| Age | Commit message (Collapse) | Author |
|---|
|
I've been using the linked list cursor for a few different things, and I
find it inconvenient to use because all of the functions have signatures
along the lines of `Self -> Option<Self>`. The root cause of these
signatures is that the cursor points *at* an element, rather than
*between* two elements.
Thus, change the cursor API to point between two elements. This is
inspired by the stdlib linked list (well, really by this guy [1]), which
also uses cursors that point between elements.
The `peek_next` method returns a helper that lets you look at and
optionally remove the element, as one common use-case of cursors is to
iterate a list to look for an element, then remove that element.
For many of the methods, this will reduce how many we need since they
now just need a prev/next method, instead of the current state where you
may end up needing all of curr/prev/next. Also, if we decide to add a
function for splitting a list into two lists at the cursor, then a
cursor that points between elements is exactly what makes the most
sense.
Another advantage is that this means you can now have a cursor into an
empty list.
Link: https://rust-unofficial.github.io/too-many-lists/sixth-cursors-intro.html [1]
Reviewed-by: Andreas Hindborg <a.hindborg@kernel.org>
Reviewed-by: Boqun Feng <boqun.feng@gmail.com>
Signed-off-by: Alice Ryhl <aliceryhl@google.com>
Link: https://lore.kernel.org/r/20250210-cursor-between-v7-2-36f0215181ed@google.com
Signed-off-by: Miguel Ojeda <ojeda@kernel.org>
|
|
To prepare for a new cursor API that has the ability to insert elements
into the list, extract the common code needed for this operation into a
new `insert_inner` method.
Both `push_back` and `push_front` are updated to use the new function.
Reviewed-by: Andreas Hindborg <a.hindborg@kernel.org>
Reviewed-by: Boqun Feng <boqun.feng@gmail.com>
Signed-off-by: Alice Ryhl <aliceryhl@google.com>
Link: https://lore.kernel.org/r/20250210-cursor-between-v7-1-36f0215181ed@google.com
Signed-off-by: Miguel Ojeda <ojeda@kernel.org>
|
|
Checking that we are not missing any `// SAFETY` comments in our `unsafe`
blocks is something we have wanted to do for a long time, as well as
cleaning up the remaining cases that were not documented [1].
Back when Rust for Linux started, this was something that could have
been done via a script, like Rust's `tidy`. Soon after, in Rust 1.58.0,
Clippy implemented the `undocumented_unsafe_blocks` lint [2].
Even though the lint has a few false positives, e.g. in some cases where
attributes appear between the comment and the `unsafe` block [3], there
are workarounds and the lint seems quite usable already.
Thus enable the lint now.
We still have a few cases to clean up, so just allow those for the moment
by writing a `TODO` comment -- some of those may be good candidates for
new contributors.
Link: https://github.com/Rust-for-Linux/linux/issues/351 [1]
Link: https://rust-lang.github.io/rust-clippy/master/#/undocumented_unsafe_blocks [2]
Link: https://github.com/rust-lang/rust-clippy/issues/13189 [3]
Reviewed-by: Alice Ryhl <aliceryhl@google.com>
Reviewed-by: Trevor Gross <tmgross@umich.edu>
Tested-by: Gary Guo <gary@garyguo.net>
Reviewed-by: Gary Guo <gary@garyguo.net>
Link: https://lore.kernel.org/r/20240904204347.168520-5-ojeda@kernel.org
Signed-off-by: Miguel Ojeda <ojeda@kernel.org>
|
|
One way to explain what `ListArc` does is that it controls exclusive
access to the prev/next pointer field in a refcounted object. The
feature of having a special reference to a refcounted object with
exclusive access to specific fields is useful for other things, so
provide a general utility for that.
This is used by Rust Binder to keep track of which processes have a
reference to a given node. This involves an object for each process/node
pair, that is referenced by both the process and the node. For some
fields in this object, only the process's reference needs to access
them (and it needs mutable access), so Binder uses a ListArc to give the
process's reference exclusive access.
Reviewed-by: Benno Lossin <benno.lossin@proton.me>
Signed-off-by: Alice Ryhl <aliceryhl@google.com>
Link: https://lore.kernel.org/r/20240814-linked-list-v5-10-f5f5e8075da0@google.com
Signed-off-by: Miguel Ojeda <ojeda@kernel.org>
|
|
Support linked lists that can hold many different structs at once. This
is generally done using trait objects. The main challenge is figuring
what the struct is given only a pointer to the ListLinks.
We do this by storing a pointer to the struct next to the ListLinks
field. The container_of operation will then just read that pointer. When
the type is a trait object, that pointer will be a fat pointer whose
metadata is a vtable that tells you what kind of struct it is.
Heterogeneous lists are heavily used by Rust Binder. There are a lot of
so-called todo lists containing various events that need to be delivered
to userspace next time userspace calls into the driver. And there are
quite a few different todo item types: incoming transaction, changes to
refcounts, death notifications, and more.
Reviewed-by: Benno Lossin <benno.lossin@proton.me>
Signed-off-by: Alice Ryhl <aliceryhl@google.com>
Link: https://lore.kernel.org/r/20240814-linked-list-v5-9-f5f5e8075da0@google.com
Signed-off-by: Miguel Ojeda <ojeda@kernel.org>
|
|
The cursor is very similar to the list iterator, but it has one
important feature that the iterator doesn't: it can be used to remove
items from the linked list.
This feature cannot be added to the iterator because the references you
get from the iterator are considered borrows of the original list,
rather than borrows of the iterator. This means that there's no way to
prevent code like this:
let item = iter.next();
iter.remove();
use(item);
If `iter` was a cursor instead of an iterator, then `item` will be
considered a borrow of `iter`. Since `remove` destroys `iter`, this
means that the borrow-checker will prevent uses of `item` after the call
to `remove`.
So there is a trade-off between supporting use in traditional for loops,
and supporting removal of elements as you iterate. Iterators and cursors
represents two different choices on that spectrum.
Rust Binder needs cursors for the list of death notifications that a
process is currently handling. When userspace tells Binder that it has
finished processing the death notification, Binder will iterate the list
to search for the relevant item and remove it.
Reviewed-by: Benno Lossin <benno.lossin@proton.me>
Signed-off-by: Alice Ryhl <aliceryhl@google.com>
Link: https://lore.kernel.org/r/20240814-linked-list-v5-8-f5f5e8075da0@google.com
Signed-off-by: Miguel Ojeda <ojeda@kernel.org>
|
|
Rust Binder has lists containing stuff such as all contexts or all
processes, and sometimes needs to iterate over them. This patch enables
Rust Binder to do that using a normal for loop.
The iterator returns the ArcBorrow type, so it is possible to grab a
refcount to values while iterating.
Reviewed-by: Benno Lossin <benno.lossin@proton.me>
Signed-off-by: Alice Ryhl <aliceryhl@google.com>
Link: https://lore.kernel.org/r/20240814-linked-list-v5-7-f5f5e8075da0@google.com
Signed-off-by: Miguel Ojeda <ojeda@kernel.org>
|
|
Add the actual linked list itself.
The linked list uses the following design: The List type itself just has
a single pointer to the first element of the list. And the actual list
items then form a cycle. So the last item is `first->prev`.
This is slightly different from the usual kernel linked list. Matching
that exactly would amount to giving List two pointers, and having it be
part of the cycle of items. This alternate design has the advantage that
the cycle is never completely empty, which can reduce the number of
branches in some cases. However, it also has the disadvantage that List
must be pinned, which this design is trying to avoid.
Having the list items form a cycle rather than having null pointers at
the beginning/end is convenient for several reasons. For one, it lets us
store only one pointer in List, and it simplifies the implementation of
several functions.
Unfortunately, the `remove` function that removes an arbitrary element
from the list has to be unsafe. This is needed because there is no way
to handle the case where you pass an element from the wrong list. For
example, if it is the first element of some other list, then that other
list's `first` pointer would not be updated. Similarly, it could be a
data race if you try to remove it from two different lists in parallel.
(There's no problem with passing `remove` an item that's not in any
list. Additionally, other removal methods such as `pop_front` need not
be unsafe, as they can't be used to remove items from another list.)
A future patch in this series will introduce support for cursors that
can be used to remove arbitrary items without unsafe code.
Reviewed-by: Benno Lossin <benno.lossin@proton.me>
Signed-off-by: Alice Ryhl <aliceryhl@google.com>
Link: https://lore.kernel.org/r/20240814-linked-list-v5-6-f5f5e8075da0@google.com
[ Fixed a few typos. - Miguel ]
Signed-off-by: Miguel Ojeda <ojeda@kernel.org>
|
|
Adds a macro for safely implementing the ListItem trait. As part of the
implementation of the macro, we also provide a HasListLinks trait
similar to the workqueue's HasWorkItem trait.
The HasListLinks trait is only necessary if you are implementing
ListItem using the impl_list_item macro.
Reviewed-by: Benno Lossin <benno.lossin@proton.me>
Signed-off-by: Alice Ryhl <aliceryhl@google.com>
Link: https://lore.kernel.org/r/20240814-linked-list-v5-5-f5f5e8075da0@google.com
Signed-off-by: Miguel Ojeda <ojeda@kernel.org>
|
|
Define the ListLinks struct, which wraps the prev/next pointers that
will be used to insert values into a List in a future patch. Also
define the ListItem trait, which is implemented by structs that have a
ListLinks field.
The ListItem trait provides four different methods that are all
essentially container_of or the reverse of container_of. Two of them are
used before inserting/after removing an item from the list, and the two
others are used when looking at a value without changing whether it is
in a list. This distinction is introduced because it is needed for the
patch that adds support for heterogeneous lists, which are implemented
by adding a third pointer field with a fat pointer to the full struct.
When inserting into the heterogeneous list, the pointer-to-self is
updated to have the right vtable, and the container_of operation is
implemented by just returning that pointer instead of using the real
container_of operation.
Reviewed-by: Benno Lossin <benno.lossin@proton.me>
Signed-off-by: Alice Ryhl <aliceryhl@google.com>
Link: https://lore.kernel.org/r/20240814-linked-list-v5-4-f5f5e8075da0@google.com
Signed-off-by: Miguel Ojeda <ojeda@kernel.org>
|
|
Add the ability to track whether a ListArc exists for a given value,
allowing for the creation of ListArcs without going through UniqueArc.
The `impl_list_arc_safe!` macro is extended with a `tracked_by` strategy
that defers the tracking of ListArcs to a field of the struct.
Additionally, the AtomicListArcTracker type is introduced, which can
track whether a ListArc exists using an atomic. By deferring the
tracking to a field of type AtomicListArcTracker, structs gain the
ability to create ListArcs without going through a UniqueArc.
Rust Binder uses this for some objects where we want to be able to
insert them into a linked list at any time. Using the
AtomicListArcTracker, we are able to check whether an item is already in
the list, and if not, we can create a `ListArc` and push it.
The macro has the ability to defer the tracking of ListArcs to a field,
using whatever strategy that field has. Since we don't add any
strategies other than AtomicListArcTracker, another similar option would
be to hard-code that the field should be an AtomicListArcTracker.
However, Rust Binder has a case where the AtomicListArcTracker is not
stored directly in the struct, but in a sub-struct. Furthermore, the
outer struct is generic:
struct Wrapper<T: ?Sized> {
links: ListLinks,
inner: T,
}
Here, the Wrapper struct implements ListArcSafe with `tracked_by inner`,
and then the various types used with `inner` also uses the macro to
implement ListArcSafe. Some of them use the untracked strategy, and some
of them use tracked_by with an AtomicListArcTracker. This way, Wrapper
just inherits whichever choice `inner` has made.
Reviewed-by: Benno Lossin <benno.lossin@proton.me>
Signed-off-by: Alice Ryhl <aliceryhl@google.com>
Link: https://lore.kernel.org/r/20240814-linked-list-v5-3-f5f5e8075da0@google.com
Signed-off-by: Miguel Ojeda <ojeda@kernel.org>
|
|
The `ListArc` type can be thought of as a special reference to a
refcounted object that owns the permission to manipulate the
`next`/`prev` pointers stored in the refcounted object. By ensuring that
each object has only one `ListArc` reference, the owner of that
reference is assured exclusive access to the `next`/`prev` pointers.
When a `ListArc` is inserted into a `List`, the `List` takes ownership
of the `ListArc` reference.
There are various strategies for ensuring that a value has only one
`ListArc` reference. The simplest is to convert a `UniqueArc` into a
`ListArc`. However, the refcounted object could also keep track of
whether a `ListArc` exists using a boolean, which could allow for the
creation of new `ListArc` references from an `Arc` reference. Whatever
strategy is used, the relevant tracking is referred to as "the tracking
inside `T`", and the `ListArcSafe` trait (and its subtraits) are used to
update the tracking when a `ListArc` is created or destroyed.
Note that we allow the case where the tracking inside `T` thinks that a
`ListArc` exists, but actually, there isn't a `ListArc`. However, we do
not allow the opposite situation where a `ListArc` exists, but the
tracking thinks it doesn't. This is because the former can at most
result in us failing to create a `ListArc` when the operation could
succeed, whereas the latter can result in the creation of two `ListArc`
references. Only the latter situation can lead to memory safety issues.
This patch introduces the `impl_list_arc_safe!` macro that allows you to
implement `ListArcSafe` for types using the strategy where a `ListArc`
can only be created from a `UniqueArc`. Other strategies are introduced
in later patches.
This is part of the linked list that Rust Binder will use for many
different things. The strategy where a `ListArc` can only be created
from a `UniqueArc` is actually sufficient for most of the objects that
Rust Binder needs to insert into linked lists. Usually, these are todo
items that are created and then immediately inserted into a queue.
The const generic ID allows objects to have several prev/next pointer
pairs so that the same object can be inserted into several different
lists. You are able to have several `ListArc` references as long as they
correspond to different pointer pairs. The ID itself is purely a
compile-time concept and will not be present in the final binary. Both
the `List` and the `ListArc` will need to agree on the ID for them to
work together. Rust Binder uses this in a few places (e.g. death
recipients) where the same object can be inserted into both generic todo
lists and some other lists for tracking the status of the object.
The ID is a const generic rather than a type parameter because the
`pair_from_unique` method needs to be able to assert that the two ids
are different. There's no easy way to assert that when using types
instead of integers.
Reviewed-by: Benno Lossin <benno.lossin@proton.me>
Signed-off-by: Alice Ryhl <aliceryhl@google.com>
Link: https://lore.kernel.org/r/20240814-linked-list-v5-2-f5f5e8075da0@google.com
Signed-off-by: Miguel Ojeda <ojeda@kernel.org>
|
