Messages in this thread | | | From | Atish Patra <> | Subject | Re: [PATCH v3 1/3] RISC-V: Issue a local tlbflush if possible. | Date | Thu, 22 Aug 2019 04:01:24 +0000 |
| |
On Thu, 2019-08-22 at 03:46 +0200, Christoph Hellwig wrote: > On Wed, Aug 21, 2019 at 05:46:42PM -0700, Atish Patra wrote: > > In RISC-V, tlb flush happens via SBI which is expensive. If the > > local > > cpu is the only cpu in cpumask, there is no need to invoke a SBI > > call. > > > > Just do a local flush and return. > > > > Signed-off-by: Atish Patra <atish.patra@wdc.com> > > --- > > arch/riscv/mm/tlbflush.c | 15 +++++++++++++++ > > 1 file changed, 15 insertions(+) > > > > diff --git a/arch/riscv/mm/tlbflush.c b/arch/riscv/mm/tlbflush.c > > index df93b26f1b9d..36430ee3bed9 100644 > > --- a/arch/riscv/mm/tlbflush.c > > +++ b/arch/riscv/mm/tlbflush.c > > @@ -2,6 +2,7 @@ > > > > #include <linux/mm.h> > > #include <linux/smp.h> > > +#include <linux/sched.h> > > #include <asm/sbi.h> > > > > void flush_tlb_all(void) > > @@ -13,9 +14,23 @@ static void __sbi_tlb_flush_range(struct cpumask > > *cmask, unsigned long start, > > unsigned long size) > > { > > struct cpumask hmask; > > + unsigned int cpuid = get_cpu(); > > > > + if (!cmask) { > > + riscv_cpuid_to_hartid_mask(cpu_online_mask, &hmask); > > + goto issue_sfence; > > + } > > + > > + if (cpumask_test_cpu(cpuid, cmask) && cpumask_weight(cmask) == > > 1) { > > + local_flush_tlb_all(); > > + goto done; > > + } > > I think a single core on a SMP kernel is a valid enough use case > given > how litte distros still have UP kernels. So Maybe this shiuld rather > be: > > if (!cmask) > cmask = cpu_online_mask; > > if (cpumask_test_cpu(cpuid, cmask) && cpumask_weight(cmask) == > 1) { > local_flush_tlb_all(); > } else { > riscv_cpuid_to_hartid_mask(cmask, &hmask); > sbi_remote_sfence_vma(hmask.bits, start, size); > }
The downside of this is that for every !cmask case in true SMP (more common probably) it will execute 2 extra cpumask instructions. As tlbflush path is in performance critical path, I think we should favor more common case (SMP with more than 1 core).
Thoughts ?
-- Regards, Atish
| |