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

// SPDX-License-Identifier: GPL-2.0-only
/*
 * 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/initrd.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/fs.h>
#include <linux/slab.h>
#include <linux/backing-dev.h>

#include <linux/uaccess.h>

#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;

	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.
	 */
	gfp_flags = GFP_NOIO | __GFP_ZERO | __GFP_HIGHMEM;
	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;
	page->index = idx;
	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);
	}
	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++;

		/*
		 * It takes 3.4 seconds to remove 80GiB ramdisk.
		 * So, we need cond_resched to avoid stalling the CPU.
		 */
		cond_resched();

		/*
		 * 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 -ENOSPC;
	if (copy < n) {
		sector += copy >> SECTOR_SHIFT;
		if (!brd_insert_page(brd, sector))
			return -ENOSPC;
	}
	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);
	memcpy(dst + offset, src, copy);
	kunmap_atomic(dst);

	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);
		memcpy(dst, src, copy);
		kunmap_atomic(dst);
	}
}

/*
 * 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);
		memcpy(dst, src + offset, copy);
		kunmap_atomic(src);
	} 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);
			memcpy(dst, src, copy);
			kunmap_atomic(src);
		} 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, unsigned int op,
			sector_t sector)
{
	void *mem;
	int err = 0;

	if (op_is_write(op)) {
		err = copy_to_brd_setup(brd, sector, len);
		if (err)
			goto out;
	}

	mem = kmap_atomic(page);
	if (!op_is_write(op)) {
		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);

out:
	return err;
}

static blk_qc_t brd_submit_bio(struct bio *bio)
{
	struct brd_device *brd = bio->bi_disk->private_data;
	sector_t sector = bio->bi_iter.bi_sector;
	struct bio_vec bvec;
	struct bvec_iter iter;

	bio_for_each_segment(bvec, bio, iter) {
		unsigned int len = bvec.bv_len;
		int err;

		/* Don't support un-aligned buffer */
		WARN_ON_ONCE((bvec.bv_offset & (SECTOR_SIZE - 1)) ||
				(len & (SECTOR_SIZE - 1)));

		err = brd_do_bvec(brd, bvec.bv_page, len, bvec.bv_offset,
				  bio_op(bio), sector);
		if (err)
			goto io_error;
		sector += len >> SECTOR_SHIFT;
	}

	bio_endio(bio);
	return BLK_QC_T_NONE;
io_error:
	bio_io_error(bio);
	return BLK_QC_T_NONE;
}

static int brd_rw_page(struct block_device *bdev, sector_t sector,
		       struct page *page, unsigned int op)
{
	struct brd_device *brd = bdev->bd_disk->private_data;
	int err;

	if (PageTransHuge(page))
		return -ENOTSUPP;
	err = brd_do_bvec(brd, page, PAGE_SIZE, 0, op, sector);
	page_endio(page, op_is_write(op), err);
	return err;
}

static const struct block_device_operations brd_fops = {
	.owner =		THIS_MODULE,
	.submit_bio =		brd_submit_bio,
	.rw_page =		brd_rw_page,
};

/*
 * And now the modules code and kernel interface.
 */
static int rd_nr = CONFIG_BLK_DEV_RAM_COUNT;
module_param(rd_nr, int, 0444);
MODULE_PARM_DESC(rd_nr, "Maximum number of brd devices");

unsigned long rd_size = CONFIG_BLK_DEV_RAM_SIZE;
module_param(rd_size, ulong, 0444);
MODULE_PARM_DESC(rd_size, "Size of each RAM disk in kbytes.");

static int max_part = 1;
module_param(max_part, int, 0444);
MODULE_PARM_DESC(max_part, "Num Minors to reserve between devices");

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(NUMA_NO_NODE);
	if (!brd->brd_queue)
		goto out_free_dev;

	/* This is so fdisk will align partitions on 4k, because of
	 * direct_access API needing 4k alignment, returning a PFN
	 * (This is only a problem on very small devices <= 4M,
	 *  otherwise fdisk will align on 1M. Regardless this call
	 *  is harmless)
	 */
	blk_queue_physical_block_size(brd->brd_queue, PAGE_SIZE);
	disk = brd->brd_disk = alloc_disk(max_part);
	if (!disk)
		goto out_free_queue;
	disk->major		= RAMDISK_MAJOR;
	disk->first_minor	= i * max_part;
	disk->fops		= &brd_fops;
	disk->private_data	= brd;
	disk->flags		= GENHD_FL_EXT_DEVT;
	sprintf(disk->disk_name, "ram%d", i);
	set_capacity(disk, rd_size * 2);

	/* Tell the block layer that this is not a rotational device */
	blk_queue_flag_set(QUEUE_FLAG_NONROT, brd->brd_queue);
	blk_queue_flag_clear(QUEUE_FLAG_ADD_RANDOM, brd->brd_queue);

	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 void brd_probe(dev_t dev)
{
	struct brd_device *brd;
	int i = MINOR(dev) / max_part;

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

	brd = brd_alloc(i);
	if (brd) {
		brd->brd_disk->queue = brd->brd_queue;
		add_disk(brd->brd_disk);
		list_add_tail(&brd->brd_list, &brd_devices);
	}

out_unlock:
	mutex_unlock(&brd_devices_mutex);
}

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

static inline void brd_check_and_reset_par(void)
{
	if (unlikely(!max_part))
		max_part = 1;

	/*
	 * make sure 'max_part' can be divided exactly by (1U << MINORBITS),
	 * otherwise, it is possiable to get same dev_t when adding partitions.
	 */
	if ((1U << MINORBITS) % max_part != 0)
		max_part = 1UL << fls(max_part);

	if (max_part > DISK_MAX_PARTS) {
		pr_info("brd: max_part can't be larger than %d, reset max_part = %d.\n",
			DISK_MAX_PARTS, DISK_MAX_PARTS);
		max_part = DISK_MAX_PARTS;
	}
}

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

	/*
	 * brd module now has a feature to instantiate underlying device
	 * structure on-demand, provided that there is an access dev node.
	 *
	 * (1) if rd_nr is specified, create that many upfront. else
	 *     it defaults to CONFIG_BLK_DEV_RAM_COUNT
	 * (2) User can further extend brd devices by create dev node themselves
	 *     and have kernel automatically instantiate actual device
	 *     on-demand. Example:
	 *		mknod /path/devnod_name b 1 X	# 1 is the rd major
	 *		fdisk -l /path/devnod_name
	 *	If (X / max_part) was not already created it will be created
	 *	dynamically.
	 */

	if (__register_blkdev(RAMDISK_MAJOR, "ramdisk", brd_probe))
		return -EIO;

	brd_check_and_reset_par();

	mutex_lock(&brd_devices_mutex);
	for (i = 0; i < rd_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) {
		/*
		 * associate with queue just before adding disk for
		 * avoiding to mess up failure path
		 */
		brd->brd_disk->queue = brd->brd_queue;
		add_disk(brd->brd_disk);
	}
	mutex_unlock(&brd_devices_mutex);

	pr_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);
	}
	mutex_unlock(&brd_devices_mutex);
	unregister_blkdev(RAMDISK_MAJOR, "ramdisk");

	pr_info("brd: module NOT loaded !!!\n");
	return -ENOMEM;
}

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

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

	unregister_blkdev(RAMDISK_MAJOR, "ramdisk");

	pr_info("brd: module unloaded\n");
}

module_init(brd_init);
module_exit(brd_exit);
Commit	Line	Data
09c434b8	1	// SPDX-License-Identifier: GPL-2.0-only
9db5579b NP	2	/*
	3	* Ram backed block device driver.
	4	*
	5	* Copyright (C) 2007 Nick Piggin
	6	* Copyright (C) 2007 Novell Inc.
	7	*
	8	* Parts derived from drivers/block/rd.c, and drivers/block/loop.c, copyright
	9	* of their respective owners.
	10	*/
	11
	12	#include <linux/init.h>
287f3ca5	13	#include <linux/initrd.h>
9db5579b NP	14	#include <linux/module.h>
	15	#include <linux/moduleparam.h>
	16	#include <linux/major.h>
	17	#include <linux/blkdev.h>
	18	#include <linux/bio.h>
	19	#include <linux/highmem.h>
2a48fc0a	20	#include <linux/mutex.h>
9db5579b	21	#include <linux/radix-tree.h>
ff01bb48	22	#include <linux/fs.h>
5a0e3ad6	23	#include <linux/slab.h>
23c47d2a	24	#include <linux/backing-dev.h>
9db5579b	25
7c0f6ba6	26	#include <linux/uaccess.h>
9db5579b	27
9db5579b NP	28	#define PAGE_SECTORS_SHIFT (PAGE_SHIFT - SECTOR_SHIFT)
	29	#define PAGE_SECTORS (1 << PAGE_SECTORS_SHIFT)
	30
	31	/*
	32	* Each block ramdisk device has a radix_tree brd_pages of pages that stores
	33	* the pages containing the block device's contents. A brd page's ->index is
	34	* its offset in PAGE_SIZE units. This is similar to, but in no way connected
	35	* with, the kernel's pagecache or buffer cache (which sit above our block
	36	* device).
	37	*/
	38	struct brd_device {
	39	int brd_number;
9db5579b NP	40
	41	struct request_queue *brd_queue;
	42	struct gendisk *brd_disk;
	43	struct list_head brd_list;
	44
	45	/*
	46	* Backing store of pages and lock to protect it. This is the contents
	47	* of the block device.
	48	*/
	49	spinlock_t brd_lock;
	50	struct radix_tree_root brd_pages;
	51	};
	52
	53	/*
	54	* Look up and return a brd's page for a given sector.
	55	*/
	56	static struct page brd_lookup_page(struct brd_device brd, sector_t sector)
	57	{
	58	pgoff_t idx;
	59	struct page *page;
	60
	61	/*
	62	* The page lifetime is protected by the fact that we have opened the
	63	* device node -- brd pages will never be deleted under us, so we
	64	* don't need any further locking or refcounting.
	65	*
	66	* This is strictly true for the radix-tree nodes as well (ie. we
	67	* don't actually need the rcu_read_lock()), however that is not a
	68	* documented feature of the radix-tree API so it is better to be
	69	* safe here (we don't have total exclusion from radix tree updates
	70	* here, only deletes).
	71	*/
	72	rcu_read_lock();
	73	idx = sector >> PAGE_SECTORS_SHIFT; /* sector to page index */
	74	page = radix_tree_lookup(&brd->brd_pages, idx);
	75	rcu_read_unlock();
	76
	77	BUG_ON(page && page->index != idx);
	78
	79	return page;
	80	}
	81
	82	/*
	83	* Look up and return a brd's page for a given sector.
	84	* If one does not exist, allocate an empty page, and insert that. Then
	85	* return it.
	86	*/
	87	static struct page brd_insert_page(struct brd_device brd, sector_t sector)
	88	{
	89	pgoff_t idx;
	90	struct page *page;
75acb9cd	91	gfp_t gfp_flags;
9db5579b NP	92
	93	page = brd_lookup_page(brd, sector);
	94	if (page)
	95	return page;
	96
	97	/*
	98	* Must use NOIO because we don't want to recurse back into the
	99	* block or filesystem layers from page reclaim.
	100	*/
f6b50160	101	gfp_flags = GFP_NOIO \| __GFP_ZERO \| __GFP_HIGHMEM;
26defe34	102	page = alloc_page(gfp_flags);
9db5579b NP	103	if (!page)
	104	return NULL;
	105
	106	if (radix_tree_preload(GFP_NOIO)) {
	107	__free_page(page);
	108	return NULL;
	109	}
	110
	111	spin_lock(&brd->brd_lock);
	112	idx = sector >> PAGE_SECTORS_SHIFT;
dfd20b2b	113	page->index = idx;
9db5579b NP	114	if (radix_tree_insert(&brd->brd_pages, idx, page)) {
	115	__free_page(page);
	116	page = radix_tree_lookup(&brd->brd_pages, idx);
	117	BUG_ON(!page);
	118	BUG_ON(page->index != idx);
dfd20b2b	119	}
9db5579b NP	120	spin_unlock(&brd->brd_lock);
	121
	122	radix_tree_preload_end();
	123
	124	return page;
	125	}
	126
	127	/*
	128	* Free all backing store pages and radix tree. This must only be called when
	129	* there are no other users of the device.
	130	*/
	131	#define FREE_BATCH 16
	132	static void brd_free_pages(struct brd_device *brd)
	133	{
	134	unsigned long pos = 0;
	135	struct page *pages[FREE_BATCH];
	136	int nr_pages;
	137
	138	do {
	139	int i;
	140
	141	nr_pages = radix_tree_gang_lookup(&brd->brd_pages,
	142	(void **)pages, pos, FREE_BATCH);
	143
	144	for (i = 0; i < nr_pages; i++) {
	145	void *ret;
	146
	147	BUG_ON(pages[i]->index < pos);
	148	pos = pages[i]->index;
	149	ret = radix_tree_delete(&brd->brd_pages, pos);
	150	BUG_ON(!ret \|\| ret != pages[i]);
	151	__free_page(pages[i]);
	152	}
	153
	154	pos++;
	155
936b33f7 MP	156	/*
	157	* It takes 3.4 seconds to remove 80GiB ramdisk.
	158	* So, we need cond_resched to avoid stalling the CPU.
	159	*/
	160	cond_resched();
	161
9db5579b NP	162	/*
	163	* This assumes radix_tree_gang_lookup always returns as
	164	* many pages as possible. If the radix-tree code changes,
	165	* so will this have to.
	166	*/
	167	} while (nr_pages == FREE_BATCH);
	168	}
	169
	170	/*
	171	* copy_to_brd_setup must be called before copy_to_brd. It may sleep.
	172	*/
	173	static int copy_to_brd_setup(struct brd_device *brd, sector_t sector, size_t n)
	174	{
	175	unsigned int offset = (sector & (PAGE_SECTORS-1)) << SECTOR_SHIFT;
	176	size_t copy;
	177
	178	copy = min_t(size_t, n, PAGE_SIZE - offset);
	179	if (!brd_insert_page(brd, sector))
96f8d8e0	180	return -ENOSPC;
9db5579b NP	181	if (copy < n) {
	182	sector += copy >> SECTOR_SHIFT;
	183	if (!brd_insert_page(brd, sector))
96f8d8e0	184	return -ENOSPC;
9db5579b NP	185	}
	186	return 0;
	187	}
	188
	189	/*
	190	* Copy n bytes from src to the brd starting at sector. Does not sleep.
	191	*/
	192	static void copy_to_brd(struct brd_device brd, const void src,
	193	sector_t sector, size_t n)
	194	{
	195	struct page *page;
	196	void *dst;
	197	unsigned int offset = (sector & (PAGE_SECTORS-1)) << SECTOR_SHIFT;
	198	size_t copy;
	199
	200	copy = min_t(size_t, n, PAGE_SIZE - offset);
	201	page = brd_lookup_page(brd, sector);
	202	BUG_ON(!page);
	203
cfd8005c	204	dst = kmap_atomic(page);
9db5579b	205	memcpy(dst + offset, src, copy);
cfd8005c	206	kunmap_atomic(dst);
9db5579b NP	207
	208	if (copy < n) {
	209	src += copy;
	210	sector += copy >> SECTOR_SHIFT;
	211	copy = n - copy;
	212	page = brd_lookup_page(brd, sector);
	213	BUG_ON(!page);
	214
cfd8005c	215	dst = kmap_atomic(page);
9db5579b	216	memcpy(dst, src, copy);
cfd8005c	217	kunmap_atomic(dst);
9db5579b NP	218	}
	219	}
	220
	221	/*
	222	* Copy n bytes to dst from the brd starting at sector. Does not sleep.
	223	*/
	224	static void copy_from_brd(void dst, struct brd_device brd,
	225	sector_t sector, size_t n)
	226	{
	227	struct page *page;
	228	void *src;
	229	unsigned int offset = (sector & (PAGE_SECTORS-1)) << SECTOR_SHIFT;
	230	size_t copy;
	231
	232	copy = min_t(size_t, n, PAGE_SIZE - offset);
	233	page = brd_lookup_page(brd, sector);
	234	if (page) {
cfd8005c	235	src = kmap_atomic(page);
9db5579b	236	memcpy(dst, src + offset, copy);
cfd8005c	237	kunmap_atomic(src);
9db5579b NP	238	} else
	239	memset(dst, 0, copy);
	240
	241	if (copy < n) {
	242	dst += copy;
	243	sector += copy >> SECTOR_SHIFT;
	244	copy = n - copy;
	245	page = brd_lookup_page(brd, sector);
	246	if (page) {
cfd8005c	247	src = kmap_atomic(page);
9db5579b	248	memcpy(dst, src, copy);
cfd8005c	249	kunmap_atomic(src);
9db5579b NP	250	} else
	251	memset(dst, 0, copy);
	252	}
	253	}
	254
	255	/*
	256	* Process a single bvec of a bio.
	257	*/
	258	static int brd_do_bvec(struct brd_device brd, struct page page,
3f289dcb	259	unsigned int len, unsigned int off, unsigned int op,
9db5579b NP	260	sector_t sector)
	261	{
	262	void *mem;
	263	int err = 0;
	264
3f289dcb	265	if (op_is_write(op)) {
9db5579b NP	266	err = copy_to_brd_setup(brd, sector, len);
	267	if (err)
	268	goto out;
	269	}
	270
cfd8005c	271	mem = kmap_atomic(page);
3f289dcb	272	if (!op_is_write(op)) {
9db5579b NP	273	copy_from_brd(mem + off, brd, sector, len);
9db5579b NP	274	flush_dcache_page(page);
c2572f2b NP	275	} else {
c2572f2b NP	276	flush_dcache_page(page);
9db5579b	277	copy_to_brd(brd, mem + off, sector, len);
c2572f2b	278	}
cfd8005c	279	kunmap_atomic(mem);
9db5579b NP	280
	281	out:
	282	return err;
	283	}
	284
c62b37d9	285	static blk_qc_t brd_submit_bio(struct bio *bio)
9db5579b	286	{
74d46992	287	struct brd_device *brd = bio->bi_disk->private_data;
74cb8994	288	sector_t sector = bio->bi_iter.bi_sector;
7988613b	289	struct bio_vec bvec;
7988613b	290	struct bvec_iter iter;
9db5579b	291
7988613b KO	292	bio_for_each_segment(bvec, bio, iter) {
7988613b KO	293	unsigned int len = bvec.bv_len;
4246a0b6 CH	294	int err;
4246a0b6 CH	295
f1acbf21 ML	296	/* Don't support un-aligned buffer */
	297	WARN_ON_ONCE((bvec.bv_offset & (SECTOR_SIZE - 1)) \|\|
	298	(len & (SECTOR_SIZE - 1)));
	299
c11f0c0b	300	err = brd_do_bvec(brd, bvec.bv_page, len, bvec.bv_offset,
3f289dcb	301	bio_op(bio), sector);
9db5579b	302	if (err)
4246a0b6	303	goto io_error;
9db5579b NP	304	sector += len >> SECTOR_SHIFT;
	305	}
	306
4246a0b6	307	bio_endio(bio);
dece1635	308	return BLK_QC_T_NONE;
4246a0b6 CH	309	io_error:
4246a0b6 CH	310	bio_io_error(bio);
dece1635	311	return BLK_QC_T_NONE;
9db5579b NP	312	}
9db5579b NP	313
a72132c3	314	static int brd_rw_page(struct block_device *bdev, sector_t sector,
3f289dcb	315	struct page *page, unsigned int op)
a72132c3 MW	316	{
a72132c3 MW	317	struct brd_device *brd = bdev->bd_disk->private_data;
98cc093c HY	318	int err;
	319
	320	if (PageTransHuge(page))
	321	return -ENOTSUPP;
3f289dcb TH	322	err = brd_do_bvec(brd, page, PAGE_SIZE, 0, op, sector);
3f289dcb TH	323	page_endio(page, op_is_write(op), err);
a72132c3 MW	324	return err;
	325	}
	326
83d5cde4	327	static const struct block_device_operations brd_fops = {
75acb9cd	328	.owner = THIS_MODULE,
c62b37d9	329	.submit_bio = brd_submit_bio,
a72132c3	330	.rw_page = brd_rw_page,
9db5579b NP	331	};
	332
	333	/*
	334	* And now the modules code and kernel interface.
	335	*/
937af5ec	336	static int rd_nr = CONFIG_BLK_DEV_RAM_COUNT;
5657a819	337	module_param(rd_nr, int, 0444);
9db5579b	338	MODULE_PARM_DESC(rd_nr, "Maximum number of brd devices");
937af5ec	339
366f4aea	340	unsigned long rd_size = CONFIG_BLK_DEV_RAM_SIZE;
5657a819	341	module_param(rd_size, ulong, 0444);
9db5579b	342	MODULE_PARM_DESC(rd_size, "Size of each RAM disk in kbytes.");
937af5ec BH	343
937af5ec BH	344	static int max_part = 1;
5657a819	345	module_param(max_part, int, 0444);
937af5ec BH	346	MODULE_PARM_DESC(max_part, "Num Minors to reserve between devices");
937af5ec BH	347
9db5579b NP	348	MODULE_LICENSE("GPL");
9db5579b NP	349	MODULE_ALIAS_BLOCKDEV_MAJOR(RAMDISK_MAJOR);
efedf51c	350	MODULE_ALIAS("rd");
9db5579b NP	351
	352	#ifndef MODULE
	353	/* Legacy boot options - nonmodular */
	354	static int __init ramdisk_size(char *str)
	355	{
	356	rd_size = simple_strtol(str, NULL, 0);
	357	return 1;
	358	}
1adbee50	359	__setup("ramdisk_size=", ramdisk_size);
9db5579b NP	360	#endif
	361
	362	/*
	363	* The device scheme is derived from loop.c. Keep them in synch where possible
	364	* (should share code eventually).
	365	*/
	366	static LIST_HEAD(brd_devices);
	367	static DEFINE_MUTEX(brd_devices_mutex);
	368
	369	static struct brd_device *brd_alloc(int i)
	370	{
	371	struct brd_device *brd;
	372	struct gendisk *disk;
	373
	374	brd = kzalloc(sizeof(*brd), GFP_KERNEL);
	375	if (!brd)
	376	goto out;
	377	brd->brd_number = i;
	378	spin_lock_init(&brd->brd_lock);
	379	INIT_RADIX_TREE(&brd->brd_pages, GFP_ATOMIC);
	380
c62b37d9	381	brd->brd_queue = blk_alloc_queue(NUMA_NO_NODE);
9db5579b NP	382	if (!brd->brd_queue)
9db5579b NP	383	goto out_free_dev;
c8fa3173	384
c8fa3173 BH	385	/* This is so fdisk will align partitions on 4k, because of
	386	* direct_access API needing 4k alignment, returning a PFN
	387	* (This is only a problem on very small devices <= 4M,
	388	* otherwise fdisk will align on 1M. Regardless this call
	389	* is harmless)
	390	*/
	391	blk_queue_physical_block_size(brd->brd_queue, PAGE_SIZE);
937af5ec	392	disk = brd->brd_disk = alloc_disk(max_part);
9db5579b NP	393	if (!disk)
	394	goto out_free_queue;
	395	disk->major = RAMDISK_MAJOR;
937af5ec	396	disk->first_minor = i * max_part;
9db5579b NP	397	disk->fops = &brd_fops;
9db5579b NP	398	disk->private_data = brd;
937af5ec	399	disk->flags = GENHD_FL_EXT_DEVT;
9db5579b NP	400	sprintf(disk->disk_name, "ram%d", i);
	401	set_capacity(disk, rd_size * 2);
	402
316ba573	403	/* Tell the block layer that this is not a rotational device */
153fcd5f ML	404	blk_queue_flag_set(QUEUE_FLAG_NONROT, brd->brd_queue);
153fcd5f ML	405	blk_queue_flag_clear(QUEUE_FLAG_ADD_RANDOM, brd->brd_queue);
316ba573	406
9db5579b NP	407	return brd;
	408
	409	out_free_queue:
	410	blk_cleanup_queue(brd->brd_queue);
	411	out_free_dev:
	412	kfree(brd);
	413	out:
	414	return NULL;
	415	}
	416
	417	static void brd_free(struct brd_device *brd)
	418	{
	419	put_disk(brd->brd_disk);
	420	blk_cleanup_queue(brd->brd_queue);
	421	brd_free_pages(brd);
	422	kfree(brd);
	423	}
	424
7cc178a6	425	static void brd_probe(dev_t dev)
9db5579b NP	426	{
9db5579b NP	427	struct brd_device *brd;
7cc178a6	428	int i = MINOR(dev) / max_part;
9db5579b	429
7cc178a6	430	mutex_lock(&brd_devices_mutex);
9db5579b NP	431	list_for_each_entry(brd, &brd_devices, brd_list) {
9db5579b NP	432	if (brd->brd_number == i)
7cc178a6	433	goto out_unlock;
9db5579b NP	434	}
	435
	436	brd = brd_alloc(i);
	437	if (brd) {
153fcd5f	438	brd->brd_disk->queue = brd->brd_queue;
9db5579b NP	439	add_disk(brd->brd_disk);
	440	list_add_tail(&brd->brd_list, &brd_devices);
	441	}
7cc178a6 CH	442
	443	out_unlock:
	444	mutex_unlock(&brd_devices_mutex);
9db5579b NP	445	}
	446
	447	static void brd_del_one(struct brd_device *brd)
	448	{
	449	list_del(&brd->brd_list);
	450	del_gendisk(brd->brd_disk);
	451	brd_free(brd);
	452	}
	453
c8ab4225 ZL	454	static inline void brd_check_and_reset_par(void)
	455	{
	456	if (unlikely(!max_part))
	457	max_part = 1;
	458
	459	/*
	460	* make sure 'max_part' can be divided exactly by (1U << MINORBITS),
	461	* otherwise, it is possiable to get same dev_t when adding partitions.
	462	*/
	463	if ((1U << MINORBITS) % max_part != 0)
	464	max_part = 1UL << fls(max_part);
	465
	466	if (max_part > DISK_MAX_PARTS) {
	467	pr_info("brd: max_part can't be larger than %d, reset max_part = %d.\n",
	468	DISK_MAX_PARTS, DISK_MAX_PARTS);
	469	max_part = DISK_MAX_PARTS;
	470	}
	471	}
	472
9db5579b NP	473	static int __init brd_init(void)
9db5579b NP	474	{
9db5579b	475	struct brd_device brd, next;
937af5ec	476	int i;
9db5579b NP	477
	478	/*
	479	* brd module now has a feature to instantiate underlying device
	480	* structure on-demand, provided that there is an access dev node.
9db5579b	481	*
937af5ec BH	482	* (1) if rd_nr is specified, create that many upfront. else
	483	* it defaults to CONFIG_BLK_DEV_RAM_COUNT
	484	* (2) User can further extend brd devices by create dev node themselves
	485	* and have kernel automatically instantiate actual device
	486	* on-demand. Example:
	487	* mknod /path/devnod_name b 1 X # 1 is the rd major
	488	* fdisk -l /path/devnod_name
	489	* If (X / max_part) was not already created it will be created
	490	* dynamically.
9db5579b	491	*/
d7853d1f	492
7cc178a6	493	if (__register_blkdev(RAMDISK_MAJOR, "ramdisk", brd_probe))
9db5579b NP	494	return -EIO;
9db5579b NP	495
c8ab4225	496	brd_check_and_reset_par();
937af5ec	497
7cc178a6	498	mutex_lock(&brd_devices_mutex);
937af5ec	499	for (i = 0; i < rd_nr; i++) {
9db5579b NP	500	brd = brd_alloc(i);
	501	if (!brd)
	502	goto out_free;
	503	list_add_tail(&brd->brd_list, &brd_devices);
	504	}
	505
	506	/* point of no return */
	507
153fcd5f ML	508	list_for_each_entry(brd, &brd_devices, brd_list) {
	509	/*
	510	* associate with queue just before adding disk for
	511	* avoiding to mess up failure path
	512	*/
	513	brd->brd_disk->queue = brd->brd_queue;
9db5579b	514	add_disk(brd->brd_disk);
153fcd5f	515	}
7cc178a6	516	mutex_unlock(&brd_devices_mutex);
9db5579b	517
937af5ec	518	pr_info("brd: module loaded\n");
9db5579b NP	519	return 0;
	520
	521	out_free:
	522	list_for_each_entry_safe(brd, next, &brd_devices, brd_list) {
	523	list_del(&brd->brd_list);
	524	brd_free(brd);
	525	}
7cc178a6	526	mutex_unlock(&brd_devices_mutex);
c82f2966	527	unregister_blkdev(RAMDISK_MAJOR, "ramdisk");
9db5579b	528
937af5ec	529	pr_info("brd: module NOT loaded !!!\n");
9db5579b NP	530	return -ENOMEM;
	531	}
	532
	533	static void __exit brd_exit(void)
	534	{
9db5579b NP	535	struct brd_device brd, next;
9db5579b NP	536
9db5579b NP	537	list_for_each_entry_safe(brd, next, &brd_devices, brd_list)
	538	brd_del_one(brd);
	539
9db5579b	540	unregister_blkdev(RAMDISK_MAJOR, "ramdisk");
937af5ec BH	541
937af5ec BH	542	pr_info("brd: module unloaded\n");
9db5579b NP	543	}
	544
	545	module_init(brd_init);
	546	module_exit(brd_exit);
	547