freebsd-dev/sys/x86/xen/hvm.c
Roger Pau Monné 27c36a12f1 xen: introduce a new way to setup event channel upcall
The main differences with the currently implemented method are:

 - Requires a local APIC EOI, since it doesn't bypass the local APIC
   as the previous method used to do.
 - Can be set to use different IDT vectors on each vCPU. Note that
   FreeBSD doesn't make use of this feature since the event channel
   IDT vector is reserved system wide.

Note that the old method of setting the event channel upcall is
not removed, and will be used as a fallback if this newly introduced
method is not available.

MFC after:	1 month
Sponsored by:	Citrix Systems R&D
2019-01-30 11:34:52 +00:00

536 lines
13 KiB
C

/*-
* SPDX-License-Identifier: BSD-2-Clause-FreeBSD
*
* Copyright (c) 2008, 2013 Citrix Systems, Inc.
* 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.
*/
#include <sys/cdefs.h>
__FBSDID("$FreeBSD$");
#include <sys/param.h>
#include <sys/bus.h>
#include <sys/kernel.h>
#include <sys/malloc.h>
#include <sys/proc.h>
#include <sys/smp.h>
#include <sys/systm.h>
#include <vm/vm.h>
#include <vm/pmap.h>
#include <vm/vm_param.h>
#include <dev/pci/pcivar.h>
#include <machine/cpufunc.h>
#include <machine/cpu.h>
#include <machine/smp.h>
#include <x86/apicreg.h>
#include <xen/xen-os.h>
#include <xen/error.h>
#include <xen/features.h>
#include <xen/gnttab.h>
#include <xen/hypervisor.h>
#include <xen/hvm.h>
#include <xen/xen_intr.h>
#include <xen/interface/arch-x86/cpuid.h>
#include <xen/interface/hvm/params.h>
#include <xen/interface/vcpu.h>
/*--------------------------- Forward Declarations ---------------------------*/
static void xen_hvm_cpu_init(void);
/*-------------------------------- Global Data -------------------------------*/
enum xen_domain_type xen_domain_type = XEN_NATIVE;
#ifdef SMP
struct cpu_ops xen_hvm_cpu_ops = {
.cpu_init = xen_hvm_cpu_init,
.cpu_resume = xen_hvm_cpu_init
};
#endif
static MALLOC_DEFINE(M_XENHVM, "xen_hvm", "Xen HVM PV Support");
/**
* If non-zero, the hypervisor has been configured to use a direct
* IDT event callback for interrupt injection.
*/
int xen_vector_callback_enabled;
/**
* Start info flags. ATM this only used to store the initial domain flag for
* PVHv2, and it's always empty for HVM guests.
*/
uint32_t hvm_start_flags;
/**
* Signal whether the vector injected for the event channel upcall requires to
* be EOI'ed on the local APIC.
*/
bool xen_evtchn_needs_ack;
/*------------------------------- Per-CPU Data -------------------------------*/
DPCPU_DEFINE(struct vcpu_info, vcpu_local_info);
DPCPU_DEFINE(struct vcpu_info *, vcpu_info);
/*------------------ Hypervisor Access Shared Memory Regions -----------------*/
shared_info_t *HYPERVISOR_shared_info;
/*------------------------------ Sysctl tunables -----------------------------*/
int xen_disable_pv_disks = 0;
int xen_disable_pv_nics = 0;
TUNABLE_INT("hw.xen.disable_pv_disks", &xen_disable_pv_disks);
TUNABLE_INT("hw.xen.disable_pv_nics", &xen_disable_pv_nics);
/*---------------------- XEN Hypervisor Probe and Setup ----------------------*/
static uint32_t cpuid_base;
static uint32_t
xen_hvm_cpuid_base(void)
{
uint32_t base, regs[4];
for (base = 0x40000000; base < 0x40010000; base += 0x100) {
do_cpuid(base, regs);
if (!memcmp("XenVMMXenVMM", &regs[1], 12)
&& (regs[0] - base) >= 2)
return (base);
}
return (0);
}
static void
hypervisor_quirks(unsigned int major, unsigned int minor)
{
#ifdef SMP
if (((major < 4) || (major == 4 && minor <= 5)) &&
msix_disable_migration == -1) {
/*
* Xen hypervisors prior to 4.6.0 do not properly
* handle updates to enabled MSI-X table entries,
* so disable MSI-X interrupt migration in that
* case.
*/
if (bootverbose)
printf(
"Disabling MSI-X interrupt migration due to Xen hypervisor bug.\n"
"Set machdep.msix_disable_migration=0 to forcefully enable it.\n");
msix_disable_migration = 1;
}
#endif
}
static void
hypervisor_version(void)
{
uint32_t regs[4];
int major, minor;
do_cpuid(cpuid_base + 1, regs);
major = regs[0] >> 16;
minor = regs[0] & 0xffff;
printf("XEN: Hypervisor version %d.%d detected.\n", major, minor);
hypervisor_quirks(major, minor);
}
/*
* Allocate and fill in the hypcall page.
*/
int
xen_hvm_init_hypercall_stubs(enum xen_hvm_init_type init_type)
{
uint32_t regs[4];
/* Legacy PVH will get here without the cpuid leaf being set. */
if (cpuid_base == 0)
cpuid_base = xen_hvm_cpuid_base();
if (cpuid_base == 0)
return (ENXIO);
if (xen_domain() && init_type == XEN_HVM_INIT_LATE) {
/*
* If the domain type is already set we can assume that the
* hypercall page has been populated too, so just print the
* version (and apply any quirks) and exit.
*/
hypervisor_version();
return 0;
}
if (init_type == XEN_HVM_INIT_LATE)
hypervisor_version();
/*
* Find the hypercall pages.
*/
do_cpuid(cpuid_base + 2, regs);
if (regs[0] != 1)
return (EINVAL);
wrmsr(regs[1], (init_type == XEN_HVM_INIT_EARLY)
? ((vm_paddr_t)&hypercall_page - KERNBASE)
: vtophys(&hypercall_page));
return (0);
}
static void
xen_hvm_init_shared_info_page(void)
{
struct xen_add_to_physmap xatp;
if (xen_pv_domain()) {
/*
* Already setup in the PV case, shared_info is passed inside
* of the start_info struct at start of day.
*/
return;
}
if (HYPERVISOR_shared_info == NULL) {
HYPERVISOR_shared_info = malloc(PAGE_SIZE, M_XENHVM, M_NOWAIT);
if (HYPERVISOR_shared_info == NULL)
panic("Unable to allocate Xen shared info page");
}
xatp.domid = DOMID_SELF;
xatp.idx = 0;
xatp.space = XENMAPSPACE_shared_info;
xatp.gpfn = vtophys(HYPERVISOR_shared_info) >> PAGE_SHIFT;
if (HYPERVISOR_memory_op(XENMEM_add_to_physmap, &xatp))
panic("HYPERVISOR_memory_op failed");
}
static int
set_percpu_callback(unsigned int vcpu)
{
struct xen_hvm_evtchn_upcall_vector vec;
int error;
vec.vcpu = vcpu;
vec.vector = IDT_EVTCHN;
error = HYPERVISOR_hvm_op(HVMOP_set_evtchn_upcall_vector, &vec);
return (error != 0 ? xen_translate_error(error) : 0);
}
/*
* Tell the hypervisor how to contact us for event channel callbacks.
*/
void
xen_hvm_set_callback(device_t dev)
{
struct xen_hvm_param xhp;
int irq;
if (xen_vector_callback_enabled)
return;
xhp.domid = DOMID_SELF;
xhp.index = HVM_PARAM_CALLBACK_IRQ;
if (xen_feature(XENFEAT_hvm_callback_vector) != 0) {
int error;
error = set_percpu_callback(0);
if (error == 0) {
xen_evtchn_needs_ack = true;
/* Trick toolstack to think we are enlightened */
xhp.value = 1;
} else
xhp.value = HVM_CALLBACK_VECTOR(IDT_EVTCHN);
error = HYPERVISOR_hvm_op(HVMOP_set_param, &xhp);
if (error == 0) {
xen_vector_callback_enabled = 1;
return;
} else if (xen_evtchn_needs_ack)
panic("Unable to setup fake HVM param: %d", error);
printf("Xen HVM callback vector registration failed (%d). "
"Falling back to emulated device interrupt\n", error);
}
xen_vector_callback_enabled = 0;
if (dev == NULL) {
/*
* Called from early boot or resume.
* xenpci will invoke us again later.
*/
return;
}
irq = pci_get_irq(dev);
if (irq < 16) {
xhp.value = HVM_CALLBACK_GSI(irq);
} else {
u_int slot;
u_int pin;
slot = pci_get_slot(dev);
pin = pci_get_intpin(dev) - 1;
xhp.value = HVM_CALLBACK_PCI_INTX(slot, pin);
}
if (HYPERVISOR_hvm_op(HVMOP_set_param, &xhp) != 0)
panic("Can't set evtchn callback");
}
#define XEN_MAGIC_IOPORT 0x10
enum {
XMI_MAGIC = 0x49d2,
XMI_UNPLUG_IDE_DISKS = 0x01,
XMI_UNPLUG_NICS = 0x02,
XMI_UNPLUG_IDE_EXCEPT_PRI_MASTER = 0x04
};
static void
xen_hvm_disable_emulated_devices(void)
{
u_short disable_devs = 0;
if (xen_pv_domain()) {
/*
* No emulated devices in the PV case, so no need to unplug
* anything.
*/
if (xen_disable_pv_disks != 0 || xen_disable_pv_nics != 0)
printf("PV devices cannot be disabled in PV guests\n");
return;
}
if (inw(XEN_MAGIC_IOPORT) != XMI_MAGIC)
return;
if (xen_disable_pv_disks == 0) {
if (bootverbose)
printf("XEN: disabling emulated disks\n");
disable_devs |= XMI_UNPLUG_IDE_DISKS;
}
if (xen_disable_pv_nics == 0) {
if (bootverbose)
printf("XEN: disabling emulated nics\n");
disable_devs |= XMI_UNPLUG_NICS;
}
if (disable_devs != 0)
outw(XEN_MAGIC_IOPORT, disable_devs);
}
static void
xen_hvm_init(enum xen_hvm_init_type init_type)
{
int error;
int i;
if (init_type == XEN_HVM_INIT_CANCELLED_SUSPEND)
return;
error = xen_hvm_init_hypercall_stubs(init_type);
switch (init_type) {
case XEN_HVM_INIT_LATE:
if (error != 0)
return;
/*
* If xen_domain_type is not set at this point
* it means we are inside a (PV)HVM guest, because
* for PVH the guest type is set much earlier
* (see hammer_time_xen).
*/
if (!xen_domain()) {
xen_domain_type = XEN_HVM_DOMAIN;
vm_guest = VM_GUEST_XEN;
}
setup_xen_features();
#ifdef SMP
cpu_ops = xen_hvm_cpu_ops;
#endif
break;
case XEN_HVM_INIT_RESUME:
if (error != 0)
panic("Unable to init Xen hypercall stubs on resume");
/* Clear stale vcpu_info. */
CPU_FOREACH(i)
DPCPU_ID_SET(i, vcpu_info, NULL);
break;
default:
panic("Unsupported HVM initialization type");
}
xen_vector_callback_enabled = 0;
xen_evtchn_needs_ack = false;
xen_hvm_set_callback(NULL);
/*
* On (PV)HVM domains we need to request the hypervisor to
* fill the shared info page, for PVH guest the shared_info page
* is passed inside the start_info struct and is already set, so this
* functions are no-ops.
*/
xen_hvm_init_shared_info_page();
xen_hvm_disable_emulated_devices();
}
void
xen_hvm_suspend(void)
{
}
void
xen_hvm_resume(bool suspend_cancelled)
{
xen_hvm_init(suspend_cancelled ?
XEN_HVM_INIT_CANCELLED_SUSPEND : XEN_HVM_INIT_RESUME);
/* Register vcpu_info area for CPU#0. */
xen_hvm_cpu_init();
}
static void
xen_hvm_sysinit(void *arg __unused)
{
xen_hvm_init(XEN_HVM_INIT_LATE);
}
SYSINIT(xen_hvm_init, SI_SUB_HYPERVISOR, SI_ORDER_FIRST, xen_hvm_sysinit, NULL);
static void
xen_hvm_cpu_init(void)
{
struct vcpu_register_vcpu_info info;
struct vcpu_info *vcpu_info;
uint32_t regs[4];
int cpu, rc;
if (!xen_domain())
return;
if (DPCPU_GET(vcpu_info) != NULL) {
/*
* vcpu_info is already set. We're resuming
* from a failed migration and our pre-suspend
* configuration is still valid.
*/
return;
}
/*
* Set vCPU ID. If available fetch the ID from CPUID, if not just use
* the ACPI ID.
*/
KASSERT(cpuid_base != 0, ("Invalid base Xen CPUID leaf"));
cpuid_count(cpuid_base + 4, 0, regs);
KASSERT((regs[0] & XEN_HVM_CPUID_VCPU_ID_PRESENT) ||
!xen_pv_domain(),
("Xen PV domain without vcpu_id in cpuid"));
PCPU_SET(vcpu_id, (regs[0] & XEN_HVM_CPUID_VCPU_ID_PRESENT) ?
regs[1] : PCPU_GET(acpi_id));
if (xen_evtchn_needs_ack && !IS_BSP()) {
/*
* Setup the per-vpcu event channel upcall vector. This is only
* required when using the new HVMOP_set_evtchn_upcall_vector
* hypercall, which allows using a different vector for each
* vCPU. Note that FreeBSD uses the same vector for all vCPUs
* because it's not dynamically allocated.
*/
rc = set_percpu_callback(PCPU_GET(vcpu_id));
if (rc != 0)
panic("Event channel upcall vector setup failed: %d",
rc);
}
/*
* Set the vCPU info.
*
* NB: the vCPU info for vCPUs < 32 can be fetched from the shared info
* page, but in order to make sure the mapping code is correct always
* attempt to map the vCPU info at a custom place.
*/
vcpu_info = DPCPU_PTR(vcpu_local_info);
cpu = PCPU_GET(vcpu_id);
info.mfn = vtophys(vcpu_info) >> PAGE_SHIFT;
info.offset = vtophys(vcpu_info) - trunc_page(vtophys(vcpu_info));
rc = HYPERVISOR_vcpu_op(VCPUOP_register_vcpu_info, cpu, &info);
if (rc != 0)
DPCPU_SET(vcpu_info, &HYPERVISOR_shared_info->vcpu_info[cpu]);
else
DPCPU_SET(vcpu_info, vcpu_info);
}
SYSINIT(xen_hvm_cpu_init, SI_SUB_INTR, SI_ORDER_FIRST, xen_hvm_cpu_init, NULL);
/* HVM/PVH start_info accessors */
static vm_paddr_t
hvm_get_xenstore_mfn(void)
{
return (hvm_get_parameter(HVM_PARAM_STORE_PFN));
}
static evtchn_port_t
hvm_get_xenstore_evtchn(void)
{
return (hvm_get_parameter(HVM_PARAM_STORE_EVTCHN));
}
static vm_paddr_t
hvm_get_console_mfn(void)
{
return (hvm_get_parameter(HVM_PARAM_CONSOLE_PFN));
}
static evtchn_port_t
hvm_get_console_evtchn(void)
{
return (hvm_get_parameter(HVM_PARAM_CONSOLE_EVTCHN));
}
static uint32_t
hvm_get_start_flags(void)
{
return (hvm_start_flags);
}
struct hypervisor_info hypervisor_info = {
.get_xenstore_mfn = hvm_get_xenstore_mfn,
.get_xenstore_evtchn = hvm_get_xenstore_evtchn,
.get_console_mfn = hvm_get_console_mfn,
.get_console_evtchn = hvm_get_console_evtchn,
.get_start_flags = hvm_get_start_flags,
};