1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804 805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845 846 847 848 849 850 851 852 853 854 855 856 857 858 859 860 861 862 863 864 865 866 867 868 869 870 871 872 873 874 875 876 877 878 879 880 881 882 883 884 885 886 887 888 889 890 891 892 893 894 895 896 897 898 899 900 901 902 903 904 905 906 907 908 909 910 911 912 913 914 915 916 917 918 919 920 921 922 923 924 925 926 927 928 929 930 931 932 933 934 935 936 937 938
//! Provides an allocator for virtual memory pages.
//! The minimum unit of allocation is a single page.
//!
//! This also supports early allocation of pages (up to 32 separate chunks)
//! before heap allocation is available, and does so behind the scenes using the same single interface.
//!
//! Once heap allocation is available, it uses a dynamically-allocated list of page chunks to track allocations.
//!
//! The core allocation function is [`allocate_pages_deferred()`](fn.allocate_pages_deferred.html),
//! but there are several convenience functions that offer simpler interfaces for general usage.
//!
//! # Notes
//! This allocator only makes one attempt to merge deallocated pages into existing
//! free chunks for de-fragmentation. It does not iteratively merge adjacent chunks in order to
//! maximally combine separate chunks into the biggest single chunk.
//! Instead, free chunks are lazily merged only when running out of address space
//! or when needed to fulfill a specific request.
#![no_std]
extern crate alloc;
#[macro_use] extern crate log;
extern crate kernel_config;
extern crate memory_structs;
extern crate spin;
#[macro_use] extern crate static_assertions;
extern crate intrusive_collections;
use intrusive_collections::Bound;
mod static_array_rb_tree;
// mod static_array_linked_list;
use core::{borrow::Borrow, cmp::{Ordering, max, min}, fmt, ops::{Deref, DerefMut}};
use kernel_config::memory::*;
use memory_structs::{VirtualAddress, Page, PageRange, PageSize, Page4K, Page2M, Page1G};
use spin::{Mutex, Once};
use static_array_rb_tree::*;
/// Certain regions are pre-designated for special usage, specifically the kernel's initial identity mapping.
/// They will be allocated from if an address within them is specifically
/// otherwise, they will only be allocated from as a "last resort" if all other non-designated address ranges are exhausted.
///
/// Any virtual addresses **less than or equal** to this address are considered "designated".
/// This lower part of the address range that's designated covers from 0x0 to this address.
static DESIGNATED_PAGES_LOW_END: Once<Page> = Once::new();
/// Defines the upper part of the address space that's designated, similar to `DESIGNATED_PAGES_LOW_END`.
/// Any virtual addresses **greater than or equal to** this address is considered "designated".
/// This higher part of the address range covers from:
/// the beginning of the recursive P4 entry used for modifying upcoming page tables
/// to the very end of the address space.
///
/// TODO: once the heap is fully dynamic and not dependent on static addresses,
/// we can exclude the heap from the designated region.
static DESIGNATED_PAGES_HIGH_START: Page = Page::containing_address(
VirtualAddress::new_canonical(UPCOMING_PAGE_TABLE_RECURSIVE_P4_START)
);
const MIN_PAGE: Page = Page::containing_address(VirtualAddress::zero());
const MAX_PAGE: Page = Page::containing_address(VirtualAddress::new_canonical(MAX_VIRTUAL_ADDRESS));
/// The single, system-wide list of free chunks of virtual memory pages.
static FREE_PAGE_LIST: Mutex<StaticArrayRBTree<Chunk>> = Mutex::new(StaticArrayRBTree::empty());
/// Initialize the page allocator.
///
/// # Arguments
/// * `end_vaddr_of_low_designated_region`: the `VirtualAddress` that marks the end of the
/// lower designated region, which should be the ending address of the initial kernel image
/// (a lower-half identity address).
///
/// The page allocator considers two regions as "designated" regions. It will only allocate pages
/// within these designated regions if the specifically-requested address falls within them.
/// 1. The lower designated region is for identity-mapped bootloader content
/// and base kernel image sections, which is used during OS initialization.
/// 2. The higher designated region is for the same content, mapped to the higher half
/// of the address space. It also excludes the address ranges for the P4 entries that
/// Theseus uses for recursive page table mapping.
/// * See [`RECURSIVE_P4_INDEX`] and [`UPCOMING_PAGE_TABLE_RECURSIVE_P4_INDEX`].
///
/// General allocation requests for pages at any virtual address will not use
/// addresses within designated regions unless the entire address space is already in use,
/// which is an extraordinarily unlikely (i.e., basically impossible) situation.
pub fn init(end_vaddr_of_low_designated_region: VirtualAddress) -> Result<(), &'static str> {
assert!(end_vaddr_of_low_designated_region < DESIGNATED_PAGES_HIGH_START.start_address());
let designated_low_end_page = DESIGNATED_PAGES_LOW_END.call_once(
|| Page::containing_address(end_vaddr_of_low_designated_region)
);
let designated_low_end = *designated_low_end_page;
let initial_free_chunks = [
// The first region contains all pages from address zero to the end of the low designated region,
// which is generally reserved for identity-mapped bootloader stuff and base kernel image sections.
Some(Chunk {
pages: PageRange::new(
Page::containing_address(VirtualAddress::zero()),
designated_low_end,
)
}),
// The second region contains the massive range from the end of the low designated region
// to the beginning of the high designated region, which comprises the majority of the address space.
// The beginning of the high designated region starts at the reserved P4 entry used to
// recursively map the "upcoming" page table (i.e., UPCOMING_PAGE_TABLE_RECURSIVE_P4_INDEX).
Some(Chunk {
pages: PageRange::new(
designated_low_end + 1,
DESIGNATED_PAGES_HIGH_START - 1,
)
}),
// Here, we skip the addresses covered by the `UPCOMING_PAGE_TABLE_RECURSIVE_P4_INDEX`.
// The third region contains the range of addresses reserved for the heap,
// which ends at the beginning of the addresses covered by the `RECURSIVE_P4_INDEX`,
Some(Chunk {
pages: PageRange::new(
Page::containing_address(VirtualAddress::new_canonical(KERNEL_HEAP_START)),
// This is the page right below the beginning of the 510th entry of the top-level P4 page table.
Page::containing_address(VirtualAddress::new_canonical(RECURSIVE_P4_START - 1)),
)
}),
// Here, we skip the addresses covered by the `RECURSIVE_P4_INDEX`.
// The fourth region contains all pages in the 511th (last) entry of P4.
Some(Chunk {
pages: PageRange::new(
Page::containing_address(VirtualAddress::new_canonical(KERNEL_TEXT_START)),
MAX_PAGE,
)
}),
None, None, None, None,
None, None, None, None, None, None, None, None,
None, None, None, None, None, None, None, None,
None, None, None, None, None, None, None, None,
];
*FREE_PAGE_LIST.lock() = StaticArrayRBTree::new(initial_free_chunks);
Ok(())
}
/// A range of contiguous 4K-sized pages.
///
/// # Ordering and Equality
///
/// `Chunk` implements the `Ord` trait, and its total ordering is ONLY based on
/// its **starting** `Page`. This is useful so we can store `Chunk`s in a sorted collection.
///
/// Similarly, `Chunk` implements equality traits, `Eq` and `PartialEq`,
/// both of which are also based ONLY on the **starting** `Page` of the `Chunk`.
/// Thus, comparing two `Chunk`s with the `==` or `!=` operators may not work as expected.
/// since it ignores their actual range of pages.
#[derive(Debug, Clone, Eq)]
struct Chunk {
/// The Pages covered by this chunk, an inclusive range.
pages: PageRange<Page4K>,
}
impl Chunk {
fn as_allocated_pages(&self) -> AllocatedPages<Page4K> {
AllocatedPages::<Page4K> {
pages: self.pages.clone(),
}
}
/// Returns a new `Chunk` with an empty range of pages.
fn empty() -> Chunk {
Chunk {
pages: PageRange::<Page4K>::empty(),
}
}
}
impl Deref for Chunk {
type Target = PageRange<Page4K>;
fn deref(&self) -> &PageRange<Page4K> {
&self.pages
}
}
impl Ord for Chunk {
fn cmp(&self, other: &Self) -> Ordering {
self.pages.start().cmp(other.pages.start())
}
}
impl PartialOrd for Chunk {
fn partial_cmp(&self, other: &Self) -> Option<Ordering> {
Some(self.cmp(other))
}
}
impl PartialEq for Chunk {
fn eq(&self, other: &Self) -> bool {
self.pages.start() == other.pages.start()
}
}
impl Borrow<Page<Page4K>> for &'_ Chunk {
fn borrow(&self) -> &Page<Page4K> {
self.pages.start()
}
}
/// Represents a range of allocated `VirtualAddress`es, specified in `Page`s.
///
/// These pages are not initially mapped to any physical memory frames, you must do that separately
/// in order to actually use their memory; see the `MappedPages` type for more.
///
/// This object represents ownership of the allocated virtual pages;
/// if this object falls out of scope, its allocated pages will be auto-deallocated upon drop.
pub struct AllocatedPages<P: PageSize = Page4K> {
pages: PageRange<P>,
}
// AllocatedPages must not be Cloneable, and it must not expose its inner pages as mutable.
assert_not_impl_any!(AllocatedPages<Page4K>: DerefMut, Clone);
assert_not_impl_any!(AllocatedPages<Page2M>: DerefMut, Clone);
assert_not_impl_any!(AllocatedPages<Page1G>: DerefMut, Clone);
impl<P: PageSize> fmt::Debug for AllocatedPages<P> {
fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result {
write!(f, "AllocatedPages({:?})", self.pages)
}
}
impl<P: PageSize> Default for AllocatedPages<P> {
fn default() -> AllocatedPages<P> {
Self::empty()
}
}
impl<P: PageSize> AllocatedPages<P> {
/// Returns an empty AllocatedPages object that performs no page allocation.
/// Can be used as a placeholder, but will not permit any real usage.
pub const fn empty() -> AllocatedPages<P> {
AllocatedPages {
pages: PageRange::<P>::empty()
}
}
/// Returns the starting `VirtualAddress` in this range of pages.
pub const fn start_address(&self) -> VirtualAddress {
self.pages.start_address()
}
/// Returns the size in bytes of this range of pages.
pub const fn size_in_bytes(&self) -> usize {
self.pages.size_in_bytes()
}
/// Returns the size in number of pages of this range of pages.
pub const fn size_in_pages(&self) -> usize {
self.pages.size_in_pages()
}
/// Returns the starting `Page` in this range of pages.
pub const fn start(&self) -> &Page<P> {
self.pages.start()
}
/// Returns the ending `Page` (inclusive) in this range of pages.
pub const fn end(&self) -> &Page<P> {
self.pages.end()
}
/// Returns a reference to the inner `PageRange`, which is cloneable/iterable.
pub const fn range(&self) -> &PageRange<P> {
&self.pages
}
/// Returns the offset of the given `VirtualAddress` within this range of pages,
/// i.e., `addr - self.start_address()`.
///
/// If the given `addr` is not covered by this range of pages, this returns `None`.
///
/// ## Examples
/// If the range covers addresses `0x2000` to `0x4000`,
/// then `offset_of_address(0x3500)` would return `Some(0x1500)`.
pub const fn offset_of_address(&self, addr: VirtualAddress) -> Option<usize> {
self.pages.offset_of_address(addr)
}
/// Returns the `VirtualAddress` at the given offset into this range of pages,
/// i.e., `self.start_address() + offset`.
///
/// If the given `offset` is not within this range of pages, this returns `None`.
///
/// ## Examples
/// If the range covers addresses `0x2000` through `0x3FFF`,
/// then `address_at_offset(0x1500)` would return `Some(0x3500)`,
/// and `address_at_offset(0x2000)` would return `None`.
pub const fn address_at_offset(&self, offset: usize) -> Option<VirtualAddress> {
self.pages.address_at_offset(offset)
}
/// Merges the given `AllocatedPages` object `ap` into this `AllocatedPages` object (`self`).
/// This is just for convenience and usability purposes, it performs no allocation or remapping.
///
/// The `ap` must be virtually contiguous and come immediately after `self`,
/// that is, `self.end` must equal `ap.start`.
/// If this condition is met, `self` is modified and `Ok(())` is returned,
/// otherwise `Err(ap)` is returned.
pub fn merge(&mut self, ap: AllocatedPages<P>) -> Result<(), AllocatedPages<P>> {
// make sure the pages are contiguous
if *ap.start() != (*self.end() + 1) {
return Err(ap);
}
self.pages = PageRange::<P>::new(*self.start(), *ap.end());
// ensure the now-merged AllocatedPages doesn't run its drop handler and free its pages.
core::mem::forget(ap);
Ok(())
}
/// Splits this `AllocatedPages` into two separate `AllocatedPages` objects:
/// * `[beginning : at_page - 1]`
/// * `[at_page : end]`
///
/// This function follows the behavior of [`core::slice::split_at()`],
/// thus, either one of the returned `AllocatedPages` objects may be empty.
/// * If `at_page == self.start`, the first returned `AllocatedPages` object will be empty.
/// * If `at_page == self.end + 1`, the second returned `AllocatedPages` object will be empty.
///
/// Returns an `Err` containing this `AllocatedPages` if `at_page` is otherwise out of bounds.
///
/// [`core::slice::split_at()`]: https://doc.rust-lang.org/core/primitive.slice.html#method.split_at
pub fn split(
self,
at_page: Page<P>,
) -> Result<(AllocatedPages<P>, AllocatedPages<P>), AllocatedPages<P>> {
let end_of_first = at_page - 1;
let (first, second) = if at_page == *self.start() && at_page <= *self.end() {
let first = PageRange::<P>::empty();
let second = PageRange::<P>::new(at_page, *self.end());
(first, second)
}
else if at_page == (*self.end() + 1) && end_of_first >= *self.start() {
let first = PageRange::<P>::new(*self.start(), *self.end());
let second = PageRange::<P>::empty();
(first, second)
}
else if at_page > *self.start() && end_of_first <= *self.end() {
let first = PageRange::<P>::new(*self.start(), end_of_first);
let second = PageRange::<P>::new(at_page, *self.end());
(first, second)
}
else {
return Err(self);
};
// ensure the original AllocatedPages doesn't run its drop handler and free its pages.
core::mem::forget(self);
Ok((
AllocatedPages::<P> { pages: first },
AllocatedPages::<P> { pages: second },
))
}
}
impl<P: PageSize> Drop for AllocatedPages<P> {
fn drop(&mut self) {
if self.size_in_pages() == 0 { return; }
// trace!("page_allocator: deallocating {:?}", self);
let chunk = Chunk {
pages: self.pages.clone().into_4k_pages(),
};
let mut list = FREE_PAGE_LIST.lock();
match &mut list.0 {
// For early allocations, just add the deallocated chunk to the free pages list.
Inner::Array(_) => {
if list.insert(chunk).is_ok() {
return;
}
}
// For full-fledged deallocations, use the entry API to efficiently determine if
// we can merge the deallocated pages with an existing contiguously-adjactent chunk
// or if we need to insert a new chunk.
Inner::RBTree(ref mut tree) => {
let mut cursor_mut = tree.lower_bound_mut(Bound::Included(chunk.start()));
if let Some(next_chunk) = cursor_mut.get() {
if *chunk.end() + 1 == *next_chunk.start() {
// trace!("Prepending {:?} onto beg of next {:?}", chunk, next_chunk.deref());
if cursor_mut.replace_with(Wrapper::new_link(Chunk {
pages: PageRange::new(*chunk.start(), *next_chunk.end()),
})).is_ok() {
return;
}
}
}
if let Some(prev_chunk) = cursor_mut.peek_prev().get() {
if *prev_chunk.end() + 1 == *chunk.start() {
// trace!("Appending {:?} onto end of prev {:?}", chunk, prev_chunk.deref());
let new_page_range = PageRange::new(*prev_chunk.start(), *chunk.end());
cursor_mut.move_prev();
if cursor_mut.replace_with(Wrapper::new_link(Chunk {
pages: new_page_range,
})).is_ok() {
return;
}
}
}
// trace!("Inserting new chunk for deallocated {:?} ", chunk.pages);
cursor_mut.insert(Wrapper::new_link(chunk));
return;
}
}
log::error!("BUG: couldn't insert deallocated {:?} into free page list", self.pages);
}
}
/// A series of pending actions related to page allocator bookkeeping,
/// which may result in heap allocation.
///
/// The actions are triggered upon dropping this struct.
/// This struct can be returned from the `allocate_pages()` family of functions
/// in order to allow the caller to precisely control when those actions
/// that may result in heap allocation should occur.
/// Such actions include adding chunks to lists of free pages or pages in use.
///
/// The vast majority of use cases don't care about such precise control,
/// so you can simply drop this struct at any time or ignore it
/// with a `let _ = ...` binding to instantly drop it.
pub struct DeferredAllocAction<'list> {
/// A reference to the list into which we will insert the free `Chunk`s.
free_list: &'list Mutex<StaticArrayRBTree<Chunk>>,
/// A free chunk that needs to be added back to the free list.
free1: Chunk,
/// Another free chunk that needs to be added back to the free list.
free2: Chunk,
}
impl<'list> DeferredAllocAction<'list> {
fn new<F1, F2>(free1: F1, free2: F2) -> DeferredAllocAction<'list>
where F1: Into<Option<Chunk>>,
F2: Into<Option<Chunk>>,
{
let free_list = &FREE_PAGE_LIST;
let free1 = free1.into().unwrap_or_else(Chunk::empty);
let free2 = free2.into().unwrap_or_else(Chunk::empty);
DeferredAllocAction { free_list, free1, free2 }
}
}
impl<'list> Drop for DeferredAllocAction<'list> {
fn drop(&mut self) {
// Insert all of the chunks, both allocated and free ones, into the list.
if self.free1.size_in_pages() > 0 {
self.free_list.lock().insert(self.free1.clone()).unwrap();
}
if self.free2.size_in_pages() > 0 {
self.free_list.lock().insert(self.free2.clone()).unwrap();
}
}
}
/// Possible errors returned by the page allocator.
#[derive(Debug)]
pub enum AllocationError {
/// The requested address was not free: it was already allocated, or is outside the range of this allocator.
AddressNotFree(Page<Page4K>, usize),
/// The address space was full, or there was not a large-enough chunk
/// or enough remaining chunks (within the given `PageRange`, if any)
/// that could satisfy the requested allocation size.
OutOfAddressSpace(usize, Option<PageRange<Page4K>>),
/// The allocator has not yet been initialized.
NotInitialized,
}
impl From<AllocationError> for &'static str {
fn from(alloc_err: AllocationError) -> &'static str {
match alloc_err {
AllocationError::AddressNotFree(..) => "address was in use or outside of this page allocator's range",
AllocationError::OutOfAddressSpace(_, Some(_range)) => "out of virtual address space in specified range",
AllocationError::OutOfAddressSpace(_, None) => "out of virtual address space",
AllocationError::NotInitialized => "the page allocator has not yet been initialized",
}
}
}
/// Searches the given `list` for the chunk that contains the range of pages from
/// `requested_page` to `requested_page + num_pages`.
fn find_specific_chunk(
list: &mut StaticArrayRBTree<Chunk>,
requested_page: Page<Page4K>,
num_pages: usize
) -> Result<(AllocatedPages<Page4K>, DeferredAllocAction<'static>), AllocationError> {
// The end page is an inclusive bound, hence the -1. Parentheses are needed to avoid overflow.
let requested_end_page = requested_page + (num_pages - 1);
match &mut list.0 {
Inner::Array(ref mut arr) => {
for elem in arr.iter_mut() {
if let Some(chunk) = elem {
if requested_page >= *chunk.start() && requested_end_page <= *chunk.end() {
// Here: `chunk` was big enough and did contain the requested address.
return adjust_chosen_chunk(requested_page, num_pages, &chunk.clone(), ValueRefMut::Array(elem));
}
}
}
}
Inner::RBTree(ref mut tree) => {
let mut cursor_mut = tree.upper_bound_mut(Bound::Included(&requested_page));
if let Some(chunk) = cursor_mut.get().map(|w| w.deref()) {
if requested_page >= *chunk.start() {
if requested_end_page <= *chunk.end() {
return adjust_chosen_chunk(requested_page, num_pages, &chunk.clone(), ValueRefMut::RBTree(cursor_mut));
} else {
// Here, we've found a chunk that includes the requested start page, but it's too small
// to cover the number of requested pages.
// Thus, we attempt to merge this chunk with the next contiguous chunk(s) to create one single larger chunk.
let chunk = chunk.clone(); // ends the above borrow on `cursor_mut`
let mut new_end_page = *chunk.end();
cursor_mut.move_next();
while let Some(next_chunk) = cursor_mut.get().map(|w| w.deref()) {
if *next_chunk.start() - 1 == new_end_page {
new_end_page = *next_chunk.end();
cursor_mut.remove().expect("BUG: page_allocator failed to merge contiguous chunks.");
// The above call to `cursor_mut.remove()` advances the cursor to the next chunk.
} else {
break; // the next chunk wasn't contiguous, so stop iterating.
}
}
if new_end_page > *chunk.end() {
cursor_mut.move_prev(); // move the cursor back to the original chunk
let _removed_chunk = cursor_mut.replace_with(Wrapper::new_link(Chunk { pages: PageRange::new(*chunk.start(), new_end_page) }))
.expect("BUG: page_allocator failed to replace the current chunk while merging contiguous chunks.");
return adjust_chosen_chunk(requested_page, num_pages, &chunk, ValueRefMut::RBTree(cursor_mut));
}
}
}
}
}
}
Err(AllocationError::AddressNotFree(requested_page, num_pages))
}
/// Searches the given `list` for any chunk large enough to hold at least `num_pages`.
///
/// If a given range is specified, the returned `AllocatedPages` *must* exist
/// fully within that inclusive range of pages.
///
/// If no range is specified, this function first attempts to find a suitable chunk
/// that is **not** within the designated regions,
/// and only allocates from the designated regions as a backup option.
///
/// If an alignment is specified (in terms of number of 4KiB pages), then the starting page
/// in the allocated range must be aligned to that number of pages.
/// If no specific alignment is needed, the default aligment of 1 page should be used.
fn find_any_chunk(
list: &mut StaticArrayRBTree<Chunk>,
num_pages: usize,
within_range: Option<&PageRange<Page4K>>,
alignment_4k_pages: usize,
) -> Result<(AllocatedPages, DeferredAllocAction<'static>), AllocationError> {
let designated_low_end = DESIGNATED_PAGES_LOW_END.get()
.ok_or(AllocationError::NotInitialized)?;
let full_range = PageRange::<Page4K>::new(*designated_low_end + 1, DESIGNATED_PAGES_HIGH_START - 1);
let range = within_range.unwrap_or(&full_range);
// During the first pass, we only search within the given range.
// If no range was given, we search from the end of the low designated region
// to the start of the high designated region.
match list.0 {
Inner::Array(ref mut arr) => {
for elem in arr.iter_mut() {
if let Some(chunk) = elem {
// Use max and min below to ensure that the range of pages we allocate from
// is within *both* the current chunk's bounds and the range's bounds.
let lowest_possible_start_page = max(chunk.start(), range.start())
.align_up(alignment_4k_pages);
let highest_possible_end_page = *min(chunk.end(), range.end());
if lowest_possible_start_page + num_pages <= highest_possible_end_page {
return adjust_chosen_chunk(
lowest_possible_start_page,
num_pages,
&chunk.clone(),
ValueRefMut::Array(elem),
);
}
// The early static array is not sorted, so we must iterate over all elements.
}
}
}
Inner::RBTree(ref mut tree) => {
// NOTE: if RBTree had a `range_mut()` method, we could simply do the following:
// ```
// let eligible_chunks = tree.range_mut(
// Bound::Included(range.start()),
// Bound::Included(range.end())
// );
// for c in eligible_chunks { ... }
// ```
//
// However, RBTree doesn't have a `range_mut()` method, so we use cursors for manual iteration.
//
// Because we allocate new pages by peeling them off from the beginning part of a chunk,
// it's MUCH faster to start the search for free pages from higher addresses moving down.
// This results in an O(1) allocation time in the general case, until all address ranges are already in use.
let mut cursor = tree.upper_bound_mut(Bound::Included(range.end()));
while let Some(chunk) = cursor.get().map(|w| w.deref()) {
// Use max and min below to ensure that the range of pages we allocate from
// is within *both* the current chunk's bounds and the range's bounds.
let lowest_possible_start_page = max(chunk.start(), range.start())
.align_up(alignment_4k_pages);
let highest_possible_end_page = *min(chunk.end(), range.end());
if lowest_possible_start_page + num_pages <= highest_possible_end_page {
return adjust_chosen_chunk(
lowest_possible_start_page,
num_pages,
&chunk.clone(),
ValueRefMut::RBTree(cursor)
);
}
if chunk.start() <= range.start() {
break; // move on to searching through the designated regions
}
warn!("page_allocator: unlikely scenario: had to search multiple chunks while trying to allocate {} pages in {:?}.", num_pages, range);
cursor.move_prev();
}
}
}
// If we failed to find suitable pages within the given range, return an error.
if let Some(range) = within_range {
return Err(AllocationError::OutOfAddressSpace(num_pages, Some(range.clone())));
}
// If we can't find any suitable chunks in the non-designated regions, then look in both designated regions.
warn!("PageAllocator: unlikely scenario: non-designated chunks are all allocated, \
falling back to allocating {} pages from designated regions!", num_pages);
match list.0 {
Inner::Array(ref mut arr) => {
for elem in arr.iter_mut() {
if let Some(chunk) = elem {
let lowest_possible_start_page = chunk.start().align_up(alignment_4k_pages);
if lowest_possible_start_page + num_pages <= *chunk.end() {
return adjust_chosen_chunk(
lowest_possible_start_page,
num_pages,
&chunk.clone(),
ValueRefMut::Array(elem),
);
}
}
}
}
Inner::RBTree(ref mut tree) => {
// NOTE: if RBTree had a `range_mut()` method, we could simply do the following:
// ```
// let eligible_chunks = tree.range(
// Bound::<&Page>::Unbounded,
// Bound::Included(&DESIGNATED_PAGES_LOW_END)
// ).chain(tree.range(
// Bound::Included(&DESIGNATED_PAGES_HIGH_START),
// Bound::<&Page>::Unbounded
// ));
// for c in eligible_chunks { ... }
// ```
//
// RBTree doesn't have a `range_mut()` method, so we use cursors for two rounds of iteration.
// The first iterates over the lower designated region, from higher addresses to lower, down to zero.
let mut cursor = tree.upper_bound_mut(Bound::Included(designated_low_end));
while let Some(chunk) = cursor.get().map(|w| w.deref()) {
let lowest_possible_start_page = chunk.start().align_up(alignment_4k_pages);
if lowest_possible_start_page + num_pages <= *chunk.end() {
return adjust_chosen_chunk(
lowest_possible_start_page,
num_pages,
&chunk.clone(),
ValueRefMut::RBTree(cursor),
);
}
cursor.move_prev();
}
// The second iterates over the higher designated region, from the highest (max) address down to the designated region boundary.
let mut cursor = tree.upper_bound_mut::<Chunk>(Bound::Unbounded);
while let Some(chunk) = cursor.get().map(|w| w.deref()) {
if chunk.start() < &DESIGNATED_PAGES_HIGH_START {
// we already iterated over non-designated pages in the first match statement above, so we're out of memory.
break;
}
let lowest_possible_start_page = chunk.start().align_up(alignment_4k_pages);
if lowest_possible_start_page + num_pages <= *chunk.end() {
return adjust_chosen_chunk(
lowest_possible_start_page,
num_pages,
&chunk.clone(),
ValueRefMut::RBTree(cursor),
);
}
cursor.move_prev();
}
}
}
Err(AllocationError::OutOfAddressSpace(num_pages, None))
}
/// The final part of the main allocation routine.
///
/// The given chunk is the one we've chosen to allocate from.
/// This function breaks up that chunk into multiple ones and returns an `AllocatedPages`
/// from (part of) that chunk, ranging from `start_page` to `start_page + num_pages`.
fn adjust_chosen_chunk(
start_page: Page<Page4K>,
num_pages: usize,
chosen_chunk: &Chunk,
mut chosen_chunk_ref: ValueRefMut<Chunk>,
) -> Result<(AllocatedPages<Page4K>, DeferredAllocAction<'static>), AllocationError> {
// The new allocated chunk might start in the middle of an existing chunk,
// so we need to break up that existing chunk into 3 possible chunks: before, newly-allocated, and after.
//
// Because Pages and VirtualAddresses use saturating add and subtract, we need to double-check that we're not creating
// an overlapping duplicate Chunk at either the very minimum or the very maximum of the address space.
let new_allocation = Chunk {
// The end page is an inclusive bound, hence the -1. Parentheses are needed to avoid overflow.
pages: PageRange::<Page4K>::new(start_page, start_page + (num_pages - 1)),
};
let before = if start_page == MIN_PAGE {
None
} else {
Some(Chunk {
pages: PageRange::<Page4K>::new(*chosen_chunk.start(), *new_allocation.start() - 1),
})
};
let after = if new_allocation.end() == &MAX_PAGE {
None
} else {
Some(Chunk {
pages: PageRange::<Page4K>::new(*new_allocation.end() + 1, *chosen_chunk.end()),
})
};
// some sanity checks -- these can be removed or disabled for better performance
if let Some(ref b) = before {
assert!(!new_allocation.contains(b.end()));
assert!(!b.contains(new_allocation.start()));
}
if let Some(ref a) = after {
assert!(!new_allocation.contains(a.start()));
assert!(!a.contains(new_allocation.end()));
}
// Remove the chosen chunk from the free page list.
let _removed_chunk = chosen_chunk_ref.remove();
assert_eq!(Some(chosen_chunk), _removed_chunk.as_ref()); // sanity check
// TODO: Re-use the allocated wrapper if possible, rather than allocate a new one entirely.
// if let RemovedValue::RBTree(Some(wrapper_adapter)) = _removed_chunk { ... }
Ok((
new_allocation.as_allocated_pages(),
DeferredAllocAction::new(before, after),
))
}
/// Possible options when requesting pages from the page allocator.
pub enum AllocationRequest<'r> {
/// The allocated pages must start exactly at the given `VirtualAddress`.
AtVirtualAddress(VirtualAddress),
/// The allocated pages may be located at any virtual address,
/// but the starting page must be aligned to a multiple of `alignment_4k_pages`.
/// An alignment of `1` page is equivalent to specifying no alignment requirement.
///
/// Note: alignment is specified in number of 4KiB pages, not number of bytes.
AlignedTo { alignment_4k_pages: usize },
/// The allocated pages can be located anywhere within the given range.
WithinRange(&'r PageRange<Page4K>),
/// The allocated pages can be located at any virtual address
/// and have no special alignment requirements beyond a single page.
Any,
}
/// The core page allocation routine that allocates the given number of virtual pages,
/// optionally at the requested starting `VirtualAddress`.
///
/// This simply reserves a range of virtual addresses, it does not allocate
/// actual physical memory frames nor do any memory mapping.
/// Thus, the returned `AllocatedPages` aren't directly usable until they are mapped to physical frames.
///
/// Allocation is based on a red-black tree and is thus `O(log(n))`.
/// Fragmentation isn't cleaned up until we're out of address space, but that's not really a big deal.
///
/// # Arguments
/// * `request`: whether to allocate `num_pages` pages at any address,
/// at a specific virtual address, or withing a specified range.
/// * `num_pages`: the number of `Page`s to be allocated.
///
/// # Return
/// If successful, returns a tuple of two items:
/// * the pages that were allocated, and
/// * an opaque struct representing details of bookkeeping-related actions that may cause heap allocation.
/// Those actions are deferred until this returned `DeferredAllocAction` struct object is dropped,
/// allowing the caller (such as the heap implementation itself) to control when heap allocation may occur.
pub fn allocate_pages_deferred(
request: AllocationRequest,
num_pages: usize,
) -> Result<(AllocatedPages<Page4K>, DeferredAllocAction<'static>), &'static str> {
if num_pages == 0 {
warn!("PageAllocator: requested an allocation of 0 pages... stupid!");
return Err("cannot allocate zero pages");
}
let mut locked_list = FREE_PAGE_LIST.lock();
// The main logic of the allocator is to find an appropriate chunk that can satisfy the allocation request.
// An appropriate chunk satisfies the following conditions:
// - Can fit the requested size (starting at the requested address) within the chunk.
// - The chunk can only be within in a designated region if a specific address was requested,
// or all other non-designated chunks are already in use.
let res = match request {
AllocationRequest::AtVirtualAddress(vaddr) => {
find_specific_chunk(&mut locked_list, Page::containing_address(vaddr), num_pages)
}
AllocationRequest::AlignedTo { alignment_4k_pages } => {
find_any_chunk(&mut locked_list, num_pages, None, alignment_4k_pages)
}
AllocationRequest::WithinRange(range) => {
find_any_chunk(&mut locked_list, num_pages, Some(range), 1)
}
AllocationRequest::Any => {
find_any_chunk(&mut locked_list, num_pages, None, 1)
}
};
res.map_err(From::from) // convert from AllocationError to &str
}
/// Similar to [`allocated_pages_deferred()`](fn.allocate_pages_deferred.html),
/// but accepts a size value for the allocated pages in number of bytes instead of number of pages.
///
/// This function still allocates whole pages by rounding up the number of bytes.
pub fn allocate_pages_by_bytes_deferred(
request: AllocationRequest,
num_bytes: usize,
) -> Result<(AllocatedPages<Page4K>, DeferredAllocAction<'static>), &'static str> {
let actual_num_bytes = if let AllocationRequest::AtVirtualAddress(vaddr) = request {
num_bytes + (vaddr.value() % PAGE_SIZE)
} else {
num_bytes
};
let num_pages = (actual_num_bytes + PAGE_SIZE - 1) / PAGE_SIZE; // round up
allocate_pages_deferred(request, num_pages)
}
/// Allocates the given number of pages with no constraints on the starting virtual address.
///
/// See [`allocate_pages_deferred()`](fn.allocate_pages_deferred.html) for more details.
pub fn allocate_pages(num_pages: usize) -> Option<AllocatedPages<Page4K>> {
allocate_pages_deferred(AllocationRequest::Any, num_pages)
.map(|(ap, _action)| ap)
.ok()
}
/// Allocates pages with no constraints on the starting virtual address,
/// with a size given by the number of bytes.
///
/// This function still allocates whole pages by rounding up the number of bytes.
/// See [`allocate_pages_deferred()`](fn.allocate_pages_deferred.html) for more details.
pub fn allocate_pages_by_bytes(num_bytes: usize) -> Option<AllocatedPages<Page4K>> {
allocate_pages_by_bytes_deferred(AllocationRequest::Any, num_bytes)
.map(|(ap, _action)| ap)
.ok()
}
/// Allocates pages starting at the given `VirtualAddress` with a size given in number of bytes.
///
/// This function still allocates whole pages by rounding up the number of bytes.
/// See [`allocate_pages_deferred()`](fn.allocate_pages_deferred.html) for more details.
pub fn allocate_pages_by_bytes_at(vaddr: VirtualAddress, num_bytes: usize) -> Result<AllocatedPages<Page4K>, &'static str> {
allocate_pages_by_bytes_deferred(AllocationRequest::AtVirtualAddress(vaddr), num_bytes)
.map(|(ap, _action)| ap)
}
/// Allocates the given number of pages starting at (inclusive of) the page containing the given `VirtualAddress`.
///
/// See [`allocate_pages_deferred()`](fn.allocate_pages_deferred.html) for more details.
pub fn allocate_pages_at(vaddr: VirtualAddress, num_pages: usize) -> Result<AllocatedPages<Page4K>, &'static str> {
allocate_pages_deferred(AllocationRequest::AtVirtualAddress(vaddr), num_pages)
.map(|(ap, _action)| ap)
}
/// Allocates the given number of pages with the constraint that
/// they must be within the given inclusive `range` of pages.
pub fn allocate_pages_in_range(
num_pages: usize,
range: &PageRange<Page4K>,
) -> Result<AllocatedPages<Page4K>, &'static str> {
allocate_pages_deferred(AllocationRequest::WithinRange(range), num_pages)
.map(|(ap, _action)| ap)
}
/// Allocates pages with a size given in number of bytes with the constraint that
/// they must be within the given inclusive `range` of pages.
pub fn allocate_pages_by_bytes_in_range(
num_bytes: usize,
range: &PageRange<Page4K>,
) -> Result<AllocatedPages<Page4K>, &'static str> {
allocate_pages_by_bytes_deferred(AllocationRequest::WithinRange(range), num_bytes)
.map(|(ap, _action)| ap)
}
/// Converts the page allocator from using static memory (a primitive array) to dynamically-allocated memory.
///
/// Call this function once heap allocation is available.
/// Calling this multiple times is unnecessary but harmless, as it will do nothing after the first invocation.
#[doc(hidden)]
pub fn convert_page_allocator_to_heap_based() {
FREE_PAGE_LIST.lock().convert_to_heap_allocated();
}
/// A debugging function used to dump the full internal state of the page allocator.
#[doc(hidden)]
pub fn dump_page_allocator_state() {
debug!("--------------- FREE PAGES LIST ---------------");
for c in FREE_PAGE_LIST.lock().iter() {
debug!("{:X?}", c);
}
debug!("---------------------------------------------------");
}