vfio.cpp 20 KB
Newer Older
1
2
3
4
/** Virtual Function IO wrapper around kernel API
 *
 * @author Steffen Vogel <stvogel@eonerc.rwth-aachen.de>
 * @author Daniel Krebs <github@daniel-krebs.net>
Steffen Vogel's avatar
Steffen Vogel committed
5
 * @copyright 2014-2019, Steffen Vogel
6
7
8
 * @copyright 2018, Daniel Krebs
 * @license GNU General Public License (version 3)
 *
9
 * VILLAScommon
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
 *
 * This program is free software: you can redistribute it and/or modify
 * it under the terms of the GNU General Public License as published by
 * the Free Software Foundation, either version 3 of the License, or
 * any later version.
 *
 * This program is distributed in the hope that it will be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
 * GNU General Public License for more details.
 *
 * You should have received a copy of the GNU General Public License
 * along with this program.  If not, see <http://www.gnu.org/licenses/>.
 *********************************************************************************/

#define _DEFAULT_SOURCE

27
28
29
30
31
#ifdef __arm__
  #define _LARGEFILE64_SOURCE 1
  #define _FILE_OFFSET_BITS 64
#endif

32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
#include <algorithm>
#include <string>
#include <sstream>
#include <limits>

#include <cstdlib>
#include <cstring>

#include <stdint.h>
#include <unistd.h>
#include <fcntl.h>
#include <sys/ioctl.h>
#include <sys/eventfd.h>
#include <linux/pci_regs.h>

#include <villas/log.hpp>
#include <villas/kernel/pci.h>
#include <villas/kernel/kernel.h>
#include <villas/kernel/vfio.hpp>

52
53
using namespace villas;

54
static const char *vfio_pci_region_names[] = {
55
56
57
58
59
60
61
62
63
    "PCI_BAR0",		/* VFIO_PCI_BAR0_REGION_INDEX */
    "PCI_BAR1",		/* VFIO_PCI_BAR1_REGION_INDEX */
    "PCI_BAR2",		/* VFIO_PCI_BAR2_REGION_INDEX */
    "PCI_BAR3",		/* VFIO_PCI_BAR3_REGION_INDEX */
    "PCI_BAR4",		/* VFIO_PCI_BAR4_REGION_INDEX */
    "PCI_BAR5",		/* VFIO_PCI_BAR5_REGION_INDEX */
    "PCI_ROM",		/* VFIO_PCI_ROM_REGION_INDEX */
    "PCI_CONFIG",	/* VFIO_PCI_CONFIG_REGION_INDEX */
    "PCI_VGA"		/* VFIO_PCI_INTX_IRQ_INDEX */
64
65
66
};

static const char *vfio_pci_irq_names[] = {
67
68
69
70
71
    "PCI_INTX",		/* VFIO_PCI_INTX_IRQ_INDEX */
    "PCI_MSI", 		/* VFIO_PCI_MSI_IRQ_INDEX */
    "PCI_MSIX",		/* VFIO_PCI_MSIX_IRQ_INDEX */
    "PCI_ERR", 		/* VFIO_PCI_ERR_IRQ_INDEX */
    "PCI_REQ"		/* VFIO_PCI_REQ_IRQ_INDEX */
72
73
74
75
76
77
78
79
80
};

namespace villas {


VfioContainer::VfioContainer()
    : iova_next(0)
{

81
82
	Logger logger = logging.get("kernel:vfio");

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
	static constexpr const char* requiredKernelModules[] = {
	    "vfio", "vfio_pci", "vfio_iommu_type1"
	};

	for(const char* module : requiredKernelModules) {
		if(kernel_module_loaded(module) != 0) {
			logger->error("Kernel module '{}' required but not loaded. "
			              "Please load manually!", module);
			throw std::exception();
		}
	}

	/* Open VFIO API */
	fd = open(VFIO_DEV, O_RDWR);
	if (fd < 0) {
		logger->error("Failed to open VFIO container");
		throw std::exception();
	}

	/* Check VFIO API version */
	version = ioctl(fd, VFIO_GET_API_VERSION);
	if (version < 0 || version != VFIO_API_VERSION) {
		logger->error("Failed to get VFIO version");
		throw std::exception();
	}

	/* Check available VFIO extensions (IOMMU types) */
	extensions = 0;
	for (unsigned int i = VFIO_TYPE1_IOMMU; i <= VFIO_NOIOMMU_IOMMU; i++) {
		int ret = ioctl(fd, VFIO_CHECK_EXTENSION, i);
		if (ret < 0) {
			logger->error("Failed to get VFIO extensions");
			throw std::exception();
		}
		else if (ret > 0) {
			extensions |= (1 << i);
		}
	}

	hasIommu = false;

	if(not (extensions & (1 << VFIO_NOIOMMU_IOMMU))) {
		if(not (extensions & (1 << VFIO_TYPE1_IOMMU))) {
			logger->error("No supported IOMMU extension found");
			throw std::exception();
		} else {
			hasIommu = true;
		}
	}

	logger->debug("Version:    {:#x}", version);
	logger->debug("Extensions: {:#x}", extensions);
	logger->debug("IOMMU:      {}", hasIommu ? "yes" : "no");
}


VfioContainer::~VfioContainer()
{
141
142
	Logger logger = logging.get("kernel:vfio");

143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
	logger->debug("Clean up container with fd {}", fd);

	/* Release memory and close fds */
	groups.clear();

	/* Close container */
	int ret = close(fd);
	if (ret < 0) {
		logger->error("Cannot close vfio container fd {}", fd);
	}
}


std::shared_ptr<VfioContainer>
VfioContainer::create()
{
	std::shared_ptr<VfioContainer> container { new VfioContainer };
	return container;
}


void
VfioContainer::dump()
{
167
168
	Logger logger = logging.get("kernel:vfio");

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
	logger->info("File descriptor: {}", fd);
	logger->info("Version: {}", version);
	logger->info("Extensions: 0x{:x}", extensions);

	for(auto& group : groups) {
		logger->info("VFIO Group {}, viable={}, container={}",
		             group->index,
		             (group->status.flags & VFIO_GROUP_FLAGS_VIABLE) > 0,
		             (group->status.flags & VFIO_GROUP_FLAGS_CONTAINER_SET) > 0
		);

		for(auto& device : group->devices) {
			logger->info("Device {}: regions={}, irqs={}, flags={}",
			             device->name,
			             device->info.num_regions,
			             device->info.num_irqs,
			             device->info.flags
			);

			for (size_t i = 0; i < device->info.num_regions && i < 8; i++) {
				struct vfio_region_info *region = &device->regions[i];

				if (region->size > 0)
					logger->info("Region {} {}: size={}, offset={}, flags={}",
					             region->index,
					             (device->info.flags & VFIO_DEVICE_FLAGS_PCI) ?
					                 vfio_pci_region_names[i] : "",
					             region->size,
					             region->offset,
					             region->flags
					);
			}

			for (size_t i = 0; i < device->info.num_irqs; i++) {
				struct vfio_irq_info *irq = &device->irqs[i];

				if (irq->count > 0)
					logger->info("IRQ {} {}: count={}, flags={}",
					             irq->index,
					             (device->info.flags & VFIO_DEVICE_FLAGS_PCI ) ?
					                 vfio_pci_irq_names[i] : "",
					             irq->count,
					             irq->flags
				);
			}
		}
	}
}


VfioDevice&
VfioContainer::attachDevice(const char* name, int index)
{
222
223
	Logger logger = logging.get("kernel:vfio");

224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
	VfioGroup& group = getOrAttachGroup(index);
	auto device = std::make_unique<VfioDevice>(name, group);

	/* Open device fd */
	device->fd = ioctl(group.fd, VFIO_GROUP_GET_DEVICE_FD, name);
	if (device->fd < 0) {
		logger->error("Failed to open VFIO device: {}", device->name);
		throw std::exception();
	}

	/* Get device info */
	device->info.argsz = sizeof(device->info);

	int ret = ioctl(device->fd, VFIO_DEVICE_GET_INFO, &device->info);
	if (ret < 0) {
		logger->error("Failed to get VFIO device info for: {}", device->name);
		throw std::exception();
	}

	logger->debug("Device has {} regions", device->info.num_regions);
	logger->debug("Device has {} IRQs", device->info.num_irqs);

246
	/* Reserve slots already so that we can use the []-operator for access */
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
	device->irqs.resize(device->info.num_irqs);
	device->regions.resize(device->info.num_regions);
	device->mappings.resize(device->info.num_regions);

	/* Get device regions */
	for (size_t i = 0; i < device->info.num_regions && i < 8; i++) {
		struct vfio_region_info region;
		memset(&region, 0, sizeof (region));

		region.argsz = sizeof(region);
		region.index = i;

		ret = ioctl(device->fd, VFIO_DEVICE_GET_REGION_INFO, &region);
		if (ret < 0) {
			logger->error("Failed to get region of VFIO device: {}", device->name);
			throw std::exception();
		}

		device->regions[i] = region;
	}


	/* Get device irqs */
	for (size_t i = 0; i < device->info.num_irqs; i++) {
		struct vfio_irq_info irq;
		memset(&irq, 0, sizeof (irq));

		irq.argsz = sizeof(irq);
		irq.index = i;

		ret = ioctl(device->fd, VFIO_DEVICE_GET_IRQ_INFO, &irq);
		if (ret < 0) {
			logger->error("Failed to get IRQs of VFIO device: {}", device->name);
			throw std::exception();
		}

		device->irqs[i] = irq;
	}

	group.devices.push_back(std::move(device));

	return *group.devices.back().get();
}


VfioDevice&
VfioContainer::attachDevice(const pci_device* pdev)
{
	int ret;
	char name[32];
	static constexpr const char* kernelDriver = "vfio-pci";

299
300
	Logger logger = logging.get("kernel:vfio");

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
	/* Load PCI bus driver for VFIO */
	if (kernel_module_load("vfio_pci")) {
		logger->error("Failed to load kernel driver: vfio_pci");
		throw std::exception();
	}

	/* Bind PCI card to vfio-pci driver if not already bound */
	ret = pci_get_driver(pdev, name, sizeof(name));
	if (ret || strcmp(name, kernelDriver)) {
		logger->debug("Bind PCI card to kernel driver '{}'", kernelDriver);
		ret = pci_attach_driver(pdev, kernelDriver);
		if (ret) {
			logger->error("Failed to attach device to driver");
			throw std::exception();
		}
	}

	/* Get IOMMU group of device */
	int index = isIommuEnabled() ? pci_get_iommu_group(pdev) : 0;
	if (index < 0) {
		logger->error("Failed to get IOMMU group of device");
		throw std::exception();
	}

	/* VFIO device name consists of PCI BDF */
	snprintf(name, sizeof(name), "%04x:%02x:%02x.%x", pdev->slot.domain,
	         pdev->slot.bus, pdev->slot.device, pdev->slot.function);

	logger->info("Attach to device {} with index {}", std::string(name), index);
	auto& device = attachDevice(name, index);

	device.pci_device = pdev;

	/* Check if this is really a vfio-pci device */
	if(not device.isVfioPciDevice()) {
		logger->error("Device is not a vfio-pci device");
		throw std::exception();
	}

	return device;
}


uintptr_t
VfioContainer::memoryMap(uintptr_t virt, uintptr_t phys, size_t length)
{
	int ret;

349
350
351
	Logger logger = logging.get("kernel:vfio");

	if (not hasIommu) {
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
		logger->error("DMA mapping not supported without IOMMU");
		return UINTPTR_MAX;
	}

	if (length & 0xFFF) {
		length += 0x1000;
		length &= ~0xFFF;
	}

	/* Super stupid allocator */
	size_t iovaIncrement = 0;
	if (phys == UINTPTR_MAX) {
		phys = this->iova_next;
		iovaIncrement = length;
	}

	struct vfio_iommu_type1_dma_map dmaMap;
	memset(&dmaMap, 0, sizeof(dmaMap));

	dmaMap.argsz = sizeof(dmaMap);
	dmaMap.vaddr = virt;
	dmaMap.iova  = phys;
	dmaMap.size  = length;
	dmaMap.flags = VFIO_DMA_MAP_FLAG_READ | VFIO_DMA_MAP_FLAG_WRITE;

	ret = ioctl(this->fd, VFIO_IOMMU_MAP_DMA, &dmaMap);
	if (ret) {
		logger->error("Failed to create DMA mapping: {}", ret);
		return UINTPTR_MAX;
	}

	logger->debug("DMA map size={:#x}, iova={:#x}, vaddr={:#x}",
	              dmaMap.size, dmaMap.iova, dmaMap.vaddr);

386
	/* Mapping successful, advance IOVA allocator */
387
388
	this->iova_next += iovaIncrement;

389
390
391
	/* We intentionally don't return the actual mapped length, the users are
	 * only guaranteed to have their demanded memory mapped correctly
	 */
392
393
394
395
396
397
398
399
400
	return dmaMap.iova;
}


bool
VfioContainer::memoryUnmap(uintptr_t phys, size_t length)
{
	int ret;

401
402
403
	Logger logger = logging.get("kernel:vfio");

	if (not hasIommu)
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
		return true;

	struct vfio_iommu_type1_dma_unmap dmaUnmap;
	dmaUnmap.argsz = sizeof(struct vfio_iommu_type1_dma_unmap);
	dmaUnmap.flags = 0;
	dmaUnmap.iova  = phys;
	dmaUnmap.size  = length;

	ret = ioctl(this->fd, VFIO_IOMMU_UNMAP_DMA, &dmaUnmap);
	if (ret) {
		logger->error("Failed to unmap DMA mapping");
		return false;
	}

	return true;
}


VfioGroup&
VfioContainer::getOrAttachGroup(int index)
{
425
426
	Logger logger = logging.get("kernel:vfio");

427
	/* Search if group with index already exists */
428
	for (auto& group : groups) {
429
430
431
432
433
		if(group->index == index) {
			return *group;
		}
	}

434
	/* Group not yet part of this container, so acquire ownership */
435
	auto group = VfioGroup::attach(*this, index);
436
	if (not group) {
437
438
439
440
441
442
		logger->error("Failed to attach to IOMMU group: {}", index);
		throw std::exception();
	} else {
		logger->debug("Attached new group {} to VFIO container", index);
	}

443
	/* Push to our list */
444
445
446
447
448
449
450
451
	groups.push_back(std::move(group));

	return *groups.back();
}


VfioDevice::~VfioDevice()
{
452
453
	Logger logger = logging.get("kernel:vfio");

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
	logger->debug("Clean up device {} with fd {}", this->name, this->fd);

	for(auto& region : regions) {
		regionUnmap(region.index);
	}

	int ret = close(fd);
	if (ret != 0) {
		logger->error("Closing device fd {} failed", fd);
	}
}


bool
VfioDevice::reset()
{
	if (this->info.flags & VFIO_DEVICE_FLAGS_RESET)
		return ioctl(this->fd, VFIO_DEVICE_RESET) == 0;
	else
		return false; /* not supported by this device */
}


void*
VfioDevice::regionMap(size_t index)
{
	struct vfio_region_info *r = &regions[index];

	if (!(r->flags & VFIO_REGION_INFO_FLAG_MMAP))
		return MAP_FAILED;

485
486
487
488
489
490
	int flags = MAP_SHARED;

#ifndef __arm__
	flags |= MAP_SHARED;
#endif

491
492
	mappings[index] = mmap(nullptr, r->size,
	                       PROT_READ | PROT_WRITE,
493
	                       flags, fd, r->offset);
494
495
496
497
498
499
500
501
502
503
504

	return mappings[index];
}


bool
VfioDevice::regionUnmap(size_t index)
{
	int ret;
	struct vfio_region_info *r = &regions[index];

505
506
	Logger logger = logging.get("kernel:vfio");

507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
	if (!mappings[index])
		return false; /* was not mapped */

	logger->debug("Unmap region {} from device {}", index, name);

	ret = munmap(mappings[index], r->size);
	if (ret)
		return false;

	mappings[index] = nullptr;

	return true;
}


size_t
VfioDevice::regionGetSize(size_t index)
{
525
526
	Logger logger = logging.get("kernel:vfio");

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
	if(index >= regions.size()) {
		logger->error("Index out of range: {} >= {}", index, regions.size());
		throw std::out_of_range("Index out of range");
	}

	return regions[index].size;
}


bool
VfioDevice::pciEnable()
{
	int ret;
	uint32_t reg;
	const off_t offset = PCI_COMMAND +
	                     (static_cast<off_t>(VFIO_PCI_CONFIG_REGION_INDEX) << 40);

	/* Check if this is really a vfio-pci device */
	if (!(this->info.flags & VFIO_DEVICE_FLAGS_PCI))
		return false;

	ret = pread(this->fd, &reg, sizeof(reg), offset);
	if (ret != sizeof(reg))
		return false;

	/* Enable memory access and PCI bus mastering which is required for DMA */
	reg |= PCI_COMMAND_MEMORY | PCI_COMMAND_MASTER;

	ret = pwrite(this->fd, &reg, sizeof(reg), offset);
	if (ret != sizeof(reg))
		return false;

	return true;
}


bool
VfioDevice::pciHotReset()
{
566
567
	Logger logger = logging.get("kernel:vfio");

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
	/* Check if this is really a vfio-pci device */
	if (not isVfioPciDevice())
		return false;

	const size_t reset_infolen = sizeof(struct vfio_pci_hot_reset_info) +
	                             sizeof(struct vfio_pci_dependent_device) * 64;

	auto reset_info = reinterpret_cast<struct vfio_pci_hot_reset_info*>
	                    (calloc(1, reset_infolen));

	reset_info->argsz = reset_infolen;


	if (ioctl(this->fd, VFIO_DEVICE_GET_PCI_HOT_RESET_INFO, reset_info) != 0) {
		free(reset_info);
		return false;
	}

	logger->debug("Dependent devices for hot-reset:");
	for (size_t i = 0; i < reset_info->count; i++) {
		struct vfio_pci_dependent_device *dd = &reset_info->devices[i];
		logger->debug("  {:04x}:{:02x}:{:02x}.{:01x}: iommu_group={}",
		              dd->segment, dd->bus,
		              PCI_SLOT(dd->devfn), PCI_FUNC(dd->devfn), dd->group_id);

		if (static_cast<int>(dd->group_id) != this->group.index) {
			free(reset_info);
			return false;
		}
	}

	free(reset_info);

	const size_t resetlen = sizeof(struct vfio_pci_hot_reset) +
	                        sizeof(int32_t) * 1;
	auto reset = reinterpret_cast<struct vfio_pci_hot_reset*>
	                (calloc(1, resetlen));

	reset->argsz = resetlen;
	reset->count = 1;
	reset->group_fds[0] = this->group.fd;

	int ret = ioctl(this->fd, VFIO_DEVICE_PCI_HOT_RESET, reset);
	const bool success = (ret == 0);

	free(reset);

	if(not success and not group.container->isIommuEnabled()) {
		logger->info("PCI hot reset failed, but this is expected without IOMMU");
		return true;
	}

	return success;
}


int
VfioDevice::pciMsiInit(int efds[])
{
	/* Check if this is really a vfio-pci device */
	if(not isVfioPciDevice())
		return -1;

	const size_t irqCount = irqs[VFIO_PCI_MSI_IRQ_INDEX].count;
	const size_t irqSetSize = sizeof(struct vfio_irq_set) +
	                          sizeof(int) * irqCount;

	auto irqSet = reinterpret_cast<struct vfio_irq_set*>(calloc(1, irqSetSize));
	if(irqSet == nullptr)
		return -1;

	irqSet->argsz = irqSetSize;
	irqSet->flags = VFIO_IRQ_SET_DATA_EVENTFD | VFIO_IRQ_SET_ACTION_TRIGGER;
	irqSet->index = VFIO_PCI_MSI_IRQ_INDEX;
	irqSet->start = 0;
	irqSet->count = irqCount;

	/* Now set the new eventfds */
	for (size_t i = 0; i < irqCount; i++) {
		efds[i] = eventfd(0, 0);
		if (efds[i] < 0) {
			free(irqSet);
			return -1;
		}
	}

	memcpy(irqSet->data, efds, sizeof(int) * irqCount);

	if(ioctl(fd, VFIO_DEVICE_SET_IRQS, irqSet) != 0) {
		free(irqSet);
		return -1;
	}

	free(irqSet);

	return irqCount;
}


int
VfioDevice::pciMsiDeinit(int efds[])
{
	/* Check if this is really a vfio-pci device */
	if(not isVfioPciDevice())
		return -1;

	const size_t irqCount = irqs[VFIO_PCI_MSI_IRQ_INDEX].count;
	const size_t irqSetSize = sizeof(struct vfio_irq_set) +
	                            sizeof(int) * irqCount;

	auto irqSet = reinterpret_cast<struct vfio_irq_set*>(calloc(1, irqSetSize));
	if(irqSet == nullptr)
		return -1;

	irqSet->argsz = irqSetSize;
	irqSet->flags = VFIO_IRQ_SET_DATA_EVENTFD | VFIO_IRQ_SET_ACTION_TRIGGER;
	irqSet->index = VFIO_PCI_MSI_IRQ_INDEX;
	irqSet->count = irqCount;
	irqSet->start = 0;

	for (size_t i = 0; i < irqCount; i++) {
		close(efds[i]);
		efds[i] = -1;
	}

	memcpy(irqSet->data, efds, sizeof(int) * irqCount);

	if (ioctl(fd, VFIO_DEVICE_SET_IRQS, irqSet) != 0) {
		free(irqSet);
		return -1;
	}

	free(irqSet);

	return irqCount;
}


bool
VfioDevice::pciMsiFind(int nos[])
{
	int ret, idx, irq;
	char *end, *col, *last, line[1024], name[13];
	FILE *f;

	f = fopen("/proc/interrupts", "r");
	if (!f)
		return false;

	for (int i = 0; i < 32; i++)
		nos[i] = -1;

	/* For each line in /proc/interrupts */
	while (fgets(line, sizeof(line), f)) {
		col = strtok(line, " ");

		/* IRQ number is in first column */
		irq = strtol(col, &end, 10);
		if (col == end)
			continue;

		/* Find last column of line */
		do {
			last = col;
		} while ((col = strtok(nullptr, " ")));


		ret = sscanf(last, "vfio-msi[%u](%12[0-9:])", &idx, name);
		if (ret == 2) {
			if (strstr(this->name.c_str(), name) == this->name.c_str())
				nos[idx] = irq;
		}
	}

	fclose(f);

	return true;
}


bool
VfioDevice::isVfioPciDevice() const
{
	return info.flags & VFIO_DEVICE_FLAGS_PCI;
}


VfioGroup::~VfioGroup()
{
757
758
	Logger logger = logging.get("kernel:vfio");

759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
	logger->debug("Clean up group {} with fd {}", this->index, this->fd);

	/* Release memory and close fds */
	devices.clear();

	if(fd < 0) {
		logger->debug("Destructing group that has not been attached");
	} else {
		int ret = ioctl(fd, VFIO_GROUP_UNSET_CONTAINER);
		if (ret != 0) {
			logger->error("Cannot unset container for group fd {}", fd);
		}

		ret = close(fd);
		if (ret != 0) {
			logger->error("Cannot close group fd {}", fd);
		}
	}
}


std::unique_ptr<VfioGroup>
VfioGroup::attach(VfioContainer& container, int groupIndex)
{
783
784
	Logger logger = logging.get("kernel:vfio");

785
786
787
788
789
790
791
792
793
794
	std::unique_ptr<VfioGroup> group { new VfioGroup(groupIndex) };

	group->container = &container;

	/* Open group fd */
	std::stringstream groupPath;
	groupPath << VFIO_PATH
	          << (container.isIommuEnabled() ? "" : "noiommu-")
	          << groupIndex;

795
	logger->debug("path: {}", groupPath.str().c_str());
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
	group->fd = open(groupPath.str().c_str(), O_RDWR);
	if (group->fd < 0) {
		logger->error("Failed to open VFIO group {}", group->index);
		return nullptr;
	}

	logger->debug("VFIO group {} (fd {}) has path {}",
	              groupIndex, group->fd, groupPath.str());

	/* Claim group ownership */
	int ret = ioctl(group->fd, VFIO_GROUP_SET_CONTAINER, &container.getFd());
	if (ret < 0) {
		logger->error("Failed to attach VFIO group {} to container fd {} (error {})",
		              group->index, container.getFd(), ret);
		return nullptr;
	}

	/* Set IOMMU type */
	int iommu_type = container.isIommuEnabled() ?
	                     VFIO_TYPE1_IOMMU :
	                     VFIO_NOIOMMU_IOMMU;

	ret = ioctl(container.getFd(), VFIO_SET_IOMMU, iommu_type);
	if (ret < 0) {
		logger->error("Failed to set IOMMU type of container: {}", ret);
		return nullptr;
	}

	/* Check group viability and features */
	group->status.argsz = sizeof(group->status);

	ret = ioctl(group->fd, VFIO_GROUP_GET_STATUS, &group->status);
	if (ret < 0) {
		logger->error("Failed to get VFIO group status");
		return nullptr;
	}

	if (!(group->status.flags & VFIO_GROUP_FLAGS_VIABLE)) {
		logger->error("VFIO group is not available: bind all devices to the VFIO driver!");
		return nullptr;
	}

	return group;
}

841
} /* namespace villas */
842