[PATCH 2/2] riscv: add ASID-based tlbflushing methods

[Guo Ren]

Rebase with "THP supprt for RISCV" & "Add DMA_COHERENT v2" on linux-5.13-rc4.

Tested-by: Guo Ren <guoren at kernel.org>

On Sun, Jun 6, 2021 at 11:21 PM Christoph Hellwig <hch at lst.de> wrote:
>
> From: Guo Ren <guoren at linux.alibaba.com>
>
> Implement optimized version of the tlb flushing routines for systems
> using ASIDs. These are behind the use_asid_allocator static branch to
> not affect existing systems not using ASIDs.
>
> Signed-off-by: Guo Ren <guoren at linux.alibaba.com>
> [hch: rebased on top of previous cleanups, use the same algorithm as
>       the non-ASID based code for local vs global flushes, keep functions
>       as local as possible]
> Signed-off-by: Christoph Hellwig <hch at lst.de>
> ---
>  arch/riscv/include/asm/mmu_context.h |  2 ++
>  arch/riscv/mm/context.c              |  2 +-
>  arch/riscv/mm/tlbflush.c             | 47 +++++++++++++++++++++++-----
>  3 files changed, 43 insertions(+), 8 deletions(-)
>
> diff --git a/arch/riscv/include/asm/mmu_context.h b/arch/riscv/include/asm/mmu_context.h
> index b0659413a080..7030837adc1a 100644
> --- a/arch/riscv/include/asm/mmu_context.h
> +++ b/arch/riscv/include/asm/mmu_context.h
> @@ -33,6 +33,8 @@ static inline int init_new_context(struct task_struct *tsk,
>         return 0;
>  }
>
> +DECLARE_STATIC_KEY_FALSE(use_asid_allocator);
> +
>  #include <asm-generic/mmu_context.h>
>
>  #endif /* _ASM_RISCV_MMU_CONTEXT_H */
> diff --git a/arch/riscv/mm/context.c b/arch/riscv/mm/context.c
> index 9bc46ab01c25..6ed696bad558 100644
> --- a/arch/riscv/mm/context.c
> +++ b/arch/riscv/mm/context.c
> @@ -18,7 +18,7 @@
>
>  #ifdef CONFIG_MMU
>
> -static DEFINE_STATIC_KEY_FALSE(use_asid_allocator);
> +DEFINE_STATIC_KEY_FALSE(use_asid_allocator);
>
>  static unsigned long asid_bits;
>  static unsigned long num_asids;
> diff --git a/arch/riscv/mm/tlbflush.c b/arch/riscv/mm/tlbflush.c
> index b458949fa8df..64f8201237c2 100644
> --- a/arch/riscv/mm/tlbflush.c
> +++ b/arch/riscv/mm/tlbflush.c
> @@ -4,6 +4,24 @@
>  #include <linux/smp.h>
>  #include <linux/sched.h>
>  #include <asm/sbi.h>
> +#include <asm/mmu_context.h>
> +
> +static inline void local_flush_tlb_all_asid(unsigned long asid)
> +{
> +       __asm__ __volatile__ ("sfence.vma x0, %0"
> +                       :
> +                       : "r" (asid)
> +                       : "memory");
> +}
> +
> +static inline void local_flush_tlb_page_asid(unsigned long addr,
> +               unsigned long asid)
> +{
> +       __asm__ __volatile__ ("sfence.vma %0, %1"
> +                       :
> +                       : "r" (addr), "r" (asid)
> +                       : "memory");
> +}
>
>  void flush_tlb_all(void)
>  {
> @@ -16,21 +34,36 @@ static void __sbi_tlb_flush_range(struct mm_struct *mm, unsigned long start,
>         struct cpumask *cmask = mm_cpumask(mm);
>         struct cpumask hmask;
>         unsigned int cpuid;
> +       bool broadcast;
>
>         if (cpumask_empty(cmask))
>                 return;
>
>         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)) {
> +               unsigned long asid = atomic_long_read(&mm->context.id);
>
> -       if (cpumask_any_but(cmask, cpuid) >= nr_cpu_ids) {
> -               /* local cpu is the only cpu present in cpumask */
> -               if (size <= stride)
> +               if (broadcast) {
> +                       riscv_cpuid_to_hartid_mask(cmask, &hmask);
> +                       sbi_remote_sfence_vma_asid(cpumask_bits(&hmask),
> +                                                  start, size, asid);
> +               } else if (size <= stride) {
> +                       local_flush_tlb_page_asid(start, asid);
> +               } else {
> +                       local_flush_tlb_all_asid(asid);
> +               }
It will reduce efficiency in our system, but I'll add
"local_flush_tlb_range_asid(start, start+size, stride, size)" later.

> +       } else {
> +               if (broadcast) {
> +                       riscv_cpuid_to_hartid_mask(cmask, &hmask);
> +                       sbi_remote_sfence_vma(cpumask_bits(&hmask),
> +                                             start, size);
> +               } else if (size <= stride) {
This logic is from the Atish's patch.
https://git.kernel.org/pub/scm/linux/kernel/git/next/linux-next.git/commit/arch/riscv/mm/tlbflush.c?h=next-20210604&id=6efb16b1d5514865d0f7a01910648568ad3225d8

I don't know why not use the range? Maybe fix some hardware issues.

>                         local_flush_tlb_page(start);
> -               else
> +               } else {
>                         local_flush_tlb_all();
> -       } else {
> -               riscv_cpuid_to_hartid_mask(cmask, &hmask);
> -               sbi_remote_sfence_vma(cpumask_bits(&hmask), start, size);
> +               }
>         }
>
>         put_cpu();
> --
> 2.30.2
>


-- 
Best Regards
 Guo Ren

ML: https://lore.kernel.org/linux-csky/