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;
110 * The IMMUTABLE and APPEND_ONLY flags can only be changed by
111 * the relevant capability.
114 if ((oldflags & OCFS2_IMMUTABLE_FL) || ((flags ^ oldflags) &
115 (OCFS2_APPEND_FL | OCFS2_IMMUTABLE_FL))) {
116 if (!capable(CAP_LINUX_IMMUTABLE))
120 handle = ocfs2_start_trans(osb, OCFS2_INODE_UPDATE_CREDITS);
121 if (IS_ERR(handle)) {
122 status = PTR_ERR(handle);
127 ocfs2_inode->ip_attr = flags;
128 ocfs2_set_inode_flags(inode);
130 status = ocfs2_mark_inode_dirty(handle, inode, bh);
134 ocfs2_commit_trans(osb, handle);
137 ocfs2_inode_unlock(inode, 1);
146 static int ocfs2_info_handle_blocksize(struct inode *inode,
147 struct ocfs2_info_request __user *req)
149 struct ocfs2_info_blocksize oib;
151 if (o2info_from_user(oib, req))
154 oib.ib_blocksize = inode->i_sb->s_blocksize;
156 o2info_set_request_filled(&oib.ib_req);
158 if (o2info_to_user(oib, req))
164 static int ocfs2_info_handle_clustersize(struct inode *inode,
165 struct ocfs2_info_request __user *req)
167 struct ocfs2_info_clustersize oic;
168 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
170 if (o2info_from_user(oic, req))
173 oic.ic_clustersize = osb->s_clustersize;
175 o2info_set_request_filled(&oic.ic_req);
177 if (o2info_to_user(oic, req))
183 static int ocfs2_info_handle_maxslots(struct inode *inode,
184 struct ocfs2_info_request __user *req)
186 struct ocfs2_info_maxslots oim;
187 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
189 if (o2info_from_user(oim, req))
192 oim.im_max_slots = osb->max_slots;
194 o2info_set_request_filled(&oim.im_req);
196 if (o2info_to_user(oim, req))
202 static int ocfs2_info_handle_label(struct inode *inode,
203 struct ocfs2_info_request __user *req)
205 struct ocfs2_info_label oil;
206 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
208 if (o2info_from_user(oil, req))
211 memcpy(oil.il_label, osb->vol_label, OCFS2_MAX_VOL_LABEL_LEN);
213 o2info_set_request_filled(&oil.il_req);
215 if (o2info_to_user(oil, req))
221 static int ocfs2_info_handle_uuid(struct inode *inode,
222 struct ocfs2_info_request __user *req)
224 struct ocfs2_info_uuid oiu;
225 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
227 if (o2info_from_user(oiu, req))
230 memcpy(oiu.iu_uuid_str, osb->uuid_str, OCFS2_TEXT_UUID_LEN + 1);
232 o2info_set_request_filled(&oiu.iu_req);
234 if (o2info_to_user(oiu, req))
240 static int ocfs2_info_handle_fs_features(struct inode *inode,
241 struct ocfs2_info_request __user *req)
243 struct ocfs2_info_fs_features oif;
244 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
246 if (o2info_from_user(oif, req))
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;
253 o2info_set_request_filled(&oif.if_req);
255 if (o2info_to_user(oif, req))
261 static int ocfs2_info_handle_journal_size(struct inode *inode,
262 struct ocfs2_info_request __user *req)
264 struct ocfs2_info_journal_size oij;
265 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
267 if (o2info_from_user(oij, req))
270 oij.ij_journal_size = i_size_read(osb->journal->j_inode);
272 o2info_set_request_filled(&oij.ij_req);
274 if (o2info_to_user(oij, req))
280 static int ocfs2_info_scan_inode_alloc(struct ocfs2_super *osb,
281 struct inode *inode_alloc, u64 blkno,
282 struct ocfs2_info_freeinode *fi,
285 int status = 0, unlock = 0;
287 struct buffer_head *bh = NULL;
288 struct ocfs2_dinode *dinode_alloc = NULL;
291 inode_lock(inode_alloc);
293 if (inode_alloc && o2info_coherent(&fi->ifi_req)) {
294 status = ocfs2_inode_lock(inode_alloc, &bh, 0);
301 status = ocfs2_read_blocks_sync(osb, blkno, 1, &bh);
308 dinode_alloc = (struct ocfs2_dinode *)bh->b_data;
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);
318 ocfs2_inode_unlock(inode_alloc, 0);
321 inode_unlock(inode_alloc);
328 static int ocfs2_info_handle_freeinode(struct inode *inode,
329 struct ocfs2_info_request __user *req)
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;
339 oifi = kzalloc(sizeof(struct ocfs2_info_freeinode), GFP_KERNEL);
346 if (o2info_from_user(*oifi, req)) {
351 oifi->ifi_slotnum = osb->max_slots;
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);
357 mlog(ML_ERROR, "unable to get alloc inode in "
363 ocfs2_sprintf_system_inode_name(namebuf,
366 status = ocfs2_lookup_ino_from_name(osb->sys_root_inode,
376 status = ocfs2_info_scan_inode_alloc(osb, inode_alloc, blkno, oifi, i);
385 o2info_set_request_filled(&oifi->ifi_req);
387 if (o2info_to_user(*oifi, req)) {
395 o2info_set_request_error(&oifi->ifi_req, req);
402 static void o2ffg_update_histogram(struct ocfs2_info_free_chunk_list *hist,
403 unsigned int chunksize)
407 index = __ilog2_u32(chunksize);
408 if (index >= OCFS2_INFO_MAX_HIST)
409 index = OCFS2_INFO_MAX_HIST - 1;
411 hist->fc_chunks[index]++;
412 hist->fc_clusters[index] += chunksize;
415 static void o2ffg_update_stats(struct ocfs2_info_freefrag_stats *stats,
416 unsigned int chunksize)
418 if (chunksize > stats->ffs_max)
419 stats->ffs_max = chunksize;
421 if (chunksize < stats->ffs_min)
422 stats->ffs_min = chunksize;
424 stats->ffs_avg += chunksize;
425 stats->ffs_free_chunks_real++;
428 static void ocfs2_info_update_ffg(struct ocfs2_info_freefrag *ffg,
429 unsigned int chunksize)
431 o2ffg_update_histogram(&(ffg->iff_ffs.ffs_fc_hist), chunksize);
432 o2ffg_update_stats(&(ffg->iff_ffs), chunksize);
435 static 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,
442 int status = 0, used;
445 struct buffer_head *bh = NULL;
446 struct ocfs2_group_desc *bg = NULL;
448 unsigned int max_bits, num_clusters;
449 unsigned int offset = 0, cluster, chunk;
450 unsigned int chunk_free, last_chunksize = 0;
452 if (!le32_to_cpu(rec->c_free))
457 blkno = le64_to_cpu(rec->c_blkno);
459 blkno = le64_to_cpu(bg->bg_next_group);
466 if (o2info_coherent(&ffg->iff_req))
467 status = ocfs2_read_group_descriptor(gb_inode,
471 status = ocfs2_read_blocks_sync(osb, blkno, 1, &bh);
474 mlog(ML_ERROR, "Can't read the group descriptor # "
475 "%llu from device.", (unsigned long long)blkno);
480 bg = (struct ocfs2_group_desc *)bh->b_data;
482 if (!le16_to_cpu(bg->bg_free_bits_count))
485 max_bits = le16_to_cpu(bg->bg_bits);
488 for (chunk = 0; chunk < chunks_in_group; chunk++) {
490 * last chunk may be not an entire one.
492 if ((offset + ffg->iff_chunksize) > max_bits)
493 num_clusters = max_bits - offset;
495 num_clusters = ffg->iff_chunksize;
498 for (cluster = 0; cluster < num_clusters; cluster++) {
499 used = ocfs2_test_bit(offset,
500 (unsigned long *)bg->bg_bitmap);
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.
511 if (used && last_chunksize) {
512 ocfs2_info_update_ffg(ffg,
520 if (chunk_free == ffg->iff_chunksize)
521 ffg->iff_ffs.ffs_free_chunks++;
525 * need to update the info for last free chunk.
528 ocfs2_info_update_ffg(ffg, last_chunksize);
530 } while (le64_to_cpu(bg->bg_next_group));
538 static int ocfs2_info_freefrag_scan_bitmap(struct ocfs2_super *osb,
539 struct inode *gb_inode, u64 blkno,
540 struct ocfs2_info_freefrag *ffg)
543 int status = 0, unlock = 0, i;
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;
551 inode_lock(gb_inode);
553 if (o2info_coherent(&ffg->iff_req)) {
554 status = ocfs2_inode_lock(gb_inode, &bh, 0);
561 status = ocfs2_read_blocks_sync(osb, blkno, 1, &bh);
568 gb_dinode = (struct ocfs2_dinode *)bh->b_data;
569 cl = &(gb_dinode->id2.i_chain);
572 * Chunksize(in) clusters from userspace should be
573 * less than clusters in a group.
575 if (ffg->iff_chunksize > le16_to_cpu(cl->cl_cpg)) {
580 memset(&ffg->iff_ffs, 0, sizeof(struct ocfs2_info_freefrag_stats));
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);
588 chunks_in_group = le16_to_cpu(cl->cl_cpg) / ffg->iff_chunksize + 1;
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,
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);
605 ocfs2_inode_unlock(gb_inode, 0);
608 inode_unlock(gb_inode);
616 static int ocfs2_info_handle_freefrag(struct inode *inode,
617 struct ocfs2_info_request __user *req)
621 int status, type = GLOBAL_BITMAP_SYSTEM_INODE;
623 struct ocfs2_info_freefrag *oiff;
624 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
625 struct inode *gb_inode = NULL;
627 oiff = kzalloc(sizeof(struct ocfs2_info_freefrag), GFP_KERNEL);
634 if (o2info_from_user(*oiff, req)) {
639 * chunksize from userspace should be power of 2.
641 if ((oiff->iff_chunksize & (oiff->iff_chunksize - 1)) ||
642 (!oiff->iff_chunksize)) {
647 if (o2info_coherent(&oiff->iff_req)) {
648 gb_inode = ocfs2_get_system_file_inode(osb, type,
651 mlog(ML_ERROR, "unable to get global_bitmap inode\n");
656 ocfs2_sprintf_system_inode_name(namebuf, sizeof(namebuf), type,
658 status = ocfs2_lookup_ino_from_name(osb->sys_root_inode,
668 status = ocfs2_info_freefrag_scan_bitmap(osb, gb_inode, blkno, oiff);
672 o2info_set_request_filled(&oiff->iff_req);
674 if (o2info_to_user(*oiff, req)) {
682 o2info_set_request_error(&oiff->iff_req, req);
689 static int ocfs2_info_handle_unknown(struct inode *inode,
690 struct ocfs2_info_request __user *req)
692 struct ocfs2_info_request oir;
694 if (o2info_from_user(oir, req))
697 o2info_clear_request_filled(&oir);
699 if (o2info_to_user(oir, req))
706 * Validate and distinguish OCFS2_IOC_INFO requests.
708 * - validate the magic number.
709 * - distinguish different requests.
710 * - validate size of different requests.
712 static int ocfs2_info_handle_request(struct inode *inode,
713 struct ocfs2_info_request __user *req)
715 int status = -EFAULT;
716 struct ocfs2_info_request oir;
718 if (o2info_from_user(oir, req))
722 if (oir.ir_magic != OCFS2_INFO_MAGIC)
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);
730 case OCFS2_INFO_CLUSTERSIZE:
731 if (oir.ir_size == sizeof(struct ocfs2_info_clustersize))
732 status = ocfs2_info_handle_clustersize(inode, req);
734 case OCFS2_INFO_MAXSLOTS:
735 if (oir.ir_size == sizeof(struct ocfs2_info_maxslots))
736 status = ocfs2_info_handle_maxslots(inode, req);
738 case OCFS2_INFO_LABEL:
739 if (oir.ir_size == sizeof(struct ocfs2_info_label))
740 status = ocfs2_info_handle_label(inode, req);
742 case OCFS2_INFO_UUID:
743 if (oir.ir_size == sizeof(struct ocfs2_info_uuid))
744 status = ocfs2_info_handle_uuid(inode, req);
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);
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);
754 case OCFS2_INFO_FREEINODE:
755 if (oir.ir_size == sizeof(struct ocfs2_info_freeinode))
756 status = ocfs2_info_handle_freeinode(inode, req);
758 case OCFS2_INFO_FREEFRAG:
759 if (oir.ir_size == sizeof(struct ocfs2_info_freefrag))
760 status = ocfs2_info_handle_freefrag(inode, req);
763 status = ocfs2_info_handle_unknown(inode, req);
771 static int ocfs2_get_request_ptr(struct ocfs2_info *info, int idx,
772 u64 *req_addr, int compat_flag)
774 int status = -EFAULT;
775 u64 __user *bp = NULL;
780 * pointer bp stores the base address of a pointers array,
781 * which collects all addresses of separate request.
783 bp = (u64 __user *)(unsigned long)compat_ptr(info->oi_requests);
788 bp = (u64 __user *)(unsigned long)(info->oi_requests);
790 if (o2info_from_user(*req_addr, bp + idx))
799 * OCFS2_IOC_INFO handles an array of requests passed from userspace.
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.
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.
809 static int ocfs2_info_handle(struct inode *inode, struct ocfs2_info *info,
814 struct ocfs2_info_request __user *reqp;
816 if ((info->oi_count > OCFS2_INFO_MAX_REQUEST) ||
817 (!info->oi_requests)) {
822 for (i = 0; i < info->oi_count; i++) {
824 status = ocfs2_get_request_ptr(info, i, &req_addr, compat_flag);
828 reqp = (struct ocfs2_info_request __user *)(unsigned long)req_addr;
834 status = ocfs2_info_handle_request(inode, reqp);
843 long ocfs2_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
845 struct inode *inode = file_inode(filp);
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;
855 struct ocfs2_info info;
856 void __user *argp = (void __user *)arg;
859 case OCFS2_IOC_GETFLAGS:
860 status = ocfs2_get_inode_attr(inode, &flags);
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))
870 status = mnt_want_write_file(filp);
873 status = ocfs2_set_inode_attr(inode, flags,
874 OCFS2_FL_MODIFIABLE);
875 mnt_drop_write_file(filp);
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)))
884 return ocfs2_change_file_space(filp, cmd, &sr);
885 case OCFS2_IOC_GROUP_EXTEND:
886 if (!capable(CAP_SYS_RESOURCE))
889 if (get_user(new_clusters, (int __user *)arg))
892 status = mnt_want_write_file(filp);
895 status = ocfs2_group_extend(inode, new_clusters);
896 mnt_drop_write_file(filp);
898 case OCFS2_IOC_GROUP_ADD:
899 case OCFS2_IOC_GROUP_ADD64:
900 if (!capable(CAP_SYS_RESOURCE))
903 if (copy_from_user(&input, (int __user *) arg, sizeof(input)))
906 status = mnt_want_write_file(filp);
909 status = ocfs2_group_add(inode, &input);
910 mnt_drop_write_file(filp);
912 case OCFS2_IOC_REFLINK:
913 if (copy_from_user(&args, argp, sizeof(args)))
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);
919 return ocfs2_reflink_ioctl(inode, old_path, new_path, preserve);
921 if (copy_from_user(&info, argp, sizeof(struct ocfs2_info)))
924 return ocfs2_info_handle(inode, &info, 0);
927 struct super_block *sb = inode->i_sb;
928 struct request_queue *q = bdev_get_queue(sb->s_bdev);
929 struct fstrim_range range;
932 if (!capable(CAP_SYS_ADMIN))
935 if (!blk_queue_discard(q))
938 if (copy_from_user(&range, argp, sizeof(range)))
941 range.minlen = max_t(u64, q->limits.discard_granularity,
943 ret = ocfs2_trim_fs(sb, &range);
947 if (copy_to_user(argp, &range, sizeof(range)))
952 case OCFS2_IOC_MOVE_EXT:
953 return ocfs2_ioctl_move_extents(filp, argp);
960 long ocfs2_compat_ioctl(struct file *file, unsigned cmd, unsigned long arg)
963 struct reflink_arguments args;
964 struct inode *inode = file_inode(file);
965 struct ocfs2_info info;
966 void __user *argp = (void __user *)arg;
969 case OCFS2_IOC32_GETFLAGS:
970 cmd = OCFS2_IOC_GETFLAGS;
972 case OCFS2_IOC32_SETFLAGS:
973 cmd = OCFS2_IOC_SETFLAGS;
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:
983 case OCFS2_IOC_REFLINK:
984 if (copy_from_user(&args, argp, sizeof(args)))
986 preserve = (args.preserve != 0);
988 return ocfs2_reflink_ioctl(inode, compat_ptr(args.old_path),
989 compat_ptr(args.new_path), preserve);
991 if (copy_from_user(&info, argp, sizeof(struct ocfs2_info)))
994 return ocfs2_info_handle(inode, &info, 1);
995 case OCFS2_IOC_MOVE_EXT:
1001 return ocfs2_ioctl(file, cmd, arg);