On Wed, Dec 31, 2025 at 12:22:29PM +0000, Alice Ryhl wrote:
> Using `READ_ONCE` is the correct way to read the `f_flags` field.
>
> Signed-off-by: Alice Ryhl <aliceryhl@google.com>
> ---
> rust/kernel/fs/file.rs | 8 ++------
> 1 file changed, 2 insertions(+), 6 deletions(-)
>
> diff --git a/rust/kernel/fs/file.rs b/rust/kernel/fs/file.rs
> index 23ee689bd2400565223181645157d832a836589f..6b07f08e7012f512e53743266096ce0076d29e1c 100644
> --- a/rust/kernel/fs/file.rs
> +++ b/rust/kernel/fs/file.rs
> @@ -335,12 +335,8 @@ pub fn cred(&self) -> &Credential {
> /// The flags are a combination of the constants in [`flags`].
> #[inline]
> pub fn flags(&self) -> u32 {
> - // This `read_volatile` is intended to correspond to a READ_ONCE call.
> - //
> - // SAFETY: The file is valid because the shared reference guarantees a nonzero refcount.
> - //
> - // FIXME(read_once): Replace with `read_once` when available on the Rust side.
> - unsafe { core::ptr::addr_of!((*self.as_ptr()).f_flags).read_volatile() }
> + // SAFETY: The `f_flags` field of `struct file` is readable with `READ_ONCE`.
> + unsafe { kernel::sync::READ_ONCE(&raw const (*self.as_ptr()).f_flags) }
Not a question directly to this patch, but for FS folks: I see we read
and write `f_flags` normally (i.e. without *_ONCE() or any atomic), and
I don't see any synchronization between these read and write (maybe I'm
missing something?), if read and write can happen at the same time, it's
data race. So I assume we must have some assumption on the atomicity of
these accesses to `f_flags`, could you may share or confirm this? Thanks
Regards,
Boqun
> }
> }
>
>
> --
> 2.52.0.351.gbe84eed79e-goog
>