x86/amd_node: Fix PCI device reference counting in amd_smn_init()grafted

The local "root" pointer is a temporary variable used during the device
search. Therefore, refcount related to the search iterators should be cleaned
up after the search is complete.

Use the __free() cleanup macro to ensure the refcount is decremented when the
temporary pointer goes out of scope.

Additionally, increment the refcount when caching a root pointer. This ensures
the in-use refcount is separate from the temporary search refcounting.

Finally, drop the redundant "root = NULL" before the second search loop. The
pci_get_class() iterator always decrements the refcount of its "from"
argument, so the first loop can only fall through with "root" already NULL.

Fixes: 0a4b61d9c2e4 ("x86/amd_node: Fix AMD root device caching")
Closes: https://sashiko.dev/#/patchset/20260806160159.230453-1-jason.andryuk%40amd.com
Reported-by: Sashiko <sashiko-bot@kernel.org>
Assisted-by: LLM
Signed-off-by: Yazen Ghannam <yazen.ghannam@amd.com>
Signed-off-by: Borislav Petkov (AMD) <bp@alien8.de>
Reviewed-by: Mario Limonciello (AMD) <superm1@kernel.org>
Cc: <stable@kernel.org>
Link: https://patch.msgid.link/20260903154325.74343-1-yazen.ghannam@amd.com

1 files changed, 120 insertions, 0 deletions

diff --git a/rust/syn/spanned.rs b/rust/syn/spanned.rs
new file mode 100644
index 000000000..fcc93699f
--- /dev/null
+++ b/rust/syn/spanned.rs
@@ -0,0 +1,120 @@
+// SPDX-License-Identifier: Apache-2.0 OR MIT
+
+//! A trait that can provide the `Span` of the complete contents of a syntax
+//! tree node.
+//!
+//! <br>
+//!
+//! # Example
+//!
+//! Suppose in a procedural macro we have a [`Type`] that we want to assert
+//! implements the [`Sync`] trait. Maybe this is the type of one of the fields
+//! of a struct for which we are deriving a trait implementation, and we need to
+//! be able to pass a reference to one of those fields across threads.
+//!
+//! [`Type`]: crate::Type
+//! [`Sync`]: std::marker::Sync
+//!
+//! If the field type does *not* implement `Sync` as required, we want the
+//! compiler to report an error pointing out exactly which type it was.
+//!
+//! The following macro code takes a variable `ty` of type `Type` and produces a
+//! static assertion that `Sync` is implemented for that type.
+//!
+//! ```
+//! # extern crate proc_macro;
+//! #
+//! use proc_macro::TokenStream;
+//! use proc_macro2::Span;
+//! use quote::quote_spanned;
+//! use syn::Type;
+//! use syn::spanned::Spanned;
+//!
+//! # const IGNORE_TOKENS: &str = stringify! {
+//! #[proc_macro_derive(MyMacro)]
+//! # };
+//! pub fn my_macro(input: TokenStream) -> TokenStream {
+//!     # let ty = get_a_type();
+//!     /* ... */
+//!
+//!     let assert_sync = quote_spanned! {ty.span()=>
+//!         struct _AssertSync where #ty: Sync;
+//!     };
+//!
+//!     /* ... */
+//!     # input
+//! }
+//! #
+//! # fn get_a_type() -> Type {
+//! #     unimplemented!()
+//! # }
+//! ```
+//!
+//! By inserting this `assert_sync` fragment into the output code generated by
+//! our macro, the user's code will fail to compile if `ty` does not implement
+//! `Sync`. The errors they would see look like the following.
+//!
+//! ```text
+//! error[E0277]: the trait bound `*const i32: std::marker::Sync` is not satisfied
+//!   --> src/main.rs:10:21
+//!    |
+//! 10 |     bad_field: *const i32,
+//!    |                ^^^^^^^^^^ `*const i32` cannot be shared between threads safely
+//! ```
+//!
+//! In this technique, using the `Type`'s span for the error message makes the
+//! error appear in the correct place underlining the right type.
+//!
+//! <br>
+//!
+//! # Limitations
+//!
+//! The underlying [`proc_macro::Span::join`] method is nightly-only. When
+//! called from within a procedural macro in a nightly compiler, `Spanned` will
+//! use `join` to produce the intended span. When not using a nightly compiler,
+//! only the span of the *first token* of the syntax tree node is returned.
+//!
+//! In the common case of wanting to use the joined span as the span of a
+//! `syn::Error`, consider instead using [`syn::Error::new_spanned`] which is
+//! able to span the error correctly under the complete syntax tree node without
+//! needing the unstable `join`.
+//!
+//! [`syn::Error::new_spanned`]: crate::Error::new_spanned
+
+use proc_macro2::Span;
+use quote::spanned::Spanned as ToTokens;
+
+/// A trait that can provide the `Span` of the complete contents of a syntax
+/// tree node.
+///
+/// This trait is automatically implemented for all types that implement
+/// [`ToTokens`] from the `quote` crate, as well as for `Span` itself.
+///
+/// [`ToTokens`]: quote::ToTokens
+///
+/// See the [module documentation] for an example.
+///
+/// [module documentation]: self
+pub trait Spanned: private::Sealed {
+    /// Returns a `Span` covering the complete contents of this syntax tree
+    /// node, or [`Span::call_site()`] if this node is empty.
+    ///
+    /// [`Span::call_site()`]: proc_macro2::Span::call_site
+    fn span(&self) -> Span;
+}
+
+impl<T: ?Sized + ToTokens> Spanned for T {
+    fn span(&self) -> Span {
+        self.__span()
+    }
+}
+
+mod private {
+    use crate::spanned::ToTokens;
+
+    pub trait Sealed {}
+    impl<T: ?Sized + ToTokens> Sealed for T {}
+
+    #[cfg(any(feature = "full", feature = "derive"))]
+    impl Sealed for crate::QSelf {}
+}