Formatter`].
487 pub(crate) unsafe fn from_buffer(buf: *mut u8, len: usize) -> Self {
488 // SAFETY: The safety requirements of this function satisfy those of the callee.
489 Self(unsafe { RawFormatter::from_buffer(buf, len) })
490 }
491}
492
493impl Deref for Formatter {
494 type Target = RawFormatter;
495
496 fn deref(&self) -> &Self::Target {
497 &self.0
498 }
499}
500
501impl fmt::Write for Formatter {
502 fn write_str(&mut self, s: &str) -> fmt::Result {
503 self.0.write_str(s)?;
504
505 // Fail the request if we go past the end of the buffer.
506 if self.0.pos > self.0.end {
507 Err(fmt::Error)
508 } else {
509 Ok(())
510 }
511 }
512}
513
514/// An owned string that is guaranteed to have exactly one `NUL` byte, which is at the end.
515///
516/// Used for interoperability with kernel APIs that take C strings.
517///
518/// # Invariants
519///
520/// The string is always `NUL`-terminated and contains no other `NUL` bytes.
521///
522/// # Examples
523///
524/// ```
525/// use kernel::str::CString;
526///
527/// let s = CString::try_from_fmt(fmt!("{}{}{}", "abc", 10, 20)).unwrap();
528/// assert_eq!(s.as_bytes_with_nul(), "abc1020\0".as_bytes());
529///
530/// let tmp = "testing";
531/// let s = CString::try_from_fmt(fmt!("{tmp}{}", 123)).unwrap();
532/// assert_eq!(s.as_bytes_with_nul(), "testing123\0".as_bytes());
533///
534/// // This fails because it has an embedded `NUL` byte.
535/// let s = CString::try_from_fmt(fmt!("a\0b{}", 123));
536/// assert_eq!(s.is_ok(), false);
537/// ```
538pub struct CString {
539 buf: Vec<u8>,
540}
541
542impl CString {
543 /// Creates an instance of [`CString`] from the given formatted arguments.
544 pub fn try_from_fmt(args: fmt::Arguments<'_>) -> Result<Self, Error> {
545 // Calculate the size needed (formatted string plus `NUL` terminator).
546 let mut f = RawFormatter::new();
547 f.write_fmt(args)?;
548 f.write_str("\0")?;
549 let size = f.bytes_written();
550
551 // Allocate a vector with the required number of bytes, and write to it.
552 let mut buf = Vec::try_with_capacity(size)?;
553 // SAFETY: The buffer stored in `buf` is at least of size `size` and is valid for writes.
554 let mut f = unsafe { Formatter::from_buffer(buf.as_mut_ptr(), size) };
555 f.write_fmt(args)?;
556 f.write_str("\0")?;
557
558 // SAFETY: The number of bytes that can be written to `f` is bounded by `size`, which is
559 // `buf`'s capacity. The contents of the buffer have been initialised by writes to `f`.
560 unsafe { buf.set_len(f.bytes_written()) };
561
562 // Check that there are no `NUL` bytes before the end.
563 // SAFETY: The buffer is valid for read because `f.bytes_written()` is bounded by `size`
564 // (which the minimum buffer size) and is non-zero (we wrote at least the `NUL` terminator)
565 // so `f.bytes_written() - 1` doesn't underflow.
566 let ptr = unsafe { bindings::memchr(buf.as_ptr().cast(), 0, (f.bytes_written() - 1) as _) };
567 if !ptr.is_null() {
568 return Err(EINVAL);
569 }
570
571 // INVARIANT: We wrote the `NUL` terminator and checked above that no other `NUL` bytes
572 // exist in the buffer.
573 Ok(Self { buf })
574 }
575}
576
577impl Deref for CString {
578 type Target = CStr;
579
580 fn deref(&self) -> &Self::Target {
581 // SAFETY: The type invariants guarantee that the string is `NUL`-terminated and that no
582 // other `NUL` bytes exist.
583 unsafe { CStr::from_bytes_with_nul_unchecked(self.buf.as_slice()) }
584 }
585}
586
587/// A convenience alias for [`core::format_args`].
588#[macro_export]
589macro_rules! fmt {
590 ($($f:tt)*) => ( core::format_args!($($f)*) )
591}