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/sunrpc/metrics.h>
30 #include <linux/nfs_fs.h>
31 #include <linux/nfs_mount.h>
32 #include <linux/nfs4_mount.h>
33 #include <linux/lockd/bind.h>
34 #include <linux/smp_lock.h>
35 #include <linux/seq_file.h>
36 #include <linux/mount.h>
37 #include <linux/nfs_idmap.h>
38 #include <linux/vfs.h>
40 #include <asm/system.h>
41 #include <asm/uaccess.h>
45 #include "delegation.h"
48 #define NFSDBG_FACILITY NFSDBG_VFS
49 #define NFS_PARANOIA 1
51 /* Maximum number of readahead requests
52 * FIXME: this should really be a sysctl so that users may tune it to suit
53 * their needs. People that do NFS over a slow network, might for
54 * instance want to reduce it to something closer to 1 for improved
55 * interactive response.
57 #define NFS_MAX_READAHEAD (RPC_DEF_SLOT_TABLE - 1)
59 static void nfs_invalidate_inode(struct inode
*);
60 static int nfs_update_inode(struct inode
*, struct nfs_fattr
*);
62 static struct inode
*nfs_alloc_inode(struct super_block
*sb
);
63 static void nfs_destroy_inode(struct inode
*);
64 static int nfs_write_inode(struct inode
*,int);
65 static void nfs_clear_inode(struct inode
*);
66 static void nfs_umount_begin(struct vfsmount
*, int);
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
,
79 .clear_inode
= nfs_clear_inode
,
80 .umount_begin
= nfs_umount_begin
,
81 .show_options
= nfs_show_options
,
82 .show_stats
= nfs_show_stats
,
88 static struct rpc_stat nfs_rpcstat
= {
89 .program
= &nfs_program
91 static struct rpc_version
* nfs_version
[] = {
95 #if defined(CONFIG_NFS_V3)
97 #elif defined(CONFIG_NFS_V4)
100 #if defined(CONFIG_NFS_V4)
105 static struct rpc_program nfs_program
= {
107 .number
= NFS_PROGRAM
,
108 .nrvers
= ARRAY_SIZE(nfs_version
),
109 .version
= nfs_version
,
110 .stats
= &nfs_rpcstat
,
111 .pipe_dir_name
= "/nfs",
114 #ifdef CONFIG_NFS_V3_ACL
115 static struct rpc_stat nfsacl_rpcstat
= { &nfsacl_program
};
116 static struct rpc_version
* nfsacl_version
[] = {
117 [3] = &nfsacl_version3
,
120 struct rpc_program nfsacl_program
= {
122 .number
= NFS_ACL_PROGRAM
,
123 .nrvers
= ARRAY_SIZE(nfsacl_version
),
124 .version
= nfsacl_version
,
125 .stats
= &nfsacl_rpcstat
,
127 #endif /* CONFIG_NFS_V3_ACL */
129 static inline unsigned long
130 nfs_fattr_to_ino_t(struct nfs_fattr
*fattr
)
132 return nfs_fileid_to_ino_t(fattr
->fileid
);
136 nfs_write_inode(struct inode
*inode
, int sync
)
138 int flags
= sync
? FLUSH_SYNC
: 0;
141 ret
= nfs_commit_inode(inode
, flags
);
148 nfs_clear_inode(struct inode
*inode
)
150 struct nfs_inode
*nfsi
= NFS_I(inode
);
151 struct rpc_cred
*cred
;
154 * The following should never happen...
156 BUG_ON(nfs_have_writebacks(inode
));
157 BUG_ON (!list_empty(&nfsi
->open_files
));
158 nfs_zap_acl_cache(inode
);
159 cred
= nfsi
->cache_access
.cred
;
162 BUG_ON(atomic_read(&nfsi
->data_updates
) != 0);
165 static void nfs_umount_begin(struct vfsmount
*vfsmnt
, int flags
)
167 struct nfs_server
*server
;
168 struct rpc_clnt
*rpc
;
170 if (!(flags
& MNT_FORCE
))
172 /* -EIO all pending I/O */
173 server
= NFS_SB(vfsmnt
->mnt_sb
);
174 rpc
= server
->client
;
176 rpc_killall_tasks(rpc
);
177 rpc
= server
->client_acl
;
179 rpc_killall_tasks(rpc
);
183 static inline unsigned long
184 nfs_block_bits(unsigned long bsize
, unsigned char *nrbitsp
)
186 /* make sure blocksize is a power of two */
187 if ((bsize
& (bsize
- 1)) || nrbitsp
) {
188 unsigned char nrbits
;
190 for (nrbits
= 31; nrbits
&& !(bsize
& (1 << nrbits
)); nrbits
--)
201 * Calculate the number of 512byte blocks used.
203 static inline unsigned long
204 nfs_calc_block_size(u64 tsize
)
206 loff_t used
= (tsize
+ 511) >> 9;
207 return (used
> ULONG_MAX
) ? ULONG_MAX
: used
;
211 * Compute and set NFS server blocksize
213 static inline unsigned long
214 nfs_block_size(unsigned long bsize
, unsigned char *nrbitsp
)
216 if (bsize
< NFS_MIN_FILE_IO_SIZE
)
217 bsize
= NFS_DEF_FILE_IO_SIZE
;
218 else if (bsize
>= NFS_MAX_FILE_IO_SIZE
)
219 bsize
= NFS_MAX_FILE_IO_SIZE
;
221 return nfs_block_bits(bsize
, nrbitsp
);
225 * Obtain the root inode of the file system.
227 static struct inode
*
228 nfs_get_root(struct super_block
*sb
, struct nfs_fh
*rootfh
, struct nfs_fsinfo
*fsinfo
)
230 struct nfs_server
*server
= NFS_SB(sb
);
233 error
= server
->rpc_ops
->getroot(server
, rootfh
, fsinfo
);
235 dprintk("nfs_get_root: getattr error = %d\n", -error
);
236 return ERR_PTR(error
);
239 server
->fsid
= fsinfo
->fattr
->fsid
;
240 return nfs_fhget(sb
, rootfh
, fsinfo
->fattr
);
244 * Do NFS version-independent mount processing, and sanity checking
247 nfs_sb_init(struct super_block
*sb
, rpc_authflavor_t authflavor
)
249 struct nfs_server
*server
;
250 struct inode
*root_inode
;
251 struct nfs_fattr fattr
;
252 struct nfs_fsinfo fsinfo
= {
255 struct nfs_pathconf pathinfo
= {
258 int no_root_error
= 0;
259 unsigned long max_rpc_payload
;
261 /* We probably want something more informative here */
262 snprintf(sb
->s_id
, sizeof(sb
->s_id
), "%x:%x", MAJOR(sb
->s_dev
), MINOR(sb
->s_dev
));
266 sb
->s_magic
= NFS_SUPER_MAGIC
;
268 server
->io_stats
= nfs_alloc_iostats();
269 if (server
->io_stats
== NULL
)
272 root_inode
= nfs_get_root(sb
, &server
->fh
, &fsinfo
);
273 /* Did getting the root inode fail? */
274 if (IS_ERR(root_inode
)) {
275 no_root_error
= PTR_ERR(root_inode
);
278 sb
->s_root
= d_alloc_root(root_inode
);
280 no_root_error
= -ENOMEM
;
283 sb
->s_root
->d_op
= server
->rpc_ops
->dentry_ops
;
285 /* mount time stamp, in seconds */
286 server
->mount_time
= jiffies
;
288 /* Get some general file system info */
289 if (server
->namelen
== 0 &&
290 server
->rpc_ops
->pathconf(server
, &server
->fh
, &pathinfo
) >= 0)
291 server
->namelen
= pathinfo
.max_namelen
;
292 /* Work out a lot of parameters */
293 if (server
->rsize
== 0)
294 server
->rsize
= nfs_block_size(fsinfo
.rtpref
, NULL
);
295 if (server
->wsize
== 0)
296 server
->wsize
= nfs_block_size(fsinfo
.wtpref
, NULL
);
298 if (fsinfo
.rtmax
>= 512 && server
->rsize
> fsinfo
.rtmax
)
299 server
->rsize
= nfs_block_size(fsinfo
.rtmax
, NULL
);
300 if (fsinfo
.wtmax
>= 512 && server
->wsize
> fsinfo
.wtmax
)
301 server
->wsize
= nfs_block_size(fsinfo
.wtmax
, NULL
);
303 max_rpc_payload
= nfs_block_size(rpc_max_payload(server
->client
), NULL
);
304 if (server
->rsize
> max_rpc_payload
)
305 server
->rsize
= max_rpc_payload
;
306 if (server
->rsize
> NFS_MAX_FILE_IO_SIZE
)
307 server
->rsize
= NFS_MAX_FILE_IO_SIZE
;
308 server
->rpages
= (server
->rsize
+ PAGE_CACHE_SIZE
- 1) >> PAGE_CACHE_SHIFT
;
310 if (server
->wsize
> max_rpc_payload
)
311 server
->wsize
= max_rpc_payload
;
312 if (server
->wsize
> NFS_MAX_FILE_IO_SIZE
)
313 server
->wsize
= NFS_MAX_FILE_IO_SIZE
;
314 server
->wpages
= (server
->wsize
+ PAGE_CACHE_SIZE
- 1) >> PAGE_CACHE_SHIFT
;
316 if (sb
->s_blocksize
== 0)
317 sb
->s_blocksize
= nfs_block_bits(server
->wsize
,
318 &sb
->s_blocksize_bits
);
319 server
->wtmult
= nfs_block_bits(fsinfo
.wtmult
, NULL
);
321 server
->dtsize
= nfs_block_size(fsinfo
.dtpref
, NULL
);
322 if (server
->dtsize
> PAGE_CACHE_SIZE
)
323 server
->dtsize
= PAGE_CACHE_SIZE
;
324 if (server
->dtsize
> server
->rsize
)
325 server
->dtsize
= server
->rsize
;
327 if (server
->flags
& NFS_MOUNT_NOAC
) {
328 server
->acregmin
= server
->acregmax
= 0;
329 server
->acdirmin
= server
->acdirmax
= 0;
330 sb
->s_flags
|= MS_SYNCHRONOUS
;
332 server
->backing_dev_info
.ra_pages
= server
->rpages
* NFS_MAX_READAHEAD
;
334 sb
->s_maxbytes
= fsinfo
.maxfilesize
;
335 if (sb
->s_maxbytes
> MAX_LFS_FILESIZE
)
336 sb
->s_maxbytes
= MAX_LFS_FILESIZE
;
338 server
->client
->cl_intr
= (server
->flags
& NFS_MOUNT_INTR
) ? 1 : 0;
339 server
->client
->cl_softrtry
= (server
->flags
& NFS_MOUNT_SOFT
) ? 1 : 0;
341 /* We're airborne Set socket buffersize */
342 rpc_setbufsize(server
->client
, server
->wsize
+ 100, server
->rsize
+ 100);
344 /* Yargs. It didn't work out. */
346 dprintk("nfs_sb_init: get root inode failed: errno %d\n", -no_root_error
);
347 if (!IS_ERR(root_inode
))
349 return no_root_error
;
352 static void nfs_init_timeout_values(struct rpc_timeout
*to
, int proto
, unsigned int timeo
, unsigned int retrans
)
354 to
->to_initval
= timeo
* HZ
/ 10;
355 to
->to_retries
= retrans
;
362 to
->to_initval
= 60 * HZ
;
363 if (to
->to_initval
> NFS_MAX_TCP_TIMEOUT
)
364 to
->to_initval
= NFS_MAX_TCP_TIMEOUT
;
365 to
->to_increment
= to
->to_initval
;
366 to
->to_maxval
= to
->to_initval
+ (to
->to_increment
* to
->to_retries
);
367 to
->to_exponential
= 0;
372 to
->to_initval
= 11 * HZ
/ 10;
373 if (to
->to_initval
> NFS_MAX_UDP_TIMEOUT
)
374 to
->to_initval
= NFS_MAX_UDP_TIMEOUT
;
375 to
->to_maxval
= NFS_MAX_UDP_TIMEOUT
;
376 to
->to_exponential
= 1;
382 * Create an RPC client handle.
384 static struct rpc_clnt
*
385 nfs_create_client(struct nfs_server
*server
, const struct nfs_mount_data
*data
)
387 struct rpc_timeout timeparms
;
388 struct rpc_xprt
*xprt
= NULL
;
389 struct rpc_clnt
*clnt
= NULL
;
390 int proto
= (data
->flags
& NFS_MOUNT_TCP
) ? IPPROTO_TCP
: IPPROTO_UDP
;
392 nfs_init_timeout_values(&timeparms
, proto
, data
->timeo
, data
->retrans
);
394 server
->retrans_timeo
= timeparms
.to_initval
;
395 server
->retrans_count
= timeparms
.to_retries
;
397 /* create transport and client */
398 xprt
= xprt_create_proto(proto
, &server
->addr
, &timeparms
);
400 dprintk("%s: cannot create RPC transport. Error = %ld\n",
401 __FUNCTION__
, PTR_ERR(xprt
));
402 return (struct rpc_clnt
*)xprt
;
404 clnt
= rpc_create_client(xprt
, server
->hostname
, &nfs_program
,
405 server
->rpc_ops
->version
, data
->pseudoflavor
);
407 dprintk("%s: cannot create RPC client. Error = %ld\n",
408 __FUNCTION__
, PTR_ERR(xprt
));
413 clnt
->cl_softrtry
= 1;
422 * The way this works is that the mount process passes a structure
423 * in the data argument which contains the server's IP address
424 * and the root file handle obtained from the server's mount
425 * daemon. We stash these away in the private superblock fields.
428 nfs_fill_super(struct super_block
*sb
, struct nfs_mount_data
*data
, int silent
)
430 struct nfs_server
*server
;
431 rpc_authflavor_t authflavor
;
434 sb
->s_blocksize_bits
= 0;
437 sb
->s_blocksize
= nfs_block_size(data
->bsize
, &sb
->s_blocksize_bits
);
439 server
->rsize
= nfs_block_size(data
->rsize
, NULL
);
441 server
->wsize
= nfs_block_size(data
->wsize
, NULL
);
442 server
->flags
= data
->flags
& NFS_MOUNT_FLAGMASK
;
444 server
->acregmin
= data
->acregmin
*HZ
;
445 server
->acregmax
= data
->acregmax
*HZ
;
446 server
->acdirmin
= data
->acdirmin
*HZ
;
447 server
->acdirmax
= data
->acdirmax
*HZ
;
449 /* Start lockd here, before we might error out */
450 if (!(server
->flags
& NFS_MOUNT_NONLM
))
453 server
->namelen
= data
->namlen
;
454 server
->hostname
= kmalloc(strlen(data
->hostname
) + 1, GFP_KERNEL
);
455 if (!server
->hostname
)
457 strcpy(server
->hostname
, data
->hostname
);
459 /* Check NFS protocol revision and initialize RPC op vector
460 * and file handle pool. */
462 if (server
->flags
& NFS_MOUNT_VER3
) {
463 server
->rpc_ops
= &nfs_v3_clientops
;
464 server
->caps
|= NFS_CAP_READDIRPLUS
;
466 server
->rpc_ops
= &nfs_v2_clientops
;
469 server
->rpc_ops
= &nfs_v2_clientops
;
472 /* Fill in pseudoflavor for mount version < 5 */
473 if (!(data
->flags
& NFS_MOUNT_SECFLAVOUR
))
474 data
->pseudoflavor
= RPC_AUTH_UNIX
;
475 authflavor
= data
->pseudoflavor
; /* save for sb_init() */
476 /* XXX maybe we want to add a server->pseudoflavor field */
478 /* Create RPC client handles */
479 server
->client
= nfs_create_client(server
, data
);
480 if (IS_ERR(server
->client
))
481 return PTR_ERR(server
->client
);
482 /* RFC 2623, sec 2.3.2 */
483 if (authflavor
!= RPC_AUTH_UNIX
) {
484 struct rpc_auth
*auth
;
486 server
->client_sys
= rpc_clone_client(server
->client
);
487 if (IS_ERR(server
->client_sys
))
488 return PTR_ERR(server
->client_sys
);
489 auth
= rpcauth_create(RPC_AUTH_UNIX
, server
->client_sys
);
491 return PTR_ERR(auth
);
493 atomic_inc(&server
->client
->cl_count
);
494 server
->client_sys
= server
->client
;
496 if (server
->flags
& NFS_MOUNT_VER3
) {
497 #ifdef CONFIG_NFS_V3_ACL
498 if (!(server
->flags
& NFS_MOUNT_NOACL
)) {
499 server
->client_acl
= rpc_bind_new_program(server
->client
, &nfsacl_program
, 3);
500 /* No errors! Assume that Sun nfsacls are supported */
501 if (!IS_ERR(server
->client_acl
))
502 server
->caps
|= NFS_CAP_ACLS
;
505 server
->flags
&= ~NFS_MOUNT_NOACL
;
506 #endif /* CONFIG_NFS_V3_ACL */
508 * The VFS shouldn't apply the umask to mode bits. We will
509 * do so ourselves when necessary.
511 sb
->s_flags
|= MS_POSIXACL
;
512 if (server
->namelen
== 0 || server
->namelen
> NFS3_MAXNAMLEN
)
513 server
->namelen
= NFS3_MAXNAMLEN
;
516 if (server
->namelen
== 0 || server
->namelen
> NFS2_MAXNAMLEN
)
517 server
->namelen
= NFS2_MAXNAMLEN
;
520 sb
->s_op
= &nfs_sops
;
521 return nfs_sb_init(sb
, authflavor
);
525 nfs_statfs(struct super_block
*sb
, struct kstatfs
*buf
)
527 struct nfs_server
*server
= NFS_SB(sb
);
528 unsigned char blockbits
;
529 unsigned long blockres
;
530 struct nfs_fh
*rootfh
= NFS_FH(sb
->s_root
->d_inode
);
531 struct nfs_fattr fattr
;
532 struct nfs_fsstat res
= {
539 error
= server
->rpc_ops
->statfs(server
, rootfh
, &res
);
540 buf
->f_type
= NFS_SUPER_MAGIC
;
545 * Current versions of glibc do not correctly handle the
546 * case where f_frsize != f_bsize. Eventually we want to
547 * report the value of wtmult in this field.
549 buf
->f_frsize
= sb
->s_blocksize
;
552 * On most *nix systems, f_blocks, f_bfree, and f_bavail
553 * are reported in units of f_frsize. Linux hasn't had
554 * an f_frsize field in its statfs struct until recently,
555 * thus historically Linux's sys_statfs reports these
556 * fields in units of f_bsize.
558 buf
->f_bsize
= sb
->s_blocksize
;
559 blockbits
= sb
->s_blocksize_bits
;
560 blockres
= (1 << blockbits
) - 1;
561 buf
->f_blocks
= (res
.tbytes
+ blockres
) >> blockbits
;
562 buf
->f_bfree
= (res
.fbytes
+ blockres
) >> blockbits
;
563 buf
->f_bavail
= (res
.abytes
+ blockres
) >> blockbits
;
565 buf
->f_files
= res
.tfiles
;
566 buf
->f_ffree
= res
.afiles
;
568 buf
->f_namelen
= server
->namelen
;
574 dprintk("%s: statfs error = %d\n", __FUNCTION__
, -error
);
575 buf
->f_bsize
= buf
->f_blocks
= buf
->f_bfree
= buf
->f_bavail
= -1;
580 static void nfs_show_mount_options(struct seq_file
*m
, struct nfs_server
*nfss
, int showdefaults
)
582 static struct proc_nfs_info
{
587 { NFS_MOUNT_SOFT
, ",soft", ",hard" },
588 { NFS_MOUNT_INTR
, ",intr", "" },
589 { NFS_MOUNT_NOCTO
, ",nocto", "" },
590 { NFS_MOUNT_NOAC
, ",noac", "" },
591 { NFS_MOUNT_NONLM
, ",nolock", "" },
592 { NFS_MOUNT_NOACL
, ",noacl", "" },
595 struct proc_nfs_info
*nfs_infop
;
599 seq_printf(m
, ",vers=%d", nfss
->rpc_ops
->version
);
600 seq_printf(m
, ",rsize=%d", nfss
->rsize
);
601 seq_printf(m
, ",wsize=%d", nfss
->wsize
);
602 if (nfss
->acregmin
!= 3*HZ
|| showdefaults
)
603 seq_printf(m
, ",acregmin=%d", nfss
->acregmin
/HZ
);
604 if (nfss
->acregmax
!= 60*HZ
|| showdefaults
)
605 seq_printf(m
, ",acregmax=%d", nfss
->acregmax
/HZ
);
606 if (nfss
->acdirmin
!= 30*HZ
|| showdefaults
)
607 seq_printf(m
, ",acdirmin=%d", nfss
->acdirmin
/HZ
);
608 if (nfss
->acdirmax
!= 60*HZ
|| showdefaults
)
609 seq_printf(m
, ",acdirmax=%d", nfss
->acdirmax
/HZ
);
610 for (nfs_infop
= nfs_info
; nfs_infop
->flag
; nfs_infop
++) {
611 if (nfss
->flags
& nfs_infop
->flag
)
612 seq_puts(m
, nfs_infop
->str
);
614 seq_puts(m
, nfs_infop
->nostr
);
616 switch (nfss
->client
->cl_xprt
->prot
) {
624 snprintf(buf
, sizeof(buf
), "%u", nfss
->client
->cl_xprt
->prot
);
627 seq_printf(m
, ",proto=%s", proto
);
628 seq_printf(m
, ",timeo=%lu", 10U * nfss
->retrans_timeo
/ HZ
);
629 seq_printf(m
, ",retrans=%u", nfss
->retrans_count
);
632 static int nfs_show_options(struct seq_file
*m
, struct vfsmount
*mnt
)
634 struct nfs_server
*nfss
= NFS_SB(mnt
->mnt_sb
);
636 nfs_show_mount_options(m
, nfss
, 0);
638 seq_puts(m
, ",addr=");
639 seq_escape(m
, nfss
->hostname
, " \t\n\\");
644 static int nfs_show_stats(struct seq_file
*m
, struct vfsmount
*mnt
)
647 struct nfs_server
*nfss
= NFS_SB(mnt
->mnt_sb
);
648 struct rpc_auth
*auth
= nfss
->client
->cl_auth
;
649 struct nfs_iostats totals
= { };
651 seq_printf(m
, "statvers=%s", NFS_IOSTAT_VERS
);
654 * Display all mount option settings
656 seq_printf(m
, "\n\topts:\t");
657 seq_puts(m
, mnt
->mnt_sb
->s_flags
& MS_RDONLY
? "ro" : "rw");
658 seq_puts(m
, mnt
->mnt_sb
->s_flags
& MS_SYNCHRONOUS
? ",sync" : "");
659 seq_puts(m
, mnt
->mnt_sb
->s_flags
& MS_NOATIME
? ",noatime" : "");
660 seq_puts(m
, mnt
->mnt_sb
->s_flags
& MS_NODIRATIME
? ",nodiratime" : "");
661 nfs_show_mount_options(m
, nfss
, 1);
663 seq_printf(m
, "\n\tage:\t%lu", (jiffies
- nfss
->mount_time
) / HZ
);
665 seq_printf(m
, "\n\tcaps:\t");
666 seq_printf(m
, "caps=0x%x", nfss
->caps
);
667 seq_printf(m
, ",wtmult=%d", nfss
->wtmult
);
668 seq_printf(m
, ",dtsize=%d", nfss
->dtsize
);
669 seq_printf(m
, ",bsize=%d", nfss
->bsize
);
670 seq_printf(m
, ",namelen=%d", nfss
->namelen
);
673 if (nfss
->rpc_ops
->version
== 4) {
674 seq_printf(m
, "\n\tnfsv4:\t");
675 seq_printf(m
, "bm0=0x%x", nfss
->attr_bitmask
[0]);
676 seq_printf(m
, ",bm1=0x%x", nfss
->attr_bitmask
[1]);
677 seq_printf(m
, ",acl=0x%x", nfss
->acl_bitmask
);
682 * Display security flavor in effect for this mount
684 seq_printf(m
, "\n\tsec:\tflavor=%d", auth
->au_ops
->au_flavor
);
686 seq_printf(m
, ",pseudoflavor=%d", auth
->au_flavor
);
689 * Display superblock I/O counters
691 for_each_possible_cpu(cpu
) {
692 struct nfs_iostats
*stats
;
695 stats
= per_cpu_ptr(nfss
->io_stats
, cpu
);
697 for (i
= 0; i
< __NFSIOS_COUNTSMAX
; i
++)
698 totals
.events
[i
] += stats
->events
[i
];
699 for (i
= 0; i
< __NFSIOS_BYTESMAX
; i
++)
700 totals
.bytes
[i
] += stats
->bytes
[i
];
705 seq_printf(m
, "\n\tevents:\t");
706 for (i
= 0; i
< __NFSIOS_COUNTSMAX
; i
++)
707 seq_printf(m
, "%lu ", totals
.events
[i
]);
708 seq_printf(m
, "\n\tbytes:\t");
709 for (i
= 0; i
< __NFSIOS_BYTESMAX
; i
++)
710 seq_printf(m
, "%Lu ", totals
.bytes
[i
]);
713 rpc_print_iostats(m
, nfss
->client
);
719 * nfs_sync_mapping - helper to flush all mmapped dirty data to disk
721 int nfs_sync_mapping(struct address_space
*mapping
)
725 if (mapping
->nrpages
== 0)
727 unmap_mapping_range(mapping
, 0, 0, 0);
728 ret
= filemap_write_and_wait(mapping
);
731 ret
= nfs_wb_all(mapping
->host
);
737 * Invalidate the local caches
739 static void nfs_zap_caches_locked(struct inode
*inode
)
741 struct nfs_inode
*nfsi
= NFS_I(inode
);
742 int mode
= inode
->i_mode
;
744 nfs_inc_stats(inode
, NFSIOS_ATTRINVALIDATE
);
746 NFS_ATTRTIMEO(inode
) = NFS_MINATTRTIMEO(inode
);
747 NFS_ATTRTIMEO_UPDATE(inode
) = jiffies
;
749 memset(NFS_COOKIEVERF(inode
), 0, sizeof(NFS_COOKIEVERF(inode
)));
750 if (S_ISREG(mode
) || S_ISDIR(mode
) || S_ISLNK(mode
))
751 nfsi
->cache_validity
|= NFS_INO_INVALID_ATTR
|NFS_INO_INVALID_DATA
|NFS_INO_INVALID_ACCESS
|NFS_INO_INVALID_ACL
|NFS_INO_REVAL_PAGECACHE
;
753 nfsi
->cache_validity
|= NFS_INO_INVALID_ATTR
|NFS_INO_INVALID_ACCESS
|NFS_INO_INVALID_ACL
|NFS_INO_REVAL_PAGECACHE
;
756 void nfs_zap_caches(struct inode
*inode
)
758 spin_lock(&inode
->i_lock
);
759 nfs_zap_caches_locked(inode
);
760 spin_unlock(&inode
->i_lock
);
763 static void nfs_zap_acl_cache(struct inode
*inode
)
765 void (*clear_acl_cache
)(struct inode
*);
767 clear_acl_cache
= NFS_PROTO(inode
)->clear_acl_cache
;
768 if (clear_acl_cache
!= NULL
)
769 clear_acl_cache(inode
);
770 spin_lock(&inode
->i_lock
);
771 NFS_I(inode
)->cache_validity
&= ~NFS_INO_INVALID_ACL
;
772 spin_unlock(&inode
->i_lock
);
776 * Invalidate, but do not unhash, the inode.
777 * NB: must be called with inode->i_lock held!
779 static void nfs_invalidate_inode(struct inode
*inode
)
781 set_bit(NFS_INO_STALE
, &NFS_FLAGS(inode
));
782 nfs_zap_caches_locked(inode
);
785 struct nfs_find_desc
{
787 struct nfs_fattr
*fattr
;
791 * In NFSv3 we can have 64bit inode numbers. In order to support
792 * this, and re-exported directories (also seen in NFSv2)
793 * we are forced to allow 2 different inodes to have the same
797 nfs_find_actor(struct inode
*inode
, void *opaque
)
799 struct nfs_find_desc
*desc
= (struct nfs_find_desc
*)opaque
;
800 struct nfs_fh
*fh
= desc
->fh
;
801 struct nfs_fattr
*fattr
= desc
->fattr
;
803 if (NFS_FILEID(inode
) != fattr
->fileid
)
805 if (nfs_compare_fh(NFS_FH(inode
), fh
))
807 if (is_bad_inode(inode
) || NFS_STALE(inode
))
813 nfs_init_locked(struct inode
*inode
, void *opaque
)
815 struct nfs_find_desc
*desc
= (struct nfs_find_desc
*)opaque
;
816 struct nfs_fattr
*fattr
= desc
->fattr
;
818 NFS_FILEID(inode
) = fattr
->fileid
;
819 nfs_copy_fh(NFS_FH(inode
), desc
->fh
);
823 /* Don't use READDIRPLUS on directories that we believe are too large */
824 #define NFS_LIMIT_READDIRPLUS (8*PAGE_SIZE)
827 * This is our front-end to iget that looks up inodes by file handle
828 * instead of inode number.
831 nfs_fhget(struct super_block
*sb
, struct nfs_fh
*fh
, struct nfs_fattr
*fattr
)
833 struct nfs_find_desc desc
= {
837 struct inode
*inode
= ERR_PTR(-ENOENT
);
840 if ((fattr
->valid
& NFS_ATTR_FATTR
) == 0)
844 printk("NFS: Buggy server - nlink == 0!\n");
848 hash
= nfs_fattr_to_ino_t(fattr
);
850 inode
= iget5_locked(sb
, hash
, nfs_find_actor
, nfs_init_locked
, &desc
);
852 inode
= ERR_PTR(-ENOMEM
);
856 if (inode
->i_state
& I_NEW
) {
857 struct nfs_inode
*nfsi
= NFS_I(inode
);
859 /* We set i_ino for the few things that still rely on it,
863 /* We can't support update_atime(), since the server will reset it */
864 inode
->i_flags
|= S_NOATIME
|S_NOCMTIME
;
865 inode
->i_mode
= fattr
->mode
;
866 /* Why so? Because we want revalidate for devices/FIFOs, and
867 * that's precisely what we have in nfs_file_inode_operations.
869 inode
->i_op
= NFS_SB(sb
)->rpc_ops
->file_inode_ops
;
870 if (S_ISREG(inode
->i_mode
)) {
871 inode
->i_fop
= &nfs_file_operations
;
872 inode
->i_data
.a_ops
= &nfs_file_aops
;
873 inode
->i_data
.backing_dev_info
= &NFS_SB(sb
)->backing_dev_info
;
874 } else if (S_ISDIR(inode
->i_mode
)) {
875 inode
->i_op
= NFS_SB(sb
)->rpc_ops
->dir_inode_ops
;
876 inode
->i_fop
= &nfs_dir_operations
;
877 if (nfs_server_capable(inode
, NFS_CAP_READDIRPLUS
)
878 && fattr
->size
<= NFS_LIMIT_READDIRPLUS
)
879 set_bit(NFS_INO_ADVISE_RDPLUS
, &NFS_FLAGS(inode
));
880 } else if (S_ISLNK(inode
->i_mode
))
881 inode
->i_op
= &nfs_symlink_inode_operations
;
883 init_special_inode(inode
, inode
->i_mode
, fattr
->rdev
);
885 nfsi
->read_cache_jiffies
= fattr
->time_start
;
886 nfsi
->last_updated
= jiffies
;
887 inode
->i_atime
= fattr
->atime
;
888 inode
->i_mtime
= fattr
->mtime
;
889 inode
->i_ctime
= fattr
->ctime
;
890 if (fattr
->valid
& NFS_ATTR_FATTR_V4
)
891 nfsi
->change_attr
= fattr
->change_attr
;
892 inode
->i_size
= nfs_size_to_loff_t(fattr
->size
);
893 inode
->i_nlink
= fattr
->nlink
;
894 inode
->i_uid
= fattr
->uid
;
895 inode
->i_gid
= fattr
->gid
;
896 if (fattr
->valid
& (NFS_ATTR_FATTR_V3
| NFS_ATTR_FATTR_V4
)) {
898 * report the blocks in 512byte units
900 inode
->i_blocks
= nfs_calc_block_size(fattr
->du
.nfs3
.used
);
901 inode
->i_blksize
= inode
->i_sb
->s_blocksize
;
903 inode
->i_blocks
= fattr
->du
.nfs2
.blocks
;
904 inode
->i_blksize
= fattr
->du
.nfs2
.blocksize
;
906 nfsi
->attrtimeo
= NFS_MINATTRTIMEO(inode
);
907 nfsi
->attrtimeo_timestamp
= jiffies
;
908 memset(nfsi
->cookieverf
, 0, sizeof(nfsi
->cookieverf
));
909 nfsi
->cache_access
.cred
= NULL
;
911 unlock_new_inode(inode
);
913 nfs_refresh_inode(inode
, fattr
);
914 dprintk("NFS: nfs_fhget(%s/%Ld ct=%d)\n",
916 (long long)NFS_FILEID(inode
),
917 atomic_read(&inode
->i_count
));
923 dprintk("nfs_fhget: iget failed with error %ld\n", PTR_ERR(inode
));
927 #define NFS_VALID_ATTRS (ATTR_MODE|ATTR_UID|ATTR_GID|ATTR_SIZE|ATTR_ATIME|ATTR_ATIME_SET|ATTR_MTIME|ATTR_MTIME_SET)
930 nfs_setattr(struct dentry
*dentry
, struct iattr
*attr
)
932 struct inode
*inode
= dentry
->d_inode
;
933 struct nfs_fattr fattr
;
936 nfs_inc_stats(inode
, NFSIOS_VFSSETATTR
);
938 if (attr
->ia_valid
& ATTR_SIZE
) {
939 if (!S_ISREG(inode
->i_mode
) || attr
->ia_size
== i_size_read(inode
))
940 attr
->ia_valid
&= ~ATTR_SIZE
;
943 /* Optimization: if the end result is no change, don't RPC */
944 attr
->ia_valid
&= NFS_VALID_ATTRS
;
945 if (attr
->ia_valid
== 0)
949 nfs_begin_data_update(inode
);
950 /* Write all dirty data */
951 filemap_write_and_wait(inode
->i_mapping
);
954 * Return any delegations if we're going to change ACLs
956 if ((attr
->ia_valid
& (ATTR_MODE
|ATTR_UID
|ATTR_GID
)) != 0)
957 nfs_inode_return_delegation(inode
);
958 error
= NFS_PROTO(inode
)->setattr(dentry
, &fattr
, attr
);
960 nfs_refresh_inode(inode
, &fattr
);
961 nfs_end_data_update(inode
);
967 * nfs_setattr_update_inode - Update inode metadata after a setattr call.
968 * @inode: pointer to struct inode
969 * @attr: pointer to struct iattr
971 * Note: we do this in the *proc.c in order to ensure that
972 * it works for things like exclusive creates too.
974 void nfs_setattr_update_inode(struct inode
*inode
, struct iattr
*attr
)
976 if ((attr
->ia_valid
& (ATTR_MODE
|ATTR_UID
|ATTR_GID
)) != 0) {
977 if ((attr
->ia_valid
& ATTR_MODE
) != 0) {
978 int mode
= attr
->ia_mode
& S_IALLUGO
;
979 mode
|= inode
->i_mode
& ~S_IALLUGO
;
980 inode
->i_mode
= mode
;
982 if ((attr
->ia_valid
& ATTR_UID
) != 0)
983 inode
->i_uid
= attr
->ia_uid
;
984 if ((attr
->ia_valid
& ATTR_GID
) != 0)
985 inode
->i_gid
= attr
->ia_gid
;
986 spin_lock(&inode
->i_lock
);
987 NFS_I(inode
)->cache_validity
|= NFS_INO_INVALID_ACCESS
|NFS_INO_INVALID_ACL
;
988 spin_unlock(&inode
->i_lock
);
990 if ((attr
->ia_valid
& ATTR_SIZE
) != 0) {
991 nfs_inc_stats(inode
, NFSIOS_SETATTRTRUNC
);
992 inode
->i_size
= attr
->ia_size
;
993 vmtruncate(inode
, attr
->ia_size
);
997 static int nfs_wait_schedule(void *word
)
999 if (signal_pending(current
))
1000 return -ERESTARTSYS
;
1006 * Wait for the inode to get unlocked.
1008 static int nfs_wait_on_inode(struct inode
*inode
)
1010 struct rpc_clnt
*clnt
= NFS_CLIENT(inode
);
1011 struct nfs_inode
*nfsi
= NFS_I(inode
);
1015 rpc_clnt_sigmask(clnt
, &oldmask
);
1016 error
= wait_on_bit_lock(&nfsi
->flags
, NFS_INO_REVALIDATING
,
1017 nfs_wait_schedule
, TASK_INTERRUPTIBLE
);
1018 rpc_clnt_sigunmask(clnt
, &oldmask
);
1023 static void nfs_wake_up_inode(struct inode
*inode
)
1025 struct nfs_inode
*nfsi
= NFS_I(inode
);
1027 clear_bit(NFS_INO_REVALIDATING
, &nfsi
->flags
);
1028 smp_mb__after_clear_bit();
1029 wake_up_bit(&nfsi
->flags
, NFS_INO_REVALIDATING
);
1032 int nfs_getattr(struct vfsmount
*mnt
, struct dentry
*dentry
, struct kstat
*stat
)
1034 struct inode
*inode
= dentry
->d_inode
;
1035 int need_atime
= NFS_I(inode
)->cache_validity
& NFS_INO_INVALID_ATIME
;
1038 /* Flush out writes to the server in order to update c/mtime */
1039 nfs_sync_inode_wait(inode
, 0, 0, FLUSH_NOCOMMIT
);
1042 * We may force a getattr if the user cares about atime.
1044 * Note that we only have to check the vfsmount flags here:
1045 * - NFS always sets S_NOATIME by so checking it would give a
1047 * - NFS never sets MS_NOATIME or MS_NODIRATIME so there is
1048 * no point in checking those.
1050 if ((mnt
->mnt_flags
& MNT_NOATIME
) ||
1051 ((mnt
->mnt_flags
& MNT_NODIRATIME
) && S_ISDIR(inode
->i_mode
)))
1055 err
= __nfs_revalidate_inode(NFS_SERVER(inode
), inode
);
1057 err
= nfs_revalidate_inode(NFS_SERVER(inode
), inode
);
1059 generic_fillattr(inode
, stat
);
1063 static struct nfs_open_context
*alloc_nfs_open_context(struct vfsmount
*mnt
, struct dentry
*dentry
, struct rpc_cred
*cred
)
1065 struct nfs_open_context
*ctx
;
1067 ctx
= (struct nfs_open_context
*)kmalloc(sizeof(*ctx
), GFP_KERNEL
);
1069 atomic_set(&ctx
->count
, 1);
1070 ctx
->dentry
= dget(dentry
);
1071 ctx
->vfsmnt
= mntget(mnt
);
1072 ctx
->cred
= get_rpccred(cred
);
1074 ctx
->lockowner
= current
->files
;
1076 ctx
->dir_cookie
= 0;
1081 struct nfs_open_context
*get_nfs_open_context(struct nfs_open_context
*ctx
)
1084 atomic_inc(&ctx
->count
);
1088 void put_nfs_open_context(struct nfs_open_context
*ctx
)
1090 if (atomic_dec_and_test(&ctx
->count
)) {
1091 if (!list_empty(&ctx
->list
)) {
1092 struct inode
*inode
= ctx
->dentry
->d_inode
;
1093 spin_lock(&inode
->i_lock
);
1094 list_del(&ctx
->list
);
1095 spin_unlock(&inode
->i_lock
);
1097 if (ctx
->state
!= NULL
)
1098 nfs4_close_state(ctx
->state
, ctx
->mode
);
1099 if (ctx
->cred
!= NULL
)
1100 put_rpccred(ctx
->cred
);
1102 mntput(ctx
->vfsmnt
);
1108 * Ensure that mmap has a recent RPC credential for use when writing out
1111 static void nfs_file_set_open_context(struct file
*filp
, struct nfs_open_context
*ctx
)
1113 struct inode
*inode
= filp
->f_dentry
->d_inode
;
1114 struct nfs_inode
*nfsi
= NFS_I(inode
);
1116 filp
->private_data
= get_nfs_open_context(ctx
);
1117 spin_lock(&inode
->i_lock
);
1118 list_add(&ctx
->list
, &nfsi
->open_files
);
1119 spin_unlock(&inode
->i_lock
);
1123 * Given an inode, search for an open context with the desired characteristics
1125 struct nfs_open_context
*nfs_find_open_context(struct inode
*inode
, struct rpc_cred
*cred
, int mode
)
1127 struct nfs_inode
*nfsi
= NFS_I(inode
);
1128 struct nfs_open_context
*pos
, *ctx
= NULL
;
1130 spin_lock(&inode
->i_lock
);
1131 list_for_each_entry(pos
, &nfsi
->open_files
, list
) {
1132 if (cred
!= NULL
&& pos
->cred
!= cred
)
1134 if ((pos
->mode
& mode
) == mode
) {
1135 ctx
= get_nfs_open_context(pos
);
1139 spin_unlock(&inode
->i_lock
);
1143 static void nfs_file_clear_open_context(struct file
*filp
)
1145 struct inode
*inode
= filp
->f_dentry
->d_inode
;
1146 struct nfs_open_context
*ctx
= (struct nfs_open_context
*)filp
->private_data
;
1149 filp
->private_data
= NULL
;
1150 spin_lock(&inode
->i_lock
);
1151 list_move_tail(&ctx
->list
, &NFS_I(inode
)->open_files
);
1152 spin_unlock(&inode
->i_lock
);
1153 put_nfs_open_context(ctx
);
1158 * These allocate and release file read/write context information.
1160 int nfs_open(struct inode
*inode
, struct file
*filp
)
1162 struct nfs_open_context
*ctx
;
1163 struct rpc_cred
*cred
;
1165 cred
= rpcauth_lookupcred(NFS_CLIENT(inode
)->cl_auth
, 0);
1167 return PTR_ERR(cred
);
1168 ctx
= alloc_nfs_open_context(filp
->f_vfsmnt
, filp
->f_dentry
, cred
);
1172 ctx
->mode
= filp
->f_mode
;
1173 nfs_file_set_open_context(filp
, ctx
);
1174 put_nfs_open_context(ctx
);
1178 int nfs_release(struct inode
*inode
, struct file
*filp
)
1180 nfs_file_clear_open_context(filp
);
1185 * This function is called whenever some part of NFS notices that
1186 * the cached attributes have to be refreshed.
1189 __nfs_revalidate_inode(struct nfs_server
*server
, struct inode
*inode
)
1191 int status
= -ESTALE
;
1192 struct nfs_fattr fattr
;
1193 struct nfs_inode
*nfsi
= NFS_I(inode
);
1195 dfprintk(PAGECACHE
, "NFS: revalidating (%s/%Ld)\n",
1196 inode
->i_sb
->s_id
, (long long)NFS_FILEID(inode
));
1198 nfs_inc_stats(inode
, NFSIOS_INODEREVALIDATE
);
1200 if (!inode
|| is_bad_inode(inode
))
1202 if (NFS_STALE(inode
))
1205 status
= nfs_wait_on_inode(inode
);
1208 if (NFS_STALE(inode
)) {
1210 /* Do we trust the cached ESTALE? */
1211 if (NFS_ATTRTIMEO(inode
) != 0) {
1212 if (nfsi
->cache_validity
& (NFS_INO_INVALID_ATTR
|NFS_INO_INVALID_ATIME
)) {
1219 status
= NFS_PROTO(inode
)->getattr(server
, NFS_FH(inode
), &fattr
);
1221 dfprintk(PAGECACHE
, "nfs_revalidate_inode: (%s/%Ld) getattr failed, error=%d\n",
1223 (long long)NFS_FILEID(inode
), status
);
1224 if (status
== -ESTALE
) {
1225 nfs_zap_caches(inode
);
1226 if (!S_ISDIR(inode
->i_mode
))
1227 set_bit(NFS_INO_STALE
, &NFS_FLAGS(inode
));
1232 spin_lock(&inode
->i_lock
);
1233 status
= nfs_update_inode(inode
, &fattr
);
1235 spin_unlock(&inode
->i_lock
);
1236 dfprintk(PAGECACHE
, "nfs_revalidate_inode: (%s/%Ld) refresh failed, error=%d\n",
1238 (long long)NFS_FILEID(inode
), status
);
1241 spin_unlock(&inode
->i_lock
);
1243 if (nfsi
->cache_validity
& NFS_INO_INVALID_ACL
)
1244 nfs_zap_acl_cache(inode
);
1246 dfprintk(PAGECACHE
, "NFS: (%s/%Ld) revalidation complete\n",
1248 (long long)NFS_FILEID(inode
));
1251 nfs_wake_up_inode(inode
);
1258 int nfs_attribute_timeout(struct inode
*inode
)
1260 struct nfs_inode
*nfsi
= NFS_I(inode
);
1262 if (nfs_have_delegation(inode
, FMODE_READ
))
1264 return time_after(jiffies
, nfsi
->read_cache_jiffies
+nfsi
->attrtimeo
);
1268 * nfs_revalidate_inode - Revalidate the inode attributes
1269 * @server - pointer to nfs_server struct
1270 * @inode - pointer to inode struct
1272 * Updates inode attribute information by retrieving the data from the server.
1274 int nfs_revalidate_inode(struct nfs_server
*server
, struct inode
*inode
)
1276 if (!(NFS_I(inode
)->cache_validity
& NFS_INO_INVALID_ATTR
)
1277 && !nfs_attribute_timeout(inode
))
1278 return NFS_STALE(inode
) ? -ESTALE
: 0;
1279 return __nfs_revalidate_inode(server
, inode
);
1283 * nfs_revalidate_mapping - Revalidate the pagecache
1284 * @inode - pointer to host inode
1285 * @mapping - pointer to mapping
1287 int nfs_revalidate_mapping(struct inode
*inode
, struct address_space
*mapping
)
1289 struct nfs_inode
*nfsi
= NFS_I(inode
);
1292 if (NFS_STALE(inode
))
1294 if ((nfsi
->cache_validity
& NFS_INO_REVAL_PAGECACHE
)
1295 || nfs_attribute_timeout(inode
))
1296 ret
= __nfs_revalidate_inode(NFS_SERVER(inode
), inode
);
1298 if (nfsi
->cache_validity
& NFS_INO_INVALID_DATA
) {
1299 nfs_inc_stats(inode
, NFSIOS_DATAINVALIDATE
);
1300 if (S_ISREG(inode
->i_mode
))
1301 nfs_sync_mapping(mapping
);
1302 invalidate_inode_pages2(mapping
);
1304 spin_lock(&inode
->i_lock
);
1305 nfsi
->cache_validity
&= ~NFS_INO_INVALID_DATA
;
1306 if (S_ISDIR(inode
->i_mode
)) {
1307 memset(nfsi
->cookieverf
, 0, sizeof(nfsi
->cookieverf
));
1308 /* This ensures we revalidate child dentries */
1309 nfsi
->cache_change_attribute
= jiffies
;
1311 spin_unlock(&inode
->i_lock
);
1313 dfprintk(PAGECACHE
, "NFS: (%s/%Ld) data cache invalidated\n",
1315 (long long)NFS_FILEID(inode
));
1321 * nfs_begin_data_update
1322 * @inode - pointer to inode
1323 * Declare that a set of operations will update file data on the server
1325 void nfs_begin_data_update(struct inode
*inode
)
1327 atomic_inc(&NFS_I(inode
)->data_updates
);
1331 * nfs_end_data_update
1332 * @inode - pointer to inode
1333 * Declare end of the operations that will update file data
1334 * This will mark the inode as immediately needing revalidation
1335 * of its attribute cache.
1337 void nfs_end_data_update(struct inode
*inode
)
1339 struct nfs_inode
*nfsi
= NFS_I(inode
);
1341 if (!nfs_have_delegation(inode
, FMODE_READ
)) {
1342 /* Directories and symlinks: invalidate page cache */
1343 if (S_ISDIR(inode
->i_mode
) || S_ISLNK(inode
->i_mode
)) {
1344 spin_lock(&inode
->i_lock
);
1345 nfsi
->cache_validity
|= NFS_INO_INVALID_DATA
;
1346 spin_unlock(&inode
->i_lock
);
1349 nfsi
->cache_change_attribute
= jiffies
;
1350 atomic_dec(&nfsi
->data_updates
);
1353 static void nfs_wcc_update_inode(struct inode
*inode
, struct nfs_fattr
*fattr
)
1355 struct nfs_inode
*nfsi
= NFS_I(inode
);
1357 /* If we have atomic WCC data, we may update some attributes */
1358 if ((fattr
->valid
& NFS_ATTR_WCC
) != 0) {
1359 if (timespec_equal(&inode
->i_ctime
, &fattr
->pre_ctime
)) {
1360 memcpy(&inode
->i_ctime
, &fattr
->ctime
, sizeof(inode
->i_ctime
));
1361 nfsi
->cache_change_attribute
= jiffies
;
1363 if (timespec_equal(&inode
->i_mtime
, &fattr
->pre_mtime
)) {
1364 memcpy(&inode
->i_mtime
, &fattr
->mtime
, sizeof(inode
->i_mtime
));
1365 nfsi
->cache_change_attribute
= jiffies
;
1367 if (inode
->i_size
== fattr
->pre_size
&& nfsi
->npages
== 0) {
1368 inode
->i_size
= fattr
->size
;
1369 nfsi
->cache_change_attribute
= jiffies
;
1375 * nfs_check_inode_attributes - verify consistency of the inode attribute cache
1376 * @inode - pointer to inode
1377 * @fattr - updated attributes
1379 * Verifies the attribute cache. If we have just changed the attributes,
1380 * so that fattr carries weak cache consistency data, then it may
1381 * also update the ctime/mtime/change_attribute.
1383 static int nfs_check_inode_attributes(struct inode
*inode
, struct nfs_fattr
*fattr
)
1385 struct nfs_inode
*nfsi
= NFS_I(inode
);
1386 loff_t cur_size
, new_isize
;
1390 /* Has the inode gone and changed behind our back? */
1391 if (nfsi
->fileid
!= fattr
->fileid
1392 || (inode
->i_mode
& S_IFMT
) != (fattr
->mode
& S_IFMT
)) {
1396 /* Are we in the process of updating data on the server? */
1397 data_unstable
= nfs_caches_unstable(inode
);
1399 /* Do atomic weak cache consistency updates */
1400 nfs_wcc_update_inode(inode
, fattr
);
1402 if ((fattr
->valid
& NFS_ATTR_FATTR_V4
) != 0 &&
1403 nfsi
->change_attr
!= fattr
->change_attr
)
1404 nfsi
->cache_validity
|= NFS_INO_INVALID_ATTR
|NFS_INO_REVAL_PAGECACHE
;
1406 /* Verify a few of the more important attributes */
1407 if (!timespec_equal(&inode
->i_mtime
, &fattr
->mtime
))
1408 nfsi
->cache_validity
|= NFS_INO_INVALID_ATTR
|NFS_INO_REVAL_PAGECACHE
;
1410 cur_size
= i_size_read(inode
);
1411 new_isize
= nfs_size_to_loff_t(fattr
->size
);
1412 if (cur_size
!= new_isize
&& nfsi
->npages
== 0)
1413 nfsi
->cache_validity
|= NFS_INO_INVALID_ATTR
|NFS_INO_REVAL_PAGECACHE
;
1415 /* Have any file permissions changed? */
1416 if ((inode
->i_mode
& S_IALLUGO
) != (fattr
->mode
& S_IALLUGO
)
1417 || inode
->i_uid
!= fattr
->uid
1418 || inode
->i_gid
!= fattr
->gid
)
1419 nfsi
->cache_validity
|= NFS_INO_INVALID_ATTR
| NFS_INO_INVALID_ACCESS
| NFS_INO_INVALID_ACL
;
1421 /* Has the link count changed? */
1422 if (inode
->i_nlink
!= fattr
->nlink
)
1423 nfsi
->cache_validity
|= NFS_INO_INVALID_ATTR
;
1425 if (!timespec_equal(&inode
->i_atime
, &fattr
->atime
))
1426 nfsi
->cache_validity
|= NFS_INO_INVALID_ATIME
;
1428 nfsi
->read_cache_jiffies
= fattr
->time_start
;
1433 * nfs_refresh_inode - try to update the inode attribute cache
1434 * @inode - pointer to inode
1435 * @fattr - updated attributes
1437 * Check that an RPC call that returned attributes has not overlapped with
1438 * other recent updates of the inode metadata, then decide whether it is
1439 * safe to do a full update of the inode attributes, or whether just to
1440 * call nfs_check_inode_attributes.
1442 int nfs_refresh_inode(struct inode
*inode
, struct nfs_fattr
*fattr
)
1444 struct nfs_inode
*nfsi
= NFS_I(inode
);
1447 if ((fattr
->valid
& NFS_ATTR_FATTR
) == 0)
1449 spin_lock(&inode
->i_lock
);
1450 if (time_after(fattr
->time_start
, nfsi
->last_updated
))
1451 status
= nfs_update_inode(inode
, fattr
);
1453 status
= nfs_check_inode_attributes(inode
, fattr
);
1455 spin_unlock(&inode
->i_lock
);
1460 * nfs_post_op_update_inode - try to update the inode attribute cache
1461 * @inode - pointer to inode
1462 * @fattr - updated attributes
1464 * After an operation that has changed the inode metadata, mark the
1465 * attribute cache as being invalid, then try to update it.
1467 int nfs_post_op_update_inode(struct inode
*inode
, struct nfs_fattr
*fattr
)
1469 struct nfs_inode
*nfsi
= NFS_I(inode
);
1472 spin_lock(&inode
->i_lock
);
1473 if (unlikely((fattr
->valid
& NFS_ATTR_FATTR
) == 0)) {
1474 nfsi
->cache_validity
|= NFS_INO_INVALID_ACCESS
|NFS_INO_INVALID_ATTR
|NFS_INO_REVAL_PAGECACHE
;
1477 status
= nfs_update_inode(inode
, fattr
);
1479 spin_unlock(&inode
->i_lock
);
1484 * Many nfs protocol calls return the new file attributes after
1485 * an operation. Here we update the inode to reflect the state
1486 * of the server's inode.
1488 * This is a bit tricky because we have to make sure all dirty pages
1489 * have been sent off to the server before calling invalidate_inode_pages.
1490 * To make sure no other process adds more write requests while we try
1491 * our best to flush them, we make them sleep during the attribute refresh.
1493 * A very similar scenario holds for the dir cache.
1495 static int nfs_update_inode(struct inode
*inode
, struct nfs_fattr
*fattr
)
1497 struct nfs_server
*server
;
1498 struct nfs_inode
*nfsi
= NFS_I(inode
);
1499 loff_t cur_isize
, new_isize
;
1500 unsigned int invalid
= 0;
1503 dfprintk(VFS
, "NFS: %s(%s/%ld ct=%d info=0x%x)\n",
1504 __FUNCTION__
, inode
->i_sb
->s_id
, inode
->i_ino
,
1505 atomic_read(&inode
->i_count
), fattr
->valid
);
1507 if (nfsi
->fileid
!= fattr
->fileid
)
1511 * Make sure the inode's type hasn't changed.
1513 if ((inode
->i_mode
& S_IFMT
) != (fattr
->mode
& S_IFMT
))
1516 server
= NFS_SERVER(inode
);
1517 /* Update the fsid if and only if this is the root directory */
1518 if (inode
== inode
->i_sb
->s_root
->d_inode
1519 && !nfs_fsid_equal(&server
->fsid
, &fattr
->fsid
))
1520 server
->fsid
= fattr
->fsid
;
1523 * Update the read time so we don't revalidate too often.
1525 nfsi
->read_cache_jiffies
= fattr
->time_start
;
1526 nfsi
->last_updated
= jiffies
;
1528 /* Are we racing with known updates of the metadata on the server? */
1529 data_stable
= nfs_verify_change_attribute(inode
, fattr
->time_start
);
1531 nfsi
->cache_validity
&= ~(NFS_INO_INVALID_ATTR
|NFS_INO_REVAL_PAGECACHE
|NFS_INO_INVALID_ATIME
);
1533 /* Do atomic weak cache consistency updates */
1534 nfs_wcc_update_inode(inode
, fattr
);
1536 /* Check if our cached file size is stale */
1537 new_isize
= nfs_size_to_loff_t(fattr
->size
);
1538 cur_isize
= i_size_read(inode
);
1539 if (new_isize
!= cur_isize
) {
1540 /* Do we perhaps have any outstanding writes? */
1541 if (nfsi
->npages
== 0) {
1542 /* No, but did we race with nfs_end_data_update()? */
1544 inode
->i_size
= new_isize
;
1545 invalid
|= NFS_INO_INVALID_DATA
;
1547 invalid
|= NFS_INO_INVALID_ATTR
;
1548 } else if (new_isize
> cur_isize
) {
1549 inode
->i_size
= new_isize
;
1550 invalid
|= NFS_INO_INVALID_ATTR
|NFS_INO_INVALID_DATA
;
1552 nfsi
->cache_change_attribute
= jiffies
;
1553 dprintk("NFS: isize change on server for file %s/%ld\n",
1554 inode
->i_sb
->s_id
, inode
->i_ino
);
1557 /* Check if the mtime agrees */
1558 if (!timespec_equal(&inode
->i_mtime
, &fattr
->mtime
)) {
1559 memcpy(&inode
->i_mtime
, &fattr
->mtime
, sizeof(inode
->i_mtime
));
1560 dprintk("NFS: mtime change on server for file %s/%ld\n",
1561 inode
->i_sb
->s_id
, inode
->i_ino
);
1562 invalid
|= NFS_INO_INVALID_ATTR
|NFS_INO_INVALID_DATA
;
1563 nfsi
->cache_change_attribute
= jiffies
;
1566 /* If ctime has changed we should definitely clear access+acl caches */
1567 if (!timespec_equal(&inode
->i_ctime
, &fattr
->ctime
)) {
1568 invalid
|= NFS_INO_INVALID_ACCESS
|NFS_INO_INVALID_ACL
;
1569 memcpy(&inode
->i_ctime
, &fattr
->ctime
, sizeof(inode
->i_ctime
));
1570 nfsi
->cache_change_attribute
= jiffies
;
1572 memcpy(&inode
->i_atime
, &fattr
->atime
, sizeof(inode
->i_atime
));
1574 if ((inode
->i_mode
& S_IALLUGO
) != (fattr
->mode
& S_IALLUGO
) ||
1575 inode
->i_uid
!= fattr
->uid
||
1576 inode
->i_gid
!= fattr
->gid
)
1577 invalid
|= NFS_INO_INVALID_ATTR
|NFS_INO_INVALID_ACCESS
|NFS_INO_INVALID_ACL
;
1579 inode
->i_mode
= fattr
->mode
;
1580 inode
->i_nlink
= fattr
->nlink
;
1581 inode
->i_uid
= fattr
->uid
;
1582 inode
->i_gid
= fattr
->gid
;
1584 if (fattr
->valid
& (NFS_ATTR_FATTR_V3
| NFS_ATTR_FATTR_V4
)) {
1586 * report the blocks in 512byte units
1588 inode
->i_blocks
= nfs_calc_block_size(fattr
->du
.nfs3
.used
);
1589 inode
->i_blksize
= inode
->i_sb
->s_blocksize
;
1591 inode
->i_blocks
= fattr
->du
.nfs2
.blocks
;
1592 inode
->i_blksize
= fattr
->du
.nfs2
.blocksize
;
1595 if ((fattr
->valid
& NFS_ATTR_FATTR_V4
) != 0 &&
1596 nfsi
->change_attr
!= fattr
->change_attr
) {
1597 dprintk("NFS: change_attr change on server for file %s/%ld\n",
1598 inode
->i_sb
->s_id
, inode
->i_ino
);
1599 nfsi
->change_attr
= fattr
->change_attr
;
1600 invalid
|= NFS_INO_INVALID_ATTR
|NFS_INO_INVALID_DATA
|NFS_INO_INVALID_ACCESS
|NFS_INO_INVALID_ACL
;
1601 nfsi
->cache_change_attribute
= jiffies
;
1604 /* Update attrtimeo value if we're out of the unstable period */
1605 if (invalid
& NFS_INO_INVALID_ATTR
) {
1606 nfs_inc_stats(inode
, NFSIOS_ATTRINVALIDATE
);
1607 nfsi
->attrtimeo
= NFS_MINATTRTIMEO(inode
);
1608 nfsi
->attrtimeo_timestamp
= jiffies
;
1609 } else if (time_after(jiffies
, nfsi
->attrtimeo_timestamp
+nfsi
->attrtimeo
)) {
1610 if ((nfsi
->attrtimeo
<<= 1) > NFS_MAXATTRTIMEO(inode
))
1611 nfsi
->attrtimeo
= NFS_MAXATTRTIMEO(inode
);
1612 nfsi
->attrtimeo_timestamp
= jiffies
;
1614 /* Don't invalidate the data if we were to blame */
1615 if (!(S_ISREG(inode
->i_mode
) || S_ISDIR(inode
->i_mode
)
1616 || S_ISLNK(inode
->i_mode
)))
1617 invalid
&= ~NFS_INO_INVALID_DATA
;
1619 invalid
&= ~(NFS_INO_INVALID_ATTR
|NFS_INO_INVALID_ATIME
|NFS_INO_REVAL_PAGECACHE
);
1620 if (!nfs_have_delegation(inode
, FMODE_READ
))
1621 nfsi
->cache_validity
|= invalid
;
1626 * Big trouble! The inode has become a different object.
1629 printk(KERN_DEBUG
"%s: inode %ld mode changed, %07o to %07o\n",
1630 __FUNCTION__
, inode
->i_ino
, inode
->i_mode
, fattr
->mode
);
1634 * No need to worry about unhashing the dentry, as the
1635 * lookup validation will know that the inode is bad.
1636 * (But we fall through to invalidate the caches.)
1638 nfs_invalidate_inode(inode
);
1642 printk(KERN_ERR
"NFS: server %s error: fileid changed\n"
1643 "fsid %s: expected fileid 0x%Lx, got 0x%Lx\n",
1644 NFS_SERVER(inode
)->hostname
, inode
->i_sb
->s_id
,
1645 (long long)nfsi
->fileid
, (long long)fattr
->fileid
);
1650 * File system information
1653 static int nfs_set_super(struct super_block
*s
, void *data
)
1655 s
->s_fs_info
= data
;
1656 return set_anon_super(s
, data
);
1659 static int nfs_compare_super(struct super_block
*sb
, void *data
)
1661 struct nfs_server
*server
= data
;
1662 struct nfs_server
*old
= NFS_SB(sb
);
1664 if (old
->addr
.sin_addr
.s_addr
!= server
->addr
.sin_addr
.s_addr
)
1666 if (old
->addr
.sin_port
!= server
->addr
.sin_port
)
1668 return !nfs_compare_fh(&old
->fh
, &server
->fh
);
1671 static struct super_block
*nfs_get_sb(struct file_system_type
*fs_type
,
1672 int flags
, const char *dev_name
, void *raw_data
)
1675 struct nfs_server
*server
= NULL
;
1676 struct super_block
*s
;
1677 struct nfs_fh
*root
;
1678 struct nfs_mount_data
*data
= raw_data
;
1680 s
= ERR_PTR(-EINVAL
);
1682 dprintk("%s: missing data argument\n", __FUNCTION__
);
1685 if (data
->version
<= 0 || data
->version
> NFS_MOUNT_VERSION
) {
1686 dprintk("%s: bad mount version\n", __FUNCTION__
);
1689 switch (data
->version
) {
1695 if (data
->flags
& NFS_MOUNT_VER3
) {
1696 dprintk("%s: mount structure version %d does not support NFSv3\n",
1701 data
->root
.size
= NFS2_FHSIZE
;
1702 memcpy(data
->root
.data
, data
->old_root
.data
, NFS2_FHSIZE
);
1704 if (data
->flags
& NFS_MOUNT_SECFLAVOUR
) {
1705 dprintk("%s: mount structure version %d does not support strong security\n",
1711 memset(data
->context
, 0, sizeof(data
->context
));
1713 #ifndef CONFIG_NFS_V3
1714 /* If NFSv3 is not compiled in, return -EPROTONOSUPPORT */
1715 s
= ERR_PTR(-EPROTONOSUPPORT
);
1716 if (data
->flags
& NFS_MOUNT_VER3
) {
1717 dprintk("%s: NFSv3 not compiled into kernel\n", __FUNCTION__
);
1720 #endif /* CONFIG_NFS_V3 */
1722 s
= ERR_PTR(-ENOMEM
);
1723 server
= kzalloc(sizeof(struct nfs_server
), GFP_KERNEL
);
1726 /* Zero out the NFS state stuff */
1727 init_nfsv4_state(server
);
1728 server
->client
= server
->client_sys
= server
->client_acl
= ERR_PTR(-EINVAL
);
1731 if (data
->flags
& NFS_MOUNT_VER3
)
1732 root
->size
= data
->root
.size
;
1734 root
->size
= NFS2_FHSIZE
;
1735 s
= ERR_PTR(-EINVAL
);
1736 if (root
->size
> sizeof(root
->data
)) {
1737 dprintk("%s: invalid root filehandle\n", __FUNCTION__
);
1740 memcpy(root
->data
, data
->root
.data
, root
->size
);
1742 /* We now require that the mount process passes the remote address */
1743 memcpy(&server
->addr
, &data
->addr
, sizeof(server
->addr
));
1744 if (server
->addr
.sin_addr
.s_addr
== INADDR_ANY
) {
1745 dprintk("%s: mount program didn't pass remote address!\n",
1750 /* Fire up rpciod if not yet running */
1751 s
= ERR_PTR(rpciod_up());
1753 dprintk("%s: couldn't start rpciod! Error = %ld\n",
1754 __FUNCTION__
, PTR_ERR(s
));
1758 s
= sget(fs_type
, nfs_compare_super
, nfs_set_super
, server
);
1759 if (IS_ERR(s
) || s
->s_root
)
1760 goto out_rpciod_down
;
1764 error
= nfs_fill_super(s
, data
, flags
& MS_SILENT
? 1 : 0);
1766 up_write(&s
->s_umount
);
1767 deactivate_super(s
);
1768 return ERR_PTR(error
);
1770 s
->s_flags
|= MS_ACTIVE
;
1779 static void nfs_kill_super(struct super_block
*s
)
1781 struct nfs_server
*server
= NFS_SB(s
);
1785 if (!IS_ERR(server
->client
))
1786 rpc_shutdown_client(server
->client
);
1787 if (!IS_ERR(server
->client_sys
))
1788 rpc_shutdown_client(server
->client_sys
);
1789 if (!IS_ERR(server
->client_acl
))
1790 rpc_shutdown_client(server
->client_acl
);
1792 if (!(server
->flags
& NFS_MOUNT_NONLM
))
1793 lockd_down(); /* release rpc.lockd */
1795 rpciod_down(); /* release rpciod */
1797 nfs_free_iostats(server
->io_stats
);
1798 kfree(server
->hostname
);
1802 static struct file_system_type nfs_fs_type
= {
1803 .owner
= THIS_MODULE
,
1805 .get_sb
= nfs_get_sb
,
1806 .kill_sb
= nfs_kill_super
,
1807 .fs_flags
= FS_ODD_RENAME
|FS_REVAL_DOT
|FS_BINARY_MOUNTDATA
,
1810 #ifdef CONFIG_NFS_V4
1812 static void nfs4_clear_inode(struct inode
*);
1815 static struct super_operations nfs4_sops
= {
1816 .alloc_inode
= nfs_alloc_inode
,
1817 .destroy_inode
= nfs_destroy_inode
,
1818 .write_inode
= nfs_write_inode
,
1819 .statfs
= nfs_statfs
,
1820 .clear_inode
= nfs4_clear_inode
,
1821 .umount_begin
= nfs_umount_begin
,
1822 .show_options
= nfs_show_options
,
1823 .show_stats
= nfs_show_stats
,
1827 * Clean out any remaining NFSv4 state that might be left over due
1828 * to open() calls that passed nfs_atomic_lookup, but failed to call
1831 static void nfs4_clear_inode(struct inode
*inode
)
1833 struct nfs_inode
*nfsi
= NFS_I(inode
);
1835 /* If we are holding a delegation, return it! */
1836 nfs_inode_return_delegation(inode
);
1837 /* First call standard NFS clear_inode() code */
1838 nfs_clear_inode(inode
);
1839 /* Now clear out any remaining state */
1840 while (!list_empty(&nfsi
->open_states
)) {
1841 struct nfs4_state
*state
;
1843 state
= list_entry(nfsi
->open_states
.next
,
1846 dprintk("%s(%s/%Ld): found unclaimed NFSv4 state %p\n",
1849 (long long)NFS_FILEID(inode
),
1851 BUG_ON(atomic_read(&state
->count
) != 1);
1852 nfs4_close_state(state
, state
->state
);
1857 static int nfs4_fill_super(struct super_block
*sb
, struct nfs4_mount_data
*data
, int silent
)
1859 struct nfs_server
*server
;
1860 struct nfs4_client
*clp
= NULL
;
1861 struct rpc_xprt
*xprt
= NULL
;
1862 struct rpc_clnt
*clnt
= NULL
;
1863 struct rpc_timeout timeparms
;
1864 rpc_authflavor_t authflavour
;
1867 sb
->s_blocksize_bits
= 0;
1868 sb
->s_blocksize
= 0;
1869 server
= NFS_SB(sb
);
1870 if (data
->rsize
!= 0)
1871 server
->rsize
= nfs_block_size(data
->rsize
, NULL
);
1872 if (data
->wsize
!= 0)
1873 server
->wsize
= nfs_block_size(data
->wsize
, NULL
);
1874 server
->flags
= data
->flags
& NFS_MOUNT_FLAGMASK
;
1875 server
->caps
= NFS_CAP_ATOMIC_OPEN
;
1877 server
->acregmin
= data
->acregmin
*HZ
;
1878 server
->acregmax
= data
->acregmax
*HZ
;
1879 server
->acdirmin
= data
->acdirmin
*HZ
;
1880 server
->acdirmax
= data
->acdirmax
*HZ
;
1882 server
->rpc_ops
= &nfs_v4_clientops
;
1884 nfs_init_timeout_values(&timeparms
, data
->proto
, data
->timeo
, data
->retrans
);
1886 server
->retrans_timeo
= timeparms
.to_initval
;
1887 server
->retrans_count
= timeparms
.to_retries
;
1889 clp
= nfs4_get_client(&server
->addr
.sin_addr
);
1891 dprintk("%s: failed to create NFS4 client.\n", __FUNCTION__
);
1895 /* Now create transport and client */
1896 authflavour
= RPC_AUTH_UNIX
;
1897 if (data
->auth_flavourlen
!= 0) {
1898 if (data
->auth_flavourlen
!= 1) {
1899 dprintk("%s: Invalid number of RPC auth flavours %d.\n",
1900 __FUNCTION__
, data
->auth_flavourlen
);
1904 if (copy_from_user(&authflavour
, data
->auth_flavours
, sizeof(authflavour
))) {
1910 down_write(&clp
->cl_sem
);
1911 if (IS_ERR(clp
->cl_rpcclient
)) {
1912 xprt
= xprt_create_proto(data
->proto
, &server
->addr
, &timeparms
);
1914 up_write(&clp
->cl_sem
);
1915 err
= PTR_ERR(xprt
);
1916 dprintk("%s: cannot create RPC transport. Error = %d\n",
1920 clnt
= rpc_create_client(xprt
, server
->hostname
, &nfs_program
,
1921 server
->rpc_ops
->version
, authflavour
);
1923 up_write(&clp
->cl_sem
);
1924 err
= PTR_ERR(clnt
);
1925 dprintk("%s: cannot create RPC client. Error = %d\n",
1930 clnt
->cl_softrtry
= 1;
1931 clp
->cl_rpcclient
= clnt
;
1932 memcpy(clp
->cl_ipaddr
, server
->ip_addr
, sizeof(clp
->cl_ipaddr
));
1935 list_add_tail(&server
->nfs4_siblings
, &clp
->cl_superblocks
);
1936 clnt
= rpc_clone_client(clp
->cl_rpcclient
);
1938 server
->nfs4_state
= clp
;
1939 up_write(&clp
->cl_sem
);
1943 err
= PTR_ERR(clnt
);
1944 dprintk("%s: cannot create RPC client. Error = %d\n",
1949 server
->client
= clnt
;
1951 if (server
->nfs4_state
->cl_idmap
== NULL
) {
1952 dprintk("%s: failed to create idmapper.\n", __FUNCTION__
);
1956 if (clnt
->cl_auth
->au_flavor
!= authflavour
) {
1957 struct rpc_auth
*auth
;
1959 auth
= rpcauth_create(authflavour
, clnt
);
1961 dprintk("%s: couldn't create credcache!\n", __FUNCTION__
);
1962 return PTR_ERR(auth
);
1966 sb
->s_time_gran
= 1;
1968 sb
->s_op
= &nfs4_sops
;
1969 err
= nfs_sb_init(sb
, authflavour
);
1974 nfs4_put_client(clp
);
1978 static int nfs4_compare_super(struct super_block
*sb
, void *data
)
1980 struct nfs_server
*server
= data
;
1981 struct nfs_server
*old
= NFS_SB(sb
);
1983 if (strcmp(server
->hostname
, old
->hostname
) != 0)
1985 if (strcmp(server
->mnt_path
, old
->mnt_path
) != 0)
1991 nfs_copy_user_string(char *dst
, struct nfs_string
*src
, int maxlen
)
1996 return ERR_PTR(-EINVAL
);
1997 if (src
->len
< maxlen
)
2000 p
= dst
= kmalloc(maxlen
+ 1, GFP_KERNEL
);
2002 return ERR_PTR(-ENOMEM
);
2004 if (copy_from_user(dst
, src
->data
, maxlen
)) {
2006 return ERR_PTR(-EFAULT
);
2012 static struct super_block
*nfs4_get_sb(struct file_system_type
*fs_type
,
2013 int flags
, const char *dev_name
, void *raw_data
)
2016 struct nfs_server
*server
;
2017 struct super_block
*s
;
2018 struct nfs4_mount_data
*data
= raw_data
;
2022 dprintk("%s: missing data argument\n", __FUNCTION__
);
2023 return ERR_PTR(-EINVAL
);
2025 if (data
->version
<= 0 || data
->version
> NFS4_MOUNT_VERSION
) {
2026 dprintk("%s: bad mount version\n", __FUNCTION__
);
2027 return ERR_PTR(-EINVAL
);
2030 server
= kzalloc(sizeof(struct nfs_server
), GFP_KERNEL
);
2032 return ERR_PTR(-ENOMEM
);
2033 /* Zero out the NFS state stuff */
2034 init_nfsv4_state(server
);
2035 server
->client
= server
->client_sys
= server
->client_acl
= ERR_PTR(-EINVAL
);
2037 p
= nfs_copy_user_string(NULL
, &data
->hostname
, 256);
2040 server
->hostname
= p
;
2042 p
= nfs_copy_user_string(NULL
, &data
->mnt_path
, 1024);
2045 server
->mnt_path
= p
;
2047 p
= nfs_copy_user_string(server
->ip_addr
, &data
->client_addr
,
2048 sizeof(server
->ip_addr
) - 1);
2052 /* We now require that the mount process passes the remote address */
2053 if (data
->host_addrlen
!= sizeof(server
->addr
)) {
2054 s
= ERR_PTR(-EINVAL
);
2057 if (copy_from_user(&server
->addr
, data
->host_addr
, sizeof(server
->addr
))) {
2058 s
= ERR_PTR(-EFAULT
);
2061 if (server
->addr
.sin_family
!= AF_INET
||
2062 server
->addr
.sin_addr
.s_addr
== INADDR_ANY
) {
2063 dprintk("%s: mount program didn't pass remote IP address!\n",
2065 s
= ERR_PTR(-EINVAL
);
2069 /* Fire up rpciod if not yet running */
2070 s
= ERR_PTR(rpciod_up());
2072 dprintk("%s: couldn't start rpciod! Error = %ld\n",
2073 __FUNCTION__
, PTR_ERR(s
));
2077 s
= sget(fs_type
, nfs4_compare_super
, nfs_set_super
, server
);
2079 if (IS_ERR(s
) || s
->s_root
)
2084 error
= nfs4_fill_super(s
, data
, flags
& MS_SILENT
? 1 : 0);
2086 up_write(&s
->s_umount
);
2087 deactivate_super(s
);
2088 return ERR_PTR(error
);
2090 s
->s_flags
|= MS_ACTIVE
;
2093 s
= (struct super_block
*)p
;
2095 kfree(server
->mnt_path
);
2096 kfree(server
->hostname
);
2101 static void nfs4_kill_super(struct super_block
*sb
)
2103 struct nfs_server
*server
= NFS_SB(sb
);
2105 nfs_return_all_delegations(sb
);
2106 kill_anon_super(sb
);
2108 nfs4_renewd_prepare_shutdown(server
);
2110 if (server
->client
!= NULL
&& !IS_ERR(server
->client
))
2111 rpc_shutdown_client(server
->client
);
2113 destroy_nfsv4_state(server
);
2117 nfs_free_iostats(server
->io_stats
);
2118 kfree(server
->hostname
);
2122 static struct file_system_type nfs4_fs_type
= {
2123 .owner
= THIS_MODULE
,
2125 .get_sb
= nfs4_get_sb
,
2126 .kill_sb
= nfs4_kill_super
,
2127 .fs_flags
= FS_ODD_RENAME
|FS_REVAL_DOT
|FS_BINARY_MOUNTDATA
,
2130 static const int nfs_set_port_min
= 0;
2131 static const int nfs_set_port_max
= 65535;
2132 static int param_set_port(const char *val
, struct kernel_param
*kp
)
2135 int num
= simple_strtol(val
, &endp
, 0);
2136 if (endp
== val
|| *endp
|| num
< nfs_set_port_min
|| num
> nfs_set_port_max
)
2138 *((int *)kp
->arg
) = num
;
2142 module_param_call(callback_tcpport
, param_set_port
, param_get_int
,
2143 &nfs_callback_set_tcpport
, 0644);
2145 static int param_set_idmap_timeout(const char *val
, struct kernel_param
*kp
)
2148 int num
= simple_strtol(val
, &endp
, 0);
2150 if (endp
== val
|| *endp
|| num
< 0 || jif
< num
)
2152 *((int *)kp
->arg
) = jif
;
2156 module_param_call(idmap_cache_timeout
, param_set_idmap_timeout
, param_get_int
,
2157 &nfs_idmap_cache_timeout
, 0644);
2159 #define nfs4_init_once(nfsi) \
2161 INIT_LIST_HEAD(&(nfsi)->open_states); \
2162 nfsi->delegation = NULL; \
2163 nfsi->delegation_state = 0; \
2164 init_rwsem(&nfsi->rwsem); \
2167 static inline int register_nfs4fs(void)
2171 ret
= nfs_register_sysctl();
2174 ret
= register_filesystem(&nfs4_fs_type
);
2176 nfs_unregister_sysctl();
2180 static inline void unregister_nfs4fs(void)
2182 unregister_filesystem(&nfs4_fs_type
);
2183 nfs_unregister_sysctl();
2186 #define nfs4_init_once(nfsi) \
2188 #define register_nfs4fs() (0)
2189 #define unregister_nfs4fs()
2192 extern int nfs_init_nfspagecache(void);
2193 extern void nfs_destroy_nfspagecache(void);
2194 extern int nfs_init_readpagecache(void);
2195 extern void nfs_destroy_readpagecache(void);
2196 extern int nfs_init_writepagecache(void);
2197 extern void nfs_destroy_writepagecache(void);
2198 #ifdef CONFIG_NFS_DIRECTIO
2199 extern int nfs_init_directcache(void);
2200 extern void nfs_destroy_directcache(void);
2203 static kmem_cache_t
* nfs_inode_cachep
;
2205 static struct inode
*nfs_alloc_inode(struct super_block
*sb
)
2207 struct nfs_inode
*nfsi
;
2208 nfsi
= (struct nfs_inode
*)kmem_cache_alloc(nfs_inode_cachep
, SLAB_KERNEL
);
2212 nfsi
->cache_validity
= 0UL;
2213 nfsi
->cache_change_attribute
= jiffies
;
2214 #ifdef CONFIG_NFS_V3_ACL
2215 nfsi
->acl_access
= ERR_PTR(-EAGAIN
);
2216 nfsi
->acl_default
= ERR_PTR(-EAGAIN
);
2218 #ifdef CONFIG_NFS_V4
2219 nfsi
->nfs4_acl
= NULL
;
2220 #endif /* CONFIG_NFS_V4 */
2221 return &nfsi
->vfs_inode
;
2224 static void nfs_destroy_inode(struct inode
*inode
)
2226 kmem_cache_free(nfs_inode_cachep
, NFS_I(inode
));
2229 static void init_once(void * foo
, kmem_cache_t
* cachep
, unsigned long flags
)
2231 struct nfs_inode
*nfsi
= (struct nfs_inode
*) foo
;
2233 if ((flags
& (SLAB_CTOR_VERIFY
|SLAB_CTOR_CONSTRUCTOR
)) ==
2234 SLAB_CTOR_CONSTRUCTOR
) {
2235 inode_init_once(&nfsi
->vfs_inode
);
2236 spin_lock_init(&nfsi
->req_lock
);
2237 INIT_LIST_HEAD(&nfsi
->dirty
);
2238 INIT_LIST_HEAD(&nfsi
->commit
);
2239 INIT_LIST_HEAD(&nfsi
->open_files
);
2240 INIT_RADIX_TREE(&nfsi
->nfs_page_tree
, GFP_ATOMIC
);
2241 atomic_set(&nfsi
->data_updates
, 0);
2245 nfs4_init_once(nfsi
);
2249 static int nfs_init_inodecache(void)
2251 nfs_inode_cachep
= kmem_cache_create("nfs_inode_cache",
2252 sizeof(struct nfs_inode
),
2253 0, (SLAB_RECLAIM_ACCOUNT
|
2256 if (nfs_inode_cachep
== NULL
)
2262 static void nfs_destroy_inodecache(void)
2264 if (kmem_cache_destroy(nfs_inode_cachep
))
2265 printk(KERN_INFO
"nfs_inode_cache: not all structures were freed\n");
2271 static int __init
init_nfs_fs(void)
2275 err
= nfs_init_nfspagecache();
2279 err
= nfs_init_inodecache();
2283 err
= nfs_init_readpagecache();
2287 err
= nfs_init_writepagecache();
2291 #ifdef CONFIG_NFS_DIRECTIO
2292 err
= nfs_init_directcache();
2297 #ifdef CONFIG_PROC_FS
2298 rpc_proc_register(&nfs_rpcstat
);
2300 err
= register_filesystem(&nfs_fs_type
);
2303 if ((err
= register_nfs4fs()) != 0)
2307 #ifdef CONFIG_PROC_FS
2308 rpc_proc_unregister("nfs");
2310 #ifdef CONFIG_NFS_DIRECTIO
2311 nfs_destroy_directcache();
2314 nfs_destroy_writepagecache();
2316 nfs_destroy_readpagecache();
2318 nfs_destroy_inodecache();
2320 nfs_destroy_nfspagecache();
2325 static void __exit
exit_nfs_fs(void)
2327 #ifdef CONFIG_NFS_DIRECTIO
2328 nfs_destroy_directcache();
2330 nfs_destroy_writepagecache();
2331 nfs_destroy_readpagecache();
2332 nfs_destroy_inodecache();
2333 nfs_destroy_nfspagecache();
2334 #ifdef CONFIG_PROC_FS
2335 rpc_proc_unregister("nfs");
2337 unregister_filesystem(&nfs_fs_type
);
2338 unregister_nfs4fs();
2341 /* Not quite true; I just maintain it */
2342 MODULE_AUTHOR("Olaf Kirch <okir@monad.swb.de>");
2343 MODULE_LICENSE("GPL");
2345 module_init(init_nfs_fs
)
2346 module_exit(exit_nfs_fs
)