Loading...
1// SPDX-License-Identifier: GPL-2.0
2
3//! String representations.
4
5use alloc::alloc::AllocError;
6use alloc::vec::Vec;
7use core::fmt::{self, Write};
8use core::ops::{self, Deref, Index};
9
10use crate::{
11 bindings,
12 error::{code::*, Error},
13};
14
15/// Byte string without UTF-8 validity guarantee.
16#[repr(transparent)]
17pub struct BStr([u8]);
18
19impl BStr {
20 /// Returns the length of this string.
21 #[inline]
22 pub const fn len(&self) -> usize {
23 self.0.len()
24 }
25
26 /// Returns `true` if the string is empty.
27 #[inline]
28 pub const fn is_empty(&self) -> bool {
29 self.len() == 0
30 }
31
32 /// Creates a [`BStr`] from a `[u8]`.
33 #[inline]
34 pub const fn from_bytes(bytes: &[u8]) -> &Self {
35 // SAFETY: `BStr` is transparent to `[u8]`.
36 unsafe { &*(bytes as *const [u8] as *const BStr) }
37 }
38}
39
40impl fmt::Display for BStr {
41 /// Formats printable ASCII characters, escaping the rest.
42 ///
43 /// ```
44 /// # use kernel::{fmt, b_str, str::{BStr, CString}};
45 /// let ascii = b_str!("Hello, BStr!");
46 /// let s = CString::try_from_fmt(fmt!("{}", ascii)).unwrap();
47 /// assert_eq!(s.as_bytes(), "Hello, BStr!".as_bytes());
48 ///
49 /// let non_ascii = b_str!("🦀");
50 /// let s = CString::try_from_fmt(fmt!("{}", non_ascii)).unwrap();
51 /// assert_eq!(s.as_bytes(), "\\xf0\\x9f\\xa6\\x80".as_bytes());
52 /// ```
53 fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
54 for &b in &self.0 {
55 match b {
56 // Common escape codes.
57 b'\t' => f.write_str("\\t")?,
58 b'\n' => f.write_str("\\n")?,
59 b'\r' => f.write_str("\\r")?,
60 // Printable characters.
61 0x20..=0x7e => f.write_char(b as char)?,
62 _ => write!(f, "\\x{:02x}", b)?,
63 }
64 }
65 Ok(())
66 }
67}
68
69impl fmt::Debug for BStr {
70 /// Formats printable ASCII characters with a double quote on either end,
71 /// escaping the rest.
72 ///
73 /// ```
74 /// # use kernel::{fmt, b_str, str::{BStr, CString}};
75 /// // Embedded double quotes are escaped.
76 /// let ascii = b_str!("Hello, \"BStr\"!");
77 /// let s = CString::try_from_fmt(fmt!("{:?}", ascii)).unwrap();
78 /// assert_eq!(s.as_bytes(), "\"Hello, \\\"BStr\\\"!\"".as_bytes());
79 ///
80 /// let non_ascii = b_str!("😺");
81 /// let s = CString::try_from_fmt(fmt!("{:?}", non_ascii)).unwrap();
82 /// assert_eq!(s.as_bytes(), "\"\\xf0\\x9f\\x98\\xba\"".as_bytes());
83 /// ```
84 fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
85 f.write_char('"')?;
86 for &b in &self.0 {
87 match b {
88 // Common escape codes.
89 b'\t' => f.write_str("\\t")?,
90 b'\n' => f.write_str("\\n")?,
91 b'\r' => f.write_str("\\r")?,
92 // String escape characters.
93 b'\"' => f.write_str("\\\"")?,
94 b'\\' => f.write_str("\\\\")?,
95 // Printable characters.
96 0x20..=0x7e => f.write_char(b as char)?,
97 _ => write!(f, "\\x{:02x}", b)?,
98 }
99 }
100 f.write_char('"')
101 }
102}
103
104impl Deref for BStr {
105 type Target = [u8];
106
107 #[inline]
108 fn deref(&self) -> &Self::Target {
109 &self.0
110 }
111}
112
113/// Creates a new [`BStr`] from a string literal.
114///
115/// `b_str!` converts the supplied string literal to byte string, so non-ASCII
116/// characters can be included.
117///
118/// # Examples
119///
120/// ```
121/// # use kernel::b_str;
122/// # use kernel::str::BStr;
123/// const MY_BSTR: &BStr = b_str!("My awesome BStr!");
124/// ```
125#[macro_export]
126macro_rules! b_str {
127 ($str:literal) => {{
128 const S: &'static str = $str;
129 const C: &'static $crate::str::BStr = $crate::str::BStr::from_bytes(S.as_bytes());
130 C
131 }};
132}
133
134/// Possible errors when using conversion functions in [`CStr`].
135#[derive(Debug, Clone, Copy)]
136pub enum CStrConvertError {
137 /// Supplied bytes contain an interior `NUL`.
138 InteriorNul,
139
140 /// Supplied bytes are not terminated by `NUL`.
141 NotNulTerminated,
142}
143
144impl From<CStrConvertError> for Error {
145 #[inline]
146 fn from(_: CStrConvertError) -> Error {
147 EINVAL
148 }
149}
150
151/// A string that is guaranteed to have exactly one `NUL` byte, which is at the
152/// end.
153///
154/// Used for interoperability with kernel APIs that take C strings.
155#[repr(transparent)]
156pub struct CStr([u8]);
157
158impl CStr {
159 /// Returns the length of this string excluding `NUL`.
160 #[inline]
161 pub const fn len(&self) -> usize {
162 self.len_with_nul() - 1
163 }
164
165 /// Returns the length of this string with `NUL`.
166 #[inline]
167 pub const fn len_with_nul(&self) -> usize {
168 // SAFETY: This is one of the invariant of `CStr`.
169 // We add a `unreachable_unchecked` here to hint the optimizer that
170 // the value returned from this function is non-zero.
171 if self.0.is_empty() {
172 unsafe { core::hint::unreachable_unchecked() };
173 }
174 self.0.len()
175 }
176
177 /// Returns `true` if the string only includes `NUL`.
178 #[inline]
179 pub const fn is_empty(&self) -> bool {
180 self.len() == 0
181 }
182
183 /// Wraps a raw C string pointer.
184 ///
185 /// # Safety
186 ///
187 /// `ptr` must be a valid pointer to a `NUL`-terminated C string, and it must
188 /// last at least `'a`. When `CStr` is alive, the memory pointed by `ptr`
189 /// must not be mutated.
190 #[inline]
191 pub unsafe fn from_char_ptr<'a>(ptr: *const core::ffi::c_char) -> &'a Self {
192 // SAFETY: The safety precondition guarantees `ptr` is a valid pointer
193 // to a `NUL`-terminated C string.
194 let len = unsafe { bindings::strlen(ptr) } + 1;
195 // SAFETY: Lifetime guaranteed by the safety precondition.
196 let bytes = unsafe { core::slice::from_raw_parts(ptr as _, len as _) };
197 // SAFETY: As `len` is returned by `strlen`, `bytes` does not contain interior `NUL`.
198 // As we have added 1 to `len`, the last byte is known to be `NUL`.
199 unsafe { Self::from_bytes_with_nul_unchecked(bytes) }
200 }
201
202 /// Creates a [`CStr`] from a `[u8]`.
203 ///
204 /// The provided slice must be `NUL`-terminated, does not contain any
205 /// interior `NUL` bytes.
206 pub const fn from_bytes_with_nul(bytes: &[u8]) -> Result<&Self, CStrConvertError> {
207 if bytes.is_empty() {
208 return Err(CStrConvertError::NotNulTerminated);
209 }
210 if bytes[bytes.len() - 1] != 0 {
211 return Err(CStrConvertError::NotNulTerminated);
212 }
213 let mut i = 0;
214 // `i + 1 < bytes.len()` allows LLVM to optimize away bounds checking,
215 // while it couldn't optimize away bounds checks for `i < bytes.len() - 1`.
216 while i + 1 < bytes.len() {
217 if bytes[i] == 0 {
218 return Err(CStrConvertError::InteriorNul);
219 }
220 i += 1;
221 }
222 // SAFETY: We just checked that all properties hold.
223 Ok(unsafe { Self::from_bytes_with_nul_unchecked(bytes) })
224 }
225
226 /// Creates a [`CStr`] from a `[u8]` without performing any additional
227 /// checks.
228 ///
229 /// # Safety
230 ///
231 /// `bytes` *must* end with a `NUL` byte, and should only have a single
232 /// `NUL` byte (or the string will be truncated).
233 #[inline]
234 pub const unsafe fn from_bytes_with_nul_unchecked(bytes: &[u8]) -> &CStr {
235 // SAFETY: Properties of `bytes` guaranteed by the safety precondition.
236 unsafe { core::mem::transmute(bytes) }
237 }
238
239 /// Returns a C pointer to the string.
240 #[inline]
241 pub const fn as_char_ptr(&self) -> *const core::ffi::c_char {
242 self.0.as_ptr() as _
243 }
244
245 /// Convert the string to a byte slice without the trailing `NUL` byte.
246 #[inline]
247 pub fn as_bytes(&self) -> &[u8] {
248 &self.0[..self.len()]
249 }
250
251 /// Convert the string to a byte slice containing the trailing `NUL` byte.
252 #[inline]
253 pub const fn as_bytes_with_nul(&self) -> &[u8] {
254 &self.0
255 }
256
257 /// Yields a [`&str`] slice if the [`CStr`] contains valid UTF-8.
258 ///
259 /// If the contents of the [`CStr`] are valid UTF-8 data, this
260 /// function will return the corresponding [`&str`] slice. Otherwise,
261 /// it will return an error with details of where UTF-8 validation failed.
262 ///
263 /// # Examples
264 ///
265 /// ```
266 /// # use kernel::str::CStr;
267 /// let cstr = CStr::from_bytes_with_nul(b"foo\0").unwrap();
268 /// assert_eq!(cstr.to_str(), Ok("foo"));
269 /// ```
270 #[inline]
271 pub fn to_str(&self) -> Result<&str, core::str::Utf8Error> {
272 core::str::from_utf8(self.as_bytes())
273 }
274
275 /// Unsafely convert this [`CStr`] into a [`&str`], without checking for
276 /// valid UTF-8.
277 ///
278 /// # Safety
279 ///
280 /// The contents must be valid UTF-8.
281 ///
282 /// # Examples
283 ///
284 /// ```
285 /// # use kernel::c_str;
286 /// # use kernel::str::CStr;
287 /// let bar = c_str!("ツ");
288 /// // SAFETY: String literals are guaranteed to be valid UTF-8
289 /// // by the Rust compiler.
290 /// assert_eq!(unsafe { bar.as_str_unchecked() }, "ツ");
291 /// ```
292 #[inline]
293 pub unsafe fn as_str_unchecked(&self) -> &str {
294 unsafe { core::str::from_utf8_unchecked(self.as_bytes()) }
295 }
296
297 /// Convert this [`CStr`] into a [`CString`] by allocating memory and
298 /// copying over the string data.
299 pub fn to_cstring(&self) -> Result<CString, AllocError> {
300 CString::try_from(self)
301 }
302}
303
304impl fmt::Display for CStr {
305 /// Formats printable ASCII characters, escaping the rest.
306 ///
307 /// ```
308 /// # use kernel::c_str;
309 /// # use kernel::fmt;
310 /// # use kernel::str::CStr;
311 /// # use kernel::str::CString;
312 /// let penguin = c_str!("🐧");
313 /// let s = CString::try_from_fmt(fmt!("{}", penguin)).unwrap();
314 /// assert_eq!(s.as_bytes_with_nul(), "\\xf0\\x9f\\x90\\xa7\0".as_bytes());
315 ///
316 /// let ascii = c_str!("so \"cool\"");
317 /// let s = CString::try_from_fmt(fmt!("{}", ascii)).unwrap();
318 /// assert_eq!(s.as_bytes_with_nul(), "so \"cool\"\0".as_bytes());
319 /// ```
320 fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
321 for &c in self.as_bytes() {
322 if (0x20..0x7f).contains(&c) {
323 // Printable character.
324 f.write_char(c as char)?;
325 } else {
326 write!(f, "\\x{:02x}", c)?;
327 }
328 }
329 Ok(())
330 }
331}
332
333impl fmt::Debug for CStr {
334 /// Formats printable ASCII characters with a double quote on either end, escaping the rest.
335 ///
336 /// ```
337 /// # use kernel::c_str;
338 /// # use kernel::fmt;
339 /// # use kernel::str::CStr;
340 /// # use kernel::str::CString;
341 /// let penguin = c_str!("🐧");
342 /// let s = CString::try_from_fmt(fmt!("{:?}", penguin)).unwrap();
343 /// assert_eq!(s.as_bytes_with_nul(), "\"\\xf0\\x9f\\x90\\xa7\"\0".as_bytes());
344 ///
345 /// // Embedded double quotes are escaped.
346 /// let ascii = c_str!("so \"cool\"");
347 /// let s = CString::try_from_fmt(fmt!("{:?}", ascii)).unwrap();
348 /// assert_eq!(s.as_bytes_with_nul(), "\"so \\\"cool\\\"\"\0".as_bytes());
349 /// ```
350 fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
351 f.write_str("\"")?;
352 for &c in self.as_bytes() {
353 match c {
354 // Printable characters.
355 b'\"' => f.write_str("\\\"")?,
356 0x20..=0x7e => f.write_char(c as char)?,
357 _ => write!(f, "\\x{:02x}", c)?,
358 }
359 }
360 f.write_str("\"")
361 }
362}
363
364impl AsRef<BStr> for CStr {
365 #[inline]
366 fn as_ref(&self) -> &BStr {
367 BStr::from_bytes(self.as_bytes())
368 }
369}
370
371impl Deref for CStr {
372 type Target = BStr;
373
374 #[inline]
375 fn deref(&self) -> &Self::Target {
376 self.as_ref()
377 }
378}
379
380impl Index<ops::RangeFrom<usize>> for CStr {
381 type Output = CStr;
382
383 #[inline]
384 fn index(&self, index: ops::RangeFrom<usize>) -> &Self::Output {
385 // Delegate bounds checking to slice.
386 // Assign to _ to mute clippy's unnecessary operation warning.
387 let _ = &self.as_bytes()[index.start..];
388 // SAFETY: We just checked the bounds.
389 unsafe { Self::from_bytes_with_nul_unchecked(&self.0[index.start..]) }
390 }
391}
392
393impl Index<ops::RangeFull> for CStr {
394 type Output = CStr;
395
396 #[inline]
397 fn index(&self, _index: ops::RangeFull) -> &Self::Output {
398 self
399 }
400}
401
402mod private {
403 use core::ops;
404
405 // Marker trait for index types that can be forward to `BStr`.
406 pub trait CStrIndex {}
407
408 impl CStrIndex for usize {}
409 impl CStrIndex for ops::Range<usize> {}
410 impl CStrIndex for ops::RangeInclusive<usize> {}
411 impl CStrIndex for ops::RangeToInclusive<usize> {}
412}
413
414impl<Idx> Index<Idx> for CStr
415where
416 Idx: private::CStrIndex,
417 BStr: Index<Idx>,
418{
419 type Output = <BStr as Index<Idx>>::Output;
420
421 #[inline]
422 fn index(&self, index: Idx) -> &Self::Output {
423 &self.as_ref()[index]
424 }
425}
426
427/// Creates a new [`CStr`] from a string literal.
428///
429/// The string literal should not contain any `NUL` bytes.
430///
431/// # Examples
432///
433/// ```
434/// # use kernel::c_str;
435/// # use kernel::str::CStr;
436/// const MY_CSTR: &CStr = c_str!("My awesome CStr!");
437/// ```
438#[macro_export]
439macro_rules! c_str {
440 ($str:expr) => {{
441 const S: &str = concat!($str, "\0");
442 const C: &$crate::str::CStr = match $crate::str::CStr::from_bytes_with_nul(S.as_bytes()) {
443 Ok(v) => v,
444 Err(_) => panic!("string contains interior NUL"),
445 };
446 C
447 }};
448}
449
450#[cfg(test)]
451mod tests {
452 use super::*;
453 use alloc::format;
454
455 const ALL_ASCII_CHARS: &'static str =
456 "\\x01\\x02\\x03\\x04\\x05\\x06\\x07\\x08\\x09\\x0a\\x0b\\x0c\\x0d\\x0e\\x0f\
457 \\x10\\x11\\x12\\x13\\x14\\x15\\x16\\x17\\x18\\x19\\x1a\\x1b\\x1c\\x1d\\x1e\\x1f \
458 !\"#$%&'()*+,-./0123456789:;<=>?@\
459 ABCDEFGHIJKLMNOPQRSTUVWXYZ[\\]^_`abcdefghijklmnopqrstuvwxyz{|}~\\x7f\
460 \\x80\\x81\\x82\\x83\\x84\\x85\\x86\\x87\\x88\\x89\\x8a\\x8b\\x8c\\x8d\\x8e\\x8f\
461 \\x90\\x91\\x92\\x93\\x94\\x95\\x96\\x97\\x98\\x99\\x9a\\x9b\\x9c\\x9d\\x9e\\x9f\
462 \\xa0\\xa1\\xa2\\xa3\\xa4\\xa5\\xa6\\xa7\\xa8\\xa9\\xaa\\xab\\xac\\xad\\xae\\xaf\
463 \\xb0\\xb1\\xb2\\xb3\\xb4\\xb5\\xb6\\xb7\\xb8\\xb9\\xba\\xbb\\xbc\\xbd\\xbe\\xbf\
464 \\xc0\\xc1\\xc2\\xc3\\xc4\\xc5\\xc6\\xc7\\xc8\\xc9\\xca\\xcb\\xcc\\xcd\\xce\\xcf\
465 \\xd0\\xd1\\xd2\\xd3\\xd4\\xd5\\xd6\\xd7\\xd8\\xd9\\xda\\xdb\\xdc\\xdd\\xde\\xdf\
466 \\xe0\\xe1\\xe2\\xe3\\xe4\\xe5\\xe6\\xe7\\xe8\\xe9\\xea\\xeb\\xec\\xed\\xee\\xef\
467 \\xf0\\xf1\\xf2\\xf3\\xf4\\xf5\\xf6\\xf7\\xf8\\xf9\\xfa\\xfb\\xfc\\xfd\\xfe\\xff";
468
469 #[test]
470 fn test_cstr_to_str() {
471 let good_bytes = b"\xf0\x9f\xa6\x80\0";
472 let checked_cstr = CStr::from_bytes_with_nul(good_bytes).unwrap();
473 let checked_str = checked_cstr.to_str().unwrap();
474 assert_eq!(checked_str, "🦀");
475 }
476
477 #[test]
478 #[should_panic]
479 fn test_cstr_to_str_panic() {
480 let bad_bytes = b"\xc3\x28\0";
481 let checked_cstr = CStr::from_bytes_with_nul(bad_bytes).unwrap();
482 checked_cstr.to_str().unwrap();
483 }
484
485 #[test]
486 fn test_cstr_as_str_unchecked() {
487 let good_bytes = b"\xf0\x9f\x90\xA7\0";
488 let checked_cstr = CStr::from_bytes_with_nul(good_bytes).unwrap();
489 let unchecked_str = unsafe { checked_cstr.as_str_unchecked() };
490 assert_eq!(unchecked_str, "🐧");
491 }
492
493 #[test]
494 fn test_cstr_display() {
495 let hello_world = CStr::from_bytes_with_nul(b"hello, world!\0").unwrap();
496 assert_eq!(format!("{}", hello_world), "hello, world!");
497 let non_printables = CStr::from_bytes_with_nul(b"\x01\x09\x0a\0").unwrap();
498 assert_eq!(format!("{}", non_printables), "\\x01\\x09\\x0a");
499 let non_ascii = CStr::from_bytes_with_nul(b"d\xe9j\xe0 vu\0").unwrap();
500 assert_eq!(format!("{}", non_ascii), "d\\xe9j\\xe0 vu");
501 let good_bytes = CStr::from_bytes_with_nul(b"\xf0\x9f\xa6\x80\0").unwrap();
502 assert_eq!(format!("{}", good_bytes), "\\xf0\\x9f\\xa6\\x80");
503 }
504
505 #[test]
506 fn test_cstr_display_all_bytes() {
507 let mut bytes: [u8; 256] = [0; 256];
508 // fill `bytes` with [1..=255] + [0]
509 for i in u8::MIN..=u8::MAX {
510 bytes[i as usize] = i.wrapping_add(1);
511 }
512 let cstr = CStr::from_bytes_with_nul(&bytes).unwrap();
513 assert_eq!(format!("{}", cstr), ALL_ASCII_CHARS);
514 }
515
516 #[test]
517 fn test_cstr_debug() {
518 let hello_world = CStr::from_bytes_with_nul(b"hello, world!\0").unwrap();
519 assert_eq!(format!("{:?}", hello_world), "\"hello, world!\"");
520 let non_printables = CStr::from_bytes_with_nul(b"\x01\x09\x0a\0").unwrap();
521 assert_eq!(format!("{:?}", non_printables), "\"\\x01\\x09\\x0a\"");
522 let non_ascii = CStr::from_bytes_with_nul(b"d\xe9j\xe0 vu\0").unwrap();
523 assert_eq!(format!("{:?}", non_ascii), "\"d\\xe9j\\xe0 vu\"");
524 let good_bytes = CStr::from_bytes_with_nul(b"\xf0\x9f\xa6\x80\0").unwrap();
525 assert_eq!(format!("{:?}", good_bytes), "\"\\xf0\\x9f\\xa6\\x80\"");
526 }
527
528 #[test]
529 fn test_bstr_display() {
530 let hello_world = BStr::from_bytes(b"hello, world!");
531 assert_eq!(format!("{}", hello_world), "hello, world!");
532 let escapes = BStr::from_bytes(b"_\t_\n_\r_\\_\'_\"_");
533 assert_eq!(format!("{}", escapes), "_\\t_\\n_\\r_\\_'_\"_");
534 let others = BStr::from_bytes(b"\x01");
535 assert_eq!(format!("{}", others), "\\x01");
536 let non_ascii = BStr::from_bytes(b"d\xe9j\xe0 vu");
537 assert_eq!(format!("{}", non_ascii), "d\\xe9j\\xe0 vu");
538 let good_bytes = BStr::from_bytes(b"\xf0\x9f\xa6\x80");
539 assert_eq!(format!("{}", good_bytes), "\\xf0\\x9f\\xa6\\x80");
540 }
541
542 #[test]
543 fn test_bstr_debug() {
544 let hello_world = BStr::from_bytes(b"hello, world!");
545 assert_eq!(format!("{:?}", hello_world), "\"hello, world!\"");
546 let escapes = BStr::from_bytes(b"_\t_\n_\r_\\_\'_\"_");
547 assert_eq!(format!("{:?}", escapes), "\"_\\t_\\n_\\r_\\\\_'_\\\"_\"");
548 let others = BStr::from_bytes(b"\x01");
549 assert_eq!(format!("{:?}", others), "\"\\x01\"");
550 let non_ascii = BStr::from_bytes(b"d\xe9j\xe0 vu");
551 assert_eq!(format!("{:?}", non_ascii), "\"d\\xe9j\\xe0 vu\"");
552 let good_bytes = BStr::from_bytes(b"\xf0\x9f\xa6\x80");
553 assert_eq!(format!("{:?}", good_bytes), "\"\\xf0\\x9f\\xa6\\x80\"");
554 }
555}
556
557/// Allows formatting of [`fmt::Arguments`] into a raw buffer.
558///
559/// It does not fail if callers write past the end of the buffer so that they can calculate the
560/// size required to fit everything.
561///
562/// # Invariants
563///
564/// The memory region between `pos` (inclusive) and `end` (exclusive) is valid for writes if `pos`
565/// is less than `end`.
566pub(crate) struct RawFormatter {
567 // Use `usize` to use `saturating_*` functions.
568 beg: usize,
569 pos: usize,
570 end: usize,
571}
572
573impl RawFormatter {
574 /// Creates a new instance of [`RawFormatter`] with an empty buffer.
575 fn new() -> Self {
576 // INVARIANT: The buffer is empty, so the region that needs to be writable is empty.
577 Self {
578 beg: 0,
579 pos: 0,
580 end: 0,
581 }
582 }
583
584 /// Creates a new instance of [`RawFormatter`] with the given buffer pointers.
585 ///
586 /// # Safety
587 ///
588 /// If `pos` is less than `end`, then the region between `pos` (inclusive) and `end`
589 /// (exclusive) must be valid for writes for the lifetime of the returned [`RawFormatter`].
590 pub(crate) unsafe fn from_ptrs(pos: *mut u8, end: *mut u8) -> Self {
591 // INVARIANT: The safety requirements guarantee the type invariants.
592 Self {
593 beg: pos as _,
594 pos: pos as _,
595 end: end as _,
596 }
597 }
598
599 /// Creates a new instance of [`RawFormatter`] with the given buffer.
600 ///
601 /// # Safety
602 ///
603 /// The memory region starting at `buf` and extending for `len` bytes must be valid for writes
604 /// for the lifetime of the returned [`RawFormatter`].
605 pub(crate) unsafe fn from_buffer(buf: *mut u8, len: usize) -> Self {
606 let pos = buf as usize;
607 // INVARIANT: We ensure that `end` is never less then `buf`, and the safety requirements
608 // guarantees that the memory region is valid for writes.
609 Self {
610 pos,
611 beg: pos,
612 end: pos.saturating_add(len),
613 }
614 }
615
616 /// Returns the current insert position.
617 ///
618 /// N.B. It may point to invalid memory.
619 pub(crate) fn pos(&self) -> *mut u8 {
620 self.pos as _
621 }
622
623 /// Returns the number of bytes written to the formatter.
624 pub(crate) fn bytes_written(&self) -> usize {
625 self.pos - self.beg
626 }
627}
628
629impl fmt::Write for RawFormatter {
630 fn write_str(&mut self, s: &str) -> fmt::Result {
631 // `pos` value after writing `len` bytes. This does not have to be bounded by `end`, but we
632 // don't want it to wrap around to 0.
633 let pos_new = self.pos.saturating_add(s.len());
634
635 // Amount that we can copy. `saturating_sub` ensures we get 0 if `pos` goes past `end`.
636 let len_to_copy = core::cmp::min(pos_new, self.end).saturating_sub(self.pos);
637
638 if len_to_copy > 0 {
639 // SAFETY: If `len_to_copy` is non-zero, then we know `pos` has not gone past `end`
640 // yet, so it is valid for write per the type invariants.
641 unsafe {
642 core::ptr::copy_nonoverlapping(
643 s.as_bytes().as_ptr(),
644 self.pos as *mut u8,
645 len_to_copy,
646 )
647 };
648 }
649
650 self.pos = pos_new;
651 Ok(())
652 }
653}
654
655/// Allows formatting of [`fmt::Arguments`] into a raw buffer.
656///
657/// Fails if callers attempt to write more than will fit in the buffer.
658pub(crate) struct Formatter(RawFormatter);
659
660impl Formatter {
661 /// Creates a new instance of [`Formatter`] with the given buffer.
662 ///
663 /// # Safety
664 ///
665 /// The memory region starting at `buf` and extending for `len` bytes must be valid for writes
666 /// for the lifetime of the returned [`Formatter`].
667 pub(crate) unsafe fn from_buffer(buf: *mut u8, len: usize) -> Self {
668 // SAFETY: The safety requirements of this function satisfy those of the callee.
669 Self(unsafe { RawFormatter::from_buffer(buf, len) })
670 }
671}
672
673impl Deref for Formatter {
674 type Target = RawFormatter;
675
676 fn deref(&self) -> &Self::Target {
677 &self.0
678 }
679}
680
681impl fmt::Write for Formatter {
682 fn write_str(&mut self, s: &str) -> fmt::Result {
683 self.0.write_str(s)?;
684
685 // Fail the request if we go past the end of the buffer.
686 if self.0.pos > self.0.end {
687 Err(fmt::Error)
688 } else {
689 Ok(())
690 }
691 }
692}
693
694/// An owned string that is guaranteed to have exactly one `NUL` byte, which is at the end.
695///
696/// Used for interoperability with kernel APIs that take C strings.
697///
698/// # Invariants
699///
700/// The string is always `NUL`-terminated and contains no other `NUL` bytes.
701///
702/// # Examples
703///
704/// ```
705/// use kernel::{str::CString, fmt};
706///
707/// let s = CString::try_from_fmt(fmt!("{}{}{}", "abc", 10, 20)).unwrap();
708/// assert_eq!(s.as_bytes_with_nul(), "abc1020\0".as_bytes());
709///
710/// let tmp = "testing";
711/// let s = CString::try_from_fmt(fmt!("{tmp}{}", 123)).unwrap();
712/// assert_eq!(s.as_bytes_with_nul(), "testing123\0".as_bytes());
713///
714/// // This fails because it has an embedded `NUL` byte.
715/// let s = CString::try_from_fmt(fmt!("a\0b{}", 123));
716/// assert_eq!(s.is_ok(), false);
717/// ```
718pub struct CString {
719 buf: Vec<u8>,
720}
721
722impl CString {
723 /// Creates an instance of [`CString`] from the given formatted arguments.
724 pub fn try_from_fmt(args: fmt::Arguments<'_>) -> Result<Self, Error> {
725 // Calculate the size needed (formatted string plus `NUL` terminator).
726 let mut f = RawFormatter::new();
727 f.write_fmt(args)?;
728 f.write_str("\0")?;
729 let size = f.bytes_written();
730
731 // Allocate a vector with the required number of bytes, and write to it.
732 let mut buf = Vec::try_with_capacity(size)?;
733 // SAFETY: The buffer stored in `buf` is at least of size `size` and is valid for writes.
734 let mut f = unsafe { Formatter::from_buffer(buf.as_mut_ptr(), size) };
735 f.write_fmt(args)?;
736 f.write_str("\0")?;
737
738 // SAFETY: The number of bytes that can be written to `f` is bounded by `size`, which is
739 // `buf`'s capacity. The contents of the buffer have been initialised by writes to `f`.
740 unsafe { buf.set_len(f.bytes_written()) };
741
742 // Check that there are no `NUL` bytes before the end.
743 // SAFETY: The buffer is valid for read because `f.bytes_written()` is bounded by `size`
744 // (which the minimum buffer size) and is non-zero (we wrote at least the `NUL` terminator)
745 // so `f.bytes_written() - 1` doesn't underflow.
746 let ptr = unsafe { bindings::memchr(buf.as_ptr().cast(), 0, (f.bytes_written() - 1) as _) };
747 if !ptr.is_null() {
748 return Err(EINVAL);
749 }
750
751 // INVARIANT: We wrote the `NUL` terminator and checked above that no other `NUL` bytes
752 // exist in the buffer.
753 Ok(Self { buf })
754 }
755}
756
757impl Deref for CString {
758 type Target = CStr;
759
760 fn deref(&self) -> &Self::Target {
761 // SAFETY: The type invariants guarantee that the string is `NUL`-terminated and that no
762 // other `NUL` bytes exist.
763 unsafe { CStr::from_bytes_with_nul_unchecked(self.buf.as_slice()) }
764 }
765}
766
767impl<'a> TryFrom<&'a CStr> for CString {
768 type Error = AllocError;
769
770 fn try_from(cstr: &'a CStr) -> Result<CString, AllocError> {
771 let mut buf = Vec::new();
772
773 buf.try_extend_from_slice(cstr.as_bytes_with_nul())
774 .map_err(|_| AllocError)?;
775
776 // INVARIANT: The `CStr` and `CString` types have the same invariants for
777 // the string data, and we copied it over without changes.
778 Ok(CString { buf })
779 }
780}
781
782impl fmt::Debug for CString {
783 fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
784 fmt::Debug::fmt(&**self, f)
785 }
786}
787
788/// A convenience alias for [`core::format_args`].
789#[macro_export]
790macro_rules! fmt {
791 ($($f:tt)*) => ( core::format_args!($($f)*) )
792}
1// SPDX-License-Identifier: GPL-2.0
2
3//! String representations.
4
5use crate::alloc::{flags::*, AllocError, KVec};
6use core::fmt::{self, Write};
7use core::ops::{self, Deref, DerefMut, Index};
8
9use crate::error::{code::*, Error};
10
11/// Byte string without UTF-8 validity guarantee.
12#[repr(transparent)]
13pub struct BStr([u8]);
14
15impl BStr {
16 /// Returns the length of this string.
17 #[inline]
18 pub const fn len(&self) -> usize {
19 self.0.len()
20 }
21
22 /// Returns `true` if the string is empty.
23 #[inline]
24 pub const fn is_empty(&self) -> bool {
25 self.len() == 0
26 }
27
28 /// Creates a [`BStr`] from a `[u8]`.
29 #[inline]
30 pub const fn from_bytes(bytes: &[u8]) -> &Self {
31 // SAFETY: `BStr` is transparent to `[u8]`.
32 unsafe { &*(bytes as *const [u8] as *const BStr) }
33 }
34}
35
36impl fmt::Display for BStr {
37 /// Formats printable ASCII characters, escaping the rest.
38 ///
39 /// ```
40 /// # use kernel::{fmt, b_str, str::{BStr, CString}};
41 /// let ascii = b_str!("Hello, BStr!");
42 /// let s = CString::try_from_fmt(fmt!("{}", ascii)).unwrap();
43 /// assert_eq!(s.as_bytes(), "Hello, BStr!".as_bytes());
44 ///
45 /// let non_ascii = b_str!("🦀");
46 /// let s = CString::try_from_fmt(fmt!("{}", non_ascii)).unwrap();
47 /// assert_eq!(s.as_bytes(), "\\xf0\\x9f\\xa6\\x80".as_bytes());
48 /// ```
49 fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
50 for &b in &self.0 {
51 match b {
52 // Common escape codes.
53 b'\t' => f.write_str("\\t")?,
54 b'\n' => f.write_str("\\n")?,
55 b'\r' => f.write_str("\\r")?,
56 // Printable characters.
57 0x20..=0x7e => f.write_char(b as char)?,
58 _ => write!(f, "\\x{:02x}", b)?,
59 }
60 }
61 Ok(())
62 }
63}
64
65impl fmt::Debug for BStr {
66 /// Formats printable ASCII characters with a double quote on either end,
67 /// escaping the rest.
68 ///
69 /// ```
70 /// # use kernel::{fmt, b_str, str::{BStr, CString}};
71 /// // Embedded double quotes are escaped.
72 /// let ascii = b_str!("Hello, \"BStr\"!");
73 /// let s = CString::try_from_fmt(fmt!("{:?}", ascii)).unwrap();
74 /// assert_eq!(s.as_bytes(), "\"Hello, \\\"BStr\\\"!\"".as_bytes());
75 ///
76 /// let non_ascii = b_str!("😺");
77 /// let s = CString::try_from_fmt(fmt!("{:?}", non_ascii)).unwrap();
78 /// assert_eq!(s.as_bytes(), "\"\\xf0\\x9f\\x98\\xba\"".as_bytes());
79 /// ```
80 fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
81 f.write_char('"')?;
82 for &b in &self.0 {
83 match b {
84 // Common escape codes.
85 b'\t' => f.write_str("\\t")?,
86 b'\n' => f.write_str("\\n")?,
87 b'\r' => f.write_str("\\r")?,
88 // String escape characters.
89 b'\"' => f.write_str("\\\"")?,
90 b'\\' => f.write_str("\\\\")?,
91 // Printable characters.
92 0x20..=0x7e => f.write_char(b as char)?,
93 _ => write!(f, "\\x{:02x}", b)?,
94 }
95 }
96 f.write_char('"')
97 }
98}
99
100impl Deref for BStr {
101 type Target = [u8];
102
103 #[inline]
104 fn deref(&self) -> &Self::Target {
105 &self.0
106 }
107}
108
109/// Creates a new [`BStr`] from a string literal.
110///
111/// `b_str!` converts the supplied string literal to byte string, so non-ASCII
112/// characters can be included.
113///
114/// # Examples
115///
116/// ```
117/// # use kernel::b_str;
118/// # use kernel::str::BStr;
119/// const MY_BSTR: &BStr = b_str!("My awesome BStr!");
120/// ```
121#[macro_export]
122macro_rules! b_str {
123 ($str:literal) => {{
124 const S: &'static str = $str;
125 const C: &'static $crate::str::BStr = $crate::str::BStr::from_bytes(S.as_bytes());
126 C
127 }};
128}
129
130/// Possible errors when using conversion functions in [`CStr`].
131#[derive(Debug, Clone, Copy)]
132pub enum CStrConvertError {
133 /// Supplied bytes contain an interior `NUL`.
134 InteriorNul,
135
136 /// Supplied bytes are not terminated by `NUL`.
137 NotNulTerminated,
138}
139
140impl From<CStrConvertError> for Error {
141 #[inline]
142 fn from(_: CStrConvertError) -> Error {
143 EINVAL
144 }
145}
146
147/// A string that is guaranteed to have exactly one `NUL` byte, which is at the
148/// end.
149///
150/// Used for interoperability with kernel APIs that take C strings.
151#[repr(transparent)]
152pub struct CStr([u8]);
153
154impl CStr {
155 /// Returns the length of this string excluding `NUL`.
156 #[inline]
157 pub const fn len(&self) -> usize {
158 self.len_with_nul() - 1
159 }
160
161 /// Returns the length of this string with `NUL`.
162 #[inline]
163 pub const fn len_with_nul(&self) -> usize {
164 if self.0.is_empty() {
165 // SAFETY: This is one of the invariant of `CStr`.
166 // We add a `unreachable_unchecked` here to hint the optimizer that
167 // the value returned from this function is non-zero.
168 unsafe { core::hint::unreachable_unchecked() };
169 }
170 self.0.len()
171 }
172
173 /// Returns `true` if the string only includes `NUL`.
174 #[inline]
175 pub const fn is_empty(&self) -> bool {
176 self.len() == 0
177 }
178
179 /// Wraps a raw C string pointer.
180 ///
181 /// # Safety
182 ///
183 /// `ptr` must be a valid pointer to a `NUL`-terminated C string, and it must
184 /// last at least `'a`. When `CStr` is alive, the memory pointed by `ptr`
185 /// must not be mutated.
186 #[inline]
187 pub unsafe fn from_char_ptr<'a>(ptr: *const crate::ffi::c_char) -> &'a Self {
188 // SAFETY: The safety precondition guarantees `ptr` is a valid pointer
189 // to a `NUL`-terminated C string.
190 let len = unsafe { bindings::strlen(ptr) } + 1;
191 // SAFETY: Lifetime guaranteed by the safety precondition.
192 let bytes = unsafe { core::slice::from_raw_parts(ptr as _, len) };
193 // SAFETY: As `len` is returned by `strlen`, `bytes` does not contain interior `NUL`.
194 // As we have added 1 to `len`, the last byte is known to be `NUL`.
195 unsafe { Self::from_bytes_with_nul_unchecked(bytes) }
196 }
197
198 /// Creates a [`CStr`] from a `[u8]`.
199 ///
200 /// The provided slice must be `NUL`-terminated, does not contain any
201 /// interior `NUL` bytes.
202 pub const fn from_bytes_with_nul(bytes: &[u8]) -> Result<&Self, CStrConvertError> {
203 if bytes.is_empty() {
204 return Err(CStrConvertError::NotNulTerminated);
205 }
206 if bytes[bytes.len() - 1] != 0 {
207 return Err(CStrConvertError::NotNulTerminated);
208 }
209 let mut i = 0;
210 // `i + 1 < bytes.len()` allows LLVM to optimize away bounds checking,
211 // while it couldn't optimize away bounds checks for `i < bytes.len() - 1`.
212 while i + 1 < bytes.len() {
213 if bytes[i] == 0 {
214 return Err(CStrConvertError::InteriorNul);
215 }
216 i += 1;
217 }
218 // SAFETY: We just checked that all properties hold.
219 Ok(unsafe { Self::from_bytes_with_nul_unchecked(bytes) })
220 }
221
222 /// Creates a [`CStr`] from a `[u8]` without performing any additional
223 /// checks.
224 ///
225 /// # Safety
226 ///
227 /// `bytes` *must* end with a `NUL` byte, and should only have a single
228 /// `NUL` byte (or the string will be truncated).
229 #[inline]
230 pub const unsafe fn from_bytes_with_nul_unchecked(bytes: &[u8]) -> &CStr {
231 // SAFETY: Properties of `bytes` guaranteed by the safety precondition.
232 unsafe { core::mem::transmute(bytes) }
233 }
234
235 /// Creates a mutable [`CStr`] from a `[u8]` without performing any
236 /// additional checks.
237 ///
238 /// # Safety
239 ///
240 /// `bytes` *must* end with a `NUL` byte, and should only have a single
241 /// `NUL` byte (or the string will be truncated).
242 #[inline]
243 pub unsafe fn from_bytes_with_nul_unchecked_mut(bytes: &mut [u8]) -> &mut CStr {
244 // SAFETY: Properties of `bytes` guaranteed by the safety precondition.
245 unsafe { &mut *(bytes as *mut [u8] as *mut CStr) }
246 }
247
248 /// Returns a C pointer to the string.
249 #[inline]
250 pub const fn as_char_ptr(&self) -> *const crate::ffi::c_char {
251 self.0.as_ptr()
252 }
253
254 /// Convert the string to a byte slice without the trailing `NUL` byte.
255 #[inline]
256 pub fn as_bytes(&self) -> &[u8] {
257 &self.0[..self.len()]
258 }
259
260 /// Convert the string to a byte slice containing the trailing `NUL` byte.
261 #[inline]
262 pub const fn as_bytes_with_nul(&self) -> &[u8] {
263 &self.0
264 }
265
266 /// Yields a [`&str`] slice if the [`CStr`] contains valid UTF-8.
267 ///
268 /// If the contents of the [`CStr`] are valid UTF-8 data, this
269 /// function will return the corresponding [`&str`] slice. Otherwise,
270 /// it will return an error with details of where UTF-8 validation failed.
271 ///
272 /// # Examples
273 ///
274 /// ```
275 /// # use kernel::str::CStr;
276 /// let cstr = CStr::from_bytes_with_nul(b"foo\0").unwrap();
277 /// assert_eq!(cstr.to_str(), Ok("foo"));
278 /// ```
279 #[inline]
280 pub fn to_str(&self) -> Result<&str, core::str::Utf8Error> {
281 core::str::from_utf8(self.as_bytes())
282 }
283
284 /// Unsafely convert this [`CStr`] into a [`&str`], without checking for
285 /// valid UTF-8.
286 ///
287 /// # Safety
288 ///
289 /// The contents must be valid UTF-8.
290 ///
291 /// # Examples
292 ///
293 /// ```
294 /// # use kernel::c_str;
295 /// # use kernel::str::CStr;
296 /// let bar = c_str!("ツ");
297 /// // SAFETY: String literals are guaranteed to be valid UTF-8
298 /// // by the Rust compiler.
299 /// assert_eq!(unsafe { bar.as_str_unchecked() }, "ツ");
300 /// ```
301 #[inline]
302 pub unsafe fn as_str_unchecked(&self) -> &str {
303 // SAFETY: TODO.
304 unsafe { core::str::from_utf8_unchecked(self.as_bytes()) }
305 }
306
307 /// Convert this [`CStr`] into a [`CString`] by allocating memory and
308 /// copying over the string data.
309 pub fn to_cstring(&self) -> Result<CString, AllocError> {
310 CString::try_from(self)
311 }
312
313 /// Converts this [`CStr`] to its ASCII lower case equivalent in-place.
314 ///
315 /// ASCII letters 'A' to 'Z' are mapped to 'a' to 'z',
316 /// but non-ASCII letters are unchanged.
317 ///
318 /// To return a new lowercased value without modifying the existing one, use
319 /// [`to_ascii_lowercase()`].
320 ///
321 /// [`to_ascii_lowercase()`]: #method.to_ascii_lowercase
322 pub fn make_ascii_lowercase(&mut self) {
323 // INVARIANT: This doesn't introduce or remove NUL bytes in the C
324 // string.
325 self.0.make_ascii_lowercase();
326 }
327
328 /// Converts this [`CStr`] to its ASCII upper case equivalent in-place.
329 ///
330 /// ASCII letters 'a' to 'z' are mapped to 'A' to 'Z',
331 /// but non-ASCII letters are unchanged.
332 ///
333 /// To return a new uppercased value without modifying the existing one, use
334 /// [`to_ascii_uppercase()`].
335 ///
336 /// [`to_ascii_uppercase()`]: #method.to_ascii_uppercase
337 pub fn make_ascii_uppercase(&mut self) {
338 // INVARIANT: This doesn't introduce or remove NUL bytes in the C
339 // string.
340 self.0.make_ascii_uppercase();
341 }
342
343 /// Returns a copy of this [`CString`] where each character is mapped to its
344 /// ASCII lower case equivalent.
345 ///
346 /// ASCII letters 'A' to 'Z' are mapped to 'a' to 'z',
347 /// but non-ASCII letters are unchanged.
348 ///
349 /// To lowercase the value in-place, use [`make_ascii_lowercase`].
350 ///
351 /// [`make_ascii_lowercase`]: str::make_ascii_lowercase
352 pub fn to_ascii_lowercase(&self) -> Result<CString, AllocError> {
353 let mut s = self.to_cstring()?;
354
355 s.make_ascii_lowercase();
356
357 Ok(s)
358 }
359
360 /// Returns a copy of this [`CString`] where each character is mapped to its
361 /// ASCII upper case equivalent.
362 ///
363 /// ASCII letters 'a' to 'z' are mapped to 'A' to 'Z',
364 /// but non-ASCII letters are unchanged.
365 ///
366 /// To uppercase the value in-place, use [`make_ascii_uppercase`].
367 ///
368 /// [`make_ascii_uppercase`]: str::make_ascii_uppercase
369 pub fn to_ascii_uppercase(&self) -> Result<CString, AllocError> {
370 let mut s = self.to_cstring()?;
371
372 s.make_ascii_uppercase();
373
374 Ok(s)
375 }
376}
377
378impl fmt::Display for CStr {
379 /// Formats printable ASCII characters, escaping the rest.
380 ///
381 /// ```
382 /// # use kernel::c_str;
383 /// # use kernel::fmt;
384 /// # use kernel::str::CStr;
385 /// # use kernel::str::CString;
386 /// let penguin = c_str!("🐧");
387 /// let s = CString::try_from_fmt(fmt!("{}", penguin)).unwrap();
388 /// assert_eq!(s.as_bytes_with_nul(), "\\xf0\\x9f\\x90\\xa7\0".as_bytes());
389 ///
390 /// let ascii = c_str!("so \"cool\"");
391 /// let s = CString::try_from_fmt(fmt!("{}", ascii)).unwrap();
392 /// assert_eq!(s.as_bytes_with_nul(), "so \"cool\"\0".as_bytes());
393 /// ```
394 fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
395 for &c in self.as_bytes() {
396 if (0x20..0x7f).contains(&c) {
397 // Printable character.
398 f.write_char(c as char)?;
399 } else {
400 write!(f, "\\x{:02x}", c)?;
401 }
402 }
403 Ok(())
404 }
405}
406
407impl fmt::Debug for CStr {
408 /// Formats printable ASCII characters with a double quote on either end, escaping the rest.
409 ///
410 /// ```
411 /// # use kernel::c_str;
412 /// # use kernel::fmt;
413 /// # use kernel::str::CStr;
414 /// # use kernel::str::CString;
415 /// let penguin = c_str!("🐧");
416 /// let s = CString::try_from_fmt(fmt!("{:?}", penguin)).unwrap();
417 /// assert_eq!(s.as_bytes_with_nul(), "\"\\xf0\\x9f\\x90\\xa7\"\0".as_bytes());
418 ///
419 /// // Embedded double quotes are escaped.
420 /// let ascii = c_str!("so \"cool\"");
421 /// let s = CString::try_from_fmt(fmt!("{:?}", ascii)).unwrap();
422 /// assert_eq!(s.as_bytes_with_nul(), "\"so \\\"cool\\\"\"\0".as_bytes());
423 /// ```
424 fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
425 f.write_str("\"")?;
426 for &c in self.as_bytes() {
427 match c {
428 // Printable characters.
429 b'\"' => f.write_str("\\\"")?,
430 0x20..=0x7e => f.write_char(c as char)?,
431 _ => write!(f, "\\x{:02x}", c)?,
432 }
433 }
434 f.write_str("\"")
435 }
436}
437
438impl AsRef<BStr> for CStr {
439 #[inline]
440 fn as_ref(&self) -> &BStr {
441 BStr::from_bytes(self.as_bytes())
442 }
443}
444
445impl Deref for CStr {
446 type Target = BStr;
447
448 #[inline]
449 fn deref(&self) -> &Self::Target {
450 self.as_ref()
451 }
452}
453
454impl Index<ops::RangeFrom<usize>> for CStr {
455 type Output = CStr;
456
457 #[inline]
458 fn index(&self, index: ops::RangeFrom<usize>) -> &Self::Output {
459 // Delegate bounds checking to slice.
460 // Assign to _ to mute clippy's unnecessary operation warning.
461 let _ = &self.as_bytes()[index.start..];
462 // SAFETY: We just checked the bounds.
463 unsafe { Self::from_bytes_with_nul_unchecked(&self.0[index.start..]) }
464 }
465}
466
467impl Index<ops::RangeFull> for CStr {
468 type Output = CStr;
469
470 #[inline]
471 fn index(&self, _index: ops::RangeFull) -> &Self::Output {
472 self
473 }
474}
475
476mod private {
477 use core::ops;
478
479 // Marker trait for index types that can be forward to `BStr`.
480 pub trait CStrIndex {}
481
482 impl CStrIndex for usize {}
483 impl CStrIndex for ops::Range<usize> {}
484 impl CStrIndex for ops::RangeInclusive<usize> {}
485 impl CStrIndex for ops::RangeToInclusive<usize> {}
486}
487
488impl<Idx> Index<Idx> for CStr
489where
490 Idx: private::CStrIndex,
491 BStr: Index<Idx>,
492{
493 type Output = <BStr as Index<Idx>>::Output;
494
495 #[inline]
496 fn index(&self, index: Idx) -> &Self::Output {
497 &self.as_ref()[index]
498 }
499}
500
501/// Creates a new [`CStr`] from a string literal.
502///
503/// The string literal should not contain any `NUL` bytes.
504///
505/// # Examples
506///
507/// ```
508/// # use kernel::c_str;
509/// # use kernel::str::CStr;
510/// const MY_CSTR: &CStr = c_str!("My awesome CStr!");
511/// ```
512#[macro_export]
513macro_rules! c_str {
514 ($str:expr) => {{
515 const S: &str = concat!($str, "\0");
516 const C: &$crate::str::CStr = match $crate::str::CStr::from_bytes_with_nul(S.as_bytes()) {
517 Ok(v) => v,
518 Err(_) => panic!("string contains interior NUL"),
519 };
520 C
521 }};
522}
523
524#[cfg(test)]
525mod tests {
526 use super::*;
527
528 struct String(CString);
529
530 impl String {
531 fn from_fmt(args: fmt::Arguments<'_>) -> Self {
532 String(CString::try_from_fmt(args).unwrap())
533 }
534 }
535
536 impl Deref for String {
537 type Target = str;
538
539 fn deref(&self) -> &str {
540 self.0.to_str().unwrap()
541 }
542 }
543
544 macro_rules! format {
545 ($($f:tt)*) => ({
546 &*String::from_fmt(kernel::fmt!($($f)*))
547 })
548 }
549
550 const ALL_ASCII_CHARS: &'static str =
551 "\\x01\\x02\\x03\\x04\\x05\\x06\\x07\\x08\\x09\\x0a\\x0b\\x0c\\x0d\\x0e\\x0f\
552 \\x10\\x11\\x12\\x13\\x14\\x15\\x16\\x17\\x18\\x19\\x1a\\x1b\\x1c\\x1d\\x1e\\x1f \
553 !\"#$%&'()*+,-./0123456789:;<=>?@\
554 ABCDEFGHIJKLMNOPQRSTUVWXYZ[\\]^_`abcdefghijklmnopqrstuvwxyz{|}~\\x7f\
555 \\x80\\x81\\x82\\x83\\x84\\x85\\x86\\x87\\x88\\x89\\x8a\\x8b\\x8c\\x8d\\x8e\\x8f\
556 \\x90\\x91\\x92\\x93\\x94\\x95\\x96\\x97\\x98\\x99\\x9a\\x9b\\x9c\\x9d\\x9e\\x9f\
557 \\xa0\\xa1\\xa2\\xa3\\xa4\\xa5\\xa6\\xa7\\xa8\\xa9\\xaa\\xab\\xac\\xad\\xae\\xaf\
558 \\xb0\\xb1\\xb2\\xb3\\xb4\\xb5\\xb6\\xb7\\xb8\\xb9\\xba\\xbb\\xbc\\xbd\\xbe\\xbf\
559 \\xc0\\xc1\\xc2\\xc3\\xc4\\xc5\\xc6\\xc7\\xc8\\xc9\\xca\\xcb\\xcc\\xcd\\xce\\xcf\
560 \\xd0\\xd1\\xd2\\xd3\\xd4\\xd5\\xd6\\xd7\\xd8\\xd9\\xda\\xdb\\xdc\\xdd\\xde\\xdf\
561 \\xe0\\xe1\\xe2\\xe3\\xe4\\xe5\\xe6\\xe7\\xe8\\xe9\\xea\\xeb\\xec\\xed\\xee\\xef\
562 \\xf0\\xf1\\xf2\\xf3\\xf4\\xf5\\xf6\\xf7\\xf8\\xf9\\xfa\\xfb\\xfc\\xfd\\xfe\\xff";
563
564 #[test]
565 fn test_cstr_to_str() {
566 let good_bytes = b"\xf0\x9f\xa6\x80\0";
567 let checked_cstr = CStr::from_bytes_with_nul(good_bytes).unwrap();
568 let checked_str = checked_cstr.to_str().unwrap();
569 assert_eq!(checked_str, "🦀");
570 }
571
572 #[test]
573 #[should_panic]
574 fn test_cstr_to_str_panic() {
575 let bad_bytes = b"\xc3\x28\0";
576 let checked_cstr = CStr::from_bytes_with_nul(bad_bytes).unwrap();
577 checked_cstr.to_str().unwrap();
578 }
579
580 #[test]
581 fn test_cstr_as_str_unchecked() {
582 let good_bytes = b"\xf0\x9f\x90\xA7\0";
583 let checked_cstr = CStr::from_bytes_with_nul(good_bytes).unwrap();
584 let unchecked_str = unsafe { checked_cstr.as_str_unchecked() };
585 assert_eq!(unchecked_str, "🐧");
586 }
587
588 #[test]
589 fn test_cstr_display() {
590 let hello_world = CStr::from_bytes_with_nul(b"hello, world!\0").unwrap();
591 assert_eq!(format!("{}", hello_world), "hello, world!");
592 let non_printables = CStr::from_bytes_with_nul(b"\x01\x09\x0a\0").unwrap();
593 assert_eq!(format!("{}", non_printables), "\\x01\\x09\\x0a");
594 let non_ascii = CStr::from_bytes_with_nul(b"d\xe9j\xe0 vu\0").unwrap();
595 assert_eq!(format!("{}", non_ascii), "d\\xe9j\\xe0 vu");
596 let good_bytes = CStr::from_bytes_with_nul(b"\xf0\x9f\xa6\x80\0").unwrap();
597 assert_eq!(format!("{}", good_bytes), "\\xf0\\x9f\\xa6\\x80");
598 }
599
600 #[test]
601 fn test_cstr_display_all_bytes() {
602 let mut bytes: [u8; 256] = [0; 256];
603 // fill `bytes` with [1..=255] + [0]
604 for i in u8::MIN..=u8::MAX {
605 bytes[i as usize] = i.wrapping_add(1);
606 }
607 let cstr = CStr::from_bytes_with_nul(&bytes).unwrap();
608 assert_eq!(format!("{}", cstr), ALL_ASCII_CHARS);
609 }
610
611 #[test]
612 fn test_cstr_debug() {
613 let hello_world = CStr::from_bytes_with_nul(b"hello, world!\0").unwrap();
614 assert_eq!(format!("{:?}", hello_world), "\"hello, world!\"");
615 let non_printables = CStr::from_bytes_with_nul(b"\x01\x09\x0a\0").unwrap();
616 assert_eq!(format!("{:?}", non_printables), "\"\\x01\\x09\\x0a\"");
617 let non_ascii = CStr::from_bytes_with_nul(b"d\xe9j\xe0 vu\0").unwrap();
618 assert_eq!(format!("{:?}", non_ascii), "\"d\\xe9j\\xe0 vu\"");
619 let good_bytes = CStr::from_bytes_with_nul(b"\xf0\x9f\xa6\x80\0").unwrap();
620 assert_eq!(format!("{:?}", good_bytes), "\"\\xf0\\x9f\\xa6\\x80\"");
621 }
622
623 #[test]
624 fn test_bstr_display() {
625 let hello_world = BStr::from_bytes(b"hello, world!");
626 assert_eq!(format!("{}", hello_world), "hello, world!");
627 let escapes = BStr::from_bytes(b"_\t_\n_\r_\\_\'_\"_");
628 assert_eq!(format!("{}", escapes), "_\\t_\\n_\\r_\\_'_\"_");
629 let others = BStr::from_bytes(b"\x01");
630 assert_eq!(format!("{}", others), "\\x01");
631 let non_ascii = BStr::from_bytes(b"d\xe9j\xe0 vu");
632 assert_eq!(format!("{}", non_ascii), "d\\xe9j\\xe0 vu");
633 let good_bytes = BStr::from_bytes(b"\xf0\x9f\xa6\x80");
634 assert_eq!(format!("{}", good_bytes), "\\xf0\\x9f\\xa6\\x80");
635 }
636
637 #[test]
638 fn test_bstr_debug() {
639 let hello_world = BStr::from_bytes(b"hello, world!");
640 assert_eq!(format!("{:?}", hello_world), "\"hello, world!\"");
641 let escapes = BStr::from_bytes(b"_\t_\n_\r_\\_\'_\"_");
642 assert_eq!(format!("{:?}", escapes), "\"_\\t_\\n_\\r_\\\\_'_\\\"_\"");
643 let others = BStr::from_bytes(b"\x01");
644 assert_eq!(format!("{:?}", others), "\"\\x01\"");
645 let non_ascii = BStr::from_bytes(b"d\xe9j\xe0 vu");
646 assert_eq!(format!("{:?}", non_ascii), "\"d\\xe9j\\xe0 vu\"");
647 let good_bytes = BStr::from_bytes(b"\xf0\x9f\xa6\x80");
648 assert_eq!(format!("{:?}", good_bytes), "\"\\xf0\\x9f\\xa6\\x80\"");
649 }
650}
651
652/// Allows formatting of [`fmt::Arguments`] into a raw buffer.
653///
654/// It does not fail if callers write past the end of the buffer so that they can calculate the
655/// size required to fit everything.
656///
657/// # Invariants
658///
659/// The memory region between `pos` (inclusive) and `end` (exclusive) is valid for writes if `pos`
660/// is less than `end`.
661pub(crate) struct RawFormatter {
662 // Use `usize` to use `saturating_*` functions.
663 beg: usize,
664 pos: usize,
665 end: usize,
666}
667
668impl RawFormatter {
669 /// Creates a new instance of [`RawFormatter`] with an empty buffer.
670 fn new() -> Self {
671 // INVARIANT: The buffer is empty, so the region that needs to be writable is empty.
672 Self {
673 beg: 0,
674 pos: 0,
675 end: 0,
676 }
677 }
678
679 /// Creates a new instance of [`RawFormatter`] with the given buffer pointers.
680 ///
681 /// # Safety
682 ///
683 /// If `pos` is less than `end`, then the region between `pos` (inclusive) and `end`
684 /// (exclusive) must be valid for writes for the lifetime of the returned [`RawFormatter`].
685 pub(crate) unsafe fn from_ptrs(pos: *mut u8, end: *mut u8) -> Self {
686 // INVARIANT: The safety requirements guarantee the type invariants.
687 Self {
688 beg: pos as _,
689 pos: pos as _,
690 end: end as _,
691 }
692 }
693
694 /// Creates a new instance of [`RawFormatter`] with the given buffer.
695 ///
696 /// # Safety
697 ///
698 /// The memory region starting at `buf` and extending for `len` bytes must be valid for writes
699 /// for the lifetime of the returned [`RawFormatter`].
700 pub(crate) unsafe fn from_buffer(buf: *mut u8, len: usize) -> Self {
701 let pos = buf as usize;
702 // INVARIANT: We ensure that `end` is never less then `buf`, and the safety requirements
703 // guarantees that the memory region is valid for writes.
704 Self {
705 pos,
706 beg: pos,
707 end: pos.saturating_add(len),
708 }
709 }
710
711 /// Returns the current insert position.
712 ///
713 /// N.B. It may point to invalid memory.
714 pub(crate) fn pos(&self) -> *mut u8 {
715 self.pos as _
716 }
717
718 /// Returns the number of bytes written to the formatter.
719 pub(crate) fn bytes_written(&self) -> usize {
720 self.pos - self.beg
721 }
722}
723
724impl fmt::Write for RawFormatter {
725 fn write_str(&mut self, s: &str) -> fmt::Result {
726 // `pos` value after writing `len` bytes. This does not have to be bounded by `end`, but we
727 // don't want it to wrap around to 0.
728 let pos_new = self.pos.saturating_add(s.len());
729
730 // Amount that we can copy. `saturating_sub` ensures we get 0 if `pos` goes past `end`.
731 let len_to_copy = core::cmp::min(pos_new, self.end).saturating_sub(self.pos);
732
733 if len_to_copy > 0 {
734 // SAFETY: If `len_to_copy` is non-zero, then we know `pos` has not gone past `end`
735 // yet, so it is valid for write per the type invariants.
736 unsafe {
737 core::ptr::copy_nonoverlapping(
738 s.as_bytes().as_ptr(),
739 self.pos as *mut u8,
740 len_to_copy,
741 )
742 };
743 }
744
745 self.pos = pos_new;
746 Ok(())
747 }
748}
749
750/// Allows formatting of [`fmt::Arguments`] into a raw buffer.
751///
752/// Fails if callers attempt to write more than will fit in the buffer.
753pub(crate) struct Formatter(RawFormatter);
754
755impl Formatter {
756 /// Creates a new instance of [`Formatter`] with the given buffer.
757 ///
758 /// # Safety
759 ///
760 /// The memory region starting at `buf` and extending for `len` bytes must be valid for writes
761 /// for the lifetime of the returned [`Formatter`].
762 pub(crate) unsafe fn from_buffer(buf: *mut u8, len: usize) -> Self {
763 // SAFETY: The safety requirements of this function satisfy those of the callee.
764 Self(unsafe { RawFormatter::from_buffer(buf, len) })
765 }
766}
767
768impl Deref for Formatter {
769 type Target = RawFormatter;
770
771 fn deref(&self) -> &Self::Target {
772 &self.0
773 }
774}
775
776impl fmt::Write for Formatter {
777 fn write_str(&mut self, s: &str) -> fmt::Result {
778 self.0.write_str(s)?;
779
780 // Fail the request if we go past the end of the buffer.
781 if self.0.pos > self.0.end {
782 Err(fmt::Error)
783 } else {
784 Ok(())
785 }
786 }
787}
788
789/// An owned string that is guaranteed to have exactly one `NUL` byte, which is at the end.
790///
791/// Used for interoperability with kernel APIs that take C strings.
792///
793/// # Invariants
794///
795/// The string is always `NUL`-terminated and contains no other `NUL` bytes.
796///
797/// # Examples
798///
799/// ```
800/// use kernel::{str::CString, fmt};
801///
802/// let s = CString::try_from_fmt(fmt!("{}{}{}", "abc", 10, 20)).unwrap();
803/// assert_eq!(s.as_bytes_with_nul(), "abc1020\0".as_bytes());
804///
805/// let tmp = "testing";
806/// let s = CString::try_from_fmt(fmt!("{tmp}{}", 123)).unwrap();
807/// assert_eq!(s.as_bytes_with_nul(), "testing123\0".as_bytes());
808///
809/// // This fails because it has an embedded `NUL` byte.
810/// let s = CString::try_from_fmt(fmt!("a\0b{}", 123));
811/// assert_eq!(s.is_ok(), false);
812/// ```
813pub struct CString {
814 buf: KVec<u8>,
815}
816
817impl CString {
818 /// Creates an instance of [`CString`] from the given formatted arguments.
819 pub fn try_from_fmt(args: fmt::Arguments<'_>) -> Result<Self, Error> {
820 // Calculate the size needed (formatted string plus `NUL` terminator).
821 let mut f = RawFormatter::new();
822 f.write_fmt(args)?;
823 f.write_str("\0")?;
824 let size = f.bytes_written();
825
826 // Allocate a vector with the required number of bytes, and write to it.
827 let mut buf = KVec::with_capacity(size, GFP_KERNEL)?;
828 // SAFETY: The buffer stored in `buf` is at least of size `size` and is valid for writes.
829 let mut f = unsafe { Formatter::from_buffer(buf.as_mut_ptr(), size) };
830 f.write_fmt(args)?;
831 f.write_str("\0")?;
832
833 // SAFETY: The number of bytes that can be written to `f` is bounded by `size`, which is
834 // `buf`'s capacity. The contents of the buffer have been initialised by writes to `f`.
835 unsafe { buf.set_len(f.bytes_written()) };
836
837 // Check that there are no `NUL` bytes before the end.
838 // SAFETY: The buffer is valid for read because `f.bytes_written()` is bounded by `size`
839 // (which the minimum buffer size) and is non-zero (we wrote at least the `NUL` terminator)
840 // so `f.bytes_written() - 1` doesn't underflow.
841 let ptr = unsafe { bindings::memchr(buf.as_ptr().cast(), 0, f.bytes_written() - 1) };
842 if !ptr.is_null() {
843 return Err(EINVAL);
844 }
845
846 // INVARIANT: We wrote the `NUL` terminator and checked above that no other `NUL` bytes
847 // exist in the buffer.
848 Ok(Self { buf })
849 }
850}
851
852impl Deref for CString {
853 type Target = CStr;
854
855 fn deref(&self) -> &Self::Target {
856 // SAFETY: The type invariants guarantee that the string is `NUL`-terminated and that no
857 // other `NUL` bytes exist.
858 unsafe { CStr::from_bytes_with_nul_unchecked(self.buf.as_slice()) }
859 }
860}
861
862impl DerefMut for CString {
863 fn deref_mut(&mut self) -> &mut Self::Target {
864 // SAFETY: A `CString` is always NUL-terminated and contains no other
865 // NUL bytes.
866 unsafe { CStr::from_bytes_with_nul_unchecked_mut(self.buf.as_mut_slice()) }
867 }
868}
869
870impl<'a> TryFrom<&'a CStr> for CString {
871 type Error = AllocError;
872
873 fn try_from(cstr: &'a CStr) -> Result<CString, AllocError> {
874 let mut buf = KVec::new();
875
876 buf.extend_from_slice(cstr.as_bytes_with_nul(), GFP_KERNEL)?;
877
878 // INVARIANT: The `CStr` and `CString` types have the same invariants for
879 // the string data, and we copied it over without changes.
880 Ok(CString { buf })
881 }
882}
883
884impl fmt::Debug for CString {
885 fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
886 fmt::Debug::fmt(&**self, f)
887 }
888}
889
890/// A convenience alias for [`core::format_args`].
891#[macro_export]
892macro_rules! fmt {
893 ($($f:tt)*) => ( core::format_args!($($f)*) )
894}