ptp: Add support for the AMZNC10C 'vmclock' device

The vmclock device addresses the problem of live migration with
precision clocks. The tolerances of a hardware counter (e.g. TSC) are
typically around ±50PPM. A guest will use NTP/PTP/PPS to discipline that
counter against an external source of 'real' time, and track the precise
frequency of the counter as it changes with environmental conditions.

When a guest is live migrated, anything it knows about the frequency of
the underlying counter becomes invalid. It may move from a host where
the counter running at -50PPM of its nominal frequency, to a host where
it runs at +50PPM. There will also be a step change in the value of the
counter, as the correctness of its absolute value at migration is
limited by the accuracy of the source and destination host's time
synchronization.

In its simplest form, the device merely advertises a 'disruption_marker'
which indicates that the guest should throw away any NTP synchronization
it thinks it has, and start again.

Because the shared memory region can be exposed all the way to userspace
through the /dev/vmclock0 node, applications can still use time from a
fast vDSO 'system call', and check the disruption marker to be sure that
their timestamp is indeed truthful.

The structure also allows for the precise time, as known by the host, to
be exposed directly to guests so that they don't have to wait for NTP to
resync from scratch. The PTP driver consumes this information if present.
Like the KVM PTP clock, this PTP driver can convert TSC-based cross
timestamps into KVM clock values. Unlike the KVM PTP clock, it does so
only when such is actually helpful.

The values and fields are based on the nascent virtio-rtc specification,
and the intent is that a version (hopefully precisely this version) of
this structure will be included as an optional part of that spec. In the
meantime, this driver supports the simple ACPI form of the device which
is being shipped in certain commercial hypervisors (and submitted for
inclusion in QEMU).

Signed-off-by: David Woodhouse <[e-mail]>
Acked-by: Richard Cochran <[e-mail]>
Signed-off-by: David S. Miller <[e-mail]>

MAINTAINERS

···
 F:	drivers/ptp/ptp_vclock.c
 F:	net/ethtool/phc_vclocks.c
 
+PTP VMCLOCK SUPPORT
+M:	David Woodhouse <[e-mail]>
+L:	[e-mail]
+S:	Maintained
+F:	drivers/ptp/ptp_vmclock.c
+F:	include/uapi/linux/vmclock-abi.h
+
 PTRACE SUPPORT
 M:	Oleg Nesterov <[e-mail]>
 S:	Maintained

drivers/ptp/Kconfig

···
 	  To compile this driver as a module, choose M here: the module
 	  will be called ptp_kvm.
 
+config PTP_1588_CLOCK_VMCLOCK
+	tristate "Virtual machine PTP clock"
+	depends on X86_TSC || ARM_ARCH_TIMER
+	depends on PTP_1588_CLOCK && ACPI && ARCH_SUPPORTS_INT128
+	default y
+	help
+	  This driver adds support for using a virtual precision clock
+	  advertised by the hypervisor. This clock is only useful in virtual
+	  machines where such a device is present.
+
+	  To compile this driver as a module, choose M here: the module
+	  will be called ptp_vmclock.
+
 config PTP_1588_CLOCK_IDT82P33
 	tristate "IDT 82P33xxx PTP clock"
 	depends on PTP_1588_CLOCK && I2C

drivers/ptp/Makefile

···
 obj-$(CONFIG_PTP_1588_CLOCK_INES)	+= ptp_ines.o
 obj-$(CONFIG_PTP_1588_CLOCK_PCH)	+= ptp_pch.o
 obj-$(CONFIG_PTP_1588_CLOCK_KVM)	+= ptp_kvm.o
+obj-$(CONFIG_PTP_1588_CLOCK_VMCLOCK)	+= ptp_vmclock.o
 obj-$(CONFIG_PTP_1588_CLOCK_QORIQ)	+= ptp-qoriq.o
 ptp-qoriq-y				+= ptp_qoriq.o
 ptp-qoriq-$(CONFIG_DEBUG_FS)		+= ptp_qoriq_debugfs.o

drivers/ptp/ptp_vmclock.c

···
+// SPDX-License-Identifier: GPL-2.0-or-later
+/*
+ * Virtual PTP 1588 clock for use with LM-safe VMclock device.
+ *
+ * Copyright © 2024 Amazon.com, Inc. or its affiliates.
+ */
+
+#include <linux/acpi.h>
+#include <linux/device.h>
+#include <linux/err.h>
+#include <linux/file.h>
+#include <linux/fs.h>
+#include <linux/init.h>
+#include <linux/kernel.h>
+#include <linux/miscdevice.h>
+#include <linux/mm.h>
+#include <linux/module.h>
+#include <linux/platform_device.h>
+#include <linux/slab.h>
+
+#include <uapi/linux/vmclock-abi.h>
+
+#include <linux/ptp_clock_kernel.h>
+
+#ifdef CONFIG_X86
+#include <asm/pvclock.h>
+#include <asm/kvmclock.h>
+#endif
+
+#ifdef CONFIG_KVM_GUEST
+#define SUPPORT_KVMCLOCK
+#endif
+
+static DEFINE_IDA(vmclock_ida);
+
+ACPI_MODULE_NAME("vmclock");
+
+struct vmclock_state {
+	struct resource res;
+	struct vmclock_abi *clk;
+	struct miscdevice miscdev;
+	struct ptp_clock_info ptp_clock_info;
+	struct ptp_clock *ptp_clock;
+	enum clocksource_ids cs_id, sys_cs_id;
+	int index;
+	char *name;
+};
+
+#define VMCLOCK_MAX_WAIT ms_to_ktime(100)
+
+/* Require at least the flags field to be present. All else can be optional. */
+#define VMCLOCK_MIN_SIZE offsetof(struct vmclock_abi, pad)
+
+#define VMCLOCK_FIELD_PRESENT(_c, _f)			  \
+	(le32_to_cpu((_c)->size) >= (offsetof(struct vmclock_abi, _f) +	\
+				     sizeof((_c)->_f)))
+
+/*
+ * Multiply a 64-bit count by a 64-bit tick 'period' in units of seconds >> 64
+ * and add the fractional second part of the reference time.
+ *
+ * The result is a 128-bit value, the top 64 bits of which are seconds, and
+ * the low 64 bits are (seconds >> 64).
+ */
+static uint64_t mul_u64_u64_shr_add_u64(uint64_t *res_hi, uint64_t delta,
+					uint64_t period, uint8_t shift,
+					uint64_t frac_sec)
+{
+	unsigned __int128 res = (unsigned __int128)delta * period;
+
+	res >>= shift;
+	res += frac_sec;
+	*res_hi = res >> 64;
+	return (uint64_t)res;
+}
+
+static bool tai_adjust(struct vmclock_abi *clk, uint64_t *sec)
+{
+	if (likely(clk->time_type == VMCLOCK_TIME_UTC))
+		return true;
+
+	if (clk->time_type == VMCLOCK_TIME_TAI &&
+	    (le64_to_cpu(clk->flags) & VMCLOCK_FLAG_TAI_OFFSET_VALID)) {
+		if (sec)
+			*sec += (int16_t)le16_to_cpu(clk->tai_offset_sec);
+		return true;
+	}
+	return false;
+}
+
+static int vmclock_get_crosststamp(struct vmclock_state *st,
+				   struct ptp_system_timestamp *sts,
+				   struct system_counterval_t *system_counter,
+				   struct timespec64 *tspec)
+{
+	ktime_t deadline = ktime_add(ktime_get(), VMCLOCK_MAX_WAIT);
+	struct system_time_snapshot systime_snapshot;
+	uint64_t cycle, delta, seq, frac_sec;
+
+#ifdef CONFIG_X86
+	/*
+	 * We'd expect the hypervisor to know this and to report the clock
+	 * status as VMCLOCK_STATUS_UNRELIABLE. But be paranoid.
+	 */
+	if (check_tsc_unstable())
+		return -EINVAL;
+#endif
+
+	while (1) {
+		seq = le32_to_cpu(st->clk->seq_count) & ~1ULL;
+
+		/*
+		 * This pairs with a write barrier in the hypervisor
+		 * which populates this structure.
+		 */
+		virt_rmb();
+
+		if (st->clk->clock_status == VMCLOCK_STATUS_UNRELIABLE)
+			return -EINVAL;
+
+		/*
+		 * When invoked for gettimex64(), fill in the pre/post system
+		 * times. The simple case is when system time is based on the
+		 * same counter as st->cs_id, in which case all three times
+		 * will be derived from the *same* counter value.
+		 *
+		 * If the system isn't using the same counter, then the value
+		 * from ktime_get_snapshot() will still be used as pre_ts, and
+		 * ptp_read_system_postts() is called to populate postts after
+		 * calling get_cycles().
+		 *
+		 * The conversion to timespec64 happens further down, outside
+		 * the seq_count loop.
+		 */
+		if (sts) {
+			ktime_get_snapshot(&systime_snapshot);
+			if (systime_snapshot.cs_id == st->cs_id) {
+				cycle = systime_snapshot.cycles;
+			} else {
+				cycle = get_cycles();
+				ptp_read_system_postts(sts);
+			}
+		} else {
+			cycle = get_cycles();
+		}
+
+		delta = cycle - le64_to_cpu(st->clk->counter_value);
+
+		frac_sec = mul_u64_u64_shr_add_u64(&tspec->tv_sec, delta,
+						   le64_to_cpu(st->clk->counter_period_frac_sec),
+						   st->clk->counter_period_shift,
+						   le64_to_cpu(st->clk->time_frac_sec));
+		tspec->tv_nsec = mul_u64_u64_shr(frac_sec, NSEC_PER_SEC, 64);
+		tspec->tv_sec += le64_to_cpu(st->clk->time_sec);
+
+		if (!tai_adjust(st->clk, &tspec->tv_sec))
+			return -EINVAL;
+
+		/*
+		 * This pairs with a write barrier in the hypervisor
+		 * which populates this structure.
+		 */
+		virt_rmb();
+		if (seq == le32_to_cpu(st->clk->seq_count))
+			break;
+
+		if (ktime_after(ktime_get(), deadline))
+			return -ETIMEDOUT;
+	}
+
+	if (system_counter) {
+		system_counter->cycles = cycle;
+		system_counter->cs_id = st->cs_id;
+	}
+
+	if (sts) {
+		sts->pre_ts = ktime_to_timespec64(systime_snapshot.real);
+		if (systime_snapshot.cs_id == st->cs_id)
+			sts->post_ts = sts->pre_ts;
+	}
+
+	return 0;
+}
+
+#ifdef SUPPORT_KVMCLOCK
+/*
+ * In the case where the system is using the KVM clock for timekeeping, convert
+ * the TSC value into a KVM clock time in order to return a paired reading that
+ * get_device_system_crosststamp() can cope with.
+ */
+static int vmclock_get_crosststamp_kvmclock(struct vmclock_state *st,
+					    struct ptp_system_timestamp *sts,
+					    struct system_counterval_t *system_counter,
+					    struct timespec64 *tspec)
+{
+	struct pvclock_vcpu_time_info *pvti = this_cpu_pvti();
+	unsigned int pvti_ver;
+	int ret;
+
+	preempt_disable_notrace();
+
+	do {
+		pvti_ver = pvclock_read_begin(pvti);
+
+		ret = vmclock_get_crosststamp(st, sts, system_counter, tspec);
+		if (ret)
+			break;
+
+		system_counter->cycles = __pvclock_read_cycles(pvti,
+							       system_counter->cycles);
+		system_counter->cs_id = CSID_X86_KVM_CLK;
+
+		/*
+		 * This retry should never really happen; if the TSC is
+		 * stable and reliable enough across vCPUS that it is sane
+		 * for the hypervisor to expose a VMCLOCK device which uses
+		 * it as the reference counter, then the KVM clock sohuld be
+		 * in 'master clock mode' and basically never changed. But
+		 * the KVM clock is a fickle and often broken thing, so do
+		 * it "properly" just in case.
+		 */
+	} while (pvclock_read_retry(pvti, pvti_ver));
+
+	preempt_enable_notrace();
+
+	return ret;
+}
+#endif
+
+static int ptp_vmclock_get_time_fn(ktime_t *device_time,
+				   struct system_counterval_t *system_counter,
+				   void *ctx)
+{
+	struct vmclock_state *st = ctx;
+	struct timespec64 tspec;
+	int ret;
+
+#ifdef SUPPORT_KVMCLOCK
+	if (READ_ONCE(st->sys_cs_id) == CSID_X86_KVM_CLK)
+		ret = vmclock_get_crosststamp_kvmclock(st, NULL, system_counter,
+						       &tspec);
+	else
+#endif
+		ret = vmclock_get_crosststamp(st, NULL, system_counter, &tspec);
+
+	if (!ret)
+		*device_time = timespec64_to_ktime(tspec);
+
+	return ret;
+}
+
+static int ptp_vmclock_getcrosststamp(struct ptp_clock_info *ptp,
+				      struct system_device_crosststamp *xtstamp)
+{
+	struct vmclock_state *st = container_of(ptp, struct vmclock_state,
+						ptp_clock_info);
+	int ret = get_device_system_crosststamp(ptp_vmclock_get_time_fn, st,
+						NULL, xtstamp);
+#ifdef SUPPORT_KVMCLOCK
+	/*
+	 * On x86, the KVM clock may be used for the system time. We can
+	 * actually convert a TSC reading to that, and return a paired
+	 * timestamp that get_device_system_crosststamp() *can* handle.
+	 */
+	if (ret == -ENODEV) {
+		struct system_time_snapshot systime_snapshot;
+
+		ktime_get_snapshot(&systime_snapshot);
+
+		if (systime_snapshot.cs_id == CSID_X86_TSC ||
+		    systime_snapshot.cs_id == CSID_X86_KVM_CLK) {
+			WRITE_ONCE(st->sys_cs_id, systime_snapshot.cs_id);
+			ret = get_device_system_crosststamp(ptp_vmclock_get_time_fn,
+							    st, NULL, xtstamp);
+		}
+	}
+#endif
+	return ret;
+}
+
+/*
+ * PTP clock operations
+ */
+
+static int ptp_vmclock_adjfine(struct ptp_clock_info *ptp, long delta)
+{
+	return -EOPNOTSUPP;
+}
+
+static int ptp_vmclock_adjtime(struct ptp_clock_info *ptp, s64 delta)
+{
+	return -EOPNOTSUPP;
+}
+
+static int ptp_vmclock_settime(struct ptp_clock_info *ptp,
+			   const struct timespec64 *ts)
+{
+	return -EOPNOTSUPP;
+}
+
+static int ptp_vmclock_gettimex(struct ptp_clock_info *ptp, struct timespec64 *ts,
+				struct ptp_system_timestamp *sts)
+{
+	struct vmclock_state *st = container_of(ptp, struct vmclock_state,
+						ptp_clock_info);
+
+	return vmclock_get_crosststamp(st, sts, NULL, ts);
+}
+
+static int ptp_vmclock_enable(struct ptp_clock_info *ptp,
+			  struct ptp_clock_request *rq, int on)
+{
+	return -EOPNOTSUPP;
+}
+
+static const struct ptp_clock_info ptp_vmclock_info = {
+	.owner		= THIS_MODULE,
+	.max_adj	= 0,
+	.n_ext_ts	= 0,
+	.n_pins		= 0,
+	.pps		= 0,
+	.adjfine	= ptp_vmclock_adjfine,
+	.adjtime	= ptp_vmclock_adjtime,
+	.gettimex64	= ptp_vmclock_gettimex,
+	.settime64	= ptp_vmclock_settime,
+	.enable		= ptp_vmclock_enable,
+	.getcrosststamp = ptp_vmclock_getcrosststamp,
+};
+
+static struct ptp_clock *vmclock_ptp_register(struct device *dev,
+					      struct vmclock_state *st)
+{
+	enum clocksource_ids cs_id;
+
+	if (IS_ENABLED(CONFIG_ARM64) &&
+	    st->clk->counter_id == VMCLOCK_COUNTER_ARM_VCNT) {
+		/* Can we check it's the virtual counter? */
+		cs_id = CSID_ARM_ARCH_COUNTER;
+	} else if (IS_ENABLED(CONFIG_X86) &&
+		   st->clk->counter_id == VMCLOCK_COUNTER_X86_TSC) {
+		cs_id = CSID_X86_TSC;
+	} else {
+		return NULL;
+	}
+
+	/* Only UTC, or TAI with offset */
+	if (!tai_adjust(st->clk, NULL)) {
+		dev_info(dev, "vmclock does not provide unambiguous UTC\n");
+		return NULL;
+	}
+
+	st->sys_cs_id = cs_id;
+	st->cs_id = cs_id;
+	st->ptp_clock_info = ptp_vmclock_info;
+	strscpy(st->ptp_clock_info.name, st->name);
+
+	return ptp_clock_register(&st->ptp_clock_info, dev);
+}
+
+static int vmclock_miscdev_mmap(struct file *fp, struct vm_area_struct *vma)
+{
+	struct vmclock_state *st = container_of(fp->private_data,
+						struct vmclock_state, miscdev);
+
+	if ((vma->vm_flags & (VM_READ|VM_WRITE)) != VM_READ)
+		return -EROFS;
+
+	if (vma->vm_end - vma->vm_start != PAGE_SIZE || vma->vm_pgoff)
+		return -EINVAL;
+
+	if (io_remap_pfn_range(vma, vma->vm_start,
+			       st->res.start >> PAGE_SHIFT, PAGE_SIZE,
+			       vma->vm_page_prot))
+		return -EAGAIN;
+
+	return 0;
+}
+
+static ssize_t vmclock_miscdev_read(struct file *fp, char __user *buf,
+				    size_t count, loff_t *ppos)
+{
+	struct vmclock_state *st = container_of(fp->private_data,
+						struct vmclock_state, miscdev);
+	ktime_t deadline = ktime_add(ktime_get(), VMCLOCK_MAX_WAIT);
+	size_t max_count;
+	uint32_t seq;
+
+	if (*ppos >= PAGE_SIZE)
+		return 0;
+
+	max_count = PAGE_SIZE - *ppos;
+	if (count > max_count)
+		count = max_count;
+
+	while (1) {
+		seq = le32_to_cpu(st->clk->seq_count) & ~1U;
+		/* Pairs with hypervisor wmb */
+		virt_rmb();
+
+		if (copy_to_user(buf, ((char *)st->clk) + *ppos, count))
+			return -EFAULT;
+
+		/* Pairs with hypervisor wmb */
+		virt_rmb();
+		if (seq == le32_to_cpu(st->clk->seq_count))
+			break;
+
+		if (ktime_after(ktime_get(), deadline))
+			return -ETIMEDOUT;
+	}
+
+	*ppos += count;
+	return count;
+}
+
+static const struct file_operations vmclock_miscdev_fops = {
+	.mmap = vmclock_miscdev_mmap,
+	.read = vmclock_miscdev_read,
+};
+
+/* module operations */
+
+static void vmclock_remove(struct platform_device *pdev)
+{
+	struct device *dev = &pdev->dev;
+	struct vmclock_state *st = dev_get_drvdata(dev);
+
+	if (st->ptp_clock)
+		ptp_clock_unregister(st->ptp_clock);
+
+	if (st->miscdev.minor != MISC_DYNAMIC_MINOR)
+		misc_deregister(&st->miscdev);
+}
+
+static acpi_status vmclock_acpi_resources(struct acpi_resource *ares, void *data)
+{
+	struct vmclock_state *st = data;
+	struct resource_win win;
+	struct resource *res = &win.res;
+
+	if (ares->type == ACPI_RESOURCE_TYPE_END_TAG)
+		return AE_OK;
+
+	/* There can be only one */
+	if (resource_type(&st->res) == IORESOURCE_MEM)
+		return AE_ERROR;
+
+	if (acpi_dev_resource_memory(ares, res) ||
+	    acpi_dev_resource_address_space(ares, &win)) {
+
+		if (resource_type(res) != IORESOURCE_MEM ||
+		    resource_size(res) < sizeof(st->clk))
+			return AE_ERROR;
+
+		st->res = *res;
+		return AE_OK;
+	}
+
+	return AE_ERROR;
+}
+
+static int vmclock_probe_acpi(struct device *dev, struct vmclock_state *st)
+{
+	struct acpi_device *adev = ACPI_COMPANION(dev);
+	acpi_status status;
+
+	/*
+	 * This should never happen as this function is only called when
+	 * has_acpi_companion(dev) is true, but the logic is sufficiently
+	 * complex that Coverity can't see the tautology.
+	 */
+	if (!adev)
+		return -ENODEV;
+
+	status = acpi_walk_resources(adev->handle, METHOD_NAME__CRS,
+				     vmclock_acpi_resources, st);
+	if (ACPI_FAILURE(status) || resource_type(&st->res) != IORESOURCE_MEM) {
+		dev_err(dev, "failed to get resources\n");
+		return -ENODEV;
+	}
+
+	return 0;
+}
+
+static void vmclock_put_idx(void *data)
+{
+	struct vmclock_state *st = data;
+
+	ida_free(&vmclock_ida, st->index);
+}
+
+static int vmclock_probe(struct platform_device *pdev)
+{
+	struct device *dev = &pdev->dev;
+	struct vmclock_state *st;
+	int ret;
+
+	st = devm_kzalloc(dev, sizeof(*st), GFP_KERNEL);
+	if (!st)
+		return -ENOMEM;
+
+	if (has_acpi_companion(dev))
+		ret = vmclock_probe_acpi(dev, st);
+	else
+		ret = -EINVAL; /* Only ACPI for now */
+
+	if (ret) {
+		dev_info(dev, "Failed to obtain physical address: %d\n", ret);
+		goto out;
+	}
+
+	if (resource_size(&st->res) < VMCLOCK_MIN_SIZE) {
+		dev_info(dev, "Region too small (0x%llx)\n",
+			 resource_size(&st->res));
+		ret = -EINVAL;
+		goto out;
+	}
+	st->clk = devm_memremap(dev, st->res.start, resource_size(&st->res),
+				MEMREMAP_WB | MEMREMAP_DEC);
+	if (IS_ERR(st->clk)) {
+		ret = PTR_ERR(st->clk);
+		dev_info(dev, "failed to map shared memory\n");
+		st->clk = NULL;
+		goto out;
+	}
+
+	if (le32_to_cpu(st->clk->magic) != VMCLOCK_MAGIC ||
+	    le32_to_cpu(st->clk->size) > resource_size(&st->res) ||
+	    le16_to_cpu(st->clk->version) != 1) {
+		dev_info(dev, "vmclock magic fields invalid\n");
+		ret = -EINVAL;
+		goto out;
+	}
+
+	ret = ida_alloc(&vmclock_ida, GFP_KERNEL);
+	if (ret < 0)
+		goto out;
+
+	st->index = ret;
+	ret = devm_add_action_or_reset(&pdev->dev, vmclock_put_idx, st);
+	if (ret)
+		goto out;
+
+	st->name = devm_kasprintf(&pdev->dev, GFP_KERNEL, "vmclock%d", st->index);
+	if (!st->name) {
+		ret = -ENOMEM;
+		goto out;
+	}
+
+	/*
+	 * If the structure is big enough, it can be mapped to userspace.
+	 * Theoretically a guest OS even using larger pages could still
+	 * use 4KiB PTEs to map smaller MMIO regions like this, but let's
+	 * cross that bridge if/when we come to it.
+	 */
+	if (le32_to_cpu(st->clk->size) >= PAGE_SIZE) {
+		st->miscdev.minor = MISC_DYNAMIC_MINOR;
+		st->miscdev.fops = &vmclock_miscdev_fops;
+		st->miscdev.name = st->name;
+
+		ret = misc_register(&st->miscdev);
+		if (ret)
+			goto out;
+	}
+
+	/* If there is valid clock information, register a PTP clock */
+	if (VMCLOCK_FIELD_PRESENT(st->clk, time_frac_sec)) {
+		/* Can return a silent NULL, or an error. */
+		st->ptp_clock = vmclock_ptp_register(dev, st);
+		if (IS_ERR(st->ptp_clock)) {
+			ret = PTR_ERR(st->ptp_clock);
+			st->ptp_clock = NULL;
+			vmclock_remove(pdev);
+			goto out;
+		}
+	}
+
+	if (!st->miscdev.minor && !st->ptp_clock) {
+		/* Neither miscdev nor PTP registered */
+		dev_info(dev, "vmclock: Neither miscdev nor PTP available; not registering\n");
+		ret = -ENODEV;
+		goto out;
+	}
+
+	dev_info(dev, "%s: registered %s%s%s\n", st->name,
+		 st->miscdev.minor ? "miscdev" : "",
+		 (st->miscdev.minor && st->ptp_clock) ? ", " : "",
+		 st->ptp_clock ? "PTP" : "");
+
+	dev_set_drvdata(dev, st);
+
+ out:
+	return ret;
+}
+
+static const struct acpi_device_id vmclock_acpi_ids[] = {
+	{ "AMZNC10C", 0 },
+	{}
+};
+MODULE_DEVICE_TABLE(acpi, vmclock_acpi_ids);
+
+static struct platform_driver vmclock_platform_driver = {
+	.probe		= vmclock_probe,
+	.remove_new	= vmclock_remove,
+	.driver	= {
+		.name	= "vmclock",
+		.acpi_match_table = vmclock_acpi_ids,
+	},
+};
+
+module_platform_driver(vmclock_platform_driver)
+
+MODULE_AUTHOR("David Woodhouse <[e-mail]>");
+MODULE_DESCRIPTION("PTP clock using VMCLOCK");
+MODULE_LICENSE("GPL");

include/uapi/linux/vmclock-abi.h

···
+/* SPDX-License-Identifier: ((GPL-2.0 WITH Linux-syscall-note) OR BSD-2-Clause) */
+
+/*
+ * This structure provides a vDSO-style clock to VM guests, exposing the
+ * relationship (or lack thereof) between the CPU clock (TSC, timebase, arch
+ * counter, etc.) and real time. It is designed to address the problem of
+ * live migration, which other clock enlightenments do not.
+ *
+ * When a guest is live migrated, this affects the clock in two ways.
+ *
+ * First, even between identical hosts the actual frequency of the underlying
+ * counter will change within the tolerances of its specification (typically
+ * ±50PPM, or 4 seconds a day). This frequency also varies over time on the
+ * same host, but can be tracked by NTP as it generally varies slowly. With
+ * live migration there is a step change in the frequency, with no warning.
+ *
+ * Second, there may be a step change in the value of the counter itself, as
+ * its accuracy is limited by the precision of the NTP synchronization on the
+ * source and destination hosts.
+ *
+ * So any calibration (NTP, PTP, etc.) which the guest has done on the source
+ * host before migration is invalid, and needs to be redone on the new host.
+ *
+ * In its most basic mode, this structure provides only an indication to the
+ * guest that live migration has occurred. This allows the guest to know that
+ * its clock is invalid and take remedial action. For applications that need
+ * reliable accurate timestamps (e.g. distributed databases), the structure
+ * can be mapped all the way to userspace. This allows the application to see
+ * directly for itself that the clock is disrupted and take appropriate
+ * action, even when using a vDSO-style method to get the time instead of a
+ * system call.
+ *
+ * In its more advanced mode. this structure can also be used to expose the
+ * precise relationship of the CPU counter to real time, as calibrated by the
+ * host. This means that userspace applications can have accurate time
+ * immediately after live migration, rather than having to pause operations
+ * and wait for NTP to recover. This mode does, of course, rely on the
+ * counter being reliable and consistent across CPUs.
+ *
+ * Note that this must be true UTC, never with smeared leap seconds. If a
+ * guest wishes to construct a smeared clock, it can do so. Presenting a
+ * smeared clock through this interface would be problematic because it
+ * actually messes with the apparent counter *period*. A linear smearing
+ * of 1 ms per second would effectively tweak the counter period by 1000PPM
+ * at the start/end of the smearing period, while a sinusoidal smear would
+ * basically be impossible to represent.
+ *
+ * This structure is offered with the intent that it be adopted into the
+ * nascent virtio-rtc standard, as a virtio-rtc that does not address the live
+ * migration problem seems a little less than fit for purpose. For that
+ * reason, certain fields use precisely the same numeric definitions as in
+ * the virtio-rtc proposal. The structure can also be exposed through an ACPI
+ * device with the CID "VMCLOCK", modelled on the "VMGENID" device except for
+ * the fact that it uses a real _CRS to convey the address of the structure
+ * (which should be a full page, to allow for mapping directly to userspace).
+ */
+
+#ifndef __VMCLOCK_ABI_H__
+#define __VMCLOCK_ABI_H__
+
+#include <linux/types.h>
+
+struct vmclock_abi {
+	/* CONSTANT FIELDS */
+	__le32 magic;
+#define VMCLOCK_MAGIC	0x4b4c4356 /* "VCLK" */
+	__le32 size;		/* Size of region containing this structure */
+	__le16 version;	/* 1 */
+	__u8 counter_id; /* Matches VIRTIO_RTC_COUNTER_xxx except INVALID */
+#define VMCLOCK_COUNTER_ARM_VCNT	0
+#define VMCLOCK_COUNTER_X86_TSC		1
+#define VMCLOCK_COUNTER_INVALID		0xff
+	__u8 time_type; /* Matches VIRTIO_RTC_TYPE_xxx */
+#define VMCLOCK_TIME_UTC			0	/* Since 1970-01-01 00:00:00z */
+#define VMCLOCK_TIME_TAI			1	/* Since 1970-01-01 00:00:00z */
+#define VMCLOCK_TIME_MONOTONIC			2	/* Since undefined epoch */
+#define VMCLOCK_TIME_INVALID_SMEARED		3	/* Not supported */
+#define VMCLOCK_TIME_INVALID_MAYBE_SMEARED	4	/* Not supported */
+
+	/* NON-CONSTANT FIELDS PROTECTED BY SEQCOUNT LOCK */
+	__le32 seq_count;	/* Low bit means an update is in progress */
+	/*
+	 * This field changes to another non-repeating value when the CPU
+	 * counter is disrupted, for example on live migration. This lets
+	 * the guest know that it should discard any calibration it has
+	 * performed of the counter against external sources (NTP/PTP/etc.).
+	 */
+	__le64 disruption_marker;
+	__le64 flags;
+	/* Indicates that the tai_offset_sec field is valid */
+#define VMCLOCK_FLAG_TAI_OFFSET_VALID		(1 << 0)
+	/*
+	 * Optionally used to notify guests of pending maintenance events.
+	 * A guest which provides latency-sensitive services may wish to
+	 * remove itself from service if an event is coming up. Two flags
+	 * indicate the approximate imminence of the event.
+	 */
+#define VMCLOCK_FLAG_DISRUPTION_SOON		(1 << 1) /* About a day */
+#define VMCLOCK_FLAG_DISRUPTION_IMMINENT	(1 << 2) /* About an hour */
+#define VMCLOCK_FLAG_PERIOD_ESTERROR_VALID	(1 << 3)
+#define VMCLOCK_FLAG_PERIOD_MAXERROR_VALID	(1 << 4)
+#define VMCLOCK_FLAG_TIME_ESTERROR_VALID	(1 << 5)
+#define VMCLOCK_FLAG_TIME_MAXERROR_VALID	(1 << 6)
+	/*
+	 * If the MONOTONIC flag is set then (other than leap seconds) it is
+	 * guaranteed that the time calculated according this structure at
+	 * any given moment shall never appear to be later than the time
+	 * calculated via the structure at any *later* moment.
+	 *
+	 * In particular, a timestamp based on a counter reading taken
+	 * immediately after setting the low bit of seq_count (and the
+	 * associated memory barrier), using the previously-valid time and
+	 * period fields, shall never be later than a timestamp based on
+	 * a counter reading taken immediately before *clearing* the low
+	 * bit again after the update, using the about-to-be-valid fields.
+	 */
+#define VMCLOCK_FLAG_TIME_MONOTONIC		(1 << 7)
+
+	__u8 pad[2];
+	__u8 clock_status;
+#define VMCLOCK_STATUS_UNKNOWN		0
+#define VMCLOCK_STATUS_INITIALIZING	1
+#define VMCLOCK_STATUS_SYNCHRONIZED	2
+#define VMCLOCK_STATUS_FREERUNNING	3
+#define VMCLOCK_STATUS_UNRELIABLE	4
+
+	/*
+	 * The time exposed through this device is never smeared. This field
+	 * corresponds to the 'subtype' field in virtio-rtc, which indicates
+	 * the smearing method. However in this case it provides a *hint* to
+	 * the guest operating system, such that *if* the guest OS wants to
+	 * provide its users with an alternative clock which does not follow
+	 * UTC, it may do so in a fashion consistent with the other systems
+	 * in the nearby environment.
+	 */
+	__u8 leap_second_smearing_hint; /* Matches VIRTIO_RTC_SUBTYPE_xxx */
+#define VMCLOCK_SMEARING_STRICT		0
+#define VMCLOCK_SMEARING_NOON_LINEAR	1
+#define VMCLOCK_SMEARING_UTC_SLS	2
+	__le16 tai_offset_sec; /* Actually two's complement signed */
+	__u8 leap_indicator;
+	/*
+	 * This field is based on the VIRTIO_RTC_LEAP_xxx values as defined
+	 * in the current draft of virtio-rtc, but since smearing cannot be
+	 * used with the shared memory device, some values are not used.
+	 *
+	 * The _POST_POS and _POST_NEG values allow the guest to perform
+	 * its own smearing during the day or so after a leap second when
+	 * such smearing may need to continue being applied for a leap
+	 * second which is now theoretically "historical".
+	 */
+#define VMCLOCK_LEAP_NONE	0x00	/* No known nearby leap second */
+#define VMCLOCK_LEAP_PRE_POS	0x01	/* Positive leap second at EOM */
+#define VMCLOCK_LEAP_PRE_NEG	0x02	/* Negative leap second at EOM */
+#define VMCLOCK_LEAP_POS	0x03	/* Set during 23:59:60 second */
+#define VMCLOCK_LEAP_POST_POS	0x04
+#define VMCLOCK_LEAP_POST_NEG	0x05
+
+	/* Bit shift for counter_period_frac_sec and its error rate */
+	__u8 counter_period_shift;
+	/*
+	 * Paired values of counter and UTC at a given point in time.
+	 */
+	__le64 counter_value;
+	/*
+	 * Counter period, and error margin of same. The unit of these
+	 * fields is 1/2^(64 + counter_period_shift) of a second.
+	 */
+	__le64 counter_period_frac_sec;
+	__le64 counter_period_esterror_rate_frac_sec;
+	__le64 counter_period_maxerror_rate_frac_sec;
+
+	/*
+	 * Time according to time_type field above.
+	 */
+	__le64 time_sec;		/* Seconds since time_type epoch */
+	__le64 time_frac_sec;		/* Units of 1/2^64 of a second */
+	__le64 time_esterror_nanosec;
+	__le64 time_maxerror_nanosec;
+};
+
+#endif /*  __VMCLOCK_ABI_H__ */