riscv: Add support for BATCHED_UNMAP_TLB_FLUSH

Allow to defer the flushing of the TLB when unmapping pages, which allows
to reduce the numbers of IPI and the number of sfence.vma.

The ubenchmarch used in commit 43b3dfdd04 ("arm64: support
batched/deferred tlb shootdown during page reclamation/migration") that
was multithreaded to force the usage of IPI shows good performance
improvement on all platforms:

* Unmatched: ~34%
* TH1520   : ~78%
* Qemu     : ~81%

In addition, perf on qemu reports an important decrease in time spent
dealing with IPIs:

Before:  68.17%  main     [kernel.kallsyms]            [k] __sbi_rfence_v02_call
After :   8.64%  main     [kernel.kallsyms]            [k] __sbi_rfence_v02_call

* Benchmark:

int stick_this_thread_to_core(int core_id) {
        int num_cores = sysconf(_SC_NPROCESSORS_ONLN);
        if (core_id < 0 || core_id >= num_cores)
           return EINVAL;

        cpu_set_t cpuset;
        CPU_ZERO(&cpuset);
        CPU_SET(core_id, &cpuset);

        pthread_t current_thread = pthread_self();
        return pthread_setaffinity_np(current_thread,
sizeof(cpu_set_t), &cpuset);
}

static void *fn_thread (void *p_data)
{
        int ret;
        pthread_t thread;

        stick_this_thread_to_core((int)p_data);

        while (1) {
                sleep(1);
        }

        return NULL;
}

int main()
{
        volatile unsigned char *p = mmap(NULL, SIZE, PROT_READ | PROT_WRITE,
                                         MAP_SHARED | MAP_ANONYMOUS, -1, 0);
        pthread_t threads[4];
        int ret;

        for (int i = 0; i < 4; ++i) {
                ret = pthread_create(&threads[i], NULL, fn_thread, (void *)i);
                if (ret)
                {
                        printf("%s", strerror (ret));
                }
        }

        memset(p, 0x88, SIZE);

        for (int k = 0; k < 10000; k++) {
                /* swap in */
                for (int i = 0; i < SIZE; i += 4096) {
                        (void)p[i];
                }

                /* swap out */
                madvise(p, SIZE, MADV_PAGEOUT);
        }

        for (int i = 0; i < 4; i++)
        {
                pthread_cancel(threads[i]);
        }

        for (int i = 0; i < 4; i++)
        {
                pthread_join(threads[i], NULL);
        }

        return 0;
}

Signed-off-by: Alexandre Ghiti <alexghiti@rivosinc.com>
Reviewed-by: Jisheng Zhang <jszhang@kernel.org>
Tested-by: Jisheng Zhang <jszhang@kernel.org> # Tested on TH1520
Tested-by: Nam Cao <namcao@linutronix.de>
Link: https://lore.kernel.org/r/20240108193640.344929-1-alexghiti@rivosinc.com
Signed-off-by: Palmer Dabbelt <palmer@rivosinc.com>

Documentation/features/vm/TLB/arch-support.txt

@@ -20,7 +20,7 @@
     |    openrisc: |  ..  |
     |      parisc: | TODO |
     |     powerpc: | TODO |
-    |       riscv: | TODO |
+    |       riscv: |  ok  |
     |        s390: | TODO |
     |          sh: | TODO |
     |       sparc: | TODO |

arch/riscv/Kconfig

@@ -53,6 +53,7 @@ config RISCV
 	select ARCH_USE_MEMTEST
 	select ARCH_USE_QUEUED_RWLOCKS
 	select ARCH_USES_CFI_TRAPS if CFI_CLANG
+	select ARCH_WANT_BATCHED_UNMAP_TLB_FLUSH if SMP && MMU
 	select ARCH_WANT_DEFAULT_TOPDOWN_MMAP_LAYOUT if MMU
 	select ARCH_WANT_FRAME_POINTERS
 	select ARCH_WANT_GENERAL_HUGETLB if !RISCV_ISA_SVNAPOT

arch/riscv/include/asm/tlbbatch.h

@@ -0,0 +1,15 @@
+/* SPDX-License-Identifier: GPL-2.0-only */
+/*
+ * Copyright (C) 2023 Rivos Inc.
+ */
+
+#ifndef _ASM_RISCV_TLBBATCH_H
+#define _ASM_RISCV_TLBBATCH_H
+
+#include <linux/cpumask.h>
+
+struct arch_tlbflush_unmap_batch {
+	struct cpumask cpumask;
+};
+
+#endif /* _ASM_RISCV_TLBBATCH_H */

arch/riscv/include/asm/tlbflush.h

@@ -46,6 +46,14 @@ void flush_tlb_kernel_range(unsigned long start, unsigned long end);
 void flush_pmd_tlb_range(struct vm_area_struct *vma, unsigned long start,
 			unsigned long end);
 #endif
+
+bool arch_tlbbatch_should_defer(struct mm_struct *mm);
+void arch_tlbbatch_add_pending(struct arch_tlbflush_unmap_batch *batch,
+			       struct mm_struct *mm,
+			       unsigned long uaddr);
+void arch_flush_tlb_batched_pending(struct mm_struct *mm);
+void arch_tlbbatch_flush(struct arch_tlbflush_unmap_batch *batch);
+
 #else /* CONFIG_SMP && CONFIG_MMU */
 
 #define flush_tlb_all() local_flush_tlb_all()

arch/riscv/mm/tlbflush.c

@@ -93,29 +93,23 @@ static void __ipi_flush_tlb_range_asid(void *info)
 	local_flush_tlb_range_asid(d->start, d->size, d->stride, d->asid);
 }
 
-static void __flush_tlb_range(struct mm_struct *mm, unsigned long start,
-			      unsigned long size, unsigned long stride)
+static void __flush_tlb_range(struct cpumask *cmask, unsigned long asid,
+			      unsigned long start, unsigned long size,
+			      unsigned long stride)
 {
 	struct flush_tlb_range_data ftd;
-	const struct cpumask *cmask;
-	unsigned long asid = FLUSH_TLB_NO_ASID;
 	bool broadcast;
 
-	if (mm) {
-		unsigned int cpuid;
+	if (cpumask_empty(cmask))
+		return;
 
-		cmask = mm_cpumask(mm);
-		if (cpumask_empty(cmask))
-			return;
+	if (cmask != cpu_online_mask) {
+		unsigned int cpuid;
 
 		cpuid = get_cpu();
 		/* check if the tlbflush needs to be sent to other CPUs */
 		broadcast = cpumask_any_but(cmask, cpuid) < nr_cpu_ids;
-
-		if (static_branch_unlikely(&use_asid_allocator))
-			asid = atomic_long_read(&mm->context.id) & asid_mask;
 	} else {
-		cmask = cpu_online_mask;
 		broadcast = true;
 	}
 
@@ -135,25 +129,34 @@ static void __flush_tlb_range(struct mm_struct *mm, unsigned long start,
 		local_flush_tlb_range_asid(start, size, stride, asid);
 	}
 
-	if (mm)
+	if (cmask != cpu_online_mask)
 		put_cpu();
 }
 
+static inline unsigned long get_mm_asid(struct mm_struct *mm)
+{
+	return static_branch_unlikely(&use_asid_allocator) ?
+			atomic_long_read(&mm->context.id) & asid_mask : FLUSH_TLB_NO_ASID;
+}
+
 void flush_tlb_mm(struct mm_struct *mm)
 {
-	__flush_tlb_range(mm, 0, FLUSH_TLB_MAX_SIZE, PAGE_SIZE);
+	__flush_tlb_range(mm_cpumask(mm), get_mm_asid(mm),
+			  0, FLUSH_TLB_MAX_SIZE, PAGE_SIZE);
 }
 
 void flush_tlb_mm_range(struct mm_struct *mm,
 			unsigned long start, unsigned long end,
 			unsigned int page_size)
 {
-	__flush_tlb_range(mm, start, end - start, page_size);
+	__flush_tlb_range(mm_cpumask(mm), get_mm_asid(mm),
+			  start, end - start, page_size);
 }
 
 void flush_tlb_page(struct vm_area_struct *vma, unsigned long addr)
 {
-	__flush_tlb_range(vma->vm_mm, addr, PAGE_SIZE, PAGE_SIZE);
+	__flush_tlb_range(mm_cpumask(vma->vm_mm), get_mm_asid(vma->vm_mm),
+			  addr, PAGE_SIZE, PAGE_SIZE);
 }
 
 void flush_tlb_range(struct vm_area_struct *vma, unsigned long start,
@@ -185,18 +188,44 @@ void flush_tlb_range(struct vm_area_struct *vma, unsigned long start,
 		}
 	}
 
-	__flush_tlb_range(vma->vm_mm, start, end - start, stride_size);
+	__flush_tlb_range(mm_cpumask(vma->vm_mm), get_mm_asid(vma->vm_mm),
+			  start, end - start, stride_size);
 }
 
 void flush_tlb_kernel_range(unsigned long start, unsigned long end)
 {
-	__flush_tlb_range(NULL, start, end - start, PAGE_SIZE);
+	__flush_tlb_range((struct cpumask *)cpu_online_mask, FLUSH_TLB_NO_ASID,
+			  start, end - start, PAGE_SIZE);
 }
 
 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
 void flush_pmd_tlb_range(struct vm_area_struct *vma, unsigned long start,
 			unsigned long end)
 {
-	__flush_tlb_range(vma->vm_mm, start, end - start, PMD_SIZE);
+	__flush_tlb_range(mm_cpumask(vma->vm_mm), get_mm_asid(vma->vm_mm),
+			  start, end - start, PMD_SIZE);
 }
 #endif
+
+bool arch_tlbbatch_should_defer(struct mm_struct *mm)
+{
+	return true;
+}
+
+void arch_tlbbatch_add_pending(struct arch_tlbflush_unmap_batch *batch,
+			       struct mm_struct *mm,
+			       unsigned long uaddr)
+{
+	cpumask_or(&batch->cpumask, &batch->cpumask, mm_cpumask(mm));
+}
+
+void arch_flush_tlb_batched_pending(struct mm_struct *mm)
+{
+	flush_tlb_mm(mm);
+}
+
+void arch_tlbbatch_flush(struct arch_tlbflush_unmap_batch *batch)
+{
+	__flush_tlb_range(&batch->cpumask, FLUSH_TLB_NO_ASID, 0,
+			  FLUSH_TLB_MAX_SIZE, PAGE_SIZE);
+}