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>
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>
47 static const struct vm_operations_struct v9fs_file_vm_ops;
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)
59 struct v9fs_inode *v9inode;
60 struct v9fs_session_info *v9ses;
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);
70 omode = v9fs_uflags2omode(file->f_flags,
71 v9fs_proto_dotu(v9ses));
72 fid = file->private_data;
74 fid = v9fs_fid_clone(file->f_path.dentry);
78 err = p9_client_open(fid, omode);
83 if ((file->f_flags & O_APPEND) &&
84 (!v9fs_proto_dotu(v9ses) && !v9fs_proto_dotl(v9ses)))
85 generic_file_llseek(file, 0, SEEK_END);
88 file->private_data = fid;
89 mutex_lock(&v9inode->v_mutex);
90 if (v9ses->cache && !v9inode->writeback_fid &&
91 ((file->f_flags & O_ACCMODE) != O_RDONLY)) {
93 * clone a fid and add it to writeback_fid
94 * we do it during open time instead of
95 * page dirty time via write_begin/page_mkwrite
96 * because we want write after unlink usecase
99 fid = v9fs_writeback_fid(file->f_path.dentry);
102 mutex_unlock(&v9inode->v_mutex);
105 v9inode->writeback_fid = (void *) fid;
107 mutex_unlock(&v9inode->v_mutex);
108 #ifdef CONFIG_9P_FSCACHE
110 v9fs_cache_inode_set_cookie(inode, file);
114 p9_client_clunk(file->private_data);
115 file->private_data = NULL;
120 * v9fs_file_lock - lock a file (or directory)
121 * @filp: file to be locked
123 * @fl: file lock structure
125 * Bugs: this looks like a local only lock, we should extend into 9P
126 * by using open exclusive
129 static int v9fs_file_lock(struct file *filp, int cmd, struct file_lock *fl)
132 struct inode *inode = filp->f_path.dentry->d_inode;
134 p9_debug(P9_DEBUG_VFS, "filp: %p lock: %p\n", filp, fl);
136 /* No mandatory locks */
137 if (__mandatory_lock(inode) && fl->fl_type != F_UNLCK)
140 if ((IS_SETLK(cmd) || IS_SETLKW(cmd)) && fl->fl_type != F_UNLCK) {
141 filemap_write_and_wait(inode->i_mapping);
142 invalidate_mapping_pages(&inode->i_data, 0, -1);
148 static int v9fs_file_do_lock(struct file *filp, int cmd, struct file_lock *fl)
150 struct p9_flock flock;
154 unsigned char fl_type;
156 fid = filp->private_data;
159 if ((fl->fl_flags & FL_POSIX) != FL_POSIX)
162 res = posix_lock_file_wait(filp, fl);
166 /* convert posix lock to p9 tlock args */
167 memset(&flock, 0, sizeof(flock));
168 /* map the lock type */
169 switch (fl->fl_type) {
171 flock.type = P9_LOCK_TYPE_RDLCK;
174 flock.type = P9_LOCK_TYPE_WRLCK;
177 flock.type = P9_LOCK_TYPE_UNLCK;
180 flock.start = fl->fl_start;
181 if (fl->fl_end == OFFSET_MAX)
184 flock.length = fl->fl_end - fl->fl_start + 1;
185 flock.proc_id = fl->fl_pid;
186 flock.client_id = utsname()->nodename;
188 flock.flags = P9_LOCK_FLAGS_BLOCK;
191 * if its a blocked request and we get P9_LOCK_BLOCKED as the status
192 * for lock request, keep on trying
195 res = p9_client_lock_dotl(fid, &flock, &status);
199 if (status != P9_LOCK_BLOCKED)
201 if (status == P9_LOCK_BLOCKED && !IS_SETLKW(cmd))
203 if (schedule_timeout_interruptible(P9_LOCK_TIMEOUT) != 0)
207 /* map 9p status to VFS status */
209 case P9_LOCK_SUCCESS:
212 case P9_LOCK_BLOCKED:
224 * incase server returned error for lock request, revert
227 if (res < 0 && fl->fl_type != F_UNLCK) {
228 fl_type = fl->fl_type;
229 fl->fl_type = F_UNLCK;
230 res = posix_lock_file_wait(filp, fl);
231 fl->fl_type = fl_type;
237 static int v9fs_file_getlock(struct file *filp, struct file_lock *fl)
239 struct p9_getlock glock;
243 fid = filp->private_data;
246 posix_test_lock(filp, fl);
248 * if we have a conflicting lock locally, no need to validate
251 if (fl->fl_type != F_UNLCK)
254 /* convert posix lock to p9 tgetlock args */
255 memset(&glock, 0, sizeof(glock));
256 glock.type = P9_LOCK_TYPE_UNLCK;
257 glock.start = fl->fl_start;
258 if (fl->fl_end == OFFSET_MAX)
261 glock.length = fl->fl_end - fl->fl_start + 1;
262 glock.proc_id = fl->fl_pid;
263 glock.client_id = utsname()->nodename;
265 res = p9_client_getlock_dotl(fid, &glock);
268 /* map 9p lock type to os lock type */
269 switch (glock.type) {
270 case P9_LOCK_TYPE_RDLCK:
271 fl->fl_type = F_RDLCK;
273 case P9_LOCK_TYPE_WRLCK:
274 fl->fl_type = F_WRLCK;
276 case P9_LOCK_TYPE_UNLCK:
277 fl->fl_type = F_UNLCK;
280 if (glock.type != P9_LOCK_TYPE_UNLCK) {
281 fl->fl_start = glock.start;
282 if (glock.length == 0)
283 fl->fl_end = OFFSET_MAX;
285 fl->fl_end = glock.start + glock.length - 1;
286 fl->fl_pid = glock.proc_id;
292 * v9fs_file_lock_dotl - lock a file (or directory)
293 * @filp: file to be locked
295 * @fl: file lock structure
299 static int v9fs_file_lock_dotl(struct file *filp, int cmd, struct file_lock *fl)
301 struct inode *inode = filp->f_path.dentry->d_inode;
304 p9_debug(P9_DEBUG_VFS, "filp: %p cmd:%d lock: %p name: %s\n",
305 filp, cmd, fl, filp->f_path.dentry->d_name.name);
307 /* No mandatory locks */
308 if (__mandatory_lock(inode) && fl->fl_type != F_UNLCK)
311 if ((IS_SETLK(cmd) || IS_SETLKW(cmd)) && fl->fl_type != F_UNLCK) {
312 filemap_write_and_wait(inode->i_mapping);
313 invalidate_mapping_pages(&inode->i_data, 0, -1);
316 if (IS_SETLK(cmd) || IS_SETLKW(cmd))
317 ret = v9fs_file_do_lock(filp, cmd, fl);
318 else if (IS_GETLK(cmd))
319 ret = v9fs_file_getlock(filp, fl);
327 * v9fs_file_flock_dotl - lock a file
328 * @filp: file to be locked
330 * @fl: file lock structure
334 static int v9fs_file_flock_dotl(struct file *filp, int cmd,
335 struct file_lock *fl)
337 struct inode *inode = filp->f_path.dentry->d_inode;
340 p9_debug(P9_DEBUG_VFS, "filp: %p cmd:%d lock: %p name: %s\n",
341 filp, cmd, fl, filp->f_path.dentry->d_name.name);
343 /* No mandatory locks */
344 if (__mandatory_lock(inode) && fl->fl_type != F_UNLCK)
347 if (!(fl->fl_flags & FL_FLOCK))
350 if ((IS_SETLK(cmd) || IS_SETLKW(cmd)) && fl->fl_type != F_UNLCK) {
351 filemap_write_and_wait(inode->i_mapping);
352 invalidate_mapping_pages(&inode->i_data, 0, -1);
354 /* Convert flock to posix lock */
355 fl->fl_owner = (fl_owner_t)filp;
357 fl->fl_end = OFFSET_MAX;
358 fl->fl_flags |= FL_POSIX;
359 fl->fl_flags ^= FL_FLOCK;
361 if (IS_SETLK(cmd) | IS_SETLKW(cmd))
362 ret = v9fs_file_do_lock(filp, cmd, fl);
370 * v9fs_fid_readn - read from a fid
372 * @data: data buffer to read data into
373 * @udata: user data buffer to read data into
374 * @count: size of buffer
375 * @offset: offset at which to read data
379 v9fs_fid_readn(struct p9_fid *fid, char *data, char __user *udata, u32 count,
384 p9_debug(P9_DEBUG_VFS, "fid %d offset %llu count %d\n",
385 fid->fid, (long long unsigned)offset, count);
388 size = fid->iounit ? fid->iounit : fid->clnt->msize - P9_IOHDRSZ;
390 n = p9_client_read(fid, data, udata, offset, count);
402 } while (count > 0 && n == size);
411 * v9fs_file_readn - read from a file
412 * @filp: file pointer to read
413 * @data: data buffer to read data into
414 * @udata: user data buffer to read data into
415 * @count: size of buffer
416 * @offset: offset at which to read data
420 v9fs_file_readn(struct file *filp, char *data, char __user *udata, u32 count,
423 return v9fs_fid_readn(filp->private_data, data, udata, count, offset);
427 * v9fs_file_read - read from a file
428 * @filp: file pointer to read
429 * @udata: user data buffer to read data into
430 * @count: size of buffer
431 * @offset: offset at which to read data
436 v9fs_file_read(struct file *filp, char __user *udata, size_t count,
443 p9_debug(P9_DEBUG_VFS, "count %zu offset %lld\n", count, *offset);
444 fid = filp->private_data;
446 size = fid->iounit ? fid->iounit : fid->clnt->msize - P9_IOHDRSZ;
448 ret = v9fs_file_readn(filp, NULL, udata, count, *offset);
450 ret = p9_client_read(fid, NULL, udata, *offset, count);
459 v9fs_file_write_internal(struct inode *inode, struct p9_fid *fid,
460 const char __user *data, size_t count,
461 loff_t *offset, int invalidate)
466 struct p9_client *clnt;
467 loff_t origin = *offset;
468 unsigned long pg_start, pg_end;
470 p9_debug(P9_DEBUG_VFS, "data %p count %d offset %x\n",
471 data, (int)count, (int)*offset);
475 n = p9_client_write(fid, NULL, data+total, origin+total, count);
482 if (invalidate && (total > 0)) {
483 pg_start = origin >> PAGE_CACHE_SHIFT;
484 pg_end = (origin + total - 1) >> PAGE_CACHE_SHIFT;
485 if (inode->i_mapping && inode->i_mapping->nrpages)
486 invalidate_inode_pages2_range(inode->i_mapping,
489 i_size = i_size_read(inode);
490 if (*offset > i_size) {
491 inode_add_bytes(inode, *offset - i_size);
492 i_size_write(inode, *offset);
502 * v9fs_file_write - write to a file
503 * @filp: file pointer to write
504 * @data: data buffer to write data from
505 * @count: size of buffer
506 * @offset: offset at which to write data
510 v9fs_file_write(struct file *filp, const char __user * data,
511 size_t count, loff_t *offset)
514 loff_t origin = *offset;
517 retval = generic_write_checks(filp, &origin, &count, 0);
522 if ((ssize_t) count < 0)
528 retval = v9fs_file_write_internal(filp->f_path.dentry->d_inode,
530 data, count, &origin, 1);
531 /* update offset on successful write */
539 static int v9fs_file_fsync(struct file *filp, loff_t start, loff_t end,
543 struct inode *inode = filp->f_mapping->host;
544 struct p9_wstat wstat;
547 retval = filemap_write_and_wait_range(inode->i_mapping, start, end);
551 mutex_lock(&inode->i_mutex);
552 p9_debug(P9_DEBUG_VFS, "filp %p datasync %x\n", filp, datasync);
554 fid = filp->private_data;
555 v9fs_blank_wstat(&wstat);
557 retval = p9_client_wstat(fid, &wstat);
558 mutex_unlock(&inode->i_mutex);
563 int v9fs_file_fsync_dotl(struct file *filp, loff_t start, loff_t end,
567 struct inode *inode = filp->f_mapping->host;
570 retval = filemap_write_and_wait_range(inode->i_mapping, start, end);
574 mutex_lock(&inode->i_mutex);
575 p9_debug(P9_DEBUG_VFS, "filp %p datasync %x\n", filp, datasync);
577 fid = filp->private_data;
579 retval = p9_client_fsync(fid, datasync);
580 mutex_unlock(&inode->i_mutex);
586 v9fs_file_mmap(struct file *file, struct vm_area_struct *vma)
590 retval = generic_file_mmap(file, vma);
592 vma->vm_ops = &v9fs_file_vm_ops;
598 v9fs_vm_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf)
600 struct v9fs_inode *v9inode;
601 struct page *page = vmf->page;
602 struct file *filp = vma->vm_file;
603 struct inode *inode = filp->f_path.dentry->d_inode;
606 p9_debug(P9_DEBUG_VFS, "page %p fid %lx\n",
607 page, (unsigned long)filp->private_data);
609 /* Update file times before taking page lock */
610 file_update_time(filp);
612 v9inode = V9FS_I(inode);
613 /* make sure the cache has finished storing the page */
614 v9fs_fscache_wait_on_page_write(inode, page);
615 BUG_ON(!v9inode->writeback_fid);
617 if (page->mapping != inode->i_mapping)
620 return VM_FAULT_LOCKED;
623 return VM_FAULT_NOPAGE;
627 v9fs_direct_read(struct file *filp, char __user *udata, size_t count,
632 struct address_space *mapping;
635 mapping = filp->f_mapping;
636 inode = mapping->host;
639 size = i_size_read(inode);
641 filemap_write_and_wait_range(mapping, offset,
644 return v9fs_file_read(filp, udata, count, offsetp);
648 * v9fs_cached_file_read - read from a file
649 * @filp: file pointer to read
650 * @udata: user data buffer to read data into
651 * @count: size of buffer
652 * @offset: offset at which to read data
656 v9fs_cached_file_read(struct file *filp, char __user *data, size_t count,
659 if (filp->f_flags & O_DIRECT)
660 return v9fs_direct_read(filp, data, count, offset);
661 return do_sync_read(filp, data, count, offset);
665 v9fs_direct_write(struct file *filp, const char __user * data,
666 size_t count, loff_t *offsetp)
671 struct address_space *mapping;
674 mapping = filp->f_mapping;
675 inode = mapping->host;
679 mutex_lock(&inode->i_mutex);
680 retval = filemap_write_and_wait_range(mapping, offset,
685 * After a write we want buffered reads to be sure to go to disk to get
686 * the new data. We invalidate clean cached page from the region we're
687 * about to write. We do this *before* the write so that if we fail
688 * here we fall back to buffered write
690 if (mapping->nrpages) {
691 pgoff_t pg_start = offset >> PAGE_CACHE_SHIFT;
692 pgoff_t pg_end = (offset + count - 1) >> PAGE_CACHE_SHIFT;
694 retval = invalidate_inode_pages2_range(mapping,
697 * If a page can not be invalidated, fall back
701 if (retval == -EBUSY)
706 retval = v9fs_file_write(filp, data, count, offsetp);
708 mutex_unlock(&inode->i_mutex);
712 mutex_unlock(&inode->i_mutex);
713 return do_sync_write(filp, data, count, offsetp);
717 * v9fs_cached_file_write - write to a file
718 * @filp: file pointer to write
719 * @data: data buffer to write data from
720 * @count: size of buffer
721 * @offset: offset at which to write data
725 v9fs_cached_file_write(struct file *filp, const char __user * data,
726 size_t count, loff_t *offset)
729 if (filp->f_flags & O_DIRECT)
730 return v9fs_direct_write(filp, data, count, offset);
731 return do_sync_write(filp, data, count, offset);
734 static const struct vm_operations_struct v9fs_file_vm_ops = {
735 .fault = filemap_fault,
736 .page_mkwrite = v9fs_vm_page_mkwrite,
737 .remap_pages = generic_file_remap_pages,
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,