1 //! The virtual memory representation of the MIR interpreter.
4 use std::convert::TryFrom;
6 use std::ops::{Deref, DerefMut, Range};
9 use rustc_ast::Mutability;
10 use rustc_data_structures::sorted_map::SortedMap;
11 use rustc_target::abi::{Align, HasDataLayout, Size};
14 read_target_uint, write_target_uint, AllocId, InterpError, Pointer, Scalar, ScalarMaybeUninit,
15 UndefinedBehaviorInfo, UninitBytesAccess, UnsupportedOpInfo,
18 /// This type represents an Allocation in the Miri/CTFE core engine.
20 /// Its public API is rather low-level, working directly with allocation offsets and a custom error
21 /// type to account for the lack of an AllocId on this level. The Miri/CTFE core engine `memory`
22 /// module provides higher-level access.
23 #[derive(Clone, Debug, Eq, PartialEq, PartialOrd, Ord, Hash, TyEncodable, TyDecodable)]
25 pub struct Allocation<Tag = (), Extra = ()> {
26 /// The actual bytes of the allocation.
27 /// Note that the bytes of a pointer represent the offset of the pointer.
29 /// Maps from byte addresses to extra data for each pointer.
30 /// Only the first byte of a pointer is inserted into the map; i.e.,
31 /// every entry in this map applies to `pointer_size` consecutive bytes starting
32 /// at the given offset.
33 relocations: Relocations<Tag>,
34 /// Denotes which part of this allocation is initialized.
36 /// The alignment of the allocation to detect unaligned reads.
37 /// (`Align` guarantees that this is a power of two.)
39 /// `true` if the allocation is mutable.
40 /// Also used by codegen to determine if a static should be put into mutable memory,
41 /// which happens for `static mut` and `static` with interior mutability.
42 pub mutability: Mutability,
43 /// Extra state for the machine.
47 /// We have our own error type that does not know about the `AllocId`; that information
48 /// is added when converting to `InterpError`.
51 /// Encountered a pointer where we needed raw bytes.
53 /// Using uninitialized data where it is not allowed.
54 InvalidUninitBytes(Option<UninitBytesAccess>),
56 pub type AllocResult<T = ()> = Result<T, AllocError>;
59 pub fn to_interp_error<'tcx>(self, alloc_id: AllocId) -> InterpError<'tcx> {
61 AllocError::ReadPointerAsBytes => {
62 InterpError::Unsupported(UnsupportedOpInfo::ReadPointerAsBytes)
64 AllocError::InvalidUninitBytes(info) => InterpError::UndefinedBehavior(
65 UndefinedBehaviorInfo::InvalidUninitBytes(info.map(|b| (alloc_id, b))),
71 /// The information that makes up a memory access: offset and size.
72 #[derive(Copy, Clone, Debug)]
73 pub struct AllocRange {
78 /// Free-starting constructor for less syntactic overhead.
80 pub fn alloc_range(start: Size, size: Size) -> AllocRange {
81 AllocRange { start, size }
86 pub fn end(self) -> Size {
87 self.start + self.size // This does overflow checking.
90 /// Returns the `subrange` within this range; panics if it is not a subrange.
92 pub fn subrange(self, subrange: AllocRange) -> AllocRange {
93 let sub_start = self.start + subrange.start;
94 let range = alloc_range(sub_start, subrange.size);
95 assert!(range.end() <= self.end(), "access outside the bounds for given AllocRange");
100 // The constructors are all without extra; the extra gets added by a machine hook later.
101 impl<Tag> Allocation<Tag> {
102 /// Creates a read-only allocation initialized by the given bytes
103 pub fn from_bytes<'a>(slice: impl Into<Cow<'a, [u8]>>, align: Align) -> Self {
104 let bytes = slice.into().into_owned();
105 let size = Size::from_bytes(bytes.len());
108 relocations: Relocations::new(),
109 init_mask: InitMask::new(size, true),
111 mutability: Mutability::Not,
116 pub fn from_byte_aligned_bytes<'a>(slice: impl Into<Cow<'a, [u8]>>) -> Self {
117 Allocation::from_bytes(slice, Align::ONE)
120 pub fn uninit(size: Size, align: Align) -> Self {
122 bytes: vec![0; size.bytes_usize()],
123 relocations: Relocations::new(),
124 init_mask: InitMask::new(size, false),
126 mutability: Mutability::Mut,
132 impl Allocation<()> {
133 /// Add Tag and Extra fields
134 pub fn with_tags_and_extra<T, E>(
136 mut tagger: impl FnMut(AllocId) -> T,
138 ) -> Allocation<T, E> {
141 relocations: Relocations::from_presorted(
144 // The allocations in the relocations (pointers stored *inside* this allocation)
145 // all get the base pointer tag.
146 .map(|&(offset, ((), alloc))| {
147 let tag = tagger(alloc);
148 (offset, (tag, alloc))
152 init_mask: self.init_mask,
154 mutability: self.mutability,
160 /// Raw accessors. Provide access to otherwise private bytes.
161 impl<Tag, Extra> Allocation<Tag, Extra> {
162 pub fn len(&self) -> usize {
166 pub fn size(&self) -> Size {
167 Size::from_bytes(self.len())
170 /// Looks at a slice which may describe uninitialized bytes or describe a relocation. This differs
171 /// from `get_bytes_with_uninit_and_ptr` in that it does no relocation checks (even on the
173 /// This must not be used for reads affecting the interpreter execution.
174 pub fn inspect_with_uninit_and_ptr_outside_interpreter(&self, range: Range<usize>) -> &[u8] {
178 /// Returns the mask indicating which bytes are initialized.
179 pub fn init_mask(&self) -> &InitMask {
183 /// Returns the relocation list.
184 pub fn relocations(&self) -> &Relocations<Tag> {
190 impl<Tag: Copy, Extra> Allocation<Tag, Extra> {
191 /// The last argument controls whether we error out when there are uninitialized
192 /// or pointer bytes. You should never call this, call `get_bytes` or
193 /// `get_bytes_with_uninit_and_ptr` instead,
195 /// This function also guarantees that the resulting pointer will remain stable
196 /// even when new allocations are pushed to the `HashMap`. `copy_repeatedly` relies
199 /// It is the caller's responsibility to check bounds and alignment beforehand.
200 fn get_bytes_internal(
202 cx: &impl HasDataLayout,
204 check_init_and_ptr: bool,
205 ) -> AllocResult<&[u8]> {
206 if check_init_and_ptr {
207 self.check_init(range)?;
208 self.check_relocations(cx, range)?;
210 // We still don't want relocations on the *edges*.
211 self.check_relocation_edges(cx, range)?;
214 Ok(&self.bytes[range.start.bytes_usize()..range.end().bytes_usize()])
217 /// Checks that these bytes are initialized and not pointer bytes, and then return them
220 /// It is the caller's responsibility to check bounds and alignment beforehand.
221 /// Most likely, you want to use the `PlaceTy` and `OperandTy`-based methods
222 /// on `InterpCx` instead.
224 pub fn get_bytes(&self, cx: &impl HasDataLayout, range: AllocRange) -> AllocResult<&[u8]> {
225 self.get_bytes_internal(cx, range, true)
228 /// It is the caller's responsibility to handle uninitialized and pointer bytes.
229 /// However, this still checks that there are no relocations on the *edges*.
231 /// It is the caller's responsibility to check bounds and alignment beforehand.
233 pub fn get_bytes_with_uninit_and_ptr(
235 cx: &impl HasDataLayout,
237 ) -> AllocResult<&[u8]> {
238 self.get_bytes_internal(cx, range, false)
241 /// Just calling this already marks everything as defined and removes relocations,
242 /// so be sure to actually put data there!
244 /// It is the caller's responsibility to check bounds and alignment beforehand.
245 /// Most likely, you want to use the `PlaceTy` and `OperandTy`-based methods
246 /// on `InterpCx` instead.
247 pub fn get_bytes_mut(&mut self, cx: &impl HasDataLayout, range: AllocRange) -> &mut [u8] {
248 self.mark_init(range, true);
249 self.clear_relocations(cx, range);
251 &mut self.bytes[range.start.bytes_usize()..range.end().bytes_usize()]
254 /// A raw pointer variant of `get_bytes_mut` that avoids invalidating existing aliases into this memory.
255 pub fn get_bytes_mut_ptr(&mut self, cx: &impl HasDataLayout, range: AllocRange) -> *mut [u8] {
256 self.mark_init(range, true);
257 self.clear_relocations(cx, range);
259 assert!(range.end().bytes_usize() <= self.bytes.len()); // need to do our own bounds-check
260 let begin_ptr = self.bytes.as_mut_ptr().wrapping_add(range.start.bytes_usize());
261 let len = range.end().bytes_usize() - range.start.bytes_usize();
262 ptr::slice_from_raw_parts_mut(begin_ptr, len)
266 /// Reading and writing.
267 impl<Tag: Copy, Extra> Allocation<Tag, Extra> {
268 /// Validates that `ptr.offset` and `ptr.offset + size` do not point to the middle of a
269 /// relocation. If `allow_uninit_and_ptr` is `false`, also enforces that the memory in the
270 /// given range contains neither relocations nor uninitialized bytes.
273 cx: &impl HasDataLayout,
275 allow_uninit_and_ptr: bool,
277 // Check bounds and relocations on the edges.
278 self.get_bytes_with_uninit_and_ptr(cx, range)?;
279 // Check uninit and ptr.
280 if !allow_uninit_and_ptr {
281 self.check_init(range)?;
282 self.check_relocations(cx, range)?;
287 /// Reads a *non-ZST* scalar.
289 /// ZSTs can't be read because in order to obtain a `Pointer`, we need to check
290 /// for ZSTness anyway due to integer pointers being valid for ZSTs.
292 /// It is the caller's responsibility to check bounds and alignment beforehand.
293 /// Most likely, you want to call `InterpCx::read_scalar` instead of this method.
296 cx: &impl HasDataLayout,
298 ) -> AllocResult<ScalarMaybeUninit<Tag>> {
299 // `get_bytes_unchecked` tests relocation edges.
300 let bytes = self.get_bytes_with_uninit_and_ptr(cx, range)?;
301 // Uninit check happens *after* we established that the alignment is correct.
302 // We must not return `Ok()` for unaligned pointers!
303 if self.is_init(range).is_err() {
304 // This inflates uninitialized bytes to the entire scalar, even if only a few
305 // bytes are uninitialized.
306 return Ok(ScalarMaybeUninit::Uninit);
308 // Now we do the actual reading.
309 let bits = read_target_uint(cx.data_layout().endian, bytes).unwrap();
310 // See if we got a pointer.
311 if range.size != cx.data_layout().pointer_size {
313 // *Now*, we better make sure that the inside is free of relocations too.
314 self.check_relocations(cx, range)?;
317 if let Some(&(tag, alloc_id)) = self.relocations.get(&range.start) {
318 let ptr = Pointer::new_with_tag(alloc_id, Size::from_bytes(bits), tag);
319 return Ok(ScalarMaybeUninit::Scalar(ptr.into()));
322 // We don't. Just return the bits.
323 Ok(ScalarMaybeUninit::Scalar(Scalar::from_uint(bits, range.size)))
326 /// Writes a *non-ZST* scalar.
328 /// ZSTs can't be read because in order to obtain a `Pointer`, we need to check
329 /// for ZSTness anyway due to integer pointers being valid for ZSTs.
331 /// It is the caller's responsibility to check bounds and alignment beforehand.
332 /// Most likely, you want to call `InterpCx::write_scalar` instead of this method.
335 cx: &impl HasDataLayout,
337 val: ScalarMaybeUninit<Tag>,
339 let val = match val {
340 ScalarMaybeUninit::Scalar(scalar) => scalar,
341 ScalarMaybeUninit::Uninit => {
342 self.mark_init(range, false);
347 let bytes = match val.to_bits_or_ptr(range.size, cx) {
348 Err(val) => u128::from(val.offset.bytes()),
352 let endian = cx.data_layout().endian;
353 let dst = self.get_bytes_mut(cx, range);
354 write_target_uint(endian, dst, bytes).unwrap();
356 // See if we have to also write a relocation.
357 if let Scalar::Ptr(val) = val {
358 self.relocations.insert(range.start, (val.tag, val.alloc_id));
366 impl<Tag: Copy, Extra> Allocation<Tag, Extra> {
367 /// Returns all relocations overlapping with the given pointer-offset pair.
368 pub fn get_relocations(
370 cx: &impl HasDataLayout,
372 ) -> &[(Size, (Tag, AllocId))] {
373 // We have to go back `pointer_size - 1` bytes, as that one would still overlap with
374 // the beginning of this range.
375 let start = range.start.bytes().saturating_sub(cx.data_layout().pointer_size.bytes() - 1);
376 self.relocations.range(Size::from_bytes(start)..range.end())
379 /// Checks that there are no relocations overlapping with the given range.
381 fn check_relocations(&self, cx: &impl HasDataLayout, range: AllocRange) -> AllocResult {
382 if self.get_relocations(cx, range).is_empty() {
385 Err(AllocError::ReadPointerAsBytes)
389 /// Removes all relocations inside the given range.
390 /// If there are relocations overlapping with the edges, they
391 /// are removed as well *and* the bytes they cover are marked as
392 /// uninitialized. This is a somewhat odd "spooky action at a distance",
393 /// but it allows strictly more code to run than if we would just error
394 /// immediately in that case.
395 fn clear_relocations(&mut self, cx: &impl HasDataLayout, range: AllocRange) {
396 // Find the start and end of the given range and its outermost relocations.
397 let (first, last) = {
398 // Find all relocations overlapping the given range.
399 let relocations = self.get_relocations(cx, range);
400 if relocations.is_empty() {
405 relocations.first().unwrap().0,
406 relocations.last().unwrap().0 + cx.data_layout().pointer_size,
409 let start = range.start;
410 let end = range.end();
412 // Mark parts of the outermost relocations as uninitialized if they partially fall outside the
415 self.init_mask.set_range(first, start, false);
418 self.init_mask.set_range(end, last, false);
421 // Forget all the relocations.
422 self.relocations.remove_range(first..last);
425 /// Errors if there are relocations overlapping with the edges of the
426 /// given memory range.
428 fn check_relocation_edges(&self, cx: &impl HasDataLayout, range: AllocRange) -> AllocResult {
429 self.check_relocations(cx, alloc_range(range.start, Size::ZERO))?;
430 self.check_relocations(cx, alloc_range(range.end(), Size::ZERO))?;
435 /// Uninitialized bytes.
436 impl<Tag: Copy, Extra> Allocation<Tag, Extra> {
437 /// Checks whether the given range is entirely initialized.
439 /// Returns `Ok(())` if it's initialized. Otherwise returns the range of byte
440 /// indexes of the first contiguous uninitialized access.
441 fn is_init(&self, range: AllocRange) -> Result<(), Range<Size>> {
442 self.init_mask.is_range_initialized(range.start, range.end()) // `Size` addition
445 /// Checks that a range of bytes is initialized. If not, returns the `InvalidUninitBytes`
446 /// error which will report the first range of bytes which is uninitialized.
447 fn check_init(&self, range: AllocRange) -> AllocResult {
448 self.is_init(range).or_else(|idx_range| {
449 Err(AllocError::InvalidUninitBytes(Some(UninitBytesAccess {
450 access_offset: range.start,
451 access_size: range.size,
452 uninit_offset: idx_range.start,
453 uninit_size: idx_range.end - idx_range.start, // `Size` subtraction
458 pub fn mark_init(&mut self, range: AllocRange, is_init: bool) {
459 if range.size.bytes() == 0 {
462 self.init_mask.set_range(range.start, range.end(), is_init);
466 /// Run-length encoding of the uninit mask.
467 /// Used to copy parts of a mask multiple times to another allocation.
468 pub struct InitMaskCompressed {
469 /// Whether the first range is initialized.
471 /// The lengths of ranges that are run-length encoded.
472 /// The initialization state of the ranges alternate starting with `initial`.
473 ranges: smallvec::SmallVec<[u64; 1]>,
476 impl InitMaskCompressed {
477 pub fn no_bytes_init(&self) -> bool {
478 // The `ranges` are run-length encoded and of alternating initialization state.
479 // So if `ranges.len() > 1` then the second block is an initialized range.
480 !self.initial && self.ranges.len() == 1
484 /// Transferring the initialization mask to other allocations.
485 impl<Tag, Extra> Allocation<Tag, Extra> {
486 /// Creates a run-length encoding of the initialization mask.
487 pub fn compress_uninit_range(&self, src: Pointer<Tag>, size: Size) -> InitMaskCompressed {
488 // Since we are copying `size` bytes from `src` to `dest + i * size` (`for i in 0..repeat`),
489 // a naive initialization mask copying algorithm would repeatedly have to read the initialization mask from
490 // the source and write it to the destination. Even if we optimized the memory accesses,
491 // we'd be doing all of this `repeat` times.
492 // Therefore we precompute a compressed version of the initialization mask of the source value and
493 // then write it back `repeat` times without computing any more information from the source.
495 // A precomputed cache for ranges of initialized / uninitialized bits
496 // 0000010010001110 will become
497 // `[5, 1, 2, 1, 3, 3, 1]`,
498 // where each element toggles the state.
500 let mut ranges = smallvec::SmallVec::<[u64; 1]>::new();
501 let initial = self.init_mask.get(src.offset);
503 let mut cur = initial;
505 for i in 1..size.bytes() {
506 // FIXME: optimize to bitshift the current uninitialized block's bits and read the top bit.
507 if self.init_mask.get(src.offset + Size::from_bytes(i)) == cur {
510 ranges.push(cur_len);
516 ranges.push(cur_len);
518 InitMaskCompressed { ranges, initial }
521 /// Applies multiple instances of the run-length encoding to the initialization mask.
522 pub fn mark_compressed_init_range(
524 defined: &InitMaskCompressed,
529 // An optimization where we can just overwrite an entire range of initialization
530 // bits if they are going to be uniformly `1` or `0`.
531 if defined.ranges.len() <= 1 {
532 self.init_mask.set_range_inbounds(
534 dest.offset + size * repeat, // `Size` operations
540 for mut j in 0..repeat {
542 j += dest.offset.bytes();
543 let mut cur = defined.initial;
544 for range in &defined.ranges {
547 self.init_mask.set_range_inbounds(
548 Size::from_bytes(old_j),
559 #[derive(Clone, PartialEq, Eq, PartialOrd, Ord, Hash, Debug, TyEncodable, TyDecodable)]
560 pub struct Relocations<Tag = (), Id = AllocId>(SortedMap<Size, (Tag, Id)>);
562 impl<Tag, Id> Relocations<Tag, Id> {
563 pub fn new() -> Self {
564 Relocations(SortedMap::new())
567 // The caller must guarantee that the given relocations are already sorted
568 // by address and contain no duplicates.
569 pub fn from_presorted(r: Vec<(Size, (Tag, Id))>) -> Self {
570 Relocations(SortedMap::from_presorted_elements(r))
574 impl<Tag> Deref for Relocations<Tag> {
575 type Target = SortedMap<Size, (Tag, AllocId)>;
577 fn deref(&self) -> &Self::Target {
582 impl<Tag> DerefMut for Relocations<Tag> {
583 fn deref_mut(&mut self) -> &mut Self::Target {
588 /// A partial, owned list of relocations to transfer into another allocation.
589 pub struct AllocationRelocations<Tag> {
590 relative_relocations: Vec<(Size, (Tag, AllocId))>,
593 impl<Tag: Copy, Extra> Allocation<Tag, Extra> {
594 pub fn prepare_relocation_copy(
596 cx: &impl HasDataLayout,
600 ) -> AllocationRelocations<Tag> {
601 let relocations = self.get_relocations(cx, src);
602 if relocations.is_empty() {
603 return AllocationRelocations { relative_relocations: Vec::new() };
607 let mut new_relocations = Vec::with_capacity(relocations.len() * (count as usize));
610 new_relocations.extend(relocations.iter().map(|&(offset, reloc)| {
611 // compute offset for current repetition
612 let dest_offset = dest + size * i; // `Size` operations
614 // shift offsets from source allocation to destination allocation
615 (offset + dest_offset) - src.start, // `Size` operations
621 AllocationRelocations { relative_relocations: new_relocations }
624 /// Applies a relocation copy.
625 /// The affected range, as defined in the parameters to `prepare_relocation_copy` is expected
626 /// to be clear of relocations.
627 pub fn mark_relocation_range(&mut self, relocations: AllocationRelocations<Tag>) {
628 self.relocations.insert_presorted(relocations.relative_relocations);
632 ////////////////////////////////////////////////////////////////////////////////
633 // Uninitialized byte tracking
634 ////////////////////////////////////////////////////////////////////////////////
638 /// A bitmask where each bit refers to the byte with the same index. If the bit is `true`, the byte
639 /// is initialized. If it is `false` the byte is uninitialized.
640 #[derive(Clone, Debug, Eq, PartialEq, PartialOrd, Ord, Hash, TyEncodable, TyDecodable)]
641 #[derive(HashStable)]
642 pub struct InitMask {
648 pub const BLOCK_SIZE: u64 = 64;
650 pub fn new(size: Size, state: bool) -> Self {
651 let mut m = InitMask { blocks: vec![], len: Size::ZERO };
656 /// Checks whether the range `start..end` (end-exclusive) is entirely initialized.
658 /// Returns `Ok(())` if it's initialized. Otherwise returns a range of byte
659 /// indexes for the first contiguous span of the uninitialized access.
661 pub fn is_range_initialized(&self, start: Size, end: Size) -> Result<(), Range<Size>> {
663 return Err(self.len..end);
666 // FIXME(oli-obk): optimize this for allocations larger than a block.
667 let idx = (start.bytes()..end.bytes()).map(Size::from_bytes).find(|&i| !self.get(i));
671 let uninit_end = (idx.bytes()..end.bytes())
672 .map(Size::from_bytes)
673 .find(|&i| self.get(i))
681 pub fn set_range(&mut self, start: Size, end: Size, new_state: bool) {
684 self.grow(end - len, new_state);
686 self.set_range_inbounds(start, end, new_state);
689 pub fn set_range_inbounds(&mut self, start: Size, end: Size, new_state: bool) {
690 let (blocka, bita) = bit_index(start);
691 let (blockb, bitb) = bit_index(end);
692 if blocka == blockb {
693 // First set all bits except the first `bita`,
694 // then unset the last `64 - bitb` bits.
695 let range = if bitb == 0 {
698 (u64::MAX << bita) & (u64::MAX >> (64 - bitb))
701 self.blocks[blocka] |= range;
703 self.blocks[blocka] &= !range;
707 // across block boundaries
709 // Set `bita..64` to `1`.
710 self.blocks[blocka] |= u64::MAX << bita;
711 // Set `0..bitb` to `1`.
713 self.blocks[blockb] |= u64::MAX >> (64 - bitb);
715 // Fill in all the other blocks (much faster than one bit at a time).
716 for block in (blocka + 1)..blockb {
717 self.blocks[block] = u64::MAX;
720 // Set `bita..64` to `0`.
721 self.blocks[blocka] &= !(u64::MAX << bita);
722 // Set `0..bitb` to `0`.
724 self.blocks[blockb] &= !(u64::MAX >> (64 - bitb));
726 // Fill in all the other blocks (much faster than one bit at a time).
727 for block in (blocka + 1)..blockb {
728 self.blocks[block] = 0;
734 pub fn get(&self, i: Size) -> bool {
735 let (block, bit) = bit_index(i);
736 (self.blocks[block] & (1 << bit)) != 0
740 pub fn set(&mut self, i: Size, new_state: bool) {
741 let (block, bit) = bit_index(i);
742 self.set_bit(block, bit, new_state);
746 fn set_bit(&mut self, block: usize, bit: usize, new_state: bool) {
748 self.blocks[block] |= 1 << bit;
750 self.blocks[block] &= !(1 << bit);
754 pub fn grow(&mut self, amount: Size, new_state: bool) {
755 if amount.bytes() == 0 {
758 let unused_trailing_bits =
759 u64::try_from(self.blocks.len()).unwrap() * Self::BLOCK_SIZE - self.len.bytes();
760 if amount.bytes() > unused_trailing_bits {
761 let additional_blocks = amount.bytes() / Self::BLOCK_SIZE + 1;
763 // FIXME(oli-obk): optimize this by repeating `new_state as Block`.
764 iter::repeat(0).take(usize::try_from(additional_blocks).unwrap()),
767 let start = self.len;
769 self.set_range_inbounds(start, start + amount, new_state); // `Size` operation
774 fn bit_index(bits: Size) -> (usize, usize) {
775 let bits = bits.bytes();
776 let a = bits / InitMask::BLOCK_SIZE;
777 let b = bits % InitMask::BLOCK_SIZE;
778 (usize::try_from(a).unwrap(), usize::try_from(b).unwrap())