Linux Audio

Check our new training course

Loading...
v5.9
  1// SPDX-License-Identifier: GPL-2.0
  2/*
  3 * linux/fs/ocfs2/ioctl.c
  4 *
  5 * Copyright (C) 2006 Herbert Poetzl
  6 * adapted from Remy Card's ext2/ioctl.c
  7 */
  8
  9#include <linux/fs.h>
 10#include <linux/mount.h>
 11#include <linux/blkdev.h>
 12#include <linux/compat.h>
 13
 14#include <cluster/masklog.h>
 15
 16#include "ocfs2.h"
 17#include "alloc.h"
 18#include "dlmglue.h"
 19#include "file.h"
 20#include "inode.h"
 21#include "journal.h"
 22
 23#include "ocfs2_fs.h"
 24#include "ioctl.h"
 25#include "resize.h"
 26#include "refcounttree.h"
 27#include "sysfile.h"
 28#include "dir.h"
 29#include "buffer_head_io.h"
 30#include "suballoc.h"
 31#include "move_extents.h"
 32
 33#define o2info_from_user(a, b)	\
 34		copy_from_user(&(a), (b), sizeof(a))
 35#define o2info_to_user(a, b)	\
 36		copy_to_user((typeof(a) __user *)b, &(a), sizeof(a))
 37
 38/*
 39 * This is just a best-effort to tell userspace that this request
 40 * caused the error.
 41 */
 42static inline void o2info_set_request_error(struct ocfs2_info_request *kreq,
 43					struct ocfs2_info_request __user *req)
 44{
 45	kreq->ir_flags |= OCFS2_INFO_FL_ERROR;
 46	(void)put_user(kreq->ir_flags, (__u32 __user *)&(req->ir_flags));
 47}
 48
 49static inline void o2info_set_request_filled(struct ocfs2_info_request *req)
 50{
 51	req->ir_flags |= OCFS2_INFO_FL_FILLED;
 52}
 53
 54static inline void o2info_clear_request_filled(struct ocfs2_info_request *req)
 55{
 56	req->ir_flags &= ~OCFS2_INFO_FL_FILLED;
 57}
 58
 59static inline int o2info_coherent(struct ocfs2_info_request *req)
 60{
 61	return (!(req->ir_flags & OCFS2_INFO_FL_NON_COHERENT));
 62}
 63
 64static int ocfs2_get_inode_attr(struct inode *inode, unsigned *flags)
 65{
 66	int status;
 67
 68	status = ocfs2_inode_lock(inode, NULL, 0);
 69	if (status < 0) {
 70		mlog_errno(status);
 71		return status;
 72	}
 73	ocfs2_get_inode_flags(OCFS2_I(inode));
 74	*flags = OCFS2_I(inode)->ip_attr;
 75	ocfs2_inode_unlock(inode, 0);
 76
 77	return status;
 78}
 79
 80static int ocfs2_set_inode_attr(struct inode *inode, unsigned flags,
 81				unsigned mask)
 82{
 83	struct ocfs2_inode_info *ocfs2_inode = OCFS2_I(inode);
 84	struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
 85	handle_t *handle = NULL;
 86	struct buffer_head *bh = NULL;
 87	unsigned oldflags;
 88	int status;
 89
 90	inode_lock(inode);
 91
 92	status = ocfs2_inode_lock(inode, &bh, 1);
 93	if (status < 0) {
 94		mlog_errno(status);
 95		goto bail;
 96	}
 97
 98	status = -EACCES;
 99	if (!inode_owner_or_capable(inode))
100		goto bail_unlock;
101
102	if (!S_ISDIR(inode->i_mode))
103		flags &= ~OCFS2_DIRSYNC_FL;
104
105	oldflags = ocfs2_inode->ip_attr;
106	flags = flags & mask;
107	flags |= oldflags & ~mask;
108
109	status = vfs_ioc_setflags_prepare(inode, oldflags, flags);
110	if (status)
111		goto bail_unlock;
 
 
 
 
 
 
 
112
113	handle = ocfs2_start_trans(osb, OCFS2_INODE_UPDATE_CREDITS);
114	if (IS_ERR(handle)) {
115		status = PTR_ERR(handle);
116		mlog_errno(status);
117		goto bail_unlock;
118	}
119
120	ocfs2_inode->ip_attr = flags;
121	ocfs2_set_inode_flags(inode);
122
123	status = ocfs2_mark_inode_dirty(handle, inode, bh);
124	if (status < 0)
125		mlog_errno(status);
126
127	ocfs2_commit_trans(osb, handle);
128
129bail_unlock:
130	ocfs2_inode_unlock(inode, 1);
131bail:
132	inode_unlock(inode);
133
134	brelse(bh);
135
136	return status;
137}
138
139static int ocfs2_info_handle_blocksize(struct inode *inode,
140				       struct ocfs2_info_request __user *req)
141{
142	struct ocfs2_info_blocksize oib;
143
144	if (o2info_from_user(oib, req))
145		return -EFAULT;
146
147	oib.ib_blocksize = inode->i_sb->s_blocksize;
148
149	o2info_set_request_filled(&oib.ib_req);
150
151	if (o2info_to_user(oib, req))
152		return -EFAULT;
153
154	return 0;
155}
156
157static int ocfs2_info_handle_clustersize(struct inode *inode,
158					 struct ocfs2_info_request __user *req)
159{
160	struct ocfs2_info_clustersize oic;
161	struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
162
163	if (o2info_from_user(oic, req))
164		return -EFAULT;
165
166	oic.ic_clustersize = osb->s_clustersize;
167
168	o2info_set_request_filled(&oic.ic_req);
169
170	if (o2info_to_user(oic, req))
171		return -EFAULT;
172
173	return 0;
174}
175
176static int ocfs2_info_handle_maxslots(struct inode *inode,
177				      struct ocfs2_info_request __user *req)
178{
179	struct ocfs2_info_maxslots oim;
180	struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
181
182	if (o2info_from_user(oim, req))
183		return -EFAULT;
184
185	oim.im_max_slots = osb->max_slots;
186
187	o2info_set_request_filled(&oim.im_req);
188
189	if (o2info_to_user(oim, req))
190		return -EFAULT;
191
192	return 0;
193}
194
195static int ocfs2_info_handle_label(struct inode *inode,
196				   struct ocfs2_info_request __user *req)
197{
198	struct ocfs2_info_label oil;
199	struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
200
201	if (o2info_from_user(oil, req))
202		return -EFAULT;
203
204	memcpy(oil.il_label, osb->vol_label, OCFS2_MAX_VOL_LABEL_LEN);
205
206	o2info_set_request_filled(&oil.il_req);
207
208	if (o2info_to_user(oil, req))
209		return -EFAULT;
210
211	return 0;
212}
213
214static int ocfs2_info_handle_uuid(struct inode *inode,
215				  struct ocfs2_info_request __user *req)
216{
217	struct ocfs2_info_uuid oiu;
218	struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
219
220	if (o2info_from_user(oiu, req))
221		return -EFAULT;
222
223	memcpy(oiu.iu_uuid_str, osb->uuid_str, OCFS2_TEXT_UUID_LEN + 1);
224
225	o2info_set_request_filled(&oiu.iu_req);
226
227	if (o2info_to_user(oiu, req))
228		return -EFAULT;
229
230	return 0;
231}
232
233static int ocfs2_info_handle_fs_features(struct inode *inode,
234					 struct ocfs2_info_request __user *req)
235{
236	struct ocfs2_info_fs_features oif;
237	struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
238
239	if (o2info_from_user(oif, req))
240		return -EFAULT;
241
242	oif.if_compat_features = osb->s_feature_compat;
243	oif.if_incompat_features = osb->s_feature_incompat;
244	oif.if_ro_compat_features = osb->s_feature_ro_compat;
245
246	o2info_set_request_filled(&oif.if_req);
247
248	if (o2info_to_user(oif, req))
249		return -EFAULT;
250
251	return 0;
252}
253
254static int ocfs2_info_handle_journal_size(struct inode *inode,
255					  struct ocfs2_info_request __user *req)
256{
257	struct ocfs2_info_journal_size oij;
258	struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
259
260	if (o2info_from_user(oij, req))
261		return -EFAULT;
262
263	oij.ij_journal_size = i_size_read(osb->journal->j_inode);
264
265	o2info_set_request_filled(&oij.ij_req);
266
267	if (o2info_to_user(oij, req))
268		return -EFAULT;
269
270	return 0;
271}
272
273static int ocfs2_info_scan_inode_alloc(struct ocfs2_super *osb,
274				       struct inode *inode_alloc, u64 blkno,
275				       struct ocfs2_info_freeinode *fi,
276				       u32 slot)
277{
278	int status = 0, unlock = 0;
279
280	struct buffer_head *bh = NULL;
281	struct ocfs2_dinode *dinode_alloc = NULL;
282
283	if (inode_alloc)
284		inode_lock(inode_alloc);
285
286	if (inode_alloc && o2info_coherent(&fi->ifi_req)) {
287		status = ocfs2_inode_lock(inode_alloc, &bh, 0);
288		if (status < 0) {
289			mlog_errno(status);
290			goto bail;
291		}
292		unlock = 1;
293	} else {
294		status = ocfs2_read_blocks_sync(osb, blkno, 1, &bh);
295		if (status < 0) {
296			mlog_errno(status);
297			goto bail;
298		}
299	}
300
301	dinode_alloc = (struct ocfs2_dinode *)bh->b_data;
302
303	fi->ifi_stat[slot].lfi_total =
304		le32_to_cpu(dinode_alloc->id1.bitmap1.i_total);
305	fi->ifi_stat[slot].lfi_free =
306		le32_to_cpu(dinode_alloc->id1.bitmap1.i_total) -
307		le32_to_cpu(dinode_alloc->id1.bitmap1.i_used);
308
309bail:
310	if (unlock)
311		ocfs2_inode_unlock(inode_alloc, 0);
312
313	if (inode_alloc)
314		inode_unlock(inode_alloc);
315
316	brelse(bh);
317
318	return status;
319}
320
321static int ocfs2_info_handle_freeinode(struct inode *inode,
322				       struct ocfs2_info_request __user *req)
323{
324	u32 i;
325	u64 blkno = -1;
326	char namebuf[40];
327	int status, type = INODE_ALLOC_SYSTEM_INODE;
328	struct ocfs2_info_freeinode *oifi = NULL;
329	struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
330	struct inode *inode_alloc = NULL;
331
332	oifi = kzalloc(sizeof(struct ocfs2_info_freeinode), GFP_KERNEL);
333	if (!oifi) {
334		status = -ENOMEM;
335		mlog_errno(status);
336		goto out_err;
337	}
338
339	if (o2info_from_user(*oifi, req)) {
340		status = -EFAULT;
341		goto out_free;
342	}
343
344	oifi->ifi_slotnum = osb->max_slots;
345
346	for (i = 0; i < oifi->ifi_slotnum; i++) {
347		if (o2info_coherent(&oifi->ifi_req)) {
348			inode_alloc = ocfs2_get_system_file_inode(osb, type, i);
349			if (!inode_alloc) {
350				mlog(ML_ERROR, "unable to get alloc inode in "
351				    "slot %u\n", i);
352				status = -EIO;
353				goto bail;
354			}
355		} else {
356			ocfs2_sprintf_system_inode_name(namebuf,
357							sizeof(namebuf),
358							type, i);
359			status = ocfs2_lookup_ino_from_name(osb->sys_root_inode,
360							    namebuf,
361							    strlen(namebuf),
362							    &blkno);
363			if (status < 0) {
364				status = -ENOENT;
365				goto bail;
366			}
367		}
368
369		status = ocfs2_info_scan_inode_alloc(osb, inode_alloc, blkno, oifi, i);
370
371		iput(inode_alloc);
372		inode_alloc = NULL;
373
374		if (status < 0)
375			goto bail;
376	}
377
378	o2info_set_request_filled(&oifi->ifi_req);
379
380	if (o2info_to_user(*oifi, req)) {
381		status = -EFAULT;
382		goto out_free;
383	}
384
385	status = 0;
386bail:
387	if (status)
388		o2info_set_request_error(&oifi->ifi_req, req);
389out_free:
390	kfree(oifi);
391out_err:
392	return status;
393}
394
395static void o2ffg_update_histogram(struct ocfs2_info_free_chunk_list *hist,
396				   unsigned int chunksize)
397{
398	u32 index;
399
400	index = __ilog2_u32(chunksize);
401	if (index >= OCFS2_INFO_MAX_HIST)
402		index = OCFS2_INFO_MAX_HIST - 1;
403
404	hist->fc_chunks[index]++;
405	hist->fc_clusters[index] += chunksize;
406}
407
408static void o2ffg_update_stats(struct ocfs2_info_freefrag_stats *stats,
409			       unsigned int chunksize)
410{
411	if (chunksize > stats->ffs_max)
412		stats->ffs_max = chunksize;
413
414	if (chunksize < stats->ffs_min)
415		stats->ffs_min = chunksize;
416
417	stats->ffs_avg += chunksize;
418	stats->ffs_free_chunks_real++;
419}
420
421static void ocfs2_info_update_ffg(struct ocfs2_info_freefrag *ffg,
422				  unsigned int chunksize)
423{
424	o2ffg_update_histogram(&(ffg->iff_ffs.ffs_fc_hist), chunksize);
425	o2ffg_update_stats(&(ffg->iff_ffs), chunksize);
426}
427
428static int ocfs2_info_freefrag_scan_chain(struct ocfs2_super *osb,
429					  struct inode *gb_inode,
430					  struct ocfs2_dinode *gb_dinode,
431					  struct ocfs2_chain_rec *rec,
432					  struct ocfs2_info_freefrag *ffg,
433					  u32 chunks_in_group)
434{
435	int status = 0, used;
436	u64 blkno;
437
438	struct buffer_head *bh = NULL;
439	struct ocfs2_group_desc *bg = NULL;
440
441	unsigned int max_bits, num_clusters;
442	unsigned int offset = 0, cluster, chunk;
443	unsigned int chunk_free, last_chunksize = 0;
444
445	if (!le32_to_cpu(rec->c_free))
446		goto bail;
447
448	do {
449		if (!bg)
450			blkno = le64_to_cpu(rec->c_blkno);
451		else
452			blkno = le64_to_cpu(bg->bg_next_group);
453
454		if (bh) {
455			brelse(bh);
456			bh = NULL;
457		}
458
459		if (o2info_coherent(&ffg->iff_req))
460			status = ocfs2_read_group_descriptor(gb_inode,
461							     gb_dinode,
462							     blkno, &bh);
463		else
464			status = ocfs2_read_blocks_sync(osb, blkno, 1, &bh);
465
466		if (status < 0) {
467			mlog(ML_ERROR, "Can't read the group descriptor # "
468			     "%llu from device.", (unsigned long long)blkno);
469			status = -EIO;
470			goto bail;
471		}
472
473		bg = (struct ocfs2_group_desc *)bh->b_data;
474
475		if (!le16_to_cpu(bg->bg_free_bits_count))
476			continue;
477
478		max_bits = le16_to_cpu(bg->bg_bits);
479		offset = 0;
480
481		for (chunk = 0; chunk < chunks_in_group; chunk++) {
482			/*
483			 * last chunk may be not an entire one.
484			 */
485			if ((offset + ffg->iff_chunksize) > max_bits)
486				num_clusters = max_bits - offset;
487			else
488				num_clusters = ffg->iff_chunksize;
489
490			chunk_free = 0;
491			for (cluster = 0; cluster < num_clusters; cluster++) {
492				used = ocfs2_test_bit(offset,
493						(unsigned long *)bg->bg_bitmap);
494				/*
495				 * - chunk_free counts free clusters in #N chunk.
496				 * - last_chunksize records the size(in) clusters
497				 *   for the last real free chunk being counted.
498				 */
499				if (!used) {
500					last_chunksize++;
501					chunk_free++;
502				}
503
504				if (used && last_chunksize) {
505					ocfs2_info_update_ffg(ffg,
506							      last_chunksize);
507					last_chunksize = 0;
508				}
509
510				offset++;
511			}
512
513			if (chunk_free == ffg->iff_chunksize)
514				ffg->iff_ffs.ffs_free_chunks++;
515		}
516
517		/*
518		 * need to update the info for last free chunk.
519		 */
520		if (last_chunksize)
521			ocfs2_info_update_ffg(ffg, last_chunksize);
522
523	} while (le64_to_cpu(bg->bg_next_group));
524
525bail:
526	brelse(bh);
527
528	return status;
529}
530
531static int ocfs2_info_freefrag_scan_bitmap(struct ocfs2_super *osb,
532					   struct inode *gb_inode, u64 blkno,
533					   struct ocfs2_info_freefrag *ffg)
534{
535	u32 chunks_in_group;
536	int status = 0, unlock = 0, i;
537
538	struct buffer_head *bh = NULL;
539	struct ocfs2_chain_list *cl = NULL;
540	struct ocfs2_chain_rec *rec = NULL;
541	struct ocfs2_dinode *gb_dinode = NULL;
542
543	if (gb_inode)
544		inode_lock(gb_inode);
545
546	if (o2info_coherent(&ffg->iff_req)) {
547		status = ocfs2_inode_lock(gb_inode, &bh, 0);
548		if (status < 0) {
549			mlog_errno(status);
550			goto bail;
551		}
552		unlock = 1;
553	} else {
554		status = ocfs2_read_blocks_sync(osb, blkno, 1, &bh);
555		if (status < 0) {
556			mlog_errno(status);
557			goto bail;
558		}
559	}
560
561	gb_dinode = (struct ocfs2_dinode *)bh->b_data;
562	cl = &(gb_dinode->id2.i_chain);
563
564	/*
565	 * Chunksize(in) clusters from userspace should be
566	 * less than clusters in a group.
567	 */
568	if (ffg->iff_chunksize > le16_to_cpu(cl->cl_cpg)) {
569		status = -EINVAL;
570		goto bail;
571	}
572
573	memset(&ffg->iff_ffs, 0, sizeof(struct ocfs2_info_freefrag_stats));
574
575	ffg->iff_ffs.ffs_min = ~0U;
576	ffg->iff_ffs.ffs_clusters =
577			le32_to_cpu(gb_dinode->id1.bitmap1.i_total);
578	ffg->iff_ffs.ffs_free_clusters = ffg->iff_ffs.ffs_clusters -
579			le32_to_cpu(gb_dinode->id1.bitmap1.i_used);
580
581	chunks_in_group = le16_to_cpu(cl->cl_cpg) / ffg->iff_chunksize + 1;
582
583	for (i = 0; i < le16_to_cpu(cl->cl_next_free_rec); i++) {
584		rec = &(cl->cl_recs[i]);
585		status = ocfs2_info_freefrag_scan_chain(osb, gb_inode,
586							gb_dinode,
587							rec, ffg,
588							chunks_in_group);
589		if (status)
590			goto bail;
591	}
592
593	if (ffg->iff_ffs.ffs_free_chunks_real)
594		ffg->iff_ffs.ffs_avg = (ffg->iff_ffs.ffs_avg /
595					ffg->iff_ffs.ffs_free_chunks_real);
596bail:
597	if (unlock)
598		ocfs2_inode_unlock(gb_inode, 0);
599
600	if (gb_inode)
601		inode_unlock(gb_inode);
602
603	iput(gb_inode);
604	brelse(bh);
605
606	return status;
607}
608
609static int ocfs2_info_handle_freefrag(struct inode *inode,
610				      struct ocfs2_info_request __user *req)
611{
612	u64 blkno = -1;
613	char namebuf[40];
614	int status, type = GLOBAL_BITMAP_SYSTEM_INODE;
615
616	struct ocfs2_info_freefrag *oiff;
617	struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
618	struct inode *gb_inode = NULL;
619
620	oiff = kzalloc(sizeof(struct ocfs2_info_freefrag), GFP_KERNEL);
621	if (!oiff) {
622		status = -ENOMEM;
623		mlog_errno(status);
624		goto out_err;
625	}
626
627	if (o2info_from_user(*oiff, req)) {
628		status = -EFAULT;
629		goto out_free;
630	}
631	/*
632	 * chunksize from userspace should be power of 2.
633	 */
634	if ((oiff->iff_chunksize & (oiff->iff_chunksize - 1)) ||
635	    (!oiff->iff_chunksize)) {
636		status = -EINVAL;
637		goto bail;
638	}
639
640	if (o2info_coherent(&oiff->iff_req)) {
641		gb_inode = ocfs2_get_system_file_inode(osb, type,
642						       OCFS2_INVALID_SLOT);
643		if (!gb_inode) {
644			mlog(ML_ERROR, "unable to get global_bitmap inode\n");
645			status = -EIO;
646			goto bail;
647		}
648	} else {
649		ocfs2_sprintf_system_inode_name(namebuf, sizeof(namebuf), type,
650						OCFS2_INVALID_SLOT);
651		status = ocfs2_lookup_ino_from_name(osb->sys_root_inode,
652						    namebuf,
653						    strlen(namebuf),
654						    &blkno);
655		if (status < 0) {
656			status = -ENOENT;
657			goto bail;
658		}
659	}
660
661	status = ocfs2_info_freefrag_scan_bitmap(osb, gb_inode, blkno, oiff);
662	if (status < 0)
663		goto bail;
664
665	o2info_set_request_filled(&oiff->iff_req);
666
667	if (o2info_to_user(*oiff, req)) {
668		status = -EFAULT;
669		goto out_free;
670	}
671
672	status = 0;
673bail:
674	if (status)
675		o2info_set_request_error(&oiff->iff_req, req);
676out_free:
677	kfree(oiff);
678out_err:
679	return status;
680}
681
682static int ocfs2_info_handle_unknown(struct inode *inode,
683				     struct ocfs2_info_request __user *req)
684{
685	struct ocfs2_info_request oir;
686
687	if (o2info_from_user(oir, req))
688		return -EFAULT;
689
690	o2info_clear_request_filled(&oir);
691
692	if (o2info_to_user(oir, req))
693		return -EFAULT;
694
695	return 0;
696}
697
698/*
699 * Validate and distinguish OCFS2_IOC_INFO requests.
700 *
701 * - validate the magic number.
702 * - distinguish different requests.
703 * - validate size of different requests.
704 */
705static int ocfs2_info_handle_request(struct inode *inode,
706				     struct ocfs2_info_request __user *req)
707{
708	int status = -EFAULT;
709	struct ocfs2_info_request oir;
710
711	if (o2info_from_user(oir, req))
712		goto bail;
713
714	status = -EINVAL;
715	if (oir.ir_magic != OCFS2_INFO_MAGIC)
716		goto bail;
717
718	switch (oir.ir_code) {
719	case OCFS2_INFO_BLOCKSIZE:
720		if (oir.ir_size == sizeof(struct ocfs2_info_blocksize))
721			status = ocfs2_info_handle_blocksize(inode, req);
722		break;
723	case OCFS2_INFO_CLUSTERSIZE:
724		if (oir.ir_size == sizeof(struct ocfs2_info_clustersize))
725			status = ocfs2_info_handle_clustersize(inode, req);
726		break;
727	case OCFS2_INFO_MAXSLOTS:
728		if (oir.ir_size == sizeof(struct ocfs2_info_maxslots))
729			status = ocfs2_info_handle_maxslots(inode, req);
730		break;
731	case OCFS2_INFO_LABEL:
732		if (oir.ir_size == sizeof(struct ocfs2_info_label))
733			status = ocfs2_info_handle_label(inode, req);
734		break;
735	case OCFS2_INFO_UUID:
736		if (oir.ir_size == sizeof(struct ocfs2_info_uuid))
737			status = ocfs2_info_handle_uuid(inode, req);
738		break;
739	case OCFS2_INFO_FS_FEATURES:
740		if (oir.ir_size == sizeof(struct ocfs2_info_fs_features))
741			status = ocfs2_info_handle_fs_features(inode, req);
742		break;
743	case OCFS2_INFO_JOURNAL_SIZE:
744		if (oir.ir_size == sizeof(struct ocfs2_info_journal_size))
745			status = ocfs2_info_handle_journal_size(inode, req);
746		break;
747	case OCFS2_INFO_FREEINODE:
748		if (oir.ir_size == sizeof(struct ocfs2_info_freeinode))
749			status = ocfs2_info_handle_freeinode(inode, req);
750		break;
751	case OCFS2_INFO_FREEFRAG:
752		if (oir.ir_size == sizeof(struct ocfs2_info_freefrag))
753			status = ocfs2_info_handle_freefrag(inode, req);
754		break;
755	default:
756		status = ocfs2_info_handle_unknown(inode, req);
757		break;
758	}
759
760bail:
761	return status;
762}
763
764static int ocfs2_get_request_ptr(struct ocfs2_info *info, int idx,
765				 u64 *req_addr, int compat_flag)
766{
767	int status = -EFAULT;
768	u64 __user *bp = NULL;
769
770	if (compat_flag) {
771#ifdef CONFIG_COMPAT
772		/*
773		 * pointer bp stores the base address of a pointers array,
774		 * which collects all addresses of separate request.
775		 */
776		bp = (u64 __user *)(unsigned long)compat_ptr(info->oi_requests);
777#else
778		BUG();
779#endif
780	} else
781		bp = (u64 __user *)(unsigned long)(info->oi_requests);
782
783	if (o2info_from_user(*req_addr, bp + idx))
784		goto bail;
785
786	status = 0;
787bail:
788	return status;
789}
790
791/*
792 * OCFS2_IOC_INFO handles an array of requests passed from userspace.
793 *
794 * ocfs2_info_handle() recevies a large info aggregation, grab and
795 * validate the request count from header, then break it into small
796 * pieces, later specific handlers can handle them one by one.
797 *
798 * Idea here is to make each separate request small enough to ensure
799 * a better backward&forward compatibility, since a small piece of
800 * request will be less likely to be broken if disk layout get changed.
801 */
802static int ocfs2_info_handle(struct inode *inode, struct ocfs2_info *info,
803			     int compat_flag)
804{
805	int i, status = 0;
806	u64 req_addr;
807	struct ocfs2_info_request __user *reqp;
808
809	if ((info->oi_count > OCFS2_INFO_MAX_REQUEST) ||
810	    (!info->oi_requests)) {
811		status = -EINVAL;
812		goto bail;
813	}
814
815	for (i = 0; i < info->oi_count; i++) {
816
817		status = ocfs2_get_request_ptr(info, i, &req_addr, compat_flag);
818		if (status)
819			break;
820
821		reqp = (struct ocfs2_info_request __user *)(unsigned long)req_addr;
822		if (!reqp) {
823			status = -EINVAL;
824			goto bail;
825		}
826
827		status = ocfs2_info_handle_request(inode, reqp);
828		if (status)
829			break;
830	}
831
832bail:
833	return status;
834}
835
836long ocfs2_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
837{
838	struct inode *inode = file_inode(filp);
839	unsigned int flags;
840	int new_clusters;
841	int status;
842	struct ocfs2_space_resv sr;
843	struct ocfs2_new_group_input input;
844	struct reflink_arguments args;
845	const char __user *old_path;
846	const char __user *new_path;
847	bool preserve;
848	struct ocfs2_info info;
849	void __user *argp = (void __user *)arg;
850
851	switch (cmd) {
852	case OCFS2_IOC_GETFLAGS:
853		status = ocfs2_get_inode_attr(inode, &flags);
854		if (status < 0)
855			return status;
856
857		flags &= OCFS2_FL_VISIBLE;
858		return put_user(flags, (int __user *) arg);
859	case OCFS2_IOC_SETFLAGS:
860		if (get_user(flags, (int __user *) arg))
861			return -EFAULT;
862
863		status = mnt_want_write_file(filp);
864		if (status)
865			return status;
866		status = ocfs2_set_inode_attr(inode, flags,
867			OCFS2_FL_MODIFIABLE);
868		mnt_drop_write_file(filp);
869		return status;
870	case OCFS2_IOC_RESVSP:
871	case OCFS2_IOC_RESVSP64:
872	case OCFS2_IOC_UNRESVSP:
873	case OCFS2_IOC_UNRESVSP64:
874		if (copy_from_user(&sr, (int __user *) arg, sizeof(sr)))
875			return -EFAULT;
876
877		return ocfs2_change_file_space(filp, cmd, &sr);
878	case OCFS2_IOC_GROUP_EXTEND:
879		if (!capable(CAP_SYS_RESOURCE))
880			return -EPERM;
881
882		if (get_user(new_clusters, (int __user *)arg))
883			return -EFAULT;
884
885		status = mnt_want_write_file(filp);
886		if (status)
887			return status;
888		status = ocfs2_group_extend(inode, new_clusters);
889		mnt_drop_write_file(filp);
890		return status;
891	case OCFS2_IOC_GROUP_ADD:
892	case OCFS2_IOC_GROUP_ADD64:
893		if (!capable(CAP_SYS_RESOURCE))
894			return -EPERM;
895
896		if (copy_from_user(&input, (int __user *) arg, sizeof(input)))
897			return -EFAULT;
898
899		status = mnt_want_write_file(filp);
900		if (status)
901			return status;
902		status = ocfs2_group_add(inode, &input);
903		mnt_drop_write_file(filp);
904		return status;
905	case OCFS2_IOC_REFLINK:
906		if (copy_from_user(&args, argp, sizeof(args)))
907			return -EFAULT;
908		old_path = (const char __user *)(unsigned long)args.old_path;
909		new_path = (const char __user *)(unsigned long)args.new_path;
910		preserve = (args.preserve != 0);
911
912		return ocfs2_reflink_ioctl(inode, old_path, new_path, preserve);
913	case OCFS2_IOC_INFO:
914		if (copy_from_user(&info, argp, sizeof(struct ocfs2_info)))
915			return -EFAULT;
916
917		return ocfs2_info_handle(inode, &info, 0);
918	case FITRIM:
919	{
920		struct super_block *sb = inode->i_sb;
921		struct request_queue *q = bdev_get_queue(sb->s_bdev);
922		struct fstrim_range range;
923		int ret = 0;
924
925		if (!capable(CAP_SYS_ADMIN))
926			return -EPERM;
927
928		if (!blk_queue_discard(q))
929			return -EOPNOTSUPP;
930
931		if (copy_from_user(&range, argp, sizeof(range)))
932			return -EFAULT;
933
934		range.minlen = max_t(u64, q->limits.discard_granularity,
935				     range.minlen);
936		ret = ocfs2_trim_fs(sb, &range);
937		if (ret < 0)
938			return ret;
939
940		if (copy_to_user(argp, &range, sizeof(range)))
941			return -EFAULT;
942
943		return 0;
944	}
945	case OCFS2_IOC_MOVE_EXT:
946		return ocfs2_ioctl_move_extents(filp, argp);
947	default:
948		return -ENOTTY;
949	}
950}
951
952#ifdef CONFIG_COMPAT
953long ocfs2_compat_ioctl(struct file *file, unsigned cmd, unsigned long arg)
954{
955	bool preserve;
956	struct reflink_arguments args;
957	struct inode *inode = file_inode(file);
958	struct ocfs2_info info;
959	void __user *argp = (void __user *)arg;
960
961	switch (cmd) {
962	case OCFS2_IOC32_GETFLAGS:
963		cmd = OCFS2_IOC_GETFLAGS;
964		break;
965	case OCFS2_IOC32_SETFLAGS:
966		cmd = OCFS2_IOC_SETFLAGS;
967		break;
968	case OCFS2_IOC_RESVSP:
969	case OCFS2_IOC_RESVSP64:
970	case OCFS2_IOC_UNRESVSP:
971	case OCFS2_IOC_UNRESVSP64:
972	case OCFS2_IOC_GROUP_EXTEND:
973	case OCFS2_IOC_GROUP_ADD:
974	case OCFS2_IOC_GROUP_ADD64:
975		break;
976	case OCFS2_IOC_REFLINK:
977		if (copy_from_user(&args, argp, sizeof(args)))
978			return -EFAULT;
979		preserve = (args.preserve != 0);
980
981		return ocfs2_reflink_ioctl(inode, compat_ptr(args.old_path),
982					   compat_ptr(args.new_path), preserve);
983	case OCFS2_IOC_INFO:
984		if (copy_from_user(&info, argp, sizeof(struct ocfs2_info)))
985			return -EFAULT;
986
987		return ocfs2_info_handle(inode, &info, 1);
988	case FITRIM:
989	case OCFS2_IOC_MOVE_EXT:
990		break;
991	default:
992		return -ENOIOCTLCMD;
993	}
994
995	return ocfs2_ioctl(file, cmd, arg);
996}
997#endif
v4.17
   1// SPDX-License-Identifier: GPL-2.0
   2/*
   3 * linux/fs/ocfs2/ioctl.c
   4 *
   5 * Copyright (C) 2006 Herbert Poetzl
   6 * adapted from Remy Card's ext2/ioctl.c
   7 */
   8
   9#include <linux/fs.h>
  10#include <linux/mount.h>
  11#include <linux/blkdev.h>
  12#include <linux/compat.h>
  13
  14#include <cluster/masklog.h>
  15
  16#include "ocfs2.h"
  17#include "alloc.h"
  18#include "dlmglue.h"
  19#include "file.h"
  20#include "inode.h"
  21#include "journal.h"
  22
  23#include "ocfs2_fs.h"
  24#include "ioctl.h"
  25#include "resize.h"
  26#include "refcounttree.h"
  27#include "sysfile.h"
  28#include "dir.h"
  29#include "buffer_head_io.h"
  30#include "suballoc.h"
  31#include "move_extents.h"
  32
  33#define o2info_from_user(a, b)	\
  34		copy_from_user(&(a), (b), sizeof(a))
  35#define o2info_to_user(a, b)	\
  36		copy_to_user((typeof(a) __user *)b, &(a), sizeof(a))
  37
  38/*
  39 * This is just a best-effort to tell userspace that this request
  40 * caused the error.
  41 */
  42static inline void o2info_set_request_error(struct ocfs2_info_request *kreq,
  43					struct ocfs2_info_request __user *req)
  44{
  45	kreq->ir_flags |= OCFS2_INFO_FL_ERROR;
  46	(void)put_user(kreq->ir_flags, (__u32 __user *)&(req->ir_flags));
  47}
  48
  49static inline void o2info_set_request_filled(struct ocfs2_info_request *req)
  50{
  51	req->ir_flags |= OCFS2_INFO_FL_FILLED;
  52}
  53
  54static inline void o2info_clear_request_filled(struct ocfs2_info_request *req)
  55{
  56	req->ir_flags &= ~OCFS2_INFO_FL_FILLED;
  57}
  58
  59static inline int o2info_coherent(struct ocfs2_info_request *req)
  60{
  61	return (!(req->ir_flags & OCFS2_INFO_FL_NON_COHERENT));
  62}
  63
  64static int ocfs2_get_inode_attr(struct inode *inode, unsigned *flags)
  65{
  66	int status;
  67
  68	status = ocfs2_inode_lock(inode, NULL, 0);
  69	if (status < 0) {
  70		mlog_errno(status);
  71		return status;
  72	}
  73	ocfs2_get_inode_flags(OCFS2_I(inode));
  74	*flags = OCFS2_I(inode)->ip_attr;
  75	ocfs2_inode_unlock(inode, 0);
  76
  77	return status;
  78}
  79
  80static int ocfs2_set_inode_attr(struct inode *inode, unsigned flags,
  81				unsigned mask)
  82{
  83	struct ocfs2_inode_info *ocfs2_inode = OCFS2_I(inode);
  84	struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
  85	handle_t *handle = NULL;
  86	struct buffer_head *bh = NULL;
  87	unsigned oldflags;
  88	int status;
  89
  90	inode_lock(inode);
  91
  92	status = ocfs2_inode_lock(inode, &bh, 1);
  93	if (status < 0) {
  94		mlog_errno(status);
  95		goto bail;
  96	}
  97
  98	status = -EACCES;
  99	if (!inode_owner_or_capable(inode))
 100		goto bail_unlock;
 101
 102	if (!S_ISDIR(inode->i_mode))
 103		flags &= ~OCFS2_DIRSYNC_FL;
 104
 105	oldflags = ocfs2_inode->ip_attr;
 106	flags = flags & mask;
 107	flags |= oldflags & ~mask;
 108
 109	/*
 110	 * The IMMUTABLE and APPEND_ONLY flags can only be changed by
 111	 * the relevant capability.
 112	 */
 113	status = -EPERM;
 114	if ((oldflags & OCFS2_IMMUTABLE_FL) || ((flags ^ oldflags) &
 115		(OCFS2_APPEND_FL | OCFS2_IMMUTABLE_FL))) {
 116		if (!capable(CAP_LINUX_IMMUTABLE))
 117			goto bail_unlock;
 118	}
 119
 120	handle = ocfs2_start_trans(osb, OCFS2_INODE_UPDATE_CREDITS);
 121	if (IS_ERR(handle)) {
 122		status = PTR_ERR(handle);
 123		mlog_errno(status);
 124		goto bail_unlock;
 125	}
 126
 127	ocfs2_inode->ip_attr = flags;
 128	ocfs2_set_inode_flags(inode);
 129
 130	status = ocfs2_mark_inode_dirty(handle, inode, bh);
 131	if (status < 0)
 132		mlog_errno(status);
 133
 134	ocfs2_commit_trans(osb, handle);
 135
 136bail_unlock:
 137	ocfs2_inode_unlock(inode, 1);
 138bail:
 139	inode_unlock(inode);
 140
 141	brelse(bh);
 142
 143	return status;
 144}
 145
 146static int ocfs2_info_handle_blocksize(struct inode *inode,
 147				       struct ocfs2_info_request __user *req)
 148{
 149	struct ocfs2_info_blocksize oib;
 150
 151	if (o2info_from_user(oib, req))
 152		return -EFAULT;
 153
 154	oib.ib_blocksize = inode->i_sb->s_blocksize;
 155
 156	o2info_set_request_filled(&oib.ib_req);
 157
 158	if (o2info_to_user(oib, req))
 159		return -EFAULT;
 160
 161	return 0;
 162}
 163
 164static int ocfs2_info_handle_clustersize(struct inode *inode,
 165					 struct ocfs2_info_request __user *req)
 166{
 167	struct ocfs2_info_clustersize oic;
 168	struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
 169
 170	if (o2info_from_user(oic, req))
 171		return -EFAULT;
 172
 173	oic.ic_clustersize = osb->s_clustersize;
 174
 175	o2info_set_request_filled(&oic.ic_req);
 176
 177	if (o2info_to_user(oic, req))
 178		return -EFAULT;
 179
 180	return 0;
 181}
 182
 183static int ocfs2_info_handle_maxslots(struct inode *inode,
 184				      struct ocfs2_info_request __user *req)
 185{
 186	struct ocfs2_info_maxslots oim;
 187	struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
 188
 189	if (o2info_from_user(oim, req))
 190		return -EFAULT;
 191
 192	oim.im_max_slots = osb->max_slots;
 193
 194	o2info_set_request_filled(&oim.im_req);
 195
 196	if (o2info_to_user(oim, req))
 197		return -EFAULT;
 198
 199	return 0;
 200}
 201
 202static int ocfs2_info_handle_label(struct inode *inode,
 203				   struct ocfs2_info_request __user *req)
 204{
 205	struct ocfs2_info_label oil;
 206	struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
 207
 208	if (o2info_from_user(oil, req))
 209		return -EFAULT;
 210
 211	memcpy(oil.il_label, osb->vol_label, OCFS2_MAX_VOL_LABEL_LEN);
 212
 213	o2info_set_request_filled(&oil.il_req);
 214
 215	if (o2info_to_user(oil, req))
 216		return -EFAULT;
 217
 218	return 0;
 219}
 220
 221static int ocfs2_info_handle_uuid(struct inode *inode,
 222				  struct ocfs2_info_request __user *req)
 223{
 224	struct ocfs2_info_uuid oiu;
 225	struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
 226
 227	if (o2info_from_user(oiu, req))
 228		return -EFAULT;
 229
 230	memcpy(oiu.iu_uuid_str, osb->uuid_str, OCFS2_TEXT_UUID_LEN + 1);
 231
 232	o2info_set_request_filled(&oiu.iu_req);
 233
 234	if (o2info_to_user(oiu, req))
 235		return -EFAULT;
 236
 237	return 0;
 238}
 239
 240static int ocfs2_info_handle_fs_features(struct inode *inode,
 241					 struct ocfs2_info_request __user *req)
 242{
 243	struct ocfs2_info_fs_features oif;
 244	struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
 245
 246	if (o2info_from_user(oif, req))
 247		return -EFAULT;
 248
 249	oif.if_compat_features = osb->s_feature_compat;
 250	oif.if_incompat_features = osb->s_feature_incompat;
 251	oif.if_ro_compat_features = osb->s_feature_ro_compat;
 252
 253	o2info_set_request_filled(&oif.if_req);
 254
 255	if (o2info_to_user(oif, req))
 256		return -EFAULT;
 257
 258	return 0;
 259}
 260
 261static int ocfs2_info_handle_journal_size(struct inode *inode,
 262					  struct ocfs2_info_request __user *req)
 263{
 264	struct ocfs2_info_journal_size oij;
 265	struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
 266
 267	if (o2info_from_user(oij, req))
 268		return -EFAULT;
 269
 270	oij.ij_journal_size = i_size_read(osb->journal->j_inode);
 271
 272	o2info_set_request_filled(&oij.ij_req);
 273
 274	if (o2info_to_user(oij, req))
 275		return -EFAULT;
 276
 277	return 0;
 278}
 279
 280static int ocfs2_info_scan_inode_alloc(struct ocfs2_super *osb,
 281				       struct inode *inode_alloc, u64 blkno,
 282				       struct ocfs2_info_freeinode *fi,
 283				       u32 slot)
 284{
 285	int status = 0, unlock = 0;
 286
 287	struct buffer_head *bh = NULL;
 288	struct ocfs2_dinode *dinode_alloc = NULL;
 289
 290	if (inode_alloc)
 291		inode_lock(inode_alloc);
 292
 293	if (o2info_coherent(&fi->ifi_req)) {
 294		status = ocfs2_inode_lock(inode_alloc, &bh, 0);
 295		if (status < 0) {
 296			mlog_errno(status);
 297			goto bail;
 298		}
 299		unlock = 1;
 300	} else {
 301		status = ocfs2_read_blocks_sync(osb, blkno, 1, &bh);
 302		if (status < 0) {
 303			mlog_errno(status);
 304			goto bail;
 305		}
 306	}
 307
 308	dinode_alloc = (struct ocfs2_dinode *)bh->b_data;
 309
 310	fi->ifi_stat[slot].lfi_total =
 311		le32_to_cpu(dinode_alloc->id1.bitmap1.i_total);
 312	fi->ifi_stat[slot].lfi_free =
 313		le32_to_cpu(dinode_alloc->id1.bitmap1.i_total) -
 314		le32_to_cpu(dinode_alloc->id1.bitmap1.i_used);
 315
 316bail:
 317	if (unlock)
 318		ocfs2_inode_unlock(inode_alloc, 0);
 319
 320	if (inode_alloc)
 321		inode_unlock(inode_alloc);
 322
 323	brelse(bh);
 324
 325	return status;
 326}
 327
 328static int ocfs2_info_handle_freeinode(struct inode *inode,
 329				       struct ocfs2_info_request __user *req)
 330{
 331	u32 i;
 332	u64 blkno = -1;
 333	char namebuf[40];
 334	int status, type = INODE_ALLOC_SYSTEM_INODE;
 335	struct ocfs2_info_freeinode *oifi = NULL;
 336	struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
 337	struct inode *inode_alloc = NULL;
 338
 339	oifi = kzalloc(sizeof(struct ocfs2_info_freeinode), GFP_KERNEL);
 340	if (!oifi) {
 341		status = -ENOMEM;
 342		mlog_errno(status);
 343		goto out_err;
 344	}
 345
 346	if (o2info_from_user(*oifi, req)) {
 347		status = -EFAULT;
 348		goto out_free;
 349	}
 350
 351	oifi->ifi_slotnum = osb->max_slots;
 352
 353	for (i = 0; i < oifi->ifi_slotnum; i++) {
 354		if (o2info_coherent(&oifi->ifi_req)) {
 355			inode_alloc = ocfs2_get_system_file_inode(osb, type, i);
 356			if (!inode_alloc) {
 357				mlog(ML_ERROR, "unable to get alloc inode in "
 358				    "slot %u\n", i);
 359				status = -EIO;
 360				goto bail;
 361			}
 362		} else {
 363			ocfs2_sprintf_system_inode_name(namebuf,
 364							sizeof(namebuf),
 365							type, i);
 366			status = ocfs2_lookup_ino_from_name(osb->sys_root_inode,
 367							    namebuf,
 368							    strlen(namebuf),
 369							    &blkno);
 370			if (status < 0) {
 371				status = -ENOENT;
 372				goto bail;
 373			}
 374		}
 375
 376		status = ocfs2_info_scan_inode_alloc(osb, inode_alloc, blkno, oifi, i);
 377
 378		iput(inode_alloc);
 379		inode_alloc = NULL;
 380
 381		if (status < 0)
 382			goto bail;
 383	}
 384
 385	o2info_set_request_filled(&oifi->ifi_req);
 386
 387	if (o2info_to_user(*oifi, req)) {
 388		status = -EFAULT;
 389		goto out_free;
 390	}
 391
 392	status = 0;
 393bail:
 394	if (status)
 395		o2info_set_request_error(&oifi->ifi_req, req);
 396out_free:
 397	kfree(oifi);
 398out_err:
 399	return status;
 400}
 401
 402static void o2ffg_update_histogram(struct ocfs2_info_free_chunk_list *hist,
 403				   unsigned int chunksize)
 404{
 405	int index;
 406
 407	index = __ilog2_u32(chunksize);
 408	if (index >= OCFS2_INFO_MAX_HIST)
 409		index = OCFS2_INFO_MAX_HIST - 1;
 410
 411	hist->fc_chunks[index]++;
 412	hist->fc_clusters[index] += chunksize;
 413}
 414
 415static void o2ffg_update_stats(struct ocfs2_info_freefrag_stats *stats,
 416			       unsigned int chunksize)
 417{
 418	if (chunksize > stats->ffs_max)
 419		stats->ffs_max = chunksize;
 420
 421	if (chunksize < stats->ffs_min)
 422		stats->ffs_min = chunksize;
 423
 424	stats->ffs_avg += chunksize;
 425	stats->ffs_free_chunks_real++;
 426}
 427
 428static void ocfs2_info_update_ffg(struct ocfs2_info_freefrag *ffg,
 429				  unsigned int chunksize)
 430{
 431	o2ffg_update_histogram(&(ffg->iff_ffs.ffs_fc_hist), chunksize);
 432	o2ffg_update_stats(&(ffg->iff_ffs), chunksize);
 433}
 434
 435static int ocfs2_info_freefrag_scan_chain(struct ocfs2_super *osb,
 436					  struct inode *gb_inode,
 437					  struct ocfs2_dinode *gb_dinode,
 438					  struct ocfs2_chain_rec *rec,
 439					  struct ocfs2_info_freefrag *ffg,
 440					  u32 chunks_in_group)
 441{
 442	int status = 0, used;
 443	u64 blkno;
 444
 445	struct buffer_head *bh = NULL;
 446	struct ocfs2_group_desc *bg = NULL;
 447
 448	unsigned int max_bits, num_clusters;
 449	unsigned int offset = 0, cluster, chunk;
 450	unsigned int chunk_free, last_chunksize = 0;
 451
 452	if (!le32_to_cpu(rec->c_free))
 453		goto bail;
 454
 455	do {
 456		if (!bg)
 457			blkno = le64_to_cpu(rec->c_blkno);
 458		else
 459			blkno = le64_to_cpu(bg->bg_next_group);
 460
 461		if (bh) {
 462			brelse(bh);
 463			bh = NULL;
 464		}
 465
 466		if (o2info_coherent(&ffg->iff_req))
 467			status = ocfs2_read_group_descriptor(gb_inode,
 468							     gb_dinode,
 469							     blkno, &bh);
 470		else
 471			status = ocfs2_read_blocks_sync(osb, blkno, 1, &bh);
 472
 473		if (status < 0) {
 474			mlog(ML_ERROR, "Can't read the group descriptor # "
 475			     "%llu from device.", (unsigned long long)blkno);
 476			status = -EIO;
 477			goto bail;
 478		}
 479
 480		bg = (struct ocfs2_group_desc *)bh->b_data;
 481
 482		if (!le16_to_cpu(bg->bg_free_bits_count))
 483			continue;
 484
 485		max_bits = le16_to_cpu(bg->bg_bits);
 486		offset = 0;
 487
 488		for (chunk = 0; chunk < chunks_in_group; chunk++) {
 489			/*
 490			 * last chunk may be not an entire one.
 491			 */
 492			if ((offset + ffg->iff_chunksize) > max_bits)
 493				num_clusters = max_bits - offset;
 494			else
 495				num_clusters = ffg->iff_chunksize;
 496
 497			chunk_free = 0;
 498			for (cluster = 0; cluster < num_clusters; cluster++) {
 499				used = ocfs2_test_bit(offset,
 500						(unsigned long *)bg->bg_bitmap);
 501				/*
 502				 * - chunk_free counts free clusters in #N chunk.
 503				 * - last_chunksize records the size(in) clusters
 504				 *   for the last real free chunk being counted.
 505				 */
 506				if (!used) {
 507					last_chunksize++;
 508					chunk_free++;
 509				}
 510
 511				if (used && last_chunksize) {
 512					ocfs2_info_update_ffg(ffg,
 513							      last_chunksize);
 514					last_chunksize = 0;
 515				}
 516
 517				offset++;
 518			}
 519
 520			if (chunk_free == ffg->iff_chunksize)
 521				ffg->iff_ffs.ffs_free_chunks++;
 522		}
 523
 524		/*
 525		 * need to update the info for last free chunk.
 526		 */
 527		if (last_chunksize)
 528			ocfs2_info_update_ffg(ffg, last_chunksize);
 529
 530	} while (le64_to_cpu(bg->bg_next_group));
 531
 532bail:
 533	brelse(bh);
 534
 535	return status;
 536}
 537
 538static int ocfs2_info_freefrag_scan_bitmap(struct ocfs2_super *osb,
 539					   struct inode *gb_inode, u64 blkno,
 540					   struct ocfs2_info_freefrag *ffg)
 541{
 542	u32 chunks_in_group;
 543	int status = 0, unlock = 0, i;
 544
 545	struct buffer_head *bh = NULL;
 546	struct ocfs2_chain_list *cl = NULL;
 547	struct ocfs2_chain_rec *rec = NULL;
 548	struct ocfs2_dinode *gb_dinode = NULL;
 549
 550	if (gb_inode)
 551		inode_lock(gb_inode);
 552
 553	if (o2info_coherent(&ffg->iff_req)) {
 554		status = ocfs2_inode_lock(gb_inode, &bh, 0);
 555		if (status < 0) {
 556			mlog_errno(status);
 557			goto bail;
 558		}
 559		unlock = 1;
 560	} else {
 561		status = ocfs2_read_blocks_sync(osb, blkno, 1, &bh);
 562		if (status < 0) {
 563			mlog_errno(status);
 564			goto bail;
 565		}
 566	}
 567
 568	gb_dinode = (struct ocfs2_dinode *)bh->b_data;
 569	cl = &(gb_dinode->id2.i_chain);
 570
 571	/*
 572	 * Chunksize(in) clusters from userspace should be
 573	 * less than clusters in a group.
 574	 */
 575	if (ffg->iff_chunksize > le16_to_cpu(cl->cl_cpg)) {
 576		status = -EINVAL;
 577		goto bail;
 578	}
 579
 580	memset(&ffg->iff_ffs, 0, sizeof(struct ocfs2_info_freefrag_stats));
 581
 582	ffg->iff_ffs.ffs_min = ~0U;
 583	ffg->iff_ffs.ffs_clusters =
 584			le32_to_cpu(gb_dinode->id1.bitmap1.i_total);
 585	ffg->iff_ffs.ffs_free_clusters = ffg->iff_ffs.ffs_clusters -
 586			le32_to_cpu(gb_dinode->id1.bitmap1.i_used);
 587
 588	chunks_in_group = le16_to_cpu(cl->cl_cpg) / ffg->iff_chunksize + 1;
 589
 590	for (i = 0; i < le16_to_cpu(cl->cl_next_free_rec); i++) {
 591		rec = &(cl->cl_recs[i]);
 592		status = ocfs2_info_freefrag_scan_chain(osb, gb_inode,
 593							gb_dinode,
 594							rec, ffg,
 595							chunks_in_group);
 596		if (status)
 597			goto bail;
 598	}
 599
 600	if (ffg->iff_ffs.ffs_free_chunks_real)
 601		ffg->iff_ffs.ffs_avg = (ffg->iff_ffs.ffs_avg /
 602					ffg->iff_ffs.ffs_free_chunks_real);
 603bail:
 604	if (unlock)
 605		ocfs2_inode_unlock(gb_inode, 0);
 606
 607	if (gb_inode)
 608		inode_unlock(gb_inode);
 609
 610	iput(gb_inode);
 611	brelse(bh);
 612
 613	return status;
 614}
 615
 616static int ocfs2_info_handle_freefrag(struct inode *inode,
 617				      struct ocfs2_info_request __user *req)
 618{
 619	u64 blkno = -1;
 620	char namebuf[40];
 621	int status, type = GLOBAL_BITMAP_SYSTEM_INODE;
 622
 623	struct ocfs2_info_freefrag *oiff;
 624	struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
 625	struct inode *gb_inode = NULL;
 626
 627	oiff = kzalloc(sizeof(struct ocfs2_info_freefrag), GFP_KERNEL);
 628	if (!oiff) {
 629		status = -ENOMEM;
 630		mlog_errno(status);
 631		goto out_err;
 632	}
 633
 634	if (o2info_from_user(*oiff, req)) {
 635		status = -EFAULT;
 636		goto out_free;
 637	}
 638	/*
 639	 * chunksize from userspace should be power of 2.
 640	 */
 641	if ((oiff->iff_chunksize & (oiff->iff_chunksize - 1)) ||
 642	    (!oiff->iff_chunksize)) {
 643		status = -EINVAL;
 644		goto bail;
 645	}
 646
 647	if (o2info_coherent(&oiff->iff_req)) {
 648		gb_inode = ocfs2_get_system_file_inode(osb, type,
 649						       OCFS2_INVALID_SLOT);
 650		if (!gb_inode) {
 651			mlog(ML_ERROR, "unable to get global_bitmap inode\n");
 652			status = -EIO;
 653			goto bail;
 654		}
 655	} else {
 656		ocfs2_sprintf_system_inode_name(namebuf, sizeof(namebuf), type,
 657						OCFS2_INVALID_SLOT);
 658		status = ocfs2_lookup_ino_from_name(osb->sys_root_inode,
 659						    namebuf,
 660						    strlen(namebuf),
 661						    &blkno);
 662		if (status < 0) {
 663			status = -ENOENT;
 664			goto bail;
 665		}
 666	}
 667
 668	status = ocfs2_info_freefrag_scan_bitmap(osb, gb_inode, blkno, oiff);
 669	if (status < 0)
 670		goto bail;
 671
 672	o2info_set_request_filled(&oiff->iff_req);
 673
 674	if (o2info_to_user(*oiff, req)) {
 675		status = -EFAULT;
 676		goto out_free;
 677	}
 678
 679	status = 0;
 680bail:
 681	if (status)
 682		o2info_set_request_error(&oiff->iff_req, req);
 683out_free:
 684	kfree(oiff);
 685out_err:
 686	return status;
 687}
 688
 689static int ocfs2_info_handle_unknown(struct inode *inode,
 690				     struct ocfs2_info_request __user *req)
 691{
 692	struct ocfs2_info_request oir;
 693
 694	if (o2info_from_user(oir, req))
 695		return -EFAULT;
 696
 697	o2info_clear_request_filled(&oir);
 698
 699	if (o2info_to_user(oir, req))
 700		return -EFAULT;
 701
 702	return 0;
 703}
 704
 705/*
 706 * Validate and distinguish OCFS2_IOC_INFO requests.
 707 *
 708 * - validate the magic number.
 709 * - distinguish different requests.
 710 * - validate size of different requests.
 711 */
 712static int ocfs2_info_handle_request(struct inode *inode,
 713				     struct ocfs2_info_request __user *req)
 714{
 715	int status = -EFAULT;
 716	struct ocfs2_info_request oir;
 717
 718	if (o2info_from_user(oir, req))
 719		goto bail;
 720
 721	status = -EINVAL;
 722	if (oir.ir_magic != OCFS2_INFO_MAGIC)
 723		goto bail;
 724
 725	switch (oir.ir_code) {
 726	case OCFS2_INFO_BLOCKSIZE:
 727		if (oir.ir_size == sizeof(struct ocfs2_info_blocksize))
 728			status = ocfs2_info_handle_blocksize(inode, req);
 729		break;
 730	case OCFS2_INFO_CLUSTERSIZE:
 731		if (oir.ir_size == sizeof(struct ocfs2_info_clustersize))
 732			status = ocfs2_info_handle_clustersize(inode, req);
 733		break;
 734	case OCFS2_INFO_MAXSLOTS:
 735		if (oir.ir_size == sizeof(struct ocfs2_info_maxslots))
 736			status = ocfs2_info_handle_maxslots(inode, req);
 737		break;
 738	case OCFS2_INFO_LABEL:
 739		if (oir.ir_size == sizeof(struct ocfs2_info_label))
 740			status = ocfs2_info_handle_label(inode, req);
 741		break;
 742	case OCFS2_INFO_UUID:
 743		if (oir.ir_size == sizeof(struct ocfs2_info_uuid))
 744			status = ocfs2_info_handle_uuid(inode, req);
 745		break;
 746	case OCFS2_INFO_FS_FEATURES:
 747		if (oir.ir_size == sizeof(struct ocfs2_info_fs_features))
 748			status = ocfs2_info_handle_fs_features(inode, req);
 749		break;
 750	case OCFS2_INFO_JOURNAL_SIZE:
 751		if (oir.ir_size == sizeof(struct ocfs2_info_journal_size))
 752			status = ocfs2_info_handle_journal_size(inode, req);
 753		break;
 754	case OCFS2_INFO_FREEINODE:
 755		if (oir.ir_size == sizeof(struct ocfs2_info_freeinode))
 756			status = ocfs2_info_handle_freeinode(inode, req);
 757		break;
 758	case OCFS2_INFO_FREEFRAG:
 759		if (oir.ir_size == sizeof(struct ocfs2_info_freefrag))
 760			status = ocfs2_info_handle_freefrag(inode, req);
 761		break;
 762	default:
 763		status = ocfs2_info_handle_unknown(inode, req);
 764		break;
 765	}
 766
 767bail:
 768	return status;
 769}
 770
 771static int ocfs2_get_request_ptr(struct ocfs2_info *info, int idx,
 772				 u64 *req_addr, int compat_flag)
 773{
 774	int status = -EFAULT;
 775	u64 __user *bp = NULL;
 776
 777	if (compat_flag) {
 778#ifdef CONFIG_COMPAT
 779		/*
 780		 * pointer bp stores the base address of a pointers array,
 781		 * which collects all addresses of separate request.
 782		 */
 783		bp = (u64 __user *)(unsigned long)compat_ptr(info->oi_requests);
 784#else
 785		BUG();
 786#endif
 787	} else
 788		bp = (u64 __user *)(unsigned long)(info->oi_requests);
 789
 790	if (o2info_from_user(*req_addr, bp + idx))
 791		goto bail;
 792
 793	status = 0;
 794bail:
 795	return status;
 796}
 797
 798/*
 799 * OCFS2_IOC_INFO handles an array of requests passed from userspace.
 800 *
 801 * ocfs2_info_handle() recevies a large info aggregation, grab and
 802 * validate the request count from header, then break it into small
 803 * pieces, later specific handlers can handle them one by one.
 804 *
 805 * Idea here is to make each separate request small enough to ensure
 806 * a better backward&forward compatibility, since a small piece of
 807 * request will be less likely to be broken if disk layout get changed.
 808 */
 809static int ocfs2_info_handle(struct inode *inode, struct ocfs2_info *info,
 810			     int compat_flag)
 811{
 812	int i, status = 0;
 813	u64 req_addr;
 814	struct ocfs2_info_request __user *reqp;
 815
 816	if ((info->oi_count > OCFS2_INFO_MAX_REQUEST) ||
 817	    (!info->oi_requests)) {
 818		status = -EINVAL;
 819		goto bail;
 820	}
 821
 822	for (i = 0; i < info->oi_count; i++) {
 823
 824		status = ocfs2_get_request_ptr(info, i, &req_addr, compat_flag);
 825		if (status)
 826			break;
 827
 828		reqp = (struct ocfs2_info_request __user *)(unsigned long)req_addr;
 829		if (!reqp) {
 830			status = -EINVAL;
 831			goto bail;
 832		}
 833
 834		status = ocfs2_info_handle_request(inode, reqp);
 835		if (status)
 836			break;
 837	}
 838
 839bail:
 840	return status;
 841}
 842
 843long ocfs2_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
 844{
 845	struct inode *inode = file_inode(filp);
 846	unsigned int flags;
 847	int new_clusters;
 848	int status;
 849	struct ocfs2_space_resv sr;
 850	struct ocfs2_new_group_input input;
 851	struct reflink_arguments args;
 852	const char __user *old_path;
 853	const char __user *new_path;
 854	bool preserve;
 855	struct ocfs2_info info;
 856	void __user *argp = (void __user *)arg;
 857
 858	switch (cmd) {
 859	case OCFS2_IOC_GETFLAGS:
 860		status = ocfs2_get_inode_attr(inode, &flags);
 861		if (status < 0)
 862			return status;
 863
 864		flags &= OCFS2_FL_VISIBLE;
 865		return put_user(flags, (int __user *) arg);
 866	case OCFS2_IOC_SETFLAGS:
 867		if (get_user(flags, (int __user *) arg))
 868			return -EFAULT;
 869
 870		status = mnt_want_write_file(filp);
 871		if (status)
 872			return status;
 873		status = ocfs2_set_inode_attr(inode, flags,
 874			OCFS2_FL_MODIFIABLE);
 875		mnt_drop_write_file(filp);
 876		return status;
 877	case OCFS2_IOC_RESVSP:
 878	case OCFS2_IOC_RESVSP64:
 879	case OCFS2_IOC_UNRESVSP:
 880	case OCFS2_IOC_UNRESVSP64:
 881		if (copy_from_user(&sr, (int __user *) arg, sizeof(sr)))
 882			return -EFAULT;
 883
 884		return ocfs2_change_file_space(filp, cmd, &sr);
 885	case OCFS2_IOC_GROUP_EXTEND:
 886		if (!capable(CAP_SYS_RESOURCE))
 887			return -EPERM;
 888
 889		if (get_user(new_clusters, (int __user *)arg))
 890			return -EFAULT;
 891
 892		status = mnt_want_write_file(filp);
 893		if (status)
 894			return status;
 895		status = ocfs2_group_extend(inode, new_clusters);
 896		mnt_drop_write_file(filp);
 897		return status;
 898	case OCFS2_IOC_GROUP_ADD:
 899	case OCFS2_IOC_GROUP_ADD64:
 900		if (!capable(CAP_SYS_RESOURCE))
 901			return -EPERM;
 902
 903		if (copy_from_user(&input, (int __user *) arg, sizeof(input)))
 904			return -EFAULT;
 905
 906		status = mnt_want_write_file(filp);
 907		if (status)
 908			return status;
 909		status = ocfs2_group_add(inode, &input);
 910		mnt_drop_write_file(filp);
 911		return status;
 912	case OCFS2_IOC_REFLINK:
 913		if (copy_from_user(&args, argp, sizeof(args)))
 914			return -EFAULT;
 915		old_path = (const char __user *)(unsigned long)args.old_path;
 916		new_path = (const char __user *)(unsigned long)args.new_path;
 917		preserve = (args.preserve != 0);
 918
 919		return ocfs2_reflink_ioctl(inode, old_path, new_path, preserve);
 920	case OCFS2_IOC_INFO:
 921		if (copy_from_user(&info, argp, sizeof(struct ocfs2_info)))
 922			return -EFAULT;
 923
 924		return ocfs2_info_handle(inode, &info, 0);
 925	case FITRIM:
 926	{
 927		struct super_block *sb = inode->i_sb;
 928		struct request_queue *q = bdev_get_queue(sb->s_bdev);
 929		struct fstrim_range range;
 930		int ret = 0;
 931
 932		if (!capable(CAP_SYS_ADMIN))
 933			return -EPERM;
 934
 935		if (!blk_queue_discard(q))
 936			return -EOPNOTSUPP;
 937
 938		if (copy_from_user(&range, argp, sizeof(range)))
 939			return -EFAULT;
 940
 941		range.minlen = max_t(u64, q->limits.discard_granularity,
 942				     range.minlen);
 943		ret = ocfs2_trim_fs(sb, &range);
 944		if (ret < 0)
 945			return ret;
 946
 947		if (copy_to_user(argp, &range, sizeof(range)))
 948			return -EFAULT;
 949
 950		return 0;
 951	}
 952	case OCFS2_IOC_MOVE_EXT:
 953		return ocfs2_ioctl_move_extents(filp, argp);
 954	default:
 955		return -ENOTTY;
 956	}
 957}
 958
 959#ifdef CONFIG_COMPAT
 960long ocfs2_compat_ioctl(struct file *file, unsigned cmd, unsigned long arg)
 961{
 962	bool preserve;
 963	struct reflink_arguments args;
 964	struct inode *inode = file_inode(file);
 965	struct ocfs2_info info;
 966	void __user *argp = (void __user *)arg;
 967
 968	switch (cmd) {
 969	case OCFS2_IOC32_GETFLAGS:
 970		cmd = OCFS2_IOC_GETFLAGS;
 971		break;
 972	case OCFS2_IOC32_SETFLAGS:
 973		cmd = OCFS2_IOC_SETFLAGS;
 974		break;
 975	case OCFS2_IOC_RESVSP:
 976	case OCFS2_IOC_RESVSP64:
 977	case OCFS2_IOC_UNRESVSP:
 978	case OCFS2_IOC_UNRESVSP64:
 979	case OCFS2_IOC_GROUP_EXTEND:
 980	case OCFS2_IOC_GROUP_ADD:
 981	case OCFS2_IOC_GROUP_ADD64:
 982		break;
 983	case OCFS2_IOC_REFLINK:
 984		if (copy_from_user(&args, argp, sizeof(args)))
 985			return -EFAULT;
 986		preserve = (args.preserve != 0);
 987
 988		return ocfs2_reflink_ioctl(inode, compat_ptr(args.old_path),
 989					   compat_ptr(args.new_path), preserve);
 990	case OCFS2_IOC_INFO:
 991		if (copy_from_user(&info, argp, sizeof(struct ocfs2_info)))
 992			return -EFAULT;
 993
 994		return ocfs2_info_handle(inode, &info, 1);
 
 995	case OCFS2_IOC_MOVE_EXT:
 996		break;
 997	default:
 998		return -ENOIOCTLCMD;
 999	}
1000
1001	return ocfs2_ioctl(file, cmd, arg);
1002}
1003#endif