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>
14 #include <cluster/masklog.h>
26 #include "refcounttree.h"
29 #include "buffer_head_io.h"
31 #include "move_extents.h"
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))
39 * This is just a best-effort to tell userspace that this request
42 static inline void o2info_set_request_error(struct ocfs2_info_request *kreq,
43 struct ocfs2_info_request __user *req)
45 kreq->ir_flags |= OCFS2_INFO_FL_ERROR;
46 (void)put_user(kreq->ir_flags, (__u32 __user *)&(req->ir_flags));
49 static inline void o2info_set_request_filled(struct ocfs2_info_request *req)
51 req->ir_flags |= OCFS2_INFO_FL_FILLED;
54 static inline void o2info_clear_request_filled(struct ocfs2_info_request *req)
56 req->ir_flags &= ~OCFS2_INFO_FL_FILLED;
59 static inline int o2info_coherent(struct ocfs2_info_request *req)
61 return (!(req->ir_flags & OCFS2_INFO_FL_NON_COHERENT));
64 static int ocfs2_get_inode_attr(struct inode *inode, unsigned *flags)
68 status = ocfs2_inode_lock(inode, NULL, 0);
73 ocfs2_get_inode_flags(OCFS2_I(inode));
74 *flags = OCFS2_I(inode)->ip_attr;
75 ocfs2_inode_unlock(inode, 0);
80 static int ocfs2_set_inode_attr(struct inode *inode, unsigned flags,
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;
92 status = ocfs2_inode_lock(inode, &bh, 1);
99 if (!inode_owner_or_capable(inode))
102 if (!S_ISDIR(inode->i_mode))
103 flags &= ~OCFS2_DIRSYNC_FL;
105 oldflags = ocfs2_inode->ip_attr;
106 flags = flags & mask;
107 flags |= oldflags & ~mask;
109 status = vfs_ioc_setflags_prepare(inode, oldflags, flags);
113 handle = ocfs2_start_trans(osb, OCFS2_INODE_UPDATE_CREDITS);
114 if (IS_ERR(handle)) {
115 status = PTR_ERR(handle);
120 ocfs2_inode->ip_attr = flags;
121 ocfs2_set_inode_flags(inode);
123 status = ocfs2_mark_inode_dirty(handle, inode, bh);
127 ocfs2_commit_trans(osb, handle);
130 ocfs2_inode_unlock(inode, 1);
139 static int ocfs2_info_handle_blocksize(struct inode *inode,
140 struct ocfs2_info_request __user *req)
142 struct ocfs2_info_blocksize oib;
144 if (o2info_from_user(oib, req))
147 oib.ib_blocksize = inode->i_sb->s_blocksize;
149 o2info_set_request_filled(&oib.ib_req);
151 if (o2info_to_user(oib, req))
157 static int ocfs2_info_handle_clustersize(struct inode *inode,
158 struct ocfs2_info_request __user *req)
160 struct ocfs2_info_clustersize oic;
161 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
163 if (o2info_from_user(oic, req))
166 oic.ic_clustersize = osb->s_clustersize;
168 o2info_set_request_filled(&oic.ic_req);
170 if (o2info_to_user(oic, req))
176 static int ocfs2_info_handle_maxslots(struct inode *inode,
177 struct ocfs2_info_request __user *req)
179 struct ocfs2_info_maxslots oim;
180 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
182 if (o2info_from_user(oim, req))
185 oim.im_max_slots = osb->max_slots;
187 o2info_set_request_filled(&oim.im_req);
189 if (o2info_to_user(oim, req))
195 static int ocfs2_info_handle_label(struct inode *inode,
196 struct ocfs2_info_request __user *req)
198 struct ocfs2_info_label oil;
199 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
201 if (o2info_from_user(oil, req))
204 memcpy(oil.il_label, osb->vol_label, OCFS2_MAX_VOL_LABEL_LEN);
206 o2info_set_request_filled(&oil.il_req);
208 if (o2info_to_user(oil, req))
214 static int ocfs2_info_handle_uuid(struct inode *inode,
215 struct ocfs2_info_request __user *req)
217 struct ocfs2_info_uuid oiu;
218 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
220 if (o2info_from_user(oiu, req))
223 memcpy(oiu.iu_uuid_str, osb->uuid_str, OCFS2_TEXT_UUID_LEN + 1);
225 o2info_set_request_filled(&oiu.iu_req);
227 if (o2info_to_user(oiu, req))
233 static int ocfs2_info_handle_fs_features(struct inode *inode,
234 struct ocfs2_info_request __user *req)
236 struct ocfs2_info_fs_features oif;
237 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
239 if (o2info_from_user(oif, req))
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;
246 o2info_set_request_filled(&oif.if_req);
248 if (o2info_to_user(oif, req))
254 static int ocfs2_info_handle_journal_size(struct inode *inode,
255 struct ocfs2_info_request __user *req)
257 struct ocfs2_info_journal_size oij;
258 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
260 if (o2info_from_user(oij, req))
263 oij.ij_journal_size = i_size_read(osb->journal->j_inode);
265 o2info_set_request_filled(&oij.ij_req);
267 if (o2info_to_user(oij, req))
273 static int ocfs2_info_scan_inode_alloc(struct ocfs2_super *osb,
274 struct inode *inode_alloc, u64 blkno,
275 struct ocfs2_info_freeinode *fi,
278 int status = 0, unlock = 0;
280 struct buffer_head *bh = NULL;
281 struct ocfs2_dinode *dinode_alloc = NULL;
284 inode_lock(inode_alloc);
286 if (inode_alloc && o2info_coherent(&fi->ifi_req)) {
287 status = ocfs2_inode_lock(inode_alloc, &bh, 0);
294 status = ocfs2_read_blocks_sync(osb, blkno, 1, &bh);
301 dinode_alloc = (struct ocfs2_dinode *)bh->b_data;
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);
311 ocfs2_inode_unlock(inode_alloc, 0);
314 inode_unlock(inode_alloc);
321 static int ocfs2_info_handle_freeinode(struct inode *inode,
322 struct ocfs2_info_request __user *req)
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;
332 oifi = kzalloc(sizeof(struct ocfs2_info_freeinode), GFP_KERNEL);
339 if (o2info_from_user(*oifi, req)) {
344 oifi->ifi_slotnum = osb->max_slots;
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);
350 mlog(ML_ERROR, "unable to get alloc inode in "
356 ocfs2_sprintf_system_inode_name(namebuf,
359 status = ocfs2_lookup_ino_from_name(osb->sys_root_inode,
369 status = ocfs2_info_scan_inode_alloc(osb, inode_alloc, blkno, oifi, i);
378 o2info_set_request_filled(&oifi->ifi_req);
380 if (o2info_to_user(*oifi, req)) {
388 o2info_set_request_error(&oifi->ifi_req, req);
395 static void o2ffg_update_histogram(struct ocfs2_info_free_chunk_list *hist,
396 unsigned int chunksize)
400 index = __ilog2_u32(chunksize);
401 if (index >= OCFS2_INFO_MAX_HIST)
402 index = OCFS2_INFO_MAX_HIST - 1;
404 hist->fc_chunks[index]++;
405 hist->fc_clusters[index] += chunksize;
408 static void o2ffg_update_stats(struct ocfs2_info_freefrag_stats *stats,
409 unsigned int chunksize)
411 if (chunksize > stats->ffs_max)
412 stats->ffs_max = chunksize;
414 if (chunksize < stats->ffs_min)
415 stats->ffs_min = chunksize;
417 stats->ffs_avg += chunksize;
418 stats->ffs_free_chunks_real++;
421 static void ocfs2_info_update_ffg(struct ocfs2_info_freefrag *ffg,
422 unsigned int chunksize)
424 o2ffg_update_histogram(&(ffg->iff_ffs.ffs_fc_hist), chunksize);
425 o2ffg_update_stats(&(ffg->iff_ffs), chunksize);
428 static 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,
435 int status = 0, used;
438 struct buffer_head *bh = NULL;
439 struct ocfs2_group_desc *bg = NULL;
441 unsigned int max_bits, num_clusters;
442 unsigned int offset = 0, cluster, chunk;
443 unsigned int chunk_free, last_chunksize = 0;
445 if (!le32_to_cpu(rec->c_free))
450 blkno = le64_to_cpu(rec->c_blkno);
452 blkno = le64_to_cpu(bg->bg_next_group);
459 if (o2info_coherent(&ffg->iff_req))
460 status = ocfs2_read_group_descriptor(gb_inode,
464 status = ocfs2_read_blocks_sync(osb, blkno, 1, &bh);
467 mlog(ML_ERROR, "Can't read the group descriptor # "
468 "%llu from device.", (unsigned long long)blkno);
473 bg = (struct ocfs2_group_desc *)bh->b_data;
475 if (!le16_to_cpu(bg->bg_free_bits_count))
478 max_bits = le16_to_cpu(bg->bg_bits);
481 for (chunk = 0; chunk < chunks_in_group; chunk++) {
483 * last chunk may be not an entire one.
485 if ((offset + ffg->iff_chunksize) > max_bits)
486 num_clusters = max_bits - offset;
488 num_clusters = ffg->iff_chunksize;
491 for (cluster = 0; cluster < num_clusters; cluster++) {
492 used = ocfs2_test_bit(offset,
493 (unsigned long *)bg->bg_bitmap);
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.
504 if (used && last_chunksize) {
505 ocfs2_info_update_ffg(ffg,
513 if (chunk_free == ffg->iff_chunksize)
514 ffg->iff_ffs.ffs_free_chunks++;
518 * need to update the info for last free chunk.
521 ocfs2_info_update_ffg(ffg, last_chunksize);
523 } while (le64_to_cpu(bg->bg_next_group));
531 static int ocfs2_info_freefrag_scan_bitmap(struct ocfs2_super *osb,
532 struct inode *gb_inode, u64 blkno,
533 struct ocfs2_info_freefrag *ffg)
536 int status = 0, unlock = 0, i;
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;
544 inode_lock(gb_inode);
546 if (o2info_coherent(&ffg->iff_req)) {
547 status = ocfs2_inode_lock(gb_inode, &bh, 0);
554 status = ocfs2_read_blocks_sync(osb, blkno, 1, &bh);
561 gb_dinode = (struct ocfs2_dinode *)bh->b_data;
562 cl = &(gb_dinode->id2.i_chain);
565 * Chunksize(in) clusters from userspace should be
566 * less than clusters in a group.
568 if (ffg->iff_chunksize > le16_to_cpu(cl->cl_cpg)) {
573 memset(&ffg->iff_ffs, 0, sizeof(struct ocfs2_info_freefrag_stats));
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);
581 chunks_in_group = le16_to_cpu(cl->cl_cpg) / ffg->iff_chunksize + 1;
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,
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);
598 ocfs2_inode_unlock(gb_inode, 0);
601 inode_unlock(gb_inode);
609 static int ocfs2_info_handle_freefrag(struct inode *inode,
610 struct ocfs2_info_request __user *req)
614 int status, type = GLOBAL_BITMAP_SYSTEM_INODE;
616 struct ocfs2_info_freefrag *oiff;
617 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
618 struct inode *gb_inode = NULL;
620 oiff = kzalloc(sizeof(struct ocfs2_info_freefrag), GFP_KERNEL);
627 if (o2info_from_user(*oiff, req)) {
632 * chunksize from userspace should be power of 2.
634 if ((oiff->iff_chunksize & (oiff->iff_chunksize - 1)) ||
635 (!oiff->iff_chunksize)) {
640 if (o2info_coherent(&oiff->iff_req)) {
641 gb_inode = ocfs2_get_system_file_inode(osb, type,
644 mlog(ML_ERROR, "unable to get global_bitmap inode\n");
649 ocfs2_sprintf_system_inode_name(namebuf, sizeof(namebuf), type,
651 status = ocfs2_lookup_ino_from_name(osb->sys_root_inode,
661 status = ocfs2_info_freefrag_scan_bitmap(osb, gb_inode, blkno, oiff);
665 o2info_set_request_filled(&oiff->iff_req);
667 if (o2info_to_user(*oiff, req)) {
675 o2info_set_request_error(&oiff->iff_req, req);
682 static int ocfs2_info_handle_unknown(struct inode *inode,
683 struct ocfs2_info_request __user *req)
685 struct ocfs2_info_request oir;
687 if (o2info_from_user(oir, req))
690 o2info_clear_request_filled(&oir);
692 if (o2info_to_user(oir, req))
699 * Validate and distinguish OCFS2_IOC_INFO requests.
701 * - validate the magic number.
702 * - distinguish different requests.
703 * - validate size of different requests.
705 static int ocfs2_info_handle_request(struct inode *inode,
706 struct ocfs2_info_request __user *req)
708 int status = -EFAULT;
709 struct ocfs2_info_request oir;
711 if (o2info_from_user(oir, req))
715 if (oir.ir_magic != OCFS2_INFO_MAGIC)
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);
723 case OCFS2_INFO_CLUSTERSIZE:
724 if (oir.ir_size == sizeof(struct ocfs2_info_clustersize))
725 status = ocfs2_info_handle_clustersize(inode, req);
727 case OCFS2_INFO_MAXSLOTS:
728 if (oir.ir_size == sizeof(struct ocfs2_info_maxslots))
729 status = ocfs2_info_handle_maxslots(inode, req);
731 case OCFS2_INFO_LABEL:
732 if (oir.ir_size == sizeof(struct ocfs2_info_label))
733 status = ocfs2_info_handle_label(inode, req);
735 case OCFS2_INFO_UUID:
736 if (oir.ir_size == sizeof(struct ocfs2_info_uuid))
737 status = ocfs2_info_handle_uuid(inode, req);
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);
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);
747 case OCFS2_INFO_FREEINODE:
748 if (oir.ir_size == sizeof(struct ocfs2_info_freeinode))
749 status = ocfs2_info_handle_freeinode(inode, req);
751 case OCFS2_INFO_FREEFRAG:
752 if (oir.ir_size == sizeof(struct ocfs2_info_freefrag))
753 status = ocfs2_info_handle_freefrag(inode, req);
756 status = ocfs2_info_handle_unknown(inode, req);
764 static int ocfs2_get_request_ptr(struct ocfs2_info *info, int idx,
765 u64 *req_addr, int compat_flag)
767 int status = -EFAULT;
768 u64 __user *bp = NULL;
773 * pointer bp stores the base address of a pointers array,
774 * which collects all addresses of separate request.
776 bp = (u64 __user *)(unsigned long)compat_ptr(info->oi_requests);
781 bp = (u64 __user *)(unsigned long)(info->oi_requests);
783 if (o2info_from_user(*req_addr, bp + idx))
792 * OCFS2_IOC_INFO handles an array of requests passed from userspace.
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.
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.
802 static int ocfs2_info_handle(struct inode *inode, struct ocfs2_info *info,
807 struct ocfs2_info_request __user *reqp;
809 if ((info->oi_count > OCFS2_INFO_MAX_REQUEST) ||
810 (!info->oi_requests)) {
815 for (i = 0; i < info->oi_count; i++) {
817 status = ocfs2_get_request_ptr(info, i, &req_addr, compat_flag);
821 reqp = (struct ocfs2_info_request __user *)(unsigned long)req_addr;
827 status = ocfs2_info_handle_request(inode, reqp);
836 long ocfs2_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
838 struct inode *inode = file_inode(filp);
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;
848 struct ocfs2_info info;
849 void __user *argp = (void __user *)arg;
852 case OCFS2_IOC_GETFLAGS:
853 status = ocfs2_get_inode_attr(inode, &flags);
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))
863 status = mnt_want_write_file(filp);
866 status = ocfs2_set_inode_attr(inode, flags,
867 OCFS2_FL_MODIFIABLE);
868 mnt_drop_write_file(filp);
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)))
877 return ocfs2_change_file_space(filp, cmd, &sr);
878 case OCFS2_IOC_GROUP_EXTEND:
879 if (!capable(CAP_SYS_RESOURCE))
882 if (get_user(new_clusters, (int __user *)arg))
885 status = mnt_want_write_file(filp);
888 status = ocfs2_group_extend(inode, new_clusters);
889 mnt_drop_write_file(filp);
891 case OCFS2_IOC_GROUP_ADD:
892 case OCFS2_IOC_GROUP_ADD64:
893 if (!capable(CAP_SYS_RESOURCE))
896 if (copy_from_user(&input, (int __user *) arg, sizeof(input)))
899 status = mnt_want_write_file(filp);
902 status = ocfs2_group_add(inode, &input);
903 mnt_drop_write_file(filp);
905 case OCFS2_IOC_REFLINK:
906 if (copy_from_user(&args, argp, sizeof(args)))
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);
912 return ocfs2_reflink_ioctl(inode, old_path, new_path, preserve);
914 if (copy_from_user(&info, argp, sizeof(struct ocfs2_info)))
917 return ocfs2_info_handle(inode, &info, 0);
920 struct super_block *sb = inode->i_sb;
921 struct request_queue *q = bdev_get_queue(sb->s_bdev);
922 struct fstrim_range range;
925 if (!capable(CAP_SYS_ADMIN))
928 if (!blk_queue_discard(q))
931 if (copy_from_user(&range, argp, sizeof(range)))
934 range.minlen = max_t(u64, q->limits.discard_granularity,
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_IOC32_GETFLAGS:
963 cmd = OCFS2_IOC_GETFLAGS;
965 case OCFS2_IOC32_SETFLAGS:
966 cmd = OCFS2_IOC_SETFLAGS;
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:
976 case OCFS2_IOC_REFLINK:
977 if (copy_from_user(&args, argp, sizeof(args)))
979 preserve = (args.preserve != 0);
981 return ocfs2_reflink_ioctl(inode, compat_ptr(args.old_path),
982 compat_ptr(args.new_path), preserve);
984 if (copy_from_user(&info, argp, sizeof(struct ocfs2_info)))
987 return ocfs2_info_handle(inode, &info, 1);
988 case OCFS2_IOC_MOVE_EXT:
994 return ocfs2_ioctl(file, cmd, arg);