blob: bb7ca022bcb2b62df3a19fa44a167ff411899a0b [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001/*
2 * linux/fs/nfs/inode.c
3 *
4 * Copyright (C) 1992 Rick Sladkey
5 *
6 * nfs inode and superblock handling functions
7 *
8 * Modularised by Alan Cox <Alan.Cox@linux.org>, while hacking some
9 * experimental NFS changes. Modularisation taken straight from SYS5 fs.
10 *
11 * Change to nfs_read_super() to permit NFS mounts to multi-homed hosts.
12 * J.S.Peatfield@damtp.cam.ac.uk
13 *
14 */
15
16#include <linux/config.h>
17#include <linux/module.h>
18#include <linux/init.h>
19
20#include <linux/time.h>
21#include <linux/kernel.h>
22#include <linux/mm.h>
23#include <linux/string.h>
24#include <linux/stat.h>
25#include <linux/errno.h>
26#include <linux/unistd.h>
27#include <linux/sunrpc/clnt.h>
28#include <linux/sunrpc/stats.h>
29#include <linux/nfs_fs.h>
30#include <linux/nfs_mount.h>
31#include <linux/nfs4_mount.h>
32#include <linux/lockd/bind.h>
33#include <linux/smp_lock.h>
34#include <linux/seq_file.h>
35#include <linux/mount.h>
36#include <linux/nfs_idmap.h>
37#include <linux/vfs.h>
38
39#include <asm/system.h>
40#include <asm/uaccess.h>
41
Trond Myklebust4ce79712005-06-22 17:16:21 +000042#include "nfs4_fs.h"
Linus Torvalds1da177e2005-04-16 15:20:36 -070043#include "delegation.h"
44
45#define NFSDBG_FACILITY NFSDBG_VFS
46#define NFS_PARANOIA 1
47
48/* Maximum number of readahead requests
49 * FIXME: this should really be a sysctl so that users may tune it to suit
50 * their needs. People that do NFS over a slow network, might for
51 * instance want to reduce it to something closer to 1 for improved
52 * interactive response.
53 */
54#define NFS_MAX_READAHEAD (RPC_DEF_SLOT_TABLE - 1)
55
56static void nfs_invalidate_inode(struct inode *);
57static int nfs_update_inode(struct inode *, struct nfs_fattr *, unsigned long);
58
59static struct inode *nfs_alloc_inode(struct super_block *sb);
60static void nfs_destroy_inode(struct inode *);
61static int nfs_write_inode(struct inode *,int);
62static void nfs_delete_inode(struct inode *);
63static void nfs_clear_inode(struct inode *);
64static void nfs_umount_begin(struct super_block *);
65static int nfs_statfs(struct super_block *, struct kstatfs *);
66static int nfs_show_options(struct seq_file *, struct vfsmount *);
Trond Myklebustada70d92005-06-22 17:16:22 +000067static void nfs_zap_acl_cache(struct inode *);
Linus Torvalds1da177e2005-04-16 15:20:36 -070068
69static struct rpc_program nfs_program;
70
71static struct super_operations nfs_sops = {
72 .alloc_inode = nfs_alloc_inode,
73 .destroy_inode = nfs_destroy_inode,
74 .write_inode = nfs_write_inode,
75 .delete_inode = nfs_delete_inode,
76 .statfs = nfs_statfs,
77 .clear_inode = nfs_clear_inode,
78 .umount_begin = nfs_umount_begin,
79 .show_options = nfs_show_options,
80};
81
82/*
83 * RPC cruft for NFS
84 */
85static struct rpc_stat nfs_rpcstat = {
86 .program = &nfs_program
87};
88static struct rpc_version * nfs_version[] = {
89 NULL,
90 NULL,
91 &nfs_version2,
92#if defined(CONFIG_NFS_V3)
93 &nfs_version3,
94#elif defined(CONFIG_NFS_V4)
95 NULL,
96#endif
97#if defined(CONFIG_NFS_V4)
98 &nfs_version4,
99#endif
100};
101
102static struct rpc_program nfs_program = {
103 .name = "nfs",
104 .number = NFS_PROGRAM,
105 .nrvers = sizeof(nfs_version) / sizeof(nfs_version[0]),
106 .version = nfs_version,
107 .stats = &nfs_rpcstat,
108 .pipe_dir_name = "/nfs",
109};
110
Andreas Gruenbacherb7fa0552005-06-22 17:16:27 +0000111#ifdef CONFIG_NFS_V3_ACL
112static struct rpc_stat nfsacl_rpcstat = { &nfsacl_program };
113static struct rpc_version * nfsacl_version[] = {
114 [3] = &nfsacl_version3,
115};
116
117struct rpc_program nfsacl_program = {
118 .name = "nfsacl",
119 .number = NFS_ACL_PROGRAM,
120 .nrvers = sizeof(nfsacl_version) / sizeof(nfsacl_version[0]),
121 .version = nfsacl_version,
122 .stats = &nfsacl_rpcstat,
123};
124#endif /* CONFIG_NFS_V3_ACL */
125
Linus Torvalds1da177e2005-04-16 15:20:36 -0700126static inline unsigned long
127nfs_fattr_to_ino_t(struct nfs_fattr *fattr)
128{
129 return nfs_fileid_to_ino_t(fattr->fileid);
130}
131
132static int
133nfs_write_inode(struct inode *inode, int sync)
134{
135 int flags = sync ? FLUSH_WAIT : 0;
136 int ret;
137
Trond Myklebust3da28eb2005-06-22 17:16:31 +0000138 ret = nfs_commit_inode(inode, flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700139 if (ret < 0)
140 return ret;
141 return 0;
142}
143
144static void
145nfs_delete_inode(struct inode * inode)
146{
147 dprintk("NFS: delete_inode(%s/%ld)\n", inode->i_sb->s_id, inode->i_ino);
148
149 nfs_wb_all(inode);
150 /*
151 * The following should never happen...
152 */
153 if (nfs_have_writebacks(inode)) {
154 printk(KERN_ERR "nfs_delete_inode: inode %ld has pending RPC requests\n", inode->i_ino);
155 }
156
157 clear_inode(inode);
158}
159
Linus Torvalds1da177e2005-04-16 15:20:36 -0700160static void
161nfs_clear_inode(struct inode *inode)
162{
163 struct nfs_inode *nfsi = NFS_I(inode);
164 struct rpc_cred *cred;
165
166 nfs_wb_all(inode);
167 BUG_ON (!list_empty(&nfsi->open_files));
Trond Myklebustada70d92005-06-22 17:16:22 +0000168 nfs_zap_acl_cache(inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700169 cred = nfsi->cache_access.cred;
170 if (cred)
171 put_rpccred(cred);
172 BUG_ON(atomic_read(&nfsi->data_updates) != 0);
173}
174
175void
176nfs_umount_begin(struct super_block *sb)
177{
J. Bruce Fields6a192752005-06-22 17:16:23 +0000178 struct rpc_clnt *rpc = NFS_SB(sb)->client;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700179
180 /* -EIO all pending I/O */
J. Bruce Fields6a192752005-06-22 17:16:23 +0000181 if (!IS_ERR(rpc))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700182 rpc_killall_tasks(rpc);
Andreas Gruenbacherb7fa0552005-06-22 17:16:27 +0000183 rpc = NFS_SB(sb)->client_acl;
184 if (!IS_ERR(rpc))
185 rpc_killall_tasks(rpc);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700186}
187
188
189static inline unsigned long
190nfs_block_bits(unsigned long bsize, unsigned char *nrbitsp)
191{
192 /* make sure blocksize is a power of two */
193 if ((bsize & (bsize - 1)) || nrbitsp) {
194 unsigned char nrbits;
195
196 for (nrbits = 31; nrbits && !(bsize & (1 << nrbits)); nrbits--)
197 ;
198 bsize = 1 << nrbits;
199 if (nrbitsp)
200 *nrbitsp = nrbits;
201 }
202
203 return bsize;
204}
205
206/*
207 * Calculate the number of 512byte blocks used.
208 */
209static inline unsigned long
210nfs_calc_block_size(u64 tsize)
211{
212 loff_t used = (tsize + 511) >> 9;
213 return (used > ULONG_MAX) ? ULONG_MAX : used;
214}
215
216/*
217 * Compute and set NFS server blocksize
218 */
219static inline unsigned long
220nfs_block_size(unsigned long bsize, unsigned char *nrbitsp)
221{
222 if (bsize < 1024)
223 bsize = NFS_DEF_FILE_IO_BUFFER_SIZE;
224 else if (bsize >= NFS_MAX_FILE_IO_BUFFER_SIZE)
225 bsize = NFS_MAX_FILE_IO_BUFFER_SIZE;
226
227 return nfs_block_bits(bsize, nrbitsp);
228}
229
230/*
231 * Obtain the root inode of the file system.
232 */
233static struct inode *
234nfs_get_root(struct super_block *sb, struct nfs_fh *rootfh, struct nfs_fsinfo *fsinfo)
235{
236 struct nfs_server *server = NFS_SB(sb);
237 struct inode *rooti;
238 int error;
239
240 error = server->rpc_ops->getroot(server, rootfh, fsinfo);
241 if (error < 0) {
242 dprintk("nfs_get_root: getattr error = %d\n", -error);
243 return ERR_PTR(error);
244 }
245
246 rooti = nfs_fhget(sb, rootfh, fsinfo->fattr);
247 if (!rooti)
248 return ERR_PTR(-ENOMEM);
249 return rooti;
250}
251
252/*
253 * Do NFS version-independent mount processing, and sanity checking
254 */
255static int
256nfs_sb_init(struct super_block *sb, rpc_authflavor_t authflavor)
257{
258 struct nfs_server *server;
259 struct inode *root_inode;
260 struct nfs_fattr fattr;
261 struct nfs_fsinfo fsinfo = {
262 .fattr = &fattr,
263 };
264 struct nfs_pathconf pathinfo = {
265 .fattr = &fattr,
266 };
267 int no_root_error = 0;
268 unsigned long max_rpc_payload;
269
270 /* We probably want something more informative here */
271 snprintf(sb->s_id, sizeof(sb->s_id), "%x:%x", MAJOR(sb->s_dev), MINOR(sb->s_dev));
272
273 server = NFS_SB(sb);
274
275 sb->s_magic = NFS_SUPER_MAGIC;
276
277 root_inode = nfs_get_root(sb, &server->fh, &fsinfo);
278 /* Did getting the root inode fail? */
279 if (IS_ERR(root_inode)) {
280 no_root_error = PTR_ERR(root_inode);
281 goto out_no_root;
282 }
283 sb->s_root = d_alloc_root(root_inode);
284 if (!sb->s_root) {
285 no_root_error = -ENOMEM;
286 goto out_no_root;
287 }
288 sb->s_root->d_op = server->rpc_ops->dentry_ops;
289
290 /* Get some general file system info */
291 if (server->namelen == 0 &&
292 server->rpc_ops->pathconf(server, &server->fh, &pathinfo) >= 0)
293 server->namelen = pathinfo.max_namelen;
294 /* Work out a lot of parameters */
295 if (server->rsize == 0)
296 server->rsize = nfs_block_size(fsinfo.rtpref, NULL);
297 if (server->wsize == 0)
298 server->wsize = nfs_block_size(fsinfo.wtpref, NULL);
299
300 if (fsinfo.rtmax >= 512 && server->rsize > fsinfo.rtmax)
301 server->rsize = nfs_block_size(fsinfo.rtmax, NULL);
302 if (fsinfo.wtmax >= 512 && server->wsize > fsinfo.wtmax)
303 server->wsize = nfs_block_size(fsinfo.wtmax, NULL);
304
305 max_rpc_payload = nfs_block_size(rpc_max_payload(server->client), NULL);
306 if (server->rsize > max_rpc_payload)
307 server->rsize = max_rpc_payload;
308 if (server->wsize > max_rpc_payload)
309 server->wsize = max_rpc_payload;
310
311 server->rpages = (server->rsize + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT;
312 if (server->rpages > NFS_READ_MAXIOV) {
313 server->rpages = NFS_READ_MAXIOV;
314 server->rsize = server->rpages << PAGE_CACHE_SHIFT;
315 }
316
317 server->wpages = (server->wsize + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT;
318 if (server->wpages > NFS_WRITE_MAXIOV) {
319 server->wpages = NFS_WRITE_MAXIOV;
320 server->wsize = server->wpages << PAGE_CACHE_SHIFT;
321 }
322
323 if (sb->s_blocksize == 0)
324 sb->s_blocksize = nfs_block_bits(server->wsize,
325 &sb->s_blocksize_bits);
326 server->wtmult = nfs_block_bits(fsinfo.wtmult, NULL);
327
328 server->dtsize = nfs_block_size(fsinfo.dtpref, NULL);
329 if (server->dtsize > PAGE_CACHE_SIZE)
330 server->dtsize = PAGE_CACHE_SIZE;
331 if (server->dtsize > server->rsize)
332 server->dtsize = server->rsize;
333
334 if (server->flags & NFS_MOUNT_NOAC) {
335 server->acregmin = server->acregmax = 0;
336 server->acdirmin = server->acdirmax = 0;
337 sb->s_flags |= MS_SYNCHRONOUS;
338 }
339 server->backing_dev_info.ra_pages = server->rpages * NFS_MAX_READAHEAD;
340
341 sb->s_maxbytes = fsinfo.maxfilesize;
342 if (sb->s_maxbytes > MAX_LFS_FILESIZE)
343 sb->s_maxbytes = MAX_LFS_FILESIZE;
344
345 server->client->cl_intr = (server->flags & NFS_MOUNT_INTR) ? 1 : 0;
346 server->client->cl_softrtry = (server->flags & NFS_MOUNT_SOFT) ? 1 : 0;
347
348 /* We're airborne Set socket buffersize */
349 rpc_setbufsize(server->client, server->wsize + 100, server->rsize + 100);
350 return 0;
351 /* Yargs. It didn't work out. */
352out_no_root:
353 dprintk("nfs_sb_init: get root inode failed: errno %d\n", -no_root_error);
354 if (!IS_ERR(root_inode))
355 iput(root_inode);
356 return no_root_error;
357}
358
359/*
360 * Create an RPC client handle.
361 */
362static struct rpc_clnt *
363nfs_create_client(struct nfs_server *server, const struct nfs_mount_data *data)
364{
365 struct rpc_timeout timeparms;
366 struct rpc_xprt *xprt = NULL;
367 struct rpc_clnt *clnt = NULL;
368 int tcp = (data->flags & NFS_MOUNT_TCP);
369
370 /* Initialize timeout values */
371 timeparms.to_initval = data->timeo * HZ / 10;
372 timeparms.to_retries = data->retrans;
373 timeparms.to_maxval = tcp ? RPC_MAX_TCP_TIMEOUT : RPC_MAX_UDP_TIMEOUT;
374 timeparms.to_exponential = 1;
375
376 if (!timeparms.to_initval)
377 timeparms.to_initval = (tcp ? 600 : 11) * HZ / 10;
378 if (!timeparms.to_retries)
379 timeparms.to_retries = 5;
380
381 /* create transport and client */
382 xprt = xprt_create_proto(tcp ? IPPROTO_TCP : IPPROTO_UDP,
383 &server->addr, &timeparms);
384 if (IS_ERR(xprt)) {
Trond Myklebust9085bbc2005-06-22 17:16:20 +0000385 dprintk("%s: cannot create RPC transport. Error = %ld\n",
386 __FUNCTION__, PTR_ERR(xprt));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700387 return (struct rpc_clnt *)xprt;
388 }
389 clnt = rpc_create_client(xprt, server->hostname, &nfs_program,
390 server->rpc_ops->version, data->pseudoflavor);
391 if (IS_ERR(clnt)) {
Trond Myklebust9085bbc2005-06-22 17:16:20 +0000392 dprintk("%s: cannot create RPC client. Error = %ld\n",
393 __FUNCTION__, PTR_ERR(xprt));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700394 goto out_fail;
395 }
396
397 clnt->cl_intr = 1;
398 clnt->cl_softrtry = 1;
399 clnt->cl_chatty = 1;
400
401 return clnt;
402
403out_fail:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700404 return clnt;
405}
406
407/*
408 * The way this works is that the mount process passes a structure
409 * in the data argument which contains the server's IP address
410 * and the root file handle obtained from the server's mount
411 * daemon. We stash these away in the private superblock fields.
412 */
413static int
414nfs_fill_super(struct super_block *sb, struct nfs_mount_data *data, int silent)
415{
416 struct nfs_server *server;
417 rpc_authflavor_t authflavor;
418
419 server = NFS_SB(sb);
420 sb->s_blocksize_bits = 0;
421 sb->s_blocksize = 0;
422 if (data->bsize)
423 sb->s_blocksize = nfs_block_size(data->bsize, &sb->s_blocksize_bits);
424 if (data->rsize)
425 server->rsize = nfs_block_size(data->rsize, NULL);
426 if (data->wsize)
427 server->wsize = nfs_block_size(data->wsize, NULL);
428 server->flags = data->flags & NFS_MOUNT_FLAGMASK;
429
430 server->acregmin = data->acregmin*HZ;
431 server->acregmax = data->acregmax*HZ;
432 server->acdirmin = data->acdirmin*HZ;
433 server->acdirmax = data->acdirmax*HZ;
434
435 /* Start lockd here, before we might error out */
436 if (!(server->flags & NFS_MOUNT_NONLM))
437 lockd_up();
438
439 server->namelen = data->namlen;
440 server->hostname = kmalloc(strlen(data->hostname) + 1, GFP_KERNEL);
441 if (!server->hostname)
442 return -ENOMEM;
443 strcpy(server->hostname, data->hostname);
444
445 /* Check NFS protocol revision and initialize RPC op vector
446 * and file handle pool. */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700447#ifdef CONFIG_NFS_V3
Trond Myklebust9085bbc2005-06-22 17:16:20 +0000448 if (server->flags & NFS_MOUNT_VER3) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700449 server->rpc_ops = &nfs_v3_clientops;
450 server->caps |= NFS_CAP_READDIRPLUS;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700451 } else {
452 server->rpc_ops = &nfs_v2_clientops;
453 }
Trond Myklebust9085bbc2005-06-22 17:16:20 +0000454#else
455 server->rpc_ops = &nfs_v2_clientops;
456#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700457
458 /* Fill in pseudoflavor for mount version < 5 */
459 if (!(data->flags & NFS_MOUNT_SECFLAVOUR))
460 data->pseudoflavor = RPC_AUTH_UNIX;
461 authflavor = data->pseudoflavor; /* save for sb_init() */
462 /* XXX maybe we want to add a server->pseudoflavor field */
463
464 /* Create RPC client handles */
465 server->client = nfs_create_client(server, data);
466 if (IS_ERR(server->client))
467 return PTR_ERR(server->client);
468 /* RFC 2623, sec 2.3.2 */
469 if (authflavor != RPC_AUTH_UNIX) {
J. Bruce Fields6a192752005-06-22 17:16:23 +0000470 struct rpc_auth *auth;
471
Linus Torvalds1da177e2005-04-16 15:20:36 -0700472 server->client_sys = rpc_clone_client(server->client);
473 if (IS_ERR(server->client_sys))
474 return PTR_ERR(server->client_sys);
J. Bruce Fields6a192752005-06-22 17:16:23 +0000475 auth = rpcauth_create(RPC_AUTH_UNIX, server->client_sys);
476 if (IS_ERR(auth))
477 return PTR_ERR(auth);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700478 } else {
479 atomic_inc(&server->client->cl_count);
480 server->client_sys = server->client;
481 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700482 if (server->flags & NFS_MOUNT_VER3) {
Andreas Gruenbacherb7fa0552005-06-22 17:16:27 +0000483#ifdef CONFIG_NFS_V3_ACL
484 if (!(server->flags & NFS_MOUNT_NOACL)) {
485 server->client_acl = rpc_bind_new_program(server->client, &nfsacl_program, 3);
486 /* No errors! Assume that Sun nfsacls are supported */
487 if (!IS_ERR(server->client_acl))
488 server->caps |= NFS_CAP_ACLS;
489 }
490#else
491 server->flags &= ~NFS_MOUNT_NOACL;
492#endif /* CONFIG_NFS_V3_ACL */
Andreas Gruenbacher055ffbe2005-06-22 17:16:27 +0000493 /*
494 * The VFS shouldn't apply the umask to mode bits. We will
495 * do so ourselves when necessary.
496 */
497 sb->s_flags |= MS_POSIXACL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700498 if (server->namelen == 0 || server->namelen > NFS3_MAXNAMLEN)
499 server->namelen = NFS3_MAXNAMLEN;
500 sb->s_time_gran = 1;
501 } else {
502 if (server->namelen == 0 || server->namelen > NFS2_MAXNAMLEN)
503 server->namelen = NFS2_MAXNAMLEN;
504 }
505
506 sb->s_op = &nfs_sops;
507 return nfs_sb_init(sb, authflavor);
508}
509
510static int
511nfs_statfs(struct super_block *sb, struct kstatfs *buf)
512{
513 struct nfs_server *server = NFS_SB(sb);
514 unsigned char blockbits;
515 unsigned long blockres;
516 struct nfs_fh *rootfh = NFS_FH(sb->s_root->d_inode);
517 struct nfs_fattr fattr;
518 struct nfs_fsstat res = {
519 .fattr = &fattr,
520 };
521 int error;
522
523 lock_kernel();
524
525 error = server->rpc_ops->statfs(server, rootfh, &res);
526 buf->f_type = NFS_SUPER_MAGIC;
527 if (error < 0)
528 goto out_err;
529
530 /*
531 * Current versions of glibc do not correctly handle the
532 * case where f_frsize != f_bsize. Eventually we want to
533 * report the value of wtmult in this field.
534 */
535 buf->f_frsize = sb->s_blocksize;
536
537 /*
538 * On most *nix systems, f_blocks, f_bfree, and f_bavail
539 * are reported in units of f_frsize. Linux hasn't had
540 * an f_frsize field in its statfs struct until recently,
541 * thus historically Linux's sys_statfs reports these
542 * fields in units of f_bsize.
543 */
544 buf->f_bsize = sb->s_blocksize;
545 blockbits = sb->s_blocksize_bits;
546 blockres = (1 << blockbits) - 1;
547 buf->f_blocks = (res.tbytes + blockres) >> blockbits;
548 buf->f_bfree = (res.fbytes + blockres) >> blockbits;
549 buf->f_bavail = (res.abytes + blockres) >> blockbits;
550
551 buf->f_files = res.tfiles;
552 buf->f_ffree = res.afiles;
553
554 buf->f_namelen = server->namelen;
555 out:
556 unlock_kernel();
557
558 return 0;
559
560 out_err:
561 printk(KERN_WARNING "nfs_statfs: statfs error = %d\n", -error);
562 buf->f_bsize = buf->f_blocks = buf->f_bfree = buf->f_bavail = -1;
563 goto out;
564
565}
566
567static int nfs_show_options(struct seq_file *m, struct vfsmount *mnt)
568{
569 static struct proc_nfs_info {
570 int flag;
571 char *str;
572 char *nostr;
573 } nfs_info[] = {
574 { NFS_MOUNT_SOFT, ",soft", ",hard" },
575 { NFS_MOUNT_INTR, ",intr", "" },
576 { NFS_MOUNT_POSIX, ",posix", "" },
577 { NFS_MOUNT_TCP, ",tcp", ",udp" },
578 { NFS_MOUNT_NOCTO, ",nocto", "" },
579 { NFS_MOUNT_NOAC, ",noac", "" },
580 { NFS_MOUNT_NONLM, ",nolock", ",lock" },
Andreas Gruenbacherb7fa0552005-06-22 17:16:27 +0000581 { NFS_MOUNT_NOACL, ",noacl", "" },
Linus Torvalds1da177e2005-04-16 15:20:36 -0700582 { 0, NULL, NULL }
583 };
584 struct proc_nfs_info *nfs_infop;
585 struct nfs_server *nfss = NFS_SB(mnt->mnt_sb);
586
587 seq_printf(m, ",v%d", nfss->rpc_ops->version);
588 seq_printf(m, ",rsize=%d", nfss->rsize);
589 seq_printf(m, ",wsize=%d", nfss->wsize);
590 if (nfss->acregmin != 3*HZ)
591 seq_printf(m, ",acregmin=%d", nfss->acregmin/HZ);
592 if (nfss->acregmax != 60*HZ)
593 seq_printf(m, ",acregmax=%d", nfss->acregmax/HZ);
594 if (nfss->acdirmin != 30*HZ)
595 seq_printf(m, ",acdirmin=%d", nfss->acdirmin/HZ);
596 if (nfss->acdirmax != 60*HZ)
597 seq_printf(m, ",acdirmax=%d", nfss->acdirmax/HZ);
598 for (nfs_infop = nfs_info; nfs_infop->flag; nfs_infop++) {
599 if (nfss->flags & nfs_infop->flag)
600 seq_puts(m, nfs_infop->str);
601 else
602 seq_puts(m, nfs_infop->nostr);
603 }
604 seq_puts(m, ",addr=");
605 seq_escape(m, nfss->hostname, " \t\n\\");
606 return 0;
607}
608
609/*
610 * Invalidate the local caches
611 */
612void
613nfs_zap_caches(struct inode *inode)
614{
615 struct nfs_inode *nfsi = NFS_I(inode);
616 int mode = inode->i_mode;
617
618 NFS_ATTRTIMEO(inode) = NFS_MINATTRTIMEO(inode);
619 NFS_ATTRTIMEO_UPDATE(inode) = jiffies;
620
621 memset(NFS_COOKIEVERF(inode), 0, sizeof(NFS_COOKIEVERF(inode)));
622 if (S_ISREG(mode) || S_ISDIR(mode) || S_ISLNK(mode))
Trond Myklebustfe51beec2005-06-22 17:16:30 +0000623 nfsi->flags |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_DATA|NFS_INO_INVALID_ACCESS|NFS_INO_INVALID_ACL|NFS_INO_REVAL_PAGECACHE;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700624 else
Trond Myklebustfe51beec2005-06-22 17:16:30 +0000625 nfsi->flags |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_ACCESS|NFS_INO_INVALID_ACL|NFS_INO_REVAL_PAGECACHE;
Trond Myklebustada70d92005-06-22 17:16:22 +0000626}
627
628static void nfs_zap_acl_cache(struct inode *inode)
629{
630 void (*clear_acl_cache)(struct inode *);
631
632 clear_acl_cache = NFS_PROTO(inode)->clear_acl_cache;
633 if (clear_acl_cache != NULL)
634 clear_acl_cache(inode);
635 NFS_I(inode)->flags &= ~NFS_INO_INVALID_ACL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700636}
637
638/*
639 * Invalidate, but do not unhash, the inode
640 */
641static void
642nfs_invalidate_inode(struct inode *inode)
643{
644 umode_t save_mode = inode->i_mode;
645
646 make_bad_inode(inode);
647 inode->i_mode = save_mode;
648 nfs_zap_caches(inode);
649}
650
651struct nfs_find_desc {
652 struct nfs_fh *fh;
653 struct nfs_fattr *fattr;
654};
655
656/*
657 * In NFSv3 we can have 64bit inode numbers. In order to support
658 * this, and re-exported directories (also seen in NFSv2)
659 * we are forced to allow 2 different inodes to have the same
660 * i_ino.
661 */
662static int
663nfs_find_actor(struct inode *inode, void *opaque)
664{
665 struct nfs_find_desc *desc = (struct nfs_find_desc *)opaque;
666 struct nfs_fh *fh = desc->fh;
667 struct nfs_fattr *fattr = desc->fattr;
668
669 if (NFS_FILEID(inode) != fattr->fileid)
670 return 0;
671 if (nfs_compare_fh(NFS_FH(inode), fh))
672 return 0;
673 if (is_bad_inode(inode) || NFS_STALE(inode))
674 return 0;
675 return 1;
676}
677
678static int
679nfs_init_locked(struct inode *inode, void *opaque)
680{
681 struct nfs_find_desc *desc = (struct nfs_find_desc *)opaque;
682 struct nfs_fattr *fattr = desc->fattr;
683
684 NFS_FILEID(inode) = fattr->fileid;
685 nfs_copy_fh(NFS_FH(inode), desc->fh);
686 return 0;
687}
688
689/* Don't use READDIRPLUS on directories that we believe are too large */
690#define NFS_LIMIT_READDIRPLUS (8*PAGE_SIZE)
691
692/*
693 * This is our front-end to iget that looks up inodes by file handle
694 * instead of inode number.
695 */
696struct inode *
697nfs_fhget(struct super_block *sb, struct nfs_fh *fh, struct nfs_fattr *fattr)
698{
699 struct nfs_find_desc desc = {
700 .fh = fh,
701 .fattr = fattr
702 };
703 struct inode *inode = NULL;
704 unsigned long hash;
705
706 if ((fattr->valid & NFS_ATTR_FATTR) == 0)
707 goto out_no_inode;
708
709 if (!fattr->nlink) {
710 printk("NFS: Buggy server - nlink == 0!\n");
711 goto out_no_inode;
712 }
713
714 hash = nfs_fattr_to_ino_t(fattr);
715
716 if (!(inode = iget5_locked(sb, hash, nfs_find_actor, nfs_init_locked, &desc)))
717 goto out_no_inode;
718
719 if (inode->i_state & I_NEW) {
720 struct nfs_inode *nfsi = NFS_I(inode);
721
722 /* We set i_ino for the few things that still rely on it,
723 * such as stat(2) */
724 inode->i_ino = hash;
725
726 /* We can't support update_atime(), since the server will reset it */
727 inode->i_flags |= S_NOATIME|S_NOCMTIME;
728 inode->i_mode = fattr->mode;
729 /* Why so? Because we want revalidate for devices/FIFOs, and
730 * that's precisely what we have in nfs_file_inode_operations.
731 */
J. Bruce Fields92cfc622005-06-22 17:16:22 +0000732 inode->i_op = NFS_SB(sb)->rpc_ops->file_inode_ops;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700733 if (S_ISREG(inode->i_mode)) {
734 inode->i_fop = &nfs_file_operations;
735 inode->i_data.a_ops = &nfs_file_aops;
736 inode->i_data.backing_dev_info = &NFS_SB(sb)->backing_dev_info;
737 } else if (S_ISDIR(inode->i_mode)) {
738 inode->i_op = NFS_SB(sb)->rpc_ops->dir_inode_ops;
739 inode->i_fop = &nfs_dir_operations;
740 if (nfs_server_capable(inode, NFS_CAP_READDIRPLUS)
741 && fattr->size <= NFS_LIMIT_READDIRPLUS)
742 NFS_FLAGS(inode) |= NFS_INO_ADVISE_RDPLUS;
743 } else if (S_ISLNK(inode->i_mode))
744 inode->i_op = &nfs_symlink_inode_operations;
745 else
746 init_special_inode(inode, inode->i_mode, fattr->rdev);
747
748 nfsi->read_cache_jiffies = fattr->timestamp;
749 inode->i_atime = fattr->atime;
750 inode->i_mtime = fattr->mtime;
751 inode->i_ctime = fattr->ctime;
752 if (fattr->valid & NFS_ATTR_FATTR_V4)
753 nfsi->change_attr = fattr->change_attr;
754 inode->i_size = nfs_size_to_loff_t(fattr->size);
755 inode->i_nlink = fattr->nlink;
756 inode->i_uid = fattr->uid;
757 inode->i_gid = fattr->gid;
758 if (fattr->valid & (NFS_ATTR_FATTR_V3 | NFS_ATTR_FATTR_V4)) {
759 /*
760 * report the blocks in 512byte units
761 */
762 inode->i_blocks = nfs_calc_block_size(fattr->du.nfs3.used);
763 inode->i_blksize = inode->i_sb->s_blocksize;
764 } else {
765 inode->i_blocks = fattr->du.nfs2.blocks;
766 inode->i_blksize = fattr->du.nfs2.blocksize;
767 }
768 nfsi->attrtimeo = NFS_MINATTRTIMEO(inode);
769 nfsi->attrtimeo_timestamp = jiffies;
770 memset(nfsi->cookieverf, 0, sizeof(nfsi->cookieverf));
771 nfsi->cache_access.cred = NULL;
772
773 unlock_new_inode(inode);
774 } else
775 nfs_refresh_inode(inode, fattr);
776 dprintk("NFS: nfs_fhget(%s/%Ld ct=%d)\n",
777 inode->i_sb->s_id,
778 (long long)NFS_FILEID(inode),
779 atomic_read(&inode->i_count));
780
781out:
782 return inode;
783
784out_no_inode:
785 printk("nfs_fhget: iget failed\n");
786 goto out;
787}
788
789#define NFS_VALID_ATTRS (ATTR_MODE|ATTR_UID|ATTR_GID|ATTR_SIZE|ATTR_ATIME|ATTR_ATIME_SET|ATTR_MTIME|ATTR_MTIME_SET)
790
791int
792nfs_setattr(struct dentry *dentry, struct iattr *attr)
793{
794 struct inode *inode = dentry->d_inode;
795 struct nfs_fattr fattr;
796 int error;
797
798 if (attr->ia_valid & ATTR_SIZE) {
799 if (!S_ISREG(inode->i_mode) || attr->ia_size == i_size_read(inode))
800 attr->ia_valid &= ~ATTR_SIZE;
801 }
802
803 /* Optimization: if the end result is no change, don't RPC */
804 attr->ia_valid &= NFS_VALID_ATTRS;
805 if (attr->ia_valid == 0)
806 return 0;
807
808 lock_kernel();
809 nfs_begin_data_update(inode);
810 /* Write all dirty data if we're changing file permissions or size */
811 if ((attr->ia_valid & (ATTR_MODE|ATTR_UID|ATTR_GID|ATTR_SIZE)) != 0) {
812 if (filemap_fdatawrite(inode->i_mapping) == 0)
813 filemap_fdatawait(inode->i_mapping);
814 nfs_wb_all(inode);
815 }
816 error = NFS_PROTO(inode)->setattr(dentry, &fattr, attr);
Trond Myklebust65e43082005-08-16 11:49:44 -0400817 if (error == 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700818 nfs_refresh_inode(inode, &fattr);
Trond Myklebust65e43082005-08-16 11:49:44 -0400819 nfs_end_data_update(inode);
820 unlock_kernel();
821 return error;
822}
823
824/**
825 * nfs_setattr_update_inode - Update inode metadata after a setattr call.
826 * @inode: pointer to struct inode
827 * @attr: pointer to struct iattr
828 *
829 * Note: we do this in the *proc.c in order to ensure that
830 * it works for things like exclusive creates too.
831 */
832void nfs_setattr_update_inode(struct inode *inode, struct iattr *attr)
833{
834 if ((attr->ia_valid & (ATTR_MODE|ATTR_UID|ATTR_GID)) != 0) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700835 if ((attr->ia_valid & ATTR_MODE) != 0) {
Trond Myklebust65e43082005-08-16 11:49:44 -0400836 int mode = attr->ia_mode & S_IALLUGO;
837 mode |= inode->i_mode & ~S_IALLUGO;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700838 inode->i_mode = mode;
839 }
840 if ((attr->ia_valid & ATTR_UID) != 0)
841 inode->i_uid = attr->ia_uid;
842 if ((attr->ia_valid & ATTR_GID) != 0)
843 inode->i_gid = attr->ia_gid;
Trond Myklebustada70d92005-06-22 17:16:22 +0000844 NFS_FLAGS(inode) |= NFS_INO_INVALID_ACCESS|NFS_INO_INVALID_ACL;
Trond Myklebust65e43082005-08-16 11:49:44 -0400845 }
846 if ((attr->ia_valid & ATTR_SIZE) != 0) {
847 inode->i_size = attr->ia_size;
848 vmtruncate(inode, attr->ia_size);
849 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700850}
851
852/*
853 * Wait for the inode to get unlocked.
854 * (Used for NFS_INO_LOCKED and NFS_INO_REVALIDATING).
855 */
856static int
857nfs_wait_on_inode(struct inode *inode, int flag)
858{
859 struct rpc_clnt *clnt = NFS_CLIENT(inode);
860 struct nfs_inode *nfsi = NFS_I(inode);
861
862 int error;
863 if (!(NFS_FLAGS(inode) & flag))
864 return 0;
865 atomic_inc(&inode->i_count);
866 error = nfs_wait_event(clnt, nfsi->nfs_i_wait,
867 !(NFS_FLAGS(inode) & flag));
868 iput(inode);
869 return error;
870}
871
872int nfs_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
873{
874 struct inode *inode = dentry->d_inode;
875 struct nfs_inode *nfsi = NFS_I(inode);
876 int need_atime = nfsi->flags & NFS_INO_INVALID_ATIME;
877 int err;
878
879 if (__IS_FLG(inode, MS_NOATIME))
880 need_atime = 0;
881 else if (__IS_FLG(inode, MS_NODIRATIME) && S_ISDIR(inode->i_mode))
882 need_atime = 0;
883 /* We may force a getattr if the user cares about atime */
884 if (need_atime)
885 err = __nfs_revalidate_inode(NFS_SERVER(inode), inode);
886 else
887 err = nfs_revalidate_inode(NFS_SERVER(inode), inode);
888 if (!err)
889 generic_fillattr(inode, stat);
890 return err;
891}
892
893struct nfs_open_context *alloc_nfs_open_context(struct dentry *dentry, struct rpc_cred *cred)
894{
895 struct nfs_open_context *ctx;
896
897 ctx = (struct nfs_open_context *)kmalloc(sizeof(*ctx), GFP_KERNEL);
898 if (ctx != NULL) {
899 atomic_set(&ctx->count, 1);
900 ctx->dentry = dget(dentry);
901 ctx->cred = get_rpccred(cred);
902 ctx->state = NULL;
903 ctx->lockowner = current->files;
904 ctx->error = 0;
Olivier Galibert00a92642005-06-22 17:16:29 +0000905 ctx->dir_cookie = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700906 }
907 return ctx;
908}
909
910struct nfs_open_context *get_nfs_open_context(struct nfs_open_context *ctx)
911{
912 if (ctx != NULL)
913 atomic_inc(&ctx->count);
914 return ctx;
915}
916
917void put_nfs_open_context(struct nfs_open_context *ctx)
918{
919 if (atomic_dec_and_test(&ctx->count)) {
920 if (!list_empty(&ctx->list)) {
921 struct inode *inode = ctx->dentry->d_inode;
922 spin_lock(&inode->i_lock);
923 list_del(&ctx->list);
924 spin_unlock(&inode->i_lock);
925 }
926 if (ctx->state != NULL)
927 nfs4_close_state(ctx->state, ctx->mode);
928 if (ctx->cred != NULL)
929 put_rpccred(ctx->cred);
930 dput(ctx->dentry);
931 kfree(ctx);
932 }
933}
934
935/*
936 * Ensure that mmap has a recent RPC credential for use when writing out
937 * shared pages
938 */
939void nfs_file_set_open_context(struct file *filp, struct nfs_open_context *ctx)
940{
941 struct inode *inode = filp->f_dentry->d_inode;
942 struct nfs_inode *nfsi = NFS_I(inode);
943
944 filp->private_data = get_nfs_open_context(ctx);
945 spin_lock(&inode->i_lock);
946 list_add(&ctx->list, &nfsi->open_files);
947 spin_unlock(&inode->i_lock);
948}
949
950struct nfs_open_context *nfs_find_open_context(struct inode *inode, int mode)
951{
952 struct nfs_inode *nfsi = NFS_I(inode);
953 struct nfs_open_context *pos, *ctx = NULL;
954
955 spin_lock(&inode->i_lock);
956 list_for_each_entry(pos, &nfsi->open_files, list) {
957 if ((pos->mode & mode) == mode) {
958 ctx = get_nfs_open_context(pos);
959 break;
960 }
961 }
962 spin_unlock(&inode->i_lock);
963 return ctx;
964}
965
966void nfs_file_clear_open_context(struct file *filp)
967{
968 struct inode *inode = filp->f_dentry->d_inode;
969 struct nfs_open_context *ctx = (struct nfs_open_context *)filp->private_data;
970
971 if (ctx) {
972 filp->private_data = NULL;
973 spin_lock(&inode->i_lock);
974 list_move_tail(&ctx->list, &NFS_I(inode)->open_files);
975 spin_unlock(&inode->i_lock);
976 put_nfs_open_context(ctx);
977 }
978}
979
980/*
981 * These allocate and release file read/write context information.
982 */
983int nfs_open(struct inode *inode, struct file *filp)
984{
985 struct nfs_open_context *ctx;
986 struct rpc_cred *cred;
987
988 cred = rpcauth_lookupcred(NFS_CLIENT(inode)->cl_auth, 0);
989 if (IS_ERR(cred))
990 return PTR_ERR(cred);
991 ctx = alloc_nfs_open_context(filp->f_dentry, cred);
992 put_rpccred(cred);
993 if (ctx == NULL)
994 return -ENOMEM;
995 ctx->mode = filp->f_mode;
996 nfs_file_set_open_context(filp, ctx);
997 put_nfs_open_context(ctx);
998 if ((filp->f_mode & FMODE_WRITE) != 0)
999 nfs_begin_data_update(inode);
1000 return 0;
1001}
1002
1003int nfs_release(struct inode *inode, struct file *filp)
1004{
1005 if ((filp->f_mode & FMODE_WRITE) != 0)
1006 nfs_end_data_update(inode);
1007 nfs_file_clear_open_context(filp);
1008 return 0;
1009}
1010
1011/*
1012 * This function is called whenever some part of NFS notices that
1013 * the cached attributes have to be refreshed.
1014 */
1015int
1016__nfs_revalidate_inode(struct nfs_server *server, struct inode *inode)
1017{
1018 int status = -ESTALE;
1019 struct nfs_fattr fattr;
1020 struct nfs_inode *nfsi = NFS_I(inode);
1021 unsigned long verifier;
1022 unsigned int flags;
1023
1024 dfprintk(PAGECACHE, "NFS: revalidating (%s/%Ld)\n",
1025 inode->i_sb->s_id, (long long)NFS_FILEID(inode));
1026
1027 lock_kernel();
1028 if (!inode || is_bad_inode(inode))
1029 goto out_nowait;
1030 if (NFS_STALE(inode))
1031 goto out_nowait;
1032
1033 while (NFS_REVALIDATING(inode)) {
1034 status = nfs_wait_on_inode(inode, NFS_INO_REVALIDATING);
1035 if (status < 0)
1036 goto out_nowait;
1037 if (NFS_ATTRTIMEO(inode) == 0)
1038 continue;
1039 if (NFS_FLAGS(inode) & (NFS_INO_INVALID_ATTR|NFS_INO_INVALID_DATA|NFS_INO_INVALID_ATIME))
1040 continue;
1041 status = NFS_STALE(inode) ? -ESTALE : 0;
1042 goto out_nowait;
1043 }
1044 NFS_FLAGS(inode) |= NFS_INO_REVALIDATING;
1045
1046 /* Protect against RPC races by saving the change attribute */
1047 verifier = nfs_save_change_attribute(inode);
1048 status = NFS_PROTO(inode)->getattr(server, NFS_FH(inode), &fattr);
1049 if (status != 0) {
1050 dfprintk(PAGECACHE, "nfs_revalidate_inode: (%s/%Ld) getattr failed, error=%d\n",
1051 inode->i_sb->s_id,
1052 (long long)NFS_FILEID(inode), status);
1053 if (status == -ESTALE) {
1054 nfs_zap_caches(inode);
1055 if (!S_ISDIR(inode->i_mode))
1056 NFS_FLAGS(inode) |= NFS_INO_STALE;
1057 }
1058 goto out;
1059 }
1060
1061 status = nfs_update_inode(inode, &fattr, verifier);
1062 if (status) {
1063 dfprintk(PAGECACHE, "nfs_revalidate_inode: (%s/%Ld) refresh failed, error=%d\n",
1064 inode->i_sb->s_id,
1065 (long long)NFS_FILEID(inode), status);
1066 goto out;
1067 }
1068 flags = nfsi->flags;
Trond Myklebustfe51beec2005-06-22 17:16:30 +00001069 nfsi->flags &= ~NFS_INO_REVAL_PAGECACHE;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001070 /*
1071 * We may need to keep the attributes marked as invalid if
1072 * we raced with nfs_end_attr_update().
1073 */
1074 if (verifier == nfsi->cache_change_attribute)
1075 nfsi->flags &= ~(NFS_INO_INVALID_ATTR|NFS_INO_INVALID_ATIME);
1076 /* Do the page cache invalidation */
Trond Myklebust7d52e862005-06-22 17:16:30 +00001077 nfs_revalidate_mapping(inode, inode->i_mapping);
Trond Myklebustada70d92005-06-22 17:16:22 +00001078 if (flags & NFS_INO_INVALID_ACL)
1079 nfs_zap_acl_cache(inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001080 dfprintk(PAGECACHE, "NFS: (%s/%Ld) revalidation complete\n",
1081 inode->i_sb->s_id,
1082 (long long)NFS_FILEID(inode));
1083
1084out:
1085 NFS_FLAGS(inode) &= ~NFS_INO_REVALIDATING;
1086 wake_up(&nfsi->nfs_i_wait);
1087 out_nowait:
1088 unlock_kernel();
1089 return status;
1090}
1091
1092int nfs_attribute_timeout(struct inode *inode)
1093{
1094 struct nfs_inode *nfsi = NFS_I(inode);
1095
1096 if (nfs_have_delegation(inode, FMODE_READ))
1097 return 0;
1098 return time_after(jiffies, nfsi->read_cache_jiffies+nfsi->attrtimeo);
1099}
1100
1101/**
1102 * nfs_revalidate_inode - Revalidate the inode attributes
1103 * @server - pointer to nfs_server struct
1104 * @inode - pointer to inode struct
1105 *
1106 * Updates inode attribute information by retrieving the data from the server.
1107 */
1108int nfs_revalidate_inode(struct nfs_server *server, struct inode *inode)
1109{
1110 if (!(NFS_FLAGS(inode) & (NFS_INO_INVALID_ATTR|NFS_INO_INVALID_DATA))
1111 && !nfs_attribute_timeout(inode))
1112 return NFS_STALE(inode) ? -ESTALE : 0;
1113 return __nfs_revalidate_inode(server, inode);
1114}
1115
1116/**
Trond Myklebust7d52e862005-06-22 17:16:30 +00001117 * nfs_revalidate_mapping - Revalidate the pagecache
1118 * @inode - pointer to host inode
1119 * @mapping - pointer to mapping
1120 */
1121void nfs_revalidate_mapping(struct inode *inode, struct address_space *mapping)
1122{
1123 struct nfs_inode *nfsi = NFS_I(inode);
1124
1125 if (nfsi->flags & NFS_INO_INVALID_DATA) {
1126 if (S_ISREG(inode->i_mode)) {
1127 if (filemap_fdatawrite(mapping) == 0)
1128 filemap_fdatawait(mapping);
1129 nfs_wb_all(inode);
1130 }
1131 invalidate_inode_pages2(mapping);
1132 nfsi->flags &= ~NFS_INO_INVALID_DATA;
1133 if (S_ISDIR(inode->i_mode)) {
1134 memset(nfsi->cookieverf, 0, sizeof(nfsi->cookieverf));
1135 /* This ensures we revalidate child dentries */
1136 nfsi->cache_change_attribute++;
1137 }
1138 dfprintk(PAGECACHE, "NFS: (%s/%Ld) data cache invalidated\n",
1139 inode->i_sb->s_id,
1140 (long long)NFS_FILEID(inode));
1141 }
1142}
1143
1144/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001145 * nfs_begin_data_update
1146 * @inode - pointer to inode
1147 * Declare that a set of operations will update file data on the server
1148 */
1149void nfs_begin_data_update(struct inode *inode)
1150{
1151 atomic_inc(&NFS_I(inode)->data_updates);
1152}
1153
1154/**
1155 * nfs_end_data_update
1156 * @inode - pointer to inode
1157 * Declare end of the operations that will update file data
1158 * This will mark the inode as immediately needing revalidation
1159 * of its attribute cache.
1160 */
1161void nfs_end_data_update(struct inode *inode)
1162{
1163 struct nfs_inode *nfsi = NFS_I(inode);
1164
1165 if (!nfs_have_delegation(inode, FMODE_READ)) {
1166 /* Mark the attribute cache for revalidation */
1167 nfsi->flags |= NFS_INO_INVALID_ATTR;
1168 /* Directories and symlinks: invalidate page cache too */
1169 if (S_ISDIR(inode->i_mode) || S_ISLNK(inode->i_mode))
1170 nfsi->flags |= NFS_INO_INVALID_DATA;
1171 }
1172 nfsi->cache_change_attribute ++;
1173 atomic_dec(&nfsi->data_updates);
1174}
1175
1176/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001177 * nfs_refresh_inode - verify consistency of the inode attribute cache
1178 * @inode - pointer to inode
1179 * @fattr - updated attributes
1180 *
1181 * Verifies the attribute cache. If we have just changed the attributes,
1182 * so that fattr carries weak cache consistency data, then it may
1183 * also update the ctime/mtime/change_attribute.
1184 */
1185int nfs_refresh_inode(struct inode *inode, struct nfs_fattr *fattr)
1186{
1187 struct nfs_inode *nfsi = NFS_I(inode);
1188 loff_t cur_size, new_isize;
1189 int data_unstable;
1190
1191 /* Do we hold a delegation? */
1192 if (nfs_have_delegation(inode, FMODE_READ))
1193 return 0;
1194
1195 /* Are we in the process of updating data on the server? */
1196 data_unstable = nfs_caches_unstable(inode);
1197
1198 if (fattr->valid & NFS_ATTR_FATTR_V4) {
1199 if ((fattr->valid & NFS_ATTR_PRE_CHANGE) != 0
1200 && nfsi->change_attr == fattr->pre_change_attr)
1201 nfsi->change_attr = fattr->change_attr;
Trond Myklebustfe51beec2005-06-22 17:16:30 +00001202 if (nfsi->change_attr != fattr->change_attr) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001203 nfsi->flags |= NFS_INO_INVALID_ATTR;
Trond Myklebustfe51beec2005-06-22 17:16:30 +00001204 if (!data_unstable)
1205 nfsi->flags |= NFS_INO_REVAL_PAGECACHE;
1206 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001207 }
1208
1209 if ((fattr->valid & NFS_ATTR_FATTR) == 0)
1210 return 0;
1211
1212 /* Has the inode gone and changed behind our back? */
1213 if (nfsi->fileid != fattr->fileid
1214 || (inode->i_mode & S_IFMT) != (fattr->mode & S_IFMT))
1215 return -EIO;
1216
1217 cur_size = i_size_read(inode);
1218 new_isize = nfs_size_to_loff_t(fattr->size);
1219
1220 /* If we have atomic WCC data, we may update some attributes */
1221 if ((fattr->valid & NFS_ATTR_WCC) != 0) {
1222 if (timespec_equal(&inode->i_ctime, &fattr->pre_ctime))
1223 memcpy(&inode->i_ctime, &fattr->ctime, sizeof(inode->i_ctime));
1224 if (timespec_equal(&inode->i_mtime, &fattr->pre_mtime))
1225 memcpy(&inode->i_mtime, &fattr->mtime, sizeof(inode->i_mtime));
1226 }
1227
1228 /* Verify a few of the more important attributes */
Trond Myklebustfe51beec2005-06-22 17:16:30 +00001229 if (!timespec_equal(&inode->i_mtime, &fattr->mtime)) {
Trond Myklebust951a1432005-06-22 17:16:30 +00001230 nfsi->flags |= NFS_INO_INVALID_ATTR;
Trond Myklebustfe51beec2005-06-22 17:16:30 +00001231 if (!data_unstable)
1232 nfsi->flags |= NFS_INO_REVAL_PAGECACHE;
1233 }
1234 if (cur_size != new_isize) {
1235 nfsi->flags |= NFS_INO_INVALID_ATTR;
1236 if (nfsi->npages == 0)
1237 nfsi->flags |= NFS_INO_REVAL_PAGECACHE;
1238 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001239
1240 /* Have any file permissions changed? */
1241 if ((inode->i_mode & S_IALLUGO) != (fattr->mode & S_IALLUGO)
1242 || inode->i_uid != fattr->uid
1243 || inode->i_gid != fattr->gid)
Trond Myklebustada70d92005-06-22 17:16:22 +00001244 nfsi->flags |= NFS_INO_INVALID_ATTR | NFS_INO_INVALID_ACCESS | NFS_INO_INVALID_ACL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001245
1246 /* Has the link count changed? */
1247 if (inode->i_nlink != fattr->nlink)
1248 nfsi->flags |= NFS_INO_INVALID_ATTR;
1249
1250 if (!timespec_equal(&inode->i_atime, &fattr->atime))
1251 nfsi->flags |= NFS_INO_INVALID_ATIME;
1252
1253 nfsi->read_cache_jiffies = fattr->timestamp;
1254 return 0;
1255}
1256
1257/*
1258 * Many nfs protocol calls return the new file attributes after
1259 * an operation. Here we update the inode to reflect the state
1260 * of the server's inode.
1261 *
1262 * This is a bit tricky because we have to make sure all dirty pages
1263 * have been sent off to the server before calling invalidate_inode_pages.
1264 * To make sure no other process adds more write requests while we try
1265 * our best to flush them, we make them sleep during the attribute refresh.
1266 *
1267 * A very similar scenario holds for the dir cache.
1268 */
1269static int nfs_update_inode(struct inode *inode, struct nfs_fattr *fattr, unsigned long verifier)
1270{
1271 struct nfs_inode *nfsi = NFS_I(inode);
Trond Myklebust951a1432005-06-22 17:16:30 +00001272 loff_t cur_isize, new_isize;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001273 unsigned int invalid = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001274 int data_unstable;
1275
1276 dfprintk(VFS, "NFS: %s(%s/%ld ct=%d info=0x%x)\n",
1277 __FUNCTION__, inode->i_sb->s_id, inode->i_ino,
1278 atomic_read(&inode->i_count), fattr->valid);
1279
1280 if ((fattr->valid & NFS_ATTR_FATTR) == 0)
1281 return 0;
1282
1283 if (nfsi->fileid != fattr->fileid) {
1284 printk(KERN_ERR "%s: inode number mismatch\n"
1285 "expected (%s/0x%Lx), got (%s/0x%Lx)\n",
1286 __FUNCTION__,
1287 inode->i_sb->s_id, (long long)nfsi->fileid,
1288 inode->i_sb->s_id, (long long)fattr->fileid);
1289 goto out_err;
1290 }
1291
1292 /*
1293 * Make sure the inode's type hasn't changed.
1294 */
1295 if ((inode->i_mode & S_IFMT) != (fattr->mode & S_IFMT))
1296 goto out_changed;
1297
1298 /*
1299 * Update the read time so we don't revalidate too often.
1300 */
1301 nfsi->read_cache_jiffies = fattr->timestamp;
1302
1303 /* Are we racing with known updates of the metadata on the server? */
1304 data_unstable = ! nfs_verify_change_attribute(inode, verifier);
1305
Trond Myklebust951a1432005-06-22 17:16:30 +00001306 /* Check if our cached file size is stale */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001307 new_isize = nfs_size_to_loff_t(fattr->size);
1308 cur_isize = i_size_read(inode);
Trond Myklebust951a1432005-06-22 17:16:30 +00001309 if (new_isize != cur_isize) {
1310 /* Do we perhaps have any outstanding writes? */
1311 if (nfsi->npages == 0) {
1312 /* No, but did we race with nfs_end_data_update()? */
1313 if (verifier == nfsi->cache_change_attribute) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001314 inode->i_size = new_isize;
Trond Myklebust951a1432005-06-22 17:16:30 +00001315 invalid |= NFS_INO_INVALID_DATA;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001316 }
Trond Myklebust951a1432005-06-22 17:16:30 +00001317 invalid |= NFS_INO_INVALID_ATTR;
1318 } else if (new_isize > cur_isize) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001319 inode->i_size = new_isize;
1320 invalid |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_DATA;
1321 }
Trond Myklebust951a1432005-06-22 17:16:30 +00001322 dprintk("NFS: isize change on server for file %s/%ld\n",
1323 inode->i_sb->s_id, inode->i_ino);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001324 }
1325
Trond Myklebust951a1432005-06-22 17:16:30 +00001326 /* Check if the mtime agrees */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001327 if (!timespec_equal(&inode->i_mtime, &fattr->mtime)) {
1328 memcpy(&inode->i_mtime, &fattr->mtime, sizeof(inode->i_mtime));
Trond Myklebust951a1432005-06-22 17:16:30 +00001329 dprintk("NFS: mtime change on server for file %s/%ld\n",
1330 inode->i_sb->s_id, inode->i_ino);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001331 if (!data_unstable)
1332 invalid |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_DATA;
1333 }
1334
1335 if ((fattr->valid & NFS_ATTR_FATTR_V4)
1336 && nfsi->change_attr != fattr->change_attr) {
Trond Myklebust951a1432005-06-22 17:16:30 +00001337 dprintk("NFS: change_attr change on server for file %s/%ld\n",
Linus Torvalds1da177e2005-04-16 15:20:36 -07001338 inode->i_sb->s_id, inode->i_ino);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001339 nfsi->change_attr = fattr->change_attr;
1340 if (!data_unstable)
Trond Myklebustada70d92005-06-22 17:16:22 +00001341 invalid |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_DATA|NFS_INO_INVALID_ACCESS|NFS_INO_INVALID_ACL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001342 }
1343
Trond Myklebustada70d92005-06-22 17:16:22 +00001344 /* If ctime has changed we should definitely clear access+acl caches */
1345 if (!timespec_equal(&inode->i_ctime, &fattr->ctime)) {
1346 if (!data_unstable)
1347 invalid |= NFS_INO_INVALID_ACCESS|NFS_INO_INVALID_ACL;
1348 memcpy(&inode->i_ctime, &fattr->ctime, sizeof(inode->i_ctime));
1349 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001350 memcpy(&inode->i_atime, &fattr->atime, sizeof(inode->i_atime));
1351
1352 if ((inode->i_mode & S_IALLUGO) != (fattr->mode & S_IALLUGO) ||
1353 inode->i_uid != fattr->uid ||
1354 inode->i_gid != fattr->gid)
Trond Myklebustada70d92005-06-22 17:16:22 +00001355 invalid |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_ACCESS|NFS_INO_INVALID_ACL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001356
1357 inode->i_mode = fattr->mode;
1358 inode->i_nlink = fattr->nlink;
1359 inode->i_uid = fattr->uid;
1360 inode->i_gid = fattr->gid;
1361
1362 if (fattr->valid & (NFS_ATTR_FATTR_V3 | NFS_ATTR_FATTR_V4)) {
1363 /*
1364 * report the blocks in 512byte units
1365 */
1366 inode->i_blocks = nfs_calc_block_size(fattr->du.nfs3.used);
1367 inode->i_blksize = inode->i_sb->s_blocksize;
1368 } else {
1369 inode->i_blocks = fattr->du.nfs2.blocks;
1370 inode->i_blksize = fattr->du.nfs2.blocksize;
1371 }
1372
1373 /* Update attrtimeo value if we're out of the unstable period */
1374 if (invalid & NFS_INO_INVALID_ATTR) {
1375 nfsi->attrtimeo = NFS_MINATTRTIMEO(inode);
1376 nfsi->attrtimeo_timestamp = jiffies;
1377 } else if (time_after(jiffies, nfsi->attrtimeo_timestamp+nfsi->attrtimeo)) {
1378 if ((nfsi->attrtimeo <<= 1) > NFS_MAXATTRTIMEO(inode))
1379 nfsi->attrtimeo = NFS_MAXATTRTIMEO(inode);
1380 nfsi->attrtimeo_timestamp = jiffies;
1381 }
1382 /* Don't invalidate the data if we were to blame */
1383 if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode)
1384 || S_ISLNK(inode->i_mode)))
1385 invalid &= ~NFS_INO_INVALID_DATA;
1386 if (!nfs_have_delegation(inode, FMODE_READ))
1387 nfsi->flags |= invalid;
1388
1389 return 0;
1390 out_changed:
1391 /*
1392 * Big trouble! The inode has become a different object.
1393 */
1394#ifdef NFS_PARANOIA
1395 printk(KERN_DEBUG "%s: inode %ld mode changed, %07o to %07o\n",
1396 __FUNCTION__, inode->i_ino, inode->i_mode, fattr->mode);
1397#endif
1398 /*
1399 * No need to worry about unhashing the dentry, as the
1400 * lookup validation will know that the inode is bad.
1401 * (But we fall through to invalidate the caches.)
1402 */
1403 nfs_invalidate_inode(inode);
1404 out_err:
1405 NFS_FLAGS(inode) |= NFS_INO_STALE;
1406 return -ESTALE;
1407}
1408
1409/*
1410 * File system information
1411 */
1412
1413static int nfs_set_super(struct super_block *s, void *data)
1414{
1415 s->s_fs_info = data;
1416 return set_anon_super(s, data);
1417}
1418
1419static int nfs_compare_super(struct super_block *sb, void *data)
1420{
1421 struct nfs_server *server = data;
1422 struct nfs_server *old = NFS_SB(sb);
1423
1424 if (old->addr.sin_addr.s_addr != server->addr.sin_addr.s_addr)
1425 return 0;
1426 if (old->addr.sin_port != server->addr.sin_port)
1427 return 0;
1428 return !nfs_compare_fh(&old->fh, &server->fh);
1429}
1430
1431static struct super_block *nfs_get_sb(struct file_system_type *fs_type,
1432 int flags, const char *dev_name, void *raw_data)
1433{
1434 int error;
Trond Myklebust9085bbc2005-06-22 17:16:20 +00001435 struct nfs_server *server = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001436 struct super_block *s;
1437 struct nfs_fh *root;
1438 struct nfs_mount_data *data = raw_data;
1439
Trond Myklebust9085bbc2005-06-22 17:16:20 +00001440 s = ERR_PTR(-EINVAL);
1441 if (data == NULL) {
1442 dprintk("%s: missing data argument\n", __FUNCTION__);
1443 goto out_err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001444 }
Trond Myklebust9085bbc2005-06-22 17:16:20 +00001445 if (data->version <= 0 || data->version > NFS_MOUNT_VERSION) {
1446 dprintk("%s: bad mount version\n", __FUNCTION__);
1447 goto out_err;
1448 }
1449 switch (data->version) {
1450 case 1:
1451 data->namlen = 0;
1452 case 2:
1453 data->bsize = 0;
1454 case 3:
1455 if (data->flags & NFS_MOUNT_VER3) {
1456 dprintk("%s: mount structure version %d does not support NFSv3\n",
1457 __FUNCTION__,
1458 data->version);
1459 goto out_err;
1460 }
1461 data->root.size = NFS2_FHSIZE;
1462 memcpy(data->root.data, data->old_root.data, NFS2_FHSIZE);
1463 case 4:
1464 if (data->flags & NFS_MOUNT_SECFLAVOUR) {
1465 dprintk("%s: mount structure version %d does not support strong security\n",
1466 __FUNCTION__,
1467 data->version);
1468 goto out_err;
1469 }
1470 case 5:
1471 memset(data->context, 0, sizeof(data->context));
1472 }
1473#ifndef CONFIG_NFS_V3
1474 /* If NFSv3 is not compiled in, return -EPROTONOSUPPORT */
1475 s = ERR_PTR(-EPROTONOSUPPORT);
1476 if (data->flags & NFS_MOUNT_VER3) {
1477 dprintk("%s: NFSv3 not compiled into kernel\n", __FUNCTION__);
1478 goto out_err;
1479 }
1480#endif /* CONFIG_NFS_V3 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001481
Trond Myklebust9085bbc2005-06-22 17:16:20 +00001482 s = ERR_PTR(-ENOMEM);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001483 server = kmalloc(sizeof(struct nfs_server), GFP_KERNEL);
1484 if (!server)
Trond Myklebust9085bbc2005-06-22 17:16:20 +00001485 goto out_err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001486 memset(server, 0, sizeof(struct nfs_server));
1487 /* Zero out the NFS state stuff */
1488 init_nfsv4_state(server);
Andreas Gruenbacherb7fa0552005-06-22 17:16:27 +00001489 server->client = server->client_sys = server->client_acl = ERR_PTR(-EINVAL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001490
Linus Torvalds1da177e2005-04-16 15:20:36 -07001491 root = &server->fh;
1492 if (data->flags & NFS_MOUNT_VER3)
1493 root->size = data->root.size;
1494 else
1495 root->size = NFS2_FHSIZE;
Trond Myklebust9085bbc2005-06-22 17:16:20 +00001496 s = ERR_PTR(-EINVAL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001497 if (root->size > sizeof(root->data)) {
Trond Myklebust9085bbc2005-06-22 17:16:20 +00001498 dprintk("%s: invalid root filehandle\n", __FUNCTION__);
1499 goto out_err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001500 }
1501 memcpy(root->data, data->root.data, root->size);
1502
1503 /* We now require that the mount process passes the remote address */
1504 memcpy(&server->addr, &data->addr, sizeof(server->addr));
1505 if (server->addr.sin_addr.s_addr == INADDR_ANY) {
Trond Myklebust9085bbc2005-06-22 17:16:20 +00001506 dprintk("%s: mount program didn't pass remote address!\n",
1507 __FUNCTION__);
1508 goto out_err;
1509 }
1510
1511 /* Fire up rpciod if not yet running */
1512 s = ERR_PTR(rpciod_up());
1513 if (IS_ERR(s)) {
1514 dprintk("%s: couldn't start rpciod! Error = %ld\n",
1515 __FUNCTION__, PTR_ERR(s));
1516 goto out_err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001517 }
1518
1519 s = sget(fs_type, nfs_compare_super, nfs_set_super, server);
Trond Myklebust9085bbc2005-06-22 17:16:20 +00001520 if (IS_ERR(s) || s->s_root)
1521 goto out_rpciod_down;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001522
1523 s->s_flags = flags;
1524
Linus Torvalds1da177e2005-04-16 15:20:36 -07001525 error = nfs_fill_super(s, data, flags & MS_VERBOSE ? 1 : 0);
1526 if (error) {
1527 up_write(&s->s_umount);
1528 deactivate_super(s);
1529 return ERR_PTR(error);
1530 }
1531 s->s_flags |= MS_ACTIVE;
1532 return s;
Trond Myklebust9085bbc2005-06-22 17:16:20 +00001533out_rpciod_down:
1534 rpciod_down();
1535out_err:
1536 kfree(server);
1537 return s;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001538}
1539
1540static void nfs_kill_super(struct super_block *s)
1541{
1542 struct nfs_server *server = NFS_SB(s);
1543
1544 kill_anon_super(s);
1545
J. Bruce Fields6a192752005-06-22 17:16:23 +00001546 if (!IS_ERR(server->client))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001547 rpc_shutdown_client(server->client);
J. Bruce Fields6a192752005-06-22 17:16:23 +00001548 if (!IS_ERR(server->client_sys))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001549 rpc_shutdown_client(server->client_sys);
Andreas Gruenbacherb7fa0552005-06-22 17:16:27 +00001550 if (!IS_ERR(server->client_acl))
1551 rpc_shutdown_client(server->client_acl);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001552
1553 if (!(server->flags & NFS_MOUNT_NONLM))
1554 lockd_down(); /* release rpc.lockd */
1555
1556 rpciod_down(); /* release rpciod */
1557
1558 if (server->hostname != NULL)
1559 kfree(server->hostname);
1560 kfree(server);
1561}
1562
1563static struct file_system_type nfs_fs_type = {
1564 .owner = THIS_MODULE,
1565 .name = "nfs",
1566 .get_sb = nfs_get_sb,
1567 .kill_sb = nfs_kill_super,
1568 .fs_flags = FS_ODD_RENAME|FS_REVAL_DOT|FS_BINARY_MOUNTDATA,
1569};
1570
1571#ifdef CONFIG_NFS_V4
1572
1573static void nfs4_clear_inode(struct inode *);
1574
1575
1576static struct super_operations nfs4_sops = {
1577 .alloc_inode = nfs_alloc_inode,
1578 .destroy_inode = nfs_destroy_inode,
1579 .write_inode = nfs_write_inode,
1580 .delete_inode = nfs_delete_inode,
1581 .statfs = nfs_statfs,
1582 .clear_inode = nfs4_clear_inode,
1583 .umount_begin = nfs_umount_begin,
1584 .show_options = nfs_show_options,
1585};
1586
1587/*
1588 * Clean out any remaining NFSv4 state that might be left over due
1589 * to open() calls that passed nfs_atomic_lookup, but failed to call
1590 * nfs_open().
1591 */
1592static void nfs4_clear_inode(struct inode *inode)
1593{
1594 struct nfs_inode *nfsi = NFS_I(inode);
1595
1596 /* If we are holding a delegation, return it! */
1597 if (nfsi->delegation != NULL)
1598 nfs_inode_return_delegation(inode);
1599 /* First call standard NFS clear_inode() code */
1600 nfs_clear_inode(inode);
1601 /* Now clear out any remaining state */
1602 while (!list_empty(&nfsi->open_states)) {
1603 struct nfs4_state *state;
1604
1605 state = list_entry(nfsi->open_states.next,
1606 struct nfs4_state,
1607 inode_states);
1608 dprintk("%s(%s/%Ld): found unclaimed NFSv4 state %p\n",
1609 __FUNCTION__,
1610 inode->i_sb->s_id,
1611 (long long)NFS_FILEID(inode),
1612 state);
1613 BUG_ON(atomic_read(&state->count) != 1);
1614 nfs4_close_state(state, state->state);
1615 }
1616}
1617
1618
1619static int nfs4_fill_super(struct super_block *sb, struct nfs4_mount_data *data, int silent)
1620{
1621 struct nfs_server *server;
1622 struct nfs4_client *clp = NULL;
1623 struct rpc_xprt *xprt = NULL;
1624 struct rpc_clnt *clnt = NULL;
1625 struct rpc_timeout timeparms;
1626 rpc_authflavor_t authflavour;
1627 int proto, err = -EIO;
1628
1629 sb->s_blocksize_bits = 0;
1630 sb->s_blocksize = 0;
1631 server = NFS_SB(sb);
1632 if (data->rsize != 0)
1633 server->rsize = nfs_block_size(data->rsize, NULL);
1634 if (data->wsize != 0)
1635 server->wsize = nfs_block_size(data->wsize, NULL);
1636 server->flags = data->flags & NFS_MOUNT_FLAGMASK;
1637 server->caps = NFS_CAP_ATOMIC_OPEN;
1638
1639 server->acregmin = data->acregmin*HZ;
1640 server->acregmax = data->acregmax*HZ;
1641 server->acdirmin = data->acdirmin*HZ;
1642 server->acdirmax = data->acdirmax*HZ;
1643
1644 server->rpc_ops = &nfs_v4_clientops;
1645 /* Initialize timeout values */
1646
1647 timeparms.to_initval = data->timeo * HZ / 10;
1648 timeparms.to_retries = data->retrans;
1649 timeparms.to_exponential = 1;
1650 if (!timeparms.to_retries)
1651 timeparms.to_retries = 5;
1652
1653 proto = data->proto;
1654 /* Which IP protocol do we use? */
1655 switch (proto) {
1656 case IPPROTO_TCP:
1657 timeparms.to_maxval = RPC_MAX_TCP_TIMEOUT;
1658 if (!timeparms.to_initval)
1659 timeparms.to_initval = 600 * HZ / 10;
1660 break;
1661 case IPPROTO_UDP:
1662 timeparms.to_maxval = RPC_MAX_UDP_TIMEOUT;
1663 if (!timeparms.to_initval)
1664 timeparms.to_initval = 11 * HZ / 10;
1665 break;
1666 default:
1667 return -EINVAL;
1668 }
1669
1670 clp = nfs4_get_client(&server->addr.sin_addr);
1671 if (!clp) {
Trond Myklebust9085bbc2005-06-22 17:16:20 +00001672 dprintk("%s: failed to create NFS4 client.\n", __FUNCTION__);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001673 return -EIO;
1674 }
1675
1676 /* Now create transport and client */
1677 authflavour = RPC_AUTH_UNIX;
1678 if (data->auth_flavourlen != 0) {
Trond Myklebust9085bbc2005-06-22 17:16:20 +00001679 if (data->auth_flavourlen != 1) {
1680 dprintk("%s: Invalid number of RPC auth flavours %d.\n",
1681 __FUNCTION__, data->auth_flavourlen);
1682 err = -EINVAL;
1683 goto out_fail;
1684 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001685 if (copy_from_user(&authflavour, data->auth_flavours, sizeof(authflavour))) {
1686 err = -EFAULT;
1687 goto out_fail;
1688 }
1689 }
1690
1691 down_write(&clp->cl_sem);
J. Bruce Fields6a192752005-06-22 17:16:23 +00001692 if (IS_ERR(clp->cl_rpcclient)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001693 xprt = xprt_create_proto(proto, &server->addr, &timeparms);
1694 if (IS_ERR(xprt)) {
1695 up_write(&clp->cl_sem);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001696 err = PTR_ERR(xprt);
Trond Myklebust9085bbc2005-06-22 17:16:20 +00001697 dprintk("%s: cannot create RPC transport. Error = %d\n",
1698 __FUNCTION__, err);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001699 goto out_fail;
1700 }
1701 clnt = rpc_create_client(xprt, server->hostname, &nfs_program,
1702 server->rpc_ops->version, authflavour);
1703 if (IS_ERR(clnt)) {
1704 up_write(&clp->cl_sem);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001705 err = PTR_ERR(clnt);
Trond Myklebust9085bbc2005-06-22 17:16:20 +00001706 dprintk("%s: cannot create RPC client. Error = %d\n",
1707 __FUNCTION__, err);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001708 goto out_fail;
1709 }
1710 clnt->cl_intr = 1;
1711 clnt->cl_softrtry = 1;
1712 clnt->cl_chatty = 1;
1713 clp->cl_rpcclient = clnt;
1714 clp->cl_cred = rpcauth_lookupcred(clnt->cl_auth, 0);
1715 if (IS_ERR(clp->cl_cred)) {
1716 up_write(&clp->cl_sem);
1717 err = PTR_ERR(clp->cl_cred);
1718 clp->cl_cred = NULL;
1719 goto out_fail;
1720 }
1721 memcpy(clp->cl_ipaddr, server->ip_addr, sizeof(clp->cl_ipaddr));
1722 nfs_idmap_new(clp);
1723 }
1724 if (list_empty(&clp->cl_superblocks)) {
1725 err = nfs4_init_client(clp);
1726 if (err != 0) {
1727 up_write(&clp->cl_sem);
1728 goto out_fail;
1729 }
1730 }
1731 list_add_tail(&server->nfs4_siblings, &clp->cl_superblocks);
1732 clnt = rpc_clone_client(clp->cl_rpcclient);
1733 if (!IS_ERR(clnt))
1734 server->nfs4_state = clp;
1735 up_write(&clp->cl_sem);
1736 clp = NULL;
1737
1738 if (IS_ERR(clnt)) {
Trond Myklebust9085bbc2005-06-22 17:16:20 +00001739 err = PTR_ERR(clnt);
1740 dprintk("%s: cannot create RPC client. Error = %d\n",
1741 __FUNCTION__, err);
1742 return err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001743 }
1744
1745 server->client = clnt;
1746
1747 if (server->nfs4_state->cl_idmap == NULL) {
Trond Myklebust9085bbc2005-06-22 17:16:20 +00001748 dprintk("%s: failed to create idmapper.\n", __FUNCTION__);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001749 return -ENOMEM;
1750 }
1751
1752 if (clnt->cl_auth->au_flavor != authflavour) {
J. Bruce Fields6a192752005-06-22 17:16:23 +00001753 struct rpc_auth *auth;
1754
1755 auth = rpcauth_create(authflavour, clnt);
1756 if (IS_ERR(auth)) {
Trond Myklebust9085bbc2005-06-22 17:16:20 +00001757 dprintk("%s: couldn't create credcache!\n", __FUNCTION__);
J. Bruce Fields6a192752005-06-22 17:16:23 +00001758 return PTR_ERR(auth);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001759 }
1760 }
1761
1762 sb->s_time_gran = 1;
1763
1764 sb->s_op = &nfs4_sops;
1765 err = nfs_sb_init(sb, authflavour);
1766 if (err == 0)
1767 return 0;
1768out_fail:
1769 if (clp)
1770 nfs4_put_client(clp);
1771 return err;
1772}
1773
1774static int nfs4_compare_super(struct super_block *sb, void *data)
1775{
1776 struct nfs_server *server = data;
1777 struct nfs_server *old = NFS_SB(sb);
1778
1779 if (strcmp(server->hostname, old->hostname) != 0)
1780 return 0;
1781 if (strcmp(server->mnt_path, old->mnt_path) != 0)
1782 return 0;
1783 return 1;
1784}
1785
1786static void *
1787nfs_copy_user_string(char *dst, struct nfs_string *src, int maxlen)
1788{
1789 void *p = NULL;
1790
1791 if (!src->len)
1792 return ERR_PTR(-EINVAL);
1793 if (src->len < maxlen)
1794 maxlen = src->len;
1795 if (dst == NULL) {
1796 p = dst = kmalloc(maxlen + 1, GFP_KERNEL);
1797 if (p == NULL)
1798 return ERR_PTR(-ENOMEM);
1799 }
1800 if (copy_from_user(dst, src->data, maxlen)) {
1801 if (p != NULL)
1802 kfree(p);
1803 return ERR_PTR(-EFAULT);
1804 }
1805 dst[maxlen] = '\0';
1806 return dst;
1807}
1808
1809static struct super_block *nfs4_get_sb(struct file_system_type *fs_type,
1810 int flags, const char *dev_name, void *raw_data)
1811{
1812 int error;
1813 struct nfs_server *server;
1814 struct super_block *s;
1815 struct nfs4_mount_data *data = raw_data;
1816 void *p;
1817
Trond Myklebust9085bbc2005-06-22 17:16:20 +00001818 if (data == NULL) {
1819 dprintk("%s: missing data argument\n", __FUNCTION__);
1820 return ERR_PTR(-EINVAL);
1821 }
1822 if (data->version <= 0 || data->version > NFS4_MOUNT_VERSION) {
1823 dprintk("%s: bad mount version\n", __FUNCTION__);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001824 return ERR_PTR(-EINVAL);
1825 }
1826
1827 server = kmalloc(sizeof(struct nfs_server), GFP_KERNEL);
1828 if (!server)
1829 return ERR_PTR(-ENOMEM);
1830 memset(server, 0, sizeof(struct nfs_server));
1831 /* Zero out the NFS state stuff */
1832 init_nfsv4_state(server);
Andreas Gruenbacherb7fa0552005-06-22 17:16:27 +00001833 server->client = server->client_sys = server->client_acl = ERR_PTR(-EINVAL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001834
Linus Torvalds1da177e2005-04-16 15:20:36 -07001835 p = nfs_copy_user_string(NULL, &data->hostname, 256);
1836 if (IS_ERR(p))
1837 goto out_err;
1838 server->hostname = p;
1839
1840 p = nfs_copy_user_string(NULL, &data->mnt_path, 1024);
1841 if (IS_ERR(p))
1842 goto out_err;
1843 server->mnt_path = p;
1844
1845 p = nfs_copy_user_string(server->ip_addr, &data->client_addr,
1846 sizeof(server->ip_addr) - 1);
1847 if (IS_ERR(p))
1848 goto out_err;
1849
1850 /* We now require that the mount process passes the remote address */
1851 if (data->host_addrlen != sizeof(server->addr)) {
1852 s = ERR_PTR(-EINVAL);
1853 goto out_free;
1854 }
1855 if (copy_from_user(&server->addr, data->host_addr, sizeof(server->addr))) {
1856 s = ERR_PTR(-EFAULT);
1857 goto out_free;
1858 }
1859 if (server->addr.sin_family != AF_INET ||
1860 server->addr.sin_addr.s_addr == INADDR_ANY) {
Trond Myklebust9085bbc2005-06-22 17:16:20 +00001861 dprintk("%s: mount program didn't pass remote IP address!\n",
1862 __FUNCTION__);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001863 s = ERR_PTR(-EINVAL);
1864 goto out_free;
1865 }
1866
Trond Myklebust9085bbc2005-06-22 17:16:20 +00001867 /* Fire up rpciod if not yet running */
1868 s = ERR_PTR(rpciod_up());
1869 if (IS_ERR(s)) {
1870 dprintk("%s: couldn't start rpciod! Error = %ld\n",
1871 __FUNCTION__, PTR_ERR(s));
1872 goto out_free;
1873 }
1874
Linus Torvalds1da177e2005-04-16 15:20:36 -07001875 s = sget(fs_type, nfs4_compare_super, nfs_set_super, server);
1876
1877 if (IS_ERR(s) || s->s_root)
1878 goto out_free;
1879
1880 s->s_flags = flags;
1881
Linus Torvalds1da177e2005-04-16 15:20:36 -07001882 error = nfs4_fill_super(s, data, flags & MS_VERBOSE ? 1 : 0);
1883 if (error) {
1884 up_write(&s->s_umount);
1885 deactivate_super(s);
1886 return ERR_PTR(error);
1887 }
1888 s->s_flags |= MS_ACTIVE;
1889 return s;
1890out_err:
1891 s = (struct super_block *)p;
1892out_free:
1893 if (server->mnt_path)
1894 kfree(server->mnt_path);
1895 if (server->hostname)
1896 kfree(server->hostname);
1897 kfree(server);
1898 return s;
1899}
1900
1901static void nfs4_kill_super(struct super_block *sb)
1902{
1903 struct nfs_server *server = NFS_SB(sb);
1904
1905 nfs_return_all_delegations(sb);
1906 kill_anon_super(sb);
1907
1908 nfs4_renewd_prepare_shutdown(server);
1909
1910 if (server->client != NULL && !IS_ERR(server->client))
1911 rpc_shutdown_client(server->client);
1912 rpciod_down(); /* release rpciod */
1913
1914 destroy_nfsv4_state(server);
1915
1916 if (server->hostname != NULL)
1917 kfree(server->hostname);
1918 kfree(server);
1919}
1920
1921static struct file_system_type nfs4_fs_type = {
1922 .owner = THIS_MODULE,
1923 .name = "nfs4",
1924 .get_sb = nfs4_get_sb,
1925 .kill_sb = nfs4_kill_super,
1926 .fs_flags = FS_ODD_RENAME|FS_REVAL_DOT|FS_BINARY_MOUNTDATA,
1927};
1928
1929#define nfs4_init_once(nfsi) \
1930 do { \
1931 INIT_LIST_HEAD(&(nfsi)->open_states); \
1932 nfsi->delegation = NULL; \
1933 nfsi->delegation_state = 0; \
1934 init_rwsem(&nfsi->rwsem); \
1935 } while(0)
1936#define register_nfs4fs() register_filesystem(&nfs4_fs_type)
1937#define unregister_nfs4fs() unregister_filesystem(&nfs4_fs_type)
1938#else
1939#define nfs4_init_once(nfsi) \
1940 do { } while (0)
1941#define register_nfs4fs() (0)
1942#define unregister_nfs4fs()
1943#endif
1944
1945extern int nfs_init_nfspagecache(void);
1946extern void nfs_destroy_nfspagecache(void);
1947extern int nfs_init_readpagecache(void);
1948extern void nfs_destroy_readpagecache(void);
1949extern int nfs_init_writepagecache(void);
1950extern void nfs_destroy_writepagecache(void);
1951#ifdef CONFIG_NFS_DIRECTIO
1952extern int nfs_init_directcache(void);
1953extern void nfs_destroy_directcache(void);
1954#endif
1955
1956static kmem_cache_t * nfs_inode_cachep;
1957
1958static struct inode *nfs_alloc_inode(struct super_block *sb)
1959{
1960 struct nfs_inode *nfsi;
1961 nfsi = (struct nfs_inode *)kmem_cache_alloc(nfs_inode_cachep, SLAB_KERNEL);
1962 if (!nfsi)
1963 return NULL;
1964 nfsi->flags = 0;
Trond Myklebust458818e2005-06-22 17:16:27 +00001965#ifdef CONFIG_NFS_V3_ACL
1966 nfsi->acl_access = ERR_PTR(-EAGAIN);
1967 nfsi->acl_default = ERR_PTR(-EAGAIN);
1968#endif
J. Bruce Fieldse50a1c22005-06-22 17:16:23 +00001969#ifdef CONFIG_NFS_V4
1970 nfsi->nfs4_acl = NULL;
1971#endif /* CONFIG_NFS_V4 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001972 return &nfsi->vfs_inode;
1973}
1974
1975static void nfs_destroy_inode(struct inode *inode)
1976{
1977 kmem_cache_free(nfs_inode_cachep, NFS_I(inode));
1978}
1979
1980static void init_once(void * foo, kmem_cache_t * cachep, unsigned long flags)
1981{
1982 struct nfs_inode *nfsi = (struct nfs_inode *) foo;
1983
1984 if ((flags & (SLAB_CTOR_VERIFY|SLAB_CTOR_CONSTRUCTOR)) ==
1985 SLAB_CTOR_CONSTRUCTOR) {
1986 inode_init_once(&nfsi->vfs_inode);
1987 spin_lock_init(&nfsi->req_lock);
1988 INIT_LIST_HEAD(&nfsi->dirty);
1989 INIT_LIST_HEAD(&nfsi->commit);
1990 INIT_LIST_HEAD(&nfsi->open_files);
1991 INIT_RADIX_TREE(&nfsi->nfs_page_tree, GFP_ATOMIC);
1992 atomic_set(&nfsi->data_updates, 0);
1993 nfsi->ndirty = 0;
1994 nfsi->ncommit = 0;
1995 nfsi->npages = 0;
1996 init_waitqueue_head(&nfsi->nfs_i_wait);
1997 nfs4_init_once(nfsi);
1998 }
1999}
2000
Adrian Bunk75c96f82005-05-05 16:16:09 -07002001static int nfs_init_inodecache(void)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002002{
2003 nfs_inode_cachep = kmem_cache_create("nfs_inode_cache",
2004 sizeof(struct nfs_inode),
2005 0, SLAB_RECLAIM_ACCOUNT,
2006 init_once, NULL);
2007 if (nfs_inode_cachep == NULL)
2008 return -ENOMEM;
2009
2010 return 0;
2011}
2012
Adrian Bunk75c96f82005-05-05 16:16:09 -07002013static void nfs_destroy_inodecache(void)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002014{
2015 if (kmem_cache_destroy(nfs_inode_cachep))
2016 printk(KERN_INFO "nfs_inode_cache: not all structures were freed\n");
2017}
2018
2019/*
2020 * Initialize NFS
2021 */
2022static int __init init_nfs_fs(void)
2023{
2024 int err;
2025
2026 err = nfs_init_nfspagecache();
2027 if (err)
2028 goto out4;
2029
2030 err = nfs_init_inodecache();
2031 if (err)
2032 goto out3;
2033
2034 err = nfs_init_readpagecache();
2035 if (err)
2036 goto out2;
2037
2038 err = nfs_init_writepagecache();
2039 if (err)
2040 goto out1;
2041
2042#ifdef CONFIG_NFS_DIRECTIO
2043 err = nfs_init_directcache();
2044 if (err)
2045 goto out0;
2046#endif
2047
2048#ifdef CONFIG_PROC_FS
2049 rpc_proc_register(&nfs_rpcstat);
2050#endif
2051 err = register_filesystem(&nfs_fs_type);
2052 if (err)
2053 goto out;
2054 if ((err = register_nfs4fs()) != 0)
2055 goto out;
2056 return 0;
2057out:
2058#ifdef CONFIG_PROC_FS
2059 rpc_proc_unregister("nfs");
2060#endif
2061 nfs_destroy_writepagecache();
2062#ifdef CONFIG_NFS_DIRECTIO
2063out0:
2064 nfs_destroy_directcache();
2065#endif
2066out1:
2067 nfs_destroy_readpagecache();
2068out2:
2069 nfs_destroy_inodecache();
2070out3:
2071 nfs_destroy_nfspagecache();
2072out4:
2073 return err;
2074}
2075
2076static void __exit exit_nfs_fs(void)
2077{
2078#ifdef CONFIG_NFS_DIRECTIO
2079 nfs_destroy_directcache();
2080#endif
2081 nfs_destroy_writepagecache();
2082 nfs_destroy_readpagecache();
2083 nfs_destroy_inodecache();
2084 nfs_destroy_nfspagecache();
2085#ifdef CONFIG_PROC_FS
2086 rpc_proc_unregister("nfs");
2087#endif
2088 unregister_filesystem(&nfs_fs_type);
2089 unregister_nfs4fs();
2090}
2091
2092/* Not quite true; I just maintain it */
2093MODULE_AUTHOR("Olaf Kirch <okir@monad.swb.de>");
2094MODULE_LICENSE("GPL");
2095
2096module_init(init_nfs_fs)
2097module_exit(exit_nfs_fs)