1 // SPDX-License-Identifier: GPL-2.0-or-later 1 // SPDX-License-Identifier: GPL-2.0-or-later
2 /* !! 2 /* -*- mode: c; c-basic-offset: 8; -*-
>> 3 * vim: noexpandtab sw=8 ts=8 sts=0:
>> 4 *
3 * file.c 5 * file.c
4 * 6 *
5 * File open, close, extend, truncate 7 * File open, close, extend, truncate
6 * 8 *
7 * Copyright (C) 2002, 2004 Oracle. All right 9 * Copyright (C) 2002, 2004 Oracle. All rights reserved.
8 */ 10 */
9 11
10 #include <linux/capability.h> 12 #include <linux/capability.h>
11 #include <linux/fs.h> 13 #include <linux/fs.h>
12 #include <linux/types.h> 14 #include <linux/types.h>
13 #include <linux/slab.h> 15 #include <linux/slab.h>
14 #include <linux/highmem.h> 16 #include <linux/highmem.h>
15 #include <linux/pagemap.h> 17 #include <linux/pagemap.h>
16 #include <linux/uio.h> 18 #include <linux/uio.h>
17 #include <linux/sched.h> 19 #include <linux/sched.h>
18 #include <linux/splice.h> 20 #include <linux/splice.h>
19 #include <linux/mount.h> 21 #include <linux/mount.h>
20 #include <linux/writeback.h> 22 #include <linux/writeback.h>
21 #include <linux/falloc.h> 23 #include <linux/falloc.h>
22 #include <linux/quotaops.h> 24 #include <linux/quotaops.h>
23 #include <linux/blkdev.h> 25 #include <linux/blkdev.h>
24 #include <linux/backing-dev.h> 26 #include <linux/backing-dev.h>
25 27
26 #include <cluster/masklog.h> 28 #include <cluster/masklog.h>
27 29
28 #include "ocfs2.h" 30 #include "ocfs2.h"
29 31
30 #include "alloc.h" 32 #include "alloc.h"
31 #include "aops.h" 33 #include "aops.h"
32 #include "dir.h" 34 #include "dir.h"
33 #include "dlmglue.h" 35 #include "dlmglue.h"
34 #include "extent_map.h" 36 #include "extent_map.h"
35 #include "file.h" 37 #include "file.h"
36 #include "sysfile.h" 38 #include "sysfile.h"
37 #include "inode.h" 39 #include "inode.h"
38 #include "ioctl.h" 40 #include "ioctl.h"
39 #include "journal.h" 41 #include "journal.h"
40 #include "locks.h" 42 #include "locks.h"
41 #include "mmap.h" 43 #include "mmap.h"
42 #include "suballoc.h" 44 #include "suballoc.h"
43 #include "super.h" 45 #include "super.h"
44 #include "xattr.h" 46 #include "xattr.h"
45 #include "acl.h" 47 #include "acl.h"
46 #include "quota.h" 48 #include "quota.h"
47 #include "refcounttree.h" 49 #include "refcounttree.h"
48 #include "ocfs2_trace.h" 50 #include "ocfs2_trace.h"
49 51
50 #include "buffer_head_io.h" 52 #include "buffer_head_io.h"
51 53
52 static int ocfs2_init_file_private(struct inod 54 static int ocfs2_init_file_private(struct inode *inode, struct file *file)
53 { 55 {
54 struct ocfs2_file_private *fp; 56 struct ocfs2_file_private *fp;
55 57
56 fp = kzalloc(sizeof(struct ocfs2_file_ 58 fp = kzalloc(sizeof(struct ocfs2_file_private), GFP_KERNEL);
57 if (!fp) 59 if (!fp)
58 return -ENOMEM; 60 return -ENOMEM;
59 61
60 fp->fp_file = file; 62 fp->fp_file = file;
61 mutex_init(&fp->fp_mutex); 63 mutex_init(&fp->fp_mutex);
62 ocfs2_file_lock_res_init(&fp->fp_flock 64 ocfs2_file_lock_res_init(&fp->fp_flock, fp);
63 file->private_data = fp; 65 file->private_data = fp;
64 66
65 return 0; 67 return 0;
66 } 68 }
67 69
68 static void ocfs2_free_file_private(struct ino 70 static void ocfs2_free_file_private(struct inode *inode, struct file *file)
69 { 71 {
70 struct ocfs2_file_private *fp = file-> 72 struct ocfs2_file_private *fp = file->private_data;
71 struct ocfs2_super *osb = OCFS2_SB(ino 73 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
72 74
73 if (fp) { 75 if (fp) {
74 ocfs2_simple_drop_lockres(osb, 76 ocfs2_simple_drop_lockres(osb, &fp->fp_flock);
75 ocfs2_lock_res_free(&fp->fp_fl 77 ocfs2_lock_res_free(&fp->fp_flock);
76 kfree(fp); 78 kfree(fp);
77 file->private_data = NULL; 79 file->private_data = NULL;
78 } 80 }
79 } 81 }
80 82
81 static int ocfs2_file_open(struct inode *inode 83 static int ocfs2_file_open(struct inode *inode, struct file *file)
82 { 84 {
83 int status; 85 int status;
84 int mode = file->f_flags; 86 int mode = file->f_flags;
85 struct ocfs2_inode_info *oi = OCFS2_I( 87 struct ocfs2_inode_info *oi = OCFS2_I(inode);
86 88
87 trace_ocfs2_file_open(inode, file, fil 89 trace_ocfs2_file_open(inode, file, file->f_path.dentry,
88 (unsigned long l 90 (unsigned long long)oi->ip_blkno,
89 file->f_path.den 91 file->f_path.dentry->d_name.len,
90 file->f_path.den 92 file->f_path.dentry->d_name.name, mode);
91 93
92 if (file->f_mode & FMODE_WRITE) { 94 if (file->f_mode & FMODE_WRITE) {
93 status = dquot_initialize(inod 95 status = dquot_initialize(inode);
94 if (status) 96 if (status)
95 goto leave; 97 goto leave;
96 } 98 }
97 99
98 spin_lock(&oi->ip_lock); 100 spin_lock(&oi->ip_lock);
99 101
100 /* Check that the inode hasn't been wi 102 /* Check that the inode hasn't been wiped from disk by another
101 * node. If it hasn't then we're safe 103 * node. If it hasn't then we're safe as long as we hold the
102 * spin lock until our increment of op 104 * spin lock until our increment of open count. */
103 if (oi->ip_flags & OCFS2_INODE_DELETED 105 if (oi->ip_flags & OCFS2_INODE_DELETED) {
104 spin_unlock(&oi->ip_lock); 106 spin_unlock(&oi->ip_lock);
105 107
106 status = -ENOENT; 108 status = -ENOENT;
107 goto leave; 109 goto leave;
108 } 110 }
109 111
110 if (mode & O_DIRECT) 112 if (mode & O_DIRECT)
111 oi->ip_flags |= OCFS2_INODE_OP 113 oi->ip_flags |= OCFS2_INODE_OPEN_DIRECT;
112 114
113 oi->ip_open_count++; 115 oi->ip_open_count++;
114 spin_unlock(&oi->ip_lock); 116 spin_unlock(&oi->ip_lock);
115 117
116 status = ocfs2_init_file_private(inode 118 status = ocfs2_init_file_private(inode, file);
117 if (status) { 119 if (status) {
118 /* 120 /*
119 * We want to set open count b 121 * We want to set open count back if we're failing the
120 * open. 122 * open.
121 */ 123 */
122 spin_lock(&oi->ip_lock); 124 spin_lock(&oi->ip_lock);
123 oi->ip_open_count--; 125 oi->ip_open_count--;
124 spin_unlock(&oi->ip_lock); 126 spin_unlock(&oi->ip_lock);
125 } 127 }
126 128
127 file->f_mode |= FMODE_NOWAIT; 129 file->f_mode |= FMODE_NOWAIT;
128 130
129 leave: 131 leave:
130 return status; 132 return status;
131 } 133 }
132 134
133 static int ocfs2_file_release(struct inode *in 135 static int ocfs2_file_release(struct inode *inode, struct file *file)
134 { 136 {
135 struct ocfs2_inode_info *oi = OCFS2_I( 137 struct ocfs2_inode_info *oi = OCFS2_I(inode);
136 138
137 spin_lock(&oi->ip_lock); 139 spin_lock(&oi->ip_lock);
138 if (!--oi->ip_open_count) 140 if (!--oi->ip_open_count)
139 oi->ip_flags &= ~OCFS2_INODE_O 141 oi->ip_flags &= ~OCFS2_INODE_OPEN_DIRECT;
140 142
141 trace_ocfs2_file_release(inode, file, 143 trace_ocfs2_file_release(inode, file, file->f_path.dentry,
142 oi->ip_blkno, 144 oi->ip_blkno,
143 file->f_path. 145 file->f_path.dentry->d_name.len,
144 file->f_path. 146 file->f_path.dentry->d_name.name,
145 oi->ip_open_c 147 oi->ip_open_count);
146 spin_unlock(&oi->ip_lock); 148 spin_unlock(&oi->ip_lock);
147 149
148 ocfs2_free_file_private(inode, file); 150 ocfs2_free_file_private(inode, file);
149 151
150 return 0; 152 return 0;
151 } 153 }
152 154
153 static int ocfs2_dir_open(struct inode *inode, 155 static int ocfs2_dir_open(struct inode *inode, struct file *file)
154 { 156 {
155 return ocfs2_init_file_private(inode, 157 return ocfs2_init_file_private(inode, file);
156 } 158 }
157 159
158 static int ocfs2_dir_release(struct inode *ino 160 static int ocfs2_dir_release(struct inode *inode, struct file *file)
159 { 161 {
160 ocfs2_free_file_private(inode, file); 162 ocfs2_free_file_private(inode, file);
161 return 0; 163 return 0;
162 } 164 }
163 165
164 static int ocfs2_sync_file(struct file *file, 166 static int ocfs2_sync_file(struct file *file, loff_t start, loff_t end,
165 int datasync) 167 int datasync)
166 { 168 {
167 int err = 0; 169 int err = 0;
168 struct inode *inode = file->f_mapping- 170 struct inode *inode = file->f_mapping->host;
169 struct ocfs2_super *osb = OCFS2_SB(ino 171 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
170 struct ocfs2_inode_info *oi = OCFS2_I( 172 struct ocfs2_inode_info *oi = OCFS2_I(inode);
171 journal_t *journal = osb->journal->j_j 173 journal_t *journal = osb->journal->j_journal;
172 int ret; 174 int ret;
173 tid_t commit_tid; 175 tid_t commit_tid;
174 bool needs_barrier = false; 176 bool needs_barrier = false;
175 177
176 trace_ocfs2_sync_file(inode, file, fil 178 trace_ocfs2_sync_file(inode, file, file->f_path.dentry,
177 oi->ip_blkno, 179 oi->ip_blkno,
178 file->f_path.den 180 file->f_path.dentry->d_name.len,
179 file->f_path.den 181 file->f_path.dentry->d_name.name,
180 (unsigned long l 182 (unsigned long long)datasync);
181 183
182 if (ocfs2_is_hard_readonly(osb) || ocf 184 if (ocfs2_is_hard_readonly(osb) || ocfs2_is_soft_readonly(osb))
183 return -EROFS; 185 return -EROFS;
184 186
185 err = file_write_and_wait_range(file, 187 err = file_write_and_wait_range(file, start, end);
186 if (err) 188 if (err)
187 return err; 189 return err;
188 190
189 commit_tid = datasync ? oi->i_datasync 191 commit_tid = datasync ? oi->i_datasync_tid : oi->i_sync_tid;
190 if (journal->j_flags & JBD2_BARRIER && 192 if (journal->j_flags & JBD2_BARRIER &&
191 !jbd2_trans_will_send_data_barrier 193 !jbd2_trans_will_send_data_barrier(journal, commit_tid))
192 needs_barrier = true; 194 needs_barrier = true;
193 err = jbd2_complete_transaction(journa 195 err = jbd2_complete_transaction(journal, commit_tid);
194 if (needs_barrier) { 196 if (needs_barrier) {
195 ret = blkdev_issue_flush(inode !! 197 ret = blkdev_issue_flush(inode->i_sb->s_bdev, GFP_KERNEL, NULL);
196 if (!err) 198 if (!err)
197 err = ret; 199 err = ret;
198 } 200 }
199 201
200 if (err) 202 if (err)
201 mlog_errno(err); 203 mlog_errno(err);
202 204
203 return (err < 0) ? -EIO : 0; 205 return (err < 0) ? -EIO : 0;
204 } 206 }
205 207
206 int ocfs2_should_update_atime(struct inode *in 208 int ocfs2_should_update_atime(struct inode *inode,
207 struct vfsmount 209 struct vfsmount *vfsmnt)
208 { 210 {
209 struct timespec64 now; 211 struct timespec64 now;
210 struct ocfs2_super *osb = OCFS2_SB(ino 212 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
211 213
212 if (ocfs2_is_hard_readonly(osb) || ocf 214 if (ocfs2_is_hard_readonly(osb) || ocfs2_is_soft_readonly(osb))
213 return 0; 215 return 0;
214 216
215 if ((inode->i_flags & S_NOATIME) || 217 if ((inode->i_flags & S_NOATIME) ||
216 ((inode->i_sb->s_flags & SB_NODIRA 218 ((inode->i_sb->s_flags & SB_NODIRATIME) && S_ISDIR(inode->i_mode)))
217 return 0; 219 return 0;
218 220
219 /* 221 /*
220 * We can be called with no vfsmnt str 222 * We can be called with no vfsmnt structure - NFSD will
221 * sometimes do this. 223 * sometimes do this.
222 * 224 *
223 * Note that our action here is differ 225 * Note that our action here is different than touch_atime() -
224 * if we can't tell whether this is a 226 * if we can't tell whether this is a noatime mount, then we
225 * don't know whether to trust the val 227 * don't know whether to trust the value of s_atime_quantum.
226 */ 228 */
227 if (vfsmnt == NULL) 229 if (vfsmnt == NULL)
228 return 0; 230 return 0;
229 231
230 if ((vfsmnt->mnt_flags & MNT_NOATIME) 232 if ((vfsmnt->mnt_flags & MNT_NOATIME) ||
231 ((vfsmnt->mnt_flags & MNT_NODIRATI 233 ((vfsmnt->mnt_flags & MNT_NODIRATIME) && S_ISDIR(inode->i_mode)))
232 return 0; 234 return 0;
233 235
234 if (vfsmnt->mnt_flags & MNT_RELATIME) 236 if (vfsmnt->mnt_flags & MNT_RELATIME) {
235 struct timespec64 ctime = inod !! 237 if ((timespec64_compare(&inode->i_atime, &inode->i_mtime) <= 0) ||
236 struct timespec64 atime = inod !! 238 (timespec64_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; 239 return 1;
242 240
243 return 0; 241 return 0;
244 } 242 }
245 243
246 now = current_time(inode); 244 now = current_time(inode);
247 if ((now.tv_sec - inode_get_atime_sec( !! 245 if ((now.tv_sec - inode->i_atime.tv_sec <= osb->s_atime_quantum))
248 return 0; 246 return 0;
249 else 247 else
250 return 1; 248 return 1;
251 } 249 }
252 250
253 int ocfs2_update_inode_atime(struct inode *ino 251 int ocfs2_update_inode_atime(struct inode *inode,
254 struct buffer_hea 252 struct buffer_head *bh)
255 { 253 {
256 int ret; 254 int ret;
257 struct ocfs2_super *osb = OCFS2_SB(ino 255 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
258 handle_t *handle; 256 handle_t *handle;
259 struct ocfs2_dinode *di = (struct ocfs 257 struct ocfs2_dinode *di = (struct ocfs2_dinode *) bh->b_data;
260 258
261 handle = ocfs2_start_trans(osb, OCFS2_ 259 handle = ocfs2_start_trans(osb, OCFS2_INODE_UPDATE_CREDITS);
262 if (IS_ERR(handle)) { 260 if (IS_ERR(handle)) {
263 ret = PTR_ERR(handle); 261 ret = PTR_ERR(handle);
264 mlog_errno(ret); 262 mlog_errno(ret);
265 goto out; 263 goto out;
266 } 264 }
267 265
268 ret = ocfs2_journal_access_di(handle, 266 ret = ocfs2_journal_access_di(handle, INODE_CACHE(inode), bh,
269 OCFS2_JO 267 OCFS2_JOURNAL_ACCESS_WRITE);
270 if (ret) { 268 if (ret) {
271 mlog_errno(ret); 269 mlog_errno(ret);
272 goto out_commit; 270 goto out_commit;
273 } 271 }
274 272
275 /* 273 /*
276 * Don't use ocfs2_mark_inode_dirty() 274 * Don't use ocfs2_mark_inode_dirty() here as we don't always
277 * have i_rwsem to guard against concu !! 275 * have i_mutex to guard against concurrent changes to other
278 * inode fields. 276 * inode fields.
279 */ 277 */
280 inode_set_atime_to_ts(inode, current_t !! 278 inode->i_atime = current_time(inode);
281 di->i_atime = cpu_to_le64(inode_get_at !! 279 di->i_atime = cpu_to_le64(inode->i_atime.tv_sec);
282 di->i_atime_nsec = cpu_to_le32(inode_g !! 280 di->i_atime_nsec = cpu_to_le32(inode->i_atime.tv_nsec);
283 ocfs2_update_inode_fsync_trans(handle, 281 ocfs2_update_inode_fsync_trans(handle, inode, 0);
284 ocfs2_journal_dirty(handle, bh); 282 ocfs2_journal_dirty(handle, bh);
285 283
286 out_commit: 284 out_commit:
287 ocfs2_commit_trans(osb, handle); 285 ocfs2_commit_trans(osb, handle);
288 out: 286 out:
289 return ret; 287 return ret;
290 } 288 }
291 289
292 int ocfs2_set_inode_size(handle_t *handle, 290 int ocfs2_set_inode_size(handle_t *handle,
293 struct inode * 291 struct inode *inode,
294 struct buffer_ 292 struct buffer_head *fe_bh,
295 u64 new_i_size 293 u64 new_i_size)
296 { 294 {
297 int status; 295 int status;
298 296
299 i_size_write(inode, new_i_size); 297 i_size_write(inode, new_i_size);
300 inode->i_blocks = ocfs2_inode_sector_c 298 inode->i_blocks = ocfs2_inode_sector_count(inode);
301 inode_set_mtime_to_ts(inode, inode_set !! 299 inode->i_ctime = inode->i_mtime = current_time(inode);
302 300
303 status = ocfs2_mark_inode_dirty(handle 301 status = ocfs2_mark_inode_dirty(handle, inode, fe_bh);
304 if (status < 0) { 302 if (status < 0) {
305 mlog_errno(status); 303 mlog_errno(status);
306 goto bail; 304 goto bail;
307 } 305 }
308 306
309 bail: 307 bail:
310 return status; 308 return status;
311 } 309 }
312 310
313 int ocfs2_simple_size_update(struct inode *ino 311 int ocfs2_simple_size_update(struct inode *inode,
314 struct buffer_hea 312 struct buffer_head *di_bh,
315 u64 new_i_size) 313 u64 new_i_size)
316 { 314 {
317 int ret; 315 int ret;
318 struct ocfs2_super *osb = OCFS2_SB(ino 316 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
319 handle_t *handle = NULL; 317 handle_t *handle = NULL;
320 318
321 handle = ocfs2_start_trans(osb, OCFS2_ 319 handle = ocfs2_start_trans(osb, OCFS2_INODE_UPDATE_CREDITS);
322 if (IS_ERR(handle)) { 320 if (IS_ERR(handle)) {
323 ret = PTR_ERR(handle); 321 ret = PTR_ERR(handle);
324 mlog_errno(ret); 322 mlog_errno(ret);
325 goto out; 323 goto out;
326 } 324 }
327 325
328 ret = ocfs2_set_inode_size(handle, ino 326 ret = ocfs2_set_inode_size(handle, inode, di_bh,
329 new_i_size) 327 new_i_size);
330 if (ret < 0) 328 if (ret < 0)
331 mlog_errno(ret); 329 mlog_errno(ret);
332 330
333 ocfs2_update_inode_fsync_trans(handle, 331 ocfs2_update_inode_fsync_trans(handle, inode, 0);
334 ocfs2_commit_trans(osb, handle); 332 ocfs2_commit_trans(osb, handle);
335 out: 333 out:
336 return ret; 334 return ret;
337 } 335 }
338 336
339 static int ocfs2_cow_file_pos(struct inode *in 337 static int ocfs2_cow_file_pos(struct inode *inode,
340 struct buffer_he 338 struct buffer_head *fe_bh,
341 u64 offset) 339 u64 offset)
342 { 340 {
343 int status; 341 int status;
344 u32 phys, cpos = offset >> OCFS2_SB(in 342 u32 phys, cpos = offset >> OCFS2_SB(inode->i_sb)->s_clustersize_bits;
345 unsigned int num_clusters = 0; 343 unsigned int num_clusters = 0;
346 unsigned int ext_flags = 0; 344 unsigned int ext_flags = 0;
347 345
348 /* 346 /*
349 * If the new offset is aligned to the 347 * If the new offset is aligned to the range of the cluster, there is
350 * no space for ocfs2_zero_range_for_t 348 * no space for ocfs2_zero_range_for_truncate to fill, so no need to
351 * CoW either. 349 * CoW either.
352 */ 350 */
353 if ((offset & (OCFS2_SB(inode->i_sb)-> 351 if ((offset & (OCFS2_SB(inode->i_sb)->s_clustersize - 1)) == 0)
354 return 0; 352 return 0;
355 353
356 status = ocfs2_get_clusters(inode, cpo 354 status = ocfs2_get_clusters(inode, cpos, &phys,
357 &num_clust 355 &num_clusters, &ext_flags);
358 if (status) { 356 if (status) {
359 mlog_errno(status); 357 mlog_errno(status);
360 goto out; 358 goto out;
361 } 359 }
362 360
363 if (!(ext_flags & OCFS2_EXT_REFCOUNTED 361 if (!(ext_flags & OCFS2_EXT_REFCOUNTED))
364 goto out; 362 goto out;
365 363
366 return ocfs2_refcount_cow(inode, fe_bh 364 return ocfs2_refcount_cow(inode, fe_bh, cpos, 1, cpos+1);
367 365
368 out: 366 out:
369 return status; 367 return status;
370 } 368 }
371 369
372 static int ocfs2_orphan_for_truncate(struct oc 370 static int ocfs2_orphan_for_truncate(struct ocfs2_super *osb,
373 struct in 371 struct inode *inode,
374 struct bu 372 struct buffer_head *fe_bh,
375 u64 new_i 373 u64 new_i_size)
376 { 374 {
377 int status; 375 int status;
378 handle_t *handle; 376 handle_t *handle;
379 struct ocfs2_dinode *di; 377 struct ocfs2_dinode *di;
380 u64 cluster_bytes; 378 u64 cluster_bytes;
381 379
382 /* 380 /*
383 * We need to CoW the cluster contains 381 * We need to CoW the cluster contains the offset if it is reflinked
384 * since we will call ocfs2_zero_range 382 * since we will call ocfs2_zero_range_for_truncate later which will
385 * write "" from offset to the end of 383 * write "" from offset to the end of the cluster.
386 */ 384 */
387 status = ocfs2_cow_file_pos(inode, fe_ 385 status = ocfs2_cow_file_pos(inode, fe_bh, new_i_size);
388 if (status) { 386 if (status) {
389 mlog_errno(status); 387 mlog_errno(status);
390 return status; 388 return status;
391 } 389 }
392 390
393 /* TODO: This needs to actually orphan 391 /* TODO: This needs to actually orphan the inode in this
394 * transaction. */ 392 * transaction. */
395 393
396 handle = ocfs2_start_trans(osb, OCFS2_ 394 handle = ocfs2_start_trans(osb, OCFS2_INODE_UPDATE_CREDITS);
397 if (IS_ERR(handle)) { 395 if (IS_ERR(handle)) {
398 status = PTR_ERR(handle); 396 status = PTR_ERR(handle);
399 mlog_errno(status); 397 mlog_errno(status);
400 goto out; 398 goto out;
401 } 399 }
402 400
403 status = ocfs2_journal_access_di(handl 401 status = ocfs2_journal_access_di(handle, INODE_CACHE(inode), fe_bh,
404 OCFS2 402 OCFS2_JOURNAL_ACCESS_WRITE);
405 if (status < 0) { 403 if (status < 0) {
406 mlog_errno(status); 404 mlog_errno(status);
407 goto out_commit; 405 goto out_commit;
408 } 406 }
409 407
410 /* 408 /*
411 * Do this before setting i_size. 409 * Do this before setting i_size.
412 */ 410 */
413 cluster_bytes = ocfs2_align_bytes_to_c 411 cluster_bytes = ocfs2_align_bytes_to_clusters(inode->i_sb, new_i_size);
414 status = ocfs2_zero_range_for_truncate 412 status = ocfs2_zero_range_for_truncate(inode, handle, new_i_size,
415 413 cluster_bytes);
416 if (status) { 414 if (status) {
417 mlog_errno(status); 415 mlog_errno(status);
418 goto out_commit; 416 goto out_commit;
419 } 417 }
420 418
421 i_size_write(inode, new_i_size); 419 i_size_write(inode, new_i_size);
422 inode_set_mtime_to_ts(inode, inode_set !! 420 inode->i_ctime = inode->i_mtime = current_time(inode);
423 421
424 di = (struct ocfs2_dinode *) fe_bh->b_ 422 di = (struct ocfs2_dinode *) fe_bh->b_data;
425 di->i_size = cpu_to_le64(new_i_size); 423 di->i_size = cpu_to_le64(new_i_size);
426 di->i_ctime = di->i_mtime = cpu_to_le6 !! 424 di->i_ctime = di->i_mtime = cpu_to_le64(inode->i_ctime.tv_sec);
427 di->i_ctime_nsec = di->i_mtime_nsec = !! 425 di->i_ctime_nsec = di->i_mtime_nsec = cpu_to_le32(inode->i_ctime.tv_nsec);
428 ocfs2_update_inode_fsync_trans(handle, 426 ocfs2_update_inode_fsync_trans(handle, inode, 0);
429 427
430 ocfs2_journal_dirty(handle, fe_bh); 428 ocfs2_journal_dirty(handle, fe_bh);
431 429
432 out_commit: 430 out_commit:
433 ocfs2_commit_trans(osb, handle); 431 ocfs2_commit_trans(osb, handle);
434 out: 432 out:
435 return status; 433 return status;
436 } 434 }
437 435
438 int ocfs2_truncate_file(struct inode *inode, 436 int ocfs2_truncate_file(struct inode *inode,
439 struct buffer_h 437 struct buffer_head *di_bh,
440 u64 new_i_size) 438 u64 new_i_size)
441 { 439 {
442 int status = 0; 440 int status = 0;
443 struct ocfs2_dinode *fe = NULL; 441 struct ocfs2_dinode *fe = NULL;
444 struct ocfs2_super *osb = OCFS2_SB(ino 442 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
445 443
446 /* We trust di_bh because it comes fro 444 /* We trust di_bh because it comes from ocfs2_inode_lock(), which
447 * already validated it */ 445 * already validated it */
448 fe = (struct ocfs2_dinode *) di_bh->b_ 446 fe = (struct ocfs2_dinode *) di_bh->b_data;
449 447
450 trace_ocfs2_truncate_file((unsigned lo 448 trace_ocfs2_truncate_file((unsigned long long)OCFS2_I(inode)->ip_blkno,
451 (unsigned lo 449 (unsigned long long)le64_to_cpu(fe->i_size),
452 (unsigned lo 450 (unsigned long long)new_i_size);
453 451
454 mlog_bug_on_msg(le64_to_cpu(fe->i_size 452 mlog_bug_on_msg(le64_to_cpu(fe->i_size) != i_size_read(inode),
455 "Inode %llu, inode i_s 453 "Inode %llu, inode i_size = %lld != di "
456 "i_size = %llu, i_flag 454 "i_size = %llu, i_flags = 0x%x\n",
457 (unsigned long long)OC 455 (unsigned long long)OCFS2_I(inode)->ip_blkno,
458 i_size_read(inode), 456 i_size_read(inode),
459 (unsigned long long)le 457 (unsigned long long)le64_to_cpu(fe->i_size),
460 le32_to_cpu(fe->i_flag 458 le32_to_cpu(fe->i_flags));
461 459
462 if (new_i_size > le64_to_cpu(fe->i_siz 460 if (new_i_size > le64_to_cpu(fe->i_size)) {
463 trace_ocfs2_truncate_file_erro 461 trace_ocfs2_truncate_file_error(
464 (unsigned long long)le 462 (unsigned long long)le64_to_cpu(fe->i_size),
465 (unsigned long long)ne 463 (unsigned long long)new_i_size);
466 status = -EINVAL; 464 status = -EINVAL;
467 mlog_errno(status); 465 mlog_errno(status);
468 goto bail; 466 goto bail;
469 } 467 }
470 468
471 down_write(&OCFS2_I(inode)->ip_alloc_s 469 down_write(&OCFS2_I(inode)->ip_alloc_sem);
472 470
473 ocfs2_resv_discard(&osb->osb_la_resmap 471 ocfs2_resv_discard(&osb->osb_la_resmap,
474 &OCFS2_I(inode)->ip 472 &OCFS2_I(inode)->ip_la_data_resv);
475 473
476 /* 474 /*
477 * The inode lock forced other nodes t 475 * The inode lock forced other nodes to sync and drop their
478 * pages, which (correctly) happens ev 476 * pages, which (correctly) happens even if we have a truncate
479 * without allocation change - ocfs2 c 477 * without allocation change - ocfs2 cluster sizes can be much
480 * greater than page size, so we have 478 * greater than page size, so we have to truncate them
481 * anyway. 479 * anyway.
482 */ 480 */
>> 481 unmap_mapping_range(inode->i_mapping, new_i_size + PAGE_SIZE - 1, 0, 1);
>> 482 truncate_inode_pages(inode->i_mapping, new_i_size);
483 483
484 if (OCFS2_I(inode)->ip_dyn_features & 484 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 485 status = ocfs2_truncate_inline(inode, di_bh, new_i_size,
489 486 i_size_read(inode), 1);
490 if (status) 487 if (status)
491 mlog_errno(status); 488 mlog_errno(status);
492 489
493 goto bail_unlock_sem; 490 goto bail_unlock_sem;
494 } 491 }
495 492
496 /* alright, we're going to need to do 493 /* alright, we're going to need to do a full blown alloc size
497 * change. Orphan the inode so that re 494 * change. Orphan the inode so that recovery can complete the
498 * truncate if necessary. This does th 495 * truncate if necessary. This does the task of marking
499 * i_size. */ 496 * i_size. */
500 status = ocfs2_orphan_for_truncate(osb 497 status = ocfs2_orphan_for_truncate(osb, inode, di_bh, new_i_size);
501 if (status < 0) { 498 if (status < 0) {
502 mlog_errno(status); 499 mlog_errno(status);
503 goto bail_unlock_sem; 500 goto bail_unlock_sem;
504 } 501 }
505 502
506 unmap_mapping_range(inode->i_mapping, <<
507 truncate_inode_pages(inode->i_mapping, <<
508 <<
509 status = ocfs2_commit_truncate(osb, in 503 status = ocfs2_commit_truncate(osb, inode, di_bh);
510 if (status < 0) { 504 if (status < 0) {
511 mlog_errno(status); 505 mlog_errno(status);
512 goto bail_unlock_sem; 506 goto bail_unlock_sem;
513 } 507 }
514 508
515 /* TODO: orphan dir cleanup here. */ 509 /* TODO: orphan dir cleanup here. */
516 bail_unlock_sem: 510 bail_unlock_sem:
517 up_write(&OCFS2_I(inode)->ip_alloc_sem 511 up_write(&OCFS2_I(inode)->ip_alloc_sem);
518 512
519 bail: 513 bail:
520 if (!status && OCFS2_I(inode)->ip_clus 514 if (!status && OCFS2_I(inode)->ip_clusters == 0)
521 status = ocfs2_try_remove_refc 515 status = ocfs2_try_remove_refcount_tree(inode, di_bh);
522 516
523 return status; 517 return status;
524 } 518 }
525 519
526 /* 520 /*
527 * extend file allocation only here. 521 * extend file allocation only here.
528 * we'll update all the disk stuff, and oip->a 522 * we'll update all the disk stuff, and oip->alloc_size
529 * 523 *
530 * expect stuff to be locked, a transaction st 524 * expect stuff to be locked, a transaction started and enough data /
531 * metadata reservations in the contexts. 525 * metadata reservations in the contexts.
532 * 526 *
533 * Will return -EAGAIN, and a reason if a rest 527 * Will return -EAGAIN, and a reason if a restart is needed.
534 * If passed in, *reason will always be set, e 528 * If passed in, *reason will always be set, even in error.
535 */ 529 */
536 int ocfs2_add_inode_data(struct ocfs2_super *o 530 int ocfs2_add_inode_data(struct ocfs2_super *osb,
537 struct inode *inode, 531 struct inode *inode,
538 u32 *logical_offset, 532 u32 *logical_offset,
539 u32 clusters_to_add, 533 u32 clusters_to_add,
540 int mark_unwritten, 534 int mark_unwritten,
541 struct buffer_head *f 535 struct buffer_head *fe_bh,
542 handle_t *handle, 536 handle_t *handle,
543 struct ocfs2_alloc_co 537 struct ocfs2_alloc_context *data_ac,
544 struct ocfs2_alloc_co 538 struct ocfs2_alloc_context *meta_ac,
545 enum ocfs2_alloc_rest 539 enum ocfs2_alloc_restarted *reason_ret)
546 { 540 {
>> 541 int ret;
547 struct ocfs2_extent_tree et; 542 struct ocfs2_extent_tree et;
548 543
549 ocfs2_init_dinode_extent_tree(&et, INO 544 ocfs2_init_dinode_extent_tree(&et, INODE_CACHE(inode), fe_bh);
550 return ocfs2_add_clusters_in_btree(han !! 545 ret = ocfs2_add_clusters_in_btree(handle, &et, logical_offset,
551 clu !! 546 clusters_to_add, mark_unwritten,
552 dat !! 547 data_ac, meta_ac, reason_ret);
>> 548
>> 549 return ret;
553 } 550 }
554 551
555 static int ocfs2_extend_allocation(struct inod 552 static int ocfs2_extend_allocation(struct inode *inode, u32 logical_start,
556 u32 cluster 553 u32 clusters_to_add, int mark_unwritten)
557 { 554 {
558 int status = 0; 555 int status = 0;
559 int restart_func = 0; 556 int restart_func = 0;
560 int credits; 557 int credits;
561 u32 prev_clusters; 558 u32 prev_clusters;
562 struct buffer_head *bh = NULL; 559 struct buffer_head *bh = NULL;
563 struct ocfs2_dinode *fe = NULL; 560 struct ocfs2_dinode *fe = NULL;
564 handle_t *handle = NULL; 561 handle_t *handle = NULL;
565 struct ocfs2_alloc_context *data_ac = 562 struct ocfs2_alloc_context *data_ac = NULL;
566 struct ocfs2_alloc_context *meta_ac = 563 struct ocfs2_alloc_context *meta_ac = NULL;
567 enum ocfs2_alloc_restarted why = RESTA 564 enum ocfs2_alloc_restarted why = RESTART_NONE;
568 struct ocfs2_super *osb = OCFS2_SB(ino 565 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
569 struct ocfs2_extent_tree et; 566 struct ocfs2_extent_tree et;
570 int did_quota = 0; 567 int did_quota = 0;
571 568
572 /* 569 /*
573 * Unwritten extent only exists for fi 570 * Unwritten extent only exists for file systems which
574 * support holes. 571 * support holes.
575 */ 572 */
576 BUG_ON(mark_unwritten && !ocfs2_sparse 573 BUG_ON(mark_unwritten && !ocfs2_sparse_alloc(osb));
577 574
578 status = ocfs2_read_inode_block(inode, 575 status = ocfs2_read_inode_block(inode, &bh);
579 if (status < 0) { 576 if (status < 0) {
580 mlog_errno(status); 577 mlog_errno(status);
581 goto leave; 578 goto leave;
582 } 579 }
583 fe = (struct ocfs2_dinode *) bh->b_dat 580 fe = (struct ocfs2_dinode *) bh->b_data;
584 581
585 restart_all: 582 restart_all:
586 BUG_ON(le32_to_cpu(fe->i_clusters) != 583 BUG_ON(le32_to_cpu(fe->i_clusters) != OCFS2_I(inode)->ip_clusters);
587 584
588 ocfs2_init_dinode_extent_tree(&et, INO 585 ocfs2_init_dinode_extent_tree(&et, INODE_CACHE(inode), bh);
589 status = ocfs2_lock_allocators(inode, 586 status = ocfs2_lock_allocators(inode, &et, clusters_to_add, 0,
590 &data_a 587 &data_ac, &meta_ac);
591 if (status) { 588 if (status) {
592 mlog_errno(status); 589 mlog_errno(status);
593 goto leave; 590 goto leave;
594 } 591 }
595 592
596 credits = ocfs2_calc_extend_credits(os 593 credits = ocfs2_calc_extend_credits(osb->sb, &fe->id2.i_list);
597 handle = ocfs2_start_trans(osb, credit 594 handle = ocfs2_start_trans(osb, credits);
598 if (IS_ERR(handle)) { 595 if (IS_ERR(handle)) {
599 status = PTR_ERR(handle); 596 status = PTR_ERR(handle);
600 handle = NULL; 597 handle = NULL;
601 mlog_errno(status); 598 mlog_errno(status);
602 goto leave; 599 goto leave;
603 } 600 }
604 601
605 restarted_transaction: 602 restarted_transaction:
606 trace_ocfs2_extend_allocation( 603 trace_ocfs2_extend_allocation(
607 (unsigned long long)OCFS2_I(in 604 (unsigned long long)OCFS2_I(inode)->ip_blkno,
608 (unsigned long long)i_size_rea 605 (unsigned long long)i_size_read(inode),
609 le32_to_cpu(fe->i_clusters), c 606 le32_to_cpu(fe->i_clusters), clusters_to_add,
610 why, restart_func); 607 why, restart_func);
611 608
612 status = dquot_alloc_space_nodirty(ino 609 status = dquot_alloc_space_nodirty(inode,
613 ocfs2_clusters_to_byte 610 ocfs2_clusters_to_bytes(osb->sb, clusters_to_add));
614 if (status) 611 if (status)
615 goto leave; 612 goto leave;
616 did_quota = 1; 613 did_quota = 1;
617 614
618 /* reserve a write to the file entry e 615 /* reserve a write to the file entry early on - that we if we
619 * run out of credits in the allocatio 616 * run out of credits in the allocation path, we can still
620 * update i_size. */ 617 * update i_size. */
621 status = ocfs2_journal_access_di(handl 618 status = ocfs2_journal_access_di(handle, INODE_CACHE(inode), bh,
622 OCFS2 619 OCFS2_JOURNAL_ACCESS_WRITE);
623 if (status < 0) { 620 if (status < 0) {
624 mlog_errno(status); 621 mlog_errno(status);
625 goto leave; 622 goto leave;
626 } 623 }
627 624
628 prev_clusters = OCFS2_I(inode)->ip_clu 625 prev_clusters = OCFS2_I(inode)->ip_clusters;
629 626
630 status = ocfs2_add_inode_data(osb, 627 status = ocfs2_add_inode_data(osb,
631 inode, 628 inode,
632 &logical 629 &logical_start,
633 clusters 630 clusters_to_add,
634 mark_unw 631 mark_unwritten,
635 bh, 632 bh,
636 handle, 633 handle,
637 data_ac, 634 data_ac,
638 meta_ac, 635 meta_ac,
639 &why); 636 &why);
640 if ((status < 0) && (status != -EAGAIN 637 if ((status < 0) && (status != -EAGAIN)) {
641 if (status != -ENOSPC) 638 if (status != -ENOSPC)
642 mlog_errno(status); 639 mlog_errno(status);
643 goto leave; 640 goto leave;
644 } 641 }
645 ocfs2_update_inode_fsync_trans(handle, 642 ocfs2_update_inode_fsync_trans(handle, inode, 1);
646 ocfs2_journal_dirty(handle, bh); 643 ocfs2_journal_dirty(handle, bh);
647 644
648 spin_lock(&OCFS2_I(inode)->ip_lock); 645 spin_lock(&OCFS2_I(inode)->ip_lock);
649 clusters_to_add -= (OCFS2_I(inode)->ip 646 clusters_to_add -= (OCFS2_I(inode)->ip_clusters - prev_clusters);
650 spin_unlock(&OCFS2_I(inode)->ip_lock); 647 spin_unlock(&OCFS2_I(inode)->ip_lock);
651 /* Release unused quota reservation */ 648 /* Release unused quota reservation */
652 dquot_free_space(inode, 649 dquot_free_space(inode,
653 ocfs2_clusters_to_byte 650 ocfs2_clusters_to_bytes(osb->sb, clusters_to_add));
654 did_quota = 0; 651 did_quota = 0;
655 652
656 if (why != RESTART_NONE && clusters_to 653 if (why != RESTART_NONE && clusters_to_add) {
657 if (why == RESTART_META) { 654 if (why == RESTART_META) {
658 restart_func = 1; 655 restart_func = 1;
659 status = 0; 656 status = 0;
660 } else { 657 } else {
661 BUG_ON(why != RESTART_ 658 BUG_ON(why != RESTART_TRANS);
662 659
663 status = ocfs2_allocat 660 status = ocfs2_allocate_extend_trans(handle, 1);
664 if (status < 0) { 661 if (status < 0) {
665 /* handle stil 662 /* handle still has to be committed at
666 * this point. 663 * this point. */
667 status = -ENOM 664 status = -ENOMEM;
668 mlog_errno(sta 665 mlog_errno(status);
669 goto leave; 666 goto leave;
670 } 667 }
671 goto restarted_transac 668 goto restarted_transaction;
672 } 669 }
673 } 670 }
674 671
675 trace_ocfs2_extend_allocation_end(OCFS 672 trace_ocfs2_extend_allocation_end(OCFS2_I(inode)->ip_blkno,
676 le32_to_cpu(fe->i_clusters), 673 le32_to_cpu(fe->i_clusters),
677 (unsigned long long)le64_to_cpu(f 674 (unsigned long long)le64_to_cpu(fe->i_size),
678 OCFS2_I(inode)->ip_clusters, 675 OCFS2_I(inode)->ip_clusters,
679 (unsigned long long)i_size_read(i 676 (unsigned long long)i_size_read(inode));
680 677
681 leave: 678 leave:
682 if (status < 0 && did_quota) 679 if (status < 0 && did_quota)
683 dquot_free_space(inode, 680 dquot_free_space(inode,
684 ocfs2_clusters_to_byte 681 ocfs2_clusters_to_bytes(osb->sb, clusters_to_add));
685 if (handle) { 682 if (handle) {
686 ocfs2_commit_trans(osb, handle 683 ocfs2_commit_trans(osb, handle);
687 handle = NULL; 684 handle = NULL;
688 } 685 }
689 if (data_ac) { 686 if (data_ac) {
690 ocfs2_free_alloc_context(data_ 687 ocfs2_free_alloc_context(data_ac);
691 data_ac = NULL; 688 data_ac = NULL;
692 } 689 }
693 if (meta_ac) { 690 if (meta_ac) {
694 ocfs2_free_alloc_context(meta_ 691 ocfs2_free_alloc_context(meta_ac);
695 meta_ac = NULL; 692 meta_ac = NULL;
696 } 693 }
697 if ((!status) && restart_func) { 694 if ((!status) && restart_func) {
698 restart_func = 0; 695 restart_func = 0;
699 goto restart_all; 696 goto restart_all;
700 } 697 }
701 brelse(bh); 698 brelse(bh);
702 bh = NULL; 699 bh = NULL;
703 700
704 return status; 701 return status;
705 } 702 }
706 703
707 /* 704 /*
708 * While a write will already be ordering the 705 * While a write will already be ordering the data, a truncate will not.
709 * Thus, we need to explicitly order the zeroe 706 * Thus, we need to explicitly order the zeroed pages.
710 */ 707 */
711 static handle_t *ocfs2_zero_start_ordered_tran 708 static handle_t *ocfs2_zero_start_ordered_transaction(struct inode *inode,
712 !! 709 struct buffer_head *di_bh)
713 <<
714 <<
715 { 710 {
716 struct ocfs2_super *osb = OCFS2_SB(ino 711 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
717 handle_t *handle = NULL; 712 handle_t *handle = NULL;
718 int ret = 0; 713 int ret = 0;
719 714
720 if (!ocfs2_should_order_data(inode)) 715 if (!ocfs2_should_order_data(inode))
721 goto out; 716 goto out;
722 717
723 handle = ocfs2_start_trans(osb, OCFS2_ 718 handle = ocfs2_start_trans(osb, OCFS2_INODE_UPDATE_CREDITS);
724 if (IS_ERR(handle)) { 719 if (IS_ERR(handle)) {
725 ret = -ENOMEM; 720 ret = -ENOMEM;
726 mlog_errno(ret); 721 mlog_errno(ret);
727 goto out; 722 goto out;
728 } 723 }
729 724
730 ret = ocfs2_jbd2_inode_add_write(handl !! 725 ret = ocfs2_jbd2_file_inode(handle, inode);
731 if (ret < 0) { 726 if (ret < 0) {
732 mlog_errno(ret); 727 mlog_errno(ret);
733 goto out; 728 goto out;
734 } 729 }
735 730
736 ret = ocfs2_journal_access_di(handle, 731 ret = ocfs2_journal_access_di(handle, INODE_CACHE(inode), di_bh,
737 OCFS2_JO 732 OCFS2_JOURNAL_ACCESS_WRITE);
738 if (ret) 733 if (ret)
739 mlog_errno(ret); 734 mlog_errno(ret);
740 ocfs2_update_inode_fsync_trans(handle, 735 ocfs2_update_inode_fsync_trans(handle, inode, 1);
741 736
742 out: 737 out:
743 if (ret) { 738 if (ret) {
744 if (!IS_ERR(handle)) 739 if (!IS_ERR(handle))
745 ocfs2_commit_trans(osb 740 ocfs2_commit_trans(osb, handle);
746 handle = ERR_PTR(ret); 741 handle = ERR_PTR(ret);
747 } 742 }
748 return handle; 743 return handle;
749 } 744 }
750 745
751 /* Some parts of this taken from generic_cont_ 746 /* Some parts of this taken from generic_cont_expand, which turned out
752 * to be too fragile to do exactly what we nee 747 * to be too fragile to do exactly what we need without us having to
753 * worry about recursive locking in ->write_be 748 * worry about recursive locking in ->write_begin() and ->write_end(). */
754 static int ocfs2_write_zero_page(struct inode 749 static int ocfs2_write_zero_page(struct inode *inode, u64 abs_from,
755 u64 abs_to, s 750 u64 abs_to, struct buffer_head *di_bh)
756 { 751 {
757 struct address_space *mapping = inode- 752 struct address_space *mapping = inode->i_mapping;
758 struct folio *folio; !! 753 struct page *page;
759 unsigned long index = abs_from >> PAGE 754 unsigned long index = abs_from >> PAGE_SHIFT;
760 handle_t *handle; 755 handle_t *handle;
761 int ret = 0; 756 int ret = 0;
762 unsigned zero_from, zero_to, block_sta 757 unsigned zero_from, zero_to, block_start, block_end;
763 struct ocfs2_dinode *di = (struct ocfs 758 struct ocfs2_dinode *di = (struct ocfs2_dinode *)di_bh->b_data;
764 759
765 BUG_ON(abs_from >= abs_to); 760 BUG_ON(abs_from >= abs_to);
766 BUG_ON(abs_to > (((u64)index + 1) << P 761 BUG_ON(abs_to > (((u64)index + 1) << PAGE_SHIFT));
767 BUG_ON(abs_from & (inode->i_blkbits - 762 BUG_ON(abs_from & (inode->i_blkbits - 1));
768 763
769 handle = ocfs2_zero_start_ordered_tran !! 764 handle = ocfs2_zero_start_ordered_transaction(inode, di_bh);
770 <<
771 <<
772 if (IS_ERR(handle)) { 765 if (IS_ERR(handle)) {
773 ret = PTR_ERR(handle); 766 ret = PTR_ERR(handle);
774 goto out; 767 goto out;
775 } 768 }
776 769
777 folio = __filemap_get_folio(mapping, i !! 770 page = find_or_create_page(mapping, index, GFP_NOFS);
778 FGP_LOCK | FGP_ACCESSE !! 771 if (!page) {
779 if (IS_ERR(folio)) { !! 772 ret = -ENOMEM;
780 ret = PTR_ERR(folio); <<
781 mlog_errno(ret); 773 mlog_errno(ret);
782 goto out_commit_trans; 774 goto out_commit_trans;
783 } 775 }
784 776
785 /* Get the offsets within the page tha 777 /* Get the offsets within the page that we want to zero */
786 zero_from = abs_from & (PAGE_SIZE - 1) 778 zero_from = abs_from & (PAGE_SIZE - 1);
787 zero_to = abs_to & (PAGE_SIZE - 1); 779 zero_to = abs_to & (PAGE_SIZE - 1);
788 if (!zero_to) 780 if (!zero_to)
789 zero_to = PAGE_SIZE; 781 zero_to = PAGE_SIZE;
790 782
791 trace_ocfs2_write_zero_page( 783 trace_ocfs2_write_zero_page(
792 (unsigned long long)OC 784 (unsigned long long)OCFS2_I(inode)->ip_blkno,
793 (unsigned long long)ab 785 (unsigned long long)abs_from,
794 (unsigned long long)ab 786 (unsigned long long)abs_to,
795 index, zero_from, zero 787 index, zero_from, zero_to);
796 788
797 /* We know that zero_from is block ali 789 /* We know that zero_from is block aligned */
798 for (block_start = zero_from; block_st 790 for (block_start = zero_from; block_start < zero_to;
799 block_start = block_end) { 791 block_start = block_end) {
800 block_end = block_start + i_bl 792 block_end = block_start + i_blocksize(inode);
801 793
802 /* 794 /*
803 * block_start is block-aligne 795 * block_start is block-aligned. Bump it by one to force
804 * __block_write_begin and blo 796 * __block_write_begin and block_commit_write to zero the
805 * whole block. 797 * whole block.
806 */ 798 */
807 ret = __block_write_begin(foli !! 799 ret = __block_write_begin(page, block_start + 1, 0,
808 ocfs 800 ocfs2_get_block);
809 if (ret < 0) { 801 if (ret < 0) {
810 mlog_errno(ret); 802 mlog_errno(ret);
811 goto out_unlock; 803 goto out_unlock;
812 } 804 }
813 805
814 806
815 /* must not update i_size! */ 807 /* must not update i_size! */
816 block_commit_write(&folio->pag !! 808 ret = block_commit_write(page, block_start + 1,
>> 809 block_start + 1);
>> 810 if (ret < 0)
>> 811 mlog_errno(ret);
>> 812 else
>> 813 ret = 0;
817 } 814 }
818 815
819 /* 816 /*
820 * fs-writeback will release the dirty 817 * fs-writeback will release the dirty pages without page lock
821 * whose offset are over inode size, t 818 * whose offset are over inode size, the release happens at
822 * block_write_full_folio(). !! 819 * block_write_full_page().
823 */ 820 */
824 i_size_write(inode, abs_to); 821 i_size_write(inode, abs_to);
825 inode->i_blocks = ocfs2_inode_sector_c 822 inode->i_blocks = ocfs2_inode_sector_count(inode);
826 di->i_size = cpu_to_le64((u64)i_size_r 823 di->i_size = cpu_to_le64((u64)i_size_read(inode));
827 inode_set_mtime_to_ts(inode, inode_set !! 824 inode->i_mtime = inode->i_ctime = current_time(inode);
828 di->i_mtime = di->i_ctime = cpu_to_le6 !! 825 di->i_mtime = di->i_ctime = cpu_to_le64(inode->i_mtime.tv_sec);
829 di->i_ctime_nsec = cpu_to_le32(inode_g !! 826 di->i_ctime_nsec = cpu_to_le32(inode->i_mtime.tv_nsec);
830 di->i_mtime_nsec = di->i_ctime_nsec; 827 di->i_mtime_nsec = di->i_ctime_nsec;
831 if (handle) { 828 if (handle) {
832 ocfs2_journal_dirty(handle, di 829 ocfs2_journal_dirty(handle, di_bh);
833 ocfs2_update_inode_fsync_trans 830 ocfs2_update_inode_fsync_trans(handle, inode, 1);
834 } 831 }
835 832
836 out_unlock: 833 out_unlock:
837 folio_unlock(folio); !! 834 unlock_page(page);
838 folio_put(folio); !! 835 put_page(page);
839 out_commit_trans: 836 out_commit_trans:
840 if (handle) 837 if (handle)
841 ocfs2_commit_trans(OCFS2_SB(in 838 ocfs2_commit_trans(OCFS2_SB(inode->i_sb), handle);
842 out: 839 out:
843 return ret; 840 return ret;
844 } 841 }
845 842
846 /* 843 /*
847 * Find the next range to zero. We do this in 844 * Find the next range to zero. We do this in terms of bytes because
848 * that's what ocfs2_zero_extend() wants, and 845 * that's what ocfs2_zero_extend() wants, and it is dealing with the
849 * pagecache. We may return multiple extents. 846 * pagecache. We may return multiple extents.
850 * 847 *
851 * zero_start and zero_end are ocfs2_zero_exte 848 * zero_start and zero_end are ocfs2_zero_extend()s current idea of what
852 * needs to be zeroed. range_start and range_ 849 * needs to be zeroed. range_start and range_end return the next zeroing
853 * range. A subsequent call should pass the p 850 * range. A subsequent call should pass the previous range_end as its
854 * zero_start. If range_end is 0, there's not 851 * zero_start. If range_end is 0, there's nothing to do.
855 * 852 *
856 * Unwritten extents are skipped over. Refcou 853 * Unwritten extents are skipped over. Refcounted extents are CoWd.
857 */ 854 */
858 static int ocfs2_zero_extend_get_range(struct 855 static int ocfs2_zero_extend_get_range(struct inode *inode,
859 struct 856 struct buffer_head *di_bh,
860 u64 zer 857 u64 zero_start, u64 zero_end,
861 u64 *ra 858 u64 *range_start, u64 *range_end)
862 { 859 {
863 int rc = 0, needs_cow = 0; 860 int rc = 0, needs_cow = 0;
864 u32 p_cpos, zero_clusters = 0; 861 u32 p_cpos, zero_clusters = 0;
865 u32 zero_cpos = 862 u32 zero_cpos =
866 zero_start >> OCFS2_SB(inode-> 863 zero_start >> OCFS2_SB(inode->i_sb)->s_clustersize_bits;
867 u32 last_cpos = ocfs2_clusters_for_byt 864 u32 last_cpos = ocfs2_clusters_for_bytes(inode->i_sb, zero_end);
868 unsigned int num_clusters = 0; 865 unsigned int num_clusters = 0;
869 unsigned int ext_flags = 0; 866 unsigned int ext_flags = 0;
870 867
871 while (zero_cpos < last_cpos) { 868 while (zero_cpos < last_cpos) {
872 rc = ocfs2_get_clusters(inode, 869 rc = ocfs2_get_clusters(inode, zero_cpos, &p_cpos,
873 &num_c 870 &num_clusters, &ext_flags);
874 if (rc) { 871 if (rc) {
875 mlog_errno(rc); 872 mlog_errno(rc);
876 goto out; 873 goto out;
877 } 874 }
878 875
879 if (p_cpos && !(ext_flags & OC 876 if (p_cpos && !(ext_flags & OCFS2_EXT_UNWRITTEN)) {
880 zero_clusters = num_cl 877 zero_clusters = num_clusters;
881 if (ext_flags & OCFS2_ 878 if (ext_flags & OCFS2_EXT_REFCOUNTED)
882 needs_cow = 1; 879 needs_cow = 1;
883 break; 880 break;
884 } 881 }
885 882
886 zero_cpos += num_clusters; 883 zero_cpos += num_clusters;
887 } 884 }
888 if (!zero_clusters) { 885 if (!zero_clusters) {
889 *range_end = 0; 886 *range_end = 0;
890 goto out; 887 goto out;
891 } 888 }
892 889
893 while ((zero_cpos + zero_clusters) < l 890 while ((zero_cpos + zero_clusters) < last_cpos) {
894 rc = ocfs2_get_clusters(inode, 891 rc = ocfs2_get_clusters(inode, zero_cpos + zero_clusters,
895 &p_cpo 892 &p_cpos, &num_clusters,
896 &ext_f 893 &ext_flags);
897 if (rc) { 894 if (rc) {
898 mlog_errno(rc); 895 mlog_errno(rc);
899 goto out; 896 goto out;
900 } 897 }
901 898
902 if (!p_cpos || (ext_flags & OC 899 if (!p_cpos || (ext_flags & OCFS2_EXT_UNWRITTEN))
903 break; 900 break;
904 if (ext_flags & OCFS2_EXT_REFC 901 if (ext_flags & OCFS2_EXT_REFCOUNTED)
905 needs_cow = 1; 902 needs_cow = 1;
906 zero_clusters += num_clusters; 903 zero_clusters += num_clusters;
907 } 904 }
908 if ((zero_cpos + zero_clusters) > last 905 if ((zero_cpos + zero_clusters) > last_cpos)
909 zero_clusters = last_cpos - ze 906 zero_clusters = last_cpos - zero_cpos;
910 907
911 if (needs_cow) { 908 if (needs_cow) {
912 rc = ocfs2_refcount_cow(inode, 909 rc = ocfs2_refcount_cow(inode, di_bh, zero_cpos,
913 zero_c 910 zero_clusters, UINT_MAX);
914 if (rc) { 911 if (rc) {
915 mlog_errno(rc); 912 mlog_errno(rc);
916 goto out; 913 goto out;
917 } 914 }
918 } 915 }
919 916
920 *range_start = ocfs2_clusters_to_bytes 917 *range_start = ocfs2_clusters_to_bytes(inode->i_sb, zero_cpos);
921 *range_end = ocfs2_clusters_to_bytes(i 918 *range_end = ocfs2_clusters_to_bytes(inode->i_sb,
922 z 919 zero_cpos + zero_clusters);
923 920
924 out: 921 out:
925 return rc; 922 return rc;
926 } 923 }
927 924
928 /* 925 /*
929 * Zero one range returned from ocfs2_zero_ext 926 * Zero one range returned from ocfs2_zero_extend_get_range(). The caller
930 * has made sure that the entire range needs z 927 * has made sure that the entire range needs zeroing.
931 */ 928 */
932 static int ocfs2_zero_extend_range(struct inod 929 static int ocfs2_zero_extend_range(struct inode *inode, u64 range_start,
933 u64 range_e 930 u64 range_end, struct buffer_head *di_bh)
934 { 931 {
935 int rc = 0; 932 int rc = 0;
936 u64 next_pos; 933 u64 next_pos;
937 u64 zero_pos = range_start; 934 u64 zero_pos = range_start;
938 935
939 trace_ocfs2_zero_extend_range( 936 trace_ocfs2_zero_extend_range(
940 (unsigned long long)OC 937 (unsigned long long)OCFS2_I(inode)->ip_blkno,
941 (unsigned long long)ra 938 (unsigned long long)range_start,
942 (unsigned long long)ra 939 (unsigned long long)range_end);
943 BUG_ON(range_start >= range_end); 940 BUG_ON(range_start >= range_end);
944 941
945 while (zero_pos < range_end) { 942 while (zero_pos < range_end) {
946 next_pos = (zero_pos & PAGE_MA 943 next_pos = (zero_pos & PAGE_MASK) + PAGE_SIZE;
947 if (next_pos > range_end) 944 if (next_pos > range_end)
948 next_pos = range_end; 945 next_pos = range_end;
949 rc = ocfs2_write_zero_page(ino 946 rc = ocfs2_write_zero_page(inode, zero_pos, next_pos, di_bh);
950 if (rc < 0) { 947 if (rc < 0) {
951 mlog_errno(rc); 948 mlog_errno(rc);
952 break; 949 break;
953 } 950 }
954 zero_pos = next_pos; 951 zero_pos = next_pos;
955 952
956 /* 953 /*
957 * Very large extends have the 954 * Very large extends have the potential to lock up
958 * the cpu for extended period 955 * the cpu for extended periods of time.
959 */ 956 */
960 cond_resched(); 957 cond_resched();
961 } 958 }
962 959
963 return rc; 960 return rc;
964 } 961 }
965 962
966 int ocfs2_zero_extend(struct inode *inode, str 963 int ocfs2_zero_extend(struct inode *inode, struct buffer_head *di_bh,
967 loff_t zero_to_size) 964 loff_t zero_to_size)
968 { 965 {
969 int ret = 0; 966 int ret = 0;
970 u64 zero_start, range_start = 0, range 967 u64 zero_start, range_start = 0, range_end = 0;
971 struct super_block *sb = inode->i_sb; 968 struct super_block *sb = inode->i_sb;
972 969
973 zero_start = ocfs2_align_bytes_to_bloc 970 zero_start = ocfs2_align_bytes_to_blocks(sb, i_size_read(inode));
974 trace_ocfs2_zero_extend((unsigned long 971 trace_ocfs2_zero_extend((unsigned long long)OCFS2_I(inode)->ip_blkno,
975 (unsigned long 972 (unsigned long long)zero_start,
976 (unsigned long 973 (unsigned long long)i_size_read(inode));
977 while (zero_start < zero_to_size) { 974 while (zero_start < zero_to_size) {
978 ret = ocfs2_zero_extend_get_ra 975 ret = ocfs2_zero_extend_get_range(inode, di_bh, zero_start,
979 976 zero_to_size,
980 977 &range_start,
981 978 &range_end);
982 if (ret) { 979 if (ret) {
983 mlog_errno(ret); 980 mlog_errno(ret);
984 break; 981 break;
985 } 982 }
986 if (!range_end) 983 if (!range_end)
987 break; 984 break;
988 /* Trim the ends */ 985 /* Trim the ends */
989 if (range_start < zero_start) 986 if (range_start < zero_start)
990 range_start = zero_sta 987 range_start = zero_start;
991 if (range_end > zero_to_size) 988 if (range_end > zero_to_size)
992 range_end = zero_to_si 989 range_end = zero_to_size;
993 990
994 ret = ocfs2_zero_extend_range( 991 ret = ocfs2_zero_extend_range(inode, range_start,
995 992 range_end, di_bh);
996 if (ret) { 993 if (ret) {
997 mlog_errno(ret); 994 mlog_errno(ret);
998 break; 995 break;
999 } 996 }
1000 zero_start = range_end; 997 zero_start = range_end;
1001 } 998 }
1002 999
1003 return ret; 1000 return ret;
1004 } 1001 }
1005 1002
1006 int ocfs2_extend_no_holes(struct inode *inode 1003 int ocfs2_extend_no_holes(struct inode *inode, struct buffer_head *di_bh,
1007 u64 new_i_size, u64 1004 u64 new_i_size, u64 zero_to)
1008 { 1005 {
1009 int ret; 1006 int ret;
1010 u32 clusters_to_add; 1007 u32 clusters_to_add;
1011 struct ocfs2_inode_info *oi = OCFS2_I 1008 struct ocfs2_inode_info *oi = OCFS2_I(inode);
1012 1009
1013 /* 1010 /*
1014 * Only quota files call this without 1011 * Only quota files call this without a bh, and they can't be
1015 * refcounted. 1012 * refcounted.
1016 */ 1013 */
1017 BUG_ON(!di_bh && ocfs2_is_refcount_in 1014 BUG_ON(!di_bh && ocfs2_is_refcount_inode(inode));
1018 BUG_ON(!di_bh && !(oi->ip_flags & OCF 1015 BUG_ON(!di_bh && !(oi->ip_flags & OCFS2_INODE_SYSTEM_FILE));
1019 1016
1020 clusters_to_add = ocfs2_clusters_for_ 1017 clusters_to_add = ocfs2_clusters_for_bytes(inode->i_sb, new_i_size);
1021 if (clusters_to_add < oi->ip_clusters 1018 if (clusters_to_add < oi->ip_clusters)
1022 clusters_to_add = 0; 1019 clusters_to_add = 0;
1023 else 1020 else
1024 clusters_to_add -= oi->ip_clu 1021 clusters_to_add -= oi->ip_clusters;
1025 1022
1026 if (clusters_to_add) { 1023 if (clusters_to_add) {
1027 ret = ocfs2_extend_allocation 1024 ret = ocfs2_extend_allocation(inode, oi->ip_clusters,
1028 1025 clusters_to_add, 0);
1029 if (ret) { 1026 if (ret) {
1030 mlog_errno(ret); 1027 mlog_errno(ret);
1031 goto out; 1028 goto out;
1032 } 1029 }
1033 } 1030 }
1034 1031
1035 /* 1032 /*
1036 * Call this even if we don't add any 1033 * Call this even if we don't add any clusters to the tree. We
1037 * still need to zero the area betwee 1034 * still need to zero the area between the old i_size and the
1038 * new i_size. 1035 * new i_size.
1039 */ 1036 */
1040 ret = ocfs2_zero_extend(inode, di_bh, 1037 ret = ocfs2_zero_extend(inode, di_bh, zero_to);
1041 if (ret < 0) 1038 if (ret < 0)
1042 mlog_errno(ret); 1039 mlog_errno(ret);
1043 1040
1044 out: 1041 out:
1045 return ret; 1042 return ret;
1046 } 1043 }
1047 1044
1048 static int ocfs2_extend_file(struct inode *in 1045 static int ocfs2_extend_file(struct inode *inode,
1049 struct buffer_he 1046 struct buffer_head *di_bh,
1050 u64 new_i_size) 1047 u64 new_i_size)
1051 { 1048 {
1052 int ret = 0; 1049 int ret = 0;
1053 struct ocfs2_inode_info *oi = OCFS2_I 1050 struct ocfs2_inode_info *oi = OCFS2_I(inode);
1054 1051
1055 BUG_ON(!di_bh); 1052 BUG_ON(!di_bh);
1056 1053
1057 /* setattr sometimes calls us like th 1054 /* setattr sometimes calls us like this. */
1058 if (new_i_size == 0) 1055 if (new_i_size == 0)
1059 goto out; 1056 goto out;
1060 1057
1061 if (i_size_read(inode) == new_i_size) 1058 if (i_size_read(inode) == new_i_size)
1062 goto out; 1059 goto out;
1063 BUG_ON(new_i_size < i_size_read(inode 1060 BUG_ON(new_i_size < i_size_read(inode));
1064 1061
1065 /* 1062 /*
1066 * The alloc sem blocks people in rea 1063 * The alloc sem blocks people in read/write from reading our
1067 * allocation until we're done changi 1064 * allocation until we're done changing it. We depend on
1068 * i_rwsem to block other extend/trun !! 1065 * i_mutex to block other extend/truncate calls while we're
1069 * here. We even have to hold it for 1066 * here. We even have to hold it for sparse files because there
1070 * might be some tail zeroing. 1067 * might be some tail zeroing.
1071 */ 1068 */
1072 down_write(&oi->ip_alloc_sem); 1069 down_write(&oi->ip_alloc_sem);
1073 1070
1074 if (oi->ip_dyn_features & OCFS2_INLIN 1071 if (oi->ip_dyn_features & OCFS2_INLINE_DATA_FL) {
1075 /* 1072 /*
1076 * We can optimize small exte 1073 * We can optimize small extends by keeping the inodes
1077 * inline data. 1074 * inline data.
1078 */ 1075 */
1079 if (ocfs2_size_fits_inline_da 1076 if (ocfs2_size_fits_inline_data(di_bh, new_i_size)) {
1080 up_write(&oi->ip_allo 1077 up_write(&oi->ip_alloc_sem);
1081 goto out_update_size; 1078 goto out_update_size;
1082 } 1079 }
1083 1080
1084 ret = ocfs2_convert_inline_da 1081 ret = ocfs2_convert_inline_data_to_extents(inode, di_bh);
1085 if (ret) { 1082 if (ret) {
1086 up_write(&oi->ip_allo 1083 up_write(&oi->ip_alloc_sem);
1087 mlog_errno(ret); 1084 mlog_errno(ret);
1088 goto out; 1085 goto out;
1089 } 1086 }
1090 } 1087 }
1091 1088
1092 if (ocfs2_sparse_alloc(OCFS2_SB(inode 1089 if (ocfs2_sparse_alloc(OCFS2_SB(inode->i_sb)))
1093 ret = ocfs2_zero_extend(inode 1090 ret = ocfs2_zero_extend(inode, di_bh, new_i_size);
1094 else 1091 else
1095 ret = ocfs2_extend_no_holes(i 1092 ret = ocfs2_extend_no_holes(inode, di_bh, new_i_size,
1096 n 1093 new_i_size);
1097 1094
1098 up_write(&oi->ip_alloc_sem); 1095 up_write(&oi->ip_alloc_sem);
1099 1096
1100 if (ret < 0) { 1097 if (ret < 0) {
1101 mlog_errno(ret); 1098 mlog_errno(ret);
1102 goto out; 1099 goto out;
1103 } 1100 }
1104 1101
1105 out_update_size: 1102 out_update_size:
1106 ret = ocfs2_simple_size_update(inode, 1103 ret = ocfs2_simple_size_update(inode, di_bh, new_i_size);
1107 if (ret < 0) 1104 if (ret < 0)
1108 mlog_errno(ret); 1105 mlog_errno(ret);
1109 1106
1110 out: 1107 out:
1111 return ret; 1108 return ret;
1112 } 1109 }
1113 1110
1114 int ocfs2_setattr(struct mnt_idmap *idmap, st !! 1111 int ocfs2_setattr(struct dentry *dentry, struct iattr *attr)
1115 struct iattr *attr) <<
1116 { 1112 {
1117 int status = 0, size_change; 1113 int status = 0, size_change;
1118 int inode_locked = 0; 1114 int inode_locked = 0;
1119 struct inode *inode = d_inode(dentry) 1115 struct inode *inode = d_inode(dentry);
1120 struct super_block *sb = inode->i_sb; 1116 struct super_block *sb = inode->i_sb;
1121 struct ocfs2_super *osb = OCFS2_SB(sb 1117 struct ocfs2_super *osb = OCFS2_SB(sb);
1122 struct buffer_head *bh = NULL; 1118 struct buffer_head *bh = NULL;
1123 handle_t *handle = NULL; 1119 handle_t *handle = NULL;
1124 struct dquot *transfer_to[MAXQUOTAS] 1120 struct dquot *transfer_to[MAXQUOTAS] = { };
1125 int qtype; 1121 int qtype;
1126 int had_lock; 1122 int had_lock;
1127 struct ocfs2_lock_holder oh; 1123 struct ocfs2_lock_holder oh;
1128 1124
1129 trace_ocfs2_setattr(inode, dentry, 1125 trace_ocfs2_setattr(inode, dentry,
1130 (unsigned long lo 1126 (unsigned long long)OCFS2_I(inode)->ip_blkno,
1131 dentry->d_name.le 1127 dentry->d_name.len, dentry->d_name.name,
1132 attr->ia_valid, !! 1128 attr->ia_valid, attr->ia_mode,
1133 attr->ia_vali !! 1129 from_kuid(&init_user_ns, attr->ia_uid),
1134 attr->ia_vali !! 1130 from_kgid(&init_user_ns, attr->ia_gid));
1135 from_ <<
1136 attr->ia_vali <<
1137 from_ <<
1138 1131
1139 /* ensuring we don't even attempt to 1132 /* ensuring we don't even attempt to truncate a symlink */
1140 if (S_ISLNK(inode->i_mode)) 1133 if (S_ISLNK(inode->i_mode))
1141 attr->ia_valid &= ~ATTR_SIZE; 1134 attr->ia_valid &= ~ATTR_SIZE;
1142 1135
1143 #define OCFS2_VALID_ATTRS (ATTR_ATIME | ATTR_ 1136 #define OCFS2_VALID_ATTRS (ATTR_ATIME | ATTR_MTIME | ATTR_CTIME | ATTR_SIZE \
1144 | ATTR_GID | ATTR_ 1137 | ATTR_GID | ATTR_UID | ATTR_MODE)
1145 if (!(attr->ia_valid & OCFS2_VALID_AT 1138 if (!(attr->ia_valid & OCFS2_VALID_ATTRS))
1146 return 0; 1139 return 0;
1147 1140
1148 status = setattr_prepare(&nop_mnt_idm !! 1141 status = setattr_prepare(dentry, attr);
1149 if (status) 1142 if (status)
1150 return status; 1143 return status;
1151 1144
1152 if (is_quota_modification(&nop_mnt_id !! 1145 if (is_quota_modification(inode, attr)) {
1153 status = dquot_initialize(ino 1146 status = dquot_initialize(inode);
1154 if (status) 1147 if (status)
1155 return status; 1148 return status;
1156 } 1149 }
1157 size_change = S_ISREG(inode->i_mode) 1150 size_change = S_ISREG(inode->i_mode) && attr->ia_valid & ATTR_SIZE;
1158 if (size_change) { 1151 if (size_change) {
1159 /* 1152 /*
1160 * Here we should wait dio to 1153 * Here we should wait dio to finish before inode lock
1161 * to avoid a deadlock betwee 1154 * to avoid a deadlock between ocfs2_setattr() and
1162 * ocfs2_dio_end_io_write() 1155 * ocfs2_dio_end_io_write()
1163 */ 1156 */
1164 inode_dio_wait(inode); 1157 inode_dio_wait(inode);
1165 1158
1166 status = ocfs2_rw_lock(inode, 1159 status = ocfs2_rw_lock(inode, 1);
1167 if (status < 0) { 1160 if (status < 0) {
1168 mlog_errno(status); 1161 mlog_errno(status);
1169 goto bail; 1162 goto bail;
1170 } 1163 }
1171 } 1164 }
1172 1165
1173 had_lock = ocfs2_inode_lock_tracker(i 1166 had_lock = ocfs2_inode_lock_tracker(inode, &bh, 1, &oh);
1174 if (had_lock < 0) { 1167 if (had_lock < 0) {
1175 status = had_lock; 1168 status = had_lock;
1176 goto bail_unlock_rw; 1169 goto bail_unlock_rw;
1177 } else if (had_lock) { 1170 } else if (had_lock) {
1178 /* 1171 /*
1179 * As far as we know, ocfs2_s 1172 * As far as we know, ocfs2_setattr() could only be the first
1180 * VFS entry point in the cal 1173 * VFS entry point in the call chain of recursive cluster
1181 * locking issue. 1174 * locking issue.
1182 * 1175 *
1183 * For instance: 1176 * For instance:
1184 * chmod_common() 1177 * chmod_common()
1185 * notify_change() 1178 * notify_change()
1186 * ocfs2_setattr() 1179 * ocfs2_setattr()
1187 * posix_acl_chmod() 1180 * posix_acl_chmod()
1188 * ocfs2_iop_get_acl() 1181 * ocfs2_iop_get_acl()
1189 * 1182 *
1190 * But, we're not 100% sure i 1183 * But, we're not 100% sure if it's always true, because the
1191 * ordering of the VFS entry 1184 * ordering of the VFS entry points in the call chain is out
1192 * of our control. So, we'd b 1185 * of our control. So, we'd better dump the stack here to
1193 * catch the other cases of r 1186 * catch the other cases of recursive locking.
1194 */ 1187 */
1195 mlog(ML_ERROR, "Another case 1188 mlog(ML_ERROR, "Another case of recursive locking:\n");
1196 dump_stack(); 1189 dump_stack();
1197 } 1190 }
1198 inode_locked = 1; 1191 inode_locked = 1;
1199 1192
1200 if (size_change) { 1193 if (size_change) {
1201 status = inode_newsize_ok(ino 1194 status = inode_newsize_ok(inode, attr->ia_size);
1202 if (status) 1195 if (status)
1203 goto bail_unlock; 1196 goto bail_unlock;
1204 1197
1205 if (i_size_read(inode) >= att 1198 if (i_size_read(inode) >= attr->ia_size) {
1206 if (ocfs2_should_orde 1199 if (ocfs2_should_order_data(inode)) {
1207 status = ocfs 1200 status = ocfs2_begin_ordered_truncate(inode,
1208 1201 attr->ia_size);
1209 if (status) 1202 if (status)
1210 goto 1203 goto bail_unlock;
1211 } 1204 }
1212 status = ocfs2_trunca 1205 status = ocfs2_truncate_file(inode, bh, attr->ia_size);
1213 } else 1206 } else
1214 status = ocfs2_extend 1207 status = ocfs2_extend_file(inode, bh, attr->ia_size);
1215 if (status < 0) { 1208 if (status < 0) {
1216 if (status != -ENOSPC 1209 if (status != -ENOSPC)
1217 mlog_errno(st 1210 mlog_errno(status);
1218 status = -ENOSPC; 1211 status = -ENOSPC;
1219 goto bail_unlock; 1212 goto bail_unlock;
1220 } 1213 }
1221 } 1214 }
1222 1215
1223 if ((attr->ia_valid & ATTR_UID && !ui 1216 if ((attr->ia_valid & ATTR_UID && !uid_eq(attr->ia_uid, inode->i_uid)) ||
1224 (attr->ia_valid & ATTR_GID && !gi 1217 (attr->ia_valid & ATTR_GID && !gid_eq(attr->ia_gid, inode->i_gid))) {
1225 /* 1218 /*
1226 * Gather pointers to quota s 1219 * Gather pointers to quota structures so that allocation /
1227 * freeing of quota structure 1220 * freeing of quota structures happens here and not inside
1228 * dquot_transfer() where we 1221 * dquot_transfer() where we have problems with lock ordering
1229 */ 1222 */
1230 if (attr->ia_valid & ATTR_UID 1223 if (attr->ia_valid & ATTR_UID && !uid_eq(attr->ia_uid, inode->i_uid)
1231 && OCFS2_HAS_RO_COMPAT_FE 1224 && OCFS2_HAS_RO_COMPAT_FEATURE(sb,
1232 OCFS2_FEATURE_RO_COMPAT_U 1225 OCFS2_FEATURE_RO_COMPAT_USRQUOTA)) {
1233 transfer_to[USRQUOTA] 1226 transfer_to[USRQUOTA] = dqget(sb, make_kqid_uid(attr->ia_uid));
1234 if (IS_ERR(transfer_t 1227 if (IS_ERR(transfer_to[USRQUOTA])) {
1235 status = PTR_ 1228 status = PTR_ERR(transfer_to[USRQUOTA]);
1236 transfer_to[U <<
1237 goto bail_unl 1229 goto bail_unlock;
1238 } 1230 }
1239 } 1231 }
1240 if (attr->ia_valid & ATTR_GID 1232 if (attr->ia_valid & ATTR_GID && !gid_eq(attr->ia_gid, inode->i_gid)
1241 && OCFS2_HAS_RO_COMPAT_FE 1233 && OCFS2_HAS_RO_COMPAT_FEATURE(sb,
1242 OCFS2_FEATURE_RO_COMPAT_G 1234 OCFS2_FEATURE_RO_COMPAT_GRPQUOTA)) {
1243 transfer_to[GRPQUOTA] 1235 transfer_to[GRPQUOTA] = dqget(sb, make_kqid_gid(attr->ia_gid));
1244 if (IS_ERR(transfer_t 1236 if (IS_ERR(transfer_to[GRPQUOTA])) {
1245 status = PTR_ 1237 status = PTR_ERR(transfer_to[GRPQUOTA]);
1246 transfer_to[G <<
1247 goto bail_unl 1238 goto bail_unlock;
1248 } 1239 }
1249 } 1240 }
1250 down_write(&OCFS2_I(inode)->i <<
1251 handle = ocfs2_start_trans(os 1241 handle = ocfs2_start_trans(osb, OCFS2_INODE_UPDATE_CREDITS +
1252 2 1242 2 * ocfs2_quota_trans_credits(sb));
1253 if (IS_ERR(handle)) { 1243 if (IS_ERR(handle)) {
1254 status = PTR_ERR(hand 1244 status = PTR_ERR(handle);
1255 mlog_errno(status); 1245 mlog_errno(status);
1256 goto bail_unlock_allo !! 1246 goto bail_unlock;
1257 } 1247 }
1258 status = __dquot_transfer(ino 1248 status = __dquot_transfer(inode, transfer_to);
1259 if (status < 0) 1249 if (status < 0)
1260 goto bail_commit; 1250 goto bail_commit;
1261 } else { 1251 } else {
1262 down_write(&OCFS2_I(inode)->i <<
1263 handle = ocfs2_start_trans(os 1252 handle = ocfs2_start_trans(osb, OCFS2_INODE_UPDATE_CREDITS);
1264 if (IS_ERR(handle)) { 1253 if (IS_ERR(handle)) {
1265 status = PTR_ERR(hand 1254 status = PTR_ERR(handle);
1266 mlog_errno(status); 1255 mlog_errno(status);
1267 goto bail_unlock_allo !! 1256 goto bail_unlock;
1268 } 1257 }
1269 } 1258 }
1270 1259
1271 setattr_copy(&nop_mnt_idmap, inode, a !! 1260 setattr_copy(inode, attr);
1272 mark_inode_dirty(inode); 1261 mark_inode_dirty(inode);
1273 1262
1274 status = ocfs2_mark_inode_dirty(handl 1263 status = ocfs2_mark_inode_dirty(handle, inode, bh);
1275 if (status < 0) 1264 if (status < 0)
1276 mlog_errno(status); 1265 mlog_errno(status);
1277 1266
1278 bail_commit: 1267 bail_commit:
1279 ocfs2_commit_trans(osb, handle); 1268 ocfs2_commit_trans(osb, handle);
1280 bail_unlock_alloc: <<
1281 up_write(&OCFS2_I(inode)->ip_alloc_se <<
1282 bail_unlock: 1269 bail_unlock:
1283 if (status && inode_locked) { 1270 if (status && inode_locked) {
1284 ocfs2_inode_unlock_tracker(in 1271 ocfs2_inode_unlock_tracker(inode, 1, &oh, had_lock);
1285 inode_locked = 0; 1272 inode_locked = 0;
1286 } 1273 }
1287 bail_unlock_rw: 1274 bail_unlock_rw:
1288 if (size_change) 1275 if (size_change)
1289 ocfs2_rw_unlock(inode, 1); 1276 ocfs2_rw_unlock(inode, 1);
1290 bail: 1277 bail:
1291 1278
1292 /* Release quota pointers in case we 1279 /* Release quota pointers in case we acquired them */
1293 for (qtype = 0; qtype < OCFS2_MAXQUOT 1280 for (qtype = 0; qtype < OCFS2_MAXQUOTAS; qtype++)
1294 dqput(transfer_to[qtype]); 1281 dqput(transfer_to[qtype]);
1295 1282
1296 if (!status && attr->ia_valid & ATTR_ 1283 if (!status && attr->ia_valid & ATTR_MODE) {
1297 status = ocfs2_acl_chmod(inod 1284 status = ocfs2_acl_chmod(inode, bh);
1298 if (status < 0) 1285 if (status < 0)
1299 mlog_errno(status); 1286 mlog_errno(status);
1300 } 1287 }
1301 if (inode_locked) 1288 if (inode_locked)
1302 ocfs2_inode_unlock_tracker(in 1289 ocfs2_inode_unlock_tracker(inode, 1, &oh, had_lock);
1303 1290
1304 brelse(bh); 1291 brelse(bh);
1305 return status; 1292 return status;
1306 } 1293 }
1307 1294
1308 int ocfs2_getattr(struct mnt_idmap *idmap, co !! 1295 int ocfs2_getattr(const struct path *path, struct kstat *stat,
1309 struct kstat *stat, u32 req !! 1296 u32 request_mask, unsigned int flags)
1310 { 1297 {
1311 struct inode *inode = d_inode(path->d 1298 struct inode *inode = d_inode(path->dentry);
1312 struct super_block *sb = path->dentry 1299 struct super_block *sb = path->dentry->d_sb;
1313 struct ocfs2_super *osb = sb->s_fs_in 1300 struct ocfs2_super *osb = sb->s_fs_info;
1314 int err; 1301 int err;
1315 1302
1316 err = ocfs2_inode_revalidate(path->de 1303 err = ocfs2_inode_revalidate(path->dentry);
1317 if (err) { 1304 if (err) {
1318 if (err != -ENOENT) 1305 if (err != -ENOENT)
1319 mlog_errno(err); 1306 mlog_errno(err);
1320 goto bail; 1307 goto bail;
1321 } 1308 }
1322 1309
1323 generic_fillattr(&nop_mnt_idmap, requ !! 1310 generic_fillattr(inode, stat);
1324 /* 1311 /*
1325 * If there is inline data in the ino 1312 * If there is inline data in the inode, the inode will normally not
1326 * have data blocks allocated (it may 1313 * have data blocks allocated (it may have an external xattr block).
1327 * Report at least one sector for suc 1314 * Report at least one sector for such files, so tools like tar, rsync,
1328 * others don't incorrectly think the 1315 * others don't incorrectly think the file is completely sparse.
1329 */ 1316 */
1330 if (unlikely(OCFS2_I(inode)->ip_dyn_f 1317 if (unlikely(OCFS2_I(inode)->ip_dyn_features & OCFS2_INLINE_DATA_FL))
1331 stat->blocks += (stat->size + 1318 stat->blocks += (stat->size + 511)>>9;
1332 1319
1333 /* We set the blksize from the cluste 1320 /* We set the blksize from the cluster size for performance */
1334 stat->blksize = osb->s_clustersize; 1321 stat->blksize = osb->s_clustersize;
1335 1322
1336 bail: 1323 bail:
1337 return err; 1324 return err;
1338 } 1325 }
1339 1326
1340 int ocfs2_permission(struct mnt_idmap *idmap, !! 1327 int ocfs2_permission(struct inode *inode, int mask)
1341 int mask) <<
1342 { 1328 {
1343 int ret, had_lock; 1329 int ret, had_lock;
1344 struct ocfs2_lock_holder oh; 1330 struct ocfs2_lock_holder oh;
1345 1331
1346 if (mask & MAY_NOT_BLOCK) 1332 if (mask & MAY_NOT_BLOCK)
1347 return -ECHILD; 1333 return -ECHILD;
1348 1334
1349 had_lock = ocfs2_inode_lock_tracker(i 1335 had_lock = ocfs2_inode_lock_tracker(inode, NULL, 0, &oh);
1350 if (had_lock < 0) { 1336 if (had_lock < 0) {
1351 ret = had_lock; 1337 ret = had_lock;
1352 goto out; 1338 goto out;
1353 } else if (had_lock) { 1339 } else if (had_lock) {
1354 /* See comments in ocfs2_seta 1340 /* See comments in ocfs2_setattr() for details.
1355 * The call chain of this cas 1341 * The call chain of this case could be:
1356 * do_sys_open() 1342 * do_sys_open()
1357 * may_open() 1343 * may_open()
1358 * inode_permission() 1344 * inode_permission()
1359 * ocfs2_permission() 1345 * ocfs2_permission()
1360 * ocfs2_iop_get_acl() 1346 * ocfs2_iop_get_acl()
1361 */ 1347 */
1362 mlog(ML_ERROR, "Another case 1348 mlog(ML_ERROR, "Another case of recursive locking:\n");
1363 dump_stack(); 1349 dump_stack();
1364 } 1350 }
1365 1351
1366 ret = generic_permission(&nop_mnt_idm !! 1352 ret = generic_permission(inode, mask);
1367 1353
1368 ocfs2_inode_unlock_tracker(inode, 0, 1354 ocfs2_inode_unlock_tracker(inode, 0, &oh, had_lock);
1369 out: 1355 out:
1370 return ret; 1356 return ret;
1371 } 1357 }
1372 1358
1373 static int __ocfs2_write_remove_suid(struct i 1359 static int __ocfs2_write_remove_suid(struct inode *inode,
1374 struct b 1360 struct buffer_head *bh)
1375 { 1361 {
1376 int ret; 1362 int ret;
1377 handle_t *handle; 1363 handle_t *handle;
1378 struct ocfs2_super *osb = OCFS2_SB(in 1364 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
1379 struct ocfs2_dinode *di; 1365 struct ocfs2_dinode *di;
1380 1366
1381 trace_ocfs2_write_remove_suid( 1367 trace_ocfs2_write_remove_suid(
1382 (unsigned long long)O 1368 (unsigned long long)OCFS2_I(inode)->ip_blkno,
1383 inode->i_mode); 1369 inode->i_mode);
1384 1370
1385 handle = ocfs2_start_trans(osb, OCFS2 1371 handle = ocfs2_start_trans(osb, OCFS2_INODE_UPDATE_CREDITS);
1386 if (IS_ERR(handle)) { 1372 if (IS_ERR(handle)) {
1387 ret = PTR_ERR(handle); 1373 ret = PTR_ERR(handle);
1388 mlog_errno(ret); 1374 mlog_errno(ret);
1389 goto out; 1375 goto out;
1390 } 1376 }
1391 1377
1392 ret = ocfs2_journal_access_di(handle, 1378 ret = ocfs2_journal_access_di(handle, INODE_CACHE(inode), bh,
1393 OCFS2_J 1379 OCFS2_JOURNAL_ACCESS_WRITE);
1394 if (ret < 0) { 1380 if (ret < 0) {
1395 mlog_errno(ret); 1381 mlog_errno(ret);
1396 goto out_trans; 1382 goto out_trans;
1397 } 1383 }
1398 1384
1399 inode->i_mode &= ~S_ISUID; 1385 inode->i_mode &= ~S_ISUID;
1400 if ((inode->i_mode & S_ISGID) && (ino 1386 if ((inode->i_mode & S_ISGID) && (inode->i_mode & S_IXGRP))
1401 inode->i_mode &= ~S_ISGID; 1387 inode->i_mode &= ~S_ISGID;
1402 1388
1403 di = (struct ocfs2_dinode *) bh->b_da 1389 di = (struct ocfs2_dinode *) bh->b_data;
1404 di->i_mode = cpu_to_le16(inode->i_mod 1390 di->i_mode = cpu_to_le16(inode->i_mode);
1405 ocfs2_update_inode_fsync_trans(handle 1391 ocfs2_update_inode_fsync_trans(handle, inode, 0);
1406 1392
1407 ocfs2_journal_dirty(handle, bh); 1393 ocfs2_journal_dirty(handle, bh);
1408 1394
1409 out_trans: 1395 out_trans:
1410 ocfs2_commit_trans(osb, handle); 1396 ocfs2_commit_trans(osb, handle);
1411 out: 1397 out:
1412 return ret; 1398 return ret;
1413 } 1399 }
1414 1400
1415 static int ocfs2_write_remove_suid(struct ino 1401 static int ocfs2_write_remove_suid(struct inode *inode)
1416 { 1402 {
1417 int ret; 1403 int ret;
1418 struct buffer_head *bh = NULL; 1404 struct buffer_head *bh = NULL;
1419 1405
1420 ret = ocfs2_read_inode_block(inode, & 1406 ret = ocfs2_read_inode_block(inode, &bh);
1421 if (ret < 0) { 1407 if (ret < 0) {
1422 mlog_errno(ret); 1408 mlog_errno(ret);
1423 goto out; 1409 goto out;
1424 } 1410 }
1425 1411
1426 ret = __ocfs2_write_remove_suid(inod 1412 ret = __ocfs2_write_remove_suid(inode, bh);
1427 out: 1413 out:
1428 brelse(bh); 1414 brelse(bh);
1429 return ret; 1415 return ret;
1430 } 1416 }
1431 1417
1432 /* 1418 /*
1433 * Allocate enough extents to cover the regio 1419 * Allocate enough extents to cover the region starting at byte offset
1434 * start for len bytes. Existing extents are 1420 * start for len bytes. Existing extents are skipped, any extents
1435 * added are marked as "unwritten". 1421 * added are marked as "unwritten".
1436 */ 1422 */
1437 static int ocfs2_allocate_unwritten_extents(s 1423 static int ocfs2_allocate_unwritten_extents(struct inode *inode,
1438 u 1424 u64 start, u64 len)
1439 { 1425 {
1440 int ret; 1426 int ret;
1441 u32 cpos, phys_cpos, clusters, alloc_ 1427 u32 cpos, phys_cpos, clusters, alloc_size;
1442 u64 end = start + len; 1428 u64 end = start + len;
1443 struct buffer_head *di_bh = NULL; 1429 struct buffer_head *di_bh = NULL;
1444 1430
1445 if (OCFS2_I(inode)->ip_dyn_features & 1431 if (OCFS2_I(inode)->ip_dyn_features & OCFS2_INLINE_DATA_FL) {
1446 ret = ocfs2_read_inode_block( 1432 ret = ocfs2_read_inode_block(inode, &di_bh);
1447 if (ret) { 1433 if (ret) {
1448 mlog_errno(ret); 1434 mlog_errno(ret);
1449 goto out; 1435 goto out;
1450 } 1436 }
1451 1437
1452 /* 1438 /*
1453 * Nothing to do if the reque 1439 * Nothing to do if the requested reservation range
1454 * fits within the inode. 1440 * fits within the inode.
1455 */ 1441 */
1456 if (ocfs2_size_fits_inline_da 1442 if (ocfs2_size_fits_inline_data(di_bh, end))
1457 goto out; 1443 goto out;
1458 1444
1459 ret = ocfs2_convert_inline_da 1445 ret = ocfs2_convert_inline_data_to_extents(inode, di_bh);
1460 if (ret) { 1446 if (ret) {
1461 mlog_errno(ret); 1447 mlog_errno(ret);
1462 goto out; 1448 goto out;
1463 } 1449 }
1464 } 1450 }
1465 1451
1466 /* 1452 /*
1467 * We consider both start and len to 1453 * We consider both start and len to be inclusive.
1468 */ 1454 */
1469 cpos = start >> OCFS2_SB(inode->i_sb) 1455 cpos = start >> OCFS2_SB(inode->i_sb)->s_clustersize_bits;
1470 clusters = ocfs2_clusters_for_bytes(i 1456 clusters = ocfs2_clusters_for_bytes(inode->i_sb, start + len);
1471 clusters -= cpos; 1457 clusters -= cpos;
1472 1458
1473 while (clusters) { 1459 while (clusters) {
1474 ret = ocfs2_get_clusters(inod 1460 ret = ocfs2_get_clusters(inode, cpos, &phys_cpos,
1475 &all 1461 &alloc_size, NULL);
1476 if (ret) { 1462 if (ret) {
1477 mlog_errno(ret); 1463 mlog_errno(ret);
1478 goto out; 1464 goto out;
1479 } 1465 }
1480 1466
1481 /* 1467 /*
1482 * Hole or existing extent le 1468 * Hole or existing extent len can be arbitrary, so
1483 * cap it to our own allocati 1469 * cap it to our own allocation request.
1484 */ 1470 */
1485 if (alloc_size > clusters) 1471 if (alloc_size > clusters)
1486 alloc_size = clusters 1472 alloc_size = clusters;
1487 1473
1488 if (phys_cpos) { 1474 if (phys_cpos) {
1489 /* 1475 /*
1490 * We already have an 1476 * We already have an allocation at this
1491 * region so we can s 1477 * region so we can safely skip it.
1492 */ 1478 */
1493 goto next; 1479 goto next;
1494 } 1480 }
1495 1481
1496 ret = ocfs2_extend_allocation 1482 ret = ocfs2_extend_allocation(inode, cpos, alloc_size, 1);
1497 if (ret) { 1483 if (ret) {
1498 if (ret != -ENOSPC) 1484 if (ret != -ENOSPC)
1499 mlog_errno(re 1485 mlog_errno(ret);
1500 goto out; 1486 goto out;
1501 } 1487 }
1502 1488
1503 next: 1489 next:
1504 cpos += alloc_size; 1490 cpos += alloc_size;
1505 clusters -= alloc_size; 1491 clusters -= alloc_size;
1506 } 1492 }
1507 1493
1508 ret = 0; 1494 ret = 0;
1509 out: 1495 out:
1510 1496
1511 brelse(di_bh); 1497 brelse(di_bh);
1512 return ret; 1498 return ret;
1513 } 1499 }
1514 1500
1515 /* 1501 /*
1516 * Truncate a byte range, avoiding pages with 1502 * Truncate a byte range, avoiding pages within partial clusters. This
1517 * preserves those pages for the zeroing code 1503 * preserves those pages for the zeroing code to write to.
1518 */ 1504 */
1519 static void ocfs2_truncate_cluster_pages(stru 1505 static void ocfs2_truncate_cluster_pages(struct inode *inode, u64 byte_start,
1520 u64 1506 u64 byte_len)
1521 { 1507 {
1522 struct ocfs2_super *osb = OCFS2_SB(in 1508 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
1523 loff_t start, end; 1509 loff_t start, end;
1524 struct address_space *mapping = inode 1510 struct address_space *mapping = inode->i_mapping;
1525 1511
1526 start = (loff_t)ocfs2_align_bytes_to_ 1512 start = (loff_t)ocfs2_align_bytes_to_clusters(inode->i_sb, byte_start);
1527 end = byte_start + byte_len; 1513 end = byte_start + byte_len;
1528 end = end & ~(osb->s_clustersize - 1) 1514 end = end & ~(osb->s_clustersize - 1);
1529 1515
1530 if (start < end) { 1516 if (start < end) {
1531 unmap_mapping_range(mapping, 1517 unmap_mapping_range(mapping, start, end - start, 0);
1532 truncate_inode_pages_range(ma 1518 truncate_inode_pages_range(mapping, start, end - 1);
1533 } 1519 }
1534 } 1520 }
1535 1521
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 1522 static int ocfs2_zero_partial_clusters(struct inode *inode,
1576 u64 st 1523 u64 start, u64 len)
1577 { 1524 {
1578 int ret = 0; 1525 int ret = 0;
1579 u64 tmpend = 0; 1526 u64 tmpend = 0;
1580 u64 end = start + len; 1527 u64 end = start + len;
1581 struct ocfs2_super *osb = OCFS2_SB(in 1528 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
1582 unsigned int csize = osb->s_clustersi 1529 unsigned int csize = osb->s_clustersize;
1583 handle_t *handle; 1530 handle_t *handle;
1584 loff_t isize = i_size_read(inode); <<
1585 1531
1586 /* 1532 /*
1587 * The "start" and "end" values are N 1533 * The "start" and "end" values are NOT necessarily part of
1588 * the range whose allocation is bein 1534 * the range whose allocation is being deleted. Rather, this
1589 * is what the user passed in with th 1535 * is what the user passed in with the request. We must zero
1590 * partial clusters here. There's no 1536 * partial clusters here. There's no need to worry about
1591 * physical allocation - the zeroing 1537 * physical allocation - the zeroing code knows to skip holes.
1592 */ 1538 */
1593 trace_ocfs2_zero_partial_clusters( 1539 trace_ocfs2_zero_partial_clusters(
1594 (unsigned long long)OCFS2_I(i 1540 (unsigned long long)OCFS2_I(inode)->ip_blkno,
1595 (unsigned long long)start, (u 1541 (unsigned long long)start, (unsigned long long)end);
1596 1542
1597 /* 1543 /*
1598 * If both edges are on a cluster bou 1544 * If both edges are on a cluster boundary then there's no
1599 * zeroing required as the region is 1545 * zeroing required as the region is part of the allocation to
1600 * be truncated. 1546 * be truncated.
1601 */ 1547 */
1602 if ((start & (csize - 1)) == 0 && (en 1548 if ((start & (csize - 1)) == 0 && (end & (csize - 1)) == 0)
1603 goto out; 1549 goto out;
1604 1550
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 1551 handle = ocfs2_start_trans(osb, OCFS2_INODE_UPDATE_CREDITS);
1626 if (IS_ERR(handle)) { 1552 if (IS_ERR(handle)) {
1627 ret = PTR_ERR(handle); 1553 ret = PTR_ERR(handle);
1628 mlog_errno(ret); 1554 mlog_errno(ret);
1629 goto out; 1555 goto out;
1630 } 1556 }
1631 1557
1632 /* 1558 /*
1633 * If start is on a cluster boundary 1559 * If start is on a cluster boundary and end is somewhere in another
1634 * cluster, we have not COWed the clu 1560 * cluster, we have not COWed the cluster starting at start, unless
1635 * end is also within the same cluste 1561 * end is also within the same cluster. So, in this case, we skip this
1636 * first call to ocfs2_zero_range_for 1562 * first call to ocfs2_zero_range_for_truncate() truncate and move on
1637 * to the next one. 1563 * to the next one.
1638 */ 1564 */
1639 if ((start & (csize - 1)) != 0) { 1565 if ((start & (csize - 1)) != 0) {
1640 /* 1566 /*
1641 * We want to get the byte of 1567 * We want to get the byte offset of the end of the 1st
1642 * cluster. 1568 * cluster.
1643 */ 1569 */
1644 tmpend = (u64)osb->s_clusters 1570 tmpend = (u64)osb->s_clustersize +
1645 (start & ~(osb->s_clu 1571 (start & ~(osb->s_clustersize - 1));
1646 if (tmpend > end) 1572 if (tmpend > end)
1647 tmpend = end; 1573 tmpend = end;
1648 1574
1649 trace_ocfs2_zero_partial_clus 1575 trace_ocfs2_zero_partial_clusters_range1(
1650 (unsigned long long)s 1576 (unsigned long long)start,
1651 (unsigned long long)t 1577 (unsigned long long)tmpend);
1652 1578
1653 ret = ocfs2_zero_range_for_tr 1579 ret = ocfs2_zero_range_for_truncate(inode, handle, start,
1654 1580 tmpend);
1655 if (ret) 1581 if (ret)
1656 mlog_errno(ret); 1582 mlog_errno(ret);
1657 } 1583 }
1658 1584
1659 if (tmpend < end) { 1585 if (tmpend < end) {
1660 /* 1586 /*
1661 * This may make start and en 1587 * This may make start and end equal, but the zeroing
1662 * code will skip any work in 1588 * code will skip any work in that case so there's no
1663 * need to catch it up here. 1589 * need to catch it up here.
1664 */ 1590 */
1665 start = end & ~(osb->s_cluste 1591 start = end & ~(osb->s_clustersize - 1);
1666 1592
1667 trace_ocfs2_zero_partial_clus 1593 trace_ocfs2_zero_partial_clusters_range2(
1668 (unsigned long long)s 1594 (unsigned long long)start, (unsigned long long)end);
1669 1595
1670 ret = ocfs2_zero_range_for_tr 1596 ret = ocfs2_zero_range_for_truncate(inode, handle, start, end);
1671 if (ret) 1597 if (ret)
1672 mlog_errno(ret); 1598 mlog_errno(ret);
1673 } 1599 }
1674 ocfs2_update_inode_fsync_trans(handle 1600 ocfs2_update_inode_fsync_trans(handle, inode, 1);
1675 1601
1676 ocfs2_commit_trans(osb, handle); 1602 ocfs2_commit_trans(osb, handle);
1677 out: 1603 out:
1678 return ret; 1604 return ret;
1679 } 1605 }
1680 1606
1681 static int ocfs2_find_rec(struct ocfs2_extent 1607 static int ocfs2_find_rec(struct ocfs2_extent_list *el, u32 pos)
1682 { 1608 {
1683 int i; 1609 int i;
1684 struct ocfs2_extent_rec *rec = NULL; 1610 struct ocfs2_extent_rec *rec = NULL;
1685 1611
1686 for (i = le16_to_cpu(el->l_next_free_ 1612 for (i = le16_to_cpu(el->l_next_free_rec) - 1; i >= 0; i--) {
1687 1613
1688 rec = &el->l_recs[i]; 1614 rec = &el->l_recs[i];
1689 1615
1690 if (le32_to_cpu(rec->e_cpos) 1616 if (le32_to_cpu(rec->e_cpos) < pos)
1691 break; 1617 break;
1692 } 1618 }
1693 1619
1694 return i; 1620 return i;
1695 } 1621 }
1696 1622
1697 /* 1623 /*
1698 * Helper to calculate the punching pos and l 1624 * Helper to calculate the punching pos and length in one run, we handle the
1699 * following three cases in order: 1625 * following three cases in order:
1700 * 1626 *
1701 * - remove the entire record 1627 * - remove the entire record
1702 * - remove a partial record 1628 * - remove a partial record
1703 * - no record needs to be removed (hole-punc 1629 * - no record needs to be removed (hole-punching completed)
1704 */ 1630 */
1705 static void ocfs2_calc_trunc_pos(struct inode 1631 static void ocfs2_calc_trunc_pos(struct inode *inode,
1706 struct ocfs2 1632 struct ocfs2_extent_list *el,
1707 struct ocfs2 1633 struct ocfs2_extent_rec *rec,
1708 u32 trunc_st 1634 u32 trunc_start, u32 *trunc_cpos,
1709 u32 *trunc_l 1635 u32 *trunc_len, u32 *trunc_end,
1710 u64 *blkno, 1636 u64 *blkno, int *done)
1711 { 1637 {
1712 int ret = 0; 1638 int ret = 0;
1713 u32 coff, range; 1639 u32 coff, range;
1714 1640
1715 range = le32_to_cpu(rec->e_cpos) + oc 1641 range = le32_to_cpu(rec->e_cpos) + ocfs2_rec_clusters(el, rec);
1716 1642
1717 if (le32_to_cpu(rec->e_cpos) >= trunc 1643 if (le32_to_cpu(rec->e_cpos) >= trunc_start) {
1718 /* 1644 /*
1719 * remove an entire extent re 1645 * remove an entire extent record.
1720 */ 1646 */
1721 *trunc_cpos = le32_to_cpu(rec 1647 *trunc_cpos = le32_to_cpu(rec->e_cpos);
1722 /* 1648 /*
1723 * Skip holes if any. 1649 * Skip holes if any.
1724 */ 1650 */
1725 if (range < *trunc_end) 1651 if (range < *trunc_end)
1726 *trunc_end = range; 1652 *trunc_end = range;
1727 *trunc_len = *trunc_end - le3 1653 *trunc_len = *trunc_end - le32_to_cpu(rec->e_cpos);
1728 *blkno = le64_to_cpu(rec->e_b 1654 *blkno = le64_to_cpu(rec->e_blkno);
1729 *trunc_end = le32_to_cpu(rec- 1655 *trunc_end = le32_to_cpu(rec->e_cpos);
1730 } else if (range > trunc_start) { 1656 } else if (range > trunc_start) {
1731 /* 1657 /*
1732 * remove a partial extent re 1658 * remove a partial extent record, which means we're
1733 * removing the last extent r 1659 * removing the last extent record.
1734 */ 1660 */
1735 *trunc_cpos = trunc_start; 1661 *trunc_cpos = trunc_start;
1736 /* 1662 /*
1737 * skip hole if any. 1663 * skip hole if any.
1738 */ 1664 */
1739 if (range < *trunc_end) 1665 if (range < *trunc_end)
1740 *trunc_end = range; 1666 *trunc_end = range;
1741 *trunc_len = *trunc_end - tru 1667 *trunc_len = *trunc_end - trunc_start;
1742 coff = trunc_start - le32_to_ 1668 coff = trunc_start - le32_to_cpu(rec->e_cpos);
1743 *blkno = le64_to_cpu(rec->e_b 1669 *blkno = le64_to_cpu(rec->e_blkno) +
1744 ocfs2_cluster 1670 ocfs2_clusters_to_blocks(inode->i_sb, coff);
1745 *trunc_end = trunc_start; 1671 *trunc_end = trunc_start;
1746 } else { 1672 } else {
1747 /* 1673 /*
1748 * It may have two following 1674 * It may have two following possibilities:
1749 * 1675 *
1750 * - last record has been rem 1676 * - last record has been removed
1751 * - trunc_start was within a 1677 * - trunc_start was within a hole
1752 * 1678 *
1753 * both two cases mean the co 1679 * both two cases mean the completion of hole punching.
1754 */ 1680 */
1755 ret = 1; 1681 ret = 1;
1756 } 1682 }
1757 1683
1758 *done = ret; 1684 *done = ret;
1759 } 1685 }
1760 1686
1761 int ocfs2_remove_inode_range(struct inode *in 1687 int ocfs2_remove_inode_range(struct inode *inode,
1762 struct buffer_he 1688 struct buffer_head *di_bh, u64 byte_start,
1763 u64 byte_len) 1689 u64 byte_len)
1764 { 1690 {
1765 int ret = 0, flags = 0, done = 0, i; 1691 int ret = 0, flags = 0, done = 0, i;
1766 u32 trunc_start, trunc_len, trunc_end 1692 u32 trunc_start, trunc_len, trunc_end, trunc_cpos, phys_cpos;
1767 u32 cluster_in_el; 1693 u32 cluster_in_el;
1768 struct ocfs2_super *osb = OCFS2_SB(in 1694 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
1769 struct ocfs2_cached_dealloc_ctxt deal 1695 struct ocfs2_cached_dealloc_ctxt dealloc;
1770 struct address_space *mapping = inode 1696 struct address_space *mapping = inode->i_mapping;
1771 struct ocfs2_extent_tree et; 1697 struct ocfs2_extent_tree et;
1772 struct ocfs2_path *path = NULL; 1698 struct ocfs2_path *path = NULL;
1773 struct ocfs2_extent_list *el = NULL; 1699 struct ocfs2_extent_list *el = NULL;
1774 struct ocfs2_extent_rec *rec = NULL; 1700 struct ocfs2_extent_rec *rec = NULL;
1775 struct ocfs2_dinode *di = (struct ocf 1701 struct ocfs2_dinode *di = (struct ocfs2_dinode *)di_bh->b_data;
1776 u64 blkno, refcount_loc = le64_to_cpu 1702 u64 blkno, refcount_loc = le64_to_cpu(di->i_refcount_loc);
1777 1703
1778 ocfs2_init_dinode_extent_tree(&et, IN 1704 ocfs2_init_dinode_extent_tree(&et, INODE_CACHE(inode), di_bh);
1779 ocfs2_init_dealloc_ctxt(&dealloc); 1705 ocfs2_init_dealloc_ctxt(&dealloc);
1780 1706
1781 trace_ocfs2_remove_inode_range( 1707 trace_ocfs2_remove_inode_range(
1782 (unsigned long long)O 1708 (unsigned long long)OCFS2_I(inode)->ip_blkno,
1783 (unsigned long long)b 1709 (unsigned long long)byte_start,
1784 (unsigned long long)b 1710 (unsigned long long)byte_len);
1785 1711
1786 if (byte_len == 0) 1712 if (byte_len == 0)
1787 return 0; 1713 return 0;
1788 1714
1789 if (OCFS2_I(inode)->ip_dyn_features & 1715 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 1716 ret = ocfs2_truncate_inline(inode, di_bh, byte_start,
1799 b 1717 byte_start + byte_len, 0);
1800 if (ret) { 1718 if (ret) {
1801 mlog_errno(ret); 1719 mlog_errno(ret);
1802 goto out; 1720 goto out;
1803 } 1721 }
1804 /* 1722 /*
1805 * There's no need to get fan 1723 * There's no need to get fancy with the page cache
1806 * truncate of an inline-data 1724 * truncate of an inline-data inode. We're talking
1807 * about less than a page her 1725 * about less than a page here, which will be cached
1808 * in the dinode buffer anywa 1726 * in the dinode buffer anyway.
1809 */ 1727 */
1810 unmap_mapping_range(mapping, 1728 unmap_mapping_range(mapping, 0, 0, 0);
1811 truncate_inode_pages(mapping, 1729 truncate_inode_pages(mapping, 0);
1812 goto out; 1730 goto out;
1813 } 1731 }
1814 1732
1815 /* 1733 /*
1816 * For reflinks, we may need to CoW 2 1734 * For reflinks, we may need to CoW 2 clusters which might be
1817 * partially zero'd later, if hole's 1735 * partially zero'd later, if hole's start and end offset were
1818 * within one cluster(means is not ex 1736 * within one cluster(means is not exactly aligned to clustersize).
1819 */ 1737 */
1820 1738
1821 if (ocfs2_is_refcount_inode(inode)) { 1739 if (ocfs2_is_refcount_inode(inode)) {
1822 ret = ocfs2_cow_file_pos(inod 1740 ret = ocfs2_cow_file_pos(inode, di_bh, byte_start);
1823 if (ret) { 1741 if (ret) {
1824 mlog_errno(ret); 1742 mlog_errno(ret);
1825 goto out; 1743 goto out;
1826 } 1744 }
1827 1745
1828 ret = ocfs2_cow_file_pos(inod 1746 ret = ocfs2_cow_file_pos(inode, di_bh, byte_start + byte_len);
1829 if (ret) { 1747 if (ret) {
1830 mlog_errno(ret); 1748 mlog_errno(ret);
1831 goto out; 1749 goto out;
1832 } 1750 }
1833 } 1751 }
1834 1752
1835 trunc_start = ocfs2_clusters_for_byte 1753 trunc_start = ocfs2_clusters_for_bytes(osb->sb, byte_start);
1836 trunc_end = (byte_start + byte_len) > 1754 trunc_end = (byte_start + byte_len) >> osb->s_clustersize_bits;
1837 cluster_in_el = trunc_end; 1755 cluster_in_el = trunc_end;
1838 1756
1839 ret = ocfs2_zero_partial_clusters(ino 1757 ret = ocfs2_zero_partial_clusters(inode, byte_start, byte_len);
1840 if (ret) { 1758 if (ret) {
1841 mlog_errno(ret); 1759 mlog_errno(ret);
1842 goto out; 1760 goto out;
1843 } 1761 }
1844 1762
1845 path = ocfs2_new_path_from_et(&et); 1763 path = ocfs2_new_path_from_et(&et);
1846 if (!path) { 1764 if (!path) {
1847 ret = -ENOMEM; 1765 ret = -ENOMEM;
1848 mlog_errno(ret); 1766 mlog_errno(ret);
1849 goto out; 1767 goto out;
1850 } 1768 }
1851 1769
1852 while (trunc_end > trunc_start) { 1770 while (trunc_end > trunc_start) {
1853 1771
1854 ret = ocfs2_find_path(INODE_C 1772 ret = ocfs2_find_path(INODE_CACHE(inode), path,
1855 cluster 1773 cluster_in_el);
1856 if (ret) { 1774 if (ret) {
1857 mlog_errno(ret); 1775 mlog_errno(ret);
1858 goto out; 1776 goto out;
1859 } 1777 }
1860 1778
1861 el = path_leaf_el(path); 1779 el = path_leaf_el(path);
1862 1780
1863 i = ocfs2_find_rec(el, trunc_ 1781 i = ocfs2_find_rec(el, trunc_end);
1864 /* 1782 /*
1865 * Need to go to previous ext 1783 * Need to go to previous extent block.
1866 */ 1784 */
1867 if (i < 0) { 1785 if (i < 0) {
1868 if (path->p_tree_dept 1786 if (path->p_tree_depth == 0)
1869 break; 1787 break;
1870 1788
1871 ret = ocfs2_find_cpos 1789 ret = ocfs2_find_cpos_for_left_leaf(inode->i_sb,
1872 1790 path,
1873 1791 &cluster_in_el);
1874 if (ret) { 1792 if (ret) {
1875 mlog_errno(re 1793 mlog_errno(ret);
1876 goto out; 1794 goto out;
1877 } 1795 }
1878 1796
1879 /* 1797 /*
1880 * We've reached the 1798 * We've reached the leftmost extent block,
1881 * it's safe to leave 1799 * it's safe to leave.
1882 */ 1800 */
1883 if (cluster_in_el == 1801 if (cluster_in_el == 0)
1884 break; 1802 break;
1885 1803
1886 /* 1804 /*
1887 * The 'pos' searched 1805 * The 'pos' searched for previous extent block is
1888 * always one cluster 1806 * always one cluster less than actual trunc_end.
1889 */ 1807 */
1890 trunc_end = cluster_i 1808 trunc_end = cluster_in_el + 1;
1891 1809
1892 ocfs2_reinit_path(pat 1810 ocfs2_reinit_path(path, 1);
1893 1811
1894 continue; 1812 continue;
1895 1813
1896 } else 1814 } else
1897 rec = &el->l_recs[i]; 1815 rec = &el->l_recs[i];
1898 1816
1899 ocfs2_calc_trunc_pos(inode, e 1817 ocfs2_calc_trunc_pos(inode, el, rec, trunc_start, &trunc_cpos,
1900 &trunc_l 1818 &trunc_len, &trunc_end, &blkno, &done);
1901 if (done) 1819 if (done)
1902 break; 1820 break;
1903 1821
1904 flags = rec->e_flags; 1822 flags = rec->e_flags;
1905 phys_cpos = ocfs2_blocks_to_c 1823 phys_cpos = ocfs2_blocks_to_clusters(inode->i_sb, blkno);
1906 1824
1907 ret = ocfs2_remove_btree_rang 1825 ret = ocfs2_remove_btree_range(inode, &et, trunc_cpos,
1908 1826 phys_cpos, trunc_len, flags,
1909 1827 &dealloc, refcount_loc, false);
1910 if (ret < 0) { 1828 if (ret < 0) {
1911 mlog_errno(ret); 1829 mlog_errno(ret);
1912 goto out; 1830 goto out;
1913 } 1831 }
1914 1832
1915 cluster_in_el = trunc_end; 1833 cluster_in_el = trunc_end;
1916 1834
1917 ocfs2_reinit_path(path, 1); 1835 ocfs2_reinit_path(path, 1);
1918 } 1836 }
1919 1837
1920 ocfs2_truncate_cluster_pages(inode, b 1838 ocfs2_truncate_cluster_pages(inode, byte_start, byte_len);
1921 1839
1922 out: 1840 out:
1923 ocfs2_free_path(path); 1841 ocfs2_free_path(path);
1924 ocfs2_schedule_truncate_log_flush(osb 1842 ocfs2_schedule_truncate_log_flush(osb, 1);
1925 ocfs2_run_deallocs(osb, &dealloc); 1843 ocfs2_run_deallocs(osb, &dealloc);
1926 1844
1927 return ret; 1845 return ret;
1928 } 1846 }
1929 1847
1930 /* 1848 /*
1931 * Parts of this function taken from xfs_chan 1849 * Parts of this function taken from xfs_change_file_space()
1932 */ 1850 */
1933 static int __ocfs2_change_file_space(struct f 1851 static int __ocfs2_change_file_space(struct file *file, struct inode *inode,
1934 loff_t f 1852 loff_t f_pos, unsigned int cmd,
1935 struct o 1853 struct ocfs2_space_resv *sr,
1936 int chan 1854 int change_size)
1937 { 1855 {
1938 int ret; 1856 int ret;
1939 s64 llen; 1857 s64 llen;
1940 loff_t size, orig_isize; !! 1858 loff_t size;
1941 struct ocfs2_super *osb = OCFS2_SB(in 1859 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
1942 struct buffer_head *di_bh = NULL; 1860 struct buffer_head *di_bh = NULL;
1943 handle_t *handle; 1861 handle_t *handle;
1944 unsigned long long max_off = inode->i 1862 unsigned long long max_off = inode->i_sb->s_maxbytes;
1945 1863
1946 if (ocfs2_is_hard_readonly(osb) || oc 1864 if (ocfs2_is_hard_readonly(osb) || ocfs2_is_soft_readonly(osb))
1947 return -EROFS; 1865 return -EROFS;
1948 1866
1949 inode_lock(inode); 1867 inode_lock(inode);
1950 1868
1951 /* Wait all existing dio workers, new <<
1952 inode_dio_wait(inode); <<
1953 /* 1869 /*
1954 * This prevents concurrent writes on 1870 * This prevents concurrent writes on other nodes
1955 */ 1871 */
1956 ret = ocfs2_rw_lock(inode, 1); 1872 ret = ocfs2_rw_lock(inode, 1);
1957 if (ret) { 1873 if (ret) {
1958 mlog_errno(ret); 1874 mlog_errno(ret);
1959 goto out; 1875 goto out;
1960 } 1876 }
1961 1877
1962 ret = ocfs2_inode_lock(inode, &di_bh, 1878 ret = ocfs2_inode_lock(inode, &di_bh, 1);
1963 if (ret) { 1879 if (ret) {
1964 mlog_errno(ret); 1880 mlog_errno(ret);
1965 goto out_rw_unlock; 1881 goto out_rw_unlock;
1966 } 1882 }
1967 1883
1968 if (inode->i_flags & (S_IMMUTABLE|S_A 1884 if (inode->i_flags & (S_IMMUTABLE|S_APPEND)) {
1969 ret = -EPERM; 1885 ret = -EPERM;
1970 goto out_inode_unlock; 1886 goto out_inode_unlock;
1971 } 1887 }
1972 1888
1973 switch (sr->l_whence) { 1889 switch (sr->l_whence) {
1974 case 0: /*SEEK_SET*/ 1890 case 0: /*SEEK_SET*/
1975 break; 1891 break;
1976 case 1: /*SEEK_CUR*/ 1892 case 1: /*SEEK_CUR*/
1977 sr->l_start += f_pos; 1893 sr->l_start += f_pos;
1978 break; 1894 break;
1979 case 2: /*SEEK_END*/ 1895 case 2: /*SEEK_END*/
1980 sr->l_start += i_size_read(in 1896 sr->l_start += i_size_read(inode);
1981 break; 1897 break;
1982 default: 1898 default:
1983 ret = -EINVAL; 1899 ret = -EINVAL;
1984 goto out_inode_unlock; 1900 goto out_inode_unlock;
1985 } 1901 }
1986 sr->l_whence = 0; 1902 sr->l_whence = 0;
1987 1903
1988 llen = sr->l_len > 0 ? sr->l_len - 1 1904 llen = sr->l_len > 0 ? sr->l_len - 1 : sr->l_len;
1989 1905
1990 if (sr->l_start < 0 1906 if (sr->l_start < 0
1991 || sr->l_start > max_off 1907 || sr->l_start > max_off
1992 || (sr->l_start + llen) < 0 1908 || (sr->l_start + llen) < 0
1993 || (sr->l_start + llen) > max_off 1909 || (sr->l_start + llen) > max_off) {
1994 ret = -EINVAL; 1910 ret = -EINVAL;
1995 goto out_inode_unlock; 1911 goto out_inode_unlock;
1996 } 1912 }
1997 size = sr->l_start + sr->l_len; 1913 size = sr->l_start + sr->l_len;
1998 1914
1999 if (cmd == OCFS2_IOC_RESVSP || cmd == 1915 if (cmd == OCFS2_IOC_RESVSP || cmd == OCFS2_IOC_RESVSP64 ||
2000 cmd == OCFS2_IOC_UNRESVSP || cmd 1916 cmd == OCFS2_IOC_UNRESVSP || cmd == OCFS2_IOC_UNRESVSP64) {
2001 if (sr->l_len <= 0) { 1917 if (sr->l_len <= 0) {
2002 ret = -EINVAL; 1918 ret = -EINVAL;
2003 goto out_inode_unlock 1919 goto out_inode_unlock;
2004 } 1920 }
2005 } 1921 }
2006 1922
2007 if (file && setattr_should_drop_suidg !! 1923 if (file && should_remove_suid(file->f_path.dentry)) {
2008 ret = __ocfs2_write_remove_su 1924 ret = __ocfs2_write_remove_suid(inode, di_bh);
2009 if (ret) { 1925 if (ret) {
2010 mlog_errno(ret); 1926 mlog_errno(ret);
2011 goto out_inode_unlock 1927 goto out_inode_unlock;
2012 } 1928 }
2013 } 1929 }
2014 1930
2015 down_write(&OCFS2_I(inode)->ip_alloc_ 1931 down_write(&OCFS2_I(inode)->ip_alloc_sem);
2016 switch (cmd) { 1932 switch (cmd) {
2017 case OCFS2_IOC_RESVSP: 1933 case OCFS2_IOC_RESVSP:
2018 case OCFS2_IOC_RESVSP64: 1934 case OCFS2_IOC_RESVSP64:
2019 /* 1935 /*
2020 * This takes unsigned offset 1936 * This takes unsigned offsets, but the signed ones we
2021 * pass have been checked aga 1937 * pass have been checked against overflow above.
2022 */ 1938 */
2023 ret = ocfs2_allocate_unwritte 1939 ret = ocfs2_allocate_unwritten_extents(inode, sr->l_start,
2024 1940 sr->l_len);
2025 break; 1941 break;
2026 case OCFS2_IOC_UNRESVSP: 1942 case OCFS2_IOC_UNRESVSP:
2027 case OCFS2_IOC_UNRESVSP64: 1943 case OCFS2_IOC_UNRESVSP64:
2028 ret = ocfs2_remove_inode_rang 1944 ret = ocfs2_remove_inode_range(inode, di_bh, sr->l_start,
2029 1945 sr->l_len);
2030 break; 1946 break;
2031 default: 1947 default:
2032 ret = -EINVAL; 1948 ret = -EINVAL;
2033 } 1949 }
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 1950 up_write(&OCFS2_I(inode)->ip_alloc_sem);
2044 if (ret) { 1951 if (ret) {
2045 mlog_errno(ret); 1952 mlog_errno(ret);
2046 goto out_inode_unlock; 1953 goto out_inode_unlock;
2047 } 1954 }
2048 1955
2049 /* 1956 /*
2050 * We update c/mtime for these change 1957 * We update c/mtime for these changes
2051 */ 1958 */
2052 handle = ocfs2_start_trans(osb, OCFS2 1959 handle = ocfs2_start_trans(osb, OCFS2_INODE_UPDATE_CREDITS);
2053 if (IS_ERR(handle)) { 1960 if (IS_ERR(handle)) {
2054 ret = PTR_ERR(handle); 1961 ret = PTR_ERR(handle);
2055 mlog_errno(ret); 1962 mlog_errno(ret);
2056 goto out_inode_unlock; 1963 goto out_inode_unlock;
2057 } 1964 }
2058 1965
2059 inode_set_mtime_to_ts(inode, inode_se !! 1966 if (change_size && i_size_read(inode) < size)
>> 1967 i_size_write(inode, size);
>> 1968
>> 1969 inode->i_ctime = inode->i_mtime = current_time(inode);
2060 ret = ocfs2_mark_inode_dirty(handle, 1970 ret = ocfs2_mark_inode_dirty(handle, inode, di_bh);
2061 if (ret < 0) 1971 if (ret < 0)
2062 mlog_errno(ret); 1972 mlog_errno(ret);
2063 1973
2064 if (file && (file->f_flags & O_SYNC)) 1974 if (file && (file->f_flags & O_SYNC))
2065 handle->h_sync = 1; 1975 handle->h_sync = 1;
2066 1976
2067 ocfs2_commit_trans(osb, handle); 1977 ocfs2_commit_trans(osb, handle);
2068 1978
2069 out_inode_unlock: 1979 out_inode_unlock:
2070 brelse(di_bh); 1980 brelse(di_bh);
2071 ocfs2_inode_unlock(inode, 1); 1981 ocfs2_inode_unlock(inode, 1);
2072 out_rw_unlock: 1982 out_rw_unlock:
2073 ocfs2_rw_unlock(inode, 1); 1983 ocfs2_rw_unlock(inode, 1);
2074 1984
2075 out: 1985 out:
2076 inode_unlock(inode); 1986 inode_unlock(inode);
2077 return ret; 1987 return ret;
2078 } 1988 }
2079 1989
2080 int ocfs2_change_file_space(struct file *file 1990 int ocfs2_change_file_space(struct file *file, unsigned int cmd,
2081 struct ocfs2_spac 1991 struct ocfs2_space_resv *sr)
2082 { 1992 {
2083 struct inode *inode = file_inode(file 1993 struct inode *inode = file_inode(file);
2084 struct ocfs2_super *osb = OCFS2_SB(in 1994 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
2085 int ret; 1995 int ret;
2086 1996
2087 if ((cmd == OCFS2_IOC_RESVSP || cmd = 1997 if ((cmd == OCFS2_IOC_RESVSP || cmd == OCFS2_IOC_RESVSP64) &&
2088 !ocfs2_writes_unwritten_extents(o 1998 !ocfs2_writes_unwritten_extents(osb))
2089 return -ENOTTY; 1999 return -ENOTTY;
2090 else if ((cmd == OCFS2_IOC_UNRESVSP | 2000 else if ((cmd == OCFS2_IOC_UNRESVSP || cmd == OCFS2_IOC_UNRESVSP64) &&
2091 !ocfs2_sparse_alloc(osb)) 2001 !ocfs2_sparse_alloc(osb))
2092 return -ENOTTY; 2002 return -ENOTTY;
2093 2003
2094 if (!S_ISREG(inode->i_mode)) 2004 if (!S_ISREG(inode->i_mode))
2095 return -EINVAL; 2005 return -EINVAL;
2096 2006
2097 if (!(file->f_mode & FMODE_WRITE)) 2007 if (!(file->f_mode & FMODE_WRITE))
2098 return -EBADF; 2008 return -EBADF;
2099 2009
2100 ret = mnt_want_write_file(file); 2010 ret = mnt_want_write_file(file);
2101 if (ret) 2011 if (ret)
2102 return ret; 2012 return ret;
2103 ret = __ocfs2_change_file_space(file, 2013 ret = __ocfs2_change_file_space(file, inode, file->f_pos, cmd, sr, 0);
2104 mnt_drop_write_file(file); 2014 mnt_drop_write_file(file);
2105 return ret; 2015 return ret;
2106 } 2016 }
2107 2017
2108 static long ocfs2_fallocate(struct file *file 2018 static long ocfs2_fallocate(struct file *file, int mode, loff_t offset,
2109 loff_t len) 2019 loff_t len)
2110 { 2020 {
2111 struct inode *inode = file_inode(file 2021 struct inode *inode = file_inode(file);
2112 struct ocfs2_super *osb = OCFS2_SB(in 2022 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
2113 struct ocfs2_space_resv sr; 2023 struct ocfs2_space_resv sr;
2114 int change_size = 1; 2024 int change_size = 1;
2115 int cmd = OCFS2_IOC_RESVSP64; 2025 int cmd = OCFS2_IOC_RESVSP64;
2116 int ret = 0; <<
2117 2026
2118 if (mode & ~(FALLOC_FL_KEEP_SIZE | FA 2027 if (mode & ~(FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE))
2119 return -EOPNOTSUPP; 2028 return -EOPNOTSUPP;
2120 if (!ocfs2_writes_unwritten_extents(o 2029 if (!ocfs2_writes_unwritten_extents(osb))
2121 return -EOPNOTSUPP; 2030 return -EOPNOTSUPP;
2122 2031
2123 if (mode & FALLOC_FL_KEEP_SIZE) { !! 2032 if (mode & FALLOC_FL_KEEP_SIZE)
2124 change_size = 0; 2033 change_size = 0;
2125 } else { <<
2126 ret = inode_newsize_ok(inode, <<
2127 if (ret) <<
2128 return ret; <<
2129 } <<
2130 2034
2131 if (mode & FALLOC_FL_PUNCH_HOLE) 2035 if (mode & FALLOC_FL_PUNCH_HOLE)
2132 cmd = OCFS2_IOC_UNRESVSP64; 2036 cmd = OCFS2_IOC_UNRESVSP64;
2133 2037
2134 sr.l_whence = 0; 2038 sr.l_whence = 0;
2135 sr.l_start = (s64)offset; 2039 sr.l_start = (s64)offset;
2136 sr.l_len = (s64)len; 2040 sr.l_len = (s64)len;
2137 2041
2138 return __ocfs2_change_file_space(NULL 2042 return __ocfs2_change_file_space(NULL, inode, offset, cmd, &sr,
2139 chan 2043 change_size);
2140 } 2044 }
2141 2045
2142 int ocfs2_check_range_for_refcount(struct ino 2046 int ocfs2_check_range_for_refcount(struct inode *inode, loff_t pos,
2143 size_t cou 2047 size_t count)
2144 { 2048 {
2145 int ret = 0; 2049 int ret = 0;
2146 unsigned int extent_flags; 2050 unsigned int extent_flags;
2147 u32 cpos, clusters, extent_len, phys_ 2051 u32 cpos, clusters, extent_len, phys_cpos;
2148 struct super_block *sb = inode->i_sb; 2052 struct super_block *sb = inode->i_sb;
2149 2053
2150 if (!ocfs2_refcount_tree(OCFS2_SB(ino 2054 if (!ocfs2_refcount_tree(OCFS2_SB(inode->i_sb)) ||
2151 !ocfs2_is_refcount_inode(inode) | 2055 !ocfs2_is_refcount_inode(inode) ||
2152 OCFS2_I(inode)->ip_dyn_features & 2056 OCFS2_I(inode)->ip_dyn_features & OCFS2_INLINE_DATA_FL)
2153 return 0; 2057 return 0;
2154 2058
2155 cpos = pos >> OCFS2_SB(sb)->s_cluster 2059 cpos = pos >> OCFS2_SB(sb)->s_clustersize_bits;
2156 clusters = ocfs2_clusters_for_bytes(s 2060 clusters = ocfs2_clusters_for_bytes(sb, pos + count) - cpos;
2157 2061
2158 while (clusters) { 2062 while (clusters) {
2159 ret = ocfs2_get_clusters(inod 2063 ret = ocfs2_get_clusters(inode, cpos, &phys_cpos, &extent_len,
2160 &ext 2064 &extent_flags);
2161 if (ret < 0) { 2065 if (ret < 0) {
2162 mlog_errno(ret); 2066 mlog_errno(ret);
2163 goto out; 2067 goto out;
2164 } 2068 }
2165 2069
2166 if (phys_cpos && (extent_flag 2070 if (phys_cpos && (extent_flags & OCFS2_EXT_REFCOUNTED)) {
2167 ret = 1; 2071 ret = 1;
2168 break; 2072 break;
2169 } 2073 }
2170 2074
2171 if (extent_len > clusters) 2075 if (extent_len > clusters)
2172 extent_len = clusters 2076 extent_len = clusters;
2173 2077
2174 clusters -= extent_len; 2078 clusters -= extent_len;
2175 cpos += extent_len; 2079 cpos += extent_len;
2176 } 2080 }
2177 out: 2081 out:
2178 return ret; 2082 return ret;
2179 } 2083 }
2180 2084
2181 static int ocfs2_is_io_unaligned(struct inode 2085 static int ocfs2_is_io_unaligned(struct inode *inode, size_t count, loff_t pos)
2182 { 2086 {
2183 int blockmask = inode->i_sb->s_blocks 2087 int blockmask = inode->i_sb->s_blocksize - 1;
2184 loff_t final_size = pos + count; 2088 loff_t final_size = pos + count;
2185 2089
2186 if ((pos & blockmask) || (final_size 2090 if ((pos & blockmask) || (final_size & blockmask))
2187 return 1; 2091 return 1;
2188 return 0; 2092 return 0;
2189 } 2093 }
2190 2094
2191 static int ocfs2_inode_lock_for_extent_tree(s 2095 static int ocfs2_inode_lock_for_extent_tree(struct inode *inode,
2192 s 2096 struct buffer_head **di_bh,
2193 i 2097 int meta_level,
>> 2098 int overwrite_io,
2194 i 2099 int write_sem,
2195 i 2100 int wait)
2196 { 2101 {
2197 int ret = 0; 2102 int ret = 0;
2198 2103
2199 if (wait) 2104 if (wait)
2200 ret = ocfs2_inode_lock(inode, !! 2105 ret = ocfs2_inode_lock(inode, NULL, meta_level);
2201 else 2106 else
2202 ret = ocfs2_try_inode_lock(in !! 2107 ret = ocfs2_try_inode_lock(inode,
>> 2108 overwrite_io ? NULL : di_bh, meta_level);
2203 if (ret < 0) 2109 if (ret < 0)
2204 goto out; 2110 goto out;
2205 2111
2206 if (wait) { 2112 if (wait) {
2207 if (write_sem) 2113 if (write_sem)
2208 down_write(&OCFS2_I(i 2114 down_write(&OCFS2_I(inode)->ip_alloc_sem);
2209 else 2115 else
2210 down_read(&OCFS2_I(in 2116 down_read(&OCFS2_I(inode)->ip_alloc_sem);
2211 } else { 2117 } else {
2212 if (write_sem) 2118 if (write_sem)
2213 ret = down_write_tryl 2119 ret = down_write_trylock(&OCFS2_I(inode)->ip_alloc_sem);
2214 else 2120 else
2215 ret = down_read_trylo 2121 ret = down_read_trylock(&OCFS2_I(inode)->ip_alloc_sem);
2216 2122
2217 if (!ret) { 2123 if (!ret) {
2218 ret = -EAGAIN; 2124 ret = -EAGAIN;
2219 goto out_unlock; 2125 goto out_unlock;
2220 } 2126 }
2221 } 2127 }
2222 2128
2223 return ret; 2129 return ret;
2224 2130
2225 out_unlock: 2131 out_unlock:
2226 brelse(*di_bh); 2132 brelse(*di_bh);
2227 *di_bh = NULL; <<
2228 ocfs2_inode_unlock(inode, meta_level) 2133 ocfs2_inode_unlock(inode, meta_level);
2229 out: 2134 out:
2230 return ret; 2135 return ret;
2231 } 2136 }
2232 2137
2233 static void ocfs2_inode_unlock_for_extent_tre 2138 static void ocfs2_inode_unlock_for_extent_tree(struct inode *inode,
2234 2139 struct buffer_head **di_bh,
2235 2140 int meta_level,
2236 2141 int write_sem)
2237 { 2142 {
2238 if (write_sem) 2143 if (write_sem)
2239 up_write(&OCFS2_I(inode)->ip_ 2144 up_write(&OCFS2_I(inode)->ip_alloc_sem);
2240 else 2145 else
2241 up_read(&OCFS2_I(inode)->ip_a 2146 up_read(&OCFS2_I(inode)->ip_alloc_sem);
2242 2147
2243 brelse(*di_bh); 2148 brelse(*di_bh);
2244 *di_bh = NULL; 2149 *di_bh = NULL;
2245 2150
2246 if (meta_level >= 0) 2151 if (meta_level >= 0)
2247 ocfs2_inode_unlock(inode, met 2152 ocfs2_inode_unlock(inode, meta_level);
2248 } 2153 }
2249 2154
2250 static int ocfs2_prepare_inode_for_write(stru 2155 static int ocfs2_prepare_inode_for_write(struct file *file,
2251 loff 2156 loff_t pos, size_t count, int wait)
2252 { 2157 {
2253 int ret = 0, meta_level = 0, overwrit 2158 int ret = 0, meta_level = 0, overwrite_io = 0;
2254 int write_sem = 0; 2159 int write_sem = 0;
2255 struct dentry *dentry = file->f_path. 2160 struct dentry *dentry = file->f_path.dentry;
2256 struct inode *inode = d_inode(dentry) 2161 struct inode *inode = d_inode(dentry);
2257 struct buffer_head *di_bh = NULL; 2162 struct buffer_head *di_bh = NULL;
>> 2163 loff_t end;
2258 u32 cpos; 2164 u32 cpos;
2259 u32 clusters; 2165 u32 clusters;
2260 2166
2261 /* 2167 /*
2262 * We start with a read level meta lo 2168 * We start with a read level meta lock and only jump to an ex
2263 * if we need to make modifications h 2169 * if we need to make modifications here.
2264 */ 2170 */
2265 for(;;) { 2171 for(;;) {
2266 ret = ocfs2_inode_lock_for_ex 2172 ret = ocfs2_inode_lock_for_extent_tree(inode,
2267 2173 &di_bh,
2268 2174 meta_level,
>> 2175 overwrite_io,
2269 2176 write_sem,
2270 2177 wait);
2271 if (ret < 0) { 2178 if (ret < 0) {
2272 if (ret != -EAGAIN) 2179 if (ret != -EAGAIN)
2273 mlog_errno(re 2180 mlog_errno(ret);
2274 goto out; 2181 goto out;
2275 } 2182 }
2276 2183
2277 /* 2184 /*
2278 * Check if IO will overwrite 2185 * Check if IO will overwrite allocated blocks in case
2279 * IOCB_NOWAIT flag is set. 2186 * IOCB_NOWAIT flag is set.
2280 */ 2187 */
2281 if (!wait && !overwrite_io) { 2188 if (!wait && !overwrite_io) {
2282 overwrite_io = 1; 2189 overwrite_io = 1;
2283 2190
2284 ret = ocfs2_overwrite 2191 ret = ocfs2_overwrite_io(inode, di_bh, pos, count);
2285 if (ret < 0) { 2192 if (ret < 0) {
2286 if (ret != -E 2193 if (ret != -EAGAIN)
2287 mlog_ 2194 mlog_errno(ret);
2288 goto out_unlo 2195 goto out_unlock;
2289 } 2196 }
2290 } 2197 }
2291 2198
2292 /* Clear suid / sgid if neces 2199 /* Clear suid / sgid if necessary. We do this here
2293 * instead of later in the wr 2200 * instead of later in the write path because
2294 * remove_suid() calls ->seta 2201 * remove_suid() calls ->setattr without any hint that
2295 * we may have already done o 2202 * we may have already done our cluster locking. Since
2296 * ocfs2_setattr() *must* tak 2203 * ocfs2_setattr() *must* take cluster locks to
2297 * proceed, this will lead us 2204 * proceed, this will lead us to recursively lock the
2298 * inode. There's also the di 2205 * inode. There's also the dinode i_size state which
2299 * can be lost via setattr du 2206 * can be lost via setattr during extending writes (we
2300 * set inode->i_size at the e 2207 * set inode->i_size at the end of a write. */
2301 if (setattr_should_drop_suidg !! 2208 if (should_remove_suid(dentry)) {
2302 if (meta_level == 0) 2209 if (meta_level == 0) {
2303 ocfs2_inode_u 2210 ocfs2_inode_unlock_for_extent_tree(inode,
2304 2211 &di_bh,
2305 2212 meta_level,
2306 2213 write_sem);
2307 meta_level = 2214 meta_level = 1;
2308 continue; 2215 continue;
2309 } 2216 }
2310 2217
2311 ret = ocfs2_write_rem 2218 ret = ocfs2_write_remove_suid(inode);
2312 if (ret < 0) { 2219 if (ret < 0) {
2313 mlog_errno(re 2220 mlog_errno(ret);
2314 goto out_unlo 2221 goto out_unlock;
2315 } 2222 }
2316 } 2223 }
2317 2224
>> 2225 end = pos + count;
>> 2226
2318 ret = ocfs2_check_range_for_r 2227 ret = ocfs2_check_range_for_refcount(inode, pos, count);
2319 if (ret == 1) { 2228 if (ret == 1) {
2320 ocfs2_inode_unlock_fo 2229 ocfs2_inode_unlock_for_extent_tree(inode,
2321 2230 &di_bh,
2322 2231 meta_level,
2323 2232 write_sem);
2324 meta_level = 1; <<
2325 write_sem = 1; <<
2326 ret = ocfs2_inode_loc 2233 ret = ocfs2_inode_lock_for_extent_tree(inode,
2327 2234 &di_bh,
2328 2235 meta_level,
2329 !! 2236 overwrite_io,
>> 2237 1,
2330 2238 wait);
>> 2239 write_sem = 1;
2331 if (ret < 0) { 2240 if (ret < 0) {
2332 if (ret != -E 2241 if (ret != -EAGAIN)
2333 mlog_ 2242 mlog_errno(ret);
2334 goto out; 2243 goto out;
2335 } 2244 }
2336 2245
2337 cpos = pos >> OCFS2_S 2246 cpos = pos >> OCFS2_SB(inode->i_sb)->s_clustersize_bits;
2338 clusters = 2247 clusters =
2339 ocfs2_cluster 2248 ocfs2_clusters_for_bytes(inode->i_sb, pos + count) - cpos;
2340 ret = ocfs2_refcount_ 2249 ret = ocfs2_refcount_cow(inode, di_bh, cpos, clusters, UINT_MAX);
2341 } 2250 }
2342 2251
2343 if (ret < 0) { 2252 if (ret < 0) {
2344 if (ret != -EAGAIN) 2253 if (ret != -EAGAIN)
2345 mlog_errno(re 2254 mlog_errno(ret);
2346 goto out_unlock; 2255 goto out_unlock;
2347 } 2256 }
2348 2257
2349 break; 2258 break;
2350 } 2259 }
2351 2260
2352 out_unlock: 2261 out_unlock:
2353 trace_ocfs2_prepare_inode_for_write(O 2262 trace_ocfs2_prepare_inode_for_write(OCFS2_I(inode)->ip_blkno,
2354 p 2263 pos, count, wait);
2355 2264
2356 ocfs2_inode_unlock_for_extent_tree(in 2265 ocfs2_inode_unlock_for_extent_tree(inode,
2357 &d 2266 &di_bh,
2358 me 2267 meta_level,
2359 wr 2268 write_sem);
2360 2269
2361 out: 2270 out:
2362 return ret; 2271 return ret;
2363 } 2272 }
2364 2273
2365 static ssize_t ocfs2_file_write_iter(struct k 2274 static ssize_t ocfs2_file_write_iter(struct kiocb *iocb,
2366 struct io 2275 struct iov_iter *from)
2367 { 2276 {
2368 int rw_level; 2277 int rw_level;
2369 ssize_t written = 0; 2278 ssize_t written = 0;
2370 ssize_t ret; 2279 ssize_t ret;
2371 size_t count = iov_iter_count(from); 2280 size_t count = iov_iter_count(from);
2372 struct file *file = iocb->ki_filp; 2281 struct file *file = iocb->ki_filp;
2373 struct inode *inode = file_inode(file 2282 struct inode *inode = file_inode(file);
2374 struct ocfs2_super *osb = OCFS2_SB(in 2283 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
2375 int full_coherency = !(osb->s_mount_o 2284 int full_coherency = !(osb->s_mount_opt &
2376 OCFS2_MOUNT_CO 2285 OCFS2_MOUNT_COHERENCY_BUFFERED);
2377 void *saved_ki_complete = NULL; 2286 void *saved_ki_complete = NULL;
2378 int append_write = ((iocb->ki_pos + c 2287 int append_write = ((iocb->ki_pos + count) >=
2379 i_size_read(inode) ? 2288 i_size_read(inode) ? 1 : 0);
2380 int direct_io = iocb->ki_flags & IOCB 2289 int direct_io = iocb->ki_flags & IOCB_DIRECT ? 1 : 0;
2381 int nowait = iocb->ki_flags & IOCB_NO 2290 int nowait = iocb->ki_flags & IOCB_NOWAIT ? 1 : 0;
2382 2291
2383 trace_ocfs2_file_write_iter(inode, fi 2292 trace_ocfs2_file_write_iter(inode, file, file->f_path.dentry,
2384 (unsigned long long)OCFS2_I(i 2293 (unsigned long long)OCFS2_I(inode)->ip_blkno,
2385 file->f_path.dentry->d_name.l 2294 file->f_path.dentry->d_name.len,
2386 file->f_path.dentry->d_name.n 2295 file->f_path.dentry->d_name.name,
2387 (unsigned int)from->nr_segs); 2296 (unsigned int)from->nr_segs); /* GRRRRR */
2388 2297
2389 if (!direct_io && nowait) 2298 if (!direct_io && nowait)
2390 return -EOPNOTSUPP; 2299 return -EOPNOTSUPP;
2391 2300
2392 if (count == 0) 2301 if (count == 0)
2393 return 0; 2302 return 0;
2394 2303
2395 if (nowait) { 2304 if (nowait) {
2396 if (!inode_trylock(inode)) 2305 if (!inode_trylock(inode))
2397 return -EAGAIN; 2306 return -EAGAIN;
2398 } else 2307 } else
2399 inode_lock(inode); 2308 inode_lock(inode);
2400 2309
2401 /* 2310 /*
2402 * Concurrent O_DIRECT writes are all 2311 * Concurrent O_DIRECT writes are allowed with
2403 * mount_option "coherency=buffered". 2312 * mount_option "coherency=buffered".
2404 * For append write, we must take rw 2313 * For append write, we must take rw EX.
2405 */ 2314 */
2406 rw_level = (!direct_io || full_cohere 2315 rw_level = (!direct_io || full_coherency || append_write);
2407 2316
2408 if (nowait) 2317 if (nowait)
2409 ret = ocfs2_try_rw_lock(inode 2318 ret = ocfs2_try_rw_lock(inode, rw_level);
2410 else 2319 else
2411 ret = ocfs2_rw_lock(inode, rw 2320 ret = ocfs2_rw_lock(inode, rw_level);
2412 if (ret < 0) { 2321 if (ret < 0) {
2413 if (ret != -EAGAIN) 2322 if (ret != -EAGAIN)
2414 mlog_errno(ret); 2323 mlog_errno(ret);
2415 goto out_mutex; 2324 goto out_mutex;
2416 } 2325 }
2417 2326
2418 /* 2327 /*
2419 * O_DIRECT writes with "coherency=fu 2328 * O_DIRECT writes with "coherency=full" need to take EX cluster
2420 * inode_lock to guarantee coherency. 2329 * inode_lock to guarantee coherency.
2421 */ 2330 */
2422 if (direct_io && full_coherency) { 2331 if (direct_io && full_coherency) {
2423 /* 2332 /*
2424 * We need to take and drop t 2333 * We need to take and drop the inode lock to force
2425 * other nodes to drop their 2334 * other nodes to drop their caches. Buffered I/O
2426 * already does this in write 2335 * already does this in write_begin().
2427 */ 2336 */
2428 if (nowait) 2337 if (nowait)
2429 ret = ocfs2_try_inode 2338 ret = ocfs2_try_inode_lock(inode, NULL, 1);
2430 else 2339 else
2431 ret = ocfs2_inode_loc 2340 ret = ocfs2_inode_lock(inode, NULL, 1);
2432 if (ret < 0) { 2341 if (ret < 0) {
2433 if (ret != -EAGAIN) 2342 if (ret != -EAGAIN)
2434 mlog_errno(re 2343 mlog_errno(ret);
2435 goto out; 2344 goto out;
2436 } 2345 }
2437 2346
2438 ocfs2_inode_unlock(inode, 1); 2347 ocfs2_inode_unlock(inode, 1);
2439 } 2348 }
2440 2349
2441 ret = generic_write_checks(iocb, from 2350 ret = generic_write_checks(iocb, from);
2442 if (ret <= 0) { 2351 if (ret <= 0) {
2443 if (ret) 2352 if (ret)
2444 mlog_errno(ret); 2353 mlog_errno(ret);
2445 goto out; 2354 goto out;
2446 } 2355 }
2447 count = ret; 2356 count = ret;
2448 2357
2449 ret = ocfs2_prepare_inode_for_write(f 2358 ret = ocfs2_prepare_inode_for_write(file, iocb->ki_pos, count, !nowait);
2450 if (ret < 0) { 2359 if (ret < 0) {
2451 if (ret != -EAGAIN) 2360 if (ret != -EAGAIN)
2452 mlog_errno(ret); 2361 mlog_errno(ret);
2453 goto out; 2362 goto out;
2454 } 2363 }
2455 2364
2456 if (direct_io && !is_sync_kiocb(iocb) 2365 if (direct_io && !is_sync_kiocb(iocb) &&
2457 ocfs2_is_io_unaligned(inode, coun 2366 ocfs2_is_io_unaligned(inode, count, iocb->ki_pos)) {
2458 /* 2367 /*
2459 * Make it a sync io if it's 2368 * Make it a sync io if it's an unaligned aio.
2460 */ 2369 */
2461 saved_ki_complete = xchg(&ioc 2370 saved_ki_complete = xchg(&iocb->ki_complete, NULL);
2462 } 2371 }
2463 2372
2464 /* communicate with ocfs2_dio_end_io 2373 /* communicate with ocfs2_dio_end_io */
2465 ocfs2_iocb_set_rw_locked(iocb, rw_lev 2374 ocfs2_iocb_set_rw_locked(iocb, rw_level);
2466 2375
2467 written = __generic_file_write_iter(i 2376 written = __generic_file_write_iter(iocb, from);
2468 /* buffered aio wouldn't have proper 2377 /* buffered aio wouldn't have proper lock coverage today */
2469 BUG_ON(written == -EIOCBQUEUED && !di 2378 BUG_ON(written == -EIOCBQUEUED && !direct_io);
2470 2379
2471 /* 2380 /*
2472 * deep in g_f_a_w_n()->ocfs2_direct_ 2381 * 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 2382 * function pointer which is called when o_direct io completes so that
2474 * it can unlock our rw lock. 2383 * it can unlock our rw lock.
2475 * Unfortunately there are error case 2384 * Unfortunately there are error cases which call end_io and others
2476 * that don't. so we don't have to u 2385 * 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 2386 * async dio is going to do it in the future or an end_io after an
2478 * error has already done it. 2387 * error has already done it.
2479 */ 2388 */
2480 if ((written == -EIOCBQUEUED) || (!oc 2389 if ((written == -EIOCBQUEUED) || (!ocfs2_iocb_is_rw_locked(iocb))) {
2481 rw_level = -1; 2390 rw_level = -1;
2482 } 2391 }
2483 2392
2484 if (unlikely(written <= 0)) 2393 if (unlikely(written <= 0))
2485 goto out; 2394 goto out;
2486 2395
2487 if (((file->f_flags & O_DSYNC) && !di 2396 if (((file->f_flags & O_DSYNC) && !direct_io) ||
2488 IS_SYNC(inode)) { 2397 IS_SYNC(inode)) {
2489 ret = filemap_fdatawrite_rang 2398 ret = filemap_fdatawrite_range(file->f_mapping,
2490 2399 iocb->ki_pos - written,
2491 2400 iocb->ki_pos - 1);
2492 if (ret < 0) 2401 if (ret < 0)
2493 written = ret; 2402 written = ret;
2494 2403
2495 if (!ret) { 2404 if (!ret) {
2496 ret = jbd2_journal_fo 2405 ret = jbd2_journal_force_commit(osb->journal->j_journal);
2497 if (ret < 0) 2406 if (ret < 0)
2498 written = ret 2407 written = ret;
2499 } 2408 }
2500 2409
2501 if (!ret) 2410 if (!ret)
2502 ret = filemap_fdatawa 2411 ret = filemap_fdatawait_range(file->f_mapping,
2503 2412 iocb->ki_pos - written,
2504 2413 iocb->ki_pos - 1);
2505 } 2414 }
2506 2415
2507 out: 2416 out:
2508 if (saved_ki_complete) 2417 if (saved_ki_complete)
2509 xchg(&iocb->ki_complete, save 2418 xchg(&iocb->ki_complete, saved_ki_complete);
2510 2419
2511 if (rw_level != -1) 2420 if (rw_level != -1)
2512 ocfs2_rw_unlock(inode, rw_lev 2421 ocfs2_rw_unlock(inode, rw_level);
2513 2422
2514 out_mutex: 2423 out_mutex:
2515 inode_unlock(inode); 2424 inode_unlock(inode);
2516 2425
2517 if (written) 2426 if (written)
2518 ret = written; 2427 ret = written;
2519 return ret; 2428 return ret;
2520 } 2429 }
2521 2430
2522 static ssize_t ocfs2_file_read_iter(struct ki 2431 static ssize_t ocfs2_file_read_iter(struct kiocb *iocb,
2523 struct iov 2432 struct iov_iter *to)
2524 { 2433 {
2525 int ret = 0, rw_level = -1, lock_leve 2434 int ret = 0, rw_level = -1, lock_level = 0;
2526 struct file *filp = iocb->ki_filp; 2435 struct file *filp = iocb->ki_filp;
2527 struct inode *inode = file_inode(filp 2436 struct inode *inode = file_inode(filp);
2528 int direct_io = iocb->ki_flags & IOCB 2437 int direct_io = iocb->ki_flags & IOCB_DIRECT ? 1 : 0;
2529 int nowait = iocb->ki_flags & IOCB_NO 2438 int nowait = iocb->ki_flags & IOCB_NOWAIT ? 1 : 0;
2530 2439
2531 trace_ocfs2_file_read_iter(inode, fil 2440 trace_ocfs2_file_read_iter(inode, filp, filp->f_path.dentry,
2532 (unsigned long long)O 2441 (unsigned long long)OCFS2_I(inode)->ip_blkno,
2533 filp->f_path.dentry-> 2442 filp->f_path.dentry->d_name.len,
2534 filp->f_path.dentry-> 2443 filp->f_path.dentry->d_name.name,
2535 to->nr_segs); /* GR 2444 to->nr_segs); /* GRRRRR */
2536 2445
2537 2446
2538 if (!inode) { 2447 if (!inode) {
2539 ret = -EINVAL; 2448 ret = -EINVAL;
2540 mlog_errno(ret); 2449 mlog_errno(ret);
2541 goto bail; 2450 goto bail;
2542 } 2451 }
2543 2452
2544 if (!direct_io && nowait) 2453 if (!direct_io && nowait)
2545 return -EOPNOTSUPP; 2454 return -EOPNOTSUPP;
2546 2455
2547 /* 2456 /*
2548 * buffered reads protect themselves !! 2457 * buffered reads protect themselves in ->readpage(). O_DIRECT reads
2549 * need locks to protect pending read 2458 * need locks to protect pending reads from racing with truncate.
2550 */ 2459 */
2551 if (direct_io) { 2460 if (direct_io) {
2552 if (nowait) 2461 if (nowait)
2553 ret = ocfs2_try_rw_lo 2462 ret = ocfs2_try_rw_lock(inode, 0);
2554 else 2463 else
2555 ret = ocfs2_rw_lock(i 2464 ret = ocfs2_rw_lock(inode, 0);
2556 2465
2557 if (ret < 0) { 2466 if (ret < 0) {
2558 if (ret != -EAGAIN) 2467 if (ret != -EAGAIN)
2559 mlog_errno(re 2468 mlog_errno(ret);
2560 goto bail; 2469 goto bail;
2561 } 2470 }
2562 rw_level = 0; 2471 rw_level = 0;
2563 /* communicate with ocfs2_dio 2472 /* communicate with ocfs2_dio_end_io */
2564 ocfs2_iocb_set_rw_locked(iocb 2473 ocfs2_iocb_set_rw_locked(iocb, rw_level);
2565 } 2474 }
2566 2475
2567 /* 2476 /*
2568 * We're fine letting folks race trun 2477 * We're fine letting folks race truncates and extending
2569 * writes with read across the cluste 2478 * writes with read across the cluster, just like they can
2570 * locally. Hence no rw_lock during r 2479 * locally. Hence no rw_lock during read.
2571 * 2480 *
2572 * Take and drop the meta data lock t 2481 * Take and drop the meta data lock to update inode fields
2573 * like i_size. This allows the check 2482 * like i_size. This allows the checks down below
2574 * copy_splice_read() a chance of act !! 2483 * generic_file_read_iter() a chance of actually working.
2575 */ 2484 */
2576 ret = ocfs2_inode_lock_atime(inode, f 2485 ret = ocfs2_inode_lock_atime(inode, filp->f_path.mnt, &lock_level,
2577 !nowait) 2486 !nowait);
2578 if (ret < 0) { 2487 if (ret < 0) {
2579 if (ret != -EAGAIN) 2488 if (ret != -EAGAIN)
2580 mlog_errno(ret); 2489 mlog_errno(ret);
2581 goto bail; 2490 goto bail;
2582 } 2491 }
2583 ocfs2_inode_unlock(inode, lock_level) 2492 ocfs2_inode_unlock(inode, lock_level);
2584 2493
2585 ret = generic_file_read_iter(iocb, to 2494 ret = generic_file_read_iter(iocb, to);
2586 trace_generic_file_read_iter_ret(ret) 2495 trace_generic_file_read_iter_ret(ret);
2587 2496
2588 /* buffered aio wouldn't have proper 2497 /* buffered aio wouldn't have proper lock coverage today */
2589 BUG_ON(ret == -EIOCBQUEUED && !direct 2498 BUG_ON(ret == -EIOCBQUEUED && !direct_io);
2590 2499
2591 /* see ocfs2_file_write_iter */ 2500 /* see ocfs2_file_write_iter */
2592 if (ret == -EIOCBQUEUED || !ocfs2_ioc 2501 if (ret == -EIOCBQUEUED || !ocfs2_iocb_is_rw_locked(iocb)) {
2593 rw_level = -1; 2502 rw_level = -1;
2594 } 2503 }
2595 2504
2596 bail: 2505 bail:
2597 if (rw_level != -1) 2506 if (rw_level != -1)
2598 ocfs2_rw_unlock(inode, rw_lev 2507 ocfs2_rw_unlock(inode, rw_level);
2599 2508
2600 return ret; 2509 return ret;
2601 } 2510 }
2602 2511
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() */ 2512 /* Refer generic_file_llseek_unlocked() */
2641 static loff_t ocfs2_file_llseek(struct file * 2513 static loff_t ocfs2_file_llseek(struct file *file, loff_t offset, int whence)
2642 { 2514 {
2643 struct inode *inode = file->f_mapping 2515 struct inode *inode = file->f_mapping->host;
2644 int ret = 0; 2516 int ret = 0;
2645 2517
2646 inode_lock(inode); 2518 inode_lock(inode);
2647 2519
2648 switch (whence) { 2520 switch (whence) {
2649 case SEEK_SET: 2521 case SEEK_SET:
2650 break; 2522 break;
2651 case SEEK_END: 2523 case SEEK_END:
2652 /* SEEK_END requires the OCFS 2524 /* SEEK_END requires the OCFS2 inode lock for the file
2653 * because it references the 2525 * because it references the file's size.
2654 */ 2526 */
2655 ret = ocfs2_inode_lock(inode, 2527 ret = ocfs2_inode_lock(inode, NULL, 0);
2656 if (ret < 0) { 2528 if (ret < 0) {
2657 mlog_errno(ret); 2529 mlog_errno(ret);
2658 goto out; 2530 goto out;
2659 } 2531 }
2660 offset += i_size_read(inode); 2532 offset += i_size_read(inode);
2661 ocfs2_inode_unlock(inode, 0); 2533 ocfs2_inode_unlock(inode, 0);
2662 break; 2534 break;
2663 case SEEK_CUR: 2535 case SEEK_CUR:
2664 if (offset == 0) { 2536 if (offset == 0) {
2665 offset = file->f_pos; 2537 offset = file->f_pos;
2666 goto out; 2538 goto out;
2667 } 2539 }
2668 offset += file->f_pos; 2540 offset += file->f_pos;
2669 break; 2541 break;
2670 case SEEK_DATA: 2542 case SEEK_DATA:
2671 case SEEK_HOLE: 2543 case SEEK_HOLE:
2672 ret = ocfs2_seek_data_hole_of 2544 ret = ocfs2_seek_data_hole_offset(file, &offset, whence);
2673 if (ret) 2545 if (ret)
2674 goto out; 2546 goto out;
2675 break; 2547 break;
2676 default: 2548 default:
2677 ret = -EINVAL; 2549 ret = -EINVAL;
2678 goto out; 2550 goto out;
2679 } 2551 }
2680 2552
2681 offset = vfs_setpos(file, offset, ino 2553 offset = vfs_setpos(file, offset, inode->i_sb->s_maxbytes);
2682 2554
2683 out: 2555 out:
2684 inode_unlock(inode); 2556 inode_unlock(inode);
2685 if (ret) 2557 if (ret)
2686 return ret; 2558 return ret;
2687 return offset; 2559 return offset;
2688 } 2560 }
2689 2561
2690 static loff_t ocfs2_remap_file_range(struct f 2562 static loff_t ocfs2_remap_file_range(struct file *file_in, loff_t pos_in,
2691 struct f 2563 struct file *file_out, loff_t pos_out,
2692 loff_t l 2564 loff_t len, unsigned int remap_flags)
2693 { 2565 {
2694 struct inode *inode_in = file_inode(f 2566 struct inode *inode_in = file_inode(file_in);
2695 struct inode *inode_out = file_inode( 2567 struct inode *inode_out = file_inode(file_out);
2696 struct ocfs2_super *osb = OCFS2_SB(in 2568 struct ocfs2_super *osb = OCFS2_SB(inode_in->i_sb);
2697 struct buffer_head *in_bh = NULL, *ou 2569 struct buffer_head *in_bh = NULL, *out_bh = NULL;
2698 bool same_inode = (inode_in == inode_ 2570 bool same_inode = (inode_in == inode_out);
2699 loff_t remapped = 0; 2571 loff_t remapped = 0;
2700 ssize_t ret; 2572 ssize_t ret;
2701 2573
2702 if (remap_flags & ~(REMAP_FILE_DEDUP 2574 if (remap_flags & ~(REMAP_FILE_DEDUP | REMAP_FILE_ADVISORY))
2703 return -EINVAL; 2575 return -EINVAL;
2704 if (!ocfs2_refcount_tree(osb)) 2576 if (!ocfs2_refcount_tree(osb))
2705 return -EOPNOTSUPP; 2577 return -EOPNOTSUPP;
2706 if (ocfs2_is_hard_readonly(osb) || oc 2578 if (ocfs2_is_hard_readonly(osb) || ocfs2_is_soft_readonly(osb))
2707 return -EROFS; 2579 return -EROFS;
2708 2580
2709 /* Lock both files against IO */ 2581 /* Lock both files against IO */
2710 ret = ocfs2_reflink_inodes_lock(inode 2582 ret = ocfs2_reflink_inodes_lock(inode_in, &in_bh, inode_out, &out_bh);
2711 if (ret) 2583 if (ret)
2712 return ret; 2584 return ret;
2713 2585
2714 /* Check file eligibility and prepare 2586 /* Check file eligibility and prepare for block sharing. */
2715 ret = -EINVAL; 2587 ret = -EINVAL;
2716 if ((OCFS2_I(inode_in)->ip_flags & OC 2588 if ((OCFS2_I(inode_in)->ip_flags & OCFS2_INODE_SYSTEM_FILE) ||
2717 (OCFS2_I(inode_out)->ip_flags & O 2589 (OCFS2_I(inode_out)->ip_flags & OCFS2_INODE_SYSTEM_FILE))
2718 goto out_unlock; 2590 goto out_unlock;
2719 2591
2720 ret = generic_remap_file_range_prep(f 2592 ret = generic_remap_file_range_prep(file_in, pos_in, file_out, pos_out,
2721 &len, remap_flags); 2593 &len, remap_flags);
2722 if (ret < 0 || len == 0) 2594 if (ret < 0 || len == 0)
2723 goto out_unlock; 2595 goto out_unlock;
2724 2596
2725 /* Lock out changes to the allocation 2597 /* Lock out changes to the allocation maps and remap. */
2726 down_write(&OCFS2_I(inode_in)->ip_all 2598 down_write(&OCFS2_I(inode_in)->ip_alloc_sem);
2727 if (!same_inode) 2599 if (!same_inode)
2728 down_write_nested(&OCFS2_I(in 2600 down_write_nested(&OCFS2_I(inode_out)->ip_alloc_sem,
2729 SINGLE_DEPT 2601 SINGLE_DEPTH_NESTING);
2730 2602
2731 /* Zap any page cache for the destina 2603 /* Zap any page cache for the destination file's range. */
2732 truncate_inode_pages_range(&inode_out 2604 truncate_inode_pages_range(&inode_out->i_data,
2733 round_down 2605 round_down(pos_out, PAGE_SIZE),
2734 round_up(p 2606 round_up(pos_out + len, PAGE_SIZE) - 1);
2735 2607
2736 remapped = ocfs2_reflink_remap_blocks 2608 remapped = ocfs2_reflink_remap_blocks(inode_in, in_bh, pos_in,
2737 inode_out, out_bh, po 2609 inode_out, out_bh, pos_out, len);
2738 up_write(&OCFS2_I(inode_in)->ip_alloc 2610 up_write(&OCFS2_I(inode_in)->ip_alloc_sem);
2739 if (!same_inode) 2611 if (!same_inode)
2740 up_write(&OCFS2_I(inode_out)- 2612 up_write(&OCFS2_I(inode_out)->ip_alloc_sem);
2741 if (remapped < 0) { 2613 if (remapped < 0) {
2742 ret = remapped; 2614 ret = remapped;
2743 mlog_errno(ret); 2615 mlog_errno(ret);
2744 goto out_unlock; 2616 goto out_unlock;
2745 } 2617 }
2746 2618
2747 /* 2619 /*
2748 * Empty the extent map so that we ma 2620 * Empty the extent map so that we may get the right extent
2749 * record from the disk. 2621 * record from the disk.
2750 */ 2622 */
2751 ocfs2_extent_map_trunc(inode_in, 0); 2623 ocfs2_extent_map_trunc(inode_in, 0);
2752 ocfs2_extent_map_trunc(inode_out, 0); 2624 ocfs2_extent_map_trunc(inode_out, 0);
2753 2625
2754 ret = ocfs2_reflink_update_dest(inode 2626 ret = ocfs2_reflink_update_dest(inode_out, out_bh, pos_out + len);
2755 if (ret) { 2627 if (ret) {
2756 mlog_errno(ret); 2628 mlog_errno(ret);
2757 goto out_unlock; 2629 goto out_unlock;
2758 } 2630 }
2759 2631
2760 out_unlock: 2632 out_unlock:
2761 ocfs2_reflink_inodes_unlock(inode_in, 2633 ocfs2_reflink_inodes_unlock(inode_in, in_bh, inode_out, out_bh);
2762 return remapped > 0 ? remapped : ret; 2634 return remapped > 0 ? remapped : ret;
2763 } 2635 }
2764 2636
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 } <<
2771 <<
2772 const struct inode_operations ocfs2_file_iops 2637 const struct inode_operations ocfs2_file_iops = {
2773 .setattr = ocfs2_setattr, 2638 .setattr = ocfs2_setattr,
2774 .getattr = ocfs2_getattr, 2639 .getattr = ocfs2_getattr,
2775 .permission = ocfs2_permission, 2640 .permission = ocfs2_permission,
2776 .listxattr = ocfs2_listxattr, 2641 .listxattr = ocfs2_listxattr,
2777 .fiemap = ocfs2_fiemap, 2642 .fiemap = ocfs2_fiemap,
2778 .get_inode_acl = ocfs2_iop_get_acl, !! 2643 .get_acl = ocfs2_iop_get_acl,
2779 .set_acl = ocfs2_iop_set_acl, 2644 .set_acl = ocfs2_iop_set_acl,
2780 .fileattr_get = ocfs2_fileattr_get, <<
2781 .fileattr_set = ocfs2_fileattr_set, <<
2782 }; 2645 };
2783 2646
2784 const struct inode_operations ocfs2_special_f 2647 const struct inode_operations ocfs2_special_file_iops = {
2785 .setattr = ocfs2_setattr, 2648 .setattr = ocfs2_setattr,
2786 .getattr = ocfs2_getattr, 2649 .getattr = ocfs2_getattr,
2787 .listxattr = ocfs2_listxattr, <<
2788 .permission = ocfs2_permission, 2650 .permission = ocfs2_permission,
2789 .get_inode_acl = ocfs2_iop_get_acl, !! 2651 .get_acl = ocfs2_iop_get_acl,
2790 .set_acl = ocfs2_iop_set_acl, 2652 .set_acl = ocfs2_iop_set_acl,
2791 }; 2653 };
2792 2654
2793 /* 2655 /*
2794 * Other than ->lock, keep ocfs2_fops and ocf 2656 * Other than ->lock, keep ocfs2_fops and ocfs2_dops in sync with
2795 * ocfs2_fops_no_plocks and ocfs2_dops_no_plo 2657 * ocfs2_fops_no_plocks and ocfs2_dops_no_plocks!
2796 */ 2658 */
2797 const struct file_operations ocfs2_fops = { 2659 const struct file_operations ocfs2_fops = {
2798 .llseek = ocfs2_file_llseek, 2660 .llseek = ocfs2_file_llseek,
2799 .mmap = ocfs2_mmap, 2661 .mmap = ocfs2_mmap,
2800 .fsync = ocfs2_sync_file, 2662 .fsync = ocfs2_sync_file,
2801 .release = ocfs2_file_release, 2663 .release = ocfs2_file_release,
2802 .open = ocfs2_file_open, 2664 .open = ocfs2_file_open,
2803 .read_iter = ocfs2_file_read_ite 2665 .read_iter = ocfs2_file_read_iter,
2804 .write_iter = ocfs2_file_write_it 2666 .write_iter = ocfs2_file_write_iter,
2805 .unlocked_ioctl = ocfs2_ioctl, 2667 .unlocked_ioctl = ocfs2_ioctl,
2806 #ifdef CONFIG_COMPAT 2668 #ifdef CONFIG_COMPAT
2807 .compat_ioctl = ocfs2_compat_ioctl, 2669 .compat_ioctl = ocfs2_compat_ioctl,
2808 #endif 2670 #endif
2809 .lock = ocfs2_lock, 2671 .lock = ocfs2_lock,
2810 .flock = ocfs2_flock, 2672 .flock = ocfs2_flock,
2811 .splice_read = ocfs2_file_splice_r !! 2673 .splice_read = generic_file_splice_read,
2812 .splice_write = iter_file_splice_wr 2674 .splice_write = iter_file_splice_write,
2813 .fallocate = ocfs2_fallocate, 2675 .fallocate = ocfs2_fallocate,
2814 .remap_file_range = ocfs2_remap_file_ 2676 .remap_file_range = ocfs2_remap_file_range,
2815 }; 2677 };
2816 2678
2817 WRAP_DIR_ITER(ocfs2_readdir) // FIXME! <<
2818 const struct file_operations ocfs2_dops = { 2679 const struct file_operations ocfs2_dops = {
2819 .llseek = ocfs2_dir_llseek, !! 2680 .llseek = generic_file_llseek,
2820 .read = generic_read_dir, 2681 .read = generic_read_dir,
2821 .iterate_shared = shared_ocfs2_readdi !! 2682 .iterate = ocfs2_readdir,
2822 .fsync = ocfs2_sync_file, 2683 .fsync = ocfs2_sync_file,
2823 .release = ocfs2_dir_release, 2684 .release = ocfs2_dir_release,
2824 .open = ocfs2_dir_open, 2685 .open = ocfs2_dir_open,
2825 .unlocked_ioctl = ocfs2_ioctl, 2686 .unlocked_ioctl = ocfs2_ioctl,
2826 #ifdef CONFIG_COMPAT 2687 #ifdef CONFIG_COMPAT
2827 .compat_ioctl = ocfs2_compat_ioctl, 2688 .compat_ioctl = ocfs2_compat_ioctl,
2828 #endif 2689 #endif
2829 .lock = ocfs2_lock, 2690 .lock = ocfs2_lock,
2830 .flock = ocfs2_flock, 2691 .flock = ocfs2_flock,
2831 }; 2692 };
2832 2693
2833 /* 2694 /*
2834 * POSIX-lockless variants of our file_operat 2695 * POSIX-lockless variants of our file_operations.
2835 * 2696 *
2836 * These will be used if the underlying clust 2697 * These will be used if the underlying cluster stack does not support
2837 * posix file locking, if the user passes the 2698 * posix file locking, if the user passes the "localflocks" mount
2838 * option, or if we have a local-only fs. 2699 * option, or if we have a local-only fs.
2839 * 2700 *
2840 * ocfs2_flock is in here because all stacks 2701 * ocfs2_flock is in here because all stacks handle UNIX file locks,
2841 * so we still want it in the case of no stac 2702 * so we still want it in the case of no stack support for
2842 * plocks. Internally, it will do the right t 2703 * plocks. Internally, it will do the right thing when asked to ignore
2843 * the cluster. 2704 * the cluster.
2844 */ 2705 */
2845 const struct file_operations ocfs2_fops_no_pl 2706 const struct file_operations ocfs2_fops_no_plocks = {
2846 .llseek = ocfs2_file_llseek, 2707 .llseek = ocfs2_file_llseek,
2847 .mmap = ocfs2_mmap, 2708 .mmap = ocfs2_mmap,
2848 .fsync = ocfs2_sync_file, 2709 .fsync = ocfs2_sync_file,
2849 .release = ocfs2_file_release, 2710 .release = ocfs2_file_release,
2850 .open = ocfs2_file_open, 2711 .open = ocfs2_file_open,
2851 .read_iter = ocfs2_file_read_ite 2712 .read_iter = ocfs2_file_read_iter,
2852 .write_iter = ocfs2_file_write_it 2713 .write_iter = ocfs2_file_write_iter,
2853 .unlocked_ioctl = ocfs2_ioctl, 2714 .unlocked_ioctl = ocfs2_ioctl,
2854 #ifdef CONFIG_COMPAT 2715 #ifdef CONFIG_COMPAT
2855 .compat_ioctl = ocfs2_compat_ioctl, 2716 .compat_ioctl = ocfs2_compat_ioctl,
2856 #endif 2717 #endif
2857 .flock = ocfs2_flock, 2718 .flock = ocfs2_flock,
2858 .splice_read = filemap_splice_read !! 2719 .splice_read = generic_file_splice_read,
2859 .splice_write = iter_file_splice_wr 2720 .splice_write = iter_file_splice_write,
2860 .fallocate = ocfs2_fallocate, 2721 .fallocate = ocfs2_fallocate,
2861 .remap_file_range = ocfs2_remap_file_ 2722 .remap_file_range = ocfs2_remap_file_range,
2862 }; 2723 };
2863 2724
2864 const struct file_operations ocfs2_dops_no_pl 2725 const struct file_operations ocfs2_dops_no_plocks = {
2865 .llseek = ocfs2_dir_llseek, !! 2726 .llseek = generic_file_llseek,
2866 .read = generic_read_dir, 2727 .read = generic_read_dir,
2867 .iterate_shared = shared_ocfs2_readdi !! 2728 .iterate = ocfs2_readdir,
2868 .fsync = ocfs2_sync_file, 2729 .fsync = ocfs2_sync_file,
2869 .release = ocfs2_dir_release, 2730 .release = ocfs2_dir_release,
2870 .open = ocfs2_dir_open, 2731 .open = ocfs2_dir_open,
2871 .unlocked_ioctl = ocfs2_ioctl, 2732 .unlocked_ioctl = ocfs2_ioctl,
2872 #ifdef CONFIG_COMPAT 2733 #ifdef CONFIG_COMPAT
2873 .compat_ioctl = ocfs2_compat_ioctl, 2734 .compat_ioctl = ocfs2_compat_ioctl,
2874 #endif 2735 #endif
2875 .flock = ocfs2_flock, 2736 .flock = ocfs2_flock,
2876 }; 2737 };
2877 2738
Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.