[linux-block.git] / drivers / dax / device.c

// SPDX-License-Identifier: GPL-2.0
/* Copyright(c) 2016-2018 Intel Corporation. All rights reserved. */
#include <linux/memremap.h>
#include <linux/pagemap.h>
#include <linux/module.h>
#include <linux/device.h>
#include <linux/pfn_t.h>
#include <linux/cdev.h>
#include <linux/slab.h>
#include <linux/dax.h>
#include <linux/fs.h>
#include <linux/mm.h>
#include <linux/mman.h>
#include "dax-private.h"
#include "bus.h"

static int check_vma(struct dev_dax *dev_dax, struct vm_area_struct *vma,
		const char *func)
{
	struct dax_region *dax_region = dev_dax->region;
	struct device *dev = &dev_dax->dev;
	unsigned long mask;

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

	/* prevent private mappings from being established */
	if ((vma->vm_flags & VM_MAYSHARE) != VM_MAYSHARE) {
		dev_info_ratelimited(dev,
				"%s: %s: fail, attempted private mapping\n",
				current->comm, func);
		return -EINVAL;
	}

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

	if (!vma_is_dax(vma)) {
		dev_info_ratelimited(dev,
				"%s: %s: fail, vma is not DAX capable\n",
				current->comm, func);
		return -EINVAL;
	}

	return 0;
}

/* see "strong" declaration in tools/testing/nvdimm/dax-dev.c */
__weak phys_addr_t dax_pgoff_to_phys(struct dev_dax *dev_dax, pgoff_t pgoff,
		unsigned long size)
{
	struct resource *res = &dev_dax->region->res;
	phys_addr_t phys;

	phys = pgoff * PAGE_SIZE + res->start;
	if (phys >= res->start && phys <= res->end) {
		if (phys + size - 1 <= res->end)
			return phys;
	}

	return -1;
}

static vm_fault_t __dev_dax_pte_fault(struct dev_dax *dev_dax,
				struct vm_fault *vmf, pfn_t *pfn)
{
	struct device *dev = &dev_dax->dev;
	struct dax_region *dax_region;
	phys_addr_t phys;
	unsigned int fault_size = PAGE_SIZE;

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

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

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

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

	*pfn = phys_to_pfn_t(phys, PFN_DEV|PFN_MAP);

	return vmf_insert_mixed(vmf->vma, vmf->address, *pfn);
}

static vm_fault_t __dev_dax_pmd_fault(struct dev_dax *dev_dax,
				struct vm_fault *vmf, pfn_t *pfn)
{
	unsigned long pmd_addr = vmf->address & PMD_MASK;
	struct device *dev = &dev_dax->dev;
	struct dax_region *dax_region;
	phys_addr_t phys;
	pgoff_t pgoff;
	unsigned int fault_size = PMD_SIZE;

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

	dax_region = dev_dax->region;
	if (dax_region->align > PMD_SIZE) {
		dev_dbg(dev, "alignment (%#x) > fault size (%#x)\n",
			dax_region->align, fault_size);
		return VM_FAULT_SIGBUS;
	}

	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;

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

	*pfn = phys_to_pfn_t(phys, PFN_DEV|PFN_MAP);

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

#ifdef CONFIG_HAVE_ARCH_TRANSPARENT_HUGEPAGE_PUD
static vm_fault_t __dev_dax_pud_fault(struct dev_dax *dev_dax,
				struct vm_fault *vmf, pfn_t *pfn)
{
	unsigned long pud_addr = vmf->address & PUD_MASK;
	struct device *dev = &dev_dax->dev;
	struct dax_region *dax_region;
	phys_addr_t phys;
	pgoff_t pgoff;
	unsigned int fault_size = PUD_SIZE;


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

	dax_region = dev_dax->region;
	if (dax_region->align > PUD_SIZE) {
		dev_dbg(dev, "alignment (%#x) > fault size (%#x)\n",
			dax_region->align, fault_size);
		return VM_FAULT_SIGBUS;
	}

	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;

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

	*pfn = phys_to_pfn_t(phys, PFN_DEV|PFN_MAP);

	return vmf_insert_pfn_pud(vmf, *pfn, vmf->flags & FAULT_FLAG_WRITE);
}
#else
static vm_fault_t __dev_dax_pud_fault(struct dev_dax *dev_dax,
				struct vm_fault *vmf, pfn_t *pfn)
{
	return VM_FAULT_FALLBACK;
}
#endif /* !CONFIG_HAVE_ARCH_TRANSPARENT_HUGEPAGE_PUD */

static vm_fault_t dev_dax_huge_fault(struct vm_fault *vmf,
		enum page_entry_size pe_size)
{
	struct file *filp = vmf->vma->vm_file;
	unsigned long fault_size;
	vm_fault_t rc = VM_FAULT_SIGBUS;
	int id;
	pfn_t pfn;
	struct dev_dax *dev_dax = filp->private_data;

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

	id = dax_read_lock();
	switch (pe_size) {
	case PE_SIZE_PTE:
		fault_size = PAGE_SIZE;
		rc = __dev_dax_pte_fault(dev_dax, vmf, &pfn);
		break;
	case PE_SIZE_PMD:
		fault_size = PMD_SIZE;
		rc = __dev_dax_pmd_fault(dev_dax, vmf, &pfn);
		break;
	case PE_SIZE_PUD:
		fault_size = PUD_SIZE;
		rc = __dev_dax_pud_fault(dev_dax, vmf, &pfn);
		break;
	default:
		rc = VM_FAULT_SIGBUS;
	}

	if (rc == VM_FAULT_NOPAGE) {
		unsigned long i;
		pgoff_t pgoff;

		/*
		 * 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.
		 */
		pgoff = linear_page_index(vmf->vma, vmf->address
				& ~(fault_size - 1));
		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;
			page->index = pgoff + i;
		}
	}
	dax_read_unlock(id);

	return rc;
}

static vm_fault_t dev_dax_fault(struct vm_fault *vmf)
{
	return dev_dax_huge_fault(vmf, PE_SIZE_PTE);
}

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;
}

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;
}

static const struct vm_operations_struct dax_vm_ops = {
	.fault = dev_dax_fault,
	.huge_fault = dev_dax_huge_fault,
	.split = dev_dax_split,
	.pagesize = dev_dax_pagesize,
};

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

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

	/*
	 * We lock to check dax_dev liveness and will re-check at
	 * fault time.
	 */
	id = dax_read_lock();
	rc = check_vma(dev_dax, vma, __func__);
	dax_read_unlock(id);
	if (rc)
		return rc;

	vma->vm_ops = &dax_vm_ops;
	vma->vm_flags |= VM_HUGEPAGE;
	return 0;
}

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

	if (!dev_dax || addr)
		goto out;

	dax_region = dev_dax->region;
	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;

	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);
}

static const struct address_space_operations dev_dax_aops = {
	.set_page_dirty		= noop_set_page_dirty,
	.invalidatepage		= noop_invalidatepage,
};

static int dax_open(struct inode *inode, struct file *filp)
{
	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);

	dev_dbg(&dev_dax->dev, "trace\n");
	inode->i_mapping = __dax_inode->i_mapping;
	inode->i_mapping->host = __dax_inode;
	inode->i_mapping->a_ops = &dev_dax_aops;
	filp->f_mapping = inode->i_mapping;
	filp->f_wb_err = filemap_sample_wb_err(filp->f_mapping);
	filp->f_sb_err = file_sample_sb_err(filp);
	filp->private_data = dev_dax;
	inode->i_flags = S_DAX;

	return 0;
}

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

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

static const struct file_operations dax_fops = {
	.llseek = noop_llseek,
	.owner = THIS_MODULE,
	.open = dax_open,
	.release = dax_release,
	.get_unmapped_area = dax_get_unmapped_area,
	.mmap = dax_mmap,
	.mmap_supported_flags = MAP_SYNC,
};

static void dev_dax_cdev_del(void *cdev)
{
	cdev_del(cdev);
}

static void dev_dax_kill(void *dev_dax)
{
	kill_dev_dax(dev_dax);
}

int dev_dax_probe(struct device *dev)
{
	struct dev_dax *dev_dax = to_dev_dax(dev);
	struct dax_device *dax_dev = dev_dax->dax_dev;
	struct resource *res = &dev_dax->region->res;
	struct inode *inode;
	struct cdev *cdev;
	void *addr;
	int rc;

	/* 1:1 map region resource range to device-dax instance range */
	if (!devm_request_mem_region(dev, res->start, resource_size(res),
				dev_name(dev))) {
		dev_warn(dev, "could not reserve region %pR\n", res);
		return -EBUSY;
	}

	dev_dax->pgmap.type = MEMORY_DEVICE_GENERIC;
	addr = devm_memremap_pages(dev, &dev_dax->pgmap);
	if (IS_ERR(addr))
		return PTR_ERR(addr);

	inode = dax_inode(dax_dev);
	cdev = inode->i_cdev;
	cdev_init(cdev, &dax_fops);
	if (dev->class) {
		/* for the CONFIG_DEV_DAX_PMEM_COMPAT case */
		cdev->owner = dev->parent->driver->owner;
	} else
		cdev->owner = dev->driver->owner;
	cdev_set_parent(cdev, &dev->kobj);
	rc = cdev_add(cdev, dev->devt, 1);
	if (rc)
		return rc;

	rc = devm_add_action_or_reset(dev, dev_dax_cdev_del, cdev);
	if (rc)
		return rc;

	run_dax(dax_dev);
	return devm_add_action_or_reset(dev, dev_dax_kill, dev_dax);
}
EXPORT_SYMBOL_GPL(dev_dax_probe);

static int dev_dax_remove(struct device *dev)
{
	/* all probe actions are unwound by devm */
	return 0;
}

static struct dax_device_driver device_dax_driver = {
	.drv = {
		.probe = dev_dax_probe,
		.remove = dev_dax_remove,
	},
	.match_always = 1,
};

static int __init dax_init(void)
{
	return dax_driver_register(&device_dax_driver);
}

static void __exit dax_exit(void)
{
	dax_driver_unregister(&device_dax_driver);
}

MODULE_AUTHOR("Intel Corporation");
MODULE_LICENSE("GPL v2");
module_init(dax_init);
module_exit(dax_exit);
MODULE_ALIAS_DAX_DEVICE(0);
Commit	Line	Data
51cf784c DW	1	// SPDX-License-Identifier: GPL-2.0
51cf784c DW	2	/* Copyright(c) 2016-2018 Intel Corporation. All rights reserved. */
89ec9f2c	3	#include <linux/memremap.h>
ab68f262 DW	4	#include <linux/pagemap.h>
	5	#include <linux/module.h>
	6	#include <linux/device.h>
	7	#include <linux/pfn_t.h>
ba09c01d	8	#include <linux/cdev.h>
ab68f262 DW	9	#include <linux/slab.h>
	10	#include <linux/dax.h>
	11	#include <linux/fs.h>
	12	#include <linux/mm.h>
ef842302	13	#include <linux/mman.h>
efebc711	14	#include "dax-private.h"
51cf784c	15	#include "bus.h"
ab68f262	16
5f0694b3	17	static int check_vma(struct dev_dax dev_dax, struct vm_area_struct vma,
dee41079 DW	18	const char *func)
dee41079 DW	19	{
5f0694b3 DW	20	struct dax_region *dax_region = dev_dax->region;
5f0694b3 DW	21	struct device *dev = &dev_dax->dev;
dee41079 DW	22	unsigned long mask;
dee41079 DW	23
7b6be844	24	if (!dax_alive(dev_dax->dax_dev))
dee41079 DW	25	return -ENXIO;
dee41079 DW	26
4cb19355	27	/* prevent private mappings from being established */
325896ff	28	if ((vma->vm_flags & VM_MAYSHARE) != VM_MAYSHARE) {
5a14e91d JM	29	dev_info_ratelimited(dev,
5a14e91d JM	30	"%s: %s: fail, attempted private mapping\n",
dee41079 DW	31	current->comm, func);
	32	return -EINVAL;
	33	}
	34
	35	mask = dax_region->align - 1;
	36	if (vma->vm_start & mask \|\| vma->vm_end & mask) {
5a14e91d JM	37	dev_info_ratelimited(dev,
5a14e91d JM	38	"%s: %s: fail, unaligned vma (%#lx - %#lx, %#lx)\n",
dee41079 DW	39	current->comm, func, vma->vm_start, vma->vm_end,
	40	mask);
	41	return -EINVAL;
	42	}
	43
dee41079	44	if (!vma_is_dax(vma)) {
5a14e91d JM	45	dev_info_ratelimited(dev,
5a14e91d JM	46	"%s: %s: fail, vma is not DAX capable\n",
dee41079 DW	47	current->comm, func);
	48	return -EINVAL;
	49	}
	50
	51	return 0;
	52	}
	53
efebc711	54	/* see "strong" declaration in tools/testing/nvdimm/dax-dev.c */
73616367	55	__weak phys_addr_t dax_pgoff_to_phys(struct dev_dax *dev_dax, pgoff_t pgoff,
dee41079 DW	56	unsigned long size)
dee41079 DW	57	{
753a0850 DW	58	struct resource *res = &dev_dax->region->res;
753a0850 DW	59	phys_addr_t phys;
dee41079	60
753a0850 DW	61	phys = pgoff * PAGE_SIZE + res->start;
753a0850 DW	62	if (phys >= res->start && phys <= res->end) {
dee41079 DW	63	if (phys + size - 1 <= res->end)
	64	return phys;
	65	}
	66
	67	return -1;
	68	}
	69
226ab561	70	static vm_fault_t __dev_dax_pte_fault(struct dev_dax *dev_dax,
2232c638	71	struct vm_fault vmf, pfn_t pfn)
dee41079	72	{
5f0694b3	73	struct device *dev = &dev_dax->dev;
dee41079	74	struct dax_region *dax_region;
dee41079	75	phys_addr_t phys;
0134ed4f	76	unsigned int fault_size = PAGE_SIZE;
dee41079	77
5f0694b3	78	if (check_vma(dev_dax, vmf->vma, __func__))
dee41079 DW	79	return VM_FAULT_SIGBUS;
dee41079 DW	80
5f0694b3	81	dax_region = dev_dax->region;
dee41079	82	if (dax_region->align > PAGE_SIZE) {
6daaca52 DW	83	dev_dbg(dev, "alignment (%#x) > fault size (%#x)\n",
6daaca52 DW	84	dax_region->align, fault_size);
dee41079 DW	85	return VM_FAULT_SIGBUS;
	86	}
	87
0134ed4f DJ	88	if (fault_size != dax_region->align)
	89	return VM_FAULT_SIGBUS;
	90
73616367	91	phys = dax_pgoff_to_phys(dev_dax, vmf->pgoff, PAGE_SIZE);
dee41079	92	if (phys == -1) {
6daaca52	93	dev_dbg(dev, "pgoff_to_phys(%#lx) failed\n", vmf->pgoff);
dee41079 DW	94	return VM_FAULT_SIGBUS;
	95	}
	96
ec826909	97	*pfn = phys_to_pfn_t(phys, PFN_DEV\|PFN_MAP);
dee41079	98
2232c638	99	return vmf_insert_mixed(vmf->vma, vmf->address, *pfn);
dee41079 DW	100	}
dee41079 DW	101
226ab561	102	static vm_fault_t __dev_dax_pmd_fault(struct dev_dax *dev_dax,
2232c638	103	struct vm_fault vmf, pfn_t pfn)
dee41079	104	{
d8a849e1	105	unsigned long pmd_addr = vmf->address & PMD_MASK;
5f0694b3	106	struct device *dev = &dev_dax->dev;
dee41079 DW	107	struct dax_region *dax_region;
	108	phys_addr_t phys;
	109	pgoff_t pgoff;
0134ed4f	110	unsigned int fault_size = PMD_SIZE;
dee41079	111
5f0694b3	112	if (check_vma(dev_dax, vmf->vma, __func__))
dee41079 DW	113	return VM_FAULT_SIGBUS;
dee41079 DW	114
5f0694b3	115	dax_region = dev_dax->region;
dee41079	116	if (dax_region->align > PMD_SIZE) {
6daaca52 DW	117	dev_dbg(dev, "alignment (%#x) > fault size (%#x)\n",
6daaca52 DW	118	dax_region->align, fault_size);
dee41079 DW	119	return VM_FAULT_SIGBUS;
	120	}
	121
0134ed4f DJ	122	if (fault_size < dax_region->align)
	123	return VM_FAULT_SIGBUS;
	124	else if (fault_size > dax_region->align)
	125	return VM_FAULT_FALLBACK;
	126
	127	/* if we are outside of the VMA */
	128	if (pmd_addr < vmf->vma->vm_start \|\|
	129	(pmd_addr + PMD_SIZE) > vmf->vma->vm_end)
	130	return VM_FAULT_SIGBUS;
	131
f4200391	132	pgoff = linear_page_index(vmf->vma, pmd_addr);
73616367	133	phys = dax_pgoff_to_phys(dev_dax, pgoff, PMD_SIZE);
dee41079	134	if (phys == -1) {
6daaca52	135	dev_dbg(dev, "pgoff_to_phys(%#lx) failed\n", pgoff);
dee41079 DW	136	return VM_FAULT_SIGBUS;
	137	}
	138
ec826909	139	*pfn = phys_to_pfn_t(phys, PFN_DEV\|PFN_MAP);
dee41079	140
fce86ff5	141	return vmf_insert_pfn_pmd(vmf, *pfn, vmf->flags & FAULT_FLAG_WRITE);
dee41079 DW	142	}
dee41079 DW	143
9557feee	144	#ifdef CONFIG_HAVE_ARCH_TRANSPARENT_HUGEPAGE_PUD
226ab561	145	static vm_fault_t __dev_dax_pud_fault(struct dev_dax *dev_dax,
2232c638	146	struct vm_fault vmf, pfn_t pfn)
9557feee DJ	147	{
9557feee DJ	148	unsigned long pud_addr = vmf->address & PUD_MASK;
5f0694b3	149	struct device *dev = &dev_dax->dev;
9557feee DJ	150	struct dax_region *dax_region;
	151	phys_addr_t phys;
	152	pgoff_t pgoff;
70b085b0 DJ	153	unsigned int fault_size = PUD_SIZE;
70b085b0 DJ	154
9557feee	155
5f0694b3	156	if (check_vma(dev_dax, vmf->vma, __func__))
9557feee DJ	157	return VM_FAULT_SIGBUS;
9557feee DJ	158
5f0694b3	159	dax_region = dev_dax->region;
9557feee	160	if (dax_region->align > PUD_SIZE) {
6daaca52 DW	161	dev_dbg(dev, "alignment (%#x) > fault size (%#x)\n",
6daaca52 DW	162	dax_region->align, fault_size);
9557feee DJ	163	return VM_FAULT_SIGBUS;
	164	}
	165
70b085b0 DJ	166	if (fault_size < dax_region->align)
	167	return VM_FAULT_SIGBUS;
	168	else if (fault_size > dax_region->align)
	169	return VM_FAULT_FALLBACK;
	170
	171	/* if we are outside of the VMA */
	172	if (pud_addr < vmf->vma->vm_start \|\|
	173	(pud_addr + PUD_SIZE) > vmf->vma->vm_end)
	174	return VM_FAULT_SIGBUS;
	175
9557feee	176	pgoff = linear_page_index(vmf->vma, pud_addr);
73616367	177	phys = dax_pgoff_to_phys(dev_dax, pgoff, PUD_SIZE);
9557feee	178	if (phys == -1) {
6daaca52	179	dev_dbg(dev, "pgoff_to_phys(%#lx) failed\n", pgoff);
9557feee DJ	180	return VM_FAULT_SIGBUS;
	181	}
	182
ec826909	183	*pfn = phys_to_pfn_t(phys, PFN_DEV\|PFN_MAP);
9557feee	184
fce86ff5	185	return vmf_insert_pfn_pud(vmf, *pfn, vmf->flags & FAULT_FLAG_WRITE);
9557feee DJ	186	}
9557feee DJ	187	#else
226ab561	188	static vm_fault_t __dev_dax_pud_fault(struct dev_dax *dev_dax,
2232c638	189	struct vm_fault vmf, pfn_t pfn)
9557feee DJ	190	{
	191	return VM_FAULT_FALLBACK;
	192	}
	193	#endif /* !CONFIG_HAVE_ARCH_TRANSPARENT_HUGEPAGE_PUD */
	194
226ab561	195	static vm_fault_t dev_dax_huge_fault(struct vm_fault *vmf,
c791ace1	196	enum page_entry_size pe_size)
dee41079	197	{
f4200391	198	struct file *filp = vmf->vma->vm_file;
2232c638	199	unsigned long fault_size;
36bdac1e SJ	200	vm_fault_t rc = VM_FAULT_SIGBUS;
36bdac1e SJ	201	int id;
2232c638	202	pfn_t pfn;
5f0694b3	203	struct dev_dax *dev_dax = filp->private_data;
dee41079	204
6daaca52 DW	205	dev_dbg(&dev_dax->dev, "%s: %s (%#lx - %#lx) size = %d\n", current->comm,
6daaca52 DW	206	(vmf->flags & FAULT_FLAG_WRITE) ? "write" : "read",
76202620	207	vmf->vma->vm_start, vmf->vma->vm_end, pe_size);
dee41079	208
7b6be844	209	id = dax_read_lock();
c791ace1 DJ	210	switch (pe_size) {
c791ace1 DJ	211	case PE_SIZE_PTE:
2232c638 DW	212	fault_size = PAGE_SIZE;
2232c638 DW	213	rc = __dev_dax_pte_fault(dev_dax, vmf, &pfn);
a2d58167	214	break;
c791ace1	215	case PE_SIZE_PMD:
2232c638 DW	216	fault_size = PMD_SIZE;
2232c638 DW	217	rc = __dev_dax_pmd_fault(dev_dax, vmf, &pfn);
9557feee	218	break;
c791ace1	219	case PE_SIZE_PUD:
2232c638 DW	220	fault_size = PUD_SIZE;
2232c638 DW	221	rc = __dev_dax_pud_fault(dev_dax, vmf, &pfn);
a2d58167 DJ	222	break;
a2d58167 DJ	223	default:
54eafcc9	224	rc = VM_FAULT_SIGBUS;
a2d58167	225	}
2232c638 DW	226
	227	if (rc == VM_FAULT_NOPAGE) {
	228	unsigned long i;
35de2995	229	pgoff_t pgoff;
2232c638 DW	230
	231	/*
	232	* In the device-dax case the only possibility for a
	233	* VM_FAULT_NOPAGE result is when device-dax capacity is
	234	* mapped. No need to consider the zero page, or racing
	235	* conflicting mappings.
	236	*/
35de2995 DW	237	pgoff = linear_page_index(vmf->vma, vmf->address
35de2995 DW	238	& ~(fault_size - 1));
2232c638 DW	239	for (i = 0; i < fault_size / PAGE_SIZE; i++) {
	240	struct page *page;
	241
	242	page = pfn_to_page(pfn_t_to_pfn(pfn) + i);
	243	if (page->mapping)
	244	continue;
	245	page->mapping = filp->f_mapping;
35de2995	246	page->index = pgoff + i;
2232c638 DW	247	}
2232c638 DW	248	}
7b6be844	249	dax_read_unlock(id);
dee41079 DW	250
	251	return rc;
	252	}
	253
226ab561	254	static vm_fault_t dev_dax_fault(struct vm_fault *vmf)
c791ace1	255	{
5f0694b3	256	return dev_dax_huge_fault(vmf, PE_SIZE_PTE);
c791ace1 DJ	257	}
c791ace1 DJ	258
9702cffd DW	259	static int dev_dax_split(struct vm_area_struct *vma, unsigned long addr)
	260	{
	261	struct file *filp = vma->vm_file;
	262	struct dev_dax *dev_dax = filp->private_data;
	263	struct dax_region *dax_region = dev_dax->region;
	264
	265	if (!IS_ALIGNED(addr, dax_region->align))
	266	return -EINVAL;
	267	return 0;
	268	}
	269
c1d53b92 DW	270	static unsigned long dev_dax_pagesize(struct vm_area_struct *vma)
	271	{
	272	struct file *filp = vma->vm_file;
	273	struct dev_dax *dev_dax = filp->private_data;
	274	struct dax_region *dax_region = dev_dax->region;
	275
	276	return dax_region->align;
	277	}
	278
5f0694b3 DW	279	static const struct vm_operations_struct dax_vm_ops = {
	280	.fault = dev_dax_fault,
	281	.huge_fault = dev_dax_huge_fault,
9702cffd	282	.split = dev_dax_split,
c1d53b92	283	.pagesize = dev_dax_pagesize,
dee41079 DW	284	};
dee41079 DW	285
af69f51e	286	static int dax_mmap(struct file filp, struct vm_area_struct vma)
dee41079	287	{
5f0694b3	288	struct dev_dax *dev_dax = filp->private_data;
7b6be844	289	int rc, id;
dee41079	290
6daaca52	291	dev_dbg(&dev_dax->dev, "trace\n");
dee41079	292
7b6be844 DW	293	/*
	294	* We lock to check dax_dev liveness and will re-check at
	295	* fault time.
	296	*/
	297	id = dax_read_lock();
5f0694b3	298	rc = check_vma(dev_dax, vma, __func__);
7b6be844	299	dax_read_unlock(id);
dee41079 DW	300	if (rc)
	301	return rc;
	302
5f0694b3	303	vma->vm_ops = &dax_vm_ops;
e1fb4a08	304	vma->vm_flags \|= VM_HUGEPAGE;
dee41079	305	return 0;
043a9255 DW	306	}
	307
	308	/* return an unmapped area aligned to the dax region specified alignment */
af69f51e	309	static unsigned long dax_get_unmapped_area(struct file *filp,
043a9255 DW	310	unsigned long addr, unsigned long len, unsigned long pgoff,
	311	unsigned long flags)
	312	{
	313	unsigned long off, off_end, off_align, len_align, addr_align, align;
5f0694b3	314	struct dev_dax *dev_dax = filp ? filp->private_data : NULL;
043a9255 DW	315	struct dax_region *dax_region;
043a9255 DW	316
5f0694b3	317	if (!dev_dax \|\| addr)
043a9255 DW	318	goto out;
043a9255 DW	319
5f0694b3	320	dax_region = dev_dax->region;
043a9255 DW	321	align = dax_region->align;
	322	off = pgoff << PAGE_SHIFT;
	323	off_end = off + len;
	324	off_align = round_up(off, align);
	325
	326	if ((off_end <= off_align) \|\| ((off_end - off_align) < align))
	327	goto out;
	328
	329	len_align = len + align;
	330	if ((off + len_align) < off)
	331	goto out;
dee41079	332
043a9255 DW	333	addr_align = current->mm->get_unmapped_area(filp, addr, len_align,
	334	pgoff, flags);
	335	if (!IS_ERR_VALUE(addr_align)) {
	336	addr_align += (off - addr_align) & (align - 1);
	337	return addr_align;
	338	}
	339	out:
	340	return current->mm->get_unmapped_area(filp, addr, len, pgoff, flags);
	341	}
	342
41c9b1be DJ	343	static const struct address_space_operations dev_dax_aops = {
	344	.set_page_dirty = noop_set_page_dirty,
	345	.invalidatepage = noop_invalidatepage,
	346	};
	347
af69f51e	348	static int dax_open(struct inode inode, struct file filp)
043a9255	349	{
7b6be844 DW	350	struct dax_device *dax_dev = inode_dax(inode);
	351	struct inode *__dax_inode = dax_inode(dax_dev);
	352	struct dev_dax *dev_dax = dax_get_private(dax_dev);
043a9255	353
6daaca52	354	dev_dbg(&dev_dax->dev, "trace\n");
7b6be844 DW	355	inode->i_mapping = __dax_inode->i_mapping;
7b6be844 DW	356	inode->i_mapping->host = __dax_inode;
41c9b1be	357	inode->i_mapping->a_ops = &dev_dax_aops;
3bc52c45	358	filp->f_mapping = inode->i_mapping;
5660e13d	359	filp->f_wb_err = filemap_sample_wb_err(filp->f_mapping);
735e4ae5	360	filp->f_sb_err = file_sample_sb_err(filp);
5f0694b3	361	filp->private_data = dev_dax;
ebd84d72	362	inode->i_flags = S_DAX;
043a9255	363
043a9255 DW	364	return 0;
043a9255 DW	365	}
dee41079	366
af69f51e	367	static int dax_release(struct inode inode, struct file filp)
043a9255	368	{
5f0694b3	369	struct dev_dax *dev_dax = filp->private_data;
043a9255	370
6daaca52	371	dev_dbg(&dev_dax->dev, "trace\n");
043a9255	372	return 0;
dee41079 DW	373	}
dee41079 DW	374
ab68f262 DW	375	static const struct file_operations dax_fops = {
	376	.llseek = noop_llseek,
	377	.owner = THIS_MODULE,
af69f51e DW	378	.open = dax_open,
	379	.release = dax_release,
	380	.get_unmapped_area = dax_get_unmapped_area,
	381	.mmap = dax_mmap,
ef842302	382	.mmap_supported_flags = MAP_SYNC,
ab68f262 DW	383	};
ab68f262 DW	384
9567da0b	385	static void dev_dax_cdev_del(void *cdev)
043a9255	386	{
9567da0b DW	387	cdev_del(cdev);
9567da0b DW	388	}
043a9255	389
9567da0b DW	390	static void dev_dax_kill(void *dev_dax)
	391	{
	392	kill_dev_dax(dev_dax);
ebd84d72 DW	393	}
ebd84d72 DW	394
730926c3	395	int dev_dax_probe(struct device *dev)
043a9255	396	{
9567da0b DW	397	struct dev_dax *dev_dax = to_dev_dax(dev);
9567da0b DW	398	struct dax_device *dax_dev = dev_dax->dax_dev;
89ec9f2c	399	struct resource *res = &dev_dax->region->res;
7b6be844	400	struct inode *inode;
ba09c01d	401	struct cdev *cdev;
89ec9f2c	402	void *addr;
753a0850	403	int rc;
43fe51e1	404
89ec9f2c DW	405	/* 1:1 map region resource range to device-dax instance range */
	406	if (!devm_request_mem_region(dev, res->start, resource_size(res),
	407	dev_name(dev))) {
	408	dev_warn(dev, "could not reserve region %pR\n", res);
	409	return -EBUSY;
	410	}
	411
4533d3ae	412	dev_dax->pgmap.type = MEMORY_DEVICE_GENERIC;
89ec9f2c	413	addr = devm_memremap_pages(dev, &dev_dax->pgmap);
50f44ee7	414	if (IS_ERR(addr))
89ec9f2c	415	return PTR_ERR(addr);
89ec9f2c	416
7b6be844 DW	417	inode = dax_inode(dax_dev);
7b6be844 DW	418	cdev = inode->i_cdev;
ba09c01d	419	cdev_init(cdev, &dax_fops);
730926c3 DW	420	if (dev->class) {
	421	/* for the CONFIG_DEV_DAX_PMEM_COMPAT case */
	422	cdev->owner = dev->parent->driver->owner;
	423	} else
	424	cdev->owner = dev->driver->owner;
9567da0b DW	425	cdev_set_parent(cdev, &dev->kobj);
9567da0b DW	426	rc = cdev_add(cdev, dev->devt, 1);
d76911ee	427	if (rc)
9567da0b	428	return rc;
d76911ee	429
9567da0b DW	430	rc = devm_add_action_or_reset(dev, dev_dax_cdev_del, cdev);
	431	if (rc)
	432	return rc;
043a9255	433
9567da0b DW	434	run_dax(dax_dev);
	435	return devm_add_action_or_reset(dev, dev_dax_kill, dev_dax);
	436	}
730926c3	437	EXPORT_SYMBOL_GPL(dev_dax_probe);
043a9255	438
9567da0b DW	439	static int dev_dax_remove(struct device *dev)
	440	{
	441	/* all probe actions are unwound by devm */
	442	return 0;
043a9255	443	}
9567da0b	444
d200781e DW	445	static struct dax_device_driver device_dax_driver = {
	446	.drv = {
	447	.probe = dev_dax_probe,
	448	.remove = dev_dax_remove,
	449	},
	450	.match_always = 1,
9567da0b	451	};
043a9255	452
ab68f262 DW	453	static int __init dax_init(void)
ab68f262 DW	454	{
9567da0b	455	return dax_driver_register(&device_dax_driver);
ab68f262 DW	456	}
	457
	458	static void __exit dax_exit(void)
	459	{
d200781e	460	dax_driver_unregister(&device_dax_driver);
ab68f262 DW	461	}
	462
	463	MODULE_AUTHOR("Intel Corporation");
	464	MODULE_LICENSE("GPL v2");
9567da0b	465	module_init(dax_init);
ab68f262	466	module_exit(dax_exit);
9567da0b	467	MODULE_ALIAS_DAX_DEVICE(0);