aboutsummaryrefslogtreecommitdiffstats
diff options
context:
space:
mode:
-rw-r--r--Documentation/filesystems/xfs.txt17
-rw-r--r--fs/xfs/kmem.c10
-rw-r--r--fs/xfs/kmem.h5
-rw-r--r--fs/xfs/libxfs/xfs_bmap.c6
-rw-r--r--fs/xfs/libxfs/xfs_bmap.h33
-rw-r--r--fs/xfs/libxfs/xfs_format.h10
-rw-r--r--fs/xfs/libxfs/xfs_fs.h (renamed from fs/xfs/xfs_fs.h)0
-rw-r--r--fs/xfs/libxfs/xfs_symlink_remote.c2
-rw-r--r--fs/xfs/libxfs/xfs_types.h (renamed from fs/xfs/xfs_types.h)0
-rw-r--r--fs/xfs/xfs_aops.c149
-rw-r--r--fs/xfs/xfs_aops.h3
-rw-r--r--fs/xfs/xfs_bmap_util.h37
-rw-r--r--fs/xfs/xfs_buf_item.c6
-rw-r--r--fs/xfs/xfs_dquot.h2
-rw-r--r--fs/xfs/xfs_file.c64
-rw-r--r--fs/xfs/xfs_inode.c136
-rw-r--r--fs/xfs/xfs_inode.h11
-rw-r--r--fs/xfs/xfs_ioctl.c50
-rw-r--r--fs/xfs/xfs_ioctl32.c2
-rw-r--r--fs/xfs/xfs_iomap.c2
-rw-r--r--fs/xfs/xfs_iomap.h2
-rw-r--r--fs/xfs/xfs_iops.c21
-rw-r--r--fs/xfs/xfs_log.c10
-rw-r--r--fs/xfs/xfs_super.c2
-rw-r--r--fs/xfs/xfs_sysctl.c18
-rw-r--r--fs/xfs/xfs_trans.c1
26 files changed, 348 insertions, 251 deletions
diff --git a/Documentation/filesystems/xfs.txt b/Documentation/filesystems/xfs.txt
index 2e5684596090..0bfafe108357 100644
--- a/Documentation/filesystems/xfs.txt
+++ b/Documentation/filesystems/xfs.txt
@@ -349,16 +349,13 @@ The following sysctls are available for the XFS filesystem:
349Deprecated Sysctls 349Deprecated Sysctls
350================== 350==================
351 351
352 fs.xfs.xfsbufd_centisecs (Min: 50 Default: 100 Max: 3000) 352None at present.
353 Dirty metadata is now tracked by the log subsystem and
354 flushing is driven by log space and idling demands. The
355 xfsbufd no longer exists, so this syctl does nothing.
356 353
357 Due for removal in 3.14.
358 354
359 fs.xfs.age_buffer_centisecs (Min: 100 Default: 1500 Max: 720000) 355Removed Sysctls
360 Dirty metadata is now tracked by the log subsystem and 356===============
361 flushing is driven by log space and idling demands. The
362 xfsbufd no longer exists, so this syctl does nothing.
363 357
364 Due for removal in 3.14. 358 Name Removed
359 ---- -------
360 fs.xfs.xfsbufd_centisec v3.20
361 fs.xfs.age_buffer_centisecs v3.20
diff --git a/fs/xfs/kmem.c b/fs/xfs/kmem.c
index 53e95b2a1369..a7a3a63bb360 100644
--- a/fs/xfs/kmem.c
+++ b/fs/xfs/kmem.c
@@ -91,16 +91,6 @@ kmem_zalloc_large(size_t size, xfs_km_flags_t flags)
91 return ptr; 91 return ptr;
92} 92}
93 93
94void
95kmem_free(const void *ptr)
96{
97 if (!is_vmalloc_addr(ptr)) {
98 kfree(ptr);
99 } else {
100 vfree(ptr);
101 }
102}
103
104void * 94void *
105kmem_realloc(const void *ptr, size_t newsize, size_t oldsize, 95kmem_realloc(const void *ptr, size_t newsize, size_t oldsize,
106 xfs_km_flags_t flags) 96 xfs_km_flags_t flags)
diff --git a/fs/xfs/kmem.h b/fs/xfs/kmem.h
index 64db0e53edea..cc6b768fc068 100644
--- a/fs/xfs/kmem.h
+++ b/fs/xfs/kmem.h
@@ -63,7 +63,10 @@ kmem_flags_convert(xfs_km_flags_t flags)
63extern void *kmem_alloc(size_t, xfs_km_flags_t); 63extern void *kmem_alloc(size_t, xfs_km_flags_t);
64extern void *kmem_zalloc_large(size_t size, xfs_km_flags_t); 64extern void *kmem_zalloc_large(size_t size, xfs_km_flags_t);
65extern void *kmem_realloc(const void *, size_t, size_t, xfs_km_flags_t); 65extern void *kmem_realloc(const void *, size_t, size_t, xfs_km_flags_t);
66extern void kmem_free(const void *); 66static inline void kmem_free(const void *ptr)
67{
68 kvfree(ptr);
69}
67 70
68 71
69extern void *kmem_zalloc_greedy(size_t *, size_t, size_t); 72extern void *kmem_zalloc_greedy(size_t *, size_t, size_t);
diff --git a/fs/xfs/libxfs/xfs_bmap.c b/fs/xfs/libxfs/xfs_bmap.c
index 63a5bb9113ee..61ec015dca16 100644
--- a/fs/xfs/libxfs/xfs_bmap.c
+++ b/fs/xfs/libxfs/xfs_bmap.c
@@ -973,7 +973,11 @@ xfs_bmap_local_to_extents(
973 *firstblock = args.fsbno; 973 *firstblock = args.fsbno;
974 bp = xfs_btree_get_bufl(args.mp, tp, args.fsbno, 0); 974 bp = xfs_btree_get_bufl(args.mp, tp, args.fsbno, 0);
975 975
976 /* initialise the block and copy the data */ 976 /*
977 * Initialise the block and copy the data
978 *
979 * Note: init_fn must set the buffer log item type correctly!
980 */
977 init_fn(tp, bp, ip, ifp); 981 init_fn(tp, bp, ip, ifp);
978 982
979 /* account for the change in fork size and log everything */ 983 /* account for the change in fork size and log everything */
diff --git a/fs/xfs/libxfs/xfs_bmap.h b/fs/xfs/libxfs/xfs_bmap.h
index 44db6db86402..b9d8a499d2c4 100644
--- a/fs/xfs/libxfs/xfs_bmap.h
+++ b/fs/xfs/libxfs/xfs_bmap.h
@@ -28,6 +28,37 @@ struct xfs_trans;
28extern kmem_zone_t *xfs_bmap_free_item_zone; 28extern kmem_zone_t *xfs_bmap_free_item_zone;
29 29
30/* 30/*
31 * Argument structure for xfs_bmap_alloc.
32 */
33struct xfs_bmalloca {
34 xfs_fsblock_t *firstblock; /* i/o first block allocated */
35 struct xfs_bmap_free *flist; /* bmap freelist */
36 struct xfs_trans *tp; /* transaction pointer */
37 struct xfs_inode *ip; /* incore inode pointer */
38 struct xfs_bmbt_irec prev; /* extent before the new one */
39 struct xfs_bmbt_irec got; /* extent after, or delayed */
40
41 xfs_fileoff_t offset; /* offset in file filling in */
42 xfs_extlen_t length; /* i/o length asked/allocated */
43 xfs_fsblock_t blkno; /* starting block of new extent */
44
45 struct xfs_btree_cur *cur; /* btree cursor */
46 xfs_extnum_t idx; /* current extent index */
47 int nallocs;/* number of extents alloc'd */
48 int logflags;/* flags for transaction logging */
49
50 xfs_extlen_t total; /* total blocks needed for xaction */
51 xfs_extlen_t minlen; /* minimum allocation size (blocks) */
52 xfs_extlen_t minleft; /* amount must be left after alloc */
53 bool eof; /* set if allocating past last extent */
54 bool wasdel; /* replacing a delayed allocation */
55 bool userdata;/* set if is user data */
56 bool aeof; /* allocated space at eof */
57 bool conv; /* overwriting unwritten extents */
58 int flags;
59};
60
61/*
31 * List of extents to be free "later". 62 * List of extents to be free "later".
32 * The list is kept sorted on xbf_startblock. 63 * The list is kept sorted on xbf_startblock.
33 */ 64 */
@@ -149,6 +180,8 @@ void xfs_bmap_local_to_extents_empty(struct xfs_inode *ip, int whichfork);
149void xfs_bmap_add_free(xfs_fsblock_t bno, xfs_filblks_t len, 180void xfs_bmap_add_free(xfs_fsblock_t bno, xfs_filblks_t len,
150 struct xfs_bmap_free *flist, struct xfs_mount *mp); 181 struct xfs_bmap_free *flist, struct xfs_mount *mp);
151void xfs_bmap_cancel(struct xfs_bmap_free *flist); 182void xfs_bmap_cancel(struct xfs_bmap_free *flist);
183int xfs_bmap_finish(struct xfs_trans **tp, struct xfs_bmap_free *flist,
184 int *committed);
152void xfs_bmap_compute_maxlevels(struct xfs_mount *mp, int whichfork); 185void xfs_bmap_compute_maxlevels(struct xfs_mount *mp, int whichfork);
153int xfs_bmap_first_unused(struct xfs_trans *tp, struct xfs_inode *ip, 186int xfs_bmap_first_unused(struct xfs_trans *tp, struct xfs_inode *ip,
154 xfs_extlen_t len, xfs_fileoff_t *unused, int whichfork); 187 xfs_extlen_t len, xfs_fileoff_t *unused, int whichfork);
diff --git a/fs/xfs/libxfs/xfs_format.h b/fs/xfs/libxfs/xfs_format.h
index 749c86102794..8eb718979383 100644
--- a/fs/xfs/libxfs/xfs_format.h
+++ b/fs/xfs/libxfs/xfs_format.h
@@ -307,8 +307,8 @@ typedef enum {
307#define XFS_SB_ICOUNT XFS_SB_MVAL(ICOUNT) 307#define XFS_SB_ICOUNT XFS_SB_MVAL(ICOUNT)
308#define XFS_SB_IFREE XFS_SB_MVAL(IFREE) 308#define XFS_SB_IFREE XFS_SB_MVAL(IFREE)
309#define XFS_SB_FDBLOCKS XFS_SB_MVAL(FDBLOCKS) 309#define XFS_SB_FDBLOCKS XFS_SB_MVAL(FDBLOCKS)
310#define XFS_SB_FEATURES2 XFS_SB_MVAL(FEATURES2) 310#define XFS_SB_FEATURES2 (XFS_SB_MVAL(FEATURES2) | \
311#define XFS_SB_BAD_FEATURES2 XFS_SB_MVAL(BAD_FEATURES2) 311 XFS_SB_MVAL(BAD_FEATURES2))
312#define XFS_SB_FEATURES_COMPAT XFS_SB_MVAL(FEATURES_COMPAT) 312#define XFS_SB_FEATURES_COMPAT XFS_SB_MVAL(FEATURES_COMPAT)
313#define XFS_SB_FEATURES_RO_COMPAT XFS_SB_MVAL(FEATURES_RO_COMPAT) 313#define XFS_SB_FEATURES_RO_COMPAT XFS_SB_MVAL(FEATURES_RO_COMPAT)
314#define XFS_SB_FEATURES_INCOMPAT XFS_SB_MVAL(FEATURES_INCOMPAT) 314#define XFS_SB_FEATURES_INCOMPAT XFS_SB_MVAL(FEATURES_INCOMPAT)
@@ -322,9 +322,9 @@ typedef enum {
322 XFS_SB_VERSIONNUM | XFS_SB_UQUOTINO | XFS_SB_GQUOTINO | \ 322 XFS_SB_VERSIONNUM | XFS_SB_UQUOTINO | XFS_SB_GQUOTINO | \
323 XFS_SB_QFLAGS | XFS_SB_SHARED_VN | XFS_SB_UNIT | XFS_SB_WIDTH | \ 323 XFS_SB_QFLAGS | XFS_SB_SHARED_VN | XFS_SB_UNIT | XFS_SB_WIDTH | \
324 XFS_SB_ICOUNT | XFS_SB_IFREE | XFS_SB_FDBLOCKS | XFS_SB_FEATURES2 | \ 324 XFS_SB_ICOUNT | XFS_SB_IFREE | XFS_SB_FDBLOCKS | XFS_SB_FEATURES2 | \
325 XFS_SB_BAD_FEATURES2 | XFS_SB_FEATURES_COMPAT | \ 325 XFS_SB_FEATURES_COMPAT | XFS_SB_FEATURES_RO_COMPAT | \
326 XFS_SB_FEATURES_RO_COMPAT | XFS_SB_FEATURES_INCOMPAT | \ 326 XFS_SB_FEATURES_INCOMPAT | XFS_SB_FEATURES_LOG_INCOMPAT | \
327 XFS_SB_FEATURES_LOG_INCOMPAT | XFS_SB_PQUOTINO) 327 XFS_SB_PQUOTINO)
328 328
329 329
330/* 330/*
diff --git a/fs/xfs/xfs_fs.h b/fs/xfs/libxfs/xfs_fs.h
index 18dc721ca19f..18dc721ca19f 100644
--- a/fs/xfs/xfs_fs.h
+++ b/fs/xfs/libxfs/xfs_fs.h
diff --git a/fs/xfs/libxfs/xfs_symlink_remote.c b/fs/xfs/libxfs/xfs_symlink_remote.c
index c80c5236c3da..e7e26bd6468f 100644
--- a/fs/xfs/libxfs/xfs_symlink_remote.c
+++ b/fs/xfs/libxfs/xfs_symlink_remote.c
@@ -178,6 +178,8 @@ xfs_symlink_local_to_remote(
178 struct xfs_mount *mp = ip->i_mount; 178 struct xfs_mount *mp = ip->i_mount;
179 char *buf; 179 char *buf;
180 180
181 xfs_trans_buf_set_type(tp, bp, XFS_BLFT_SYMLINK_BUF);
182
181 if (!xfs_sb_version_hascrc(&mp->m_sb)) { 183 if (!xfs_sb_version_hascrc(&mp->m_sb)) {
182 bp->b_ops = NULL; 184 bp->b_ops = NULL;
183 memcpy(bp->b_addr, ifp->if_u1.if_data, ifp->if_bytes); 185 memcpy(bp->b_addr, ifp->if_u1.if_data, ifp->if_bytes);
diff --git a/fs/xfs/xfs_types.h b/fs/xfs/libxfs/xfs_types.h
index b79dc66b2ecd..b79dc66b2ecd 100644
--- a/fs/xfs/xfs_types.h
+++ b/fs/xfs/libxfs/xfs_types.h
diff --git a/fs/xfs/xfs_aops.c b/fs/xfs/xfs_aops.c
index 18e2f3bbae5e..3a9b7a1b8704 100644
--- a/fs/xfs/xfs_aops.c
+++ b/fs/xfs/xfs_aops.c
@@ -135,30 +135,22 @@ xfs_setfilesize_trans_alloc(
135 */ 135 */
136STATIC int 136STATIC int
137xfs_setfilesize( 137xfs_setfilesize(
138 struct xfs_ioend *ioend) 138 struct xfs_inode *ip,
139 struct xfs_trans *tp,
140 xfs_off_t offset,
141 size_t size)
139{ 142{
140 struct xfs_inode *ip = XFS_I(ioend->io_inode);
141 struct xfs_trans *tp = ioend->io_append_trans;
142 xfs_fsize_t isize; 143 xfs_fsize_t isize;
143 144
144 /*
145 * The transaction may have been allocated in the I/O submission thread,
146 * thus we need to mark ourselves as beeing in a transaction manually.
147 * Similarly for freeze protection.
148 */
149 current_set_flags_nested(&tp->t_pflags, PF_FSTRANS);
150 rwsem_acquire_read(&VFS_I(ip)->i_sb->s_writers.lock_map[SB_FREEZE_FS-1],
151 0, 1, _THIS_IP_);
152
153 xfs_ilock(ip, XFS_ILOCK_EXCL); 145 xfs_ilock(ip, XFS_ILOCK_EXCL);
154 isize = xfs_new_eof(ip, ioend->io_offset + ioend->io_size); 146 isize = xfs_new_eof(ip, offset + size);
155 if (!isize) { 147 if (!isize) {
156 xfs_iunlock(ip, XFS_ILOCK_EXCL); 148 xfs_iunlock(ip, XFS_ILOCK_EXCL);
157 xfs_trans_cancel(tp, 0); 149 xfs_trans_cancel(tp, 0);
158 return 0; 150 return 0;
159 } 151 }
160 152
161 trace_xfs_setfilesize(ip, ioend->io_offset, ioend->io_size); 153 trace_xfs_setfilesize(ip, offset, size);
162 154
163 ip->i_d.di_size = isize; 155 ip->i_d.di_size = isize;
164 xfs_trans_ijoin(tp, ip, XFS_ILOCK_EXCL); 156 xfs_trans_ijoin(tp, ip, XFS_ILOCK_EXCL);
@@ -167,6 +159,25 @@ xfs_setfilesize(
167 return xfs_trans_commit(tp, 0); 159 return xfs_trans_commit(tp, 0);
168} 160}
169 161
162STATIC int
163xfs_setfilesize_ioend(
164 struct xfs_ioend *ioend)
165{
166 struct xfs_inode *ip = XFS_I(ioend->io_inode);
167 struct xfs_trans *tp = ioend->io_append_trans;
168
169 /*
170 * The transaction may have been allocated in the I/O submission thread,
171 * thus we need to mark ourselves as being in a transaction manually.
172 * Similarly for freeze protection.
173 */
174 current_set_flags_nested(&tp->t_pflags, PF_FSTRANS);
175 rwsem_acquire_read(&VFS_I(ip)->i_sb->s_writers.lock_map[SB_FREEZE_FS-1],
176 0, 1, _THIS_IP_);
177
178 return xfs_setfilesize(ip, tp, ioend->io_offset, ioend->io_size);
179}
180
170/* 181/*
171 * Schedule IO completion handling on the final put of an ioend. 182 * Schedule IO completion handling on the final put of an ioend.
172 * 183 *
@@ -182,8 +193,7 @@ xfs_finish_ioend(
182 193
183 if (ioend->io_type == XFS_IO_UNWRITTEN) 194 if (ioend->io_type == XFS_IO_UNWRITTEN)
184 queue_work(mp->m_unwritten_workqueue, &ioend->io_work); 195 queue_work(mp->m_unwritten_workqueue, &ioend->io_work);
185 else if (ioend->io_append_trans || 196 else if (ioend->io_append_trans)
186 (ioend->io_isdirect && xfs_ioend_is_append(ioend)))
187 queue_work(mp->m_data_workqueue, &ioend->io_work); 197 queue_work(mp->m_data_workqueue, &ioend->io_work);
188 else 198 else
189 xfs_destroy_ioend(ioend); 199 xfs_destroy_ioend(ioend);
@@ -215,22 +225,8 @@ xfs_end_io(
215 if (ioend->io_type == XFS_IO_UNWRITTEN) { 225 if (ioend->io_type == XFS_IO_UNWRITTEN) {
216 error = xfs_iomap_write_unwritten(ip, ioend->io_offset, 226 error = xfs_iomap_write_unwritten(ip, ioend->io_offset,
217 ioend->io_size); 227 ioend->io_size);
218 } else if (ioend->io_isdirect && xfs_ioend_is_append(ioend)) {
219 /*
220 * For direct I/O we do not know if we need to allocate blocks
221 * or not so we can't preallocate an append transaction as that
222 * results in nested reservations and log space deadlocks. Hence
223 * allocate the transaction here. While this is sub-optimal and
224 * can block IO completion for some time, we're stuck with doing
225 * it this way until we can pass the ioend to the direct IO
226 * allocation callbacks and avoid nesting that way.
227 */
228 error = xfs_setfilesize_trans_alloc(ioend);
229 if (error)
230 goto done;
231 error = xfs_setfilesize(ioend);
232 } else if (ioend->io_append_trans) { 228 } else if (ioend->io_append_trans) {
233 error = xfs_setfilesize(ioend); 229 error = xfs_setfilesize_ioend(ioend);
234 } else { 230 } else {
235 ASSERT(!xfs_ioend_is_append(ioend)); 231 ASSERT(!xfs_ioend_is_append(ioend));
236 } 232 }
@@ -242,17 +238,6 @@ done:
242} 238}
243 239
244/* 240/*
245 * Call IO completion handling in caller context on the final put of an ioend.
246 */
247STATIC void
248xfs_finish_ioend_sync(
249 struct xfs_ioend *ioend)
250{
251 if (atomic_dec_and_test(&ioend->io_remaining))
252 xfs_end_io(&ioend->io_work);
253}
254
255/*
256 * Allocate and initialise an IO completion structure. 241 * Allocate and initialise an IO completion structure.
257 * We need to track unwritten extent write completion here initially. 242 * We need to track unwritten extent write completion here initially.
258 * We'll need to extend this for updating the ondisk inode size later 243 * We'll need to extend this for updating the ondisk inode size later
@@ -273,7 +258,6 @@ xfs_alloc_ioend(
273 * all the I/O from calling the completion routine too early. 258 * all the I/O from calling the completion routine too early.
274 */ 259 */
275 atomic_set(&ioend->io_remaining, 1); 260 atomic_set(&ioend->io_remaining, 1);
276 ioend->io_isdirect = 0;
277 ioend->io_error = 0; 261 ioend->io_error = 0;
278 ioend->io_list = NULL; 262 ioend->io_list = NULL;
279 ioend->io_type = type; 263 ioend->io_type = type;
@@ -1459,11 +1443,7 @@ xfs_get_blocks_direct(
1459 * 1443 *
1460 * If the private argument is non-NULL __xfs_get_blocks signals us that we 1444 * If the private argument is non-NULL __xfs_get_blocks signals us that we
1461 * need to issue a transaction to convert the range from unwritten to written 1445 * need to issue a transaction to convert the range from unwritten to written
1462 * extents. In case this is regular synchronous I/O we just call xfs_end_io 1446 * extents.
1463 * to do this and we are done. But in case this was a successful AIO
1464 * request this handler is called from interrupt context, from which we
1465 * can't start transactions. In that case offload the I/O completion to
1466 * the workqueues we also use for buffered I/O completion.
1467 */ 1447 */
1468STATIC void 1448STATIC void
1469xfs_end_io_direct_write( 1449xfs_end_io_direct_write(
@@ -1472,7 +1452,12 @@ xfs_end_io_direct_write(
1472 ssize_t size, 1452 ssize_t size,
1473 void *private) 1453 void *private)
1474{ 1454{
1475 struct xfs_ioend *ioend = iocb->private; 1455 struct inode *inode = file_inode(iocb->ki_filp);
1456 struct xfs_inode *ip = XFS_I(inode);
1457 struct xfs_mount *mp = ip->i_mount;
1458
1459 if (XFS_FORCED_SHUTDOWN(mp))
1460 return;
1476 1461
1477 /* 1462 /*
1478 * While the generic direct I/O code updates the inode size, it does 1463 * While the generic direct I/O code updates the inode size, it does
@@ -1480,22 +1465,33 @@ xfs_end_io_direct_write(
1480 * end_io handler thinks the on-disk size is outside the in-core 1465 * end_io handler thinks the on-disk size is outside the in-core
1481 * size. To prevent this just update it a little bit earlier here. 1466 * size. To prevent this just update it a little bit earlier here.
1482 */ 1467 */
1483 if (offset + size > i_size_read(ioend->io_inode)) 1468 if (offset + size > i_size_read(inode))
1484 i_size_write(ioend->io_inode, offset + size); 1469 i_size_write(inode, offset + size);
1485 1470
1486 /* 1471 /*
1487 * blockdev_direct_IO can return an error even after the I/O 1472 * For direct I/O we do not know if we need to allocate blocks or not,
1488 * completion handler was called. Thus we need to protect 1473 * so we can't preallocate an append transaction, as that results in
1489 * against double-freeing. 1474 * nested reservations and log space deadlocks. Hence allocate the
1475 * transaction here. While this is sub-optimal and can block IO
1476 * completion for some time, we're stuck with doing it this way until
1477 * we can pass the ioend to the direct IO allocation callbacks and
1478 * avoid nesting that way.
1490 */ 1479 */
1491 iocb->private = NULL; 1480 if (private && size > 0) {
1492 1481 xfs_iomap_write_unwritten(ip, offset, size);
1493 ioend->io_offset = offset; 1482 } else if (offset + size > ip->i_d.di_size) {
1494 ioend->io_size = size; 1483 struct xfs_trans *tp;
1495 if (private && size > 0) 1484 int error;
1496 ioend->io_type = XFS_IO_UNWRITTEN; 1485
1486 tp = xfs_trans_alloc(mp, XFS_TRANS_FSYNC_TS);
1487 error = xfs_trans_reserve(tp, &M_RES(mp)->tr_fsyncts, 0, 0);
1488 if (error) {
1489 xfs_trans_cancel(tp, 0);
1490 return;
1491 }
1497 1492
1498 xfs_finish_ioend_sync(ioend); 1493 xfs_setfilesize(ip, tp, offset, size);
1494 }
1499} 1495}
1500 1496
1501STATIC ssize_t 1497STATIC ssize_t
@@ -1507,39 +1503,16 @@ xfs_vm_direct_IO(
1507{ 1503{
1508 struct inode *inode = iocb->ki_filp->f_mapping->host; 1504 struct inode *inode = iocb->ki_filp->f_mapping->host;
1509 struct block_device *bdev = xfs_find_bdev_for_inode(inode); 1505 struct block_device *bdev = xfs_find_bdev_for_inode(inode);
1510 struct xfs_ioend *ioend = NULL;
1511 ssize_t ret;
1512 1506
1513 if (rw & WRITE) { 1507 if (rw & WRITE) {
1514 size_t size = iov_iter_count(iter); 1508 return __blockdev_direct_IO(rw, iocb, inode, bdev, iter,
1515
1516 /*
1517 * We cannot preallocate a size update transaction here as we
1518 * don't know whether allocation is necessary or not. Hence we
1519 * can only tell IO completion that one is necessary if we are
1520 * not doing unwritten extent conversion.
1521 */
1522 iocb->private = ioend = xfs_alloc_ioend(inode, XFS_IO_DIRECT);
1523 if (offset + size > XFS_I(inode)->i_d.di_size)
1524 ioend->io_isdirect = 1;
1525
1526 ret = __blockdev_direct_IO(rw, iocb, inode, bdev, iter,
1527 offset, xfs_get_blocks_direct, 1509 offset, xfs_get_blocks_direct,
1528 xfs_end_io_direct_write, NULL, 1510 xfs_end_io_direct_write, NULL,
1529 DIO_ASYNC_EXTEND); 1511 DIO_ASYNC_EXTEND);
1530 if (ret != -EIOCBQUEUED && iocb->private)
1531 goto out_destroy_ioend;
1532 } else {
1533 ret = __blockdev_direct_IO(rw, iocb, inode, bdev, iter,
1534 offset, xfs_get_blocks_direct,
1535 NULL, NULL, 0);
1536 } 1512 }
1537 1513 return __blockdev_direct_IO(rw, iocb, inode, bdev, iter,
1538 return ret; 1514 offset, xfs_get_blocks_direct,
1539 1515 NULL, NULL, 0);
1540out_destroy_ioend:
1541 xfs_destroy_ioend(ioend);
1542 return ret;
1543} 1516}
1544 1517
1545/* 1518/*
diff --git a/fs/xfs/xfs_aops.h b/fs/xfs/xfs_aops.h
index f94dd459dff9..ac644e0137a4 100644
--- a/fs/xfs/xfs_aops.h
+++ b/fs/xfs/xfs_aops.h
@@ -24,14 +24,12 @@ extern mempool_t *xfs_ioend_pool;
24 * Types of I/O for bmap clustering and I/O completion tracking. 24 * Types of I/O for bmap clustering and I/O completion tracking.
25 */ 25 */
26enum { 26enum {
27 XFS_IO_DIRECT = 0, /* special case for direct I/O ioends */
28 XFS_IO_DELALLOC, /* covers delalloc region */ 27 XFS_IO_DELALLOC, /* covers delalloc region */
29 XFS_IO_UNWRITTEN, /* covers allocated but uninitialized data */ 28 XFS_IO_UNWRITTEN, /* covers allocated but uninitialized data */
30 XFS_IO_OVERWRITE, /* covers already allocated extent */ 29 XFS_IO_OVERWRITE, /* covers already allocated extent */
31}; 30};
32 31
33#define XFS_IO_TYPES \ 32#define XFS_IO_TYPES \
34 { 0, "" }, \
35 { XFS_IO_DELALLOC, "delalloc" }, \ 33 { XFS_IO_DELALLOC, "delalloc" }, \
36 { XFS_IO_UNWRITTEN, "unwritten" }, \ 34 { XFS_IO_UNWRITTEN, "unwritten" }, \
37 { XFS_IO_OVERWRITE, "overwrite" } 35 { XFS_IO_OVERWRITE, "overwrite" }
@@ -45,7 +43,6 @@ typedef struct xfs_ioend {
45 unsigned int io_type; /* delalloc / unwritten */ 43 unsigned int io_type; /* delalloc / unwritten */
46 int io_error; /* I/O error code */ 44 int io_error; /* I/O error code */
47 atomic_t io_remaining; /* hold count */ 45 atomic_t io_remaining; /* hold count */
48 unsigned int io_isdirect : 1;/* direct I/O */
49 struct inode *io_inode; /* file being written to */ 46 struct inode *io_inode; /* file being written to */
50 struct buffer_head *io_buffer_head;/* buffer linked list head */ 47 struct buffer_head *io_buffer_head;/* buffer linked list head */
51 struct buffer_head *io_buffer_tail;/* buffer linked list tail */ 48 struct buffer_head *io_buffer_tail;/* buffer linked list tail */
diff --git a/fs/xfs/xfs_bmap_util.h b/fs/xfs/xfs_bmap_util.h
index 2fdb72d2c908..736429a72a12 100644
--- a/fs/xfs/xfs_bmap_util.h
+++ b/fs/xfs/xfs_bmap_util.h
@@ -26,43 +26,8 @@ struct xfs_ifork;
26struct xfs_inode; 26struct xfs_inode;
27struct xfs_mount; 27struct xfs_mount;
28struct xfs_trans; 28struct xfs_trans;
29struct xfs_bmalloca;
29 30
30/*
31 * Argument structure for xfs_bmap_alloc.
32 */
33struct xfs_bmalloca {
34 xfs_fsblock_t *firstblock; /* i/o first block allocated */
35 struct xfs_bmap_free *flist; /* bmap freelist */
36 struct xfs_trans *tp; /* transaction pointer */
37 struct xfs_inode *ip; /* incore inode pointer */
38 struct xfs_bmbt_irec prev; /* extent before the new one */
39 struct xfs_bmbt_irec got; /* extent after, or delayed */
40
41 xfs_fileoff_t offset; /* offset in file filling in */
42 xfs_extlen_t length; /* i/o length asked/allocated */
43 xfs_fsblock_t blkno; /* starting block of new extent */
44
45 struct xfs_btree_cur *cur; /* btree cursor */
46 xfs_extnum_t idx; /* current extent index */
47 int nallocs;/* number of extents alloc'd */
48 int logflags;/* flags for transaction logging */
49
50 xfs_extlen_t total; /* total blocks needed for xaction */
51 xfs_extlen_t minlen; /* minimum allocation size (blocks) */
52 xfs_extlen_t minleft; /* amount must be left after alloc */
53 bool eof; /* set if allocating past last extent */
54 bool wasdel; /* replacing a delayed allocation */
55 bool userdata;/* set if is user data */
56 bool aeof; /* allocated space at eof */
57 bool conv; /* overwriting unwritten extents */
58 int flags;
59 struct completion *done;
60 struct work_struct work;
61 int result;
62};
63
64int xfs_bmap_finish(struct xfs_trans **tp, struct xfs_bmap_free *flist,
65 int *committed);
66int xfs_bmap_rtalloc(struct xfs_bmalloca *ap); 31int xfs_bmap_rtalloc(struct xfs_bmalloca *ap);
67int xfs_bmap_eof(struct xfs_inode *ip, xfs_fileoff_t endoff, 32int xfs_bmap_eof(struct xfs_inode *ip, xfs_fileoff_t endoff,
68 int whichfork, int *eof); 33 int whichfork, int *eof);
diff --git a/fs/xfs/xfs_buf_item.c b/fs/xfs/xfs_buf_item.c
index 3f9bd58edec7..507d96a57ac7 100644
--- a/fs/xfs/xfs_buf_item.c
+++ b/fs/xfs/xfs_buf_item.c
@@ -319,6 +319,10 @@ xfs_buf_item_format(
319 ASSERT(atomic_read(&bip->bli_refcount) > 0); 319 ASSERT(atomic_read(&bip->bli_refcount) > 0);
320 ASSERT((bip->bli_flags & XFS_BLI_LOGGED) || 320 ASSERT((bip->bli_flags & XFS_BLI_LOGGED) ||
321 (bip->bli_flags & XFS_BLI_STALE)); 321 (bip->bli_flags & XFS_BLI_STALE));
322 ASSERT((bip->bli_flags & XFS_BLI_STALE) ||
323 (xfs_blft_from_flags(&bip->__bli_format) > XFS_BLFT_UNKNOWN_BUF
324 && xfs_blft_from_flags(&bip->__bli_format) < XFS_BLFT_MAX_BUF));
325
322 326
323 /* 327 /*
324 * If it is an inode buffer, transfer the in-memory state to the 328 * If it is an inode buffer, transfer the in-memory state to the
@@ -535,7 +539,7 @@ xfs_buf_item_push(
535 if ((bp->b_flags & XBF_WRITE_FAIL) && 539 if ((bp->b_flags & XBF_WRITE_FAIL) &&
536 ___ratelimit(&xfs_buf_write_fail_rl_state, "XFS:")) { 540 ___ratelimit(&xfs_buf_write_fail_rl_state, "XFS:")) {
537 xfs_warn(bp->b_target->bt_mount, 541 xfs_warn(bp->b_target->bt_mount,
538"Detected failing async write on buffer block 0x%llx. Retrying async write.\n", 542"Detected failing async write on buffer block 0x%llx. Retrying async write.",
539 (long long)bp->b_bn); 543 (long long)bp->b_bn);
540 } 544 }
541 545
diff --git a/fs/xfs/xfs_dquot.h b/fs/xfs/xfs_dquot.h
index c24c67e22a2a..2f536f33cd26 100644
--- a/fs/xfs/xfs_dquot.h
+++ b/fs/xfs/xfs_dquot.h
@@ -86,7 +86,7 @@ static inline void xfs_dqflock(xfs_dquot_t *dqp)
86 wait_for_completion(&dqp->q_flush); 86 wait_for_completion(&dqp->q_flush);
87} 87}
88 88
89static inline int xfs_dqflock_nowait(xfs_dquot_t *dqp) 89static inline bool xfs_dqflock_nowait(xfs_dquot_t *dqp)
90{ 90{
91 return try_wait_for_completion(&dqp->q_flush); 91 return try_wait_for_completion(&dqp->q_flush);
92} 92}
diff --git a/fs/xfs/xfs_file.c b/fs/xfs/xfs_file.c
index 13e974e6a889..712d312d8e3e 100644
--- a/fs/xfs/xfs_file.c
+++ b/fs/xfs/xfs_file.c
@@ -127,6 +127,42 @@ xfs_iozero(
127 return (-status); 127 return (-status);
128} 128}
129 129
130int
131xfs_update_prealloc_flags(
132 struct xfs_inode *ip,
133 enum xfs_prealloc_flags flags)
134{
135 struct xfs_trans *tp;
136 int error;
137
138 tp = xfs_trans_alloc(ip->i_mount, XFS_TRANS_WRITEID);
139 error = xfs_trans_reserve(tp, &M_RES(ip->i_mount)->tr_writeid, 0, 0);
140 if (error) {
141 xfs_trans_cancel(tp, 0);
142 return error;
143 }
144
145 xfs_ilock(ip, XFS_ILOCK_EXCL);
146 xfs_trans_ijoin(tp, ip, XFS_ILOCK_EXCL);
147
148 if (!(flags & XFS_PREALLOC_INVISIBLE)) {
149 ip->i_d.di_mode &= ~S_ISUID;
150 if (ip->i_d.di_mode & S_IXGRP)
151 ip->i_d.di_mode &= ~S_ISGID;
152 xfs_trans_ichgtime(tp, ip, XFS_ICHGTIME_MOD | XFS_ICHGTIME_CHG);
153 }
154
155 if (flags & XFS_PREALLOC_SET)
156 ip->i_d.di_flags |= XFS_DIFLAG_PREALLOC;
157 if (flags & XFS_PREALLOC_CLEAR)
158 ip->i_d.di_flags &= ~XFS_DIFLAG_PREALLOC;
159
160 xfs_trans_log_inode(tp, ip, XFS_ILOG_CORE);
161 if (flags & XFS_PREALLOC_SYNC)
162 xfs_trans_set_sync(tp);
163 return xfs_trans_commit(tp, 0);
164}
165
130/* 166/*
131 * Fsync operations on directories are much simpler than on regular files, 167 * Fsync operations on directories are much simpler than on regular files,
132 * as there is no file data to flush, and thus also no need for explicit 168 * as there is no file data to flush, and thus also no need for explicit
@@ -784,8 +820,8 @@ xfs_file_fallocate(
784{ 820{
785 struct inode *inode = file_inode(file); 821 struct inode *inode = file_inode(file);
786 struct xfs_inode *ip = XFS_I(inode); 822 struct xfs_inode *ip = XFS_I(inode);
787 struct xfs_trans *tp;
788 long error; 823 long error;
824 enum xfs_prealloc_flags flags = 0;
789 loff_t new_size = 0; 825 loff_t new_size = 0;
790 826
791 if (!S_ISREG(inode->i_mode)) 827 if (!S_ISREG(inode->i_mode))
@@ -822,6 +858,8 @@ xfs_file_fallocate(
822 if (error) 858 if (error)
823 goto out_unlock; 859 goto out_unlock;
824 } else { 860 } else {
861 flags |= XFS_PREALLOC_SET;
862
825 if (!(mode & FALLOC_FL_KEEP_SIZE) && 863 if (!(mode & FALLOC_FL_KEEP_SIZE) &&
826 offset + len > i_size_read(inode)) { 864 offset + len > i_size_read(inode)) {
827 new_size = offset + len; 865 new_size = offset + len;
@@ -839,28 +877,10 @@ xfs_file_fallocate(
839 goto out_unlock; 877 goto out_unlock;
840 } 878 }
841 879
842 tp = xfs_trans_alloc(ip->i_mount, XFS_TRANS_WRITEID);
843 error = xfs_trans_reserve(tp, &M_RES(ip->i_mount)->tr_writeid, 0, 0);
844 if (error) {
845 xfs_trans_cancel(tp, 0);
846 goto out_unlock;
847 }
848
849 xfs_ilock(ip, XFS_ILOCK_EXCL);
850 xfs_trans_ijoin(tp, ip, XFS_ILOCK_EXCL);
851 ip->i_d.di_mode &= ~S_ISUID;
852 if (ip->i_d.di_mode & S_IXGRP)
853 ip->i_d.di_mode &= ~S_ISGID;
854
855 if (!(mode & (FALLOC_FL_PUNCH_HOLE | FALLOC_FL_COLLAPSE_RANGE)))
856 ip->i_d.di_flags |= XFS_DIFLAG_PREALLOC;
857
858 xfs_trans_ichgtime(tp, ip, XFS_ICHGTIME_MOD | XFS_ICHGTIME_CHG);
859 xfs_trans_log_inode(tp, ip, XFS_ILOG_CORE);
860
861 if (file->f_flags & O_DSYNC) 880 if (file->f_flags & O_DSYNC)
862 xfs_trans_set_sync(tp); 881 flags |= XFS_PREALLOC_SYNC;
863 error = xfs_trans_commit(tp, 0); 882
883 error = xfs_update_prealloc_flags(ip, flags);
864 if (error) 884 if (error)
865 goto out_unlock; 885 goto out_unlock;
866 886
diff --git a/fs/xfs/xfs_inode.c b/fs/xfs/xfs_inode.c
index 41f804e740d7..daafa1f6d260 100644
--- a/fs/xfs/xfs_inode.c
+++ b/fs/xfs/xfs_inode.c
@@ -1995,6 +1995,7 @@ xfs_iunlink(
1995 agi->agi_unlinked[bucket_index] = cpu_to_be32(agino); 1995 agi->agi_unlinked[bucket_index] = cpu_to_be32(agino);
1996 offset = offsetof(xfs_agi_t, agi_unlinked) + 1996 offset = offsetof(xfs_agi_t, agi_unlinked) +
1997 (sizeof(xfs_agino_t) * bucket_index); 1997 (sizeof(xfs_agino_t) * bucket_index);
1998 xfs_trans_buf_set_type(tp, agibp, XFS_BLFT_AGI_BUF);
1998 xfs_trans_log_buf(tp, agibp, offset, 1999 xfs_trans_log_buf(tp, agibp, offset,
1999 (offset + sizeof(xfs_agino_t) - 1)); 2000 (offset + sizeof(xfs_agino_t) - 1));
2000 return 0; 2001 return 0;
@@ -2086,6 +2087,7 @@ xfs_iunlink_remove(
2086 agi->agi_unlinked[bucket_index] = cpu_to_be32(next_agino); 2087 agi->agi_unlinked[bucket_index] = cpu_to_be32(next_agino);
2087 offset = offsetof(xfs_agi_t, agi_unlinked) + 2088 offset = offsetof(xfs_agi_t, agi_unlinked) +
2088 (sizeof(xfs_agino_t) * bucket_index); 2089 (sizeof(xfs_agino_t) * bucket_index);
2090 xfs_trans_buf_set_type(tp, agibp, XFS_BLFT_AGI_BUF);
2089 xfs_trans_log_buf(tp, agibp, offset, 2091 xfs_trans_log_buf(tp, agibp, offset,
2090 (offset + sizeof(xfs_agino_t) - 1)); 2092 (offset + sizeof(xfs_agino_t) - 1));
2091 } else { 2093 } else {
@@ -2656,6 +2658,124 @@ xfs_sort_for_rename(
2656} 2658}
2657 2659
2658/* 2660/*
2661 * xfs_cross_rename()
2662 *
2663 * responsible for handling RENAME_EXCHANGE flag in renameat2() sytemcall
2664 */
2665STATIC int
2666xfs_cross_rename(
2667 struct xfs_trans *tp,
2668 struct xfs_inode *dp1,
2669 struct xfs_name *name1,
2670 struct xfs_inode *ip1,
2671 struct xfs_inode *dp2,
2672 struct xfs_name *name2,
2673 struct xfs_inode *ip2,
2674 struct xfs_bmap_free *free_list,
2675 xfs_fsblock_t *first_block,
2676 int spaceres)
2677{
2678 int error = 0;
2679 int ip1_flags = 0;
2680 int ip2_flags = 0;
2681 int dp2_flags = 0;
2682
2683 /* Swap inode number for dirent in first parent */
2684 error = xfs_dir_replace(tp, dp1, name1,
2685 ip2->i_ino,
2686 first_block, free_list, spaceres);
2687 if (error)
2688 goto out;
2689
2690 /* Swap inode number for dirent in second parent */
2691 error = xfs_dir_replace(tp, dp2, name2,
2692 ip1->i_ino,
2693 first_block, free_list, spaceres);
2694 if (error)
2695 goto out;
2696
2697 /*
2698 * If we're renaming one or more directories across different parents,
2699 * update the respective ".." entries (and link counts) to match the new
2700 * parents.
2701 */
2702 if (dp1 != dp2) {
2703 dp2_flags = XFS_ICHGTIME_MOD | XFS_ICHGTIME_CHG;
2704
2705 if (S_ISDIR(ip2->i_d.di_mode)) {
2706 error = xfs_dir_replace(tp, ip2, &xfs_name_dotdot,
2707 dp1->i_ino, first_block,
2708 free_list, spaceres);
2709 if (error)
2710 goto out;
2711
2712 /* transfer ip2 ".." reference to dp1 */
2713 if (!S_ISDIR(ip1->i_d.di_mode)) {
2714 error = xfs_droplink(tp, dp2);
2715 if (error)
2716 goto out;
2717 error = xfs_bumplink(tp, dp1);
2718 if (error)
2719 goto out;
2720 }
2721
2722 /*
2723 * Although ip1 isn't changed here, userspace needs
2724 * to be warned about the change, so that applications
2725 * relying on it (like backup ones), will properly
2726 * notify the change
2727 */
2728 ip1_flags |= XFS_ICHGTIME_CHG;
2729 ip2_flags |= XFS_ICHGTIME_MOD | XFS_ICHGTIME_CHG;
2730 }
2731
2732 if (S_ISDIR(ip1->i_d.di_mode)) {
2733 error = xfs_dir_replace(tp, ip1, &xfs_name_dotdot,
2734 dp2->i_ino, first_block,
2735 free_list, spaceres);
2736 if (error)
2737 goto out;
2738
2739 /* transfer ip1 ".." reference to dp2 */
2740 if (!S_ISDIR(ip2->i_d.di_mode)) {
2741 error = xfs_droplink(tp, dp1);
2742 if (error)
2743 goto out;
2744 error = xfs_bumplink(tp, dp2);
2745 if (error)
2746 goto out;
2747 }
2748
2749 /*
2750 * Although ip2 isn't changed here, userspace needs
2751 * to be warned about the change, so that applications
2752 * relying on it (like backup ones), will properly
2753 * notify the change
2754 */
2755 ip1_flags |= XFS_ICHGTIME_MOD | XFS_ICHGTIME_CHG;
2756 ip2_flags |= XFS_ICHGTIME_CHG;
2757 }
2758 }
2759
2760 if (ip1_flags) {
2761 xfs_trans_ichgtime(tp, ip1, ip1_flags);
2762 xfs_trans_log_inode(tp, ip1, XFS_ILOG_CORE);
2763 }
2764 if (ip2_flags) {
2765 xfs_trans_ichgtime(tp, ip2, ip2_flags);
2766 xfs_trans_log_inode(tp, ip2, XFS_ILOG_CORE);
2767 }
2768 if (dp2_flags) {
2769 xfs_trans_ichgtime(tp, dp2, dp2_flags);
2770 xfs_trans_log_inode(tp, dp2, XFS_ILOG_CORE);
2771 }
2772 xfs_trans_ichgtime(tp, dp1, XFS_ICHGTIME_MOD | XFS_ICHGTIME_CHG);
2773 xfs_trans_log_inode(tp, dp1, XFS_ILOG_CORE);
2774out:
2775 return error;
2776}
2777
2778/*
2659 * xfs_rename 2779 * xfs_rename
2660 */ 2780 */
2661int 2781int
@@ -2665,7 +2785,8 @@ xfs_rename(
2665 xfs_inode_t *src_ip, 2785 xfs_inode_t *src_ip,
2666 xfs_inode_t *target_dp, 2786 xfs_inode_t *target_dp,
2667 struct xfs_name *target_name, 2787 struct xfs_name *target_name,
2668 xfs_inode_t *target_ip) 2788 xfs_inode_t *target_ip,
2789 unsigned int flags)
2669{ 2790{
2670 xfs_trans_t *tp = NULL; 2791 xfs_trans_t *tp = NULL;
2671 xfs_mount_t *mp = src_dp->i_mount; 2792 xfs_mount_t *mp = src_dp->i_mount;
@@ -2743,6 +2864,18 @@ xfs_rename(
2743 } 2864 }
2744 2865
2745 /* 2866 /*
2867 * Handle RENAME_EXCHANGE flags
2868 */
2869 if (flags & RENAME_EXCHANGE) {
2870 error = xfs_cross_rename(tp, src_dp, src_name, src_ip,
2871 target_dp, target_name, target_ip,
2872 &free_list, &first_block, spaceres);
2873 if (error)
2874 goto abort_return;
2875 goto finish_rename;
2876 }
2877
2878 /*
2746 * Set up the target. 2879 * Set up the target.
2747 */ 2880 */
2748 if (target_ip == NULL) { 2881 if (target_ip == NULL) {
@@ -2881,6 +3014,7 @@ xfs_rename(
2881 if (new_parent) 3014 if (new_parent)
2882 xfs_trans_log_inode(tp, target_dp, XFS_ILOG_CORE); 3015 xfs_trans_log_inode(tp, target_dp, XFS_ILOG_CORE);
2883 3016
3017finish_rename:
2884 /* 3018 /*
2885 * If this is a synchronous mount, make sure that the 3019 * If this is a synchronous mount, make sure that the
2886 * rename transaction goes to disk before returning to 3020 * rename transaction goes to disk before returning to
diff --git a/fs/xfs/xfs_inode.h b/fs/xfs/xfs_inode.h
index 4ed2ba9342dc..86cd6b39bed7 100644
--- a/fs/xfs/xfs_inode.h
+++ b/fs/xfs/xfs_inode.h
@@ -338,7 +338,7 @@ int xfs_link(struct xfs_inode *tdp, struct xfs_inode *sip,
338int xfs_rename(struct xfs_inode *src_dp, struct xfs_name *src_name, 338int xfs_rename(struct xfs_inode *src_dp, struct xfs_name *src_name,
339 struct xfs_inode *src_ip, struct xfs_inode *target_dp, 339 struct xfs_inode *src_ip, struct xfs_inode *target_dp,
340 struct xfs_name *target_name, 340 struct xfs_name *target_name,
341 struct xfs_inode *target_ip); 341 struct xfs_inode *target_ip, unsigned int flags);
342 342
343void xfs_ilock(xfs_inode_t *, uint); 343void xfs_ilock(xfs_inode_t *, uint);
344int xfs_ilock_nowait(xfs_inode_t *, uint); 344int xfs_ilock_nowait(xfs_inode_t *, uint);
@@ -377,6 +377,15 @@ int xfs_droplink(struct xfs_trans *, struct xfs_inode *);
377int xfs_bumplink(struct xfs_trans *, struct xfs_inode *); 377int xfs_bumplink(struct xfs_trans *, struct xfs_inode *);
378 378
379/* from xfs_file.c */ 379/* from xfs_file.c */
380enum xfs_prealloc_flags {
381 XFS_PREALLOC_SET = (1 << 1),
382 XFS_PREALLOC_CLEAR = (1 << 2),
383 XFS_PREALLOC_SYNC = (1 << 3),
384 XFS_PREALLOC_INVISIBLE = (1 << 4),
385};
386
387int xfs_update_prealloc_flags(struct xfs_inode *,
388 enum xfs_prealloc_flags);
380int xfs_zero_eof(struct xfs_inode *, xfs_off_t, xfs_fsize_t); 389int xfs_zero_eof(struct xfs_inode *, xfs_off_t, xfs_fsize_t);
381int xfs_iozero(struct xfs_inode *, loff_t, size_t); 390int xfs_iozero(struct xfs_inode *, loff_t, size_t);
382 391
diff --git a/fs/xfs/xfs_ioctl.c b/fs/xfs/xfs_ioctl.c
index f1dc90be508c..f7afb86c9148 100644
--- a/fs/xfs/xfs_ioctl.c
+++ b/fs/xfs/xfs_ioctl.c
@@ -606,11 +606,8 @@ xfs_ioc_space(
606 unsigned int cmd, 606 unsigned int cmd,
607 xfs_flock64_t *bf) 607 xfs_flock64_t *bf)
608{ 608{
609 struct xfs_mount *mp = ip->i_mount;
610 struct xfs_trans *tp;
611 struct iattr iattr; 609 struct iattr iattr;
612 bool setprealloc = false; 610 enum xfs_prealloc_flags flags = 0;
613 bool clrprealloc = false;
614 int error; 611 int error;
615 612
616 /* 613 /*
@@ -630,6 +627,11 @@ xfs_ioc_space(
630 if (!S_ISREG(inode->i_mode)) 627 if (!S_ISREG(inode->i_mode))
631 return -EINVAL; 628 return -EINVAL;
632 629
630 if (filp->f_flags & O_DSYNC)
631 flags |= XFS_PREALLOC_SYNC;
632 if (ioflags & XFS_IO_INVIS)
633 flags |= XFS_PREALLOC_INVISIBLE;
634
633 error = mnt_want_write_file(filp); 635 error = mnt_want_write_file(filp);
634 if (error) 636 if (error)
635 return error; 637 return error;
@@ -673,25 +675,23 @@ xfs_ioc_space(
673 } 675 }
674 676
675 if (bf->l_start < 0 || 677 if (bf->l_start < 0 ||
676 bf->l_start > mp->m_super->s_maxbytes || 678 bf->l_start > inode->i_sb->s_maxbytes ||
677 bf->l_start + bf->l_len < 0 || 679 bf->l_start + bf->l_len < 0 ||
678 bf->l_start + bf->l_len >= mp->m_super->s_maxbytes) { 680 bf->l_start + bf->l_len >= inode->i_sb->s_maxbytes) {
679 error = -EINVAL; 681 error = -EINVAL;
680 goto out_unlock; 682 goto out_unlock;
681 } 683 }
682 684
683 switch (cmd) { 685 switch (cmd) {
684 case XFS_IOC_ZERO_RANGE: 686 case XFS_IOC_ZERO_RANGE:
687 flags |= XFS_PREALLOC_SET;
685 error = xfs_zero_file_space(ip, bf->l_start, bf->l_len); 688 error = xfs_zero_file_space(ip, bf->l_start, bf->l_len);
686 if (!error)
687 setprealloc = true;
688 break; 689 break;
689 case XFS_IOC_RESVSP: 690 case XFS_IOC_RESVSP:
690 case XFS_IOC_RESVSP64: 691 case XFS_IOC_RESVSP64:
692 flags |= XFS_PREALLOC_SET;
691 error = xfs_alloc_file_space(ip, bf->l_start, bf->l_len, 693 error = xfs_alloc_file_space(ip, bf->l_start, bf->l_len,
692 XFS_BMAPI_PREALLOC); 694 XFS_BMAPI_PREALLOC);
693 if (!error)
694 setprealloc = true;
695 break; 695 break;
696 case XFS_IOC_UNRESVSP: 696 case XFS_IOC_UNRESVSP:
697 case XFS_IOC_UNRESVSP64: 697 case XFS_IOC_UNRESVSP64:
@@ -701,6 +701,7 @@ xfs_ioc_space(
701 case XFS_IOC_ALLOCSP64: 701 case XFS_IOC_ALLOCSP64:
702 case XFS_IOC_FREESP: 702 case XFS_IOC_FREESP:
703 case XFS_IOC_FREESP64: 703 case XFS_IOC_FREESP64:
704 flags |= XFS_PREALLOC_CLEAR;
704 if (bf->l_start > XFS_ISIZE(ip)) { 705 if (bf->l_start > XFS_ISIZE(ip)) {
705 error = xfs_alloc_file_space(ip, XFS_ISIZE(ip), 706 error = xfs_alloc_file_space(ip, XFS_ISIZE(ip),
706 bf->l_start - XFS_ISIZE(ip), 0); 707 bf->l_start - XFS_ISIZE(ip), 0);
@@ -712,8 +713,6 @@ xfs_ioc_space(
712 iattr.ia_size = bf->l_start; 713 iattr.ia_size = bf->l_start;
713 714
714 error = xfs_setattr_size(ip, &iattr); 715 error = xfs_setattr_size(ip, &iattr);
715 if (!error)
716 clrprealloc = true;
717 break; 716 break;
718 default: 717 default:
719 ASSERT(0); 718 ASSERT(0);
@@ -723,32 +722,7 @@ xfs_ioc_space(
723 if (error) 722 if (error)
724 goto out_unlock; 723 goto out_unlock;
725 724
726 tp = xfs_trans_alloc(mp, XFS_TRANS_WRITEID); 725 error = xfs_update_prealloc_flags(ip, flags);
727 error = xfs_trans_reserve(tp, &M_RES(mp)->tr_writeid, 0, 0);
728 if (error) {
729 xfs_trans_cancel(tp, 0);
730 goto out_unlock;
731 }
732
733 xfs_ilock(ip, XFS_ILOCK_EXCL);
734 xfs_trans_ijoin(tp, ip, XFS_ILOCK_EXCL);
735
736 if (!(ioflags & XFS_IO_INVIS)) {
737 ip->i_d.di_mode &= ~S_ISUID;
738 if (ip->i_d.di_mode & S_IXGRP)
739 ip->i_d.di_mode &= ~S_ISGID;
740 xfs_trans_ichgtime(tp, ip, XFS_ICHGTIME_MOD | XFS_ICHGTIME_CHG);
741 }
742
743 if (setprealloc)
744 ip->i_d.di_flags |= XFS_DIFLAG_PREALLOC;
745 else if (clrprealloc)
746 ip->i_d.di_flags &= ~XFS_DIFLAG_PREALLOC;
747
748 xfs_trans_log_inode(tp, ip, XFS_ILOG_CORE);
749 if (filp->f_flags & O_DSYNC)
750 xfs_trans_set_sync(tp);
751 error = xfs_trans_commit(tp, 0);
752 726
753out_unlock: 727out_unlock:
754 xfs_iunlock(ip, XFS_IOLOCK_EXCL); 728 xfs_iunlock(ip, XFS_IOLOCK_EXCL);
diff --git a/fs/xfs/xfs_ioctl32.c b/fs/xfs/xfs_ioctl32.c
index ec6772866f3d..bfc7c7c8a0c8 100644
--- a/fs/xfs/xfs_ioctl32.c
+++ b/fs/xfs/xfs_ioctl32.c
@@ -423,7 +423,7 @@ xfs_compat_attrmulti_by_handle(
423 423
424 ops = memdup_user(compat_ptr(am_hreq.ops), size); 424 ops = memdup_user(compat_ptr(am_hreq.ops), size);
425 if (IS_ERR(ops)) { 425 if (IS_ERR(ops)) {
426 error = -PTR_ERR(ops); 426 error = PTR_ERR(ops);
427 goto out_dput; 427 goto out_dput;
428 } 428 }
429 429
diff --git a/fs/xfs/xfs_iomap.c b/fs/xfs/xfs_iomap.c
index c980e2a5086b..ccb1dd0d509e 100644
--- a/fs/xfs/xfs_iomap.c
+++ b/fs/xfs/xfs_iomap.c
@@ -802,7 +802,7 @@ int
802xfs_iomap_write_unwritten( 802xfs_iomap_write_unwritten(
803 xfs_inode_t *ip, 803 xfs_inode_t *ip,
804 xfs_off_t offset, 804 xfs_off_t offset,
805 size_t count) 805 xfs_off_t count)
806{ 806{
807 xfs_mount_t *mp = ip->i_mount; 807 xfs_mount_t *mp = ip->i_mount;
808 xfs_fileoff_t offset_fsb; 808 xfs_fileoff_t offset_fsb;
diff --git a/fs/xfs/xfs_iomap.h b/fs/xfs/xfs_iomap.h
index 411fbb8919ef..8688e663d744 100644
--- a/fs/xfs/xfs_iomap.h
+++ b/fs/xfs/xfs_iomap.h
@@ -27,6 +27,6 @@ int xfs_iomap_write_delay(struct xfs_inode *, xfs_off_t, size_t,
27 struct xfs_bmbt_irec *); 27 struct xfs_bmbt_irec *);
28int xfs_iomap_write_allocate(struct xfs_inode *, xfs_off_t, 28int xfs_iomap_write_allocate(struct xfs_inode *, xfs_off_t,
29 struct xfs_bmbt_irec *); 29 struct xfs_bmbt_irec *);
30int xfs_iomap_write_unwritten(struct xfs_inode *, xfs_off_t, size_t); 30int xfs_iomap_write_unwritten(struct xfs_inode *, xfs_off_t, xfs_off_t);
31 31
32#endif /* __XFS_IOMAP_H__*/ 32#endif /* __XFS_IOMAP_H__*/
diff --git a/fs/xfs/xfs_iops.c b/fs/xfs/xfs_iops.c
index c50311cae1b1..ce80eeb8faa4 100644
--- a/fs/xfs/xfs_iops.c
+++ b/fs/xfs/xfs_iops.c
@@ -380,18 +380,27 @@ xfs_vn_rename(
380 struct inode *odir, 380 struct inode *odir,
381 struct dentry *odentry, 381 struct dentry *odentry,
382 struct inode *ndir, 382 struct inode *ndir,
383 struct dentry *ndentry) 383 struct dentry *ndentry,
384 unsigned int flags)
384{ 385{
385 struct inode *new_inode = ndentry->d_inode; 386 struct inode *new_inode = ndentry->d_inode;
387 int omode = 0;
386 struct xfs_name oname; 388 struct xfs_name oname;
387 struct xfs_name nname; 389 struct xfs_name nname;
388 390
389 xfs_dentry_to_name(&oname, odentry, 0); 391 if (flags & ~(RENAME_NOREPLACE | RENAME_EXCHANGE))
392 return -EINVAL;
393
394 /* if we are exchanging files, we need to set i_mode of both files */
395 if (flags & RENAME_EXCHANGE)
396 omode = ndentry->d_inode->i_mode;
397
398 xfs_dentry_to_name(&oname, odentry, omode);
390 xfs_dentry_to_name(&nname, ndentry, odentry->d_inode->i_mode); 399 xfs_dentry_to_name(&nname, ndentry, odentry->d_inode->i_mode);
391 400
392 return xfs_rename(XFS_I(odir), &oname, XFS_I(odentry->d_inode), 401 return xfs_rename(XFS_I(odir), &oname, XFS_I(odentry->d_inode),
393 XFS_I(ndir), &nname, new_inode ? 402 XFS_I(ndir), &nname,
394 XFS_I(new_inode) : NULL); 403 new_inode ? XFS_I(new_inode) : NULL, flags);
395} 404}
396 405
397/* 406/*
@@ -1144,7 +1153,7 @@ static const struct inode_operations xfs_dir_inode_operations = {
1144 */ 1153 */
1145 .rmdir = xfs_vn_unlink, 1154 .rmdir = xfs_vn_unlink,
1146 .mknod = xfs_vn_mknod, 1155 .mknod = xfs_vn_mknod,
1147 .rename = xfs_vn_rename, 1156 .rename2 = xfs_vn_rename,
1148 .get_acl = xfs_get_acl, 1157 .get_acl = xfs_get_acl,
1149 .set_acl = xfs_set_acl, 1158 .set_acl = xfs_set_acl,
1150 .getattr = xfs_vn_getattr, 1159 .getattr = xfs_vn_getattr,
@@ -1172,7 +1181,7 @@ static const struct inode_operations xfs_dir_ci_inode_operations = {
1172 */ 1181 */
1173 .rmdir = xfs_vn_unlink, 1182 .rmdir = xfs_vn_unlink,
1174 .mknod = xfs_vn_mknod, 1183 .mknod = xfs_vn_mknod,
1175 .rename = xfs_vn_rename, 1184 .rename2 = xfs_vn_rename,
1176 .get_acl = xfs_get_acl, 1185 .get_acl = xfs_get_acl,
1177 .set_acl = xfs_set_acl, 1186 .set_acl = xfs_set_acl,
1178 .getattr = xfs_vn_getattr, 1187 .getattr = xfs_vn_getattr,
diff --git a/fs/xfs/xfs_log.c b/fs/xfs/xfs_log.c
index 2b8dcf2b3dd1..bcc7cfabb787 100644
--- a/fs/xfs/xfs_log.c
+++ b/fs/xfs/xfs_log.c
@@ -1407,6 +1407,8 @@ xlog_alloc_log(
1407 ASSERT(xfs_buf_islocked(bp)); 1407 ASSERT(xfs_buf_islocked(bp));
1408 xfs_buf_unlock(bp); 1408 xfs_buf_unlock(bp);
1409 1409
1410 /* use high priority wq for log I/O completion */
1411 bp->b_ioend_wq = mp->m_log_workqueue;
1410 bp->b_iodone = xlog_iodone; 1412 bp->b_iodone = xlog_iodone;
1411 log->l_xbuf = bp; 1413 log->l_xbuf = bp;
1412 1414
@@ -1439,6 +1441,8 @@ xlog_alloc_log(
1439 ASSERT(xfs_buf_islocked(bp)); 1441 ASSERT(xfs_buf_islocked(bp));
1440 xfs_buf_unlock(bp); 1442 xfs_buf_unlock(bp);
1441 1443
1444 /* use high priority wq for log I/O completion */
1445 bp->b_ioend_wq = mp->m_log_workqueue;
1442 bp->b_iodone = xlog_iodone; 1446 bp->b_iodone = xlog_iodone;
1443 iclog->ic_bp = bp; 1447 iclog->ic_bp = bp;
1444 iclog->ic_data = bp->b_addr; 1448 iclog->ic_data = bp->b_addr;
@@ -1818,8 +1822,6 @@ xlog_sync(
1818 XFS_BUF_ZEROFLAGS(bp); 1822 XFS_BUF_ZEROFLAGS(bp);
1819 XFS_BUF_ASYNC(bp); 1823 XFS_BUF_ASYNC(bp);
1820 bp->b_flags |= XBF_SYNCIO; 1824 bp->b_flags |= XBF_SYNCIO;
1821 /* use high priority completion wq */
1822 bp->b_ioend_wq = log->l_mp->m_log_workqueue;
1823 1825
1824 if (log->l_mp->m_flags & XFS_MOUNT_BARRIER) { 1826 if (log->l_mp->m_flags & XFS_MOUNT_BARRIER) {
1825 bp->b_flags |= XBF_FUA; 1827 bp->b_flags |= XBF_FUA;
@@ -1868,8 +1870,6 @@ xlog_sync(
1868 bp->b_flags |= XBF_SYNCIO; 1870 bp->b_flags |= XBF_SYNCIO;
1869 if (log->l_mp->m_flags & XFS_MOUNT_BARRIER) 1871 if (log->l_mp->m_flags & XFS_MOUNT_BARRIER)
1870 bp->b_flags |= XBF_FUA; 1872 bp->b_flags |= XBF_FUA;
1871 /* use high priority completion wq */
1872 bp->b_ioend_wq = log->l_mp->m_log_workqueue;
1873 1873
1874 ASSERT(XFS_BUF_ADDR(bp) <= log->l_logBBsize-1); 1874 ASSERT(XFS_BUF_ADDR(bp) <= log->l_logBBsize-1);
1875 ASSERT(XFS_BUF_ADDR(bp) + BTOBB(count) <= log->l_logBBsize); 1875 ASSERT(XFS_BUF_ADDR(bp) + BTOBB(count) <= log->l_logBBsize);
@@ -2039,7 +2039,7 @@ xlog_print_tic_res(
2039 " total reg = %u bytes (o/flow = %u bytes)\n" 2039 " total reg = %u bytes (o/flow = %u bytes)\n"
2040 " ophdrs = %u (ophdr space = %u bytes)\n" 2040 " ophdrs = %u (ophdr space = %u bytes)\n"
2041 " ophdr + reg = %u bytes\n" 2041 " ophdr + reg = %u bytes\n"
2042 " num regions = %u\n", 2042 " num regions = %u",
2043 ((ticket->t_trans_type <= 0 || 2043 ((ticket->t_trans_type <= 0 ||
2044 ticket->t_trans_type > XFS_TRANS_TYPE_MAX) ? 2044 ticket->t_trans_type > XFS_TRANS_TYPE_MAX) ?
2045 "bad-trans-type" : trans_type_str[ticket->t_trans_type-1]), 2045 "bad-trans-type" : trans_type_str[ticket->t_trans_type-1]),
diff --git a/fs/xfs/xfs_super.c b/fs/xfs/xfs_super.c
index 26afa436fafe..f2449fd86926 100644
--- a/fs/xfs/xfs_super.c
+++ b/fs/xfs/xfs_super.c
@@ -685,7 +685,7 @@ xfs_blkdev_get(
685 mp); 685 mp);
686 if (IS_ERR(*bdevp)) { 686 if (IS_ERR(*bdevp)) {
687 error = PTR_ERR(*bdevp); 687 error = PTR_ERR(*bdevp);
688 xfs_warn(mp, "Invalid device [%s], error=%d\n", name, error); 688 xfs_warn(mp, "Invalid device [%s], error=%d", name, error);
689 } 689 }
690 690
691 return error; 691 return error;
diff --git a/fs/xfs/xfs_sysctl.c b/fs/xfs/xfs_sysctl.c
index 1743b9f8e23d..a0c8067cea6f 100644
--- a/fs/xfs/xfs_sysctl.c
+++ b/fs/xfs/xfs_sysctl.c
@@ -149,24 +149,6 @@ static struct ctl_table xfs_table[] = {
149 .extra2 = &xfs_params.inherit_noatim.max 149 .extra2 = &xfs_params.inherit_noatim.max
150 }, 150 },
151 { 151 {
152 .procname = "xfsbufd_centisecs",
153 .data = &xfs_params.xfs_buf_timer.val,
154 .maxlen = sizeof(int),
155 .mode = 0644,
156 .proc_handler = proc_dointvec_minmax,
157 .extra1 = &xfs_params.xfs_buf_timer.min,
158 .extra2 = &xfs_params.xfs_buf_timer.max
159 },
160 {
161 .procname = "age_buffer_centisecs",
162 .data = &xfs_params.xfs_buf_age.val,
163 .maxlen = sizeof(int),
164 .mode = 0644,
165 .proc_handler = proc_dointvec_minmax,
166 .extra1 = &xfs_params.xfs_buf_age.min,
167 .extra2 = &xfs_params.xfs_buf_age.max
168 },
169 {
170 .procname = "inherit_nosymlinks", 152 .procname = "inherit_nosymlinks",
171 .data = &xfs_params.inherit_nosym.val, 153 .data = &xfs_params.inherit_nosym.val,
172 .maxlen = sizeof(int), 154 .maxlen = sizeof(int),
diff --git a/fs/xfs/xfs_trans.c b/fs/xfs/xfs_trans.c
index fa3135b9bf04..eb90cd59a0ec 100644
--- a/fs/xfs/xfs_trans.c
+++ b/fs/xfs/xfs_trans.c
@@ -472,6 +472,7 @@ xfs_trans_apply_sb_deltas(
472 whole = 1; 472 whole = 1;
473 } 473 }
474 474
475 xfs_trans_buf_set_type(tp, bp, XFS_BLFT_SB_BUF);
475 if (whole) 476 if (whole)
476 /* 477 /*
477 * Log the whole thing, the fields are noncontiguous. 478 * Log the whole thing, the fields are noncontiguous.