From: Marc Zyngier <maz@kernel.org>
To: Reiji Watanabe <reijiw@google.com>
Cc: kvmarm@lists.linux.dev, kvm@vger.kernel.org,
linux-arm-kernel@lists.infradead.org,
James Morse <james.morse@arm.com>,
Suzuki K Poulose <suzuki.poulose@arm.com>,
Oliver Upton <oliver.upton@linux.dev>,
Zenghui Yu <yuzenghui@huawei.com>,
Ricardo Koller <ricarkol@google.com>,
Simon Veith <sveith@amazon.de>,
dwmw2@infradead.org
Subject: Re: [PATCH 05/16] KVM: arm64: timers: Convert per-vcpu virtual offset to a global value
Date: Wed, 22 Feb 2023 10:54:46 +0000 [thread overview]
Message-ID: <86cz62x955.wl-maz@kernel.org> (raw)
In-Reply-To: <CAAeT=FymvmeQepDBp_r1XVg4-jGT-KejPDbrbf2yV=FtPFYWbg@mail.gmail.com>
On Wed, 22 Feb 2023 06:15:56 +0000,
Reiji Watanabe <reijiw@google.com> wrote:
>
> Hi Marc,
>
> On Thu, Feb 16, 2023 at 6:21 AM Marc Zyngier <maz@kernel.org> wrote:
> >
> > Having a per-vcpu virtual offset is a pain. It needs to be synchronized
> > on each update, and expands badly to a setup where different timers can
> > have different offsets, or have composite offsets (as with NV).
> >
> > So let's start by replacing the use of the CNTVOFF_EL2 shadow register
> > (which we want to reclaim for NV anyway), and make the virtual timer
> > carry a pointer to a VM-wide offset.
>
> That's nice!
>
> >
> > This simplifies the code significantly.
> >
> > Signed-off-by: Marc Zyngier <maz@kernel.org>
> > ---
> > arch/arm64/include/asm/kvm_host.h | 3 +++
> > arch/arm64/kvm/arch_timer.c | 45 +++++++------------------------
> > arch/arm64/kvm/hypercalls.c | 2 +-
> > include/kvm/arm_arch_timer.h | 15 +++++++++++
> > 4 files changed, 29 insertions(+), 36 deletions(-)
> >
> > diff --git a/arch/arm64/include/asm/kvm_host.h b/arch/arm64/include/asm/kvm_host.h
> > index 2a8175f38439..3adac0c5e175 100644
> > --- a/arch/arm64/include/asm/kvm_host.h
> > +++ b/arch/arm64/include/asm/kvm_host.h
> > @@ -193,6 +193,9 @@ struct kvm_arch {
> > /* Interrupt controller */
> > struct vgic_dist vgic;
> >
> > + /* Timers */
> > + struct arch_timer_vm_offsets offsets;
>
> Nit: Perhaps using a more explicit name for the 'offsets' field might
> be better than simply 'offsets', since it is a field of kvm_arch (not
> a field of arch timer related struct).
> (e.g. timer_offsets ?)
Good point. Actually, maybe this should be a more generic structure
(arch_timer_global, or something similar), containing the offset(s).
I'll have a think.
>
> > +
> > /* Mandated version of PSCI */
> > u32 psci_version;
> >
> > diff --git a/arch/arm64/kvm/arch_timer.c b/arch/arm64/kvm/arch_timer.c
> > index 329a8444724f..1bb44f668608 100644
> > --- a/arch/arm64/kvm/arch_timer.c
> > +++ b/arch/arm64/kvm/arch_timer.c
> > @@ -84,14 +84,10 @@ u64 timer_get_cval(struct arch_timer_context *ctxt)
> >
> > static u64 timer_get_offset(struct arch_timer_context *ctxt)
> > {
> > - struct kvm_vcpu *vcpu = ctxt->vcpu;
> > + if (ctxt->offset.vm_offset)
> > + return *ctxt->offset.vm_offset;
> >
> > - switch(arch_timer_ctx_index(ctxt)) {
> > - case TIMER_VTIMER:
> > - return __vcpu_sys_reg(vcpu, CNTVOFF_EL2);
> > - default:
> > - return 0;
> > - }
> > + return 0;
> > }
> >
> > static void timer_set_ctl(struct arch_timer_context *ctxt, u32 ctl)
> > @@ -128,15 +124,12 @@ static void timer_set_cval(struct arch_timer_context *ctxt, u64 cval)
> >
> > static void timer_set_offset(struct arch_timer_context *ctxt, u64 offset)
> > {
> > - struct kvm_vcpu *vcpu = ctxt->vcpu;
> > -
> > - switch(arch_timer_ctx_index(ctxt)) {
> > - case TIMER_VTIMER:
> > - __vcpu_sys_reg(vcpu, CNTVOFF_EL2) = offset;
> > - break;
> > - default:
> > + if (!ctxt->offset.vm_offset) {
> > WARN(offset, "timer %ld\n", arch_timer_ctx_index(ctxt));
> > + return;
> > }
> > +
> > + WRITE_ONCE(*ctxt->offset.vm_offset, offset);
> > }
> >
> > u64 kvm_phys_timer_read(void)
> > @@ -765,25 +758,6 @@ int kvm_timer_vcpu_reset(struct kvm_vcpu *vcpu)
> > return 0;
> > }
> >
> > -/* Make the updates of cntvoff for all vtimer contexts atomic */
> > -static void update_vtimer_cntvoff(struct kvm_vcpu *vcpu, u64 cntvoff)
> > -{
> > - unsigned long i;
> > - struct kvm *kvm = vcpu->kvm;
> > - struct kvm_vcpu *tmp;
> > -
> > - mutex_lock(&kvm->lock);
> > - kvm_for_each_vcpu(i, tmp, kvm)
> > - timer_set_offset(vcpu_vtimer(tmp), cntvoff);
> > -
> > - /*
> > - * When called from the vcpu create path, the CPU being created is not
> > - * included in the loop above, so we just set it here as well.
> > - */
> > - timer_set_offset(vcpu_vtimer(vcpu), cntvoff);
> > - mutex_unlock(&kvm->lock);
> > -}
> > -
> > void kvm_timer_vcpu_init(struct kvm_vcpu *vcpu)
> > {
> > struct arch_timer_cpu *timer = vcpu_timer(vcpu);
> > @@ -791,10 +765,11 @@ void kvm_timer_vcpu_init(struct kvm_vcpu *vcpu)
> > struct arch_timer_context *ptimer = vcpu_ptimer(vcpu);
> >
> > vtimer->vcpu = vcpu;
> > + vtimer->offset.vm_offset = &vcpu->kvm->arch.offsets.voffset;
> > ptimer->vcpu = vcpu;
> >
> > /* Synchronize cntvoff across all vtimers of a VM. */
> > - update_vtimer_cntvoff(vcpu, kvm_phys_timer_read());
> > + timer_set_offset(vtimer, kvm_phys_timer_read());
> > timer_set_offset(ptimer, 0);
> >
> > hrtimer_init(&timer->bg_timer, CLOCK_MONOTONIC, HRTIMER_MODE_ABS_HARD);
> > @@ -840,7 +815,7 @@ int kvm_arm_timer_set_reg(struct kvm_vcpu *vcpu, u64 regid, u64 value)
> > break;
> > case KVM_REG_ARM_TIMER_CNT:
> > timer = vcpu_vtimer(vcpu);
> > - update_vtimer_cntvoff(vcpu, kvm_phys_timer_read() - value);
> > + timer_set_offset(timer, kvm_phys_timer_read() - value);
> > break;
> > case KVM_REG_ARM_TIMER_CVAL:
> > timer = vcpu_vtimer(vcpu);
> > diff --git a/arch/arm64/kvm/hypercalls.c b/arch/arm64/kvm/hypercalls.c
> > index 64c086c02c60..169d629f97cb 100644
> > --- a/arch/arm64/kvm/hypercalls.c
> > +++ b/arch/arm64/kvm/hypercalls.c
> > @@ -44,7 +44,7 @@ static void kvm_ptp_get_time(struct kvm_vcpu *vcpu, u64 *val)
> > feature = smccc_get_arg1(vcpu);
> > switch (feature) {
> > case KVM_PTP_VIRT_COUNTER:
> > - cycles = systime_snapshot.cycles - vcpu_read_sys_reg(vcpu, CNTVOFF_EL2);
> > + cycles = systime_snapshot.cycles - vcpu->kvm->arch.offsets.voffset;
> > break;
> > case KVM_PTP_PHYS_COUNTER:
> > cycles = systime_snapshot.cycles;
> > diff --git a/include/kvm/arm_arch_timer.h b/include/kvm/arm_arch_timer.h
> > index 76174f4fb646..41fda6255174 100644
> > --- a/include/kvm/arm_arch_timer.h
> > +++ b/include/kvm/arm_arch_timer.h
> > @@ -23,6 +23,19 @@ enum kvm_arch_timer_regs {
> > TIMER_REG_CTL,
> > };
> >
> > +struct arch_timer_offset {
> > + /*
> > + * If set, pointer to one of the offsets in the kvm's offset
> > + * structure. If NULL, assume a zero offset.
> > + */
> > + u64 *vm_offset;
> > +};
> > +
> > +struct arch_timer_vm_offsets {
> > + /* Offset applied to the virtual timer/counter */
> > + u64 voffset;
> > +};
> > +
> > struct arch_timer_context {
> > struct kvm_vcpu *vcpu;
> >
> > @@ -33,6 +46,8 @@ struct arch_timer_context {
> > struct hrtimer hrtimer;
> > u64 ns_frac;
> >
> > + /* Offset for this counter/timer */
> > + struct arch_timer_offset offset;
> > /*
> > * We have multiple paths which can save/restore the timer state onto
> > * the hardware, so we need some way of keeping track of where the
> > --
> > 2.34.1
> >
> >
>
> It appears that we can remove CNTVOFF_EL2 from the sysreg file.
> Having said that, I don't think it matters anyway since the
> following patch appears to use that again.
Indeed, and this is why I didn't remove it the first place (we need it
for NV).
>
> Reviewed-by: Reiji Watanabe <reijiw@google.com>
Thanks!
M.
--
Without deviation from the norm, progress is not possible.
next prev parent reply other threads:[~2023-02-22 10:54 UTC|newest]
Thread overview: 55+ messages / expand[flat|nested] mbox.gz Atom feed top
2023-02-16 14:21 [PATCH 00/16] KVM: arm64: Rework timer offsetting for fun and profit Marc Zyngier
2023-02-16 14:21 ` [PATCH 01/16] arm64: Add CNTPOFF_EL2 register definition Marc Zyngier
2023-02-16 14:21 ` [PATCH 02/16] arm64: Add HAS_ECV_CNTPOFF capability Marc Zyngier
2023-02-22 4:30 ` Reiji Watanabe
2023-02-22 10:47 ` Marc Zyngier
2023-02-16 14:21 ` [PATCH 03/16] kvm: arm64: Expose {un,}lock_all_vcpus() to the reset of KVM Marc Zyngier
2023-02-23 22:30 ` Colton Lewis
2023-02-16 14:21 ` [PATCH 04/16] KVM: arm64: timers: Use a per-vcpu, per-timer accumulator for fractional ns Marc Zyngier
2023-02-23 22:30 ` Colton Lewis
2023-02-16 14:21 ` [PATCH 05/16] KVM: arm64: timers: Convert per-vcpu virtual offset to a global value Marc Zyngier
2023-02-22 6:15 ` Reiji Watanabe
2023-02-22 10:54 ` Marc Zyngier [this message]
2023-02-16 14:21 ` [PATCH 06/16] KVM: arm64: timers: Use CNTPOFF_EL2 to offset the physical timer Marc Zyngier
2023-02-23 22:34 ` Colton Lewis
2023-02-24 8:59 ` Marc Zyngier
2023-02-16 14:21 ` [PATCH 07/16] KVM: arm64: timers: Allow physical offset without CNTPOFF_EL2 Marc Zyngier
2023-02-23 22:40 ` Colton Lewis
2023-02-24 10:54 ` Marc Zyngier
2023-02-16 14:21 ` [PATCH 08/16] KVM: arm64: timers: Allow userspace to set the counter offsets Marc Zyngier
2023-02-16 22:09 ` Oliver Upton
2023-02-17 10:17 ` Marc Zyngier
2023-02-17 22:11 ` Oliver Upton
2023-02-22 11:56 ` Marc Zyngier
2023-02-22 16:34 ` Oliver Upton
2023-02-23 18:25 ` Marc Zyngier
2023-03-08 7:46 ` Oliver Upton
2023-03-08 7:53 ` Oliver Upton
2023-03-09 8:29 ` Marc Zyngier
2023-03-09 8:25 ` Marc Zyngier
2023-02-23 22:41 ` Colton Lewis
2023-02-24 11:24 ` Marc Zyngier
2023-02-16 14:21 ` [PATCH 09/16] KVM: arm64: timers: Allow save/restoring of the physical timer Marc Zyngier
2023-02-16 14:21 ` [PATCH 10/16] KVM: arm64: timers: Rationalise per-vcpu timer init Marc Zyngier
2023-02-16 14:21 ` [PATCH 11/16] KVM: arm64: Document KVM_ARM_SET_CNT_OFFSETS and co Marc Zyngier
2023-02-16 14:21 ` [PATCH 12/16] KVM: arm64: nv: timers: Add a per-timer, per-vcpu offset Marc Zyngier
2023-02-24 20:07 ` Colton Lewis
2023-02-25 10:32 ` Marc Zyngier
2023-02-16 14:21 ` [PATCH 13/16] KVM: arm64: nv: timers: Support hyp timer emulation Marc Zyngier
2023-02-24 20:08 ` Colton Lewis
2023-02-25 10:34 ` Marc Zyngier
2023-02-16 14:21 ` [PATCH 14/16] KVM: arm64: selftests: Add physical timer registers to the sysreg list Marc Zyngier
2023-02-16 14:21 ` [PATCH 15/16] KVM: arm64: selftests: Augment existing timer test to handle variable offsets Marc Zyngier
2023-03-06 22:08 ` Colton Lewis
2023-03-09 9:01 ` Marc Zyngier
2023-03-10 19:26 ` Colton Lewis
2023-03-12 15:53 ` Marc Zyngier
2023-03-13 11:43 ` Marc Zyngier
2023-03-14 17:47 ` Colton Lewis
2023-03-14 18:18 ` Marc Zyngier
2023-02-16 14:21 ` [PATCH 16/16] KVM: arm64: selftests: Deal with spurious timer interrupts Marc Zyngier
2023-02-21 16:28 ` [PATCH 00/16] KVM: arm64: Rework timer offsetting for fun and profit Veith, Simon
2023-02-21 22:17 ` Marc Zyngier
2023-02-23 22:29 ` Colton Lewis
2023-02-24 8:45 ` Marc Zyngier
2023-02-24 20:07 ` Colton Lewis
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=86cz62x955.wl-maz@kernel.org \
--to=maz@kernel.org \
--cc=dwmw2@infradead.org \
--cc=james.morse@arm.com \
--cc=kvm@vger.kernel.org \
--cc=kvmarm@lists.linux.dev \
--cc=linux-arm-kernel@lists.infradead.org \
--cc=oliver.upton@linux.dev \
--cc=reijiw@google.com \
--cc=ricarkol@google.com \
--cc=suzuki.poulose@arm.com \
--cc=sveith@amazon.de \
--cc=yuzenghui@huawei.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).