diff --git a/Cargo.lock b/Cargo.lock
index 34fdcba..b3c2b21 100644
--- a/Cargo.lock
+++ b/Cargo.lock
@@ -2,6 +2,12 @@
 # It is not intended for manual editing.
 version = 4
 
+[[package]]
+name = "autocfg"
+version = "1.5.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "f2032f911046de80f0a198e0901378627c33f59ea0ac00e363d481118bd70a53"
+
 [[package]]
 name = "bit_field"
 version = "0.10.3"
@@ -14,13 +20,22 @@ version = "2.13.1"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "b588b76d00fde79687d7646a9b5bdf3cc0f655e0bbd080335a95d7e96f3587da"
 
+[[package]]
+name = "foundation"
+version = "0.1.0"
+dependencies = [
+ "num-traits",
+]
+
 [[package]]
 name = "kernel"
 version = "0.1.0"
 dependencies = [
  "bit_field",
  "bitflags",
+ "foundation",
  "log",
+ "num-traits",
  "plain",
  "rbtree",
  "rustc-demangle",
@@ -33,6 +48,15 @@ version = "0.4.33"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "0ceec5bc11778974d1bcb055b18002eba7f4b3518b6a0081b3af5f21666da9ad"
 
+[[package]]
+name = "num-traits"
+version = "0.2.19"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "071dfc062690e90b734c0b2273ce72ad0ffa95f0c74596bc250dcfd960262841"
+dependencies = [
+ "autocfg",
+]
+
 [[package]]
 name = "plain"
 version = "0.2.3"
diff --git a/crates/foundation/Cargo.toml b/crates/foundation/Cargo.toml
new file mode 100644
index 0000000..9ffa7a3
--- /dev/null
+++ b/crates/foundation/Cargo.toml
@@ -0,0 +1,7 @@
+[package]
+name = "foundation"
+version = "0.1.0"
+edition = "2024"
+
+[dependencies]
+num-traits = { version = "0.2.19", default-features = false }
diff --git a/crates/foundation/src/alloc/bump.rs b/crates/foundation/src/alloc/bump.rs
new file mode 100644
index 0000000..e4b3c64
--- /dev/null
+++ b/crates/foundation/src/alloc/bump.rs
@@ -0,0 +1,426 @@
+
+//! A bump allocator inspired by / taken from the `stumpalo` crate
+
+use core::{
+    alloc::{Allocator, Layout},
+    cell::Cell,
+    marker::PhantomData,
+    ops::{Deref, DerefMut},
+    ptr::{self, NonNull},
+};
+
+#[repr(align(16))]
+struct Chunk {
+    next: Option<NonNull<Chunk>>,
+    size: usize,
+}
+
+const EMPTY_CHUNK: Chunk = Chunk {
+    next: None,
+    size: 0,
+};
+
+struct RestorePoint {
+    top: *mut u8,
+    chunk: NonNull<Chunk>,
+}
+
+pub struct Bump<A: Allocator> {
+    top: Cell<*mut u8>,
+    bottom: Cell<*mut Chunk>,
+    next_chunk: Cell<Option<NonNull<Chunk>>>,
+    backing_alloc: A,
+}
+
+unsafe impl<A: Allocator + Send> Send for Bump<A> {}
+
+unsafe impl<A: Allocator> Allocator for BumpScope<'_, '_, A> {
+    fn allocate(
+        &self,
+        layout: Layout,
+    ) -> Result<core::ptr::NonNull<[u8]>, core::alloc::AllocError> {
+        let virt = unsafe { self.alloc_layout(layout).ok_or(core::alloc::AllocError) }?;
+
+        Ok(virt.cast_slice(layout.size()))
+    }
+
+    unsafe fn deallocate(&self, _ptr: core::ptr::NonNull<u8>, _layout: Layout) {}
+}
+
+impl<A: Allocator> Bump<A> {
+    pub fn new_in(backing_alloc: A) -> Self {
+        let chunk = NonNull::from(&EMPTY_CHUNK);
+        let bottom = unsafe { chunk.as_ptr().add(1) };
+
+        Self {
+            top: Cell::new(bottom.cast::<u8>()),
+            bottom: Cell::new(bottom),
+            next_chunk: Cell::new(None),
+            backing_alloc,
+        }
+    }
+
+    /// # Safety
+    /// The caller must ensure that the chunk is aligned to `Chunk` alignment.
+    pub unsafe fn from_raw_chunk_in(chunk: NonNull<[u8]>, backing_alloc: A) -> Self {
+        assert!(
+            chunk.as_ptr().addr().is_multiple_of(align_of::<Chunk>()),
+            "Chunk must be aligned to Chunk alignment"
+        );
+        assert!(
+            chunk.len() >= size_of::<Chunk>(),
+            "Chunk must be at least the size of Chunk"
+        );
+
+        let len = chunk.len();
+        let chunk = chunk.cast::<Chunk>();
+        let bottom = unsafe { chunk.as_ptr().add(1) };
+        let top = unsafe { bottom.byte_add(len - size_of::<Chunk>()) };
+
+        Self {
+            top: Cell::new(top.cast::<u8>()),
+            bottom: Cell::new(bottom),
+            next_chunk: Cell::new(None),
+            backing_alloc,
+        }
+    }
+
+    pub fn as_scope<'env>(&'env mut self) -> BumpScope<'env, 'env, A> {
+        BumpScope {
+            bump: self,
+            _env: PhantomData,
+            _scope: PhantomData,
+        }
+    }
+
+    pub fn scope<'env, F, R>(&'env mut self, f: F) -> R
+    where
+        F: for<'scope> FnOnce(&'scope mut BumpScope<'env, 'scope, A>) -> R + 'env,
+        R: 'env,
+    {
+        let restore = self.restore_point();
+
+        let mut scope = BumpScope {
+            bump: self,
+            _env: PhantomData,
+            _scope: PhantomData,
+        };
+
+        let result = f(&mut scope);
+
+        unsafe { self.restore(restore) };
+
+        result
+    }
+
+    fn restore_point(&self) -> RestorePoint {
+        RestorePoint {
+            top: self.top.get(),
+            chunk: NonNull::from(self.chunk()),
+        }
+    }
+
+    unsafe fn restore(&self, restore_point: RestorePoint) {
+        // get current chunk and next chunk
+        let mut chunk = Some(NonNull::from(self.chunk()));
+        let mut head = self.next_chunk.get();
+
+        // walk the linked list of chunks used since the restore point and
+        // re-link them onto the free list.
+        while let Some(cnk) = chunk
+            && cnk != restore_point.chunk
+        {
+            let next = unsafe { ptr::replace(&raw mut (*cnk.as_ptr()).next, head) };
+
+            head = Some(cnk);
+            chunk = next;
+        }
+
+        self.top.set(restore_point.top);
+        let bot = unsafe { restore_point.chunk.as_ptr().add(1) };
+        self.bottom.set(bot);
+        self.next_chunk.set(head);
+    }
+
+    #[inline]
+    pub unsafe fn alloc_layout(&self, layout: Layout) -> Option<NonNull<u8>> {
+        let top = self.top.get();
+        let bottom = self.bottom.get().addr();
+        let extra = Self::extra_bytes(top, layout.align());
+
+        let slow_path = Self::out_of_mem(layout, top, bottom, false);
+
+        if slow_path {
+            self.alloc_layout_slow_cold(layout)
+        } else {
+            let new_top = unsafe { top.byte_sub(extra).byte_sub(layout.size()) };
+            self.top.set(new_top);
+
+            Some(unsafe { NonNull::new_unchecked(new_top) })
+        }
+    }
+
+    #[inline(never)]
+    fn try_alloc_slow_with_no_inline<F, T>(&self, f: F) -> Option<NonNull<T>>
+    where
+        F: FnOnce() -> T,
+    {
+        let p = self.alloc_layout_slow(Layout::new::<T>())?;
+        let p = p.cast::<T>();
+        unsafe { p.write(f()) };
+        Some(p)
+    }
+
+    #[cold]
+    fn alloc_layout_slow_cold(&self, layout: Layout) -> Option<NonNull<u8>> {
+        self.alloc_layout_slow(layout)
+    }
+
+    // #[inline(never)]
+    // fn alloc_layout_slow_no_inline(&self, layout: Layout) -> Option<NonNull<u8>> {
+    //     self.alloc_layout_slow(layout)
+    // }
+
+    fn alloc_layout_slow(&self, layout: Layout) -> Option<NonNull<u8>> {
+        let extra_with_chunk = {
+            assert!(size_of::<Chunk>() == align_of::<Chunk>());
+            // if the allocation is greater than the size of the chunk header, we need more bytes past the chunk header to align the allocation.
+            if layout.align() > size_of::<Chunk>() {
+                layout.align() - size_of::<Chunk>()
+            } else {
+                0
+            }
+        };
+
+        let min_size = layout.size() + extra_with_chunk;
+
+        while let Some(mut chunk) = self.next_chunk.get() {
+            // SAFETY: `Bump` is not `Sync` and `alloc_layout_slow is not
+            // reentrant, we have exclusive access to `chunk`
+            let chunk = unsafe { chunk.as_mut() };
+            let cap = chunk.size;
+
+            if cap >= min_size {
+                let prev = self.chunk();
+                self.next_chunk.set(chunk.next);
+
+                // add the current chunk to the linked list of chunks, so it can be freed later.
+                chunk.next = Some(NonNull::from(prev));
+
+                unsafe {
+                    // bottom is just past the chunk header
+                    let bot = (&raw const *chunk).add(1).cast_mut();
+                    self.bottom.set(bot);
+
+                    // calculate top of chunk
+                    let top = bot.byte_add(cap).cast::<u8>();
+
+                    // sub allocation
+                    let extra = Self::extra_bytes(top, layout.align());
+                    let top = top.byte_sub(extra).byte_sub(layout.size());
+                    self.top.set(top);
+
+                    return Some(NonNull::new_unchecked(top));
+                };
+            } else {
+                // chunk is too small, try the next one
+                self.next_chunk.set(chunk.next);
+                unsafe {
+                    let layout = Layout::from_size_align_unchecked(
+                        cap + size_of::<Chunk>(),
+                        align_of::<Chunk>(),
+                    );
+                    self.backing_alloc
+                        .deallocate(NonNull::from(chunk).cast(), layout);
+                }
+            }
+        }
+
+        let chunk = self.chunk();
+        let prev = if chunk.size == 0 {
+            None
+        } else {
+            Some(NonNull::from(chunk))
+        };
+
+        let size = self.next_chunk_size().max(min_size);
+        let new_chunk = self
+            .backing_alloc
+            .allocate(unsafe {
+                Layout::from_size_align_unchecked(size + size_of::<Chunk>(), align_of::<Chunk>())
+            })
+            .ok()?
+            .cast::<Chunk>();
+
+        unsafe {
+            new_chunk.write(Chunk { next: prev, size });
+
+            let bot = new_chunk.add(1).as_ptr();
+            self.bottom.set(bot);
+
+            let top = bot.byte_add(size).cast::<u8>();
+            let extra = Self::extra_bytes(top, layout.align());
+            let top = top.byte_sub(extra).byte_sub(layout.size());
+            self.top.set(top);
+
+            Some(NonNull::new_unchecked(top))
+        }
+    }
+
+    fn next_chunk_size(&self) -> usize {
+        let chunk = self.chunk();
+        let cap = (chunk.size + size_of::<Chunk>()).saturating_mul(2) - size_of::<Chunk>();
+
+        cap.max(0x1000)
+    }
+
+    pub unsafe fn alloc_raw<T>(&self) -> Option<NonNull<T>> {
+        let layout = core::alloc::Layout::new::<T>();
+        unsafe { self.alloc_layout(layout).map(|p| p.cast::<T>()) }
+    }
+
+    fn out_of_mem(layout: Layout, top: *mut u8, bottom: usize, comptime: bool) -> bool {
+        let extra = Self::extra_bytes(top, layout.align());
+
+        let max_padding = layout.align() - 1;
+        let max_size = layout.size() + max_padding;
+        let top = top.addr();
+
+        // biggest possible virtual address on x86_64 is 2^57 - 1
+        const MAX_ADDR: usize = ((1u64 << 57) - 1) as usize;
+        // offsets of < SAFE_SIZE are guaranteed not to overflow the address space.
+        const SAFE_SIZE: usize = usize::MAX - MAX_ADDR;
+
+        if comptime && max_size < SAFE_SIZE {
+            if max_size <= 16 {
+                let top = top - extra;
+                let top = top - layout.size();
+                bottom > top
+            } else if max_padding < 16 {
+                let top = top - extra;
+                bottom + layout.size() > top
+            } else {
+                bottom + layout.size() + max_padding > top
+            }
+        } else if max_padding
+            .checked_add(isize::MAX as usize)
+            .is_some_and(|p| p < SAFE_SIZE)
+        {
+            if max_size < 16 {
+                let top = top - extra;
+                bottom + layout.size() > top
+            } else {
+                bottom + extra + layout.size() > top
+            }
+        } else {
+            if max_padding < 16 {
+                let top = top - extra;
+                top.checked_sub(layout.size()).is_none_or(|t| t < bottom)
+            } else {
+                top.checked_sub(extra)
+                    .and_then(|t| t.checked_sub(layout.size()))
+                    .is_none_or(|t| t < bottom)
+            }
+        }
+    }
+
+    fn extra_bytes(ptr: *mut u8, align: usize) -> usize {
+        assert!(align.is_power_of_two(), "Alignment must be a power of two");
+        ptr.addr() & (align - 1)
+    }
+
+    fn chunk(&self) -> &Chunk {
+        unsafe { self.bottom.get().sub(1).as_ref_unchecked() }
+    }
+}
+
+#[repr(transparent)]
+pub struct BumpScope<'env, 'scope, A: Allocator> {
+    bump: &'scope mut Bump<A>,
+    _env: PhantomData<&'env &'env mut ()>,
+    _scope: PhantomData<&'scope &'scope mut ()>,
+}
+
+impl<'env, 'scope, A: Allocator> Deref for BumpScope<'env, 'scope, A> {
+    type Target = Bump<A>;
+
+    fn deref(&self) -> &Self::Target {
+        self.bump
+    }
+}
+
+impl<'env, 'scope, A: Allocator> DerefMut for BumpScope<'env, 'scope, A> {
+    fn deref_mut(&mut self) -> &mut Self::Target {
+        self.bump
+    }
+}
+
+impl<'env, 'scope, A: Allocator> BumpScope<'env, 'scope, A> {
+    pub fn alloc_with<F, T>(&self, f: F) -> &'scope mut T
+    where
+        F: FnOnce() -> T,
+    {
+        self.try_alloc_with(f).unwrap_or_else(|| {
+            panic!(
+                "Bump allocator out of memory when allocating {} bytes with alignment {}",
+                size_of::<T>(),
+                align_of::<T>()
+            )
+        })
+    }
+
+    pub fn try_alloc_with<F, T>(&self, f: F) -> Option<&'scope mut T>
+    where
+        F: FnOnce() -> T,
+    {
+        let layout = Layout::new::<T>();
+        let top = self.top.get();
+        let bottom = self.bottom.get().addr();
+        let extra = Bump::<A>::extra_bytes(top, layout.align());
+
+        let slow_path = Bump::<A>::out_of_mem(layout, top, bottom, false);
+
+        let mut ptr = if layout.size() <= 16 {
+            if slow_path {
+                self.try_alloc_slow_with_no_inline(f)?
+            } else {
+                let new_top = unsafe { top.byte_sub(extra).byte_sub(layout.size()) };
+                self.top.set(new_top);
+                unsafe {
+                    ptr::write(new_top.cast::<T>(), f());
+                    NonNull::new_unchecked(new_top.cast::<T>())
+                }
+            }
+        } else {
+            // in-place
+            let ptr = if slow_path {
+                self.alloc_layout_slow_cold(layout)?
+            } else {
+                let new_top = unsafe { top.byte_sub(extra).byte_sub(layout.size()) };
+                self.top.set(new_top);
+
+                unsafe { NonNull::new_unchecked(new_top) }
+            };
+
+            let ptr = ptr.cast::<T>();
+            unsafe {
+                ptr.write(f());
+            }
+
+            ptr
+        };
+
+        Some(unsafe { ptr.as_mut() })
+    }
+}
+
+fn _asdf(bump: &mut Bump<super::PanicingAllocator>) {
+    let mut bump = bump.as_scope();
+    let x = bump.alloc_with(|| 3u64);
+    bump.scope(|bump| {
+        let _y = bump.alloc_with(|| 4u64);
+        let _z = bump.alloc_with(|| 5u64);
+    });
+
+    assert!(*x == 3);
+}
diff --git a/crates/foundation/src/alloc/mod.rs b/crates/foundation/src/alloc/mod.rs
new file mode 100644
index 0000000..a6258f1
--- /dev/null
+++ b/crates/foundation/src/alloc/mod.rs
@@ -0,0 +1,23 @@
+use core::ptr::NonNull;
+
+use liballoc::alloc::Allocator;
+
+pub mod bump;
+pub mod slab;
+
+#[derive(Debug, Default)]
+pub struct PanicingAllocator;
+
+unsafe impl Allocator for PanicingAllocator {
+    fn allocate(
+        &self,
+        layout: core::alloc::Layout,
+    ) -> Result<NonNull<[u8]>, core::alloc::AllocError> {
+        match layout.size() {
+            0 => Ok(NonNull::slice_from_raw_parts(NonNull::dangling(), 0)),
+            _ => panic!("PanicingAllocator cannot allocate memory"),
+        }
+    }
+
+    unsafe fn deallocate(&self, _ptr: NonNull<u8>, _layout: core::alloc::Layout) {}
+}
diff --git a/crates/foundation/src/alloc/slab.rs b/crates/foundation/src/alloc/slab.rs
new file mode 100644
index 0000000..9fb65f2
--- /dev/null
+++ b/crates/foundation/src/alloc/slab.rs
@@ -0,0 +1,379 @@
+//! A slab allocator
+
+use core::{alloc::Layout, cell::Cell, hint::unlikely, mem::offset_of, num::NonZero, ptr::NonNull};
+
+const UNLINKED: NonNull<()> = unsafe { NonNull::new_unchecked(!0 as *mut ()) };
+
+use liballoc::alloc::Allocator;
+
+use crate::mem;
+
+const PAGE_SIZE: usize = 4096;
+
+pub struct Slab<A: Allocator + Clone> {
+    /// Size and alignment of each element in the slab.
+    element_size: usize,
+    /// Pointer to the first chunk in the slab.
+    head: Option<NonNull<SlabChunk<A>>>,
+    alloc: A,
+}
+
+struct SlabChunk<A: Allocator + Clone> {
+    /// Pointer to the next chunk in the slab.
+    next: Option<NonNull<SlabChunk<A>>>,
+    /// Pointer to the slab that owns this chunk.
+    slab: NonNull<Slab<A>>,
+    /// Linked list of free elements in this slab. When this is `None`, the
+    /// slab is full.
+    free: Cell<Option<NonNull<ChunkSlot>>>,
+    /// Number of outstanding allocations from this slab. When this reaches
+    /// zero, the slab can be freed.
+    count: Cell<usize>,
+}
+
+struct ChunkSlot(Option<NonNull<Self>>);
+
+enum SlotResult {
+    Some(NonNull<u8>),
+    Last(NonNull<u8>),
+    None,
+}
+
+impl<A: Allocator + Clone> SlabChunk<A> {
+    fn pop_free_slot(&self) -> SlotResult {
+        let Some(slot) = self.free.get() else {
+            return SlotResult::None;
+        };
+
+        self.free.set(unsafe { slot.as_ref() }.next());
+        self.count.update(|count| count + 1);
+
+        match self.free.get() {
+            Some(_) => SlotResult::Some(slot.cast()),
+            None => SlotResult::Last(slot.cast()),
+        }
+    }
+
+    /// returns `true` if the slab is now empty and can be freed
+    fn push_free_slot(&self, slot: NonNull<u8>) -> bool {
+        let slot = slot.cast::<ChunkSlot>();
+        let next = self.free.get();
+        unsafe { slot.as_ptr().write(ChunkSlot(next)) };
+        self.free.set(Some(slot));
+        self.count.update(|count| count - 1);
+
+        self.count.get() == 0
+    }
+}
+
+impl ChunkSlot {
+    fn next(&self) -> Option<NonNull<Self>> {
+        self.0
+    }
+}
+
+impl<A: Allocator + Clone> Slab<A> {
+    fn new(element_size: usize, alloc: A) -> Self {
+        assert!(
+            element_size.is_power_of_two(),
+            "element_size must be a power of two"
+        );
+
+        Self {
+            element_size,
+            head: None,
+            alloc,
+        }
+    }
+
+    fn first_slot_offset(&self) -> usize {
+        mem::align_up(core::mem::size_of::<SlabChunk<A>>(), self.element_size)
+    }
+
+    fn count_and_layout(&self) -> (usize, Layout) {
+        let (count, size, align) = {
+            let one_page_count = (PAGE_SIZE - self.first_slot_offset()) / self.element_size;
+
+            if one_page_count < 3 {
+                let count = 3;
+                let size =
+                    (self.first_slot_offset() + count * self.element_size).next_power_of_two();
+
+                assert!(size.is_multiple_of(PAGE_SIZE));
+                assert!(size >= PAGE_SIZE);
+                assert!(size.is_multiple_of(self.element_size));
+
+                (count, size, size)
+            } else {
+                let count = one_page_count;
+                let size = PAGE_SIZE;
+                (count, size, self.element_size)
+            }
+        };
+
+        (count, unsafe {
+            Layout::from_size_align_unchecked(size, align)
+        })
+    }
+
+    fn alloc_chunk(&mut self) -> NonNull<SlabChunk<A>> {
+        // we want to limit chunks to 1 page unless the element size is so
+        // large that we can fit fewer than 3 elements in a page.
+        let (count, layout) = self.count_and_layout();
+
+        let Some(bytes) = self.alloc.allocate(layout).ok() else {
+            panic!()
+        };
+
+        let chunk = bytes.as_non_null_ptr().cast::<SlabChunk<A>>();
+        unsafe {
+            let first_slot = chunk
+                .as_ptr()
+                .byte_add(self.first_slot_offset())
+                .cast::<ChunkSlot>();
+
+            for i in 0..(count - 1) {
+                let chunk = first_slot.byte_add(i * self.element_size);
+                let next = first_slot.byte_add((i + 1) * self.element_size);
+                chunk.write(ChunkSlot(Some(NonNull::new_unchecked(next))));
+            }
+            first_slot
+                .byte_add((count - 1) * self.element_size)
+                .write(ChunkSlot(None));
+
+            chunk.write(SlabChunk {
+                next: self.head,
+                slab: NonNull::from(self),
+                free: Cell::new(Some(NonNull::new_unchecked(first_slot))),
+                count: Cell::new(0),
+            });
+        }
+
+        chunk
+    }
+
+    #[cold]
+    fn alloc_chunk_cold(&mut self) -> NonNull<SlabChunk<A>> {
+        self.alloc_chunk()
+    }
+
+    fn alloc_slot(&mut self) -> NonNull<[u8]> {
+        let mut chunk = match self.head {
+            Some(chunk) => chunk,
+            None => {
+                let chunk = self.alloc_chunk_cold();
+                self.head = Some(chunk);
+                chunk
+            }
+        };
+
+        let chunk = unsafe { chunk.as_mut() };
+
+        let ptr = match chunk.pop_free_slot() {
+            SlotResult::Some(non_null) => non_null,
+            SlotResult::Last(non_null) => {
+                self.head = chunk.next.replace(UNLINKED.cast());
+                non_null
+            }
+            SlotResult::None => {
+                panic!("SlabChunk is full, but it is still the head of the slab");
+            }
+        };
+
+        ptr.cast_slice(self.element_size)
+    }
+
+    fn free_slot(&mut self, slot: NonNull<u8>) {
+        let (_, layout) = self.count_and_layout();
+        let mut chunk = slot
+            .map_addr(|addr| unsafe {
+                NonZero::new_unchecked(mem::align_down(addr.get(), layout.align()))
+            })
+            .cast::<SlabChunk<A>>();
+
+        let chunk = unsafe { chunk.as_mut() };
+        let linked = chunk.next != Some(UNLINKED.cast());
+
+        if chunk.push_free_slot(slot) {
+            if linked {
+                let mut head = self.head.expect("chunk is linked, so head exists");
+                while let Some(next) = unsafe { head.as_ref().next } {
+                    if next == chunk.into() {
+                        unsafe { head.as_mut().next = chunk.next };
+                        break;
+                    }
+                    head = next;
+                }
+            }
+
+            unsafe {
+                self.alloc
+                    .deallocate(NonNull::from_mut(chunk).cast(), layout)
+            };
+        } else if !linked {
+            chunk.next = self.head;
+            self.head = Some(chunk.into());
+        }
+    }
+}
+
+impl<A: Allocator + Clone> Drop for Slab<A> {
+    fn drop(&mut self) {
+        let (_, layout) = self.count_and_layout();
+
+        let mut chunk = self.head.take();
+        while let Some(mut chunk_ptr) = chunk {
+            let chunk_ref = unsafe { chunk_ptr.as_mut() };
+            chunk = chunk_ref.next.take();
+
+            unsafe {
+                self.alloc.deallocate(chunk_ptr.cast(), layout);
+            }
+        }
+    }
+}
+
+const SLAB_ALLOCATOR_BUCKETS: usize = 8;
+pub struct SlabAllocator<A: Allocator + Clone> {
+    /// Slabs for each power-of-two from 16 bytes to 2048 bytes (inclusive).
+    slabs: [Slab<A>; SLAB_ALLOCATOR_BUCKETS],
+    alloc: A,
+}
+
+impl<A: Allocator + Clone> SlabAllocator<A> {
+    pub fn new(alloc: A) -> Self {
+        let slabs = [
+            Slab::new(16, alloc.clone()),
+            Slab::new(32, alloc.clone()),
+            Slab::new(64, alloc.clone()),
+            Slab::new(128, alloc.clone()),
+            Slab::new(256, alloc.clone()),
+            Slab::new(512, alloc.clone()),
+            Slab::new(1024, alloc.clone()),
+            Slab::new(2048, alloc.clone()),
+        ];
+
+        Self { slabs, alloc }
+    }
+
+    fn slab_index_for_size(size: usize) -> Option<usize> {
+        if size == 0 {
+            return None;
+        }
+
+        // the smallest slab is 16 bytes
+        let size = size.max(16);
+
+        // get the index of the slab by calculating the log2 of the size and
+        // subtracting 4 (since 2^4 = 16)
+        let index = (size.next_power_of_two().trailing_zeros() - 16usize.trailing_zeros()) as usize;
+
+        // we have 8 slabs, so the index must be less than 8
+        if index < SLAB_ALLOCATOR_BUCKETS {
+            Some(index)
+        } else {
+            None
+        }
+    }
+
+    pub fn alloc(&mut self, layout: Layout) -> Option<NonNull<[u8]>> {
+        if unlikely(layout.size() == 0) {
+            return Some(NonNull::dangling().cast_slice(0));
+        }
+
+        let size = layout.size().max(layout.align());
+
+        match Self::slab_index_for_size(size) {
+            Some(slab_index) => {
+                Some(unsafe { self.slabs.get_unchecked_mut(slab_index).alloc_slot() })
+            }
+            None => {
+                // allocate directly from the backing allocator
+                self.alloc.allocate(layout).ok()
+            }
+        }
+    }
+
+    pub fn dealloc(&mut self, ptr: NonNull<u8>, layout: Layout) {
+        if unlikely(layout.size() == 0) {
+            return;
+        }
+
+        let size = layout.size().max(layout.align());
+
+        match Self::slab_index_for_size(size) {
+            Some(_) => unsafe {
+                let chunk = ptr
+                    .map_addr(|addr| {
+                        NonZero::new_unchecked(mem::align_down(addr.get(), layout.align()))
+                    })
+                    .cast::<SlabChunk<A>>();
+
+                let mut slab = (&raw const (*chunk.as_ptr()).slab).read();
+                slab.as_mut().free_slot(ptr);
+            },
+            None => {
+                // deallocate directly to the backing allocator
+                unsafe { self.alloc.deallocate(ptr, layout) }
+            }
+        }
+    }
+}
+
+impl<A: Allocator + Clone> Drop for SlabAllocator<A> {
+    fn drop(&mut self) {}
+}
+
+#[cfg(test)]
+mod tests {
+    use super::*;
+
+    use std::{prelude::rust_2024::*, sync::Mutex};
+
+    struct SlabAllocatorWrapper<A: Allocator + Clone>(Mutex<SlabAllocator<A>>);
+
+    impl<A: Allocator + Clone> SlabAllocatorWrapper<A> {
+        fn new(alloc: A) -> Self {
+            Self(Mutex::new(SlabAllocator::new(alloc)))
+        }
+    }
+    impl SlabAllocatorWrapper<std::alloc::System> {
+        fn new_system() -> SlabAllocatorWrapper<std::alloc::System> {
+            SlabAllocatorWrapper::new(std::alloc::System)
+        }
+    }
+
+    unsafe impl<A: Allocator + Clone> Allocator for SlabAllocatorWrapper<A> {
+        fn allocate(
+            &self,
+            layout: core::alloc::Layout,
+        ) -> Result<NonNull<[u8]>, core::alloc::AllocError> {
+            self.0
+                .lock()
+                .unwrap()
+                .alloc(layout)
+                .ok_or(core::alloc::AllocError)
+        }
+
+        unsafe fn deallocate(&self, ptr: NonNull<u8>, layout: core::alloc::Layout) {
+            self.0.lock().unwrap().dealloc(ptr, layout)
+        }
+    }
+
+    #[test]
+    fn test_slab_allocator() {
+        let alloc = SlabAllocatorWrapper::new_system();
+
+        let layout = Layout::from_size_align(16, 8).unwrap();
+        let ptr = alloc.allocate(layout).unwrap();
+        assert_eq!(ptr.len(), 16);
+        assert_eq!(ptr.addr().get() % 8, 0);
+
+        let mut boxed = Box::new_in(42u32, &alloc);
+        assert_eq!(*boxed, 42);
+        assert_ne!(Box::as_non_null(&mut boxed), ptr.cast());
+
+        unsafe { alloc.deallocate(ptr.cast(), layout) };
+        _ = boxed;
+    }
+}
diff --git a/crates/foundation/src/lib.rs b/crates/foundation/src/lib.rs
new file mode 100644
index 0000000..5c005aa
--- /dev/null
+++ b/crates/foundation/src/lib.rs
@@ -0,0 +1,99 @@
+#![no_std]
+#![feature(
+    const_trait_impl,
+    const_default,
+    allocator_api,
+    ptr_cast_slice,
+    likely_unlikely,
+    slice_ptr_get
+)]
+#![cfg_attr(test, feature(box_as_non_null))]
+
+extern crate alloc as liballoc;
+
+#[cfg(test)]
+extern crate std;
+
+pub mod alloc;
+pub mod sync;
+pub mod mem {
+    use num_traits::PrimInt;
+
+    pub fn replace<T, R>(v: &mut T, f: impl FnOnce(T) -> (T, R)) -> R {
+        struct Guard;
+        impl Drop for Guard {
+            fn drop(&mut self) {
+                panic!("replace() panicked");
+            }
+        }
+        let guard = Guard;
+        let value = unsafe { core::ptr::read(v) };
+        let (new_value, ret) = f(value);
+        unsafe { core::ptr::write(v, new_value) };
+        core::mem::forget(guard);
+        ret
+    }
+
+    pub fn align_up<T: PrimInt>(value: T, alignment: T) -> T {
+        (value + alignment - T::one()) & !(alignment - T::one())
+    }
+
+    pub fn align_down<T: PrimInt>(value: T, alignment: T) -> T {
+        value & !(alignment - T::one())
+    }
+
+    /// # Safety
+    pub unsafe fn volatile_copy<T: Sized>(src: *const T, dst: *mut T, count: usize) {
+        unsafe {
+            for i in 0..count {
+                let src_ptr = src.add(i);
+                let dst_ptr = dst.add(i);
+                let val = src_ptr.read();
+                dst_ptr.write_volatile(val);
+            }
+        }
+    }
+}
+
+pub struct DropGuard<F: FnOnce()>(::core::mem::ManuallyDrop<F>);
+impl<F: FnOnce()> DropGuard<F> {
+    pub fn new(f: F) -> Self {
+        DropGuard(::core::mem::ManuallyDrop::new(f))
+    }
+    #[allow(dead_code)]
+    pub fn forget(self) {
+        let mut this = ::core::mem::ManuallyDrop::new(self);
+        unsafe {
+            ::core::mem::ManuallyDrop::drop(&mut this.0);
+        }
+    }
+}
+
+impl<F: FnOnce()> Drop for DropGuard<F> {
+    fn drop(&mut self) {
+        unsafe { ::core::ptr::read(&*self.0)() }
+    }
+}
+
+#[derive(Default)]
+pub struct PanicGuard;
+
+impl PanicGuard {
+    pub fn new() -> Self {
+        Self
+    }
+
+    pub fn forget(self) {
+        core::mem::forget(self);
+    }
+}
+
+impl Drop for PanicGuard {
+    #[track_caller]
+    fn drop(&mut self) {
+        panic!(
+            "PanicGuard dropped without being forgotten: {}",
+            core::panic::Location::caller()
+        );
+    }
+}
diff --git a/crates/foundation/src/sync/mod.rs b/crates/foundation/src/sync/mod.rs
new file mode 100644
index 0000000..d00f9ba
--- /dev/null
+++ b/crates/foundation/src/sync/mod.rs
@@ -0,0 +1,487 @@
+mod spin_mutex {
+    use core::{
+        cell::UnsafeCell,
+        sync::atomic::{AtomicBool, Ordering},
+    };
+
+    pub struct SpinMutex<T: ?Sized> {
+        lock: AtomicBool,
+        data: UnsafeCell<T>,
+    }
+
+    pub struct SpinMutexGuard<'a, T: ?Sized + 'a> {
+        mutex: &'a SpinMutex<T>,
+    }
+
+    impl<T: ?Sized> core::ops::Deref for SpinMutexGuard<'_, T> {
+        type Target = T;
+
+        fn deref(&self) -> &Self::Target {
+            unsafe { &*self.mutex.data.get() }
+        }
+    }
+
+    impl<T: ?Sized> core::ops::DerefMut for SpinMutexGuard<'_, T> {
+        fn deref_mut(&mut self) -> &mut Self::Target {
+            unsafe { &mut *self.mutex.data.get() }
+        }
+    }
+
+    impl<T: ?Sized> Drop for SpinMutexGuard<'_, T> {
+        fn drop(&mut self) {
+            unsafe { self.mutex.unlock() }
+        }
+    }
+
+    unsafe impl<T: ?Sized + Send> Send for SpinMutex<T> {}
+    // SAFETY: SpinMutex properly synchronises access to T, but T must be Send
+    // to be safely shared between threads.
+    unsafe impl<T: ?Sized + Send> Sync for SpinMutex<T> {}
+
+    // SAFETY: &SpinMutexGuard cannot yield a &mut T, and the SpinMutexGuard
+    // holds the lock for the duration of its lifetime, so it is safe to share
+    // between threads.
+    unsafe impl<T: ?Sized + Sync> Sync for SpinMutexGuard<'_, T> {}
+    // SAFETY: SpinMutex may be unlocked on any thread, so it is safe to send
+    // the guard to another thread as long as T is Send.
+    unsafe impl<T: ?Sized + Send> Send for SpinMutexGuard<'_, T> {}
+
+    impl<T> SpinMutex<T> {
+        pub const fn new(data: T) -> Self {
+            Self {
+                lock: AtomicBool::new(false),
+                data: UnsafeCell::new(data),
+            }
+        }
+
+        pub fn into_inner(self) -> T {
+            self.data.into_inner()
+        }
+    }
+
+    impl<T: ?Sized> SpinMutex<T> {
+        pub fn as_mut_ptr(&self) -> *mut T {
+            self.data.get()
+        }
+
+        /// # Safety
+        /// The caller must be the logical owner of the locked mutex.
+        #[expect(clippy::mut_from_ref, reason = "unsafe mutex api")]
+        pub unsafe fn get_mut_unchecked(&self) -> &mut T {
+            unsafe { &mut *self.data.get() }
+        }
+
+        /// # Safety
+        /// The caller must be the logical owner of the locked mutex.
+        pub unsafe fn get_unchecked(&self) -> &T {
+            unsafe { &*self.data.get() }
+        }
+
+        pub fn try_lock(&self) -> bool {
+            self.lock
+                .compare_exchange(false, true, Ordering::Acquire, Ordering::Relaxed)
+                .is_ok()
+        }
+        pub fn try_lock_weak(&self) -> bool {
+            self.lock
+                .compare_exchange_weak(false, true, Ordering::Acquire, Ordering::Relaxed)
+                .is_ok()
+        }
+        pub fn lock(&self) -> SpinMutexGuard<'_, T> {
+            while !self.try_lock() {
+                core::hint::spin_loop();
+            }
+            SpinMutexGuard { mutex: self }
+        }
+
+        /// # Safety
+        /// The caller must be the logical owner of the locked mutex.
+        pub unsafe fn unlock(&self) {
+            self.lock.store(false, Ordering::Release);
+        }
+    }
+}
+
+pub use spin_mutex::SpinMutex;
+
+mod once {
+    use core::{
+        cell::{Cell, UnsafeCell},
+        mem::{ManuallyDrop, MaybeUninit},
+        sync::atomic::{AtomicU8, Ordering},
+    };
+
+    const UNINITIALIZED: u8 = 0;
+    const INITIALIZING: u8 = 1;
+    const INITIALIZED: u8 = 2;
+    const POISONED: u8 = 3;
+
+    pub struct OnceState {
+        poisoned: bool,
+        state_to_set: Cell<u8>,
+    }
+
+    impl OnceState {
+        pub fn is_poisoned(&self) -> bool {
+            self.poisoned
+        }
+        pub fn poison(&self) {
+            self.state_to_set.set(POISONED);
+        }
+    }
+
+    pub struct Once {
+        state: AtomicU8,
+    }
+
+    const impl Default for Once {
+        fn default() -> Self {
+            Self::new()
+        }
+    }
+
+    impl Once {
+        pub const fn new() -> Self {
+            Self {
+                state: AtomicU8::new(UNINITIALIZED),
+            }
+        }
+
+        pub fn state(&self) -> u8 {
+            self.state.load(Ordering::Acquire)
+        }
+
+        pub fn set_state(&self, state: u8) {
+            self.state.store(state, Ordering::Release);
+        }
+
+        pub fn is_completed(&self) -> bool {
+            self.state.load(Ordering::Acquire) == INITIALIZED
+        }
+
+        pub fn call_once<F>(&self, f: F)
+        where
+            F: FnOnce(&OnceState),
+        {
+            if self.is_completed() {
+                return;
+            }
+
+            self.call_once_slow(false, f);
+        }
+
+        pub fn call_once_force<F>(&self, f: F)
+        where
+            F: FnOnce(&OnceState),
+        {
+            if self.is_completed() {
+                return;
+            }
+
+            self.call_once_slow(true, f);
+        }
+
+        #[cold]
+        pub fn call_once_slow<F>(&self, ignore_poison: bool, f: F)
+        where
+            F: FnOnce(&OnceState),
+        {
+            let mut state = self.state.load(Ordering::Acquire);
+            loop {
+                match state {
+                    INITIALIZED => return,
+                    POISONED if !ignore_poison => {
+                        panic!("Once instance has previously been poisoned")
+                    }
+                    POISONED | UNINITIALIZED => {
+                        match self.state.compare_exchange(
+                            state,
+                            INITIALIZING,
+                            Ordering::Acquire,
+                            // if we get `Err(INITIALIZED)`, we want to have
+                            // acquired what the lock is protecting.
+                            Ordering::Acquire,
+                        ) {
+                            Err(new) => {
+                                state = new;
+                                continue;
+                            }
+                            Ok(_) => {
+                                // even though we don't have unwinding, for
+                                // completeness sake we'll poison the lock if
+                                // the closure panics.
+                                let guard = crate::DropGuard::new(|| {
+                                    self.state.store(POISONED, Ordering::Release)
+                                });
+
+                                let state = OnceState {
+                                    poisoned: state == POISONED,
+                                    state_to_set: Cell::new(INITIALIZED),
+                                };
+
+                                f(&state);
+                                guard.forget();
+
+                                self.state
+                                    .store(state.state_to_set.take(), Ordering::Release);
+                                return;
+                            }
+                        }
+                    }
+                    _ => {
+                        assert_eq!(state, INITIALIZING);
+                        loop {
+                            state = self.state.load(Ordering::Acquire);
+                            if state != INITIALIZING {
+                                break;
+                            }
+                            core::hint::spin_loop();
+                        }
+                    }
+                }
+            }
+        }
+    }
+
+    pub struct OnceLock<T> {
+        once: Once,
+        data: UnsafeCell<MaybeUninit<T>>,
+    }
+
+    unsafe impl<T: Send + Sync> Sync for OnceLock<T> {}
+
+    const impl<T> Default for OnceLock<T> {
+        fn default() -> Self {
+            Self {
+                once: Once::new(),
+                data: UnsafeCell::new(MaybeUninit::uninit()),
+            }
+        }
+    }
+
+    impl<T> OnceLock<T> {
+        pub const fn new() -> Self {
+            Self {
+                once: Once::new(),
+                data: UnsafeCell::new(MaybeUninit::uninit()),
+            }
+        }
+
+        pub const fn from(t: T) -> Self {
+            Self {
+                once: Once::new(),
+                data: UnsafeCell::new(MaybeUninit::new(t)),
+            }
+        }
+
+        /// # Safety
+        /// The caller must ensure that the `OnceLock` has been initialized before calling this method.
+        pub unsafe fn get_unchecked(&self) -> &T {
+            unsafe { self.data.get().as_ref_unchecked().assume_init_ref() }
+        }
+
+        /// # Safety
+        /// The caller must ensure that the `OnceLock` has been initialized before calling this method.
+        pub unsafe fn get_mut_unchecked(&mut self) -> &mut T {
+            unsafe { self.data.get().as_mut_unchecked().assume_init_mut() }
+        }
+
+        pub fn get(&self) -> Option<&T> {
+            if self.is_completed() {
+                Some(unsafe { self.get_unchecked() })
+            } else {
+                None
+            }
+        }
+
+        pub fn get_mut(&mut self) -> Option<&mut T> {
+            if self.is_completed_mut() {
+                Some(unsafe { self.get_mut_unchecked() })
+            } else {
+                None
+            }
+        }
+
+        pub fn is_completed(&self) -> bool {
+            self.once.is_completed()
+        }
+
+        pub fn is_completed_mut(&mut self) -> bool {
+            // avoid atomic load when we have mutable access to self
+            *self.once.state.get_mut() == INITIALIZED
+        }
+
+        pub fn get_spinning(&self) -> Option<&T> {
+            loop {
+                match self.once.state() {
+                    INITIALIZED => return Some(unsafe { self.get_unchecked() }),
+                    INITIALIZING => core::hint::spin_loop(),
+                    UNINITIALIZED | POISONED => return None,
+                    _ => unreachable!(),
+                }
+            }
+        }
+
+        pub fn initialize<F, E>(&self, f: F) -> Result<(), E>
+        where
+            F: FnOnce() -> Result<T, E>,
+        {
+            let mut res: Result<(), E> = Ok(());
+            let cell = unsafe { &mut *self.data.get() };
+
+            self.once.call_once_force(|state| match f() {
+                Ok(value) => {
+                    unsafe { cell.as_mut_ptr().write(value) };
+                }
+                Err(e) => {
+                    state.poison();
+                    res = Err(e);
+                }
+            });
+
+            res
+        }
+
+        pub fn get_or_init<F>(&self, f: F) -> &T
+        where
+            F: FnOnce() -> T,
+        {
+            match self.get_or_try_init(|| Ok::<T, core::convert::Infallible>(f())) {
+                Ok(value) => value,
+                Err(_) => unreachable!(),
+            }
+        }
+
+        pub fn get_mut_or_init<F>(&mut self, f: F) -> &mut T
+        where
+            F: FnOnce() -> T,
+        {
+            match self.get_mut_or_try_init(|| Ok::<T, core::convert::Infallible>(f())) {
+                Ok(value) => value,
+                Err(_) => unreachable!(),
+            }
+        }
+
+        pub fn get_or_try_init<F, E>(&self, f: F) -> Result<&T, E>
+        where
+            F: FnOnce() -> Result<T, E>,
+        {
+            if !self.is_completed() {
+                self.initialize(f)?;
+            }
+
+            unsafe { Ok(self.get_unchecked()) }
+        }
+
+        pub fn get_mut_or_try_init<F, E>(&mut self, f: F) -> Result<&mut T, E>
+        where
+            F: FnOnce() -> Result<T, E>,
+        {
+            if self.get_mut().is_none() {
+                self.initialize(f)?;
+            }
+
+            Ok(unsafe { self.get_mut_unchecked() })
+        }
+
+        pub fn try_insert(&self, value: T) -> Result<&T, (&T, T)> {
+            let mut val = Some(value);
+            let res = self.get_or_init(|| unsafe { val.take().unwrap_unchecked() });
+            match val {
+                Some(value) => Err((res, value)),
+                None => Ok(res),
+            }
+        }
+    }
+
+    impl<T> Drop for OnceLock<T> {
+        fn drop(&mut self) {
+            if self.is_completed() {
+                unsafe { self.data.get().as_mut_unchecked().assume_init_drop() }
+            }
+        }
+    }
+
+    pub struct LazyLock<T, F = fn() -> T> {
+        once: Once,
+        f: UnsafeCell<ManuallyDrop<F>>,
+        t: UnsafeCell<MaybeUninit<T>>,
+    }
+
+    unsafe impl<T: Sync + Send, F: Send> Sync for LazyLock<T, F> {}
+
+    impl<T, F> LazyLock<T, F>
+    where
+        F: FnOnce() -> T,
+    {
+        pub const fn new(f: F) -> Self {
+            Self {
+                once: Once::new(),
+                f: UnsafeCell::new(ManuallyDrop::new(f)),
+                t: UnsafeCell::new(MaybeUninit::uninit()),
+            }
+        }
+
+        pub fn get(&self) -> Option<&T> {
+            if self.once.is_completed() {
+                Some(unsafe { (&*self.t.get()).assume_init_ref() })
+            } else {
+                None
+            }
+        }
+
+        fn force(this: &Self) -> &T {
+            this.once.call_once(|_| {
+                // SAFETY: because `call_once` will panic if poisoned, this
+                // closure will only be called once.
+                let f = unsafe { ManuallyDrop::take(&mut *this.f.get()) };
+                let val = f();
+                unsafe { (&mut *this.t.get()).write(val) };
+            });
+
+            unsafe { (&*this.t.get()).assume_init_ref() }
+        }
+
+        fn force_mut(this: &mut Self) -> &mut T {
+            this.once.call_once(|_| {
+                // SAFETY: because `call_once` will panic if poisoned, this
+                // closure will only be called once.
+                let f = unsafe { ManuallyDrop::take(&mut *this.f.get()) };
+                let val = f();
+                unsafe { (&mut *this.t.get()).write(val) };
+            });
+
+            unsafe { (&mut *this.t.get()).assume_init_mut() }
+        }
+    }
+
+    impl<T, F: FnOnce() -> T> core::ops::Deref for LazyLock<T, F> {
+        type Target = T;
+
+        fn deref(&self) -> &Self::Target {
+            LazyLock::force(self)
+        }
+    }
+
+    impl<T, F: FnOnce() -> T> core::ops::DerefMut for LazyLock<T, F> {
+        fn deref_mut(&mut self) -> &mut Self::Target {
+            LazyLock::force_mut(self)
+        }
+    }
+
+    impl<T, F> Drop for LazyLock<T, F> {
+        fn drop(&mut self) {
+            match self.once.state() {
+                UNINITIALIZED => unsafe {
+                    ManuallyDrop::drop(self.f.get_mut());
+                },
+                INITIALIZED => unsafe {
+                    (&mut *self.t.get()).assume_init_drop();
+                },
+                _ => {}
+            }
+        }
+    }
+}
+
+pub use once::{LazyLock, Once, OnceLock};
diff --git a/kernel/Cargo.toml b/kernel/Cargo.toml
index 27aecf6..e883032 100644
--- a/kernel/Cargo.toml
+++ b/kernel/Cargo.toml
@@ -31,3 +31,5 @@ seq-macro = "0.3.6"
 rbtree = { path = "../crates/rbtree" }
 plain = "0.2.3"
 rustc-demangle = "0.1.28"
+num-traits = { version = "0.2.19", default-features = false }
+foundation = { version = "0.1.0", path = "../crates/foundation" }