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
);
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(lo
->lo_disk
, 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 write_begin and write_end.
209 static int do_lo_send_aops(struct loop_device
*lo
, struct bio_vec
*bvec
,
210 int bsize
, loff_t pos
, struct page
*unused
)
212 struct file
*file
= lo
->lo_backing_file
; /* kudos to NFsckingS */
213 struct address_space
*mapping
= file
->f_mapping
;
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
;
225 unsigned size
, copied
;
230 IV
= ((sector_t
)index
<< (PAGE_CACHE_SHIFT
- 9))+(offset
>> 9);
231 size
= PAGE_CACHE_SIZE
- offset
;
235 ret
= pagecache_write_begin(file
, mapping
, pos
, size
, 0,
240 transfer_result
= lo_do_transfer(lo
, WRITE
, page
, offset
,
241 bvec
->bv_page
, bv_offs
, size
, IV
);
243 if (unlikely(transfer_result
))
246 ret
= pagecache_write_end(file
, mapping
, pos
, size
, copied
,
248 if (ret
< 0 || ret
!= copied
)
251 if (unlikely(transfer_result
))
262 mutex_unlock(&mapping
->host
->i_mutex
);
270 * __do_lo_send_write - helper for writing data to a loop device
272 * This helper just factors out common code between do_lo_send_direct_write()
273 * and do_lo_send_write().
275 static int __do_lo_send_write(struct file
*file
,
276 u8
*buf
, const int len
, loff_t pos
)
279 mm_segment_t old_fs
= get_fs();
282 bw
= file
->f_op
->write(file
, buf
, len
, &pos
);
284 if (likely(bw
== len
))
286 printk(KERN_ERR
"loop: Write error at byte offset %llu, length %i.\n",
287 (unsigned long long)pos
, len
);
294 * do_lo_send_direct_write - helper for writing data to a loop device
296 * This is the fast, non-transforming version for backing filesystems which do
297 * not implement the address space operations write_begin and write_end.
298 * It uses the write file operation which should be present on all writeable
301 static int do_lo_send_direct_write(struct loop_device
*lo
,
302 struct bio_vec
*bvec
, int bsize
, loff_t pos
, struct page
*page
)
304 ssize_t bw
= __do_lo_send_write(lo
->lo_backing_file
,
305 kmap(bvec
->bv_page
) + bvec
->bv_offset
,
307 kunmap(bvec
->bv_page
);
313 * do_lo_send_write - helper for writing data to a loop device
315 * This is the slow, transforming version for filesystems which do not
316 * implement the address space operations write_begin and write_end. It
317 * uses the write file operation which should be present on all writeable
320 * Using fops->write is slower than using aops->{prepare,commit}_write in the
321 * transforming case because we need to double buffer the data as we cannot do
322 * the transformations in place as we do not have direct access to the
323 * destination pages of the backing file.
325 static int do_lo_send_write(struct loop_device
*lo
, struct bio_vec
*bvec
,
326 int bsize
, loff_t pos
, struct page
*page
)
328 int ret
= lo_do_transfer(lo
, WRITE
, page
, 0, bvec
->bv_page
,
329 bvec
->bv_offset
, bvec
->bv_len
, pos
>> 9);
331 return __do_lo_send_write(lo
->lo_backing_file
,
332 page_address(page
), bvec
->bv_len
,
334 printk(KERN_ERR
"loop: Transfer error at byte offset %llu, "
335 "length %i.\n", (unsigned long long)pos
, bvec
->bv_len
);
341 static int lo_send(struct loop_device
*lo
, struct bio
*bio
, int bsize
,
344 int (*do_lo_send
)(struct loop_device
*, struct bio_vec
*, int, loff_t
,
346 struct bio_vec
*bvec
;
347 struct page
*page
= NULL
;
350 do_lo_send
= do_lo_send_aops
;
351 if (!(lo
->lo_flags
& LO_FLAGS_USE_AOPS
)) {
352 do_lo_send
= do_lo_send_direct_write
;
353 if (lo
->transfer
!= transfer_none
) {
354 page
= alloc_page(GFP_NOIO
| __GFP_HIGHMEM
);
358 do_lo_send
= do_lo_send_write
;
361 bio_for_each_segment(bvec
, bio
, i
) {
362 ret
= do_lo_send(lo
, bvec
, bsize
, pos
, page
);
374 printk(KERN_ERR
"loop: Failed to allocate temporary page for write.\n");
379 struct lo_read_data
{
380 struct loop_device
*lo
;
387 lo_splice_actor(struct pipe_inode_info
*pipe
, struct pipe_buffer
*buf
,
388 struct splice_desc
*sd
)
390 struct lo_read_data
*p
= sd
->u
.data
;
391 struct loop_device
*lo
= p
->lo
;
392 struct page
*page
= buf
->page
;
397 ret
= buf
->ops
->confirm(pipe
, buf
);
401 IV
= ((sector_t
) page
->index
<< (PAGE_CACHE_SHIFT
- 9)) +
407 if (lo_do_transfer(lo
, READ
, page
, buf
->offset
, p
->page
, p
->offset
, size
, IV
)) {
408 printk(KERN_ERR
"loop: transfer error block %ld\n",
413 flush_dcache_page(p
->page
);
422 lo_direct_splice_actor(struct pipe_inode_info
*pipe
, struct splice_desc
*sd
)
424 return __splice_from_pipe(pipe
, sd
, lo_splice_actor
);
428 do_lo_receive(struct loop_device
*lo
,
429 struct bio_vec
*bvec
, int bsize
, loff_t pos
)
431 struct lo_read_data cookie
;
432 struct splice_desc sd
;
437 cookie
.page
= bvec
->bv_page
;
438 cookie
.offset
= bvec
->bv_offset
;
439 cookie
.bsize
= bsize
;
442 sd
.total_len
= bvec
->bv_len
;
447 file
= lo
->lo_backing_file
;
448 retval
= splice_direct_to_actor(file
, &sd
, lo_direct_splice_actor
);
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(struct request_queue
*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
)))
528 loop_add_bio(lo
, old_bio
);
529 wake_up(&lo
->lo_event
);
530 spin_unlock_irq(&lo
->lo_lock
);
534 spin_unlock_irq(&lo
->lo_lock
);
535 bio_io_error(old_bio
);
540 * kick off io on the underlying address space
542 static void loop_unplug(struct request_queue
*q
)
544 struct loop_device
*lo
= q
->queuedata
;
546 clear_bit(QUEUE_FLAG_PLUGGED
, &q
->queue_flags
);
547 blk_run_address_space(lo
->lo_backing_file
->f_mapping
);
550 struct switch_request
{
552 struct completion wait
;
555 static void do_loop_switch(struct loop_device
*, struct switch_request
*);
557 static inline void loop_handle_bio(struct loop_device
*lo
, struct bio
*bio
)
559 if (unlikely(!bio
->bi_bdev
)) {
560 do_loop_switch(lo
, bio
->bi_private
);
563 int ret
= do_bio_filebacked(lo
, bio
);
569 * worker thread that handles reads/writes to file backed loop devices,
570 * to avoid blocking in our make_request_fn. it also does loop decrypting
571 * on reads for block backed loop, as that is too heavy to do from
572 * b_end_io context where irqs may be disabled.
574 * Loop explanation: loop_clr_fd() sets lo_state to Lo_rundown before
575 * calling kthread_stop(). Therefore once kthread_should_stop() is
576 * true, make_request will not place any more requests. Therefore
577 * once kthread_should_stop() is true and lo_bio is NULL, we are
578 * done with the loop.
580 static int loop_thread(void *data
)
582 struct loop_device
*lo
= data
;
585 set_user_nice(current
, -20);
587 while (!kthread_should_stop() || lo
->lo_bio
) {
589 wait_event_interruptible(lo
->lo_event
,
590 lo
->lo_bio
|| kthread_should_stop());
594 spin_lock_irq(&lo
->lo_lock
);
595 bio
= loop_get_bio(lo
);
596 spin_unlock_irq(&lo
->lo_lock
);
599 loop_handle_bio(lo
, bio
);
606 * loop_switch performs the hard work of switching a backing store.
607 * First it needs to flush existing IO, it does this by sending a magic
608 * BIO down the pipe. The completion of this BIO does the actual switch.
610 static int loop_switch(struct loop_device
*lo
, struct file
*file
)
612 struct switch_request w
;
613 struct bio
*bio
= bio_alloc(GFP_KERNEL
, 0);
616 init_completion(&w
.wait
);
618 bio
->bi_private
= &w
;
620 loop_make_request(lo
->lo_queue
, bio
);
621 wait_for_completion(&w
.wait
);
626 * Do the actual switch; called from the BIO completion routine
628 static void do_loop_switch(struct loop_device
*lo
, struct switch_request
*p
)
630 struct file
*file
= p
->file
;
631 struct file
*old_file
= lo
->lo_backing_file
;
632 struct address_space
*mapping
= file
->f_mapping
;
634 mapping_set_gfp_mask(old_file
->f_mapping
, lo
->old_gfp_mask
);
635 lo
->lo_backing_file
= file
;
636 lo
->lo_blocksize
= S_ISBLK(mapping
->host
->i_mode
) ?
637 mapping
->host
->i_bdev
->bd_block_size
: PAGE_SIZE
;
638 lo
->old_gfp_mask
= mapping_gfp_mask(mapping
);
639 mapping_set_gfp_mask(mapping
, lo
->old_gfp_mask
& ~(__GFP_IO
|__GFP_FS
));
645 * loop_change_fd switched the backing store of a loopback device to
646 * a new file. This is useful for operating system installers to free up
647 * the original file and in High Availability environments to switch to
648 * an alternative location for the content in case of server meltdown.
649 * This can only work if the loop device is used read-only, and if the
650 * new backing store is the same size and type as the old backing store.
652 static int loop_change_fd(struct loop_device
*lo
, struct file
*lo_file
,
653 struct block_device
*bdev
, unsigned int arg
)
655 struct file
*file
, *old_file
;
660 if (lo
->lo_state
!= Lo_bound
)
663 /* the loop device has to be read-only */
665 if (!(lo
->lo_flags
& LO_FLAGS_READ_ONLY
))
673 inode
= file
->f_mapping
->host
;
674 old_file
= lo
->lo_backing_file
;
678 if (!S_ISREG(inode
->i_mode
) && !S_ISBLK(inode
->i_mode
))
681 /* new backing store needs to support loop (eg splice_read) */
682 if (!inode
->i_fop
->splice_read
)
685 /* size of the new backing store needs to be the same */
686 if (get_loop_size(lo
, file
) != get_loop_size(lo
, old_file
))
690 error
= loop_switch(lo
, file
);
703 static inline int is_loop_device(struct file
*file
)
705 struct inode
*i
= file
->f_mapping
->host
;
707 return i
&& S_ISBLK(i
->i_mode
) && MAJOR(i
->i_rdev
) == LOOP_MAJOR
;
710 static int loop_set_fd(struct loop_device
*lo
, struct file
*lo_file
,
711 struct block_device
*bdev
, unsigned int arg
)
713 struct file
*file
, *f
;
715 struct address_space
*mapping
;
716 unsigned lo_blocksize
;
721 /* This is safe, since we have a reference from open(). */
722 __module_get(THIS_MODULE
);
730 if (lo
->lo_state
!= Lo_unbound
)
733 /* Avoid recursion */
735 while (is_loop_device(f
)) {
736 struct loop_device
*l
;
738 if (f
->f_mapping
->host
->i_rdev
== lo_file
->f_mapping
->host
->i_rdev
)
741 l
= f
->f_mapping
->host
->i_bdev
->bd_disk
->private_data
;
742 if (l
->lo_state
== Lo_unbound
) {
746 f
= l
->lo_backing_file
;
749 mapping
= file
->f_mapping
;
750 inode
= mapping
->host
;
752 if (!(file
->f_mode
& FMODE_WRITE
))
753 lo_flags
|= LO_FLAGS_READ_ONLY
;
756 if (S_ISREG(inode
->i_mode
) || S_ISBLK(inode
->i_mode
)) {
757 const struct address_space_operations
*aops
= mapping
->a_ops
;
759 * If we can't read - sorry. If we only can't write - well,
760 * it's going to be read-only.
762 if (!file
->f_op
->splice_read
)
764 if (aops
->prepare_write
|| aops
->write_begin
)
765 lo_flags
|= LO_FLAGS_USE_AOPS
;
766 if (!(lo_flags
& LO_FLAGS_USE_AOPS
) && !file
->f_op
->write
)
767 lo_flags
|= LO_FLAGS_READ_ONLY
;
769 lo_blocksize
= S_ISBLK(inode
->i_mode
) ?
770 inode
->i_bdev
->bd_block_size
: PAGE_SIZE
;
777 size
= get_loop_size(lo
, file
);
779 if ((loff_t
)(sector_t
)size
!= size
) {
784 if (!(lo_file
->f_mode
& FMODE_WRITE
))
785 lo_flags
|= LO_FLAGS_READ_ONLY
;
787 set_device_ro(bdev
, (lo_flags
& LO_FLAGS_READ_ONLY
) != 0);
789 lo
->lo_blocksize
= lo_blocksize
;
790 lo
->lo_device
= bdev
;
791 lo
->lo_flags
= lo_flags
;
792 lo
->lo_backing_file
= file
;
793 lo
->transfer
= transfer_none
;
795 lo
->lo_sizelimit
= 0;
796 lo
->old_gfp_mask
= mapping_gfp_mask(mapping
);
797 mapping_set_gfp_mask(mapping
, lo
->old_gfp_mask
& ~(__GFP_IO
|__GFP_FS
));
799 lo
->lo_bio
= lo
->lo_biotail
= NULL
;
802 * set queue make_request_fn, and add limits based on lower level
805 blk_queue_make_request(lo
->lo_queue
, loop_make_request
);
806 lo
->lo_queue
->queuedata
= lo
;
807 lo
->lo_queue
->unplug_fn
= loop_unplug
;
809 set_capacity(lo
->lo_disk
, size
);
810 bd_set_size(bdev
, size
<< 9);
812 set_blocksize(bdev
, lo_blocksize
);
814 lo
->lo_thread
= kthread_create(loop_thread
, lo
, "loop%d",
816 if (IS_ERR(lo
->lo_thread
)) {
817 error
= PTR_ERR(lo
->lo_thread
);
820 lo
->lo_state
= Lo_bound
;
821 wake_up_process(lo
->lo_thread
);
825 lo
->lo_thread
= NULL
;
826 lo
->lo_device
= NULL
;
827 lo
->lo_backing_file
= NULL
;
829 set_capacity(lo
->lo_disk
, 0);
830 invalidate_bdev(bdev
);
831 bd_set_size(bdev
, 0);
832 mapping_set_gfp_mask(mapping
, lo
->old_gfp_mask
);
833 lo
->lo_state
= Lo_unbound
;
837 /* This is safe: open() is still holding a reference. */
838 module_put(THIS_MODULE
);
843 loop_release_xfer(struct loop_device
*lo
)
846 struct loop_func_table
*xfer
= lo
->lo_encryption
;
850 err
= xfer
->release(lo
);
852 lo
->lo_encryption
= NULL
;
853 module_put(xfer
->owner
);
859 loop_init_xfer(struct loop_device
*lo
, struct loop_func_table
*xfer
,
860 const struct loop_info64
*i
)
865 struct module
*owner
= xfer
->owner
;
867 if (!try_module_get(owner
))
870 err
= xfer
->init(lo
, i
);
874 lo
->lo_encryption
= xfer
;
879 static int loop_clr_fd(struct loop_device
*lo
, struct block_device
*bdev
)
881 struct file
*filp
= lo
->lo_backing_file
;
882 gfp_t gfp
= lo
->old_gfp_mask
;
884 if (lo
->lo_state
!= Lo_bound
)
887 if (lo
->lo_refcnt
> 1) /* we needed one fd for the ioctl */
893 spin_lock_irq(&lo
->lo_lock
);
894 lo
->lo_state
= Lo_rundown
;
895 spin_unlock_irq(&lo
->lo_lock
);
897 kthread_stop(lo
->lo_thread
);
899 lo
->lo_backing_file
= NULL
;
901 loop_release_xfer(lo
);
904 lo
->lo_device
= NULL
;
905 lo
->lo_encryption
= NULL
;
907 lo
->lo_sizelimit
= 0;
908 lo
->lo_encrypt_key_size
= 0;
910 lo
->lo_thread
= NULL
;
911 memset(lo
->lo_encrypt_key
, 0, LO_KEY_SIZE
);
912 memset(lo
->lo_crypt_name
, 0, LO_NAME_SIZE
);
913 memset(lo
->lo_file_name
, 0, LO_NAME_SIZE
);
914 invalidate_bdev(bdev
);
915 set_capacity(lo
->lo_disk
, 0);
916 bd_set_size(bdev
, 0);
917 mapping_set_gfp_mask(filp
->f_mapping
, gfp
);
918 lo
->lo_state
= Lo_unbound
;
920 /* This is safe: open() is still holding a reference. */
921 module_put(THIS_MODULE
);
926 loop_set_status(struct loop_device
*lo
, const struct loop_info64
*info
)
929 struct loop_func_table
*xfer
;
931 if (lo
->lo_encrypt_key_size
&& lo
->lo_key_owner
!= current
->uid
&&
932 !capable(CAP_SYS_ADMIN
))
934 if (lo
->lo_state
!= Lo_bound
)
936 if ((unsigned int) info
->lo_encrypt_key_size
> LO_KEY_SIZE
)
939 err
= loop_release_xfer(lo
);
943 if (info
->lo_encrypt_type
) {
944 unsigned int type
= info
->lo_encrypt_type
;
946 if (type
>= MAX_LO_CRYPT
)
948 xfer
= xfer_funcs
[type
];
954 err
= loop_init_xfer(lo
, xfer
, info
);
958 if (lo
->lo_offset
!= info
->lo_offset
||
959 lo
->lo_sizelimit
!= info
->lo_sizelimit
) {
960 lo
->lo_offset
= info
->lo_offset
;
961 lo
->lo_sizelimit
= info
->lo_sizelimit
;
962 if (figure_loop_size(lo
))
966 memcpy(lo
->lo_file_name
, info
->lo_file_name
, LO_NAME_SIZE
);
967 memcpy(lo
->lo_crypt_name
, info
->lo_crypt_name
, LO_NAME_SIZE
);
968 lo
->lo_file_name
[LO_NAME_SIZE
-1] = 0;
969 lo
->lo_crypt_name
[LO_NAME_SIZE
-1] = 0;
973 lo
->transfer
= xfer
->transfer
;
974 lo
->ioctl
= xfer
->ioctl
;
976 if ((lo
->lo_flags
& LO_FLAGS_AUTOCLEAR
) !=
977 (info
->lo_flags
& LO_FLAGS_AUTOCLEAR
))
978 lo
->lo_flags
^= LO_FLAGS_AUTOCLEAR
;
980 lo
->lo_encrypt_key_size
= info
->lo_encrypt_key_size
;
981 lo
->lo_init
[0] = info
->lo_init
[0];
982 lo
->lo_init
[1] = info
->lo_init
[1];
983 if (info
->lo_encrypt_key_size
) {
984 memcpy(lo
->lo_encrypt_key
, info
->lo_encrypt_key
,
985 info
->lo_encrypt_key_size
);
986 lo
->lo_key_owner
= current
->uid
;
993 loop_get_status(struct loop_device
*lo
, struct loop_info64
*info
)
995 struct file
*file
= lo
->lo_backing_file
;
999 if (lo
->lo_state
!= Lo_bound
)
1001 error
= vfs_getattr(file
->f_path
.mnt
, file
->f_path
.dentry
, &stat
);
1004 memset(info
, 0, sizeof(*info
));
1005 info
->lo_number
= lo
->lo_number
;
1006 info
->lo_device
= huge_encode_dev(stat
.dev
);
1007 info
->lo_inode
= stat
.ino
;
1008 info
->lo_rdevice
= huge_encode_dev(lo
->lo_device
? stat
.rdev
: stat
.dev
);
1009 info
->lo_offset
= lo
->lo_offset
;
1010 info
->lo_sizelimit
= lo
->lo_sizelimit
;
1011 info
->lo_flags
= lo
->lo_flags
;
1012 memcpy(info
->lo_file_name
, lo
->lo_file_name
, LO_NAME_SIZE
);
1013 memcpy(info
->lo_crypt_name
, lo
->lo_crypt_name
, LO_NAME_SIZE
);
1014 info
->lo_encrypt_type
=
1015 lo
->lo_encryption
? lo
->lo_encryption
->number
: 0;
1016 if (lo
->lo_encrypt_key_size
&& capable(CAP_SYS_ADMIN
)) {
1017 info
->lo_encrypt_key_size
= lo
->lo_encrypt_key_size
;
1018 memcpy(info
->lo_encrypt_key
, lo
->lo_encrypt_key
,
1019 lo
->lo_encrypt_key_size
);
1025 loop_info64_from_old(const struct loop_info
*info
, struct loop_info64
*info64
)
1027 memset(info64
, 0, sizeof(*info64
));
1028 info64
->lo_number
= info
->lo_number
;
1029 info64
->lo_device
= info
->lo_device
;
1030 info64
->lo_inode
= info
->lo_inode
;
1031 info64
->lo_rdevice
= info
->lo_rdevice
;
1032 info64
->lo_offset
= info
->lo_offset
;
1033 info64
->lo_sizelimit
= 0;
1034 info64
->lo_encrypt_type
= info
->lo_encrypt_type
;
1035 info64
->lo_encrypt_key_size
= info
->lo_encrypt_key_size
;
1036 info64
->lo_flags
= info
->lo_flags
;
1037 info64
->lo_init
[0] = info
->lo_init
[0];
1038 info64
->lo_init
[1] = info
->lo_init
[1];
1039 if (info
->lo_encrypt_type
== LO_CRYPT_CRYPTOAPI
)
1040 memcpy(info64
->lo_crypt_name
, info
->lo_name
, LO_NAME_SIZE
);
1042 memcpy(info64
->lo_file_name
, info
->lo_name
, LO_NAME_SIZE
);
1043 memcpy(info64
->lo_encrypt_key
, info
->lo_encrypt_key
, LO_KEY_SIZE
);
1047 loop_info64_to_old(const struct loop_info64
*info64
, struct loop_info
*info
)
1049 memset(info
, 0, sizeof(*info
));
1050 info
->lo_number
= info64
->lo_number
;
1051 info
->lo_device
= info64
->lo_device
;
1052 info
->lo_inode
= info64
->lo_inode
;
1053 info
->lo_rdevice
= info64
->lo_rdevice
;
1054 info
->lo_offset
= info64
->lo_offset
;
1055 info
->lo_encrypt_type
= info64
->lo_encrypt_type
;
1056 info
->lo_encrypt_key_size
= info64
->lo_encrypt_key_size
;
1057 info
->lo_flags
= info64
->lo_flags
;
1058 info
->lo_init
[0] = info64
->lo_init
[0];
1059 info
->lo_init
[1] = info64
->lo_init
[1];
1060 if (info
->lo_encrypt_type
== LO_CRYPT_CRYPTOAPI
)
1061 memcpy(info
->lo_name
, info64
->lo_crypt_name
, LO_NAME_SIZE
);
1063 memcpy(info
->lo_name
, info64
->lo_file_name
, LO_NAME_SIZE
);
1064 memcpy(info
->lo_encrypt_key
, info64
->lo_encrypt_key
, LO_KEY_SIZE
);
1066 /* error in case values were truncated */
1067 if (info
->lo_device
!= info64
->lo_device
||
1068 info
->lo_rdevice
!= info64
->lo_rdevice
||
1069 info
->lo_inode
!= info64
->lo_inode
||
1070 info
->lo_offset
!= info64
->lo_offset
)
1077 loop_set_status_old(struct loop_device
*lo
, const struct loop_info __user
*arg
)
1079 struct loop_info info
;
1080 struct loop_info64 info64
;
1082 if (copy_from_user(&info
, arg
, sizeof (struct loop_info
)))
1084 loop_info64_from_old(&info
, &info64
);
1085 return loop_set_status(lo
, &info64
);
1089 loop_set_status64(struct loop_device
*lo
, const struct loop_info64 __user
*arg
)
1091 struct loop_info64 info64
;
1093 if (copy_from_user(&info64
, arg
, sizeof (struct loop_info64
)))
1095 return loop_set_status(lo
, &info64
);
1099 loop_get_status_old(struct loop_device
*lo
, struct loop_info __user
*arg
) {
1100 struct loop_info info
;
1101 struct loop_info64 info64
;
1107 err
= loop_get_status(lo
, &info64
);
1109 err
= loop_info64_to_old(&info64
, &info
);
1110 if (!err
&& copy_to_user(arg
, &info
, sizeof(info
)))
1117 loop_get_status64(struct loop_device
*lo
, struct loop_info64 __user
*arg
) {
1118 struct loop_info64 info64
;
1124 err
= loop_get_status(lo
, &info64
);
1125 if (!err
&& copy_to_user(arg
, &info64
, sizeof(info64
)))
1131 static int lo_ioctl(struct inode
* inode
, struct file
* file
,
1132 unsigned int cmd
, unsigned long arg
)
1134 struct loop_device
*lo
= inode
->i_bdev
->bd_disk
->private_data
;
1137 mutex_lock(&lo
->lo_ctl_mutex
);
1140 err
= loop_set_fd(lo
, file
, inode
->i_bdev
, arg
);
1142 case LOOP_CHANGE_FD
:
1143 err
= loop_change_fd(lo
, file
, inode
->i_bdev
, arg
);
1146 err
= loop_clr_fd(lo
, inode
->i_bdev
);
1148 case LOOP_SET_STATUS
:
1149 err
= loop_set_status_old(lo
, (struct loop_info __user
*) arg
);
1151 case LOOP_GET_STATUS
:
1152 err
= loop_get_status_old(lo
, (struct loop_info __user
*) arg
);
1154 case LOOP_SET_STATUS64
:
1155 err
= loop_set_status64(lo
, (struct loop_info64 __user
*) arg
);
1157 case LOOP_GET_STATUS64
:
1158 err
= loop_get_status64(lo
, (struct loop_info64 __user
*) arg
);
1161 err
= lo
->ioctl
? lo
->ioctl(lo
, cmd
, arg
) : -EINVAL
;
1163 mutex_unlock(&lo
->lo_ctl_mutex
);
1167 #ifdef CONFIG_COMPAT
1168 struct compat_loop_info
{
1169 compat_int_t lo_number
; /* ioctl r/o */
1170 compat_dev_t lo_device
; /* ioctl r/o */
1171 compat_ulong_t lo_inode
; /* ioctl r/o */
1172 compat_dev_t lo_rdevice
; /* ioctl r/o */
1173 compat_int_t lo_offset
;
1174 compat_int_t lo_encrypt_type
;
1175 compat_int_t lo_encrypt_key_size
; /* ioctl w/o */
1176 compat_int_t lo_flags
; /* ioctl r/o */
1177 char lo_name
[LO_NAME_SIZE
];
1178 unsigned char lo_encrypt_key
[LO_KEY_SIZE
]; /* ioctl w/o */
1179 compat_ulong_t lo_init
[2];
1184 * Transfer 32-bit compatibility structure in userspace to 64-bit loop info
1185 * - noinlined to reduce stack space usage in main part of driver
1188 loop_info64_from_compat(const struct compat_loop_info __user
*arg
,
1189 struct loop_info64
*info64
)
1191 struct compat_loop_info info
;
1193 if (copy_from_user(&info
, arg
, sizeof(info
)))
1196 memset(info64
, 0, sizeof(*info64
));
1197 info64
->lo_number
= info
.lo_number
;
1198 info64
->lo_device
= info
.lo_device
;
1199 info64
->lo_inode
= info
.lo_inode
;
1200 info64
->lo_rdevice
= info
.lo_rdevice
;
1201 info64
->lo_offset
= info
.lo_offset
;
1202 info64
->lo_sizelimit
= 0;
1203 info64
->lo_encrypt_type
= info
.lo_encrypt_type
;
1204 info64
->lo_encrypt_key_size
= info
.lo_encrypt_key_size
;
1205 info64
->lo_flags
= info
.lo_flags
;
1206 info64
->lo_init
[0] = info
.lo_init
[0];
1207 info64
->lo_init
[1] = info
.lo_init
[1];
1208 if (info
.lo_encrypt_type
== LO_CRYPT_CRYPTOAPI
)
1209 memcpy(info64
->lo_crypt_name
, info
.lo_name
, LO_NAME_SIZE
);
1211 memcpy(info64
->lo_file_name
, info
.lo_name
, LO_NAME_SIZE
);
1212 memcpy(info64
->lo_encrypt_key
, info
.lo_encrypt_key
, LO_KEY_SIZE
);
1217 * Transfer 64-bit loop info to 32-bit compatibility structure in userspace
1218 * - noinlined to reduce stack space usage in main part of driver
1221 loop_info64_to_compat(const struct loop_info64
*info64
,
1222 struct compat_loop_info __user
*arg
)
1224 struct compat_loop_info info
;
1226 memset(&info
, 0, sizeof(info
));
1227 info
.lo_number
= info64
->lo_number
;
1228 info
.lo_device
= info64
->lo_device
;
1229 info
.lo_inode
= info64
->lo_inode
;
1230 info
.lo_rdevice
= info64
->lo_rdevice
;
1231 info
.lo_offset
= info64
->lo_offset
;
1232 info
.lo_encrypt_type
= info64
->lo_encrypt_type
;
1233 info
.lo_encrypt_key_size
= info64
->lo_encrypt_key_size
;
1234 info
.lo_flags
= info64
->lo_flags
;
1235 info
.lo_init
[0] = info64
->lo_init
[0];
1236 info
.lo_init
[1] = info64
->lo_init
[1];
1237 if (info
.lo_encrypt_type
== LO_CRYPT_CRYPTOAPI
)
1238 memcpy(info
.lo_name
, info64
->lo_crypt_name
, LO_NAME_SIZE
);
1240 memcpy(info
.lo_name
, info64
->lo_file_name
, LO_NAME_SIZE
);
1241 memcpy(info
.lo_encrypt_key
, info64
->lo_encrypt_key
, LO_KEY_SIZE
);
1243 /* error in case values were truncated */
1244 if (info
.lo_device
!= info64
->lo_device
||
1245 info
.lo_rdevice
!= info64
->lo_rdevice
||
1246 info
.lo_inode
!= info64
->lo_inode
||
1247 info
.lo_offset
!= info64
->lo_offset
||
1248 info
.lo_init
[0] != info64
->lo_init
[0] ||
1249 info
.lo_init
[1] != info64
->lo_init
[1])
1252 if (copy_to_user(arg
, &info
, sizeof(info
)))
1258 loop_set_status_compat(struct loop_device
*lo
,
1259 const struct compat_loop_info __user
*arg
)
1261 struct loop_info64 info64
;
1264 ret
= loop_info64_from_compat(arg
, &info64
);
1267 return loop_set_status(lo
, &info64
);
1271 loop_get_status_compat(struct loop_device
*lo
,
1272 struct compat_loop_info __user
*arg
)
1274 struct loop_info64 info64
;
1280 err
= loop_get_status(lo
, &info64
);
1282 err
= loop_info64_to_compat(&info64
, arg
);
1286 static long lo_compat_ioctl(struct file
*file
, unsigned int cmd
, unsigned long arg
)
1288 struct inode
*inode
= file
->f_path
.dentry
->d_inode
;
1289 struct loop_device
*lo
= inode
->i_bdev
->bd_disk
->private_data
;
1293 case LOOP_SET_STATUS
:
1294 mutex_lock(&lo
->lo_ctl_mutex
);
1295 err
= loop_set_status_compat(
1296 lo
, (const struct compat_loop_info __user
*) arg
);
1297 mutex_unlock(&lo
->lo_ctl_mutex
);
1299 case LOOP_GET_STATUS
:
1300 mutex_lock(&lo
->lo_ctl_mutex
);
1301 err
= loop_get_status_compat(
1302 lo
, (struct compat_loop_info __user
*) arg
);
1303 mutex_unlock(&lo
->lo_ctl_mutex
);
1306 case LOOP_GET_STATUS64
:
1307 case LOOP_SET_STATUS64
:
1308 arg
= (unsigned long) compat_ptr(arg
);
1310 case LOOP_CHANGE_FD
:
1311 err
= lo_ioctl(inode
, file
, cmd
, arg
);
1321 static int lo_open(struct inode
*inode
, struct file
*file
)
1323 struct loop_device
*lo
= inode
->i_bdev
->bd_disk
->private_data
;
1325 mutex_lock(&lo
->lo_ctl_mutex
);
1327 mutex_unlock(&lo
->lo_ctl_mutex
);
1332 static int lo_release(struct inode
*inode
, struct file
*file
)
1334 struct loop_device
*lo
= inode
->i_bdev
->bd_disk
->private_data
;
1336 mutex_lock(&lo
->lo_ctl_mutex
);
1339 if ((lo
->lo_flags
& LO_FLAGS_AUTOCLEAR
) && !lo
->lo_refcnt
)
1340 loop_clr_fd(lo
, inode
->i_bdev
);
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 static int max_loop
;
1361 module_param(max_loop
, int, 0);
1362 MODULE_PARM_DESC(max_loop
, "Maximum number of loop devices");
1363 MODULE_LICENSE("GPL");
1364 MODULE_ALIAS_BLOCKDEV_MAJOR(LOOP_MAJOR
);
1366 int loop_register_transfer(struct loop_func_table
*funcs
)
1368 unsigned int n
= funcs
->number
;
1370 if (n
>= MAX_LO_CRYPT
|| xfer_funcs
[n
])
1372 xfer_funcs
[n
] = funcs
;
1376 int loop_unregister_transfer(int number
)
1378 unsigned int n
= number
;
1379 struct loop_device
*lo
;
1380 struct loop_func_table
*xfer
;
1382 if (n
== 0 || n
>= MAX_LO_CRYPT
|| (xfer
= xfer_funcs
[n
]) == NULL
)
1385 xfer_funcs
[n
] = NULL
;
1387 list_for_each_entry(lo
, &loop_devices
, lo_list
) {
1388 mutex_lock(&lo
->lo_ctl_mutex
);
1390 if (lo
->lo_encryption
== xfer
)
1391 loop_release_xfer(lo
);
1393 mutex_unlock(&lo
->lo_ctl_mutex
);
1399 EXPORT_SYMBOL(loop_register_transfer
);
1400 EXPORT_SYMBOL(loop_unregister_transfer
);
1402 static struct loop_device
*loop_alloc(int i
)
1404 struct loop_device
*lo
;
1405 struct gendisk
*disk
;
1407 lo
= kzalloc(sizeof(*lo
), GFP_KERNEL
);
1411 lo
->lo_queue
= blk_alloc_queue(GFP_KERNEL
);
1415 disk
= lo
->lo_disk
= alloc_disk(1);
1417 goto out_free_queue
;
1419 mutex_init(&lo
->lo_ctl_mutex
);
1421 lo
->lo_thread
= NULL
;
1422 init_waitqueue_head(&lo
->lo_event
);
1423 spin_lock_init(&lo
->lo_lock
);
1424 disk
->major
= LOOP_MAJOR
;
1425 disk
->first_minor
= i
;
1426 disk
->fops
= &lo_fops
;
1427 disk
->private_data
= lo
;
1428 disk
->queue
= lo
->lo_queue
;
1429 sprintf(disk
->disk_name
, "loop%d", i
);
1433 blk_cleanup_queue(lo
->lo_queue
);
1440 static void loop_free(struct loop_device
*lo
)
1442 blk_cleanup_queue(lo
->lo_queue
);
1443 put_disk(lo
->lo_disk
);
1444 list_del(&lo
->lo_list
);
1448 static struct loop_device
*loop_init_one(int i
)
1450 struct loop_device
*lo
;
1452 list_for_each_entry(lo
, &loop_devices
, lo_list
) {
1453 if (lo
->lo_number
== i
)
1459 add_disk(lo
->lo_disk
);
1460 list_add_tail(&lo
->lo_list
, &loop_devices
);
1465 static void loop_del_one(struct loop_device
*lo
)
1467 del_gendisk(lo
->lo_disk
);
1471 static struct kobject
*loop_probe(dev_t dev
, int *part
, void *data
)
1473 struct loop_device
*lo
;
1474 struct kobject
*kobj
;
1476 mutex_lock(&loop_devices_mutex
);
1477 lo
= loop_init_one(dev
& MINORMASK
);
1478 kobj
= lo
? get_disk(lo
->lo_disk
) : ERR_PTR(-ENOMEM
);
1479 mutex_unlock(&loop_devices_mutex
);
1485 static int __init
loop_init(void)
1488 unsigned long range
;
1489 struct loop_device
*lo
, *next
;
1492 * loop module now has a feature to instantiate underlying device
1493 * structure on-demand, provided that there is an access dev node.
1494 * However, this will not work well with user space tool that doesn't
1495 * know about such "feature". In order to not break any existing
1496 * tool, we do the following:
1498 * (1) if max_loop is specified, create that many upfront, and this
1499 * also becomes a hard limit.
1500 * (2) if max_loop is not specified, create 8 loop device on module
1501 * load, user can further extend loop device by create dev node
1502 * themselves and have kernel automatically instantiate actual
1505 if (max_loop
> 1UL << MINORBITS
)
1513 range
= 1UL << MINORBITS
;
1516 if (register_blkdev(LOOP_MAJOR
, "loop"))
1519 for (i
= 0; i
< nr
; i
++) {
1523 list_add_tail(&lo
->lo_list
, &loop_devices
);
1526 /* point of no return */
1528 list_for_each_entry(lo
, &loop_devices
, lo_list
)
1529 add_disk(lo
->lo_disk
);
1531 blk_register_region(MKDEV(LOOP_MAJOR
, 0), range
,
1532 THIS_MODULE
, loop_probe
, NULL
, NULL
);
1534 printk(KERN_INFO
"loop: module loaded\n");
1538 printk(KERN_INFO
"loop: out of memory\n");
1540 list_for_each_entry_safe(lo
, next
, &loop_devices
, lo_list
)
1543 unregister_blkdev(LOOP_MAJOR
, "loop");
1547 static void __exit
loop_exit(void)
1549 unsigned long range
;
1550 struct loop_device
*lo
, *next
;
1552 range
= max_loop
? max_loop
: 1UL << MINORBITS
;
1554 list_for_each_entry_safe(lo
, next
, &loop_devices
, lo_list
)
1557 blk_unregister_region(MKDEV(LOOP_MAJOR
, 0), range
);
1558 unregister_blkdev(LOOP_MAJOR
, "loop");
1561 module_init(loop_init
);
1562 module_exit(loop_exit
);
1565 static int __init
max_loop_setup(char *str
)
1567 max_loop
= simple_strtol(str
, NULL
, 0);
1571 __setup("max_loop=", max_loop_setup
);