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_delete_inode(struct inode
*);
66 static void nfs_clear_inode(struct inode
*);
67 static void nfs_umount_begin(struct super_block
*);
68 static int nfs_statfs(struct dentry
*, struct kstatfs
*);
69 static int nfs_show_options(struct seq_file
*, struct vfsmount
*);
70 static int nfs_show_stats(struct seq_file
*, struct vfsmount
*);
71 static void nfs_zap_acl_cache(struct inode
*);
73 static struct rpc_program nfs_program
;
75 static struct super_operations nfs_sops
= {
76 .alloc_inode
= nfs_alloc_inode
,
77 .destroy_inode
= nfs_destroy_inode
,
78 .write_inode
= nfs_write_inode
,
79 .delete_inode
= nfs_delete_inode
,
81 .clear_inode
= nfs_clear_inode
,
82 .umount_begin
= nfs_umount_begin
,
83 .show_options
= nfs_show_options
,
84 .show_stats
= nfs_show_stats
,
90 static struct rpc_stat nfs_rpcstat
= {
91 .program
= &nfs_program
93 static struct rpc_version
* nfs_version
[] = {
97 #if defined(CONFIG_NFS_V3)
99 #elif defined(CONFIG_NFS_V4)
102 #if defined(CONFIG_NFS_V4)
107 static struct rpc_program nfs_program
= {
109 .number
= NFS_PROGRAM
,
110 .nrvers
= ARRAY_SIZE(nfs_version
),
111 .version
= nfs_version
,
112 .stats
= &nfs_rpcstat
,
113 .pipe_dir_name
= "/nfs",
116 #ifdef CONFIG_NFS_V3_ACL
117 static struct rpc_stat nfsacl_rpcstat
= { &nfsacl_program
};
118 static struct rpc_version
* nfsacl_version
[] = {
119 [3] = &nfsacl_version3
,
122 struct rpc_program nfsacl_program
= {
124 .number
= NFS_ACL_PROGRAM
,
125 .nrvers
= ARRAY_SIZE(nfsacl_version
),
126 .version
= nfsacl_version
,
127 .stats
= &nfsacl_rpcstat
,
129 #endif /* CONFIG_NFS_V3_ACL */
131 static inline unsigned long
132 nfs_fattr_to_ino_t(struct nfs_fattr
*fattr
)
134 return nfs_fileid_to_ino_t(fattr
->fileid
);
138 nfs_write_inode(struct inode
*inode
, int sync
)
140 int flags
= sync
? FLUSH_SYNC
: 0;
143 ret
= nfs_commit_inode(inode
, flags
);
150 nfs_delete_inode(struct inode
* inode
)
152 dprintk("NFS: delete_inode(%s/%ld)\n", inode
->i_sb
->s_id
, inode
->i_ino
);
154 truncate_inode_pages(&inode
->i_data
, 0);
158 * The following should never happen...
160 if (nfs_have_writebacks(inode
)) {
161 printk(KERN_ERR
"nfs_delete_inode: inode %ld has pending RPC requests\n", inode
->i_ino
);
168 nfs_clear_inode(struct inode
*inode
)
170 struct nfs_inode
*nfsi
= NFS_I(inode
);
171 struct rpc_cred
*cred
;
174 BUG_ON (!list_empty(&nfsi
->open_files
));
175 nfs_zap_acl_cache(inode
);
176 cred
= nfsi
->cache_access
.cred
;
179 BUG_ON(atomic_read(&nfsi
->data_updates
) != 0);
183 nfs_umount_begin(struct super_block
*sb
)
185 struct rpc_clnt
*rpc
= NFS_SB(sb
)->client
;
187 /* -EIO all pending I/O */
189 rpc_killall_tasks(rpc
);
190 rpc
= NFS_SB(sb
)->client_acl
;
192 rpc_killall_tasks(rpc
);
196 static inline unsigned long
197 nfs_block_bits(unsigned long bsize
, unsigned char *nrbitsp
)
199 /* make sure blocksize is a power of two */
200 if ((bsize
& (bsize
- 1)) || nrbitsp
) {
201 unsigned char nrbits
;
203 for (nrbits
= 31; nrbits
&& !(bsize
& (1 << nrbits
)); nrbits
--)
214 * Calculate the number of 512byte blocks used.
216 static inline unsigned long
217 nfs_calc_block_size(u64 tsize
)
219 loff_t used
= (tsize
+ 511) >> 9;
220 return (used
> ULONG_MAX
) ? ULONG_MAX
: used
;
224 * Compute and set NFS server blocksize
226 static inline unsigned long
227 nfs_block_size(unsigned long bsize
, unsigned char *nrbitsp
)
229 if (bsize
< NFS_MIN_FILE_IO_SIZE
)
230 bsize
= NFS_DEF_FILE_IO_SIZE
;
231 else if (bsize
>= NFS_MAX_FILE_IO_SIZE
)
232 bsize
= NFS_MAX_FILE_IO_SIZE
;
234 return nfs_block_bits(bsize
, nrbitsp
);
238 * Obtain the root inode of the file system.
240 static struct inode
*
241 nfs_get_root(struct super_block
*sb
, struct nfs_fh
*rootfh
, struct nfs_fsinfo
*fsinfo
)
243 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 return nfs_fhget(sb
, rootfh
, fsinfo
->fattr
);
256 * Do NFS version-independent mount processing, and sanity checking
259 nfs_sb_init(struct super_block
*sb
, rpc_authflavor_t authflavor
)
261 struct nfs_server
*server
;
262 struct inode
*root_inode
;
263 struct nfs_fattr fattr
;
264 struct nfs_fsinfo fsinfo
= {
267 struct nfs_pathconf pathinfo
= {
270 int no_root_error
= 0;
271 unsigned long max_rpc_payload
;
273 /* We probably want something more informative here */
274 snprintf(sb
->s_id
, sizeof(sb
->s_id
), "%x:%x", MAJOR(sb
->s_dev
), MINOR(sb
->s_dev
));
278 sb
->s_magic
= NFS_SUPER_MAGIC
;
280 server
->io_stats
= nfs_alloc_iostats();
281 if (server
->io_stats
== NULL
)
284 root_inode
= nfs_get_root(sb
, &server
->fh
, &fsinfo
);
285 /* Did getting the root inode fail? */
286 if (IS_ERR(root_inode
)) {
287 no_root_error
= PTR_ERR(root_inode
);
290 sb
->s_root
= d_alloc_root(root_inode
);
292 no_root_error
= -ENOMEM
;
295 sb
->s_root
->d_op
= server
->rpc_ops
->dentry_ops
;
297 /* mount time stamp, in seconds */
298 server
->mount_time
= jiffies
;
300 /* Get some general file system info */
301 if (server
->namelen
== 0 &&
302 server
->rpc_ops
->pathconf(server
, &server
->fh
, &pathinfo
) >= 0)
303 server
->namelen
= pathinfo
.max_namelen
;
304 /* Work out a lot of parameters */
305 if (server
->rsize
== 0)
306 server
->rsize
= nfs_block_size(fsinfo
.rtpref
, NULL
);
307 if (server
->wsize
== 0)
308 server
->wsize
= nfs_block_size(fsinfo
.wtpref
, NULL
);
310 if (fsinfo
.rtmax
>= 512 && server
->rsize
> fsinfo
.rtmax
)
311 server
->rsize
= nfs_block_size(fsinfo
.rtmax
, NULL
);
312 if (fsinfo
.wtmax
>= 512 && server
->wsize
> fsinfo
.wtmax
)
313 server
->wsize
= nfs_block_size(fsinfo
.wtmax
, NULL
);
315 max_rpc_payload
= nfs_block_size(rpc_max_payload(server
->client
), NULL
);
316 if (server
->rsize
> max_rpc_payload
)
317 server
->rsize
= max_rpc_payload
;
318 if (server
->rsize
> NFS_MAX_FILE_IO_SIZE
)
319 server
->rsize
= NFS_MAX_FILE_IO_SIZE
;
320 server
->rpages
= (server
->rsize
+ PAGE_CACHE_SIZE
- 1) >> PAGE_CACHE_SHIFT
;
322 if (server
->wsize
> max_rpc_payload
)
323 server
->wsize
= max_rpc_payload
;
324 if (server
->wsize
> NFS_MAX_FILE_IO_SIZE
)
325 server
->wsize
= NFS_MAX_FILE_IO_SIZE
;
326 server
->wpages
= (server
->wsize
+ PAGE_CACHE_SIZE
- 1) >> PAGE_CACHE_SHIFT
;
328 if (sb
->s_blocksize
== 0)
329 sb
->s_blocksize
= nfs_block_bits(server
->wsize
,
330 &sb
->s_blocksize_bits
);
331 server
->wtmult
= nfs_block_bits(fsinfo
.wtmult
, NULL
);
333 server
->dtsize
= nfs_block_size(fsinfo
.dtpref
, NULL
);
334 if (server
->dtsize
> PAGE_CACHE_SIZE
)
335 server
->dtsize
= PAGE_CACHE_SIZE
;
336 if (server
->dtsize
> server
->rsize
)
337 server
->dtsize
= server
->rsize
;
339 if (server
->flags
& NFS_MOUNT_NOAC
) {
340 server
->acregmin
= server
->acregmax
= 0;
341 server
->acdirmin
= server
->acdirmax
= 0;
342 sb
->s_flags
|= MS_SYNCHRONOUS
;
344 server
->backing_dev_info
.ra_pages
= server
->rpages
* NFS_MAX_READAHEAD
;
346 sb
->s_maxbytes
= fsinfo
.maxfilesize
;
347 if (sb
->s_maxbytes
> MAX_LFS_FILESIZE
)
348 sb
->s_maxbytes
= MAX_LFS_FILESIZE
;
350 server
->client
->cl_intr
= (server
->flags
& NFS_MOUNT_INTR
) ? 1 : 0;
351 server
->client
->cl_softrtry
= (server
->flags
& NFS_MOUNT_SOFT
) ? 1 : 0;
353 /* We're airborne Set socket buffersize */
354 rpc_setbufsize(server
->client
, server
->wsize
+ 100, server
->rsize
+ 100);
356 /* Yargs. It didn't work out. */
358 dprintk("nfs_sb_init: get root inode failed: errno %d\n", -no_root_error
);
359 if (!IS_ERR(root_inode
))
361 return no_root_error
;
364 static void nfs_init_timeout_values(struct rpc_timeout
*to
, int proto
, unsigned int timeo
, unsigned int retrans
)
366 to
->to_initval
= timeo
* HZ
/ 10;
367 to
->to_retries
= retrans
;
374 to
->to_initval
= 60 * HZ
;
375 if (to
->to_initval
> NFS_MAX_TCP_TIMEOUT
)
376 to
->to_initval
= NFS_MAX_TCP_TIMEOUT
;
377 to
->to_increment
= to
->to_initval
;
378 to
->to_maxval
= to
->to_initval
+ (to
->to_increment
* to
->to_retries
);
379 to
->to_exponential
= 0;
384 to
->to_initval
= 11 * HZ
/ 10;
385 if (to
->to_initval
> NFS_MAX_UDP_TIMEOUT
)
386 to
->to_initval
= NFS_MAX_UDP_TIMEOUT
;
387 to
->to_maxval
= NFS_MAX_UDP_TIMEOUT
;
388 to
->to_exponential
= 1;
394 * Create an RPC client handle.
396 static struct rpc_clnt
*
397 nfs_create_client(struct nfs_server
*server
, const struct nfs_mount_data
*data
)
399 struct rpc_timeout timeparms
;
400 struct rpc_xprt
*xprt
= NULL
;
401 struct rpc_clnt
*clnt
= NULL
;
402 int proto
= (data
->flags
& NFS_MOUNT_TCP
) ? IPPROTO_TCP
: IPPROTO_UDP
;
404 nfs_init_timeout_values(&timeparms
, proto
, data
->timeo
, data
->retrans
);
406 server
->retrans_timeo
= timeparms
.to_initval
;
407 server
->retrans_count
= timeparms
.to_retries
;
409 /* create transport and client */
410 xprt
= xprt_create_proto(proto
, &server
->addr
, &timeparms
);
412 dprintk("%s: cannot create RPC transport. Error = %ld\n",
413 __FUNCTION__
, PTR_ERR(xprt
));
414 return (struct rpc_clnt
*)xprt
;
416 clnt
= rpc_create_client(xprt
, server
->hostname
, &nfs_program
,
417 server
->rpc_ops
->version
, data
->pseudoflavor
);
419 dprintk("%s: cannot create RPC client. Error = %ld\n",
420 __FUNCTION__
, PTR_ERR(xprt
));
425 clnt
->cl_softrtry
= 1;
434 * The way this works is that the mount process passes a structure
435 * in the data argument which contains the server's IP address
436 * and the root file handle obtained from the server's mount
437 * daemon. We stash these away in the private superblock fields.
440 nfs_fill_super(struct super_block
*sb
, struct nfs_mount_data
*data
, int silent
)
442 struct nfs_server
*server
;
443 rpc_authflavor_t authflavor
;
446 sb
->s_blocksize_bits
= 0;
449 sb
->s_blocksize
= nfs_block_size(data
->bsize
, &sb
->s_blocksize_bits
);
451 server
->rsize
= nfs_block_size(data
->rsize
, NULL
);
453 server
->wsize
= nfs_block_size(data
->wsize
, NULL
);
454 server
->flags
= data
->flags
& NFS_MOUNT_FLAGMASK
;
456 server
->acregmin
= data
->acregmin
*HZ
;
457 server
->acregmax
= data
->acregmax
*HZ
;
458 server
->acdirmin
= data
->acdirmin
*HZ
;
459 server
->acdirmax
= data
->acdirmax
*HZ
;
461 /* Start lockd here, before we might error out */
462 if (!(server
->flags
& NFS_MOUNT_NONLM
))
465 server
->namelen
= data
->namlen
;
466 server
->hostname
= kmalloc(strlen(data
->hostname
) + 1, GFP_KERNEL
);
467 if (!server
->hostname
)
469 strcpy(server
->hostname
, data
->hostname
);
471 /* Check NFS protocol revision and initialize RPC op vector
472 * and file handle pool. */
474 if (server
->flags
& NFS_MOUNT_VER3
) {
475 server
->rpc_ops
= &nfs_v3_clientops
;
476 server
->caps
|= NFS_CAP_READDIRPLUS
;
478 server
->rpc_ops
= &nfs_v2_clientops
;
481 server
->rpc_ops
= &nfs_v2_clientops
;
484 /* Fill in pseudoflavor for mount version < 5 */
485 if (!(data
->flags
& NFS_MOUNT_SECFLAVOUR
))
486 data
->pseudoflavor
= RPC_AUTH_UNIX
;
487 authflavor
= data
->pseudoflavor
; /* save for sb_init() */
488 /* XXX maybe we want to add a server->pseudoflavor field */
490 /* Create RPC client handles */
491 server
->client
= nfs_create_client(server
, data
);
492 if (IS_ERR(server
->client
))
493 return PTR_ERR(server
->client
);
494 /* RFC 2623, sec 2.3.2 */
495 if (authflavor
!= RPC_AUTH_UNIX
) {
496 struct rpc_auth
*auth
;
498 server
->client_sys
= rpc_clone_client(server
->client
);
499 if (IS_ERR(server
->client_sys
))
500 return PTR_ERR(server
->client_sys
);
501 auth
= rpcauth_create(RPC_AUTH_UNIX
, server
->client_sys
);
503 return PTR_ERR(auth
);
505 atomic_inc(&server
->client
->cl_count
);
506 server
->client_sys
= server
->client
;
508 if (server
->flags
& NFS_MOUNT_VER3
) {
509 #ifdef CONFIG_NFS_V3_ACL
510 if (!(server
->flags
& NFS_MOUNT_NOACL
)) {
511 server
->client_acl
= rpc_bind_new_program(server
->client
, &nfsacl_program
, 3);
512 /* No errors! Assume that Sun nfsacls are supported */
513 if (!IS_ERR(server
->client_acl
))
514 server
->caps
|= NFS_CAP_ACLS
;
517 server
->flags
&= ~NFS_MOUNT_NOACL
;
518 #endif /* CONFIG_NFS_V3_ACL */
520 * The VFS shouldn't apply the umask to mode bits. We will
521 * do so ourselves when necessary.
523 sb
->s_flags
|= MS_POSIXACL
;
524 if (server
->namelen
== 0 || server
->namelen
> NFS3_MAXNAMLEN
)
525 server
->namelen
= NFS3_MAXNAMLEN
;
528 if (server
->namelen
== 0 || server
->namelen
> NFS2_MAXNAMLEN
)
529 server
->namelen
= NFS2_MAXNAMLEN
;
532 sb
->s_op
= &nfs_sops
;
533 return nfs_sb_init(sb
, authflavor
);
537 nfs_statfs(struct dentry
*dentry
, struct kstatfs
*buf
)
539 struct super_block
*sb
= dentry
->d_sb
;
540 struct nfs_server
*server
= NFS_SB(sb
);
541 unsigned char blockbits
;
542 unsigned long blockres
;
543 struct nfs_fh
*rootfh
= NFS_FH(sb
->s_root
->d_inode
);
544 struct nfs_fattr fattr
;
545 struct nfs_fsstat res
= {
552 error
= server
->rpc_ops
->statfs(server
, rootfh
, &res
);
553 buf
->f_type
= NFS_SUPER_MAGIC
;
558 * Current versions of glibc do not correctly handle the
559 * case where f_frsize != f_bsize. Eventually we want to
560 * report the value of wtmult in this field.
562 buf
->f_frsize
= sb
->s_blocksize
;
565 * On most *nix systems, f_blocks, f_bfree, and f_bavail
566 * are reported in units of f_frsize. Linux hasn't had
567 * an f_frsize field in its statfs struct until recently,
568 * thus historically Linux's sys_statfs reports these
569 * fields in units of f_bsize.
571 buf
->f_bsize
= sb
->s_blocksize
;
572 blockbits
= sb
->s_blocksize_bits
;
573 blockres
= (1 << blockbits
) - 1;
574 buf
->f_blocks
= (res
.tbytes
+ blockres
) >> blockbits
;
575 buf
->f_bfree
= (res
.fbytes
+ blockres
) >> blockbits
;
576 buf
->f_bavail
= (res
.abytes
+ blockres
) >> blockbits
;
578 buf
->f_files
= res
.tfiles
;
579 buf
->f_ffree
= res
.afiles
;
581 buf
->f_namelen
= server
->namelen
;
587 dprintk("%s: statfs error = %d\n", __FUNCTION__
, -error
);
588 buf
->f_bsize
= buf
->f_blocks
= buf
->f_bfree
= buf
->f_bavail
= -1;
593 static void nfs_show_mount_options(struct seq_file
*m
, struct nfs_server
*nfss
, int showdefaults
)
595 static struct proc_nfs_info
{
600 { NFS_MOUNT_SOFT
, ",soft", ",hard" },
601 { NFS_MOUNT_INTR
, ",intr", "" },
602 { NFS_MOUNT_NOCTO
, ",nocto", "" },
603 { NFS_MOUNT_NOAC
, ",noac", "" },
604 { NFS_MOUNT_NONLM
, ",nolock", "" },
605 { NFS_MOUNT_NOACL
, ",noacl", "" },
608 struct proc_nfs_info
*nfs_infop
;
612 seq_printf(m
, ",vers=%d", nfss
->rpc_ops
->version
);
613 seq_printf(m
, ",rsize=%d", nfss
->rsize
);
614 seq_printf(m
, ",wsize=%d", nfss
->wsize
);
615 if (nfss
->acregmin
!= 3*HZ
|| showdefaults
)
616 seq_printf(m
, ",acregmin=%d", nfss
->acregmin
/HZ
);
617 if (nfss
->acregmax
!= 60*HZ
|| showdefaults
)
618 seq_printf(m
, ",acregmax=%d", nfss
->acregmax
/HZ
);
619 if (nfss
->acdirmin
!= 30*HZ
|| showdefaults
)
620 seq_printf(m
, ",acdirmin=%d", nfss
->acdirmin
/HZ
);
621 if (nfss
->acdirmax
!= 60*HZ
|| showdefaults
)
622 seq_printf(m
, ",acdirmax=%d", nfss
->acdirmax
/HZ
);
623 for (nfs_infop
= nfs_info
; nfs_infop
->flag
; nfs_infop
++) {
624 if (nfss
->flags
& nfs_infop
->flag
)
625 seq_puts(m
, nfs_infop
->str
);
627 seq_puts(m
, nfs_infop
->nostr
);
629 switch (nfss
->client
->cl_xprt
->prot
) {
637 snprintf(buf
, sizeof(buf
), "%u", nfss
->client
->cl_xprt
->prot
);
640 seq_printf(m
, ",proto=%s", proto
);
641 seq_printf(m
, ",timeo=%lu", 10U * nfss
->retrans_timeo
/ HZ
);
642 seq_printf(m
, ",retrans=%u", nfss
->retrans_count
);
645 static int nfs_show_options(struct seq_file
*m
, struct vfsmount
*mnt
)
647 struct nfs_server
*nfss
= NFS_SB(mnt
->mnt_sb
);
649 nfs_show_mount_options(m
, nfss
, 0);
651 seq_puts(m
, ",addr=");
652 seq_escape(m
, nfss
->hostname
, " \t\n\\");
657 static int nfs_show_stats(struct seq_file
*m
, struct vfsmount
*mnt
)
660 struct nfs_server
*nfss
= NFS_SB(mnt
->mnt_sb
);
661 struct rpc_auth
*auth
= nfss
->client
->cl_auth
;
662 struct nfs_iostats totals
= { };
664 seq_printf(m
, "statvers=%s", NFS_IOSTAT_VERS
);
667 * Display all mount option settings
669 seq_printf(m
, "\n\topts:\t");
670 seq_puts(m
, mnt
->mnt_sb
->s_flags
& MS_RDONLY
? "ro" : "rw");
671 seq_puts(m
, mnt
->mnt_sb
->s_flags
& MS_SYNCHRONOUS
? ",sync" : "");
672 seq_puts(m
, mnt
->mnt_sb
->s_flags
& MS_NOATIME
? ",noatime" : "");
673 seq_puts(m
, mnt
->mnt_sb
->s_flags
& MS_NODIRATIME
? ",nodiratime" : "");
674 nfs_show_mount_options(m
, nfss
, 1);
676 seq_printf(m
, "\n\tage:\t%lu", (jiffies
- nfss
->mount_time
) / HZ
);
678 seq_printf(m
, "\n\tcaps:\t");
679 seq_printf(m
, "caps=0x%x", nfss
->caps
);
680 seq_printf(m
, ",wtmult=%d", nfss
->wtmult
);
681 seq_printf(m
, ",dtsize=%d", nfss
->dtsize
);
682 seq_printf(m
, ",bsize=%d", nfss
->bsize
);
683 seq_printf(m
, ",namelen=%d", nfss
->namelen
);
686 if (nfss
->rpc_ops
->version
== 4) {
687 seq_printf(m
, "\n\tnfsv4:\t");
688 seq_printf(m
, "bm0=0x%x", nfss
->attr_bitmask
[0]);
689 seq_printf(m
, ",bm1=0x%x", nfss
->attr_bitmask
[1]);
690 seq_printf(m
, ",acl=0x%x", nfss
->acl_bitmask
);
695 * Display security flavor in effect for this mount
697 seq_printf(m
, "\n\tsec:\tflavor=%d", auth
->au_ops
->au_flavor
);
699 seq_printf(m
, ",pseudoflavor=%d", auth
->au_flavor
);
702 * Display superblock I/O counters
704 for_each_possible_cpu(cpu
) {
705 struct nfs_iostats
*stats
;
708 stats
= per_cpu_ptr(nfss
->io_stats
, cpu
);
710 for (i
= 0; i
< __NFSIOS_COUNTSMAX
; i
++)
711 totals
.events
[i
] += stats
->events
[i
];
712 for (i
= 0; i
< __NFSIOS_BYTESMAX
; i
++)
713 totals
.bytes
[i
] += stats
->bytes
[i
];
718 seq_printf(m
, "\n\tevents:\t");
719 for (i
= 0; i
< __NFSIOS_COUNTSMAX
; i
++)
720 seq_printf(m
, "%lu ", totals
.events
[i
]);
721 seq_printf(m
, "\n\tbytes:\t");
722 for (i
= 0; i
< __NFSIOS_BYTESMAX
; i
++)
723 seq_printf(m
, "%Lu ", totals
.bytes
[i
]);
726 rpc_print_iostats(m
, nfss
->client
);
732 * nfs_sync_mapping - helper to flush all mmapped dirty data to disk
734 int nfs_sync_mapping(struct address_space
*mapping
)
738 if (mapping
->nrpages
== 0)
740 unmap_mapping_range(mapping
, 0, 0, 0);
741 ret
= filemap_write_and_wait(mapping
);
744 ret
= nfs_wb_all(mapping
->host
);
750 * Invalidate the local caches
752 static void nfs_zap_caches_locked(struct inode
*inode
)
754 struct nfs_inode
*nfsi
= NFS_I(inode
);
755 int mode
= inode
->i_mode
;
757 nfs_inc_stats(inode
, NFSIOS_ATTRINVALIDATE
);
759 NFS_ATTRTIMEO(inode
) = NFS_MINATTRTIMEO(inode
);
760 NFS_ATTRTIMEO_UPDATE(inode
) = jiffies
;
762 memset(NFS_COOKIEVERF(inode
), 0, sizeof(NFS_COOKIEVERF(inode
)));
763 if (S_ISREG(mode
) || S_ISDIR(mode
) || S_ISLNK(mode
))
764 nfsi
->cache_validity
|= NFS_INO_INVALID_ATTR
|NFS_INO_INVALID_DATA
|NFS_INO_INVALID_ACCESS
|NFS_INO_INVALID_ACL
|NFS_INO_REVAL_PAGECACHE
;
766 nfsi
->cache_validity
|= NFS_INO_INVALID_ATTR
|NFS_INO_INVALID_ACCESS
|NFS_INO_INVALID_ACL
|NFS_INO_REVAL_PAGECACHE
;
769 void nfs_zap_caches(struct inode
*inode
)
771 spin_lock(&inode
->i_lock
);
772 nfs_zap_caches_locked(inode
);
773 spin_unlock(&inode
->i_lock
);
776 static void nfs_zap_acl_cache(struct inode
*inode
)
778 void (*clear_acl_cache
)(struct inode
*);
780 clear_acl_cache
= NFS_PROTO(inode
)->clear_acl_cache
;
781 if (clear_acl_cache
!= NULL
)
782 clear_acl_cache(inode
);
783 spin_lock(&inode
->i_lock
);
784 NFS_I(inode
)->cache_validity
&= ~NFS_INO_INVALID_ACL
;
785 spin_unlock(&inode
->i_lock
);
789 * Invalidate, but do not unhash, the inode.
790 * NB: must be called with inode->i_lock held!
792 static void nfs_invalidate_inode(struct inode
*inode
)
794 set_bit(NFS_INO_STALE
, &NFS_FLAGS(inode
));
795 nfs_zap_caches_locked(inode
);
798 struct nfs_find_desc
{
800 struct nfs_fattr
*fattr
;
804 * In NFSv3 we can have 64bit inode numbers. In order to support
805 * this, and re-exported directories (also seen in NFSv2)
806 * we are forced to allow 2 different inodes to have the same
810 nfs_find_actor(struct inode
*inode
, void *opaque
)
812 struct nfs_find_desc
*desc
= (struct nfs_find_desc
*)opaque
;
813 struct nfs_fh
*fh
= desc
->fh
;
814 struct nfs_fattr
*fattr
= desc
->fattr
;
816 if (NFS_FILEID(inode
) != fattr
->fileid
)
818 if (nfs_compare_fh(NFS_FH(inode
), fh
))
820 if (is_bad_inode(inode
) || NFS_STALE(inode
))
826 nfs_init_locked(struct inode
*inode
, void *opaque
)
828 struct nfs_find_desc
*desc
= (struct nfs_find_desc
*)opaque
;
829 struct nfs_fattr
*fattr
= desc
->fattr
;
831 NFS_FILEID(inode
) = fattr
->fileid
;
832 nfs_copy_fh(NFS_FH(inode
), desc
->fh
);
836 /* Don't use READDIRPLUS on directories that we believe are too large */
837 #define NFS_LIMIT_READDIRPLUS (8*PAGE_SIZE)
840 * This is our front-end to iget that looks up inodes by file handle
841 * instead of inode number.
844 nfs_fhget(struct super_block
*sb
, struct nfs_fh
*fh
, struct nfs_fattr
*fattr
)
846 struct nfs_find_desc desc
= {
850 struct inode
*inode
= ERR_PTR(-ENOENT
);
853 if ((fattr
->valid
& NFS_ATTR_FATTR
) == 0)
857 printk("NFS: Buggy server - nlink == 0!\n");
861 hash
= nfs_fattr_to_ino_t(fattr
);
863 inode
= iget5_locked(sb
, hash
, nfs_find_actor
, nfs_init_locked
, &desc
);
865 inode
= ERR_PTR(-ENOMEM
);
869 if (inode
->i_state
& I_NEW
) {
870 struct nfs_inode
*nfsi
= NFS_I(inode
);
872 /* We set i_ino for the few things that still rely on it,
876 /* We can't support update_atime(), since the server will reset it */
877 inode
->i_flags
|= S_NOATIME
|S_NOCMTIME
;
878 inode
->i_mode
= fattr
->mode
;
879 /* Why so? Because we want revalidate for devices/FIFOs, and
880 * that's precisely what we have in nfs_file_inode_operations.
882 inode
->i_op
= NFS_SB(sb
)->rpc_ops
->file_inode_ops
;
883 if (S_ISREG(inode
->i_mode
)) {
884 inode
->i_fop
= &nfs_file_operations
;
885 inode
->i_data
.a_ops
= &nfs_file_aops
;
886 inode
->i_data
.backing_dev_info
= &NFS_SB(sb
)->backing_dev_info
;
887 } else if (S_ISDIR(inode
->i_mode
)) {
888 inode
->i_op
= NFS_SB(sb
)->rpc_ops
->dir_inode_ops
;
889 inode
->i_fop
= &nfs_dir_operations
;
890 if (nfs_server_capable(inode
, NFS_CAP_READDIRPLUS
)
891 && fattr
->size
<= NFS_LIMIT_READDIRPLUS
)
892 set_bit(NFS_INO_ADVISE_RDPLUS
, &NFS_FLAGS(inode
));
893 } else if (S_ISLNK(inode
->i_mode
))
894 inode
->i_op
= &nfs_symlink_inode_operations
;
896 init_special_inode(inode
, inode
->i_mode
, fattr
->rdev
);
898 nfsi
->read_cache_jiffies
= fattr
->time_start
;
899 nfsi
->last_updated
= jiffies
;
900 inode
->i_atime
= fattr
->atime
;
901 inode
->i_mtime
= fattr
->mtime
;
902 inode
->i_ctime
= fattr
->ctime
;
903 if (fattr
->valid
& NFS_ATTR_FATTR_V4
)
904 nfsi
->change_attr
= fattr
->change_attr
;
905 inode
->i_size
= nfs_size_to_loff_t(fattr
->size
);
906 inode
->i_nlink
= fattr
->nlink
;
907 inode
->i_uid
= fattr
->uid
;
908 inode
->i_gid
= fattr
->gid
;
909 if (fattr
->valid
& (NFS_ATTR_FATTR_V3
| NFS_ATTR_FATTR_V4
)) {
911 * report the blocks in 512byte units
913 inode
->i_blocks
= nfs_calc_block_size(fattr
->du
.nfs3
.used
);
914 inode
->i_blksize
= inode
->i_sb
->s_blocksize
;
916 inode
->i_blocks
= fattr
->du
.nfs2
.blocks
;
917 inode
->i_blksize
= fattr
->du
.nfs2
.blocksize
;
919 nfsi
->attrtimeo
= NFS_MINATTRTIMEO(inode
);
920 nfsi
->attrtimeo_timestamp
= jiffies
;
921 memset(nfsi
->cookieverf
, 0, sizeof(nfsi
->cookieverf
));
922 nfsi
->cache_access
.cred
= NULL
;
924 unlock_new_inode(inode
);
926 nfs_refresh_inode(inode
, fattr
);
927 dprintk("NFS: nfs_fhget(%s/%Ld ct=%d)\n",
929 (long long)NFS_FILEID(inode
),
930 atomic_read(&inode
->i_count
));
936 dprintk("nfs_fhget: iget failed with error %ld\n", PTR_ERR(inode
));
940 #define NFS_VALID_ATTRS (ATTR_MODE|ATTR_UID|ATTR_GID|ATTR_SIZE|ATTR_ATIME|ATTR_ATIME_SET|ATTR_MTIME|ATTR_MTIME_SET)
943 nfs_setattr(struct dentry
*dentry
, struct iattr
*attr
)
945 struct inode
*inode
= dentry
->d_inode
;
946 struct nfs_fattr fattr
;
949 nfs_inc_stats(inode
, NFSIOS_VFSSETATTR
);
951 if (attr
->ia_valid
& ATTR_SIZE
) {
952 if (!S_ISREG(inode
->i_mode
) || attr
->ia_size
== i_size_read(inode
))
953 attr
->ia_valid
&= ~ATTR_SIZE
;
956 /* Optimization: if the end result is no change, don't RPC */
957 attr
->ia_valid
&= NFS_VALID_ATTRS
;
958 if (attr
->ia_valid
== 0)
962 nfs_begin_data_update(inode
);
963 /* Write all dirty data */
964 filemap_write_and_wait(inode
->i_mapping
);
967 * Return any delegations if we're going to change ACLs
969 if ((attr
->ia_valid
& (ATTR_MODE
|ATTR_UID
|ATTR_GID
)) != 0)
970 nfs_inode_return_delegation(inode
);
971 error
= NFS_PROTO(inode
)->setattr(dentry
, &fattr
, attr
);
973 nfs_refresh_inode(inode
, &fattr
);
974 nfs_end_data_update(inode
);
980 * nfs_setattr_update_inode - Update inode metadata after a setattr call.
981 * @inode: pointer to struct inode
982 * @attr: pointer to struct iattr
984 * Note: we do this in the *proc.c in order to ensure that
985 * it works for things like exclusive creates too.
987 void nfs_setattr_update_inode(struct inode
*inode
, struct iattr
*attr
)
989 if ((attr
->ia_valid
& (ATTR_MODE
|ATTR_UID
|ATTR_GID
)) != 0) {
990 if ((attr
->ia_valid
& ATTR_MODE
) != 0) {
991 int mode
= attr
->ia_mode
& S_IALLUGO
;
992 mode
|= inode
->i_mode
& ~S_IALLUGO
;
993 inode
->i_mode
= mode
;
995 if ((attr
->ia_valid
& ATTR_UID
) != 0)
996 inode
->i_uid
= attr
->ia_uid
;
997 if ((attr
->ia_valid
& ATTR_GID
) != 0)
998 inode
->i_gid
= attr
->ia_gid
;
999 spin_lock(&inode
->i_lock
);
1000 NFS_I(inode
)->cache_validity
|= NFS_INO_INVALID_ACCESS
|NFS_INO_INVALID_ACL
;
1001 spin_unlock(&inode
->i_lock
);
1003 if ((attr
->ia_valid
& ATTR_SIZE
) != 0) {
1004 nfs_inc_stats(inode
, NFSIOS_SETATTRTRUNC
);
1005 inode
->i_size
= attr
->ia_size
;
1006 vmtruncate(inode
, attr
->ia_size
);
1010 static int nfs_wait_schedule(void *word
)
1012 if (signal_pending(current
))
1013 return -ERESTARTSYS
;
1019 * Wait for the inode to get unlocked.
1021 static int nfs_wait_on_inode(struct inode
*inode
)
1023 struct rpc_clnt
*clnt
= NFS_CLIENT(inode
);
1024 struct nfs_inode
*nfsi
= NFS_I(inode
);
1028 rpc_clnt_sigmask(clnt
, &oldmask
);
1029 error
= wait_on_bit_lock(&nfsi
->flags
, NFS_INO_REVALIDATING
,
1030 nfs_wait_schedule
, TASK_INTERRUPTIBLE
);
1031 rpc_clnt_sigunmask(clnt
, &oldmask
);
1036 static void nfs_wake_up_inode(struct inode
*inode
)
1038 struct nfs_inode
*nfsi
= NFS_I(inode
);
1040 clear_bit(NFS_INO_REVALIDATING
, &nfsi
->flags
);
1041 smp_mb__after_clear_bit();
1042 wake_up_bit(&nfsi
->flags
, NFS_INO_REVALIDATING
);
1045 int nfs_getattr(struct vfsmount
*mnt
, struct dentry
*dentry
, struct kstat
*stat
)
1047 struct inode
*inode
= dentry
->d_inode
;
1048 int need_atime
= NFS_I(inode
)->cache_validity
& NFS_INO_INVALID_ATIME
;
1051 /* Flush out writes to the server in order to update c/mtime */
1052 nfs_sync_inode_wait(inode
, 0, 0, FLUSH_NOCOMMIT
);
1055 * We may force a getattr if the user cares about atime.
1057 * Note that we only have to check the vfsmount flags here:
1058 * - NFS always sets S_NOATIME by so checking it would give a
1060 * - NFS never sets MS_NOATIME or MS_NODIRATIME so there is
1061 * no point in checking those.
1063 if ((mnt
->mnt_flags
& MNT_NOATIME
) ||
1064 ((mnt
->mnt_flags
& MNT_NODIRATIME
) && S_ISDIR(inode
->i_mode
)))
1068 err
= __nfs_revalidate_inode(NFS_SERVER(inode
), inode
);
1070 err
= nfs_revalidate_inode(NFS_SERVER(inode
), inode
);
1072 generic_fillattr(inode
, stat
);
1076 static struct nfs_open_context
*alloc_nfs_open_context(struct vfsmount
*mnt
, struct dentry
*dentry
, struct rpc_cred
*cred
)
1078 struct nfs_open_context
*ctx
;
1080 ctx
= (struct nfs_open_context
*)kmalloc(sizeof(*ctx
), GFP_KERNEL
);
1082 atomic_set(&ctx
->count
, 1);
1083 ctx
->dentry
= dget(dentry
);
1084 ctx
->vfsmnt
= mntget(mnt
);
1085 ctx
->cred
= get_rpccred(cred
);
1087 ctx
->lockowner
= current
->files
;
1089 ctx
->dir_cookie
= 0;
1094 struct nfs_open_context
*get_nfs_open_context(struct nfs_open_context
*ctx
)
1097 atomic_inc(&ctx
->count
);
1101 void put_nfs_open_context(struct nfs_open_context
*ctx
)
1103 if (atomic_dec_and_test(&ctx
->count
)) {
1104 if (!list_empty(&ctx
->list
)) {
1105 struct inode
*inode
= ctx
->dentry
->d_inode
;
1106 spin_lock(&inode
->i_lock
);
1107 list_del(&ctx
->list
);
1108 spin_unlock(&inode
->i_lock
);
1110 if (ctx
->state
!= NULL
)
1111 nfs4_close_state(ctx
->state
, ctx
->mode
);
1112 if (ctx
->cred
!= NULL
)
1113 put_rpccred(ctx
->cred
);
1115 mntput(ctx
->vfsmnt
);
1121 * Ensure that mmap has a recent RPC credential for use when writing out
1124 static void nfs_file_set_open_context(struct file
*filp
, struct nfs_open_context
*ctx
)
1126 struct inode
*inode
= filp
->f_dentry
->d_inode
;
1127 struct nfs_inode
*nfsi
= NFS_I(inode
);
1129 filp
->private_data
= get_nfs_open_context(ctx
);
1130 spin_lock(&inode
->i_lock
);
1131 list_add(&ctx
->list
, &nfsi
->open_files
);
1132 spin_unlock(&inode
->i_lock
);
1136 * Given an inode, search for an open context with the desired characteristics
1138 struct nfs_open_context
*nfs_find_open_context(struct inode
*inode
, struct rpc_cred
*cred
, int mode
)
1140 struct nfs_inode
*nfsi
= NFS_I(inode
);
1141 struct nfs_open_context
*pos
, *ctx
= NULL
;
1143 spin_lock(&inode
->i_lock
);
1144 list_for_each_entry(pos
, &nfsi
->open_files
, list
) {
1145 if (cred
!= NULL
&& pos
->cred
!= cred
)
1147 if ((pos
->mode
& mode
) == mode
) {
1148 ctx
= get_nfs_open_context(pos
);
1152 spin_unlock(&inode
->i_lock
);
1156 static void nfs_file_clear_open_context(struct file
*filp
)
1158 struct inode
*inode
= filp
->f_dentry
->d_inode
;
1159 struct nfs_open_context
*ctx
= (struct nfs_open_context
*)filp
->private_data
;
1162 filp
->private_data
= NULL
;
1163 spin_lock(&inode
->i_lock
);
1164 list_move_tail(&ctx
->list
, &NFS_I(inode
)->open_files
);
1165 spin_unlock(&inode
->i_lock
);
1166 put_nfs_open_context(ctx
);
1171 * These allocate and release file read/write context information.
1173 int nfs_open(struct inode
*inode
, struct file
*filp
)
1175 struct nfs_open_context
*ctx
;
1176 struct rpc_cred
*cred
;
1178 cred
= rpcauth_lookupcred(NFS_CLIENT(inode
)->cl_auth
, 0);
1180 return PTR_ERR(cred
);
1181 ctx
= alloc_nfs_open_context(filp
->f_vfsmnt
, filp
->f_dentry
, cred
);
1185 ctx
->mode
= filp
->f_mode
;
1186 nfs_file_set_open_context(filp
, ctx
);
1187 put_nfs_open_context(ctx
);
1191 int nfs_release(struct inode
*inode
, struct file
*filp
)
1193 nfs_file_clear_open_context(filp
);
1198 * This function is called whenever some part of NFS notices that
1199 * the cached attributes have to be refreshed.
1202 __nfs_revalidate_inode(struct nfs_server
*server
, struct inode
*inode
)
1204 int status
= -ESTALE
;
1205 struct nfs_fattr fattr
;
1206 struct nfs_inode
*nfsi
= NFS_I(inode
);
1208 dfprintk(PAGECACHE
, "NFS: revalidating (%s/%Ld)\n",
1209 inode
->i_sb
->s_id
, (long long)NFS_FILEID(inode
));
1212 if (!inode
|| is_bad_inode(inode
))
1214 if (NFS_STALE(inode
))
1217 status
= nfs_wait_on_inode(inode
);
1220 if (NFS_STALE(inode
)) {
1222 /* Do we trust the cached ESTALE? */
1223 if (NFS_ATTRTIMEO(inode
) != 0) {
1224 if (nfsi
->cache_validity
& (NFS_INO_INVALID_ATTR
|NFS_INO_INVALID_DATA
|NFS_INO_INVALID_ATIME
)) {
1231 status
= NFS_PROTO(inode
)->getattr(server
, NFS_FH(inode
), &fattr
);
1233 dfprintk(PAGECACHE
, "nfs_revalidate_inode: (%s/%Ld) getattr failed, error=%d\n",
1235 (long long)NFS_FILEID(inode
), status
);
1236 if (status
== -ESTALE
) {
1237 nfs_zap_caches(inode
);
1238 if (!S_ISDIR(inode
->i_mode
))
1239 set_bit(NFS_INO_STALE
, &NFS_FLAGS(inode
));
1244 spin_lock(&inode
->i_lock
);
1245 status
= nfs_update_inode(inode
, &fattr
);
1247 spin_unlock(&inode
->i_lock
);
1248 dfprintk(PAGECACHE
, "nfs_revalidate_inode: (%s/%Ld) refresh failed, error=%d\n",
1250 (long long)NFS_FILEID(inode
), status
);
1253 spin_unlock(&inode
->i_lock
);
1255 nfs_revalidate_mapping(inode
, inode
->i_mapping
);
1257 if (nfsi
->cache_validity
& NFS_INO_INVALID_ACL
)
1258 nfs_zap_acl_cache(inode
);
1260 dfprintk(PAGECACHE
, "NFS: (%s/%Ld) revalidation complete\n",
1262 (long long)NFS_FILEID(inode
));
1265 nfs_wake_up_inode(inode
);
1272 int nfs_attribute_timeout(struct inode
*inode
)
1274 struct nfs_inode
*nfsi
= NFS_I(inode
);
1276 if (nfs_have_delegation(inode
, FMODE_READ
))
1278 return time_after(jiffies
, nfsi
->read_cache_jiffies
+nfsi
->attrtimeo
);
1282 * nfs_revalidate_inode - Revalidate the inode attributes
1283 * @server - pointer to nfs_server struct
1284 * @inode - pointer to inode struct
1286 * Updates inode attribute information by retrieving the data from the server.
1288 int nfs_revalidate_inode(struct nfs_server
*server
, struct inode
*inode
)
1290 nfs_inc_stats(inode
, NFSIOS_INODEREVALIDATE
);
1291 if (!(NFS_I(inode
)->cache_validity
& (NFS_INO_INVALID_ATTR
|NFS_INO_INVALID_DATA
))
1292 && !nfs_attribute_timeout(inode
))
1293 return NFS_STALE(inode
) ? -ESTALE
: 0;
1294 return __nfs_revalidate_inode(server
, inode
);
1298 * nfs_revalidate_mapping - Revalidate the pagecache
1299 * @inode - pointer to host inode
1300 * @mapping - pointer to mapping
1302 void nfs_revalidate_mapping(struct inode
*inode
, struct address_space
*mapping
)
1304 struct nfs_inode
*nfsi
= NFS_I(inode
);
1306 if (nfsi
->cache_validity
& NFS_INO_INVALID_DATA
) {
1307 nfs_inc_stats(inode
, NFSIOS_DATAINVALIDATE
);
1308 if (S_ISREG(inode
->i_mode
))
1309 nfs_sync_mapping(mapping
);
1310 invalidate_inode_pages2(mapping
);
1312 spin_lock(&inode
->i_lock
);
1313 nfsi
->cache_validity
&= ~NFS_INO_INVALID_DATA
;
1314 if (S_ISDIR(inode
->i_mode
)) {
1315 memset(nfsi
->cookieverf
, 0, sizeof(nfsi
->cookieverf
));
1316 /* This ensures we revalidate child dentries */
1317 nfsi
->cache_change_attribute
= jiffies
;
1319 spin_unlock(&inode
->i_lock
);
1321 dfprintk(PAGECACHE
, "NFS: (%s/%Ld) data cache invalidated\n",
1323 (long long)NFS_FILEID(inode
));
1328 * nfs_begin_data_update
1329 * @inode - pointer to inode
1330 * Declare that a set of operations will update file data on the server
1332 void nfs_begin_data_update(struct inode
*inode
)
1334 atomic_inc(&NFS_I(inode
)->data_updates
);
1338 * nfs_end_data_update
1339 * @inode - pointer to inode
1340 * Declare end of the operations that will update file data
1341 * This will mark the inode as immediately needing revalidation
1342 * of its attribute cache.
1344 void nfs_end_data_update(struct inode
*inode
)
1346 struct nfs_inode
*nfsi
= NFS_I(inode
);
1348 if (!nfs_have_delegation(inode
, FMODE_READ
)) {
1349 /* Directories and symlinks: invalidate page cache */
1350 if (S_ISDIR(inode
->i_mode
) || S_ISLNK(inode
->i_mode
)) {
1351 spin_lock(&inode
->i_lock
);
1352 nfsi
->cache_validity
|= NFS_INO_INVALID_DATA
;
1353 spin_unlock(&inode
->i_lock
);
1356 nfsi
->cache_change_attribute
= jiffies
;
1357 atomic_dec(&nfsi
->data_updates
);
1360 static void nfs_wcc_update_inode(struct inode
*inode
, struct nfs_fattr
*fattr
)
1362 struct nfs_inode
*nfsi
= NFS_I(inode
);
1364 if ((fattr
->valid
& NFS_ATTR_PRE_CHANGE
) != 0
1365 && nfsi
->change_attr
== fattr
->pre_change_attr
) {
1366 nfsi
->change_attr
= fattr
->change_attr
;
1367 nfsi
->cache_change_attribute
= jiffies
;
1370 /* If we have atomic WCC data, we may update some attributes */
1371 if ((fattr
->valid
& NFS_ATTR_WCC
) != 0) {
1372 if (timespec_equal(&inode
->i_ctime
, &fattr
->pre_ctime
)) {
1373 memcpy(&inode
->i_ctime
, &fattr
->ctime
, sizeof(inode
->i_ctime
));
1374 nfsi
->cache_change_attribute
= jiffies
;
1376 if (timespec_equal(&inode
->i_mtime
, &fattr
->pre_mtime
)) {
1377 memcpy(&inode
->i_mtime
, &fattr
->mtime
, sizeof(inode
->i_mtime
));
1378 nfsi
->cache_change_attribute
= jiffies
;
1380 if (inode
->i_size
== fattr
->pre_size
&& nfsi
->npages
== 0) {
1381 inode
->i_size
= fattr
->size
;
1382 nfsi
->cache_change_attribute
= jiffies
;
1388 * nfs_check_inode_attributes - verify consistency of the inode attribute cache
1389 * @inode - pointer to inode
1390 * @fattr - updated attributes
1392 * Verifies the attribute cache. If we have just changed the attributes,
1393 * so that fattr carries weak cache consistency data, then it may
1394 * also update the ctime/mtime/change_attribute.
1396 static int nfs_check_inode_attributes(struct inode
*inode
, struct nfs_fattr
*fattr
)
1398 struct nfs_inode
*nfsi
= NFS_I(inode
);
1399 loff_t cur_size
, new_isize
;
1403 if ((fattr
->valid
& NFS_ATTR_FATTR
) == 0)
1406 /* Has the inode gone and changed behind our back? */
1407 if (nfsi
->fileid
!= fattr
->fileid
1408 || (inode
->i_mode
& S_IFMT
) != (fattr
->mode
& S_IFMT
)) {
1412 /* Are we in the process of updating data on the server? */
1413 data_unstable
= nfs_caches_unstable(inode
);
1415 /* Do atomic weak cache consistency updates */
1416 nfs_wcc_update_inode(inode
, fattr
);
1418 if ((fattr
->valid
& NFS_ATTR_FATTR_V4
) != 0) {
1419 if (nfsi
->change_attr
== fattr
->change_attr
)
1421 nfsi
->cache_validity
|= NFS_INO_INVALID_ATTR
;
1423 nfsi
->cache_validity
|= NFS_INO_REVAL_PAGECACHE
;
1426 /* Verify a few of the more important attributes */
1427 if (!timespec_equal(&inode
->i_mtime
, &fattr
->mtime
)) {
1428 nfsi
->cache_validity
|= NFS_INO_INVALID_ATTR
;
1430 nfsi
->cache_validity
|= NFS_INO_REVAL_PAGECACHE
;
1433 cur_size
= i_size_read(inode
);
1434 new_isize
= nfs_size_to_loff_t(fattr
->size
);
1435 if (cur_size
!= new_isize
&& nfsi
->npages
== 0)
1436 nfsi
->cache_validity
|= NFS_INO_INVALID_ATTR
|NFS_INO_REVAL_PAGECACHE
;
1438 /* Have any file permissions changed? */
1439 if ((inode
->i_mode
& S_IALLUGO
) != (fattr
->mode
& S_IALLUGO
)
1440 || inode
->i_uid
!= fattr
->uid
1441 || inode
->i_gid
!= fattr
->gid
)
1442 nfsi
->cache_validity
|= NFS_INO_INVALID_ATTR
| NFS_INO_INVALID_ACCESS
| NFS_INO_INVALID_ACL
;
1444 /* Has the link count changed? */
1445 if (inode
->i_nlink
!= fattr
->nlink
)
1446 nfsi
->cache_validity
|= NFS_INO_INVALID_ATTR
;
1449 if (!timespec_equal(&inode
->i_atime
, &fattr
->atime
))
1450 nfsi
->cache_validity
|= NFS_INO_INVALID_ATIME
;
1452 nfsi
->read_cache_jiffies
= fattr
->time_start
;
1457 * nfs_refresh_inode - try to update the inode attribute cache
1458 * @inode - pointer to inode
1459 * @fattr - updated attributes
1461 * Check that an RPC call that returned attributes has not overlapped with
1462 * other recent updates of the inode metadata, then decide whether it is
1463 * safe to do a full update of the inode attributes, or whether just to
1464 * call nfs_check_inode_attributes.
1466 int nfs_refresh_inode(struct inode
*inode
, struct nfs_fattr
*fattr
)
1468 struct nfs_inode
*nfsi
= NFS_I(inode
);
1471 if ((fattr
->valid
& NFS_ATTR_FATTR
) == 0)
1473 spin_lock(&inode
->i_lock
);
1474 nfsi
->cache_validity
&= ~NFS_INO_REVAL_PAGECACHE
;
1475 if (time_after(fattr
->time_start
, nfsi
->last_updated
))
1476 status
= nfs_update_inode(inode
, fattr
);
1478 status
= nfs_check_inode_attributes(inode
, fattr
);
1480 spin_unlock(&inode
->i_lock
);
1485 * nfs_post_op_update_inode - try to update the inode attribute cache
1486 * @inode - pointer to inode
1487 * @fattr - updated attributes
1489 * After an operation that has changed the inode metadata, mark the
1490 * attribute cache as being invalid, then try to update it.
1492 int nfs_post_op_update_inode(struct inode
*inode
, struct nfs_fattr
*fattr
)
1494 struct nfs_inode
*nfsi
= NFS_I(inode
);
1497 spin_lock(&inode
->i_lock
);
1498 if (unlikely((fattr
->valid
& NFS_ATTR_FATTR
) == 0)) {
1499 nfsi
->cache_validity
|= NFS_INO_INVALID_ATTR
| NFS_INO_INVALID_ACCESS
;
1502 status
= nfs_update_inode(inode
, fattr
);
1504 spin_unlock(&inode
->i_lock
);
1509 * Many nfs protocol calls return the new file attributes after
1510 * an operation. Here we update the inode to reflect the state
1511 * of the server's inode.
1513 * This is a bit tricky because we have to make sure all dirty pages
1514 * have been sent off to the server before calling invalidate_inode_pages.
1515 * To make sure no other process adds more write requests while we try
1516 * our best to flush them, we make them sleep during the attribute refresh.
1518 * A very similar scenario holds for the dir cache.
1520 static int nfs_update_inode(struct inode
*inode
, struct nfs_fattr
*fattr
)
1522 struct nfs_inode
*nfsi
= NFS_I(inode
);
1523 loff_t cur_isize
, new_isize
;
1524 unsigned int invalid
= 0;
1527 dfprintk(VFS
, "NFS: %s(%s/%ld ct=%d info=0x%x)\n",
1528 __FUNCTION__
, inode
->i_sb
->s_id
, inode
->i_ino
,
1529 atomic_read(&inode
->i_count
), fattr
->valid
);
1531 if ((fattr
->valid
& NFS_ATTR_FATTR
) == 0)
1534 if (nfsi
->fileid
!= fattr
->fileid
)
1538 * Make sure the inode's type hasn't changed.
1540 if ((inode
->i_mode
& S_IFMT
) != (fattr
->mode
& S_IFMT
))
1544 * Update the read time so we don't revalidate too often.
1546 nfsi
->read_cache_jiffies
= fattr
->time_start
;
1547 nfsi
->last_updated
= jiffies
;
1549 /* Are we racing with known updates of the metadata on the server? */
1550 data_stable
= nfs_verify_change_attribute(inode
, fattr
->time_start
);
1552 nfsi
->cache_validity
&= ~(NFS_INO_INVALID_ATTR
|NFS_INO_INVALID_ATIME
);
1554 /* Do atomic weak cache consistency updates */
1555 nfs_wcc_update_inode(inode
, fattr
);
1557 /* Check if our cached file size is stale */
1558 new_isize
= nfs_size_to_loff_t(fattr
->size
);
1559 cur_isize
= i_size_read(inode
);
1560 if (new_isize
!= cur_isize
) {
1561 /* Do we perhaps have any outstanding writes? */
1562 if (nfsi
->npages
== 0) {
1563 /* No, but did we race with nfs_end_data_update()? */
1565 inode
->i_size
= new_isize
;
1566 invalid
|= NFS_INO_INVALID_DATA
;
1568 invalid
|= NFS_INO_INVALID_ATTR
;
1569 } else if (new_isize
> cur_isize
) {
1570 inode
->i_size
= new_isize
;
1571 invalid
|= NFS_INO_INVALID_ATTR
|NFS_INO_INVALID_DATA
;
1573 nfsi
->cache_change_attribute
= jiffies
;
1574 dprintk("NFS: isize change on server for file %s/%ld\n",
1575 inode
->i_sb
->s_id
, inode
->i_ino
);
1578 /* Check if the mtime agrees */
1579 if (!timespec_equal(&inode
->i_mtime
, &fattr
->mtime
)) {
1580 memcpy(&inode
->i_mtime
, &fattr
->mtime
, sizeof(inode
->i_mtime
));
1581 dprintk("NFS: mtime change on server for file %s/%ld\n",
1582 inode
->i_sb
->s_id
, inode
->i_ino
);
1583 invalid
|= NFS_INO_INVALID_ATTR
|NFS_INO_INVALID_DATA
;
1584 nfsi
->cache_change_attribute
= jiffies
;
1587 /* If ctime has changed we should definitely clear access+acl caches */
1588 if (!timespec_equal(&inode
->i_ctime
, &fattr
->ctime
)) {
1589 invalid
|= NFS_INO_INVALID_ACCESS
|NFS_INO_INVALID_ACL
;
1590 memcpy(&inode
->i_ctime
, &fattr
->ctime
, sizeof(inode
->i_ctime
));
1591 nfsi
->cache_change_attribute
= jiffies
;
1593 memcpy(&inode
->i_atime
, &fattr
->atime
, sizeof(inode
->i_atime
));
1595 if ((inode
->i_mode
& S_IALLUGO
) != (fattr
->mode
& S_IALLUGO
) ||
1596 inode
->i_uid
!= fattr
->uid
||
1597 inode
->i_gid
!= fattr
->gid
)
1598 invalid
|= NFS_INO_INVALID_ATTR
|NFS_INO_INVALID_ACCESS
|NFS_INO_INVALID_ACL
;
1600 inode
->i_mode
= fattr
->mode
;
1601 inode
->i_nlink
= fattr
->nlink
;
1602 inode
->i_uid
= fattr
->uid
;
1603 inode
->i_gid
= fattr
->gid
;
1605 if (fattr
->valid
& (NFS_ATTR_FATTR_V3
| NFS_ATTR_FATTR_V4
)) {
1607 * report the blocks in 512byte units
1609 inode
->i_blocks
= nfs_calc_block_size(fattr
->du
.nfs3
.used
);
1610 inode
->i_blksize
= inode
->i_sb
->s_blocksize
;
1612 inode
->i_blocks
= fattr
->du
.nfs2
.blocks
;
1613 inode
->i_blksize
= fattr
->du
.nfs2
.blocksize
;
1616 if ((fattr
->valid
& NFS_ATTR_FATTR_V4
)) {
1617 if (nfsi
->change_attr
!= fattr
->change_attr
) {
1618 dprintk("NFS: change_attr change on server for file %s/%ld\n",
1619 inode
->i_sb
->s_id
, inode
->i_ino
);
1620 nfsi
->change_attr
= fattr
->change_attr
;
1621 invalid
|= NFS_INO_INVALID_ATTR
|NFS_INO_INVALID_DATA
|NFS_INO_INVALID_ACCESS
|NFS_INO_INVALID_ACL
;
1622 nfsi
->cache_change_attribute
= jiffies
;
1624 invalid
&= ~(NFS_INO_INVALID_ATTR
|NFS_INO_INVALID_DATA
);
1627 /* Update attrtimeo value if we're out of the unstable period */
1628 if (invalid
& NFS_INO_INVALID_ATTR
) {
1629 nfs_inc_stats(inode
, NFSIOS_ATTRINVALIDATE
);
1630 nfsi
->attrtimeo
= NFS_MINATTRTIMEO(inode
);
1631 nfsi
->attrtimeo_timestamp
= jiffies
;
1632 } else if (time_after(jiffies
, nfsi
->attrtimeo_timestamp
+nfsi
->attrtimeo
)) {
1633 if ((nfsi
->attrtimeo
<<= 1) > NFS_MAXATTRTIMEO(inode
))
1634 nfsi
->attrtimeo
= NFS_MAXATTRTIMEO(inode
);
1635 nfsi
->attrtimeo_timestamp
= jiffies
;
1637 /* Don't invalidate the data if we were to blame */
1638 if (!(S_ISREG(inode
->i_mode
) || S_ISDIR(inode
->i_mode
)
1639 || S_ISLNK(inode
->i_mode
)))
1640 invalid
&= ~NFS_INO_INVALID_DATA
;
1642 invalid
&= ~(NFS_INO_INVALID_ATTR
|NFS_INO_INVALID_ATIME
|NFS_INO_REVAL_PAGECACHE
);
1643 if (!nfs_have_delegation(inode
, FMODE_READ
))
1644 nfsi
->cache_validity
|= invalid
;
1649 * Big trouble! The inode has become a different object.
1652 printk(KERN_DEBUG
"%s: inode %ld mode changed, %07o to %07o\n",
1653 __FUNCTION__
, inode
->i_ino
, inode
->i_mode
, fattr
->mode
);
1657 * No need to worry about unhashing the dentry, as the
1658 * lookup validation will know that the inode is bad.
1659 * (But we fall through to invalidate the caches.)
1661 nfs_invalidate_inode(inode
);
1665 printk(KERN_ERR
"NFS: server %s error: fileid changed\n"
1666 "fsid %s: expected fileid 0x%Lx, got 0x%Lx\n",
1667 NFS_SERVER(inode
)->hostname
, inode
->i_sb
->s_id
,
1668 (long long)nfsi
->fileid
, (long long)fattr
->fileid
);
1673 * File system information
1676 static int nfs_set_super(struct super_block
*s
, void *data
)
1678 s
->s_fs_info
= data
;
1679 return set_anon_super(s
, data
);
1682 static int nfs_compare_super(struct super_block
*sb
, void *data
)
1684 struct nfs_server
*server
= data
;
1685 struct nfs_server
*old
= NFS_SB(sb
);
1687 if (old
->addr
.sin_addr
.s_addr
!= server
->addr
.sin_addr
.s_addr
)
1689 if (old
->addr
.sin_port
!= server
->addr
.sin_port
)
1691 return !nfs_compare_fh(&old
->fh
, &server
->fh
);
1694 static int nfs_get_sb(struct file_system_type
*fs_type
,
1695 int flags
, const char *dev_name
, void *raw_data
, struct vfsmount
*mnt
)
1698 struct nfs_server
*server
= NULL
;
1699 struct super_block
*s
;
1700 struct nfs_fh
*root
;
1701 struct nfs_mount_data
*data
= raw_data
;
1705 dprintk("%s: missing data argument\n", __FUNCTION__
);
1706 goto out_err_noserver
;
1708 if (data
->version
<= 0 || data
->version
> NFS_MOUNT_VERSION
) {
1709 dprintk("%s: bad mount version\n", __FUNCTION__
);
1710 goto out_err_noserver
;
1712 switch (data
->version
) {
1718 if (data
->flags
& NFS_MOUNT_VER3
) {
1719 dprintk("%s: mount structure version %d does not support NFSv3\n",
1722 goto out_err_noserver
;
1724 data
->root
.size
= NFS2_FHSIZE
;
1725 memcpy(data
->root
.data
, data
->old_root
.data
, NFS2_FHSIZE
);
1727 if (data
->flags
& NFS_MOUNT_SECFLAVOUR
) {
1728 dprintk("%s: mount structure version %d does not support strong security\n",
1731 goto out_err_noserver
;
1734 memset(data
->context
, 0, sizeof(data
->context
));
1736 #ifndef CONFIG_NFS_V3
1737 /* If NFSv3 is not compiled in, return -EPROTONOSUPPORT */
1738 error
= -EPROTONOSUPPORT
;
1739 if (data
->flags
& NFS_MOUNT_VER3
) {
1740 dprintk("%s: NFSv3 not compiled into kernel\n", __FUNCTION__
);
1741 goto out_err_noserver
;
1743 #endif /* CONFIG_NFS_V3 */
1746 server
= kzalloc(sizeof(struct nfs_server
), GFP_KERNEL
);
1748 goto out_err_noserver
;
1749 /* Zero out the NFS state stuff */
1750 init_nfsv4_state(server
);
1751 server
->client
= server
->client_sys
= server
->client_acl
= ERR_PTR(-EINVAL
);
1754 if (data
->flags
& NFS_MOUNT_VER3
)
1755 root
->size
= data
->root
.size
;
1757 root
->size
= NFS2_FHSIZE
;
1759 if (root
->size
> sizeof(root
->data
)) {
1760 dprintk("%s: invalid root filehandle\n", __FUNCTION__
);
1763 memcpy(root
->data
, data
->root
.data
, root
->size
);
1765 /* We now require that the mount process passes the remote address */
1766 memcpy(&server
->addr
, &data
->addr
, sizeof(server
->addr
));
1767 if (server
->addr
.sin_addr
.s_addr
== INADDR_ANY
) {
1768 dprintk("%s: mount program didn't pass remote address!\n",
1773 /* Fire up rpciod if not yet running */
1774 error
= rpciod_up();
1776 dprintk("%s: couldn't start rpciod! Error = %d\n",
1777 __FUNCTION__
, error
);
1781 s
= sget(fs_type
, nfs_compare_super
, nfs_set_super
, server
);
1784 goto out_err_rpciod
;
1788 goto out_rpciod_down
;
1792 error
= nfs_fill_super(s
, data
, flags
& MS_SILENT
? 1 : 0);
1794 up_write(&s
->s_umount
);
1795 deactivate_super(s
);
1798 s
->s_flags
|= MS_ACTIVE
;
1799 return simple_set_mnt(mnt
, s
);
1804 return simple_set_mnt(mnt
, s
);
1814 static void nfs_kill_super(struct super_block
*s
)
1816 struct nfs_server
*server
= NFS_SB(s
);
1820 if (!IS_ERR(server
->client
))
1821 rpc_shutdown_client(server
->client
);
1822 if (!IS_ERR(server
->client_sys
))
1823 rpc_shutdown_client(server
->client_sys
);
1824 if (!IS_ERR(server
->client_acl
))
1825 rpc_shutdown_client(server
->client_acl
);
1827 if (!(server
->flags
& NFS_MOUNT_NONLM
))
1828 lockd_down(); /* release rpc.lockd */
1830 rpciod_down(); /* release rpciod */
1832 nfs_free_iostats(server
->io_stats
);
1833 kfree(server
->hostname
);
1837 static struct file_system_type nfs_fs_type
= {
1838 .owner
= THIS_MODULE
,
1840 .get_sb
= nfs_get_sb
,
1841 .kill_sb
= nfs_kill_super
,
1842 .fs_flags
= FS_ODD_RENAME
|FS_REVAL_DOT
|FS_BINARY_MOUNTDATA
,
1845 #ifdef CONFIG_NFS_V4
1847 static void nfs4_clear_inode(struct inode
*);
1850 static struct super_operations nfs4_sops
= {
1851 .alloc_inode
= nfs_alloc_inode
,
1852 .destroy_inode
= nfs_destroy_inode
,
1853 .write_inode
= nfs_write_inode
,
1854 .delete_inode
= nfs_delete_inode
,
1855 .statfs
= nfs_statfs
,
1856 .clear_inode
= nfs4_clear_inode
,
1857 .umount_begin
= nfs_umount_begin
,
1858 .show_options
= nfs_show_options
,
1859 .show_stats
= nfs_show_stats
,
1863 * Clean out any remaining NFSv4 state that might be left over due
1864 * to open() calls that passed nfs_atomic_lookup, but failed to call
1867 static void nfs4_clear_inode(struct inode
*inode
)
1869 struct nfs_inode
*nfsi
= NFS_I(inode
);
1871 /* If we are holding a delegation, return it! */
1872 nfs_inode_return_delegation(inode
);
1873 /* First call standard NFS clear_inode() code */
1874 nfs_clear_inode(inode
);
1875 /* Now clear out any remaining state */
1876 while (!list_empty(&nfsi
->open_states
)) {
1877 struct nfs4_state
*state
;
1879 state
= list_entry(nfsi
->open_states
.next
,
1882 dprintk("%s(%s/%Ld): found unclaimed NFSv4 state %p\n",
1885 (long long)NFS_FILEID(inode
),
1887 BUG_ON(atomic_read(&state
->count
) != 1);
1888 nfs4_close_state(state
, state
->state
);
1893 static int nfs4_fill_super(struct super_block
*sb
, struct nfs4_mount_data
*data
, int silent
)
1895 struct nfs_server
*server
;
1896 struct nfs4_client
*clp
= NULL
;
1897 struct rpc_xprt
*xprt
= NULL
;
1898 struct rpc_clnt
*clnt
= NULL
;
1899 struct rpc_timeout timeparms
;
1900 rpc_authflavor_t authflavour
;
1903 sb
->s_blocksize_bits
= 0;
1904 sb
->s_blocksize
= 0;
1905 server
= NFS_SB(sb
);
1906 if (data
->rsize
!= 0)
1907 server
->rsize
= nfs_block_size(data
->rsize
, NULL
);
1908 if (data
->wsize
!= 0)
1909 server
->wsize
= nfs_block_size(data
->wsize
, NULL
);
1910 server
->flags
= data
->flags
& NFS_MOUNT_FLAGMASK
;
1911 server
->caps
= NFS_CAP_ATOMIC_OPEN
;
1913 server
->acregmin
= data
->acregmin
*HZ
;
1914 server
->acregmax
= data
->acregmax
*HZ
;
1915 server
->acdirmin
= data
->acdirmin
*HZ
;
1916 server
->acdirmax
= data
->acdirmax
*HZ
;
1918 server
->rpc_ops
= &nfs_v4_clientops
;
1920 nfs_init_timeout_values(&timeparms
, data
->proto
, data
->timeo
, data
->retrans
);
1922 server
->retrans_timeo
= timeparms
.to_initval
;
1923 server
->retrans_count
= timeparms
.to_retries
;
1925 clp
= nfs4_get_client(&server
->addr
.sin_addr
);
1927 dprintk("%s: failed to create NFS4 client.\n", __FUNCTION__
);
1931 /* Now create transport and client */
1932 authflavour
= RPC_AUTH_UNIX
;
1933 if (data
->auth_flavourlen
!= 0) {
1934 if (data
->auth_flavourlen
!= 1) {
1935 dprintk("%s: Invalid number of RPC auth flavours %d.\n",
1936 __FUNCTION__
, data
->auth_flavourlen
);
1940 if (copy_from_user(&authflavour
, data
->auth_flavours
, sizeof(authflavour
))) {
1946 down_write(&clp
->cl_sem
);
1947 if (IS_ERR(clp
->cl_rpcclient
)) {
1948 xprt
= xprt_create_proto(data
->proto
, &server
->addr
, &timeparms
);
1950 up_write(&clp
->cl_sem
);
1951 err
= PTR_ERR(xprt
);
1952 dprintk("%s: cannot create RPC transport. Error = %d\n",
1956 clnt
= rpc_create_client(xprt
, server
->hostname
, &nfs_program
,
1957 server
->rpc_ops
->version
, authflavour
);
1959 up_write(&clp
->cl_sem
);
1960 err
= PTR_ERR(clnt
);
1961 dprintk("%s: cannot create RPC client. Error = %d\n",
1966 clnt
->cl_softrtry
= 1;
1967 clp
->cl_rpcclient
= clnt
;
1968 memcpy(clp
->cl_ipaddr
, server
->ip_addr
, sizeof(clp
->cl_ipaddr
));
1971 list_add_tail(&server
->nfs4_siblings
, &clp
->cl_superblocks
);
1972 clnt
= rpc_clone_client(clp
->cl_rpcclient
);
1974 server
->nfs4_state
= clp
;
1975 up_write(&clp
->cl_sem
);
1979 err
= PTR_ERR(clnt
);
1980 dprintk("%s: cannot create RPC client. Error = %d\n",
1985 server
->client
= clnt
;
1987 if (server
->nfs4_state
->cl_idmap
== NULL
) {
1988 dprintk("%s: failed to create idmapper.\n", __FUNCTION__
);
1992 if (clnt
->cl_auth
->au_flavor
!= authflavour
) {
1993 struct rpc_auth
*auth
;
1995 auth
= rpcauth_create(authflavour
, clnt
);
1997 dprintk("%s: couldn't create credcache!\n", __FUNCTION__
);
1998 return PTR_ERR(auth
);
2002 sb
->s_time_gran
= 1;
2004 sb
->s_op
= &nfs4_sops
;
2005 err
= nfs_sb_init(sb
, authflavour
);
2010 nfs4_put_client(clp
);
2014 static int nfs4_compare_super(struct super_block
*sb
, void *data
)
2016 struct nfs_server
*server
= data
;
2017 struct nfs_server
*old
= NFS_SB(sb
);
2019 if (strcmp(server
->hostname
, old
->hostname
) != 0)
2021 if (strcmp(server
->mnt_path
, old
->mnt_path
) != 0)
2027 nfs_copy_user_string(char *dst
, struct nfs_string
*src
, int maxlen
)
2032 return ERR_PTR(-EINVAL
);
2033 if (src
->len
< maxlen
)
2036 p
= dst
= kmalloc(maxlen
+ 1, GFP_KERNEL
);
2038 return ERR_PTR(-ENOMEM
);
2040 if (copy_from_user(dst
, src
->data
, maxlen
)) {
2042 return ERR_PTR(-EFAULT
);
2048 static int nfs4_get_sb(struct file_system_type
*fs_type
,
2049 int flags
, const char *dev_name
, void *raw_data
, struct vfsmount
*mnt
)
2052 struct nfs_server
*server
;
2053 struct super_block
*s
;
2054 struct nfs4_mount_data
*data
= raw_data
;
2058 dprintk("%s: missing data argument\n", __FUNCTION__
);
2061 if (data
->version
<= 0 || data
->version
> NFS4_MOUNT_VERSION
) {
2062 dprintk("%s: bad mount version\n", __FUNCTION__
);
2066 server
= kzalloc(sizeof(struct nfs_server
), GFP_KERNEL
);
2069 /* Zero out the NFS state stuff */
2070 init_nfsv4_state(server
);
2071 server
->client
= server
->client_sys
= server
->client_acl
= ERR_PTR(-EINVAL
);
2073 p
= nfs_copy_user_string(NULL
, &data
->hostname
, 256);
2076 server
->hostname
= p
;
2078 p
= nfs_copy_user_string(NULL
, &data
->mnt_path
, 1024);
2081 server
->mnt_path
= p
;
2083 p
= nfs_copy_user_string(server
->ip_addr
, &data
->client_addr
,
2084 sizeof(server
->ip_addr
) - 1);
2088 /* We now require that the mount process passes the remote address */
2089 if (data
->host_addrlen
!= sizeof(server
->addr
)) {
2093 if (copy_from_user(&server
->addr
, data
->host_addr
, sizeof(server
->addr
))) {
2097 if (server
->addr
.sin_family
!= AF_INET
||
2098 server
->addr
.sin_addr
.s_addr
== INADDR_ANY
) {
2099 dprintk("%s: mount program didn't pass remote IP address!\n",
2105 /* Fire up rpciod if not yet running */
2106 error
= rpciod_up();
2108 dprintk("%s: couldn't start rpciod! Error = %d\n",
2109 __FUNCTION__
, error
);
2113 s
= sget(fs_type
, nfs4_compare_super
, nfs_set_super
, server
);
2120 kfree(server
->mnt_path
);
2121 kfree(server
->hostname
);
2123 return simple_set_mnt(mnt
, s
);
2128 error
= nfs4_fill_super(s
, data
, flags
& MS_SILENT
? 1 : 0);
2130 up_write(&s
->s_umount
);
2131 deactivate_super(s
);
2134 s
->s_flags
|= MS_ACTIVE
;
2135 return simple_set_mnt(mnt
, s
);
2139 kfree(server
->mnt_path
);
2140 kfree(server
->hostname
);
2145 static void nfs4_kill_super(struct super_block
*sb
)
2147 struct nfs_server
*server
= NFS_SB(sb
);
2149 nfs_return_all_delegations(sb
);
2150 kill_anon_super(sb
);
2152 nfs4_renewd_prepare_shutdown(server
);
2154 if (server
->client
!= NULL
&& !IS_ERR(server
->client
))
2155 rpc_shutdown_client(server
->client
);
2157 destroy_nfsv4_state(server
);
2161 nfs_free_iostats(server
->io_stats
);
2162 kfree(server
->hostname
);
2166 static struct file_system_type nfs4_fs_type
= {
2167 .owner
= THIS_MODULE
,
2169 .get_sb
= nfs4_get_sb
,
2170 .kill_sb
= nfs4_kill_super
,
2171 .fs_flags
= FS_ODD_RENAME
|FS_REVAL_DOT
|FS_BINARY_MOUNTDATA
,
2174 static const int nfs_set_port_min
= 0;
2175 static const int nfs_set_port_max
= 65535;
2176 static int param_set_port(const char *val
, struct kernel_param
*kp
)
2179 int num
= simple_strtol(val
, &endp
, 0);
2180 if (endp
== val
|| *endp
|| num
< nfs_set_port_min
|| num
> nfs_set_port_max
)
2182 *((int *)kp
->arg
) = num
;
2186 module_param_call(callback_tcpport
, param_set_port
, param_get_int
,
2187 &nfs_callback_set_tcpport
, 0644);
2189 static int param_set_idmap_timeout(const char *val
, struct kernel_param
*kp
)
2192 int num
= simple_strtol(val
, &endp
, 0);
2194 if (endp
== val
|| *endp
|| num
< 0 || jif
< num
)
2196 *((int *)kp
->arg
) = jif
;
2200 module_param_call(idmap_cache_timeout
, param_set_idmap_timeout
, param_get_int
,
2201 &nfs_idmap_cache_timeout
, 0644);
2203 #define nfs4_init_once(nfsi) \
2205 INIT_LIST_HEAD(&(nfsi)->open_states); \
2206 nfsi->delegation = NULL; \
2207 nfsi->delegation_state = 0; \
2208 init_rwsem(&nfsi->rwsem); \
2211 static inline int register_nfs4fs(void)
2215 ret
= nfs_register_sysctl();
2218 ret
= register_filesystem(&nfs4_fs_type
);
2220 nfs_unregister_sysctl();
2224 static inline void unregister_nfs4fs(void)
2226 unregister_filesystem(&nfs4_fs_type
);
2227 nfs_unregister_sysctl();
2230 #define nfs4_init_once(nfsi) \
2232 #define register_nfs4fs() (0)
2233 #define unregister_nfs4fs()
2236 extern int nfs_init_nfspagecache(void);
2237 extern void nfs_destroy_nfspagecache(void);
2238 extern int nfs_init_readpagecache(void);
2239 extern void nfs_destroy_readpagecache(void);
2240 extern int nfs_init_writepagecache(void);
2241 extern void nfs_destroy_writepagecache(void);
2242 #ifdef CONFIG_NFS_DIRECTIO
2243 extern int nfs_init_directcache(void);
2244 extern void nfs_destroy_directcache(void);
2247 static kmem_cache_t
* nfs_inode_cachep
;
2249 static struct inode
*nfs_alloc_inode(struct super_block
*sb
)
2251 struct nfs_inode
*nfsi
;
2252 nfsi
= (struct nfs_inode
*)kmem_cache_alloc(nfs_inode_cachep
, SLAB_KERNEL
);
2256 nfsi
->cache_validity
= 0UL;
2257 nfsi
->cache_change_attribute
= jiffies
;
2258 #ifdef CONFIG_NFS_V3_ACL
2259 nfsi
->acl_access
= ERR_PTR(-EAGAIN
);
2260 nfsi
->acl_default
= ERR_PTR(-EAGAIN
);
2262 #ifdef CONFIG_NFS_V4
2263 nfsi
->nfs4_acl
= NULL
;
2264 #endif /* CONFIG_NFS_V4 */
2265 return &nfsi
->vfs_inode
;
2268 static void nfs_destroy_inode(struct inode
*inode
)
2270 kmem_cache_free(nfs_inode_cachep
, NFS_I(inode
));
2273 static void init_once(void * foo
, kmem_cache_t
* cachep
, unsigned long flags
)
2275 struct nfs_inode
*nfsi
= (struct nfs_inode
*) foo
;
2277 if ((flags
& (SLAB_CTOR_VERIFY
|SLAB_CTOR_CONSTRUCTOR
)) ==
2278 SLAB_CTOR_CONSTRUCTOR
) {
2279 inode_init_once(&nfsi
->vfs_inode
);
2280 spin_lock_init(&nfsi
->req_lock
);
2281 INIT_LIST_HEAD(&nfsi
->dirty
);
2282 INIT_LIST_HEAD(&nfsi
->commit
);
2283 INIT_LIST_HEAD(&nfsi
->open_files
);
2284 INIT_RADIX_TREE(&nfsi
->nfs_page_tree
, GFP_ATOMIC
);
2285 atomic_set(&nfsi
->data_updates
, 0);
2289 nfs4_init_once(nfsi
);
2293 static int nfs_init_inodecache(void)
2295 nfs_inode_cachep
= kmem_cache_create("nfs_inode_cache",
2296 sizeof(struct nfs_inode
),
2297 0, (SLAB_RECLAIM_ACCOUNT
|
2300 if (nfs_inode_cachep
== NULL
)
2306 static void nfs_destroy_inodecache(void)
2308 if (kmem_cache_destroy(nfs_inode_cachep
))
2309 printk(KERN_INFO
"nfs_inode_cache: not all structures were freed\n");
2315 static int __init
init_nfs_fs(void)
2319 err
= nfs_init_nfspagecache();
2323 err
= nfs_init_inodecache();
2327 err
= nfs_init_readpagecache();
2331 err
= nfs_init_writepagecache();
2335 #ifdef CONFIG_NFS_DIRECTIO
2336 err
= nfs_init_directcache();
2341 #ifdef CONFIG_PROC_FS
2342 rpc_proc_register(&nfs_rpcstat
);
2344 err
= register_filesystem(&nfs_fs_type
);
2347 if ((err
= register_nfs4fs()) != 0)
2351 #ifdef CONFIG_PROC_FS
2352 rpc_proc_unregister("nfs");
2354 #ifdef CONFIG_NFS_DIRECTIO
2355 nfs_destroy_directcache();
2358 nfs_destroy_writepagecache();
2360 nfs_destroy_readpagecache();
2362 nfs_destroy_inodecache();
2364 nfs_destroy_nfspagecache();
2369 static void __exit
exit_nfs_fs(void)
2371 #ifdef CONFIG_NFS_DIRECTIO
2372 nfs_destroy_directcache();
2374 nfs_destroy_writepagecache();
2375 nfs_destroy_readpagecache();
2376 nfs_destroy_inodecache();
2377 nfs_destroy_nfspagecache();
2378 #ifdef CONFIG_PROC_FS
2379 rpc_proc_unregister("nfs");
2381 unregister_filesystem(&nfs_fs_type
);
2382 unregister_nfs4fs();
2385 /* Not quite true; I just maintain it */
2386 MODULE_AUTHOR("Olaf Kirch <okir@monad.swb.de>");
2387 MODULE_LICENSE("GPL");
2389 module_init(init_nfs_fs
)
2390 module_exit(exit_nfs_fs
)