Loading...
1/*
2 * linux/fs/9p/vfs_file.c
3 *
4 * This file contians vfs file ops for 9P2000.
5 *
6 * Copyright (C) 2004 by Eric Van Hensbergen <ericvh@gmail.com>
7 * Copyright (C) 2002 by Ron Minnich <rminnich@lanl.gov>
8 *
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.
12 *
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.
17 *
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
23 *
24 */
25
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 <linux/uio.h>
40#include <linux/slab.h>
41#include <net/9p/9p.h>
42#include <net/9p/client.h>
43
44#include "v9fs.h"
45#include "v9fs_vfs.h"
46#include "fid.h"
47#include "cache.h"
48
49static const struct vm_operations_struct v9fs_file_vm_ops;
50static const struct vm_operations_struct v9fs_mmap_file_vm_ops;
51
52/**
53 * v9fs_file_open - open a file (or directory)
54 * @inode: inode to be opened
55 * @file: file being opened
56 *
57 */
58
59int v9fs_file_open(struct inode *inode, struct file *file)
60{
61 int err;
62 struct v9fs_inode *v9inode;
63 struct v9fs_session_info *v9ses;
64 struct p9_fid *fid;
65 int omode;
66
67 p9_debug(P9_DEBUG_VFS, "inode: %p file: %p\n", inode, file);
68 v9inode = V9FS_I(inode);
69 v9ses = v9fs_inode2v9ses(inode);
70 if (v9fs_proto_dotl(v9ses))
71 omode = v9fs_open_to_dotl_flags(file->f_flags);
72 else
73 omode = v9fs_uflags2omode(file->f_flags,
74 v9fs_proto_dotu(v9ses));
75 fid = file->private_data;
76 if (!fid) {
77 fid = v9fs_fid_clone(file->f_path.dentry);
78 if (IS_ERR(fid))
79 return PTR_ERR(fid);
80
81 err = p9_client_open(fid, omode);
82 if (err < 0) {
83 p9_client_clunk(fid);
84 return err;
85 }
86 if ((file->f_flags & O_APPEND) &&
87 (!v9fs_proto_dotu(v9ses) && !v9fs_proto_dotl(v9ses)))
88 generic_file_llseek(file, 0, SEEK_END);
89 }
90
91 file->private_data = fid;
92 mutex_lock(&v9inode->v_mutex);
93 if ((v9ses->cache == CACHE_LOOSE || v9ses->cache == CACHE_FSCACHE) &&
94 !v9inode->writeback_fid &&
95 ((file->f_flags & O_ACCMODE) != O_RDONLY)) {
96 /*
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.
102 */
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;
108 }
109 v9inode->writeback_fid = (void *) fid;
110 }
111 mutex_unlock(&v9inode->v_mutex);
112 if (v9ses->cache == CACHE_LOOSE || v9ses->cache == CACHE_FSCACHE)
113 v9fs_cache_inode_set_cookie(inode, file);
114 return 0;
115out_error:
116 p9_client_clunk(file->private_data);
117 file->private_data = NULL;
118 return err;
119}
120
121/**
122 * v9fs_file_lock - lock a file (or directory)
123 * @filp: file to be locked
124 * @cmd: lock command
125 * @fl: file lock structure
126 *
127 * Bugs: this looks like a local only lock, we should extend into 9P
128 * by using open exclusive
129 */
130
131static int v9fs_file_lock(struct file *filp, int cmd, struct file_lock *fl)
132{
133 int res = 0;
134 struct inode *inode = file_inode(filp);
135
136 p9_debug(P9_DEBUG_VFS, "filp: %p lock: %p\n", filp, fl);
137
138 /* No mandatory locks */
139 if (__mandatory_lock(inode) && fl->fl_type != F_UNLCK)
140 return -ENOLCK;
141
142 if ((IS_SETLK(cmd) || IS_SETLKW(cmd)) && fl->fl_type != F_UNLCK) {
143 filemap_write_and_wait(inode->i_mapping);
144 invalidate_mapping_pages(&inode->i_data, 0, -1);
145 }
146
147 return res;
148}
149
150static int v9fs_file_do_lock(struct file *filp, int cmd, struct file_lock *fl)
151{
152 struct p9_flock flock;
153 struct p9_fid *fid;
154 uint8_t status = P9_LOCK_ERROR;
155 int res = 0;
156 unsigned char fl_type;
157
158 fid = filp->private_data;
159 BUG_ON(fid == NULL);
160
161 if ((fl->fl_flags & FL_POSIX) != FL_POSIX)
162 BUG();
163
164 res = locks_lock_file_wait(filp, fl);
165 if (res < 0)
166 goto out;
167
168 /* convert posix lock to p9 tlock args */
169 memset(&flock, 0, sizeof(flock));
170 /* map the lock type */
171 switch (fl->fl_type) {
172 case F_RDLCK:
173 flock.type = P9_LOCK_TYPE_RDLCK;
174 break;
175 case F_WRLCK:
176 flock.type = P9_LOCK_TYPE_WRLCK;
177 break;
178 case F_UNLCK:
179 flock.type = P9_LOCK_TYPE_UNLCK;
180 break;
181 }
182 flock.start = fl->fl_start;
183 if (fl->fl_end == OFFSET_MAX)
184 flock.length = 0;
185 else
186 flock.length = fl->fl_end - fl->fl_start + 1;
187 flock.proc_id = fl->fl_pid;
188 flock.client_id = fid->clnt->name;
189 if (IS_SETLKW(cmd))
190 flock.flags = P9_LOCK_FLAGS_BLOCK;
191
192 /*
193 * if its a blocked request and we get P9_LOCK_BLOCKED as the status
194 * for lock request, keep on trying
195 */
196 for (;;) {
197 res = p9_client_lock_dotl(fid, &flock, &status);
198 if (res < 0)
199 goto out_unlock;
200
201 if (status != P9_LOCK_BLOCKED)
202 break;
203 if (status == P9_LOCK_BLOCKED && !IS_SETLKW(cmd))
204 break;
205 if (schedule_timeout_interruptible(P9_LOCK_TIMEOUT) != 0)
206 break;
207 }
208
209 /* map 9p status to VFS status */
210 switch (status) {
211 case P9_LOCK_SUCCESS:
212 res = 0;
213 break;
214 case P9_LOCK_BLOCKED:
215 res = -EAGAIN;
216 break;
217 default:
218 WARN_ONCE(1, "unknown lock status code: %d\n", status);
219 /* fallthough */
220 case P9_LOCK_ERROR:
221 case P9_LOCK_GRACE:
222 res = -ENOLCK;
223 break;
224 }
225
226out_unlock:
227 /*
228 * incase server returned error for lock request, revert
229 * it locally
230 */
231 if (res < 0 && fl->fl_type != F_UNLCK) {
232 fl_type = fl->fl_type;
233 fl->fl_type = F_UNLCK;
234 /* Even if this fails we want to return the remote error */
235 locks_lock_file_wait(filp, fl);
236 fl->fl_type = fl_type;
237 }
238out:
239 return res;
240}
241
242static int v9fs_file_getlock(struct file *filp, struct file_lock *fl)
243{
244 struct p9_getlock glock;
245 struct p9_fid *fid;
246 int res = 0;
247
248 fid = filp->private_data;
249 BUG_ON(fid == NULL);
250
251 posix_test_lock(filp, fl);
252 /*
253 * if we have a conflicting lock locally, no need to validate
254 * with server
255 */
256 if (fl->fl_type != F_UNLCK)
257 return res;
258
259 /* convert posix lock to p9 tgetlock args */
260 memset(&glock, 0, sizeof(glock));
261 glock.type = P9_LOCK_TYPE_UNLCK;
262 glock.start = fl->fl_start;
263 if (fl->fl_end == OFFSET_MAX)
264 glock.length = 0;
265 else
266 glock.length = fl->fl_end - fl->fl_start + 1;
267 glock.proc_id = fl->fl_pid;
268 glock.client_id = fid->clnt->name;
269
270 res = p9_client_getlock_dotl(fid, &glock);
271 if (res < 0)
272 return res;
273 /* map 9p lock type to os lock type */
274 switch (glock.type) {
275 case P9_LOCK_TYPE_RDLCK:
276 fl->fl_type = F_RDLCK;
277 break;
278 case P9_LOCK_TYPE_WRLCK:
279 fl->fl_type = F_WRLCK;
280 break;
281 case P9_LOCK_TYPE_UNLCK:
282 fl->fl_type = F_UNLCK;
283 break;
284 }
285 if (glock.type != P9_LOCK_TYPE_UNLCK) {
286 fl->fl_start = glock.start;
287 if (glock.length == 0)
288 fl->fl_end = OFFSET_MAX;
289 else
290 fl->fl_end = glock.start + glock.length - 1;
291 fl->fl_pid = glock.proc_id;
292 }
293 kfree(glock.client_id);
294 return res;
295}
296
297/**
298 * v9fs_file_lock_dotl - lock a file (or directory)
299 * @filp: file to be locked
300 * @cmd: lock command
301 * @fl: file lock structure
302 *
303 */
304
305static int v9fs_file_lock_dotl(struct file *filp, int cmd, struct file_lock *fl)
306{
307 struct inode *inode = file_inode(filp);
308 int ret = -ENOLCK;
309
310 p9_debug(P9_DEBUG_VFS, "filp: %p cmd:%d lock: %p name: %pD\n",
311 filp, cmd, fl, filp);
312
313 /* No mandatory locks */
314 if (__mandatory_lock(inode) && fl->fl_type != F_UNLCK)
315 goto out_err;
316
317 if ((IS_SETLK(cmd) || IS_SETLKW(cmd)) && fl->fl_type != F_UNLCK) {
318 filemap_write_and_wait(inode->i_mapping);
319 invalidate_mapping_pages(&inode->i_data, 0, -1);
320 }
321
322 if (IS_SETLK(cmd) || IS_SETLKW(cmd))
323 ret = v9fs_file_do_lock(filp, cmd, fl);
324 else if (IS_GETLK(cmd))
325 ret = v9fs_file_getlock(filp, fl);
326 else
327 ret = -EINVAL;
328out_err:
329 return ret;
330}
331
332/**
333 * v9fs_file_flock_dotl - lock a file
334 * @filp: file to be locked
335 * @cmd: lock command
336 * @fl: file lock structure
337 *
338 */
339
340static int v9fs_file_flock_dotl(struct file *filp, int cmd,
341 struct file_lock *fl)
342{
343 struct inode *inode = file_inode(filp);
344 int ret = -ENOLCK;
345
346 p9_debug(P9_DEBUG_VFS, "filp: %p cmd:%d lock: %p name: %pD\n",
347 filp, cmd, fl, filp);
348
349 /* No mandatory locks */
350 if (__mandatory_lock(inode) && fl->fl_type != F_UNLCK)
351 goto out_err;
352
353 if (!(fl->fl_flags & FL_FLOCK))
354 goto out_err;
355
356 if ((IS_SETLK(cmd) || IS_SETLKW(cmd)) && fl->fl_type != F_UNLCK) {
357 filemap_write_and_wait(inode->i_mapping);
358 invalidate_mapping_pages(&inode->i_data, 0, -1);
359 }
360 /* Convert flock to posix lock */
361 fl->fl_flags |= FL_POSIX;
362 fl->fl_flags ^= FL_FLOCK;
363
364 if (IS_SETLK(cmd) | IS_SETLKW(cmd))
365 ret = v9fs_file_do_lock(filp, cmd, fl);
366 else
367 ret = -EINVAL;
368out_err:
369 return ret;
370}
371
372/**
373 * v9fs_file_read - read from a file
374 * @filp: file pointer to read
375 * @udata: user data buffer to read data into
376 * @count: size of buffer
377 * @offset: offset at which to read data
378 *
379 */
380
381static ssize_t
382v9fs_file_read_iter(struct kiocb *iocb, struct iov_iter *to)
383{
384 struct p9_fid *fid = iocb->ki_filp->private_data;
385 int ret, err = 0;
386
387 p9_debug(P9_DEBUG_VFS, "count %zu offset %lld\n",
388 iov_iter_count(to), iocb->ki_pos);
389
390 ret = p9_client_read(fid, iocb->ki_pos, to, &err);
391 if (!ret)
392 return err;
393
394 iocb->ki_pos += ret;
395 return ret;
396}
397
398/**
399 * v9fs_file_write - write to a file
400 * @filp: file pointer to write
401 * @data: data buffer to write data from
402 * @count: size of buffer
403 * @offset: offset at which to write data
404 *
405 */
406static ssize_t
407v9fs_file_write_iter(struct kiocb *iocb, struct iov_iter *from)
408{
409 struct file *file = iocb->ki_filp;
410 ssize_t retval;
411 loff_t origin;
412 int err = 0;
413
414 retval = generic_write_checks(iocb, from);
415 if (retval <= 0)
416 return retval;
417
418 origin = iocb->ki_pos;
419 retval = p9_client_write(file->private_data, iocb->ki_pos, from, &err);
420 if (retval > 0) {
421 struct inode *inode = file_inode(file);
422 loff_t i_size;
423 unsigned long pg_start, pg_end;
424 pg_start = origin >> PAGE_SHIFT;
425 pg_end = (origin + retval - 1) >> PAGE_SHIFT;
426 if (inode->i_mapping && inode->i_mapping->nrpages)
427 invalidate_inode_pages2_range(inode->i_mapping,
428 pg_start, pg_end);
429 iocb->ki_pos += retval;
430 i_size = i_size_read(inode);
431 if (iocb->ki_pos > i_size) {
432 inode_add_bytes(inode, iocb->ki_pos - i_size);
433 i_size_write(inode, iocb->ki_pos);
434 }
435 return retval;
436 }
437 return err;
438}
439
440static int v9fs_file_fsync(struct file *filp, loff_t start, loff_t end,
441 int datasync)
442{
443 struct p9_fid *fid;
444 struct inode *inode = filp->f_mapping->host;
445 struct p9_wstat wstat;
446 int retval;
447
448 retval = filemap_write_and_wait_range(inode->i_mapping, start, end);
449 if (retval)
450 return retval;
451
452 inode_lock(inode);
453 p9_debug(P9_DEBUG_VFS, "filp %p datasync %x\n", filp, datasync);
454
455 fid = filp->private_data;
456 v9fs_blank_wstat(&wstat);
457
458 retval = p9_client_wstat(fid, &wstat);
459 inode_unlock(inode);
460
461 return retval;
462}
463
464int v9fs_file_fsync_dotl(struct file *filp, loff_t start, loff_t end,
465 int datasync)
466{
467 struct p9_fid *fid;
468 struct inode *inode = filp->f_mapping->host;
469 int retval;
470
471 retval = filemap_write_and_wait_range(inode->i_mapping, start, end);
472 if (retval)
473 return retval;
474
475 inode_lock(inode);
476 p9_debug(P9_DEBUG_VFS, "filp %p datasync %x\n", filp, datasync);
477
478 fid = filp->private_data;
479
480 retval = p9_client_fsync(fid, datasync);
481 inode_unlock(inode);
482
483 return retval;
484}
485
486static int
487v9fs_file_mmap(struct file *filp, struct vm_area_struct *vma)
488{
489 int retval;
490
491
492 retval = generic_file_mmap(filp, vma);
493 if (!retval)
494 vma->vm_ops = &v9fs_file_vm_ops;
495
496 return retval;
497}
498
499static int
500v9fs_mmap_file_mmap(struct file *filp, struct vm_area_struct *vma)
501{
502 int retval;
503 struct inode *inode;
504 struct v9fs_inode *v9inode;
505 struct p9_fid *fid;
506
507 inode = file_inode(filp);
508 v9inode = V9FS_I(inode);
509 mutex_lock(&v9inode->v_mutex);
510 if (!v9inode->writeback_fid &&
511 (vma->vm_flags & VM_WRITE)) {
512 /*
513 * clone a fid and add it to writeback_fid
514 * we do it during mmap instead of
515 * page dirty time via write_begin/page_mkwrite
516 * because we want write after unlink usecase
517 * to work.
518 */
519 fid = v9fs_writeback_fid(filp->f_path.dentry);
520 if (IS_ERR(fid)) {
521 retval = PTR_ERR(fid);
522 mutex_unlock(&v9inode->v_mutex);
523 return retval;
524 }
525 v9inode->writeback_fid = (void *) fid;
526 }
527 mutex_unlock(&v9inode->v_mutex);
528
529 retval = generic_file_mmap(filp, vma);
530 if (!retval)
531 vma->vm_ops = &v9fs_mmap_file_vm_ops;
532
533 return retval;
534}
535
536static int
537v9fs_vm_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf)
538{
539 struct v9fs_inode *v9inode;
540 struct page *page = vmf->page;
541 struct file *filp = vma->vm_file;
542 struct inode *inode = file_inode(filp);
543
544
545 p9_debug(P9_DEBUG_VFS, "page %p fid %lx\n",
546 page, (unsigned long)filp->private_data);
547
548 /* Update file times before taking page lock */
549 file_update_time(filp);
550
551 v9inode = V9FS_I(inode);
552 /* make sure the cache has finished storing the page */
553 v9fs_fscache_wait_on_page_write(inode, page);
554 BUG_ON(!v9inode->writeback_fid);
555 lock_page(page);
556 if (page->mapping != inode->i_mapping)
557 goto out_unlock;
558 wait_for_stable_page(page);
559
560 return VM_FAULT_LOCKED;
561out_unlock:
562 unlock_page(page);
563 return VM_FAULT_NOPAGE;
564}
565
566/**
567 * v9fs_mmap_file_read - read from a file
568 * @filp: file pointer to read
569 * @data: user data buffer to read data into
570 * @count: size of buffer
571 * @offset: offset at which to read data
572 *
573 */
574static ssize_t
575v9fs_mmap_file_read_iter(struct kiocb *iocb, struct iov_iter *to)
576{
577 /* TODO: Check if there are dirty pages */
578 return v9fs_file_read_iter(iocb, to);
579}
580
581/**
582 * v9fs_mmap_file_write - write to a file
583 * @filp: file pointer to write
584 * @data: data buffer to write data from
585 * @count: size of buffer
586 * @offset: offset at which to write data
587 *
588 */
589static ssize_t
590v9fs_mmap_file_write_iter(struct kiocb *iocb, struct iov_iter *from)
591{
592 /*
593 * TODO: invalidate mmaps on filp's inode between
594 * offset and offset+count
595 */
596 return v9fs_file_write_iter(iocb, from);
597}
598
599static void v9fs_mmap_vm_close(struct vm_area_struct *vma)
600{
601 struct inode *inode;
602
603 struct writeback_control wbc = {
604 .nr_to_write = LONG_MAX,
605 .sync_mode = WB_SYNC_ALL,
606 .range_start = vma->vm_pgoff * PAGE_SIZE,
607 /* absolute end, byte at end included */
608 .range_end = vma->vm_pgoff * PAGE_SIZE +
609 (vma->vm_end - vma->vm_start - 1),
610 };
611
612
613 p9_debug(P9_DEBUG_VFS, "9p VMA close, %p, flushing", vma);
614
615 inode = file_inode(vma->vm_file);
616
617 if (!mapping_cap_writeback_dirty(inode->i_mapping))
618 wbc.nr_to_write = 0;
619
620 might_sleep();
621 sync_inode(inode, &wbc);
622}
623
624
625static const struct vm_operations_struct v9fs_file_vm_ops = {
626 .fault = filemap_fault,
627 .map_pages = filemap_map_pages,
628 .page_mkwrite = v9fs_vm_page_mkwrite,
629};
630
631static const struct vm_operations_struct v9fs_mmap_file_vm_ops = {
632 .close = v9fs_mmap_vm_close,
633 .fault = filemap_fault,
634 .map_pages = filemap_map_pages,
635 .page_mkwrite = v9fs_vm_page_mkwrite,
636};
637
638
639const struct file_operations v9fs_cached_file_operations = {
640 .llseek = generic_file_llseek,
641 .read_iter = generic_file_read_iter,
642 .write_iter = generic_file_write_iter,
643 .open = v9fs_file_open,
644 .release = v9fs_dir_release,
645 .lock = v9fs_file_lock,
646 .mmap = v9fs_file_mmap,
647 .fsync = v9fs_file_fsync,
648};
649
650const struct file_operations v9fs_cached_file_operations_dotl = {
651 .llseek = generic_file_llseek,
652 .read_iter = generic_file_read_iter,
653 .write_iter = generic_file_write_iter,
654 .open = v9fs_file_open,
655 .release = v9fs_dir_release,
656 .lock = v9fs_file_lock_dotl,
657 .flock = v9fs_file_flock_dotl,
658 .mmap = v9fs_file_mmap,
659 .fsync = v9fs_file_fsync_dotl,
660};
661
662const struct file_operations v9fs_file_operations = {
663 .llseek = generic_file_llseek,
664 .read_iter = v9fs_file_read_iter,
665 .write_iter = v9fs_file_write_iter,
666 .open = v9fs_file_open,
667 .release = v9fs_dir_release,
668 .lock = v9fs_file_lock,
669 .mmap = generic_file_readonly_mmap,
670 .fsync = v9fs_file_fsync,
671};
672
673const struct file_operations v9fs_file_operations_dotl = {
674 .llseek = generic_file_llseek,
675 .read_iter = v9fs_file_read_iter,
676 .write_iter = v9fs_file_write_iter,
677 .open = v9fs_file_open,
678 .release = v9fs_dir_release,
679 .lock = v9fs_file_lock_dotl,
680 .flock = v9fs_file_flock_dotl,
681 .mmap = generic_file_readonly_mmap,
682 .fsync = v9fs_file_fsync_dotl,
683};
684
685const struct file_operations v9fs_mmap_file_operations = {
686 .llseek = generic_file_llseek,
687 .read_iter = v9fs_mmap_file_read_iter,
688 .write_iter = v9fs_mmap_file_write_iter,
689 .open = v9fs_file_open,
690 .release = v9fs_dir_release,
691 .lock = v9fs_file_lock,
692 .mmap = v9fs_mmap_file_mmap,
693 .fsync = v9fs_file_fsync,
694};
695
696const struct file_operations v9fs_mmap_file_operations_dotl = {
697 .llseek = generic_file_llseek,
698 .read_iter = v9fs_mmap_file_read_iter,
699 .write_iter = v9fs_mmap_file_write_iter,
700 .open = v9fs_file_open,
701 .release = v9fs_dir_release,
702 .lock = v9fs_file_lock_dotl,
703 .flock = v9fs_file_flock_dotl,
704 .mmap = v9fs_mmap_file_mmap,
705 .fsync = v9fs_file_fsync_dotl,
706};
1/*
2 * linux/fs/9p/vfs_file.c
3 *
4 * This file contians vfs file ops for 9P2000.
5 *
6 * Copyright (C) 2004 by Eric Van Hensbergen <ericvh@gmail.com>
7 * Copyright (C) 2002 by Ron Minnich <rminnich@lanl.gov>
8 *
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.
12 *
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.
17 *
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
23 *
24 */
25
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>
41
42#include "v9fs.h"
43#include "v9fs_vfs.h"
44#include "fid.h"
45#include "cache.h"
46
47static const struct vm_operations_struct v9fs_file_vm_ops;
48
49/**
50 * v9fs_file_open - open a file (or directory)
51 * @inode: inode to be opened
52 * @file: file being opened
53 *
54 */
55
56int v9fs_file_open(struct inode *inode, struct file *file)
57{
58 int err;
59 struct v9fs_inode *v9inode;
60 struct v9fs_session_info *v9ses;
61 struct p9_fid *fid;
62 int omode;
63
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);
77
78 err = p9_client_open(fid, omode);
79 if (err < 0) {
80 p9_client_clunk(fid);
81 return err;
82 }
83 if (file->f_flags & O_TRUNC) {
84 i_size_write(inode, 0);
85 inode->i_blocks = 0;
86 }
87 if ((file->f_flags & O_APPEND) &&
88 (!v9fs_proto_dotu(v9ses) && !v9fs_proto_dotl(v9ses)))
89 generic_file_llseek(file, 0, SEEK_END);
90 }
91
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)) {
96 /*
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.
102 */
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;
108 }
109 v9inode->writeback_fid = (void *) fid;
110 }
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;
117out_error:
118 p9_client_clunk(file->private_data);
119 file->private_data = NULL;
120 return err;
121}
122
123/**
124 * v9fs_file_lock - lock a file (or directory)
125 * @filp: file to be locked
126 * @cmd: lock command
127 * @fl: file lock structure
128 *
129 * Bugs: this looks like a local only lock, we should extend into 9P
130 * by using open exclusive
131 */
132
133static int v9fs_file_lock(struct file *filp, int cmd, struct file_lock *fl)
134{
135 int res = 0;
136 struct inode *inode = filp->f_path.dentry->d_inode;
137
138 p9_debug(P9_DEBUG_VFS, "filp: %p lock: %p\n", filp, fl);
139
140 /* No mandatory locks */
141 if (__mandatory_lock(inode) && fl->fl_type != F_UNLCK)
142 return -ENOLCK;
143
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);
147 }
148
149 return res;
150}
151
152static int v9fs_file_do_lock(struct file *filp, int cmd, struct file_lock *fl)
153{
154 struct p9_flock flock;
155 struct p9_fid *fid;
156 uint8_t status;
157 int res = 0;
158 unsigned char fl_type;
159
160 fid = filp->private_data;
161 BUG_ON(fid == NULL);
162
163 if ((fl->fl_flags & FL_POSIX) != FL_POSIX)
164 BUG();
165
166 res = posix_lock_file_wait(filp, fl);
167 if (res < 0)
168 goto out;
169
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;
183 }
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;
193
194 /*
195 * if its a blocked request and we get P9_LOCK_BLOCKED as the status
196 * for lock request, keep on trying
197 */
198 for (;;) {
199 res = p9_client_lock_dotl(fid, &flock, &status);
200 if (res < 0)
201 break;
202
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;
209 }
210
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();
225 }
226
227 /*
228 * incase server returned error for lock request, revert
229 * it locally
230 */
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;
236 }
237out:
238 return res;
239}
240
241static int v9fs_file_getlock(struct file *filp, struct file_lock *fl)
242{
243 struct p9_getlock glock;
244 struct p9_fid *fid;
245 int res = 0;
246
247 fid = filp->private_data;
248 BUG_ON(fid == NULL);
249
250 posix_test_lock(filp, fl);
251 /*
252 * if we have a conflicting lock locally, no need to validate
253 * with server
254 */
255 if (fl->fl_type != F_UNLCK)
256 return res;
257
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;
268
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;
283 }
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;
291 }
292 return res;
293}
294
295/**
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
300 *
301 */
302
303static int v9fs_file_lock_dotl(struct file *filp, int cmd, struct file_lock *fl)
304{
305 struct inode *inode = filp->f_path.dentry->d_inode;
306 int ret = -ENOLCK;
307
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);
310
311 /* No mandatory locks */
312 if (__mandatory_lock(inode) && fl->fl_type != F_UNLCK)
313 goto out_err;
314
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);
318 }
319
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;
326out_err:
327 return ret;
328}
329
330/**
331 * v9fs_file_flock_dotl - lock a file
332 * @filp: file to be locked
333 * @cmd: lock command
334 * @fl: file lock structure
335 *
336 */
337
338static int v9fs_file_flock_dotl(struct file *filp, int cmd,
339 struct file_lock *fl)
340{
341 struct inode *inode = filp->f_path.dentry->d_inode;
342 int ret = -ENOLCK;
343
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);
346
347 /* No mandatory locks */
348 if (__mandatory_lock(inode) && fl->fl_type != F_UNLCK)
349 goto out_err;
350
351 if (!(fl->fl_flags & FL_FLOCK))
352 goto out_err;
353
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);
357 }
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;
364
365 if (IS_SETLK(cmd) | IS_SETLKW(cmd))
366 ret = v9fs_file_do_lock(filp, cmd, fl);
367 else
368 ret = -EINVAL;
369out_err:
370 return ret;
371}
372
373/**
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
380 *
381 */
382ssize_t
383v9fs_fid_readn(struct p9_fid *fid, char *data, char __user *udata, u32 count,
384 u64 offset)
385{
386 int n, total, size;
387
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;
397
398 if (data)
399 data += n;
400 if (udata)
401 udata += n;
402
403 offset += n;
404 count -= n;
405 total += n;
406 } while (count > 0 && n == size);
407
408 if (n < 0)
409 total = n;
410
411 return total;
412}
413
414/**
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
421 *
422 */
423ssize_t
424v9fs_file_readn(struct file *filp, char *data, char __user *udata, u32 count,
425 u64 offset)
426{
427 return v9fs_fid_readn(filp->private_data, data, udata, count, offset);
428}
429
430/**
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
436 *
437 */
438
439static ssize_t
440v9fs_file_read(struct file *filp, char __user *udata, size_t count,
441 loff_t * offset)
442{
443 int ret;
444 struct p9_fid *fid;
445 size_t size;
446
447 p9_debug(P9_DEBUG_VFS, "count %zu offset %lld\n", count, *offset);
448 fid = filp->private_data;
449
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);
455
456 if (ret > 0)
457 *offset += ret;
458
459 return ret;
460}
461
462ssize_t
463v9fs_file_write_internal(struct inode *inode, struct p9_fid *fid,
464 const char __user *data, size_t count,
465 loff_t *offset, int invalidate)
466{
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;
473
474 p9_debug(P9_DEBUG_VFS, "data %p count %d offset %x\n",
475 data, (int)count, (int)*offset);
476
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);
485
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);
497 }
498 }
499 if (n < 0)
500 return n;
501
502 return total;
503}
504
505/**
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
511 *
512 */
513static ssize_t
514v9fs_file_write(struct file *filp, const char __user * data,
515 size_t count, loff_t *offset)
516{
517 ssize_t retval = 0;
518 loff_t origin = *offset;
519
520
521 retval = generic_write_checks(filp, &origin, &count, 0);
522 if (retval)
523 goto out;
524
525 retval = -EINVAL;
526 if ((ssize_t) count < 0)
527 goto out;
528 retval = 0;
529 if (!count)
530 goto out;
531
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;
538out:
539 return retval;
540}
541
542
543static int v9fs_file_fsync(struct file *filp, loff_t start, loff_t end,
544 int datasync)
545{
546 struct p9_fid *fid;
547 struct inode *inode = filp->f_mapping->host;
548 struct p9_wstat wstat;
549 int retval;
550
551 retval = filemap_write_and_wait_range(inode->i_mapping, start, end);
552 if (retval)
553 return retval;
554
555 mutex_lock(&inode->i_mutex);
556 p9_debug(P9_DEBUG_VFS, "filp %p datasync %x\n", filp, datasync);
557
558 fid = filp->private_data;
559 v9fs_blank_wstat(&wstat);
560
561 retval = p9_client_wstat(fid, &wstat);
562 mutex_unlock(&inode->i_mutex);
563
564 return retval;
565}
566
567int v9fs_file_fsync_dotl(struct file *filp, loff_t start, loff_t end,
568 int datasync)
569{
570 struct p9_fid *fid;
571 struct inode *inode = filp->f_mapping->host;
572 int retval;
573
574 retval = filemap_write_and_wait_range(inode->i_mapping, start, end);
575 if (retval)
576 return retval;
577
578 mutex_lock(&inode->i_mutex);
579 p9_debug(P9_DEBUG_VFS, "filp %p datasync %x\n", filp, datasync);
580
581 fid = filp->private_data;
582
583 retval = p9_client_fsync(fid, datasync);
584 mutex_unlock(&inode->i_mutex);
585
586 return retval;
587}
588
589static int
590v9fs_file_mmap(struct file *file, struct vm_area_struct *vma)
591{
592 int retval;
593
594 retval = generic_file_mmap(file, vma);
595 if (!retval)
596 vma->vm_ops = &v9fs_file_vm_ops;
597
598 return retval;
599}
600
601static int
602v9fs_vm_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf)
603{
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;
608
609
610 p9_debug(P9_DEBUG_VFS, "page %p fid %lx\n",
611 page, (unsigned long)filp->private_data);
612
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;
620
621 return VM_FAULT_LOCKED;
622out_unlock:
623 unlock_page(page);
624 return VM_FAULT_NOPAGE;
625}
626
627static ssize_t
628v9fs_direct_read(struct file *filp, char __user *udata, size_t count,
629 loff_t *offsetp)
630{
631 loff_t size, offset;
632 struct inode *inode;
633 struct address_space *mapping;
634
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);
644
645 return v9fs_file_read(filp, udata, count, offsetp);
646}
647
648/**
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
654 *
655 */
656static ssize_t
657v9fs_cached_file_read(struct file *filp, char __user *data, size_t count,
658 loff_t *offset)
659{
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);
663}
664
665static ssize_t
666v9fs_direct_write(struct file *filp, const char __user * data,
667 size_t count, loff_t *offsetp)
668{
669 loff_t offset;
670 ssize_t retval;
671 struct inode *inode;
672 struct address_space *mapping;
673
674 offset = *offsetp;
675 mapping = filp->f_mapping;
676 inode = mapping->host;
677 if (!count)
678 return 0;
679
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;
685 /*
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
690 */
691 if (mapping->nrpages) {
692 pgoff_t pg_start = offset >> PAGE_CACHE_SHIFT;
693 pgoff_t pg_end = (offset + count - 1) >> PAGE_CACHE_SHIFT;
694
695 retval = invalidate_inode_pages2_range(mapping,
696 pg_start, pg_end);
697 /*
698 * If a page can not be invalidated, fall back
699 * to buffered write.
700 */
701 if (retval) {
702 if (retval == -EBUSY)
703 goto buff_write;
704 goto err_out;
705 }
706 }
707 retval = v9fs_file_write(filp, data, count, offsetp);
708err_out:
709 mutex_unlock(&inode->i_mutex);
710 return retval;
711
712buff_write:
713 mutex_unlock(&inode->i_mutex);
714 return do_sync_write(filp, data, count, offsetp);
715}
716
717/**
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
723 *
724 */
725static ssize_t
726v9fs_cached_file_write(struct file *filp, const char __user * data,
727 size_t count, loff_t *offset)
728{
729
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);
733}
734
735static const struct vm_operations_struct v9fs_file_vm_ops = {
736 .fault = filemap_fault,
737 .page_mkwrite = v9fs_vm_page_mkwrite,
738};
739
740
741const 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,
752};
753
754const 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,
766};
767
768const 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,
777};
778
779const 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,
789};