Loading...
1/*
2 * linux/fs/ocfs2/ioctl.c
3 *
4 * Copyright (C) 2006 Herbert Poetzl
5 * adapted from Remy Card's ext2/ioctl.c
6 */
7
8#include <linux/fs.h>
9#include <linux/mount.h>
10#include <linux/blkdev.h>
11#include <linux/compat.h>
12
13#include <cluster/masklog.h>
14
15#include "ocfs2.h"
16#include "alloc.h"
17#include "dlmglue.h"
18#include "file.h"
19#include "inode.h"
20#include "journal.h"
21
22#include "ocfs2_fs.h"
23#include "ioctl.h"
24#include "resize.h"
25#include "refcounttree.h"
26#include "sysfile.h"
27#include "dir.h"
28#include "buffer_head_io.h"
29#include "suballoc.h"
30#include "move_extents.h"
31
32#define o2info_from_user(a, b) \
33 copy_from_user(&(a), (b), sizeof(a))
34#define o2info_to_user(a, b) \
35 copy_to_user((typeof(a) __user *)b, &(a), sizeof(a))
36
37/*
38 * This is just a best-effort to tell userspace that this request
39 * caused the error.
40 */
41static inline void o2info_set_request_error(struct ocfs2_info_request *kreq,
42 struct ocfs2_info_request __user *req)
43{
44 kreq->ir_flags |= OCFS2_INFO_FL_ERROR;
45 (void)put_user(kreq->ir_flags, (__u32 __user *)&(req->ir_flags));
46}
47
48static inline void o2info_set_request_filled(struct ocfs2_info_request *req)
49{
50 req->ir_flags |= OCFS2_INFO_FL_FILLED;
51}
52
53static inline void o2info_clear_request_filled(struct ocfs2_info_request *req)
54{
55 req->ir_flags &= ~OCFS2_INFO_FL_FILLED;
56}
57
58static inline int o2info_coherent(struct ocfs2_info_request *req)
59{
60 return (!(req->ir_flags & OCFS2_INFO_FL_NON_COHERENT));
61}
62
63static int ocfs2_get_inode_attr(struct inode *inode, unsigned *flags)
64{
65 int status;
66
67 status = ocfs2_inode_lock(inode, NULL, 0);
68 if (status < 0) {
69 mlog_errno(status);
70 return status;
71 }
72 ocfs2_get_inode_flags(OCFS2_I(inode));
73 *flags = OCFS2_I(inode)->ip_attr;
74 ocfs2_inode_unlock(inode, 0);
75
76 return status;
77}
78
79static int ocfs2_set_inode_attr(struct inode *inode, unsigned flags,
80 unsigned mask)
81{
82 struct ocfs2_inode_info *ocfs2_inode = OCFS2_I(inode);
83 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
84 handle_t *handle = NULL;
85 struct buffer_head *bh = NULL;
86 unsigned oldflags;
87 int status;
88
89 inode_lock(inode);
90
91 status = ocfs2_inode_lock(inode, &bh, 1);
92 if (status < 0) {
93 mlog_errno(status);
94 goto bail;
95 }
96
97 status = -EACCES;
98 if (!inode_owner_or_capable(inode))
99 goto bail_unlock;
100
101 if (!S_ISDIR(inode->i_mode))
102 flags &= ~OCFS2_DIRSYNC_FL;
103
104 oldflags = ocfs2_inode->ip_attr;
105 flags = flags & mask;
106 flags |= oldflags & ~mask;
107
108 /*
109 * The IMMUTABLE and APPEND_ONLY flags can only be changed by
110 * the relevant capability.
111 */
112 status = -EPERM;
113 if ((oldflags & OCFS2_IMMUTABLE_FL) || ((flags ^ oldflags) &
114 (OCFS2_APPEND_FL | OCFS2_IMMUTABLE_FL))) {
115 if (!capable(CAP_LINUX_IMMUTABLE))
116 goto bail_unlock;
117 }
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
129 status = ocfs2_mark_inode_dirty(handle, inode, bh);
130 if (status < 0)
131 mlog_errno(status);
132
133 ocfs2_commit_trans(osb, handle);
134
135bail_unlock:
136 ocfs2_inode_unlock(inode, 1);
137bail:
138 inode_unlock(inode);
139
140 brelse(bh);
141
142 return status;
143}
144
145static int ocfs2_info_handle_blocksize(struct inode *inode,
146 struct ocfs2_info_request __user *req)
147{
148 struct ocfs2_info_blocksize oib;
149
150 if (o2info_from_user(oib, req))
151 return -EFAULT;
152
153 oib.ib_blocksize = inode->i_sb->s_blocksize;
154
155 o2info_set_request_filled(&oib.ib_req);
156
157 if (o2info_to_user(oib, req))
158 return -EFAULT;
159
160 return 0;
161}
162
163static int ocfs2_info_handle_clustersize(struct inode *inode,
164 struct ocfs2_info_request __user *req)
165{
166 struct ocfs2_info_clustersize oic;
167 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
168
169 if (o2info_from_user(oic, req))
170 return -EFAULT;
171
172 oic.ic_clustersize = osb->s_clustersize;
173
174 o2info_set_request_filled(&oic.ic_req);
175
176 if (o2info_to_user(oic, req))
177 return -EFAULT;
178
179 return 0;
180}
181
182static int ocfs2_info_handle_maxslots(struct inode *inode,
183 struct ocfs2_info_request __user *req)
184{
185 struct ocfs2_info_maxslots oim;
186 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
187
188 if (o2info_from_user(oim, req))
189 return -EFAULT;
190
191 oim.im_max_slots = osb->max_slots;
192
193 o2info_set_request_filled(&oim.im_req);
194
195 if (o2info_to_user(oim, req))
196 return -EFAULT;
197
198 return 0;
199}
200
201static int ocfs2_info_handle_label(struct inode *inode,
202 struct ocfs2_info_request __user *req)
203{
204 struct ocfs2_info_label oil;
205 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
206
207 if (o2info_from_user(oil, req))
208 return -EFAULT;
209
210 memcpy(oil.il_label, osb->vol_label, OCFS2_MAX_VOL_LABEL_LEN);
211
212 o2info_set_request_filled(&oil.il_req);
213
214 if (o2info_to_user(oil, req))
215 return -EFAULT;
216
217 return 0;
218}
219
220static int ocfs2_info_handle_uuid(struct inode *inode,
221 struct ocfs2_info_request __user *req)
222{
223 struct ocfs2_info_uuid oiu;
224 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
225
226 if (o2info_from_user(oiu, req))
227 return -EFAULT;
228
229 memcpy(oiu.iu_uuid_str, osb->uuid_str, OCFS2_TEXT_UUID_LEN + 1);
230
231 o2info_set_request_filled(&oiu.iu_req);
232
233 if (o2info_to_user(oiu, req))
234 return -EFAULT;
235
236 return 0;
237}
238
239static int ocfs2_info_handle_fs_features(struct inode *inode,
240 struct ocfs2_info_request __user *req)
241{
242 struct ocfs2_info_fs_features oif;
243 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
244
245 if (o2info_from_user(oif, req))
246 return -EFAULT;
247
248 oif.if_compat_features = osb->s_feature_compat;
249 oif.if_incompat_features = osb->s_feature_incompat;
250 oif.if_ro_compat_features = osb->s_feature_ro_compat;
251
252 o2info_set_request_filled(&oif.if_req);
253
254 if (o2info_to_user(oif, req))
255 return -EFAULT;
256
257 return 0;
258}
259
260static int ocfs2_info_handle_journal_size(struct inode *inode,
261 struct ocfs2_info_request __user *req)
262{
263 struct ocfs2_info_journal_size oij;
264 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
265
266 if (o2info_from_user(oij, req))
267 return -EFAULT;
268
269 oij.ij_journal_size = i_size_read(osb->journal->j_inode);
270
271 o2info_set_request_filled(&oij.ij_req);
272
273 if (o2info_to_user(oij, req))
274 return -EFAULT;
275
276 return 0;
277}
278
279static int ocfs2_info_scan_inode_alloc(struct ocfs2_super *osb,
280 struct inode *inode_alloc, u64 blkno,
281 struct ocfs2_info_freeinode *fi,
282 u32 slot)
283{
284 int status = 0, unlock = 0;
285
286 struct buffer_head *bh = NULL;
287 struct ocfs2_dinode *dinode_alloc = NULL;
288
289 if (inode_alloc)
290 inode_lock(inode_alloc);
291
292 if (o2info_coherent(&fi->ifi_req)) {
293 status = ocfs2_inode_lock(inode_alloc, &bh, 0);
294 if (status < 0) {
295 mlog_errno(status);
296 goto bail;
297 }
298 unlock = 1;
299 } else {
300 status = ocfs2_read_blocks_sync(osb, blkno, 1, &bh);
301 if (status < 0) {
302 mlog_errno(status);
303 goto bail;
304 }
305 }
306
307 dinode_alloc = (struct ocfs2_dinode *)bh->b_data;
308
309 fi->ifi_stat[slot].lfi_total =
310 le32_to_cpu(dinode_alloc->id1.bitmap1.i_total);
311 fi->ifi_stat[slot].lfi_free =
312 le32_to_cpu(dinode_alloc->id1.bitmap1.i_total) -
313 le32_to_cpu(dinode_alloc->id1.bitmap1.i_used);
314
315bail:
316 if (unlock)
317 ocfs2_inode_unlock(inode_alloc, 0);
318
319 if (inode_alloc)
320 inode_unlock(inode_alloc);
321
322 brelse(bh);
323
324 return status;
325}
326
327static int ocfs2_info_handle_freeinode(struct inode *inode,
328 struct ocfs2_info_request __user *req)
329{
330 u32 i;
331 u64 blkno = -1;
332 char namebuf[40];
333 int status, type = INODE_ALLOC_SYSTEM_INODE;
334 struct ocfs2_info_freeinode *oifi = NULL;
335 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
336 struct inode *inode_alloc = NULL;
337
338 oifi = kzalloc(sizeof(struct ocfs2_info_freeinode), GFP_KERNEL);
339 if (!oifi) {
340 status = -ENOMEM;
341 mlog_errno(status);
342 goto out_err;
343 }
344
345 if (o2info_from_user(*oifi, req)) {
346 status = -EFAULT;
347 goto out_free;
348 }
349
350 oifi->ifi_slotnum = osb->max_slots;
351
352 for (i = 0; i < oifi->ifi_slotnum; i++) {
353 if (o2info_coherent(&oifi->ifi_req)) {
354 inode_alloc = ocfs2_get_system_file_inode(osb, type, i);
355 if (!inode_alloc) {
356 mlog(ML_ERROR, "unable to get alloc inode in "
357 "slot %u\n", i);
358 status = -EIO;
359 goto bail;
360 }
361 } else {
362 ocfs2_sprintf_system_inode_name(namebuf,
363 sizeof(namebuf),
364 type, i);
365 status = ocfs2_lookup_ino_from_name(osb->sys_root_inode,
366 namebuf,
367 strlen(namebuf),
368 &blkno);
369 if (status < 0) {
370 status = -ENOENT;
371 goto bail;
372 }
373 }
374
375 status = ocfs2_info_scan_inode_alloc(osb, inode_alloc, blkno, oifi, i);
376
377 iput(inode_alloc);
378 inode_alloc = NULL;
379
380 if (status < 0)
381 goto bail;
382 }
383
384 o2info_set_request_filled(&oifi->ifi_req);
385
386 if (o2info_to_user(*oifi, req)) {
387 status = -EFAULT;
388 goto out_free;
389 }
390
391 status = 0;
392bail:
393 if (status)
394 o2info_set_request_error(&oifi->ifi_req, req);
395out_free:
396 kfree(oifi);
397out_err:
398 return status;
399}
400
401static void o2ffg_update_histogram(struct ocfs2_info_free_chunk_list *hist,
402 unsigned int chunksize)
403{
404 int index;
405
406 index = __ilog2_u32(chunksize);
407 if (index >= OCFS2_INFO_MAX_HIST)
408 index = OCFS2_INFO_MAX_HIST - 1;
409
410 hist->fc_chunks[index]++;
411 hist->fc_clusters[index] += chunksize;
412}
413
414static void o2ffg_update_stats(struct ocfs2_info_freefrag_stats *stats,
415 unsigned int chunksize)
416{
417 if (chunksize > stats->ffs_max)
418 stats->ffs_max = chunksize;
419
420 if (chunksize < stats->ffs_min)
421 stats->ffs_min = chunksize;
422
423 stats->ffs_avg += chunksize;
424 stats->ffs_free_chunks_real++;
425}
426
427static void ocfs2_info_update_ffg(struct ocfs2_info_freefrag *ffg,
428 unsigned int chunksize)
429{
430 o2ffg_update_histogram(&(ffg->iff_ffs.ffs_fc_hist), chunksize);
431 o2ffg_update_stats(&(ffg->iff_ffs), chunksize);
432}
433
434static int ocfs2_info_freefrag_scan_chain(struct ocfs2_super *osb,
435 struct inode *gb_inode,
436 struct ocfs2_dinode *gb_dinode,
437 struct ocfs2_chain_rec *rec,
438 struct ocfs2_info_freefrag *ffg,
439 u32 chunks_in_group)
440{
441 int status = 0, used;
442 u64 blkno;
443
444 struct buffer_head *bh = NULL;
445 struct ocfs2_group_desc *bg = NULL;
446
447 unsigned int max_bits, num_clusters;
448 unsigned int offset = 0, cluster, chunk;
449 unsigned int chunk_free, last_chunksize = 0;
450
451 if (!le32_to_cpu(rec->c_free))
452 goto bail;
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 offset = 0;
486
487 for (chunk = 0; chunk < chunks_in_group; chunk++) {
488 /*
489 * last chunk may be not an entire one.
490 */
491 if ((offset + ffg->iff_chunksize) > max_bits)
492 num_clusters = max_bits - offset;
493 else
494 num_clusters = ffg->iff_chunksize;
495
496 chunk_free = 0;
497 for (cluster = 0; cluster < num_clusters; cluster++) {
498 used = ocfs2_test_bit(offset,
499 (unsigned long *)bg->bg_bitmap);
500 /*
501 * - chunk_free counts free clusters in #N chunk.
502 * - last_chunksize records the size(in) clusters
503 * for the last real free chunk being counted.
504 */
505 if (!used) {
506 last_chunksize++;
507 chunk_free++;
508 }
509
510 if (used && last_chunksize) {
511 ocfs2_info_update_ffg(ffg,
512 last_chunksize);
513 last_chunksize = 0;
514 }
515
516 offset++;
517 }
518
519 if (chunk_free == ffg->iff_chunksize)
520 ffg->iff_ffs.ffs_free_chunks++;
521 }
522
523 /*
524 * need to update the info for last free chunk.
525 */
526 if (last_chunksize)
527 ocfs2_info_update_ffg(ffg, last_chunksize);
528
529 } while (le64_to_cpu(bg->bg_next_group));
530
531bail:
532 brelse(bh);
533
534 return status;
535}
536
537static int ocfs2_info_freefrag_scan_bitmap(struct ocfs2_super *osb,
538 struct inode *gb_inode, u64 blkno,
539 struct ocfs2_info_freefrag *ffg)
540{
541 u32 chunks_in_group;
542 int status = 0, unlock = 0, i;
543
544 struct buffer_head *bh = NULL;
545 struct ocfs2_chain_list *cl = NULL;
546 struct ocfs2_chain_rec *rec = NULL;
547 struct ocfs2_dinode *gb_dinode = NULL;
548
549 if (gb_inode)
550 inode_lock(gb_inode);
551
552 if (o2info_coherent(&ffg->iff_req)) {
553 status = ocfs2_inode_lock(gb_inode, &bh, 0);
554 if (status < 0) {
555 mlog_errno(status);
556 goto bail;
557 }
558 unlock = 1;
559 } else {
560 status = ocfs2_read_blocks_sync(osb, blkno, 1, &bh);
561 if (status < 0) {
562 mlog_errno(status);
563 goto bail;
564 }
565 }
566
567 gb_dinode = (struct ocfs2_dinode *)bh->b_data;
568 cl = &(gb_dinode->id2.i_chain);
569
570 /*
571 * Chunksize(in) clusters from userspace should be
572 * less than clusters in a group.
573 */
574 if (ffg->iff_chunksize > le16_to_cpu(cl->cl_cpg)) {
575 status = -EINVAL;
576 goto bail;
577 }
578
579 memset(&ffg->iff_ffs, 0, sizeof(struct ocfs2_info_freefrag_stats));
580
581 ffg->iff_ffs.ffs_min = ~0U;
582 ffg->iff_ffs.ffs_clusters =
583 le32_to_cpu(gb_dinode->id1.bitmap1.i_total);
584 ffg->iff_ffs.ffs_free_clusters = ffg->iff_ffs.ffs_clusters -
585 le32_to_cpu(gb_dinode->id1.bitmap1.i_used);
586
587 chunks_in_group = le16_to_cpu(cl->cl_cpg) / ffg->iff_chunksize + 1;
588
589 for (i = 0; i < le16_to_cpu(cl->cl_next_free_rec); i++) {
590 rec = &(cl->cl_recs[i]);
591 status = ocfs2_info_freefrag_scan_chain(osb, gb_inode,
592 gb_dinode,
593 rec, ffg,
594 chunks_in_group);
595 if (status)
596 goto bail;
597 }
598
599 if (ffg->iff_ffs.ffs_free_chunks_real)
600 ffg->iff_ffs.ffs_avg = (ffg->iff_ffs.ffs_avg /
601 ffg->iff_ffs.ffs_free_chunks_real);
602bail:
603 if (unlock)
604 ocfs2_inode_unlock(gb_inode, 0);
605
606 if (gb_inode)
607 inode_unlock(gb_inode);
608
609 iput(gb_inode);
610 brelse(bh);
611
612 return status;
613}
614
615static int ocfs2_info_handle_freefrag(struct inode *inode,
616 struct ocfs2_info_request __user *req)
617{
618 u64 blkno = -1;
619 char namebuf[40];
620 int status, type = GLOBAL_BITMAP_SYSTEM_INODE;
621
622 struct ocfs2_info_freefrag *oiff;
623 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
624 struct inode *gb_inode = NULL;
625
626 oiff = kzalloc(sizeof(struct ocfs2_info_freefrag), GFP_KERNEL);
627 if (!oiff) {
628 status = -ENOMEM;
629 mlog_errno(status);
630 goto out_err;
631 }
632
633 if (o2info_from_user(*oiff, req)) {
634 status = -EFAULT;
635 goto out_free;
636 }
637 /*
638 * chunksize from userspace should be power of 2.
639 */
640 if ((oiff->iff_chunksize & (oiff->iff_chunksize - 1)) ||
641 (!oiff->iff_chunksize)) {
642 status = -EINVAL;
643 goto bail;
644 }
645
646 if (o2info_coherent(&oiff->iff_req)) {
647 gb_inode = ocfs2_get_system_file_inode(osb, type,
648 OCFS2_INVALID_SLOT);
649 if (!gb_inode) {
650 mlog(ML_ERROR, "unable to get global_bitmap inode\n");
651 status = -EIO;
652 goto bail;
653 }
654 } else {
655 ocfs2_sprintf_system_inode_name(namebuf, sizeof(namebuf), type,
656 OCFS2_INVALID_SLOT);
657 status = ocfs2_lookup_ino_from_name(osb->sys_root_inode,
658 namebuf,
659 strlen(namebuf),
660 &blkno);
661 if (status < 0) {
662 status = -ENOENT;
663 goto bail;
664 }
665 }
666
667 status = ocfs2_info_freefrag_scan_bitmap(osb, gb_inode, blkno, oiff);
668 if (status < 0)
669 goto bail;
670
671 o2info_set_request_filled(&oiff->iff_req);
672
673 if (o2info_to_user(*oiff, req)) {
674 status = -EFAULT;
675 goto out_free;
676 }
677
678 status = 0;
679bail:
680 if (status)
681 o2info_set_request_error(&oiff->iff_req, req);
682out_free:
683 kfree(oiff);
684out_err:
685 return status;
686}
687
688static int ocfs2_info_handle_unknown(struct inode *inode,
689 struct ocfs2_info_request __user *req)
690{
691 struct ocfs2_info_request oir;
692
693 if (o2info_from_user(oir, req))
694 return -EFAULT;
695
696 o2info_clear_request_filled(&oir);
697
698 if (o2info_to_user(oir, req))
699 return -EFAULT;
700
701 return 0;
702}
703
704/*
705 * Validate and distinguish OCFS2_IOC_INFO requests.
706 *
707 * - validate the magic number.
708 * - distinguish different requests.
709 * - validate size of different requests.
710 */
711static int ocfs2_info_handle_request(struct inode *inode,
712 struct ocfs2_info_request __user *req)
713{
714 int status = -EFAULT;
715 struct ocfs2_info_request oir;
716
717 if (o2info_from_user(oir, req))
718 goto bail;
719
720 status = -EINVAL;
721 if (oir.ir_magic != OCFS2_INFO_MAGIC)
722 goto bail;
723
724 switch (oir.ir_code) {
725 case OCFS2_INFO_BLOCKSIZE:
726 if (oir.ir_size == sizeof(struct ocfs2_info_blocksize))
727 status = ocfs2_info_handle_blocksize(inode, req);
728 break;
729 case OCFS2_INFO_CLUSTERSIZE:
730 if (oir.ir_size == sizeof(struct ocfs2_info_clustersize))
731 status = ocfs2_info_handle_clustersize(inode, req);
732 break;
733 case OCFS2_INFO_MAXSLOTS:
734 if (oir.ir_size == sizeof(struct ocfs2_info_maxslots))
735 status = ocfs2_info_handle_maxslots(inode, req);
736 break;
737 case OCFS2_INFO_LABEL:
738 if (oir.ir_size == sizeof(struct ocfs2_info_label))
739 status = ocfs2_info_handle_label(inode, req);
740 break;
741 case OCFS2_INFO_UUID:
742 if (oir.ir_size == sizeof(struct ocfs2_info_uuid))
743 status = ocfs2_info_handle_uuid(inode, req);
744 break;
745 case OCFS2_INFO_FS_FEATURES:
746 if (oir.ir_size == sizeof(struct ocfs2_info_fs_features))
747 status = ocfs2_info_handle_fs_features(inode, req);
748 break;
749 case OCFS2_INFO_JOURNAL_SIZE:
750 if (oir.ir_size == sizeof(struct ocfs2_info_journal_size))
751 status = ocfs2_info_handle_journal_size(inode, req);
752 break;
753 case OCFS2_INFO_FREEINODE:
754 if (oir.ir_size == sizeof(struct ocfs2_info_freeinode))
755 status = ocfs2_info_handle_freeinode(inode, req);
756 break;
757 case OCFS2_INFO_FREEFRAG:
758 if (oir.ir_size == sizeof(struct ocfs2_info_freefrag))
759 status = ocfs2_info_handle_freefrag(inode, req);
760 break;
761 default:
762 status = ocfs2_info_handle_unknown(inode, req);
763 break;
764 }
765
766bail:
767 return status;
768}
769
770static int ocfs2_get_request_ptr(struct ocfs2_info *info, int idx,
771 u64 *req_addr, int compat_flag)
772{
773 int status = -EFAULT;
774 u64 __user *bp = NULL;
775
776 if (compat_flag) {
777#ifdef CONFIG_COMPAT
778 /*
779 * pointer bp stores the base address of a pointers array,
780 * which collects all addresses of separate request.
781 */
782 bp = (u64 __user *)(unsigned long)compat_ptr(info->oi_requests);
783#else
784 BUG();
785#endif
786 } else
787 bp = (u64 __user *)(unsigned long)(info->oi_requests);
788
789 if (o2info_from_user(*req_addr, bp + idx))
790 goto bail;
791
792 status = 0;
793bail:
794 return status;
795}
796
797/*
798 * OCFS2_IOC_INFO handles an array of requests passed from userspace.
799 *
800 * ocfs2_info_handle() recevies a large info aggregation, grab and
801 * validate the request count from header, then break it into small
802 * pieces, later specific handlers can handle them one by one.
803 *
804 * Idea here is to make each separate request small enough to ensure
805 * a better backward&forward compatibility, since a small piece of
806 * request will be less likely to be broken if disk layout get changed.
807 */
808static int ocfs2_info_handle(struct inode *inode, struct ocfs2_info *info,
809 int compat_flag)
810{
811 int i, status = 0;
812 u64 req_addr;
813 struct ocfs2_info_request __user *reqp;
814
815 if ((info->oi_count > OCFS2_INFO_MAX_REQUEST) ||
816 (!info->oi_requests)) {
817 status = -EINVAL;
818 goto bail;
819 }
820
821 for (i = 0; i < info->oi_count; i++) {
822
823 status = ocfs2_get_request_ptr(info, i, &req_addr, compat_flag);
824 if (status)
825 break;
826
827 reqp = (struct ocfs2_info_request __user *)(unsigned long)req_addr;
828 if (!reqp) {
829 status = -EINVAL;
830 goto bail;
831 }
832
833 status = ocfs2_info_handle_request(inode, reqp);
834 if (status)
835 break;
836 }
837
838bail:
839 return status;
840}
841
842long ocfs2_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
843{
844 struct inode *inode = file_inode(filp);
845 unsigned int flags;
846 int new_clusters;
847 int status;
848 struct ocfs2_space_resv sr;
849 struct ocfs2_new_group_input input;
850 struct reflink_arguments args;
851 const char __user *old_path;
852 const char __user *new_path;
853 bool preserve;
854 struct ocfs2_info info;
855 void __user *argp = (void __user *)arg;
856
857 switch (cmd) {
858 case OCFS2_IOC_GETFLAGS:
859 status = ocfs2_get_inode_attr(inode, &flags);
860 if (status < 0)
861 return status;
862
863 flags &= OCFS2_FL_VISIBLE;
864 return put_user(flags, (int __user *) arg);
865 case OCFS2_IOC_SETFLAGS:
866 if (get_user(flags, (int __user *) arg))
867 return -EFAULT;
868
869 status = mnt_want_write_file(filp);
870 if (status)
871 return status;
872 status = ocfs2_set_inode_attr(inode, flags,
873 OCFS2_FL_MODIFIABLE);
874 mnt_drop_write_file(filp);
875 return status;
876 case OCFS2_IOC_RESVSP:
877 case OCFS2_IOC_RESVSP64:
878 case OCFS2_IOC_UNRESVSP:
879 case OCFS2_IOC_UNRESVSP64:
880 if (copy_from_user(&sr, (int __user *) arg, sizeof(sr)))
881 return -EFAULT;
882
883 return ocfs2_change_file_space(filp, cmd, &sr);
884 case OCFS2_IOC_GROUP_EXTEND:
885 if (!capable(CAP_SYS_RESOURCE))
886 return -EPERM;
887
888 if (get_user(new_clusters, (int __user *)arg))
889 return -EFAULT;
890
891 status = mnt_want_write_file(filp);
892 if (status)
893 return status;
894 status = ocfs2_group_extend(inode, new_clusters);
895 mnt_drop_write_file(filp);
896 return status;
897 case OCFS2_IOC_GROUP_ADD:
898 case OCFS2_IOC_GROUP_ADD64:
899 if (!capable(CAP_SYS_RESOURCE))
900 return -EPERM;
901
902 if (copy_from_user(&input, (int __user *) arg, sizeof(input)))
903 return -EFAULT;
904
905 status = mnt_want_write_file(filp);
906 if (status)
907 return status;
908 status = ocfs2_group_add(inode, &input);
909 mnt_drop_write_file(filp);
910 return status;
911 case OCFS2_IOC_REFLINK:
912 if (copy_from_user(&args, argp, sizeof(args)))
913 return -EFAULT;
914 old_path = (const char __user *)(unsigned long)args.old_path;
915 new_path = (const char __user *)(unsigned long)args.new_path;
916 preserve = (args.preserve != 0);
917
918 return ocfs2_reflink_ioctl(inode, old_path, new_path, preserve);
919 case OCFS2_IOC_INFO:
920 if (copy_from_user(&info, argp, sizeof(struct ocfs2_info)))
921 return -EFAULT;
922
923 return ocfs2_info_handle(inode, &info, 0);
924 case FITRIM:
925 {
926 struct super_block *sb = inode->i_sb;
927 struct request_queue *q = bdev_get_queue(sb->s_bdev);
928 struct fstrim_range range;
929 int ret = 0;
930
931 if (!capable(CAP_SYS_ADMIN))
932 return -EPERM;
933
934 if (!blk_queue_discard(q))
935 return -EOPNOTSUPP;
936
937 if (copy_from_user(&range, argp, sizeof(range)))
938 return -EFAULT;
939
940 range.minlen = max_t(u64, q->limits.discard_granularity,
941 range.minlen);
942 ret = ocfs2_trim_fs(sb, &range);
943 if (ret < 0)
944 return ret;
945
946 if (copy_to_user(argp, &range, sizeof(range)))
947 return -EFAULT;
948
949 return 0;
950 }
951 case OCFS2_IOC_MOVE_EXT:
952 return ocfs2_ioctl_move_extents(filp, argp);
953 default:
954 return -ENOTTY;
955 }
956}
957
958#ifdef CONFIG_COMPAT
959long ocfs2_compat_ioctl(struct file *file, unsigned cmd, unsigned long arg)
960{
961 bool preserve;
962 struct reflink_arguments args;
963 struct inode *inode = file_inode(file);
964 struct ocfs2_info info;
965 void __user *argp = (void __user *)arg;
966
967 switch (cmd) {
968 case OCFS2_IOC32_GETFLAGS:
969 cmd = OCFS2_IOC_GETFLAGS;
970 break;
971 case OCFS2_IOC32_SETFLAGS:
972 cmd = OCFS2_IOC_SETFLAGS;
973 break;
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 OCFS2_IOC_MOVE_EXT:
995 break;
996 default:
997 return -ENOIOCTLCMD;
998 }
999
1000 return ocfs2_ioctl(file, cmd, arg);
1001}
1002#endif
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 */
43static 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
50static inline void o2info_set_request_filled(struct ocfs2_info_request *req)
51{
52 req->ir_flags |= OCFS2_INFO_FL_FILLED;
53}
54
55static inline void o2info_clear_request_filled(struct ocfs2_info_request *req)
56{
57 req->ir_flags &= ~OCFS2_INFO_FL_FILLED;
58}
59
60static inline int o2info_coherent(struct ocfs2_info_request *req)
61{
62 return (!(req->ir_flags & OCFS2_INFO_FL_NON_COHERENT));
63}
64
65int ocfs2_fileattr_get(struct dentry *dentry, struct fileattr *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
85int ocfs2_fileattr_set(struct mnt_idmap *idmap,
86 struct dentry *dentry, struct fileattr *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
136bail_unlock:
137 ocfs2_inode_unlock(inode, 1);
138bail:
139 brelse(bh);
140
141 return status;
142}
143
144static 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
162static 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
181static 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
200static 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
219static 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
238static 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
259static 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
278static 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
314bail:
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
326static 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(sizeof(struct ocfs2_info_freeinode), GFP_KERNEL);
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 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,
366 strlen(namebuf),
367 &blkno);
368 if (status < 0) {
369 status = -ENOENT;
370 goto bail;
371 }
372 }
373
374 status = ocfs2_info_scan_inode_alloc(osb, inode_alloc, blkno, oifi, i);
375
376 iput(inode_alloc);
377 inode_alloc = NULL;
378
379 if (status < 0)
380 goto bail;
381 }
382
383 o2info_set_request_filled(&oifi->ifi_req);
384
385 if (o2info_to_user(*oifi, req)) {
386 status = -EFAULT;
387 goto out_free;
388 }
389
390 status = 0;
391bail:
392 if (status)
393 o2info_set_request_error(&oifi->ifi_req, req);
394out_free:
395 kfree(oifi);
396out_err:
397 return status;
398}
399
400static void o2ffg_update_histogram(struct ocfs2_info_free_chunk_list *hist,
401 unsigned int chunksize)
402{
403 u32 index;
404
405 index = __ilog2_u32(chunksize);
406 if (index >= OCFS2_INFO_MAX_HIST)
407 index = OCFS2_INFO_MAX_HIST - 1;
408
409 hist->fc_chunks[index]++;
410 hist->fc_clusters[index] += chunksize;
411}
412
413static void o2ffg_update_stats(struct ocfs2_info_freefrag_stats *stats,
414 unsigned int chunksize)
415{
416 if (chunksize > stats->ffs_max)
417 stats->ffs_max = chunksize;
418
419 if (chunksize < stats->ffs_min)
420 stats->ffs_min = chunksize;
421
422 stats->ffs_avg += chunksize;
423 stats->ffs_free_chunks_real++;
424}
425
426static void ocfs2_info_update_ffg(struct ocfs2_info_freefrag *ffg,
427 unsigned int chunksize)
428{
429 o2ffg_update_histogram(&(ffg->iff_ffs.ffs_fc_hist), chunksize);
430 o2ffg_update_stats(&(ffg->iff_ffs), chunksize);
431}
432
433static int ocfs2_info_freefrag_scan_chain(struct ocfs2_super *osb,
434 struct inode *gb_inode,
435 struct ocfs2_dinode *gb_dinode,
436 struct ocfs2_chain_rec *rec,
437 struct ocfs2_info_freefrag *ffg,
438 u32 chunks_in_group)
439{
440 int status = 0, used;
441 u64 blkno;
442
443 struct buffer_head *bh = NULL;
444 struct ocfs2_group_desc *bg = NULL;
445
446 unsigned int max_bits, num_clusters;
447 unsigned int offset = 0, cluster, chunk;
448 unsigned int chunk_free, last_chunksize = 0;
449
450 if (!le32_to_cpu(rec->c_free))
451 goto bail;
452
453 do {
454 if (!bg)
455 blkno = le64_to_cpu(rec->c_blkno);
456 else
457 blkno = le64_to_cpu(bg->bg_next_group);
458
459 if (bh) {
460 brelse(bh);
461 bh = NULL;
462 }
463
464 if (o2info_coherent(&ffg->iff_req))
465 status = ocfs2_read_group_descriptor(gb_inode,
466 gb_dinode,
467 blkno, &bh);
468 else
469 status = ocfs2_read_blocks_sync(osb, blkno, 1, &bh);
470
471 if (status < 0) {
472 mlog(ML_ERROR, "Can't read the group descriptor # "
473 "%llu from device.", (unsigned long long)blkno);
474 status = -EIO;
475 goto bail;
476 }
477
478 bg = (struct ocfs2_group_desc *)bh->b_data;
479
480 if (!le16_to_cpu(bg->bg_free_bits_count))
481 continue;
482
483 max_bits = le16_to_cpu(bg->bg_bits);
484 offset = 0;
485
486 for (chunk = 0; chunk < chunks_in_group; chunk++) {
487 /*
488 * last chunk may be not an entire one.
489 */
490 if ((offset + ffg->iff_chunksize) > max_bits)
491 num_clusters = max_bits - offset;
492 else
493 num_clusters = ffg->iff_chunksize;
494
495 chunk_free = 0;
496 for (cluster = 0; cluster < num_clusters; cluster++) {
497 used = ocfs2_test_bit(offset,
498 (unsigned long *)bg->bg_bitmap);
499 /*
500 * - chunk_free counts free clusters in #N chunk.
501 * - last_chunksize records the size(in) clusters
502 * for the last real free chunk being counted.
503 */
504 if (!used) {
505 last_chunksize++;
506 chunk_free++;
507 }
508
509 if (used && last_chunksize) {
510 ocfs2_info_update_ffg(ffg,
511 last_chunksize);
512 last_chunksize = 0;
513 }
514
515 offset++;
516 }
517
518 if (chunk_free == ffg->iff_chunksize)
519 ffg->iff_ffs.ffs_free_chunks++;
520 }
521
522 /*
523 * need to update the info for last free chunk.
524 */
525 if (last_chunksize)
526 ocfs2_info_update_ffg(ffg, last_chunksize);
527
528 } while (le64_to_cpu(bg->bg_next_group));
529
530bail:
531 brelse(bh);
532
533 return status;
534}
535
536static int ocfs2_info_freefrag_scan_bitmap(struct ocfs2_super *osb,
537 struct inode *gb_inode, u64 blkno,
538 struct ocfs2_info_freefrag *ffg)
539{
540 u32 chunks_in_group;
541 int status = 0, unlock = 0, i;
542
543 struct buffer_head *bh = NULL;
544 struct ocfs2_chain_list *cl = NULL;
545 struct ocfs2_chain_rec *rec = NULL;
546 struct ocfs2_dinode *gb_dinode = NULL;
547
548 if (gb_inode)
549 inode_lock(gb_inode);
550
551 if (o2info_coherent(&ffg->iff_req)) {
552 status = ocfs2_inode_lock(gb_inode, &bh, 0);
553 if (status < 0) {
554 mlog_errno(status);
555 goto bail;
556 }
557 unlock = 1;
558 } else {
559 status = ocfs2_read_blocks_sync(osb, blkno, 1, &bh);
560 if (status < 0) {
561 mlog_errno(status);
562 goto bail;
563 }
564 }
565
566 gb_dinode = (struct ocfs2_dinode *)bh->b_data;
567 cl = &(gb_dinode->id2.i_chain);
568
569 /*
570 * Chunksize(in) clusters from userspace should be
571 * less than clusters in a group.
572 */
573 if (ffg->iff_chunksize > le16_to_cpu(cl->cl_cpg)) {
574 status = -EINVAL;
575 goto bail;
576 }
577
578 memset(&ffg->iff_ffs, 0, sizeof(struct ocfs2_info_freefrag_stats));
579
580 ffg->iff_ffs.ffs_min = ~0U;
581 ffg->iff_ffs.ffs_clusters =
582 le32_to_cpu(gb_dinode->id1.bitmap1.i_total);
583 ffg->iff_ffs.ffs_free_clusters = ffg->iff_ffs.ffs_clusters -
584 le32_to_cpu(gb_dinode->id1.bitmap1.i_used);
585
586 chunks_in_group = le16_to_cpu(cl->cl_cpg) / ffg->iff_chunksize + 1;
587
588 for (i = 0; i < le16_to_cpu(cl->cl_next_free_rec); i++) {
589 rec = &(cl->cl_recs[i]);
590 status = ocfs2_info_freefrag_scan_chain(osb, gb_inode,
591 gb_dinode,
592 rec, ffg,
593 chunks_in_group);
594 if (status)
595 goto bail;
596 }
597
598 if (ffg->iff_ffs.ffs_free_chunks_real)
599 ffg->iff_ffs.ffs_avg = (ffg->iff_ffs.ffs_avg /
600 ffg->iff_ffs.ffs_free_chunks_real);
601bail:
602 if (unlock)
603 ocfs2_inode_unlock(gb_inode, 0);
604
605 if (gb_inode)
606 inode_unlock(gb_inode);
607
608 iput(gb_inode);
609 brelse(bh);
610
611 return status;
612}
613
614static int ocfs2_info_handle_freefrag(struct inode *inode,
615 struct ocfs2_info_request __user *req)
616{
617 u64 blkno = -1;
618 char namebuf[40];
619 int status, type = GLOBAL_BITMAP_SYSTEM_INODE;
620
621 struct ocfs2_info_freefrag *oiff;
622 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
623 struct inode *gb_inode = NULL;
624
625 oiff = kzalloc(sizeof(struct ocfs2_info_freefrag), GFP_KERNEL);
626 if (!oiff) {
627 status = -ENOMEM;
628 mlog_errno(status);
629 goto out_err;
630 }
631
632 if (o2info_from_user(*oiff, req)) {
633 status = -EFAULT;
634 goto out_free;
635 }
636 /*
637 * chunksize from userspace should be power of 2.
638 */
639 if ((oiff->iff_chunksize & (oiff->iff_chunksize - 1)) ||
640 (!oiff->iff_chunksize)) {
641 status = -EINVAL;
642 goto bail;
643 }
644
645 if (o2info_coherent(&oiff->iff_req)) {
646 gb_inode = ocfs2_get_system_file_inode(osb, type,
647 OCFS2_INVALID_SLOT);
648 if (!gb_inode) {
649 mlog(ML_ERROR, "unable to get global_bitmap inode\n");
650 status = -EIO;
651 goto bail;
652 }
653 } else {
654 ocfs2_sprintf_system_inode_name(namebuf, sizeof(namebuf), type,
655 OCFS2_INVALID_SLOT);
656 status = ocfs2_lookup_ino_from_name(osb->sys_root_inode,
657 namebuf,
658 strlen(namebuf),
659 &blkno);
660 if (status < 0) {
661 status = -ENOENT;
662 goto bail;
663 }
664 }
665
666 status = ocfs2_info_freefrag_scan_bitmap(osb, gb_inode, blkno, oiff);
667 if (status < 0)
668 goto bail;
669
670 o2info_set_request_filled(&oiff->iff_req);
671
672 if (o2info_to_user(*oiff, req)) {
673 status = -EFAULT;
674 goto out_free;
675 }
676
677 status = 0;
678bail:
679 if (status)
680 o2info_set_request_error(&oiff->iff_req, req);
681out_free:
682 kfree(oiff);
683out_err:
684 return status;
685}
686
687static int ocfs2_info_handle_unknown(struct inode *inode,
688 struct ocfs2_info_request __user *req)
689{
690 struct ocfs2_info_request oir;
691
692 if (o2info_from_user(oir, req))
693 return -EFAULT;
694
695 o2info_clear_request_filled(&oir);
696
697 if (o2info_to_user(oir, req))
698 return -EFAULT;
699
700 return 0;
701}
702
703/*
704 * Validate and distinguish OCFS2_IOC_INFO requests.
705 *
706 * - validate the magic number.
707 * - distinguish different requests.
708 * - validate size of different requests.
709 */
710static int ocfs2_info_handle_request(struct inode *inode,
711 struct ocfs2_info_request __user *req)
712{
713 int status = -EFAULT;
714 struct ocfs2_info_request oir;
715
716 if (o2info_from_user(oir, req))
717 goto bail;
718
719 status = -EINVAL;
720 if (oir.ir_magic != OCFS2_INFO_MAGIC)
721 goto bail;
722
723 switch (oir.ir_code) {
724 case OCFS2_INFO_BLOCKSIZE:
725 if (oir.ir_size == sizeof(struct ocfs2_info_blocksize))
726 status = ocfs2_info_handle_blocksize(inode, req);
727 break;
728 case OCFS2_INFO_CLUSTERSIZE:
729 if (oir.ir_size == sizeof(struct ocfs2_info_clustersize))
730 status = ocfs2_info_handle_clustersize(inode, req);
731 break;
732 case OCFS2_INFO_MAXSLOTS:
733 if (oir.ir_size == sizeof(struct ocfs2_info_maxslots))
734 status = ocfs2_info_handle_maxslots(inode, req);
735 break;
736 case OCFS2_INFO_LABEL:
737 if (oir.ir_size == sizeof(struct ocfs2_info_label))
738 status = ocfs2_info_handle_label(inode, req);
739 break;
740 case OCFS2_INFO_UUID:
741 if (oir.ir_size == sizeof(struct ocfs2_info_uuid))
742 status = ocfs2_info_handle_uuid(inode, req);
743 break;
744 case OCFS2_INFO_FS_FEATURES:
745 if (oir.ir_size == sizeof(struct ocfs2_info_fs_features))
746 status = ocfs2_info_handle_fs_features(inode, req);
747 break;
748 case OCFS2_INFO_JOURNAL_SIZE:
749 if (oir.ir_size == sizeof(struct ocfs2_info_journal_size))
750 status = ocfs2_info_handle_journal_size(inode, req);
751 break;
752 case OCFS2_INFO_FREEINODE:
753 if (oir.ir_size == sizeof(struct ocfs2_info_freeinode))
754 status = ocfs2_info_handle_freeinode(inode, req);
755 break;
756 case OCFS2_INFO_FREEFRAG:
757 if (oir.ir_size == sizeof(struct ocfs2_info_freefrag))
758 status = ocfs2_info_handle_freefrag(inode, req);
759 break;
760 default:
761 status = ocfs2_info_handle_unknown(inode, req);
762 break;
763 }
764
765bail:
766 return status;
767}
768
769static int ocfs2_get_request_ptr(struct ocfs2_info *info, int idx,
770 u64 *req_addr, int compat_flag)
771{
772 int status = -EFAULT;
773 u64 __user *bp = NULL;
774
775 if (compat_flag) {
776#ifdef CONFIG_COMPAT
777 /*
778 * pointer bp stores the base address of a pointers array,
779 * which collects all addresses of separate request.
780 */
781 bp = (u64 __user *)(unsigned long)compat_ptr(info->oi_requests);
782#else
783 BUG();
784#endif
785 } else
786 bp = (u64 __user *)(unsigned long)(info->oi_requests);
787
788 if (o2info_from_user(*req_addr, bp + idx))
789 goto bail;
790
791 status = 0;
792bail:
793 return status;
794}
795
796/*
797 * OCFS2_IOC_INFO handles an array of requests passed from userspace.
798 *
799 * ocfs2_info_handle() recevies a large info aggregation, grab and
800 * validate the request count from header, then break it into small
801 * pieces, later specific handlers can handle them one by one.
802 *
803 * Idea here is to make each separate request small enough to ensure
804 * a better backward&forward compatibility, since a small piece of
805 * request will be less likely to be broken if disk layout get changed.
806 */
807static noinline_for_stack int
808ocfs2_info_handle(struct inode *inode, struct ocfs2_info *info, int compat_flag)
809{
810 int i, status = 0;
811 u64 req_addr;
812 struct ocfs2_info_request __user *reqp;
813
814 if ((info->oi_count > OCFS2_INFO_MAX_REQUEST) ||
815 (!info->oi_requests)) {
816 status = -EINVAL;
817 goto bail;
818 }
819
820 for (i = 0; i < info->oi_count; i++) {
821
822 status = ocfs2_get_request_ptr(info, i, &req_addr, compat_flag);
823 if (status)
824 break;
825
826 reqp = (struct ocfs2_info_request __user *)(unsigned long)req_addr;
827 if (!reqp) {
828 status = -EINVAL;
829 goto bail;
830 }
831
832 status = ocfs2_info_handle_request(inode, reqp);
833 if (status)
834 break;
835 }
836
837bail:
838 return status;
839}
840
841long ocfs2_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
842{
843 struct inode *inode = file_inode(filp);
844 void __user *argp = (void __user *)arg;
845 int status;
846
847 switch (cmd) {
848 case OCFS2_IOC_RESVSP:
849 case OCFS2_IOC_RESVSP64:
850 case OCFS2_IOC_UNRESVSP:
851 case OCFS2_IOC_UNRESVSP64:
852 {
853 struct ocfs2_space_resv sr;
854
855 if (copy_from_user(&sr, (int __user *) arg, sizeof(sr)))
856 return -EFAULT;
857
858 return ocfs2_change_file_space(filp, cmd, &sr);
859 }
860 case OCFS2_IOC_GROUP_EXTEND:
861 {
862 int new_clusters;
863
864 if (!capable(CAP_SYS_RESOURCE))
865 return -EPERM;
866
867 if (get_user(new_clusters, (int __user *)arg))
868 return -EFAULT;
869
870 status = mnt_want_write_file(filp);
871 if (status)
872 return status;
873 status = ocfs2_group_extend(inode, new_clusters);
874 mnt_drop_write_file(filp);
875 return status;
876 }
877 case OCFS2_IOC_GROUP_ADD:
878 case OCFS2_IOC_GROUP_ADD64:
879 {
880 struct ocfs2_new_group_input input;
881
882 if (!capable(CAP_SYS_RESOURCE))
883 return -EPERM;
884
885 if (copy_from_user(&input, (int __user *) arg, sizeof(input)))
886 return -EFAULT;
887
888 status = mnt_want_write_file(filp);
889 if (status)
890 return status;
891 status = ocfs2_group_add(inode, &input);
892 mnt_drop_write_file(filp);
893 return status;
894 }
895 case OCFS2_IOC_REFLINK:
896 {
897 struct reflink_arguments args;
898 const char __user *old_path;
899 const char __user *new_path;
900 bool preserve;
901
902 if (copy_from_user(&args, argp, sizeof(args)))
903 return -EFAULT;
904 old_path = (const char __user *)(unsigned long)args.old_path;
905 new_path = (const char __user *)(unsigned long)args.new_path;
906 preserve = (args.preserve != 0);
907
908 return ocfs2_reflink_ioctl(inode, old_path, new_path, preserve);
909 }
910 case OCFS2_IOC_INFO:
911 {
912 struct ocfs2_info info;
913
914 if (copy_from_user(&info, argp, sizeof(struct ocfs2_info)))
915 return -EFAULT;
916
917 return ocfs2_info_handle(inode, &info, 0);
918 }
919 case FITRIM:
920 {
921 struct super_block *sb = inode->i_sb;
922 struct fstrim_range range;
923 int ret = 0;
924
925 if (!capable(CAP_SYS_ADMIN))
926 return -EPERM;
927
928 if (!bdev_max_discard_sectors(sb->s_bdev))
929 return -EOPNOTSUPP;
930
931 if (copy_from_user(&range, argp, sizeof(range)))
932 return -EFAULT;
933
934 range.minlen = max_t(u64, bdev_discard_granularity(sb->s_bdev),
935 range.minlen);
936 ret = ocfs2_trim_fs(sb, &range);
937 if (ret < 0)
938 return ret;
939
940 if (copy_to_user(argp, &range, sizeof(range)))
941 return -EFAULT;
942
943 return 0;
944 }
945 case OCFS2_IOC_MOVE_EXT:
946 return ocfs2_ioctl_move_extents(filp, argp);
947 default:
948 return -ENOTTY;
949 }
950}
951
952#ifdef CONFIG_COMPAT
953long ocfs2_compat_ioctl(struct file *file, unsigned cmd, unsigned long arg)
954{
955 bool preserve;
956 struct reflink_arguments args;
957 struct inode *inode = file_inode(file);
958 struct ocfs2_info info;
959 void __user *argp = (void __user *)arg;
960
961 switch (cmd) {
962 case OCFS2_IOC_RESVSP:
963 case OCFS2_IOC_RESVSP64:
964 case OCFS2_IOC_UNRESVSP:
965 case OCFS2_IOC_UNRESVSP64:
966 case OCFS2_IOC_GROUP_EXTEND:
967 case OCFS2_IOC_GROUP_ADD:
968 case OCFS2_IOC_GROUP_ADD64:
969 break;
970 case OCFS2_IOC_REFLINK:
971 if (copy_from_user(&args, argp, sizeof(args)))
972 return -EFAULT;
973 preserve = (args.preserve != 0);
974
975 return ocfs2_reflink_ioctl(inode, compat_ptr(args.old_path),
976 compat_ptr(args.new_path), preserve);
977 case OCFS2_IOC_INFO:
978 if (copy_from_user(&info, argp, sizeof(struct ocfs2_info)))
979 return -EFAULT;
980
981 return ocfs2_info_handle(inode, &info, 1);
982 case FITRIM:
983 case OCFS2_IOC_MOVE_EXT:
984 break;
985 default:
986 return -ENOIOCTLCMD;
987 }
988
989 return ocfs2_ioctl(file, cmd, arg);
990}
991#endif