mirror of https://lore.kernel.org/lkml/
 help / color / mirror / Atom feed
From: Alice Ryhl <aliceryhl@google.com>
To: boqun.feng@gmail.com
Cc: alex.gaynor@gmail.com, aliceryhl@google.com,
	benno.lossin@proton.me, bjorn3_gh@protonmail.com,
	gary@garyguo.net, jiangshanlai@gmail.com,
	linux-kernel@vger.kernel.org, ojeda@kernel.org,
	patches@lists.linux.dev, rust-for-linux@vger.kernel.org,
	tj@kernel.org, wedsonaf@gmail.com
Subject: Re: [PATCH v2 5/8] rust: workqueue: add helper for defining work_struct fields
Date: Fri,  2 Jun 2023 09:37:41 +0000	[thread overview]
Message-ID: <20230602093741.1040283-1-aliceryhl@google.com> (raw)
In-Reply-To: <ZHkI8LXrqUOJ7G5E@boqun-archlinux>

Boqun Feng <boqun.feng@gmail.com> writes:
> On Thu, Jun 01, 2023 at 01:49:43PM +0000, Alice Ryhl wrote:
>> +/// Defines the method that should be called directly when a work item is executed.
>> +///
>> +/// Typically you would implement [`WorkItem`] instead. The `run` method on this trait will
>> +/// usually just perform the appropriate `container_of` translation and then call into the `run`
>> +/// method from the [`WorkItem`] trait.
>> +///
>> +/// This trait is used when the `work_struct` field is defined using the [`Work`] helper.
>> +///
>> +/// # Safety
>> +///
>> +/// Implementers must ensure that [`__enqueue`] uses a `work_struct` initialized with the [`run`]
>> +/// method of this trait as the function pointer.
>> +///
>> +/// [`__enqueue`]: RawWorkItem::__enqueue
>> +/// [`run`]: WorkItemPointer::run
>> +pub unsafe trait WorkItemPointer<const ID: u64>: RawWorkItem<ID> {
>> +    /// Run this work item.
>> +    ///
>> +    /// # Safety
>> +    ///
>> +    /// The provided `work_struct` pointer must originate from a previous call to `__enqueue` where
>> +    /// the `queue_work_on` closure returned true, and the pointer must still be valid.
>> +    unsafe extern "C" fn run(ptr: *mut bindings::work_struct);
>> +}
>> +
>> +/// Defines the method that should be called when this work item is executed.
>> +///
>> +/// This trait is used when the `work_struct` field is defined using the [`Work`] helper.
>> +pub trait WorkItem<const ID: u64 = 0> {
>> +    /// The pointer type that this struct is wrapped in. This will typically be `Arc<Self>` or
>> +    /// `Pin<Box<Self>>`.
>> +    type Pointer: WorkItemPointer<ID>;
> 
> This being an associate type makes me wonder how do we want to support
> the following (totally made-up by me, but I think it makes sense)?:
> 
> Say we have a struct
> 
> 	pub struct Foo {
> 		work: Work<Foo>,
> 		data: Data,
> 	}
> 
> 	impl Foo {
> 		pub fn do_sth(&self) {
> 			...
> 		}
> 	}
> 
> and we want to queue both Pin<Box<Foo>> and Arc<Foo> as work items, but
> the following doesn't work:
> 
> 	// Pin<Box<Foo>> can be queued.
> 	impl WorkItem for Foo {
> 		type Pointer = Pin<Box<Foo>>;
> 		fn run(ptr: Self::Pointer) {
> 			ptr.do_sth();
> 		}
> 	}
> 
> 	// Arc<Foo> can also be queued.
> 	impl WorkItem for Foo {
> 		type Pointer = Arc<Foo>;
> 		fn run(ptr: Self::Pointer) {
> 			ptr.do_sth();
> 		}
> 	}
> 
> Of course, we can use new type idiom, but that's not really great, and
> we may have more smart pointer types in the future.
> 
> Am I missing something here?

Basically, you're asking "is it possible to use the same `work_struct`
field for several different pointer types"?

When creating the function pointer to store in the `work_struct`, the
function pointer _must_ know what the pointer type is. Otherwise it
cannot call the right `WorkItem::run` method or perform the correct
`container_of` operation. (E.g. an `Arc` embeds a `refcount_t` before
the struct, but a `Box` does not.)

With an associated type there is no problem with that. Associated types
force you to make a choice, which means that the `work_struct` knows
what the pointer type is when you create it. Supporting what you suggest
means that we must be able to change the function pointer stored in the
`work_struct` after initializing it.

This is rather tricky because you can call `enqueue` from several
threads in parallel; just setting the function pointer before you call
`queue_work_on` would be a data race. I suppose you could do it by
implementing our own `queue_work_on` that sets the function pointer
_after_ successfully setting the `WORK_STRUCT_PENDING_BIT`, but I don't
think this patchset should do that.

Alice

  reply	other threads:[~2023-06-02  9:37 UTC|newest]

Thread overview: 53+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2023-06-01 13:49 [PATCH v2 0/8] rust: workqueue: add bindings for the workqueue Alice Ryhl
2023-06-01 13:49 ` [PATCH v2 1/8] rust: workqueue: add low-level workqueue bindings Alice Ryhl
2023-06-01 14:29   ` Martin Rodriguez Reboredo
2023-06-02 14:39   ` Andreas Hindborg (Samsung)
2023-06-06 23:36   ` Boqun Feng
2023-06-07 15:18     ` Alice Ryhl
2023-06-11 15:45   ` Benno Lossin
2023-06-01 13:49 ` [PATCH v2 2/8] rust: add offset_of! macro Alice Ryhl
2023-06-01 17:17   ` Gary Guo
2023-06-02 10:33   ` Andreas Hindborg (Samsung)
2023-06-07 15:15     ` Alice Ryhl
2023-06-11 15:47   ` Benno Lossin
2023-06-01 13:49 ` [PATCH v2 3/8] rust: sync: add `Arc::{from_raw, into_raw}` Alice Ryhl
2023-06-01 17:26   ` Gary Guo
2023-06-02 10:51   ` Andreas Hindborg (Samsung)
2023-06-05 14:31     ` Gary Guo
2023-06-05 14:49       ` Andreas Hindborg (Samsung)
2023-06-05 15:00       ` Boqun Feng
2023-06-05 15:20         ` Boqun Feng
2023-06-05 18:34         ` Andreas Hindborg (Samsung)
2023-06-11 15:48   ` Benno Lossin
2023-06-01 13:49 ` [PATCH v2 4/8] rust: workqueue: define built-in queues Alice Ryhl
2023-06-01 17:30   ` Gary Guo
2023-06-01 17:52     ` Martin Rodriguez Reboredo
2023-06-02  8:32     ` Alice Ryhl
2023-06-02 14:46   ` Andreas Hindborg (Samsung)
2023-06-11 15:49   ` Benno Lossin
2023-06-01 13:49 ` [PATCH v2 5/8] rust: workqueue: add helper for defining work_struct fields Alice Ryhl
2023-06-01 14:50   ` Martin Rodriguez Reboredo
2023-06-01 18:44   ` Boqun Feng
2023-06-02  8:38     ` Alice Ryhl
2023-06-02 16:32       ` Boqun Feng
2023-06-01 21:09   ` Boqun Feng
2023-06-02  9:37     ` Alice Ryhl [this message]
2023-06-02 14:41   ` Andreas Hindborg (Samsung)
2023-06-11 15:59   ` Benno Lossin
2023-06-27  8:42     ` Alice Ryhl
2023-06-01 13:49 ` [PATCH v2 6/8] rust: workqueue: implement `WorkItemPointer` for pointer types Alice Ryhl
2023-06-01 14:51   ` Martin Rodriguez Reboredo
2023-06-02 14:42   ` Andreas Hindborg (Samsung)
2023-06-11 16:01   ` Benno Lossin
2023-06-01 13:49 ` [PATCH v2 7/8] rust: workqueue: add `try_spawn` helper method Alice Ryhl
2023-06-01 14:53   ` Martin Rodriguez Reboredo
2023-06-02 14:43   ` Andreas Hindborg (Samsung)
2023-06-11 16:10   ` Benno Lossin
2023-06-01 13:49 ` [PATCH v2 8/8] rust: workqueue: add examples Alice Ryhl
2023-06-01 14:58   ` Martin Rodriguez Reboredo
2023-06-01 17:32   ` Gary Guo
2023-06-02  9:39     ` Alice Ryhl
2023-06-02 14:44   ` Andreas Hindborg (Samsung)
2023-06-02 14:48   ` Andreas Hindborg (Samsung)
2023-06-11 16:15   ` Benno Lossin
2023-06-27  8:38     ` Alice Ryhl

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20230602093741.1040283-1-aliceryhl@google.com \
    --to=aliceryhl@google.com \
    --cc=alex.gaynor@gmail.com \
    --cc=benno.lossin@proton.me \
    --cc=bjorn3_gh@protonmail.com \
    --cc=boqun.feng@gmail.com \
    --cc=gary@garyguo.net \
    --cc=jiangshanlai@gmail.com \
    --cc=linux-kernel@vger.kernel.org \
    --cc=ojeda@kernel.org \
    --cc=patches@lists.linux.dev \
    --cc=rust-for-linux@vger.kernel.org \
    --cc=tj@kernel.org \
    --cc=wedsonaf@gmail.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox

all inboxes | Powered by JetHome®