[PATCH] rust: sync: rcu: Mark Guard methods as inline

[PATCH] rust: sync: rcu: Mark Guard methods as inline

[I Hsin Cheng]

Currenyly the implementation of "Guard" methods are basically wrappers
around rcu's function within kernel. Building the kernel with llvm
18.1.8 on x86_64 machine will generate the following symbols:

$ nm vmlinux | grep ' _R'.*Guard | rustfilt
ffffffff817b6c90 T <kernel::sync::rcu::Guard>::new
ffffffff817b6cb0 T <kernel::sync::rcu::Guard>::unlock
ffffffff817b6cd0 T <kernel::sync::rcu::Guard as core::ops::drop::Drop>::drop
ffffffff817b6c90 T <kernel::sync::rcu::Guard as core::default::Default>::default

These Rust symbols are basically wrappers around functions
"rcu_read_lock" and "rcu_read_unlock". Marking them as inline can
reduce the generation of these symbols, and saves the size of code
generation for 100 bytes.

$ ./scripts/bloat-o-meter vmlinux_old vmlinux_new
add/remove: 0/8 grow/shrink: 0/0 up/down: 0/-100 (-100)
Function                                     old     new   delta
_RNvXs_NtNtCsaYBeKL739Xz_6kernel4sync3rcuNtB4_5GuardNtNtCsdaXADs8PRFB_4core7default7Default7default       9       -      -9
_RNvXs0_NtNtCsaYBeKL739Xz_6kernel4sync3rcuNtB5_5GuardNtNtNtCsdaXADs8PRFB_4core3ops4drop4Drop4drop       9       -      -9
_RNvMNtNtCsaYBeKL739Xz_6kernel4sync3rcuNtB2_5Guard6unlock       9       -      -9
_RNvMNtNtCsaYBeKL739Xz_6kernel4sync3rcuNtB2_5Guard3new       9       -      -9
__pfx__RNvXs_NtNtCsaYBeKL739Xz_6kernel4sync3rcuNtB4_5GuardNtNtCsdaXADs8PRFB_4core7default7Default7default      16       -     -16
__pfx__RNvXs0_NtNtCsaYBeKL739Xz_6kernel4sync3rcuNtB5_5GuardNtNtNtCsdaXADs8PRFB_4core3ops4drop4Drop4drop      16       -     -16
__pfx__RNvMNtNtCsaYBeKL739Xz_6kernel4sync3rcuNtB2_5Guard6unlock      16       -     -16
__pfx__RNvMNtNtCsaYBeKL739Xz_6kernel4sync3rcuNtB2_5Guard3new      16       -     -16
Total: Before=23385830, After=23385730, chg -0.00%

Link: https://github.com/Rust-for-Linux/linux/issues/1145
Signed-off-by: I Hsin Cheng <richard120310@gmail.com>
---
 rust/kernel/sync/rcu.rs | 4 ++++
 1 file changed, 4 insertions(+)

diff --git a/rust/kernel/sync/rcu.rs b/rust/kernel/sync/rcu.rs
index b51d9150ffe2..26df2a6cb09f 100644
--- a/rust/kernel/sync/rcu.rs
+++ b/rust/kernel/sync/rcu.rs
@@ -17,6 +17,7 @@
 
 impl Guard {
     /// Acquires the RCU read side lock and returns a guard.
+    #[inline]
     pub fn new() -> Self {
         // SAFETY: An FFI call with no additional requirements.
         unsafe { bindings::rcu_read_lock() };
@@ -25,16 +26,19 @@ pub fn new() -> Self {
     }
 
     /// Explicitly releases the RCU read side lock.
+    #[inline]
     pub fn unlock(self) {}
 }
 
 impl Default for Guard {
+    #[inline]
     fn default() -> Self {
         Self::new()
     }
 }
 
 impl Drop for Guard {
+    #[inline]
     fn drop(&mut self) {
         // SAFETY: By the type invariants, the RCU read side is locked, so it is ok to unlock it.
         unsafe { bindings::rcu_read_unlock() };
-- 
2.43.0

Re: [PATCH] rust: sync: rcu: Mark Guard methods as inline

[Benno Lossin]

On Wed Mar 12, 2025 at 11:17 AM CET, I Hsin Cheng wrote:
> Currenyly the implementation of "Guard" methods are basically wrappers
> around rcu's function within kernel. Building the kernel with llvm
> 18.1.8 on x86_64 machine will generate the following symbols:
>
> $ nm vmlinux | grep ' _R'.*Guard | rustfilt
> ffffffff817b6c90 T <kernel::sync::rcu::Guard>::new
> ffffffff817b6cb0 T <kernel::sync::rcu::Guard>::unlock
> ffffffff817b6cd0 T <kernel::sync::rcu::Guard as core::ops::drop::Drop>::drop
> ffffffff817b6c90 T <kernel::sync::rcu::Guard as core::default::Default>::default
>
> These Rust symbols are basically wrappers around functions
> "rcu_read_lock" and "rcu_read_unlock". Marking them as inline can
> reduce the generation of these symbols, and saves the size of code
> generation for 100 bytes.
>
> $ ./scripts/bloat-o-meter vmlinux_old vmlinux_new
> add/remove: 0/8 grow/shrink: 0/0 up/down: 0/-100 (-100)
> Function                                     old     new   delta
> _RNvXs_NtNtCsaYBeKL739Xz_6kernel4sync3rcuNtB4_5GuardNtNtCsdaXADs8PRFB_4core7default7Default7default       9       -      -9
> _RNvXs0_NtNtCsaYBeKL739Xz_6kernel4sync3rcuNtB5_5GuardNtNtNtCsdaXADs8PRFB_4core3ops4drop4Drop4drop       9       -      -9
> _RNvMNtNtCsaYBeKL739Xz_6kernel4sync3rcuNtB2_5Guard6unlock       9       -      -9
> _RNvMNtNtCsaYBeKL739Xz_6kernel4sync3rcuNtB2_5Guard3new       9       -      -9
> __pfx__RNvXs_NtNtCsaYBeKL739Xz_6kernel4sync3rcuNtB4_5GuardNtNtCsdaXADs8PRFB_4core7default7Default7default      16       -     -16
> __pfx__RNvXs0_NtNtCsaYBeKL739Xz_6kernel4sync3rcuNtB5_5GuardNtNtNtCsdaXADs8PRFB_4core3ops4drop4Drop4drop      16       -     -16
> __pfx__RNvMNtNtCsaYBeKL739Xz_6kernel4sync3rcuNtB2_5Guard6unlock      16       -     -16
> __pfx__RNvMNtNtCsaYBeKL739Xz_6kernel4sync3rcuNtB2_5Guard3new      16       -     -16
> Total: Before=23385830, After=23385730, chg -0.00%
>
> Link: https://github.com/Rust-for-Linux/linux/issues/1145
> Signed-off-by: I Hsin Cheng <richard120310@gmail.com>

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

---
Cheers,
Benno

Re: [PATCH] rust: sync: rcu: Mark Guard methods as inline

[Charalampos Mitrodimas]

I Hsin Cheng <richard120310@gmail.com> writes:

> Currenyly the implementation of "Guard" methods are basically wrappers

"Currenyly" -> "Currently".

> around rcu's function within kernel. Building the kernel with llvm
> 18.1.8 on x86_64 machine will generate the following symbols:
>
> $ nm vmlinux | grep ' _R'.*Guard | rustfilt
> ffffffff817b6c90 T <kernel::sync::rcu::Guard>::new
> ffffffff817b6cb0 T <kernel::sync::rcu::Guard>::unlock
> ffffffff817b6cd0 T <kernel::sync::rcu::Guard as core::ops::drop::Drop>::drop
> ffffffff817b6c90 T <kernel::sync::rcu::Guard as core::default::Default>::default
>
> These Rust symbols are basically wrappers around functions
> "rcu_read_lock" and "rcu_read_unlock". Marking them as inline can
> reduce the generation of these symbols, and saves the size of code
> generation for 100 bytes.
>
> $ ./scripts/bloat-o-meter vmlinux_old vmlinux_new
> add/remove: 0/8 grow/shrink: 0/0 up/down: 0/-100 (-100)
> Function                                     old     new   delta
> _RNvXs_NtNtCsaYBeKL739Xz_6kernel4sync3rcuNtB4_5GuardNtNtCsdaXADs8PRFB_4core7default7Default7default       9       -      -9
> _RNvXs0_NtNtCsaYBeKL739Xz_6kernel4sync3rcuNtB5_5GuardNtNtNtCsdaXADs8PRFB_4core3ops4drop4Drop4drop       9       -      -9
> _RNvMNtNtCsaYBeKL739Xz_6kernel4sync3rcuNtB2_5Guard6unlock       9       -      -9
> _RNvMNtNtCsaYBeKL739Xz_6kernel4sync3rcuNtB2_5Guard3new       9       -      -9
> __pfx__RNvXs_NtNtCsaYBeKL739Xz_6kernel4sync3rcuNtB4_5GuardNtNtCsdaXADs8PRFB_4core7default7Default7default      16       -     -16
> __pfx__RNvXs0_NtNtCsaYBeKL739Xz_6kernel4sync3rcuNtB5_5GuardNtNtNtCsdaXADs8PRFB_4core3ops4drop4Drop4drop      16       -     -16
> __pfx__RNvMNtNtCsaYBeKL739Xz_6kernel4sync3rcuNtB2_5Guard6unlock      16       -     -16
> __pfx__RNvMNtNtCsaYBeKL739Xz_6kernel4sync3rcuNtB2_5Guard3new      16       -     -16
> Total: Before=23385830, After=23385730, chg -0.00%
>
> Link: https://github.com/Rust-for-Linux/linux/issues/1145
> Signed-off-by: I Hsin Cheng <richard120310@gmail.com>

Reviewed-by: Charalampos Mitrodimas <charmitro@posteo.net>

Re: [PATCH] rust: sync: rcu: Mark Guard methods as inline

[Joel Fernandes]

On Wed, Mar 12, 2025 at 06:17:23PM +0800, I Hsin Cheng wrote:
> Currenyly the implementation of "Guard" methods are basically wrappers
> around rcu's function within kernel. Building the kernel with llvm
> 18.1.8 on x86_64 machine will generate the following symbols:
> 
> $ nm vmlinux | grep ' _R'.*Guard | rustfilt
> ffffffff817b6c90 T <kernel::sync::rcu::Guard>::new
> ffffffff817b6cb0 T <kernel::sync::rcu::Guard>::unlock
> ffffffff817b6cd0 T <kernel::sync::rcu::Guard as core::ops::drop::Drop>::drop
> ffffffff817b6c90 T <kernel::sync::rcu::Guard as core::default::Default>::default
> 
> These Rust symbols are basically wrappers around functions
> "rcu_read_lock" and "rcu_read_unlock". Marking them as inline can
> reduce the generation of these symbols, and saves the size of code
> generation for 100 bytes.
> 
> $ ./scripts/bloat-o-meter vmlinux_old vmlinux_new
> add/remove: 0/8 grow/shrink: 0/0 up/down: 0/-100 (-100)
> Function                                     old     new   delta
> _RNvXs_NtNtCsaYBeKL739Xz_6kernel4sync3rcuNtB4_5GuardNtNtCsdaXADs8PRFB_4core7default7Default7default       9       -      -9
> _RNvXs0_NtNtCsaYBeKL739Xz_6kernel4sync3rcuNtB5_5GuardNtNtNtCsdaXADs8PRFB_4core3ops4drop4Drop4drop       9       -      -9
> _RNvMNtNtCsaYBeKL739Xz_6kernel4sync3rcuNtB2_5Guard6unlock       9       -      -9
> _RNvMNtNtCsaYBeKL739Xz_6kernel4sync3rcuNtB2_5Guard3new       9       -      -9
> __pfx__RNvXs_NtNtCsaYBeKL739Xz_6kernel4sync3rcuNtB4_5GuardNtNtCsdaXADs8PRFB_4core7default7Default7default      16       -     -16
> __pfx__RNvXs0_NtNtCsaYBeKL739Xz_6kernel4sync3rcuNtB5_5GuardNtNtNtCsdaXADs8PRFB_4core3ops4drop4Drop4drop      16       -     -16
> __pfx__RNvMNtNtCsaYBeKL739Xz_6kernel4sync3rcuNtB2_5Guard6unlock      16       -     -16
> __pfx__RNvMNtNtCsaYBeKL739Xz_6kernel4sync3rcuNtB2_5Guard3new      16       -     -16
> Total: Before=23385830, After=23385730, chg -0.00%
> 
> Link: https://github.com/Rust-for-Linux/linux/issues/1145
> Signed-off-by: I Hsin Cheng <richard120310@gmail.com>
> ---
>  rust/kernel/sync/rcu.rs | 4 ++++
>  1 file changed, 4 insertions(+)
> 
> diff --git a/rust/kernel/sync/rcu.rs b/rust/kernel/sync/rcu.rs
> index b51d9150ffe2..26df2a6cb09f 100644
> --- a/rust/kernel/sync/rcu.rs
> +++ b/rust/kernel/sync/rcu.rs
> @@ -17,6 +17,7 @@
>  
>  impl Guard {
>      /// Acquires the RCU read side lock and returns a guard.
> +    #[inline]
>      pub fn new() -> Self {
>          // SAFETY: An FFI call with no additional requirements.
>          unsafe { bindings::rcu_read_lock() };
> @@ -25,16 +26,19 @@ pub fn new() -> Self {
>      }
>  
>      /// Explicitly releases the RCU read side lock.
> +    #[inline]
>      pub fn unlock(self) {}
>  }
>  
>  impl Default for Guard {
> +    #[inline]
>      fn default() -> Self {
>          Self::new()
>      }
>  }
>  
>  impl Drop for Guard {
> +    #[inline]
>      fn drop(&mut self) {
>          // SAFETY: By the type invariants, the RCU read side is locked, so it is ok to unlock it.
>          unsafe { bindings::rcu_read_unlock() };

Reviewed-by: Joel Fernandes <joelagnelf@nvidia.com>

If no objections, I can queue it for 6.16 but let me know if Rust maintainers
prefer to take it.

thanks,

 - Joel

Re: [PATCH] rust: sync: rcu: Mark Guard methods as inline

[Christian Schrefl]

On 12.03.25 11:17 AM, I Hsin Cheng wrote:
> Currenyly the implementation of "Guard" methods are basically wrappers
> around rcu's function within kernel. Building the kernel with llvm
> 18.1.8 on x86_64 machine will generate the following symbols:
> 
> $ nm vmlinux | grep ' _R'.*Guard | rustfilt
> ffffffff817b6c90 T <kernel::sync::rcu::Guard>::new
> ffffffff817b6cb0 T <kernel::sync::rcu::Guard>::unlock
> ffffffff817b6cd0 T <kernel::sync::rcu::Guard as core::ops::drop::Drop>::drop
> ffffffff817b6c90 T <kernel::sync::rcu::Guard as core::default::Default>::default
> 
> These Rust symbols are basically wrappers around functions
> "rcu_read_lock" and "rcu_read_unlock". Marking them as inline can
> reduce the generation of these symbols, and saves the size of code
> generation for 100 bytes.
> 
> $ ./scripts/bloat-o-meter vmlinux_old vmlinux_new
> add/remove: 0/8 grow/shrink: 0/0 up/down: 0/-100 (-100)
> Function                                     old     new   delta
> _RNvXs_NtNtCsaYBeKL739Xz_6kernel4sync3rcuNtB4_5GuardNtNtCsdaXADs8PRFB_4core7default7Default7default       9       -      -9
> _RNvXs0_NtNtCsaYBeKL739Xz_6kernel4sync3rcuNtB5_5GuardNtNtNtCsdaXADs8PRFB_4core3ops4drop4Drop4drop       9       -      -9
> _RNvMNtNtCsaYBeKL739Xz_6kernel4sync3rcuNtB2_5Guard6unlock       9       -      -9
> _RNvMNtNtCsaYBeKL739Xz_6kernel4sync3rcuNtB2_5Guard3new       9       -      -9
> __pfx__RNvXs_NtNtCsaYBeKL739Xz_6kernel4sync3rcuNtB4_5GuardNtNtCsdaXADs8PRFB_4core7default7Default7default      16       -     -16
> __pfx__RNvXs0_NtNtCsaYBeKL739Xz_6kernel4sync3rcuNtB5_5GuardNtNtNtCsdaXADs8PRFB_4core3ops4drop4Drop4drop      16       -     -16
> __pfx__RNvMNtNtCsaYBeKL739Xz_6kernel4sync3rcuNtB2_5Guard6unlock      16       -     -16
> __pfx__RNvMNtNtCsaYBeKL739Xz_6kernel4sync3rcuNtB2_5Guard3new      16       -     -16
> Total: Before=23385830, After=23385730, chg -0.00%
> 
> Link: https://github.com/Rust-for-Linux/linux/issues/1145
> Signed-off-by: I Hsin Cheng <richard120310@gmail.com>
> ---Reviewed-by: Christian Schrefl <chrisi.schrefl@gmail.com>

Cheers
Christian

Re: [PATCH] rust: sync: rcu: Mark Guard methods as inline

[Miguel Ojeda]

On Thu, Mar 13, 2025 at 5:30 AM Joel Fernandes <joelagnelf@nvidia.com> wrote:
>
> Reviewed-by: Joel Fernandes <joelagnelf@nvidia.com>
>
> If no objections, I can queue it for 6.16 but let me know if Rust maintainers
> prefer to take it.

Thanks! That would be great -- FWIW:

Acked-by: Miguel Ojeda <ojeda@kernel.org>

By the way, should `read_lock()` be marked too? It just calls another
(now `#[inline]` one).

Cheers,
Miguel

Re: [PATCH] rust: sync: rcu: Mark Guard methods as inline

[I Hsin Cheng]

On Fri, Mar 14, 2025 at 03:43:23PM +0100, Miguel Ojeda wrote:
> On Thu, Mar 13, 2025 at 5:30 AM Joel Fernandes <joelagnelf@nvidia.com> wrote:
> >
> > Reviewed-by: Joel Fernandes <joelagnelf@nvidia.com>
> >
> > If no objections, I can queue it for 6.16 but let me know if Rust maintainers
> > prefer to take it.
> 
> Thanks! That would be great -- FWIW:
> 
> Acked-by: Miguel Ojeda <ojeda@kernel.org>
> 
> By the way, should `read_lock()` be marked too? It just calls another
> (now `#[inline]` one).
> 
> Cheers,
> Miguel

Hi all,

Thanks for your kindly review and comments!

> By the way, should `read_lock()` be marked too? It just calls another
> (now `#[inline]` one).

I'll do it and fix the typo mentioned by Charalampos. I'll send v2 to
you ASAP. Thanks !

Best regards,
I Hsin Cheng