4 * Copyright (C) 1992 Rick Sladkey
6 * nfs inode and superblock handling functions
8 * Modularised by Alan Cox <Alan.Cox@linux.org>, while hacking some
9 * experimental NFS changes. Modularisation taken straight from SYS5 fs.
11 * Change to nfs_read_super() to permit NFS mounts to multi-homed hosts.
12 * J.S.Peatfield@damtp.cam.ac.uk
16 #include <linux/config.h>
17 #include <linux/module.h>
18 #include <linux/init.h>
20 #include <linux/time.h>
21 #include <linux/kernel.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>
39 #include <asm/system.h>
40 #include <asm/uaccess.h>
43 #include "delegation.h"
45 #define NFSDBG_FACILITY NFSDBG_VFS
46 #define NFS_PARANOIA 1
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.
54 #define NFS_MAX_READAHEAD (RPC_DEF_SLOT_TABLE - 1)
56 static void nfs_invalidate_inode(struct inode
*);
57 static int nfs_update_inode(struct inode
*, struct nfs_fattr
*, unsigned long);
59 static struct inode
*nfs_alloc_inode(struct super_block
*sb
);
60 static void nfs_destroy_inode(struct inode
*);
61 static int nfs_write_inode(struct inode
*,int);
62 static void nfs_delete_inode(struct inode
*);
63 static void nfs_clear_inode(struct inode
*);
64 static void nfs_umount_begin(struct super_block
*);
65 static int nfs_statfs(struct super_block
*, struct kstatfs
*);
66 static int nfs_show_options(struct seq_file
*, struct vfsmount
*);
67 static void nfs_zap_acl_cache(struct inode
*);
69 static struct rpc_program nfs_program
;
71 static 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
,
77 .clear_inode
= nfs_clear_inode
,
78 .umount_begin
= nfs_umount_begin
,
79 .show_options
= nfs_show_options
,
85 static struct rpc_stat nfs_rpcstat
= {
86 .program
= &nfs_program
88 static struct rpc_version
* nfs_version
[] = {
92 #if defined(CONFIG_NFS_V3)
94 #elif defined(CONFIG_NFS_V4)
97 #if defined(CONFIG_NFS_V4)
102 static struct rpc_program nfs_program
= {
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",
111 static inline unsigned long
112 nfs_fattr_to_ino_t(struct nfs_fattr
*fattr
)
114 return nfs_fileid_to_ino_t(fattr
->fileid
);
118 nfs_write_inode(struct inode
*inode
, int sync
)
120 int flags
= sync
? FLUSH_WAIT
: 0;
123 ret
= nfs_commit_inode(inode
, 0, 0, flags
);
130 nfs_delete_inode(struct inode
* inode
)
132 dprintk("NFS: delete_inode(%s/%ld)\n", inode
->i_sb
->s_id
, inode
->i_ino
);
136 * The following should never happen...
138 if (nfs_have_writebacks(inode
)) {
139 printk(KERN_ERR
"nfs_delete_inode: inode %ld has pending RPC requests\n", inode
->i_ino
);
146 nfs_clear_inode(struct inode
*inode
)
148 struct nfs_inode
*nfsi
= NFS_I(inode
);
149 struct rpc_cred
*cred
;
152 BUG_ON (!list_empty(&nfsi
->open_files
));
153 nfs_zap_acl_cache(inode
);
154 cred
= nfsi
->cache_access
.cred
;
157 BUG_ON(atomic_read(&nfsi
->data_updates
) != 0);
161 nfs_umount_begin(struct super_block
*sb
)
163 struct nfs_server
*server
= NFS_SB(sb
);
164 struct rpc_clnt
*rpc
;
166 /* -EIO all pending I/O */
167 if ((rpc
= server
->client
) != NULL
)
168 rpc_killall_tasks(rpc
);
172 static inline unsigned long
173 nfs_block_bits(unsigned long bsize
, unsigned char *nrbitsp
)
175 /* make sure blocksize is a power of two */
176 if ((bsize
& (bsize
- 1)) || nrbitsp
) {
177 unsigned char nrbits
;
179 for (nrbits
= 31; nrbits
&& !(bsize
& (1 << nrbits
)); nrbits
--)
190 * Calculate the number of 512byte blocks used.
192 static inline unsigned long
193 nfs_calc_block_size(u64 tsize
)
195 loff_t used
= (tsize
+ 511) >> 9;
196 return (used
> ULONG_MAX
) ? ULONG_MAX
: used
;
200 * Compute and set NFS server blocksize
202 static inline unsigned long
203 nfs_block_size(unsigned long bsize
, unsigned char *nrbitsp
)
206 bsize
= NFS_DEF_FILE_IO_BUFFER_SIZE
;
207 else if (bsize
>= NFS_MAX_FILE_IO_BUFFER_SIZE
)
208 bsize
= NFS_MAX_FILE_IO_BUFFER_SIZE
;
210 return nfs_block_bits(bsize
, nrbitsp
);
214 * Obtain the root inode of the file system.
216 static struct inode
*
217 nfs_get_root(struct super_block
*sb
, struct nfs_fh
*rootfh
, struct nfs_fsinfo
*fsinfo
)
219 struct nfs_server
*server
= NFS_SB(sb
);
223 error
= server
->rpc_ops
->getroot(server
, rootfh
, fsinfo
);
225 dprintk("nfs_get_root: getattr error = %d\n", -error
);
226 return ERR_PTR(error
);
229 rooti
= nfs_fhget(sb
, rootfh
, fsinfo
->fattr
);
231 return ERR_PTR(-ENOMEM
);
236 * Do NFS version-independent mount processing, and sanity checking
239 nfs_sb_init(struct super_block
*sb
, rpc_authflavor_t authflavor
)
241 struct nfs_server
*server
;
242 struct inode
*root_inode
;
243 struct nfs_fattr fattr
;
244 struct nfs_fsinfo fsinfo
= {
247 struct nfs_pathconf pathinfo
= {
250 int no_root_error
= 0;
251 unsigned long max_rpc_payload
;
253 /* We probably want something more informative here */
254 snprintf(sb
->s_id
, sizeof(sb
->s_id
), "%x:%x", MAJOR(sb
->s_dev
), MINOR(sb
->s_dev
));
258 sb
->s_magic
= NFS_SUPER_MAGIC
;
260 root_inode
= nfs_get_root(sb
, &server
->fh
, &fsinfo
);
261 /* Did getting the root inode fail? */
262 if (IS_ERR(root_inode
)) {
263 no_root_error
= PTR_ERR(root_inode
);
266 sb
->s_root
= d_alloc_root(root_inode
);
268 no_root_error
= -ENOMEM
;
271 sb
->s_root
->d_op
= server
->rpc_ops
->dentry_ops
;
273 /* Get some general file system info */
274 if (server
->namelen
== 0 &&
275 server
->rpc_ops
->pathconf(server
, &server
->fh
, &pathinfo
) >= 0)
276 server
->namelen
= pathinfo
.max_namelen
;
277 /* Work out a lot of parameters */
278 if (server
->rsize
== 0)
279 server
->rsize
= nfs_block_size(fsinfo
.rtpref
, NULL
);
280 if (server
->wsize
== 0)
281 server
->wsize
= nfs_block_size(fsinfo
.wtpref
, NULL
);
283 if (fsinfo
.rtmax
>= 512 && server
->rsize
> fsinfo
.rtmax
)
284 server
->rsize
= nfs_block_size(fsinfo
.rtmax
, NULL
);
285 if (fsinfo
.wtmax
>= 512 && server
->wsize
> fsinfo
.wtmax
)
286 server
->wsize
= nfs_block_size(fsinfo
.wtmax
, NULL
);
288 max_rpc_payload
= nfs_block_size(rpc_max_payload(server
->client
), NULL
);
289 if (server
->rsize
> max_rpc_payload
)
290 server
->rsize
= max_rpc_payload
;
291 if (server
->wsize
> max_rpc_payload
)
292 server
->wsize
= max_rpc_payload
;
294 server
->rpages
= (server
->rsize
+ PAGE_CACHE_SIZE
- 1) >> PAGE_CACHE_SHIFT
;
295 if (server
->rpages
> NFS_READ_MAXIOV
) {
296 server
->rpages
= NFS_READ_MAXIOV
;
297 server
->rsize
= server
->rpages
<< PAGE_CACHE_SHIFT
;
300 server
->wpages
= (server
->wsize
+ PAGE_CACHE_SIZE
- 1) >> PAGE_CACHE_SHIFT
;
301 if (server
->wpages
> NFS_WRITE_MAXIOV
) {
302 server
->wpages
= NFS_WRITE_MAXIOV
;
303 server
->wsize
= server
->wpages
<< PAGE_CACHE_SHIFT
;
306 if (sb
->s_blocksize
== 0)
307 sb
->s_blocksize
= nfs_block_bits(server
->wsize
,
308 &sb
->s_blocksize_bits
);
309 server
->wtmult
= nfs_block_bits(fsinfo
.wtmult
, NULL
);
311 server
->dtsize
= nfs_block_size(fsinfo
.dtpref
, NULL
);
312 if (server
->dtsize
> PAGE_CACHE_SIZE
)
313 server
->dtsize
= PAGE_CACHE_SIZE
;
314 if (server
->dtsize
> server
->rsize
)
315 server
->dtsize
= server
->rsize
;
317 if (server
->flags
& NFS_MOUNT_NOAC
) {
318 server
->acregmin
= server
->acregmax
= 0;
319 server
->acdirmin
= server
->acdirmax
= 0;
320 sb
->s_flags
|= MS_SYNCHRONOUS
;
322 server
->backing_dev_info
.ra_pages
= server
->rpages
* NFS_MAX_READAHEAD
;
324 sb
->s_maxbytes
= fsinfo
.maxfilesize
;
325 if (sb
->s_maxbytes
> MAX_LFS_FILESIZE
)
326 sb
->s_maxbytes
= MAX_LFS_FILESIZE
;
328 server
->client
->cl_intr
= (server
->flags
& NFS_MOUNT_INTR
) ? 1 : 0;
329 server
->client
->cl_softrtry
= (server
->flags
& NFS_MOUNT_SOFT
) ? 1 : 0;
331 /* We're airborne Set socket buffersize */
332 rpc_setbufsize(server
->client
, server
->wsize
+ 100, server
->rsize
+ 100);
334 /* Yargs. It didn't work out. */
336 dprintk("nfs_sb_init: get root inode failed: errno %d\n", -no_root_error
);
337 if (!IS_ERR(root_inode
))
339 return no_root_error
;
343 * Create an RPC client handle.
345 static struct rpc_clnt
*
346 nfs_create_client(struct nfs_server
*server
, const struct nfs_mount_data
*data
)
348 struct rpc_timeout timeparms
;
349 struct rpc_xprt
*xprt
= NULL
;
350 struct rpc_clnt
*clnt
= NULL
;
351 int tcp
= (data
->flags
& NFS_MOUNT_TCP
);
353 /* Initialize timeout values */
354 timeparms
.to_initval
= data
->timeo
* HZ
/ 10;
355 timeparms
.to_retries
= data
->retrans
;
356 timeparms
.to_maxval
= tcp
? RPC_MAX_TCP_TIMEOUT
: RPC_MAX_UDP_TIMEOUT
;
357 timeparms
.to_exponential
= 1;
359 if (!timeparms
.to_initval
)
360 timeparms
.to_initval
= (tcp
? 600 : 11) * HZ
/ 10;
361 if (!timeparms
.to_retries
)
362 timeparms
.to_retries
= 5;
364 /* create transport and client */
365 xprt
= xprt_create_proto(tcp
? IPPROTO_TCP
: IPPROTO_UDP
,
366 &server
->addr
, &timeparms
);
368 dprintk("%s: cannot create RPC transport. Error = %ld\n",
369 __FUNCTION__
, PTR_ERR(xprt
));
370 return (struct rpc_clnt
*)xprt
;
372 clnt
= rpc_create_client(xprt
, server
->hostname
, &nfs_program
,
373 server
->rpc_ops
->version
, data
->pseudoflavor
);
375 dprintk("%s: cannot create RPC client. Error = %ld\n",
376 __FUNCTION__
, PTR_ERR(xprt
));
381 clnt
->cl_softrtry
= 1;
391 * The way this works is that the mount process passes a structure
392 * in the data argument which contains the server's IP address
393 * and the root file handle obtained from the server's mount
394 * daemon. We stash these away in the private superblock fields.
397 nfs_fill_super(struct super_block
*sb
, struct nfs_mount_data
*data
, int silent
)
399 struct nfs_server
*server
;
400 rpc_authflavor_t authflavor
;
403 sb
->s_blocksize_bits
= 0;
406 sb
->s_blocksize
= nfs_block_size(data
->bsize
, &sb
->s_blocksize_bits
);
408 server
->rsize
= nfs_block_size(data
->rsize
, NULL
);
410 server
->wsize
= nfs_block_size(data
->wsize
, NULL
);
411 server
->flags
= data
->flags
& NFS_MOUNT_FLAGMASK
;
413 server
->acregmin
= data
->acregmin
*HZ
;
414 server
->acregmax
= data
->acregmax
*HZ
;
415 server
->acdirmin
= data
->acdirmin
*HZ
;
416 server
->acdirmax
= data
->acdirmax
*HZ
;
418 /* Start lockd here, before we might error out */
419 if (!(server
->flags
& NFS_MOUNT_NONLM
))
422 server
->namelen
= data
->namlen
;
423 server
->hostname
= kmalloc(strlen(data
->hostname
) + 1, GFP_KERNEL
);
424 if (!server
->hostname
)
426 strcpy(server
->hostname
, data
->hostname
);
428 /* Check NFS protocol revision and initialize RPC op vector
429 * and file handle pool. */
431 if (server
->flags
& NFS_MOUNT_VER3
) {
432 server
->rpc_ops
= &nfs_v3_clientops
;
433 server
->caps
|= NFS_CAP_READDIRPLUS
;
435 server
->rpc_ops
= &nfs_v2_clientops
;
438 server
->rpc_ops
= &nfs_v2_clientops
;
441 /* Fill in pseudoflavor for mount version < 5 */
442 if (!(data
->flags
& NFS_MOUNT_SECFLAVOUR
))
443 data
->pseudoflavor
= RPC_AUTH_UNIX
;
444 authflavor
= data
->pseudoflavor
; /* save for sb_init() */
445 /* XXX maybe we want to add a server->pseudoflavor field */
447 /* Create RPC client handles */
448 server
->client
= nfs_create_client(server
, data
);
449 if (IS_ERR(server
->client
))
450 return PTR_ERR(server
->client
);
451 /* RFC 2623, sec 2.3.2 */
452 if (authflavor
!= RPC_AUTH_UNIX
) {
453 server
->client_sys
= rpc_clone_client(server
->client
);
454 if (IS_ERR(server
->client_sys
))
455 return PTR_ERR(server
->client_sys
);
456 if (!rpcauth_create(RPC_AUTH_UNIX
, server
->client_sys
))
459 atomic_inc(&server
->client
->cl_count
);
460 server
->client_sys
= server
->client
;
463 if (server
->flags
& NFS_MOUNT_VER3
) {
464 if (server
->namelen
== 0 || server
->namelen
> NFS3_MAXNAMLEN
)
465 server
->namelen
= NFS3_MAXNAMLEN
;
468 if (server
->namelen
== 0 || server
->namelen
> NFS2_MAXNAMLEN
)
469 server
->namelen
= NFS2_MAXNAMLEN
;
472 sb
->s_op
= &nfs_sops
;
473 return nfs_sb_init(sb
, authflavor
);
477 nfs_statfs(struct super_block
*sb
, struct kstatfs
*buf
)
479 struct nfs_server
*server
= NFS_SB(sb
);
480 unsigned char blockbits
;
481 unsigned long blockres
;
482 struct nfs_fh
*rootfh
= NFS_FH(sb
->s_root
->d_inode
);
483 struct nfs_fattr fattr
;
484 struct nfs_fsstat res
= {
491 error
= server
->rpc_ops
->statfs(server
, rootfh
, &res
);
492 buf
->f_type
= NFS_SUPER_MAGIC
;
497 * Current versions of glibc do not correctly handle the
498 * case where f_frsize != f_bsize. Eventually we want to
499 * report the value of wtmult in this field.
501 buf
->f_frsize
= sb
->s_blocksize
;
504 * On most *nix systems, f_blocks, f_bfree, and f_bavail
505 * are reported in units of f_frsize. Linux hasn't had
506 * an f_frsize field in its statfs struct until recently,
507 * thus historically Linux's sys_statfs reports these
508 * fields in units of f_bsize.
510 buf
->f_bsize
= sb
->s_blocksize
;
511 blockbits
= sb
->s_blocksize_bits
;
512 blockres
= (1 << blockbits
) - 1;
513 buf
->f_blocks
= (res
.tbytes
+ blockres
) >> blockbits
;
514 buf
->f_bfree
= (res
.fbytes
+ blockres
) >> blockbits
;
515 buf
->f_bavail
= (res
.abytes
+ blockres
) >> blockbits
;
517 buf
->f_files
= res
.tfiles
;
518 buf
->f_ffree
= res
.afiles
;
520 buf
->f_namelen
= server
->namelen
;
527 printk(KERN_WARNING
"nfs_statfs: statfs error = %d\n", -error
);
528 buf
->f_bsize
= buf
->f_blocks
= buf
->f_bfree
= buf
->f_bavail
= -1;
533 static int nfs_show_options(struct seq_file
*m
, struct vfsmount
*mnt
)
535 static struct proc_nfs_info
{
540 { NFS_MOUNT_SOFT
, ",soft", ",hard" },
541 { NFS_MOUNT_INTR
, ",intr", "" },
542 { NFS_MOUNT_POSIX
, ",posix", "" },
543 { NFS_MOUNT_TCP
, ",tcp", ",udp" },
544 { NFS_MOUNT_NOCTO
, ",nocto", "" },
545 { NFS_MOUNT_NOAC
, ",noac", "" },
546 { NFS_MOUNT_NONLM
, ",nolock", ",lock" },
549 struct proc_nfs_info
*nfs_infop
;
550 struct nfs_server
*nfss
= NFS_SB(mnt
->mnt_sb
);
552 seq_printf(m
, ",v%d", nfss
->rpc_ops
->version
);
553 seq_printf(m
, ",rsize=%d", nfss
->rsize
);
554 seq_printf(m
, ",wsize=%d", nfss
->wsize
);
555 if (nfss
->acregmin
!= 3*HZ
)
556 seq_printf(m
, ",acregmin=%d", nfss
->acregmin
/HZ
);
557 if (nfss
->acregmax
!= 60*HZ
)
558 seq_printf(m
, ",acregmax=%d", nfss
->acregmax
/HZ
);
559 if (nfss
->acdirmin
!= 30*HZ
)
560 seq_printf(m
, ",acdirmin=%d", nfss
->acdirmin
/HZ
);
561 if (nfss
->acdirmax
!= 60*HZ
)
562 seq_printf(m
, ",acdirmax=%d", nfss
->acdirmax
/HZ
);
563 for (nfs_infop
= nfs_info
; nfs_infop
->flag
; nfs_infop
++) {
564 if (nfss
->flags
& nfs_infop
->flag
)
565 seq_puts(m
, nfs_infop
->str
);
567 seq_puts(m
, nfs_infop
->nostr
);
569 seq_puts(m
, ",addr=");
570 seq_escape(m
, nfss
->hostname
, " \t\n\\");
575 * Invalidate the local caches
578 nfs_zap_caches(struct inode
*inode
)
580 struct nfs_inode
*nfsi
= NFS_I(inode
);
581 int mode
= inode
->i_mode
;
583 NFS_ATTRTIMEO(inode
) = NFS_MINATTRTIMEO(inode
);
584 NFS_ATTRTIMEO_UPDATE(inode
) = jiffies
;
586 memset(NFS_COOKIEVERF(inode
), 0, sizeof(NFS_COOKIEVERF(inode
)));
587 if (S_ISREG(mode
) || S_ISDIR(mode
) || S_ISLNK(mode
))
588 nfsi
->flags
|= NFS_INO_INVALID_ATTR
|NFS_INO_INVALID_DATA
|NFS_INO_INVALID_ACCESS
|NFS_INO_INVALID_ACL
;
590 nfsi
->flags
|= NFS_INO_INVALID_ATTR
|NFS_INO_INVALID_ACCESS
|NFS_INO_INVALID_ACL
;
593 static void nfs_zap_acl_cache(struct inode
*inode
)
595 void (*clear_acl_cache
)(struct inode
*);
597 clear_acl_cache
= NFS_PROTO(inode
)->clear_acl_cache
;
598 if (clear_acl_cache
!= NULL
)
599 clear_acl_cache(inode
);
600 NFS_I(inode
)->flags
&= ~NFS_INO_INVALID_ACL
;
604 * Invalidate, but do not unhash, the inode
607 nfs_invalidate_inode(struct inode
*inode
)
609 umode_t save_mode
= inode
->i_mode
;
611 make_bad_inode(inode
);
612 inode
->i_mode
= save_mode
;
613 nfs_zap_caches(inode
);
616 struct nfs_find_desc
{
618 struct nfs_fattr
*fattr
;
622 * In NFSv3 we can have 64bit inode numbers. In order to support
623 * this, and re-exported directories (also seen in NFSv2)
624 * we are forced to allow 2 different inodes to have the same
628 nfs_find_actor(struct inode
*inode
, void *opaque
)
630 struct nfs_find_desc
*desc
= (struct nfs_find_desc
*)opaque
;
631 struct nfs_fh
*fh
= desc
->fh
;
632 struct nfs_fattr
*fattr
= desc
->fattr
;
634 if (NFS_FILEID(inode
) != fattr
->fileid
)
636 if (nfs_compare_fh(NFS_FH(inode
), fh
))
638 if (is_bad_inode(inode
) || NFS_STALE(inode
))
644 nfs_init_locked(struct inode
*inode
, void *opaque
)
646 struct nfs_find_desc
*desc
= (struct nfs_find_desc
*)opaque
;
647 struct nfs_fattr
*fattr
= desc
->fattr
;
649 NFS_FILEID(inode
) = fattr
->fileid
;
650 nfs_copy_fh(NFS_FH(inode
), desc
->fh
);
654 /* Don't use READDIRPLUS on directories that we believe are too large */
655 #define NFS_LIMIT_READDIRPLUS (8*PAGE_SIZE)
658 * This is our front-end to iget that looks up inodes by file handle
659 * instead of inode number.
662 nfs_fhget(struct super_block
*sb
, struct nfs_fh
*fh
, struct nfs_fattr
*fattr
)
664 struct nfs_find_desc desc
= {
668 struct inode
*inode
= NULL
;
671 if ((fattr
->valid
& NFS_ATTR_FATTR
) == 0)
675 printk("NFS: Buggy server - nlink == 0!\n");
679 hash
= nfs_fattr_to_ino_t(fattr
);
681 if (!(inode
= iget5_locked(sb
, hash
, nfs_find_actor
, nfs_init_locked
, &desc
)))
684 if (inode
->i_state
& I_NEW
) {
685 struct nfs_inode
*nfsi
= NFS_I(inode
);
687 /* We set i_ino for the few things that still rely on it,
691 /* We can't support update_atime(), since the server will reset it */
692 inode
->i_flags
|= S_NOATIME
|S_NOCMTIME
;
693 inode
->i_mode
= fattr
->mode
;
694 /* Why so? Because we want revalidate for devices/FIFOs, and
695 * that's precisely what we have in nfs_file_inode_operations.
697 inode
->i_op
= NFS_SB(sb
)->rpc_ops
->file_inode_ops
;
698 if (S_ISREG(inode
->i_mode
)) {
699 inode
->i_fop
= &nfs_file_operations
;
700 inode
->i_data
.a_ops
= &nfs_file_aops
;
701 inode
->i_data
.backing_dev_info
= &NFS_SB(sb
)->backing_dev_info
;
702 } else if (S_ISDIR(inode
->i_mode
)) {
703 inode
->i_op
= NFS_SB(sb
)->rpc_ops
->dir_inode_ops
;
704 inode
->i_fop
= &nfs_dir_operations
;
705 if (nfs_server_capable(inode
, NFS_CAP_READDIRPLUS
)
706 && fattr
->size
<= NFS_LIMIT_READDIRPLUS
)
707 NFS_FLAGS(inode
) |= NFS_INO_ADVISE_RDPLUS
;
708 } else if (S_ISLNK(inode
->i_mode
))
709 inode
->i_op
= &nfs_symlink_inode_operations
;
711 init_special_inode(inode
, inode
->i_mode
, fattr
->rdev
);
713 nfsi
->read_cache_jiffies
= fattr
->timestamp
;
714 inode
->i_atime
= fattr
->atime
;
715 inode
->i_mtime
= fattr
->mtime
;
716 inode
->i_ctime
= fattr
->ctime
;
717 if (fattr
->valid
& NFS_ATTR_FATTR_V4
)
718 nfsi
->change_attr
= fattr
->change_attr
;
719 inode
->i_size
= nfs_size_to_loff_t(fattr
->size
);
720 inode
->i_nlink
= fattr
->nlink
;
721 inode
->i_uid
= fattr
->uid
;
722 inode
->i_gid
= fattr
->gid
;
723 if (fattr
->valid
& (NFS_ATTR_FATTR_V3
| NFS_ATTR_FATTR_V4
)) {
725 * report the blocks in 512byte units
727 inode
->i_blocks
= nfs_calc_block_size(fattr
->du
.nfs3
.used
);
728 inode
->i_blksize
= inode
->i_sb
->s_blocksize
;
730 inode
->i_blocks
= fattr
->du
.nfs2
.blocks
;
731 inode
->i_blksize
= fattr
->du
.nfs2
.blocksize
;
733 nfsi
->attrtimeo
= NFS_MINATTRTIMEO(inode
);
734 nfsi
->attrtimeo_timestamp
= jiffies
;
735 memset(nfsi
->cookieverf
, 0, sizeof(nfsi
->cookieverf
));
736 nfsi
->cache_access
.cred
= NULL
;
738 unlock_new_inode(inode
);
740 nfs_refresh_inode(inode
, fattr
);
741 dprintk("NFS: nfs_fhget(%s/%Ld ct=%d)\n",
743 (long long)NFS_FILEID(inode
),
744 atomic_read(&inode
->i_count
));
750 printk("nfs_fhget: iget failed\n");
754 #define NFS_VALID_ATTRS (ATTR_MODE|ATTR_UID|ATTR_GID|ATTR_SIZE|ATTR_ATIME|ATTR_ATIME_SET|ATTR_MTIME|ATTR_MTIME_SET)
757 nfs_setattr(struct dentry
*dentry
, struct iattr
*attr
)
759 struct inode
*inode
= dentry
->d_inode
;
760 struct nfs_fattr fattr
;
763 if (attr
->ia_valid
& ATTR_SIZE
) {
764 if (!S_ISREG(inode
->i_mode
) || attr
->ia_size
== i_size_read(inode
))
765 attr
->ia_valid
&= ~ATTR_SIZE
;
768 /* Optimization: if the end result is no change, don't RPC */
769 attr
->ia_valid
&= NFS_VALID_ATTRS
;
770 if (attr
->ia_valid
== 0)
774 nfs_begin_data_update(inode
);
775 /* Write all dirty data if we're changing file permissions or size */
776 if ((attr
->ia_valid
& (ATTR_MODE
|ATTR_UID
|ATTR_GID
|ATTR_SIZE
)) != 0) {
777 if (filemap_fdatawrite(inode
->i_mapping
) == 0)
778 filemap_fdatawait(inode
->i_mapping
);
781 error
= NFS_PROTO(inode
)->setattr(dentry
, &fattr
, attr
);
783 nfs_refresh_inode(inode
, &fattr
);
784 if ((attr
->ia_valid
& ATTR_MODE
) != 0) {
786 mode
= inode
->i_mode
& ~S_IALLUGO
;
787 mode
|= attr
->ia_mode
& S_IALLUGO
;
788 inode
->i_mode
= mode
;
790 if ((attr
->ia_valid
& ATTR_UID
) != 0)
791 inode
->i_uid
= attr
->ia_uid
;
792 if ((attr
->ia_valid
& ATTR_GID
) != 0)
793 inode
->i_gid
= attr
->ia_gid
;
794 if ((attr
->ia_valid
& ATTR_SIZE
) != 0) {
795 inode
->i_size
= attr
->ia_size
;
796 vmtruncate(inode
, attr
->ia_size
);
799 if ((attr
->ia_valid
& (ATTR_MODE
|ATTR_UID
|ATTR_GID
)) != 0)
800 NFS_FLAGS(inode
) |= NFS_INO_INVALID_ACCESS
|NFS_INO_INVALID_ACL
;
801 nfs_end_data_update(inode
);
807 * Wait for the inode to get unlocked.
808 * (Used for NFS_INO_LOCKED and NFS_INO_REVALIDATING).
811 nfs_wait_on_inode(struct inode
*inode
, int flag
)
813 struct rpc_clnt
*clnt
= NFS_CLIENT(inode
);
814 struct nfs_inode
*nfsi
= NFS_I(inode
);
817 if (!(NFS_FLAGS(inode
) & flag
))
819 atomic_inc(&inode
->i_count
);
820 error
= nfs_wait_event(clnt
, nfsi
->nfs_i_wait
,
821 !(NFS_FLAGS(inode
) & flag
));
826 int nfs_getattr(struct vfsmount
*mnt
, struct dentry
*dentry
, struct kstat
*stat
)
828 struct inode
*inode
= dentry
->d_inode
;
829 struct nfs_inode
*nfsi
= NFS_I(inode
);
830 int need_atime
= nfsi
->flags
& NFS_INO_INVALID_ATIME
;
833 if (__IS_FLG(inode
, MS_NOATIME
))
835 else if (__IS_FLG(inode
, MS_NODIRATIME
) && S_ISDIR(inode
->i_mode
))
837 /* We may force a getattr if the user cares about atime */
839 err
= __nfs_revalidate_inode(NFS_SERVER(inode
), inode
);
841 err
= nfs_revalidate_inode(NFS_SERVER(inode
), inode
);
843 generic_fillattr(inode
, stat
);
847 struct nfs_open_context
*alloc_nfs_open_context(struct dentry
*dentry
, struct rpc_cred
*cred
)
849 struct nfs_open_context
*ctx
;
851 ctx
= (struct nfs_open_context
*)kmalloc(sizeof(*ctx
), GFP_KERNEL
);
853 atomic_set(&ctx
->count
, 1);
854 ctx
->dentry
= dget(dentry
);
855 ctx
->cred
= get_rpccred(cred
);
857 ctx
->lockowner
= current
->files
;
863 struct nfs_open_context
*get_nfs_open_context(struct nfs_open_context
*ctx
)
866 atomic_inc(&ctx
->count
);
870 void put_nfs_open_context(struct nfs_open_context
*ctx
)
872 if (atomic_dec_and_test(&ctx
->count
)) {
873 if (!list_empty(&ctx
->list
)) {
874 struct inode
*inode
= ctx
->dentry
->d_inode
;
875 spin_lock(&inode
->i_lock
);
876 list_del(&ctx
->list
);
877 spin_unlock(&inode
->i_lock
);
879 if (ctx
->state
!= NULL
)
880 nfs4_close_state(ctx
->state
, ctx
->mode
);
881 if (ctx
->cred
!= NULL
)
882 put_rpccred(ctx
->cred
);
889 * Ensure that mmap has a recent RPC credential for use when writing out
892 void nfs_file_set_open_context(struct file
*filp
, struct nfs_open_context
*ctx
)
894 struct inode
*inode
= filp
->f_dentry
->d_inode
;
895 struct nfs_inode
*nfsi
= NFS_I(inode
);
897 filp
->private_data
= get_nfs_open_context(ctx
);
898 spin_lock(&inode
->i_lock
);
899 list_add(&ctx
->list
, &nfsi
->open_files
);
900 spin_unlock(&inode
->i_lock
);
903 struct nfs_open_context
*nfs_find_open_context(struct inode
*inode
, int mode
)
905 struct nfs_inode
*nfsi
= NFS_I(inode
);
906 struct nfs_open_context
*pos
, *ctx
= NULL
;
908 spin_lock(&inode
->i_lock
);
909 list_for_each_entry(pos
, &nfsi
->open_files
, list
) {
910 if ((pos
->mode
& mode
) == mode
) {
911 ctx
= get_nfs_open_context(pos
);
915 spin_unlock(&inode
->i_lock
);
919 void nfs_file_clear_open_context(struct file
*filp
)
921 struct inode
*inode
= filp
->f_dentry
->d_inode
;
922 struct nfs_open_context
*ctx
= (struct nfs_open_context
*)filp
->private_data
;
925 filp
->private_data
= NULL
;
926 spin_lock(&inode
->i_lock
);
927 list_move_tail(&ctx
->list
, &NFS_I(inode
)->open_files
);
928 spin_unlock(&inode
->i_lock
);
929 put_nfs_open_context(ctx
);
934 * These allocate and release file read/write context information.
936 int nfs_open(struct inode
*inode
, struct file
*filp
)
938 struct nfs_open_context
*ctx
;
939 struct rpc_cred
*cred
;
941 cred
= rpcauth_lookupcred(NFS_CLIENT(inode
)->cl_auth
, 0);
943 return PTR_ERR(cred
);
944 ctx
= alloc_nfs_open_context(filp
->f_dentry
, cred
);
948 ctx
->mode
= filp
->f_mode
;
949 nfs_file_set_open_context(filp
, ctx
);
950 put_nfs_open_context(ctx
);
951 if ((filp
->f_mode
& FMODE_WRITE
) != 0)
952 nfs_begin_data_update(inode
);
956 int nfs_release(struct inode
*inode
, struct file
*filp
)
958 if ((filp
->f_mode
& FMODE_WRITE
) != 0)
959 nfs_end_data_update(inode
);
960 nfs_file_clear_open_context(filp
);
965 * This function is called whenever some part of NFS notices that
966 * the cached attributes have to be refreshed.
969 __nfs_revalidate_inode(struct nfs_server
*server
, struct inode
*inode
)
971 int status
= -ESTALE
;
972 struct nfs_fattr fattr
;
973 struct nfs_inode
*nfsi
= NFS_I(inode
);
974 unsigned long verifier
;
977 dfprintk(PAGECACHE
, "NFS: revalidating (%s/%Ld)\n",
978 inode
->i_sb
->s_id
, (long long)NFS_FILEID(inode
));
981 if (!inode
|| is_bad_inode(inode
))
983 if (NFS_STALE(inode
))
986 while (NFS_REVALIDATING(inode
)) {
987 status
= nfs_wait_on_inode(inode
, NFS_INO_REVALIDATING
);
990 if (NFS_ATTRTIMEO(inode
) == 0)
992 if (NFS_FLAGS(inode
) & (NFS_INO_INVALID_ATTR
|NFS_INO_INVALID_DATA
|NFS_INO_INVALID_ATIME
))
994 status
= NFS_STALE(inode
) ? -ESTALE
: 0;
997 NFS_FLAGS(inode
) |= NFS_INO_REVALIDATING
;
999 /* Protect against RPC races by saving the change attribute */
1000 verifier
= nfs_save_change_attribute(inode
);
1001 status
= NFS_PROTO(inode
)->getattr(server
, NFS_FH(inode
), &fattr
);
1003 dfprintk(PAGECACHE
, "nfs_revalidate_inode: (%s/%Ld) getattr failed, error=%d\n",
1005 (long long)NFS_FILEID(inode
), status
);
1006 if (status
== -ESTALE
) {
1007 nfs_zap_caches(inode
);
1008 if (!S_ISDIR(inode
->i_mode
))
1009 NFS_FLAGS(inode
) |= NFS_INO_STALE
;
1014 status
= nfs_update_inode(inode
, &fattr
, verifier
);
1016 dfprintk(PAGECACHE
, "nfs_revalidate_inode: (%s/%Ld) refresh failed, error=%d\n",
1018 (long long)NFS_FILEID(inode
), status
);
1021 flags
= nfsi
->flags
;
1023 * We may need to keep the attributes marked as invalid if
1024 * we raced with nfs_end_attr_update().
1026 if (verifier
== nfsi
->cache_change_attribute
)
1027 nfsi
->flags
&= ~(NFS_INO_INVALID_ATTR
|NFS_INO_INVALID_ATIME
);
1028 /* Do the page cache invalidation */
1029 if (flags
& NFS_INO_INVALID_DATA
) {
1030 if (S_ISREG(inode
->i_mode
)) {
1031 if (filemap_fdatawrite(inode
->i_mapping
) == 0)
1032 filemap_fdatawait(inode
->i_mapping
);
1035 nfsi
->flags
&= ~NFS_INO_INVALID_DATA
;
1036 invalidate_inode_pages2(inode
->i_mapping
);
1037 memset(NFS_COOKIEVERF(inode
), 0, sizeof(NFS_COOKIEVERF(inode
)));
1038 dfprintk(PAGECACHE
, "NFS: (%s/%Ld) data cache invalidated\n",
1040 (long long)NFS_FILEID(inode
));
1041 /* This ensures we revalidate dentries */
1042 nfsi
->cache_change_attribute
++;
1044 if (flags
& NFS_INO_INVALID_ACL
)
1045 nfs_zap_acl_cache(inode
);
1046 dfprintk(PAGECACHE
, "NFS: (%s/%Ld) revalidation complete\n",
1048 (long long)NFS_FILEID(inode
));
1051 NFS_FLAGS(inode
) &= ~NFS_INO_REVALIDATING
;
1052 wake_up(&nfsi
->nfs_i_wait
);
1058 int nfs_attribute_timeout(struct inode
*inode
)
1060 struct nfs_inode
*nfsi
= NFS_I(inode
);
1062 if (nfs_have_delegation(inode
, FMODE_READ
))
1064 return time_after(jiffies
, nfsi
->read_cache_jiffies
+nfsi
->attrtimeo
);
1068 * nfs_revalidate_inode - Revalidate the inode attributes
1069 * @server - pointer to nfs_server struct
1070 * @inode - pointer to inode struct
1072 * Updates inode attribute information by retrieving the data from the server.
1074 int nfs_revalidate_inode(struct nfs_server
*server
, struct inode
*inode
)
1076 if (!(NFS_FLAGS(inode
) & (NFS_INO_INVALID_ATTR
|NFS_INO_INVALID_DATA
))
1077 && !nfs_attribute_timeout(inode
))
1078 return NFS_STALE(inode
) ? -ESTALE
: 0;
1079 return __nfs_revalidate_inode(server
, inode
);
1083 * nfs_begin_data_update
1084 * @inode - pointer to inode
1085 * Declare that a set of operations will update file data on the server
1087 void nfs_begin_data_update(struct inode
*inode
)
1089 atomic_inc(&NFS_I(inode
)->data_updates
);
1093 * nfs_end_data_update
1094 * @inode - pointer to inode
1095 * Declare end of the operations that will update file data
1096 * This will mark the inode as immediately needing revalidation
1097 * of its attribute cache.
1099 void nfs_end_data_update(struct inode
*inode
)
1101 struct nfs_inode
*nfsi
= NFS_I(inode
);
1103 if (!nfs_have_delegation(inode
, FMODE_READ
)) {
1104 /* Mark the attribute cache for revalidation */
1105 nfsi
->flags
|= NFS_INO_INVALID_ATTR
;
1106 /* Directories and symlinks: invalidate page cache too */
1107 if (S_ISDIR(inode
->i_mode
) || S_ISLNK(inode
->i_mode
))
1108 nfsi
->flags
|= NFS_INO_INVALID_DATA
;
1110 nfsi
->cache_change_attribute
++;
1111 atomic_dec(&nfsi
->data_updates
);
1115 * nfs_end_data_update_defer
1116 * @inode - pointer to inode
1117 * Declare end of the operations that will update file data
1118 * This will defer marking the inode as needing revalidation
1119 * unless there are no other pending updates.
1121 void nfs_end_data_update_defer(struct inode
*inode
)
1123 struct nfs_inode
*nfsi
= NFS_I(inode
);
1125 if (atomic_dec_and_test(&nfsi
->data_updates
)) {
1126 /* Mark the attribute cache for revalidation */
1127 nfsi
->flags
|= NFS_INO_INVALID_ATTR
;
1128 /* Directories and symlinks: invalidate page cache too */
1129 if (S_ISDIR(inode
->i_mode
) || S_ISLNK(inode
->i_mode
))
1130 nfsi
->flags
|= NFS_INO_INVALID_DATA
;
1131 nfsi
->cache_change_attribute
++;
1136 * nfs_refresh_inode - verify consistency of the inode attribute cache
1137 * @inode - pointer to inode
1138 * @fattr - updated attributes
1140 * Verifies the attribute cache. If we have just changed the attributes,
1141 * so that fattr carries weak cache consistency data, then it may
1142 * also update the ctime/mtime/change_attribute.
1144 int nfs_refresh_inode(struct inode
*inode
, struct nfs_fattr
*fattr
)
1146 struct nfs_inode
*nfsi
= NFS_I(inode
);
1147 loff_t cur_size
, new_isize
;
1150 /* Do we hold a delegation? */
1151 if (nfs_have_delegation(inode
, FMODE_READ
))
1154 /* Are we in the process of updating data on the server? */
1155 data_unstable
= nfs_caches_unstable(inode
);
1157 if (fattr
->valid
& NFS_ATTR_FATTR_V4
) {
1158 if ((fattr
->valid
& NFS_ATTR_PRE_CHANGE
) != 0
1159 && nfsi
->change_attr
== fattr
->pre_change_attr
)
1160 nfsi
->change_attr
= fattr
->change_attr
;
1161 if (!data_unstable
&& nfsi
->change_attr
!= fattr
->change_attr
)
1162 nfsi
->flags
|= NFS_INO_INVALID_ATTR
;
1165 if ((fattr
->valid
& NFS_ATTR_FATTR
) == 0)
1168 /* Has the inode gone and changed behind our back? */
1169 if (nfsi
->fileid
!= fattr
->fileid
1170 || (inode
->i_mode
& S_IFMT
) != (fattr
->mode
& S_IFMT
))
1173 cur_size
= i_size_read(inode
);
1174 new_isize
= nfs_size_to_loff_t(fattr
->size
);
1176 /* If we have atomic WCC data, we may update some attributes */
1177 if ((fattr
->valid
& NFS_ATTR_WCC
) != 0) {
1178 if (timespec_equal(&inode
->i_ctime
, &fattr
->pre_ctime
))
1179 memcpy(&inode
->i_ctime
, &fattr
->ctime
, sizeof(inode
->i_ctime
));
1180 if (timespec_equal(&inode
->i_mtime
, &fattr
->pre_mtime
))
1181 memcpy(&inode
->i_mtime
, &fattr
->mtime
, sizeof(inode
->i_mtime
));
1184 /* Verify a few of the more important attributes */
1185 if (!data_unstable
) {
1186 if (!timespec_equal(&inode
->i_mtime
, &fattr
->mtime
)
1187 || cur_size
!= new_isize
)
1188 nfsi
->flags
|= NFS_INO_INVALID_ATTR
;
1189 } else if (S_ISREG(inode
->i_mode
) && new_isize
> cur_size
)
1190 nfsi
->flags
|= NFS_INO_INVALID_ATTR
;
1192 /* Have any file permissions changed? */
1193 if ((inode
->i_mode
& S_IALLUGO
) != (fattr
->mode
& S_IALLUGO
)
1194 || inode
->i_uid
!= fattr
->uid
1195 || inode
->i_gid
!= fattr
->gid
)
1196 nfsi
->flags
|= NFS_INO_INVALID_ATTR
| NFS_INO_INVALID_ACCESS
| NFS_INO_INVALID_ACL
;
1198 /* Has the link count changed? */
1199 if (inode
->i_nlink
!= fattr
->nlink
)
1200 nfsi
->flags
|= NFS_INO_INVALID_ATTR
;
1202 if (!timespec_equal(&inode
->i_atime
, &fattr
->atime
))
1203 nfsi
->flags
|= NFS_INO_INVALID_ATIME
;
1205 nfsi
->read_cache_jiffies
= fattr
->timestamp
;
1210 * Many nfs protocol calls return the new file attributes after
1211 * an operation. Here we update the inode to reflect the state
1212 * of the server's inode.
1214 * This is a bit tricky because we have to make sure all dirty pages
1215 * have been sent off to the server before calling invalidate_inode_pages.
1216 * To make sure no other process adds more write requests while we try
1217 * our best to flush them, we make them sleep during the attribute refresh.
1219 * A very similar scenario holds for the dir cache.
1221 static int nfs_update_inode(struct inode
*inode
, struct nfs_fattr
*fattr
, unsigned long verifier
)
1223 struct nfs_inode
*nfsi
= NFS_I(inode
);
1226 unsigned int invalid
= 0;
1230 dfprintk(VFS
, "NFS: %s(%s/%ld ct=%d info=0x%x)\n",
1231 __FUNCTION__
, inode
->i_sb
->s_id
, inode
->i_ino
,
1232 atomic_read(&inode
->i_count
), fattr
->valid
);
1234 if ((fattr
->valid
& NFS_ATTR_FATTR
) == 0)
1237 if (nfsi
->fileid
!= fattr
->fileid
) {
1238 printk(KERN_ERR
"%s: inode number mismatch\n"
1239 "expected (%s/0x%Lx), got (%s/0x%Lx)\n",
1241 inode
->i_sb
->s_id
, (long long)nfsi
->fileid
,
1242 inode
->i_sb
->s_id
, (long long)fattr
->fileid
);
1247 * Make sure the inode's type hasn't changed.
1249 if ((inode
->i_mode
& S_IFMT
) != (fattr
->mode
& S_IFMT
))
1253 * Update the read time so we don't revalidate too often.
1255 nfsi
->read_cache_jiffies
= fattr
->timestamp
;
1257 /* Are we racing with known updates of the metadata on the server? */
1258 data_unstable
= ! nfs_verify_change_attribute(inode
, verifier
);
1260 /* Check if the file size agrees */
1261 new_size
= fattr
->size
;
1262 new_isize
= nfs_size_to_loff_t(fattr
->size
);
1263 cur_isize
= i_size_read(inode
);
1264 if (cur_isize
!= new_size
) {
1265 #ifdef NFS_DEBUG_VERBOSE
1266 printk(KERN_DEBUG
"NFS: isize change on %s/%ld\n", inode
->i_sb
->s_id
, inode
->i_ino
);
1269 * If we have pending writebacks, things can get
1272 if (S_ISREG(inode
->i_mode
) && data_unstable
) {
1273 if (new_isize
> cur_isize
) {
1274 inode
->i_size
= new_isize
;
1275 invalid
|= NFS_INO_INVALID_ATTR
|NFS_INO_INVALID_DATA
;
1278 inode
->i_size
= new_isize
;
1279 invalid
|= NFS_INO_INVALID_ATTR
|NFS_INO_INVALID_DATA
;
1284 * Note: we don't check inode->i_mtime since pipes etc.
1285 * can change this value in VFS without requiring a
1286 * cache revalidation.
1288 if (!timespec_equal(&inode
->i_mtime
, &fattr
->mtime
)) {
1289 memcpy(&inode
->i_mtime
, &fattr
->mtime
, sizeof(inode
->i_mtime
));
1290 #ifdef NFS_DEBUG_VERBOSE
1291 printk(KERN_DEBUG
"NFS: mtime change on %s/%ld\n", inode
->i_sb
->s_id
, inode
->i_ino
);
1294 invalid
|= NFS_INO_INVALID_ATTR
|NFS_INO_INVALID_DATA
;
1297 if ((fattr
->valid
& NFS_ATTR_FATTR_V4
)
1298 && nfsi
->change_attr
!= fattr
->change_attr
) {
1299 #ifdef NFS_DEBUG_VERBOSE
1300 printk(KERN_DEBUG
"NFS: change_attr change on %s/%ld\n",
1301 inode
->i_sb
->s_id
, inode
->i_ino
);
1303 nfsi
->change_attr
= fattr
->change_attr
;
1305 invalid
|= NFS_INO_INVALID_ATTR
|NFS_INO_INVALID_DATA
|NFS_INO_INVALID_ACCESS
|NFS_INO_INVALID_ACL
;
1308 /* If ctime has changed we should definitely clear access+acl caches */
1309 if (!timespec_equal(&inode
->i_ctime
, &fattr
->ctime
)) {
1311 invalid
|= NFS_INO_INVALID_ACCESS
|NFS_INO_INVALID_ACL
;
1312 memcpy(&inode
->i_ctime
, &fattr
->ctime
, sizeof(inode
->i_ctime
));
1314 memcpy(&inode
->i_atime
, &fattr
->atime
, sizeof(inode
->i_atime
));
1316 if ((inode
->i_mode
& S_IALLUGO
) != (fattr
->mode
& S_IALLUGO
) ||
1317 inode
->i_uid
!= fattr
->uid
||
1318 inode
->i_gid
!= fattr
->gid
)
1319 invalid
|= NFS_INO_INVALID_ATTR
|NFS_INO_INVALID_ACCESS
|NFS_INO_INVALID_ACL
;
1321 inode
->i_mode
= fattr
->mode
;
1322 inode
->i_nlink
= fattr
->nlink
;
1323 inode
->i_uid
= fattr
->uid
;
1324 inode
->i_gid
= fattr
->gid
;
1326 if (fattr
->valid
& (NFS_ATTR_FATTR_V3
| NFS_ATTR_FATTR_V4
)) {
1328 * report the blocks in 512byte units
1330 inode
->i_blocks
= nfs_calc_block_size(fattr
->du
.nfs3
.used
);
1331 inode
->i_blksize
= inode
->i_sb
->s_blocksize
;
1333 inode
->i_blocks
= fattr
->du
.nfs2
.blocks
;
1334 inode
->i_blksize
= fattr
->du
.nfs2
.blocksize
;
1337 /* Update attrtimeo value if we're out of the unstable period */
1338 if (invalid
& NFS_INO_INVALID_ATTR
) {
1339 nfsi
->attrtimeo
= NFS_MINATTRTIMEO(inode
);
1340 nfsi
->attrtimeo_timestamp
= jiffies
;
1341 } else if (time_after(jiffies
, nfsi
->attrtimeo_timestamp
+nfsi
->attrtimeo
)) {
1342 if ((nfsi
->attrtimeo
<<= 1) > NFS_MAXATTRTIMEO(inode
))
1343 nfsi
->attrtimeo
= NFS_MAXATTRTIMEO(inode
);
1344 nfsi
->attrtimeo_timestamp
= jiffies
;
1346 /* Don't invalidate the data if we were to blame */
1347 if (!(S_ISREG(inode
->i_mode
) || S_ISDIR(inode
->i_mode
)
1348 || S_ISLNK(inode
->i_mode
)))
1349 invalid
&= ~NFS_INO_INVALID_DATA
;
1350 if (!nfs_have_delegation(inode
, FMODE_READ
))
1351 nfsi
->flags
|= invalid
;
1356 * Big trouble! The inode has become a different object.
1359 printk(KERN_DEBUG
"%s: inode %ld mode changed, %07o to %07o\n",
1360 __FUNCTION__
, inode
->i_ino
, inode
->i_mode
, fattr
->mode
);
1363 * No need to worry about unhashing the dentry, as the
1364 * lookup validation will know that the inode is bad.
1365 * (But we fall through to invalidate the caches.)
1367 nfs_invalidate_inode(inode
);
1369 NFS_FLAGS(inode
) |= NFS_INO_STALE
;
1374 * File system information
1377 static int nfs_set_super(struct super_block
*s
, void *data
)
1379 s
->s_fs_info
= data
;
1380 return set_anon_super(s
, data
);
1383 static int nfs_compare_super(struct super_block
*sb
, void *data
)
1385 struct nfs_server
*server
= data
;
1386 struct nfs_server
*old
= NFS_SB(sb
);
1388 if (old
->addr
.sin_addr
.s_addr
!= server
->addr
.sin_addr
.s_addr
)
1390 if (old
->addr
.sin_port
!= server
->addr
.sin_port
)
1392 return !nfs_compare_fh(&old
->fh
, &server
->fh
);
1395 static struct super_block
*nfs_get_sb(struct file_system_type
*fs_type
,
1396 int flags
, const char *dev_name
, void *raw_data
)
1399 struct nfs_server
*server
= NULL
;
1400 struct super_block
*s
;
1401 struct nfs_fh
*root
;
1402 struct nfs_mount_data
*data
= raw_data
;
1404 s
= ERR_PTR(-EINVAL
);
1406 dprintk("%s: missing data argument\n", __FUNCTION__
);
1409 if (data
->version
<= 0 || data
->version
> NFS_MOUNT_VERSION
) {
1410 dprintk("%s: bad mount version\n", __FUNCTION__
);
1413 switch (data
->version
) {
1419 if (data
->flags
& NFS_MOUNT_VER3
) {
1420 dprintk("%s: mount structure version %d does not support NFSv3\n",
1425 data
->root
.size
= NFS2_FHSIZE
;
1426 memcpy(data
->root
.data
, data
->old_root
.data
, NFS2_FHSIZE
);
1428 if (data
->flags
& NFS_MOUNT_SECFLAVOUR
) {
1429 dprintk("%s: mount structure version %d does not support strong security\n",
1435 memset(data
->context
, 0, sizeof(data
->context
));
1437 #ifndef CONFIG_NFS_V3
1438 /* If NFSv3 is not compiled in, return -EPROTONOSUPPORT */
1439 s
= ERR_PTR(-EPROTONOSUPPORT
);
1440 if (data
->flags
& NFS_MOUNT_VER3
) {
1441 dprintk("%s: NFSv3 not compiled into kernel\n", __FUNCTION__
);
1444 #endif /* CONFIG_NFS_V3 */
1446 s
= ERR_PTR(-ENOMEM
);
1447 server
= kmalloc(sizeof(struct nfs_server
), GFP_KERNEL
);
1450 memset(server
, 0, sizeof(struct nfs_server
));
1451 /* Zero out the NFS state stuff */
1452 init_nfsv4_state(server
);
1455 if (data
->flags
& NFS_MOUNT_VER3
)
1456 root
->size
= data
->root
.size
;
1458 root
->size
= NFS2_FHSIZE
;
1459 s
= ERR_PTR(-EINVAL
);
1460 if (root
->size
> sizeof(root
->data
)) {
1461 dprintk("%s: invalid root filehandle\n", __FUNCTION__
);
1464 memcpy(root
->data
, data
->root
.data
, root
->size
);
1466 /* We now require that the mount process passes the remote address */
1467 memcpy(&server
->addr
, &data
->addr
, sizeof(server
->addr
));
1468 if (server
->addr
.sin_addr
.s_addr
== INADDR_ANY
) {
1469 dprintk("%s: mount program didn't pass remote address!\n",
1474 /* Fire up rpciod if not yet running */
1475 s
= ERR_PTR(rpciod_up());
1477 dprintk("%s: couldn't start rpciod! Error = %ld\n",
1478 __FUNCTION__
, PTR_ERR(s
));
1482 s
= sget(fs_type
, nfs_compare_super
, nfs_set_super
, server
);
1483 if (IS_ERR(s
) || s
->s_root
)
1484 goto out_rpciod_down
;
1488 error
= nfs_fill_super(s
, data
, flags
& MS_VERBOSE
? 1 : 0);
1490 up_write(&s
->s_umount
);
1491 deactivate_super(s
);
1492 return ERR_PTR(error
);
1494 s
->s_flags
|= MS_ACTIVE
;
1503 static void nfs_kill_super(struct super_block
*s
)
1505 struct nfs_server
*server
= NFS_SB(s
);
1509 if (server
->client
!= NULL
&& !IS_ERR(server
->client
))
1510 rpc_shutdown_client(server
->client
);
1511 if (server
->client_sys
!= NULL
&& !IS_ERR(server
->client_sys
))
1512 rpc_shutdown_client(server
->client_sys
);
1514 if (!(server
->flags
& NFS_MOUNT_NONLM
))
1515 lockd_down(); /* release rpc.lockd */
1517 rpciod_down(); /* release rpciod */
1519 if (server
->hostname
!= NULL
)
1520 kfree(server
->hostname
);
1524 static struct file_system_type nfs_fs_type
= {
1525 .owner
= THIS_MODULE
,
1527 .get_sb
= nfs_get_sb
,
1528 .kill_sb
= nfs_kill_super
,
1529 .fs_flags
= FS_ODD_RENAME
|FS_REVAL_DOT
|FS_BINARY_MOUNTDATA
,
1532 #ifdef CONFIG_NFS_V4
1534 static void nfs4_clear_inode(struct inode
*);
1537 static struct super_operations nfs4_sops
= {
1538 .alloc_inode
= nfs_alloc_inode
,
1539 .destroy_inode
= nfs_destroy_inode
,
1540 .write_inode
= nfs_write_inode
,
1541 .delete_inode
= nfs_delete_inode
,
1542 .statfs
= nfs_statfs
,
1543 .clear_inode
= nfs4_clear_inode
,
1544 .umount_begin
= nfs_umount_begin
,
1545 .show_options
= nfs_show_options
,
1549 * Clean out any remaining NFSv4 state that might be left over due
1550 * to open() calls that passed nfs_atomic_lookup, but failed to call
1553 static void nfs4_clear_inode(struct inode
*inode
)
1555 struct nfs_inode
*nfsi
= NFS_I(inode
);
1557 /* If we are holding a delegation, return it! */
1558 if (nfsi
->delegation
!= NULL
)
1559 nfs_inode_return_delegation(inode
);
1560 /* First call standard NFS clear_inode() code */
1561 nfs_clear_inode(inode
);
1562 /* Now clear out any remaining state */
1563 while (!list_empty(&nfsi
->open_states
)) {
1564 struct nfs4_state
*state
;
1566 state
= list_entry(nfsi
->open_states
.next
,
1569 dprintk("%s(%s/%Ld): found unclaimed NFSv4 state %p\n",
1572 (long long)NFS_FILEID(inode
),
1574 BUG_ON(atomic_read(&state
->count
) != 1);
1575 nfs4_close_state(state
, state
->state
);
1580 static int nfs4_fill_super(struct super_block
*sb
, struct nfs4_mount_data
*data
, int silent
)
1582 struct nfs_server
*server
;
1583 struct nfs4_client
*clp
= NULL
;
1584 struct rpc_xprt
*xprt
= NULL
;
1585 struct rpc_clnt
*clnt
= NULL
;
1586 struct rpc_timeout timeparms
;
1587 rpc_authflavor_t authflavour
;
1588 int proto
, err
= -EIO
;
1590 sb
->s_blocksize_bits
= 0;
1591 sb
->s_blocksize
= 0;
1592 server
= NFS_SB(sb
);
1593 if (data
->rsize
!= 0)
1594 server
->rsize
= nfs_block_size(data
->rsize
, NULL
);
1595 if (data
->wsize
!= 0)
1596 server
->wsize
= nfs_block_size(data
->wsize
, NULL
);
1597 server
->flags
= data
->flags
& NFS_MOUNT_FLAGMASK
;
1598 server
->caps
= NFS_CAP_ATOMIC_OPEN
;
1600 server
->acregmin
= data
->acregmin
*HZ
;
1601 server
->acregmax
= data
->acregmax
*HZ
;
1602 server
->acdirmin
= data
->acdirmin
*HZ
;
1603 server
->acdirmax
= data
->acdirmax
*HZ
;
1605 server
->rpc_ops
= &nfs_v4_clientops
;
1606 /* Initialize timeout values */
1608 timeparms
.to_initval
= data
->timeo
* HZ
/ 10;
1609 timeparms
.to_retries
= data
->retrans
;
1610 timeparms
.to_exponential
= 1;
1611 if (!timeparms
.to_retries
)
1612 timeparms
.to_retries
= 5;
1614 proto
= data
->proto
;
1615 /* Which IP protocol do we use? */
1618 timeparms
.to_maxval
= RPC_MAX_TCP_TIMEOUT
;
1619 if (!timeparms
.to_initval
)
1620 timeparms
.to_initval
= 600 * HZ
/ 10;
1623 timeparms
.to_maxval
= RPC_MAX_UDP_TIMEOUT
;
1624 if (!timeparms
.to_initval
)
1625 timeparms
.to_initval
= 11 * HZ
/ 10;
1631 clp
= nfs4_get_client(&server
->addr
.sin_addr
);
1633 dprintk("%s: failed to create NFS4 client.\n", __FUNCTION__
);
1637 /* Now create transport and client */
1638 authflavour
= RPC_AUTH_UNIX
;
1639 if (data
->auth_flavourlen
!= 0) {
1640 if (data
->auth_flavourlen
!= 1) {
1641 dprintk("%s: Invalid number of RPC auth flavours %d.\n",
1642 __FUNCTION__
, data
->auth_flavourlen
);
1646 if (copy_from_user(&authflavour
, data
->auth_flavours
, sizeof(authflavour
))) {
1652 down_write(&clp
->cl_sem
);
1653 if (clp
->cl_rpcclient
== NULL
) {
1654 xprt
= xprt_create_proto(proto
, &server
->addr
, &timeparms
);
1656 up_write(&clp
->cl_sem
);
1657 err
= PTR_ERR(xprt
);
1658 dprintk("%s: cannot create RPC transport. Error = %d\n",
1662 clnt
= rpc_create_client(xprt
, server
->hostname
, &nfs_program
,
1663 server
->rpc_ops
->version
, authflavour
);
1665 up_write(&clp
->cl_sem
);
1666 err
= PTR_ERR(clnt
);
1667 dprintk("%s: cannot create RPC client. Error = %d\n",
1672 clnt
->cl_softrtry
= 1;
1673 clnt
->cl_chatty
= 1;
1674 clp
->cl_rpcclient
= clnt
;
1675 clp
->cl_cred
= rpcauth_lookupcred(clnt
->cl_auth
, 0);
1676 if (IS_ERR(clp
->cl_cred
)) {
1677 up_write(&clp
->cl_sem
);
1678 err
= PTR_ERR(clp
->cl_cred
);
1679 clp
->cl_cred
= NULL
;
1682 memcpy(clp
->cl_ipaddr
, server
->ip_addr
, sizeof(clp
->cl_ipaddr
));
1685 if (list_empty(&clp
->cl_superblocks
)) {
1686 err
= nfs4_init_client(clp
);
1688 up_write(&clp
->cl_sem
);
1692 list_add_tail(&server
->nfs4_siblings
, &clp
->cl_superblocks
);
1693 clnt
= rpc_clone_client(clp
->cl_rpcclient
);
1695 server
->nfs4_state
= clp
;
1696 up_write(&clp
->cl_sem
);
1700 err
= PTR_ERR(clnt
);
1701 dprintk("%s: cannot create RPC client. Error = %d\n",
1706 server
->client
= clnt
;
1708 if (server
->nfs4_state
->cl_idmap
== NULL
) {
1709 dprintk("%s: failed to create idmapper.\n", __FUNCTION__
);
1713 if (clnt
->cl_auth
->au_flavor
!= authflavour
) {
1714 if (rpcauth_create(authflavour
, clnt
) == NULL
) {
1715 dprintk("%s: couldn't create credcache!\n", __FUNCTION__
);
1720 sb
->s_time_gran
= 1;
1722 sb
->s_op
= &nfs4_sops
;
1723 err
= nfs_sb_init(sb
, authflavour
);
1728 nfs4_put_client(clp
);
1732 static int nfs4_compare_super(struct super_block
*sb
, void *data
)
1734 struct nfs_server
*server
= data
;
1735 struct nfs_server
*old
= NFS_SB(sb
);
1737 if (strcmp(server
->hostname
, old
->hostname
) != 0)
1739 if (strcmp(server
->mnt_path
, old
->mnt_path
) != 0)
1745 nfs_copy_user_string(char *dst
, struct nfs_string
*src
, int maxlen
)
1750 return ERR_PTR(-EINVAL
);
1751 if (src
->len
< maxlen
)
1754 p
= dst
= kmalloc(maxlen
+ 1, GFP_KERNEL
);
1756 return ERR_PTR(-ENOMEM
);
1758 if (copy_from_user(dst
, src
->data
, maxlen
)) {
1761 return ERR_PTR(-EFAULT
);
1767 static struct super_block
*nfs4_get_sb(struct file_system_type
*fs_type
,
1768 int flags
, const char *dev_name
, void *raw_data
)
1771 struct nfs_server
*server
;
1772 struct super_block
*s
;
1773 struct nfs4_mount_data
*data
= raw_data
;
1777 dprintk("%s: missing data argument\n", __FUNCTION__
);
1778 return ERR_PTR(-EINVAL
);
1780 if (data
->version
<= 0 || data
->version
> NFS4_MOUNT_VERSION
) {
1781 dprintk("%s: bad mount version\n", __FUNCTION__
);
1782 return ERR_PTR(-EINVAL
);
1785 server
= kmalloc(sizeof(struct nfs_server
), GFP_KERNEL
);
1787 return ERR_PTR(-ENOMEM
);
1788 memset(server
, 0, sizeof(struct nfs_server
));
1789 /* Zero out the NFS state stuff */
1790 init_nfsv4_state(server
);
1792 p
= nfs_copy_user_string(NULL
, &data
->hostname
, 256);
1795 server
->hostname
= p
;
1797 p
= nfs_copy_user_string(NULL
, &data
->mnt_path
, 1024);
1800 server
->mnt_path
= p
;
1802 p
= nfs_copy_user_string(server
->ip_addr
, &data
->client_addr
,
1803 sizeof(server
->ip_addr
) - 1);
1807 /* We now require that the mount process passes the remote address */
1808 if (data
->host_addrlen
!= sizeof(server
->addr
)) {
1809 s
= ERR_PTR(-EINVAL
);
1812 if (copy_from_user(&server
->addr
, data
->host_addr
, sizeof(server
->addr
))) {
1813 s
= ERR_PTR(-EFAULT
);
1816 if (server
->addr
.sin_family
!= AF_INET
||
1817 server
->addr
.sin_addr
.s_addr
== INADDR_ANY
) {
1818 dprintk("%s: mount program didn't pass remote IP address!\n",
1820 s
= ERR_PTR(-EINVAL
);
1824 /* Fire up rpciod if not yet running */
1825 s
= ERR_PTR(rpciod_up());
1827 dprintk("%s: couldn't start rpciod! Error = %ld\n",
1828 __FUNCTION__
, PTR_ERR(s
));
1832 s
= sget(fs_type
, nfs4_compare_super
, nfs_set_super
, server
);
1834 if (IS_ERR(s
) || s
->s_root
)
1839 error
= nfs4_fill_super(s
, data
, flags
& MS_VERBOSE
? 1 : 0);
1841 up_write(&s
->s_umount
);
1842 deactivate_super(s
);
1843 return ERR_PTR(error
);
1845 s
->s_flags
|= MS_ACTIVE
;
1848 s
= (struct super_block
*)p
;
1850 if (server
->mnt_path
)
1851 kfree(server
->mnt_path
);
1852 if (server
->hostname
)
1853 kfree(server
->hostname
);
1858 static void nfs4_kill_super(struct super_block
*sb
)
1860 struct nfs_server
*server
= NFS_SB(sb
);
1862 nfs_return_all_delegations(sb
);
1863 kill_anon_super(sb
);
1865 nfs4_renewd_prepare_shutdown(server
);
1867 if (server
->client
!= NULL
&& !IS_ERR(server
->client
))
1868 rpc_shutdown_client(server
->client
);
1869 rpciod_down(); /* release rpciod */
1871 destroy_nfsv4_state(server
);
1873 if (server
->hostname
!= NULL
)
1874 kfree(server
->hostname
);
1878 static struct file_system_type nfs4_fs_type
= {
1879 .owner
= THIS_MODULE
,
1881 .get_sb
= nfs4_get_sb
,
1882 .kill_sb
= nfs4_kill_super
,
1883 .fs_flags
= FS_ODD_RENAME
|FS_REVAL_DOT
|FS_BINARY_MOUNTDATA
,
1886 #define nfs4_init_once(nfsi) \
1888 INIT_LIST_HEAD(&(nfsi)->open_states); \
1889 nfsi->delegation = NULL; \
1890 nfsi->delegation_state = 0; \
1891 init_rwsem(&nfsi->rwsem); \
1893 #define register_nfs4fs() register_filesystem(&nfs4_fs_type)
1894 #define unregister_nfs4fs() unregister_filesystem(&nfs4_fs_type)
1896 #define nfs4_init_once(nfsi) \
1898 #define register_nfs4fs() (0)
1899 #define unregister_nfs4fs()
1902 extern int nfs_init_nfspagecache(void);
1903 extern void nfs_destroy_nfspagecache(void);
1904 extern int nfs_init_readpagecache(void);
1905 extern void nfs_destroy_readpagecache(void);
1906 extern int nfs_init_writepagecache(void);
1907 extern void nfs_destroy_writepagecache(void);
1908 #ifdef CONFIG_NFS_DIRECTIO
1909 extern int nfs_init_directcache(void);
1910 extern void nfs_destroy_directcache(void);
1913 static kmem_cache_t
* nfs_inode_cachep
;
1915 static struct inode
*nfs_alloc_inode(struct super_block
*sb
)
1917 struct nfs_inode
*nfsi
;
1918 nfsi
= (struct nfs_inode
*)kmem_cache_alloc(nfs_inode_cachep
, SLAB_KERNEL
);
1922 #ifdef CONFIG_NFS_V4
1923 nfsi
->nfs4_acl
= NULL
;
1924 #endif /* CONFIG_NFS_V4 */
1925 return &nfsi
->vfs_inode
;
1928 static void nfs_destroy_inode(struct inode
*inode
)
1930 kmem_cache_free(nfs_inode_cachep
, NFS_I(inode
));
1933 static void init_once(void * foo
, kmem_cache_t
* cachep
, unsigned long flags
)
1935 struct nfs_inode
*nfsi
= (struct nfs_inode
*) foo
;
1937 if ((flags
& (SLAB_CTOR_VERIFY
|SLAB_CTOR_CONSTRUCTOR
)) ==
1938 SLAB_CTOR_CONSTRUCTOR
) {
1939 inode_init_once(&nfsi
->vfs_inode
);
1940 spin_lock_init(&nfsi
->req_lock
);
1941 INIT_LIST_HEAD(&nfsi
->dirty
);
1942 INIT_LIST_HEAD(&nfsi
->commit
);
1943 INIT_LIST_HEAD(&nfsi
->open_files
);
1944 INIT_RADIX_TREE(&nfsi
->nfs_page_tree
, GFP_ATOMIC
);
1945 atomic_set(&nfsi
->data_updates
, 0);
1949 init_waitqueue_head(&nfsi
->nfs_i_wait
);
1950 nfs4_init_once(nfsi
);
1954 static int nfs_init_inodecache(void)
1956 nfs_inode_cachep
= kmem_cache_create("nfs_inode_cache",
1957 sizeof(struct nfs_inode
),
1958 0, SLAB_RECLAIM_ACCOUNT
,
1960 if (nfs_inode_cachep
== NULL
)
1966 static void nfs_destroy_inodecache(void)
1968 if (kmem_cache_destroy(nfs_inode_cachep
))
1969 printk(KERN_INFO
"nfs_inode_cache: not all structures were freed\n");
1975 static int __init
init_nfs_fs(void)
1979 err
= nfs_init_nfspagecache();
1983 err
= nfs_init_inodecache();
1987 err
= nfs_init_readpagecache();
1991 err
= nfs_init_writepagecache();
1995 #ifdef CONFIG_NFS_DIRECTIO
1996 err
= nfs_init_directcache();
2001 #ifdef CONFIG_PROC_FS
2002 rpc_proc_register(&nfs_rpcstat
);
2004 err
= register_filesystem(&nfs_fs_type
);
2007 if ((err
= register_nfs4fs()) != 0)
2011 #ifdef CONFIG_PROC_FS
2012 rpc_proc_unregister("nfs");
2014 nfs_destroy_writepagecache();
2015 #ifdef CONFIG_NFS_DIRECTIO
2017 nfs_destroy_directcache();
2020 nfs_destroy_readpagecache();
2022 nfs_destroy_inodecache();
2024 nfs_destroy_nfspagecache();
2029 static void __exit
exit_nfs_fs(void)
2031 #ifdef CONFIG_NFS_DIRECTIO
2032 nfs_destroy_directcache();
2034 nfs_destroy_writepagecache();
2035 nfs_destroy_readpagecache();
2036 nfs_destroy_inodecache();
2037 nfs_destroy_nfspagecache();
2038 #ifdef CONFIG_PROC_FS
2039 rpc_proc_unregister("nfs");
2041 unregister_filesystem(&nfs_fs_type
);
2042 unregister_nfs4fs();
2045 /* Not quite true; I just maintain it */
2046 MODULE_AUTHOR("Olaf Kirch <okir@monad.swb.de>");
2047 MODULE_LICENSE("GPL");
2049 module_init(init_nfs_fs
)
2050 module_exit(exit_nfs_fs
)