2 * Copyright (c) 2011 NetApp, Inc.
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
8 * 1. Redistributions of source code must retain the above copyright
9 * notice, this list of conditions and the following disclaimer.
10 * 2. Redistributions in binary form must reproduce the above copyright
11 * notice, this list of conditions and the following disclaimer in the
12 * documentation and/or other materials provided with the distribution.
14 * THIS SOFTWARE IS PROVIDED BY NETAPP, INC ``AS IS'' AND
15 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
16 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
17 * ARE DISCLAIMED. IN NO EVENT SHALL NETAPP, INC OR CONTRIBUTORS BE LIABLE
18 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
19 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
20 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
21 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
22 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
23 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29 #include <sys/cdefs.h>
30 __FBSDID("$FreeBSD$");
32 #include <sys/param.h>
33 #include <sys/kernel.h>
34 #include <sys/queue.h>
36 #include <sys/mutex.h>
37 #include <sys/malloc.h>
39 #include <sys/sysctl.h>
40 #include <sys/libkern.h>
41 #include <sys/ioccom.h>
47 #include <vm/vm_map.h>
49 #include <machine/vmparam.h>
50 #include <machine/vmm.h>
51 #include <machine/vmm_instruction_emul.h>
52 #include <machine/vmm_dev.h>
54 #include "vmm_lapic.h"
58 #include "io/vatpic.h"
59 #include "io/vioapic.h"
64 struct vm *vm; /* vm instance cookie */
66 SLIST_ENTRY(vmmdev_softc) link;
69 #define VSC_LINKED 0x01
71 static SLIST_HEAD(, vmmdev_softc) head;
73 static struct mtx vmmdev_mtx;
75 static MALLOC_DEFINE(M_VMMDEV, "vmmdev", "vmmdev");
79 static struct vmmdev_softc *
80 vmmdev_lookup(const char *name)
82 struct vmmdev_softc *sc;
84 #ifdef notyet /* XXX kernel is not compiled with invariants */
85 mtx_assert(&vmmdev_mtx, MA_OWNED);
88 SLIST_FOREACH(sc, &head, link) {
89 if (strcmp(name, vm_name(sc->vm)) == 0)
96 static struct vmmdev_softc *
97 vmmdev_lookup2(struct cdev *cdev)
100 return (cdev->si_drv1);
104 vmmdev_rw(struct cdev *cdev, struct uio *uio, int flags)
106 int error, off, c, prot;
109 struct vmmdev_softc *sc;
111 static char zerobuf[PAGE_SIZE];
114 sc = vmmdev_lookup2(cdev);
118 prot = (uio->uio_rw == UIO_WRITE ? VM_PROT_WRITE : VM_PROT_READ);
119 while (uio->uio_resid > 0 && error == 0) {
120 gpa = uio->uio_offset;
121 off = gpa & PAGE_MASK;
122 c = min(uio->uio_resid, PAGE_SIZE - off);
125 * The VM has a hole in its physical memory map. If we want to
126 * use 'dd' to inspect memory beyond the hole we need to
127 * provide bogus data for memory that lies in the hole.
129 * Since this device does not support lseek(2), dd(1) will
130 * read(2) blocks of data to simulate the lseek(2).
132 hpa = vm_gpa_hold(sc->vm, gpa, c, prot, &cookie);
134 if (uio->uio_rw == UIO_READ)
135 error = uiomove(zerobuf, c, uio);
139 error = uiomove(hpa, c, uio);
140 vm_gpa_release(cookie);
147 vmmdev_ioctl(struct cdev *cdev, u_long cmd, caddr_t data, int fflag,
150 int error, vcpu, state_changed, size;
152 struct vmmdev_softc *sc;
153 struct vm_memory_segment *seg;
154 struct vm_register *vmreg;
155 struct vm_seg_desc *vmsegdesc;
156 struct vm_run *vmrun;
157 struct vm_exception *vmexc;
158 struct vm_lapic_irq *vmirq;
159 struct vm_lapic_msi *vmmsi;
160 struct vm_ioapic_irq *ioapic_irq;
161 struct vm_isa_irq *isa_irq;
162 struct vm_isa_irq_trigger *isa_irq_trigger;
163 struct vm_capability *vmcap;
164 struct vm_pptdev *pptdev;
165 struct vm_pptdev_mmio *pptmmio;
166 struct vm_pptdev_msi *pptmsi;
167 struct vm_pptdev_msix *pptmsix;
168 struct vm_nmi *vmnmi;
169 struct vm_stats *vmstats;
170 struct vm_stat_desc *statdesc;
171 struct vm_x2apic *x2apic;
172 struct vm_gpa_pte *gpapte;
173 struct vm_suspend *vmsuspend;
174 struct vm_gla2gpa *gg;
175 struct vm_activate_cpu *vac;
176 struct vm_cpuset *vm_cpuset;
177 struct vm_intinfo *vmii;
178 struct vm_rtc_time *rtctime;
179 struct vm_rtc_data *rtcdata;
181 sc = vmmdev_lookup2(cdev);
190 * Some VMM ioctls can operate only on vcpus that are not running.
194 case VM_GET_REGISTER:
195 case VM_SET_REGISTER:
196 case VM_GET_SEGMENT_DESCRIPTOR:
197 case VM_SET_SEGMENT_DESCRIPTOR:
198 case VM_INJECT_EXCEPTION:
199 case VM_GET_CAPABILITY:
200 case VM_SET_CAPABILITY:
203 case VM_SET_X2APIC_STATE:
205 case VM_ACTIVATE_CPU:
209 * XXX fragile, handle with care
210 * Assumes that the first field of the ioctl data is the vcpu.
213 if (vcpu < 0 || vcpu >= VM_MAXCPU) {
218 error = vcpu_set_state(sc->vm, vcpu, VCPU_FROZEN, true);
225 case VM_MAP_PPTDEV_MMIO:
227 case VM_UNBIND_PPTDEV:
231 * ioctls that operate on the entire virtual machine must
232 * prevent all vcpus from running.
235 for (vcpu = 0; vcpu < VM_MAXCPU; vcpu++) {
236 error = vcpu_set_state(sc->vm, vcpu, VCPU_FROZEN, true);
243 vcpu_set_state(sc->vm, vcpu, VCPU_IDLE, false);
256 vmrun = (struct vm_run *)data;
257 error = vm_run(sc->vm, vmrun);
260 vmsuspend = (struct vm_suspend *)data;
261 error = vm_suspend(sc->vm, vmsuspend->how);
264 error = vm_reinit(sc->vm);
267 statdesc = (struct vm_stat_desc *)data;
268 error = vmm_stat_desc_copy(statdesc->index,
269 statdesc->desc, sizeof(statdesc->desc));
273 CTASSERT(MAX_VM_STATS >= MAX_VMM_STAT_ELEMS);
274 vmstats = (struct vm_stats *)data;
275 getmicrotime(&vmstats->tv);
276 error = vmm_stat_copy(sc->vm, vmstats->cpuid,
277 &vmstats->num_entries, vmstats->statbuf);
281 pptmsi = (struct vm_pptdev_msi *)data;
282 error = ppt_setup_msi(sc->vm, pptmsi->vcpu,
283 pptmsi->bus, pptmsi->slot, pptmsi->func,
284 pptmsi->addr, pptmsi->msg,
288 pptmsix = (struct vm_pptdev_msix *)data;
289 error = ppt_setup_msix(sc->vm, pptmsix->vcpu,
290 pptmsix->bus, pptmsix->slot,
291 pptmsix->func, pptmsix->idx,
292 pptmsix->addr, pptmsix->msg,
293 pptmsix->vector_control);
295 case VM_MAP_PPTDEV_MMIO:
296 pptmmio = (struct vm_pptdev_mmio *)data;
297 error = ppt_map_mmio(sc->vm, pptmmio->bus, pptmmio->slot,
298 pptmmio->func, pptmmio->gpa, pptmmio->len,
302 pptdev = (struct vm_pptdev *)data;
303 error = vm_assign_pptdev(sc->vm, pptdev->bus, pptdev->slot,
306 case VM_UNBIND_PPTDEV:
307 pptdev = (struct vm_pptdev *)data;
308 error = vm_unassign_pptdev(sc->vm, pptdev->bus, pptdev->slot,
311 case VM_INJECT_EXCEPTION:
312 vmexc = (struct vm_exception *)data;
313 error = vm_inject_exception(sc->vm, vmexc->cpuid, vmexc);
316 vmnmi = (struct vm_nmi *)data;
317 error = vm_inject_nmi(sc->vm, vmnmi->cpuid);
320 vmirq = (struct vm_lapic_irq *)data;
321 error = lapic_intr_edge(sc->vm, vmirq->cpuid, vmirq->vector);
323 case VM_LAPIC_LOCAL_IRQ:
324 vmirq = (struct vm_lapic_irq *)data;
325 error = lapic_set_local_intr(sc->vm, vmirq->cpuid,
329 vmmsi = (struct vm_lapic_msi *)data;
330 error = lapic_intr_msi(sc->vm, vmmsi->addr, vmmsi->msg);
332 case VM_IOAPIC_ASSERT_IRQ:
333 ioapic_irq = (struct vm_ioapic_irq *)data;
334 error = vioapic_assert_irq(sc->vm, ioapic_irq->irq);
336 case VM_IOAPIC_DEASSERT_IRQ:
337 ioapic_irq = (struct vm_ioapic_irq *)data;
338 error = vioapic_deassert_irq(sc->vm, ioapic_irq->irq);
340 case VM_IOAPIC_PULSE_IRQ:
341 ioapic_irq = (struct vm_ioapic_irq *)data;
342 error = vioapic_pulse_irq(sc->vm, ioapic_irq->irq);
344 case VM_IOAPIC_PINCOUNT:
345 *(int *)data = vioapic_pincount(sc->vm);
347 case VM_ISA_ASSERT_IRQ:
348 isa_irq = (struct vm_isa_irq *)data;
349 error = vatpic_assert_irq(sc->vm, isa_irq->atpic_irq);
350 if (error == 0 && isa_irq->ioapic_irq != -1)
351 error = vioapic_assert_irq(sc->vm,
352 isa_irq->ioapic_irq);
354 case VM_ISA_DEASSERT_IRQ:
355 isa_irq = (struct vm_isa_irq *)data;
356 error = vatpic_deassert_irq(sc->vm, isa_irq->atpic_irq);
357 if (error == 0 && isa_irq->ioapic_irq != -1)
358 error = vioapic_deassert_irq(sc->vm,
359 isa_irq->ioapic_irq);
361 case VM_ISA_PULSE_IRQ:
362 isa_irq = (struct vm_isa_irq *)data;
363 error = vatpic_pulse_irq(sc->vm, isa_irq->atpic_irq);
364 if (error == 0 && isa_irq->ioapic_irq != -1)
365 error = vioapic_pulse_irq(sc->vm, isa_irq->ioapic_irq);
367 case VM_ISA_SET_IRQ_TRIGGER:
368 isa_irq_trigger = (struct vm_isa_irq_trigger *)data;
369 error = vatpic_set_irq_trigger(sc->vm,
370 isa_irq_trigger->atpic_irq, isa_irq_trigger->trigger);
373 seg = (struct vm_memory_segment *)data;
374 error = vm_malloc(sc->vm, seg->gpa, seg->len);
376 case VM_GET_MEMORY_SEG:
377 seg = (struct vm_memory_segment *)data;
379 (void)vm_gpabase2memseg(sc->vm, seg->gpa, seg);
382 case VM_GET_REGISTER:
383 vmreg = (struct vm_register *)data;
384 error = vm_get_register(sc->vm, vmreg->cpuid, vmreg->regnum,
387 case VM_SET_REGISTER:
388 vmreg = (struct vm_register *)data;
389 error = vm_set_register(sc->vm, vmreg->cpuid, vmreg->regnum,
392 case VM_SET_SEGMENT_DESCRIPTOR:
393 vmsegdesc = (struct vm_seg_desc *)data;
394 error = vm_set_seg_desc(sc->vm, vmsegdesc->cpuid,
398 case VM_GET_SEGMENT_DESCRIPTOR:
399 vmsegdesc = (struct vm_seg_desc *)data;
400 error = vm_get_seg_desc(sc->vm, vmsegdesc->cpuid,
404 case VM_GET_CAPABILITY:
405 vmcap = (struct vm_capability *)data;
406 error = vm_get_capability(sc->vm, vmcap->cpuid,
410 case VM_SET_CAPABILITY:
411 vmcap = (struct vm_capability *)data;
412 error = vm_set_capability(sc->vm, vmcap->cpuid,
416 case VM_SET_X2APIC_STATE:
417 x2apic = (struct vm_x2apic *)data;
418 error = vm_set_x2apic_state(sc->vm,
419 x2apic->cpuid, x2apic->state);
421 case VM_GET_X2APIC_STATE:
422 x2apic = (struct vm_x2apic *)data;
423 error = vm_get_x2apic_state(sc->vm,
424 x2apic->cpuid, &x2apic->state);
426 case VM_GET_GPA_PMAP:
427 gpapte = (struct vm_gpa_pte *)data;
428 pmap_get_mapping(vmspace_pmap(vm_get_vmspace(sc->vm)),
429 gpapte->gpa, gpapte->pte, &gpapte->ptenum);
432 case VM_GET_HPET_CAPABILITIES:
433 error = vhpet_getcap((struct vm_hpet_cap *)data);
436 CTASSERT(PROT_READ == VM_PROT_READ);
437 CTASSERT(PROT_WRITE == VM_PROT_WRITE);
438 CTASSERT(PROT_EXEC == VM_PROT_EXECUTE);
439 gg = (struct vm_gla2gpa *)data;
440 error = vmm_gla2gpa(sc->vm, gg->vcpuid, &gg->paging, gg->gla,
442 KASSERT(error == 0 || error == 1 || error == -1,
443 ("%s: vmm_gla2gpa unknown error %d", __func__, error));
446 * error = 0: the translation was successful
447 * error = 1: a fault was injected into the guest
456 case VM_ACTIVATE_CPU:
457 vac = (struct vm_activate_cpu *)data;
458 error = vm_activate_cpu(sc->vm, vac->vcpuid);
462 vm_cpuset = (struct vm_cpuset *)data;
463 size = vm_cpuset->cpusetsize;
464 if (size < sizeof(cpuset_t) || size > CPU_MAXSIZE / NBBY) {
468 cpuset = malloc(size, M_TEMP, M_WAITOK | M_ZERO);
469 if (vm_cpuset->which == VM_ACTIVE_CPUS)
470 *cpuset = vm_active_cpus(sc->vm);
471 else if (vm_cpuset->which == VM_SUSPENDED_CPUS)
472 *cpuset = vm_suspended_cpus(sc->vm);
476 error = copyout(cpuset, vm_cpuset->cpus, size);
477 free(cpuset, M_TEMP);
480 vmii = (struct vm_intinfo *)data;
481 error = vm_exit_intinfo(sc->vm, vmii->vcpuid, vmii->info1);
484 vmii = (struct vm_intinfo *)data;
485 error = vm_get_intinfo(sc->vm, vmii->vcpuid, &vmii->info1,
489 rtcdata = (struct vm_rtc_data *)data;
490 error = vrtc_nvram_write(sc->vm, rtcdata->offset,
494 rtcdata = (struct vm_rtc_data *)data;
495 error = vrtc_nvram_read(sc->vm, rtcdata->offset,
499 rtctime = (struct vm_rtc_time *)data;
500 error = vrtc_set_time(sc->vm, rtctime->secs);
504 rtctime = (struct vm_rtc_time *)data;
505 rtctime->secs = vrtc_get_time(sc->vm);
512 if (state_changed == 1) {
513 vcpu_set_state(sc->vm, vcpu, VCPU_IDLE, false);
514 } else if (state_changed == 2) {
515 for (vcpu = 0; vcpu < VM_MAXCPU; vcpu++)
516 vcpu_set_state(sc->vm, vcpu, VCPU_IDLE, false);
520 /* Make sure that no handler returns a bogus value like ERESTART */
521 KASSERT(error >= 0, ("vmmdev_ioctl: invalid error return %d", error));
526 vmmdev_mmap_single(struct cdev *cdev, vm_ooffset_t *offset,
527 vm_size_t size, struct vm_object **object, int nprot)
530 struct vmmdev_softc *sc;
532 sc = vmmdev_lookup2(cdev);
533 if (sc != NULL && (nprot & PROT_EXEC) == 0)
534 error = vm_get_memobj(sc->vm, *offset, size, offset, object);
542 vmmdev_destroy(void *arg)
545 struct vmmdev_softc *sc = arg;
547 if (sc->cdev != NULL)
548 destroy_dev(sc->cdev);
553 if ((sc->flags & VSC_LINKED) != 0) {
554 mtx_lock(&vmmdev_mtx);
555 SLIST_REMOVE(&head, sc, vmmdev_softc, link);
556 mtx_unlock(&vmmdev_mtx);
563 sysctl_vmm_destroy(SYSCTL_HANDLER_ARGS)
566 char buf[VM_MAX_NAMELEN];
567 struct vmmdev_softc *sc;
570 strlcpy(buf, "beavis", sizeof(buf));
571 error = sysctl_handle_string(oidp, buf, sizeof(buf), req);
572 if (error != 0 || req->newptr == NULL)
575 mtx_lock(&vmmdev_mtx);
576 sc = vmmdev_lookup(buf);
577 if (sc == NULL || sc->cdev == NULL) {
578 mtx_unlock(&vmmdev_mtx);
583 * The 'cdev' will be destroyed asynchronously when 'si_threadcount'
584 * goes down to 0 so we should not do it again in the callback.
588 mtx_unlock(&vmmdev_mtx);
591 * Schedule the 'cdev' to be destroyed:
593 * - any new operations on this 'cdev' will return an error (ENXIO).
595 * - when the 'si_threadcount' dwindles down to zero the 'cdev' will
596 * be destroyed and the callback will be invoked in a taskqueue
599 destroy_dev_sched_cb(cdev, vmmdev_destroy, sc);
603 SYSCTL_PROC(_hw_vmm, OID_AUTO, destroy, CTLTYPE_STRING | CTLFLAG_RW,
604 NULL, 0, sysctl_vmm_destroy, "A", NULL);
606 static struct cdevsw vmmdevsw = {
608 .d_version = D_VERSION,
609 .d_ioctl = vmmdev_ioctl,
610 .d_mmap_single = vmmdev_mmap_single,
612 .d_write = vmmdev_rw,
616 sysctl_vmm_create(SYSCTL_HANDLER_ARGS)
621 struct vmmdev_softc *sc, *sc2;
622 char buf[VM_MAX_NAMELEN];
624 strlcpy(buf, "beavis", sizeof(buf));
625 error = sysctl_handle_string(oidp, buf, sizeof(buf), req);
626 if (error != 0 || req->newptr == NULL)
629 mtx_lock(&vmmdev_mtx);
630 sc = vmmdev_lookup(buf);
631 mtx_unlock(&vmmdev_mtx);
635 error = vm_create(buf, &vm);
639 sc = malloc(sizeof(struct vmmdev_softc), M_VMMDEV, M_WAITOK | M_ZERO);
643 * Lookup the name again just in case somebody sneaked in when we
646 mtx_lock(&vmmdev_mtx);
647 sc2 = vmmdev_lookup(buf);
649 SLIST_INSERT_HEAD(&head, sc, link);
650 sc->flags |= VSC_LINKED;
652 mtx_unlock(&vmmdev_mtx);
659 error = make_dev_p(MAKEDEV_CHECKNAME, &cdev, &vmmdevsw, NULL,
660 UID_ROOT, GID_WHEEL, 0600, "vmm/%s", buf);
666 mtx_lock(&vmmdev_mtx);
668 sc->cdev->si_drv1 = sc;
669 mtx_unlock(&vmmdev_mtx);
673 SYSCTL_PROC(_hw_vmm, OID_AUTO, create, CTLTYPE_STRING | CTLFLAG_RW,
674 NULL, 0, sysctl_vmm_create, "A", NULL);
679 mtx_init(&vmmdev_mtx, "vmm device mutex", NULL, MTX_DEF);
687 if (SLIST_EMPTY(&head))