1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26#include <linux/capability.h>
27#include <linux/fs.h>
28#include <linux/types.h>
29#include <linux/slab.h>
30#include <linux/highmem.h>
31#include <linux/pagemap.h>
32#include <linux/uio.h>
33#include <linux/sched.h>
34#include <linux/splice.h>
35#include <linux/mount.h>
36#include <linux/writeback.h>
37#include <linux/falloc.h>
38#include <linux/quotaops.h>
39#include <linux/blkdev.h>
40#include <linux/backing-dev.h>
41
42#include <cluster/masklog.h>
43
44#include "ocfs2.h"
45
46#include "alloc.h"
47#include "aops.h"
48#include "dir.h"
49#include "dlmglue.h"
50#include "extent_map.h"
51#include "file.h"
52#include "sysfile.h"
53#include "inode.h"
54#include "ioctl.h"
55#include "journal.h"
56#include "locks.h"
57#include "mmap.h"
58#include "suballoc.h"
59#include "super.h"
60#include "xattr.h"
61#include "acl.h"
62#include "quota.h"
63#include "refcounttree.h"
64#include "ocfs2_trace.h"
65
66#include "buffer_head_io.h"
67
68static int ocfs2_init_file_private(struct inode *inode, struct file *file)
69{
70 struct ocfs2_file_private *fp;
71
72 fp = kzalloc(sizeof(struct ocfs2_file_private), GFP_KERNEL);
73 if (!fp)
74 return -ENOMEM;
75
76 fp->fp_file = file;
77 mutex_init(&fp->fp_mutex);
78 ocfs2_file_lock_res_init(&fp->fp_flock, fp);
79 file->private_data = fp;
80
81 return 0;
82}
83
84static void ocfs2_free_file_private(struct inode *inode, struct file *file)
85{
86 struct ocfs2_file_private *fp = file->private_data;
87 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
88
89 if (fp) {
90 ocfs2_simple_drop_lockres(osb, &fp->fp_flock);
91 ocfs2_lock_res_free(&fp->fp_flock);
92 kfree(fp);
93 file->private_data = NULL;
94 }
95}
96
97static int ocfs2_file_open(struct inode *inode, struct file *file)
98{
99 int status;
100 int mode = file->f_flags;
101 struct ocfs2_inode_info *oi = OCFS2_I(inode);
102
103 trace_ocfs2_file_open(inode, file, file->f_path.dentry,
104 (unsigned long long)OCFS2_I(inode)->ip_blkno,
105 file->f_path.dentry->d_name.len,
106 file->f_path.dentry->d_name.name, mode);
107
108 if (file->f_mode & FMODE_WRITE) {
109 status = dquot_initialize(inode);
110 if (status)
111 goto leave;
112 }
113
114 spin_lock(&oi->ip_lock);
115
116
117
118
119 if (OCFS2_I(inode)->ip_flags & OCFS2_INODE_DELETED) {
120 spin_unlock(&oi->ip_lock);
121
122 status = -ENOENT;
123 goto leave;
124 }
125
126 if (mode & O_DIRECT)
127 oi->ip_flags |= OCFS2_INODE_OPEN_DIRECT;
128
129 oi->ip_open_count++;
130 spin_unlock(&oi->ip_lock);
131
132 status = ocfs2_init_file_private(inode, file);
133 if (status) {
134
135
136
137
138 spin_lock(&oi->ip_lock);
139 oi->ip_open_count--;
140 spin_unlock(&oi->ip_lock);
141 }
142
143leave:
144 return status;
145}
146
147static int ocfs2_file_release(struct inode *inode, struct file *file)
148{
149 struct ocfs2_inode_info *oi = OCFS2_I(inode);
150
151 spin_lock(&oi->ip_lock);
152 if (!--oi->ip_open_count)
153 oi->ip_flags &= ~OCFS2_INODE_OPEN_DIRECT;
154
155 trace_ocfs2_file_release(inode, file, file->f_path.dentry,
156 oi->ip_blkno,
157 file->f_path.dentry->d_name.len,
158 file->f_path.dentry->d_name.name,
159 oi->ip_open_count);
160 spin_unlock(&oi->ip_lock);
161
162 ocfs2_free_file_private(inode, file);
163
164 return 0;
165}
166
167static int ocfs2_dir_open(struct inode *inode, struct file *file)
168{
169 return ocfs2_init_file_private(inode, file);
170}
171
172static int ocfs2_dir_release(struct inode *inode, struct file *file)
173{
174 ocfs2_free_file_private(inode, file);
175 return 0;
176}
177
178static int ocfs2_sync_file(struct file *file, loff_t start, loff_t end,
179 int datasync)
180{
181 int err = 0;
182 struct inode *inode = file->f_mapping->host;
183 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
184 struct ocfs2_inode_info *oi = OCFS2_I(inode);
185 journal_t *journal = osb->journal->j_journal;
186 int ret;
187 tid_t commit_tid;
188 bool needs_barrier = false;
189
190 trace_ocfs2_sync_file(inode, file, file->f_path.dentry,
191 OCFS2_I(inode)->ip_blkno,
192 file->f_path.dentry->d_name.len,
193 file->f_path.dentry->d_name.name,
194 (unsigned long long)datasync);
195
196 if (ocfs2_is_hard_readonly(osb) || ocfs2_is_soft_readonly(osb))
197 return -EROFS;
198
199 err = filemap_write_and_wait_range(inode->i_mapping, start, end);
200 if (err)
201 return err;
202
203 commit_tid = datasync ? oi->i_datasync_tid : oi->i_sync_tid;
204 if (journal->j_flags & JBD2_BARRIER &&
205 !jbd2_trans_will_send_data_barrier(journal, commit_tid))
206 needs_barrier = true;
207 err = jbd2_complete_transaction(journal, commit_tid);
208 if (needs_barrier) {
209 ret = blkdev_issue_flush(inode->i_sb->s_bdev, GFP_KERNEL, NULL);
210 if (!err)
211 err = ret;
212 }
213
214 if (err)
215 mlog_errno(err);
216
217 return (err < 0) ? -EIO : 0;
218}
219
220int ocfs2_should_update_atime(struct inode *inode,
221 struct vfsmount *vfsmnt)
222{
223 struct timespec now;
224 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
225
226 if (ocfs2_is_hard_readonly(osb) || ocfs2_is_soft_readonly(osb))
227 return 0;
228
229 if ((inode->i_flags & S_NOATIME) ||
230 ((inode->i_sb->s_flags & MS_NODIRATIME) && S_ISDIR(inode->i_mode)))
231 return 0;
232
233
234
235
236
237
238
239
240
241 if (vfsmnt == NULL)
242 return 0;
243
244 if ((vfsmnt->mnt_flags & MNT_NOATIME) ||
245 ((vfsmnt->mnt_flags & MNT_NODIRATIME) && S_ISDIR(inode->i_mode)))
246 return 0;
247
248 if (vfsmnt->mnt_flags & MNT_RELATIME) {
249 if ((timespec_compare(&inode->i_atime, &inode->i_mtime) <= 0) ||
250 (timespec_compare(&inode->i_atime, &inode->i_ctime) <= 0))
251 return 1;
252
253 return 0;
254 }
255
256 now = CURRENT_TIME;
257 if ((now.tv_sec - inode->i_atime.tv_sec <= osb->s_atime_quantum))
258 return 0;
259 else
260 return 1;
261}
262
263int ocfs2_update_inode_atime(struct inode *inode,
264 struct buffer_head *bh)
265{
266 int ret;
267 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
268 handle_t *handle;
269 struct ocfs2_dinode *di = (struct ocfs2_dinode *) bh->b_data;
270
271 handle = ocfs2_start_trans(osb, OCFS2_INODE_UPDATE_CREDITS);
272 if (IS_ERR(handle)) {
273 ret = PTR_ERR(handle);
274 mlog_errno(ret);
275 goto out;
276 }
277
278 ret = ocfs2_journal_access_di(handle, INODE_CACHE(inode), bh,
279 OCFS2_JOURNAL_ACCESS_WRITE);
280 if (ret) {
281 mlog_errno(ret);
282 goto out_commit;
283 }
284
285
286
287
288
289
290 inode->i_atime = CURRENT_TIME;
291 di->i_atime = cpu_to_le64(inode->i_atime.tv_sec);
292 di->i_atime_nsec = cpu_to_le32(inode->i_atime.tv_nsec);
293 ocfs2_update_inode_fsync_trans(handle, inode, 0);
294 ocfs2_journal_dirty(handle, bh);
295
296out_commit:
297 ocfs2_commit_trans(OCFS2_SB(inode->i_sb), handle);
298out:
299 return ret;
300}
301
302int ocfs2_set_inode_size(handle_t *handle,
303 struct inode *inode,
304 struct buffer_head *fe_bh,
305 u64 new_i_size)
306{
307 int status;
308
309 i_size_write(inode, new_i_size);
310 inode->i_blocks = ocfs2_inode_sector_count(inode);
311 inode->i_ctime = inode->i_mtime = CURRENT_TIME;
312
313 status = ocfs2_mark_inode_dirty(handle, inode, fe_bh);
314 if (status < 0) {
315 mlog_errno(status);
316 goto bail;
317 }
318
319bail:
320 return status;
321}
322
323int ocfs2_simple_size_update(struct inode *inode,
324 struct buffer_head *di_bh,
325 u64 new_i_size)
326{
327 int ret;
328 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
329 handle_t *handle = NULL;
330
331 handle = ocfs2_start_trans(osb, OCFS2_INODE_UPDATE_CREDITS);
332 if (IS_ERR(handle)) {
333 ret = PTR_ERR(handle);
334 mlog_errno(ret);
335 goto out;
336 }
337
338 ret = ocfs2_set_inode_size(handle, inode, di_bh,
339 new_i_size);
340 if (ret < 0)
341 mlog_errno(ret);
342
343 ocfs2_update_inode_fsync_trans(handle, inode, 0);
344 ocfs2_commit_trans(osb, handle);
345out:
346 return ret;
347}
348
349static int ocfs2_cow_file_pos(struct inode *inode,
350 struct buffer_head *fe_bh,
351 u64 offset)
352{
353 int status;
354 u32 phys, cpos = offset >> OCFS2_SB(inode->i_sb)->s_clustersize_bits;
355 unsigned int num_clusters = 0;
356 unsigned int ext_flags = 0;
357
358
359
360
361
362
363 if ((offset & (OCFS2_SB(inode->i_sb)->s_clustersize - 1)) == 0)
364 return 0;
365
366 status = ocfs2_get_clusters(inode, cpos, &phys,
367 &num_clusters, &ext_flags);
368 if (status) {
369 mlog_errno(status);
370 goto out;
371 }
372
373 if (!(ext_flags & OCFS2_EXT_REFCOUNTED))
374 goto out;
375
376 return ocfs2_refcount_cow(inode, fe_bh, cpos, 1, cpos+1);
377
378out:
379 return status;
380}
381
382static int ocfs2_orphan_for_truncate(struct ocfs2_super *osb,
383 struct inode *inode,
384 struct buffer_head *fe_bh,
385 u64 new_i_size)
386{
387 int status;
388 handle_t *handle;
389 struct ocfs2_dinode *di;
390 u64 cluster_bytes;
391
392
393
394
395
396
397 status = ocfs2_cow_file_pos(inode, fe_bh, new_i_size);
398 if (status) {
399 mlog_errno(status);
400 return status;
401 }
402
403
404
405
406 handle = ocfs2_start_trans(osb, OCFS2_INODE_UPDATE_CREDITS);
407 if (IS_ERR(handle)) {
408 status = PTR_ERR(handle);
409 mlog_errno(status);
410 goto out;
411 }
412
413 status = ocfs2_journal_access_di(handle, INODE_CACHE(inode), fe_bh,
414 OCFS2_JOURNAL_ACCESS_WRITE);
415 if (status < 0) {
416 mlog_errno(status);
417 goto out_commit;
418 }
419
420
421
422
423 cluster_bytes = ocfs2_align_bytes_to_clusters(inode->i_sb, new_i_size);
424 status = ocfs2_zero_range_for_truncate(inode, handle, new_i_size,
425 cluster_bytes);
426 if (status) {
427 mlog_errno(status);
428 goto out_commit;
429 }
430
431 i_size_write(inode, new_i_size);
432 inode->i_ctime = inode->i_mtime = CURRENT_TIME;
433
434 di = (struct ocfs2_dinode *) fe_bh->b_data;
435 di->i_size = cpu_to_le64(new_i_size);
436 di->i_ctime = di->i_mtime = cpu_to_le64(inode->i_ctime.tv_sec);
437 di->i_ctime_nsec = di->i_mtime_nsec = cpu_to_le32(inode->i_ctime.tv_nsec);
438 ocfs2_update_inode_fsync_trans(handle, inode, 0);
439
440 ocfs2_journal_dirty(handle, fe_bh);
441
442out_commit:
443 ocfs2_commit_trans(osb, handle);
444out:
445 return status;
446}
447
448int ocfs2_truncate_file(struct inode *inode,
449 struct buffer_head *di_bh,
450 u64 new_i_size)
451{
452 int status = 0;
453 struct ocfs2_dinode *fe = NULL;
454 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
455
456
457
458 fe = (struct ocfs2_dinode *) di_bh->b_data;
459
460 trace_ocfs2_truncate_file((unsigned long long)OCFS2_I(inode)->ip_blkno,
461 (unsigned long long)le64_to_cpu(fe->i_size),
462 (unsigned long long)new_i_size);
463
464 mlog_bug_on_msg(le64_to_cpu(fe->i_size) != i_size_read(inode),
465 "Inode %llu, inode i_size = %lld != di "
466 "i_size = %llu, i_flags = 0x%x\n",
467 (unsigned long long)OCFS2_I(inode)->ip_blkno,
468 i_size_read(inode),
469 (unsigned long long)le64_to_cpu(fe->i_size),
470 le32_to_cpu(fe->i_flags));
471
472 if (new_i_size > le64_to_cpu(fe->i_size)) {
473 trace_ocfs2_truncate_file_error(
474 (unsigned long long)le64_to_cpu(fe->i_size),
475 (unsigned long long)new_i_size);
476 status = -EINVAL;
477 mlog_errno(status);
478 goto bail;
479 }
480
481 down_write(&OCFS2_I(inode)->ip_alloc_sem);
482
483 ocfs2_resv_discard(&osb->osb_la_resmap,
484 &OCFS2_I(inode)->ip_la_data_resv);
485
486
487
488
489
490
491
492
493 unmap_mapping_range(inode->i_mapping, new_i_size + PAGE_SIZE - 1, 0, 1);
494 truncate_inode_pages(inode->i_mapping, new_i_size);
495
496 if (OCFS2_I(inode)->ip_dyn_features & OCFS2_INLINE_DATA_FL) {
497 status = ocfs2_truncate_inline(inode, di_bh, new_i_size,
498 i_size_read(inode), 1);
499 if (status)
500 mlog_errno(status);
501
502 goto bail_unlock_sem;
503 }
504
505
506
507
508
509 status = ocfs2_orphan_for_truncate(osb, inode, di_bh, new_i_size);
510 if (status < 0) {
511 mlog_errno(status);
512 goto bail_unlock_sem;
513 }
514
515 status = ocfs2_commit_truncate(osb, inode, di_bh);
516 if (status < 0) {
517 mlog_errno(status);
518 goto bail_unlock_sem;
519 }
520
521
522bail_unlock_sem:
523 up_write(&OCFS2_I(inode)->ip_alloc_sem);
524
525bail:
526 if (!status && OCFS2_I(inode)->ip_clusters == 0)
527 status = ocfs2_try_remove_refcount_tree(inode, di_bh);
528
529 return status;
530}
531
532
533
534
535
536
537
538
539
540
541
542int ocfs2_add_inode_data(struct ocfs2_super *osb,
543 struct inode *inode,
544 u32 *logical_offset,
545 u32 clusters_to_add,
546 int mark_unwritten,
547 struct buffer_head *fe_bh,
548 handle_t *handle,
549 struct ocfs2_alloc_context *data_ac,
550 struct ocfs2_alloc_context *meta_ac,
551 enum ocfs2_alloc_restarted *reason_ret)
552{
553 int ret;
554 struct ocfs2_extent_tree et;
555
556 ocfs2_init_dinode_extent_tree(&et, INODE_CACHE(inode), fe_bh);
557 ret = ocfs2_add_clusters_in_btree(handle, &et, logical_offset,
558 clusters_to_add, mark_unwritten,
559 data_ac, meta_ac, reason_ret);
560
561 return ret;
562}
563
564static int __ocfs2_extend_allocation(struct inode *inode, u32 logical_start,
565 u32 clusters_to_add, int mark_unwritten)
566{
567 int status = 0;
568 int restart_func = 0;
569 int credits;
570 u32 prev_clusters;
571 struct buffer_head *bh = NULL;
572 struct ocfs2_dinode *fe = NULL;
573 handle_t *handle = NULL;
574 struct ocfs2_alloc_context *data_ac = NULL;
575 struct ocfs2_alloc_context *meta_ac = NULL;
576 enum ocfs2_alloc_restarted why = RESTART_NONE;
577 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
578 struct ocfs2_extent_tree et;
579 int did_quota = 0;
580
581
582
583
584
585 BUG_ON(mark_unwritten && !ocfs2_sparse_alloc(osb));
586
587 status = ocfs2_read_inode_block(inode, &bh);
588 if (status < 0) {
589 mlog_errno(status);
590 goto leave;
591 }
592 fe = (struct ocfs2_dinode *) bh->b_data;
593
594restart_all:
595 BUG_ON(le32_to_cpu(fe->i_clusters) != OCFS2_I(inode)->ip_clusters);
596
597 ocfs2_init_dinode_extent_tree(&et, INODE_CACHE(inode), bh);
598 status = ocfs2_lock_allocators(inode, &et, clusters_to_add, 0,
599 &data_ac, &meta_ac);
600 if (status) {
601 mlog_errno(status);
602 goto leave;
603 }
604
605 credits = ocfs2_calc_extend_credits(osb->sb, &fe->id2.i_list);
606 handle = ocfs2_start_trans(osb, credits);
607 if (IS_ERR(handle)) {
608 status = PTR_ERR(handle);
609 handle = NULL;
610 mlog_errno(status);
611 goto leave;
612 }
613
614restarted_transaction:
615 trace_ocfs2_extend_allocation(
616 (unsigned long long)OCFS2_I(inode)->ip_blkno,
617 (unsigned long long)i_size_read(inode),
618 le32_to_cpu(fe->i_clusters), clusters_to_add,
619 why, restart_func);
620
621 status = dquot_alloc_space_nodirty(inode,
622 ocfs2_clusters_to_bytes(osb->sb, clusters_to_add));
623 if (status)
624 goto leave;
625 did_quota = 1;
626
627
628
629
630 status = ocfs2_journal_access_di(handle, INODE_CACHE(inode), bh,
631 OCFS2_JOURNAL_ACCESS_WRITE);
632 if (status < 0) {
633 mlog_errno(status);
634 goto leave;
635 }
636
637 prev_clusters = OCFS2_I(inode)->ip_clusters;
638
639 status = ocfs2_add_inode_data(osb,
640 inode,
641 &logical_start,
642 clusters_to_add,
643 mark_unwritten,
644 bh,
645 handle,
646 data_ac,
647 meta_ac,
648 &why);
649 if ((status < 0) && (status != -EAGAIN)) {
650 if (status != -ENOSPC)
651 mlog_errno(status);
652 goto leave;
653 }
654 ocfs2_update_inode_fsync_trans(handle, inode, 1);
655 ocfs2_journal_dirty(handle, bh);
656
657 spin_lock(&OCFS2_I(inode)->ip_lock);
658 clusters_to_add -= (OCFS2_I(inode)->ip_clusters - prev_clusters);
659 spin_unlock(&OCFS2_I(inode)->ip_lock);
660
661 dquot_free_space(inode,
662 ocfs2_clusters_to_bytes(osb->sb, clusters_to_add));
663 did_quota = 0;
664
665 if (why != RESTART_NONE && clusters_to_add) {
666 if (why == RESTART_META) {
667 restart_func = 1;
668 status = 0;
669 } else {
670 BUG_ON(why != RESTART_TRANS);
671
672 status = ocfs2_allocate_extend_trans(handle, 1);
673 if (status < 0) {
674
675
676 status = -ENOMEM;
677 mlog_errno(status);
678 goto leave;
679 }
680 goto restarted_transaction;
681 }
682 }
683
684 trace_ocfs2_extend_allocation_end(OCFS2_I(inode)->ip_blkno,
685 le32_to_cpu(fe->i_clusters),
686 (unsigned long long)le64_to_cpu(fe->i_size),
687 OCFS2_I(inode)->ip_clusters,
688 (unsigned long long)i_size_read(inode));
689
690leave:
691 if (status < 0 && did_quota)
692 dquot_free_space(inode,
693 ocfs2_clusters_to_bytes(osb->sb, clusters_to_add));
694 if (handle) {
695 ocfs2_commit_trans(osb, handle);
696 handle = NULL;
697 }
698 if (data_ac) {
699 ocfs2_free_alloc_context(data_ac);
700 data_ac = NULL;
701 }
702 if (meta_ac) {
703 ocfs2_free_alloc_context(meta_ac);
704 meta_ac = NULL;
705 }
706 if ((!status) && restart_func) {
707 restart_func = 0;
708 goto restart_all;
709 }
710 brelse(bh);
711 bh = NULL;
712
713 return status;
714}
715
716int ocfs2_extend_allocation(struct inode *inode, u32 logical_start,
717 u32 clusters_to_add, int mark_unwritten)
718{
719 return __ocfs2_extend_allocation(inode, logical_start,
720 clusters_to_add, mark_unwritten);
721}
722
723
724
725
726
727static handle_t *ocfs2_zero_start_ordered_transaction(struct inode *inode,
728 struct buffer_head *di_bh)
729{
730 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
731 handle_t *handle = NULL;
732 int ret = 0;
733
734 if (!ocfs2_should_order_data(inode))
735 goto out;
736
737 handle = ocfs2_start_trans(osb, OCFS2_INODE_UPDATE_CREDITS);
738 if (IS_ERR(handle)) {
739 ret = -ENOMEM;
740 mlog_errno(ret);
741 goto out;
742 }
743
744 ret = ocfs2_jbd2_file_inode(handle, inode);
745 if (ret < 0) {
746 mlog_errno(ret);
747 goto out;
748 }
749
750 ret = ocfs2_journal_access_di(handle, INODE_CACHE(inode), di_bh,
751 OCFS2_JOURNAL_ACCESS_WRITE);
752 if (ret)
753 mlog_errno(ret);
754 ocfs2_update_inode_fsync_trans(handle, inode, 1);
755
756out:
757 if (ret) {
758 if (!IS_ERR(handle))
759 ocfs2_commit_trans(osb, handle);
760 handle = ERR_PTR(ret);
761 }
762 return handle;
763}
764
765
766
767
768static int ocfs2_write_zero_page(struct inode *inode, u64 abs_from,
769 u64 abs_to, struct buffer_head *di_bh)
770{
771 struct address_space *mapping = inode->i_mapping;
772 struct page *page;
773 unsigned long index = abs_from >> PAGE_CACHE_SHIFT;
774 handle_t *handle;
775 int ret = 0;
776 unsigned zero_from, zero_to, block_start, block_end;
777 struct ocfs2_dinode *di = (struct ocfs2_dinode *)di_bh->b_data;
778
779 BUG_ON(abs_from >= abs_to);
780 BUG_ON(abs_to > (((u64)index + 1) << PAGE_CACHE_SHIFT));
781 BUG_ON(abs_from & (inode->i_blkbits - 1));
782
783 handle = ocfs2_zero_start_ordered_transaction(inode, di_bh);
784 if (IS_ERR(handle)) {
785 ret = PTR_ERR(handle);
786 goto out;
787 }
788
789 page = find_or_create_page(mapping, index, GFP_NOFS);
790 if (!page) {
791 ret = -ENOMEM;
792 mlog_errno(ret);
793 goto out_commit_trans;
794 }
795
796
797 zero_from = abs_from & (PAGE_CACHE_SIZE - 1);
798 zero_to = abs_to & (PAGE_CACHE_SIZE - 1);
799 if (!zero_to)
800 zero_to = PAGE_CACHE_SIZE;
801
802 trace_ocfs2_write_zero_page(
803 (unsigned long long)OCFS2_I(inode)->ip_blkno,
804 (unsigned long long)abs_from,
805 (unsigned long long)abs_to,
806 index, zero_from, zero_to);
807
808
809 for (block_start = zero_from; block_start < zero_to;
810 block_start = block_end) {
811 block_end = block_start + (1 << inode->i_blkbits);
812
813
814
815
816
817
818 ret = __block_write_begin(page, block_start + 1, 0,
819 ocfs2_get_block);
820 if (ret < 0) {
821 mlog_errno(ret);
822 goto out_unlock;
823 }
824
825
826
827 ret = block_commit_write(page, block_start + 1,
828 block_start + 1);
829 if (ret < 0)
830 mlog_errno(ret);
831 else
832 ret = 0;
833 }
834
835
836
837
838
839
840 i_size_write(inode, abs_to);
841 inode->i_blocks = ocfs2_inode_sector_count(inode);
842 di->i_size = cpu_to_le64((u64)i_size_read(inode));
843 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
844 di->i_mtime = di->i_ctime = cpu_to_le64(inode->i_mtime.tv_sec);
845 di->i_ctime_nsec = cpu_to_le32(inode->i_mtime.tv_nsec);
846 di->i_mtime_nsec = di->i_ctime_nsec;
847 if (handle) {
848 ocfs2_journal_dirty(handle, di_bh);
849 ocfs2_update_inode_fsync_trans(handle, inode, 1);
850 }
851
852out_unlock:
853 unlock_page(page);
854 page_cache_release(page);
855out_commit_trans:
856 if (handle)
857 ocfs2_commit_trans(OCFS2_SB(inode->i_sb), handle);
858out:
859 return ret;
860}
861
862
863
864
865
866
867
868
869
870
871
872
873
874static int ocfs2_zero_extend_get_range(struct inode *inode,
875 struct buffer_head *di_bh,
876 u64 zero_start, u64 zero_end,
877 u64 *range_start, u64 *range_end)
878{
879 int rc = 0, needs_cow = 0;
880 u32 p_cpos, zero_clusters = 0;
881 u32 zero_cpos =
882 zero_start >> OCFS2_SB(inode->i_sb)->s_clustersize_bits;
883 u32 last_cpos = ocfs2_clusters_for_bytes(inode->i_sb, zero_end);
884 unsigned int num_clusters = 0;
885 unsigned int ext_flags = 0;
886
887 while (zero_cpos < last_cpos) {
888 rc = ocfs2_get_clusters(inode, zero_cpos, &p_cpos,
889 &num_clusters, &ext_flags);
890 if (rc) {
891 mlog_errno(rc);
892 goto out;
893 }
894
895 if (p_cpos && !(ext_flags & OCFS2_EXT_UNWRITTEN)) {
896 zero_clusters = num_clusters;
897 if (ext_flags & OCFS2_EXT_REFCOUNTED)
898 needs_cow = 1;
899 break;
900 }
901
902 zero_cpos += num_clusters;
903 }
904 if (!zero_clusters) {
905 *range_end = 0;
906 goto out;
907 }
908
909 while ((zero_cpos + zero_clusters) < last_cpos) {
910 rc = ocfs2_get_clusters(inode, zero_cpos + zero_clusters,
911 &p_cpos, &num_clusters,
912 &ext_flags);
913 if (rc) {
914 mlog_errno(rc);
915 goto out;
916 }
917
918 if (!p_cpos || (ext_flags & OCFS2_EXT_UNWRITTEN))
919 break;
920 if (ext_flags & OCFS2_EXT_REFCOUNTED)
921 needs_cow = 1;
922 zero_clusters += num_clusters;
923 }
924 if ((zero_cpos + zero_clusters) > last_cpos)
925 zero_clusters = last_cpos - zero_cpos;
926
927 if (needs_cow) {
928 rc = ocfs2_refcount_cow(inode, di_bh, zero_cpos,
929 zero_clusters, UINT_MAX);
930 if (rc) {
931 mlog_errno(rc);
932 goto out;
933 }
934 }
935
936 *range_start = ocfs2_clusters_to_bytes(inode->i_sb, zero_cpos);
937 *range_end = ocfs2_clusters_to_bytes(inode->i_sb,
938 zero_cpos + zero_clusters);
939
940out:
941 return rc;
942}
943
944
945
946
947
948static int ocfs2_zero_extend_range(struct inode *inode, u64 range_start,
949 u64 range_end, struct buffer_head *di_bh)
950{
951 int rc = 0;
952 u64 next_pos;
953 u64 zero_pos = range_start;
954
955 trace_ocfs2_zero_extend_range(
956 (unsigned long long)OCFS2_I(inode)->ip_blkno,
957 (unsigned long long)range_start,
958 (unsigned long long)range_end);
959 BUG_ON(range_start >= range_end);
960
961 while (zero_pos < range_end) {
962 next_pos = (zero_pos & PAGE_CACHE_MASK) + PAGE_CACHE_SIZE;
963 if (next_pos > range_end)
964 next_pos = range_end;
965 rc = ocfs2_write_zero_page(inode, zero_pos, next_pos, di_bh);
966 if (rc < 0) {
967 mlog_errno(rc);
968 break;
969 }
970 zero_pos = next_pos;
971
972
973
974
975
976 cond_resched();
977 }
978
979 return rc;
980}
981
982int ocfs2_zero_extend(struct inode *inode, struct buffer_head *di_bh,
983 loff_t zero_to_size)
984{
985 int ret = 0;
986 u64 zero_start, range_start = 0, range_end = 0;
987 struct super_block *sb = inode->i_sb;
988
989 zero_start = ocfs2_align_bytes_to_blocks(sb, i_size_read(inode));
990 trace_ocfs2_zero_extend((unsigned long long)OCFS2_I(inode)->ip_blkno,
991 (unsigned long long)zero_start,
992 (unsigned long long)i_size_read(inode));
993 while (zero_start < zero_to_size) {
994 ret = ocfs2_zero_extend_get_range(inode, di_bh, zero_start,
995 zero_to_size,
996 &range_start,
997 &range_end);
998 if (ret) {
999 mlog_errno(ret);
1000 break;
1001 }
1002 if (!range_end)
1003 break;
1004
1005 if (range_start < zero_start)
1006 range_start = zero_start;
1007 if (range_end > zero_to_size)
1008 range_end = zero_to_size;
1009
1010 ret = ocfs2_zero_extend_range(inode, range_start,
1011 range_end, di_bh);
1012 if (ret) {
1013 mlog_errno(ret);
1014 break;
1015 }
1016 zero_start = range_end;
1017 }
1018
1019 return ret;
1020}
1021
1022int ocfs2_extend_no_holes(struct inode *inode, struct buffer_head *di_bh,
1023 u64 new_i_size, u64 zero_to)
1024{
1025 int ret;
1026 u32 clusters_to_add;
1027 struct ocfs2_inode_info *oi = OCFS2_I(inode);
1028
1029
1030
1031
1032
1033 BUG_ON(!di_bh && (oi->ip_dyn_features & OCFS2_HAS_REFCOUNT_FL));
1034 BUG_ON(!di_bh && !(oi->ip_flags & OCFS2_INODE_SYSTEM_FILE));
1035
1036 clusters_to_add = ocfs2_clusters_for_bytes(inode->i_sb, new_i_size);
1037 if (clusters_to_add < oi->ip_clusters)
1038 clusters_to_add = 0;
1039 else
1040 clusters_to_add -= oi->ip_clusters;
1041
1042 if (clusters_to_add) {
1043 ret = __ocfs2_extend_allocation(inode, oi->ip_clusters,
1044 clusters_to_add, 0);
1045 if (ret) {
1046 mlog_errno(ret);
1047 goto out;
1048 }
1049 }
1050
1051
1052
1053
1054
1055
1056 ret = ocfs2_zero_extend(inode, di_bh, zero_to);
1057 if (ret < 0)
1058 mlog_errno(ret);
1059
1060out:
1061 return ret;
1062}
1063
1064static int ocfs2_extend_file(struct inode *inode,
1065 struct buffer_head *di_bh,
1066 u64 new_i_size)
1067{
1068 int ret = 0;
1069 struct ocfs2_inode_info *oi = OCFS2_I(inode);
1070
1071 BUG_ON(!di_bh);
1072
1073
1074 if (new_i_size == 0)
1075 goto out;
1076
1077 if (i_size_read(inode) == new_i_size)
1078 goto out;
1079 BUG_ON(new_i_size < i_size_read(inode));
1080
1081
1082
1083
1084
1085
1086
1087
1088 down_write(&oi->ip_alloc_sem);
1089
1090 if (oi->ip_dyn_features & OCFS2_INLINE_DATA_FL) {
1091
1092
1093
1094
1095 if (ocfs2_size_fits_inline_data(di_bh, new_i_size)) {
1096 up_write(&oi->ip_alloc_sem);
1097 goto out_update_size;
1098 }
1099
1100 ret = ocfs2_convert_inline_data_to_extents(inode, di_bh);
1101 if (ret) {
1102 up_write(&oi->ip_alloc_sem);
1103 mlog_errno(ret);
1104 goto out;
1105 }
1106 }
1107
1108 if (ocfs2_sparse_alloc(OCFS2_SB(inode->i_sb)))
1109 ret = ocfs2_zero_extend(inode, di_bh, new_i_size);
1110 else
1111 ret = ocfs2_extend_no_holes(inode, di_bh, new_i_size,
1112 new_i_size);
1113
1114 up_write(&oi->ip_alloc_sem);
1115
1116 if (ret < 0) {
1117 mlog_errno(ret);
1118 goto out;
1119 }
1120
1121out_update_size:
1122 ret = ocfs2_simple_size_update(inode, di_bh, new_i_size);
1123 if (ret < 0)
1124 mlog_errno(ret);
1125
1126out:
1127 return ret;
1128}
1129
1130int ocfs2_setattr(struct dentry *dentry, struct iattr *attr)
1131{
1132 int status = 0, size_change;
1133 int inode_locked = 0;
1134 struct inode *inode = d_inode(dentry);
1135 struct super_block *sb = inode->i_sb;
1136 struct ocfs2_super *osb = OCFS2_SB(sb);
1137 struct buffer_head *bh = NULL;
1138 handle_t *handle = NULL;
1139 struct dquot *transfer_to[MAXQUOTAS] = { };
1140 int qtype;
1141
1142 trace_ocfs2_setattr(inode, dentry,
1143 (unsigned long long)OCFS2_I(inode)->ip_blkno,
1144 dentry->d_name.len, dentry->d_name.name,
1145 attr->ia_valid, attr->ia_mode,
1146 from_kuid(&init_user_ns, attr->ia_uid),
1147 from_kgid(&init_user_ns, attr->ia_gid));
1148
1149
1150 if (S_ISLNK(inode->i_mode))
1151 attr->ia_valid &= ~ATTR_SIZE;
1152
1153#define OCFS2_VALID_ATTRS (ATTR_ATIME | ATTR_MTIME | ATTR_CTIME | ATTR_SIZE \
1154 | ATTR_GID | ATTR_UID | ATTR_MODE)
1155 if (!(attr->ia_valid & OCFS2_VALID_ATTRS))
1156 return 0;
1157
1158 status = inode_change_ok(inode, attr);
1159 if (status)
1160 return status;
1161
1162 if (is_quota_modification(inode, attr)) {
1163 status = dquot_initialize(inode);
1164 if (status)
1165 return status;
1166 }
1167 size_change = S_ISREG(inode->i_mode) && attr->ia_valid & ATTR_SIZE;
1168 if (size_change) {
1169 status = ocfs2_rw_lock(inode, 1);
1170 if (status < 0) {
1171 mlog_errno(status);
1172 goto bail;
1173 }
1174 }
1175
1176 status = ocfs2_inode_lock(inode, &bh, 1);
1177 if (status < 0) {
1178 if (status != -ENOENT)
1179 mlog_errno(status);
1180 goto bail_unlock_rw;
1181 }
1182 inode_locked = 1;
1183
1184 if (size_change) {
1185 status = inode_newsize_ok(inode, attr->ia_size);
1186 if (status)
1187 goto bail_unlock;
1188
1189 inode_dio_wait(inode);
1190
1191 if (i_size_read(inode) >= attr->ia_size) {
1192 if (ocfs2_should_order_data(inode)) {
1193 status = ocfs2_begin_ordered_truncate(inode,
1194 attr->ia_size);
1195 if (status)
1196 goto bail_unlock;
1197 }
1198 status = ocfs2_truncate_file(inode, bh, attr->ia_size);
1199 } else
1200 status = ocfs2_extend_file(inode, bh, attr->ia_size);
1201 if (status < 0) {
1202 if (status != -ENOSPC)
1203 mlog_errno(status);
1204 status = -ENOSPC;
1205 goto bail_unlock;
1206 }
1207 }
1208
1209 if ((attr->ia_valid & ATTR_UID && !uid_eq(attr->ia_uid, inode->i_uid)) ||
1210 (attr->ia_valid & ATTR_GID && !gid_eq(attr->ia_gid, inode->i_gid))) {
1211
1212
1213
1214
1215
1216 if (attr->ia_valid & ATTR_UID && !uid_eq(attr->ia_uid, inode->i_uid)
1217 && OCFS2_HAS_RO_COMPAT_FEATURE(sb,
1218 OCFS2_FEATURE_RO_COMPAT_USRQUOTA)) {
1219 transfer_to[USRQUOTA] = dqget(sb, make_kqid_uid(attr->ia_uid));
1220 if (IS_ERR(transfer_to[USRQUOTA])) {
1221 status = PTR_ERR(transfer_to[USRQUOTA]);
1222 goto bail_unlock;
1223 }
1224 }
1225 if (attr->ia_valid & ATTR_GID && !gid_eq(attr->ia_gid, inode->i_gid)
1226 && OCFS2_HAS_RO_COMPAT_FEATURE(sb,
1227 OCFS2_FEATURE_RO_COMPAT_GRPQUOTA)) {
1228 transfer_to[GRPQUOTA] = dqget(sb, make_kqid_gid(attr->ia_gid));
1229 if (IS_ERR(transfer_to[GRPQUOTA])) {
1230 status = PTR_ERR(transfer_to[GRPQUOTA]);
1231 goto bail_unlock;
1232 }
1233 }
1234 handle = ocfs2_start_trans(osb, OCFS2_INODE_UPDATE_CREDITS +
1235 2 * ocfs2_quota_trans_credits(sb));
1236 if (IS_ERR(handle)) {
1237 status = PTR_ERR(handle);
1238 mlog_errno(status);
1239 goto bail_unlock;
1240 }
1241 status = __dquot_transfer(inode, transfer_to);
1242 if (status < 0)
1243 goto bail_commit;
1244 } else {
1245 handle = ocfs2_start_trans(osb, OCFS2_INODE_UPDATE_CREDITS);
1246 if (IS_ERR(handle)) {
1247 status = PTR_ERR(handle);
1248 mlog_errno(status);
1249 goto bail_unlock;
1250 }
1251 }
1252
1253 setattr_copy(inode, attr);
1254 mark_inode_dirty(inode);
1255
1256 status = ocfs2_mark_inode_dirty(handle, inode, bh);
1257 if (status < 0)
1258 mlog_errno(status);
1259
1260bail_commit:
1261 ocfs2_commit_trans(osb, handle);
1262bail_unlock:
1263 if (status) {
1264 ocfs2_inode_unlock(inode, 1);
1265 inode_locked = 0;
1266 }
1267bail_unlock_rw:
1268 if (size_change)
1269 ocfs2_rw_unlock(inode, 1);
1270bail:
1271 brelse(bh);
1272
1273
1274 for (qtype = 0; qtype < OCFS2_MAXQUOTAS; qtype++)
1275 dqput(transfer_to[qtype]);
1276
1277 if (!status && attr->ia_valid & ATTR_MODE) {
1278 status = posix_acl_chmod(inode, inode->i_mode);
1279 if (status < 0)
1280 mlog_errno(status);
1281 }
1282 if (inode_locked)
1283 ocfs2_inode_unlock(inode, 1);
1284
1285 return status;
1286}
1287
1288int ocfs2_getattr(struct vfsmount *mnt,
1289 struct dentry *dentry,
1290 struct kstat *stat)
1291{
1292 struct inode *inode = d_inode(dentry);
1293 struct super_block *sb = d_inode(dentry)->i_sb;
1294 struct ocfs2_super *osb = sb->s_fs_info;
1295 int err;
1296
1297 err = ocfs2_inode_revalidate(dentry);
1298 if (err) {
1299 if (err != -ENOENT)
1300 mlog_errno(err);
1301 goto bail;
1302 }
1303
1304 generic_fillattr(inode, stat);
1305
1306
1307 stat->blksize = osb->s_clustersize;
1308
1309bail:
1310 return err;
1311}
1312
1313int ocfs2_permission(struct inode *inode, int mask)
1314{
1315 int ret;
1316
1317 if (mask & MAY_NOT_BLOCK)
1318 return -ECHILD;
1319
1320 ret = ocfs2_inode_lock(inode, NULL, 0);
1321 if (ret) {
1322 if (ret != -ENOENT)
1323 mlog_errno(ret);
1324 goto out;
1325 }
1326
1327 ret = generic_permission(inode, mask);
1328
1329 ocfs2_inode_unlock(inode, 0);
1330out:
1331 return ret;
1332}
1333
1334static int __ocfs2_write_remove_suid(struct inode *inode,
1335 struct buffer_head *bh)
1336{
1337 int ret;
1338 handle_t *handle;
1339 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
1340 struct ocfs2_dinode *di;
1341
1342 trace_ocfs2_write_remove_suid(
1343 (unsigned long long)OCFS2_I(inode)->ip_blkno,
1344 inode->i_mode);
1345
1346 handle = ocfs2_start_trans(osb, OCFS2_INODE_UPDATE_CREDITS);
1347 if (IS_ERR(handle)) {
1348 ret = PTR_ERR(handle);
1349 mlog_errno(ret);
1350 goto out;
1351 }
1352
1353 ret = ocfs2_journal_access_di(handle, INODE_CACHE(inode), bh,
1354 OCFS2_JOURNAL_ACCESS_WRITE);
1355 if (ret < 0) {
1356 mlog_errno(ret);
1357 goto out_trans;
1358 }
1359
1360 inode->i_mode &= ~S_ISUID;
1361 if ((inode->i_mode & S_ISGID) && (inode->i_mode & S_IXGRP))
1362 inode->i_mode &= ~S_ISGID;
1363
1364 di = (struct ocfs2_dinode *) bh->b_data;
1365 di->i_mode = cpu_to_le16(inode->i_mode);
1366 ocfs2_update_inode_fsync_trans(handle, inode, 0);
1367
1368 ocfs2_journal_dirty(handle, bh);
1369
1370out_trans:
1371 ocfs2_commit_trans(osb, handle);
1372out:
1373 return ret;
1374}
1375
1376
1377
1378
1379
1380static int ocfs2_check_range_for_holes(struct inode *inode, loff_t pos,
1381 size_t count)
1382{
1383 int ret = 0;
1384 unsigned int extent_flags;
1385 u32 cpos, clusters, extent_len, phys_cpos;
1386 struct super_block *sb = inode->i_sb;
1387
1388 cpos = pos >> OCFS2_SB(sb)->s_clustersize_bits;
1389 clusters = ocfs2_clusters_for_bytes(sb, pos + count) - cpos;
1390
1391 while (clusters) {
1392 ret = ocfs2_get_clusters(inode, cpos, &phys_cpos, &extent_len,
1393 &extent_flags);
1394 if (ret < 0) {
1395 mlog_errno(ret);
1396 goto out;
1397 }
1398
1399 if (phys_cpos == 0 || (extent_flags & OCFS2_EXT_UNWRITTEN)) {
1400 ret = 1;
1401 break;
1402 }
1403
1404 if (extent_len > clusters)
1405 extent_len = clusters;
1406
1407 clusters -= extent_len;
1408 cpos += extent_len;
1409 }
1410out:
1411 return ret;
1412}
1413
1414static int ocfs2_write_remove_suid(struct inode *inode)
1415{
1416 int ret;
1417 struct buffer_head *bh = NULL;
1418
1419 ret = ocfs2_read_inode_block(inode, &bh);
1420 if (ret < 0) {
1421 mlog_errno(ret);
1422 goto out;
1423 }
1424
1425 ret = __ocfs2_write_remove_suid(inode, bh);
1426out:
1427 brelse(bh);
1428 return ret;
1429}
1430
1431
1432
1433
1434
1435
1436static int ocfs2_allocate_unwritten_extents(struct inode *inode,
1437 u64 start, u64 len)
1438{
1439 int ret;
1440 u32 cpos, phys_cpos, clusters, alloc_size;
1441 u64 end = start + len;
1442 struct buffer_head *di_bh = NULL;
1443
1444 if (OCFS2_I(inode)->ip_dyn_features & OCFS2_INLINE_DATA_FL) {
1445 ret = ocfs2_read_inode_block(inode, &di_bh);
1446 if (ret) {
1447 mlog_errno(ret);
1448 goto out;
1449 }
1450
1451
1452
1453
1454
1455 if (ocfs2_size_fits_inline_data(di_bh, end))
1456 goto out;
1457
1458 ret = ocfs2_convert_inline_data_to_extents(inode, di_bh);
1459 if (ret) {
1460 mlog_errno(ret);
1461 goto out;
1462 }
1463 }
1464
1465
1466
1467
1468 cpos = start >> OCFS2_SB(inode->i_sb)->s_clustersize_bits;
1469 clusters = ocfs2_clusters_for_bytes(inode->i_sb, start + len);
1470 clusters -= cpos;
1471
1472 while (clusters) {
1473 ret = ocfs2_get_clusters(inode, cpos, &phys_cpos,
1474 &alloc_size, NULL);
1475 if (ret) {
1476 mlog_errno(ret);
1477 goto out;
1478 }
1479
1480
1481
1482
1483
1484 if (alloc_size > clusters)
1485 alloc_size = clusters;
1486
1487 if (phys_cpos) {
1488
1489
1490
1491
1492 goto next;
1493 }
1494
1495 ret = __ocfs2_extend_allocation(inode, cpos, alloc_size, 1);
1496 if (ret) {
1497 if (ret != -ENOSPC)
1498 mlog_errno(ret);
1499 goto out;
1500 }
1501
1502next:
1503 cpos += alloc_size;
1504 clusters -= alloc_size;
1505 }
1506
1507 ret = 0;
1508out:
1509
1510 brelse(di_bh);
1511 return ret;
1512}
1513
1514
1515
1516
1517
1518static void ocfs2_truncate_cluster_pages(struct inode *inode, u64 byte_start,
1519 u64 byte_len)
1520{
1521 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
1522 loff_t start, end;
1523 struct address_space *mapping = inode->i_mapping;
1524
1525 start = (loff_t)ocfs2_align_bytes_to_clusters(inode->i_sb, byte_start);
1526 end = byte_start + byte_len;
1527 end = end & ~(osb->s_clustersize - 1);
1528
1529 if (start < end) {
1530 unmap_mapping_range(mapping, start, end - start, 0);
1531 truncate_inode_pages_range(mapping, start, end - 1);
1532 }
1533}
1534
1535static int ocfs2_zero_partial_clusters(struct inode *inode,
1536 u64 start, u64 len)
1537{
1538 int ret = 0;
1539 u64 tmpend, end = start + len;
1540 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
1541 unsigned int csize = osb->s_clustersize;
1542 handle_t *handle;
1543
1544
1545
1546
1547
1548
1549
1550
1551 trace_ocfs2_zero_partial_clusters(
1552 (unsigned long long)OCFS2_I(inode)->ip_blkno,
1553 (unsigned long long)start, (unsigned long long)end);
1554
1555
1556
1557
1558
1559
1560 if ((start & (csize - 1)) == 0 && (end & (csize - 1)) == 0)
1561 goto out;
1562
1563 handle = ocfs2_start_trans(osb, OCFS2_INODE_UPDATE_CREDITS);
1564 if (IS_ERR(handle)) {
1565 ret = PTR_ERR(handle);
1566 mlog_errno(ret);
1567 goto out;
1568 }
1569
1570
1571
1572
1573 tmpend = (u64)osb->s_clustersize + (start & ~(osb->s_clustersize - 1));
1574 if (tmpend > end)
1575 tmpend = end;
1576
1577 trace_ocfs2_zero_partial_clusters_range1((unsigned long long)start,
1578 (unsigned long long)tmpend);
1579
1580 ret = ocfs2_zero_range_for_truncate(inode, handle, start, tmpend);
1581 if (ret)
1582 mlog_errno(ret);
1583
1584 if (tmpend < end) {
1585
1586
1587
1588
1589
1590 start = end & ~(osb->s_clustersize - 1);
1591
1592 trace_ocfs2_zero_partial_clusters_range2(
1593 (unsigned long long)start, (unsigned long long)end);
1594
1595 ret = ocfs2_zero_range_for_truncate(inode, handle, start, end);
1596 if (ret)
1597 mlog_errno(ret);
1598 }
1599 ocfs2_update_inode_fsync_trans(handle, inode, 1);
1600
1601 ocfs2_commit_trans(osb, handle);
1602out:
1603 return ret;
1604}
1605
1606static int ocfs2_find_rec(struct ocfs2_extent_list *el, u32 pos)
1607{
1608 int i;
1609 struct ocfs2_extent_rec *rec = NULL;
1610
1611 for (i = le16_to_cpu(el->l_next_free_rec) - 1; i >= 0; i--) {
1612
1613 rec = &el->l_recs[i];
1614
1615 if (le32_to_cpu(rec->e_cpos) < pos)
1616 break;
1617 }
1618
1619 return i;
1620}
1621
1622
1623
1624
1625
1626
1627
1628
1629
1630static void ocfs2_calc_trunc_pos(struct inode *inode,
1631 struct ocfs2_extent_list *el,
1632 struct ocfs2_extent_rec *rec,
1633 u32 trunc_start, u32 *trunc_cpos,
1634 u32 *trunc_len, u32 *trunc_end,
1635 u64 *blkno, int *done)
1636{
1637 int ret = 0;
1638 u32 coff, range;
1639
1640 range = le32_to_cpu(rec->e_cpos) + ocfs2_rec_clusters(el, rec);
1641
1642 if (le32_to_cpu(rec->e_cpos) >= trunc_start) {
1643
1644
1645
1646 *trunc_cpos = le32_to_cpu(rec->e_cpos);
1647
1648
1649
1650 if (range < *trunc_end)
1651 *trunc_end = range;
1652 *trunc_len = *trunc_end - le32_to_cpu(rec->e_cpos);
1653 *blkno = le64_to_cpu(rec->e_blkno);
1654 *trunc_end = le32_to_cpu(rec->e_cpos);
1655 } else if (range > trunc_start) {
1656
1657
1658
1659
1660 *trunc_cpos = trunc_start;
1661
1662
1663
1664 if (range < *trunc_end)
1665 *trunc_end = range;
1666 *trunc_len = *trunc_end - trunc_start;
1667 coff = trunc_start - le32_to_cpu(rec->e_cpos);
1668 *blkno = le64_to_cpu(rec->e_blkno) +
1669 ocfs2_clusters_to_blocks(inode->i_sb, coff);
1670 *trunc_end = trunc_start;
1671 } else {
1672
1673
1674
1675
1676
1677
1678
1679
1680 ret = 1;
1681 }
1682
1683 *done = ret;
1684}
1685
1686static int ocfs2_remove_inode_range(struct inode *inode,
1687 struct buffer_head *di_bh, u64 byte_start,
1688 u64 byte_len)
1689{
1690 int ret = 0, flags = 0, done = 0, i;
1691 u32 trunc_start, trunc_len, trunc_end, trunc_cpos, phys_cpos;
1692 u32 cluster_in_el;
1693 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
1694 struct ocfs2_cached_dealloc_ctxt dealloc;
1695 struct address_space *mapping = inode->i_mapping;
1696 struct ocfs2_extent_tree et;
1697 struct ocfs2_path *path = NULL;
1698 struct ocfs2_extent_list *el = NULL;
1699 struct ocfs2_extent_rec *rec = NULL;
1700 struct ocfs2_dinode *di = (struct ocfs2_dinode *)di_bh->b_data;
1701 u64 blkno, refcount_loc = le64_to_cpu(di->i_refcount_loc);
1702
1703 ocfs2_init_dinode_extent_tree(&et, INODE_CACHE(inode), di_bh);
1704 ocfs2_init_dealloc_ctxt(&dealloc);
1705
1706 trace_ocfs2_remove_inode_range(
1707 (unsigned long long)OCFS2_I(inode)->ip_blkno,
1708 (unsigned long long)byte_start,
1709 (unsigned long long)byte_len);
1710
1711 if (byte_len == 0)
1712 return 0;
1713
1714 if (OCFS2_I(inode)->ip_dyn_features & OCFS2_INLINE_DATA_FL) {
1715 ret = ocfs2_truncate_inline(inode, di_bh, byte_start,
1716 byte_start + byte_len, 0);
1717 if (ret) {
1718 mlog_errno(ret);
1719 goto out;
1720 }
1721
1722
1723
1724
1725
1726
1727 unmap_mapping_range(mapping, 0, 0, 0);
1728 truncate_inode_pages(mapping, 0);
1729 goto out;
1730 }
1731
1732
1733
1734
1735
1736
1737
1738 if (OCFS2_I(inode)->ip_dyn_features & OCFS2_HAS_REFCOUNT_FL) {
1739
1740 ret = ocfs2_cow_file_pos(inode, di_bh, byte_start);
1741 if (ret) {
1742 mlog_errno(ret);
1743 goto out;
1744 }
1745
1746 ret = ocfs2_cow_file_pos(inode, di_bh, byte_start + byte_len);
1747 if (ret) {
1748 mlog_errno(ret);
1749 goto out;
1750 }
1751 }
1752
1753 trunc_start = ocfs2_clusters_for_bytes(osb->sb, byte_start);
1754 trunc_end = (byte_start + byte_len) >> osb->s_clustersize_bits;
1755 cluster_in_el = trunc_end;
1756
1757 ret = ocfs2_zero_partial_clusters(inode, byte_start, byte_len);
1758 if (ret) {
1759 mlog_errno(ret);
1760 goto out;
1761 }
1762
1763 path = ocfs2_new_path_from_et(&et);
1764 if (!path) {
1765 ret = -ENOMEM;
1766 mlog_errno(ret);
1767 goto out;
1768 }
1769
1770 while (trunc_end > trunc_start) {
1771
1772 ret = ocfs2_find_path(INODE_CACHE(inode), path,
1773 cluster_in_el);
1774 if (ret) {
1775 mlog_errno(ret);
1776 goto out;
1777 }
1778
1779 el = path_leaf_el(path);
1780
1781 i = ocfs2_find_rec(el, trunc_end);
1782
1783
1784
1785 if (i < 0) {
1786 if (path->p_tree_depth == 0)
1787 break;
1788
1789 ret = ocfs2_find_cpos_for_left_leaf(inode->i_sb,
1790 path,
1791 &cluster_in_el);
1792 if (ret) {
1793 mlog_errno(ret);
1794 goto out;
1795 }
1796
1797
1798
1799
1800
1801 if (cluster_in_el == 0)
1802 break;
1803
1804
1805
1806
1807
1808 trunc_end = cluster_in_el + 1;
1809
1810 ocfs2_reinit_path(path, 1);
1811
1812 continue;
1813
1814 } else
1815 rec = &el->l_recs[i];
1816
1817 ocfs2_calc_trunc_pos(inode, el, rec, trunc_start, &trunc_cpos,
1818 &trunc_len, &trunc_end, &blkno, &done);
1819 if (done)
1820 break;
1821
1822 flags = rec->e_flags;
1823 phys_cpos = ocfs2_blocks_to_clusters(inode->i_sb, blkno);
1824
1825 ret = ocfs2_remove_btree_range(inode, &et, trunc_cpos,
1826 phys_cpos, trunc_len, flags,
1827 &dealloc, refcount_loc, false);
1828 if (ret < 0) {
1829 mlog_errno(ret);
1830 goto out;
1831 }
1832
1833 cluster_in_el = trunc_end;
1834
1835 ocfs2_reinit_path(path, 1);
1836 }
1837
1838 ocfs2_truncate_cluster_pages(inode, byte_start, byte_len);
1839
1840out:
1841 ocfs2_free_path(path);
1842 ocfs2_schedule_truncate_log_flush(osb, 1);
1843 ocfs2_run_deallocs(osb, &dealloc);
1844
1845 return ret;
1846}
1847
1848
1849
1850
1851static int __ocfs2_change_file_space(struct file *file, struct inode *inode,
1852 loff_t f_pos, unsigned int cmd,
1853 struct ocfs2_space_resv *sr,
1854 int change_size)
1855{
1856 int ret;
1857 s64 llen;
1858 loff_t size;
1859 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
1860 struct buffer_head *di_bh = NULL;
1861 handle_t *handle;
1862 unsigned long long max_off = inode->i_sb->s_maxbytes;
1863
1864 if (ocfs2_is_hard_readonly(osb) || ocfs2_is_soft_readonly(osb))
1865 return -EROFS;
1866
1867 mutex_lock(&inode->i_mutex);
1868
1869
1870
1871
1872 ret = ocfs2_rw_lock(inode, 1);
1873 if (ret) {
1874 mlog_errno(ret);
1875 goto out;
1876 }
1877
1878 ret = ocfs2_inode_lock(inode, &di_bh, 1);
1879 if (ret) {
1880 mlog_errno(ret);
1881 goto out_rw_unlock;
1882 }
1883
1884 if (inode->i_flags & (S_IMMUTABLE|S_APPEND)) {
1885 ret = -EPERM;
1886 goto out_inode_unlock;
1887 }
1888
1889 switch (sr->l_whence) {
1890 case 0:
1891 break;
1892 case 1:
1893 sr->l_start += f_pos;
1894 break;
1895 case 2:
1896 sr->l_start += i_size_read(inode);
1897 break;
1898 default:
1899 ret = -EINVAL;
1900 goto out_inode_unlock;
1901 }
1902 sr->l_whence = 0;
1903
1904 llen = sr->l_len > 0 ? sr->l_len - 1 : sr->l_len;
1905
1906 if (sr->l_start < 0
1907 || sr->l_start > max_off
1908 || (sr->l_start + llen) < 0
1909 || (sr->l_start + llen) > max_off) {
1910 ret = -EINVAL;
1911 goto out_inode_unlock;
1912 }
1913 size = sr->l_start + sr->l_len;
1914
1915 if (cmd == OCFS2_IOC_RESVSP || cmd == OCFS2_IOC_RESVSP64 ||
1916 cmd == OCFS2_IOC_UNRESVSP || cmd == OCFS2_IOC_UNRESVSP64) {
1917 if (sr->l_len <= 0) {
1918 ret = -EINVAL;
1919 goto out_inode_unlock;
1920 }
1921 }
1922
1923 if (file && should_remove_suid(file->f_path.dentry)) {
1924 ret = __ocfs2_write_remove_suid(inode, di_bh);
1925 if (ret) {
1926 mlog_errno(ret);
1927 goto out_inode_unlock;
1928 }
1929 }
1930
1931 down_write(&OCFS2_I(inode)->ip_alloc_sem);
1932 switch (cmd) {
1933 case OCFS2_IOC_RESVSP:
1934 case OCFS2_IOC_RESVSP64:
1935
1936
1937
1938
1939 ret = ocfs2_allocate_unwritten_extents(inode, sr->l_start,
1940 sr->l_len);
1941 break;
1942 case OCFS2_IOC_UNRESVSP:
1943 case OCFS2_IOC_UNRESVSP64:
1944 ret = ocfs2_remove_inode_range(inode, di_bh, sr->l_start,
1945 sr->l_len);
1946 break;
1947 default:
1948 ret = -EINVAL;
1949 }
1950 up_write(&OCFS2_I(inode)->ip_alloc_sem);
1951 if (ret) {
1952 mlog_errno(ret);
1953 goto out_inode_unlock;
1954 }
1955
1956
1957
1958
1959 handle = ocfs2_start_trans(osb, OCFS2_INODE_UPDATE_CREDITS);
1960 if (IS_ERR(handle)) {
1961 ret = PTR_ERR(handle);
1962 mlog_errno(ret);
1963 goto out_inode_unlock;
1964 }
1965
1966 if (change_size && i_size_read(inode) < size)
1967 i_size_write(inode, size);
1968
1969 inode->i_ctime = inode->i_mtime = CURRENT_TIME;
1970 ret = ocfs2_mark_inode_dirty(handle, inode, di_bh);
1971 if (ret < 0)
1972 mlog_errno(ret);
1973
1974 if (file && (file->f_flags & O_SYNC))
1975 handle->h_sync = 1;
1976
1977 ocfs2_commit_trans(osb, handle);
1978
1979out_inode_unlock:
1980 brelse(di_bh);
1981 ocfs2_inode_unlock(inode, 1);
1982out_rw_unlock:
1983 ocfs2_rw_unlock(inode, 1);
1984
1985out:
1986 mutex_unlock(&inode->i_mutex);
1987 return ret;
1988}
1989
1990int ocfs2_change_file_space(struct file *file, unsigned int cmd,
1991 struct ocfs2_space_resv *sr)
1992{
1993 struct inode *inode = file_inode(file);
1994 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
1995 int ret;
1996
1997 if ((cmd == OCFS2_IOC_RESVSP || cmd == OCFS2_IOC_RESVSP64) &&
1998 !ocfs2_writes_unwritten_extents(osb))
1999 return -ENOTTY;
2000 else if ((cmd == OCFS2_IOC_UNRESVSP || cmd == OCFS2_IOC_UNRESVSP64) &&
2001 !ocfs2_sparse_alloc(osb))
2002 return -ENOTTY;
2003
2004 if (!S_ISREG(inode->i_mode))
2005 return -EINVAL;
2006
2007 if (!(file->f_mode & FMODE_WRITE))
2008 return -EBADF;
2009
2010 ret = mnt_want_write_file(file);
2011 if (ret)
2012 return ret;
2013 ret = __ocfs2_change_file_space(file, inode, file->f_pos, cmd, sr, 0);
2014 mnt_drop_write_file(file);
2015 return ret;
2016}
2017
2018static long ocfs2_fallocate(struct file *file, int mode, loff_t offset,
2019 loff_t len)
2020{
2021 struct inode *inode = file_inode(file);
2022 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
2023 struct ocfs2_space_resv sr;
2024 int change_size = 1;
2025 int cmd = OCFS2_IOC_RESVSP64;
2026
2027 if (mode & ~(FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE))
2028 return -EOPNOTSUPP;
2029 if (!ocfs2_writes_unwritten_extents(osb))
2030 return -EOPNOTSUPP;
2031
2032 if (mode & FALLOC_FL_KEEP_SIZE)
2033 change_size = 0;
2034
2035 if (mode & FALLOC_FL_PUNCH_HOLE)
2036 cmd = OCFS2_IOC_UNRESVSP64;
2037
2038 sr.l_whence = 0;
2039 sr.l_start = (s64)offset;
2040 sr.l_len = (s64)len;
2041
2042 return __ocfs2_change_file_space(NULL, inode, offset, cmd, &sr,
2043 change_size);
2044}
2045
2046int ocfs2_check_range_for_refcount(struct inode *inode, loff_t pos,
2047 size_t count)
2048{
2049 int ret = 0;
2050 unsigned int extent_flags;
2051 u32 cpos, clusters, extent_len, phys_cpos;
2052 struct super_block *sb = inode->i_sb;
2053
2054 if (!ocfs2_refcount_tree(OCFS2_SB(inode->i_sb)) ||
2055 !(OCFS2_I(inode)->ip_dyn_features & OCFS2_HAS_REFCOUNT_FL) ||
2056 OCFS2_I(inode)->ip_dyn_features & OCFS2_INLINE_DATA_FL)
2057 return 0;
2058
2059 cpos = pos >> OCFS2_SB(sb)->s_clustersize_bits;
2060 clusters = ocfs2_clusters_for_bytes(sb, pos + count) - cpos;
2061
2062 while (clusters) {
2063 ret = ocfs2_get_clusters(inode, cpos, &phys_cpos, &extent_len,
2064 &extent_flags);
2065 if (ret < 0) {
2066 mlog_errno(ret);
2067 goto out;
2068 }
2069
2070 if (phys_cpos && (extent_flags & OCFS2_EXT_REFCOUNTED)) {
2071 ret = 1;
2072 break;
2073 }
2074
2075 if (extent_len > clusters)
2076 extent_len = clusters;
2077
2078 clusters -= extent_len;
2079 cpos += extent_len;
2080 }
2081out:
2082 return ret;
2083}
2084
2085static int ocfs2_is_io_unaligned(struct inode *inode, size_t count, loff_t pos)
2086{
2087 int blockmask = inode->i_sb->s_blocksize - 1;
2088 loff_t final_size = pos + count;
2089
2090 if ((pos & blockmask) || (final_size & blockmask))
2091 return 1;
2092 return 0;
2093}
2094
2095static int ocfs2_prepare_inode_for_refcount(struct inode *inode,
2096 struct file *file,
2097 loff_t pos, size_t count,
2098 int *meta_level)
2099{
2100 int ret;
2101 struct buffer_head *di_bh = NULL;
2102 u32 cpos = pos >> OCFS2_SB(inode->i_sb)->s_clustersize_bits;
2103 u32 clusters =
2104 ocfs2_clusters_for_bytes(inode->i_sb, pos + count) - cpos;
2105
2106 ret = ocfs2_inode_lock(inode, &di_bh, 1);
2107 if (ret) {
2108 mlog_errno(ret);
2109 goto out;
2110 }
2111
2112 *meta_level = 1;
2113
2114 ret = ocfs2_refcount_cow(inode, di_bh, cpos, clusters, UINT_MAX);
2115 if (ret)
2116 mlog_errno(ret);
2117out:
2118 brelse(di_bh);
2119 return ret;
2120}
2121
2122static int ocfs2_prepare_inode_for_write(struct file *file,
2123 loff_t pos,
2124 size_t count,
2125 int appending,
2126 int *direct_io,
2127 int *has_refcount)
2128{
2129 int ret = 0, meta_level = 0;
2130 struct dentry *dentry = file->f_path.dentry;
2131 struct inode *inode = d_inode(dentry);
2132 loff_t end;
2133 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
2134 int full_coherency = !(osb->s_mount_opt &
2135 OCFS2_MOUNT_COHERENCY_BUFFERED);
2136
2137
2138
2139
2140
2141 for(;;) {
2142 ret = ocfs2_inode_lock(inode, NULL, meta_level);
2143 if (ret < 0) {
2144 meta_level = -1;
2145 mlog_errno(ret);
2146 goto out;
2147 }
2148
2149
2150
2151
2152
2153
2154
2155
2156
2157
2158 if (should_remove_suid(dentry)) {
2159 if (meta_level == 0) {
2160 ocfs2_inode_unlock(inode, meta_level);
2161 meta_level = 1;
2162 continue;
2163 }
2164
2165 ret = ocfs2_write_remove_suid(inode);
2166 if (ret < 0) {
2167 mlog_errno(ret);
2168 goto out_unlock;
2169 }
2170 }
2171
2172 end = pos + count;
2173
2174 ret = ocfs2_check_range_for_refcount(inode, pos, count);
2175 if (ret == 1) {
2176 ocfs2_inode_unlock(inode, meta_level);
2177 meta_level = -1;
2178
2179 ret = ocfs2_prepare_inode_for_refcount(inode,
2180 file,
2181 pos,
2182 count,
2183 &meta_level);
2184 if (has_refcount)
2185 *has_refcount = 1;
2186 if (direct_io)
2187 *direct_io = 0;
2188 }
2189
2190 if (ret < 0) {
2191 mlog_errno(ret);
2192 goto out_unlock;
2193 }
2194
2195
2196
2197
2198
2199 if (!direct_io || !(*direct_io))
2200 break;
2201
2202
2203
2204
2205
2206 if (OCFS2_I(inode)->ip_dyn_features & OCFS2_INLINE_DATA_FL) {
2207 *direct_io = 0;
2208 break;
2209 }
2210
2211
2212
2213
2214
2215
2216
2217 if (end > i_size_read(inode) && !full_coherency) {
2218 *direct_io = 0;
2219 break;
2220 }
2221
2222
2223
2224
2225 if (end > i_size_read(inode) &&
2226 !ocfs2_supports_append_dio(osb)) {
2227 *direct_io = 0;
2228 break;
2229 }
2230
2231
2232
2233
2234
2235
2236
2237 ret = ocfs2_check_range_for_holes(inode, pos, count);
2238 if (ret == 1) {
2239
2240
2241
2242
2243
2244 if (!ocfs2_supports_append_dio(osb))
2245 *direct_io = 0;
2246 ret = 0;
2247 } else if (ret < 0)
2248 mlog_errno(ret);
2249 break;
2250 }
2251
2252out_unlock:
2253 trace_ocfs2_prepare_inode_for_write(OCFS2_I(inode)->ip_blkno,
2254 pos, appending, count,
2255 direct_io, has_refcount);
2256
2257 if (meta_level >= 0)
2258 ocfs2_inode_unlock(inode, meta_level);
2259
2260out:
2261 return ret;
2262}
2263
2264static ssize_t ocfs2_file_write_iter(struct kiocb *iocb,
2265 struct iov_iter *from)
2266{
2267 int direct_io, appending, rw_level;
2268 int can_do_direct, has_refcount = 0;
2269 ssize_t written = 0;
2270 ssize_t ret;
2271 size_t count = iov_iter_count(from), orig_count;
2272 struct file *file = iocb->ki_filp;
2273 struct inode *inode = file_inode(file);
2274 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
2275 int full_coherency = !(osb->s_mount_opt &
2276 OCFS2_MOUNT_COHERENCY_BUFFERED);
2277 int unaligned_dio = 0;
2278 int dropped_dio = 0;
2279 int append_write = ((iocb->ki_pos + count) >=
2280 i_size_read(inode) ? 1 : 0);
2281
2282 trace_ocfs2_file_aio_write(inode, file, file->f_path.dentry,
2283 (unsigned long long)OCFS2_I(inode)->ip_blkno,
2284 file->f_path.dentry->d_name.len,
2285 file->f_path.dentry->d_name.name,
2286 (unsigned int)from->nr_segs);
2287
2288 if (count == 0)
2289 return 0;
2290
2291 appending = iocb->ki_flags & IOCB_APPEND ? 1 : 0;
2292 direct_io = iocb->ki_flags & IOCB_DIRECT ? 1 : 0;
2293
2294 mutex_lock(&inode->i_mutex);
2295
2296relock:
2297
2298
2299
2300
2301
2302 rw_level = (!direct_io || full_coherency || append_write);
2303
2304 ret = ocfs2_rw_lock(inode, rw_level);
2305 if (ret < 0) {
2306 mlog_errno(ret);
2307 goto out_mutex;
2308 }
2309
2310
2311
2312
2313
2314 if (direct_io && full_coherency) {
2315
2316
2317
2318
2319
2320 ret = ocfs2_inode_lock(inode, NULL, 1);
2321 if (ret < 0) {
2322 mlog_errno(ret);
2323 goto out;
2324 }
2325
2326 ocfs2_inode_unlock(inode, 1);
2327 }
2328
2329 orig_count = iov_iter_count(from);
2330 ret = generic_write_checks(iocb, from);
2331 if (ret <= 0) {
2332 if (ret)
2333 mlog_errno(ret);
2334 goto out;
2335 }
2336 count = ret;
2337
2338 can_do_direct = direct_io;
2339 ret = ocfs2_prepare_inode_for_write(file, iocb->ki_pos, count, appending,
2340 &can_do_direct, &has_refcount);
2341 if (ret < 0) {
2342 mlog_errno(ret);
2343 goto out;
2344 }
2345
2346 if (direct_io && !is_sync_kiocb(iocb))
2347 unaligned_dio = ocfs2_is_io_unaligned(inode, count, iocb->ki_pos);
2348
2349
2350
2351
2352
2353 if (direct_io && !can_do_direct) {
2354 ocfs2_rw_unlock(inode, rw_level);
2355
2356 rw_level = -1;
2357
2358 direct_io = 0;
2359 iocb->ki_flags &= ~IOCB_DIRECT;
2360 iov_iter_reexpand(from, orig_count);
2361 dropped_dio = 1;
2362 goto relock;
2363 }
2364
2365 if (unaligned_dio) {
2366
2367
2368
2369
2370 mutex_lock(&OCFS2_I(inode)->ip_unaligned_aio);
2371
2372 ocfs2_iocb_set_unaligned_aio(iocb);
2373 }
2374
2375
2376 ocfs2_iocb_set_rw_locked(iocb, rw_level);
2377
2378 written = __generic_file_write_iter(iocb, from);
2379
2380 BUG_ON(written == -EIOCBQUEUED && !(iocb->ki_flags & IOCB_DIRECT));
2381
2382
2383
2384
2385
2386
2387
2388
2389
2390
2391 if ((written == -EIOCBQUEUED) || (!ocfs2_iocb_is_rw_locked(iocb))) {
2392 rw_level = -1;
2393 unaligned_dio = 0;
2394 }
2395
2396 if (unlikely(written <= 0))
2397 goto no_sync;
2398
2399 if (((file->f_flags & O_DSYNC) && !direct_io) ||
2400 IS_SYNC(inode) || dropped_dio) {
2401 ret = filemap_fdatawrite_range(file->f_mapping,
2402 iocb->ki_pos - written,
2403 iocb->ki_pos - 1);
2404 if (ret < 0)
2405 written = ret;
2406
2407 if (!ret) {
2408 ret = jbd2_journal_force_commit(osb->journal->j_journal);
2409 if (ret < 0)
2410 written = ret;
2411 }
2412
2413 if (!ret)
2414 ret = filemap_fdatawait_range(file->f_mapping,
2415 iocb->ki_pos - written,
2416 iocb->ki_pos - 1);
2417 }
2418
2419no_sync:
2420 if (unaligned_dio && ocfs2_iocb_is_unaligned_aio(iocb)) {
2421 ocfs2_iocb_clear_unaligned_aio(iocb);
2422 mutex_unlock(&OCFS2_I(inode)->ip_unaligned_aio);
2423 }
2424
2425out:
2426 if (rw_level != -1)
2427 ocfs2_rw_unlock(inode, rw_level);
2428
2429out_mutex:
2430 mutex_unlock(&inode->i_mutex);
2431
2432 if (written)
2433 ret = written;
2434 return ret;
2435}
2436
2437static ssize_t ocfs2_file_splice_read(struct file *in,
2438 loff_t *ppos,
2439 struct pipe_inode_info *pipe,
2440 size_t len,
2441 unsigned int flags)
2442{
2443 int ret = 0, lock_level = 0;
2444 struct inode *inode = file_inode(in);
2445
2446 trace_ocfs2_file_splice_read(inode, in, in->f_path.dentry,
2447 (unsigned long long)OCFS2_I(inode)->ip_blkno,
2448 in->f_path.dentry->d_name.len,
2449 in->f_path.dentry->d_name.name, len);
2450
2451
2452
2453
2454 ret = ocfs2_inode_lock_atime(inode, in->f_path.mnt, &lock_level);
2455 if (ret < 0) {
2456 mlog_errno(ret);
2457 goto bail;
2458 }
2459 ocfs2_inode_unlock(inode, lock_level);
2460
2461 ret = generic_file_splice_read(in, ppos, pipe, len, flags);
2462
2463bail:
2464 return ret;
2465}
2466
2467static ssize_t ocfs2_file_read_iter(struct kiocb *iocb,
2468 struct iov_iter *to)
2469{
2470 int ret = 0, rw_level = -1, lock_level = 0;
2471 struct file *filp = iocb->ki_filp;
2472 struct inode *inode = file_inode(filp);
2473
2474 trace_ocfs2_file_aio_read(inode, filp, filp->f_path.dentry,
2475 (unsigned long long)OCFS2_I(inode)->ip_blkno,
2476 filp->f_path.dentry->d_name.len,
2477 filp->f_path.dentry->d_name.name,
2478 to->nr_segs);
2479
2480
2481 if (!inode) {
2482 ret = -EINVAL;
2483 mlog_errno(ret);
2484 goto bail;
2485 }
2486
2487
2488
2489
2490
2491 if (iocb->ki_flags & IOCB_DIRECT) {
2492 ret = ocfs2_rw_lock(inode, 0);
2493 if (ret < 0) {
2494 mlog_errno(ret);
2495 goto bail;
2496 }
2497 rw_level = 0;
2498
2499 ocfs2_iocb_set_rw_locked(iocb, rw_level);
2500 }
2501
2502
2503
2504
2505
2506
2507
2508
2509
2510
2511 ret = ocfs2_inode_lock_atime(inode, filp->f_path.mnt, &lock_level);
2512 if (ret < 0) {
2513 mlog_errno(ret);
2514 goto bail;
2515 }
2516 ocfs2_inode_unlock(inode, lock_level);
2517
2518 ret = generic_file_read_iter(iocb, to);
2519 trace_generic_file_aio_read_ret(ret);
2520
2521
2522 BUG_ON(ret == -EIOCBQUEUED && !(iocb->ki_flags & IOCB_DIRECT));
2523
2524
2525 if (ret == -EIOCBQUEUED || !ocfs2_iocb_is_rw_locked(iocb)) {
2526 rw_level = -1;
2527 }
2528
2529bail:
2530 if (rw_level != -1)
2531 ocfs2_rw_unlock(inode, rw_level);
2532
2533 return ret;
2534}
2535
2536
2537static loff_t ocfs2_file_llseek(struct file *file, loff_t offset, int whence)
2538{
2539 struct inode *inode = file->f_mapping->host;
2540 int ret = 0;
2541
2542 mutex_lock(&inode->i_mutex);
2543
2544 switch (whence) {
2545 case SEEK_SET:
2546 break;
2547 case SEEK_END:
2548
2549
2550
2551 ret = ocfs2_inode_lock(inode, NULL, 0);
2552 if (ret < 0) {
2553 mlog_errno(ret);
2554 goto out;
2555 }
2556 offset += i_size_read(inode);
2557 ocfs2_inode_unlock(inode, 0);
2558 break;
2559 case SEEK_CUR:
2560 if (offset == 0) {
2561 offset = file->f_pos;
2562 goto out;
2563 }
2564 offset += file->f_pos;
2565 break;
2566 case SEEK_DATA:
2567 case SEEK_HOLE:
2568 ret = ocfs2_seek_data_hole_offset(file, &offset, whence);
2569 if (ret)
2570 goto out;
2571 break;
2572 default:
2573 ret = -EINVAL;
2574 goto out;
2575 }
2576
2577 offset = vfs_setpos(file, offset, inode->i_sb->s_maxbytes);
2578
2579out:
2580 mutex_unlock(&inode->i_mutex);
2581 if (ret)
2582 return ret;
2583 return offset;
2584}
2585
2586const struct inode_operations ocfs2_file_iops = {
2587 .setattr = ocfs2_setattr,
2588 .getattr = ocfs2_getattr,
2589 .permission = ocfs2_permission,
2590 .setxattr = generic_setxattr,
2591 .getxattr = generic_getxattr,
2592 .listxattr = ocfs2_listxattr,
2593 .removexattr = generic_removexattr,
2594 .fiemap = ocfs2_fiemap,
2595 .get_acl = ocfs2_iop_get_acl,
2596 .set_acl = ocfs2_iop_set_acl,
2597};
2598
2599const struct inode_operations ocfs2_special_file_iops = {
2600 .setattr = ocfs2_setattr,
2601 .getattr = ocfs2_getattr,
2602 .permission = ocfs2_permission,
2603 .get_acl = ocfs2_iop_get_acl,
2604 .set_acl = ocfs2_iop_set_acl,
2605};
2606
2607
2608
2609
2610
2611const struct file_operations ocfs2_fops = {
2612 .llseek = ocfs2_file_llseek,
2613 .mmap = ocfs2_mmap,
2614 .fsync = ocfs2_sync_file,
2615 .release = ocfs2_file_release,
2616 .open = ocfs2_file_open,
2617 .read_iter = ocfs2_file_read_iter,
2618 .write_iter = ocfs2_file_write_iter,
2619 .unlocked_ioctl = ocfs2_ioctl,
2620#ifdef CONFIG_COMPAT
2621 .compat_ioctl = ocfs2_compat_ioctl,
2622#endif
2623 .lock = ocfs2_lock,
2624 .flock = ocfs2_flock,
2625 .splice_read = ocfs2_file_splice_read,
2626 .splice_write = iter_file_splice_write,
2627 .fallocate = ocfs2_fallocate,
2628};
2629
2630const struct file_operations ocfs2_dops = {
2631 .llseek = generic_file_llseek,
2632 .read = generic_read_dir,
2633 .iterate = ocfs2_readdir,
2634 .fsync = ocfs2_sync_file,
2635 .release = ocfs2_dir_release,
2636 .open = ocfs2_dir_open,
2637 .unlocked_ioctl = ocfs2_ioctl,
2638#ifdef CONFIG_COMPAT
2639 .compat_ioctl = ocfs2_compat_ioctl,
2640#endif
2641 .lock = ocfs2_lock,
2642 .flock = ocfs2_flock,
2643};
2644
2645
2646
2647
2648
2649
2650
2651
2652
2653
2654
2655
2656
2657const struct file_operations ocfs2_fops_no_plocks = {
2658 .llseek = ocfs2_file_llseek,
2659 .mmap = ocfs2_mmap,
2660 .fsync = ocfs2_sync_file,
2661 .release = ocfs2_file_release,
2662 .open = ocfs2_file_open,
2663 .read_iter = ocfs2_file_read_iter,
2664 .write_iter = ocfs2_file_write_iter,
2665 .unlocked_ioctl = ocfs2_ioctl,
2666#ifdef CONFIG_COMPAT
2667 .compat_ioctl = ocfs2_compat_ioctl,
2668#endif
2669 .flock = ocfs2_flock,
2670 .splice_read = ocfs2_file_splice_read,
2671 .splice_write = iter_file_splice_write,
2672 .fallocate = ocfs2_fallocate,
2673};
2674
2675const struct file_operations ocfs2_dops_no_plocks = {
2676 .llseek = generic_file_llseek,
2677 .read = generic_read_dir,
2678 .iterate = ocfs2_readdir,
2679 .fsync = ocfs2_sync_file,
2680 .release = ocfs2_dir_release,
2681 .open = ocfs2_dir_open,
2682 .unlocked_ioctl = ocfs2_ioctl,
2683#ifdef CONFIG_COMPAT
2684 .compat_ioctl = ocfs2_compat_ioctl,
2685#endif
2686 .flock = ocfs2_flock,
2687};
2688