Jetpack/kernel/nvidia/drivers/char/hmm_dmirror.c

/*
 * Copyright 2013 Red Hat Inc.
 * Copyright (c) 2017-2018, NVIDIA CORPORATION.  All rights reserved.
 *
 * 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 2 of
 * the License, or (at your option) 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.
 *
 * Authors: Jérôme Glisse <jglisse@redhat.com>
 */
/*
 * This is a dummy driver to exercice the HMM (heterogeneous memory management)
 * mirror API of the kernel. Userspace program register with the dummy device
 * to mirror their own address space and can use the device to read/write to
 * any valid virtual address.
 *
 * In some way it can also serve as an example driver for people wanting to use
 * HMM inside there own device driver.
 */
#include <linux/init.h>
#include <linux/fs.h>
#include <linux/mm.h>
#include <linux/module.h>
#include <linux/kernel.h>
#include <linux/cdev.h>
#include <linux/device.h>
#include <linux/mutex.h>
#include <linux/rwsem.h>
#include <linux/slab.h>
#include <linux/highmem.h>
#include <linux/delay.h>
#include <linux/pagemap.h>
#include <linux/hmm.h>
#include <linux/vmalloc.h>
#include <linux/swap.h>
#include <linux/swapops.h>
#include <linux/platform_device.h>
#include <linux/version.h>
#if LINUX_VERSION_CODE >= KERNEL_VERSION(4, 14, 0)
#include <linux/sched/mm.h>
#else
#include <linux/sched.h>
#endif

#include <uapi/linux/hmm_dmirror.h>

struct dmirror_device;

struct dummy_bounce {
	void			*ptr;
	unsigned long		size;
	unsigned long		addr;
	unsigned long		cpages;
};
#define FLAG_HMM_PFN_VALID (1 << 0)
#define FLAG_HMM_PFN_WRITE (1 << 1)
#define VALUE_HMM_PFN_NONE (1 << 4)
#define HPFN_SHIFT			7

#define DPT_SHIFT PAGE_SHIFT
#define DPT_VALID (1 << 0)
#define DPT_WRITE (1 << 1)
#define DPT_DPAGE (1 << 2)

struct dmirror_pt {
	unsigned long		pgd[PTRS_PER_PGD];
	struct rw_semaphore	lock;
};

struct dmirror {
	struct dmirror_device	*mdevice;
	struct file		*filp;
	struct hmm_mirror	mirror;
	struct mm_struct	*mm;
	struct dmirror_pt	pt;
};

struct dmirror_device {
	dev_t			dev;
	struct cdev		cdevice;
	struct class *cl;
	struct hmm_devmem	*devmem;
	struct platform_device	*pdevice;
	struct hmm_device	*hmm_device;
	struct page		*frees;
	spinlock_t		lock;

	unsigned long		calloc;
	unsigned long		cfree;
};

static inline unsigned long dmirror_pt_pgd(unsigned long addr)
{
	return (addr >> PGDIR_SHIFT) & (PTRS_PER_PGD - 1);
}

static inline unsigned long dmirror_pt_pud(unsigned long addr)
{
	return (addr >> PUD_SHIFT) & (PTRS_PER_PUD - 1);
}

static inline unsigned long dmirror_pt_pmd(unsigned long addr)
{
	return (addr >> PMD_SHIFT) & (PTRS_PER_PMD - 1);
}

static inline unsigned long dmirror_pt_pte(unsigned long addr)
{
	return (addr >> PAGE_SHIFT) & (PTRS_PER_PTE - 1);
}

static inline struct page *dmirror_pt_page(unsigned long dpte)
{
	if (!(dpte & DPT_VALID))
		return NULL;
	return pfn_to_page(dpte >> DPT_SHIFT);
}

static inline unsigned long dmirror_pt_from_page(struct page *page)
{
	if (!page)
		return 0;
	return (page_to_pfn(page) << DPT_SHIFT) | DPT_VALID;
}

static inline unsigned long dmirror_pt_pud_end(unsigned long addr)
{
	return (addr & PGDIR_MASK) + ((long)PTRS_PER_PUD << PUD_SHIFT);
}

static inline unsigned long dmirror_pt_pmd_end(unsigned long addr)
{
	return (addr & PUD_MASK) + ((long)PTRS_PER_PMD << PMD_SHIFT);
}

static inline unsigned long dmirror_pt_pte_end(unsigned long addr)
{
	return (addr & PMD_MASK) + ((long)PTRS_PER_PTE << PAGE_SHIFT);
}

typedef int (*dmirror_walk_cb_t)(struct dmirror *dmirror,
				struct hmm_range *range,
				unsigned long *dpte,
				void *private);

static int dummy_pt_walk(struct dmirror *dmirror,
			 dmirror_walk_cb_t cb,
			 struct hmm_range *range,
			 void *private,
			 bool populate)
{
	unsigned long start = range->start;
	unsigned long *dpgd = &dmirror->pt.pgd[dmirror_pt_pgd(start)];
	unsigned long addr = start & PAGE_MASK;
	unsigned long end = range->end;

	BUG_ON(start == end);

	for (; addr != end; dpgd++) {
		unsigned long pud_end, *dpud;
		struct page *pud_page;

		pud_end = min(end, dmirror_pt_pud_end(addr));
		pud_page = dmirror_pt_page(*dpgd);
		if (!pud_page) {
			if (!populate) {
				addr = pud_end;
				continue;
			}
			pud_page = alloc_page(GFP_HIGHUSER | __GFP_ZERO);
			if (!pud_page) {
				return -ENOMEM;
			}
			*dpgd = dmirror_pt_from_page(pud_page);
		}
		dpud = kmap(pud_page);
		dpud = &dpud[dmirror_pt_pud(addr)];
		for (; addr != pud_end; dpud++) {
			unsigned long pmd_end, *dpmd;
			struct page *pmd_page;

			pmd_end = min(end, dmirror_pt_pmd_end(addr));
			pmd_page = dmirror_pt_page(*dpud);
			if (!pmd_page) {
				if (!populate) {
					addr = pmd_end;
					continue;
				}
				pmd_page = alloc_page(GFP_HIGHUSER | __GFP_ZERO);
				if (!pmd_page) {
					kunmap(pud_page);
					return -ENOMEM;
				}
				*dpud = dmirror_pt_from_page(pmd_page);
			}
			dpmd = kmap(pmd_page);
			dpmd = &dpmd[dmirror_pt_pmd(addr)];
			for (; addr != pmd_end; dpmd++) {
				unsigned long *dpte, pte_end;
				struct hmm_range pte_range;
				struct page *pte_page;
				int ret;

				memcpy(&pte_range, range,
						sizeof(struct hmm_range));
				pte_range.flags = range->flags;
				pte_range.values = range->values;
				pte_end = min(end, dmirror_pt_pte_end(addr));
				pte_range.start = addr;
				pte_range.end = pte_end;
				pte_page = dmirror_pt_page(*dpmd);
				if (!pte_page) {
					if (!populate) {
						addr = pte_end;
						continue;
					}
					pte_page = alloc_page(GFP_HIGHUSER | __GFP_ZERO);
					if (!pte_page) {
						kunmap(pmd_page);
						kunmap(pud_page);
						return -ENOMEM;
					}
					*dpmd = dmirror_pt_from_page(pte_page);
				}
				dpte = kmap(pte_page);
				dpte = &dpte[dmirror_pt_pte(addr)];
				ret = cb(dmirror, &pte_range, dpte, private);
				kunmap(pte_page);
				addr = pte_end;
				if (ret) {
					kunmap(pmd_page);
					kunmap(pud_page);
					return ret;
				}
			}
			kunmap(pmd_page);
			addr = pmd_end;
		}
		kunmap(pud_page);
		addr = pud_end;
	}

	return 0;
}

int dummy_bounce_init(struct dummy_bounce *bounce,
		      unsigned long size,
		      unsigned long addr)
{
	bounce->addr = addr;
	bounce->size = size;
	bounce->ptr = vmalloc(size);
	if (!bounce->ptr)
		return -ENOMEM;
	return 0;
}

int dummy_bounce_copy_from(struct dummy_bounce *bounce, unsigned long addr)
{
	unsigned long end = (addr & PAGE_MASK) + bounce->size;
	char __user *uptr = (void __user *)(addr & PAGE_MASK);
	void *ptr = bounce->ptr;

	for (; addr < end; addr += PAGE_SIZE, ptr += PAGE_SIZE, uptr += PAGE_SIZE) {
		int ret;

		ret = copy_from_user(ptr, uptr, PAGE_SIZE);
		if (ret)
			return ret;
	}

	return 0;
}

int dummy_bounce_copy_to(struct dummy_bounce *bounce, unsigned long addr)
{
	unsigned long end = (addr & PAGE_MASK) + bounce->size;
	char __user *uptr = (void __user *)(addr & PAGE_MASK);
	void *ptr = bounce->ptr;

	for (; addr < end; addr += PAGE_SIZE, ptr += PAGE_SIZE, uptr += PAGE_SIZE) {
		int ret;

		ret = copy_to_user(uptr, ptr, PAGE_SIZE);
		if (ret)
			return ret;
	}

	return 0;
}

void dummy_bounce_fini(struct dummy_bounce *bounce)
{
	vfree(bounce->ptr);
}

static int dummy_do_update(struct dmirror *dmirror,
			struct hmm_range *range,
			unsigned long *dpte,
			void *private)
{
	unsigned long addr = range->start;
	unsigned long end = range->end;

	for (; addr < end; addr += PAGE_SIZE, ++dpte) {
		/* Clear pte */
		*dpte = 0;
	}

	return 0;
}

static void dummy_update(struct hmm_mirror *mirror,
			 enum hmm_update_type update,
			 unsigned long start,
			 unsigned long end)
{
	struct dmirror *dmirror = container_of(mirror, struct dmirror, mirror);
	struct hmm_range range;

	range.start = start;
	range.end = end;
	down_write(&dmirror->pt.lock);
	dummy_pt_walk(dmirror, dummy_do_update, &range, NULL, false);
	up_write(&dmirror->pt.lock);
}

static const struct hmm_mirror_ops dmirror_ops = {
	.sync_cpu_device_pagetables	= &dummy_update,
};


static int dmirror_pt_init(struct dmirror *dmirror)
{
	init_rwsem(&dmirror->pt.lock);
	return 0;
}

/* dmirror_new() - allocate and initialize dummy mirror struct.
 *
 * @mdevice: The dummy device this mirror is associated with.
 * @filp: The active device file descriptor this mirror is associated with.
 */
static struct dmirror *dmirror_new(struct dmirror_device *mdevice,
				   struct file *filp)
{
	struct mm_struct *mm = get_task_mm(current);
	struct dmirror *dmirror;
	int r;

	if (!mm)
		return NULL;

	/* Mirror this process address space */
	dmirror = kzalloc(sizeof(*dmirror), GFP_KERNEL);
	if (dmirror == NULL)
		return NULL;
	dmirror->mdevice = mdevice;
	dmirror->filp = filp;
	if (dmirror_pt_init(dmirror)) {
		kfree(dmirror);
		return NULL;
	}

	dmirror->mm = mm;
	dmirror->mirror.ops = &dmirror_ops;
	down_write(&mm->mmap_sem);
	r = hmm_mirror_register(&dmirror->mirror, mm);
	up_write(&mm->mmap_sem);
	mmput(mm);

	if (r) {
		kfree(dmirror);
		return NULL;
	}

	return dmirror;
}

static void dmirror_del(struct dmirror *dmirror)
{
	hmm_mirror_unregister(&dmirror->mirror);
	kfree(dmirror);
}


/*
 * Below are the file operation for the dummy device file. Only ioctl matter.
 *
 * Note this is highly specific to the dummy device driver and should not be
 * construed as an example on how to design the API a real device driver would
 * expose to userspace.
 */
static ssize_t dummy_fops_read(struct file *filp,
			       char __user *buf,
			       size_t count,
			       loff_t *ppos)
{
	return -EINVAL;
}

static ssize_t dummy_fops_write(struct file *filp,
				const char __user *buf,
				size_t count,
				loff_t *ppos)
{
	return -EINVAL;
}

static int dummy_fops_mmap(struct file *filp, struct vm_area_struct *vma)
{
	/* Forbid mmap of the dummy device file. */
	return -EINVAL;
}

static int dummy_fops_open(struct inode *inode, struct file *filp)
{
	struct cdev *cdev = inode->i_cdev;
	struct dmirror_device *mdevice;
	struct dmirror *dmirror;

	/* No exclusive opens. */
	if (filp->f_flags & O_EXCL)
		return -EINVAL;

	mdevice = container_of(cdev, struct dmirror_device, cdevice);
	dmirror = dmirror_new(mdevice, filp);
	filp->private_data = dmirror;

	return dmirror ? 0 : -ENOMEM;
}

static int dummy_fops_release(struct inode *inode, struct file *filp)
{
	struct dmirror_device *mdevice;
	struct dmirror *dmirror;

	if (!filp->private_data)
		return 0;

	dmirror = filp->private_data;
	mdevice = dmirror->mdevice;
	printk(KERN_INFO "DEVICE PAGE %ld %ld (%ld)\n", mdevice->calloc, mdevice->cfree, mdevice->calloc - mdevice->cfree);

	dmirror_del(dmirror);
	filp->private_data = NULL;

	return 0;
}

struct dummy_fault {
	uint64_t		*pfns;
	unsigned long		start;
	unsigned long		missing;
	bool			write;
	bool			invalid;
};

static int dummy_do_fault(struct dmirror *dmirror,
			  struct hmm_range *range,
			  unsigned long *dpte,
			  void *private)
{
	unsigned long addr = range->start;
	struct dummy_fault *dfault = private;
	unsigned long idx = (addr - dfault->start) >> PAGE_SHIFT;
	unsigned long end = range->end;
	uint64_t *pfns = dfault->pfns;

	for (; addr < end; addr += PAGE_SIZE, ++dpte, ++idx) {
		struct page *page;
		/*
		 * Special pfn are device memory ie page inserted inside the
		 * CPU page table with either vm_insert_pfn or vm_insert_page
		 * in both case we assume that device can not access this
		 * memory safely.
		 *
		 * The HMM_PFN_ERROR is if it is accessing invalid memory
		 * either because of memory error (hardware detected memory
		 * corruption) or more likely because of truncate on mmap
		 * file.
		 */
		if ((pfns[idx] & (range->values[HMM_PFN_SPECIAL] |
					range->values[HMM_PFN_ERROR]))) {
			dfault->invalid = true;
			continue;
		}

		if (!(pfns[idx] & range->flags[HMM_PFN_VALID])) {
			dfault->missing++;
			continue;
		}

		page = hmm_pfn_to_page(range, pfns[idx]);

		*dpte = dmirror_pt_from_page(page);

		if (pfns[idx] & HMM_PFN_WRITE) {
			*dpte |= DPT_WRITE;
		} else if (dfault->write) {
			dfault->missing++;
		}
	}

	return 0;
}

static int dummy_fault(struct dmirror *dmirror,
		       unsigned long start,
		       unsigned long end,
		       bool write)
{
	struct mm_struct *mm = dmirror->mm;
	unsigned long addr = start;
	uint64_t pfns[64];
	uint64_t flags[64];
	uint64_t values[64];

	memset(pfns, 0, sizeof(pfns));
	memset(flags, 0, sizeof(flags));
	memset(values, 0, sizeof(values));

	flags[HMM_PFN_VALID] = FLAG_HMM_PFN_VALID;
	flags[HMM_PFN_WRITE] = FLAG_HMM_PFN_WRITE;

	values[HMM_PFN_NONE] = VALUE_HMM_PFN_NONE;

	do {
		struct vm_area_struct *vma;
		struct dummy_fault dfault;
		struct hmm_range range;
		unsigned long next;
		int ret;

		down_read(&mm->mmap_sem);
		next = min(addr + (64 << PAGE_SHIFT), end);
		vma = find_vma_intersection(mm, addr, end);
		if (!vma) {
			up_read(&mm->mmap_sem);
			return -EFAULT;
		}
		if (!(vma->vm_flags & VM_READ)) {
			up_read(&mm->mmap_sem);
			return -EFAULT;
		}
		if (write && !(vma->vm_flags & VM_WRITE)) {
			up_read(&mm->mmap_sem);
			return -EFAULT;
		}
		addr = max(vma->vm_start, addr);
		next = min(min(addr + (64 << PAGE_SHIFT), end), vma->vm_end);
		range.vma = vma;
		range.start = addr;
		range.end = next;
		range.pfns = pfns;
		range.flags = flags;
		range.values = values;
		range.pfn_shift = HPFN_SHIFT;

		ret = hmm_vma_fault(&range, false);

		switch (ret) {
		case 0:
			break;
		case -EAGAIN:
			continue;
		default:
			up_read(&mm->mmap_sem);
			return ret;
		}
		down_read(&dmirror->pt.lock);
		if (!hmm_vma_range_done(&range)) {
			up_read(&dmirror->pt.lock);
			up_read(&mm->mmap_sem);
			continue;
		}

		dfault.invalid = false;
		dfault.write = write;
		dfault.missing = 0;
		dfault.start = addr;
		dfault.pfns = pfns;
		ret = dummy_pt_walk(dmirror, dummy_do_fault,
				    &range, &dfault, true);
		up_read(&dmirror->pt.lock);
		up_read(&mm->mmap_sem);
		if (ret)
			return ret;
		if (dfault.invalid)
			return -EFAULT;

		if (!dfault.missing) {
			addr = next;
		} else {
			return -EFAULT;
		}
	} while (addr != end);

	return 0;
}

static bool dummy_device_is_mine(struct dmirror_device *mdevice,
				 struct page *page)
{
	if (!is_zone_device_page(page))
		return false;
	return page->pgmap->data == mdevice->devmem;
}

static int dummy_do_read(struct dmirror *dmirror,
			struct hmm_range *range,
			unsigned long *dpte,
			void *private)
{
	struct dmirror_device *mdevice = dmirror->mdevice;
	struct dummy_bounce *bounce = private;
	unsigned long addr = range->start;
	unsigned long end = range->end;
	void *ptr;

	ptr = bounce->ptr + ((addr - bounce->addr) & PAGE_MASK);

	for (; addr < end; addr += PAGE_SIZE, ++dpte) {
		struct page *page;
		void *tmp;

		page = dmirror_pt_page(*dpte);
		if (!page) {
			return -ENOENT;
		}
		if (is_zone_device_page(page)) {
			if (!dummy_device_is_mine(mdevice, page))
				return -ENOENT;
			page = (void *)hmm_devmem_page_get_drvdata(page);
		}

		tmp = kmap(page);
		memcpy(ptr, tmp, PAGE_SIZE);
		kunmap(page);

		ptr += PAGE_SIZE;
		bounce->cpages++;
	}

	return 0;
}

static int dummy_read(struct dmirror *dmirror,
		      struct hmm_dmirror_read *dread)
{
	struct dummy_bounce bounce;
	struct hmm_range range;
	unsigned long start, end;
	unsigned long size;
	int ret;

	if ((dread->ptr & (~PAGE_MASK)) || (dread->addr & (~PAGE_MASK)))
		return -EINVAL;
	if (dread->addr >= (dread->addr + (dread->npages << PAGE_SHIFT)))
		return -EINVAL;

	start = dread->addr & PAGE_MASK;
	size = (dread->npages << PAGE_SHIFT);
	end = start + (dread->npages << PAGE_SHIFT);

	ret = dummy_bounce_init(&bounce, size, start);
	if (ret)
		return ret;

again:
	dread->dpages = 0;
	bounce.cpages = 0;
	range.start = start;
	range.end = end;
	down_read(&dmirror->pt.lock);
	ret = dummy_pt_walk(dmirror, dummy_do_read,
			    &range, &bounce, true);
	up_read(&dmirror->pt.lock);

	if (ret == -ENOENT) {
		ret = dummy_fault(dmirror, start, end, false);
		if (ret) {
			dummy_bounce_fini(&bounce);
			return ret;
		}
		goto again;
	}

	ret = dummy_bounce_copy_to(&bounce, dread->ptr);
	dread->cpages = bounce.cpages;
	dummy_bounce_fini(&bounce);
	return ret;
}

static int dummy_do_write(struct dmirror *dmirror,
			struct hmm_range *range,
			unsigned long *dpte,
			void *private)
{
	struct dmirror_device *mdevice = dmirror->mdevice;
	struct dummy_bounce *bounce = private;
	unsigned long addr = range->start;
	unsigned long end = range->end;
	void *ptr;

	ptr = bounce->ptr + ((addr - bounce->addr) & PAGE_MASK);

	for (; addr < end; addr += PAGE_SIZE, ++dpte) {
		struct page *page;
		void *tmp;

		page = dmirror_pt_page(*dpte);
		if (!page || !(*dpte & DPT_WRITE))
			return -ENOENT;
		if (is_zone_device_page(page)) {
			if (!dummy_device_is_mine(mdevice, page))
				return -ENOENT;
			page = (void *)hmm_devmem_page_get_drvdata(page);
		}

		tmp = kmap(page);
		memcpy(tmp, ptr, PAGE_SIZE);
		kunmap(page);

		ptr += PAGE_SIZE;
		bounce->cpages++;
	}

	return 0;
}

static int dummy_write(struct dmirror *dmirror,
		       struct hmm_dmirror_write *dwrite)
{
	struct dummy_bounce bounce;
	struct hmm_range range;
	unsigned long start, end;
	unsigned long size;
	int ret;

	if ((dwrite->ptr & (~PAGE_MASK)) || (dwrite->addr & (~PAGE_MASK)))
		return -EINVAL;
	if (dwrite->addr >= (dwrite->addr + (dwrite->npages << PAGE_SHIFT)))
		return -EINVAL;

	start = (unsigned long)dwrite->addr & PAGE_MASK;
	size = (unsigned long)(dwrite->npages << PAGE_SHIFT);
	end = (unsigned long)(start + (dwrite->npages << PAGE_SHIFT));

	ret = dummy_bounce_init(&bounce, size, dwrite->addr & PAGE_MASK);
	if (ret)
		return ret;
	ret = dummy_bounce_copy_from(&bounce, dwrite->ptr);
	if (ret)
		return ret;

again:
	bounce.cpages = 0;
	dwrite->dpages = 0;
	range.start = start;
	range.end = end;
	down_read(&dmirror->pt.lock);
	ret = dummy_pt_walk(dmirror, dummy_do_write,
			    &range, &bounce, true);
	up_read(&dmirror->pt.lock);

	if (ret == -ENOENT) {
		ret = dummy_fault(dmirror, start, end, true);
		if (ret) {
			dummy_bounce_fini(&bounce);
			return ret;
		}
		goto again;
	}

	dwrite->cpages = bounce.cpages;
	dummy_bounce_fini(&bounce);
	return 0;
}

static struct page *dummy_device_alloc_page(struct dmirror_device *mdevice)
{
	struct page *dpage = NULL, *rpage;

	/*
	 * This is a fake device so we alloc real system memory to fake
	 * our device memory
	 */
	rpage = alloc_page(GFP_HIGHUSER | __GFP_ZERO);
	if (!rpage)
		return NULL;

	spin_lock(&mdevice->lock);
	if (mdevice->frees) {
		dpage = mdevice->frees;
		mdevice->frees = dpage->s_mem;
	} else {
		spin_unlock(&mdevice->lock);
		__free_page(rpage);
		return NULL;
	}

	if (!trylock_page(dpage)) {
		dpage->s_mem = mdevice->frees;
		mdevice->frees = dpage;
		spin_unlock(&mdevice->lock);
		__free_page(rpage);
		return NULL;
	}
	mdevice->calloc++;
	spin_unlock(&mdevice->lock);

	hmm_devmem_page_set_drvdata(dpage, (unsigned long)rpage);
	get_page(dpage);
	return dpage;
}

struct dummy_migrate {
	struct dmirror_device		*mdevice;
	struct hmm_dmirror_migrate	*dmigrate;
};

static void dummy_migrate_alloc_and_copy(struct vm_area_struct *vma,
					 const unsigned long *src_pfns,
					 unsigned long *dst_pfns,
					 unsigned long start,
					 unsigned long end,
					 void *private)
{
	struct dummy_migrate *dmigrate = private;
	struct dmirror_device *mdevice;
	unsigned long addr;

	if (!dmigrate)
		return;

	mdevice = dmigrate->mdevice;

	for (addr = start; addr < end; addr += PAGE_SIZE, src_pfns++, dst_pfns++) {
		struct page *spage = migrate_pfn_to_page(*src_pfns);
		struct page *dpage, *rpage;

		*dst_pfns = 0;

		if (!spage && !(*src_pfns & MIGRATE_PFN_MIGRATE))
			continue;

		if (spage && !(*src_pfns & MIGRATE_PFN_MIGRATE))
			continue;

		if (spage && (*src_pfns & MIGRATE_PFN_DEVICE)) {
			if (!dummy_device_is_mine(mdevice, spage)) {
				continue;
			}
			spage = (void *)hmm_devmem_page_get_drvdata(spage);
		}

		dpage = dummy_device_alloc_page(mdevice);
		if (!dpage) {
			*dst_pfns = 0;
			continue;
		}

		rpage = (void *)hmm_devmem_page_get_drvdata(dpage);

		if (spage)
			copy_highpage(rpage, spage);
		*dst_pfns = migrate_pfn(page_to_pfn(dpage)) |
			    MIGRATE_PFN_DEVICE |
			    MIGRATE_PFN_LOCKED;
	}
}

static void dummy_migrate_finalize_and_map(struct vm_area_struct *vma,
					   const unsigned long *src_pfns,
					   const unsigned long *dst_pfns,
					   unsigned long start,
					   unsigned long end,
					   void *private)
{
	struct dummy_migrate *dmigrate = private;
	unsigned long addr;

	if (!dmigrate || !dmigrate->dmigrate)
		return;

	for (addr = start; addr < end; addr+= PAGE_SIZE, src_pfns++, dst_pfns++) {
		struct page *page = migrate_pfn_to_page(*dst_pfns);

		if (!page)
			continue;

		if (!(*src_pfns & MIGRATE_PFN_MIGRATE))
			continue;
		if (!dummy_device_is_mine(dmigrate->mdevice, page))
			continue;
		dmigrate->dmigrate->npages++;
	}
}

static const struct migrate_vma_ops dmirror_migrate_ops = {
	.alloc_and_copy		= dummy_migrate_alloc_and_copy,
	.finalize_and_map	= dummy_migrate_finalize_and_map,
};

static int dummy_migrate(struct dmirror *dmirror,
			 struct hmm_dmirror_migrate *dmigrate)
{
	unsigned long addr = dmigrate->addr, end;
	struct mm_struct *mm = dmirror->mm;
	struct vm_area_struct *vma;
	struct dummy_migrate tmp;
	int ret;

	tmp.mdevice = dmirror->mdevice;
	tmp.dmigrate = dmigrate;

	down_read(&mm->mmap_sem);
	end = addr + (dmigrate->npages << PAGE_SHIFT);
	vma = find_vma_intersection(mm, addr, end);
	if (!vma || vma->vm_start > addr || vma->vm_end < end) {
		ret = -EINVAL;
		goto out;
	}

	for (dmigrate->npages = 0; addr < end;) {
		unsigned long src_pfns[64];
		unsigned long dst_pfns[64];
		unsigned long next;

		next = min(end, addr + (64 << PAGE_SHIFT));

		ret = migrate_vma(&dmirror_migrate_ops, vma, addr,
				  next, src_pfns, dst_pfns, &tmp);
		if (ret)
			goto out;

		addr = next;
	}

out:
	up_read(&mm->mmap_sem);
	return ret;
}

static long dummy_fops_unlocked_ioctl(struct file *filp,
				      unsigned int command,
				      unsigned long arg)
{
	void __user *uarg = (void __user *)arg;
	struct hmm_dmirror_migrate dmigrate;
	struct hmm_dmirror_write dwrite;
	struct hmm_dmirror_read dread;
	struct dmirror *dmirror;
	int ret;

	dmirror = filp->private_data;
	if (!dmirror)
		return -EINVAL;

	switch (command) {
	case HMM_DMIRROR_READ:
		ret = copy_from_user(&dread, uarg, sizeof(dread));
		if (ret)
			return ret;

		ret = dummy_read(dmirror, &dread);
		if (ret)
			return ret;

		return copy_to_user(uarg, &dread, sizeof(dread));
	case HMM_DMIRROR_WRITE:
		ret = copy_from_user(&dwrite, uarg, sizeof(dwrite));
		if (ret)
			return ret;

		ret = dummy_write(dmirror, &dwrite);
		if (ret)
			return ret;

		return copy_to_user(uarg, &dwrite, sizeof(dwrite));
	case HMM_DMIRROR_MIGRATE:
		ret = copy_from_user(&dmigrate, uarg, sizeof(dmigrate));
		if (ret)
			return ret;

		ret = dummy_migrate(dmirror, &dmigrate);
		if (ret)
			return ret;

		return copy_to_user(uarg, &dmigrate, sizeof(dmigrate));
	default:
		ret = -EINVAL;
		break;
	}

	return ret;
}

static const struct file_operations dmirror_fops = {
	.read		= dummy_fops_read,
	.write		= dummy_fops_write,
	.mmap		= dummy_fops_mmap,
	.open		= dummy_fops_open,
	.release	= dummy_fops_release,
	.unlocked_ioctl = dummy_fops_unlocked_ioctl,
	.llseek		= default_llseek,
	.owner		= THIS_MODULE,
};

static void dummy_devmem_free(struct hmm_devmem *devmem,
			      struct page *page)
{
	struct dmirror_device *mdevice;
	struct page *rpage;

	rpage = (struct page *)hmm_devmem_page_get_drvdata(page);
	mdevice = dev_get_drvdata(devmem->device);
	hmm_devmem_page_set_drvdata(page, 0);
	__free_page(rpage);

	spin_lock(&mdevice->lock);
	mdevice->cfree++;
	page->s_mem = mdevice->frees;
	mdevice->frees = page;
	spin_unlock(&mdevice->lock);
}

struct dummy_devmem_fault {
	struct dmirror_device	*mdevice;
};

static void dummy_devmem_fault_alloc_and_copy(struct vm_area_struct *vma,
					      const unsigned long *src_pfns,
					      unsigned long *dst_pfns,
					      unsigned long start,
					      unsigned long end,
					      void *private)
{
	struct dummy_devmem_fault *fault = private;
	unsigned long addr;

	for (addr = start; addr < end; addr += PAGE_SIZE, src_pfns++, dst_pfns++) {
		struct page *dpage, *spage;

		*dst_pfns = MIGRATE_PFN_ERROR;

		spage = migrate_pfn_to_page(*src_pfns);
		if (!spage || !(*src_pfns & MIGRATE_PFN_MIGRATE))
			continue;
		if (!dummy_device_is_mine(fault->mdevice, spage))
			continue;
		spage = (void *)hmm_devmem_page_get_drvdata(spage);

		dpage = hmm_vma_alloc_locked_page(vma, addr);
		if (!dpage) {
			*dst_pfns = MIGRATE_PFN_ERROR;
			continue;
		}

		copy_highpage(dpage, spage);
		*dst_pfns = migrate_pfn(page_to_pfn(dpage)) |
			    MIGRATE_PFN_LOCKED;
	}
}

void dummy_devmem_fault_finalize_and_map(struct vm_area_struct *vma,
					 const unsigned long *src_pfns,
					 const unsigned long *dst_pfns,
					 unsigned long start,
					 unsigned long end,
					 void *private)
{
}

static const struct migrate_vma_ops dummy_devmem_migrate = {
	.alloc_and_copy		= dummy_devmem_fault_alloc_and_copy,
	.finalize_and_map	= dummy_devmem_fault_finalize_and_map,
};

/*
 * hmm_devmem_fault_range() - migrate back a virtual range of memory
 *
 * @devmem: hmm_devmem struct use to track and manage the ZONE_DEVICE memory
 * @vma: virtual memory area containing the range to be migrated
 * @ops: migration callback for allocating destination memory and copying
 * @src: array of unsigned long containing source pfns
 * @dst: array of unsigned long containing destination pfns
 * @start: start address of the range to migrate (inclusive)
 * @addr: fault address (must be inside the range)
 * @end: end address of the range to migrate (exclusive)
 * @private: pointer passed back to each of the callback
 * Returns: 0 on success, VM_FAULT_SIGBUS on error
 *
 * This is a wrapper around migrate_vma() which checks the migration status
 * for a given fault address and returns the corresponding page fault handler
 * status. That will be 0 on success, or VM_FAULT_SIGBUS if migration failed
 * for the faulting address.
 *
 * This is a helper intendend to be used by the ZONE_DEVICE fault handler.
 */
int hmm_devmem_fault_range(struct hmm_devmem *devmem,
			   struct vm_area_struct *vma,
			   const struct migrate_vma_ops *ops,
			   unsigned long *src,
			   unsigned long *dst,
			   unsigned long start,
			   unsigned long addr,
			   unsigned long end,
			   void *private)
{
	if (migrate_vma(ops, vma, start, end, src, dst, private))
		return VM_FAULT_SIGBUS;

	if (dst[(addr - start) >> PAGE_SHIFT] & MIGRATE_PFN_ERROR)
		return VM_FAULT_SIGBUS;

	return 0;
}
EXPORT_SYMBOL(hmm_devmem_fault_range);

static int dummy_devmem_fault(struct hmm_devmem *devmem,
			      struct vm_area_struct *vma,
			      unsigned long addr,
			      const struct page *page,
			      unsigned flags,
			      pmd_t *pmdp)
{
	unsigned long src_pfns, dst_pfns = 0;
	struct dummy_devmem_fault fault;
	unsigned long start, end;

	fault.mdevice = dev_get_drvdata(devmem->device);

	/* FIXME demonstrate how we can adjust migrate range */
	start = addr;
	end = addr + PAGE_SIZE;
	return hmm_devmem_fault_range(devmem, vma, &dummy_devmem_migrate,
				      &src_pfns, &dst_pfns, start,
				      addr, end, &fault);
}

static const struct hmm_devmem_ops dmirror_devmem_ops = {
	.free		= dummy_devmem_free,
	.fault		= dummy_devmem_fault,
};

static int dmirror_probe(struct platform_device *pdev)
{
	struct dmirror_device *mdevice = platform_get_drvdata(pdev);
	struct device *dev;
	unsigned long pfn;
	int ret;

	mdevice->hmm_device = hmm_device_new(mdevice);
	if (IS_ERR(mdevice->hmm_device))
		return PTR_ERR(mdevice->hmm_device);

	mdevice->devmem = hmm_devmem_add(&dmirror_devmem_ops,
					 &mdevice->hmm_device->device,
					 64 << 20);
	if (IS_ERR(mdevice->devmem)) {
		hmm_device_put(mdevice->hmm_device);
		return PTR_ERR(mdevice->devmem);
	}

	ret = alloc_chrdev_region(&mdevice->dev, 0, 1, "HMM_DMIRROR");
	if (ret < 0) {
		hmm_devmem_remove(mdevice->devmem);
		hmm_device_put(mdevice->hmm_device);
		return ret;
	}

	mdevice->cl = class_create(THIS_MODULE, "chardrv");
	if (IS_ERR_OR_NULL(mdevice->cl)) {
		unregister_chrdev_region(mdevice->dev, 1);
		hmm_devmem_remove(mdevice->devmem);
		hmm_device_put(mdevice->hmm_device);
		return PTR_ERR(mdevice->cl);
	}

	dev = device_create(mdevice->cl, NULL, mdevice->dev, NULL,
			"hmm_dummy_device");
	if (IS_ERR_OR_NULL(dev)) {
		class_destroy(mdevice->cl);
		unregister_chrdev_region(mdevice->dev, 1);
		hmm_devmem_remove(mdevice->devmem);
		hmm_device_put(mdevice->hmm_device);
		return PTR_ERR(dev);
	}

	cdev_init(&mdevice->cdevice, &dmirror_fops);
	ret = cdev_add(&mdevice->cdevice, mdevice->dev, 1);
	if (ret) {
		device_destroy(mdevice->cl, mdevice->dev);
		class_destroy(mdevice->cl);
		unregister_chrdev_region(mdevice->dev, 1);
		hmm_devmem_remove(mdevice->devmem);
		hmm_device_put(mdevice->hmm_device);
		return ret;
	}

	/* Build list of free struct page */
	spin_lock_init(&mdevice->lock);

	spin_lock(&mdevice->lock);
	mdevice->frees = NULL;

	for (pfn = mdevice->devmem->pfn_first; pfn < mdevice->devmem->pfn_last; pfn++) {
		struct page *page = pfn_to_page(pfn);

		page->s_mem = mdevice->frees;
		mdevice->frees = page;
	}

	mdevice->calloc = 0;
	mdevice->cfree = 0;
	spin_unlock(&mdevice->lock);

	return 0;
}

static int dmirror_remove(struct platform_device *pdev)
{
	struct dmirror_device *mdevice = platform_get_drvdata(pdev);

	cdev_del(&mdevice->cdevice);
	device_destroy(mdevice->cl, mdevice->dev);
	class_destroy(mdevice->cl);
	unregister_chrdev_region(mdevice->dev, 1);
	hmm_devmem_remove(mdevice->devmem);
	hmm_device_put(mdevice->hmm_device);
	return 0;
}


static struct platform_device *dmirror_platform_device;

static struct platform_driver dmirror_device_driver = {
	.probe		= dmirror_probe,
	.remove		= dmirror_remove,
	.driver		= {
		.name	= "HMM_DMIRROR",
	},
};

static int __init hmm_dmirror_init(void)
{
	struct dmirror_device *mdevice;
	int ret;

	mdevice = kzalloc(sizeof(*mdevice), GFP_KERNEL);
	if (!mdevice)
		return -ENOMEM;

	dmirror_platform_device = platform_device_alloc("HMM_DMIRROR", -1);
	if (!dmirror_platform_device) {
		kfree(mdevice);
		return -ENOMEM;
	}
	platform_set_drvdata(dmirror_platform_device, mdevice);
	mdevice->pdevice = dmirror_platform_device;

	ret = platform_device_add(dmirror_platform_device);
	if (ret < 0) {
		platform_device_put(dmirror_platform_device);
		return ret;
	}

	ret = platform_driver_register(&dmirror_device_driver);
	if (ret < 0) {
		platform_device_unregister(dmirror_platform_device);
		return ret;
	}

	pr_debug("hmm_dmirror loaded THIS IS A DANGEROUS MODULE !!!\n");
	return 0;
}

static void __exit hmm_dmirror_exit(void)
{
	struct dmirror_device *mdevice;

	mdevice = platform_get_drvdata(dmirror_platform_device);
	platform_driver_unregister(&dmirror_device_driver);
	platform_device_unregister(dmirror_platform_device);
	kfree(mdevice);
}

module_init(hmm_dmirror_init);
module_exit(hmm_dmirror_exit);
MODULE_LICENSE("GPL");