Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.37 as permitted sender) client-ip=23.128.96.37;
MIME-Version: 1.0
References: <20231116023036.2324371-1-maobibo@loongson.cn>
In-Reply-To: <20231116023036.2324371-1-maobibo@loongson.cn>
From:   Huacai Chen <chenhuacai@kernel.org>
Date:   Thu, 16 Nov 2023 10:54:52 +0800
Message-ID: <CAAhV-H4Wyp-6_gFSfm8uWUMiEJnebk9n4JxQrx_nBdxkTF5wUA@mail.gmail.com>
Subject: Re: [PATCH v4 0/3] LoongArch: KVM: Remove SW timer switch when vcpu
 is halt polling
To:     Bibo Mao <maobibo@loongson.cn>
Cc:     Tianrui Zhao <zhaotianrui@loongson.cn>,
        WANG Xuerui <kernel@xen0n.name>, kvm@vger.kernel.org,
        loongarch@lists.linux.dev, linux-kernel@vger.kernel.org
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable
Precedence: bulk

Hi, Bibo,

I suggest submitting this series to the internal repo, too. Because we
don't have enough resources to test the stability for the upstream
version, while this is a fundamental change. On the other hand, the
patch "LoongArch:LSVZ: set timer offset at first time once" can be
submitted first because it is already in the internal repo.

Huacai

On Thu, Nov 16, 2023 at 10:33=E2=80=AFAM Bibo Mao <maobibo@loongson.cn> wro=
te:
>
> This patches removes SW timer switch during vcpu block stage. VM uses HW
> timer rather than SW PV timer on LoongArch system, it can check pending
> HW timer interrupt status directly, rather than switch to SW timer and
> check injected SW timer interrupt.
>
> When SW timer is not used in vcpu halt-polling mode, the relative
> SW timer handling before entering guest can be removed also. Timer
> emulation is simpler than before, SW timer emuation is only used in vcpu
> thread context switch.
> ---
> Changes in v4:
>   If vcpu is scheduled out since there is no pending event, and timer is
> fired during sched-out period. SW hrtimer is used to wake up vcpu thread
> in time, rather than inject pending timer irq.
>
> Changes in v3:
>   Add kvm_arch_vcpu_runnable checking before kvm_vcpu_halt.
>
> Changes in v2:
>   Add halt polling support for idle instruction emulation, using api
> kvm_vcpu_halt rather than kvm_vcpu_block in function kvm_emu_idle.
>
> ---
> Bibo Mao (3):
>   LoongArch: KVM: Remove SW timer switch when vcpu is halt polling
>   LoongArch: KVM: Allow to access HW timer CSR registers always
>   LoongArch: KVM: Remove kvm_acquire_timer before entering guest
>
>  arch/loongarch/include/asm/kvm_vcpu.h |  1 -
>  arch/loongarch/kvm/exit.c             | 13 +-----
>  arch/loongarch/kvm/main.c             |  1 -
>  arch/loongarch/kvm/timer.c            | 62 ++++++++++-----------------
>  arch/loongarch/kvm/vcpu.c             | 38 ++++------------
>  5 files changed, 32 insertions(+), 83 deletions(-)
>
>
> base-commit: c42d9eeef8e5ba9292eda36fd8e3c11f35ee065c
> --
> 2.39.3
>