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/compat.h>
11
12#include <cluster/masklog.h>
13
14#include "ocfs2.h"
15#include "alloc.h"
16#include "dlmglue.h"
17#include "file.h"
18#include "inode.h"
19#include "journal.h"
20
21#include "ocfs2_fs.h"
22#include "ioctl.h"
23#include "resize.h"
24#include "refcounttree.h"
25#include "sysfile.h"
26#include "dir.h"
27#include "buffer_head_io.h"
28#include "suballoc.h"
29#include "move_extents.h"
30
31#include <linux/ext2_fs.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 call is void because we are already reporting an error that may
40 * be -EFAULT. The error will be returned from the ioctl(2) call. It's
41 * just a best-effort to tell userspace that this request 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
65static int ocfs2_get_inode_attr(struct inode *inode, unsigned *flags)
66{
67 int status;
68
69 status = ocfs2_inode_lock(inode, NULL, 0);
70 if (status < 0) {
71 mlog_errno(status);
72 return status;
73 }
74 ocfs2_get_inode_flags(OCFS2_I(inode));
75 *flags = OCFS2_I(inode)->ip_attr;
76 ocfs2_inode_unlock(inode, 0);
77
78 return status;
79}
80
81static int ocfs2_set_inode_attr(struct inode *inode, unsigned flags,
82 unsigned mask)
83{
84 struct ocfs2_inode_info *ocfs2_inode = OCFS2_I(inode);
85 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
86 handle_t *handle = NULL;
87 struct buffer_head *bh = NULL;
88 unsigned oldflags;
89 int status;
90
91 mutex_lock(&inode->i_mutex);
92
93 status = ocfs2_inode_lock(inode, &bh, 1);
94 if (status < 0) {
95 mlog_errno(status);
96 goto bail;
97 }
98
99 status = -EACCES;
100 if (!inode_owner_or_capable(inode))
101 goto bail_unlock;
102
103 if (!S_ISDIR(inode->i_mode))
104 flags &= ~OCFS2_DIRSYNC_FL;
105
106 handle = ocfs2_start_trans(osb, OCFS2_INODE_UPDATE_CREDITS);
107 if (IS_ERR(handle)) {
108 status = PTR_ERR(handle);
109 mlog_errno(status);
110 goto bail_unlock;
111 }
112
113 oldflags = ocfs2_inode->ip_attr;
114 flags = flags & mask;
115 flags |= oldflags & ~mask;
116
117 /*
118 * The IMMUTABLE and APPEND_ONLY flags can only be changed by
119 * the relevant capability.
120 */
121 status = -EPERM;
122 if ((oldflags & OCFS2_IMMUTABLE_FL) || ((flags ^ oldflags) &
123 (OCFS2_APPEND_FL | OCFS2_IMMUTABLE_FL))) {
124 if (!capable(CAP_LINUX_IMMUTABLE))
125 goto bail_unlock;
126 }
127
128 ocfs2_inode->ip_attr = flags;
129 ocfs2_set_inode_flags(inode);
130
131 status = ocfs2_mark_inode_dirty(handle, inode, bh);
132 if (status < 0)
133 mlog_errno(status);
134
135 ocfs2_commit_trans(osb, handle);
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 = osb->journal->j_inode->i_size;
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 bail;
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 if (status < 0)
417 goto bail;
418
419 iput(inode_alloc);
420 inode_alloc = NULL;
421 }
422
423 o2info_set_request_filled(&oifi->ifi_req);
424
425 if (o2info_to_user(*oifi, req))
426 goto bail;
427
428 status = 0;
429bail:
430 if (status)
431 o2info_set_request_error(&oifi->ifi_req, req);
432
433 kfree(oifi);
434
435 return status;
436}
437
438static void o2ffg_update_histogram(struct ocfs2_info_free_chunk_list *hist,
439 unsigned int chunksize)
440{
441 int index;
442
443 index = __ilog2_u32(chunksize);
444 if (index >= OCFS2_INFO_MAX_HIST)
445 index = OCFS2_INFO_MAX_HIST - 1;
446
447 hist->fc_chunks[index]++;
448 hist->fc_clusters[index] += chunksize;
449}
450
451static void o2ffg_update_stats(struct ocfs2_info_freefrag_stats *stats,
452 unsigned int chunksize)
453{
454 if (chunksize > stats->ffs_max)
455 stats->ffs_max = chunksize;
456
457 if (chunksize < stats->ffs_min)
458 stats->ffs_min = chunksize;
459
460 stats->ffs_avg += chunksize;
461 stats->ffs_free_chunks_real++;
462}
463
464void ocfs2_info_update_ffg(struct ocfs2_info_freefrag *ffg,
465 unsigned int chunksize)
466{
467 o2ffg_update_histogram(&(ffg->iff_ffs.ffs_fc_hist), chunksize);
468 o2ffg_update_stats(&(ffg->iff_ffs), chunksize);
469}
470
471int ocfs2_info_freefrag_scan_chain(struct ocfs2_super *osb,
472 struct inode *gb_inode,
473 struct ocfs2_dinode *gb_dinode,
474 struct ocfs2_chain_rec *rec,
475 struct ocfs2_info_freefrag *ffg,
476 u32 chunks_in_group)
477{
478 int status = 0, used;
479 u64 blkno;
480
481 struct buffer_head *bh = NULL;
482 struct ocfs2_group_desc *bg = NULL;
483
484 unsigned int max_bits, num_clusters;
485 unsigned int offset = 0, cluster, chunk;
486 unsigned int chunk_free, last_chunksize = 0;
487
488 if (!le32_to_cpu(rec->c_free))
489 goto bail;
490
491 do {
492 if (!bg)
493 blkno = le64_to_cpu(rec->c_blkno);
494 else
495 blkno = le64_to_cpu(bg->bg_next_group);
496
497 if (bh) {
498 brelse(bh);
499 bh = NULL;
500 }
501
502 if (o2info_coherent(&ffg->iff_req))
503 status = ocfs2_read_group_descriptor(gb_inode,
504 gb_dinode,
505 blkno, &bh);
506 else
507 status = ocfs2_read_blocks_sync(osb, blkno, 1, &bh);
508
509 if (status < 0) {
510 mlog(ML_ERROR, "Can't read the group descriptor # "
511 "%llu from device.", (unsigned long long)blkno);
512 status = -EIO;
513 goto bail;
514 }
515
516 bg = (struct ocfs2_group_desc *)bh->b_data;
517
518 if (!le16_to_cpu(bg->bg_free_bits_count))
519 continue;
520
521 max_bits = le16_to_cpu(bg->bg_bits);
522 offset = 0;
523
524 for (chunk = 0; chunk < chunks_in_group; chunk++) {
525 /*
526 * last chunk may be not an entire one.
527 */
528 if ((offset + ffg->iff_chunksize) > max_bits)
529 num_clusters = max_bits - offset;
530 else
531 num_clusters = ffg->iff_chunksize;
532
533 chunk_free = 0;
534 for (cluster = 0; cluster < num_clusters; cluster++) {
535 used = ocfs2_test_bit(offset,
536 (unsigned long *)bg->bg_bitmap);
537 /*
538 * - chunk_free counts free clusters in #N chunk.
539 * - last_chunksize records the size(in) clusters
540 * for the last real free chunk being counted.
541 */
542 if (!used) {
543 last_chunksize++;
544 chunk_free++;
545 }
546
547 if (used && last_chunksize) {
548 ocfs2_info_update_ffg(ffg,
549 last_chunksize);
550 last_chunksize = 0;
551 }
552
553 offset++;
554 }
555
556 if (chunk_free == ffg->iff_chunksize)
557 ffg->iff_ffs.ffs_free_chunks++;
558 }
559
560 /*
561 * need to update the info for last free chunk.
562 */
563 if (last_chunksize)
564 ocfs2_info_update_ffg(ffg, last_chunksize);
565
566 } while (le64_to_cpu(bg->bg_next_group));
567
568bail:
569 brelse(bh);
570
571 return status;
572}
573
574int ocfs2_info_freefrag_scan_bitmap(struct ocfs2_super *osb,
575 struct inode *gb_inode, u64 blkno,
576 struct ocfs2_info_freefrag *ffg)
577{
578 u32 chunks_in_group;
579 int status = 0, unlock = 0, i;
580
581 struct buffer_head *bh = NULL;
582 struct ocfs2_chain_list *cl = NULL;
583 struct ocfs2_chain_rec *rec = NULL;
584 struct ocfs2_dinode *gb_dinode = NULL;
585
586 if (gb_inode)
587 mutex_lock(&gb_inode->i_mutex);
588
589 if (o2info_coherent(&ffg->iff_req)) {
590 status = ocfs2_inode_lock(gb_inode, &bh, 0);
591 if (status < 0) {
592 mlog_errno(status);
593 goto bail;
594 }
595 unlock = 1;
596 } else {
597 status = ocfs2_read_blocks_sync(osb, blkno, 1, &bh);
598 if (status < 0) {
599 mlog_errno(status);
600 goto bail;
601 }
602 }
603
604 gb_dinode = (struct ocfs2_dinode *)bh->b_data;
605 cl = &(gb_dinode->id2.i_chain);
606
607 /*
608 * Chunksize(in) clusters from userspace should be
609 * less than clusters in a group.
610 */
611 if (ffg->iff_chunksize > le16_to_cpu(cl->cl_cpg)) {
612 status = -EINVAL;
613 goto bail;
614 }
615
616 memset(&ffg->iff_ffs, 0, sizeof(struct ocfs2_info_freefrag_stats));
617
618 ffg->iff_ffs.ffs_min = ~0U;
619 ffg->iff_ffs.ffs_clusters =
620 le32_to_cpu(gb_dinode->id1.bitmap1.i_total);
621 ffg->iff_ffs.ffs_free_clusters = ffg->iff_ffs.ffs_clusters -
622 le32_to_cpu(gb_dinode->id1.bitmap1.i_used);
623
624 chunks_in_group = le16_to_cpu(cl->cl_cpg) / ffg->iff_chunksize + 1;
625
626 for (i = 0; i < le16_to_cpu(cl->cl_next_free_rec); i++) {
627 rec = &(cl->cl_recs[i]);
628 status = ocfs2_info_freefrag_scan_chain(osb, gb_inode,
629 gb_dinode,
630 rec, ffg,
631 chunks_in_group);
632 if (status)
633 goto bail;
634 }
635
636 if (ffg->iff_ffs.ffs_free_chunks_real)
637 ffg->iff_ffs.ffs_avg = (ffg->iff_ffs.ffs_avg /
638 ffg->iff_ffs.ffs_free_chunks_real);
639bail:
640 if (unlock)
641 ocfs2_inode_unlock(gb_inode, 0);
642
643 if (gb_inode)
644 mutex_unlock(&gb_inode->i_mutex);
645
646 if (gb_inode)
647 iput(gb_inode);
648
649 brelse(bh);
650
651 return status;
652}
653
654int ocfs2_info_handle_freefrag(struct inode *inode,
655 struct ocfs2_info_request __user *req)
656{
657 u64 blkno = -1;
658 char namebuf[40];
659 int status = -EFAULT, type = GLOBAL_BITMAP_SYSTEM_INODE;
660
661 struct ocfs2_info_freefrag *oiff;
662 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
663 struct inode *gb_inode = NULL;
664
665 oiff = kzalloc(sizeof(struct ocfs2_info_freefrag), GFP_KERNEL);
666 if (!oiff) {
667 status = -ENOMEM;
668 mlog_errno(status);
669 goto bail;
670 }
671
672 if (o2info_from_user(*oiff, req))
673 goto bail;
674 /*
675 * chunksize from userspace should be power of 2.
676 */
677 if ((oiff->iff_chunksize & (oiff->iff_chunksize - 1)) ||
678 (!oiff->iff_chunksize)) {
679 status = -EINVAL;
680 goto bail;
681 }
682
683 if (o2info_coherent(&oiff->iff_req)) {
684 gb_inode = ocfs2_get_system_file_inode(osb, type,
685 OCFS2_INVALID_SLOT);
686 if (!gb_inode) {
687 mlog(ML_ERROR, "unable to get global_bitmap inode\n");
688 status = -EIO;
689 goto bail;
690 }
691 } else {
692 ocfs2_sprintf_system_inode_name(namebuf, sizeof(namebuf), type,
693 OCFS2_INVALID_SLOT);
694 status = ocfs2_lookup_ino_from_name(osb->sys_root_inode,
695 namebuf,
696 strlen(namebuf),
697 &blkno);
698 if (status < 0) {
699 status = -ENOENT;
700 goto bail;
701 }
702 }
703
704 status = ocfs2_info_freefrag_scan_bitmap(osb, gb_inode, blkno, oiff);
705 if (status < 0)
706 goto bail;
707
708 o2info_set_request_filled(&oiff->iff_req);
709
710 if (o2info_to_user(*oiff, req))
711 goto bail;
712
713 status = 0;
714bail:
715 if (status)
716 o2info_set_request_error(&oiff->iff_req, req);
717
718 kfree(oiff);
719
720 return status;
721}
722
723int ocfs2_info_handle_unknown(struct inode *inode,
724 struct ocfs2_info_request __user *req)
725{
726 int status = -EFAULT;
727 struct ocfs2_info_request oir;
728
729 if (o2info_from_user(oir, req))
730 goto bail;
731
732 o2info_clear_request_filled(&oir);
733
734 if (o2info_to_user(oir, req))
735 goto bail;
736
737 status = 0;
738bail:
739 if (status)
740 o2info_set_request_error(&oir, req);
741
742 return status;
743}
744
745/*
746 * Validate and distinguish OCFS2_IOC_INFO requests.
747 *
748 * - validate the magic number.
749 * - distinguish different requests.
750 * - validate size of different requests.
751 */
752int ocfs2_info_handle_request(struct inode *inode,
753 struct ocfs2_info_request __user *req)
754{
755 int status = -EFAULT;
756 struct ocfs2_info_request oir;
757
758 if (o2info_from_user(oir, req))
759 goto bail;
760
761 status = -EINVAL;
762 if (oir.ir_magic != OCFS2_INFO_MAGIC)
763 goto bail;
764
765 switch (oir.ir_code) {
766 case OCFS2_INFO_BLOCKSIZE:
767 if (oir.ir_size == sizeof(struct ocfs2_info_blocksize))
768 status = ocfs2_info_handle_blocksize(inode, req);
769 break;
770 case OCFS2_INFO_CLUSTERSIZE:
771 if (oir.ir_size == sizeof(struct ocfs2_info_clustersize))
772 status = ocfs2_info_handle_clustersize(inode, req);
773 break;
774 case OCFS2_INFO_MAXSLOTS:
775 if (oir.ir_size == sizeof(struct ocfs2_info_maxslots))
776 status = ocfs2_info_handle_maxslots(inode, req);
777 break;
778 case OCFS2_INFO_LABEL:
779 if (oir.ir_size == sizeof(struct ocfs2_info_label))
780 status = ocfs2_info_handle_label(inode, req);
781 break;
782 case OCFS2_INFO_UUID:
783 if (oir.ir_size == sizeof(struct ocfs2_info_uuid))
784 status = ocfs2_info_handle_uuid(inode, req);
785 break;
786 case OCFS2_INFO_FS_FEATURES:
787 if (oir.ir_size == sizeof(struct ocfs2_info_fs_features))
788 status = ocfs2_info_handle_fs_features(inode, req);
789 break;
790 case OCFS2_INFO_JOURNAL_SIZE:
791 if (oir.ir_size == sizeof(struct ocfs2_info_journal_size))
792 status = ocfs2_info_handle_journal_size(inode, req);
793 break;
794 case OCFS2_INFO_FREEINODE:
795 if (oir.ir_size == sizeof(struct ocfs2_info_freeinode))
796 status = ocfs2_info_handle_freeinode(inode, req);
797 break;
798 case OCFS2_INFO_FREEFRAG:
799 if (oir.ir_size == sizeof(struct ocfs2_info_freefrag))
800 status = ocfs2_info_handle_freefrag(inode, req);
801 break;
802 default:
803 status = ocfs2_info_handle_unknown(inode, req);
804 break;
805 }
806
807bail:
808 return status;
809}
810
811int ocfs2_get_request_ptr(struct ocfs2_info *info, int idx,
812 u64 *req_addr, int compat_flag)
813{
814 int status = -EFAULT;
815 u64 __user *bp = NULL;
816
817 if (compat_flag) {
818#ifdef CONFIG_COMPAT
819 /*
820 * pointer bp stores the base address of a pointers array,
821 * which collects all addresses of separate request.
822 */
823 bp = (u64 __user *)(unsigned long)compat_ptr(info->oi_requests);
824#else
825 BUG();
826#endif
827 } else
828 bp = (u64 __user *)(unsigned long)(info->oi_requests);
829
830 if (o2info_from_user(*req_addr, bp + idx))
831 goto bail;
832
833 status = 0;
834bail:
835 return status;
836}
837
838/*
839 * OCFS2_IOC_INFO handles an array of requests passed from userspace.
840 *
841 * ocfs2_info_handle() recevies a large info aggregation, grab and
842 * validate the request count from header, then break it into small
843 * pieces, later specific handlers can handle them one by one.
844 *
845 * Idea here is to make each separate request small enough to ensure
846 * a better backward&forward compatibility, since a small piece of
847 * request will be less likely to be broken if disk layout get changed.
848 */
849int ocfs2_info_handle(struct inode *inode, struct ocfs2_info *info,
850 int compat_flag)
851{
852 int i, status = 0;
853 u64 req_addr;
854 struct ocfs2_info_request __user *reqp;
855
856 if ((info->oi_count > OCFS2_INFO_MAX_REQUEST) ||
857 (!info->oi_requests)) {
858 status = -EINVAL;
859 goto bail;
860 }
861
862 for (i = 0; i < info->oi_count; i++) {
863
864 status = ocfs2_get_request_ptr(info, i, &req_addr, compat_flag);
865 if (status)
866 break;
867
868 reqp = (struct ocfs2_info_request *)(unsigned long)req_addr;
869 if (!reqp) {
870 status = -EINVAL;
871 goto bail;
872 }
873
874 status = ocfs2_info_handle_request(inode, reqp);
875 if (status)
876 break;
877 }
878
879bail:
880 return status;
881}
882
883long ocfs2_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
884{
885 struct inode *inode = filp->f_path.dentry->d_inode;
886 unsigned int flags;
887 int new_clusters;
888 int status;
889 struct ocfs2_space_resv sr;
890 struct ocfs2_new_group_input input;
891 struct reflink_arguments args;
892 const char *old_path, *new_path;
893 bool preserve;
894 struct ocfs2_info info;
895
896 switch (cmd) {
897 case OCFS2_IOC_GETFLAGS:
898 status = ocfs2_get_inode_attr(inode, &flags);
899 if (status < 0)
900 return status;
901
902 flags &= OCFS2_FL_VISIBLE;
903 return put_user(flags, (int __user *) arg);
904 case OCFS2_IOC_SETFLAGS:
905 if (get_user(flags, (int __user *) arg))
906 return -EFAULT;
907
908 status = mnt_want_write(filp->f_path.mnt);
909 if (status)
910 return status;
911 status = ocfs2_set_inode_attr(inode, flags,
912 OCFS2_FL_MODIFIABLE);
913 mnt_drop_write(filp->f_path.mnt);
914 return status;
915 case OCFS2_IOC_RESVSP:
916 case OCFS2_IOC_RESVSP64:
917 case OCFS2_IOC_UNRESVSP:
918 case OCFS2_IOC_UNRESVSP64:
919 if (copy_from_user(&sr, (int __user *) arg, sizeof(sr)))
920 return -EFAULT;
921
922 return ocfs2_change_file_space(filp, cmd, &sr);
923 case OCFS2_IOC_GROUP_EXTEND:
924 if (!capable(CAP_SYS_RESOURCE))
925 return -EPERM;
926
927 if (get_user(new_clusters, (int __user *)arg))
928 return -EFAULT;
929
930 return ocfs2_group_extend(inode, new_clusters);
931 case OCFS2_IOC_GROUP_ADD:
932 case OCFS2_IOC_GROUP_ADD64:
933 if (!capable(CAP_SYS_RESOURCE))
934 return -EPERM;
935
936 if (copy_from_user(&input, (int __user *) arg, sizeof(input)))
937 return -EFAULT;
938
939 return ocfs2_group_add(inode, &input);
940 case OCFS2_IOC_REFLINK:
941 if (copy_from_user(&args, (struct reflink_arguments *)arg,
942 sizeof(args)))
943 return -EFAULT;
944 old_path = (const char *)(unsigned long)args.old_path;
945 new_path = (const char *)(unsigned long)args.new_path;
946 preserve = (args.preserve != 0);
947
948 return ocfs2_reflink_ioctl(inode, old_path, new_path, preserve);
949 case OCFS2_IOC_INFO:
950 if (copy_from_user(&info, (struct ocfs2_info __user *)arg,
951 sizeof(struct ocfs2_info)))
952 return -EFAULT;
953
954 return ocfs2_info_handle(inode, &info, 0);
955 case FITRIM:
956 {
957 struct super_block *sb = inode->i_sb;
958 struct fstrim_range range;
959 int ret = 0;
960
961 if (!capable(CAP_SYS_ADMIN))
962 return -EPERM;
963
964 if (copy_from_user(&range, (struct fstrim_range *)arg,
965 sizeof(range)))
966 return -EFAULT;
967
968 ret = ocfs2_trim_fs(sb, &range);
969 if (ret < 0)
970 return ret;
971
972 if (copy_to_user((struct fstrim_range *)arg, &range,
973 sizeof(range)))
974 return -EFAULT;
975
976 return 0;
977 }
978 case OCFS2_IOC_MOVE_EXT:
979 return ocfs2_ioctl_move_extents(filp, (void __user *)arg);
980 default:
981 return -ENOTTY;
982 }
983}
984
985#ifdef CONFIG_COMPAT
986long ocfs2_compat_ioctl(struct file *file, unsigned cmd, unsigned long arg)
987{
988 bool preserve;
989 struct reflink_arguments args;
990 struct inode *inode = file->f_path.dentry->d_inode;
991 struct ocfs2_info info;
992
993 switch (cmd) {
994 case OCFS2_IOC32_GETFLAGS:
995 cmd = OCFS2_IOC_GETFLAGS;
996 break;
997 case OCFS2_IOC32_SETFLAGS:
998 cmd = OCFS2_IOC_SETFLAGS;
999 break;
1000 case OCFS2_IOC_RESVSP:
1001 case OCFS2_IOC_RESVSP64:
1002 case OCFS2_IOC_UNRESVSP:
1003 case OCFS2_IOC_UNRESVSP64:
1004 case OCFS2_IOC_GROUP_EXTEND:
1005 case OCFS2_IOC_GROUP_ADD:
1006 case OCFS2_IOC_GROUP_ADD64:
1007 case FITRIM:
1008 break;
1009 case OCFS2_IOC_REFLINK:
1010 if (copy_from_user(&args, (struct reflink_arguments *)arg,
1011 sizeof(args)))
1012 return -EFAULT;
1013 preserve = (args.preserve != 0);
1014
1015 return ocfs2_reflink_ioctl(inode, compat_ptr(args.old_path),
1016 compat_ptr(args.new_path), preserve);
1017 case OCFS2_IOC_INFO:
1018 if (copy_from_user(&info, (struct ocfs2_info __user *)arg,
1019 sizeof(struct ocfs2_info)))
1020 return -EFAULT;
1021
1022 return ocfs2_info_handle(inode, &info, 1);
1023 case OCFS2_IOC_MOVE_EXT:
1024 break;
1025 default:
1026 return -ENOIOCTLCMD;
1027 }
1028
1029 return ocfs2_ioctl(file, cmd, arg);
1030}
1031#endif
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