summaryrefslogtreecommitdiffstats
path: root/usr.sbin/bhyve/virtio.c
blob: 41a9e428980d07e85bd1f8ba6e8b9229555e1695 (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
/*-
 * Copyright (c) 2013  Chris Torek <torek @ torek net>
 * 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/uio.h>

#include <stdio.h>
#include <stdint.h>
#include <pthread.h>
#include <pthread_np.h>

#include "bhyverun.h"
#include "pci_emul.h"
#include "virtio.h"

/*
 * Functions for dealing with generalized "virtual devices" as
 * defined by <https://www.google.com/#output=search&q=virtio+spec>
 */

/*
 * In case we decide to relax the "virtio softc comes at the
 * front of virtio-based device softc" constraint, let's use
 * this to convert.
 */
#define DEV_SOFTC(vs) ((void *)(vs))

/*
 * Link a virtio_softc to its constants, the device softc, and
 * the PCI emulation.
 */
void
vi_softc_linkup(struct virtio_softc *vs, struct virtio_consts *vc,
		void *dev_softc, struct pci_devinst *pi,
		struct vqueue_info *queues)
{
	int i;

	/* vs and dev_softc addresses must match */
	assert((void *)vs == dev_softc);
	vs->vs_vc = vc;
	vs->vs_pi = pi;
	pi->pi_arg = vs;

	vs->vs_queues = queues;
	for (i = 0; i < vc->vc_nvq; i++) {
		queues[i].vq_vs = vs;
		queues[i].vq_num = i;
	}
}

/*
 * Reset device (device-wide).  This erases all queues, i.e.,
 * all the queues become invalid (though we don't wipe out the
 * internal pointers, we just clear the VQ_ALLOC flag).
 *
 * It resets negotiated features to "none".
 *
 * If MSI-X is enabled, this also resets all the vectors to NO_VECTOR.
 */
void
vi_reset_dev(struct virtio_softc *vs)
{
	struct vqueue_info *vq;
	int i, nvq;

	if (vs->vs_mtx)
		assert(pthread_mutex_isowned_np(vs->vs_mtx));

	nvq = vs->vs_vc->vc_nvq;
	for (vq = vs->vs_queues, i = 0; i < nvq; vq++, i++) {
		vq->vq_flags = 0;
		vq->vq_last_avail = 0;
		vq->vq_save_used = 0;
		vq->vq_pfn = 0;
		vq->vq_msix_idx = VIRTIO_MSI_NO_VECTOR;
	}
	vs->vs_negotiated_caps = 0;
	vs->vs_curq = 0;
	/* vs->vs_status = 0; -- redundant */
	if (vs->vs_isr)
		pci_lintr_deassert(vs->vs_pi);
	vs->vs_isr = 0;
	vs->vs_msix_cfg_idx = VIRTIO_MSI_NO_VECTOR;
}

/*
 * Set I/O BAR (usually 0) to map PCI config registers.
 */
void
vi_set_io_bar(struct virtio_softc *vs, int barnum)
{
	size_t size;

	/*
	 * ??? should we use CFG0 if MSI-X is disabled?
	 * Existing code did not...
	 */
	size = VTCFG_R_CFG1 + vs->vs_vc->vc_cfgsize;
	pci_emul_alloc_bar(vs->vs_pi, barnum, PCIBAR_IO, size);
}

/*
 * Initialize MSI-X vector capabilities if we're to use MSI-X,
 * or MSI capabilities if not.
 *
 * We assume we want one MSI-X vector per queue, here, plus one
 * for the config vec.
 */
int
vi_intr_init(struct virtio_softc *vs, int barnum, int use_msix)
{
	int nvec;

	if (use_msix) {
		vs->vs_flags |= VIRTIO_USE_MSIX;
		VS_LOCK(vs);
		vi_reset_dev(vs); /* set all vectors to NO_VECTOR */
		VS_UNLOCK(vs);
		nvec = vs->vs_vc->vc_nvq + 1;
		if (pci_emul_add_msixcap(vs->vs_pi, nvec, barnum))
			return (1);
	} else
		vs->vs_flags &= ~VIRTIO_USE_MSIX;

	/* Only 1 MSI vector for bhyve */
	pci_emul_add_msicap(vs->vs_pi, 1);

	/* Legacy interrupts are mandatory for virtio devices */
	pci_lintr_request(vs->vs_pi);

	return (0);
}

/*
 * Initialize the currently-selected virtio queue (vs->vs_curq).
 * The guest just gave us a page frame number, from which we can
 * calculate the addresses of the queue.
 */
void
vi_vq_init(struct virtio_softc *vs, uint32_t pfn)
{
	struct vqueue_info *vq;
	uint64_t phys;
	size_t size;
	char *base;

	vq = &vs->vs_queues[vs->vs_curq];
	vq->vq_pfn = pfn;
	phys = (uint64_t)pfn << VRING_PFN;
	size = vring_size(vq->vq_qsize);
	base = paddr_guest2host(vs->vs_pi->pi_vmctx, phys, size);

	/* First page(s) are descriptors... */
	vq->vq_desc = (struct virtio_desc *)base;
	base += vq->vq_qsize * sizeof(struct virtio_desc);

	/* ... immediately followed by "avail" ring (entirely uint16_t's) */
	vq->vq_avail = (struct vring_avail *)base;
	base += (2 + vq->vq_qsize + 1) * sizeof(uint16_t);

	/* Then it's rounded up to the next page... */
	base = (char *)roundup2((uintptr_t)base, VRING_ALIGN);

	/* ... and the last page(s) are the used ring. */
	vq->vq_used = (struct vring_used *)base;

	/* Mark queue as allocated, and start at 0 when we use it. */
	vq->vq_flags = VQ_ALLOC;
	vq->vq_last_avail = 0;
	vq->vq_save_used = 0;
}

/*
 * Helper inline for vq_getchain(): record the i'th "real"
 * descriptor.
 */
static inline void
_vq_record(int i, volatile struct virtio_desc *vd, struct vmctx *ctx,
	   struct iovec *iov, int n_iov, uint16_t *flags) {

	if (i >= n_iov)
		return;
	iov[i].iov_base = paddr_guest2host(ctx, vd->vd_addr, vd->vd_len);
	iov[i].iov_len = vd->vd_len;
	if (flags != NULL)
		flags[i] = vd->vd_flags;
}
#define	VQ_MAX_DESCRIPTORS	512	/* see below */

/*
 * Examine the chain of descriptors starting at the "next one" to
 * make sure that they describe a sensible request.  If so, return
 * the number of "real" descriptors that would be needed/used in
 * acting on this request.  This may be smaller than the number of
 * available descriptors, e.g., if there are two available but
 * they are two separate requests, this just returns 1.  Or, it
 * may be larger: if there are indirect descriptors involved,
 * there may only be one descriptor available but it may be an
 * indirect pointing to eight more.  We return 8 in this case,
 * i.e., we do not count the indirect descriptors, only the "real"
 * ones.
 *
 * Basically, this vets the vd_flags and vd_next field of each
 * descriptor and tells you how many are involved.  Since some may
 * be indirect, this also needs the vmctx (in the pci_devinst
 * at vs->vs_pi) so that it can find indirect descriptors.
 *
 * As we process each descriptor, we copy and adjust it (guest to
 * host address wise, also using the vmtctx) into the given iov[]
 * array (of the given size).  If the array overflows, we stop
 * placing values into the array but keep processing descriptors,
 * up to VQ_MAX_DESCRIPTORS, before giving up and returning -1.
 * So you, the caller, must not assume that iov[] is as big as the
 * return value (you can process the same thing twice to allocate
 * a larger iov array if needed, or supply a zero length to find
 * out how much space is needed).
 *
 * If you want to verify the WRITE flag on each descriptor, pass a
 * non-NULL "flags" pointer to an array of "uint16_t" of the same size
 * as n_iov and we'll copy each vd_flags field after unwinding any
 * indirects.
 *
 * If some descriptor(s) are invalid, this prints a diagnostic message
 * and returns -1.  If no descriptors are ready now it simply returns 0.
 *
 * You are assumed to have done a vq_ring_ready() if needed (note
 * that vq_has_descs() does one).
 */
int
vq_getchain(struct vqueue_info *vq, uint16_t *pidx,
	    struct iovec *iov, int n_iov, uint16_t *flags)
{
	int i;
	u_int ndesc, n_indir;
	u_int idx, next;
	volatile struct virtio_desc *vdir, *vindir, *vp;
	struct vmctx *ctx;
	struct virtio_softc *vs;
	const char *name;

	vs = vq->vq_vs;
	name = vs->vs_vc->vc_name;

	/*
	 * Note: it's the responsibility of the guest not to
	 * update vq->vq_avail->va_idx until all of the descriptors
         * the guest has written are valid (including all their
         * vd_next fields and vd_flags).
	 *
	 * Compute (last_avail - va_idx) in integers mod 2**16.  This is
	 * the number of descriptors the device has made available
	 * since the last time we updated vq->vq_last_avail.
	 *
	 * We just need to do the subtraction as an unsigned int,
	 * then trim off excess bits.
	 */
	idx = vq->vq_last_avail;
	ndesc = (uint16_t)((u_int)vq->vq_avail->va_idx - idx);
	if (ndesc == 0)
		return (0);
	if (ndesc > vq->vq_qsize) {
		/* XXX need better way to diagnose issues */
		fprintf(stderr,
		    "%s: ndesc (%u) out of range, driver confused?\r\n",
		    name, (u_int)ndesc);
		return (-1);
	}

	/*
	 * Now count/parse "involved" descriptors starting from
	 * the head of the chain.
	 *
	 * To prevent loops, we could be more complicated and
	 * check whether we're re-visiting a previously visited
	 * index, but we just abort if the count gets excessive.
	 */
	ctx = vs->vs_pi->pi_vmctx;
	*pidx = next = vq->vq_avail->va_ring[idx & (vq->vq_qsize - 1)];
	vq->vq_last_avail++;
	for (i = 0; i < VQ_MAX_DESCRIPTORS; next = vdir->vd_next) {
		if (next >= vq->vq_qsize) {
			fprintf(stderr,
			    "%s: descriptor index %u out of range, "
			    "driver confused?\r\n",
			    name, next);
			return (-1);
		}
		vdir = &vq->vq_desc[next];
		if ((vdir->vd_flags & VRING_DESC_F_INDIRECT) == 0) {
			_vq_record(i, vdir, ctx, iov, n_iov, flags);
			i++;
		} else if ((vs->vs_negotiated_caps &
		    VIRTIO_RING_F_INDIRECT_DESC) == 0) {
			fprintf(stderr,
			    "%s: descriptor has forbidden INDIRECT flag, "
			    "driver confused?\r\n",
			    name);
			return (-1);
		} else {
			n_indir = vdir->vd_len / 16;
			if ((vdir->vd_len & 0xf) || n_indir == 0) {
				fprintf(stderr,
				    "%s: invalid indir len 0x%x, "
				    "driver confused?\r\n",
				    name, (u_int)vdir->vd_len);
				return (-1);
			}
			vindir = paddr_guest2host(ctx,
			    vdir->vd_addr, vdir->vd_len);
			/*
			 * Indirects start at the 0th, then follow
			 * their own embedded "next"s until those run
			 * out.  Each one's indirect flag must be off
			 * (we don't really have to check, could just
			 * ignore errors...).
			 */
			next = 0;
			for (;;) {
				vp = &vindir[next];
				if (vp->vd_flags & VRING_DESC_F_INDIRECT) {
					fprintf(stderr,
					    "%s: indirect desc has INDIR flag,"
					    " driver confused?\r\n",
					    name);
					return (-1);
				}
				_vq_record(i, vp, ctx, iov, n_iov, flags);
				if (++i > VQ_MAX_DESCRIPTORS)
					goto loopy;
				if ((vp->vd_flags & VRING_DESC_F_NEXT) == 0)
					break;
				next = vp->vd_next;
				if (next >= n_indir) {
					fprintf(stderr,
					    "%s: invalid next %u > %u, "
					    "driver confused?\r\n",
					    name, (u_int)next, n_indir);
					return (-1);
				}
			}
		}
		if ((vdir->vd_flags & VRING_DESC_F_NEXT) == 0)
			return (i);
	}
loopy:
	fprintf(stderr,
	    "%s: descriptor loop? count > %d - driver confused?\r\n",
	    name, i);
	return (-1);
}

/*
 * Return the currently-first request chain back to the available queue.
 *
 * (This chain is the one you handled when you called vq_getchain()
 * and used its positive return value.)
 */
void
vq_retchain(struct vqueue_info *vq)
{

	vq->vq_last_avail--;
}

/*
 * Return specified request chain to the guest, setting its I/O length
 * to the provided value.
 *
 * (This chain is the one you handled when you called vq_getchain()
 * and used its positive return value.)
 */
void
vq_relchain(struct vqueue_info *vq, uint16_t idx, uint32_t iolen)
{
	uint16_t uidx, mask;
	volatile struct vring_used *vuh;
	volatile struct virtio_used *vue;

	/*
	 * Notes:
	 *  - mask is N-1 where N is a power of 2 so computes x % N
	 *  - vuh points to the "used" data shared with guest
	 *  - vue points to the "used" ring entry we want to update
	 *  - head is the same value we compute in vq_iovecs().
	 *
	 * (I apologize for the two fields named vu_idx; the
	 * virtio spec calls the one that vue points to, "id"...)
	 */
	mask = vq->vq_qsize - 1;
	vuh = vq->vq_used;

	uidx = vuh->vu_idx;
	vue = &vuh->vu_ring[uidx++ & mask];
	vue->vu_idx = idx;
	vue->vu_tlen = iolen;
	vuh->vu_idx = uidx;
}

/*
 * Driver has finished processing "available" chains and calling
 * vq_relchain on each one.  If driver used all the available
 * chains, used_all should be set.
 *
 * If the "used" index moved we may need to inform the guest, i.e.,
 * deliver an interrupt.  Even if the used index did NOT move we
 * may need to deliver an interrupt, if the avail ring is empty and
 * we are supposed to interrupt on empty.
 *
 * Note that used_all_avail is provided by the caller because it's
 * a snapshot of the ring state when he decided to finish interrupt
 * processing -- it's possible that descriptors became available after
 * that point.  (It's also typically a constant 1/True as well.)
 */
void
vq_endchains(struct vqueue_info *vq, int used_all_avail)
{
	struct virtio_softc *vs;
	uint16_t event_idx, new_idx, old_idx;
	int intr;

	/*
	 * Interrupt generation: if we're using EVENT_IDX,
	 * interrupt if we've crossed the event threshold.
	 * Otherwise interrupt is generated if we added "used" entries,
	 * but suppressed by VRING_AVAIL_F_NO_INTERRUPT.
	 *
	 * In any case, though, if NOTIFY_ON_EMPTY is set and the
	 * entire avail was processed, we need to interrupt always.
	 */
	vs = vq->vq_vs;
	old_idx = vq->vq_save_used;
	vq->vq_save_used = new_idx = vq->vq_used->vu_idx;
	if (used_all_avail &&
	    (vs->vs_negotiated_caps & VIRTIO_F_NOTIFY_ON_EMPTY))
		intr = 1;
	else if (vs->vs_negotiated_caps & VIRTIO_RING_F_EVENT_IDX) {
		event_idx = VQ_USED_EVENT_IDX(vq);
		/*
		 * This calculation is per docs and the kernel
		 * (see src/sys/dev/virtio/virtio_ring.h).
		 */
		intr = (uint16_t)(new_idx - event_idx - 1) <
			(uint16_t)(new_idx - old_idx);
	} else {
		intr = new_idx != old_idx &&
		    !(vq->vq_avail->va_flags & VRING_AVAIL_F_NO_INTERRUPT);
	}
	if (intr)
		vq_interrupt(vs, vq);
}

/* Note: these are in sorted order to make for a fast search */
static struct config_reg {
	uint16_t	cr_offset;	/* register offset */
	uint8_t		cr_size;	/* size (bytes) */
	uint8_t		cr_ro;		/* true => reg is read only */
	const char	*cr_name;	/* name of reg */
} config_regs[] = {
	{ VTCFG_R_HOSTCAP,	4, 1, "HOSTCAP" },
	{ VTCFG_R_GUESTCAP,	4, 0, "GUESTCAP" },
	{ VTCFG_R_PFN,		4, 0, "PFN" },
	{ VTCFG_R_QNUM,		2, 1, "QNUM" },
	{ VTCFG_R_QSEL,		2, 0, "QSEL" },
	{ VTCFG_R_QNOTIFY,	2, 0, "QNOTIFY" },
	{ VTCFG_R_STATUS,	1, 0, "STATUS" },
	{ VTCFG_R_ISR,		1, 0, "ISR" },
	{ VTCFG_R_CFGVEC,	2, 0, "CFGVEC" },
	{ VTCFG_R_QVEC,		2, 0, "QVEC" },
};

static inline struct config_reg *
vi_find_cr(int offset) {
	u_int hi, lo, mid;
	struct config_reg *cr;

	lo = 0;
	hi = sizeof(config_regs) / sizeof(*config_regs) - 1;
	while (hi >= lo) {
		mid = (hi + lo) >> 1;
		cr = &config_regs[mid];
		if (cr->cr_offset == offset)
			return (cr);
		if (cr->cr_offset < offset)
			lo = mid + 1;
		else
			hi = mid - 1;
	}
	return (NULL);
}

/*
 * Handle pci config space reads.
 * If it's to the MSI-X info, do that.
 * If it's part of the virtio standard stuff, do that.
 * Otherwise dispatch to the actual driver.
 */
uint64_t
vi_pci_read(struct vmctx *ctx, int vcpu, struct pci_devinst *pi,
	    int baridx, uint64_t offset, int size)
{
	struct virtio_softc *vs = pi->pi_arg;
	struct virtio_consts *vc;
	struct config_reg *cr;
	uint64_t virtio_config_size, max;
	const char *name;
	uint32_t newoff;
	uint32_t value;
	int error;

	if (vs->vs_flags & VIRTIO_USE_MSIX) {
		if (baridx == pci_msix_table_bar(pi) ||
		    baridx == pci_msix_pba_bar(pi)) {
			return (pci_emul_msix_tread(pi, offset, size));
		}
	}

	/* XXX probably should do something better than just assert() */
	assert(baridx == 0);

	if (vs->vs_mtx)
		pthread_mutex_lock(vs->vs_mtx);

	vc = vs->vs_vc;
	name = vc->vc_name;
	value = size == 1 ? 0xff : size == 2 ? 0xffff : 0xffffffff;

	if (size != 1 && size != 2 && size != 4)
		goto bad;

	if (pci_msix_enabled(pi))
		virtio_config_size = VTCFG_R_CFG1;
	else
		virtio_config_size = VTCFG_R_CFG0;

	if (offset >= virtio_config_size) {
		/*
		 * Subtract off the standard size (including MSI-X
		 * registers if enabled) and dispatch to underlying driver.
		 * If that fails, fall into general code.
		 */
		newoff = offset - virtio_config_size;
		max = vc->vc_cfgsize ? vc->vc_cfgsize : 0x100000000;
		if (newoff + size > max)
			goto bad;
		error = (*vc->vc_cfgread)(DEV_SOFTC(vs), newoff, size, &value);
		if (!error)
			goto done;
	}

bad:
	cr = vi_find_cr(offset);
	if (cr == NULL || cr->cr_size != size) {
		if (cr != NULL) {
			/* offset must be OK, so size must be bad */
			fprintf(stderr,
			    "%s: read from %s: bad size %d\r\n",
			    name, cr->cr_name, size);
		} else {
			fprintf(stderr,
			    "%s: read from bad offset/size %jd/%d\r\n",
			    name, (uintmax_t)offset, size);
		}
		goto done;
	}

	switch (offset) {
	case VTCFG_R_HOSTCAP:
		value = vc->vc_hv_caps;
		break;
	case VTCFG_R_GUESTCAP:
		value = vs->vs_negotiated_caps;
		break;
	case VTCFG_R_PFN:
		if (vs->vs_curq < vc->vc_nvq)
			value = vs->vs_queues[vs->vs_curq].vq_pfn;
		break;
	case VTCFG_R_QNUM:
		value = vs->vs_curq < vc->vc_nvq ?
		    vs->vs_queues[vs->vs_curq].vq_qsize : 0;
		break;
	case VTCFG_R_QSEL:
		value = vs->vs_curq;
		break;
	case VTCFG_R_QNOTIFY:
		value = 0;	/* XXX */
		break;
	case VTCFG_R_STATUS:
		value = vs->vs_status;
		break;
	case VTCFG_R_ISR:
		value = vs->vs_isr;
		vs->vs_isr = 0;		/* a read clears this flag */
		if (value)
			pci_lintr_deassert(pi);
		break;
	case VTCFG_R_CFGVEC:
		value = vs->vs_msix_cfg_idx;
		break;
	case VTCFG_R_QVEC:
		value = vs->vs_curq < vc->vc_nvq ?
		    vs->vs_queues[vs->vs_curq].vq_msix_idx :
		    VIRTIO_MSI_NO_VECTOR;
		break;
	}
done:
	if (vs->vs_mtx)
		pthread_mutex_unlock(vs->vs_mtx);
	return (value);
}

/*
 * Handle pci config space writes.
 * If it's to the MSI-X info, do that.
 * If it's part of the virtio standard stuff, do that.
 * Otherwise dispatch to the actual driver.
 */
void
vi_pci_write(struct vmctx *ctx, int vcpu, struct pci_devinst *pi,
	     int baridx, uint64_t offset, int size, uint64_t value)
{
	struct virtio_softc *vs = pi->pi_arg;
	struct vqueue_info *vq;
	struct virtio_consts *vc;
	struct config_reg *cr;
	uint64_t virtio_config_size, max;
	const char *name;
	uint32_t newoff;
	int error;

	if (vs->vs_flags & VIRTIO_USE_MSIX) {
		if (baridx == pci_msix_table_bar(pi) ||
		    baridx == pci_msix_pba_bar(pi)) {
			pci_emul_msix_twrite(pi, offset, size, value);
			return;
		}
	}

	/* XXX probably should do something better than just assert() */
	assert(baridx == 0);

	if (vs->vs_mtx)
		pthread_mutex_lock(vs->vs_mtx);

	vc = vs->vs_vc;
	name = vc->vc_name;

	if (size != 1 && size != 2 && size != 4)
		goto bad;

	if (pci_msix_enabled(pi))
		virtio_config_size = VTCFG_R_CFG1;
	else
		virtio_config_size = VTCFG_R_CFG0;

	if (offset >= virtio_config_size) {
		/*
		 * Subtract off the standard size (including MSI-X
		 * registers if enabled) and dispatch to underlying driver.
		 */
		newoff = offset - virtio_config_size;
		max = vc->vc_cfgsize ? vc->vc_cfgsize : 0x100000000;
		if (newoff + size > max)
			goto bad;
		error = (*vc->vc_cfgwrite)(DEV_SOFTC(vs), newoff, size, value);
		if (!error)
			goto done;
	}

bad:
	cr = vi_find_cr(offset);
	if (cr == NULL || cr->cr_size != size || cr->cr_ro) {
		if (cr != NULL) {
			/* offset must be OK, wrong size and/or reg is R/O */
			if (cr->cr_size != size)
				fprintf(stderr,
				    "%s: write to %s: bad size %d\r\n",
				    name, cr->cr_name, size);
			if (cr->cr_ro)
				fprintf(stderr,
				    "%s: write to read-only reg %s\r\n",
				    name, cr->cr_name);
		} else {
			fprintf(stderr,
			    "%s: write to bad offset/size %jd/%d\r\n",
			    name, (uintmax_t)offset, size);
		}
		goto done;
	}

	switch (offset) {
	case VTCFG_R_GUESTCAP:
		vs->vs_negotiated_caps = value & vc->vc_hv_caps;
		if (vc->vc_apply_features)
			(*vc->vc_apply_features)(DEV_SOFTC(vs),
			    vs->vs_negotiated_caps);
		break;
	case VTCFG_R_PFN:
		if (vs->vs_curq >= vc->vc_nvq)
			goto bad_qindex;
		vi_vq_init(vs, value);
		break;
	case VTCFG_R_QSEL:
		/*
		 * Note that the guest is allowed to select an
		 * invalid queue; we just need to return a QNUM
		 * of 0 while the bad queue is selected.
		 */
		vs->vs_curq = value;
		break;
	case VTCFG_R_QNOTIFY:
		if (value >= vc->vc_nvq) {
			fprintf(stderr, "%s: queue %d notify out of range\r\n",
				name, (int)value);
			goto done;
		}
		vq = &vs->vs_queues[value];
		if (vq->vq_notify)
			(*vq->vq_notify)(DEV_SOFTC(vs), vq);
		else if (vc->vc_qnotify)
			(*vc->vc_qnotify)(DEV_SOFTC(vs), vq);
		else
			fprintf(stderr,
			    "%s: qnotify queue %d: missing vq/vc notify\r\n",
				name, (int)value);
		break;
	case VTCFG_R_STATUS:
		vs->vs_status = value;
		if (value == 0)
			(*vc->vc_reset)(DEV_SOFTC(vs));
		break;
	case VTCFG_R_CFGVEC:
		vs->vs_msix_cfg_idx = value;
		break;
	case VTCFG_R_QVEC:
		if (vs->vs_curq >= vc->vc_nvq)
			goto bad_qindex;
		vq = &vs->vs_queues[vs->vs_curq];
		vq->vq_msix_idx = value;
		break;
	}
	goto done;

bad_qindex:
	fprintf(stderr,
	    "%s: write config reg %s: curq %d >= max %d\r\n",
	    name, cr->cr_name, vs->vs_curq, vc->vc_nvq);
done:
	if (vs->vs_mtx)
		pthread_mutex_unlock(vs->vs_mtx);
}
OpenPOWER on IntegriCloud