[RFC,v2,05/11] RISC-V: Use IPIs for remote TLB flush when possible

Commit Message

Anup Patel June 18, 2021, 12:38 p.m. UTC

If IPI calls are injected using SBI IPI calls then remote TLB flush
using SBI RFENCE calls is much faster because using IPIs for remote
TLB flush would still endup as SBI IPI calls with extra processing
on kernel side.

It is now possible to have specialized hardware (such as RISC-V AIA)
which allows S-mode software to directly inject IPIs without any
assistance from M-mode runtime firmware.

This patch extends remote TLB flush functions to use IPIs whenever
underlying IPI operations are suitable for remote FENCEs.

Signed-off-by: Anup Patel <[e-mail]>
---
 arch/riscv/mm/tlbflush.c | 62 +++++++++++++++++++++++++++++++---------
 1 file changed, 48 insertions(+), 14 deletions(-)

Patch

diff --git a/arch/riscv/mm/tlbflush.c b/arch/riscv/mm/tlbflush.c
index 720b443c4528..009c56fa102d 100644
--- a/arch/riscv/mm/tlbflush.c
+++ b/arch/riscv/mm/tlbflush.c
@@ -1,39 +1,73 @@ 
 // SPDX-License-Identifier: GPL-2.0
+/*
+ * TLB flush implementation.
+ *
+ * Copyright (c) 2021 Western Digital Corporation or its affiliates.
+ */
 
 #include <linux/mm.h>
 #include <linux/smp.h>
 #include <linux/sched.h>
 #include <asm/sbi.h>
 
+static void ipi_flush_tlb_all(void *info)
+{
+	local_flush_tlb_all();
+}
+
 void flush_tlb_all(void)
 {
-	sbi_remote_sfence_vma(NULL, 0, -1);
+	if (!riscv_use_ipi_for_rfence())
+		sbi_remote_sfence_vma(NULL, 0, -1);
+	else
+		on_each_cpu(ipi_flush_tlb_all, NULL, 1);
+}
+
+struct flush_range_data {
+	unsigned long start;
+	unsigned long size;
+};
+
+static void ipi_flush_range(void *info)
+{
+	struct flush_range_data *data = info;
+
+	/* local cpu is the only cpu present in cpumask */
+	if (data->size <= PAGE_SIZE)
+		local_flush_tlb_page(data->start);
+	else
+		local_flush_tlb_all();
 }
 
 /*
- * This function must not be called with cmask being null.
+ * This function must not be called with NULL cpumask.
  * Kernel may panic if cmask is NULL.
  */
-static void __sbi_tlb_flush_range(struct cpumask *cmask, unsigned long start,
-				  unsigned long size)
+static void flush_range(struct cpumask *cmask, unsigned long start,
+			unsigned long size)
 {
+	struct flush_range_data info;
 	struct cpumask hmask;
 	unsigned int cpuid;
 
 	if (cpumask_empty(cmask))
 		return;
 
+	info.start = start;
+	info.size = size;
+
 	cpuid = get_cpu();
 
 	if (cpumask_any_but(cmask, cpuid) >= nr_cpu_ids) {
-		/* local cpu is the only cpu present in cpumask */
-		if (size <= PAGE_SIZE)
-			local_flush_tlb_page(start);
-		else
-			local_flush_tlb_all();
+		ipi_flush_range(&info);
 	} else {
-		riscv_cpuid_to_hartid_mask(cmask, &hmask);
-		sbi_remote_sfence_vma(cpumask_bits(&hmask), start, size);
+		if (!riscv_use_ipi_for_rfence()) {
+			riscv_cpuid_to_hartid_mask(cmask, &hmask);
+			sbi_remote_sfence_vma(cpumask_bits(&hmask),
+					      start, size);
+		} else {
+			on_each_cpu_mask(cmask, ipi_flush_range, &info, 1);
+		}
 	}
 
 	put_cpu();
@@ -41,16 +75,16 @@  static void __sbi_tlb_flush_range(struct cpumask *cmask, unsigned long start,
 
 void flush_tlb_mm(struct mm_struct *mm)
 {
-	__sbi_tlb_flush_range(mm_cpumask(mm), 0, -1);
+	flush_range(mm_cpumask(mm), 0, -1);
 }
 
 void flush_tlb_page(struct vm_area_struct *vma, unsigned long addr)
 {
-	__sbi_tlb_flush_range(mm_cpumask(vma->vm_mm), addr, PAGE_SIZE);
+	flush_range(mm_cpumask(vma->vm_mm), addr, PAGE_SIZE);
 }
 
 void flush_tlb_range(struct vm_area_struct *vma, unsigned long start,
 		     unsigned long end)
 {
-	__sbi_tlb_flush_range(mm_cpumask(vma->vm_mm), start, end - start);
+	flush_range(mm_cpumask(vma->vm_mm), start, end - start);
 }