blob: 28b8c3f3cda3af002e99044fee907f119ac78afb [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>
Linus Torvalds1da177e2005-04-16 15:20:36 -070027#include <linux/pagemap.h>
Alexey Dobriyane8edc6e2007-05-21 01:22:52 +040028#include <linux/aio.h>
Tejun Heo5a0e3ad2010-03-24 17:04:11 +090029#include <linux/gfp.h>
Trond Myklebustb608b282010-07-30 15:31:54 -040030#include <linux/swap.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070031
32#include <asm/uaccess.h>
33#include <asm/system.h>
34
35#include "delegation.h"
Trond Myklebust94387fb2007-07-22 17:09:05 -040036#include "internal.h"
Chuck Lever91d5b472006-03-20 13:44:14 -050037#include "iostat.h"
David Howells545db452009-04-03 16:42:44 +010038#include "fscache.h"
Benny Halevye5e94012010-10-20 00:18:01 -040039#include "pnfs.h"
Linus Torvalds1da177e2005-04-16 15:20:36 -070040
41#define NFSDBG_FACILITY NFSDBG_FILE
42
43static int nfs_file_open(struct inode *, struct file *);
44static int nfs_file_release(struct inode *, struct file *);
Trond Myklebust980802e2005-06-13 11:14:01 -040045static loff_t nfs_file_llseek(struct file *file, loff_t offset, int origin);
Linus Torvalds1da177e2005-04-16 15:20:36 -070046static int nfs_file_mmap(struct file *, struct vm_area_struct *);
Jens Axboef0930ff2007-06-01 11:51:43 +020047static ssize_t nfs_file_splice_read(struct file *filp, loff_t *ppos,
48 struct pipe_inode_info *pipe,
49 size_t count, unsigned int flags);
Badari Pulavarty027445c2006-09-30 23:28:46 -070050static ssize_t nfs_file_read(struct kiocb *, const struct iovec *iov,
51 unsigned long nr_segs, loff_t pos);
Suresh Jayaramanbf40d342009-06-17 18:02:09 -070052static ssize_t nfs_file_splice_write(struct pipe_inode_info *pipe,
53 struct file *filp, loff_t *ppos,
54 size_t count, unsigned int flags);
Badari Pulavarty027445c2006-09-30 23:28:46 -070055static ssize_t nfs_file_write(struct kiocb *, const struct iovec *iov,
56 unsigned long nr_segs, loff_t pos);
Miklos Szeredi75e1fcc2006-06-23 02:05:12 -070057static int nfs_file_flush(struct file *, fl_owner_t id);
Josef Bacik02c24a82011-07-16 20:44:56 -040058static int nfs_file_fsync(struct file *, loff_t, loff_t, int datasync);
Linus Torvalds1da177e2005-04-16 15:20:36 -070059static int nfs_check_flags(int flags);
60static int nfs_lock(struct file *filp, int cmd, struct file_lock *fl);
61static int nfs_flock(struct file *filp, int cmd, struct file_lock *fl);
J. Bruce Fields370f6592007-06-08 15:23:34 -040062static int nfs_setlease(struct file *file, long arg, struct file_lock **fl);
Linus Torvalds1da177e2005-04-16 15:20:36 -070063
Alexey Dobriyanf0f37e22009-09-27 22:29:37 +040064static const struct vm_operations_struct nfs_file_vm_ops;
Trond Myklebust94387fb2007-07-22 17:09:05 -040065
Arjan van de Ven4b6f5d22006-03-28 01:56:42 -080066const struct file_operations nfs_file_operations = {
Trond Myklebust980802e2005-06-13 11:14:01 -040067 .llseek = nfs_file_llseek,
Linus Torvalds1da177e2005-04-16 15:20:36 -070068 .read = do_sync_read,
69 .write = do_sync_write,
Badari Pulavarty027445c2006-09-30 23:28:46 -070070 .aio_read = nfs_file_read,
71 .aio_write = nfs_file_write,
Linus Torvalds1da177e2005-04-16 15:20:36 -070072 .mmap = nfs_file_mmap,
73 .open = nfs_file_open,
74 .flush = nfs_file_flush,
75 .release = nfs_file_release,
Chuck Lever54917782008-05-27 16:29:07 -040076 .fsync = nfs_file_fsync,
Linus Torvalds1da177e2005-04-16 15:20:36 -070077 .lock = nfs_lock,
78 .flock = nfs_flock,
Jens Axboef0930ff2007-06-01 11:51:43 +020079 .splice_read = nfs_file_splice_read,
Suresh Jayaramanbf40d342009-06-17 18:02:09 -070080 .splice_write = nfs_file_splice_write,
Linus Torvalds1da177e2005-04-16 15:20:36 -070081 .check_flags = nfs_check_flags,
J. Bruce Fields370f6592007-06-08 15:23:34 -040082 .setlease = nfs_setlease,
Linus Torvalds1da177e2005-04-16 15:20:36 -070083};
84
Arjan van de Ven92e1d5b2007-02-12 00:55:39 -080085const struct inode_operations nfs_file_inode_operations = {
Linus Torvalds1da177e2005-04-16 15:20:36 -070086 .permission = nfs_permission,
87 .getattr = nfs_getattr,
88 .setattr = nfs_setattr,
89};
90
Andreas Gruenbacherb7fa0552005-06-22 17:16:27 +000091#ifdef CONFIG_NFS_V3
Arjan van de Ven92e1d5b2007-02-12 00:55:39 -080092const struct inode_operations nfs3_file_inode_operations = {
Andreas Gruenbacherb7fa0552005-06-22 17:16:27 +000093 .permission = nfs_permission,
94 .getattr = nfs_getattr,
95 .setattr = nfs_setattr,
96 .listxattr = nfs3_listxattr,
97 .getxattr = nfs3_getxattr,
98 .setxattr = nfs3_setxattr,
99 .removexattr = nfs3_removexattr,
100};
101#endif /* CONFIG_NFS_v3 */
102
Linus Torvalds1da177e2005-04-16 15:20:36 -0700103/* Hack for future NFS swap support */
104#ifndef IS_SWAPFILE
105# define IS_SWAPFILE(inode) (0)
106#endif
107
108static int nfs_check_flags(int flags)
109{
110 if ((flags & (O_APPEND | O_DIRECT)) == (O_APPEND | O_DIRECT))
111 return -EINVAL;
112
113 return 0;
114}
115
116/*
117 * Open file
118 */
119static int
120nfs_file_open(struct inode *inode, struct file *filp)
121{
Linus Torvalds1da177e2005-04-16 15:20:36 -0700122 int res;
123
Chuck Lever6da24bc2008-06-11 17:55:58 -0400124 dprintk("NFS: open file(%s/%s)\n",
Chuck Levercc0dd2d2008-06-11 17:55:42 -0400125 filp->f_path.dentry->d_parent->d_name.name,
126 filp->f_path.dentry->d_name.name);
127
Chuck Leverc2459dc2010-02-01 14:17:14 -0500128 nfs_inc_stats(inode, NFSIOS_VFSOPEN);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700129 res = nfs_check_flags(filp->f_flags);
130 if (res)
131 return res;
132
Trond Myklebust46cb6502008-06-11 16:32:46 -0400133 res = nfs_open(inode, filp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700134 return res;
135}
136
137static int
138nfs_file_release(struct inode *inode, struct file *filp)
139{
Chuck Lever6da24bc2008-06-11 17:55:58 -0400140 struct dentry *dentry = filp->f_path.dentry;
141
142 dprintk("NFS: release(%s/%s)\n",
143 dentry->d_parent->d_name.name,
144 dentry->d_name.name);
145
Chuck Lever91d5b472006-03-20 13:44:14 -0500146 nfs_inc_stats(inode, NFSIOS_VFSRELEASE);
Trond Myklebust46cb6502008-06-11 16:32:46 -0400147 return nfs_release(inode, filp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700148}
149
Trond Myklebust980802e2005-06-13 11:14:01 -0400150/**
151 * nfs_revalidate_size - Revalidate the file size
152 * @inode - pointer to inode struct
153 * @file - pointer to struct file
154 *
155 * Revalidates the file length. This is basically a wrapper around
156 * nfs_revalidate_inode() that takes into account the fact that we may
157 * have cached writes (in which case we don't care about the server's
158 * idea of what the file length is), or O_DIRECT (in which case we
159 * shouldn't trust the cache).
160 */
161static int nfs_revalidate_file_size(struct inode *inode, struct file *filp)
162{
163 struct nfs_server *server = NFS_SERVER(inode);
164 struct nfs_inode *nfsi = NFS_I(inode);
165
Trond Myklebustd7cf8dd2010-04-16 16:42:46 -0400166 if (nfs_have_delegated_attributes(inode))
167 goto out_noreval;
168
Trond Myklebust980802e2005-06-13 11:14:01 -0400169 if (filp->f_flags & O_DIRECT)
170 goto force_reval;
Trond Myklebustd7cf8dd2010-04-16 16:42:46 -0400171 if (nfsi->cache_validity & NFS_INO_REVAL_PAGECACHE)
172 goto force_reval;
173 if (nfs_attribute_timeout(inode))
174 goto force_reval;
175out_noreval:
176 return 0;
Trond Myklebust980802e2005-06-13 11:14:01 -0400177force_reval:
178 return __nfs_revalidate_inode(server, inode);
179}
180
181static loff_t nfs_file_llseek(struct file *filp, loff_t offset, int origin)
182{
Andi Kleen9465efc2008-06-27 11:05:24 +0200183 loff_t loff;
Trond Myklebuste89e8962008-07-15 18:34:16 -0400184
Chuck Lever6da24bc2008-06-11 17:55:58 -0400185 dprintk("NFS: llseek file(%s/%s, %lld, %d)\n",
Chuck Leverb84e06c2008-06-11 17:55:34 -0400186 filp->f_path.dentry->d_parent->d_name.name,
187 filp->f_path.dentry->d_name.name,
188 offset, origin);
189
Josef Bacik06222e42011-07-18 13:21:38 -0400190 /*
191 * origin == SEEK_END || SEEK_DATA || SEEK_HOLE => we must revalidate
192 * the cached file length
193 */
194 if (origin != SEEK_SET || origin != SEEK_CUR) {
Trond Myklebust980802e2005-06-13 11:14:01 -0400195 struct inode *inode = filp->f_mapping->host;
Trond Myklebustd5e66342008-09-23 17:28:35 -0400196
Trond Myklebust980802e2005-06-13 11:14:01 -0400197 int retval = nfs_revalidate_file_size(inode, filp);
198 if (retval < 0)
199 return (loff_t)retval;
Trond Myklebustd5e66342008-09-23 17:28:35 -0400200
201 spin_lock(&inode->i_lock);
202 loff = generic_file_llseek_unlocked(filp, offset, origin);
203 spin_unlock(&inode->i_lock);
204 } else
205 loff = generic_file_llseek_unlocked(filp, offset, origin);
Andi Kleen9465efc2008-06-27 11:05:24 +0200206 return loff;
Trond Myklebust980802e2005-06-13 11:14:01 -0400207}
208
Linus Torvalds1da177e2005-04-16 15:20:36 -0700209/*
210 * Flush all dirty pages, and check for write errors.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700211 */
212static int
Miklos Szeredi75e1fcc2006-06-23 02:05:12 -0700213nfs_file_flush(struct file *file, fl_owner_t id)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700214{
Chuck Lever6da24bc2008-06-11 17:55:58 -0400215 struct dentry *dentry = file->f_path.dentry;
216 struct inode *inode = dentry->d_inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700217
Chuck Lever6da24bc2008-06-11 17:55:58 -0400218 dprintk("NFS: flush(%s/%s)\n",
219 dentry->d_parent->d_name.name,
220 dentry->d_name.name);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700221
Chuck Leverc2459dc2010-02-01 14:17:14 -0500222 nfs_inc_stats(inode, NFSIOS_VFSFLUSH);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700223 if ((file->f_mode & FMODE_WRITE) == 0)
224 return 0;
Trond Myklebust7b159fc2007-07-25 14:09:54 -0400225
Trond Myklebust7fe5c392009-03-19 15:35:50 -0400226 /* Flush writes to the server and return any errors */
Trond Myklebustaf7fa162010-07-31 14:29:06 -0400227 return vfs_fsync(file, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700228}
229
230static ssize_t
Badari Pulavarty027445c2006-09-30 23:28:46 -0700231nfs_file_read(struct kiocb *iocb, const struct iovec *iov,
232 unsigned long nr_segs, loff_t pos)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700233{
Josef "Jeff" Sipek01cce932006-12-08 02:36:40 -0800234 struct dentry * dentry = iocb->ki_filp->f_path.dentry;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700235 struct inode * inode = dentry->d_inode;
236 ssize_t result;
Badari Pulavarty027445c2006-09-30 23:28:46 -0700237 size_t count = iov_length(iov, nr_segs);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700238
Linus Torvalds1da177e2005-04-16 15:20:36 -0700239 if (iocb->ki_filp->f_flags & O_DIRECT)
Badari Pulavarty027445c2006-09-30 23:28:46 -0700240 return nfs_file_direct_read(iocb, iov, nr_segs, pos);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700241
Chuck Lever6da24bc2008-06-11 17:55:58 -0400242 dprintk("NFS: read(%s/%s, %lu@%lu)\n",
Linus Torvalds1da177e2005-04-16 15:20:36 -0700243 dentry->d_parent->d_name.name, dentry->d_name.name,
244 (unsigned long) count, (unsigned long) pos);
245
Trond Myklebust44b11872006-05-25 01:40:59 -0400246 result = nfs_revalidate_mapping(inode, iocb->ki_filp->f_mapping);
Chuck Lever4184dcf2010-02-01 14:17:23 -0500247 if (!result) {
Badari Pulavarty027445c2006-09-30 23:28:46 -0700248 result = generic_file_aio_read(iocb, iov, nr_segs, pos);
Chuck Lever4184dcf2010-02-01 14:17:23 -0500249 if (result > 0)
250 nfs_add_stats(inode, NFSIOS_NORMALREADBYTES, result);
251 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700252 return result;
253}
254
255static ssize_t
Jens Axboef0930ff2007-06-01 11:51:43 +0200256nfs_file_splice_read(struct file *filp, loff_t *ppos,
257 struct pipe_inode_info *pipe, size_t count,
258 unsigned int flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700259{
Josef "Jeff" Sipek01cce932006-12-08 02:36:40 -0800260 struct dentry *dentry = filp->f_path.dentry;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700261 struct inode *inode = dentry->d_inode;
262 ssize_t res;
263
Chuck Lever6da24bc2008-06-11 17:55:58 -0400264 dprintk("NFS: splice_read(%s/%s, %lu@%Lu)\n",
Linus Torvalds1da177e2005-04-16 15:20:36 -0700265 dentry->d_parent->d_name.name, dentry->d_name.name,
266 (unsigned long) count, (unsigned long long) *ppos);
267
Trond Myklebust44b11872006-05-25 01:40:59 -0400268 res = nfs_revalidate_mapping(inode, filp->f_mapping);
Chuck Leveraa2f1ef2010-02-01 14:17:32 -0500269 if (!res) {
Jens Axboef0930ff2007-06-01 11:51:43 +0200270 res = generic_file_splice_read(filp, ppos, pipe, count, flags);
Chuck Leveraa2f1ef2010-02-01 14:17:32 -0500271 if (res > 0)
272 nfs_add_stats(inode, NFSIOS_NORMALREADBYTES, res);
273 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700274 return res;
275}
276
277static int
278nfs_file_mmap(struct file * file, struct vm_area_struct * vma)
279{
Josef "Jeff" Sipek01cce932006-12-08 02:36:40 -0800280 struct dentry *dentry = file->f_path.dentry;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700281 struct inode *inode = dentry->d_inode;
282 int status;
283
Chuck Lever6da24bc2008-06-11 17:55:58 -0400284 dprintk("NFS: mmap(%s/%s)\n",
Linus Torvalds1da177e2005-04-16 15:20:36 -0700285 dentry->d_parent->d_name.name, dentry->d_name.name);
286
Trond Myklebuste1ebfd32009-03-11 14:37:54 -0400287 /* Note: generic_file_mmap() returns ENOSYS on nommu systems
288 * so we call that before revalidating the mapping
289 */
290 status = generic_file_mmap(file, vma);
Trond Myklebust94387fb2007-07-22 17:09:05 -0400291 if (!status) {
292 vma->vm_ops = &nfs_file_vm_ops;
Trond Myklebuste1ebfd32009-03-11 14:37:54 -0400293 status = nfs_revalidate_mapping(inode, file->f_mapping);
Trond Myklebust94387fb2007-07-22 17:09:05 -0400294 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700295 return status;
296}
297
298/*
299 * Flush any dirty pages for this process, and check for write errors.
300 * The return status from this call provides a reliable indication of
301 * whether any write errors occurred for this process.
Trond Myklebustaf7fa162010-07-31 14:29:06 -0400302 *
303 * Notice that it clears the NFS_CONTEXT_ERROR_WRITE before synching to
304 * disk, but it retrieves and clears ctx->error after synching, despite
305 * the two being set at the same time in nfs_context_set_write_error().
306 * This is because the former is used to notify the _next_ call to
Lucas De Marchi25985ed2011-03-30 22:57:33 -0300307 * nfs_file_write() that a write error occurred, and hence cause it to
Trond Myklebustaf7fa162010-07-31 14:29:06 -0400308 * fall back to doing a synchronous write.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700309 */
310static int
Josef Bacik02c24a82011-07-16 20:44:56 -0400311nfs_file_fsync(struct file *file, loff_t start, loff_t end, int datasync)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700312{
Christoph Hellwig7ea80852010-05-26 17:53:25 +0200313 struct dentry *dentry = file->f_path.dentry;
Trond Myklebustcd3758e2007-08-10 17:44:32 -0400314 struct nfs_open_context *ctx = nfs_file_open_context(file);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700315 struct inode *inode = dentry->d_inode;
Trond Myklebustaf7fa162010-07-31 14:29:06 -0400316 int have_error, status;
317 int ret = 0;
318
Chuck Lever6da24bc2008-06-11 17:55:58 -0400319 dprintk("NFS: fsync file(%s/%s) datasync %d\n",
Chuck Lever54917782008-05-27 16:29:07 -0400320 dentry->d_parent->d_name.name, dentry->d_name.name,
321 datasync);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700322
Josef Bacik02c24a82011-07-16 20:44:56 -0400323 ret = filemap_write_and_wait_range(inode->i_mapping, start, end);
324 if (ret)
325 return ret;
326 mutex_lock(&inode->i_mutex);
327
Chuck Lever91d5b472006-03-20 13:44:14 -0500328 nfs_inc_stats(inode, NFSIOS_VFSFSYNC);
Trond Myklebustaf7fa162010-07-31 14:29:06 -0400329 have_error = test_and_clear_bit(NFS_CONTEXT_ERROR_WRITE, &ctx->flags);
330 status = nfs_commit_inode(inode, FLUSH_SYNC);
331 have_error |= test_bit(NFS_CONTEXT_ERROR_WRITE, &ctx->flags);
332 if (have_error)
333 ret = xchg(&ctx->error, 0);
J. R. Okajima07020992010-08-11 13:10:16 -0400334 if (!ret && status < 0)
Trond Myklebustaf7fa162010-07-31 14:29:06 -0400335 ret = status;
Andy Adamson863a3c62011-03-23 13:27:54 +0000336 if (!ret && !datasync)
337 /* application has asked for meta-data sync */
Andy Adamsonef311532011-03-12 02:58:10 -0500338 ret = pnfs_layoutcommit_inode(inode, true);
Josef Bacik02c24a82011-07-16 20:44:56 -0400339 mutex_unlock(&inode->i_mutex);
Trond Myklebustaf7fa162010-07-31 14:29:06 -0400340 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700341}
342
343/*
Peter Staubach38c73042009-08-10 08:54:16 -0400344 * Decide whether a read/modify/write cycle may be more efficient
345 * then a modify/write/read cycle when writing to a page in the
346 * page cache.
347 *
348 * The modify/write/read cycle may occur if a page is read before
349 * being completely filled by the writer. In this situation, the
350 * page must be completely written to stable storage on the server
351 * before it can be refilled by reading in the page from the server.
352 * This can lead to expensive, small, FILE_SYNC mode writes being
353 * done.
354 *
355 * It may be more efficient to read the page first if the file is
356 * open for reading in addition to writing, the page is not marked
357 * as Uptodate, it is not dirty or waiting to be committed,
358 * indicating that it was previously allocated and then modified,
359 * that there were valid bytes of data in that range of the file,
360 * and that the new data won't completely replace the old data in
361 * that range of the file.
362 */
363static int nfs_want_read_modify_write(struct file *file, struct page *page,
364 loff_t pos, unsigned len)
365{
366 unsigned int pglen = nfs_page_length(page);
367 unsigned int offset = pos & (PAGE_CACHE_SIZE - 1);
368 unsigned int end = offset + len;
369
370 if ((file->f_mode & FMODE_READ) && /* open for read? */
371 !PageUptodate(page) && /* Uptodate? */
372 !PagePrivate(page) && /* i/o request already? */
373 pglen && /* valid bytes of file? */
374 (end < pglen || offset)) /* replace all valid bytes? */
375 return 1;
376 return 0;
377}
378
379/*
Nick Piggin4899f9c2007-10-16 01:25:16 -0700380 * This does the "real" work of the write. We must allocate and lock the
381 * page to be sent back to the generic routine, which then copies the
382 * data from user space.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700383 *
384 * If the writer ends up delaying the write, the writer needs to
385 * increment the page use counts until he is done with the page.
386 */
Nick Piggin4899f9c2007-10-16 01:25:16 -0700387static int nfs_write_begin(struct file *file, struct address_space *mapping,
388 loff_t pos, unsigned len, unsigned flags,
389 struct page **pagep, void **fsdata)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700390{
Nick Piggin4899f9c2007-10-16 01:25:16 -0700391 int ret;
Peter Staubach38c73042009-08-10 08:54:16 -0400392 pgoff_t index = pos >> PAGE_CACHE_SHIFT;
Nick Piggin4899f9c2007-10-16 01:25:16 -0700393 struct page *page;
Peter Staubach38c73042009-08-10 08:54:16 -0400394 int once_thru = 0;
Nick Piggin4899f9c2007-10-16 01:25:16 -0700395
Chuck Leverb7eaefa2008-06-11 17:55:50 -0400396 dfprintk(PAGECACHE, "NFS: write_begin(%s/%s(%ld), %u@%lld)\n",
397 file->f_path.dentry->d_parent->d_name.name,
398 file->f_path.dentry->d_name.name,
399 mapping->host->i_ino, len, (long long) pos);
400
Peter Staubach38c73042009-08-10 08:54:16 -0400401start:
Trond Myklebust72cb77f2009-03-11 14:10:30 -0400402 /*
403 * Prevent starvation issues if someone is doing a consistency
404 * sync-to-disk
405 */
406 ret = wait_on_bit(&NFS_I(mapping->host)->flags, NFS_INO_FLUSHING,
407 nfs_wait_bit_killable, TASK_KILLABLE);
408 if (ret)
409 return ret;
410
Nick Piggin54566b22009-01-04 12:00:53 -0800411 page = grab_cache_page_write_begin(mapping, index, flags);
Nick Piggin4899f9c2007-10-16 01:25:16 -0700412 if (!page)
413 return -ENOMEM;
414 *pagep = page;
415
416 ret = nfs_flush_incompatible(file, page);
417 if (ret) {
418 unlock_page(page);
419 page_cache_release(page);
Peter Staubach38c73042009-08-10 08:54:16 -0400420 } else if (!once_thru &&
421 nfs_want_read_modify_write(file, page, pos, len)) {
422 once_thru = 1;
423 ret = nfs_readpage(file, page);
424 page_cache_release(page);
425 if (!ret)
426 goto start;
Nick Piggin4899f9c2007-10-16 01:25:16 -0700427 }
428 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700429}
430
Nick Piggin4899f9c2007-10-16 01:25:16 -0700431static int nfs_write_end(struct file *file, struct address_space *mapping,
432 loff_t pos, unsigned len, unsigned copied,
433 struct page *page, void *fsdata)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700434{
Nick Piggin4899f9c2007-10-16 01:25:16 -0700435 unsigned offset = pos & (PAGE_CACHE_SIZE - 1);
436 int status;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700437
Chuck Leverb7eaefa2008-06-11 17:55:50 -0400438 dfprintk(PAGECACHE, "NFS: write_end(%s/%s(%ld), %u@%lld)\n",
439 file->f_path.dentry->d_parent->d_name.name,
440 file->f_path.dentry->d_name.name,
441 mapping->host->i_ino, len, (long long) pos);
442
Trond Myklebustefc91ed2008-06-10 18:31:00 -0400443 /*
444 * Zero any uninitialised parts of the page, and then mark the page
445 * as up to date if it turns out that we're extending the file.
446 */
447 if (!PageUptodate(page)) {
448 unsigned pglen = nfs_page_length(page);
449 unsigned end = offset + len;
450
451 if (pglen == 0) {
452 zero_user_segments(page, 0, offset,
453 end, PAGE_CACHE_SIZE);
454 SetPageUptodate(page);
455 } else if (end >= pglen) {
456 zero_user_segment(page, end, PAGE_CACHE_SIZE);
457 if (offset == 0)
458 SetPageUptodate(page);
459 } else
460 zero_user_segment(page, pglen, PAGE_CACHE_SIZE);
461 }
462
Nick Piggin4899f9c2007-10-16 01:25:16 -0700463 status = nfs_updatepage(file, page, offset, copied);
Nick Piggin4899f9c2007-10-16 01:25:16 -0700464
465 unlock_page(page);
466 page_cache_release(page);
467
Chuck Lever3d509e52007-12-20 14:55:04 -0500468 if (status < 0)
469 return status;
470 return copied;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700471}
472
David Howells6b9b3512009-04-03 16:42:41 +0100473/*
474 * Partially or wholly invalidate a page
475 * - Release the private state associated with a page if undergoing complete
476 * page invalidation
David Howells545db452009-04-03 16:42:44 +0100477 * - Called if either PG_private or PG_fscache is set on the page
David Howells6b9b3512009-04-03 16:42:41 +0100478 * - Caller holds page lock
479 */
NeilBrown2ff28e22006-03-26 01:37:18 -0800480static void nfs_invalidate_page(struct page *page, unsigned long offset)
Trond Myklebustcd52ed32006-03-20 13:44:04 -0500481{
Chuck Leverb7eaefa2008-06-11 17:55:50 -0400482 dfprintk(PAGECACHE, "NFS: invalidate_page(%p, %lu)\n", page, offset);
483
Trond Myklebust1c759502006-10-09 16:18:38 -0400484 if (offset != 0)
485 return;
Trond Myklebustd2ccddf2006-05-31 01:13:38 -0400486 /* Cancel any unstarted writes on this page */
Trond Myklebust1b3b4a12007-08-28 10:29:36 -0400487 nfs_wb_page_cancel(page->mapping->host, page);
David Howells545db452009-04-03 16:42:44 +0100488
489 nfs_fscache_invalidate_page(page, page->mapping->host);
Trond Myklebustcd52ed32006-03-20 13:44:04 -0500490}
491
David Howells6b9b3512009-04-03 16:42:41 +0100492/*
493 * Attempt to release the private state associated with a page
David Howells545db452009-04-03 16:42:44 +0100494 * - Called if either PG_private or PG_fscache is set on the page
David Howells6b9b3512009-04-03 16:42:41 +0100495 * - Caller holds page lock
496 * - Return true (may release page) or false (may not)
497 */
Trond Myklebustcd52ed32006-03-20 13:44:04 -0500498static int nfs_release_page(struct page *page, gfp_t gfp)
499{
Trond Myklebustb608b282010-07-30 15:31:54 -0400500 struct address_space *mapping = page->mapping;
501
Chuck Leverb7eaefa2008-06-11 17:55:50 -0400502 dfprintk(PAGECACHE, "NFS: release_page(%p)\n", page);
503
Trond Myklebustd812e572010-03-19 13:55:17 -0400504 /* Only do I/O if gfp is a superset of GFP_KERNEL */
Trond Myklebustb608b282010-07-30 15:31:54 -0400505 if (mapping && (gfp & GFP_KERNEL) == GFP_KERNEL) {
506 int how = FLUSH_SYNC;
507
508 /* Don't let kswapd deadlock waiting for OOM RPC calls */
509 if (current_is_kswapd())
510 how = 0;
511 nfs_commit_inode(mapping->host, how);
512 }
Trond Myklebuste3db7692007-01-10 23:15:39 -0800513 /* If PagePrivate() is set, then the page is not freeable */
David Howells545db452009-04-03 16:42:44 +0100514 if (PagePrivate(page))
515 return 0;
516 return nfs_fscache_release_page(page, gfp);
Trond Myklebuste3db7692007-01-10 23:15:39 -0800517}
518
David Howells6b9b3512009-04-03 16:42:41 +0100519/*
520 * Attempt to clear the private state associated with a page when an error
521 * occurs that requires the cached contents of an inode to be written back or
522 * destroyed
David Howells545db452009-04-03 16:42:44 +0100523 * - Called if either PG_private or fscache is set on the page
David Howells6b9b3512009-04-03 16:42:41 +0100524 * - Caller holds page lock
525 * - Return 0 if successful, -error otherwise
526 */
Trond Myklebuste3db7692007-01-10 23:15:39 -0800527static int nfs_launder_page(struct page *page)
528{
Chuck Leverb7eaefa2008-06-11 17:55:50 -0400529 struct inode *inode = page->mapping->host;
David Howells545db452009-04-03 16:42:44 +0100530 struct nfs_inode *nfsi = NFS_I(inode);
Chuck Leverb7eaefa2008-06-11 17:55:50 -0400531
532 dfprintk(PAGECACHE, "NFS: launder_page(%ld, %llu)\n",
533 inode->i_ino, (long long)page_offset(page));
534
David Howells545db452009-04-03 16:42:44 +0100535 nfs_fscache_wait_on_page_write(nfsi, page);
Chuck Leverb7eaefa2008-06-11 17:55:50 -0400536 return nfs_wb_page(inode, page);
Trond Myklebustcd52ed32006-03-20 13:44:04 -0500537}
538
Christoph Hellwigf5e54d62006-06-28 04:26:44 -0700539const struct address_space_operations nfs_file_aops = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700540 .readpage = nfs_readpage,
541 .readpages = nfs_readpages,
Trond Myklebust9cccef92007-07-22 17:09:05 -0400542 .set_page_dirty = __set_page_dirty_nobuffers,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700543 .writepage = nfs_writepage,
544 .writepages = nfs_writepages,
Nick Piggin4899f9c2007-10-16 01:25:16 -0700545 .write_begin = nfs_write_begin,
546 .write_end = nfs_write_end,
Trond Myklebustcd52ed32006-03-20 13:44:04 -0500547 .invalidatepage = nfs_invalidate_page,
548 .releasepage = nfs_release_page,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700549 .direct_IO = nfs_direct_IO,
Trond Myklebust074cc1d2009-08-10 08:54:13 -0400550 .migratepage = nfs_migrate_page,
Trond Myklebuste3db7692007-01-10 23:15:39 -0800551 .launder_page = nfs_launder_page,
Andi Kleenf590f332009-09-16 11:50:17 +0200552 .error_remove_page = generic_error_remove_page,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700553};
554
David Howells6b9b3512009-04-03 16:42:41 +0100555/*
556 * Notification that a PTE pointing to an NFS page is about to be made
557 * writable, implying that someone is about to modify the page through a
558 * shared-writable mapping
559 */
Nick Pigginc2ec1752009-03-31 15:23:21 -0700560static int nfs_vm_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf)
Trond Myklebust94387fb2007-07-22 17:09:05 -0400561{
Nick Pigginc2ec1752009-03-31 15:23:21 -0700562 struct page *page = vmf->page;
Trond Myklebust94387fb2007-07-22 17:09:05 -0400563 struct file *filp = vma->vm_file;
Chuck Leverb7eaefa2008-06-11 17:55:50 -0400564 struct dentry *dentry = filp->f_path.dentry;
Trond Myklebust94387fb2007-07-22 17:09:05 -0400565 unsigned pagelen;
Trond Myklebustbc4866b2010-10-04 17:59:08 -0400566 int ret = VM_FAULT_NOPAGE;
Nick Piggin4899f9c2007-10-16 01:25:16 -0700567 struct address_space *mapping;
Trond Myklebust94387fb2007-07-22 17:09:05 -0400568
Chuck Leverb7eaefa2008-06-11 17:55:50 -0400569 dfprintk(PAGECACHE, "NFS: vm_page_mkwrite(%s/%s(%ld), offset %lld)\n",
570 dentry->d_parent->d_name.name, dentry->d_name.name,
571 filp->f_mapping->host->i_ino,
572 (long long)page_offset(page));
573
David Howells545db452009-04-03 16:42:44 +0100574 /* make sure the cache has finished storing the page */
575 nfs_fscache_wait_on_page_write(NFS_I(dentry->d_inode), page);
576
Trond Myklebust94387fb2007-07-22 17:09:05 -0400577 lock_page(page);
Nick Piggin4899f9c2007-10-16 01:25:16 -0700578 mapping = page->mapping;
Chuck Leverb7eaefa2008-06-11 17:55:50 -0400579 if (mapping != dentry->d_inode->i_mapping)
Trond Myklebust8b1f9ee2008-01-22 17:13:06 -0500580 goto out_unlock;
581
Trond Myklebust94387fb2007-07-22 17:09:05 -0400582 pagelen = nfs_page_length(page);
Trond Myklebust8b1f9ee2008-01-22 17:13:06 -0500583 if (pagelen == 0)
584 goto out_unlock;
585
Trond Myklebustbc4866b2010-10-04 17:59:08 -0400586 ret = VM_FAULT_LOCKED;
587 if (nfs_flush_incompatible(filp, page) == 0 &&
588 nfs_updatepage(filp, page, 0, pagelen) == 0)
589 goto out;
Trond Myklebust8b1f9ee2008-01-22 17:13:06 -0500590
Trond Myklebustbc4866b2010-10-04 17:59:08 -0400591 ret = VM_FAULT_SIGBUS;
Trond Myklebust8b1f9ee2008-01-22 17:13:06 -0500592out_unlock:
Trond Myklebust94387fb2007-07-22 17:09:05 -0400593 unlock_page(page);
Trond Myklebustbc4866b2010-10-04 17:59:08 -0400594out:
595 return ret;
Trond Myklebust94387fb2007-07-22 17:09:05 -0400596}
597
Alexey Dobriyanf0f37e22009-09-27 22:29:37 +0400598static const struct vm_operations_struct nfs_file_vm_ops = {
Trond Myklebust94387fb2007-07-22 17:09:05 -0400599 .fault = filemap_fault,
600 .page_mkwrite = nfs_vm_page_mkwrite,
601};
602
Trond Myklebust7b159fc2007-07-25 14:09:54 -0400603static int nfs_need_sync_write(struct file *filp, struct inode *inode)
604{
605 struct nfs_open_context *ctx;
606
Christoph Hellwig6b2f3d12009-10-27 11:05:28 +0100607 if (IS_SYNC(inode) || (filp->f_flags & O_DSYNC))
Trond Myklebust7b159fc2007-07-25 14:09:54 -0400608 return 1;
Trond Myklebustcd3758e2007-08-10 17:44:32 -0400609 ctx = nfs_file_open_context(filp);
Trond Myklebust7b159fc2007-07-25 14:09:54 -0400610 if (test_bit(NFS_CONTEXT_ERROR_WRITE, &ctx->flags))
611 return 1;
612 return 0;
613}
614
Badari Pulavarty027445c2006-09-30 23:28:46 -0700615static ssize_t nfs_file_write(struct kiocb *iocb, const struct iovec *iov,
616 unsigned long nr_segs, loff_t pos)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700617{
Josef "Jeff" Sipek01cce932006-12-08 02:36:40 -0800618 struct dentry * dentry = iocb->ki_filp->f_path.dentry;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700619 struct inode * inode = dentry->d_inode;
Chuck Lever7e381172010-02-01 14:17:41 -0500620 unsigned long written = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700621 ssize_t result;
Badari Pulavarty027445c2006-09-30 23:28:46 -0700622 size_t count = iov_length(iov, nr_segs);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700623
Linus Torvalds1da177e2005-04-16 15:20:36 -0700624 if (iocb->ki_filp->f_flags & O_DIRECT)
Badari Pulavarty027445c2006-09-30 23:28:46 -0700625 return nfs_file_direct_write(iocb, iov, nr_segs, pos);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700626
Chuck Lever6da24bc2008-06-11 17:55:58 -0400627 dprintk("NFS: write(%s/%s, %lu@%Ld)\n",
Linus Torvalds1da177e2005-04-16 15:20:36 -0700628 dentry->d_parent->d_name.name, dentry->d_name.name,
Chuck Lever6da24bc2008-06-11 17:55:58 -0400629 (unsigned long) count, (long long) pos);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700630
631 result = -EBUSY;
632 if (IS_SWAPFILE(inode))
633 goto out_swapfile;
Trond Myklebust7d52e862005-06-22 17:16:30 +0000634 /*
635 * O_APPEND implies that we must revalidate the file length.
636 */
637 if (iocb->ki_filp->f_flags & O_APPEND) {
638 result = nfs_revalidate_file_size(inode, iocb->ki_filp);
639 if (result)
640 goto out;
Trond Myklebustfe51beec2005-06-22 17:16:30 +0000641 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700642
643 result = count;
644 if (!count)
645 goto out;
646
Badari Pulavarty027445c2006-09-30 23:28:46 -0700647 result = generic_file_aio_write(iocb, iov, nr_segs, pos);
Chuck Lever7e381172010-02-01 14:17:41 -0500648 if (result > 0)
649 written = result;
650
Christoph Hellwig6b2f3d12009-10-27 11:05:28 +0100651 /* Return error values for O_DSYNC and IS_SYNC() */
Trond Myklebust7b159fc2007-07-25 14:09:54 -0400652 if (result >= 0 && nfs_need_sync_write(iocb->ki_filp, inode)) {
Trond Myklebustaf7fa162010-07-31 14:29:06 -0400653 int err = vfs_fsync(iocb->ki_filp, 0);
Trond Myklebust200baa22006-12-05 00:35:40 -0500654 if (err < 0)
655 result = err;
656 }
Chuck Lever7e381172010-02-01 14:17:41 -0500657 if (result > 0)
658 nfs_add_stats(inode, NFSIOS_NORMALWRITTENBYTES, written);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700659out:
660 return result;
661
662out_swapfile:
663 printk(KERN_INFO "NFS: attempt to write to active swap file!\n");
664 goto out;
665}
666
Suresh Jayaramanbf40d342009-06-17 18:02:09 -0700667static ssize_t nfs_file_splice_write(struct pipe_inode_info *pipe,
668 struct file *filp, loff_t *ppos,
669 size_t count, unsigned int flags)
670{
671 struct dentry *dentry = filp->f_path.dentry;
672 struct inode *inode = dentry->d_inode;
Chuck Lever7e381172010-02-01 14:17:41 -0500673 unsigned long written = 0;
Suresh Jayaramanbf40d342009-06-17 18:02:09 -0700674 ssize_t ret;
675
676 dprintk("NFS splice_write(%s/%s, %lu@%llu)\n",
677 dentry->d_parent->d_name.name, dentry->d_name.name,
678 (unsigned long) count, (unsigned long long) *ppos);
679
680 /*
681 * The combination of splice and an O_APPEND destination is disallowed.
682 */
683
Suresh Jayaramanbf40d342009-06-17 18:02:09 -0700684 ret = generic_file_splice_write(pipe, filp, ppos, count, flags);
Chuck Lever7e381172010-02-01 14:17:41 -0500685 if (ret > 0)
686 written = ret;
687
Suresh Jayaramanbf40d342009-06-17 18:02:09 -0700688 if (ret >= 0 && nfs_need_sync_write(filp, inode)) {
Trond Myklebustaf7fa162010-07-31 14:29:06 -0400689 int err = vfs_fsync(filp, 0);
Suresh Jayaramanbf40d342009-06-17 18:02:09 -0700690 if (err < 0)
691 ret = err;
692 }
Chuck Lever7e381172010-02-01 14:17:41 -0500693 if (ret > 0)
694 nfs_add_stats(inode, NFSIOS_NORMALWRITTENBYTES, written);
Suresh Jayaramanbf40d342009-06-17 18:02:09 -0700695 return ret;
696}
697
Suresh Jayaraman5eebde22010-09-23 08:55:58 -0400698static int
699do_getlk(struct file *filp, int cmd, struct file_lock *fl, int is_local)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700700{
701 struct inode *inode = filp->f_mapping->host;
702 int status = 0;
Sergey Vlasov21ac19d2010-11-28 21:04:05 +0000703 unsigned int saved_type = fl->fl_type;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700704
Trond Myklebust039c4d72005-10-18 14:20:16 -0700705 /* Try local locking first */
J. Bruce Fields6d34ac12007-05-11 16:09:32 -0400706 posix_test_lock(filp, fl);
707 if (fl->fl_type != F_UNLCK) {
708 /* found a conflict */
Trond Myklebust039c4d72005-10-18 14:20:16 -0700709 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700710 }
Sergey Vlasov21ac19d2010-11-28 21:04:05 +0000711 fl->fl_type = saved_type;
Trond Myklebust039c4d72005-10-18 14:20:16 -0700712
713 if (nfs_have_delegation(inode, FMODE_READ))
714 goto out_noconflict;
715
Suresh Jayaraman5eebde22010-09-23 08:55:58 -0400716 if (is_local)
Trond Myklebust039c4d72005-10-18 14:20:16 -0700717 goto out_noconflict;
718
719 status = NFS_PROTO(inode)->lock(filp, cmd, fl);
720out:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700721 return status;
Trond Myklebust039c4d72005-10-18 14:20:16 -0700722out_noconflict:
723 fl->fl_type = F_UNLCK;
724 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700725}
726
727static int do_vfs_lock(struct file *file, struct file_lock *fl)
728{
729 int res = 0;
730 switch (fl->fl_flags & (FL_POSIX|FL_FLOCK)) {
731 case FL_POSIX:
732 res = posix_lock_file_wait(file, fl);
733 break;
734 case FL_FLOCK:
735 res = flock_lock_file_wait(file, fl);
736 break;
737 default:
738 BUG();
739 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700740 return res;
741}
742
Suresh Jayaraman5eebde22010-09-23 08:55:58 -0400743static int
744do_unlk(struct file *filp, int cmd, struct file_lock *fl, int is_local)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700745{
746 struct inode *inode = filp->f_mapping->host;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700747 int status;
748
Linus Torvalds1da177e2005-04-16 15:20:36 -0700749 /*
750 * Flush all pending writes before doing anything
751 * with locks..
752 */
Trond Myklebust29884df2005-12-13 16:13:54 -0500753 nfs_sync_mapping(filp->f_mapping);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700754
755 /* NOTE: special case
756 * If we're signalled while cleaning up locks on process exit, we
757 * still need to complete the unlock.
758 */
Suresh Jayaraman5eebde22010-09-23 08:55:58 -0400759 /*
760 * Use local locking if mounted with "-onolock" or with appropriate
761 * "-olocal_lock="
762 */
763 if (!is_local)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700764 status = NFS_PROTO(inode)->lock(filp, cmd, fl);
765 else
766 status = do_vfs_lock(filp, fl);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700767 return status;
768}
769
Suresh Jayaraman5eebde22010-09-23 08:55:58 -0400770static int
Ricardo Labiaga6b967242010-10-12 16:30:05 -0700771is_time_granular(struct timespec *ts) {
772 return ((ts->tv_sec == 0) && (ts->tv_nsec <= 1000));
773}
774
775static int
Suresh Jayaraman5eebde22010-09-23 08:55:58 -0400776do_setlk(struct file *filp, int cmd, struct file_lock *fl, int is_local)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700777{
778 struct inode *inode = filp->f_mapping->host;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700779 int status;
780
Linus Torvalds1da177e2005-04-16 15:20:36 -0700781 /*
782 * Flush all pending writes before doing anything
783 * with locks..
784 */
Trond Myklebust29884df2005-12-13 16:13:54 -0500785 status = nfs_sync_mapping(filp->f_mapping);
786 if (status != 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700787 goto out;
788
Suresh Jayaraman5eebde22010-09-23 08:55:58 -0400789 /*
790 * Use local locking if mounted with "-onolock" or with appropriate
791 * "-olocal_lock="
792 */
793 if (!is_local)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700794 status = NFS_PROTO(inode)->lock(filp, cmd, fl);
Trond Myklebustc4d7c402008-04-01 20:26:52 -0400795 else
Linus Torvalds1da177e2005-04-16 15:20:36 -0700796 status = do_vfs_lock(filp, fl);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700797 if (status < 0)
798 goto out;
Ricardo Labiaga6b967242010-10-12 16:30:05 -0700799
Linus Torvalds1da177e2005-04-16 15:20:36 -0700800 /*
Ricardo Labiaga6b967242010-10-12 16:30:05 -0700801 * Revalidate the cache if the server has time stamps granular
802 * enough to detect subsecond changes. Otherwise, clear the
803 * cache to prevent missing any changes.
804 *
Linus Torvalds1da177e2005-04-16 15:20:36 -0700805 * This makes locking act as a cache coherency point.
806 */
Trond Myklebust29884df2005-12-13 16:13:54 -0500807 nfs_sync_mapping(filp->f_mapping);
Ricardo Labiaga6b967242010-10-12 16:30:05 -0700808 if (!nfs_have_delegation(inode, FMODE_READ)) {
809 if (is_time_granular(&NFS_SERVER(inode)->time_delta))
810 __nfs_revalidate_inode(NFS_SERVER(inode), inode);
811 else
812 nfs_zap_caches(inode);
813 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700814out:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700815 return status;
816}
817
818/*
819 * Lock a (portion of) a file
820 */
821static int nfs_lock(struct file *filp, int cmd, struct file_lock *fl)
822{
Chuck Lever6da24bc2008-06-11 17:55:58 -0400823 struct inode *inode = filp->f_mapping->host;
Trond Myklebust21162712008-05-20 19:34:39 -0400824 int ret = -ENOLCK;
Suresh Jayaraman5eebde22010-09-23 08:55:58 -0400825 int is_local = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700826
Chuck Lever6da24bc2008-06-11 17:55:58 -0400827 dprintk("NFS: lock(%s/%s, t=%x, fl=%x, r=%lld:%lld)\n",
828 filp->f_path.dentry->d_parent->d_name.name,
829 filp->f_path.dentry->d_name.name,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700830 fl->fl_type, fl->fl_flags,
831 (long long)fl->fl_start, (long long)fl->fl_end);
Chuck Lever6da24bc2008-06-11 17:55:58 -0400832
Chuck Lever91d5b472006-03-20 13:44:14 -0500833 nfs_inc_stats(inode, NFSIOS_VFSLOCK);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700834
835 /* No mandatory locks over NFS */
Pavel Emelyanovdfad9442007-10-01 14:41:15 -0700836 if (__mandatory_lock(inode) && fl->fl_type != F_UNLCK)
Trond Myklebust21162712008-05-20 19:34:39 -0400837 goto out_err;
838
Suresh Jayaraman5eebde22010-09-23 08:55:58 -0400839 if (NFS_SERVER(inode)->flags & NFS_MOUNT_LOCAL_FCNTL)
840 is_local = 1;
841
Trond Myklebust21162712008-05-20 19:34:39 -0400842 if (NFS_PROTO(inode)->lock_check_bounds != NULL) {
843 ret = NFS_PROTO(inode)->lock_check_bounds(fl);
844 if (ret < 0)
845 goto out_err;
846 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700847
848 if (IS_GETLK(cmd))
Suresh Jayaraman5eebde22010-09-23 08:55:58 -0400849 ret = do_getlk(filp, cmd, fl, is_local);
Trond Myklebust21162712008-05-20 19:34:39 -0400850 else if (fl->fl_type == F_UNLCK)
Suresh Jayaraman5eebde22010-09-23 08:55:58 -0400851 ret = do_unlk(filp, cmd, fl, is_local);
Trond Myklebust21162712008-05-20 19:34:39 -0400852 else
Suresh Jayaraman5eebde22010-09-23 08:55:58 -0400853 ret = do_setlk(filp, cmd, fl, is_local);
Trond Myklebust21162712008-05-20 19:34:39 -0400854out_err:
855 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700856}
857
858/*
859 * Lock a (portion of) a file
860 */
861static int nfs_flock(struct file *filp, int cmd, struct file_lock *fl)
862{
Suresh Jayaraman5eebde22010-09-23 08:55:58 -0400863 struct inode *inode = filp->f_mapping->host;
864 int is_local = 0;
865
Chuck Lever6da24bc2008-06-11 17:55:58 -0400866 dprintk("NFS: flock(%s/%s, t=%x, fl=%x)\n",
867 filp->f_path.dentry->d_parent->d_name.name,
868 filp->f_path.dentry->d_name.name,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700869 fl->fl_type, fl->fl_flags);
870
Linus Torvalds1da177e2005-04-16 15:20:36 -0700871 if (!(fl->fl_flags & FL_FLOCK))
872 return -ENOLCK;
873
Suresh Jayaraman5eebde22010-09-23 08:55:58 -0400874 if (NFS_SERVER(inode)->flags & NFS_MOUNT_LOCAL_FLOCK)
875 is_local = 1;
876
Linus Torvalds1da177e2005-04-16 15:20:36 -0700877 /* We're simulating flock() locks using posix locks on the server */
878 fl->fl_owner = (fl_owner_t)filp;
879 fl->fl_start = 0;
880 fl->fl_end = OFFSET_MAX;
881
882 if (fl->fl_type == F_UNLCK)
Suresh Jayaraman5eebde22010-09-23 08:55:58 -0400883 return do_unlk(filp, cmd, fl, is_local);
884 return do_setlk(filp, cmd, fl, is_local);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700885}
J. Bruce Fields370f6592007-06-08 15:23:34 -0400886
Chuck Lever6da24bc2008-06-11 17:55:58 -0400887/*
888 * There is no protocol support for leases, so we have no way to implement
889 * them correctly in the face of opens by other clients.
890 */
J. Bruce Fields370f6592007-06-08 15:23:34 -0400891static int nfs_setlease(struct file *file, long arg, struct file_lock **fl)
892{
Chuck Lever6da24bc2008-06-11 17:55:58 -0400893 dprintk("NFS: setlease(%s/%s, arg=%ld)\n",
894 file->f_path.dentry->d_parent->d_name.name,
895 file->f_path.dentry->d_name.name, arg);
J. Bruce Fields370f6592007-06-08 15:23:34 -0400896 return -EINVAL;
897}