Re: [PATCH v2 1/2] RISCV: KVM: Introduce mp_state_lock to avoid lock inversion in SBI_EXT_HSM_HART_START

From: Anup Patel
Date: Sun Apr 21 2024 - 23:57:28 EST

Next message: Anup Patel: "Re: [PATCH v2 2/2] RISCV: KVM: Introduce vcpu->reset_cntx_lock"
Previous message: Gavin Shan: "Re: [PATCH v5 01/18] cpu: Do not warn on arch_register_cpu() returning -EPROBE_DEFER"
In reply to: Yong-Xuan Wang: "[PATCH v2 1/2] RISCV: KVM: Introduce mp_state_lock to avoid lock inversion in SBI_EXT_HSM_HART_START"
Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]

On Wed, Apr 17, 2024 at 1:15 PM Yong-Xuan Wang <yongxuan.wang@xxxxxxxxxx> wrote:
>
> Documentation/virt/kvm/locking.rst advises that kvm->lock should be
> acquired outside vcpu->mutex and kvm->srcu. However, when KVM/RISC-V
> handling SBI_EXT_HSM_HART_START, the lock ordering is vcpu->mutex,
> kvm->srcu then kvm->lock.
>
> Although the lockdep checking no longer complains about this after commit
> f0f44752f5f6 ("rcu: Annotate SRCU's update-side lockdep dependencies"),
> it's necessary to replace kvm->lock with a new dedicated lock to ensure
> only one hart can execute the SBI_EXT_HSM_HART_START call for the target
> hart simultaneously.
>
> Additionally, this patch also rename "power_off" to "mp_state" with two
> possible values. The vcpu->mp_state_lock also protects the access of
> vcpu->mp_state.
>
> Signed-off-by: Yong-Xuan Wang <yongxuan.wang@xxxxxxxxxx>
> ---
> arch/riscv/include/asm/kvm_host.h | 7 ++--
> arch/riscv/kvm/vcpu.c | 56 ++++++++++++++++++++++++-------
> arch/riscv/kvm/vcpu_sbi.c | 7 ++--
> arch/riscv/kvm/vcpu_sbi_hsm.c | 23 ++++++++-----
> 4 files changed, 68 insertions(+), 25 deletions(-)
>
> diff --git a/arch/riscv/include/asm/kvm_host.h b/arch/riscv/include/asm/kvm_host.h
> index 484d04a92fa6..64d35a8c908c 100644
> --- a/arch/riscv/include/asm/kvm_host.h
> +++ b/arch/riscv/include/asm/kvm_host.h
> @@ -252,8 +252,9 @@ struct kvm_vcpu_arch {
> /* Cache pages needed to program page tables with spinlock held */
> struct kvm_mmu_memory_cache mmu_page_cache;
>
> - /* VCPU power-off state */
> - bool power_off;
> + /* VCPU power state */
> + struct kvm_mp_state mp_state;
> + spinlock_t mp_state_lock;
>
> /* Don't run the VCPU (blocked) */
> bool pause;
> @@ -375,7 +376,9 @@ void kvm_riscv_vcpu_flush_interrupts(struct kvm_vcpu *vcpu);
> void kvm_riscv_vcpu_sync_interrupts(struct kvm_vcpu *vcpu);
> bool kvm_riscv_vcpu_has_interrupts(struct kvm_vcpu *vcpu, u64 mask);
> void kvm_riscv_vcpu_power_off(struct kvm_vcpu *vcpu);
> +void __kvm_riscv_vcpu_power_on(struct kvm_vcpu *vcpu);
> void kvm_riscv_vcpu_power_on(struct kvm_vcpu *vcpu);
> +bool kvm_riscv_vcpu_stopped(struct kvm_vcpu *vcpu);
>
> void kvm_riscv_vcpu_sbi_sta_reset(struct kvm_vcpu *vcpu);
> void kvm_riscv_vcpu_record_steal_time(struct kvm_vcpu *vcpu);
> diff --git a/arch/riscv/kvm/vcpu.c b/arch/riscv/kvm/vcpu.c
> index b5ca9f2e98ac..70937f71c3c4 100644
> --- a/arch/riscv/kvm/vcpu.c
> +++ b/arch/riscv/kvm/vcpu.c
> @@ -102,6 +102,8 @@ int kvm_arch_vcpu_create(struct kvm_vcpu *vcpu)
> struct kvm_cpu_context *cntx;
> struct kvm_vcpu_csr *reset_csr = &vcpu->arch.guest_reset_csr;
>
> + spin_lock_init(&vcpu->arch.mp_state_lock);
> +
> /* Mark this VCPU never ran */
> vcpu->arch.ran_atleast_once = false;
> vcpu->arch.mmu_page_cache.gfp_zero = __GFP_ZERO;
> @@ -201,7 +203,7 @@ void kvm_arch_vcpu_unblocking(struct kvm_vcpu *vcpu)
> int kvm_arch_vcpu_runnable(struct kvm_vcpu *vcpu)
> {
> return (kvm_riscv_vcpu_has_interrupts(vcpu, -1UL) &&
> - !vcpu->arch.power_off && !vcpu->arch.pause);
> + !kvm_riscv_vcpu_stopped(vcpu) && !vcpu->arch.pause);
> }
>
> int kvm_arch_vcpu_should_kick(struct kvm_vcpu *vcpu)
> @@ -429,26 +431,50 @@ bool kvm_riscv_vcpu_has_interrupts(struct kvm_vcpu *vcpu, u64 mask)
> return kvm_riscv_vcpu_aia_has_interrupts(vcpu, mask);
> }
>
> -void kvm_riscv_vcpu_power_off(struct kvm_vcpu *vcpu)
> +static void __kvm_riscv_vcpu_power_off(struct kvm_vcpu *vcpu)
> {
> - vcpu->arch.power_off = true;
> + vcpu->arch.mp_state.mp_state = KVM_MP_STATE_STOPPED;
> kvm_make_request(KVM_REQ_SLEEP, vcpu);
> kvm_vcpu_kick(vcpu);
> }
>
> -void kvm_riscv_vcpu_power_on(struct kvm_vcpu *vcpu)
> +void kvm_riscv_vcpu_power_off(struct kvm_vcpu *vcpu)
> +{
> + spin_lock(&vcpu->arch.mp_state_lock);
> + __kvm_riscv_vcpu_power_off(vcpu);
> + spin_unlock(&vcpu->arch.mp_state_lock);
> +}
> +
> +void __kvm_riscv_vcpu_power_on(struct kvm_vcpu *vcpu)
> {
> - vcpu->arch.power_off = false;
> + vcpu->arch.mp_state.mp_state = KVM_MP_STATE_RUNNABLE;
> kvm_vcpu_wake_up(vcpu);
> }
>
> +void kvm_riscv_vcpu_power_on(struct kvm_vcpu *vcpu)
> +{
> + spin_lock(&vcpu->arch.mp_state_lock);
> + __kvm_riscv_vcpu_power_on(vcpu);
> + spin_unlock(&vcpu->arch.mp_state_lock);
> +}
> +
> +bool kvm_riscv_vcpu_stopped(struct kvm_vcpu *vcpu)
> +{
> + bool ret;
> +
> + spin_lock(&vcpu->arch.mp_state_lock);
> + ret = vcpu->arch.mp_state.mp_state == KVM_MP_STATE_STOPPED;
> + spin_unlock(&vcpu->arch.mp_state_lock);
> +
> + return ret;

Checking mp_state is very expensive this way because spin locks
are implicit fences.

Instead of this, we can simply use READ_ONCE() here and we
use WRITE_ONCE() + spin lock to serialize writes.

I will take care of this at the time of merging this patch.

> +}
> +
> int kvm_arch_vcpu_ioctl_get_mpstate(struct kvm_vcpu *vcpu,
> struct kvm_mp_state *mp_state)
> {
> - if (vcpu->arch.power_off)
> - mp_state->mp_state = KVM_MP_STATE_STOPPED;
> - else
> - mp_state->mp_state = KVM_MP_STATE_RUNNABLE;
> + spin_lock(&vcpu->arch.mp_state_lock);
> + *mp_state = vcpu->arch.mp_state;
> + spin_unlock(&vcpu->arch.mp_state_lock);
>
> return 0;
> }
> @@ -458,17 +484,21 @@ int kvm_arch_vcpu_ioctl_set_mpstate(struct kvm_vcpu *vcpu,
> {
> int ret = 0;
>
> + spin_lock(&vcpu->arch.mp_state_lock);
> +
> switch (mp_state->mp_state) {
> case KVM_MP_STATE_RUNNABLE:
> - vcpu->arch.power_off = false;
> + vcpu->arch.mp_state.mp_state = KVM_MP_STATE_RUNNABLE;
> break;
> case KVM_MP_STATE_STOPPED:
> - kvm_riscv_vcpu_power_off(vcpu);
> + __kvm_riscv_vcpu_power_off(vcpu);
> break;
> default:
> ret = -EINVAL;
> }
>
> + spin_unlock(&vcpu->arch.mp_state_lock);
> +
> return ret;
> }
>
> @@ -584,11 +614,11 @@ static void kvm_riscv_check_vcpu_requests(struct kvm_vcpu *vcpu)
> if (kvm_check_request(KVM_REQ_SLEEP, vcpu)) {
> kvm_vcpu_srcu_read_unlock(vcpu);
> rcuwait_wait_event(wait,
> - (!vcpu->arch.power_off) && (!vcpu->arch.pause),
> + (!kvm_riscv_vcpu_stopped(vcpu)) && (!vcpu->arch.pause),
> TASK_INTERRUPTIBLE);
> kvm_vcpu_srcu_read_lock(vcpu);
>
> - if (vcpu->arch.power_off || vcpu->arch.pause) {
> + if (kvm_riscv_vcpu_stopped(vcpu) || vcpu->arch.pause) {
> /*
> * Awaken to handle a signal, request to
> * sleep again later.
> diff --git a/arch/riscv/kvm/vcpu_sbi.c b/arch/riscv/kvm/vcpu_sbi.c
> index 72a2ffb8dcd1..1851fc979bd2 100644
> --- a/arch/riscv/kvm/vcpu_sbi.c
> +++ b/arch/riscv/kvm/vcpu_sbi.c
> @@ -138,8 +138,11 @@ void kvm_riscv_vcpu_sbi_system_reset(struct kvm_vcpu *vcpu,
> unsigned long i;
> struct kvm_vcpu *tmp;
>
> - kvm_for_each_vcpu(i, tmp, vcpu->kvm)
> - tmp->arch.power_off = true;
> + kvm_for_each_vcpu(i, tmp, vcpu->kvm) {
> + spin_lock(&vcpu->arch.mp_state_lock);
> + tmp->arch.mp_state.mp_state = KVM_MP_STATE_STOPPED;
> + spin_unlock(&vcpu->arch.mp_state_lock);
> + }
> kvm_make_all_cpus_request(vcpu->kvm, KVM_REQ_SLEEP);
>
> memset(&run->system_event, 0, sizeof(run->system_event));
> diff --git a/arch/riscv/kvm/vcpu_sbi_hsm.c b/arch/riscv/kvm/vcpu_sbi_hsm.c
> index 7dca0e9381d9..115a6c6525fd 100644
> --- a/arch/riscv/kvm/vcpu_sbi_hsm.c
> +++ b/arch/riscv/kvm/vcpu_sbi_hsm.c
> @@ -18,12 +18,18 @@ static int kvm_sbi_hsm_vcpu_start(struct kvm_vcpu *vcpu)
> struct kvm_cpu_context *cp = &vcpu->arch.guest_context;
> struct kvm_vcpu *target_vcpu;
> unsigned long target_vcpuid = cp->a0;
> + int ret = 0;
>
> target_vcpu = kvm_get_vcpu_by_id(vcpu->kvm, target_vcpuid);
> if (!target_vcpu)
> return SBI_ERR_INVALID_PARAM;
> - if (!target_vcpu->arch.power_off)
> - return SBI_ERR_ALREADY_AVAILABLE;
> +
> + spin_lock(&target_vcpu->arch.mp_state_lock);
> +
> + if (target_vcpu->arch.mp_state.mp_state != KVM_MP_STATE_STOPPED) {
> + ret = SBI_ERR_ALREADY_AVAILABLE;
> + goto out;
> + }
>
> reset_cntx = &target_vcpu->arch.guest_reset_context;
> /* start address */
> @@ -34,14 +40,18 @@ static int kvm_sbi_hsm_vcpu_start(struct kvm_vcpu *vcpu)
> reset_cntx->a1 = cp->a2;
> kvm_make_request(KVM_REQ_VCPU_RESET, target_vcpu);
>
> - kvm_riscv_vcpu_power_on(target_vcpu);
> + __kvm_riscv_vcpu_power_on(target_vcpu);
> +
> +out:
> + spin_unlock(&target_vcpu->arch.mp_state_lock);
> +
>
> return 0;
> }
>
> static int kvm_sbi_hsm_vcpu_stop(struct kvm_vcpu *vcpu)
> {
> - if (vcpu->arch.power_off)
> + if (kvm_riscv_vcpu_stopped(vcpu))
> return SBI_ERR_FAILURE;
>
> kvm_riscv_vcpu_power_off(vcpu);
> @@ -58,7 +68,7 @@ static int kvm_sbi_hsm_vcpu_get_status(struct kvm_vcpu *vcpu)
> target_vcpu = kvm_get_vcpu_by_id(vcpu->kvm, target_vcpuid);
> if (!target_vcpu)
> return SBI_ERR_INVALID_PARAM;
> - if (!target_vcpu->arch.power_off)
> + if (!kvm_riscv_vcpu_stopped(target_vcpu))
> return SBI_HSM_STATE_STARTED;
> else if (vcpu->stat.generic.blocking)
> return SBI_HSM_STATE_SUSPENDED;
> @@ -71,14 +81,11 @@ static int kvm_sbi_ext_hsm_handler(struct kvm_vcpu *vcpu, struct kvm_run *run,
> {
> int ret = 0;
> struct kvm_cpu_context *cp = &vcpu->arch.guest_context;
> - struct kvm *kvm = vcpu->kvm;
> unsigned long funcid = cp->a6;
>
> switch (funcid) {
> case SBI_EXT_HSM_HART_START:
> - mutex_lock(&kvm->lock);
> ret = kvm_sbi_hsm_vcpu_start(vcpu);
> - mutex_unlock(&kvm->lock);
> break;
> case SBI_EXT_HSM_HART_STOP:
> ret = kvm_sbi_hsm_vcpu_stop(vcpu);
> --
> 2.17.1
>

Reviewed-by: Anup Patel <anup@xxxxxxxxxxxxxx>

Regards,
Anup

Next message: Anup Patel: "Re: [PATCH v2 2/2] RISCV: KVM: Introduce vcpu->reset_cntx_lock"
Previous message: Gavin Shan: "Re: [PATCH v5 01/18] cpu: Do not warn on arch_register_cpu() returning -EPROBE_DEFER"
In reply to: Yong-Xuan Wang: "[PATCH v2 1/2] RISCV: KVM: Introduce mp_state_lock to avoid lock inversion in SBI_EXT_HSM_HART_START"
Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]