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>
78 #include <asm/uaccess.h>
80 static int max_loop = 8;
81 static struct loop_device *loop_dev;
82 static struct gendisk **disks;
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(disks[lo->lo_number], 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)) {
250 * The transfer failed, but we still write the data to
251 * keep prepare/commit calls balanced.
253 printk(KERN_ERR "loop: transfer error block %llu\n",
254 (unsigned long long)index);
255 kaddr = kmap_atomic(page, KM_USER0);
256 memset(kaddr + offset, 0, size);
257 kunmap_atomic(kaddr, KM_USER0);
259 flush_dcache_page(page);
260 ret = aops->commit_write(file, page, offset,
263 if (ret == AOP_TRUNCATED_PAGE) {
264 page_cache_release(page);
269 if (unlikely(transfer_result))
277 page_cache_release(page);
281 mutex_unlock(&mapping->host->i_mutex);
285 page_cache_release(page);
292 * __do_lo_send_write - helper for writing data to a loop device
294 * This helper just factors out common code between do_lo_send_direct_write()
295 * and do_lo_send_write().
297 static int __do_lo_send_write(struct file *file,
298 u8 __user *buf, const int len, loff_t pos)
301 mm_segment_t old_fs = get_fs();
304 bw = file->f_op->write(file, buf, len, &pos);
306 if (likely(bw == len))
308 printk(KERN_ERR "loop: Write error at byte offset %llu, length %i.\n",
309 (unsigned long long)pos, len);
316 * do_lo_send_direct_write - helper for writing data to a loop device
318 * This is the fast, non-transforming version for backing filesystems which do
319 * not implement the address space operations prepare_write and commit_write.
320 * It uses the write file operation which should be present on all writeable
323 static int do_lo_send_direct_write(struct loop_device *lo,
324 struct bio_vec *bvec, int bsize, loff_t pos, struct page *page)
326 ssize_t bw = __do_lo_send_write(lo->lo_backing_file,
327 (u8 __user *)kmap(bvec->bv_page) + bvec->bv_offset,
329 kunmap(bvec->bv_page);
335 * do_lo_send_write - helper for writing data to a loop device
337 * This is the slow, transforming version for filesystems which do not
338 * implement the address space operations prepare_write and commit_write. It
339 * uses the write file operation which should be present on all writeable
342 * Using fops->write is slower than using aops->{prepare,commit}_write in the
343 * transforming case because we need to double buffer the data as we cannot do
344 * the transformations in place as we do not have direct access to the
345 * destination pages of the backing file.
347 static int do_lo_send_write(struct loop_device *lo, struct bio_vec *bvec,
348 int bsize, loff_t pos, struct page *page)
350 int ret = lo_do_transfer(lo, WRITE, page, 0, bvec->bv_page,
351 bvec->bv_offset, bvec->bv_len, pos >> 9);
353 return __do_lo_send_write(lo->lo_backing_file,
354 (u8 __user *)page_address(page), bvec->bv_len,
356 printk(KERN_ERR "loop: Transfer error at byte offset %llu, "
357 "length %i.\n", (unsigned long long)pos, bvec->bv_len);
363 static int lo_send(struct loop_device *lo, struct bio *bio, int bsize,
366 int (*do_lo_send)(struct loop_device *, struct bio_vec *, int, loff_t,
368 struct bio_vec *bvec;
369 struct page *page = NULL;
372 do_lo_send = do_lo_send_aops;
373 if (!(lo->lo_flags & LO_FLAGS_USE_AOPS)) {
374 do_lo_send = do_lo_send_direct_write;
375 if (lo->transfer != transfer_none) {
376 page = alloc_page(GFP_NOIO | __GFP_HIGHMEM);
380 do_lo_send = do_lo_send_write;
383 bio_for_each_segment(bvec, bio, i) {
384 ret = do_lo_send(lo, bvec, bsize, pos, page);
396 printk(KERN_ERR "loop: Failed to allocate temporary page for write.\n");
401 struct lo_read_data {
402 struct loop_device *lo;
409 lo_read_actor(read_descriptor_t *desc, struct page *page,
410 unsigned long offset, unsigned long size)
412 unsigned long count = desc->count;
413 struct lo_read_data *p = desc->arg.data;
414 struct loop_device *lo = p->lo;
417 IV = ((sector_t) page->index << (PAGE_CACHE_SHIFT - 9))+(offset >> 9);
422 if (lo_do_transfer(lo, READ, page, offset, p->page, p->offset, size, IV)) {
424 printk(KERN_ERR "loop: transfer error block %ld\n",
426 desc->error = -EINVAL;
429 flush_dcache_page(p->page);
431 desc->count = count - size;
432 desc->written += size;
438 do_lo_receive(struct loop_device *lo,
439 struct bio_vec *bvec, int bsize, loff_t pos)
441 struct lo_read_data cookie;
446 cookie.page = bvec->bv_page;
447 cookie.offset = bvec->bv_offset;
448 cookie.bsize = bsize;
449 file = lo->lo_backing_file;
450 retval = file->f_op->sendfile(file, &pos, bvec->bv_len,
451 lo_read_actor, &cookie);
452 return (retval < 0)? retval: 0;
456 lo_receive(struct loop_device *lo, struct bio *bio, int bsize, loff_t pos)
458 struct bio_vec *bvec;
461 bio_for_each_segment(bvec, bio, i) {
462 ret = do_lo_receive(lo, bvec, bsize, pos);
470 static int do_bio_filebacked(struct loop_device *lo, struct bio *bio)
475 pos = ((loff_t) bio->bi_sector << 9) + lo->lo_offset;
476 if (bio_rw(bio) == WRITE)
477 ret = lo_send(lo, bio, lo->lo_blocksize, pos);
479 ret = lo_receive(lo, bio, lo->lo_blocksize, pos);
484 * Add bio to back of pending list
486 static void loop_add_bio(struct loop_device *lo, struct bio *bio)
488 if (lo->lo_biotail) {
489 lo->lo_biotail->bi_next = bio;
490 lo->lo_biotail = bio;
492 lo->lo_bio = lo->lo_biotail = bio;
496 * Grab first pending buffer
498 static struct bio *loop_get_bio(struct loop_device *lo)
502 if ((bio = lo->lo_bio)) {
503 if (bio == lo->lo_biotail)
504 lo->lo_biotail = NULL;
505 lo->lo_bio = bio->bi_next;
512 static int loop_make_request(request_queue_t *q, struct bio *old_bio)
514 struct loop_device *lo = q->queuedata;
515 int rw = bio_rw(old_bio);
520 BUG_ON(!lo || (rw != READ && rw != WRITE));
522 spin_lock_irq(&lo->lo_lock);
523 if (lo->lo_state != Lo_bound)
525 if (unlikely(rw == WRITE && (lo->lo_flags & LO_FLAGS_READ_ONLY)))
527 loop_add_bio(lo, old_bio);
528 wake_up(&lo->lo_event);
529 spin_unlock_irq(&lo->lo_lock);
533 spin_unlock_irq(&lo->lo_lock);
534 bio_io_error(old_bio, old_bio->bi_size);
539 * kick off io on the underlying address space
541 static void loop_unplug(request_queue_t *q)
543 struct loop_device *lo = q->queuedata;
545 clear_bit(QUEUE_FLAG_PLUGGED, &q->queue_flags);
546 blk_run_address_space(lo->lo_backing_file->f_mapping);
549 struct switch_request {
551 struct completion wait;
554 static void do_loop_switch(struct loop_device *, struct switch_request *);
556 static inline void loop_handle_bio(struct loop_device *lo, struct bio *bio)
558 if (unlikely(!bio->bi_bdev)) {
559 do_loop_switch(lo, bio->bi_private);
562 int ret = do_bio_filebacked(lo, bio);
563 bio_endio(bio, bio->bi_size, ret);
568 * worker thread that handles reads/writes to file backed loop devices,
569 * to avoid blocking in our make_request_fn. it also does loop decrypting
570 * on reads for block backed loop, as that is too heavy to do from
571 * b_end_io context where irqs may be disabled.
573 * Loop explanation: loop_clr_fd() sets lo_state to Lo_rundown before
574 * calling kthread_stop(). Therefore once kthread_should_stop() is
575 * true, make_request will not place any more requests. Therefore
576 * once kthread_should_stop() is true and lo_bio is NULL, we are
577 * done with the loop.
579 static int loop_thread(void *data)
581 struct loop_device *lo = data;
585 * loop can be used in an encrypted device,
586 * hence, it mustn't be stopped at all
587 * because it could be indirectly used during suspension
589 current->flags |= PF_NOFREEZE;
591 set_user_nice(current, -20);
593 while (!kthread_should_stop() || lo->lo_bio) {
595 wait_event_interruptible(lo->lo_event,
596 lo->lo_bio || kthread_should_stop());
600 spin_lock_irq(&lo->lo_lock);
601 bio = loop_get_bio(lo);
602 spin_unlock_irq(&lo->lo_lock);
605 loop_handle_bio(lo, bio);
612 * loop_switch performs the hard work of switching a backing store.
613 * First it needs to flush existing IO, it does this by sending a magic
614 * BIO down the pipe. The completion of this BIO does the actual switch.
616 static int loop_switch(struct loop_device *lo, struct file *file)
618 struct switch_request w;
619 struct bio *bio = bio_alloc(GFP_KERNEL, 1);
622 init_completion(&w.wait);
624 bio->bi_private = &w;
626 loop_make_request(lo->lo_queue, bio);
627 wait_for_completion(&w.wait);
632 * Do the actual switch; called from the BIO completion routine
634 static void do_loop_switch(struct loop_device *lo, struct switch_request *p)
636 struct file *file = p->file;
637 struct file *old_file = lo->lo_backing_file;
638 struct address_space *mapping = file->f_mapping;
640 mapping_set_gfp_mask(old_file->f_mapping, lo->old_gfp_mask);
641 lo->lo_backing_file = file;
642 lo->lo_blocksize = S_ISBLK(mapping->host->i_mode) ?
643 mapping->host->i_bdev->bd_block_size : PAGE_SIZE;
644 lo->old_gfp_mask = mapping_gfp_mask(mapping);
645 mapping_set_gfp_mask(mapping, lo->old_gfp_mask & ~(__GFP_IO|__GFP_FS));
651 * loop_change_fd switched the backing store of a loopback device to
652 * a new file. This is useful for operating system installers to free up
653 * the original file and in High Availability environments to switch to
654 * an alternative location for the content in case of server meltdown.
655 * This can only work if the loop device is used read-only, and if the
656 * new backing store is the same size and type as the old backing store.
658 static int loop_change_fd(struct loop_device *lo, struct file *lo_file,
659 struct block_device *bdev, unsigned int arg)
661 struct file *file, *old_file;
666 if (lo->lo_state != Lo_bound)
669 /* the loop device has to be read-only */
671 if (!(lo->lo_flags & LO_FLAGS_READ_ONLY))
679 inode = file->f_mapping->host;
680 old_file = lo->lo_backing_file;
684 if (!S_ISREG(inode->i_mode) && !S_ISBLK(inode->i_mode))
687 /* new backing store needs to support loop (eg sendfile) */
688 if (!inode->i_fop->sendfile)
691 /* size of the new backing store needs to be the same */
692 if (get_loop_size(lo, file) != get_loop_size(lo, old_file))
696 error = loop_switch(lo, file);
709 static inline int is_loop_device(struct file *file)
711 struct inode *i = file->f_mapping->host;
713 return i && S_ISBLK(i->i_mode) && MAJOR(i->i_rdev) == LOOP_MAJOR;
716 static int loop_set_fd(struct loop_device *lo, struct file *lo_file,
717 struct block_device *bdev, unsigned int arg)
719 struct file *file, *f;
721 struct address_space *mapping;
722 unsigned lo_blocksize;
727 /* This is safe, since we have a reference from open(). */
728 __module_get(THIS_MODULE);
736 if (lo->lo_state != Lo_unbound)
739 /* Avoid recursion */
741 while (is_loop_device(f)) {
742 struct loop_device *l;
744 if (f->f_mapping->host->i_rdev == lo_file->f_mapping->host->i_rdev)
747 l = f->f_mapping->host->i_bdev->bd_disk->private_data;
748 if (l->lo_state == Lo_unbound) {
752 f = l->lo_backing_file;
755 mapping = file->f_mapping;
756 inode = mapping->host;
758 if (!(file->f_mode & FMODE_WRITE))
759 lo_flags |= LO_FLAGS_READ_ONLY;
762 if (S_ISREG(inode->i_mode) || S_ISBLK(inode->i_mode)) {
763 const struct address_space_operations *aops = mapping->a_ops;
765 * If we can't read - sorry. If we only can't write - well,
766 * it's going to be read-only.
768 if (!file->f_op->sendfile)
770 if (aops->prepare_write && aops->commit_write)
771 lo_flags |= LO_FLAGS_USE_AOPS;
772 if (!(lo_flags & LO_FLAGS_USE_AOPS) && !file->f_op->write)
773 lo_flags |= LO_FLAGS_READ_ONLY;
775 lo_blocksize = S_ISBLK(inode->i_mode) ?
776 inode->i_bdev->bd_block_size : PAGE_SIZE;
783 size = get_loop_size(lo, file);
785 if ((loff_t)(sector_t)size != size) {
790 if (!(lo_file->f_mode & FMODE_WRITE))
791 lo_flags |= LO_FLAGS_READ_ONLY;
793 set_device_ro(bdev, (lo_flags & LO_FLAGS_READ_ONLY) != 0);
795 lo->lo_blocksize = lo_blocksize;
796 lo->lo_device = bdev;
797 lo->lo_flags = lo_flags;
798 lo->lo_backing_file = file;
799 lo->transfer = transfer_none;
801 lo->lo_sizelimit = 0;
802 lo->old_gfp_mask = mapping_gfp_mask(mapping);
803 mapping_set_gfp_mask(mapping, lo->old_gfp_mask & ~(__GFP_IO|__GFP_FS));
805 lo->lo_bio = lo->lo_biotail = NULL;
808 * set queue make_request_fn, and add limits based on lower level
811 blk_queue_make_request(lo->lo_queue, loop_make_request);
812 lo->lo_queue->queuedata = lo;
813 lo->lo_queue->unplug_fn = loop_unplug;
815 set_capacity(disks[lo->lo_number], size);
816 bd_set_size(bdev, size << 9);
818 set_blocksize(bdev, lo_blocksize);
820 lo->lo_thread = kthread_create(loop_thread, lo, "loop%d",
822 if (IS_ERR(lo->lo_thread)) {
823 error = PTR_ERR(lo->lo_thread);
826 lo->lo_state = Lo_bound;
827 wake_up_process(lo->lo_thread);
831 lo->lo_thread = NULL;
832 lo->lo_device = NULL;
833 lo->lo_backing_file = NULL;
835 set_capacity(disks[lo->lo_number], 0);
836 invalidate_bdev(bdev, 0);
837 bd_set_size(bdev, 0);
838 mapping_set_gfp_mask(mapping, lo->old_gfp_mask);
839 lo->lo_state = Lo_unbound;
843 /* This is safe: open() is still holding a reference. */
844 module_put(THIS_MODULE);
849 loop_release_xfer(struct loop_device *lo)
852 struct loop_func_table *xfer = lo->lo_encryption;
856 err = xfer->release(lo);
858 lo->lo_encryption = NULL;
859 module_put(xfer->owner);
865 loop_init_xfer(struct loop_device *lo, struct loop_func_table *xfer,
866 const struct loop_info64 *i)
871 struct module *owner = xfer->owner;
873 if (!try_module_get(owner))
876 err = xfer->init(lo, i);
880 lo->lo_encryption = xfer;
885 static int loop_clr_fd(struct loop_device *lo, struct block_device *bdev)
887 struct file *filp = lo->lo_backing_file;
888 gfp_t gfp = lo->old_gfp_mask;
890 if (lo->lo_state != Lo_bound)
893 if (lo->lo_refcnt > 1) /* we needed one fd for the ioctl */
899 spin_lock_irq(&lo->lo_lock);
900 lo->lo_state = Lo_rundown;
901 spin_unlock_irq(&lo->lo_lock);
903 kthread_stop(lo->lo_thread);
905 lo->lo_backing_file = NULL;
907 loop_release_xfer(lo);
910 lo->lo_device = NULL;
911 lo->lo_encryption = NULL;
913 lo->lo_sizelimit = 0;
914 lo->lo_encrypt_key_size = 0;
916 lo->lo_thread = NULL;
917 memset(lo->lo_encrypt_key, 0, LO_KEY_SIZE);
918 memset(lo->lo_crypt_name, 0, LO_NAME_SIZE);
919 memset(lo->lo_file_name, 0, LO_NAME_SIZE);
920 invalidate_bdev(bdev, 0);
921 set_capacity(disks[lo->lo_number], 0);
922 bd_set_size(bdev, 0);
923 mapping_set_gfp_mask(filp->f_mapping, gfp);
924 lo->lo_state = Lo_unbound;
926 /* This is safe: open() is still holding a reference. */
927 module_put(THIS_MODULE);
932 loop_set_status(struct loop_device *lo, const struct loop_info64 *info)
935 struct loop_func_table *xfer;
937 if (lo->lo_encrypt_key_size && lo->lo_key_owner != current->uid &&
938 !capable(CAP_SYS_ADMIN))
940 if (lo->lo_state != Lo_bound)
942 if ((unsigned int) info->lo_encrypt_key_size > LO_KEY_SIZE)
945 err = loop_release_xfer(lo);
949 if (info->lo_encrypt_type) {
950 unsigned int type = info->lo_encrypt_type;
952 if (type >= MAX_LO_CRYPT)
954 xfer = xfer_funcs[type];
960 err = loop_init_xfer(lo, xfer, info);
964 if (lo->lo_offset != info->lo_offset ||
965 lo->lo_sizelimit != info->lo_sizelimit) {
966 lo->lo_offset = info->lo_offset;
967 lo->lo_sizelimit = info->lo_sizelimit;
968 if (figure_loop_size(lo))
972 memcpy(lo->lo_file_name, info->lo_file_name, LO_NAME_SIZE);
973 memcpy(lo->lo_crypt_name, info->lo_crypt_name, LO_NAME_SIZE);
974 lo->lo_file_name[LO_NAME_SIZE-1] = 0;
975 lo->lo_crypt_name[LO_NAME_SIZE-1] = 0;
979 lo->transfer = xfer->transfer;
980 lo->ioctl = xfer->ioctl;
982 lo->lo_encrypt_key_size = info->lo_encrypt_key_size;
983 lo->lo_init[0] = info->lo_init[0];
984 lo->lo_init[1] = info->lo_init[1];
985 if (info->lo_encrypt_key_size) {
986 memcpy(lo->lo_encrypt_key, info->lo_encrypt_key,
987 info->lo_encrypt_key_size);
988 lo->lo_key_owner = current->uid;
995 loop_get_status(struct loop_device *lo, struct loop_info64 *info)
997 struct file *file = lo->lo_backing_file;
1001 if (lo->lo_state != Lo_bound)
1003 error = vfs_getattr(file->f_vfsmnt, file->f_dentry, &stat);
1006 memset(info, 0, sizeof(*info));
1007 info->lo_number = lo->lo_number;
1008 info->lo_device = huge_encode_dev(stat.dev);
1009 info->lo_inode = stat.ino;
1010 info->lo_rdevice = huge_encode_dev(lo->lo_device ? stat.rdev : stat.dev);
1011 info->lo_offset = lo->lo_offset;
1012 info->lo_sizelimit = lo->lo_sizelimit;
1013 info->lo_flags = lo->lo_flags;
1014 memcpy(info->lo_file_name, lo->lo_file_name, LO_NAME_SIZE);
1015 memcpy(info->lo_crypt_name, lo->lo_crypt_name, LO_NAME_SIZE);
1016 info->lo_encrypt_type =
1017 lo->lo_encryption ? lo->lo_encryption->number : 0;
1018 if (lo->lo_encrypt_key_size && capable(CAP_SYS_ADMIN)) {
1019 info->lo_encrypt_key_size = lo->lo_encrypt_key_size;
1020 memcpy(info->lo_encrypt_key, lo->lo_encrypt_key,
1021 lo->lo_encrypt_key_size);
1027 loop_info64_from_old(const struct loop_info *info, struct loop_info64 *info64)
1029 memset(info64, 0, sizeof(*info64));
1030 info64->lo_number = info->lo_number;
1031 info64->lo_device = info->lo_device;
1032 info64->lo_inode = info->lo_inode;
1033 info64->lo_rdevice = info->lo_rdevice;
1034 info64->lo_offset = info->lo_offset;
1035 info64->lo_sizelimit = 0;
1036 info64->lo_encrypt_type = info->lo_encrypt_type;
1037 info64->lo_encrypt_key_size = info->lo_encrypt_key_size;
1038 info64->lo_flags = info->lo_flags;
1039 info64->lo_init[0] = info->lo_init[0];
1040 info64->lo_init[1] = info->lo_init[1];
1041 if (info->lo_encrypt_type == LO_CRYPT_CRYPTOAPI)
1042 memcpy(info64->lo_crypt_name, info->lo_name, LO_NAME_SIZE);
1044 memcpy(info64->lo_file_name, info->lo_name, LO_NAME_SIZE);
1045 memcpy(info64->lo_encrypt_key, info->lo_encrypt_key, LO_KEY_SIZE);
1049 loop_info64_to_old(const struct loop_info64 *info64, struct loop_info *info)
1051 memset(info, 0, sizeof(*info));
1052 info->lo_number = info64->lo_number;
1053 info->lo_device = info64->lo_device;
1054 info->lo_inode = info64->lo_inode;
1055 info->lo_rdevice = info64->lo_rdevice;
1056 info->lo_offset = info64->lo_offset;
1057 info->lo_encrypt_type = info64->lo_encrypt_type;
1058 info->lo_encrypt_key_size = info64->lo_encrypt_key_size;
1059 info->lo_flags = info64->lo_flags;
1060 info->lo_init[0] = info64->lo_init[0];
1061 info->lo_init[1] = info64->lo_init[1];
1062 if (info->lo_encrypt_type == LO_CRYPT_CRYPTOAPI)
1063 memcpy(info->lo_name, info64->lo_crypt_name, LO_NAME_SIZE);
1065 memcpy(info->lo_name, info64->lo_file_name, LO_NAME_SIZE);
1066 memcpy(info->lo_encrypt_key, info64->lo_encrypt_key, LO_KEY_SIZE);
1068 /* error in case values were truncated */
1069 if (info->lo_device != info64->lo_device ||
1070 info->lo_rdevice != info64->lo_rdevice ||
1071 info->lo_inode != info64->lo_inode ||
1072 info->lo_offset != info64->lo_offset)
1079 loop_set_status_old(struct loop_device *lo, const struct loop_info __user *arg)
1081 struct loop_info info;
1082 struct loop_info64 info64;
1084 if (copy_from_user(&info, arg, sizeof (struct loop_info)))
1086 loop_info64_from_old(&info, &info64);
1087 return loop_set_status(lo, &info64);
1091 loop_set_status64(struct loop_device *lo, const struct loop_info64 __user *arg)
1093 struct loop_info64 info64;
1095 if (copy_from_user(&info64, arg, sizeof (struct loop_info64)))
1097 return loop_set_status(lo, &info64);
1101 loop_get_status_old(struct loop_device *lo, struct loop_info __user *arg) {
1102 struct loop_info info;
1103 struct loop_info64 info64;
1109 err = loop_get_status(lo, &info64);
1111 err = loop_info64_to_old(&info64, &info);
1112 if (!err && copy_to_user(arg, &info, sizeof(info)))
1119 loop_get_status64(struct loop_device *lo, struct loop_info64 __user *arg) {
1120 struct loop_info64 info64;
1126 err = loop_get_status(lo, &info64);
1127 if (!err && copy_to_user(arg, &info64, sizeof(info64)))
1133 static int lo_ioctl(struct inode * inode, struct file * file,
1134 unsigned int cmd, unsigned long arg)
1136 struct loop_device *lo = inode->i_bdev->bd_disk->private_data;
1139 mutex_lock(&lo->lo_ctl_mutex);
1142 err = loop_set_fd(lo, file, inode->i_bdev, arg);
1144 case LOOP_CHANGE_FD:
1145 err = loop_change_fd(lo, file, inode->i_bdev, arg);
1148 err = loop_clr_fd(lo, inode->i_bdev);
1150 case LOOP_SET_STATUS:
1151 err = loop_set_status_old(lo, (struct loop_info __user *) arg);
1153 case LOOP_GET_STATUS:
1154 err = loop_get_status_old(lo, (struct loop_info __user *) arg);
1156 case LOOP_SET_STATUS64:
1157 err = loop_set_status64(lo, (struct loop_info64 __user *) arg);
1159 case LOOP_GET_STATUS64:
1160 err = loop_get_status64(lo, (struct loop_info64 __user *) arg);
1163 err = lo->ioctl ? lo->ioctl(lo, cmd, arg) : -EINVAL;
1165 mutex_unlock(&lo->lo_ctl_mutex);
1169 #ifdef CONFIG_COMPAT
1170 struct compat_loop_info {
1171 compat_int_t lo_number; /* ioctl r/o */
1172 compat_dev_t lo_device; /* ioctl r/o */
1173 compat_ulong_t lo_inode; /* ioctl r/o */
1174 compat_dev_t lo_rdevice; /* ioctl r/o */
1175 compat_int_t lo_offset;
1176 compat_int_t lo_encrypt_type;
1177 compat_int_t lo_encrypt_key_size; /* ioctl w/o */
1178 compat_int_t lo_flags; /* ioctl r/o */
1179 char lo_name[LO_NAME_SIZE];
1180 unsigned char lo_encrypt_key[LO_KEY_SIZE]; /* ioctl w/o */
1181 compat_ulong_t lo_init[2];
1186 * Transfer 32-bit compatibility structure in userspace to 64-bit loop info
1187 * - noinlined to reduce stack space usage in main part of driver
1190 loop_info64_from_compat(const struct compat_loop_info *arg,
1191 struct loop_info64 *info64)
1193 struct compat_loop_info info;
1195 if (copy_from_user(&info, arg, sizeof(info)))
1198 memset(info64, 0, sizeof(*info64));
1199 info64->lo_number = info.lo_number;
1200 info64->lo_device = info.lo_device;
1201 info64->lo_inode = info.lo_inode;
1202 info64->lo_rdevice = info.lo_rdevice;
1203 info64->lo_offset = info.lo_offset;
1204 info64->lo_sizelimit = 0;
1205 info64->lo_encrypt_type = info.lo_encrypt_type;
1206 info64->lo_encrypt_key_size = info.lo_encrypt_key_size;
1207 info64->lo_flags = info.lo_flags;
1208 info64->lo_init[0] = info.lo_init[0];
1209 info64->lo_init[1] = info.lo_init[1];
1210 if (info.lo_encrypt_type == LO_CRYPT_CRYPTOAPI)
1211 memcpy(info64->lo_crypt_name, info.lo_name, LO_NAME_SIZE);
1213 memcpy(info64->lo_file_name, info.lo_name, LO_NAME_SIZE);
1214 memcpy(info64->lo_encrypt_key, info.lo_encrypt_key, LO_KEY_SIZE);
1219 * Transfer 64-bit loop info to 32-bit compatibility structure in userspace
1220 * - noinlined to reduce stack space usage in main part of driver
1223 loop_info64_to_compat(const struct loop_info64 *info64,
1224 struct compat_loop_info __user *arg)
1226 struct compat_loop_info info;
1228 memset(&info, 0, sizeof(info));
1229 info.lo_number = info64->lo_number;
1230 info.lo_device = info64->lo_device;
1231 info.lo_inode = info64->lo_inode;
1232 info.lo_rdevice = info64->lo_rdevice;
1233 info.lo_offset = info64->lo_offset;
1234 info.lo_encrypt_type = info64->lo_encrypt_type;
1235 info.lo_encrypt_key_size = info64->lo_encrypt_key_size;
1236 info.lo_flags = info64->lo_flags;
1237 info.lo_init[0] = info64->lo_init[0];
1238 info.lo_init[1] = info64->lo_init[1];
1239 if (info.lo_encrypt_type == LO_CRYPT_CRYPTOAPI)
1240 memcpy(info.lo_name, info64->lo_crypt_name, LO_NAME_SIZE);
1242 memcpy(info.lo_name, info64->lo_file_name, LO_NAME_SIZE);
1243 memcpy(info.lo_encrypt_key, info64->lo_encrypt_key, LO_KEY_SIZE);
1245 /* error in case values were truncated */
1246 if (info.lo_device != info64->lo_device ||
1247 info.lo_rdevice != info64->lo_rdevice ||
1248 info.lo_inode != info64->lo_inode ||
1249 info.lo_offset != info64->lo_offset ||
1250 info.lo_init[0] != info64->lo_init[0] ||
1251 info.lo_init[1] != info64->lo_init[1])
1254 if (copy_to_user(arg, &info, sizeof(info)))
1260 loop_set_status_compat(struct loop_device *lo,
1261 const struct compat_loop_info __user *arg)
1263 struct loop_info64 info64;
1266 ret = loop_info64_from_compat(arg, &info64);
1269 return loop_set_status(lo, &info64);
1273 loop_get_status_compat(struct loop_device *lo,
1274 struct compat_loop_info __user *arg)
1276 struct loop_info64 info64;
1282 err = loop_get_status(lo, &info64);
1284 err = loop_info64_to_compat(&info64, arg);
1288 static long lo_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
1290 struct inode *inode = file->f_dentry->d_inode;
1291 struct loop_device *lo = inode->i_bdev->bd_disk->private_data;
1296 case LOOP_SET_STATUS:
1297 mutex_lock(&lo->lo_ctl_mutex);
1298 err = loop_set_status_compat(
1299 lo, (const struct compat_loop_info __user *) arg);
1300 mutex_unlock(&lo->lo_ctl_mutex);
1302 case LOOP_GET_STATUS:
1303 mutex_lock(&lo->lo_ctl_mutex);
1304 err = loop_get_status_compat(
1305 lo, (struct compat_loop_info __user *) arg);
1306 mutex_unlock(&lo->lo_ctl_mutex);
1309 case LOOP_GET_STATUS64:
1310 case LOOP_SET_STATUS64:
1311 arg = (unsigned long) compat_ptr(arg);
1313 case LOOP_CHANGE_FD:
1314 err = lo_ioctl(inode, file, cmd, arg);
1325 static int lo_open(struct inode *inode, struct file *file)
1327 struct loop_device *lo = inode->i_bdev->bd_disk->private_data;
1329 mutex_lock(&lo->lo_ctl_mutex);
1331 mutex_unlock(&lo->lo_ctl_mutex);
1336 static int lo_release(struct inode *inode, struct file *file)
1338 struct loop_device *lo = inode->i_bdev->bd_disk->private_data;
1340 mutex_lock(&lo->lo_ctl_mutex);
1342 mutex_unlock(&lo->lo_ctl_mutex);
1347 static struct block_device_operations lo_fops = {
1348 .owner = THIS_MODULE,
1350 .release = lo_release,
1352 #ifdef CONFIG_COMPAT
1353 .compat_ioctl = lo_compat_ioctl,
1358 * And now the modules code and kernel interface.
1360 module_param(max_loop, int, 0);
1361 MODULE_PARM_DESC(max_loop, "Maximum number of loop devices (1-256)");
1362 MODULE_LICENSE("GPL");
1363 MODULE_ALIAS_BLOCKDEV_MAJOR(LOOP_MAJOR);
1365 int loop_register_transfer(struct loop_func_table *funcs)
1367 unsigned int n = funcs->number;
1369 if (n >= MAX_LO_CRYPT || xfer_funcs[n])
1371 xfer_funcs[n] = funcs;
1375 int loop_unregister_transfer(int number)
1377 unsigned int n = number;
1378 struct loop_device *lo;
1379 struct loop_func_table *xfer;
1381 if (n == 0 || n >= MAX_LO_CRYPT || (xfer = xfer_funcs[n]) == NULL)
1384 xfer_funcs[n] = NULL;
1386 for (lo = &loop_dev[0]; lo < &loop_dev[max_loop]; lo++) {
1387 mutex_lock(&lo->lo_ctl_mutex);
1389 if (lo->lo_encryption == xfer)
1390 loop_release_xfer(lo);
1392 mutex_unlock(&lo->lo_ctl_mutex);
1398 EXPORT_SYMBOL(loop_register_transfer);
1399 EXPORT_SYMBOL(loop_unregister_transfer);
1401 static int __init loop_init(void)
1405 if (max_loop < 1 || max_loop > 256) {
1406 printk(KERN_WARNING "loop: invalid max_loop (must be between"
1407 " 1 and 256), using default (8)\n");
1411 if (register_blkdev(LOOP_MAJOR, "loop"))
1414 loop_dev = kmalloc(max_loop * sizeof(struct loop_device), GFP_KERNEL);
1417 memset(loop_dev, 0, max_loop * sizeof(struct loop_device));
1419 disks = kmalloc(max_loop * sizeof(struct gendisk *), GFP_KERNEL);
1423 for (i = 0; i < max_loop; i++) {
1424 disks[i] = alloc_disk(1);
1429 for (i = 0; i < max_loop; i++) {
1430 struct loop_device *lo = &loop_dev[i];
1431 struct gendisk *disk = disks[i];
1433 memset(lo, 0, sizeof(*lo));
1434 lo->lo_queue = blk_alloc_queue(GFP_KERNEL);
1437 mutex_init(&lo->lo_ctl_mutex);
1439 lo->lo_thread = NULL;
1440 init_waitqueue_head(&lo->lo_event);
1441 spin_lock_init(&lo->lo_lock);
1442 disk->major = LOOP_MAJOR;
1443 disk->first_minor = i;
1444 disk->fops = &lo_fops;
1445 sprintf(disk->disk_name, "loop%d", i);
1446 disk->private_data = lo;
1447 disk->queue = lo->lo_queue;
1450 /* We cannot fail after we call this, so another loop!*/
1451 for (i = 0; i < max_loop; i++)
1453 printk(KERN_INFO "loop: loaded (max %d devices)\n", max_loop);
1458 blk_cleanup_queue(loop_dev[i].lo_queue);
1467 unregister_blkdev(LOOP_MAJOR, "loop");
1468 printk(KERN_ERR "loop: ran out of memory\n");
1472 static void loop_exit(void)
1476 for (i = 0; i < max_loop; i++) {
1477 del_gendisk(disks[i]);
1478 blk_cleanup_queue(loop_dev[i].lo_queue);
1481 if (unregister_blkdev(LOOP_MAJOR, "loop"))
1482 printk(KERN_WARNING "loop: cannot unregister blkdev\n");
1488 module_init(loop_init);
1489 module_exit(loop_exit);
1492 static int __init max_loop_setup(char *str)
1494 max_loop = simple_strtol(str, NULL, 0);
1498 __setup("max_loop=", max_loop_setup);