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>
44 #include "delegation.h"
47 #define NFSDBG_FACILITY NFSDBG_VFS
48 #define NFS_PARANOIA 1
50 /* Maximum number of readahead requests
51 * FIXME: this should really be a sysctl so that users may tune it to suit
52 * their needs. People that do NFS over a slow network, might for
53 * instance want to reduce it to something closer to 1 for improved
54 * interactive response.
56 #define NFS_MAX_READAHEAD (RPC_DEF_SLOT_TABLE - 1)
58 static void nfs_invalidate_inode(struct inode
*);
59 static int nfs_update_inode(struct inode
*, struct nfs_fattr
*);
61 static struct inode
*nfs_alloc_inode(struct super_block
*sb
);
62 static void nfs_destroy_inode(struct inode
*);
63 static int nfs_write_inode(struct inode
*,int);
64 static void nfs_delete_inode(struct inode
*);
65 static void nfs_clear_inode(struct inode
*);
66 static void nfs_umount_begin(struct super_block
*);
67 static int nfs_statfs(struct super_block
*, struct kstatfs
*);
68 static int nfs_show_options(struct seq_file
*, struct vfsmount
*);
69 static int nfs_show_stats(struct seq_file
*, struct vfsmount
*);
70 static void nfs_zap_acl_cache(struct inode
*);
72 static struct rpc_program nfs_program
;
74 static struct super_operations nfs_sops
= {
75 .alloc_inode
= nfs_alloc_inode
,
76 .destroy_inode
= nfs_destroy_inode
,
77 .write_inode
= nfs_write_inode
,
78 .delete_inode
= nfs_delete_inode
,
80 .clear_inode
= nfs_clear_inode
,
81 .umount_begin
= nfs_umount_begin
,
82 .show_options
= nfs_show_options
,
83 .show_stats
= nfs_show_stats
,
89 static struct rpc_stat nfs_rpcstat
= {
90 .program
= &nfs_program
92 static struct rpc_version
* nfs_version
[] = {
96 #if defined(CONFIG_NFS_V3)
98 #elif defined(CONFIG_NFS_V4)
101 #if defined(CONFIG_NFS_V4)
106 static struct rpc_program nfs_program
= {
108 .number
= NFS_PROGRAM
,
109 .nrvers
= sizeof(nfs_version
) / sizeof(nfs_version
[0]),
110 .version
= nfs_version
,
111 .stats
= &nfs_rpcstat
,
112 .pipe_dir_name
= "/nfs",
115 #ifdef CONFIG_NFS_V3_ACL
116 static struct rpc_stat nfsacl_rpcstat
= { &nfsacl_program
};
117 static struct rpc_version
* nfsacl_version
[] = {
118 [3] = &nfsacl_version3
,
121 struct rpc_program nfsacl_program
= {
123 .number
= NFS_ACL_PROGRAM
,
124 .nrvers
= sizeof(nfsacl_version
) / sizeof(nfsacl_version
[0]),
125 .version
= nfsacl_version
,
126 .stats
= &nfsacl_rpcstat
,
128 #endif /* CONFIG_NFS_V3_ACL */
130 static inline unsigned long
131 nfs_fattr_to_ino_t(struct nfs_fattr
*fattr
)
133 return nfs_fileid_to_ino_t(fattr
->fileid
);
137 nfs_write_inode(struct inode
*inode
, int sync
)
139 int flags
= sync
? FLUSH_WAIT
: 0;
142 ret
= nfs_commit_inode(inode
, flags
);
149 nfs_delete_inode(struct inode
* inode
)
151 dprintk("NFS: delete_inode(%s/%ld)\n", inode
->i_sb
->s_id
, inode
->i_ino
);
153 truncate_inode_pages(&inode
->i_data
, 0);
157 * The following should never happen...
159 if (nfs_have_writebacks(inode
)) {
160 printk(KERN_ERR
"nfs_delete_inode: inode %ld has pending RPC requests\n", inode
->i_ino
);
167 nfs_clear_inode(struct inode
*inode
)
169 struct nfs_inode
*nfsi
= NFS_I(inode
);
170 struct rpc_cred
*cred
;
173 BUG_ON (!list_empty(&nfsi
->open_files
));
174 nfs_zap_acl_cache(inode
);
175 cred
= nfsi
->cache_access
.cred
;
178 BUG_ON(atomic_read(&nfsi
->data_updates
) != 0);
182 nfs_umount_begin(struct super_block
*sb
)
184 struct rpc_clnt
*rpc
= NFS_SB(sb
)->client
;
186 /* -EIO all pending I/O */
188 rpc_killall_tasks(rpc
);
189 rpc
= NFS_SB(sb
)->client_acl
;
191 rpc_killall_tasks(rpc
);
195 static inline unsigned long
196 nfs_block_bits(unsigned long bsize
, unsigned char *nrbitsp
)
198 /* make sure blocksize is a power of two */
199 if ((bsize
& (bsize
- 1)) || nrbitsp
) {
200 unsigned char nrbits
;
202 for (nrbits
= 31; nrbits
&& !(bsize
& (1 << nrbits
)); nrbits
--)
213 * Calculate the number of 512byte blocks used.
215 static inline unsigned long
216 nfs_calc_block_size(u64 tsize
)
218 loff_t used
= (tsize
+ 511) >> 9;
219 return (used
> ULONG_MAX
) ? ULONG_MAX
: used
;
223 * Compute and set NFS server blocksize
225 static inline unsigned long
226 nfs_block_size(unsigned long bsize
, unsigned char *nrbitsp
)
228 if (bsize
< NFS_MIN_FILE_IO_SIZE
)
229 bsize
= NFS_DEF_FILE_IO_SIZE
;
230 else if (bsize
>= NFS_MAX_FILE_IO_SIZE
)
231 bsize
= NFS_MAX_FILE_IO_SIZE
;
233 return nfs_block_bits(bsize
, nrbitsp
);
237 * Obtain the root inode of the file system.
239 static struct inode
*
240 nfs_get_root(struct super_block
*sb
, struct nfs_fh
*rootfh
, struct nfs_fsinfo
*fsinfo
)
242 struct nfs_server
*server
= NFS_SB(sb
);
246 error
= server
->rpc_ops
->getroot(server
, rootfh
, fsinfo
);
248 dprintk("nfs_get_root: getattr error = %d\n", -error
);
249 return ERR_PTR(error
);
252 rooti
= nfs_fhget(sb
, rootfh
, fsinfo
->fattr
);
254 return ERR_PTR(-ENOMEM
);
259 * Do NFS version-independent mount processing, and sanity checking
262 nfs_sb_init(struct super_block
*sb
, rpc_authflavor_t authflavor
)
264 struct nfs_server
*server
;
265 struct inode
*root_inode
;
266 struct nfs_fattr fattr
;
267 struct nfs_fsinfo fsinfo
= {
270 struct nfs_pathconf pathinfo
= {
273 int no_root_error
= 0;
274 unsigned long max_rpc_payload
;
276 /* We probably want something more informative here */
277 snprintf(sb
->s_id
, sizeof(sb
->s_id
), "%x:%x", MAJOR(sb
->s_dev
), MINOR(sb
->s_dev
));
281 sb
->s_magic
= NFS_SUPER_MAGIC
;
283 root_inode
= nfs_get_root(sb
, &server
->fh
, &fsinfo
);
284 /* Did getting the root inode fail? */
285 if (IS_ERR(root_inode
)) {
286 no_root_error
= PTR_ERR(root_inode
);
289 sb
->s_root
= d_alloc_root(root_inode
);
291 no_root_error
= -ENOMEM
;
294 sb
->s_root
->d_op
= server
->rpc_ops
->dentry_ops
;
296 server
->io_stats
= nfs_alloc_iostats();
297 if (!server
->io_stats
) {
298 no_root_error
= -ENOMEM
;
302 /* Get some general file system info */
303 if (server
->namelen
== 0 &&
304 server
->rpc_ops
->pathconf(server
, &server
->fh
, &pathinfo
) >= 0)
305 server
->namelen
= pathinfo
.max_namelen
;
306 /* Work out a lot of parameters */
307 if (server
->rsize
== 0)
308 server
->rsize
= nfs_block_size(fsinfo
.rtpref
, NULL
);
309 if (server
->wsize
== 0)
310 server
->wsize
= nfs_block_size(fsinfo
.wtpref
, NULL
);
312 if (fsinfo
.rtmax
>= 512 && server
->rsize
> fsinfo
.rtmax
)
313 server
->rsize
= nfs_block_size(fsinfo
.rtmax
, NULL
);
314 if (fsinfo
.wtmax
>= 512 && server
->wsize
> fsinfo
.wtmax
)
315 server
->wsize
= nfs_block_size(fsinfo
.wtmax
, NULL
);
317 max_rpc_payload
= nfs_block_size(rpc_max_payload(server
->client
), NULL
);
318 if (server
->rsize
> max_rpc_payload
)
319 server
->rsize
= max_rpc_payload
;
320 if (server
->rsize
> NFS_MAX_FILE_IO_SIZE
)
321 server
->rsize
= NFS_MAX_FILE_IO_SIZE
;
322 server
->rpages
= (server
->rsize
+ PAGE_CACHE_SIZE
- 1) >> PAGE_CACHE_SHIFT
;
324 if (server
->wsize
> max_rpc_payload
)
325 server
->wsize
= max_rpc_payload
;
326 if (server
->wsize
> NFS_MAX_FILE_IO_SIZE
)
327 server
->wsize
= NFS_MAX_FILE_IO_SIZE
;
328 server
->wpages
= (server
->wsize
+ PAGE_CACHE_SIZE
- 1) >> PAGE_CACHE_SHIFT
;
330 if (sb
->s_blocksize
== 0)
331 sb
->s_blocksize
= nfs_block_bits(server
->wsize
,
332 &sb
->s_blocksize_bits
);
333 server
->wtmult
= nfs_block_bits(fsinfo
.wtmult
, NULL
);
335 server
->dtsize
= nfs_block_size(fsinfo
.dtpref
, NULL
);
336 if (server
->dtsize
> PAGE_CACHE_SIZE
)
337 server
->dtsize
= PAGE_CACHE_SIZE
;
338 if (server
->dtsize
> server
->rsize
)
339 server
->dtsize
= server
->rsize
;
341 if (server
->flags
& NFS_MOUNT_NOAC
) {
342 server
->acregmin
= server
->acregmax
= 0;
343 server
->acdirmin
= server
->acdirmax
= 0;
344 sb
->s_flags
|= MS_SYNCHRONOUS
;
346 server
->backing_dev_info
.ra_pages
= server
->rpages
* NFS_MAX_READAHEAD
;
348 sb
->s_maxbytes
= fsinfo
.maxfilesize
;
349 if (sb
->s_maxbytes
> MAX_LFS_FILESIZE
)
350 sb
->s_maxbytes
= MAX_LFS_FILESIZE
;
352 server
->client
->cl_intr
= (server
->flags
& NFS_MOUNT_INTR
) ? 1 : 0;
353 server
->client
->cl_softrtry
= (server
->flags
& NFS_MOUNT_SOFT
) ? 1 : 0;
355 /* We're airborne Set socket buffersize */
356 rpc_setbufsize(server
->client
, server
->wsize
+ 100, server
->rsize
+ 100);
358 /* Yargs. It didn't work out. */
360 dprintk("nfs_sb_init: get root inode failed: errno %d\n", -no_root_error
);
361 if (!IS_ERR(root_inode
))
363 return no_root_error
;
366 static void nfs_init_timeout_values(struct rpc_timeout
*to
, int proto
, unsigned int timeo
, unsigned int retrans
)
368 to
->to_initval
= timeo
* HZ
/ 10;
369 to
->to_retries
= retrans
;
376 to
->to_initval
= 60 * HZ
;
377 if (to
->to_initval
> NFS_MAX_TCP_TIMEOUT
)
378 to
->to_initval
= NFS_MAX_TCP_TIMEOUT
;
379 to
->to_increment
= to
->to_initval
;
380 to
->to_maxval
= to
->to_initval
+ (to
->to_increment
* to
->to_retries
);
381 to
->to_exponential
= 0;
386 to
->to_initval
= 11 * HZ
/ 10;
387 if (to
->to_initval
> NFS_MAX_UDP_TIMEOUT
)
388 to
->to_initval
= NFS_MAX_UDP_TIMEOUT
;
389 to
->to_maxval
= NFS_MAX_UDP_TIMEOUT
;
390 to
->to_exponential
= 1;
396 * Create an RPC client handle.
398 static struct rpc_clnt
*
399 nfs_create_client(struct nfs_server
*server
, const struct nfs_mount_data
*data
)
401 struct rpc_timeout timeparms
;
402 struct rpc_xprt
*xprt
= NULL
;
403 struct rpc_clnt
*clnt
= NULL
;
404 int proto
= (data
->flags
& NFS_MOUNT_TCP
) ? IPPROTO_TCP
: IPPROTO_UDP
;
406 nfs_init_timeout_values(&timeparms
, proto
, data
->timeo
, data
->retrans
);
408 server
->retrans_timeo
= timeparms
.to_initval
;
409 server
->retrans_count
= timeparms
.to_retries
;
411 /* create transport and client */
412 xprt
= xprt_create_proto(proto
, &server
->addr
, &timeparms
);
414 dprintk("%s: cannot create RPC transport. Error = %ld\n",
415 __FUNCTION__
, PTR_ERR(xprt
));
416 return (struct rpc_clnt
*)xprt
;
418 clnt
= rpc_create_client(xprt
, server
->hostname
, &nfs_program
,
419 server
->rpc_ops
->version
, data
->pseudoflavor
);
421 dprintk("%s: cannot create RPC client. Error = %ld\n",
422 __FUNCTION__
, PTR_ERR(xprt
));
427 clnt
->cl_softrtry
= 1;
436 * The way this works is that the mount process passes a structure
437 * in the data argument which contains the server's IP address
438 * and the root file handle obtained from the server's mount
439 * daemon. We stash these away in the private superblock fields.
442 nfs_fill_super(struct super_block
*sb
, struct nfs_mount_data
*data
, int silent
)
444 struct nfs_server
*server
;
445 rpc_authflavor_t authflavor
;
448 sb
->s_blocksize_bits
= 0;
451 sb
->s_blocksize
= nfs_block_size(data
->bsize
, &sb
->s_blocksize_bits
);
453 server
->rsize
= nfs_block_size(data
->rsize
, NULL
);
455 server
->wsize
= nfs_block_size(data
->wsize
, NULL
);
456 server
->flags
= data
->flags
& NFS_MOUNT_FLAGMASK
;
458 server
->acregmin
= data
->acregmin
*HZ
;
459 server
->acregmax
= data
->acregmax
*HZ
;
460 server
->acdirmin
= data
->acdirmin
*HZ
;
461 server
->acdirmax
= data
->acdirmax
*HZ
;
463 /* Start lockd here, before we might error out */
464 if (!(server
->flags
& NFS_MOUNT_NONLM
))
467 server
->namelen
= data
->namlen
;
468 server
->hostname
= kmalloc(strlen(data
->hostname
) + 1, GFP_KERNEL
);
469 if (!server
->hostname
)
471 strcpy(server
->hostname
, data
->hostname
);
473 /* Check NFS protocol revision and initialize RPC op vector
474 * and file handle pool. */
476 if (server
->flags
& NFS_MOUNT_VER3
) {
477 server
->rpc_ops
= &nfs_v3_clientops
;
478 server
->caps
|= NFS_CAP_READDIRPLUS
;
480 server
->rpc_ops
= &nfs_v2_clientops
;
483 server
->rpc_ops
= &nfs_v2_clientops
;
486 /* Fill in pseudoflavor for mount version < 5 */
487 if (!(data
->flags
& NFS_MOUNT_SECFLAVOUR
))
488 data
->pseudoflavor
= RPC_AUTH_UNIX
;
489 authflavor
= data
->pseudoflavor
; /* save for sb_init() */
490 /* XXX maybe we want to add a server->pseudoflavor field */
492 /* Create RPC client handles */
493 server
->client
= nfs_create_client(server
, data
);
494 if (IS_ERR(server
->client
))
495 return PTR_ERR(server
->client
);
496 /* RFC 2623, sec 2.3.2 */
497 if (authflavor
!= RPC_AUTH_UNIX
) {
498 struct rpc_auth
*auth
;
500 server
->client_sys
= rpc_clone_client(server
->client
);
501 if (IS_ERR(server
->client_sys
))
502 return PTR_ERR(server
->client_sys
);
503 auth
= rpcauth_create(RPC_AUTH_UNIX
, server
->client_sys
);
505 return PTR_ERR(auth
);
507 atomic_inc(&server
->client
->cl_count
);
508 server
->client_sys
= server
->client
;
510 if (server
->flags
& NFS_MOUNT_VER3
) {
511 #ifdef CONFIG_NFS_V3_ACL
512 if (!(server
->flags
& NFS_MOUNT_NOACL
)) {
513 server
->client_acl
= rpc_bind_new_program(server
->client
, &nfsacl_program
, 3);
514 /* No errors! Assume that Sun nfsacls are supported */
515 if (!IS_ERR(server
->client_acl
))
516 server
->caps
|= NFS_CAP_ACLS
;
519 server
->flags
&= ~NFS_MOUNT_NOACL
;
520 #endif /* CONFIG_NFS_V3_ACL */
522 * The VFS shouldn't apply the umask to mode bits. We will
523 * do so ourselves when necessary.
525 sb
->s_flags
|= MS_POSIXACL
;
526 if (server
->namelen
== 0 || server
->namelen
> NFS3_MAXNAMLEN
)
527 server
->namelen
= NFS3_MAXNAMLEN
;
530 if (server
->namelen
== 0 || server
->namelen
> NFS2_MAXNAMLEN
)
531 server
->namelen
= NFS2_MAXNAMLEN
;
534 sb
->s_op
= &nfs_sops
;
535 return nfs_sb_init(sb
, authflavor
);
539 nfs_statfs(struct super_block
*sb
, struct kstatfs
*buf
)
541 struct nfs_server
*server
= NFS_SB(sb
);
542 unsigned char blockbits
;
543 unsigned long blockres
;
544 struct nfs_fh
*rootfh
= NFS_FH(sb
->s_root
->d_inode
);
545 struct nfs_fattr fattr
;
546 struct nfs_fsstat res
= {
553 error
= server
->rpc_ops
->statfs(server
, rootfh
, &res
);
554 buf
->f_type
= NFS_SUPER_MAGIC
;
559 * Current versions of glibc do not correctly handle the
560 * case where f_frsize != f_bsize. Eventually we want to
561 * report the value of wtmult in this field.
563 buf
->f_frsize
= sb
->s_blocksize
;
566 * On most *nix systems, f_blocks, f_bfree, and f_bavail
567 * are reported in units of f_frsize. Linux hasn't had
568 * an f_frsize field in its statfs struct until recently,
569 * thus historically Linux's sys_statfs reports these
570 * fields in units of f_bsize.
572 buf
->f_bsize
= sb
->s_blocksize
;
573 blockbits
= sb
->s_blocksize_bits
;
574 blockres
= (1 << blockbits
) - 1;
575 buf
->f_blocks
= (res
.tbytes
+ blockres
) >> blockbits
;
576 buf
->f_bfree
= (res
.fbytes
+ blockres
) >> blockbits
;
577 buf
->f_bavail
= (res
.abytes
+ blockres
) >> blockbits
;
579 buf
->f_files
= res
.tfiles
;
580 buf
->f_ffree
= res
.afiles
;
582 buf
->f_namelen
= server
->namelen
;
588 dprintk("%s: statfs error = %d\n", __FUNCTION__
, -error
);
589 buf
->f_bsize
= buf
->f_blocks
= buf
->f_bfree
= buf
->f_bavail
= -1;
594 static void nfs_show_mount_options(struct seq_file
*m
, struct nfs_server
*nfss
, int showdefaults
)
596 static struct proc_nfs_info
{
601 { NFS_MOUNT_SOFT
, ",soft", ",hard" },
602 { NFS_MOUNT_INTR
, ",intr", "" },
603 { NFS_MOUNT_NOCTO
, ",nocto", "" },
604 { NFS_MOUNT_NOAC
, ",noac", "" },
605 { NFS_MOUNT_NONLM
, ",nolock", "" },
606 { NFS_MOUNT_NOACL
, ",noacl", "" },
609 struct proc_nfs_info
*nfs_infop
;
613 seq_printf(m
, ",vers=%d", nfss
->rpc_ops
->version
);
614 seq_printf(m
, ",rsize=%d", nfss
->rsize
);
615 seq_printf(m
, ",wsize=%d", nfss
->wsize
);
616 if (nfss
->acregmin
!= 3*HZ
|| showdefaults
)
617 seq_printf(m
, ",acregmin=%d", nfss
->acregmin
/HZ
);
618 if (nfss
->acregmax
!= 60*HZ
|| showdefaults
)
619 seq_printf(m
, ",acregmax=%d", nfss
->acregmax
/HZ
);
620 if (nfss
->acdirmin
!= 30*HZ
|| showdefaults
)
621 seq_printf(m
, ",acdirmin=%d", nfss
->acdirmin
/HZ
);
622 if (nfss
->acdirmax
!= 60*HZ
|| showdefaults
)
623 seq_printf(m
, ",acdirmax=%d", nfss
->acdirmax
/HZ
);
624 for (nfs_infop
= nfs_info
; nfs_infop
->flag
; nfs_infop
++) {
625 if (nfss
->flags
& nfs_infop
->flag
)
626 seq_puts(m
, nfs_infop
->str
);
628 seq_puts(m
, nfs_infop
->nostr
);
630 switch (nfss
->client
->cl_xprt
->prot
) {
638 snprintf(buf
, sizeof(buf
), "%u", nfss
->client
->cl_xprt
->prot
);
641 seq_printf(m
, ",proto=%s", proto
);
642 seq_printf(m
, ",timeo=%lu", 10U * nfss
->retrans_timeo
/ HZ
);
643 seq_printf(m
, ",retrans=%u", nfss
->retrans_count
);
646 static int nfs_show_options(struct seq_file
*m
, struct vfsmount
*mnt
)
648 struct nfs_server
*nfss
= NFS_SB(mnt
->mnt_sb
);
650 nfs_show_mount_options(m
, nfss
, 0);
652 seq_puts(m
, ",addr=");
653 seq_escape(m
, nfss
->hostname
, " \t\n\\");
658 static int nfs_show_stats(struct seq_file
*m
, struct vfsmount
*mnt
)
661 struct nfs_server
*nfss
= NFS_SB(mnt
->mnt_sb
);
662 struct rpc_auth
*auth
= nfss
->client
->cl_auth
;
663 struct nfs_iostats totals
= { };
665 seq_printf(m
, "statvers=%s", NFS_IOSTAT_VERS
);
668 * Display all mount option settings
670 seq_printf(m
, "\n\topts:\t");
671 seq_puts(m
, mnt
->mnt_sb
->s_flags
& MS_RDONLY
? "ro" : "rw");
672 seq_puts(m
, mnt
->mnt_sb
->s_flags
& MS_SYNCHRONOUS
? ",sync" : "");
673 seq_puts(m
, mnt
->mnt_sb
->s_flags
& MS_NOATIME
? ",noatime" : "");
674 seq_puts(m
, mnt
->mnt_sb
->s_flags
& MS_NODIRATIME
? ",nodiratime" : "");
675 nfs_show_mount_options(m
, nfss
, 1);
677 seq_printf(m
, "\n\tcaps:\t");
678 seq_printf(m
, "caps=0x%x", nfss
->caps
);
679 seq_printf(m
, ",wtmult=%d", nfss
->wtmult
);
680 seq_printf(m
, ",dtsize=%d", nfss
->dtsize
);
681 seq_printf(m
, ",bsize=%d", nfss
->bsize
);
682 seq_printf(m
, ",namelen=%d", nfss
->namelen
);
685 if (nfss
->rpc_ops
->version
== 4) {
686 seq_printf(m
, "\n\tnfsv4:\t");
687 seq_printf(m
, "bm0=0x%x", nfss
->attr_bitmask
[0]);
688 seq_printf(m
, ",bm1=0x%x", nfss
->attr_bitmask
[1]);
689 seq_printf(m
, ",acl=0x%x", nfss
->acl_bitmask
);
694 * Display security flavor in effect for this mount
696 seq_printf(m
, "\n\tsec:\tflavor=%d", auth
->au_ops
->au_flavor
);
698 seq_printf(m
, ",pseudoflavor=%d", auth
->au_flavor
);
701 * Display superblock I/O counters
703 for (cpu
= 0; cpu
< NR_CPUS
; cpu
++) {
704 struct nfs_iostats
*stats
;
706 if (!cpu_possible(cpu
))
710 stats
= per_cpu_ptr(nfss
->io_stats
, cpu
);
712 for (i
= 0; i
< __NFSIOS_COUNTSMAX
; i
++)
713 totals
.events
[i
] += stats
->events
[i
];
714 for (i
= 0; i
< __NFSIOS_BYTESMAX
; i
++)
715 totals
.bytes
[i
] += stats
->bytes
[i
];
720 seq_printf(m
, "\n\tevents:\t");
721 for (i
= 0; i
< __NFSIOS_COUNTSMAX
; i
++)
722 seq_printf(m
, "%lu ", totals
.events
[i
]);
723 seq_printf(m
, "\n\tbytes:\t");
724 for (i
= 0; i
< __NFSIOS_BYTESMAX
; i
++)
725 seq_printf(m
, "%Lu ", totals
.bytes
[i
]);
731 * nfs_sync_mapping - helper to flush all mmapped dirty data to disk
733 int nfs_sync_mapping(struct address_space
*mapping
)
737 if (mapping
->nrpages
== 0)
739 unmap_mapping_range(mapping
, 0, 0, 0);
740 ret
= filemap_write_and_wait(mapping
);
743 ret
= nfs_wb_all(mapping
->host
);
749 * Invalidate the local caches
751 static void nfs_zap_caches_locked(struct inode
*inode
)
753 struct nfs_inode
*nfsi
= NFS_I(inode
);
754 int mode
= inode
->i_mode
;
756 NFS_ATTRTIMEO(inode
) = NFS_MINATTRTIMEO(inode
);
757 NFS_ATTRTIMEO_UPDATE(inode
) = jiffies
;
759 memset(NFS_COOKIEVERF(inode
), 0, sizeof(NFS_COOKIEVERF(inode
)));
760 if (S_ISREG(mode
) || S_ISDIR(mode
) || S_ISLNK(mode
))
761 nfsi
->cache_validity
|= NFS_INO_INVALID_ATTR
|NFS_INO_INVALID_DATA
|NFS_INO_INVALID_ACCESS
|NFS_INO_INVALID_ACL
|NFS_INO_REVAL_PAGECACHE
;
763 nfsi
->cache_validity
|= NFS_INO_INVALID_ATTR
|NFS_INO_INVALID_ACCESS
|NFS_INO_INVALID_ACL
|NFS_INO_REVAL_PAGECACHE
;
766 void nfs_zap_caches(struct inode
*inode
)
768 spin_lock(&inode
->i_lock
);
769 nfs_zap_caches_locked(inode
);
770 spin_unlock(&inode
->i_lock
);
773 static void nfs_zap_acl_cache(struct inode
*inode
)
775 void (*clear_acl_cache
)(struct inode
*);
777 clear_acl_cache
= NFS_PROTO(inode
)->clear_acl_cache
;
778 if (clear_acl_cache
!= NULL
)
779 clear_acl_cache(inode
);
780 spin_lock(&inode
->i_lock
);
781 NFS_I(inode
)->cache_validity
&= ~NFS_INO_INVALID_ACL
;
782 spin_unlock(&inode
->i_lock
);
786 * Invalidate, but do not unhash, the inode.
787 * NB: must be called with inode->i_lock held!
789 static void nfs_invalidate_inode(struct inode
*inode
)
791 set_bit(NFS_INO_STALE
, &NFS_FLAGS(inode
));
792 nfs_zap_caches_locked(inode
);
795 struct nfs_find_desc
{
797 struct nfs_fattr
*fattr
;
801 * In NFSv3 we can have 64bit inode numbers. In order to support
802 * this, and re-exported directories (also seen in NFSv2)
803 * we are forced to allow 2 different inodes to have the same
807 nfs_find_actor(struct inode
*inode
, void *opaque
)
809 struct nfs_find_desc
*desc
= (struct nfs_find_desc
*)opaque
;
810 struct nfs_fh
*fh
= desc
->fh
;
811 struct nfs_fattr
*fattr
= desc
->fattr
;
813 if (NFS_FILEID(inode
) != fattr
->fileid
)
815 if (nfs_compare_fh(NFS_FH(inode
), fh
))
817 if (is_bad_inode(inode
) || NFS_STALE(inode
))
823 nfs_init_locked(struct inode
*inode
, void *opaque
)
825 struct nfs_find_desc
*desc
= (struct nfs_find_desc
*)opaque
;
826 struct nfs_fattr
*fattr
= desc
->fattr
;
828 NFS_FILEID(inode
) = fattr
->fileid
;
829 nfs_copy_fh(NFS_FH(inode
), desc
->fh
);
833 /* Don't use READDIRPLUS on directories that we believe are too large */
834 #define NFS_LIMIT_READDIRPLUS (8*PAGE_SIZE)
837 * This is our front-end to iget that looks up inodes by file handle
838 * instead of inode number.
841 nfs_fhget(struct super_block
*sb
, struct nfs_fh
*fh
, struct nfs_fattr
*fattr
)
843 struct nfs_find_desc desc
= {
847 struct inode
*inode
= NULL
;
850 if ((fattr
->valid
& NFS_ATTR_FATTR
) == 0)
854 printk("NFS: Buggy server - nlink == 0!\n");
858 hash
= nfs_fattr_to_ino_t(fattr
);
860 if (!(inode
= iget5_locked(sb
, hash
, nfs_find_actor
, nfs_init_locked
, &desc
)))
863 if (inode
->i_state
& I_NEW
) {
864 struct nfs_inode
*nfsi
= NFS_I(inode
);
866 /* We set i_ino for the few things that still rely on it,
870 /* We can't support update_atime(), since the server will reset it */
871 inode
->i_flags
|= S_NOATIME
|S_NOCMTIME
;
872 inode
->i_mode
= fattr
->mode
;
873 /* Why so? Because we want revalidate for devices/FIFOs, and
874 * that's precisely what we have in nfs_file_inode_operations.
876 inode
->i_op
= NFS_SB(sb
)->rpc_ops
->file_inode_ops
;
877 if (S_ISREG(inode
->i_mode
)) {
878 inode
->i_fop
= &nfs_file_operations
;
879 inode
->i_data
.a_ops
= &nfs_file_aops
;
880 inode
->i_data
.backing_dev_info
= &NFS_SB(sb
)->backing_dev_info
;
881 } else if (S_ISDIR(inode
->i_mode
)) {
882 inode
->i_op
= NFS_SB(sb
)->rpc_ops
->dir_inode_ops
;
883 inode
->i_fop
= &nfs_dir_operations
;
884 if (nfs_server_capable(inode
, NFS_CAP_READDIRPLUS
)
885 && fattr
->size
<= NFS_LIMIT_READDIRPLUS
)
886 set_bit(NFS_INO_ADVISE_RDPLUS
, &NFS_FLAGS(inode
));
887 } else if (S_ISLNK(inode
->i_mode
))
888 inode
->i_op
= &nfs_symlink_inode_operations
;
890 init_special_inode(inode
, inode
->i_mode
, fattr
->rdev
);
892 nfsi
->read_cache_jiffies
= fattr
->time_start
;
893 nfsi
->last_updated
= jiffies
;
894 inode
->i_atime
= fattr
->atime
;
895 inode
->i_mtime
= fattr
->mtime
;
896 inode
->i_ctime
= fattr
->ctime
;
897 if (fattr
->valid
& NFS_ATTR_FATTR_V4
)
898 nfsi
->change_attr
= fattr
->change_attr
;
899 inode
->i_size
= nfs_size_to_loff_t(fattr
->size
);
900 inode
->i_nlink
= fattr
->nlink
;
901 inode
->i_uid
= fattr
->uid
;
902 inode
->i_gid
= fattr
->gid
;
903 if (fattr
->valid
& (NFS_ATTR_FATTR_V3
| NFS_ATTR_FATTR_V4
)) {
905 * report the blocks in 512byte units
907 inode
->i_blocks
= nfs_calc_block_size(fattr
->du
.nfs3
.used
);
908 inode
->i_blksize
= inode
->i_sb
->s_blocksize
;
910 inode
->i_blocks
= fattr
->du
.nfs2
.blocks
;
911 inode
->i_blksize
= fattr
->du
.nfs2
.blocksize
;
913 nfsi
->attrtimeo
= NFS_MINATTRTIMEO(inode
);
914 nfsi
->attrtimeo_timestamp
= jiffies
;
915 memset(nfsi
->cookieverf
, 0, sizeof(nfsi
->cookieverf
));
916 nfsi
->cache_access
.cred
= NULL
;
918 unlock_new_inode(inode
);
920 nfs_refresh_inode(inode
, fattr
);
921 dprintk("NFS: nfs_fhget(%s/%Ld ct=%d)\n",
923 (long long)NFS_FILEID(inode
),
924 atomic_read(&inode
->i_count
));
930 printk("nfs_fhget: iget failed\n");
934 #define NFS_VALID_ATTRS (ATTR_MODE|ATTR_UID|ATTR_GID|ATTR_SIZE|ATTR_ATIME|ATTR_ATIME_SET|ATTR_MTIME|ATTR_MTIME_SET)
937 nfs_setattr(struct dentry
*dentry
, struct iattr
*attr
)
939 struct inode
*inode
= dentry
->d_inode
;
940 struct nfs_fattr fattr
;
943 if (attr
->ia_valid
& ATTR_SIZE
) {
944 if (!S_ISREG(inode
->i_mode
) || attr
->ia_size
== i_size_read(inode
))
945 attr
->ia_valid
&= ~ATTR_SIZE
;
948 /* Optimization: if the end result is no change, don't RPC */
949 attr
->ia_valid
&= NFS_VALID_ATTRS
;
950 if (attr
->ia_valid
== 0)
954 nfs_begin_data_update(inode
);
955 /* Write all dirty data */
956 filemap_write_and_wait(inode
->i_mapping
);
959 * Return any delegations if we're going to change ACLs
961 if ((attr
->ia_valid
& (ATTR_MODE
|ATTR_UID
|ATTR_GID
)) != 0)
962 nfs_inode_return_delegation(inode
);
963 error
= NFS_PROTO(inode
)->setattr(dentry
, &fattr
, attr
);
965 nfs_refresh_inode(inode
, &fattr
);
966 nfs_end_data_update(inode
);
972 * nfs_setattr_update_inode - Update inode metadata after a setattr call.
973 * @inode: pointer to struct inode
974 * @attr: pointer to struct iattr
976 * Note: we do this in the *proc.c in order to ensure that
977 * it works for things like exclusive creates too.
979 void nfs_setattr_update_inode(struct inode
*inode
, struct iattr
*attr
)
981 if ((attr
->ia_valid
& (ATTR_MODE
|ATTR_UID
|ATTR_GID
)) != 0) {
982 if ((attr
->ia_valid
& ATTR_MODE
) != 0) {
983 int mode
= attr
->ia_mode
& S_IALLUGO
;
984 mode
|= inode
->i_mode
& ~S_IALLUGO
;
985 inode
->i_mode
= mode
;
987 if ((attr
->ia_valid
& ATTR_UID
) != 0)
988 inode
->i_uid
= attr
->ia_uid
;
989 if ((attr
->ia_valid
& ATTR_GID
) != 0)
990 inode
->i_gid
= attr
->ia_gid
;
991 spin_lock(&inode
->i_lock
);
992 NFS_I(inode
)->cache_validity
|= NFS_INO_INVALID_ACCESS
|NFS_INO_INVALID_ACL
;
993 spin_unlock(&inode
->i_lock
);
995 if ((attr
->ia_valid
& ATTR_SIZE
) != 0) {
996 inode
->i_size
= attr
->ia_size
;
997 vmtruncate(inode
, attr
->ia_size
);
1001 static int nfs_wait_schedule(void *word
)
1003 if (signal_pending(current
))
1004 return -ERESTARTSYS
;
1010 * Wait for the inode to get unlocked.
1012 static int nfs_wait_on_inode(struct inode
*inode
)
1014 struct rpc_clnt
*clnt
= NFS_CLIENT(inode
);
1015 struct nfs_inode
*nfsi
= NFS_I(inode
);
1019 rpc_clnt_sigmask(clnt
, &oldmask
);
1020 error
= wait_on_bit_lock(&nfsi
->flags
, NFS_INO_REVALIDATING
,
1021 nfs_wait_schedule
, TASK_INTERRUPTIBLE
);
1022 rpc_clnt_sigunmask(clnt
, &oldmask
);
1027 static void nfs_wake_up_inode(struct inode
*inode
)
1029 struct nfs_inode
*nfsi
= NFS_I(inode
);
1031 clear_bit(NFS_INO_REVALIDATING
, &nfsi
->flags
);
1032 smp_mb__after_clear_bit();
1033 wake_up_bit(&nfsi
->flags
, NFS_INO_REVALIDATING
);
1036 int nfs_getattr(struct vfsmount
*mnt
, struct dentry
*dentry
, struct kstat
*stat
)
1038 struct inode
*inode
= dentry
->d_inode
;
1039 int need_atime
= NFS_I(inode
)->cache_validity
& NFS_INO_INVALID_ATIME
;
1042 /* Flush out writes to the server in order to update c/mtime */
1043 nfs_sync_inode(inode
, 0, 0, FLUSH_WAIT
|FLUSH_NOCOMMIT
);
1046 * We may force a getattr if the user cares about atime.
1048 * Note that we only have to check the vfsmount flags here:
1049 * - NFS always sets S_NOATIME by so checking it would give a
1051 * - NFS never sets MS_NOATIME or MS_NODIRATIME so there is
1052 * no point in checking those.
1054 if ((mnt
->mnt_flags
& MNT_NOATIME
) ||
1055 ((mnt
->mnt_flags
& MNT_NODIRATIME
) && S_ISDIR(inode
->i_mode
)))
1059 err
= __nfs_revalidate_inode(NFS_SERVER(inode
), inode
);
1061 err
= nfs_revalidate_inode(NFS_SERVER(inode
), inode
);
1063 generic_fillattr(inode
, stat
);
1067 static struct nfs_open_context
*alloc_nfs_open_context(struct vfsmount
*mnt
, struct dentry
*dentry
, struct rpc_cred
*cred
)
1069 struct nfs_open_context
*ctx
;
1071 ctx
= (struct nfs_open_context
*)kmalloc(sizeof(*ctx
), GFP_KERNEL
);
1073 atomic_set(&ctx
->count
, 1);
1074 ctx
->dentry
= dget(dentry
);
1075 ctx
->vfsmnt
= mntget(mnt
);
1076 ctx
->cred
= get_rpccred(cred
);
1078 ctx
->lockowner
= current
->files
;
1080 ctx
->dir_cookie
= 0;
1085 struct nfs_open_context
*get_nfs_open_context(struct nfs_open_context
*ctx
)
1088 atomic_inc(&ctx
->count
);
1092 void put_nfs_open_context(struct nfs_open_context
*ctx
)
1094 if (atomic_dec_and_test(&ctx
->count
)) {
1095 if (!list_empty(&ctx
->list
)) {
1096 struct inode
*inode
= ctx
->dentry
->d_inode
;
1097 spin_lock(&inode
->i_lock
);
1098 list_del(&ctx
->list
);
1099 spin_unlock(&inode
->i_lock
);
1101 if (ctx
->state
!= NULL
)
1102 nfs4_close_state(ctx
->state
, ctx
->mode
);
1103 if (ctx
->cred
!= NULL
)
1104 put_rpccred(ctx
->cred
);
1106 mntput(ctx
->vfsmnt
);
1112 * Ensure that mmap has a recent RPC credential for use when writing out
1115 static void nfs_file_set_open_context(struct file
*filp
, struct nfs_open_context
*ctx
)
1117 struct inode
*inode
= filp
->f_dentry
->d_inode
;
1118 struct nfs_inode
*nfsi
= NFS_I(inode
);
1120 filp
->private_data
= get_nfs_open_context(ctx
);
1121 spin_lock(&inode
->i_lock
);
1122 list_add(&ctx
->list
, &nfsi
->open_files
);
1123 spin_unlock(&inode
->i_lock
);
1127 * Given an inode, search for an open context with the desired characteristics
1129 struct nfs_open_context
*nfs_find_open_context(struct inode
*inode
, struct rpc_cred
*cred
, int mode
)
1131 struct nfs_inode
*nfsi
= NFS_I(inode
);
1132 struct nfs_open_context
*pos
, *ctx
= NULL
;
1134 spin_lock(&inode
->i_lock
);
1135 list_for_each_entry(pos
, &nfsi
->open_files
, list
) {
1136 if (cred
!= NULL
&& pos
->cred
!= cred
)
1138 if ((pos
->mode
& mode
) == mode
) {
1139 ctx
= get_nfs_open_context(pos
);
1143 spin_unlock(&inode
->i_lock
);
1147 static void nfs_file_clear_open_context(struct file
*filp
)
1149 struct inode
*inode
= filp
->f_dentry
->d_inode
;
1150 struct nfs_open_context
*ctx
= (struct nfs_open_context
*)filp
->private_data
;
1153 filp
->private_data
= NULL
;
1154 spin_lock(&inode
->i_lock
);
1155 list_move_tail(&ctx
->list
, &NFS_I(inode
)->open_files
);
1156 spin_unlock(&inode
->i_lock
);
1157 put_nfs_open_context(ctx
);
1162 * These allocate and release file read/write context information.
1164 int nfs_open(struct inode
*inode
, struct file
*filp
)
1166 struct nfs_open_context
*ctx
;
1167 struct rpc_cred
*cred
;
1169 cred
= rpcauth_lookupcred(NFS_CLIENT(inode
)->cl_auth
, 0);
1171 return PTR_ERR(cred
);
1172 ctx
= alloc_nfs_open_context(filp
->f_vfsmnt
, filp
->f_dentry
, cred
);
1176 ctx
->mode
= filp
->f_mode
;
1177 nfs_file_set_open_context(filp
, ctx
);
1178 put_nfs_open_context(ctx
);
1182 int nfs_release(struct inode
*inode
, struct file
*filp
)
1184 nfs_file_clear_open_context(filp
);
1189 * This function is called whenever some part of NFS notices that
1190 * the cached attributes have to be refreshed.
1193 __nfs_revalidate_inode(struct nfs_server
*server
, struct inode
*inode
)
1195 int status
= -ESTALE
;
1196 struct nfs_fattr fattr
;
1197 struct nfs_inode
*nfsi
= NFS_I(inode
);
1199 dfprintk(PAGECACHE
, "NFS: revalidating (%s/%Ld)\n",
1200 inode
->i_sb
->s_id
, (long long)NFS_FILEID(inode
));
1203 if (!inode
|| is_bad_inode(inode
))
1205 if (NFS_STALE(inode
))
1208 status
= nfs_wait_on_inode(inode
);
1211 if (NFS_STALE(inode
)) {
1213 /* Do we trust the cached ESTALE? */
1214 if (NFS_ATTRTIMEO(inode
) != 0) {
1215 if (nfsi
->cache_validity
& (NFS_INO_INVALID_ATTR
|NFS_INO_INVALID_DATA
|NFS_INO_INVALID_ATIME
)) {
1222 status
= NFS_PROTO(inode
)->getattr(server
, NFS_FH(inode
), &fattr
);
1224 dfprintk(PAGECACHE
, "nfs_revalidate_inode: (%s/%Ld) getattr failed, error=%d\n",
1226 (long long)NFS_FILEID(inode
), status
);
1227 if (status
== -ESTALE
) {
1228 nfs_zap_caches(inode
);
1229 if (!S_ISDIR(inode
->i_mode
))
1230 set_bit(NFS_INO_STALE
, &NFS_FLAGS(inode
));
1235 spin_lock(&inode
->i_lock
);
1236 status
= nfs_update_inode(inode
, &fattr
);
1238 spin_unlock(&inode
->i_lock
);
1239 dfprintk(PAGECACHE
, "nfs_revalidate_inode: (%s/%Ld) refresh failed, error=%d\n",
1241 (long long)NFS_FILEID(inode
), status
);
1244 spin_unlock(&inode
->i_lock
);
1246 nfs_revalidate_mapping(inode
, inode
->i_mapping
);
1248 if (nfsi
->cache_validity
& NFS_INO_INVALID_ACL
)
1249 nfs_zap_acl_cache(inode
);
1251 dfprintk(PAGECACHE
, "NFS: (%s/%Ld) revalidation complete\n",
1253 (long long)NFS_FILEID(inode
));
1256 nfs_wake_up_inode(inode
);
1263 int nfs_attribute_timeout(struct inode
*inode
)
1265 struct nfs_inode
*nfsi
= NFS_I(inode
);
1267 if (nfs_have_delegation(inode
, FMODE_READ
))
1269 return time_after(jiffies
, nfsi
->read_cache_jiffies
+nfsi
->attrtimeo
);
1273 * nfs_revalidate_inode - Revalidate the inode attributes
1274 * @server - pointer to nfs_server struct
1275 * @inode - pointer to inode struct
1277 * Updates inode attribute information by retrieving the data from the server.
1279 int nfs_revalidate_inode(struct nfs_server
*server
, struct inode
*inode
)
1281 if (!(NFS_I(inode
)->cache_validity
& (NFS_INO_INVALID_ATTR
|NFS_INO_INVALID_DATA
))
1282 && !nfs_attribute_timeout(inode
))
1283 return NFS_STALE(inode
) ? -ESTALE
: 0;
1284 return __nfs_revalidate_inode(server
, inode
);
1288 * nfs_revalidate_mapping - Revalidate the pagecache
1289 * @inode - pointer to host inode
1290 * @mapping - pointer to mapping
1292 void nfs_revalidate_mapping(struct inode
*inode
, struct address_space
*mapping
)
1294 struct nfs_inode
*nfsi
= NFS_I(inode
);
1296 if (nfsi
->cache_validity
& NFS_INO_INVALID_DATA
) {
1297 if (S_ISREG(inode
->i_mode
))
1298 nfs_sync_mapping(mapping
);
1299 invalidate_inode_pages2(mapping
);
1301 spin_lock(&inode
->i_lock
);
1302 nfsi
->cache_validity
&= ~NFS_INO_INVALID_DATA
;
1303 if (S_ISDIR(inode
->i_mode
)) {
1304 memset(nfsi
->cookieverf
, 0, sizeof(nfsi
->cookieverf
));
1305 /* This ensures we revalidate child dentries */
1306 nfsi
->cache_change_attribute
= jiffies
;
1308 spin_unlock(&inode
->i_lock
);
1310 dfprintk(PAGECACHE
, "NFS: (%s/%Ld) data cache invalidated\n",
1312 (long long)NFS_FILEID(inode
));
1317 * nfs_begin_data_update
1318 * @inode - pointer to inode
1319 * Declare that a set of operations will update file data on the server
1321 void nfs_begin_data_update(struct inode
*inode
)
1323 atomic_inc(&NFS_I(inode
)->data_updates
);
1327 * nfs_end_data_update
1328 * @inode - pointer to inode
1329 * Declare end of the operations that will update file data
1330 * This will mark the inode as immediately needing revalidation
1331 * of its attribute cache.
1333 void nfs_end_data_update(struct inode
*inode
)
1335 struct nfs_inode
*nfsi
= NFS_I(inode
);
1337 if (!nfs_have_delegation(inode
, FMODE_READ
)) {
1338 /* Directories and symlinks: invalidate page cache */
1339 if (S_ISDIR(inode
->i_mode
) || S_ISLNK(inode
->i_mode
)) {
1340 spin_lock(&inode
->i_lock
);
1341 nfsi
->cache_validity
|= NFS_INO_INVALID_DATA
;
1342 spin_unlock(&inode
->i_lock
);
1345 nfsi
->cache_change_attribute
= jiffies
;
1346 atomic_dec(&nfsi
->data_updates
);
1349 static void nfs_wcc_update_inode(struct inode
*inode
, struct nfs_fattr
*fattr
)
1351 struct nfs_inode
*nfsi
= NFS_I(inode
);
1353 if ((fattr
->valid
& NFS_ATTR_PRE_CHANGE
) != 0
1354 && nfsi
->change_attr
== fattr
->pre_change_attr
) {
1355 nfsi
->change_attr
= fattr
->change_attr
;
1356 nfsi
->cache_change_attribute
= jiffies
;
1359 /* If we have atomic WCC data, we may update some attributes */
1360 if ((fattr
->valid
& NFS_ATTR_WCC
) != 0) {
1361 if (timespec_equal(&inode
->i_ctime
, &fattr
->pre_ctime
)) {
1362 memcpy(&inode
->i_ctime
, &fattr
->ctime
, sizeof(inode
->i_ctime
));
1363 nfsi
->cache_change_attribute
= jiffies
;
1365 if (timespec_equal(&inode
->i_mtime
, &fattr
->pre_mtime
)) {
1366 memcpy(&inode
->i_mtime
, &fattr
->mtime
, sizeof(inode
->i_mtime
));
1367 nfsi
->cache_change_attribute
= jiffies
;
1369 if (inode
->i_size
== fattr
->pre_size
&& nfsi
->npages
== 0) {
1370 inode
->i_size
= fattr
->size
;
1371 nfsi
->cache_change_attribute
= jiffies
;
1377 * nfs_check_inode_attributes - verify consistency of the inode attribute cache
1378 * @inode - pointer to inode
1379 * @fattr - updated attributes
1381 * Verifies the attribute cache. If we have just changed the attributes,
1382 * so that fattr carries weak cache consistency data, then it may
1383 * also update the ctime/mtime/change_attribute.
1385 static int nfs_check_inode_attributes(struct inode
*inode
, struct nfs_fattr
*fattr
)
1387 struct nfs_inode
*nfsi
= NFS_I(inode
);
1388 loff_t cur_size
, new_isize
;
1392 if ((fattr
->valid
& NFS_ATTR_FATTR
) == 0)
1395 /* Has the inode gone and changed behind our back? */
1396 if (nfsi
->fileid
!= fattr
->fileid
1397 || (inode
->i_mode
& S_IFMT
) != (fattr
->mode
& S_IFMT
)) {
1401 /* Are we in the process of updating data on the server? */
1402 data_unstable
= nfs_caches_unstable(inode
);
1404 /* Do atomic weak cache consistency updates */
1405 nfs_wcc_update_inode(inode
, fattr
);
1407 if ((fattr
->valid
& NFS_ATTR_FATTR_V4
) != 0) {
1408 if (nfsi
->change_attr
== fattr
->change_attr
)
1410 nfsi
->cache_validity
|= NFS_INO_INVALID_ATTR
;
1412 nfsi
->cache_validity
|= NFS_INO_REVAL_PAGECACHE
;
1415 /* Verify a few of the more important attributes */
1416 if (!timespec_equal(&inode
->i_mtime
, &fattr
->mtime
)) {
1417 nfsi
->cache_validity
|= NFS_INO_INVALID_ATTR
;
1419 nfsi
->cache_validity
|= NFS_INO_REVAL_PAGECACHE
;
1422 cur_size
= i_size_read(inode
);
1423 new_isize
= nfs_size_to_loff_t(fattr
->size
);
1424 if (cur_size
!= new_isize
&& nfsi
->npages
== 0)
1425 nfsi
->cache_validity
|= NFS_INO_INVALID_ATTR
|NFS_INO_REVAL_PAGECACHE
;
1427 /* Have any file permissions changed? */
1428 if ((inode
->i_mode
& S_IALLUGO
) != (fattr
->mode
& S_IALLUGO
)
1429 || inode
->i_uid
!= fattr
->uid
1430 || inode
->i_gid
!= fattr
->gid
)
1431 nfsi
->cache_validity
|= NFS_INO_INVALID_ATTR
| NFS_INO_INVALID_ACCESS
| NFS_INO_INVALID_ACL
;
1433 /* Has the link count changed? */
1434 if (inode
->i_nlink
!= fattr
->nlink
)
1435 nfsi
->cache_validity
|= NFS_INO_INVALID_ATTR
;
1438 if (!timespec_equal(&inode
->i_atime
, &fattr
->atime
))
1439 nfsi
->cache_validity
|= NFS_INO_INVALID_ATIME
;
1441 nfsi
->read_cache_jiffies
= fattr
->time_start
;
1446 * nfs_refresh_inode - try to update the inode attribute cache
1447 * @inode - pointer to inode
1448 * @fattr - updated attributes
1450 * Check that an RPC call that returned attributes has not overlapped with
1451 * other recent updates of the inode metadata, then decide whether it is
1452 * safe to do a full update of the inode attributes, or whether just to
1453 * call nfs_check_inode_attributes.
1455 int nfs_refresh_inode(struct inode
*inode
, struct nfs_fattr
*fattr
)
1457 struct nfs_inode
*nfsi
= NFS_I(inode
);
1460 if ((fattr
->valid
& NFS_ATTR_FATTR
) == 0)
1462 spin_lock(&inode
->i_lock
);
1463 nfsi
->cache_validity
&= ~NFS_INO_REVAL_PAGECACHE
;
1464 if (time_after(fattr
->time_start
, nfsi
->last_updated
))
1465 status
= nfs_update_inode(inode
, fattr
);
1467 status
= nfs_check_inode_attributes(inode
, fattr
);
1469 spin_unlock(&inode
->i_lock
);
1474 * nfs_post_op_update_inode - try to update the inode attribute cache
1475 * @inode - pointer to inode
1476 * @fattr - updated attributes
1478 * After an operation that has changed the inode metadata, mark the
1479 * attribute cache as being invalid, then try to update it.
1481 int nfs_post_op_update_inode(struct inode
*inode
, struct nfs_fattr
*fattr
)
1483 struct nfs_inode
*nfsi
= NFS_I(inode
);
1486 spin_lock(&inode
->i_lock
);
1487 if (unlikely((fattr
->valid
& NFS_ATTR_FATTR
) == 0)) {
1488 nfsi
->cache_validity
|= NFS_INO_INVALID_ATTR
| NFS_INO_INVALID_ACCESS
;
1491 status
= nfs_update_inode(inode
, fattr
);
1493 spin_unlock(&inode
->i_lock
);
1498 * Many nfs protocol calls return the new file attributes after
1499 * an operation. Here we update the inode to reflect the state
1500 * of the server's inode.
1502 * This is a bit tricky because we have to make sure all dirty pages
1503 * have been sent off to the server before calling invalidate_inode_pages.
1504 * To make sure no other process adds more write requests while we try
1505 * our best to flush them, we make them sleep during the attribute refresh.
1507 * A very similar scenario holds for the dir cache.
1509 static int nfs_update_inode(struct inode
*inode
, struct nfs_fattr
*fattr
)
1511 struct nfs_inode
*nfsi
= NFS_I(inode
);
1512 loff_t cur_isize
, new_isize
;
1513 unsigned int invalid
= 0;
1516 dfprintk(VFS
, "NFS: %s(%s/%ld ct=%d info=0x%x)\n",
1517 __FUNCTION__
, inode
->i_sb
->s_id
, inode
->i_ino
,
1518 atomic_read(&inode
->i_count
), fattr
->valid
);
1520 if ((fattr
->valid
& NFS_ATTR_FATTR
) == 0)
1523 if (nfsi
->fileid
!= fattr
->fileid
)
1527 * Make sure the inode's type hasn't changed.
1529 if ((inode
->i_mode
& S_IFMT
) != (fattr
->mode
& S_IFMT
))
1533 * Update the read time so we don't revalidate too often.
1535 nfsi
->read_cache_jiffies
= fattr
->time_start
;
1536 nfsi
->last_updated
= jiffies
;
1538 /* Are we racing with known updates of the metadata on the server? */
1539 data_stable
= nfs_verify_change_attribute(inode
, fattr
->time_start
);
1541 nfsi
->cache_validity
&= ~(NFS_INO_INVALID_ATTR
|NFS_INO_INVALID_ATIME
);
1543 /* Do atomic weak cache consistency updates */
1544 nfs_wcc_update_inode(inode
, fattr
);
1546 /* Check if our cached file size is stale */
1547 new_isize
= nfs_size_to_loff_t(fattr
->size
);
1548 cur_isize
= i_size_read(inode
);
1549 if (new_isize
!= cur_isize
) {
1550 /* Do we perhaps have any outstanding writes? */
1551 if (nfsi
->npages
== 0) {
1552 /* No, but did we race with nfs_end_data_update()? */
1554 inode
->i_size
= new_isize
;
1555 invalid
|= NFS_INO_INVALID_DATA
;
1557 invalid
|= NFS_INO_INVALID_ATTR
;
1558 } else if (new_isize
> cur_isize
) {
1559 inode
->i_size
= new_isize
;
1560 invalid
|= NFS_INO_INVALID_ATTR
|NFS_INO_INVALID_DATA
;
1562 nfsi
->cache_change_attribute
= jiffies
;
1563 dprintk("NFS: isize change on server for file %s/%ld\n",
1564 inode
->i_sb
->s_id
, inode
->i_ino
);
1567 /* Check if the mtime agrees */
1568 if (!timespec_equal(&inode
->i_mtime
, &fattr
->mtime
)) {
1569 memcpy(&inode
->i_mtime
, &fattr
->mtime
, sizeof(inode
->i_mtime
));
1570 dprintk("NFS: mtime change on server for file %s/%ld\n",
1571 inode
->i_sb
->s_id
, inode
->i_ino
);
1572 invalid
|= NFS_INO_INVALID_ATTR
|NFS_INO_INVALID_DATA
;
1573 nfsi
->cache_change_attribute
= jiffies
;
1576 /* If ctime has changed we should definitely clear access+acl caches */
1577 if (!timespec_equal(&inode
->i_ctime
, &fattr
->ctime
)) {
1578 invalid
|= NFS_INO_INVALID_ACCESS
|NFS_INO_INVALID_ACL
;
1579 memcpy(&inode
->i_ctime
, &fattr
->ctime
, sizeof(inode
->i_ctime
));
1580 nfsi
->cache_change_attribute
= jiffies
;
1582 memcpy(&inode
->i_atime
, &fattr
->atime
, sizeof(inode
->i_atime
));
1584 if ((inode
->i_mode
& S_IALLUGO
) != (fattr
->mode
& S_IALLUGO
) ||
1585 inode
->i_uid
!= fattr
->uid
||
1586 inode
->i_gid
!= fattr
->gid
)
1587 invalid
|= NFS_INO_INVALID_ATTR
|NFS_INO_INVALID_ACCESS
|NFS_INO_INVALID_ACL
;
1589 inode
->i_mode
= fattr
->mode
;
1590 inode
->i_nlink
= fattr
->nlink
;
1591 inode
->i_uid
= fattr
->uid
;
1592 inode
->i_gid
= fattr
->gid
;
1594 if (fattr
->valid
& (NFS_ATTR_FATTR_V3
| NFS_ATTR_FATTR_V4
)) {
1596 * report the blocks in 512byte units
1598 inode
->i_blocks
= nfs_calc_block_size(fattr
->du
.nfs3
.used
);
1599 inode
->i_blksize
= inode
->i_sb
->s_blocksize
;
1601 inode
->i_blocks
= fattr
->du
.nfs2
.blocks
;
1602 inode
->i_blksize
= fattr
->du
.nfs2
.blocksize
;
1605 if ((fattr
->valid
& NFS_ATTR_FATTR_V4
)) {
1606 if (nfsi
->change_attr
!= fattr
->change_attr
) {
1607 dprintk("NFS: change_attr change on server for file %s/%ld\n",
1608 inode
->i_sb
->s_id
, inode
->i_ino
);
1609 nfsi
->change_attr
= fattr
->change_attr
;
1610 invalid
|= NFS_INO_INVALID_ATTR
|NFS_INO_INVALID_DATA
|NFS_INO_INVALID_ACCESS
|NFS_INO_INVALID_ACL
;
1611 nfsi
->cache_change_attribute
= jiffies
;
1613 invalid
&= ~(NFS_INO_INVALID_ATTR
|NFS_INO_INVALID_DATA
);
1616 /* Update attrtimeo value if we're out of the unstable period */
1617 if (invalid
& NFS_INO_INVALID_ATTR
) {
1618 nfsi
->attrtimeo
= NFS_MINATTRTIMEO(inode
);
1619 nfsi
->attrtimeo_timestamp
= jiffies
;
1620 } else if (time_after(jiffies
, nfsi
->attrtimeo_timestamp
+nfsi
->attrtimeo
)) {
1621 if ((nfsi
->attrtimeo
<<= 1) > NFS_MAXATTRTIMEO(inode
))
1622 nfsi
->attrtimeo
= NFS_MAXATTRTIMEO(inode
);
1623 nfsi
->attrtimeo_timestamp
= jiffies
;
1625 /* Don't invalidate the data if we were to blame */
1626 if (!(S_ISREG(inode
->i_mode
) || S_ISDIR(inode
->i_mode
)
1627 || S_ISLNK(inode
->i_mode
)))
1628 invalid
&= ~NFS_INO_INVALID_DATA
;
1630 invalid
&= ~(NFS_INO_INVALID_ATTR
|NFS_INO_INVALID_ATIME
|NFS_INO_REVAL_PAGECACHE
);
1631 if (!nfs_have_delegation(inode
, FMODE_READ
))
1632 nfsi
->cache_validity
|= invalid
;
1637 * Big trouble! The inode has become a different object.
1640 printk(KERN_DEBUG
"%s: inode %ld mode changed, %07o to %07o\n",
1641 __FUNCTION__
, inode
->i_ino
, inode
->i_mode
, fattr
->mode
);
1645 * No need to worry about unhashing the dentry, as the
1646 * lookup validation will know that the inode is bad.
1647 * (But we fall through to invalidate the caches.)
1649 nfs_invalidate_inode(inode
);
1653 printk(KERN_ERR
"NFS: server %s error: fileid changed\n"
1654 "fsid %s: expected fileid 0x%Lx, got 0x%Lx\n",
1655 NFS_SERVER(inode
)->hostname
, inode
->i_sb
->s_id
,
1656 (long long)nfsi
->fileid
, (long long)fattr
->fileid
);
1661 * File system information
1664 static int nfs_set_super(struct super_block
*s
, void *data
)
1666 s
->s_fs_info
= data
;
1667 return set_anon_super(s
, data
);
1670 static int nfs_compare_super(struct super_block
*sb
, void *data
)
1672 struct nfs_server
*server
= data
;
1673 struct nfs_server
*old
= NFS_SB(sb
);
1675 if (old
->addr
.sin_addr
.s_addr
!= server
->addr
.sin_addr
.s_addr
)
1677 if (old
->addr
.sin_port
!= server
->addr
.sin_port
)
1679 return !nfs_compare_fh(&old
->fh
, &server
->fh
);
1682 static struct super_block
*nfs_get_sb(struct file_system_type
*fs_type
,
1683 int flags
, const char *dev_name
, void *raw_data
)
1686 struct nfs_server
*server
= NULL
;
1687 struct super_block
*s
;
1688 struct nfs_fh
*root
;
1689 struct nfs_mount_data
*data
= raw_data
;
1691 s
= ERR_PTR(-EINVAL
);
1693 dprintk("%s: missing data argument\n", __FUNCTION__
);
1696 if (data
->version
<= 0 || data
->version
> NFS_MOUNT_VERSION
) {
1697 dprintk("%s: bad mount version\n", __FUNCTION__
);
1700 switch (data
->version
) {
1706 if (data
->flags
& NFS_MOUNT_VER3
) {
1707 dprintk("%s: mount structure version %d does not support NFSv3\n",
1712 data
->root
.size
= NFS2_FHSIZE
;
1713 memcpy(data
->root
.data
, data
->old_root
.data
, NFS2_FHSIZE
);
1715 if (data
->flags
& NFS_MOUNT_SECFLAVOUR
) {
1716 dprintk("%s: mount structure version %d does not support strong security\n",
1722 memset(data
->context
, 0, sizeof(data
->context
));
1724 #ifndef CONFIG_NFS_V3
1725 /* If NFSv3 is not compiled in, return -EPROTONOSUPPORT */
1726 s
= ERR_PTR(-EPROTONOSUPPORT
);
1727 if (data
->flags
& NFS_MOUNT_VER3
) {
1728 dprintk("%s: NFSv3 not compiled into kernel\n", __FUNCTION__
);
1731 #endif /* CONFIG_NFS_V3 */
1733 s
= ERR_PTR(-ENOMEM
);
1734 server
= kzalloc(sizeof(struct nfs_server
), GFP_KERNEL
);
1737 /* Zero out the NFS state stuff */
1738 init_nfsv4_state(server
);
1739 server
->client
= server
->client_sys
= server
->client_acl
= ERR_PTR(-EINVAL
);
1742 if (data
->flags
& NFS_MOUNT_VER3
)
1743 root
->size
= data
->root
.size
;
1745 root
->size
= NFS2_FHSIZE
;
1746 s
= ERR_PTR(-EINVAL
);
1747 if (root
->size
> sizeof(root
->data
)) {
1748 dprintk("%s: invalid root filehandle\n", __FUNCTION__
);
1751 memcpy(root
->data
, data
->root
.data
, root
->size
);
1753 /* We now require that the mount process passes the remote address */
1754 memcpy(&server
->addr
, &data
->addr
, sizeof(server
->addr
));
1755 if (server
->addr
.sin_addr
.s_addr
== INADDR_ANY
) {
1756 dprintk("%s: mount program didn't pass remote address!\n",
1761 /* Fire up rpciod if not yet running */
1762 s
= ERR_PTR(rpciod_up());
1764 dprintk("%s: couldn't start rpciod! Error = %ld\n",
1765 __FUNCTION__
, PTR_ERR(s
));
1769 s
= sget(fs_type
, nfs_compare_super
, nfs_set_super
, server
);
1770 if (IS_ERR(s
) || s
->s_root
)
1771 goto out_rpciod_down
;
1775 error
= nfs_fill_super(s
, data
, flags
& MS_VERBOSE
? 1 : 0);
1777 up_write(&s
->s_umount
);
1778 deactivate_super(s
);
1779 return ERR_PTR(error
);
1781 s
->s_flags
|= MS_ACTIVE
;
1790 static void nfs_kill_super(struct super_block
*s
)
1792 struct nfs_server
*server
= NFS_SB(s
);
1796 if (!IS_ERR(server
->client
))
1797 rpc_shutdown_client(server
->client
);
1798 if (!IS_ERR(server
->client_sys
))
1799 rpc_shutdown_client(server
->client_sys
);
1800 if (!IS_ERR(server
->client_acl
))
1801 rpc_shutdown_client(server
->client_acl
);
1803 if (!(server
->flags
& NFS_MOUNT_NONLM
))
1804 lockd_down(); /* release rpc.lockd */
1806 rpciod_down(); /* release rpciod */
1808 kfree(server
->hostname
);
1812 static struct file_system_type nfs_fs_type
= {
1813 .owner
= THIS_MODULE
,
1815 .get_sb
= nfs_get_sb
,
1816 .kill_sb
= nfs_kill_super
,
1817 .fs_flags
= FS_ODD_RENAME
|FS_REVAL_DOT
|FS_BINARY_MOUNTDATA
,
1820 #ifdef CONFIG_NFS_V4
1822 static void nfs4_clear_inode(struct inode
*);
1825 static struct super_operations nfs4_sops
= {
1826 .alloc_inode
= nfs_alloc_inode
,
1827 .destroy_inode
= nfs_destroy_inode
,
1828 .write_inode
= nfs_write_inode
,
1829 .delete_inode
= nfs_delete_inode
,
1830 .statfs
= nfs_statfs
,
1831 .clear_inode
= nfs4_clear_inode
,
1832 .umount_begin
= nfs_umount_begin
,
1833 .show_options
= nfs_show_options
,
1834 .show_stats
= nfs_show_stats
,
1838 * Clean out any remaining NFSv4 state that might be left over due
1839 * to open() calls that passed nfs_atomic_lookup, but failed to call
1842 static void nfs4_clear_inode(struct inode
*inode
)
1844 struct nfs_inode
*nfsi
= NFS_I(inode
);
1846 /* If we are holding a delegation, return it! */
1847 nfs_inode_return_delegation(inode
);
1848 /* First call standard NFS clear_inode() code */
1849 nfs_clear_inode(inode
);
1850 /* Now clear out any remaining state */
1851 while (!list_empty(&nfsi
->open_states
)) {
1852 struct nfs4_state
*state
;
1854 state
= list_entry(nfsi
->open_states
.next
,
1857 dprintk("%s(%s/%Ld): found unclaimed NFSv4 state %p\n",
1860 (long long)NFS_FILEID(inode
),
1862 BUG_ON(atomic_read(&state
->count
) != 1);
1863 nfs4_close_state(state
, state
->state
);
1868 static int nfs4_fill_super(struct super_block
*sb
, struct nfs4_mount_data
*data
, int silent
)
1870 struct nfs_server
*server
;
1871 struct nfs4_client
*clp
= NULL
;
1872 struct rpc_xprt
*xprt
= NULL
;
1873 struct rpc_clnt
*clnt
= NULL
;
1874 struct rpc_timeout timeparms
;
1875 rpc_authflavor_t authflavour
;
1878 sb
->s_blocksize_bits
= 0;
1879 sb
->s_blocksize
= 0;
1880 server
= NFS_SB(sb
);
1881 if (data
->rsize
!= 0)
1882 server
->rsize
= nfs_block_size(data
->rsize
, NULL
);
1883 if (data
->wsize
!= 0)
1884 server
->wsize
= nfs_block_size(data
->wsize
, NULL
);
1885 server
->flags
= data
->flags
& NFS_MOUNT_FLAGMASK
;
1886 server
->caps
= NFS_CAP_ATOMIC_OPEN
;
1888 server
->acregmin
= data
->acregmin
*HZ
;
1889 server
->acregmax
= data
->acregmax
*HZ
;
1890 server
->acdirmin
= data
->acdirmin
*HZ
;
1891 server
->acdirmax
= data
->acdirmax
*HZ
;
1893 server
->rpc_ops
= &nfs_v4_clientops
;
1895 nfs_init_timeout_values(&timeparms
, data
->proto
, data
->timeo
, data
->retrans
);
1897 server
->retrans_timeo
= timeparms
.to_initval
;
1898 server
->retrans_count
= timeparms
.to_retries
;
1900 clp
= nfs4_get_client(&server
->addr
.sin_addr
);
1902 dprintk("%s: failed to create NFS4 client.\n", __FUNCTION__
);
1906 /* Now create transport and client */
1907 authflavour
= RPC_AUTH_UNIX
;
1908 if (data
->auth_flavourlen
!= 0) {
1909 if (data
->auth_flavourlen
!= 1) {
1910 dprintk("%s: Invalid number of RPC auth flavours %d.\n",
1911 __FUNCTION__
, data
->auth_flavourlen
);
1915 if (copy_from_user(&authflavour
, data
->auth_flavours
, sizeof(authflavour
))) {
1921 down_write(&clp
->cl_sem
);
1922 if (IS_ERR(clp
->cl_rpcclient
)) {
1923 xprt
= xprt_create_proto(data
->proto
, &server
->addr
, &timeparms
);
1925 up_write(&clp
->cl_sem
);
1926 err
= PTR_ERR(xprt
);
1927 dprintk("%s: cannot create RPC transport. Error = %d\n",
1931 clnt
= rpc_create_client(xprt
, server
->hostname
, &nfs_program
,
1932 server
->rpc_ops
->version
, authflavour
);
1934 up_write(&clp
->cl_sem
);
1935 err
= PTR_ERR(clnt
);
1936 dprintk("%s: cannot create RPC client. Error = %d\n",
1941 clnt
->cl_softrtry
= 1;
1942 clp
->cl_rpcclient
= clnt
;
1943 memcpy(clp
->cl_ipaddr
, server
->ip_addr
, sizeof(clp
->cl_ipaddr
));
1946 list_add_tail(&server
->nfs4_siblings
, &clp
->cl_superblocks
);
1947 clnt
= rpc_clone_client(clp
->cl_rpcclient
);
1949 server
->nfs4_state
= clp
;
1950 up_write(&clp
->cl_sem
);
1954 err
= PTR_ERR(clnt
);
1955 dprintk("%s: cannot create RPC client. Error = %d\n",
1960 server
->client
= clnt
;
1962 if (server
->nfs4_state
->cl_idmap
== NULL
) {
1963 dprintk("%s: failed to create idmapper.\n", __FUNCTION__
);
1967 if (clnt
->cl_auth
->au_flavor
!= authflavour
) {
1968 struct rpc_auth
*auth
;
1970 auth
= rpcauth_create(authflavour
, clnt
);
1972 dprintk("%s: couldn't create credcache!\n", __FUNCTION__
);
1973 return PTR_ERR(auth
);
1977 sb
->s_time_gran
= 1;
1979 sb
->s_op
= &nfs4_sops
;
1980 err
= nfs_sb_init(sb
, authflavour
);
1985 nfs4_put_client(clp
);
1989 static int nfs4_compare_super(struct super_block
*sb
, void *data
)
1991 struct nfs_server
*server
= data
;
1992 struct nfs_server
*old
= NFS_SB(sb
);
1994 if (strcmp(server
->hostname
, old
->hostname
) != 0)
1996 if (strcmp(server
->mnt_path
, old
->mnt_path
) != 0)
2002 nfs_copy_user_string(char *dst
, struct nfs_string
*src
, int maxlen
)
2007 return ERR_PTR(-EINVAL
);
2008 if (src
->len
< maxlen
)
2011 p
= dst
= kmalloc(maxlen
+ 1, GFP_KERNEL
);
2013 return ERR_PTR(-ENOMEM
);
2015 if (copy_from_user(dst
, src
->data
, maxlen
)) {
2017 return ERR_PTR(-EFAULT
);
2023 static struct super_block
*nfs4_get_sb(struct file_system_type
*fs_type
,
2024 int flags
, const char *dev_name
, void *raw_data
)
2027 struct nfs_server
*server
;
2028 struct super_block
*s
;
2029 struct nfs4_mount_data
*data
= raw_data
;
2033 dprintk("%s: missing data argument\n", __FUNCTION__
);
2034 return ERR_PTR(-EINVAL
);
2036 if (data
->version
<= 0 || data
->version
> NFS4_MOUNT_VERSION
) {
2037 dprintk("%s: bad mount version\n", __FUNCTION__
);
2038 return ERR_PTR(-EINVAL
);
2041 server
= kzalloc(sizeof(struct nfs_server
), GFP_KERNEL
);
2043 return ERR_PTR(-ENOMEM
);
2044 /* Zero out the NFS state stuff */
2045 init_nfsv4_state(server
);
2046 server
->client
= server
->client_sys
= server
->client_acl
= ERR_PTR(-EINVAL
);
2048 p
= nfs_copy_user_string(NULL
, &data
->hostname
, 256);
2051 server
->hostname
= p
;
2053 p
= nfs_copy_user_string(NULL
, &data
->mnt_path
, 1024);
2056 server
->mnt_path
= p
;
2058 p
= nfs_copy_user_string(server
->ip_addr
, &data
->client_addr
,
2059 sizeof(server
->ip_addr
) - 1);
2063 /* We now require that the mount process passes the remote address */
2064 if (data
->host_addrlen
!= sizeof(server
->addr
)) {
2065 s
= ERR_PTR(-EINVAL
);
2068 if (copy_from_user(&server
->addr
, data
->host_addr
, sizeof(server
->addr
))) {
2069 s
= ERR_PTR(-EFAULT
);
2072 if (server
->addr
.sin_family
!= AF_INET
||
2073 server
->addr
.sin_addr
.s_addr
== INADDR_ANY
) {
2074 dprintk("%s: mount program didn't pass remote IP address!\n",
2076 s
= ERR_PTR(-EINVAL
);
2080 /* Fire up rpciod if not yet running */
2081 s
= ERR_PTR(rpciod_up());
2083 dprintk("%s: couldn't start rpciod! Error = %ld\n",
2084 __FUNCTION__
, PTR_ERR(s
));
2088 s
= sget(fs_type
, nfs4_compare_super
, nfs_set_super
, server
);
2090 if (IS_ERR(s
) || s
->s_root
)
2095 error
= nfs4_fill_super(s
, data
, flags
& MS_VERBOSE
? 1 : 0);
2097 up_write(&s
->s_umount
);
2098 deactivate_super(s
);
2099 return ERR_PTR(error
);
2101 s
->s_flags
|= MS_ACTIVE
;
2104 s
= (struct super_block
*)p
;
2106 kfree(server
->mnt_path
);
2107 kfree(server
->hostname
);
2108 nfs_free_iostats(server
->io_stats
);
2113 static void nfs4_kill_super(struct super_block
*sb
)
2115 struct nfs_server
*server
= NFS_SB(sb
);
2117 nfs_return_all_delegations(sb
);
2118 kill_anon_super(sb
);
2120 nfs4_renewd_prepare_shutdown(server
);
2122 if (server
->client
!= NULL
&& !IS_ERR(server
->client
))
2123 rpc_shutdown_client(server
->client
);
2125 destroy_nfsv4_state(server
);
2129 kfree(server
->hostname
);
2133 static struct file_system_type nfs4_fs_type
= {
2134 .owner
= THIS_MODULE
,
2136 .get_sb
= nfs4_get_sb
,
2137 .kill_sb
= nfs4_kill_super
,
2138 .fs_flags
= FS_ODD_RENAME
|FS_REVAL_DOT
|FS_BINARY_MOUNTDATA
,
2141 static const int nfs_set_port_min
= 0;
2142 static const int nfs_set_port_max
= 65535;
2143 static int param_set_port(const char *val
, struct kernel_param
*kp
)
2146 int num
= simple_strtol(val
, &endp
, 0);
2147 if (endp
== val
|| *endp
|| num
< nfs_set_port_min
|| num
> nfs_set_port_max
)
2149 *((int *)kp
->arg
) = num
;
2153 module_param_call(callback_tcpport
, param_set_port
, param_get_int
,
2154 &nfs_callback_set_tcpport
, 0644);
2156 static int param_set_idmap_timeout(const char *val
, struct kernel_param
*kp
)
2159 int num
= simple_strtol(val
, &endp
, 0);
2161 if (endp
== val
|| *endp
|| num
< 0 || jif
< num
)
2163 *((int *)kp
->arg
) = jif
;
2167 module_param_call(idmap_cache_timeout
, param_set_idmap_timeout
, param_get_int
,
2168 &nfs_idmap_cache_timeout
, 0644);
2170 #define nfs4_init_once(nfsi) \
2172 INIT_LIST_HEAD(&(nfsi)->open_states); \
2173 nfsi->delegation = NULL; \
2174 nfsi->delegation_state = 0; \
2175 init_rwsem(&nfsi->rwsem); \
2178 static inline int register_nfs4fs(void)
2182 ret
= nfs_register_sysctl();
2185 ret
= register_filesystem(&nfs4_fs_type
);
2187 nfs_unregister_sysctl();
2191 static inline void unregister_nfs4fs(void)
2193 unregister_filesystem(&nfs4_fs_type
);
2194 nfs_unregister_sysctl();
2197 #define nfs4_init_once(nfsi) \
2199 #define register_nfs4fs() (0)
2200 #define unregister_nfs4fs()
2203 extern int nfs_init_nfspagecache(void);
2204 extern void nfs_destroy_nfspagecache(void);
2205 extern int nfs_init_readpagecache(void);
2206 extern void nfs_destroy_readpagecache(void);
2207 extern int nfs_init_writepagecache(void);
2208 extern void nfs_destroy_writepagecache(void);
2209 #ifdef CONFIG_NFS_DIRECTIO
2210 extern int nfs_init_directcache(void);
2211 extern void nfs_destroy_directcache(void);
2214 static kmem_cache_t
* nfs_inode_cachep
;
2216 static struct inode
*nfs_alloc_inode(struct super_block
*sb
)
2218 struct nfs_inode
*nfsi
;
2219 nfsi
= (struct nfs_inode
*)kmem_cache_alloc(nfs_inode_cachep
, SLAB_KERNEL
);
2223 nfsi
->cache_validity
= 0UL;
2224 nfsi
->cache_change_attribute
= jiffies
;
2225 #ifdef CONFIG_NFS_V3_ACL
2226 nfsi
->acl_access
= ERR_PTR(-EAGAIN
);
2227 nfsi
->acl_default
= ERR_PTR(-EAGAIN
);
2229 #ifdef CONFIG_NFS_V4
2230 nfsi
->nfs4_acl
= NULL
;
2231 #endif /* CONFIG_NFS_V4 */
2232 return &nfsi
->vfs_inode
;
2235 static void nfs_destroy_inode(struct inode
*inode
)
2237 kmem_cache_free(nfs_inode_cachep
, NFS_I(inode
));
2240 static void init_once(void * foo
, kmem_cache_t
* cachep
, unsigned long flags
)
2242 struct nfs_inode
*nfsi
= (struct nfs_inode
*) foo
;
2244 if ((flags
& (SLAB_CTOR_VERIFY
|SLAB_CTOR_CONSTRUCTOR
)) ==
2245 SLAB_CTOR_CONSTRUCTOR
) {
2246 inode_init_once(&nfsi
->vfs_inode
);
2247 spin_lock_init(&nfsi
->req_lock
);
2248 INIT_LIST_HEAD(&nfsi
->dirty
);
2249 INIT_LIST_HEAD(&nfsi
->commit
);
2250 INIT_LIST_HEAD(&nfsi
->open_files
);
2251 INIT_RADIX_TREE(&nfsi
->nfs_page_tree
, GFP_ATOMIC
);
2252 atomic_set(&nfsi
->data_updates
, 0);
2256 nfs4_init_once(nfsi
);
2260 static int nfs_init_inodecache(void)
2262 nfs_inode_cachep
= kmem_cache_create("nfs_inode_cache",
2263 sizeof(struct nfs_inode
),
2264 0, SLAB_RECLAIM_ACCOUNT
,
2266 if (nfs_inode_cachep
== NULL
)
2272 static void nfs_destroy_inodecache(void)
2274 if (kmem_cache_destroy(nfs_inode_cachep
))
2275 printk(KERN_INFO
"nfs_inode_cache: not all structures were freed\n");
2281 static int __init
init_nfs_fs(void)
2285 err
= nfs_init_nfspagecache();
2289 err
= nfs_init_inodecache();
2293 err
= nfs_init_readpagecache();
2297 err
= nfs_init_writepagecache();
2301 #ifdef CONFIG_NFS_DIRECTIO
2302 err
= nfs_init_directcache();
2307 #ifdef CONFIG_PROC_FS
2308 rpc_proc_register(&nfs_rpcstat
);
2310 err
= register_filesystem(&nfs_fs_type
);
2313 if ((err
= register_nfs4fs()) != 0)
2317 #ifdef CONFIG_PROC_FS
2318 rpc_proc_unregister("nfs");
2320 #ifdef CONFIG_NFS_DIRECTIO
2321 nfs_destroy_directcache();
2324 nfs_destroy_writepagecache();
2326 nfs_destroy_readpagecache();
2328 nfs_destroy_inodecache();
2330 nfs_destroy_nfspagecache();
2335 static void __exit
exit_nfs_fs(void)
2337 #ifdef CONFIG_NFS_DIRECTIO
2338 nfs_destroy_directcache();
2340 nfs_destroy_writepagecache();
2341 nfs_destroy_readpagecache();
2342 nfs_destroy_inodecache();
2343 nfs_destroy_nfspagecache();
2344 #ifdef CONFIG_PROC_FS
2345 rpc_proc_unregister("nfs");
2347 unregister_filesystem(&nfs_fs_type
);
2348 unregister_nfs4fs();
2351 /* Not quite true; I just maintain it */
2352 MODULE_AUTHOR("Olaf Kirch <okir@monad.swb.de>");
2353 MODULE_LICENSE("GPL");
2355 module_init(init_nfs_fs
)
2356 module_exit(exit_nfs_fs
)