Re: [PATCH] RISC-V: Issue a local tlb flush if possible.

From: Anup Patel <anup@brainfault.org>
To: Atish Patra <atish.patra@wdc.com>
Cc: "linux-kernel@vger.kernel.org List"
	<linux-kernel@vger.kernel.org>, Albert Ou <aou@eecs.berkeley.edu>,
	Alexios Zavras <alexios.zavras@intel.com>,
	Greg Kroah-Hartman <gregkh@linuxfoundation.org>,
	Palmer Dabbelt <palmer@sifive.com>,
	Anup Patel <anup.patel@wdc.com>,
	Paul Walmsley <paul.walmsley@sifive.com>,
	linux-riscv <linux-riscv@lists.infradead.org>,
	Allison Randal <allison@lohutok.net>
Subject: Re: [PATCH] RISC-V: Issue a local tlb flush if possible.
Date: Sat, 10 Aug 2019 09:00:29 +0530	[thread overview]
Message-ID: <CAAhSdy1bnBoOdYJHm97JyG5oiY6PuLqamedx4BnfbrhVvmv6Xw@mail.gmail.com> (raw)
In-Reply-To: <20190810014309.20838-1-atish.patra@wdc.com>

On Sat, Aug 10, 2019 at 7:13 AM Atish Patra <atish.patra@wdc.com> wrote:
>
> In RISC-V, tlb flush happens via SBI which is expensive.
> If the target cpumask contains a local hartid, some cost
> can be saved by issuing a local tlb flush as we do that
> in OpenSBI anyways.
>
> Signed-off-by: Atish Patra <atish.patra@wdc.com>
> ---
>  arch/riscv/include/asm/tlbflush.h | 33 +++++++++++++++++++++++++++----
>  1 file changed, 29 insertions(+), 4 deletions(-)
>
> diff --git a/arch/riscv/include/asm/tlbflush.h b/arch/riscv/include/asm/tlbflush.h
> index 687dd19735a7..b32ba4fa5888 100644
> --- a/arch/riscv/include/asm/tlbflush.h
> +++ b/arch/riscv/include/asm/tlbflush.h
> @@ -8,6 +8,7 @@
>  #define _ASM_RISCV_TLBFLUSH_H
>
>  #include <linux/mm_types.h>
> +#include <linux/sched.h>
>  #include <asm/smp.h>
>
>  /*
> @@ -46,14 +47,38 @@ static inline void remote_sfence_vma(struct cpumask *cmask, unsigned long start,
>                                      unsigned long size)
>  {
>         struct cpumask hmask;
> +       struct cpumask tmask;
> +       int cpuid = smp_processor_id();
>
>         cpumask_clear(&hmask);
> -       riscv_cpuid_to_hartid_mask(cmask, &hmask);
> -       sbi_remote_sfence_vma(hmask.bits, start, size);
> +       cpumask_clear(&tmask);
> +
> +       if (cmask)
> +               cpumask_copy(&tmask, cmask);
> +       else
> +               cpumask_copy(&tmask, cpu_online_mask);

This can be further simplified.

We can totally avoid tmask, cpumask_copy(), and cpumask_clear()
by directly updating hmask.

In addition to this patch, we should also handle the case of
empty hart mask in OpenSBI.

> +
> +       if (cpumask_test_cpu(cpuid, &tmask)) {
> +               /* Save trap cost by issuing a local tlb flush here */
> +               if ((start == 0 && size == -1) || (size > PAGE_SIZE))
> +                       local_flush_tlb_all();
> +               else if (size == PAGE_SIZE)
> +                       local_flush_tlb_page(start);
> +               cpumask_clear_cpu(cpuid, &tmask);
> +       } else if (cpumask_empty(&tmask)) {
> +               /* cpumask is empty. So just do a local flush */
> +               local_flush_tlb_all();
> +               return;
> +       }
> +
> +       if (!cpumask_empty(&tmask)) {
> +               riscv_cpuid_to_hartid_mask(&tmask, &hmask);
> +               sbi_remote_sfence_vma(hmask.bits, start, size);
> +       }
>  }
>
> -#define flush_tlb_all() sbi_remote_sfence_vma(NULL, 0, -1)
> -#define flush_tlb_page(vma, addr) flush_tlb_range(vma, addr, 0)
> +#define flush_tlb_all() remote_sfence_vma(NULL, 0, -1)
> +#define flush_tlb_page(vma, addr) flush_tlb_range(vma, addr, (addr) + PAGE_SIZE)
>  #define flush_tlb_range(vma, start, end) \
>         remote_sfence_vma(mm_cpumask((vma)->vm_mm), start, (end) - (start))
>  #define flush_tlb_mm(mm) \
> --
> 2.21.0
>
>
> _______________________________________________
> linux-riscv mailing list
> linux-riscv@lists.infradead.org
> http://lists.infradead.org/mailman/listinfo/linux-riscv

Regards,
Anup