Date:   Fri, 18 Oct 2019 08:25:17 +0530
From:   Anup Patel <anup@...infault.org>
To:     Christoph Hellwig <hch@....de>
Cc:     Palmer Dabbelt <palmer@...ive.com>,
        Paul Walmsley <paul.walmsley@...ive.com>,
        Damien Le Moal <damien.lemoal@....com>,
        linux-riscv <linux-riscv@...ts.infradead.org>,
        "linux-kernel@...r.kernel.org List" <linux-kernel@...r.kernel.org>
Subject: Re: [PATCH 07/15] riscv: implement remote sfence.i using IPIs

On Thu, Oct 17, 2019 at 11:08 PM Christoph Hellwig <hch@....de> wrote:
>
> The RISC-V ISA only supports flushing the instruction cache for the
> local CPU core.  Currently we always offload the remote TLB flushing to
> the SBI, which then issues an IPI under the hoods.  But with M-mode
> we do not have an SBI so we have to do it ourselves.   IPI to the
> other nodes using the existing kernel helpers instead if we have
> native clint support and thus can IPI directly from the kernel.
>
> Signed-off-by: Christoph Hellwig <hch@....de>
> ---
>  arch/riscv/include/asm/sbi.h |  3 +++
>  arch/riscv/mm/cacheflush.c   | 24 ++++++++++++++++++------
>  2 files changed, 21 insertions(+), 6 deletions(-)
>
> diff --git a/arch/riscv/include/asm/sbi.h b/arch/riscv/include/asm/sbi.h
> index b167af3e7470..0cb74eccc73f 100644
> --- a/arch/riscv/include/asm/sbi.h
> +++ b/arch/riscv/include/asm/sbi.h
> @@ -94,5 +94,8 @@ static inline void sbi_remote_sfence_vma_asid(const unsigned long *hart_mask,
>  {
>         SBI_CALL_4(SBI_REMOTE_SFENCE_VMA_ASID, hart_mask, start, size, asid);
>  }
> +#else /* CONFIG_RISCV_SBI */
> +/* stub to for code is only reachable under IS_ENABLED(CONFIG_RISCV_SBI): */
> +void sbi_remote_fence_i(const unsigned long *hart_mask);
>  #endif /* CONFIG_RISCV_SBI */
>  #endif /* _ASM_RISCV_SBI_H */
> diff --git a/arch/riscv/mm/cacheflush.c b/arch/riscv/mm/cacheflush.c
> index 3f15938dec89..794c9ab256eb 100644
> --- a/arch/riscv/mm/cacheflush.c
> +++ b/arch/riscv/mm/cacheflush.c
> @@ -10,9 +10,17 @@
>
>  #include <asm/sbi.h>
>
> +static void ipi_remote_fence_i(void *info)
> +{
> +       return local_flush_icache_all();
> +}
> +
>  void flush_icache_all(void)
>  {
> -       sbi_remote_fence_i(NULL);
> +       if (IS_ENABLED(CONFIG_RISCV_SBI))
> +               sbi_remote_fence_i(NULL);
> +       else
> +               on_each_cpu(ipi_remote_fence_i, NULL, 1);
>  }
>
>  /*
> @@ -28,7 +36,7 @@ void flush_icache_all(void)
>  void flush_icache_mm(struct mm_struct *mm, bool local)
>  {
>         unsigned int cpu;
> -       cpumask_t others, hmask, *mask;
> +       cpumask_t others, *mask;
>
>         preempt_disable();
>
> @@ -46,10 +54,7 @@ void flush_icache_mm(struct mm_struct *mm, bool local)
>          */
>         cpumask_andnot(&others, mm_cpumask(mm), cpumask_of(cpu));
>         local |= cpumask_empty(&others);
> -       if (mm != current->active_mm || !local) {
> -               riscv_cpuid_to_hartid_mask(&others, &hmask);
> -               sbi_remote_fence_i(hmask.bits);
> -       } else {
> +       if (mm == current->active_mm && local) {
>                 /*
>                  * It's assumed that at least one strongly ordered operation is
>                  * performed on this hart between setting a hart's cpumask bit
> @@ -59,6 +64,13 @@ void flush_icache_mm(struct mm_struct *mm, bool local)
>                  * with flush_icache_deferred().
>                  */
>                 smp_mb();
> +       } else if (IS_ENABLED(CONFIG_RISCV_SBI)) {
> +               cpumask_t hartid_mask;
> +
> +               riscv_cpuid_to_hartid_mask(&others, &hartid_mask);
> +               sbi_remote_fence_i(cpumask_bits(&hartid_mask));
> +       } else {
> +               on_each_cpu_mask(&others, ipi_remote_fence_i, NULL, 1);
>         }
>
>         preempt_enable();
> --
> 2.20.1
>
>
> _______________________________________________
> linux-riscv mailing list
> linux-riscv@...ts.infradead.org
> http://lists.infradead.org/mailman/listinfo/linux-riscv

LGTM.

Reviewed-by: Anup Patel <anup@...infault.org>

Regards,
Anup

Powered by blists - more mailing lists