On Mon, Jun 17, 2019 at 11:32:55PM +0900, Takao Indoh wrote:
> From: Takao Indoh <indou.takao@xxxxxxxxxxx>
>
> This patch adds new boot parameter 'disable_tlbflush_is' to disable TLB
> flush within the same inner shareable domain for performance tuning.
>
> In the case of flush_tlb_mm() *without* this parameter, TLB entry is
> invalidated by __tlbi(aside1is, asid). By this instruction, all CPUs within
> the same inner shareable domain check if there are TLB entries which have
> this ASID, this causes performance noise, especially at large-scale HPC
> environment, which has more than thousand nodes with low latency
> interconnect.
>
> When this new parameter is specified, TLB entry is invalidated by
> __tlbi(aside1, asid) only on the CPUs specified by mm_cpumask(mm).
> Therefore TLB flush is done on minimal CPUs and performance problem does
> not occur.
>
> Signed-off-by: QI Fuli <qi.fuli@xxxxxxxxxxx>
> Signed-off-by: Takao Indoh <indou.takao@xxxxxxxxxxx>
[...]
> +void flush_tlb_mm(struct mm_struct *mm)
> +{
> + if (disable_tlbflush_is)
> + on_each_cpu_mask(mm_cpumask(mm), ipi_flush_tlb_mm,
> + (void *)mm, true);
> + else
> + __flush_tlb_mm(mm);
> +}
Could we try instead to call a _nosync variant here when
cpumask_weight() is 1 or the *IS if greater than 1 and avoid the IPI?
Will tried this in the past but because of the task placement after
fork()+execve(), I think we always ended up with a weight of 2 in the
child process. Your first patch "solves" this by flushing the TLBs on
context switch (bar the CnP case). Can you give it a try to see if it
improves things? At least it's a starting point for further
investigation.
I fully agree with Will that we don't want two different TLB handling
implementations in the arm64 kernel and even less desirable to have a
command line option.
Thanks.
--
Catalin
