kvm: Use host timekeeping in guest.

This is used to synchronize time between host and guest. The guest can request the (guest) physical address it wants the data in through the MSR_KVM_TIMEKEEPER_EN MSR.

It currently assumes the host timekeeper is "tsc".

Signed-off-by Suleiman Souhlal <[email protected]>

---
 arch/x86/include/asm/kvm_host.h      |   3 +
 arch/x86/include/asm/pvclock-abi.h   |  27 ++++++
 arch/x86/include/uapi/asm/kvm_para.h |   1 +
 arch/x86/kvm/x86.c                   | 121 +++++++++++++++++++++++++++
 4 files changed, 152 insertions(+)

arch/x86/include/asm/kvm_host.h +3 -0

index 50eb430b0ad8..4d622450cb4a 100644
@@ -659,7 +659,10 @@ struct kvm_vcpu_arch {
 	struct pvclock_vcpu_time_info hv_clock;
 	unsigned int hw_tsc_khz;
 	struct gfn_to_hva_cache pv_time;
+	struct gfn_to_hva_cache pv_timekeeper_g2h;
+	struct pvclock_timekeeper pv_timekeeper;
 	bool pv_time_enabled;
+	bool pv_timekeeper_enabled;
 	/* set guest stopped flag in pvclock flags field */
 	bool pvclock_set_guest_stopped_request;
 

arch/x86/include/asm/pvclock-abi.h +15 -0

index 1436226efe3e..2809008b9b26 100644
@@ -40,6 +40,33 @@ struct pvclock_wall_clock {
 	u32   nsec;
 } __attribute__((__packed__));
 
+struct pvclock_read_base {
+	u64 mask;
+	u64 cycle_last;
+	u32 mult;
+	u32 shift;
+	u64 xtime_nsec;
+	u64 base;
+} __attribute__((__packed__));
+
+struct pvclock_timekeeper {
+	u64 gen;
+	u64 flags;
+	struct pvclock_read_base tkr_mono;
+	struct pvclock_read_base tkr_raw;
+	u64 xtime_sec;

arch/x86/include/asm/pvclock-abi.h +2 -0

+#define	PVCLOCK_TIMEKEEPER_ENABLED (1 << 0)
+
 #define PVCLOCK_TSC_STABLE_BIT	(1 << 0)
 #define PVCLOCK_GUEST_STOPPED	(1 << 1)
 /* PVCLOCK_COUNTS_FROM_ZERO broke ABI and can't be used anymore. */

arch/x86/include/uapi/asm/kvm_para.h +1 -0

index 2a8e0b6b9805..3ebb1d87db3a 100644
@@ -50,6 +50,7 @@
 #define MSR_KVM_STEAL_TIME  0x4b564d03
 #define MSR_KVM_PV_EOI_EN      0x4b564d04
 #define MSR_KVM_POLL_CONTROL	0x4b564d05
+#define MSR_KVM_TIMEKEEPER_EN	0x4b564d06
 
 struct kvm_steal_time {
 	__u64 steal;

arch/x86/kvm/x86.c +116 -0

index 661e2bf38526..937f83cdda4b 100644
@@ -157,6 +157,8 @@ module_param(force_emulation_prefix, bool, S_IRUGO);
 int __read_mostly pi_inject_timer = -1;
 module_param(pi_inject_timer, bint, S_IRUGO | S_IWUSR);
 
+static atomic_t pv_timekeepers_nr;
+
 #define KVM_NR_SHARED_MSRS 16
 
 struct kvm_shared_msrs_global {

@@ -2729,6 +2731,16 @@ int kvm_set_msr_common(struct kvm_vcpu *vcpu, struct msr_data *msr_info)
 
 		break;
 	}
+	case MSR_KVM_TIMEKEEPER_EN:
+		if (kvm_gfn_to_hva_cache_init(vcpu->kvm,
+		    &vcpu->arch.pv_timekeeper_g2h, data,
+		    sizeof(struct pvclock_timekeeper)))
+			vcpu->arch.pv_timekeeper_enabled = false;
+		else {
+			vcpu->arch.pv_timekeeper_enabled = true;
+			atomic_inc(&pv_timekeepers_nr);
+		}
+		break;
 	case MSR_KVM_ASYNC_PF_EN:
 		if (kvm_pv_enable_async_pf(vcpu, data))
 			return 1;

@@ -7097,6 +7109,109 @@ static struct perf_guest_info_callbacks kvm_guest_cbs = {
 };
 
 #ifdef CONFIG_X86_64
+static DEFINE_SPINLOCK(shadow_pvtk_lock);
+static struct pvclock_timekeeper shadow_pvtk;
+
+static void
+pvclock_copy_read_base(struct pvclock_read_base *pvtkr,
+    struct tk_read_base *tkr)
+{
+	pvtkr->cycle_last = tkr->cycle_last;
+	pvtkr->mult = tkr->mult;
+	pvtkr->shift = tkr->shift;
+	pvtkr->mask = tkr->mask;
+	pvtkr->xtime_nsec = tkr->xtime_nsec;
+	pvtkr->base = tkr->base;
+}
+
+static void
+kvm_copy_into_pvtk(struct kvm_vcpu *vcpu)
+{
+	struct pvclock_timekeeper *pvtk;
+	unsigned long flags;
+
+	if (!vcpu->arch.pv_timekeeper_enabled)
+		return;
+
+	pvtk = &vcpu->arch.pv_timekeeper;
+	if (pvclock_gtod_data.clock.vclock_mode == VCLOCK_TSC) {
+		pvtk->flags |= PVCLOCK_TIMEKEEPER_ENABLED;
+		spin_lock_irqsave(&shadow_pvtk_lock, flags);
+		pvtk->tkr_mono = shadow_pvtk.tkr_mono;
+		pvtk->tkr_raw = shadow_pvtk.tkr_raw;
+
+		pvtk->xtime_sec = shadow_pvtk.xtime_sec;
+		pvtk->ktime_sec = shadow_pvtk.ktime_sec;
+		pvtk->wall_to_monotonic_sec =
+		    shadow_pvtk.wall_to_monotonic_sec;
+		pvtk->wall_to_monotonic_nsec =
+		    shadow_pvtk.wall_to_monotonic_nsec;
+		pvtk->offs_real = shadow_pvtk.offs_real;
+		pvtk->offs_boot = shadow_pvtk.offs_boot;
+		pvtk->offs_tai = shadow_pvtk.offs_tai;
+		pvtk->raw_sec = shadow_pvtk.raw_sec;
+		spin_unlock_irqrestore(&shadow_pvtk_lock, flags);
+
+		pvtk->tsc_offset = kvm_x86_ops->read_l1_tsc_offset(vcpu);
+	} else
+		pvtk->flags &= ~PVCLOCK_TIMEKEEPER_ENABLED;
+
+	BUILD_BUG_ON(offsetof(struct pvclock_timekeeper, gen) != 0);
+
+	/*
+	 * Make the gen count odd to indicate we are in the process of
+	 * updating.
+	 */
+	vcpu->arch.pv_timekeeper.gen++;
+	vcpu->arch.pv_timekeeper.gen |= 1;
+
+	/*
+	 * See comment in kvm_guest_time_update() for why we have to do
+	 * multiple writes.
+	 */
+	kvm_write_guest_cached(vcpu->kvm, &vcpu->arch.pv_timekeeper_g2h,
+	    &vcpu->arch.pv_timekeeper, sizeof(vcpu->arch.pv_timekeeper.gen));
+
+	smp_wmb();
+
+	kvm_write_guest_cached(vcpu->kvm, &vcpu->arch.pv_timekeeper_g2h,
+	    &vcpu->arch.pv_timekeeper, sizeof(vcpu->arch.pv_timekeeper));
+
+	smp_wmb();
+
+	vcpu->arch.pv_timekeeper.gen++;
+
+	kvm_write_guest_cached(vcpu->kvm, &vcpu->arch.pv_timekeeper_g2h,
+	    &vcpu->arch.pv_timekeeper, sizeof(vcpu->arch.pv_timekeeper.gen));
+}
+
+static void
+update_shadow_pvtk(struct timekeeper *tk)
+{
+	struct pvclock_timekeeper *pvtk;
+	unsigned long flags;
+
+	pvtk = &shadow_pvtk;
+
+	if (atomic_read(&pv_timekeepers_nr) == 0 ||
+	    pvclock_gtod_data.clock.vclock_mode != VCLOCK_TSC)
+		return;
+
+	spin_lock_irqsave(&shadow_pvtk_lock, flags);
+	pvclock_copy_read_base(&pvtk->tkr_mono, &tk->tkr_mono);
+	pvclock_copy_read_base(&pvtk->tkr_raw, &tk->tkr_raw);
+
+	pvtk->xtime_sec = tk->xtime_sec;
+	pvtk->ktime_sec = tk->ktime_sec;
+	pvtk->wall_to_monotonic_sec = tk->wall_to_monotonic.tv_sec;
+	pvtk->wall_to_monotonic_nsec = tk->wall_to_monotonic.tv_nsec;
+	pvtk->offs_real = tk->offs_real;
+	pvtk->offs_boot = tk->offs_boot;
+	pvtk->offs_tai = tk->offs_tai;
+	pvtk->raw_sec = tk->raw_sec;
+	spin_unlock_irqrestore(&shadow_pvtk_lock, flags);
+}
+
 static void pvclock_gtod_update_fn(struct work_struct *work)
 {
 	struct kvm *kvm;

@@ -7124,6 +7239,7 @@ static int pvclock_gtod_notify(struct notifier_block *nb, unsigned long unused,
 	struct timekeeper *tk = priv;
 
 	update_pvclock_gtod(tk);
+	update_shadow_pvtk(tk);
 
 	/* disable master clock if host does not trust, or does not
 	 * use, TSC based clocksource.

arch/x86/kvm/x86.c +4 -0

+
 	if (kvm_request_pending(vcpu)) {
 		if (kvm_check_request(KVM_REQ_GET_VMCS12_PAGES, vcpu))
 			kvm_x86_ops->get_vmcs12_pages(vcpu);

@@ -9020,6 +9138,9 @@ void kvm_arch_vcpu_free(struct kvm_vcpu *vcpu)
 
 	kvmclock_reset(vcpu);
 
+	if (vcpu->arch.pv_timekeeper_enabled)
+		atomic_dec(&pv_timekeepers_nr);
+
 	kvm_x86_ops->vcpu_free(vcpu);
 	free_cpumask_var(wbinvd_dirty_mask);
 }

When kvm-hostclock is selected, and the host supports it, update our timekeeping parameters to be the same as the host.
This lets us have our time synchronized with the host's, even in the presence of host NTP or suspend.

Signed-off-by Suleiman Souhlal <[email protected]>

---
 arch/x86/Kconfig                    |   9 ++
 arch/x86/include/asm/kvmclock.h     |  12 +++
 arch/x86/kernel/Makefile            |   2 +
 arch/x86/kernel/kvmclock.c          |   5 +-
 arch/x86/kernel/kvmhostclock.c      | 130 ++++++++++++++++++++++++++++
 include/linux/timekeeper_internal.h |   8 ++
 kernel/time/timekeeping.c           |   2 +
 7 files changed, 167 insertions(+), 1 deletion(-)
 create mode 100644 arch/x86/kernel/kvmhostclock.c

arch/x86/Kconfig +9 -0

index d6e1faa28c58..c5b1257ea969 100644
@@ -839,6 +839,15 @@ config PARAVIRT_TIME_ACCOUNTING
 config PARAVIRT_CLOCK
 	bool
 
+config KVM_HOSTCLOCK
+	bool "kvmclock uses host timekeeping"
+	depends on KVM_GUEST
+	help
+	  Select this option to make the guest use the same timekeeping
+	  parameters as the host. This means that time will be almost
+	  exactly the same between the two. Only works if the host uses "tsc"
+	  clocksource.
+
 config JAILHOUSE_GUEST
 	bool "Jailhouse non-root cell support"
 	depends on X86_64 && PCI

arch/x86/include/asm/kvmclock.h +12 -0

index eceea9299097..de1a590ff97e 100644
@@ -2,6 +2,18 @@
 #ifndef _ASM_X86_KVM_CLOCK_H
 #define _ASM_X86_KVM_CLOCK_H
 
+#include <linux/timekeeper_internal.h>
+
 extern struct clocksource kvm_clock;
 
+unsigned long kvm_get_tsc_khz(void);
+
+#ifdef CONFIG_KVM_HOSTCLOCK
+void kvm_hostclock_init(void);
+#else
+static inline void kvm_hostclock_init(void)
+{
+}
+#endif /* KVM_HOSTCLOCK */
+
 #endif /* _ASM_X86_KVM_CLOCK_H */

arch/x86/kernel/Makefile +2 -0

index 3578ad248bc9..bc7be935fc5e 100644
@@ -17,6 +17,7 @@ CFLAGS_REMOVE_tsc.o = -pg
 CFLAGS_REMOVE_paravirt-spinlocks.o = -pg
 CFLAGS_REMOVE_pvclock.o = -pg
 CFLAGS_REMOVE_kvmclock.o = -pg
+CFLAGS_REMOVE_kvmhostclock.o = -pg
 CFLAGS_REMOVE_ftrace.o = -pg
 CFLAGS_REMOVE_early_printk.o = -pg
 CFLAGS_REMOVE_head64.o = -pg

@@ -112,6 +113,7 @@ obj-$(CONFIG_AMD_NB)		+= amd_nb.o
 obj-$(CONFIG_DEBUG_NMI_SELFTEST) += nmi_selftest.o
 
 obj-$(CONFIG_KVM_GUEST)		+= kvm.o kvmclock.o
+obj-$(CONFIG_KVM_HOSTCLOCK)	+= kvmhostclock.o
 obj-$(CONFIG_PARAVIRT)		+= paravirt.o paravirt_patch.o
 obj-$(CONFIG_PARAVIRT_SPINLOCKS)+= paravirt-spinlocks.o
 obj-$(CONFIG_PARAVIRT_CLOCK)	+= pvclock.o

arch/x86/kernel/kvmclock.c +4 -1

index 904494b924c1..4ab862de9777 100644
@@ -125,7 +125,7 @@ static inline void kvm_sched_clock_init(bool stable)
  * poll of guests can be running and trouble each other. So we preset
  * lpj here
  */
-static unsigned long kvm_get_tsc_khz(void)
+unsigned long kvm_get_tsc_khz(void)
 {
 	setup_force_cpu_cap(X86_FEATURE_TSC_KNOWN_FREQ);
 	return pvclock_tsc_khz(this_cpu_pvti());

@@ -366,5 +366,8 @@ void __init kvmclock_init(void)
 		kvm_clock.rating = 299;
 
 	clocksource_register_hz(&kvm_clock, NSEC_PER_SEC);
+
+	kvm_hostclock_init();
+
 	pv_info.name = "KVM";
 }

arch/x86/kernel/kvmhostclock.c new file +38 -0

new file mode 100644
index 000000000000..9971343c2bed
@@ -0,0 +1,130 @@
+// SPDX-License-Identifier: GPL-2.0-only
+/*
+ * KVM clocksource that uses host timekeeping.
+ * Copyright (c) 2019 Suleiman Souhlal, Google LLC
+ */
+
+#include <linux/clocksource.h>
+#include <linux/kvm_para.h>
+#include <asm/pvclock.h>
+#include <asm/msr.h>
+#include <asm/kvmclock.h>
+#include <linux/timekeeper_internal.h>
+
+struct pvclock_timekeeper pv_timekeeper;
+
+static bool pv_timekeeper_enabled;
+static bool pv_timekeeper_present;
+static int old_vclock_mode;
+
+static u64
+kvm_hostclock_get_cycles(struct clocksource *cs)
+{
+	return rdtsc_ordered();
+}
+
+static int
+kvm_hostclock_enable(struct clocksource *cs)
+{
+	pv_timekeeper_enabled = 1;
+
+	old_vclock_mode = kvm_clock.archdata.vclock_mode;
+	kvm_clock.archdata.vclock_mode = VCLOCK_TSC;
+	return 0;
+}
+
+static void
+kvm_hostclock_disable(struct clocksource *cs)
+{

arch/x86/kernel/kvmhostclock.c new file +66 -0

+	.mask = CLOCKSOURCE_MASK(64),
+	.flags = CLOCK_SOURCE_IS_CONTINUOUS,
+};
+
+static void
+pvclock_copy_into_read_base(struct pvclock_timekeeper *pvtk,
+    struct tk_read_base *tkr, struct pvclock_read_base *pvtkr)
+{
+	int shift_diff;
+
+	tkr->mask = pvtkr->mask;
+	tkr->cycle_last = pvtkr->cycle_last + pvtk->tsc_offset;
+	tkr->mult = pvtkr->mult;
+	shift_diff = tkr->shift - pvtkr->shift;
+	tkr->shift = pvtkr->shift;
+	tkr->xtime_nsec = pvtkr->xtime_nsec;
+	tkr->base = pvtkr->base;
+}
+
+static u64
+pvtk_read_begin(struct pvclock_timekeeper *pvtk)
+{
+	u64 gen;
+
+	gen = pvtk->gen & ~1;
+	/* Make sure that the gen count is read before the data. */
+	virt_rmb();
+
+	return gen;
+}
+
+static bool
+pvtk_read_retry(struct pvclock_timekeeper *pvtk, u64 gen)
+{
+	/* Make sure that the gen count is re-read after the data. */
+	virt_rmb();
+	return unlikely(gen != pvtk->gen);
+}
+
+void
+kvm_clock_copy_into_tk(struct timekeeper *tk)
+{
+	struct pvclock_timekeeper *pvtk;
+	u64 gen;
+
+	if (!pv_timekeeper_enabled)
+		return;
+
+	pvtk = &pv_timekeeper;
+	do {
+		gen = pvtk_read_begin(pvtk);
+		if (!(pv_timekeeper.flags & PVCLOCK_TIMEKEEPER_ENABLED))
+			return;
+
+		pvclock_copy_into_read_base(pvtk, &tk->tkr_mono,
+		    &pvtk->tkr_mono);
+		pvclock_copy_into_read_base(pvtk, &tk->tkr_raw, &pvtk->tkr_raw);
+
+		tk->xtime_sec = pvtk->xtime_sec;
+		tk->ktime_sec = pvtk->ktime_sec;
+		tk->wall_to_monotonic.tv_sec = pvtk->wall_to_monotonic_sec;
+		tk->wall_to_monotonic.tv_nsec = pvtk->wall_to_monotonic_nsec;
+		tk->offs_real = pvtk->offs_real;
+		tk->offs_boot = pvtk->offs_boot;
+		tk->offs_tai = pvtk->offs_tai;
+		tk->raw_sec = pvtk->raw_sec;

arch/x86/kernel/kvmhostclock.c new file +6 -0

+	if (pv_timekeeper.flags & PVCLOCK_TIMEKEEPER_ENABLED) {
+		pv_timekeeper_present = 1;
+
+		clocksource_register_khz(&kvm_hostclock, kvm_get_tsc_khz());
+	}
+}

include/linux/timekeeper_internal.h +8 -0

index 84ff2844df2a..43b036375cdc 100644
@@ -153,4 +153,12 @@ static inline void update_vsyscall_tz(void)
 }
 #endif
 
+#ifdef CONFIG_KVM_HOSTCLOCK
+void kvm_clock_copy_into_tk(struct timekeeper *tk);
+#else
+static inline void kvm_clock_copy_into_tk(struct timekeeper *tk)
+{
+}
+#endif
+
 #endif /* _LINUX_TIMEKEEPER_INTERNAL_H */

kernel/time/timekeeping.c +2 -0

index ca69290bee2a..09bcf13b2334 100644
@@ -2107,6 +2107,8 @@ static void timekeeping_advance(enum timekeeping_adv_mode mode)
 	clock_set |= accumulate_nsecs_to_secs(tk);
 
 	write_seqcount_begin(&tk_core.seq);
+	kvm_clock_copy_into_tk(tk);
+
 	/*
 	 * Update the real timekeeper.
 	 *

arch/x86/kernel/kvmhostclock.c new file +6 -0

+kvm_hostclock_enable(struct clocksource *cs)
+{
+	pv_timekeeper_enabled = 1;
+
+	old_vclock_mode = kvm_clock.archdata.vclock_mode;
+	kvm_clock.archdata.vclock_mode = VCLOCK_TSC;

arch/x86/kernel/kvmhostclock.c new file +11 -0

+	pvtk = &pv_timekeeper;
+	do {
+		gen = pvtk_read_begin(pvtk);
+		if (!(pv_timekeeper.flags & PVCLOCK_TIMEKEEPER_ENABLED))
+			return;
+
+		pvclock_copy_into_read_base(pvtk, &tk->tkr_mono,
+		    &pvtk->tkr_mono);
+		pvclock_copy_into_read_base(pvtk, &tk->tkr_raw, &pvtk->tkr_raw);
+
+		tk->xtime_sec = pvtk->xtime_sec;

Quoted 6 lines · Suleiman Souhlal

arch/x86/kernel/kvmhostclock.c new file +6 -0

+kvm_hostclock_enable(struct clocksource *cs)
+{
+     pv_timekeeper_enabled = 1;
+
+     old_vclock_mode = kvm_clock.archdata.vclock_mode;
+     kvm_clock.archdata.vclock_mode = VCLOCK_TSC;

arch/x86/kernel/kvmhostclock.c line 42

Quoted 10 lines

arch/x86/kernel/kvmhostclock.c new file +10 -0

+     return 0;
+}
+
+static void
+kvm_hostclock_disable(struct clocksource *cs)
+{
+     pv_timekeeper_enabled = 0;
+     kvm_clock.archdata.vclock_mode = old_vclock_mode;
+}
+

Why do you poke at kvm_clock? Instead you should add

.archdata = { .vclock_mode = VCLOCK_TSC },

to the kvm_hostclock declaration.

Please also check that the invariant TSC CPUID bit CPUID[0x80000007].EDX[8] is set before enabling this feature.

Paolo

Quoted 15 lines

arch/x86/kernel/kvmhostclock.c new file +15 -0

+     pvtk = &pv_timekeeper;
+     do {
+             gen = pvtk_read_begin(pvtk);
+             if (!(pv_timekeeper.flags & PVCLOCK_TIMEKEEPER_ENABLED))
+                     return;
+
+             pvclock_copy_into_read_base(pvtk, &tk->tkr_mono,
+                 &pvtk->tkr_mono);
+             pvclock_copy_into_read_base(pvtk, &tk->tkr_raw, &pvtk->tkr_raw);
+
+             tk->xtime_sec = pvtk->xtime_sec;
+             tk->ktime_sec = pvtk->ktime_sec;
+             tk->wall_to_monotonic.tv_sec = pvtk->wall_to_monotonic_sec;
+             tk->wall_to_monotonic.tv_nsec = pvtk->wall_to_monotonic_nsec;
+             tk->offs_real = pvtk->offs_real;

arch/x86/kernel/kvmhostclock.c line 117

Quoted 6 lines

arch/x86/kernel/kvmhostclock.c new file +6 -0

+             tk->offs_boot = pvtk->offs_boot;
+             tk->offs_tai = pvtk->offs_tai;
+             tk->raw_sec = pvtk->raw_sec;
+     } while (pvtk_read_retry(pvtk, gen));
+}
+

Should you write an "enabled value" (basically the flags) into pvtk as well?

arch/x86/kernel/kvmhostclock.c line 124

Quoted 6 lines

arch/x86/kernel/kvmhostclock.c new file +6 -0

+kvm_hostclock_init(void)
+{
+     unsigned long pa;
+
+     pa = __pa(&pv_timekeeper);
+     wrmsrl(MSR_KVM_TIMEKEEPER_EN, pa);

As Vitaly said, a new CPUID bit must be defined in Documentation/virt/kvm/cpuid.txt, and used here. Also please make bit 0 an enable bit.

I think we have that already (pvtk->flags).
I'll change the if statement above to use pvtk instead of pv_timekeeper.

Quoted 11 lines

arch/x86/kernel/kvmhostclock.c line 124

Quoted 6 lines

arch/x86/kernel/kvmhostclock.c new file +6 -0

+kvm_hostclock_init(void)
+{
+     unsigned long pa;
+
+     pa = __pa(&pv_timekeeper);
+     wrmsrl(MSR_KVM_TIMEKEEPER_EN, pa);

As Vitaly said, a new CPUID bit must be defined in Documentation/virt/kvm/cpuid.txt, and used here. Also please make bit 0 an enable bit.

I think I might not be able to move the enable bit to 0 because we need the generation count (pvclock_timekeeper.gen) to be the first word of the struct due to the way we copy the data to userspace, similarly to how kvm_setup_pvclock_page() does it.

This RFC is to try to solve the following problem:

We have some applications that are currently running in their own namespace, that still talk to other processes on the machine, using IPC, and expect to run on the same machine.

We want to move them into a virtual machine, for the usual benefits of virtualization.

However, some of these programs use CLOCK_MONOTONIC and CLOCK_BOOTTIME timestamps, as part of their protocol, when talking to the host.

Generally speaking, we have multiple event sources, for example sensors, input devices, display controller vsync, etc and we would like to rely on them in the guest for various scenarios.

As a specific example, we are trying to run some wayland clients (in the guest) who talk to the server (in the host), and the server gives input events based on host time. Additionally, there are also vsync events that the clients use for timing their rendering.

Another use case we have are timestamps from IIO sensors and cameras. There are applications that need to determine how the timestamps relate to the current time and the only way to get current time is clock_gettime(), which would return a value from a different time domain than the timestamps.

In this case, it is not feasible to change these programs, due to the number of the places we would have to change.

We spent some time thinking about this, and the best solution we could come up with was the following:

Make the guest kernel return the same CLOCK_MONOTONIC and CLOCK_GETTIME timestamps as the host.

To do that, I am changing kvmclock to request to the host to copy its timekeeping parameters (mult, base, cycle_last, etc), so that the guest timekeeper can use the same values, so that time can be synchronized between the guest and the host.

I wonder how feasible it is to map the host's vdso into the guest and thus make the guest use the *same* (as opposed to "synchronized") clock as the host's userspace? Another benefit is that it's essentially an ABI so is not changed as liberally as internal structures like timekeeper, etc. There is probably certain complication in handling the syscall fallback in the vdso when used in the guest kernel, though.

You'll also need to ensure neither tsc scaling and nor offsetting is applied to the VM once this clock is enabled.

Roman.

I wonder how feasible it is to map the host's vdso into the guest and thus make the guest use the *same* (as opposed to "synchronized") clock as the host's userspace? Another benefit is that it's essentially an ABI so is not changed as liberally as internal structures like timekeeper, etc. There is probably certain complication in handling the syscall fallback in the vdso when used in the guest kernel, though.

You'll also need to ensure neither tsc scaling and nor offsetting is applied to the VM once this clock is enabled.

Changes in v2:
- Move out of kvmclock and into its own clocksource and file.
- Remove timekeeping.c #ifdefs.
- Fix i386 build.