blob: 8e66b5a2d490129bb918a10ecf5e811a7488783e [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001/*
2 * linux/fs/nfs/file.c
3 *
4 * Copyright (C) 1992 Rick Sladkey
5 *
6 * Changes Copyright (C) 1994 by Florian La Roche
7 * - Do not copy data too often around in the kernel.
8 * - In nfs_file_read the return value of kmalloc wasn't checked.
9 * - Put in a better version of read look-ahead buffering. Original idea
10 * and implementation by Wai S Kok elekokws@ee.nus.sg.
11 *
12 * Expire cache on write to a file by Wai S Kok (Oct 1994).
13 *
14 * Total rewrite of read side for new NFS buffer cache.. Linus.
15 *
16 * nfs regular file handling functions
17 */
18
19#include <linux/time.h>
20#include <linux/kernel.h>
21#include <linux/errno.h>
22#include <linux/fcntl.h>
23#include <linux/stat.h>
24#include <linux/nfs_fs.h>
25#include <linux/nfs_mount.h>
26#include <linux/mm.h>
27#include <linux/slab.h>
28#include <linux/pagemap.h>
29#include <linux/smp_lock.h>
30
31#include <asm/uaccess.h>
32#include <asm/system.h>
33
34#include "delegation.h"
Chuck Lever91d5b472006-03-20 13:44:14 -050035#include "iostat.h"
Linus Torvalds1da177e2005-04-16 15:20:36 -070036
37#define NFSDBG_FACILITY NFSDBG_FILE
38
39static int nfs_file_open(struct inode *, struct file *);
40static int nfs_file_release(struct inode *, struct file *);
Trond Myklebust980802e2005-06-13 11:14:01 -040041static loff_t nfs_file_llseek(struct file *file, loff_t offset, int origin);
Linus Torvalds1da177e2005-04-16 15:20:36 -070042static int nfs_file_mmap(struct file *, struct vm_area_struct *);
43static ssize_t nfs_file_sendfile(struct file *, loff_t *, size_t, read_actor_t, void *);
Badari Pulavarty027445c2006-09-30 23:28:46 -070044static ssize_t nfs_file_read(struct kiocb *, const struct iovec *iov,
45 unsigned long nr_segs, loff_t pos);
46static ssize_t nfs_file_write(struct kiocb *, const struct iovec *iov,
47 unsigned long nr_segs, loff_t pos);
Miklos Szeredi75e1fcc2006-06-23 02:05:12 -070048static int nfs_file_flush(struct file *, fl_owner_t id);
Linus Torvalds1da177e2005-04-16 15:20:36 -070049static int nfs_fsync(struct file *, struct dentry *dentry, int datasync);
50static int nfs_check_flags(int flags);
51static int nfs_lock(struct file *filp, int cmd, struct file_lock *fl);
52static int nfs_flock(struct file *filp, int cmd, struct file_lock *fl);
53
Arjan van de Ven4b6f5d22006-03-28 01:56:42 -080054const struct file_operations nfs_file_operations = {
Trond Myklebust980802e2005-06-13 11:14:01 -040055 .llseek = nfs_file_llseek,
Linus Torvalds1da177e2005-04-16 15:20:36 -070056 .read = do_sync_read,
57 .write = do_sync_write,
Badari Pulavarty027445c2006-09-30 23:28:46 -070058 .aio_read = nfs_file_read,
59 .aio_write = nfs_file_write,
Linus Torvalds1da177e2005-04-16 15:20:36 -070060 .mmap = nfs_file_mmap,
61 .open = nfs_file_open,
62 .flush = nfs_file_flush,
63 .release = nfs_file_release,
64 .fsync = nfs_fsync,
65 .lock = nfs_lock,
66 .flock = nfs_flock,
67 .sendfile = nfs_file_sendfile,
68 .check_flags = nfs_check_flags,
69};
70
Arjan van de Ven92e1d5b2007-02-12 00:55:39 -080071const struct inode_operations nfs_file_inode_operations = {
Linus Torvalds1da177e2005-04-16 15:20:36 -070072 .permission = nfs_permission,
73 .getattr = nfs_getattr,
74 .setattr = nfs_setattr,
75};
76
Andreas Gruenbacherb7fa0552005-06-22 17:16:27 +000077#ifdef CONFIG_NFS_V3
Arjan van de Ven92e1d5b2007-02-12 00:55:39 -080078const struct inode_operations nfs3_file_inode_operations = {
Andreas Gruenbacherb7fa0552005-06-22 17:16:27 +000079 .permission = nfs_permission,
80 .getattr = nfs_getattr,
81 .setattr = nfs_setattr,
82 .listxattr = nfs3_listxattr,
83 .getxattr = nfs3_getxattr,
84 .setxattr = nfs3_setxattr,
85 .removexattr = nfs3_removexattr,
86};
87#endif /* CONFIG_NFS_v3 */
88
Linus Torvalds1da177e2005-04-16 15:20:36 -070089/* Hack for future NFS swap support */
90#ifndef IS_SWAPFILE
91# define IS_SWAPFILE(inode) (0)
92#endif
93
94static int nfs_check_flags(int flags)
95{
96 if ((flags & (O_APPEND | O_DIRECT)) == (O_APPEND | O_DIRECT))
97 return -EINVAL;
98
99 return 0;
100}
101
102/*
103 * Open file
104 */
105static int
106nfs_file_open(struct inode *inode, struct file *filp)
107{
Linus Torvalds1da177e2005-04-16 15:20:36 -0700108 int res;
109
110 res = nfs_check_flags(filp->f_flags);
111 if (res)
112 return res;
113
Chuck Lever91d5b472006-03-20 13:44:14 -0500114 nfs_inc_stats(inode, NFSIOS_VFSOPEN);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700115 lock_kernel();
David Howells1f163412006-08-22 20:06:11 -0400116 res = NFS_PROTO(inode)->file_open(inode, filp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700117 unlock_kernel();
118 return res;
119}
120
121static int
122nfs_file_release(struct inode *inode, struct file *filp)
123{
124 /* Ensure that dirty pages are flushed out with the right creds */
125 if (filp->f_mode & FMODE_WRITE)
126 filemap_fdatawrite(filp->f_mapping);
Chuck Lever91d5b472006-03-20 13:44:14 -0500127 nfs_inc_stats(inode, NFSIOS_VFSRELEASE);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700128 return NFS_PROTO(inode)->file_release(inode, filp);
129}
130
Trond Myklebust980802e2005-06-13 11:14:01 -0400131/**
132 * nfs_revalidate_size - Revalidate the file size
133 * @inode - pointer to inode struct
134 * @file - pointer to struct file
135 *
136 * Revalidates the file length. This is basically a wrapper around
137 * nfs_revalidate_inode() that takes into account the fact that we may
138 * have cached writes (in which case we don't care about the server's
139 * idea of what the file length is), or O_DIRECT (in which case we
140 * shouldn't trust the cache).
141 */
142static int nfs_revalidate_file_size(struct inode *inode, struct file *filp)
143{
144 struct nfs_server *server = NFS_SERVER(inode);
145 struct nfs_inode *nfsi = NFS_I(inode);
146
147 if (server->flags & NFS_MOUNT_NOAC)
148 goto force_reval;
149 if (filp->f_flags & O_DIRECT)
150 goto force_reval;
151 if (nfsi->npages != 0)
152 return 0;
Chuck Lever55296802005-08-18 11:24:09 -0700153 if (!(nfsi->cache_validity & NFS_INO_REVAL_PAGECACHE) && !nfs_attribute_timeout(inode))
Trond Myklebustfe51bee2005-06-22 17:16:30 +0000154 return 0;
Trond Myklebust980802e2005-06-13 11:14:01 -0400155force_reval:
156 return __nfs_revalidate_inode(server, inode);
157}
158
159static loff_t nfs_file_llseek(struct file *filp, loff_t offset, int origin)
160{
161 /* origin == SEEK_END => we must revalidate the cached file length */
Josef 'Jeff' Sipekaec5e172006-09-16 21:09:32 -0400162 if (origin == SEEK_END) {
Trond Myklebust980802e2005-06-13 11:14:01 -0400163 struct inode *inode = filp->f_mapping->host;
164 int retval = nfs_revalidate_file_size(inode, filp);
165 if (retval < 0)
166 return (loff_t)retval;
167 }
168 return remote_llseek(filp, offset, origin);
169}
170
Linus Torvalds1da177e2005-04-16 15:20:36 -0700171/*
172 * Flush all dirty pages, and check for write errors.
173 *
174 */
175static int
Miklos Szeredi75e1fcc2006-06-23 02:05:12 -0700176nfs_file_flush(struct file *file, fl_owner_t id)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700177{
178 struct nfs_open_context *ctx = (struct nfs_open_context *)file->private_data;
Josef "Jeff" Sipek01cce932006-12-08 02:36:40 -0800179 struct inode *inode = file->f_path.dentry->d_inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700180 int status;
181
182 dfprintk(VFS, "nfs: flush(%s/%ld)\n", inode->i_sb->s_id, inode->i_ino);
183
184 if ((file->f_mode & FMODE_WRITE) == 0)
185 return 0;
Chuck Lever91d5b472006-03-20 13:44:14 -0500186 nfs_inc_stats(inode, NFSIOS_VFSFLUSH);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700187 lock_kernel();
188 /* Ensure that data+attribute caches are up to date after close() */
189 status = nfs_wb_all(inode);
190 if (!status) {
191 status = ctx->error;
192 ctx->error = 0;
Trond Myklebust3338c142005-10-27 22:12:41 -0400193 if (!status)
194 nfs_revalidate_inode(NFS_SERVER(inode), inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700195 }
196 unlock_kernel();
197 return status;
198}
199
200static ssize_t
Badari Pulavarty027445c2006-09-30 23:28:46 -0700201nfs_file_read(struct kiocb *iocb, const struct iovec *iov,
202 unsigned long nr_segs, loff_t pos)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700203{
Josef "Jeff" Sipek01cce932006-12-08 02:36:40 -0800204 struct dentry * dentry = iocb->ki_filp->f_path.dentry;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700205 struct inode * inode = dentry->d_inode;
206 ssize_t result;
Badari Pulavarty027445c2006-09-30 23:28:46 -0700207 size_t count = iov_length(iov, nr_segs);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700208
209#ifdef CONFIG_NFS_DIRECTIO
210 if (iocb->ki_filp->f_flags & O_DIRECT)
Badari Pulavarty027445c2006-09-30 23:28:46 -0700211 return nfs_file_direct_read(iocb, iov, nr_segs, pos);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700212#endif
213
214 dfprintk(VFS, "nfs: read(%s/%s, %lu@%lu)\n",
215 dentry->d_parent->d_name.name, dentry->d_name.name,
216 (unsigned long) count, (unsigned long) pos);
217
Trond Myklebust44b11872006-05-25 01:40:59 -0400218 result = nfs_revalidate_mapping(inode, iocb->ki_filp->f_mapping);
Chuck Lever91d5b472006-03-20 13:44:14 -0500219 nfs_add_stats(inode, NFSIOS_NORMALREADBYTES, count);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700220 if (!result)
Badari Pulavarty027445c2006-09-30 23:28:46 -0700221 result = generic_file_aio_read(iocb, iov, nr_segs, pos);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700222 return result;
223}
224
225static ssize_t
226nfs_file_sendfile(struct file *filp, loff_t *ppos, size_t count,
227 read_actor_t actor, void *target)
228{
Josef "Jeff" Sipek01cce932006-12-08 02:36:40 -0800229 struct dentry *dentry = filp->f_path.dentry;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700230 struct inode *inode = dentry->d_inode;
231 ssize_t res;
232
233 dfprintk(VFS, "nfs: sendfile(%s/%s, %lu@%Lu)\n",
234 dentry->d_parent->d_name.name, dentry->d_name.name,
235 (unsigned long) count, (unsigned long long) *ppos);
236
Trond Myklebust44b11872006-05-25 01:40:59 -0400237 res = nfs_revalidate_mapping(inode, filp->f_mapping);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700238 if (!res)
239 res = generic_file_sendfile(filp, ppos, count, actor, target);
240 return res;
241}
242
243static int
244nfs_file_mmap(struct file * file, struct vm_area_struct * vma)
245{
Josef "Jeff" Sipek01cce932006-12-08 02:36:40 -0800246 struct dentry *dentry = file->f_path.dentry;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700247 struct inode *inode = dentry->d_inode;
248 int status;
249
250 dfprintk(VFS, "nfs: mmap(%s/%s)\n",
251 dentry->d_parent->d_name.name, dentry->d_name.name);
252
Trond Myklebust44b11872006-05-25 01:40:59 -0400253 status = nfs_revalidate_mapping(inode, file->f_mapping);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700254 if (!status)
255 status = generic_file_mmap(file, vma);
256 return status;
257}
258
259/*
260 * Flush any dirty pages for this process, and check for write errors.
261 * The return status from this call provides a reliable indication of
262 * whether any write errors occurred for this process.
263 */
264static int
265nfs_fsync(struct file *file, struct dentry *dentry, int datasync)
266{
267 struct nfs_open_context *ctx = (struct nfs_open_context *)file->private_data;
268 struct inode *inode = dentry->d_inode;
269 int status;
270
271 dfprintk(VFS, "nfs: fsync(%s/%ld)\n", inode->i_sb->s_id, inode->i_ino);
272
Chuck Lever91d5b472006-03-20 13:44:14 -0500273 nfs_inc_stats(inode, NFSIOS_VFSFSYNC);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700274 lock_kernel();
275 status = nfs_wb_all(inode);
276 if (!status) {
277 status = ctx->error;
278 ctx->error = 0;
279 }
280 unlock_kernel();
281 return status;
282}
283
284/*
285 * This does the "real" work of the write. The generic routine has
286 * allocated the page, locked it, done all the page alignment stuff
287 * calculations etc. Now we should just copy the data from user
288 * space and write it back to the real medium..
289 *
290 * If the writer ends up delaying the write, the writer needs to
291 * increment the page use counts until he is done with the page.
292 */
293static int nfs_prepare_write(struct file *file, struct page *page, unsigned offset, unsigned to)
294{
295 return nfs_flush_incompatible(file, page);
296}
297
298static int nfs_commit_write(struct file *file, struct page *page, unsigned offset, unsigned to)
299{
300 long status;
301
302 lock_kernel();
303 status = nfs_updatepage(file, page, offset, to-offset);
304 unlock_kernel();
305 return status;
306}
307
NeilBrown2ff28e22006-03-26 01:37:18 -0800308static void nfs_invalidate_page(struct page *page, unsigned long offset)
Trond Myklebustcd52ed32006-03-20 13:44:04 -0500309{
Trond Myklebust1c759502006-10-09 16:18:38 -0400310 if (offset != 0)
311 return;
Trond Myklebustd2ccddf2006-05-31 01:13:38 -0400312 /* Cancel any unstarted writes on this page */
Trond Myklebust61822ab2006-12-05 00:35:42 -0500313 nfs_wb_page_priority(page->mapping->host, page, FLUSH_INVALIDATE);
Trond Myklebustcd52ed32006-03-20 13:44:04 -0500314}
315
316static int nfs_release_page(struct page *page, gfp_t gfp)
317{
Trond Myklebuste3db7692007-01-10 23:15:39 -0800318 /* If PagePrivate() is set, then the page is not freeable */
319 return 0;
320}
321
322static int nfs_launder_page(struct page *page)
323{
324 return nfs_wb_page(page->mapping->host, page);
Trond Myklebustcd52ed32006-03-20 13:44:04 -0500325}
326
Christoph Hellwigf5e54d62006-06-28 04:26:44 -0700327const struct address_space_operations nfs_file_aops = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700328 .readpage = nfs_readpage,
329 .readpages = nfs_readpages,
Trond Myklebust1a545332006-12-05 00:35:40 -0500330 .set_page_dirty = nfs_set_page_dirty,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700331 .writepage = nfs_writepage,
332 .writepages = nfs_writepages,
333 .prepare_write = nfs_prepare_write,
334 .commit_write = nfs_commit_write,
Trond Myklebustcd52ed32006-03-20 13:44:04 -0500335 .invalidatepage = nfs_invalidate_page,
336 .releasepage = nfs_release_page,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700337#ifdef CONFIG_NFS_DIRECTIO
338 .direct_IO = nfs_direct_IO,
339#endif
Trond Myklebuste3db7692007-01-10 23:15:39 -0800340 .launder_page = nfs_launder_page,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700341};
342
Badari Pulavarty027445c2006-09-30 23:28:46 -0700343static ssize_t nfs_file_write(struct kiocb *iocb, const struct iovec *iov,
344 unsigned long nr_segs, loff_t pos)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700345{
Josef "Jeff" Sipek01cce932006-12-08 02:36:40 -0800346 struct dentry * dentry = iocb->ki_filp->f_path.dentry;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700347 struct inode * inode = dentry->d_inode;
348 ssize_t result;
Badari Pulavarty027445c2006-09-30 23:28:46 -0700349 size_t count = iov_length(iov, nr_segs);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700350
351#ifdef CONFIG_NFS_DIRECTIO
352 if (iocb->ki_filp->f_flags & O_DIRECT)
Badari Pulavarty027445c2006-09-30 23:28:46 -0700353 return nfs_file_direct_write(iocb, iov, nr_segs, pos);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700354#endif
355
Badari Pulavarty027445c2006-09-30 23:28:46 -0700356 dfprintk(VFS, "nfs: write(%s/%s(%ld), %lu@%Ld)\n",
Linus Torvalds1da177e2005-04-16 15:20:36 -0700357 dentry->d_parent->d_name.name, dentry->d_name.name,
Badari Pulavarty027445c2006-09-30 23:28:46 -0700358 inode->i_ino, (unsigned long) count, (long long) pos);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700359
360 result = -EBUSY;
361 if (IS_SWAPFILE(inode))
362 goto out_swapfile;
Trond Myklebust7d52e862005-06-22 17:16:30 +0000363 /*
364 * O_APPEND implies that we must revalidate the file length.
365 */
366 if (iocb->ki_filp->f_flags & O_APPEND) {
367 result = nfs_revalidate_file_size(inode, iocb->ki_filp);
368 if (result)
369 goto out;
Trond Myklebustfe51bee2005-06-22 17:16:30 +0000370 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700371
372 result = count;
373 if (!count)
374 goto out;
375
Chuck Lever91d5b472006-03-20 13:44:14 -0500376 nfs_add_stats(inode, NFSIOS_NORMALWRITTENBYTES, count);
Badari Pulavarty027445c2006-09-30 23:28:46 -0700377 result = generic_file_aio_write(iocb, iov, nr_segs, pos);
Trond Myklebust200baa22006-12-05 00:35:40 -0500378 /* Return error values for O_SYNC and IS_SYNC() */
379 if (result >= 0 && (IS_SYNC(inode) || (iocb->ki_filp->f_flags & O_SYNC))) {
380 int err = nfs_fsync(iocb->ki_filp, dentry, 1);
381 if (err < 0)
382 result = err;
383 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700384out:
385 return result;
386
387out_swapfile:
388 printk(KERN_INFO "NFS: attempt to write to active swap file!\n");
389 goto out;
390}
391
392static int do_getlk(struct file *filp, int cmd, struct file_lock *fl)
393{
Andy Adamson8dc7c312006-03-20 13:44:26 -0500394 struct file_lock cfl;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700395 struct inode *inode = filp->f_mapping->host;
396 int status = 0;
397
398 lock_kernel();
Trond Myklebust039c4d72005-10-18 14:20:16 -0700399 /* Try local locking first */
Andy Adamson8dc7c312006-03-20 13:44:26 -0500400 if (posix_test_lock(filp, fl, &cfl)) {
Trond Myklebuste4cd0382006-03-20 13:44:44 -0500401 fl->fl_start = cfl.fl_start;
402 fl->fl_end = cfl.fl_end;
403 fl->fl_type = cfl.fl_type;
404 fl->fl_pid = cfl.fl_pid;
Trond Myklebust039c4d72005-10-18 14:20:16 -0700405 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700406 }
Trond Myklebust039c4d72005-10-18 14:20:16 -0700407
408 if (nfs_have_delegation(inode, FMODE_READ))
409 goto out_noconflict;
410
411 if (NFS_SERVER(inode)->flags & NFS_MOUNT_NONLM)
412 goto out_noconflict;
413
414 status = NFS_PROTO(inode)->lock(filp, cmd, fl);
415out:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700416 unlock_kernel();
417 return status;
Trond Myklebust039c4d72005-10-18 14:20:16 -0700418out_noconflict:
419 fl->fl_type = F_UNLCK;
420 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700421}
422
423static int do_vfs_lock(struct file *file, struct file_lock *fl)
424{
425 int res = 0;
426 switch (fl->fl_flags & (FL_POSIX|FL_FLOCK)) {
427 case FL_POSIX:
428 res = posix_lock_file_wait(file, fl);
429 break;
430 case FL_FLOCK:
431 res = flock_lock_file_wait(file, fl);
432 break;
433 default:
434 BUG();
435 }
436 if (res < 0)
Neil Brown46bae1a2007-01-30 14:36:01 -0800437 dprintk(KERN_WARNING "%s: VFS is out of sync with lock manager"
438 " - error %d!\n",
439 __FUNCTION__, res);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700440 return res;
441}
442
443static int do_unlk(struct file *filp, int cmd, struct file_lock *fl)
444{
445 struct inode *inode = filp->f_mapping->host;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700446 int status;
447
Linus Torvalds1da177e2005-04-16 15:20:36 -0700448 /*
449 * Flush all pending writes before doing anything
450 * with locks..
451 */
Trond Myklebust29884df2005-12-13 16:13:54 -0500452 nfs_sync_mapping(filp->f_mapping);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700453
454 /* NOTE: special case
455 * If we're signalled while cleaning up locks on process exit, we
456 * still need to complete the unlock.
457 */
458 lock_kernel();
459 /* Use local locking if mounted with "-onolock" */
460 if (!(NFS_SERVER(inode)->flags & NFS_MOUNT_NONLM))
461 status = NFS_PROTO(inode)->lock(filp, cmd, fl);
462 else
463 status = do_vfs_lock(filp, fl);
464 unlock_kernel();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700465 return status;
466}
467
468static int do_setlk(struct file *filp, int cmd, struct file_lock *fl)
469{
470 struct inode *inode = filp->f_mapping->host;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700471 int status;
472
Linus Torvalds1da177e2005-04-16 15:20:36 -0700473 /*
474 * Flush all pending writes before doing anything
475 * with locks..
476 */
Trond Myklebust29884df2005-12-13 16:13:54 -0500477 status = nfs_sync_mapping(filp->f_mapping);
478 if (status != 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700479 goto out;
480
481 lock_kernel();
482 /* Use local locking if mounted with "-onolock" */
483 if (!(NFS_SERVER(inode)->flags & NFS_MOUNT_NONLM)) {
484 status = NFS_PROTO(inode)->lock(filp, cmd, fl);
485 /* If we were signalled we still need to ensure that
486 * we clean up any state on the server. We therefore
487 * record the lock call as having succeeded in order to
488 * ensure that locks_remove_posix() cleans it out when
489 * the process exits.
490 */
491 if (status == -EINTR || status == -ERESTARTSYS)
492 do_vfs_lock(filp, fl);
493 } else
494 status = do_vfs_lock(filp, fl);
495 unlock_kernel();
496 if (status < 0)
497 goto out;
498 /*
499 * Make sure we clear the cache whenever we try to get the lock.
500 * This makes locking act as a cache coherency point.
501 */
Trond Myklebust29884df2005-12-13 16:13:54 -0500502 nfs_sync_mapping(filp->f_mapping);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700503 nfs_zap_caches(inode);
504out:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700505 return status;
506}
507
508/*
509 * Lock a (portion of) a file
510 */
511static int nfs_lock(struct file *filp, int cmd, struct file_lock *fl)
512{
513 struct inode * inode = filp->f_mapping->host;
514
515 dprintk("NFS: nfs_lock(f=%s/%ld, t=%x, fl=%x, r=%Ld:%Ld)\n",
516 inode->i_sb->s_id, inode->i_ino,
517 fl->fl_type, fl->fl_flags,
518 (long long)fl->fl_start, (long long)fl->fl_end);
Chuck Lever91d5b472006-03-20 13:44:14 -0500519 nfs_inc_stats(inode, NFSIOS_VFSLOCK);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700520
521 /* No mandatory locks over NFS */
ASANO Masahiro0800c5f2005-12-22 13:24:54 +0900522 if ((inode->i_mode & (S_ISGID | S_IXGRP)) == S_ISGID &&
523 fl->fl_type != F_UNLCK)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700524 return -ENOLCK;
525
526 if (IS_GETLK(cmd))
527 return do_getlk(filp, cmd, fl);
528 if (fl->fl_type == F_UNLCK)
529 return do_unlk(filp, cmd, fl);
530 return do_setlk(filp, cmd, fl);
531}
532
533/*
534 * Lock a (portion of) a file
535 */
536static int nfs_flock(struct file *filp, int cmd, struct file_lock *fl)
537{
Linus Torvalds1da177e2005-04-16 15:20:36 -0700538 dprintk("NFS: nfs_flock(f=%s/%ld, t=%x, fl=%x)\n",
Josef "Jeff" Sipek01cce932006-12-08 02:36:40 -0800539 filp->f_path.dentry->d_inode->i_sb->s_id,
540 filp->f_path.dentry->d_inode->i_ino,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700541 fl->fl_type, fl->fl_flags);
542
Linus Torvalds1da177e2005-04-16 15:20:36 -0700543 /*
544 * No BSD flocks over NFS allowed.
545 * Note: we could try to fake a POSIX lock request here by
546 * using ((u32) filp | 0x80000000) or some such as the pid.
547 * Not sure whether that would be unique, though, or whether
548 * that would break in other places.
549 */
550 if (!(fl->fl_flags & FL_FLOCK))
551 return -ENOLCK;
552
553 /* We're simulating flock() locks using posix locks on the server */
554 fl->fl_owner = (fl_owner_t)filp;
555 fl->fl_start = 0;
556 fl->fl_end = OFFSET_MAX;
557
558 if (fl->fl_type == F_UNLCK)
559 return do_unlk(filp, cmd, fl);
560 return do_setlk(filp, cmd, fl);
561}