[RFC PATCH] rust: Provide correct provenance when constructing THIS_MODULE

[RFC PATCH] rust: Provide correct provenance when constructing THIS_MODULE

[Boqun Feng]

Currently while defining `THIS_MODULE` symbol in `module!()`, the
pointer used to contruct `ThisModule` is derived from an immutable
reference of `__this_module`, which means the pointer doesn't have
the provenance for writting, and that means any write to that pointer
is UB regardless of data races or not. However, the usage of
`THIS_MODULE` includes passing this pointer to functions that may write
to it (probably in unsafe code), and this will create soundness issues.

One way to fix this is using `addr_of_mut!()` but that requires the
unstable feature "const_mut_refs". So instead of `addr_of_mut()!`,
an extern static `Opaque` is used here: since `Opaque<T>` is transparent
to `T`, an extern static `Opaque` will just wrap the C symbol (defined
in a C compile unit) in an `Opaque`, which provides a pointer with
writable provenance via `Opaque::get()`. This fix the potential UBs
because of pointer provenance unmatched.

Reported-by: Alice Ryhl <aliceryhl@google.com>
Signed-off-by: Boqun Feng <boqun.feng@gmail.com>
---
 rust/macros/module.rs | 6 +++++-
 1 file changed, 5 insertions(+), 1 deletion(-)

diff --git a/rust/macros/module.rs b/rust/macros/module.rs
index 571ffa2e189c..aef3b132f32b 100644
--- a/rust/macros/module.rs
+++ b/rust/macros/module.rs
@@ -217,7 +217,11 @@ pub(crate) fn module(ts: TokenStream) -> TokenStream {
             // freed until the module is unloaded.
             #[cfg(MODULE)]
             static THIS_MODULE: kernel::ThisModule = unsafe {{
-                kernel::ThisModule::from_ptr(&kernel::bindings::__this_module as *const _ as *mut _)
+                extern \"C\" {{
+                    static __this_module: kernel::types::Opaque<kernel::bindings::module>;
+                }}
+
+                kernel::ThisModule::from_ptr(__this_module.get())
             }};
             #[cfg(not(MODULE))]
             static THIS_MODULE: kernel::ThisModule = unsafe {{
-- 
2.45.2

Re: [RFC PATCH] rust: Provide correct provenance when constructing THIS_MODULE

[Alice Ryhl]

On Wed, Aug 28, 2024 at 8:02 PM Boqun Feng <boqun.feng@gmail.com> wrote:
>
> Currently while defining `THIS_MODULE` symbol in `module!()`, the
> pointer used to contruct `ThisModule` is derived from an immutable
> reference of `__this_module`, which means the pointer doesn't have
> the provenance for writting, and that means any write to that pointer
> is UB regardless of data races or not. However, the usage of
> `THIS_MODULE` includes passing this pointer to functions that may write
> to it (probably in unsafe code), and this will create soundness issues.
>
> One way to fix this is using `addr_of_mut!()` but that requires the
> unstable feature "const_mut_refs". So instead of `addr_of_mut()!`,
> an extern static `Opaque` is used here: since `Opaque<T>` is transparent
> to `T`, an extern static `Opaque` will just wrap the C symbol (defined
> in a C compile unit) in an `Opaque`, which provides a pointer with
> writable provenance via `Opaque::get()`. This fix the potential UBs
> because of pointer provenance unmatched.
>
> Reported-by: Alice Ryhl <aliceryhl@google.com>
> Signed-off-by: Boqun Feng <boqun.feng@gmail.com>

Thanks. This probably didn't need to be an RFC.

Reviewed-by: Alice Ryhl <aliceryhl@google.com>

Re: [RFC PATCH] rust: Provide correct provenance when constructing THIS_MODULE

[Trevor Gross]

On Wed, Aug 28, 2024 at 1:02 PM Boqun Feng <boqun.feng@gmail.com> wrote:
>
> Currently while defining `THIS_MODULE` symbol in `module!()`, the
> pointer used to contruct `ThisModule` is derived from an immutable
> reference of `__this_module`, which means the pointer doesn't have
> the provenance for writting, and that means any write to that pointer
> is UB regardless of data races or not. However, the usage of
> `THIS_MODULE` includes passing this pointer to functions that may write
> to it (probably in unsafe code), and this will create soundness issues.
>
> One way to fix this is using `addr_of_mut!()` but that requires the
> unstable feature "const_mut_refs". So instead of `addr_of_mut()!`,
> an extern static `Opaque` is used here: since `Opaque<T>` is transparent
> to `T`, an extern static `Opaque` will just wrap the C symbol (defined
> in a C compile unit) in an `Opaque`, which provides a pointer with
> writable provenance via `Opaque::get()`. This fix the potential UBs
> because of pointer provenance unmatched.
>
> Reported-by: Alice Ryhl <aliceryhl@google.com>
> Signed-off-by: Boqun Feng <boqun.feng@gmail.com>

Reviewed-by: Trevor Gross <tmgross@umich.edu>

Thanks for the fix!

Re: [RFC PATCH] rust: Provide correct provenance when constructing THIS_MODULE

[Gary Guo]

On Wed, 28 Aug 2024 11:01:29 -0700
Boqun Feng <boqun.feng@gmail.com> wrote:

> Currently while defining `THIS_MODULE` symbol in `module!()`, the
> pointer used to contruct `ThisModule` is derived from an immutable
> reference of `__this_module`, which means the pointer doesn't have
> the provenance for writting, and that means any write to that pointer
> is UB regardless of data races or not. However, the usage of
> `THIS_MODULE` includes passing this pointer to functions that may write
> to it (probably in unsafe code), and this will create soundness issues.
> 
> One way to fix this is using `addr_of_mut!()` but that requires the
> unstable feature "const_mut_refs". So instead of `addr_of_mut()!`,
> an extern static `Opaque` is used here: since `Opaque<T>` is transparent
> to `T`, an extern static `Opaque` will just wrap the C symbol (defined
> in a C compile unit) in an `Opaque`, which provides a pointer with
> writable provenance via `Opaque::get()`. This fix the potential UBs
> because of pointer provenance unmatched.

`const_mut_refs` is getting stablised so we should be able to use it
soon. I am slightly inclined to use `addr_of_mut!()` over `Opaque` in
this case so we can use it directly from bindgen.

That said, the current approach also LGTM.

Reviewed-by: Gary Guo <gary@garyguo.net>

> 
> Reported-by: Alice Ryhl <aliceryhl@google.com>
> Signed-off-by: Boqun Feng <boqun.feng@gmail.com>
> ---
>  rust/macros/module.rs | 6 +++++-
>  1 file changed, 5 insertions(+), 1 deletion(-)
> 
> diff --git a/rust/macros/module.rs b/rust/macros/module.rs
> index 571ffa2e189c..aef3b132f32b 100644
> --- a/rust/macros/module.rs
> +++ b/rust/macros/module.rs
> @@ -217,7 +217,11 @@ pub(crate) fn module(ts: TokenStream) -> TokenStream {
>              // freed until the module is unloaded.
>              #[cfg(MODULE)]
>              static THIS_MODULE: kernel::ThisModule = unsafe {{
> -                kernel::ThisModule::from_ptr(&kernel::bindings::__this_module as *const _ as *mut _)
> +                extern \"C\" {{
> +                    static __this_module: kernel::types::Opaque<kernel::bindings::module>;
> +                }}
> +
> +                kernel::ThisModule::from_ptr(__this_module.get())
>              }};
>              #[cfg(not(MODULE))]
>              static THIS_MODULE: kernel::ThisModule = unsafe {{

Re: [RFC PATCH] rust: Provide correct provenance when constructing THIS_MODULE

[Boqun Feng]

On Fri, Aug 30, 2024 at 04:05:32PM +0100, Gary Guo wrote:
> On Wed, 28 Aug 2024 11:01:29 -0700
> Boqun Feng <boqun.feng@gmail.com> wrote:
> 
> > Currently while defining `THIS_MODULE` symbol in `module!()`, the
> > pointer used to contruct `ThisModule` is derived from an immutable
> > reference of `__this_module`, which means the pointer doesn't have
> > the provenance for writting, and that means any write to that pointer
> > is UB regardless of data races or not. However, the usage of
> > `THIS_MODULE` includes passing this pointer to functions that may write
> > to it (probably in unsafe code), and this will create soundness issues.
> > 
> > One way to fix this is using `addr_of_mut!()` but that requires the
> > unstable feature "const_mut_refs". So instead of `addr_of_mut()!`,
> > an extern static `Opaque` is used here: since `Opaque<T>` is transparent
> > to `T`, an extern static `Opaque` will just wrap the C symbol (defined
> > in a C compile unit) in an `Opaque`, which provides a pointer with
> > writable provenance via `Opaque::get()`. This fix the potential UBs
> > because of pointer provenance unmatched.
> 
> `const_mut_refs` is getting stablised so we should be able to use it
> soon. I am slightly inclined to use `addr_of_mut!()` over `Opaque` in
> this case so we can use it directly from bindgen.
> 

That's the reason why I put "RFC" in the title, although I feel `Opaque`
is better, one of the reasons is you can easily provide a `*mut T` with
wrong provenance (e.g. casting from a `&T`), but it's harder to
construct a (or an? ;-)) `&Opaque<T>` incorrectly. So if an API takes
`&Opaque<T>` instead of a `*mut T`, it can reduce some user errors.
Therefore personally, I prefer `&Opaque<T>` (or `&UnsafeCell<T>`). But
of course, I don't think this is something really strong, and I might
miss something, so I don't feel bad using `*mut T`.

Moreover, besides `const_mut_refs`, we also may also want to wait for

	https://github.com/rust-lang/rust/pull/125834

to be in stable to avoid unnecessary unsafe.

> That said, the current approach also LGTM.
> 

Thanks!

Regards,
Boqun

> Reviewed-by: Gary Guo <gary@garyguo.net>
> 
> > 
> > Reported-by: Alice Ryhl <aliceryhl@google.com>
> > Signed-off-by: Boqun Feng <boqun.feng@gmail.com>
> > ---
> >  rust/macros/module.rs | 6 +++++-
> >  1 file changed, 5 insertions(+), 1 deletion(-)
> > 
> > diff --git a/rust/macros/module.rs b/rust/macros/module.rs
> > index 571ffa2e189c..aef3b132f32b 100644
> > --- a/rust/macros/module.rs
> > +++ b/rust/macros/module.rs
> > @@ -217,7 +217,11 @@ pub(crate) fn module(ts: TokenStream) -> TokenStream {
> >              // freed until the module is unloaded.
> >              #[cfg(MODULE)]
> >              static THIS_MODULE: kernel::ThisModule = unsafe {{
> > -                kernel::ThisModule::from_ptr(&kernel::bindings::__this_module as *const _ as *mut _)
> > +                extern \"C\" {{
> > +                    static __this_module: kernel::types::Opaque<kernel::bindings::module>;
> > +                }}
> > +
> > +                kernel::ThisModule::from_ptr(__this_module.get())
> >              }};
> >              #[cfg(not(MODULE))]
> >              static THIS_MODULE: kernel::ThisModule = unsafe {{
> 

Re: [RFC PATCH] rust: Provide correct provenance when constructing THIS_MODULE

[Benno Lossin]

On 28.08.24 20:01, Boqun Feng wrote:
> Currently while defining `THIS_MODULE` symbol in `module!()`, the
> pointer used to contruct `ThisModule` is derived from an immutable
> reference of `__this_module`, which means the pointer doesn't have
> the provenance for writting, and that means any write to that pointer
> is UB regardless of data races or not. However, the usage of
> `THIS_MODULE` includes passing this pointer to functions that may write
> to it (probably in unsafe code), and this will create soundness issues.
> 
> One way to fix this is using `addr_of_mut!()` but that requires the
> unstable feature "const_mut_refs". So instead of `addr_of_mut()!`,
> an extern static `Opaque` is used here: since `Opaque<T>` is transparent
> to `T`, an extern static `Opaque` will just wrap the C symbol (defined
> in a C compile unit) in an `Opaque`, which provides a pointer with
> writable provenance via `Opaque::get()`. This fix the potential UBs
> because of pointer provenance unmatched.
> 
> Reported-by: Alice Ryhl <aliceryhl@google.com>
> Signed-off-by: Boqun Feng <boqun.feng@gmail.com>

Reviewed-by: Benno Lossin <benno.lossin@proton.me>

---
Cheers,
Benno

> ---
>  rust/macros/module.rs | 6 +++++-
>  1 file changed, 5 insertions(+), 1 deletion(-)
> 
> diff --git a/rust/macros/module.rs b/rust/macros/module.rs
> index 571ffa2e189c..aef3b132f32b 100644
> --- a/rust/macros/module.rs
> +++ b/rust/macros/module.rs
> @@ -217,7 +217,11 @@ pub(crate) fn module(ts: TokenStream) -> TokenStream {
>              // freed until the module is unloaded.
>              #[cfg(MODULE)]
>              static THIS_MODULE: kernel::ThisModule = unsafe {{
> -                kernel::ThisModule::from_ptr(&kernel::bindings::__this_module as *const _ as *mut _)
> +                extern \"C\" {{
> +                    static __this_module: kernel::types::Opaque<kernel::bindings::module>;
> +                }}
> +
> +                kernel::ThisModule::from_ptr(__this_module.get())
>              }};
>              #[cfg(not(MODULE))]
>              static THIS_MODULE: kernel::ThisModule = unsafe {{
> --
> 2.45.2
> 

Re: [RFC PATCH] rust: Provide correct provenance when constructing THIS_MODULE

[Miguel Ojeda]

On Wed, Aug 28, 2024 at 8:02 PM Boqun Feng <boqun.feng@gmail.com> wrote:
>
> Currently while defining `THIS_MODULE` symbol in `module!()`, the
> pointer used to contruct `ThisModule` is derived from an immutable
> reference of `__this_module`, which means the pointer doesn't have
> the provenance for writting, and that means any write to that pointer
> is UB regardless of data races or not. However, the usage of
> `THIS_MODULE` includes passing this pointer to functions that may write
> to it (probably in unsafe code), and this will create soundness issues.
>
> One way to fix this is using `addr_of_mut!()` but that requires the
> unstable feature "const_mut_refs". So instead of `addr_of_mut()!`,
> an extern static `Opaque` is used here: since `Opaque<T>` is transparent
> to `T`, an extern static `Opaque` will just wrap the C symbol (defined
> in a C compile unit) in an `Opaque`, which provides a pointer with
> writable provenance via `Opaque::get()`. This fix the potential UBs
> because of pointer provenance unmatched.
>
> Reported-by: Alice Ryhl <aliceryhl@google.com>
> Signed-off-by: Boqun Feng <boqun.feng@gmail.com>

Applied to `rust-fixes` -- thanks everyone!

    Closes: https://rust-for-linux.zulipchat.com/#narrow/stream/x/topic/x/near/465412664
    Fixes: 1fbde52bde73 ("rust: add `macros` crate")
    Cc: stable@vger.kernel.org # 6.6.x: be2ca1e03965: ("rust: types:
Make Opaque::get const")

    [ Fixed two typos, reworded title. - Miguel ]

For 6.1 we will send a targeted patch via Option 3 instead of
cherry-picking `Opaque` and a bunch of other bits that would be
required.

Cheers,
Miguel