RISC-V: Limit the scope of TLB shootdowns
authorAndrew Waterman <andrew@sifive.com>
Thu, 26 Oct 2017 08:53:40 +0000 (01:53 -0700)
committerPalmer Dabbelt <palmer@dabbelt.com>
Wed, 31 Jan 2018 03:13:33 +0000 (19:13 -0800)
commitf1b65f20fb05d1dd94656904848b96cc6df52bc0
tree274e7cc83e11045eef2e5883e1e664629f910dce
parentfe9b842f72921fb18b93cf47a255f374289ef242
RISC-V: Limit the scope of TLB shootdowns

RISC-V systems perform TLB shootdows via the SBI, which currently
performs an IPI to each of the remote harts which then performs a local
TLB flush.  This process is a bit on the slow side, but we can at least
speed it up for some common cases by restricting the set of harts to
shoot down to the actual set of harts that are currently participating
in the given mm context, as opposed to the entire system.

This should provide a measurable performance increase, but we haven't
measured it.  Regardless, it seems like obviously the right thing to do
here.

Signed-off-by: Andrew Waterman <andrew@sifive.com>
Signed-off-by: Palmer Dabbelt <palmer@dabbelt.com>
arch/riscv/include/asm/tlbflush.h