btree: fix tree corruption in btree_get_prev()
[linux-2.6/linux-acpi-2.6/ibm-acpi-2.6.git] / fs / 9p / vfs_file.c
blob9d6e1685d9226c388ac5cac0f0adf2fc3d601c32
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_DPRINTK(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_DPRINTK(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 schedule_timeout_interruptible(P9_LOCK_TIMEOUT);
210 /* map 9p status to VFS status */
211 switch (status) {
212 case P9_LOCK_SUCCESS:
213 res = 0;
214 break;
215 case P9_LOCK_BLOCKED:
216 res = -EAGAIN;
217 break;
218 case P9_LOCK_ERROR:
219 case P9_LOCK_GRACE:
220 res = -ENOLCK;
221 break;
222 default:
223 BUG();
227 * incase server returned error for lock request, revert
228 * it locally
230 if (res < 0 && fl->fl_type != F_UNLCK) {
231 fl_type = fl->fl_type;
232 fl->fl_type = F_UNLCK;
233 res = posix_lock_file_wait(filp, fl);
234 fl->fl_type = fl_type;
236 out:
237 return res;
240 static int v9fs_file_getlock(struct file *filp, struct file_lock *fl)
242 struct p9_getlock glock;
243 struct p9_fid *fid;
244 int res = 0;
246 fid = filp->private_data;
247 BUG_ON(fid == NULL);
249 posix_test_lock(filp, fl);
251 * if we have a conflicting lock locally, no need to validate
252 * with server
254 if (fl->fl_type != F_UNLCK)
255 return res;
257 /* convert posix lock to p9 tgetlock args */
258 memset(&glock, 0, sizeof(glock));
259 glock.type = P9_LOCK_TYPE_UNLCK;
260 glock.start = fl->fl_start;
261 if (fl->fl_end == OFFSET_MAX)
262 glock.length = 0;
263 else
264 glock.length = fl->fl_end - fl->fl_start + 1;
265 glock.proc_id = fl->fl_pid;
266 glock.client_id = utsname()->nodename;
268 res = p9_client_getlock_dotl(fid, &glock);
269 if (res < 0)
270 return res;
271 /* map 9p lock type to os lock type */
272 switch (glock.type) {
273 case P9_LOCK_TYPE_RDLCK:
274 fl->fl_type = F_RDLCK;
275 break;
276 case P9_LOCK_TYPE_WRLCK:
277 fl->fl_type = F_WRLCK;
278 break;
279 case P9_LOCK_TYPE_UNLCK:
280 fl->fl_type = F_UNLCK;
281 break;
283 if (glock.type != P9_LOCK_TYPE_UNLCK) {
284 fl->fl_start = glock.start;
285 if (glock.length == 0)
286 fl->fl_end = OFFSET_MAX;
287 else
288 fl->fl_end = glock.start + glock.length - 1;
289 fl->fl_pid = glock.proc_id;
291 return res;
295 * v9fs_file_lock_dotl - lock a file (or directory)
296 * @filp: file to be locked
297 * @cmd: lock command
298 * @fl: file lock structure
302 static int v9fs_file_lock_dotl(struct file *filp, int cmd, struct file_lock *fl)
304 struct inode *inode = filp->f_path.dentry->d_inode;
305 int ret = -ENOLCK;
307 P9_DPRINTK(P9_DEBUG_VFS, "filp: %p cmd:%d lock: %p name: %s\n", filp,
308 cmd, fl, filp->f_path.dentry->d_name.name);
310 /* No mandatory locks */
311 if (__mandatory_lock(inode) && fl->fl_type != F_UNLCK)
312 goto out_err;
314 if ((IS_SETLK(cmd) || IS_SETLKW(cmd)) && fl->fl_type != F_UNLCK) {
315 filemap_write_and_wait(inode->i_mapping);
316 invalidate_mapping_pages(&inode->i_data, 0, -1);
319 if (IS_SETLK(cmd) || IS_SETLKW(cmd))
320 ret = v9fs_file_do_lock(filp, cmd, fl);
321 else if (IS_GETLK(cmd))
322 ret = v9fs_file_getlock(filp, fl);
323 else
324 ret = -EINVAL;
325 out_err:
326 return ret;
330 * v9fs_file_flock_dotl - lock a file
331 * @filp: file to be locked
332 * @cmd: lock command
333 * @fl: file lock structure
337 static int v9fs_file_flock_dotl(struct file *filp, int cmd,
338 struct file_lock *fl)
340 struct inode *inode = filp->f_path.dentry->d_inode;
341 int ret = -ENOLCK;
343 P9_DPRINTK(P9_DEBUG_VFS, "filp: %p cmd:%d lock: %p name: %s\n", filp,
344 cmd, fl, filp->f_path.dentry->d_name.name);
346 /* No mandatory locks */
347 if (__mandatory_lock(inode) && fl->fl_type != F_UNLCK)
348 goto out_err;
350 if (!(fl->fl_flags & FL_FLOCK))
351 goto out_err;
353 if ((IS_SETLK(cmd) || IS_SETLKW(cmd)) && fl->fl_type != F_UNLCK) {
354 filemap_write_and_wait(inode->i_mapping);
355 invalidate_mapping_pages(&inode->i_data, 0, -1);
357 /* Convert flock to posix lock */
358 fl->fl_owner = (fl_owner_t)filp;
359 fl->fl_start = 0;
360 fl->fl_end = OFFSET_MAX;
361 fl->fl_flags |= FL_POSIX;
362 fl->fl_flags ^= FL_FLOCK;
364 if (IS_SETLK(cmd) | IS_SETLKW(cmd))
365 ret = v9fs_file_do_lock(filp, cmd, fl);
366 else
367 ret = -EINVAL;
368 out_err:
369 return ret;
373 * v9fs_fid_readn - read from a fid
374 * @fid: fid to read
375 * @data: data buffer to read data into
376 * @udata: user data buffer to read data into
377 * @count: size of buffer
378 * @offset: offset at which to read data
381 ssize_t
382 v9fs_fid_readn(struct p9_fid *fid, char *data, char __user *udata, u32 count,
383 u64 offset)
385 int n, total, size;
387 P9_DPRINTK(P9_DEBUG_VFS, "fid %d offset %llu count %d\n", fid->fid,
388 (long long unsigned) offset, count);
389 n = 0;
390 total = 0;
391 size = fid->iounit ? fid->iounit : fid->clnt->msize - P9_IOHDRSZ;
392 do {
393 n = p9_client_read(fid, data, udata, offset, count);
394 if (n <= 0)
395 break;
397 if (data)
398 data += n;
399 if (udata)
400 udata += n;
402 offset += n;
403 count -= n;
404 total += n;
405 } while (count > 0 && n == size);
407 if (n < 0)
408 total = n;
410 return total;
414 * v9fs_file_readn - read from a file
415 * @filp: file pointer to read
416 * @data: data buffer to read data into
417 * @udata: user data buffer to read data into
418 * @count: size of buffer
419 * @offset: offset at which to read data
422 ssize_t
423 v9fs_file_readn(struct file *filp, char *data, char __user *udata, u32 count,
424 u64 offset)
426 return v9fs_fid_readn(filp->private_data, data, udata, count, offset);
430 * v9fs_file_read - read from a file
431 * @filp: file pointer to read
432 * @udata: user data buffer to read data into
433 * @count: size of buffer
434 * @offset: offset at which to read data
438 static ssize_t
439 v9fs_file_read(struct file *filp, char __user *udata, size_t count,
440 loff_t * offset)
442 int ret;
443 struct p9_fid *fid;
444 size_t size;
446 P9_DPRINTK(P9_DEBUG_VFS, "count %zu offset %lld\n", count, *offset);
447 fid = filp->private_data;
449 size = fid->iounit ? fid->iounit : fid->clnt->msize - P9_IOHDRSZ;
450 if (count > size)
451 ret = v9fs_file_readn(filp, NULL, udata, count, *offset);
452 else
453 ret = p9_client_read(fid, NULL, udata, *offset, count);
455 if (ret > 0)
456 *offset += ret;
458 return ret;
461 ssize_t
462 v9fs_file_write_internal(struct inode *inode, struct p9_fid *fid,
463 const char __user *data, size_t count,
464 loff_t *offset, int invalidate)
466 int n;
467 loff_t i_size;
468 size_t total = 0;
469 struct p9_client *clnt;
470 loff_t origin = *offset;
471 unsigned long pg_start, pg_end;
473 P9_DPRINTK(P9_DEBUG_VFS, "data %p count %d offset %x\n", data,
474 (int)count, (int)*offset);
476 clnt = fid->clnt;
477 do {
478 n = p9_client_write(fid, NULL, data+total, origin+total, count);
479 if (n <= 0)
480 break;
481 count -= n;
482 total += n;
483 } while (count > 0);
485 if (invalidate && (total > 0)) {
486 pg_start = origin >> PAGE_CACHE_SHIFT;
487 pg_end = (origin + total - 1) >> PAGE_CACHE_SHIFT;
488 if (inode->i_mapping && inode->i_mapping->nrpages)
489 invalidate_inode_pages2_range(inode->i_mapping,
490 pg_start, pg_end);
491 *offset += total;
492 i_size = i_size_read(inode);
493 if (*offset > i_size) {
494 inode_add_bytes(inode, *offset - i_size);
495 i_size_write(inode, *offset);
498 if (n < 0)
499 return n;
501 return total;
505 * v9fs_file_write - write to a file
506 * @filp: file pointer to write
507 * @data: data buffer to write data from
508 * @count: size of buffer
509 * @offset: offset at which to write data
512 static ssize_t
513 v9fs_file_write(struct file *filp, const char __user * data,
514 size_t count, loff_t *offset)
516 ssize_t retval = 0;
517 loff_t origin = *offset;
520 retval = generic_write_checks(filp, &origin, &count, 0);
521 if (retval)
522 goto out;
524 retval = -EINVAL;
525 if ((ssize_t) count < 0)
526 goto out;
527 retval = 0;
528 if (!count)
529 goto out;
531 retval = v9fs_file_write_internal(filp->f_path.dentry->d_inode,
532 filp->private_data,
533 data, count, &origin, 1);
534 /* update offset on successful write */
535 if (retval > 0)
536 *offset = origin;
537 out:
538 return retval;
542 static int v9fs_file_fsync(struct file *filp, int datasync)
544 struct p9_fid *fid;
545 struct p9_wstat wstat;
546 int retval;
548 P9_DPRINTK(P9_DEBUG_VFS, "filp %p datasync %x\n", filp, datasync);
550 fid = filp->private_data;
551 v9fs_blank_wstat(&wstat);
553 retval = p9_client_wstat(fid, &wstat);
554 return retval;
557 int v9fs_file_fsync_dotl(struct file *filp, int datasync)
559 struct p9_fid *fid;
560 int retval;
562 P9_DPRINTK(P9_DEBUG_VFS, "v9fs_file_fsync_dotl: filp %p datasync %x\n",
563 filp, datasync);
565 fid = filp->private_data;
567 retval = p9_client_fsync(fid, datasync);
568 return retval;
571 static int
572 v9fs_file_mmap(struct file *file, struct vm_area_struct *vma)
574 int retval;
576 retval = generic_file_mmap(file, vma);
577 if (!retval)
578 vma->vm_ops = &v9fs_file_vm_ops;
580 return retval;
583 static int
584 v9fs_vm_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf)
586 struct v9fs_inode *v9inode;
587 struct page *page = vmf->page;
588 struct file *filp = vma->vm_file;
589 struct inode *inode = filp->f_path.dentry->d_inode;
592 P9_DPRINTK(P9_DEBUG_VFS, "page %p fid %lx\n",
593 page, (unsigned long)filp->private_data);
595 v9inode = V9FS_I(inode);
596 /* make sure the cache has finished storing the page */
597 v9fs_fscache_wait_on_page_write(inode, page);
598 BUG_ON(!v9inode->writeback_fid);
599 lock_page(page);
600 if (page->mapping != inode->i_mapping)
601 goto out_unlock;
603 return VM_FAULT_LOCKED;
604 out_unlock:
605 unlock_page(page);
606 return VM_FAULT_NOPAGE;
609 static ssize_t
610 v9fs_direct_read(struct file *filp, char __user *udata, size_t count,
611 loff_t *offsetp)
613 loff_t size, offset;
614 struct inode *inode;
615 struct address_space *mapping;
617 offset = *offsetp;
618 mapping = filp->f_mapping;
619 inode = mapping->host;
620 if (!count)
621 return 0;
622 size = i_size_read(inode);
623 if (offset < size)
624 filemap_write_and_wait_range(mapping, offset,
625 offset + count - 1);
627 return v9fs_file_read(filp, udata, count, offsetp);
631 * v9fs_cached_file_read - read from a file
632 * @filp: file pointer to read
633 * @udata: user data buffer to read data into
634 * @count: size of buffer
635 * @offset: offset at which to read data
638 static ssize_t
639 v9fs_cached_file_read(struct file *filp, char __user *data, size_t count,
640 loff_t *offset)
642 if (filp->f_flags & O_DIRECT)
643 return v9fs_direct_read(filp, data, count, offset);
644 return do_sync_read(filp, data, count, offset);
647 static ssize_t
648 v9fs_direct_write(struct file *filp, const char __user * data,
649 size_t count, loff_t *offsetp)
651 loff_t offset;
652 ssize_t retval;
653 struct inode *inode;
654 struct address_space *mapping;
656 offset = *offsetp;
657 mapping = filp->f_mapping;
658 inode = mapping->host;
659 if (!count)
660 return 0;
662 mutex_lock(&inode->i_mutex);
663 retval = filemap_write_and_wait_range(mapping, offset,
664 offset + count - 1);
665 if (retval)
666 goto err_out;
668 * After a write we want buffered reads to be sure to go to disk to get
669 * the new data. We invalidate clean cached page from the region we're
670 * about to write. We do this *before* the write so that if we fail
671 * here we fall back to buffered write
673 if (mapping->nrpages) {
674 pgoff_t pg_start = offset >> PAGE_CACHE_SHIFT;
675 pgoff_t pg_end = (offset + count - 1) >> PAGE_CACHE_SHIFT;
677 retval = invalidate_inode_pages2_range(mapping,
678 pg_start, pg_end);
680 * If a page can not be invalidated, fall back
681 * to buffered write.
683 if (retval) {
684 if (retval == -EBUSY)
685 goto buff_write;
686 goto err_out;
689 retval = v9fs_file_write(filp, data, count, offsetp);
690 err_out:
691 mutex_unlock(&inode->i_mutex);
692 return retval;
694 buff_write:
695 mutex_unlock(&inode->i_mutex);
696 return do_sync_write(filp, data, count, offsetp);
700 * v9fs_cached_file_write - write to a file
701 * @filp: file pointer to write
702 * @data: data buffer to write data from
703 * @count: size of buffer
704 * @offset: offset at which to write data
707 static ssize_t
708 v9fs_cached_file_write(struct file *filp, const char __user * data,
709 size_t count, loff_t *offset)
712 if (filp->f_flags & O_DIRECT)
713 return v9fs_direct_write(filp, data, count, offset);
714 return do_sync_write(filp, data, count, offset);
717 static const struct vm_operations_struct v9fs_file_vm_ops = {
718 .fault = filemap_fault,
719 .page_mkwrite = v9fs_vm_page_mkwrite,
723 const struct file_operations v9fs_cached_file_operations = {
724 .llseek = generic_file_llseek,
725 .read = v9fs_cached_file_read,
726 .write = v9fs_cached_file_write,
727 .aio_read = generic_file_aio_read,
728 .aio_write = generic_file_aio_write,
729 .open = v9fs_file_open,
730 .release = v9fs_dir_release,
731 .lock = v9fs_file_lock,
732 .mmap = v9fs_file_mmap,
733 .fsync = v9fs_file_fsync,
736 const struct file_operations v9fs_cached_file_operations_dotl = {
737 .llseek = generic_file_llseek,
738 .read = v9fs_cached_file_read,
739 .write = v9fs_cached_file_write,
740 .aio_read = generic_file_aio_read,
741 .aio_write = generic_file_aio_write,
742 .open = v9fs_file_open,
743 .release = v9fs_dir_release,
744 .lock = v9fs_file_lock_dotl,
745 .flock = v9fs_file_flock_dotl,
746 .mmap = v9fs_file_mmap,
747 .fsync = v9fs_file_fsync_dotl,
750 const struct file_operations v9fs_file_operations = {
751 .llseek = generic_file_llseek,
752 .read = v9fs_file_read,
753 .write = v9fs_file_write,
754 .open = v9fs_file_open,
755 .release = v9fs_dir_release,
756 .lock = v9fs_file_lock,
757 .mmap = generic_file_readonly_mmap,
758 .fsync = v9fs_file_fsync,
761 const struct file_operations v9fs_file_operations_dotl = {
762 .llseek = generic_file_llseek,
763 .read = v9fs_file_read,
764 .write = v9fs_file_write,
765 .open = v9fs_file_open,
766 .release = v9fs_dir_release,
767 .lock = v9fs_file_lock_dotl,
768 .flock = v9fs_file_flock_dotl,
769 .mmap = generic_file_readonly_mmap,
770 .fsync = v9fs_file_fsync_dotl,