projects / linux-2.6-block.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
[PATCH] NFSv4: ACL support for the NFSv4 client: write
[linux-2.6-block.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
111static inline unsigned long
112nfs_fattr_to_ino_t(struct nfs_fattr *fattr)
113{
114 return nfs_fileid_to_ino_t(fattr->fileid);
115}
116
117static int
118nfs_write_inode(struct inode *inode, int sync)
119{
120 int flags = sync ? FLUSH_WAIT : 0;
121 int ret;
122
123 ret = nfs_commit_inode(inode, 0, 0, flags);
124 if (ret < 0)
125 return ret;
126 return 0;
127}
128
129static void
130nfs_delete_inode(struct inode * inode)
131{
132 dprintk("NFS: delete_inode(%s/%ld)\n", inode->i_sb->s_id, inode->i_ino);
133
134 nfs_wb_all(inode);
135 /*
136 * The following should never happen...
137 */
138 if (nfs_have_writebacks(inode)) {
139 printk(KERN_ERR "nfs_delete_inode: inode %ld has pending RPC requests\n", inode->i_ino);
140 }
141
142 clear_inode(inode);
143}
144
145/*
146 * For the moment, the only task for the NFS clear_inode method is to
147 * release the mmap credential
148 */
149static void
150nfs_clear_inode(struct inode *inode)
151{
152 struct nfs_inode *nfsi = NFS_I(inode);
153 struct rpc_cred *cred;
154
155 nfs_wb_all(inode);
156 BUG_ON (!list_empty(&nfsi->open_files));
ada70d94 157 nfs_zap_acl_cache(inode);
1da177e4
LT		 158 cred = nfsi->cache_access.cred;
159 if (cred)
160 put_rpccred(cred);
161 BUG_ON(atomic_read(&nfsi->data_updates) != 0);
162}
163
164void
165nfs_umount_begin(struct super_block *sb)
166{
167 struct nfs_server *server = NFS_SB(sb);
168 struct rpc_clnt *rpc;
169
170 /* -EIO all pending I/O */
171 if ((rpc = server->client) != NULL)
172 rpc_killall_tasks(rpc);
173}
174
175
176static inline unsigned long
177nfs_block_bits(unsigned long bsize, unsigned char *nrbitsp)
178{
179 /* make sure blocksize is a power of two */
180 if ((bsize & (bsize - 1)) || nrbitsp) {
181 unsigned char nrbits;
182
183 for (nrbits = 31; nrbits && !(bsize & (1 << nrbits)); nrbits--)
184 ;
185 bsize = 1 << nrbits;
186 if (nrbitsp)
187 *nrbitsp = nrbits;
188 }
189
190 return bsize;
191}
192
193/*
194 * Calculate the number of 512byte blocks used.
195 */
196static inline unsigned long
197nfs_calc_block_size(u64 tsize)
198{
199 loff_t used = (tsize + 511) >> 9;
200 return (used > ULONG_MAX) ? ULONG_MAX : used;
201}
202
203/*
204 * Compute and set NFS server blocksize
205 */
206static inline unsigned long
207nfs_block_size(unsigned long bsize, unsigned char *nrbitsp)
208{
209 if (bsize < 1024)
210 bsize = NFS_DEF_FILE_IO_BUFFER_SIZE;
211 else if (bsize >= NFS_MAX_FILE_IO_BUFFER_SIZE)
212 bsize = NFS_MAX_FILE_IO_BUFFER_SIZE;
213
214 return nfs_block_bits(bsize, nrbitsp);
215}
216
217/*
218 * Obtain the root inode of the file system.
219 */
220static struct inode *
221nfs_get_root(struct super_block *sb, struct nfs_fh *rootfh, struct nfs_fsinfo *fsinfo)
222{
223 struct nfs_server *server = NFS_SB(sb);
224 struct inode *rooti;
225 int error;
226
227 error = server->rpc_ops->getroot(server, rootfh, fsinfo);
228 if (error < 0) {
229 dprintk("nfs_get_root: getattr error = %d\n", -error);
230 return ERR_PTR(error);
231 }
232
233 rooti = nfs_fhget(sb, rootfh, fsinfo->fattr);
234 if (!rooti)
235 return ERR_PTR(-ENOMEM);
236 return rooti;
237}
238
239/*
240 * Do NFS version-independent mount processing, and sanity checking
241 */
242static int
243nfs_sb_init(struct super_block *sb, rpc_authflavor_t authflavor)
244{
245 struct nfs_server *server;
246 struct inode *root_inode;
247 struct nfs_fattr fattr;
248 struct nfs_fsinfo fsinfo = {
249 .fattr = &fattr,
250 };
251 struct nfs_pathconf pathinfo = {
252 .fattr = &fattr,
253 };
254 int no_root_error = 0;
255 unsigned long max_rpc_payload;
256
257 /* We probably want something more informative here */
258 snprintf(sb->s_id, sizeof(sb->s_id), "%x:%x", MAJOR(sb->s_dev), MINOR(sb->s_dev));
259
260 server = NFS_SB(sb);
261
262 sb->s_magic = NFS_SUPER_MAGIC;
263
264 root_inode = nfs_get_root(sb, &server->fh, &fsinfo);
265 /* Did getting the root inode fail? */
266 if (IS_ERR(root_inode)) {
267 no_root_error = PTR_ERR(root_inode);
268 goto out_no_root;
269 }
270 sb->s_root = d_alloc_root(root_inode);
271 if (!sb->s_root) {
272 no_root_error = -ENOMEM;
273 goto out_no_root;
274 }
275 sb->s_root->d_op = server->rpc_ops->dentry_ops;
276
277 /* Get some general file system info */
278 if (server->namelen == 0 &&
279 server->rpc_ops->pathconf(server, &server->fh, &pathinfo) >= 0)
280 server->namelen = pathinfo.max_namelen;
281 /* Work out a lot of parameters */
282 if (server->rsize == 0)
283 server->rsize = nfs_block_size(fsinfo.rtpref, NULL);
284 if (server->wsize == 0)
285 server->wsize = nfs_block_size(fsinfo.wtpref, NULL);
286
287 if (fsinfo.rtmax >= 512 && server->rsize > fsinfo.rtmax)
288 server->rsize = nfs_block_size(fsinfo.rtmax, NULL);
289 if (fsinfo.wtmax >= 512 && server->wsize > fsinfo.wtmax)
290 server->wsize = nfs_block_size(fsinfo.wtmax, NULL);
291
292 max_rpc_payload = nfs_block_size(rpc_max_payload(server->client), NULL);
293 if (server->rsize > max_rpc_payload)
294 server->rsize = max_rpc_payload;
295 if (server->wsize > max_rpc_payload)
296 server->wsize = max_rpc_payload;
297
298 server->rpages = (server->rsize + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT;
299 if (server->rpages > NFS_READ_MAXIOV) {
300 server->rpages = NFS_READ_MAXIOV;
301 server->rsize = server->rpages << PAGE_CACHE_SHIFT;
302 }
303
304 server->wpages = (server->wsize + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT;
305 if (server->wpages > NFS_WRITE_MAXIOV) {
306 server->wpages = NFS_WRITE_MAXIOV;
307 server->wsize = server->wpages << PAGE_CACHE_SHIFT;
308 }
309
310 if (sb->s_blocksize == 0)
311 sb->s_blocksize = nfs_block_bits(server->wsize,
312 &sb->s_blocksize_bits);
313 server->wtmult = nfs_block_bits(fsinfo.wtmult, NULL);
314
315 server->dtsize = nfs_block_size(fsinfo.dtpref, NULL);
316 if (server->dtsize > PAGE_CACHE_SIZE)
317 server->dtsize = PAGE_CACHE_SIZE;
318 if (server->dtsize > server->rsize)
319 server->dtsize = server->rsize;
320
321 if (server->flags & NFS_MOUNT_NOAC) {
322 server->acregmin = server->acregmax = 0;
323 server->acdirmin = server->acdirmax = 0;
324 sb->s_flags |= MS_SYNCHRONOUS;
325 }
326 server->backing_dev_info.ra_pages = server->rpages * NFS_MAX_READAHEAD;
327
328 sb->s_maxbytes = fsinfo.maxfilesize;
329 if (sb->s_maxbytes > MAX_LFS_FILESIZE)
330 sb->s_maxbytes = MAX_LFS_FILESIZE;
331
332 server->client->cl_intr = (server->flags & NFS_MOUNT_INTR) ? 1 : 0;
333 server->client->cl_softrtry = (server->flags & NFS_MOUNT_SOFT) ? 1 : 0;
334
335 /* We're airborne Set socket buffersize */
336 rpc_setbufsize(server->client, server->wsize + 100, server->rsize + 100);
337 return 0;
338 /* Yargs. It didn't work out. */
339out_no_root:
340 dprintk("nfs_sb_init: get root inode failed: errno %d\n", -no_root_error);
341 if (!IS_ERR(root_inode))
342 iput(root_inode);
343 return no_root_error;
344}
345
346/*
347 * Create an RPC client handle.
348 */
349static struct rpc_clnt *
350nfs_create_client(struct nfs_server *server, const struct nfs_mount_data *data)
351{
352 struct rpc_timeout timeparms;
353 struct rpc_xprt *xprt = NULL;
354 struct rpc_clnt *clnt = NULL;
355 int tcp = (data->flags & NFS_MOUNT_TCP);
356
357 /* Initialize timeout values */
358 timeparms.to_initval = data->timeo * HZ / 10;
359 timeparms.to_retries = data->retrans;
360 timeparms.to_maxval = tcp ? RPC_MAX_TCP_TIMEOUT : RPC_MAX_UDP_TIMEOUT;
361 timeparms.to_exponential = 1;
362
363 if (!timeparms.to_initval)
364 timeparms.to_initval = (tcp ? 600 : 11) * HZ / 10;
365 if (!timeparms.to_retries)
366 timeparms.to_retries = 5;
367
368 /* create transport and client */
369 xprt = xprt_create_proto(tcp ? IPPROTO_TCP : IPPROTO_UDP,
370 &server->addr, &timeparms);
371 if (IS_ERR(xprt)) {
9085bbcb
TM		 372 dprintk("%s: cannot create RPC transport. Error = %ld\n",
373 __FUNCTION__, PTR_ERR(xprt));
1da177e4
LT		 374 return (struct rpc_clnt *)xprt;
375 }
376 clnt = rpc_create_client(xprt, server->hostname, &nfs_program,
377 server->rpc_ops->version, data->pseudoflavor);
378 if (IS_ERR(clnt)) {
9085bbcb
TM		 379 dprintk("%s: cannot create RPC client. Error = %ld\n",
380 __FUNCTION__, PTR_ERR(xprt));
1da177e4
LT		 381 goto out_fail;
382 }
383
384 clnt->cl_intr = 1;
385 clnt->cl_softrtry = 1;
386 clnt->cl_chatty = 1;
387
388 return clnt;
389
390out_fail:
1da177e4
LT		 391 return clnt;
392}
393
394/*
395 * The way this works is that the mount process passes a structure
396 * in the data argument which contains the server's IP address
397 * and the root file handle obtained from the server's mount
398 * daemon. We stash these away in the private superblock fields.
399 */
400static int
401nfs_fill_super(struct super_block *sb, struct nfs_mount_data *data, int silent)
402{
403 struct nfs_server *server;
404 rpc_authflavor_t authflavor;
405
406 server = NFS_SB(sb);
407 sb->s_blocksize_bits = 0;
408 sb->s_blocksize = 0;
409 if (data->bsize)
410 sb->s_blocksize = nfs_block_size(data->bsize, &sb->s_blocksize_bits);
411 if (data->rsize)
412 server->rsize = nfs_block_size(data->rsize, NULL);
413 if (data->wsize)
414 server->wsize = nfs_block_size(data->wsize, NULL);
415 server->flags = data->flags & NFS_MOUNT_FLAGMASK;
416
417 server->acregmin = data->acregmin*HZ;
418 server->acregmax = data->acregmax*HZ;
419 server->acdirmin = data->acdirmin*HZ;
420 server->acdirmax = data->acdirmax*HZ;
421
422 /* Start lockd here, before we might error out */
423 if (!(server->flags & NFS_MOUNT_NONLM))
424 lockd_up();
425
426 server->namelen = data->namlen;
427 server->hostname = kmalloc(strlen(data->hostname) + 1, GFP_KERNEL);
428 if (!server->hostname)
429 return -ENOMEM;
430 strcpy(server->hostname, data->hostname);
431
432 /* Check NFS protocol revision and initialize RPC op vector
433 * and file handle pool. */
1da177e4 434#ifdef CONFIG_NFS_V3
9085bbcb 435 if (server->flags & NFS_MOUNT_VER3) {
1da177e4
LT		 436 server->rpc_ops = &nfs_v3_clientops;
437 server->caps |= NFS_CAP_READDIRPLUS;
1da177e4
LT		 438 } else {
439 server->rpc_ops = &nfs_v2_clientops;
440 }
9085bbcb
TM		 441#else
442 server->rpc_ops = &nfs_v2_clientops;
443#endif
1da177e4
LT		 444
445 /* Fill in pseudoflavor for mount version < 5 */
446 if (!(data->flags & NFS_MOUNT_SECFLAVOUR))
447 data->pseudoflavor = RPC_AUTH_UNIX;
448 authflavor = data->pseudoflavor; /* save for sb_init() */
449 /* XXX maybe we want to add a server->pseudoflavor field */
450
451 /* Create RPC client handles */
452 server->client = nfs_create_client(server, data);
453 if (IS_ERR(server->client))
454 return PTR_ERR(server->client);
455 /* RFC 2623, sec 2.3.2 */
456 if (authflavor != RPC_AUTH_UNIX) {
457 server->client_sys = rpc_clone_client(server->client);
458 if (IS_ERR(server->client_sys))
459 return PTR_ERR(server->client_sys);
460 if (!rpcauth_create(RPC_AUTH_UNIX, server->client_sys))
461 return -ENOMEM;
462 } else {
463 atomic_inc(&server->client->cl_count);
464 server->client_sys = server->client;
465 }
466
467 if (server->flags & NFS_MOUNT_VER3) {
468 if (server->namelen == 0 || server->namelen > NFS3_MAXNAMLEN)
469 server->namelen = NFS3_MAXNAMLEN;
470 sb->s_time_gran = 1;
471 } else {
472 if (server->namelen == 0 || server->namelen > NFS2_MAXNAMLEN)
473 server->namelen = NFS2_MAXNAMLEN;
474 }
475
476 sb->s_op = &nfs_sops;
477 return nfs_sb_init(sb, authflavor);
478}
479
480static int
481nfs_statfs(struct super_block *sb, struct kstatfs *buf)
482{
483 struct nfs_server *server = NFS_SB(sb);
484 unsigned char blockbits;
485 unsigned long blockres;
486 struct nfs_fh *rootfh = NFS_FH(sb->s_root->d_inode);
487 struct nfs_fattr fattr;
488 struct nfs_fsstat res = {
489 .fattr = &fattr,
490 };
491 int error;
492
493 lock_kernel();
494
495 error = server->rpc_ops->statfs(server, rootfh, &res);
496 buf->f_type = NFS_SUPER_MAGIC;
497 if (error < 0)
498 goto out_err;
499
500 /*
501 * Current versions of glibc do not correctly handle the
502 * case where f_frsize != f_bsize. Eventually we want to
503 * report the value of wtmult in this field.
504 */
505 buf->f_frsize = sb->s_blocksize;
506
507 /*
508 * On most *nix systems, f_blocks, f_bfree, and f_bavail
509 * are reported in units of f_frsize. Linux hasn't had
510 * an f_frsize field in its statfs struct until recently,
511 * thus historically Linux's sys_statfs reports these
512 * fields in units of f_bsize.
513 */
514 buf->f_bsize = sb->s_blocksize;
515 blockbits = sb->s_blocksize_bits;
516 blockres = (1 << blockbits) - 1;
517 buf->f_blocks = (res.tbytes + blockres) >> blockbits;
518 buf->f_bfree = (res.fbytes + blockres) >> blockbits;
519 buf->f_bavail = (res.abytes + blockres) >> blockbits;
520
521 buf->f_files = res.tfiles;
522 buf->f_ffree = res.afiles;
523
524 buf->f_namelen = server->namelen;
525 out:
526 unlock_kernel();
527
528 return 0;
529
530 out_err:
531 printk(KERN_WARNING "nfs_statfs: statfs error = %d\n", -error);
532 buf->f_bsize = buf->f_blocks = buf->f_bfree = buf->f_bavail = -1;
533 goto out;
534
535}
536
537static int nfs_show_options(struct seq_file *m, struct vfsmount *mnt)
538{
539 static struct proc_nfs_info {
540 int flag;
541 char *str;
542 char *nostr;
543 } nfs_info[] = {
544 { NFS_MOUNT_SOFT, ",soft", ",hard" },
545 { NFS_MOUNT_INTR, ",intr", "" },
546 { NFS_MOUNT_POSIX, ",posix", "" },
547 { NFS_MOUNT_TCP, ",tcp", ",udp" },
548 { NFS_MOUNT_NOCTO, ",nocto", "" },
549 { NFS_MOUNT_NOAC, ",noac", "" },
550 { NFS_MOUNT_NONLM, ",nolock", ",lock" },
551 { 0, NULL, NULL }
552 };
553 struct proc_nfs_info *nfs_infop;
554 struct nfs_server *nfss = NFS_SB(mnt->mnt_sb);
555
556 seq_printf(m, ",v%d", nfss->rpc_ops->version);
557 seq_printf(m, ",rsize=%d", nfss->rsize);
558 seq_printf(m, ",wsize=%d", nfss->wsize);
559 if (nfss->acregmin != 3*HZ)
560 seq_printf(m, ",acregmin=%d", nfss->acregmin/HZ);
561 if (nfss->acregmax != 60*HZ)
562 seq_printf(m, ",acregmax=%d", nfss->acregmax/HZ);
563 if (nfss->acdirmin != 30*HZ)
564 seq_printf(m, ",acdirmin=%d", nfss->acdirmin/HZ);
565 if (nfss->acdirmax != 60*HZ)
566 seq_printf(m, ",acdirmax=%d", nfss->acdirmax/HZ);
567 for (nfs_infop = nfs_info; nfs_infop->flag; nfs_infop++) {
568 if (nfss->flags & nfs_infop->flag)
569 seq_puts(m, nfs_infop->str);
570 else
571 seq_puts(m, nfs_infop->nostr);
572 }
573 seq_puts(m, ",addr=");
574 seq_escape(m, nfss->hostname, " \t\n\\");
575 return 0;
576}
577
578/*
579 * Invalidate the local caches
580 */
581void
582nfs_zap_caches(struct inode *inode)
583{
584 struct nfs_inode *nfsi = NFS_I(inode);
585 int mode = inode->i_mode;
586
587 NFS_ATTRTIMEO(inode) = NFS_MINATTRTIMEO(inode);
588 NFS_ATTRTIMEO_UPDATE(inode) = jiffies;
589
590 memset(NFS_COOKIEVERF(inode), 0, sizeof(NFS_COOKIEVERF(inode)));
591 if (S_ISREG(mode) || S_ISDIR(mode) || S_ISLNK(mode))
ada70d94 592 nfsi->flags |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_DATA|NFS_INO_INVALID_ACCESS|NFS_INO_INVALID_ACL;
1da177e4 593 else
ada70d94
TM		 594 nfsi->flags |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_ACCESS|NFS_INO_INVALID_ACL;
595}
596
597static void nfs_zap_acl_cache(struct inode *inode)
598{
599 void (*clear_acl_cache)(struct inode *);
600
601 clear_acl_cache = NFS_PROTO(inode)->clear_acl_cache;
602 if (clear_acl_cache != NULL)
603 clear_acl_cache(inode);
604 NFS_I(inode)->flags &= ~NFS_INO_INVALID_ACL;
1da177e4
LT		 605}
606
607/*
608 * Invalidate, but do not unhash, the inode
609 */
610static void
611nfs_invalidate_inode(struct inode *inode)
612{
613 umode_t save_mode = inode->i_mode;
614
615 make_bad_inode(inode);
616 inode->i_mode = save_mode;
617 nfs_zap_caches(inode);
618}
619
620struct nfs_find_desc {
621 struct nfs_fh *fh;
622 struct nfs_fattr *fattr;
623};
624
625/*
626 * In NFSv3 we can have 64bit inode numbers. In order to support
627 * this, and re-exported directories (also seen in NFSv2)
628 * we are forced to allow 2 different inodes to have the same
629 * i_ino.
630 */
631static int
632nfs_find_actor(struct inode *inode, void *opaque)
633{
634 struct nfs_find_desc *desc = (struct nfs_find_desc *)opaque;
635 struct nfs_fh *fh = desc->fh;
636 struct nfs_fattr *fattr = desc->fattr;
637
638 if (NFS_FILEID(inode) != fattr->fileid)
639 return 0;
640 if (nfs_compare_fh(NFS_FH(inode), fh))
641 return 0;
642 if (is_bad_inode(inode) || NFS_STALE(inode))
643 return 0;
644 return 1;
645}
646
647static int
648nfs_init_locked(struct inode *inode, void *opaque)
649{
650 struct nfs_find_desc *desc = (struct nfs_find_desc *)opaque;
651 struct nfs_fattr *fattr = desc->fattr;
652
653 NFS_FILEID(inode) = fattr->fileid;
654 nfs_copy_fh(NFS_FH(inode), desc->fh);
655 return 0;
656}
657
658/* Don't use READDIRPLUS on directories that we believe are too large */
659#define NFS_LIMIT_READDIRPLUS (8*PAGE_SIZE)
660
661/*
662 * This is our front-end to iget that looks up inodes by file handle
663 * instead of inode number.
664 */
665struct inode *
666nfs_fhget(struct super_block *sb, struct nfs_fh *fh, struct nfs_fattr *fattr)
667{
668 struct nfs_find_desc desc = {
669 .fh = fh,
670 .fattr = fattr
671 };
672 struct inode *inode = NULL;
673 unsigned long hash;
674
675 if ((fattr->valid & NFS_ATTR_FATTR) == 0)
676 goto out_no_inode;
677
678 if (!fattr->nlink) {
679 printk("NFS: Buggy server - nlink == 0!\n");
680 goto out_no_inode;
681 }
682
683 hash = nfs_fattr_to_ino_t(fattr);
684
685 if (!(inode = iget5_locked(sb, hash, nfs_find_actor, nfs_init_locked, &desc)))
686 goto out_no_inode;
687
688 if (inode->i_state & I_NEW) {
689 struct nfs_inode *nfsi = NFS_I(inode);
690
691 /* We set i_ino for the few things that still rely on it,
692 * such as stat(2) */
693 inode->i_ino = hash;
694
695 /* We can't support update_atime(), since the server will reset it */
696 inode->i_flags |= S_NOATIME|S_NOCMTIME;
697 inode->i_mode = fattr->mode;
698 /* Why so? Because we want revalidate for devices/FIFOs, and
699 * that's precisely what we have in nfs_file_inode_operations.
700 */
92cfc62c 701 inode->i_op = NFS_SB(sb)->rpc_ops->file_inode_ops;
1da177e4
LT		 702 if (S_ISREG(inode->i_mode)) {
703 inode->i_fop = &nfs_file_operations;
704 inode->i_data.a_ops = &nfs_file_aops;
705 inode->i_data.backing_dev_info = &NFS_SB(sb)->backing_dev_info;
706 } else if (S_ISDIR(inode->i_mode)) {
707 inode->i_op = NFS_SB(sb)->rpc_ops->dir_inode_ops;
708 inode->i_fop = &nfs_dir_operations;
709 if (nfs_server_capable(inode, NFS_CAP_READDIRPLUS)
710 && fattr->size <= NFS_LIMIT_READDIRPLUS)
711 NFS_FLAGS(inode) |= NFS_INO_ADVISE_RDPLUS;
712 } else if (S_ISLNK(inode->i_mode))
713 inode->i_op = &nfs_symlink_inode_operations;
714 else
715 init_special_inode(inode, inode->i_mode, fattr->rdev);
716
717 nfsi->read_cache_jiffies = fattr->timestamp;
718 inode->i_atime = fattr->atime;
719 inode->i_mtime = fattr->mtime;
720 inode->i_ctime = fattr->ctime;
721 if (fattr->valid & NFS_ATTR_FATTR_V4)
722 nfsi->change_attr = fattr->change_attr;
723 inode->i_size = nfs_size_to_loff_t(fattr->size);
724 inode->i_nlink = fattr->nlink;
725 inode->i_uid = fattr->uid;
726 inode->i_gid = fattr->gid;
727 if (fattr->valid & (NFS_ATTR_FATTR_V3 | NFS_ATTR_FATTR_V4)) {
728 /*
729 * report the blocks in 512byte units
730 */
731 inode->i_blocks = nfs_calc_block_size(fattr->du.nfs3.used);
732 inode->i_blksize = inode->i_sb->s_blocksize;
733 } else {
734 inode->i_blocks = fattr->du.nfs2.blocks;
735 inode->i_blksize = fattr->du.nfs2.blocksize;
736 }
737 nfsi->attrtimeo = NFS_MINATTRTIMEO(inode);
738 nfsi->attrtimeo_timestamp = jiffies;
739 memset(nfsi->cookieverf, 0, sizeof(nfsi->cookieverf));
740 nfsi->cache_access.cred = NULL;
741
742 unlock_new_inode(inode);
743 } else
744 nfs_refresh_inode(inode, fattr);
745 dprintk("NFS: nfs_fhget(%s/%Ld ct=%d)\n",
746 inode->i_sb->s_id,
747 (long long)NFS_FILEID(inode),
748 atomic_read(&inode->i_count));
749
750out:
751 return inode;
752
753out_no_inode:
754 printk("nfs_fhget: iget failed\n");
755 goto out;
756}
757
758#define NFS_VALID_ATTRS (ATTR_MODE|ATTR_UID|ATTR_GID|ATTR_SIZE|ATTR_ATIME|ATTR_ATIME_SET|ATTR_MTIME|ATTR_MTIME_SET)
759
760int
761nfs_setattr(struct dentry *dentry, struct iattr *attr)
762{
763 struct inode *inode = dentry->d_inode;
764 struct nfs_fattr fattr;
765 int error;
766
767 if (attr->ia_valid & ATTR_SIZE) {
768 if (!S_ISREG(inode->i_mode) || attr->ia_size == i_size_read(inode))
769 attr->ia_valid &= ~ATTR_SIZE;
770 }
771
772 /* Optimization: if the end result is no change, don't RPC */
773 attr->ia_valid &= NFS_VALID_ATTRS;
774 if (attr->ia_valid == 0)
775 return 0;
776
777 lock_kernel();
778 nfs_begin_data_update(inode);
779 /* Write all dirty data if we're changing file permissions or size */
780 if ((attr->ia_valid & (ATTR_MODE|ATTR_UID|ATTR_GID|ATTR_SIZE)) != 0) {
781 if (filemap_fdatawrite(inode->i_mapping) == 0)
782 filemap_fdatawait(inode->i_mapping);
783 nfs_wb_all(inode);
784 }
785 error = NFS_PROTO(inode)->setattr(dentry, &fattr, attr);
786 if (error == 0) {
787 nfs_refresh_inode(inode, &fattr);
788 if ((attr->ia_valid & ATTR_MODE) != 0) {
789 int mode;
790 mode = inode->i_mode & ~S_IALLUGO;
791 mode |= attr->ia_mode & S_IALLUGO;
792 inode->i_mode = mode;
793 }
794 if ((attr->ia_valid & ATTR_UID) != 0)
795 inode->i_uid = attr->ia_uid;
796 if ((attr->ia_valid & ATTR_GID) != 0)
797 inode->i_gid = attr->ia_gid;
798 if ((attr->ia_valid & ATTR_SIZE) != 0) {
799 inode->i_size = attr->ia_size;
800 vmtruncate(inode, attr->ia_size);
801 }
802 }
803 if ((attr->ia_valid & (ATTR_MODE|ATTR_UID|ATTR_GID)) != 0)
ada70d94 804 NFS_FLAGS(inode) |= NFS_INO_INVALID_ACCESS|NFS_INO_INVALID_ACL;
1da177e4
LT		 805 nfs_end_data_update(inode);
806 unlock_kernel();
807 return error;
808}
809
810/*
811 * Wait for the inode to get unlocked.
812 * (Used for NFS_INO_LOCKED and NFS_INO_REVALIDATING).
813 */
814static int
815nfs_wait_on_inode(struct inode *inode, int flag)
816{
817 struct rpc_clnt *clnt = NFS_CLIENT(inode);
818 struct nfs_inode *nfsi = NFS_I(inode);
819
820 int error;
821 if (!(NFS_FLAGS(inode) & flag))
822 return 0;
823 atomic_inc(&inode->i_count);
824 error = nfs_wait_event(clnt, nfsi->nfs_i_wait,
825 !(NFS_FLAGS(inode) & flag));
826 iput(inode);
827 return error;
828}
829
830int nfs_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
831{
832 struct inode *inode = dentry->d_inode;
833 struct nfs_inode *nfsi = NFS_I(inode);
834 int need_atime = nfsi->flags & NFS_INO_INVALID_ATIME;
835 int err;
836
837 if (__IS_FLG(inode, MS_NOATIME))
838 need_atime = 0;
839 else if (__IS_FLG(inode, MS_NODIRATIME) && S_ISDIR(inode->i_mode))
840 need_atime = 0;
841 /* We may force a getattr if the user cares about atime */
842 if (need_atime)
843 err = __nfs_revalidate_inode(NFS_SERVER(inode), inode);
844 else
845 err = nfs_revalidate_inode(NFS_SERVER(inode), inode);
846 if (!err)
847 generic_fillattr(inode, stat);
848 return err;
849}
850
851struct nfs_open_context *alloc_nfs_open_context(struct dentry *dentry, struct rpc_cred *cred)
852{
853 struct nfs_open_context *ctx;
854
855 ctx = (struct nfs_open_context *)kmalloc(sizeof(*ctx), GFP_KERNEL);
856 if (ctx != NULL) {
857 atomic_set(&ctx->count, 1);
858 ctx->dentry = dget(dentry);
859 ctx->cred = get_rpccred(cred);
860 ctx->state = NULL;
861 ctx->lockowner = current->files;
862 ctx->error = 0;
1da177e4
LT		 863 }
864 return ctx;
865}
866
867struct nfs_open_context *get_nfs_open_context(struct nfs_open_context *ctx)
868{
869 if (ctx != NULL)
870 atomic_inc(&ctx->count);
871 return ctx;
872}
873
874void put_nfs_open_context(struct nfs_open_context *ctx)
875{
876 if (atomic_dec_and_test(&ctx->count)) {
877 if (!list_empty(&ctx->list)) {
878 struct inode *inode = ctx->dentry->d_inode;
879 spin_lock(&inode->i_lock);
880 list_del(&ctx->list);
881 spin_unlock(&inode->i_lock);
882 }
883 if (ctx->state != NULL)
884 nfs4_close_state(ctx->state, ctx->mode);
885 if (ctx->cred != NULL)
886 put_rpccred(ctx->cred);
887 dput(ctx->dentry);
888 kfree(ctx);
889 }
890}
891
892/*
893 * Ensure that mmap has a recent RPC credential for use when writing out
894 * shared pages
895 */
896void nfs_file_set_open_context(struct file *filp, struct nfs_open_context *ctx)
897{
898 struct inode *inode = filp->f_dentry->d_inode;
899 struct nfs_inode *nfsi = NFS_I(inode);
900
901 filp->private_data = get_nfs_open_context(ctx);
902 spin_lock(&inode->i_lock);
903 list_add(&ctx->list, &nfsi->open_files);
904 spin_unlock(&inode->i_lock);
905}
906
907struct nfs_open_context *nfs_find_open_context(struct inode *inode, int mode)
908{
909 struct nfs_inode *nfsi = NFS_I(inode);
910 struct nfs_open_context *pos, *ctx = NULL;
911
912 spin_lock(&inode->i_lock);
913 list_for_each_entry(pos, &nfsi->open_files, list) {
914 if ((pos->mode & mode) == mode) {
915 ctx = get_nfs_open_context(pos);
916 break;
917 }
918 }
919 spin_unlock(&inode->i_lock);
920 return ctx;
921}
922
923void nfs_file_clear_open_context(struct file *filp)
924{
925 struct inode *inode = filp->f_dentry->d_inode;
926 struct nfs_open_context *ctx = (struct nfs_open_context *)filp->private_data;
927
928 if (ctx) {
929 filp->private_data = NULL;
930 spin_lock(&inode->i_lock);
931 list_move_tail(&ctx->list, &NFS_I(inode)->open_files);
932 spin_unlock(&inode->i_lock);
933 put_nfs_open_context(ctx);
934 }
935}
936
937/*
938 * These allocate and release file read/write context information.
939 */
940int nfs_open(struct inode *inode, struct file *filp)
941{
942 struct nfs_open_context *ctx;
943 struct rpc_cred *cred;
944
945 cred = rpcauth_lookupcred(NFS_CLIENT(inode)->cl_auth, 0);
946 if (IS_ERR(cred))
947 return PTR_ERR(cred);
948 ctx = alloc_nfs_open_context(filp->f_dentry, cred);
949 put_rpccred(cred);
950 if (ctx == NULL)
951 return -ENOMEM;
952 ctx->mode = filp->f_mode;
953 nfs_file_set_open_context(filp, ctx);
954 put_nfs_open_context(ctx);
955 if ((filp->f_mode & FMODE_WRITE) != 0)
956 nfs_begin_data_update(inode);
957 return 0;
958}
959
960int nfs_release(struct inode *inode, struct file *filp)
961{
962 if ((filp->f_mode & FMODE_WRITE) != 0)
963 nfs_end_data_update(inode);
964 nfs_file_clear_open_context(filp);
965 return 0;
966}
967
968/*
969 * This function is called whenever some part of NFS notices that
970 * the cached attributes have to be refreshed.
971 */
972int
973__nfs_revalidate_inode(struct nfs_server *server, struct inode *inode)
974{
975 int status = -ESTALE;
976 struct nfs_fattr fattr;
977 struct nfs_inode *nfsi = NFS_I(inode);
978 unsigned long verifier;
979 unsigned int flags;
980
981 dfprintk(PAGECACHE, "NFS: revalidating (%s/%Ld)\n",
982 inode->i_sb->s_id, (long long)NFS_FILEID(inode));
983
984 lock_kernel();
985 if (!inode || is_bad_inode(inode))
986 goto out_nowait;
987 if (NFS_STALE(inode))
988 goto out_nowait;
989
990 while (NFS_REVALIDATING(inode)) {
991 status = nfs_wait_on_inode(inode, NFS_INO_REVALIDATING);
992 if (status < 0)
993 goto out_nowait;
994 if (NFS_ATTRTIMEO(inode) == 0)
995 continue;
996 if (NFS_FLAGS(inode) & (NFS_INO_INVALID_ATTR|NFS_INO_INVALID_DATA|NFS_INO_INVALID_ATIME))
997 continue;
998 status = NFS_STALE(inode) ? -ESTALE : 0;
999 goto out_nowait;
1000 }
1001 NFS_FLAGS(inode) |= NFS_INO_REVALIDATING;
1002
1003 /* Protect against RPC races by saving the change attribute */
1004 verifier = nfs_save_change_attribute(inode);
1005 status = NFS_PROTO(inode)->getattr(server, NFS_FH(inode), &fattr);
1006 if (status != 0) {
1007 dfprintk(PAGECACHE, "nfs_revalidate_inode: (%s/%Ld) getattr failed, error=%d\n",
1008 inode->i_sb->s_id,
1009 (long long)NFS_FILEID(inode), status);
1010 if (status == -ESTALE) {
1011 nfs_zap_caches(inode);
1012 if (!S_ISDIR(inode->i_mode))
1013 NFS_FLAGS(inode) |= NFS_INO_STALE;
1014 }
1015 goto out;
1016 }
1017
1018 status = nfs_update_inode(inode, &fattr, verifier);
1019 if (status) {
1020 dfprintk(PAGECACHE, "nfs_revalidate_inode: (%s/%Ld) refresh failed, error=%d\n",
1021 inode->i_sb->s_id,
1022 (long long)NFS_FILEID(inode), status);
1023 goto out;
1024 }
1025 flags = nfsi->flags;
1026 /*
1027 * We may need to keep the attributes marked as invalid if
1028 * we raced with nfs_end_attr_update().
1029 */
1030 if (verifier == nfsi->cache_change_attribute)
1031 nfsi->flags &= ~(NFS_INO_INVALID_ATTR|NFS_INO_INVALID_ATIME);
1032 /* Do the page cache invalidation */
1033 if (flags & NFS_INO_INVALID_DATA) {
1034 if (S_ISREG(inode->i_mode)) {
1035 if (filemap_fdatawrite(inode->i_mapping) == 0)
1036 filemap_fdatawait(inode->i_mapping);
1037 nfs_wb_all(inode);
1038 }
1039 nfsi->flags &= ~NFS_INO_INVALID_DATA;
1040 invalidate_inode_pages2(inode->i_mapping);
1041 memset(NFS_COOKIEVERF(inode), 0, sizeof(NFS_COOKIEVERF(inode)));
1042 dfprintk(PAGECACHE, "NFS: (%s/%Ld) data cache invalidated\n",
1043 inode->i_sb->s_id,
1044 (long long)NFS_FILEID(inode));
1045 /* This ensures we revalidate dentries */
1046 nfsi->cache_change_attribute++;
1047 }
ada70d94
TM		 1048 if (flags & NFS_INO_INVALID_ACL)
1049 nfs_zap_acl_cache(inode);
1da177e4
LT		 1050 dfprintk(PAGECACHE, "NFS: (%s/%Ld) revalidation complete\n",
1051 inode->i_sb->s_id,
1052 (long long)NFS_FILEID(inode));
1053
1054out:
1055 NFS_FLAGS(inode) &= ~NFS_INO_REVALIDATING;
1056 wake_up(&nfsi->nfs_i_wait);
1057 out_nowait:
1058 unlock_kernel();
1059 return status;
1060}
1061
1062int nfs_attribute_timeout(struct inode *inode)
1063{
1064 struct nfs_inode *nfsi = NFS_I(inode);
1065
1066 if (nfs_have_delegation(inode, FMODE_READ))
1067 return 0;
1068 return time_after(jiffies, nfsi->read_cache_jiffies+nfsi->attrtimeo);
1069}
1070
1071/**
1072 * nfs_revalidate_inode - Revalidate the inode attributes
1073 * @server - pointer to nfs_server struct
1074 * @inode - pointer to inode struct
1075 *
1076 * Updates inode attribute information by retrieving the data from the server.
1077 */
1078int nfs_revalidate_inode(struct nfs_server *server, struct inode *inode)
1079{
1080 if (!(NFS_FLAGS(inode) & (NFS_INO_INVALID_ATTR|NFS_INO_INVALID_DATA))
1081 && !nfs_attribute_timeout(inode))
1082 return NFS_STALE(inode) ? -ESTALE : 0;
1083 return __nfs_revalidate_inode(server, inode);
1084}
1085
1086/**
1087 * nfs_begin_data_update
1088 * @inode - pointer to inode
1089 * Declare that a set of operations will update file data on the server
1090 */
1091void nfs_begin_data_update(struct inode *inode)
1092{
1093 atomic_inc(&NFS_I(inode)->data_updates);
1094}
1095
1096/**
1097 * nfs_end_data_update
1098 * @inode - pointer to inode
1099 * Declare end of the operations that will update file data
1100 * This will mark the inode as immediately needing revalidation
1101 * of its attribute cache.
1102 */
1103void nfs_end_data_update(struct inode *inode)
1104{
1105 struct nfs_inode *nfsi = NFS_I(inode);
1106
1107 if (!nfs_have_delegation(inode, FMODE_READ)) {
1108 /* Mark the attribute cache for revalidation */
1109 nfsi->flags |= NFS_INO_INVALID_ATTR;
1110 /* Directories and symlinks: invalidate page cache too */
1111 if (S_ISDIR(inode->i_mode) || S_ISLNK(inode->i_mode))
1112 nfsi->flags |= NFS_INO_INVALID_DATA;
1113 }
1114 nfsi->cache_change_attribute ++;
1115 atomic_dec(&nfsi->data_updates);
1116}
1117
1118/**
1119 * nfs_end_data_update_defer
1120 * @inode - pointer to inode
1121 * Declare end of the operations that will update file data
1122 * This will defer marking the inode as needing revalidation
1123 * unless there are no other pending updates.
1124 */
1125void nfs_end_data_update_defer(struct inode *inode)
1126{
1127 struct nfs_inode *nfsi = NFS_I(inode);
1128
1129 if (atomic_dec_and_test(&nfsi->data_updates)) {
1130 /* Mark the attribute cache for revalidation */
1131 nfsi->flags |= NFS_INO_INVALID_ATTR;
1132 /* Directories and symlinks: invalidate page cache too */
1133 if (S_ISDIR(inode->i_mode) || S_ISLNK(inode->i_mode))
1134 nfsi->flags |= NFS_INO_INVALID_DATA;
1135 nfsi->cache_change_attribute ++;
1136 }
1137}
1138
1139/**
1140 * nfs_refresh_inode - verify consistency of the inode attribute cache
1141 * @inode - pointer to inode
1142 * @fattr - updated attributes
1143 *
1144 * Verifies the attribute cache. If we have just changed the attributes,
1145 * so that fattr carries weak cache consistency data, then it may
1146 * also update the ctime/mtime/change_attribute.
1147 */
1148int nfs_refresh_inode(struct inode *inode, struct nfs_fattr *fattr)
1149{
1150 struct nfs_inode *nfsi = NFS_I(inode);
1151 loff_t cur_size, new_isize;
1152 int data_unstable;
1153
1154 /* Do we hold a delegation? */
1155 if (nfs_have_delegation(inode, FMODE_READ))
1156 return 0;
1157
1158 /* Are we in the process of updating data on the server? */
1159 data_unstable = nfs_caches_unstable(inode);
1160
1161 if (fattr->valid & NFS_ATTR_FATTR_V4) {
1162 if ((fattr->valid & NFS_ATTR_PRE_CHANGE) != 0
1163 && nfsi->change_attr == fattr->pre_change_attr)
1164 nfsi->change_attr = fattr->change_attr;
1165 if (!data_unstable && nfsi->change_attr != fattr->change_attr)
1166 nfsi->flags |= NFS_INO_INVALID_ATTR;
1167 }
1168
1169 if ((fattr->valid & NFS_ATTR_FATTR) == 0)
1170 return 0;
1171
1172 /* Has the inode gone and changed behind our back? */
1173 if (nfsi->fileid != fattr->fileid
1174 || (inode->i_mode & S_IFMT) != (fattr->mode & S_IFMT))
1175 return -EIO;
1176
1177 cur_size = i_size_read(inode);
1178 new_isize = nfs_size_to_loff_t(fattr->size);
1179
1180 /* If we have atomic WCC data, we may update some attributes */
1181 if ((fattr->valid & NFS_ATTR_WCC) != 0) {
1182 if (timespec_equal(&inode->i_ctime, &fattr->pre_ctime))
1183 memcpy(&inode->i_ctime, &fattr->ctime, sizeof(inode->i_ctime));
1184 if (timespec_equal(&inode->i_mtime, &fattr->pre_mtime))
1185 memcpy(&inode->i_mtime, &fattr->mtime, sizeof(inode->i_mtime));
1186 }
1187
1188 /* Verify a few of the more important attributes */
1189 if (!data_unstable) {
1190 if (!timespec_equal(&inode->i_mtime, &fattr->mtime)
1191 || cur_size != new_isize)
1192 nfsi->flags |= NFS_INO_INVALID_ATTR;
1193 } else if (S_ISREG(inode->i_mode) && new_isize > cur_size)
1194 nfsi->flags |= NFS_INO_INVALID_ATTR;
1195
1196 /* Have any file permissions changed? */
1197 if ((inode->i_mode & S_IALLUGO) != (fattr->mode & S_IALLUGO)
1198 || inode->i_uid != fattr->uid
1199 || inode->i_gid != fattr->gid)
ada70d94 1200 nfsi->flags |= NFS_INO_INVALID_ATTR | NFS_INO_INVALID_ACCESS | NFS_INO_INVALID_ACL;
1da177e4
LT		 1201
1202 /* Has the link count changed? */
1203 if (inode->i_nlink != fattr->nlink)
1204 nfsi->flags |= NFS_INO_INVALID_ATTR;
1205
1206 if (!timespec_equal(&inode->i_atime, &fattr->atime))
1207 nfsi->flags |= NFS_INO_INVALID_ATIME;
1208
1209 nfsi->read_cache_jiffies = fattr->timestamp;
1210 return 0;
1211}
1212
1213/*
1214 * Many nfs protocol calls return the new file attributes after
1215 * an operation. Here we update the inode to reflect the state
1216 * of the server's inode.
1217 *
1218 * This is a bit tricky because we have to make sure all dirty pages
1219 * have been sent off to the server before calling invalidate_inode_pages.
1220 * To make sure no other process adds more write requests while we try
1221 * our best to flush them, we make them sleep during the attribute refresh.
1222 *
1223 * A very similar scenario holds for the dir cache.
1224 */
1225static int nfs_update_inode(struct inode *inode, struct nfs_fattr *fattr, unsigned long verifier)
1226{
1227 struct nfs_inode *nfsi = NFS_I(inode);
1228 __u64 new_size;
1229 loff_t new_isize;
1230 unsigned int invalid = 0;
1231 loff_t cur_isize;
1232 int data_unstable;
1233
1234 dfprintk(VFS, "NFS: %s(%s/%ld ct=%d info=0x%x)\n",
1235 __FUNCTION__, inode->i_sb->s_id, inode->i_ino,
1236 atomic_read(&inode->i_count), fattr->valid);
1237
1238 if ((fattr->valid & NFS_ATTR_FATTR) == 0)
1239 return 0;
1240
1241 if (nfsi->fileid != fattr->fileid) {
1242 printk(KERN_ERR "%s: inode number mismatch\n"
1243 "expected (%s/0x%Lx), got (%s/0x%Lx)\n",
1244 __FUNCTION__,
1245 inode->i_sb->s_id, (long long)nfsi->fileid,
1246 inode->i_sb->s_id, (long long)fattr->fileid);
1247 goto out_err;
1248 }
1249
1250 /*
1251 * Make sure the inode's type hasn't changed.
1252 */
1253 if ((inode->i_mode & S_IFMT) != (fattr->mode & S_IFMT))
1254 goto out_changed;
1255
1256 /*
1257 * Update the read time so we don't revalidate too often.
1258 */
1259 nfsi->read_cache_jiffies = fattr->timestamp;
1260
1261 /* Are we racing with known updates of the metadata on the server? */
1262 data_unstable = ! nfs_verify_change_attribute(inode, verifier);
1263
1264 /* Check if the file size agrees */
1265 new_size = fattr->size;
1266 new_isize = nfs_size_to_loff_t(fattr->size);
1267 cur_isize = i_size_read(inode);
1268 if (cur_isize != new_size) {
1269#ifdef NFS_DEBUG_VERBOSE
1270 printk(KERN_DEBUG "NFS: isize change on %s/%ld\n", inode->i_sb->s_id, inode->i_ino);
1271#endif
1272 /*
1273 * If we have pending writebacks, things can get
1274 * messy.
1275 */
1276 if (S_ISREG(inode->i_mode) && data_unstable) {
1277 if (new_isize > cur_isize) {
1278 inode->i_size = new_isize;
1279 invalid |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_DATA;
1280 }
1281 } else {
1282 inode->i_size = new_isize;
1283 invalid |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_DATA;
1284 }
1285 }
1286
1287 /*
1288 * Note: we don't check inode->i_mtime since pipes etc.
1289 * can change this value in VFS without requiring a
1290 * cache revalidation.
1291 */
1292 if (!timespec_equal(&inode->i_mtime, &fattr->mtime)) {
1293 memcpy(&inode->i_mtime, &fattr->mtime, sizeof(inode->i_mtime));
1294#ifdef NFS_DEBUG_VERBOSE
1295 printk(KERN_DEBUG "NFS: mtime change on %s/%ld\n", inode->i_sb->s_id, inode->i_ino);
1296#endif
1297 if (!data_unstable)
1298 invalid |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_DATA;
1299 }
1300
1301 if ((fattr->valid & NFS_ATTR_FATTR_V4)
1302 && nfsi->change_attr != fattr->change_attr) {
1303#ifdef NFS_DEBUG_VERBOSE
1304 printk(KERN_DEBUG "NFS: change_attr change on %s/%ld\n",
1305 inode->i_sb->s_id, inode->i_ino);
1306#endif
1307 nfsi->change_attr = fattr->change_attr;
1308 if (!data_unstable)
ada70d94 1309 invalid |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_DATA|NFS_INO_INVALID_ACCESS|NFS_INO_INVALID_ACL;
1da177e4
LT		 1310 }
1311
ada70d94
TM		 1312 /* If ctime has changed we should definitely clear access+acl caches */
1313 if (!timespec_equal(&inode->i_ctime, &fattr->ctime)) {
1314 if (!data_unstable)
1315 invalid |= NFS_INO_INVALID_ACCESS|NFS_INO_INVALID_ACL;
1316 memcpy(&inode->i_ctime, &fattr->ctime, sizeof(inode->i_ctime));
1317 }
1da177e4
LT		 1318 memcpy(&inode->i_atime, &fattr->atime, sizeof(inode->i_atime));
1319
1320 if ((inode->i_mode & S_IALLUGO) != (fattr->mode & S_IALLUGO) ||
1321 inode->i_uid != fattr->uid ||
1322 inode->i_gid != fattr->gid)
ada70d94 1323 invalid |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_ACCESS|NFS_INO_INVALID_ACL;
1da177e4
LT		 1324
1325 inode->i_mode = fattr->mode;
1326 inode->i_nlink = fattr->nlink;
1327 inode->i_uid = fattr->uid;
1328 inode->i_gid = fattr->gid;
1329
1330 if (fattr->valid & (NFS_ATTR_FATTR_V3 | NFS_ATTR_FATTR_V4)) {
1331 /*
1332 * report the blocks in 512byte units
1333 */
1334 inode->i_blocks = nfs_calc_block_size(fattr->du.nfs3.used);
1335 inode->i_blksize = inode->i_sb->s_blocksize;
1336 } else {
1337 inode->i_blocks = fattr->du.nfs2.blocks;
1338 inode->i_blksize = fattr->du.nfs2.blocksize;
1339 }
1340
1341 /* Update attrtimeo value if we're out of the unstable period */
1342 if (invalid & NFS_INO_INVALID_ATTR) {
1343 nfsi->attrtimeo = NFS_MINATTRTIMEO(inode);
1344 nfsi->attrtimeo_timestamp = jiffies;
1345 } else if (time_after(jiffies, nfsi->attrtimeo_timestamp+nfsi->attrtimeo)) {
1346 if ((nfsi->attrtimeo <<= 1) > NFS_MAXATTRTIMEO(inode))
1347 nfsi->attrtimeo = NFS_MAXATTRTIMEO(inode);
1348 nfsi->attrtimeo_timestamp = jiffies;
1349 }
1350 /* Don't invalidate the data if we were to blame */
1351 if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode)
1352 || S_ISLNK(inode->i_mode)))
1353 invalid &= ~NFS_INO_INVALID_DATA;
1354 if (!nfs_have_delegation(inode, FMODE_READ))
1355 nfsi->flags |= invalid;
1356
1357 return 0;
1358 out_changed:
1359 /*
1360 * Big trouble! The inode has become a different object.
1361 */
1362#ifdef NFS_PARANOIA
1363 printk(KERN_DEBUG "%s: inode %ld mode changed, %07o to %07o\n",
1364 __FUNCTION__, inode->i_ino, inode->i_mode, fattr->mode);
1365#endif
1366 /*
1367 * No need to worry about unhashing the dentry, as the
1368 * lookup validation will know that the inode is bad.
1369 * (But we fall through to invalidate the caches.)
1370 */
1371 nfs_invalidate_inode(inode);
1372 out_err:
1373 NFS_FLAGS(inode) |= NFS_INO_STALE;
1374 return -ESTALE;
1375}
1376
1377/*
1378 * File system information
1379 */
1380
1381static int nfs_set_super(struct super_block *s, void *data)
1382{
1383 s->s_fs_info = data;
1384 return set_anon_super(s, data);
1385}
1386
1387static int nfs_compare_super(struct super_block *sb, void *data)
1388{
1389 struct nfs_server *server = data;
1390 struct nfs_server *old = NFS_SB(sb);
1391
1392 if (old->addr.sin_addr.s_addr != server->addr.sin_addr.s_addr)
1393 return 0;
1394 if (old->addr.sin_port != server->addr.sin_port)
1395 return 0;
1396 return !nfs_compare_fh(&old->fh, &server->fh);
1397}
1398
1399static struct super_block *nfs_get_sb(struct file_system_type *fs_type,
1400 int flags, const char *dev_name, void *raw_data)
1401{
1402 int error;
9085bbcb 1403 struct nfs_server *server = NULL;
1da177e4
LT		 1404 struct super_block *s;
1405 struct nfs_fh *root;
1406 struct nfs_mount_data *data = raw_data;
1407
9085bbcb
TM		 1408 s = ERR_PTR(-EINVAL);
1409 if (data == NULL) {
1410 dprintk("%s: missing data argument\n", __FUNCTION__);
1411 goto out_err;
1412 }
1413 if (data->version <= 0 || data->version > NFS_MOUNT_VERSION) {
1414 dprintk("%s: bad mount version\n", __FUNCTION__);
1415 goto out_err;
1da177e4 1416 }
9085bbcb
TM		 1417 switch (data->version) {
1418 case 1:
1419 data->namlen = 0;
1420 case 2:
1421 data->bsize = 0;
1422 case 3:
1423 if (data->flags & NFS_MOUNT_VER3) {
1424 dprintk("%s: mount structure version %d does not support NFSv3\n",
1425 __FUNCTION__,
1426 data->version);
1427 goto out_err;
1428 }
1429 data->root.size = NFS2_FHSIZE;
1430 memcpy(data->root.data, data->old_root.data, NFS2_FHSIZE);
1431 case 4:
1432 if (data->flags & NFS_MOUNT_SECFLAVOUR) {
1433 dprintk("%s: mount structure version %d does not support strong security\n",
1434 __FUNCTION__,
1435 data->version);
1436 goto out_err;
1437 }
1438 case 5:
1439 memset(data->context, 0, sizeof(data->context));
1440 }
1441#ifndef CONFIG_NFS_V3
1442 /* If NFSv3 is not compiled in, return -EPROTONOSUPPORT */
1443 s = ERR_PTR(-EPROTONOSUPPORT);
1444 if (data->flags & NFS_MOUNT_VER3) {
1445 dprintk("%s: NFSv3 not compiled into kernel\n", __FUNCTION__);
1446 goto out_err;
1447 }
1448#endif /* CONFIG_NFS_V3 */
1da177e4 1449
9085bbcb 1450 s = ERR_PTR(-ENOMEM);
1da177e4
LT		 1451 server = kmalloc(sizeof(struct nfs_server), GFP_KERNEL);
1452 if (!server)
9085bbcb 1453 goto out_err;
1da177e4
LT		 1454 memset(server, 0, sizeof(struct nfs_server));
1455 /* Zero out the NFS state stuff */
1456 init_nfsv4_state(server);
1457
1da177e4
LT		 1458 root = &server->fh;
1459 if (data->flags & NFS_MOUNT_VER3)
1460 root->size = data->root.size;
1461 else
1462 root->size = NFS2_FHSIZE;
9085bbcb 1463 s = ERR_PTR(-EINVAL);
1da177e4 1464 if (root->size > sizeof(root->data)) {
9085bbcb
TM		 1465 dprintk("%s: invalid root filehandle\n", __FUNCTION__);
1466 goto out_err;
1da177e4
LT		 1467 }
1468 memcpy(root->data, data->root.data, root->size);
1469
1470 /* We now require that the mount process passes the remote address */
1471 memcpy(&server->addr, &data->addr, sizeof(server->addr));
1472 if (server->addr.sin_addr.s_addr == INADDR_ANY) {
9085bbcb
TM		 1473 dprintk("%s: mount program didn't pass remote address!\n",
1474 __FUNCTION__);
1475 goto out_err;
1da177e4
LT		 1476 }
1477
9085bbcb
TM		 1478 /* Fire up rpciod if not yet running */
1479 s = ERR_PTR(rpciod_up());
1480 if (IS_ERR(s)) {
1481 dprintk("%s: couldn't start rpciod! Error = %ld\n",
1482 __FUNCTION__, PTR_ERR(s));
1483 goto out_err;
1da177e4
LT		 1484 }
1485
9085bbcb
TM		 1486 s = sget(fs_type, nfs_compare_super, nfs_set_super, server);
1487 if (IS_ERR(s) || s->s_root)
1488 goto out_rpciod_down;
1da177e4 1489
9085bbcb 1490 s->s_flags = flags;
1da177e4
LT		 1491
1492 error = nfs_fill_super(s, data, flags & MS_VERBOSE ? 1 : 0);
1493 if (error) {
1494 up_write(&s->s_umount);
1495 deactivate_super(s);
1496 return ERR_PTR(error);
1497 }
1498 s->s_flags |= MS_ACTIVE;
1499 return s;
9085bbcb
TM		 1500out_rpciod_down:
1501 rpciod_down();
1502out_err:
1503 kfree(server);
1504 return s;
1da177e4
LT		 1505}
1506
1507static void nfs_kill_super(struct super_block *s)
1508{
1509 struct nfs_server *server = NFS_SB(s);
1510
1511 kill_anon_super(s);
1512
1513 if (server->client != NULL && !IS_ERR(server->client))
1514 rpc_shutdown_client(server->client);
1515 if (server->client_sys != NULL && !IS_ERR(server->client_sys))
1516 rpc_shutdown_client(server->client_sys);
1517
1518 if (!(server->flags & NFS_MOUNT_NONLM))
1519 lockd_down(); /* release rpc.lockd */
1520
1521 rpciod_down(); /* release rpciod */
1522
1523 if (server->hostname != NULL)
1524 kfree(server->hostname);
1525 kfree(server);
1526}
1527
1528static struct file_system_type nfs_fs_type = {
1529 .owner = THIS_MODULE,
1530 .name = "nfs",
1531 .get_sb = nfs_get_sb,
1532 .kill_sb = nfs_kill_super,
1533 .fs_flags = FS_ODD_RENAME|FS_REVAL_DOT|FS_BINARY_MOUNTDATA,
1534};
1535
1536#ifdef CONFIG_NFS_V4
1537
1538static void nfs4_clear_inode(struct inode *);
1539
1540
1541static struct super_operations nfs4_sops = {
1542 .alloc_inode = nfs_alloc_inode,
1543 .destroy_inode = nfs_destroy_inode,
1544 .write_inode = nfs_write_inode,
1545 .delete_inode = nfs_delete_inode,
1546 .statfs = nfs_statfs,
1547 .clear_inode = nfs4_clear_inode,
1548 .umount_begin = nfs_umount_begin,
1549 .show_options = nfs_show_options,
1550};
1551
1552/*
1553 * Clean out any remaining NFSv4 state that might be left over due
1554 * to open() calls that passed nfs_atomic_lookup, but failed to call
1555 * nfs_open().
1556 */
1557static void nfs4_clear_inode(struct inode *inode)
1558{
1559 struct nfs_inode *nfsi = NFS_I(inode);
1560
1561 /* If we are holding a delegation, return it! */
1562 if (nfsi->delegation != NULL)
1563 nfs_inode_return_delegation(inode);
1564 /* First call standard NFS clear_inode() code */
1565 nfs_clear_inode(inode);
1566 /* Now clear out any remaining state */
1567 while (!list_empty(&nfsi->open_states)) {
1568 struct nfs4_state *state;
1569
1570 state = list_entry(nfsi->open_states.next,
1571 struct nfs4_state,
1572 inode_states);
1573 dprintk("%s(%s/%Ld): found unclaimed NFSv4 state %p\n",
1574 __FUNCTION__,
1575 inode->i_sb->s_id,
1576 (long long)NFS_FILEID(inode),
1577 state);
1578 BUG_ON(atomic_read(&state->count) != 1);
1579 nfs4_close_state(state, state->state);
1580 }
1581}
1582
1583
1584static int nfs4_fill_super(struct super_block *sb, struct nfs4_mount_data *data, int silent)
1585{
1586 struct nfs_server *server;
1587 struct nfs4_client *clp = NULL;
1588 struct rpc_xprt *xprt = NULL;
1589 struct rpc_clnt *clnt = NULL;
1590 struct rpc_timeout timeparms;
1591 rpc_authflavor_t authflavour;
1592 int proto, err = -EIO;
1593
1594 sb->s_blocksize_bits = 0;
1595 sb->s_blocksize = 0;
1596 server = NFS_SB(sb);
1597 if (data->rsize != 0)
1598 server->rsize = nfs_block_size(data->rsize, NULL);
1599 if (data->wsize != 0)
1600 server->wsize = nfs_block_size(data->wsize, NULL);
1601 server->flags = data->flags & NFS_MOUNT_FLAGMASK;
1602 server->caps = NFS_CAP_ATOMIC_OPEN;
1603
1604 server->acregmin = data->acregmin*HZ;
1605 server->acregmax = data->acregmax*HZ;
1606 server->acdirmin = data->acdirmin*HZ;
1607 server->acdirmax = data->acdirmax*HZ;
1608
1609 server->rpc_ops = &nfs_v4_clientops;
1610 /* Initialize timeout values */
1611
1612 timeparms.to_initval = data->timeo * HZ / 10;
1613 timeparms.to_retries = data->retrans;
1614 timeparms.to_exponential = 1;
1615 if (!timeparms.to_retries)
1616 timeparms.to_retries = 5;
1617
1618 proto = data->proto;
1619 /* Which IP protocol do we use? */
1620 switch (proto) {
1621 case IPPROTO_TCP:
1622 timeparms.to_maxval = RPC_MAX_TCP_TIMEOUT;
1623 if (!timeparms.to_initval)
1624 timeparms.to_initval = 600 * HZ / 10;
1625 break;
1626 case IPPROTO_UDP:
1627 timeparms.to_maxval = RPC_MAX_UDP_TIMEOUT;
1628 if (!timeparms.to_initval)
1629 timeparms.to_initval = 11 * HZ / 10;
1630 break;
1631 default:
1632 return -EINVAL;
1633 }
1634
1635 clp = nfs4_get_client(&server->addr.sin_addr);
1636 if (!clp) {
9085bbcb 1637 dprintk("%s: failed to create NFS4 client.\n", __FUNCTION__);
1da177e4
LT		 1638 return -EIO;
1639 }
1640
1641 /* Now create transport and client */
1642 authflavour = RPC_AUTH_UNIX;
1643 if (data->auth_flavourlen != 0) {
9085bbcb
TM		 1644 if (data->auth_flavourlen != 1) {
1645 dprintk("%s: Invalid number of RPC auth flavours %d.\n",
1646 __FUNCTION__, data->auth_flavourlen);
1647 err = -EINVAL;
1648 goto out_fail;
1649 }
1da177e4
LT		 1650 if (copy_from_user(&authflavour, data->auth_flavours, sizeof(authflavour))) {
1651 err = -EFAULT;
1652 goto out_fail;
1653 }
1654 }
1655
1656 down_write(&clp->cl_sem);
1657 if (clp->cl_rpcclient == NULL) {
1658 xprt = xprt_create_proto(proto, &server->addr, &timeparms);
1659 if (IS_ERR(xprt)) {
1660 up_write(&clp->cl_sem);
1da177e4 1661 err = PTR_ERR(xprt);
9085bbcb
TM		 1662 dprintk("%s: cannot create RPC transport. Error = %d\n",
1663 __FUNCTION__, err);
1da177e4
LT		 1664 goto out_fail;
1665 }
1666 clnt = rpc_create_client(xprt, server->hostname, &nfs_program,
1667 server->rpc_ops->version, authflavour);
1668 if (IS_ERR(clnt)) {
1669 up_write(&clp->cl_sem);
1da177e4 1670 err = PTR_ERR(clnt);
9085bbcb
TM		 1671 dprintk("%s: cannot create RPC client. Error = %d\n",
1672 __FUNCTION__, err);
1da177e4
LT		 1673 goto out_fail;
1674 }
1675 clnt->cl_intr = 1;
1676 clnt->cl_softrtry = 1;
1677 clnt->cl_chatty = 1;
1678 clp->cl_rpcclient = clnt;
1679 clp->cl_cred = rpcauth_lookupcred(clnt->cl_auth, 0);
1680 if (IS_ERR(clp->cl_cred)) {
1681 up_write(&clp->cl_sem);
1682 err = PTR_ERR(clp->cl_cred);
1683 clp->cl_cred = NULL;
1684 goto out_fail;
1685 }
1686 memcpy(clp->cl_ipaddr, server->ip_addr, sizeof(clp->cl_ipaddr));
1687 nfs_idmap_new(clp);
1688 }
1689 if (list_empty(&clp->cl_superblocks)) {
1690 err = nfs4_init_client(clp);
1691 if (err != 0) {
1692 up_write(&clp->cl_sem);
1693 goto out_fail;
1694 }
1695 }
1696 list_add_tail(&server->nfs4_siblings, &clp->cl_superblocks);
1697 clnt = rpc_clone_client(clp->cl_rpcclient);
1698 if (!IS_ERR(clnt))
1699 server->nfs4_state = clp;
1700 up_write(&clp->cl_sem);
1701 clp = NULL;
1702
1703 if (IS_ERR(clnt)) {
9085bbcb
TM		 1704 err = PTR_ERR(clnt);
1705 dprintk("%s: cannot create RPC client. Error = %d\n",
1706 __FUNCTION__, err);
1707 return err;
1da177e4
LT		 1708 }
1709
1710 server->client = clnt;
1711
1712 if (server->nfs4_state->cl_idmap == NULL) {
9085bbcb 1713 dprintk("%s: failed to create idmapper.\n", __FUNCTION__);
1da177e4
LT		 1714 return -ENOMEM;
1715 }
1716
1717 if (clnt->cl_auth->au_flavor != authflavour) {
1718 if (rpcauth_create(authflavour, clnt) == NULL) {
9085bbcb 1719 dprintk("%s: couldn't create credcache!\n", __FUNCTION__);
1da177e4
LT		 1720 return -ENOMEM;
1721 }
1722 }
1723
1724 sb->s_time_gran = 1;
1725
1726 sb->s_op = &nfs4_sops;
1727 err = nfs_sb_init(sb, authflavour);
1728 if (err == 0)
1729 return 0;
1730out_fail:
1731 if (clp)
1732 nfs4_put_client(clp);
1733 return err;
1734}
1735
1736static int nfs4_compare_super(struct super_block *sb, void *data)
1737{
1738 struct nfs_server *server = data;
1739 struct nfs_server *old = NFS_SB(sb);
1740
1741 if (strcmp(server->hostname, old->hostname) != 0)
1742 return 0;
1743 if (strcmp(server->mnt_path, old->mnt_path) != 0)
1744 return 0;
1745 return 1;
1746}
1747
1748static void *
1749nfs_copy_user_string(char *dst, struct nfs_string *src, int maxlen)
1750{
1751 void *p = NULL;
1752
1753 if (!src->len)
1754 return ERR_PTR(-EINVAL);
1755 if (src->len < maxlen)
1756 maxlen = src->len;
1757 if (dst == NULL) {
1758 p = dst = kmalloc(maxlen + 1, GFP_KERNEL);
1759 if (p == NULL)
1760 return ERR_PTR(-ENOMEM);
1761 }
1762 if (copy_from_user(dst, src->data, maxlen)) {
1763 if (p != NULL)
1764 kfree(p);
1765 return ERR_PTR(-EFAULT);
1766 }
1767 dst[maxlen] = '\0';
1768 return dst;
1769}
1770
1771static struct super_block *nfs4_get_sb(struct file_system_type *fs_type,
1772 int flags, const char *dev_name, void *raw_data)
1773{
1774 int error;
1775 struct nfs_server *server;
1776 struct super_block *s;
1777 struct nfs4_mount_data *data = raw_data;
1778 void *p;
1779
9085bbcb
TM		 1780 if (data == NULL) {
1781 dprintk("%s: missing data argument\n", __FUNCTION__);
1782 return ERR_PTR(-EINVAL);
1783 }
1784 if (data->version <= 0 || data->version > NFS4_MOUNT_VERSION) {
1785 dprintk("%s: bad mount version\n", __FUNCTION__);
1da177e4
LT		 1786 return ERR_PTR(-EINVAL);
1787 }
1788
1789 server = kmalloc(sizeof(struct nfs_server), GFP_KERNEL);
1790 if (!server)
1791 return ERR_PTR(-ENOMEM);
1792 memset(server, 0, sizeof(struct nfs_server));
1793 /* Zero out the NFS state stuff */
1794 init_nfsv4_state(server);
1795
1da177e4
LT		 1796 p = nfs_copy_user_string(NULL, &data->hostname, 256);
1797 if (IS_ERR(p))
1798 goto out_err;
1799 server->hostname = p;
1800
1801 p = nfs_copy_user_string(NULL, &data->mnt_path, 1024);
1802 if (IS_ERR(p))
1803 goto out_err;
1804 server->mnt_path = p;
1805
1806 p = nfs_copy_user_string(server->ip_addr, &data->client_addr,
1807 sizeof(server->ip_addr) - 1);
1808 if (IS_ERR(p))
1809 goto out_err;
1810
1811 /* We now require that the mount process passes the remote address */
1812 if (data->host_addrlen != sizeof(server->addr)) {
1813 s = ERR_PTR(-EINVAL);
1814 goto out_free;
1815 }
1816 if (copy_from_user(&server->addr, data->host_addr, sizeof(server->addr))) {
1817 s = ERR_PTR(-EFAULT);
1818 goto out_free;
1819 }
1820 if (server->addr.sin_family != AF_INET ||
1821 server->addr.sin_addr.s_addr == INADDR_ANY) {
9085bbcb
TM		 1822 dprintk("%s: mount program didn't pass remote IP address!\n",
1823 __FUNCTION__);
1da177e4
LT		 1824 s = ERR_PTR(-EINVAL);
1825 goto out_free;
1826 }
1827
9085bbcb
TM		 1828 /* Fire up rpciod if not yet running */
1829 s = ERR_PTR(rpciod_up());
1830 if (IS_ERR(s)) {
1831 dprintk("%s: couldn't start rpciod! Error = %ld\n",
1832 __FUNCTION__, PTR_ERR(s));
1833 goto out_free;
1834 }
1835
1da177e4
LT		 1836 s = sget(fs_type, nfs4_compare_super, nfs_set_super, server);
1837
1838 if (IS_ERR(s) || s->s_root)
1839 goto out_free;
1840
1841 s->s_flags = flags;
1842
1da177e4
LT		 1843 error = nfs4_fill_super(s, data, flags & MS_VERBOSE ? 1 : 0);
1844 if (error) {
1845 up_write(&s->s_umount);
1846 deactivate_super(s);
1847 return ERR_PTR(error);
1848 }
1849 s->s_flags |= MS_ACTIVE;
1850 return s;
1851out_err:
1852 s = (struct super_block *)p;
1853out_free:
1854 if (server->mnt_path)
1855 kfree(server->mnt_path);
1856 if (server->hostname)
1857 kfree(server->hostname);
1858 kfree(server);
1859 return s;
1860}
1861
1862static void nfs4_kill_super(struct super_block *sb)
1863{
1864 struct nfs_server *server = NFS_SB(sb);
1865
1866 nfs_return_all_delegations(sb);
1867 kill_anon_super(sb);
1868
1869 nfs4_renewd_prepare_shutdown(server);
1870
1871 if (server->client != NULL && !IS_ERR(server->client))
1872 rpc_shutdown_client(server->client);
1873 rpciod_down(); /* release rpciod */
1874
1875 destroy_nfsv4_state(server);
1876
1877 if (server->hostname != NULL)
1878 kfree(server->hostname);
1879 kfree(server);
1880}
1881
1882static struct file_system_type nfs4_fs_type = {
1883 .owner = THIS_MODULE,
1884 .name = "nfs4",
1885 .get_sb = nfs4_get_sb,
1886 .kill_sb = nfs4_kill_super,
1887 .fs_flags = FS_ODD_RENAME|FS_REVAL_DOT|FS_BINARY_MOUNTDATA,
1888};
1889
1890#define nfs4_init_once(nfsi) \
1891 do { \
1892 INIT_LIST_HEAD(&(nfsi)->open_states); \
1893 nfsi->delegation = NULL; \
1894 nfsi->delegation_state = 0; \
1895 init_rwsem(&nfsi->rwsem); \
1896 } while(0)
1897#define register_nfs4fs() register_filesystem(&nfs4_fs_type)
1898#define unregister_nfs4fs() unregister_filesystem(&nfs4_fs_type)
1899#else
1900#define nfs4_init_once(nfsi) \
1901 do { } while (0)
1902#define register_nfs4fs() (0)
1903#define unregister_nfs4fs()
1904#endif
1905
1906extern int nfs_init_nfspagecache(void);
1907extern void nfs_destroy_nfspagecache(void);
1908extern int nfs_init_readpagecache(void);
1909extern void nfs_destroy_readpagecache(void);
1910extern int nfs_init_writepagecache(void);
1911extern void nfs_destroy_writepagecache(void);
1912#ifdef CONFIG_NFS_DIRECTIO
1913extern int nfs_init_directcache(void);
1914extern void nfs_destroy_directcache(void);
1915#endif
1916
1917static kmem_cache_t * nfs_inode_cachep;
1918
1919static struct inode *nfs_alloc_inode(struct super_block *sb)
1920{
1921 struct nfs_inode *nfsi;
1922 nfsi = (struct nfs_inode *)kmem_cache_alloc(nfs_inode_cachep, SLAB_KERNEL);
1923 if (!nfsi)
1924 return NULL;
1925 nfsi->flags = 0;
1926 return &nfsi->vfs_inode;
1927}
1928
1929static void nfs_destroy_inode(struct inode *inode)
1930{
1931 kmem_cache_free(nfs_inode_cachep, NFS_I(inode));
1932}
1933
1934static void init_once(void * foo, kmem_cache_t * cachep, unsigned long flags)
1935{
1936 struct nfs_inode *nfsi = (struct nfs_inode *) foo;
1937
1938 if ((flags & (SLAB_CTOR_VERIFY|SLAB_CTOR_CONSTRUCTOR)) ==
1939 SLAB_CTOR_CONSTRUCTOR) {
1940 inode_init_once(&nfsi->vfs_inode);
1941 spin_lock_init(&nfsi->req_lock);
1942 INIT_LIST_HEAD(&nfsi->dirty);
1943 INIT_LIST_HEAD(&nfsi->commit);
1944 INIT_LIST_HEAD(&nfsi->open_files);
1945 INIT_RADIX_TREE(&nfsi->nfs_page_tree, GFP_ATOMIC);
1946 atomic_set(&nfsi->data_updates, 0);
1947 nfsi->ndirty = 0;
1948 nfsi->ncommit = 0;
1949 nfsi->npages = 0;
1950 init_waitqueue_head(&nfsi->nfs_i_wait);
1951 nfs4_init_once(nfsi);
1952 }
1953}
1954
75c96f85 1955static int nfs_init_inodecache(void)
1da177e4
LT		 1956{
1957 nfs_inode_cachep = kmem_cache_create("nfs_inode_cache",
1958 sizeof(struct nfs_inode),
1959 0, SLAB_RECLAIM_ACCOUNT,
1960 init_once, NULL);
1961 if (nfs_inode_cachep == NULL)
1962 return -ENOMEM;
1963
1964 return 0;
1965}
1966
75c96f85 1967static void nfs_destroy_inodecache(void)
1da177e4
LT		 1968{
1969 if (kmem_cache_destroy(nfs_inode_cachep))
1970 printk(KERN_INFO "nfs_inode_cache: not all structures were freed\n");
1971}
1972
1973/*
1974 * Initialize NFS
1975 */
1976static int __init init_nfs_fs(void)
1977{
1978 int err;
1979
1980 err = nfs_init_nfspagecache();
1981 if (err)
1982 goto out4;
1983
1984 err = nfs_init_inodecache();
1985 if (err)
1986 goto out3;
1987
1988 err = nfs_init_readpagecache();
1989 if (err)
1990 goto out2;
1991
1992 err = nfs_init_writepagecache();
1993 if (err)
1994 goto out1;
1995
1996#ifdef CONFIG_NFS_DIRECTIO
1997 err = nfs_init_directcache();
1998 if (err)
1999 goto out0;
2000#endif
2001
2002#ifdef CONFIG_PROC_FS
2003 rpc_proc_register(&nfs_rpcstat);
2004#endif
2005 err = register_filesystem(&nfs_fs_type);
2006 if (err)
2007 goto out;
2008 if ((err = register_nfs4fs()) != 0)
2009 goto out;
2010 return 0;
2011out:
2012#ifdef CONFIG_PROC_FS
2013 rpc_proc_unregister("nfs");
2014#endif
2015 nfs_destroy_writepagecache();
2016#ifdef CONFIG_NFS_DIRECTIO
2017out0:
2018 nfs_destroy_directcache();
2019#endif
2020out1:
2021 nfs_destroy_readpagecache();
2022out2:
2023 nfs_destroy_inodecache();
2024out3:
2025 nfs_destroy_nfspagecache();
2026out4:
2027 return err;
2028}
2029
2030static void __exit exit_nfs_fs(void)
2031{
2032#ifdef CONFIG_NFS_DIRECTIO
2033 nfs_destroy_directcache();
2034#endif
2035 nfs_destroy_writepagecache();
2036 nfs_destroy_readpagecache();
2037 nfs_destroy_inodecache();
2038 nfs_destroy_nfspagecache();
2039#ifdef CONFIG_PROC_FS
2040 rpc_proc_unregister("nfs");
2041#endif
2042 unregister_filesystem(&nfs_fs_type);
2043 unregister_nfs4fs();
2044}
2045
2046/* Not quite true; I just maintain it */
2047MODULE_AUTHOR("Olaf Kirch <okir@monad.swb.de>");
2048MODULE_LICENSE("GPL");
2049
2050module_init(init_nfs_fs)
2051module_exit(exit_nfs_fs)
Linux 6.x block layer and io_uring tree(s)