lists.openwall.net   lists  /  announce  owl-users  owl-dev  john-users  john-dev  passwdqc-users  yescrypt  popa3d-users  /  oss-security  kernel-hardening  musl  sabotage  tlsify  passwords  /  crypt-dev  xvendor  /  Bugtraq  Full-Disclosure  linux-kernel  linux-netdev  linux-ext4  linux-hardening  linux-cve-announce  PHC 
Open Source and information security mailing list archives
 
Hash Suite: Windows password security audit tool. GUI, reports in PDF.
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <CADUfDZoBrnDpnTOxiDq6pBkctJ3NDJq7Wcqm2pUu_ooqMy8yyw@mail.gmail.com>
Date: Sun, 20 Jul 2025 15:10:28 -0400
From: Caleb Sander Mateos <csander@...estorage.com>
To: Sidong Yang <sidong.yang@...iosa.ai>
Cc: Miguel Ojeda <ojeda@...nel.org>, Arnd Bergmann <arnd@...db.de>, Jens Axboe <axboe@...nel.dk>, 
	rust-for-linux@...r.kernel.org, linux-kernel@...r.kernel.org, 
	io-uring@...r.kernel.org
Subject: Re: [PATCH 2/4] rust: io_uring: introduce rust abstraction for
 io-uring cmd

On Sat, Jul 19, 2025 at 10:34 AM Sidong Yang <sidong.yang@...iosa.ai> wrote:
>
> This patch introduces rust abstraction for io-uring sqe, cmd. IoUringSqe
> abstracts io_uring_sqe and it has cmd_data(). and IoUringCmd is
> abstraction for io_uring_cmd. From this, user can get cmd_op, flags,
> pdu and also sqe.
>
> Signed-off-by: Sidong Yang <sidong.yang@...iosa.ai>
> ---
>  rust/kernel/io_uring.rs | 114 ++++++++++++++++++++++++++++++++++++++++
>  rust/kernel/lib.rs      |   1 +
>  2 files changed, 115 insertions(+)
>  create mode 100644 rust/kernel/io_uring.rs
>
> diff --git a/rust/kernel/io_uring.rs b/rust/kernel/io_uring.rs
> new file mode 100644
> index 000000000000..7843effbedb4
> --- /dev/null
> +++ b/rust/kernel/io_uring.rs
> @@ -0,0 +1,114 @@
> +// SPDX-License-Identifier: GPL-2.0
> +
> +// Copyright (C) 2025 Furiosa AI.
> +
> +//! Files and file descriptors.
> +//!
> +//! C headers: [`include/linux/io_uring/cmd.h`](srctree/include/linux/io_uring/cmd.h) and
> +//! [`include/linux/file.h`](srctree/include/linux/file.h)
> +
> +use core::mem::MaybeUninit;
> +
> +use crate::{fs::File, types::Opaque};
> +
> +pub mod flags {
> +    pub const COMPLETE_DEFER: i32 = bindings::io_uring_cmd_flags_IO_URING_F_COMPLETE_DEFER;
> +    pub const UNLOCKED: i32 = bindings::io_uring_cmd_flags_IO_URING_F_UNLOCKED;
> +
> +    pub const MULTISHOT: i32 = bindings::io_uring_cmd_flags_IO_URING_F_MULTISHOT;
> +    pub const IOWQ: i32 = bindings::io_uring_cmd_flags_IO_URING_F_IOWQ;
> +    pub const NONBLOCK: i32 = bindings::io_uring_cmd_flags_IO_URING_F_NONBLOCK;
> +
> +    pub const SQE128: i32 = bindings::io_uring_cmd_flags_IO_URING_F_SQE128;
> +    pub const CQE32: i32 = bindings::io_uring_cmd_flags_IO_URING_F_CQE32;
> +    pub const IOPOLL: i32 = bindings::io_uring_cmd_flags_IO_URING_F_IOPOLL;
> +
> +    pub const CANCEL: i32 = bindings::io_uring_cmd_flags_IO_URING_F_CANCEL;
> +    pub const COMPAT: i32 = bindings::io_uring_cmd_flags_IO_URING_F_COMPAT;
> +    pub const TASK_DEAD: i32 = bindings::io_uring_cmd_flags_IO_URING_F_TASK_DEAD;
> +}
> +
> +#[repr(transparent)]
> +pub struct IoUringCmd {
> +    inner: Opaque<bindings::io_uring_cmd>,
> +}
> +
> +impl IoUringCmd {
> +    /// Returns the cmd_op with associated with the io_uring_cmd.
> +    #[inline]
> +    pub fn cmd_op(&self) -> u32 {
> +        // SAFETY: The call guarantees that the pointer is not dangling and stays valid
> +        unsafe { (*self.inner.get()).cmd_op }
> +    }
> +
> +    /// Returns the flags with associated with the io_uring_cmd.
> +    #[inline]
> +    pub fn flags(&self) -> u32 {
> +        // SAFETY: The call guarantees that the pointer is not dangling and stays valid
> +        unsafe { (*self.inner.get()).flags }
> +    }
> +
> +    /// Returns the ref pdu for free use.
> +    #[inline]
> +    pub fn pdu(&mut self) -> MaybeUninit<&mut [u8; 32]> {

Should be &mut MaybeUninit, right? It's the bytes that may be
uninitialized, not the reference.

> +        // SAFETY: The call guarantees that the pointer is not dangling and stays valid
> +        unsafe { MaybeUninit::new(&mut (*self.inner.get()).pdu) }
> +    }
> +
> +    /// Constructs a new `struct io_uring_cmd` wrapper from a file descriptor.

Why "from a file descriptor"?

Also, missing a comment documenting the safety preconditions?

> +    #[inline]
> +    pub unsafe fn from_raw<'a>(ptr: *const bindings::io_uring_cmd) -> &'a IoUringCmd {

Could take NonNull instead of a raw pointer.

> +        // SAFETY: The caller guarantees that the pointer is not dangling and stays valid for the
> +        // duration of 'a. The cast is okay because `File` is `repr(transparent)`.

"File" -> "IoUringCmd"?

> +        unsafe { &*ptr.cast() }
> +    }
> +
> +    // Returns the file that referenced by uring cmd self.

I had a hard time parsing this comment. How about "Returns a reference
to the uring cmd's file object"?

> +    #[inline]
> +    pub fn file<'a>(&'a self) -> &'a File {

Could elide the lifetime.

> +        // SAFETY: The call guarantees that the pointer is not dangling and stays valid
> +        let file = unsafe { (*self.inner.get()).file };
> +        unsafe { File::from_raw_file(file) }

Missing a SAFETY comment for File::from_raw_file()? I would expect
something about io_uring_cmd's file field storing a non-null pointer
to a struct file on which a reference is held for the duration of the
uring cmd.

> +    }
> +
> +    // Returns the sqe  that referenced by uring cmd self.

"Returns a reference to the uring cmd's SQE"?

> +    #[inline]
> +    pub fn sqe(&self) -> &IoUringSqe {
> +        // SAFETY: The call guarantees that the pointer is not dangling and stays valid
> +        let ptr = unsafe { (*self.inner.get()).sqe };

"ptr" isn't very descriptive. How about "sqe"?

> +        unsafe { IoUringSqe::from_raw(ptr) }

Similar, missing SAFETY comment for IoUringSqe::from_raw()?

> +    }
> +
> +    // Called by consumers of io_uring_cmd, if they originally returned -EIOCBQUEUED upon receiving the command
> +    #[inline]
> +    pub fn done(self, ret: isize, res2: u64, issue_flags: u32) {

I don't think it's safe to move io_uring_cmd. io_uring_cmd_done(), for
example, calls cmd_to_io_kiocb() to turn struct io_uring_cmd *ioucmd
into struct io_kiocb *req via a pointer cast. And struct io_kiocb's
definitely need to be pinned in memory. For example,
io_req_normal_work_add() inserts the struct io_kiocb into a linked
list. Probably some sort of pinning is necessary for IoUringCmd.

> +        // SAFETY: The call guarantees that the pointer is not dangling and stays valid
> +        unsafe {
> +            bindings::io_uring_cmd_done(self.inner.get(), ret, res2, issue_flags);
> +        }
> +    }
> +}
> +
> +#[repr(transparent)]
> +pub struct IoUringSqe {
> +    inner: Opaque<bindings::io_uring_sqe>,
> +}
> +
> +impl<'a> IoUringSqe {
> +    pub fn cmd_data(&'a self) -> &'a [Opaque<u8>] {
> +        // SAFETY: The call guarantees that the pointer is not dangling and stays valid
> +        unsafe {
> +            let cmd = (*self.inner.get()).__bindgen_anon_6.cmd.as_ref();
> +            core::slice::from_raw_parts(cmd.as_ptr() as *const Opaque<u8>, 8)

Why 8? Should be 16 bytes for a 64-byte SQE and 80 bytes for a
128-byte SQE, right?

> +        }
> +    }
> +
> +    #[inline]
> +    pub unsafe fn from_raw(ptr: *const bindings::io_uring_sqe) -> &'a IoUringSqe {

Take NonNull here too?

> +        // SAFETY: The caller guarantees that the pointer is not dangling and stays valid for the
> +        // duration of 'a. The cast is okay because `File` is `repr(transparent)`.
> +        //
> +        // INVARIANT: The caller guarantees that there are no problematic `fdget_pos` calls.

Why "File" and "fdget_pos"?

Best,
Caleb

> +        unsafe { &*ptr.cast() }
> +    }
> +}
> diff --git a/rust/kernel/lib.rs b/rust/kernel/lib.rs
> index 6b4774b2b1c3..fb310e78d51d 100644
> --- a/rust/kernel/lib.rs
> +++ b/rust/kernel/lib.rs
> @@ -80,6 +80,7 @@
>  pub mod fs;
>  pub mod init;
>  pub mod io;
> +pub mod io_uring;
>  pub mod ioctl;
>  pub mod jump_label;
>  #[cfg(CONFIG_KUNIT)]
> --
> 2.43.0
>
>

Powered by blists - more mailing lists

Powered by Openwall GNU/*/Linux Powered by OpenVZ