xref: /linux/fs/ocfs2/ioctl.c (revision 5ea5880764cbb164afb17a62e76ca75dc371409d)
1 // SPDX-License-Identifier: GPL-2.0
2 /*
3  * linux/fs/ocfs2/ioctl.c
4  *
5  * Copyright (C) 2006 Herbert Poetzl
6  * adapted from Remy Card's ext2/ioctl.c
7  */
8 
9 #include <linux/fs.h>
10 #include <linux/mount.h>
11 #include <linux/blkdev.h>
12 #include <linux/compat.h>
13 #include <linux/fileattr.h>
14 
15 #include <cluster/masklog.h>
16 
17 #include "ocfs2.h"
18 #include "alloc.h"
19 #include "dlmglue.h"
20 #include "file.h"
21 #include "inode.h"
22 #include "journal.h"
23 
24 #include "ocfs2_fs.h"
25 #include "ioctl.h"
26 #include "resize.h"
27 #include "refcounttree.h"
28 #include "sysfile.h"
29 #include "dir.h"
30 #include "buffer_head_io.h"
31 #include "suballoc.h"
32 #include "move_extents.h"
33 
34 #define o2info_from_user(a, b)	\
35 		copy_from_user(&(a), (b), sizeof(a))
36 #define o2info_to_user(a, b)	\
37 		copy_to_user((typeof(a) __user *)b, &(a), sizeof(a))
38 
39 /*
40  * This is just a best-effort to tell userspace that this request
41  * caused the error.
42  */
43 static inline void o2info_set_request_error(struct ocfs2_info_request *kreq,
44 					struct ocfs2_info_request __user *req)
45 {
46 	kreq->ir_flags |= OCFS2_INFO_FL_ERROR;
47 	(void)put_user(kreq->ir_flags, (__u32 __user *)&(req->ir_flags));
48 }
49 
50 static inline void o2info_set_request_filled(struct ocfs2_info_request *req)
51 {
52 	req->ir_flags |= OCFS2_INFO_FL_FILLED;
53 }
54 
55 static inline void o2info_clear_request_filled(struct ocfs2_info_request *req)
56 {
57 	req->ir_flags &= ~OCFS2_INFO_FL_FILLED;
58 }
59 
60 static inline int o2info_coherent(struct ocfs2_info_request *req)
61 {
62 	return (!(req->ir_flags & OCFS2_INFO_FL_NON_COHERENT));
63 }
64 
65 int ocfs2_fileattr_get(struct dentry *dentry, struct file_kattr *fa)
66 {
67 	struct inode *inode = d_inode(dentry);
68 	unsigned int flags;
69 	int status;
70 
71 	status = ocfs2_inode_lock(inode, NULL, 0);
72 	if (status < 0) {
73 		mlog_errno(status);
74 		return status;
75 	}
76 	ocfs2_get_inode_flags(OCFS2_I(inode));
77 	flags = OCFS2_I(inode)->ip_attr;
78 	ocfs2_inode_unlock(inode, 0);
79 
80 	fileattr_fill_flags(fa, flags & OCFS2_FL_VISIBLE);
81 
82 	return status;
83 }
84 
85 int ocfs2_fileattr_set(struct mnt_idmap *idmap,
86 		       struct dentry *dentry, struct file_kattr *fa)
87 {
88 	struct inode *inode = d_inode(dentry);
89 	unsigned int flags = fa->flags;
90 	struct ocfs2_inode_info *ocfs2_inode = OCFS2_I(inode);
91 	struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
92 	handle_t *handle = NULL;
93 	struct buffer_head *bh = NULL;
94 	unsigned oldflags;
95 	int status;
96 
97 	if (fileattr_has_fsx(fa))
98 		return -EOPNOTSUPP;
99 
100 	status = ocfs2_inode_lock(inode, &bh, 1);
101 	if (status < 0) {
102 		mlog_errno(status);
103 		goto bail;
104 	}
105 
106 	if (!S_ISDIR(inode->i_mode))
107 		flags &= ~OCFS2_DIRSYNC_FL;
108 
109 	oldflags = ocfs2_inode->ip_attr;
110 	flags = flags & OCFS2_FL_MODIFIABLE;
111 	flags |= oldflags & ~OCFS2_FL_MODIFIABLE;
112 
113 	/* Check already done by VFS, but repeat with ocfs lock */
114 	status = -EPERM;
115 	if ((flags ^ oldflags) & (FS_APPEND_FL | FS_IMMUTABLE_FL) &&
116 	    !capable(CAP_LINUX_IMMUTABLE))
117 		goto bail_unlock;
118 
119 	handle = ocfs2_start_trans(osb, OCFS2_INODE_UPDATE_CREDITS);
120 	if (IS_ERR(handle)) {
121 		status = PTR_ERR(handle);
122 		mlog_errno(status);
123 		goto bail_unlock;
124 	}
125 
126 	ocfs2_inode->ip_attr = flags;
127 	ocfs2_set_inode_flags(inode);
128 	inode_set_ctime_current(inode);
129 
130 	status = ocfs2_mark_inode_dirty(handle, inode, bh);
131 	if (status < 0)
132 		mlog_errno(status);
133 
134 	ocfs2_commit_trans(osb, handle);
135 
136 bail_unlock:
137 	ocfs2_inode_unlock(inode, 1);
138 bail:
139 	brelse(bh);
140 
141 	return status;
142 }
143 
144 static int ocfs2_info_handle_blocksize(struct inode *inode,
145 				       struct ocfs2_info_request __user *req)
146 {
147 	struct ocfs2_info_blocksize oib;
148 
149 	if (o2info_from_user(oib, req))
150 		return -EFAULT;
151 
152 	oib.ib_blocksize = inode->i_sb->s_blocksize;
153 
154 	o2info_set_request_filled(&oib.ib_req);
155 
156 	if (o2info_to_user(oib, req))
157 		return -EFAULT;
158 
159 	return 0;
160 }
161 
162 static int ocfs2_info_handle_clustersize(struct inode *inode,
163 					 struct ocfs2_info_request __user *req)
164 {
165 	struct ocfs2_info_clustersize oic;
166 	struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
167 
168 	if (o2info_from_user(oic, req))
169 		return -EFAULT;
170 
171 	oic.ic_clustersize = osb->s_clustersize;
172 
173 	o2info_set_request_filled(&oic.ic_req);
174 
175 	if (o2info_to_user(oic, req))
176 		return -EFAULT;
177 
178 	return 0;
179 }
180 
181 static int ocfs2_info_handle_maxslots(struct inode *inode,
182 				      struct ocfs2_info_request __user *req)
183 {
184 	struct ocfs2_info_maxslots oim;
185 	struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
186 
187 	if (o2info_from_user(oim, req))
188 		return -EFAULT;
189 
190 	oim.im_max_slots = osb->max_slots;
191 
192 	o2info_set_request_filled(&oim.im_req);
193 
194 	if (o2info_to_user(oim, req))
195 		return -EFAULT;
196 
197 	return 0;
198 }
199 
200 static int ocfs2_info_handle_label(struct inode *inode,
201 				   struct ocfs2_info_request __user *req)
202 {
203 	struct ocfs2_info_label oil;
204 	struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
205 
206 	if (o2info_from_user(oil, req))
207 		return -EFAULT;
208 
209 	memcpy(oil.il_label, osb->vol_label, OCFS2_MAX_VOL_LABEL_LEN);
210 
211 	o2info_set_request_filled(&oil.il_req);
212 
213 	if (o2info_to_user(oil, req))
214 		return -EFAULT;
215 
216 	return 0;
217 }
218 
219 static int ocfs2_info_handle_uuid(struct inode *inode,
220 				  struct ocfs2_info_request __user *req)
221 {
222 	struct ocfs2_info_uuid oiu;
223 	struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
224 
225 	if (o2info_from_user(oiu, req))
226 		return -EFAULT;
227 
228 	memcpy(oiu.iu_uuid_str, osb->uuid_str, OCFS2_TEXT_UUID_LEN + 1);
229 
230 	o2info_set_request_filled(&oiu.iu_req);
231 
232 	if (o2info_to_user(oiu, req))
233 		return -EFAULT;
234 
235 	return 0;
236 }
237 
238 static int ocfs2_info_handle_fs_features(struct inode *inode,
239 					 struct ocfs2_info_request __user *req)
240 {
241 	struct ocfs2_info_fs_features oif;
242 	struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
243 
244 	if (o2info_from_user(oif, req))
245 		return -EFAULT;
246 
247 	oif.if_compat_features = osb->s_feature_compat;
248 	oif.if_incompat_features = osb->s_feature_incompat;
249 	oif.if_ro_compat_features = osb->s_feature_ro_compat;
250 
251 	o2info_set_request_filled(&oif.if_req);
252 
253 	if (o2info_to_user(oif, req))
254 		return -EFAULT;
255 
256 	return 0;
257 }
258 
259 static int ocfs2_info_handle_journal_size(struct inode *inode,
260 					  struct ocfs2_info_request __user *req)
261 {
262 	struct ocfs2_info_journal_size oij;
263 	struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
264 
265 	if (o2info_from_user(oij, req))
266 		return -EFAULT;
267 
268 	oij.ij_journal_size = i_size_read(osb->journal->j_inode);
269 
270 	o2info_set_request_filled(&oij.ij_req);
271 
272 	if (o2info_to_user(oij, req))
273 		return -EFAULT;
274 
275 	return 0;
276 }
277 
278 static int ocfs2_info_scan_inode_alloc(struct ocfs2_super *osb,
279 				       struct inode *inode_alloc, u64 blkno,
280 				       struct ocfs2_info_freeinode *fi,
281 				       u32 slot)
282 {
283 	int status = 0, unlock = 0;
284 
285 	struct buffer_head *bh = NULL;
286 	struct ocfs2_dinode *dinode_alloc = NULL;
287 
288 	if (inode_alloc)
289 		inode_lock(inode_alloc);
290 
291 	if (inode_alloc && o2info_coherent(&fi->ifi_req)) {
292 		status = ocfs2_inode_lock(inode_alloc, &bh, 0);
293 		if (status < 0) {
294 			mlog_errno(status);
295 			goto bail;
296 		}
297 		unlock = 1;
298 	} else {
299 		status = ocfs2_read_blocks_sync(osb, blkno, 1, &bh);
300 		if (status < 0) {
301 			mlog_errno(status);
302 			goto bail;
303 		}
304 	}
305 
306 	dinode_alloc = (struct ocfs2_dinode *)bh->b_data;
307 
308 	fi->ifi_stat[slot].lfi_total =
309 		le32_to_cpu(dinode_alloc->id1.bitmap1.i_total);
310 	fi->ifi_stat[slot].lfi_free =
311 		le32_to_cpu(dinode_alloc->id1.bitmap1.i_total) -
312 		le32_to_cpu(dinode_alloc->id1.bitmap1.i_used);
313 
314 bail:
315 	if (unlock)
316 		ocfs2_inode_unlock(inode_alloc, 0);
317 
318 	if (inode_alloc)
319 		inode_unlock(inode_alloc);
320 
321 	brelse(bh);
322 
323 	return status;
324 }
325 
326 static int ocfs2_info_handle_freeinode(struct inode *inode,
327 				       struct ocfs2_info_request __user *req)
328 {
329 	u32 i;
330 	u64 blkno = -1;
331 	char namebuf[40];
332 	int status, type = INODE_ALLOC_SYSTEM_INODE;
333 	struct ocfs2_info_freeinode *oifi = NULL;
334 	struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
335 	struct inode *inode_alloc = NULL;
336 
337 	oifi = kzalloc_obj(struct ocfs2_info_freeinode);
338 	if (!oifi) {
339 		status = -ENOMEM;
340 		mlog_errno(status);
341 		goto out_err;
342 	}
343 
344 	if (o2info_from_user(*oifi, req)) {
345 		status = -EFAULT;
346 		goto out_free;
347 	}
348 
349 	oifi->ifi_slotnum = osb->max_slots;
350 
351 	for (i = 0; i < oifi->ifi_slotnum; i++) {
352 		if (o2info_coherent(&oifi->ifi_req)) {
353 			inode_alloc = ocfs2_get_system_file_inode(osb, type, i);
354 			if (!inode_alloc) {
355 				mlog(ML_ERROR, "unable to get alloc inode in "
356 				    "slot %u\n", i);
357 				status = -EIO;
358 				goto bail;
359 			}
360 		} else {
361 			int len = ocfs2_sprintf_system_inode_name(namebuf,
362 								  sizeof(namebuf),
363 								  type, i);
364 			status = ocfs2_lookup_ino_from_name(osb->sys_root_inode,
365 							    namebuf, len, &blkno);
366 			if (status < 0) {
367 				status = -ENOENT;
368 				goto bail;
369 			}
370 		}
371 
372 		status = ocfs2_info_scan_inode_alloc(osb, inode_alloc, blkno, oifi, i);
373 
374 		iput(inode_alloc);
375 		inode_alloc = NULL;
376 
377 		if (status < 0)
378 			goto bail;
379 	}
380 
381 	o2info_set_request_filled(&oifi->ifi_req);
382 
383 	if (o2info_to_user(*oifi, req)) {
384 		status = -EFAULT;
385 		goto out_free;
386 	}
387 
388 	status = 0;
389 bail:
390 	if (status)
391 		o2info_set_request_error(&oifi->ifi_req, req);
392 out_free:
393 	kfree(oifi);
394 out_err:
395 	return status;
396 }
397 
398 static void o2ffg_update_histogram(struct ocfs2_info_free_chunk_list *hist,
399 				   unsigned int chunksize)
400 {
401 	u32 index;
402 
403 	index = __ilog2_u32(chunksize);
404 	if (index >= OCFS2_INFO_MAX_HIST)
405 		index = OCFS2_INFO_MAX_HIST - 1;
406 
407 	hist->fc_chunks[index]++;
408 	hist->fc_clusters[index] += chunksize;
409 }
410 
411 static void o2ffg_update_stats(struct ocfs2_info_freefrag_stats *stats,
412 			       unsigned int chunksize)
413 {
414 	if (chunksize > stats->ffs_max)
415 		stats->ffs_max = chunksize;
416 
417 	if (chunksize < stats->ffs_min)
418 		stats->ffs_min = chunksize;
419 
420 	stats->ffs_avg += chunksize;
421 	stats->ffs_free_chunks_real++;
422 }
423 
424 static void ocfs2_info_update_ffg(struct ocfs2_info_freefrag *ffg,
425 				  unsigned int chunksize)
426 {
427 	o2ffg_update_histogram(&(ffg->iff_ffs.ffs_fc_hist), chunksize);
428 	o2ffg_update_stats(&(ffg->iff_ffs), chunksize);
429 }
430 
431 static int ocfs2_info_freefrag_scan_chain(struct ocfs2_super *osb,
432 					  struct inode *gb_inode,
433 					  struct ocfs2_dinode *gb_dinode,
434 					  struct ocfs2_chain_rec *rec,
435 					  struct ocfs2_info_freefrag *ffg,
436 					  u32 chunks_in_group)
437 {
438 	int status = 0, used;
439 	u64 blkno;
440 
441 	struct buffer_head *bh = NULL;
442 	struct ocfs2_group_desc *bg = NULL;
443 
444 	unsigned int max_bits, max_bitmap_bits, num_clusters;
445 	unsigned int offset = 0, cluster, chunk;
446 	unsigned int chunk_free, last_chunksize = 0;
447 
448 	if (!le32_to_cpu(rec->c_free))
449 		goto bail;
450 
451 	max_bitmap_bits = 8 * ocfs2_group_bitmap_size(osb->sb, 0,
452 					      osb->s_feature_incompat);
453 
454 	do {
455 		if (!bg)
456 			blkno = le64_to_cpu(rec->c_blkno);
457 		else
458 			blkno = le64_to_cpu(bg->bg_next_group);
459 
460 		if (bh) {
461 			brelse(bh);
462 			bh = NULL;
463 		}
464 
465 		if (o2info_coherent(&ffg->iff_req))
466 			status = ocfs2_read_group_descriptor(gb_inode,
467 							     gb_dinode,
468 							     blkno, &bh);
469 		else
470 			status = ocfs2_read_blocks_sync(osb, blkno, 1, &bh);
471 
472 		if (status < 0) {
473 			mlog(ML_ERROR, "Can't read the group descriptor # "
474 			     "%llu from device.", (unsigned long long)blkno);
475 			status = -EIO;
476 			goto bail;
477 		}
478 
479 		bg = (struct ocfs2_group_desc *)bh->b_data;
480 
481 		if (!le16_to_cpu(bg->bg_free_bits_count))
482 			continue;
483 
484 		max_bits = le16_to_cpu(bg->bg_bits);
485 
486 		/*
487 		 * Non-coherent scans read raw blocks and do not get the
488 		 * bg_bits validation from
489 		 * ocfs2_read_group_descriptor().
490 		 */
491 		if (max_bits > max_bitmap_bits) {
492 			mlog(ML_ERROR,
493 			     "Group desc #%llu has %u bits, max bitmap bits %u\n",
494 			     (unsigned long long)blkno, max_bits, max_bitmap_bits);
495 			max_bits = max_bitmap_bits;
496 		}
497 
498 		offset = 0;
499 
500 		for (chunk = 0; chunk < chunks_in_group; chunk++) {
501 			/*
502 			 * last chunk may be not an entire one.
503 			 */
504 			if ((offset + ffg->iff_chunksize) > max_bits)
505 				num_clusters = max_bits - offset;
506 			else
507 				num_clusters = ffg->iff_chunksize;
508 
509 			chunk_free = 0;
510 			for (cluster = 0; cluster < num_clusters; cluster++) {
511 				used = ocfs2_test_bit(offset,
512 						(unsigned long *)bg->bg_bitmap);
513 				/*
514 				 * - chunk_free counts free clusters in #N chunk.
515 				 * - last_chunksize records the size(in) clusters
516 				 *   for the last real free chunk being counted.
517 				 */
518 				if (!used) {
519 					last_chunksize++;
520 					chunk_free++;
521 				}
522 
523 				if (used && last_chunksize) {
524 					ocfs2_info_update_ffg(ffg,
525 							      last_chunksize);
526 					last_chunksize = 0;
527 				}
528 
529 				offset++;
530 			}
531 
532 			if (chunk_free == ffg->iff_chunksize)
533 				ffg->iff_ffs.ffs_free_chunks++;
534 		}
535 
536 		/*
537 		 * need to update the info for last free chunk.
538 		 */
539 		if (last_chunksize)
540 			ocfs2_info_update_ffg(ffg, last_chunksize);
541 
542 	} while (le64_to_cpu(bg->bg_next_group));
543 
544 bail:
545 	brelse(bh);
546 
547 	return status;
548 }
549 
550 static int ocfs2_info_freefrag_scan_bitmap(struct ocfs2_super *osb,
551 					   struct inode *gb_inode, u64 blkno,
552 					   struct ocfs2_info_freefrag *ffg)
553 {
554 	u32 chunks_in_group;
555 	int status = 0, unlock = 0, i;
556 
557 	struct buffer_head *bh = NULL;
558 	struct ocfs2_chain_list *cl = NULL;
559 	struct ocfs2_chain_rec *rec = NULL;
560 	struct ocfs2_dinode *gb_dinode = NULL;
561 
562 	if (gb_inode)
563 		inode_lock(gb_inode);
564 
565 	if (o2info_coherent(&ffg->iff_req)) {
566 		status = ocfs2_inode_lock(gb_inode, &bh, 0);
567 		if (status < 0) {
568 			mlog_errno(status);
569 			goto bail;
570 		}
571 		unlock = 1;
572 	} else {
573 		status = ocfs2_read_blocks_sync(osb, blkno, 1, &bh);
574 		if (status < 0) {
575 			mlog_errno(status);
576 			goto bail;
577 		}
578 	}
579 
580 	gb_dinode = (struct ocfs2_dinode *)bh->b_data;
581 	cl = &(gb_dinode->id2.i_chain);
582 
583 	/*
584 	 * Chunksize(in) clusters from userspace should be
585 	 * less than clusters in a group.
586 	 */
587 	if (ffg->iff_chunksize > le16_to_cpu(cl->cl_cpg)) {
588 		status = -EINVAL;
589 		goto bail;
590 	}
591 
592 	memset(&ffg->iff_ffs, 0, sizeof(struct ocfs2_info_freefrag_stats));
593 
594 	ffg->iff_ffs.ffs_min = ~0U;
595 	ffg->iff_ffs.ffs_clusters =
596 			le32_to_cpu(gb_dinode->id1.bitmap1.i_total);
597 	ffg->iff_ffs.ffs_free_clusters = ffg->iff_ffs.ffs_clusters -
598 			le32_to_cpu(gb_dinode->id1.bitmap1.i_used);
599 
600 	chunks_in_group = le16_to_cpu(cl->cl_cpg) / ffg->iff_chunksize + 1;
601 
602 	for (i = 0; i < le16_to_cpu(cl->cl_next_free_rec); i++) {
603 		rec = &(cl->cl_recs[i]);
604 		status = ocfs2_info_freefrag_scan_chain(osb, gb_inode,
605 							gb_dinode,
606 							rec, ffg,
607 							chunks_in_group);
608 		if (status)
609 			goto bail;
610 	}
611 
612 	if (ffg->iff_ffs.ffs_free_chunks_real)
613 		ffg->iff_ffs.ffs_avg = (ffg->iff_ffs.ffs_avg /
614 					ffg->iff_ffs.ffs_free_chunks_real);
615 bail:
616 	if (unlock)
617 		ocfs2_inode_unlock(gb_inode, 0);
618 
619 	if (gb_inode)
620 		inode_unlock(gb_inode);
621 
622 	iput(gb_inode);
623 	brelse(bh);
624 
625 	return status;
626 }
627 
628 static int ocfs2_info_handle_freefrag(struct inode *inode,
629 				      struct ocfs2_info_request __user *req)
630 {
631 	u64 blkno = -1;
632 	char namebuf[40];
633 	int status, type = GLOBAL_BITMAP_SYSTEM_INODE;
634 
635 	struct ocfs2_info_freefrag *oiff;
636 	struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
637 	struct inode *gb_inode = NULL;
638 
639 	oiff = kzalloc_obj(struct ocfs2_info_freefrag);
640 	if (!oiff) {
641 		status = -ENOMEM;
642 		mlog_errno(status);
643 		goto out_err;
644 	}
645 
646 	if (o2info_from_user(*oiff, req)) {
647 		status = -EFAULT;
648 		goto out_free;
649 	}
650 	/*
651 	 * chunksize from userspace should be power of 2.
652 	 */
653 	if ((oiff->iff_chunksize & (oiff->iff_chunksize - 1)) ||
654 	    (!oiff->iff_chunksize)) {
655 		status = -EINVAL;
656 		goto bail;
657 	}
658 
659 	if (o2info_coherent(&oiff->iff_req)) {
660 		gb_inode = ocfs2_get_system_file_inode(osb, type,
661 						       OCFS2_INVALID_SLOT);
662 		if (!gb_inode) {
663 			mlog(ML_ERROR, "unable to get global_bitmap inode\n");
664 			status = -EIO;
665 			goto bail;
666 		}
667 	} else {
668 		int len = ocfs2_sprintf_system_inode_name(namebuf, sizeof(namebuf),
669 							  type, OCFS2_INVALID_SLOT);
670 		status = ocfs2_lookup_ino_from_name(osb->sys_root_inode,
671 						    namebuf, len, &blkno);
672 		if (status < 0) {
673 			status = -ENOENT;
674 			goto bail;
675 		}
676 	}
677 
678 	status = ocfs2_info_freefrag_scan_bitmap(osb, gb_inode, blkno, oiff);
679 	if (status < 0)
680 		goto bail;
681 
682 	o2info_set_request_filled(&oiff->iff_req);
683 
684 	if (o2info_to_user(*oiff, req)) {
685 		status = -EFAULT;
686 		goto out_free;
687 	}
688 
689 	status = 0;
690 bail:
691 	if (status)
692 		o2info_set_request_error(&oiff->iff_req, req);
693 out_free:
694 	kfree(oiff);
695 out_err:
696 	return status;
697 }
698 
699 static int ocfs2_info_handle_unknown(struct inode *inode,
700 				     struct ocfs2_info_request __user *req)
701 {
702 	struct ocfs2_info_request oir;
703 
704 	if (o2info_from_user(oir, req))
705 		return -EFAULT;
706 
707 	o2info_clear_request_filled(&oir);
708 
709 	if (o2info_to_user(oir, req))
710 		return -EFAULT;
711 
712 	return 0;
713 }
714 
715 /*
716  * Validate and distinguish OCFS2_IOC_INFO requests.
717  *
718  * - validate the magic number.
719  * - distinguish different requests.
720  * - validate size of different requests.
721  */
722 static int ocfs2_info_handle_request(struct inode *inode,
723 				     struct ocfs2_info_request __user *req)
724 {
725 	int status = -EFAULT;
726 	struct ocfs2_info_request oir;
727 
728 	if (o2info_from_user(oir, req))
729 		goto bail;
730 
731 	status = -EINVAL;
732 	if (oir.ir_magic != OCFS2_INFO_MAGIC)
733 		goto bail;
734 
735 	switch (oir.ir_code) {
736 	case OCFS2_INFO_BLOCKSIZE:
737 		if (oir.ir_size == sizeof(struct ocfs2_info_blocksize))
738 			status = ocfs2_info_handle_blocksize(inode, req);
739 		break;
740 	case OCFS2_INFO_CLUSTERSIZE:
741 		if (oir.ir_size == sizeof(struct ocfs2_info_clustersize))
742 			status = ocfs2_info_handle_clustersize(inode, req);
743 		break;
744 	case OCFS2_INFO_MAXSLOTS:
745 		if (oir.ir_size == sizeof(struct ocfs2_info_maxslots))
746 			status = ocfs2_info_handle_maxslots(inode, req);
747 		break;
748 	case OCFS2_INFO_LABEL:
749 		if (oir.ir_size == sizeof(struct ocfs2_info_label))
750 			status = ocfs2_info_handle_label(inode, req);
751 		break;
752 	case OCFS2_INFO_UUID:
753 		if (oir.ir_size == sizeof(struct ocfs2_info_uuid))
754 			status = ocfs2_info_handle_uuid(inode, req);
755 		break;
756 	case OCFS2_INFO_FS_FEATURES:
757 		if (oir.ir_size == sizeof(struct ocfs2_info_fs_features))
758 			status = ocfs2_info_handle_fs_features(inode, req);
759 		break;
760 	case OCFS2_INFO_JOURNAL_SIZE:
761 		if (oir.ir_size == sizeof(struct ocfs2_info_journal_size))
762 			status = ocfs2_info_handle_journal_size(inode, req);
763 		break;
764 	case OCFS2_INFO_FREEINODE:
765 		if (oir.ir_size == sizeof(struct ocfs2_info_freeinode))
766 			status = ocfs2_info_handle_freeinode(inode, req);
767 		break;
768 	case OCFS2_INFO_FREEFRAG:
769 		if (oir.ir_size == sizeof(struct ocfs2_info_freefrag))
770 			status = ocfs2_info_handle_freefrag(inode, req);
771 		break;
772 	default:
773 		status = ocfs2_info_handle_unknown(inode, req);
774 		break;
775 	}
776 
777 bail:
778 	return status;
779 }
780 
781 static int ocfs2_get_request_ptr(struct ocfs2_info *info, int idx,
782 				 u64 *req_addr, int compat_flag)
783 {
784 	int status = -EFAULT;
785 	u64 __user *bp = NULL;
786 
787 	if (compat_flag) {
788 #ifdef CONFIG_COMPAT
789 		/*
790 		 * pointer bp stores the base address of a pointers array,
791 		 * which collects all addresses of separate request.
792 		 */
793 		bp = (u64 __user *)(unsigned long)compat_ptr(info->oi_requests);
794 #else
795 		BUG();
796 #endif
797 	} else
798 		bp = (u64 __user *)(unsigned long)(info->oi_requests);
799 
800 	if (o2info_from_user(*req_addr, bp + idx))
801 		goto bail;
802 
803 	status = 0;
804 bail:
805 	return status;
806 }
807 
808 /*
809  * OCFS2_IOC_INFO handles an array of requests passed from userspace.
810  *
811  * ocfs2_info_handle() receives a large info aggregation, grab and
812  * validate the request count from header, then break it into small
813  * pieces, later specific handlers can handle them one by one.
814  *
815  * Idea here is to make each separate request small enough to ensure
816  * a better backward&forward compatibility, since a small piece of
817  * request will be less likely to be broken if disk layout get changed.
818  */
819 static noinline_for_stack int
820 ocfs2_info_handle(struct inode *inode, struct ocfs2_info *info, int compat_flag)
821 {
822 	int i, status = 0;
823 	u64 req_addr;
824 	struct ocfs2_info_request __user *reqp;
825 
826 	if ((info->oi_count > OCFS2_INFO_MAX_REQUEST) ||
827 	    (!info->oi_requests)) {
828 		status = -EINVAL;
829 		goto bail;
830 	}
831 
832 	for (i = 0; i < info->oi_count; i++) {
833 
834 		status = ocfs2_get_request_ptr(info, i, &req_addr, compat_flag);
835 		if (status)
836 			break;
837 
838 		reqp = (struct ocfs2_info_request __user *)(unsigned long)req_addr;
839 		if (!reqp) {
840 			status = -EINVAL;
841 			goto bail;
842 		}
843 
844 		status = ocfs2_info_handle_request(inode, reqp);
845 		if (status)
846 			break;
847 	}
848 
849 bail:
850 	return status;
851 }
852 
853 long ocfs2_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
854 {
855 	struct inode *inode = file_inode(filp);
856 	void __user *argp = (void __user *)arg;
857 	int status;
858 
859 	switch (cmd) {
860 	case OCFS2_IOC_RESVSP:
861 	case OCFS2_IOC_RESVSP64:
862 	case OCFS2_IOC_UNRESVSP:
863 	case OCFS2_IOC_UNRESVSP64:
864 	{
865 		struct ocfs2_space_resv sr;
866 
867 		if (copy_from_user(&sr, (int __user *) arg, sizeof(sr)))
868 			return -EFAULT;
869 
870 		return ocfs2_change_file_space(filp, cmd, &sr);
871 	}
872 	case OCFS2_IOC_GROUP_EXTEND:
873 	{
874 		int new_clusters;
875 
876 		if (!capable(CAP_SYS_RESOURCE))
877 			return -EPERM;
878 
879 		if (get_user(new_clusters, (int __user *)arg))
880 			return -EFAULT;
881 
882 		status = mnt_want_write_file(filp);
883 		if (status)
884 			return status;
885 		status = ocfs2_group_extend(inode, new_clusters);
886 		mnt_drop_write_file(filp);
887 		return status;
888 	}
889 	case OCFS2_IOC_GROUP_ADD:
890 	case OCFS2_IOC_GROUP_ADD64:
891 	{
892 		struct ocfs2_new_group_input input;
893 
894 		if (!capable(CAP_SYS_RESOURCE))
895 			return -EPERM;
896 
897 		if (copy_from_user(&input, (int __user *) arg, sizeof(input)))
898 			return -EFAULT;
899 
900 		status = mnt_want_write_file(filp);
901 		if (status)
902 			return status;
903 		status = ocfs2_group_add(inode, &input);
904 		mnt_drop_write_file(filp);
905 		return status;
906 	}
907 	case OCFS2_IOC_REFLINK:
908 	{
909 		struct reflink_arguments args;
910 		const char __user *old_path;
911 		const char __user *new_path;
912 		bool preserve;
913 
914 		if (copy_from_user(&args, argp, sizeof(args)))
915 			return -EFAULT;
916 		old_path = (const char __user *)(unsigned long)args.old_path;
917 		new_path = (const char __user *)(unsigned long)args.new_path;
918 		preserve = (args.preserve != 0);
919 
920 		return ocfs2_reflink_ioctl(inode, old_path, new_path, preserve);
921 	}
922 	case OCFS2_IOC_INFO:
923 	{
924 		struct ocfs2_info info;
925 
926 		if (copy_from_user(&info, argp, sizeof(struct ocfs2_info)))
927 			return -EFAULT;
928 
929 		return ocfs2_info_handle(inode, &info, 0);
930 	}
931 	case FITRIM:
932 	{
933 		struct super_block *sb = inode->i_sb;
934 		struct fstrim_range range;
935 		int ret = 0;
936 
937 		if (!capable(CAP_SYS_ADMIN))
938 			return -EPERM;
939 
940 		if (!bdev_max_discard_sectors(sb->s_bdev))
941 			return -EOPNOTSUPP;
942 
943 		if (copy_from_user(&range, argp, sizeof(range)))
944 			return -EFAULT;
945 
946 		range.minlen = max_t(u64, bdev_discard_granularity(sb->s_bdev),
947 				     range.minlen);
948 		ret = ocfs2_trim_fs(sb, &range);
949 		if (ret < 0)
950 			return ret;
951 
952 		if (copy_to_user(argp, &range, sizeof(range)))
953 			return -EFAULT;
954 
955 		return 0;
956 	}
957 	case OCFS2_IOC_MOVE_EXT:
958 		return ocfs2_ioctl_move_extents(filp, argp);
959 	default:
960 		return -ENOTTY;
961 	}
962 }
963 
964 #ifdef CONFIG_COMPAT
965 long ocfs2_compat_ioctl(struct file *file, unsigned cmd, unsigned long arg)
966 {
967 	bool preserve;
968 	struct reflink_arguments args;
969 	struct inode *inode = file_inode(file);
970 	struct ocfs2_info info;
971 	void __user *argp = (void __user *)arg;
972 
973 	switch (cmd) {
974 	case OCFS2_IOC_RESVSP:
975 	case OCFS2_IOC_RESVSP64:
976 	case OCFS2_IOC_UNRESVSP:
977 	case OCFS2_IOC_UNRESVSP64:
978 	case OCFS2_IOC_GROUP_EXTEND:
979 	case OCFS2_IOC_GROUP_ADD:
980 	case OCFS2_IOC_GROUP_ADD64:
981 		break;
982 	case OCFS2_IOC_REFLINK:
983 		if (copy_from_user(&args, argp, sizeof(args)))
984 			return -EFAULT;
985 		preserve = (args.preserve != 0);
986 
987 		return ocfs2_reflink_ioctl(inode, compat_ptr(args.old_path),
988 					   compat_ptr(args.new_path), preserve);
989 	case OCFS2_IOC_INFO:
990 		if (copy_from_user(&info, argp, sizeof(struct ocfs2_info)))
991 			return -EFAULT;
992 
993 		return ocfs2_info_handle(inode, &info, 1);
994 	case FITRIM:
995 	case OCFS2_IOC_MOVE_EXT:
996 		break;
997 	default:
998 		return -ENOIOCTLCMD;
999 	}
1000 
1001 	return ocfs2_ioctl(file, cmd, arg);
1002 }
1003 #endif
1004