1 // SPDX-License-Identifier: GPL-2.0
3 * linux/fs/ocfs2/ioctl.c
5 * Copyright (C) 2006 Herbert Poetzl
6 * adapted from Remy Card's ext2/ioctl.c
10 #include <linux/mount.h>
11 #include <linux/blkdev.h>
12 #include <linux/compat.h>
13 #include <linux/fileattr.h>
15 #include <cluster/masklog.h>
27 #include "refcounttree.h"
30 #include "buffer_head_io.h"
32 #include "move_extents.h"
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))
40 * This is just a best-effort to tell userspace that this request
43 static inline void o2info_set_request_error(struct ocfs2_info_request *kreq,
44 struct ocfs2_info_request __user *req)
46 kreq->ir_flags |= OCFS2_INFO_FL_ERROR;
47 (void)put_user(kreq->ir_flags, (__u32 __user *)&(req->ir_flags));
50 static inline void o2info_set_request_filled(struct ocfs2_info_request *req)
52 req->ir_flags |= OCFS2_INFO_FL_FILLED;
55 static inline void o2info_clear_request_filled(struct ocfs2_info_request *req)
57 req->ir_flags &= ~OCFS2_INFO_FL_FILLED;
60 static inline int o2info_coherent(struct ocfs2_info_request *req)
62 return (!(req->ir_flags & OCFS2_INFO_FL_NON_COHERENT));
65 int ocfs2_fileattr_get(struct dentry *dentry, struct fileattr *fa)
67 struct inode *inode = d_inode(dentry);
71 status = ocfs2_inode_lock(inode, NULL, 0);
76 ocfs2_get_inode_flags(OCFS2_I(inode));
77 flags = OCFS2_I(inode)->ip_attr;
78 ocfs2_inode_unlock(inode, 0);
80 fileattr_fill_flags(fa, flags & OCFS2_FL_VISIBLE);
85 int ocfs2_fileattr_set(struct mnt_idmap *idmap,
86 struct dentry *dentry, struct fileattr *fa)
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;
97 if (fileattr_has_fsx(fa))
100 status = ocfs2_inode_lock(inode, &bh, 1);
106 if (!S_ISDIR(inode->i_mode))
107 flags &= ~OCFS2_DIRSYNC_FL;
109 oldflags = ocfs2_inode->ip_attr;
110 flags = flags & OCFS2_FL_MODIFIABLE;
111 flags |= oldflags & ~OCFS2_FL_MODIFIABLE;
113 /* Check already done by VFS, but repeat with ocfs lock */
115 if ((flags ^ oldflags) & (FS_APPEND_FL | FS_IMMUTABLE_FL) &&
116 !capable(CAP_LINUX_IMMUTABLE))
119 handle = ocfs2_start_trans(osb, OCFS2_INODE_UPDATE_CREDITS);
120 if (IS_ERR(handle)) {
121 status = PTR_ERR(handle);
126 ocfs2_inode->ip_attr = flags;
127 ocfs2_set_inode_flags(inode);
128 inode_set_ctime_current(inode);
130 status = ocfs2_mark_inode_dirty(handle, inode, bh);
134 ocfs2_commit_trans(osb, handle);
137 ocfs2_inode_unlock(inode, 1);
144 static int ocfs2_info_handle_blocksize(struct inode *inode,
145 struct ocfs2_info_request __user *req)
147 struct ocfs2_info_blocksize oib;
149 if (o2info_from_user(oib, req))
152 oib.ib_blocksize = inode->i_sb->s_blocksize;
154 o2info_set_request_filled(&oib.ib_req);
156 if (o2info_to_user(oib, req))
162 static int ocfs2_info_handle_clustersize(struct inode *inode,
163 struct ocfs2_info_request __user *req)
165 struct ocfs2_info_clustersize oic;
166 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
168 if (o2info_from_user(oic, req))
171 oic.ic_clustersize = osb->s_clustersize;
173 o2info_set_request_filled(&oic.ic_req);
175 if (o2info_to_user(oic, req))
181 static int ocfs2_info_handle_maxslots(struct inode *inode,
182 struct ocfs2_info_request __user *req)
184 struct ocfs2_info_maxslots oim;
185 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
187 if (o2info_from_user(oim, req))
190 oim.im_max_slots = osb->max_slots;
192 o2info_set_request_filled(&oim.im_req);
194 if (o2info_to_user(oim, req))
200 static int ocfs2_info_handle_label(struct inode *inode,
201 struct ocfs2_info_request __user *req)
203 struct ocfs2_info_label oil;
204 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
206 if (o2info_from_user(oil, req))
209 memcpy(oil.il_label, osb->vol_label, OCFS2_MAX_VOL_LABEL_LEN);
211 o2info_set_request_filled(&oil.il_req);
213 if (o2info_to_user(oil, req))
219 static int ocfs2_info_handle_uuid(struct inode *inode,
220 struct ocfs2_info_request __user *req)
222 struct ocfs2_info_uuid oiu;
223 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
225 if (o2info_from_user(oiu, req))
228 memcpy(oiu.iu_uuid_str, osb->uuid_str, OCFS2_TEXT_UUID_LEN + 1);
230 o2info_set_request_filled(&oiu.iu_req);
232 if (o2info_to_user(oiu, req))
238 static int ocfs2_info_handle_fs_features(struct inode *inode,
239 struct ocfs2_info_request __user *req)
241 struct ocfs2_info_fs_features oif;
242 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
244 if (o2info_from_user(oif, req))
247 oif.if_compat_features = osb->s_feature_compat;
248 oif.if_incompat_features = osb->s_feature_incompat;
249 oif.if_ro_compat_features = osb->s_feature_ro_compat;
251 o2info_set_request_filled(&oif.if_req);
253 if (o2info_to_user(oif, req))
259 static int ocfs2_info_handle_journal_size(struct inode *inode,
260 struct ocfs2_info_request __user *req)
262 struct ocfs2_info_journal_size oij;
263 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
265 if (o2info_from_user(oij, req))
268 oij.ij_journal_size = i_size_read(osb->journal->j_inode);
270 o2info_set_request_filled(&oij.ij_req);
272 if (o2info_to_user(oij, req))
278 static int ocfs2_info_scan_inode_alloc(struct ocfs2_super *osb,
279 struct inode *inode_alloc, u64 blkno,
280 struct ocfs2_info_freeinode *fi,
283 int status = 0, unlock = 0;
285 struct buffer_head *bh = NULL;
286 struct ocfs2_dinode *dinode_alloc = NULL;
289 inode_lock(inode_alloc);
291 if (inode_alloc && o2info_coherent(&fi->ifi_req)) {
292 status = ocfs2_inode_lock(inode_alloc, &bh, 0);
299 status = ocfs2_read_blocks_sync(osb, blkno, 1, &bh);
306 dinode_alloc = (struct ocfs2_dinode *)bh->b_data;
308 fi->ifi_stat[slot].lfi_total =
309 le32_to_cpu(dinode_alloc->id1.bitmap1.i_total);
310 fi->ifi_stat[slot].lfi_free =
311 le32_to_cpu(dinode_alloc->id1.bitmap1.i_total) -
312 le32_to_cpu(dinode_alloc->id1.bitmap1.i_used);
316 ocfs2_inode_unlock(inode_alloc, 0);
319 inode_unlock(inode_alloc);
326 static int ocfs2_info_handle_freeinode(struct inode *inode,
327 struct ocfs2_info_request __user *req)
332 int status, type = INODE_ALLOC_SYSTEM_INODE;
333 struct ocfs2_info_freeinode *oifi = NULL;
334 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
335 struct inode *inode_alloc = NULL;
337 oifi = kzalloc(sizeof(struct ocfs2_info_freeinode), GFP_KERNEL);
344 if (o2info_from_user(*oifi, req)) {
349 oifi->ifi_slotnum = osb->max_slots;
351 for (i = 0; i < oifi->ifi_slotnum; i++) {
352 if (o2info_coherent(&oifi->ifi_req)) {
353 inode_alloc = ocfs2_get_system_file_inode(osb, type, i);
355 mlog(ML_ERROR, "unable to get alloc inode in "
361 ocfs2_sprintf_system_inode_name(namebuf,
364 status = ocfs2_lookup_ino_from_name(osb->sys_root_inode,
374 status = ocfs2_info_scan_inode_alloc(osb, inode_alloc, blkno, oifi, i);
383 o2info_set_request_filled(&oifi->ifi_req);
385 if (o2info_to_user(*oifi, req)) {
393 o2info_set_request_error(&oifi->ifi_req, req);
400 static void o2ffg_update_histogram(struct ocfs2_info_free_chunk_list *hist,
401 unsigned int chunksize)
405 index = __ilog2_u32(chunksize);
406 if (index >= OCFS2_INFO_MAX_HIST)
407 index = OCFS2_INFO_MAX_HIST - 1;
409 hist->fc_chunks[index]++;
410 hist->fc_clusters[index] += chunksize;
413 static void o2ffg_update_stats(struct ocfs2_info_freefrag_stats *stats,
414 unsigned int chunksize)
416 if (chunksize > stats->ffs_max)
417 stats->ffs_max = chunksize;
419 if (chunksize < stats->ffs_min)
420 stats->ffs_min = chunksize;
422 stats->ffs_avg += chunksize;
423 stats->ffs_free_chunks_real++;
426 static void ocfs2_info_update_ffg(struct ocfs2_info_freefrag *ffg,
427 unsigned int chunksize)
429 o2ffg_update_histogram(&(ffg->iff_ffs.ffs_fc_hist), chunksize);
430 o2ffg_update_stats(&(ffg->iff_ffs), chunksize);
433 static int ocfs2_info_freefrag_scan_chain(struct ocfs2_super *osb,
434 struct inode *gb_inode,
435 struct ocfs2_dinode *gb_dinode,
436 struct ocfs2_chain_rec *rec,
437 struct ocfs2_info_freefrag *ffg,
440 int status = 0, used;
443 struct buffer_head *bh = NULL;
444 struct ocfs2_group_desc *bg = NULL;
446 unsigned int max_bits, num_clusters;
447 unsigned int offset = 0, cluster, chunk;
448 unsigned int chunk_free, last_chunksize = 0;
450 if (!le32_to_cpu(rec->c_free))
455 blkno = le64_to_cpu(rec->c_blkno);
457 blkno = le64_to_cpu(bg->bg_next_group);
464 if (o2info_coherent(&ffg->iff_req))
465 status = ocfs2_read_group_descriptor(gb_inode,
469 status = ocfs2_read_blocks_sync(osb, blkno, 1, &bh);
472 mlog(ML_ERROR, "Can't read the group descriptor # "
473 "%llu from device.", (unsigned long long)blkno);
478 bg = (struct ocfs2_group_desc *)bh->b_data;
480 if (!le16_to_cpu(bg->bg_free_bits_count))
483 max_bits = le16_to_cpu(bg->bg_bits);
486 for (chunk = 0; chunk < chunks_in_group; chunk++) {
488 * last chunk may be not an entire one.
490 if ((offset + ffg->iff_chunksize) > max_bits)
491 num_clusters = max_bits - offset;
493 num_clusters = ffg->iff_chunksize;
496 for (cluster = 0; cluster < num_clusters; cluster++) {
497 used = ocfs2_test_bit(offset,
498 (unsigned long *)bg->bg_bitmap);
500 * - chunk_free counts free clusters in #N chunk.
501 * - last_chunksize records the size(in) clusters
502 * for the last real free chunk being counted.
509 if (used && last_chunksize) {
510 ocfs2_info_update_ffg(ffg,
518 if (chunk_free == ffg->iff_chunksize)
519 ffg->iff_ffs.ffs_free_chunks++;
523 * need to update the info for last free chunk.
526 ocfs2_info_update_ffg(ffg, last_chunksize);
528 } while (le64_to_cpu(bg->bg_next_group));
536 static int ocfs2_info_freefrag_scan_bitmap(struct ocfs2_super *osb,
537 struct inode *gb_inode, u64 blkno,
538 struct ocfs2_info_freefrag *ffg)
541 int status = 0, unlock = 0, i;
543 struct buffer_head *bh = NULL;
544 struct ocfs2_chain_list *cl = NULL;
545 struct ocfs2_chain_rec *rec = NULL;
546 struct ocfs2_dinode *gb_dinode = NULL;
549 inode_lock(gb_inode);
551 if (o2info_coherent(&ffg->iff_req)) {
552 status = ocfs2_inode_lock(gb_inode, &bh, 0);
559 status = ocfs2_read_blocks_sync(osb, blkno, 1, &bh);
566 gb_dinode = (struct ocfs2_dinode *)bh->b_data;
567 cl = &(gb_dinode->id2.i_chain);
570 * Chunksize(in) clusters from userspace should be
571 * less than clusters in a group.
573 if (ffg->iff_chunksize > le16_to_cpu(cl->cl_cpg)) {
578 memset(&ffg->iff_ffs, 0, sizeof(struct ocfs2_info_freefrag_stats));
580 ffg->iff_ffs.ffs_min = ~0U;
581 ffg->iff_ffs.ffs_clusters =
582 le32_to_cpu(gb_dinode->id1.bitmap1.i_total);
583 ffg->iff_ffs.ffs_free_clusters = ffg->iff_ffs.ffs_clusters -
584 le32_to_cpu(gb_dinode->id1.bitmap1.i_used);
586 chunks_in_group = le16_to_cpu(cl->cl_cpg) / ffg->iff_chunksize + 1;
588 for (i = 0; i < le16_to_cpu(cl->cl_next_free_rec); i++) {
589 rec = &(cl->cl_recs[i]);
590 status = ocfs2_info_freefrag_scan_chain(osb, gb_inode,
598 if (ffg->iff_ffs.ffs_free_chunks_real)
599 ffg->iff_ffs.ffs_avg = (ffg->iff_ffs.ffs_avg /
600 ffg->iff_ffs.ffs_free_chunks_real);
603 ocfs2_inode_unlock(gb_inode, 0);
606 inode_unlock(gb_inode);
614 static int ocfs2_info_handle_freefrag(struct inode *inode,
615 struct ocfs2_info_request __user *req)
619 int status, type = GLOBAL_BITMAP_SYSTEM_INODE;
621 struct ocfs2_info_freefrag *oiff;
622 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
623 struct inode *gb_inode = NULL;
625 oiff = kzalloc(sizeof(struct ocfs2_info_freefrag), GFP_KERNEL);
632 if (o2info_from_user(*oiff, req)) {
637 * chunksize from userspace should be power of 2.
639 if ((oiff->iff_chunksize & (oiff->iff_chunksize - 1)) ||
640 (!oiff->iff_chunksize)) {
645 if (o2info_coherent(&oiff->iff_req)) {
646 gb_inode = ocfs2_get_system_file_inode(osb, type,
649 mlog(ML_ERROR, "unable to get global_bitmap inode\n");
654 ocfs2_sprintf_system_inode_name(namebuf, sizeof(namebuf), type,
656 status = ocfs2_lookup_ino_from_name(osb->sys_root_inode,
666 status = ocfs2_info_freefrag_scan_bitmap(osb, gb_inode, blkno, oiff);
670 o2info_set_request_filled(&oiff->iff_req);
672 if (o2info_to_user(*oiff, req)) {
680 o2info_set_request_error(&oiff->iff_req, req);
687 static int ocfs2_info_handle_unknown(struct inode *inode,
688 struct ocfs2_info_request __user *req)
690 struct ocfs2_info_request oir;
692 if (o2info_from_user(oir, req))
695 o2info_clear_request_filled(&oir);
697 if (o2info_to_user(oir, req))
704 * Validate and distinguish OCFS2_IOC_INFO requests.
706 * - validate the magic number.
707 * - distinguish different requests.
708 * - validate size of different requests.
710 static int ocfs2_info_handle_request(struct inode *inode,
711 struct ocfs2_info_request __user *req)
713 int status = -EFAULT;
714 struct ocfs2_info_request oir;
716 if (o2info_from_user(oir, req))
720 if (oir.ir_magic != OCFS2_INFO_MAGIC)
723 switch (oir.ir_code) {
724 case OCFS2_INFO_BLOCKSIZE:
725 if (oir.ir_size == sizeof(struct ocfs2_info_blocksize))
726 status = ocfs2_info_handle_blocksize(inode, req);
728 case OCFS2_INFO_CLUSTERSIZE:
729 if (oir.ir_size == sizeof(struct ocfs2_info_clustersize))
730 status = ocfs2_info_handle_clustersize(inode, req);
732 case OCFS2_INFO_MAXSLOTS:
733 if (oir.ir_size == sizeof(struct ocfs2_info_maxslots))
734 status = ocfs2_info_handle_maxslots(inode, req);
736 case OCFS2_INFO_LABEL:
737 if (oir.ir_size == sizeof(struct ocfs2_info_label))
738 status = ocfs2_info_handle_label(inode, req);
740 case OCFS2_INFO_UUID:
741 if (oir.ir_size == sizeof(struct ocfs2_info_uuid))
742 status = ocfs2_info_handle_uuid(inode, req);
744 case OCFS2_INFO_FS_FEATURES:
745 if (oir.ir_size == sizeof(struct ocfs2_info_fs_features))
746 status = ocfs2_info_handle_fs_features(inode, req);
748 case OCFS2_INFO_JOURNAL_SIZE:
749 if (oir.ir_size == sizeof(struct ocfs2_info_journal_size))
750 status = ocfs2_info_handle_journal_size(inode, req);
752 case OCFS2_INFO_FREEINODE:
753 if (oir.ir_size == sizeof(struct ocfs2_info_freeinode))
754 status = ocfs2_info_handle_freeinode(inode, req);
756 case OCFS2_INFO_FREEFRAG:
757 if (oir.ir_size == sizeof(struct ocfs2_info_freefrag))
758 status = ocfs2_info_handle_freefrag(inode, req);
761 status = ocfs2_info_handle_unknown(inode, req);
769 static int ocfs2_get_request_ptr(struct ocfs2_info *info, int idx,
770 u64 *req_addr, int compat_flag)
772 int status = -EFAULT;
773 u64 __user *bp = NULL;
778 * pointer bp stores the base address of a pointers array,
779 * which collects all addresses of separate request.
781 bp = (u64 __user *)(unsigned long)compat_ptr(info->oi_requests);
786 bp = (u64 __user *)(unsigned long)(info->oi_requests);
788 if (o2info_from_user(*req_addr, bp + idx))
797 * OCFS2_IOC_INFO handles an array of requests passed from userspace.
799 * ocfs2_info_handle() recevies a large info aggregation, grab and
800 * validate the request count from header, then break it into small
801 * pieces, later specific handlers can handle them one by one.
803 * Idea here is to make each separate request small enough to ensure
804 * a better backward&forward compatibility, since a small piece of
805 * request will be less likely to be broken if disk layout get changed.
807 static noinline_for_stack int
808 ocfs2_info_handle(struct inode *inode, struct ocfs2_info *info, int compat_flag)
812 struct ocfs2_info_request __user *reqp;
814 if ((info->oi_count > OCFS2_INFO_MAX_REQUEST) ||
815 (!info->oi_requests)) {
820 for (i = 0; i < info->oi_count; i++) {
822 status = ocfs2_get_request_ptr(info, i, &req_addr, compat_flag);
826 reqp = (struct ocfs2_info_request __user *)(unsigned long)req_addr;
832 status = ocfs2_info_handle_request(inode, reqp);
841 long ocfs2_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
843 struct inode *inode = file_inode(filp);
844 void __user *argp = (void __user *)arg;
848 case OCFS2_IOC_RESVSP:
849 case OCFS2_IOC_RESVSP64:
850 case OCFS2_IOC_UNRESVSP:
851 case OCFS2_IOC_UNRESVSP64:
853 struct ocfs2_space_resv sr;
855 if (copy_from_user(&sr, (int __user *) arg, sizeof(sr)))
858 return ocfs2_change_file_space(filp, cmd, &sr);
860 case OCFS2_IOC_GROUP_EXTEND:
864 if (!capable(CAP_SYS_RESOURCE))
867 if (get_user(new_clusters, (int __user *)arg))
870 status = mnt_want_write_file(filp);
873 status = ocfs2_group_extend(inode, new_clusters);
874 mnt_drop_write_file(filp);
877 case OCFS2_IOC_GROUP_ADD:
878 case OCFS2_IOC_GROUP_ADD64:
880 struct ocfs2_new_group_input input;
882 if (!capable(CAP_SYS_RESOURCE))
885 if (copy_from_user(&input, (int __user *) arg, sizeof(input)))
888 status = mnt_want_write_file(filp);
891 status = ocfs2_group_add(inode, &input);
892 mnt_drop_write_file(filp);
895 case OCFS2_IOC_REFLINK:
897 struct reflink_arguments args;
898 const char __user *old_path;
899 const char __user *new_path;
902 if (copy_from_user(&args, argp, sizeof(args)))
904 old_path = (const char __user *)(unsigned long)args.old_path;
905 new_path = (const char __user *)(unsigned long)args.new_path;
906 preserve = (args.preserve != 0);
908 return ocfs2_reflink_ioctl(inode, old_path, new_path, preserve);
912 struct ocfs2_info info;
914 if (copy_from_user(&info, argp, sizeof(struct ocfs2_info)))
917 return ocfs2_info_handle(inode, &info, 0);
921 struct super_block *sb = inode->i_sb;
922 struct fstrim_range range;
925 if (!capable(CAP_SYS_ADMIN))
928 if (!bdev_max_discard_sectors(sb->s_bdev))
931 if (copy_from_user(&range, argp, sizeof(range)))
934 range.minlen = max_t(u64, bdev_discard_granularity(sb->s_bdev),
936 ret = ocfs2_trim_fs(sb, &range);
940 if (copy_to_user(argp, &range, sizeof(range)))
945 case OCFS2_IOC_MOVE_EXT:
946 return ocfs2_ioctl_move_extents(filp, argp);
953 long ocfs2_compat_ioctl(struct file *file, unsigned cmd, unsigned long arg)
956 struct reflink_arguments args;
957 struct inode *inode = file_inode(file);
958 struct ocfs2_info info;
959 void __user *argp = (void __user *)arg;
962 case OCFS2_IOC_RESVSP:
963 case OCFS2_IOC_RESVSP64:
964 case OCFS2_IOC_UNRESVSP:
965 case OCFS2_IOC_UNRESVSP64:
966 case OCFS2_IOC_GROUP_EXTEND:
967 case OCFS2_IOC_GROUP_ADD:
968 case OCFS2_IOC_GROUP_ADD64:
970 case OCFS2_IOC_REFLINK:
971 if (copy_from_user(&args, argp, sizeof(args)))
973 preserve = (args.preserve != 0);
975 return ocfs2_reflink_ioctl(inode, compat_ptr(args.old_path),
976 compat_ptr(args.new_path), preserve);
978 if (copy_from_user(&info, argp, sizeof(struct ocfs2_info)))
981 return ocfs2_info_handle(inode, &info, 1);
983 case OCFS2_IOC_MOVE_EXT:
989 return ocfs2_ioctl(file, cmd, arg);