diff mbox series

[RFC,2/2] x86/kvmclock: Use host timekeeping.

Message ID 20190920062713.78503-3-suleiman@google.com (mailing list archive)
State New, archived
Headers show
Series kvm: Use host timekeeping in guest. | expand

Commit Message

Suleiman Souhlal Sept. 20, 2019, 6:27 a.m. UTC
When CONFIG_KVMCLOCK_HOST_TIMEKEEPING is enabled, 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.

When enabled, kvmclock uses raw tsc instead of pvclock.

When enabled, syscalls that can change time, such as settimeofday(2)
or adj_timex(2) are disabled in the guest.

Signed-off-by: Suleiman Souhlal <suleiman@google.com>
---
 arch/x86/Kconfig                |   9 +++
 arch/x86/include/asm/kvmclock.h |   2 +
 arch/x86/kernel/kvmclock.c      | 127 +++++++++++++++++++++++++++++++-
 kernel/time/timekeeping.c       |  21 ++++++
 4 files changed, 155 insertions(+), 4 deletions(-)

Comments

Vitaly Kuznetsov Sept. 20, 2019, 1:33 p.m. UTC | #1
Suleiman Souhlal <suleiman@google.com> writes:

> When CONFIG_KVMCLOCK_HOST_TIMEKEEPING is enabled, 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.
>
> When enabled, kvmclock uses raw tsc instead of pvclock.
>
> When enabled, syscalls that can change time, such as settimeofday(2)
> or adj_timex(2) are disabled in the guest.
>
> Signed-off-by: Suleiman Souhlal <suleiman@google.com>
> ---
>  arch/x86/Kconfig                |   9 +++
>  arch/x86/include/asm/kvmclock.h |   2 +
>  arch/x86/kernel/kvmclock.c      | 127 +++++++++++++++++++++++++++++++-
>  kernel/time/timekeeping.c       |  21 ++++++
>  4 files changed, 155 insertions(+), 4 deletions(-)
>
> diff --git a/arch/x86/Kconfig b/arch/x86/Kconfig
> index 4195f44c6a09..37299377d9d7 100644
> --- a/arch/x86/Kconfig
> +++ b/arch/x86/Kconfig
> @@ -837,6 +837,15 @@ config PARAVIRT_TIME_ACCOUNTING
>  config PARAVIRT_CLOCK
>  	bool
>  
> +config KVMCLOCK_HOST_TIMEKEEPING
> +	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.
> +

I'd also like to speak up against this config, it is confusing. In case
the goal is to come up with a TSC-based clock for guests which will
return the same as clock_gettime() on the host (or, is the goal to just
have the same reading for all guests on the host?) I'd suggest we create
a separate (from KVMCLOCK) clocksource (mirroring host timekeeper) and
guests will be free to pick the one they like.
Suleiman Souhlal Sept. 24, 2019, 8:10 a.m. UTC | #2
On Fri, Sep 20, 2019 at 10:33 PM Vitaly Kuznetsov <vkuznets@redhat.com> wrote:
>
> Suleiman Souhlal <suleiman@google.com> writes:
>
> > When CONFIG_KVMCLOCK_HOST_TIMEKEEPING is enabled, 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.
> >
> > When enabled, kvmclock uses raw tsc instead of pvclock.
> >
> > When enabled, syscalls that can change time, such as settimeofday(2)
> > or adj_timex(2) are disabled in the guest.
> >
> > Signed-off-by: Suleiman Souhlal <suleiman@google.com>
> > ---
> >  arch/x86/Kconfig                |   9 +++
> >  arch/x86/include/asm/kvmclock.h |   2 +
> >  arch/x86/kernel/kvmclock.c      | 127 +++++++++++++++++++++++++++++++-
> >  kernel/time/timekeeping.c       |  21 ++++++
> >  4 files changed, 155 insertions(+), 4 deletions(-)
> >
> > diff --git a/arch/x86/Kconfig b/arch/x86/Kconfig
> > index 4195f44c6a09..37299377d9d7 100644
> > --- a/arch/x86/Kconfig
> > +++ b/arch/x86/Kconfig
> > @@ -837,6 +837,15 @@ config PARAVIRT_TIME_ACCOUNTING
> >  config PARAVIRT_CLOCK
> >       bool
> >
> > +config KVMCLOCK_HOST_TIMEKEEPING
> > +     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.
> > +
>
> I'd also like to speak up against this config, it is confusing. In case
> the goal is to come up with a TSC-based clock for guests which will
> return the same as clock_gettime() on the host (or, is the goal to just
> have the same reading for all guests on the host?) I'd suggest we create
> a separate (from KVMCLOCK) clocksource (mirroring host timekeeper) and
> guests will be free to pick the one they like.

Fair enough. I'll do that in the next version of the patch.

The goal is to have a guest clock that gives the same
clock_gettime(CLOCK_MONOTONIC) as the host.

Thanks,
-- Suleiman
Vitaly Kuznetsov Sept. 24, 2019, 11:14 a.m. UTC | #3
Suleiman Souhlal <suleiman@google.com> writes:

> On Fri, Sep 20, 2019 at 10:33 PM Vitaly Kuznetsov <vkuznets@redhat.com> wrote:
>>
>> Suleiman Souhlal <suleiman@google.com> writes:
>>
>> > When CONFIG_KVMCLOCK_HOST_TIMEKEEPING is enabled, 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.
>> >
>> > When enabled, kvmclock uses raw tsc instead of pvclock.
>> >
>> > When enabled, syscalls that can change time, such as settimeofday(2)
>> > or adj_timex(2) are disabled in the guest.
>> >
>> > Signed-off-by: Suleiman Souhlal <suleiman@google.com>
>> > ---
>> >  arch/x86/Kconfig                |   9 +++
>> >  arch/x86/include/asm/kvmclock.h |   2 +
>> >  arch/x86/kernel/kvmclock.c      | 127 +++++++++++++++++++++++++++++++-
>> >  kernel/time/timekeeping.c       |  21 ++++++
>> >  4 files changed, 155 insertions(+), 4 deletions(-)
>> >
>> > diff --git a/arch/x86/Kconfig b/arch/x86/Kconfig
>> > index 4195f44c6a09..37299377d9d7 100644
>> > --- a/arch/x86/Kconfig
>> > +++ b/arch/x86/Kconfig
>> > @@ -837,6 +837,15 @@ config PARAVIRT_TIME_ACCOUNTING
>> >  config PARAVIRT_CLOCK
>> >       bool
>> >
>> > +config KVMCLOCK_HOST_TIMEKEEPING
>> > +     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.
>> > +
>>
>> I'd also like to speak up against this config, it is confusing. In case
>> the goal is to come up with a TSC-based clock for guests which will
>> return the same as clock_gettime() on the host (or, is the goal to just
>> have the same reading for all guests on the host?) I'd suggest we create
>> a separate (from KVMCLOCK) clocksource (mirroring host timekeeper) and
>> guests will be free to pick the one they like.
>
> Fair enough. I'll do that in the next version of the patch.
>
> The goal is to have a guest clock that gives the same
> clock_gettime(CLOCK_MONOTONIC) as the host.
>

KVMCLOCK has lots of legacy derived from times when TSC synchronization
was not a given (I heard that this is still sometimes problematic with
multi-socket systems but oh well). If I was to design a new clock I'd
probably mirror Hyper-V's TSC page clocksource invalidating the page
when host timekeeper values are updated (and making guest spin).

The tricky part with this approach is probably tsc scaling/tsc
offsetting which is still going to be controlled by kvmclock (so the
guest has no option to read 'pure TSC'). As an alternative, you can make
kvmclock un-pluggable so when it's not enabled TSC frequency/offset can
remain intact. You can, of course, try to update timekeeper values to
match the new frequency/offset every time they change but rounding
errors may bite.
diff mbox series

Patch

diff --git a/arch/x86/Kconfig b/arch/x86/Kconfig
index 4195f44c6a09..37299377d9d7 100644
--- a/arch/x86/Kconfig
+++ b/arch/x86/Kconfig
@@ -837,6 +837,15 @@  config PARAVIRT_TIME_ACCOUNTING
 config PARAVIRT_CLOCK
 	bool
 
+config KVMCLOCK_HOST_TIMEKEEPING
+	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
diff --git a/arch/x86/include/asm/kvmclock.h b/arch/x86/include/asm/kvmclock.h
index eceea9299097..cad7eae62b54 100644
--- a/arch/x86/include/asm/kvmclock.h
+++ b/arch/x86/include/asm/kvmclock.h
@@ -4,4 +4,6 @@ 
 
 extern struct clocksource kvm_clock;
 
+bool kvm_clock_copy_into_tk(struct timekeeper *tk);
+
 #endif /* _ASM_X86_KVM_CLOCK_H */
diff --git a/arch/x86/kernel/kvmclock.c b/arch/x86/kernel/kvmclock.c
index 904494b924c1..e3a33dcaffc8 100644
--- a/arch/x86/kernel/kvmclock.c
+++ b/arch/x86/kernel/kvmclock.c
@@ -16,6 +16,7 @@ 
 #include <linux/mm.h>
 #include <linux/slab.h>
 #include <linux/set_memory.h>
+#include <linux/timekeeper_internal.h>
 
 #include <asm/hypervisor.h>
 #include <asm/mem_encrypt.h>
@@ -29,6 +30,9 @@  static int msr_kvm_system_time __ro_after_init = MSR_KVM_SYSTEM_TIME;
 static int msr_kvm_wall_clock __ro_after_init = MSR_KVM_WALL_CLOCK;
 static u64 kvm_sched_clock_offset __ro_after_init;
 
+static bool pv_timekeeper_enabled;
+static bool pv_timekeeper_present;
+
 static int __init parse_no_kvmclock(char *arg)
 {
 	kvmclock = 0;
@@ -54,6 +58,8 @@  static struct pvclock_wall_clock wall_clock __bss_decrypted;
 static DEFINE_PER_CPU(struct pvclock_vsyscall_time_info *, hv_clock_per_cpu);
 static struct pvclock_vsyscall_time_info *hvclock_mem;
 
+struct pvclock_timekeeper pv_timekeeper;
+
 static inline struct pvclock_vcpu_time_info *this_cpu_pvti(void)
 {
 	return &this_cpu_read(hv_clock_per_cpu)->pvti;
@@ -94,7 +100,10 @@  static u64 kvm_clock_read(void)
 
 static u64 kvm_clock_get_cycles(struct clocksource *cs)
 {
-	return kvm_clock_read();
+	if (pv_timekeeper_present)
+		return rdtsc_ordered();
+	else
+		return kvm_clock_read();
 }
 
 static u64 kvm_sched_clock_read(void)
@@ -159,9 +168,26 @@  bool kvm_check_and_clear_guest_paused(void)
 	return ret;
 }
 
+int
+kvm_clock_enable(struct clocksource *cs)
+{
+	if (pv_timekeeper_present)
+		pv_timekeeper_enabled = 1;
+
+	return 0;
+}
+
+void
+kvm_clock_disable(struct clocksource *cs)
+{
+	pv_timekeeper_enabled = 0;
+}
+
 struct clocksource kvm_clock = {
 	.name	= "kvm-clock",
 	.read	= kvm_clock_get_cycles,
+	.enable = kvm_clock_enable,
+	.disable = kvm_clock_disable,
 	.rating	= 400,
 	.mask	= CLOCKSOURCE_MASK(64),
 	.flags	= CLOCK_SOURCE_IS_CONTINUOUS,
@@ -197,6 +223,86 @@  static void kvm_setup_secondary_clock(void)
 }
 #endif
 
+#ifdef CONFIG_KVMCLOCK_HOST_TIMEKEEPING
+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);
+}
+
+bool
+kvm_clock_copy_into_tk(struct timekeeper *tk)
+{
+	struct pvclock_timekeeper *pvtk;
+	u64 gen;
+
+	if (!pv_timekeeper_enabled)
+		return false;
+
+	pvtk = &pv_timekeeper;
+	do {
+		gen = pvtk_read_begin(pvtk);
+		if (!(pv_timekeeper.flags & PVCLOCK_TIMEKEEPER_ENABLED))
+			return false;
+
+		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;
+	} while (pvtk_read_retry(pvtk, gen));
+
+	return true;
+}
+
+static void
+kvm_register_timekeeper(void)
+{
+	unsigned long pa;
+
+	pa = __pa(&pv_timekeeper);
+	wrmsrl(MSR_KVM_TIMEKEEPER_EN, pa);
+	if (pv_timekeeper.flags & PVCLOCK_TIMEKEEPER_ENABLED)
+		pv_timekeeper_present = 1;
+}
+#endif /* CONFIG_KVMCLOCK_HOST_TIMEKEEPING */
+
 /*
  * After the clock is registered, the host will keep writing to the
  * registered memory location. If the guest happens to shutdown, this memory
@@ -271,8 +377,10 @@  static int __init kvm_setup_vsyscall_timeinfo(void)
 	flags = pvclock_read_flags(&hv_clock_boot[0].pvti);
 	if (!(flags & PVCLOCK_TSC_STABLE_BIT))
 		return 0;
-
-	kvm_clock.archdata.vclock_mode = VCLOCK_PVCLOCK;
+	if (pv_timekeeper_present)
+		kvm_clock.archdata.vclock_mode = VCLOCK_TSC;
+	else
+		kvm_clock.archdata.vclock_mode = VCLOCK_PVCLOCK;
 #endif
 
 	kvmclock_init_mem();
@@ -312,6 +420,10 @@  void __init kvmclock_init(void)
 	if (!kvm_para_available() || !kvmclock)
 		return;
 
+#ifdef CONFIG_KVMCLOCK_HOST_TIMEKEEPING
+	kvm_register_timekeeper();
+#endif
+
 	if (kvm_para_has_feature(KVM_FEATURE_CLOCKSOURCE2)) {
 		msr_kvm_system_time = MSR_KVM_SYSTEM_TIME_NEW;
 		msr_kvm_wall_clock = MSR_KVM_WALL_CLOCK_NEW;
@@ -360,11 +472,18 @@  void __init kvmclock_init(void)
 	 * can use TSC as clocksource.
 	 *
 	 */
+#ifndef CONFIG_KVMCLOCK_HOST_TIMEKEEPING
 	if (boot_cpu_has(X86_FEATURE_CONSTANT_TSC) &&
 	    boot_cpu_has(X86_FEATURE_NONSTOP_TSC) &&
 	    !check_tsc_unstable())
 		kvm_clock.rating = 299;
+#endif
+
+	if (pv_timekeeper_present)
+		clocksource_register_khz(&kvm_clock,
+		    pvclock_tsc_khz(this_cpu_pvti()));
+	else
+		clocksource_register_hz(&kvm_clock, NSEC_PER_SEC);
 
-	clocksource_register_hz(&kvm_clock, NSEC_PER_SEC);
 	pv_info.name = "KVM";
 }
diff --git a/kernel/time/timekeeping.c b/kernel/time/timekeeping.c
index ca69290bee2a..c6e4785babb1 100644
--- a/kernel/time/timekeeping.c
+++ b/kernel/time/timekeeping.c
@@ -23,6 +23,10 @@ 
 #include <linux/compiler.h>
 #include <linux/audit.h>
 
+#ifdef CONFIG_KVMCLOCK_HOST_TIMEKEEPING
+#include <asm/kvmclock.h>
+#endif
+
 #include "tick-internal.h"
 #include "ntp_internal.h"
 #include "timekeeping_internal.h"
@@ -1223,6 +1227,7 @@  EXPORT_SYMBOL_GPL(get_device_system_crosststamp);
  */
 int do_settimeofday64(const struct timespec64 *ts)
 {
+#ifndef CONFIG_KVMCLOCK_HOST_TIMEKEEPING
 	struct timekeeper *tk = &tk_core.timekeeper;
 	struct timespec64 ts_delta, xt;
 	unsigned long flags;
@@ -1261,9 +1266,13 @@  int do_settimeofday64(const struct timespec64 *ts)
 		audit_tk_injoffset(ts_delta);
 
 	return ret;
+#else /* CONFIG_KVMCLOCK_HOST_TIMEKEEPING */
+	return -EINVAL;
+#endif /* CONFIG_KVMCLOCK_HOST_TIMEKEEPING */
 }
 EXPORT_SYMBOL(do_settimeofday64);
 
+#ifndef CONFIG_KVMCLOCK_HOST_TIMEKEEPING
 /**
  * timekeeping_inject_offset - Adds or subtracts from the current time.
  * @tv:		pointer to the timespec variable containing the offset
@@ -1307,6 +1316,7 @@  static int timekeeping_inject_offset(const struct timespec64 *ts)
 
 	return ret;
 }
+#endif /* CONFIG_KVMCLOCK_HOST_TIMEKEEPING */
 
 /*
  * Indicates if there is an offset between the system clock and the hardware
@@ -1332,6 +1342,7 @@  int persistent_clock_is_local;
  */
 void timekeeping_warp_clock(void)
 {
+#ifndef CONFIG_KVMCLOCK_HOST_TIMEKEEPING
 	if (sys_tz.tz_minuteswest != 0) {
 		struct timespec64 adjust;
 
@@ -1340,6 +1351,7 @@  void timekeeping_warp_clock(void)
 		adjust.tv_nsec = 0;
 		timekeeping_inject_offset(&adjust);
 	}
+#endif /* CONFIG_KVMCLOCK_HOST_TIMEKEEPING */
 }
 
 /**
@@ -2107,6 +2119,9 @@  static void timekeeping_advance(enum timekeeping_adv_mode mode)
 	clock_set |= accumulate_nsecs_to_secs(tk);
 
 	write_seqcount_begin(&tk_core.seq);
+#ifdef CONFIG_KVMCLOCK_HOST_TIMEKEEPING
+	kvm_clock_copy_into_tk(tk);
+#endif
 	/*
 	 * Update the real timekeeper.
 	 *
@@ -2241,6 +2256,7 @@  ktime_t ktime_get_update_offsets_now(unsigned int *cwsseq, ktime_t *offs_real,
 	return base;
 }
 
+#ifndef CONFIG_KVMCLOCK_HOST_TIMEKEEPING
 /**
  * timekeeping_validate_timex - Ensures the timex is ok for use in do_adjtimex
  */
@@ -2305,6 +2321,7 @@  static int timekeeping_validate_timex(const struct __kernel_timex *txc)
 
 	return 0;
 }
+#endif /* CONFIG_KVMCLOCK_HOST_TIMEKEEPING */
 
 
 /**
@@ -2312,6 +2329,7 @@  static int timekeeping_validate_timex(const struct __kernel_timex *txc)
  */
 int do_adjtimex(struct __kernel_timex *txc)
 {
+#ifndef CONFIG_KVMCLOCK_HOST_TIMEKEEPING
 	struct timekeeper *tk = &tk_core.timekeeper;
 	struct audit_ntp_data ad;
 	unsigned long flags;
@@ -2368,6 +2386,9 @@  int do_adjtimex(struct __kernel_timex *txc)
 	ntp_notify_cmos_timer();
 
 	return ret;
+#else /* CONFIG_KVMCLOCK_HOST_TIMEKEEPING */
+	return -EINVAL;
+#endif /* CONFIG_KVMCLOCK_HOST_TIMEKEEPING */
 }
 
 #ifdef CONFIG_NTP_PPS