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 write_begin is not available on the backing filesystem.
44 * Anton Altaparmakov, 16 Feb 2005
47 * - Advisory locking is ignored here.
48 * - Should use an own CAP_* category instead of CAP_SYS_ADMIN
52 #include <linux/module.h>
53 #include <linux/moduleparam.h>
54 #include <linux/sched.h>
56 #include <linux/file.h>
57 #include <linux/stat.h>
58 #include <linux/errno.h>
59 #include <linux/major.h>
60 #include <linux/wait.h>
61 #include <linux/blkdev.h>
62 #include <linux/blkpg.h>
63 #include <linux/init.h>
64 #include <linux/swap.h>
65 #include <linux/slab.h>
66 #include <linux/loop.h>
67 #include <linux/compat.h>
68 #include <linux/suspend.h>
69 #include <linux/freezer.h>
70 #include <linux/mutex.h>
71 #include <linux/writeback.h>
72 #include <linux/buffer_head.h> /* for invalidate_bdev() */
73 #include <linux/completion.h>
74 #include <linux/highmem.h>
75 #include <linux/kthread.h>
76 #include <linux/splice.h>
77 #include <linux/sysfs.h>
79 #include <asm/uaccess.h>
81 static DEFINE_MUTEX(loop_mutex
);
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(loop_buf
, KM_USER1
);
105 kunmap_atomic(raw_buf
, KM_USER0
);
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(loop_buf
, KM_USER1
);
134 kunmap_atomic(raw_buf
, KM_USER0
);
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 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 file_update_time(file
);
245 transfer_result
= lo_do_transfer(lo
, WRITE
, page
, offset
,
246 bvec
->bv_page
, bv_offs
, size
, IV
);
248 if (unlikely(transfer_result
))
251 ret
= pagecache_write_end(file
, mapping
, pos
, size
, copied
,
253 if (ret
< 0 || ret
!= copied
)
256 if (unlikely(transfer_result
))
267 mutex_unlock(&mapping
->host
->i_mutex
);
275 * __do_lo_send_write - helper for writing data to a loop device
277 * This helper just factors out common code between do_lo_send_direct_write()
278 * and do_lo_send_write().
280 static int __do_lo_send_write(struct file
*file
,
281 u8
*buf
, const int len
, loff_t pos
)
284 mm_segment_t old_fs
= get_fs();
287 bw
= file
->f_op
->write(file
, buf
, len
, &pos
);
289 if (likely(bw
== len
))
291 printk(KERN_ERR
"loop: Write error at byte offset %llu, length %i.\n",
292 (unsigned long long)pos
, len
);
299 * do_lo_send_direct_write - helper for writing data to a loop device
301 * This is the fast, non-transforming version for backing filesystems which do
302 * not implement the address space operations write_begin and write_end.
303 * It uses the write file operation which should be present on all writeable
306 static int do_lo_send_direct_write(struct loop_device
*lo
,
307 struct bio_vec
*bvec
, loff_t pos
, struct page
*page
)
309 ssize_t bw
= __do_lo_send_write(lo
->lo_backing_file
,
310 kmap(bvec
->bv_page
) + bvec
->bv_offset
,
312 kunmap(bvec
->bv_page
);
318 * do_lo_send_write - helper for writing data to a loop device
320 * This is the slow, transforming version for filesystems which do not
321 * implement the address space operations write_begin and write_end. It
322 * uses the write file operation which should be present on all writeable
325 * Using fops->write is slower than using aops->{prepare,commit}_write in the
326 * transforming case because we need to double buffer the data as we cannot do
327 * the transformations in place as we do not have direct access to the
328 * destination pages of the backing file.
330 static int do_lo_send_write(struct loop_device
*lo
, struct bio_vec
*bvec
,
331 loff_t pos
, struct page
*page
)
333 int ret
= lo_do_transfer(lo
, WRITE
, page
, 0, bvec
->bv_page
,
334 bvec
->bv_offset
, bvec
->bv_len
, pos
>> 9);
336 return __do_lo_send_write(lo
->lo_backing_file
,
337 page_address(page
), bvec
->bv_len
,
339 printk(KERN_ERR
"loop: Transfer error at byte offset %llu, "
340 "length %i.\n", (unsigned long long)pos
, bvec
->bv_len
);
346 static int lo_send(struct loop_device
*lo
, struct bio
*bio
, loff_t pos
)
348 int (*do_lo_send
)(struct loop_device
*, struct bio_vec
*, loff_t
,
350 struct bio_vec
*bvec
;
351 struct page
*page
= NULL
;
354 do_lo_send
= do_lo_send_aops
;
355 if (!(lo
->lo_flags
& LO_FLAGS_USE_AOPS
)) {
356 do_lo_send
= do_lo_send_direct_write
;
357 if (lo
->transfer
!= transfer_none
) {
358 page
= alloc_page(GFP_NOIO
| __GFP_HIGHMEM
);
362 do_lo_send
= do_lo_send_write
;
365 bio_for_each_segment(bvec
, bio
, i
) {
366 ret
= do_lo_send(lo
, bvec
, pos
, page
);
378 printk(KERN_ERR
"loop: Failed to allocate temporary page for write.\n");
383 struct lo_read_data
{
384 struct loop_device
*lo
;
391 lo_splice_actor(struct pipe_inode_info
*pipe
, struct pipe_buffer
*buf
,
392 struct splice_desc
*sd
)
394 struct lo_read_data
*p
= sd
->u
.data
;
395 struct loop_device
*lo
= p
->lo
;
396 struct page
*page
= buf
->page
;
400 IV
= ((sector_t
) page
->index
<< (PAGE_CACHE_SHIFT
- 9)) +
406 if (lo_do_transfer(lo
, READ
, page
, buf
->offset
, p
->page
, p
->offset
, size
, IV
)) {
407 printk(KERN_ERR
"loop: transfer error block %ld\n",
412 flush_dcache_page(p
->page
);
421 lo_direct_splice_actor(struct pipe_inode_info
*pipe
, struct splice_desc
*sd
)
423 return __splice_from_pipe(pipe
, sd
, lo_splice_actor
);
427 do_lo_receive(struct loop_device
*lo
,
428 struct bio_vec
*bvec
, int bsize
, loff_t pos
)
430 struct lo_read_data cookie
;
431 struct splice_desc sd
;
436 cookie
.page
= bvec
->bv_page
;
437 cookie
.offset
= bvec
->bv_offset
;
438 cookie
.bsize
= bsize
;
441 sd
.total_len
= bvec
->bv_len
;
446 file
= lo
->lo_backing_file
;
447 retval
= splice_direct_to_actor(file
, &sd
, lo_direct_splice_actor
);
456 lo_receive(struct loop_device
*lo
, struct bio
*bio
, int bsize
, loff_t pos
)
458 struct bio_vec
*bvec
;
461 bio_for_each_segment(bvec
, bio
, i
) {
462 ret
= do_lo_receive(lo
, bvec
, bsize
, pos
);
470 static int do_bio_filebacked(struct loop_device
*lo
, struct bio
*bio
)
475 pos
= ((loff_t
) bio
->bi_sector
<< 9) + lo
->lo_offset
;
477 if (bio_rw(bio
) == WRITE
) {
478 struct file
*file
= lo
->lo_backing_file
;
480 if (bio
->bi_rw
& REQ_FLUSH
) {
481 ret
= vfs_fsync(file
, 0);
482 if (unlikely(ret
&& ret
!= -EINVAL
)) {
488 ret
= lo_send(lo
, bio
, pos
);
490 if ((bio
->bi_rw
& REQ_FUA
) && !ret
) {
491 ret
= vfs_fsync(file
, 0);
492 if (unlikely(ret
&& ret
!= -EINVAL
))
496 ret
= lo_receive(lo
, bio
, lo
->lo_blocksize
, pos
);
503 * Add bio to back of pending list
505 static void loop_add_bio(struct loop_device
*lo
, struct bio
*bio
)
507 bio_list_add(&lo
->lo_bio_list
, bio
);
511 * Grab first pending buffer
513 static struct bio
*loop_get_bio(struct loop_device
*lo
)
515 return bio_list_pop(&lo
->lo_bio_list
);
518 static int loop_make_request(struct request_queue
*q
, struct bio
*old_bio
)
520 struct loop_device
*lo
= q
->queuedata
;
521 int rw
= bio_rw(old_bio
);
526 BUG_ON(!lo
|| (rw
!= READ
&& rw
!= WRITE
));
528 spin_lock_irq(&lo
->lo_lock
);
529 if (lo
->lo_state
!= Lo_bound
)
531 if (unlikely(rw
== WRITE
&& (lo
->lo_flags
& LO_FLAGS_READ_ONLY
)))
533 loop_add_bio(lo
, old_bio
);
534 wake_up(&lo
->lo_event
);
535 spin_unlock_irq(&lo
->lo_lock
);
539 spin_unlock_irq(&lo
->lo_lock
);
540 bio_io_error(old_bio
);
545 * kick off io on the underlying address space
547 static void loop_unplug(struct request_queue
*q
)
549 struct loop_device
*lo
= q
->queuedata
;
551 queue_flag_clear_unlocked(QUEUE_FLAG_PLUGGED
, q
);
552 blk_run_address_space(lo
->lo_backing_file
->f_mapping
);
555 struct switch_request
{
557 struct completion wait
;
560 static void do_loop_switch(struct loop_device
*, struct switch_request
*);
562 static inline void loop_handle_bio(struct loop_device
*lo
, struct bio
*bio
)
564 if (unlikely(!bio
->bi_bdev
)) {
565 do_loop_switch(lo
, bio
->bi_private
);
568 int ret
= do_bio_filebacked(lo
, bio
);
574 * worker thread that handles reads/writes to file backed loop devices,
575 * to avoid blocking in our make_request_fn. it also does loop decrypting
576 * on reads for block backed loop, as that is too heavy to do from
577 * b_end_io context where irqs may be disabled.
579 * Loop explanation: loop_clr_fd() sets lo_state to Lo_rundown before
580 * calling kthread_stop(). Therefore once kthread_should_stop() is
581 * true, make_request will not place any more requests. Therefore
582 * once kthread_should_stop() is true and lo_bio is NULL, we are
583 * done with the loop.
585 static int loop_thread(void *data
)
587 struct loop_device
*lo
= data
;
590 set_user_nice(current
, -20);
592 while (!kthread_should_stop() || !bio_list_empty(&lo
->lo_bio_list
)) {
594 wait_event_interruptible(lo
->lo_event
,
595 !bio_list_empty(&lo
->lo_bio_list
) ||
596 kthread_should_stop());
598 if (bio_list_empty(&lo
->lo_bio_list
))
600 spin_lock_irq(&lo
->lo_lock
);
601 bio
= loop_get_bio(lo
);
602 spin_unlock_irq(&lo
->lo_lock
);
605 loop_handle_bio(lo
, bio
);
612 * loop_switch performs the hard work of switching a backing store.
613 * First it needs to flush existing IO, it does this by sending a magic
614 * BIO down the pipe. The completion of this BIO does the actual switch.
616 static int loop_switch(struct loop_device
*lo
, struct file
*file
)
618 struct switch_request w
;
619 struct bio
*bio
= bio_alloc(GFP_KERNEL
, 0);
622 init_completion(&w
.wait
);
624 bio
->bi_private
= &w
;
626 loop_make_request(lo
->lo_queue
, bio
);
627 wait_for_completion(&w
.wait
);
632 * Helper to flush the IOs in loop, but keeping loop thread running
634 static int loop_flush(struct loop_device
*lo
)
636 /* loop not yet configured, no running thread, nothing to flush */
640 return loop_switch(lo
, NULL
);
644 * Do the actual switch; called from the BIO completion routine
646 static void do_loop_switch(struct loop_device
*lo
, struct switch_request
*p
)
648 struct file
*file
= p
->file
;
649 struct file
*old_file
= lo
->lo_backing_file
;
650 struct address_space
*mapping
;
652 /* if no new file, only flush of queued bios requested */
656 mapping
= file
->f_mapping
;
657 mapping_set_gfp_mask(old_file
->f_mapping
, lo
->old_gfp_mask
);
658 lo
->lo_backing_file
= file
;
659 lo
->lo_blocksize
= S_ISBLK(mapping
->host
->i_mode
) ?
660 mapping
->host
->i_bdev
->bd_block_size
: PAGE_SIZE
;
661 lo
->old_gfp_mask
= mapping_gfp_mask(mapping
);
662 mapping_set_gfp_mask(mapping
, lo
->old_gfp_mask
& ~(__GFP_IO
|__GFP_FS
));
669 * loop_change_fd switched the backing store of a loopback device to
670 * a new file. This is useful for operating system installers to free up
671 * the original file and in High Availability environments to switch to
672 * an alternative location for the content in case of server meltdown.
673 * This can only work if the loop device is used read-only, and if the
674 * new backing store is the same size and type as the old backing store.
676 static int loop_change_fd(struct loop_device
*lo
, struct block_device
*bdev
,
679 struct file
*file
, *old_file
;
684 if (lo
->lo_state
!= Lo_bound
)
687 /* the loop device has to be read-only */
689 if (!(lo
->lo_flags
& LO_FLAGS_READ_ONLY
))
697 inode
= file
->f_mapping
->host
;
698 old_file
= lo
->lo_backing_file
;
702 if (!S_ISREG(inode
->i_mode
) && !S_ISBLK(inode
->i_mode
))
705 /* size of the new backing store needs to be the same */
706 if (get_loop_size(lo
, file
) != get_loop_size(lo
, old_file
))
710 error
= loop_switch(lo
, file
);
716 ioctl_by_bdev(bdev
, BLKRRPART
, 0);
725 static inline int is_loop_device(struct file
*file
)
727 struct inode
*i
= file
->f_mapping
->host
;
729 return i
&& S_ISBLK(i
->i_mode
) && MAJOR(i
->i_rdev
) == LOOP_MAJOR
;
732 /* loop sysfs attributes */
734 static ssize_t
loop_attr_show(struct device
*dev
, char *page
,
735 ssize_t (*callback
)(struct loop_device
*, char *))
737 struct loop_device
*l
, *lo
= NULL
;
739 mutex_lock(&loop_devices_mutex
);
740 list_for_each_entry(l
, &loop_devices
, lo_list
)
741 if (disk_to_dev(l
->lo_disk
) == dev
) {
745 mutex_unlock(&loop_devices_mutex
);
747 return lo
? callback(lo
, page
) : -EIO
;
750 #define LOOP_ATTR_RO(_name) \
751 static ssize_t loop_attr_##_name##_show(struct loop_device *, char *); \
752 static ssize_t loop_attr_do_show_##_name(struct device *d, \
753 struct device_attribute *attr, char *b) \
755 return loop_attr_show(d, b, loop_attr_##_name##_show); \
757 static struct device_attribute loop_attr_##_name = \
758 __ATTR(_name, S_IRUGO, loop_attr_do_show_##_name, NULL);
760 static ssize_t
loop_attr_backing_file_show(struct loop_device
*lo
, char *buf
)
765 mutex_lock(&lo
->lo_ctl_mutex
);
766 if (lo
->lo_backing_file
)
767 p
= d_path(&lo
->lo_backing_file
->f_path
, buf
, PAGE_SIZE
- 1);
768 mutex_unlock(&lo
->lo_ctl_mutex
);
770 if (IS_ERR_OR_NULL(p
))
774 memmove(buf
, p
, ret
);
782 static ssize_t
loop_attr_offset_show(struct loop_device
*lo
, char *buf
)
784 return sprintf(buf
, "%llu\n", (unsigned long long)lo
->lo_offset
);
787 static ssize_t
loop_attr_sizelimit_show(struct loop_device
*lo
, char *buf
)
789 return sprintf(buf
, "%llu\n", (unsigned long long)lo
->lo_sizelimit
);
792 static ssize_t
loop_attr_autoclear_show(struct loop_device
*lo
, char *buf
)
794 int autoclear
= (lo
->lo_flags
& LO_FLAGS_AUTOCLEAR
);
796 return sprintf(buf
, "%s\n", autoclear
? "1" : "0");
799 LOOP_ATTR_RO(backing_file
);
800 LOOP_ATTR_RO(offset
);
801 LOOP_ATTR_RO(sizelimit
);
802 LOOP_ATTR_RO(autoclear
);
804 static struct attribute
*loop_attrs
[] = {
805 &loop_attr_backing_file
.attr
,
806 &loop_attr_offset
.attr
,
807 &loop_attr_sizelimit
.attr
,
808 &loop_attr_autoclear
.attr
,
812 static struct attribute_group loop_attribute_group
= {
817 static int loop_sysfs_init(struct loop_device
*lo
)
819 return sysfs_create_group(&disk_to_dev(lo
->lo_disk
)->kobj
,
820 &loop_attribute_group
);
823 static void loop_sysfs_exit(struct loop_device
*lo
)
825 sysfs_remove_group(&disk_to_dev(lo
->lo_disk
)->kobj
,
826 &loop_attribute_group
);
829 static int loop_set_fd(struct loop_device
*lo
, fmode_t mode
,
830 struct block_device
*bdev
, unsigned int arg
)
832 struct file
*file
, *f
;
834 struct address_space
*mapping
;
835 unsigned lo_blocksize
;
840 /* This is safe, since we have a reference from open(). */
841 __module_get(THIS_MODULE
);
849 if (lo
->lo_state
!= Lo_unbound
)
852 /* Avoid recursion */
854 while (is_loop_device(f
)) {
855 struct loop_device
*l
;
857 if (f
->f_mapping
->host
->i_bdev
== bdev
)
860 l
= f
->f_mapping
->host
->i_bdev
->bd_disk
->private_data
;
861 if (l
->lo_state
== Lo_unbound
) {
865 f
= l
->lo_backing_file
;
868 mapping
= file
->f_mapping
;
869 inode
= mapping
->host
;
871 if (!(file
->f_mode
& FMODE_WRITE
))
872 lo_flags
|= LO_FLAGS_READ_ONLY
;
875 if (S_ISREG(inode
->i_mode
) || S_ISBLK(inode
->i_mode
)) {
876 const struct address_space_operations
*aops
= mapping
->a_ops
;
878 if (aops
->write_begin
)
879 lo_flags
|= LO_FLAGS_USE_AOPS
;
880 if (!(lo_flags
& LO_FLAGS_USE_AOPS
) && !file
->f_op
->write
)
881 lo_flags
|= LO_FLAGS_READ_ONLY
;
883 lo_blocksize
= S_ISBLK(inode
->i_mode
) ?
884 inode
->i_bdev
->bd_block_size
: PAGE_SIZE
;
891 size
= get_loop_size(lo
, file
);
893 if ((loff_t
)(sector_t
)size
!= size
) {
898 if (!(mode
& FMODE_WRITE
))
899 lo_flags
|= LO_FLAGS_READ_ONLY
;
901 set_device_ro(bdev
, (lo_flags
& LO_FLAGS_READ_ONLY
) != 0);
903 lo
->lo_blocksize
= lo_blocksize
;
904 lo
->lo_device
= bdev
;
905 lo
->lo_flags
= lo_flags
;
906 lo
->lo_backing_file
= file
;
907 lo
->transfer
= transfer_none
;
909 lo
->lo_sizelimit
= 0;
910 lo
->old_gfp_mask
= mapping_gfp_mask(mapping
);
911 mapping_set_gfp_mask(mapping
, lo
->old_gfp_mask
& ~(__GFP_IO
|__GFP_FS
));
913 bio_list_init(&lo
->lo_bio_list
);
916 * set queue make_request_fn, and add limits based on lower level
919 blk_queue_make_request(lo
->lo_queue
, loop_make_request
);
920 lo
->lo_queue
->queuedata
= lo
;
921 lo
->lo_queue
->unplug_fn
= loop_unplug
;
923 if (!(lo_flags
& LO_FLAGS_READ_ONLY
) && file
->f_op
->fsync
)
924 blk_queue_flush(lo
->lo_queue
, REQ_FLUSH
);
926 set_capacity(lo
->lo_disk
, size
);
927 bd_set_size(bdev
, size
<< 9);
929 /* let user-space know about the new size */
930 kobject_uevent(&disk_to_dev(bdev
->bd_disk
)->kobj
, KOBJ_CHANGE
);
932 set_blocksize(bdev
, lo_blocksize
);
934 lo
->lo_thread
= kthread_create(loop_thread
, lo
, "loop%d",
936 if (IS_ERR(lo
->lo_thread
)) {
937 error
= PTR_ERR(lo
->lo_thread
);
940 lo
->lo_state
= Lo_bound
;
941 wake_up_process(lo
->lo_thread
);
943 ioctl_by_bdev(bdev
, BLKRRPART
, 0);
948 lo
->lo_thread
= NULL
;
949 lo
->lo_device
= NULL
;
950 lo
->lo_backing_file
= NULL
;
952 set_capacity(lo
->lo_disk
, 0);
953 invalidate_bdev(bdev
);
954 bd_set_size(bdev
, 0);
955 kobject_uevent(&disk_to_dev(bdev
->bd_disk
)->kobj
, KOBJ_CHANGE
);
956 mapping_set_gfp_mask(mapping
, lo
->old_gfp_mask
);
957 lo
->lo_state
= Lo_unbound
;
961 /* This is safe: open() is still holding a reference. */
962 module_put(THIS_MODULE
);
967 loop_release_xfer(struct loop_device
*lo
)
970 struct loop_func_table
*xfer
= lo
->lo_encryption
;
974 err
= xfer
->release(lo
);
976 lo
->lo_encryption
= NULL
;
977 module_put(xfer
->owner
);
983 loop_init_xfer(struct loop_device
*lo
, struct loop_func_table
*xfer
,
984 const struct loop_info64
*i
)
989 struct module
*owner
= xfer
->owner
;
991 if (!try_module_get(owner
))
994 err
= xfer
->init(lo
, i
);
998 lo
->lo_encryption
= xfer
;
1003 static int loop_clr_fd(struct loop_device
*lo
, struct block_device
*bdev
)
1005 struct file
*filp
= lo
->lo_backing_file
;
1006 gfp_t gfp
= lo
->old_gfp_mask
;
1008 if (lo
->lo_state
!= Lo_bound
)
1011 if (lo
->lo_refcnt
> 1) /* we needed one fd for the ioctl */
1017 spin_lock_irq(&lo
->lo_lock
);
1018 lo
->lo_state
= Lo_rundown
;
1019 spin_unlock_irq(&lo
->lo_lock
);
1021 kthread_stop(lo
->lo_thread
);
1023 lo
->lo_queue
->unplug_fn
= NULL
;
1024 lo
->lo_backing_file
= NULL
;
1026 loop_release_xfer(lo
);
1027 lo
->transfer
= NULL
;
1029 lo
->lo_device
= NULL
;
1030 lo
->lo_encryption
= NULL
;
1032 lo
->lo_sizelimit
= 0;
1033 lo
->lo_encrypt_key_size
= 0;
1035 lo
->lo_thread
= NULL
;
1036 memset(lo
->lo_encrypt_key
, 0, LO_KEY_SIZE
);
1037 memset(lo
->lo_crypt_name
, 0, LO_NAME_SIZE
);
1038 memset(lo
->lo_file_name
, 0, LO_NAME_SIZE
);
1040 invalidate_bdev(bdev
);
1041 set_capacity(lo
->lo_disk
, 0);
1042 loop_sysfs_exit(lo
);
1044 bd_set_size(bdev
, 0);
1045 /* let user-space know about this change */
1046 kobject_uevent(&disk_to_dev(bdev
->bd_disk
)->kobj
, KOBJ_CHANGE
);
1048 mapping_set_gfp_mask(filp
->f_mapping
, gfp
);
1049 lo
->lo_state
= Lo_unbound
;
1050 /* This is safe: open() is still holding a reference. */
1051 module_put(THIS_MODULE
);
1052 if (max_part
> 0 && bdev
)
1053 ioctl_by_bdev(bdev
, BLKRRPART
, 0);
1054 mutex_unlock(&lo
->lo_ctl_mutex
);
1056 * Need not hold lo_ctl_mutex to fput backing file.
1057 * Calling fput holding lo_ctl_mutex triggers a circular
1058 * lock dependency possibility warning as fput can take
1059 * bd_mutex which is usually taken before lo_ctl_mutex.
1066 loop_set_status(struct loop_device
*lo
, const struct loop_info64
*info
)
1069 struct loop_func_table
*xfer
;
1070 uid_t uid
= current_uid();
1072 if (lo
->lo_encrypt_key_size
&&
1073 lo
->lo_key_owner
!= uid
&&
1074 !capable(CAP_SYS_ADMIN
))
1076 if (lo
->lo_state
!= Lo_bound
)
1078 if ((unsigned int) info
->lo_encrypt_key_size
> LO_KEY_SIZE
)
1081 err
= loop_release_xfer(lo
);
1085 if (info
->lo_encrypt_type
) {
1086 unsigned int type
= info
->lo_encrypt_type
;
1088 if (type
>= MAX_LO_CRYPT
)
1090 xfer
= xfer_funcs
[type
];
1096 err
= loop_init_xfer(lo
, xfer
, info
);
1100 if (lo
->lo_offset
!= info
->lo_offset
||
1101 lo
->lo_sizelimit
!= info
->lo_sizelimit
) {
1102 lo
->lo_offset
= info
->lo_offset
;
1103 lo
->lo_sizelimit
= info
->lo_sizelimit
;
1104 if (figure_loop_size(lo
))
1108 memcpy(lo
->lo_file_name
, info
->lo_file_name
, LO_NAME_SIZE
);
1109 memcpy(lo
->lo_crypt_name
, info
->lo_crypt_name
, LO_NAME_SIZE
);
1110 lo
->lo_file_name
[LO_NAME_SIZE
-1] = 0;
1111 lo
->lo_crypt_name
[LO_NAME_SIZE
-1] = 0;
1115 lo
->transfer
= xfer
->transfer
;
1116 lo
->ioctl
= xfer
->ioctl
;
1118 if ((lo
->lo_flags
& LO_FLAGS_AUTOCLEAR
) !=
1119 (info
->lo_flags
& LO_FLAGS_AUTOCLEAR
))
1120 lo
->lo_flags
^= LO_FLAGS_AUTOCLEAR
;
1122 lo
->lo_encrypt_key_size
= info
->lo_encrypt_key_size
;
1123 lo
->lo_init
[0] = info
->lo_init
[0];
1124 lo
->lo_init
[1] = info
->lo_init
[1];
1125 if (info
->lo_encrypt_key_size
) {
1126 memcpy(lo
->lo_encrypt_key
, info
->lo_encrypt_key
,
1127 info
->lo_encrypt_key_size
);
1128 lo
->lo_key_owner
= uid
;
1135 loop_get_status(struct loop_device
*lo
, struct loop_info64
*info
)
1137 struct file
*file
= lo
->lo_backing_file
;
1141 if (lo
->lo_state
!= Lo_bound
)
1143 error
= vfs_getattr(file
->f_path
.mnt
, file
->f_path
.dentry
, &stat
);
1146 memset(info
, 0, sizeof(*info
));
1147 info
->lo_number
= lo
->lo_number
;
1148 info
->lo_device
= huge_encode_dev(stat
.dev
);
1149 info
->lo_inode
= stat
.ino
;
1150 info
->lo_rdevice
= huge_encode_dev(lo
->lo_device
? stat
.rdev
: stat
.dev
);
1151 info
->lo_offset
= lo
->lo_offset
;
1152 info
->lo_sizelimit
= lo
->lo_sizelimit
;
1153 info
->lo_flags
= lo
->lo_flags
;
1154 memcpy(info
->lo_file_name
, lo
->lo_file_name
, LO_NAME_SIZE
);
1155 memcpy(info
->lo_crypt_name
, lo
->lo_crypt_name
, LO_NAME_SIZE
);
1156 info
->lo_encrypt_type
=
1157 lo
->lo_encryption
? lo
->lo_encryption
->number
: 0;
1158 if (lo
->lo_encrypt_key_size
&& capable(CAP_SYS_ADMIN
)) {
1159 info
->lo_encrypt_key_size
= lo
->lo_encrypt_key_size
;
1160 memcpy(info
->lo_encrypt_key
, lo
->lo_encrypt_key
,
1161 lo
->lo_encrypt_key_size
);
1167 loop_info64_from_old(const struct loop_info
*info
, struct loop_info64
*info64
)
1169 memset(info64
, 0, sizeof(*info64
));
1170 info64
->lo_number
= info
->lo_number
;
1171 info64
->lo_device
= info
->lo_device
;
1172 info64
->lo_inode
= info
->lo_inode
;
1173 info64
->lo_rdevice
= info
->lo_rdevice
;
1174 info64
->lo_offset
= info
->lo_offset
;
1175 info64
->lo_sizelimit
= 0;
1176 info64
->lo_encrypt_type
= info
->lo_encrypt_type
;
1177 info64
->lo_encrypt_key_size
= info
->lo_encrypt_key_size
;
1178 info64
->lo_flags
= info
->lo_flags
;
1179 info64
->lo_init
[0] = info
->lo_init
[0];
1180 info64
->lo_init
[1] = info
->lo_init
[1];
1181 if (info
->lo_encrypt_type
== LO_CRYPT_CRYPTOAPI
)
1182 memcpy(info64
->lo_crypt_name
, info
->lo_name
, LO_NAME_SIZE
);
1184 memcpy(info64
->lo_file_name
, info
->lo_name
, LO_NAME_SIZE
);
1185 memcpy(info64
->lo_encrypt_key
, info
->lo_encrypt_key
, LO_KEY_SIZE
);
1189 loop_info64_to_old(const struct loop_info64
*info64
, struct loop_info
*info
)
1191 memset(info
, 0, sizeof(*info
));
1192 info
->lo_number
= info64
->lo_number
;
1193 info
->lo_device
= info64
->lo_device
;
1194 info
->lo_inode
= info64
->lo_inode
;
1195 info
->lo_rdevice
= info64
->lo_rdevice
;
1196 info
->lo_offset
= info64
->lo_offset
;
1197 info
->lo_encrypt_type
= info64
->lo_encrypt_type
;
1198 info
->lo_encrypt_key_size
= info64
->lo_encrypt_key_size
;
1199 info
->lo_flags
= info64
->lo_flags
;
1200 info
->lo_init
[0] = info64
->lo_init
[0];
1201 info
->lo_init
[1] = info64
->lo_init
[1];
1202 if (info
->lo_encrypt_type
== LO_CRYPT_CRYPTOAPI
)
1203 memcpy(info
->lo_name
, info64
->lo_crypt_name
, LO_NAME_SIZE
);
1205 memcpy(info
->lo_name
, info64
->lo_file_name
, LO_NAME_SIZE
);
1206 memcpy(info
->lo_encrypt_key
, info64
->lo_encrypt_key
, LO_KEY_SIZE
);
1208 /* error in case values were truncated */
1209 if (info
->lo_device
!= info64
->lo_device
||
1210 info
->lo_rdevice
!= info64
->lo_rdevice
||
1211 info
->lo_inode
!= info64
->lo_inode
||
1212 info
->lo_offset
!= info64
->lo_offset
)
1219 loop_set_status_old(struct loop_device
*lo
, const struct loop_info __user
*arg
)
1221 struct loop_info info
;
1222 struct loop_info64 info64
;
1224 if (copy_from_user(&info
, arg
, sizeof (struct loop_info
)))
1226 loop_info64_from_old(&info
, &info64
);
1227 return loop_set_status(lo
, &info64
);
1231 loop_set_status64(struct loop_device
*lo
, const struct loop_info64 __user
*arg
)
1233 struct loop_info64 info64
;
1235 if (copy_from_user(&info64
, arg
, sizeof (struct loop_info64
)))
1237 return loop_set_status(lo
, &info64
);
1241 loop_get_status_old(struct loop_device
*lo
, struct loop_info __user
*arg
) {
1242 struct loop_info info
;
1243 struct loop_info64 info64
;
1249 err
= loop_get_status(lo
, &info64
);
1251 err
= loop_info64_to_old(&info64
, &info
);
1252 if (!err
&& copy_to_user(arg
, &info
, sizeof(info
)))
1259 loop_get_status64(struct loop_device
*lo
, struct loop_info64 __user
*arg
) {
1260 struct loop_info64 info64
;
1266 err
= loop_get_status(lo
, &info64
);
1267 if (!err
&& copy_to_user(arg
, &info64
, sizeof(info64
)))
1273 static int loop_set_capacity(struct loop_device
*lo
, struct block_device
*bdev
)
1280 if (unlikely(lo
->lo_state
!= Lo_bound
))
1282 err
= figure_loop_size(lo
);
1285 sec
= get_capacity(lo
->lo_disk
);
1286 /* the width of sector_t may be narrow for bit-shift */
1289 mutex_lock(&bdev
->bd_mutex
);
1290 bd_set_size(bdev
, sz
);
1291 /* let user-space know about the new size */
1292 kobject_uevent(&disk_to_dev(bdev
->bd_disk
)->kobj
, KOBJ_CHANGE
);
1293 mutex_unlock(&bdev
->bd_mutex
);
1299 static int lo_ioctl(struct block_device
*bdev
, fmode_t mode
,
1300 unsigned int cmd
, unsigned long arg
)
1302 struct loop_device
*lo
= bdev
->bd_disk
->private_data
;
1305 mutex_lock_nested(&lo
->lo_ctl_mutex
, 1);
1308 err
= loop_set_fd(lo
, mode
, bdev
, arg
);
1310 case LOOP_CHANGE_FD
:
1311 err
= loop_change_fd(lo
, bdev
, arg
);
1314 /* loop_clr_fd would have unlocked lo_ctl_mutex on success */
1315 err
= loop_clr_fd(lo
, bdev
);
1319 case LOOP_SET_STATUS
:
1320 err
= loop_set_status_old(lo
, (struct loop_info __user
*) arg
);
1322 case LOOP_GET_STATUS
:
1323 err
= loop_get_status_old(lo
, (struct loop_info __user
*) arg
);
1325 case LOOP_SET_STATUS64
:
1326 err
= loop_set_status64(lo
, (struct loop_info64 __user
*) arg
);
1328 case LOOP_GET_STATUS64
:
1329 err
= loop_get_status64(lo
, (struct loop_info64 __user
*) arg
);
1331 case LOOP_SET_CAPACITY
:
1333 if ((mode
& FMODE_WRITE
) || capable(CAP_SYS_ADMIN
))
1334 err
= loop_set_capacity(lo
, bdev
);
1337 err
= lo
->ioctl
? lo
->ioctl(lo
, cmd
, arg
) : -EINVAL
;
1339 mutex_unlock(&lo
->lo_ctl_mutex
);
1345 #ifdef CONFIG_COMPAT
1346 struct compat_loop_info
{
1347 compat_int_t lo_number
; /* ioctl r/o */
1348 compat_dev_t lo_device
; /* ioctl r/o */
1349 compat_ulong_t lo_inode
; /* ioctl r/o */
1350 compat_dev_t lo_rdevice
; /* ioctl r/o */
1351 compat_int_t lo_offset
;
1352 compat_int_t lo_encrypt_type
;
1353 compat_int_t lo_encrypt_key_size
; /* ioctl w/o */
1354 compat_int_t lo_flags
; /* ioctl r/o */
1355 char lo_name
[LO_NAME_SIZE
];
1356 unsigned char lo_encrypt_key
[LO_KEY_SIZE
]; /* ioctl w/o */
1357 compat_ulong_t lo_init
[2];
1362 * Transfer 32-bit compatibility structure in userspace to 64-bit loop info
1363 * - noinlined to reduce stack space usage in main part of driver
1366 loop_info64_from_compat(const struct compat_loop_info __user
*arg
,
1367 struct loop_info64
*info64
)
1369 struct compat_loop_info info
;
1371 if (copy_from_user(&info
, arg
, sizeof(info
)))
1374 memset(info64
, 0, sizeof(*info64
));
1375 info64
->lo_number
= info
.lo_number
;
1376 info64
->lo_device
= info
.lo_device
;
1377 info64
->lo_inode
= info
.lo_inode
;
1378 info64
->lo_rdevice
= info
.lo_rdevice
;
1379 info64
->lo_offset
= info
.lo_offset
;
1380 info64
->lo_sizelimit
= 0;
1381 info64
->lo_encrypt_type
= info
.lo_encrypt_type
;
1382 info64
->lo_encrypt_key_size
= info
.lo_encrypt_key_size
;
1383 info64
->lo_flags
= info
.lo_flags
;
1384 info64
->lo_init
[0] = info
.lo_init
[0];
1385 info64
->lo_init
[1] = info
.lo_init
[1];
1386 if (info
.lo_encrypt_type
== LO_CRYPT_CRYPTOAPI
)
1387 memcpy(info64
->lo_crypt_name
, info
.lo_name
, LO_NAME_SIZE
);
1389 memcpy(info64
->lo_file_name
, info
.lo_name
, LO_NAME_SIZE
);
1390 memcpy(info64
->lo_encrypt_key
, info
.lo_encrypt_key
, LO_KEY_SIZE
);
1395 * Transfer 64-bit loop info to 32-bit compatibility structure in userspace
1396 * - noinlined to reduce stack space usage in main part of driver
1399 loop_info64_to_compat(const struct loop_info64
*info64
,
1400 struct compat_loop_info __user
*arg
)
1402 struct compat_loop_info info
;
1404 memset(&info
, 0, sizeof(info
));
1405 info
.lo_number
= info64
->lo_number
;
1406 info
.lo_device
= info64
->lo_device
;
1407 info
.lo_inode
= info64
->lo_inode
;
1408 info
.lo_rdevice
= info64
->lo_rdevice
;
1409 info
.lo_offset
= info64
->lo_offset
;
1410 info
.lo_encrypt_type
= info64
->lo_encrypt_type
;
1411 info
.lo_encrypt_key_size
= info64
->lo_encrypt_key_size
;
1412 info
.lo_flags
= info64
->lo_flags
;
1413 info
.lo_init
[0] = info64
->lo_init
[0];
1414 info
.lo_init
[1] = info64
->lo_init
[1];
1415 if (info
.lo_encrypt_type
== LO_CRYPT_CRYPTOAPI
)
1416 memcpy(info
.lo_name
, info64
->lo_crypt_name
, LO_NAME_SIZE
);
1418 memcpy(info
.lo_name
, info64
->lo_file_name
, LO_NAME_SIZE
);
1419 memcpy(info
.lo_encrypt_key
, info64
->lo_encrypt_key
, LO_KEY_SIZE
);
1421 /* error in case values were truncated */
1422 if (info
.lo_device
!= info64
->lo_device
||
1423 info
.lo_rdevice
!= info64
->lo_rdevice
||
1424 info
.lo_inode
!= info64
->lo_inode
||
1425 info
.lo_offset
!= info64
->lo_offset
||
1426 info
.lo_init
[0] != info64
->lo_init
[0] ||
1427 info
.lo_init
[1] != info64
->lo_init
[1])
1430 if (copy_to_user(arg
, &info
, sizeof(info
)))
1436 loop_set_status_compat(struct loop_device
*lo
,
1437 const struct compat_loop_info __user
*arg
)
1439 struct loop_info64 info64
;
1442 ret
= loop_info64_from_compat(arg
, &info64
);
1445 return loop_set_status(lo
, &info64
);
1449 loop_get_status_compat(struct loop_device
*lo
,
1450 struct compat_loop_info __user
*arg
)
1452 struct loop_info64 info64
;
1458 err
= loop_get_status(lo
, &info64
);
1460 err
= loop_info64_to_compat(&info64
, arg
);
1464 static int lo_compat_ioctl(struct block_device
*bdev
, fmode_t mode
,
1465 unsigned int cmd
, unsigned long arg
)
1467 struct loop_device
*lo
= bdev
->bd_disk
->private_data
;
1471 case LOOP_SET_STATUS
:
1472 mutex_lock(&lo
->lo_ctl_mutex
);
1473 err
= loop_set_status_compat(
1474 lo
, (const struct compat_loop_info __user
*) arg
);
1475 mutex_unlock(&lo
->lo_ctl_mutex
);
1477 case LOOP_GET_STATUS
:
1478 mutex_lock(&lo
->lo_ctl_mutex
);
1479 err
= loop_get_status_compat(
1480 lo
, (struct compat_loop_info __user
*) arg
);
1481 mutex_unlock(&lo
->lo_ctl_mutex
);
1483 case LOOP_SET_CAPACITY
:
1485 case LOOP_GET_STATUS64
:
1486 case LOOP_SET_STATUS64
:
1487 arg
= (unsigned long) compat_ptr(arg
);
1489 case LOOP_CHANGE_FD
:
1490 err
= lo_ioctl(bdev
, mode
, cmd
, arg
);
1500 static int lo_open(struct block_device
*bdev
, fmode_t mode
)
1502 struct loop_device
*lo
= bdev
->bd_disk
->private_data
;
1504 mutex_lock(&loop_mutex
);
1505 mutex_lock(&lo
->lo_ctl_mutex
);
1507 mutex_unlock(&lo
->lo_ctl_mutex
);
1508 mutex_unlock(&loop_mutex
);
1513 static int lo_release(struct gendisk
*disk
, fmode_t mode
)
1515 struct loop_device
*lo
= disk
->private_data
;
1518 mutex_lock(&loop_mutex
);
1519 mutex_lock(&lo
->lo_ctl_mutex
);
1521 if (--lo
->lo_refcnt
)
1524 if (lo
->lo_flags
& LO_FLAGS_AUTOCLEAR
) {
1526 * In autoclear mode, stop the loop thread
1527 * and remove configuration after last close.
1529 err
= loop_clr_fd(lo
, NULL
);
1534 * Otherwise keep thread (if running) and config,
1535 * but flush possible ongoing bios in thread.
1541 mutex_unlock(&lo
->lo_ctl_mutex
);
1543 mutex_unlock(&loop_mutex
);
1547 static const struct block_device_operations lo_fops
= {
1548 .owner
= THIS_MODULE
,
1550 .release
= lo_release
,
1552 #ifdef CONFIG_COMPAT
1553 .compat_ioctl
= lo_compat_ioctl
,
1558 * And now the modules code and kernel interface.
1560 static int max_loop
;
1561 module_param(max_loop
, int, 0);
1562 MODULE_PARM_DESC(max_loop
, "Maximum number of loop devices");
1563 module_param(max_part
, int, 0);
1564 MODULE_PARM_DESC(max_part
, "Maximum number of partitions per loop device");
1565 MODULE_LICENSE("GPL");
1566 MODULE_ALIAS_BLOCKDEV_MAJOR(LOOP_MAJOR
);
1568 int loop_register_transfer(struct loop_func_table
*funcs
)
1570 unsigned int n
= funcs
->number
;
1572 if (n
>= MAX_LO_CRYPT
|| xfer_funcs
[n
])
1574 xfer_funcs
[n
] = funcs
;
1578 int loop_unregister_transfer(int number
)
1580 unsigned int n
= number
;
1581 struct loop_device
*lo
;
1582 struct loop_func_table
*xfer
;
1584 if (n
== 0 || n
>= MAX_LO_CRYPT
|| (xfer
= xfer_funcs
[n
]) == NULL
)
1587 xfer_funcs
[n
] = NULL
;
1589 list_for_each_entry(lo
, &loop_devices
, lo_list
) {
1590 mutex_lock(&lo
->lo_ctl_mutex
);
1592 if (lo
->lo_encryption
== xfer
)
1593 loop_release_xfer(lo
);
1595 mutex_unlock(&lo
->lo_ctl_mutex
);
1601 EXPORT_SYMBOL(loop_register_transfer
);
1602 EXPORT_SYMBOL(loop_unregister_transfer
);
1604 static struct loop_device
*loop_alloc(int i
)
1606 struct loop_device
*lo
;
1607 struct gendisk
*disk
;
1609 lo
= kzalloc(sizeof(*lo
), GFP_KERNEL
);
1613 lo
->lo_queue
= blk_alloc_queue(GFP_KERNEL
);
1617 disk
= lo
->lo_disk
= alloc_disk(1 << part_shift
);
1619 goto out_free_queue
;
1621 mutex_init(&lo
->lo_ctl_mutex
);
1623 lo
->lo_thread
= NULL
;
1624 init_waitqueue_head(&lo
->lo_event
);
1625 spin_lock_init(&lo
->lo_lock
);
1626 disk
->major
= LOOP_MAJOR
;
1627 disk
->first_minor
= i
<< part_shift
;
1628 disk
->fops
= &lo_fops
;
1629 disk
->private_data
= lo
;
1630 disk
->queue
= lo
->lo_queue
;
1631 sprintf(disk
->disk_name
, "loop%d", i
);
1635 blk_cleanup_queue(lo
->lo_queue
);
1642 static void loop_free(struct loop_device
*lo
)
1644 if (!lo
->lo_queue
->queue_lock
)
1645 lo
->lo_queue
->queue_lock
= &lo
->lo_queue
->__queue_lock
;
1647 blk_cleanup_queue(lo
->lo_queue
);
1648 put_disk(lo
->lo_disk
);
1649 list_del(&lo
->lo_list
);
1653 static struct loop_device
*loop_init_one(int i
)
1655 struct loop_device
*lo
;
1657 list_for_each_entry(lo
, &loop_devices
, lo_list
) {
1658 if (lo
->lo_number
== i
)
1664 add_disk(lo
->lo_disk
);
1665 list_add_tail(&lo
->lo_list
, &loop_devices
);
1670 static void loop_del_one(struct loop_device
*lo
)
1672 del_gendisk(lo
->lo_disk
);
1676 static struct kobject
*loop_probe(dev_t dev
, int *part
, void *data
)
1678 struct loop_device
*lo
;
1679 struct kobject
*kobj
;
1681 mutex_lock(&loop_devices_mutex
);
1682 lo
= loop_init_one(dev
& MINORMASK
);
1683 kobj
= lo
? get_disk(lo
->lo_disk
) : ERR_PTR(-ENOMEM
);
1684 mutex_unlock(&loop_devices_mutex
);
1690 static int __init
loop_init(void)
1693 unsigned long range
;
1694 struct loop_device
*lo
, *next
;
1697 * loop module now has a feature to instantiate underlying device
1698 * structure on-demand, provided that there is an access dev node.
1699 * However, this will not work well with user space tool that doesn't
1700 * know about such "feature". In order to not break any existing
1701 * tool, we do the following:
1703 * (1) if max_loop is specified, create that many upfront, and this
1704 * also becomes a hard limit.
1705 * (2) if max_loop is not specified, create 8 loop device on module
1706 * load, user can further extend loop device by create dev node
1707 * themselves and have kernel automatically instantiate actual
1713 part_shift
= fls(max_part
);
1715 if (max_loop
> 1UL << (MINORBITS
- part_shift
))
1723 range
= 1UL << (MINORBITS
- part_shift
);
1726 if (register_blkdev(LOOP_MAJOR
, "loop"))
1729 for (i
= 0; i
< nr
; i
++) {
1733 list_add_tail(&lo
->lo_list
, &loop_devices
);
1736 /* point of no return */
1738 list_for_each_entry(lo
, &loop_devices
, lo_list
)
1739 add_disk(lo
->lo_disk
);
1741 blk_register_region(MKDEV(LOOP_MAJOR
, 0), range
,
1742 THIS_MODULE
, loop_probe
, NULL
, NULL
);
1744 printk(KERN_INFO
"loop: module loaded\n");
1748 printk(KERN_INFO
"loop: out of memory\n");
1750 list_for_each_entry_safe(lo
, next
, &loop_devices
, lo_list
)
1753 unregister_blkdev(LOOP_MAJOR
, "loop");
1757 static void __exit
loop_exit(void)
1759 unsigned long range
;
1760 struct loop_device
*lo
, *next
;
1762 range
= max_loop
? max_loop
: 1UL << (MINORBITS
- part_shift
);
1764 list_for_each_entry_safe(lo
, next
, &loop_devices
, lo_list
)
1767 blk_unregister_region(MKDEV(LOOP_MAJOR
, 0), range
);
1768 unregister_blkdev(LOOP_MAJOR
, "loop");
1771 module_init(loop_init
);
1772 module_exit(loop_exit
);
1775 static int __init
max_loop_setup(char *str
)
1777 max_loop
= simple_strtol(str
, NULL
, 0);
1781 __setup("max_loop=", max_loop_setup
);