[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 struct dev_dax *ref_to_dev_dax(struct percpu_ref *ref)
{
	return container_of(ref, struct dev_dax, ref);
}

static void dev_dax_percpu_release(struct percpu_ref *ref)
{
	struct dev_dax *dev_dax = ref_to_dev_dax(ref);

	dev_dbg(&dev_dax->dev, "%s\n", __func__);
	complete(&dev_dax->cmp);
}

static void dev_dax_percpu_exit(void *data)
{
	struct percpu_ref *ref = data;
	struct dev_dax *dev_dax = ref_to_dev_dax(ref);

	dev_dbg(&dev_dax->dev, "%s\n", __func__);
	wait_for_completion(&dev_dax->cmp);
	percpu_ref_exit(ref);
}

static void dev_dax_percpu_kill(struct percpu_ref *data)
{
	struct percpu_ref *ref = data;
	struct dev_dax *dev_dax = ref_to_dev_dax(ref);

	dev_dbg(&dev_dax->dev, "%s\n", __func__);
	percpu_ref_kill(ref);
}

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 ((dax_region->pfn_flags & (PFN_DEV|PFN_MAP)) == PFN_DEV
			&& (vma->vm_flags & VM_DONTCOPY) == 0) {
		dev_info_ratelimited(dev,
				"%s: %s: fail, dax range requires MADV_DONTFORK\n",
				current->comm, func);
		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, dax_region->pfn_flags);

	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;
	}

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

	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;
	}

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

	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->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;
	}

	init_completion(&dev_dax->cmp);
	rc = percpu_ref_init(&dev_dax->ref, dev_dax_percpu_release, 0,
			GFP_KERNEL);
	if (rc)
		return rc;

	rc = devm_add_action_or_reset(dev, dev_dax_percpu_exit, &dev_dax->ref);
	if (rc)
		return rc;

	dev_dax->pgmap.ref = &dev_dax->ref;
	dev_dax->pgmap.kill = dev_dax_percpu_kill;
	addr = devm_memremap_pages(dev, &dev_dax->pgmap);
	if (IS_ERR(addr)) {
		devm_remove_action(dev, dev_dax_percpu_exit, &dev_dax->ref);
		percpu_ref_exit(&dev_dax->ref);
		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
89ec9f2c DW	17	static struct dev_dax ref_to_dev_dax(struct percpu_ref ref)
	18	{
	19	return container_of(ref, struct dev_dax, ref);
	20	}
	21
	22	static void dev_dax_percpu_release(struct percpu_ref *ref)
	23	{
	24	struct dev_dax *dev_dax = ref_to_dev_dax(ref);
	25
	26	dev_dbg(&dev_dax->dev, "%s\n", __func__);
	27	complete(&dev_dax->cmp);
	28	}
	29
	30	static void dev_dax_percpu_exit(void *data)
	31	{
	32	struct percpu_ref *ref = data;
	33	struct dev_dax *dev_dax = ref_to_dev_dax(ref);
	34
	35	dev_dbg(&dev_dax->dev, "%s\n", __func__);
	36	wait_for_completion(&dev_dax->cmp);
	37	percpu_ref_exit(ref);
	38	}
	39
	40	static void dev_dax_percpu_kill(struct percpu_ref *data)
	41	{
	42	struct percpu_ref *ref = data;
	43	struct dev_dax *dev_dax = ref_to_dev_dax(ref);
	44
	45	dev_dbg(&dev_dax->dev, "%s\n", __func__);
	46	percpu_ref_kill(ref);
	47	}
	48
5f0694b3	49	static int check_vma(struct dev_dax dev_dax, struct vm_area_struct vma,
dee41079 DW	50	const char *func)
dee41079 DW	51	{
5f0694b3 DW	52	struct dax_region *dax_region = dev_dax->region;
5f0694b3 DW	53	struct device *dev = &dev_dax->dev;
dee41079 DW	54	unsigned long mask;
dee41079 DW	55
7b6be844	56	if (!dax_alive(dev_dax->dax_dev))
dee41079 DW	57	return -ENXIO;
dee41079 DW	58
4cb19355	59	/* prevent private mappings from being established */
325896ff	60	if ((vma->vm_flags & VM_MAYSHARE) != VM_MAYSHARE) {
5a14e91d JM	61	dev_info_ratelimited(dev,
5a14e91d JM	62	"%s: %s: fail, attempted private mapping\n",
dee41079 DW	63	current->comm, func);
	64	return -EINVAL;
	65	}
	66
	67	mask = dax_region->align - 1;
	68	if (vma->vm_start & mask \|\| vma->vm_end & mask) {
5a14e91d JM	69	dev_info_ratelimited(dev,
5a14e91d JM	70	"%s: %s: fail, unaligned vma (%#lx - %#lx, %#lx)\n",
dee41079 DW	71	current->comm, func, vma->vm_start, vma->vm_end,
	72	mask);
	73	return -EINVAL;
	74	}
	75
	76	if ((dax_region->pfn_flags & (PFN_DEV\|PFN_MAP)) == PFN_DEV
	77	&& (vma->vm_flags & VM_DONTCOPY) == 0) {
5a14e91d JM	78	dev_info_ratelimited(dev,
5a14e91d JM	79	"%s: %s: fail, dax range requires MADV_DONTFORK\n",
dee41079 DW	80	current->comm, func);
	81	return -EINVAL;
	82	}
	83
	84	if (!vma_is_dax(vma)) {
5a14e91d JM	85	dev_info_ratelimited(dev,
5a14e91d JM	86	"%s: %s: fail, vma is not DAX capable\n",
dee41079 DW	87	current->comm, func);
	88	return -EINVAL;
	89	}
	90
	91	return 0;
	92	}
	93
efebc711	94	/* see "strong" declaration in tools/testing/nvdimm/dax-dev.c */
73616367	95	__weak phys_addr_t dax_pgoff_to_phys(struct dev_dax *dev_dax, pgoff_t pgoff,
dee41079 DW	96	unsigned long size)
dee41079 DW	97	{
753a0850 DW	98	struct resource *res = &dev_dax->region->res;
753a0850 DW	99	phys_addr_t phys;
dee41079	100
753a0850 DW	101	phys = pgoff * PAGE_SIZE + res->start;
753a0850 DW	102	if (phys >= res->start && phys <= res->end) {
dee41079 DW	103	if (phys + size - 1 <= res->end)
	104	return phys;
	105	}
	106
	107	return -1;
	108	}
	109
226ab561	110	static vm_fault_t __dev_dax_pte_fault(struct dev_dax *dev_dax,
2232c638	111	struct vm_fault vmf, pfn_t pfn)
dee41079	112	{
5f0694b3	113	struct device *dev = &dev_dax->dev;
dee41079	114	struct dax_region *dax_region;
dee41079	115	phys_addr_t phys;
0134ed4f	116	unsigned int fault_size = PAGE_SIZE;
dee41079	117
5f0694b3	118	if (check_vma(dev_dax, vmf->vma, __func__))
dee41079 DW	119	return VM_FAULT_SIGBUS;
dee41079 DW	120
5f0694b3	121	dax_region = dev_dax->region;
dee41079	122	if (dax_region->align > PAGE_SIZE) {
6daaca52 DW	123	dev_dbg(dev, "alignment (%#x) > fault size (%#x)\n",
6daaca52 DW	124	dax_region->align, fault_size);
dee41079 DW	125	return VM_FAULT_SIGBUS;
	126	}
	127
0134ed4f DJ	128	if (fault_size != dax_region->align)
	129	return VM_FAULT_SIGBUS;
	130
73616367	131	phys = dax_pgoff_to_phys(dev_dax, vmf->pgoff, PAGE_SIZE);
dee41079	132	if (phys == -1) {
6daaca52	133	dev_dbg(dev, "pgoff_to_phys(%#lx) failed\n", vmf->pgoff);
dee41079 DW	134	return VM_FAULT_SIGBUS;
	135	}
	136
2232c638	137	*pfn = phys_to_pfn_t(phys, dax_region->pfn_flags);
dee41079	138
2232c638	139	return vmf_insert_mixed(vmf->vma, vmf->address, *pfn);
dee41079 DW	140	}
dee41079 DW	141
226ab561	142	static vm_fault_t __dev_dax_pmd_fault(struct dev_dax *dev_dax,
2232c638	143	struct vm_fault vmf, pfn_t pfn)
dee41079	144	{
d8a849e1	145	unsigned long pmd_addr = vmf->address & PMD_MASK;
5f0694b3	146	struct device *dev = &dev_dax->dev;
dee41079 DW	147	struct dax_region *dax_region;
	148	phys_addr_t phys;
	149	pgoff_t pgoff;
0134ed4f	150	unsigned int fault_size = PMD_SIZE;
dee41079	151
5f0694b3	152	if (check_vma(dev_dax, vmf->vma, __func__))
dee41079 DW	153	return VM_FAULT_SIGBUS;
dee41079 DW	154
5f0694b3	155	dax_region = dev_dax->region;
dee41079	156	if (dax_region->align > PMD_SIZE) {
6daaca52 DW	157	dev_dbg(dev, "alignment (%#x) > fault size (%#x)\n",
6daaca52 DW	158	dax_region->align, fault_size);
dee41079 DW	159	return VM_FAULT_SIGBUS;
	160	}
	161
	162	/* dax pmd mappings require pfn_t_devmap() */
	163	if ((dax_region->pfn_flags & (PFN_DEV\|PFN_MAP)) != (PFN_DEV\|PFN_MAP)) {
6daaca52	164	dev_dbg(dev, "region lacks devmap flags\n");
dee41079 DW	165	return VM_FAULT_SIGBUS;
	166	}
	167
0134ed4f DJ	168	if (fault_size < dax_region->align)
	169	return VM_FAULT_SIGBUS;
	170	else if (fault_size > dax_region->align)
	171	return VM_FAULT_FALLBACK;
	172
	173	/* if we are outside of the VMA */
	174	if (pmd_addr < vmf->vma->vm_start \|\|
	175	(pmd_addr + PMD_SIZE) > vmf->vma->vm_end)
	176	return VM_FAULT_SIGBUS;
	177
f4200391	178	pgoff = linear_page_index(vmf->vma, pmd_addr);
73616367	179	phys = dax_pgoff_to_phys(dev_dax, pgoff, PMD_SIZE);
dee41079	180	if (phys == -1) {
6daaca52	181	dev_dbg(dev, "pgoff_to_phys(%#lx) failed\n", pgoff);
dee41079 DW	182	return VM_FAULT_SIGBUS;
	183	}
	184
2232c638	185	*pfn = phys_to_pfn_t(phys, dax_region->pfn_flags);
dee41079	186
fce86ff5	187	return vmf_insert_pfn_pmd(vmf, *pfn, vmf->flags & FAULT_FLAG_WRITE);
dee41079 DW	188	}
dee41079 DW	189
9557feee	190	#ifdef CONFIG_HAVE_ARCH_TRANSPARENT_HUGEPAGE_PUD
226ab561	191	static vm_fault_t __dev_dax_pud_fault(struct dev_dax *dev_dax,
2232c638	192	struct vm_fault vmf, pfn_t pfn)
9557feee DJ	193	{
9557feee DJ	194	unsigned long pud_addr = vmf->address & PUD_MASK;
5f0694b3	195	struct device *dev = &dev_dax->dev;
9557feee DJ	196	struct dax_region *dax_region;
	197	phys_addr_t phys;
	198	pgoff_t pgoff;
70b085b0 DJ	199	unsigned int fault_size = PUD_SIZE;
70b085b0 DJ	200
9557feee	201
5f0694b3	202	if (check_vma(dev_dax, vmf->vma, __func__))
9557feee DJ	203	return VM_FAULT_SIGBUS;
9557feee DJ	204
5f0694b3	205	dax_region = dev_dax->region;
9557feee	206	if (dax_region->align > PUD_SIZE) {
6daaca52 DW	207	dev_dbg(dev, "alignment (%#x) > fault size (%#x)\n",
6daaca52 DW	208	dax_region->align, fault_size);
9557feee DJ	209	return VM_FAULT_SIGBUS;
	210	}
	211
	212	/* dax pud mappings require pfn_t_devmap() */
	213	if ((dax_region->pfn_flags & (PFN_DEV\|PFN_MAP)) != (PFN_DEV\|PFN_MAP)) {
6daaca52	214	dev_dbg(dev, "region lacks devmap flags\n");
9557feee DJ	215	return VM_FAULT_SIGBUS;
	216	}
	217
70b085b0 DJ	218	if (fault_size < dax_region->align)
	219	return VM_FAULT_SIGBUS;
	220	else if (fault_size > dax_region->align)
	221	return VM_FAULT_FALLBACK;
	222
	223	/* if we are outside of the VMA */
	224	if (pud_addr < vmf->vma->vm_start \|\|
	225	(pud_addr + PUD_SIZE) > vmf->vma->vm_end)
	226	return VM_FAULT_SIGBUS;
	227
9557feee	228	pgoff = linear_page_index(vmf->vma, pud_addr);
73616367	229	phys = dax_pgoff_to_phys(dev_dax, pgoff, PUD_SIZE);
9557feee	230	if (phys == -1) {
6daaca52	231	dev_dbg(dev, "pgoff_to_phys(%#lx) failed\n", pgoff);
9557feee DJ	232	return VM_FAULT_SIGBUS;
	233	}
	234
2232c638	235	*pfn = phys_to_pfn_t(phys, dax_region->pfn_flags);
9557feee	236
fce86ff5	237	return vmf_insert_pfn_pud(vmf, *pfn, vmf->flags & FAULT_FLAG_WRITE);
9557feee DJ	238	}
9557feee DJ	239	#else
226ab561	240	static vm_fault_t __dev_dax_pud_fault(struct dev_dax *dev_dax,
2232c638	241	struct vm_fault vmf, pfn_t pfn)
9557feee DJ	242	{
	243	return VM_FAULT_FALLBACK;
	244	}
	245	#endif /* !CONFIG_HAVE_ARCH_TRANSPARENT_HUGEPAGE_PUD */
	246
226ab561	247	static vm_fault_t dev_dax_huge_fault(struct vm_fault *vmf,
c791ace1	248	enum page_entry_size pe_size)
dee41079	249	{
f4200391	250	struct file *filp = vmf->vma->vm_file;
2232c638	251	unsigned long fault_size;
36bdac1e SJ	252	vm_fault_t rc = VM_FAULT_SIGBUS;
36bdac1e SJ	253	int id;
2232c638	254	pfn_t pfn;
5f0694b3	255	struct dev_dax *dev_dax = filp->private_data;
dee41079	256
6daaca52 DW	257	dev_dbg(&dev_dax->dev, "%s: %s (%#lx - %#lx) size = %d\n", current->comm,
6daaca52 DW	258	(vmf->flags & FAULT_FLAG_WRITE) ? "write" : "read",
76202620	259	vmf->vma->vm_start, vmf->vma->vm_end, pe_size);
dee41079	260
7b6be844	261	id = dax_read_lock();
c791ace1 DJ	262	switch (pe_size) {
c791ace1 DJ	263	case PE_SIZE_PTE:
2232c638 DW	264	fault_size = PAGE_SIZE;
2232c638 DW	265	rc = __dev_dax_pte_fault(dev_dax, vmf, &pfn);
a2d58167	266	break;
c791ace1	267	case PE_SIZE_PMD:
2232c638 DW	268	fault_size = PMD_SIZE;
2232c638 DW	269	rc = __dev_dax_pmd_fault(dev_dax, vmf, &pfn);
9557feee	270	break;
c791ace1	271	case PE_SIZE_PUD:
2232c638 DW	272	fault_size = PUD_SIZE;
2232c638 DW	273	rc = __dev_dax_pud_fault(dev_dax, vmf, &pfn);
a2d58167 DJ	274	break;
a2d58167 DJ	275	default:
54eafcc9	276	rc = VM_FAULT_SIGBUS;
a2d58167	277	}
2232c638 DW	278
	279	if (rc == VM_FAULT_NOPAGE) {
	280	unsigned long i;
35de2995	281	pgoff_t pgoff;
2232c638 DW	282
	283	/*
	284	* In the device-dax case the only possibility for a
	285	* VM_FAULT_NOPAGE result is when device-dax capacity is
	286	* mapped. No need to consider the zero page, or racing
	287	* conflicting mappings.
	288	*/
35de2995 DW	289	pgoff = linear_page_index(vmf->vma, vmf->address
35de2995 DW	290	& ~(fault_size - 1));
2232c638 DW	291	for (i = 0; i < fault_size / PAGE_SIZE; i++) {
	292	struct page *page;
	293
	294	page = pfn_to_page(pfn_t_to_pfn(pfn) + i);
	295	if (page->mapping)
	296	continue;
	297	page->mapping = filp->f_mapping;
35de2995	298	page->index = pgoff + i;
2232c638 DW	299	}
2232c638 DW	300	}
7b6be844	301	dax_read_unlock(id);
dee41079 DW	302
	303	return rc;
	304	}
	305
226ab561	306	static vm_fault_t dev_dax_fault(struct vm_fault *vmf)
c791ace1	307	{
5f0694b3	308	return dev_dax_huge_fault(vmf, PE_SIZE_PTE);
c791ace1 DJ	309	}
c791ace1 DJ	310
9702cffd DW	311	static int dev_dax_split(struct vm_area_struct *vma, unsigned long addr)
	312	{
	313	struct file *filp = vma->vm_file;
	314	struct dev_dax *dev_dax = filp->private_data;
	315	struct dax_region *dax_region = dev_dax->region;
	316
	317	if (!IS_ALIGNED(addr, dax_region->align))
	318	return -EINVAL;
	319	return 0;
	320	}
	321
c1d53b92 DW	322	static unsigned long dev_dax_pagesize(struct vm_area_struct *vma)
	323	{
	324	struct file *filp = vma->vm_file;
	325	struct dev_dax *dev_dax = filp->private_data;
	326	struct dax_region *dax_region = dev_dax->region;
	327
	328	return dax_region->align;
	329	}
	330
5f0694b3 DW	331	static const struct vm_operations_struct dax_vm_ops = {
	332	.fault = dev_dax_fault,
	333	.huge_fault = dev_dax_huge_fault,
9702cffd	334	.split = dev_dax_split,
c1d53b92	335	.pagesize = dev_dax_pagesize,
dee41079 DW	336	};
dee41079 DW	337
af69f51e	338	static int dax_mmap(struct file filp, struct vm_area_struct vma)
dee41079	339	{
5f0694b3	340	struct dev_dax *dev_dax = filp->private_data;
7b6be844	341	int rc, id;
dee41079	342
6daaca52	343	dev_dbg(&dev_dax->dev, "trace\n");
dee41079	344
7b6be844 DW	345	/*
	346	* We lock to check dax_dev liveness and will re-check at
	347	* fault time.
	348	*/
	349	id = dax_read_lock();
5f0694b3	350	rc = check_vma(dev_dax, vma, __func__);
7b6be844	351	dax_read_unlock(id);
dee41079 DW	352	if (rc)
	353	return rc;
	354
5f0694b3	355	vma->vm_ops = &dax_vm_ops;
e1fb4a08	356	vma->vm_flags \|= VM_HUGEPAGE;
dee41079	357	return 0;
043a9255 DW	358	}
	359
	360	/* return an unmapped area aligned to the dax region specified alignment */
af69f51e	361	static unsigned long dax_get_unmapped_area(struct file *filp,
043a9255 DW	362	unsigned long addr, unsigned long len, unsigned long pgoff,
	363	unsigned long flags)
	364	{
	365	unsigned long off, off_end, off_align, len_align, addr_align, align;
5f0694b3	366	struct dev_dax *dev_dax = filp ? filp->private_data : NULL;
043a9255 DW	367	struct dax_region *dax_region;
043a9255 DW	368
5f0694b3	369	if (!dev_dax \|\| addr)
043a9255 DW	370	goto out;
043a9255 DW	371
5f0694b3	372	dax_region = dev_dax->region;
043a9255 DW	373	align = dax_region->align;
	374	off = pgoff << PAGE_SHIFT;
	375	off_end = off + len;
	376	off_align = round_up(off, align);
	377
	378	if ((off_end <= off_align) \|\| ((off_end - off_align) < align))
	379	goto out;
	380
	381	len_align = len + align;
	382	if ((off + len_align) < off)
	383	goto out;
dee41079	384
043a9255 DW	385	addr_align = current->mm->get_unmapped_area(filp, addr, len_align,
	386	pgoff, flags);
	387	if (!IS_ERR_VALUE(addr_align)) {
	388	addr_align += (off - addr_align) & (align - 1);
	389	return addr_align;
	390	}
	391	out:
	392	return current->mm->get_unmapped_area(filp, addr, len, pgoff, flags);
	393	}
	394
41c9b1be DJ	395	static const struct address_space_operations dev_dax_aops = {
	396	.set_page_dirty = noop_set_page_dirty,
	397	.invalidatepage = noop_invalidatepage,
	398	};
	399
af69f51e	400	static int dax_open(struct inode inode, struct file filp)
043a9255	401	{
7b6be844 DW	402	struct dax_device *dax_dev = inode_dax(inode);
	403	struct inode *__dax_inode = dax_inode(dax_dev);
	404	struct dev_dax *dev_dax = dax_get_private(dax_dev);
043a9255	405
6daaca52	406	dev_dbg(&dev_dax->dev, "trace\n");
7b6be844 DW	407	inode->i_mapping = __dax_inode->i_mapping;
7b6be844 DW	408	inode->i_mapping->host = __dax_inode;
41c9b1be	409	inode->i_mapping->a_ops = &dev_dax_aops;
3bc52c45	410	filp->f_mapping = inode->i_mapping;
5660e13d	411	filp->f_wb_err = filemap_sample_wb_err(filp->f_mapping);
5f0694b3	412	filp->private_data = dev_dax;
ebd84d72	413	inode->i_flags = S_DAX;
043a9255	414
043a9255 DW	415	return 0;
043a9255 DW	416	}
dee41079	417
af69f51e	418	static int dax_release(struct inode inode, struct file filp)
043a9255	419	{
5f0694b3	420	struct dev_dax *dev_dax = filp->private_data;
043a9255	421
6daaca52	422	dev_dbg(&dev_dax->dev, "trace\n");
043a9255	423	return 0;
dee41079 DW	424	}
dee41079 DW	425
ab68f262 DW	426	static const struct file_operations dax_fops = {
	427	.llseek = noop_llseek,
	428	.owner = THIS_MODULE,
af69f51e DW	429	.open = dax_open,
	430	.release = dax_release,
	431	.get_unmapped_area = dax_get_unmapped_area,
	432	.mmap = dax_mmap,
ef842302	433	.mmap_supported_flags = MAP_SYNC,
ab68f262 DW	434	};
ab68f262 DW	435
9567da0b	436	static void dev_dax_cdev_del(void *cdev)
043a9255	437	{
9567da0b DW	438	cdev_del(cdev);
9567da0b DW	439	}
043a9255	440
9567da0b DW	441	static void dev_dax_kill(void *dev_dax)
	442	{
	443	kill_dev_dax(dev_dax);
ebd84d72 DW	444	}
ebd84d72 DW	445
730926c3	446	int dev_dax_probe(struct device *dev)
043a9255	447	{
9567da0b DW	448	struct dev_dax *dev_dax = to_dev_dax(dev);
9567da0b DW	449	struct dax_device *dax_dev = dev_dax->dax_dev;
89ec9f2c	450	struct resource *res = &dev_dax->region->res;
7b6be844	451	struct inode *inode;
ba09c01d	452	struct cdev *cdev;
89ec9f2c	453	void *addr;
753a0850	454	int rc;
43fe51e1	455
89ec9f2c DW	456	/* 1:1 map region resource range to device-dax instance range */
	457	if (!devm_request_mem_region(dev, res->start, resource_size(res),
	458	dev_name(dev))) {
	459	dev_warn(dev, "could not reserve region %pR\n", res);
	460	return -EBUSY;
	461	}
	462
	463	init_completion(&dev_dax->cmp);
	464	rc = percpu_ref_init(&dev_dax->ref, dev_dax_percpu_release, 0,
	465	GFP_KERNEL);
	466	if (rc)
	467	return rc;
	468
	469	rc = devm_add_action_or_reset(dev, dev_dax_percpu_exit, &dev_dax->ref);
	470	if (rc)
	471	return rc;
	472
	473	dev_dax->pgmap.ref = &dev_dax->ref;
	474	dev_dax->pgmap.kill = dev_dax_percpu_kill;
	475	addr = devm_memremap_pages(dev, &dev_dax->pgmap);
	476	if (IS_ERR(addr)) {
	477	devm_remove_action(dev, dev_dax_percpu_exit, &dev_dax->ref);
	478	percpu_ref_exit(&dev_dax->ref);
	479	return PTR_ERR(addr);
	480	}
	481
7b6be844 DW	482	inode = dax_inode(dax_dev);
7b6be844 DW	483	cdev = inode->i_cdev;
ba09c01d	484	cdev_init(cdev, &dax_fops);
730926c3 DW	485	if (dev->class) {
	486	/* for the CONFIG_DEV_DAX_PMEM_COMPAT case */
	487	cdev->owner = dev->parent->driver->owner;
	488	} else
	489	cdev->owner = dev->driver->owner;
9567da0b DW	490	cdev_set_parent(cdev, &dev->kobj);
9567da0b DW	491	rc = cdev_add(cdev, dev->devt, 1);
d76911ee	492	if (rc)
9567da0b	493	return rc;
d76911ee	494
9567da0b DW	495	rc = devm_add_action_or_reset(dev, dev_dax_cdev_del, cdev);
	496	if (rc)
	497	return rc;
043a9255	498
9567da0b DW	499	run_dax(dax_dev);
	500	return devm_add_action_or_reset(dev, dev_dax_kill, dev_dax);
	501	}
730926c3	502	EXPORT_SYMBOL_GPL(dev_dax_probe);
043a9255	503
9567da0b DW	504	static int dev_dax_remove(struct device *dev)
	505	{
	506	/* all probe actions are unwound by devm */
	507	return 0;
043a9255	508	}
9567da0b	509
d200781e DW	510	static struct dax_device_driver device_dax_driver = {
	511	.drv = {
	512	.probe = dev_dax_probe,
	513	.remove = dev_dax_remove,
	514	},
	515	.match_always = 1,
9567da0b	516	};
043a9255	517
ab68f262 DW	518	static int __init dax_init(void)
ab68f262 DW	519	{
9567da0b	520	return dax_driver_register(&device_dax_driver);
ab68f262 DW	521	}
	522
	523	static void __exit dax_exit(void)
	524	{
d200781e	525	dax_driver_unregister(&device_dax_driver);
ab68f262 DW	526	}
	527
	528	MODULE_AUTHOR("Intel Corporation");
	529	MODULE_LICENSE("GPL v2");
9567da0b	530	module_init(dax_init);
ab68f262	531	module_exit(dax_exit);
9567da0b	532	MODULE_ALIAS_DAX_DEVICE(0);