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 call is void because we are already reporting an error that may
39 * be -EFAULT. The error will be returned from the ioctl(2) call. It's
40 * just a best-effort to tell userspace that this request caused the error.
41 */
42static inline void o2info_set_request_error(struct ocfs2_info_request *kreq,
43 struct ocfs2_info_request __user *req)
44{
45 kreq->ir_flags |= OCFS2_INFO_FL_ERROR;
46 (void)put_user(kreq->ir_flags, (__u32 __user *)&(req->ir_flags));
47}
48
49static inline void o2info_set_request_filled(struct ocfs2_info_request *req)
50{
51 req->ir_flags |= OCFS2_INFO_FL_FILLED;
52}
53
54static inline void o2info_clear_request_filled(struct ocfs2_info_request *req)
55{
56 req->ir_flags &= ~OCFS2_INFO_FL_FILLED;
57}
58
59static inline int o2info_coherent(struct ocfs2_info_request *req)
60{
61 return (!(req->ir_flags & OCFS2_INFO_FL_NON_COHERENT));
62}
63
64static int ocfs2_get_inode_attr(struct inode *inode, unsigned *flags)
65{
66 int status;
67
68 status = ocfs2_inode_lock(inode, NULL, 0);
69 if (status < 0) {
70 mlog_errno(status);
71 return status;
72 }
73 ocfs2_get_inode_flags(OCFS2_I(inode));
74 *flags = OCFS2_I(inode)->ip_attr;
75 ocfs2_inode_unlock(inode, 0);
76
77 return status;
78}
79
80static int ocfs2_set_inode_attr(struct inode *inode, unsigned flags,
81 unsigned mask)
82{
83 struct ocfs2_inode_info *ocfs2_inode = OCFS2_I(inode);
84 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
85 handle_t *handle = NULL;
86 struct buffer_head *bh = NULL;
87 unsigned oldflags;
88 int status;
89
90 mutex_lock(&inode->i_mutex);
91
92 status = ocfs2_inode_lock(inode, &bh, 1);
93 if (status < 0) {
94 mlog_errno(status);
95 goto bail;
96 }
97
98 status = -EACCES;
99 if (!inode_owner_or_capable(inode))
100 goto bail_unlock;
101
102 if (!S_ISDIR(inode->i_mode))
103 flags &= ~OCFS2_DIRSYNC_FL;
104
105 oldflags = ocfs2_inode->ip_attr;
106 flags = flags & mask;
107 flags |= oldflags & ~mask;
108
109 /*
110 * The IMMUTABLE and APPEND_ONLY flags can only be changed by
111 * the relevant capability.
112 */
113 status = -EPERM;
114 if ((oldflags & OCFS2_IMMUTABLE_FL) || ((flags ^ oldflags) &
115 (OCFS2_APPEND_FL | OCFS2_IMMUTABLE_FL))) {
116 if (!capable(CAP_LINUX_IMMUTABLE))
117 goto bail_unlock;
118 }
119
120 handle = ocfs2_start_trans(osb, OCFS2_INODE_UPDATE_CREDITS);
121 if (IS_ERR(handle)) {
122 status = PTR_ERR(handle);
123 mlog_errno(status);
124 goto bail_unlock;
125 }
126
127 ocfs2_inode->ip_attr = flags;
128 ocfs2_set_inode_flags(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 mutex_unlock(&inode->i_mutex);
140
141 brelse(bh);
142
143 return status;
144}
145
146int ocfs2_info_handle_blocksize(struct inode *inode,
147 struct ocfs2_info_request __user *req)
148{
149 int status = -EFAULT;
150 struct ocfs2_info_blocksize oib;
151
152 if (o2info_from_user(oib, req))
153 goto bail;
154
155 oib.ib_blocksize = inode->i_sb->s_blocksize;
156
157 o2info_set_request_filled(&oib.ib_req);
158
159 if (o2info_to_user(oib, req))
160 goto bail;
161
162 status = 0;
163bail:
164 if (status)
165 o2info_set_request_error(&oib.ib_req, req);
166
167 return status;
168}
169
170int ocfs2_info_handle_clustersize(struct inode *inode,
171 struct ocfs2_info_request __user *req)
172{
173 int status = -EFAULT;
174 struct ocfs2_info_clustersize oic;
175 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
176
177 if (o2info_from_user(oic, req))
178 goto bail;
179
180 oic.ic_clustersize = osb->s_clustersize;
181
182 o2info_set_request_filled(&oic.ic_req);
183
184 if (o2info_to_user(oic, req))
185 goto bail;
186
187 status = 0;
188bail:
189 if (status)
190 o2info_set_request_error(&oic.ic_req, req);
191
192 return status;
193}
194
195int ocfs2_info_handle_maxslots(struct inode *inode,
196 struct ocfs2_info_request __user *req)
197{
198 int status = -EFAULT;
199 struct ocfs2_info_maxslots oim;
200 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
201
202 if (o2info_from_user(oim, req))
203 goto bail;
204
205 oim.im_max_slots = osb->max_slots;
206
207 o2info_set_request_filled(&oim.im_req);
208
209 if (o2info_to_user(oim, req))
210 goto bail;
211
212 status = 0;
213bail:
214 if (status)
215 o2info_set_request_error(&oim.im_req, req);
216
217 return status;
218}
219
220int ocfs2_info_handle_label(struct inode *inode,
221 struct ocfs2_info_request __user *req)
222{
223 int status = -EFAULT;
224 struct ocfs2_info_label oil;
225 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
226
227 if (o2info_from_user(oil, req))
228 goto bail;
229
230 memcpy(oil.il_label, osb->vol_label, OCFS2_MAX_VOL_LABEL_LEN);
231
232 o2info_set_request_filled(&oil.il_req);
233
234 if (o2info_to_user(oil, req))
235 goto bail;
236
237 status = 0;
238bail:
239 if (status)
240 o2info_set_request_error(&oil.il_req, req);
241
242 return status;
243}
244
245int ocfs2_info_handle_uuid(struct inode *inode,
246 struct ocfs2_info_request __user *req)
247{
248 int status = -EFAULT;
249 struct ocfs2_info_uuid oiu;
250 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
251
252 if (o2info_from_user(oiu, req))
253 goto bail;
254
255 memcpy(oiu.iu_uuid_str, osb->uuid_str, OCFS2_TEXT_UUID_LEN + 1);
256
257 o2info_set_request_filled(&oiu.iu_req);
258
259 if (o2info_to_user(oiu, req))
260 goto bail;
261
262 status = 0;
263bail:
264 if (status)
265 o2info_set_request_error(&oiu.iu_req, req);
266
267 return status;
268}
269
270int ocfs2_info_handle_fs_features(struct inode *inode,
271 struct ocfs2_info_request __user *req)
272{
273 int status = -EFAULT;
274 struct ocfs2_info_fs_features oif;
275 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
276
277 if (o2info_from_user(oif, req))
278 goto bail;
279
280 oif.if_compat_features = osb->s_feature_compat;
281 oif.if_incompat_features = osb->s_feature_incompat;
282 oif.if_ro_compat_features = osb->s_feature_ro_compat;
283
284 o2info_set_request_filled(&oif.if_req);
285
286 if (o2info_to_user(oif, req))
287 goto bail;
288
289 status = 0;
290bail:
291 if (status)
292 o2info_set_request_error(&oif.if_req, req);
293
294 return status;
295}
296
297int ocfs2_info_handle_journal_size(struct inode *inode,
298 struct ocfs2_info_request __user *req)
299{
300 int status = -EFAULT;
301 struct ocfs2_info_journal_size oij;
302 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
303
304 if (o2info_from_user(oij, req))
305 goto bail;
306
307 oij.ij_journal_size = i_size_read(osb->journal->j_inode);
308
309 o2info_set_request_filled(&oij.ij_req);
310
311 if (o2info_to_user(oij, req))
312 goto bail;
313
314 status = 0;
315bail:
316 if (status)
317 o2info_set_request_error(&oij.ij_req, req);
318
319 return status;
320}
321
322int ocfs2_info_scan_inode_alloc(struct ocfs2_super *osb,
323 struct inode *inode_alloc, u64 blkno,
324 struct ocfs2_info_freeinode *fi, u32 slot)
325{
326 int status = 0, unlock = 0;
327
328 struct buffer_head *bh = NULL;
329 struct ocfs2_dinode *dinode_alloc = NULL;
330
331 if (inode_alloc)
332 mutex_lock(&inode_alloc->i_mutex);
333
334 if (o2info_coherent(&fi->ifi_req)) {
335 status = ocfs2_inode_lock(inode_alloc, &bh, 0);
336 if (status < 0) {
337 mlog_errno(status);
338 goto bail;
339 }
340 unlock = 1;
341 } else {
342 status = ocfs2_read_blocks_sync(osb, blkno, 1, &bh);
343 if (status < 0) {
344 mlog_errno(status);
345 goto bail;
346 }
347 }
348
349 dinode_alloc = (struct ocfs2_dinode *)bh->b_data;
350
351 fi->ifi_stat[slot].lfi_total =
352 le32_to_cpu(dinode_alloc->id1.bitmap1.i_total);
353 fi->ifi_stat[slot].lfi_free =
354 le32_to_cpu(dinode_alloc->id1.bitmap1.i_total) -
355 le32_to_cpu(dinode_alloc->id1.bitmap1.i_used);
356
357bail:
358 if (unlock)
359 ocfs2_inode_unlock(inode_alloc, 0);
360
361 if (inode_alloc)
362 mutex_unlock(&inode_alloc->i_mutex);
363
364 brelse(bh);
365
366 return status;
367}
368
369int ocfs2_info_handle_freeinode(struct inode *inode,
370 struct ocfs2_info_request __user *req)
371{
372 u32 i;
373 u64 blkno = -1;
374 char namebuf[40];
375 int status = -EFAULT, type = INODE_ALLOC_SYSTEM_INODE;
376 struct ocfs2_info_freeinode *oifi = NULL;
377 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
378 struct inode *inode_alloc = NULL;
379
380 oifi = kzalloc(sizeof(struct ocfs2_info_freeinode), GFP_KERNEL);
381 if (!oifi) {
382 status = -ENOMEM;
383 mlog_errno(status);
384 goto out_err;
385 }
386
387 if (o2info_from_user(*oifi, req))
388 goto bail;
389
390 oifi->ifi_slotnum = osb->max_slots;
391
392 for (i = 0; i < oifi->ifi_slotnum; i++) {
393 if (o2info_coherent(&oifi->ifi_req)) {
394 inode_alloc = ocfs2_get_system_file_inode(osb, type, i);
395 if (!inode_alloc) {
396 mlog(ML_ERROR, "unable to get alloc inode in "
397 "slot %u\n", i);
398 status = -EIO;
399 goto bail;
400 }
401 } else {
402 ocfs2_sprintf_system_inode_name(namebuf,
403 sizeof(namebuf),
404 type, i);
405 status = ocfs2_lookup_ino_from_name(osb->sys_root_inode,
406 namebuf,
407 strlen(namebuf),
408 &blkno);
409 if (status < 0) {
410 status = -ENOENT;
411 goto bail;
412 }
413 }
414
415 status = ocfs2_info_scan_inode_alloc(osb, inode_alloc, blkno, oifi, i);
416
417 iput(inode_alloc);
418 inode_alloc = NULL;
419
420 if (status < 0)
421 goto bail;
422 }
423
424 o2info_set_request_filled(&oifi->ifi_req);
425
426 if (o2info_to_user(*oifi, req))
427 goto bail;
428
429 status = 0;
430bail:
431 if (status)
432 o2info_set_request_error(&oifi->ifi_req, req);
433
434 kfree(oifi);
435out_err:
436 return status;
437}
438
439static void o2ffg_update_histogram(struct ocfs2_info_free_chunk_list *hist,
440 unsigned int chunksize)
441{
442 int index;
443
444 index = __ilog2_u32(chunksize);
445 if (index >= OCFS2_INFO_MAX_HIST)
446 index = OCFS2_INFO_MAX_HIST - 1;
447
448 hist->fc_chunks[index]++;
449 hist->fc_clusters[index] += chunksize;
450}
451
452static void o2ffg_update_stats(struct ocfs2_info_freefrag_stats *stats,
453 unsigned int chunksize)
454{
455 if (chunksize > stats->ffs_max)
456 stats->ffs_max = chunksize;
457
458 if (chunksize < stats->ffs_min)
459 stats->ffs_min = chunksize;
460
461 stats->ffs_avg += chunksize;
462 stats->ffs_free_chunks_real++;
463}
464
465void ocfs2_info_update_ffg(struct ocfs2_info_freefrag *ffg,
466 unsigned int chunksize)
467{
468 o2ffg_update_histogram(&(ffg->iff_ffs.ffs_fc_hist), chunksize);
469 o2ffg_update_stats(&(ffg->iff_ffs), chunksize);
470}
471
472int ocfs2_info_freefrag_scan_chain(struct ocfs2_super *osb,
473 struct inode *gb_inode,
474 struct ocfs2_dinode *gb_dinode,
475 struct ocfs2_chain_rec *rec,
476 struct ocfs2_info_freefrag *ffg,
477 u32 chunks_in_group)
478{
479 int status = 0, used;
480 u64 blkno;
481
482 struct buffer_head *bh = NULL;
483 struct ocfs2_group_desc *bg = NULL;
484
485 unsigned int max_bits, num_clusters;
486 unsigned int offset = 0, cluster, chunk;
487 unsigned int chunk_free, last_chunksize = 0;
488
489 if (!le32_to_cpu(rec->c_free))
490 goto bail;
491
492 do {
493 if (!bg)
494 blkno = le64_to_cpu(rec->c_blkno);
495 else
496 blkno = le64_to_cpu(bg->bg_next_group);
497
498 if (bh) {
499 brelse(bh);
500 bh = NULL;
501 }
502
503 if (o2info_coherent(&ffg->iff_req))
504 status = ocfs2_read_group_descriptor(gb_inode,
505 gb_dinode,
506 blkno, &bh);
507 else
508 status = ocfs2_read_blocks_sync(osb, blkno, 1, &bh);
509
510 if (status < 0) {
511 mlog(ML_ERROR, "Can't read the group descriptor # "
512 "%llu from device.", (unsigned long long)blkno);
513 status = -EIO;
514 goto bail;
515 }
516
517 bg = (struct ocfs2_group_desc *)bh->b_data;
518
519 if (!le16_to_cpu(bg->bg_free_bits_count))
520 continue;
521
522 max_bits = le16_to_cpu(bg->bg_bits);
523 offset = 0;
524
525 for (chunk = 0; chunk < chunks_in_group; chunk++) {
526 /*
527 * last chunk may be not an entire one.
528 */
529 if ((offset + ffg->iff_chunksize) > max_bits)
530 num_clusters = max_bits - offset;
531 else
532 num_clusters = ffg->iff_chunksize;
533
534 chunk_free = 0;
535 for (cluster = 0; cluster < num_clusters; cluster++) {
536 used = ocfs2_test_bit(offset,
537 (unsigned long *)bg->bg_bitmap);
538 /*
539 * - chunk_free counts free clusters in #N chunk.
540 * - last_chunksize records the size(in) clusters
541 * for the last real free chunk being counted.
542 */
543 if (!used) {
544 last_chunksize++;
545 chunk_free++;
546 }
547
548 if (used && last_chunksize) {
549 ocfs2_info_update_ffg(ffg,
550 last_chunksize);
551 last_chunksize = 0;
552 }
553
554 offset++;
555 }
556
557 if (chunk_free == ffg->iff_chunksize)
558 ffg->iff_ffs.ffs_free_chunks++;
559 }
560
561 /*
562 * need to update the info for last free chunk.
563 */
564 if (last_chunksize)
565 ocfs2_info_update_ffg(ffg, last_chunksize);
566
567 } while (le64_to_cpu(bg->bg_next_group));
568
569bail:
570 brelse(bh);
571
572 return status;
573}
574
575int ocfs2_info_freefrag_scan_bitmap(struct ocfs2_super *osb,
576 struct inode *gb_inode, u64 blkno,
577 struct ocfs2_info_freefrag *ffg)
578{
579 u32 chunks_in_group;
580 int status = 0, unlock = 0, i;
581
582 struct buffer_head *bh = NULL;
583 struct ocfs2_chain_list *cl = NULL;
584 struct ocfs2_chain_rec *rec = NULL;
585 struct ocfs2_dinode *gb_dinode = NULL;
586
587 if (gb_inode)
588 mutex_lock(&gb_inode->i_mutex);
589
590 if (o2info_coherent(&ffg->iff_req)) {
591 status = ocfs2_inode_lock(gb_inode, &bh, 0);
592 if (status < 0) {
593 mlog_errno(status);
594 goto bail;
595 }
596 unlock = 1;
597 } else {
598 status = ocfs2_read_blocks_sync(osb, blkno, 1, &bh);
599 if (status < 0) {
600 mlog_errno(status);
601 goto bail;
602 }
603 }
604
605 gb_dinode = (struct ocfs2_dinode *)bh->b_data;
606 cl = &(gb_dinode->id2.i_chain);
607
608 /*
609 * Chunksize(in) clusters from userspace should be
610 * less than clusters in a group.
611 */
612 if (ffg->iff_chunksize > le16_to_cpu(cl->cl_cpg)) {
613 status = -EINVAL;
614 goto bail;
615 }
616
617 memset(&ffg->iff_ffs, 0, sizeof(struct ocfs2_info_freefrag_stats));
618
619 ffg->iff_ffs.ffs_min = ~0U;
620 ffg->iff_ffs.ffs_clusters =
621 le32_to_cpu(gb_dinode->id1.bitmap1.i_total);
622 ffg->iff_ffs.ffs_free_clusters = ffg->iff_ffs.ffs_clusters -
623 le32_to_cpu(gb_dinode->id1.bitmap1.i_used);
624
625 chunks_in_group = le16_to_cpu(cl->cl_cpg) / ffg->iff_chunksize + 1;
626
627 for (i = 0; i < le16_to_cpu(cl->cl_next_free_rec); i++) {
628 rec = &(cl->cl_recs[i]);
629 status = ocfs2_info_freefrag_scan_chain(osb, gb_inode,
630 gb_dinode,
631 rec, ffg,
632 chunks_in_group);
633 if (status)
634 goto bail;
635 }
636
637 if (ffg->iff_ffs.ffs_free_chunks_real)
638 ffg->iff_ffs.ffs_avg = (ffg->iff_ffs.ffs_avg /
639 ffg->iff_ffs.ffs_free_chunks_real);
640bail:
641 if (unlock)
642 ocfs2_inode_unlock(gb_inode, 0);
643
644 if (gb_inode)
645 mutex_unlock(&gb_inode->i_mutex);
646
647 if (gb_inode)
648 iput(gb_inode);
649
650 brelse(bh);
651
652 return status;
653}
654
655int ocfs2_info_handle_freefrag(struct inode *inode,
656 struct ocfs2_info_request __user *req)
657{
658 u64 blkno = -1;
659 char namebuf[40];
660 int status = -EFAULT, type = GLOBAL_BITMAP_SYSTEM_INODE;
661
662 struct ocfs2_info_freefrag *oiff;
663 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
664 struct inode *gb_inode = NULL;
665
666 oiff = kzalloc(sizeof(struct ocfs2_info_freefrag), GFP_KERNEL);
667 if (!oiff) {
668 status = -ENOMEM;
669 mlog_errno(status);
670 goto out_err;
671 }
672
673 if (o2info_from_user(*oiff, req))
674 goto bail;
675 /*
676 * chunksize from userspace should be power of 2.
677 */
678 if ((oiff->iff_chunksize & (oiff->iff_chunksize - 1)) ||
679 (!oiff->iff_chunksize)) {
680 status = -EINVAL;
681 goto bail;
682 }
683
684 if (o2info_coherent(&oiff->iff_req)) {
685 gb_inode = ocfs2_get_system_file_inode(osb, type,
686 OCFS2_INVALID_SLOT);
687 if (!gb_inode) {
688 mlog(ML_ERROR, "unable to get global_bitmap inode\n");
689 status = -EIO;
690 goto bail;
691 }
692 } else {
693 ocfs2_sprintf_system_inode_name(namebuf, sizeof(namebuf), type,
694 OCFS2_INVALID_SLOT);
695 status = ocfs2_lookup_ino_from_name(osb->sys_root_inode,
696 namebuf,
697 strlen(namebuf),
698 &blkno);
699 if (status < 0) {
700 status = -ENOENT;
701 goto bail;
702 }
703 }
704
705 status = ocfs2_info_freefrag_scan_bitmap(osb, gb_inode, blkno, oiff);
706 if (status < 0)
707 goto bail;
708
709 o2info_set_request_filled(&oiff->iff_req);
710
711 if (o2info_to_user(*oiff, req)) {
712 status = -EFAULT;
713 goto bail;
714 }
715
716 status = 0;
717bail:
718 if (status)
719 o2info_set_request_error(&oiff->iff_req, req);
720
721 kfree(oiff);
722out_err:
723 return status;
724}
725
726int ocfs2_info_handle_unknown(struct inode *inode,
727 struct ocfs2_info_request __user *req)
728{
729 int status = -EFAULT;
730 struct ocfs2_info_request oir;
731
732 if (o2info_from_user(oir, req))
733 goto bail;
734
735 o2info_clear_request_filled(&oir);
736
737 if (o2info_to_user(oir, req))
738 goto bail;
739
740 status = 0;
741bail:
742 if (status)
743 o2info_set_request_error(&oir, req);
744
745 return status;
746}
747
748/*
749 * Validate and distinguish OCFS2_IOC_INFO requests.
750 *
751 * - validate the magic number.
752 * - distinguish different requests.
753 * - validate size of different requests.
754 */
755int ocfs2_info_handle_request(struct inode *inode,
756 struct ocfs2_info_request __user *req)
757{
758 int status = -EFAULT;
759 struct ocfs2_info_request oir;
760
761 if (o2info_from_user(oir, req))
762 goto bail;
763
764 status = -EINVAL;
765 if (oir.ir_magic != OCFS2_INFO_MAGIC)
766 goto bail;
767
768 switch (oir.ir_code) {
769 case OCFS2_INFO_BLOCKSIZE:
770 if (oir.ir_size == sizeof(struct ocfs2_info_blocksize))
771 status = ocfs2_info_handle_blocksize(inode, req);
772 break;
773 case OCFS2_INFO_CLUSTERSIZE:
774 if (oir.ir_size == sizeof(struct ocfs2_info_clustersize))
775 status = ocfs2_info_handle_clustersize(inode, req);
776 break;
777 case OCFS2_INFO_MAXSLOTS:
778 if (oir.ir_size == sizeof(struct ocfs2_info_maxslots))
779 status = ocfs2_info_handle_maxslots(inode, req);
780 break;
781 case OCFS2_INFO_LABEL:
782 if (oir.ir_size == sizeof(struct ocfs2_info_label))
783 status = ocfs2_info_handle_label(inode, req);
784 break;
785 case OCFS2_INFO_UUID:
786 if (oir.ir_size == sizeof(struct ocfs2_info_uuid))
787 status = ocfs2_info_handle_uuid(inode, req);
788 break;
789 case OCFS2_INFO_FS_FEATURES:
790 if (oir.ir_size == sizeof(struct ocfs2_info_fs_features))
791 status = ocfs2_info_handle_fs_features(inode, req);
792 break;
793 case OCFS2_INFO_JOURNAL_SIZE:
794 if (oir.ir_size == sizeof(struct ocfs2_info_journal_size))
795 status = ocfs2_info_handle_journal_size(inode, req);
796 break;
797 case OCFS2_INFO_FREEINODE:
798 if (oir.ir_size == sizeof(struct ocfs2_info_freeinode))
799 status = ocfs2_info_handle_freeinode(inode, req);
800 break;
801 case OCFS2_INFO_FREEFRAG:
802 if (oir.ir_size == sizeof(struct ocfs2_info_freefrag))
803 status = ocfs2_info_handle_freefrag(inode, req);
804 break;
805 default:
806 status = ocfs2_info_handle_unknown(inode, req);
807 break;
808 }
809
810bail:
811 return status;
812}
813
814int ocfs2_get_request_ptr(struct ocfs2_info *info, int idx,
815 u64 *req_addr, int compat_flag)
816{
817 int status = -EFAULT;
818 u64 __user *bp = NULL;
819
820 if (compat_flag) {
821#ifdef CONFIG_COMPAT
822 /*
823 * pointer bp stores the base address of a pointers array,
824 * which collects all addresses of separate request.
825 */
826 bp = (u64 __user *)(unsigned long)compat_ptr(info->oi_requests);
827#else
828 BUG();
829#endif
830 } else
831 bp = (u64 __user *)(unsigned long)(info->oi_requests);
832
833 if (o2info_from_user(*req_addr, bp + idx))
834 goto bail;
835
836 status = 0;
837bail:
838 return status;
839}
840
841/*
842 * OCFS2_IOC_INFO handles an array of requests passed from userspace.
843 *
844 * ocfs2_info_handle() recevies a large info aggregation, grab and
845 * validate the request count from header, then break it into small
846 * pieces, later specific handlers can handle them one by one.
847 *
848 * Idea here is to make each separate request small enough to ensure
849 * a better backward&forward compatibility, since a small piece of
850 * request will be less likely to be broken if disk layout get changed.
851 */
852int ocfs2_info_handle(struct inode *inode, struct ocfs2_info *info,
853 int compat_flag)
854{
855 int i, status = 0;
856 u64 req_addr;
857 struct ocfs2_info_request __user *reqp;
858
859 if ((info->oi_count > OCFS2_INFO_MAX_REQUEST) ||
860 (!info->oi_requests)) {
861 status = -EINVAL;
862 goto bail;
863 }
864
865 for (i = 0; i < info->oi_count; i++) {
866
867 status = ocfs2_get_request_ptr(info, i, &req_addr, compat_flag);
868 if (status)
869 break;
870
871 reqp = (struct ocfs2_info_request __user *)(unsigned long)req_addr;
872 if (!reqp) {
873 status = -EINVAL;
874 goto bail;
875 }
876
877 status = ocfs2_info_handle_request(inode, reqp);
878 if (status)
879 break;
880 }
881
882bail:
883 return status;
884}
885
886long ocfs2_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
887{
888 struct inode *inode = file_inode(filp);
889 unsigned int flags;
890 int new_clusters;
891 int status;
892 struct ocfs2_space_resv sr;
893 struct ocfs2_new_group_input input;
894 struct reflink_arguments args;
895 const char __user *old_path;
896 const char __user *new_path;
897 bool preserve;
898 struct ocfs2_info info;
899 void __user *argp = (void __user *)arg;
900
901 switch (cmd) {
902 case OCFS2_IOC_GETFLAGS:
903 status = ocfs2_get_inode_attr(inode, &flags);
904 if (status < 0)
905 return status;
906
907 flags &= OCFS2_FL_VISIBLE;
908 return put_user(flags, (int __user *) arg);
909 case OCFS2_IOC_SETFLAGS:
910 if (get_user(flags, (int __user *) arg))
911 return -EFAULT;
912
913 status = mnt_want_write_file(filp);
914 if (status)
915 return status;
916 status = ocfs2_set_inode_attr(inode, flags,
917 OCFS2_FL_MODIFIABLE);
918 mnt_drop_write_file(filp);
919 return status;
920 case OCFS2_IOC_RESVSP:
921 case OCFS2_IOC_RESVSP64:
922 case OCFS2_IOC_UNRESVSP:
923 case OCFS2_IOC_UNRESVSP64:
924 if (copy_from_user(&sr, (int __user *) arg, sizeof(sr)))
925 return -EFAULT;
926
927 return ocfs2_change_file_space(filp, cmd, &sr);
928 case OCFS2_IOC_GROUP_EXTEND:
929 if (!capable(CAP_SYS_RESOURCE))
930 return -EPERM;
931
932 if (get_user(new_clusters, (int __user *)arg))
933 return -EFAULT;
934
935 status = mnt_want_write_file(filp);
936 if (status)
937 return status;
938 status = ocfs2_group_extend(inode, new_clusters);
939 mnt_drop_write_file(filp);
940 return status;
941 case OCFS2_IOC_GROUP_ADD:
942 case OCFS2_IOC_GROUP_ADD64:
943 if (!capable(CAP_SYS_RESOURCE))
944 return -EPERM;
945
946 if (copy_from_user(&input, (int __user *) arg, sizeof(input)))
947 return -EFAULT;
948
949 status = mnt_want_write_file(filp);
950 if (status)
951 return status;
952 status = ocfs2_group_add(inode, &input);
953 mnt_drop_write_file(filp);
954 return status;
955 case OCFS2_IOC_REFLINK:
956 if (copy_from_user(&args, argp, sizeof(args)))
957 return -EFAULT;
958 old_path = (const char __user *)(unsigned long)args.old_path;
959 new_path = (const char __user *)(unsigned long)args.new_path;
960 preserve = (args.preserve != 0);
961
962 return ocfs2_reflink_ioctl(inode, old_path, new_path, preserve);
963 case OCFS2_IOC_INFO:
964 if (copy_from_user(&info, argp, sizeof(struct ocfs2_info)))
965 return -EFAULT;
966
967 return ocfs2_info_handle(inode, &info, 0);
968 case FITRIM:
969 {
970 struct super_block *sb = inode->i_sb;
971 struct request_queue *q = bdev_get_queue(sb->s_bdev);
972 struct fstrim_range range;
973 int ret = 0;
974
975 if (!capable(CAP_SYS_ADMIN))
976 return -EPERM;
977
978 if (!blk_queue_discard(q))
979 return -EOPNOTSUPP;
980
981 if (copy_from_user(&range, argp, sizeof(range)))
982 return -EFAULT;
983
984 range.minlen = max_t(u64, q->limits.discard_granularity,
985 range.minlen);
986 ret = ocfs2_trim_fs(sb, &range);
987 if (ret < 0)
988 return ret;
989
990 if (copy_to_user(argp, &range, sizeof(range)))
991 return -EFAULT;
992
993 return 0;
994 }
995 case OCFS2_IOC_MOVE_EXT:
996 return ocfs2_ioctl_move_extents(filp, argp);
997 default:
998 return -ENOTTY;
999 }
1000}
1001
1002#ifdef CONFIG_COMPAT
1003long ocfs2_compat_ioctl(struct file *file, unsigned cmd, unsigned long arg)
1004{
1005 bool preserve;
1006 struct reflink_arguments args;
1007 struct inode *inode = file_inode(file);
1008 struct ocfs2_info info;
1009 void __user *argp = (void __user *)arg;
1010
1011 switch (cmd) {
1012 case OCFS2_IOC32_GETFLAGS:
1013 cmd = OCFS2_IOC_GETFLAGS;
1014 break;
1015 case OCFS2_IOC32_SETFLAGS:
1016 cmd = OCFS2_IOC_SETFLAGS;
1017 break;
1018 case OCFS2_IOC_RESVSP:
1019 case OCFS2_IOC_RESVSP64:
1020 case OCFS2_IOC_UNRESVSP:
1021 case OCFS2_IOC_UNRESVSP64:
1022 case OCFS2_IOC_GROUP_EXTEND:
1023 case OCFS2_IOC_GROUP_ADD:
1024 case OCFS2_IOC_GROUP_ADD64:
1025 case FITRIM:
1026 break;
1027 case OCFS2_IOC_REFLINK:
1028 if (copy_from_user(&args, argp, sizeof(args)))
1029 return -EFAULT;
1030 preserve = (args.preserve != 0);
1031
1032 return ocfs2_reflink_ioctl(inode, compat_ptr(args.old_path),
1033 compat_ptr(args.new_path), preserve);
1034 case OCFS2_IOC_INFO:
1035 if (copy_from_user(&info, argp, sizeof(struct ocfs2_info)))
1036 return -EFAULT;
1037
1038 return ocfs2_info_handle(inode, &info, 1);
1039 case OCFS2_IOC_MOVE_EXT:
1040 break;
1041 default:
1042 return -ENOIOCTLCMD;
1043 }
1044
1045 return ocfs2_ioctl(file, cmd, arg);
1046}
1047#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
14#include <cluster/masklog.h>
15
16#include "ocfs2.h"
17#include "alloc.h"
18#include "dlmglue.h"
19#include "file.h"
20#include "inode.h"
21#include "journal.h"
22
23#include "ocfs2_fs.h"
24#include "ioctl.h"
25#include "resize.h"
26#include "refcounttree.h"
27#include "sysfile.h"
28#include "dir.h"
29#include "buffer_head_io.h"
30#include "suballoc.h"
31#include "move_extents.h"
32
33#define o2info_from_user(a, b) \
34 copy_from_user(&(a), (b), sizeof(a))
35#define o2info_to_user(a, b) \
36 copy_to_user((typeof(a) __user *)b, &(a), sizeof(a))
37
38/*
39 * This is just a best-effort to tell userspace that this request
40 * caused the error.
41 */
42static inline void o2info_set_request_error(struct ocfs2_info_request *kreq,
43 struct ocfs2_info_request __user *req)
44{
45 kreq->ir_flags |= OCFS2_INFO_FL_ERROR;
46 (void)put_user(kreq->ir_flags, (__u32 __user *)&(req->ir_flags));
47}
48
49static inline void o2info_set_request_filled(struct ocfs2_info_request *req)
50{
51 req->ir_flags |= OCFS2_INFO_FL_FILLED;
52}
53
54static inline void o2info_clear_request_filled(struct ocfs2_info_request *req)
55{
56 req->ir_flags &= ~OCFS2_INFO_FL_FILLED;
57}
58
59static inline int o2info_coherent(struct ocfs2_info_request *req)
60{
61 return (!(req->ir_flags & OCFS2_INFO_FL_NON_COHERENT));
62}
63
64static int ocfs2_get_inode_attr(struct inode *inode, unsigned *flags)
65{
66 int status;
67
68 status = ocfs2_inode_lock(inode, NULL, 0);
69 if (status < 0) {
70 mlog_errno(status);
71 return status;
72 }
73 ocfs2_get_inode_flags(OCFS2_I(inode));
74 *flags = OCFS2_I(inode)->ip_attr;
75 ocfs2_inode_unlock(inode, 0);
76
77 return status;
78}
79
80static int ocfs2_set_inode_attr(struct inode *inode, unsigned flags,
81 unsigned mask)
82{
83 struct ocfs2_inode_info *ocfs2_inode = OCFS2_I(inode);
84 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
85 handle_t *handle = NULL;
86 struct buffer_head *bh = NULL;
87 unsigned oldflags;
88 int status;
89
90 inode_lock(inode);
91
92 status = ocfs2_inode_lock(inode, &bh, 1);
93 if (status < 0) {
94 mlog_errno(status);
95 goto bail;
96 }
97
98 status = -EACCES;
99 if (!inode_owner_or_capable(inode))
100 goto bail_unlock;
101
102 if (!S_ISDIR(inode->i_mode))
103 flags &= ~OCFS2_DIRSYNC_FL;
104
105 oldflags = ocfs2_inode->ip_attr;
106 flags = flags & mask;
107 flags |= oldflags & ~mask;
108
109 /*
110 * The IMMUTABLE and APPEND_ONLY flags can only be changed by
111 * the relevant capability.
112 */
113 status = -EPERM;
114 if ((oldflags & OCFS2_IMMUTABLE_FL) || ((flags ^ oldflags) &
115 (OCFS2_APPEND_FL | OCFS2_IMMUTABLE_FL))) {
116 if (!capable(CAP_LINUX_IMMUTABLE))
117 goto bail_unlock;
118 }
119
120 handle = ocfs2_start_trans(osb, OCFS2_INODE_UPDATE_CREDITS);
121 if (IS_ERR(handle)) {
122 status = PTR_ERR(handle);
123 mlog_errno(status);
124 goto bail_unlock;
125 }
126
127 ocfs2_inode->ip_attr = flags;
128 ocfs2_set_inode_flags(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 inode_unlock(inode);
140
141 brelse(bh);
142
143 return status;
144}
145
146static int ocfs2_info_handle_blocksize(struct inode *inode,
147 struct ocfs2_info_request __user *req)
148{
149 struct ocfs2_info_blocksize oib;
150
151 if (o2info_from_user(oib, req))
152 return -EFAULT;
153
154 oib.ib_blocksize = inode->i_sb->s_blocksize;
155
156 o2info_set_request_filled(&oib.ib_req);
157
158 if (o2info_to_user(oib, req))
159 return -EFAULT;
160
161 return 0;
162}
163
164static int ocfs2_info_handle_clustersize(struct inode *inode,
165 struct ocfs2_info_request __user *req)
166{
167 struct ocfs2_info_clustersize oic;
168 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
169
170 if (o2info_from_user(oic, req))
171 return -EFAULT;
172
173 oic.ic_clustersize = osb->s_clustersize;
174
175 o2info_set_request_filled(&oic.ic_req);
176
177 if (o2info_to_user(oic, req))
178 return -EFAULT;
179
180 return 0;
181}
182
183static int ocfs2_info_handle_maxslots(struct inode *inode,
184 struct ocfs2_info_request __user *req)
185{
186 struct ocfs2_info_maxslots oim;
187 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
188
189 if (o2info_from_user(oim, req))
190 return -EFAULT;
191
192 oim.im_max_slots = osb->max_slots;
193
194 o2info_set_request_filled(&oim.im_req);
195
196 if (o2info_to_user(oim, req))
197 return -EFAULT;
198
199 return 0;
200}
201
202static int ocfs2_info_handle_label(struct inode *inode,
203 struct ocfs2_info_request __user *req)
204{
205 struct ocfs2_info_label oil;
206 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
207
208 if (o2info_from_user(oil, req))
209 return -EFAULT;
210
211 memcpy(oil.il_label, osb->vol_label, OCFS2_MAX_VOL_LABEL_LEN);
212
213 o2info_set_request_filled(&oil.il_req);
214
215 if (o2info_to_user(oil, req))
216 return -EFAULT;
217
218 return 0;
219}
220
221static int ocfs2_info_handle_uuid(struct inode *inode,
222 struct ocfs2_info_request __user *req)
223{
224 struct ocfs2_info_uuid oiu;
225 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
226
227 if (o2info_from_user(oiu, req))
228 return -EFAULT;
229
230 memcpy(oiu.iu_uuid_str, osb->uuid_str, OCFS2_TEXT_UUID_LEN + 1);
231
232 o2info_set_request_filled(&oiu.iu_req);
233
234 if (o2info_to_user(oiu, req))
235 return -EFAULT;
236
237 return 0;
238}
239
240static int ocfs2_info_handle_fs_features(struct inode *inode,
241 struct ocfs2_info_request __user *req)
242{
243 struct ocfs2_info_fs_features oif;
244 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
245
246 if (o2info_from_user(oif, req))
247 return -EFAULT;
248
249 oif.if_compat_features = osb->s_feature_compat;
250 oif.if_incompat_features = osb->s_feature_incompat;
251 oif.if_ro_compat_features = osb->s_feature_ro_compat;
252
253 o2info_set_request_filled(&oif.if_req);
254
255 if (o2info_to_user(oif, req))
256 return -EFAULT;
257
258 return 0;
259}
260
261static int ocfs2_info_handle_journal_size(struct inode *inode,
262 struct ocfs2_info_request __user *req)
263{
264 struct ocfs2_info_journal_size oij;
265 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
266
267 if (o2info_from_user(oij, req))
268 return -EFAULT;
269
270 oij.ij_journal_size = i_size_read(osb->journal->j_inode);
271
272 o2info_set_request_filled(&oij.ij_req);
273
274 if (o2info_to_user(oij, req))
275 return -EFAULT;
276
277 return 0;
278}
279
280static int ocfs2_info_scan_inode_alloc(struct ocfs2_super *osb,
281 struct inode *inode_alloc, u64 blkno,
282 struct ocfs2_info_freeinode *fi,
283 u32 slot)
284{
285 int status = 0, unlock = 0;
286
287 struct buffer_head *bh = NULL;
288 struct ocfs2_dinode *dinode_alloc = NULL;
289
290 if (inode_alloc)
291 inode_lock(inode_alloc);
292
293 if (o2info_coherent(&fi->ifi_req)) {
294 status = ocfs2_inode_lock(inode_alloc, &bh, 0);
295 if (status < 0) {
296 mlog_errno(status);
297 goto bail;
298 }
299 unlock = 1;
300 } else {
301 status = ocfs2_read_blocks_sync(osb, blkno, 1, &bh);
302 if (status < 0) {
303 mlog_errno(status);
304 goto bail;
305 }
306 }
307
308 dinode_alloc = (struct ocfs2_dinode *)bh->b_data;
309
310 fi->ifi_stat[slot].lfi_total =
311 le32_to_cpu(dinode_alloc->id1.bitmap1.i_total);
312 fi->ifi_stat[slot].lfi_free =
313 le32_to_cpu(dinode_alloc->id1.bitmap1.i_total) -
314 le32_to_cpu(dinode_alloc->id1.bitmap1.i_used);
315
316bail:
317 if (unlock)
318 ocfs2_inode_unlock(inode_alloc, 0);
319
320 if (inode_alloc)
321 inode_unlock(inode_alloc);
322
323 brelse(bh);
324
325 return status;
326}
327
328static int ocfs2_info_handle_freeinode(struct inode *inode,
329 struct ocfs2_info_request __user *req)
330{
331 u32 i;
332 u64 blkno = -1;
333 char namebuf[40];
334 int status, type = INODE_ALLOC_SYSTEM_INODE;
335 struct ocfs2_info_freeinode *oifi = NULL;
336 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
337 struct inode *inode_alloc = NULL;
338
339 oifi = kzalloc(sizeof(struct ocfs2_info_freeinode), GFP_KERNEL);
340 if (!oifi) {
341 status = -ENOMEM;
342 mlog_errno(status);
343 goto out_err;
344 }
345
346 if (o2info_from_user(*oifi, req)) {
347 status = -EFAULT;
348 goto out_free;
349 }
350
351 oifi->ifi_slotnum = osb->max_slots;
352
353 for (i = 0; i < oifi->ifi_slotnum; i++) {
354 if (o2info_coherent(&oifi->ifi_req)) {
355 inode_alloc = ocfs2_get_system_file_inode(osb, type, i);
356 if (!inode_alloc) {
357 mlog(ML_ERROR, "unable to get alloc inode in "
358 "slot %u\n", i);
359 status = -EIO;
360 goto bail;
361 }
362 } else {
363 ocfs2_sprintf_system_inode_name(namebuf,
364 sizeof(namebuf),
365 type, i);
366 status = ocfs2_lookup_ino_from_name(osb->sys_root_inode,
367 namebuf,
368 strlen(namebuf),
369 &blkno);
370 if (status < 0) {
371 status = -ENOENT;
372 goto bail;
373 }
374 }
375
376 status = ocfs2_info_scan_inode_alloc(osb, inode_alloc, blkno, oifi, i);
377
378 iput(inode_alloc);
379 inode_alloc = NULL;
380
381 if (status < 0)
382 goto bail;
383 }
384
385 o2info_set_request_filled(&oifi->ifi_req);
386
387 if (o2info_to_user(*oifi, req)) {
388 status = -EFAULT;
389 goto out_free;
390 }
391
392 status = 0;
393bail:
394 if (status)
395 o2info_set_request_error(&oifi->ifi_req, req);
396out_free:
397 kfree(oifi);
398out_err:
399 return status;
400}
401
402static void o2ffg_update_histogram(struct ocfs2_info_free_chunk_list *hist,
403 unsigned int chunksize)
404{
405 int index;
406
407 index = __ilog2_u32(chunksize);
408 if (index >= OCFS2_INFO_MAX_HIST)
409 index = OCFS2_INFO_MAX_HIST - 1;
410
411 hist->fc_chunks[index]++;
412 hist->fc_clusters[index] += chunksize;
413}
414
415static void o2ffg_update_stats(struct ocfs2_info_freefrag_stats *stats,
416 unsigned int chunksize)
417{
418 if (chunksize > stats->ffs_max)
419 stats->ffs_max = chunksize;
420
421 if (chunksize < stats->ffs_min)
422 stats->ffs_min = chunksize;
423
424 stats->ffs_avg += chunksize;
425 stats->ffs_free_chunks_real++;
426}
427
428static void ocfs2_info_update_ffg(struct ocfs2_info_freefrag *ffg,
429 unsigned int chunksize)
430{
431 o2ffg_update_histogram(&(ffg->iff_ffs.ffs_fc_hist), chunksize);
432 o2ffg_update_stats(&(ffg->iff_ffs), chunksize);
433}
434
435static int ocfs2_info_freefrag_scan_chain(struct ocfs2_super *osb,
436 struct inode *gb_inode,
437 struct ocfs2_dinode *gb_dinode,
438 struct ocfs2_chain_rec *rec,
439 struct ocfs2_info_freefrag *ffg,
440 u32 chunks_in_group)
441{
442 int status = 0, used;
443 u64 blkno;
444
445 struct buffer_head *bh = NULL;
446 struct ocfs2_group_desc *bg = NULL;
447
448 unsigned int max_bits, num_clusters;
449 unsigned int offset = 0, cluster, chunk;
450 unsigned int chunk_free, last_chunksize = 0;
451
452 if (!le32_to_cpu(rec->c_free))
453 goto bail;
454
455 do {
456 if (!bg)
457 blkno = le64_to_cpu(rec->c_blkno);
458 else
459 blkno = le64_to_cpu(bg->bg_next_group);
460
461 if (bh) {
462 brelse(bh);
463 bh = NULL;
464 }
465
466 if (o2info_coherent(&ffg->iff_req))
467 status = ocfs2_read_group_descriptor(gb_inode,
468 gb_dinode,
469 blkno, &bh);
470 else
471 status = ocfs2_read_blocks_sync(osb, blkno, 1, &bh);
472
473 if (status < 0) {
474 mlog(ML_ERROR, "Can't read the group descriptor # "
475 "%llu from device.", (unsigned long long)blkno);
476 status = -EIO;
477 goto bail;
478 }
479
480 bg = (struct ocfs2_group_desc *)bh->b_data;
481
482 if (!le16_to_cpu(bg->bg_free_bits_count))
483 continue;
484
485 max_bits = le16_to_cpu(bg->bg_bits);
486 offset = 0;
487
488 for (chunk = 0; chunk < chunks_in_group; chunk++) {
489 /*
490 * last chunk may be not an entire one.
491 */
492 if ((offset + ffg->iff_chunksize) > max_bits)
493 num_clusters = max_bits - offset;
494 else
495 num_clusters = ffg->iff_chunksize;
496
497 chunk_free = 0;
498 for (cluster = 0; cluster < num_clusters; cluster++) {
499 used = ocfs2_test_bit(offset,
500 (unsigned long *)bg->bg_bitmap);
501 /*
502 * - chunk_free counts free clusters in #N chunk.
503 * - last_chunksize records the size(in) clusters
504 * for the last real free chunk being counted.
505 */
506 if (!used) {
507 last_chunksize++;
508 chunk_free++;
509 }
510
511 if (used && last_chunksize) {
512 ocfs2_info_update_ffg(ffg,
513 last_chunksize);
514 last_chunksize = 0;
515 }
516
517 offset++;
518 }
519
520 if (chunk_free == ffg->iff_chunksize)
521 ffg->iff_ffs.ffs_free_chunks++;
522 }
523
524 /*
525 * need to update the info for last free chunk.
526 */
527 if (last_chunksize)
528 ocfs2_info_update_ffg(ffg, last_chunksize);
529
530 } while (le64_to_cpu(bg->bg_next_group));
531
532bail:
533 brelse(bh);
534
535 return status;
536}
537
538static int ocfs2_info_freefrag_scan_bitmap(struct ocfs2_super *osb,
539 struct inode *gb_inode, u64 blkno,
540 struct ocfs2_info_freefrag *ffg)
541{
542 u32 chunks_in_group;
543 int status = 0, unlock = 0, i;
544
545 struct buffer_head *bh = NULL;
546 struct ocfs2_chain_list *cl = NULL;
547 struct ocfs2_chain_rec *rec = NULL;
548 struct ocfs2_dinode *gb_dinode = NULL;
549
550 if (gb_inode)
551 inode_lock(gb_inode);
552
553 if (o2info_coherent(&ffg->iff_req)) {
554 status = ocfs2_inode_lock(gb_inode, &bh, 0);
555 if (status < 0) {
556 mlog_errno(status);
557 goto bail;
558 }
559 unlock = 1;
560 } else {
561 status = ocfs2_read_blocks_sync(osb, blkno, 1, &bh);
562 if (status < 0) {
563 mlog_errno(status);
564 goto bail;
565 }
566 }
567
568 gb_dinode = (struct ocfs2_dinode *)bh->b_data;
569 cl = &(gb_dinode->id2.i_chain);
570
571 /*
572 * Chunksize(in) clusters from userspace should be
573 * less than clusters in a group.
574 */
575 if (ffg->iff_chunksize > le16_to_cpu(cl->cl_cpg)) {
576 status = -EINVAL;
577 goto bail;
578 }
579
580 memset(&ffg->iff_ffs, 0, sizeof(struct ocfs2_info_freefrag_stats));
581
582 ffg->iff_ffs.ffs_min = ~0U;
583 ffg->iff_ffs.ffs_clusters =
584 le32_to_cpu(gb_dinode->id1.bitmap1.i_total);
585 ffg->iff_ffs.ffs_free_clusters = ffg->iff_ffs.ffs_clusters -
586 le32_to_cpu(gb_dinode->id1.bitmap1.i_used);
587
588 chunks_in_group = le16_to_cpu(cl->cl_cpg) / ffg->iff_chunksize + 1;
589
590 for (i = 0; i < le16_to_cpu(cl->cl_next_free_rec); i++) {
591 rec = &(cl->cl_recs[i]);
592 status = ocfs2_info_freefrag_scan_chain(osb, gb_inode,
593 gb_dinode,
594 rec, ffg,
595 chunks_in_group);
596 if (status)
597 goto bail;
598 }
599
600 if (ffg->iff_ffs.ffs_free_chunks_real)
601 ffg->iff_ffs.ffs_avg = (ffg->iff_ffs.ffs_avg /
602 ffg->iff_ffs.ffs_free_chunks_real);
603bail:
604 if (unlock)
605 ocfs2_inode_unlock(gb_inode, 0);
606
607 if (gb_inode)
608 inode_unlock(gb_inode);
609
610 iput(gb_inode);
611 brelse(bh);
612
613 return status;
614}
615
616static int ocfs2_info_handle_freefrag(struct inode *inode,
617 struct ocfs2_info_request __user *req)
618{
619 u64 blkno = -1;
620 char namebuf[40];
621 int status, type = GLOBAL_BITMAP_SYSTEM_INODE;
622
623 struct ocfs2_info_freefrag *oiff;
624 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
625 struct inode *gb_inode = NULL;
626
627 oiff = kzalloc(sizeof(struct ocfs2_info_freefrag), GFP_KERNEL);
628 if (!oiff) {
629 status = -ENOMEM;
630 mlog_errno(status);
631 goto out_err;
632 }
633
634 if (o2info_from_user(*oiff, req)) {
635 status = -EFAULT;
636 goto out_free;
637 }
638 /*
639 * chunksize from userspace should be power of 2.
640 */
641 if ((oiff->iff_chunksize & (oiff->iff_chunksize - 1)) ||
642 (!oiff->iff_chunksize)) {
643 status = -EINVAL;
644 goto bail;
645 }
646
647 if (o2info_coherent(&oiff->iff_req)) {
648 gb_inode = ocfs2_get_system_file_inode(osb, type,
649 OCFS2_INVALID_SLOT);
650 if (!gb_inode) {
651 mlog(ML_ERROR, "unable to get global_bitmap inode\n");
652 status = -EIO;
653 goto bail;
654 }
655 } else {
656 ocfs2_sprintf_system_inode_name(namebuf, sizeof(namebuf), type,
657 OCFS2_INVALID_SLOT);
658 status = ocfs2_lookup_ino_from_name(osb->sys_root_inode,
659 namebuf,
660 strlen(namebuf),
661 &blkno);
662 if (status < 0) {
663 status = -ENOENT;
664 goto bail;
665 }
666 }
667
668 status = ocfs2_info_freefrag_scan_bitmap(osb, gb_inode, blkno, oiff);
669 if (status < 0)
670 goto bail;
671
672 o2info_set_request_filled(&oiff->iff_req);
673
674 if (o2info_to_user(*oiff, req)) {
675 status = -EFAULT;
676 goto out_free;
677 }
678
679 status = 0;
680bail:
681 if (status)
682 o2info_set_request_error(&oiff->iff_req, req);
683out_free:
684 kfree(oiff);
685out_err:
686 return status;
687}
688
689static int ocfs2_info_handle_unknown(struct inode *inode,
690 struct ocfs2_info_request __user *req)
691{
692 struct ocfs2_info_request oir;
693
694 if (o2info_from_user(oir, req))
695 return -EFAULT;
696
697 o2info_clear_request_filled(&oir);
698
699 if (o2info_to_user(oir, req))
700 return -EFAULT;
701
702 return 0;
703}
704
705/*
706 * Validate and distinguish OCFS2_IOC_INFO requests.
707 *
708 * - validate the magic number.
709 * - distinguish different requests.
710 * - validate size of different requests.
711 */
712static int ocfs2_info_handle_request(struct inode *inode,
713 struct ocfs2_info_request __user *req)
714{
715 int status = -EFAULT;
716 struct ocfs2_info_request oir;
717
718 if (o2info_from_user(oir, req))
719 goto bail;
720
721 status = -EINVAL;
722 if (oir.ir_magic != OCFS2_INFO_MAGIC)
723 goto bail;
724
725 switch (oir.ir_code) {
726 case OCFS2_INFO_BLOCKSIZE:
727 if (oir.ir_size == sizeof(struct ocfs2_info_blocksize))
728 status = ocfs2_info_handle_blocksize(inode, req);
729 break;
730 case OCFS2_INFO_CLUSTERSIZE:
731 if (oir.ir_size == sizeof(struct ocfs2_info_clustersize))
732 status = ocfs2_info_handle_clustersize(inode, req);
733 break;
734 case OCFS2_INFO_MAXSLOTS:
735 if (oir.ir_size == sizeof(struct ocfs2_info_maxslots))
736 status = ocfs2_info_handle_maxslots(inode, req);
737 break;
738 case OCFS2_INFO_LABEL:
739 if (oir.ir_size == sizeof(struct ocfs2_info_label))
740 status = ocfs2_info_handle_label(inode, req);
741 break;
742 case OCFS2_INFO_UUID:
743 if (oir.ir_size == sizeof(struct ocfs2_info_uuid))
744 status = ocfs2_info_handle_uuid(inode, req);
745 break;
746 case OCFS2_INFO_FS_FEATURES:
747 if (oir.ir_size == sizeof(struct ocfs2_info_fs_features))
748 status = ocfs2_info_handle_fs_features(inode, req);
749 break;
750 case OCFS2_INFO_JOURNAL_SIZE:
751 if (oir.ir_size == sizeof(struct ocfs2_info_journal_size))
752 status = ocfs2_info_handle_journal_size(inode, req);
753 break;
754 case OCFS2_INFO_FREEINODE:
755 if (oir.ir_size == sizeof(struct ocfs2_info_freeinode))
756 status = ocfs2_info_handle_freeinode(inode, req);
757 break;
758 case OCFS2_INFO_FREEFRAG:
759 if (oir.ir_size == sizeof(struct ocfs2_info_freefrag))
760 status = ocfs2_info_handle_freefrag(inode, req);
761 break;
762 default:
763 status = ocfs2_info_handle_unknown(inode, req);
764 break;
765 }
766
767bail:
768 return status;
769}
770
771static int ocfs2_get_request_ptr(struct ocfs2_info *info, int idx,
772 u64 *req_addr, int compat_flag)
773{
774 int status = -EFAULT;
775 u64 __user *bp = NULL;
776
777 if (compat_flag) {
778#ifdef CONFIG_COMPAT
779 /*
780 * pointer bp stores the base address of a pointers array,
781 * which collects all addresses of separate request.
782 */
783 bp = (u64 __user *)(unsigned long)compat_ptr(info->oi_requests);
784#else
785 BUG();
786#endif
787 } else
788 bp = (u64 __user *)(unsigned long)(info->oi_requests);
789
790 if (o2info_from_user(*req_addr, bp + idx))
791 goto bail;
792
793 status = 0;
794bail:
795 return status;
796}
797
798/*
799 * OCFS2_IOC_INFO handles an array of requests passed from userspace.
800 *
801 * ocfs2_info_handle() recevies a large info aggregation, grab and
802 * validate the request count from header, then break it into small
803 * pieces, later specific handlers can handle them one by one.
804 *
805 * Idea here is to make each separate request small enough to ensure
806 * a better backward&forward compatibility, since a small piece of
807 * request will be less likely to be broken if disk layout get changed.
808 */
809static int ocfs2_info_handle(struct inode *inode, struct ocfs2_info *info,
810 int compat_flag)
811{
812 int i, status = 0;
813 u64 req_addr;
814 struct ocfs2_info_request __user *reqp;
815
816 if ((info->oi_count > OCFS2_INFO_MAX_REQUEST) ||
817 (!info->oi_requests)) {
818 status = -EINVAL;
819 goto bail;
820 }
821
822 for (i = 0; i < info->oi_count; i++) {
823
824 status = ocfs2_get_request_ptr(info, i, &req_addr, compat_flag);
825 if (status)
826 break;
827
828 reqp = (struct ocfs2_info_request __user *)(unsigned long)req_addr;
829 if (!reqp) {
830 status = -EINVAL;
831 goto bail;
832 }
833
834 status = ocfs2_info_handle_request(inode, reqp);
835 if (status)
836 break;
837 }
838
839bail:
840 return status;
841}
842
843long ocfs2_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
844{
845 struct inode *inode = file_inode(filp);
846 unsigned int flags;
847 int new_clusters;
848 int status;
849 struct ocfs2_space_resv sr;
850 struct ocfs2_new_group_input input;
851 struct reflink_arguments args;
852 const char __user *old_path;
853 const char __user *new_path;
854 bool preserve;
855 struct ocfs2_info info;
856 void __user *argp = (void __user *)arg;
857
858 switch (cmd) {
859 case OCFS2_IOC_GETFLAGS:
860 status = ocfs2_get_inode_attr(inode, &flags);
861 if (status < 0)
862 return status;
863
864 flags &= OCFS2_FL_VISIBLE;
865 return put_user(flags, (int __user *) arg);
866 case OCFS2_IOC_SETFLAGS:
867 if (get_user(flags, (int __user *) arg))
868 return -EFAULT;
869
870 status = mnt_want_write_file(filp);
871 if (status)
872 return status;
873 status = ocfs2_set_inode_attr(inode, flags,
874 OCFS2_FL_MODIFIABLE);
875 mnt_drop_write_file(filp);
876 return status;
877 case OCFS2_IOC_RESVSP:
878 case OCFS2_IOC_RESVSP64:
879 case OCFS2_IOC_UNRESVSP:
880 case OCFS2_IOC_UNRESVSP64:
881 if (copy_from_user(&sr, (int __user *) arg, sizeof(sr)))
882 return -EFAULT;
883
884 return ocfs2_change_file_space(filp, cmd, &sr);
885 case OCFS2_IOC_GROUP_EXTEND:
886 if (!capable(CAP_SYS_RESOURCE))
887 return -EPERM;
888
889 if (get_user(new_clusters, (int __user *)arg))
890 return -EFAULT;
891
892 status = mnt_want_write_file(filp);
893 if (status)
894 return status;
895 status = ocfs2_group_extend(inode, new_clusters);
896 mnt_drop_write_file(filp);
897 return status;
898 case OCFS2_IOC_GROUP_ADD:
899 case OCFS2_IOC_GROUP_ADD64:
900 if (!capable(CAP_SYS_RESOURCE))
901 return -EPERM;
902
903 if (copy_from_user(&input, (int __user *) arg, sizeof(input)))
904 return -EFAULT;
905
906 status = mnt_want_write_file(filp);
907 if (status)
908 return status;
909 status = ocfs2_group_add(inode, &input);
910 mnt_drop_write_file(filp);
911 return status;
912 case OCFS2_IOC_REFLINK:
913 if (copy_from_user(&args, argp, sizeof(args)))
914 return -EFAULT;
915 old_path = (const char __user *)(unsigned long)args.old_path;
916 new_path = (const char __user *)(unsigned long)args.new_path;
917 preserve = (args.preserve != 0);
918
919 return ocfs2_reflink_ioctl(inode, old_path, new_path, preserve);
920 case OCFS2_IOC_INFO:
921 if (copy_from_user(&info, argp, sizeof(struct ocfs2_info)))
922 return -EFAULT;
923
924 return ocfs2_info_handle(inode, &info, 0);
925 case FITRIM:
926 {
927 struct super_block *sb = inode->i_sb;
928 struct request_queue *q = bdev_get_queue(sb->s_bdev);
929 struct fstrim_range range;
930 int ret = 0;
931
932 if (!capable(CAP_SYS_ADMIN))
933 return -EPERM;
934
935 if (!blk_queue_discard(q))
936 return -EOPNOTSUPP;
937
938 if (copy_from_user(&range, argp, sizeof(range)))
939 return -EFAULT;
940
941 range.minlen = max_t(u64, q->limits.discard_granularity,
942 range.minlen);
943 ret = ocfs2_trim_fs(sb, &range);
944 if (ret < 0)
945 return ret;
946
947 if (copy_to_user(argp, &range, sizeof(range)))
948 return -EFAULT;
949
950 return 0;
951 }
952 case OCFS2_IOC_MOVE_EXT:
953 return ocfs2_ioctl_move_extents(filp, argp);
954 default:
955 return -ENOTTY;
956 }
957}
958
959#ifdef CONFIG_COMPAT
960long ocfs2_compat_ioctl(struct file *file, unsigned cmd, unsigned long arg)
961{
962 bool preserve;
963 struct reflink_arguments args;
964 struct inode *inode = file_inode(file);
965 struct ocfs2_info info;
966 void __user *argp = (void __user *)arg;
967
968 switch (cmd) {
969 case OCFS2_IOC32_GETFLAGS:
970 cmd = OCFS2_IOC_GETFLAGS;
971 break;
972 case OCFS2_IOC32_SETFLAGS:
973 cmd = OCFS2_IOC_SETFLAGS;
974 break;
975 case OCFS2_IOC_RESVSP:
976 case OCFS2_IOC_RESVSP64:
977 case OCFS2_IOC_UNRESVSP:
978 case OCFS2_IOC_UNRESVSP64:
979 case OCFS2_IOC_GROUP_EXTEND:
980 case OCFS2_IOC_GROUP_ADD:
981 case OCFS2_IOC_GROUP_ADD64:
982 break;
983 case OCFS2_IOC_REFLINK:
984 if (copy_from_user(&args, argp, sizeof(args)))
985 return -EFAULT;
986 preserve = (args.preserve != 0);
987
988 return ocfs2_reflink_ioctl(inode, compat_ptr(args.old_path),
989 compat_ptr(args.new_path), preserve);
990 case OCFS2_IOC_INFO:
991 if (copy_from_user(&info, argp, sizeof(struct ocfs2_info)))
992 return -EFAULT;
993
994 return ocfs2_info_handle(inode, &info, 1);
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