blob: ef57a5ae5904663d4ab20c974e920e711f921436 [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>
Alexey Dobriyane8edc6e2007-05-21 01:22:52 +040030#include <linux/aio.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"
Linus Torvalds1da177e2005-04-16 15:20:36 -070038
39#define NFSDBG_FACILITY NFSDBG_FILE
40
41static int nfs_file_open(struct inode *, struct file *);
42static int nfs_file_release(struct inode *, struct file *);
Trond Myklebust980802e2005-06-13 11:14:01 -040043static loff_t nfs_file_llseek(struct file *file, loff_t offset, int origin);
Linus Torvalds1da177e2005-04-16 15:20:36 -070044static int nfs_file_mmap(struct file *, struct vm_area_struct *);
Jens Axboef0930ff2007-06-01 11:51:43 +020045static ssize_t nfs_file_splice_read(struct file *filp, loff_t *ppos,
46 struct pipe_inode_info *pipe,
47 size_t count, unsigned int flags);
Badari Pulavarty027445c2006-09-30 23:28:46 -070048static ssize_t nfs_file_read(struct kiocb *, const struct iovec *iov,
49 unsigned long nr_segs, loff_t pos);
50static ssize_t nfs_file_write(struct kiocb *, const struct iovec *iov,
51 unsigned long nr_segs, loff_t pos);
Miklos Szeredi75e1fcc2006-06-23 02:05:12 -070052static int nfs_file_flush(struct file *, fl_owner_t id);
Linus Torvalds1da177e2005-04-16 15:20:36 -070053static int nfs_fsync(struct file *, struct dentry *dentry, int datasync);
54static int nfs_check_flags(int flags);
55static int nfs_lock(struct file *filp, int cmd, struct file_lock *fl);
56static int nfs_flock(struct file *filp, int cmd, struct file_lock *fl);
J. Bruce Fields370f6592007-06-08 15:23:34 -040057static int nfs_setlease(struct file *file, long arg, struct file_lock **fl);
Linus Torvalds1da177e2005-04-16 15:20:36 -070058
Trond Myklebust94387fb2007-07-22 17:09:05 -040059static struct vm_operations_struct nfs_file_vm_ops;
60
Arjan van de Ven4b6f5d22006-03-28 01:56:42 -080061const struct file_operations nfs_file_operations = {
Trond Myklebust980802e2005-06-13 11:14:01 -040062 .llseek = nfs_file_llseek,
Linus Torvalds1da177e2005-04-16 15:20:36 -070063 .read = do_sync_read,
64 .write = do_sync_write,
Badari Pulavarty027445c2006-09-30 23:28:46 -070065 .aio_read = nfs_file_read,
66 .aio_write = nfs_file_write,
Linus Torvalds1da177e2005-04-16 15:20:36 -070067 .mmap = nfs_file_mmap,
68 .open = nfs_file_open,
69 .flush = nfs_file_flush,
70 .release = nfs_file_release,
71 .fsync = nfs_fsync,
72 .lock = nfs_lock,
73 .flock = nfs_flock,
Jens Axboef0930ff2007-06-01 11:51:43 +020074 .splice_read = nfs_file_splice_read,
Linus Torvalds1da177e2005-04-16 15:20:36 -070075 .check_flags = nfs_check_flags,
J. Bruce Fields370f6592007-06-08 15:23:34 -040076 .setlease = nfs_setlease,
Linus Torvalds1da177e2005-04-16 15:20:36 -070077};
78
Arjan van de Ven92e1d5b2007-02-12 00:55:39 -080079const struct inode_operations nfs_file_inode_operations = {
Linus Torvalds1da177e2005-04-16 15:20:36 -070080 .permission = nfs_permission,
81 .getattr = nfs_getattr,
82 .setattr = nfs_setattr,
83};
84
Andreas Gruenbacherb7fa0552005-06-22 17:16:27 +000085#ifdef CONFIG_NFS_V3
Arjan van de Ven92e1d5b2007-02-12 00:55:39 -080086const struct inode_operations nfs3_file_inode_operations = {
Andreas Gruenbacherb7fa0552005-06-22 17:16:27 +000087 .permission = nfs_permission,
88 .getattr = nfs_getattr,
89 .setattr = nfs_setattr,
90 .listxattr = nfs3_listxattr,
91 .getxattr = nfs3_getxattr,
92 .setxattr = nfs3_setxattr,
93 .removexattr = nfs3_removexattr,
94};
95#endif /* CONFIG_NFS_v3 */
96
Linus Torvalds1da177e2005-04-16 15:20:36 -070097/* Hack for future NFS swap support */
98#ifndef IS_SWAPFILE
99# define IS_SWAPFILE(inode) (0)
100#endif
101
102static int nfs_check_flags(int flags)
103{
104 if ((flags & (O_APPEND | O_DIRECT)) == (O_APPEND | O_DIRECT))
105 return -EINVAL;
106
107 return 0;
108}
109
110/*
111 * Open file
112 */
113static int
114nfs_file_open(struct inode *inode, struct file *filp)
115{
Linus Torvalds1da177e2005-04-16 15:20:36 -0700116 int res;
117
118 res = nfs_check_flags(filp->f_flags);
119 if (res)
120 return res;
121
Chuck Lever91d5b472006-03-20 13:44:14 -0500122 nfs_inc_stats(inode, NFSIOS_VFSOPEN);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700123 lock_kernel();
David Howells1f163412006-08-22 20:06:11 -0400124 res = NFS_PROTO(inode)->file_open(inode, filp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700125 unlock_kernel();
126 return res;
127}
128
129static int
130nfs_file_release(struct inode *inode, struct file *filp)
131{
132 /* Ensure that dirty pages are flushed out with the right creds */
133 if (filp->f_mode & FMODE_WRITE)
Trond Myklebusta49c3c72007-10-18 18:03:27 -0400134 nfs_wb_all(filp->f_path.dentry->d_inode);
Chuck Lever91d5b472006-03-20 13:44:14 -0500135 nfs_inc_stats(inode, NFSIOS_VFSRELEASE);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700136 return NFS_PROTO(inode)->file_release(inode, filp);
137}
138
Trond Myklebust980802e2005-06-13 11:14:01 -0400139/**
140 * nfs_revalidate_size - Revalidate the file size
141 * @inode - pointer to inode struct
142 * @file - pointer to struct file
143 *
144 * Revalidates the file length. This is basically a wrapper around
145 * nfs_revalidate_inode() that takes into account the fact that we may
146 * have cached writes (in which case we don't care about the server's
147 * idea of what the file length is), or O_DIRECT (in which case we
148 * shouldn't trust the cache).
149 */
150static int nfs_revalidate_file_size(struct inode *inode, struct file *filp)
151{
152 struct nfs_server *server = NFS_SERVER(inode);
153 struct nfs_inode *nfsi = NFS_I(inode);
154
155 if (server->flags & NFS_MOUNT_NOAC)
156 goto force_reval;
157 if (filp->f_flags & O_DIRECT)
158 goto force_reval;
159 if (nfsi->npages != 0)
160 return 0;
Chuck Lever55296802005-08-18 11:24:09 -0700161 if (!(nfsi->cache_validity & NFS_INO_REVAL_PAGECACHE) && !nfs_attribute_timeout(inode))
Trond Myklebustfe51bee2005-06-22 17:16:30 +0000162 return 0;
Trond Myklebust980802e2005-06-13 11:14:01 -0400163force_reval:
164 return __nfs_revalidate_inode(server, inode);
165}
166
167static loff_t nfs_file_llseek(struct file *filp, loff_t offset, int origin)
168{
169 /* origin == SEEK_END => we must revalidate the cached file length */
Josef 'Jeff' Sipekaec5e172006-09-16 21:09:32 -0400170 if (origin == SEEK_END) {
Trond Myklebust980802e2005-06-13 11:14:01 -0400171 struct inode *inode = filp->f_mapping->host;
172 int retval = nfs_revalidate_file_size(inode, filp);
173 if (retval < 0)
174 return (loff_t)retval;
175 }
176 return remote_llseek(filp, offset, origin);
177}
178
Linus Torvalds1da177e2005-04-16 15:20:36 -0700179/*
Trond Myklebust7b159fc2007-07-25 14:09:54 -0400180 * Helper for nfs_file_flush() and nfs_fsync()
181 *
182 * Notice that it clears the NFS_CONTEXT_ERROR_WRITE before synching to
183 * disk, but it retrieves and clears ctx->error after synching, despite
184 * the two being set at the same time in nfs_context_set_write_error().
185 * This is because the former is used to notify the _next_ call to
186 * nfs_file_write() that a write error occured, and hence cause it to
187 * fall back to doing a synchronous write.
188 */
189static int nfs_do_fsync(struct nfs_open_context *ctx, struct inode *inode)
190{
191 int have_error, status;
192 int ret = 0;
193
194 have_error = test_and_clear_bit(NFS_CONTEXT_ERROR_WRITE, &ctx->flags);
195 status = nfs_wb_all(inode);
196 have_error |= test_bit(NFS_CONTEXT_ERROR_WRITE, &ctx->flags);
197 if (have_error)
198 ret = xchg(&ctx->error, 0);
199 if (!ret)
200 ret = status;
201 return ret;
202}
203
204/*
Linus Torvalds1da177e2005-04-16 15:20:36 -0700205 * Flush all dirty pages, and check for write errors.
206 *
207 */
208static int
Miklos Szeredi75e1fcc2006-06-23 02:05:12 -0700209nfs_file_flush(struct file *file, fl_owner_t id)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700210{
Trond Myklebustcd3758e2007-08-10 17:44:32 -0400211 struct nfs_open_context *ctx = nfs_file_open_context(file);
Josef "Jeff" Sipek01cce932006-12-08 02:36:40 -0800212 struct inode *inode = file->f_path.dentry->d_inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700213 int status;
214
215 dfprintk(VFS, "nfs: flush(%s/%ld)\n", inode->i_sb->s_id, inode->i_ino);
216
217 if ((file->f_mode & FMODE_WRITE) == 0)
218 return 0;
Chuck Lever91d5b472006-03-20 13:44:14 -0500219 nfs_inc_stats(inode, NFSIOS_VFSFLUSH);
Trond Myklebust7b159fc2007-07-25 14:09:54 -0400220
Linus Torvalds1da177e2005-04-16 15:20:36 -0700221 /* Ensure that data+attribute caches are up to date after close() */
Trond Myklebust7b159fc2007-07-25 14:09:54 -0400222 status = nfs_do_fsync(ctx, inode);
223 if (!status)
224 nfs_revalidate_inode(NFS_SERVER(inode), inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700225 return status;
226}
227
228static ssize_t
Badari Pulavarty027445c2006-09-30 23:28:46 -0700229nfs_file_read(struct kiocb *iocb, const struct iovec *iov,
230 unsigned long nr_segs, loff_t pos)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700231{
Josef "Jeff" Sipek01cce932006-12-08 02:36:40 -0800232 struct dentry * dentry = iocb->ki_filp->f_path.dentry;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700233 struct inode * inode = dentry->d_inode;
234 ssize_t result;
Badari Pulavarty027445c2006-09-30 23:28:46 -0700235 size_t count = iov_length(iov, nr_segs);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700236
237#ifdef CONFIG_NFS_DIRECTIO
238 if (iocb->ki_filp->f_flags & O_DIRECT)
Badari Pulavarty027445c2006-09-30 23:28:46 -0700239 return nfs_file_direct_read(iocb, iov, nr_segs, pos);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700240#endif
241
242 dfprintk(VFS, "nfs: read(%s/%s, %lu@%lu)\n",
243 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 Lever91d5b472006-03-20 13:44:14 -0500247 nfs_add_stats(inode, NFSIOS_NORMALREADBYTES, count);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700248 if (!result)
Badari Pulavarty027445c2006-09-30 23:28:46 -0700249 result = generic_file_aio_read(iocb, iov, nr_segs, pos);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700250 return result;
251}
252
253static ssize_t
Jens Axboef0930ff2007-06-01 11:51:43 +0200254nfs_file_splice_read(struct file *filp, loff_t *ppos,
255 struct pipe_inode_info *pipe, size_t count,
256 unsigned int flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700257{
Josef "Jeff" Sipek01cce932006-12-08 02:36:40 -0800258 struct dentry *dentry = filp->f_path.dentry;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700259 struct inode *inode = dentry->d_inode;
260 ssize_t res;
261
Jens Axboef0930ff2007-06-01 11:51:43 +0200262 dfprintk(VFS, "nfs: splice_read(%s/%s, %lu@%Lu)\n",
Linus Torvalds1da177e2005-04-16 15:20:36 -0700263 dentry->d_parent->d_name.name, dentry->d_name.name,
264 (unsigned long) count, (unsigned long long) *ppos);
265
Trond Myklebust44b11872006-05-25 01:40:59 -0400266 res = nfs_revalidate_mapping(inode, filp->f_mapping);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700267 if (!res)
Jens Axboef0930ff2007-06-01 11:51:43 +0200268 res = generic_file_splice_read(filp, ppos, pipe, count, flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700269 return res;
270}
271
272static int
273nfs_file_mmap(struct file * file, struct vm_area_struct * vma)
274{
Josef "Jeff" Sipek01cce932006-12-08 02:36:40 -0800275 struct dentry *dentry = file->f_path.dentry;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700276 struct inode *inode = dentry->d_inode;
277 int status;
278
279 dfprintk(VFS, "nfs: mmap(%s/%s)\n",
280 dentry->d_parent->d_name.name, dentry->d_name.name);
281
Trond Myklebust44b11872006-05-25 01:40:59 -0400282 status = nfs_revalidate_mapping(inode, file->f_mapping);
Trond Myklebust94387fb2007-07-22 17:09:05 -0400283 if (!status) {
284 vma->vm_ops = &nfs_file_vm_ops;
285 vma->vm_flags |= VM_CAN_NONLINEAR;
286 file_accessed(file);
287 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700288 return status;
289}
290
291/*
292 * Flush any dirty pages for this process, and check for write errors.
293 * The return status from this call provides a reliable indication of
294 * whether any write errors occurred for this process.
295 */
296static int
297nfs_fsync(struct file *file, struct dentry *dentry, int datasync)
298{
Trond Myklebustcd3758e2007-08-10 17:44:32 -0400299 struct nfs_open_context *ctx = nfs_file_open_context(file);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700300 struct inode *inode = dentry->d_inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700301
302 dfprintk(VFS, "nfs: fsync(%s/%ld)\n", inode->i_sb->s_id, inode->i_ino);
303
Chuck Lever91d5b472006-03-20 13:44:14 -0500304 nfs_inc_stats(inode, NFSIOS_VFSFSYNC);
Trond Myklebust7b159fc2007-07-25 14:09:54 -0400305 return nfs_do_fsync(ctx, inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700306}
307
308/*
Nick Piggin4899f9c2007-10-16 01:25:16 -0700309 * This does the "real" work of the write. We must allocate and lock the
310 * page to be sent back to the generic routine, which then copies the
311 * data from user space.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700312 *
313 * If the writer ends up delaying the write, the writer needs to
314 * increment the page use counts until he is done with the page.
315 */
Nick Piggin4899f9c2007-10-16 01:25:16 -0700316static int nfs_write_begin(struct file *file, struct address_space *mapping,
317 loff_t pos, unsigned len, unsigned flags,
318 struct page **pagep, void **fsdata)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700319{
Nick Piggin4899f9c2007-10-16 01:25:16 -0700320 int ret;
321 pgoff_t index;
322 struct page *page;
323 index = pos >> PAGE_CACHE_SHIFT;
324
325 page = __grab_cache_page(mapping, index);
326 if (!page)
327 return -ENOMEM;
328 *pagep = page;
329
330 ret = nfs_flush_incompatible(file, page);
331 if (ret) {
332 unlock_page(page);
333 page_cache_release(page);
334 }
335 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700336}
337
Nick Piggin4899f9c2007-10-16 01:25:16 -0700338static int nfs_write_end(struct file *file, struct address_space *mapping,
339 loff_t pos, unsigned len, unsigned copied,
340 struct page *page, void *fsdata)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700341{
Nick Piggin4899f9c2007-10-16 01:25:16 -0700342 unsigned offset = pos & (PAGE_CACHE_SIZE - 1);
343 int status;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700344
345 lock_kernel();
Nick Piggin4899f9c2007-10-16 01:25:16 -0700346 status = nfs_updatepage(file, page, offset, copied);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700347 unlock_kernel();
Nick Piggin4899f9c2007-10-16 01:25:16 -0700348
349 unlock_page(page);
350 page_cache_release(page);
351
Chuck Lever3d509e52007-12-20 14:55:04 -0500352 if (status < 0)
353 return status;
354 return copied;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700355}
356
NeilBrown2ff28e22006-03-26 01:37:18 -0800357static void nfs_invalidate_page(struct page *page, unsigned long offset)
Trond Myklebustcd52ed32006-03-20 13:44:04 -0500358{
Trond Myklebust1c759502006-10-09 16:18:38 -0400359 if (offset != 0)
360 return;
Trond Myklebustd2ccddf2006-05-31 01:13:38 -0400361 /* Cancel any unstarted writes on this page */
Trond Myklebust1b3b4a12007-08-28 10:29:36 -0400362 nfs_wb_page_cancel(page->mapping->host, page);
Trond Myklebustcd52ed32006-03-20 13:44:04 -0500363}
364
365static int nfs_release_page(struct page *page, gfp_t gfp)
366{
Trond Myklebuste3db7692007-01-10 23:15:39 -0800367 /* If PagePrivate() is set, then the page is not freeable */
368 return 0;
369}
370
371static int nfs_launder_page(struct page *page)
372{
373 return nfs_wb_page(page->mapping->host, page);
Trond Myklebustcd52ed32006-03-20 13:44:04 -0500374}
375
Christoph Hellwigf5e54d62006-06-28 04:26:44 -0700376const struct address_space_operations nfs_file_aops = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700377 .readpage = nfs_readpage,
378 .readpages = nfs_readpages,
Trond Myklebust9cccef92007-07-22 17:09:05 -0400379 .set_page_dirty = __set_page_dirty_nobuffers,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700380 .writepage = nfs_writepage,
381 .writepages = nfs_writepages,
Nick Piggin4899f9c2007-10-16 01:25:16 -0700382 .write_begin = nfs_write_begin,
383 .write_end = nfs_write_end,
Trond Myklebustcd52ed32006-03-20 13:44:04 -0500384 .invalidatepage = nfs_invalidate_page,
385 .releasepage = nfs_release_page,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700386#ifdef CONFIG_NFS_DIRECTIO
387 .direct_IO = nfs_direct_IO,
388#endif
Trond Myklebuste3db7692007-01-10 23:15:39 -0800389 .launder_page = nfs_launder_page,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700390};
391
Trond Myklebust94387fb2007-07-22 17:09:05 -0400392static int nfs_vm_page_mkwrite(struct vm_area_struct *vma, struct page *page)
393{
394 struct file *filp = vma->vm_file;
395 unsigned pagelen;
396 int ret = -EINVAL;
Nick Piggin4899f9c2007-10-16 01:25:16 -0700397 struct address_space *mapping;
Trond Myklebust94387fb2007-07-22 17:09:05 -0400398
399 lock_page(page);
Nick Piggin4899f9c2007-10-16 01:25:16 -0700400 mapping = page->mapping;
Trond Myklebust8b1f9ee2008-01-22 17:13:06 -0500401 if (mapping != vma->vm_file->f_path.dentry->d_inode->i_mapping)
402 goto out_unlock;
403
404 ret = 0;
Trond Myklebust94387fb2007-07-22 17:09:05 -0400405 pagelen = nfs_page_length(page);
Trond Myklebust8b1f9ee2008-01-22 17:13:06 -0500406 if (pagelen == 0)
407 goto out_unlock;
408
409 ret = nfs_flush_incompatible(filp, page);
410 if (ret != 0)
411 goto out_unlock;
412
413 ret = nfs_updatepage(filp, page, 0, pagelen);
414 if (ret == 0)
415 ret = pagelen;
416out_unlock:
Trond Myklebust94387fb2007-07-22 17:09:05 -0400417 unlock_page(page);
418 return ret;
419}
420
421static struct vm_operations_struct nfs_file_vm_ops = {
422 .fault = filemap_fault,
423 .page_mkwrite = nfs_vm_page_mkwrite,
424};
425
Trond Myklebust7b159fc2007-07-25 14:09:54 -0400426static int nfs_need_sync_write(struct file *filp, struct inode *inode)
427{
428 struct nfs_open_context *ctx;
429
430 if (IS_SYNC(inode) || (filp->f_flags & O_SYNC))
431 return 1;
Trond Myklebustcd3758e2007-08-10 17:44:32 -0400432 ctx = nfs_file_open_context(filp);
Trond Myklebust7b159fc2007-07-25 14:09:54 -0400433 if (test_bit(NFS_CONTEXT_ERROR_WRITE, &ctx->flags))
434 return 1;
435 return 0;
436}
437
Badari Pulavarty027445c2006-09-30 23:28:46 -0700438static ssize_t nfs_file_write(struct kiocb *iocb, const struct iovec *iov,
439 unsigned long nr_segs, loff_t pos)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700440{
Josef "Jeff" Sipek01cce932006-12-08 02:36:40 -0800441 struct dentry * dentry = iocb->ki_filp->f_path.dentry;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700442 struct inode * inode = dentry->d_inode;
443 ssize_t result;
Badari Pulavarty027445c2006-09-30 23:28:46 -0700444 size_t count = iov_length(iov, nr_segs);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700445
446#ifdef CONFIG_NFS_DIRECTIO
447 if (iocb->ki_filp->f_flags & O_DIRECT)
Badari Pulavarty027445c2006-09-30 23:28:46 -0700448 return nfs_file_direct_write(iocb, iov, nr_segs, pos);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700449#endif
450
Badari Pulavarty027445c2006-09-30 23:28:46 -0700451 dfprintk(VFS, "nfs: write(%s/%s(%ld), %lu@%Ld)\n",
Linus Torvalds1da177e2005-04-16 15:20:36 -0700452 dentry->d_parent->d_name.name, dentry->d_name.name,
Badari Pulavarty027445c2006-09-30 23:28:46 -0700453 inode->i_ino, (unsigned long) count, (long long) pos);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700454
455 result = -EBUSY;
456 if (IS_SWAPFILE(inode))
457 goto out_swapfile;
Trond Myklebust7d52e862005-06-22 17:16:30 +0000458 /*
459 * O_APPEND implies that we must revalidate the file length.
460 */
461 if (iocb->ki_filp->f_flags & O_APPEND) {
462 result = nfs_revalidate_file_size(inode, iocb->ki_filp);
463 if (result)
464 goto out;
Trond Myklebustfe51bee2005-06-22 17:16:30 +0000465 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700466
467 result = count;
468 if (!count)
469 goto out;
470
Chuck Lever91d5b472006-03-20 13:44:14 -0500471 nfs_add_stats(inode, NFSIOS_NORMALWRITTENBYTES, count);
Badari Pulavarty027445c2006-09-30 23:28:46 -0700472 result = generic_file_aio_write(iocb, iov, nr_segs, pos);
Trond Myklebust200baa22006-12-05 00:35:40 -0500473 /* Return error values for O_SYNC and IS_SYNC() */
Trond Myklebust7b159fc2007-07-25 14:09:54 -0400474 if (result >= 0 && nfs_need_sync_write(iocb->ki_filp, inode)) {
Trond Myklebustcd3758e2007-08-10 17:44:32 -0400475 int err = nfs_do_fsync(nfs_file_open_context(iocb->ki_filp), inode);
Trond Myklebust200baa22006-12-05 00:35:40 -0500476 if (err < 0)
477 result = err;
478 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700479out:
480 return result;
481
482out_swapfile:
483 printk(KERN_INFO "NFS: attempt to write to active swap file!\n");
484 goto out;
485}
486
487static int do_getlk(struct file *filp, int cmd, struct file_lock *fl)
488{
489 struct inode *inode = filp->f_mapping->host;
490 int status = 0;
491
492 lock_kernel();
Trond Myklebust039c4d72005-10-18 14:20:16 -0700493 /* Try local locking first */
J. Bruce Fields6d34ac12007-05-11 16:09:32 -0400494 posix_test_lock(filp, fl);
495 if (fl->fl_type != F_UNLCK) {
496 /* found a conflict */
Trond Myklebust039c4d72005-10-18 14:20:16 -0700497 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700498 }
Trond Myklebust039c4d72005-10-18 14:20:16 -0700499
500 if (nfs_have_delegation(inode, FMODE_READ))
501 goto out_noconflict;
502
503 if (NFS_SERVER(inode)->flags & NFS_MOUNT_NONLM)
504 goto out_noconflict;
505
506 status = NFS_PROTO(inode)->lock(filp, cmd, fl);
507out:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700508 unlock_kernel();
509 return status;
Trond Myklebust039c4d72005-10-18 14:20:16 -0700510out_noconflict:
511 fl->fl_type = F_UNLCK;
512 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700513}
514
515static int do_vfs_lock(struct file *file, struct file_lock *fl)
516{
517 int res = 0;
518 switch (fl->fl_flags & (FL_POSIX|FL_FLOCK)) {
519 case FL_POSIX:
520 res = posix_lock_file_wait(file, fl);
521 break;
522 case FL_FLOCK:
523 res = flock_lock_file_wait(file, fl);
524 break;
525 default:
526 BUG();
527 }
528 if (res < 0)
Neil Brown46bae1a2007-01-30 14:36:01 -0800529 dprintk(KERN_WARNING "%s: VFS is out of sync with lock manager"
530 " - error %d!\n",
531 __FUNCTION__, res);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700532 return res;
533}
534
535static int do_unlk(struct file *filp, int cmd, struct file_lock *fl)
536{
537 struct inode *inode = filp->f_mapping->host;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700538 int status;
539
Linus Torvalds1da177e2005-04-16 15:20:36 -0700540 /*
541 * Flush all pending writes before doing anything
542 * with locks..
543 */
Trond Myklebust29884df2005-12-13 16:13:54 -0500544 nfs_sync_mapping(filp->f_mapping);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700545
546 /* NOTE: special case
547 * If we're signalled while cleaning up locks on process exit, we
548 * still need to complete the unlock.
549 */
550 lock_kernel();
551 /* Use local locking if mounted with "-onolock" */
552 if (!(NFS_SERVER(inode)->flags & NFS_MOUNT_NONLM))
553 status = NFS_PROTO(inode)->lock(filp, cmd, fl);
554 else
555 status = do_vfs_lock(filp, fl);
556 unlock_kernel();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700557 return status;
558}
559
560static int do_setlk(struct file *filp, int cmd, struct file_lock *fl)
561{
562 struct inode *inode = filp->f_mapping->host;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700563 int status;
564
Linus Torvalds1da177e2005-04-16 15:20:36 -0700565 /*
566 * Flush all pending writes before doing anything
567 * with locks..
568 */
Trond Myklebust29884df2005-12-13 16:13:54 -0500569 status = nfs_sync_mapping(filp->f_mapping);
570 if (status != 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700571 goto out;
572
573 lock_kernel();
574 /* Use local locking if mounted with "-onolock" */
575 if (!(NFS_SERVER(inode)->flags & NFS_MOUNT_NONLM)) {
576 status = NFS_PROTO(inode)->lock(filp, cmd, fl);
577 /* If we were signalled we still need to ensure that
578 * we clean up any state on the server. We therefore
579 * record the lock call as having succeeded in order to
580 * ensure that locks_remove_posix() cleans it out when
581 * the process exits.
582 */
583 if (status == -EINTR || status == -ERESTARTSYS)
584 do_vfs_lock(filp, fl);
585 } else
586 status = do_vfs_lock(filp, fl);
587 unlock_kernel();
588 if (status < 0)
589 goto out;
590 /*
591 * Make sure we clear the cache whenever we try to get the lock.
592 * This makes locking act as a cache coherency point.
593 */
Trond Myklebust29884df2005-12-13 16:13:54 -0500594 nfs_sync_mapping(filp->f_mapping);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700595 nfs_zap_caches(inode);
596out:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700597 return status;
598}
599
600/*
601 * Lock a (portion of) a file
602 */
603static int nfs_lock(struct file *filp, int cmd, struct file_lock *fl)
604{
605 struct inode * inode = filp->f_mapping->host;
606
607 dprintk("NFS: nfs_lock(f=%s/%ld, t=%x, fl=%x, r=%Ld:%Ld)\n",
608 inode->i_sb->s_id, inode->i_ino,
609 fl->fl_type, fl->fl_flags,
610 (long long)fl->fl_start, (long long)fl->fl_end);
Chuck Lever91d5b472006-03-20 13:44:14 -0500611 nfs_inc_stats(inode, NFSIOS_VFSLOCK);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700612
613 /* No mandatory locks over NFS */
Pavel Emelyanovdfad9442007-10-01 14:41:15 -0700614 if (__mandatory_lock(inode) && fl->fl_type != F_UNLCK)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700615 return -ENOLCK;
616
617 if (IS_GETLK(cmd))
618 return do_getlk(filp, cmd, fl);
619 if (fl->fl_type == F_UNLCK)
620 return do_unlk(filp, cmd, fl);
621 return do_setlk(filp, cmd, fl);
622}
623
624/*
625 * Lock a (portion of) a file
626 */
627static int nfs_flock(struct file *filp, int cmd, struct file_lock *fl)
628{
Linus Torvalds1da177e2005-04-16 15:20:36 -0700629 dprintk("NFS: nfs_flock(f=%s/%ld, t=%x, fl=%x)\n",
Josef "Jeff" Sipek01cce932006-12-08 02:36:40 -0800630 filp->f_path.dentry->d_inode->i_sb->s_id,
631 filp->f_path.dentry->d_inode->i_ino,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700632 fl->fl_type, fl->fl_flags);
633
Linus Torvalds1da177e2005-04-16 15:20:36 -0700634 /*
635 * No BSD flocks over NFS allowed.
636 * Note: we could try to fake a POSIX lock request here by
637 * using ((u32) filp | 0x80000000) or some such as the pid.
638 * Not sure whether that would be unique, though, or whether
639 * that would break in other places.
640 */
641 if (!(fl->fl_flags & FL_FLOCK))
642 return -ENOLCK;
643
644 /* We're simulating flock() locks using posix locks on the server */
645 fl->fl_owner = (fl_owner_t)filp;
646 fl->fl_start = 0;
647 fl->fl_end = OFFSET_MAX;
648
649 if (fl->fl_type == F_UNLCK)
650 return do_unlk(filp, cmd, fl);
651 return do_setlk(filp, cmd, fl);
652}
J. Bruce Fields370f6592007-06-08 15:23:34 -0400653
654static int nfs_setlease(struct file *file, long arg, struct file_lock **fl)
655{
656 /*
657 * There is no protocol support for leases, so we have no way
658 * to implement them correctly in the face of opens by other
659 * clients.
660 */
661 return -EINVAL;
662}