device.c 17.5 KB
Newer Older
1
/*
2
 * Copyright(c) 2016 - 2017 Intel Corporation. All rights reserved.
3 4 5 6 7 8 9 10 11 12 13 14 15 16
 *
 * This program is free software; you can redistribute it and/or modify
 * it under the terms of version 2 of the GNU General Public License as
 * published by the Free Software Foundation.
 *
 * 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.
 */
#include <linux/pagemap.h>
#include <linux/module.h>
#include <linux/device.h>
#include <linux/pfn_t.h>
D
Dan Williams 已提交
17
#include <linux/cdev.h>
18 19 20 21
#include <linux/slab.h>
#include <linux/dax.h>
#include <linux/fs.h>
#include <linux/mm.h>
22
#include <linux/mman.h>
23
#include "dax-private.h"
24
#include "dax.h"
25 26 27

static struct class *dax_class;

28 29 30 31 32
/*
 * Rely on the fact that drvdata is set before the attributes are
 * registered, and that the attributes are unregistered before drvdata
 * is cleared to assume that drvdata is always valid.
 */
33 34 35
static ssize_t id_show(struct device *dev,
		struct device_attribute *attr, char *buf)
{
36
	struct dax_region *dax_region = dev_get_drvdata(dev);
37

38
	return sprintf(buf, "%d\n", dax_region->id);
39 40 41 42 43 44
}
static DEVICE_ATTR_RO(id);

static ssize_t region_size_show(struct device *dev,
		struct device_attribute *attr, char *buf)
{
45
	struct dax_region *dax_region = dev_get_drvdata(dev);
46

47 48
	return sprintf(buf, "%llu\n", (unsigned long long)
			resource_size(&dax_region->res));
49 50 51 52 53 54 55
}
static struct device_attribute dev_attr_region_size = __ATTR(size, 0444,
		region_size_show, NULL);

static ssize_t align_show(struct device *dev,
		struct device_attribute *attr, char *buf)
{
56
	struct dax_region *dax_region = dev_get_drvdata(dev);
57

58
	return sprintf(buf, "%u\n", dax_region->align);
59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78
}
static DEVICE_ATTR_RO(align);

static struct attribute *dax_region_attributes[] = {
	&dev_attr_region_size.attr,
	&dev_attr_align.attr,
	&dev_attr_id.attr,
	NULL,
};

static const struct attribute_group dax_region_attribute_group = {
	.name = "dax_region",
	.attrs = dax_region_attributes,
};

static const struct attribute_group *dax_region_attribute_groups[] = {
	&dax_region_attribute_group,
	NULL,
};

79 80 81 82 83 84 85 86 87 88 89 90 91 92
static void dax_region_free(struct kref *kref)
{
	struct dax_region *dax_region;

	dax_region = container_of(kref, struct dax_region, kref);
	kfree(dax_region);
}

void dax_region_put(struct dax_region *dax_region)
{
	kref_put(&dax_region->kref, dax_region_free);
}
EXPORT_SYMBOL_GPL(dax_region_put);

93 94 95 96 97 98 99 100 101
static void dax_region_unregister(void *region)
{
	struct dax_region *dax_region = region;

	sysfs_remove_groups(&dax_region->dev->kobj,
			dax_region_attribute_groups);
	dax_region_put(dax_region);
}

102
struct dax_region *alloc_dax_region(struct device *parent, int region_id,
D
Dan Williams 已提交
103
		struct resource *res, unsigned int align,
104 105 106 107
		unsigned long pfn_flags)
{
	struct dax_region *dax_region;

108 109 110 111 112 113 114 115 116 117
	/*
	 * The DAX core assumes that it can store its private data in
	 * parent->driver_data. This WARN is a reminder / safeguard for
	 * developers of device-dax drivers.
	 */
	if (dev_get_drvdata(parent)) {
		dev_WARN(parent, "dax core failed to setup private data\n");
		return NULL;
	}

118 119 120
	if (!IS_ALIGNED(res->start, align)
			|| !IS_ALIGNED(resource_size(res), align))
		return NULL;
121

122
	dax_region = kzalloc(sizeof(*dax_region), GFP_KERNEL);
123 124 125
	if (!dax_region)
		return NULL;

126
	dev_set_drvdata(parent, dax_region);
127 128 129 130 131 132
	memcpy(&dax_region->res, res, sizeof(*res));
	dax_region->pfn_flags = pfn_flags;
	kref_init(&dax_region->kref);
	dax_region->id = region_id;
	dax_region->align = align;
	dax_region->dev = parent;
133 134
	if (sysfs_create_groups(&parent->kobj, dax_region_attribute_groups)) {
		kfree(dax_region);
135
		return NULL;
136
	}
137

138 139 140
	kref_get(&dax_region->kref);
	if (devm_add_action_or_reset(parent, dax_region_unregister, dax_region))
		return NULL;
141 142 143 144
	return dax_region;
}
EXPORT_SYMBOL_GPL(alloc_dax_region);

145
static struct dev_dax *to_dev_dax(struct device *dev)
146
{
147
	return container_of(dev, struct dev_dax, dev);
148 149
}

150 151 152
static ssize_t size_show(struct device *dev,
		struct device_attribute *attr, char *buf)
{
153
	struct dev_dax *dev_dax = to_dev_dax(dev);
154 155 156
	unsigned long long size = 0;
	int i;

157 158
	for (i = 0; i < dev_dax->num_resources; i++)
		size += resource_size(&dev_dax->res[i]);
159 160 161 162 163

	return sprintf(buf, "%llu\n", size);
}
static DEVICE_ATTR_RO(size);

164
static struct attribute *dev_dax_attributes[] = {
165 166 167 168
	&dev_attr_size.attr,
	NULL,
};

169 170
static const struct attribute_group dev_dax_attribute_group = {
	.attrs = dev_dax_attributes,
171 172 173
};

static const struct attribute_group *dax_attribute_groups[] = {
174
	&dev_dax_attribute_group,
175 176 177
	NULL,
};

178
static int check_vma(struct dev_dax *dev_dax, struct vm_area_struct *vma,
179 180
		const char *func)
{
181 182
	struct dax_region *dax_region = dev_dax->region;
	struct device *dev = &dev_dax->dev;
183 184
	unsigned long mask;

185
	if (!dax_alive(dev_dax->dax_dev))
186 187
		return -ENXIO;

188
	/* prevent private mappings from being established */
189
	if ((vma->vm_flags & VM_MAYSHARE) != VM_MAYSHARE) {
190 191
		dev_info_ratelimited(dev,
				"%s: %s: fail, attempted private mapping\n",
192 193 194 195 196 197
				current->comm, func);
		return -EINVAL;
	}

	mask = dax_region->align - 1;
	if (vma->vm_start & mask || vma->vm_end & mask) {
198 199
		dev_info_ratelimited(dev,
				"%s: %s: fail, unaligned vma (%#lx - %#lx, %#lx)\n",
200 201 202 203 204 205 206
				current->comm, func, vma->vm_start, vma->vm_end,
				mask);
		return -EINVAL;
	}

	if ((dax_region->pfn_flags & (PFN_DEV|PFN_MAP)) == PFN_DEV
			&& (vma->vm_flags & VM_DONTCOPY) == 0) {
207 208
		dev_info_ratelimited(dev,
				"%s: %s: fail, dax range requires MADV_DONTFORK\n",
209 210 211 212 213
				current->comm, func);
		return -EINVAL;
	}

	if (!vma_is_dax(vma)) {
214 215
		dev_info_ratelimited(dev,
				"%s: %s: fail, vma is not DAX capable\n",
216 217 218 219 220 221 222
				current->comm, func);
		return -EINVAL;
	}

	return 0;
}

223
/* see "strong" declaration in tools/testing/nvdimm/dax-dev.c */
224
__weak phys_addr_t dax_pgoff_to_phys(struct dev_dax *dev_dax, pgoff_t pgoff,
225 226 227
		unsigned long size)
{
	struct resource *res;
228 229
	/* gcc-4.6.3-nolibc for i386 complains that this is uninitialized */
	phys_addr_t uninitialized_var(phys);
230 231
	int i;

232 233
	for (i = 0; i < dev_dax->num_resources; i++) {
		res = &dev_dax->res[i];
234 235 236 237 238 239
		phys = pgoff * PAGE_SIZE + res->start;
		if (phys >= res->start && phys <= res->end)
			break;
		pgoff -= PHYS_PFN(resource_size(res));
	}

240 241
	if (i < dev_dax->num_resources) {
		res = &dev_dax->res[i];
242 243 244 245 246 247 248
		if (phys + size - 1 <= res->end)
			return phys;
	}

	return -1;
}

249
static vm_fault_t __dev_dax_pte_fault(struct dev_dax *dev_dax,
D
Dan Williams 已提交
250
				struct vm_fault *vmf, pfn_t *pfn)
251
{
252
	struct device *dev = &dev_dax->dev;
253 254
	struct dax_region *dax_region;
	phys_addr_t phys;
255
	unsigned int fault_size = PAGE_SIZE;
256

257
	if (check_vma(dev_dax, vmf->vma, __func__))
258 259
		return VM_FAULT_SIGBUS;

260
	dax_region = dev_dax->region;
261
	if (dax_region->align > PAGE_SIZE) {
262 263
		dev_dbg(dev, "alignment (%#x) > fault size (%#x)\n",
			dax_region->align, fault_size);
264 265 266
		return VM_FAULT_SIGBUS;
	}

267 268 269
	if (fault_size != dax_region->align)
		return VM_FAULT_SIGBUS;

270
	phys = dax_pgoff_to_phys(dev_dax, vmf->pgoff, PAGE_SIZE);
271
	if (phys == -1) {
272
		dev_dbg(dev, "pgoff_to_phys(%#lx) failed\n", vmf->pgoff);
273 274 275
		return VM_FAULT_SIGBUS;
	}

D
Dan Williams 已提交
276
	*pfn = phys_to_pfn_t(phys, dax_region->pfn_flags);
277

D
Dan Williams 已提交
278
	return vmf_insert_mixed(vmf->vma, vmf->address, *pfn);
279 280
}

281
static vm_fault_t __dev_dax_pmd_fault(struct dev_dax *dev_dax,
D
Dan Williams 已提交
282
				struct vm_fault *vmf, pfn_t *pfn)
283
{
284
	unsigned long pmd_addr = vmf->address & PMD_MASK;
285
	struct device *dev = &dev_dax->dev;
286 287 288
	struct dax_region *dax_region;
	phys_addr_t phys;
	pgoff_t pgoff;
289
	unsigned int fault_size = PMD_SIZE;
290

291
	if (check_vma(dev_dax, vmf->vma, __func__))
292 293
		return VM_FAULT_SIGBUS;

294
	dax_region = dev_dax->region;
295
	if (dax_region->align > PMD_SIZE) {
296 297
		dev_dbg(dev, "alignment (%#x) > fault size (%#x)\n",
			dax_region->align, fault_size);
298 299 300 301 302
		return VM_FAULT_SIGBUS;
	}

	/* dax pmd mappings require pfn_t_devmap() */
	if ((dax_region->pfn_flags & (PFN_DEV|PFN_MAP)) != (PFN_DEV|PFN_MAP)) {
303
		dev_dbg(dev, "region lacks devmap flags\n");
304 305 306
		return VM_FAULT_SIGBUS;
	}

307 308 309 310 311 312 313 314 315 316
	if (fault_size < dax_region->align)
		return VM_FAULT_SIGBUS;
	else if (fault_size > dax_region->align)
		return VM_FAULT_FALLBACK;

	/* if we are outside of the VMA */
	if (pmd_addr < vmf->vma->vm_start ||
			(pmd_addr + PMD_SIZE) > vmf->vma->vm_end)
		return VM_FAULT_SIGBUS;

317
	pgoff = linear_page_index(vmf->vma, pmd_addr);
318
	phys = dax_pgoff_to_phys(dev_dax, pgoff, PMD_SIZE);
319
	if (phys == -1) {
320
		dev_dbg(dev, "pgoff_to_phys(%#lx) failed\n", pgoff);
321 322 323
		return VM_FAULT_SIGBUS;
	}

D
Dan Williams 已提交
324
	*pfn = phys_to_pfn_t(phys, dax_region->pfn_flags);
325

326
	return vmf_insert_pfn_pmd(vmf, *pfn, vmf->flags & FAULT_FLAG_WRITE);
327 328
}

329
#ifdef CONFIG_HAVE_ARCH_TRANSPARENT_HUGEPAGE_PUD
330
static vm_fault_t __dev_dax_pud_fault(struct dev_dax *dev_dax,
D
Dan Williams 已提交
331
				struct vm_fault *vmf, pfn_t *pfn)
332 333
{
	unsigned long pud_addr = vmf->address & PUD_MASK;
334
	struct device *dev = &dev_dax->dev;
335 336 337
	struct dax_region *dax_region;
	phys_addr_t phys;
	pgoff_t pgoff;
338 339
	unsigned int fault_size = PUD_SIZE;

340

341
	if (check_vma(dev_dax, vmf->vma, __func__))
342 343
		return VM_FAULT_SIGBUS;

344
	dax_region = dev_dax->region;
345
	if (dax_region->align > PUD_SIZE) {
346 347
		dev_dbg(dev, "alignment (%#x) > fault size (%#x)\n",
			dax_region->align, fault_size);
348 349 350 351 352
		return VM_FAULT_SIGBUS;
	}

	/* dax pud mappings require pfn_t_devmap() */
	if ((dax_region->pfn_flags & (PFN_DEV|PFN_MAP)) != (PFN_DEV|PFN_MAP)) {
353
		dev_dbg(dev, "region lacks devmap flags\n");
354 355 356
		return VM_FAULT_SIGBUS;
	}

357 358 359 360 361 362 363 364 365 366
	if (fault_size < dax_region->align)
		return VM_FAULT_SIGBUS;
	else if (fault_size > dax_region->align)
		return VM_FAULT_FALLBACK;

	/* if we are outside of the VMA */
	if (pud_addr < vmf->vma->vm_start ||
			(pud_addr + PUD_SIZE) > vmf->vma->vm_end)
		return VM_FAULT_SIGBUS;

367
	pgoff = linear_page_index(vmf->vma, pud_addr);
368
	phys = dax_pgoff_to_phys(dev_dax, pgoff, PUD_SIZE);
369
	if (phys == -1) {
370
		dev_dbg(dev, "pgoff_to_phys(%#lx) failed\n", pgoff);
371 372 373
		return VM_FAULT_SIGBUS;
	}

D
Dan Williams 已提交
374
	*pfn = phys_to_pfn_t(phys, dax_region->pfn_flags);
375

376
	return vmf_insert_pfn_pud(vmf, *pfn, vmf->flags & FAULT_FLAG_WRITE);
377 378
}
#else
379
static vm_fault_t __dev_dax_pud_fault(struct dev_dax *dev_dax,
D
Dan Williams 已提交
380
				struct vm_fault *vmf, pfn_t *pfn)
381 382 383 384 385
{
	return VM_FAULT_FALLBACK;
}
#endif /* !CONFIG_HAVE_ARCH_TRANSPARENT_HUGEPAGE_PUD */

386
static vm_fault_t dev_dax_huge_fault(struct vm_fault *vmf,
387
		enum page_entry_size pe_size)
388
{
389
	struct file *filp = vmf->vma->vm_file;
D
Dan Williams 已提交
390
	unsigned long fault_size;
391 392
	vm_fault_t rc = VM_FAULT_SIGBUS;
	int id;
D
Dan Williams 已提交
393
	pfn_t pfn;
394
	struct dev_dax *dev_dax = filp->private_data;
395

396 397
	dev_dbg(&dev_dax->dev, "%s: %s (%#lx - %#lx) size = %d\n", current->comm,
			(vmf->flags & FAULT_FLAG_WRITE) ? "write" : "read",
398
			vmf->vma->vm_start, vmf->vma->vm_end, pe_size);
399

400
	id = dax_read_lock();
401 402
	switch (pe_size) {
	case PE_SIZE_PTE:
D
Dan Williams 已提交
403 404
		fault_size = PAGE_SIZE;
		rc = __dev_dax_pte_fault(dev_dax, vmf, &pfn);
405
		break;
406
	case PE_SIZE_PMD:
D
Dan Williams 已提交
407 408
		fault_size = PMD_SIZE;
		rc = __dev_dax_pmd_fault(dev_dax, vmf, &pfn);
409
		break;
410
	case PE_SIZE_PUD:
D
Dan Williams 已提交
411 412
		fault_size = PUD_SIZE;
		rc = __dev_dax_pud_fault(dev_dax, vmf, &pfn);
413 414
		break;
	default:
415
		rc = VM_FAULT_SIGBUS;
416
	}
D
Dan Williams 已提交
417 418 419

	if (rc == VM_FAULT_NOPAGE) {
		unsigned long i;
D
Dan Williams 已提交
420
		pgoff_t pgoff;
D
Dan Williams 已提交
421 422 423 424 425 426 427

		/*
		 * In the device-dax case the only possibility for a
		 * VM_FAULT_NOPAGE result is when device-dax capacity is
		 * mapped. No need to consider the zero page, or racing
		 * conflicting mappings.
		 */
D
Dan Williams 已提交
428 429
		pgoff = linear_page_index(vmf->vma, vmf->address
				& ~(fault_size - 1));
D
Dan Williams 已提交
430 431 432 433 434 435 436
		for (i = 0; i < fault_size / PAGE_SIZE; i++) {
			struct page *page;

			page = pfn_to_page(pfn_t_to_pfn(pfn) + i);
			if (page->mapping)
				continue;
			page->mapping = filp->f_mapping;
D
Dan Williams 已提交
437
			page->index = pgoff + i;
D
Dan Williams 已提交
438 439
		}
	}
440
	dax_read_unlock(id);
441 442 443 444

	return rc;
}

445
static vm_fault_t dev_dax_fault(struct vm_fault *vmf)
446
{
447
	return dev_dax_huge_fault(vmf, PE_SIZE_PTE);
448 449
}

450 451 452 453 454 455 456 457 458 459 460
static int dev_dax_split(struct vm_area_struct *vma, unsigned long addr)
{
	struct file *filp = vma->vm_file;
	struct dev_dax *dev_dax = filp->private_data;
	struct dax_region *dax_region = dev_dax->region;

	if (!IS_ALIGNED(addr, dax_region->align))
		return -EINVAL;
	return 0;
}

461 462 463 464 465 466 467 468 469
static unsigned long dev_dax_pagesize(struct vm_area_struct *vma)
{
	struct file *filp = vma->vm_file;
	struct dev_dax *dev_dax = filp->private_data;
	struct dax_region *dax_region = dev_dax->region;

	return dax_region->align;
}

470 471 472
static const struct vm_operations_struct dax_vm_ops = {
	.fault = dev_dax_fault,
	.huge_fault = dev_dax_huge_fault,
473
	.split = dev_dax_split,
474
	.pagesize = dev_dax_pagesize,
475 476
};

477
static int dax_mmap(struct file *filp, struct vm_area_struct *vma)
478
{
479
	struct dev_dax *dev_dax = filp->private_data;
480
	int rc, id;
481

482
	dev_dbg(&dev_dax->dev, "trace\n");
483

484 485 486 487 488
	/*
	 * We lock to check dax_dev liveness and will re-check at
	 * fault time.
	 */
	id = dax_read_lock();
489
	rc = check_vma(dev_dax, vma, __func__);
490
	dax_read_unlock(id);
491 492 493
	if (rc)
		return rc;

494
	vma->vm_ops = &dax_vm_ops;
495
	vma->vm_flags |= VM_HUGEPAGE;
496
	return 0;
497 498 499
}

/* return an unmapped area aligned to the dax region specified alignment */
500
static unsigned long dax_get_unmapped_area(struct file *filp,
501 502 503 504
		unsigned long addr, unsigned long len, unsigned long pgoff,
		unsigned long flags)
{
	unsigned long off, off_end, off_align, len_align, addr_align, align;
505
	struct dev_dax *dev_dax = filp ? filp->private_data : NULL;
506 507
	struct dax_region *dax_region;

508
	if (!dev_dax || addr)
509 510
		goto out;

511
	dax_region = dev_dax->region;
512 513 514 515 516 517 518 519 520 521 522
	align = dax_region->align;
	off = pgoff << PAGE_SHIFT;
	off_end = off + len;
	off_align = round_up(off, align);

	if ((off_end <= off_align) || ((off_end - off_align) < align))
		goto out;

	len_align = len + align;
	if ((off + len_align) < off)
		goto out;
523

524 525 526 527 528 529 530 531 532 533
	addr_align = current->mm->get_unmapped_area(filp, addr, len_align,
			pgoff, flags);
	if (!IS_ERR_VALUE(addr_align)) {
		addr_align += (off - addr_align) & (align - 1);
		return addr_align;
	}
 out:
	return current->mm->get_unmapped_area(filp, addr, len, pgoff, flags);
}

534 535 536 537 538
static const struct address_space_operations dev_dax_aops = {
	.set_page_dirty		= noop_set_page_dirty,
	.invalidatepage		= noop_invalidatepage,
};

539
static int dax_open(struct inode *inode, struct file *filp)
540
{
541 542 543
	struct dax_device *dax_dev = inode_dax(inode);
	struct inode *__dax_inode = dax_inode(dax_dev);
	struct dev_dax *dev_dax = dax_get_private(dax_dev);
544

545
	dev_dbg(&dev_dax->dev, "trace\n");
546 547
	inode->i_mapping = __dax_inode->i_mapping;
	inode->i_mapping->host = __dax_inode;
548
	inode->i_mapping->a_ops = &dev_dax_aops;
549
	filp->f_mapping = inode->i_mapping;
550
	filp->f_wb_err = filemap_sample_wb_err(filp->f_mapping);
551
	filp->private_data = dev_dax;
552
	inode->i_flags = S_DAX;
553 554 555

	return 0;
}
556

557
static int dax_release(struct inode *inode, struct file *filp)
558
{
559
	struct dev_dax *dev_dax = filp->private_data;
560

561
	dev_dbg(&dev_dax->dev, "trace\n");
562
	return 0;
563 564
}

565 566 567
static const struct file_operations dax_fops = {
	.llseek = noop_llseek,
	.owner = THIS_MODULE,
568 569 570 571
	.open = dax_open,
	.release = dax_release,
	.get_unmapped_area = dax_get_unmapped_area,
	.mmap = dax_mmap,
572
	.mmap_supported_flags = MAP_SYNC,
573 574
};

575
static void dev_dax_release(struct device *dev)
576
{
577 578
	struct dev_dax *dev_dax = to_dev_dax(dev);
	struct dax_region *dax_region = dev_dax->region;
579
	struct dax_device *dax_dev = dev_dax->dax_dev;
580

581
	dax_region_put(dax_region);
582
	put_dax(dax_dev);
583
	kfree(dev_dax);
584 585
}

586
static void kill_dev_dax(struct dev_dax *dev_dax)
587
{
588 589 590 591 592
	struct dax_device *dax_dev = dev_dax->dax_dev;
	struct inode *inode = dax_inode(dax_dev);

	kill_dax(dax_dev);
	unmap_mapping_range(inode->i_mapping, 0, 0, 1);
593 594
}

595
static void unregister_dev_dax(void *dev)
596
{
597
	struct dev_dax *dev_dax = to_dev_dax(dev);
598 599 600
	struct dax_device *dax_dev = dev_dax->dax_dev;
	struct inode *inode = dax_inode(dax_dev);
	struct cdev *cdev = inode->i_cdev;
601

602
	dev_dbg(dev, "trace\n");
603

604
	kill_dev_dax(dev_dax);
605
	cdev_device_del(cdev, dev);
606
	put_device(dev);
607 608
}

609
struct dev_dax *devm_create_dev_dax(struct dax_region *dax_region,
610
		int id, struct resource *res, int count)
611 612
{
	struct device *parent = dax_region->dev;
613
	struct dax_device *dax_dev;
614
	struct dev_dax *dev_dax;
615
	struct inode *inode;
616
	struct device *dev;
D
Dan Williams 已提交
617
	struct cdev *cdev;
618 619 620 621
	int rc, i;

	if (!count)
		return ERR_PTR(-EINVAL);
622

623
	dev_dax = kzalloc(struct_size(dev_dax, res, count), GFP_KERNEL);
624
	if (!dev_dax)
625
		return ERR_PTR(-ENOMEM);
626

627 628 629 630 631 632 633
	for (i = 0; i < count; i++) {
		if (!IS_ALIGNED(res[i].start, dax_region->align)
				|| !IS_ALIGNED(resource_size(&res[i]),
					dax_region->align)) {
			rc = -EINVAL;
			break;
		}
634 635
		dev_dax->res[i].start = res[i].start;
		dev_dax->res[i].end = res[i].end;
636 637 638
	}

	if (i < count)
D
Dan Williams 已提交
639
		goto err;
640

D
Dan Williams 已提交
641 642 643 644 645
	/*
	 * No 'host' or dax_operations since there is no access to this
	 * device outside of mmap of the resulting character device.
	 */
	dax_dev = alloc_dax(dev_dax, NULL, NULL);
646 647
	if (!dax_dev) {
		rc = -ENOMEM;
D
Dan Williams 已提交
648
		goto err;
649
	}
650

651
	/* from here on we're committed to teardown via dax_dev_release() */
652
	dev = &dev_dax->dev;
653
	device_initialize(dev);
D
Dan Williams 已提交
654

655 656
	inode = dax_inode(dax_dev);
	cdev = inode->i_cdev;
D
Dan Williams 已提交
657 658 659
	cdev_init(cdev, &dax_fops);
	cdev->owner = parent->driver->owner;

660
	dev_dax->num_resources = count;
661
	dev_dax->dax_dev = dax_dev;
662
	dev_dax->region = dax_region;
D
Dan Williams 已提交
663 664
	kref_get(&dax_region->kref);

665
	dev->devt = inode->i_rdev;
666 667 668
	dev->class = dax_class;
	dev->parent = parent;
	dev->groups = dax_attribute_groups;
669
	dev->release = dev_dax_release;
670
	dev_set_name(dev, "dax%d.%d", dax_region->id, id);
671 672

	rc = cdev_device_add(cdev, dev);
673
	if (rc) {
674
		kill_dev_dax(dev_dax);
675
		put_device(dev);
676
		return ERR_PTR(rc);
677
	}
678

679
	rc = devm_add_action_or_reset(dax_region->dev, unregister_dev_dax, dev);
680 681 682
	if (rc)
		return ERR_PTR(rc);

683
	return dev_dax;
684

D
Dan Williams 已提交
685
 err:
686
	kfree(dev_dax);
687

688
	return ERR_PTR(rc);
689
}
690
EXPORT_SYMBOL_GPL(devm_create_dev_dax);
691

692 693 694
static int __init dax_init(void)
{
	dax_class = class_create(THIS_MODULE, "dax");
695
	return PTR_ERR_OR_ZERO(dax_class);
696 697 698 699 700 701 702 703 704 705 706
}

static void __exit dax_exit(void)
{
	class_destroy(dax_class);
}

MODULE_AUTHOR("Intel Corporation");
MODULE_LICENSE("GPL v2");
subsys_initcall(dax_init);
module_exit(dax_exit);