blob: a3922f4cc0a88e7837c0e594d60cf6f321dbd49a [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
138 ret = nfs_commit_inode(inode, 0, 0, flags);
139 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 Myklebustada70d92005-06-22 17:16:22 +0000623 nfsi->flags |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_DATA|NFS_INO_INVALID_ACCESS|NFS_INO_INVALID_ACL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700624 else
Trond Myklebustada70d92005-06-22 17:16:22 +0000625 nfsi->flags |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_ACCESS|NFS_INO_INVALID_ACL;
626}
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);
817 if (error == 0) {
818 nfs_refresh_inode(inode, &fattr);
819 if ((attr->ia_valid & ATTR_MODE) != 0) {
820 int mode;
821 mode = inode->i_mode & ~S_IALLUGO;
822 mode |= attr->ia_mode & S_IALLUGO;
823 inode->i_mode = mode;
824 }
825 if ((attr->ia_valid & ATTR_UID) != 0)
826 inode->i_uid = attr->ia_uid;
827 if ((attr->ia_valid & ATTR_GID) != 0)
828 inode->i_gid = attr->ia_gid;
829 if ((attr->ia_valid & ATTR_SIZE) != 0) {
830 inode->i_size = attr->ia_size;
831 vmtruncate(inode, attr->ia_size);
832 }
833 }
834 if ((attr->ia_valid & (ATTR_MODE|ATTR_UID|ATTR_GID)) != 0)
Trond Myklebustada70d92005-06-22 17:16:22 +0000835 NFS_FLAGS(inode) |= NFS_INO_INVALID_ACCESS|NFS_INO_INVALID_ACL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700836 nfs_end_data_update(inode);
837 unlock_kernel();
838 return error;
839}
840
841/*
842 * Wait for the inode to get unlocked.
843 * (Used for NFS_INO_LOCKED and NFS_INO_REVALIDATING).
844 */
845static int
846nfs_wait_on_inode(struct inode *inode, int flag)
847{
848 struct rpc_clnt *clnt = NFS_CLIENT(inode);
849 struct nfs_inode *nfsi = NFS_I(inode);
850
851 int error;
852 if (!(NFS_FLAGS(inode) & flag))
853 return 0;
854 atomic_inc(&inode->i_count);
855 error = nfs_wait_event(clnt, nfsi->nfs_i_wait,
856 !(NFS_FLAGS(inode) & flag));
857 iput(inode);
858 return error;
859}
860
861int nfs_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
862{
863 struct inode *inode = dentry->d_inode;
864 struct nfs_inode *nfsi = NFS_I(inode);
865 int need_atime = nfsi->flags & NFS_INO_INVALID_ATIME;
866 int err;
867
868 if (__IS_FLG(inode, MS_NOATIME))
869 need_atime = 0;
870 else if (__IS_FLG(inode, MS_NODIRATIME) && S_ISDIR(inode->i_mode))
871 need_atime = 0;
872 /* We may force a getattr if the user cares about atime */
873 if (need_atime)
874 err = __nfs_revalidate_inode(NFS_SERVER(inode), inode);
875 else
876 err = nfs_revalidate_inode(NFS_SERVER(inode), inode);
877 if (!err)
878 generic_fillattr(inode, stat);
879 return err;
880}
881
882struct nfs_open_context *alloc_nfs_open_context(struct dentry *dentry, struct rpc_cred *cred)
883{
884 struct nfs_open_context *ctx;
885
886 ctx = (struct nfs_open_context *)kmalloc(sizeof(*ctx), GFP_KERNEL);
887 if (ctx != NULL) {
888 atomic_set(&ctx->count, 1);
889 ctx->dentry = dget(dentry);
890 ctx->cred = get_rpccred(cred);
891 ctx->state = NULL;
892 ctx->lockowner = current->files;
893 ctx->error = 0;
Olivier Galibert00a92642005-06-22 17:16:29 +0000894 ctx->dir_cookie = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700895 }
896 return ctx;
897}
898
899struct nfs_open_context *get_nfs_open_context(struct nfs_open_context *ctx)
900{
901 if (ctx != NULL)
902 atomic_inc(&ctx->count);
903 return ctx;
904}
905
906void put_nfs_open_context(struct nfs_open_context *ctx)
907{
908 if (atomic_dec_and_test(&ctx->count)) {
909 if (!list_empty(&ctx->list)) {
910 struct inode *inode = ctx->dentry->d_inode;
911 spin_lock(&inode->i_lock);
912 list_del(&ctx->list);
913 spin_unlock(&inode->i_lock);
914 }
915 if (ctx->state != NULL)
916 nfs4_close_state(ctx->state, ctx->mode);
917 if (ctx->cred != NULL)
918 put_rpccred(ctx->cred);
919 dput(ctx->dentry);
920 kfree(ctx);
921 }
922}
923
924/*
925 * Ensure that mmap has a recent RPC credential for use when writing out
926 * shared pages
927 */
928void nfs_file_set_open_context(struct file *filp, struct nfs_open_context *ctx)
929{
930 struct inode *inode = filp->f_dentry->d_inode;
931 struct nfs_inode *nfsi = NFS_I(inode);
932
933 filp->private_data = get_nfs_open_context(ctx);
934 spin_lock(&inode->i_lock);
935 list_add(&ctx->list, &nfsi->open_files);
936 spin_unlock(&inode->i_lock);
937}
938
939struct nfs_open_context *nfs_find_open_context(struct inode *inode, int mode)
940{
941 struct nfs_inode *nfsi = NFS_I(inode);
942 struct nfs_open_context *pos, *ctx = NULL;
943
944 spin_lock(&inode->i_lock);
945 list_for_each_entry(pos, &nfsi->open_files, list) {
946 if ((pos->mode & mode) == mode) {
947 ctx = get_nfs_open_context(pos);
948 break;
949 }
950 }
951 spin_unlock(&inode->i_lock);
952 return ctx;
953}
954
955void nfs_file_clear_open_context(struct file *filp)
956{
957 struct inode *inode = filp->f_dentry->d_inode;
958 struct nfs_open_context *ctx = (struct nfs_open_context *)filp->private_data;
959
960 if (ctx) {
961 filp->private_data = NULL;
962 spin_lock(&inode->i_lock);
963 list_move_tail(&ctx->list, &NFS_I(inode)->open_files);
964 spin_unlock(&inode->i_lock);
965 put_nfs_open_context(ctx);
966 }
967}
968
969/*
970 * These allocate and release file read/write context information.
971 */
972int nfs_open(struct inode *inode, struct file *filp)
973{
974 struct nfs_open_context *ctx;
975 struct rpc_cred *cred;
976
977 cred = rpcauth_lookupcred(NFS_CLIENT(inode)->cl_auth, 0);
978 if (IS_ERR(cred))
979 return PTR_ERR(cred);
980 ctx = alloc_nfs_open_context(filp->f_dentry, cred);
981 put_rpccred(cred);
982 if (ctx == NULL)
983 return -ENOMEM;
984 ctx->mode = filp->f_mode;
985 nfs_file_set_open_context(filp, ctx);
986 put_nfs_open_context(ctx);
987 if ((filp->f_mode & FMODE_WRITE) != 0)
988 nfs_begin_data_update(inode);
989 return 0;
990}
991
992int nfs_release(struct inode *inode, struct file *filp)
993{
994 if ((filp->f_mode & FMODE_WRITE) != 0)
995 nfs_end_data_update(inode);
996 nfs_file_clear_open_context(filp);
997 return 0;
998}
999
1000/*
1001 * This function is called whenever some part of NFS notices that
1002 * the cached attributes have to be refreshed.
1003 */
1004int
1005__nfs_revalidate_inode(struct nfs_server *server, struct inode *inode)
1006{
1007 int status = -ESTALE;
1008 struct nfs_fattr fattr;
1009 struct nfs_inode *nfsi = NFS_I(inode);
1010 unsigned long verifier;
1011 unsigned int flags;
1012
1013 dfprintk(PAGECACHE, "NFS: revalidating (%s/%Ld)\n",
1014 inode->i_sb->s_id, (long long)NFS_FILEID(inode));
1015
1016 lock_kernel();
1017 if (!inode || is_bad_inode(inode))
1018 goto out_nowait;
1019 if (NFS_STALE(inode))
1020 goto out_nowait;
1021
1022 while (NFS_REVALIDATING(inode)) {
1023 status = nfs_wait_on_inode(inode, NFS_INO_REVALIDATING);
1024 if (status < 0)
1025 goto out_nowait;
1026 if (NFS_ATTRTIMEO(inode) == 0)
1027 continue;
1028 if (NFS_FLAGS(inode) & (NFS_INO_INVALID_ATTR|NFS_INO_INVALID_DATA|NFS_INO_INVALID_ATIME))
1029 continue;
1030 status = NFS_STALE(inode) ? -ESTALE : 0;
1031 goto out_nowait;
1032 }
1033 NFS_FLAGS(inode) |= NFS_INO_REVALIDATING;
1034
1035 /* Protect against RPC races by saving the change attribute */
1036 verifier = nfs_save_change_attribute(inode);
1037 status = NFS_PROTO(inode)->getattr(server, NFS_FH(inode), &fattr);
1038 if (status != 0) {
1039 dfprintk(PAGECACHE, "nfs_revalidate_inode: (%s/%Ld) getattr failed, error=%d\n",
1040 inode->i_sb->s_id,
1041 (long long)NFS_FILEID(inode), status);
1042 if (status == -ESTALE) {
1043 nfs_zap_caches(inode);
1044 if (!S_ISDIR(inode->i_mode))
1045 NFS_FLAGS(inode) |= NFS_INO_STALE;
1046 }
1047 goto out;
1048 }
1049
1050 status = nfs_update_inode(inode, &fattr, verifier);
1051 if (status) {
1052 dfprintk(PAGECACHE, "nfs_revalidate_inode: (%s/%Ld) refresh failed, error=%d\n",
1053 inode->i_sb->s_id,
1054 (long long)NFS_FILEID(inode), status);
1055 goto out;
1056 }
1057 flags = nfsi->flags;
1058 /*
1059 * We may need to keep the attributes marked as invalid if
1060 * we raced with nfs_end_attr_update().
1061 */
1062 if (verifier == nfsi->cache_change_attribute)
1063 nfsi->flags &= ~(NFS_INO_INVALID_ATTR|NFS_INO_INVALID_ATIME);
1064 /* Do the page cache invalidation */
Trond Myklebust7d52e862005-06-22 17:16:30 +00001065 nfs_revalidate_mapping(inode, inode->i_mapping);
Trond Myklebustada70d92005-06-22 17:16:22 +00001066 if (flags & NFS_INO_INVALID_ACL)
1067 nfs_zap_acl_cache(inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001068 dfprintk(PAGECACHE, "NFS: (%s/%Ld) revalidation complete\n",
1069 inode->i_sb->s_id,
1070 (long long)NFS_FILEID(inode));
1071
1072out:
1073 NFS_FLAGS(inode) &= ~NFS_INO_REVALIDATING;
1074 wake_up(&nfsi->nfs_i_wait);
1075 out_nowait:
1076 unlock_kernel();
1077 return status;
1078}
1079
1080int nfs_attribute_timeout(struct inode *inode)
1081{
1082 struct nfs_inode *nfsi = NFS_I(inode);
1083
1084 if (nfs_have_delegation(inode, FMODE_READ))
1085 return 0;
1086 return time_after(jiffies, nfsi->read_cache_jiffies+nfsi->attrtimeo);
1087}
1088
1089/**
1090 * nfs_revalidate_inode - Revalidate the inode attributes
1091 * @server - pointer to nfs_server struct
1092 * @inode - pointer to inode struct
1093 *
1094 * Updates inode attribute information by retrieving the data from the server.
1095 */
1096int nfs_revalidate_inode(struct nfs_server *server, struct inode *inode)
1097{
1098 if (!(NFS_FLAGS(inode) & (NFS_INO_INVALID_ATTR|NFS_INO_INVALID_DATA))
1099 && !nfs_attribute_timeout(inode))
1100 return NFS_STALE(inode) ? -ESTALE : 0;
1101 return __nfs_revalidate_inode(server, inode);
1102}
1103
1104/**
Trond Myklebust7d52e862005-06-22 17:16:30 +00001105 * nfs_revalidate_mapping - Revalidate the pagecache
1106 * @inode - pointer to host inode
1107 * @mapping - pointer to mapping
1108 */
1109void nfs_revalidate_mapping(struct inode *inode, struct address_space *mapping)
1110{
1111 struct nfs_inode *nfsi = NFS_I(inode);
1112
1113 if (nfsi->flags & NFS_INO_INVALID_DATA) {
1114 if (S_ISREG(inode->i_mode)) {
1115 if (filemap_fdatawrite(mapping) == 0)
1116 filemap_fdatawait(mapping);
1117 nfs_wb_all(inode);
1118 }
1119 invalidate_inode_pages2(mapping);
1120 nfsi->flags &= ~NFS_INO_INVALID_DATA;
1121 if (S_ISDIR(inode->i_mode)) {
1122 memset(nfsi->cookieverf, 0, sizeof(nfsi->cookieverf));
1123 /* This ensures we revalidate child dentries */
1124 nfsi->cache_change_attribute++;
1125 }
1126 dfprintk(PAGECACHE, "NFS: (%s/%Ld) data cache invalidated\n",
1127 inode->i_sb->s_id,
1128 (long long)NFS_FILEID(inode));
1129 }
1130}
1131
1132/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001133 * nfs_begin_data_update
1134 * @inode - pointer to inode
1135 * Declare that a set of operations will update file data on the server
1136 */
1137void nfs_begin_data_update(struct inode *inode)
1138{
1139 atomic_inc(&NFS_I(inode)->data_updates);
1140}
1141
1142/**
1143 * nfs_end_data_update
1144 * @inode - pointer to inode
1145 * Declare end of the operations that will update file data
1146 * This will mark the inode as immediately needing revalidation
1147 * of its attribute cache.
1148 */
1149void nfs_end_data_update(struct inode *inode)
1150{
1151 struct nfs_inode *nfsi = NFS_I(inode);
1152
1153 if (!nfs_have_delegation(inode, FMODE_READ)) {
1154 /* Mark the attribute cache for revalidation */
1155 nfsi->flags |= NFS_INO_INVALID_ATTR;
1156 /* Directories and symlinks: invalidate page cache too */
1157 if (S_ISDIR(inode->i_mode) || S_ISLNK(inode->i_mode))
1158 nfsi->flags |= NFS_INO_INVALID_DATA;
1159 }
1160 nfsi->cache_change_attribute ++;
1161 atomic_dec(&nfsi->data_updates);
1162}
1163
1164/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001165 * nfs_refresh_inode - verify consistency of the inode attribute cache
1166 * @inode - pointer to inode
1167 * @fattr - updated attributes
1168 *
1169 * Verifies the attribute cache. If we have just changed the attributes,
1170 * so that fattr carries weak cache consistency data, then it may
1171 * also update the ctime/mtime/change_attribute.
1172 */
1173int nfs_refresh_inode(struct inode *inode, struct nfs_fattr *fattr)
1174{
1175 struct nfs_inode *nfsi = NFS_I(inode);
1176 loff_t cur_size, new_isize;
1177 int data_unstable;
1178
1179 /* Do we hold a delegation? */
1180 if (nfs_have_delegation(inode, FMODE_READ))
1181 return 0;
1182
1183 /* Are we in the process of updating data on the server? */
1184 data_unstable = nfs_caches_unstable(inode);
1185
1186 if (fattr->valid & NFS_ATTR_FATTR_V4) {
1187 if ((fattr->valid & NFS_ATTR_PRE_CHANGE) != 0
1188 && nfsi->change_attr == fattr->pre_change_attr)
1189 nfsi->change_attr = fattr->change_attr;
1190 if (!data_unstable && nfsi->change_attr != fattr->change_attr)
1191 nfsi->flags |= NFS_INO_INVALID_ATTR;
1192 }
1193
1194 if ((fattr->valid & NFS_ATTR_FATTR) == 0)
1195 return 0;
1196
1197 /* Has the inode gone and changed behind our back? */
1198 if (nfsi->fileid != fattr->fileid
1199 || (inode->i_mode & S_IFMT) != (fattr->mode & S_IFMT))
1200 return -EIO;
1201
1202 cur_size = i_size_read(inode);
1203 new_isize = nfs_size_to_loff_t(fattr->size);
1204
1205 /* If we have atomic WCC data, we may update some attributes */
1206 if ((fattr->valid & NFS_ATTR_WCC) != 0) {
1207 if (timespec_equal(&inode->i_ctime, &fattr->pre_ctime))
1208 memcpy(&inode->i_ctime, &fattr->ctime, sizeof(inode->i_ctime));
1209 if (timespec_equal(&inode->i_mtime, &fattr->pre_mtime))
1210 memcpy(&inode->i_mtime, &fattr->mtime, sizeof(inode->i_mtime));
1211 }
1212
1213 /* Verify a few of the more important attributes */
1214 if (!data_unstable) {
1215 if (!timespec_equal(&inode->i_mtime, &fattr->mtime)
1216 || cur_size != new_isize)
1217 nfsi->flags |= NFS_INO_INVALID_ATTR;
Trond Myklebust951a1432005-06-22 17:16:30 +00001218 } else if (new_isize != cur_size && nfsi->npages == 0)
1219 nfsi->flags |= NFS_INO_INVALID_ATTR;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001220
1221 /* Have any file permissions changed? */
1222 if ((inode->i_mode & S_IALLUGO) != (fattr->mode & S_IALLUGO)
1223 || inode->i_uid != fattr->uid
1224 || inode->i_gid != fattr->gid)
Trond Myklebustada70d92005-06-22 17:16:22 +00001225 nfsi->flags |= NFS_INO_INVALID_ATTR | NFS_INO_INVALID_ACCESS | NFS_INO_INVALID_ACL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001226
1227 /* Has the link count changed? */
1228 if (inode->i_nlink != fattr->nlink)
1229 nfsi->flags |= NFS_INO_INVALID_ATTR;
1230
1231 if (!timespec_equal(&inode->i_atime, &fattr->atime))
1232 nfsi->flags |= NFS_INO_INVALID_ATIME;
1233
1234 nfsi->read_cache_jiffies = fattr->timestamp;
1235 return 0;
1236}
1237
1238/*
1239 * Many nfs protocol calls return the new file attributes after
1240 * an operation. Here we update the inode to reflect the state
1241 * of the server's inode.
1242 *
1243 * This is a bit tricky because we have to make sure all dirty pages
1244 * have been sent off to the server before calling invalidate_inode_pages.
1245 * To make sure no other process adds more write requests while we try
1246 * our best to flush them, we make them sleep during the attribute refresh.
1247 *
1248 * A very similar scenario holds for the dir cache.
1249 */
1250static int nfs_update_inode(struct inode *inode, struct nfs_fattr *fattr, unsigned long verifier)
1251{
1252 struct nfs_inode *nfsi = NFS_I(inode);
Trond Myklebust951a1432005-06-22 17:16:30 +00001253 loff_t cur_isize, new_isize;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001254 unsigned int invalid = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001255 int data_unstable;
1256
1257 dfprintk(VFS, "NFS: %s(%s/%ld ct=%d info=0x%x)\n",
1258 __FUNCTION__, inode->i_sb->s_id, inode->i_ino,
1259 atomic_read(&inode->i_count), fattr->valid);
1260
1261 if ((fattr->valid & NFS_ATTR_FATTR) == 0)
1262 return 0;
1263
1264 if (nfsi->fileid != fattr->fileid) {
1265 printk(KERN_ERR "%s: inode number mismatch\n"
1266 "expected (%s/0x%Lx), got (%s/0x%Lx)\n",
1267 __FUNCTION__,
1268 inode->i_sb->s_id, (long long)nfsi->fileid,
1269 inode->i_sb->s_id, (long long)fattr->fileid);
1270 goto out_err;
1271 }
1272
1273 /*
1274 * Make sure the inode's type hasn't changed.
1275 */
1276 if ((inode->i_mode & S_IFMT) != (fattr->mode & S_IFMT))
1277 goto out_changed;
1278
1279 /*
1280 * Update the read time so we don't revalidate too often.
1281 */
1282 nfsi->read_cache_jiffies = fattr->timestamp;
1283
1284 /* Are we racing with known updates of the metadata on the server? */
1285 data_unstable = ! nfs_verify_change_attribute(inode, verifier);
1286
Trond Myklebust951a1432005-06-22 17:16:30 +00001287 /* Check if our cached file size is stale */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001288 new_isize = nfs_size_to_loff_t(fattr->size);
1289 cur_isize = i_size_read(inode);
Trond Myklebust951a1432005-06-22 17:16:30 +00001290 if (new_isize != cur_isize) {
1291 /* Do we perhaps have any outstanding writes? */
1292 if (nfsi->npages == 0) {
1293 /* No, but did we race with nfs_end_data_update()? */
1294 if (verifier == nfsi->cache_change_attribute) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001295 inode->i_size = new_isize;
Trond Myklebust951a1432005-06-22 17:16:30 +00001296 invalid |= NFS_INO_INVALID_DATA;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001297 }
Trond Myklebust951a1432005-06-22 17:16:30 +00001298 invalid |= NFS_INO_INVALID_ATTR;
1299 } else if (new_isize > cur_isize) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001300 inode->i_size = new_isize;
1301 invalid |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_DATA;
1302 }
Trond Myklebust951a1432005-06-22 17:16:30 +00001303 dprintk("NFS: isize change on server for file %s/%ld\n",
1304 inode->i_sb->s_id, inode->i_ino);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001305 }
1306
Trond Myklebust951a1432005-06-22 17:16:30 +00001307 /* Check if the mtime agrees */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001308 if (!timespec_equal(&inode->i_mtime, &fattr->mtime)) {
1309 memcpy(&inode->i_mtime, &fattr->mtime, sizeof(inode->i_mtime));
Trond Myklebust951a1432005-06-22 17:16:30 +00001310 dprintk("NFS: mtime change on server for file %s/%ld\n",
1311 inode->i_sb->s_id, inode->i_ino);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001312 if (!data_unstable)
1313 invalid |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_DATA;
1314 }
1315
1316 if ((fattr->valid & NFS_ATTR_FATTR_V4)
1317 && nfsi->change_attr != fattr->change_attr) {
Trond Myklebust951a1432005-06-22 17:16:30 +00001318 dprintk("NFS: change_attr change on server for file %s/%ld\n",
Linus Torvalds1da177e2005-04-16 15:20:36 -07001319 inode->i_sb->s_id, inode->i_ino);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001320 nfsi->change_attr = fattr->change_attr;
1321 if (!data_unstable)
Trond Myklebustada70d92005-06-22 17:16:22 +00001322 invalid |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_DATA|NFS_INO_INVALID_ACCESS|NFS_INO_INVALID_ACL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001323 }
1324
Trond Myklebustada70d92005-06-22 17:16:22 +00001325 /* If ctime has changed we should definitely clear access+acl caches */
1326 if (!timespec_equal(&inode->i_ctime, &fattr->ctime)) {
1327 if (!data_unstable)
1328 invalid |= NFS_INO_INVALID_ACCESS|NFS_INO_INVALID_ACL;
1329 memcpy(&inode->i_ctime, &fattr->ctime, sizeof(inode->i_ctime));
1330 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001331 memcpy(&inode->i_atime, &fattr->atime, sizeof(inode->i_atime));
1332
1333 if ((inode->i_mode & S_IALLUGO) != (fattr->mode & S_IALLUGO) ||
1334 inode->i_uid != fattr->uid ||
1335 inode->i_gid != fattr->gid)
Trond Myklebustada70d92005-06-22 17:16:22 +00001336 invalid |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_ACCESS|NFS_INO_INVALID_ACL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001337
1338 inode->i_mode = fattr->mode;
1339 inode->i_nlink = fattr->nlink;
1340 inode->i_uid = fattr->uid;
1341 inode->i_gid = fattr->gid;
1342
1343 if (fattr->valid & (NFS_ATTR_FATTR_V3 | NFS_ATTR_FATTR_V4)) {
1344 /*
1345 * report the blocks in 512byte units
1346 */
1347 inode->i_blocks = nfs_calc_block_size(fattr->du.nfs3.used);
1348 inode->i_blksize = inode->i_sb->s_blocksize;
1349 } else {
1350 inode->i_blocks = fattr->du.nfs2.blocks;
1351 inode->i_blksize = fattr->du.nfs2.blocksize;
1352 }
1353
1354 /* Update attrtimeo value if we're out of the unstable period */
1355 if (invalid & NFS_INO_INVALID_ATTR) {
1356 nfsi->attrtimeo = NFS_MINATTRTIMEO(inode);
1357 nfsi->attrtimeo_timestamp = jiffies;
1358 } else if (time_after(jiffies, nfsi->attrtimeo_timestamp+nfsi->attrtimeo)) {
1359 if ((nfsi->attrtimeo <<= 1) > NFS_MAXATTRTIMEO(inode))
1360 nfsi->attrtimeo = NFS_MAXATTRTIMEO(inode);
1361 nfsi->attrtimeo_timestamp = jiffies;
1362 }
1363 /* Don't invalidate the data if we were to blame */
1364 if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode)
1365 || S_ISLNK(inode->i_mode)))
1366 invalid &= ~NFS_INO_INVALID_DATA;
1367 if (!nfs_have_delegation(inode, FMODE_READ))
1368 nfsi->flags |= invalid;
1369
1370 return 0;
1371 out_changed:
1372 /*
1373 * Big trouble! The inode has become a different object.
1374 */
1375#ifdef NFS_PARANOIA
1376 printk(KERN_DEBUG "%s: inode %ld mode changed, %07o to %07o\n",
1377 __FUNCTION__, inode->i_ino, inode->i_mode, fattr->mode);
1378#endif
1379 /*
1380 * No need to worry about unhashing the dentry, as the
1381 * lookup validation will know that the inode is bad.
1382 * (But we fall through to invalidate the caches.)
1383 */
1384 nfs_invalidate_inode(inode);
1385 out_err:
1386 NFS_FLAGS(inode) |= NFS_INO_STALE;
1387 return -ESTALE;
1388}
1389
1390/*
1391 * File system information
1392 */
1393
1394static int nfs_set_super(struct super_block *s, void *data)
1395{
1396 s->s_fs_info = data;
1397 return set_anon_super(s, data);
1398}
1399
1400static int nfs_compare_super(struct super_block *sb, void *data)
1401{
1402 struct nfs_server *server = data;
1403 struct nfs_server *old = NFS_SB(sb);
1404
1405 if (old->addr.sin_addr.s_addr != server->addr.sin_addr.s_addr)
1406 return 0;
1407 if (old->addr.sin_port != server->addr.sin_port)
1408 return 0;
1409 return !nfs_compare_fh(&old->fh, &server->fh);
1410}
1411
1412static struct super_block *nfs_get_sb(struct file_system_type *fs_type,
1413 int flags, const char *dev_name, void *raw_data)
1414{
1415 int error;
Trond Myklebust9085bbc2005-06-22 17:16:20 +00001416 struct nfs_server *server = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001417 struct super_block *s;
1418 struct nfs_fh *root;
1419 struct nfs_mount_data *data = raw_data;
1420
Trond Myklebust9085bbc2005-06-22 17:16:20 +00001421 s = ERR_PTR(-EINVAL);
1422 if (data == NULL) {
1423 dprintk("%s: missing data argument\n", __FUNCTION__);
1424 goto out_err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001425 }
Trond Myklebust9085bbc2005-06-22 17:16:20 +00001426 if (data->version <= 0 || data->version > NFS_MOUNT_VERSION) {
1427 dprintk("%s: bad mount version\n", __FUNCTION__);
1428 goto out_err;
1429 }
1430 switch (data->version) {
1431 case 1:
1432 data->namlen = 0;
1433 case 2:
1434 data->bsize = 0;
1435 case 3:
1436 if (data->flags & NFS_MOUNT_VER3) {
1437 dprintk("%s: mount structure version %d does not support NFSv3\n",
1438 __FUNCTION__,
1439 data->version);
1440 goto out_err;
1441 }
1442 data->root.size = NFS2_FHSIZE;
1443 memcpy(data->root.data, data->old_root.data, NFS2_FHSIZE);
1444 case 4:
1445 if (data->flags & NFS_MOUNT_SECFLAVOUR) {
1446 dprintk("%s: mount structure version %d does not support strong security\n",
1447 __FUNCTION__,
1448 data->version);
1449 goto out_err;
1450 }
1451 case 5:
1452 memset(data->context, 0, sizeof(data->context));
1453 }
1454#ifndef CONFIG_NFS_V3
1455 /* If NFSv3 is not compiled in, return -EPROTONOSUPPORT */
1456 s = ERR_PTR(-EPROTONOSUPPORT);
1457 if (data->flags & NFS_MOUNT_VER3) {
1458 dprintk("%s: NFSv3 not compiled into kernel\n", __FUNCTION__);
1459 goto out_err;
1460 }
1461#endif /* CONFIG_NFS_V3 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001462
Trond Myklebust9085bbc2005-06-22 17:16:20 +00001463 s = ERR_PTR(-ENOMEM);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001464 server = kmalloc(sizeof(struct nfs_server), GFP_KERNEL);
1465 if (!server)
Trond Myklebust9085bbc2005-06-22 17:16:20 +00001466 goto out_err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001467 memset(server, 0, sizeof(struct nfs_server));
1468 /* Zero out the NFS state stuff */
1469 init_nfsv4_state(server);
Andreas Gruenbacherb7fa0552005-06-22 17:16:27 +00001470 server->client = server->client_sys = server->client_acl = ERR_PTR(-EINVAL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001471
Linus Torvalds1da177e2005-04-16 15:20:36 -07001472 root = &server->fh;
1473 if (data->flags & NFS_MOUNT_VER3)
1474 root->size = data->root.size;
1475 else
1476 root->size = NFS2_FHSIZE;
Trond Myklebust9085bbc2005-06-22 17:16:20 +00001477 s = ERR_PTR(-EINVAL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001478 if (root->size > sizeof(root->data)) {
Trond Myklebust9085bbc2005-06-22 17:16:20 +00001479 dprintk("%s: invalid root filehandle\n", __FUNCTION__);
1480 goto out_err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001481 }
1482 memcpy(root->data, data->root.data, root->size);
1483
1484 /* We now require that the mount process passes the remote address */
1485 memcpy(&server->addr, &data->addr, sizeof(server->addr));
1486 if (server->addr.sin_addr.s_addr == INADDR_ANY) {
Trond Myklebust9085bbc2005-06-22 17:16:20 +00001487 dprintk("%s: mount program didn't pass remote address!\n",
1488 __FUNCTION__);
1489 goto out_err;
1490 }
1491
1492 /* Fire up rpciod if not yet running */
1493 s = ERR_PTR(rpciod_up());
1494 if (IS_ERR(s)) {
1495 dprintk("%s: couldn't start rpciod! Error = %ld\n",
1496 __FUNCTION__, PTR_ERR(s));
1497 goto out_err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001498 }
1499
1500 s = sget(fs_type, nfs_compare_super, nfs_set_super, server);
Trond Myklebust9085bbc2005-06-22 17:16:20 +00001501 if (IS_ERR(s) || s->s_root)
1502 goto out_rpciod_down;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001503
1504 s->s_flags = flags;
1505
Linus Torvalds1da177e2005-04-16 15:20:36 -07001506 error = nfs_fill_super(s, data, flags & MS_VERBOSE ? 1 : 0);
1507 if (error) {
1508 up_write(&s->s_umount);
1509 deactivate_super(s);
1510 return ERR_PTR(error);
1511 }
1512 s->s_flags |= MS_ACTIVE;
1513 return s;
Trond Myklebust9085bbc2005-06-22 17:16:20 +00001514out_rpciod_down:
1515 rpciod_down();
1516out_err:
1517 kfree(server);
1518 return s;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001519}
1520
1521static void nfs_kill_super(struct super_block *s)
1522{
1523 struct nfs_server *server = NFS_SB(s);
1524
1525 kill_anon_super(s);
1526
J. Bruce Fields6a192752005-06-22 17:16:23 +00001527 if (!IS_ERR(server->client))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001528 rpc_shutdown_client(server->client);
J. Bruce Fields6a192752005-06-22 17:16:23 +00001529 if (!IS_ERR(server->client_sys))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001530 rpc_shutdown_client(server->client_sys);
Andreas Gruenbacherb7fa0552005-06-22 17:16:27 +00001531 if (!IS_ERR(server->client_acl))
1532 rpc_shutdown_client(server->client_acl);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001533
1534 if (!(server->flags & NFS_MOUNT_NONLM))
1535 lockd_down(); /* release rpc.lockd */
1536
1537 rpciod_down(); /* release rpciod */
1538
1539 if (server->hostname != NULL)
1540 kfree(server->hostname);
1541 kfree(server);
1542}
1543
1544static struct file_system_type nfs_fs_type = {
1545 .owner = THIS_MODULE,
1546 .name = "nfs",
1547 .get_sb = nfs_get_sb,
1548 .kill_sb = nfs_kill_super,
1549 .fs_flags = FS_ODD_RENAME|FS_REVAL_DOT|FS_BINARY_MOUNTDATA,
1550};
1551
1552#ifdef CONFIG_NFS_V4
1553
1554static void nfs4_clear_inode(struct inode *);
1555
1556
1557static struct super_operations nfs4_sops = {
1558 .alloc_inode = nfs_alloc_inode,
1559 .destroy_inode = nfs_destroy_inode,
1560 .write_inode = nfs_write_inode,
1561 .delete_inode = nfs_delete_inode,
1562 .statfs = nfs_statfs,
1563 .clear_inode = nfs4_clear_inode,
1564 .umount_begin = nfs_umount_begin,
1565 .show_options = nfs_show_options,
1566};
1567
1568/*
1569 * Clean out any remaining NFSv4 state that might be left over due
1570 * to open() calls that passed nfs_atomic_lookup, but failed to call
1571 * nfs_open().
1572 */
1573static void nfs4_clear_inode(struct inode *inode)
1574{
1575 struct nfs_inode *nfsi = NFS_I(inode);
1576
1577 /* If we are holding a delegation, return it! */
1578 if (nfsi->delegation != NULL)
1579 nfs_inode_return_delegation(inode);
1580 /* First call standard NFS clear_inode() code */
1581 nfs_clear_inode(inode);
1582 /* Now clear out any remaining state */
1583 while (!list_empty(&nfsi->open_states)) {
1584 struct nfs4_state *state;
1585
1586 state = list_entry(nfsi->open_states.next,
1587 struct nfs4_state,
1588 inode_states);
1589 dprintk("%s(%s/%Ld): found unclaimed NFSv4 state %p\n",
1590 __FUNCTION__,
1591 inode->i_sb->s_id,
1592 (long long)NFS_FILEID(inode),
1593 state);
1594 BUG_ON(atomic_read(&state->count) != 1);
1595 nfs4_close_state(state, state->state);
1596 }
1597}
1598
1599
1600static int nfs4_fill_super(struct super_block *sb, struct nfs4_mount_data *data, int silent)
1601{
1602 struct nfs_server *server;
1603 struct nfs4_client *clp = NULL;
1604 struct rpc_xprt *xprt = NULL;
1605 struct rpc_clnt *clnt = NULL;
1606 struct rpc_timeout timeparms;
1607 rpc_authflavor_t authflavour;
1608 int proto, err = -EIO;
1609
1610 sb->s_blocksize_bits = 0;
1611 sb->s_blocksize = 0;
1612 server = NFS_SB(sb);
1613 if (data->rsize != 0)
1614 server->rsize = nfs_block_size(data->rsize, NULL);
1615 if (data->wsize != 0)
1616 server->wsize = nfs_block_size(data->wsize, NULL);
1617 server->flags = data->flags & NFS_MOUNT_FLAGMASK;
1618 server->caps = NFS_CAP_ATOMIC_OPEN;
1619
1620 server->acregmin = data->acregmin*HZ;
1621 server->acregmax = data->acregmax*HZ;
1622 server->acdirmin = data->acdirmin*HZ;
1623 server->acdirmax = data->acdirmax*HZ;
1624
1625 server->rpc_ops = &nfs_v4_clientops;
1626 /* Initialize timeout values */
1627
1628 timeparms.to_initval = data->timeo * HZ / 10;
1629 timeparms.to_retries = data->retrans;
1630 timeparms.to_exponential = 1;
1631 if (!timeparms.to_retries)
1632 timeparms.to_retries = 5;
1633
1634 proto = data->proto;
1635 /* Which IP protocol do we use? */
1636 switch (proto) {
1637 case IPPROTO_TCP:
1638 timeparms.to_maxval = RPC_MAX_TCP_TIMEOUT;
1639 if (!timeparms.to_initval)
1640 timeparms.to_initval = 600 * HZ / 10;
1641 break;
1642 case IPPROTO_UDP:
1643 timeparms.to_maxval = RPC_MAX_UDP_TIMEOUT;
1644 if (!timeparms.to_initval)
1645 timeparms.to_initval = 11 * HZ / 10;
1646 break;
1647 default:
1648 return -EINVAL;
1649 }
1650
1651 clp = nfs4_get_client(&server->addr.sin_addr);
1652 if (!clp) {
Trond Myklebust9085bbc2005-06-22 17:16:20 +00001653 dprintk("%s: failed to create NFS4 client.\n", __FUNCTION__);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001654 return -EIO;
1655 }
1656
1657 /* Now create transport and client */
1658 authflavour = RPC_AUTH_UNIX;
1659 if (data->auth_flavourlen != 0) {
Trond Myklebust9085bbc2005-06-22 17:16:20 +00001660 if (data->auth_flavourlen != 1) {
1661 dprintk("%s: Invalid number of RPC auth flavours %d.\n",
1662 __FUNCTION__, data->auth_flavourlen);
1663 err = -EINVAL;
1664 goto out_fail;
1665 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001666 if (copy_from_user(&authflavour, data->auth_flavours, sizeof(authflavour))) {
1667 err = -EFAULT;
1668 goto out_fail;
1669 }
1670 }
1671
1672 down_write(&clp->cl_sem);
J. Bruce Fields6a192752005-06-22 17:16:23 +00001673 if (IS_ERR(clp->cl_rpcclient)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001674 xprt = xprt_create_proto(proto, &server->addr, &timeparms);
1675 if (IS_ERR(xprt)) {
1676 up_write(&clp->cl_sem);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001677 err = PTR_ERR(xprt);
Trond Myklebust9085bbc2005-06-22 17:16:20 +00001678 dprintk("%s: cannot create RPC transport. Error = %d\n",
1679 __FUNCTION__, err);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001680 goto out_fail;
1681 }
1682 clnt = rpc_create_client(xprt, server->hostname, &nfs_program,
1683 server->rpc_ops->version, authflavour);
1684 if (IS_ERR(clnt)) {
1685 up_write(&clp->cl_sem);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001686 err = PTR_ERR(clnt);
Trond Myklebust9085bbc2005-06-22 17:16:20 +00001687 dprintk("%s: cannot create RPC client. Error = %d\n",
1688 __FUNCTION__, err);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001689 goto out_fail;
1690 }
1691 clnt->cl_intr = 1;
1692 clnt->cl_softrtry = 1;
1693 clnt->cl_chatty = 1;
1694 clp->cl_rpcclient = clnt;
1695 clp->cl_cred = rpcauth_lookupcred(clnt->cl_auth, 0);
1696 if (IS_ERR(clp->cl_cred)) {
1697 up_write(&clp->cl_sem);
1698 err = PTR_ERR(clp->cl_cred);
1699 clp->cl_cred = NULL;
1700 goto out_fail;
1701 }
1702 memcpy(clp->cl_ipaddr, server->ip_addr, sizeof(clp->cl_ipaddr));
1703 nfs_idmap_new(clp);
1704 }
1705 if (list_empty(&clp->cl_superblocks)) {
1706 err = nfs4_init_client(clp);
1707 if (err != 0) {
1708 up_write(&clp->cl_sem);
1709 goto out_fail;
1710 }
1711 }
1712 list_add_tail(&server->nfs4_siblings, &clp->cl_superblocks);
1713 clnt = rpc_clone_client(clp->cl_rpcclient);
1714 if (!IS_ERR(clnt))
1715 server->nfs4_state = clp;
1716 up_write(&clp->cl_sem);
1717 clp = NULL;
1718
1719 if (IS_ERR(clnt)) {
Trond Myklebust9085bbc2005-06-22 17:16:20 +00001720 err = PTR_ERR(clnt);
1721 dprintk("%s: cannot create RPC client. Error = %d\n",
1722 __FUNCTION__, err);
1723 return err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001724 }
1725
1726 server->client = clnt;
1727
1728 if (server->nfs4_state->cl_idmap == NULL) {
Trond Myklebust9085bbc2005-06-22 17:16:20 +00001729 dprintk("%s: failed to create idmapper.\n", __FUNCTION__);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001730 return -ENOMEM;
1731 }
1732
1733 if (clnt->cl_auth->au_flavor != authflavour) {
J. Bruce Fields6a192752005-06-22 17:16:23 +00001734 struct rpc_auth *auth;
1735
1736 auth = rpcauth_create(authflavour, clnt);
1737 if (IS_ERR(auth)) {
Trond Myklebust9085bbc2005-06-22 17:16:20 +00001738 dprintk("%s: couldn't create credcache!\n", __FUNCTION__);
J. Bruce Fields6a192752005-06-22 17:16:23 +00001739 return PTR_ERR(auth);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001740 }
1741 }
1742
1743 sb->s_time_gran = 1;
1744
1745 sb->s_op = &nfs4_sops;
1746 err = nfs_sb_init(sb, authflavour);
1747 if (err == 0)
1748 return 0;
1749out_fail:
1750 if (clp)
1751 nfs4_put_client(clp);
1752 return err;
1753}
1754
1755static int nfs4_compare_super(struct super_block *sb, void *data)
1756{
1757 struct nfs_server *server = data;
1758 struct nfs_server *old = NFS_SB(sb);
1759
1760 if (strcmp(server->hostname, old->hostname) != 0)
1761 return 0;
1762 if (strcmp(server->mnt_path, old->mnt_path) != 0)
1763 return 0;
1764 return 1;
1765}
1766
1767static void *
1768nfs_copy_user_string(char *dst, struct nfs_string *src, int maxlen)
1769{
1770 void *p = NULL;
1771
1772 if (!src->len)
1773 return ERR_PTR(-EINVAL);
1774 if (src->len < maxlen)
1775 maxlen = src->len;
1776 if (dst == NULL) {
1777 p = dst = kmalloc(maxlen + 1, GFP_KERNEL);
1778 if (p == NULL)
1779 return ERR_PTR(-ENOMEM);
1780 }
1781 if (copy_from_user(dst, src->data, maxlen)) {
1782 if (p != NULL)
1783 kfree(p);
1784 return ERR_PTR(-EFAULT);
1785 }
1786 dst[maxlen] = '\0';
1787 return dst;
1788}
1789
1790static struct super_block *nfs4_get_sb(struct file_system_type *fs_type,
1791 int flags, const char *dev_name, void *raw_data)
1792{
1793 int error;
1794 struct nfs_server *server;
1795 struct super_block *s;
1796 struct nfs4_mount_data *data = raw_data;
1797 void *p;
1798
Trond Myklebust9085bbc2005-06-22 17:16:20 +00001799 if (data == NULL) {
1800 dprintk("%s: missing data argument\n", __FUNCTION__);
1801 return ERR_PTR(-EINVAL);
1802 }
1803 if (data->version <= 0 || data->version > NFS4_MOUNT_VERSION) {
1804 dprintk("%s: bad mount version\n", __FUNCTION__);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001805 return ERR_PTR(-EINVAL);
1806 }
1807
1808 server = kmalloc(sizeof(struct nfs_server), GFP_KERNEL);
1809 if (!server)
1810 return ERR_PTR(-ENOMEM);
1811 memset(server, 0, sizeof(struct nfs_server));
1812 /* Zero out the NFS state stuff */
1813 init_nfsv4_state(server);
Andreas Gruenbacherb7fa0552005-06-22 17:16:27 +00001814 server->client = server->client_sys = server->client_acl = ERR_PTR(-EINVAL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001815
Linus Torvalds1da177e2005-04-16 15:20:36 -07001816 p = nfs_copy_user_string(NULL, &data->hostname, 256);
1817 if (IS_ERR(p))
1818 goto out_err;
1819 server->hostname = p;
1820
1821 p = nfs_copy_user_string(NULL, &data->mnt_path, 1024);
1822 if (IS_ERR(p))
1823 goto out_err;
1824 server->mnt_path = p;
1825
1826 p = nfs_copy_user_string(server->ip_addr, &data->client_addr,
1827 sizeof(server->ip_addr) - 1);
1828 if (IS_ERR(p))
1829 goto out_err;
1830
1831 /* We now require that the mount process passes the remote address */
1832 if (data->host_addrlen != sizeof(server->addr)) {
1833 s = ERR_PTR(-EINVAL);
1834 goto out_free;
1835 }
1836 if (copy_from_user(&server->addr, data->host_addr, sizeof(server->addr))) {
1837 s = ERR_PTR(-EFAULT);
1838 goto out_free;
1839 }
1840 if (server->addr.sin_family != AF_INET ||
1841 server->addr.sin_addr.s_addr == INADDR_ANY) {
Trond Myklebust9085bbc2005-06-22 17:16:20 +00001842 dprintk("%s: mount program didn't pass remote IP address!\n",
1843 __FUNCTION__);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001844 s = ERR_PTR(-EINVAL);
1845 goto out_free;
1846 }
1847
Trond Myklebust9085bbc2005-06-22 17:16:20 +00001848 /* Fire up rpciod if not yet running */
1849 s = ERR_PTR(rpciod_up());
1850 if (IS_ERR(s)) {
1851 dprintk("%s: couldn't start rpciod! Error = %ld\n",
1852 __FUNCTION__, PTR_ERR(s));
1853 goto out_free;
1854 }
1855
Linus Torvalds1da177e2005-04-16 15:20:36 -07001856 s = sget(fs_type, nfs4_compare_super, nfs_set_super, server);
1857
1858 if (IS_ERR(s) || s->s_root)
1859 goto out_free;
1860
1861 s->s_flags = flags;
1862
Linus Torvalds1da177e2005-04-16 15:20:36 -07001863 error = nfs4_fill_super(s, data, flags & MS_VERBOSE ? 1 : 0);
1864 if (error) {
1865 up_write(&s->s_umount);
1866 deactivate_super(s);
1867 return ERR_PTR(error);
1868 }
1869 s->s_flags |= MS_ACTIVE;
1870 return s;
1871out_err:
1872 s = (struct super_block *)p;
1873out_free:
1874 if (server->mnt_path)
1875 kfree(server->mnt_path);
1876 if (server->hostname)
1877 kfree(server->hostname);
1878 kfree(server);
1879 return s;
1880}
1881
1882static void nfs4_kill_super(struct super_block *sb)
1883{
1884 struct nfs_server *server = NFS_SB(sb);
1885
1886 nfs_return_all_delegations(sb);
1887 kill_anon_super(sb);
1888
1889 nfs4_renewd_prepare_shutdown(server);
1890
1891 if (server->client != NULL && !IS_ERR(server->client))
1892 rpc_shutdown_client(server->client);
1893 rpciod_down(); /* release rpciod */
1894
1895 destroy_nfsv4_state(server);
1896
1897 if (server->hostname != NULL)
1898 kfree(server->hostname);
1899 kfree(server);
1900}
1901
1902static struct file_system_type nfs4_fs_type = {
1903 .owner = THIS_MODULE,
1904 .name = "nfs4",
1905 .get_sb = nfs4_get_sb,
1906 .kill_sb = nfs4_kill_super,
1907 .fs_flags = FS_ODD_RENAME|FS_REVAL_DOT|FS_BINARY_MOUNTDATA,
1908};
1909
1910#define nfs4_init_once(nfsi) \
1911 do { \
1912 INIT_LIST_HEAD(&(nfsi)->open_states); \
1913 nfsi->delegation = NULL; \
1914 nfsi->delegation_state = 0; \
1915 init_rwsem(&nfsi->rwsem); \
1916 } while(0)
1917#define register_nfs4fs() register_filesystem(&nfs4_fs_type)
1918#define unregister_nfs4fs() unregister_filesystem(&nfs4_fs_type)
1919#else
1920#define nfs4_init_once(nfsi) \
1921 do { } while (0)
1922#define register_nfs4fs() (0)
1923#define unregister_nfs4fs()
1924#endif
1925
1926extern int nfs_init_nfspagecache(void);
1927extern void nfs_destroy_nfspagecache(void);
1928extern int nfs_init_readpagecache(void);
1929extern void nfs_destroy_readpagecache(void);
1930extern int nfs_init_writepagecache(void);
1931extern void nfs_destroy_writepagecache(void);
1932#ifdef CONFIG_NFS_DIRECTIO
1933extern int nfs_init_directcache(void);
1934extern void nfs_destroy_directcache(void);
1935#endif
1936
1937static kmem_cache_t * nfs_inode_cachep;
1938
1939static struct inode *nfs_alloc_inode(struct super_block *sb)
1940{
1941 struct nfs_inode *nfsi;
1942 nfsi = (struct nfs_inode *)kmem_cache_alloc(nfs_inode_cachep, SLAB_KERNEL);
1943 if (!nfsi)
1944 return NULL;
1945 nfsi->flags = 0;
Trond Myklebust458818e2005-06-22 17:16:27 +00001946#ifdef CONFIG_NFS_V3_ACL
1947 nfsi->acl_access = ERR_PTR(-EAGAIN);
1948 nfsi->acl_default = ERR_PTR(-EAGAIN);
1949#endif
J. Bruce Fieldse50a1c22005-06-22 17:16:23 +00001950#ifdef CONFIG_NFS_V4
1951 nfsi->nfs4_acl = NULL;
1952#endif /* CONFIG_NFS_V4 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001953 return &nfsi->vfs_inode;
1954}
1955
1956static void nfs_destroy_inode(struct inode *inode)
1957{
1958 kmem_cache_free(nfs_inode_cachep, NFS_I(inode));
1959}
1960
1961static void init_once(void * foo, kmem_cache_t * cachep, unsigned long flags)
1962{
1963 struct nfs_inode *nfsi = (struct nfs_inode *) foo;
1964
1965 if ((flags & (SLAB_CTOR_VERIFY|SLAB_CTOR_CONSTRUCTOR)) ==
1966 SLAB_CTOR_CONSTRUCTOR) {
1967 inode_init_once(&nfsi->vfs_inode);
1968 spin_lock_init(&nfsi->req_lock);
1969 INIT_LIST_HEAD(&nfsi->dirty);
1970 INIT_LIST_HEAD(&nfsi->commit);
1971 INIT_LIST_HEAD(&nfsi->open_files);
1972 INIT_RADIX_TREE(&nfsi->nfs_page_tree, GFP_ATOMIC);
1973 atomic_set(&nfsi->data_updates, 0);
1974 nfsi->ndirty = 0;
1975 nfsi->ncommit = 0;
1976 nfsi->npages = 0;
1977 init_waitqueue_head(&nfsi->nfs_i_wait);
1978 nfs4_init_once(nfsi);
1979 }
1980}
1981
Adrian Bunk75c96f82005-05-05 16:16:09 -07001982static int nfs_init_inodecache(void)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001983{
1984 nfs_inode_cachep = kmem_cache_create("nfs_inode_cache",
1985 sizeof(struct nfs_inode),
1986 0, SLAB_RECLAIM_ACCOUNT,
1987 init_once, NULL);
1988 if (nfs_inode_cachep == NULL)
1989 return -ENOMEM;
1990
1991 return 0;
1992}
1993
Adrian Bunk75c96f82005-05-05 16:16:09 -07001994static void nfs_destroy_inodecache(void)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001995{
1996 if (kmem_cache_destroy(nfs_inode_cachep))
1997 printk(KERN_INFO "nfs_inode_cache: not all structures were freed\n");
1998}
1999
2000/*
2001 * Initialize NFS
2002 */
2003static int __init init_nfs_fs(void)
2004{
2005 int err;
2006
2007 err = nfs_init_nfspagecache();
2008 if (err)
2009 goto out4;
2010
2011 err = nfs_init_inodecache();
2012 if (err)
2013 goto out3;
2014
2015 err = nfs_init_readpagecache();
2016 if (err)
2017 goto out2;
2018
2019 err = nfs_init_writepagecache();
2020 if (err)
2021 goto out1;
2022
2023#ifdef CONFIG_NFS_DIRECTIO
2024 err = nfs_init_directcache();
2025 if (err)
2026 goto out0;
2027#endif
2028
2029#ifdef CONFIG_PROC_FS
2030 rpc_proc_register(&nfs_rpcstat);
2031#endif
2032 err = register_filesystem(&nfs_fs_type);
2033 if (err)
2034 goto out;
2035 if ((err = register_nfs4fs()) != 0)
2036 goto out;
2037 return 0;
2038out:
2039#ifdef CONFIG_PROC_FS
2040 rpc_proc_unregister("nfs");
2041#endif
2042 nfs_destroy_writepagecache();
2043#ifdef CONFIG_NFS_DIRECTIO
2044out0:
2045 nfs_destroy_directcache();
2046#endif
2047out1:
2048 nfs_destroy_readpagecache();
2049out2:
2050 nfs_destroy_inodecache();
2051out3:
2052 nfs_destroy_nfspagecache();
2053out4:
2054 return err;
2055}
2056
2057static void __exit exit_nfs_fs(void)
2058{
2059#ifdef CONFIG_NFS_DIRECTIO
2060 nfs_destroy_directcache();
2061#endif
2062 nfs_destroy_writepagecache();
2063 nfs_destroy_readpagecache();
2064 nfs_destroy_inodecache();
2065 nfs_destroy_nfspagecache();
2066#ifdef CONFIG_PROC_FS
2067 rpc_proc_unregister("nfs");
2068#endif
2069 unregister_filesystem(&nfs_fs_type);
2070 unregister_nfs4fs();
2071}
2072
2073/* Not quite true; I just maintain it */
2074MODULE_AUTHOR("Olaf Kirch <okir@monad.swb.de>");
2075MODULE_LICENSE("GPL");
2076
2077module_init(init_nfs_fs)
2078module_exit(exit_nfs_fs)