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/freezer.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>
78 #include <linux/splice.h>
80 #include <asm/uaccess.h>
82 static LIST_HEAD(loop_devices
);
83 static DEFINE_MUTEX(loop_devices_mutex
);
86 static int part_shift
;
91 static int transfer_none(struct loop_device
*lo
, int cmd
,
92 struct page
*raw_page
, unsigned raw_off
,
93 struct page
*loop_page
, unsigned loop_off
,
94 int size
, sector_t real_block
)
96 char *raw_buf
= kmap_atomic(raw_page
, KM_USER0
) + raw_off
;
97 char *loop_buf
= kmap_atomic(loop_page
, KM_USER1
) + loop_off
;
100 memcpy(loop_buf
, raw_buf
, size
);
102 memcpy(raw_buf
, loop_buf
, size
);
104 kunmap_atomic(raw_buf
, KM_USER0
);
105 kunmap_atomic(loop_buf
, KM_USER1
);
110 static int transfer_xor(struct loop_device
*lo
, int cmd
,
111 struct page
*raw_page
, unsigned raw_off
,
112 struct page
*loop_page
, unsigned loop_off
,
113 int size
, sector_t real_block
)
115 char *raw_buf
= kmap_atomic(raw_page
, KM_USER0
) + raw_off
;
116 char *loop_buf
= kmap_atomic(loop_page
, KM_USER1
) + loop_off
;
117 char *in
, *out
, *key
;
128 key
= lo
->lo_encrypt_key
;
129 keysize
= lo
->lo_encrypt_key_size
;
130 for (i
= 0; i
< size
; i
++)
131 *out
++ = *in
++ ^ key
[(i
& 511) % keysize
];
133 kunmap_atomic(raw_buf
, KM_USER0
);
134 kunmap_atomic(loop_buf
, KM_USER1
);
139 static int xor_init(struct loop_device
*lo
, const struct loop_info64
*info
)
141 if (unlikely(info
->lo_encrypt_key_size
<= 0))
146 static struct loop_func_table none_funcs
= {
147 .number
= LO_CRYPT_NONE
,
148 .transfer
= transfer_none
,
151 static struct loop_func_table xor_funcs
= {
152 .number
= LO_CRYPT_XOR
,
153 .transfer
= transfer_xor
,
157 /* xfer_funcs[0] is special - its release function is never called */
158 static struct loop_func_table
*xfer_funcs
[MAX_LO_CRYPT
] = {
163 static loff_t
get_loop_size(struct loop_device
*lo
, struct file
*file
)
165 loff_t size
, offset
, loopsize
;
167 /* Compute loopsize in bytes */
168 size
= i_size_read(file
->f_mapping
->host
);
169 offset
= lo
->lo_offset
;
170 loopsize
= size
- offset
;
171 if (lo
->lo_sizelimit
> 0 && lo
->lo_sizelimit
< loopsize
)
172 loopsize
= lo
->lo_sizelimit
;
175 * Unfortunately, if we want to do I/O on the device,
176 * the number of 512-byte sectors has to fit into a sector_t.
178 return loopsize
>> 9;
182 figure_loop_size(struct loop_device
*lo
)
184 loff_t size
= get_loop_size(lo
, lo
->lo_backing_file
);
185 sector_t x
= (sector_t
)size
;
187 if (unlikely((loff_t
)x
!= size
))
190 set_capacity(lo
->lo_disk
, x
);
195 lo_do_transfer(struct loop_device
*lo
, int cmd
,
196 struct page
*rpage
, unsigned roffs
,
197 struct page
*lpage
, unsigned loffs
,
198 int size
, sector_t rblock
)
200 if (unlikely(!lo
->transfer
))
203 return lo
->transfer(lo
, cmd
, rpage
, roffs
, lpage
, loffs
, size
, rblock
);
207 * do_lo_send_aops - helper for writing data to a loop device
209 * This is the fast version for backing filesystems which implement the address
210 * space operations write_begin and write_end.
212 static int do_lo_send_aops(struct loop_device
*lo
, struct bio_vec
*bvec
,
213 int bsize
, loff_t pos
, struct page
*unused
)
215 struct file
*file
= lo
->lo_backing_file
; /* kudos to NFsckingS */
216 struct address_space
*mapping
= file
->f_mapping
;
218 unsigned offset
, bv_offs
;
221 mutex_lock(&mapping
->host
->i_mutex
);
222 index
= pos
>> PAGE_CACHE_SHIFT
;
223 offset
= pos
& ((pgoff_t
)PAGE_CACHE_SIZE
- 1);
224 bv_offs
= bvec
->bv_offset
;
228 unsigned size
, copied
;
233 IV
= ((sector_t
)index
<< (PAGE_CACHE_SHIFT
- 9))+(offset
>> 9);
234 size
= PAGE_CACHE_SIZE
- offset
;
238 ret
= pagecache_write_begin(file
, mapping
, pos
, size
, 0,
243 transfer_result
= lo_do_transfer(lo
, WRITE
, page
, offset
,
244 bvec
->bv_page
, bv_offs
, size
, IV
);
246 if (unlikely(transfer_result
))
249 ret
= pagecache_write_end(file
, mapping
, pos
, size
, copied
,
251 if (ret
< 0 || ret
!= copied
)
254 if (unlikely(transfer_result
))
265 mutex_unlock(&mapping
->host
->i_mutex
);
273 * __do_lo_send_write - helper for writing data to a loop device
275 * This helper just factors out common code between do_lo_send_direct_write()
276 * and do_lo_send_write().
278 static int __do_lo_send_write(struct file
*file
,
279 u8
*buf
, const int len
, loff_t pos
)
282 mm_segment_t old_fs
= get_fs();
285 bw
= file
->f_op
->write(file
, buf
, len
, &pos
);
287 if (likely(bw
== len
))
289 printk(KERN_ERR
"loop: Write error at byte offset %llu, length %i.\n",
290 (unsigned long long)pos
, len
);
297 * do_lo_send_direct_write - helper for writing data to a loop device
299 * This is the fast, non-transforming version for backing filesystems which do
300 * not implement the address space operations write_begin and write_end.
301 * It uses the write file operation which should be present on all writeable
304 static int do_lo_send_direct_write(struct loop_device
*lo
,
305 struct bio_vec
*bvec
, int bsize
, loff_t pos
, struct page
*page
)
307 ssize_t bw
= __do_lo_send_write(lo
->lo_backing_file
,
308 kmap(bvec
->bv_page
) + bvec
->bv_offset
,
310 kunmap(bvec
->bv_page
);
316 * do_lo_send_write - helper for writing data to a loop device
318 * This is the slow, transforming version for filesystems which do not
319 * implement the address space operations write_begin and write_end. It
320 * uses the write file operation which should be present on all writeable
323 * Using fops->write is slower than using aops->{prepare,commit}_write in the
324 * transforming case because we need to double buffer the data as we cannot do
325 * the transformations in place as we do not have direct access to the
326 * destination pages of the backing file.
328 static int do_lo_send_write(struct loop_device
*lo
, struct bio_vec
*bvec
,
329 int bsize
, loff_t pos
, struct page
*page
)
331 int ret
= lo_do_transfer(lo
, WRITE
, page
, 0, bvec
->bv_page
,
332 bvec
->bv_offset
, bvec
->bv_len
, pos
>> 9);
334 return __do_lo_send_write(lo
->lo_backing_file
,
335 page_address(page
), bvec
->bv_len
,
337 printk(KERN_ERR
"loop: Transfer error at byte offset %llu, "
338 "length %i.\n", (unsigned long long)pos
, bvec
->bv_len
);
344 static int lo_send(struct loop_device
*lo
, struct bio
*bio
, int bsize
,
347 int (*do_lo_send
)(struct loop_device
*, struct bio_vec
*, int, loff_t
,
349 struct bio_vec
*bvec
;
350 struct page
*page
= NULL
;
353 do_lo_send
= do_lo_send_aops
;
354 if (!(lo
->lo_flags
& LO_FLAGS_USE_AOPS
)) {
355 do_lo_send
= do_lo_send_direct_write
;
356 if (lo
->transfer
!= transfer_none
) {
357 page
= alloc_page(GFP_NOIO
| __GFP_HIGHMEM
);
361 do_lo_send
= do_lo_send_write
;
364 bio_for_each_segment(bvec
, bio
, i
) {
365 ret
= do_lo_send(lo
, bvec
, bsize
, pos
, page
);
377 printk(KERN_ERR
"loop: Failed to allocate temporary page for write.\n");
382 struct lo_read_data
{
383 struct loop_device
*lo
;
390 lo_splice_actor(struct pipe_inode_info
*pipe
, struct pipe_buffer
*buf
,
391 struct splice_desc
*sd
)
393 struct lo_read_data
*p
= sd
->u
.data
;
394 struct loop_device
*lo
= p
->lo
;
395 struct page
*page
= buf
->page
;
400 ret
= buf
->ops
->confirm(pipe
, buf
);
404 IV
= ((sector_t
) page
->index
<< (PAGE_CACHE_SHIFT
- 9)) +
410 if (lo_do_transfer(lo
, READ
, page
, buf
->offset
, p
->page
, p
->offset
, size
, IV
)) {
411 printk(KERN_ERR
"loop: transfer error block %ld\n",
416 flush_dcache_page(p
->page
);
425 lo_direct_splice_actor(struct pipe_inode_info
*pipe
, struct splice_desc
*sd
)
427 return __splice_from_pipe(pipe
, sd
, lo_splice_actor
);
431 do_lo_receive(struct loop_device
*lo
,
432 struct bio_vec
*bvec
, int bsize
, loff_t pos
)
434 struct lo_read_data cookie
;
435 struct splice_desc sd
;
440 cookie
.page
= bvec
->bv_page
;
441 cookie
.offset
= bvec
->bv_offset
;
442 cookie
.bsize
= bsize
;
445 sd
.total_len
= bvec
->bv_len
;
450 file
= lo
->lo_backing_file
;
451 retval
= splice_direct_to_actor(file
, &sd
, lo_direct_splice_actor
);
460 lo_receive(struct loop_device
*lo
, struct bio
*bio
, int bsize
, loff_t pos
)
462 struct bio_vec
*bvec
;
465 bio_for_each_segment(bvec
, bio
, i
) {
466 ret
= do_lo_receive(lo
, bvec
, bsize
, pos
);
474 static int do_bio_filebacked(struct loop_device
*lo
, struct bio
*bio
)
479 pos
= ((loff_t
) bio
->bi_sector
<< 9) + lo
->lo_offset
;
480 if (bio_rw(bio
) == WRITE
)
481 ret
= lo_send(lo
, bio
, lo
->lo_blocksize
, pos
);
483 ret
= lo_receive(lo
, bio
, lo
->lo_blocksize
, pos
);
488 * Add bio to back of pending list
490 static void loop_add_bio(struct loop_device
*lo
, struct bio
*bio
)
492 if (lo
->lo_biotail
) {
493 lo
->lo_biotail
->bi_next
= bio
;
494 lo
->lo_biotail
= bio
;
496 lo
->lo_bio
= lo
->lo_biotail
= bio
;
500 * Grab first pending buffer
502 static struct bio
*loop_get_bio(struct loop_device
*lo
)
506 if ((bio
= lo
->lo_bio
)) {
507 if (bio
== lo
->lo_biotail
)
508 lo
->lo_biotail
= NULL
;
509 lo
->lo_bio
= bio
->bi_next
;
516 static int loop_make_request(struct request_queue
*q
, struct bio
*old_bio
)
518 struct loop_device
*lo
= q
->queuedata
;
519 int rw
= bio_rw(old_bio
);
524 BUG_ON(!lo
|| (rw
!= READ
&& rw
!= WRITE
));
526 spin_lock_irq(&lo
->lo_lock
);
527 if (lo
->lo_state
!= Lo_bound
)
529 if (unlikely(rw
== WRITE
&& (lo
->lo_flags
& LO_FLAGS_READ_ONLY
)))
531 loop_add_bio(lo
, old_bio
);
532 wake_up(&lo
->lo_event
);
533 spin_unlock_irq(&lo
->lo_lock
);
537 spin_unlock_irq(&lo
->lo_lock
);
538 bio_io_error(old_bio
);
543 * kick off io on the underlying address space
545 static void loop_unplug(struct request_queue
*q
)
547 struct loop_device
*lo
= q
->queuedata
;
549 queue_flag_clear_unlocked(QUEUE_FLAG_PLUGGED
, q
);
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
);
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 * Loop explanation: loop_clr_fd() sets lo_state to Lo_rundown before
578 * calling kthread_stop(). Therefore once kthread_should_stop() is
579 * true, make_request will not place any more requests. Therefore
580 * once kthread_should_stop() is true and lo_bio is NULL, we are
581 * done with the loop.
583 static int loop_thread(void *data
)
585 struct loop_device
*lo
= data
;
588 set_user_nice(current
, -20);
590 while (!kthread_should_stop() || lo
->lo_bio
) {
592 wait_event_interruptible(lo
->lo_event
,
593 lo
->lo_bio
|| kthread_should_stop());
597 spin_lock_irq(&lo
->lo_lock
);
598 bio
= loop_get_bio(lo
);
599 spin_unlock_irq(&lo
->lo_lock
);
602 loop_handle_bio(lo
, bio
);
609 * loop_switch performs the hard work of switching a backing store.
610 * First it needs to flush existing IO, it does this by sending a magic
611 * BIO down the pipe. The completion of this BIO does the actual switch.
613 static int loop_switch(struct loop_device
*lo
, struct file
*file
)
615 struct switch_request w
;
616 struct bio
*bio
= bio_alloc(GFP_KERNEL
, 0);
619 init_completion(&w
.wait
);
621 bio
->bi_private
= &w
;
623 loop_make_request(lo
->lo_queue
, bio
);
624 wait_for_completion(&w
.wait
);
629 * Do the actual switch; called from the BIO completion routine
631 static void do_loop_switch(struct loop_device
*lo
, struct switch_request
*p
)
633 struct file
*file
= p
->file
;
634 struct file
*old_file
= lo
->lo_backing_file
;
635 struct address_space
*mapping
= file
->f_mapping
;
637 mapping_set_gfp_mask(old_file
->f_mapping
, lo
->old_gfp_mask
);
638 lo
->lo_backing_file
= file
;
639 lo
->lo_blocksize
= S_ISBLK(mapping
->host
->i_mode
) ?
640 mapping
->host
->i_bdev
->bd_block_size
: PAGE_SIZE
;
641 lo
->old_gfp_mask
= mapping_gfp_mask(mapping
);
642 mapping_set_gfp_mask(mapping
, lo
->old_gfp_mask
& ~(__GFP_IO
|__GFP_FS
));
648 * loop_change_fd switched the backing store of a loopback device to
649 * a new file. This is useful for operating system installers to free up
650 * the original file and in High Availability environments to switch to
651 * an alternative location for the content in case of server meltdown.
652 * This can only work if the loop device is used read-only, and if the
653 * new backing store is the same size and type as the old backing store.
655 static int loop_change_fd(struct loop_device
*lo
, struct file
*lo_file
,
656 struct block_device
*bdev
, unsigned int arg
)
658 struct file
*file
, *old_file
;
663 if (lo
->lo_state
!= Lo_bound
)
666 /* the loop device has to be read-only */
668 if (!(lo
->lo_flags
& LO_FLAGS_READ_ONLY
))
676 inode
= file
->f_mapping
->host
;
677 old_file
= lo
->lo_backing_file
;
681 if (!S_ISREG(inode
->i_mode
) && !S_ISBLK(inode
->i_mode
))
684 /* new backing store needs to support loop (eg splice_read) */
685 if (!inode
->i_fop
->splice_read
)
688 /* size of the new backing store needs to be the same */
689 if (get_loop_size(lo
, file
) != get_loop_size(lo
, old_file
))
693 error
= loop_switch(lo
, file
);
699 ioctl_by_bdev(bdev
, BLKRRPART
, 0);
708 static inline int is_loop_device(struct file
*file
)
710 struct inode
*i
= file
->f_mapping
->host
;
712 return i
&& S_ISBLK(i
->i_mode
) && MAJOR(i
->i_rdev
) == LOOP_MAJOR
;
715 static int loop_set_fd(struct loop_device
*lo
, struct file
*lo_file
,
716 struct block_device
*bdev
, unsigned int arg
)
718 struct file
*file
, *f
;
720 struct address_space
*mapping
;
721 unsigned lo_blocksize
;
726 /* This is safe, since we have a reference from open(). */
727 __module_get(THIS_MODULE
);
735 if (lo
->lo_state
!= Lo_unbound
)
738 /* Avoid recursion */
740 while (is_loop_device(f
)) {
741 struct loop_device
*l
;
743 if (f
->f_mapping
->host
->i_rdev
== lo_file
->f_mapping
->host
->i_rdev
)
746 l
= f
->f_mapping
->host
->i_bdev
->bd_disk
->private_data
;
747 if (l
->lo_state
== Lo_unbound
) {
751 f
= l
->lo_backing_file
;
754 mapping
= file
->f_mapping
;
755 inode
= mapping
->host
;
757 if (!(file
->f_mode
& FMODE_WRITE
))
758 lo_flags
|= LO_FLAGS_READ_ONLY
;
761 if (S_ISREG(inode
->i_mode
) || S_ISBLK(inode
->i_mode
)) {
762 const struct address_space_operations
*aops
= mapping
->a_ops
;
764 * If we can't read - sorry. If we only can't write - well,
765 * it's going to be read-only.
767 if (!file
->f_op
->splice_read
)
769 if (aops
->prepare_write
|| aops
->write_begin
)
770 lo_flags
|= LO_FLAGS_USE_AOPS
;
771 if (!(lo_flags
& LO_FLAGS_USE_AOPS
) && !file
->f_op
->write
)
772 lo_flags
|= LO_FLAGS_READ_ONLY
;
774 lo_blocksize
= S_ISBLK(inode
->i_mode
) ?
775 inode
->i_bdev
->bd_block_size
: PAGE_SIZE
;
782 size
= get_loop_size(lo
, file
);
784 if ((loff_t
)(sector_t
)size
!= size
) {
789 if (!(lo_file
->f_mode
& FMODE_WRITE
))
790 lo_flags
|= LO_FLAGS_READ_ONLY
;
792 set_device_ro(bdev
, (lo_flags
& LO_FLAGS_READ_ONLY
) != 0);
794 lo
->lo_blocksize
= lo_blocksize
;
795 lo
->lo_device
= bdev
;
796 lo
->lo_flags
= lo_flags
;
797 lo
->lo_backing_file
= file
;
798 lo
->transfer
= transfer_none
;
800 lo
->lo_sizelimit
= 0;
801 lo
->old_gfp_mask
= mapping_gfp_mask(mapping
);
802 mapping_set_gfp_mask(mapping
, lo
->old_gfp_mask
& ~(__GFP_IO
|__GFP_FS
));
804 lo
->lo_bio
= lo
->lo_biotail
= NULL
;
807 * set queue make_request_fn, and add limits based on lower level
810 blk_queue_make_request(lo
->lo_queue
, loop_make_request
);
811 lo
->lo_queue
->queuedata
= lo
;
812 lo
->lo_queue
->unplug_fn
= loop_unplug
;
814 set_capacity(lo
->lo_disk
, size
);
815 bd_set_size(bdev
, size
<< 9);
817 set_blocksize(bdev
, lo_blocksize
);
819 lo
->lo_thread
= kthread_create(loop_thread
, lo
, "loop%d",
821 if (IS_ERR(lo
->lo_thread
)) {
822 error
= PTR_ERR(lo
->lo_thread
);
825 lo
->lo_state
= Lo_bound
;
826 wake_up_process(lo
->lo_thread
);
828 ioctl_by_bdev(bdev
, BLKRRPART
, 0);
832 lo
->lo_thread
= NULL
;
833 lo
->lo_device
= NULL
;
834 lo
->lo_backing_file
= NULL
;
836 set_capacity(lo
->lo_disk
, 0);
837 invalidate_bdev(bdev
);
838 bd_set_size(bdev
, 0);
839 mapping_set_gfp_mask(mapping
, lo
->old_gfp_mask
);
840 lo
->lo_state
= Lo_unbound
;
844 /* This is safe: open() is still holding a reference. */
845 module_put(THIS_MODULE
);
850 loop_release_xfer(struct loop_device
*lo
)
853 struct loop_func_table
*xfer
= lo
->lo_encryption
;
857 err
= xfer
->release(lo
);
859 lo
->lo_encryption
= NULL
;
860 module_put(xfer
->owner
);
866 loop_init_xfer(struct loop_device
*lo
, struct loop_func_table
*xfer
,
867 const struct loop_info64
*i
)
872 struct module
*owner
= xfer
->owner
;
874 if (!try_module_get(owner
))
877 err
= xfer
->init(lo
, i
);
881 lo
->lo_encryption
= xfer
;
886 static int loop_clr_fd(struct loop_device
*lo
, struct block_device
*bdev
)
888 struct file
*filp
= lo
->lo_backing_file
;
889 gfp_t gfp
= lo
->old_gfp_mask
;
891 if (lo
->lo_state
!= Lo_bound
)
894 if (lo
->lo_refcnt
> 1) /* we needed one fd for the ioctl */
900 spin_lock_irq(&lo
->lo_lock
);
901 lo
->lo_state
= Lo_rundown
;
902 spin_unlock_irq(&lo
->lo_lock
);
904 kthread_stop(lo
->lo_thread
);
906 lo
->lo_backing_file
= NULL
;
908 loop_release_xfer(lo
);
911 lo
->lo_device
= NULL
;
912 lo
->lo_encryption
= NULL
;
914 lo
->lo_sizelimit
= 0;
915 lo
->lo_encrypt_key_size
= 0;
917 lo
->lo_thread
= NULL
;
918 memset(lo
->lo_encrypt_key
, 0, LO_KEY_SIZE
);
919 memset(lo
->lo_crypt_name
, 0, LO_NAME_SIZE
);
920 memset(lo
->lo_file_name
, 0, LO_NAME_SIZE
);
921 invalidate_bdev(bdev
);
922 set_capacity(lo
->lo_disk
, 0);
923 bd_set_size(bdev
, 0);
924 mapping_set_gfp_mask(filp
->f_mapping
, gfp
);
925 lo
->lo_state
= Lo_unbound
;
927 /* This is safe: open() is still holding a reference. */
928 module_put(THIS_MODULE
);
930 ioctl_by_bdev(bdev
, BLKRRPART
, 0);
935 loop_set_status(struct loop_device
*lo
, const struct loop_info64
*info
)
938 struct loop_func_table
*xfer
;
940 if (lo
->lo_encrypt_key_size
&& lo
->lo_key_owner
!= current
->uid
&&
941 !capable(CAP_SYS_ADMIN
))
943 if (lo
->lo_state
!= Lo_bound
)
945 if ((unsigned int) info
->lo_encrypt_key_size
> LO_KEY_SIZE
)
948 err
= loop_release_xfer(lo
);
952 if (info
->lo_encrypt_type
) {
953 unsigned int type
= info
->lo_encrypt_type
;
955 if (type
>= MAX_LO_CRYPT
)
957 xfer
= xfer_funcs
[type
];
963 err
= loop_init_xfer(lo
, xfer
, info
);
967 if (lo
->lo_offset
!= info
->lo_offset
||
968 lo
->lo_sizelimit
!= info
->lo_sizelimit
) {
969 lo
->lo_offset
= info
->lo_offset
;
970 lo
->lo_sizelimit
= info
->lo_sizelimit
;
971 if (figure_loop_size(lo
))
975 memcpy(lo
->lo_file_name
, info
->lo_file_name
, LO_NAME_SIZE
);
976 memcpy(lo
->lo_crypt_name
, info
->lo_crypt_name
, LO_NAME_SIZE
);
977 lo
->lo_file_name
[LO_NAME_SIZE
-1] = 0;
978 lo
->lo_crypt_name
[LO_NAME_SIZE
-1] = 0;
982 lo
->transfer
= xfer
->transfer
;
983 lo
->ioctl
= xfer
->ioctl
;
985 if ((lo
->lo_flags
& LO_FLAGS_AUTOCLEAR
) !=
986 (info
->lo_flags
& LO_FLAGS_AUTOCLEAR
))
987 lo
->lo_flags
^= LO_FLAGS_AUTOCLEAR
;
989 lo
->lo_encrypt_key_size
= info
->lo_encrypt_key_size
;
990 lo
->lo_init
[0] = info
->lo_init
[0];
991 lo
->lo_init
[1] = info
->lo_init
[1];
992 if (info
->lo_encrypt_key_size
) {
993 memcpy(lo
->lo_encrypt_key
, info
->lo_encrypt_key
,
994 info
->lo_encrypt_key_size
);
995 lo
->lo_key_owner
= current
->uid
;
1002 loop_get_status(struct loop_device
*lo
, struct loop_info64
*info
)
1004 struct file
*file
= lo
->lo_backing_file
;
1008 if (lo
->lo_state
!= Lo_bound
)
1010 error
= vfs_getattr(file
->f_path
.mnt
, file
->f_path
.dentry
, &stat
);
1013 memset(info
, 0, sizeof(*info
));
1014 info
->lo_number
= lo
->lo_number
;
1015 info
->lo_device
= huge_encode_dev(stat
.dev
);
1016 info
->lo_inode
= stat
.ino
;
1017 info
->lo_rdevice
= huge_encode_dev(lo
->lo_device
? stat
.rdev
: stat
.dev
);
1018 info
->lo_offset
= lo
->lo_offset
;
1019 info
->lo_sizelimit
= lo
->lo_sizelimit
;
1020 info
->lo_flags
= lo
->lo_flags
;
1021 memcpy(info
->lo_file_name
, lo
->lo_file_name
, LO_NAME_SIZE
);
1022 memcpy(info
->lo_crypt_name
, lo
->lo_crypt_name
, LO_NAME_SIZE
);
1023 info
->lo_encrypt_type
=
1024 lo
->lo_encryption
? lo
->lo_encryption
->number
: 0;
1025 if (lo
->lo_encrypt_key_size
&& capable(CAP_SYS_ADMIN
)) {
1026 info
->lo_encrypt_key_size
= lo
->lo_encrypt_key_size
;
1027 memcpy(info
->lo_encrypt_key
, lo
->lo_encrypt_key
,
1028 lo
->lo_encrypt_key_size
);
1034 loop_info64_from_old(const struct loop_info
*info
, struct loop_info64
*info64
)
1036 memset(info64
, 0, sizeof(*info64
));
1037 info64
->lo_number
= info
->lo_number
;
1038 info64
->lo_device
= info
->lo_device
;
1039 info64
->lo_inode
= info
->lo_inode
;
1040 info64
->lo_rdevice
= info
->lo_rdevice
;
1041 info64
->lo_offset
= info
->lo_offset
;
1042 info64
->lo_sizelimit
= 0;
1043 info64
->lo_encrypt_type
= info
->lo_encrypt_type
;
1044 info64
->lo_encrypt_key_size
= info
->lo_encrypt_key_size
;
1045 info64
->lo_flags
= info
->lo_flags
;
1046 info64
->lo_init
[0] = info
->lo_init
[0];
1047 info64
->lo_init
[1] = info
->lo_init
[1];
1048 if (info
->lo_encrypt_type
== LO_CRYPT_CRYPTOAPI
)
1049 memcpy(info64
->lo_crypt_name
, info
->lo_name
, LO_NAME_SIZE
);
1051 memcpy(info64
->lo_file_name
, info
->lo_name
, LO_NAME_SIZE
);
1052 memcpy(info64
->lo_encrypt_key
, info
->lo_encrypt_key
, LO_KEY_SIZE
);
1056 loop_info64_to_old(const struct loop_info64
*info64
, struct loop_info
*info
)
1058 memset(info
, 0, sizeof(*info
));
1059 info
->lo_number
= info64
->lo_number
;
1060 info
->lo_device
= info64
->lo_device
;
1061 info
->lo_inode
= info64
->lo_inode
;
1062 info
->lo_rdevice
= info64
->lo_rdevice
;
1063 info
->lo_offset
= info64
->lo_offset
;
1064 info
->lo_encrypt_type
= info64
->lo_encrypt_type
;
1065 info
->lo_encrypt_key_size
= info64
->lo_encrypt_key_size
;
1066 info
->lo_flags
= info64
->lo_flags
;
1067 info
->lo_init
[0] = info64
->lo_init
[0];
1068 info
->lo_init
[1] = info64
->lo_init
[1];
1069 if (info
->lo_encrypt_type
== LO_CRYPT_CRYPTOAPI
)
1070 memcpy(info
->lo_name
, info64
->lo_crypt_name
, LO_NAME_SIZE
);
1072 memcpy(info
->lo_name
, info64
->lo_file_name
, LO_NAME_SIZE
);
1073 memcpy(info
->lo_encrypt_key
, info64
->lo_encrypt_key
, LO_KEY_SIZE
);
1075 /* error in case values were truncated */
1076 if (info
->lo_device
!= info64
->lo_device
||
1077 info
->lo_rdevice
!= info64
->lo_rdevice
||
1078 info
->lo_inode
!= info64
->lo_inode
||
1079 info
->lo_offset
!= info64
->lo_offset
)
1086 loop_set_status_old(struct loop_device
*lo
, const struct loop_info __user
*arg
)
1088 struct loop_info info
;
1089 struct loop_info64 info64
;
1091 if (copy_from_user(&info
, arg
, sizeof (struct loop_info
)))
1093 loop_info64_from_old(&info
, &info64
);
1094 return loop_set_status(lo
, &info64
);
1098 loop_set_status64(struct loop_device
*lo
, const struct loop_info64 __user
*arg
)
1100 struct loop_info64 info64
;
1102 if (copy_from_user(&info64
, arg
, sizeof (struct loop_info64
)))
1104 return loop_set_status(lo
, &info64
);
1108 loop_get_status_old(struct loop_device
*lo
, struct loop_info __user
*arg
) {
1109 struct loop_info info
;
1110 struct loop_info64 info64
;
1116 err
= loop_get_status(lo
, &info64
);
1118 err
= loop_info64_to_old(&info64
, &info
);
1119 if (!err
&& copy_to_user(arg
, &info
, sizeof(info
)))
1126 loop_get_status64(struct loop_device
*lo
, struct loop_info64 __user
*arg
) {
1127 struct loop_info64 info64
;
1133 err
= loop_get_status(lo
, &info64
);
1134 if (!err
&& copy_to_user(arg
, &info64
, sizeof(info64
)))
1140 static int lo_ioctl(struct inode
* inode
, struct file
* file
,
1141 unsigned int cmd
, unsigned long arg
)
1143 struct loop_device
*lo
= inode
->i_bdev
->bd_disk
->private_data
;
1146 mutex_lock(&lo
->lo_ctl_mutex
);
1149 err
= loop_set_fd(lo
, file
, inode
->i_bdev
, arg
);
1151 case LOOP_CHANGE_FD
:
1152 err
= loop_change_fd(lo
, file
, inode
->i_bdev
, arg
);
1155 err
= loop_clr_fd(lo
, inode
->i_bdev
);
1157 case LOOP_SET_STATUS
:
1158 err
= loop_set_status_old(lo
, (struct loop_info __user
*) arg
);
1160 case LOOP_GET_STATUS
:
1161 err
= loop_get_status_old(lo
, (struct loop_info __user
*) arg
);
1163 case LOOP_SET_STATUS64
:
1164 err
= loop_set_status64(lo
, (struct loop_info64 __user
*) arg
);
1166 case LOOP_GET_STATUS64
:
1167 err
= loop_get_status64(lo
, (struct loop_info64 __user
*) arg
);
1170 err
= lo
->ioctl
? lo
->ioctl(lo
, cmd
, arg
) : -EINVAL
;
1172 mutex_unlock(&lo
->lo_ctl_mutex
);
1176 #ifdef CONFIG_COMPAT
1177 struct compat_loop_info
{
1178 compat_int_t lo_number
; /* ioctl r/o */
1179 compat_dev_t lo_device
; /* ioctl r/o */
1180 compat_ulong_t lo_inode
; /* ioctl r/o */
1181 compat_dev_t lo_rdevice
; /* ioctl r/o */
1182 compat_int_t lo_offset
;
1183 compat_int_t lo_encrypt_type
;
1184 compat_int_t lo_encrypt_key_size
; /* ioctl w/o */
1185 compat_int_t lo_flags
; /* ioctl r/o */
1186 char lo_name
[LO_NAME_SIZE
];
1187 unsigned char lo_encrypt_key
[LO_KEY_SIZE
]; /* ioctl w/o */
1188 compat_ulong_t lo_init
[2];
1193 * Transfer 32-bit compatibility structure in userspace to 64-bit loop info
1194 * - noinlined to reduce stack space usage in main part of driver
1197 loop_info64_from_compat(const struct compat_loop_info __user
*arg
,
1198 struct loop_info64
*info64
)
1200 struct compat_loop_info info
;
1202 if (copy_from_user(&info
, arg
, sizeof(info
)))
1205 memset(info64
, 0, sizeof(*info64
));
1206 info64
->lo_number
= info
.lo_number
;
1207 info64
->lo_device
= info
.lo_device
;
1208 info64
->lo_inode
= info
.lo_inode
;
1209 info64
->lo_rdevice
= info
.lo_rdevice
;
1210 info64
->lo_offset
= info
.lo_offset
;
1211 info64
->lo_sizelimit
= 0;
1212 info64
->lo_encrypt_type
= info
.lo_encrypt_type
;
1213 info64
->lo_encrypt_key_size
= info
.lo_encrypt_key_size
;
1214 info64
->lo_flags
= info
.lo_flags
;
1215 info64
->lo_init
[0] = info
.lo_init
[0];
1216 info64
->lo_init
[1] = info
.lo_init
[1];
1217 if (info
.lo_encrypt_type
== LO_CRYPT_CRYPTOAPI
)
1218 memcpy(info64
->lo_crypt_name
, info
.lo_name
, LO_NAME_SIZE
);
1220 memcpy(info64
->lo_file_name
, info
.lo_name
, LO_NAME_SIZE
);
1221 memcpy(info64
->lo_encrypt_key
, info
.lo_encrypt_key
, LO_KEY_SIZE
);
1226 * Transfer 64-bit loop info to 32-bit compatibility structure in userspace
1227 * - noinlined to reduce stack space usage in main part of driver
1230 loop_info64_to_compat(const struct loop_info64
*info64
,
1231 struct compat_loop_info __user
*arg
)
1233 struct compat_loop_info info
;
1235 memset(&info
, 0, sizeof(info
));
1236 info
.lo_number
= info64
->lo_number
;
1237 info
.lo_device
= info64
->lo_device
;
1238 info
.lo_inode
= info64
->lo_inode
;
1239 info
.lo_rdevice
= info64
->lo_rdevice
;
1240 info
.lo_offset
= info64
->lo_offset
;
1241 info
.lo_encrypt_type
= info64
->lo_encrypt_type
;
1242 info
.lo_encrypt_key_size
= info64
->lo_encrypt_key_size
;
1243 info
.lo_flags
= info64
->lo_flags
;
1244 info
.lo_init
[0] = info64
->lo_init
[0];
1245 info
.lo_init
[1] = info64
->lo_init
[1];
1246 if (info
.lo_encrypt_type
== LO_CRYPT_CRYPTOAPI
)
1247 memcpy(info
.lo_name
, info64
->lo_crypt_name
, LO_NAME_SIZE
);
1249 memcpy(info
.lo_name
, info64
->lo_file_name
, LO_NAME_SIZE
);
1250 memcpy(info
.lo_encrypt_key
, info64
->lo_encrypt_key
, LO_KEY_SIZE
);
1252 /* error in case values were truncated */
1253 if (info
.lo_device
!= info64
->lo_device
||
1254 info
.lo_rdevice
!= info64
->lo_rdevice
||
1255 info
.lo_inode
!= info64
->lo_inode
||
1256 info
.lo_offset
!= info64
->lo_offset
||
1257 info
.lo_init
[0] != info64
->lo_init
[0] ||
1258 info
.lo_init
[1] != info64
->lo_init
[1])
1261 if (copy_to_user(arg
, &info
, sizeof(info
)))
1267 loop_set_status_compat(struct loop_device
*lo
,
1268 const struct compat_loop_info __user
*arg
)
1270 struct loop_info64 info64
;
1273 ret
= loop_info64_from_compat(arg
, &info64
);
1276 return loop_set_status(lo
, &info64
);
1280 loop_get_status_compat(struct loop_device
*lo
,
1281 struct compat_loop_info __user
*arg
)
1283 struct loop_info64 info64
;
1289 err
= loop_get_status(lo
, &info64
);
1291 err
= loop_info64_to_compat(&info64
, arg
);
1295 static long lo_compat_ioctl(struct file
*file
, unsigned int cmd
, unsigned long arg
)
1297 struct inode
*inode
= file
->f_path
.dentry
->d_inode
;
1298 struct loop_device
*lo
= inode
->i_bdev
->bd_disk
->private_data
;
1302 case LOOP_SET_STATUS
:
1303 mutex_lock(&lo
->lo_ctl_mutex
);
1304 err
= loop_set_status_compat(
1305 lo
, (const struct compat_loop_info __user
*) arg
);
1306 mutex_unlock(&lo
->lo_ctl_mutex
);
1308 case LOOP_GET_STATUS
:
1309 mutex_lock(&lo
->lo_ctl_mutex
);
1310 err
= loop_get_status_compat(
1311 lo
, (struct compat_loop_info __user
*) arg
);
1312 mutex_unlock(&lo
->lo_ctl_mutex
);
1315 case LOOP_GET_STATUS64
:
1316 case LOOP_SET_STATUS64
:
1317 arg
= (unsigned long) compat_ptr(arg
);
1319 case LOOP_CHANGE_FD
:
1320 err
= lo_ioctl(inode
, file
, cmd
, arg
);
1330 static int lo_open(struct inode
*inode
, struct file
*file
)
1332 struct loop_device
*lo
= inode
->i_bdev
->bd_disk
->private_data
;
1334 mutex_lock(&lo
->lo_ctl_mutex
);
1336 mutex_unlock(&lo
->lo_ctl_mutex
);
1341 static int lo_release(struct inode
*inode
, struct file
*file
)
1343 struct loop_device
*lo
= inode
->i_bdev
->bd_disk
->private_data
;
1345 mutex_lock(&lo
->lo_ctl_mutex
);
1348 if ((lo
->lo_flags
& LO_FLAGS_AUTOCLEAR
) && !lo
->lo_refcnt
)
1349 loop_clr_fd(lo
, inode
->i_bdev
);
1351 mutex_unlock(&lo
->lo_ctl_mutex
);
1356 static struct block_device_operations lo_fops
= {
1357 .owner
= THIS_MODULE
,
1359 .release
= lo_release
,
1361 #ifdef CONFIG_COMPAT
1362 .compat_ioctl
= lo_compat_ioctl
,
1367 * And now the modules code and kernel interface.
1369 static int max_loop
;
1370 module_param(max_loop
, int, 0);
1371 MODULE_PARM_DESC(max_loop
, "Maximum number of loop devices");
1372 module_param(max_part
, int, 0);
1373 MODULE_PARM_DESC(max_part
, "Maximum number of partitions per loop device");
1374 MODULE_LICENSE("GPL");
1375 MODULE_ALIAS_BLOCKDEV_MAJOR(LOOP_MAJOR
);
1377 int loop_register_transfer(struct loop_func_table
*funcs
)
1379 unsigned int n
= funcs
->number
;
1381 if (n
>= MAX_LO_CRYPT
|| xfer_funcs
[n
])
1383 xfer_funcs
[n
] = funcs
;
1387 int loop_unregister_transfer(int number
)
1389 unsigned int n
= number
;
1390 struct loop_device
*lo
;
1391 struct loop_func_table
*xfer
;
1393 if (n
== 0 || n
>= MAX_LO_CRYPT
|| (xfer
= xfer_funcs
[n
]) == NULL
)
1396 xfer_funcs
[n
] = NULL
;
1398 list_for_each_entry(lo
, &loop_devices
, lo_list
) {
1399 mutex_lock(&lo
->lo_ctl_mutex
);
1401 if (lo
->lo_encryption
== xfer
)
1402 loop_release_xfer(lo
);
1404 mutex_unlock(&lo
->lo_ctl_mutex
);
1410 EXPORT_SYMBOL(loop_register_transfer
);
1411 EXPORT_SYMBOL(loop_unregister_transfer
);
1413 static struct loop_device
*loop_alloc(int i
)
1415 struct loop_device
*lo
;
1416 struct gendisk
*disk
;
1418 lo
= kzalloc(sizeof(*lo
), GFP_KERNEL
);
1422 lo
->lo_queue
= blk_alloc_queue(GFP_KERNEL
);
1426 disk
= lo
->lo_disk
= alloc_disk(1 << part_shift
);
1428 goto out_free_queue
;
1430 mutex_init(&lo
->lo_ctl_mutex
);
1432 lo
->lo_thread
= NULL
;
1433 init_waitqueue_head(&lo
->lo_event
);
1434 spin_lock_init(&lo
->lo_lock
);
1435 disk
->major
= LOOP_MAJOR
;
1436 disk
->first_minor
= i
<< part_shift
;
1437 disk
->fops
= &lo_fops
;
1438 disk
->private_data
= lo
;
1439 disk
->queue
= lo
->lo_queue
;
1440 sprintf(disk
->disk_name
, "loop%d", i
);
1444 blk_cleanup_queue(lo
->lo_queue
);
1451 static void loop_free(struct loop_device
*lo
)
1453 blk_cleanup_queue(lo
->lo_queue
);
1454 put_disk(lo
->lo_disk
);
1455 list_del(&lo
->lo_list
);
1459 static struct loop_device
*loop_init_one(int i
)
1461 struct loop_device
*lo
;
1463 list_for_each_entry(lo
, &loop_devices
, lo_list
) {
1464 if (lo
->lo_number
== i
)
1470 add_disk(lo
->lo_disk
);
1471 list_add_tail(&lo
->lo_list
, &loop_devices
);
1476 static void loop_del_one(struct loop_device
*lo
)
1478 del_gendisk(lo
->lo_disk
);
1482 static struct kobject
*loop_probe(dev_t dev
, int *part
, void *data
)
1484 struct loop_device
*lo
;
1485 struct kobject
*kobj
;
1487 mutex_lock(&loop_devices_mutex
);
1488 lo
= loop_init_one(dev
& MINORMASK
);
1489 kobj
= lo
? get_disk(lo
->lo_disk
) : ERR_PTR(-ENOMEM
);
1490 mutex_unlock(&loop_devices_mutex
);
1496 static int __init
loop_init(void)
1499 unsigned long range
;
1500 struct loop_device
*lo
, *next
;
1503 * loop module now has a feature to instantiate underlying device
1504 * structure on-demand, provided that there is an access dev node.
1505 * However, this will not work well with user space tool that doesn't
1506 * know about such "feature". In order to not break any existing
1507 * tool, we do the following:
1509 * (1) if max_loop is specified, create that many upfront, and this
1510 * also becomes a hard limit.
1511 * (2) if max_loop is not specified, create 8 loop device on module
1512 * load, user can further extend loop device by create dev node
1513 * themselves and have kernel automatically instantiate actual
1519 part_shift
= fls(max_part
);
1521 if (max_loop
> 1UL << (MINORBITS
- part_shift
))
1529 range
= 1UL << (MINORBITS
- part_shift
);
1532 if (register_blkdev(LOOP_MAJOR
, "loop"))
1535 for (i
= 0; i
< nr
; i
++) {
1539 list_add_tail(&lo
->lo_list
, &loop_devices
);
1542 /* point of no return */
1544 list_for_each_entry(lo
, &loop_devices
, lo_list
)
1545 add_disk(lo
->lo_disk
);
1547 blk_register_region(MKDEV(LOOP_MAJOR
, 0), range
,
1548 THIS_MODULE
, loop_probe
, NULL
, NULL
);
1550 printk(KERN_INFO
"loop: module loaded\n");
1554 printk(KERN_INFO
"loop: out of memory\n");
1556 list_for_each_entry_safe(lo
, next
, &loop_devices
, lo_list
)
1559 unregister_blkdev(LOOP_MAJOR
, "loop");
1563 static void __exit
loop_exit(void)
1565 unsigned long range
;
1566 struct loop_device
*lo
, *next
;
1568 range
= max_loop
? max_loop
: 1UL << (MINORBITS
- part_shift
);
1570 list_for_each_entry_safe(lo
, next
, &loop_devices
, lo_list
)
1573 blk_unregister_region(MKDEV(LOOP_MAJOR
, 0), range
);
1574 unregister_blkdev(LOOP_MAJOR
, "loop");
1577 module_init(loop_init
);
1578 module_exit(loop_exit
);
1581 static int __init
max_loop_setup(char *str
)
1583 max_loop
= simple_strtol(str
, NULL
, 0);
1587 __setup("max_loop=", max_loop_setup
);