2 * Copyright (c) 2008, 2013 Citrix Systems, Inc.
3 * Copyright (c) 2012 Spectra Logic Corporation
6 * Redistribution and use in source and binary forms, with or without
7 * modification, are permitted provided that the following conditions
9 * 1. Redistributions of source code must retain the above copyright
10 * notice, this list of conditions and the following disclaimer.
11 * 2. Redistributions in binary form must reproduce the above copyright
12 * notice, this list of conditions and the following disclaimer in the
13 * documentation and/or other materials provided with the distribution.
15 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS AS IS'' AND
16 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
17 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
18 * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
19 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
20 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
21 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
22 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
23 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
24 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
28 #include <sys/cdefs.h>
29 __FBSDID("$FreeBSD$");
31 #include <sys/param.h>
33 #include <sys/kernel.h>
34 #include <sys/malloc.h>
37 #include <sys/systm.h>
42 #include <dev/pci/pcivar.h>
44 #include <machine/cpufunc.h>
45 #include <machine/cpu.h>
46 #include <machine/smp.h>
48 #include <x86/apicreg.h>
50 #include <xen/xen-os.h>
51 #include <xen/features.h>
52 #include <xen/gnttab.h>
53 #include <xen/hypervisor.h>
55 #include <xen/xen_intr.h>
57 #include <xen/interface/hvm/params.h>
58 #include <xen/interface/vcpu.h>
60 /*--------------------------- Forward Declarations ---------------------------*/
62 static void xen_hvm_cpu_resume(void);
64 static void xen_hvm_cpu_init(void);
66 /*---------------------------- Extern Declarations ---------------------------*/
67 /* Variables used by mp_machdep to perform the bitmap IPI */
68 extern volatile u_int cpu_ipi_pending[MAXCPU];
70 /*-------------------------------- Local Types -------------------------------*/
71 enum xen_hvm_init_type {
73 XEN_HVM_INIT_CANCELLED_SUSPEND,
77 /*-------------------------------- Global Data -------------------------------*/
78 enum xen_domain_type xen_domain_type = XEN_NATIVE;
81 struct cpu_ops xen_hvm_cpu_ops = {
82 .cpu_init = xen_hvm_cpu_init,
83 .cpu_resume = xen_hvm_cpu_resume
87 static MALLOC_DEFINE(M_XENHVM, "xen_hvm", "Xen HVM PV Support");
90 * If non-zero, the hypervisor has been configured to use a direct
91 * IDT event callback for interrupt injection.
93 int xen_vector_callback_enabled;
95 /*------------------------------- Per-CPU Data -------------------------------*/
96 DPCPU_DEFINE(struct vcpu_info, vcpu_local_info);
97 DPCPU_DEFINE(struct vcpu_info *, vcpu_info);
99 /*------------------ Hypervisor Access Shared Memory Regions -----------------*/
100 shared_info_t *HYPERVISOR_shared_info;
101 start_info_t *HYPERVISOR_start_info;
104 /* XEN diverged cpu operations */
106 xen_hvm_cpu_resume(void)
108 u_int cpuid = PCPU_GET(cpuid);
111 * Reset pending bitmap IPIs, because Xen doesn't preserve pending
112 * event channels on migration.
114 cpu_ipi_pending[cpuid] = 0;
116 /* register vcpu_info area */
120 /*---------------------- XEN Hypervisor Probe and Setup ----------------------*/
122 xen_hvm_cpuid_base(void)
124 uint32_t base, regs[4];
126 for (base = 0x40000000; base < 0x40010000; base += 0x100) {
127 do_cpuid(base, regs);
128 if (!memcmp("XenVMMXenVMM", ®s[1], 12)
129 && (regs[0] - base) >= 2)
136 * Allocate and fill in the hypcall page.
139 xen_hvm_init_hypercall_stubs(enum xen_hvm_init_type init_type)
141 uint32_t base, regs[4];
144 if (xen_pv_domain()) {
145 /* hypercall page is already set in the PV case */
149 base = xen_hvm_cpuid_base();
153 if (init_type == XEN_HVM_INIT_COLD) {
154 do_cpuid(base + 1, regs);
155 printf("XEN: Hypervisor version %d.%d detected.\n",
156 regs[0] >> 16, regs[0] & 0xffff);
160 * Find the hypercall pages.
162 do_cpuid(base + 2, regs);
164 for (i = 0; i < regs[0]; i++)
165 wrmsr(regs[1], vtophys(&hypercall_page + i * PAGE_SIZE) + i);
171 xen_hvm_init_shared_info_page(void)
173 struct xen_add_to_physmap xatp;
175 if (xen_pv_domain()) {
177 * Already setup in the PV case, shared_info is passed inside
178 * of the start_info struct at start of day.
183 if (HYPERVISOR_shared_info == NULL) {
184 HYPERVISOR_shared_info = malloc(PAGE_SIZE, M_XENHVM, M_NOWAIT);
185 if (HYPERVISOR_shared_info == NULL)
186 panic("Unable to allocate Xen shared info page");
189 xatp.domid = DOMID_SELF;
191 xatp.space = XENMAPSPACE_shared_info;
192 xatp.gpfn = vtophys(HYPERVISOR_shared_info) >> PAGE_SHIFT;
193 if (HYPERVISOR_memory_op(XENMEM_add_to_physmap, &xatp))
194 panic("HYPERVISOR_memory_op failed");
198 * Tell the hypervisor how to contact us for event channel callbacks.
201 xen_hvm_set_callback(device_t dev)
203 struct xen_hvm_param xhp;
206 if (xen_vector_callback_enabled)
209 xhp.domid = DOMID_SELF;
210 xhp.index = HVM_PARAM_CALLBACK_IRQ;
211 if (xen_feature(XENFEAT_hvm_callback_vector) != 0) {
214 xhp.value = HVM_CALLBACK_VECTOR(IDT_EVTCHN);
215 error = HYPERVISOR_hvm_op(HVMOP_set_param, &xhp);
217 xen_vector_callback_enabled = 1;
220 printf("Xen HVM callback vector registration failed (%d). "
221 "Falling back to emulated device interrupt\n", error);
223 xen_vector_callback_enabled = 0;
226 * Called from early boot or resume.
227 * xenpci will invoke us again later.
232 irq = pci_get_irq(dev);
234 xhp.value = HVM_CALLBACK_GSI(irq);
239 slot = pci_get_slot(dev);
240 pin = pci_get_intpin(dev) - 1;
241 xhp.value = HVM_CALLBACK_PCI_INTX(slot, pin);
244 if (HYPERVISOR_hvm_op(HVMOP_set_param, &xhp) != 0)
245 panic("Can't set evtchn callback");
248 #define XEN_MAGIC_IOPORT 0x10
251 XMI_UNPLUG_IDE_DISKS = 0x01,
252 XMI_UNPLUG_NICS = 0x02,
253 XMI_UNPLUG_IDE_EXCEPT_PRI_MASTER = 0x04
257 xen_hvm_disable_emulated_devices(void)
260 if (xen_pv_domain()) {
262 * No emulated devices in the PV case, so no need to unplug
268 if (inw(XEN_MAGIC_IOPORT) != XMI_MAGIC)
272 printf("XEN: Disabling emulated block and network devices\n");
273 outw(XEN_MAGIC_IOPORT, XMI_UNPLUG_IDE_DISKS|XMI_UNPLUG_NICS);
277 xen_hvm_init(enum xen_hvm_init_type init_type)
282 if (init_type == XEN_HVM_INIT_CANCELLED_SUSPEND)
285 error = xen_hvm_init_hypercall_stubs(init_type);
288 case XEN_HVM_INIT_COLD:
293 * If xen_domain_type is not set at this point
294 * it means we are inside a (PV)HVM guest, because
295 * for PVH the guest type is set much earlier
296 * (see hammer_time_xen).
299 xen_domain_type = XEN_HVM_DOMAIN;
300 vm_guest = VM_GUEST_XEN;
303 setup_xen_features();
305 cpu_ops = xen_hvm_cpu_ops;
308 case XEN_HVM_INIT_RESUME:
310 panic("Unable to init Xen hypercall stubs on resume");
312 /* Clear stale vcpu_info. */
314 DPCPU_ID_SET(i, vcpu_info, NULL);
317 panic("Unsupported HVM initialization type");
320 xen_vector_callback_enabled = 0;
321 xen_hvm_set_callback(NULL);
324 * On (PV)HVM domains we need to request the hypervisor to
325 * fill the shared info page, for PVH guest the shared_info page
326 * is passed inside the start_info struct and is already set, so this
327 * functions are no-ops.
329 xen_hvm_init_shared_info_page();
330 xen_hvm_disable_emulated_devices();
334 xen_hvm_suspend(void)
339 xen_hvm_resume(bool suspend_cancelled)
342 xen_hvm_init(suspend_cancelled ?
343 XEN_HVM_INIT_CANCELLED_SUSPEND : XEN_HVM_INIT_RESUME);
345 /* Register vcpu_info area for CPU#0. */
350 xen_hvm_sysinit(void *arg __unused)
352 xen_hvm_init(XEN_HVM_INIT_COLD);
356 xen_set_vcpu_id(void)
361 if (!xen_hvm_domain())
364 /* Set vcpu_id to acpi_id */
367 pc->pc_vcpu_id = pc->pc_acpi_id;
369 printf("XEN: CPU %u has VCPU ID %u\n",
375 xen_hvm_cpu_init(void)
377 struct vcpu_register_vcpu_info info;
378 struct vcpu_info *vcpu_info;
384 if (DPCPU_GET(vcpu_info) != NULL) {
386 * vcpu_info is already set. We're resuming
387 * from a failed migration and our pre-suspend
388 * configuration is still valid.
393 vcpu_info = DPCPU_PTR(vcpu_local_info);
394 cpu = PCPU_GET(vcpu_id);
395 info.mfn = vtophys(vcpu_info) >> PAGE_SHIFT;
396 info.offset = vtophys(vcpu_info) - trunc_page(vtophys(vcpu_info));
398 rc = HYPERVISOR_vcpu_op(VCPUOP_register_vcpu_info, cpu, &info);
400 DPCPU_SET(vcpu_info, &HYPERVISOR_shared_info->vcpu_info[cpu]);
402 DPCPU_SET(vcpu_info, vcpu_info);
405 SYSINIT(xen_hvm_init, SI_SUB_HYPERVISOR, SI_ORDER_FIRST, xen_hvm_sysinit, NULL);
406 SYSINIT(xen_hvm_cpu_init, SI_SUB_INTR, SI_ORDER_FIRST, xen_hvm_cpu_init, NULL);
407 SYSINIT(xen_set_vcpu_id, SI_SUB_CPU, SI_ORDER_ANY, xen_set_vcpu_id, NULL);