]> bbs.cooldavid.org Git - net-next-2.6.git/blame - fs/xfs/linux-2.6/xfs_file.c
mm: fix fault vs invalidate race for linear mappings
[net-next-2.6.git] / fs / xfs / linux-2.6 / xfs_file.c
CommitLineData
1da177e4 1/*
7b718769
NS
2 * Copyright (c) 2000-2005 Silicon Graphics, Inc.
3 * All Rights Reserved.
1da177e4 4 *
7b718769
NS
5 * This program is free software; you can redistribute it and/or
6 * modify it under the terms of the GNU General Public License as
1da177e4
LT
7 * published by the Free Software Foundation.
8 *
7b718769
NS
9 * This program is distributed in the hope that it would be useful,
10 * but WITHOUT ANY WARRANTY; without even the implied warranty of
11 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
12 * GNU General Public License for more details.
1da177e4 13 *
7b718769
NS
14 * You should have received a copy of the GNU General Public License
15 * along with this program; if not, write the Free Software Foundation,
16 * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
1da177e4 17 */
1da177e4 18#include "xfs.h"
a844f451 19#include "xfs_bit.h"
1da177e4 20#include "xfs_log.h"
a844f451 21#include "xfs_inum.h"
1da177e4 22#include "xfs_sb.h"
a844f451 23#include "xfs_ag.h"
1da177e4
LT
24#include "xfs_dir2.h"
25#include "xfs_trans.h"
26#include "xfs_dmapi.h"
27#include "xfs_mount.h"
28#include "xfs_bmap_btree.h"
29#include "xfs_alloc_btree.h"
30#include "xfs_ialloc_btree.h"
31#include "xfs_alloc.h"
32#include "xfs_btree.h"
33#include "xfs_attr_sf.h"
1da177e4
LT
34#include "xfs_dir2_sf.h"
35#include "xfs_dinode.h"
36#include "xfs_inode.h"
37#include "xfs_error.h"
38#include "xfs_rw.h"
39#include "xfs_ioctl32.h"
40
41#include <linux/dcache.h>
42#include <linux/smp_lock.h>
43
3562fd45 44static struct vm_operations_struct xfs_file_vm_ops;
6fac0cb4 45#ifdef CONFIG_XFS_DMAPI
3562fd45 46static struct vm_operations_struct xfs_dmapi_file_vm_ops;
6fac0cb4 47#endif
1da177e4 48
7989cb8e 49STATIC_INLINE ssize_t
3562fd45 50__xfs_file_read(
1da177e4 51 struct kiocb *iocb,
027445c3
BP
52 const struct iovec *iov,
53 unsigned long nr_segs,
1da177e4 54 int ioflags,
1da177e4
LT
55 loff_t pos)
56{
1da177e4 57 struct file *file = iocb->ki_filp;
e678fb0d 58 bhv_vnode_t *vp = vn_from_inode(file->f_path.dentry->d_inode);
1da177e4
LT
59
60 BUG_ON(iocb->ki_pos != pos);
1da177e4
LT
61 if (unlikely(file->f_flags & O_DIRECT))
62 ioflags |= IO_ISDIRECT;
027445c3
BP
63 return bhv_vop_read(vp, iocb, iov, nr_segs, &iocb->ki_pos,
64 ioflags, NULL);
1da177e4
LT
65}
66
1da177e4 67STATIC ssize_t
3562fd45 68xfs_file_aio_read(
1da177e4 69 struct kiocb *iocb,
027445c3
BP
70 const struct iovec *iov,
71 unsigned long nr_segs,
1da177e4
LT
72 loff_t pos)
73{
027445c3 74 return __xfs_file_read(iocb, iov, nr_segs, IO_ISAIO, pos);
1da177e4
LT
75}
76
77STATIC ssize_t
3562fd45 78xfs_file_aio_read_invis(
1da177e4 79 struct kiocb *iocb,
027445c3
BP
80 const struct iovec *iov,
81 unsigned long nr_segs,
1da177e4
LT
82 loff_t pos)
83{
027445c3 84 return __xfs_file_read(iocb, iov, nr_segs, IO_ISAIO|IO_INVIS, pos);
1da177e4
LT
85}
86
7989cb8e 87STATIC_INLINE ssize_t
3562fd45 88__xfs_file_write(
027445c3
BP
89 struct kiocb *iocb,
90 const struct iovec *iov,
91 unsigned long nr_segs,
92 int ioflags,
93 loff_t pos)
1da177e4 94{
1da177e4
LT
95 struct file *file = iocb->ki_filp;
96 struct inode *inode = file->f_mapping->host;
67fcaa73 97 bhv_vnode_t *vp = vn_from_inode(inode);
1da177e4
LT
98
99 BUG_ON(iocb->ki_pos != pos);
100 if (unlikely(file->f_flags & O_DIRECT))
101 ioflags |= IO_ISDIRECT;
027445c3
BP
102 return bhv_vop_write(vp, iocb, iov, nr_segs, &iocb->ki_pos,
103 ioflags, NULL);
1da177e4
LT
104}
105
1da177e4 106STATIC ssize_t
3562fd45 107xfs_file_aio_write(
1da177e4 108 struct kiocb *iocb,
027445c3
BP
109 const struct iovec *iov,
110 unsigned long nr_segs,
1da177e4
LT
111 loff_t pos)
112{
027445c3 113 return __xfs_file_write(iocb, iov, nr_segs, IO_ISAIO, pos);
1da177e4
LT
114}
115
116STATIC ssize_t
3562fd45 117xfs_file_aio_write_invis(
1da177e4 118 struct kiocb *iocb,
027445c3
BP
119 const struct iovec *iov,
120 unsigned long nr_segs,
1da177e4
LT
121 loff_t pos)
122{
027445c3 123 return __xfs_file_write(iocb, iov, nr_segs, IO_ISAIO|IO_INVIS, pos);
1da177e4
LT
124}
125
1b895840
NS
126STATIC ssize_t
127xfs_file_splice_read(
128 struct file *infilp,
cbb7e577 129 loff_t *ppos,
3a326a2c 130 struct pipe_inode_info *pipe,
1b895840
NS
131 size_t len,
132 unsigned int flags)
133{
e678fb0d 134 return bhv_vop_splice_read(vn_from_inode(infilp->f_path.dentry->d_inode),
67fcaa73 135 infilp, ppos, pipe, len, flags, 0, NULL);
1b895840
NS
136}
137
138STATIC ssize_t
139xfs_file_splice_read_invis(
140 struct file *infilp,
cbb7e577 141 loff_t *ppos,
3a326a2c 142 struct pipe_inode_info *pipe,
1b895840
NS
143 size_t len,
144 unsigned int flags)
145{
e678fb0d 146 return bhv_vop_splice_read(vn_from_inode(infilp->f_path.dentry->d_inode),
67fcaa73
NS
147 infilp, ppos, pipe, len, flags, IO_INVIS,
148 NULL);
1b895840
NS
149}
150
151STATIC ssize_t
152xfs_file_splice_write(
3a326a2c 153 struct pipe_inode_info *pipe,
1b895840 154 struct file *outfilp,
cbb7e577 155 loff_t *ppos,
1b895840
NS
156 size_t len,
157 unsigned int flags)
158{
e678fb0d 159 return bhv_vop_splice_write(vn_from_inode(outfilp->f_path.dentry->d_inode),
67fcaa73 160 pipe, outfilp, ppos, len, flags, 0, NULL);
1b895840
NS
161}
162
163STATIC ssize_t
164xfs_file_splice_write_invis(
3a326a2c 165 struct pipe_inode_info *pipe,
1b895840 166 struct file *outfilp,
cbb7e577 167 loff_t *ppos,
1b895840
NS
168 size_t len,
169 unsigned int flags)
170{
e678fb0d 171 return bhv_vop_splice_write(vn_from_inode(outfilp->f_path.dentry->d_inode),
67fcaa73
NS
172 pipe, outfilp, ppos, len, flags, IO_INVIS,
173 NULL);
1b895840 174}
1da177e4
LT
175
176STATIC int
3562fd45 177xfs_file_open(
1da177e4
LT
178 struct inode *inode,
179 struct file *filp)
180{
1da177e4
LT
181 if (!(filp->f_flags & O_LARGEFILE) && i_size_read(inode) > MAX_NON_LFS)
182 return -EFBIG;
67fcaa73 183 return -bhv_vop_open(vn_from_inode(inode), NULL);
1da177e4
LT
184}
185
1da177e4 186STATIC int
3562fd45 187xfs_file_release(
1da177e4
LT
188 struct inode *inode,
189 struct file *filp)
190{
67fcaa73 191 bhv_vnode_t *vp = vn_from_inode(inode);
1da177e4
LT
192
193 if (vp)
67fcaa73
NS
194 return -bhv_vop_release(vp);
195 return 0;
1da177e4
LT
196}
197
1da177e4 198STATIC int
3562fd45 199xfs_file_fsync(
1da177e4
LT
200 struct file *filp,
201 struct dentry *dentry,
202 int datasync)
203{
67fcaa73 204 bhv_vnode_t *vp = vn_from_inode(dentry->d_inode);
1da177e4
LT
205 int flags = FSYNC_WAIT;
206
207 if (datasync)
208 flags |= FSYNC_DATA;
7d4fb40a
NS
209 if (VN_TRUNC(vp))
210 VUNTRUNCATE(vp);
67fcaa73 211 return -bhv_vop_fsync(vp, flags, NULL, (xfs_off_t)0, (xfs_off_t)-1);
1da177e4
LT
212}
213
bb3f724e 214#ifdef CONFIG_XFS_DMAPI
bb3f724e 215STATIC struct page *
3562fd45 216xfs_vm_nopage(
bb3f724e
DR
217 struct vm_area_struct *area,
218 unsigned long address,
219 int *type)
220{
e678fb0d 221 struct inode *inode = area->vm_file->f_path.dentry->d_inode;
67fcaa73 222 bhv_vnode_t *vp = vn_from_inode(inode);
bb3f724e
DR
223
224 ASSERT_ALWAYS(vp->v_vfsp->vfs_flag & VFS_DMI);
fbc1462b 225 if (XFS_SEND_MMAP(XFS_VFSTOM(vp->v_vfsp), area, 0))
bb3f724e 226 return NULL;
bb3f724e
DR
227 return filemap_nopage(area, address, type);
228}
bb3f724e
DR
229#endif /* CONFIG_XFS_DMAPI */
230
1da177e4 231STATIC int
3562fd45 232xfs_file_readdir(
1da177e4
LT
233 struct file *filp,
234 void *dirent,
235 filldir_t filldir)
236{
237 int error = 0;
e678fb0d 238 bhv_vnode_t *vp = vn_from_inode(filp->f_path.dentry->d_inode);
1da177e4
LT
239 uio_t uio;
240 iovec_t iov;
241 int eof = 0;
242 caddr_t read_buf;
243 int namelen, size = 0;
244 size_t rlen = PAGE_CACHE_SIZE;
245 xfs_off_t start_offset, curr_offset;
246 xfs_dirent_t *dbp = NULL;
247
1da177e4
LT
248 /* Try fairly hard to get memory */
249 do {
f52720ca 250 if ((read_buf = kmalloc(rlen, GFP_KERNEL)))
1da177e4
LT
251 break;
252 rlen >>= 1;
253 } while (rlen >= 1024);
254
255 if (read_buf == NULL)
256 return -ENOMEM;
257
258 uio.uio_iov = &iov;
259 uio.uio_segflg = UIO_SYSSPACE;
260 curr_offset = filp->f_pos;
261 if (filp->f_pos != 0x7fffffff)
262 uio.uio_offset = filp->f_pos;
263 else
264 uio.uio_offset = 0xffffffff;
265
266 while (!eof) {
267 uio.uio_resid = iov.iov_len = rlen;
268 iov.iov_base = read_buf;
269 uio.uio_iovcnt = 1;
270
271 start_offset = uio.uio_offset;
272
67fcaa73 273 error = bhv_vop_readdir(vp, &uio, NULL, &eof);
1da177e4
LT
274 if ((uio.uio_offset == start_offset) || error) {
275 size = 0;
276 break;
277 }
278
279 size = rlen - uio.uio_resid;
280 dbp = (xfs_dirent_t *)read_buf;
281 while (size > 0) {
282 namelen = strlen(dbp->d_name);
283
284 if (filldir(dirent, dbp->d_name, namelen,
285 (loff_t) curr_offset & 0x7fffffff,
286 (ino_t) dbp->d_ino,
287 DT_UNKNOWN)) {
288 goto done;
289 }
290 size -= dbp->d_reclen;
291 curr_offset = (loff_t)dbp->d_off /* & 0x7fffffff */;
1b895840 292 dbp = (xfs_dirent_t *)((char *)dbp + dbp->d_reclen);
1da177e4
LT
293 }
294 }
295done:
296 if (!error) {
297 if (size == 0)
298 filp->f_pos = uio.uio_offset & 0x7fffffff;
299 else if (dbp)
300 filp->f_pos = curr_offset;
301 }
302
303 kfree(read_buf);
304 return -error;
305}
306
1da177e4 307STATIC int
3562fd45 308xfs_file_mmap(
1da177e4
LT
309 struct file *filp,
310 struct vm_area_struct *vma)
311{
3562fd45 312 vma->vm_ops = &xfs_file_vm_ops;
d00806b1 313 vma->vm_flags |= VM_CAN_INVALIDATE;
6fac0cb4 314
6fac0cb4 315#ifdef CONFIG_XFS_DMAPI
e678fb0d 316 if (vn_from_inode(filp->f_path.dentry->d_inode)->v_vfsp->vfs_flag & VFS_DMI)
3562fd45 317 vma->vm_ops = &xfs_dmapi_file_vm_ops;
bb3f724e 318#endif /* CONFIG_XFS_DMAPI */
1da177e4 319
fbc1462b 320 file_accessed(filp);
1da177e4
LT
321 return 0;
322}
323
1da177e4 324STATIC long
3562fd45 325xfs_file_ioctl(
1da177e4
LT
326 struct file *filp,
327 unsigned int cmd,
67fcaa73 328 unsigned long p)
1da177e4
LT
329{
330 int error;
e678fb0d 331 struct inode *inode = filp->f_path.dentry->d_inode;
67fcaa73 332 bhv_vnode_t *vp = vn_from_inode(inode);
1da177e4 333
67fcaa73 334 error = bhv_vop_ioctl(vp, inode, filp, 0, cmd, (void __user *)p);
1da177e4
LT
335 VMODIFY(vp);
336
337 /* NOTE: some of the ioctl's return positive #'s as a
338 * byte count indicating success, such as
339 * readlink_by_handle. So we don't "sign flip"
340 * like most other routines. This means true
341 * errors need to be returned as a negative value.
342 */
343 return error;
344}
345
346STATIC long
3562fd45 347xfs_file_ioctl_invis(
1da177e4
LT
348 struct file *filp,
349 unsigned int cmd,
67fcaa73 350 unsigned long p)
1da177e4 351{
1b895840 352 int error;
e678fb0d 353 struct inode *inode = filp->f_path.dentry->d_inode;
67fcaa73 354 bhv_vnode_t *vp = vn_from_inode(inode);
1da177e4 355
67fcaa73 356 error = bhv_vop_ioctl(vp, inode, filp, IO_INVIS, cmd, (void __user *)p);
1da177e4
LT
357 VMODIFY(vp);
358
359 /* NOTE: some of the ioctl's return positive #'s as a
360 * byte count indicating success, such as
361 * readlink_by_handle. So we don't "sign flip"
362 * like most other routines. This means true
363 * errors need to be returned as a negative value.
364 */
365 return error;
366}
367
bb3f724e 368#ifdef CONFIG_XFS_DMAPI
1da177e4
LT
369#ifdef HAVE_VMOP_MPROTECT
370STATIC int
3562fd45 371xfs_vm_mprotect(
1da177e4
LT
372 struct vm_area_struct *vma,
373 unsigned int newflags)
374{
e678fb0d 375 bhv_vnode_t *vp = vn_from_inode(vma->vm_file->f_path.dentry->d_inode);
1da177e4
LT
376 int error = 0;
377
378 if (vp->v_vfsp->vfs_flag & VFS_DMI) {
379 if ((vma->vm_flags & VM_MAYSHARE) &&
380 (newflags & VM_WRITE) && !(vma->vm_flags & VM_WRITE)) {
381 xfs_mount_t *mp = XFS_VFSTOM(vp->v_vfsp);
382
383 error = XFS_SEND_MMAP(mp, vma, VM_WRITE);
384 }
385 }
386 return error;
387}
388#endif /* HAVE_VMOP_MPROTECT */
bb3f724e 389#endif /* CONFIG_XFS_DMAPI */
1da177e4
LT
390
391#ifdef HAVE_FOP_OPEN_EXEC
392/* If the user is attempting to execute a file that is offline then
393 * we have to trigger a DMAPI READ event before the file is marked as busy
394 * otherwise the invisible I/O will not be able to write to the file to bring
395 * it back online.
396 */
397STATIC int
3562fd45 398xfs_file_open_exec(
1da177e4
LT
399 struct inode *inode)
400{
67fcaa73 401 bhv_vnode_t *vp = vn_from_inode(inode);
1da177e4 402
1d47bec2
NS
403 if (unlikely(vp->v_vfsp->vfs_flag & VFS_DMI)) {
404 xfs_mount_t *mp = XFS_VFSTOM(vp->v_vfsp);
405 xfs_inode_t *ip = xfs_vtoi(vp);
406
407 if (!ip)
408 return -EINVAL;
409 if (DM_EVENT_ENABLED(vp->v_vfsp, ip, DM_EVENT_READ))
410 return -XFS_SEND_DATA(mp, DM_EVENT_READ, vp,
1da177e4 411 0, 0, 0, NULL);
1da177e4 412 }
1d47bec2 413 return 0;
1da177e4
LT
414}
415#endif /* HAVE_FOP_OPEN_EXEC */
416
4b6f5d20 417const struct file_operations xfs_file_operations = {
1da177e4
LT
418 .llseek = generic_file_llseek,
419 .read = do_sync_read,
bb3f724e 420 .write = do_sync_write,
3562fd45
NS
421 .aio_read = xfs_file_aio_read,
422 .aio_write = xfs_file_aio_write,
1b895840
NS
423 .splice_read = xfs_file_splice_read,
424 .splice_write = xfs_file_splice_write,
3562fd45 425 .unlocked_ioctl = xfs_file_ioctl,
1da177e4 426#ifdef CONFIG_COMPAT
3562fd45 427 .compat_ioctl = xfs_file_compat_ioctl,
1da177e4 428#endif
3562fd45
NS
429 .mmap = xfs_file_mmap,
430 .open = xfs_file_open,
431 .release = xfs_file_release,
432 .fsync = xfs_file_fsync,
1da177e4 433#ifdef HAVE_FOP_OPEN_EXEC
3562fd45 434 .open_exec = xfs_file_open_exec,
1da177e4
LT
435#endif
436};
437
4b6f5d20 438const struct file_operations xfs_invis_file_operations = {
1da177e4
LT
439 .llseek = generic_file_llseek,
440 .read = do_sync_read,
bb3f724e 441 .write = do_sync_write,
3562fd45
NS
442 .aio_read = xfs_file_aio_read_invis,
443 .aio_write = xfs_file_aio_write_invis,
1b895840
NS
444 .splice_read = xfs_file_splice_read_invis,
445 .splice_write = xfs_file_splice_write_invis,
3562fd45 446 .unlocked_ioctl = xfs_file_ioctl_invis,
1da177e4 447#ifdef CONFIG_COMPAT
3562fd45 448 .compat_ioctl = xfs_file_compat_invis_ioctl,
1da177e4 449#endif
3562fd45
NS
450 .mmap = xfs_file_mmap,
451 .open = xfs_file_open,
452 .release = xfs_file_release,
453 .fsync = xfs_file_fsync,
1da177e4
LT
454};
455
456
4b6f5d20 457const struct file_operations xfs_dir_file_operations = {
1da177e4 458 .read = generic_read_dir,
3562fd45
NS
459 .readdir = xfs_file_readdir,
460 .unlocked_ioctl = xfs_file_ioctl,
d3870398 461#ifdef CONFIG_COMPAT
3562fd45 462 .compat_ioctl = xfs_file_compat_ioctl,
d3870398 463#endif
3562fd45 464 .fsync = xfs_file_fsync,
1da177e4
LT
465};
466
3562fd45 467static struct vm_operations_struct xfs_file_vm_ops = {
1da177e4
LT
468 .nopage = filemap_nopage,
469 .populate = filemap_populate,
6fac0cb4
DR
470};
471
472#ifdef CONFIG_XFS_DMAPI
3562fd45
NS
473static struct vm_operations_struct xfs_dmapi_file_vm_ops = {
474 .nopage = xfs_vm_nopage,
6fac0cb4 475 .populate = filemap_populate,
1da177e4 476#ifdef HAVE_VMOP_MPROTECT
3562fd45 477 .mprotect = xfs_vm_mprotect,
1da177e4
LT
478#endif
479};
6fac0cb4 480#endif /* CONFIG_XFS_DMAPI */