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