On Thu, Aug 29, 2024 at 07:14:18PM +0000, Benno Lossin wrote:
> On 16.08.24 02:11, Danilo Krummrich wrote:
> > So far the kernel's `Box` and `Vec` types can't be used by userspace
> > test cases, since all users of those types (e.g. `CString`) use kernel
> > allocators for instantiation.
> >
> > In order to allow userspace test cases to make use of such types as
> > well, implement the `Cmalloc` allocator within the allocator_test module
> > and type alias all kernel allocators to `Cmalloc`. The `Cmalloc`
> > allocator uses libc's realloc() function as allocator backend.
> >
> > Signed-off-by: Danilo Krummrich <dakr@xxxxxxxxxx>
> > ---
> > rust/kernel/alloc/allocator_test.rs | 178 ++++++++++++++++++++++++++--
> > 1 file changed, 171 insertions(+), 7 deletions(-)
> >
> > diff --git a/rust/kernel/alloc/allocator_test.rs b/rust/kernel/alloc/allocator_test.rs
> > index 1b2642c547ec..7fff308d02dc 100644
> > --- a/rust/kernel/alloc/allocator_test.rs
> > +++ b/rust/kernel/alloc/allocator_test.rs
> > @@ -2,20 +2,184 @@
> >
>
> Could add a short paragraph as the module description why this module
> exists? Would probably be enough to paste the commit message.
Yes, sounds good.
>
> > #![allow(missing_docs)]
> >
> > -use super::{AllocError, Allocator, Flags};
> > +use super::{flags::*, AllocError, Allocator, Flags};
> > use core::alloc::Layout;
> > +use core::cmp;
> > +use core::mem;
> > +use core::ptr;
> > use core::ptr::NonNull;
> >
> > -pub struct Kmalloc;
> > +pub struct Cmalloc;
> > +pub type Kmalloc = Cmalloc;
> > pub type Vmalloc = Kmalloc;
> > pub type KVmalloc = Kmalloc;
> >
> > -unsafe impl Allocator for Kmalloc {
> > +extern "C" {
> > + #[link_name = "aligned_alloc"]
> > + fn libc_aligned_alloc(align: usize, size: usize) -> *mut core::ffi::c_void;
> > +
> > + #[link_name = "free"]
> > + fn libc_free(ptr: *mut core::ffi::c_void);
> > +}
> > +
> > +struct CmallocData {
> > + // The actual size as requested through `Cmalloc::alloc` or `Cmalloc::realloc`.
> > + size: usize,
> > + // The offset from the pointer returned to the caller of `Cmalloc::alloc` or `Cmalloc::realloc`
> > + // to the actual base address of the allocation.
> > + offset: usize,
> > +}
> > +
> > +impl Cmalloc {
> > + /// Adjust the size and alignment such that we can additionally store `CmallocData` right
> > + /// before the actual data described by `layout`.
> > + ///
> > + /// Example:
> > + ///
> > + /// For `CmallocData` assume an alignment of 8 and a size of 16.
> > + /// For `layout` assume and alignment of 16 and a size of 64.
>
> This looks like you want it rendered as bulletpoints (but it won't).
Actually, that wasn't my intention, but I'm fine changing that.
>
> > + ///
> > + /// 0 16 32 96
> > + /// |----------------|----------------|------------------------------------------------|
> > + /// empty CmallocData data
>
> Can you put this inside of '```'? Then it will render nicely in markdown
> (don't forget to specify the type 'text')
Sure.
>
> > + ///
> > + /// For this example the returned `Layout` has an alignment of 32 and a size of 96.
> > + fn layout_adjust(layout: Layout) -> Result<Layout, AllocError> {
> > + let layout = layout.pad_to_align();
> > +
> > + // Ensure that `CmallocData` fits into half the alignment. Additionally, this guarantees
> > + // that advancing a pointer aligned to `align` by `align / 2` we still satisfy or exceed
> > + // the alignment requested through `layout`.
> > + let align = cmp::max(
> > + layout.align(),
> > + mem::size_of::<CmallocData>().next_power_of_two(),
> > + ) * 2;
> > +
> > + // Add the additional space required for `CmallocData`.
> > + let size = layout.size() + mem::size_of::<CmallocData>();
> > +
> > + Ok(Layout::from_size_align(size, align)
> > + .map_err(|_| AllocError)?
> > + .pad_to_align())
> > + }
> > +
> > + fn alloc_store_data(layout: Layout) -> Result<NonNull<u8>, AllocError> {
> > + let requested_size = layout.size();
> > +
> > + let layout = Self::layout_adjust(layout)?;
> > + let min_align = layout.align() / 2;
> > +
> > + // SAFETY: Returns either NULL or a pointer to a memory allocation that satisfies or
> > + // exceeds the given size and alignment requirements.
> > + let raw_ptr = unsafe { libc_aligned_alloc(layout.align(), layout.size()) } as *mut u8;
> > +
> > + let priv_ptr = NonNull::new(raw_ptr).ok_or(AllocError)?;
> > +
> > + // SAFETY: Advance the pointer by `min_align`. The adjustments from `Self::layout_adjust`
> > + // ensure that after this operation the original size and alignment requirements are still
> > + // satisfied or exceeded.
>
> This SAFETY comment should address why it's OK to call `add`. You
> justify something different, namely why the allocation still satisfies
> the requirements of `layout`. That is something that this function
> should probably guarantee.
So, I guess you're arguing that instead I should say that, we're still within
the bounds of the same allocated object and don't exceed `isize`?
>
> > + let ptr = unsafe { priv_ptr.as_ptr().add(min_align) };
> > +
> > + // SAFETY: `min_align` is greater than or equal to the size of `CmallocData`, hence we
> > + // don't exceed the allocation boundaries.
> > + let data_ptr: *mut CmallocData = unsafe { ptr.sub(mem::size_of::<CmallocData>()) }.cast();
> > +
> > + let data = CmallocData {
> > + size: requested_size,
> > + offset: min_align,
> > + };
> > +
> > + // SAFETY: `data_ptr` is properly aligned and within the allocation boundaries reserved for
> > + // `CmallocData`.
> > + unsafe { data_ptr.write(data) };
> > +
> > + NonNull::new(ptr).ok_or(AllocError)
> > + }
> > +
> > + /// # Safety
> > + ///
> > + /// `ptr` must have been previously allocated with `Self::alloc_store_data`.
>
> You additionally need that you have shared access to the pointee.
>
> > + unsafe fn data<'a>(ptr: NonNull<u8>) -> &'a CmallocData {
> > + // SAFETY: `Self::alloc_store_data` stores the `CmallocData` right before the address
> > + // returned to callers of `Self::alloc_store_data`.
> > + let data_ptr: *mut CmallocData =
> > + unsafe { ptr.as_ptr().sub(mem::size_of::<CmallocData>()) }.cast();
> > +
> > + // SAFETY: The `CmallocData` has been previously stored at this offset with
> > + // `Self::alloc_store_data`.
> > + unsafe { &*data_ptr }
> > + }
> > +
> > + /// # Safety
> > + ///
> > + /// This function must not be called more than once for the same allocation.
> > + ///
> > + /// `ptr` must have been previously allocated with `Self::alloc_store_data`.
>
> You additionally need that you have exclusive access to the pointee.
>
> > + unsafe fn free_read_data(ptr: NonNull<u8>) {
> > + // SAFETY: `ptr` has been created by `Self::alloc_store_data`.
> > + let data = unsafe { Self::data(ptr) };
> > +
> > + // SAFETY: `ptr` has been created by `Self::alloc_store_data`.
> > + let priv_ptr = unsafe { ptr.as_ptr().sub(data.offset) };
> > +
> > + // SAFETY: `priv_ptr` has previously been allocatored with this `Allocator`.
> > + unsafe { libc_free(priv_ptr.cast()) };
> > + }
> > +}
> > +
> > +unsafe impl Allocator for Cmalloc {
> > + fn alloc(layout: Layout, flags: Flags) -> Result<NonNull<[u8]>, AllocError> {
> > + if layout.size() == 0 {
> > + return Ok(NonNull::slice_from_raw_parts(NonNull::dangling(), 0));
> > + }
> > +
> > + let ptr = Self::alloc_store_data(layout)?;
> > +
> > + if flags.contains(__GFP_ZERO) {
> > + // SAFETY: `Self::alloc_store_data` guarantees that `ptr` points to memory of at least
> > + // `layout.size()` bytes.
> > + unsafe { ptr.as_ptr().write_bytes(0, layout.size()) };
> > + }
>
> This makes me wonder, what other flags should we handle for this
> allocator?
I don't think there are any other flags that we can handle. The only other one
that'd make sense is __GFP_NOFAIL, but we can't guarantee that.
If any specific gfp flags are needed, I think it's simply not a candidate for a
userspace test.
If we really want to do something here, we could whitelist the flags we ignore,
since they do not matter (such as __GFP_NOWARN) and panic() for everything else.
But I don't think that's really needed.
>
> > +
> > + Ok(NonNull::slice_from_raw_parts(ptr, layout.size()))
> > + }
> > +
> > unsafe fn realloc(
> > - _ptr: Option<NonNull<u8>>,
> > - _layout: Layout,
> > - _flags: Flags,
> > + ptr: Option<NonNull<u8>>,
> > + layout: Layout,
> > + flags: Flags,
> > ) -> Result<NonNull<[u8]>, AllocError> {
> > - panic!();
> > + let src: NonNull<u8> = if let Some(src) = ptr {
> > + src.cast()
>
> Why the cast?
Probably a copy-paste mistake.
>
> > + } else {
> > + return Self::alloc(layout, flags);
> > + };
>
> You should be able to write this instead:
>
> let Some(src) = ptr else {
> return Self::alloc(layout, flags);
> };
Yes, indeed.
>
> > +
> > + if layout.size() == 0 {
> > + // SAFETY: `src` has been created by `Self::alloc_store_data`.
>
> This is not true, consider:
>
> let ptr = alloc(size = 0);
> free(ptr)
>
> Alloc will return a dangling pointer due to the first if statement and
> then this function will pass it to `free_read_data`, even though it
> wasn't created by `alloc_store_data`.
> This isn't forbidden by the `Allocator` trait function's safety
> requirements.
>
> > + unsafe { Self::free_read_data(src) };
> > +
> > + return Ok(NonNull::slice_from_raw_parts(NonNull::dangling(), 0));
> > + }
> > +
> > + let dst = Self::alloc(layout, flags)?;
> > +
> > + // SAFETY: `src` has been created by `Self::alloc_store_data`.
> > + let data = unsafe { Self::data(src) };
>
> Same issue here, if the allocation passed in is zero size. I think you
> have no other choice than to allocate even for zero size requests...
> Otherwise how would you know that they are zero-sized.
Good catch - gonna fix it.
>
> ---
> Cheers,
> Benno
>
> > +
> > + // SAFETY: `src` has previously been allocated with this `Allocator`; `dst` has just been
> > + // newly allocated. Copy up to the smaller of both sizes.
> > + unsafe {
> > + ptr::copy_nonoverlapping(
> > + src.as_ptr(),
> > + dst.as_ptr().cast(),
> > + cmp::min(layout.size(), data.size),
> > + )
> > + };
> > +
> > + // SAFETY: `src` has been created by `Self::alloc_store_data`.
> > + unsafe { Self::free_read_data(src) };
> > +
> > + Ok(dst)
> > }
> > }
> > --
> > 2.46.0
> >
>
