From: Alistair Popple <apopple@nvidia.com>
To: Gary Guo <gary@garyguo.net>
Cc: rust-for-linux@vger.kernel.org, "Miguel Ojeda" <ojeda@kernel.org>,
"Boqun Feng" <boqun@kernel.org>,
"Björn Roy Baron" <bjorn3_gh@protonmail.com>,
"Benno Lossin" <lossin@kernel.org>,
"Andreas Hindborg" <a.hindborg@kernel.org>,
"Alice Ryhl" <aliceryhl@google.com>,
"Trevor Gross" <tmgross@umich.edu>,
"Danilo Krummrich" <dakr@kernel.org>,
"Daniel Almeida" <daniel.almeida@collabora.com>,
"Tamir Duberstein" <tamird@kernel.org>,
"Alexandre Courbot" <acourbot@nvidia.com>,
"Onur Özkan" <work@onurozkan.dev>,
"Asahi Lina" <lina+kernel@asahilina.net>,
"Martin Rodriguez Reboredo" <yakoyoku@gmail.com>
Subject: Re: [PATCH] rust: kbuild: export symbols from the `uapi` crate
Date: Wed, 16 Sep 2026 11:25:13 +1000 [thread overview]
Message-ID: <aqnqBivf3is_1N7x@nvdebian.thelocal> (raw)
In-Reply-To: <DLBLTKG6KVE9.VNWH1D3896J6@garyguo.net>
On 2026-09-10 at 21:36 +1000, Gary Guo <gary@garyguo.net> wrote...
> On Thu Sep 10, 2026 at 3:21 AM BST, Alistair Popple wrote:
> > On 2026-09-10 at 00:25 +1000, Gary Guo <gary@garyguo.net> wrote...
> >> On Wed Sep 9, 2026 at 7:06 AM BST, Alistair Popple wrote:
> >> > Symbols from the `bindings` crate and the C helpers are exported so
> >> > that loadable modules can link against code the compiler chose not to
> >> > instantiate in the module itself. The `uapi` crate has the same problem
> >> > but nothing from it is exported.
> >> >
> >> > In practice this goes unnoticed because the only functions in `uapi`
> >> > are the trivial `Default` implementations generated by `bindgen`,
> >> > which rustc inlines across crates. With CONFIG_RUST_DEBUG_ASSERTIONS=y
> >> > however, the `Default` implementation for structs can grow past the
> >> > automatic cross-crate inlining threshold. The module then references the
> >> > copy in `uapi.o`, which is not exported, and the build fails:
> >> >
> >> > ERROR: modpost: drivers/gpu/nova-drm.ko: symbol '_RNvXsH_Csk9v2ZIpWbWt_4uapiNtB5_17drm_nova_info_gpuNtNtCsfr3MPOfBGpN_4core7default7Default7default' undefined!
> >>
> >> Hmm, uapi' ideally should just be type exports and produce no code.
> >>
> >> Default implementation should carry `#[inline]` annotation so they're
> >> only codegenned in actual user crate. Rust's `#[derive(Default)]` would always
> >> add that. Can you check if that's bindgen-produced manual Default impl that
> >> somehow misses the attribute?
> >
> > Right, bindgen-produced Default impl do not have that attribute. For example
> > here is the bindgen generated implementation I see:
>
> Perhaps you can open an issue upstream?
For sure - see https://github.com/rust-lang/rust-bindgen/issues/3494
I wasn't quite sure of the justification for why we don't want to just export
these, so please feel free to comment there.
> >
> > impl Default for drm_nova_info_gpu {
> > fn default() -> Self {
> > let mut s = ::core::mem::MaybeUninit::<Self>::uninit();
> > unsafe {
> > ::core::ptr::write_bytes(s.as_mut_ptr(), 0, 1);
> > s.assume_init()
> > }
> > }
> > }
> >
> > Manually adding #[inline] fixes the problem of course. I guess we could add some
> > sed Makefile hackery to manually add it. For example something like the below,
> > not sure if that would be preferred?
>
> Yeah, textual hacks are a bit too fragile (especially that it'll break when
> upstream bindgen adds these annotations). So if you need these impls, I think
> exporting these symbols are fine.
Fun though :) But yes, I agree this is pretty hacky so would rather export the
symbols, at least until bindgen is fixed.
> That said, you can achieve what you need with `drm_nova_info_gpu::zeroed()`
> instead?
We could, but I'd rather fix this here as people are inevitably going to try and
use default::Default() and the linker errors that come from it are pretty obtuse
and difficult to diagnose what the problem really is.
- Alistair
> Best,
> Gary
>
> >
> > ---
> >
> > diff --cc rust/Makefile
> > index da1a7409d984,da1a7409d984..535a3bf181bb
> > --- a/rust/Makefile
> > +++ b/rust/Makefile
> > @@@ -525,13 -525,13 +525,16 @@@ quiet_cmd_bindgen = BINDGEN $
> > $(obj)/bindings/bindings_generated.rs: private bindgen_target_flags = \
> > $(shell grep -Ev '^#|^$$' $(src)/bindgen_parameters)
> > $(obj)/bindings/bindings_generated.rs: private bindgen_target_extra = ; \
> > -- sed -Ei 's/pub const RUST_CONST_HELPER_([a-zA-Z0-9_]*)/pub const \1/g' $@
> > ++ sed -Ei 's/pub const RUST_CONST_HELPER_([a-zA-Z0-9_]*)/pub const \1/g; \
> > ++ s/^([[:space:]]*)fn default\(\) -> Self \{/\1#[inline]\n\1fn default() -> Self {/' $@
> > $(obj)/bindings/bindings_generated.rs: $(src)/bindings/bindings_helper.h \
> > $(src)/bindgen_parameters FORCE
> > $(call if_changed_dep,bindgen)
> >
> > $(obj)/uapi/uapi_generated.rs: private bindgen_target_flags = \
> > $(shell grep -Ev '^#|^$$' $(src)/bindgen_parameters)
> > ++$(obj)/uapi/uapi_generated.rs: private bindgen_target_extra = ; \
> > ++ sed -Ei 's/^([[:space:]]*)fn default\(\) -> Self \{/\1#[inline]\n\1fn default() -> Self {/' $@
> > $(obj)/uapi/uapi_generated.rs: $(src)/uapi/uapi_helper.h \
> > $(src)/bindgen_parameters FORCE
> > $(call if_changed_dep,bindgen)
> >
>
next prev parent reply other threads:[~2026-09-16 1:25 UTC|newest]
Thread overview: 7+ messages / expand[flat|nested] mbox.gz Atom feed top
2026-09-09 6:06 [PATCH] rust: kbuild: export symbols from the `uapi` crate Alistair Popple
2026-09-09 14:25 ` Gary Guo
2026-09-10 2:21 ` Alistair Popple
2026-09-10 9:20 ` Alice Ryhl
2026-09-10 11:36 ` Gary Guo
2026-09-16 1:25 ` Alistair Popple [this message]
2026-09-16 7:39 ` Gary Guo
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=aqnqBivf3is_1N7x@nvdebian.thelocal \
--to=apopple@nvidia.com \
--cc=a.hindborg@kernel.org \
--cc=acourbot@nvidia.com \
--cc=aliceryhl@google.com \
--cc=bjorn3_gh@protonmail.com \
--cc=boqun@kernel.org \
--cc=dakr@kernel.org \
--cc=daniel.almeida@collabora.com \
--cc=gary@garyguo.net \
--cc=lina+kernel@asahilina.net \
--cc=lossin@kernel.org \
--cc=ojeda@kernel.org \
--cc=rust-for-linux@vger.kernel.org \
--cc=tamird@kernel.org \
--cc=tmgross@umich.edu \
--cc=work@onurozkan.dev \
--cc=yakoyoku@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