mbox series

[v2,0/6] Rust: support `CPU_MITIGATIONS` and enable `objtool`

Message ID 20240724161501.1319115-1-ojeda@kernel.org (mailing list archive)
Headers show
Series Rust: support `CPU_MITIGATIONS` and enable `objtool` | expand

Message

Miguel Ojeda July 24, 2024, 4:14 p.m. UTC
Hi,

This is an updated series to the CPU mitigations support for Rust. It
also has the patch to enable `objtool`, so that we can start running it
for Rust.

It would be nice to get this applied soon, so that we start being
warning-free (since we already get warnings under IBT builds via
`vmlinux.o`). I am happy to take it through the Rust tree if the x86 and
objtool maintainers give an Acked-by, or through any of the other trees,
as you prefer. Otherwise, I think at this point we would need to make
Rust exclusive to the mitigations, which isn't great.

With this series, again, x86_64 is warning-free with `objtool` enabled. I
tested `-O2`/`-Os` and the Rust versions we support under `-O2` (mainly
for the `noreturn` patch, which uses heuristics), as well as IBT vs. no
IBT (i.e.  running on individual object files vs. in `vmlinux`). I also
did an arm64 build.

Testing is very welcome for this one!

Cheers,
Miguel

v2:
  - Add patch to enable `objtool` for Rust.

  - Add patch to list `noreturn` Rust functions (via heuristics) to avoid
    warnings related to that.

  - Make the `RETHUNK` patch not an RFC since the Rust compiler has
  support for
    it now.

  - Update the names of the migitation config symbols, given the changes
  at e.g.
    commit 7b75782ffd82 ("x86/bugs: Rename CONFIG_MITIGATION_SLS =>
    CONFIG_MITIGATION_SLS").

Miguel Ojeda (6):
  rust: module: add static pointer to `{init,cleanup}_module()`
  x86/rust: support MITIGATION_RETPOLINE
  x86/rust: support MITIGATION_RETHUNK
  x86/rust: support MITIGATION_SLS
  objtool: list `noreturn` Rust functions
  objtool/kbuild/rust: enable objtool for Rust

 arch/x86/Makefile               |  7 ++++++-
 rust/Makefile                   | 22 ++++++++++++--------
 rust/macros/module.rs           | 12 +++++++++++
 scripts/Makefile.build          |  9 +++++++--
 scripts/generate_rust_target.rs | 15 ++++++++++++++
 tools/objtool/check.c           | 36 ++++++++++++++++++++++++++++++++-
 tools/objtool/noreturns.h       |  2 ++
 7 files changed, 91 insertions(+), 12 deletions(-)


base-commit: b1263411112305acf2af728728591465becb45b0
--
2.45.2

Comments

Benno Lossin July 24, 2024, 9:51 p.m. UTC | #1
On 24.07.24 18:14, Miguel Ojeda wrote:
> Hi,
> 
> This is an updated series to the CPU mitigations support for Rust. It
> also has the patch to enable `objtool`, so that we can start running it
> for Rust.
> 
> It would be nice to get this applied soon, so that we start being
> warning-free (since we already get warnings under IBT builds via
> `vmlinux.o`). I am happy to take it through the Rust tree if the x86 and
> objtool maintainers give an Acked-by, or through any of the other trees,
> as you prefer. Otherwise, I think at this point we would need to make
> Rust exclusive to the mitigations, which isn't great.
> 
> With this series, again, x86_64 is warning-free with `objtool` enabled. I
> tested `-O2`/`-Os` and the Rust versions we support under `-O2` (mainly
> for the `noreturn` patch, which uses heuristics), as well as IBT vs. no
> IBT (i.e.  running on individual object files vs. in `vmlinux`). I also
> did an arm64 build.
> 
> Testing is very welcome for this one!
> 
> Cheers,
> Miguel
> 
> v2:
>   - Add patch to enable `objtool` for Rust.
> 
>   - Add patch to list `noreturn` Rust functions (via heuristics) to avoid
>     warnings related to that.
> 
>   - Make the `RETHUNK` patch not an RFC since the Rust compiler has
>   support for
>     it now.
> 
>   - Update the names of the migitation config symbols, given the changes
>   at e.g.
>     commit 7b75782ffd82 ("x86/bugs: Rename CONFIG_MITIGATION_SLS =>
>     CONFIG_MITIGATION_SLS").
> 
> Miguel Ojeda (6):
>   rust: module: add static pointer to `{init,cleanup}_module()`
>   x86/rust: support MITIGATION_RETPOLINE
>   x86/rust: support MITIGATION_RETHUNK
>   x86/rust: support MITIGATION_SLS
>   objtool: list `noreturn` Rust functions
>   objtool/kbuild/rust: enable objtool for Rust
> 
>  arch/x86/Makefile               |  7 ++++++-
>  rust/Makefile                   | 22 ++++++++++++--------
>  rust/macros/module.rs           | 12 +++++++++++
>  scripts/Makefile.build          |  9 +++++++--
>  scripts/generate_rust_target.rs | 15 ++++++++++++++
>  tools/objtool/check.c           | 36 ++++++++++++++++++++++++++++++++-
>  tools/objtool/noreturns.h       |  2 ++
>  7 files changed, 91 insertions(+), 12 deletions(-)
> 
> 
> base-commit: b1263411112305acf2af728728591465becb45b0
> --
> 2.45.2
> 

I tested this series with a config that produces the objtool warnings on
b126341 and it worked flawlessly. I also tried `-O2` and `-Os`:

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

---
Cheers,
Benno
Peter Zijlstra July 25, 2024, 8:38 a.m. UTC | #2
On Wed, Jul 24, 2024 at 06:14:53PM +0200, Miguel Ojeda wrote:
> Hi,
> 
> This is an updated series to the CPU mitigations support for Rust. It
> also has the patch to enable `objtool`, so that we can start running it
> for Rust.
> 
> It would be nice to get this applied soon, so that we start being
> warning-free (since we already get warnings under IBT builds via
> `vmlinux.o`). I am happy to take it through the Rust tree if the x86 and
> objtool maintainers give an Acked-by, or through any of the other trees,
> as you prefer. Otherwise, I think at this point we would need to make
> Rust exclusive to the mitigations, which isn't great.
> 
> With this series, again, x86_64 is warning-free with `objtool` enabled. I
> tested `-O2`/`-Os` and the Rust versions we support under `-O2` (mainly
> for the `noreturn` patch, which uses heuristics), as well as IBT vs. no
> IBT (i.e.  running on individual object files vs. in `vmlinux`). I also
> did an arm64 build.

W00t :-)

Aside from a small niggle about maybe doing a helper function for those
Rust runtime things, I don't see anything objectionable here.

Thanks!
Alice Ryhl July 25, 2024, 9:43 a.m. UTC | #3
On Wed, Jul 24, 2024 at 6:15 PM Miguel Ojeda <ojeda@kernel.org> wrote:
>
> Hi,
>
> This is an updated series to the CPU mitigations support for Rust. It
> also has the patch to enable `objtool`, so that we can start running it
> for Rust.
>
> It would be nice to get this applied soon, so that we start being
> warning-free (since we already get warnings under IBT builds via
> `vmlinux.o`). I am happy to take it through the Rust tree if the x86 and
> objtool maintainers give an Acked-by, or through any of the other trees,
> as you prefer. Otherwise, I think at this point we would need to make
> Rust exclusive to the mitigations, which isn't great.
>
> With this series, again, x86_64 is warning-free with `objtool` enabled. I
> tested `-O2`/`-Os` and the Rust versions we support under `-O2` (mainly
> for the `noreturn` patch, which uses heuristics), as well as IBT vs. no
> IBT (i.e.  running on individual object files vs. in `vmlinux`). I also
> did an arm64 build.
>
> Testing is very welcome for this one!

Verified that this eliminates the relevant warnings in an x86 build of
the android-mainline kernel.

Tested-by: Alice Ryhl <aliceryhl@google.com>
Miguel Ojeda July 25, 2024, 9:53 a.m. UTC | #4
On Thu, Jul 25, 2024 at 10:38 AM Peter Zijlstra <peterz@infradead.org> wrote:
>
> W00t :-)
>
> Aside from a small niggle about maybe doing a helper function for those
> Rust runtime things, I don't see anything objectionable here.
>
> Thanks!

Thanks for taking a look that quick, Peter, I appreciate it.

Happy to move that to a helper.

Cheers,
Miguel