[PATCH v13 2/5] rust: support formatting of foreign types

[Tamir Duberstein]

On Thu, Jul 3, 2025 at 2:55 PM Tamir Duberstein <tamird at gmail.com> wrote:
>
> On Thu, Jul 3, 2025 at 11:08 AM Benno Lossin <lossin at kernel.org> wrote:
> >
> > On Thu Jul 3, 2025 at 3:55 PM CEST, Tamir Duberstein wrote:
> > > On Thu, Jul 3, 2025 at 5:32 AM Benno Lossin <lossin at kernel.org> wrote:
> > >> On Tue Jul 1, 2025 at 6:49 PM CEST, Tamir Duberstein wrote:
> > >> > Introduce a `fmt!` macro which wraps all arguments in
> > >> > `kernel::fmt::Adapter` and a `kernel::fmt::Display` trait. This enables
> > >> > formatting of foreign types (like `core::ffi::CStr`) that do not
> > >> > implement `core::fmt::Display` due to concerns around lossy conversions which
> > >> > do not apply in the kernel.
> > >> >
> > >> > Replace all direct calls to `format_args!` with `fmt!`.
> > >> >
> > >> > Replace all implementations of `core::fmt::Display` with implementations
> > >> > of `kernel::fmt::Display`.
> > >> >
> > >> > Suggested-by: Alice Ryhl <aliceryhl at google.com>
> > >> > Link: https://rust-for-linux.zulipchat.com/#narrow/channel/288089-General/topic/Custom.20formatting/with/516476467
> > >> > Acked-by: Greg Kroah-Hartman <gregkh at linuxfoundation.org>
> > >> > Reviewed-by: Alice Ryhl <aliceryhl at google.com>
> > >> > Signed-off-by: Tamir Duberstein <tamird at gmail.com>
> > >> > ---
> > >> >  drivers/block/rnull.rs       |  2 +-
> > >> >  drivers/gpu/nova-core/gpu.rs |  4 +-
> > >> >  rust/kernel/block/mq.rs      |  2 +-
> > >> >  rust/kernel/device.rs        |  2 +-
> > >> >  rust/kernel/fmt.rs           | 89 +++++++++++++++++++++++++++++++++++++++
> > >> >  rust/kernel/kunit.rs         |  6 +--
> > >> >  rust/kernel/lib.rs           |  1 +
> > >> >  rust/kernel/prelude.rs       |  3 +-
> > >> >  rust/kernel/print.rs         |  4 +-
> > >> >  rust/kernel/seq_file.rs      |  2 +-
> > >> >  rust/kernel/str.rs           | 22 ++++------
> > >> >  rust/macros/fmt.rs           | 99 ++++++++++++++++++++++++++++++++++++++++++++
> > >> >  rust/macros/lib.rs           | 19 +++++++++
> > >> >  rust/macros/quote.rs         |  7 ++++
> > >> >  scripts/rustdoc_test_gen.rs  |  2 +-
> > >> >  15 files changed, 236 insertions(+), 28 deletions(-)
> > >>
> > >> This would be a lot easier to review if he proc-macro and the call
> > >> replacement were different patches.
> > >>
> > >> Also the `kernel/fmt.rs` file should be a different commit.
> > >
> > > Can you help me understand why? The changes you ask to be separated
> > > would all be in different files, so why would separate commits make it
> > > easier to review?
> >
> > It takes less time to go through the entire patch and give a RB. I can
> > take smaller time chunks and don't have to get back into the entire
> > context of the patch when I don't have 30-60min available.
>
> Ah, I see what you mean. Yeah, the requirement to RB the entire patch
> does mean there's a benefit to smaller patches.
>
> > In this patch the biggest problem is the rename & addition of new
> > things, maybe just adding 200 lines in those files could be okay to go
> > together, see below for more.
>
> After implementing your suggestion of re-exporting things from
> `kernel::fmt` the diffstat is
>
> 26 files changed, 253 insertions(+), 51 deletions(-)
>
> so I guess I could do all the additions in one patch, but then
> *everything* else has to go in a single patch together because the
> formatting macros either want core::fmt::Display or
> kernel::fmt::Display; they can't work in a halfway state.
>
> >
> > > I prefer to keep things in one commit because the changes are highly
> > > interdependent. The proc macro doesn't make sense without
> > > kernel/fmt.rs and kernel/fmt.rs is useless without the proc macro.
> >
> > I think that `Adapter`, the custom `Display` and their impl blocks
> > don't need to be in the same commit as the proc-macro. They are related,
> > but maybe someone is not well-versed in proc-macros and thus doesn't
> > want to review that part.
>
> Sure, I guess I will split them. But as noted above: changing the
> formatting macros and all the types' trait implementations has to be a
> "flag day" change.
>
> >
> > >> > diff --git a/rust/kernel/fmt.rs b/rust/kernel/fmt.rs
> > >> > new file mode 100644
> > >> > index 000000000000..348d16987de6
> > >> > --- /dev/null
> > >> > +++ b/rust/kernel/fmt.rs
> > >> > @@ -0,0 +1,89 @@
> > >> > +// SPDX-License-Identifier: GPL-2.0
> > >> > +
> > >> > +//! Formatting utilities.
> > >> > +
> > >> > +use core::fmt;
> > >>
> > >> I think we should pub export all types that we are still using from
> > >> `core::fmt`. For example `Result`, `Formatter`, `Debug` etc.
> > >>
> > >> That way I can still use the same pattern of importing `fmt` and then
> > >> writing
> > >>
> > >>     impl fmt::Display for MyType {
> > >>         fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result {}
> > >>     }
> > >
> > > Great idea, done for the next spin. It would be nice to be able to
> > > lint against references to `core::fmt` outside of kernel/fmt.rs.
> >
> > I think there was something in clippy that can do that globally and we
> > could allow that in this file?
>
> I didn't find anything suitable. Do you have one in mind?

I think we want https://github.com/rust-lang/rust-clippy/issues/14807.