percpu: Introduce percpu hot section

Add a subsection to the percpu data for frequently accessed variables
that should remain cached on each processor.  These varables should not
be accessed from other processors to avoid cacheline bouncing.

This will replace the pcpu_hot struct on x86, and open up similar
functionality to other architectures and the kernel core.

Signed-off-by: Brian Gerst <brgerst@gmail.com>
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Acked-by: Uros Bizjak <ubizjak@gmail.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Link: https://lore.kernel.org/r/20250303165246.2175811-2-brgerst@gmail.com

diff --git a/include/asm-generic/vmlinux.lds.h b/include/asm-generic/vmlinux.lds.h
index b32e453acf9d3fdcdce94a7bc7b54100429ec432..c4e8fac502543ee605fddd8fec57f368bfdec437 100644 (file)
--- a/include/asm-generic/vmlinux.lds.h
+++ b/include/asm-generic/vmlinux.lds.h
@@ -385,6 +385,11 @@ defined(CONFIG_AUTOFDO_CLANG) || defined(CONFIG_PROPELLER_CLANG)
        . = ALIGN(PAGE_SIZE);                                           \
        __nosave_end = .;
 
+#define CACHE_HOT_DATA(align)                                          \
+       . = ALIGN(align);                                               \
+       *(SORT_BY_ALIGNMENT(.data..hot.*))                              \
+       . = ALIGN(align);
+
 #define PAGE_ALIGNED_DATA(page_align)                                  \
        . = ALIGN(page_align);                                          \
        *(.data..page_aligned)                                          \
@@ -1065,6 +1070,11 @@ defined(CONFIG_AUTOFDO_CLANG) || defined(CONFIG_PROPELLER_CLANG)
        . = ALIGN(PAGE_SIZE);                                           \
        *(.data..percpu..page_aligned)                                  \
        . = ALIGN(cacheline);                                           \
+       __per_cpu_hot_start = .;                                        \
+       *(SORT_BY_ALIGNMENT(.data..percpu..hot.*))                      \
+       __per_cpu_hot_pad = .;                                          \
+       . = ALIGN(cacheline);                                           \
+       __per_cpu_hot_end = .;                                          \
        *(.data..percpu..read_mostly)                                   \
        . = ALIGN(cacheline);                                           \
        *(.data..percpu)                                                \
@@ -1112,6 +1122,7 @@ defined(CONFIG_AUTOFDO_CLANG) || defined(CONFIG_PROPELLER_CLANG)
                INIT_TASK_DATA(inittask)                                \
                NOSAVE_DATA                                             \
                PAGE_ALIGNED_DATA(pagealigned)                          \
+               CACHE_HOT_DATA(cacheline)                               \
                CACHELINE_ALIGNED_DATA(cacheline)                       \
                READ_MOSTLY_DATA(cacheline)                             \
                DATA_DATA                                               \

diff --git a/include/linux/percpu-defs.h b/include/linux/percpu-defs.h
index 40d34e032d5b8a4ba98699871f17d18288dbf15e..0fcacb9097789eab68bddeab22f313bfb802ebe9 100644 (file)
--- a/include/linux/percpu-defs.h
+++ b/include/linux/percpu-defs.h
@@ -112,6 +112,19 @@
 #define DEFINE_PER_CPU(type, name)                                     \
        DEFINE_PER_CPU_SECTION(type, name, "")
 
+/*
+ * Declaration/definition used for per-CPU variables that are frequently
+ * accessed and should be in a single cacheline.
+ *
+ * For use only by architecture and core code.  Only use scalar or pointer
+ * types to maximize density.
+ */
+#define DECLARE_PER_CPU_CACHE_HOT(type, name)                          \
+       DECLARE_PER_CPU_SECTION(type, name, "..hot.." #name)
+
+#define DEFINE_PER_CPU_CACHE_HOT(type, name)                           \
+       DEFINE_PER_CPU_SECTION(type, name, "..hot.." #name)
+
 /*
  * Declaration/definition used for per-CPU variables that must be cacheline
  * aligned under SMP conditions so that, whilst a particular instance of the