[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/gfp.h>
#include <linux/radix-tree.h>
#include <linux/buffer_head.h> /* invalidate_bh_lrus() */

#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 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;
}

/*
 * 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;
}

/*
 * 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;

	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(&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);

	return error;
}

static struct block_device_operations brd_fops = {
	.owner =		THIS_MODULE,
	.locked_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_ordered(brd->brd_queue, QUEUE_ORDERED_TAG, NULL);
	blk_queue_max_sectors(brd->brd_queue, 1024);
	blk_queue_bounce_limit(brd->brd_queue, BLK_BOUNCE_ANY);

	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>
	18	#include <linux/gfp.h>
	19	#include <linux/radix-tree.h>
	20	#include <linux/buffer_head.h> /* invalidate_bh_lrus() */
	21
	22	#include <asm/uaccess.h>
	23
	24	#define SECTOR_SHIFT 9
	25	#define PAGE_SECTORS_SHIFT (PAGE_SHIFT - SECTOR_SHIFT)
	26	#define PAGE_SECTORS (1 << PAGE_SECTORS_SHIFT)
	27
	28	/*
	29	* Each block ramdisk device has a radix_tree brd_pages of pages that stores
	30	* the pages containing the block device's contents. A brd page's ->index is
	31	* its offset in PAGE_SIZE units. This is similar to, but in no way connected
	32	* with, the kernel's pagecache or buffer cache (which sit above our block
	33	* device).
	34	*/
	35	struct brd_device {
	36	int brd_number;
	37	int brd_refcnt;
	38	loff_t brd_offset;
	39	loff_t brd_sizelimit;
	40	unsigned brd_blocksize;
	41
	42	struct request_queue *brd_queue;
	43	struct gendisk *brd_disk;
	44	struct list_head brd_list;
	45
	46	/*
	47	* Backing store of pages and lock to protect it. This is the contents
	48	* of the block device.
	49	*/
	50	spinlock_t brd_lock;
	51	struct radix_tree_root brd_pages;
	52	};
	53
	54	/*
	55	* Look up and return a brd's page for a given sector.
	56	*/
	57	static struct page brd_lookup_page(struct brd_device brd, sector_t sector)
	58	{
	59	pgoff_t idx;
	60	struct page *page;
	61
	62	/*
	63	* The page lifetime is protected by the fact that we have opened the
	64	* device node -- brd pages will never be deleted under us, so we
65	* don't need any further locking or refcounting.
66	*
67	* This is strictly true for the radix-tree nodes as well (ie. we
68	* don't actually need the rcu_read_lock()), however that is not a
69	* documented feature of the radix-tree API so it is better to be
70	* safe here (we don't have total exclusion from radix tree updates
71	* here, only deletes).
72	*/
73	rcu_read_lock();
74	idx = sector >> PAGE_SECTORS_SHIFT; /* sector to page index */
75	page = radix_tree_lookup(&brd->brd_pages, idx);
76	rcu_read_unlock();
77
78	BUG_ON(page && page->index != idx);
79
80	return page;
81	}
82
83	/*
84	* Look up and return a brd's page for a given sector.
85	* If one does not exist, allocate an empty page, and insert that. Then
86	* return it.
87	*/
88	static struct page brd_insert_page(struct brd_device brd, sector_t sector)
89	{
90	pgoff_t idx;
91	struct page *page;
75acb9cd	92	gfp_t gfp_flags;
9db5579b NP	93
	94	page = brd_lookup_page(brd, sector);
	95	if (page)
	96	return page;
	97
	98	/*
	99	* Must use NOIO because we don't want to recurse back into the
	100	* block or filesystem layers from page reclaim.
75acb9cd NP	101	*
	102	* Cannot support XIP and highmem, because our ->direct_access
	103	* routine for XIP must return memory that is always addressable.
	104	* If XIP was reworked to use pfns and kmap throughout, this
	105	* restriction might be able to be lifted.
9db5579b	106	*/
75acb9cd NP	107	gfp_flags = GFP_NOIO \| __GFP_ZERO;
	108	#ifndef CONFIG_BLK_DEV_XIP
	109	gfp_flags \|= __GFP_HIGHMEM;
	110	#endif
26defe34	111	page = alloc_page(gfp_flags);
9db5579b NP	112	if (!page)
	113	return NULL;
	114
	115	if (radix_tree_preload(GFP_NOIO)) {
	116	__free_page(page);
	117	return NULL;
	118	}
	119
	120	spin_lock(&brd->brd_lock);
	121	idx = sector >> PAGE_SECTORS_SHIFT;
	122	if (radix_tree_insert(&brd->brd_pages, idx, page)) {
	123	__free_page(page);
	124	page = radix_tree_lookup(&brd->brd_pages, idx);
	125	BUG_ON(!page);
	126	BUG_ON(page->index != idx);
	127	} else
	128	page->index = idx;
	129	spin_unlock(&brd->brd_lock);
	130
	131	radix_tree_preload_end();
	132
	133	return page;
	134	}
	135
	136	/*
	137	* Free all backing store pages and radix tree. This must only be called when
	138	* there are no other users of the device.
	139	*/
	140	#define FREE_BATCH 16
	141	static void brd_free_pages(struct brd_device *brd)
	142	{
	143	unsigned long pos = 0;
	144	struct page *pages[FREE_BATCH];
	145	int nr_pages;
	146
	147	do {
	148	int i;
	149
	150	nr_pages = radix_tree_gang_lookup(&brd->brd_pages,
	151	(void **)pages, pos, FREE_BATCH);
	152
	153	for (i = 0; i < nr_pages; i++) {
	154	void *ret;
	155
	156	BUG_ON(pages[i]->index < pos);
	157	pos = pages[i]->index;
	158	ret = radix_tree_delete(&brd->brd_pages, pos);
	159	BUG_ON(!ret \|\| ret != pages[i]);
	160	__free_page(pages[i]);
	161	}
	162
	163	pos++;
	164
	165	/*
	166	* This assumes radix_tree_gang_lookup always returns as
	167	* many pages as possible. If the radix-tree code changes,
	168	* so will this have to.
	169	*/
	170	} while (nr_pages == FREE_BATCH);
	171	}
	172
	173	/*
	174	* copy_to_brd_setup must be called before copy_to_brd. It may sleep.
	175	*/
176	static int copy_to_brd_setup(struct brd_device *brd, sector_t sector, size_t n)
177	{
178	unsigned int offset = (sector & (PAGE_SECTORS-1)) << SECTOR_SHIFT;
179	size_t copy;
180
181	copy = min_t(size_t, n, PAGE_SIZE - offset);
182	if (!brd_insert_page(brd, sector))
183	return -ENOMEM;
184	if (copy < n) {
185	sector += copy >> SECTOR_SHIFT;
186	if (!brd_insert_page(brd, sector))
187	return -ENOMEM;
188	}
189	return 0;
190	}
191
192	/*
193	* Copy n bytes from src to the brd starting at sector. Does not sleep.
194	*/
195	static void copy_to_brd(struct brd_device brd, const void src,
196	sector_t sector, size_t n)
197	{
198	struct page *page;
199	void *dst;
200	unsigned int offset = (sector & (PAGE_SECTORS-1)) << SECTOR_SHIFT;
201	size_t copy;
202
203	copy = min_t(size_t, n, PAGE_SIZE - offset);
204	page = brd_lookup_page(brd, sector);
205	BUG_ON(!page);
206
207	dst = kmap_atomic(page, KM_USER1);
208	memcpy(dst + offset, src, copy);
209	kunmap_atomic(dst, KM_USER1);
210
211	if (copy < n) {
212	src += copy;
213	sector += copy >> SECTOR_SHIFT;
214	copy = n - copy;
215	page = brd_lookup_page(brd, sector);
216	BUG_ON(!page);
217
218	dst = kmap_atomic(page, KM_USER1);
219	memcpy(dst, src, copy);
220	kunmap_atomic(dst, KM_USER1);
221	}
222	}
223
224	/*
225	* Copy n bytes to dst from the brd starting at sector. Does not sleep.
226	*/
227	static void copy_from_brd(void dst, struct brd_device brd,
228	sector_t sector, size_t n)
229	{
230	struct page *page;
231	void *src;
232	unsigned int offset = (sector & (PAGE_SECTORS-1)) << SECTOR_SHIFT;
233	size_t copy;
234
235	copy = min_t(size_t, n, PAGE_SIZE - offset);
236	page = brd_lookup_page(brd, sector);
237	if (page) {
238	src = kmap_atomic(page, KM_USER1);
239	memcpy(dst, src + offset, copy);
240	kunmap_atomic(src, KM_USER1);
241	} else
242	memset(dst, 0, copy);
243
244	if (copy < n) {
245	dst += copy;
246	sector += copy >> SECTOR_SHIFT;
247	copy = n - copy;
248	page = brd_lookup_page(brd, sector);
249	if (page) {
250	src = kmap_atomic(page, KM_USER1);
251	memcpy(dst, src, copy);
252	kunmap_atomic(src, KM_USER1);
253	} else
254	memset(dst, 0, copy);
255	}
256	}
257
258	/*
259	* Process a single bvec of a bio.
260	*/
261	static int brd_do_bvec(struct brd_device brd, struct page page,
262	unsigned int len, unsigned int off, int rw,
263	sector_t sector)
264	{
265	void *mem;
266	int err = 0;
267
268	if (rw != READ) {
269	err = copy_to_brd_setup(brd, sector, len);
270	if (err)
271	goto out;
272	}
273
274	mem = kmap_atomic(page, KM_USER0);
275	if (rw == READ) {
276	copy_from_brd(mem + off, brd, sector, len);
277	flush_dcache_page(page);
c2572f2b NP	278	} else {
c2572f2b NP	279	flush_dcache_page(page);
9db5579b	280	copy_to_brd(brd, mem + off, sector, len);
c2572f2b	281	}
9db5579b NP	282	kunmap_atomic(mem, KM_USER0);
	283
	284	out:
	285	return err;
	286	}
	287
	288	static int brd_make_request(struct request_queue q, struct bio bio)
	289	{
	290	struct block_device *bdev = bio->bi_bdev;
	291	struct brd_device *brd = bdev->bd_disk->private_data;
	292	int rw;
	293	struct bio_vec *bvec;
	294	sector_t sector;
	295	int i;
	296	int err = -EIO;
	297
	298	sector = bio->bi_sector;
	299	if (sector + (bio->bi_size >> SECTOR_SHIFT) >
	300	get_capacity(bdev->bd_disk))
	301	goto out;
	302
	303	rw = bio_rw(bio);
	304	if (rw == READA)
	305	rw = READ;
	306
	307	bio_for_each_segment(bvec, bio, i) {
	308	unsigned int len = bvec->bv_len;
	309	err = brd_do_bvec(brd, bvec->bv_page, len,
	310	bvec->bv_offset, rw, sector);
	311	if (err)
	312	break;
	313	sector += len >> SECTOR_SHIFT;
	314	}
	315
	316	out:
	317	bio_endio(bio, err);
	318
	319	return 0;
	320	}
	321
75acb9cd NP	322	#ifdef CONFIG_BLK_DEV_XIP
75acb9cd NP	323	static int brd_direct_access (struct block_device *bdev, sector_t sector,
30afcb4b	324	void *kaddr, unsigned long pfn)
75acb9cd NP	325	{
	326	struct brd_device *brd = bdev->bd_disk->private_data;
	327	struct page *page;
	328
	329	if (!brd)
	330	return -ENODEV;
	331	if (sector & (PAGE_SECTORS-1))
	332	return -EINVAL;
	333	if (sector + PAGE_SECTORS > get_capacity(bdev->bd_disk))
	334	return -ERANGE;
	335	page = brd_insert_page(brd, sector);
	336	if (!page)
	337	return -ENOMEM;
30afcb4b JH	338	*kaddr = page_address(page);
30afcb4b JH	339	*pfn = page_to_pfn(page);
75acb9cd NP	340
	341	return 0;
	342	}
	343	#endif
	344
2b9ecd03	345	static int brd_ioctl(struct block_device *bdev, fmode_t mode,
9db5579b NP	346	unsigned int cmd, unsigned long arg)
	347	{
	348	int error;
9db5579b NP	349	struct brd_device *brd = bdev->bd_disk->private_data;
	350
	351	if (cmd != BLKFLSBUF)
	352	return -ENOTTY;
	353
	354	/*
	355	* ram device BLKFLSBUF has special semantics, we want to actually
	356	* release and destroy the ramdisk data.
	357	*/
	358	mutex_lock(&bdev->bd_mutex);
	359	error = -EBUSY;
	360	if (bdev->bd_openers <= 1) {
	361	/*
	362	* Invalidate the cache first, so it isn't written
	363	* back to the device.
	364	*
	365	* Another thread might instantiate more buffercache here,
	366	* but there is not much we can do to close that race.
	367	*/
	368	invalidate_bh_lrus();
	369	truncate_inode_pages(bdev->bd_inode->i_mapping, 0);
	370	brd_free_pages(brd);
	371	error = 0;
	372	}
	373	mutex_unlock(&bdev->bd_mutex);
	374
	375	return error;
	376	}
	377
	378	static struct block_device_operations brd_fops = {
75acb9cd	379	.owner = THIS_MODULE,
2b9ecd03	380	.locked_ioctl = brd_ioctl,
75acb9cd NP	381	#ifdef CONFIG_BLK_DEV_XIP
	382	.direct_access = brd_direct_access,
	383	#endif
9db5579b NP	384	};
	385
	386	/*
	387	* And now the modules code and kernel interface.
	388	*/
	389	static int rd_nr;
	390	int rd_size = CONFIG_BLK_DEV_RAM_SIZE;
d7853d1f LV	391	static int max_part;
d7853d1f LV	392	static int part_shift;
9db5579b NP	393	module_param(rd_nr, int, 0);
	394	MODULE_PARM_DESC(rd_nr, "Maximum number of brd devices");
	395	module_param(rd_size, int, 0);
	396	MODULE_PARM_DESC(rd_size, "Size of each RAM disk in kbytes.");
d7853d1f LV	397	module_param(max_part, int, 0);
d7853d1f LV	398	MODULE_PARM_DESC(max_part, "Maximum number of partitions per RAM disk");
9db5579b NP	399	MODULE_LICENSE("GPL");
9db5579b NP	400	MODULE_ALIAS_BLOCKDEV_MAJOR(RAMDISK_MAJOR);
efedf51c	401	MODULE_ALIAS("rd");
9db5579b NP	402
	403	#ifndef MODULE
	404	/* Legacy boot options - nonmodular */
	405	static int __init ramdisk_size(char *str)
	406	{
	407	rd_size = simple_strtol(str, NULL, 0);
	408	return 1;
	409	}
1adbee50	410	__setup("ramdisk_size=", ramdisk_size);
9db5579b NP	411	#endif
	412
	413	/*
	414	* The device scheme is derived from loop.c. Keep them in synch where possible
	415	* (should share code eventually).
	416	*/
	417	static LIST_HEAD(brd_devices);
	418	static DEFINE_MUTEX(brd_devices_mutex);
	419
	420	static struct brd_device *brd_alloc(int i)
	421	{
	422	struct brd_device *brd;
	423	struct gendisk *disk;
	424
	425	brd = kzalloc(sizeof(*brd), GFP_KERNEL);
	426	if (!brd)
	427	goto out;
	428	brd->brd_number = i;
	429	spin_lock_init(&brd->brd_lock);
	430	INIT_RADIX_TREE(&brd->brd_pages, GFP_ATOMIC);
	431
	432	brd->brd_queue = blk_alloc_queue(GFP_KERNEL);
	433	if (!brd->brd_queue)
	434	goto out_free_dev;
	435	blk_queue_make_request(brd->brd_queue, brd_make_request);
dfbc4752	436	blk_queue_ordered(brd->brd_queue, QUEUE_ORDERED_TAG, NULL);
9db5579b NP	437	blk_queue_max_sectors(brd->brd_queue, 1024);
	438	blk_queue_bounce_limit(brd->brd_queue, BLK_BOUNCE_ANY);
	439
d7853d1f	440	disk = brd->brd_disk = alloc_disk(1 << part_shift);
9db5579b NP	441	if (!disk)
	442	goto out_free_queue;
	443	disk->major = RAMDISK_MAJOR;
d7853d1f	444	disk->first_minor = i << part_shift;
9db5579b NP	445	disk->fops = &brd_fops;
	446	disk->private_data = brd;
	447	disk->queue = brd->brd_queue;
53978d0a	448	disk->flags \|= GENHD_FL_SUPPRESS_PARTITION_INFO;
9db5579b NP	449	sprintf(disk->disk_name, "ram%d", i);
	450	set_capacity(disk, rd_size * 2);
	451
	452	return brd;
	453
	454	out_free_queue:
	455	blk_cleanup_queue(brd->brd_queue);
	456	out_free_dev:
	457	kfree(brd);
	458	out:
	459	return NULL;
	460	}
	461
	462	static void brd_free(struct brd_device *brd)
	463	{
	464	put_disk(brd->brd_disk);
	465	blk_cleanup_queue(brd->brd_queue);
	466	brd_free_pages(brd);
	467	kfree(brd);
	468	}
	469
	470	static struct brd_device *brd_init_one(int i)
	471	{
	472	struct brd_device *brd;
	473
	474	list_for_each_entry(brd, &brd_devices, brd_list) {
	475	if (brd->brd_number == i)
	476	goto out;
	477	}
	478
	479	brd = brd_alloc(i);
	480	if (brd) {
	481	add_disk(brd->brd_disk);
	482	list_add_tail(&brd->brd_list, &brd_devices);
	483	}
	484	out:
	485	return brd;
	486	}
	487
	488	static void brd_del_one(struct brd_device *brd)
	489	{
	490	list_del(&brd->brd_list);
	491	del_gendisk(brd->brd_disk);
	492	brd_free(brd);
	493	}
	494
	495	static struct kobject brd_probe(dev_t dev, int part, void *data)
	496	{
	497	struct brd_device *brd;
	498	struct kobject *kobj;
	499
	500	mutex_lock(&brd_devices_mutex);
	501	brd = brd_init_one(dev & MINORMASK);
	502	kobj = brd ? get_disk(brd->brd_disk) : ERR_PTR(-ENOMEM);
	503	mutex_unlock(&brd_devices_mutex);
	504
	505	*part = 0;
	506	return kobj;
	507	}
	508
	509	static int __init brd_init(void)
	510	{
	511	int i, nr;
	512	unsigned long range;
513	struct brd_device brd, next;
514
515	/*
516	* brd module now has a feature to instantiate underlying device
517	* structure on-demand, provided that there is an access dev node.
518	* However, this will not work well with user space tool that doesn't
519	* know about such "feature". In order to not break any existing
520	* tool, we do the following:
521	*
522	* (1) if rd_nr is specified, create that many upfront, and this
523	* also becomes a hard limit.
524	* (2) if rd_nr is not specified, create 1 rd device on module
525	* load, user can further extend brd device by create dev node
526	* themselves and have kernel automatically instantiate actual
527	* device on-demand.
528	*/
d7853d1f LV	529
	530	part_shift = 0;
	531	if (max_part > 0)
	532	part_shift = fls(max_part);
	533
	534	if (rd_nr > 1UL << (MINORBITS - part_shift))
9db5579b NP	535	return -EINVAL;
	536
	537	if (rd_nr) {
	538	nr = rd_nr;
	539	range = rd_nr;
	540	} else {
	541	nr = CONFIG_BLK_DEV_RAM_COUNT;
d7853d1f	542	range = 1UL << (MINORBITS - part_shift);
9db5579b NP	543	}
	544
	545	if (register_blkdev(RAMDISK_MAJOR, "ramdisk"))
	546	return -EIO;
	547
	548	for (i = 0; i < nr; i++) {
	549	brd = brd_alloc(i);
	550	if (!brd)
	551	goto out_free;
	552	list_add_tail(&brd->brd_list, &brd_devices);
	553	}
	554
	555	/* point of no return */
	556
	557	list_for_each_entry(brd, &brd_devices, brd_list)
	558	add_disk(brd->brd_disk);
	559
	560	blk_register_region(MKDEV(RAMDISK_MAJOR, 0), range,
	561	THIS_MODULE, brd_probe, NULL, NULL);
	562
	563	printk(KERN_INFO "brd: module loaded\n");
	564	return 0;
	565
	566	out_free:
	567	list_for_each_entry_safe(brd, next, &brd_devices, brd_list) {
	568	list_del(&brd->brd_list);
	569	brd_free(brd);
	570	}
c82f2966	571	unregister_blkdev(RAMDISK_MAJOR, "ramdisk");
9db5579b	572
9db5579b NP	573	return -ENOMEM;
	574	}
	575
	576	static void __exit brd_exit(void)
	577	{
	578	unsigned long range;
	579	struct brd_device brd, next;
	580
d7853d1f	581	range = rd_nr ? rd_nr : 1UL << (MINORBITS - part_shift);
9db5579b NP	582
	583	list_for_each_entry_safe(brd, next, &brd_devices, brd_list)
	584	brd_del_one(brd);
	585
	586	blk_unregister_region(MKDEV(RAMDISK_MAJOR, 0), range);
	587	unregister_blkdev(RAMDISK_MAJOR, "ramdisk");
	588	}
	589
	590	module_init(brd_init);
	591	module_exit(brd_exit);
	592