Avoid beyond bounds copy while caching ACL
[zen-stable.git] / fs / 9p / vfs_file.c
blobfc06fd27065eb3cade2f5e5f60f7155d3487ae80
1 /*
2 * linux/fs/9p/vfs_file.c
4 * This file contians vfs file ops for 9P2000.
6 * Copyright (C) 2004 by Eric Van Hensbergen <ericvh@gmail.com>
7 * Copyright (C) 2002 by Ron Minnich <rminnich@lanl.gov>
9 * This program is free software; you can redistribute it and/or modify
10 * it under the terms of the GNU General Public License version 2
11 * as published by the Free Software Foundation.
13 * This program is distributed in the hope that it will be useful,
14 * but WITHOUT ANY WARRANTY; without even the implied warranty of
15 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
16 * GNU General Public License for more details.
18 * You should have received a copy of the GNU General Public License
19 * along with this program; if not, write to:
20 * Free Software Foundation
21 * 51 Franklin Street, Fifth Floor
22 * Boston, MA 02111-1301 USA
26 #include <linux/module.h>
27 #include <linux/errno.h>
28 #include <linux/fs.h>
29 #include <linux/sched.h>
30 #include <linux/file.h>
31 #include <linux/stat.h>
32 #include <linux/string.h>
33 #include <linux/inet.h>
34 #include <linux/list.h>
35 #include <linux/pagemap.h>
36 #include <linux/utsname.h>
37 #include <asm/uaccess.h>
38 #include <linux/idr.h>
39 #include <net/9p/9p.h>
40 #include <net/9p/client.h>
42 #include "v9fs.h"
43 #include "v9fs_vfs.h"
44 #include "fid.h"
45 #include "cache.h"
47 static const struct vm_operations_struct v9fs_file_vm_ops;
49 /**
50 * v9fs_file_open - open a file (or directory)
51 * @inode: inode to be opened
52 * @file: file being opened
56 int v9fs_file_open(struct inode *inode, struct file *file)
58 int err;
59 struct v9fs_inode *v9inode;
60 struct v9fs_session_info *v9ses;
61 struct p9_fid *fid;
62 int omode;
64 p9_debug(P9_DEBUG_VFS, "inode: %p file: %p\n", inode, file);
65 v9inode = V9FS_I(inode);
66 v9ses = v9fs_inode2v9ses(inode);
67 if (v9fs_proto_dotl(v9ses))
68 omode = v9fs_open_to_dotl_flags(file->f_flags);
69 else
70 omode = v9fs_uflags2omode(file->f_flags,
71 v9fs_proto_dotu(v9ses));
72 fid = file->private_data;
73 if (!fid) {
74 fid = v9fs_fid_clone(file->f_path.dentry);
75 if (IS_ERR(fid))
76 return PTR_ERR(fid);
78 err = p9_client_open(fid, omode);
79 if (err < 0) {
80 p9_client_clunk(fid);
81 return err;
83 if (file->f_flags & O_TRUNC) {
84 i_size_write(inode, 0);
85 inode->i_blocks = 0;
87 if ((file->f_flags & O_APPEND) &&
88 (!v9fs_proto_dotu(v9ses) && !v9fs_proto_dotl(v9ses)))
89 generic_file_llseek(file, 0, SEEK_END);
92 file->private_data = fid;
93 mutex_lock(&v9inode->v_mutex);
94 if (v9ses->cache && !v9inode->writeback_fid &&
95 ((file->f_flags & O_ACCMODE) != O_RDONLY)) {
97 * clone a fid and add it to writeback_fid
98 * we do it during open time instead of
99 * page dirty time via write_begin/page_mkwrite
100 * because we want write after unlink usecase
101 * to work.
103 fid = v9fs_writeback_fid(file->f_path.dentry);
104 if (IS_ERR(fid)) {
105 err = PTR_ERR(fid);
106 mutex_unlock(&v9inode->v_mutex);
107 goto out_error;
109 v9inode->writeback_fid = (void *) fid;
111 mutex_unlock(&v9inode->v_mutex);
112 #ifdef CONFIG_9P_FSCACHE
113 if (v9ses->cache)
114 v9fs_cache_inode_set_cookie(inode, file);
115 #endif
116 return 0;
117 out_error:
118 p9_client_clunk(file->private_data);
119 file->private_data = NULL;
120 return err;
124 * v9fs_file_lock - lock a file (or directory)
125 * @filp: file to be locked
126 * @cmd: lock command
127 * @fl: file lock structure
129 * Bugs: this looks like a local only lock, we should extend into 9P
130 * by using open exclusive
133 static int v9fs_file_lock(struct file *filp, int cmd, struct file_lock *fl)
135 int res = 0;
136 struct inode *inode = filp->f_path.dentry->d_inode;
138 p9_debug(P9_DEBUG_VFS, "filp: %p lock: %p\n", filp, fl);
140 /* No mandatory locks */
141 if (__mandatory_lock(inode) && fl->fl_type != F_UNLCK)
142 return -ENOLCK;
144 if ((IS_SETLK(cmd) || IS_SETLKW(cmd)) && fl->fl_type != F_UNLCK) {
145 filemap_write_and_wait(inode->i_mapping);
146 invalidate_mapping_pages(&inode->i_data, 0, -1);
149 return res;
152 static int v9fs_file_do_lock(struct file *filp, int cmd, struct file_lock *fl)
154 struct p9_flock flock;
155 struct p9_fid *fid;
156 uint8_t status;
157 int res = 0;
158 unsigned char fl_type;
160 fid = filp->private_data;
161 BUG_ON(fid == NULL);
163 if ((fl->fl_flags & FL_POSIX) != FL_POSIX)
164 BUG();
166 res = posix_lock_file_wait(filp, fl);
167 if (res < 0)
168 goto out;
170 /* convert posix lock to p9 tlock args */
171 memset(&flock, 0, sizeof(flock));
172 /* map the lock type */
173 switch (fl->fl_type) {
174 case F_RDLCK:
175 flock.type = P9_LOCK_TYPE_RDLCK;
176 break;
177 case F_WRLCK:
178 flock.type = P9_LOCK_TYPE_WRLCK;
179 break;
180 case F_UNLCK:
181 flock.type = P9_LOCK_TYPE_UNLCK;
182 break;
184 flock.start = fl->fl_start;
185 if (fl->fl_end == OFFSET_MAX)
186 flock.length = 0;
187 else
188 flock.length = fl->fl_end - fl->fl_start + 1;
189 flock.proc_id = fl->fl_pid;
190 flock.client_id = utsname()->nodename;
191 if (IS_SETLKW(cmd))
192 flock.flags = P9_LOCK_FLAGS_BLOCK;
195 * if its a blocked request and we get P9_LOCK_BLOCKED as the status
196 * for lock request, keep on trying
198 for (;;) {
199 res = p9_client_lock_dotl(fid, &flock, &status);
200 if (res < 0)
201 break;
203 if (status != P9_LOCK_BLOCKED)
204 break;
205 if (status == P9_LOCK_BLOCKED && !IS_SETLKW(cmd))
206 break;
207 if (schedule_timeout_interruptible(P9_LOCK_TIMEOUT) != 0)
208 break;
211 /* map 9p status to VFS status */
212 switch (status) {
213 case P9_LOCK_SUCCESS:
214 res = 0;
215 break;
216 case P9_LOCK_BLOCKED:
217 res = -EAGAIN;
218 break;
219 case P9_LOCK_ERROR:
220 case P9_LOCK_GRACE:
221 res = -ENOLCK;
222 break;
223 default:
224 BUG();
228 * incase server returned error for lock request, revert
229 * it locally
231 if (res < 0 && fl->fl_type != F_UNLCK) {
232 fl_type = fl->fl_type;
233 fl->fl_type = F_UNLCK;
234 res = posix_lock_file_wait(filp, fl);
235 fl->fl_type = fl_type;
237 out:
238 return res;
241 static int v9fs_file_getlock(struct file *filp, struct file_lock *fl)
243 struct p9_getlock glock;
244 struct p9_fid *fid;
245 int res = 0;
247 fid = filp->private_data;
248 BUG_ON(fid == NULL);
250 posix_test_lock(filp, fl);
252 * if we have a conflicting lock locally, no need to validate
253 * with server
255 if (fl->fl_type != F_UNLCK)
256 return res;
258 /* convert posix lock to p9 tgetlock args */
259 memset(&glock, 0, sizeof(glock));
260 glock.type = P9_LOCK_TYPE_UNLCK;
261 glock.start = fl->fl_start;
262 if (fl->fl_end == OFFSET_MAX)
263 glock.length = 0;
264 else
265 glock.length = fl->fl_end - fl->fl_start + 1;
266 glock.proc_id = fl->fl_pid;
267 glock.client_id = utsname()->nodename;
269 res = p9_client_getlock_dotl(fid, &glock);
270 if (res < 0)
271 return res;
272 /* map 9p lock type to os lock type */
273 switch (glock.type) {
274 case P9_LOCK_TYPE_RDLCK:
275 fl->fl_type = F_RDLCK;
276 break;
277 case P9_LOCK_TYPE_WRLCK:
278 fl->fl_type = F_WRLCK;
279 break;
280 case P9_LOCK_TYPE_UNLCK:
281 fl->fl_type = F_UNLCK;
282 break;
284 if (glock.type != P9_LOCK_TYPE_UNLCK) {
285 fl->fl_start = glock.start;
286 if (glock.length == 0)
287 fl->fl_end = OFFSET_MAX;
288 else
289 fl->fl_end = glock.start + glock.length - 1;
290 fl->fl_pid = glock.proc_id;
292 return res;
296 * v9fs_file_lock_dotl - lock a file (or directory)
297 * @filp: file to be locked
298 * @cmd: lock command
299 * @fl: file lock structure
303 static int v9fs_file_lock_dotl(struct file *filp, int cmd, struct file_lock *fl)
305 struct inode *inode = filp->f_path.dentry->d_inode;
306 int ret = -ENOLCK;
308 p9_debug(P9_DEBUG_VFS, "filp: %p cmd:%d lock: %p name: %s\n",
309 filp, cmd, fl, filp->f_path.dentry->d_name.name);
311 /* No mandatory locks */
312 if (__mandatory_lock(inode) && fl->fl_type != F_UNLCK)
313 goto out_err;
315 if ((IS_SETLK(cmd) || IS_SETLKW(cmd)) && fl->fl_type != F_UNLCK) {
316 filemap_write_and_wait(inode->i_mapping);
317 invalidate_mapping_pages(&inode->i_data, 0, -1);
320 if (IS_SETLK(cmd) || IS_SETLKW(cmd))
321 ret = v9fs_file_do_lock(filp, cmd, fl);
322 else if (IS_GETLK(cmd))
323 ret = v9fs_file_getlock(filp, fl);
324 else
325 ret = -EINVAL;
326 out_err:
327 return ret;
331 * v9fs_file_flock_dotl - lock a file
332 * @filp: file to be locked
333 * @cmd: lock command
334 * @fl: file lock structure
338 static int v9fs_file_flock_dotl(struct file *filp, int cmd,
339 struct file_lock *fl)
341 struct inode *inode = filp->f_path.dentry->d_inode;
342 int ret = -ENOLCK;
344 p9_debug(P9_DEBUG_VFS, "filp: %p cmd:%d lock: %p name: %s\n",
345 filp, cmd, fl, filp->f_path.dentry->d_name.name);
347 /* No mandatory locks */
348 if (__mandatory_lock(inode) && fl->fl_type != F_UNLCK)
349 goto out_err;
351 if (!(fl->fl_flags & FL_FLOCK))
352 goto out_err;
354 if ((IS_SETLK(cmd) || IS_SETLKW(cmd)) && fl->fl_type != F_UNLCK) {
355 filemap_write_and_wait(inode->i_mapping);
356 invalidate_mapping_pages(&inode->i_data, 0, -1);
358 /* Convert flock to posix lock */
359 fl->fl_owner = (fl_owner_t)filp;
360 fl->fl_start = 0;
361 fl->fl_end = OFFSET_MAX;
362 fl->fl_flags |= FL_POSIX;
363 fl->fl_flags ^= FL_FLOCK;
365 if (IS_SETLK(cmd) | IS_SETLKW(cmd))
366 ret = v9fs_file_do_lock(filp, cmd, fl);
367 else
368 ret = -EINVAL;
369 out_err:
370 return ret;
374 * v9fs_fid_readn - read from a fid
375 * @fid: fid to read
376 * @data: data buffer to read data into
377 * @udata: user data buffer to read data into
378 * @count: size of buffer
379 * @offset: offset at which to read data
382 ssize_t
383 v9fs_fid_readn(struct p9_fid *fid, char *data, char __user *udata, u32 count,
384 u64 offset)
386 int n, total, size;
388 p9_debug(P9_DEBUG_VFS, "fid %d offset %llu count %d\n",
389 fid->fid, (long long unsigned)offset, count);
390 n = 0;
391 total = 0;
392 size = fid->iounit ? fid->iounit : fid->clnt->msize - P9_IOHDRSZ;
393 do {
394 n = p9_client_read(fid, data, udata, offset, count);
395 if (n <= 0)
396 break;
398 if (data)
399 data += n;
400 if (udata)
401 udata += n;
403 offset += n;
404 count -= n;
405 total += n;
406 } while (count > 0 && n == size);
408 if (n < 0)
409 total = n;
411 return total;
415 * v9fs_file_readn - read from a file
416 * @filp: file pointer to read
417 * @data: data buffer to read data into
418 * @udata: user data buffer to read data into
419 * @count: size of buffer
420 * @offset: offset at which to read data
423 ssize_t
424 v9fs_file_readn(struct file *filp, char *data, char __user *udata, u32 count,
425 u64 offset)
427 return v9fs_fid_readn(filp->private_data, data, udata, count, offset);
431 * v9fs_file_read - read from a file
432 * @filp: file pointer to read
433 * @udata: user data buffer to read data into
434 * @count: size of buffer
435 * @offset: offset at which to read data
439 static ssize_t
440 v9fs_file_read(struct file *filp, char __user *udata, size_t count,
441 loff_t * offset)
443 int ret;
444 struct p9_fid *fid;
445 size_t size;
447 p9_debug(P9_DEBUG_VFS, "count %zu offset %lld\n", count, *offset);
448 fid = filp->private_data;
450 size = fid->iounit ? fid->iounit : fid->clnt->msize - P9_IOHDRSZ;
451 if (count > size)
452 ret = v9fs_file_readn(filp, NULL, udata, count, *offset);
453 else
454 ret = p9_client_read(fid, NULL, udata, *offset, count);
456 if (ret > 0)
457 *offset += ret;
459 return ret;
462 ssize_t
463 v9fs_file_write_internal(struct inode *inode, struct p9_fid *fid,
464 const char __user *data, size_t count,
465 loff_t *offset, int invalidate)
467 int n;
468 loff_t i_size;
469 size_t total = 0;
470 struct p9_client *clnt;
471 loff_t origin = *offset;
472 unsigned long pg_start, pg_end;
474 p9_debug(P9_DEBUG_VFS, "data %p count %d offset %x\n",
475 data, (int)count, (int)*offset);
477 clnt = fid->clnt;
478 do {
479 n = p9_client_write(fid, NULL, data+total, origin+total, count);
480 if (n <= 0)
481 break;
482 count -= n;
483 total += n;
484 } while (count > 0);
486 if (invalidate && (total > 0)) {
487 pg_start = origin >> PAGE_CACHE_SHIFT;
488 pg_end = (origin + total - 1) >> PAGE_CACHE_SHIFT;
489 if (inode->i_mapping && inode->i_mapping->nrpages)
490 invalidate_inode_pages2_range(inode->i_mapping,
491 pg_start, pg_end);
492 *offset += total;
493 i_size = i_size_read(inode);
494 if (*offset > i_size) {
495 inode_add_bytes(inode, *offset - i_size);
496 i_size_write(inode, *offset);
499 if (n < 0)
500 return n;
502 return total;
506 * v9fs_file_write - write to a file
507 * @filp: file pointer to write
508 * @data: data buffer to write data from
509 * @count: size of buffer
510 * @offset: offset at which to write data
513 static ssize_t
514 v9fs_file_write(struct file *filp, const char __user * data,
515 size_t count, loff_t *offset)
517 ssize_t retval = 0;
518 loff_t origin = *offset;
521 retval = generic_write_checks(filp, &origin, &count, 0);
522 if (retval)
523 goto out;
525 retval = -EINVAL;
526 if ((ssize_t) count < 0)
527 goto out;
528 retval = 0;
529 if (!count)
530 goto out;
532 retval = v9fs_file_write_internal(filp->f_path.dentry->d_inode,
533 filp->private_data,
534 data, count, &origin, 1);
535 /* update offset on successful write */
536 if (retval > 0)
537 *offset = origin;
538 out:
539 return retval;
543 static int v9fs_file_fsync(struct file *filp, loff_t start, loff_t end,
544 int datasync)
546 struct p9_fid *fid;
547 struct inode *inode = filp->f_mapping->host;
548 struct p9_wstat wstat;
549 int retval;
551 retval = filemap_write_and_wait_range(inode->i_mapping, start, end);
552 if (retval)
553 return retval;
555 mutex_lock(&inode->i_mutex);
556 p9_debug(P9_DEBUG_VFS, "filp %p datasync %x\n", filp, datasync);
558 fid = filp->private_data;
559 v9fs_blank_wstat(&wstat);
561 retval = p9_client_wstat(fid, &wstat);
562 mutex_unlock(&inode->i_mutex);
564 return retval;
567 int v9fs_file_fsync_dotl(struct file *filp, loff_t start, loff_t end,
568 int datasync)
570 struct p9_fid *fid;
571 struct inode *inode = filp->f_mapping->host;
572 int retval;
574 retval = filemap_write_and_wait_range(inode->i_mapping, start, end);
575 if (retval)
576 return retval;
578 mutex_lock(&inode->i_mutex);
579 p9_debug(P9_DEBUG_VFS, "filp %p datasync %x\n", filp, datasync);
581 fid = filp->private_data;
583 retval = p9_client_fsync(fid, datasync);
584 mutex_unlock(&inode->i_mutex);
586 return retval;
589 static int
590 v9fs_file_mmap(struct file *file, struct vm_area_struct *vma)
592 int retval;
594 retval = generic_file_mmap(file, vma);
595 if (!retval)
596 vma->vm_ops = &v9fs_file_vm_ops;
598 return retval;
601 static int
602 v9fs_vm_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf)
604 struct v9fs_inode *v9inode;
605 struct page *page = vmf->page;
606 struct file *filp = vma->vm_file;
607 struct inode *inode = filp->f_path.dentry->d_inode;
610 p9_debug(P9_DEBUG_VFS, "page %p fid %lx\n",
611 page, (unsigned long)filp->private_data);
613 v9inode = V9FS_I(inode);
614 /* make sure the cache has finished storing the page */
615 v9fs_fscache_wait_on_page_write(inode, page);
616 BUG_ON(!v9inode->writeback_fid);
617 lock_page(page);
618 if (page->mapping != inode->i_mapping)
619 goto out_unlock;
621 return VM_FAULT_LOCKED;
622 out_unlock:
623 unlock_page(page);
624 return VM_FAULT_NOPAGE;
627 static ssize_t
628 v9fs_direct_read(struct file *filp, char __user *udata, size_t count,
629 loff_t *offsetp)
631 loff_t size, offset;
632 struct inode *inode;
633 struct address_space *mapping;
635 offset = *offsetp;
636 mapping = filp->f_mapping;
637 inode = mapping->host;
638 if (!count)
639 return 0;
640 size = i_size_read(inode);
641 if (offset < size)
642 filemap_write_and_wait_range(mapping, offset,
643 offset + count - 1);
645 return v9fs_file_read(filp, udata, count, offsetp);
649 * v9fs_cached_file_read - read from a file
650 * @filp: file pointer to read
651 * @udata: user data buffer to read data into
652 * @count: size of buffer
653 * @offset: offset at which to read data
656 static ssize_t
657 v9fs_cached_file_read(struct file *filp, char __user *data, size_t count,
658 loff_t *offset)
660 if (filp->f_flags & O_DIRECT)
661 return v9fs_direct_read(filp, data, count, offset);
662 return do_sync_read(filp, data, count, offset);
665 static ssize_t
666 v9fs_direct_write(struct file *filp, const char __user * data,
667 size_t count, loff_t *offsetp)
669 loff_t offset;
670 ssize_t retval;
671 struct inode *inode;
672 struct address_space *mapping;
674 offset = *offsetp;
675 mapping = filp->f_mapping;
676 inode = mapping->host;
677 if (!count)
678 return 0;
680 mutex_lock(&inode->i_mutex);
681 retval = filemap_write_and_wait_range(mapping, offset,
682 offset + count - 1);
683 if (retval)
684 goto err_out;
686 * After a write we want buffered reads to be sure to go to disk to get
687 * the new data. We invalidate clean cached page from the region we're
688 * about to write. We do this *before* the write so that if we fail
689 * here we fall back to buffered write
691 if (mapping->nrpages) {
692 pgoff_t pg_start = offset >> PAGE_CACHE_SHIFT;
693 pgoff_t pg_end = (offset + count - 1) >> PAGE_CACHE_SHIFT;
695 retval = invalidate_inode_pages2_range(mapping,
696 pg_start, pg_end);
698 * If a page can not be invalidated, fall back
699 * to buffered write.
701 if (retval) {
702 if (retval == -EBUSY)
703 goto buff_write;
704 goto err_out;
707 retval = v9fs_file_write(filp, data, count, offsetp);
708 err_out:
709 mutex_unlock(&inode->i_mutex);
710 return retval;
712 buff_write:
713 mutex_unlock(&inode->i_mutex);
714 return do_sync_write(filp, data, count, offsetp);
718 * v9fs_cached_file_write - write to a file
719 * @filp: file pointer to write
720 * @data: data buffer to write data from
721 * @count: size of buffer
722 * @offset: offset at which to write data
725 static ssize_t
726 v9fs_cached_file_write(struct file *filp, const char __user * data,
727 size_t count, loff_t *offset)
730 if (filp->f_flags & O_DIRECT)
731 return v9fs_direct_write(filp, data, count, offset);
732 return do_sync_write(filp, data, count, offset);
735 static const struct vm_operations_struct v9fs_file_vm_ops = {
736 .fault = filemap_fault,
737 .page_mkwrite = v9fs_vm_page_mkwrite,
741 const struct file_operations v9fs_cached_file_operations = {
742 .llseek = generic_file_llseek,
743 .read = v9fs_cached_file_read,
744 .write = v9fs_cached_file_write,
745 .aio_read = generic_file_aio_read,
746 .aio_write = generic_file_aio_write,
747 .open = v9fs_file_open,
748 .release = v9fs_dir_release,
749 .lock = v9fs_file_lock,
750 .mmap = v9fs_file_mmap,
751 .fsync = v9fs_file_fsync,
754 const struct file_operations v9fs_cached_file_operations_dotl = {
755 .llseek = generic_file_llseek,
756 .read = v9fs_cached_file_read,
757 .write = v9fs_cached_file_write,
758 .aio_read = generic_file_aio_read,
759 .aio_write = generic_file_aio_write,
760 .open = v9fs_file_open,
761 .release = v9fs_dir_release,
762 .lock = v9fs_file_lock_dotl,
763 .flock = v9fs_file_flock_dotl,
764 .mmap = v9fs_file_mmap,
765 .fsync = v9fs_file_fsync_dotl,
768 const struct file_operations v9fs_file_operations = {
769 .llseek = generic_file_llseek,
770 .read = v9fs_file_read,
771 .write = v9fs_file_write,
772 .open = v9fs_file_open,
773 .release = v9fs_dir_release,
774 .lock = v9fs_file_lock,
775 .mmap = generic_file_readonly_mmap,
776 .fsync = v9fs_file_fsync,
779 const struct file_operations v9fs_file_operations_dotl = {
780 .llseek = generic_file_llseek,
781 .read = v9fs_file_read,
782 .write = v9fs_file_write,
783 .open = v9fs_file_open,
784 .release = v9fs_dir_release,
785 .lock = v9fs_file_lock_dotl,
786 .flock = v9fs_file_flock_dotl,
787 .mmap = generic_file_readonly_mmap,
788 .fsync = v9fs_file_fsync_dotl,