2 * linux/fs/ocfs2/ioctl.c
4 * Copyright (C) 2006 Herbert Poetzl
5 * adapted from Remy Card's ext2/ioctl.c
9 #include <linux/mount.h>
10 #include <linux/blkdev.h>
11 #include <linux/compat.h>
13 #include <cluster/masklog.h>
25 #include "refcounttree.h"
28 #include "buffer_head_io.h"
30 #include "move_extents.h"
32 #define o2info_from_user(a, b) \
33 copy_from_user(&(a), (b), sizeof(a))
34 #define o2info_to_user(a, b) \
35 copy_to_user((typeof(a) __user *)b, &(a), sizeof(a))
38 * This is just a best-effort to tell userspace that this request
41 static inline void o2info_set_request_error(struct ocfs2_info_request *kreq,
42 struct ocfs2_info_request __user *req)
44 kreq->ir_flags |= OCFS2_INFO_FL_ERROR;
45 (void)put_user(kreq->ir_flags, (__u32 __user *)&(req->ir_flags));
48 static inline void o2info_set_request_filled(struct ocfs2_info_request *req)
50 req->ir_flags |= OCFS2_INFO_FL_FILLED;
53 static inline void o2info_clear_request_filled(struct ocfs2_info_request *req)
55 req->ir_flags &= ~OCFS2_INFO_FL_FILLED;
58 static inline int o2info_coherent(struct ocfs2_info_request *req)
60 return (!(req->ir_flags & OCFS2_INFO_FL_NON_COHERENT));
63 static int ocfs2_get_inode_attr(struct inode *inode, unsigned *flags)
67 status = ocfs2_inode_lock(inode, NULL, 0);
72 ocfs2_get_inode_flags(OCFS2_I(inode));
73 *flags = OCFS2_I(inode)->ip_attr;
74 ocfs2_inode_unlock(inode, 0);
79 static int ocfs2_set_inode_attr(struct inode *inode, unsigned flags,
82 struct ocfs2_inode_info *ocfs2_inode = OCFS2_I(inode);
83 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
84 handle_t *handle = NULL;
85 struct buffer_head *bh = NULL;
89 mutex_lock(&inode->i_mutex);
91 status = ocfs2_inode_lock(inode, &bh, 1);
98 if (!inode_owner_or_capable(inode))
101 if (!S_ISDIR(inode->i_mode))
102 flags &= ~OCFS2_DIRSYNC_FL;
104 oldflags = ocfs2_inode->ip_attr;
105 flags = flags & mask;
106 flags |= oldflags & ~mask;
109 * The IMMUTABLE and APPEND_ONLY flags can only be changed by
110 * the relevant capability.
113 if ((oldflags & OCFS2_IMMUTABLE_FL) || ((flags ^ oldflags) &
114 (OCFS2_APPEND_FL | OCFS2_IMMUTABLE_FL))) {
115 if (!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);
129 status = ocfs2_mark_inode_dirty(handle, inode, bh);
133 ocfs2_commit_trans(osb, handle);
136 ocfs2_inode_unlock(inode, 1);
138 mutex_unlock(&inode->i_mutex);
145 static int ocfs2_info_handle_blocksize(struct inode *inode,
146 struct ocfs2_info_request __user *req)
148 struct ocfs2_info_blocksize oib;
150 if (o2info_from_user(oib, req))
153 oib.ib_blocksize = inode->i_sb->s_blocksize;
155 o2info_set_request_filled(&oib.ib_req);
157 if (o2info_to_user(oib, req))
163 static int ocfs2_info_handle_clustersize(struct inode *inode,
164 struct ocfs2_info_request __user *req)
166 struct ocfs2_info_clustersize oic;
167 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
169 if (o2info_from_user(oic, req))
172 oic.ic_clustersize = osb->s_clustersize;
174 o2info_set_request_filled(&oic.ic_req);
176 if (o2info_to_user(oic, req))
182 static int ocfs2_info_handle_maxslots(struct inode *inode,
183 struct ocfs2_info_request __user *req)
185 struct ocfs2_info_maxslots oim;
186 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
188 if (o2info_from_user(oim, req))
191 oim.im_max_slots = osb->max_slots;
193 o2info_set_request_filled(&oim.im_req);
195 if (o2info_to_user(oim, req))
201 static int ocfs2_info_handle_label(struct inode *inode,
202 struct ocfs2_info_request __user *req)
204 struct ocfs2_info_label oil;
205 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
207 if (o2info_from_user(oil, req))
210 memcpy(oil.il_label, osb->vol_label, OCFS2_MAX_VOL_LABEL_LEN);
212 o2info_set_request_filled(&oil.il_req);
214 if (o2info_to_user(oil, req))
220 static int ocfs2_info_handle_uuid(struct inode *inode,
221 struct ocfs2_info_request __user *req)
223 struct ocfs2_info_uuid oiu;
224 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
226 if (o2info_from_user(oiu, req))
229 memcpy(oiu.iu_uuid_str, osb->uuid_str, OCFS2_TEXT_UUID_LEN + 1);
231 o2info_set_request_filled(&oiu.iu_req);
233 if (o2info_to_user(oiu, req))
239 static int ocfs2_info_handle_fs_features(struct inode *inode,
240 struct ocfs2_info_request __user *req)
242 struct ocfs2_info_fs_features oif;
243 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
245 if (o2info_from_user(oif, req))
248 oif.if_compat_features = osb->s_feature_compat;
249 oif.if_incompat_features = osb->s_feature_incompat;
250 oif.if_ro_compat_features = osb->s_feature_ro_compat;
252 o2info_set_request_filled(&oif.if_req);
254 if (o2info_to_user(oif, req))
260 static int ocfs2_info_handle_journal_size(struct inode *inode,
261 struct ocfs2_info_request __user *req)
263 struct ocfs2_info_journal_size oij;
264 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
266 if (o2info_from_user(oij, req))
269 oij.ij_journal_size = i_size_read(osb->journal->j_inode);
271 o2info_set_request_filled(&oij.ij_req);
273 if (o2info_to_user(oij, req))
279 static int ocfs2_info_scan_inode_alloc(struct ocfs2_super *osb,
280 struct inode *inode_alloc, u64 blkno,
281 struct ocfs2_info_freeinode *fi,
284 int status = 0, unlock = 0;
286 struct buffer_head *bh = NULL;
287 struct ocfs2_dinode *dinode_alloc = NULL;
290 mutex_lock(&inode_alloc->i_mutex);
292 if (o2info_coherent(&fi->ifi_req)) {
293 status = ocfs2_inode_lock(inode_alloc, &bh, 0);
300 status = ocfs2_read_blocks_sync(osb, blkno, 1, &bh);
307 dinode_alloc = (struct ocfs2_dinode *)bh->b_data;
309 fi->ifi_stat[slot].lfi_total =
310 le32_to_cpu(dinode_alloc->id1.bitmap1.i_total);
311 fi->ifi_stat[slot].lfi_free =
312 le32_to_cpu(dinode_alloc->id1.bitmap1.i_total) -
313 le32_to_cpu(dinode_alloc->id1.bitmap1.i_used);
317 ocfs2_inode_unlock(inode_alloc, 0);
320 mutex_unlock(&inode_alloc->i_mutex);
327 static int ocfs2_info_handle_freeinode(struct inode *inode,
328 struct ocfs2_info_request __user *req)
333 int status, type = INODE_ALLOC_SYSTEM_INODE;
334 struct ocfs2_info_freeinode *oifi = NULL;
335 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
336 struct inode *inode_alloc = NULL;
338 oifi = kzalloc(sizeof(struct ocfs2_info_freeinode), GFP_KERNEL);
345 if (o2info_from_user(*oifi, req)) {
350 oifi->ifi_slotnum = osb->max_slots;
352 for (i = 0; i < oifi->ifi_slotnum; i++) {
353 if (o2info_coherent(&oifi->ifi_req)) {
354 inode_alloc = ocfs2_get_system_file_inode(osb, type, i);
356 mlog(ML_ERROR, "unable to get alloc inode in "
362 ocfs2_sprintf_system_inode_name(namebuf,
365 status = ocfs2_lookup_ino_from_name(osb->sys_root_inode,
375 status = ocfs2_info_scan_inode_alloc(osb, inode_alloc, blkno, oifi, i);
384 o2info_set_request_filled(&oifi->ifi_req);
386 if (o2info_to_user(*oifi, req)) {
394 o2info_set_request_error(&oifi->ifi_req, req);
401 static void o2ffg_update_histogram(struct ocfs2_info_free_chunk_list *hist,
402 unsigned int chunksize)
406 index = __ilog2_u32(chunksize);
407 if (index >= OCFS2_INFO_MAX_HIST)
408 index = OCFS2_INFO_MAX_HIST - 1;
410 hist->fc_chunks[index]++;
411 hist->fc_clusters[index] += chunksize;
414 static void o2ffg_update_stats(struct ocfs2_info_freefrag_stats *stats,
415 unsigned int chunksize)
417 if (chunksize > stats->ffs_max)
418 stats->ffs_max = chunksize;
420 if (chunksize < stats->ffs_min)
421 stats->ffs_min = chunksize;
423 stats->ffs_avg += chunksize;
424 stats->ffs_free_chunks_real++;
427 static void ocfs2_info_update_ffg(struct ocfs2_info_freefrag *ffg,
428 unsigned int chunksize)
430 o2ffg_update_histogram(&(ffg->iff_ffs.ffs_fc_hist), chunksize);
431 o2ffg_update_stats(&(ffg->iff_ffs), chunksize);
434 static int ocfs2_info_freefrag_scan_chain(struct ocfs2_super *osb,
435 struct inode *gb_inode,
436 struct ocfs2_dinode *gb_dinode,
437 struct ocfs2_chain_rec *rec,
438 struct ocfs2_info_freefrag *ffg,
441 int status = 0, used;
444 struct buffer_head *bh = NULL;
445 struct ocfs2_group_desc *bg = NULL;
447 unsigned int max_bits, num_clusters;
448 unsigned int offset = 0, cluster, chunk;
449 unsigned int chunk_free, last_chunksize = 0;
451 if (!le32_to_cpu(rec->c_free))
456 blkno = le64_to_cpu(rec->c_blkno);
458 blkno = le64_to_cpu(bg->bg_next_group);
465 if (o2info_coherent(&ffg->iff_req))
466 status = ocfs2_read_group_descriptor(gb_inode,
470 status = ocfs2_read_blocks_sync(osb, blkno, 1, &bh);
473 mlog(ML_ERROR, "Can't read the group descriptor # "
474 "%llu from device.", (unsigned long long)blkno);
479 bg = (struct ocfs2_group_desc *)bh->b_data;
481 if (!le16_to_cpu(bg->bg_free_bits_count))
484 max_bits = le16_to_cpu(bg->bg_bits);
487 for (chunk = 0; chunk < chunks_in_group; chunk++) {
489 * last chunk may be not an entire one.
491 if ((offset + ffg->iff_chunksize) > max_bits)
492 num_clusters = max_bits - offset;
494 num_clusters = ffg->iff_chunksize;
497 for (cluster = 0; cluster < num_clusters; cluster++) {
498 used = ocfs2_test_bit(offset,
499 (unsigned long *)bg->bg_bitmap);
501 * - chunk_free counts free clusters in #N chunk.
502 * - last_chunksize records the size(in) clusters
503 * for the last real free chunk being counted.
510 if (used && last_chunksize) {
511 ocfs2_info_update_ffg(ffg,
519 if (chunk_free == ffg->iff_chunksize)
520 ffg->iff_ffs.ffs_free_chunks++;
524 * need to update the info for last free chunk.
527 ocfs2_info_update_ffg(ffg, last_chunksize);
529 } while (le64_to_cpu(bg->bg_next_group));
537 static int ocfs2_info_freefrag_scan_bitmap(struct ocfs2_super *osb,
538 struct inode *gb_inode, u64 blkno,
539 struct ocfs2_info_freefrag *ffg)
542 int status = 0, unlock = 0, i;
544 struct buffer_head *bh = NULL;
545 struct ocfs2_chain_list *cl = NULL;
546 struct ocfs2_chain_rec *rec = NULL;
547 struct ocfs2_dinode *gb_dinode = NULL;
550 mutex_lock(&gb_inode->i_mutex);
552 if (o2info_coherent(&ffg->iff_req)) {
553 status = ocfs2_inode_lock(gb_inode, &bh, 0);
560 status = ocfs2_read_blocks_sync(osb, blkno, 1, &bh);
567 gb_dinode = (struct ocfs2_dinode *)bh->b_data;
568 cl = &(gb_dinode->id2.i_chain);
571 * Chunksize(in) clusters from userspace should be
572 * less than clusters in a group.
574 if (ffg->iff_chunksize > le16_to_cpu(cl->cl_cpg)) {
579 memset(&ffg->iff_ffs, 0, sizeof(struct ocfs2_info_freefrag_stats));
581 ffg->iff_ffs.ffs_min = ~0U;
582 ffg->iff_ffs.ffs_clusters =
583 le32_to_cpu(gb_dinode->id1.bitmap1.i_total);
584 ffg->iff_ffs.ffs_free_clusters = ffg->iff_ffs.ffs_clusters -
585 le32_to_cpu(gb_dinode->id1.bitmap1.i_used);
587 chunks_in_group = le16_to_cpu(cl->cl_cpg) / ffg->iff_chunksize + 1;
589 for (i = 0; i < le16_to_cpu(cl->cl_next_free_rec); i++) {
590 rec = &(cl->cl_recs[i]);
591 status = ocfs2_info_freefrag_scan_chain(osb, gb_inode,
599 if (ffg->iff_ffs.ffs_free_chunks_real)
600 ffg->iff_ffs.ffs_avg = (ffg->iff_ffs.ffs_avg /
601 ffg->iff_ffs.ffs_free_chunks_real);
604 ocfs2_inode_unlock(gb_inode, 0);
607 mutex_unlock(&gb_inode->i_mutex);
617 static int ocfs2_info_handle_freefrag(struct inode *inode,
618 struct ocfs2_info_request __user *req)
622 int status, type = GLOBAL_BITMAP_SYSTEM_INODE;
624 struct ocfs2_info_freefrag *oiff;
625 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
626 struct inode *gb_inode = NULL;
628 oiff = kzalloc(sizeof(struct ocfs2_info_freefrag), GFP_KERNEL);
635 if (o2info_from_user(*oiff, req)) {
640 * chunksize from userspace should be power of 2.
642 if ((oiff->iff_chunksize & (oiff->iff_chunksize - 1)) ||
643 (!oiff->iff_chunksize)) {
648 if (o2info_coherent(&oiff->iff_req)) {
649 gb_inode = ocfs2_get_system_file_inode(osb, type,
652 mlog(ML_ERROR, "unable to get global_bitmap inode\n");
657 ocfs2_sprintf_system_inode_name(namebuf, sizeof(namebuf), type,
659 status = ocfs2_lookup_ino_from_name(osb->sys_root_inode,
669 status = ocfs2_info_freefrag_scan_bitmap(osb, gb_inode, blkno, oiff);
673 o2info_set_request_filled(&oiff->iff_req);
675 if (o2info_to_user(*oiff, req)) {
683 o2info_set_request_error(&oiff->iff_req, req);
690 static int ocfs2_info_handle_unknown(struct inode *inode,
691 struct ocfs2_info_request __user *req)
693 struct ocfs2_info_request oir;
695 if (o2info_from_user(oir, req))
698 o2info_clear_request_filled(&oir);
700 if (o2info_to_user(oir, req))
707 * Validate and distinguish OCFS2_IOC_INFO requests.
709 * - validate the magic number.
710 * - distinguish different requests.
711 * - validate size of different requests.
713 static int ocfs2_info_handle_request(struct inode *inode,
714 struct ocfs2_info_request __user *req)
716 int status = -EFAULT;
717 struct ocfs2_info_request oir;
719 if (o2info_from_user(oir, req))
723 if (oir.ir_magic != OCFS2_INFO_MAGIC)
726 switch (oir.ir_code) {
727 case OCFS2_INFO_BLOCKSIZE:
728 if (oir.ir_size == sizeof(struct ocfs2_info_blocksize))
729 status = ocfs2_info_handle_blocksize(inode, req);
731 case OCFS2_INFO_CLUSTERSIZE:
732 if (oir.ir_size == sizeof(struct ocfs2_info_clustersize))
733 status = ocfs2_info_handle_clustersize(inode, req);
735 case OCFS2_INFO_MAXSLOTS:
736 if (oir.ir_size == sizeof(struct ocfs2_info_maxslots))
737 status = ocfs2_info_handle_maxslots(inode, req);
739 case OCFS2_INFO_LABEL:
740 if (oir.ir_size == sizeof(struct ocfs2_info_label))
741 status = ocfs2_info_handle_label(inode, req);
743 case OCFS2_INFO_UUID:
744 if (oir.ir_size == sizeof(struct ocfs2_info_uuid))
745 status = ocfs2_info_handle_uuid(inode, req);
747 case OCFS2_INFO_FS_FEATURES:
748 if (oir.ir_size == sizeof(struct ocfs2_info_fs_features))
749 status = ocfs2_info_handle_fs_features(inode, req);
751 case OCFS2_INFO_JOURNAL_SIZE:
752 if (oir.ir_size == sizeof(struct ocfs2_info_journal_size))
753 status = ocfs2_info_handle_journal_size(inode, req);
755 case OCFS2_INFO_FREEINODE:
756 if (oir.ir_size == sizeof(struct ocfs2_info_freeinode))
757 status = ocfs2_info_handle_freeinode(inode, req);
759 case OCFS2_INFO_FREEFRAG:
760 if (oir.ir_size == sizeof(struct ocfs2_info_freefrag))
761 status = ocfs2_info_handle_freefrag(inode, req);
764 status = ocfs2_info_handle_unknown(inode, req);
772 static int ocfs2_get_request_ptr(struct ocfs2_info *info, int idx,
773 u64 *req_addr, int compat_flag)
775 int status = -EFAULT;
776 u64 __user *bp = NULL;
781 * pointer bp stores the base address of a pointers array,
782 * which collects all addresses of separate request.
784 bp = (u64 __user *)(unsigned long)compat_ptr(info->oi_requests);
789 bp = (u64 __user *)(unsigned long)(info->oi_requests);
791 if (o2info_from_user(*req_addr, bp + idx))
800 * OCFS2_IOC_INFO handles an array of requests passed from userspace.
802 * ocfs2_info_handle() recevies a large info aggregation, grab and
803 * validate the request count from header, then break it into small
804 * pieces, later specific handlers can handle them one by one.
806 * Idea here is to make each separate request small enough to ensure
807 * a better backward&forward compatibility, since a small piece of
808 * request will be less likely to be broken if disk layout get changed.
810 static int ocfs2_info_handle(struct inode *inode, struct ocfs2_info *info,
815 struct ocfs2_info_request __user *reqp;
817 if ((info->oi_count > OCFS2_INFO_MAX_REQUEST) ||
818 (!info->oi_requests)) {
823 for (i = 0; i < info->oi_count; i++) {
825 status = ocfs2_get_request_ptr(info, i, &req_addr, compat_flag);
829 reqp = (struct ocfs2_info_request __user *)(unsigned long)req_addr;
835 status = ocfs2_info_handle_request(inode, reqp);
844 long ocfs2_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
846 struct inode *inode = file_inode(filp);
850 struct ocfs2_space_resv sr;
851 struct ocfs2_new_group_input input;
852 struct reflink_arguments args;
853 const char __user *old_path;
854 const char __user *new_path;
856 struct ocfs2_info info;
857 void __user *argp = (void __user *)arg;
860 case OCFS2_IOC_GETFLAGS:
861 status = ocfs2_get_inode_attr(inode, &flags);
865 flags &= OCFS2_FL_VISIBLE;
866 return put_user(flags, (int __user *) arg);
867 case OCFS2_IOC_SETFLAGS:
868 if (get_user(flags, (int __user *) arg))
871 status = mnt_want_write_file(filp);
874 status = ocfs2_set_inode_attr(inode, flags,
875 OCFS2_FL_MODIFIABLE);
876 mnt_drop_write_file(filp);
878 case OCFS2_IOC_RESVSP:
879 case OCFS2_IOC_RESVSP64:
880 case OCFS2_IOC_UNRESVSP:
881 case OCFS2_IOC_UNRESVSP64:
882 if (copy_from_user(&sr, (int __user *) arg, sizeof(sr)))
885 return ocfs2_change_file_space(filp, cmd, &sr);
886 case OCFS2_IOC_GROUP_EXTEND:
887 if (!capable(CAP_SYS_RESOURCE))
890 if (get_user(new_clusters, (int __user *)arg))
893 status = mnt_want_write_file(filp);
896 status = ocfs2_group_extend(inode, new_clusters);
897 mnt_drop_write_file(filp);
899 case OCFS2_IOC_GROUP_ADD:
900 case OCFS2_IOC_GROUP_ADD64:
901 if (!capable(CAP_SYS_RESOURCE))
904 if (copy_from_user(&input, (int __user *) arg, sizeof(input)))
907 status = mnt_want_write_file(filp);
910 status = ocfs2_group_add(inode, &input);
911 mnt_drop_write_file(filp);
913 case OCFS2_IOC_REFLINK:
914 if (copy_from_user(&args, argp, sizeof(args)))
916 old_path = (const char __user *)(unsigned long)args.old_path;
917 new_path = (const char __user *)(unsigned long)args.new_path;
918 preserve = (args.preserve != 0);
920 return ocfs2_reflink_ioctl(inode, old_path, new_path, preserve);
922 if (copy_from_user(&info, argp, sizeof(struct ocfs2_info)))
925 return ocfs2_info_handle(inode, &info, 0);
928 struct super_block *sb = inode->i_sb;
929 struct request_queue *q = bdev_get_queue(sb->s_bdev);
930 struct fstrim_range range;
933 if (!capable(CAP_SYS_ADMIN))
936 if (!blk_queue_discard(q))
939 if (copy_from_user(&range, argp, sizeof(range)))
942 range.minlen = max_t(u64, q->limits.discard_granularity,
944 ret = ocfs2_trim_fs(sb, &range);
948 if (copy_to_user(argp, &range, sizeof(range)))
953 case OCFS2_IOC_MOVE_EXT:
954 return ocfs2_ioctl_move_extents(filp, argp);
961 long ocfs2_compat_ioctl(struct file *file, unsigned cmd, unsigned long arg)
964 struct reflink_arguments args;
965 struct inode *inode = file_inode(file);
966 struct ocfs2_info info;
967 void __user *argp = (void __user *)arg;
970 case OCFS2_IOC32_GETFLAGS:
971 cmd = OCFS2_IOC_GETFLAGS;
973 case OCFS2_IOC32_SETFLAGS:
974 cmd = OCFS2_IOC_SETFLAGS;
976 case OCFS2_IOC_RESVSP:
977 case OCFS2_IOC_RESVSP64:
978 case OCFS2_IOC_UNRESVSP:
979 case OCFS2_IOC_UNRESVSP64:
980 case OCFS2_IOC_GROUP_EXTEND:
981 case OCFS2_IOC_GROUP_ADD:
982 case OCFS2_IOC_GROUP_ADD64:
984 case OCFS2_IOC_REFLINK:
985 if (copy_from_user(&args, argp, sizeof(args)))
987 preserve = (args.preserve != 0);
989 return ocfs2_reflink_ioctl(inode, compat_ptr(args.old_path),
990 compat_ptr(args.new_path), preserve);
992 if (copy_from_user(&info, argp, sizeof(struct ocfs2_info)))
995 return ocfs2_info_handle(inode, &info, 1);
996 case OCFS2_IOC_MOVE_EXT:
1002 return ocfs2_ioctl(file, cmd, arg);