Linux Audio

Check our new training course

Loading...
v3.5.6
 
  1/*
  2 * ioctl.c - NILFS ioctl operations.
  3 *
  4 * Copyright (C) 2007, 2008 Nippon Telegraph and Telephone Corporation.
  5 *
  6 * This program is free software; you can redistribute it and/or modify
  7 * it under the terms of the GNU General Public License as published by
  8 * the Free Software Foundation; either version 2 of the License, or
  9 * (at your option) any later version.
 10 *
 11 * This program is distributed in the hope that it will be useful,
 12 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 13 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
 14 * GNU General Public License for more details.
 15 *
 16 * You should have received a copy of the GNU General Public License
 17 * along with this program; if not, write to the Free Software
 18 * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA  02110-1301  USA
 19 *
 20 * Written by Koji Sato <koji@osrg.net>.
 21 */
 22
 23#include <linux/fs.h>
 24#include <linux/wait.h>
 25#include <linux/slab.h>
 26#include <linux/capability.h>	/* capable() */
 27#include <linux/uaccess.h>	/* copy_from_user(), copy_to_user() */
 28#include <linux/vmalloc.h>
 29#include <linux/compat.h>	/* compat_ptr() */
 30#include <linux/mount.h>	/* mnt_want_write_file(), mnt_drop_write_file() */
 31#include <linux/buffer_head.h>
 32#include <linux/nilfs2_fs.h>
 33#include "nilfs.h"
 34#include "segment.h"
 35#include "bmap.h"
 36#include "cpfile.h"
 37#include "sufile.h"
 38#include "dat.h"
 39
 40
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 41static int nilfs_ioctl_wrap_copy(struct the_nilfs *nilfs,
 42				 struct nilfs_argv *argv, int dir,
 43				 ssize_t (*dofunc)(struct the_nilfs *,
 44						   __u64 *, int,
 45						   void *, size_t, size_t))
 46{
 47	void *buf;
 48	void __user *base = (void __user *)(unsigned long)argv->v_base;
 49	size_t maxmembs, total, n;
 50	ssize_t nr;
 51	int ret, i;
 52	__u64 pos, ppos;
 53
 54	if (argv->v_nmembs == 0)
 55		return 0;
 56
 57	if (argv->v_size > PAGE_SIZE)
 58		return -EINVAL;
 59
 
 
 
 
 
 
 
 
 60	buf = (void *)__get_free_pages(GFP_NOFS, 0);
 61	if (unlikely(!buf))
 62		return -ENOMEM;
 63	maxmembs = PAGE_SIZE / argv->v_size;
 64
 65	ret = 0;
 66	total = 0;
 67	pos = argv->v_index;
 68	for (i = 0; i < argv->v_nmembs; i += n) {
 69		n = (argv->v_nmembs - i < maxmembs) ?
 70			argv->v_nmembs - i : maxmembs;
 71		if ((dir & _IOC_WRITE) &&
 72		    copy_from_user(buf, base + argv->v_size * i,
 73				   argv->v_size * n)) {
 74			ret = -EFAULT;
 75			break;
 76		}
 77		ppos = pos;
 78		nr = dofunc(nilfs, &pos, argv->v_flags, buf, argv->v_size,
 79			       n);
 80		if (nr < 0) {
 81			ret = nr;
 82			break;
 83		}
 84		if ((dir & _IOC_READ) &&
 85		    copy_to_user(base + argv->v_size * i, buf,
 86				 argv->v_size * nr)) {
 87			ret = -EFAULT;
 88			break;
 89		}
 90		total += nr;
 91		if ((size_t)nr < n)
 92			break;
 93		if (pos == ppos)
 94			pos += n;
 95	}
 96	argv->v_nmembs = total;
 97
 98	free_pages((unsigned long)buf, 0);
 99	return ret;
100}
101
 
 
 
102static int nilfs_ioctl_getflags(struct inode *inode, void __user *argp)
103{
104	unsigned int flags = NILFS_I(inode)->i_flags & FS_FL_USER_VISIBLE;
105
106	return put_user(flags, (int __user *)argp);
107}
108
 
 
 
109static int nilfs_ioctl_setflags(struct inode *inode, struct file *filp,
110				void __user *argp)
111{
112	struct nilfs_transaction_info ti;
113	unsigned int flags, oldflags;
114	int ret;
115
116	if (!inode_owner_or_capable(inode))
117		return -EACCES;
118
119	if (get_user(flags, (int __user *)argp))
120		return -EFAULT;
121
122	ret = mnt_want_write_file(filp);
123	if (ret)
124		return ret;
125
126	flags = nilfs_mask_flags(inode->i_mode, flags);
127
128	mutex_lock(&inode->i_mutex);
129
130	oldflags = NILFS_I(inode)->i_flags;
131
132	/*
133	 * The IMMUTABLE and APPEND_ONLY flags can only be changed by the
134	 * relevant capability.
135	 */
136	ret = -EPERM;
137	if (((flags ^ oldflags) & (FS_APPEND_FL | FS_IMMUTABLE_FL)) &&
138	    !capable(CAP_LINUX_IMMUTABLE))
139		goto out;
140
141	ret = nilfs_transaction_begin(inode->i_sb, &ti, 0);
142	if (ret)
143		goto out;
144
145	NILFS_I(inode)->i_flags = (oldflags & ~FS_FL_USER_MODIFIABLE) |
146		(flags & FS_FL_USER_MODIFIABLE);
147
148	nilfs_set_inode_flags(inode);
149	inode->i_ctime = CURRENT_TIME;
150	if (IS_SYNC(inode))
151		nilfs_set_transaction_flag(NILFS_TI_SYNC);
152
153	nilfs_mark_inode_dirty(inode);
154	ret = nilfs_transaction_commit(inode->i_sb);
155out:
156	mutex_unlock(&inode->i_mutex);
157	mnt_drop_write_file(filp);
158	return ret;
159}
160
 
 
 
161static int nilfs_ioctl_getversion(struct inode *inode, void __user *argp)
162{
163	return put_user(inode->i_generation, (int __user *)argp);
164}
165
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
166static int nilfs_ioctl_change_cpmode(struct inode *inode, struct file *filp,
167				     unsigned int cmd, void __user *argp)
168{
169	struct the_nilfs *nilfs = inode->i_sb->s_fs_info;
170	struct nilfs_transaction_info ti;
171	struct nilfs_cpmode cpmode;
172	int ret;
173
174	if (!capable(CAP_SYS_ADMIN))
175		return -EPERM;
176
177	ret = mnt_want_write_file(filp);
178	if (ret)
179		return ret;
180
181	ret = -EFAULT;
182	if (copy_from_user(&cpmode, argp, sizeof(cpmode)))
183		goto out;
184
185	mutex_lock(&nilfs->ns_snapshot_mount_mutex);
186
187	nilfs_transaction_begin(inode->i_sb, &ti, 0);
188	ret = nilfs_cpfile_change_cpmode(
189		nilfs->ns_cpfile, cpmode.cm_cno, cpmode.cm_mode);
190	if (unlikely(ret < 0))
191		nilfs_transaction_abort(inode->i_sb);
192	else
193		nilfs_transaction_commit(inode->i_sb); /* never fails */
194
195	mutex_unlock(&nilfs->ns_snapshot_mount_mutex);
196out:
197	mnt_drop_write_file(filp);
198	return ret;
199}
200
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
201static int
202nilfs_ioctl_delete_checkpoint(struct inode *inode, struct file *filp,
203			      unsigned int cmd, void __user *argp)
204{
205	struct the_nilfs *nilfs = inode->i_sb->s_fs_info;
206	struct nilfs_transaction_info ti;
207	__u64 cno;
208	int ret;
209
210	if (!capable(CAP_SYS_ADMIN))
211		return -EPERM;
212
213	ret = mnt_want_write_file(filp);
214	if (ret)
215		return ret;
216
217	ret = -EFAULT;
218	if (copy_from_user(&cno, argp, sizeof(cno)))
219		goto out;
220
221	nilfs_transaction_begin(inode->i_sb, &ti, 0);
222	ret = nilfs_cpfile_delete_checkpoint(nilfs->ns_cpfile, cno);
223	if (unlikely(ret < 0))
224		nilfs_transaction_abort(inode->i_sb);
225	else
226		nilfs_transaction_commit(inode->i_sb); /* never fails */
227out:
228	mnt_drop_write_file(filp);
229	return ret;
230}
231
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
232static ssize_t
233nilfs_ioctl_do_get_cpinfo(struct the_nilfs *nilfs, __u64 *posp, int flags,
234			  void *buf, size_t size, size_t nmembs)
235{
236	int ret;
237
238	down_read(&nilfs->ns_segctor_sem);
239	ret = nilfs_cpfile_get_cpinfo(nilfs->ns_cpfile, posp, flags, buf,
240				      size, nmembs);
241	up_read(&nilfs->ns_segctor_sem);
242	return ret;
243}
244
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
245static int nilfs_ioctl_get_cpstat(struct inode *inode, struct file *filp,
246				  unsigned int cmd, void __user *argp)
247{
248	struct the_nilfs *nilfs = inode->i_sb->s_fs_info;
249	struct nilfs_cpstat cpstat;
250	int ret;
251
252	down_read(&nilfs->ns_segctor_sem);
253	ret = nilfs_cpfile_get_stat(nilfs->ns_cpfile, &cpstat);
254	up_read(&nilfs->ns_segctor_sem);
255	if (ret < 0)
256		return ret;
257
258	if (copy_to_user(argp, &cpstat, sizeof(cpstat)))
259		ret = -EFAULT;
260	return ret;
261}
262
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
263static ssize_t
264nilfs_ioctl_do_get_suinfo(struct the_nilfs *nilfs, __u64 *posp, int flags,
265			  void *buf, size_t size, size_t nmembs)
266{
267	int ret;
268
269	down_read(&nilfs->ns_segctor_sem);
270	ret = nilfs_sufile_get_suinfo(nilfs->ns_sufile, *posp, buf, size,
271				      nmembs);
272	up_read(&nilfs->ns_segctor_sem);
273	return ret;
274}
275
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
276static int nilfs_ioctl_get_sustat(struct inode *inode, struct file *filp,
277				  unsigned int cmd, void __user *argp)
278{
279	struct the_nilfs *nilfs = inode->i_sb->s_fs_info;
280	struct nilfs_sustat sustat;
281	int ret;
282
283	down_read(&nilfs->ns_segctor_sem);
284	ret = nilfs_sufile_get_stat(nilfs->ns_sufile, &sustat);
285	up_read(&nilfs->ns_segctor_sem);
286	if (ret < 0)
287		return ret;
288
289	if (copy_to_user(argp, &sustat, sizeof(sustat)))
290		ret = -EFAULT;
291	return ret;
292}
293
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
294static ssize_t
295nilfs_ioctl_do_get_vinfo(struct the_nilfs *nilfs, __u64 *posp, int flags,
296			 void *buf, size_t size, size_t nmembs)
297{
298	int ret;
299
300	down_read(&nilfs->ns_segctor_sem);
301	ret = nilfs_dat_get_vinfo(nilfs->ns_dat, buf, size, nmembs);
302	up_read(&nilfs->ns_segctor_sem);
303	return ret;
304}
305
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
306static ssize_t
307nilfs_ioctl_do_get_bdescs(struct the_nilfs *nilfs, __u64 *posp, int flags,
308			  void *buf, size_t size, size_t nmembs)
309{
310	struct nilfs_bmap *bmap = NILFS_I(nilfs->ns_dat)->i_bmap;
311	struct nilfs_bdesc *bdescs = buf;
312	int ret, i;
313
314	down_read(&nilfs->ns_segctor_sem);
315	for (i = 0; i < nmembs; i++) {
316		ret = nilfs_bmap_lookup_at_level(bmap,
317						 bdescs[i].bd_offset,
318						 bdescs[i].bd_level + 1,
319						 &bdescs[i].bd_blocknr);
320		if (ret < 0) {
321			if (ret != -ENOENT) {
322				up_read(&nilfs->ns_segctor_sem);
323				return ret;
324			}
325			bdescs[i].bd_blocknr = 0;
326		}
327	}
328	up_read(&nilfs->ns_segctor_sem);
329	return nmembs;
330}
331
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
332static int nilfs_ioctl_get_bdescs(struct inode *inode, struct file *filp,
333				  unsigned int cmd, void __user *argp)
334{
335	struct the_nilfs *nilfs = inode->i_sb->s_fs_info;
336	struct nilfs_argv argv;
337	int ret;
338
339	if (copy_from_user(&argv, argp, sizeof(argv)))
340		return -EFAULT;
341
342	if (argv.v_size != sizeof(struct nilfs_bdesc))
343		return -EINVAL;
344
345	ret = nilfs_ioctl_wrap_copy(nilfs, &argv, _IOC_DIR(cmd),
346				    nilfs_ioctl_do_get_bdescs);
347	if (ret < 0)
348		return ret;
349
350	if (copy_to_user(argp, &argv, sizeof(argv)))
351		ret = -EFAULT;
352	return ret;
353}
354
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
355static int nilfs_ioctl_move_inode_block(struct inode *inode,
356					struct nilfs_vdesc *vdesc,
357					struct list_head *buffers)
358{
359	struct buffer_head *bh;
360	int ret;
361
362	if (vdesc->vd_flags == 0)
363		ret = nilfs_gccache_submit_read_data(
364			inode, vdesc->vd_offset, vdesc->vd_blocknr,
365			vdesc->vd_vblocknr, &bh);
366	else
367		ret = nilfs_gccache_submit_read_node(
368			inode, vdesc->vd_blocknr, vdesc->vd_vblocknr, &bh);
369
370	if (unlikely(ret < 0)) {
371		if (ret == -ENOENT)
372			printk(KERN_CRIT
373			       "%s: invalid virtual block address (%s): "
374			       "ino=%llu, cno=%llu, offset=%llu, "
375			       "blocknr=%llu, vblocknr=%llu\n",
376			       __func__, vdesc->vd_flags ? "node" : "data",
377			       (unsigned long long)vdesc->vd_ino,
378			       (unsigned long long)vdesc->vd_cno,
379			       (unsigned long long)vdesc->vd_offset,
380			       (unsigned long long)vdesc->vd_blocknr,
381			       (unsigned long long)vdesc->vd_vblocknr);
382		return ret;
383	}
384	if (unlikely(!list_empty(&bh->b_assoc_buffers))) {
385		printk(KERN_CRIT "%s: conflicting %s buffer: ino=%llu, "
386		       "cno=%llu, offset=%llu, blocknr=%llu, vblocknr=%llu\n",
387		       __func__, vdesc->vd_flags ? "node" : "data",
388		       (unsigned long long)vdesc->vd_ino,
389		       (unsigned long long)vdesc->vd_cno,
390		       (unsigned long long)vdesc->vd_offset,
391		       (unsigned long long)vdesc->vd_blocknr,
392		       (unsigned long long)vdesc->vd_vblocknr);
393		brelse(bh);
394		return -EEXIST;
395	}
396	list_add_tail(&bh->b_assoc_buffers, buffers);
397	return 0;
398}
399
 
 
 
 
 
 
 
 
 
 
 
 
 
400static int nilfs_ioctl_move_blocks(struct super_block *sb,
401				   struct nilfs_argv *argv, void *buf)
402{
403	size_t nmembs = argv->v_nmembs;
404	struct the_nilfs *nilfs = sb->s_fs_info;
405	struct inode *inode;
406	struct nilfs_vdesc *vdesc;
407	struct buffer_head *bh, *n;
408	LIST_HEAD(buffers);
409	ino_t ino;
410	__u64 cno;
411	int i, ret;
412
413	for (i = 0, vdesc = buf; i < nmembs; ) {
414		ino = vdesc->vd_ino;
415		cno = vdesc->vd_cno;
416		inode = nilfs_iget_for_gc(sb, ino, cno);
417		if (IS_ERR(inode)) {
418			ret = PTR_ERR(inode);
419			goto failed;
420		}
421		if (list_empty(&NILFS_I(inode)->i_dirty)) {
422			/*
423			 * Add the inode to GC inode list. Garbage Collection
424			 * is serialized and no two processes manipulate the
425			 * list simultaneously.
426			 */
427			igrab(inode);
428			list_add(&NILFS_I(inode)->i_dirty,
429				 &nilfs->ns_gc_inodes);
430		}
431
432		do {
433			ret = nilfs_ioctl_move_inode_block(inode, vdesc,
434							   &buffers);
435			if (unlikely(ret < 0)) {
436				iput(inode);
437				goto failed;
438			}
439			vdesc++;
440		} while (++i < nmembs &&
441			 vdesc->vd_ino == ino && vdesc->vd_cno == cno);
442
443		iput(inode); /* The inode still remains in GC inode list */
444	}
445
446	list_for_each_entry_safe(bh, n, &buffers, b_assoc_buffers) {
447		ret = nilfs_gccache_wait_and_mark_dirty(bh);
448		if (unlikely(ret < 0)) {
449			WARN_ON(ret == -EEXIST);
450			goto failed;
451		}
452		list_del_init(&bh->b_assoc_buffers);
453		brelse(bh);
454	}
455	return nmembs;
456
457 failed:
458	list_for_each_entry_safe(bh, n, &buffers, b_assoc_buffers) {
459		list_del_init(&bh->b_assoc_buffers);
460		brelse(bh);
461	}
462	return ret;
463}
464
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
465static int nilfs_ioctl_delete_checkpoints(struct the_nilfs *nilfs,
466					  struct nilfs_argv *argv, void *buf)
467{
468	size_t nmembs = argv->v_nmembs;
469	struct inode *cpfile = nilfs->ns_cpfile;
470	struct nilfs_period *periods = buf;
471	int ret, i;
472
473	for (i = 0; i < nmembs; i++) {
474		ret = nilfs_cpfile_delete_checkpoints(
475			cpfile, periods[i].p_start, periods[i].p_end);
476		if (ret < 0)
477			return ret;
478	}
479	return nmembs;
480}
481
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
482static int nilfs_ioctl_free_vblocknrs(struct the_nilfs *nilfs,
483				      struct nilfs_argv *argv, void *buf)
484{
485	size_t nmembs = argv->v_nmembs;
486	int ret;
487
488	ret = nilfs_dat_freev(nilfs->ns_dat, buf, nmembs);
489
490	return (ret < 0) ? ret : nmembs;
491}
492
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
493static int nilfs_ioctl_mark_blocks_dirty(struct the_nilfs *nilfs,
494					 struct nilfs_argv *argv, void *buf)
495{
496	size_t nmembs = argv->v_nmembs;
497	struct nilfs_bmap *bmap = NILFS_I(nilfs->ns_dat)->i_bmap;
498	struct nilfs_bdesc *bdescs = buf;
 
499	int ret, i;
500
501	for (i = 0; i < nmembs; i++) {
502		/* XXX: use macro or inline func to check liveness */
503		ret = nilfs_bmap_lookup_at_level(bmap,
504						 bdescs[i].bd_offset,
505						 bdescs[i].bd_level + 1,
506						 &bdescs[i].bd_blocknr);
507		if (ret < 0) {
508			if (ret != -ENOENT)
509				return ret;
510			bdescs[i].bd_blocknr = 0;
511		}
512		if (bdescs[i].bd_blocknr != bdescs[i].bd_oblocknr)
513			/* skip dead block */
514			continue;
515		if (bdescs[i].bd_level == 0) {
516			ret = nilfs_mdt_mark_block_dirty(nilfs->ns_dat,
517							 bdescs[i].bd_offset);
518			if (ret < 0) {
 
519				WARN_ON(ret == -ENOENT);
520				return ret;
521			}
 
 
 
522		} else {
523			ret = nilfs_bmap_mark(bmap, bdescs[i].bd_offset,
524					      bdescs[i].bd_level);
525			if (ret < 0) {
526				WARN_ON(ret == -ENOENT);
527				return ret;
528			}
529		}
530	}
531	return nmembs;
532}
533
534int nilfs_ioctl_prepare_clean_segments(struct the_nilfs *nilfs,
535				       struct nilfs_argv *argv, void **kbufs)
536{
537	const char *msg;
538	int ret;
539
540	ret = nilfs_ioctl_delete_checkpoints(nilfs, &argv[1], kbufs[1]);
541	if (ret < 0) {
542		/*
543		 * can safely abort because checkpoints can be removed
544		 * independently.
545		 */
546		msg = "cannot delete checkpoints";
547		goto failed;
548	}
549	ret = nilfs_ioctl_free_vblocknrs(nilfs, &argv[2], kbufs[2]);
550	if (ret < 0) {
551		/*
552		 * can safely abort because DAT file is updated atomically
553		 * using a copy-on-write technique.
554		 */
555		msg = "cannot delete virtual blocks from DAT file";
556		goto failed;
557	}
558	ret = nilfs_ioctl_mark_blocks_dirty(nilfs, &argv[3], kbufs[3]);
559	if (ret < 0) {
560		/*
561		 * can safely abort because the operation is nondestructive.
562		 */
563		msg = "cannot mark copying blocks dirty";
564		goto failed;
565	}
566	return 0;
567
568 failed:
569	printk(KERN_ERR "NILFS: GC failed during preparation: %s: err=%d\n",
570	       msg, ret);
571	return ret;
572}
573
 
 
 
 
 
 
 
 
 
 
 
 
 
 
574static int nilfs_ioctl_clean_segments(struct inode *inode, struct file *filp,
575				      unsigned int cmd, void __user *argp)
576{
577	struct nilfs_argv argv[5];
578	static const size_t argsz[5] = {
579		sizeof(struct nilfs_vdesc),
580		sizeof(struct nilfs_period),
581		sizeof(__u64),
582		sizeof(struct nilfs_bdesc),
583		sizeof(__u64),
584	};
585	void __user *base;
586	void *kbufs[5];
587	struct the_nilfs *nilfs;
588	size_t len, nsegs;
589	int n, ret;
590
591	if (!capable(CAP_SYS_ADMIN))
592		return -EPERM;
593
594	ret = mnt_want_write_file(filp);
595	if (ret)
596		return ret;
597
598	ret = -EFAULT;
599	if (copy_from_user(argv, argp, sizeof(argv)))
600		goto out;
601
602	ret = -EINVAL;
603	nsegs = argv[4].v_nmembs;
604	if (argv[4].v_size != argsz[4])
605		goto out;
606	if (nsegs > UINT_MAX / sizeof(__u64))
607		goto out;
608
609	/*
610	 * argv[4] points to segment numbers this ioctl cleans.  We
611	 * use kmalloc() for its buffer because memory used for the
612	 * segment numbers is enough small.
613	 */
614	kbufs[4] = memdup_user((void __user *)(unsigned long)argv[4].v_base,
615			       nsegs * sizeof(__u64));
616	if (IS_ERR(kbufs[4])) {
617		ret = PTR_ERR(kbufs[4]);
618		goto out;
619	}
620	nilfs = inode->i_sb->s_fs_info;
621
622	for (n = 0; n < 4; n++) {
623		ret = -EINVAL;
624		if (argv[n].v_size != argsz[n])
625			goto out_free;
626
627		if (argv[n].v_nmembs > nsegs * nilfs->ns_blocks_per_segment)
628			goto out_free;
629
630		if (argv[n].v_nmembs >= UINT_MAX / argv[n].v_size)
631			goto out_free;
632
633		len = argv[n].v_size * argv[n].v_nmembs;
634		base = (void __user *)(unsigned long)argv[n].v_base;
635		if (len == 0) {
636			kbufs[n] = NULL;
637			continue;
638		}
639
640		kbufs[n] = vmalloc(len);
641		if (!kbufs[n]) {
642			ret = -ENOMEM;
643			goto out_free;
644		}
645		if (copy_from_user(kbufs[n], base, len)) {
646			ret = -EFAULT;
647			vfree(kbufs[n]);
648			goto out_free;
649		}
650	}
651
652	/*
653	 * nilfs_ioctl_move_blocks() will call nilfs_iget_for_gc(),
654	 * which will operates an inode list without blocking.
655	 * To protect the list from concurrent operations,
656	 * nilfs_ioctl_move_blocks should be atomic operation.
657	 */
658	if (test_and_set_bit(THE_NILFS_GC_RUNNING, &nilfs->ns_flags)) {
659		ret = -EBUSY;
660		goto out_free;
661	}
662
663	vfs_check_frozen(inode->i_sb, SB_FREEZE_WRITE);
664
665	ret = nilfs_ioctl_move_blocks(inode->i_sb, &argv[0], kbufs[0]);
666	if (ret < 0)
667		printk(KERN_ERR "NILFS: GC failed during preparation: "
668			"cannot read source blocks: err=%d\n", ret);
669	else
 
 
 
670		ret = nilfs_clean_segments(inode->i_sb, argv, kbufs);
 
671
672	nilfs_remove_all_gcinodes(nilfs);
673	clear_nilfs_gc_running(nilfs);
674
675out_free:
676	while (--n >= 0)
677		vfree(kbufs[n]);
678	kfree(kbufs[4]);
679out:
680	mnt_drop_write_file(filp);
681	return ret;
682}
683
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
684static int nilfs_ioctl_sync(struct inode *inode, struct file *filp,
685			    unsigned int cmd, void __user *argp)
686{
687	__u64 cno;
688	int ret;
689	struct the_nilfs *nilfs;
690
691	ret = nilfs_construct_segment(inode->i_sb);
692	if (ret < 0)
693		return ret;
694
695	nilfs = inode->i_sb->s_fs_info;
696	if (nilfs_test_opt(nilfs, BARRIER)) {
697		ret = blkdev_issue_flush(inode->i_sb->s_bdev, GFP_KERNEL, NULL);
698		if (ret == -EIO)
699			return ret;
700	}
701
702	if (argp != NULL) {
703		down_read(&nilfs->ns_segctor_sem);
704		cno = nilfs->ns_cno - 1;
705		up_read(&nilfs->ns_segctor_sem);
706		if (copy_to_user(argp, &cno, sizeof(cno)))
707			return -EFAULT;
708	}
709	return 0;
710}
711
 
 
 
 
 
 
 
 
712static int nilfs_ioctl_resize(struct inode *inode, struct file *filp,
713			      void __user *argp)
714{
715	__u64 newsize;
716	int ret = -EPERM;
717
718	if (!capable(CAP_SYS_ADMIN))
719		goto out;
720
721	ret = mnt_want_write_file(filp);
722	if (ret)
723		goto out;
724
725	ret = -EFAULT;
726	if (copy_from_user(&newsize, argp, sizeof(newsize)))
727		goto out_drop_write;
728
729	ret = nilfs_resize_fs(inode->i_sb, newsize);
730
731out_drop_write:
732	mnt_drop_write_file(filp);
733out:
734	return ret;
735}
736
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
737static int nilfs_ioctl_set_alloc_range(struct inode *inode, void __user *argp)
738{
739	struct the_nilfs *nilfs = inode->i_sb->s_fs_info;
740	__u64 range[2];
741	__u64 minseg, maxseg;
742	unsigned long segbytes;
743	int ret = -EPERM;
744
745	if (!capable(CAP_SYS_ADMIN))
746		goto out;
747
748	ret = -EFAULT;
749	if (copy_from_user(range, argp, sizeof(__u64[2])))
750		goto out;
751
752	ret = -ERANGE;
753	if (range[1] > i_size_read(inode->i_sb->s_bdev->bd_inode))
754		goto out;
755
756	segbytes = nilfs->ns_blocks_per_segment * nilfs->ns_blocksize;
757
758	minseg = range[0] + segbytes - 1;
759	do_div(minseg, segbytes);
760	maxseg = NILFS_SB2_OFFSET_BYTES(range[1]);
761	do_div(maxseg, segbytes);
762	maxseg--;
763
764	ret = nilfs_sufile_set_alloc_range(nilfs->ns_sufile, minseg, maxseg);
765out:
766	return ret;
767}
768
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
769static int nilfs_ioctl_get_info(struct inode *inode, struct file *filp,
770				unsigned int cmd, void __user *argp,
771				size_t membsz,
772				ssize_t (*dofunc)(struct the_nilfs *,
773						  __u64 *, int,
774						  void *, size_t, size_t))
775
776{
777	struct the_nilfs *nilfs = inode->i_sb->s_fs_info;
778	struct nilfs_argv argv;
779	int ret;
780
781	if (copy_from_user(&argv, argp, sizeof(argv)))
782		return -EFAULT;
783
784	if (argv.v_size < membsz)
785		return -EINVAL;
786
787	ret = nilfs_ioctl_wrap_copy(nilfs, &argv, _IOC_DIR(cmd), dofunc);
788	if (ret < 0)
789		return ret;
790
791	if (copy_to_user(argp, &argv, sizeof(argv)))
792		ret = -EFAULT;
793	return ret;
794}
795
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
796long nilfs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
797{
798	struct inode *inode = filp->f_dentry->d_inode;
799	void __user *argp = (void __user *)arg;
800
801	switch (cmd) {
802	case FS_IOC_GETFLAGS:
803		return nilfs_ioctl_getflags(inode, argp);
804	case FS_IOC_SETFLAGS:
805		return nilfs_ioctl_setflags(inode, filp, argp);
806	case FS_IOC_GETVERSION:
807		return nilfs_ioctl_getversion(inode, argp);
808	case NILFS_IOCTL_CHANGE_CPMODE:
809		return nilfs_ioctl_change_cpmode(inode, filp, cmd, argp);
810	case NILFS_IOCTL_DELETE_CHECKPOINT:
811		return nilfs_ioctl_delete_checkpoint(inode, filp, cmd, argp);
812	case NILFS_IOCTL_GET_CPINFO:
813		return nilfs_ioctl_get_info(inode, filp, cmd, argp,
814					    sizeof(struct nilfs_cpinfo),
815					    nilfs_ioctl_do_get_cpinfo);
816	case NILFS_IOCTL_GET_CPSTAT:
817		return nilfs_ioctl_get_cpstat(inode, filp, cmd, argp);
818	case NILFS_IOCTL_GET_SUINFO:
819		return nilfs_ioctl_get_info(inode, filp, cmd, argp,
820					    sizeof(struct nilfs_suinfo),
821					    nilfs_ioctl_do_get_suinfo);
 
 
822	case NILFS_IOCTL_GET_SUSTAT:
823		return nilfs_ioctl_get_sustat(inode, filp, cmd, argp);
824	case NILFS_IOCTL_GET_VINFO:
825		return nilfs_ioctl_get_info(inode, filp, cmd, argp,
826					    sizeof(struct nilfs_vinfo),
827					    nilfs_ioctl_do_get_vinfo);
828	case NILFS_IOCTL_GET_BDESCS:
829		return nilfs_ioctl_get_bdescs(inode, filp, cmd, argp);
830	case NILFS_IOCTL_CLEAN_SEGMENTS:
831		return nilfs_ioctl_clean_segments(inode, filp, cmd, argp);
832	case NILFS_IOCTL_SYNC:
833		return nilfs_ioctl_sync(inode, filp, cmd, argp);
834	case NILFS_IOCTL_RESIZE:
835		return nilfs_ioctl_resize(inode, filp, argp);
836	case NILFS_IOCTL_SET_ALLOC_RANGE:
837		return nilfs_ioctl_set_alloc_range(inode, argp);
 
 
838	default:
839		return -ENOTTY;
840	}
841}
842
843#ifdef CONFIG_COMPAT
844long nilfs_compat_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
845{
846	switch (cmd) {
847	case FS_IOC32_GETFLAGS:
848		cmd = FS_IOC_GETFLAGS;
849		break;
850	case FS_IOC32_SETFLAGS:
851		cmd = FS_IOC_SETFLAGS;
852		break;
853	case FS_IOC32_GETVERSION:
854		cmd = FS_IOC_GETVERSION;
855		break;
856	case NILFS_IOCTL_CHANGE_CPMODE:
857	case NILFS_IOCTL_DELETE_CHECKPOINT:
858	case NILFS_IOCTL_GET_CPINFO:
859	case NILFS_IOCTL_GET_CPSTAT:
860	case NILFS_IOCTL_GET_SUINFO:
 
861	case NILFS_IOCTL_GET_SUSTAT:
862	case NILFS_IOCTL_GET_VINFO:
863	case NILFS_IOCTL_GET_BDESCS:
864	case NILFS_IOCTL_CLEAN_SEGMENTS:
865	case NILFS_IOCTL_SYNC:
866	case NILFS_IOCTL_RESIZE:
867	case NILFS_IOCTL_SET_ALLOC_RANGE:
 
868		break;
869	default:
870		return -ENOIOCTLCMD;
871	}
872	return nilfs_ioctl(filp, cmd, (unsigned long)compat_ptr(arg));
873}
874#endif
v5.9
   1// SPDX-License-Identifier: GPL-2.0+
   2/*
   3 * ioctl.c - NILFS ioctl operations.
   4 *
   5 * Copyright (C) 2007, 2008 Nippon Telegraph and Telephone Corporation.
   6 *
   7 * Written by Koji Sato.
 
 
 
 
 
 
 
 
 
 
 
 
 
 
   8 */
   9
  10#include <linux/fs.h>
  11#include <linux/wait.h>
  12#include <linux/slab.h>
  13#include <linux/capability.h>	/* capable() */
  14#include <linux/uaccess.h>	/* copy_from_user(), copy_to_user() */
  15#include <linux/vmalloc.h>
  16#include <linux/compat.h>	/* compat_ptr() */
  17#include <linux/mount.h>	/* mnt_want_write_file(), mnt_drop_write_file() */
  18#include <linux/buffer_head.h>
 
  19#include "nilfs.h"
  20#include "segment.h"
  21#include "bmap.h"
  22#include "cpfile.h"
  23#include "sufile.h"
  24#include "dat.h"
  25
  26/**
  27 * nilfs_ioctl_wrap_copy - wrapping function of get/set metadata info
  28 * @nilfs: nilfs object
  29 * @argv: vector of arguments from userspace
  30 * @dir: set of direction flags
  31 * @dofunc: concrete function of get/set metadata info
  32 *
  33 * Description: nilfs_ioctl_wrap_copy() gets/sets metadata info by means of
  34 * calling dofunc() function on the basis of @argv argument.
  35 *
  36 * Return Value: On success, 0 is returned and requested metadata info
  37 * is copied into userspace. On error, one of the following
  38 * negative error codes is returned.
  39 *
  40 * %-EINVAL - Invalid arguments from userspace.
  41 *
  42 * %-ENOMEM - Insufficient amount of memory available.
  43 *
  44 * %-EFAULT - Failure during execution of requested operation.
  45 */
  46static int nilfs_ioctl_wrap_copy(struct the_nilfs *nilfs,
  47				 struct nilfs_argv *argv, int dir,
  48				 ssize_t (*dofunc)(struct the_nilfs *,
  49						   __u64 *, int,
  50						   void *, size_t, size_t))
  51{
  52	void *buf;
  53	void __user *base = (void __user *)(unsigned long)argv->v_base;
  54	size_t maxmembs, total, n;
  55	ssize_t nr;
  56	int ret, i;
  57	__u64 pos, ppos;
  58
  59	if (argv->v_nmembs == 0)
  60		return 0;
  61
  62	if (argv->v_size > PAGE_SIZE)
  63		return -EINVAL;
  64
  65	/*
  66	 * Reject pairs of a start item position (argv->v_index) and a
  67	 * total count (argv->v_nmembs) which leads position 'pos' to
  68	 * overflow by the increment at the end of the loop.
  69	 */
  70	if (argv->v_index > ~(__u64)0 - argv->v_nmembs)
  71		return -EINVAL;
  72
  73	buf = (void *)__get_free_pages(GFP_NOFS, 0);
  74	if (unlikely(!buf))
  75		return -ENOMEM;
  76	maxmembs = PAGE_SIZE / argv->v_size;
  77
  78	ret = 0;
  79	total = 0;
  80	pos = argv->v_index;
  81	for (i = 0; i < argv->v_nmembs; i += n) {
  82		n = (argv->v_nmembs - i < maxmembs) ?
  83			argv->v_nmembs - i : maxmembs;
  84		if ((dir & _IOC_WRITE) &&
  85		    copy_from_user(buf, base + argv->v_size * i,
  86				   argv->v_size * n)) {
  87			ret = -EFAULT;
  88			break;
  89		}
  90		ppos = pos;
  91		nr = dofunc(nilfs, &pos, argv->v_flags, buf, argv->v_size,
  92			       n);
  93		if (nr < 0) {
  94			ret = nr;
  95			break;
  96		}
  97		if ((dir & _IOC_READ) &&
  98		    copy_to_user(base + argv->v_size * i, buf,
  99				 argv->v_size * nr)) {
 100			ret = -EFAULT;
 101			break;
 102		}
 103		total += nr;
 104		if ((size_t)nr < n)
 105			break;
 106		if (pos == ppos)
 107			pos += n;
 108	}
 109	argv->v_nmembs = total;
 110
 111	free_pages((unsigned long)buf, 0);
 112	return ret;
 113}
 114
 115/**
 116 * nilfs_ioctl_getflags - ioctl to support lsattr
 117 */
 118static int nilfs_ioctl_getflags(struct inode *inode, void __user *argp)
 119{
 120	unsigned int flags = NILFS_I(inode)->i_flags & FS_FL_USER_VISIBLE;
 121
 122	return put_user(flags, (int __user *)argp);
 123}
 124
 125/**
 126 * nilfs_ioctl_setflags - ioctl to support chattr
 127 */
 128static int nilfs_ioctl_setflags(struct inode *inode, struct file *filp,
 129				void __user *argp)
 130{
 131	struct nilfs_transaction_info ti;
 132	unsigned int flags, oldflags;
 133	int ret;
 134
 135	if (!inode_owner_or_capable(inode))
 136		return -EACCES;
 137
 138	if (get_user(flags, (int __user *)argp))
 139		return -EFAULT;
 140
 141	ret = mnt_want_write_file(filp);
 142	if (ret)
 143		return ret;
 144
 145	flags = nilfs_mask_flags(inode->i_mode, flags);
 146
 147	inode_lock(inode);
 148
 149	oldflags = NILFS_I(inode)->i_flags;
 150
 151	ret = vfs_ioc_setflags_prepare(inode, oldflags, flags);
 152	if (ret)
 
 
 
 
 
 153		goto out;
 154
 155	ret = nilfs_transaction_begin(inode->i_sb, &ti, 0);
 156	if (ret)
 157		goto out;
 158
 159	NILFS_I(inode)->i_flags = (oldflags & ~FS_FL_USER_MODIFIABLE) |
 160		(flags & FS_FL_USER_MODIFIABLE);
 161
 162	nilfs_set_inode_flags(inode);
 163	inode->i_ctime = current_time(inode);
 164	if (IS_SYNC(inode))
 165		nilfs_set_transaction_flag(NILFS_TI_SYNC);
 166
 167	nilfs_mark_inode_dirty(inode);
 168	ret = nilfs_transaction_commit(inode->i_sb);
 169out:
 170	inode_unlock(inode);
 171	mnt_drop_write_file(filp);
 172	return ret;
 173}
 174
 175/**
 176 * nilfs_ioctl_getversion - get info about a file's version (generation number)
 177 */
 178static int nilfs_ioctl_getversion(struct inode *inode, void __user *argp)
 179{
 180	return put_user(inode->i_generation, (int __user *)argp);
 181}
 182
 183/**
 184 * nilfs_ioctl_change_cpmode - change checkpoint mode (checkpoint/snapshot)
 185 * @inode: inode object
 186 * @filp: file object
 187 * @cmd: ioctl's request code
 188 * @argp: pointer on argument from userspace
 189 *
 190 * Description: nilfs_ioctl_change_cpmode() function changes mode of
 191 * given checkpoint between checkpoint and snapshot state. This ioctl
 192 * is used in chcp and mkcp utilities.
 193 *
 194 * Return Value: On success, 0 is returned and mode of a checkpoint is
 195 * changed. On error, one of the following negative error codes
 196 * is returned.
 197 *
 198 * %-EPERM - Operation not permitted.
 199 *
 200 * %-EFAULT - Failure during checkpoint mode changing.
 201 */
 202static int nilfs_ioctl_change_cpmode(struct inode *inode, struct file *filp,
 203				     unsigned int cmd, void __user *argp)
 204{
 205	struct the_nilfs *nilfs = inode->i_sb->s_fs_info;
 206	struct nilfs_transaction_info ti;
 207	struct nilfs_cpmode cpmode;
 208	int ret;
 209
 210	if (!capable(CAP_SYS_ADMIN))
 211		return -EPERM;
 212
 213	ret = mnt_want_write_file(filp);
 214	if (ret)
 215		return ret;
 216
 217	ret = -EFAULT;
 218	if (copy_from_user(&cpmode, argp, sizeof(cpmode)))
 219		goto out;
 220
 221	mutex_lock(&nilfs->ns_snapshot_mount_mutex);
 222
 223	nilfs_transaction_begin(inode->i_sb, &ti, 0);
 224	ret = nilfs_cpfile_change_cpmode(
 225		nilfs->ns_cpfile, cpmode.cm_cno, cpmode.cm_mode);
 226	if (unlikely(ret < 0))
 227		nilfs_transaction_abort(inode->i_sb);
 228	else
 229		nilfs_transaction_commit(inode->i_sb); /* never fails */
 230
 231	mutex_unlock(&nilfs->ns_snapshot_mount_mutex);
 232out:
 233	mnt_drop_write_file(filp);
 234	return ret;
 235}
 236
 237/**
 238 * nilfs_ioctl_delete_checkpoint - remove checkpoint
 239 * @inode: inode object
 240 * @filp: file object
 241 * @cmd: ioctl's request code
 242 * @argp: pointer on argument from userspace
 243 *
 244 * Description: nilfs_ioctl_delete_checkpoint() function removes
 245 * checkpoint from NILFS2 file system. This ioctl is used in rmcp
 246 * utility.
 247 *
 248 * Return Value: On success, 0 is returned and a checkpoint is
 249 * removed. On error, one of the following negative error codes
 250 * is returned.
 251 *
 252 * %-EPERM - Operation not permitted.
 253 *
 254 * %-EFAULT - Failure during checkpoint removing.
 255 */
 256static int
 257nilfs_ioctl_delete_checkpoint(struct inode *inode, struct file *filp,
 258			      unsigned int cmd, void __user *argp)
 259{
 260	struct the_nilfs *nilfs = inode->i_sb->s_fs_info;
 261	struct nilfs_transaction_info ti;
 262	__u64 cno;
 263	int ret;
 264
 265	if (!capable(CAP_SYS_ADMIN))
 266		return -EPERM;
 267
 268	ret = mnt_want_write_file(filp);
 269	if (ret)
 270		return ret;
 271
 272	ret = -EFAULT;
 273	if (copy_from_user(&cno, argp, sizeof(cno)))
 274		goto out;
 275
 276	nilfs_transaction_begin(inode->i_sb, &ti, 0);
 277	ret = nilfs_cpfile_delete_checkpoint(nilfs->ns_cpfile, cno);
 278	if (unlikely(ret < 0))
 279		nilfs_transaction_abort(inode->i_sb);
 280	else
 281		nilfs_transaction_commit(inode->i_sb); /* never fails */
 282out:
 283	mnt_drop_write_file(filp);
 284	return ret;
 285}
 286
 287/**
 288 * nilfs_ioctl_do_get_cpinfo - callback method getting info about checkpoints
 289 * @nilfs: nilfs object
 290 * @posp: pointer on array of checkpoint's numbers
 291 * @flags: checkpoint mode (checkpoint or snapshot)
 292 * @buf: buffer for storing checkponts' info
 293 * @size: size in bytes of one checkpoint info item in array
 294 * @nmembs: number of checkpoints in array (numbers and infos)
 295 *
 296 * Description: nilfs_ioctl_do_get_cpinfo() function returns info about
 297 * requested checkpoints. The NILFS_IOCTL_GET_CPINFO ioctl is used in
 298 * lscp utility and by nilfs_cleanerd daemon.
 299 *
 300 * Return value: count of nilfs_cpinfo structures in output buffer.
 301 */
 302static ssize_t
 303nilfs_ioctl_do_get_cpinfo(struct the_nilfs *nilfs, __u64 *posp, int flags,
 304			  void *buf, size_t size, size_t nmembs)
 305{
 306	int ret;
 307
 308	down_read(&nilfs->ns_segctor_sem);
 309	ret = nilfs_cpfile_get_cpinfo(nilfs->ns_cpfile, posp, flags, buf,
 310				      size, nmembs);
 311	up_read(&nilfs->ns_segctor_sem);
 312	return ret;
 313}
 314
 315/**
 316 * nilfs_ioctl_get_cpstat - get checkpoints statistics
 317 * @inode: inode object
 318 * @filp: file object
 319 * @cmd: ioctl's request code
 320 * @argp: pointer on argument from userspace
 321 *
 322 * Description: nilfs_ioctl_get_cpstat() returns information about checkpoints.
 323 * The NILFS_IOCTL_GET_CPSTAT ioctl is used by lscp, rmcp utilities
 324 * and by nilfs_cleanerd daemon.
 325 *
 326 * Return Value: On success, 0 is returned, and checkpoints information is
 327 * copied into userspace pointer @argp. On error, one of the following
 328 * negative error codes is returned.
 329 *
 330 * %-EIO - I/O error.
 331 *
 332 * %-ENOMEM - Insufficient amount of memory available.
 333 *
 334 * %-EFAULT - Failure during getting checkpoints statistics.
 335 */
 336static int nilfs_ioctl_get_cpstat(struct inode *inode, struct file *filp,
 337				  unsigned int cmd, void __user *argp)
 338{
 339	struct the_nilfs *nilfs = inode->i_sb->s_fs_info;
 340	struct nilfs_cpstat cpstat;
 341	int ret;
 342
 343	down_read(&nilfs->ns_segctor_sem);
 344	ret = nilfs_cpfile_get_stat(nilfs->ns_cpfile, &cpstat);
 345	up_read(&nilfs->ns_segctor_sem);
 346	if (ret < 0)
 347		return ret;
 348
 349	if (copy_to_user(argp, &cpstat, sizeof(cpstat)))
 350		ret = -EFAULT;
 351	return ret;
 352}
 353
 354/**
 355 * nilfs_ioctl_do_get_suinfo - callback method getting segment usage info
 356 * @nilfs: nilfs object
 357 * @posp: pointer on array of segment numbers
 358 * @flags: *not used*
 359 * @buf: buffer for storing suinfo array
 360 * @size: size in bytes of one suinfo item in array
 361 * @nmembs: count of segment numbers and suinfos in array
 362 *
 363 * Description: nilfs_ioctl_do_get_suinfo() function returns segment usage
 364 * info about requested segments. The NILFS_IOCTL_GET_SUINFO ioctl is used
 365 * in lssu, nilfs_resize utilities and by nilfs_cleanerd daemon.
 366 *
 367 * Return value: count of nilfs_suinfo structures in output buffer.
 368 */
 369static ssize_t
 370nilfs_ioctl_do_get_suinfo(struct the_nilfs *nilfs, __u64 *posp, int flags,
 371			  void *buf, size_t size, size_t nmembs)
 372{
 373	int ret;
 374
 375	down_read(&nilfs->ns_segctor_sem);
 376	ret = nilfs_sufile_get_suinfo(nilfs->ns_sufile, *posp, buf, size,
 377				      nmembs);
 378	up_read(&nilfs->ns_segctor_sem);
 379	return ret;
 380}
 381
 382/**
 383 * nilfs_ioctl_get_sustat - get segment usage statistics
 384 * @inode: inode object
 385 * @filp: file object
 386 * @cmd: ioctl's request code
 387 * @argp: pointer on argument from userspace
 388 *
 389 * Description: nilfs_ioctl_get_sustat() returns segment usage statistics.
 390 * The NILFS_IOCTL_GET_SUSTAT ioctl is used in lssu, nilfs_resize utilities
 391 * and by nilfs_cleanerd daemon.
 392 *
 393 * Return Value: On success, 0 is returned, and segment usage information is
 394 * copied into userspace pointer @argp. On error, one of the following
 395 * negative error codes is returned.
 396 *
 397 * %-EIO - I/O error.
 398 *
 399 * %-ENOMEM - Insufficient amount of memory available.
 400 *
 401 * %-EFAULT - Failure during getting segment usage statistics.
 402 */
 403static int nilfs_ioctl_get_sustat(struct inode *inode, struct file *filp,
 404				  unsigned int cmd, void __user *argp)
 405{
 406	struct the_nilfs *nilfs = inode->i_sb->s_fs_info;
 407	struct nilfs_sustat sustat;
 408	int ret;
 409
 410	down_read(&nilfs->ns_segctor_sem);
 411	ret = nilfs_sufile_get_stat(nilfs->ns_sufile, &sustat);
 412	up_read(&nilfs->ns_segctor_sem);
 413	if (ret < 0)
 414		return ret;
 415
 416	if (copy_to_user(argp, &sustat, sizeof(sustat)))
 417		ret = -EFAULT;
 418	return ret;
 419}
 420
 421/**
 422 * nilfs_ioctl_do_get_vinfo - callback method getting virtual blocks info
 423 * @nilfs: nilfs object
 424 * @posp: *not used*
 425 * @flags: *not used*
 426 * @buf: buffer for storing array of nilfs_vinfo structures
 427 * @size: size in bytes of one vinfo item in array
 428 * @nmembs: count of vinfos in array
 429 *
 430 * Description: nilfs_ioctl_do_get_vinfo() function returns information
 431 * on virtual block addresses. The NILFS_IOCTL_GET_VINFO ioctl is used
 432 * by nilfs_cleanerd daemon.
 433 *
 434 * Return value: count of nilfs_vinfo structures in output buffer.
 435 */
 436static ssize_t
 437nilfs_ioctl_do_get_vinfo(struct the_nilfs *nilfs, __u64 *posp, int flags,
 438			 void *buf, size_t size, size_t nmembs)
 439{
 440	int ret;
 441
 442	down_read(&nilfs->ns_segctor_sem);
 443	ret = nilfs_dat_get_vinfo(nilfs->ns_dat, buf, size, nmembs);
 444	up_read(&nilfs->ns_segctor_sem);
 445	return ret;
 446}
 447
 448/**
 449 * nilfs_ioctl_do_get_bdescs - callback method getting disk block descriptors
 450 * @nilfs: nilfs object
 451 * @posp: *not used*
 452 * @flags: *not used*
 453 * @buf: buffer for storing array of nilfs_bdesc structures
 454 * @size: size in bytes of one bdesc item in array
 455 * @nmembs: count of bdescs in array
 456 *
 457 * Description: nilfs_ioctl_do_get_bdescs() function returns information
 458 * about descriptors of disk block numbers. The NILFS_IOCTL_GET_BDESCS ioctl
 459 * is used by nilfs_cleanerd daemon.
 460 *
 461 * Return value: count of nilfs_bdescs structures in output buffer.
 462 */
 463static ssize_t
 464nilfs_ioctl_do_get_bdescs(struct the_nilfs *nilfs, __u64 *posp, int flags,
 465			  void *buf, size_t size, size_t nmembs)
 466{
 467	struct nilfs_bmap *bmap = NILFS_I(nilfs->ns_dat)->i_bmap;
 468	struct nilfs_bdesc *bdescs = buf;
 469	int ret, i;
 470
 471	down_read(&nilfs->ns_segctor_sem);
 472	for (i = 0; i < nmembs; i++) {
 473		ret = nilfs_bmap_lookup_at_level(bmap,
 474						 bdescs[i].bd_offset,
 475						 bdescs[i].bd_level + 1,
 476						 &bdescs[i].bd_blocknr);
 477		if (ret < 0) {
 478			if (ret != -ENOENT) {
 479				up_read(&nilfs->ns_segctor_sem);
 480				return ret;
 481			}
 482			bdescs[i].bd_blocknr = 0;
 483		}
 484	}
 485	up_read(&nilfs->ns_segctor_sem);
 486	return nmembs;
 487}
 488
 489/**
 490 * nilfs_ioctl_get_bdescs - get disk block descriptors
 491 * @inode: inode object
 492 * @filp: file object
 493 * @cmd: ioctl's request code
 494 * @argp: pointer on argument from userspace
 495 *
 496 * Description: nilfs_ioctl_do_get_bdescs() function returns information
 497 * about descriptors of disk block numbers. The NILFS_IOCTL_GET_BDESCS ioctl
 498 * is used by nilfs_cleanerd daemon.
 499 *
 500 * Return Value: On success, 0 is returned, and disk block descriptors are
 501 * copied into userspace pointer @argp. On error, one of the following
 502 * negative error codes is returned.
 503 *
 504 * %-EINVAL - Invalid arguments from userspace.
 505 *
 506 * %-EIO - I/O error.
 507 *
 508 * %-ENOMEM - Insufficient amount of memory available.
 509 *
 510 * %-EFAULT - Failure during getting disk block descriptors.
 511 */
 512static int nilfs_ioctl_get_bdescs(struct inode *inode, struct file *filp,
 513				  unsigned int cmd, void __user *argp)
 514{
 515	struct the_nilfs *nilfs = inode->i_sb->s_fs_info;
 516	struct nilfs_argv argv;
 517	int ret;
 518
 519	if (copy_from_user(&argv, argp, sizeof(argv)))
 520		return -EFAULT;
 521
 522	if (argv.v_size != sizeof(struct nilfs_bdesc))
 523		return -EINVAL;
 524
 525	ret = nilfs_ioctl_wrap_copy(nilfs, &argv, _IOC_DIR(cmd),
 526				    nilfs_ioctl_do_get_bdescs);
 527	if (ret < 0)
 528		return ret;
 529
 530	if (copy_to_user(argp, &argv, sizeof(argv)))
 531		ret = -EFAULT;
 532	return ret;
 533}
 534
 535/**
 536 * nilfs_ioctl_move_inode_block - prepare data/node block for moving by GC
 537 * @inode: inode object
 538 * @vdesc: descriptor of virtual block number
 539 * @buffers: list of moving buffers
 540 *
 541 * Description: nilfs_ioctl_move_inode_block() function registers data/node
 542 * buffer in the GC pagecache and submit read request.
 543 *
 544 * Return Value: On success, 0 is returned. On error, one of the following
 545 * negative error codes is returned.
 546 *
 547 * %-EIO - I/O error.
 548 *
 549 * %-ENOMEM - Insufficient amount of memory available.
 550 *
 551 * %-ENOENT - Requested block doesn't exist.
 552 *
 553 * %-EEXIST - Blocks conflict is detected.
 554 */
 555static int nilfs_ioctl_move_inode_block(struct inode *inode,
 556					struct nilfs_vdesc *vdesc,
 557					struct list_head *buffers)
 558{
 559	struct buffer_head *bh;
 560	int ret;
 561
 562	if (vdesc->vd_flags == 0)
 563		ret = nilfs_gccache_submit_read_data(
 564			inode, vdesc->vd_offset, vdesc->vd_blocknr,
 565			vdesc->vd_vblocknr, &bh);
 566	else
 567		ret = nilfs_gccache_submit_read_node(
 568			inode, vdesc->vd_blocknr, vdesc->vd_vblocknr, &bh);
 569
 570	if (unlikely(ret < 0)) {
 571		if (ret == -ENOENT)
 572			nilfs_crit(inode->i_sb,
 573				   "%s: invalid virtual block address (%s): ino=%llu, cno=%llu, offset=%llu, blocknr=%llu, vblocknr=%llu",
 574				   __func__, vdesc->vd_flags ? "node" : "data",
 575				   (unsigned long long)vdesc->vd_ino,
 576				   (unsigned long long)vdesc->vd_cno,
 577				   (unsigned long long)vdesc->vd_offset,
 578				   (unsigned long long)vdesc->vd_blocknr,
 579				   (unsigned long long)vdesc->vd_vblocknr);
 
 
 580		return ret;
 581	}
 582	if (unlikely(!list_empty(&bh->b_assoc_buffers))) {
 583		nilfs_crit(inode->i_sb,
 584			   "%s: conflicting %s buffer: ino=%llu, cno=%llu, offset=%llu, blocknr=%llu, vblocknr=%llu",
 585			   __func__, vdesc->vd_flags ? "node" : "data",
 586			   (unsigned long long)vdesc->vd_ino,
 587			   (unsigned long long)vdesc->vd_cno,
 588			   (unsigned long long)vdesc->vd_offset,
 589			   (unsigned long long)vdesc->vd_blocknr,
 590			   (unsigned long long)vdesc->vd_vblocknr);
 591		brelse(bh);
 592		return -EEXIST;
 593	}
 594	list_add_tail(&bh->b_assoc_buffers, buffers);
 595	return 0;
 596}
 597
 598/**
 599 * nilfs_ioctl_move_blocks - move valid inode's blocks during garbage collection
 600 * @sb: superblock object
 601 * @argv: vector of arguments from userspace
 602 * @buf: array of nilfs_vdesc structures
 603 *
 604 * Description: nilfs_ioctl_move_blocks() function reads valid data/node
 605 * blocks that garbage collector specified with the array of nilfs_vdesc
 606 * structures and stores them into page caches of GC inodes.
 607 *
 608 * Return Value: Number of processed nilfs_vdesc structures or
 609 * error code, otherwise.
 610 */
 611static int nilfs_ioctl_move_blocks(struct super_block *sb,
 612				   struct nilfs_argv *argv, void *buf)
 613{
 614	size_t nmembs = argv->v_nmembs;
 615	struct the_nilfs *nilfs = sb->s_fs_info;
 616	struct inode *inode;
 617	struct nilfs_vdesc *vdesc;
 618	struct buffer_head *bh, *n;
 619	LIST_HEAD(buffers);
 620	ino_t ino;
 621	__u64 cno;
 622	int i, ret;
 623
 624	for (i = 0, vdesc = buf; i < nmembs; ) {
 625		ino = vdesc->vd_ino;
 626		cno = vdesc->vd_cno;
 627		inode = nilfs_iget_for_gc(sb, ino, cno);
 628		if (IS_ERR(inode)) {
 629			ret = PTR_ERR(inode);
 630			goto failed;
 631		}
 632		if (list_empty(&NILFS_I(inode)->i_dirty)) {
 633			/*
 634			 * Add the inode to GC inode list. Garbage Collection
 635			 * is serialized and no two processes manipulate the
 636			 * list simultaneously.
 637			 */
 638			igrab(inode);
 639			list_add(&NILFS_I(inode)->i_dirty,
 640				 &nilfs->ns_gc_inodes);
 641		}
 642
 643		do {
 644			ret = nilfs_ioctl_move_inode_block(inode, vdesc,
 645							   &buffers);
 646			if (unlikely(ret < 0)) {
 647				iput(inode);
 648				goto failed;
 649			}
 650			vdesc++;
 651		} while (++i < nmembs &&
 652			 vdesc->vd_ino == ino && vdesc->vd_cno == cno);
 653
 654		iput(inode); /* The inode still remains in GC inode list */
 655	}
 656
 657	list_for_each_entry_safe(bh, n, &buffers, b_assoc_buffers) {
 658		ret = nilfs_gccache_wait_and_mark_dirty(bh);
 659		if (unlikely(ret < 0)) {
 660			WARN_ON(ret == -EEXIST);
 661			goto failed;
 662		}
 663		list_del_init(&bh->b_assoc_buffers);
 664		brelse(bh);
 665	}
 666	return nmembs;
 667
 668 failed:
 669	list_for_each_entry_safe(bh, n, &buffers, b_assoc_buffers) {
 670		list_del_init(&bh->b_assoc_buffers);
 671		brelse(bh);
 672	}
 673	return ret;
 674}
 675
 676/**
 677 * nilfs_ioctl_delete_checkpoints - delete checkpoints
 678 * @nilfs: nilfs object
 679 * @argv: vector of arguments from userspace
 680 * @buf: array of periods of checkpoints numbers
 681 *
 682 * Description: nilfs_ioctl_delete_checkpoints() function deletes checkpoints
 683 * in the period from p_start to p_end, excluding p_end itself. The checkpoints
 684 * which have been already deleted are ignored.
 685 *
 686 * Return Value: Number of processed nilfs_period structures or
 687 * error code, otherwise.
 688 *
 689 * %-EIO - I/O error.
 690 *
 691 * %-ENOMEM - Insufficient amount of memory available.
 692 *
 693 * %-EINVAL - invalid checkpoints.
 694 */
 695static int nilfs_ioctl_delete_checkpoints(struct the_nilfs *nilfs,
 696					  struct nilfs_argv *argv, void *buf)
 697{
 698	size_t nmembs = argv->v_nmembs;
 699	struct inode *cpfile = nilfs->ns_cpfile;
 700	struct nilfs_period *periods = buf;
 701	int ret, i;
 702
 703	for (i = 0; i < nmembs; i++) {
 704		ret = nilfs_cpfile_delete_checkpoints(
 705			cpfile, periods[i].p_start, periods[i].p_end);
 706		if (ret < 0)
 707			return ret;
 708	}
 709	return nmembs;
 710}
 711
 712/**
 713 * nilfs_ioctl_free_vblocknrs - free virtual block numbers
 714 * @nilfs: nilfs object
 715 * @argv: vector of arguments from userspace
 716 * @buf: array of virtual block numbers
 717 *
 718 * Description: nilfs_ioctl_free_vblocknrs() function frees
 719 * the virtual block numbers specified by @buf and @argv->v_nmembs.
 720 *
 721 * Return Value: Number of processed virtual block numbers or
 722 * error code, otherwise.
 723 *
 724 * %-EIO - I/O error.
 725 *
 726 * %-ENOMEM - Insufficient amount of memory available.
 727 *
 728 * %-ENOENT - The virtual block number have not been allocated.
 729 */
 730static int nilfs_ioctl_free_vblocknrs(struct the_nilfs *nilfs,
 731				      struct nilfs_argv *argv, void *buf)
 732{
 733	size_t nmembs = argv->v_nmembs;
 734	int ret;
 735
 736	ret = nilfs_dat_freev(nilfs->ns_dat, buf, nmembs);
 737
 738	return (ret < 0) ? ret : nmembs;
 739}
 740
 741/**
 742 * nilfs_ioctl_mark_blocks_dirty - mark blocks dirty
 743 * @nilfs: nilfs object
 744 * @argv: vector of arguments from userspace
 745 * @buf: array of block descriptors
 746 *
 747 * Description: nilfs_ioctl_mark_blocks_dirty() function marks
 748 * metadata file or data blocks as dirty.
 749 *
 750 * Return Value: Number of processed block descriptors or
 751 * error code, otherwise.
 752 *
 753 * %-ENOMEM - Insufficient memory available.
 754 *
 755 * %-EIO - I/O error
 756 *
 757 * %-ENOENT - the specified block does not exist (hole block)
 758 */
 759static int nilfs_ioctl_mark_blocks_dirty(struct the_nilfs *nilfs,
 760					 struct nilfs_argv *argv, void *buf)
 761{
 762	size_t nmembs = argv->v_nmembs;
 763	struct nilfs_bmap *bmap = NILFS_I(nilfs->ns_dat)->i_bmap;
 764	struct nilfs_bdesc *bdescs = buf;
 765	struct buffer_head *bh;
 766	int ret, i;
 767
 768	for (i = 0; i < nmembs; i++) {
 769		/* XXX: use macro or inline func to check liveness */
 770		ret = nilfs_bmap_lookup_at_level(bmap,
 771						 bdescs[i].bd_offset,
 772						 bdescs[i].bd_level + 1,
 773						 &bdescs[i].bd_blocknr);
 774		if (ret < 0) {
 775			if (ret != -ENOENT)
 776				return ret;
 777			bdescs[i].bd_blocknr = 0;
 778		}
 779		if (bdescs[i].bd_blocknr != bdescs[i].bd_oblocknr)
 780			/* skip dead block */
 781			continue;
 782		if (bdescs[i].bd_level == 0) {
 783			ret = nilfs_mdt_get_block(nilfs->ns_dat,
 784						  bdescs[i].bd_offset,
 785						  false, NULL, &bh);
 786			if (unlikely(ret)) {
 787				WARN_ON(ret == -ENOENT);
 788				return ret;
 789			}
 790			mark_buffer_dirty(bh);
 791			nilfs_mdt_mark_dirty(nilfs->ns_dat);
 792			put_bh(bh);
 793		} else {
 794			ret = nilfs_bmap_mark(bmap, bdescs[i].bd_offset,
 795					      bdescs[i].bd_level);
 796			if (ret < 0) {
 797				WARN_ON(ret == -ENOENT);
 798				return ret;
 799			}
 800		}
 801	}
 802	return nmembs;
 803}
 804
 805int nilfs_ioctl_prepare_clean_segments(struct the_nilfs *nilfs,
 806				       struct nilfs_argv *argv, void **kbufs)
 807{
 808	const char *msg;
 809	int ret;
 810
 811	ret = nilfs_ioctl_delete_checkpoints(nilfs, &argv[1], kbufs[1]);
 812	if (ret < 0) {
 813		/*
 814		 * can safely abort because checkpoints can be removed
 815		 * independently.
 816		 */
 817		msg = "cannot delete checkpoints";
 818		goto failed;
 819	}
 820	ret = nilfs_ioctl_free_vblocknrs(nilfs, &argv[2], kbufs[2]);
 821	if (ret < 0) {
 822		/*
 823		 * can safely abort because DAT file is updated atomically
 824		 * using a copy-on-write technique.
 825		 */
 826		msg = "cannot delete virtual blocks from DAT file";
 827		goto failed;
 828	}
 829	ret = nilfs_ioctl_mark_blocks_dirty(nilfs, &argv[3], kbufs[3]);
 830	if (ret < 0) {
 831		/*
 832		 * can safely abort because the operation is nondestructive.
 833		 */
 834		msg = "cannot mark copying blocks dirty";
 835		goto failed;
 836	}
 837	return 0;
 838
 839 failed:
 840	nilfs_err(nilfs->ns_sb, "error %d preparing GC: %s", ret, msg);
 
 841	return ret;
 842}
 843
 844/**
 845 * nilfs_ioctl_clean_segments - clean segments
 846 * @inode: inode object
 847 * @filp: file object
 848 * @cmd: ioctl's request code
 849 * @argp: pointer on argument from userspace
 850 *
 851 * Description: nilfs_ioctl_clean_segments() function makes garbage
 852 * collection operation in the environment of requested parameters
 853 * from userspace. The NILFS_IOCTL_CLEAN_SEGMENTS ioctl is used by
 854 * nilfs_cleanerd daemon.
 855 *
 856 * Return Value: On success, 0 is returned or error code, otherwise.
 857 */
 858static int nilfs_ioctl_clean_segments(struct inode *inode, struct file *filp,
 859				      unsigned int cmd, void __user *argp)
 860{
 861	struct nilfs_argv argv[5];
 862	static const size_t argsz[5] = {
 863		sizeof(struct nilfs_vdesc),
 864		sizeof(struct nilfs_period),
 865		sizeof(__u64),
 866		sizeof(struct nilfs_bdesc),
 867		sizeof(__u64),
 868	};
 869	void __user *base;
 870	void *kbufs[5];
 871	struct the_nilfs *nilfs;
 872	size_t len, nsegs;
 873	int n, ret;
 874
 875	if (!capable(CAP_SYS_ADMIN))
 876		return -EPERM;
 877
 878	ret = mnt_want_write_file(filp);
 879	if (ret)
 880		return ret;
 881
 882	ret = -EFAULT;
 883	if (copy_from_user(argv, argp, sizeof(argv)))
 884		goto out;
 885
 886	ret = -EINVAL;
 887	nsegs = argv[4].v_nmembs;
 888	if (argv[4].v_size != argsz[4])
 889		goto out;
 890	if (nsegs > UINT_MAX / sizeof(__u64))
 891		goto out;
 892
 893	/*
 894	 * argv[4] points to segment numbers this ioctl cleans.  We
 895	 * use kmalloc() for its buffer because memory used for the
 896	 * segment numbers is enough small.
 897	 */
 898	kbufs[4] = memdup_user((void __user *)(unsigned long)argv[4].v_base,
 899			       nsegs * sizeof(__u64));
 900	if (IS_ERR(kbufs[4])) {
 901		ret = PTR_ERR(kbufs[4]);
 902		goto out;
 903	}
 904	nilfs = inode->i_sb->s_fs_info;
 905
 906	for (n = 0; n < 4; n++) {
 907		ret = -EINVAL;
 908		if (argv[n].v_size != argsz[n])
 909			goto out_free;
 910
 911		if (argv[n].v_nmembs > nsegs * nilfs->ns_blocks_per_segment)
 912			goto out_free;
 913
 914		if (argv[n].v_nmembs >= UINT_MAX / argv[n].v_size)
 915			goto out_free;
 916
 917		len = argv[n].v_size * argv[n].v_nmembs;
 918		base = (void __user *)(unsigned long)argv[n].v_base;
 919		if (len == 0) {
 920			kbufs[n] = NULL;
 921			continue;
 922		}
 923
 924		kbufs[n] = vmalloc(len);
 925		if (!kbufs[n]) {
 926			ret = -ENOMEM;
 927			goto out_free;
 928		}
 929		if (copy_from_user(kbufs[n], base, len)) {
 930			ret = -EFAULT;
 931			vfree(kbufs[n]);
 932			goto out_free;
 933		}
 934	}
 935
 936	/*
 937	 * nilfs_ioctl_move_blocks() will call nilfs_iget_for_gc(),
 938	 * which will operates an inode list without blocking.
 939	 * To protect the list from concurrent operations,
 940	 * nilfs_ioctl_move_blocks should be atomic operation.
 941	 */
 942	if (test_and_set_bit(THE_NILFS_GC_RUNNING, &nilfs->ns_flags)) {
 943		ret = -EBUSY;
 944		goto out_free;
 945	}
 946
 
 
 947	ret = nilfs_ioctl_move_blocks(inode->i_sb, &argv[0], kbufs[0]);
 948	if (ret < 0) {
 949		nilfs_err(inode->i_sb,
 950			  "error %d preparing GC: cannot read source blocks",
 951			  ret);
 952	} else {
 953		if (nilfs_sb_need_update(nilfs))
 954			set_nilfs_discontinued(nilfs);
 955		ret = nilfs_clean_segments(inode->i_sb, argv, kbufs);
 956	}
 957
 958	nilfs_remove_all_gcinodes(nilfs);
 959	clear_nilfs_gc_running(nilfs);
 960
 961out_free:
 962	while (--n >= 0)
 963		vfree(kbufs[n]);
 964	kfree(kbufs[4]);
 965out:
 966	mnt_drop_write_file(filp);
 967	return ret;
 968}
 969
 970/**
 971 * nilfs_ioctl_sync - make a checkpoint
 972 * @inode: inode object
 973 * @filp: file object
 974 * @cmd: ioctl's request code
 975 * @argp: pointer on argument from userspace
 976 *
 977 * Description: nilfs_ioctl_sync() function constructs a logical segment
 978 * for checkpointing.  This function guarantees that all modified data
 979 * and metadata are written out to the device when it successfully
 980 * returned.
 981 *
 982 * Return Value: On success, 0 is retured. On errors, one of the following
 983 * negative error code is returned.
 984 *
 985 * %-EROFS - Read only filesystem.
 986 *
 987 * %-EIO - I/O error
 988 *
 989 * %-ENOSPC - No space left on device (only in a panic state).
 990 *
 991 * %-ERESTARTSYS - Interrupted.
 992 *
 993 * %-ENOMEM - Insufficient memory available.
 994 *
 995 * %-EFAULT - Failure during execution of requested operation.
 996 */
 997static int nilfs_ioctl_sync(struct inode *inode, struct file *filp,
 998			    unsigned int cmd, void __user *argp)
 999{
1000	__u64 cno;
1001	int ret;
1002	struct the_nilfs *nilfs;
1003
1004	ret = nilfs_construct_segment(inode->i_sb);
1005	if (ret < 0)
1006		return ret;
1007
1008	nilfs = inode->i_sb->s_fs_info;
1009	ret = nilfs_flush_device(nilfs);
1010	if (ret < 0)
1011		return ret;
 
 
1012
1013	if (argp != NULL) {
1014		down_read(&nilfs->ns_segctor_sem);
1015		cno = nilfs->ns_cno - 1;
1016		up_read(&nilfs->ns_segctor_sem);
1017		if (copy_to_user(argp, &cno, sizeof(cno)))
1018			return -EFAULT;
1019	}
1020	return 0;
1021}
1022
1023/**
1024 * nilfs_ioctl_resize - resize NILFS2 volume
1025 * @inode: inode object
1026 * @filp: file object
1027 * @argp: pointer on argument from userspace
1028 *
1029 * Return Value: On success, 0 is returned or error code, otherwise.
1030 */
1031static int nilfs_ioctl_resize(struct inode *inode, struct file *filp,
1032			      void __user *argp)
1033{
1034	__u64 newsize;
1035	int ret = -EPERM;
1036
1037	if (!capable(CAP_SYS_ADMIN))
1038		goto out;
1039
1040	ret = mnt_want_write_file(filp);
1041	if (ret)
1042		goto out;
1043
1044	ret = -EFAULT;
1045	if (copy_from_user(&newsize, argp, sizeof(newsize)))
1046		goto out_drop_write;
1047
1048	ret = nilfs_resize_fs(inode->i_sb, newsize);
1049
1050out_drop_write:
1051	mnt_drop_write_file(filp);
1052out:
1053	return ret;
1054}
1055
1056/**
1057 * nilfs_ioctl_trim_fs() - trim ioctl handle function
1058 * @inode: inode object
1059 * @argp: pointer on argument from userspace
1060 *
1061 * Decription: nilfs_ioctl_trim_fs is the FITRIM ioctl handle function. It
1062 * checks the arguments from userspace and calls nilfs_sufile_trim_fs, which
1063 * performs the actual trim operation.
1064 *
1065 * Return Value: On success, 0 is returned or negative error code, otherwise.
1066 */
1067static int nilfs_ioctl_trim_fs(struct inode *inode, void __user *argp)
1068{
1069	struct the_nilfs *nilfs = inode->i_sb->s_fs_info;
1070	struct request_queue *q = bdev_get_queue(nilfs->ns_bdev);
1071	struct fstrim_range range;
1072	int ret;
1073
1074	if (!capable(CAP_SYS_ADMIN))
1075		return -EPERM;
1076
1077	if (!blk_queue_discard(q))
1078		return -EOPNOTSUPP;
1079
1080	if (copy_from_user(&range, argp, sizeof(range)))
1081		return -EFAULT;
1082
1083	range.minlen = max_t(u64, range.minlen, q->limits.discard_granularity);
1084
1085	down_read(&nilfs->ns_segctor_sem);
1086	ret = nilfs_sufile_trim_fs(nilfs->ns_sufile, &range);
1087	up_read(&nilfs->ns_segctor_sem);
1088
1089	if (ret < 0)
1090		return ret;
1091
1092	if (copy_to_user(argp, &range, sizeof(range)))
1093		return -EFAULT;
1094
1095	return 0;
1096}
1097
1098/**
1099 * nilfs_ioctl_set_alloc_range - limit range of segments to be allocated
1100 * @inode: inode object
1101 * @argp: pointer on argument from userspace
1102 *
1103 * Decription: nilfs_ioctl_set_alloc_range() function defines lower limit
1104 * of segments in bytes and upper limit of segments in bytes.
1105 * The NILFS_IOCTL_SET_ALLOC_RANGE is used by nilfs_resize utility.
1106 *
1107 * Return Value: On success, 0 is returned or error code, otherwise.
1108 */
1109static int nilfs_ioctl_set_alloc_range(struct inode *inode, void __user *argp)
1110{
1111	struct the_nilfs *nilfs = inode->i_sb->s_fs_info;
1112	__u64 range[2];
1113	__u64 minseg, maxseg;
1114	unsigned long segbytes;
1115	int ret = -EPERM;
1116
1117	if (!capable(CAP_SYS_ADMIN))
1118		goto out;
1119
1120	ret = -EFAULT;
1121	if (copy_from_user(range, argp, sizeof(__u64[2])))
1122		goto out;
1123
1124	ret = -ERANGE;
1125	if (range[1] > i_size_read(inode->i_sb->s_bdev->bd_inode))
1126		goto out;
1127
1128	segbytes = nilfs->ns_blocks_per_segment * nilfs->ns_blocksize;
1129
1130	minseg = range[0] + segbytes - 1;
1131	do_div(minseg, segbytes);
1132	maxseg = NILFS_SB2_OFFSET_BYTES(range[1]);
1133	do_div(maxseg, segbytes);
1134	maxseg--;
1135
1136	ret = nilfs_sufile_set_alloc_range(nilfs->ns_sufile, minseg, maxseg);
1137out:
1138	return ret;
1139}
1140
1141/**
1142 * nilfs_ioctl_get_info - wrapping function of get metadata info
1143 * @inode: inode object
1144 * @filp: file object
1145 * @cmd: ioctl's request code
1146 * @argp: pointer on argument from userspace
1147 * @membsz: size of an item in bytes
1148 * @dofunc: concrete function of getting metadata info
1149 *
1150 * Description: nilfs_ioctl_get_info() gets metadata info by means of
1151 * calling dofunc() function.
1152 *
1153 * Return Value: On success, 0 is returned and requested metadata info
1154 * is copied into userspace. On error, one of the following
1155 * negative error codes is returned.
1156 *
1157 * %-EINVAL - Invalid arguments from userspace.
1158 *
1159 * %-ENOMEM - Insufficient amount of memory available.
1160 *
1161 * %-EFAULT - Failure during execution of requested operation.
1162 */
1163static int nilfs_ioctl_get_info(struct inode *inode, struct file *filp,
1164				unsigned int cmd, void __user *argp,
1165				size_t membsz,
1166				ssize_t (*dofunc)(struct the_nilfs *,
1167						  __u64 *, int,
1168						  void *, size_t, size_t))
1169
1170{
1171	struct the_nilfs *nilfs = inode->i_sb->s_fs_info;
1172	struct nilfs_argv argv;
1173	int ret;
1174
1175	if (copy_from_user(&argv, argp, sizeof(argv)))
1176		return -EFAULT;
1177
1178	if (argv.v_size < membsz)
1179		return -EINVAL;
1180
1181	ret = nilfs_ioctl_wrap_copy(nilfs, &argv, _IOC_DIR(cmd), dofunc);
1182	if (ret < 0)
1183		return ret;
1184
1185	if (copy_to_user(argp, &argv, sizeof(argv)))
1186		ret = -EFAULT;
1187	return ret;
1188}
1189
1190/**
1191 * nilfs_ioctl_set_suinfo - set segment usage info
1192 * @inode: inode object
1193 * @filp: file object
1194 * @cmd: ioctl's request code
1195 * @argp: pointer on argument from userspace
1196 *
1197 * Description: Expects an array of nilfs_suinfo_update structures
1198 * encapsulated in nilfs_argv and updates the segment usage info
1199 * according to the flags in nilfs_suinfo_update.
1200 *
1201 * Return Value: On success, 0 is returned. On error, one of the
1202 * following negative error codes is returned.
1203 *
1204 * %-EPERM - Not enough permissions
1205 *
1206 * %-EFAULT - Error copying input data
1207 *
1208 * %-EIO - I/O error.
1209 *
1210 * %-ENOMEM - Insufficient amount of memory available.
1211 *
1212 * %-EINVAL - Invalid values in input (segment number, flags or nblocks)
1213 */
1214static int nilfs_ioctl_set_suinfo(struct inode *inode, struct file *filp,
1215				unsigned int cmd, void __user *argp)
1216{
1217	struct the_nilfs *nilfs = inode->i_sb->s_fs_info;
1218	struct nilfs_transaction_info ti;
1219	struct nilfs_argv argv;
1220	size_t len;
1221	void __user *base;
1222	void *kbuf;
1223	int ret;
1224
1225	if (!capable(CAP_SYS_ADMIN))
1226		return -EPERM;
1227
1228	ret = mnt_want_write_file(filp);
1229	if (ret)
1230		return ret;
1231
1232	ret = -EFAULT;
1233	if (copy_from_user(&argv, argp, sizeof(argv)))
1234		goto out;
1235
1236	ret = -EINVAL;
1237	if (argv.v_size < sizeof(struct nilfs_suinfo_update))
1238		goto out;
1239
1240	if (argv.v_nmembs > nilfs->ns_nsegments)
1241		goto out;
1242
1243	if (argv.v_nmembs >= UINT_MAX / argv.v_size)
1244		goto out;
1245
1246	len = argv.v_size * argv.v_nmembs;
1247	if (!len) {
1248		ret = 0;
1249		goto out;
1250	}
1251
1252	base = (void __user *)(unsigned long)argv.v_base;
1253	kbuf = vmalloc(len);
1254	if (!kbuf) {
1255		ret = -ENOMEM;
1256		goto out;
1257	}
1258
1259	if (copy_from_user(kbuf, base, len)) {
1260		ret = -EFAULT;
1261		goto out_free;
1262	}
1263
1264	nilfs_transaction_begin(inode->i_sb, &ti, 0);
1265	ret = nilfs_sufile_set_suinfo(nilfs->ns_sufile, kbuf, argv.v_size,
1266			argv.v_nmembs);
1267	if (unlikely(ret < 0))
1268		nilfs_transaction_abort(inode->i_sb);
1269	else
1270		nilfs_transaction_commit(inode->i_sb); /* never fails */
1271
1272out_free:
1273	vfree(kbuf);
1274out:
1275	mnt_drop_write_file(filp);
1276	return ret;
1277}
1278
1279long nilfs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
1280{
1281	struct inode *inode = file_inode(filp);
1282	void __user *argp = (void __user *)arg;
1283
1284	switch (cmd) {
1285	case FS_IOC_GETFLAGS:
1286		return nilfs_ioctl_getflags(inode, argp);
1287	case FS_IOC_SETFLAGS:
1288		return nilfs_ioctl_setflags(inode, filp, argp);
1289	case FS_IOC_GETVERSION:
1290		return nilfs_ioctl_getversion(inode, argp);
1291	case NILFS_IOCTL_CHANGE_CPMODE:
1292		return nilfs_ioctl_change_cpmode(inode, filp, cmd, argp);
1293	case NILFS_IOCTL_DELETE_CHECKPOINT:
1294		return nilfs_ioctl_delete_checkpoint(inode, filp, cmd, argp);
1295	case NILFS_IOCTL_GET_CPINFO:
1296		return nilfs_ioctl_get_info(inode, filp, cmd, argp,
1297					    sizeof(struct nilfs_cpinfo),
1298					    nilfs_ioctl_do_get_cpinfo);
1299	case NILFS_IOCTL_GET_CPSTAT:
1300		return nilfs_ioctl_get_cpstat(inode, filp, cmd, argp);
1301	case NILFS_IOCTL_GET_SUINFO:
1302		return nilfs_ioctl_get_info(inode, filp, cmd, argp,
1303					    sizeof(struct nilfs_suinfo),
1304					    nilfs_ioctl_do_get_suinfo);
1305	case NILFS_IOCTL_SET_SUINFO:
1306		return nilfs_ioctl_set_suinfo(inode, filp, cmd, argp);
1307	case NILFS_IOCTL_GET_SUSTAT:
1308		return nilfs_ioctl_get_sustat(inode, filp, cmd, argp);
1309	case NILFS_IOCTL_GET_VINFO:
1310		return nilfs_ioctl_get_info(inode, filp, cmd, argp,
1311					    sizeof(struct nilfs_vinfo),
1312					    nilfs_ioctl_do_get_vinfo);
1313	case NILFS_IOCTL_GET_BDESCS:
1314		return nilfs_ioctl_get_bdescs(inode, filp, cmd, argp);
1315	case NILFS_IOCTL_CLEAN_SEGMENTS:
1316		return nilfs_ioctl_clean_segments(inode, filp, cmd, argp);
1317	case NILFS_IOCTL_SYNC:
1318		return nilfs_ioctl_sync(inode, filp, cmd, argp);
1319	case NILFS_IOCTL_RESIZE:
1320		return nilfs_ioctl_resize(inode, filp, argp);
1321	case NILFS_IOCTL_SET_ALLOC_RANGE:
1322		return nilfs_ioctl_set_alloc_range(inode, argp);
1323	case FITRIM:
1324		return nilfs_ioctl_trim_fs(inode, argp);
1325	default:
1326		return -ENOTTY;
1327	}
1328}
1329
1330#ifdef CONFIG_COMPAT
1331long nilfs_compat_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
1332{
1333	switch (cmd) {
1334	case FS_IOC32_GETFLAGS:
1335		cmd = FS_IOC_GETFLAGS;
1336		break;
1337	case FS_IOC32_SETFLAGS:
1338		cmd = FS_IOC_SETFLAGS;
1339		break;
1340	case FS_IOC32_GETVERSION:
1341		cmd = FS_IOC_GETVERSION;
1342		break;
1343	case NILFS_IOCTL_CHANGE_CPMODE:
1344	case NILFS_IOCTL_DELETE_CHECKPOINT:
1345	case NILFS_IOCTL_GET_CPINFO:
1346	case NILFS_IOCTL_GET_CPSTAT:
1347	case NILFS_IOCTL_GET_SUINFO:
1348	case NILFS_IOCTL_SET_SUINFO:
1349	case NILFS_IOCTL_GET_SUSTAT:
1350	case NILFS_IOCTL_GET_VINFO:
1351	case NILFS_IOCTL_GET_BDESCS:
1352	case NILFS_IOCTL_CLEAN_SEGMENTS:
1353	case NILFS_IOCTL_SYNC:
1354	case NILFS_IOCTL_RESIZE:
1355	case NILFS_IOCTL_SET_ALLOC_RANGE:
1356	case FITRIM:
1357		break;
1358	default:
1359		return -ENOIOCTLCMD;
1360	}
1361	return nilfs_ioctl(filp, cmd, (unsigned long)compat_ptr(arg));
1362}
1363#endif