aboutsummaryrefslogtreecommitdiff
path: root/util/vfio-helpers.c
blob: b037d5faa51edd352250443e6ff4267a5af15ac5 (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
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
/*
 * VFIO utility
 *
 * Copyright 2016 - 2018 Red Hat, Inc.
 *
 * Authors:
 *   Fam Zheng <famz@redhat.com>
 *
 * This work is licensed under the terms of the GNU GPL, version 2 or later.
 * See the COPYING file in the top-level directory.
 */

#include "qemu/osdep.h"
#include <sys/ioctl.h>
#include <linux/vfio.h>
#include "qapi/error.h"
#include "exec/ramlist.h"
#include "exec/cpu-common.h"
#include "exec/memory.h"
#include "trace.h"
#include "qemu/error-report.h"
#include "standard-headers/linux/pci_regs.h"
#include "qemu/event_notifier.h"
#include "qemu/vfio-helpers.h"
#include "qemu/lockable.h"
#include "trace.h"

#define QEMU_VFIO_DEBUG 0

#define QEMU_VFIO_IOVA_MIN 0x10000ULL
/* XXX: Once VFIO exposes the iova bit width in the IOMMU capability interface,
 * we can use a runtime limit; alternatively it's also possible to do platform
 * specific detection by reading sysfs entries. Until then, 39 is a safe bet.
 **/
#define QEMU_VFIO_IOVA_MAX (1ULL << 39)

typedef struct {
    /* Page aligned addr. */
    void *host;
    size_t size;
    uint64_t iova;
} IOVAMapping;

struct IOVARange {
    uint64_t start;
    uint64_t end;
};

struct QEMUVFIOState {
    QemuMutex lock;

    /* These fields are protected by BQL */
    int container;
    int group;
    int device;
    RAMBlockNotifier ram_notifier;
    struct vfio_region_info config_region_info, bar_region_info[6];
    struct IOVARange *usable_iova_ranges;
    uint8_t nb_iova_ranges;

    /* These fields are protected by @lock */
    /* VFIO's IO virtual address space is managed by splitting into a few
     * sections:
     *
     * ---------------       <= 0
     * |xxxxxxxxxxxxx|
     * |-------------|       <= QEMU_VFIO_IOVA_MIN
     * |             |
     * |    Fixed    |
     * |             |
     * |-------------|       <= low_water_mark
     * |             |
     * |    Free     |
     * |             |
     * |-------------|       <= high_water_mark
     * |             |
     * |    Temp     |
     * |             |
     * |-------------|       <= QEMU_VFIO_IOVA_MAX
     * |xxxxxxxxxxxxx|
     * |xxxxxxxxxxxxx|
     * ---------------
     *
     * - Addresses lower than QEMU_VFIO_IOVA_MIN are reserved as invalid;
     *
     * - Fixed mappings of HVAs are assigned "low" IOVAs in the range of
     *   [QEMU_VFIO_IOVA_MIN, low_water_mark).  Once allocated they will not be
     *   reclaimed - low_water_mark never shrinks;
     *
     * - IOVAs in range [low_water_mark, high_water_mark) are free;
     *
     * - IOVAs in range [high_water_mark, QEMU_VFIO_IOVA_MAX) are volatile
     *   mappings. At each qemu_vfio_dma_reset_temporary() call, the whole area
     *   is recycled. The caller should make sure I/O's depending on these
     *   mappings are completed before calling.
     **/
    uint64_t low_water_mark;
    uint64_t high_water_mark;
    IOVAMapping *mappings;
    int nr_mappings;
};

/**
 * Find group file by PCI device address as specified @device, and return the
 * path. The returned string is owned by caller and should be g_free'ed later.
 */
static char *sysfs_find_group_file(const char *device, Error **errp)
{
    char *sysfs_link;
    char *sysfs_group;
    char *p;
    char *path = NULL;

    sysfs_link = g_strdup_printf("/sys/bus/pci/devices/%s/iommu_group", device);
    sysfs_group = g_malloc0(PATH_MAX);
    if (readlink(sysfs_link, sysfs_group, PATH_MAX - 1) == -1) {
        error_setg_errno(errp, errno, "Failed to find iommu group sysfs path");
        goto out;
    }
    p = strrchr(sysfs_group, '/');
    if (!p) {
        error_setg(errp, "Failed to find iommu group number");
        goto out;
    }

    path = g_strdup_printf("/dev/vfio/%s", p + 1);
out:
    g_free(sysfs_link);
    g_free(sysfs_group);
    return path;
}

static inline void assert_bar_index_valid(QEMUVFIOState *s, int index)
{
    assert(index >= 0 && index < ARRAY_SIZE(s->bar_region_info));
}

static int qemu_vfio_pci_init_bar(QEMUVFIOState *s, int index, Error **errp)
{
    g_autofree char *barname = NULL;
    assert_bar_index_valid(s, index);
    s->bar_region_info[index] = (struct vfio_region_info) {
        .index = VFIO_PCI_BAR0_REGION_INDEX + index,
        .argsz = sizeof(struct vfio_region_info),
    };
    if (ioctl(s->device, VFIO_DEVICE_GET_REGION_INFO, &s->bar_region_info[index])) {
        error_setg_errno(errp, errno, "Failed to get BAR region info");
        return -errno;
    }
    barname = g_strdup_printf("bar[%d]", index);
    trace_qemu_vfio_region_info(barname, s->bar_region_info[index].offset,
                                s->bar_region_info[index].size,
                                s->bar_region_info[index].cap_offset);

    return 0;
}

/**
 * Map a PCI bar area.
 */
void *qemu_vfio_pci_map_bar(QEMUVFIOState *s, int index,
                            uint64_t offset, uint64_t size, int prot,
                            Error **errp)
{
    void *p;
    assert(QEMU_IS_ALIGNED(offset, qemu_real_host_page_size));
    assert_bar_index_valid(s, index);
    p = mmap(NULL, MIN(size, s->bar_region_info[index].size - offset),
             prot, MAP_SHARED,
             s->device, s->bar_region_info[index].offset + offset);
    trace_qemu_vfio_pci_map_bar(index, s->bar_region_info[index].offset ,
                                size, offset, p);
    if (p == MAP_FAILED) {
        error_setg_errno(errp, errno, "Failed to map BAR region");
        p = NULL;
    }
    return p;
}

/**
 * Unmap a PCI bar area.
 */
void qemu_vfio_pci_unmap_bar(QEMUVFIOState *s, int index, void *bar,
                             uint64_t offset, uint64_t size)
{
    if (bar) {
        munmap(bar, MIN(size, s->bar_region_info[index].size - offset));
    }
}

/**
 * Initialize device IRQ with @irq_type and register an event notifier.
 */
int qemu_vfio_pci_init_irq(QEMUVFIOState *s, EventNotifier *e,
                           int irq_type, Error **errp)
{
    int r;
    struct vfio_irq_set *irq_set;
    size_t irq_set_size;
    struct vfio_irq_info irq_info = { .argsz = sizeof(irq_info) };

    irq_info.index = irq_type;
    if (ioctl(s->device, VFIO_DEVICE_GET_IRQ_INFO, &irq_info)) {
        error_setg_errno(errp, errno, "Failed to get device interrupt info");
        return -errno;
    }
    if (!(irq_info.flags & VFIO_IRQ_INFO_EVENTFD)) {
        error_setg(errp, "Device interrupt doesn't support eventfd");
        return -EINVAL;
    }

    irq_set_size = sizeof(*irq_set) + sizeof(int);
    irq_set = g_malloc0(irq_set_size);

    /* Get to a known IRQ state */
    *irq_set = (struct vfio_irq_set) {
        .argsz = irq_set_size,
        .flags = VFIO_IRQ_SET_DATA_EVENTFD | VFIO_IRQ_SET_ACTION_TRIGGER,
        .index = irq_info.index,
        .start = 0,
        .count = 1,
    };

    *(int *)&irq_set->data = event_notifier_get_fd(e);
    r = ioctl(s->device, VFIO_DEVICE_SET_IRQS, irq_set);
    g_free(irq_set);
    if (r) {
        error_setg_errno(errp, errno, "Failed to setup device interrupt");
        return -errno;
    }
    return 0;
}

static int qemu_vfio_pci_read_config(QEMUVFIOState *s, void *buf,
                                     int size, int ofs)
{
    int ret;

    trace_qemu_vfio_pci_read_config(buf, ofs, size,
                                    s->config_region_info.offset,
                                    s->config_region_info.size);
    assert(QEMU_IS_ALIGNED(s->config_region_info.offset + ofs, size));
    do {
        ret = pread(s->device, buf, size, s->config_region_info.offset + ofs);
    } while (ret == -1 && errno == EINTR);
    return ret == size ? 0 : -errno;
}

static int qemu_vfio_pci_write_config(QEMUVFIOState *s, void *buf, int size, int ofs)
{
    int ret;

    trace_qemu_vfio_pci_write_config(buf, ofs, size,
                                     s->config_region_info.offset,
                                     s->config_region_info.size);
    assert(QEMU_IS_ALIGNED(s->config_region_info.offset + ofs, size));
    do {
        ret = pwrite(s->device, buf, size, s->config_region_info.offset + ofs);
    } while (ret == -1 && errno == EINTR);
    return ret == size ? 0 : -errno;
}

static void collect_usable_iova_ranges(QEMUVFIOState *s, void *buf)
{
    struct vfio_iommu_type1_info *info = (struct vfio_iommu_type1_info *)buf;
    struct vfio_info_cap_header *cap = (void *)buf + info->cap_offset;
    struct vfio_iommu_type1_info_cap_iova_range *cap_iova_range;
    int i;

    while (cap->id != VFIO_IOMMU_TYPE1_INFO_CAP_IOVA_RANGE) {
        if (!cap->next) {
            return;
        }
        cap = (struct vfio_info_cap_header *)(buf + cap->next);
    }

    cap_iova_range = (struct vfio_iommu_type1_info_cap_iova_range *)cap;

    s->nb_iova_ranges = cap_iova_range->nr_iovas;
    if (s->nb_iova_ranges > 1) {
        s->usable_iova_ranges =
            g_renew(struct IOVARange, s->usable_iova_ranges,
                    s->nb_iova_ranges);
    }

    for (i = 0; i < s->nb_iova_ranges; i++) {
        s->usable_iova_ranges[i].start = cap_iova_range->iova_ranges[i].start;
        s->usable_iova_ranges[i].end = cap_iova_range->iova_ranges[i].end;
    }
}

static int qemu_vfio_init_pci(QEMUVFIOState *s, const char *device,
                              Error **errp)
{
    int ret;
    int i;
    uint16_t pci_cmd;
    struct vfio_group_status group_status = { .argsz = sizeof(group_status) };
    struct vfio_iommu_type1_info *iommu_info = NULL;
    size_t iommu_info_size = sizeof(*iommu_info);
    struct vfio_device_info device_info = { .argsz = sizeof(device_info) };
    char *group_file = NULL;

    s->usable_iova_ranges = NULL;

    /* Create a new container */
    s->container = open("/dev/vfio/vfio", O_RDWR);

    if (s->container == -1) {
        error_setg_errno(errp, errno, "Failed to open /dev/vfio/vfio");
        return -errno;
    }
    if (ioctl(s->container, VFIO_GET_API_VERSION) != VFIO_API_VERSION) {
        error_setg(errp, "Invalid VFIO version");
        ret = -EINVAL;
        goto fail_container;
    }

    if (!ioctl(s->container, VFIO_CHECK_EXTENSION, VFIO_TYPE1_IOMMU)) {
        error_setg_errno(errp, errno, "VFIO IOMMU Type1 is not supported");
        ret = -EINVAL;
        goto fail_container;
    }

    /* Open the group */
    group_file = sysfs_find_group_file(device, errp);
    if (!group_file) {
        ret = -EINVAL;
        goto fail_container;
    }

    s->group = open(group_file, O_RDWR);
    if (s->group == -1) {
        error_setg_errno(errp, errno, "Failed to open VFIO group file: %s",
                         group_file);
        g_free(group_file);
        ret = -errno;
        goto fail_container;
    }
    g_free(group_file);

    /* Test the group is viable and available */
    if (ioctl(s->group, VFIO_GROUP_GET_STATUS, &group_status)) {
        error_setg_errno(errp, errno, "Failed to get VFIO group status");
        ret = -errno;
        goto fail;
    }

    if (!(group_status.flags & VFIO_GROUP_FLAGS_VIABLE)) {
        error_setg(errp, "VFIO group is not viable");
        ret = -EINVAL;
        goto fail;
    }

    /* Add the group to the container */
    if (ioctl(s->group, VFIO_GROUP_SET_CONTAINER, &s->container)) {
        error_setg_errno(errp, errno, "Failed to add group to VFIO container");
        ret = -errno;
        goto fail;
    }

    /* Enable the IOMMU model we want */
    if (ioctl(s->container, VFIO_SET_IOMMU, VFIO_TYPE1_IOMMU)) {
        error_setg_errno(errp, errno, "Failed to set VFIO IOMMU type");
        ret = -errno;
        goto fail;
    }

    iommu_info = g_malloc0(iommu_info_size);
    iommu_info->argsz = iommu_info_size;

    /* Get additional IOMMU info */
    if (ioctl(s->container, VFIO_IOMMU_GET_INFO, iommu_info)) {
        error_setg_errno(errp, errno, "Failed to get IOMMU info");
        ret = -errno;
        goto fail;
    }

    /*
     * if the kernel does not report usable IOVA regions, choose
     * the legacy [QEMU_VFIO_IOVA_MIN, QEMU_VFIO_IOVA_MAX -1] region
     */
    s->nb_iova_ranges = 1;
    s->usable_iova_ranges = g_new0(struct IOVARange, 1);
    s->usable_iova_ranges[0].start = QEMU_VFIO_IOVA_MIN;
    s->usable_iova_ranges[0].end = QEMU_VFIO_IOVA_MAX - 1;

    if (iommu_info->argsz > iommu_info_size) {
        iommu_info_size = iommu_info->argsz;
        iommu_info = g_realloc(iommu_info, iommu_info_size);
        if (ioctl(s->container, VFIO_IOMMU_GET_INFO, iommu_info)) {
            ret = -errno;
            goto fail;
        }
        collect_usable_iova_ranges(s, iommu_info);
    }

    s->device = ioctl(s->group, VFIO_GROUP_GET_DEVICE_FD, device);

    if (s->device < 0) {
        error_setg_errno(errp, errno, "Failed to get device fd");
        ret = -errno;
        goto fail;
    }

    /* Test and setup the device */
    if (ioctl(s->device, VFIO_DEVICE_GET_INFO, &device_info)) {
        error_setg_errno(errp, errno, "Failed to get device info");
        ret = -errno;
        goto fail;
    }

    if (device_info.num_regions < VFIO_PCI_CONFIG_REGION_INDEX) {
        error_setg(errp, "Invalid device regions");
        ret = -EINVAL;
        goto fail;
    }

    s->config_region_info = (struct vfio_region_info) {
        .index = VFIO_PCI_CONFIG_REGION_INDEX,
        .argsz = sizeof(struct vfio_region_info),
    };
    if (ioctl(s->device, VFIO_DEVICE_GET_REGION_INFO, &s->config_region_info)) {
        error_setg_errno(errp, errno, "Failed to get config region info");
        ret = -errno;
        goto fail;
    }
    trace_qemu_vfio_region_info("config", s->config_region_info.offset,
                                s->config_region_info.size,
                                s->config_region_info.cap_offset);

    for (i = 0; i < ARRAY_SIZE(s->bar_region_info); i++) {
        ret = qemu_vfio_pci_init_bar(s, i, errp);
        if (ret) {
            goto fail;
        }
    }

    /* Enable bus master */
    ret = qemu_vfio_pci_read_config(s, &pci_cmd, sizeof(pci_cmd), PCI_COMMAND);
    if (ret) {
        goto fail;
    }
    pci_cmd |= PCI_COMMAND_MASTER;
    ret = qemu_vfio_pci_write_config(s, &pci_cmd, sizeof(pci_cmd), PCI_COMMAND);
    if (ret) {
        goto fail;
    }
    g_free(iommu_info);
    return 0;
fail:
    g_free(s->usable_iova_ranges);
    s->usable_iova_ranges = NULL;
    s->nb_iova_ranges = 0;
    g_free(iommu_info);
    close(s->group);
fail_container:
    close(s->container);
    return ret;
}

static void qemu_vfio_ram_block_added(RAMBlockNotifier *n, void *host,
                                      size_t size, size_t max_size)
{
    QEMUVFIOState *s = container_of(n, QEMUVFIOState, ram_notifier);
    Error *local_err = NULL;
    int ret;

    trace_qemu_vfio_ram_block_added(s, host, max_size);
    ret = qemu_vfio_dma_map(s, host, max_size, false, NULL, &local_err);
    if (ret) {
        error_reportf_err(local_err,
                          "qemu_vfio_dma_map(%p, %zu) failed: ",
                          host, max_size);
    }
}

static void qemu_vfio_ram_block_removed(RAMBlockNotifier *n, void *host,
                                        size_t size, size_t max_size)
{
    QEMUVFIOState *s = container_of(n, QEMUVFIOState, ram_notifier);
    if (host) {
        trace_qemu_vfio_ram_block_removed(s, host, max_size);
        qemu_vfio_dma_unmap(s, host);
    }
}

static void qemu_vfio_open_common(QEMUVFIOState *s)
{
    qemu_mutex_init(&s->lock);
    s->ram_notifier.ram_block_added = qemu_vfio_ram_block_added;
    s->ram_notifier.ram_block_removed = qemu_vfio_ram_block_removed;
    s->low_water_mark = QEMU_VFIO_IOVA_MIN;
    s->high_water_mark = QEMU_VFIO_IOVA_MAX;
    ram_block_notifier_add(&s->ram_notifier);
}

/**
 * Open a PCI device, e.g. "0000:00:01.0".
 */
QEMUVFIOState *qemu_vfio_open_pci(const char *device, Error **errp)
{
    int r;
    QEMUVFIOState *s = g_new0(QEMUVFIOState, 1);

    /*
     * VFIO may pin all memory inside mappings, resulting it in pinning
     * all memory inside RAM blocks unconditionally.
     */
    r = ram_block_discard_disable(true);
    if (r) {
        error_setg_errno(errp, -r, "Cannot set discarding of RAM broken");
        g_free(s);
        return NULL;
    }

    r = qemu_vfio_init_pci(s, device, errp);
    if (r) {
        ram_block_discard_disable(false);
        g_free(s);
        return NULL;
    }
    qemu_vfio_open_common(s);
    return s;
}

static void qemu_vfio_dump_mappings(QEMUVFIOState *s)
{
    for (int i = 0; i < s->nr_mappings; ++i) {
        trace_qemu_vfio_dump_mapping(s->mappings[i].host,
                                     s->mappings[i].iova,
                                     s->mappings[i].size);
    }
}

/**
 * Find the mapping entry that contains [host, host + size) and set @index to
 * the position. If no entry contains it, @index is the position _after_ which
 * to insert the new mapping. IOW, it is the index of the largest element that
 * is smaller than @host, or -1 if no entry is.
 */
static IOVAMapping *qemu_vfio_find_mapping(QEMUVFIOState *s, void *host,
                                           int *index)
{
    IOVAMapping *p = s->mappings;
    IOVAMapping *q = p ? p + s->nr_mappings - 1 : NULL;
    IOVAMapping *mid;
    trace_qemu_vfio_find_mapping(s, host);
    if (!p) {
        *index = -1;
        return NULL;
    }
    while (true) {
        mid = p + (q - p) / 2;
        if (mid == p) {
            break;
        }
        if (mid->host > host) {
            q = mid;
        } else if (mid->host < host) {
            p = mid;
        } else {
            break;
        }
    }
    if (mid->host > host) {
        mid--;
    } else if (mid < &s->mappings[s->nr_mappings - 1]
               && (mid + 1)->host <= host) {
        mid++;
    }
    *index = mid - &s->mappings[0];
    if (mid >= &s->mappings[0] &&
        mid->host <= host && mid->host + mid->size > host) {
        assert(mid < &s->mappings[s->nr_mappings]);
        return mid;
    }
    /* At this point *index + 1 is the right position to insert the new
     * mapping.*/
    return NULL;
}

/**
 * Allocate IOVA and create a new mapping record and insert it in @s.
 */
static IOVAMapping *qemu_vfio_add_mapping(QEMUVFIOState *s,
                                          void *host, size_t size,
                                          int index, uint64_t iova)
{
    int shift;
    IOVAMapping m = {.host = host, .size = size, .iova = iova};
    IOVAMapping *insert;

    assert(QEMU_IS_ALIGNED(size, qemu_real_host_page_size));
    assert(QEMU_IS_ALIGNED(s->low_water_mark, qemu_real_host_page_size));
    assert(QEMU_IS_ALIGNED(s->high_water_mark, qemu_real_host_page_size));
    trace_qemu_vfio_new_mapping(s, host, size, index, iova);

    assert(index >= 0);
    s->nr_mappings++;
    s->mappings = g_renew(IOVAMapping, s->mappings, s->nr_mappings);
    insert = &s->mappings[index];
    shift = s->nr_mappings - index - 1;
    if (shift) {
        memmove(insert + 1, insert, shift * sizeof(s->mappings[0]));
    }
    *insert = m;
    return insert;
}

/* Do the DMA mapping with VFIO. */
static int qemu_vfio_do_mapping(QEMUVFIOState *s, void *host, size_t size,
                                uint64_t iova, Error **errp)
{
    struct vfio_iommu_type1_dma_map dma_map = {
        .argsz = sizeof(dma_map),
        .flags = VFIO_DMA_MAP_FLAG_READ | VFIO_DMA_MAP_FLAG_WRITE,
        .iova = iova,
        .vaddr = (uintptr_t)host,
        .size = size,
    };
    trace_qemu_vfio_do_mapping(s, host, iova, size);

    if (ioctl(s->container, VFIO_IOMMU_MAP_DMA, &dma_map)) {
        error_setg_errno(errp, errno, "VFIO_MAP_DMA failed");
        return -errno;
    }
    return 0;
}

/**
 * Undo the DMA mapping from @s with VFIO, and remove from mapping list.
 */
static void qemu_vfio_undo_mapping(QEMUVFIOState *s, IOVAMapping *mapping,
                                   Error **errp)
{
    int index;
    struct vfio_iommu_type1_dma_unmap unmap = {
        .argsz = sizeof(unmap),
        .flags = 0,
        .iova = mapping->iova,
        .size = mapping->size,
    };

    index = mapping - s->mappings;
    assert(mapping->size > 0);
    assert(QEMU_IS_ALIGNED(mapping->size, qemu_real_host_page_size));
    assert(index >= 0 && index < s->nr_mappings);
    if (ioctl(s->container, VFIO_IOMMU_UNMAP_DMA, &unmap)) {
        error_setg_errno(errp, errno, "VFIO_UNMAP_DMA failed");
    }
    memmove(mapping, &s->mappings[index + 1],
            sizeof(s->mappings[0]) * (s->nr_mappings - index - 1));
    s->nr_mappings--;
    s->mappings = g_renew(IOVAMapping, s->mappings, s->nr_mappings);
}

/* Check if the mapping list is (ascending) ordered. */
static bool qemu_vfio_verify_mappings(QEMUVFIOState *s)
{
    int i;
    if (QEMU_VFIO_DEBUG) {
        for (i = 0; i < s->nr_mappings - 1; ++i) {
            if (!(s->mappings[i].host < s->mappings[i + 1].host)) {
                error_report("item %d not sorted!", i);
                qemu_vfio_dump_mappings(s);
                return false;
            }
            if (!(s->mappings[i].host + s->mappings[i].size <=
                  s->mappings[i + 1].host)) {
                error_report("item %d overlap with next!", i);
                qemu_vfio_dump_mappings(s);
                return false;
            }
        }
    }
    return true;
}

static bool qemu_vfio_find_fixed_iova(QEMUVFIOState *s, size_t size,
                                      uint64_t *iova, Error **errp)
{
    int i;

    for (i = 0; i < s->nb_iova_ranges; i++) {
        if (s->usable_iova_ranges[i].end < s->low_water_mark) {
            continue;
        }
        s->low_water_mark =
            MAX(s->low_water_mark, s->usable_iova_ranges[i].start);

        if (s->usable_iova_ranges[i].end - s->low_water_mark + 1 >= size ||
            s->usable_iova_ranges[i].end - s->low_water_mark + 1 == 0) {
            *iova = s->low_water_mark;
            s->low_water_mark += size;
            return true;
        }
    }
    error_setg(errp, "fixed iova range not found");

    return false;
}

static bool qemu_vfio_find_temp_iova(QEMUVFIOState *s, size_t size,
                                     uint64_t *iova, Error **errp)
{
    int i;

    for (i = s->nb_iova_ranges - 1; i >= 0; i--) {
        if (s->usable_iova_ranges[i].start > s->high_water_mark) {
            continue;
        }
        s->high_water_mark =
            MIN(s->high_water_mark, s->usable_iova_ranges[i].end + 1);

        if (s->high_water_mark - s->usable_iova_ranges[i].start + 1 >= size ||
            s->high_water_mark - s->usable_iova_ranges[i].start + 1 == 0) {
            *iova = s->high_water_mark - size;
            s->high_water_mark = *iova;
            return true;
        }
    }
    error_setg(errp, "temporary iova range not found");

    return false;
}

/**
 * qemu_vfio_water_mark_reached:
 *
 * Returns %true if high watermark has been reached, %false otherwise.
 */
static bool qemu_vfio_water_mark_reached(QEMUVFIOState *s, size_t size,
                                         Error **errp)
{
    if (s->high_water_mark - s->low_water_mark + 1 < size) {
        error_setg(errp, "iova exhausted (water mark reached)");
        return true;
    }
    return false;
}

/* Map [host, host + size) area into a contiguous IOVA address space, and store
 * the result in @iova if not NULL. The caller need to make sure the area is
 * aligned to page size, and mustn't overlap with existing mapping areas (split
 * mapping status within this area is not allowed).
 */
int qemu_vfio_dma_map(QEMUVFIOState *s, void *host, size_t size,
                      bool temporary, uint64_t *iova, Error **errp)
{
    int index;
    IOVAMapping *mapping;
    uint64_t iova0;

    assert(QEMU_PTR_IS_ALIGNED(host, qemu_real_host_page_size));
    assert(QEMU_IS_ALIGNED(size, qemu_real_host_page_size));
    trace_qemu_vfio_dma_map(s, host, size, temporary, iova);
    QEMU_LOCK_GUARD(&s->lock);
    mapping = qemu_vfio_find_mapping(s, host, &index);
    if (mapping) {
        iova0 = mapping->iova + ((uint8_t *)host - (uint8_t *)mapping->host);
    } else {
        int ret;

        if (qemu_vfio_water_mark_reached(s, size, errp)) {
            return -ENOMEM;
        }
        if (!temporary) {
            if (!qemu_vfio_find_fixed_iova(s, size, &iova0, errp)) {
                return -ENOMEM;
            }

            mapping = qemu_vfio_add_mapping(s, host, size, index + 1, iova0);
            assert(qemu_vfio_verify_mappings(s));
            ret = qemu_vfio_do_mapping(s, host, size, iova0, errp);
            if (ret < 0) {
                qemu_vfio_undo_mapping(s, mapping, NULL);
                return ret;
            }
            qemu_vfio_dump_mappings(s);
        } else {
            if (!qemu_vfio_find_temp_iova(s, size, &iova0, errp)) {
                return -ENOMEM;
            }
            ret = qemu_vfio_do_mapping(s, host, size, iova0, errp);
            if (ret < 0) {
                return ret;
            }
        }
    }
    trace_qemu_vfio_dma_mapped(s, host, iova0, size);
    if (iova) {
        *iova = iova0;
    }
    return 0;
}

/* Reset the high watermark and free all "temporary" mappings. */
int qemu_vfio_dma_reset_temporary(QEMUVFIOState *s)
{
    struct vfio_iommu_type1_dma_unmap unmap = {
        .argsz = sizeof(unmap),
        .flags = 0,
        .iova = s->high_water_mark,
        .size = QEMU_VFIO_IOVA_MAX - s->high_water_mark,
    };
    trace_qemu_vfio_dma_reset_temporary(s);
    QEMU_LOCK_GUARD(&s->lock);
    if (ioctl(s->container, VFIO_IOMMU_UNMAP_DMA, &unmap)) {
        error_report("VFIO_UNMAP_DMA failed: %s", strerror(errno));
        return -errno;
    }
    s->high_water_mark = QEMU_VFIO_IOVA_MAX;
    return 0;
}

/* Unmapping the whole area that was previously mapped with
 * qemu_vfio_dma_map(). */
void qemu_vfio_dma_unmap(QEMUVFIOState *s, void *host)
{
    int index = 0;
    IOVAMapping *m;

    if (!host) {
        return;
    }

    trace_qemu_vfio_dma_unmap(s, host);
    QEMU_LOCK_GUARD(&s->lock);
    m = qemu_vfio_find_mapping(s, host, &index);
    if (!m) {
        return;
    }
    qemu_vfio_undo_mapping(s, m, NULL);
}

static void qemu_vfio_reset(QEMUVFIOState *s)
{
    ioctl(s->device, VFIO_DEVICE_RESET);
}

/* Close and free the VFIO resources. */
void qemu_vfio_close(QEMUVFIOState *s)
{
    int i;

    if (!s) {
        return;
    }
    for (i = 0; i < s->nr_mappings; ++i) {
        qemu_vfio_undo_mapping(s, &s->mappings[i], NULL);
    }
    ram_block_notifier_remove(&s->ram_notifier);
    g_free(s->usable_iova_ranges);
    s->nb_iova_ranges = 0;
    qemu_vfio_reset(s);
    close(s->device);
    close(s->group);
    close(s->container);
    ram_block_discard_disable(false);
}