aboutsummaryrefslogtreecommitdiff
diff options
context:
space:
mode:
authorAdam Fenn <adam@fenn.io>2021-08-04 15:42:48 +0000
committerKonstantin Belousov <kib@FreeBSD.org>2021-10-12 16:01:26 +0000
commit6fa88a627d5e9d290022b6f463effc99f3df8ee2 (patch)
treed7d6fedd29a6157301ec08ad0f951b5f99e700c3
parent249f2478e38224df653878bc58ab549f1b7c0465 (diff)
downloadsrc-6fa88a627d5e.tar.gz
src-6fa88a627d5e.zip
kvm_clock: KVM paravirtual clock support
Add support for the KVM paravirtual clock device. Sponsored by: Juniper Networks, Inc. Sponsored by: Klara, Inc. Reviewed by: kib Differential Revision: https://reviews.freebsd.org/D29733 (cherry picked from commit 6c69c6bb4c7ffde48b130df707799d4eca21ca9f)
-rw-r--r--sys/amd64/conf/GENERIC3
-rw-r--r--sys/amd64/conf/MINIMAL3
-rw-r--r--sys/amd64/conf/NOTES3
-rw-r--r--sys/conf/files.amd643
-rw-r--r--sys/conf/files.i3863
-rw-r--r--sys/dev/kvm_clock/kvm_clock.c240
-rw-r--r--sys/i386/conf/GENERIC3
-rw-r--r--sys/i386/conf/NOTES3
-rw-r--r--sys/x86/include/kvm.h79
9 files changed, 338 insertions, 2 deletions
diff --git a/sys/amd64/conf/GENERIC b/sys/amd64/conf/GENERIC
index 0784544c9c7d..84912e695262 100644
--- a/sys/amd64/conf/GENERIC
+++ b/sys/amd64/conf/GENERIC
@@ -362,6 +362,9 @@ device virtio_blk # VirtIO Block device
device virtio_scsi # VirtIO SCSI device
device virtio_balloon # VirtIO Memory Balloon device
+# Linux KVM paravirtualization support
+device kvm_clock # KVM paravirtual clock driver
+
# HyperV drivers and enhancement support
device hyperv # HyperV drivers
diff --git a/sys/amd64/conf/MINIMAL b/sys/amd64/conf/MINIMAL
index 7c035e871e79..97a97aacff77 100644
--- a/sys/amd64/conf/MINIMAL
+++ b/sys/amd64/conf/MINIMAL
@@ -133,6 +133,9 @@ device gif # IPv6 and IPv4 tunneling
# Note that 'bpf' is required for DHCP.
device bpf # Berkeley packet filter
+# Linux KVM paravirtualization support
+device kvm_clock # KVM paravirtual clock driver
+
# Xen HVM Guest Optimizations
# NOTE: XENHVM depends on xenpci. They must be added or removed together.
options XENHVM # Xen HVM kernel infrastructure
diff --git a/sys/amd64/conf/NOTES b/sys/amd64/conf/NOTES
index 0b949acdc4a3..abf4397787cd 100644
--- a/sys/amd64/conf/NOTES
+++ b/sys/amd64/conf/NOTES
@@ -527,6 +527,9 @@ device virtio_balloon # VirtIO Memory Balloon device
device virtio_random # VirtIO Entropy device
device virtio_console # VirtIO Console device
+# Linux KVM paravirtualization support
+device kvm_clock # KVM paravirtual clock driver
+
# Microsoft Hyper-V enhancement support
device hyperv # HyperV drivers
diff --git a/sys/conf/files.amd64 b/sys/conf/files.amd64
index ad698de09fe7..425abcd8b4c7 100644
--- a/sys/conf/files.amd64
+++ b/sys/conf/files.amd64
@@ -398,6 +398,7 @@ dev/hyperv/vmbus/vmbus_res.c optional hyperv
dev/hyperv/vmbus/vmbus_xact.c optional hyperv
dev/hyperv/vmbus/amd64/hyperv_machdep.c optional hyperv
dev/hyperv/vmbus/amd64/vmbus_vector.S optional hyperv
+dev/kvm_clock/kvm_clock.c optional kvm_clock
dev/nctgpio/nctgpio.c optional nctgpio
dev/nfe/if_nfe.c optional nfe pci
dev/ntb/if_ntb/if_ntb.c optional if_ntb
@@ -819,7 +820,7 @@ x86/x86/mp_x86.c optional smp
x86/x86/mp_watchdog.c optional mp_watchdog smp
x86/x86/msi.c optional pci
x86/x86/nexus.c standard
-x86/x86/pvclock.c optional xenhvm
+x86/x86/pvclock.c optional kvm_clock | xenhvm
x86/x86/stack_machdep.c optional ddb | stack
x86/x86/tsc.c standard
x86/x86/ucode.c standard
diff --git a/sys/conf/files.i386 b/sys/conf/files.i386
index 934c72c87044..4bb09402791a 100644
--- a/sys/conf/files.i386
+++ b/sys/conf/files.i386
@@ -285,6 +285,7 @@ dev/ipmi/ipmi_ssif.c optional ipmi smbus
dev/ipmi/ipmi_pci.c optional ipmi pci
dev/ipmi/ipmi_linux.c optional ipmi compat_linux
dev/itwd/itwd.c optional itwd
+dev/kvm_clock/kvm_clock.c optional kvm_clock
dev/le/if_le_isa.c optional le isa
dev/mse/mse.c optional mse
dev/mse/mse_isa.c optional mse isa
@@ -635,7 +636,7 @@ x86/x86/nexus.c standard
x86/x86/stack_machdep.c optional ddb | stack
x86/x86/tsc.c standard
x86/x86/ucode.c standard
-x86/x86/pvclock.c optional xenhvm
+x86/x86/pvclock.c optional kvm_clock | xenhvm
x86/x86/delay.c standard
x86/xen/hvm.c optional xenhvm
x86/xen/xen_intr.c optional xenhvm
diff --git a/sys/dev/kvm_clock/kvm_clock.c b/sys/dev/kvm_clock/kvm_clock.c
new file mode 100644
index 000000000000..1a76432e417d
--- /dev/null
+++ b/sys/dev/kvm_clock/kvm_clock.c
@@ -0,0 +1,240 @@
+/*-
+ * SPDX-License-Identifier: BSD-2-Clause-FreeBSD
+ *
+ * Copyright (c) 2014 Bryan Venteicher <bryanv@FreeBSD.org>
+ * Copyright (c) 2021 Mathieu Chouquet-Stringer
+ * Copyright (c) 2021 Juniper Networks, Inc.
+ * Copyright (c) 2021 Klara, Inc.
+ *
+ * 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.
+ */
+
+/*
+ * Linux KVM paravirtual clock support
+ *
+ * References:
+ * - [1] https://www.kernel.org/doc/html/latest/virt/kvm/cpuid.html
+ * - [2] https://www.kernel.org/doc/html/latest/virt/kvm/msr.html
+ */
+
+#include <sys/cdefs.h>
+__FBSDID("$FreeBSD$");
+
+#include <sys/param.h>
+#include <sys/bus.h>
+#include <sys/domainset.h>
+#include <sys/kernel.h>
+#include <sys/malloc.h>
+#include <sys/module.h>
+#include <sys/smp.h>
+
+#include <vm/vm.h>
+#include <vm/pmap.h>
+#include <vm/vm_extern.h>
+
+#include <machine/pvclock.h>
+#include <x86/kvm.h>
+
+#include "clock_if.h"
+
+#define KVM_CLOCK_DEVNAME "kvmclock"
+/*
+ * Note: Chosen to be (1) above HPET's value (always 950), (2) above the TSC's
+ * default value of 800, and (3) below the TSC's value when it supports the
+ * "Invariant TSC" feature and is believed to be synchronized across all CPUs.
+ */
+#define KVM_CLOCK_TC_QUALITY 975
+
+struct kvm_clock_softc {
+ struct pvclock pvc;
+ struct pvclock_wall_clock wc;
+ struct pvclock_vcpu_time_info *timeinfos;
+ u_int msr_tc;
+ u_int msr_wc;
+};
+
+static devclass_t kvm_clock_devclass;
+
+static struct pvclock_wall_clock *kvm_clock_get_wallclock(void *arg);
+static void kvm_clock_system_time_enable(struct kvm_clock_softc *sc);
+static void kvm_clock_system_time_enable_pcpu(void *arg);
+
+static struct pvclock_wall_clock *
+kvm_clock_get_wallclock(void *arg)
+{
+ struct kvm_clock_softc *sc = arg;
+
+ wrmsr(sc->msr_wc, vtophys(&sc->wc));
+ return (&sc->wc);
+}
+
+static void
+kvm_clock_system_time_enable(struct kvm_clock_softc *sc)
+{
+ smp_rendezvous(NULL, kvm_clock_system_time_enable_pcpu, NULL, sc);
+}
+
+static void
+kvm_clock_system_time_enable_pcpu(void *arg)
+{
+ struct kvm_clock_softc *sc = arg;
+
+ /*
+ * See [2]; the lsb of this MSR is the system time enable bit.
+ */
+ wrmsr(sc->msr_tc, vtophys(&(sc->timeinfos)[curcpu]) | 1);
+}
+
+static void
+kvm_clock_identify(driver_t *driver, device_t parent)
+{
+ u_int regs[4];
+
+ kvm_cpuid_get_features(regs);
+ if ((regs[0] &
+ (KVM_FEATURE_CLOCKSOURCE2 | KVM_FEATURE_CLOCKSOURCE)) == 0)
+ return;
+ if (device_find_child(parent, KVM_CLOCK_DEVNAME, -1))
+ return;
+ BUS_ADD_CHILD(parent, 0, KVM_CLOCK_DEVNAME, 0);
+}
+
+static int
+kvm_clock_probe(device_t dev)
+{
+ device_set_desc(dev, "KVM paravirtual clock");
+ return (BUS_PROBE_DEFAULT);
+}
+
+static int
+kvm_clock_attach(device_t dev)
+{
+ u_int regs[4];
+ struct kvm_clock_softc *sc = device_get_softc(dev);
+ bool stable_flag_supported;
+
+ /* Process KVM "features" CPUID leaf content: */
+ kvm_cpuid_get_features(regs);
+ if ((regs[0] & KVM_FEATURE_CLOCKSOURCE2) != 0) {
+ sc->msr_tc = KVM_MSR_SYSTEM_TIME_NEW;
+ sc->msr_wc = KVM_MSR_WALL_CLOCK_NEW;
+ } else {
+ KASSERT((regs[0] & KVM_FEATURE_CLOCKSOURCE) != 0,
+ ("Clocksource feature flags disappeared since "
+ "kvm_clock_identify: regs[0] %#0x.", regs[0]));
+ sc->msr_tc = KVM_MSR_SYSTEM_TIME;
+ sc->msr_wc = KVM_MSR_WALL_CLOCK;
+ }
+ stable_flag_supported =
+ (regs[0] & KVM_FEATURE_CLOCKSOURCE_STABLE_BIT) != 0;
+
+ /* Set up 'struct pvclock_vcpu_time_info' page(s): */
+ sc->timeinfos = (struct pvclock_vcpu_time_info *)kmem_malloc(mp_ncpus *
+ sizeof(struct pvclock_vcpu_time_info), M_WAITOK | M_ZERO);
+ kvm_clock_system_time_enable(sc);
+
+ /*
+ * Init pvclock; register KVM clock wall clock, register KVM clock
+ * timecounter, and set up the requisite infrastructure for vDSO access
+ * to this timecounter.
+ * Regarding 'tc_flags': Since the KVM MSR documentation does not
+ * specifically discuss suspend/resume scenarios, conservatively
+ * leave 'TC_FLAGS_SUSPEND_SAFE' cleared and assume that the system
+ * time must be re-inited in such cases.
+ */
+ sc->pvc.get_wallclock = kvm_clock_get_wallclock;
+ sc->pvc.get_wallclock_arg = sc;
+ sc->pvc.timeinfos = sc->timeinfos;
+ sc->pvc.stable_flag_supported = stable_flag_supported;
+ pvclock_init(&sc->pvc, dev, KVM_CLOCK_DEVNAME, KVM_CLOCK_TC_QUALITY, 0);
+ return (0);
+}
+
+static int
+kvm_clock_detach(device_t dev)
+{
+ struct kvm_clock_softc *sc = device_get_softc(dev);
+
+ return (pvclock_destroy(&sc->pvc));
+}
+
+static int
+kvm_clock_suspend(device_t dev)
+{
+ return (0);
+}
+
+static int
+kvm_clock_resume(device_t dev)
+{
+ /*
+ * See note in 'kvm_clock_attach()' regarding 'TC_FLAGS_SUSPEND_SAFE';
+ * conservatively assume that the system time must be re-inited in
+ * suspend/resume scenarios.
+ */
+ kvm_clock_system_time_enable(device_get_softc(dev));
+ pvclock_resume();
+ inittodr(time_second);
+ return (0);
+}
+
+static int
+kvm_clock_gettime(device_t dev, struct timespec *ts)
+{
+ struct kvm_clock_softc *sc = device_get_softc(dev);
+
+ pvclock_gettime(&sc->pvc, ts);
+ return (0);
+}
+
+static int
+kvm_clock_settime(device_t dev, struct timespec *ts)
+{
+ /*
+ * Even though it is not possible to set the KVM clock's wall clock, to
+ * avoid the possibility of periodic benign error messages from
+ * 'settime_task_func()', report success rather than, e.g., 'ENODEV'.
+ */
+ return (0);
+}
+
+static device_method_t kvm_clock_methods[] = {
+ DEVMETHOD(device_identify, kvm_clock_identify),
+ DEVMETHOD(device_probe, kvm_clock_probe),
+ DEVMETHOD(device_attach, kvm_clock_attach),
+ DEVMETHOD(device_detach, kvm_clock_detach),
+ DEVMETHOD(device_suspend, kvm_clock_suspend),
+ DEVMETHOD(device_resume, kvm_clock_resume),
+ /* clock interface */
+ DEVMETHOD(clock_gettime, kvm_clock_gettime),
+ DEVMETHOD(clock_settime, kvm_clock_settime),
+
+ DEVMETHOD_END
+};
+
+static driver_t kvm_clock_driver = {
+ KVM_CLOCK_DEVNAME,
+ kvm_clock_methods,
+ sizeof(struct kvm_clock_softc),
+};
+
+DRIVER_MODULE(kvm_clock, nexus, kvm_clock_driver, kvm_clock_devclass, 0, 0);
diff --git a/sys/i386/conf/GENERIC b/sys/i386/conf/GENERIC
index ecb7529df250..d9cb9ed0776e 100644
--- a/sys/i386/conf/GENERIC
+++ b/sys/i386/conf/GENERIC
@@ -353,6 +353,9 @@ device virtio_blk # VirtIO Block device
device virtio_scsi # VirtIO SCSI device
device virtio_balloon # VirtIO Memory Balloon device
+# Linux KVM paravirtualization support
+device kvm_clock # KVM paravirtual clock driver
+
# HyperV drivers and enhancement support
device hyperv # HyperV drivers
diff --git a/sys/i386/conf/NOTES b/sys/i386/conf/NOTES
index 211b3a6fa7e8..34561c548622 100644
--- a/sys/i386/conf/NOTES
+++ b/sys/i386/conf/NOTES
@@ -771,6 +771,9 @@ device virtio_balloon # VirtIO Memory Balloon device
device virtio_random # VirtIO Entropy device
device virtio_console # VirtIO Console device
+# Linux KVM paravirtualization support
+device kvm_clock # KVM paravirtual clock driver
+
device hyperv # HyperV drivers
#####################################################################
diff --git a/sys/x86/include/kvm.h b/sys/x86/include/kvm.h
new file mode 100644
index 000000000000..b74d45d07981
--- /dev/null
+++ b/sys/x86/include/kvm.h
@@ -0,0 +1,79 @@
+/*-
+ * SPDX-License-Identifier: BSD-2-Clause-FreeBSD
+ *
+ * Copyright (c) 2014 Bryan Venteicher <bryanv@FreeBSD.org>
+ * Copyright (c) 2021 Mathieu Chouquet-Stringer
+ * Copyright (c) 2021 Juniper Networks, Inc.
+ * Copyright (c) 2021 Klara, Inc.
+ *
+ * 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.
+ *
+ * $FreeBSD$
+ */
+
+/*
+ * Linux KVM paravirtualization: common definitions
+ *
+ * References:
+ * - [1] https://www.kernel.org/doc/html/latest/virt/kvm/cpuid.html
+ * - [2] https://www.kernel.org/doc/html/latest/virt/kvm/msr.html
+ */
+
+#ifndef _X86_KVM_H_
+#define _X86_KVM_H_
+
+#include <sys/types.h>
+#include <sys/systm.h>
+
+#include <machine/md_var.h>
+
+#define KVM_CPUID_SIGNATURE 0x40000000
+#define KVM_CPUID_FEATURES_LEAF 0x40000001
+
+#define KVM_FEATURE_CLOCKSOURCE 0x00000001
+#define KVM_FEATURE_CLOCKSOURCE2 0x00000008
+#define KVM_FEATURE_CLOCKSOURCE_STABLE_BIT 0x01000000
+
+/* Deprecated: for the CLOCKSOURCE feature. */
+#define KVM_MSR_WALL_CLOCK 0x11
+#define KVM_MSR_SYSTEM_TIME 0x12
+
+#define KVM_MSR_WALL_CLOCK_NEW 0x4b564d00
+#define KVM_MSR_SYSTEM_TIME_NEW 0x4b564d01
+
+static inline bool
+kvm_cpuid_features_leaf_supported(void)
+{
+ return (vm_guest == VM_GUEST_KVM &&
+ KVM_CPUID_FEATURES_LEAF <= hv_high);
+}
+
+static inline void
+kvm_cpuid_get_features(u_int *regs)
+{
+ if (!kvm_cpuid_features_leaf_supported())
+ regs[0] = regs[1] = regs[2] = regs[3] = 0;
+ else
+ do_cpuid(KVM_CPUID_FEATURES_LEAF, regs);
+}
+
+#endif /* !_X86_KVM_H_ */