Skip site navigation (1)Skip section navigation (2)
Date:      Thu, 29 Aug 2013 23:11:58 +0000 (UTC)
From:      "Justin T. Gibbs" <gibbs@FreeBSD.org>
To:        src-committers@freebsd.org, svn-src-all@freebsd.org, svn-src-head@freebsd.org
Subject:   svn commit: r255046 - in head/sys: conf dev/xen/timer i386/include/xen i386/xen
Message-ID:  <201308292311.r7TNBwlC067869@svn.freebsd.org>

next in thread | raw e-mail | index | archive | help
Author: gibbs
Date: Thu Aug 29 23:11:58 2013
New Revision: 255046
URL: http://svnweb.freebsd.org/changeset/base/255046

Log:
  Introduce a new, HVM compatible, paravirtualized timer driver for Xen.
  Use this new driver for both PV and HVM instances.
  
  This driver requires a Xen hypervisor that supports vector callbacks,
  VCPUOP hypercalls, and reports that it has a "safe PV clock".
  
  New timer driver:
  Submitted by: will
  Sponsored by: Spectra Logic Corporation
  
  PV port to new driver, and bug fixes:
  Submitted by: Roger Pau Monné
  Sponsored by: Citrix Systems R&D
  
  sys/dev/xen/timer/timer.c:
  	- Register a PV timer device driver which (currently)
  	  implements device_{identify,probe,attach} and stubs
  	  device_detach.  The detach routine requires functionality
  	  not provided by timecounters(4).  The suspend and resume
  	  routines need additional work (due to Xen requiring that
  	  the hypercalls be executed on the target VCPU), and aren't
  	  needed for our purposes.
  
  	- Make sure there can only be one device instance of this
  	  driver, and that it only registers one eventtimers(4) and
  	  one timecounters(4) device interface.  Make both interfaces
  	  use PCPU data as needed.
  
  	- Match, with a few style cleanups & API differences, the
  	  Xen versions of the "fetch time" functions.
  
  	- Document the magic scale_delta() better for the i386 version.
  
  	- When registering the event timer, bind a separate event
  	  channel for the timer VIRQ to the device's event timer
  	  interrupt handler for each active VCPU.  Describe each
  	  interrupt as "xen_et:c%d", so they can be identified per
  	  CPU in "vmstat -i" or "show intrcnt" in KDB.
  
  	- When scheduling a timer into the hypervisor, try up to
  	  60 times if the hypervisor rejects the time as being in
  	  the past.  In the common case, this retry shouldn't happen,
  	  and if it does, it should only happen once.  This is
  	  because the event timer advertises a minimum period of
  	  100usec, which is only less than the usual hypercall round
  	  trip time about 1 out of every 100 tries.  (Unlike other
  	  similar drivers, this one actually checks whether the
  	  hypervisor accepted the singleshot timer set hypercall.)
  
  	- Implement a RTC PV clock based on the hypervisor wallclock.
  
  sys/conf/files:
  	- Add dev/xen/timer/timer.c if the kernel configuration
  	  includes either the XEN or XENHVM options.
  
  sys/conf/files.i386:
  sys/i386/include/xen/xen_clock_util.h:
  sys/i386/xen/clock.c:
  sys/i386/xen/xen_clock_util.c:
  sys/i386/xen/mp_machdep.c:
  sys/i386/xen/xen_rtc.c:
  	- Remove previous PV timer used in i386 XEN PV kernels, the
  	  new timer introduced in this change is used instead (so
  	  we share the same code between PVHVM and PV).
  
  MFC after: 2 weeks

Added:
  head/sys/dev/xen/timer/
  head/sys/dev/xen/timer/timer.c   (contents, props changed)
Deleted:
  head/sys/i386/include/xen/xen_clock_util.h
  head/sys/i386/xen/xen_clock_util.c
  head/sys/i386/xen/xen_rtc.c
Modified:
  head/sys/conf/files
  head/sys/conf/files.i386
  head/sys/i386/xen/clock.c
  head/sys/i386/xen/mp_machdep.c

Modified: head/sys/conf/files
==============================================================================
--- head/sys/conf/files	Thu Aug 29 23:09:34 2013	(r255045)
+++ head/sys/conf/files	Thu Aug 29 23:11:58 2013	(r255046)
@@ -2499,6 +2499,7 @@ dev/xen/control/control.c	optional xen |
 dev/xen/netback/netback.c	optional xen | xenhvm
 dev/xen/netfront/netfront.c	optional xen | xenhvm
 dev/xen/xenpci/xenpci.c		optional xenpci
+dev/xen/timer/timer.c		optional xen | xenhvm
 dev/xl/if_xl.c			optional xl pci
 dev/xl/xlphy.c			optional xl pci
 fs/deadfs/dead_vnops.c		standard

Modified: head/sys/conf/files.i386
==============================================================================
--- head/sys/conf/files.i386	Thu Aug 29 23:09:34 2013	(r255045)
+++ head/sys/conf/files.i386	Thu Aug 29 23:11:58 2013	(r255046)
@@ -483,8 +483,6 @@ i386/ibcs2/ibcs2_xenix.c	optional ibcs2
 i386/ibcs2/ibcs2_xenix_sysent.c	optional ibcs2
 i386/ibcs2/imgact_coff.c	optional ibcs2
 i386/xen/clock.c		optional xen
-i386/xen/xen_clock_util.c	optional xen
-i386/xen/xen_rtc.c		optional xen
 i386/isa/elink.c		optional ep | ie
 i386/isa/npx.c			optional npx
 i386/isa/pmtimer.c		optional pmtimer

Added: head/sys/dev/xen/timer/timer.c
==============================================================================
--- /dev/null	00:00:00 1970	(empty, because file is newly added)
+++ head/sys/dev/xen/timer/timer.c	Thu Aug 29 23:11:58 2013	(r255046)
@@ -0,0 +1,608 @@
+/**
+ * Copyright (c) 2009 Adrian Chadd
+ * Copyright (c) 2012 Spectra Logic Corporation
+ * All rights reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
+ * are met:
+ * 1. Redistributions of source code must retain the above copyright
+ *    notice, this list of conditions and the following disclaimer.
+ * 2. Redistributions in binary form must reproduce the above copyright
+ *    notice, this list of conditions and the following disclaimer in the
+ *    documentation and/or other materials provided with the distribution.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
+ * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
+ * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+ * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
+ * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
+ * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
+ * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
+ * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
+ * SUCH DAMAGE.
+ *
+ */
+
+/**
+ * \file dev/xen/timer/timer.c
+ * \brief A timer driver for the Xen hypervisor's PV clock.
+ */
+
+#include <sys/cdefs.h>
+__FBSDID("$FreeBSD$");
+
+#include <sys/param.h>
+#include <sys/systm.h>
+#include <sys/bus.h>
+#include <sys/kernel.h>
+#include <sys/module.h>
+#include <sys/time.h>
+#include <sys/timetc.h>
+#include <sys/timeet.h>
+#include <sys/smp.h>
+#include <sys/limits.h>
+#include <sys/clock.h>
+
+#include <xen/xen-os.h>
+#include <xen/features.h>
+#include <xen/xen_intr.h>
+#include <xen/hypervisor.h>
+#include <xen/interface/io/xenbus.h>
+#include <xen/interface/vcpu.h>
+
+#include <machine/cpu.h>
+#include <machine/cpufunc.h>
+#include <machine/clock.h>
+#include <machine/_inttypes.h>
+
+#include "clock_if.h"
+
+static devclass_t xentimer_devclass;
+
+#define	NSEC_IN_SEC	1000000000ULL
+#define	NSEC_IN_USEC	1000ULL
+/* 18446744073 = int(2^64 / NSEC_IN_SC) = 1 ns in 64-bit fractions */
+#define	FRAC_IN_NSEC	18446744073LL
+
+/* Xen timers may fire up to 100us off */
+#define	XENTIMER_MIN_PERIOD_IN_NSEC	100*NSEC_IN_USEC
+#define	XENCLOCK_RESOLUTION		10000000
+
+#define	ETIME	62	/* Xen "bad time" error */
+
+#define	XENTIMER_QUALITY	950
+
+struct xentimer_pcpu_data {
+	uint64_t timer;
+	uint64_t last_processed;
+	void *irq_handle;
+};
+
+DPCPU_DEFINE(struct xentimer_pcpu_data, xentimer_pcpu);
+
+DPCPU_DECLARE(struct vcpu_info *, vcpu_info);
+
+struct xentimer_softc {
+	device_t dev;
+	struct timecounter tc;
+	struct eventtimer et;
+};
+
+/* Last time; this guarantees a monotonically increasing clock. */
+volatile uint64_t xen_timer_last_time = 0;
+
+static void
+xentimer_identify(driver_t *driver, device_t parent)
+{
+	if (!xen_domain())
+		return;
+
+	/* Handle all Xen PV timers in one device instance. */
+	if (devclass_get_device(xentimer_devclass, 0))
+		return;
+
+	BUS_ADD_CHILD(parent, 0, "xen_et", 0);
+}
+
+static int
+xentimer_probe(device_t dev)
+{
+	KASSERT((xen_domain()), ("Trying to use Xen timer on bare metal"));
+	/*
+	 * In order to attach, this driver requires the following:
+	 * - Vector callback support by the hypervisor, in order to deliver
+	 *   timer interrupts to the correct CPU for CPUs other than 0.
+	 * - Access to the hypervisor shared info page, in order to look up
+	 *   each VCPU's timer information and the Xen wallclock time.
+	 * - The hypervisor must say its PV clock is "safe" to use.
+	 * - The hypervisor must support VCPUOP hypercalls.
+	 * - The maximum number of CPUs supported by FreeBSD must not exceed
+	 *   the number of VCPUs supported by the hypervisor.
+	 */
+#define	XTREQUIRES(condition, reason...)	\
+	if (!(condition)) {			\
+		device_printf(dev, ## reason);	\
+		device_detach(dev);		\
+		return (ENXIO);			\
+	}
+
+	if (xen_hvm_domain()) {
+		XTREQUIRES(xen_vector_callback_enabled,
+		           "vector callbacks unavailable\n");
+		XTREQUIRES(xen_feature(XENFEAT_hvm_safe_pvclock),
+		           "HVM safe pvclock unavailable\n");
+	}
+	XTREQUIRES(HYPERVISOR_shared_info != NULL,
+	           "shared info page unavailable\n");
+	XTREQUIRES(HYPERVISOR_vcpu_op(VCPUOP_stop_periodic_timer, 0, NULL) == 0,
+	           "VCPUOPs interface unavailable\n");
+#undef XTREQUIRES
+	device_set_desc(dev, "Xen PV Clock");
+	return (0);
+}
+
+/*
+ * Scale a 64-bit delta by scaling and multiplying by a 32-bit fraction,
+ * yielding a 64-bit result.
+ */
+static inline uint64_t
+scale_delta(uint64_t delta, uint32_t mul_frac, int shift)
+{
+	uint64_t product;
+
+	if (shift < 0)
+		delta >>= -shift;
+	else
+		delta <<= shift;
+
+#if defined(__i386__)
+	{
+		uint32_t tmp1, tmp2;
+
+		/**
+		 * For i386, the formula looks like:
+		 *
+		 *   lower = (mul_frac * (delta & UINT_MAX)) >> 32
+		 *   upper = mul_frac * (delta >> 32)
+		 *   product = lower + upper
+		 */
+		__asm__ (
+			"mul  %5       ; "
+			"mov  %4,%%eax ; "
+			"mov  %%edx,%4 ; "
+			"mul  %5       ; "
+			"xor  %5,%5    ; "
+			"add  %4,%%eax ; "
+			"adc  %5,%%edx ; "
+			: "=A" (product), "=r" (tmp1), "=r" (tmp2)
+			: "a" ((uint32_t)delta), "1" ((uint32_t)(delta >> 32)),
+			  "2" (mul_frac) );
+	}
+#elif defined(__amd64__)
+	{
+		unsigned long tmp;
+
+		__asm__ (
+			"mulq %[mul_frac] ; shrd $32, %[hi], %[lo]"
+			: [lo]"=a" (product), [hi]"=d" (tmp)
+			: "0" (delta), [mul_frac]"rm"((uint64_t)mul_frac));
+	}
+#else
+#error "xentimer: unsupported architecture"
+#endif
+
+	return (product);
+}
+
+static uint64_t
+get_nsec_offset(struct vcpu_time_info *tinfo)
+{
+
+	return (scale_delta(rdtsc() - tinfo->tsc_timestamp,
+	    tinfo->tsc_to_system_mul, tinfo->tsc_shift));
+}
+
+/*
+ * Read the current hypervisor system uptime value from Xen.
+ * See <xen/interface/xen.h> for a description of how this works.
+ */
+static uint32_t
+xen_fetch_vcpu_tinfo(struct vcpu_time_info *dst, struct vcpu_time_info *src)
+{
+
+	do {
+		dst->version = src->version;
+		rmb();
+		dst->tsc_timestamp = src->tsc_timestamp;
+		dst->system_time = src->system_time;
+		dst->tsc_to_system_mul = src->tsc_to_system_mul;
+		dst->tsc_shift = src->tsc_shift;
+		rmb();
+	} while ((src->version & 1) | (dst->version ^ src->version));
+
+	return (dst->version);
+}
+
+/**
+ * \brief Get the current time, in nanoseconds, since the hypervisor booted.
+ *
+ * \note This function returns the current CPU's idea of this value, unless
+ *       it happens to be less than another CPU's previously determined value.
+ */
+static uint64_t
+xen_fetch_vcpu_time(void)
+{
+	struct vcpu_time_info dst;
+	struct vcpu_time_info *src;
+	uint32_t pre_version;
+	uint64_t now;
+	volatile uint64_t last;
+	struct vcpu_info *vcpu = DPCPU_GET(vcpu_info);
+
+	src = &vcpu->time;
+
+	critical_enter();
+	do {
+		pre_version = xen_fetch_vcpu_tinfo(&dst, src);
+		barrier();
+		now = dst.system_time + get_nsec_offset(&dst);
+		barrier();
+	} while (pre_version != src->version);
+
+	/*
+	 * Enforce a monotonically increasing clock time across all
+	 * VCPUs.  If our time is too old, use the last time and return.
+	 * Otherwise, try to update the last time.
+	 */
+	do {
+		last = xen_timer_last_time;
+		if (last > now) {
+			now = last;
+			break;
+		}
+	} while (!atomic_cmpset_64(&xen_timer_last_time, last, now));
+
+	critical_exit();
+
+	return (now);
+}
+
+static uint32_t
+xentimer_get_timecount(struct timecounter *tc)
+{
+
+	return ((uint32_t)xen_fetch_vcpu_time() & UINT_MAX);
+}
+
+/**
+ * \brief Fetch the hypervisor boot time, known as the "Xen wallclock".
+ *
+ * \param ts		Timespec to store the current stable value.
+ * \param version	Pointer to store the corresponding wallclock version.
+ *
+ * \note This value is updated when Domain-0 shifts its clock to follow
+ *       clock drift, e.g. as detected by NTP.
+ */
+static void
+xen_fetch_wallclock(struct timespec *ts)
+{
+	shared_info_t *src = HYPERVISOR_shared_info;
+	uint32_t version = 0;
+
+	do {
+		version = src->wc_version;
+		rmb();
+		ts->tv_sec = src->wc_sec;
+		ts->tv_nsec = src->wc_nsec;
+		rmb();
+	} while ((src->wc_version & 1) | (version ^ src->wc_version));
+}
+
+static void
+xen_fetch_uptime(struct timespec *ts)
+{
+	uint64_t uptime = xen_fetch_vcpu_time();
+	ts->tv_sec = uptime / NSEC_IN_SEC;
+	ts->tv_nsec = uptime % NSEC_IN_SEC;
+}
+
+static int
+xentimer_settime(device_t dev __unused, struct timespec *ts)
+{
+	/*
+	 * Don't return EINVAL here; just silently fail if the domain isn't
+	 * privileged enough to set the TOD.
+	 */
+	return(0);
+}
+
+/**
+ * \brief Return current time according to the Xen Hypervisor wallclock.
+ *
+ * \param dev	Xentimer device.
+ * \param ts	Pointer to store the wallclock time.
+ *
+ * \note  The Xen time structures document the hypervisor start time and the
+ *        uptime-since-hypervisor-start (in nsec.) They need to be combined
+ *        in order to calculate a TOD clock.
+ */
+static int
+xentimer_gettime(device_t dev, struct timespec *ts)
+{
+	struct timespec u_ts;
+
+	timespecclear(ts);
+	xen_fetch_wallclock(ts);
+	xen_fetch_uptime(&u_ts);
+	timespecadd(ts, &u_ts);
+
+	return(0);
+}
+
+/**
+ * \brief Handle a timer interrupt for the Xen PV timer driver.
+ *
+ * \param arg	Xen timer driver softc that is expecting the interrupt.
+ */
+static int
+xentimer_intr(void *arg)
+{
+	struct xentimer_softc *sc = (struct xentimer_softc *)arg;
+	struct xentimer_pcpu_data *pcpu = DPCPU_PTR(xentimer_pcpu);
+
+	pcpu->last_processed = xen_fetch_vcpu_time();
+	if (pcpu->timer != 0 && sc->et.et_active)
+		sc->et.et_event_cb(&sc->et, sc->et.et_arg);
+
+	return (FILTER_HANDLED);
+}
+
+static int
+xentimer_vcpu_start_timer(int vcpu, uint64_t next_time)
+{
+	struct vcpu_set_singleshot_timer single;
+
+	single.timeout_abs_ns = next_time;
+	single.flags          = VCPU_SSHOTTMR_future;
+	return (HYPERVISOR_vcpu_op(VCPUOP_set_singleshot_timer, vcpu, &single));
+}
+
+static int
+xentimer_vcpu_stop_timer(int vcpu)
+{
+
+	return (HYPERVISOR_vcpu_op(VCPUOP_stop_singleshot_timer, vcpu, NULL));
+}
+
+/**
+ * \brief Set the next oneshot time for the current CPU.
+ *
+ * \param et	Xen timer driver event timer to schedule on.
+ * \param first	Delta to the next time to schedule the interrupt for.
+ * \param period Not used.
+ *
+ * \note See eventtimers(9) for more information.
+ * \note 
+ *
+ * \returns 0
+ */
+static int
+xentimer_et_start(struct eventtimer *et,
+    sbintime_t first, sbintime_t period)
+{
+	int error = 0, i = 0;
+	struct xentimer_softc *sc = et->et_priv;
+	int cpu = PCPU_GET(acpi_id);
+	struct xentimer_pcpu_data *pcpu = DPCPU_PTR(xentimer_pcpu);
+	uint64_t first_in_ns, next_time;
+
+	/* See sbttots() for this formula. */
+	first_in_ns = (((first >> 32) * NSEC_IN_SEC) +
+	               (((uint64_t)NSEC_IN_SEC * (uint32_t)first) >> 32));
+
+	/*
+	 * Retry any timer scheduling failures, where the hypervisor
+	 * returns -ETIME.  Sometimes even a 100us timer period isn't large
+	 * enough, but larger period instances are relatively uncommon.
+	 *
+	 * XXX Remove the panics once et_start() and its consumers are
+	 *     equipped to deal with start failures.
+	 */
+	do {
+		if (++i == 60)
+			panic("can't schedule timer");
+		next_time = xen_fetch_vcpu_time() + first_in_ns;
+		error = xentimer_vcpu_start_timer(cpu, next_time);
+	} while (error == -ETIME);
+
+	if (error)
+		panic("%s: Error %d setting singleshot timer to %"PRIu64"\n",
+		    device_get_nameunit(sc->dev), error, next_time);
+
+	pcpu->timer = next_time;
+	return (error);
+}
+
+/**
+ * \brief Cancel the event timer's currently running timer, if any.
+ */
+static int
+xentimer_et_stop(struct eventtimer *et)
+{
+	int cpu = PCPU_GET(acpi_id);
+	struct xentimer_pcpu_data *pcpu = DPCPU_PTR(xentimer_pcpu);
+
+	pcpu->timer = 0;
+	return (xentimer_vcpu_stop_timer(cpu));
+}
+
+/**
+ * \brief Attach a Xen PV timer driver instance.
+ * 
+ * \param dev	Bus device object to attach.
+ *
+ * \note
+ * \returns EINVAL 
+ */
+static int
+xentimer_attach(device_t dev)
+{
+	struct xentimer_softc *sc = device_get_softc(dev);
+	int error, i;
+
+	sc->dev = dev;
+
+	/* Bind an event channel to a VIRQ on each VCPU. */
+	CPU_FOREACH(i) {
+		struct xentimer_pcpu_data *pcpu = DPCPU_ID_PTR(i, xentimer_pcpu);
+
+		error = HYPERVISOR_vcpu_op(VCPUOP_stop_periodic_timer, i, NULL);
+		if (error) {
+			device_printf(dev, "Error disabling Xen periodic timer "
+			                   "on CPU %d\n", i);
+			return (error);
+		}
+
+		error = xen_intr_bind_virq(dev, VIRQ_TIMER, i, xentimer_intr,
+		    NULL, sc, INTR_TYPE_CLK, &pcpu->irq_handle);
+		if (error) {
+			device_printf(dev, "Error %d binding VIRQ_TIMER "
+			    "to VCPU %d\n", error, i);
+			return (error);
+		}
+		xen_intr_describe(pcpu->irq_handle, "c%d", i);
+	}
+
+	/* Register the event timer. */
+	sc->et.et_name = "XENTIMER";
+	sc->et.et_quality = XENTIMER_QUALITY;
+	sc->et.et_flags = ET_FLAGS_ONESHOT | ET_FLAGS_PERCPU;
+	sc->et.et_frequency = NSEC_IN_SEC;
+	/* See tstosbt() for this formula */
+	sc->et.et_min_period = (XENTIMER_MIN_PERIOD_IN_NSEC *
+	                        (((uint64_t)1 << 63) / 500000000) >> 32);
+	sc->et.et_max_period = ((sbintime_t)4 << 32);
+	sc->et.et_start = xentimer_et_start;
+	sc->et.et_stop = xentimer_et_stop;
+	sc->et.et_priv = sc;
+	et_register(&sc->et);
+
+	/* Register the timecounter. */
+	sc->tc.tc_name = "XENTIMER";
+	sc->tc.tc_quality = XENTIMER_QUALITY;
+	/*
+	 * The underlying resolution is in nanoseconds, since the timer info
+	 * scales TSC frequencies using a fraction that represents time in
+	 * terms of nanoseconds.
+	 */
+	sc->tc.tc_frequency = NSEC_IN_SEC;
+	sc->tc.tc_counter_mask = ~0u;
+	sc->tc.tc_get_timecount = xentimer_get_timecount;
+	sc->tc.tc_priv = sc;
+	tc_init(&sc->tc);
+
+	/* Register the Hypervisor wall clock */
+	clock_register(dev, XENCLOCK_RESOLUTION);
+
+	return (0);
+}
+
+static int
+xentimer_detach(device_t dev)
+{
+
+	/* Implement Xen PV clock teardown - XXX see hpet_detach ? */
+	/* If possible:
+	 * 1. need to deregister timecounter
+	 * 2. need to deregister event timer
+	 * 3. need to deregister virtual IRQ event channels
+	 */
+	return (EBUSY);
+}
+
+/**
+ * The following device methods are disabled because they wouldn't work
+ * properly.
+ */
+#ifdef NOTYET
+static int
+xentimer_resume(device_t dev)
+{
+	struct xentimer_softc *sc = device_get_softc(dev);
+	int error = 0;
+	int i;
+
+	device_printf(sc->dev, "%s", __func__);
+	CPU_FOREACH(i) {
+		struct xentimer_pcpu_data *pcpu = DPCPU_ID_PTR(i, xentimer_pcpu);
+
+		/* Skip inactive timers. */
+		if (pcpu->timer == 0)
+			continue;
+
+		/*
+		 * XXX This won't actually work, because Xen requires that
+		 *     singleshot timers be set while running on the given CPU.
+		 */
+		error = xentimer_vcpu_start_timer(i, pcpu->timer);
+		if (error == -ETIME) {
+			/* Event time has already passed; process. */
+			xentimer_intr(sc);
+		} else if (error != 0) {
+			panic("%s: error %d restarting vcpu %d\n",
+			    __func__, error, i);
+		}
+	}
+
+	return (error);
+}
+
+static int
+xentimer_suspend(device_t dev)
+{
+	struct xentimer_softc *sc = device_get_softc(dev);
+	int error = 0;
+	int i;
+
+	device_printf(sc->dev, "%s", __func__);
+	CPU_FOREACH(i) {
+		struct xentimer_pcpu_data *pcpu = DPCPU_ID_PTR(i, xentimer_pcpu);
+
+		/* Skip inactive timers. */
+		if (pcpu->timer == 0)
+			continue;
+		error = xentimer_vcpu_stop_timer(i);
+		if (error)
+			panic("Error %d stopping VCPU %d timer\n", error, i);
+	}
+
+	return (error);
+}
+#endif
+
+static device_method_t xentimer_methods[] = {
+	DEVMETHOD(device_identify, xentimer_identify),
+	DEVMETHOD(device_probe, xentimer_probe),
+	DEVMETHOD(device_attach, xentimer_attach),
+	DEVMETHOD(device_detach, xentimer_detach),
+#ifdef NOTYET
+	DEVMETHOD(device_suspend, xentimer_suspend),
+	DEVMETHOD(device_resume, xentimer_resume),
+#endif
+	/* clock interface */
+	DEVMETHOD(clock_gettime, xentimer_gettime),
+	DEVMETHOD(clock_settime, xentimer_settime),
+	DEVMETHOD_END
+};
+
+static driver_t xentimer_driver = {
+	"xen_et",
+	xentimer_methods,
+	sizeof(struct xentimer_softc),
+};
+
+DRIVER_MODULE(xentimer, nexus, xentimer_driver, xentimer_devclass, 0, 0);
+MODULE_DEPEND(xentimer, nexus, 1, 1, 1);

Modified: head/sys/i386/xen/clock.c
==============================================================================
--- head/sys/i386/xen/clock.c	Thu Aug 29 23:09:34 2013	(r255045)
+++ head/sys/i386/xen/clock.c	Thu Aug 29 23:11:58 2013	(r255046)
@@ -79,7 +79,6 @@ __FBSDID("$FreeBSD$");
 #include <x86/isa/isa.h>
 #include <isa/rtc.h>
 
-#include <xen/xen_intr.h>
 #include <vm/vm.h>
 #include <vm/pmap.h>
 #include <machine/pmap.h>
@@ -88,7 +87,7 @@ __FBSDID("$FreeBSD$");
 #include <machine/xen/xenfunc.h>
 #include <xen/interface/vcpu.h>
 #include <machine/cpu.h>
-#include <machine/xen/xen_clock_util.h>
+#include <xen/xen_intr.h>
 
 /*
  * 32-bit time_t's can't reach leap years before 1904 or after 2036, so we
@@ -117,6 +116,7 @@ struct mtx clock_lock;
 	mtx_init(&clock_lock, "clk", NULL, MTX_SPIN | MTX_NOPROFILE)
 #define	RTC_LOCK	mtx_lock_spin(&clock_lock)
 #define	RTC_UNLOCK	mtx_unlock_spin(&clock_lock)
+#define	NS_PER_TICK	(1000000000ULL/hz)
 
 int adjkerntz;		/* local offset from GMT in seconds */
 int clkintr_pending;
@@ -124,22 +124,10 @@ int pscnt = 1;
 int psdiv = 1;
 int wall_cmos_clock;
 u_int timer_freq = TIMER_FREQ;
-static int independent_wallclock;
-static int xen_disable_rtc_set;
 static u_long cyc2ns_scale; 
-static struct timespec shadow_tv;
-static uint32_t shadow_tv_version;	/* XXX: lazy locking */
 static uint64_t processed_system_time;	/* stime (ns) at last processing. */
 
-static	const u_char daysinmonth[] = {31,28,31,30,31,30,31,31,30,31,30,31};
-
-int ap_cpu_initclocks(int cpu);
-
-SYSCTL_INT(_machdep, OID_AUTO, independent_wallclock,
-    CTLFLAG_RW, &independent_wallclock, 0, "");
-SYSCTL_INT(_machdep, OID_AUTO, xen_disable_rtc_set,
-    CTLFLAG_RW, &xen_disable_rtc_set, 1, "");
-
+extern volatile uint64_t xen_timer_last_time;
 
 #define do_div(n,base) ({ \
         unsigned long __upper, __low, __high, __mod, __base; \
@@ -156,12 +144,6 @@ SYSCTL_INT(_machdep, OID_AUTO, xen_disab
 })
 
 
-#define NS_PER_TICK (1000000000ULL/hz)
-
-#define rdtscll(val) \
-    __asm__ __volatile__("rdtsc" : "=A" (val))
-
-
 /* convert from cycles(64bits) => nanoseconds (64bits)
  *  basic equation:
  *		ns = cycles / (freq / ns_per_sec)
@@ -184,208 +166,13 @@ static inline void set_cyc2ns_scale(unsi
 
 static inline unsigned long long cycles_2_ns(unsigned long long cyc)
 {
-	return (cyc * cyc2ns_scale) >> CYC2NS_SCALE_FACTOR;
-}
-
-/*
- * Scale a 64-bit delta by scaling and multiplying by a 32-bit fraction,
- * yielding a 64-bit result.
- */
-static inline uint64_t 
-scale_delta(uint64_t delta, uint32_t mul_frac, int shift)
-{
-	uint64_t product;
-	uint32_t tmp1, tmp2;
-
-	if ( shift < 0 )
-		delta >>= -shift;
-	else
-		delta <<= shift;
-
-	__asm__ (
-		"mul  %5       ; "
-		"mov  %4,%%eax ; "
-		"mov  %%edx,%4 ; "
-		"mul  %5       ; "
-		"xor  %5,%5    ; "
-		"add  %4,%%eax ; "
-		"adc  %5,%%edx ; "
-		: "=A" (product), "=r" (tmp1), "=r" (tmp2)
-		: "a" ((uint32_t)delta), "1" ((uint32_t)(delta >> 32)), "2" (mul_frac) );
-
-	return product;
-}
-
-static uint64_t
-get_nsec_offset(struct shadow_time_info *shadow)
-{
-	uint64_t now, delta;
-	rdtscll(now);
-	delta = now - shadow->tsc_timestamp;
-	return scale_delta(delta, shadow->tsc_to_nsec_mul, shadow->tsc_shift);
-}
-
-static void update_wallclock(void)
-{
-	shared_info_t *s = HYPERVISOR_shared_info;
-
-	do {
-		shadow_tv_version = s->wc_version;
-		rmb();
-		shadow_tv.tv_sec  = s->wc_sec;
-		shadow_tv.tv_nsec = s->wc_nsec;
-		rmb();
-	}
-	while ((s->wc_version & 1) | (shadow_tv_version ^ s->wc_version));
-
-}
-
-static void
-add_uptime_to_wallclock(void)
-{
-	struct timespec ut;
-
-	xen_fetch_uptime(&ut);
-	timespecadd(&shadow_tv, &ut);
+	return ((cyc * cyc2ns_scale) >> CYC2NS_SCALE_FACTOR);
 }
 
-/*
- * Reads a consistent set of time-base values from Xen, into a shadow data
- * area. Must be called with the xtime_lock held for writing.
- */
-static void __get_time_values_from_xen(void)
-{
-	shared_info_t           *s = HYPERVISOR_shared_info;
-	struct vcpu_time_info   *src;
-	struct shadow_time_info *dst;
-	uint32_t pre_version, post_version;
-	struct pcpu *pc;
-
-	pc = pcpu_find(smp_processor_id());
-	src = &s->vcpu_info[smp_processor_id()].time;
-	dst = &pc->pc_shadow_time;
-
-	spinlock_enter();
-	do {
-	        pre_version = dst->version = src->version;
-		rmb();
-		dst->tsc_timestamp     = src->tsc_timestamp;
-		dst->system_timestamp  = src->system_time;
-		dst->tsc_to_nsec_mul   = src->tsc_to_system_mul;
-		dst->tsc_shift         = src->tsc_shift;
-		rmb();
-		post_version = src->version;
-	}
-	while ((pre_version & 1) | (pre_version ^ post_version));
-
-	dst->tsc_to_usec_mul = dst->tsc_to_nsec_mul / 1000;
-	spinlock_exit();
-}
-
-
-static inline int time_values_up_to_date(int cpu)
-{
-	struct vcpu_time_info   *src;
-	struct shadow_time_info *dst;
-	struct pcpu *pc;
-
-	src = &HYPERVISOR_shared_info->vcpu_info[cpu].time; 
-	pc = pcpu_find(cpu);
-	dst = &pc->pc_shadow_time;
-
-	rmb();
-	return (dst->version == src->version);
-}
-
-static	unsigned xen_get_timecount(struct timecounter *tc);
-
-static struct timecounter xen_timecounter = {
-	xen_get_timecount,	/* get_timecount */
-	0,			/* no poll_pps */
-	~0u,			/* counter_mask */
-	0,			/* frequency */
-	"ixen",			/* name */
-	0			/* quality */
-};
-
-static struct eventtimer xen_et;
-
-struct xen_et_state {
-	int		mode;
-#define	MODE_STOP	0
-#define	MODE_PERIODIC	1
-#define	MODE_ONESHOT	2
-	int64_t		period;
-	int64_t		next;
-};
-
-static DPCPU_DEFINE(struct xen_et_state, et_state);
-
-static int
-clkintr(void *arg)
-{
-	int64_t now;
-	int cpu = smp_processor_id();
-	struct pcpu *pc = pcpu_find(cpu);
-	struct shadow_time_info *shadow = &pc->pc_shadow_time;
-	struct xen_et_state *state = DPCPU_PTR(et_state);
-
-	do {
-		__get_time_values_from_xen();
-		now = shadow->system_timestamp + get_nsec_offset(shadow);
-	} while (!time_values_up_to_date(cpu));
-
-	/* Process elapsed ticks since last call. */
-	processed_system_time = now;
-	if (state->mode == MODE_PERIODIC) {
-		while (now >= state->next) {
-		        state->next += state->period;
-			if (xen_et.et_active)
-				xen_et.et_event_cb(&xen_et, xen_et.et_arg);
-		}
-		HYPERVISOR_set_timer_op(state->next + 50000);
-	} else if (state->mode == MODE_ONESHOT) {
-		if (xen_et.et_active)
-			xen_et.et_event_cb(&xen_et, xen_et.et_arg);
-	}
-	/*
-	 * Take synchronised time from Xen once a minute if we're not
-	 * synchronised ourselves, and we haven't chosen to keep an independent
-	 * time base.
-	 */
-	
-	if (shadow_tv_version != HYPERVISOR_shared_info->wc_version &&
-	    !independent_wallclock) {
-		printf("[XEN] hypervisor wallclock nudged; nudging TOD.\n");
-		update_wallclock();
-		add_uptime_to_wallclock();
-		tc_setclock(&shadow_tv);
-	}
-	
-	/* XXX TODO */
-	return (FILTER_HANDLED);
-}
 static uint32_t
 getit(void)
 {
-	struct shadow_time_info *shadow;
-	uint64_t time;
-	uint32_t local_time_version;
-	struct pcpu *pc;
-
-	pc = pcpu_find(smp_processor_id());
-	shadow = &pc->pc_shadow_time;
-
-	do {
-	  local_time_version = shadow->version;
-	  barrier();
-	  time = shadow->system_timestamp + get_nsec_offset(shadow);
-	  if (!time_values_up_to_date(smp_processor_id()))
-	    __get_time_values_from_xen(/*cpu */);
-	  barrier();
-	} while (local_time_version != shadow->version);
-
-	  return (time);
+	return (xen_timer_last_time);
 }
 
 
@@ -489,43 +276,12 @@ DELAY(int n)
 #endif
 }
 
-
-/*
- * Restore all the timers non-atomically (XXX: should be atomically).
- *
- * This function is called from pmtimer_resume() to restore all the timers.
- * This should not be necessary, but there are broken laptops that do not
- * restore all the timers on resume.
- */
-void
-timer_restore(void)
-{
-	struct xen_et_state *state = DPCPU_PTR(et_state);
-	struct pcpu *pc;
-
-	/* Get timebases for new environment. */ 
-	__get_time_values_from_xen();
-
-	/* Reset our own concept of passage of system time. */
-	pc = pcpu_find(0);
-	processed_system_time = pc->pc_shadow_time.system_timestamp;
-	state->next = processed_system_time;
-}
-
 void
 startrtclock()
 {
-	unsigned long long alarm;
 	uint64_t __cpu_khz;
 	uint32_t cpu_khz;
 	struct vcpu_time_info *info;
-	struct pcpu *pc;
-
-	pc = pcpu_find(0);
-
-	/* initialize xen values */
-	__get_time_values_from_xen();
-	processed_system_time = pc->pc_shadow_time.system_timestamp;
 
 	__cpu_khz = 1000000ULL << 32;
 	info = &HYPERVISOR_shared_info->vcpu_info[0].time;
@@ -544,12 +300,6 @@ startrtclock()
 
 	set_cyc2ns_scale(cpu_khz/1000);
 	tsc_freq = cpu_khz * 1000;
-
-        timer_freq = 1000000000LL;
-	xen_timecounter.tc_frequency = timer_freq >> 9;
-        tc_init(&xen_timecounter);
-
-	rdtscll(alarm);
 }
 
 /*
@@ -783,116 +533,20 @@ resettodr()
 }
 #endif
 
-static int
-xen_et_start(struct eventtimer *et, sbintime_t first, sbintime_t period)
-{
-	struct xen_et_state *state = DPCPU_PTR(et_state);
-	struct shadow_time_info *shadow;
-	int64_t fperiod;
-	struct pcpu *pc;
-
-	__get_time_values_from_xen();
-
-	if (period != 0) {
-		state->mode = MODE_PERIODIC;
-		state->period = (1000000000LLU * period) >> 32;
-	} else {
-		state->mode = MODE_ONESHOT;
-		state->period = 0;
-	}
-	if (first != 0)
-		fperiod = (1000000000LLU * first) >> 32;
-	else
-		fperiod = state->period;
-
-	pc = pcpu_find(smp_processor_id());
-	shadow = &pc->pc_shadow_time;
-	state->next = shadow->system_timestamp + get_nsec_offset(shadow);
-	state->next += fperiod;
-	HYPERVISOR_set_timer_op(state->next + 50000);
-	return (0);
-}
-
-static int
-xen_et_stop(struct eventtimer *et)

*** DIFF OUTPUT TRUNCATED AT 1000 LINES ***



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?201308292311.r7TNBwlC067869>