[PATCH v6 7/8] x86/vdso: Introduce helper functions for CPU and node number

["Chang S. Bae" <chang.seok.bae@intel.com>]

The CPU initialization in vDSO is now a bit cleaned up by
the new helper functions. The helper functions will take
care of combining CPU and node number and reading each from
the combined value.

Suggested-by: Andy Lutomirski <luto@kernel.org>
Suggested-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Chang S. Bae <chang.seok.bae@intel.com>
Cc: H. Peter Anvin <hpa@zytor.com>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: Dave Hansen <dave.hansen@linux.intel.com>
---
 arch/x86/entry/vdso/vgetcpu.c  |  9 +--------
 arch/x86/entry/vdso/vma.c      | 19 +++++++------------
 arch/x86/include/asm/segment.h | 41 +++++++++++++++++++++++++++++++++++++++++
 arch/x86/include/asm/vgtod.h   | 26 --------------------------
 4 files changed, 49 insertions(+), 46 deletions(-)

diff --git a/arch/x86/entry/vdso/vgetcpu.c b/arch/x86/entry/vdso/vgetcpu.c
index 8ec3d1f..de78fc9 100644
--- a/arch/x86/entry/vdso/vgetcpu.c
+++ b/arch/x86/entry/vdso/vgetcpu.c
@@ -13,14 +13,7 @@
 notrace long
 __vdso_getcpu(unsigned *cpu, unsigned *node, struct getcpu_cache *unused)
 {
-	unsigned int p;
-
-	p = __getcpu();
-
-	if (cpu)
-		*cpu = p & VGETCPU_CPU_MASK;
-	if (node)
-		*node = p >> 12;
+	vdso_read_cpu_node(cpu, node);
 	return 0;
 }
 
diff --git a/arch/x86/entry/vdso/vma.c b/arch/x86/entry/vdso/vma.c
index 0b114aa..39b5584 100644
--- a/arch/x86/entry/vdso/vma.c
+++ b/arch/x86/entry/vdso/vma.c
@@ -339,20 +339,15 @@ static void vgetcpu_cpu_init(void *arg)
 {
 	int cpu = smp_processor_id();
 	struct desc_struct d = { };
-	unsigned long node = 0;
-#ifdef CONFIG_NUMA
-	node = cpu_to_node(cpu);
-#endif
+	unsigned long cpudata = vdso_encode_cpu_node(cpu, cpu_to_node(cpu));
+
 	if (static_cpu_has(X86_FEATURE_RDTSCP))
-		write_rdtscp_aux((node << 12) | cpu);
+		write_rdtscp_aux(cpudata);
+
+	/* Store CPU and node number in limit */
+	d.limit0 = cpudata;
+	d.limit1 = cpudata >> 16;
 
-	/*
-	 * Store cpu number in limit so that it can be loaded
-	 * quickly in user space in vgetcpu. (12 bits for the CPU
-	 * and 8 bits for the node)
-	 */
-	d.limit0 = cpu | ((node & 0xf) << 12);
-	d.limit1 = node >> 4;
 	d.type = 5;		/* RO data, expand down, accessed */
 	d.dpl = 3;		/* Visible to user code */
 	d.s = 1;		/* Not a system segment */
diff --git a/arch/x86/include/asm/segment.h b/arch/x86/include/asm/segment.h
index 3cb2aa5..d4079bd 100644
--- a/arch/x86/include/asm/segment.h
+++ b/arch/x86/include/asm/segment.h
@@ -224,6 +224,47 @@
 #define GDT_ENTRY_TLS_ENTRIES		3
 #define TLS_SIZE			(GDT_ENTRY_TLS_ENTRIES* 8)
 
+#ifdef CONFIG_X86_64
+
+/* Bit size and mask of CPU number stored in the per CPU data (and TSC_AUX) */
+#define VDSO_CPU_SIZE			12
+#define VDSO_CPU_MASK			0xfff
+
+#ifndef __ASSEMBLY__
+
+/* Helper functions to store/load CPU and node numbers */
+
+static inline unsigned long vdso_encode_cpu_node(int cpu, unsigned long node)
+{
+	return ((node << VDSO_CPU_SIZE) | cpu);
+}
+
+static inline void vdso_read_cpu_node(unsigned *cpu, unsigned *node)
+{
+	unsigned int p;
+
+	/*
+	 * Load CPU and node number from GDT.  LSL is faster than RDTSCP
+	 * and works on all CPUs.  This is volatile so that it orders
+	 * correctly with respect to barrier() and to keep GCC from cleverly
+	 * hoisting it out of the calling function.
+	 *
+	 * If RDPID is available, use it.
+	 */
+	alternative_io ("lsl %[seg],%[p]",
+			".byte 0xf3,0x0f,0xc7,0xf8", /* RDPID %eax/rax */
+			X86_FEATURE_RDPID,
+			[p] "=a" (p), [seg] "r" (__CPU_NUMBER_SEG));
+
+	if (cpu)
+		*cpu = (p & VDSO_CPU_MASK);
+	if (node)
+		*node = (p >> VDSO_CPU_SIZE);
+}
+
+#endif /* !__ASSEMBLY__ */
+#endif /* CONFIG_X86_64 */
+
 #ifdef __KERNEL__
 
 /*
diff --git a/arch/x86/include/asm/vgtod.h b/arch/x86/include/asm/vgtod.h
index 4e81ea9..056a61c 100644
--- a/arch/x86/include/asm/vgtod.h
+++ b/arch/x86/include/asm/vgtod.h
@@ -77,30 +77,4 @@ static inline void gtod_write_end(struct vsyscall_gtod_data *s)
 	++s->seq;
 }
 
-#ifdef CONFIG_X86_64
-
-#define VGETCPU_CPU_MASK 0xfff
-
-static inline unsigned int __getcpu(void)
-{
-	unsigned int p;
-
-	/*
-	 * Load CPU (and node) number from GDT.  LSL is faster than RDTSCP
-	 * and works on all CPUs.  This is volatile so that it orders
-	 * correctly with respect to barrier() and to keep GCC from cleverly
-	 * hoisting it out of the calling function.
-	 *
-	 * If RDPID is available, use it.
-	 */
-	alternative_io ("lsl %[seg],%[p]",
-			".byte 0xf3,0x0f,0xc7,0xf8", /* RDPID %eax/rax */
-			X86_FEATURE_RDPID,
-			[p] "=a" (p), [seg] "r" (__CPU_NUMBER_SEG));
-
-	return p;
-}
-
-#endif /* CONFIG_X86_64 */
-
 #endif /* _ASM_X86_VGTOD_H */
-- 
2.7.4