4 * Copyright (C) 1992 Rick Sladkey
6 * nfs inode and superblock handling functions
8 * Modularised by Alan Cox <Alan.Cox@linux.org>, while hacking some
9 * experimental NFS changes. Modularisation taken straight from SYS5 fs.
11 * Change to nfs_read_super() to permit NFS mounts to multi-homed hosts.
12 * J.S.Peatfield@damtp.cam.ac.uk
16 #include <linux/config.h>
17 #include <linux/module.h>
18 #include <linux/init.h>
20 #include <linux/time.h>
21 #include <linux/kernel.h>
23 #include <linux/string.h>
24 #include <linux/stat.h>
25 #include <linux/errno.h>
26 #include <linux/unistd.h>
27 #include <linux/sunrpc/clnt.h>
28 #include <linux/sunrpc/stats.h>
29 #include <linux/nfs_fs.h>
30 #include <linux/nfs_mount.h>
31 #include <linux/nfs4_mount.h>
32 #include <linux/lockd/bind.h>
33 #include <linux/smp_lock.h>
34 #include <linux/seq_file.h>
35 #include <linux/mount.h>
36 #include <linux/nfs_idmap.h>
37 #include <linux/vfs.h>
39 #include <asm/system.h>
40 #include <asm/uaccess.h>
42 #include "delegation.h"
44 #define NFSDBG_FACILITY NFSDBG_VFS
45 #define NFS_PARANOIA 1
47 /* Maximum number of readahead requests
48 * FIXME: this should really be a sysctl so that users may tune it to suit
49 * their needs. People that do NFS over a slow network, might for
50 * instance want to reduce it to something closer to 1 for improved
51 * interactive response.
53 #define NFS_MAX_READAHEAD (RPC_DEF_SLOT_TABLE - 1)
55 static void nfs_invalidate_inode(struct inode *);
56 static int nfs_update_inode(struct inode *, struct nfs_fattr *, unsigned long);
58 static struct inode *nfs_alloc_inode(struct super_block *sb);
59 static void nfs_destroy_inode(struct inode *);
60 static int nfs_write_inode(struct inode *,int);
61 static void nfs_delete_inode(struct inode *);
62 static void nfs_clear_inode(struct inode *);
63 static void nfs_umount_begin(struct super_block *);
64 static int nfs_statfs(struct super_block *, struct kstatfs *);
65 static int nfs_show_options(struct seq_file *, struct vfsmount *);
67 static struct super_operations nfs_sops = {
68 .alloc_inode = nfs_alloc_inode,
69 .destroy_inode = nfs_destroy_inode,
70 .write_inode = nfs_write_inode,
71 .delete_inode = nfs_delete_inode,
73 .clear_inode = nfs_clear_inode,
74 .umount_begin = nfs_umount_begin,
75 .show_options = nfs_show_options,
81 struct rpc_stat nfs_rpcstat = {
82 .program = &nfs_program
84 static struct rpc_version * nfs_version[] = {
88 #if defined(CONFIG_NFS_V3)
90 #elif defined(CONFIG_NFS_V4)
93 #if defined(CONFIG_NFS_V4)
98 struct rpc_program nfs_program = {
100 .number = NFS_PROGRAM,
101 .nrvers = sizeof(nfs_version) / sizeof(nfs_version[0]),
102 .version = nfs_version,
103 .stats = &nfs_rpcstat,
104 .pipe_dir_name = "/nfs",
107 static inline unsigned long
108 nfs_fattr_to_ino_t(struct nfs_fattr *fattr)
110 return nfs_fileid_to_ino_t(fattr->fileid);
114 nfs_write_inode(struct inode *inode, int sync)
116 int flags = sync ? FLUSH_WAIT : 0;
119 ret = nfs_commit_inode(inode, 0, 0, flags);
126 nfs_delete_inode(struct inode * inode)
128 dprintk("NFS: delete_inode(%s/%ld)\n", inode->i_sb->s_id, inode->i_ino);
132 * The following should never happen...
134 if (nfs_have_writebacks(inode)) {
135 printk(KERN_ERR "nfs_delete_inode: inode %ld has pending RPC requests\n", inode->i_ino);
142 * For the moment, the only task for the NFS clear_inode method is to
143 * release the mmap credential
146 nfs_clear_inode(struct inode *inode)
148 struct nfs_inode *nfsi = NFS_I(inode);
149 struct rpc_cred *cred;
152 BUG_ON (!list_empty(&nfsi->open_files));
153 cred = nfsi->cache_access.cred;
156 BUG_ON(atomic_read(&nfsi->data_updates) != 0);
160 nfs_umount_begin(struct super_block *sb)
162 struct nfs_server *server = NFS_SB(sb);
163 struct rpc_clnt *rpc;
165 /* -EIO all pending I/O */
166 if ((rpc = server->client) != NULL)
167 rpc_killall_tasks(rpc);
171 static inline unsigned long
172 nfs_block_bits(unsigned long bsize, unsigned char *nrbitsp)
174 /* make sure blocksize is a power of two */
175 if ((bsize & (bsize - 1)) || nrbitsp) {
176 unsigned char nrbits;
178 for (nrbits = 31; nrbits && !(bsize & (1 << nrbits)); nrbits--)
189 * Calculate the number of 512byte blocks used.
191 static inline unsigned long
192 nfs_calc_block_size(u64 tsize)
194 loff_t used = (tsize + 511) >> 9;
195 return (used > ULONG_MAX) ? ULONG_MAX : used;
199 * Compute and set NFS server blocksize
201 static inline unsigned long
202 nfs_block_size(unsigned long bsize, unsigned char *nrbitsp)
205 bsize = NFS_DEF_FILE_IO_BUFFER_SIZE;
206 else if (bsize >= NFS_MAX_FILE_IO_BUFFER_SIZE)
207 bsize = NFS_MAX_FILE_IO_BUFFER_SIZE;
209 return nfs_block_bits(bsize, nrbitsp);
213 * Obtain the root inode of the file system.
215 static struct inode *
216 nfs_get_root(struct super_block *sb, struct nfs_fh *rootfh, struct nfs_fsinfo *fsinfo)
218 struct nfs_server *server = NFS_SB(sb);
222 error = server->rpc_ops->getroot(server, rootfh, fsinfo);
224 dprintk("nfs_get_root: getattr error = %d\n", -error);
225 return ERR_PTR(error);
228 rooti = nfs_fhget(sb, rootfh, fsinfo->fattr);
230 return ERR_PTR(-ENOMEM);
235 * Do NFS version-independent mount processing, and sanity checking
238 nfs_sb_init(struct super_block *sb, rpc_authflavor_t authflavor)
240 struct nfs_server *server;
241 struct inode *root_inode;
242 struct nfs_fattr fattr;
243 struct nfs_fsinfo fsinfo = {
246 struct nfs_pathconf pathinfo = {
249 int no_root_error = 0;
251 /* We probably want something more informative here */
252 snprintf(sb->s_id, sizeof(sb->s_id), "%x:%x", MAJOR(sb->s_dev), MINOR(sb->s_dev));
256 sb->s_magic = NFS_SUPER_MAGIC;
258 root_inode = nfs_get_root(sb, &server->fh, &fsinfo);
259 /* Did getting the root inode fail? */
260 if (IS_ERR(root_inode)) {
261 no_root_error = PTR_ERR(root_inode);
264 sb->s_root = d_alloc_root(root_inode);
266 no_root_error = -ENOMEM;
269 sb->s_root->d_op = server->rpc_ops->dentry_ops;
271 /* Get some general file system info */
272 if (server->namelen == 0 &&
273 server->rpc_ops->pathconf(server, &server->fh, &pathinfo) >= 0)
274 server->namelen = pathinfo.max_namelen;
275 /* Work out a lot of parameters */
276 if (server->rsize == 0)
277 server->rsize = nfs_block_size(fsinfo.rtpref, NULL);
278 if (server->wsize == 0)
279 server->wsize = nfs_block_size(fsinfo.wtpref, NULL);
281 if (fsinfo.rtmax >= 512 && server->rsize > fsinfo.rtmax)
282 server->rsize = nfs_block_size(fsinfo.rtmax, NULL);
283 if (fsinfo.wtmax >= 512 && server->wsize > fsinfo.wtmax)
284 server->wsize = nfs_block_size(fsinfo.wtmax, NULL);
286 server->rpages = (server->rsize + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT;
287 if (server->rpages > NFS_READ_MAXIOV) {
288 server->rpages = NFS_READ_MAXIOV;
289 server->rsize = server->rpages << PAGE_CACHE_SHIFT;
292 server->wpages = (server->wsize + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT;
293 if (server->wpages > NFS_WRITE_MAXIOV) {
294 server->wpages = NFS_WRITE_MAXIOV;
295 server->wsize = server->wpages << PAGE_CACHE_SHIFT;
298 if (sb->s_blocksize == 0)
299 sb->s_blocksize = nfs_block_bits(server->wsize,
300 &sb->s_blocksize_bits);
301 server->wtmult = nfs_block_bits(fsinfo.wtmult, NULL);
303 server->dtsize = nfs_block_size(fsinfo.dtpref, NULL);
304 if (server->dtsize > PAGE_CACHE_SIZE)
305 server->dtsize = PAGE_CACHE_SIZE;
306 if (server->dtsize > server->rsize)
307 server->dtsize = server->rsize;
309 if (server->flags & NFS_MOUNT_NOAC) {
310 server->acregmin = server->acregmax = 0;
311 server->acdirmin = server->acdirmax = 0;
312 sb->s_flags |= MS_SYNCHRONOUS;
314 server->backing_dev_info.ra_pages = server->rpages * NFS_MAX_READAHEAD;
316 sb->s_maxbytes = fsinfo.maxfilesize;
317 if (sb->s_maxbytes > MAX_LFS_FILESIZE)
318 sb->s_maxbytes = MAX_LFS_FILESIZE;
320 /* We're airborne Set socket buffersize */
321 rpc_setbufsize(server->client, server->wsize + 100, server->rsize + 100);
323 /* Yargs. It didn't work out. */
325 dprintk("nfs_sb_init: get root inode failed: errno %d\n", -no_root_error);
326 if (!IS_ERR(root_inode))
328 return no_root_error;
332 * Create an RPC client handle.
334 static struct rpc_clnt *
335 nfs_create_client(struct nfs_server *server, const struct nfs_mount_data *data)
337 struct rpc_timeout timeparms;
338 struct rpc_xprt *xprt = NULL;
339 struct rpc_clnt *clnt = NULL;
340 int tcp = (data->flags & NFS_MOUNT_TCP);
342 /* Initialize timeout values */
343 timeparms.to_initval = data->timeo * HZ / 10;
344 timeparms.to_retries = data->retrans;
345 timeparms.to_maxval = tcp ? RPC_MAX_TCP_TIMEOUT : RPC_MAX_UDP_TIMEOUT;
346 timeparms.to_exponential = 1;
348 if (!timeparms.to_initval)
349 timeparms.to_initval = (tcp ? 600 : 11) * HZ / 10;
350 if (!timeparms.to_retries)
351 timeparms.to_retries = 5;
353 /* create transport and client */
354 xprt = xprt_create_proto(tcp ? IPPROTO_TCP : IPPROTO_UDP,
355 &server->addr, &timeparms);
357 printk(KERN_WARNING "NFS: cannot create RPC transport.\n");
358 return (struct rpc_clnt *)xprt;
360 clnt = rpc_create_client(xprt, server->hostname, &nfs_program,
361 server->rpc_ops->version, data->pseudoflavor);
363 printk(KERN_WARNING "NFS: cannot create RPC client.\n");
367 clnt->cl_intr = (server->flags & NFS_MOUNT_INTR) ? 1 : 0;
368 clnt->cl_softrtry = (server->flags & NFS_MOUNT_SOFT) ? 1 : 0;
369 clnt->cl_droppriv = (server->flags & NFS_MOUNT_BROKEN_SUID) ? 1 : 0;
380 * The way this works is that the mount process passes a structure
381 * in the data argument which contains the server's IP address
382 * and the root file handle obtained from the server's mount
383 * daemon. We stash these away in the private superblock fields.
386 nfs_fill_super(struct super_block *sb, struct nfs_mount_data *data, int silent)
388 struct nfs_server *server;
389 rpc_authflavor_t authflavor;
392 sb->s_blocksize_bits = 0;
395 sb->s_blocksize = nfs_block_size(data->bsize, &sb->s_blocksize_bits);
397 server->rsize = nfs_block_size(data->rsize, NULL);
399 server->wsize = nfs_block_size(data->wsize, NULL);
400 server->flags = data->flags & NFS_MOUNT_FLAGMASK;
402 server->acregmin = data->acregmin*HZ;
403 server->acregmax = data->acregmax*HZ;
404 server->acdirmin = data->acdirmin*HZ;
405 server->acdirmax = data->acdirmax*HZ;
407 /* Start lockd here, before we might error out */
408 if (!(server->flags & NFS_MOUNT_NONLM))
411 server->namelen = data->namlen;
412 server->hostname = kmalloc(strlen(data->hostname) + 1, GFP_KERNEL);
413 if (!server->hostname)
415 strcpy(server->hostname, data->hostname);
417 /* Check NFS protocol revision and initialize RPC op vector
418 * and file handle pool. */
419 if (server->flags & NFS_MOUNT_VER3) {
421 server->rpc_ops = &nfs_v3_clientops;
422 server->caps |= NFS_CAP_READDIRPLUS;
423 if (data->version < 4) {
424 printk(KERN_NOTICE "NFS: NFSv3 not supported by mount program.\n");
428 printk(KERN_NOTICE "NFS: NFSv3 not supported.\n");
432 server->rpc_ops = &nfs_v2_clientops;
435 /* Fill in pseudoflavor for mount version < 5 */
436 if (!(data->flags & NFS_MOUNT_SECFLAVOUR))
437 data->pseudoflavor = RPC_AUTH_UNIX;
438 authflavor = data->pseudoflavor; /* save for sb_init() */
439 /* XXX maybe we want to add a server->pseudoflavor field */
441 /* Create RPC client handles */
442 server->client = nfs_create_client(server, data);
443 if (IS_ERR(server->client))
444 return PTR_ERR(server->client);
445 /* RFC 2623, sec 2.3.2 */
446 if (authflavor != RPC_AUTH_UNIX) {
447 server->client_sys = rpc_clone_client(server->client);
448 if (IS_ERR(server->client_sys))
449 return PTR_ERR(server->client_sys);
450 if (!rpcauth_create(RPC_AUTH_UNIX, server->client_sys))
453 atomic_inc(&server->client->cl_count);
454 server->client_sys = server->client;
457 if (server->flags & NFS_MOUNT_VER3) {
458 if (server->namelen == 0 || server->namelen > NFS3_MAXNAMLEN)
459 server->namelen = NFS3_MAXNAMLEN;
461 if (server->namelen == 0 || server->namelen > NFS2_MAXNAMLEN)
462 server->namelen = NFS2_MAXNAMLEN;
465 sb->s_op = &nfs_sops;
466 return nfs_sb_init(sb, authflavor);
470 nfs_statfs(struct super_block *sb, struct kstatfs *buf)
472 struct nfs_server *server = NFS_SB(sb);
473 unsigned char blockbits;
474 unsigned long blockres;
475 struct nfs_fh *rootfh = NFS_FH(sb->s_root->d_inode);
476 struct nfs_fattr fattr;
477 struct nfs_fsstat res = {
484 error = server->rpc_ops->statfs(server, rootfh, &res);
485 buf->f_type = NFS_SUPER_MAGIC;
489 buf->f_frsize = server->wtmult;
490 buf->f_bsize = sb->s_blocksize;
491 blockbits = sb->s_blocksize_bits;
492 blockres = (1 << blockbits) - 1;
493 buf->f_blocks = (res.tbytes + blockres) >> blockbits;
494 buf->f_bfree = (res.fbytes + blockres) >> blockbits;
495 buf->f_bavail = (res.abytes + blockres) >> blockbits;
496 buf->f_files = res.tfiles;
497 buf->f_ffree = res.afiles;
499 buf->f_namelen = server->namelen;
506 printk(KERN_WARNING "nfs_statfs: statfs error = %d\n", -error);
507 buf->f_bsize = buf->f_blocks = buf->f_bfree = buf->f_bavail = -1;
512 static int nfs_show_options(struct seq_file *m, struct vfsmount *mnt)
514 static struct proc_nfs_info {
519 { NFS_MOUNT_SOFT, ",soft", ",hard" },
520 { NFS_MOUNT_INTR, ",intr", "" },
521 { NFS_MOUNT_POSIX, ",posix", "" },
522 { NFS_MOUNT_TCP, ",tcp", ",udp" },
523 { NFS_MOUNT_NOCTO, ",nocto", "" },
524 { NFS_MOUNT_NOAC, ",noac", "" },
525 { NFS_MOUNT_NONLM, ",nolock", ",lock" },
526 { NFS_MOUNT_BROKEN_SUID, ",broken_suid", "" },
529 struct proc_nfs_info *nfs_infop;
530 struct nfs_server *nfss = NFS_SB(mnt->mnt_sb);
532 seq_printf(m, ",v%d", nfss->rpc_ops->version);
533 seq_printf(m, ",rsize=%d", nfss->rsize);
534 seq_printf(m, ",wsize=%d", nfss->wsize);
535 if (nfss->acregmin != 3*HZ)
536 seq_printf(m, ",acregmin=%d", nfss->acregmin/HZ);
537 if (nfss->acregmax != 60*HZ)
538 seq_printf(m, ",acregmax=%d", nfss->acregmax/HZ);
539 if (nfss->acdirmin != 30*HZ)
540 seq_printf(m, ",acdirmin=%d", nfss->acdirmin/HZ);
541 if (nfss->acdirmax != 60*HZ)
542 seq_printf(m, ",acdirmax=%d", nfss->acdirmax/HZ);
543 for (nfs_infop = nfs_info; nfs_infop->flag; nfs_infop++) {
544 if (nfss->flags & nfs_infop->flag)
545 seq_puts(m, nfs_infop->str);
547 seq_puts(m, nfs_infop->nostr);
549 seq_puts(m, ",addr=");
550 seq_escape(m, nfss->hostname, " \t\n\\");
555 * Invalidate the local caches
558 nfs_zap_caches(struct inode *inode)
560 struct nfs_inode *nfsi = NFS_I(inode);
561 int mode = inode->i_mode;
563 NFS_ATTRTIMEO(inode) = NFS_MINATTRTIMEO(inode);
564 NFS_ATTRTIMEO_UPDATE(inode) = jiffies;
566 memset(NFS_COOKIEVERF(inode), 0, sizeof(NFS_COOKIEVERF(inode)));
567 if (S_ISREG(mode) || S_ISDIR(mode) || S_ISLNK(mode))
568 nfsi->flags |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_DATA;
570 nfsi->flags |= NFS_INO_INVALID_ATTR;
574 * Invalidate, but do not unhash, the inode
577 nfs_invalidate_inode(struct inode *inode)
579 umode_t save_mode = inode->i_mode;
581 make_bad_inode(inode);
582 inode->i_mode = save_mode;
583 nfs_zap_caches(inode);
586 struct nfs_find_desc {
588 struct nfs_fattr *fattr;
592 * In NFSv3 we can have 64bit inode numbers. In order to support
593 * this, and re-exported directories (also seen in NFSv2)
594 * we are forced to allow 2 different inodes to have the same
598 nfs_find_actor(struct inode *inode, void *opaque)
600 struct nfs_find_desc *desc = (struct nfs_find_desc *)opaque;
601 struct nfs_fh *fh = desc->fh;
602 struct nfs_fattr *fattr = desc->fattr;
604 if (NFS_FILEID(inode) != fattr->fileid)
606 if (nfs_compare_fh(NFS_FH(inode), fh))
608 if (is_bad_inode(inode))
614 nfs_init_locked(struct inode *inode, void *opaque)
616 struct nfs_find_desc *desc = (struct nfs_find_desc *)opaque;
617 struct nfs_fattr *fattr = desc->fattr;
619 NFS_FILEID(inode) = fattr->fileid;
620 nfs_copy_fh(NFS_FH(inode), desc->fh);
624 /* Don't use READDIRPLUS on directories that we believe are too large */
625 #define NFS_LIMIT_READDIRPLUS (8*PAGE_SIZE)
628 * This is our front-end to iget that looks up inodes by file handle
629 * instead of inode number.
632 nfs_fhget(struct super_block *sb, struct nfs_fh *fh, struct nfs_fattr *fattr)
634 struct nfs_find_desc desc = {
638 struct inode *inode = NULL;
641 if ((fattr->valid & NFS_ATTR_FATTR) == 0)
645 printk("NFS: Buggy server - nlink == 0!\n");
649 hash = nfs_fattr_to_ino_t(fattr);
651 if (!(inode = iget5_locked(sb, hash, nfs_find_actor, nfs_init_locked, &desc)))
654 if (inode->i_state & I_NEW) {
655 struct nfs_inode *nfsi = NFS_I(inode);
657 /* We set i_ino for the few things that still rely on it,
661 /* We can't support update_atime(), since the server will reset it */
662 inode->i_flags |= S_NOATIME|S_NOCMTIME;
663 inode->i_mode = fattr->mode;
664 /* Why so? Because we want revalidate for devices/FIFOs, and
665 * that's precisely what we have in nfs_file_inode_operations.
667 inode->i_op = &nfs_file_inode_operations;
668 if (S_ISREG(inode->i_mode)) {
669 inode->i_fop = &nfs_file_operations;
670 inode->i_data.a_ops = &nfs_file_aops;
671 inode->i_data.backing_dev_info = &NFS_SB(sb)->backing_dev_info;
672 } else if (S_ISDIR(inode->i_mode)) {
673 inode->i_op = NFS_SB(sb)->rpc_ops->dir_inode_ops;
674 inode->i_fop = &nfs_dir_operations;
675 if (nfs_server_capable(inode, NFS_CAP_READDIRPLUS)
676 && fattr->size <= NFS_LIMIT_READDIRPLUS)
677 NFS_FLAGS(inode) |= NFS_INO_ADVISE_RDPLUS;
678 } else if (S_ISLNK(inode->i_mode))
679 inode->i_op = &nfs_symlink_inode_operations;
681 init_special_inode(inode, inode->i_mode, fattr->rdev);
683 nfsi->read_cache_jiffies = fattr->timestamp;
684 inode->i_atime = fattr->atime;
685 inode->i_mtime = fattr->mtime;
686 inode->i_ctime = fattr->ctime;
687 if (fattr->valid & NFS_ATTR_FATTR_V4)
688 nfsi->change_attr = fattr->change_attr;
689 inode->i_size = nfs_size_to_loff_t(fattr->size);
690 inode->i_nlink = fattr->nlink;
691 inode->i_uid = fattr->uid;
692 inode->i_gid = fattr->gid;
693 if (fattr->valid & (NFS_ATTR_FATTR_V3 | NFS_ATTR_FATTR_V4)) {
695 * report the blocks in 512byte units
697 inode->i_blocks = nfs_calc_block_size(fattr->du.nfs3.used);
698 inode->i_blksize = inode->i_sb->s_blocksize;
700 inode->i_blocks = fattr->du.nfs2.blocks;
701 inode->i_blksize = fattr->du.nfs2.blocksize;
703 nfsi->attrtimeo = NFS_MINATTRTIMEO(inode);
704 nfsi->attrtimeo_timestamp = jiffies;
705 memset(nfsi->cookieverf, 0, sizeof(nfsi->cookieverf));
706 nfsi->cache_access.cred = NULL;
708 unlock_new_inode(inode);
710 nfs_refresh_inode(inode, fattr);
711 dprintk("NFS: nfs_fhget(%s/%Ld ct=%d)\n",
713 (long long)NFS_FILEID(inode),
714 atomic_read(&inode->i_count));
720 printk("nfs_fhget: iget failed\n");
724 #define NFS_VALID_ATTRS (ATTR_MODE|ATTR_UID|ATTR_GID|ATTR_SIZE|ATTR_ATIME|ATTR_ATIME_SET|ATTR_MTIME|ATTR_MTIME_SET)
727 nfs_setattr(struct dentry *dentry, struct iattr *attr)
729 struct inode *inode = dentry->d_inode;
730 struct nfs_fattr fattr;
733 if (attr->ia_valid & ATTR_SIZE) {
734 if (!S_ISREG(inode->i_mode) || attr->ia_size == i_size_read(inode))
735 attr->ia_valid &= ~ATTR_SIZE;
738 /* Optimization: if the end result is no change, don't RPC */
739 attr->ia_valid &= NFS_VALID_ATTRS;
740 if (attr->ia_valid == 0)
744 nfs_begin_data_update(inode);
745 /* Write all dirty data if we're changing file permissions or size */
746 if ((attr->ia_valid & (ATTR_MODE|ATTR_UID|ATTR_GID|ATTR_SIZE)) != 0) {
747 if (filemap_fdatawrite(inode->i_mapping) == 0)
748 filemap_fdatawait(inode->i_mapping);
751 error = NFS_PROTO(inode)->setattr(dentry, &fattr, attr);
753 nfs_refresh_inode(inode, &fattr);
754 if ((attr->ia_valid & ATTR_MODE) != 0) {
756 mode = inode->i_mode & ~S_IALLUGO;
757 mode |= attr->ia_mode & S_IALLUGO;
758 inode->i_mode = mode;
760 if ((attr->ia_valid & ATTR_UID) != 0)
761 inode->i_uid = attr->ia_uid;
762 if ((attr->ia_valid & ATTR_GID) != 0)
763 inode->i_gid = attr->ia_gid;
764 if ((attr->ia_valid & ATTR_SIZE) != 0) {
765 inode->i_size = attr->ia_size;
766 vmtruncate(inode, attr->ia_size);
769 if ((attr->ia_valid & (ATTR_MODE|ATTR_UID|ATTR_GID)) != 0) {
770 struct rpc_cred **cred = &NFS_I(inode)->cache_access.cred;
776 nfs_end_data_update(inode);
782 * Wait for the inode to get unlocked.
783 * (Used for NFS_INO_LOCKED and NFS_INO_REVALIDATING).
786 nfs_wait_on_inode(struct inode *inode, int flag)
788 struct rpc_clnt *clnt = NFS_CLIENT(inode);
789 struct nfs_inode *nfsi = NFS_I(inode);
792 if (!(NFS_FLAGS(inode) & flag))
794 atomic_inc(&inode->i_count);
795 error = nfs_wait_event(clnt, nfsi->nfs_i_wait,
796 !(NFS_FLAGS(inode) & flag));
801 int nfs_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
803 struct inode *inode = dentry->d_inode;
804 struct nfs_inode *nfsi = NFS_I(inode);
805 int need_atime = nfsi->flags & NFS_INO_INVALID_ATIME;
808 if (__IS_FLG(inode, MS_NOATIME))
810 else if (__IS_FLG(inode, MS_NODIRATIME) && S_ISDIR(inode->i_mode))
812 /* We may force a getattr if the user cares about atime */
814 err = __nfs_revalidate_inode(NFS_SERVER(inode), inode);
816 err = nfs_revalidate_inode(NFS_SERVER(inode), inode);
818 generic_fillattr(inode, stat);
822 struct nfs_open_context *alloc_nfs_open_context(struct dentry *dentry, struct rpc_cred *cred)
824 struct nfs_open_context *ctx;
826 ctx = (struct nfs_open_context *)kmalloc(sizeof(*ctx), GFP_KERNEL);
828 atomic_set(&ctx->count, 1);
829 ctx->dentry = dget(dentry);
830 ctx->cred = get_rpccred(cred);
832 ctx->lockowner = current->files;
834 init_waitqueue_head(&ctx->waitq);
839 struct nfs_open_context *get_nfs_open_context(struct nfs_open_context *ctx)
842 atomic_inc(&ctx->count);
846 void put_nfs_open_context(struct nfs_open_context *ctx)
848 if (atomic_dec_and_test(&ctx->count)) {
849 if (ctx->state != NULL)
850 nfs4_close_state(ctx->state, ctx->mode);
851 if (ctx->cred != NULL)
852 put_rpccred(ctx->cred);
859 * Ensure that mmap has a recent RPC credential for use when writing out
862 void nfs_file_set_open_context(struct file *filp, struct nfs_open_context *ctx)
864 struct inode *inode = filp->f_dentry->d_inode;
865 struct nfs_inode *nfsi = NFS_I(inode);
867 filp->private_data = get_nfs_open_context(ctx);
868 spin_lock(&inode->i_lock);
869 list_add(&ctx->list, &nfsi->open_files);
870 spin_unlock(&inode->i_lock);
873 struct nfs_open_context *nfs_find_open_context(struct inode *inode, int mode)
875 struct nfs_inode *nfsi = NFS_I(inode);
876 struct nfs_open_context *pos, *ctx = NULL;
878 spin_lock(&inode->i_lock);
879 list_for_each_entry(pos, &nfsi->open_files, list) {
880 if ((pos->mode & mode) == mode) {
881 ctx = get_nfs_open_context(pos);
885 spin_unlock(&inode->i_lock);
889 void nfs_file_clear_open_context(struct file *filp)
891 struct inode *inode = filp->f_dentry->d_inode;
892 struct nfs_open_context *ctx = (struct nfs_open_context *)filp->private_data;
895 filp->private_data = NULL;
896 spin_lock(&inode->i_lock);
897 list_del(&ctx->list);
898 spin_unlock(&inode->i_lock);
899 put_nfs_open_context(ctx);
904 * These allocate and release file read/write context information.
906 int nfs_open(struct inode *inode, struct file *filp)
908 struct nfs_open_context *ctx;
909 struct rpc_cred *cred;
911 if ((cred = rpcauth_lookupcred(NFS_CLIENT(inode)->cl_auth, 0)) == NULL)
913 ctx = alloc_nfs_open_context(filp->f_dentry, cred);
917 ctx->mode = filp->f_mode;
918 nfs_file_set_open_context(filp, ctx);
919 put_nfs_open_context(ctx);
920 if ((filp->f_mode & FMODE_WRITE) != 0)
921 nfs_begin_data_update(inode);
925 int nfs_release(struct inode *inode, struct file *filp)
927 if ((filp->f_mode & FMODE_WRITE) != 0)
928 nfs_end_data_update(inode);
929 nfs_file_clear_open_context(filp);
934 * This function is called whenever some part of NFS notices that
935 * the cached attributes have to be refreshed.
938 __nfs_revalidate_inode(struct nfs_server *server, struct inode *inode)
940 int status = -ESTALE;
941 struct nfs_fattr fattr;
942 struct nfs_inode *nfsi = NFS_I(inode);
943 unsigned long verifier;
946 dfprintk(PAGECACHE, "NFS: revalidating (%s/%Ld)\n",
947 inode->i_sb->s_id, (long long)NFS_FILEID(inode));
950 if (!inode || is_bad_inode(inode))
952 if (NFS_STALE(inode) && inode != inode->i_sb->s_root->d_inode)
955 while (NFS_REVALIDATING(inode)) {
956 status = nfs_wait_on_inode(inode, NFS_INO_REVALIDATING);
959 if (NFS_SERVER(inode)->flags & NFS_MOUNT_NOAC)
961 if (NFS_FLAGS(inode) & (NFS_INO_INVALID_ATTR|NFS_INO_INVALID_DATA|NFS_INO_INVALID_ATIME))
963 status = NFS_STALE(inode) ? -ESTALE : 0;
966 NFS_FLAGS(inode) |= NFS_INO_REVALIDATING;
968 /* Protect against RPC races by saving the change attribute */
969 verifier = nfs_save_change_attribute(inode);
970 status = NFS_PROTO(inode)->getattr(server, NFS_FH(inode), &fattr);
972 dfprintk(PAGECACHE, "nfs_revalidate_inode: (%s/%Ld) getattr failed, error=%d\n",
974 (long long)NFS_FILEID(inode), status);
975 if (status == -ESTALE) {
976 NFS_FLAGS(inode) |= NFS_INO_STALE;
977 if (inode != inode->i_sb->s_root->d_inode)
978 remove_inode_hash(inode);
983 status = nfs_update_inode(inode, &fattr, verifier);
985 dfprintk(PAGECACHE, "nfs_revalidate_inode: (%s/%Ld) refresh failed, error=%d\n",
987 (long long)NFS_FILEID(inode), status);
992 * We may need to keep the attributes marked as invalid if
993 * we raced with nfs_end_attr_update().
995 if (verifier == nfsi->cache_change_attribute)
996 nfsi->flags &= ~(NFS_INO_INVALID_ATTR|NFS_INO_INVALID_ATIME);
997 /* Do the page cache invalidation */
998 if (flags & NFS_INO_INVALID_DATA) {
999 if (S_ISREG(inode->i_mode)) {
1000 if (filemap_fdatawrite(inode->i_mapping) == 0)
1001 filemap_fdatawait(inode->i_mapping);
1004 nfsi->flags &= ~NFS_INO_INVALID_DATA;
1005 invalidate_inode_pages2(inode->i_mapping);
1006 memset(NFS_COOKIEVERF(inode), 0, sizeof(NFS_COOKIEVERF(inode)));
1007 dfprintk(PAGECACHE, "NFS: (%s/%Ld) data cache invalidated\n",
1009 (long long)NFS_FILEID(inode));
1010 /* This ensures we revalidate dentries */
1011 nfsi->cache_change_attribute++;
1013 dfprintk(PAGECACHE, "NFS: (%s/%Ld) revalidation complete\n",
1015 (long long)NFS_FILEID(inode));
1017 NFS_FLAGS(inode) &= ~NFS_INO_STALE;
1019 NFS_FLAGS(inode) &= ~NFS_INO_REVALIDATING;
1020 wake_up(&nfsi->nfs_i_wait);
1026 int nfs_attribute_timeout(struct inode *inode)
1028 struct nfs_inode *nfsi = NFS_I(inode);
1030 if (nfs_have_delegation(inode, FMODE_READ))
1032 return time_after(jiffies, nfsi->read_cache_jiffies+nfsi->attrtimeo);
1036 * nfs_revalidate_inode - Revalidate the inode attributes
1037 * @server - pointer to nfs_server struct
1038 * @inode - pointer to inode struct
1040 * Updates inode attribute information by retrieving the data from the server.
1042 int nfs_revalidate_inode(struct nfs_server *server, struct inode *inode)
1044 if (!(NFS_FLAGS(inode) & (NFS_INO_INVALID_ATTR|NFS_INO_INVALID_DATA))
1045 && !nfs_attribute_timeout(inode))
1046 return NFS_STALE(inode) ? -ESTALE : 0;
1047 return __nfs_revalidate_inode(server, inode);
1051 * nfs_begin_data_update
1052 * @inode - pointer to inode
1053 * Declare that a set of operations will update file data on the server
1055 void nfs_begin_data_update(struct inode *inode)
1057 atomic_inc(&NFS_I(inode)->data_updates);
1061 * nfs_end_data_update
1062 * @inode - pointer to inode
1063 * Declare end of the operations that will update file data
1064 * This will mark the inode as immediately needing revalidation
1065 * of its attribute cache.
1067 void nfs_end_data_update(struct inode *inode)
1069 struct nfs_inode *nfsi = NFS_I(inode);
1071 if (!nfs_have_delegation(inode, FMODE_READ)) {
1072 /* Mark the attribute cache for revalidation */
1073 nfsi->flags |= NFS_INO_INVALID_ATTR;
1074 /* Directories and symlinks: invalidate page cache too */
1075 if (S_ISDIR(inode->i_mode) || S_ISLNK(inode->i_mode))
1076 nfsi->flags |= NFS_INO_INVALID_DATA;
1078 nfsi->cache_change_attribute ++;
1079 atomic_dec(&nfsi->data_updates);
1083 * nfs_end_data_update_defer
1084 * @inode - pointer to inode
1085 * Declare end of the operations that will update file data
1086 * This will defer marking the inode as needing revalidation
1087 * unless there are no other pending updates.
1089 void nfs_end_data_update_defer(struct inode *inode)
1091 struct nfs_inode *nfsi = NFS_I(inode);
1093 if (atomic_dec_and_test(&nfsi->data_updates)) {
1094 /* Mark the attribute cache for revalidation */
1095 nfsi->flags |= NFS_INO_INVALID_ATTR;
1096 /* Directories and symlinks: invalidate page cache too */
1097 if (S_ISDIR(inode->i_mode) || S_ISLNK(inode->i_mode))
1098 nfsi->flags |= NFS_INO_INVALID_DATA;
1099 nfsi->cache_change_attribute ++;
1104 * nfs_refresh_inode - verify consistency of the inode attribute cache
1105 * @inode - pointer to inode
1106 * @fattr - updated attributes
1108 * Verifies the attribute cache. If we have just changed the attributes,
1109 * so that fattr carries weak cache consistency data, then it may
1110 * also update the ctime/mtime/change_attribute.
1112 int nfs_refresh_inode(struct inode *inode, struct nfs_fattr *fattr)
1114 struct nfs_inode *nfsi = NFS_I(inode);
1115 loff_t cur_size, new_isize;
1118 /* Do we hold a delegation? */
1119 if (nfs_have_delegation(inode, FMODE_READ))
1122 /* Are we in the process of updating data on the server? */
1123 data_unstable = nfs_caches_unstable(inode);
1125 if (fattr->valid & NFS_ATTR_FATTR_V4) {
1126 if ((fattr->valid & NFS_ATTR_PRE_CHANGE) != 0
1127 && nfsi->change_attr == fattr->pre_change_attr)
1128 nfsi->change_attr = fattr->change_attr;
1129 if (!data_unstable && nfsi->change_attr != fattr->change_attr)
1130 nfsi->flags |= NFS_INO_INVALID_ATTR;
1133 if ((fattr->valid & NFS_ATTR_FATTR) == 0)
1136 /* Has the inode gone and changed behind our back? */
1137 if (nfsi->fileid != fattr->fileid
1138 || (inode->i_mode & S_IFMT) != (fattr->mode & S_IFMT))
1141 cur_size = i_size_read(inode);
1142 new_isize = nfs_size_to_loff_t(fattr->size);
1144 /* If we have atomic WCC data, we may update some attributes */
1145 if ((fattr->valid & NFS_ATTR_WCC) != 0) {
1146 if (timespec_equal(&inode->i_ctime, &fattr->pre_ctime))
1147 memcpy(&inode->i_ctime, &fattr->ctime, sizeof(inode->i_ctime));
1148 if (timespec_equal(&inode->i_mtime, &fattr->pre_mtime))
1149 memcpy(&inode->i_mtime, &fattr->mtime, sizeof(inode->i_mtime));
1152 /* Verify a few of the more important attributes */
1153 if (!data_unstable) {
1154 if (!timespec_equal(&inode->i_mtime, &fattr->mtime)
1155 || cur_size != new_isize)
1156 nfsi->flags |= NFS_INO_INVALID_ATTR;
1157 } else if (S_ISREG(inode->i_mode) && new_isize > cur_size)
1158 nfsi->flags |= NFS_INO_INVALID_ATTR;
1160 /* Have any file permissions changed? */
1161 if ((inode->i_mode & S_IALLUGO) != (fattr->mode & S_IALLUGO)
1162 || inode->i_uid != fattr->uid
1163 || inode->i_gid != fattr->gid)
1164 nfsi->flags |= NFS_INO_INVALID_ATTR;
1166 /* Has the link count changed? */
1167 if (inode->i_nlink != fattr->nlink)
1168 nfsi->flags |= NFS_INO_INVALID_ATTR;
1170 if (!timespec_equal(&inode->i_atime, &fattr->atime))
1171 nfsi->flags |= NFS_INO_INVALID_ATIME;
1173 nfsi->read_cache_jiffies = fattr->timestamp;
1178 * Many nfs protocol calls return the new file attributes after
1179 * an operation. Here we update the inode to reflect the state
1180 * of the server's inode.
1182 * This is a bit tricky because we have to make sure all dirty pages
1183 * have been sent off to the server before calling invalidate_inode_pages.
1184 * To make sure no other process adds more write requests while we try
1185 * our best to flush them, we make them sleep during the attribute refresh.
1187 * A very similar scenario holds for the dir cache.
1189 static int nfs_update_inode(struct inode *inode, struct nfs_fattr *fattr, unsigned long verifier)
1191 struct nfs_inode *nfsi = NFS_I(inode);
1194 unsigned int invalid = 0;
1198 dfprintk(VFS, "NFS: %s(%s/%ld ct=%d info=0x%x)\n",
1199 __FUNCTION__, inode->i_sb->s_id, inode->i_ino,
1200 atomic_read(&inode->i_count), fattr->valid);
1202 if ((fattr->valid & NFS_ATTR_FATTR) == 0)
1205 if (nfsi->fileid != fattr->fileid) {
1206 printk(KERN_ERR "%s: inode number mismatch\n"
1207 "expected (%s/0x%Lx), got (%s/0x%Lx)\n",
1209 inode->i_sb->s_id, (long long)nfsi->fileid,
1210 inode->i_sb->s_id, (long long)fattr->fileid);
1215 * Make sure the inode's type hasn't changed.
1217 if ((inode->i_mode & S_IFMT) != (fattr->mode & S_IFMT))
1221 * Update the read time so we don't revalidate too often.
1223 nfsi->read_cache_jiffies = fattr->timestamp;
1225 /* Are we racing with known updates of the metadata on the server? */
1226 data_unstable = ! nfs_verify_change_attribute(inode, verifier);
1228 /* Check if the file size agrees */
1229 new_size = fattr->size;
1230 new_isize = nfs_size_to_loff_t(fattr->size);
1231 cur_isize = i_size_read(inode);
1232 if (cur_isize != new_size) {
1233 #ifdef NFS_DEBUG_VERBOSE
1234 printk(KERN_DEBUG "NFS: isize change on %s/%ld\n", inode->i_sb->s_id, inode->i_ino);
1237 * If we have pending writebacks, things can get
1240 if (S_ISREG(inode->i_mode) && data_unstable) {
1241 if (new_isize > cur_isize) {
1242 inode->i_size = new_isize;
1243 invalid |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_DATA;
1246 inode->i_size = new_isize;
1247 invalid |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_DATA;
1252 * Note: we don't check inode->i_mtime since pipes etc.
1253 * can change this value in VFS without requiring a
1254 * cache revalidation.
1256 if (!timespec_equal(&inode->i_mtime, &fattr->mtime)) {
1257 memcpy(&inode->i_mtime, &fattr->mtime, sizeof(inode->i_mtime));
1258 #ifdef NFS_DEBUG_VERBOSE
1259 printk(KERN_DEBUG "NFS: mtime change on %s/%ld\n", inode->i_sb->s_id, inode->i_ino);
1262 invalid |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_DATA;
1265 if ((fattr->valid & NFS_ATTR_FATTR_V4)
1266 && nfsi->change_attr != fattr->change_attr) {
1267 #ifdef NFS_DEBUG_VERBOSE
1268 printk(KERN_DEBUG "NFS: change_attr change on %s/%ld\n",
1269 inode->i_sb->s_id, inode->i_ino);
1271 nfsi->change_attr = fattr->change_attr;
1273 invalid |= NFS_INO_INVALID_ATTR|NFS_INO_INVALID_DATA;
1276 memcpy(&inode->i_ctime, &fattr->ctime, sizeof(inode->i_ctime));
1277 memcpy(&inode->i_atime, &fattr->atime, sizeof(inode->i_atime));
1279 if ((inode->i_mode & S_IALLUGO) != (fattr->mode & S_IALLUGO) ||
1280 inode->i_uid != fattr->uid ||
1281 inode->i_gid != fattr->gid) {
1282 struct rpc_cred **cred = &NFS_I(inode)->cache_access.cred;
1287 invalid |= NFS_INO_INVALID_ATTR;
1290 inode->i_mode = fattr->mode;
1291 inode->i_nlink = fattr->nlink;
1292 inode->i_uid = fattr->uid;
1293 inode->i_gid = fattr->gid;
1295 if (fattr->valid & (NFS_ATTR_FATTR_V3 | NFS_ATTR_FATTR_V4)) {
1297 * report the blocks in 512byte units
1299 inode->i_blocks = nfs_calc_block_size(fattr->du.nfs3.used);
1300 inode->i_blksize = inode->i_sb->s_blocksize;
1302 inode->i_blocks = fattr->du.nfs2.blocks;
1303 inode->i_blksize = fattr->du.nfs2.blocksize;
1306 /* Update attrtimeo value if we're out of the unstable period */
1307 if (invalid & NFS_INO_INVALID_ATTR) {
1308 nfsi->attrtimeo = NFS_MINATTRTIMEO(inode);
1309 nfsi->attrtimeo_timestamp = jiffies;
1310 } else if (time_after(jiffies, nfsi->attrtimeo_timestamp+nfsi->attrtimeo)) {
1311 if ((nfsi->attrtimeo <<= 1) > NFS_MAXATTRTIMEO(inode))
1312 nfsi->attrtimeo = NFS_MAXATTRTIMEO(inode);
1313 nfsi->attrtimeo_timestamp = jiffies;
1315 /* Don't invalidate the data if we were to blame */
1316 if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode)
1317 || S_ISLNK(inode->i_mode)))
1318 invalid &= ~NFS_INO_INVALID_DATA;
1319 if (!nfs_have_delegation(inode, FMODE_READ))
1320 nfsi->flags |= invalid;
1325 * Big trouble! The inode has become a different object.
1328 printk(KERN_DEBUG "%s: inode %ld mode changed, %07o to %07o\n",
1329 __FUNCTION__, inode->i_ino, inode->i_mode, fattr->mode);
1332 * No need to worry about unhashing the dentry, as the
1333 * lookup validation will know that the inode is bad.
1334 * (But we fall through to invalidate the caches.)
1336 nfs_invalidate_inode(inode);
1342 * File system information
1345 static int nfs_set_super(struct super_block *s, void *data)
1347 s->s_fs_info = data;
1348 return set_anon_super(s, data);
1351 static int nfs_compare_super(struct super_block *sb, void *data)
1353 struct nfs_server *server = data;
1354 struct nfs_server *old = NFS_SB(sb);
1356 if (old->addr.sin_addr.s_addr != server->addr.sin_addr.s_addr)
1358 if (old->addr.sin_port != server->addr.sin_port)
1360 return !nfs_compare_fh(&old->fh, &server->fh);
1363 static struct super_block *nfs_get_sb(struct file_system_type *fs_type,
1364 int flags, const char *dev_name, void *raw_data)
1367 struct nfs_server *server;
1368 struct super_block *s;
1369 struct nfs_fh *root;
1370 struct nfs_mount_data *data = raw_data;
1373 printk("nfs_read_super: missing data argument\n");
1374 return ERR_PTR(-EINVAL);
1377 server = kmalloc(sizeof(struct nfs_server), GFP_KERNEL);
1379 return ERR_PTR(-ENOMEM);
1380 memset(server, 0, sizeof(struct nfs_server));
1381 /* Zero out the NFS state stuff */
1382 init_nfsv4_state(server);
1384 if (data->version != NFS_MOUNT_VERSION) {
1385 printk("nfs warning: mount version %s than kernel\n",
1386 data->version < NFS_MOUNT_VERSION ? "older" : "newer");
1387 if (data->version < 2)
1389 if (data->version < 3)
1391 if (data->version < 4) {
1392 data->flags &= ~NFS_MOUNT_VER3;
1393 data->root.size = NFS2_FHSIZE;
1394 memcpy(data->root.data, data->old_root.data, NFS2_FHSIZE);
1396 if (data->version < 5)
1397 data->flags &= ~NFS_MOUNT_SECFLAVOUR;
1401 if (data->flags & NFS_MOUNT_VER3)
1402 root->size = data->root.size;
1404 root->size = NFS2_FHSIZE;
1405 if (root->size > sizeof(root->data)) {
1406 printk("nfs_get_sb: invalid root filehandle\n");
1408 return ERR_PTR(-EINVAL);
1410 memcpy(root->data, data->root.data, root->size);
1412 /* We now require that the mount process passes the remote address */
1413 memcpy(&server->addr, &data->addr, sizeof(server->addr));
1414 if (server->addr.sin_addr.s_addr == INADDR_ANY) {
1415 printk("NFS: mount program didn't pass remote address!\n");
1417 return ERR_PTR(-EINVAL);
1420 s = sget(fs_type, nfs_compare_super, nfs_set_super, server);
1422 if (IS_ERR(s) || s->s_root) {
1429 /* Fire up rpciod if not yet running */
1430 if (rpciod_up() != 0) {
1431 printk(KERN_WARNING "NFS: couldn't start rpciod!\n");
1433 return ERR_PTR(-EIO);
1436 error = nfs_fill_super(s, data, flags & MS_VERBOSE ? 1 : 0);
1438 up_write(&s->s_umount);
1439 deactivate_super(s);
1440 return ERR_PTR(error);
1442 s->s_flags |= MS_ACTIVE;
1446 static void nfs_kill_super(struct super_block *s)
1448 struct nfs_server *server = NFS_SB(s);
1452 nfs4_renewd_prepare_shutdown(server);
1454 if (server->client != NULL && !IS_ERR(server->client))
1455 rpc_shutdown_client(server->client);
1456 if (server->client_sys != NULL && !IS_ERR(server->client_sys))
1457 rpc_shutdown_client(server->client_sys);
1459 if (!(server->flags & NFS_MOUNT_NONLM))
1460 lockd_down(); /* release rpc.lockd */
1462 rpciod_down(); /* release rpciod */
1464 destroy_nfsv4_state(server);
1466 if (server->hostname != NULL)
1467 kfree(server->hostname);
1471 static struct file_system_type nfs_fs_type = {
1472 .owner = THIS_MODULE,
1474 .get_sb = nfs_get_sb,
1475 .kill_sb = nfs_kill_super,
1476 .fs_flags = FS_ODD_RENAME|FS_REVAL_DOT|FS_BINARY_MOUNTDATA,
1479 #ifdef CONFIG_NFS_V4
1481 static void nfs4_clear_inode(struct inode *);
1484 static struct super_operations nfs4_sops = {
1485 .alloc_inode = nfs_alloc_inode,
1486 .destroy_inode = nfs_destroy_inode,
1487 .write_inode = nfs_write_inode,
1488 .delete_inode = nfs_delete_inode,
1489 .statfs = nfs_statfs,
1490 .clear_inode = nfs4_clear_inode,
1491 .umount_begin = nfs_umount_begin,
1492 .show_options = nfs_show_options,
1496 * Clean out any remaining NFSv4 state that might be left over due
1497 * to open() calls that passed nfs_atomic_lookup, but failed to call
1500 static void nfs4_clear_inode(struct inode *inode)
1502 struct nfs_inode *nfsi = NFS_I(inode);
1504 /* If we are holding a delegation, return it! */
1505 if (nfsi->delegation != NULL)
1506 nfs_inode_return_delegation(inode);
1507 /* First call standard NFS clear_inode() code */
1508 nfs_clear_inode(inode);
1509 /* Now clear out any remaining state */
1510 while (!list_empty(&nfsi->open_states)) {
1511 struct nfs4_state *state;
1513 state = list_entry(nfsi->open_states.next,
1516 dprintk("%s(%s/%Ld): found unclaimed NFSv4 state %p\n",
1519 (long long)NFS_FILEID(inode),
1521 BUG_ON(atomic_read(&state->count) != 1);
1522 nfs4_close_state(state, state->state);
1527 static int nfs4_fill_super(struct super_block *sb, struct nfs4_mount_data *data, int silent)
1529 struct nfs_server *server;
1530 struct nfs4_client *clp = NULL;
1531 struct rpc_xprt *xprt = NULL;
1532 struct rpc_clnt *clnt = NULL;
1533 struct rpc_timeout timeparms;
1534 rpc_authflavor_t authflavour;
1535 int proto, err = -EIO;
1537 sb->s_blocksize_bits = 0;
1538 sb->s_blocksize = 0;
1539 server = NFS_SB(sb);
1540 if (data->rsize != 0)
1541 server->rsize = nfs_block_size(data->rsize, NULL);
1542 if (data->wsize != 0)
1543 server->wsize = nfs_block_size(data->wsize, NULL);
1544 server->flags = data->flags & NFS_MOUNT_FLAGMASK;
1546 /* NFSv4 doesn't use NLM locking */
1547 server->flags |= NFS_MOUNT_NONLM;
1549 server->acregmin = data->acregmin*HZ;
1550 server->acregmax = data->acregmax*HZ;
1551 server->acdirmin = data->acdirmin*HZ;
1552 server->acdirmax = data->acdirmax*HZ;
1554 server->rpc_ops = &nfs_v4_clientops;
1555 /* Initialize timeout values */
1557 timeparms.to_initval = data->timeo * HZ / 10;
1558 timeparms.to_retries = data->retrans;
1559 timeparms.to_exponential = 1;
1560 if (!timeparms.to_retries)
1561 timeparms.to_retries = 5;
1563 proto = data->proto;
1564 /* Which IP protocol do we use? */
1567 timeparms.to_maxval = RPC_MAX_TCP_TIMEOUT;
1568 if (!timeparms.to_initval)
1569 timeparms.to_initval = 600 * HZ / 10;
1572 timeparms.to_maxval = RPC_MAX_UDP_TIMEOUT;
1573 if (!timeparms.to_initval)
1574 timeparms.to_initval = 11 * HZ / 10;
1580 clp = nfs4_get_client(&server->addr.sin_addr);
1582 printk(KERN_WARNING "NFS: failed to create NFS4 client.\n");
1586 /* Now create transport and client */
1587 authflavour = RPC_AUTH_UNIX;
1588 if (data->auth_flavourlen != 0) {
1589 if (data->auth_flavourlen > 1)
1590 printk(KERN_INFO "NFS: cannot yet deal with multiple auth flavours.\n");
1591 if (copy_from_user(&authflavour, data->auth_flavours, sizeof(authflavour))) {
1597 down_write(&clp->cl_sem);
1598 if (clp->cl_rpcclient == NULL) {
1599 xprt = xprt_create_proto(proto, &server->addr, &timeparms);
1601 up_write(&clp->cl_sem);
1602 printk(KERN_WARNING "NFS: cannot create RPC transport.\n");
1603 err = PTR_ERR(xprt);
1606 clnt = rpc_create_client(xprt, server->hostname, &nfs_program,
1607 server->rpc_ops->version, authflavour);
1609 up_write(&clp->cl_sem);
1610 printk(KERN_WARNING "NFS: cannot create RPC client.\n");
1612 err = PTR_ERR(clnt);
1615 clnt->cl_chatty = 1;
1616 clp->cl_rpcclient = clnt;
1617 clp->cl_cred = rpcauth_lookupcred(clnt->cl_auth, 0);
1618 memcpy(clp->cl_ipaddr, server->ip_addr, sizeof(clp->cl_ipaddr));
1621 if (list_empty(&clp->cl_superblocks)) {
1622 err = nfs4_init_client(clp);
1624 up_write(&clp->cl_sem);
1628 list_add_tail(&server->nfs4_siblings, &clp->cl_superblocks);
1629 clnt = rpc_clone_client(clp->cl_rpcclient);
1631 server->nfs4_state = clp;
1632 up_write(&clp->cl_sem);
1636 printk(KERN_WARNING "NFS: cannot create RPC client.\n");
1637 return PTR_ERR(clnt);
1640 clnt->cl_intr = (server->flags & NFS4_MOUNT_INTR) ? 1 : 0;
1641 clnt->cl_softrtry = (server->flags & NFS4_MOUNT_SOFT) ? 1 : 0;
1642 server->client = clnt;
1644 if (server->nfs4_state->cl_idmap == NULL) {
1645 printk(KERN_WARNING "NFS: failed to create idmapper.\n");
1649 if (clnt->cl_auth->au_flavor != authflavour) {
1650 if (rpcauth_create(authflavour, clnt) == NULL) {
1651 printk(KERN_WARNING "NFS: couldn't create credcache!\n");
1656 sb->s_op = &nfs4_sops;
1657 err = nfs_sb_init(sb, authflavour);
1662 nfs4_put_client(clp);
1666 static int nfs4_compare_super(struct super_block *sb, void *data)
1668 struct nfs_server *server = data;
1669 struct nfs_server *old = NFS_SB(sb);
1671 if (strcmp(server->hostname, old->hostname) != 0)
1673 if (strcmp(server->mnt_path, old->mnt_path) != 0)
1679 nfs_copy_user_string(char *dst, struct nfs_string *src, int maxlen)
1684 return ERR_PTR(-EINVAL);
1685 if (src->len < maxlen)
1688 p = dst = kmalloc(maxlen + 1, GFP_KERNEL);
1690 return ERR_PTR(-ENOMEM);
1692 if (copy_from_user(dst, src->data, maxlen)) {
1695 return ERR_PTR(-EFAULT);
1701 static struct super_block *nfs4_get_sb(struct file_system_type *fs_type,
1702 int flags, const char *dev_name, void *raw_data)
1705 struct nfs_server *server;
1706 struct super_block *s;
1707 struct nfs4_mount_data *data = raw_data;
1711 printk("nfs_read_super: missing data argument\n");
1712 return ERR_PTR(-EINVAL);
1715 server = kmalloc(sizeof(struct nfs_server), GFP_KERNEL);
1717 return ERR_PTR(-ENOMEM);
1718 memset(server, 0, sizeof(struct nfs_server));
1719 /* Zero out the NFS state stuff */
1720 init_nfsv4_state(server);
1722 if (data->version != NFS4_MOUNT_VERSION) {
1723 printk("nfs warning: mount version %s than kernel\n",
1724 data->version < NFS4_MOUNT_VERSION ? "older" : "newer");
1727 p = nfs_copy_user_string(NULL, &data->hostname, 256);
1730 server->hostname = p;
1732 p = nfs_copy_user_string(NULL, &data->mnt_path, 1024);
1735 server->mnt_path = p;
1737 p = nfs_copy_user_string(server->ip_addr, &data->client_addr,
1738 sizeof(server->ip_addr) - 1);
1742 /* We now require that the mount process passes the remote address */
1743 if (data->host_addrlen != sizeof(server->addr)) {
1744 s = ERR_PTR(-EINVAL);
1747 if (copy_from_user(&server->addr, data->host_addr, sizeof(server->addr))) {
1748 s = ERR_PTR(-EFAULT);
1751 if (server->addr.sin_family != AF_INET ||
1752 server->addr.sin_addr.s_addr == INADDR_ANY) {
1753 printk("NFS: mount program didn't pass remote IP address!\n");
1754 s = ERR_PTR(-EINVAL);
1758 s = sget(fs_type, nfs4_compare_super, nfs_set_super, server);
1760 if (IS_ERR(s) || s->s_root)
1765 /* Fire up rpciod if not yet running */
1766 if (rpciod_up() != 0) {
1767 printk(KERN_WARNING "NFS: couldn't start rpciod!\n");
1772 error = nfs4_fill_super(s, data, flags & MS_VERBOSE ? 1 : 0);
1774 up_write(&s->s_umount);
1775 deactivate_super(s);
1776 return ERR_PTR(error);
1778 s->s_flags |= MS_ACTIVE;
1781 s = (struct super_block *)p;
1783 if (server->mnt_path)
1784 kfree(server->mnt_path);
1785 if (server->hostname)
1786 kfree(server->hostname);
1791 static void nfs4_kill_super(struct super_block *sb)
1793 nfs_return_all_delegations(sb);
1797 static struct file_system_type nfs4_fs_type = {
1798 .owner = THIS_MODULE,
1800 .get_sb = nfs4_get_sb,
1801 .kill_sb = nfs4_kill_super,
1802 .fs_flags = FS_ODD_RENAME|FS_REVAL_DOT|FS_BINARY_MOUNTDATA,
1805 #define nfs4_init_once(nfsi) \
1807 INIT_LIST_HEAD(&(nfsi)->open_states); \
1808 nfsi->delegation = NULL; \
1809 nfsi->delegation_state = 0; \
1810 init_rwsem(&nfsi->rwsem); \
1812 #define register_nfs4fs() register_filesystem(&nfs4_fs_type)
1813 #define unregister_nfs4fs() unregister_filesystem(&nfs4_fs_type)
1815 #define nfs4_init_once(nfsi) \
1817 #define register_nfs4fs() (0)
1818 #define unregister_nfs4fs()
1821 extern int nfs_init_nfspagecache(void);
1822 extern void nfs_destroy_nfspagecache(void);
1823 extern int nfs_init_readpagecache(void);
1824 extern int nfs_destroy_readpagecache(void);
1825 extern int nfs_init_writepagecache(void);
1826 extern int nfs_destroy_writepagecache(void);
1828 static kmem_cache_t * nfs_inode_cachep;
1830 static struct inode *nfs_alloc_inode(struct super_block *sb)
1832 struct nfs_inode *nfsi;
1833 nfsi = (struct nfs_inode *)kmem_cache_alloc(nfs_inode_cachep, SLAB_KERNEL);
1837 return &nfsi->vfs_inode;
1840 static void nfs_destroy_inode(struct inode *inode)
1842 kmem_cache_free(nfs_inode_cachep, NFS_I(inode));
1845 static void init_once(void * foo, kmem_cache_t * cachep, unsigned long flags)
1847 struct nfs_inode *nfsi = (struct nfs_inode *) foo;
1849 if ((flags & (SLAB_CTOR_VERIFY|SLAB_CTOR_CONSTRUCTOR)) ==
1850 SLAB_CTOR_CONSTRUCTOR) {
1851 inode_init_once(&nfsi->vfs_inode);
1852 spin_lock_init(&nfsi->req_lock);
1853 INIT_LIST_HEAD(&nfsi->dirty);
1854 INIT_LIST_HEAD(&nfsi->commit);
1855 INIT_LIST_HEAD(&nfsi->open_files);
1856 INIT_RADIX_TREE(&nfsi->nfs_page_tree, GFP_ATOMIC);
1857 atomic_set(&nfsi->data_updates, 0);
1861 init_waitqueue_head(&nfsi->nfs_i_wait);
1862 nfs4_init_once(nfsi);
1866 int nfs_init_inodecache(void)
1868 nfs_inode_cachep = kmem_cache_create("nfs_inode_cache",
1869 sizeof(struct nfs_inode),
1870 0, SLAB_RECLAIM_ACCOUNT,
1872 if (nfs_inode_cachep == NULL)
1878 void nfs_destroy_inodecache(void)
1880 if (kmem_cache_destroy(nfs_inode_cachep))
1881 printk(KERN_INFO "nfs_inode_cache: not all structures were freed\n");
1887 static int __init init_nfs_fs(void)
1891 err = nfs_init_nfspagecache();
1895 err = nfs_init_inodecache();
1899 err = nfs_init_readpagecache();
1903 err = nfs_init_writepagecache();
1907 #ifdef CONFIG_PROC_FS
1908 rpc_proc_register(&nfs_rpcstat);
1910 err = register_filesystem(&nfs_fs_type);
1913 if ((err = register_nfs4fs()) != 0)
1917 rpc_proc_unregister("nfs");
1918 nfs_destroy_writepagecache();
1920 nfs_destroy_readpagecache();
1922 nfs_destroy_inodecache();
1924 nfs_destroy_nfspagecache();
1929 static void __exit exit_nfs_fs(void)
1931 nfs_destroy_writepagecache();
1932 nfs_destroy_readpagecache();
1933 nfs_destroy_inodecache();
1934 nfs_destroy_nfspagecache();
1935 #ifdef CONFIG_PROC_FS
1936 rpc_proc_unregister("nfs");
1938 unregister_filesystem(&nfs_fs_type);
1939 unregister_nfs4fs();
1942 /* Not quite true; I just maintain it */
1943 MODULE_AUTHOR("Olaf Kirch <okir@monad.swb.de>");
1944 MODULE_LICENSE("GPL");
1946 module_init(init_nfs_fs)
1947 module_exit(exit_nfs_fs)