vmm_dev.c revision 240922 
1/*-
2 * Copyright (c) 2011 NetApp, Inc.
3 * All rights reserved.
4 *
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
7 * are met:
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.
13 *
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
24 * SUCH DAMAGE.
25 *
26 * $FreeBSD$
27 */
28
29#include <sys/cdefs.h>
30__FBSDID("$FreeBSD$");
31
32#include <sys/param.h>
33#include <sys/kernel.h>
34#include <sys/queue.h>
35#include <sys/lock.h>
36#include <sys/mutex.h>
37#include <sys/malloc.h>
38#include <sys/conf.h>
39#include <sys/sysctl.h>
40#include <sys/libkern.h>
41#include <sys/ioccom.h>
42#include <sys/mman.h>
43#include <sys/uio.h>
44
45#include <vm/vm.h>
46#include <vm/pmap.h>
47
48#include <machine/pmap.h>
49#include <machine/vmparam.h>
50
51#include <machine/vmm.h>
52#include "vmm_lapic.h"
53#include "vmm_stat.h"
54#include "vmm_mem.h"
55#include "io/ppt.h"
56#include <machine/vmm_dev.h>
57
58struct vmmdev_softc {
59	struct vm	*vm;		/* vm instance cookie */
60	struct cdev	*cdev;
61	SLIST_ENTRY(vmmdev_softc) link;
62};
63static SLIST_HEAD(, vmmdev_softc) head;
64
65static struct mtx vmmdev_mtx;
66
67static MALLOC_DEFINE(M_VMMDEV, "vmmdev", "vmmdev");
68
69SYSCTL_DECL(_hw_vmm);
70
71static struct vmmdev_softc *
72vmmdev_lookup(const char *name)
73{
74	struct vmmdev_softc *sc;
75
76#ifdef notyet	/* XXX kernel is not compiled with invariants */
77	mtx_assert(&vmmdev_mtx, MA_OWNED);
78#endif
79
80	SLIST_FOREACH(sc, &head, link) {
81		if (strcmp(name, vm_name(sc->vm)) == 0)
82			break;
83	}
84
85	return (sc);
86}
87
88static struct vmmdev_softc *
89vmmdev_lookup2(struct cdev *cdev)
90{
91	struct vmmdev_softc *sc;
92
93#ifdef notyet	/* XXX kernel is not compiled with invariants */
94	mtx_assert(&vmmdev_mtx, MA_OWNED);
95#endif
96
97	SLIST_FOREACH(sc, &head, link) {
98		if (sc->cdev == cdev)
99			break;
100	}
101
102	return (sc);
103}
104
105static int
106vmmdev_rw(struct cdev *cdev, struct uio *uio, int flags)
107{
108	int error, off, c;
109	vm_paddr_t hpa, gpa;
110	struct vmmdev_softc *sc;
111
112	static char zerobuf[PAGE_SIZE];
113
114	error = 0;
115	mtx_lock(&vmmdev_mtx);
116	sc = vmmdev_lookup2(cdev);
117
118	while (uio->uio_resid > 0 && error == 0) {
119		gpa = uio->uio_offset;
120		off = gpa & PAGE_MASK;
121		c = min(uio->uio_resid, PAGE_SIZE - off);
122
123		/*
124		 * The VM has a hole in its physical memory map. If we want to
125		 * use 'dd' to inspect memory beyond the hole we need to
126		 * provide bogus data for memory that lies in the hole.
127		 *
128		 * Since this device does not support lseek(2), dd(1) will
129		 * read(2) blocks of data to simulate the lseek(2).
130		 */
131		hpa = vm_gpa2hpa(sc->vm, gpa, c);
132		if (hpa == (vm_paddr_t)-1) {
133			if (uio->uio_rw == UIO_READ)
134				error = uiomove(zerobuf, c, uio);
135			else
136				error = EFAULT;
137		} else
138			error = uiomove((void *)PHYS_TO_DMAP(hpa), c, uio);
139	}
140
141	mtx_unlock(&vmmdev_mtx);
142	return (error);
143}
144
145static int
146vmmdev_ioctl(struct cdev *cdev, u_long cmd, caddr_t data, int fflag,
147	     struct thread *td)
148{
149	int error, vcpu;
150	struct vmmdev_softc *sc;
151	struct vm_memory_segment *seg;
152	struct vm_register *vmreg;
153	struct vm_seg_desc* vmsegdesc;
154	struct vm_pin *vmpin;
155	struct vm_run *vmrun;
156	struct vm_event *vmevent;
157	struct vm_lapic_irq *vmirq;
158	struct vm_capability *vmcap;
159	struct vm_pptdev *pptdev;
160	struct vm_pptdev_mmio *pptmmio;
161	struct vm_pptdev_msi *pptmsi;
162	struct vm_pptdev_msix *pptmsix;
163	struct vm_nmi *vmnmi;
164	struct vm_stats *vmstats;
165	struct vm_stat_desc *statdesc;
166	struct vm_x2apic *x2apic;
167
168	mtx_lock(&vmmdev_mtx);
169	sc = vmmdev_lookup2(cdev);
170	if (sc == NULL) {
171		mtx_unlock(&vmmdev_mtx);
172		return (ENXIO);
173	}
174
175	/*
176	 * Some VMM ioctls can operate only on vcpus that are not running.
177	 */
178	switch (cmd) {
179	case VM_RUN:
180	case VM_SET_PINNING:
181	case VM_GET_REGISTER:
182	case VM_SET_REGISTER:
183	case VM_GET_SEGMENT_DESCRIPTOR:
184	case VM_SET_SEGMENT_DESCRIPTOR:
185	case VM_INJECT_EVENT:
186	case VM_GET_CAPABILITY:
187	case VM_SET_CAPABILITY:
188	case VM_PPTDEV_MSI:
189	case VM_SET_X2APIC_STATE:
190		/*
191		 * XXX fragile, handle with care
192		 * Assumes that the first field of the ioctl data is the vcpu.
193		 */
194		vcpu = *(int *)data;
195		if (vcpu < 0 || vcpu >= VM_MAXCPU) {
196			error = EINVAL;
197			goto done;
198		}
199
200		if (vcpu_is_running(sc->vm, vcpu, NULL)) {
201			error = EBUSY;
202			goto done;
203		}
204		break;
205	default:
206		break;
207	}
208
209	switch(cmd) {
210	case VM_RUN:
211		vmrun = (struct vm_run *)data;
212
213		vm_set_run_state(sc->vm, vmrun->cpuid, VCPU_RUNNING);
214		mtx_unlock(&vmmdev_mtx);
215
216		error = vm_run(sc->vm, vmrun);
217
218		mtx_lock(&vmmdev_mtx);
219		vm_set_run_state(sc->vm, vmrun->cpuid, VCPU_STOPPED);
220		break;
221	case VM_STAT_DESC: {
222		const char *desc;
223		statdesc = (struct vm_stat_desc *)data;
224		desc = vmm_stat_desc(statdesc->index);
225		if (desc != NULL) {
226			error = 0;
227			strlcpy(statdesc->desc, desc, sizeof(statdesc->desc));
228		} else
229			error = EINVAL;
230		break;
231	}
232	case VM_STATS: {
233		CTASSERT(MAX_VM_STATS >= MAX_VMM_STAT_TYPES);
234		vmstats = (struct vm_stats *)data;
235		getmicrotime(&vmstats->tv);
236		error = vmm_stat_copy(sc->vm, vmstats->cpuid,
237				      &vmstats->num_entries, vmstats->statbuf);
238		break;
239	}
240	case VM_PPTDEV_MSI:
241		pptmsi = (struct vm_pptdev_msi *)data;
242		error = ppt_setup_msi(sc->vm, pptmsi->vcpu,
243				      pptmsi->bus, pptmsi->slot, pptmsi->func,
244				      pptmsi->destcpu, pptmsi->vector,
245				      pptmsi->numvec);
246		break;
247	case VM_PPTDEV_MSIX:
248		pptmsix = (struct vm_pptdev_msix *)data;
249		error = ppt_setup_msix(sc->vm, pptmsix->vcpu,
250				       pptmsix->bus, pptmsix->slot,
251				       pptmsix->func, pptmsix->idx,
252				       pptmsix->msg, pptmsix->vector_control,
253				       pptmsix->addr);
254		break;
255	case VM_MAP_PPTDEV_MMIO:
256		pptmmio = (struct vm_pptdev_mmio *)data;
257		error = ppt_map_mmio(sc->vm, pptmmio->bus, pptmmio->slot,
258				     pptmmio->func, pptmmio->gpa, pptmmio->len,
259				     pptmmio->hpa);
260		break;
261	case VM_BIND_PPTDEV:
262		pptdev = (struct vm_pptdev *)data;
263		error = ppt_assign_device(sc->vm, pptdev->bus, pptdev->slot,
264					  pptdev->func);
265		break;
266	case VM_UNBIND_PPTDEV:
267		pptdev = (struct vm_pptdev *)data;
268		error = ppt_unassign_device(sc->vm, pptdev->bus, pptdev->slot,
269					    pptdev->func);
270		break;
271	case VM_INJECT_EVENT:
272		vmevent = (struct vm_event *)data;
273		error = vm_inject_event(sc->vm, vmevent->cpuid, vmevent->type,
274					vmevent->vector,
275					vmevent->error_code,
276					vmevent->error_code_valid);
277		break;
278	case VM_INJECT_NMI:
279		vmnmi = (struct vm_nmi *)data;
280		error = vm_inject_nmi(sc->vm, vmnmi->cpuid);
281		break;
282	case VM_LAPIC_IRQ:
283		vmirq = (struct vm_lapic_irq *)data;
284		error = lapic_set_intr(sc->vm, vmirq->cpuid, vmirq->vector);
285		break;
286	case VM_SET_PINNING:
287		vmpin = (struct vm_pin *)data;
288		error = vm_set_pinning(sc->vm, vmpin->vm_cpuid,
289				       vmpin->host_cpuid);
290		break;
291	case VM_GET_PINNING:
292		vmpin = (struct vm_pin *)data;
293		error = vm_get_pinning(sc->vm, vmpin->vm_cpuid,
294				       &vmpin->host_cpuid);
295		break;
296	case VM_MAP_MEMORY:
297		seg = (struct vm_memory_segment *)data;
298		error = vm_malloc(sc->vm, seg->gpa, seg->len, &seg->hpa);
299		break;
300	case VM_GET_MEMORY_SEG:
301		seg = (struct vm_memory_segment *)data;
302		seg->hpa = seg->len = 0;
303		(void)vm_gpabase2memseg(sc->vm, seg->gpa, seg);
304		error = 0;
305		break;
306	case VM_GET_REGISTER:
307		vmreg = (struct vm_register *)data;
308		error = vm_get_register(sc->vm, vmreg->cpuid, vmreg->regnum,
309					&vmreg->regval);
310		break;
311	case VM_SET_REGISTER:
312		vmreg = (struct vm_register *)data;
313		error = vm_set_register(sc->vm, vmreg->cpuid, vmreg->regnum,
314					vmreg->regval);
315		break;
316	case VM_SET_SEGMENT_DESCRIPTOR:
317		vmsegdesc = (struct vm_seg_desc *)data;
318		error = vm_set_seg_desc(sc->vm, vmsegdesc->cpuid,
319					vmsegdesc->regnum,
320					&vmsegdesc->desc);
321		break;
322	case VM_GET_SEGMENT_DESCRIPTOR:
323		vmsegdesc = (struct vm_seg_desc *)data;
324		error = vm_get_seg_desc(sc->vm, vmsegdesc->cpuid,
325					vmsegdesc->regnum,
326					&vmsegdesc->desc);
327		break;
328	case VM_GET_CAPABILITY:
329		vmcap = (struct vm_capability *)data;
330		error = vm_get_capability(sc->vm, vmcap->cpuid,
331					  vmcap->captype,
332					  &vmcap->capval);
333		break;
334	case VM_SET_CAPABILITY:
335		vmcap = (struct vm_capability *)data;
336		error = vm_set_capability(sc->vm, vmcap->cpuid,
337					  vmcap->captype,
338					  vmcap->capval);
339		break;
340	case VM_SET_X2APIC_STATE:
341		x2apic = (struct vm_x2apic *)data;
342		error = vm_set_x2apic_state(sc->vm,
343					    x2apic->cpuid, x2apic->state);
344		break;
345	case VM_GET_X2APIC_STATE:
346		x2apic = (struct vm_x2apic *)data;
347		error = vm_get_x2apic_state(sc->vm,
348					    x2apic->cpuid, &x2apic->state);
349		break;
350	default:
351		error = ENOTTY;
352		break;
353	}
354done:
355	mtx_unlock(&vmmdev_mtx);
356
357	return (error);
358}
359
360static int
361vmmdev_mmap(struct cdev *cdev, vm_ooffset_t offset, vm_paddr_t *paddr,
362    int nprot, vm_memattr_t *memattr)
363{
364	int error;
365	struct vmmdev_softc *sc;
366
367	error = -1;
368	mtx_lock(&vmmdev_mtx);
369
370	sc = vmmdev_lookup2(cdev);
371	if (sc != NULL && (nprot & PROT_EXEC) == 0) {
372		*paddr = vm_gpa2hpa(sc->vm, (vm_paddr_t)offset, PAGE_SIZE);
373		if (*paddr != (vm_paddr_t)-1)
374			error = 0;
375	}
376
377	mtx_unlock(&vmmdev_mtx);
378
379	return (error);
380}
381
382static void
383vmmdev_destroy(struct vmmdev_softc *sc)
384{
385
386#ifdef notyet	/* XXX kernel is not compiled with invariants */
387	mtx_assert(&vmmdev_mtx, MA_OWNED);
388#endif
389
390	/*
391	 * XXX must stop virtual machine instances that may be still
392	 * running and cleanup their state.
393	 */
394	SLIST_REMOVE(&head, sc, vmmdev_softc, link);
395	destroy_dev(sc->cdev);
396	vm_destroy(sc->vm);
397	free(sc, M_VMMDEV);
398}
399
400static int
401sysctl_vmm_destroy(SYSCTL_HANDLER_ARGS)
402{
403	int error;
404	char buf[VM_MAX_NAMELEN];
405	struct vmmdev_softc *sc;
406
407	strlcpy(buf, "beavis", sizeof(buf));
408	error = sysctl_handle_string(oidp, buf, sizeof(buf), req);
409	if (error != 0 || req->newptr == NULL)
410		return (error);
411
412	mtx_lock(&vmmdev_mtx);
413	sc = vmmdev_lookup(buf);
414	if (sc == NULL) {
415		mtx_unlock(&vmmdev_mtx);
416		return (EINVAL);
417	}
418	vmmdev_destroy(sc);
419	mtx_unlock(&vmmdev_mtx);
420	return (0);
421}
422SYSCTL_PROC(_hw_vmm, OID_AUTO, destroy, CTLTYPE_STRING | CTLFLAG_RW,
423	    NULL, 0, sysctl_vmm_destroy, "A", NULL);
424
425static struct cdevsw vmmdevsw = {
426	.d_name		= "vmmdev",
427	.d_version	= D_VERSION,
428	.d_ioctl	= vmmdev_ioctl,
429	.d_mmap		= vmmdev_mmap,
430	.d_read		= vmmdev_rw,
431	.d_write	= vmmdev_rw,
432};
433
434static int
435sysctl_vmm_create(SYSCTL_HANDLER_ARGS)
436{
437	int error;
438	struct vm *vm;
439	struct vmmdev_softc *sc;
440	char buf[VM_MAX_NAMELEN];
441
442	strlcpy(buf, "beavis", sizeof(buf));
443	error = sysctl_handle_string(oidp, buf, sizeof(buf), req);
444	if (error != 0 || req->newptr == NULL)
445		return (error);
446
447	mtx_lock(&vmmdev_mtx);
448
449	sc = vmmdev_lookup(buf);
450	if (sc != NULL) {
451		mtx_unlock(&vmmdev_mtx);
452		return (EEXIST);
453	}
454
455	vm = vm_create(buf);
456	if (vm == NULL) {
457		mtx_unlock(&vmmdev_mtx);
458		return (EINVAL);
459	}
460
461	sc = malloc(sizeof(struct vmmdev_softc), M_VMMDEV, M_WAITOK | M_ZERO);
462	sc->vm = vm;
463	sc->cdev = make_dev(&vmmdevsw, 0, UID_ROOT, GID_WHEEL, 0600,
464			    "vmm/%s", buf);
465	sc->cdev->si_drv1 = sc;
466	SLIST_INSERT_HEAD(&head, sc, link);
467
468	mtx_unlock(&vmmdev_mtx);
469	return (0);
470}
471SYSCTL_PROC(_hw_vmm, OID_AUTO, create, CTLTYPE_STRING | CTLFLAG_RW,
472	    NULL, 0, sysctl_vmm_create, "A", NULL);
473
474static int
475sysctl_vmm_mem_total(SYSCTL_HANDLER_ARGS)
476{
477	size_t val = vmm_mem_get_mem_total();
478	return sysctl_handle_long(oidp, &val, 0, req);
479}
480SYSCTL_PROC(_hw_vmm, OID_AUTO, mem_total, CTLTYPE_LONG | CTLFLAG_RD,
481		0, 0, sysctl_vmm_mem_total, "LU", "Amount of Total memory");
482
483static int
484sysctl_vmm_mem_free(SYSCTL_HANDLER_ARGS)
485{
486	size_t val = vmm_mem_get_mem_free();
487	return sysctl_handle_long(oidp, &val, 0, req);
488}
489SYSCTL_PROC(_hw_vmm, OID_AUTO, mem_free, CTLTYPE_LONG | CTLFLAG_RD,
490		0, 0, sysctl_vmm_mem_free, "LU", "Amount of Free memory");
491
492void
493vmmdev_init(void)
494{
495	mtx_init(&vmmdev_mtx, "vmm device mutex", NULL, MTX_DEF);
496}
497
498void
499vmmdev_cleanup(void)
500{
501	struct vmmdev_softc *sc, *sc2;
502
503	mtx_lock(&vmmdev_mtx);
504
505	SLIST_FOREACH_SAFE(sc, &head, link, sc2)
506		vmmdev_destroy(sc);
507
508	mtx_unlock(&vmmdev_mtx);
509}
510