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 super_block
*, 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 super_block
*sb
, struct kstatfs
*buf
)
539 struct nfs_server
*server
= NFS_SB(sb
);
540 unsigned char blockbits
;
541 unsigned long blockres
;
542 struct nfs_fh
*rootfh
= NFS_FH(sb
->s_root
->d_inode
);
543 struct nfs_fattr fattr
;
544 struct nfs_fsstat res
= {
551 error
= server
->rpc_ops
->statfs(server
, rootfh
, &res
);
552 buf
->f_type
= NFS_SUPER_MAGIC
;
557 * Current versions of glibc do not correctly handle the
558 * case where f_frsize != f_bsize. Eventually we want to
559 * report the value of wtmult in this field.
561 buf
->f_frsize
= sb
->s_blocksize
;
564 * On most *nix systems, f_blocks, f_bfree, and f_bavail
565 * are reported in units of f_frsize. Linux hasn't had
566 * an f_frsize field in its statfs struct until recently,
567 * thus historically Linux's sys_statfs reports these
568 * fields in units of f_bsize.
570 buf
->f_bsize
= sb
->s_blocksize
;
571 blockbits
= sb
->s_blocksize_bits
;
572 blockres
= (1 << blockbits
) - 1;
573 buf
->f_blocks
= (res
.tbytes
+ blockres
) >> blockbits
;
574 buf
->f_bfree
= (res
.fbytes
+ blockres
) >> blockbits
;
575 buf
->f_bavail
= (res
.abytes
+ blockres
) >> blockbits
;
577 buf
->f_files
= res
.tfiles
;
578 buf
->f_ffree
= res
.afiles
;
580 buf
->f_namelen
= server
->namelen
;
586 dprintk("%s: statfs error = %d\n", __FUNCTION__
, -error
);
587 buf
->f_bsize
= buf
->f_blocks
= buf
->f_bfree
= buf
->f_bavail
= -1;
592 static void nfs_show_mount_options(struct seq_file
*m
, struct nfs_server
*nfss
, int showdefaults
)
594 static struct proc_nfs_info
{
599 { NFS_MOUNT_SOFT
, ",soft", ",hard" },
600 { NFS_MOUNT_INTR
, ",intr", "" },
601 { NFS_MOUNT_NOCTO
, ",nocto", "" },
602 { NFS_MOUNT_NOAC
, ",noac", "" },
603 { NFS_MOUNT_NONLM
, ",nolock", "" },
604 { NFS_MOUNT_NOACL
, ",noacl", "" },
607 struct proc_nfs_info
*nfs_infop
;
611 seq_printf(m
, ",vers=%d", nfss
->rpc_ops
->version
);
612 seq_printf(m
, ",rsize=%d", nfss
->rsize
);
613 seq_printf(m
, ",wsize=%d", nfss
->wsize
);
614 if (nfss
->acregmin
!= 3*HZ
|| showdefaults
)
615 seq_printf(m
, ",acregmin=%d", nfss
->acregmin
/HZ
);
616 if (nfss
->acregmax
!= 60*HZ
|| showdefaults
)
617 seq_printf(m
, ",acregmax=%d", nfss
->acregmax
/HZ
);
618 if (nfss
->acdirmin
!= 30*HZ
|| showdefaults
)
619 seq_printf(m
, ",acdirmin=%d", nfss
->acdirmin
/HZ
);
620 if (nfss
->acdirmax
!= 60*HZ
|| showdefaults
)
621 seq_printf(m
, ",acdirmax=%d", nfss
->acdirmax
/HZ
);
622 for (nfs_infop
= nfs_info
; nfs_infop
->flag
; nfs_infop
++) {
623 if (nfss
->flags
& nfs_infop
->flag
)
624 seq_puts(m
, nfs_infop
->str
);
626 seq_puts(m
, nfs_infop
->nostr
);
628 switch (nfss
->client
->cl_xprt
->prot
) {
636 snprintf(buf
, sizeof(buf
), "%u", nfss
->client
->cl_xprt
->prot
);
639 seq_printf(m
, ",proto=%s", proto
);
640 seq_printf(m
, ",timeo=%lu", 10U * nfss
->retrans_timeo
/ HZ
);
641 seq_printf(m
, ",retrans=%u", nfss
->retrans_count
);
644 static int nfs_show_options(struct seq_file
*m
, struct vfsmount
*mnt
)
646 struct nfs_server
*nfss
= NFS_SB(mnt
->mnt_sb
);
648 nfs_show_mount_options(m
, nfss
, 0);
650 seq_puts(m
, ",addr=");
651 seq_escape(m
, nfss
->hostname
, " \t\n\\");
656 static int nfs_show_stats(struct seq_file
*m
, struct vfsmount
*mnt
)
659 struct nfs_server
*nfss
= NFS_SB(mnt
->mnt_sb
);
660 struct rpc_auth
*auth
= nfss
->client
->cl_auth
;
661 struct nfs_iostats totals
= { };
663 seq_printf(m
, "statvers=%s", NFS_IOSTAT_VERS
);
666 * Display all mount option settings
668 seq_printf(m
, "\n\topts:\t");
669 seq_puts(m
, mnt
->mnt_sb
->s_flags
& MS_RDONLY
? "ro" : "rw");
670 seq_puts(m
, mnt
->mnt_sb
->s_flags
& MS_SYNCHRONOUS
? ",sync" : "");
671 seq_puts(m
, mnt
->mnt_sb
->s_flags
& MS_NOATIME
? ",noatime" : "");
672 seq_puts(m
, mnt
->mnt_sb
->s_flags
& MS_NODIRATIME
? ",nodiratime" : "");
673 nfs_show_mount_options(m
, nfss
, 1);
675 seq_printf(m
, "\n\tage:\t%lu", (jiffies
- nfss
->mount_time
) / HZ
);
677 seq_printf(m
, "\n\tcaps:\t");
678 seq_printf(m
, "caps=0x%x", nfss
->caps
);
679 seq_printf(m
, ",wtmult=%d", nfss
->wtmult
);
680 seq_printf(m
, ",dtsize=%d", nfss
->dtsize
);
681 seq_printf(m
, ",bsize=%d", nfss
->bsize
);
682 seq_printf(m
, ",namelen=%d", nfss
->namelen
);
685 if (nfss
->rpc_ops
->version
== 4) {
686 seq_printf(m
, "\n\tnfsv4:\t");
687 seq_printf(m
, "bm0=0x%x", nfss
->attr_bitmask
[0]);
688 seq_printf(m
, ",bm1=0x%x", nfss
->attr_bitmask
[1]);
689 seq_printf(m
, ",acl=0x%x", nfss
->acl_bitmask
);
694 * Display security flavor in effect for this mount
696 seq_printf(m
, "\n\tsec:\tflavor=%d", auth
->au_ops
->au_flavor
);
698 seq_printf(m
, ",pseudoflavor=%d", auth
->au_flavor
);
701 * Display superblock I/O counters
703 for_each_possible_cpu(cpu
) {
704 struct nfs_iostats
*stats
;
707 stats
= per_cpu_ptr(nfss
->io_stats
, cpu
);
709 for (i
= 0; i
< __NFSIOS_COUNTSMAX
; i
++)
710 totals
.events
[i
] += stats
->events
[i
];
711 for (i
= 0; i
< __NFSIOS_BYTESMAX
; i
++)
712 totals
.bytes
[i
] += stats
->bytes
[i
];
717 seq_printf(m
, "\n\tevents:\t");
718 for (i
= 0; i
< __NFSIOS_COUNTSMAX
; i
++)
719 seq_printf(m
, "%lu ", totals
.events
[i
]);
720 seq_printf(m
, "\n\tbytes:\t");
721 for (i
= 0; i
< __NFSIOS_BYTESMAX
; i
++)
722 seq_printf(m
, "%Lu ", totals
.bytes
[i
]);
725 rpc_print_iostats(m
, nfss
->client
);
731 * nfs_sync_mapping - helper to flush all mmapped dirty data to disk
733 int nfs_sync_mapping(struct address_space
*mapping
)
737 if (mapping
->nrpages
== 0)
739 unmap_mapping_range(mapping
, 0, 0, 0);
740 ret
= filemap_write_and_wait(mapping
);
743 ret
= nfs_wb_all(mapping
->host
);
749 * Invalidate the local caches
751 static void nfs_zap_caches_locked(struct inode
*inode
)
753 struct nfs_inode
*nfsi
= NFS_I(inode
);
754 int mode
= inode
->i_mode
;
756 nfs_inc_stats(inode
, NFSIOS_ATTRINVALIDATE
);
758 NFS_ATTRTIMEO(inode
) = NFS_MINATTRTIMEO(inode
);
759 NFS_ATTRTIMEO_UPDATE(inode
) = jiffies
;
761 memset(NFS_COOKIEVERF(inode
), 0, sizeof(NFS_COOKIEVERF(inode
)));
762 if (S_ISREG(mode
) || S_ISDIR(mode
) || S_ISLNK(mode
))
763 nfsi
->cache_validity
|= NFS_INO_INVALID_ATTR
|NFS_INO_INVALID_DATA
|NFS_INO_INVALID_ACCESS
|NFS_INO_INVALID_ACL
|NFS_INO_REVAL_PAGECACHE
;
765 nfsi
->cache_validity
|= NFS_INO_INVALID_ATTR
|NFS_INO_INVALID_ACCESS
|NFS_INO_INVALID_ACL
|NFS_INO_REVAL_PAGECACHE
;
768 void nfs_zap_caches(struct inode
*inode
)
770 spin_lock(&inode
->i_lock
);
771 nfs_zap_caches_locked(inode
);
772 spin_unlock(&inode
->i_lock
);
775 static void nfs_zap_acl_cache(struct inode
*inode
)
777 void (*clear_acl_cache
)(struct inode
*);
779 clear_acl_cache
= NFS_PROTO(inode
)->clear_acl_cache
;
780 if (clear_acl_cache
!= NULL
)
781 clear_acl_cache(inode
);
782 spin_lock(&inode
->i_lock
);
783 NFS_I(inode
)->cache_validity
&= ~NFS_INO_INVALID_ACL
;
784 spin_unlock(&inode
->i_lock
);
788 * Invalidate, but do not unhash, the inode.
789 * NB: must be called with inode->i_lock held!
791 static void nfs_invalidate_inode(struct inode
*inode
)
793 set_bit(NFS_INO_STALE
, &NFS_FLAGS(inode
));
794 nfs_zap_caches_locked(inode
);
797 struct nfs_find_desc
{
799 struct nfs_fattr
*fattr
;
803 * In NFSv3 we can have 64bit inode numbers. In order to support
804 * this, and re-exported directories (also seen in NFSv2)
805 * we are forced to allow 2 different inodes to have the same
809 nfs_find_actor(struct inode
*inode
, void *opaque
)
811 struct nfs_find_desc
*desc
= (struct nfs_find_desc
*)opaque
;
812 struct nfs_fh
*fh
= desc
->fh
;
813 struct nfs_fattr
*fattr
= desc
->fattr
;
815 if (NFS_FILEID(inode
) != fattr
->fileid
)
817 if (nfs_compare_fh(NFS_FH(inode
), fh
))
819 if (is_bad_inode(inode
) || NFS_STALE(inode
))
825 nfs_init_locked(struct inode
*inode
, void *opaque
)
827 struct nfs_find_desc
*desc
= (struct nfs_find_desc
*)opaque
;
828 struct nfs_fattr
*fattr
= desc
->fattr
;
830 NFS_FILEID(inode
) = fattr
->fileid
;
831 nfs_copy_fh(NFS_FH(inode
), desc
->fh
);
835 /* Don't use READDIRPLUS on directories that we believe are too large */
836 #define NFS_LIMIT_READDIRPLUS (8*PAGE_SIZE)
839 * This is our front-end to iget that looks up inodes by file handle
840 * instead of inode number.
843 nfs_fhget(struct super_block
*sb
, struct nfs_fh
*fh
, struct nfs_fattr
*fattr
)
845 struct nfs_find_desc desc
= {
849 struct inode
*inode
= ERR_PTR(-ENOENT
);
852 if ((fattr
->valid
& NFS_ATTR_FATTR
) == 0)
856 printk("NFS: Buggy server - nlink == 0!\n");
860 hash
= nfs_fattr_to_ino_t(fattr
);
862 inode
= iget5_locked(sb
, hash
, nfs_find_actor
, nfs_init_locked
, &desc
);
864 inode
= ERR_PTR(-ENOMEM
);
868 if (inode
->i_state
& I_NEW
) {
869 struct nfs_inode
*nfsi
= NFS_I(inode
);
871 /* We set i_ino for the few things that still rely on it,
875 /* We can't support update_atime(), since the server will reset it */
876 inode
->i_flags
|= S_NOATIME
|S_NOCMTIME
;
877 inode
->i_mode
= fattr
->mode
;
878 /* Why so? Because we want revalidate for devices/FIFOs, and
879 * that's precisely what we have in nfs_file_inode_operations.
881 inode
->i_op
= NFS_SB(sb
)->rpc_ops
->file_inode_ops
;
882 if (S_ISREG(inode
->i_mode
)) {
883 inode
->i_fop
= &nfs_file_operations
;
884 inode
->i_data
.a_ops
= &nfs_file_aops
;
885 inode
->i_data
.backing_dev_info
= &NFS_SB(sb
)->backing_dev_info
;
886 } else if (S_ISDIR(inode
->i_mode
)) {
887 inode
->i_op
= NFS_SB(sb
)->rpc_ops
->dir_inode_ops
;
888 inode
->i_fop
= &nfs_dir_operations
;
889 if (nfs_server_capable(inode
, NFS_CAP_READDIRPLUS
)
890 && fattr
->size
<= NFS_LIMIT_READDIRPLUS
)
891 set_bit(NFS_INO_ADVISE_RDPLUS
, &NFS_FLAGS(inode
));
892 } else if (S_ISLNK(inode
->i_mode
))
893 inode
->i_op
= &nfs_symlink_inode_operations
;
895 init_special_inode(inode
, inode
->i_mode
, fattr
->rdev
);
897 nfsi
->read_cache_jiffies
= fattr
->time_start
;
898 nfsi
->last_updated
= jiffies
;
899 inode
->i_atime
= fattr
->atime
;
900 inode
->i_mtime
= fattr
->mtime
;
901 inode
->i_ctime
= fattr
->ctime
;
902 if (fattr
->valid
& NFS_ATTR_FATTR_V4
)
903 nfsi
->change_attr
= fattr
->change_attr
;
904 inode
->i_size
= nfs_size_to_loff_t(fattr
->size
);
905 inode
->i_nlink
= fattr
->nlink
;
906 inode
->i_uid
= fattr
->uid
;
907 inode
->i_gid
= fattr
->gid
;
908 if (fattr
->valid
& (NFS_ATTR_FATTR_V3
| NFS_ATTR_FATTR_V4
)) {
910 * report the blocks in 512byte units
912 inode
->i_blocks
= nfs_calc_block_size(fattr
->du
.nfs3
.used
);
913 inode
->i_blksize
= inode
->i_sb
->s_blocksize
;
915 inode
->i_blocks
= fattr
->du
.nfs2
.blocks
;
916 inode
->i_blksize
= fattr
->du
.nfs2
.blocksize
;
918 nfsi
->attrtimeo
= NFS_MINATTRTIMEO(inode
);
919 nfsi
->attrtimeo_timestamp
= jiffies
;
920 memset(nfsi
->cookieverf
, 0, sizeof(nfsi
->cookieverf
));
921 nfsi
->cache_access
.cred
= NULL
;
923 unlock_new_inode(inode
);
925 nfs_refresh_inode(inode
, fattr
);
926 dprintk("NFS: nfs_fhget(%s/%Ld ct=%d)\n",
928 (long long)NFS_FILEID(inode
),
929 atomic_read(&inode
->i_count
));
935 dprintk("nfs_fhget: iget failed with error %ld\n", PTR_ERR(inode
));
939 #define NFS_VALID_ATTRS (ATTR_MODE|ATTR_UID|ATTR_GID|ATTR_SIZE|ATTR_ATIME|ATTR_ATIME_SET|ATTR_MTIME|ATTR_MTIME_SET)
942 nfs_setattr(struct dentry
*dentry
, struct iattr
*attr
)
944 struct inode
*inode
= dentry
->d_inode
;
945 struct nfs_fattr fattr
;
948 nfs_inc_stats(inode
, NFSIOS_VFSSETATTR
);
950 if (attr
->ia_valid
& ATTR_SIZE
) {
951 if (!S_ISREG(inode
->i_mode
) || attr
->ia_size
== i_size_read(inode
))
952 attr
->ia_valid
&= ~ATTR_SIZE
;
955 /* Optimization: if the end result is no change, don't RPC */
956 attr
->ia_valid
&= NFS_VALID_ATTRS
;
957 if (attr
->ia_valid
== 0)
961 nfs_begin_data_update(inode
);
962 /* Write all dirty data */
963 filemap_write_and_wait(inode
->i_mapping
);
966 * Return any delegations if we're going to change ACLs
968 if ((attr
->ia_valid
& (ATTR_MODE
|ATTR_UID
|ATTR_GID
)) != 0)
969 nfs_inode_return_delegation(inode
);
970 error
= NFS_PROTO(inode
)->setattr(dentry
, &fattr
, attr
);
972 nfs_refresh_inode(inode
, &fattr
);
973 nfs_end_data_update(inode
);
979 * nfs_setattr_update_inode - Update inode metadata after a setattr call.
980 * @inode: pointer to struct inode
981 * @attr: pointer to struct iattr
983 * Note: we do this in the *proc.c in order to ensure that
984 * it works for things like exclusive creates too.
986 void nfs_setattr_update_inode(struct inode
*inode
, struct iattr
*attr
)
988 if ((attr
->ia_valid
& (ATTR_MODE
|ATTR_UID
|ATTR_GID
)) != 0) {
989 if ((attr
->ia_valid
& ATTR_MODE
) != 0) {
990 int mode
= attr
->ia_mode
& S_IALLUGO
;
991 mode
|= inode
->i_mode
& ~S_IALLUGO
;
992 inode
->i_mode
= mode
;
994 if ((attr
->ia_valid
& ATTR_UID
) != 0)
995 inode
->i_uid
= attr
->ia_uid
;
996 if ((attr
->ia_valid
& ATTR_GID
) != 0)
997 inode
->i_gid
= attr
->ia_gid
;
998 spin_lock(&inode
->i_lock
);
999 NFS_I(inode
)->cache_validity
|= NFS_INO_INVALID_ACCESS
|NFS_INO_INVALID_ACL
;
1000 spin_unlock(&inode
->i_lock
);
1002 if ((attr
->ia_valid
& ATTR_SIZE
) != 0) {
1003 nfs_inc_stats(inode
, NFSIOS_SETATTRTRUNC
);
1004 inode
->i_size
= attr
->ia_size
;
1005 vmtruncate(inode
, attr
->ia_size
);
1009 static int nfs_wait_schedule(void *word
)
1011 if (signal_pending(current
))
1012 return -ERESTARTSYS
;
1018 * Wait for the inode to get unlocked.
1020 static int nfs_wait_on_inode(struct inode
*inode
)
1022 struct rpc_clnt
*clnt
= NFS_CLIENT(inode
);
1023 struct nfs_inode
*nfsi
= NFS_I(inode
);
1027 rpc_clnt_sigmask(clnt
, &oldmask
);
1028 error
= wait_on_bit_lock(&nfsi
->flags
, NFS_INO_REVALIDATING
,
1029 nfs_wait_schedule
, TASK_INTERRUPTIBLE
);
1030 rpc_clnt_sigunmask(clnt
, &oldmask
);
1035 static void nfs_wake_up_inode(struct inode
*inode
)
1037 struct nfs_inode
*nfsi
= NFS_I(inode
);
1039 clear_bit(NFS_INO_REVALIDATING
, &nfsi
->flags
);
1040 smp_mb__after_clear_bit();
1041 wake_up_bit(&nfsi
->flags
, NFS_INO_REVALIDATING
);
1044 int nfs_getattr(struct vfsmount
*mnt
, struct dentry
*dentry
, struct kstat
*stat
)
1046 struct inode
*inode
= dentry
->d_inode
;
1047 int need_atime
= NFS_I(inode
)->cache_validity
& NFS_INO_INVALID_ATIME
;
1050 /* Flush out writes to the server in order to update c/mtime */
1051 nfs_sync_inode_wait(inode
, 0, 0, FLUSH_NOCOMMIT
);
1054 * We may force a getattr if the user cares about atime.
1056 * Note that we only have to check the vfsmount flags here:
1057 * - NFS always sets S_NOATIME by so checking it would give a
1059 * - NFS never sets MS_NOATIME or MS_NODIRATIME so there is
1060 * no point in checking those.
1062 if ((mnt
->mnt_flags
& MNT_NOATIME
) ||
1063 ((mnt
->mnt_flags
& MNT_NODIRATIME
) && S_ISDIR(inode
->i_mode
)))
1067 err
= __nfs_revalidate_inode(NFS_SERVER(inode
), inode
);
1069 err
= nfs_revalidate_inode(NFS_SERVER(inode
), inode
);
1071 generic_fillattr(inode
, stat
);
1075 static struct nfs_open_context
*alloc_nfs_open_context(struct vfsmount
*mnt
, struct dentry
*dentry
, struct rpc_cred
*cred
)
1077 struct nfs_open_context
*ctx
;
1079 ctx
= (struct nfs_open_context
*)kmalloc(sizeof(*ctx
), GFP_KERNEL
);
1081 atomic_set(&ctx
->count
, 1);
1082 ctx
->dentry
= dget(dentry
);
1083 ctx
->vfsmnt
= mntget(mnt
);
1084 ctx
->cred
= get_rpccred(cred
);
1086 ctx
->lockowner
= current
->files
;
1088 ctx
->dir_cookie
= 0;
1093 struct nfs_open_context
*get_nfs_open_context(struct nfs_open_context
*ctx
)
1096 atomic_inc(&ctx
->count
);
1100 void put_nfs_open_context(struct nfs_open_context
*ctx
)
1102 if (atomic_dec_and_test(&ctx
->count
)) {
1103 if (!list_empty(&ctx
->list
)) {
1104 struct inode
*inode
= ctx
->dentry
->d_inode
;
1105 spin_lock(&inode
->i_lock
);
1106 list_del(&ctx
->list
);
1107 spin_unlock(&inode
->i_lock
);
1109 if (ctx
->state
!= NULL
)
1110 nfs4_close_state(ctx
->state
, ctx
->mode
);
1111 if (ctx
->cred
!= NULL
)
1112 put_rpccred(ctx
->cred
);
1114 mntput(ctx
->vfsmnt
);
1120 * Ensure that mmap has a recent RPC credential for use when writing out
1123 static void nfs_file_set_open_context(struct file
*filp
, struct nfs_open_context
*ctx
)
1125 struct inode
*inode
= filp
->f_dentry
->d_inode
;
1126 struct nfs_inode
*nfsi
= NFS_I(inode
);
1128 filp
->private_data
= get_nfs_open_context(ctx
);
1129 spin_lock(&inode
->i_lock
);
1130 list_add(&ctx
->list
, &nfsi
->open_files
);
1131 spin_unlock(&inode
->i_lock
);
1135 * Given an inode, search for an open context with the desired characteristics
1137 struct nfs_open_context
*nfs_find_open_context(struct inode
*inode
, struct rpc_cred
*cred
, int mode
)
1139 struct nfs_inode
*nfsi
= NFS_I(inode
);
1140 struct nfs_open_context
*pos
, *ctx
= NULL
;
1142 spin_lock(&inode
->i_lock
);
1143 list_for_each_entry(pos
, &nfsi
->open_files
, list
) {
1144 if (cred
!= NULL
&& pos
->cred
!= cred
)
1146 if ((pos
->mode
& mode
) == mode
) {
1147 ctx
= get_nfs_open_context(pos
);
1151 spin_unlock(&inode
->i_lock
);
1155 static void nfs_file_clear_open_context(struct file
*filp
)
1157 struct inode
*inode
= filp
->f_dentry
->d_inode
;
1158 struct nfs_open_context
*ctx
= (struct nfs_open_context
*)filp
->private_data
;
1161 filp
->private_data
= NULL
;
1162 spin_lock(&inode
->i_lock
);
1163 list_move_tail(&ctx
->list
, &NFS_I(inode
)->open_files
);
1164 spin_unlock(&inode
->i_lock
);
1165 put_nfs_open_context(ctx
);
1170 * These allocate and release file read/write context information.
1172 int nfs_open(struct inode
*inode
, struct file
*filp
)
1174 struct nfs_open_context
*ctx
;
1175 struct rpc_cred
*cred
;
1177 cred
= rpcauth_lookupcred(NFS_CLIENT(inode
)->cl_auth
, 0);
1179 return PTR_ERR(cred
);
1180 ctx
= alloc_nfs_open_context(filp
->f_vfsmnt
, filp
->f_dentry
, cred
);
1184 ctx
->mode
= filp
->f_mode
;
1185 nfs_file_set_open_context(filp
, ctx
);
1186 put_nfs_open_context(ctx
);
1190 int nfs_release(struct inode
*inode
, struct file
*filp
)
1192 nfs_file_clear_open_context(filp
);
1197 * This function is called whenever some part of NFS notices that
1198 * the cached attributes have to be refreshed.
1201 __nfs_revalidate_inode(struct nfs_server
*server
, struct inode
*inode
)
1203 int status
= -ESTALE
;
1204 struct nfs_fattr fattr
;
1205 struct nfs_inode
*nfsi
= NFS_I(inode
);
1207 dfprintk(PAGECACHE
, "NFS: revalidating (%s/%Ld)\n",
1208 inode
->i_sb
->s_id
, (long long)NFS_FILEID(inode
));
1211 if (!inode
|| is_bad_inode(inode
))
1213 if (NFS_STALE(inode
))
1216 status
= nfs_wait_on_inode(inode
);
1219 if (NFS_STALE(inode
)) {
1221 /* Do we trust the cached ESTALE? */
1222 if (NFS_ATTRTIMEO(inode
) != 0) {
1223 if (nfsi
->cache_validity
& (NFS_INO_INVALID_ATTR
|NFS_INO_INVALID_DATA
|NFS_INO_INVALID_ATIME
)) {
1230 status
= NFS_PROTO(inode
)->getattr(server
, NFS_FH(inode
), &fattr
);
1232 dfprintk(PAGECACHE
, "nfs_revalidate_inode: (%s/%Ld) getattr failed, error=%d\n",
1234 (long long)NFS_FILEID(inode
), status
);
1235 if (status
== -ESTALE
) {
1236 nfs_zap_caches(inode
);
1237 if (!S_ISDIR(inode
->i_mode
))
1238 set_bit(NFS_INO_STALE
, &NFS_FLAGS(inode
));
1243 spin_lock(&inode
->i_lock
);
1244 status
= nfs_update_inode(inode
, &fattr
);
1246 spin_unlock(&inode
->i_lock
);
1247 dfprintk(PAGECACHE
, "nfs_revalidate_inode: (%s/%Ld) refresh failed, error=%d\n",
1249 (long long)NFS_FILEID(inode
), status
);
1252 spin_unlock(&inode
->i_lock
);
1254 nfs_revalidate_mapping(inode
, inode
->i_mapping
);
1256 if (nfsi
->cache_validity
& NFS_INO_INVALID_ACL
)
1257 nfs_zap_acl_cache(inode
);
1259 dfprintk(PAGECACHE
, "NFS: (%s/%Ld) revalidation complete\n",
1261 (long long)NFS_FILEID(inode
));
1264 nfs_wake_up_inode(inode
);
1271 int nfs_attribute_timeout(struct inode
*inode
)
1273 struct nfs_inode
*nfsi
= NFS_I(inode
);
1275 if (nfs_have_delegation(inode
, FMODE_READ
))
1277 return time_after(jiffies
, nfsi
->read_cache_jiffies
+nfsi
->attrtimeo
);
1281 * nfs_revalidate_inode - Revalidate the inode attributes
1282 * @server - pointer to nfs_server struct
1283 * @inode - pointer to inode struct
1285 * Updates inode attribute information by retrieving the data from the server.
1287 int nfs_revalidate_inode(struct nfs_server
*server
, struct inode
*inode
)
1289 nfs_inc_stats(inode
, NFSIOS_INODEREVALIDATE
);
1290 if (!(NFS_I(inode
)->cache_validity
& (NFS_INO_INVALID_ATTR
|NFS_INO_INVALID_DATA
))
1291 && !nfs_attribute_timeout(inode
))
1292 return NFS_STALE(inode
) ? -ESTALE
: 0;
1293 return __nfs_revalidate_inode(server
, inode
);
1297 * nfs_revalidate_mapping - Revalidate the pagecache
1298 * @inode - pointer to host inode
1299 * @mapping - pointer to mapping
1301 void nfs_revalidate_mapping(struct inode
*inode
, struct address_space
*mapping
)
1303 struct nfs_inode
*nfsi
= NFS_I(inode
);
1305 if (nfsi
->cache_validity
& NFS_INO_INVALID_DATA
) {
1306 nfs_inc_stats(inode
, NFSIOS_DATAINVALIDATE
);
1307 if (S_ISREG(inode
->i_mode
))
1308 nfs_sync_mapping(mapping
);
1309 invalidate_inode_pages2(mapping
);
1311 spin_lock(&inode
->i_lock
);
1312 nfsi
->cache_validity
&= ~NFS_INO_INVALID_DATA
;
1313 if (S_ISDIR(inode
->i_mode
)) {
1314 memset(nfsi
->cookieverf
, 0, sizeof(nfsi
->cookieverf
));
1315 /* This ensures we revalidate child dentries */
1316 nfsi
->cache_change_attribute
= jiffies
;
1318 spin_unlock(&inode
->i_lock
);
1320 dfprintk(PAGECACHE
, "NFS: (%s/%Ld) data cache invalidated\n",
1322 (long long)NFS_FILEID(inode
));
1327 * nfs_begin_data_update
1328 * @inode - pointer to inode
1329 * Declare that a set of operations will update file data on the server
1331 void nfs_begin_data_update(struct inode
*inode
)
1333 atomic_inc(&NFS_I(inode
)->data_updates
);
1337 * nfs_end_data_update
1338 * @inode - pointer to inode
1339 * Declare end of the operations that will update file data
1340 * This will mark the inode as immediately needing revalidation
1341 * of its attribute cache.
1343 void nfs_end_data_update(struct inode
*inode
)
1345 struct nfs_inode
*nfsi
= NFS_I(inode
);
1347 if (!nfs_have_delegation(inode
, FMODE_READ
)) {
1348 /* Directories and symlinks: invalidate page cache */
1349 if (S_ISDIR(inode
->i_mode
) || S_ISLNK(inode
->i_mode
)) {
1350 spin_lock(&inode
->i_lock
);
1351 nfsi
->cache_validity
|= NFS_INO_INVALID_DATA
;
1352 spin_unlock(&inode
->i_lock
);
1355 nfsi
->cache_change_attribute
= jiffies
;
1356 atomic_dec(&nfsi
->data_updates
);
1359 static void nfs_wcc_update_inode(struct inode
*inode
, struct nfs_fattr
*fattr
)
1361 struct nfs_inode
*nfsi
= NFS_I(inode
);
1363 if ((fattr
->valid
& NFS_ATTR_PRE_CHANGE
) != 0
1364 && nfsi
->change_attr
== fattr
->pre_change_attr
) {
1365 nfsi
->change_attr
= fattr
->change_attr
;
1366 nfsi
->cache_change_attribute
= jiffies
;
1369 /* If we have atomic WCC data, we may update some attributes */
1370 if ((fattr
->valid
& NFS_ATTR_WCC
) != 0) {
1371 if (timespec_equal(&inode
->i_ctime
, &fattr
->pre_ctime
)) {
1372 memcpy(&inode
->i_ctime
, &fattr
->ctime
, sizeof(inode
->i_ctime
));
1373 nfsi
->cache_change_attribute
= jiffies
;
1375 if (timespec_equal(&inode
->i_mtime
, &fattr
->pre_mtime
)) {
1376 memcpy(&inode
->i_mtime
, &fattr
->mtime
, sizeof(inode
->i_mtime
));
1377 nfsi
->cache_change_attribute
= jiffies
;
1379 if (inode
->i_size
== fattr
->pre_size
&& nfsi
->npages
== 0) {
1380 inode
->i_size
= fattr
->size
;
1381 nfsi
->cache_change_attribute
= jiffies
;
1387 * nfs_check_inode_attributes - verify consistency of the inode attribute cache
1388 * @inode - pointer to inode
1389 * @fattr - updated attributes
1391 * Verifies the attribute cache. If we have just changed the attributes,
1392 * so that fattr carries weak cache consistency data, then it may
1393 * also update the ctime/mtime/change_attribute.
1395 static int nfs_check_inode_attributes(struct inode
*inode
, struct nfs_fattr
*fattr
)
1397 struct nfs_inode
*nfsi
= NFS_I(inode
);
1398 loff_t cur_size
, new_isize
;
1402 if ((fattr
->valid
& NFS_ATTR_FATTR
) == 0)
1405 /* Has the inode gone and changed behind our back? */
1406 if (nfsi
->fileid
!= fattr
->fileid
1407 || (inode
->i_mode
& S_IFMT
) != (fattr
->mode
& S_IFMT
)) {
1411 /* Are we in the process of updating data on the server? */
1412 data_unstable
= nfs_caches_unstable(inode
);
1414 /* Do atomic weak cache consistency updates */
1415 nfs_wcc_update_inode(inode
, fattr
);
1417 if ((fattr
->valid
& NFS_ATTR_FATTR_V4
) != 0) {
1418 if (nfsi
->change_attr
== fattr
->change_attr
)
1420 nfsi
->cache_validity
|= NFS_INO_INVALID_ATTR
;
1422 nfsi
->cache_validity
|= NFS_INO_REVAL_PAGECACHE
;
1425 /* Verify a few of the more important attributes */
1426 if (!timespec_equal(&inode
->i_mtime
, &fattr
->mtime
)) {
1427 nfsi
->cache_validity
|= NFS_INO_INVALID_ATTR
;
1429 nfsi
->cache_validity
|= NFS_INO_REVAL_PAGECACHE
;
1432 cur_size
= i_size_read(inode
);
1433 new_isize
= nfs_size_to_loff_t(fattr
->size
);
1434 if (cur_size
!= new_isize
&& nfsi
->npages
== 0)
1435 nfsi
->cache_validity
|= NFS_INO_INVALID_ATTR
|NFS_INO_REVAL_PAGECACHE
;
1437 /* Have any file permissions changed? */
1438 if ((inode
->i_mode
& S_IALLUGO
) != (fattr
->mode
& S_IALLUGO
)
1439 || inode
->i_uid
!= fattr
->uid
1440 || inode
->i_gid
!= fattr
->gid
)
1441 nfsi
->cache_validity
|= NFS_INO_INVALID_ATTR
| NFS_INO_INVALID_ACCESS
| NFS_INO_INVALID_ACL
;
1443 /* Has the link count changed? */
1444 if (inode
->i_nlink
!= fattr
->nlink
)
1445 nfsi
->cache_validity
|= NFS_INO_INVALID_ATTR
;
1448 if (!timespec_equal(&inode
->i_atime
, &fattr
->atime
))
1449 nfsi
->cache_validity
|= NFS_INO_INVALID_ATIME
;
1451 nfsi
->read_cache_jiffies
= fattr
->time_start
;
1456 * nfs_refresh_inode - try to update the inode attribute cache
1457 * @inode - pointer to inode
1458 * @fattr - updated attributes
1460 * Check that an RPC call that returned attributes has not overlapped with
1461 * other recent updates of the inode metadata, then decide whether it is
1462 * safe to do a full update of the inode attributes, or whether just to
1463 * call nfs_check_inode_attributes.
1465 int nfs_refresh_inode(struct inode
*inode
, struct nfs_fattr
*fattr
)
1467 struct nfs_inode
*nfsi
= NFS_I(inode
);
1470 if ((fattr
->valid
& NFS_ATTR_FATTR
) == 0)
1472 spin_lock(&inode
->i_lock
);
1473 nfsi
->cache_validity
&= ~NFS_INO_REVAL_PAGECACHE
;
1474 if (time_after(fattr
->time_start
, nfsi
->last_updated
))
1475 status
= nfs_update_inode(inode
, fattr
);
1477 status
= nfs_check_inode_attributes(inode
, fattr
);
1479 spin_unlock(&inode
->i_lock
);
1484 * nfs_post_op_update_inode - try to update the inode attribute cache
1485 * @inode - pointer to inode
1486 * @fattr - updated attributes
1488 * After an operation that has changed the inode metadata, mark the
1489 * attribute cache as being invalid, then try to update it.
1491 int nfs_post_op_update_inode(struct inode
*inode
, struct nfs_fattr
*fattr
)
1493 struct nfs_inode
*nfsi
= NFS_I(inode
);
1496 spin_lock(&inode
->i_lock
);
1497 if (unlikely((fattr
->valid
& NFS_ATTR_FATTR
) == 0)) {
1498 nfsi
->cache_validity
|= NFS_INO_INVALID_ATTR
| NFS_INO_INVALID_ACCESS
;
1501 status
= nfs_update_inode(inode
, fattr
);
1503 spin_unlock(&inode
->i_lock
);
1508 * Many nfs protocol calls return the new file attributes after
1509 * an operation. Here we update the inode to reflect the state
1510 * of the server's inode.
1512 * This is a bit tricky because we have to make sure all dirty pages
1513 * have been sent off to the server before calling invalidate_inode_pages.
1514 * To make sure no other process adds more write requests while we try
1515 * our best to flush them, we make them sleep during the attribute refresh.
1517 * A very similar scenario holds for the dir cache.
1519 static int nfs_update_inode(struct inode
*inode
, struct nfs_fattr
*fattr
)
1521 struct nfs_inode
*nfsi
= NFS_I(inode
);
1522 loff_t cur_isize
, new_isize
;
1523 unsigned int invalid
= 0;
1526 dfprintk(VFS
, "NFS: %s(%s/%ld ct=%d info=0x%x)\n",
1527 __FUNCTION__
, inode
->i_sb
->s_id
, inode
->i_ino
,
1528 atomic_read(&inode
->i_count
), fattr
->valid
);
1530 if ((fattr
->valid
& NFS_ATTR_FATTR
) == 0)
1533 if (nfsi
->fileid
!= fattr
->fileid
)
1537 * Make sure the inode's type hasn't changed.
1539 if ((inode
->i_mode
& S_IFMT
) != (fattr
->mode
& S_IFMT
))
1543 * Update the read time so we don't revalidate too often.
1545 nfsi
->read_cache_jiffies
= fattr
->time_start
;
1546 nfsi
->last_updated
= jiffies
;
1548 /* Are we racing with known updates of the metadata on the server? */
1549 data_stable
= nfs_verify_change_attribute(inode
, fattr
->time_start
);
1551 nfsi
->cache_validity
&= ~(NFS_INO_INVALID_ATTR
|NFS_INO_INVALID_ATIME
);
1553 /* Do atomic weak cache consistency updates */
1554 nfs_wcc_update_inode(inode
, fattr
);
1556 /* Check if our cached file size is stale */
1557 new_isize
= nfs_size_to_loff_t(fattr
->size
);
1558 cur_isize
= i_size_read(inode
);
1559 if (new_isize
!= cur_isize
) {
1560 /* Do we perhaps have any outstanding writes? */
1561 if (nfsi
->npages
== 0) {
1562 /* No, but did we race with nfs_end_data_update()? */
1564 inode
->i_size
= new_isize
;
1565 invalid
|= NFS_INO_INVALID_DATA
;
1567 invalid
|= NFS_INO_INVALID_ATTR
;
1568 } else if (new_isize
> cur_isize
) {
1569 inode
->i_size
= new_isize
;
1570 invalid
|= NFS_INO_INVALID_ATTR
|NFS_INO_INVALID_DATA
;
1572 nfsi
->cache_change_attribute
= jiffies
;
1573 dprintk("NFS: isize change on server for file %s/%ld\n",
1574 inode
->i_sb
->s_id
, inode
->i_ino
);
1577 /* Check if the mtime agrees */
1578 if (!timespec_equal(&inode
->i_mtime
, &fattr
->mtime
)) {
1579 memcpy(&inode
->i_mtime
, &fattr
->mtime
, sizeof(inode
->i_mtime
));
1580 dprintk("NFS: mtime change on server for file %s/%ld\n",
1581 inode
->i_sb
->s_id
, inode
->i_ino
);
1582 invalid
|= NFS_INO_INVALID_ATTR
|NFS_INO_INVALID_DATA
;
1583 nfsi
->cache_change_attribute
= jiffies
;
1586 /* If ctime has changed we should definitely clear access+acl caches */
1587 if (!timespec_equal(&inode
->i_ctime
, &fattr
->ctime
)) {
1588 invalid
|= NFS_INO_INVALID_ACCESS
|NFS_INO_INVALID_ACL
;
1589 memcpy(&inode
->i_ctime
, &fattr
->ctime
, sizeof(inode
->i_ctime
));
1590 nfsi
->cache_change_attribute
= jiffies
;
1592 memcpy(&inode
->i_atime
, &fattr
->atime
, sizeof(inode
->i_atime
));
1594 if ((inode
->i_mode
& S_IALLUGO
) != (fattr
->mode
& S_IALLUGO
) ||
1595 inode
->i_uid
!= fattr
->uid
||
1596 inode
->i_gid
!= fattr
->gid
)
1597 invalid
|= NFS_INO_INVALID_ATTR
|NFS_INO_INVALID_ACCESS
|NFS_INO_INVALID_ACL
;
1599 inode
->i_mode
= fattr
->mode
;
1600 inode
->i_nlink
= fattr
->nlink
;
1601 inode
->i_uid
= fattr
->uid
;
1602 inode
->i_gid
= fattr
->gid
;
1604 if (fattr
->valid
& (NFS_ATTR_FATTR_V3
| NFS_ATTR_FATTR_V4
)) {
1606 * report the blocks in 512byte units
1608 inode
->i_blocks
= nfs_calc_block_size(fattr
->du
.nfs3
.used
);
1609 inode
->i_blksize
= inode
->i_sb
->s_blocksize
;
1611 inode
->i_blocks
= fattr
->du
.nfs2
.blocks
;
1612 inode
->i_blksize
= fattr
->du
.nfs2
.blocksize
;
1615 if ((fattr
->valid
& NFS_ATTR_FATTR_V4
)) {
1616 if (nfsi
->change_attr
!= fattr
->change_attr
) {
1617 dprintk("NFS: change_attr change on server for file %s/%ld\n",
1618 inode
->i_sb
->s_id
, inode
->i_ino
);
1619 nfsi
->change_attr
= fattr
->change_attr
;
1620 invalid
|= NFS_INO_INVALID_ATTR
|NFS_INO_INVALID_DATA
|NFS_INO_INVALID_ACCESS
|NFS_INO_INVALID_ACL
;
1621 nfsi
->cache_change_attribute
= jiffies
;
1623 invalid
&= ~(NFS_INO_INVALID_ATTR
|NFS_INO_INVALID_DATA
);
1626 /* Update attrtimeo value if we're out of the unstable period */
1627 if (invalid
& NFS_INO_INVALID_ATTR
) {
1628 nfs_inc_stats(inode
, NFSIOS_ATTRINVALIDATE
);
1629 nfsi
->attrtimeo
= NFS_MINATTRTIMEO(inode
);
1630 nfsi
->attrtimeo_timestamp
= jiffies
;
1631 } else if (time_after(jiffies
, nfsi
->attrtimeo_timestamp
+nfsi
->attrtimeo
)) {
1632 if ((nfsi
->attrtimeo
<<= 1) > NFS_MAXATTRTIMEO(inode
))
1633 nfsi
->attrtimeo
= NFS_MAXATTRTIMEO(inode
);
1634 nfsi
->attrtimeo_timestamp
= jiffies
;
1636 /* Don't invalidate the data if we were to blame */
1637 if (!(S_ISREG(inode
->i_mode
) || S_ISDIR(inode
->i_mode
)
1638 || S_ISLNK(inode
->i_mode
)))
1639 invalid
&= ~NFS_INO_INVALID_DATA
;
1641 invalid
&= ~(NFS_INO_INVALID_ATTR
|NFS_INO_INVALID_ATIME
|NFS_INO_REVAL_PAGECACHE
);
1642 if (!nfs_have_delegation(inode
, FMODE_READ
))
1643 nfsi
->cache_validity
|= invalid
;
1648 * Big trouble! The inode has become a different object.
1651 printk(KERN_DEBUG
"%s: inode %ld mode changed, %07o to %07o\n",
1652 __FUNCTION__
, inode
->i_ino
, inode
->i_mode
, fattr
->mode
);
1656 * No need to worry about unhashing the dentry, as the
1657 * lookup validation will know that the inode is bad.
1658 * (But we fall through to invalidate the caches.)
1660 nfs_invalidate_inode(inode
);
1664 printk(KERN_ERR
"NFS: server %s error: fileid changed\n"
1665 "fsid %s: expected fileid 0x%Lx, got 0x%Lx\n",
1666 NFS_SERVER(inode
)->hostname
, inode
->i_sb
->s_id
,
1667 (long long)nfsi
->fileid
, (long long)fattr
->fileid
);
1672 * File system information
1675 static int nfs_set_super(struct super_block
*s
, void *data
)
1677 s
->s_fs_info
= data
;
1678 return set_anon_super(s
, data
);
1681 static int nfs_compare_super(struct super_block
*sb
, void *data
)
1683 struct nfs_server
*server
= data
;
1684 struct nfs_server
*old
= NFS_SB(sb
);
1686 if (old
->addr
.sin_addr
.s_addr
!= server
->addr
.sin_addr
.s_addr
)
1688 if (old
->addr
.sin_port
!= server
->addr
.sin_port
)
1690 return !nfs_compare_fh(&old
->fh
, &server
->fh
);
1693 static struct super_block
*nfs_get_sb(struct file_system_type
*fs_type
,
1694 int flags
, const char *dev_name
, void *raw_data
)
1697 struct nfs_server
*server
= NULL
;
1698 struct super_block
*s
;
1699 struct nfs_fh
*root
;
1700 struct nfs_mount_data
*data
= raw_data
;
1702 s
= ERR_PTR(-EINVAL
);
1704 dprintk("%s: missing data argument\n", __FUNCTION__
);
1707 if (data
->version
<= 0 || data
->version
> NFS_MOUNT_VERSION
) {
1708 dprintk("%s: bad mount version\n", __FUNCTION__
);
1711 switch (data
->version
) {
1717 if (data
->flags
& NFS_MOUNT_VER3
) {
1718 dprintk("%s: mount structure version %d does not support NFSv3\n",
1723 data
->root
.size
= NFS2_FHSIZE
;
1724 memcpy(data
->root
.data
, data
->old_root
.data
, NFS2_FHSIZE
);
1726 if (data
->flags
& NFS_MOUNT_SECFLAVOUR
) {
1727 dprintk("%s: mount structure version %d does not support strong security\n",
1733 memset(data
->context
, 0, sizeof(data
->context
));
1735 #ifndef CONFIG_NFS_V3
1736 /* If NFSv3 is not compiled in, return -EPROTONOSUPPORT */
1737 s
= ERR_PTR(-EPROTONOSUPPORT
);
1738 if (data
->flags
& NFS_MOUNT_VER3
) {
1739 dprintk("%s: NFSv3 not compiled into kernel\n", __FUNCTION__
);
1742 #endif /* CONFIG_NFS_V3 */
1744 s
= ERR_PTR(-ENOMEM
);
1745 server
= kzalloc(sizeof(struct nfs_server
), GFP_KERNEL
);
1748 /* Zero out the NFS state stuff */
1749 init_nfsv4_state(server
);
1750 server
->client
= server
->client_sys
= server
->client_acl
= ERR_PTR(-EINVAL
);
1753 if (data
->flags
& NFS_MOUNT_VER3
)
1754 root
->size
= data
->root
.size
;
1756 root
->size
= NFS2_FHSIZE
;
1757 s
= ERR_PTR(-EINVAL
);
1758 if (root
->size
> sizeof(root
->data
)) {
1759 dprintk("%s: invalid root filehandle\n", __FUNCTION__
);
1762 memcpy(root
->data
, data
->root
.data
, root
->size
);
1764 /* We now require that the mount process passes the remote address */
1765 memcpy(&server
->addr
, &data
->addr
, sizeof(server
->addr
));
1766 if (server
->addr
.sin_addr
.s_addr
== INADDR_ANY
) {
1767 dprintk("%s: mount program didn't pass remote address!\n",
1772 /* Fire up rpciod if not yet running */
1773 s
= ERR_PTR(rpciod_up());
1775 dprintk("%s: couldn't start rpciod! Error = %ld\n",
1776 __FUNCTION__
, PTR_ERR(s
));
1780 s
= sget(fs_type
, nfs_compare_super
, nfs_set_super
, server
);
1781 if (IS_ERR(s
) || s
->s_root
)
1782 goto out_rpciod_down
;
1786 error
= nfs_fill_super(s
, data
, flags
& MS_SILENT
? 1 : 0);
1788 up_write(&s
->s_umount
);
1789 deactivate_super(s
);
1790 return ERR_PTR(error
);
1792 s
->s_flags
|= MS_ACTIVE
;
1801 static void nfs_kill_super(struct super_block
*s
)
1803 struct nfs_server
*server
= NFS_SB(s
);
1807 if (!IS_ERR(server
->client
))
1808 rpc_shutdown_client(server
->client
);
1809 if (!IS_ERR(server
->client_sys
))
1810 rpc_shutdown_client(server
->client_sys
);
1811 if (!IS_ERR(server
->client_acl
))
1812 rpc_shutdown_client(server
->client_acl
);
1814 if (!(server
->flags
& NFS_MOUNT_NONLM
))
1815 lockd_down(); /* release rpc.lockd */
1817 rpciod_down(); /* release rpciod */
1819 nfs_free_iostats(server
->io_stats
);
1820 kfree(server
->hostname
);
1824 static struct file_system_type nfs_fs_type
= {
1825 .owner
= THIS_MODULE
,
1827 .get_sb
= nfs_get_sb
,
1828 .kill_sb
= nfs_kill_super
,
1829 .fs_flags
= FS_ODD_RENAME
|FS_REVAL_DOT
|FS_BINARY_MOUNTDATA
,
1832 #ifdef CONFIG_NFS_V4
1834 static void nfs4_clear_inode(struct inode
*);
1837 static struct super_operations nfs4_sops
= {
1838 .alloc_inode
= nfs_alloc_inode
,
1839 .destroy_inode
= nfs_destroy_inode
,
1840 .write_inode
= nfs_write_inode
,
1841 .delete_inode
= nfs_delete_inode
,
1842 .statfs
= nfs_statfs
,
1843 .clear_inode
= nfs4_clear_inode
,
1844 .umount_begin
= nfs_umount_begin
,
1845 .show_options
= nfs_show_options
,
1846 .show_stats
= nfs_show_stats
,
1850 * Clean out any remaining NFSv4 state that might be left over due
1851 * to open() calls that passed nfs_atomic_lookup, but failed to call
1854 static void nfs4_clear_inode(struct inode
*inode
)
1856 struct nfs_inode
*nfsi
= NFS_I(inode
);
1858 /* If we are holding a delegation, return it! */
1859 nfs_inode_return_delegation(inode
);
1860 /* First call standard NFS clear_inode() code */
1861 nfs_clear_inode(inode
);
1862 /* Now clear out any remaining state */
1863 while (!list_empty(&nfsi
->open_states
)) {
1864 struct nfs4_state
*state
;
1866 state
= list_entry(nfsi
->open_states
.next
,
1869 dprintk("%s(%s/%Ld): found unclaimed NFSv4 state %p\n",
1872 (long long)NFS_FILEID(inode
),
1874 BUG_ON(atomic_read(&state
->count
) != 1);
1875 nfs4_close_state(state
, state
->state
);
1880 static int nfs4_fill_super(struct super_block
*sb
, struct nfs4_mount_data
*data
, int silent
)
1882 struct nfs_server
*server
;
1883 struct nfs4_client
*clp
= NULL
;
1884 struct rpc_xprt
*xprt
= NULL
;
1885 struct rpc_clnt
*clnt
= NULL
;
1886 struct rpc_timeout timeparms
;
1887 rpc_authflavor_t authflavour
;
1890 sb
->s_blocksize_bits
= 0;
1891 sb
->s_blocksize
= 0;
1892 server
= NFS_SB(sb
);
1893 if (data
->rsize
!= 0)
1894 server
->rsize
= nfs_block_size(data
->rsize
, NULL
);
1895 if (data
->wsize
!= 0)
1896 server
->wsize
= nfs_block_size(data
->wsize
, NULL
);
1897 server
->flags
= data
->flags
& NFS_MOUNT_FLAGMASK
;
1898 server
->caps
= NFS_CAP_ATOMIC_OPEN
;
1900 server
->acregmin
= data
->acregmin
*HZ
;
1901 server
->acregmax
= data
->acregmax
*HZ
;
1902 server
->acdirmin
= data
->acdirmin
*HZ
;
1903 server
->acdirmax
= data
->acdirmax
*HZ
;
1905 server
->rpc_ops
= &nfs_v4_clientops
;
1907 nfs_init_timeout_values(&timeparms
, data
->proto
, data
->timeo
, data
->retrans
);
1909 server
->retrans_timeo
= timeparms
.to_initval
;
1910 server
->retrans_count
= timeparms
.to_retries
;
1912 clp
= nfs4_get_client(&server
->addr
.sin_addr
);
1914 dprintk("%s: failed to create NFS4 client.\n", __FUNCTION__
);
1918 /* Now create transport and client */
1919 authflavour
= RPC_AUTH_UNIX
;
1920 if (data
->auth_flavourlen
!= 0) {
1921 if (data
->auth_flavourlen
!= 1) {
1922 dprintk("%s: Invalid number of RPC auth flavours %d.\n",
1923 __FUNCTION__
, data
->auth_flavourlen
);
1927 if (copy_from_user(&authflavour
, data
->auth_flavours
, sizeof(authflavour
))) {
1933 down_write(&clp
->cl_sem
);
1934 if (IS_ERR(clp
->cl_rpcclient
)) {
1935 xprt
= xprt_create_proto(data
->proto
, &server
->addr
, &timeparms
);
1937 up_write(&clp
->cl_sem
);
1938 err
= PTR_ERR(xprt
);
1939 dprintk("%s: cannot create RPC transport. Error = %d\n",
1943 clnt
= rpc_create_client(xprt
, server
->hostname
, &nfs_program
,
1944 server
->rpc_ops
->version
, authflavour
);
1946 up_write(&clp
->cl_sem
);
1947 err
= PTR_ERR(clnt
);
1948 dprintk("%s: cannot create RPC client. Error = %d\n",
1953 clnt
->cl_softrtry
= 1;
1954 clp
->cl_rpcclient
= clnt
;
1955 memcpy(clp
->cl_ipaddr
, server
->ip_addr
, sizeof(clp
->cl_ipaddr
));
1958 list_add_tail(&server
->nfs4_siblings
, &clp
->cl_superblocks
);
1959 clnt
= rpc_clone_client(clp
->cl_rpcclient
);
1961 server
->nfs4_state
= clp
;
1962 up_write(&clp
->cl_sem
);
1966 err
= PTR_ERR(clnt
);
1967 dprintk("%s: cannot create RPC client. Error = %d\n",
1972 server
->client
= clnt
;
1974 if (server
->nfs4_state
->cl_idmap
== NULL
) {
1975 dprintk("%s: failed to create idmapper.\n", __FUNCTION__
);
1979 if (clnt
->cl_auth
->au_flavor
!= authflavour
) {
1980 struct rpc_auth
*auth
;
1982 auth
= rpcauth_create(authflavour
, clnt
);
1984 dprintk("%s: couldn't create credcache!\n", __FUNCTION__
);
1985 return PTR_ERR(auth
);
1989 sb
->s_time_gran
= 1;
1991 sb
->s_op
= &nfs4_sops
;
1992 err
= nfs_sb_init(sb
, authflavour
);
1997 nfs4_put_client(clp
);
2001 static int nfs4_compare_super(struct super_block
*sb
, void *data
)
2003 struct nfs_server
*server
= data
;
2004 struct nfs_server
*old
= NFS_SB(sb
);
2006 if (strcmp(server
->hostname
, old
->hostname
) != 0)
2008 if (strcmp(server
->mnt_path
, old
->mnt_path
) != 0)
2014 nfs_copy_user_string(char *dst
, struct nfs_string
*src
, int maxlen
)
2019 return ERR_PTR(-EINVAL
);
2020 if (src
->len
< maxlen
)
2023 p
= dst
= kmalloc(maxlen
+ 1, GFP_KERNEL
);
2025 return ERR_PTR(-ENOMEM
);
2027 if (copy_from_user(dst
, src
->data
, maxlen
)) {
2029 return ERR_PTR(-EFAULT
);
2035 static struct super_block
*nfs4_get_sb(struct file_system_type
*fs_type
,
2036 int flags
, const char *dev_name
, void *raw_data
)
2039 struct nfs_server
*server
;
2040 struct super_block
*s
;
2041 struct nfs4_mount_data
*data
= raw_data
;
2045 dprintk("%s: missing data argument\n", __FUNCTION__
);
2046 return ERR_PTR(-EINVAL
);
2048 if (data
->version
<= 0 || data
->version
> NFS4_MOUNT_VERSION
) {
2049 dprintk("%s: bad mount version\n", __FUNCTION__
);
2050 return ERR_PTR(-EINVAL
);
2053 server
= kzalloc(sizeof(struct nfs_server
), GFP_KERNEL
);
2055 return ERR_PTR(-ENOMEM
);
2056 /* Zero out the NFS state stuff */
2057 init_nfsv4_state(server
);
2058 server
->client
= server
->client_sys
= server
->client_acl
= ERR_PTR(-EINVAL
);
2060 p
= nfs_copy_user_string(NULL
, &data
->hostname
, 256);
2063 server
->hostname
= p
;
2065 p
= nfs_copy_user_string(NULL
, &data
->mnt_path
, 1024);
2068 server
->mnt_path
= p
;
2070 p
= nfs_copy_user_string(server
->ip_addr
, &data
->client_addr
,
2071 sizeof(server
->ip_addr
) - 1);
2075 /* We now require that the mount process passes the remote address */
2076 if (data
->host_addrlen
!= sizeof(server
->addr
)) {
2077 s
= ERR_PTR(-EINVAL
);
2080 if (copy_from_user(&server
->addr
, data
->host_addr
, sizeof(server
->addr
))) {
2081 s
= ERR_PTR(-EFAULT
);
2084 if (server
->addr
.sin_family
!= AF_INET
||
2085 server
->addr
.sin_addr
.s_addr
== INADDR_ANY
) {
2086 dprintk("%s: mount program didn't pass remote IP address!\n",
2088 s
= ERR_PTR(-EINVAL
);
2092 /* Fire up rpciod if not yet running */
2093 s
= ERR_PTR(rpciod_up());
2095 dprintk("%s: couldn't start rpciod! Error = %ld\n",
2096 __FUNCTION__
, PTR_ERR(s
));
2100 s
= sget(fs_type
, nfs4_compare_super
, nfs_set_super
, server
);
2102 if (IS_ERR(s
) || s
->s_root
)
2107 error
= nfs4_fill_super(s
, data
, flags
& MS_SILENT
? 1 : 0);
2109 up_write(&s
->s_umount
);
2110 deactivate_super(s
);
2111 return ERR_PTR(error
);
2113 s
->s_flags
|= MS_ACTIVE
;
2116 s
= (struct super_block
*)p
;
2118 kfree(server
->mnt_path
);
2119 kfree(server
->hostname
);
2124 static void nfs4_kill_super(struct super_block
*sb
)
2126 struct nfs_server
*server
= NFS_SB(sb
);
2128 nfs_return_all_delegations(sb
);
2129 kill_anon_super(sb
);
2131 nfs4_renewd_prepare_shutdown(server
);
2133 if (server
->client
!= NULL
&& !IS_ERR(server
->client
))
2134 rpc_shutdown_client(server
->client
);
2136 destroy_nfsv4_state(server
);
2140 nfs_free_iostats(server
->io_stats
);
2141 kfree(server
->hostname
);
2145 static struct file_system_type nfs4_fs_type
= {
2146 .owner
= THIS_MODULE
,
2148 .get_sb
= nfs4_get_sb
,
2149 .kill_sb
= nfs4_kill_super
,
2150 .fs_flags
= FS_ODD_RENAME
|FS_REVAL_DOT
|FS_BINARY_MOUNTDATA
,
2153 static const int nfs_set_port_min
= 0;
2154 static const int nfs_set_port_max
= 65535;
2155 static int param_set_port(const char *val
, struct kernel_param
*kp
)
2158 int num
= simple_strtol(val
, &endp
, 0);
2159 if (endp
== val
|| *endp
|| num
< nfs_set_port_min
|| num
> nfs_set_port_max
)
2161 *((int *)kp
->arg
) = num
;
2165 module_param_call(callback_tcpport
, param_set_port
, param_get_int
,
2166 &nfs_callback_set_tcpport
, 0644);
2168 static int param_set_idmap_timeout(const char *val
, struct kernel_param
*kp
)
2171 int num
= simple_strtol(val
, &endp
, 0);
2173 if (endp
== val
|| *endp
|| num
< 0 || jif
< num
)
2175 *((int *)kp
->arg
) = jif
;
2179 module_param_call(idmap_cache_timeout
, param_set_idmap_timeout
, param_get_int
,
2180 &nfs_idmap_cache_timeout
, 0644);
2182 #define nfs4_init_once(nfsi) \
2184 INIT_LIST_HEAD(&(nfsi)->open_states); \
2185 nfsi->delegation = NULL; \
2186 nfsi->delegation_state = 0; \
2187 init_rwsem(&nfsi->rwsem); \
2190 static inline int register_nfs4fs(void)
2194 ret
= nfs_register_sysctl();
2197 ret
= register_filesystem(&nfs4_fs_type
);
2199 nfs_unregister_sysctl();
2203 static inline void unregister_nfs4fs(void)
2205 unregister_filesystem(&nfs4_fs_type
);
2206 nfs_unregister_sysctl();
2209 #define nfs4_init_once(nfsi) \
2211 #define register_nfs4fs() (0)
2212 #define unregister_nfs4fs()
2215 extern int nfs_init_nfspagecache(void);
2216 extern void nfs_destroy_nfspagecache(void);
2217 extern int nfs_init_readpagecache(void);
2218 extern void nfs_destroy_readpagecache(void);
2219 extern int nfs_init_writepagecache(void);
2220 extern void nfs_destroy_writepagecache(void);
2221 #ifdef CONFIG_NFS_DIRECTIO
2222 extern int nfs_init_directcache(void);
2223 extern void nfs_destroy_directcache(void);
2226 static kmem_cache_t
* nfs_inode_cachep
;
2228 static struct inode
*nfs_alloc_inode(struct super_block
*sb
)
2230 struct nfs_inode
*nfsi
;
2231 nfsi
= (struct nfs_inode
*)kmem_cache_alloc(nfs_inode_cachep
, SLAB_KERNEL
);
2235 nfsi
->cache_validity
= 0UL;
2236 nfsi
->cache_change_attribute
= jiffies
;
2237 #ifdef CONFIG_NFS_V3_ACL
2238 nfsi
->acl_access
= ERR_PTR(-EAGAIN
);
2239 nfsi
->acl_default
= ERR_PTR(-EAGAIN
);
2241 #ifdef CONFIG_NFS_V4
2242 nfsi
->nfs4_acl
= NULL
;
2243 #endif /* CONFIG_NFS_V4 */
2244 return &nfsi
->vfs_inode
;
2247 static void nfs_destroy_inode(struct inode
*inode
)
2249 kmem_cache_free(nfs_inode_cachep
, NFS_I(inode
));
2252 static void init_once(void * foo
, kmem_cache_t
* cachep
, unsigned long flags
)
2254 struct nfs_inode
*nfsi
= (struct nfs_inode
*) foo
;
2256 if ((flags
& (SLAB_CTOR_VERIFY
|SLAB_CTOR_CONSTRUCTOR
)) ==
2257 SLAB_CTOR_CONSTRUCTOR
) {
2258 inode_init_once(&nfsi
->vfs_inode
);
2259 spin_lock_init(&nfsi
->req_lock
);
2260 INIT_LIST_HEAD(&nfsi
->dirty
);
2261 INIT_LIST_HEAD(&nfsi
->commit
);
2262 INIT_LIST_HEAD(&nfsi
->open_files
);
2263 INIT_RADIX_TREE(&nfsi
->nfs_page_tree
, GFP_ATOMIC
);
2264 atomic_set(&nfsi
->data_updates
, 0);
2268 nfs4_init_once(nfsi
);
2272 static int nfs_init_inodecache(void)
2274 nfs_inode_cachep
= kmem_cache_create("nfs_inode_cache",
2275 sizeof(struct nfs_inode
),
2276 0, (SLAB_RECLAIM_ACCOUNT
|
2279 if (nfs_inode_cachep
== NULL
)
2285 static void nfs_destroy_inodecache(void)
2287 if (kmem_cache_destroy(nfs_inode_cachep
))
2288 printk(KERN_INFO
"nfs_inode_cache: not all structures were freed\n");
2294 static int __init
init_nfs_fs(void)
2298 err
= nfs_init_nfspagecache();
2302 err
= nfs_init_inodecache();
2306 err
= nfs_init_readpagecache();
2310 err
= nfs_init_writepagecache();
2314 #ifdef CONFIG_NFS_DIRECTIO
2315 err
= nfs_init_directcache();
2320 #ifdef CONFIG_PROC_FS
2321 rpc_proc_register(&nfs_rpcstat
);
2323 err
= register_filesystem(&nfs_fs_type
);
2326 if ((err
= register_nfs4fs()) != 0)
2330 #ifdef CONFIG_PROC_FS
2331 rpc_proc_unregister("nfs");
2333 #ifdef CONFIG_NFS_DIRECTIO
2334 nfs_destroy_directcache();
2337 nfs_destroy_writepagecache();
2339 nfs_destroy_readpagecache();
2341 nfs_destroy_inodecache();
2343 nfs_destroy_nfspagecache();
2348 static void __exit
exit_nfs_fs(void)
2350 #ifdef CONFIG_NFS_DIRECTIO
2351 nfs_destroy_directcache();
2353 nfs_destroy_writepagecache();
2354 nfs_destroy_readpagecache();
2355 nfs_destroy_inodecache();
2356 nfs_destroy_nfspagecache();
2357 #ifdef CONFIG_PROC_FS
2358 rpc_proc_unregister("nfs");
2360 unregister_filesystem(&nfs_fs_type
);
2361 unregister_nfs4fs();
2364 /* Not quite true; I just maintain it */
2365 MODULE_AUTHOR("Olaf Kirch <okir@monad.swb.de>");
2366 MODULE_LICENSE("GPL");
2368 module_init(init_nfs_fs
)
2369 module_exit(exit_nfs_fs
)