2 * linux/drivers/block/loop.c
4 * Written by Theodore Ts'o, 3/29/93
6 * Copyright 1993 by Theodore Ts'o. Redistribution of this file is
7 * permitted under the GNU General Public License.
9 * DES encryption plus some minor changes by Werner Almesberger, 30-MAY-1993
10 * more DES encryption plus IDEA encryption by Nicholas J. Leon, June 20, 1996
12 * Modularized and updated for 1.1.16 kernel - Mitch Dsouza 28th May 1994
13 * Adapted for 1.3.59 kernel - Andries Brouwer, 1 Feb 1996
15 * Fixed do_loop_request() re-entrancy - Vincent.Renardias@waw.com Mar 20, 1997
17 * Added devfs support - Richard Gooch <rgooch@atnf.csiro.au> 16-Jan-1998
19 * Handle sparse backing files correctly - Kenn Humborg, Jun 28, 1998
21 * Loadable modules and other fixes by AK, 1998
23 * Make real block number available to downstream transfer functions, enables
24 * CBC (and relatives) mode encryption requiring unique IVs per data block.
25 * Reed H. Petty, rhp@draper.net
27 * Maximum number of loop devices now dynamic via max_loop module parameter.
28 * Russell Kroll <rkroll@exploits.org> 19990701
30 * Maximum number of loop devices when compiled-in now selectable by passing
31 * max_loop=<1-255> to the kernel on boot.
32 * Erik I. Bolsø, <eriki@himolde.no>, Oct 31, 1999
34 * Completely rewrite request handling to be make_request_fn style and
35 * non blocking, pushing work to a helper thread. Lots of fixes from
37 * Jens Axboe <axboe@suse.de>, Nov 2000
39 * Support up to 256 loop devices
40 * Heinz Mauelshagen <mge@sistina.com>, Feb 2002
42 * Support for falling back on the write file operation when the address space
43 * operations prepare_write and/or commit_write are not available on the
45 * Anton Altaparmakov, 16 Feb 2005
48 * - Advisory locking is ignored here.
49 * - Should use an own CAP_* category instead of CAP_SYS_ADMIN
53 #include <linux/module.h>
54 #include <linux/moduleparam.h>
55 #include <linux/sched.h>
57 #include <linux/file.h>
58 #include <linux/stat.h>
59 #include <linux/errno.h>
60 #include <linux/major.h>
61 #include <linux/wait.h>
62 #include <linux/blkdev.h>
63 #include <linux/blkpg.h>
64 #include <linux/init.h>
65 #include <linux/smp_lock.h>
66 #include <linux/swap.h>
67 #include <linux/slab.h>
68 #include <linux/loop.h>
69 #include <linux/compat.h>
70 #include <linux/suspend.h>
71 #include <linux/writeback.h>
72 #include <linux/buffer_head.h> /* for invalidate_bdev() */
73 #include <linux/completion.h>
74 #include <linux/highmem.h>
75 #include <linux/gfp.h>
76 #include <linux/kthread.h>
77 #include <linux/splice.h>
79 #include <asm/uaccess.h>
81 static LIST_HEAD(loop_devices);
82 static DEFINE_MUTEX(loop_devices_mutex);
87 static int transfer_none(struct loop_device *lo, int cmd,
88 struct page *raw_page, unsigned raw_off,
89 struct page *loop_page, unsigned loop_off,
90 int size, sector_t real_block)
92 char *raw_buf = kmap_atomic(raw_page, KM_USER0) + raw_off;
93 char *loop_buf = kmap_atomic(loop_page, KM_USER1) + loop_off;
96 memcpy(loop_buf, raw_buf, size);
98 memcpy(raw_buf, loop_buf, size);
100 kunmap_atomic(raw_buf, KM_USER0);
101 kunmap_atomic(loop_buf, KM_USER1);
106 static int transfer_xor(struct loop_device *lo, int cmd,
107 struct page *raw_page, unsigned raw_off,
108 struct page *loop_page, unsigned loop_off,
109 int size, sector_t real_block)
111 char *raw_buf = kmap_atomic(raw_page, KM_USER0) + raw_off;
112 char *loop_buf = kmap_atomic(loop_page, KM_USER1) + loop_off;
113 char *in, *out, *key;
124 key = lo->lo_encrypt_key;
125 keysize = lo->lo_encrypt_key_size;
126 for (i = 0; i < size; i++)
127 *out++ = *in++ ^ key[(i & 511) % keysize];
129 kunmap_atomic(raw_buf, KM_USER0);
130 kunmap_atomic(loop_buf, KM_USER1);
135 static int xor_init(struct loop_device *lo, const struct loop_info64 *info)
137 if (unlikely(info->lo_encrypt_key_size <= 0))
142 static struct loop_func_table none_funcs = {
143 .number = LO_CRYPT_NONE,
144 .transfer = transfer_none,
147 static struct loop_func_table xor_funcs = {
148 .number = LO_CRYPT_XOR,
149 .transfer = transfer_xor,
153 /* xfer_funcs[0] is special - its release function is never called */
154 static struct loop_func_table *xfer_funcs[MAX_LO_CRYPT] = {
159 static loff_t get_loop_size(struct loop_device *lo, struct file *file)
161 loff_t size, offset, loopsize;
163 /* Compute loopsize in bytes */
164 size = i_size_read(file->f_mapping->host);
165 offset = lo->lo_offset;
166 loopsize = size - offset;
167 if (lo->lo_sizelimit > 0 && lo->lo_sizelimit < loopsize)
168 loopsize = lo->lo_sizelimit;
171 * Unfortunately, if we want to do I/O on the device,
172 * the number of 512-byte sectors has to fit into a sector_t.
174 return loopsize >> 9;
178 figure_loop_size(struct loop_device *lo)
180 loff_t size = get_loop_size(lo, lo->lo_backing_file);
181 sector_t x = (sector_t)size;
183 if (unlikely((loff_t)x != size))
186 set_capacity(lo->lo_disk, x);
191 lo_do_transfer(struct loop_device *lo, int cmd,
192 struct page *rpage, unsigned roffs,
193 struct page *lpage, unsigned loffs,
194 int size, sector_t rblock)
196 if (unlikely(!lo->transfer))
199 return lo->transfer(lo, cmd, rpage, roffs, lpage, loffs, size, rblock);
203 * do_lo_send_aops - helper for writing data to a loop device
205 * This is the fast version for backing filesystems which implement the address
206 * space operations prepare_write and commit_write.
208 static int do_lo_send_aops(struct loop_device *lo, struct bio_vec *bvec,
209 int bsize, loff_t pos, struct page *page)
211 struct file *file = lo->lo_backing_file; /* kudos to NFsckingS */
212 struct address_space *mapping = file->f_mapping;
213 const struct address_space_operations *aops = mapping->a_ops;
215 unsigned offset, bv_offs;
218 mutex_lock(&mapping->host->i_mutex);
219 index = pos >> PAGE_CACHE_SHIFT;
220 offset = pos & ((pgoff_t)PAGE_CACHE_SIZE - 1);
221 bv_offs = bvec->bv_offset;
228 IV = ((sector_t)index << (PAGE_CACHE_SHIFT - 9))+(offset >> 9);
229 size = PAGE_CACHE_SIZE - offset;
232 page = grab_cache_page(mapping, index);
235 ret = aops->prepare_write(file, page, offset,
238 if (ret == AOP_TRUNCATED_PAGE) {
239 page_cache_release(page);
244 transfer_result = lo_do_transfer(lo, WRITE, page, offset,
245 bvec->bv_page, bv_offs, size, IV);
246 if (unlikely(transfer_result)) {
248 * The transfer failed, but we still write the data to
249 * keep prepare/commit calls balanced.
251 printk(KERN_ERR "loop: transfer error block %llu\n",
252 (unsigned long long)index);
253 zero_user_page(page, offset, size, KM_USER0);
255 flush_dcache_page(page);
256 ret = aops->commit_write(file, page, offset,
259 if (ret == AOP_TRUNCATED_PAGE) {
260 page_cache_release(page);
265 if (unlikely(transfer_result))
273 page_cache_release(page);
277 mutex_unlock(&mapping->host->i_mutex);
281 page_cache_release(page);
288 * __do_lo_send_write - helper for writing data to a loop device
290 * This helper just factors out common code between do_lo_send_direct_write()
291 * and do_lo_send_write().
293 static int __do_lo_send_write(struct file *file,
294 u8 *buf, const int len, loff_t pos)
297 mm_segment_t old_fs = get_fs();
300 bw = file->f_op->write(file, buf, len, &pos);
302 if (likely(bw == len))
304 printk(KERN_ERR "loop: Write error at byte offset %llu, length %i.\n",
305 (unsigned long long)pos, len);
312 * do_lo_send_direct_write - helper for writing data to a loop device
314 * This is the fast, non-transforming version for backing filesystems which do
315 * not implement the address space operations prepare_write and commit_write.
316 * It uses the write file operation which should be present on all writeable
319 static int do_lo_send_direct_write(struct loop_device *lo,
320 struct bio_vec *bvec, int bsize, loff_t pos, struct page *page)
322 ssize_t bw = __do_lo_send_write(lo->lo_backing_file,
323 kmap(bvec->bv_page) + bvec->bv_offset,
325 kunmap(bvec->bv_page);
331 * do_lo_send_write - helper for writing data to a loop device
333 * This is the slow, transforming version for filesystems which do not
334 * implement the address space operations prepare_write and commit_write. It
335 * uses the write file operation which should be present on all writeable
338 * Using fops->write is slower than using aops->{prepare,commit}_write in the
339 * transforming case because we need to double buffer the data as we cannot do
340 * the transformations in place as we do not have direct access to the
341 * destination pages of the backing file.
343 static int do_lo_send_write(struct loop_device *lo, struct bio_vec *bvec,
344 int bsize, loff_t pos, struct page *page)
346 int ret = lo_do_transfer(lo, WRITE, page, 0, bvec->bv_page,
347 bvec->bv_offset, bvec->bv_len, pos >> 9);
349 return __do_lo_send_write(lo->lo_backing_file,
350 page_address(page), bvec->bv_len,
352 printk(KERN_ERR "loop: Transfer error at byte offset %llu, "
353 "length %i.\n", (unsigned long long)pos, bvec->bv_len);
359 static int lo_send(struct loop_device *lo, struct bio *bio, int bsize,
362 int (*do_lo_send)(struct loop_device *, struct bio_vec *, int, loff_t,
364 struct bio_vec *bvec;
365 struct page *page = NULL;
368 do_lo_send = do_lo_send_aops;
369 if (!(lo->lo_flags & LO_FLAGS_USE_AOPS)) {
370 do_lo_send = do_lo_send_direct_write;
371 if (lo->transfer != transfer_none) {
372 page = alloc_page(GFP_NOIO | __GFP_HIGHMEM);
376 do_lo_send = do_lo_send_write;
379 bio_for_each_segment(bvec, bio, i) {
380 ret = do_lo_send(lo, bvec, bsize, pos, page);
392 printk(KERN_ERR "loop: Failed to allocate temporary page for write.\n");
397 struct lo_read_data {
398 struct loop_device *lo;
405 lo_splice_actor(struct pipe_inode_info *pipe, struct pipe_buffer *buf,
406 struct splice_desc *sd)
408 struct lo_read_data *p = sd->u.data;
409 struct loop_device *lo = p->lo;
410 struct page *page = buf->page;
415 ret = buf->ops->confirm(pipe, buf);
419 IV = ((sector_t) page->index << (PAGE_CACHE_SHIFT - 9)) +
425 if (lo_do_transfer(lo, READ, page, buf->offset, p->page, p->offset, size, IV)) {
426 printk(KERN_ERR "loop: transfer error block %ld\n",
431 flush_dcache_page(p->page);
440 lo_direct_splice_actor(struct pipe_inode_info *pipe, struct splice_desc *sd)
442 return __splice_from_pipe(pipe, sd, lo_splice_actor);
446 do_lo_receive(struct loop_device *lo,
447 struct bio_vec *bvec, int bsize, loff_t pos)
449 struct lo_read_data cookie;
450 struct splice_desc sd;
455 cookie.page = bvec->bv_page;
456 cookie.offset = bvec->bv_offset;
457 cookie.bsize = bsize;
460 sd.total_len = bvec->bv_len;
465 file = lo->lo_backing_file;
466 retval = splice_direct_to_actor(file, &sd, lo_direct_splice_actor);
475 lo_receive(struct loop_device *lo, struct bio *bio, int bsize, loff_t pos)
477 struct bio_vec *bvec;
480 bio_for_each_segment(bvec, bio, i) {
481 ret = do_lo_receive(lo, bvec, bsize, pos);
489 static int do_bio_filebacked(struct loop_device *lo, struct bio *bio)
494 pos = ((loff_t) bio->bi_sector << 9) + lo->lo_offset;
495 if (bio_rw(bio) == WRITE)
496 ret = lo_send(lo, bio, lo->lo_blocksize, pos);
498 ret = lo_receive(lo, bio, lo->lo_blocksize, pos);
503 * Add bio to back of pending list
505 static void loop_add_bio(struct loop_device *lo, struct bio *bio)
507 if (lo->lo_biotail) {
508 lo->lo_biotail->bi_next = bio;
509 lo->lo_biotail = bio;
511 lo->lo_bio = lo->lo_biotail = bio;
515 * Grab first pending buffer
517 static struct bio *loop_get_bio(struct loop_device *lo)
521 if ((bio = lo->lo_bio)) {
522 if (bio == lo->lo_biotail)
523 lo->lo_biotail = NULL;
524 lo->lo_bio = bio->bi_next;
531 static int loop_make_request(request_queue_t *q, struct bio *old_bio)
533 struct loop_device *lo = q->queuedata;
534 int rw = bio_rw(old_bio);
539 BUG_ON(!lo || (rw != READ && rw != WRITE));
541 spin_lock_irq(&lo->lo_lock);
542 if (lo->lo_state != Lo_bound)
544 if (unlikely(rw == WRITE && (lo->lo_flags & LO_FLAGS_READ_ONLY)))
546 loop_add_bio(lo, old_bio);
547 wake_up(&lo->lo_event);
548 spin_unlock_irq(&lo->lo_lock);
552 spin_unlock_irq(&lo->lo_lock);
553 bio_io_error(old_bio, old_bio->bi_size);
558 * kick off io on the underlying address space
560 static void loop_unplug(request_queue_t *q)
562 struct loop_device *lo = q->queuedata;
564 clear_bit(QUEUE_FLAG_PLUGGED, &q->queue_flags);
565 blk_run_address_space(lo->lo_backing_file->f_mapping);
568 struct switch_request {
570 struct completion wait;
573 static void do_loop_switch(struct loop_device *, struct switch_request *);
575 static inline void loop_handle_bio(struct loop_device *lo, struct bio *bio)
577 if (unlikely(!bio->bi_bdev)) {
578 do_loop_switch(lo, bio->bi_private);
581 int ret = do_bio_filebacked(lo, bio);
582 bio_endio(bio, bio->bi_size, ret);
587 * worker thread that handles reads/writes to file backed loop devices,
588 * to avoid blocking in our make_request_fn. it also does loop decrypting
589 * on reads for block backed loop, as that is too heavy to do from
590 * b_end_io context where irqs may be disabled.
592 * Loop explanation: loop_clr_fd() sets lo_state to Lo_rundown before
593 * calling kthread_stop(). Therefore once kthread_should_stop() is
594 * true, make_request will not place any more requests. Therefore
595 * once kthread_should_stop() is true and lo_bio is NULL, we are
596 * done with the loop.
598 static int loop_thread(void *data)
600 struct loop_device *lo = data;
604 * loop can be used in an encrypted device,
605 * hence, it mustn't be stopped at all
606 * because it could be indirectly used during suspension
608 current->flags |= PF_NOFREEZE;
610 set_user_nice(current, -20);
612 while (!kthread_should_stop() || lo->lo_bio) {
614 wait_event_interruptible(lo->lo_event,
615 lo->lo_bio || kthread_should_stop());
619 spin_lock_irq(&lo->lo_lock);
620 bio = loop_get_bio(lo);
621 spin_unlock_irq(&lo->lo_lock);
624 loop_handle_bio(lo, bio);
631 * loop_switch performs the hard work of switching a backing store.
632 * First it needs to flush existing IO, it does this by sending a magic
633 * BIO down the pipe. The completion of this BIO does the actual switch.
635 static int loop_switch(struct loop_device *lo, struct file *file)
637 struct switch_request w;
638 struct bio *bio = bio_alloc(GFP_KERNEL, 1);
641 init_completion(&w.wait);
643 bio->bi_private = &w;
645 loop_make_request(lo->lo_queue, bio);
646 wait_for_completion(&w.wait);
651 * Do the actual switch; called from the BIO completion routine
653 static void do_loop_switch(struct loop_device *lo, struct switch_request *p)
655 struct file *file = p->file;
656 struct file *old_file = lo->lo_backing_file;
657 struct address_space *mapping = file->f_mapping;
659 mapping_set_gfp_mask(old_file->f_mapping, lo->old_gfp_mask);
660 lo->lo_backing_file = file;
661 lo->lo_blocksize = S_ISBLK(mapping->host->i_mode) ?
662 mapping->host->i_bdev->bd_block_size : PAGE_SIZE;
663 lo->old_gfp_mask = mapping_gfp_mask(mapping);
664 mapping_set_gfp_mask(mapping, lo->old_gfp_mask & ~(__GFP_IO|__GFP_FS));
670 * loop_change_fd switched the backing store of a loopback device to
671 * a new file. This is useful for operating system installers to free up
672 * the original file and in High Availability environments to switch to
673 * an alternative location for the content in case of server meltdown.
674 * This can only work if the loop device is used read-only, and if the
675 * new backing store is the same size and type as the old backing store.
677 static int loop_change_fd(struct loop_device *lo, struct file *lo_file,
678 struct block_device *bdev, unsigned int arg)
680 struct file *file, *old_file;
685 if (lo->lo_state != Lo_bound)
688 /* the loop device has to be read-only */
690 if (!(lo->lo_flags & LO_FLAGS_READ_ONLY))
698 inode = file->f_mapping->host;
699 old_file = lo->lo_backing_file;
703 if (!S_ISREG(inode->i_mode) && !S_ISBLK(inode->i_mode))
706 /* new backing store needs to support loop (eg splice_read) */
707 if (!inode->i_fop->splice_read)
710 /* size of the new backing store needs to be the same */
711 if (get_loop_size(lo, file) != get_loop_size(lo, old_file))
715 error = loop_switch(lo, file);
728 static inline int is_loop_device(struct file *file)
730 struct inode *i = file->f_mapping->host;
732 return i && S_ISBLK(i->i_mode) && MAJOR(i->i_rdev) == LOOP_MAJOR;
735 static int loop_set_fd(struct loop_device *lo, struct file *lo_file,
736 struct block_device *bdev, unsigned int arg)
738 struct file *file, *f;
740 struct address_space *mapping;
741 unsigned lo_blocksize;
746 /* This is safe, since we have a reference from open(). */
747 __module_get(THIS_MODULE);
755 if (lo->lo_state != Lo_unbound)
758 /* Avoid recursion */
760 while (is_loop_device(f)) {
761 struct loop_device *l;
763 if (f->f_mapping->host->i_rdev == lo_file->f_mapping->host->i_rdev)
766 l = f->f_mapping->host->i_bdev->bd_disk->private_data;
767 if (l->lo_state == Lo_unbound) {
771 f = l->lo_backing_file;
774 mapping = file->f_mapping;
775 inode = mapping->host;
777 if (!(file->f_mode & FMODE_WRITE))
778 lo_flags |= LO_FLAGS_READ_ONLY;
781 if (S_ISREG(inode->i_mode) || S_ISBLK(inode->i_mode)) {
782 const struct address_space_operations *aops = mapping->a_ops;
784 * If we can't read - sorry. If we only can't write - well,
785 * it's going to be read-only.
787 if (!file->f_op->splice_read)
789 if (aops->prepare_write && aops->commit_write)
790 lo_flags |= LO_FLAGS_USE_AOPS;
791 if (!(lo_flags & LO_FLAGS_USE_AOPS) && !file->f_op->write)
792 lo_flags |= LO_FLAGS_READ_ONLY;
794 lo_blocksize = S_ISBLK(inode->i_mode) ?
795 inode->i_bdev->bd_block_size : PAGE_SIZE;
802 size = get_loop_size(lo, file);
804 if ((loff_t)(sector_t)size != size) {
809 if (!(lo_file->f_mode & FMODE_WRITE))
810 lo_flags |= LO_FLAGS_READ_ONLY;
812 set_device_ro(bdev, (lo_flags & LO_FLAGS_READ_ONLY) != 0);
814 lo->lo_blocksize = lo_blocksize;
815 lo->lo_device = bdev;
816 lo->lo_flags = lo_flags;
817 lo->lo_backing_file = file;
818 lo->transfer = transfer_none;
820 lo->lo_sizelimit = 0;
821 lo->old_gfp_mask = mapping_gfp_mask(mapping);
822 mapping_set_gfp_mask(mapping, lo->old_gfp_mask & ~(__GFP_IO|__GFP_FS));
824 lo->lo_bio = lo->lo_biotail = NULL;
827 * set queue make_request_fn, and add limits based on lower level
830 blk_queue_make_request(lo->lo_queue, loop_make_request);
831 lo->lo_queue->queuedata = lo;
832 lo->lo_queue->unplug_fn = loop_unplug;
834 set_capacity(lo->lo_disk, size);
835 bd_set_size(bdev, size << 9);
837 set_blocksize(bdev, lo_blocksize);
839 lo->lo_thread = kthread_create(loop_thread, lo, "loop%d",
841 if (IS_ERR(lo->lo_thread)) {
842 error = PTR_ERR(lo->lo_thread);
845 lo->lo_state = Lo_bound;
846 wake_up_process(lo->lo_thread);
850 lo->lo_thread = NULL;
851 lo->lo_device = NULL;
852 lo->lo_backing_file = NULL;
854 set_capacity(lo->lo_disk, 0);
855 invalidate_bdev(bdev);
856 bd_set_size(bdev, 0);
857 mapping_set_gfp_mask(mapping, lo->old_gfp_mask);
858 lo->lo_state = Lo_unbound;
862 /* This is safe: open() is still holding a reference. */
863 module_put(THIS_MODULE);
868 loop_release_xfer(struct loop_device *lo)
871 struct loop_func_table *xfer = lo->lo_encryption;
875 err = xfer->release(lo);
877 lo->lo_encryption = NULL;
878 module_put(xfer->owner);
884 loop_init_xfer(struct loop_device *lo, struct loop_func_table *xfer,
885 const struct loop_info64 *i)
890 struct module *owner = xfer->owner;
892 if (!try_module_get(owner))
895 err = xfer->init(lo, i);
899 lo->lo_encryption = xfer;
904 static int loop_clr_fd(struct loop_device *lo, struct block_device *bdev)
906 struct file *filp = lo->lo_backing_file;
907 gfp_t gfp = lo->old_gfp_mask;
909 if (lo->lo_state != Lo_bound)
912 if (lo->lo_refcnt > 1) /* we needed one fd for the ioctl */
918 spin_lock_irq(&lo->lo_lock);
919 lo->lo_state = Lo_rundown;
920 spin_unlock_irq(&lo->lo_lock);
922 kthread_stop(lo->lo_thread);
924 lo->lo_backing_file = NULL;
926 loop_release_xfer(lo);
929 lo->lo_device = NULL;
930 lo->lo_encryption = NULL;
932 lo->lo_sizelimit = 0;
933 lo->lo_encrypt_key_size = 0;
935 lo->lo_thread = NULL;
936 memset(lo->lo_encrypt_key, 0, LO_KEY_SIZE);
937 memset(lo->lo_crypt_name, 0, LO_NAME_SIZE);
938 memset(lo->lo_file_name, 0, LO_NAME_SIZE);
939 invalidate_bdev(bdev);
940 set_capacity(lo->lo_disk, 0);
941 bd_set_size(bdev, 0);
942 mapping_set_gfp_mask(filp->f_mapping, gfp);
943 lo->lo_state = Lo_unbound;
945 /* This is safe: open() is still holding a reference. */
946 module_put(THIS_MODULE);
951 loop_set_status(struct loop_device *lo, const struct loop_info64 *info)
954 struct loop_func_table *xfer;
956 if (lo->lo_encrypt_key_size && lo->lo_key_owner != current->uid &&
957 !capable(CAP_SYS_ADMIN))
959 if (lo->lo_state != Lo_bound)
961 if ((unsigned int) info->lo_encrypt_key_size > LO_KEY_SIZE)
964 err = loop_release_xfer(lo);
968 if (info->lo_encrypt_type) {
969 unsigned int type = info->lo_encrypt_type;
971 if (type >= MAX_LO_CRYPT)
973 xfer = xfer_funcs[type];
979 err = loop_init_xfer(lo, xfer, info);
983 if (lo->lo_offset != info->lo_offset ||
984 lo->lo_sizelimit != info->lo_sizelimit) {
985 lo->lo_offset = info->lo_offset;
986 lo->lo_sizelimit = info->lo_sizelimit;
987 if (figure_loop_size(lo))
991 memcpy(lo->lo_file_name, info->lo_file_name, LO_NAME_SIZE);
992 memcpy(lo->lo_crypt_name, info->lo_crypt_name, LO_NAME_SIZE);
993 lo->lo_file_name[LO_NAME_SIZE-1] = 0;
994 lo->lo_crypt_name[LO_NAME_SIZE-1] = 0;
998 lo->transfer = xfer->transfer;
999 lo->ioctl = xfer->ioctl;
1001 lo->lo_encrypt_key_size = info->lo_encrypt_key_size;
1002 lo->lo_init[0] = info->lo_init[0];
1003 lo->lo_init[1] = info->lo_init[1];
1004 if (info->lo_encrypt_key_size) {
1005 memcpy(lo->lo_encrypt_key, info->lo_encrypt_key,
1006 info->lo_encrypt_key_size);
1007 lo->lo_key_owner = current->uid;
1014 loop_get_status(struct loop_device *lo, struct loop_info64 *info)
1016 struct file *file = lo->lo_backing_file;
1020 if (lo->lo_state != Lo_bound)
1022 error = vfs_getattr(file->f_path.mnt, file->f_path.dentry, &stat);
1025 memset(info, 0, sizeof(*info));
1026 info->lo_number = lo->lo_number;
1027 info->lo_device = huge_encode_dev(stat.dev);
1028 info->lo_inode = stat.ino;
1029 info->lo_rdevice = huge_encode_dev(lo->lo_device ? stat.rdev : stat.dev);
1030 info->lo_offset = lo->lo_offset;
1031 info->lo_sizelimit = lo->lo_sizelimit;
1032 info->lo_flags = lo->lo_flags;
1033 memcpy(info->lo_file_name, lo->lo_file_name, LO_NAME_SIZE);
1034 memcpy(info->lo_crypt_name, lo->lo_crypt_name, LO_NAME_SIZE);
1035 info->lo_encrypt_type =
1036 lo->lo_encryption ? lo->lo_encryption->number : 0;
1037 if (lo->lo_encrypt_key_size && capable(CAP_SYS_ADMIN)) {
1038 info->lo_encrypt_key_size = lo->lo_encrypt_key_size;
1039 memcpy(info->lo_encrypt_key, lo->lo_encrypt_key,
1040 lo->lo_encrypt_key_size);
1046 loop_info64_from_old(const struct loop_info *info, struct loop_info64 *info64)
1048 memset(info64, 0, sizeof(*info64));
1049 info64->lo_number = info->lo_number;
1050 info64->lo_device = info->lo_device;
1051 info64->lo_inode = info->lo_inode;
1052 info64->lo_rdevice = info->lo_rdevice;
1053 info64->lo_offset = info->lo_offset;
1054 info64->lo_sizelimit = 0;
1055 info64->lo_encrypt_type = info->lo_encrypt_type;
1056 info64->lo_encrypt_key_size = info->lo_encrypt_key_size;
1057 info64->lo_flags = info->lo_flags;
1058 info64->lo_init[0] = info->lo_init[0];
1059 info64->lo_init[1] = info->lo_init[1];
1060 if (info->lo_encrypt_type == LO_CRYPT_CRYPTOAPI)
1061 memcpy(info64->lo_crypt_name, info->lo_name, LO_NAME_SIZE);
1063 memcpy(info64->lo_file_name, info->lo_name, LO_NAME_SIZE);
1064 memcpy(info64->lo_encrypt_key, info->lo_encrypt_key, LO_KEY_SIZE);
1068 loop_info64_to_old(const struct loop_info64 *info64, struct loop_info *info)
1070 memset(info, 0, sizeof(*info));
1071 info->lo_number = info64->lo_number;
1072 info->lo_device = info64->lo_device;
1073 info->lo_inode = info64->lo_inode;
1074 info->lo_rdevice = info64->lo_rdevice;
1075 info->lo_offset = info64->lo_offset;
1076 info->lo_encrypt_type = info64->lo_encrypt_type;
1077 info->lo_encrypt_key_size = info64->lo_encrypt_key_size;
1078 info->lo_flags = info64->lo_flags;
1079 info->lo_init[0] = info64->lo_init[0];
1080 info->lo_init[1] = info64->lo_init[1];
1081 if (info->lo_encrypt_type == LO_CRYPT_CRYPTOAPI)
1082 memcpy(info->lo_name, info64->lo_crypt_name, LO_NAME_SIZE);
1084 memcpy(info->lo_name, info64->lo_file_name, LO_NAME_SIZE);
1085 memcpy(info->lo_encrypt_key, info64->lo_encrypt_key, LO_KEY_SIZE);
1087 /* error in case values were truncated */
1088 if (info->lo_device != info64->lo_device ||
1089 info->lo_rdevice != info64->lo_rdevice ||
1090 info->lo_inode != info64->lo_inode ||
1091 info->lo_offset != info64->lo_offset)
1098 loop_set_status_old(struct loop_device *lo, const struct loop_info __user *arg)
1100 struct loop_info info;
1101 struct loop_info64 info64;
1103 if (copy_from_user(&info, arg, sizeof (struct loop_info)))
1105 loop_info64_from_old(&info, &info64);
1106 return loop_set_status(lo, &info64);
1110 loop_set_status64(struct loop_device *lo, const struct loop_info64 __user *arg)
1112 struct loop_info64 info64;
1114 if (copy_from_user(&info64, arg, sizeof (struct loop_info64)))
1116 return loop_set_status(lo, &info64);
1120 loop_get_status_old(struct loop_device *lo, struct loop_info __user *arg) {
1121 struct loop_info info;
1122 struct loop_info64 info64;
1128 err = loop_get_status(lo, &info64);
1130 err = loop_info64_to_old(&info64, &info);
1131 if (!err && copy_to_user(arg, &info, sizeof(info)))
1138 loop_get_status64(struct loop_device *lo, struct loop_info64 __user *arg) {
1139 struct loop_info64 info64;
1145 err = loop_get_status(lo, &info64);
1146 if (!err && copy_to_user(arg, &info64, sizeof(info64)))
1152 static int lo_ioctl(struct inode * inode, struct file * file,
1153 unsigned int cmd, unsigned long arg)
1155 struct loop_device *lo = inode->i_bdev->bd_disk->private_data;
1158 mutex_lock(&lo->lo_ctl_mutex);
1161 err = loop_set_fd(lo, file, inode->i_bdev, arg);
1163 case LOOP_CHANGE_FD:
1164 err = loop_change_fd(lo, file, inode->i_bdev, arg);
1167 err = loop_clr_fd(lo, inode->i_bdev);
1169 case LOOP_SET_STATUS:
1170 err = loop_set_status_old(lo, (struct loop_info __user *) arg);
1172 case LOOP_GET_STATUS:
1173 err = loop_get_status_old(lo, (struct loop_info __user *) arg);
1175 case LOOP_SET_STATUS64:
1176 err = loop_set_status64(lo, (struct loop_info64 __user *) arg);
1178 case LOOP_GET_STATUS64:
1179 err = loop_get_status64(lo, (struct loop_info64 __user *) arg);
1182 err = lo->ioctl ? lo->ioctl(lo, cmd, arg) : -EINVAL;
1184 mutex_unlock(&lo->lo_ctl_mutex);
1188 #ifdef CONFIG_COMPAT
1189 struct compat_loop_info {
1190 compat_int_t lo_number; /* ioctl r/o */
1191 compat_dev_t lo_device; /* ioctl r/o */
1192 compat_ulong_t lo_inode; /* ioctl r/o */
1193 compat_dev_t lo_rdevice; /* ioctl r/o */
1194 compat_int_t lo_offset;
1195 compat_int_t lo_encrypt_type;
1196 compat_int_t lo_encrypt_key_size; /* ioctl w/o */
1197 compat_int_t lo_flags; /* ioctl r/o */
1198 char lo_name[LO_NAME_SIZE];
1199 unsigned char lo_encrypt_key[LO_KEY_SIZE]; /* ioctl w/o */
1200 compat_ulong_t lo_init[2];
1205 * Transfer 32-bit compatibility structure in userspace to 64-bit loop info
1206 * - noinlined to reduce stack space usage in main part of driver
1209 loop_info64_from_compat(const struct compat_loop_info __user *arg,
1210 struct loop_info64 *info64)
1212 struct compat_loop_info info;
1214 if (copy_from_user(&info, arg, sizeof(info)))
1217 memset(info64, 0, sizeof(*info64));
1218 info64->lo_number = info.lo_number;
1219 info64->lo_device = info.lo_device;
1220 info64->lo_inode = info.lo_inode;
1221 info64->lo_rdevice = info.lo_rdevice;
1222 info64->lo_offset = info.lo_offset;
1223 info64->lo_sizelimit = 0;
1224 info64->lo_encrypt_type = info.lo_encrypt_type;
1225 info64->lo_encrypt_key_size = info.lo_encrypt_key_size;
1226 info64->lo_flags = info.lo_flags;
1227 info64->lo_init[0] = info.lo_init[0];
1228 info64->lo_init[1] = info.lo_init[1];
1229 if (info.lo_encrypt_type == LO_CRYPT_CRYPTOAPI)
1230 memcpy(info64->lo_crypt_name, info.lo_name, LO_NAME_SIZE);
1232 memcpy(info64->lo_file_name, info.lo_name, LO_NAME_SIZE);
1233 memcpy(info64->lo_encrypt_key, info.lo_encrypt_key, LO_KEY_SIZE);
1238 * Transfer 64-bit loop info to 32-bit compatibility structure in userspace
1239 * - noinlined to reduce stack space usage in main part of driver
1242 loop_info64_to_compat(const struct loop_info64 *info64,
1243 struct compat_loop_info __user *arg)
1245 struct compat_loop_info info;
1247 memset(&info, 0, sizeof(info));
1248 info.lo_number = info64->lo_number;
1249 info.lo_device = info64->lo_device;
1250 info.lo_inode = info64->lo_inode;
1251 info.lo_rdevice = info64->lo_rdevice;
1252 info.lo_offset = info64->lo_offset;
1253 info.lo_encrypt_type = info64->lo_encrypt_type;
1254 info.lo_encrypt_key_size = info64->lo_encrypt_key_size;
1255 info.lo_flags = info64->lo_flags;
1256 info.lo_init[0] = info64->lo_init[0];
1257 info.lo_init[1] = info64->lo_init[1];
1258 if (info.lo_encrypt_type == LO_CRYPT_CRYPTOAPI)
1259 memcpy(info.lo_name, info64->lo_crypt_name, LO_NAME_SIZE);
1261 memcpy(info.lo_name, info64->lo_file_name, LO_NAME_SIZE);
1262 memcpy(info.lo_encrypt_key, info64->lo_encrypt_key, LO_KEY_SIZE);
1264 /* error in case values were truncated */
1265 if (info.lo_device != info64->lo_device ||
1266 info.lo_rdevice != info64->lo_rdevice ||
1267 info.lo_inode != info64->lo_inode ||
1268 info.lo_offset != info64->lo_offset ||
1269 info.lo_init[0] != info64->lo_init[0] ||
1270 info.lo_init[1] != info64->lo_init[1])
1273 if (copy_to_user(arg, &info, sizeof(info)))
1279 loop_set_status_compat(struct loop_device *lo,
1280 const struct compat_loop_info __user *arg)
1282 struct loop_info64 info64;
1285 ret = loop_info64_from_compat(arg, &info64);
1288 return loop_set_status(lo, &info64);
1292 loop_get_status_compat(struct loop_device *lo,
1293 struct compat_loop_info __user *arg)
1295 struct loop_info64 info64;
1301 err = loop_get_status(lo, &info64);
1303 err = loop_info64_to_compat(&info64, arg);
1307 static long lo_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
1309 struct inode *inode = file->f_path.dentry->d_inode;
1310 struct loop_device *lo = inode->i_bdev->bd_disk->private_data;
1315 case LOOP_SET_STATUS:
1316 mutex_lock(&lo->lo_ctl_mutex);
1317 err = loop_set_status_compat(
1318 lo, (const struct compat_loop_info __user *) arg);
1319 mutex_unlock(&lo->lo_ctl_mutex);
1321 case LOOP_GET_STATUS:
1322 mutex_lock(&lo->lo_ctl_mutex);
1323 err = loop_get_status_compat(
1324 lo, (struct compat_loop_info __user *) arg);
1325 mutex_unlock(&lo->lo_ctl_mutex);
1328 case LOOP_GET_STATUS64:
1329 case LOOP_SET_STATUS64:
1330 arg = (unsigned long) compat_ptr(arg);
1332 case LOOP_CHANGE_FD:
1333 err = lo_ioctl(inode, file, cmd, arg);
1344 static int lo_open(struct inode *inode, struct file *file)
1346 struct loop_device *lo = inode->i_bdev->bd_disk->private_data;
1348 mutex_lock(&lo->lo_ctl_mutex);
1350 mutex_unlock(&lo->lo_ctl_mutex);
1355 static int lo_release(struct inode *inode, struct file *file)
1357 struct loop_device *lo = inode->i_bdev->bd_disk->private_data;
1359 mutex_lock(&lo->lo_ctl_mutex);
1361 mutex_unlock(&lo->lo_ctl_mutex);
1366 static struct block_device_operations lo_fops = {
1367 .owner = THIS_MODULE,
1369 .release = lo_release,
1371 #ifdef CONFIG_COMPAT
1372 .compat_ioctl = lo_compat_ioctl,
1377 * And now the modules code and kernel interface.
1379 static int max_loop;
1380 module_param(max_loop, int, 0);
1381 MODULE_PARM_DESC(max_loop, "Maximum number of loop devices");
1382 MODULE_LICENSE("GPL");
1383 MODULE_ALIAS_BLOCKDEV_MAJOR(LOOP_MAJOR);
1385 int loop_register_transfer(struct loop_func_table *funcs)
1387 unsigned int n = funcs->number;
1389 if (n >= MAX_LO_CRYPT || xfer_funcs[n])
1391 xfer_funcs[n] = funcs;
1395 int loop_unregister_transfer(int number)
1397 unsigned int n = number;
1398 struct loop_device *lo;
1399 struct loop_func_table *xfer;
1401 if (n == 0 || n >= MAX_LO_CRYPT || (xfer = xfer_funcs[n]) == NULL)
1404 xfer_funcs[n] = NULL;
1406 list_for_each_entry(lo, &loop_devices, lo_list) {
1407 mutex_lock(&lo->lo_ctl_mutex);
1409 if (lo->lo_encryption == xfer)
1410 loop_release_xfer(lo);
1412 mutex_unlock(&lo->lo_ctl_mutex);
1418 EXPORT_SYMBOL(loop_register_transfer);
1419 EXPORT_SYMBOL(loop_unregister_transfer);
1421 static struct loop_device *loop_alloc(int i)
1423 struct loop_device *lo;
1424 struct gendisk *disk;
1426 lo = kzalloc(sizeof(*lo), GFP_KERNEL);
1430 lo->lo_queue = blk_alloc_queue(GFP_KERNEL);
1434 disk = lo->lo_disk = alloc_disk(1);
1436 goto out_free_queue;
1438 mutex_init(&lo->lo_ctl_mutex);
1440 lo->lo_thread = NULL;
1441 init_waitqueue_head(&lo->lo_event);
1442 spin_lock_init(&lo->lo_lock);
1443 disk->major = LOOP_MAJOR;
1444 disk->first_minor = i;
1445 disk->fops = &lo_fops;
1446 disk->private_data = lo;
1447 disk->queue = lo->lo_queue;
1448 sprintf(disk->disk_name, "loop%d", i);
1452 blk_cleanup_queue(lo->lo_queue);
1459 static void loop_free(struct loop_device *lo)
1461 blk_cleanup_queue(lo->lo_queue);
1462 put_disk(lo->lo_disk);
1463 list_del(&lo->lo_list);
1467 static struct loop_device *loop_init_one(int i)
1469 struct loop_device *lo;
1471 list_for_each_entry(lo, &loop_devices, lo_list) {
1472 if (lo->lo_number == i)
1478 add_disk(lo->lo_disk);
1479 list_add_tail(&lo->lo_list, &loop_devices);
1484 static void loop_del_one(struct loop_device *lo)
1486 del_gendisk(lo->lo_disk);
1490 static struct kobject *loop_probe(dev_t dev, int *part, void *data)
1492 struct loop_device *lo;
1493 struct kobject *kobj;
1495 mutex_lock(&loop_devices_mutex);
1496 lo = loop_init_one(dev & MINORMASK);
1497 kobj = lo ? get_disk(lo->lo_disk) : ERR_PTR(-ENOMEM);
1498 mutex_unlock(&loop_devices_mutex);
1504 static int __init loop_init(void)
1507 unsigned long range;
1508 struct loop_device *lo, *next;
1511 * loop module now has a feature to instantiate underlying device
1512 * structure on-demand, provided that there is an access dev node.
1513 * However, this will not work well with user space tool that doesn't
1514 * know about such "feature". In order to not break any existing
1515 * tool, we do the following:
1517 * (1) if max_loop is specified, create that many upfront, and this
1518 * also becomes a hard limit.
1519 * (2) if max_loop is not specified, create 8 loop device on module
1520 * load, user can further extend loop device by create dev node
1521 * themselves and have kernel automatically instantiate actual
1524 if (max_loop > 1UL << MINORBITS)
1532 range = 1UL << MINORBITS;
1535 if (register_blkdev(LOOP_MAJOR, "loop"))
1538 for (i = 0; i < nr; i++) {
1542 list_add_tail(&lo->lo_list, &loop_devices);
1545 /* point of no return */
1547 list_for_each_entry(lo, &loop_devices, lo_list)
1548 add_disk(lo->lo_disk);
1550 blk_register_region(MKDEV(LOOP_MAJOR, 0), range,
1551 THIS_MODULE, loop_probe, NULL, NULL);
1553 printk(KERN_INFO "loop: module loaded\n");
1557 printk(KERN_INFO "loop: out of memory\n");
1559 list_for_each_entry_safe(lo, next, &loop_devices, lo_list)
1562 unregister_blkdev(LOOP_MAJOR, "loop");
1566 static void __exit loop_exit(void)
1568 unsigned long range;
1569 struct loop_device *lo, *next;
1571 range = max_loop ? max_loop : 1UL << MINORBITS;
1573 list_for_each_entry_safe(lo, next, &loop_devices, lo_list)
1576 blk_unregister_region(MKDEV(LOOP_MAJOR, 0), range);
1577 if (unregister_blkdev(LOOP_MAJOR, "loop"))
1578 printk(KERN_WARNING "loop: cannot unregister blkdev\n");
1581 module_init(loop_init);
1582 module_exit(loop_exit);
1585 static int __init max_loop_setup(char *str)
1587 max_loop = simple_strtol(str, NULL, 0);
1591 __setup("max_loop=", max_loop_setup);