blob: 5d2e9d9a4e28a390c15fd10a0c856baf2fe89058 [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,
Bryan Wu240ee832008-04-01 20:23:39 -070067#ifdef CONFIG_MMU
Linus Torvalds1da177e2005-04-16 15:20:36 -070068 .mmap = nfs_file_mmap,
Bryan Wu240ee832008-04-01 20:23:39 -070069#else
70 .mmap = generic_file_mmap,
71#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -070072 .open = nfs_file_open,
73 .flush = nfs_file_flush,
74 .release = nfs_file_release,
75 .fsync = nfs_fsync,
76 .lock = nfs_lock,
77 .flock = nfs_flock,
Jens Axboef0930ff2007-06-01 11:51:43 +020078 .splice_read = nfs_file_splice_read,
Linus Torvalds1da177e2005-04-16 15:20:36 -070079 .check_flags = nfs_check_flags,
J. Bruce Fields370f6592007-06-08 15:23:34 -040080 .setlease = nfs_setlease,
Linus Torvalds1da177e2005-04-16 15:20:36 -070081};
82
Arjan van de Ven92e1d5b2007-02-12 00:55:39 -080083const struct inode_operations nfs_file_inode_operations = {
Linus Torvalds1da177e2005-04-16 15:20:36 -070084 .permission = nfs_permission,
85 .getattr = nfs_getattr,
86 .setattr = nfs_setattr,
87};
88
Andreas Gruenbacherb7fa0552005-06-22 17:16:27 +000089#ifdef CONFIG_NFS_V3
Arjan van de Ven92e1d5b2007-02-12 00:55:39 -080090const struct inode_operations nfs3_file_inode_operations = {
Andreas Gruenbacherb7fa0552005-06-22 17:16:27 +000091 .permission = nfs_permission,
92 .getattr = nfs_getattr,
93 .setattr = nfs_setattr,
94 .listxattr = nfs3_listxattr,
95 .getxattr = nfs3_getxattr,
96 .setxattr = nfs3_setxattr,
97 .removexattr = nfs3_removexattr,
98};
99#endif /* CONFIG_NFS_v3 */
100
Linus Torvalds1da177e2005-04-16 15:20:36 -0700101/* Hack for future NFS swap support */
102#ifndef IS_SWAPFILE
103# define IS_SWAPFILE(inode) (0)
104#endif
105
106static int nfs_check_flags(int flags)
107{
108 if ((flags & (O_APPEND | O_DIRECT)) == (O_APPEND | O_DIRECT))
109 return -EINVAL;
110
111 return 0;
112}
113
114/*
115 * Open file
116 */
117static int
118nfs_file_open(struct inode *inode, struct file *filp)
119{
Linus Torvalds1da177e2005-04-16 15:20:36 -0700120 int res;
121
122 res = nfs_check_flags(filp->f_flags);
123 if (res)
124 return res;
125
Chuck Lever91d5b472006-03-20 13:44:14 -0500126 nfs_inc_stats(inode, NFSIOS_VFSOPEN);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700127 lock_kernel();
David Howells1f163412006-08-22 20:06:11 -0400128 res = NFS_PROTO(inode)->file_open(inode, filp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700129 unlock_kernel();
130 return res;
131}
132
133static int
134nfs_file_release(struct inode *inode, struct file *filp)
135{
136 /* Ensure that dirty pages are flushed out with the right creds */
137 if (filp->f_mode & FMODE_WRITE)
Trond Myklebusta49c3c72007-10-18 18:03:27 -0400138 nfs_wb_all(filp->f_path.dentry->d_inode);
Chuck Lever91d5b472006-03-20 13:44:14 -0500139 nfs_inc_stats(inode, NFSIOS_VFSRELEASE);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700140 return NFS_PROTO(inode)->file_release(inode, filp);
141}
142
Trond Myklebust980802e2005-06-13 11:14:01 -0400143/**
144 * nfs_revalidate_size - Revalidate the file size
145 * @inode - pointer to inode struct
146 * @file - pointer to struct file
147 *
148 * Revalidates the file length. This is basically a wrapper around
149 * nfs_revalidate_inode() that takes into account the fact that we may
150 * have cached writes (in which case we don't care about the server's
151 * idea of what the file length is), or O_DIRECT (in which case we
152 * shouldn't trust the cache).
153 */
154static int nfs_revalidate_file_size(struct inode *inode, struct file *filp)
155{
156 struct nfs_server *server = NFS_SERVER(inode);
157 struct nfs_inode *nfsi = NFS_I(inode);
158
159 if (server->flags & NFS_MOUNT_NOAC)
160 goto force_reval;
161 if (filp->f_flags & O_DIRECT)
162 goto force_reval;
163 if (nfsi->npages != 0)
164 return 0;
Chuck Lever55296802005-08-18 11:24:09 -0700165 if (!(nfsi->cache_validity & NFS_INO_REVAL_PAGECACHE) && !nfs_attribute_timeout(inode))
Trond Myklebustfe51beec2005-06-22 17:16:30 +0000166 return 0;
Trond Myklebust980802e2005-06-13 11:14:01 -0400167force_reval:
168 return __nfs_revalidate_inode(server, inode);
169}
170
171static loff_t nfs_file_llseek(struct file *filp, loff_t offset, int origin)
172{
173 /* origin == SEEK_END => we must revalidate the cached file length */
Josef 'Jeff' Sipekaec5e172006-09-16 21:09:32 -0400174 if (origin == SEEK_END) {
Trond Myklebust980802e2005-06-13 11:14:01 -0400175 struct inode *inode = filp->f_mapping->host;
176 int retval = nfs_revalidate_file_size(inode, filp);
177 if (retval < 0)
178 return (loff_t)retval;
179 }
180 return remote_llseek(filp, offset, origin);
181}
182
Linus Torvalds1da177e2005-04-16 15:20:36 -0700183/*
Trond Myklebust7b159fc2007-07-25 14:09:54 -0400184 * Helper for nfs_file_flush() and nfs_fsync()
185 *
186 * Notice that it clears the NFS_CONTEXT_ERROR_WRITE before synching to
187 * disk, but it retrieves and clears ctx->error after synching, despite
188 * the two being set at the same time in nfs_context_set_write_error().
189 * This is because the former is used to notify the _next_ call to
190 * nfs_file_write() that a write error occured, and hence cause it to
191 * fall back to doing a synchronous write.
192 */
193static int nfs_do_fsync(struct nfs_open_context *ctx, struct inode *inode)
194{
195 int have_error, status;
196 int ret = 0;
197
198 have_error = test_and_clear_bit(NFS_CONTEXT_ERROR_WRITE, &ctx->flags);
199 status = nfs_wb_all(inode);
200 have_error |= test_bit(NFS_CONTEXT_ERROR_WRITE, &ctx->flags);
201 if (have_error)
202 ret = xchg(&ctx->error, 0);
203 if (!ret)
204 ret = status;
205 return ret;
206}
207
208/*
Linus Torvalds1da177e2005-04-16 15:20:36 -0700209 * Flush all dirty pages, and check for write errors.
210 *
211 */
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{
Trond Myklebustcd3758e2007-08-10 17:44:32 -0400215 struct nfs_open_context *ctx = nfs_file_open_context(file);
Josef "Jeff" Sipek01cce932006-12-08 02:36:40 -0800216 struct inode *inode = file->f_path.dentry->d_inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700217 int status;
218
219 dfprintk(VFS, "nfs: flush(%s/%ld)\n", inode->i_sb->s_id, inode->i_ino);
220
221 if ((file->f_mode & FMODE_WRITE) == 0)
222 return 0;
Chuck Lever91d5b472006-03-20 13:44:14 -0500223 nfs_inc_stats(inode, NFSIOS_VFSFLUSH);
Trond Myklebust7b159fc2007-07-25 14:09:54 -0400224
Linus Torvalds1da177e2005-04-16 15:20:36 -0700225 /* Ensure that data+attribute caches are up to date after close() */
Trond Myklebust7b159fc2007-07-25 14:09:54 -0400226 status = nfs_do_fsync(ctx, inode);
227 if (!status)
228 nfs_revalidate_inode(NFS_SERVER(inode), inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700229 return status;
230}
231
232static ssize_t
Badari Pulavarty027445c2006-09-30 23:28:46 -0700233nfs_file_read(struct kiocb *iocb, const struct iovec *iov,
234 unsigned long nr_segs, loff_t pos)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700235{
Josef "Jeff" Sipek01cce932006-12-08 02:36:40 -0800236 struct dentry * dentry = iocb->ki_filp->f_path.dentry;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700237 struct inode * inode = dentry->d_inode;
238 ssize_t result;
Badari Pulavarty027445c2006-09-30 23:28:46 -0700239 size_t count = iov_length(iov, nr_segs);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700240
241#ifdef CONFIG_NFS_DIRECTIO
242 if (iocb->ki_filp->f_flags & O_DIRECT)
Badari Pulavarty027445c2006-09-30 23:28:46 -0700243 return nfs_file_direct_read(iocb, iov, nr_segs, pos);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700244#endif
245
246 dfprintk(VFS, "nfs: read(%s/%s, %lu@%lu)\n",
247 dentry->d_parent->d_name.name, dentry->d_name.name,
248 (unsigned long) count, (unsigned long) pos);
249
Trond Myklebust44b11872006-05-25 01:40:59 -0400250 result = nfs_revalidate_mapping(inode, iocb->ki_filp->f_mapping);
Chuck Lever91d5b472006-03-20 13:44:14 -0500251 nfs_add_stats(inode, NFSIOS_NORMALREADBYTES, count);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700252 if (!result)
Badari Pulavarty027445c2006-09-30 23:28:46 -0700253 result = generic_file_aio_read(iocb, iov, nr_segs, pos);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700254 return result;
255}
256
257static ssize_t
Jens Axboef0930ff2007-06-01 11:51:43 +0200258nfs_file_splice_read(struct file *filp, loff_t *ppos,
259 struct pipe_inode_info *pipe, size_t count,
260 unsigned int flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700261{
Josef "Jeff" Sipek01cce932006-12-08 02:36:40 -0800262 struct dentry *dentry = filp->f_path.dentry;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700263 struct inode *inode = dentry->d_inode;
264 ssize_t res;
265
Jens Axboef0930ff2007-06-01 11:51:43 +0200266 dfprintk(VFS, "nfs: splice_read(%s/%s, %lu@%Lu)\n",
Linus Torvalds1da177e2005-04-16 15:20:36 -0700267 dentry->d_parent->d_name.name, dentry->d_name.name,
268 (unsigned long) count, (unsigned long long) *ppos);
269
Trond Myklebust44b11872006-05-25 01:40:59 -0400270 res = nfs_revalidate_mapping(inode, filp->f_mapping);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700271 if (!res)
Jens Axboef0930ff2007-06-01 11:51:43 +0200272 res = generic_file_splice_read(filp, ppos, pipe, count, flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700273 return res;
274}
275
276static int
277nfs_file_mmap(struct file * file, struct vm_area_struct * vma)
278{
Josef "Jeff" Sipek01cce932006-12-08 02:36:40 -0800279 struct dentry *dentry = file->f_path.dentry;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700280 struct inode *inode = dentry->d_inode;
281 int status;
282
283 dfprintk(VFS, "nfs: mmap(%s/%s)\n",
284 dentry->d_parent->d_name.name, dentry->d_name.name);
285
Trond Myklebust44b11872006-05-25 01:40:59 -0400286 status = nfs_revalidate_mapping(inode, file->f_mapping);
Trond Myklebust94387fb2007-07-22 17:09:05 -0400287 if (!status) {
288 vma->vm_ops = &nfs_file_vm_ops;
289 vma->vm_flags |= VM_CAN_NONLINEAR;
290 file_accessed(file);
291 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700292 return status;
293}
294
295/*
296 * Flush any dirty pages for this process, and check for write errors.
297 * The return status from this call provides a reliable indication of
298 * whether any write errors occurred for this process.
299 */
300static int
301nfs_fsync(struct file *file, struct dentry *dentry, int datasync)
302{
Trond Myklebustcd3758e2007-08-10 17:44:32 -0400303 struct nfs_open_context *ctx = nfs_file_open_context(file);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700304 struct inode *inode = dentry->d_inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700305
306 dfprintk(VFS, "nfs: fsync(%s/%ld)\n", inode->i_sb->s_id, inode->i_ino);
307
Chuck Lever91d5b472006-03-20 13:44:14 -0500308 nfs_inc_stats(inode, NFSIOS_VFSFSYNC);
Trond Myklebust7b159fc2007-07-25 14:09:54 -0400309 return nfs_do_fsync(ctx, inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700310}
311
312/*
Nick Piggin4899f9c2007-10-16 01:25:16 -0700313 * This does the "real" work of the write. We must allocate and lock the
314 * page to be sent back to the generic routine, which then copies the
315 * data from user space.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700316 *
317 * If the writer ends up delaying the write, the writer needs to
318 * increment the page use counts until he is done with the page.
319 */
Nick Piggin4899f9c2007-10-16 01:25:16 -0700320static int nfs_write_begin(struct file *file, struct address_space *mapping,
321 loff_t pos, unsigned len, unsigned flags,
322 struct page **pagep, void **fsdata)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700323{
Nick Piggin4899f9c2007-10-16 01:25:16 -0700324 int ret;
325 pgoff_t index;
326 struct page *page;
327 index = pos >> PAGE_CACHE_SHIFT;
328
329 page = __grab_cache_page(mapping, index);
330 if (!page)
331 return -ENOMEM;
332 *pagep = page;
333
334 ret = nfs_flush_incompatible(file, page);
335 if (ret) {
336 unlock_page(page);
337 page_cache_release(page);
338 }
339 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700340}
341
Nick Piggin4899f9c2007-10-16 01:25:16 -0700342static int nfs_write_end(struct file *file, struct address_space *mapping,
343 loff_t pos, unsigned len, unsigned copied,
344 struct page *page, void *fsdata)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700345{
Nick Piggin4899f9c2007-10-16 01:25:16 -0700346 unsigned offset = pos & (PAGE_CACHE_SIZE - 1);
347 int status;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700348
349 lock_kernel();
Nick Piggin4899f9c2007-10-16 01:25:16 -0700350 status = nfs_updatepage(file, page, offset, copied);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700351 unlock_kernel();
Nick Piggin4899f9c2007-10-16 01:25:16 -0700352
353 unlock_page(page);
354 page_cache_release(page);
355
Chuck Lever3d509e52007-12-20 14:55:04 -0500356 if (status < 0)
357 return status;
358 return copied;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700359}
360
NeilBrown2ff28e22006-03-26 01:37:18 -0800361static void nfs_invalidate_page(struct page *page, unsigned long offset)
Trond Myklebustcd52ed32006-03-20 13:44:04 -0500362{
Trond Myklebust1c759502006-10-09 16:18:38 -0400363 if (offset != 0)
364 return;
Trond Myklebustd2ccddf2006-05-31 01:13:38 -0400365 /* Cancel any unstarted writes on this page */
Trond Myklebust1b3b4a12007-08-28 10:29:36 -0400366 nfs_wb_page_cancel(page->mapping->host, page);
Trond Myklebustcd52ed32006-03-20 13:44:04 -0500367}
368
369static int nfs_release_page(struct page *page, gfp_t gfp)
370{
Trond Myklebuste3db7692007-01-10 23:15:39 -0800371 /* If PagePrivate() is set, then the page is not freeable */
372 return 0;
373}
374
375static int nfs_launder_page(struct page *page)
376{
377 return nfs_wb_page(page->mapping->host, page);
Trond Myklebustcd52ed32006-03-20 13:44:04 -0500378}
379
Christoph Hellwigf5e54d62006-06-28 04:26:44 -0700380const struct address_space_operations nfs_file_aops = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700381 .readpage = nfs_readpage,
382 .readpages = nfs_readpages,
Trond Myklebust9cccef92007-07-22 17:09:05 -0400383 .set_page_dirty = __set_page_dirty_nobuffers,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700384 .writepage = nfs_writepage,
385 .writepages = nfs_writepages,
Nick Piggin4899f9c2007-10-16 01:25:16 -0700386 .write_begin = nfs_write_begin,
387 .write_end = nfs_write_end,
Trond Myklebustcd52ed32006-03-20 13:44:04 -0500388 .invalidatepage = nfs_invalidate_page,
389 .releasepage = nfs_release_page,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700390#ifdef CONFIG_NFS_DIRECTIO
391 .direct_IO = nfs_direct_IO,
392#endif
Trond Myklebuste3db7692007-01-10 23:15:39 -0800393 .launder_page = nfs_launder_page,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700394};
395
Trond Myklebust94387fb2007-07-22 17:09:05 -0400396static int nfs_vm_page_mkwrite(struct vm_area_struct *vma, struct page *page)
397{
398 struct file *filp = vma->vm_file;
399 unsigned pagelen;
400 int ret = -EINVAL;
Nick Piggin4899f9c2007-10-16 01:25:16 -0700401 struct address_space *mapping;
Trond Myklebust94387fb2007-07-22 17:09:05 -0400402
403 lock_page(page);
Nick Piggin4899f9c2007-10-16 01:25:16 -0700404 mapping = page->mapping;
Trond Myklebust8b1f9ee2008-01-22 17:13:06 -0500405 if (mapping != vma->vm_file->f_path.dentry->d_inode->i_mapping)
406 goto out_unlock;
407
408 ret = 0;
Trond Myklebust94387fb2007-07-22 17:09:05 -0400409 pagelen = nfs_page_length(page);
Trond Myklebust8b1f9ee2008-01-22 17:13:06 -0500410 if (pagelen == 0)
411 goto out_unlock;
412
413 ret = nfs_flush_incompatible(filp, page);
414 if (ret != 0)
415 goto out_unlock;
416
417 ret = nfs_updatepage(filp, page, 0, pagelen);
418 if (ret == 0)
419 ret = pagelen;
420out_unlock:
Trond Myklebust94387fb2007-07-22 17:09:05 -0400421 unlock_page(page);
422 return ret;
423}
424
425static struct vm_operations_struct nfs_file_vm_ops = {
426 .fault = filemap_fault,
427 .page_mkwrite = nfs_vm_page_mkwrite,
428};
429
Trond Myklebust7b159fc2007-07-25 14:09:54 -0400430static int nfs_need_sync_write(struct file *filp, struct inode *inode)
431{
432 struct nfs_open_context *ctx;
433
434 if (IS_SYNC(inode) || (filp->f_flags & O_SYNC))
435 return 1;
Trond Myklebustcd3758e2007-08-10 17:44:32 -0400436 ctx = nfs_file_open_context(filp);
Trond Myklebust7b159fc2007-07-25 14:09:54 -0400437 if (test_bit(NFS_CONTEXT_ERROR_WRITE, &ctx->flags))
438 return 1;
439 return 0;
440}
441
Badari Pulavarty027445c2006-09-30 23:28:46 -0700442static ssize_t nfs_file_write(struct kiocb *iocb, const struct iovec *iov,
443 unsigned long nr_segs, loff_t pos)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700444{
Josef "Jeff" Sipek01cce932006-12-08 02:36:40 -0800445 struct dentry * dentry = iocb->ki_filp->f_path.dentry;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700446 struct inode * inode = dentry->d_inode;
447 ssize_t result;
Badari Pulavarty027445c2006-09-30 23:28:46 -0700448 size_t count = iov_length(iov, nr_segs);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700449
450#ifdef CONFIG_NFS_DIRECTIO
451 if (iocb->ki_filp->f_flags & O_DIRECT)
Badari Pulavarty027445c2006-09-30 23:28:46 -0700452 return nfs_file_direct_write(iocb, iov, nr_segs, pos);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700453#endif
454
Badari Pulavarty027445c2006-09-30 23:28:46 -0700455 dfprintk(VFS, "nfs: write(%s/%s(%ld), %lu@%Ld)\n",
Linus Torvalds1da177e2005-04-16 15:20:36 -0700456 dentry->d_parent->d_name.name, dentry->d_name.name,
Badari Pulavarty027445c2006-09-30 23:28:46 -0700457 inode->i_ino, (unsigned long) count, (long long) pos);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700458
459 result = -EBUSY;
460 if (IS_SWAPFILE(inode))
461 goto out_swapfile;
Trond Myklebust7d52e862005-06-22 17:16:30 +0000462 /*
463 * O_APPEND implies that we must revalidate the file length.
464 */
465 if (iocb->ki_filp->f_flags & O_APPEND) {
466 result = nfs_revalidate_file_size(inode, iocb->ki_filp);
467 if (result)
468 goto out;
Trond Myklebustfe51beec2005-06-22 17:16:30 +0000469 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700470
471 result = count;
472 if (!count)
473 goto out;
474
Chuck Lever91d5b472006-03-20 13:44:14 -0500475 nfs_add_stats(inode, NFSIOS_NORMALWRITTENBYTES, count);
Badari Pulavarty027445c2006-09-30 23:28:46 -0700476 result = generic_file_aio_write(iocb, iov, nr_segs, pos);
Trond Myklebust200baa22006-12-05 00:35:40 -0500477 /* Return error values for O_SYNC and IS_SYNC() */
Trond Myklebust7b159fc2007-07-25 14:09:54 -0400478 if (result >= 0 && nfs_need_sync_write(iocb->ki_filp, inode)) {
Trond Myklebustcd3758e2007-08-10 17:44:32 -0400479 int err = nfs_do_fsync(nfs_file_open_context(iocb->ki_filp), inode);
Trond Myklebust200baa22006-12-05 00:35:40 -0500480 if (err < 0)
481 result = err;
482 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700483out:
484 return result;
485
486out_swapfile:
487 printk(KERN_INFO "NFS: attempt to write to active swap file!\n");
488 goto out;
489}
490
491static int do_getlk(struct file *filp, int cmd, struct file_lock *fl)
492{
493 struct inode *inode = filp->f_mapping->host;
494 int status = 0;
495
496 lock_kernel();
Trond Myklebust039c4d72005-10-18 14:20:16 -0700497 /* Try local locking first */
J. Bruce Fields6d34ac12007-05-11 16:09:32 -0400498 posix_test_lock(filp, fl);
499 if (fl->fl_type != F_UNLCK) {
500 /* found a conflict */
Trond Myklebust039c4d72005-10-18 14:20:16 -0700501 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700502 }
Trond Myklebust039c4d72005-10-18 14:20:16 -0700503
504 if (nfs_have_delegation(inode, FMODE_READ))
505 goto out_noconflict;
506
507 if (NFS_SERVER(inode)->flags & NFS_MOUNT_NONLM)
508 goto out_noconflict;
509
510 status = NFS_PROTO(inode)->lock(filp, cmd, fl);
511out:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700512 unlock_kernel();
513 return status;
Trond Myklebust039c4d72005-10-18 14:20:16 -0700514out_noconflict:
515 fl->fl_type = F_UNLCK;
516 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700517}
518
519static int do_vfs_lock(struct file *file, struct file_lock *fl)
520{
521 int res = 0;
522 switch (fl->fl_flags & (FL_POSIX|FL_FLOCK)) {
523 case FL_POSIX:
524 res = posix_lock_file_wait(file, fl);
525 break;
526 case FL_FLOCK:
527 res = flock_lock_file_wait(file, fl);
528 break;
529 default:
530 BUG();
531 }
532 if (res < 0)
Neil Brown46bae1a2007-01-30 14:36:01 -0800533 dprintk(KERN_WARNING "%s: VFS is out of sync with lock manager"
534 " - error %d!\n",
535 __FUNCTION__, res);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700536 return res;
537}
538
539static int do_unlk(struct file *filp, int cmd, struct file_lock *fl)
540{
541 struct inode *inode = filp->f_mapping->host;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700542 int status;
543
Linus Torvalds1da177e2005-04-16 15:20:36 -0700544 /*
545 * Flush all pending writes before doing anything
546 * with locks..
547 */
Trond Myklebust29884df2005-12-13 16:13:54 -0500548 nfs_sync_mapping(filp->f_mapping);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700549
550 /* NOTE: special case
551 * If we're signalled while cleaning up locks on process exit, we
552 * still need to complete the unlock.
553 */
554 lock_kernel();
555 /* Use local locking if mounted with "-onolock" */
556 if (!(NFS_SERVER(inode)->flags & NFS_MOUNT_NONLM))
557 status = NFS_PROTO(inode)->lock(filp, cmd, fl);
558 else
559 status = do_vfs_lock(filp, fl);
560 unlock_kernel();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700561 return status;
562}
563
564static int do_setlk(struct file *filp, int cmd, struct file_lock *fl)
565{
566 struct inode *inode = filp->f_mapping->host;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700567 int status;
568
Linus Torvalds1da177e2005-04-16 15:20:36 -0700569 /*
570 * Flush all pending writes before doing anything
571 * with locks..
572 */
Trond Myklebust29884df2005-12-13 16:13:54 -0500573 status = nfs_sync_mapping(filp->f_mapping);
574 if (status != 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700575 goto out;
576
577 lock_kernel();
578 /* Use local locking if mounted with "-onolock" */
579 if (!(NFS_SERVER(inode)->flags & NFS_MOUNT_NONLM)) {
580 status = NFS_PROTO(inode)->lock(filp, cmd, fl);
581 /* If we were signalled we still need to ensure that
582 * we clean up any state on the server. We therefore
583 * record the lock call as having succeeded in order to
584 * ensure that locks_remove_posix() cleans it out when
585 * the process exits.
586 */
587 if (status == -EINTR || status == -ERESTARTSYS)
588 do_vfs_lock(filp, fl);
589 } else
590 status = do_vfs_lock(filp, fl);
591 unlock_kernel();
592 if (status < 0)
593 goto out;
594 /*
595 * Make sure we clear the cache whenever we try to get the lock.
596 * This makes locking act as a cache coherency point.
597 */
Trond Myklebust29884df2005-12-13 16:13:54 -0500598 nfs_sync_mapping(filp->f_mapping);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700599 nfs_zap_caches(inode);
600out:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700601 return status;
602}
603
604/*
605 * Lock a (portion of) a file
606 */
607static int nfs_lock(struct file *filp, int cmd, struct file_lock *fl)
608{
609 struct inode * inode = filp->f_mapping->host;
610
611 dprintk("NFS: nfs_lock(f=%s/%ld, t=%x, fl=%x, r=%Ld:%Ld)\n",
612 inode->i_sb->s_id, inode->i_ino,
613 fl->fl_type, fl->fl_flags,
614 (long long)fl->fl_start, (long long)fl->fl_end);
Chuck Lever91d5b472006-03-20 13:44:14 -0500615 nfs_inc_stats(inode, NFSIOS_VFSLOCK);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700616
617 /* No mandatory locks over NFS */
Pavel Emelyanovdfad9442007-10-01 14:41:15 -0700618 if (__mandatory_lock(inode) && fl->fl_type != F_UNLCK)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700619 return -ENOLCK;
620
621 if (IS_GETLK(cmd))
622 return do_getlk(filp, cmd, fl);
623 if (fl->fl_type == F_UNLCK)
624 return do_unlk(filp, cmd, fl);
625 return do_setlk(filp, cmd, fl);
626}
627
628/*
629 * Lock a (portion of) a file
630 */
631static int nfs_flock(struct file *filp, int cmd, struct file_lock *fl)
632{
Linus Torvalds1da177e2005-04-16 15:20:36 -0700633 dprintk("NFS: nfs_flock(f=%s/%ld, t=%x, fl=%x)\n",
Josef "Jeff" Sipek01cce932006-12-08 02:36:40 -0800634 filp->f_path.dentry->d_inode->i_sb->s_id,
635 filp->f_path.dentry->d_inode->i_ino,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700636 fl->fl_type, fl->fl_flags);
637
Linus Torvalds1da177e2005-04-16 15:20:36 -0700638 /*
639 * No BSD flocks over NFS allowed.
640 * Note: we could try to fake a POSIX lock request here by
641 * using ((u32) filp | 0x80000000) or some such as the pid.
642 * Not sure whether that would be unique, though, or whether
643 * that would break in other places.
644 */
645 if (!(fl->fl_flags & FL_FLOCK))
646 return -ENOLCK;
647
648 /* We're simulating flock() locks using posix locks on the server */
649 fl->fl_owner = (fl_owner_t)filp;
650 fl->fl_start = 0;
651 fl->fl_end = OFFSET_MAX;
652
653 if (fl->fl_type == F_UNLCK)
654 return do_unlk(filp, cmd, fl);
655 return do_setlk(filp, cmd, fl);
656}
J. Bruce Fields370f6592007-06-08 15:23:34 -0400657
658static int nfs_setlease(struct file *file, long arg, struct file_lock **fl)
659{
660 /*
661 * There is no protocol support for leases, so we have no way
662 * to implement them correctly in the face of opens by other
663 * clients.
664 */
665 return -EINVAL;
666}