feat: added pvclock (kvm_clock) support

sys/arch/x86/conf/files.x86

@@ -81,6 +81,10 @@ device	hyperv
 attach	hyperv at cpufeaturebus
 file	arch/x86/x86/hyperv.c		hyperv needs-flag
 
+device	pvclock
+attach	pvclock at cpufeaturebus
+file	arch/x86/x86/pvclock.c		pvclock
+
 file	arch/x86/x86/apic.c		ioapic | lapic
 file	arch/x86/x86/bus_dma.c		machdep
 file	arch/x86/x86/bus_space.c	machdep

sys/arch/x86/x86/pvclock.c

@@ -0,0 +1,235 @@
+/*	$OpenBSD: pvclock.c,v 1.9 2023/02/04 19:19:37 cheloha Exp $	*/
+
+/*
+ * Copyright (c) 2018 Reyk Floeter <reyk@openbsd.org>
+ *
+ * Permission to use, copy, modify, and distribute this software for any
+ * purpose with or without fee is hereby granted, provided that the above
+ * copyright notice and this permission notice appear in all copies.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
+ * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
+ * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
+ * ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
+ * WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN
+ * ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
+ * OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
+ */
+
+#if !defined(__i386__) && !defined(__amd64__)
+#error pvclock(4) is only supported on i386 and amd64
+#endif
+
+#include <sys/cdefs.h>
+
+#include <sys/param.h>
+#include <sys/device.h>
+#include <sys/kernel.h>
+#include <sys/module.h>
+#include <sys/systm.h>
+#include <sys/timetc.h>
+
+#include <machine/cpu.h>
+#include <machine/atomic.h>
+#include <uvm/uvm_extern.h>
+
+#include <x86/x86/pvreg.h>
+
+uint pvclock_lastcount;
+
+struct pvclock_softc {
+	device_t		 sc_dev;
+	void			*sc_time;
+	paddr_t			 sc_paddr;
+	struct timecounter	*sc_tc;
+};
+
+static int	 pvclock_match(device_t, cfdata_t, void *);
+static void	 pvclock_attach(device_t, device_t, void *);
+static int	 pvclock_activate(device_t, devact_t);
+
+/*
+void	 pvclock_read_time_info(struct pvclock_softc *,
+	    struct pvclock_time_info *);
+*/
+static inline uint32_t
+	pvclock_read_begin(const struct pvclock_time_info *);
+static inline int
+	pvclock_read_done(const struct pvclock_time_info *, uint32_t);
+static uint
+	pvclock_get_timecount(struct timecounter *);
+
+struct timecounter pvclock_timecounter = {
+	.tc_get_timecount = pvclock_get_timecount,
+	.tc_counter_mask = ~0u,
+	.tc_frequency = 0,
+	.tc_name = NULL,
+	.tc_quality = -2000,
+	.tc_priv = NULL,
+};
+
+CFATTACH_DECL_NEW(pvclock, sizeof(struct pvclock_softc),
+	pvclock_match,
+	pvclock_attach,
+	NULL /* detach */,
+	pvclock_activate);
+
+
+static int
+pvclock_match(device_t parent, cfdata_t cf, void *aux)
+{
+	u_int regs[6];
+	/*
+	 * pvclock is provided by different hypervisors, we currently
+	 * only support the "kvmclock".
+	 */
+	x86_cpuid(0x40000000 + CPUID_OFFSET_KVM_FEATURES, regs);
+	/*
+	 * We only implement support for the 2nd version of pvclock.
+	 * The first version is basically the same but with different
+	 * non-standard MSRs and it is deprecated.
+	 */
+	if ((regs[0] & (1 << KVM_FEATURE_CLOCKSOURCE2)) == 0)
+		return (0);
+
+	/*
+	 * Only the "stable" clock with a sync'ed TSC is supported.
+	 * In this case the host guarantees that the TSC is constant
+	 * and invariant, either by the underlying TSC or by passing
+	 * on a synchronized value.
+	 */
+	if ((regs[0] &
+	    (1 << KVM_FEATURE_CLOCKSOURCE_STABLE_BIT)) == 0)
+		return (0);
+
+	return (1);
+}
+
+static inline uint32_t
+pvclock_read_begin(const struct pvclock_time_info *ti)
+{
+	uint32_t ti_version = ti->ti_version & ~0x1;
+	virtio_membar_sync();
+	return (ti_version);
+}
+
+static inline int
+pvclock_read_done(const struct pvclock_time_info *ti,
+    uint32_t ti_version)
+{
+	virtio_membar_sync();
+	return (ti->ti_version == ti_version);
+}
+
+static uint
+pvclock_get_timecount(struct timecounter *tc)
+{
+	struct pvclock_softc		*sc = tc->tc_priv;
+	struct pvclock_time_info	*ti;
+	uint64_t			 tsc_timestamp, system_time, delta, ctr;
+	uint32_t			 ti_version, mul_frac;
+	int8_t				 shift;
+	uint8_t				 flags;
+
+	ti = sc->sc_time;
+	do {
+		ti_version = pvclock_read_begin(ti);
+		system_time = ti->ti_system_time;
+		tsc_timestamp = ti->ti_tsc_timestamp;
+		mul_frac = ti->ti_tsc_to_system_mul;
+		shift = ti->ti_tsc_shift;
+		flags = ti->ti_flags;
+	} while (!pvclock_read_done(ti, ti_version));
+
+	/*
+	 * The algorithm is described in
+	 * linux/Documentation/virtual/kvm/msr.txt
+	 */
+	delta = rdtsc() - tsc_timestamp;
+	if (shift < 0)
+		delta >>= -shift;
+	else
+		delta <<= shift;
+	ctr = ((delta * mul_frac) >> 32) + system_time;
+
+	if ((flags & PVCLOCK_FLAG_TSC_STABLE) != 0)
+		return (ctr);
+
+	if (ctr < pvclock_lastcount)
+		return (pvclock_lastcount);
+
+	atomic_swap_uint(&pvclock_lastcount, ctr);
+
+	return (ctr);
+}
+
+static void
+pvclock_attach(device_t parent, device_t self, void *aux)
+{
+	struct pvclock_softc		*sc = device_private(self);
+	struct pvclock_time_info	*ti;
+	paddr_t			 	 pa;
+	uint32_t			 ti_version;
+	uint8_t				 flags;
+
+	if ((sc->sc_time = (void *)uvm_km_alloc(kernel_map,
+		PAGE_SIZE, PAGE_SIZE,
+		UVM_KMF_WIRED | UVM_KMF_ZERO)) == NULL) {
+		aprint_error("time page allocation failed\n");
+		return;
+	}
+
+	if (!pmap_extract(pmap_kernel(), (vaddr_t)sc->sc_time, &pa)) {
+		aprint_error("time page PA extraction failed\n");
+		uvm_km_free(kernel_map, (vaddr_t)sc->sc_time,
+			PAGE_SIZE, UVM_KMF_WIRED);
+		sc->sc_time = NULL;
+		return;
+	}
+
+	wrmsr(KVM_MSR_SYSTEM_TIME, pa | PVCLOCK_SYSTEM_TIME_ENABLE);
+	sc->sc_paddr = pa;
+
+	sc->sc_dev = self;
+
+	ti = sc->sc_time;
+	do {
+		ti_version = pvclock_read_begin(ti);
+		flags = ti->ti_flags;
+	} while (!pvclock_read_done(ti, ti_version));
+
+	sc->sc_tc = &pvclock_timecounter;
+	sc->sc_tc->tc_name = device_xname(sc->sc_dev);
+
+	sc->sc_tc->tc_frequency = 1000000000ULL;
+	sc->sc_tc->tc_priv = sc;
+
+	pvclock_lastcount = 0;
+
+	/* Better than HPET but below TSC */
+	sc->sc_tc->tc_quality = 1500;
+
+	if ((flags & PVCLOCK_FLAG_TSC_STABLE) == 0) {
+		/* if tsc is not stable, set a lower priority */
+		/* Better than i8254 but below HPET */
+		sc->sc_tc->tc_quality = 500;
+	}
+
+	tc_init(sc->sc_tc);
+}
+
+int
+pvclock_activate(device_t self, devact_t act)
+{
+	struct pvclock_softc	*sc = (struct pvclock_softc *)self;
+	int			 rv = 0;
+	paddr_t			 pa = sc->sc_paddr;
+
+	switch (act) {
+	case DVACT_DEACTIVATE:
+		wrmsr(KVM_MSR_SYSTEM_TIME, pa & ~PVCLOCK_SYSTEM_TIME_ENABLE);
+		break;
+	}
+
+	return (rv);
+}

sys/arch/x86/x86/pvreg.h

@@ -0,0 +1,118 @@
+/*	$OpenBSD: pvreg.h,v 1.6 2019/05/13 15:40:34 pd Exp $	*/
+
+/*
+ * Copyright (c) 2015 Reyk Floeter <reyk@openbsd.org>
+ * Copyright (c) 2015 Stefan Fritsch <sf@sfritsch.de>
+ *
+ * Permission to use, copy, modify, and distribute this software for any
+ * purpose with or without fee is hereby granted, provided that the above
+ * copyright notice and this permission notice appear in all copies.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
+ * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
+ * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
+ * ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
+ * WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN
+ * ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
+ * OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
+ */
+
+#ifndef _DEV_PV_PVBUS_H_
+#define _DEV_PV_PVBUS_H_
+
+#define	CPUID_HV_SIGNATURE_START	0x40000000
+#define	CPUID_HV_SIGNATURE_END		0x40010000
+#define	CPUID_HV_SIGNATURE_STEP		0x100
+#define CPUID_HV_SIGNATURE_STRLEN	12
+
+/*
+ * KVM
+ */
+#define	CPUID_OFFSET_KVM_FEATURES		0x1
+
+#define	KVM_FEATURE_CLOCKSOURCE			0	/* deprecated */
+#define	KVM_FEATURE_NOP_IO_DELAY		1
+#define	KVM_FEATURE_MMU_OP			2	/* deprecated */
+#define	KVM_FEATURE_CLOCKSOURCE2		3
+#define	KVM_FEATURE_ASYNC_PF			4
+#define	KVM_FEATURE_STEAL_TIME			5
+#define	KVM_FEATURE_PV_EOI			6
+#define	KVM_FEATURE_PV_UNHALT			7
+#define	KVM_FEATURE_CLOCKSOURCE_STABLE_BIT	24
+
+#define	KVM_MSR_EOI_EN				0x4b564d04
+#define KVM_PV_EOI_BIT				0
+
+#define KVM_MSR_WALL_CLOCK			0x4b564d00
+#define KVM_MSR_SYSTEM_TIME			0x4b564d01
+
+struct pvclock_wall_clock {
+	uint32_t		 wc_version;
+	uint32_t		 wc_sec;
+	uint32_t		 wc_nsec;
+} __packed;
+
+struct pvclock_time_info {
+	uint32_t		 ti_version;
+	uint32_t		 ti_pad0;
+	uint64_t		 ti_tsc_timestamp;
+	uint64_t		 ti_system_time;
+	uint32_t		 ti_tsc_to_system_mul;
+	int8_t			 ti_tsc_shift;
+	uint8_t			 ti_flags;
+	uint8_t			 ti_pad[2];
+} __packed;
+
+#define PVCLOCK_FLAG_TSC_STABLE			0x01
+#define PVCLOCK_SYSTEM_TIME_ENABLE		0x01
+
+/*
+ * Hyper-V
+ */
+#define CPUID_OFFSET_HYPERV_INTERFACE		0x1
+#define CPUID_OFFSET_HYPERV_VERSION		0x2
+#define CPUID_OFFSET_HYPERV_FEATURES		0x3
+#define CPUID_OFFSET_HYPERV_ENLIGHTENMENT_INFO	0x4
+#define CPUID_OFFSET_HYPERV_IMPL_LIMITS		0x5
+
+#define HYPERV_VERSION_EAX_BUILD_NUMBER		0
+#define HYPERV_VERSION_EBX_MAJOR_M		0xffff0000
+#define HYPERV_VERSION_EBX_MAJOR_S		16
+#define HYPERV_VERSION_EBX_MINOR_M		0x0000ffff
+#define HYPERV_VERSION_EBX_MINOR_S		0
+#define HYPERV_VERSION_ECX_SERVICE_PACK		0
+#define HYPERV_VERSION_EDX_SERVICE_BRANCH_M	0xff000000
+#define HYPERV_VERSION_EDX_SERVICE_BRANCH_S	24
+#define HYPERV_VERSION_EDX_SERVICE_NUMBER_M	0x00ffffff
+#define HYPERV_VERSION_EDX_SERVICE_NUMBER_S	0
+
+#define HYPERV_VERSION_WS2008			0x00060000
+#define HYPERV_VERSION_WIN7			0x00060001
+#define HYPERV_VERSION_WIN8			0x00060002
+#define HYPERV_VERSION_WIN8_1			0x00060003
+#define HYPERV_VERSION_WIN10			0x00100000
+
+#define HYPERV_FEATURE_EAX_VP_RUNTIME		0
+#define HYPERV_FEATURE_EAX_TIME_REF_COUNT	1
+#define HYPERV_FEATURE_EAX_SYNIC		2
+#define HYPERV_FEATURE_EAX_STIMER		3
+#define HYPERV_FEATURE_EAX_APIC			4
+#define HYPERV_FEATURE_EAX_HYPERCALL		5
+#define HYPERV_FEATURE_EAX_VP_INDEX		6
+#define HYPERV_FEATURE_EAX_MSR_RESET		7
+#define HYPERV_FEATURE_EAX_STATS_PAGES		8
+#define HYPERV_FEATURE_EAX_REF_TSC		9
+#define HYPERV_FEATURE_EAX_GUEST_IDLE		10
+#define HYPERV_FEATURE_EAX_TIMER_FREQ		11
+#define HYPERV_FEATURE_EAX_DEBUG		12
+
+/*
+ * Xen
+ */
+#define CPUID_OFFSET_XEN_VERSION		0x1
+#define CPUID_OFFSET_XEN_HYPERCALL		0x2
+
+#define XEN_VERSION_MAJOR_S			16
+#define XEN_VERSION_MINOR_M			0xffff
+
+#endif /* _DEV_PV_PVBUS_H_ */