1 // SPDX-License-Identifier: GPL-2.0-or-later !! 1 /* -*- mode: c; c-basic-offset: 8; -*-
2 /* !! 2 * vim: noexpandtab sw=8 ts=8 sts=0:
>> 3 *
3 * file.c 4 * file.c
4 * 5 *
5 * File open, close, extend, truncate 6 * File open, close, extend, truncate
6 * 7 *
7 * Copyright (C) 2002, 2004 Oracle. All right 8 * Copyright (C) 2002, 2004 Oracle. All rights reserved.
>> 9 *
>> 10 * This program is free software; you can redistribute it and/or
>> 11 * modify it under the terms of the GNU General Public
>> 12 * License as published by the Free Software Foundation; either
>> 13 * version 2 of the License, or (at your option) any later version.
>> 14 *
>> 15 * This program is distributed in the hope that it will be useful,
>> 16 * but WITHOUT ANY WARRANTY; without even the implied warranty of
>> 17 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
>> 18 * General Public License for more details.
>> 19 *
>> 20 * You should have received a copy of the GNU General Public
>> 21 * License along with this program; if not, write to the
>> 22 * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
>> 23 * Boston, MA 021110-1307, USA.
8 */ 24 */
9 25
10 #include <linux/capability.h> 26 #include <linux/capability.h>
11 #include <linux/fs.h> 27 #include <linux/fs.h>
12 #include <linux/types.h> 28 #include <linux/types.h>
13 #include <linux/slab.h> 29 #include <linux/slab.h>
14 #include <linux/highmem.h> 30 #include <linux/highmem.h>
15 #include <linux/pagemap.h> 31 #include <linux/pagemap.h>
16 #include <linux/uio.h> 32 #include <linux/uio.h>
17 #include <linux/sched.h> 33 #include <linux/sched.h>
18 #include <linux/splice.h> 34 #include <linux/splice.h>
19 #include <linux/mount.h> 35 #include <linux/mount.h>
20 #include <linux/writeback.h> 36 #include <linux/writeback.h>
21 #include <linux/falloc.h> 37 #include <linux/falloc.h>
22 #include <linux/quotaops.h> 38 #include <linux/quotaops.h>
23 #include <linux/blkdev.h> 39 #include <linux/blkdev.h>
24 #include <linux/backing-dev.h> 40 #include <linux/backing-dev.h>
25 41
26 #include <cluster/masklog.h> 42 #include <cluster/masklog.h>
27 43
28 #include "ocfs2.h" 44 #include "ocfs2.h"
29 45
30 #include "alloc.h" 46 #include "alloc.h"
31 #include "aops.h" 47 #include "aops.h"
32 #include "dir.h" 48 #include "dir.h"
33 #include "dlmglue.h" 49 #include "dlmglue.h"
34 #include "extent_map.h" 50 #include "extent_map.h"
35 #include "file.h" 51 #include "file.h"
36 #include "sysfile.h" 52 #include "sysfile.h"
37 #include "inode.h" 53 #include "inode.h"
38 #include "ioctl.h" 54 #include "ioctl.h"
39 #include "journal.h" 55 #include "journal.h"
40 #include "locks.h" 56 #include "locks.h"
41 #include "mmap.h" 57 #include "mmap.h"
42 #include "suballoc.h" 58 #include "suballoc.h"
43 #include "super.h" 59 #include "super.h"
44 #include "xattr.h" 60 #include "xattr.h"
45 #include "acl.h" 61 #include "acl.h"
46 #include "quota.h" 62 #include "quota.h"
47 #include "refcounttree.h" 63 #include "refcounttree.h"
48 #include "ocfs2_trace.h" 64 #include "ocfs2_trace.h"
49 65
50 #include "buffer_head_io.h" 66 #include "buffer_head_io.h"
51 67
52 static int ocfs2_init_file_private(struct inod 68 static int ocfs2_init_file_private(struct inode *inode, struct file *file)
53 { 69 {
54 struct ocfs2_file_private *fp; 70 struct ocfs2_file_private *fp;
55 71
56 fp = kzalloc(sizeof(struct ocfs2_file_ 72 fp = kzalloc(sizeof(struct ocfs2_file_private), GFP_KERNEL);
57 if (!fp) 73 if (!fp)
58 return -ENOMEM; 74 return -ENOMEM;
59 75
60 fp->fp_file = file; 76 fp->fp_file = file;
61 mutex_init(&fp->fp_mutex); 77 mutex_init(&fp->fp_mutex);
62 ocfs2_file_lock_res_init(&fp->fp_flock 78 ocfs2_file_lock_res_init(&fp->fp_flock, fp);
63 file->private_data = fp; 79 file->private_data = fp;
64 80
65 return 0; 81 return 0;
66 } 82 }
67 83
68 static void ocfs2_free_file_private(struct ino 84 static void ocfs2_free_file_private(struct inode *inode, struct file *file)
69 { 85 {
70 struct ocfs2_file_private *fp = file-> 86 struct ocfs2_file_private *fp = file->private_data;
71 struct ocfs2_super *osb = OCFS2_SB(ino 87 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
72 88
73 if (fp) { 89 if (fp) {
74 ocfs2_simple_drop_lockres(osb, 90 ocfs2_simple_drop_lockres(osb, &fp->fp_flock);
75 ocfs2_lock_res_free(&fp->fp_fl 91 ocfs2_lock_res_free(&fp->fp_flock);
76 kfree(fp); 92 kfree(fp);
77 file->private_data = NULL; 93 file->private_data = NULL;
78 } 94 }
79 } 95 }
80 96
81 static int ocfs2_file_open(struct inode *inode 97 static int ocfs2_file_open(struct inode *inode, struct file *file)
82 { 98 {
83 int status; 99 int status;
84 int mode = file->f_flags; 100 int mode = file->f_flags;
85 struct ocfs2_inode_info *oi = OCFS2_I( 101 struct ocfs2_inode_info *oi = OCFS2_I(inode);
86 102
87 trace_ocfs2_file_open(inode, file, fil 103 trace_ocfs2_file_open(inode, file, file->f_path.dentry,
88 (unsigned long l !! 104 (unsigned long long)OCFS2_I(inode)->ip_blkno,
89 file->f_path.den 105 file->f_path.dentry->d_name.len,
90 file->f_path.den 106 file->f_path.dentry->d_name.name, mode);
91 107
92 if (file->f_mode & FMODE_WRITE) { 108 if (file->f_mode & FMODE_WRITE) {
93 status = dquot_initialize(inod 109 status = dquot_initialize(inode);
94 if (status) 110 if (status)
95 goto leave; 111 goto leave;
96 } 112 }
97 113
98 spin_lock(&oi->ip_lock); 114 spin_lock(&oi->ip_lock);
99 115
100 /* Check that the inode hasn't been wi 116 /* Check that the inode hasn't been wiped from disk by another
101 * node. If it hasn't then we're safe 117 * node. If it hasn't then we're safe as long as we hold the
102 * spin lock until our increment of op 118 * spin lock until our increment of open count. */
103 if (oi->ip_flags & OCFS2_INODE_DELETED !! 119 if (OCFS2_I(inode)->ip_flags & OCFS2_INODE_DELETED) {
104 spin_unlock(&oi->ip_lock); 120 spin_unlock(&oi->ip_lock);
105 121
106 status = -ENOENT; 122 status = -ENOENT;
107 goto leave; 123 goto leave;
108 } 124 }
109 125
110 if (mode & O_DIRECT) 126 if (mode & O_DIRECT)
111 oi->ip_flags |= OCFS2_INODE_OP 127 oi->ip_flags |= OCFS2_INODE_OPEN_DIRECT;
112 128
113 oi->ip_open_count++; 129 oi->ip_open_count++;
114 spin_unlock(&oi->ip_lock); 130 spin_unlock(&oi->ip_lock);
115 131
116 status = ocfs2_init_file_private(inode 132 status = ocfs2_init_file_private(inode, file);
117 if (status) { 133 if (status) {
118 /* 134 /*
119 * We want to set open count b 135 * We want to set open count back if we're failing the
120 * open. 136 * open.
121 */ 137 */
122 spin_lock(&oi->ip_lock); 138 spin_lock(&oi->ip_lock);
123 oi->ip_open_count--; 139 oi->ip_open_count--;
124 spin_unlock(&oi->ip_lock); 140 spin_unlock(&oi->ip_lock);
125 } 141 }
126 142
127 file->f_mode |= FMODE_NOWAIT; <<
128 <<
129 leave: 143 leave:
130 return status; 144 return status;
131 } 145 }
132 146
133 static int ocfs2_file_release(struct inode *in 147 static int ocfs2_file_release(struct inode *inode, struct file *file)
134 { 148 {
135 struct ocfs2_inode_info *oi = OCFS2_I( 149 struct ocfs2_inode_info *oi = OCFS2_I(inode);
136 150
137 spin_lock(&oi->ip_lock); 151 spin_lock(&oi->ip_lock);
138 if (!--oi->ip_open_count) 152 if (!--oi->ip_open_count)
139 oi->ip_flags &= ~OCFS2_INODE_O 153 oi->ip_flags &= ~OCFS2_INODE_OPEN_DIRECT;
140 154
141 trace_ocfs2_file_release(inode, file, 155 trace_ocfs2_file_release(inode, file, file->f_path.dentry,
142 oi->ip_blkno, 156 oi->ip_blkno,
143 file->f_path. 157 file->f_path.dentry->d_name.len,
144 file->f_path. 158 file->f_path.dentry->d_name.name,
145 oi->ip_open_c 159 oi->ip_open_count);
146 spin_unlock(&oi->ip_lock); 160 spin_unlock(&oi->ip_lock);
147 161
148 ocfs2_free_file_private(inode, file); 162 ocfs2_free_file_private(inode, file);
149 163
150 return 0; 164 return 0;
151 } 165 }
152 166
153 static int ocfs2_dir_open(struct inode *inode, 167 static int ocfs2_dir_open(struct inode *inode, struct file *file)
154 { 168 {
155 return ocfs2_init_file_private(inode, 169 return ocfs2_init_file_private(inode, file);
156 } 170 }
157 171
158 static int ocfs2_dir_release(struct inode *ino 172 static int ocfs2_dir_release(struct inode *inode, struct file *file)
159 { 173 {
160 ocfs2_free_file_private(inode, file); 174 ocfs2_free_file_private(inode, file);
161 return 0; 175 return 0;
162 } 176 }
163 177
164 static int ocfs2_sync_file(struct file *file, 178 static int ocfs2_sync_file(struct file *file, loff_t start, loff_t end,
165 int datasync) 179 int datasync)
166 { 180 {
167 int err = 0; 181 int err = 0;
168 struct inode *inode = file->f_mapping- 182 struct inode *inode = file->f_mapping->host;
169 struct ocfs2_super *osb = OCFS2_SB(ino 183 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
170 struct ocfs2_inode_info *oi = OCFS2_I( 184 struct ocfs2_inode_info *oi = OCFS2_I(inode);
171 journal_t *journal = osb->journal->j_j 185 journal_t *journal = osb->journal->j_journal;
172 int ret; 186 int ret;
173 tid_t commit_tid; 187 tid_t commit_tid;
174 bool needs_barrier = false; 188 bool needs_barrier = false;
175 189
176 trace_ocfs2_sync_file(inode, file, fil 190 trace_ocfs2_sync_file(inode, file, file->f_path.dentry,
177 oi->ip_blkno, !! 191 OCFS2_I(inode)->ip_blkno,
178 file->f_path.den 192 file->f_path.dentry->d_name.len,
179 file->f_path.den 193 file->f_path.dentry->d_name.name,
180 (unsigned long l 194 (unsigned long long)datasync);
181 195
182 if (ocfs2_is_hard_readonly(osb) || ocf 196 if (ocfs2_is_hard_readonly(osb) || ocfs2_is_soft_readonly(osb))
183 return -EROFS; 197 return -EROFS;
184 198
185 err = file_write_and_wait_range(file, !! 199 err = filemap_write_and_wait_range(inode->i_mapping, start, end);
186 if (err) 200 if (err)
187 return err; 201 return err;
188 202
189 commit_tid = datasync ? oi->i_datasync 203 commit_tid = datasync ? oi->i_datasync_tid : oi->i_sync_tid;
190 if (journal->j_flags & JBD2_BARRIER && 204 if (journal->j_flags & JBD2_BARRIER &&
191 !jbd2_trans_will_send_data_barrier 205 !jbd2_trans_will_send_data_barrier(journal, commit_tid))
192 needs_barrier = true; 206 needs_barrier = true;
193 err = jbd2_complete_transaction(journa 207 err = jbd2_complete_transaction(journal, commit_tid);
194 if (needs_barrier) { 208 if (needs_barrier) {
195 ret = blkdev_issue_flush(inode !! 209 ret = blkdev_issue_flush(inode->i_sb->s_bdev, GFP_KERNEL, NULL);
196 if (!err) 210 if (!err)
197 err = ret; 211 err = ret;
198 } 212 }
199 213
200 if (err) 214 if (err)
201 mlog_errno(err); 215 mlog_errno(err);
202 216
203 return (err < 0) ? -EIO : 0; 217 return (err < 0) ? -EIO : 0;
204 } 218 }
205 219
206 int ocfs2_should_update_atime(struct inode *in 220 int ocfs2_should_update_atime(struct inode *inode,
207 struct vfsmount 221 struct vfsmount *vfsmnt)
208 { 222 {
209 struct timespec64 now; !! 223 struct timespec now;
210 struct ocfs2_super *osb = OCFS2_SB(ino 224 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
211 225
212 if (ocfs2_is_hard_readonly(osb) || ocf 226 if (ocfs2_is_hard_readonly(osb) || ocfs2_is_soft_readonly(osb))
213 return 0; 227 return 0;
214 228
215 if ((inode->i_flags & S_NOATIME) || 229 if ((inode->i_flags & S_NOATIME) ||
216 ((inode->i_sb->s_flags & SB_NODIRA !! 230 ((inode->i_sb->s_flags & MS_NODIRATIME) && S_ISDIR(inode->i_mode)))
217 return 0; 231 return 0;
218 232
219 /* 233 /*
220 * We can be called with no vfsmnt str 234 * We can be called with no vfsmnt structure - NFSD will
221 * sometimes do this. 235 * sometimes do this.
222 * 236 *
223 * Note that our action here is differ 237 * Note that our action here is different than touch_atime() -
224 * if we can't tell whether this is a 238 * if we can't tell whether this is a noatime mount, then we
225 * don't know whether to trust the val 239 * don't know whether to trust the value of s_atime_quantum.
226 */ 240 */
227 if (vfsmnt == NULL) 241 if (vfsmnt == NULL)
228 return 0; 242 return 0;
229 243
230 if ((vfsmnt->mnt_flags & MNT_NOATIME) 244 if ((vfsmnt->mnt_flags & MNT_NOATIME) ||
231 ((vfsmnt->mnt_flags & MNT_NODIRATI 245 ((vfsmnt->mnt_flags & MNT_NODIRATIME) && S_ISDIR(inode->i_mode)))
232 return 0; 246 return 0;
233 247
234 if (vfsmnt->mnt_flags & MNT_RELATIME) 248 if (vfsmnt->mnt_flags & MNT_RELATIME) {
235 struct timespec64 ctime = inod !! 249 if ((timespec_compare(&inode->i_atime, &inode->i_mtime) <= 0) ||
236 struct timespec64 atime = inod !! 250 (timespec_compare(&inode->i_atime, &inode->i_ctime) <= 0))
237 struct timespec64 mtime = inod <<
238 <<
239 if ((timespec64_compare(&atime <<
240 (timespec64_compare(&atime <<
241 return 1; 251 return 1;
242 252
243 return 0; 253 return 0;
244 } 254 }
245 255
246 now = current_time(inode); 256 now = current_time(inode);
247 if ((now.tv_sec - inode_get_atime_sec( !! 257 if ((now.tv_sec - inode->i_atime.tv_sec <= osb->s_atime_quantum))
248 return 0; 258 return 0;
249 else 259 else
250 return 1; 260 return 1;
251 } 261 }
252 262
253 int ocfs2_update_inode_atime(struct inode *ino 263 int ocfs2_update_inode_atime(struct inode *inode,
254 struct buffer_hea 264 struct buffer_head *bh)
255 { 265 {
256 int ret; 266 int ret;
257 struct ocfs2_super *osb = OCFS2_SB(ino 267 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
258 handle_t *handle; 268 handle_t *handle;
259 struct ocfs2_dinode *di = (struct ocfs 269 struct ocfs2_dinode *di = (struct ocfs2_dinode *) bh->b_data;
260 270
261 handle = ocfs2_start_trans(osb, OCFS2_ 271 handle = ocfs2_start_trans(osb, OCFS2_INODE_UPDATE_CREDITS);
262 if (IS_ERR(handle)) { 272 if (IS_ERR(handle)) {
263 ret = PTR_ERR(handle); 273 ret = PTR_ERR(handle);
264 mlog_errno(ret); 274 mlog_errno(ret);
265 goto out; 275 goto out;
266 } 276 }
267 277
268 ret = ocfs2_journal_access_di(handle, 278 ret = ocfs2_journal_access_di(handle, INODE_CACHE(inode), bh,
269 OCFS2_JO 279 OCFS2_JOURNAL_ACCESS_WRITE);
270 if (ret) { 280 if (ret) {
271 mlog_errno(ret); 281 mlog_errno(ret);
272 goto out_commit; 282 goto out_commit;
273 } 283 }
274 284
275 /* 285 /*
276 * Don't use ocfs2_mark_inode_dirty() 286 * Don't use ocfs2_mark_inode_dirty() here as we don't always
277 * have i_rwsem to guard against concu !! 287 * have i_mutex to guard against concurrent changes to other
278 * inode fields. 288 * inode fields.
279 */ 289 */
280 inode_set_atime_to_ts(inode, current_t !! 290 inode->i_atime = current_time(inode);
281 di->i_atime = cpu_to_le64(inode_get_at !! 291 di->i_atime = cpu_to_le64(inode->i_atime.tv_sec);
282 di->i_atime_nsec = cpu_to_le32(inode_g !! 292 di->i_atime_nsec = cpu_to_le32(inode->i_atime.tv_nsec);
283 ocfs2_update_inode_fsync_trans(handle, 293 ocfs2_update_inode_fsync_trans(handle, inode, 0);
284 ocfs2_journal_dirty(handle, bh); 294 ocfs2_journal_dirty(handle, bh);
285 295
286 out_commit: 296 out_commit:
287 ocfs2_commit_trans(osb, handle); !! 297 ocfs2_commit_trans(OCFS2_SB(inode->i_sb), handle);
288 out: 298 out:
289 return ret; 299 return ret;
290 } 300 }
291 301
292 int ocfs2_set_inode_size(handle_t *handle, 302 int ocfs2_set_inode_size(handle_t *handle,
293 struct inode * 303 struct inode *inode,
294 struct buffer_ 304 struct buffer_head *fe_bh,
295 u64 new_i_size 305 u64 new_i_size)
296 { 306 {
297 int status; 307 int status;
298 308
299 i_size_write(inode, new_i_size); 309 i_size_write(inode, new_i_size);
300 inode->i_blocks = ocfs2_inode_sector_c 310 inode->i_blocks = ocfs2_inode_sector_count(inode);
301 inode_set_mtime_to_ts(inode, inode_set !! 311 inode->i_ctime = inode->i_mtime = current_time(inode);
302 312
303 status = ocfs2_mark_inode_dirty(handle 313 status = ocfs2_mark_inode_dirty(handle, inode, fe_bh);
304 if (status < 0) { 314 if (status < 0) {
305 mlog_errno(status); 315 mlog_errno(status);
306 goto bail; 316 goto bail;
307 } 317 }
308 318
309 bail: 319 bail:
310 return status; 320 return status;
311 } 321 }
312 322
313 int ocfs2_simple_size_update(struct inode *ino 323 int ocfs2_simple_size_update(struct inode *inode,
314 struct buffer_hea 324 struct buffer_head *di_bh,
315 u64 new_i_size) 325 u64 new_i_size)
316 { 326 {
317 int ret; 327 int ret;
318 struct ocfs2_super *osb = OCFS2_SB(ino 328 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
319 handle_t *handle = NULL; 329 handle_t *handle = NULL;
320 330
321 handle = ocfs2_start_trans(osb, OCFS2_ 331 handle = ocfs2_start_trans(osb, OCFS2_INODE_UPDATE_CREDITS);
322 if (IS_ERR(handle)) { 332 if (IS_ERR(handle)) {
323 ret = PTR_ERR(handle); 333 ret = PTR_ERR(handle);
324 mlog_errno(ret); 334 mlog_errno(ret);
325 goto out; 335 goto out;
326 } 336 }
327 337
328 ret = ocfs2_set_inode_size(handle, ino 338 ret = ocfs2_set_inode_size(handle, inode, di_bh,
329 new_i_size) 339 new_i_size);
330 if (ret < 0) 340 if (ret < 0)
331 mlog_errno(ret); 341 mlog_errno(ret);
332 342
333 ocfs2_update_inode_fsync_trans(handle, 343 ocfs2_update_inode_fsync_trans(handle, inode, 0);
334 ocfs2_commit_trans(osb, handle); 344 ocfs2_commit_trans(osb, handle);
335 out: 345 out:
336 return ret; 346 return ret;
337 } 347 }
338 348
339 static int ocfs2_cow_file_pos(struct inode *in 349 static int ocfs2_cow_file_pos(struct inode *inode,
340 struct buffer_he 350 struct buffer_head *fe_bh,
341 u64 offset) 351 u64 offset)
342 { 352 {
343 int status; 353 int status;
344 u32 phys, cpos = offset >> OCFS2_SB(in 354 u32 phys, cpos = offset >> OCFS2_SB(inode->i_sb)->s_clustersize_bits;
345 unsigned int num_clusters = 0; 355 unsigned int num_clusters = 0;
346 unsigned int ext_flags = 0; 356 unsigned int ext_flags = 0;
347 357
348 /* 358 /*
349 * If the new offset is aligned to the 359 * If the new offset is aligned to the range of the cluster, there is
350 * no space for ocfs2_zero_range_for_t 360 * no space for ocfs2_zero_range_for_truncate to fill, so no need to
351 * CoW either. 361 * CoW either.
352 */ 362 */
353 if ((offset & (OCFS2_SB(inode->i_sb)-> 363 if ((offset & (OCFS2_SB(inode->i_sb)->s_clustersize - 1)) == 0)
354 return 0; 364 return 0;
355 365
356 status = ocfs2_get_clusters(inode, cpo 366 status = ocfs2_get_clusters(inode, cpos, &phys,
357 &num_clust 367 &num_clusters, &ext_flags);
358 if (status) { 368 if (status) {
359 mlog_errno(status); 369 mlog_errno(status);
360 goto out; 370 goto out;
361 } 371 }
362 372
363 if (!(ext_flags & OCFS2_EXT_REFCOUNTED 373 if (!(ext_flags & OCFS2_EXT_REFCOUNTED))
364 goto out; 374 goto out;
365 375
366 return ocfs2_refcount_cow(inode, fe_bh 376 return ocfs2_refcount_cow(inode, fe_bh, cpos, 1, cpos+1);
367 377
368 out: 378 out:
369 return status; 379 return status;
370 } 380 }
371 381
372 static int ocfs2_orphan_for_truncate(struct oc 382 static int ocfs2_orphan_for_truncate(struct ocfs2_super *osb,
373 struct in 383 struct inode *inode,
374 struct bu 384 struct buffer_head *fe_bh,
375 u64 new_i 385 u64 new_i_size)
376 { 386 {
377 int status; 387 int status;
378 handle_t *handle; 388 handle_t *handle;
379 struct ocfs2_dinode *di; 389 struct ocfs2_dinode *di;
380 u64 cluster_bytes; 390 u64 cluster_bytes;
381 391
382 /* 392 /*
383 * We need to CoW the cluster contains 393 * We need to CoW the cluster contains the offset if it is reflinked
384 * since we will call ocfs2_zero_range 394 * since we will call ocfs2_zero_range_for_truncate later which will
385 * write "" from offset to the end of 395 * write "" from offset to the end of the cluster.
386 */ 396 */
387 status = ocfs2_cow_file_pos(inode, fe_ 397 status = ocfs2_cow_file_pos(inode, fe_bh, new_i_size);
388 if (status) { 398 if (status) {
389 mlog_errno(status); 399 mlog_errno(status);
390 return status; 400 return status;
391 } 401 }
392 402
393 /* TODO: This needs to actually orphan 403 /* TODO: This needs to actually orphan the inode in this
394 * transaction. */ 404 * transaction. */
395 405
396 handle = ocfs2_start_trans(osb, OCFS2_ 406 handle = ocfs2_start_trans(osb, OCFS2_INODE_UPDATE_CREDITS);
397 if (IS_ERR(handle)) { 407 if (IS_ERR(handle)) {
398 status = PTR_ERR(handle); 408 status = PTR_ERR(handle);
399 mlog_errno(status); 409 mlog_errno(status);
400 goto out; 410 goto out;
401 } 411 }
402 412
403 status = ocfs2_journal_access_di(handl 413 status = ocfs2_journal_access_di(handle, INODE_CACHE(inode), fe_bh,
404 OCFS2 414 OCFS2_JOURNAL_ACCESS_WRITE);
405 if (status < 0) { 415 if (status < 0) {
406 mlog_errno(status); 416 mlog_errno(status);
407 goto out_commit; 417 goto out_commit;
408 } 418 }
409 419
410 /* 420 /*
411 * Do this before setting i_size. 421 * Do this before setting i_size.
412 */ 422 */
413 cluster_bytes = ocfs2_align_bytes_to_c 423 cluster_bytes = ocfs2_align_bytes_to_clusters(inode->i_sb, new_i_size);
414 status = ocfs2_zero_range_for_truncate 424 status = ocfs2_zero_range_for_truncate(inode, handle, new_i_size,
415 425 cluster_bytes);
416 if (status) { 426 if (status) {
417 mlog_errno(status); 427 mlog_errno(status);
418 goto out_commit; 428 goto out_commit;
419 } 429 }
420 430
421 i_size_write(inode, new_i_size); 431 i_size_write(inode, new_i_size);
422 inode_set_mtime_to_ts(inode, inode_set !! 432 inode->i_ctime = inode->i_mtime = current_time(inode);
423 433
424 di = (struct ocfs2_dinode *) fe_bh->b_ 434 di = (struct ocfs2_dinode *) fe_bh->b_data;
425 di->i_size = cpu_to_le64(new_i_size); 435 di->i_size = cpu_to_le64(new_i_size);
426 di->i_ctime = di->i_mtime = cpu_to_le6 !! 436 di->i_ctime = di->i_mtime = cpu_to_le64(inode->i_ctime.tv_sec);
427 di->i_ctime_nsec = di->i_mtime_nsec = !! 437 di->i_ctime_nsec = di->i_mtime_nsec = cpu_to_le32(inode->i_ctime.tv_nsec);
428 ocfs2_update_inode_fsync_trans(handle, 438 ocfs2_update_inode_fsync_trans(handle, inode, 0);
429 439
430 ocfs2_journal_dirty(handle, fe_bh); 440 ocfs2_journal_dirty(handle, fe_bh);
431 441
432 out_commit: 442 out_commit:
433 ocfs2_commit_trans(osb, handle); 443 ocfs2_commit_trans(osb, handle);
434 out: 444 out:
435 return status; 445 return status;
436 } 446 }
437 447
438 int ocfs2_truncate_file(struct inode *inode, 448 int ocfs2_truncate_file(struct inode *inode,
439 struct buffer_h 449 struct buffer_head *di_bh,
440 u64 new_i_size) 450 u64 new_i_size)
441 { 451 {
442 int status = 0; 452 int status = 0;
443 struct ocfs2_dinode *fe = NULL; 453 struct ocfs2_dinode *fe = NULL;
444 struct ocfs2_super *osb = OCFS2_SB(ino 454 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
445 455
446 /* We trust di_bh because it comes fro 456 /* We trust di_bh because it comes from ocfs2_inode_lock(), which
447 * already validated it */ 457 * already validated it */
448 fe = (struct ocfs2_dinode *) di_bh->b_ 458 fe = (struct ocfs2_dinode *) di_bh->b_data;
449 459
450 trace_ocfs2_truncate_file((unsigned lo 460 trace_ocfs2_truncate_file((unsigned long long)OCFS2_I(inode)->ip_blkno,
451 (unsigned lo 461 (unsigned long long)le64_to_cpu(fe->i_size),
452 (unsigned lo 462 (unsigned long long)new_i_size);
453 463
454 mlog_bug_on_msg(le64_to_cpu(fe->i_size 464 mlog_bug_on_msg(le64_to_cpu(fe->i_size) != i_size_read(inode),
455 "Inode %llu, inode i_s 465 "Inode %llu, inode i_size = %lld != di "
456 "i_size = %llu, i_flag 466 "i_size = %llu, i_flags = 0x%x\n",
457 (unsigned long long)OC 467 (unsigned long long)OCFS2_I(inode)->ip_blkno,
458 i_size_read(inode), 468 i_size_read(inode),
459 (unsigned long long)le 469 (unsigned long long)le64_to_cpu(fe->i_size),
460 le32_to_cpu(fe->i_flag 470 le32_to_cpu(fe->i_flags));
461 471
462 if (new_i_size > le64_to_cpu(fe->i_siz 472 if (new_i_size > le64_to_cpu(fe->i_size)) {
463 trace_ocfs2_truncate_file_erro 473 trace_ocfs2_truncate_file_error(
464 (unsigned long long)le 474 (unsigned long long)le64_to_cpu(fe->i_size),
465 (unsigned long long)ne 475 (unsigned long long)new_i_size);
466 status = -EINVAL; 476 status = -EINVAL;
467 mlog_errno(status); 477 mlog_errno(status);
468 goto bail; 478 goto bail;
469 } 479 }
470 480
471 down_write(&OCFS2_I(inode)->ip_alloc_s 481 down_write(&OCFS2_I(inode)->ip_alloc_sem);
472 482
473 ocfs2_resv_discard(&osb->osb_la_resmap 483 ocfs2_resv_discard(&osb->osb_la_resmap,
474 &OCFS2_I(inode)->ip 484 &OCFS2_I(inode)->ip_la_data_resv);
475 485
476 /* 486 /*
477 * The inode lock forced other nodes t 487 * The inode lock forced other nodes to sync and drop their
478 * pages, which (correctly) happens ev 488 * pages, which (correctly) happens even if we have a truncate
479 * without allocation change - ocfs2 c 489 * without allocation change - ocfs2 cluster sizes can be much
480 * greater than page size, so we have 490 * greater than page size, so we have to truncate them
481 * anyway. 491 * anyway.
482 */ 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);
483 495
484 if (OCFS2_I(inode)->ip_dyn_features & 496 if (OCFS2_I(inode)->ip_dyn_features & OCFS2_INLINE_DATA_FL) {
485 unmap_mapping_range(inode->i_m <<
486 new_i_size <<
487 truncate_inode_pages(inode->i_ <<
488 status = ocfs2_truncate_inline 497 status = ocfs2_truncate_inline(inode, di_bh, new_i_size,
489 498 i_size_read(inode), 1);
490 if (status) 499 if (status)
491 mlog_errno(status); 500 mlog_errno(status);
492 501
493 goto bail_unlock_sem; 502 goto bail_unlock_sem;
494 } 503 }
495 504
496 /* alright, we're going to need to do 505 /* alright, we're going to need to do a full blown alloc size
497 * change. Orphan the inode so that re 506 * change. Orphan the inode so that recovery can complete the
498 * truncate if necessary. This does th 507 * truncate if necessary. This does the task of marking
499 * i_size. */ 508 * i_size. */
500 status = ocfs2_orphan_for_truncate(osb 509 status = ocfs2_orphan_for_truncate(osb, inode, di_bh, new_i_size);
501 if (status < 0) { 510 if (status < 0) {
502 mlog_errno(status); 511 mlog_errno(status);
503 goto bail_unlock_sem; 512 goto bail_unlock_sem;
504 } 513 }
505 514
506 unmap_mapping_range(inode->i_mapping, <<
507 truncate_inode_pages(inode->i_mapping, <<
508 <<
509 status = ocfs2_commit_truncate(osb, in 515 status = ocfs2_commit_truncate(osb, inode, di_bh);
510 if (status < 0) { 516 if (status < 0) {
511 mlog_errno(status); 517 mlog_errno(status);
512 goto bail_unlock_sem; 518 goto bail_unlock_sem;
513 } 519 }
514 520
515 /* TODO: orphan dir cleanup here. */ 521 /* TODO: orphan dir cleanup here. */
516 bail_unlock_sem: 522 bail_unlock_sem:
517 up_write(&OCFS2_I(inode)->ip_alloc_sem 523 up_write(&OCFS2_I(inode)->ip_alloc_sem);
518 524
519 bail: 525 bail:
520 if (!status && OCFS2_I(inode)->ip_clus 526 if (!status && OCFS2_I(inode)->ip_clusters == 0)
521 status = ocfs2_try_remove_refc 527 status = ocfs2_try_remove_refcount_tree(inode, di_bh);
522 528
523 return status; 529 return status;
524 } 530 }
525 531
526 /* 532 /*
527 * extend file allocation only here. 533 * extend file allocation only here.
528 * we'll update all the disk stuff, and oip->a 534 * we'll update all the disk stuff, and oip->alloc_size
529 * 535 *
530 * expect stuff to be locked, a transaction st 536 * expect stuff to be locked, a transaction started and enough data /
531 * metadata reservations in the contexts. 537 * metadata reservations in the contexts.
532 * 538 *
533 * Will return -EAGAIN, and a reason if a rest 539 * Will return -EAGAIN, and a reason if a restart is needed.
534 * If passed in, *reason will always be set, e 540 * If passed in, *reason will always be set, even in error.
535 */ 541 */
536 int ocfs2_add_inode_data(struct ocfs2_super *o 542 int ocfs2_add_inode_data(struct ocfs2_super *osb,
537 struct inode *inode, 543 struct inode *inode,
538 u32 *logical_offset, 544 u32 *logical_offset,
539 u32 clusters_to_add, 545 u32 clusters_to_add,
540 int mark_unwritten, 546 int mark_unwritten,
541 struct buffer_head *f 547 struct buffer_head *fe_bh,
542 handle_t *handle, 548 handle_t *handle,
543 struct ocfs2_alloc_co 549 struct ocfs2_alloc_context *data_ac,
544 struct ocfs2_alloc_co 550 struct ocfs2_alloc_context *meta_ac,
545 enum ocfs2_alloc_rest 551 enum ocfs2_alloc_restarted *reason_ret)
546 { 552 {
>> 553 int ret;
547 struct ocfs2_extent_tree et; 554 struct ocfs2_extent_tree et;
548 555
549 ocfs2_init_dinode_extent_tree(&et, INO 556 ocfs2_init_dinode_extent_tree(&et, INODE_CACHE(inode), fe_bh);
550 return ocfs2_add_clusters_in_btree(han !! 557 ret = ocfs2_add_clusters_in_btree(handle, &et, logical_offset,
551 clu !! 558 clusters_to_add, mark_unwritten,
552 dat !! 559 data_ac, meta_ac, reason_ret);
>> 560
>> 561 return ret;
553 } 562 }
554 563
555 static int ocfs2_extend_allocation(struct inod !! 564 static int __ocfs2_extend_allocation(struct inode *inode, u32 logical_start,
556 u32 cluster !! 565 u32 clusters_to_add, int mark_unwritten)
557 { 566 {
558 int status = 0; 567 int status = 0;
559 int restart_func = 0; 568 int restart_func = 0;
560 int credits; 569 int credits;
561 u32 prev_clusters; 570 u32 prev_clusters;
562 struct buffer_head *bh = NULL; 571 struct buffer_head *bh = NULL;
563 struct ocfs2_dinode *fe = NULL; 572 struct ocfs2_dinode *fe = NULL;
564 handle_t *handle = NULL; 573 handle_t *handle = NULL;
565 struct ocfs2_alloc_context *data_ac = 574 struct ocfs2_alloc_context *data_ac = NULL;
566 struct ocfs2_alloc_context *meta_ac = 575 struct ocfs2_alloc_context *meta_ac = NULL;
567 enum ocfs2_alloc_restarted why = RESTA 576 enum ocfs2_alloc_restarted why = RESTART_NONE;
568 struct ocfs2_super *osb = OCFS2_SB(ino 577 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
569 struct ocfs2_extent_tree et; 578 struct ocfs2_extent_tree et;
570 int did_quota = 0; 579 int did_quota = 0;
571 580
572 /* 581 /*
573 * Unwritten extent only exists for fi 582 * Unwritten extent only exists for file systems which
574 * support holes. 583 * support holes.
575 */ 584 */
576 BUG_ON(mark_unwritten && !ocfs2_sparse 585 BUG_ON(mark_unwritten && !ocfs2_sparse_alloc(osb));
577 586
578 status = ocfs2_read_inode_block(inode, 587 status = ocfs2_read_inode_block(inode, &bh);
579 if (status < 0) { 588 if (status < 0) {
580 mlog_errno(status); 589 mlog_errno(status);
581 goto leave; 590 goto leave;
582 } 591 }
583 fe = (struct ocfs2_dinode *) bh->b_dat 592 fe = (struct ocfs2_dinode *) bh->b_data;
584 593
585 restart_all: 594 restart_all:
586 BUG_ON(le32_to_cpu(fe->i_clusters) != 595 BUG_ON(le32_to_cpu(fe->i_clusters) != OCFS2_I(inode)->ip_clusters);
587 596
588 ocfs2_init_dinode_extent_tree(&et, INO 597 ocfs2_init_dinode_extent_tree(&et, INODE_CACHE(inode), bh);
589 status = ocfs2_lock_allocators(inode, 598 status = ocfs2_lock_allocators(inode, &et, clusters_to_add, 0,
590 &data_a 599 &data_ac, &meta_ac);
591 if (status) { 600 if (status) {
592 mlog_errno(status); 601 mlog_errno(status);
593 goto leave; 602 goto leave;
594 } 603 }
595 604
596 credits = ocfs2_calc_extend_credits(os 605 credits = ocfs2_calc_extend_credits(osb->sb, &fe->id2.i_list);
597 handle = ocfs2_start_trans(osb, credit 606 handle = ocfs2_start_trans(osb, credits);
598 if (IS_ERR(handle)) { 607 if (IS_ERR(handle)) {
599 status = PTR_ERR(handle); 608 status = PTR_ERR(handle);
600 handle = NULL; 609 handle = NULL;
601 mlog_errno(status); 610 mlog_errno(status);
602 goto leave; 611 goto leave;
603 } 612 }
604 613
605 restarted_transaction: 614 restarted_transaction:
606 trace_ocfs2_extend_allocation( 615 trace_ocfs2_extend_allocation(
607 (unsigned long long)OCFS2_I(in 616 (unsigned long long)OCFS2_I(inode)->ip_blkno,
608 (unsigned long long)i_size_rea 617 (unsigned long long)i_size_read(inode),
609 le32_to_cpu(fe->i_clusters), c 618 le32_to_cpu(fe->i_clusters), clusters_to_add,
610 why, restart_func); 619 why, restart_func);
611 620
612 status = dquot_alloc_space_nodirty(ino 621 status = dquot_alloc_space_nodirty(inode,
613 ocfs2_clusters_to_byte 622 ocfs2_clusters_to_bytes(osb->sb, clusters_to_add));
614 if (status) 623 if (status)
615 goto leave; 624 goto leave;
616 did_quota = 1; 625 did_quota = 1;
617 626
618 /* reserve a write to the file entry e 627 /* reserve a write to the file entry early on - that we if we
619 * run out of credits in the allocatio 628 * run out of credits in the allocation path, we can still
620 * update i_size. */ 629 * update i_size. */
621 status = ocfs2_journal_access_di(handl 630 status = ocfs2_journal_access_di(handle, INODE_CACHE(inode), bh,
622 OCFS2 631 OCFS2_JOURNAL_ACCESS_WRITE);
623 if (status < 0) { 632 if (status < 0) {
624 mlog_errno(status); 633 mlog_errno(status);
625 goto leave; 634 goto leave;
626 } 635 }
627 636
628 prev_clusters = OCFS2_I(inode)->ip_clu 637 prev_clusters = OCFS2_I(inode)->ip_clusters;
629 638
630 status = ocfs2_add_inode_data(osb, 639 status = ocfs2_add_inode_data(osb,
631 inode, 640 inode,
632 &logical 641 &logical_start,
633 clusters 642 clusters_to_add,
634 mark_unw 643 mark_unwritten,
635 bh, 644 bh,
636 handle, 645 handle,
637 data_ac, 646 data_ac,
638 meta_ac, 647 meta_ac,
639 &why); 648 &why);
640 if ((status < 0) && (status != -EAGAIN 649 if ((status < 0) && (status != -EAGAIN)) {
641 if (status != -ENOSPC) 650 if (status != -ENOSPC)
642 mlog_errno(status); 651 mlog_errno(status);
643 goto leave; 652 goto leave;
644 } 653 }
645 ocfs2_update_inode_fsync_trans(handle, 654 ocfs2_update_inode_fsync_trans(handle, inode, 1);
646 ocfs2_journal_dirty(handle, bh); 655 ocfs2_journal_dirty(handle, bh);
647 656
648 spin_lock(&OCFS2_I(inode)->ip_lock); 657 spin_lock(&OCFS2_I(inode)->ip_lock);
649 clusters_to_add -= (OCFS2_I(inode)->ip 658 clusters_to_add -= (OCFS2_I(inode)->ip_clusters - prev_clusters);
650 spin_unlock(&OCFS2_I(inode)->ip_lock); 659 spin_unlock(&OCFS2_I(inode)->ip_lock);
651 /* Release unused quota reservation */ 660 /* Release unused quota reservation */
652 dquot_free_space(inode, 661 dquot_free_space(inode,
653 ocfs2_clusters_to_byte 662 ocfs2_clusters_to_bytes(osb->sb, clusters_to_add));
654 did_quota = 0; 663 did_quota = 0;
655 664
656 if (why != RESTART_NONE && clusters_to 665 if (why != RESTART_NONE && clusters_to_add) {
657 if (why == RESTART_META) { 666 if (why == RESTART_META) {
658 restart_func = 1; 667 restart_func = 1;
659 status = 0; 668 status = 0;
660 } else { 669 } else {
661 BUG_ON(why != RESTART_ 670 BUG_ON(why != RESTART_TRANS);
662 671
663 status = ocfs2_allocat 672 status = ocfs2_allocate_extend_trans(handle, 1);
664 if (status < 0) { 673 if (status < 0) {
665 /* handle stil 674 /* handle still has to be committed at
666 * this point. 675 * this point. */
667 status = -ENOM 676 status = -ENOMEM;
668 mlog_errno(sta 677 mlog_errno(status);
669 goto leave; 678 goto leave;
670 } 679 }
671 goto restarted_transac 680 goto restarted_transaction;
672 } 681 }
673 } 682 }
674 683
675 trace_ocfs2_extend_allocation_end(OCFS 684 trace_ocfs2_extend_allocation_end(OCFS2_I(inode)->ip_blkno,
676 le32_to_cpu(fe->i_clusters), 685 le32_to_cpu(fe->i_clusters),
677 (unsigned long long)le64_to_cpu(f 686 (unsigned long long)le64_to_cpu(fe->i_size),
678 OCFS2_I(inode)->ip_clusters, 687 OCFS2_I(inode)->ip_clusters,
679 (unsigned long long)i_size_read(i 688 (unsigned long long)i_size_read(inode));
680 689
681 leave: 690 leave:
682 if (status < 0 && did_quota) 691 if (status < 0 && did_quota)
683 dquot_free_space(inode, 692 dquot_free_space(inode,
684 ocfs2_clusters_to_byte 693 ocfs2_clusters_to_bytes(osb->sb, clusters_to_add));
685 if (handle) { 694 if (handle) {
686 ocfs2_commit_trans(osb, handle 695 ocfs2_commit_trans(osb, handle);
687 handle = NULL; 696 handle = NULL;
688 } 697 }
689 if (data_ac) { 698 if (data_ac) {
690 ocfs2_free_alloc_context(data_ 699 ocfs2_free_alloc_context(data_ac);
691 data_ac = NULL; 700 data_ac = NULL;
692 } 701 }
693 if (meta_ac) { 702 if (meta_ac) {
694 ocfs2_free_alloc_context(meta_ 703 ocfs2_free_alloc_context(meta_ac);
695 meta_ac = NULL; 704 meta_ac = NULL;
696 } 705 }
697 if ((!status) && restart_func) { 706 if ((!status) && restart_func) {
698 restart_func = 0; 707 restart_func = 0;
699 goto restart_all; 708 goto restart_all;
700 } 709 }
701 brelse(bh); 710 brelse(bh);
702 bh = NULL; 711 bh = NULL;
703 712
704 return status; 713 return status;
705 } 714 }
706 715
>> 716 int 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
707 /* 723 /*
708 * While a write will already be ordering the 724 * While a write will already be ordering the data, a truncate will not.
709 * Thus, we need to explicitly order the zeroe 725 * Thus, we need to explicitly order the zeroed pages.
710 */ 726 */
711 static handle_t *ocfs2_zero_start_ordered_tran 727 static handle_t *ocfs2_zero_start_ordered_transaction(struct inode *inode,
712 !! 728 struct buffer_head *di_bh)
713 <<
714 <<
715 { 729 {
716 struct ocfs2_super *osb = OCFS2_SB(ino 730 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
717 handle_t *handle = NULL; 731 handle_t *handle = NULL;
718 int ret = 0; 732 int ret = 0;
719 733
720 if (!ocfs2_should_order_data(inode)) 734 if (!ocfs2_should_order_data(inode))
721 goto out; 735 goto out;
722 736
723 handle = ocfs2_start_trans(osb, OCFS2_ 737 handle = ocfs2_start_trans(osb, OCFS2_INODE_UPDATE_CREDITS);
724 if (IS_ERR(handle)) { 738 if (IS_ERR(handle)) {
725 ret = -ENOMEM; 739 ret = -ENOMEM;
726 mlog_errno(ret); 740 mlog_errno(ret);
727 goto out; 741 goto out;
728 } 742 }
729 743
730 ret = ocfs2_jbd2_inode_add_write(handl !! 744 ret = ocfs2_jbd2_file_inode(handle, inode);
731 if (ret < 0) { 745 if (ret < 0) {
732 mlog_errno(ret); 746 mlog_errno(ret);
733 goto out; 747 goto out;
734 } 748 }
735 749
736 ret = ocfs2_journal_access_di(handle, 750 ret = ocfs2_journal_access_di(handle, INODE_CACHE(inode), di_bh,
737 OCFS2_JO 751 OCFS2_JOURNAL_ACCESS_WRITE);
738 if (ret) 752 if (ret)
739 mlog_errno(ret); 753 mlog_errno(ret);
740 ocfs2_update_inode_fsync_trans(handle, 754 ocfs2_update_inode_fsync_trans(handle, inode, 1);
741 755
742 out: 756 out:
743 if (ret) { 757 if (ret) {
744 if (!IS_ERR(handle)) 758 if (!IS_ERR(handle))
745 ocfs2_commit_trans(osb 759 ocfs2_commit_trans(osb, handle);
746 handle = ERR_PTR(ret); 760 handle = ERR_PTR(ret);
747 } 761 }
748 return handle; 762 return handle;
749 } 763 }
750 764
751 /* Some parts of this taken from generic_cont_ 765 /* Some parts of this taken from generic_cont_expand, which turned out
752 * to be too fragile to do exactly what we nee 766 * to be too fragile to do exactly what we need without us having to
753 * worry about recursive locking in ->write_be 767 * worry about recursive locking in ->write_begin() and ->write_end(). */
754 static int ocfs2_write_zero_page(struct inode 768 static int ocfs2_write_zero_page(struct inode *inode, u64 abs_from,
755 u64 abs_to, s 769 u64 abs_to, struct buffer_head *di_bh)
756 { 770 {
757 struct address_space *mapping = inode- 771 struct address_space *mapping = inode->i_mapping;
758 struct folio *folio; !! 772 struct page *page;
759 unsigned long index = abs_from >> PAGE 773 unsigned long index = abs_from >> PAGE_SHIFT;
760 handle_t *handle; 774 handle_t *handle;
761 int ret = 0; 775 int ret = 0;
762 unsigned zero_from, zero_to, block_sta 776 unsigned zero_from, zero_to, block_start, block_end;
763 struct ocfs2_dinode *di = (struct ocfs 777 struct ocfs2_dinode *di = (struct ocfs2_dinode *)di_bh->b_data;
764 778
765 BUG_ON(abs_from >= abs_to); 779 BUG_ON(abs_from >= abs_to);
766 BUG_ON(abs_to > (((u64)index + 1) << P 780 BUG_ON(abs_to > (((u64)index + 1) << PAGE_SHIFT));
767 BUG_ON(abs_from & (inode->i_blkbits - 781 BUG_ON(abs_from & (inode->i_blkbits - 1));
768 782
769 handle = ocfs2_zero_start_ordered_tran !! 783 handle = ocfs2_zero_start_ordered_transaction(inode, di_bh);
770 <<
771 <<
772 if (IS_ERR(handle)) { 784 if (IS_ERR(handle)) {
773 ret = PTR_ERR(handle); 785 ret = PTR_ERR(handle);
774 goto out; 786 goto out;
775 } 787 }
776 788
777 folio = __filemap_get_folio(mapping, i !! 789 page = find_or_create_page(mapping, index, GFP_NOFS);
778 FGP_LOCK | FGP_ACCESSE !! 790 if (!page) {
779 if (IS_ERR(folio)) { !! 791 ret = -ENOMEM;
780 ret = PTR_ERR(folio); <<
781 mlog_errno(ret); 792 mlog_errno(ret);
782 goto out_commit_trans; 793 goto out_commit_trans;
783 } 794 }
784 795
785 /* Get the offsets within the page tha 796 /* Get the offsets within the page that we want to zero */
786 zero_from = abs_from & (PAGE_SIZE - 1) 797 zero_from = abs_from & (PAGE_SIZE - 1);
787 zero_to = abs_to & (PAGE_SIZE - 1); 798 zero_to = abs_to & (PAGE_SIZE - 1);
788 if (!zero_to) 799 if (!zero_to)
789 zero_to = PAGE_SIZE; 800 zero_to = PAGE_SIZE;
790 801
791 trace_ocfs2_write_zero_page( 802 trace_ocfs2_write_zero_page(
792 (unsigned long long)OC 803 (unsigned long long)OCFS2_I(inode)->ip_blkno,
793 (unsigned long long)ab 804 (unsigned long long)abs_from,
794 (unsigned long long)ab 805 (unsigned long long)abs_to,
795 index, zero_from, zero 806 index, zero_from, zero_to);
796 807
797 /* We know that zero_from is block ali 808 /* We know that zero_from is block aligned */
798 for (block_start = zero_from; block_st 809 for (block_start = zero_from; block_start < zero_to;
799 block_start = block_end) { 810 block_start = block_end) {
800 block_end = block_start + i_bl 811 block_end = block_start + i_blocksize(inode);
801 812
802 /* 813 /*
803 * block_start is block-aligne 814 * block_start is block-aligned. Bump it by one to force
804 * __block_write_begin and blo 815 * __block_write_begin and block_commit_write to zero the
805 * whole block. 816 * whole block.
806 */ 817 */
807 ret = __block_write_begin(foli !! 818 ret = __block_write_begin(page, block_start + 1, 0,
808 ocfs 819 ocfs2_get_block);
809 if (ret < 0) { 820 if (ret < 0) {
810 mlog_errno(ret); 821 mlog_errno(ret);
811 goto out_unlock; 822 goto out_unlock;
812 } 823 }
813 824
814 825
815 /* must not update i_size! */ 826 /* must not update i_size! */
816 block_commit_write(&folio->pag !! 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;
817 } 833 }
818 834
819 /* 835 /*
820 * fs-writeback will release the dirty 836 * fs-writeback will release the dirty pages without page lock
821 * whose offset are over inode size, t 837 * whose offset are over inode size, the release happens at
822 * block_write_full_folio(). !! 838 * block_write_full_page().
823 */ 839 */
824 i_size_write(inode, abs_to); 840 i_size_write(inode, abs_to);
825 inode->i_blocks = ocfs2_inode_sector_c 841 inode->i_blocks = ocfs2_inode_sector_count(inode);
826 di->i_size = cpu_to_le64((u64)i_size_r 842 di->i_size = cpu_to_le64((u64)i_size_read(inode));
827 inode_set_mtime_to_ts(inode, inode_set !! 843 inode->i_mtime = inode->i_ctime = current_time(inode);
828 di->i_mtime = di->i_ctime = cpu_to_le6 !! 844 di->i_mtime = di->i_ctime = cpu_to_le64(inode->i_mtime.tv_sec);
829 di->i_ctime_nsec = cpu_to_le32(inode_g !! 845 di->i_ctime_nsec = cpu_to_le32(inode->i_mtime.tv_nsec);
830 di->i_mtime_nsec = di->i_ctime_nsec; 846 di->i_mtime_nsec = di->i_ctime_nsec;
831 if (handle) { 847 if (handle) {
832 ocfs2_journal_dirty(handle, di 848 ocfs2_journal_dirty(handle, di_bh);
833 ocfs2_update_inode_fsync_trans 849 ocfs2_update_inode_fsync_trans(handle, inode, 1);
834 } 850 }
835 851
836 out_unlock: 852 out_unlock:
837 folio_unlock(folio); !! 853 unlock_page(page);
838 folio_put(folio); !! 854 put_page(page);
839 out_commit_trans: 855 out_commit_trans:
840 if (handle) 856 if (handle)
841 ocfs2_commit_trans(OCFS2_SB(in 857 ocfs2_commit_trans(OCFS2_SB(inode->i_sb), handle);
842 out: 858 out:
843 return ret; 859 return ret;
844 } 860 }
845 861
846 /* 862 /*
847 * Find the next range to zero. We do this in 863 * Find the next range to zero. We do this in terms of bytes because
848 * that's what ocfs2_zero_extend() wants, and 864 * that's what ocfs2_zero_extend() wants, and it is dealing with the
849 * pagecache. We may return multiple extents. 865 * pagecache. We may return multiple extents.
850 * 866 *
851 * zero_start and zero_end are ocfs2_zero_exte 867 * zero_start and zero_end are ocfs2_zero_extend()s current idea of what
852 * needs to be zeroed. range_start and range_ 868 * needs to be zeroed. range_start and range_end return the next zeroing
853 * range. A subsequent call should pass the p 869 * range. A subsequent call should pass the previous range_end as its
854 * zero_start. If range_end is 0, there's not 870 * zero_start. If range_end is 0, there's nothing to do.
855 * 871 *
856 * Unwritten extents are skipped over. Refcou 872 * Unwritten extents are skipped over. Refcounted extents are CoWd.
857 */ 873 */
858 static int ocfs2_zero_extend_get_range(struct 874 static int ocfs2_zero_extend_get_range(struct inode *inode,
859 struct 875 struct buffer_head *di_bh,
860 u64 zer 876 u64 zero_start, u64 zero_end,
861 u64 *ra 877 u64 *range_start, u64 *range_end)
862 { 878 {
863 int rc = 0, needs_cow = 0; 879 int rc = 0, needs_cow = 0;
864 u32 p_cpos, zero_clusters = 0; 880 u32 p_cpos, zero_clusters = 0;
865 u32 zero_cpos = 881 u32 zero_cpos =
866 zero_start >> OCFS2_SB(inode-> 882 zero_start >> OCFS2_SB(inode->i_sb)->s_clustersize_bits;
867 u32 last_cpos = ocfs2_clusters_for_byt 883 u32 last_cpos = ocfs2_clusters_for_bytes(inode->i_sb, zero_end);
868 unsigned int num_clusters = 0; 884 unsigned int num_clusters = 0;
869 unsigned int ext_flags = 0; 885 unsigned int ext_flags = 0;
870 886
871 while (zero_cpos < last_cpos) { 887 while (zero_cpos < last_cpos) {
872 rc = ocfs2_get_clusters(inode, 888 rc = ocfs2_get_clusters(inode, zero_cpos, &p_cpos,
873 &num_c 889 &num_clusters, &ext_flags);
874 if (rc) { 890 if (rc) {
875 mlog_errno(rc); 891 mlog_errno(rc);
876 goto out; 892 goto out;
877 } 893 }
878 894
879 if (p_cpos && !(ext_flags & OC 895 if (p_cpos && !(ext_flags & OCFS2_EXT_UNWRITTEN)) {
880 zero_clusters = num_cl 896 zero_clusters = num_clusters;
881 if (ext_flags & OCFS2_ 897 if (ext_flags & OCFS2_EXT_REFCOUNTED)
882 needs_cow = 1; 898 needs_cow = 1;
883 break; 899 break;
884 } 900 }
885 901
886 zero_cpos += num_clusters; 902 zero_cpos += num_clusters;
887 } 903 }
888 if (!zero_clusters) { 904 if (!zero_clusters) {
889 *range_end = 0; 905 *range_end = 0;
890 goto out; 906 goto out;
891 } 907 }
892 908
893 while ((zero_cpos + zero_clusters) < l 909 while ((zero_cpos + zero_clusters) < last_cpos) {
894 rc = ocfs2_get_clusters(inode, 910 rc = ocfs2_get_clusters(inode, zero_cpos + zero_clusters,
895 &p_cpo 911 &p_cpos, &num_clusters,
896 &ext_f 912 &ext_flags);
897 if (rc) { 913 if (rc) {
898 mlog_errno(rc); 914 mlog_errno(rc);
899 goto out; 915 goto out;
900 } 916 }
901 917
902 if (!p_cpos || (ext_flags & OC 918 if (!p_cpos || (ext_flags & OCFS2_EXT_UNWRITTEN))
903 break; 919 break;
904 if (ext_flags & OCFS2_EXT_REFC 920 if (ext_flags & OCFS2_EXT_REFCOUNTED)
905 needs_cow = 1; 921 needs_cow = 1;
906 zero_clusters += num_clusters; 922 zero_clusters += num_clusters;
907 } 923 }
908 if ((zero_cpos + zero_clusters) > last 924 if ((zero_cpos + zero_clusters) > last_cpos)
909 zero_clusters = last_cpos - ze 925 zero_clusters = last_cpos - zero_cpos;
910 926
911 if (needs_cow) { 927 if (needs_cow) {
912 rc = ocfs2_refcount_cow(inode, 928 rc = ocfs2_refcount_cow(inode, di_bh, zero_cpos,
913 zero_c 929 zero_clusters, UINT_MAX);
914 if (rc) { 930 if (rc) {
915 mlog_errno(rc); 931 mlog_errno(rc);
916 goto out; 932 goto out;
917 } 933 }
918 } 934 }
919 935
920 *range_start = ocfs2_clusters_to_bytes 936 *range_start = ocfs2_clusters_to_bytes(inode->i_sb, zero_cpos);
921 *range_end = ocfs2_clusters_to_bytes(i 937 *range_end = ocfs2_clusters_to_bytes(inode->i_sb,
922 z 938 zero_cpos + zero_clusters);
923 939
924 out: 940 out:
925 return rc; 941 return rc;
926 } 942 }
927 943
928 /* 944 /*
929 * Zero one range returned from ocfs2_zero_ext 945 * Zero one range returned from ocfs2_zero_extend_get_range(). The caller
930 * has made sure that the entire range needs z 946 * has made sure that the entire range needs zeroing.
931 */ 947 */
932 static int ocfs2_zero_extend_range(struct inod 948 static int ocfs2_zero_extend_range(struct inode *inode, u64 range_start,
933 u64 range_e 949 u64 range_end, struct buffer_head *di_bh)
934 { 950 {
935 int rc = 0; 951 int rc = 0;
936 u64 next_pos; 952 u64 next_pos;
937 u64 zero_pos = range_start; 953 u64 zero_pos = range_start;
938 954
939 trace_ocfs2_zero_extend_range( 955 trace_ocfs2_zero_extend_range(
940 (unsigned long long)OC 956 (unsigned long long)OCFS2_I(inode)->ip_blkno,
941 (unsigned long long)ra 957 (unsigned long long)range_start,
942 (unsigned long long)ra 958 (unsigned long long)range_end);
943 BUG_ON(range_start >= range_end); 959 BUG_ON(range_start >= range_end);
944 960
945 while (zero_pos < range_end) { 961 while (zero_pos < range_end) {
946 next_pos = (zero_pos & PAGE_MA 962 next_pos = (zero_pos & PAGE_MASK) + PAGE_SIZE;
947 if (next_pos > range_end) 963 if (next_pos > range_end)
948 next_pos = range_end; 964 next_pos = range_end;
949 rc = ocfs2_write_zero_page(ino 965 rc = ocfs2_write_zero_page(inode, zero_pos, next_pos, di_bh);
950 if (rc < 0) { 966 if (rc < 0) {
951 mlog_errno(rc); 967 mlog_errno(rc);
952 break; 968 break;
953 } 969 }
954 zero_pos = next_pos; 970 zero_pos = next_pos;
955 971
956 /* 972 /*
957 * Very large extends have the 973 * Very large extends have the potential to lock up
958 * the cpu for extended period 974 * the cpu for extended periods of time.
959 */ 975 */
960 cond_resched(); 976 cond_resched();
961 } 977 }
962 978
963 return rc; 979 return rc;
964 } 980 }
965 981
966 int ocfs2_zero_extend(struct inode *inode, str 982 int ocfs2_zero_extend(struct inode *inode, struct buffer_head *di_bh,
967 loff_t zero_to_size) 983 loff_t zero_to_size)
968 { 984 {
969 int ret = 0; 985 int ret = 0;
970 u64 zero_start, range_start = 0, range 986 u64 zero_start, range_start = 0, range_end = 0;
971 struct super_block *sb = inode->i_sb; 987 struct super_block *sb = inode->i_sb;
972 988
973 zero_start = ocfs2_align_bytes_to_bloc 989 zero_start = ocfs2_align_bytes_to_blocks(sb, i_size_read(inode));
974 trace_ocfs2_zero_extend((unsigned long 990 trace_ocfs2_zero_extend((unsigned long long)OCFS2_I(inode)->ip_blkno,
975 (unsigned long 991 (unsigned long long)zero_start,
976 (unsigned long 992 (unsigned long long)i_size_read(inode));
977 while (zero_start < zero_to_size) { 993 while (zero_start < zero_to_size) {
978 ret = ocfs2_zero_extend_get_ra 994 ret = ocfs2_zero_extend_get_range(inode, di_bh, zero_start,
979 995 zero_to_size,
980 996 &range_start,
981 997 &range_end);
982 if (ret) { 998 if (ret) {
983 mlog_errno(ret); 999 mlog_errno(ret);
984 break; 1000 break;
985 } 1001 }
986 if (!range_end) 1002 if (!range_end)
987 break; 1003 break;
988 /* Trim the ends */ 1004 /* Trim the ends */
989 if (range_start < zero_start) 1005 if (range_start < zero_start)
990 range_start = zero_sta 1006 range_start = zero_start;
991 if (range_end > zero_to_size) 1007 if (range_end > zero_to_size)
992 range_end = zero_to_si 1008 range_end = zero_to_size;
993 1009
994 ret = ocfs2_zero_extend_range( 1010 ret = ocfs2_zero_extend_range(inode, range_start,
995 1011 range_end, di_bh);
996 if (ret) { 1012 if (ret) {
997 mlog_errno(ret); 1013 mlog_errno(ret);
998 break; 1014 break;
999 } 1015 }
1000 zero_start = range_end; 1016 zero_start = range_end;
1001 } 1017 }
1002 1018
1003 return ret; 1019 return ret;
1004 } 1020 }
1005 1021
1006 int ocfs2_extend_no_holes(struct inode *inode 1022 int ocfs2_extend_no_holes(struct inode *inode, struct buffer_head *di_bh,
1007 u64 new_i_size, u64 1023 u64 new_i_size, u64 zero_to)
1008 { 1024 {
1009 int ret; 1025 int ret;
1010 u32 clusters_to_add; 1026 u32 clusters_to_add;
1011 struct ocfs2_inode_info *oi = OCFS2_I 1027 struct ocfs2_inode_info *oi = OCFS2_I(inode);
1012 1028
1013 /* 1029 /*
1014 * Only quota files call this without 1030 * Only quota files call this without a bh, and they can't be
1015 * refcounted. 1031 * refcounted.
1016 */ 1032 */
1017 BUG_ON(!di_bh && ocfs2_is_refcount_in 1033 BUG_ON(!di_bh && ocfs2_is_refcount_inode(inode));
1018 BUG_ON(!di_bh && !(oi->ip_flags & OCF 1034 BUG_ON(!di_bh && !(oi->ip_flags & OCFS2_INODE_SYSTEM_FILE));
1019 1035
1020 clusters_to_add = ocfs2_clusters_for_ 1036 clusters_to_add = ocfs2_clusters_for_bytes(inode->i_sb, new_i_size);
1021 if (clusters_to_add < oi->ip_clusters 1037 if (clusters_to_add < oi->ip_clusters)
1022 clusters_to_add = 0; 1038 clusters_to_add = 0;
1023 else 1039 else
1024 clusters_to_add -= oi->ip_clu 1040 clusters_to_add -= oi->ip_clusters;
1025 1041
1026 if (clusters_to_add) { 1042 if (clusters_to_add) {
1027 ret = ocfs2_extend_allocation !! 1043 ret = __ocfs2_extend_allocation(inode, oi->ip_clusters,
1028 !! 1044 clusters_to_add, 0);
1029 if (ret) { 1045 if (ret) {
1030 mlog_errno(ret); 1046 mlog_errno(ret);
1031 goto out; 1047 goto out;
1032 } 1048 }
1033 } 1049 }
1034 1050
1035 /* 1051 /*
1036 * Call this even if we don't add any 1052 * Call this even if we don't add any clusters to the tree. We
1037 * still need to zero the area betwee 1053 * still need to zero the area between the old i_size and the
1038 * new i_size. 1054 * new i_size.
1039 */ 1055 */
1040 ret = ocfs2_zero_extend(inode, di_bh, 1056 ret = ocfs2_zero_extend(inode, di_bh, zero_to);
1041 if (ret < 0) 1057 if (ret < 0)
1042 mlog_errno(ret); 1058 mlog_errno(ret);
1043 1059
1044 out: 1060 out:
1045 return ret; 1061 return ret;
1046 } 1062 }
1047 1063
1048 static int ocfs2_extend_file(struct inode *in 1064 static int ocfs2_extend_file(struct inode *inode,
1049 struct buffer_he 1065 struct buffer_head *di_bh,
1050 u64 new_i_size) 1066 u64 new_i_size)
1051 { 1067 {
1052 int ret = 0; 1068 int ret = 0;
1053 struct ocfs2_inode_info *oi = OCFS2_I 1069 struct ocfs2_inode_info *oi = OCFS2_I(inode);
1054 1070
1055 BUG_ON(!di_bh); 1071 BUG_ON(!di_bh);
1056 1072
1057 /* setattr sometimes calls us like th 1073 /* setattr sometimes calls us like this. */
1058 if (new_i_size == 0) 1074 if (new_i_size == 0)
1059 goto out; 1075 goto out;
1060 1076
1061 if (i_size_read(inode) == new_i_size) 1077 if (i_size_read(inode) == new_i_size)
1062 goto out; 1078 goto out;
1063 BUG_ON(new_i_size < i_size_read(inode 1079 BUG_ON(new_i_size < i_size_read(inode));
1064 1080
1065 /* 1081 /*
1066 * The alloc sem blocks people in rea 1082 * The alloc sem blocks people in read/write from reading our
1067 * allocation until we're done changi 1083 * allocation until we're done changing it. We depend on
1068 * i_rwsem to block other extend/trun !! 1084 * i_mutex to block other extend/truncate calls while we're
1069 * here. We even have to hold it for 1085 * here. We even have to hold it for sparse files because there
1070 * might be some tail zeroing. 1086 * might be some tail zeroing.
1071 */ 1087 */
1072 down_write(&oi->ip_alloc_sem); 1088 down_write(&oi->ip_alloc_sem);
1073 1089
1074 if (oi->ip_dyn_features & OCFS2_INLIN 1090 if (oi->ip_dyn_features & OCFS2_INLINE_DATA_FL) {
1075 /* 1091 /*
1076 * We can optimize small exte 1092 * We can optimize small extends by keeping the inodes
1077 * inline data. 1093 * inline data.
1078 */ 1094 */
1079 if (ocfs2_size_fits_inline_da 1095 if (ocfs2_size_fits_inline_data(di_bh, new_i_size)) {
1080 up_write(&oi->ip_allo 1096 up_write(&oi->ip_alloc_sem);
1081 goto out_update_size; 1097 goto out_update_size;
1082 } 1098 }
1083 1099
1084 ret = ocfs2_convert_inline_da 1100 ret = ocfs2_convert_inline_data_to_extents(inode, di_bh);
1085 if (ret) { 1101 if (ret) {
1086 up_write(&oi->ip_allo 1102 up_write(&oi->ip_alloc_sem);
1087 mlog_errno(ret); 1103 mlog_errno(ret);
1088 goto out; 1104 goto out;
1089 } 1105 }
1090 } 1106 }
1091 1107
1092 if (ocfs2_sparse_alloc(OCFS2_SB(inode 1108 if (ocfs2_sparse_alloc(OCFS2_SB(inode->i_sb)))
1093 ret = ocfs2_zero_extend(inode 1109 ret = ocfs2_zero_extend(inode, di_bh, new_i_size);
1094 else 1110 else
1095 ret = ocfs2_extend_no_holes(i 1111 ret = ocfs2_extend_no_holes(inode, di_bh, new_i_size,
1096 n 1112 new_i_size);
1097 1113
1098 up_write(&oi->ip_alloc_sem); 1114 up_write(&oi->ip_alloc_sem);
1099 1115
1100 if (ret < 0) { 1116 if (ret < 0) {
1101 mlog_errno(ret); 1117 mlog_errno(ret);
1102 goto out; 1118 goto out;
1103 } 1119 }
1104 1120
1105 out_update_size: 1121 out_update_size:
1106 ret = ocfs2_simple_size_update(inode, 1122 ret = ocfs2_simple_size_update(inode, di_bh, new_i_size);
1107 if (ret < 0) 1123 if (ret < 0)
1108 mlog_errno(ret); 1124 mlog_errno(ret);
1109 1125
1110 out: 1126 out:
1111 return ret; 1127 return ret;
1112 } 1128 }
1113 1129
1114 int ocfs2_setattr(struct mnt_idmap *idmap, st !! 1130 int ocfs2_setattr(struct dentry *dentry, struct iattr *attr)
1115 struct iattr *attr) <<
1116 { 1131 {
1117 int status = 0, size_change; 1132 int status = 0, size_change;
1118 int inode_locked = 0; 1133 int inode_locked = 0;
1119 struct inode *inode = d_inode(dentry) 1134 struct inode *inode = d_inode(dentry);
1120 struct super_block *sb = inode->i_sb; 1135 struct super_block *sb = inode->i_sb;
1121 struct ocfs2_super *osb = OCFS2_SB(sb 1136 struct ocfs2_super *osb = OCFS2_SB(sb);
1122 struct buffer_head *bh = NULL; 1137 struct buffer_head *bh = NULL;
1123 handle_t *handle = NULL; 1138 handle_t *handle = NULL;
1124 struct dquot *transfer_to[MAXQUOTAS] 1139 struct dquot *transfer_to[MAXQUOTAS] = { };
1125 int qtype; 1140 int qtype;
1126 int had_lock; 1141 int had_lock;
1127 struct ocfs2_lock_holder oh; 1142 struct ocfs2_lock_holder oh;
1128 1143
1129 trace_ocfs2_setattr(inode, dentry, 1144 trace_ocfs2_setattr(inode, dentry,
1130 (unsigned long lo 1145 (unsigned long long)OCFS2_I(inode)->ip_blkno,
1131 dentry->d_name.le 1146 dentry->d_name.len, dentry->d_name.name,
1132 attr->ia_valid, !! 1147 attr->ia_valid, attr->ia_mode,
1133 attr->ia_vali !! 1148 from_kuid(&init_user_ns, attr->ia_uid),
1134 attr->ia_vali !! 1149 from_kgid(&init_user_ns, attr->ia_gid));
1135 from_ <<
1136 attr->ia_vali <<
1137 from_ <<
1138 1150
1139 /* ensuring we don't even attempt to 1151 /* ensuring we don't even attempt to truncate a symlink */
1140 if (S_ISLNK(inode->i_mode)) 1152 if (S_ISLNK(inode->i_mode))
1141 attr->ia_valid &= ~ATTR_SIZE; 1153 attr->ia_valid &= ~ATTR_SIZE;
1142 1154
1143 #define OCFS2_VALID_ATTRS (ATTR_ATIME | ATTR_ 1155 #define OCFS2_VALID_ATTRS (ATTR_ATIME | ATTR_MTIME | ATTR_CTIME | ATTR_SIZE \
1144 | ATTR_GID | ATTR_ 1156 | ATTR_GID | ATTR_UID | ATTR_MODE)
1145 if (!(attr->ia_valid & OCFS2_VALID_AT 1157 if (!(attr->ia_valid & OCFS2_VALID_ATTRS))
1146 return 0; 1158 return 0;
1147 1159
1148 status = setattr_prepare(&nop_mnt_idm !! 1160 status = setattr_prepare(dentry, attr);
1149 if (status) 1161 if (status)
1150 return status; 1162 return status;
1151 1163
1152 if (is_quota_modification(&nop_mnt_id !! 1164 if (is_quota_modification(inode, attr)) {
1153 status = dquot_initialize(ino 1165 status = dquot_initialize(inode);
1154 if (status) 1166 if (status)
1155 return status; 1167 return status;
1156 } 1168 }
1157 size_change = S_ISREG(inode->i_mode) 1169 size_change = S_ISREG(inode->i_mode) && attr->ia_valid & ATTR_SIZE;
1158 if (size_change) { 1170 if (size_change) {
1159 /* 1171 /*
1160 * Here we should wait dio to 1172 * Here we should wait dio to finish before inode lock
1161 * to avoid a deadlock betwee 1173 * to avoid a deadlock between ocfs2_setattr() and
1162 * ocfs2_dio_end_io_write() 1174 * ocfs2_dio_end_io_write()
1163 */ 1175 */
1164 inode_dio_wait(inode); 1176 inode_dio_wait(inode);
1165 1177
1166 status = ocfs2_rw_lock(inode, 1178 status = ocfs2_rw_lock(inode, 1);
1167 if (status < 0) { 1179 if (status < 0) {
1168 mlog_errno(status); 1180 mlog_errno(status);
1169 goto bail; 1181 goto bail;
1170 } 1182 }
1171 } 1183 }
1172 1184
1173 had_lock = ocfs2_inode_lock_tracker(i 1185 had_lock = ocfs2_inode_lock_tracker(inode, &bh, 1, &oh);
1174 if (had_lock < 0) { 1186 if (had_lock < 0) {
1175 status = had_lock; 1187 status = had_lock;
1176 goto bail_unlock_rw; 1188 goto bail_unlock_rw;
1177 } else if (had_lock) { 1189 } else if (had_lock) {
1178 /* 1190 /*
1179 * As far as we know, ocfs2_s 1191 * As far as we know, ocfs2_setattr() could only be the first
1180 * VFS entry point in the cal 1192 * VFS entry point in the call chain of recursive cluster
1181 * locking issue. 1193 * locking issue.
1182 * 1194 *
1183 * For instance: 1195 * For instance:
1184 * chmod_common() 1196 * chmod_common()
1185 * notify_change() 1197 * notify_change()
1186 * ocfs2_setattr() 1198 * ocfs2_setattr()
1187 * posix_acl_chmod() 1199 * posix_acl_chmod()
1188 * ocfs2_iop_get_acl() 1200 * ocfs2_iop_get_acl()
1189 * 1201 *
1190 * But, we're not 100% sure i 1202 * But, we're not 100% sure if it's always true, because the
1191 * ordering of the VFS entry 1203 * ordering of the VFS entry points in the call chain is out
1192 * of our control. So, we'd b 1204 * of our control. So, we'd better dump the stack here to
1193 * catch the other cases of r 1205 * catch the other cases of recursive locking.
1194 */ 1206 */
1195 mlog(ML_ERROR, "Another case 1207 mlog(ML_ERROR, "Another case of recursive locking:\n");
1196 dump_stack(); 1208 dump_stack();
1197 } 1209 }
1198 inode_locked = 1; 1210 inode_locked = 1;
1199 1211
1200 if (size_change) { 1212 if (size_change) {
1201 status = inode_newsize_ok(ino 1213 status = inode_newsize_ok(inode, attr->ia_size);
1202 if (status) 1214 if (status)
1203 goto bail_unlock; 1215 goto bail_unlock;
1204 1216
1205 if (i_size_read(inode) >= att 1217 if (i_size_read(inode) >= attr->ia_size) {
1206 if (ocfs2_should_orde 1218 if (ocfs2_should_order_data(inode)) {
1207 status = ocfs 1219 status = ocfs2_begin_ordered_truncate(inode,
1208 1220 attr->ia_size);
1209 if (status) 1221 if (status)
1210 goto 1222 goto bail_unlock;
1211 } 1223 }
1212 status = ocfs2_trunca 1224 status = ocfs2_truncate_file(inode, bh, attr->ia_size);
1213 } else 1225 } else
1214 status = ocfs2_extend 1226 status = ocfs2_extend_file(inode, bh, attr->ia_size);
1215 if (status < 0) { 1227 if (status < 0) {
1216 if (status != -ENOSPC 1228 if (status != -ENOSPC)
1217 mlog_errno(st 1229 mlog_errno(status);
1218 status = -ENOSPC; 1230 status = -ENOSPC;
1219 goto bail_unlock; 1231 goto bail_unlock;
1220 } 1232 }
1221 } 1233 }
1222 1234
1223 if ((attr->ia_valid & ATTR_UID && !ui 1235 if ((attr->ia_valid & ATTR_UID && !uid_eq(attr->ia_uid, inode->i_uid)) ||
1224 (attr->ia_valid & ATTR_GID && !gi 1236 (attr->ia_valid & ATTR_GID && !gid_eq(attr->ia_gid, inode->i_gid))) {
1225 /* 1237 /*
1226 * Gather pointers to quota s 1238 * Gather pointers to quota structures so that allocation /
1227 * freeing of quota structure 1239 * freeing of quota structures happens here and not inside
1228 * dquot_transfer() where we 1240 * dquot_transfer() where we have problems with lock ordering
1229 */ 1241 */
1230 if (attr->ia_valid & ATTR_UID 1242 if (attr->ia_valid & ATTR_UID && !uid_eq(attr->ia_uid, inode->i_uid)
1231 && OCFS2_HAS_RO_COMPAT_FE 1243 && OCFS2_HAS_RO_COMPAT_FEATURE(sb,
1232 OCFS2_FEATURE_RO_COMPAT_U 1244 OCFS2_FEATURE_RO_COMPAT_USRQUOTA)) {
1233 transfer_to[USRQUOTA] 1245 transfer_to[USRQUOTA] = dqget(sb, make_kqid_uid(attr->ia_uid));
1234 if (IS_ERR(transfer_t 1246 if (IS_ERR(transfer_to[USRQUOTA])) {
1235 status = PTR_ 1247 status = PTR_ERR(transfer_to[USRQUOTA]);
1236 transfer_to[U <<
1237 goto bail_unl 1248 goto bail_unlock;
1238 } 1249 }
1239 } 1250 }
1240 if (attr->ia_valid & ATTR_GID 1251 if (attr->ia_valid & ATTR_GID && !gid_eq(attr->ia_gid, inode->i_gid)
1241 && OCFS2_HAS_RO_COMPAT_FE 1252 && OCFS2_HAS_RO_COMPAT_FEATURE(sb,
1242 OCFS2_FEATURE_RO_COMPAT_G 1253 OCFS2_FEATURE_RO_COMPAT_GRPQUOTA)) {
1243 transfer_to[GRPQUOTA] 1254 transfer_to[GRPQUOTA] = dqget(sb, make_kqid_gid(attr->ia_gid));
1244 if (IS_ERR(transfer_t 1255 if (IS_ERR(transfer_to[GRPQUOTA])) {
1245 status = PTR_ 1256 status = PTR_ERR(transfer_to[GRPQUOTA]);
1246 transfer_to[G <<
1247 goto bail_unl 1257 goto bail_unlock;
1248 } 1258 }
1249 } 1259 }
1250 down_write(&OCFS2_I(inode)->i <<
1251 handle = ocfs2_start_trans(os 1260 handle = ocfs2_start_trans(osb, OCFS2_INODE_UPDATE_CREDITS +
1252 2 1261 2 * ocfs2_quota_trans_credits(sb));
1253 if (IS_ERR(handle)) { 1262 if (IS_ERR(handle)) {
1254 status = PTR_ERR(hand 1263 status = PTR_ERR(handle);
1255 mlog_errno(status); 1264 mlog_errno(status);
1256 goto bail_unlock_allo !! 1265 goto bail_unlock;
1257 } 1266 }
1258 status = __dquot_transfer(ino 1267 status = __dquot_transfer(inode, transfer_to);
1259 if (status < 0) 1268 if (status < 0)
1260 goto bail_commit; 1269 goto bail_commit;
1261 } else { 1270 } else {
1262 down_write(&OCFS2_I(inode)->i <<
1263 handle = ocfs2_start_trans(os 1271 handle = ocfs2_start_trans(osb, OCFS2_INODE_UPDATE_CREDITS);
1264 if (IS_ERR(handle)) { 1272 if (IS_ERR(handle)) {
1265 status = PTR_ERR(hand 1273 status = PTR_ERR(handle);
1266 mlog_errno(status); 1274 mlog_errno(status);
1267 goto bail_unlock_allo !! 1275 goto bail_unlock;
1268 } 1276 }
1269 } 1277 }
1270 1278
1271 setattr_copy(&nop_mnt_idmap, inode, a !! 1279 setattr_copy(inode, attr);
1272 mark_inode_dirty(inode); 1280 mark_inode_dirty(inode);
1273 1281
1274 status = ocfs2_mark_inode_dirty(handl 1282 status = ocfs2_mark_inode_dirty(handle, inode, bh);
1275 if (status < 0) 1283 if (status < 0)
1276 mlog_errno(status); 1284 mlog_errno(status);
1277 1285
1278 bail_commit: 1286 bail_commit:
1279 ocfs2_commit_trans(osb, handle); 1287 ocfs2_commit_trans(osb, handle);
1280 bail_unlock_alloc: <<
1281 up_write(&OCFS2_I(inode)->ip_alloc_se <<
1282 bail_unlock: 1288 bail_unlock:
1283 if (status && inode_locked) { 1289 if (status && inode_locked) {
1284 ocfs2_inode_unlock_tracker(in 1290 ocfs2_inode_unlock_tracker(inode, 1, &oh, had_lock);
1285 inode_locked = 0; 1291 inode_locked = 0;
1286 } 1292 }
1287 bail_unlock_rw: 1293 bail_unlock_rw:
1288 if (size_change) 1294 if (size_change)
1289 ocfs2_rw_unlock(inode, 1); 1295 ocfs2_rw_unlock(inode, 1);
1290 bail: 1296 bail:
1291 1297
1292 /* Release quota pointers in case we 1298 /* Release quota pointers in case we acquired them */
1293 for (qtype = 0; qtype < OCFS2_MAXQUOT 1299 for (qtype = 0; qtype < OCFS2_MAXQUOTAS; qtype++)
1294 dqput(transfer_to[qtype]); 1300 dqput(transfer_to[qtype]);
1295 1301
1296 if (!status && attr->ia_valid & ATTR_ 1302 if (!status && attr->ia_valid & ATTR_MODE) {
1297 status = ocfs2_acl_chmod(inod 1303 status = ocfs2_acl_chmod(inode, bh);
1298 if (status < 0) 1304 if (status < 0)
1299 mlog_errno(status); 1305 mlog_errno(status);
1300 } 1306 }
1301 if (inode_locked) 1307 if (inode_locked)
1302 ocfs2_inode_unlock_tracker(in 1308 ocfs2_inode_unlock_tracker(inode, 1, &oh, had_lock);
1303 1309
1304 brelse(bh); 1310 brelse(bh);
1305 return status; 1311 return status;
1306 } 1312 }
1307 1313
1308 int ocfs2_getattr(struct mnt_idmap *idmap, co !! 1314 int ocfs2_getattr(const struct path *path, struct kstat *stat,
1309 struct kstat *stat, u32 req !! 1315 u32 request_mask, unsigned int flags)
1310 { 1316 {
1311 struct inode *inode = d_inode(path->d 1317 struct inode *inode = d_inode(path->dentry);
1312 struct super_block *sb = path->dentry 1318 struct super_block *sb = path->dentry->d_sb;
1313 struct ocfs2_super *osb = sb->s_fs_in 1319 struct ocfs2_super *osb = sb->s_fs_info;
1314 int err; 1320 int err;
1315 1321
1316 err = ocfs2_inode_revalidate(path->de 1322 err = ocfs2_inode_revalidate(path->dentry);
1317 if (err) { 1323 if (err) {
1318 if (err != -ENOENT) 1324 if (err != -ENOENT)
1319 mlog_errno(err); 1325 mlog_errno(err);
1320 goto bail; 1326 goto bail;
1321 } 1327 }
1322 1328
1323 generic_fillattr(&nop_mnt_idmap, requ !! 1329 generic_fillattr(inode, stat);
1324 /* 1330 /*
1325 * If there is inline data in the ino 1331 * If there is inline data in the inode, the inode will normally not
1326 * have data blocks allocated (it may 1332 * have data blocks allocated (it may have an external xattr block).
1327 * Report at least one sector for suc 1333 * Report at least one sector for such files, so tools like tar, rsync,
1328 * others don't incorrectly think the 1334 * others don't incorrectly think the file is completely sparse.
1329 */ 1335 */
1330 if (unlikely(OCFS2_I(inode)->ip_dyn_f 1336 if (unlikely(OCFS2_I(inode)->ip_dyn_features & OCFS2_INLINE_DATA_FL))
1331 stat->blocks += (stat->size + 1337 stat->blocks += (stat->size + 511)>>9;
1332 1338
1333 /* We set the blksize from the cluste 1339 /* We set the blksize from the cluster size for performance */
1334 stat->blksize = osb->s_clustersize; 1340 stat->blksize = osb->s_clustersize;
1335 1341
1336 bail: 1342 bail:
1337 return err; 1343 return err;
1338 } 1344 }
1339 1345
1340 int ocfs2_permission(struct mnt_idmap *idmap, !! 1346 int ocfs2_permission(struct inode *inode, int mask)
1341 int mask) <<
1342 { 1347 {
1343 int ret, had_lock; 1348 int ret, had_lock;
1344 struct ocfs2_lock_holder oh; 1349 struct ocfs2_lock_holder oh;
1345 1350
1346 if (mask & MAY_NOT_BLOCK) 1351 if (mask & MAY_NOT_BLOCK)
1347 return -ECHILD; 1352 return -ECHILD;
1348 1353
1349 had_lock = ocfs2_inode_lock_tracker(i 1354 had_lock = ocfs2_inode_lock_tracker(inode, NULL, 0, &oh);
1350 if (had_lock < 0) { 1355 if (had_lock < 0) {
1351 ret = had_lock; 1356 ret = had_lock;
1352 goto out; 1357 goto out;
1353 } else if (had_lock) { 1358 } else if (had_lock) {
1354 /* See comments in ocfs2_seta 1359 /* See comments in ocfs2_setattr() for details.
1355 * The call chain of this cas 1360 * The call chain of this case could be:
1356 * do_sys_open() 1361 * do_sys_open()
1357 * may_open() 1362 * may_open()
1358 * inode_permission() 1363 * inode_permission()
1359 * ocfs2_permission() 1364 * ocfs2_permission()
1360 * ocfs2_iop_get_acl() 1365 * ocfs2_iop_get_acl()
1361 */ 1366 */
1362 mlog(ML_ERROR, "Another case 1367 mlog(ML_ERROR, "Another case of recursive locking:\n");
1363 dump_stack(); 1368 dump_stack();
1364 } 1369 }
1365 1370
1366 ret = generic_permission(&nop_mnt_idm !! 1371 ret = generic_permission(inode, mask);
1367 1372
1368 ocfs2_inode_unlock_tracker(inode, 0, 1373 ocfs2_inode_unlock_tracker(inode, 0, &oh, had_lock);
1369 out: 1374 out:
1370 return ret; 1375 return ret;
1371 } 1376 }
1372 1377
1373 static int __ocfs2_write_remove_suid(struct i 1378 static int __ocfs2_write_remove_suid(struct inode *inode,
1374 struct b 1379 struct buffer_head *bh)
1375 { 1380 {
1376 int ret; 1381 int ret;
1377 handle_t *handle; 1382 handle_t *handle;
1378 struct ocfs2_super *osb = OCFS2_SB(in 1383 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
1379 struct ocfs2_dinode *di; 1384 struct ocfs2_dinode *di;
1380 1385
1381 trace_ocfs2_write_remove_suid( 1386 trace_ocfs2_write_remove_suid(
1382 (unsigned long long)O 1387 (unsigned long long)OCFS2_I(inode)->ip_blkno,
1383 inode->i_mode); 1388 inode->i_mode);
1384 1389
1385 handle = ocfs2_start_trans(osb, OCFS2 1390 handle = ocfs2_start_trans(osb, OCFS2_INODE_UPDATE_CREDITS);
1386 if (IS_ERR(handle)) { 1391 if (IS_ERR(handle)) {
1387 ret = PTR_ERR(handle); 1392 ret = PTR_ERR(handle);
1388 mlog_errno(ret); 1393 mlog_errno(ret);
1389 goto out; 1394 goto out;
1390 } 1395 }
1391 1396
1392 ret = ocfs2_journal_access_di(handle, 1397 ret = ocfs2_journal_access_di(handle, INODE_CACHE(inode), bh,
1393 OCFS2_J 1398 OCFS2_JOURNAL_ACCESS_WRITE);
1394 if (ret < 0) { 1399 if (ret < 0) {
1395 mlog_errno(ret); 1400 mlog_errno(ret);
1396 goto out_trans; 1401 goto out_trans;
1397 } 1402 }
1398 1403
1399 inode->i_mode &= ~S_ISUID; 1404 inode->i_mode &= ~S_ISUID;
1400 if ((inode->i_mode & S_ISGID) && (ino 1405 if ((inode->i_mode & S_ISGID) && (inode->i_mode & S_IXGRP))
1401 inode->i_mode &= ~S_ISGID; 1406 inode->i_mode &= ~S_ISGID;
1402 1407
1403 di = (struct ocfs2_dinode *) bh->b_da 1408 di = (struct ocfs2_dinode *) bh->b_data;
1404 di->i_mode = cpu_to_le16(inode->i_mod 1409 di->i_mode = cpu_to_le16(inode->i_mode);
1405 ocfs2_update_inode_fsync_trans(handle 1410 ocfs2_update_inode_fsync_trans(handle, inode, 0);
1406 1411
1407 ocfs2_journal_dirty(handle, bh); 1412 ocfs2_journal_dirty(handle, bh);
1408 1413
1409 out_trans: 1414 out_trans:
1410 ocfs2_commit_trans(osb, handle); 1415 ocfs2_commit_trans(osb, handle);
1411 out: 1416 out:
1412 return ret; 1417 return ret;
1413 } 1418 }
1414 1419
1415 static int ocfs2_write_remove_suid(struct ino 1420 static int ocfs2_write_remove_suid(struct inode *inode)
1416 { 1421 {
1417 int ret; 1422 int ret;
1418 struct buffer_head *bh = NULL; 1423 struct buffer_head *bh = NULL;
1419 1424
1420 ret = ocfs2_read_inode_block(inode, & 1425 ret = ocfs2_read_inode_block(inode, &bh);
1421 if (ret < 0) { 1426 if (ret < 0) {
1422 mlog_errno(ret); 1427 mlog_errno(ret);
1423 goto out; 1428 goto out;
1424 } 1429 }
1425 1430
1426 ret = __ocfs2_write_remove_suid(inod 1431 ret = __ocfs2_write_remove_suid(inode, bh);
1427 out: 1432 out:
1428 brelse(bh); 1433 brelse(bh);
1429 return ret; 1434 return ret;
1430 } 1435 }
1431 1436
1432 /* 1437 /*
1433 * Allocate enough extents to cover the regio 1438 * Allocate enough extents to cover the region starting at byte offset
1434 * start for len bytes. Existing extents are 1439 * start for len bytes. Existing extents are skipped, any extents
1435 * added are marked as "unwritten". 1440 * added are marked as "unwritten".
1436 */ 1441 */
1437 static int ocfs2_allocate_unwritten_extents(s 1442 static int ocfs2_allocate_unwritten_extents(struct inode *inode,
1438 u 1443 u64 start, u64 len)
1439 { 1444 {
1440 int ret; 1445 int ret;
1441 u32 cpos, phys_cpos, clusters, alloc_ 1446 u32 cpos, phys_cpos, clusters, alloc_size;
1442 u64 end = start + len; 1447 u64 end = start + len;
1443 struct buffer_head *di_bh = NULL; 1448 struct buffer_head *di_bh = NULL;
1444 1449
1445 if (OCFS2_I(inode)->ip_dyn_features & 1450 if (OCFS2_I(inode)->ip_dyn_features & OCFS2_INLINE_DATA_FL) {
1446 ret = ocfs2_read_inode_block( 1451 ret = ocfs2_read_inode_block(inode, &di_bh);
1447 if (ret) { 1452 if (ret) {
1448 mlog_errno(ret); 1453 mlog_errno(ret);
1449 goto out; 1454 goto out;
1450 } 1455 }
1451 1456
1452 /* 1457 /*
1453 * Nothing to do if the reque 1458 * Nothing to do if the requested reservation range
1454 * fits within the inode. 1459 * fits within the inode.
1455 */ 1460 */
1456 if (ocfs2_size_fits_inline_da 1461 if (ocfs2_size_fits_inline_data(di_bh, end))
1457 goto out; 1462 goto out;
1458 1463
1459 ret = ocfs2_convert_inline_da 1464 ret = ocfs2_convert_inline_data_to_extents(inode, di_bh);
1460 if (ret) { 1465 if (ret) {
1461 mlog_errno(ret); 1466 mlog_errno(ret);
1462 goto out; 1467 goto out;
1463 } 1468 }
1464 } 1469 }
1465 1470
1466 /* 1471 /*
1467 * We consider both start and len to 1472 * We consider both start and len to be inclusive.
1468 */ 1473 */
1469 cpos = start >> OCFS2_SB(inode->i_sb) 1474 cpos = start >> OCFS2_SB(inode->i_sb)->s_clustersize_bits;
1470 clusters = ocfs2_clusters_for_bytes(i 1475 clusters = ocfs2_clusters_for_bytes(inode->i_sb, start + len);
1471 clusters -= cpos; 1476 clusters -= cpos;
1472 1477
1473 while (clusters) { 1478 while (clusters) {
1474 ret = ocfs2_get_clusters(inod 1479 ret = ocfs2_get_clusters(inode, cpos, &phys_cpos,
1475 &all 1480 &alloc_size, NULL);
1476 if (ret) { 1481 if (ret) {
1477 mlog_errno(ret); 1482 mlog_errno(ret);
1478 goto out; 1483 goto out;
1479 } 1484 }
1480 1485
1481 /* 1486 /*
1482 * Hole or existing extent le 1487 * Hole or existing extent len can be arbitrary, so
1483 * cap it to our own allocati 1488 * cap it to our own allocation request.
1484 */ 1489 */
1485 if (alloc_size > clusters) 1490 if (alloc_size > clusters)
1486 alloc_size = clusters 1491 alloc_size = clusters;
1487 1492
1488 if (phys_cpos) { 1493 if (phys_cpos) {
1489 /* 1494 /*
1490 * We already have an 1495 * We already have an allocation at this
1491 * region so we can s 1496 * region so we can safely skip it.
1492 */ 1497 */
1493 goto next; 1498 goto next;
1494 } 1499 }
1495 1500
1496 ret = ocfs2_extend_allocation !! 1501 ret = __ocfs2_extend_allocation(inode, cpos, alloc_size, 1);
1497 if (ret) { 1502 if (ret) {
1498 if (ret != -ENOSPC) 1503 if (ret != -ENOSPC)
1499 mlog_errno(re 1504 mlog_errno(ret);
1500 goto out; 1505 goto out;
1501 } 1506 }
1502 1507
1503 next: 1508 next:
1504 cpos += alloc_size; 1509 cpos += alloc_size;
1505 clusters -= alloc_size; 1510 clusters -= alloc_size;
1506 } 1511 }
1507 1512
1508 ret = 0; 1513 ret = 0;
1509 out: 1514 out:
1510 1515
1511 brelse(di_bh); 1516 brelse(di_bh);
1512 return ret; 1517 return ret;
1513 } 1518 }
1514 1519
1515 /* 1520 /*
1516 * Truncate a byte range, avoiding pages with 1521 * Truncate a byte range, avoiding pages within partial clusters. This
1517 * preserves those pages for the zeroing code 1522 * preserves those pages for the zeroing code to write to.
1518 */ 1523 */
1519 static void ocfs2_truncate_cluster_pages(stru 1524 static void ocfs2_truncate_cluster_pages(struct inode *inode, u64 byte_start,
1520 u64 1525 u64 byte_len)
1521 { 1526 {
1522 struct ocfs2_super *osb = OCFS2_SB(in 1527 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
1523 loff_t start, end; 1528 loff_t start, end;
1524 struct address_space *mapping = inode 1529 struct address_space *mapping = inode->i_mapping;
1525 1530
1526 start = (loff_t)ocfs2_align_bytes_to_ 1531 start = (loff_t)ocfs2_align_bytes_to_clusters(inode->i_sb, byte_start);
1527 end = byte_start + byte_len; 1532 end = byte_start + byte_len;
1528 end = end & ~(osb->s_clustersize - 1) 1533 end = end & ~(osb->s_clustersize - 1);
1529 1534
1530 if (start < end) { 1535 if (start < end) {
1531 unmap_mapping_range(mapping, 1536 unmap_mapping_range(mapping, start, end - start, 0);
1532 truncate_inode_pages_range(ma 1537 truncate_inode_pages_range(mapping, start, end - 1);
1533 } 1538 }
1534 } 1539 }
1535 1540
1536 /* <<
1537 * zero out partial blocks of one cluster. <<
1538 * <<
1539 * start: file offset where zero starts, will <<
1540 * len: it will be trimmed to the end of curr <<
1541 * is bigger than it. <<
1542 */ <<
1543 static int ocfs2_zeroout_partial_cluster(stru <<
1544 u64 s <<
1545 { <<
1546 int ret; <<
1547 u64 start_block, end_block, nr_blocks <<
1548 u64 p_block, offset; <<
1549 u32 cluster, p_cluster, nr_clusters; <<
1550 struct super_block *sb = inode->i_sb; <<
1551 u64 end = ocfs2_align_bytes_to_cluste <<
1552 <<
1553 if (start + len < end) <<
1554 end = start + len; <<
1555 <<
1556 start_block = ocfs2_blocks_for_bytes( <<
1557 end_block = ocfs2_blocks_for_bytes(sb <<
1558 nr_blocks = end_block - start_block; <<
1559 if (!nr_blocks) <<
1560 return 0; <<
1561 <<
1562 cluster = ocfs2_bytes_to_clusters(sb, <<
1563 ret = ocfs2_get_clusters(inode, clust <<
1564 &nr_clusters, <<
1565 if (ret) <<
1566 return ret; <<
1567 if (!p_cluster) <<
1568 return 0; <<
1569 <<
1570 offset = start_block - ocfs2_clusters <<
1571 p_block = ocfs2_clusters_to_blocks(sb <<
1572 return sb_issue_zeroout(sb, p_block, <<
1573 } <<
1574 <<
1575 static int ocfs2_zero_partial_clusters(struct 1541 static int ocfs2_zero_partial_clusters(struct inode *inode,
1576 u64 st 1542 u64 start, u64 len)
1577 { 1543 {
1578 int ret = 0; 1544 int ret = 0;
1579 u64 tmpend = 0; 1545 u64 tmpend = 0;
1580 u64 end = start + len; 1546 u64 end = start + len;
1581 struct ocfs2_super *osb = OCFS2_SB(in 1547 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
1582 unsigned int csize = osb->s_clustersi 1548 unsigned int csize = osb->s_clustersize;
1583 handle_t *handle; 1549 handle_t *handle;
1584 loff_t isize = i_size_read(inode); <<
1585 1550
1586 /* 1551 /*
1587 * The "start" and "end" values are N 1552 * The "start" and "end" values are NOT necessarily part of
1588 * the range whose allocation is bein 1553 * the range whose allocation is being deleted. Rather, this
1589 * is what the user passed in with th 1554 * is what the user passed in with the request. We must zero
1590 * partial clusters here. There's no 1555 * partial clusters here. There's no need to worry about
1591 * physical allocation - the zeroing 1556 * physical allocation - the zeroing code knows to skip holes.
1592 */ 1557 */
1593 trace_ocfs2_zero_partial_clusters( 1558 trace_ocfs2_zero_partial_clusters(
1594 (unsigned long long)OCFS2_I(i 1559 (unsigned long long)OCFS2_I(inode)->ip_blkno,
1595 (unsigned long long)start, (u 1560 (unsigned long long)start, (unsigned long long)end);
1596 1561
1597 /* 1562 /*
1598 * If both edges are on a cluster bou 1563 * If both edges are on a cluster boundary then there's no
1599 * zeroing required as the region is 1564 * zeroing required as the region is part of the allocation to
1600 * be truncated. 1565 * be truncated.
1601 */ 1566 */
1602 if ((start & (csize - 1)) == 0 && (en 1567 if ((start & (csize - 1)) == 0 && (end & (csize - 1)) == 0)
1603 goto out; 1568 goto out;
1604 1569
1605 /* No page cache for EOF blocks, issu <<
1606 if (end > isize) { <<
1607 /* <<
1608 * zeroout eof blocks in last <<
1609 * "isize" even "start" > "is <<
1610 * complicated to zeroout jus <<
1611 * may be not aligned with bl <<
1612 * would be required to do th <<
1613 * write is not supported. <<
1614 */ <<
1615 ret = ocfs2_zeroout_partial_c <<
1616 end - <<
1617 if (ret) { <<
1618 mlog_errno(ret); <<
1619 goto out; <<
1620 } <<
1621 if (start >= isize) <<
1622 goto out; <<
1623 end = isize; <<
1624 } <<
1625 handle = ocfs2_start_trans(osb, OCFS2 1570 handle = ocfs2_start_trans(osb, OCFS2_INODE_UPDATE_CREDITS);
1626 if (IS_ERR(handle)) { 1571 if (IS_ERR(handle)) {
1627 ret = PTR_ERR(handle); 1572 ret = PTR_ERR(handle);
1628 mlog_errno(ret); 1573 mlog_errno(ret);
1629 goto out; 1574 goto out;
1630 } 1575 }
1631 1576
1632 /* 1577 /*
1633 * If start is on a cluster boundary 1578 * If start is on a cluster boundary and end is somewhere in another
1634 * cluster, we have not COWed the clu 1579 * cluster, we have not COWed the cluster starting at start, unless
1635 * end is also within the same cluste 1580 * end is also within the same cluster. So, in this case, we skip this
1636 * first call to ocfs2_zero_range_for 1581 * first call to ocfs2_zero_range_for_truncate() truncate and move on
1637 * to the next one. 1582 * to the next one.
1638 */ 1583 */
1639 if ((start & (csize - 1)) != 0) { 1584 if ((start & (csize - 1)) != 0) {
1640 /* 1585 /*
1641 * We want to get the byte of 1586 * We want to get the byte offset of the end of the 1st
1642 * cluster. 1587 * cluster.
1643 */ 1588 */
1644 tmpend = (u64)osb->s_clusters 1589 tmpend = (u64)osb->s_clustersize +
1645 (start & ~(osb->s_clu 1590 (start & ~(osb->s_clustersize - 1));
1646 if (tmpend > end) 1591 if (tmpend > end)
1647 tmpend = end; 1592 tmpend = end;
1648 1593
1649 trace_ocfs2_zero_partial_clus 1594 trace_ocfs2_zero_partial_clusters_range1(
1650 (unsigned long long)s 1595 (unsigned long long)start,
1651 (unsigned long long)t 1596 (unsigned long long)tmpend);
1652 1597
1653 ret = ocfs2_zero_range_for_tr 1598 ret = ocfs2_zero_range_for_truncate(inode, handle, start,
1654 1599 tmpend);
1655 if (ret) 1600 if (ret)
1656 mlog_errno(ret); 1601 mlog_errno(ret);
1657 } 1602 }
1658 1603
1659 if (tmpend < end) { 1604 if (tmpend < end) {
1660 /* 1605 /*
1661 * This may make start and en 1606 * This may make start and end equal, but the zeroing
1662 * code will skip any work in 1607 * code will skip any work in that case so there's no
1663 * need to catch it up here. 1608 * need to catch it up here.
1664 */ 1609 */
1665 start = end & ~(osb->s_cluste 1610 start = end & ~(osb->s_clustersize - 1);
1666 1611
1667 trace_ocfs2_zero_partial_clus 1612 trace_ocfs2_zero_partial_clusters_range2(
1668 (unsigned long long)s 1613 (unsigned long long)start, (unsigned long long)end);
1669 1614
1670 ret = ocfs2_zero_range_for_tr 1615 ret = ocfs2_zero_range_for_truncate(inode, handle, start, end);
1671 if (ret) 1616 if (ret)
1672 mlog_errno(ret); 1617 mlog_errno(ret);
1673 } 1618 }
1674 ocfs2_update_inode_fsync_trans(handle 1619 ocfs2_update_inode_fsync_trans(handle, inode, 1);
1675 1620
1676 ocfs2_commit_trans(osb, handle); 1621 ocfs2_commit_trans(osb, handle);
1677 out: 1622 out:
1678 return ret; 1623 return ret;
1679 } 1624 }
1680 1625
1681 static int ocfs2_find_rec(struct ocfs2_extent 1626 static int ocfs2_find_rec(struct ocfs2_extent_list *el, u32 pos)
1682 { 1627 {
1683 int i; 1628 int i;
1684 struct ocfs2_extent_rec *rec = NULL; 1629 struct ocfs2_extent_rec *rec = NULL;
1685 1630
1686 for (i = le16_to_cpu(el->l_next_free_ 1631 for (i = le16_to_cpu(el->l_next_free_rec) - 1; i >= 0; i--) {
1687 1632
1688 rec = &el->l_recs[i]; 1633 rec = &el->l_recs[i];
1689 1634
1690 if (le32_to_cpu(rec->e_cpos) 1635 if (le32_to_cpu(rec->e_cpos) < pos)
1691 break; 1636 break;
1692 } 1637 }
1693 1638
1694 return i; 1639 return i;
1695 } 1640 }
1696 1641
1697 /* 1642 /*
1698 * Helper to calculate the punching pos and l 1643 * Helper to calculate the punching pos and length in one run, we handle the
1699 * following three cases in order: 1644 * following three cases in order:
1700 * 1645 *
1701 * - remove the entire record 1646 * - remove the entire record
1702 * - remove a partial record 1647 * - remove a partial record
1703 * - no record needs to be removed (hole-punc 1648 * - no record needs to be removed (hole-punching completed)
1704 */ 1649 */
1705 static void ocfs2_calc_trunc_pos(struct inode 1650 static void ocfs2_calc_trunc_pos(struct inode *inode,
1706 struct ocfs2 1651 struct ocfs2_extent_list *el,
1707 struct ocfs2 1652 struct ocfs2_extent_rec *rec,
1708 u32 trunc_st 1653 u32 trunc_start, u32 *trunc_cpos,
1709 u32 *trunc_l 1654 u32 *trunc_len, u32 *trunc_end,
1710 u64 *blkno, 1655 u64 *blkno, int *done)
1711 { 1656 {
1712 int ret = 0; 1657 int ret = 0;
1713 u32 coff, range; 1658 u32 coff, range;
1714 1659
1715 range = le32_to_cpu(rec->e_cpos) + oc 1660 range = le32_to_cpu(rec->e_cpos) + ocfs2_rec_clusters(el, rec);
1716 1661
1717 if (le32_to_cpu(rec->e_cpos) >= trunc 1662 if (le32_to_cpu(rec->e_cpos) >= trunc_start) {
1718 /* 1663 /*
1719 * remove an entire extent re 1664 * remove an entire extent record.
1720 */ 1665 */
1721 *trunc_cpos = le32_to_cpu(rec 1666 *trunc_cpos = le32_to_cpu(rec->e_cpos);
1722 /* 1667 /*
1723 * Skip holes if any. 1668 * Skip holes if any.
1724 */ 1669 */
1725 if (range < *trunc_end) 1670 if (range < *trunc_end)
1726 *trunc_end = range; 1671 *trunc_end = range;
1727 *trunc_len = *trunc_end - le3 1672 *trunc_len = *trunc_end - le32_to_cpu(rec->e_cpos);
1728 *blkno = le64_to_cpu(rec->e_b 1673 *blkno = le64_to_cpu(rec->e_blkno);
1729 *trunc_end = le32_to_cpu(rec- 1674 *trunc_end = le32_to_cpu(rec->e_cpos);
1730 } else if (range > trunc_start) { 1675 } else if (range > trunc_start) {
1731 /* 1676 /*
1732 * remove a partial extent re 1677 * remove a partial extent record, which means we're
1733 * removing the last extent r 1678 * removing the last extent record.
1734 */ 1679 */
1735 *trunc_cpos = trunc_start; 1680 *trunc_cpos = trunc_start;
1736 /* 1681 /*
1737 * skip hole if any. 1682 * skip hole if any.
1738 */ 1683 */
1739 if (range < *trunc_end) 1684 if (range < *trunc_end)
1740 *trunc_end = range; 1685 *trunc_end = range;
1741 *trunc_len = *trunc_end - tru 1686 *trunc_len = *trunc_end - trunc_start;
1742 coff = trunc_start - le32_to_ 1687 coff = trunc_start - le32_to_cpu(rec->e_cpos);
1743 *blkno = le64_to_cpu(rec->e_b 1688 *blkno = le64_to_cpu(rec->e_blkno) +
1744 ocfs2_cluster 1689 ocfs2_clusters_to_blocks(inode->i_sb, coff);
1745 *trunc_end = trunc_start; 1690 *trunc_end = trunc_start;
1746 } else { 1691 } else {
1747 /* 1692 /*
1748 * It may have two following 1693 * It may have two following possibilities:
1749 * 1694 *
1750 * - last record has been rem 1695 * - last record has been removed
1751 * - trunc_start was within a 1696 * - trunc_start was within a hole
1752 * 1697 *
1753 * both two cases mean the co 1698 * both two cases mean the completion of hole punching.
1754 */ 1699 */
1755 ret = 1; 1700 ret = 1;
1756 } 1701 }
1757 1702
1758 *done = ret; 1703 *done = ret;
1759 } 1704 }
1760 1705
1761 int ocfs2_remove_inode_range(struct inode *in 1706 int ocfs2_remove_inode_range(struct inode *inode,
1762 struct buffer_he 1707 struct buffer_head *di_bh, u64 byte_start,
1763 u64 byte_len) 1708 u64 byte_len)
1764 { 1709 {
1765 int ret = 0, flags = 0, done = 0, i; 1710 int ret = 0, flags = 0, done = 0, i;
1766 u32 trunc_start, trunc_len, trunc_end 1711 u32 trunc_start, trunc_len, trunc_end, trunc_cpos, phys_cpos;
1767 u32 cluster_in_el; 1712 u32 cluster_in_el;
1768 struct ocfs2_super *osb = OCFS2_SB(in 1713 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
1769 struct ocfs2_cached_dealloc_ctxt deal 1714 struct ocfs2_cached_dealloc_ctxt dealloc;
1770 struct address_space *mapping = inode 1715 struct address_space *mapping = inode->i_mapping;
1771 struct ocfs2_extent_tree et; 1716 struct ocfs2_extent_tree et;
1772 struct ocfs2_path *path = NULL; 1717 struct ocfs2_path *path = NULL;
1773 struct ocfs2_extent_list *el = NULL; 1718 struct ocfs2_extent_list *el = NULL;
1774 struct ocfs2_extent_rec *rec = NULL; 1719 struct ocfs2_extent_rec *rec = NULL;
1775 struct ocfs2_dinode *di = (struct ocf 1720 struct ocfs2_dinode *di = (struct ocfs2_dinode *)di_bh->b_data;
1776 u64 blkno, refcount_loc = le64_to_cpu 1721 u64 blkno, refcount_loc = le64_to_cpu(di->i_refcount_loc);
1777 1722
1778 ocfs2_init_dinode_extent_tree(&et, IN 1723 ocfs2_init_dinode_extent_tree(&et, INODE_CACHE(inode), di_bh);
1779 ocfs2_init_dealloc_ctxt(&dealloc); 1724 ocfs2_init_dealloc_ctxt(&dealloc);
1780 1725
1781 trace_ocfs2_remove_inode_range( 1726 trace_ocfs2_remove_inode_range(
1782 (unsigned long long)O 1727 (unsigned long long)OCFS2_I(inode)->ip_blkno,
1783 (unsigned long long)b 1728 (unsigned long long)byte_start,
1784 (unsigned long long)b 1729 (unsigned long long)byte_len);
1785 1730
1786 if (byte_len == 0) 1731 if (byte_len == 0)
1787 return 0; 1732 return 0;
1788 1733
1789 if (OCFS2_I(inode)->ip_dyn_features & 1734 if (OCFS2_I(inode)->ip_dyn_features & OCFS2_INLINE_DATA_FL) {
1790 int id_count = ocfs2_max_inli <<
1791 <<
1792 if (byte_start > id_count || <<
1793 ret = -EINVAL; <<
1794 mlog_errno(ret); <<
1795 goto out; <<
1796 } <<
1797 <<
1798 ret = ocfs2_truncate_inline(i 1735 ret = ocfs2_truncate_inline(inode, di_bh, byte_start,
1799 b 1736 byte_start + byte_len, 0);
1800 if (ret) { 1737 if (ret) {
1801 mlog_errno(ret); 1738 mlog_errno(ret);
1802 goto out; 1739 goto out;
1803 } 1740 }
1804 /* 1741 /*
1805 * There's no need to get fan 1742 * There's no need to get fancy with the page cache
1806 * truncate of an inline-data 1743 * truncate of an inline-data inode. We're talking
1807 * about less than a page her 1744 * about less than a page here, which will be cached
1808 * in the dinode buffer anywa 1745 * in the dinode buffer anyway.
1809 */ 1746 */
1810 unmap_mapping_range(mapping, 1747 unmap_mapping_range(mapping, 0, 0, 0);
1811 truncate_inode_pages(mapping, 1748 truncate_inode_pages(mapping, 0);
1812 goto out; 1749 goto out;
1813 } 1750 }
1814 1751
1815 /* 1752 /*
1816 * For reflinks, we may need to CoW 2 1753 * For reflinks, we may need to CoW 2 clusters which might be
1817 * partially zero'd later, if hole's 1754 * partially zero'd later, if hole's start and end offset were
1818 * within one cluster(means is not ex 1755 * within one cluster(means is not exactly aligned to clustersize).
1819 */ 1756 */
1820 1757
1821 if (ocfs2_is_refcount_inode(inode)) { 1758 if (ocfs2_is_refcount_inode(inode)) {
1822 ret = ocfs2_cow_file_pos(inod 1759 ret = ocfs2_cow_file_pos(inode, di_bh, byte_start);
1823 if (ret) { 1760 if (ret) {
1824 mlog_errno(ret); 1761 mlog_errno(ret);
1825 goto out; 1762 goto out;
1826 } 1763 }
1827 1764
1828 ret = ocfs2_cow_file_pos(inod 1765 ret = ocfs2_cow_file_pos(inode, di_bh, byte_start + byte_len);
1829 if (ret) { 1766 if (ret) {
1830 mlog_errno(ret); 1767 mlog_errno(ret);
1831 goto out; 1768 goto out;
1832 } 1769 }
1833 } 1770 }
1834 1771
1835 trunc_start = ocfs2_clusters_for_byte 1772 trunc_start = ocfs2_clusters_for_bytes(osb->sb, byte_start);
1836 trunc_end = (byte_start + byte_len) > 1773 trunc_end = (byte_start + byte_len) >> osb->s_clustersize_bits;
1837 cluster_in_el = trunc_end; 1774 cluster_in_el = trunc_end;
1838 1775
1839 ret = ocfs2_zero_partial_clusters(ino 1776 ret = ocfs2_zero_partial_clusters(inode, byte_start, byte_len);
1840 if (ret) { 1777 if (ret) {
1841 mlog_errno(ret); 1778 mlog_errno(ret);
1842 goto out; 1779 goto out;
1843 } 1780 }
1844 1781
1845 path = ocfs2_new_path_from_et(&et); 1782 path = ocfs2_new_path_from_et(&et);
1846 if (!path) { 1783 if (!path) {
1847 ret = -ENOMEM; 1784 ret = -ENOMEM;
1848 mlog_errno(ret); 1785 mlog_errno(ret);
1849 goto out; 1786 goto out;
1850 } 1787 }
1851 1788
1852 while (trunc_end > trunc_start) { 1789 while (trunc_end > trunc_start) {
1853 1790
1854 ret = ocfs2_find_path(INODE_C 1791 ret = ocfs2_find_path(INODE_CACHE(inode), path,
1855 cluster 1792 cluster_in_el);
1856 if (ret) { 1793 if (ret) {
1857 mlog_errno(ret); 1794 mlog_errno(ret);
1858 goto out; 1795 goto out;
1859 } 1796 }
1860 1797
1861 el = path_leaf_el(path); 1798 el = path_leaf_el(path);
1862 1799
1863 i = ocfs2_find_rec(el, trunc_ 1800 i = ocfs2_find_rec(el, trunc_end);
1864 /* 1801 /*
1865 * Need to go to previous ext 1802 * Need to go to previous extent block.
1866 */ 1803 */
1867 if (i < 0) { 1804 if (i < 0) {
1868 if (path->p_tree_dept 1805 if (path->p_tree_depth == 0)
1869 break; 1806 break;
1870 1807
1871 ret = ocfs2_find_cpos 1808 ret = ocfs2_find_cpos_for_left_leaf(inode->i_sb,
1872 1809 path,
1873 1810 &cluster_in_el);
1874 if (ret) { 1811 if (ret) {
1875 mlog_errno(re 1812 mlog_errno(ret);
1876 goto out; 1813 goto out;
1877 } 1814 }
1878 1815
1879 /* 1816 /*
1880 * We've reached the 1817 * We've reached the leftmost extent block,
1881 * it's safe to leave 1818 * it's safe to leave.
1882 */ 1819 */
1883 if (cluster_in_el == 1820 if (cluster_in_el == 0)
1884 break; 1821 break;
1885 1822
1886 /* 1823 /*
1887 * The 'pos' searched 1824 * The 'pos' searched for previous extent block is
1888 * always one cluster 1825 * always one cluster less than actual trunc_end.
1889 */ 1826 */
1890 trunc_end = cluster_i 1827 trunc_end = cluster_in_el + 1;
1891 1828
1892 ocfs2_reinit_path(pat 1829 ocfs2_reinit_path(path, 1);
1893 1830
1894 continue; 1831 continue;
1895 1832
1896 } else 1833 } else
1897 rec = &el->l_recs[i]; 1834 rec = &el->l_recs[i];
1898 1835
1899 ocfs2_calc_trunc_pos(inode, e 1836 ocfs2_calc_trunc_pos(inode, el, rec, trunc_start, &trunc_cpos,
1900 &trunc_l 1837 &trunc_len, &trunc_end, &blkno, &done);
1901 if (done) 1838 if (done)
1902 break; 1839 break;
1903 1840
1904 flags = rec->e_flags; 1841 flags = rec->e_flags;
1905 phys_cpos = ocfs2_blocks_to_c 1842 phys_cpos = ocfs2_blocks_to_clusters(inode->i_sb, blkno);
1906 1843
1907 ret = ocfs2_remove_btree_rang 1844 ret = ocfs2_remove_btree_range(inode, &et, trunc_cpos,
1908 1845 phys_cpos, trunc_len, flags,
1909 1846 &dealloc, refcount_loc, false);
1910 if (ret < 0) { 1847 if (ret < 0) {
1911 mlog_errno(ret); 1848 mlog_errno(ret);
1912 goto out; 1849 goto out;
1913 } 1850 }
1914 1851
1915 cluster_in_el = trunc_end; 1852 cluster_in_el = trunc_end;
1916 1853
1917 ocfs2_reinit_path(path, 1); 1854 ocfs2_reinit_path(path, 1);
1918 } 1855 }
1919 1856
1920 ocfs2_truncate_cluster_pages(inode, b 1857 ocfs2_truncate_cluster_pages(inode, byte_start, byte_len);
1921 1858
1922 out: 1859 out:
1923 ocfs2_free_path(path); 1860 ocfs2_free_path(path);
1924 ocfs2_schedule_truncate_log_flush(osb 1861 ocfs2_schedule_truncate_log_flush(osb, 1);
1925 ocfs2_run_deallocs(osb, &dealloc); 1862 ocfs2_run_deallocs(osb, &dealloc);
1926 1863
1927 return ret; 1864 return ret;
1928 } 1865 }
1929 1866
1930 /* 1867 /*
1931 * Parts of this function taken from xfs_chan 1868 * Parts of this function taken from xfs_change_file_space()
1932 */ 1869 */
1933 static int __ocfs2_change_file_space(struct f 1870 static int __ocfs2_change_file_space(struct file *file, struct inode *inode,
1934 loff_t f 1871 loff_t f_pos, unsigned int cmd,
1935 struct o 1872 struct ocfs2_space_resv *sr,
1936 int chan 1873 int change_size)
1937 { 1874 {
1938 int ret; 1875 int ret;
1939 s64 llen; 1876 s64 llen;
1940 loff_t size, orig_isize; !! 1877 loff_t size;
1941 struct ocfs2_super *osb = OCFS2_SB(in 1878 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
1942 struct buffer_head *di_bh = NULL; 1879 struct buffer_head *di_bh = NULL;
1943 handle_t *handle; 1880 handle_t *handle;
1944 unsigned long long max_off = inode->i 1881 unsigned long long max_off = inode->i_sb->s_maxbytes;
1945 1882
1946 if (ocfs2_is_hard_readonly(osb) || oc 1883 if (ocfs2_is_hard_readonly(osb) || ocfs2_is_soft_readonly(osb))
1947 return -EROFS; 1884 return -EROFS;
1948 1885
1949 inode_lock(inode); 1886 inode_lock(inode);
1950 1887
1951 /* Wait all existing dio workers, new <<
1952 inode_dio_wait(inode); <<
1953 /* 1888 /*
1954 * This prevents concurrent writes on 1889 * This prevents concurrent writes on other nodes
1955 */ 1890 */
1956 ret = ocfs2_rw_lock(inode, 1); 1891 ret = ocfs2_rw_lock(inode, 1);
1957 if (ret) { 1892 if (ret) {
1958 mlog_errno(ret); 1893 mlog_errno(ret);
1959 goto out; 1894 goto out;
1960 } 1895 }
1961 1896
1962 ret = ocfs2_inode_lock(inode, &di_bh, 1897 ret = ocfs2_inode_lock(inode, &di_bh, 1);
1963 if (ret) { 1898 if (ret) {
1964 mlog_errno(ret); 1899 mlog_errno(ret);
1965 goto out_rw_unlock; 1900 goto out_rw_unlock;
1966 } 1901 }
1967 1902
1968 if (inode->i_flags & (S_IMMUTABLE|S_A 1903 if (inode->i_flags & (S_IMMUTABLE|S_APPEND)) {
1969 ret = -EPERM; 1904 ret = -EPERM;
1970 goto out_inode_unlock; 1905 goto out_inode_unlock;
1971 } 1906 }
1972 1907
1973 switch (sr->l_whence) { 1908 switch (sr->l_whence) {
1974 case 0: /*SEEK_SET*/ 1909 case 0: /*SEEK_SET*/
1975 break; 1910 break;
1976 case 1: /*SEEK_CUR*/ 1911 case 1: /*SEEK_CUR*/
1977 sr->l_start += f_pos; 1912 sr->l_start += f_pos;
1978 break; 1913 break;
1979 case 2: /*SEEK_END*/ 1914 case 2: /*SEEK_END*/
1980 sr->l_start += i_size_read(in 1915 sr->l_start += i_size_read(inode);
1981 break; 1916 break;
1982 default: 1917 default:
1983 ret = -EINVAL; 1918 ret = -EINVAL;
1984 goto out_inode_unlock; 1919 goto out_inode_unlock;
1985 } 1920 }
1986 sr->l_whence = 0; 1921 sr->l_whence = 0;
1987 1922
1988 llen = sr->l_len > 0 ? sr->l_len - 1 1923 llen = sr->l_len > 0 ? sr->l_len - 1 : sr->l_len;
1989 1924
1990 if (sr->l_start < 0 1925 if (sr->l_start < 0
1991 || sr->l_start > max_off 1926 || sr->l_start > max_off
1992 || (sr->l_start + llen) < 0 1927 || (sr->l_start + llen) < 0
1993 || (sr->l_start + llen) > max_off 1928 || (sr->l_start + llen) > max_off) {
1994 ret = -EINVAL; 1929 ret = -EINVAL;
1995 goto out_inode_unlock; 1930 goto out_inode_unlock;
1996 } 1931 }
1997 size = sr->l_start + sr->l_len; 1932 size = sr->l_start + sr->l_len;
1998 1933
1999 if (cmd == OCFS2_IOC_RESVSP || cmd == 1934 if (cmd == OCFS2_IOC_RESVSP || cmd == OCFS2_IOC_RESVSP64 ||
2000 cmd == OCFS2_IOC_UNRESVSP || cmd 1935 cmd == OCFS2_IOC_UNRESVSP || cmd == OCFS2_IOC_UNRESVSP64) {
2001 if (sr->l_len <= 0) { 1936 if (sr->l_len <= 0) {
2002 ret = -EINVAL; 1937 ret = -EINVAL;
2003 goto out_inode_unlock 1938 goto out_inode_unlock;
2004 } 1939 }
2005 } 1940 }
2006 1941
2007 if (file && setattr_should_drop_suidg !! 1942 if (file && should_remove_suid(file->f_path.dentry)) {
2008 ret = __ocfs2_write_remove_su 1943 ret = __ocfs2_write_remove_suid(inode, di_bh);
2009 if (ret) { 1944 if (ret) {
2010 mlog_errno(ret); 1945 mlog_errno(ret);
2011 goto out_inode_unlock 1946 goto out_inode_unlock;
2012 } 1947 }
2013 } 1948 }
2014 1949
2015 down_write(&OCFS2_I(inode)->ip_alloc_ 1950 down_write(&OCFS2_I(inode)->ip_alloc_sem);
2016 switch (cmd) { 1951 switch (cmd) {
2017 case OCFS2_IOC_RESVSP: 1952 case OCFS2_IOC_RESVSP:
2018 case OCFS2_IOC_RESVSP64: 1953 case OCFS2_IOC_RESVSP64:
2019 /* 1954 /*
2020 * This takes unsigned offset 1955 * This takes unsigned offsets, but the signed ones we
2021 * pass have been checked aga 1956 * pass have been checked against overflow above.
2022 */ 1957 */
2023 ret = ocfs2_allocate_unwritte 1958 ret = ocfs2_allocate_unwritten_extents(inode, sr->l_start,
2024 1959 sr->l_len);
2025 break; 1960 break;
2026 case OCFS2_IOC_UNRESVSP: 1961 case OCFS2_IOC_UNRESVSP:
2027 case OCFS2_IOC_UNRESVSP64: 1962 case OCFS2_IOC_UNRESVSP64:
2028 ret = ocfs2_remove_inode_rang 1963 ret = ocfs2_remove_inode_range(inode, di_bh, sr->l_start,
2029 1964 sr->l_len);
2030 break; 1965 break;
2031 default: 1966 default:
2032 ret = -EINVAL; 1967 ret = -EINVAL;
2033 } 1968 }
2034 <<
2035 orig_isize = i_size_read(inode); <<
2036 /* zeroout eof blocks in the cluster. <<
2037 if (!ret && change_size && orig_isize <<
2038 ret = ocfs2_zeroout_partial_c <<
2039 size <<
2040 if (!ret) <<
2041 i_size_write(inode, s <<
2042 } <<
2043 up_write(&OCFS2_I(inode)->ip_alloc_se 1969 up_write(&OCFS2_I(inode)->ip_alloc_sem);
2044 if (ret) { 1970 if (ret) {
2045 mlog_errno(ret); 1971 mlog_errno(ret);
2046 goto out_inode_unlock; 1972 goto out_inode_unlock;
2047 } 1973 }
2048 1974
2049 /* 1975 /*
2050 * We update c/mtime for these change 1976 * We update c/mtime for these changes
2051 */ 1977 */
2052 handle = ocfs2_start_trans(osb, OCFS2 1978 handle = ocfs2_start_trans(osb, OCFS2_INODE_UPDATE_CREDITS);
2053 if (IS_ERR(handle)) { 1979 if (IS_ERR(handle)) {
2054 ret = PTR_ERR(handle); 1980 ret = PTR_ERR(handle);
2055 mlog_errno(ret); 1981 mlog_errno(ret);
2056 goto out_inode_unlock; 1982 goto out_inode_unlock;
2057 } 1983 }
2058 1984
2059 inode_set_mtime_to_ts(inode, inode_se !! 1985 if (change_size && i_size_read(inode) < size)
>> 1986 i_size_write(inode, size);
>> 1987
>> 1988 inode->i_ctime = inode->i_mtime = current_time(inode);
2060 ret = ocfs2_mark_inode_dirty(handle, 1989 ret = ocfs2_mark_inode_dirty(handle, inode, di_bh);
2061 if (ret < 0) 1990 if (ret < 0)
2062 mlog_errno(ret); 1991 mlog_errno(ret);
2063 1992
2064 if (file && (file->f_flags & O_SYNC)) 1993 if (file && (file->f_flags & O_SYNC))
2065 handle->h_sync = 1; 1994 handle->h_sync = 1;
2066 1995
2067 ocfs2_commit_trans(osb, handle); 1996 ocfs2_commit_trans(osb, handle);
2068 1997
2069 out_inode_unlock: 1998 out_inode_unlock:
2070 brelse(di_bh); 1999 brelse(di_bh);
2071 ocfs2_inode_unlock(inode, 1); 2000 ocfs2_inode_unlock(inode, 1);
2072 out_rw_unlock: 2001 out_rw_unlock:
2073 ocfs2_rw_unlock(inode, 1); 2002 ocfs2_rw_unlock(inode, 1);
2074 2003
2075 out: 2004 out:
2076 inode_unlock(inode); 2005 inode_unlock(inode);
2077 return ret; 2006 return ret;
2078 } 2007 }
2079 2008
2080 int ocfs2_change_file_space(struct file *file 2009 int ocfs2_change_file_space(struct file *file, unsigned int cmd,
2081 struct ocfs2_spac 2010 struct ocfs2_space_resv *sr)
2082 { 2011 {
2083 struct inode *inode = file_inode(file 2012 struct inode *inode = file_inode(file);
2084 struct ocfs2_super *osb = OCFS2_SB(in 2013 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
2085 int ret; 2014 int ret;
2086 2015
2087 if ((cmd == OCFS2_IOC_RESVSP || cmd = 2016 if ((cmd == OCFS2_IOC_RESVSP || cmd == OCFS2_IOC_RESVSP64) &&
2088 !ocfs2_writes_unwritten_extents(o 2017 !ocfs2_writes_unwritten_extents(osb))
2089 return -ENOTTY; 2018 return -ENOTTY;
2090 else if ((cmd == OCFS2_IOC_UNRESVSP | 2019 else if ((cmd == OCFS2_IOC_UNRESVSP || cmd == OCFS2_IOC_UNRESVSP64) &&
2091 !ocfs2_sparse_alloc(osb)) 2020 !ocfs2_sparse_alloc(osb))
2092 return -ENOTTY; 2021 return -ENOTTY;
2093 2022
2094 if (!S_ISREG(inode->i_mode)) 2023 if (!S_ISREG(inode->i_mode))
2095 return -EINVAL; 2024 return -EINVAL;
2096 2025
2097 if (!(file->f_mode & FMODE_WRITE)) 2026 if (!(file->f_mode & FMODE_WRITE))
2098 return -EBADF; 2027 return -EBADF;
2099 2028
2100 ret = mnt_want_write_file(file); 2029 ret = mnt_want_write_file(file);
2101 if (ret) 2030 if (ret)
2102 return ret; 2031 return ret;
2103 ret = __ocfs2_change_file_space(file, 2032 ret = __ocfs2_change_file_space(file, inode, file->f_pos, cmd, sr, 0);
2104 mnt_drop_write_file(file); 2033 mnt_drop_write_file(file);
2105 return ret; 2034 return ret;
2106 } 2035 }
2107 2036
2108 static long ocfs2_fallocate(struct file *file 2037 static long ocfs2_fallocate(struct file *file, int mode, loff_t offset,
2109 loff_t len) 2038 loff_t len)
2110 { 2039 {
2111 struct inode *inode = file_inode(file 2040 struct inode *inode = file_inode(file);
2112 struct ocfs2_super *osb = OCFS2_SB(in 2041 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
2113 struct ocfs2_space_resv sr; 2042 struct ocfs2_space_resv sr;
2114 int change_size = 1; 2043 int change_size = 1;
2115 int cmd = OCFS2_IOC_RESVSP64; 2044 int cmd = OCFS2_IOC_RESVSP64;
2116 int ret = 0; <<
2117 2045
2118 if (mode & ~(FALLOC_FL_KEEP_SIZE | FA 2046 if (mode & ~(FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE))
2119 return -EOPNOTSUPP; 2047 return -EOPNOTSUPP;
2120 if (!ocfs2_writes_unwritten_extents(o 2048 if (!ocfs2_writes_unwritten_extents(osb))
2121 return -EOPNOTSUPP; 2049 return -EOPNOTSUPP;
2122 2050
2123 if (mode & FALLOC_FL_KEEP_SIZE) { !! 2051 if (mode & FALLOC_FL_KEEP_SIZE)
2124 change_size = 0; 2052 change_size = 0;
2125 } else { <<
2126 ret = inode_newsize_ok(inode, <<
2127 if (ret) <<
2128 return ret; <<
2129 } <<
2130 2053
2131 if (mode & FALLOC_FL_PUNCH_HOLE) 2054 if (mode & FALLOC_FL_PUNCH_HOLE)
2132 cmd = OCFS2_IOC_UNRESVSP64; 2055 cmd = OCFS2_IOC_UNRESVSP64;
2133 2056
2134 sr.l_whence = 0; 2057 sr.l_whence = 0;
2135 sr.l_start = (s64)offset; 2058 sr.l_start = (s64)offset;
2136 sr.l_len = (s64)len; 2059 sr.l_len = (s64)len;
2137 2060
2138 return __ocfs2_change_file_space(NULL 2061 return __ocfs2_change_file_space(NULL, inode, offset, cmd, &sr,
2139 chan 2062 change_size);
2140 } 2063 }
2141 2064
2142 int ocfs2_check_range_for_refcount(struct ino 2065 int ocfs2_check_range_for_refcount(struct inode *inode, loff_t pos,
2143 size_t cou 2066 size_t count)
2144 { 2067 {
2145 int ret = 0; 2068 int ret = 0;
2146 unsigned int extent_flags; 2069 unsigned int extent_flags;
2147 u32 cpos, clusters, extent_len, phys_ 2070 u32 cpos, clusters, extent_len, phys_cpos;
2148 struct super_block *sb = inode->i_sb; 2071 struct super_block *sb = inode->i_sb;
2149 2072
2150 if (!ocfs2_refcount_tree(OCFS2_SB(ino 2073 if (!ocfs2_refcount_tree(OCFS2_SB(inode->i_sb)) ||
2151 !ocfs2_is_refcount_inode(inode) | 2074 !ocfs2_is_refcount_inode(inode) ||
2152 OCFS2_I(inode)->ip_dyn_features & 2075 OCFS2_I(inode)->ip_dyn_features & OCFS2_INLINE_DATA_FL)
2153 return 0; 2076 return 0;
2154 2077
2155 cpos = pos >> OCFS2_SB(sb)->s_cluster 2078 cpos = pos >> OCFS2_SB(sb)->s_clustersize_bits;
2156 clusters = ocfs2_clusters_for_bytes(s 2079 clusters = ocfs2_clusters_for_bytes(sb, pos + count) - cpos;
2157 2080
2158 while (clusters) { 2081 while (clusters) {
2159 ret = ocfs2_get_clusters(inod 2082 ret = ocfs2_get_clusters(inode, cpos, &phys_cpos, &extent_len,
2160 &ext 2083 &extent_flags);
2161 if (ret < 0) { 2084 if (ret < 0) {
2162 mlog_errno(ret); 2085 mlog_errno(ret);
2163 goto out; 2086 goto out;
2164 } 2087 }
2165 2088
2166 if (phys_cpos && (extent_flag 2089 if (phys_cpos && (extent_flags & OCFS2_EXT_REFCOUNTED)) {
2167 ret = 1; 2090 ret = 1;
2168 break; 2091 break;
2169 } 2092 }
2170 2093
2171 if (extent_len > clusters) 2094 if (extent_len > clusters)
2172 extent_len = clusters 2095 extent_len = clusters;
2173 2096
2174 clusters -= extent_len; 2097 clusters -= extent_len;
2175 cpos += extent_len; 2098 cpos += extent_len;
2176 } 2099 }
2177 out: 2100 out:
2178 return ret; 2101 return ret;
2179 } 2102 }
2180 2103
2181 static int ocfs2_is_io_unaligned(struct inode 2104 static int ocfs2_is_io_unaligned(struct inode *inode, size_t count, loff_t pos)
2182 { 2105 {
2183 int blockmask = inode->i_sb->s_blocks 2106 int blockmask = inode->i_sb->s_blocksize - 1;
2184 loff_t final_size = pos + count; 2107 loff_t final_size = pos + count;
2185 2108
2186 if ((pos & blockmask) || (final_size 2109 if ((pos & blockmask) || (final_size & blockmask))
2187 return 1; 2110 return 1;
2188 return 0; 2111 return 0;
2189 } 2112 }
2190 2113
2191 static int ocfs2_inode_lock_for_extent_tree(s !! 2114 static int ocfs2_prepare_inode_for_refcount(struct inode *inode,
2192 s !! 2115 struct file *file,
2193 i !! 2116 loff_t pos, size_t count,
2194 i !! 2117 int *meta_level)
2195 i <<
2196 { 2118 {
2197 int ret = 0; !! 2119 int ret;
>> 2120 struct buffer_head *di_bh = NULL;
>> 2121 u32 cpos = pos >> OCFS2_SB(inode->i_sb)->s_clustersize_bits;
>> 2122 u32 clusters =
>> 2123 ocfs2_clusters_for_bytes(inode->i_sb, pos + count) - cpos;
2198 2124
2199 if (wait) !! 2125 ret = ocfs2_inode_lock(inode, &di_bh, 1);
2200 ret = ocfs2_inode_lock(inode, !! 2126 if (ret) {
2201 else !! 2127 mlog_errno(ret);
2202 ret = ocfs2_try_inode_lock(in <<
2203 if (ret < 0) <<
2204 goto out; 2128 goto out;
2205 <<
2206 if (wait) { <<
2207 if (write_sem) <<
2208 down_write(&OCFS2_I(i <<
2209 else <<
2210 down_read(&OCFS2_I(in <<
2211 } else { <<
2212 if (write_sem) <<
2213 ret = down_write_tryl <<
2214 else <<
2215 ret = down_read_trylo <<
2216 <<
2217 if (!ret) { <<
2218 ret = -EAGAIN; <<
2219 goto out_unlock; <<
2220 } <<
2221 } 2129 }
2222 2130
2223 return ret; !! 2131 *meta_level = 1;
2224 2132
2225 out_unlock: !! 2133 ret = ocfs2_refcount_cow(inode, di_bh, cpos, clusters, UINT_MAX);
2226 brelse(*di_bh); !! 2134 if (ret)
2227 *di_bh = NULL; !! 2135 mlog_errno(ret);
2228 ocfs2_inode_unlock(inode, meta_level) <<
2229 out: 2136 out:
>> 2137 brelse(di_bh);
2230 return ret; 2138 return ret;
2231 } 2139 }
2232 2140
2233 static void ocfs2_inode_unlock_for_extent_tre <<
2234 <<
2235 <<
2236 <<
2237 { <<
2238 if (write_sem) <<
2239 up_write(&OCFS2_I(inode)->ip_ <<
2240 else <<
2241 up_read(&OCFS2_I(inode)->ip_a <<
2242 <<
2243 brelse(*di_bh); <<
2244 *di_bh = NULL; <<
2245 <<
2246 if (meta_level >= 0) <<
2247 ocfs2_inode_unlock(inode, met <<
2248 } <<
2249 <<
2250 static int ocfs2_prepare_inode_for_write(stru 2141 static int ocfs2_prepare_inode_for_write(struct file *file,
2251 loff !! 2142 loff_t pos,
>> 2143 size_t count)
2252 { 2144 {
2253 int ret = 0, meta_level = 0, overwrit !! 2145 int ret = 0, meta_level = 0;
2254 int write_sem = 0; <<
2255 struct dentry *dentry = file->f_path. 2146 struct dentry *dentry = file->f_path.dentry;
2256 struct inode *inode = d_inode(dentry) 2147 struct inode *inode = d_inode(dentry);
2257 struct buffer_head *di_bh = NULL; !! 2148 loff_t end;
2258 u32 cpos; <<
2259 u32 clusters; <<
2260 2149
2261 /* 2150 /*
2262 * We start with a read level meta lo 2151 * We start with a read level meta lock and only jump to an ex
2263 * if we need to make modifications h 2152 * if we need to make modifications here.
2264 */ 2153 */
2265 for(;;) { 2154 for(;;) {
2266 ret = ocfs2_inode_lock_for_ex !! 2155 ret = ocfs2_inode_lock(inode, NULL, meta_level);
2267 <<
2268 <<
2269 <<
2270 <<
2271 if (ret < 0) { 2156 if (ret < 0) {
2272 if (ret != -EAGAIN) !! 2157 meta_level = -1;
2273 mlog_errno(re !! 2158 mlog_errno(ret);
2274 goto out; 2159 goto out;
2275 } 2160 }
2276 2161
2277 /* <<
2278 * Check if IO will overwrite <<
2279 * IOCB_NOWAIT flag is set. <<
2280 */ <<
2281 if (!wait && !overwrite_io) { <<
2282 overwrite_io = 1; <<
2283 <<
2284 ret = ocfs2_overwrite <<
2285 if (ret < 0) { <<
2286 if (ret != -E <<
2287 mlog_ <<
2288 goto out_unlo <<
2289 } <<
2290 } <<
2291 <<
2292 /* Clear suid / sgid if neces 2162 /* Clear suid / sgid if necessary. We do this here
2293 * instead of later in the wr 2163 * instead of later in the write path because
2294 * remove_suid() calls ->seta 2164 * remove_suid() calls ->setattr without any hint that
2295 * we may have already done o 2165 * we may have already done our cluster locking. Since
2296 * ocfs2_setattr() *must* tak 2166 * ocfs2_setattr() *must* take cluster locks to
2297 * proceed, this will lead us 2167 * proceed, this will lead us to recursively lock the
2298 * inode. There's also the di 2168 * inode. There's also the dinode i_size state which
2299 * can be lost via setattr du 2169 * can be lost via setattr during extending writes (we
2300 * set inode->i_size at the e 2170 * set inode->i_size at the end of a write. */
2301 if (setattr_should_drop_suidg !! 2171 if (should_remove_suid(dentry)) {
2302 if (meta_level == 0) 2172 if (meta_level == 0) {
2303 ocfs2_inode_u !! 2173 ocfs2_inode_unlock(inode, meta_level);
2304 <<
2305 <<
2306 <<
2307 meta_level = 2174 meta_level = 1;
2308 continue; 2175 continue;
2309 } 2176 }
2310 2177
2311 ret = ocfs2_write_rem 2178 ret = ocfs2_write_remove_suid(inode);
2312 if (ret < 0) { 2179 if (ret < 0) {
2313 mlog_errno(re 2180 mlog_errno(ret);
2314 goto out_unlo 2181 goto out_unlock;
2315 } 2182 }
2316 } 2183 }
2317 2184
>> 2185 end = pos + count;
>> 2186
2318 ret = ocfs2_check_range_for_r 2187 ret = ocfs2_check_range_for_refcount(inode, pos, count);
2319 if (ret == 1) { 2188 if (ret == 1) {
2320 ocfs2_inode_unlock_fo !! 2189 ocfs2_inode_unlock(inode, meta_level);
2321 !! 2190 meta_level = -1;
2322 <<
2323 <<
2324 meta_level = 1; <<
2325 write_sem = 1; <<
2326 ret = ocfs2_inode_loc <<
2327 <<
2328 <<
2329 <<
2330 <<
2331 if (ret < 0) { <<
2332 if (ret != -E <<
2333 mlog_ <<
2334 goto out; <<
2335 } <<
2336 2191
2337 cpos = pos >> OCFS2_S !! 2192 ret = ocfs2_prepare_inode_for_refcount(inode,
2338 clusters = !! 2193 file,
2339 ocfs2_cluster !! 2194 pos,
2340 ret = ocfs2_refcount_ !! 2195 count,
>> 2196 &meta_level);
2341 } 2197 }
2342 2198
2343 if (ret < 0) { 2199 if (ret < 0) {
2344 if (ret != -EAGAIN) !! 2200 mlog_errno(ret);
2345 mlog_errno(re <<
2346 goto out_unlock; 2201 goto out_unlock;
2347 } 2202 }
2348 2203
2349 break; 2204 break;
2350 } 2205 }
2351 2206
2352 out_unlock: 2207 out_unlock:
2353 trace_ocfs2_prepare_inode_for_write(O 2208 trace_ocfs2_prepare_inode_for_write(OCFS2_I(inode)->ip_blkno,
2354 p !! 2209 pos, count);
2355 2210
2356 ocfs2_inode_unlock_for_extent_tree(in !! 2211 if (meta_level >= 0)
2357 &d !! 2212 ocfs2_inode_unlock(inode, meta_level);
2358 me <<
2359 wr <<
2360 2213
2361 out: 2214 out:
2362 return ret; 2215 return ret;
2363 } 2216 }
2364 2217
2365 static ssize_t ocfs2_file_write_iter(struct k 2218 static ssize_t ocfs2_file_write_iter(struct kiocb *iocb,
2366 struct io 2219 struct iov_iter *from)
2367 { 2220 {
2368 int rw_level; !! 2221 int direct_io, rw_level;
2369 ssize_t written = 0; 2222 ssize_t written = 0;
2370 ssize_t ret; 2223 ssize_t ret;
2371 size_t count = iov_iter_count(from); 2224 size_t count = iov_iter_count(from);
2372 struct file *file = iocb->ki_filp; 2225 struct file *file = iocb->ki_filp;
2373 struct inode *inode = file_inode(file 2226 struct inode *inode = file_inode(file);
2374 struct ocfs2_super *osb = OCFS2_SB(in 2227 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
2375 int full_coherency = !(osb->s_mount_o 2228 int full_coherency = !(osb->s_mount_opt &
2376 OCFS2_MOUNT_CO 2229 OCFS2_MOUNT_COHERENCY_BUFFERED);
2377 void *saved_ki_complete = NULL; 2230 void *saved_ki_complete = NULL;
2378 int append_write = ((iocb->ki_pos + c 2231 int append_write = ((iocb->ki_pos + count) >=
2379 i_size_read(inode) ? 2232 i_size_read(inode) ? 1 : 0);
2380 int direct_io = iocb->ki_flags & IOCB <<
2381 int nowait = iocb->ki_flags & IOCB_NO <<
2382 2233
2383 trace_ocfs2_file_write_iter(inode, fi !! 2234 trace_ocfs2_file_aio_write(inode, file, file->f_path.dentry,
2384 (unsigned long long)OCFS2_I(i 2235 (unsigned long long)OCFS2_I(inode)->ip_blkno,
2385 file->f_path.dentry->d_name.l 2236 file->f_path.dentry->d_name.len,
2386 file->f_path.dentry->d_name.n 2237 file->f_path.dentry->d_name.name,
2387 (unsigned int)from->nr_segs); 2238 (unsigned int)from->nr_segs); /* GRRRRR */
2388 2239
2389 if (!direct_io && nowait) <<
2390 return -EOPNOTSUPP; <<
2391 <<
2392 if (count == 0) 2240 if (count == 0)
2393 return 0; 2241 return 0;
2394 2242
2395 if (nowait) { !! 2243 direct_io = iocb->ki_flags & IOCB_DIRECT ? 1 : 0;
2396 if (!inode_trylock(inode)) !! 2244
2397 return -EAGAIN; !! 2245 inode_lock(inode);
2398 } else <<
2399 inode_lock(inode); <<
2400 2246
2401 /* 2247 /*
2402 * Concurrent O_DIRECT writes are all 2248 * Concurrent O_DIRECT writes are allowed with
2403 * mount_option "coherency=buffered". 2249 * mount_option "coherency=buffered".
2404 * For append write, we must take rw 2250 * For append write, we must take rw EX.
2405 */ 2251 */
2406 rw_level = (!direct_io || full_cohere 2252 rw_level = (!direct_io || full_coherency || append_write);
2407 2253
2408 if (nowait) !! 2254 ret = ocfs2_rw_lock(inode, rw_level);
2409 ret = ocfs2_try_rw_lock(inode <<
2410 else <<
2411 ret = ocfs2_rw_lock(inode, rw <<
2412 if (ret < 0) { 2255 if (ret < 0) {
2413 if (ret != -EAGAIN) !! 2256 mlog_errno(ret);
2414 mlog_errno(ret); <<
2415 goto out_mutex; 2257 goto out_mutex;
2416 } 2258 }
2417 2259
2418 /* 2260 /*
2419 * O_DIRECT writes with "coherency=fu 2261 * O_DIRECT writes with "coherency=full" need to take EX cluster
2420 * inode_lock to guarantee coherency. 2262 * inode_lock to guarantee coherency.
2421 */ 2263 */
2422 if (direct_io && full_coherency) { 2264 if (direct_io && full_coherency) {
2423 /* 2265 /*
2424 * We need to take and drop t 2266 * We need to take and drop the inode lock to force
2425 * other nodes to drop their 2267 * other nodes to drop their caches. Buffered I/O
2426 * already does this in write 2268 * already does this in write_begin().
2427 */ 2269 */
2428 if (nowait) !! 2270 ret = ocfs2_inode_lock(inode, NULL, 1);
2429 ret = ocfs2_try_inode <<
2430 else <<
2431 ret = ocfs2_inode_loc <<
2432 if (ret < 0) { 2271 if (ret < 0) {
2433 if (ret != -EAGAIN) !! 2272 mlog_errno(ret);
2434 mlog_errno(re <<
2435 goto out; 2273 goto out;
2436 } 2274 }
2437 2275
2438 ocfs2_inode_unlock(inode, 1); 2276 ocfs2_inode_unlock(inode, 1);
2439 } 2277 }
2440 2278
2441 ret = generic_write_checks(iocb, from 2279 ret = generic_write_checks(iocb, from);
2442 if (ret <= 0) { 2280 if (ret <= 0) {
2443 if (ret) 2281 if (ret)
2444 mlog_errno(ret); 2282 mlog_errno(ret);
2445 goto out; 2283 goto out;
2446 } 2284 }
2447 count = ret; 2285 count = ret;
2448 2286
2449 ret = ocfs2_prepare_inode_for_write(f !! 2287 ret = ocfs2_prepare_inode_for_write(file, iocb->ki_pos, count);
2450 if (ret < 0) { 2288 if (ret < 0) {
2451 if (ret != -EAGAIN) !! 2289 mlog_errno(ret);
2452 mlog_errno(ret); <<
2453 goto out; 2290 goto out;
2454 } 2291 }
2455 2292
2456 if (direct_io && !is_sync_kiocb(iocb) 2293 if (direct_io && !is_sync_kiocb(iocb) &&
2457 ocfs2_is_io_unaligned(inode, coun 2294 ocfs2_is_io_unaligned(inode, count, iocb->ki_pos)) {
2458 /* 2295 /*
2459 * Make it a sync io if it's 2296 * Make it a sync io if it's an unaligned aio.
2460 */ 2297 */
2461 saved_ki_complete = xchg(&ioc 2298 saved_ki_complete = xchg(&iocb->ki_complete, NULL);
2462 } 2299 }
2463 2300
2464 /* communicate with ocfs2_dio_end_io 2301 /* communicate with ocfs2_dio_end_io */
2465 ocfs2_iocb_set_rw_locked(iocb, rw_lev 2302 ocfs2_iocb_set_rw_locked(iocb, rw_level);
2466 2303
2467 written = __generic_file_write_iter(i 2304 written = __generic_file_write_iter(iocb, from);
2468 /* buffered aio wouldn't have proper 2305 /* buffered aio wouldn't have proper lock coverage today */
2469 BUG_ON(written == -EIOCBQUEUED && !di !! 2306 BUG_ON(written == -EIOCBQUEUED && !(iocb->ki_flags & IOCB_DIRECT));
2470 2307
2471 /* 2308 /*
2472 * deep in g_f_a_w_n()->ocfs2_direct_ 2309 * deep in g_f_a_w_n()->ocfs2_direct_IO we pass in a ocfs2_dio_end_io
2473 * function pointer which is called w 2310 * function pointer which is called when o_direct io completes so that
2474 * it can unlock our rw lock. 2311 * it can unlock our rw lock.
2475 * Unfortunately there are error case 2312 * Unfortunately there are error cases which call end_io and others
2476 * that don't. so we don't have to u 2313 * that don't. so we don't have to unlock the rw_lock if either an
2477 * async dio is going to do it in the 2314 * async dio is going to do it in the future or an end_io after an
2478 * error has already done it. 2315 * error has already done it.
2479 */ 2316 */
2480 if ((written == -EIOCBQUEUED) || (!oc 2317 if ((written == -EIOCBQUEUED) || (!ocfs2_iocb_is_rw_locked(iocb))) {
2481 rw_level = -1; 2318 rw_level = -1;
2482 } 2319 }
2483 2320
2484 if (unlikely(written <= 0)) 2321 if (unlikely(written <= 0))
2485 goto out; 2322 goto out;
2486 2323
2487 if (((file->f_flags & O_DSYNC) && !di 2324 if (((file->f_flags & O_DSYNC) && !direct_io) ||
2488 IS_SYNC(inode)) { 2325 IS_SYNC(inode)) {
2489 ret = filemap_fdatawrite_rang 2326 ret = filemap_fdatawrite_range(file->f_mapping,
2490 2327 iocb->ki_pos - written,
2491 2328 iocb->ki_pos - 1);
2492 if (ret < 0) 2329 if (ret < 0)
2493 written = ret; 2330 written = ret;
2494 2331
2495 if (!ret) { 2332 if (!ret) {
2496 ret = jbd2_journal_fo 2333 ret = jbd2_journal_force_commit(osb->journal->j_journal);
2497 if (ret < 0) 2334 if (ret < 0)
2498 written = ret 2335 written = ret;
2499 } 2336 }
2500 2337
2501 if (!ret) 2338 if (!ret)
2502 ret = filemap_fdatawa 2339 ret = filemap_fdatawait_range(file->f_mapping,
2503 2340 iocb->ki_pos - written,
2504 2341 iocb->ki_pos - 1);
2505 } 2342 }
2506 2343
2507 out: 2344 out:
2508 if (saved_ki_complete) 2345 if (saved_ki_complete)
2509 xchg(&iocb->ki_complete, save 2346 xchg(&iocb->ki_complete, saved_ki_complete);
2510 2347
2511 if (rw_level != -1) 2348 if (rw_level != -1)
2512 ocfs2_rw_unlock(inode, rw_lev 2349 ocfs2_rw_unlock(inode, rw_level);
2513 2350
2514 out_mutex: 2351 out_mutex:
2515 inode_unlock(inode); 2352 inode_unlock(inode);
2516 2353
2517 if (written) 2354 if (written)
2518 ret = written; 2355 ret = written;
2519 return ret; 2356 return ret;
2520 } 2357 }
2521 2358
2522 static ssize_t ocfs2_file_read_iter(struct ki 2359 static ssize_t ocfs2_file_read_iter(struct kiocb *iocb,
2523 struct iov 2360 struct iov_iter *to)
2524 { 2361 {
2525 int ret = 0, rw_level = -1, lock_leve 2362 int ret = 0, rw_level = -1, lock_level = 0;
2526 struct file *filp = iocb->ki_filp; 2363 struct file *filp = iocb->ki_filp;
2527 struct inode *inode = file_inode(filp 2364 struct inode *inode = file_inode(filp);
2528 int direct_io = iocb->ki_flags & IOCB <<
2529 int nowait = iocb->ki_flags & IOCB_NO <<
2530 2365
2531 trace_ocfs2_file_read_iter(inode, fil !! 2366 trace_ocfs2_file_aio_read(inode, filp, filp->f_path.dentry,
2532 (unsigned long long)O 2367 (unsigned long long)OCFS2_I(inode)->ip_blkno,
2533 filp->f_path.dentry-> 2368 filp->f_path.dentry->d_name.len,
2534 filp->f_path.dentry-> 2369 filp->f_path.dentry->d_name.name,
2535 to->nr_segs); /* GR 2370 to->nr_segs); /* GRRRRR */
2536 2371
2537 2372
2538 if (!inode) { 2373 if (!inode) {
2539 ret = -EINVAL; 2374 ret = -EINVAL;
2540 mlog_errno(ret); 2375 mlog_errno(ret);
2541 goto bail; 2376 goto bail;
2542 } 2377 }
2543 2378
2544 if (!direct_io && nowait) <<
2545 return -EOPNOTSUPP; <<
2546 <<
2547 /* 2379 /*
2548 * buffered reads protect themselves !! 2380 * buffered reads protect themselves in ->readpage(). O_DIRECT reads
2549 * need locks to protect pending read 2381 * need locks to protect pending reads from racing with truncate.
2550 */ 2382 */
2551 if (direct_io) { !! 2383 if (iocb->ki_flags & IOCB_DIRECT) {
2552 if (nowait) !! 2384 ret = ocfs2_rw_lock(inode, 0);
2553 ret = ocfs2_try_rw_lo <<
2554 else <<
2555 ret = ocfs2_rw_lock(i <<
2556 <<
2557 if (ret < 0) { 2385 if (ret < 0) {
2558 if (ret != -EAGAIN) !! 2386 mlog_errno(ret);
2559 mlog_errno(re <<
2560 goto bail; 2387 goto bail;
2561 } 2388 }
2562 rw_level = 0; 2389 rw_level = 0;
2563 /* communicate with ocfs2_dio 2390 /* communicate with ocfs2_dio_end_io */
2564 ocfs2_iocb_set_rw_locked(iocb 2391 ocfs2_iocb_set_rw_locked(iocb, rw_level);
2565 } 2392 }
2566 2393
2567 /* 2394 /*
2568 * We're fine letting folks race trun 2395 * We're fine letting folks race truncates and extending
2569 * writes with read across the cluste 2396 * writes with read across the cluster, just like they can
2570 * locally. Hence no rw_lock during r 2397 * locally. Hence no rw_lock during read.
2571 * 2398 *
2572 * Take and drop the meta data lock t 2399 * Take and drop the meta data lock to update inode fields
2573 * like i_size. This allows the check 2400 * like i_size. This allows the checks down below
2574 * copy_splice_read() a chance of act !! 2401 * generic_file_aio_read() a chance of actually working.
2575 */ 2402 */
2576 ret = ocfs2_inode_lock_atime(inode, f !! 2403 ret = ocfs2_inode_lock_atime(inode, filp->f_path.mnt, &lock_level);
2577 !nowait) <<
2578 if (ret < 0) { 2404 if (ret < 0) {
2579 if (ret != -EAGAIN) !! 2405 mlog_errno(ret);
2580 mlog_errno(ret); <<
2581 goto bail; 2406 goto bail;
2582 } 2407 }
2583 ocfs2_inode_unlock(inode, lock_level) 2408 ocfs2_inode_unlock(inode, lock_level);
2584 2409
2585 ret = generic_file_read_iter(iocb, to 2410 ret = generic_file_read_iter(iocb, to);
2586 trace_generic_file_read_iter_ret(ret) !! 2411 trace_generic_file_aio_read_ret(ret);
2587 2412
2588 /* buffered aio wouldn't have proper 2413 /* buffered aio wouldn't have proper lock coverage today */
2589 BUG_ON(ret == -EIOCBQUEUED && !direct !! 2414 BUG_ON(ret == -EIOCBQUEUED && !(iocb->ki_flags & IOCB_DIRECT));
2590 2415
2591 /* see ocfs2_file_write_iter */ 2416 /* see ocfs2_file_write_iter */
2592 if (ret == -EIOCBQUEUED || !ocfs2_ioc 2417 if (ret == -EIOCBQUEUED || !ocfs2_iocb_is_rw_locked(iocb)) {
2593 rw_level = -1; 2418 rw_level = -1;
2594 } 2419 }
2595 2420
2596 bail: 2421 bail:
2597 if (rw_level != -1) 2422 if (rw_level != -1)
2598 ocfs2_rw_unlock(inode, rw_lev 2423 ocfs2_rw_unlock(inode, rw_level);
2599 2424
2600 return ret; 2425 return ret;
2601 } 2426 }
2602 2427
2603 static ssize_t ocfs2_file_splice_read(struct <<
2604 struct <<
2605 size_t <<
2606 { <<
2607 struct inode *inode = file_inode(in); <<
2608 ssize_t ret = 0; <<
2609 int lock_level = 0; <<
2610 <<
2611 trace_ocfs2_file_splice_read(inode, i <<
2612 (unsigne <<
2613 in->f_pa <<
2614 in->f_pa <<
2615 flags); <<
2616 <<
2617 /* <<
2618 * We're fine letting folks race trun <<
2619 * read across the cluster, just like <<
2620 * rw_lock during read. <<
2621 * <<
2622 * Take and drop the meta data lock t <<
2623 * This allows the checks down below <<
2624 * actually working. <<
2625 */ <<
2626 ret = ocfs2_inode_lock_atime(inode, i <<
2627 if (ret < 0) { <<
2628 if (ret != -EAGAIN) <<
2629 mlog_errno(ret); <<
2630 goto bail; <<
2631 } <<
2632 ocfs2_inode_unlock(inode, lock_level) <<
2633 <<
2634 ret = filemap_splice_read(in, ppos, p <<
2635 trace_filemap_splice_read_ret(ret); <<
2636 bail: <<
2637 return ret; <<
2638 } <<
2639 <<
2640 /* Refer generic_file_llseek_unlocked() */ 2428 /* Refer generic_file_llseek_unlocked() */
2641 static loff_t ocfs2_file_llseek(struct file * 2429 static loff_t ocfs2_file_llseek(struct file *file, loff_t offset, int whence)
2642 { 2430 {
2643 struct inode *inode = file->f_mapping 2431 struct inode *inode = file->f_mapping->host;
2644 int ret = 0; 2432 int ret = 0;
2645 2433
2646 inode_lock(inode); 2434 inode_lock(inode);
2647 2435
2648 switch (whence) { 2436 switch (whence) {
2649 case SEEK_SET: 2437 case SEEK_SET:
2650 break; 2438 break;
2651 case SEEK_END: 2439 case SEEK_END:
2652 /* SEEK_END requires the OCFS 2440 /* SEEK_END requires the OCFS2 inode lock for the file
2653 * because it references the 2441 * because it references the file's size.
2654 */ 2442 */
2655 ret = ocfs2_inode_lock(inode, 2443 ret = ocfs2_inode_lock(inode, NULL, 0);
2656 if (ret < 0) { 2444 if (ret < 0) {
2657 mlog_errno(ret); 2445 mlog_errno(ret);
2658 goto out; 2446 goto out;
2659 } 2447 }
2660 offset += i_size_read(inode); 2448 offset += i_size_read(inode);
2661 ocfs2_inode_unlock(inode, 0); 2449 ocfs2_inode_unlock(inode, 0);
2662 break; 2450 break;
2663 case SEEK_CUR: 2451 case SEEK_CUR:
2664 if (offset == 0) { 2452 if (offset == 0) {
2665 offset = file->f_pos; 2453 offset = file->f_pos;
2666 goto out; 2454 goto out;
2667 } 2455 }
2668 offset += file->f_pos; 2456 offset += file->f_pos;
2669 break; 2457 break;
2670 case SEEK_DATA: 2458 case SEEK_DATA:
2671 case SEEK_HOLE: 2459 case SEEK_HOLE:
2672 ret = ocfs2_seek_data_hole_of 2460 ret = ocfs2_seek_data_hole_offset(file, &offset, whence);
2673 if (ret) 2461 if (ret)
2674 goto out; 2462 goto out;
2675 break; 2463 break;
2676 default: 2464 default:
2677 ret = -EINVAL; 2465 ret = -EINVAL;
2678 goto out; 2466 goto out;
2679 } 2467 }
2680 2468
2681 offset = vfs_setpos(file, offset, ino 2469 offset = vfs_setpos(file, offset, inode->i_sb->s_maxbytes);
2682 2470
2683 out: 2471 out:
2684 inode_unlock(inode); 2472 inode_unlock(inode);
2685 if (ret) 2473 if (ret)
2686 return ret; 2474 return ret;
2687 return offset; 2475 return offset;
2688 } 2476 }
2689 2477
2690 static loff_t ocfs2_remap_file_range(struct f !! 2478 static int ocfs2_file_clone_range(struct file *file_in,
2691 struct f !! 2479 loff_t pos_in,
2692 loff_t l !! 2480 struct file *file_out,
2693 { !! 2481 loff_t pos_out,
2694 struct inode *inode_in = file_inode(f !! 2482 u64 len)
2695 struct inode *inode_out = file_inode( !! 2483 {
2696 struct ocfs2_super *osb = OCFS2_SB(in !! 2484 return ocfs2_reflink_remap_range(file_in, pos_in, file_out, pos_out,
2697 struct buffer_head *in_bh = NULL, *ou !! 2485 len, false);
2698 bool same_inode = (inode_in == inode_ !! 2486 }
2699 loff_t remapped = 0; !! 2487
2700 ssize_t ret; !! 2488 static ssize_t ocfs2_file_dedupe_range(struct file *src_file,
2701 !! 2489 u64 loff,
2702 if (remap_flags & ~(REMAP_FILE_DEDUP !! 2490 u64 len,
2703 return -EINVAL; !! 2491 struct file *dst_file,
2704 if (!ocfs2_refcount_tree(osb)) !! 2492 u64 dst_loff)
2705 return -EOPNOTSUPP; !! 2493 {
2706 if (ocfs2_is_hard_readonly(osb) || oc !! 2494 int error;
2707 return -EROFS; !! 2495
2708 !! 2496 error = ocfs2_reflink_remap_range(src_file, loff, dst_file, dst_loff,
2709 /* Lock both files against IO */ !! 2497 len, true);
2710 ret = ocfs2_reflink_inodes_lock(inode !! 2498 if (error)
2711 if (ret) !! 2499 return error;
2712 return ret; !! 2500 return len;
2713 <<
2714 /* Check file eligibility and prepare <<
2715 ret = -EINVAL; <<
2716 if ((OCFS2_I(inode_in)->ip_flags & OC <<
2717 (OCFS2_I(inode_out)->ip_flags & O <<
2718 goto out_unlock; <<
2719 <<
2720 ret = generic_remap_file_range_prep(f <<
2721 &len, remap_flags); <<
2722 if (ret < 0 || len == 0) <<
2723 goto out_unlock; <<
2724 <<
2725 /* Lock out changes to the allocation <<
2726 down_write(&OCFS2_I(inode_in)->ip_all <<
2727 if (!same_inode) <<
2728 down_write_nested(&OCFS2_I(in <<
2729 SINGLE_DEPT <<
2730 <<
2731 /* Zap any page cache for the destina <<
2732 truncate_inode_pages_range(&inode_out <<
2733 round_down <<
2734 round_up(p <<
2735 <<
2736 remapped = ocfs2_reflink_remap_blocks <<
2737 inode_out, out_bh, po <<
2738 up_write(&OCFS2_I(inode_in)->ip_alloc <<
2739 if (!same_inode) <<
2740 up_write(&OCFS2_I(inode_out)- <<
2741 if (remapped < 0) { <<
2742 ret = remapped; <<
2743 mlog_errno(ret); <<
2744 goto out_unlock; <<
2745 } <<
2746 <<
2747 /* <<
2748 * Empty the extent map so that we ma <<
2749 * record from the disk. <<
2750 */ <<
2751 ocfs2_extent_map_trunc(inode_in, 0); <<
2752 ocfs2_extent_map_trunc(inode_out, 0); <<
2753 <<
2754 ret = ocfs2_reflink_update_dest(inode <<
2755 if (ret) { <<
2756 mlog_errno(ret); <<
2757 goto out_unlock; <<
2758 } <<
2759 <<
2760 out_unlock: <<
2761 ocfs2_reflink_inodes_unlock(inode_in, <<
2762 return remapped > 0 ? remapped : ret; <<
2763 } <<
2764 <<
2765 static loff_t ocfs2_dir_llseek(struct file *f <<
2766 { <<
2767 struct ocfs2_file_private *fp = file- <<
2768 <<
2769 return generic_llseek_cookie(file, of <<
2770 } 2501 }
2771 2502
2772 const struct inode_operations ocfs2_file_iops 2503 const struct inode_operations ocfs2_file_iops = {
2773 .setattr = ocfs2_setattr, 2504 .setattr = ocfs2_setattr,
2774 .getattr = ocfs2_getattr, 2505 .getattr = ocfs2_getattr,
2775 .permission = ocfs2_permission, 2506 .permission = ocfs2_permission,
2776 .listxattr = ocfs2_listxattr, 2507 .listxattr = ocfs2_listxattr,
2777 .fiemap = ocfs2_fiemap, 2508 .fiemap = ocfs2_fiemap,
2778 .get_inode_acl = ocfs2_iop_get_acl, !! 2509 .get_acl = ocfs2_iop_get_acl,
2779 .set_acl = ocfs2_iop_set_acl, 2510 .set_acl = ocfs2_iop_set_acl,
2780 .fileattr_get = ocfs2_fileattr_get, <<
2781 .fileattr_set = ocfs2_fileattr_set, <<
2782 }; 2511 };
2783 2512
2784 const struct inode_operations ocfs2_special_f 2513 const struct inode_operations ocfs2_special_file_iops = {
2785 .setattr = ocfs2_setattr, 2514 .setattr = ocfs2_setattr,
2786 .getattr = ocfs2_getattr, 2515 .getattr = ocfs2_getattr,
2787 .listxattr = ocfs2_listxattr, <<
2788 .permission = ocfs2_permission, 2516 .permission = ocfs2_permission,
2789 .get_inode_acl = ocfs2_iop_get_acl, !! 2517 .get_acl = ocfs2_iop_get_acl,
2790 .set_acl = ocfs2_iop_set_acl, 2518 .set_acl = ocfs2_iop_set_acl,
2791 }; 2519 };
2792 2520
2793 /* 2521 /*
2794 * Other than ->lock, keep ocfs2_fops and ocf 2522 * Other than ->lock, keep ocfs2_fops and ocfs2_dops in sync with
2795 * ocfs2_fops_no_plocks and ocfs2_dops_no_plo 2523 * ocfs2_fops_no_plocks and ocfs2_dops_no_plocks!
2796 */ 2524 */
2797 const struct file_operations ocfs2_fops = { 2525 const struct file_operations ocfs2_fops = {
2798 .llseek = ocfs2_file_llseek, 2526 .llseek = ocfs2_file_llseek,
2799 .mmap = ocfs2_mmap, 2527 .mmap = ocfs2_mmap,
2800 .fsync = ocfs2_sync_file, 2528 .fsync = ocfs2_sync_file,
2801 .release = ocfs2_file_release, 2529 .release = ocfs2_file_release,
2802 .open = ocfs2_file_open, 2530 .open = ocfs2_file_open,
2803 .read_iter = ocfs2_file_read_ite 2531 .read_iter = ocfs2_file_read_iter,
2804 .write_iter = ocfs2_file_write_it 2532 .write_iter = ocfs2_file_write_iter,
2805 .unlocked_ioctl = ocfs2_ioctl, 2533 .unlocked_ioctl = ocfs2_ioctl,
2806 #ifdef CONFIG_COMPAT 2534 #ifdef CONFIG_COMPAT
2807 .compat_ioctl = ocfs2_compat_ioctl, 2535 .compat_ioctl = ocfs2_compat_ioctl,
2808 #endif 2536 #endif
2809 .lock = ocfs2_lock, 2537 .lock = ocfs2_lock,
2810 .flock = ocfs2_flock, 2538 .flock = ocfs2_flock,
2811 .splice_read = ocfs2_file_splice_r !! 2539 .splice_read = generic_file_splice_read,
2812 .splice_write = iter_file_splice_wr 2540 .splice_write = iter_file_splice_write,
2813 .fallocate = ocfs2_fallocate, 2541 .fallocate = ocfs2_fallocate,
2814 .remap_file_range = ocfs2_remap_file_ !! 2542 .clone_file_range = ocfs2_file_clone_range,
>> 2543 .dedupe_file_range = ocfs2_file_dedupe_range,
2815 }; 2544 };
2816 2545
2817 WRAP_DIR_ITER(ocfs2_readdir) // FIXME! <<
2818 const struct file_operations ocfs2_dops = { 2546 const struct file_operations ocfs2_dops = {
2819 .llseek = ocfs2_dir_llseek, !! 2547 .llseek = generic_file_llseek,
2820 .read = generic_read_dir, 2548 .read = generic_read_dir,
2821 .iterate_shared = shared_ocfs2_readdi !! 2549 .iterate = ocfs2_readdir,
2822 .fsync = ocfs2_sync_file, 2550 .fsync = ocfs2_sync_file,
2823 .release = ocfs2_dir_release, 2551 .release = ocfs2_dir_release,
2824 .open = ocfs2_dir_open, 2552 .open = ocfs2_dir_open,
2825 .unlocked_ioctl = ocfs2_ioctl, 2553 .unlocked_ioctl = ocfs2_ioctl,
2826 #ifdef CONFIG_COMPAT 2554 #ifdef CONFIG_COMPAT
2827 .compat_ioctl = ocfs2_compat_ioctl, 2555 .compat_ioctl = ocfs2_compat_ioctl,
2828 #endif 2556 #endif
2829 .lock = ocfs2_lock, 2557 .lock = ocfs2_lock,
2830 .flock = ocfs2_flock, 2558 .flock = ocfs2_flock,
2831 }; 2559 };
2832 2560
2833 /* 2561 /*
2834 * POSIX-lockless variants of our file_operat 2562 * POSIX-lockless variants of our file_operations.
2835 * 2563 *
2836 * These will be used if the underlying clust 2564 * These will be used if the underlying cluster stack does not support
2837 * posix file locking, if the user passes the 2565 * posix file locking, if the user passes the "localflocks" mount
2838 * option, or if we have a local-only fs. 2566 * option, or if we have a local-only fs.
2839 * 2567 *
2840 * ocfs2_flock is in here because all stacks 2568 * ocfs2_flock is in here because all stacks handle UNIX file locks,
2841 * so we still want it in the case of no stac 2569 * so we still want it in the case of no stack support for
2842 * plocks. Internally, it will do the right t 2570 * plocks. Internally, it will do the right thing when asked to ignore
2843 * the cluster. 2571 * the cluster.
2844 */ 2572 */
2845 const struct file_operations ocfs2_fops_no_pl 2573 const struct file_operations ocfs2_fops_no_plocks = {
2846 .llseek = ocfs2_file_llseek, 2574 .llseek = ocfs2_file_llseek,
2847 .mmap = ocfs2_mmap, 2575 .mmap = ocfs2_mmap,
2848 .fsync = ocfs2_sync_file, 2576 .fsync = ocfs2_sync_file,
2849 .release = ocfs2_file_release, 2577 .release = ocfs2_file_release,
2850 .open = ocfs2_file_open, 2578 .open = ocfs2_file_open,
2851 .read_iter = ocfs2_file_read_ite 2579 .read_iter = ocfs2_file_read_iter,
2852 .write_iter = ocfs2_file_write_it 2580 .write_iter = ocfs2_file_write_iter,
2853 .unlocked_ioctl = ocfs2_ioctl, 2581 .unlocked_ioctl = ocfs2_ioctl,
2854 #ifdef CONFIG_COMPAT 2582 #ifdef CONFIG_COMPAT
2855 .compat_ioctl = ocfs2_compat_ioctl, 2583 .compat_ioctl = ocfs2_compat_ioctl,
2856 #endif 2584 #endif
2857 .flock = ocfs2_flock, 2585 .flock = ocfs2_flock,
2858 .splice_read = filemap_splice_read !! 2586 .splice_read = generic_file_splice_read,
2859 .splice_write = iter_file_splice_wr 2587 .splice_write = iter_file_splice_write,
2860 .fallocate = ocfs2_fallocate, 2588 .fallocate = ocfs2_fallocate,
2861 .remap_file_range = ocfs2_remap_file_ !! 2589 .clone_file_range = ocfs2_file_clone_range,
>> 2590 .dedupe_file_range = ocfs2_file_dedupe_range,
2862 }; 2591 };
2863 2592
2864 const struct file_operations ocfs2_dops_no_pl 2593 const struct file_operations ocfs2_dops_no_plocks = {
2865 .llseek = ocfs2_dir_llseek, !! 2594 .llseek = generic_file_llseek,
2866 .read = generic_read_dir, 2595 .read = generic_read_dir,
2867 .iterate_shared = shared_ocfs2_readdi !! 2596 .iterate = ocfs2_readdir,
2868 .fsync = ocfs2_sync_file, 2597 .fsync = ocfs2_sync_file,
2869 .release = ocfs2_dir_release, 2598 .release = ocfs2_dir_release,
2870 .open = ocfs2_dir_open, 2599 .open = ocfs2_dir_open,
2871 .unlocked_ioctl = ocfs2_ioctl, 2600 .unlocked_ioctl = ocfs2_ioctl,
2872 #ifdef CONFIG_COMPAT 2601 #ifdef CONFIG_COMPAT
2873 .compat_ioctl = ocfs2_compat_ioctl, 2602 .compat_ioctl = ocfs2_compat_ioctl,
2874 #endif 2603 #endif
2875 .flock = ocfs2_flock, 2604 .flock = ocfs2_flock,
2876 }; 2605 };
2877 2606
Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.