[PATCH] NFS: split nfsi->flags into two fields
[deliverable/linux.git] / fs / nfs / inode.c
CommitLineData
1da177e4
LT
1/*
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
4ce79717 42#include "nfs4_fs.h"
1da177e4
LT
43#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 *);
ada70d94 67static void nfs_zap_acl_cache(struct inode *);
1da177e4
LT
68
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
b7fa0554
AG
111#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
1da177e4
LT
126static 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
3da28eb1 138 ret = nfs_commit_inode(inode, flags);
1da177e4
LT
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
1da177e4
LT
160static 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));
ada70d94 168 nfs_zap_acl_cache(inode);
1da177e4
LT
169 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{
6a19275a 178 struct rpc_clnt *rpc = NFS_SB(sb)->client;
1da177e4
LT
179
180 /* -EIO all pending I/O */
6a19275a 181 if (!IS_ERR(rpc))
1da177e4 182 rpc_killall_tasks(rpc);
b7fa0554
AG
183 rpc = NFS_SB(sb)->client_acl;
184 if (!IS_ERR(rpc))
185 rpc_killall_tasks(rpc);
1da177e4
LT
186}
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)) {
9085bbcb
TM
385 dprintk("%s: cannot create RPC transport. Error = %ld\n",
386 __FUNCTION__, PTR_ERR(xprt));
1da177e4
LT
387 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)) {
9085bbcb
TM
392 dprintk("%s: cannot create RPC client. Error = %ld\n",
393 __FUNCTION__, PTR_ERR(xprt));
1da177e4
LT
394 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:
1da177e4
LT
404 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. */
1da177e4 447#ifdef CONFIG_NFS_V3
9085bbcb 448 if (server->flags & NFS_MOUNT_VER3) {
1da177e4
LT
449 server->rpc_ops = &nfs_v3_clientops;
450 server->caps |= NFS_CAP_READDIRPLUS;
1da177e4
LT
451 } else {
452 server->rpc_ops = &nfs_v2_clientops;
453 }
9085bbcb
TM
454#else
455 server->rpc_ops = &nfs_v2_clientops;
456#endif
1da177e4
LT
457
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) {
6a19275a
BF
470 struct rpc_auth *auth;
471
1da177e4
LT
472 server->client_sys = rpc_clone_client(server->client);
473 if (IS_ERR(server->client_sys))
474 return PTR_ERR(server->client_sys);
6a19275a
BF
475 auth = rpcauth_create(RPC_AUTH_UNIX, server->client_sys);
476 if (IS_ERR(auth))
477 return PTR_ERR(auth);
1da177e4
LT
478 } else {
479 atomic_inc(&server->client->cl_count);
480 server->client_sys = server->client;
481 }
1da177e4 482 if (server->flags & NFS_MOUNT_VER3) {
b7fa0554
AG
483#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 */
055ffbea
AG
493 /*
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;
1da177e4
LT
498 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" },
b7fa0554 581 { NFS_MOUNT_NOACL, ",noacl", "" },
1da177e4
LT
582 { 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))
55296809 623 nfsi->cache_validity |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_DATA|NFS_INO_INVALID_ACCESS|NFS_INO_INVALID_ACL|NFS_INO_REVAL_PAGECACHE;
1da177e4 624 else
55296809 625 nfsi->cache_validity |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_ACCESS|NFS_INO_INVALID_ACL|NFS_INO_REVAL_PAGECACHE;
ada70d94
TM
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);
55296809 635 NFS_I(inode)->cache_validity &= ~NFS_INO_INVALID_ACL;
1da177e4
LT
636}
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 */
92cfc62c 732 inode->i_op = NFS_SB(sb)->rpc_ops->file_inode_ops;
1da177e4
LT
733 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);
65e4308d 817 if (error == 0)
1da177e4 818 nfs_refresh_inode(inode, &fattr);
65e4308d
TM
819 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) {
1da177e4 835 if ((attr->ia_valid & ATTR_MODE) != 0) {
65e4308d
TM
836 int mode = attr->ia_mode & S_IALLUGO;
837 mode |= inode->i_mode & ~S_IALLUGO;
1da177e4
LT
838 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;
55296809 844 NFS_I(inode)->cache_validity |= NFS_INO_INVALID_ACCESS|NFS_INO_INVALID_ACL;
65e4308d
TM
845 }
846 if ((attr->ia_valid & ATTR_SIZE) != 0) {
847 inode->i_size = attr->ia_size;
848 vmtruncate(inode, attr->ia_size);
849 }
1da177e4
LT
850}
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;
55296809 875 int need_atime = NFS_I(inode)->cache_validity & NFS_INO_INVALID_ATIME;
1da177e4
LT
876 int err;
877
878 if (__IS_FLG(inode, MS_NOATIME))
879 need_atime = 0;
880 else if (__IS_FLG(inode, MS_NODIRATIME) && S_ISDIR(inode->i_mode))
881 need_atime = 0;
882 /* We may force a getattr if the user cares about atime */
883 if (need_atime)
884 err = __nfs_revalidate_inode(NFS_SERVER(inode), inode);
885 else
886 err = nfs_revalidate_inode(NFS_SERVER(inode), inode);
887 if (!err)
888 generic_fillattr(inode, stat);
889 return err;
890}
891
892struct nfs_open_context *alloc_nfs_open_context(struct dentry *dentry, struct rpc_cred *cred)
893{
894 struct nfs_open_context *ctx;
895
896 ctx = (struct nfs_open_context *)kmalloc(sizeof(*ctx), GFP_KERNEL);
897 if (ctx != NULL) {
898 atomic_set(&ctx->count, 1);
899 ctx->dentry = dget(dentry);
900 ctx->cred = get_rpccred(cred);
901 ctx->state = NULL;
902 ctx->lockowner = current->files;
903 ctx->error = 0;
00a92642 904 ctx->dir_cookie = 0;
1da177e4
LT
905 }
906 return ctx;
907}
908
909struct nfs_open_context *get_nfs_open_context(struct nfs_open_context *ctx)
910{
911 if (ctx != NULL)
912 atomic_inc(&ctx->count);
913 return ctx;
914}
915
916void put_nfs_open_context(struct nfs_open_context *ctx)
917{
918 if (atomic_dec_and_test(&ctx->count)) {
919 if (!list_empty(&ctx->list)) {
920 struct inode *inode = ctx->dentry->d_inode;
921 spin_lock(&inode->i_lock);
922 list_del(&ctx->list);
923 spin_unlock(&inode->i_lock);
924 }
925 if (ctx->state != NULL)
926 nfs4_close_state(ctx->state, ctx->mode);
927 if (ctx->cred != NULL)
928 put_rpccred(ctx->cred);
929 dput(ctx->dentry);
930 kfree(ctx);
931 }
932}
933
934/*
935 * Ensure that mmap has a recent RPC credential for use when writing out
936 * shared pages
937 */
938void nfs_file_set_open_context(struct file *filp, struct nfs_open_context *ctx)
939{
940 struct inode *inode = filp->f_dentry->d_inode;
941 struct nfs_inode *nfsi = NFS_I(inode);
942
943 filp->private_data = get_nfs_open_context(ctx);
944 spin_lock(&inode->i_lock);
945 list_add(&ctx->list, &nfsi->open_files);
946 spin_unlock(&inode->i_lock);
947}
948
949struct nfs_open_context *nfs_find_open_context(struct inode *inode, int mode)
950{
951 struct nfs_inode *nfsi = NFS_I(inode);
952 struct nfs_open_context *pos, *ctx = NULL;
953
954 spin_lock(&inode->i_lock);
955 list_for_each_entry(pos, &nfsi->open_files, list) {
956 if ((pos->mode & mode) == mode) {
957 ctx = get_nfs_open_context(pos);
958 break;
959 }
960 }
961 spin_unlock(&inode->i_lock);
962 return ctx;
963}
964
965void nfs_file_clear_open_context(struct file *filp)
966{
967 struct inode *inode = filp->f_dentry->d_inode;
968 struct nfs_open_context *ctx = (struct nfs_open_context *)filp->private_data;
969
970 if (ctx) {
971 filp->private_data = NULL;
972 spin_lock(&inode->i_lock);
973 list_move_tail(&ctx->list, &NFS_I(inode)->open_files);
974 spin_unlock(&inode->i_lock);
975 put_nfs_open_context(ctx);
976 }
977}
978
979/*
980 * These allocate and release file read/write context information.
981 */
982int nfs_open(struct inode *inode, struct file *filp)
983{
984 struct nfs_open_context *ctx;
985 struct rpc_cred *cred;
986
987 cred = rpcauth_lookupcred(NFS_CLIENT(inode)->cl_auth, 0);
988 if (IS_ERR(cred))
989 return PTR_ERR(cred);
990 ctx = alloc_nfs_open_context(filp->f_dentry, cred);
991 put_rpccred(cred);
992 if (ctx == NULL)
993 return -ENOMEM;
994 ctx->mode = filp->f_mode;
995 nfs_file_set_open_context(filp, ctx);
996 put_nfs_open_context(ctx);
997 if ((filp->f_mode & FMODE_WRITE) != 0)
998 nfs_begin_data_update(inode);
999 return 0;
1000}
1001
1002int nfs_release(struct inode *inode, struct file *filp)
1003{
1004 if ((filp->f_mode & FMODE_WRITE) != 0)
1005 nfs_end_data_update(inode);
1006 nfs_file_clear_open_context(filp);
1007 return 0;
1008}
1009
1010/*
1011 * This function is called whenever some part of NFS notices that
1012 * the cached attributes have to be refreshed.
1013 */
1014int
1015__nfs_revalidate_inode(struct nfs_server *server, struct inode *inode)
1016{
1017 int status = -ESTALE;
1018 struct nfs_fattr fattr;
1019 struct nfs_inode *nfsi = NFS_I(inode);
1020 unsigned long verifier;
55296809 1021 unsigned long cache_validity;
1da177e4
LT
1022
1023 dfprintk(PAGECACHE, "NFS: revalidating (%s/%Ld)\n",
1024 inode->i_sb->s_id, (long long)NFS_FILEID(inode));
1025
1026 lock_kernel();
1027 if (!inode || is_bad_inode(inode))
1028 goto out_nowait;
1029 if (NFS_STALE(inode))
1030 goto out_nowait;
1031
1032 while (NFS_REVALIDATING(inode)) {
1033 status = nfs_wait_on_inode(inode, NFS_INO_REVALIDATING);
1034 if (status < 0)
1035 goto out_nowait;
1036 if (NFS_ATTRTIMEO(inode) == 0)
1037 continue;
55296809 1038 if (nfsi->cache_validity & (NFS_INO_INVALID_ATTR|NFS_INO_INVALID_DATA|NFS_INO_INVALID_ATIME))
1da177e4
LT
1039 continue;
1040 status = NFS_STALE(inode) ? -ESTALE : 0;
1041 goto out_nowait;
1042 }
1043 NFS_FLAGS(inode) |= NFS_INO_REVALIDATING;
1044
1045 /* Protect against RPC races by saving the change attribute */
1046 verifier = nfs_save_change_attribute(inode);
1047 status = NFS_PROTO(inode)->getattr(server, NFS_FH(inode), &fattr);
1048 if (status != 0) {
1049 dfprintk(PAGECACHE, "nfs_revalidate_inode: (%s/%Ld) getattr failed, error=%d\n",
1050 inode->i_sb->s_id,
1051 (long long)NFS_FILEID(inode), status);
1052 if (status == -ESTALE) {
1053 nfs_zap_caches(inode);
1054 if (!S_ISDIR(inode->i_mode))
1055 NFS_FLAGS(inode) |= NFS_INO_STALE;
1056 }
1057 goto out;
1058 }
1059
1060 status = nfs_update_inode(inode, &fattr, verifier);
1061 if (status) {
1062 dfprintk(PAGECACHE, "nfs_revalidate_inode: (%s/%Ld) refresh failed, error=%d\n",
1063 inode->i_sb->s_id,
1064 (long long)NFS_FILEID(inode), status);
1065 goto out;
1066 }
55296809
CL
1067 cache_validity = nfsi->cache_validity;
1068 nfsi->cache_validity &= ~NFS_INO_REVAL_PAGECACHE;
1069
1da177e4
LT
1070 /*
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)
55296809
CL
1075 nfsi->cache_validity &= ~(NFS_INO_INVALID_ATTR|NFS_INO_INVALID_ATIME);
1076
7d52e862 1077 nfs_revalidate_mapping(inode, inode->i_mapping);
55296809
CL
1078
1079 if (cache_validity & NFS_INO_INVALID_ACL)
ada70d94 1080 nfs_zap_acl_cache(inode);
55296809 1081
1da177e4
LT
1082 dfprintk(PAGECACHE, "NFS: (%s/%Ld) revalidation complete\n",
1083 inode->i_sb->s_id,
1084 (long long)NFS_FILEID(inode));
1085
1086out:
1087 NFS_FLAGS(inode) &= ~NFS_INO_REVALIDATING;
1088 wake_up(&nfsi->nfs_i_wait);
1089 out_nowait:
1090 unlock_kernel();
1091 return status;
1092}
1093
1094int nfs_attribute_timeout(struct inode *inode)
1095{
1096 struct nfs_inode *nfsi = NFS_I(inode);
1097
1098 if (nfs_have_delegation(inode, FMODE_READ))
1099 return 0;
1100 return time_after(jiffies, nfsi->read_cache_jiffies+nfsi->attrtimeo);
1101}
1102
1103/**
1104 * nfs_revalidate_inode - Revalidate the inode attributes
1105 * @server - pointer to nfs_server struct
1106 * @inode - pointer to inode struct
1107 *
1108 * Updates inode attribute information by retrieving the data from the server.
1109 */
1110int nfs_revalidate_inode(struct nfs_server *server, struct inode *inode)
1111{
55296809 1112 if (!(NFS_I(inode)->cache_validity & (NFS_INO_INVALID_ATTR|NFS_INO_INVALID_DATA))
1da177e4
LT
1113 && !nfs_attribute_timeout(inode))
1114 return NFS_STALE(inode) ? -ESTALE : 0;
1115 return __nfs_revalidate_inode(server, inode);
1116}
1117
7d52e862
TM
1118/**
1119 * nfs_revalidate_mapping - Revalidate the pagecache
1120 * @inode - pointer to host inode
1121 * @mapping - pointer to mapping
1122 */
1123void nfs_revalidate_mapping(struct inode *inode, struct address_space *mapping)
1124{
1125 struct nfs_inode *nfsi = NFS_I(inode);
1126
55296809 1127 if (nfsi->cache_validity & NFS_INO_INVALID_DATA) {
7d52e862
TM
1128 if (S_ISREG(inode->i_mode)) {
1129 if (filemap_fdatawrite(mapping) == 0)
1130 filemap_fdatawait(mapping);
1131 nfs_wb_all(inode);
1132 }
1133 invalidate_inode_pages2(mapping);
55296809 1134 nfsi->cache_validity &= ~NFS_INO_INVALID_DATA;
7d52e862
TM
1135 if (S_ISDIR(inode->i_mode)) {
1136 memset(nfsi->cookieverf, 0, sizeof(nfsi->cookieverf));
1137 /* This ensures we revalidate child dentries */
1138 nfsi->cache_change_attribute++;
1139 }
1140 dfprintk(PAGECACHE, "NFS: (%s/%Ld) data cache invalidated\n",
1141 inode->i_sb->s_id,
1142 (long long)NFS_FILEID(inode));
1143 }
1144}
1145
1da177e4
LT
1146/**
1147 * nfs_begin_data_update
1148 * @inode - pointer to inode
1149 * Declare that a set of operations will update file data on the server
1150 */
1151void nfs_begin_data_update(struct inode *inode)
1152{
1153 atomic_inc(&NFS_I(inode)->data_updates);
1154}
1155
1156/**
1157 * nfs_end_data_update
1158 * @inode - pointer to inode
1159 * Declare end of the operations that will update file data
1160 * This will mark the inode as immediately needing revalidation
1161 * of its attribute cache.
1162 */
1163void nfs_end_data_update(struct inode *inode)
1164{
1165 struct nfs_inode *nfsi = NFS_I(inode);
1166
1167 if (!nfs_have_delegation(inode, FMODE_READ)) {
1168 /* Mark the attribute cache for revalidation */
55296809 1169 nfsi->cache_validity |= NFS_INO_INVALID_ATTR;
1da177e4
LT
1170 /* Directories and symlinks: invalidate page cache too */
1171 if (S_ISDIR(inode->i_mode) || S_ISLNK(inode->i_mode))
55296809 1172 nfsi->cache_validity |= NFS_INO_INVALID_DATA;
1da177e4
LT
1173 }
1174 nfsi->cache_change_attribute ++;
1175 atomic_dec(&nfsi->data_updates);
1176}
1177
1da177e4
LT
1178/**
1179 * nfs_refresh_inode - verify consistency of the inode attribute cache
1180 * @inode - pointer to inode
1181 * @fattr - updated attributes
1182 *
1183 * Verifies the attribute cache. If we have just changed the attributes,
1184 * so that fattr carries weak cache consistency data, then it may
1185 * also update the ctime/mtime/change_attribute.
1186 */
1187int nfs_refresh_inode(struct inode *inode, struct nfs_fattr *fattr)
1188{
1189 struct nfs_inode *nfsi = NFS_I(inode);
1190 loff_t cur_size, new_isize;
1191 int data_unstable;
1192
1193 /* Do we hold a delegation? */
1194 if (nfs_have_delegation(inode, FMODE_READ))
1195 return 0;
1196
1197 /* Are we in the process of updating data on the server? */
1198 data_unstable = nfs_caches_unstable(inode);
1199
1200 if (fattr->valid & NFS_ATTR_FATTR_V4) {
1201 if ((fattr->valid & NFS_ATTR_PRE_CHANGE) != 0
1202 && nfsi->change_attr == fattr->pre_change_attr)
1203 nfsi->change_attr = fattr->change_attr;
fe51beec 1204 if (nfsi->change_attr != fattr->change_attr) {
55296809 1205 nfsi->cache_validity |= NFS_INO_INVALID_ATTR;
fe51beec 1206 if (!data_unstable)
55296809 1207 nfsi->cache_validity |= NFS_INO_REVAL_PAGECACHE;
fe51beec 1208 }
1da177e4
LT
1209 }
1210
1211 if ((fattr->valid & NFS_ATTR_FATTR) == 0)
1212 return 0;
1213
1214 /* Has the inode gone and changed behind our back? */
1215 if (nfsi->fileid != fattr->fileid
1216 || (inode->i_mode & S_IFMT) != (fattr->mode & S_IFMT))
1217 return -EIO;
1218
1219 cur_size = i_size_read(inode);
1220 new_isize = nfs_size_to_loff_t(fattr->size);
1221
1222 /* If we have atomic WCC data, we may update some attributes */
1223 if ((fattr->valid & NFS_ATTR_WCC) != 0) {
1224 if (timespec_equal(&inode->i_ctime, &fattr->pre_ctime))
1225 memcpy(&inode->i_ctime, &fattr->ctime, sizeof(inode->i_ctime));
1226 if (timespec_equal(&inode->i_mtime, &fattr->pre_mtime))
1227 memcpy(&inode->i_mtime, &fattr->mtime, sizeof(inode->i_mtime));
1228 }
1229
1230 /* Verify a few of the more important attributes */
fe51beec 1231 if (!timespec_equal(&inode->i_mtime, &fattr->mtime)) {
55296809 1232 nfsi->cache_validity |= NFS_INO_INVALID_ATTR;
fe51beec 1233 if (!data_unstable)
55296809 1234 nfsi->cache_validity |= NFS_INO_REVAL_PAGECACHE;
fe51beec
TM
1235 }
1236 if (cur_size != new_isize) {
55296809 1237 nfsi->cache_validity |= NFS_INO_INVALID_ATTR;
fe51beec 1238 if (nfsi->npages == 0)
55296809 1239 nfsi->cache_validity |= NFS_INO_REVAL_PAGECACHE;
fe51beec 1240 }
1da177e4
LT
1241
1242 /* Have any file permissions changed? */
1243 if ((inode->i_mode & S_IALLUGO) != (fattr->mode & S_IALLUGO)
1244 || inode->i_uid != fattr->uid
1245 || inode->i_gid != fattr->gid)
55296809 1246 nfsi->cache_validity |= NFS_INO_INVALID_ATTR | NFS_INO_INVALID_ACCESS | NFS_INO_INVALID_ACL;
1da177e4
LT
1247
1248 /* Has the link count changed? */
1249 if (inode->i_nlink != fattr->nlink)
55296809 1250 nfsi->cache_validity |= NFS_INO_INVALID_ATTR;
1da177e4
LT
1251
1252 if (!timespec_equal(&inode->i_atime, &fattr->atime))
55296809 1253 nfsi->cache_validity |= NFS_INO_INVALID_ATIME;
1da177e4
LT
1254
1255 nfsi->read_cache_jiffies = fattr->timestamp;
1256 return 0;
1257}
1258
1259/*
1260 * Many nfs protocol calls return the new file attributes after
1261 * an operation. Here we update the inode to reflect the state
1262 * of the server's inode.
1263 *
1264 * This is a bit tricky because we have to make sure all dirty pages
1265 * have been sent off to the server before calling invalidate_inode_pages.
1266 * To make sure no other process adds more write requests while we try
1267 * our best to flush them, we make them sleep during the attribute refresh.
1268 *
1269 * A very similar scenario holds for the dir cache.
1270 */
1271static int nfs_update_inode(struct inode *inode, struct nfs_fattr *fattr, unsigned long verifier)
1272{
1273 struct nfs_inode *nfsi = NFS_I(inode);
951a143b 1274 loff_t cur_isize, new_isize;
1da177e4 1275 unsigned int invalid = 0;
1da177e4
LT
1276 int data_unstable;
1277
1278 dfprintk(VFS, "NFS: %s(%s/%ld ct=%d info=0x%x)\n",
1279 __FUNCTION__, inode->i_sb->s_id, inode->i_ino,
1280 atomic_read(&inode->i_count), fattr->valid);
1281
1282 if ((fattr->valid & NFS_ATTR_FATTR) == 0)
1283 return 0;
1284
1285 if (nfsi->fileid != fattr->fileid) {
1286 printk(KERN_ERR "%s: inode number mismatch\n"
1287 "expected (%s/0x%Lx), got (%s/0x%Lx)\n",
1288 __FUNCTION__,
1289 inode->i_sb->s_id, (long long)nfsi->fileid,
1290 inode->i_sb->s_id, (long long)fattr->fileid);
1291 goto out_err;
1292 }
1293
1294 /*
1295 * Make sure the inode's type hasn't changed.
1296 */
1297 if ((inode->i_mode & S_IFMT) != (fattr->mode & S_IFMT))
1298 goto out_changed;
1299
1300 /*
1301 * Update the read time so we don't revalidate too often.
1302 */
1303 nfsi->read_cache_jiffies = fattr->timestamp;
1304
1305 /* Are we racing with known updates of the metadata on the server? */
1306 data_unstable = ! nfs_verify_change_attribute(inode, verifier);
1307
951a143b 1308 /* Check if our cached file size is stale */
1da177e4
LT
1309 new_isize = nfs_size_to_loff_t(fattr->size);
1310 cur_isize = i_size_read(inode);
951a143b
TM
1311 if (new_isize != cur_isize) {
1312 /* Do we perhaps have any outstanding writes? */
1313 if (nfsi->npages == 0) {
1314 /* No, but did we race with nfs_end_data_update()? */
1315 if (verifier == nfsi->cache_change_attribute) {
1da177e4 1316 inode->i_size = new_isize;
951a143b 1317 invalid |= NFS_INO_INVALID_DATA;
1da177e4 1318 }
951a143b
TM
1319 invalid |= NFS_INO_INVALID_ATTR;
1320 } else if (new_isize > cur_isize) {
1da177e4
LT
1321 inode->i_size = new_isize;
1322 invalid |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_DATA;
1323 }
951a143b
TM
1324 dprintk("NFS: isize change on server for file %s/%ld\n",
1325 inode->i_sb->s_id, inode->i_ino);
1da177e4
LT
1326 }
1327
951a143b 1328 /* Check if the mtime agrees */
1da177e4
LT
1329 if (!timespec_equal(&inode->i_mtime, &fattr->mtime)) {
1330 memcpy(&inode->i_mtime, &fattr->mtime, sizeof(inode->i_mtime));
951a143b
TM
1331 dprintk("NFS: mtime change on server for file %s/%ld\n",
1332 inode->i_sb->s_id, inode->i_ino);
1da177e4
LT
1333 if (!data_unstable)
1334 invalid |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_DATA;
1335 }
1336
1337 if ((fattr->valid & NFS_ATTR_FATTR_V4)
1338 && nfsi->change_attr != fattr->change_attr) {
951a143b 1339 dprintk("NFS: change_attr change on server for file %s/%ld\n",
1da177e4 1340 inode->i_sb->s_id, inode->i_ino);
1da177e4
LT
1341 nfsi->change_attr = fattr->change_attr;
1342 if (!data_unstable)
ada70d94 1343 invalid |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_DATA|NFS_INO_INVALID_ACCESS|NFS_INO_INVALID_ACL;
1da177e4
LT
1344 }
1345
ada70d94
TM
1346 /* If ctime has changed we should definitely clear access+acl caches */
1347 if (!timespec_equal(&inode->i_ctime, &fattr->ctime)) {
1348 if (!data_unstable)
1349 invalid |= NFS_INO_INVALID_ACCESS|NFS_INO_INVALID_ACL;
1350 memcpy(&inode->i_ctime, &fattr->ctime, sizeof(inode->i_ctime));
1351 }
1da177e4
LT
1352 memcpy(&inode->i_atime, &fattr->atime, sizeof(inode->i_atime));
1353
1354 if ((inode->i_mode & S_IALLUGO) != (fattr->mode & S_IALLUGO) ||
1355 inode->i_uid != fattr->uid ||
1356 inode->i_gid != fattr->gid)
ada70d94 1357 invalid |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_ACCESS|NFS_INO_INVALID_ACL;
1da177e4
LT
1358
1359 inode->i_mode = fattr->mode;
1360 inode->i_nlink = fattr->nlink;
1361 inode->i_uid = fattr->uid;
1362 inode->i_gid = fattr->gid;
1363
1364 if (fattr->valid & (NFS_ATTR_FATTR_V3 | NFS_ATTR_FATTR_V4)) {
1365 /*
1366 * report the blocks in 512byte units
1367 */
1368 inode->i_blocks = nfs_calc_block_size(fattr->du.nfs3.used);
1369 inode->i_blksize = inode->i_sb->s_blocksize;
1370 } else {
1371 inode->i_blocks = fattr->du.nfs2.blocks;
1372 inode->i_blksize = fattr->du.nfs2.blocksize;
1373 }
1374
1375 /* Update attrtimeo value if we're out of the unstable period */
1376 if (invalid & NFS_INO_INVALID_ATTR) {
1377 nfsi->attrtimeo = NFS_MINATTRTIMEO(inode);
1378 nfsi->attrtimeo_timestamp = jiffies;
1379 } else if (time_after(jiffies, nfsi->attrtimeo_timestamp+nfsi->attrtimeo)) {
1380 if ((nfsi->attrtimeo <<= 1) > NFS_MAXATTRTIMEO(inode))
1381 nfsi->attrtimeo = NFS_MAXATTRTIMEO(inode);
1382 nfsi->attrtimeo_timestamp = jiffies;
1383 }
1384 /* Don't invalidate the data if we were to blame */
1385 if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode)
1386 || S_ISLNK(inode->i_mode)))
1387 invalid &= ~NFS_INO_INVALID_DATA;
1388 if (!nfs_have_delegation(inode, FMODE_READ))
55296809 1389 nfsi->cache_validity |= invalid;
1da177e4
LT
1390
1391 return 0;
1392 out_changed:
1393 /*
1394 * Big trouble! The inode has become a different object.
1395 */
1396#ifdef NFS_PARANOIA
1397 printk(KERN_DEBUG "%s: inode %ld mode changed, %07o to %07o\n",
1398 __FUNCTION__, inode->i_ino, inode->i_mode, fattr->mode);
1399#endif
1400 /*
1401 * No need to worry about unhashing the dentry, as the
1402 * lookup validation will know that the inode is bad.
1403 * (But we fall through to invalidate the caches.)
1404 */
1405 nfs_invalidate_inode(inode);
1406 out_err:
1407 NFS_FLAGS(inode) |= NFS_INO_STALE;
1408 return -ESTALE;
1409}
1410
1411/*
1412 * File system information
1413 */
1414
1415static int nfs_set_super(struct super_block *s, void *data)
1416{
1417 s->s_fs_info = data;
1418 return set_anon_super(s, data);
1419}
1420
1421static int nfs_compare_super(struct super_block *sb, void *data)
1422{
1423 struct nfs_server *server = data;
1424 struct nfs_server *old = NFS_SB(sb);
1425
1426 if (old->addr.sin_addr.s_addr != server->addr.sin_addr.s_addr)
1427 return 0;
1428 if (old->addr.sin_port != server->addr.sin_port)
1429 return 0;
1430 return !nfs_compare_fh(&old->fh, &server->fh);
1431}
1432
1433static struct super_block *nfs_get_sb(struct file_system_type *fs_type,
1434 int flags, const char *dev_name, void *raw_data)
1435{
1436 int error;
9085bbcb 1437 struct nfs_server *server = NULL;
1da177e4
LT
1438 struct super_block *s;
1439 struct nfs_fh *root;
1440 struct nfs_mount_data *data = raw_data;
1441
9085bbcb
TM
1442 s = ERR_PTR(-EINVAL);
1443 if (data == NULL) {
1444 dprintk("%s: missing data argument\n", __FUNCTION__);
1445 goto out_err;
1446 }
1447 if (data->version <= 0 || data->version > NFS_MOUNT_VERSION) {
1448 dprintk("%s: bad mount version\n", __FUNCTION__);
1449 goto out_err;
1da177e4 1450 }
9085bbcb
TM
1451 switch (data->version) {
1452 case 1:
1453 data->namlen = 0;
1454 case 2:
1455 data->bsize = 0;
1456 case 3:
1457 if (data->flags & NFS_MOUNT_VER3) {
1458 dprintk("%s: mount structure version %d does not support NFSv3\n",
1459 __FUNCTION__,
1460 data->version);
1461 goto out_err;
1462 }
1463 data->root.size = NFS2_FHSIZE;
1464 memcpy(data->root.data, data->old_root.data, NFS2_FHSIZE);
1465 case 4:
1466 if (data->flags & NFS_MOUNT_SECFLAVOUR) {
1467 dprintk("%s: mount structure version %d does not support strong security\n",
1468 __FUNCTION__,
1469 data->version);
1470 goto out_err;
1471 }
1472 case 5:
1473 memset(data->context, 0, sizeof(data->context));
1474 }
1475#ifndef CONFIG_NFS_V3
1476 /* If NFSv3 is not compiled in, return -EPROTONOSUPPORT */
1477 s = ERR_PTR(-EPROTONOSUPPORT);
1478 if (data->flags & NFS_MOUNT_VER3) {
1479 dprintk("%s: NFSv3 not compiled into kernel\n", __FUNCTION__);
1480 goto out_err;
1481 }
1482#endif /* CONFIG_NFS_V3 */
1da177e4 1483
9085bbcb 1484 s = ERR_PTR(-ENOMEM);
1da177e4
LT
1485 server = kmalloc(sizeof(struct nfs_server), GFP_KERNEL);
1486 if (!server)
9085bbcb 1487 goto out_err;
1da177e4
LT
1488 memset(server, 0, sizeof(struct nfs_server));
1489 /* Zero out the NFS state stuff */
1490 init_nfsv4_state(server);
b7fa0554 1491 server->client = server->client_sys = server->client_acl = ERR_PTR(-EINVAL);
1da177e4 1492
1da177e4
LT
1493 root = &server->fh;
1494 if (data->flags & NFS_MOUNT_VER3)
1495 root->size = data->root.size;
1496 else
1497 root->size = NFS2_FHSIZE;
9085bbcb 1498 s = ERR_PTR(-EINVAL);
1da177e4 1499 if (root->size > sizeof(root->data)) {
9085bbcb
TM
1500 dprintk("%s: invalid root filehandle\n", __FUNCTION__);
1501 goto out_err;
1da177e4
LT
1502 }
1503 memcpy(root->data, data->root.data, root->size);
1504
1505 /* We now require that the mount process passes the remote address */
1506 memcpy(&server->addr, &data->addr, sizeof(server->addr));
1507 if (server->addr.sin_addr.s_addr == INADDR_ANY) {
9085bbcb
TM
1508 dprintk("%s: mount program didn't pass remote address!\n",
1509 __FUNCTION__);
1510 goto out_err;
1da177e4
LT
1511 }
1512
9085bbcb
TM
1513 /* Fire up rpciod if not yet running */
1514 s = ERR_PTR(rpciod_up());
1515 if (IS_ERR(s)) {
1516 dprintk("%s: couldn't start rpciod! Error = %ld\n",
1517 __FUNCTION__, PTR_ERR(s));
1518 goto out_err;
1da177e4
LT
1519 }
1520
9085bbcb
TM
1521 s = sget(fs_type, nfs_compare_super, nfs_set_super, server);
1522 if (IS_ERR(s) || s->s_root)
1523 goto out_rpciod_down;
1da177e4 1524
9085bbcb 1525 s->s_flags = flags;
1da177e4
LT
1526
1527 error = nfs_fill_super(s, data, flags & MS_VERBOSE ? 1 : 0);
1528 if (error) {
1529 up_write(&s->s_umount);
1530 deactivate_super(s);
1531 return ERR_PTR(error);
1532 }
1533 s->s_flags |= MS_ACTIVE;
1534 return s;
9085bbcb
TM
1535out_rpciod_down:
1536 rpciod_down();
1537out_err:
1538 kfree(server);
1539 return s;
1da177e4
LT
1540}
1541
1542static void nfs_kill_super(struct super_block *s)
1543{
1544 struct nfs_server *server = NFS_SB(s);
1545
1546 kill_anon_super(s);
1547
6a19275a 1548 if (!IS_ERR(server->client))
1da177e4 1549 rpc_shutdown_client(server->client);
6a19275a 1550 if (!IS_ERR(server->client_sys))
1da177e4 1551 rpc_shutdown_client(server->client_sys);
b7fa0554
AG
1552 if (!IS_ERR(server->client_acl))
1553 rpc_shutdown_client(server->client_acl);
1da177e4
LT
1554
1555 if (!(server->flags & NFS_MOUNT_NONLM))
1556 lockd_down(); /* release rpc.lockd */
1557
1558 rpciod_down(); /* release rpciod */
1559
1560 if (server->hostname != NULL)
1561 kfree(server->hostname);
1562 kfree(server);
1563}
1564
1565static struct file_system_type nfs_fs_type = {
1566 .owner = THIS_MODULE,
1567 .name = "nfs",
1568 .get_sb = nfs_get_sb,
1569 .kill_sb = nfs_kill_super,
1570 .fs_flags = FS_ODD_RENAME|FS_REVAL_DOT|FS_BINARY_MOUNTDATA,
1571};
1572
1573#ifdef CONFIG_NFS_V4
1574
1575static void nfs4_clear_inode(struct inode *);
1576
1577
1578static struct super_operations nfs4_sops = {
1579 .alloc_inode = nfs_alloc_inode,
1580 .destroy_inode = nfs_destroy_inode,
1581 .write_inode = nfs_write_inode,
1582 .delete_inode = nfs_delete_inode,
1583 .statfs = nfs_statfs,
1584 .clear_inode = nfs4_clear_inode,
1585 .umount_begin = nfs_umount_begin,
1586 .show_options = nfs_show_options,
1587};
1588
1589/*
1590 * Clean out any remaining NFSv4 state that might be left over due
1591 * to open() calls that passed nfs_atomic_lookup, but failed to call
1592 * nfs_open().
1593 */
1594static void nfs4_clear_inode(struct inode *inode)
1595{
1596 struct nfs_inode *nfsi = NFS_I(inode);
1597
1598 /* If we are holding a delegation, return it! */
1599 if (nfsi->delegation != NULL)
1600 nfs_inode_return_delegation(inode);
1601 /* First call standard NFS clear_inode() code */
1602 nfs_clear_inode(inode);
1603 /* Now clear out any remaining state */
1604 while (!list_empty(&nfsi->open_states)) {
1605 struct nfs4_state *state;
1606
1607 state = list_entry(nfsi->open_states.next,
1608 struct nfs4_state,
1609 inode_states);
1610 dprintk("%s(%s/%Ld): found unclaimed NFSv4 state %p\n",
1611 __FUNCTION__,
1612 inode->i_sb->s_id,
1613 (long long)NFS_FILEID(inode),
1614 state);
1615 BUG_ON(atomic_read(&state->count) != 1);
1616 nfs4_close_state(state, state->state);
1617 }
1618}
1619
1620
1621static int nfs4_fill_super(struct super_block *sb, struct nfs4_mount_data *data, int silent)
1622{
1623 struct nfs_server *server;
1624 struct nfs4_client *clp = NULL;
1625 struct rpc_xprt *xprt = NULL;
1626 struct rpc_clnt *clnt = NULL;
1627 struct rpc_timeout timeparms;
1628 rpc_authflavor_t authflavour;
1629 int proto, err = -EIO;
1630
1631 sb->s_blocksize_bits = 0;
1632 sb->s_blocksize = 0;
1633 server = NFS_SB(sb);
1634 if (data->rsize != 0)
1635 server->rsize = nfs_block_size(data->rsize, NULL);
1636 if (data->wsize != 0)
1637 server->wsize = nfs_block_size(data->wsize, NULL);
1638 server->flags = data->flags & NFS_MOUNT_FLAGMASK;
1639 server->caps = NFS_CAP_ATOMIC_OPEN;
1640
1641 server->acregmin = data->acregmin*HZ;
1642 server->acregmax = data->acregmax*HZ;
1643 server->acdirmin = data->acdirmin*HZ;
1644 server->acdirmax = data->acdirmax*HZ;
1645
1646 server->rpc_ops = &nfs_v4_clientops;
1647 /* Initialize timeout values */
1648
1649 timeparms.to_initval = data->timeo * HZ / 10;
1650 timeparms.to_retries = data->retrans;
1651 timeparms.to_exponential = 1;
1652 if (!timeparms.to_retries)
1653 timeparms.to_retries = 5;
1654
1655 proto = data->proto;
1656 /* Which IP protocol do we use? */
1657 switch (proto) {
1658 case IPPROTO_TCP:
1659 timeparms.to_maxval = RPC_MAX_TCP_TIMEOUT;
1660 if (!timeparms.to_initval)
1661 timeparms.to_initval = 600 * HZ / 10;
1662 break;
1663 case IPPROTO_UDP:
1664 timeparms.to_maxval = RPC_MAX_UDP_TIMEOUT;
1665 if (!timeparms.to_initval)
1666 timeparms.to_initval = 11 * HZ / 10;
1667 break;
1668 default:
1669 return -EINVAL;
1670 }
1671
1672 clp = nfs4_get_client(&server->addr.sin_addr);
1673 if (!clp) {
9085bbcb 1674 dprintk("%s: failed to create NFS4 client.\n", __FUNCTION__);
1da177e4
LT
1675 return -EIO;
1676 }
1677
1678 /* Now create transport and client */
1679 authflavour = RPC_AUTH_UNIX;
1680 if (data->auth_flavourlen != 0) {
9085bbcb
TM
1681 if (data->auth_flavourlen != 1) {
1682 dprintk("%s: Invalid number of RPC auth flavours %d.\n",
1683 __FUNCTION__, data->auth_flavourlen);
1684 err = -EINVAL;
1685 goto out_fail;
1686 }
1da177e4
LT
1687 if (copy_from_user(&authflavour, data->auth_flavours, sizeof(authflavour))) {
1688 err = -EFAULT;
1689 goto out_fail;
1690 }
1691 }
1692
1693 down_write(&clp->cl_sem);
6a19275a 1694 if (IS_ERR(clp->cl_rpcclient)) {
1da177e4
LT
1695 xprt = xprt_create_proto(proto, &server->addr, &timeparms);
1696 if (IS_ERR(xprt)) {
1697 up_write(&clp->cl_sem);
1da177e4 1698 err = PTR_ERR(xprt);
9085bbcb
TM
1699 dprintk("%s: cannot create RPC transport. Error = %d\n",
1700 __FUNCTION__, err);
1da177e4
LT
1701 goto out_fail;
1702 }
1703 clnt = rpc_create_client(xprt, server->hostname, &nfs_program,
1704 server->rpc_ops->version, authflavour);
1705 if (IS_ERR(clnt)) {
1706 up_write(&clp->cl_sem);
1da177e4 1707 err = PTR_ERR(clnt);
9085bbcb
TM
1708 dprintk("%s: cannot create RPC client. Error = %d\n",
1709 __FUNCTION__, err);
1da177e4
LT
1710 goto out_fail;
1711 }
1712 clnt->cl_intr = 1;
1713 clnt->cl_softrtry = 1;
1714 clnt->cl_chatty = 1;
1715 clp->cl_rpcclient = clnt;
1716 clp->cl_cred = rpcauth_lookupcred(clnt->cl_auth, 0);
1717 if (IS_ERR(clp->cl_cred)) {
1718 up_write(&clp->cl_sem);
1719 err = PTR_ERR(clp->cl_cred);
1720 clp->cl_cred = NULL;
1721 goto out_fail;
1722 }
1723 memcpy(clp->cl_ipaddr, server->ip_addr, sizeof(clp->cl_ipaddr));
1724 nfs_idmap_new(clp);
1725 }
1726 if (list_empty(&clp->cl_superblocks)) {
1727 err = nfs4_init_client(clp);
1728 if (err != 0) {
1729 up_write(&clp->cl_sem);
1730 goto out_fail;
1731 }
1732 }
1733 list_add_tail(&server->nfs4_siblings, &clp->cl_superblocks);
1734 clnt = rpc_clone_client(clp->cl_rpcclient);
1735 if (!IS_ERR(clnt))
1736 server->nfs4_state = clp;
1737 up_write(&clp->cl_sem);
1738 clp = NULL;
1739
1740 if (IS_ERR(clnt)) {
9085bbcb
TM
1741 err = PTR_ERR(clnt);
1742 dprintk("%s: cannot create RPC client. Error = %d\n",
1743 __FUNCTION__, err);
1744 return err;
1da177e4
LT
1745 }
1746
1747 server->client = clnt;
1748
1749 if (server->nfs4_state->cl_idmap == NULL) {
9085bbcb 1750 dprintk("%s: failed to create idmapper.\n", __FUNCTION__);
1da177e4
LT
1751 return -ENOMEM;
1752 }
1753
1754 if (clnt->cl_auth->au_flavor != authflavour) {
6a19275a
BF
1755 struct rpc_auth *auth;
1756
1757 auth = rpcauth_create(authflavour, clnt);
1758 if (IS_ERR(auth)) {
9085bbcb 1759 dprintk("%s: couldn't create credcache!\n", __FUNCTION__);
6a19275a 1760 return PTR_ERR(auth);
1da177e4
LT
1761 }
1762 }
1763
1764 sb->s_time_gran = 1;
1765
1766 sb->s_op = &nfs4_sops;
1767 err = nfs_sb_init(sb, authflavour);
1768 if (err == 0)
1769 return 0;
1770out_fail:
1771 if (clp)
1772 nfs4_put_client(clp);
1773 return err;
1774}
1775
1776static int nfs4_compare_super(struct super_block *sb, void *data)
1777{
1778 struct nfs_server *server = data;
1779 struct nfs_server *old = NFS_SB(sb);
1780
1781 if (strcmp(server->hostname, old->hostname) != 0)
1782 return 0;
1783 if (strcmp(server->mnt_path, old->mnt_path) != 0)
1784 return 0;
1785 return 1;
1786}
1787
1788static void *
1789nfs_copy_user_string(char *dst, struct nfs_string *src, int maxlen)
1790{
1791 void *p = NULL;
1792
1793 if (!src->len)
1794 return ERR_PTR(-EINVAL);
1795 if (src->len < maxlen)
1796 maxlen = src->len;
1797 if (dst == NULL) {
1798 p = dst = kmalloc(maxlen + 1, GFP_KERNEL);
1799 if (p == NULL)
1800 return ERR_PTR(-ENOMEM);
1801 }
1802 if (copy_from_user(dst, src->data, maxlen)) {
1803 if (p != NULL)
1804 kfree(p);
1805 return ERR_PTR(-EFAULT);
1806 }
1807 dst[maxlen] = '\0';
1808 return dst;
1809}
1810
1811static struct super_block *nfs4_get_sb(struct file_system_type *fs_type,
1812 int flags, const char *dev_name, void *raw_data)
1813{
1814 int error;
1815 struct nfs_server *server;
1816 struct super_block *s;
1817 struct nfs4_mount_data *data = raw_data;
1818 void *p;
1819
9085bbcb
TM
1820 if (data == NULL) {
1821 dprintk("%s: missing data argument\n", __FUNCTION__);
1822 return ERR_PTR(-EINVAL);
1823 }
1824 if (data->version <= 0 || data->version > NFS4_MOUNT_VERSION) {
1825 dprintk("%s: bad mount version\n", __FUNCTION__);
1da177e4
LT
1826 return ERR_PTR(-EINVAL);
1827 }
1828
1829 server = kmalloc(sizeof(struct nfs_server), GFP_KERNEL);
1830 if (!server)
1831 return ERR_PTR(-ENOMEM);
1832 memset(server, 0, sizeof(struct nfs_server));
1833 /* Zero out the NFS state stuff */
1834 init_nfsv4_state(server);
b7fa0554 1835 server->client = server->client_sys = server->client_acl = ERR_PTR(-EINVAL);
1da177e4 1836
1da177e4
LT
1837 p = nfs_copy_user_string(NULL, &data->hostname, 256);
1838 if (IS_ERR(p))
1839 goto out_err;
1840 server->hostname = p;
1841
1842 p = nfs_copy_user_string(NULL, &data->mnt_path, 1024);
1843 if (IS_ERR(p))
1844 goto out_err;
1845 server->mnt_path = p;
1846
1847 p = nfs_copy_user_string(server->ip_addr, &data->client_addr,
1848 sizeof(server->ip_addr) - 1);
1849 if (IS_ERR(p))
1850 goto out_err;
1851
1852 /* We now require that the mount process passes the remote address */
1853 if (data->host_addrlen != sizeof(server->addr)) {
1854 s = ERR_PTR(-EINVAL);
1855 goto out_free;
1856 }
1857 if (copy_from_user(&server->addr, data->host_addr, sizeof(server->addr))) {
1858 s = ERR_PTR(-EFAULT);
1859 goto out_free;
1860 }
1861 if (server->addr.sin_family != AF_INET ||
1862 server->addr.sin_addr.s_addr == INADDR_ANY) {
9085bbcb
TM
1863 dprintk("%s: mount program didn't pass remote IP address!\n",
1864 __FUNCTION__);
1da177e4
LT
1865 s = ERR_PTR(-EINVAL);
1866 goto out_free;
1867 }
1868
9085bbcb
TM
1869 /* Fire up rpciod if not yet running */
1870 s = ERR_PTR(rpciod_up());
1871 if (IS_ERR(s)) {
1872 dprintk("%s: couldn't start rpciod! Error = %ld\n",
1873 __FUNCTION__, PTR_ERR(s));
1874 goto out_free;
1875 }
1876
1da177e4
LT
1877 s = sget(fs_type, nfs4_compare_super, nfs_set_super, server);
1878
1879 if (IS_ERR(s) || s->s_root)
1880 goto out_free;
1881
1882 s->s_flags = flags;
1883
1da177e4
LT
1884 error = nfs4_fill_super(s, data, flags & MS_VERBOSE ? 1 : 0);
1885 if (error) {
1886 up_write(&s->s_umount);
1887 deactivate_super(s);
1888 return ERR_PTR(error);
1889 }
1890 s->s_flags |= MS_ACTIVE;
1891 return s;
1892out_err:
1893 s = (struct super_block *)p;
1894out_free:
1895 if (server->mnt_path)
1896 kfree(server->mnt_path);
1897 if (server->hostname)
1898 kfree(server->hostname);
1899 kfree(server);
1900 return s;
1901}
1902
1903static void nfs4_kill_super(struct super_block *sb)
1904{
1905 struct nfs_server *server = NFS_SB(sb);
1906
1907 nfs_return_all_delegations(sb);
1908 kill_anon_super(sb);
1909
1910 nfs4_renewd_prepare_shutdown(server);
1911
1912 if (server->client != NULL && !IS_ERR(server->client))
1913 rpc_shutdown_client(server->client);
1914 rpciod_down(); /* release rpciod */
1915
1916 destroy_nfsv4_state(server);
1917
1918 if (server->hostname != NULL)
1919 kfree(server->hostname);
1920 kfree(server);
1921}
1922
1923static struct file_system_type nfs4_fs_type = {
1924 .owner = THIS_MODULE,
1925 .name = "nfs4",
1926 .get_sb = nfs4_get_sb,
1927 .kill_sb = nfs4_kill_super,
1928 .fs_flags = FS_ODD_RENAME|FS_REVAL_DOT|FS_BINARY_MOUNTDATA,
1929};
1930
1931#define nfs4_init_once(nfsi) \
1932 do { \
1933 INIT_LIST_HEAD(&(nfsi)->open_states); \
1934 nfsi->delegation = NULL; \
1935 nfsi->delegation_state = 0; \
1936 init_rwsem(&nfsi->rwsem); \
1937 } while(0)
1938#define register_nfs4fs() register_filesystem(&nfs4_fs_type)
1939#define unregister_nfs4fs() unregister_filesystem(&nfs4_fs_type)
1940#else
1941#define nfs4_init_once(nfsi) \
1942 do { } while (0)
1943#define register_nfs4fs() (0)
1944#define unregister_nfs4fs()
1945#endif
1946
1947extern int nfs_init_nfspagecache(void);
1948extern void nfs_destroy_nfspagecache(void);
1949extern int nfs_init_readpagecache(void);
1950extern void nfs_destroy_readpagecache(void);
1951extern int nfs_init_writepagecache(void);
1952extern void nfs_destroy_writepagecache(void);
1953#ifdef CONFIG_NFS_DIRECTIO
1954extern int nfs_init_directcache(void);
1955extern void nfs_destroy_directcache(void);
1956#endif
1957
1958static kmem_cache_t * nfs_inode_cachep;
1959
1960static struct inode *nfs_alloc_inode(struct super_block *sb)
1961{
1962 struct nfs_inode *nfsi;
1963 nfsi = (struct nfs_inode *)kmem_cache_alloc(nfs_inode_cachep, SLAB_KERNEL);
1964 if (!nfsi)
1965 return NULL;
55296809
CL
1966 nfsi->flags = 0UL;
1967 nfsi->cache_validity = 0UL;
458818ed
TM
1968#ifdef CONFIG_NFS_V3_ACL
1969 nfsi->acl_access = ERR_PTR(-EAGAIN);
1970 nfsi->acl_default = ERR_PTR(-EAGAIN);
1971#endif
e50a1c2e
BF
1972#ifdef CONFIG_NFS_V4
1973 nfsi->nfs4_acl = NULL;
1974#endif /* CONFIG_NFS_V4 */
1da177e4
LT
1975 return &nfsi->vfs_inode;
1976}
1977
1978static void nfs_destroy_inode(struct inode *inode)
1979{
1980 kmem_cache_free(nfs_inode_cachep, NFS_I(inode));
1981}
1982
1983static void init_once(void * foo, kmem_cache_t * cachep, unsigned long flags)
1984{
1985 struct nfs_inode *nfsi = (struct nfs_inode *) foo;
1986
1987 if ((flags & (SLAB_CTOR_VERIFY|SLAB_CTOR_CONSTRUCTOR)) ==
1988 SLAB_CTOR_CONSTRUCTOR) {
1989 inode_init_once(&nfsi->vfs_inode);
1990 spin_lock_init(&nfsi->req_lock);
1991 INIT_LIST_HEAD(&nfsi->dirty);
1992 INIT_LIST_HEAD(&nfsi->commit);
1993 INIT_LIST_HEAD(&nfsi->open_files);
1994 INIT_RADIX_TREE(&nfsi->nfs_page_tree, GFP_ATOMIC);
1995 atomic_set(&nfsi->data_updates, 0);
1996 nfsi->ndirty = 0;
1997 nfsi->ncommit = 0;
1998 nfsi->npages = 0;
1999 init_waitqueue_head(&nfsi->nfs_i_wait);
2000 nfs4_init_once(nfsi);
2001 }
2002}
2003
75c96f85 2004static int nfs_init_inodecache(void)
1da177e4
LT
2005{
2006 nfs_inode_cachep = kmem_cache_create("nfs_inode_cache",
2007 sizeof(struct nfs_inode),
2008 0, SLAB_RECLAIM_ACCOUNT,
2009 init_once, NULL);
2010 if (nfs_inode_cachep == NULL)
2011 return -ENOMEM;
2012
2013 return 0;
2014}
2015
75c96f85 2016static void nfs_destroy_inodecache(void)
1da177e4
LT
2017{
2018 if (kmem_cache_destroy(nfs_inode_cachep))
2019 printk(KERN_INFO "nfs_inode_cache: not all structures were freed\n");
2020}
2021
2022/*
2023 * Initialize NFS
2024 */
2025static int __init init_nfs_fs(void)
2026{
2027 int err;
2028
2029 err = nfs_init_nfspagecache();
2030 if (err)
2031 goto out4;
2032
2033 err = nfs_init_inodecache();
2034 if (err)
2035 goto out3;
2036
2037 err = nfs_init_readpagecache();
2038 if (err)
2039 goto out2;
2040
2041 err = nfs_init_writepagecache();
2042 if (err)
2043 goto out1;
2044
2045#ifdef CONFIG_NFS_DIRECTIO
2046 err = nfs_init_directcache();
2047 if (err)
2048 goto out0;
2049#endif
2050
2051#ifdef CONFIG_PROC_FS
2052 rpc_proc_register(&nfs_rpcstat);
2053#endif
2054 err = register_filesystem(&nfs_fs_type);
2055 if (err)
2056 goto out;
2057 if ((err = register_nfs4fs()) != 0)
2058 goto out;
2059 return 0;
2060out:
2061#ifdef CONFIG_PROC_FS
2062 rpc_proc_unregister("nfs");
2063#endif
2064 nfs_destroy_writepagecache();
2065#ifdef CONFIG_NFS_DIRECTIO
2066out0:
2067 nfs_destroy_directcache();
2068#endif
2069out1:
2070 nfs_destroy_readpagecache();
2071out2:
2072 nfs_destroy_inodecache();
2073out3:
2074 nfs_destroy_nfspagecache();
2075out4:
2076 return err;
2077}
2078
2079static void __exit exit_nfs_fs(void)
2080{
2081#ifdef CONFIG_NFS_DIRECTIO
2082 nfs_destroy_directcache();
2083#endif
2084 nfs_destroy_writepagecache();
2085 nfs_destroy_readpagecache();
2086 nfs_destroy_inodecache();
2087 nfs_destroy_nfspagecache();
2088#ifdef CONFIG_PROC_FS
2089 rpc_proc_unregister("nfs");
2090#endif
2091 unregister_filesystem(&nfs_fs_type);
2092 unregister_nfs4fs();
2093}
2094
2095/* Not quite true; I just maintain it */
2096MODULE_AUTHOR("Olaf Kirch <okir@monad.swb.de>");
2097MODULE_LICENSE("GPL");
2098
2099module_init(init_nfs_fs)
2100module_exit(exit_nfs_fs)
This page took 0.140349 seconds and 5 git commands to generate.