[linux.git] / drivers / block / brd.c

/*
 * Ram backed block device driver.
 *
 * Copyright (C) 2007 Nick Piggin
 * Copyright (C) 2007 Novell Inc.
 *
 * Parts derived from drivers/block/rd.c, and drivers/block/loop.c, copyright
 * of their respective owners.
 */

#include <linux/init.h>
#include <linux/module.h>
#include <linux/moduleparam.h>
#include <linux/major.h>
#include <linux/blkdev.h>
#include <linux/bio.h>
#include <linux/highmem.h>
#include <linux/mutex.h>
#include <linux/radix-tree.h>
#include <linux/buffer_head.h> /* invalidate_bh_lrus() */
#include <linux/slab.h>

#include <asm/uaccess.h>

#define SECTOR_SHIFT		9
#define PAGE_SECTORS_SHIFT	(PAGE_SHIFT - SECTOR_SHIFT)
#define PAGE_SECTORS		(1 << PAGE_SECTORS_SHIFT)

/*
 * Each block ramdisk device has a radix_tree brd_pages of pages that stores
 * the pages containing the block device's contents. A brd page's ->index is
 * its offset in PAGE_SIZE units. This is similar to, but in no way connected
 * with, the kernel's pagecache or buffer cache (which sit above our block
 * device).
 */
struct brd_device {
	int		brd_number;
	int		brd_refcnt;
	loff_t		brd_offset;
	loff_t		brd_sizelimit;
	unsigned	brd_blocksize;

	struct request_queue	*brd_queue;
	struct gendisk		*brd_disk;
	struct list_head	brd_list;

	/*
	 * Backing store of pages and lock to protect it. This is the contents
	 * of the block device.
	 */
	spinlock_t		brd_lock;
	struct radix_tree_root	brd_pages;
};

/*
 * Look up and return a brd's page for a given sector.
 */
static DEFINE_MUTEX(brd_mutex);
static struct page *brd_lookup_page(struct brd_device *brd, sector_t sector)
{
	pgoff_t idx;
	struct page *page;

	/*
	 * The page lifetime is protected by the fact that we have opened the
	 * device node -- brd pages will never be deleted under us, so we
	 * don't need any further locking or refcounting.
	 *
	 * This is strictly true for the radix-tree nodes as well (ie. we
	 * don't actually need the rcu_read_lock()), however that is not a
	 * documented feature of the radix-tree API so it is better to be
	 * safe here (we don't have total exclusion from radix tree updates
	 * here, only deletes).
	 */
	rcu_read_lock();
	idx = sector >> PAGE_SECTORS_SHIFT; /* sector to page index */
	page = radix_tree_lookup(&brd->brd_pages, idx);
	rcu_read_unlock();

	BUG_ON(page && page->index != idx);

	return page;
}

/*
 * Look up and return a brd's page for a given sector.
 * If one does not exist, allocate an empty page, and insert that. Then
 * return it.
 */
static struct page *brd_insert_page(struct brd_device *brd, sector_t sector)
{
	pgoff_t idx;
	struct page *page;
	gfp_t gfp_flags;

	page = brd_lookup_page(brd, sector);
	if (page)
		return page;

	/*
	 * Must use NOIO because we don't want to recurse back into the
	 * block or filesystem layers from page reclaim.
	 *
	 * Cannot support XIP and highmem, because our ->direct_access
	 * routine for XIP must return memory that is always addressable.
	 * If XIP was reworked to use pfns and kmap throughout, this
	 * restriction might be able to be lifted.
	 */
	gfp_flags = GFP_NOIO | __GFP_ZERO;
#ifndef CONFIG_BLK_DEV_XIP
	gfp_flags |= __GFP_HIGHMEM;
#endif
	page = alloc_page(gfp_flags);
	if (!page)
		return NULL;

	if (radix_tree_preload(GFP_NOIO)) {
		__free_page(page);
		return NULL;
	}

	spin_lock(&brd->brd_lock);
	idx = sector >> PAGE_SECTORS_SHIFT;
	if (radix_tree_insert(&brd->brd_pages, idx, page)) {
		__free_page(page);
		page = radix_tree_lookup(&brd->brd_pages, idx);
		BUG_ON(!page);
		BUG_ON(page->index != idx);
	} else
		page->index = idx;
	spin_unlock(&brd->brd_lock);

	radix_tree_preload_end();

	return page;
}

static void brd_free_page(struct brd_device *brd, sector_t sector)
{
	struct page *page;
	pgoff_t idx;

	spin_lock(&brd->brd_lock);
	idx = sector >> PAGE_SECTORS_SHIFT;
	page = radix_tree_delete(&brd->brd_pages, idx);
	spin_unlock(&brd->brd_lock);
	if (page)
		__free_page(page);
}

static void brd_zero_page(struct brd_device *brd, sector_t sector)
{
	struct page *page;

	page = brd_lookup_page(brd, sector);
	if (page)
		clear_highpage(page);
}

/*
 * Free all backing store pages and radix tree. This must only be called when
 * there are no other users of the device.
 */
#define FREE_BATCH 16
static void brd_free_pages(struct brd_device *brd)
{
	unsigned long pos = 0;
	struct page *pages[FREE_BATCH];
	int nr_pages;

	do {
		int i;

		nr_pages = radix_tree_gang_lookup(&brd->brd_pages,
				(void **)pages, pos, FREE_BATCH);

		for (i = 0; i < nr_pages; i++) {
			void *ret;

			BUG_ON(pages[i]->index < pos);
			pos = pages[i]->index;
			ret = radix_tree_delete(&brd->brd_pages, pos);
			BUG_ON(!ret || ret != pages[i]);
			__free_page(pages[i]);
		}

		pos++;

		/*
		 * This assumes radix_tree_gang_lookup always returns as
		 * many pages as possible. If the radix-tree code changes,
		 * so will this have to.
		 */
	} while (nr_pages == FREE_BATCH);
}

/*
 * copy_to_brd_setup must be called before copy_to_brd. It may sleep.
 */
static int copy_to_brd_setup(struct brd_device *brd, sector_t sector, size_t n)
{
	unsigned int offset = (sector & (PAGE_SECTORS-1)) << SECTOR_SHIFT;
	size_t copy;

	copy = min_t(size_t, n, PAGE_SIZE - offset);
	if (!brd_insert_page(brd, sector))
		return -ENOMEM;
	if (copy < n) {
		sector += copy >> SECTOR_SHIFT;
		if (!brd_insert_page(brd, sector))
			return -ENOMEM;
	}
	return 0;
}

static void discard_from_brd(struct brd_device *brd,
			sector_t sector, size_t n)
{
	while (n >= PAGE_SIZE) {
		/*
		 * Don't want to actually discard pages here because
		 * re-allocating the pages can result in writeback
		 * deadlocks under heavy load.
		 */
		if (0)
			brd_free_page(brd, sector);
		else
			brd_zero_page(brd, sector);
		sector += PAGE_SIZE >> SECTOR_SHIFT;
		n -= PAGE_SIZE;
	}
}

/*
 * Copy n bytes from src to the brd starting at sector. Does not sleep.
 */
static void copy_to_brd(struct brd_device *brd, const void *src,
			sector_t sector, size_t n)
{
	struct page *page;
	void *dst;
	unsigned int offset = (sector & (PAGE_SECTORS-1)) << SECTOR_SHIFT;
	size_t copy;

	copy = min_t(size_t, n, PAGE_SIZE - offset);
	page = brd_lookup_page(brd, sector);
	BUG_ON(!page);

	dst = kmap_atomic(page, KM_USER1);
	memcpy(dst + offset, src, copy);
	kunmap_atomic(dst, KM_USER1);

	if (copy < n) {
		src += copy;
		sector += copy >> SECTOR_SHIFT;
		copy = n - copy;
		page = brd_lookup_page(brd, sector);
		BUG_ON(!page);

		dst = kmap_atomic(page, KM_USER1);
		memcpy(dst, src, copy);
		kunmap_atomic(dst, KM_USER1);
	}
}

/*
 * Copy n bytes to dst from the brd starting at sector. Does not sleep.
 */
static void copy_from_brd(void *dst, struct brd_device *brd,
			sector_t sector, size_t n)
{
	struct page *page;
	void *src;
	unsigned int offset = (sector & (PAGE_SECTORS-1)) << SECTOR_SHIFT;
	size_t copy;

	copy = min_t(size_t, n, PAGE_SIZE - offset);
	page = brd_lookup_page(brd, sector);
	if (page) {
		src = kmap_atomic(page, KM_USER1);
		memcpy(dst, src + offset, copy);
		kunmap_atomic(src, KM_USER1);
	} else
		memset(dst, 0, copy);

	if (copy < n) {
		dst += copy;
		sector += copy >> SECTOR_SHIFT;
		copy = n - copy;
		page = brd_lookup_page(brd, sector);
		if (page) {
			src = kmap_atomic(page, KM_USER1);
			memcpy(dst, src, copy);
			kunmap_atomic(src, KM_USER1);
		} else
			memset(dst, 0, copy);
	}
}

/*
 * Process a single bvec of a bio.
 */
static int brd_do_bvec(struct brd_device *brd, struct page *page,
			unsigned int len, unsigned int off, int rw,
			sector_t sector)
{
	void *mem;
	int err = 0;

	if (rw != READ) {
		err = copy_to_brd_setup(brd, sector, len);
		if (err)
			goto out;
	}

	mem = kmap_atomic(page, KM_USER0);
	if (rw == READ) {
		copy_from_brd(mem + off, brd, sector, len);
		flush_dcache_page(page);
	} else {
		flush_dcache_page(page);
		copy_to_brd(brd, mem + off, sector, len);
	}
	kunmap_atomic(mem, KM_USER0);

out:
	return err;
}

static int brd_make_request(struct request_queue *q, struct bio *bio)
{
	struct block_device *bdev = bio->bi_bdev;
	struct brd_device *brd = bdev->bd_disk->private_data;
	int rw;
	struct bio_vec *bvec;
	sector_t sector;
	int i;
	int err = -EIO;

	sector = bio->bi_sector;
	if (sector + (bio->bi_size >> SECTOR_SHIFT) >
						get_capacity(bdev->bd_disk))
		goto out;

	if (unlikely(bio->bi_rw & REQ_DISCARD)) {
		err = 0;
		discard_from_brd(brd, sector, bio->bi_size);
		goto out;
	}

	rw = bio_rw(bio);
	if (rw == READA)
		rw = READ;

	bio_for_each_segment(bvec, bio, i) {
		unsigned int len = bvec->bv_len;
		err = brd_do_bvec(brd, bvec->bv_page, len,
					bvec->bv_offset, rw, sector);
		if (err)
			break;
		sector += len >> SECTOR_SHIFT;
	}

out:
	bio_endio(bio, err);

	return 0;
}

#ifdef CONFIG_BLK_DEV_XIP
static int brd_direct_access(struct block_device *bdev, sector_t sector,
			void **kaddr, unsigned long *pfn)
{
	struct brd_device *brd = bdev->bd_disk->private_data;
	struct page *page;

	if (!brd)
		return -ENODEV;
	if (sector & (PAGE_SECTORS-1))
		return -EINVAL;
	if (sector + PAGE_SECTORS > get_capacity(bdev->bd_disk))
		return -ERANGE;
	page = brd_insert_page(brd, sector);
	if (!page)
		return -ENOMEM;
	*kaddr = page_address(page);
	*pfn = page_to_pfn(page);

	return 0;
}
#endif

static int brd_ioctl(struct block_device *bdev, fmode_t mode,
			unsigned int cmd, unsigned long arg)
{
	int error;
	struct brd_device *brd = bdev->bd_disk->private_data;

	if (cmd != BLKFLSBUF)
		return -ENOTTY;

	/*
	 * ram device BLKFLSBUF has special semantics, we want to actually
	 * release and destroy the ramdisk data.
	 */
	mutex_lock(&brd_mutex);
	mutex_lock(&bdev->bd_mutex);
	error = -EBUSY;
	if (bdev->bd_openers <= 1) {
		/*
		 * Invalidate the cache first, so it isn't written
		 * back to the device.
		 *
		 * Another thread might instantiate more buffercache here,
		 * but there is not much we can do to close that race.
		 */
		invalidate_bh_lrus();
		truncate_inode_pages(bdev->bd_inode->i_mapping, 0);
		brd_free_pages(brd);
		error = 0;
	}
	mutex_unlock(&bdev->bd_mutex);
	mutex_unlock(&brd_mutex);

	return error;
}

static const struct block_device_operations brd_fops = {
	.owner =		THIS_MODULE,
	.ioctl =		brd_ioctl,
#ifdef CONFIG_BLK_DEV_XIP
	.direct_access =	brd_direct_access,
#endif
};

/*
 * And now the modules code and kernel interface.
 */
static int rd_nr;
int rd_size = CONFIG_BLK_DEV_RAM_SIZE;
static int max_part;
static int part_shift;
module_param(rd_nr, int, 0);
MODULE_PARM_DESC(rd_nr, "Maximum number of brd devices");
module_param(rd_size, int, 0);
MODULE_PARM_DESC(rd_size, "Size of each RAM disk in kbytes.");
module_param(max_part, int, 0);
MODULE_PARM_DESC(max_part, "Maximum number of partitions per RAM disk");
MODULE_LICENSE("GPL");
MODULE_ALIAS_BLOCKDEV_MAJOR(RAMDISK_MAJOR);
MODULE_ALIAS("rd");

#ifndef MODULE
/* Legacy boot options - nonmodular */
static int __init ramdisk_size(char *str)
{
	rd_size = simple_strtol(str, NULL, 0);
	return 1;
}
__setup("ramdisk_size=", ramdisk_size);
#endif

/*
 * The device scheme is derived from loop.c. Keep them in synch where possible
 * (should share code eventually).
 */
static LIST_HEAD(brd_devices);
static DEFINE_MUTEX(brd_devices_mutex);

static struct brd_device *brd_alloc(int i)
{
	struct brd_device *brd;
	struct gendisk *disk;

	brd = kzalloc(sizeof(*brd), GFP_KERNEL);
	if (!brd)
		goto out;
	brd->brd_number		= i;
	spin_lock_init(&brd->brd_lock);
	INIT_RADIX_TREE(&brd->brd_pages, GFP_ATOMIC);

	brd->brd_queue = blk_alloc_queue(GFP_KERNEL);
	if (!brd->brd_queue)
		goto out_free_dev;
	blk_queue_make_request(brd->brd_queue, brd_make_request);
	blk_queue_max_hw_sectors(brd->brd_queue, 1024);
	blk_queue_bounce_limit(brd->brd_queue, BLK_BOUNCE_ANY);

	brd->brd_queue->limits.discard_granularity = PAGE_SIZE;
	brd->brd_queue->limits.max_discard_sectors = UINT_MAX;
	brd->brd_queue->limits.discard_zeroes_data = 1;
	queue_flag_set_unlocked(QUEUE_FLAG_DISCARD, brd->brd_queue);

	disk = brd->brd_disk = alloc_disk(1 << part_shift);
	if (!disk)
		goto out_free_queue;
	disk->major		= RAMDISK_MAJOR;
	disk->first_minor	= i << part_shift;
	disk->fops		= &brd_fops;
	disk->private_data	= brd;
	disk->queue		= brd->brd_queue;
	disk->flags |= GENHD_FL_SUPPRESS_PARTITION_INFO;
	sprintf(disk->disk_name, "ram%d", i);
	set_capacity(disk, rd_size * 2);

	return brd;

out_free_queue:
	blk_cleanup_queue(brd->brd_queue);
out_free_dev:
	kfree(brd);
out:
	return NULL;
}

static void brd_free(struct brd_device *brd)
{
	put_disk(brd->brd_disk);
	blk_cleanup_queue(brd->brd_queue);
	brd_free_pages(brd);
	kfree(brd);
}

static struct brd_device *brd_init_one(int i)
{
	struct brd_device *brd;

	list_for_each_entry(brd, &brd_devices, brd_list) {
		if (brd->brd_number == i)
			goto out;
	}

	brd = brd_alloc(i);
	if (brd) {
		add_disk(brd->brd_disk);
		list_add_tail(&brd->brd_list, &brd_devices);
	}
out:
	return brd;
}

static void brd_del_one(struct brd_device *brd)
{
	list_del(&brd->brd_list);
	del_gendisk(brd->brd_disk);
	brd_free(brd);
}

static struct kobject *brd_probe(dev_t dev, int *part, void *data)
{
	struct brd_device *brd;
	struct kobject *kobj;

	mutex_lock(&brd_devices_mutex);
	brd = brd_init_one(dev & MINORMASK);
	kobj = brd ? get_disk(brd->brd_disk) : ERR_PTR(-ENOMEM);
	mutex_unlock(&brd_devices_mutex);

	*part = 0;
	return kobj;
}

static int __init brd_init(void)
{
	int i, nr;
	unsigned long range;
	struct brd_device *brd, *next;

	/*
	 * brd module now has a feature to instantiate underlying device
	 * structure on-demand, provided that there is an access dev node.
	 * However, this will not work well with user space tool that doesn't
	 * know about such "feature".  In order to not break any existing
	 * tool, we do the following:
	 *
	 * (1) if rd_nr is specified, create that many upfront, and this
	 *     also becomes a hard limit.
	 * (2) if rd_nr is not specified, create 1 rd device on module
	 *     load, user can further extend brd device by create dev node
	 *     themselves and have kernel automatically instantiate actual
	 *     device on-demand.
	 */

	part_shift = 0;
	if (max_part > 0)
		part_shift = fls(max_part);

	if (rd_nr > 1UL << (MINORBITS - part_shift))
		return -EINVAL;

	if (rd_nr) {
		nr = rd_nr;
		range = rd_nr;
	} else {
		nr = CONFIG_BLK_DEV_RAM_COUNT;
		range = 1UL << (MINORBITS - part_shift);
	}

	if (register_blkdev(RAMDISK_MAJOR, "ramdisk"))
		return -EIO;

	for (i = 0; i < nr; i++) {
		brd = brd_alloc(i);
		if (!brd)
			goto out_free;
		list_add_tail(&brd->brd_list, &brd_devices);
	}

	/* point of no return */

	list_for_each_entry(brd, &brd_devices, brd_list)
		add_disk(brd->brd_disk);

	blk_register_region(MKDEV(RAMDISK_MAJOR, 0), range,
				  THIS_MODULE, brd_probe, NULL, NULL);

	printk(KERN_INFO "brd: module loaded\n");
	return 0;

out_free:
	list_for_each_entry_safe(brd, next, &brd_devices, brd_list) {
		list_del(&brd->brd_list);
		brd_free(brd);
	}
	unregister_blkdev(RAMDISK_MAJOR, "ramdisk");

	return -ENOMEM;
}

static void __exit brd_exit(void)
{
	unsigned long range;
	struct brd_device *brd, *next;

	range = rd_nr ? rd_nr :  1UL << (MINORBITS - part_shift);

	list_for_each_entry_safe(brd, next, &brd_devices, brd_list)
		brd_del_one(brd);

	blk_unregister_region(MKDEV(RAMDISK_MAJOR, 0), range);
	unregister_blkdev(RAMDISK_MAJOR, "ramdisk");
}

module_init(brd_init);
module_exit(brd_exit);
Commit	Line	Data
9db5579b NP	1	/*
	2	* Ram backed block device driver.
	3	*
	4	* Copyright (C) 2007 Nick Piggin
	5	* Copyright (C) 2007 Novell Inc.
	6	*
	7	* Parts derived from drivers/block/rd.c, and drivers/block/loop.c, copyright
	8	* of their respective owners.
	9	*/
	10
	11	#include <linux/init.h>
	12	#include <linux/module.h>
	13	#include <linux/moduleparam.h>
	14	#include <linux/major.h>
	15	#include <linux/blkdev.h>
	16	#include <linux/bio.h>
	17	#include <linux/highmem.h>
2a48fc0a	18	#include <linux/mutex.h>
9db5579b NP	19	#include <linux/radix-tree.h>
9db5579b NP	20	#include <linux/buffer_head.h> /* invalidate_bh_lrus() */
5a0e3ad6	21	#include <linux/slab.h>
9db5579b NP	22
	23	#include <asm/uaccess.h>
	24
	25	#define SECTOR_SHIFT 9
	26	#define PAGE_SECTORS_SHIFT (PAGE_SHIFT - SECTOR_SHIFT)
	27	#define PAGE_SECTORS (1 << PAGE_SECTORS_SHIFT)
	28
	29	/*
	30	* Each block ramdisk device has a radix_tree brd_pages of pages that stores
	31	* the pages containing the block device's contents. A brd page's ->index is
	32	* its offset in PAGE_SIZE units. This is similar to, but in no way connected
	33	* with, the kernel's pagecache or buffer cache (which sit above our block
	34	* device).
	35	*/
	36	struct brd_device {
	37	int brd_number;
	38	int brd_refcnt;
	39	loff_t brd_offset;
	40	loff_t brd_sizelimit;
	41	unsigned brd_blocksize;
	42
	43	struct request_queue *brd_queue;
	44	struct gendisk *brd_disk;
	45	struct list_head brd_list;
	46
	47	/*
	48	* Backing store of pages and lock to protect it. This is the contents
	49	* of the block device.
	50	*/
	51	spinlock_t brd_lock;
	52	struct radix_tree_root brd_pages;
	53	};
	54
	55	/*
	56	* Look up and return a brd's page for a given sector.
	57	*/
2a48fc0a	58	static DEFINE_MUTEX(brd_mutex);
9db5579b NP	59	static struct page brd_lookup_page(struct brd_device brd, sector_t sector)
	60	{
	61	pgoff_t idx;
	62	struct page *page;
	63
	64	/*
	65	* The page lifetime is protected by the fact that we have opened the
	66	* device node -- brd pages will never be deleted under us, so we
	67	* don't need any further locking or refcounting.
	68	*
	69	* This is strictly true for the radix-tree nodes as well (ie. we
	70	* don't actually need the rcu_read_lock()), however that is not a
	71	* documented feature of the radix-tree API so it is better to be
	72	* safe here (we don't have total exclusion from radix tree updates
	73	* here, only deletes).
	74	*/
	75	rcu_read_lock();
	76	idx = sector >> PAGE_SECTORS_SHIFT; /* sector to page index */
	77	page = radix_tree_lookup(&brd->brd_pages, idx);
	78	rcu_read_unlock();
	79
	80	BUG_ON(page && page->index != idx);
	81
	82	return page;
	83	}
	84
	85	/*
	86	* Look up and return a brd's page for a given sector.
	87	* If one does not exist, allocate an empty page, and insert that. Then
	88	* return it.
	89	*/
	90	static struct page brd_insert_page(struct brd_device brd, sector_t sector)
	91	{
	92	pgoff_t idx;
	93	struct page *page;
75acb9cd	94	gfp_t gfp_flags;
9db5579b NP	95
	96	page = brd_lookup_page(brd, sector);
	97	if (page)
	98	return page;
	99
	100	/*
	101	* Must use NOIO because we don't want to recurse back into the
	102	* block or filesystem layers from page reclaim.
75acb9cd NP	103	*
	104	* Cannot support XIP and highmem, because our ->direct_access
	105	* routine for XIP must return memory that is always addressable.
	106	* If XIP was reworked to use pfns and kmap throughout, this
	107	* restriction might be able to be lifted.
9db5579b	108	*/
75acb9cd NP	109	gfp_flags = GFP_NOIO \| __GFP_ZERO;
	110	#ifndef CONFIG_BLK_DEV_XIP
	111	gfp_flags \|= __GFP_HIGHMEM;
	112	#endif
26defe34	113	page = alloc_page(gfp_flags);
9db5579b NP	114	if (!page)
	115	return NULL;
	116
	117	if (radix_tree_preload(GFP_NOIO)) {
	118	__free_page(page);
	119	return NULL;
	120	}
	121
	122	spin_lock(&brd->brd_lock);
	123	idx = sector >> PAGE_SECTORS_SHIFT;
	124	if (radix_tree_insert(&brd->brd_pages, idx, page)) {
	125	__free_page(page);
	126	page = radix_tree_lookup(&brd->brd_pages, idx);
	127	BUG_ON(!page);
	128	BUG_ON(page->index != idx);
	129	} else
	130	page->index = idx;
	131	spin_unlock(&brd->brd_lock);
	132
	133	radix_tree_preload_end();
	134
	135	return page;
	136	}
	137
b7c33571 NP	138	static void brd_free_page(struct brd_device *brd, sector_t sector)
	139	{
	140	struct page *page;
	141	pgoff_t idx;
	142
	143	spin_lock(&brd->brd_lock);
	144	idx = sector >> PAGE_SECTORS_SHIFT;
	145	page = radix_tree_delete(&brd->brd_pages, idx);
	146	spin_unlock(&brd->brd_lock);
	147	if (page)
	148	__free_page(page);
	149	}
	150
	151	static void brd_zero_page(struct brd_device *brd, sector_t sector)
	152	{
	153	struct page *page;
	154
	155	page = brd_lookup_page(brd, sector);
	156	if (page)
	157	clear_highpage(page);
	158	}
	159
9db5579b NP	160	/*
	161	* Free all backing store pages and radix tree. This must only be called when
	162	* there are no other users of the device.
	163	*/
	164	#define FREE_BATCH 16
	165	static void brd_free_pages(struct brd_device *brd)
	166	{
	167	unsigned long pos = 0;
	168	struct page *pages[FREE_BATCH];
	169	int nr_pages;
	170
	171	do {
	172	int i;
	173
	174	nr_pages = radix_tree_gang_lookup(&brd->brd_pages,
	175	(void **)pages, pos, FREE_BATCH);
	176
	177	for (i = 0; i < nr_pages; i++) {
	178	void *ret;
	179
	180	BUG_ON(pages[i]->index < pos);
	181	pos = pages[i]->index;
	182	ret = radix_tree_delete(&brd->brd_pages, pos);
	183	BUG_ON(!ret \|\| ret != pages[i]);
	184	__free_page(pages[i]);
	185	}
	186
	187	pos++;
	188
	189	/*
	190	* This assumes radix_tree_gang_lookup always returns as
	191	* many pages as possible. If the radix-tree code changes,
	192	* so will this have to.
	193	*/
	194	} while (nr_pages == FREE_BATCH);
	195	}
	196
	197	/*
	198	* copy_to_brd_setup must be called before copy_to_brd. It may sleep.
	199	*/
	200	static int copy_to_brd_setup(struct brd_device *brd, sector_t sector, size_t n)
	201	{
	202	unsigned int offset = (sector & (PAGE_SECTORS-1)) << SECTOR_SHIFT;
	203	size_t copy;
	204
	205	copy = min_t(size_t, n, PAGE_SIZE - offset);
	206	if (!brd_insert_page(brd, sector))
	207	return -ENOMEM;
	208	if (copy < n) {
	209	sector += copy >> SECTOR_SHIFT;
	210	if (!brd_insert_page(brd, sector))
	211	return -ENOMEM;
	212	}
	213	return 0;
	214	}
	215
b7c33571 NP	216	static void discard_from_brd(struct brd_device *brd,
	217	sector_t sector, size_t n)
	218	{
	219	while (n >= PAGE_SIZE) {
	220	/*
	221	* Don't want to actually discard pages here because
	222	* re-allocating the pages can result in writeback
	223	* deadlocks under heavy load.
	224	*/
	225	if (0)
	226	brd_free_page(brd, sector);
	227	else
	228	brd_zero_page(brd, sector);
	229	sector += PAGE_SIZE >> SECTOR_SHIFT;
	230	n -= PAGE_SIZE;
	231	}
	232	}
	233
9db5579b NP	234	/*
	235	* Copy n bytes from src to the brd starting at sector. Does not sleep.
	236	*/
	237	static void copy_to_brd(struct brd_device brd, const void src,
	238	sector_t sector, size_t n)
	239	{
	240	struct page *page;
	241	void *dst;
	242	unsigned int offset = (sector & (PAGE_SECTORS-1)) << SECTOR_SHIFT;
	243	size_t copy;
	244
	245	copy = min_t(size_t, n, PAGE_SIZE - offset);
	246	page = brd_lookup_page(brd, sector);
	247	BUG_ON(!page);
	248
	249	dst = kmap_atomic(page, KM_USER1);
	250	memcpy(dst + offset, src, copy);
	251	kunmap_atomic(dst, KM_USER1);
	252
	253	if (copy < n) {
	254	src += copy;
	255	sector += copy >> SECTOR_SHIFT;
	256	copy = n - copy;
	257	page = brd_lookup_page(brd, sector);
	258	BUG_ON(!page);
	259
	260	dst = kmap_atomic(page, KM_USER1);
	261	memcpy(dst, src, copy);
	262	kunmap_atomic(dst, KM_USER1);
	263	}
	264	}
	265
	266	/*
	267	* Copy n bytes to dst from the brd starting at sector. Does not sleep.
	268	*/
	269	static void copy_from_brd(void dst, struct brd_device brd,
	270	sector_t sector, size_t n)
	271	{
	272	struct page *page;
	273	void *src;
	274	unsigned int offset = (sector & (PAGE_SECTORS-1)) << SECTOR_SHIFT;
	275	size_t copy;
	276
	277	copy = min_t(size_t, n, PAGE_SIZE - offset);
	278	page = brd_lookup_page(brd, sector);
	279	if (page) {
	280	src = kmap_atomic(page, KM_USER1);
	281	memcpy(dst, src + offset, copy);
	282	kunmap_atomic(src, KM_USER1);
	283	} else
	284	memset(dst, 0, copy);
	285
	286	if (copy < n) {
	287	dst += copy;
	288	sector += copy >> SECTOR_SHIFT;
	289	copy = n - copy;
	290	page = brd_lookup_page(brd, sector);
	291	if (page) {
	292	src = kmap_atomic(page, KM_USER1);
	293	memcpy(dst, src, copy);
	294	kunmap_atomic(src, KM_USER1);
	295	} else
	296	memset(dst, 0, copy);
	297	}
298	}
299
300	/*
301	* Process a single bvec of a bio.
302	*/
303	static int brd_do_bvec(struct brd_device brd, struct page page,
304	unsigned int len, unsigned int off, int rw,
305	sector_t sector)
306	{
307	void *mem;
308	int err = 0;
309
310	if (rw != READ) {
311	err = copy_to_brd_setup(brd, sector, len);
312	if (err)
313	goto out;
314	}
315
316	mem = kmap_atomic(page, KM_USER0);
317	if (rw == READ) {
318	copy_from_brd(mem + off, brd, sector, len);
319	flush_dcache_page(page);
c2572f2b NP	320	} else {
c2572f2b NP	321	flush_dcache_page(page);
9db5579b	322	copy_to_brd(brd, mem + off, sector, len);
c2572f2b	323	}
9db5579b NP	324	kunmap_atomic(mem, KM_USER0);
	325
	326	out:
	327	return err;
	328	}
	329
	330	static int brd_make_request(struct request_queue q, struct bio bio)
	331	{
	332	struct block_device *bdev = bio->bi_bdev;
	333	struct brd_device *brd = bdev->bd_disk->private_data;
	334	int rw;
	335	struct bio_vec *bvec;
	336	sector_t sector;
	337	int i;
	338	int err = -EIO;
	339
	340	sector = bio->bi_sector;
	341	if (sector + (bio->bi_size >> SECTOR_SHIFT) >
	342	get_capacity(bdev->bd_disk))
	343	goto out;
	344
7b6d91da	345	if (unlikely(bio->bi_rw & REQ_DISCARD)) {
b7c33571 NP	346	err = 0;
	347	discard_from_brd(brd, sector, bio->bi_size);
	348	goto out;
	349	}
	350
9db5579b NP	351	rw = bio_rw(bio);
	352	if (rw == READA)
	353	rw = READ;
	354
	355	bio_for_each_segment(bvec, bio, i) {
	356	unsigned int len = bvec->bv_len;
	357	err = brd_do_bvec(brd, bvec->bv_page, len,
	358	bvec->bv_offset, rw, sector);
	359	if (err)
	360	break;
	361	sector += len >> SECTOR_SHIFT;
	362	}
	363
	364	out:
	365	bio_endio(bio, err);
	366
	367	return 0;
	368	}
	369
75acb9cd	370	#ifdef CONFIG_BLK_DEV_XIP
b7c33571	371	static int brd_direct_access(struct block_device *bdev, sector_t sector,
30afcb4b	372	void *kaddr, unsigned long pfn)
75acb9cd NP	373	{
	374	struct brd_device *brd = bdev->bd_disk->private_data;
	375	struct page *page;
	376
	377	if (!brd)
	378	return -ENODEV;
	379	if (sector & (PAGE_SECTORS-1))
	380	return -EINVAL;
	381	if (sector + PAGE_SECTORS > get_capacity(bdev->bd_disk))
	382	return -ERANGE;
	383	page = brd_insert_page(brd, sector);
	384	if (!page)
	385	return -ENOMEM;
30afcb4b JH	386	*kaddr = page_address(page);
30afcb4b JH	387	*pfn = page_to_pfn(page);
75acb9cd NP	388
	389	return 0;
	390	}
	391	#endif
	392
2b9ecd03	393	static int brd_ioctl(struct block_device *bdev, fmode_t mode,
9db5579b NP	394	unsigned int cmd, unsigned long arg)
	395	{
	396	int error;
9db5579b NP	397	struct brd_device *brd = bdev->bd_disk->private_data;
	398
	399	if (cmd != BLKFLSBUF)
	400	return -ENOTTY;
	401
	402	/*
	403	* ram device BLKFLSBUF has special semantics, we want to actually
	404	* release and destroy the ramdisk data.
	405	*/
2a48fc0a	406	mutex_lock(&brd_mutex);
9db5579b NP	407	mutex_lock(&bdev->bd_mutex);
	408	error = -EBUSY;
	409	if (bdev->bd_openers <= 1) {
	410	/*
	411	* Invalidate the cache first, so it isn't written
	412	* back to the device.
	413	*
	414	* Another thread might instantiate more buffercache here,
	415	* but there is not much we can do to close that race.
	416	*/
	417	invalidate_bh_lrus();
	418	truncate_inode_pages(bdev->bd_inode->i_mapping, 0);
	419	brd_free_pages(brd);
	420	error = 0;
	421	}
	422	mutex_unlock(&bdev->bd_mutex);
2a48fc0a	423	mutex_unlock(&brd_mutex);
9db5579b NP	424
	425	return error;
	426	}
	427
83d5cde4	428	static const struct block_device_operations brd_fops = {
75acb9cd	429	.owner = THIS_MODULE,
8a6cfeb6	430	.ioctl = brd_ioctl,
75acb9cd NP	431	#ifdef CONFIG_BLK_DEV_XIP
	432	.direct_access = brd_direct_access,
	433	#endif
9db5579b NP	434	};
	435
	436	/*
	437	* And now the modules code and kernel interface.
	438	*/
	439	static int rd_nr;
	440	int rd_size = CONFIG_BLK_DEV_RAM_SIZE;
d7853d1f LV	441	static int max_part;
d7853d1f LV	442	static int part_shift;
9db5579b NP	443	module_param(rd_nr, int, 0);
	444	MODULE_PARM_DESC(rd_nr, "Maximum number of brd devices");
	445	module_param(rd_size, int, 0);
	446	MODULE_PARM_DESC(rd_size, "Size of each RAM disk in kbytes.");
d7853d1f LV	447	module_param(max_part, int, 0);
d7853d1f LV	448	MODULE_PARM_DESC(max_part, "Maximum number of partitions per RAM disk");
9db5579b NP	449	MODULE_LICENSE("GPL");
9db5579b NP	450	MODULE_ALIAS_BLOCKDEV_MAJOR(RAMDISK_MAJOR);
efedf51c	451	MODULE_ALIAS("rd");
9db5579b NP	452
	453	#ifndef MODULE
	454	/* Legacy boot options - nonmodular */
	455	static int __init ramdisk_size(char *str)
	456	{
	457	rd_size = simple_strtol(str, NULL, 0);
	458	return 1;
	459	}
1adbee50	460	__setup("ramdisk_size=", ramdisk_size);
9db5579b NP	461	#endif
	462
	463	/*
	464	* The device scheme is derived from loop.c. Keep them in synch where possible
	465	* (should share code eventually).
	466	*/
	467	static LIST_HEAD(brd_devices);
	468	static DEFINE_MUTEX(brd_devices_mutex);
	469
	470	static struct brd_device *brd_alloc(int i)
	471	{
	472	struct brd_device *brd;
	473	struct gendisk *disk;
	474
	475	brd = kzalloc(sizeof(*brd), GFP_KERNEL);
	476	if (!brd)
	477	goto out;
	478	brd->brd_number = i;
	479	spin_lock_init(&brd->brd_lock);
	480	INIT_RADIX_TREE(&brd->brd_pages, GFP_ATOMIC);
	481
	482	brd->brd_queue = blk_alloc_queue(GFP_KERNEL);
	483	if (!brd->brd_queue)
	484	goto out_free_dev;
	485	blk_queue_make_request(brd->brd_queue, brd_make_request);
086fa5ff	486	blk_queue_max_hw_sectors(brd->brd_queue, 1024);
9db5579b NP	487	blk_queue_bounce_limit(brd->brd_queue, BLK_BOUNCE_ANY);
9db5579b NP	488
b7c33571 NP	489	brd->brd_queue->limits.discard_granularity = PAGE_SIZE;
	490	brd->brd_queue->limits.max_discard_sectors = UINT_MAX;
	491	brd->brd_queue->limits.discard_zeroes_data = 1;
	492	queue_flag_set_unlocked(QUEUE_FLAG_DISCARD, brd->brd_queue);
	493
d7853d1f	494	disk = brd->brd_disk = alloc_disk(1 << part_shift);
9db5579b NP	495	if (!disk)
	496	goto out_free_queue;
	497	disk->major = RAMDISK_MAJOR;
d7853d1f	498	disk->first_minor = i << part_shift;
9db5579b NP	499	disk->fops = &brd_fops;
	500	disk->private_data = brd;
	501	disk->queue = brd->brd_queue;
53978d0a	502	disk->flags \|= GENHD_FL_SUPPRESS_PARTITION_INFO;
9db5579b NP	503	sprintf(disk->disk_name, "ram%d", i);
	504	set_capacity(disk, rd_size * 2);
	505
	506	return brd;
	507
	508	out_free_queue:
	509	blk_cleanup_queue(brd->brd_queue);
	510	out_free_dev:
	511	kfree(brd);
	512	out:
	513	return NULL;
	514	}
	515
	516	static void brd_free(struct brd_device *brd)
	517	{
	518	put_disk(brd->brd_disk);
	519	blk_cleanup_queue(brd->brd_queue);
	520	brd_free_pages(brd);
	521	kfree(brd);
	522	}
	523
	524	static struct brd_device *brd_init_one(int i)
	525	{
	526	struct brd_device *brd;
	527
	528	list_for_each_entry(brd, &brd_devices, brd_list) {
	529	if (brd->brd_number == i)
	530	goto out;
	531	}
	532
	533	brd = brd_alloc(i);
	534	if (brd) {
	535	add_disk(brd->brd_disk);
	536	list_add_tail(&brd->brd_list, &brd_devices);
	537	}
	538	out:
	539	return brd;
	540	}
	541
	542	static void brd_del_one(struct brd_device *brd)
	543	{
	544	list_del(&brd->brd_list);
	545	del_gendisk(brd->brd_disk);
	546	brd_free(brd);
	547	}
	548
	549	static struct kobject brd_probe(dev_t dev, int part, void *data)
	550	{
	551	struct brd_device *brd;
	552	struct kobject *kobj;
	553
	554	mutex_lock(&brd_devices_mutex);
	555	brd = brd_init_one(dev & MINORMASK);
	556	kobj = brd ? get_disk(brd->brd_disk) : ERR_PTR(-ENOMEM);
	557	mutex_unlock(&brd_devices_mutex);
	558
	559	*part = 0;
	560	return kobj;
	561	}
	562
	563	static int __init brd_init(void)
	564	{
	565	int i, nr;
	566	unsigned long range;
567	struct brd_device brd, next;
568
569	/*
570	* brd module now has a feature to instantiate underlying device
571	* structure on-demand, provided that there is an access dev node.
572	* However, this will not work well with user space tool that doesn't
573	* know about such "feature". In order to not break any existing
574	* tool, we do the following:
575	*
576	* (1) if rd_nr is specified, create that many upfront, and this
577	* also becomes a hard limit.
578	* (2) if rd_nr is not specified, create 1 rd device on module
579	* load, user can further extend brd device by create dev node
580	* themselves and have kernel automatically instantiate actual
581	* device on-demand.
582	*/
d7853d1f LV	583
	584	part_shift = 0;
	585	if (max_part > 0)
	586	part_shift = fls(max_part);
	587
	588	if (rd_nr > 1UL << (MINORBITS - part_shift))
9db5579b NP	589	return -EINVAL;
	590
	591	if (rd_nr) {
	592	nr = rd_nr;
	593	range = rd_nr;
	594	} else {
	595	nr = CONFIG_BLK_DEV_RAM_COUNT;
d7853d1f	596	range = 1UL << (MINORBITS - part_shift);
9db5579b NP	597	}
	598
	599	if (register_blkdev(RAMDISK_MAJOR, "ramdisk"))
	600	return -EIO;
	601
	602	for (i = 0; i < nr; i++) {
	603	brd = brd_alloc(i);
	604	if (!brd)
	605	goto out_free;
	606	list_add_tail(&brd->brd_list, &brd_devices);
	607	}
	608
	609	/* point of no return */
	610
	611	list_for_each_entry(brd, &brd_devices, brd_list)
	612	add_disk(brd->brd_disk);
	613
	614	blk_register_region(MKDEV(RAMDISK_MAJOR, 0), range,
	615	THIS_MODULE, brd_probe, NULL, NULL);
	616
	617	printk(KERN_INFO "brd: module loaded\n");
	618	return 0;
	619
	620	out_free:
	621	list_for_each_entry_safe(brd, next, &brd_devices, brd_list) {
	622	list_del(&brd->brd_list);
	623	brd_free(brd);
	624	}
c82f2966	625	unregister_blkdev(RAMDISK_MAJOR, "ramdisk");
9db5579b	626
9db5579b NP	627	return -ENOMEM;
	628	}
	629
	630	static void __exit brd_exit(void)
	631	{
	632	unsigned long range;
	633	struct brd_device brd, next;
	634
d7853d1f	635	range = rd_nr ? rd_nr : 1UL << (MINORBITS - part_shift);
9db5579b NP	636
	637	list_for_each_entry_safe(brd, next, &brd_devices, brd_list)
	638	brd_del_one(brd);
	639
	640	blk_unregister_region(MKDEV(RAMDISK_MAJOR, 0), range);
	641	unregister_blkdev(RAMDISK_MAJOR, "ramdisk");
	642	}
	643
	644	module_init(brd_init);
	645	module_exit(brd_exit);
	646