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>
39 #include <linux/vserver/xid.h>
41 #include <asm/system.h>
42 #include <asm/uaccess.h>
46 #include "delegation.h"
49 #define NFSDBG_FACILITY NFSDBG_VFS
50 #define NFS_PARANOIA 1
52 /* Maximum number of readahead requests
53 * FIXME: this should really be a sysctl so that users may tune it to suit
54 * their needs. People that do NFS over a slow network, might for
55 * instance want to reduce it to something closer to 1 for improved
56 * interactive response.
58 #define NFS_MAX_READAHEAD (RPC_DEF_SLOT_TABLE - 1)
60 static void nfs_invalidate_inode(struct inode *);
61 static int nfs_update_inode(struct inode *, struct nfs_fattr *);
63 static struct inode *nfs_alloc_inode(struct super_block *sb);
64 static void nfs_destroy_inode(struct inode *);
65 static int nfs_write_inode(struct inode *,int);
66 static void nfs_delete_inode(struct inode *);
67 static void nfs_clear_inode(struct inode *);
68 static void nfs_umount_begin(struct super_block *);
69 static int nfs_statfs(struct super_block *, struct kstatfs *);
70 static int nfs_show_options(struct seq_file *, struct vfsmount *);
71 static int nfs_show_stats(struct seq_file *, struct vfsmount *);
72 static void nfs_zap_acl_cache(struct inode *);
74 static struct rpc_program nfs_program;
76 static struct super_operations nfs_sops = {
77 .alloc_inode = nfs_alloc_inode,
78 .destroy_inode = nfs_destroy_inode,
79 .write_inode = nfs_write_inode,
80 .delete_inode = nfs_delete_inode,
82 .clear_inode = nfs_clear_inode,
83 .umount_begin = nfs_umount_begin,
84 .show_options = nfs_show_options,
85 .show_stats = nfs_show_stats,
91 static struct rpc_stat nfs_rpcstat = {
92 .program = &nfs_program
94 static struct rpc_version * nfs_version[] = {
98 #if defined(CONFIG_NFS_V3)
100 #elif defined(CONFIG_NFS_V4)
103 #if defined(CONFIG_NFS_V4)
108 static struct rpc_program nfs_program = {
110 .number = NFS_PROGRAM,
111 .nrvers = ARRAY_SIZE(nfs_version),
112 .version = nfs_version,
113 .stats = &nfs_rpcstat,
114 .pipe_dir_name = "/nfs",
117 #ifdef CONFIG_NFS_V3_ACL
118 static struct rpc_stat nfsacl_rpcstat = { &nfsacl_program };
119 static struct rpc_version * nfsacl_version[] = {
120 [3] = &nfsacl_version3,
123 struct rpc_program nfsacl_program = {
125 .number = NFS_ACL_PROGRAM,
126 .nrvers = ARRAY_SIZE(nfsacl_version),
127 .version = nfsacl_version,
128 .stats = &nfsacl_rpcstat,
130 #endif /* CONFIG_NFS_V3_ACL */
132 static inline unsigned long
133 nfs_fattr_to_ino_t(struct nfs_fattr *fattr)
135 return nfs_fileid_to_ino_t(fattr->fileid);
139 nfs_write_inode(struct inode *inode, int sync)
141 int flags = sync ? FLUSH_SYNC : 0;
144 ret = nfs_commit_inode(inode, flags);
151 nfs_delete_inode(struct inode * inode)
153 dprintk("NFS: delete_inode(%s/%ld)\n", inode->i_sb->s_id, inode->i_ino);
155 truncate_inode_pages(&inode->i_data, 0);
159 * The following should never happen...
161 if (nfs_have_writebacks(inode)) {
162 printk(KERN_ERR "nfs_delete_inode: inode %ld has pending RPC requests\n", inode->i_ino);
169 nfs_clear_inode(struct inode *inode)
171 struct nfs_inode *nfsi = NFS_I(inode);
172 struct rpc_cred *cred;
175 BUG_ON (!list_empty(&nfsi->open_files));
176 nfs_zap_acl_cache(inode);
177 cred = nfsi->cache_access.cred;
180 BUG_ON(atomic_read(&nfsi->data_updates) != 0);
184 nfs_umount_begin(struct super_block *sb)
186 struct rpc_clnt *rpc = NFS_SB(sb)->client;
188 /* -EIO all pending I/O */
190 rpc_killall_tasks(rpc);
191 rpc = NFS_SB(sb)->client_acl;
193 rpc_killall_tasks(rpc);
197 static inline unsigned long
198 nfs_block_bits(unsigned long bsize, unsigned char *nrbitsp)
200 /* make sure blocksize is a power of two */
201 if ((bsize & (bsize - 1)) || nrbitsp) {
202 unsigned char nrbits;
204 for (nrbits = 31; nrbits && !(bsize & (1 << nrbits)); nrbits--)
215 * Calculate the number of 512byte blocks used.
217 static inline unsigned long
218 nfs_calc_block_size(u64 tsize)
220 loff_t used = (tsize + 511) >> 9;
221 return (used > ULONG_MAX) ? ULONG_MAX : used;
225 * Compute and set NFS server blocksize
227 static inline unsigned long
228 nfs_block_size(unsigned long bsize, unsigned char *nrbitsp)
230 if (bsize < NFS_MIN_FILE_IO_SIZE)
231 bsize = NFS_DEF_FILE_IO_SIZE;
232 else if (bsize >= NFS_MAX_FILE_IO_SIZE)
233 bsize = NFS_MAX_FILE_IO_SIZE;
235 return nfs_block_bits(bsize, nrbitsp);
239 * Obtain the root inode of the file system.
241 static struct inode *
242 nfs_get_root(struct super_block *sb, struct nfs_fh *rootfh, struct nfs_fsinfo *fsinfo)
244 struct nfs_server *server = NFS_SB(sb);
247 error = server->rpc_ops->getroot(server, rootfh, fsinfo);
249 dprintk("nfs_get_root: getattr error = %d\n", -error);
250 return ERR_PTR(error);
253 return nfs_fhget(sb, rootfh, fsinfo->fattr);
257 * Do NFS version-independent mount processing, and sanity checking
260 nfs_sb_init(struct super_block *sb, rpc_authflavor_t authflavor)
262 struct nfs_server *server;
263 struct inode *root_inode;
264 struct nfs_fattr fattr;
265 struct nfs_fsinfo fsinfo = {
268 struct nfs_pathconf pathinfo = {
271 int no_root_error = 0;
272 unsigned long max_rpc_payload;
274 /* We probably want something more informative here */
275 snprintf(sb->s_id, sizeof(sb->s_id), "%x:%x", MAJOR(sb->s_dev), MINOR(sb->s_dev));
279 sb->s_magic = NFS_SUPER_MAGIC;
281 server->io_stats = nfs_alloc_iostats();
282 if (server->io_stats == NULL)
285 root_inode = nfs_get_root(sb, &server->fh, &fsinfo);
286 /* Did getting the root inode fail? */
287 if (IS_ERR(root_inode)) {
288 no_root_error = PTR_ERR(root_inode);
291 sb->s_root = d_alloc_root(root_inode);
293 no_root_error = -ENOMEM;
296 sb->s_root->d_op = server->rpc_ops->dentry_ops;
298 /* mount time stamp, in seconds */
299 server->mount_time = jiffies;
301 /* Get some general file system info */
302 if (server->namelen == 0 &&
303 server->rpc_ops->pathconf(server, &server->fh, &pathinfo) >= 0)
304 server->namelen = pathinfo.max_namelen;
305 /* Work out a lot of parameters */
306 if (server->rsize == 0)
307 server->rsize = nfs_block_size(fsinfo.rtpref, NULL);
308 if (server->wsize == 0)
309 server->wsize = nfs_block_size(fsinfo.wtpref, NULL);
311 if (fsinfo.rtmax >= 512 && server->rsize > fsinfo.rtmax)
312 server->rsize = nfs_block_size(fsinfo.rtmax, NULL);
313 if (fsinfo.wtmax >= 512 && server->wsize > fsinfo.wtmax)
314 server->wsize = nfs_block_size(fsinfo.wtmax, NULL);
316 max_rpc_payload = nfs_block_size(rpc_max_payload(server->client), NULL);
317 if (server->rsize > max_rpc_payload)
318 server->rsize = max_rpc_payload;
319 if (server->rsize > NFS_MAX_FILE_IO_SIZE)
320 server->rsize = NFS_MAX_FILE_IO_SIZE;
321 server->rpages = (server->rsize + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT;
323 if (server->wsize > max_rpc_payload)
324 server->wsize = max_rpc_payload;
325 if (server->wsize > NFS_MAX_FILE_IO_SIZE)
326 server->wsize = NFS_MAX_FILE_IO_SIZE;
327 server->wpages = (server->wsize + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT;
329 if (sb->s_blocksize == 0)
330 sb->s_blocksize = nfs_block_bits(server->wsize,
331 &sb->s_blocksize_bits);
332 server->wtmult = nfs_block_bits(fsinfo.wtmult, NULL);
334 server->dtsize = nfs_block_size(fsinfo.dtpref, NULL);
335 if (server->dtsize > PAGE_CACHE_SIZE)
336 server->dtsize = PAGE_CACHE_SIZE;
337 if (server->dtsize > server->rsize)
338 server->dtsize = server->rsize;
340 if (server->flags & NFS_MOUNT_NOAC) {
341 server->acregmin = server->acregmax = 0;
342 server->acdirmin = server->acdirmax = 0;
343 sb->s_flags |= MS_SYNCHRONOUS;
345 server->backing_dev_info.ra_pages = server->rpages * NFS_MAX_READAHEAD;
347 if (server->flags & NFS_MOUNT_TAGXID)
348 sb->s_flags |= MS_TAGXID;
350 sb->s_maxbytes = fsinfo.maxfilesize;
351 if (sb->s_maxbytes > MAX_LFS_FILESIZE)
352 sb->s_maxbytes = MAX_LFS_FILESIZE;
354 server->client->cl_intr = (server->flags & NFS_MOUNT_INTR) ? 1 : 0;
355 server->client->cl_softrtry = (server->flags & NFS_MOUNT_SOFT) ? 1 : 0;
356 server->client->cl_tagxid = (server->flags & NFS_MOUNT_TAGXID) ? 1 : 0;
358 /* We're airborne Set socket buffersize */
359 rpc_setbufsize(server->client, server->wsize + 100, server->rsize + 100);
361 /* Yargs. It didn't work out. */
363 dprintk("nfs_sb_init: get root inode failed: errno %d\n", -no_root_error);
364 if (!IS_ERR(root_inode))
366 return no_root_error;
369 static void nfs_init_timeout_values(struct rpc_timeout *to, int proto, unsigned int timeo, unsigned int retrans)
371 to->to_initval = timeo * HZ / 10;
372 to->to_retries = retrans;
379 to->to_initval = 60 * HZ;
380 if (to->to_initval > NFS_MAX_TCP_TIMEOUT)
381 to->to_initval = NFS_MAX_TCP_TIMEOUT;
382 to->to_increment = to->to_initval;
383 to->to_maxval = to->to_initval + (to->to_increment * to->to_retries);
384 to->to_exponential = 0;
389 to->to_initval = 11 * HZ / 10;
390 if (to->to_initval > NFS_MAX_UDP_TIMEOUT)
391 to->to_initval = NFS_MAX_UDP_TIMEOUT;
392 to->to_maxval = NFS_MAX_UDP_TIMEOUT;
393 to->to_exponential = 1;
399 * Create an RPC client handle.
401 static struct rpc_clnt *
402 nfs_create_client(struct nfs_server *server, const struct nfs_mount_data *data)
404 struct rpc_timeout timeparms;
405 struct rpc_xprt *xprt = NULL;
406 struct rpc_clnt *clnt = NULL;
407 int proto = (data->flags & NFS_MOUNT_TCP) ? IPPROTO_TCP : IPPROTO_UDP;
409 nfs_init_timeout_values(&timeparms, proto, data->timeo, data->retrans);
411 server->retrans_timeo = timeparms.to_initval;
412 server->retrans_count = timeparms.to_retries;
414 /* create transport and client */
415 xprt = xprt_create_proto(proto, &server->addr, &timeparms);
417 dprintk("%s: cannot create RPC transport. Error = %ld\n",
418 __FUNCTION__, PTR_ERR(xprt));
419 return (struct rpc_clnt *)xprt;
421 clnt = rpc_create_client(xprt, server->hostname, &nfs_program,
422 server->rpc_ops->version, data->pseudoflavor);
424 dprintk("%s: cannot create RPC client. Error = %ld\n",
425 __FUNCTION__, PTR_ERR(xprt));
430 clnt->cl_softrtry = 1;
440 * The way this works is that the mount process passes a structure
441 * in the data argument which contains the server's IP address
442 * and the root file handle obtained from the server's mount
443 * daemon. We stash these away in the private superblock fields.
446 nfs_fill_super(struct super_block *sb, struct nfs_mount_data *data, int silent)
448 struct nfs_server *server;
449 rpc_authflavor_t authflavor;
452 sb->s_blocksize_bits = 0;
455 sb->s_blocksize = nfs_block_size(data->bsize, &sb->s_blocksize_bits);
457 server->rsize = nfs_block_size(data->rsize, NULL);
459 server->wsize = nfs_block_size(data->wsize, NULL);
460 server->flags = data->flags & NFS_MOUNT_FLAGMASK;
462 server->acregmin = data->acregmin*HZ;
463 server->acregmax = data->acregmax*HZ;
464 server->acdirmin = data->acdirmin*HZ;
465 server->acdirmax = data->acdirmax*HZ;
467 /* Start lockd here, before we might error out */
468 if (!(server->flags & NFS_MOUNT_NONLM))
471 server->namelen = data->namlen;
472 server->hostname = kmalloc(strlen(data->hostname) + 1, GFP_KERNEL);
473 if (!server->hostname)
475 strcpy(server->hostname, data->hostname);
477 /* Check NFS protocol revision and initialize RPC op vector
478 * and file handle pool. */
480 if (server->flags & NFS_MOUNT_VER3) {
481 server->rpc_ops = &nfs_v3_clientops;
482 server->caps |= NFS_CAP_READDIRPLUS;
484 server->rpc_ops = &nfs_v2_clientops;
487 server->rpc_ops = &nfs_v2_clientops;
490 /* Fill in pseudoflavor for mount version < 5 */
491 if (!(data->flags & NFS_MOUNT_SECFLAVOUR))
492 data->pseudoflavor = RPC_AUTH_UNIX;
493 authflavor = data->pseudoflavor; /* save for sb_init() */
494 /* XXX maybe we want to add a server->pseudoflavor field */
496 /* Create RPC client handles */
497 server->client = nfs_create_client(server, data);
498 if (IS_ERR(server->client))
499 return PTR_ERR(server->client);
500 /* RFC 2623, sec 2.3.2 */
501 if (authflavor != RPC_AUTH_UNIX) {
502 struct rpc_auth *auth;
504 server->client_sys = rpc_clone_client(server->client);
505 if (IS_ERR(server->client_sys))
506 return PTR_ERR(server->client_sys);
507 auth = rpcauth_create(RPC_AUTH_UNIX, server->client_sys);
509 return PTR_ERR(auth);
511 atomic_inc(&server->client->cl_count);
512 server->client_sys = server->client;
514 if (server->flags & NFS_MOUNT_VER3) {
515 #ifdef CONFIG_NFS_V3_ACL
516 if (!(server->flags & NFS_MOUNT_NOACL)) {
517 server->client_acl = rpc_bind_new_program(server->client, &nfsacl_program, 3);
518 /* No errors! Assume that Sun nfsacls are supported */
519 if (!IS_ERR(server->client_acl))
520 server->caps |= NFS_CAP_ACLS;
523 server->flags &= ~NFS_MOUNT_NOACL;
524 #endif /* CONFIG_NFS_V3_ACL */
526 * The VFS shouldn't apply the umask to mode bits. We will
527 * do so ourselves when necessary.
529 sb->s_flags |= MS_POSIXACL;
530 if (server->namelen == 0 || server->namelen > NFS3_MAXNAMLEN)
531 server->namelen = NFS3_MAXNAMLEN;
534 if (server->namelen == 0 || server->namelen > NFS2_MAXNAMLEN)
535 server->namelen = NFS2_MAXNAMLEN;
538 sb->s_op = &nfs_sops;
539 return nfs_sb_init(sb, authflavor);
543 nfs_statfs(struct super_block *sb, struct kstatfs *buf)
545 struct nfs_server *server = NFS_SB(sb);
546 unsigned char blockbits;
547 unsigned long blockres;
548 struct nfs_fh *rootfh = NFS_FH(sb->s_root->d_inode);
549 struct nfs_fattr fattr;
550 struct nfs_fsstat res = {
557 error = server->rpc_ops->statfs(server, rootfh, &res);
558 buf->f_type = NFS_SUPER_MAGIC;
563 * Current versions of glibc do not correctly handle the
564 * case where f_frsize != f_bsize. Eventually we want to
565 * report the value of wtmult in this field.
567 buf->f_frsize = sb->s_blocksize;
570 * On most *nix systems, f_blocks, f_bfree, and f_bavail
571 * are reported in units of f_frsize. Linux hasn't had
572 * an f_frsize field in its statfs struct until recently,
573 * thus historically Linux's sys_statfs reports these
574 * fields in units of f_bsize.
576 buf->f_bsize = sb->s_blocksize;
577 blockbits = sb->s_blocksize_bits;
578 blockres = (1 << blockbits) - 1;
579 buf->f_blocks = (res.tbytes + blockres) >> blockbits;
580 buf->f_bfree = (res.fbytes + blockres) >> blockbits;
581 buf->f_bavail = (res.abytes + blockres) >> blockbits;
583 buf->f_files = res.tfiles;
584 buf->f_ffree = res.afiles;
586 buf->f_namelen = server->namelen;
592 dprintk("%s: statfs error = %d\n", __FUNCTION__, -error);
593 buf->f_bsize = buf->f_blocks = buf->f_bfree = buf->f_bavail = -1;
598 static void nfs_show_mount_options(struct seq_file *m, struct nfs_server *nfss, int showdefaults)
600 static struct proc_nfs_info {
605 { NFS_MOUNT_SOFT, ",soft", ",hard" },
606 { NFS_MOUNT_INTR, ",intr", "" },
607 { NFS_MOUNT_NOCTO, ",nocto", "" },
608 { NFS_MOUNT_NOAC, ",noac", "" },
609 { NFS_MOUNT_NONLM, ",nolock", "" },
610 { NFS_MOUNT_NOACL, ",noacl", "" },
611 { NFS_MOUNT_TAGXID, ",tagxid", "" },
614 struct proc_nfs_info *nfs_infop;
618 seq_printf(m, ",vers=%d", nfss->rpc_ops->version);
619 seq_printf(m, ",rsize=%d", nfss->rsize);
620 seq_printf(m, ",wsize=%d", nfss->wsize);
621 if (nfss->acregmin != 3*HZ || showdefaults)
622 seq_printf(m, ",acregmin=%d", nfss->acregmin/HZ);
623 if (nfss->acregmax != 60*HZ || showdefaults)
624 seq_printf(m, ",acregmax=%d", nfss->acregmax/HZ);
625 if (nfss->acdirmin != 30*HZ || showdefaults)
626 seq_printf(m, ",acdirmin=%d", nfss->acdirmin/HZ);
627 if (nfss->acdirmax != 60*HZ || showdefaults)
628 seq_printf(m, ",acdirmax=%d", nfss->acdirmax/HZ);
629 for (nfs_infop = nfs_info; nfs_infop->flag; nfs_infop++) {
630 if (nfss->flags & nfs_infop->flag)
631 seq_puts(m, nfs_infop->str);
633 seq_puts(m, nfs_infop->nostr);
635 switch (nfss->client->cl_xprt->prot) {
643 snprintf(buf, sizeof(buf), "%u", nfss->client->cl_xprt->prot);
646 seq_printf(m, ",proto=%s", proto);
647 seq_printf(m, ",timeo=%lu", 10U * nfss->retrans_timeo / HZ);
648 seq_printf(m, ",retrans=%u", nfss->retrans_count);
651 static int nfs_show_options(struct seq_file *m, struct vfsmount *mnt)
653 struct nfs_server *nfss = NFS_SB(mnt->mnt_sb);
655 nfs_show_mount_options(m, nfss, 0);
657 seq_puts(m, ",addr=");
658 seq_escape(m, nfss->hostname, " \t\n\\");
663 static int nfs_show_stats(struct seq_file *m, struct vfsmount *mnt)
666 struct nfs_server *nfss = NFS_SB(mnt->mnt_sb);
667 struct rpc_auth *auth = nfss->client->cl_auth;
668 struct nfs_iostats totals = { };
670 seq_printf(m, "statvers=%s", NFS_IOSTAT_VERS);
673 * Display all mount option settings
675 seq_printf(m, "\n\topts:\t");
676 seq_puts(m, mnt->mnt_sb->s_flags & MS_RDONLY ? "ro" : "rw");
677 seq_puts(m, mnt->mnt_sb->s_flags & MS_SYNCHRONOUS ? ",sync" : "");
678 seq_puts(m, mnt->mnt_sb->s_flags & MS_NOATIME ? ",noatime" : "");
679 seq_puts(m, mnt->mnt_sb->s_flags & MS_NODIRATIME ? ",nodiratime" : "");
680 nfs_show_mount_options(m, nfss, 1);
682 seq_printf(m, "\n\tage:\t%lu", (jiffies - nfss->mount_time) / HZ);
684 seq_printf(m, "\n\tcaps:\t");
685 seq_printf(m, "caps=0x%x", nfss->caps);
686 seq_printf(m, ",wtmult=%d", nfss->wtmult);
687 seq_printf(m, ",dtsize=%d", nfss->dtsize);
688 seq_printf(m, ",bsize=%d", nfss->bsize);
689 seq_printf(m, ",namelen=%d", nfss->namelen);
692 if (nfss->rpc_ops->version == 4) {
693 seq_printf(m, "\n\tnfsv4:\t");
694 seq_printf(m, "bm0=0x%x", nfss->attr_bitmask[0]);
695 seq_printf(m, ",bm1=0x%x", nfss->attr_bitmask[1]);
696 seq_printf(m, ",acl=0x%x", nfss->acl_bitmask);
701 * Display security flavor in effect for this mount
703 seq_printf(m, "\n\tsec:\tflavor=%d", auth->au_ops->au_flavor);
705 seq_printf(m, ",pseudoflavor=%d", auth->au_flavor);
708 * Display superblock I/O counters
710 for_each_possible_cpu(cpu) {
711 struct nfs_iostats *stats;
714 stats = per_cpu_ptr(nfss->io_stats, cpu);
716 for (i = 0; i < __NFSIOS_COUNTSMAX; i++)
717 totals.events[i] += stats->events[i];
718 for (i = 0; i < __NFSIOS_BYTESMAX; i++)
719 totals.bytes[i] += stats->bytes[i];
724 seq_printf(m, "\n\tevents:\t");
725 for (i = 0; i < __NFSIOS_COUNTSMAX; i++)
726 seq_printf(m, "%lu ", totals.events[i]);
727 seq_printf(m, "\n\tbytes:\t");
728 for (i = 0; i < __NFSIOS_BYTESMAX; i++)
729 seq_printf(m, "%Lu ", totals.bytes[i]);
732 rpc_print_iostats(m, nfss->client);
738 * nfs_sync_mapping - helper to flush all mmapped dirty data to disk
740 int nfs_sync_mapping(struct address_space *mapping)
744 if (mapping->nrpages == 0)
746 unmap_mapping_range(mapping, 0, 0, 0);
747 ret = filemap_write_and_wait(mapping);
750 ret = nfs_wb_all(mapping->host);
756 * Invalidate the local caches
758 static void nfs_zap_caches_locked(struct inode *inode)
760 struct nfs_inode *nfsi = NFS_I(inode);
761 int mode = inode->i_mode;
763 nfs_inc_stats(inode, NFSIOS_ATTRINVALIDATE);
765 NFS_ATTRTIMEO(inode) = NFS_MINATTRTIMEO(inode);
766 NFS_ATTRTIMEO_UPDATE(inode) = jiffies;
768 memset(NFS_COOKIEVERF(inode), 0, sizeof(NFS_COOKIEVERF(inode)));
769 if (S_ISREG(mode) || S_ISDIR(mode) || S_ISLNK(mode))
770 nfsi->cache_validity |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_DATA|NFS_INO_INVALID_ACCESS|NFS_INO_INVALID_ACL|NFS_INO_REVAL_PAGECACHE;
772 nfsi->cache_validity |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_ACCESS|NFS_INO_INVALID_ACL|NFS_INO_REVAL_PAGECACHE;
775 void nfs_zap_caches(struct inode *inode)
777 spin_lock(&inode->i_lock);
778 nfs_zap_caches_locked(inode);
779 spin_unlock(&inode->i_lock);
782 static void nfs_zap_acl_cache(struct inode *inode)
784 void (*clear_acl_cache)(struct inode *);
786 clear_acl_cache = NFS_PROTO(inode)->clear_acl_cache;
787 if (clear_acl_cache != NULL)
788 clear_acl_cache(inode);
789 spin_lock(&inode->i_lock);
790 NFS_I(inode)->cache_validity &= ~NFS_INO_INVALID_ACL;
791 spin_unlock(&inode->i_lock);
795 * Invalidate, but do not unhash, the inode.
796 * NB: must be called with inode->i_lock held!
798 static void nfs_invalidate_inode(struct inode *inode)
800 set_bit(NFS_INO_STALE, &NFS_FLAGS(inode));
801 nfs_zap_caches_locked(inode);
804 struct nfs_find_desc {
806 struct nfs_fattr *fattr;
810 * In NFSv3 we can have 64bit inode numbers. In order to support
811 * this, and re-exported directories (also seen in NFSv2)
812 * we are forced to allow 2 different inodes to have the same
816 nfs_find_actor(struct inode *inode, void *opaque)
818 struct nfs_find_desc *desc = (struct nfs_find_desc *)opaque;
819 struct nfs_fh *fh = desc->fh;
820 struct nfs_fattr *fattr = desc->fattr;
822 if (NFS_FILEID(inode) != fattr->fileid)
824 if (nfs_compare_fh(NFS_FH(inode), fh))
826 if (is_bad_inode(inode) || NFS_STALE(inode))
832 nfs_init_locked(struct inode *inode, void *opaque)
834 struct nfs_find_desc *desc = (struct nfs_find_desc *)opaque;
835 struct nfs_fattr *fattr = desc->fattr;
837 NFS_FILEID(inode) = fattr->fileid;
838 nfs_copy_fh(NFS_FH(inode), desc->fh);
842 /* Don't use READDIRPLUS on directories that we believe are too large */
843 #define NFS_LIMIT_READDIRPLUS (8*PAGE_SIZE)
846 * This is our front-end to iget that looks up inodes by file handle
847 * instead of inode number.
850 nfs_fhget(struct super_block *sb, struct nfs_fh *fh, struct nfs_fattr *fattr)
852 struct nfs_find_desc desc = {
856 struct inode *inode = ERR_PTR(-ENOENT);
859 if ((fattr->valid & NFS_ATTR_FATTR) == 0)
863 printk("NFS: Buggy server - nlink == 0!\n");
867 hash = nfs_fattr_to_ino_t(fattr);
869 inode = iget5_locked(sb, hash, nfs_find_actor, nfs_init_locked, &desc);
871 inode = ERR_PTR(-ENOMEM);
875 if (inode->i_state & I_NEW) {
876 struct nfs_inode *nfsi = NFS_I(inode);
878 /* We set i_ino for the few things that still rely on it,
882 /* We can't support update_atime(), since the server will reset it */
883 inode->i_flags |= S_NOATIME|S_NOCMTIME;
884 inode->i_mode = fattr->mode;
885 /* Why so? Because we want revalidate for devices/FIFOs, and
886 * that's precisely what we have in nfs_file_inode_operations.
888 inode->i_op = NFS_SB(sb)->rpc_ops->file_inode_ops;
889 if (S_ISREG(inode->i_mode)) {
890 inode->i_fop = &nfs_file_operations;
891 inode->i_data.a_ops = &nfs_file_aops;
892 inode->i_data.backing_dev_info = &NFS_SB(sb)->backing_dev_info;
893 } else if (S_ISDIR(inode->i_mode)) {
894 inode->i_op = NFS_SB(sb)->rpc_ops->dir_inode_ops;
895 inode->i_fop = &nfs_dir_operations;
896 if (nfs_server_capable(inode, NFS_CAP_READDIRPLUS)
897 && fattr->size <= NFS_LIMIT_READDIRPLUS)
898 set_bit(NFS_INO_ADVISE_RDPLUS, &NFS_FLAGS(inode));
899 } else if (S_ISLNK(inode->i_mode))
900 inode->i_op = &nfs_symlink_inode_operations;
902 init_special_inode(inode, inode->i_mode, fattr->rdev);
904 nfsi->read_cache_jiffies = fattr->time_start;
905 nfsi->last_updated = jiffies;
906 inode->i_atime = fattr->atime;
907 inode->i_mtime = fattr->mtime;
908 inode->i_ctime = fattr->ctime;
909 if (fattr->valid & NFS_ATTR_FATTR_V4)
910 nfsi->change_attr = fattr->change_attr;
911 inode->i_size = nfs_size_to_loff_t(fattr->size);
912 inode->i_nlink = fattr->nlink;
913 inode->i_uid = INOXID_UID(XID_TAG(inode), fattr->uid, fattr->gid);
914 inode->i_gid = INOXID_GID(XID_TAG(inode), fattr->uid, fattr->gid);
915 inode->i_xid = INOXID_XID(XID_TAG(inode), fattr->uid, fattr->gid, 0);
916 /* maybe fattr->xid someday */
917 if (fattr->valid & (NFS_ATTR_FATTR_V3 | NFS_ATTR_FATTR_V4)) {
919 * report the blocks in 512byte units
921 inode->i_blocks = nfs_calc_block_size(fattr->du.nfs3.used);
922 inode->i_blksize = inode->i_sb->s_blocksize;
924 inode->i_blocks = fattr->du.nfs2.blocks;
925 inode->i_blksize = fattr->du.nfs2.blocksize;
927 nfsi->attrtimeo = NFS_MINATTRTIMEO(inode);
928 nfsi->attrtimeo_timestamp = jiffies;
929 memset(nfsi->cookieverf, 0, sizeof(nfsi->cookieverf));
930 nfsi->cache_access.cred = NULL;
932 unlock_new_inode(inode);
934 nfs_refresh_inode(inode, fattr);
935 dprintk("NFS: nfs_fhget(%s/%Ld ct=%d)\n",
937 (long long)NFS_FILEID(inode),
938 atomic_read(&inode->i_count));
944 dprintk("nfs_fhget: iget failed with error %ld\n", PTR_ERR(inode));
948 #define NFS_VALID_ATTRS (ATTR_MODE|ATTR_UID|ATTR_GID|ATTR_SIZE|ATTR_ATIME|ATTR_ATIME_SET|ATTR_MTIME|ATTR_MTIME_SET)
951 nfs_setattr(struct dentry *dentry, struct iattr *attr)
953 struct inode *inode = dentry->d_inode;
954 struct nfs_fattr fattr;
957 nfs_inc_stats(inode, NFSIOS_VFSSETATTR);
959 if (attr->ia_valid & ATTR_SIZE) {
960 if (!S_ISREG(inode->i_mode) || attr->ia_size == i_size_read(inode))
961 attr->ia_valid &= ~ATTR_SIZE;
964 /* Optimization: if the end result is no change, don't RPC */
965 attr->ia_valid &= NFS_VALID_ATTRS;
966 if (attr->ia_valid == 0)
970 nfs_begin_data_update(inode);
971 /* Write all dirty data */
972 filemap_write_and_wait(inode->i_mapping);
975 * Return any delegations if we're going to change ACLs
977 if ((attr->ia_valid & (ATTR_MODE|ATTR_UID|ATTR_GID)) != 0)
978 nfs_inode_return_delegation(inode);
979 error = NFS_PROTO(inode)->setattr(dentry, &fattr, attr);
981 nfs_refresh_inode(inode, &fattr);
982 nfs_end_data_update(inode);
988 * nfs_setattr_update_inode - Update inode metadata after a setattr call.
989 * @inode: pointer to struct inode
990 * @attr: pointer to struct iattr
992 * Note: we do this in the *proc.c in order to ensure that
993 * it works for things like exclusive creates too.
995 void nfs_setattr_update_inode(struct inode *inode, struct iattr *attr)
997 if ((attr->ia_valid & (ATTR_MODE|ATTR_UID|ATTR_GID)) != 0) {
998 if ((attr->ia_valid & ATTR_MODE) != 0) {
999 int mode = attr->ia_mode & S_IALLUGO;
1000 mode |= inode->i_mode & ~S_IALLUGO;
1001 inode->i_mode = mode;
1003 if ((attr->ia_valid & ATTR_UID) != 0)
1004 inode->i_uid = attr->ia_uid;
1005 if ((attr->ia_valid & ATTR_GID) != 0)
1006 inode->i_gid = attr->ia_gid;
1007 if ((attr->ia_valid & ATTR_XID) && IS_TAGXID(inode))
1008 inode->i_xid = attr->ia_xid;
1009 spin_lock(&inode->i_lock);
1010 NFS_I(inode)->cache_validity |= NFS_INO_INVALID_ACCESS|NFS_INO_INVALID_ACL;
1011 spin_unlock(&inode->i_lock);
1013 if ((attr->ia_valid & ATTR_SIZE) != 0) {
1014 nfs_inc_stats(inode, NFSIOS_SETATTRTRUNC);
1015 inode->i_size = attr->ia_size;
1016 vmtruncate(inode, attr->ia_size);
1020 static int nfs_wait_schedule(void *word)
1022 if (signal_pending(current))
1023 return -ERESTARTSYS;
1029 * Wait for the inode to get unlocked.
1031 static int nfs_wait_on_inode(struct inode *inode)
1033 struct rpc_clnt *clnt = NFS_CLIENT(inode);
1034 struct nfs_inode *nfsi = NFS_I(inode);
1038 rpc_clnt_sigmask(clnt, &oldmask);
1039 error = wait_on_bit_lock(&nfsi->flags, NFS_INO_REVALIDATING,
1040 nfs_wait_schedule, TASK_INTERRUPTIBLE);
1041 rpc_clnt_sigunmask(clnt, &oldmask);
1046 static void nfs_wake_up_inode(struct inode *inode)
1048 struct nfs_inode *nfsi = NFS_I(inode);
1050 clear_bit(NFS_INO_REVALIDATING, &nfsi->flags);
1051 smp_mb__after_clear_bit();
1052 wake_up_bit(&nfsi->flags, NFS_INO_REVALIDATING);
1055 int nfs_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
1057 struct inode *inode = dentry->d_inode;
1058 int need_atime = NFS_I(inode)->cache_validity & NFS_INO_INVALID_ATIME;
1061 /* Flush out writes to the server in order to update c/mtime */
1062 nfs_sync_inode_wait(inode, 0, 0, FLUSH_NOCOMMIT);
1065 * We may force a getattr if the user cares about atime.
1067 * Note that we only have to check the vfsmount flags here:
1068 * - NFS always sets S_NOATIME by so checking it would give a
1070 * - NFS never sets MS_NOATIME or MS_NODIRATIME so there is
1071 * no point in checking those.
1073 if ((mnt->mnt_flags & MNT_NOATIME) ||
1074 ((mnt->mnt_flags & MNT_NODIRATIME) && S_ISDIR(inode->i_mode)))
1078 err = __nfs_revalidate_inode(NFS_SERVER(inode), inode);
1080 err = nfs_revalidate_inode(NFS_SERVER(inode), inode);
1082 generic_fillattr(inode, stat);
1086 static struct nfs_open_context *alloc_nfs_open_context(struct vfsmount *mnt, struct dentry *dentry, struct rpc_cred *cred)
1088 struct nfs_open_context *ctx;
1090 ctx = (struct nfs_open_context *)kmalloc(sizeof(*ctx), GFP_KERNEL);
1092 atomic_set(&ctx->count, 1);
1093 ctx->dentry = dget(dentry);
1094 ctx->vfsmnt = mntget(mnt);
1095 ctx->cred = get_rpccred(cred);
1097 ctx->lockowner = current->files;
1099 ctx->dir_cookie = 0;
1104 struct nfs_open_context *get_nfs_open_context(struct nfs_open_context *ctx)
1107 atomic_inc(&ctx->count);
1111 void put_nfs_open_context(struct nfs_open_context *ctx)
1113 if (atomic_dec_and_test(&ctx->count)) {
1114 if (!list_empty(&ctx->list)) {
1115 struct inode *inode = ctx->dentry->d_inode;
1116 spin_lock(&inode->i_lock);
1117 list_del(&ctx->list);
1118 spin_unlock(&inode->i_lock);
1120 if (ctx->state != NULL)
1121 nfs4_close_state(ctx->state, ctx->mode);
1122 if (ctx->cred != NULL)
1123 put_rpccred(ctx->cred);
1125 mntput(ctx->vfsmnt);
1131 * Ensure that mmap has a recent RPC credential for use when writing out
1134 static void nfs_file_set_open_context(struct file *filp, struct nfs_open_context *ctx)
1136 struct inode *inode = filp->f_dentry->d_inode;
1137 struct nfs_inode *nfsi = NFS_I(inode);
1139 filp->private_data = get_nfs_open_context(ctx);
1140 spin_lock(&inode->i_lock);
1141 list_add(&ctx->list, &nfsi->open_files);
1142 spin_unlock(&inode->i_lock);
1146 * Given an inode, search for an open context with the desired characteristics
1148 struct nfs_open_context *nfs_find_open_context(struct inode *inode, struct rpc_cred *cred, int mode)
1150 struct nfs_inode *nfsi = NFS_I(inode);
1151 struct nfs_open_context *pos, *ctx = NULL;
1153 spin_lock(&inode->i_lock);
1154 list_for_each_entry(pos, &nfsi->open_files, list) {
1155 if (cred != NULL && pos->cred != cred)
1157 if ((pos->mode & mode) == mode) {
1158 ctx = get_nfs_open_context(pos);
1162 spin_unlock(&inode->i_lock);
1166 static void nfs_file_clear_open_context(struct file *filp)
1168 struct inode *inode = filp->f_dentry->d_inode;
1169 struct nfs_open_context *ctx = (struct nfs_open_context *)filp->private_data;
1172 filp->private_data = NULL;
1173 spin_lock(&inode->i_lock);
1174 list_move_tail(&ctx->list, &NFS_I(inode)->open_files);
1175 spin_unlock(&inode->i_lock);
1176 put_nfs_open_context(ctx);
1181 * These allocate and release file read/write context information.
1183 int nfs_open(struct inode *inode, struct file *filp)
1185 struct nfs_open_context *ctx;
1186 struct rpc_cred *cred;
1188 cred = rpcauth_lookupcred(NFS_CLIENT(inode)->cl_auth, 0);
1190 return PTR_ERR(cred);
1191 ctx = alloc_nfs_open_context(filp->f_vfsmnt, filp->f_dentry, cred);
1195 ctx->mode = filp->f_mode;
1196 nfs_file_set_open_context(filp, ctx);
1197 put_nfs_open_context(ctx);
1201 int nfs_release(struct inode *inode, struct file *filp)
1203 nfs_file_clear_open_context(filp);
1208 * This function is called whenever some part of NFS notices that
1209 * the cached attributes have to be refreshed.
1212 __nfs_revalidate_inode(struct nfs_server *server, struct inode *inode)
1214 int status = -ESTALE;
1215 struct nfs_fattr fattr;
1216 struct nfs_inode *nfsi = NFS_I(inode);
1218 dfprintk(PAGECACHE, "NFS: revalidating (%s/%Ld)\n",
1219 inode->i_sb->s_id, (long long)NFS_FILEID(inode));
1222 if (!inode || is_bad_inode(inode))
1224 if (NFS_STALE(inode))
1227 status = nfs_wait_on_inode(inode);
1230 if (NFS_STALE(inode)) {
1232 /* Do we trust the cached ESTALE? */
1233 if (NFS_ATTRTIMEO(inode) != 0) {
1234 if (nfsi->cache_validity & (NFS_INO_INVALID_ATTR|NFS_INO_INVALID_DATA|NFS_INO_INVALID_ATIME)) {
1241 status = NFS_PROTO(inode)->getattr(server, NFS_FH(inode), &fattr);
1243 dfprintk(PAGECACHE, "nfs_revalidate_inode: (%s/%Ld) getattr failed, error=%d\n",
1245 (long long)NFS_FILEID(inode), status);
1246 if (status == -ESTALE) {
1247 nfs_zap_caches(inode);
1248 if (!S_ISDIR(inode->i_mode))
1249 set_bit(NFS_INO_STALE, &NFS_FLAGS(inode));
1254 spin_lock(&inode->i_lock);
1255 status = nfs_update_inode(inode, &fattr);
1257 spin_unlock(&inode->i_lock);
1258 dfprintk(PAGECACHE, "nfs_revalidate_inode: (%s/%Ld) refresh failed, error=%d\n",
1260 (long long)NFS_FILEID(inode), status);
1263 spin_unlock(&inode->i_lock);
1265 nfs_revalidate_mapping(inode, inode->i_mapping);
1267 if (nfsi->cache_validity & NFS_INO_INVALID_ACL)
1268 nfs_zap_acl_cache(inode);
1270 dfprintk(PAGECACHE, "NFS: (%s/%Ld) revalidation complete\n",
1272 (long long)NFS_FILEID(inode));
1275 nfs_wake_up_inode(inode);
1282 int nfs_attribute_timeout(struct inode *inode)
1284 struct nfs_inode *nfsi = NFS_I(inode);
1286 if (nfs_have_delegation(inode, FMODE_READ))
1288 return time_after(jiffies, nfsi->read_cache_jiffies+nfsi->attrtimeo);
1292 * nfs_revalidate_inode - Revalidate the inode attributes
1293 * @server - pointer to nfs_server struct
1294 * @inode - pointer to inode struct
1296 * Updates inode attribute information by retrieving the data from the server.
1298 int nfs_revalidate_inode(struct nfs_server *server, struct inode *inode)
1300 nfs_inc_stats(inode, NFSIOS_INODEREVALIDATE);
1301 if (!(NFS_I(inode)->cache_validity & (NFS_INO_INVALID_ATTR|NFS_INO_INVALID_DATA))
1302 && !nfs_attribute_timeout(inode))
1303 return NFS_STALE(inode) ? -ESTALE : 0;
1304 return __nfs_revalidate_inode(server, inode);
1308 * nfs_revalidate_mapping - Revalidate the pagecache
1309 * @inode - pointer to host inode
1310 * @mapping - pointer to mapping
1312 void nfs_revalidate_mapping(struct inode *inode, struct address_space *mapping)
1314 struct nfs_inode *nfsi = NFS_I(inode);
1316 if (nfsi->cache_validity & NFS_INO_INVALID_DATA) {
1317 nfs_inc_stats(inode, NFSIOS_DATAINVALIDATE);
1318 if (S_ISREG(inode->i_mode))
1319 nfs_sync_mapping(mapping);
1320 invalidate_inode_pages2(mapping);
1322 spin_lock(&inode->i_lock);
1323 nfsi->cache_validity &= ~NFS_INO_INVALID_DATA;
1324 if (S_ISDIR(inode->i_mode)) {
1325 memset(nfsi->cookieverf, 0, sizeof(nfsi->cookieverf));
1326 /* This ensures we revalidate child dentries */
1327 nfsi->cache_change_attribute = jiffies;
1329 spin_unlock(&inode->i_lock);
1331 dfprintk(PAGECACHE, "NFS: (%s/%Ld) data cache invalidated\n",
1333 (long long)NFS_FILEID(inode));
1338 * nfs_begin_data_update
1339 * @inode - pointer to inode
1340 * Declare that a set of operations will update file data on the server
1342 void nfs_begin_data_update(struct inode *inode)
1344 atomic_inc(&NFS_I(inode)->data_updates);
1348 * nfs_end_data_update
1349 * @inode - pointer to inode
1350 * Declare end of the operations that will update file data
1351 * This will mark the inode as immediately needing revalidation
1352 * of its attribute cache.
1354 void nfs_end_data_update(struct inode *inode)
1356 struct nfs_inode *nfsi = NFS_I(inode);
1358 if (!nfs_have_delegation(inode, FMODE_READ)) {
1359 /* Directories and symlinks: invalidate page cache */
1360 if (S_ISDIR(inode->i_mode) || S_ISLNK(inode->i_mode)) {
1361 spin_lock(&inode->i_lock);
1362 nfsi->cache_validity |= NFS_INO_INVALID_DATA;
1363 spin_unlock(&inode->i_lock);
1366 nfsi->cache_change_attribute = jiffies;
1367 atomic_dec(&nfsi->data_updates);
1370 static void nfs_wcc_update_inode(struct inode *inode, struct nfs_fattr *fattr)
1372 struct nfs_inode *nfsi = NFS_I(inode);
1374 if ((fattr->valid & NFS_ATTR_PRE_CHANGE) != 0
1375 && nfsi->change_attr == fattr->pre_change_attr) {
1376 nfsi->change_attr = fattr->change_attr;
1377 nfsi->cache_change_attribute = jiffies;
1380 /* If we have atomic WCC data, we may update some attributes */
1381 if ((fattr->valid & NFS_ATTR_WCC) != 0) {
1382 if (timespec_equal(&inode->i_ctime, &fattr->pre_ctime)) {
1383 memcpy(&inode->i_ctime, &fattr->ctime, sizeof(inode->i_ctime));
1384 nfsi->cache_change_attribute = jiffies;
1386 if (timespec_equal(&inode->i_mtime, &fattr->pre_mtime)) {
1387 memcpy(&inode->i_mtime, &fattr->mtime, sizeof(inode->i_mtime));
1388 nfsi->cache_change_attribute = jiffies;
1390 if (inode->i_size == fattr->pre_size && nfsi->npages == 0) {
1391 inode->i_size = fattr->size;
1392 nfsi->cache_change_attribute = jiffies;
1398 * nfs_check_inode_attributes - verify consistency of the inode attribute cache
1399 * @inode - pointer to inode
1400 * @fattr - updated attributes
1402 * Verifies the attribute cache. If we have just changed the attributes,
1403 * so that fattr carries weak cache consistency data, then it may
1404 * also update the ctime/mtime/change_attribute.
1406 static int nfs_check_inode_attributes(struct inode *inode, struct nfs_fattr *fattr)
1408 struct nfs_inode *nfsi = NFS_I(inode);
1409 loff_t cur_size, new_isize;
1416 if ((fattr->valid & NFS_ATTR_FATTR) == 0)
1419 /* Has the inode gone and changed behind our back? */
1420 if (nfsi->fileid != fattr->fileid
1421 || (inode->i_mode & S_IFMT) != (fattr->mode & S_IFMT)) {
1425 /* Are we in the process of updating data on the server? */
1426 data_unstable = nfs_caches_unstable(inode);
1428 /* Do atomic weak cache consistency updates */
1429 nfs_wcc_update_inode(inode, fattr);
1431 if ((fattr->valid & NFS_ATTR_FATTR_V4) != 0) {
1432 if (nfsi->change_attr == fattr->change_attr)
1434 nfsi->cache_validity |= NFS_INO_INVALID_ATTR;
1436 nfsi->cache_validity |= NFS_INO_REVAL_PAGECACHE;
1439 /* Verify a few of the more important attributes */
1440 if (!timespec_equal(&inode->i_mtime, &fattr->mtime)) {
1441 nfsi->cache_validity |= NFS_INO_INVALID_ATTR;
1443 nfsi->cache_validity |= NFS_INO_REVAL_PAGECACHE;
1446 cur_size = i_size_read(inode);
1447 new_isize = nfs_size_to_loff_t(fattr->size);
1448 if (cur_size != new_isize && nfsi->npages == 0)
1449 nfsi->cache_validity |= NFS_INO_INVALID_ATTR|NFS_INO_REVAL_PAGECACHE;
1451 uid = INOXID_UID(XID_TAG(inode), fattr->uid, fattr->gid);
1452 gid = INOXID_GID(XID_TAG(inode), fattr->uid, fattr->gid);
1453 xid = INOXID_XID(XID_TAG(inode), fattr->uid, fattr->gid, 0);
1455 /* Have any file permissions changed? */
1456 if ((inode->i_mode & S_IALLUGO) != (fattr->mode & S_IALLUGO)
1457 || inode->i_uid != uid
1458 || inode->i_gid != gid
1459 || inode->i_xid != xid)
1460 nfsi->cache_validity |= NFS_INO_INVALID_ATTR | NFS_INO_INVALID_ACCESS | NFS_INO_INVALID_ACL;
1462 /* Has the link count changed? */
1463 if (inode->i_nlink != fattr->nlink)
1464 nfsi->cache_validity |= NFS_INO_INVALID_ATTR;
1467 if (!timespec_equal(&inode->i_atime, &fattr->atime))
1468 nfsi->cache_validity |= NFS_INO_INVALID_ATIME;
1470 nfsi->read_cache_jiffies = fattr->time_start;
1475 * nfs_refresh_inode - try to update the inode attribute cache
1476 * @inode - pointer to inode
1477 * @fattr - updated attributes
1479 * Check that an RPC call that returned attributes has not overlapped with
1480 * other recent updates of the inode metadata, then decide whether it is
1481 * safe to do a full update of the inode attributes, or whether just to
1482 * call nfs_check_inode_attributes.
1484 int nfs_refresh_inode(struct inode *inode, struct nfs_fattr *fattr)
1486 struct nfs_inode *nfsi = NFS_I(inode);
1489 if ((fattr->valid & NFS_ATTR_FATTR) == 0)
1491 spin_lock(&inode->i_lock);
1492 nfsi->cache_validity &= ~NFS_INO_REVAL_PAGECACHE;
1493 if (time_after(fattr->time_start, nfsi->last_updated))
1494 status = nfs_update_inode(inode, fattr);
1496 status = nfs_check_inode_attributes(inode, fattr);
1498 spin_unlock(&inode->i_lock);
1503 * nfs_post_op_update_inode - try to update the inode attribute cache
1504 * @inode - pointer to inode
1505 * @fattr - updated attributes
1507 * After an operation that has changed the inode metadata, mark the
1508 * attribute cache as being invalid, then try to update it.
1510 int nfs_post_op_update_inode(struct inode *inode, struct nfs_fattr *fattr)
1512 struct nfs_inode *nfsi = NFS_I(inode);
1515 spin_lock(&inode->i_lock);
1516 if (unlikely((fattr->valid & NFS_ATTR_FATTR) == 0)) {
1517 nfsi->cache_validity |= NFS_INO_INVALID_ATTR | NFS_INO_INVALID_ACCESS;
1520 status = nfs_update_inode(inode, fattr);
1522 spin_unlock(&inode->i_lock);
1527 * Many nfs protocol calls return the new file attributes after
1528 * an operation. Here we update the inode to reflect the state
1529 * of the server's inode.
1531 * This is a bit tricky because we have to make sure all dirty pages
1532 * have been sent off to the server before calling invalidate_inode_pages.
1533 * To make sure no other process adds more write requests while we try
1534 * our best to flush them, we make them sleep during the attribute refresh.
1536 * A very similar scenario holds for the dir cache.
1538 static int nfs_update_inode(struct inode *inode, struct nfs_fattr *fattr)
1540 struct nfs_inode *nfsi = NFS_I(inode);
1541 loff_t cur_isize, new_isize;
1542 unsigned int invalid = 0;
1548 dfprintk(VFS, "NFS: %s(%s/%ld ct=%d info=0x%x)\n",
1549 __FUNCTION__, inode->i_sb->s_id, inode->i_ino,
1550 atomic_read(&inode->i_count), fattr->valid);
1552 if ((fattr->valid & NFS_ATTR_FATTR) == 0)
1555 if (nfsi->fileid != fattr->fileid)
1559 * Make sure the inode's type hasn't changed.
1561 if ((inode->i_mode & S_IFMT) != (fattr->mode & S_IFMT))
1565 * Update the read time so we don't revalidate too often.
1567 nfsi->read_cache_jiffies = fattr->time_start;
1568 nfsi->last_updated = jiffies;
1570 /* Are we racing with known updates of the metadata on the server? */
1571 data_stable = nfs_verify_change_attribute(inode, fattr->time_start);
1573 nfsi->cache_validity &= ~(NFS_INO_INVALID_ATTR|NFS_INO_INVALID_ATIME);
1575 /* Do atomic weak cache consistency updates */
1576 nfs_wcc_update_inode(inode, fattr);
1578 /* Check if our cached file size is stale */
1579 new_isize = nfs_size_to_loff_t(fattr->size);
1580 cur_isize = i_size_read(inode);
1581 if (new_isize != cur_isize) {
1582 /* Do we perhaps have any outstanding writes? */
1583 if (nfsi->npages == 0) {
1584 /* No, but did we race with nfs_end_data_update()? */
1586 inode->i_size = new_isize;
1587 invalid |= NFS_INO_INVALID_DATA;
1589 invalid |= NFS_INO_INVALID_ATTR;
1590 } else if (new_isize > cur_isize) {
1591 inode->i_size = new_isize;
1592 invalid |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_DATA;
1594 nfsi->cache_change_attribute = jiffies;
1595 dprintk("NFS: isize change on server for file %s/%ld\n",
1596 inode->i_sb->s_id, inode->i_ino);
1599 /* Check if the mtime agrees */
1600 if (!timespec_equal(&inode->i_mtime, &fattr->mtime)) {
1601 memcpy(&inode->i_mtime, &fattr->mtime, sizeof(inode->i_mtime));
1602 dprintk("NFS: mtime change on server for file %s/%ld\n",
1603 inode->i_sb->s_id, inode->i_ino);
1604 invalid |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_DATA;
1605 nfsi->cache_change_attribute = jiffies;
1608 /* If ctime has changed we should definitely clear access+acl caches */
1609 if (!timespec_equal(&inode->i_ctime, &fattr->ctime)) {
1610 invalid |= NFS_INO_INVALID_ACCESS|NFS_INO_INVALID_ACL;
1611 memcpy(&inode->i_ctime, &fattr->ctime, sizeof(inode->i_ctime));
1612 nfsi->cache_change_attribute = jiffies;
1614 memcpy(&inode->i_atime, &fattr->atime, sizeof(inode->i_atime));
1616 uid = INOXID_UID(XID_TAG(inode), fattr->uid, fattr->gid);
1617 gid = INOXID_GID(XID_TAG(inode), fattr->uid, fattr->gid);
1618 xid = INOXID_XID(XID_TAG(inode), fattr->uid, fattr->gid, 0);
1620 if ((inode->i_mode & S_IALLUGO) != (fattr->mode & S_IALLUGO) ||
1621 inode->i_uid != uid ||
1622 inode->i_gid != gid ||
1623 inode->i_xid != xid)
1624 invalid |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_ACCESS|NFS_INO_INVALID_ACL;
1626 inode->i_mode = fattr->mode;
1627 inode->i_nlink = fattr->nlink;
1632 if (fattr->valid & (NFS_ATTR_FATTR_V3 | NFS_ATTR_FATTR_V4)) {
1634 * report the blocks in 512byte units
1636 inode->i_blocks = nfs_calc_block_size(fattr->du.nfs3.used);
1637 inode->i_blksize = inode->i_sb->s_blocksize;
1639 inode->i_blocks = fattr->du.nfs2.blocks;
1640 inode->i_blksize = fattr->du.nfs2.blocksize;
1643 if ((fattr->valid & NFS_ATTR_FATTR_V4)) {
1644 if (nfsi->change_attr != fattr->change_attr) {
1645 dprintk("NFS: change_attr change on server for file %s/%ld\n",
1646 inode->i_sb->s_id, inode->i_ino);
1647 nfsi->change_attr = fattr->change_attr;
1648 invalid |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_DATA|NFS_INO_INVALID_ACCESS|NFS_INO_INVALID_ACL;
1649 nfsi->cache_change_attribute = jiffies;
1651 invalid &= ~(NFS_INO_INVALID_ATTR|NFS_INO_INVALID_DATA);
1654 /* Update attrtimeo value if we're out of the unstable period */
1655 if (invalid & NFS_INO_INVALID_ATTR) {
1656 nfs_inc_stats(inode, NFSIOS_ATTRINVALIDATE);
1657 nfsi->attrtimeo = NFS_MINATTRTIMEO(inode);
1658 nfsi->attrtimeo_timestamp = jiffies;
1659 } else if (time_after(jiffies, nfsi->attrtimeo_timestamp+nfsi->attrtimeo)) {
1660 if ((nfsi->attrtimeo <<= 1) > NFS_MAXATTRTIMEO(inode))
1661 nfsi->attrtimeo = NFS_MAXATTRTIMEO(inode);
1662 nfsi->attrtimeo_timestamp = jiffies;
1664 /* Don't invalidate the data if we were to blame */
1665 if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode)
1666 || S_ISLNK(inode->i_mode)))
1667 invalid &= ~NFS_INO_INVALID_DATA;
1669 invalid &= ~(NFS_INO_INVALID_ATTR|NFS_INO_INVALID_ATIME|NFS_INO_REVAL_PAGECACHE);
1670 if (!nfs_have_delegation(inode, FMODE_READ))
1671 nfsi->cache_validity |= invalid;
1676 * Big trouble! The inode has become a different object.
1679 printk(KERN_DEBUG "%s: inode %ld mode changed, %07o to %07o\n",
1680 __FUNCTION__, inode->i_ino, inode->i_mode, fattr->mode);
1684 * No need to worry about unhashing the dentry, as the
1685 * lookup validation will know that the inode is bad.
1686 * (But we fall through to invalidate the caches.)
1688 nfs_invalidate_inode(inode);
1692 printk(KERN_ERR "NFS: server %s error: fileid changed\n"
1693 "fsid %s: expected fileid 0x%Lx, got 0x%Lx\n",
1694 NFS_SERVER(inode)->hostname, inode->i_sb->s_id,
1695 (long long)nfsi->fileid, (long long)fattr->fileid);
1700 * File system information
1703 static int nfs_set_super(struct super_block *s, void *data)
1705 s->s_fs_info = data;
1706 return set_anon_super(s, data);
1709 static int nfs_compare_super(struct super_block *sb, void *data)
1711 struct nfs_server *server = data;
1712 struct nfs_server *old = NFS_SB(sb);
1714 if (old->addr.sin_addr.s_addr != server->addr.sin_addr.s_addr)
1716 if (old->addr.sin_port != server->addr.sin_port)
1718 return !nfs_compare_fh(&old->fh, &server->fh);
1721 static struct super_block *nfs_get_sb(struct file_system_type *fs_type,
1722 int flags, const char *dev_name, void *raw_data)
1725 struct nfs_server *server = NULL;
1726 struct super_block *s;
1727 struct nfs_fh *root;
1728 struct nfs_mount_data *data = raw_data;
1730 s = ERR_PTR(-EINVAL);
1732 dprintk("%s: missing data argument\n", __FUNCTION__);
1735 if (data->version <= 0 || data->version > NFS_MOUNT_VERSION) {
1736 dprintk("%s: bad mount version\n", __FUNCTION__);
1739 switch (data->version) {
1745 if (data->flags & NFS_MOUNT_VER3) {
1746 dprintk("%s: mount structure version %d does not support NFSv3\n",
1751 data->root.size = NFS2_FHSIZE;
1752 memcpy(data->root.data, data->old_root.data, NFS2_FHSIZE);
1754 if (data->flags & NFS_MOUNT_SECFLAVOUR) {
1755 dprintk("%s: mount structure version %d does not support strong security\n",
1761 memset(data->context, 0, sizeof(data->context));
1763 #ifndef CONFIG_NFS_V3
1764 /* If NFSv3 is not compiled in, return -EPROTONOSUPPORT */
1765 s = ERR_PTR(-EPROTONOSUPPORT);
1766 if (data->flags & NFS_MOUNT_VER3) {
1767 dprintk("%s: NFSv3 not compiled into kernel\n", __FUNCTION__);
1770 #endif /* CONFIG_NFS_V3 */
1772 s = ERR_PTR(-ENOMEM);
1773 server = kzalloc(sizeof(struct nfs_server), GFP_KERNEL);
1776 /* Zero out the NFS state stuff */
1777 init_nfsv4_state(server);
1778 server->client = server->client_sys = server->client_acl = ERR_PTR(-EINVAL);
1781 if (data->flags & NFS_MOUNT_VER3)
1782 root->size = data->root.size;
1784 root->size = NFS2_FHSIZE;
1785 s = ERR_PTR(-EINVAL);
1786 if (root->size > sizeof(root->data)) {
1787 dprintk("%s: invalid root filehandle\n", __FUNCTION__);
1790 memcpy(root->data, data->root.data, root->size);
1792 /* We now require that the mount process passes the remote address */
1793 memcpy(&server->addr, &data->addr, sizeof(server->addr));
1794 if (server->addr.sin_addr.s_addr == INADDR_ANY) {
1795 dprintk("%s: mount program didn't pass remote address!\n",
1800 /* Fire up rpciod if not yet running */
1801 s = ERR_PTR(rpciod_up());
1803 dprintk("%s: couldn't start rpciod! Error = %ld\n",
1804 __FUNCTION__, PTR_ERR(s));
1808 s = sget(fs_type, nfs_compare_super, nfs_set_super, server);
1809 if (IS_ERR(s) || s->s_root)
1810 goto out_rpciod_down;
1814 error = nfs_fill_super(s, data, flags & MS_SILENT ? 1 : 0);
1816 up_write(&s->s_umount);
1817 deactivate_super(s);
1818 return ERR_PTR(error);
1820 s->s_flags |= MS_ACTIVE;
1829 static void nfs_kill_super(struct super_block *s)
1831 struct nfs_server *server = NFS_SB(s);
1835 if (!IS_ERR(server->client))
1836 rpc_shutdown_client(server->client);
1837 if (!IS_ERR(server->client_sys))
1838 rpc_shutdown_client(server->client_sys);
1839 if (!IS_ERR(server->client_acl))
1840 rpc_shutdown_client(server->client_acl);
1842 if (!(server->flags & NFS_MOUNT_NONLM))
1843 lockd_down(); /* release rpc.lockd */
1845 rpciod_down(); /* release rpciod */
1847 nfs_free_iostats(server->io_stats);
1848 kfree(server->hostname);
1852 static struct file_system_type nfs_fs_type = {
1853 .owner = THIS_MODULE,
1855 .get_sb = nfs_get_sb,
1856 .kill_sb = nfs_kill_super,
1857 .fs_flags = FS_ODD_RENAME|FS_REVAL_DOT|FS_BINARY_MOUNTDATA,
1860 #ifdef CONFIG_NFS_V4
1862 static void nfs4_clear_inode(struct inode *);
1865 static struct super_operations nfs4_sops = {
1866 .alloc_inode = nfs_alloc_inode,
1867 .destroy_inode = nfs_destroy_inode,
1868 .write_inode = nfs_write_inode,
1869 .delete_inode = nfs_delete_inode,
1870 .statfs = nfs_statfs,
1871 .clear_inode = nfs4_clear_inode,
1872 .umount_begin = nfs_umount_begin,
1873 .show_options = nfs_show_options,
1874 .show_stats = nfs_show_stats,
1878 * Clean out any remaining NFSv4 state that might be left over due
1879 * to open() calls that passed nfs_atomic_lookup, but failed to call
1882 static void nfs4_clear_inode(struct inode *inode)
1884 struct nfs_inode *nfsi = NFS_I(inode);
1886 /* If we are holding a delegation, return it! */
1887 nfs_inode_return_delegation(inode);
1888 /* First call standard NFS clear_inode() code */
1889 nfs_clear_inode(inode);
1890 /* Now clear out any remaining state */
1891 while (!list_empty(&nfsi->open_states)) {
1892 struct nfs4_state *state;
1894 state = list_entry(nfsi->open_states.next,
1897 dprintk("%s(%s/%Ld): found unclaimed NFSv4 state %p\n",
1900 (long long)NFS_FILEID(inode),
1902 BUG_ON(atomic_read(&state->count) != 1);
1903 nfs4_close_state(state, state->state);
1908 static int nfs4_fill_super(struct super_block *sb, struct nfs4_mount_data *data, int silent)
1910 struct nfs_server *server;
1911 struct nfs4_client *clp = NULL;
1912 struct rpc_xprt *xprt = NULL;
1913 struct rpc_clnt *clnt = NULL;
1914 struct rpc_timeout timeparms;
1915 rpc_authflavor_t authflavour;
1918 sb->s_blocksize_bits = 0;
1919 sb->s_blocksize = 0;
1920 server = NFS_SB(sb);
1921 if (data->rsize != 0)
1922 server->rsize = nfs_block_size(data->rsize, NULL);
1923 if (data->wsize != 0)
1924 server->wsize = nfs_block_size(data->wsize, NULL);
1925 server->flags = data->flags & NFS_MOUNT_FLAGMASK;
1926 server->caps = NFS_CAP_ATOMIC_OPEN;
1928 server->acregmin = data->acregmin*HZ;
1929 server->acregmax = data->acregmax*HZ;
1930 server->acdirmin = data->acdirmin*HZ;
1931 server->acdirmax = data->acdirmax*HZ;
1933 server->rpc_ops = &nfs_v4_clientops;
1935 nfs_init_timeout_values(&timeparms, data->proto, data->timeo, data->retrans);
1937 server->retrans_timeo = timeparms.to_initval;
1938 server->retrans_count = timeparms.to_retries;
1940 clp = nfs4_get_client(&server->addr.sin_addr);
1942 dprintk("%s: failed to create NFS4 client.\n", __FUNCTION__);
1946 /* Now create transport and client */
1947 authflavour = RPC_AUTH_UNIX;
1948 if (data->auth_flavourlen != 0) {
1949 if (data->auth_flavourlen != 1) {
1950 dprintk("%s: Invalid number of RPC auth flavours %d.\n",
1951 __FUNCTION__, data->auth_flavourlen);
1955 if (copy_from_user(&authflavour, data->auth_flavours, sizeof(authflavour))) {
1961 down_write(&clp->cl_sem);
1962 if (IS_ERR(clp->cl_rpcclient)) {
1963 xprt = xprt_create_proto(data->proto, &server->addr, &timeparms);
1965 up_write(&clp->cl_sem);
1966 err = PTR_ERR(xprt);
1967 dprintk("%s: cannot create RPC transport. Error = %d\n",
1971 clnt = rpc_create_client(xprt, server->hostname, &nfs_program,
1972 server->rpc_ops->version, authflavour);
1974 up_write(&clp->cl_sem);
1975 err = PTR_ERR(clnt);
1976 dprintk("%s: cannot create RPC client. Error = %d\n",
1981 clnt->cl_softrtry = 1;
1982 clp->cl_rpcclient = clnt;
1983 memcpy(clp->cl_ipaddr, server->ip_addr, sizeof(clp->cl_ipaddr));
1986 list_add_tail(&server->nfs4_siblings, &clp->cl_superblocks);
1987 clnt = rpc_clone_client(clp->cl_rpcclient);
1989 server->nfs4_state = clp;
1990 up_write(&clp->cl_sem);
1994 err = PTR_ERR(clnt);
1995 dprintk("%s: cannot create RPC client. Error = %d\n",
2000 server->client = clnt;
2002 if (server->nfs4_state->cl_idmap == NULL) {
2003 dprintk("%s: failed to create idmapper.\n", __FUNCTION__);
2007 if (clnt->cl_auth->au_flavor != authflavour) {
2008 struct rpc_auth *auth;
2010 auth = rpcauth_create(authflavour, clnt);
2012 dprintk("%s: couldn't create credcache!\n", __FUNCTION__);
2013 return PTR_ERR(auth);
2017 sb->s_time_gran = 1;
2019 sb->s_op = &nfs4_sops;
2020 err = nfs_sb_init(sb, authflavour);
2025 nfs4_put_client(clp);
2029 static int nfs4_compare_super(struct super_block *sb, void *data)
2031 struct nfs_server *server = data;
2032 struct nfs_server *old = NFS_SB(sb);
2034 if (strcmp(server->hostname, old->hostname) != 0)
2036 if (strcmp(server->mnt_path, old->mnt_path) != 0)
2042 nfs_copy_user_string(char *dst, struct nfs_string *src, int maxlen)
2047 return ERR_PTR(-EINVAL);
2048 if (src->len < maxlen)
2051 p = dst = kmalloc(maxlen + 1, GFP_KERNEL);
2053 return ERR_PTR(-ENOMEM);
2055 if (copy_from_user(dst, src->data, maxlen)) {
2057 return ERR_PTR(-EFAULT);
2063 static struct super_block *nfs4_get_sb(struct file_system_type *fs_type,
2064 int flags, const char *dev_name, void *raw_data)
2067 struct nfs_server *server;
2068 struct super_block *s;
2069 struct nfs4_mount_data *data = raw_data;
2073 dprintk("%s: missing data argument\n", __FUNCTION__);
2074 return ERR_PTR(-EINVAL);
2076 if (data->version <= 0 || data->version > NFS4_MOUNT_VERSION) {
2077 dprintk("%s: bad mount version\n", __FUNCTION__);
2078 return ERR_PTR(-EINVAL);
2081 server = kzalloc(sizeof(struct nfs_server), GFP_KERNEL);
2083 return ERR_PTR(-ENOMEM);
2084 /* Zero out the NFS state stuff */
2085 init_nfsv4_state(server);
2086 server->client = server->client_sys = server->client_acl = ERR_PTR(-EINVAL);
2088 p = nfs_copy_user_string(NULL, &data->hostname, 256);
2091 server->hostname = p;
2093 p = nfs_copy_user_string(NULL, &data->mnt_path, 1024);
2096 server->mnt_path = p;
2098 p = nfs_copy_user_string(server->ip_addr, &data->client_addr,
2099 sizeof(server->ip_addr) - 1);
2103 /* We now require that the mount process passes the remote address */
2104 if (data->host_addrlen != sizeof(server->addr)) {
2105 s = ERR_PTR(-EINVAL);
2108 if (copy_from_user(&server->addr, data->host_addr, sizeof(server->addr))) {
2109 s = ERR_PTR(-EFAULT);
2112 if (server->addr.sin_family != AF_INET ||
2113 server->addr.sin_addr.s_addr == INADDR_ANY) {
2114 dprintk("%s: mount program didn't pass remote IP address!\n",
2116 s = ERR_PTR(-EINVAL);
2120 /* Fire up rpciod if not yet running */
2121 s = ERR_PTR(rpciod_up());
2123 dprintk("%s: couldn't start rpciod! Error = %ld\n",
2124 __FUNCTION__, PTR_ERR(s));
2128 s = sget(fs_type, nfs4_compare_super, nfs_set_super, server);
2130 if (IS_ERR(s) || s->s_root)
2135 error = nfs4_fill_super(s, data, flags & MS_SILENT ? 1 : 0);
2137 up_write(&s->s_umount);
2138 deactivate_super(s);
2139 return ERR_PTR(error);
2141 s->s_flags |= MS_ACTIVE;
2144 s = (struct super_block *)p;
2146 kfree(server->mnt_path);
2147 kfree(server->hostname);
2152 static void nfs4_kill_super(struct super_block *sb)
2154 struct nfs_server *server = NFS_SB(sb);
2156 nfs_return_all_delegations(sb);
2157 kill_anon_super(sb);
2159 nfs4_renewd_prepare_shutdown(server);
2161 if (server->client != NULL && !IS_ERR(server->client))
2162 rpc_shutdown_client(server->client);
2164 destroy_nfsv4_state(server);
2168 nfs_free_iostats(server->io_stats);
2169 kfree(server->hostname);
2173 static struct file_system_type nfs4_fs_type = {
2174 .owner = THIS_MODULE,
2176 .get_sb = nfs4_get_sb,
2177 .kill_sb = nfs4_kill_super,
2178 .fs_flags = FS_ODD_RENAME|FS_REVAL_DOT|FS_BINARY_MOUNTDATA,
2181 static const int nfs_set_port_min = 0;
2182 static const int nfs_set_port_max = 65535;
2183 static int param_set_port(const char *val, struct kernel_param *kp)
2186 int num = simple_strtol(val, &endp, 0);
2187 if (endp == val || *endp || num < nfs_set_port_min || num > nfs_set_port_max)
2189 *((int *)kp->arg) = num;
2193 module_param_call(callback_tcpport, param_set_port, param_get_int,
2194 &nfs_callback_set_tcpport, 0644);
2196 static int param_set_idmap_timeout(const char *val, struct kernel_param *kp)
2199 int num = simple_strtol(val, &endp, 0);
2201 if (endp == val || *endp || num < 0 || jif < num)
2203 *((int *)kp->arg) = jif;
2207 module_param_call(idmap_cache_timeout, param_set_idmap_timeout, param_get_int,
2208 &nfs_idmap_cache_timeout, 0644);
2210 #define nfs4_init_once(nfsi) \
2212 INIT_LIST_HEAD(&(nfsi)->open_states); \
2213 nfsi->delegation = NULL; \
2214 nfsi->delegation_state = 0; \
2215 init_rwsem(&nfsi->rwsem); \
2218 static inline int register_nfs4fs(void)
2222 ret = nfs_register_sysctl();
2225 ret = register_filesystem(&nfs4_fs_type);
2227 nfs_unregister_sysctl();
2231 static inline void unregister_nfs4fs(void)
2233 unregister_filesystem(&nfs4_fs_type);
2234 nfs_unregister_sysctl();
2237 #define nfs4_init_once(nfsi) \
2239 #define register_nfs4fs() (0)
2240 #define unregister_nfs4fs()
2243 extern int nfs_init_nfspagecache(void);
2244 extern void nfs_destroy_nfspagecache(void);
2245 extern int nfs_init_readpagecache(void);
2246 extern void nfs_destroy_readpagecache(void);
2247 extern int nfs_init_writepagecache(void);
2248 extern void nfs_destroy_writepagecache(void);
2249 #ifdef CONFIG_NFS_DIRECTIO
2250 extern int nfs_init_directcache(void);
2251 extern void nfs_destroy_directcache(void);
2254 static kmem_cache_t * nfs_inode_cachep;
2256 static struct inode *nfs_alloc_inode(struct super_block *sb)
2258 struct nfs_inode *nfsi;
2259 nfsi = (struct nfs_inode *)kmem_cache_alloc(nfs_inode_cachep, SLAB_KERNEL);
2263 nfsi->cache_validity = 0UL;
2264 nfsi->cache_change_attribute = jiffies;
2265 #ifdef CONFIG_NFS_V3_ACL
2266 nfsi->acl_access = ERR_PTR(-EAGAIN);
2267 nfsi->acl_default = ERR_PTR(-EAGAIN);
2269 #ifdef CONFIG_NFS_V4
2270 nfsi->nfs4_acl = NULL;
2271 #endif /* CONFIG_NFS_V4 */
2272 return &nfsi->vfs_inode;
2275 static void nfs_destroy_inode(struct inode *inode)
2277 kmem_cache_free(nfs_inode_cachep, NFS_I(inode));
2280 static void init_once(void * foo, kmem_cache_t * cachep, unsigned long flags)
2282 struct nfs_inode *nfsi = (struct nfs_inode *) foo;
2284 if ((flags & (SLAB_CTOR_VERIFY|SLAB_CTOR_CONSTRUCTOR)) ==
2285 SLAB_CTOR_CONSTRUCTOR) {
2286 inode_init_once(&nfsi->vfs_inode);
2287 spin_lock_init(&nfsi->req_lock);
2288 INIT_LIST_HEAD(&nfsi->dirty);
2289 INIT_LIST_HEAD(&nfsi->commit);
2290 INIT_LIST_HEAD(&nfsi->open_files);
2291 INIT_RADIX_TREE(&nfsi->nfs_page_tree, GFP_ATOMIC);
2292 atomic_set(&nfsi->data_updates, 0);
2296 nfs4_init_once(nfsi);
2300 static int nfs_init_inodecache(void)
2302 nfs_inode_cachep = kmem_cache_create("nfs_inode_cache",
2303 sizeof(struct nfs_inode),
2304 0, (SLAB_RECLAIM_ACCOUNT|
2307 if (nfs_inode_cachep == NULL)
2313 static void nfs_destroy_inodecache(void)
2315 if (kmem_cache_destroy(nfs_inode_cachep))
2316 printk(KERN_INFO "nfs_inode_cache: not all structures were freed\n");
2322 static int __init init_nfs_fs(void)
2326 err = nfs_init_nfspagecache();
2330 err = nfs_init_inodecache();
2334 err = nfs_init_readpagecache();
2338 err = nfs_init_writepagecache();
2342 #ifdef CONFIG_NFS_DIRECTIO
2343 err = nfs_init_directcache();
2348 #ifdef CONFIG_PROC_FS
2349 rpc_proc_register(&nfs_rpcstat);
2351 err = register_filesystem(&nfs_fs_type);
2354 if ((err = register_nfs4fs()) != 0)
2358 #ifdef CONFIG_PROC_FS
2359 rpc_proc_unregister("nfs");
2361 #ifdef CONFIG_NFS_DIRECTIO
2362 nfs_destroy_directcache();
2365 nfs_destroy_writepagecache();
2367 nfs_destroy_readpagecache();
2369 nfs_destroy_inodecache();
2371 nfs_destroy_nfspagecache();
2376 static void __exit exit_nfs_fs(void)
2378 #ifdef CONFIG_NFS_DIRECTIO
2379 nfs_destroy_directcache();
2381 nfs_destroy_writepagecache();
2382 nfs_destroy_readpagecache();
2383 nfs_destroy_inodecache();
2384 nfs_destroy_nfspagecache();
2385 #ifdef CONFIG_PROC_FS
2386 rpc_proc_unregister("nfs");
2388 unregister_filesystem(&nfs_fs_type);
2389 unregister_nfs4fs();
2392 /* Not quite true; I just maintain it */
2393 MODULE_AUTHOR("Olaf Kirch <okir@monad.swb.de>");
2394 MODULE_LICENSE("GPL");
2396 module_init(init_nfs_fs)
2397 module_exit(exit_nfs_fs)