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/config.h>
54 #include <linux/module.h>
55 #include <linux/moduleparam.h>
56 #include <linux/sched.h>
58 #include <linux/file.h>
59 #include <linux/stat.h>
60 #include <linux/errno.h>
61 #include <linux/major.h>
62 #include <linux/wait.h>
63 #include <linux/blkdev.h>
64 #include <linux/blkpg.h>
65 #include <linux/init.h>
66 #include <linux/devfs_fs_kernel.h>
67 #include <linux/smp_lock.h>
68 #include <linux/swap.h>
69 #include <linux/slab.h>
70 #include <linux/loop.h>
71 #include <linux/suspend.h>
72 #include <linux/writeback.h>
73 #include <linux/buffer_head.h> /* for invalidate_bdev() */
74 #include <linux/completion.h>
75 #include <linux/highmem.h>
76 #include <linux/gfp.h>
77 #include <linux/kthread.h>
79 #include <asm/uaccess.h>
81 static int max_loop = 8;
82 static struct loop_device *loop_dev;
83 static struct gendisk **disks;
88 static int transfer_none(struct loop_device *lo, int cmd,
89 struct page *raw_page, unsigned raw_off,
90 struct page *loop_page, unsigned loop_off,
91 int size, sector_t real_block)
93 char *raw_buf = kmap_atomic(raw_page, KM_USER0) + raw_off;
94 char *loop_buf = kmap_atomic(loop_page, KM_USER1) + loop_off;
97 memcpy(loop_buf, raw_buf, size);
99 memcpy(raw_buf, loop_buf, size);
101 kunmap_atomic(raw_buf, KM_USER0);
102 kunmap_atomic(loop_buf, KM_USER1);
107 static int transfer_xor(struct loop_device *lo, int cmd,
108 struct page *raw_page, unsigned raw_off,
109 struct page *loop_page, unsigned loop_off,
110 int size, sector_t real_block)
112 char *raw_buf = kmap_atomic(raw_page, KM_USER0) + raw_off;
113 char *loop_buf = kmap_atomic(loop_page, KM_USER1) + loop_off;
114 char *in, *out, *key;
125 key = lo->lo_encrypt_key;
126 keysize = lo->lo_encrypt_key_size;
127 for (i = 0; i < size; i++)
128 *out++ = *in++ ^ key[(i & 511) % keysize];
130 kunmap_atomic(raw_buf, KM_USER0);
131 kunmap_atomic(loop_buf, KM_USER1);
136 static int xor_init(struct loop_device *lo, const struct loop_info64 *info)
138 if (unlikely(info->lo_encrypt_key_size <= 0))
143 static struct loop_func_table none_funcs = {
144 .number = LO_CRYPT_NONE,
145 .transfer = transfer_none,
148 static struct loop_func_table xor_funcs = {
149 .number = LO_CRYPT_XOR,
150 .transfer = transfer_xor,
154 /* xfer_funcs[0] is special - its release function is never called */
155 static struct loop_func_table *xfer_funcs[MAX_LO_CRYPT] = {
160 static loff_t get_loop_size(struct loop_device *lo, struct file *file)
162 loff_t size, offset, loopsize;
164 /* Compute loopsize in bytes */
165 size = i_size_read(file->f_mapping->host);
166 offset = lo->lo_offset;
167 loopsize = size - offset;
168 if (lo->lo_sizelimit > 0 && lo->lo_sizelimit < loopsize)
169 loopsize = lo->lo_sizelimit;
172 * Unfortunately, if we want to do I/O on the device,
173 * the number of 512-byte sectors has to fit into a sector_t.
175 return loopsize >> 9;
179 figure_loop_size(struct loop_device *lo)
181 loff_t size = get_loop_size(lo, lo->lo_backing_file);
182 sector_t x = (sector_t)size;
184 if (unlikely((loff_t)x != size))
187 set_capacity(disks[lo->lo_number], x);
192 lo_do_transfer(struct loop_device *lo, int cmd,
193 struct page *rpage, unsigned roffs,
194 struct page *lpage, unsigned loffs,
195 int size, sector_t rblock)
197 if (unlikely(!lo->transfer))
200 return lo->transfer(lo, cmd, rpage, roffs, lpage, loffs, size, rblock);
204 * do_lo_send_aops - helper for writing data to a loop device
206 * This is the fast version for backing filesystems which implement the address
207 * space operations prepare_write and commit_write.
209 static int do_lo_send_aops(struct loop_device *lo, struct bio_vec *bvec,
210 int bsize, loff_t pos, struct page *page)
212 struct file *file = lo->lo_backing_file; /* kudos to NFsckingS */
213 struct address_space *mapping = file->f_mapping;
214 struct address_space_operations *aops = mapping->a_ops;
216 unsigned offset, bv_offs;
219 mutex_lock(&mapping->host->i_mutex);
220 index = pos >> PAGE_CACHE_SHIFT;
221 offset = pos & ((pgoff_t)PAGE_CACHE_SIZE - 1);
222 bv_offs = bvec->bv_offset;
229 IV = ((sector_t)index << (PAGE_CACHE_SHIFT - 9))+(offset >> 9);
230 size = PAGE_CACHE_SIZE - offset;
233 page = grab_cache_page(mapping, index);
236 ret = aops->prepare_write(file, page, offset,
239 if (ret == AOP_TRUNCATED_PAGE) {
240 page_cache_release(page);
245 transfer_result = lo_do_transfer(lo, WRITE, page, offset,
246 bvec->bv_page, bv_offs, size, IV);
247 if (unlikely(transfer_result)) {
251 * The transfer failed, but we still write the data to
252 * keep prepare/commit calls balanced.
254 printk(KERN_ERR "loop: transfer error block %llu\n",
255 (unsigned long long)index);
256 kaddr = kmap_atomic(page, KM_USER0);
257 memset(kaddr + offset, 0, size);
258 kunmap_atomic(kaddr, KM_USER0);
260 flush_dcache_page(page);
261 ret = aops->commit_write(file, page, offset,
264 if (ret == AOP_TRUNCATED_PAGE) {
265 page_cache_release(page);
270 if (unlikely(transfer_result))
278 page_cache_release(page);
282 mutex_unlock(&mapping->host->i_mutex);
286 page_cache_release(page);
293 * __do_lo_send_write - helper for writing data to a loop device
295 * This helper just factors out common code between do_lo_send_direct_write()
296 * and do_lo_send_write().
298 static int __do_lo_send_write(struct file *file,
299 u8 __user *buf, const int len, loff_t pos)
302 mm_segment_t old_fs = get_fs();
305 bw = file->f_op->write(file, buf, len, &pos);
307 if (likely(bw == len))
309 printk(KERN_ERR "loop: Write error at byte offset %llu, length %i.\n",
310 (unsigned long long)pos, len);
317 * do_lo_send_direct_write - helper for writing data to a loop device
319 * This is the fast, non-transforming version for backing filesystems which do
320 * not implement the address space operations prepare_write and commit_write.
321 * It uses the write file operation which should be present on all writeable
324 static int do_lo_send_direct_write(struct loop_device *lo,
325 struct bio_vec *bvec, int bsize, loff_t pos, struct page *page)
327 ssize_t bw = __do_lo_send_write(lo->lo_backing_file,
328 (u8 __user *)kmap(bvec->bv_page) + bvec->bv_offset,
330 kunmap(bvec->bv_page);
336 * do_lo_send_write - helper for writing data to a loop device
338 * This is the slow, transforming version for filesystems which do not
339 * implement the address space operations prepare_write and commit_write. It
340 * uses the write file operation which should be present on all writeable
343 * Using fops->write is slower than using aops->{prepare,commit}_write in the
344 * transforming case because we need to double buffer the data as we cannot do
345 * the transformations in place as we do not have direct access to the
346 * destination pages of the backing file.
348 static int do_lo_send_write(struct loop_device *lo, struct bio_vec *bvec,
349 int bsize, loff_t pos, struct page *page)
351 int ret = lo_do_transfer(lo, WRITE, page, 0, bvec->bv_page,
352 bvec->bv_offset, bvec->bv_len, pos >> 9);
354 return __do_lo_send_write(lo->lo_backing_file,
355 (u8 __user *)page_address(page), bvec->bv_len,
357 printk(KERN_ERR "loop: Transfer error at byte offset %llu, "
358 "length %i.\n", (unsigned long long)pos, bvec->bv_len);
364 static int lo_send(struct loop_device *lo, struct bio *bio, int bsize,
367 int (*do_lo_send)(struct loop_device *, struct bio_vec *, int, loff_t,
369 struct bio_vec *bvec;
370 struct page *page = NULL;
373 do_lo_send = do_lo_send_aops;
374 if (!(lo->lo_flags & LO_FLAGS_USE_AOPS)) {
375 do_lo_send = do_lo_send_direct_write;
376 if (lo->transfer != transfer_none) {
377 page = alloc_page(GFP_NOIO | __GFP_HIGHMEM);
381 do_lo_send = do_lo_send_write;
384 bio_for_each_segment(bvec, bio, i) {
385 ret = do_lo_send(lo, bvec, bsize, pos, page);
397 printk(KERN_ERR "loop: Failed to allocate temporary page for write.\n");
402 struct lo_read_data {
403 struct loop_device *lo;
410 lo_read_actor(read_descriptor_t *desc, struct page *page,
411 unsigned long offset, unsigned long size)
413 unsigned long count = desc->count;
414 struct lo_read_data *p = desc->arg.data;
415 struct loop_device *lo = p->lo;
418 IV = ((sector_t) page->index << (PAGE_CACHE_SHIFT - 9))+(offset >> 9);
423 if (lo_do_transfer(lo, READ, page, offset, p->page, p->offset, size, IV)) {
425 printk(KERN_ERR "loop: transfer error block %ld\n",
427 desc->error = -EINVAL;
430 flush_dcache_page(p->page);
432 desc->count = count - size;
433 desc->written += size;
439 do_lo_receive(struct loop_device *lo,
440 struct bio_vec *bvec, int bsize, loff_t pos)
442 struct lo_read_data cookie;
447 cookie.page = bvec->bv_page;
448 cookie.offset = bvec->bv_offset;
449 cookie.bsize = bsize;
450 file = lo->lo_backing_file;
451 retval = file->f_op->sendfile(file, &pos, bvec->bv_len,
452 lo_read_actor, &cookie);
453 return (retval < 0)? retval: 0;
457 lo_receive(struct loop_device *lo, struct bio *bio, int bsize, loff_t pos)
459 struct bio_vec *bvec;
462 bio_for_each_segment(bvec, bio, i) {
463 ret = do_lo_receive(lo, bvec, bsize, pos);
471 static int do_bio_filebacked(struct loop_device *lo, struct bio *bio)
476 pos = ((loff_t) bio->bi_sector << 9) + lo->lo_offset;
477 if (bio_rw(bio) == WRITE)
478 ret = lo_send(lo, bio, lo->lo_blocksize, pos);
480 ret = lo_receive(lo, bio, lo->lo_blocksize, pos);
485 * Add bio to back of pending list
487 static void loop_add_bio(struct loop_device *lo, struct bio *bio)
489 if (lo->lo_biotail) {
490 lo->lo_biotail->bi_next = bio;
491 lo->lo_biotail = bio;
493 lo->lo_bio = lo->lo_biotail = bio;
497 * Grab first pending buffer
499 static struct bio *loop_get_bio(struct loop_device *lo)
503 if ((bio = lo->lo_bio)) {
504 if (bio == lo->lo_biotail)
505 lo->lo_biotail = NULL;
506 lo->lo_bio = bio->bi_next;
513 static int loop_make_request(request_queue_t *q, struct bio *old_bio)
515 struct loop_device *lo = q->queuedata;
516 int rw = bio_rw(old_bio);
521 BUG_ON(!lo || (rw != READ && rw != WRITE));
523 spin_lock_irq(&lo->lo_lock);
524 if (lo->lo_state != Lo_bound)
526 if (unlikely(rw == WRITE && (lo->lo_flags & LO_FLAGS_READ_ONLY)))
529 loop_add_bio(lo, old_bio);
530 spin_unlock_irq(&lo->lo_lock);
531 complete(&lo->lo_bh_done);
535 if (lo->lo_pending == 0)
536 complete(&lo->lo_bh_done);
537 spin_unlock_irq(&lo->lo_lock);
538 bio_io_error(old_bio, old_bio->bi_size);
543 * kick off io on the underlying address space
545 static void loop_unplug(request_queue_t *q)
547 struct loop_device *lo = q->queuedata;
549 clear_bit(QUEUE_FLAG_PLUGGED, &q->queue_flags);
550 blk_run_address_space(lo->lo_backing_file->f_mapping);
553 struct switch_request {
555 struct completion wait;
558 static void do_loop_switch(struct loop_device *, struct switch_request *);
560 static inline void loop_handle_bio(struct loop_device *lo, struct bio *bio)
562 if (unlikely(!bio->bi_bdev)) {
563 do_loop_switch(lo, bio->bi_private);
566 int ret = do_bio_filebacked(lo, bio);
567 bio_endio(bio, bio->bi_size, ret);
572 * worker thread that handles reads/writes to file backed loop devices,
573 * to avoid blocking in our make_request_fn. it also does loop decrypting
574 * on reads for block backed loop, as that is too heavy to do from
575 * b_end_io context where irqs may be disabled.
577 static int loop_thread(void *data)
579 struct loop_device *lo = data;
583 * loop can be used in an encrypted device,
584 * hence, it mustn't be stopped at all
585 * because it could be indirectly used during suspension
587 current->flags |= PF_NOFREEZE;
589 set_user_nice(current, -20);
591 lo->lo_state = Lo_bound;
597 if (wait_for_completion_interruptible(&lo->lo_bh_done))
600 spin_lock_irq(&lo->lo_lock);
603 * could be completed because of tear-down, not pending work
605 if (unlikely(!lo->lo_pending)) {
606 spin_unlock_irq(&lo->lo_lock);
610 bio = loop_get_bio(lo);
612 pending = lo->lo_pending;
613 spin_unlock_irq(&lo->lo_lock);
616 loop_handle_bio(lo, bio);
619 * upped both for pending work and tear-down, lo_pending
622 if (unlikely(!pending))
630 * loop_switch performs the hard work of switching a backing store.
631 * First it needs to flush existing IO, it does this by sending a magic
632 * BIO down the pipe. The completion of this BIO does the actual switch.
634 static int loop_switch(struct loop_device *lo, struct file *file)
636 struct switch_request w;
637 struct bio *bio = bio_alloc(GFP_KERNEL, 1);
640 init_completion(&w.wait);
642 bio->bi_private = &w;
644 loop_make_request(lo->lo_queue, bio);
645 wait_for_completion(&w.wait);
650 * Do the actual switch; called from the BIO completion routine
652 static void do_loop_switch(struct loop_device *lo, struct switch_request *p)
654 struct file *file = p->file;
655 struct file *old_file = lo->lo_backing_file;
656 struct address_space *mapping = file->f_mapping;
658 mapping_set_gfp_mask(old_file->f_mapping, lo->old_gfp_mask);
659 lo->lo_backing_file = file;
660 lo->lo_blocksize = mapping->host->i_blksize;
661 lo->old_gfp_mask = mapping_gfp_mask(mapping);
662 mapping_set_gfp_mask(mapping, lo->old_gfp_mask & ~(__GFP_IO|__GFP_FS));
668 * loop_change_fd switched the backing store of a loopback device to
669 * a new file. This is useful for operating system installers to free up
670 * the original file and in High Availability environments to switch to
671 * an alternative location for the content in case of server meltdown.
672 * This can only work if the loop device is used read-only, and if the
673 * new backing store is the same size and type as the old backing store.
675 static int loop_change_fd(struct loop_device *lo, struct file *lo_file,
676 struct block_device *bdev, unsigned int arg)
678 struct file *file, *old_file;
683 if (lo->lo_state != Lo_bound)
686 /* the loop device has to be read-only */
688 if (!(lo->lo_flags & LO_FLAGS_READ_ONLY))
696 inode = file->f_mapping->host;
697 old_file = lo->lo_backing_file;
701 if (!S_ISREG(inode->i_mode) && !S_ISBLK(inode->i_mode))
704 /* new backing store needs to support loop (eg sendfile) */
705 if (!inode->i_fop->sendfile)
708 /* size of the new backing store needs to be the same */
709 if (get_loop_size(lo, file) != get_loop_size(lo, old_file))
713 error = loop_switch(lo, file);
726 static inline int is_loop_device(struct file *file)
728 struct inode *i = file->f_mapping->host;
730 return i && S_ISBLK(i->i_mode) && MAJOR(i->i_rdev) == LOOP_MAJOR;
733 static int loop_set_fd(struct loop_device *lo, struct file *lo_file,
734 struct block_device *bdev, unsigned int arg)
736 struct file *file, *f;
738 struct address_space *mapping;
739 unsigned lo_blocksize;
742 struct task_struct *tsk;
745 /* This is safe, since we have a reference from open(). */
746 __module_get(THIS_MODULE);
754 if (lo->lo_state != Lo_unbound)
757 /* Avoid recursion */
759 while (is_loop_device(f)) {
760 struct loop_device *l;
762 if (f->f_mapping->host->i_rdev == lo_file->f_mapping->host->i_rdev)
765 l = f->f_mapping->host->i_bdev->bd_disk->private_data;
766 if (l->lo_state == Lo_unbound) {
770 f = l->lo_backing_file;
773 mapping = file->f_mapping;
774 inode = mapping->host;
776 if (!(file->f_mode & FMODE_WRITE))
777 lo_flags |= LO_FLAGS_READ_ONLY;
780 if (S_ISREG(inode->i_mode) || S_ISBLK(inode->i_mode)) {
781 struct address_space_operations *aops = mapping->a_ops;
783 * If we can't read - sorry. If we only can't write - well,
784 * it's going to be read-only.
786 if (!file->f_op->sendfile)
788 if (aops->prepare_write && aops->commit_write)
789 lo_flags |= LO_FLAGS_USE_AOPS;
790 if (!(lo_flags & LO_FLAGS_USE_AOPS) && !file->f_op->write)
791 lo_flags |= LO_FLAGS_READ_ONLY;
793 lo_blocksize = inode->i_blksize;
799 size = get_loop_size(lo, file);
801 if ((loff_t)(sector_t)size != size) {
806 if (!(lo_file->f_mode & FMODE_WRITE))
807 lo_flags |= LO_FLAGS_READ_ONLY;
809 set_device_ro(bdev, (lo_flags & LO_FLAGS_READ_ONLY) != 0);
811 lo->lo_blocksize = lo_blocksize;
812 lo->lo_device = bdev;
813 lo->lo_flags = lo_flags;
814 lo->lo_backing_file = file;
815 lo->transfer = transfer_none;
817 lo->lo_sizelimit = 0;
818 lo->old_gfp_mask = mapping_gfp_mask(mapping);
819 mapping_set_gfp_mask(mapping, lo->old_gfp_mask & ~(__GFP_IO|__GFP_FS));
821 lo->lo_bio = lo->lo_biotail = NULL;
824 * set queue make_request_fn, and add limits based on lower level
827 blk_queue_make_request(lo->lo_queue, loop_make_request);
828 lo->lo_queue->queuedata = lo;
829 lo->lo_queue->unplug_fn = loop_unplug;
831 set_capacity(disks[lo->lo_number], size);
832 bd_set_size(bdev, size << 9);
834 set_blocksize(bdev, lo_blocksize);
836 tsk = kthread_run(loop_thread, lo, "loop%d", lo->lo_number);
838 error = PTR_ERR(tsk);
846 /* This is safe: open() is still holding a reference. */
847 module_put(THIS_MODULE);
852 loop_release_xfer(struct loop_device *lo)
855 struct loop_func_table *xfer = lo->lo_encryption;
859 err = xfer->release(lo);
861 lo->lo_encryption = NULL;
862 module_put(xfer->owner);
868 loop_init_xfer(struct loop_device *lo, struct loop_func_table *xfer,
869 const struct loop_info64 *i)
874 struct module *owner = xfer->owner;
876 if (!try_module_get(owner))
879 err = xfer->init(lo, i);
883 lo->lo_encryption = xfer;
888 static int loop_clr_fd(struct loop_device *lo, struct block_device *bdev)
890 struct file *filp = lo->lo_backing_file;
891 gfp_t gfp = lo->old_gfp_mask;
893 if (lo->lo_state != Lo_bound)
899 if (lo->lo_refcnt > 1) /* we needed one fd for the ioctl */
905 spin_lock_irq(&lo->lo_lock);
906 lo->lo_state = Lo_rundown;
909 complete(&lo->lo_bh_done);
910 spin_unlock_irq(&lo->lo_lock);
912 kthread_stop(lo->lo_thread);
914 lo->lo_backing_file = NULL;
916 loop_release_xfer(lo);
919 lo->lo_device = NULL;
920 lo->lo_encryption = NULL;
922 lo->lo_sizelimit = 0;
923 lo->lo_encrypt_key_size = 0;
925 lo->lo_thread = NULL;
926 memset(lo->lo_encrypt_key, 0, LO_KEY_SIZE);
927 memset(lo->lo_crypt_name, 0, LO_NAME_SIZE);
928 memset(lo->lo_file_name, 0, LO_NAME_SIZE);
929 invalidate_bdev(bdev, 0);
930 set_capacity(disks[lo->lo_number], 0);
931 bd_set_size(bdev, 0);
932 mapping_set_gfp_mask(filp->f_mapping, gfp);
933 lo->lo_state = Lo_unbound;
935 /* This is safe: open() is still holding a reference. */
936 module_put(THIS_MODULE);
941 loop_set_status(struct loop_device *lo, const struct loop_info64 *info)
944 struct loop_func_table *xfer;
946 if (lo->lo_encrypt_key_size && lo->lo_key_owner != current->uid &&
947 !capable(CAP_SYS_ADMIN))
949 if (lo->lo_state != Lo_bound)
951 if ((unsigned int) info->lo_encrypt_key_size > LO_KEY_SIZE)
954 err = loop_release_xfer(lo);
958 if (info->lo_encrypt_type) {
959 unsigned int type = info->lo_encrypt_type;
961 if (type >= MAX_LO_CRYPT)
963 xfer = xfer_funcs[type];
969 err = loop_init_xfer(lo, xfer, info);
973 if (lo->lo_offset != info->lo_offset ||
974 lo->lo_sizelimit != info->lo_sizelimit) {
975 lo->lo_offset = info->lo_offset;
976 lo->lo_sizelimit = info->lo_sizelimit;
977 if (figure_loop_size(lo))
981 memcpy(lo->lo_file_name, info->lo_file_name, LO_NAME_SIZE);
982 memcpy(lo->lo_crypt_name, info->lo_crypt_name, LO_NAME_SIZE);
983 lo->lo_file_name[LO_NAME_SIZE-1] = 0;
984 lo->lo_crypt_name[LO_NAME_SIZE-1] = 0;
988 lo->transfer = xfer->transfer;
989 lo->ioctl = xfer->ioctl;
991 lo->lo_encrypt_key_size = info->lo_encrypt_key_size;
992 lo->lo_init[0] = info->lo_init[0];
993 lo->lo_init[1] = info->lo_init[1];
994 if (info->lo_encrypt_key_size) {
995 memcpy(lo->lo_encrypt_key, info->lo_encrypt_key,
996 info->lo_encrypt_key_size);
997 lo->lo_key_owner = current->uid;
1004 loop_get_status(struct loop_device *lo, struct loop_info64 *info)
1006 struct file *file = lo->lo_backing_file;
1010 if (lo->lo_state != Lo_bound)
1012 error = vfs_getattr(file->f_vfsmnt, file->f_dentry, &stat);
1015 memset(info, 0, sizeof(*info));
1016 info->lo_number = lo->lo_number;
1017 info->lo_device = huge_encode_dev(stat.dev);
1018 info->lo_inode = stat.ino;
1019 info->lo_rdevice = huge_encode_dev(lo->lo_device ? stat.rdev : stat.dev);
1020 info->lo_offset = lo->lo_offset;
1021 info->lo_sizelimit = lo->lo_sizelimit;
1022 info->lo_flags = lo->lo_flags;
1023 memcpy(info->lo_file_name, lo->lo_file_name, LO_NAME_SIZE);
1024 memcpy(info->lo_crypt_name, lo->lo_crypt_name, LO_NAME_SIZE);
1025 info->lo_encrypt_type =
1026 lo->lo_encryption ? lo->lo_encryption->number : 0;
1027 if (lo->lo_encrypt_key_size && capable(CAP_SYS_ADMIN)) {
1028 info->lo_encrypt_key_size = lo->lo_encrypt_key_size;
1029 memcpy(info->lo_encrypt_key, lo->lo_encrypt_key,
1030 lo->lo_encrypt_key_size);
1036 loop_info64_from_old(const struct loop_info *info, struct loop_info64 *info64)
1038 memset(info64, 0, sizeof(*info64));
1039 info64->lo_number = info->lo_number;
1040 info64->lo_device = info->lo_device;
1041 info64->lo_inode = info->lo_inode;
1042 info64->lo_rdevice = info->lo_rdevice;
1043 info64->lo_offset = info->lo_offset;
1044 info64->lo_sizelimit = 0;
1045 info64->lo_encrypt_type = info->lo_encrypt_type;
1046 info64->lo_encrypt_key_size = info->lo_encrypt_key_size;
1047 info64->lo_flags = info->lo_flags;
1048 info64->lo_init[0] = info->lo_init[0];
1049 info64->lo_init[1] = info->lo_init[1];
1050 if (info->lo_encrypt_type == LO_CRYPT_CRYPTOAPI)
1051 memcpy(info64->lo_crypt_name, info->lo_name, LO_NAME_SIZE);
1053 memcpy(info64->lo_file_name, info->lo_name, LO_NAME_SIZE);
1054 memcpy(info64->lo_encrypt_key, info->lo_encrypt_key, LO_KEY_SIZE);
1058 loop_info64_to_old(const struct loop_info64 *info64, struct loop_info *info)
1060 memset(info, 0, sizeof(*info));
1061 info->lo_number = info64->lo_number;
1062 info->lo_device = info64->lo_device;
1063 info->lo_inode = info64->lo_inode;
1064 info->lo_rdevice = info64->lo_rdevice;
1065 info->lo_offset = info64->lo_offset;
1066 info->lo_encrypt_type = info64->lo_encrypt_type;
1067 info->lo_encrypt_key_size = info64->lo_encrypt_key_size;
1068 info->lo_flags = info64->lo_flags;
1069 info->lo_init[0] = info64->lo_init[0];
1070 info->lo_init[1] = info64->lo_init[1];
1071 if (info->lo_encrypt_type == LO_CRYPT_CRYPTOAPI)
1072 memcpy(info->lo_name, info64->lo_crypt_name, LO_NAME_SIZE);
1074 memcpy(info->lo_name, info64->lo_file_name, LO_NAME_SIZE);
1075 memcpy(info->lo_encrypt_key, info64->lo_encrypt_key, LO_KEY_SIZE);
1077 /* error in case values were truncated */
1078 if (info->lo_device != info64->lo_device ||
1079 info->lo_rdevice != info64->lo_rdevice ||
1080 info->lo_inode != info64->lo_inode ||
1081 info->lo_offset != info64->lo_offset)
1088 loop_set_status_old(struct loop_device *lo, const struct loop_info __user *arg)
1090 struct loop_info info;
1091 struct loop_info64 info64;
1093 if (copy_from_user(&info, arg, sizeof (struct loop_info)))
1095 loop_info64_from_old(&info, &info64);
1096 return loop_set_status(lo, &info64);
1100 loop_set_status64(struct loop_device *lo, const struct loop_info64 __user *arg)
1102 struct loop_info64 info64;
1104 if (copy_from_user(&info64, arg, sizeof (struct loop_info64)))
1106 return loop_set_status(lo, &info64);
1110 loop_get_status_old(struct loop_device *lo, struct loop_info __user *arg) {
1111 struct loop_info info;
1112 struct loop_info64 info64;
1118 err = loop_get_status(lo, &info64);
1120 err = loop_info64_to_old(&info64, &info);
1121 if (!err && copy_to_user(arg, &info, sizeof(info)))
1128 loop_get_status64(struct loop_device *lo, struct loop_info64 __user *arg) {
1129 struct loop_info64 info64;
1135 err = loop_get_status(lo, &info64);
1136 if (!err && copy_to_user(arg, &info64, sizeof(info64)))
1142 static int lo_ioctl(struct inode * inode, struct file * file,
1143 unsigned int cmd, unsigned long arg)
1145 struct loop_device *lo = inode->i_bdev->bd_disk->private_data;
1148 mutex_lock(&lo->lo_ctl_mutex);
1151 err = loop_set_fd(lo, file, inode->i_bdev, arg);
1153 case LOOP_CHANGE_FD:
1154 err = loop_change_fd(lo, file, inode->i_bdev, arg);
1157 err = loop_clr_fd(lo, inode->i_bdev);
1159 case LOOP_SET_STATUS:
1160 err = loop_set_status_old(lo, (struct loop_info __user *) arg);
1162 case LOOP_GET_STATUS:
1163 err = loop_get_status_old(lo, (struct loop_info __user *) arg);
1165 case LOOP_SET_STATUS64:
1166 err = loop_set_status64(lo, (struct loop_info64 __user *) arg);
1168 case LOOP_GET_STATUS64:
1169 err = loop_get_status64(lo, (struct loop_info64 __user *) arg);
1172 err = lo->ioctl ? lo->ioctl(lo, cmd, arg) : -EINVAL;
1174 mutex_unlock(&lo->lo_ctl_mutex);
1178 static int lo_open(struct inode *inode, struct file *file)
1180 struct loop_device *lo = inode->i_bdev->bd_disk->private_data;
1182 mutex_lock(&lo->lo_ctl_mutex);
1184 mutex_unlock(&lo->lo_ctl_mutex);
1189 static int lo_release(struct inode *inode, struct file *file)
1191 struct loop_device *lo = inode->i_bdev->bd_disk->private_data;
1193 mutex_lock(&lo->lo_ctl_mutex);
1195 mutex_unlock(&lo->lo_ctl_mutex);
1200 static struct block_device_operations lo_fops = {
1201 .owner = THIS_MODULE,
1203 .release = lo_release,
1208 * And now the modules code and kernel interface.
1210 module_param(max_loop, int, 0);
1211 MODULE_PARM_DESC(max_loop, "Maximum number of loop devices (1-256)");
1212 MODULE_LICENSE("GPL");
1213 MODULE_ALIAS_BLOCKDEV_MAJOR(LOOP_MAJOR);
1215 int loop_register_transfer(struct loop_func_table *funcs)
1217 unsigned int n = funcs->number;
1219 if (n >= MAX_LO_CRYPT || xfer_funcs[n])
1221 xfer_funcs[n] = funcs;
1225 int loop_unregister_transfer(int number)
1227 unsigned int n = number;
1228 struct loop_device *lo;
1229 struct loop_func_table *xfer;
1231 if (n == 0 || n >= MAX_LO_CRYPT || (xfer = xfer_funcs[n]) == NULL)
1234 xfer_funcs[n] = NULL;
1236 for (lo = &loop_dev[0]; lo < &loop_dev[max_loop]; lo++) {
1237 mutex_lock(&lo->lo_ctl_mutex);
1239 if (lo->lo_encryption == xfer)
1240 loop_release_xfer(lo);
1242 mutex_unlock(&lo->lo_ctl_mutex);
1248 EXPORT_SYMBOL(loop_register_transfer);
1249 EXPORT_SYMBOL(loop_unregister_transfer);
1251 static int __init loop_init(void)
1255 if (max_loop < 1 || max_loop > 256) {
1256 printk(KERN_WARNING "loop: invalid max_loop (must be between"
1257 " 1 and 256), using default (8)\n");
1261 if (register_blkdev(LOOP_MAJOR, "loop"))
1264 loop_dev = kmalloc(max_loop * sizeof(struct loop_device), GFP_KERNEL);
1267 memset(loop_dev, 0, max_loop * sizeof(struct loop_device));
1269 disks = kmalloc(max_loop * sizeof(struct gendisk *), GFP_KERNEL);
1273 for (i = 0; i < max_loop; i++) {
1274 disks[i] = alloc_disk(1);
1279 devfs_mk_dir("loop");
1281 for (i = 0; i < max_loop; i++) {
1282 struct loop_device *lo = &loop_dev[i];
1283 struct gendisk *disk = disks[i];
1285 memset(lo, 0, sizeof(*lo));
1286 lo->lo_queue = blk_alloc_queue(GFP_KERNEL);
1289 mutex_init(&lo->lo_ctl_mutex);
1290 init_completion(&lo->lo_bh_done);
1292 spin_lock_init(&lo->lo_lock);
1293 disk->major = LOOP_MAJOR;
1294 disk->first_minor = i;
1295 disk->fops = &lo_fops;
1296 sprintf(disk->disk_name, "loop%d", i);
1297 sprintf(disk->devfs_name, "loop/%d", i);
1298 disk->private_data = lo;
1299 disk->queue = lo->lo_queue;
1302 /* We cannot fail after we call this, so another loop!*/
1303 for (i = 0; i < max_loop; i++)
1305 printk(KERN_INFO "loop: loaded (max %d devices)\n", max_loop);
1310 blk_cleanup_queue(loop_dev[i].lo_queue);
1311 devfs_remove("loop");
1320 unregister_blkdev(LOOP_MAJOR, "loop");
1321 printk(KERN_ERR "loop: ran out of memory\n");
1325 static void loop_exit(void)
1329 for (i = 0; i < max_loop; i++) {
1330 del_gendisk(disks[i]);
1331 blk_cleanup_queue(loop_dev[i].lo_queue);
1334 devfs_remove("loop");
1335 if (unregister_blkdev(LOOP_MAJOR, "loop"))
1336 printk(KERN_WARNING "loop: cannot unregister blkdev\n");
1342 module_init(loop_init);
1343 module_exit(loop_exit);
1346 static int __init max_loop_setup(char *str)
1348 max_loop = simple_strtol(str, NULL, 0);
1352 __setup("max_loop=", max_loop_setup);