4 * Copyright (C) 1992 Rick Sladkey
6 * nfs 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
14 * Split from inode.c by David Howells <dhowells@redhat.com>
18 #include <linux/config.h>
19 #include <linux/module.h>
20 #include <linux/init.h>
22 #include <linux/time.h>
23 #include <linux/kernel.h>
25 #include <linux/string.h>
26 #include <linux/stat.h>
27 #include <linux/errno.h>
28 #include <linux/unistd.h>
29 #include <linux/sunrpc/clnt.h>
30 #include <linux/sunrpc/stats.h>
31 #include <linux/sunrpc/metrics.h>
32 #include <linux/nfs_fs.h>
33 #include <linux/nfs_mount.h>
34 #include <linux/nfs4_mount.h>
35 #include <linux/lockd/bind.h>
36 #include <linux/smp_lock.h>
37 #include <linux/seq_file.h>
38 #include <linux/mount.h>
39 #include <linux/nfs_idmap.h>
40 #include <linux/vfs.h>
41 #include <linux/inet.h>
42 #include <linux/nfs_xdr.h>
44 #include <asm/system.h>
45 #include <asm/uaccess.h>
49 #include "delegation.h"
53 #define NFSDBG_FACILITY NFSDBG_VFS
55 /* Maximum number of readahead requests
56 * FIXME: this should really be a sysctl so that users may tune it to suit
57 * their needs. People that do NFS over a slow network, might for
58 * instance want to reduce it to something closer to 1 for improved
59 * interactive response.
61 #define NFS_MAX_READAHEAD (RPC_DEF_SLOT_TABLE - 1)
66 static struct rpc_version * nfs_version[] = {
70 #if defined(CONFIG_NFS_V3)
72 #elif defined(CONFIG_NFS_V4)
75 #if defined(CONFIG_NFS_V4)
80 static struct rpc_program nfs_program = {
82 .number = NFS_PROGRAM,
83 .nrvers = ARRAY_SIZE(nfs_version),
84 .version = nfs_version,
85 .stats = &nfs_rpcstat,
86 .pipe_dir_name = "/nfs",
89 struct rpc_stat nfs_rpcstat = {
90 .program = &nfs_program
94 #ifdef CONFIG_NFS_V3_ACL
95 static struct rpc_stat nfsacl_rpcstat = { &nfsacl_program };
96 static struct rpc_version * nfsacl_version[] = {
97 [3] = &nfsacl_version3,
100 struct rpc_program nfsacl_program = {
102 .number = NFS_ACL_PROGRAM,
103 .nrvers = ARRAY_SIZE(nfsacl_version),
104 .version = nfsacl_version,
105 .stats = &nfsacl_rpcstat,
107 #endif /* CONFIG_NFS_V3_ACL */
109 static void nfs_umount_begin(struct vfsmount *, int);
110 static int nfs_statfs(struct dentry *, struct kstatfs *);
111 static int nfs_show_options(struct seq_file *, struct vfsmount *);
112 static int nfs_show_stats(struct seq_file *, struct vfsmount *);
113 static int nfs_get_sb(struct file_system_type *, int, const char *, void *, struct vfsmount *);
114 static int nfs_clone_nfs_sb(struct file_system_type *fs_type,
115 int flags, const char *dev_name, void *raw_data, struct vfsmount *mnt);
116 static void nfs_kill_super(struct super_block *);
118 static struct file_system_type nfs_fs_type = {
119 .owner = THIS_MODULE,
121 .get_sb = nfs_get_sb,
122 .kill_sb = nfs_kill_super,
123 .fs_flags = FS_ODD_RENAME|FS_REVAL_DOT|FS_BINARY_MOUNTDATA,
126 struct file_system_type clone_nfs_fs_type = {
127 .owner = THIS_MODULE,
129 .get_sb = nfs_clone_nfs_sb,
130 .kill_sb = nfs_kill_super,
131 .fs_flags = FS_ODD_RENAME|FS_REVAL_DOT|FS_BINARY_MOUNTDATA,
134 static struct super_operations nfs_sops = {
135 .alloc_inode = nfs_alloc_inode,
136 .destroy_inode = nfs_destroy_inode,
137 .write_inode = nfs_write_inode,
138 .statfs = nfs_statfs,
139 .clear_inode = nfs_clear_inode,
140 .umount_begin = nfs_umount_begin,
141 .show_options = nfs_show_options,
142 .show_stats = nfs_show_stats,
146 static int nfs4_get_sb(struct file_system_type *fs_type,
147 int flags, const char *dev_name, void *raw_data, struct vfsmount *mnt);
148 static int nfs_clone_nfs4_sb(struct file_system_type *fs_type,
149 int flags, const char *dev_name, void *raw_data, struct vfsmount *mnt);
150 static int nfs_referral_nfs4_sb(struct file_system_type *fs_type,
151 int flags, const char *dev_name, void *raw_data, struct vfsmount *mnt);
152 static void nfs4_kill_super(struct super_block *sb);
154 static struct file_system_type nfs4_fs_type = {
155 .owner = THIS_MODULE,
157 .get_sb = nfs4_get_sb,
158 .kill_sb = nfs4_kill_super,
159 .fs_flags = FS_ODD_RENAME|FS_REVAL_DOT|FS_BINARY_MOUNTDATA,
162 struct file_system_type clone_nfs4_fs_type = {
163 .owner = THIS_MODULE,
165 .get_sb = nfs_clone_nfs4_sb,
166 .kill_sb = nfs4_kill_super,
167 .fs_flags = FS_ODD_RENAME|FS_REVAL_DOT|FS_BINARY_MOUNTDATA,
170 struct file_system_type nfs_referral_nfs4_fs_type = {
171 .owner = THIS_MODULE,
173 .get_sb = nfs_referral_nfs4_sb,
174 .kill_sb = nfs4_kill_super,
175 .fs_flags = FS_ODD_RENAME|FS_REVAL_DOT|FS_BINARY_MOUNTDATA,
178 static struct super_operations nfs4_sops = {
179 .alloc_inode = nfs_alloc_inode,
180 .destroy_inode = nfs_destroy_inode,
181 .write_inode = nfs_write_inode,
182 .statfs = nfs_statfs,
183 .clear_inode = nfs4_clear_inode,
184 .umount_begin = nfs_umount_begin,
185 .show_options = nfs_show_options,
186 .show_stats = nfs_show_stats,
190 static struct shrinker *acl_shrinker;
193 * Register the NFS filesystems
195 int __init register_nfs_fs(void)
199 ret = register_filesystem(&nfs_fs_type);
204 ret = nfs_register_sysctl();
207 ret = register_filesystem(&nfs4_fs_type);
211 acl_shrinker = set_shrinker(DEFAULT_SEEKS, nfs_access_cache_shrinker);
216 nfs_unregister_sysctl();
218 unregister_filesystem(&nfs_fs_type);
225 * Unregister the NFS filesystems
227 void __exit unregister_nfs_fs(void)
229 if (acl_shrinker != NULL)
230 remove_shrinker(acl_shrinker);
232 unregister_filesystem(&nfs4_fs_type);
233 nfs_unregister_sysctl();
235 unregister_filesystem(&nfs_fs_type);
239 * Deliver file system statistics to userspace
241 static int nfs_statfs(struct dentry *dentry, struct kstatfs *buf)
243 struct nfs_server *server = NFS_SB(dentry->d_sb);
244 unsigned char blockbits;
245 unsigned long blockres;
246 struct nfs_fh *fh = NFS_FH(dentry->d_inode);
247 struct nfs_fattr fattr;
248 struct nfs_fsstat res = {
255 error = server->nfs_client->rpc_ops->statfs(server, fh, &res);
256 buf->f_type = NFS_SUPER_MAGIC;
261 * Current versions of glibc do not correctly handle the
262 * case where f_frsize != f_bsize. Eventually we want to
263 * report the value of wtmult in this field.
265 buf->f_frsize = dentry->d_sb->s_blocksize;
268 * On most *nix systems, f_blocks, f_bfree, and f_bavail
269 * are reported in units of f_frsize. Linux hasn't had
270 * an f_frsize field in its statfs struct until recently,
271 * thus historically Linux's sys_statfs reports these
272 * fields in units of f_bsize.
274 buf->f_bsize = dentry->d_sb->s_blocksize;
275 blockbits = dentry->d_sb->s_blocksize_bits;
276 blockres = (1 << blockbits) - 1;
277 buf->f_blocks = (res.tbytes + blockres) >> blockbits;
278 buf->f_bfree = (res.fbytes + blockres) >> blockbits;
279 buf->f_bavail = (res.abytes + blockres) >> blockbits;
281 buf->f_files = res.tfiles;
282 buf->f_ffree = res.afiles;
284 buf->f_namelen = server->namelen;
290 dprintk("%s: statfs error = %d\n", __FUNCTION__, -error);
291 buf->f_bsize = buf->f_blocks = buf->f_bfree = buf->f_bavail = -1;
297 * Map the security flavour number to a name
299 static const char *nfs_pseudoflavour_to_name(rpc_authflavor_t flavour)
301 static const struct {
302 rpc_authflavor_t flavour;
305 { RPC_AUTH_NULL, "null" },
306 { RPC_AUTH_UNIX, "sys" },
307 { RPC_AUTH_GSS_KRB5, "krb5" },
308 { RPC_AUTH_GSS_KRB5I, "krb5i" },
309 { RPC_AUTH_GSS_KRB5P, "krb5p" },
310 { RPC_AUTH_GSS_LKEY, "lkey" },
311 { RPC_AUTH_GSS_LKEYI, "lkeyi" },
312 { RPC_AUTH_GSS_LKEYP, "lkeyp" },
313 { RPC_AUTH_GSS_SPKM, "spkm" },
314 { RPC_AUTH_GSS_SPKMI, "spkmi" },
315 { RPC_AUTH_GSS_SPKMP, "spkmp" },
320 for (i=0; sec_flavours[i].flavour != -1; i++) {
321 if (sec_flavours[i].flavour == flavour)
324 return sec_flavours[i].str;
328 * Describe the mount options in force on this server representation
330 static void nfs_show_mount_options(struct seq_file *m, struct nfs_server *nfss, int showdefaults)
332 static const struct proc_nfs_info {
337 { NFS_MOUNT_SOFT, ",soft", ",hard" },
338 { NFS_MOUNT_INTR, ",intr", "" },
339 { NFS_MOUNT_NOCTO, ",nocto", "" },
340 { NFS_MOUNT_NOAC, ",noac", "" },
341 { NFS_MOUNT_NONLM, ",nolock", "" },
342 { NFS_MOUNT_NOACL, ",noacl", "" },
345 const struct proc_nfs_info *nfs_infop;
346 struct nfs_client *clp = nfss->nfs_client;
350 seq_printf(m, ",vers=%d", clp->rpc_ops->version);
351 seq_printf(m, ",rsize=%d", nfss->rsize);
352 seq_printf(m, ",wsize=%d", nfss->wsize);
353 if (nfss->acregmin != 3*HZ || showdefaults)
354 seq_printf(m, ",acregmin=%d", nfss->acregmin/HZ);
355 if (nfss->acregmax != 60*HZ || showdefaults)
356 seq_printf(m, ",acregmax=%d", nfss->acregmax/HZ);
357 if (nfss->acdirmin != 30*HZ || showdefaults)
358 seq_printf(m, ",acdirmin=%d", nfss->acdirmin/HZ);
359 if (nfss->acdirmax != 60*HZ || showdefaults)
360 seq_printf(m, ",acdirmax=%d", nfss->acdirmax/HZ);
361 for (nfs_infop = nfs_info; nfs_infop->flag; nfs_infop++) {
362 if (nfss->flags & nfs_infop->flag)
363 seq_puts(m, nfs_infop->str);
365 seq_puts(m, nfs_infop->nostr);
367 switch (nfss->client->cl_xprt->prot) {
375 snprintf(buf, sizeof(buf), "%u", nfss->client->cl_xprt->prot);
378 seq_printf(m, ",proto=%s", proto);
379 seq_printf(m, ",timeo=%lu", 10U * nfss->retrans_timeo / HZ);
380 seq_printf(m, ",retrans=%u", nfss->retrans_count);
381 seq_printf(m, ",sec=%s", nfs_pseudoflavour_to_name(nfss->client->cl_auth->au_flavor));
385 * Describe the mount options on this VFS mountpoint
387 static int nfs_show_options(struct seq_file *m, struct vfsmount *mnt)
389 struct nfs_server *nfss = NFS_SB(mnt->mnt_sb);
391 nfs_show_mount_options(m, nfss, 0);
393 seq_puts(m, ",addr=");
394 seq_escape(m, nfss->hostname, " \t\n\\");
400 * Present statistical information for this VFS mountpoint
402 static int nfs_show_stats(struct seq_file *m, struct vfsmount *mnt)
405 struct nfs_server *nfss = NFS_SB(mnt->mnt_sb);
406 struct rpc_auth *auth = nfss->client->cl_auth;
407 struct nfs_iostats totals = { };
409 seq_printf(m, "statvers=%s", NFS_IOSTAT_VERS);
412 * Display all mount option settings
414 seq_printf(m, "\n\topts:\t");
415 seq_puts(m, mnt->mnt_sb->s_flags & MS_RDONLY ? "ro" : "rw");
416 seq_puts(m, mnt->mnt_sb->s_flags & MS_SYNCHRONOUS ? ",sync" : "");
417 seq_puts(m, mnt->mnt_sb->s_flags & MS_NOATIME ? ",noatime" : "");
418 seq_puts(m, mnt->mnt_sb->s_flags & MS_NODIRATIME ? ",nodiratime" : "");
419 nfs_show_mount_options(m, nfss, 1);
421 seq_printf(m, "\n\tage:\t%lu", (jiffies - nfss->mount_time) / HZ);
423 seq_printf(m, "\n\tcaps:\t");
424 seq_printf(m, "caps=0x%x", nfss->caps);
425 seq_printf(m, ",wtmult=%d", nfss->wtmult);
426 seq_printf(m, ",dtsize=%d", nfss->dtsize);
427 seq_printf(m, ",bsize=%d", nfss->bsize);
428 seq_printf(m, ",namelen=%d", nfss->namelen);
431 if (nfss->nfs_client->cl_nfsversion == 4) {
432 seq_printf(m, "\n\tnfsv4:\t");
433 seq_printf(m, "bm0=0x%x", nfss->attr_bitmask[0]);
434 seq_printf(m, ",bm1=0x%x", nfss->attr_bitmask[1]);
435 seq_printf(m, ",acl=0x%x", nfss->acl_bitmask);
440 * Display security flavor in effect for this mount
442 seq_printf(m, "\n\tsec:\tflavor=%d", auth->au_ops->au_flavor);
444 seq_printf(m, ",pseudoflavor=%d", auth->au_flavor);
447 * Display superblock I/O counters
449 for_each_possible_cpu(cpu) {
450 struct nfs_iostats *stats;
453 stats = per_cpu_ptr(nfss->io_stats, cpu);
455 for (i = 0; i < __NFSIOS_COUNTSMAX; i++)
456 totals.events[i] += stats->events[i];
457 for (i = 0; i < __NFSIOS_BYTESMAX; i++)
458 totals.bytes[i] += stats->bytes[i];
463 seq_printf(m, "\n\tevents:\t");
464 for (i = 0; i < __NFSIOS_COUNTSMAX; i++)
465 seq_printf(m, "%lu ", totals.events[i]);
466 seq_printf(m, "\n\tbytes:\t");
467 for (i = 0; i < __NFSIOS_BYTESMAX; i++)
468 seq_printf(m, "%Lu ", totals.bytes[i]);
471 rpc_print_iostats(m, nfss->client);
477 * Begin unmount by attempting to remove all automounted mountpoints we added
478 * in response to traversals
480 static void nfs_umount_begin(struct vfsmount *vfsmnt, int flags)
482 struct nfs_server *server;
483 struct rpc_clnt *rpc;
485 shrink_submounts(vfsmnt, &nfs_automount_list);
486 if (!(flags & MNT_FORCE))
488 /* -EIO all pending I/O */
489 server = NFS_SB(vfsmnt->mnt_sb);
490 rpc = server->client;
492 rpc_killall_tasks(rpc);
493 rpc = server->client_acl;
495 rpc_killall_tasks(rpc);
499 * Obtain the root inode of the file system.
501 static struct inode *
502 nfs_get_root(struct super_block *sb, struct nfs_fh *rootfh, struct nfs_fsinfo *fsinfo)
504 struct nfs_server *server = NFS_SB(sb);
507 error = server->nfs_client->rpc_ops->getroot(server, rootfh, fsinfo);
509 dprintk("nfs_get_root: getattr error = %d\n", -error);
510 return ERR_PTR(error);
513 server->fsid = fsinfo->fattr->fsid;
514 return nfs_fhget(sb, rootfh, fsinfo->fattr);
518 * Do NFS version-independent mount processing, and sanity checking
521 nfs_sb_init(struct super_block *sb, rpc_authflavor_t authflavor)
523 struct nfs_server *server;
524 struct inode *root_inode;
525 struct nfs_fattr fattr;
526 struct nfs_fsinfo fsinfo = {
529 struct nfs_pathconf pathinfo = {
532 int no_root_error = 0;
533 unsigned long max_rpc_payload;
535 /* We probably want something more informative here */
536 snprintf(sb->s_id, sizeof(sb->s_id), "%x:%x", MAJOR(sb->s_dev), MINOR(sb->s_dev));
540 sb->s_magic = NFS_SUPER_MAGIC;
542 server->io_stats = nfs_alloc_iostats();
543 if (server->io_stats == NULL)
546 root_inode = nfs_get_root(sb, &server->fh, &fsinfo);
547 /* Did getting the root inode fail? */
548 if (IS_ERR(root_inode)) {
549 no_root_error = PTR_ERR(root_inode);
552 sb->s_root = d_alloc_root(root_inode);
554 no_root_error = -ENOMEM;
557 sb->s_root->d_op = server->nfs_client->rpc_ops->dentry_ops;
559 /* mount time stamp, in seconds */
560 server->mount_time = jiffies;
562 /* Get some general file system info */
563 if (server->namelen == 0 &&
564 server->nfs_client->rpc_ops->pathconf(server, &server->fh, &pathinfo) >= 0)
565 server->namelen = pathinfo.max_namelen;
566 /* Work out a lot of parameters */
567 if (server->rsize == 0)
568 server->rsize = nfs_block_size(fsinfo.rtpref, NULL);
569 if (server->wsize == 0)
570 server->wsize = nfs_block_size(fsinfo.wtpref, NULL);
572 if (fsinfo.rtmax >= 512 && server->rsize > fsinfo.rtmax)
573 server->rsize = nfs_block_size(fsinfo.rtmax, NULL);
574 if (fsinfo.wtmax >= 512 && server->wsize > fsinfo.wtmax)
575 server->wsize = nfs_block_size(fsinfo.wtmax, NULL);
577 max_rpc_payload = nfs_block_size(rpc_max_payload(server->client), NULL);
578 if (server->rsize > max_rpc_payload)
579 server->rsize = max_rpc_payload;
580 if (server->rsize > NFS_MAX_FILE_IO_SIZE)
581 server->rsize = NFS_MAX_FILE_IO_SIZE;
582 server->rpages = (server->rsize + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT;
584 if (server->wsize > max_rpc_payload)
585 server->wsize = max_rpc_payload;
586 if (server->wsize > NFS_MAX_FILE_IO_SIZE)
587 server->wsize = NFS_MAX_FILE_IO_SIZE;
588 server->wpages = (server->wsize + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT;
590 if (sb->s_blocksize == 0)
591 sb->s_blocksize = nfs_block_bits(server->wsize,
592 &sb->s_blocksize_bits);
593 server->wtmult = nfs_block_bits(fsinfo.wtmult, NULL);
595 server->dtsize = nfs_block_size(fsinfo.dtpref, NULL);
596 if (server->dtsize > PAGE_CACHE_SIZE)
597 server->dtsize = PAGE_CACHE_SIZE;
598 if (server->dtsize > server->rsize)
599 server->dtsize = server->rsize;
601 if (server->flags & NFS_MOUNT_NOAC) {
602 server->acregmin = server->acregmax = 0;
603 server->acdirmin = server->acdirmax = 0;
604 sb->s_flags |= MS_SYNCHRONOUS;
606 server->backing_dev_info.ra_pages = server->rpages * NFS_MAX_READAHEAD;
608 nfs_super_set_maxbytes(sb, fsinfo.maxfilesize);
610 server->client->cl_intr = (server->flags & NFS_MOUNT_INTR) ? 1 : 0;
611 server->client->cl_softrtry = (server->flags & NFS_MOUNT_SOFT) ? 1 : 0;
613 /* We're airborne Set socket buffersize */
614 rpc_setbufsize(server->client, server->wsize + 100, server->rsize + 100);
616 /* Yargs. It didn't work out. */
618 dprintk("nfs_sb_init: get root inode failed: errno %d\n", -no_root_error);
619 if (!IS_ERR(root_inode))
621 return no_root_error;
625 * Initialise the timeout values for a connection
627 static void nfs_init_timeout_values(struct rpc_timeout *to, int proto, unsigned int timeo, unsigned int retrans)
629 to->to_initval = timeo * HZ / 10;
630 to->to_retries = retrans;
637 to->to_initval = 60 * HZ;
638 if (to->to_initval > NFS_MAX_TCP_TIMEOUT)
639 to->to_initval = NFS_MAX_TCP_TIMEOUT;
640 to->to_increment = to->to_initval;
641 to->to_maxval = to->to_initval + (to->to_increment * to->to_retries);
642 to->to_exponential = 0;
647 to->to_initval = 11 * HZ / 10;
648 if (to->to_initval > NFS_MAX_UDP_TIMEOUT)
649 to->to_initval = NFS_MAX_UDP_TIMEOUT;
650 to->to_maxval = NFS_MAX_UDP_TIMEOUT;
651 to->to_exponential = 1;
657 * Create an RPC client handle.
659 static struct rpc_clnt *
660 nfs_create_client(struct nfs_server *server, const struct nfs_mount_data *data)
662 struct nfs_client *clp;
663 struct rpc_timeout timeparms;
664 struct rpc_xprt *xprt = NULL;
665 struct rpc_clnt *clnt = NULL;
666 int proto = (data->flags & NFS_MOUNT_TCP) ? IPPROTO_TCP : IPPROTO_UDP;
670 if (server->flags & NFS_MOUNT_VER3)
674 clp = nfs_get_client(server->hostname, &server->addr, nfsversion);
676 dprintk("%s: failed to create NFS4 client.\n", __FUNCTION__);
677 return ERR_PTR(PTR_ERR(clp));
680 nfs_init_timeout_values(&timeparms, proto, data->timeo, data->retrans);
682 server->retrans_timeo = timeparms.to_initval;
683 server->retrans_count = timeparms.to_retries;
685 /* Check NFS protocol revision and initialize RPC op vector
686 * and file handle pool. */
688 if (nfsversion == 3) {
689 clp->rpc_ops = &nfs_v3_clientops;
690 server->caps |= NFS_CAP_READDIRPLUS;
692 clp->rpc_ops = &nfs_v2_clientops;
695 clp->rpc_ops = &nfs_v2_clientops;
698 /* create transport and client */
699 xprt = xprt_create_proto(proto, &server->addr, &timeparms);
701 dprintk("%s: cannot create RPC transport. Error = %ld\n",
702 __FUNCTION__, PTR_ERR(xprt));
703 nfs_mark_client_ready(clp, PTR_ERR(xprt));
705 return (struct rpc_clnt *)xprt;
707 clnt = rpc_create_client(xprt, server->hostname, &nfs_program,
708 clp->cl_nfsversion, data->pseudoflavor);
710 dprintk("%s: cannot create RPC client. Error = %ld\n",
711 __FUNCTION__, PTR_ERR(xprt));
716 clnt->cl_softrtry = 1;
718 nfs_mark_client_ready(clp, 0);
719 server->nfs_client = clp;
723 nfs_mark_client_ready(clp, PTR_ERR(xprt));
729 * Clone a server record
731 static struct nfs_server *nfs_clone_server(struct super_block *sb, struct nfs_clone_mount *data)
733 struct nfs_server *server = NFS_SB(sb);
734 struct nfs_server *parent = NFS_SB(data->sb);
735 struct inode *root_inode;
736 struct nfs_fsinfo fsinfo;
737 void *err = ERR_PTR(-ENOMEM);
739 sb->s_op = data->sb->s_op;
740 sb->s_blocksize = data->sb->s_blocksize;
741 sb->s_blocksize_bits = data->sb->s_blocksize_bits;
742 sb->s_maxbytes = data->sb->s_maxbytes;
744 server->client_sys = server->client_acl = ERR_PTR(-EINVAL);
745 server->io_stats = nfs_alloc_iostats();
746 if (server->io_stats == NULL)
749 server->client = rpc_clone_client(parent->client);
750 if (IS_ERR((err = server->client)))
753 if (!IS_ERR(parent->client_sys)) {
754 server->client_sys = rpc_clone_client(parent->client_sys);
755 if (IS_ERR((err = server->client_sys)))
758 if (!IS_ERR(parent->client_acl)) {
759 server->client_acl = rpc_clone_client(parent->client_acl);
760 if (IS_ERR((err = server->client_acl)))
763 root_inode = nfs_fhget(sb, data->fh, data->fattr);
766 sb->s_root = d_alloc_root(root_inode);
769 fsinfo.fattr = data->fattr;
770 if (NFS_PROTO(root_inode)->fsinfo(server, data->fh, &fsinfo) == 0)
771 nfs_super_set_maxbytes(sb, fsinfo.maxfilesize);
772 sb->s_root->d_op = server->nfs_client->rpc_ops->dentry_ops;
773 sb->s_flags |= MS_ACTIVE;
782 * Copy an existing superblock and attach revised data
784 static int nfs_clone_generic_sb(struct nfs_clone_mount *data,
785 struct super_block *(*fill_sb)(struct nfs_server *, struct nfs_clone_mount *),
786 struct nfs_server *(*fill_server)(struct super_block *, struct nfs_clone_mount *),
787 struct vfsmount *mnt)
789 struct nfs_server *server;
790 struct nfs_server *parent = NFS_SB(data->sb);
791 struct super_block *sb = ERR_PTR(-EINVAL);
796 server = kmalloc(sizeof(struct nfs_server), GFP_KERNEL);
799 memcpy(server, parent, sizeof(*server));
800 atomic_inc(&server->nfs_client->cl_count);
801 hostname = (data->hostname != NULL) ? data->hostname : parent->hostname;
802 len = strlen(hostname) + 1;
803 server->hostname = kmalloc(len, GFP_KERNEL);
804 if (server->hostname == NULL)
806 memcpy(server->hostname, hostname, len);
811 sb = fill_sb(server, data);
818 goto out_rpciod_down;
820 server = fill_server(sb, data);
821 if (IS_ERR(server)) {
822 error = PTR_ERR(server);
825 return simple_set_mnt(mnt, sb);
827 up_write(&sb->s_umount);
828 deactivate_super(sb);
832 kfree(server->hostname);
833 nfs_put_client(server->nfs_client);
835 return simple_set_mnt(mnt, sb);
839 kfree(server->hostname);
841 nfs_put_client(server->nfs_client);
848 * Set up an NFS2/3 superblock
850 * The way this works is that the mount process passes a structure
851 * in the data argument which contains the server's IP address
852 * and the root file handle obtained from the server's mount
853 * daemon. We stash these away in the private superblock fields.
856 nfs_fill_super(struct super_block *sb, struct nfs_mount_data *data, int silent)
858 struct nfs_server *server;
859 rpc_authflavor_t authflavor;
862 sb->s_blocksize_bits = 0;
865 sb->s_blocksize = nfs_block_size(data->bsize, &sb->s_blocksize_bits);
867 server->rsize = nfs_block_size(data->rsize, NULL);
869 server->wsize = nfs_block_size(data->wsize, NULL);
870 server->flags = data->flags & NFS_MOUNT_FLAGMASK;
872 server->acregmin = data->acregmin*HZ;
873 server->acregmax = data->acregmax*HZ;
874 server->acdirmin = data->acdirmin*HZ;
875 server->acdirmax = data->acdirmax*HZ;
877 /* Start lockd here, before we might error out */
878 if (!(server->flags & NFS_MOUNT_NONLM))
881 server->namelen = data->namlen;
882 server->hostname = kmalloc(strlen(data->hostname) + 1, GFP_KERNEL);
883 if (!server->hostname)
885 strcpy(server->hostname, data->hostname);
887 /* Fill in pseudoflavor for mount version < 5 */
888 if (!(data->flags & NFS_MOUNT_SECFLAVOUR))
889 data->pseudoflavor = RPC_AUTH_UNIX;
890 authflavor = data->pseudoflavor; /* save for sb_init() */
891 /* XXX maybe we want to add a server->pseudoflavor field */
893 /* Create RPC client handles */
894 server->client = nfs_create_client(server, data);
895 if (IS_ERR(server->client))
896 return PTR_ERR(server->client);
898 /* RFC 2623, sec 2.3.2 */
899 if (authflavor != RPC_AUTH_UNIX) {
900 struct rpc_auth *auth;
902 server->client_sys = rpc_clone_client(server->client);
903 if (IS_ERR(server->client_sys))
904 return PTR_ERR(server->client_sys);
905 auth = rpcauth_create(RPC_AUTH_UNIX, server->client_sys);
907 return PTR_ERR(auth);
909 atomic_inc(&server->client->cl_count);
910 server->client_sys = server->client;
912 if (server->flags & NFS_MOUNT_VER3) {
913 #ifdef CONFIG_NFS_V3_ACL
914 if (!(server->flags & NFS_MOUNT_NOACL)) {
915 server->client_acl = rpc_bind_new_program(server->client, &nfsacl_program, 3);
916 /* No errors! Assume that Sun nfsacls are supported */
917 if (!IS_ERR(server->client_acl))
918 server->caps |= NFS_CAP_ACLS;
921 server->flags &= ~NFS_MOUNT_NOACL;
922 #endif /* CONFIG_NFS_V3_ACL */
924 * The VFS shouldn't apply the umask to mode bits. We will
925 * do so ourselves when necessary.
927 sb->s_flags |= MS_POSIXACL;
928 if (server->namelen == 0 || server->namelen > NFS3_MAXNAMLEN)
929 server->namelen = NFS3_MAXNAMLEN;
932 if (server->namelen == 0 || server->namelen > NFS2_MAXNAMLEN)
933 server->namelen = NFS2_MAXNAMLEN;
936 sb->s_op = &nfs_sops;
937 return nfs_sb_init(sb, authflavor);
940 static int nfs_set_super(struct super_block *s, void *data)
943 return set_anon_super(s, data);
946 static int nfs_compare_super(struct super_block *sb, void *data)
948 struct nfs_server *server = data;
949 struct nfs_server *old = NFS_SB(sb);
951 if (old->addr.sin_addr.s_addr != server->addr.sin_addr.s_addr)
953 if (old->addr.sin_port != server->addr.sin_port)
955 return !nfs_compare_fh(&old->fh, &server->fh);
958 static int nfs_get_sb(struct file_system_type *fs_type,
959 int flags, const char *dev_name, void *raw_data, struct vfsmount *mnt)
962 struct nfs_server *server = NULL;
963 struct super_block *s;
965 struct nfs_mount_data *data = raw_data;
969 dprintk("%s: missing data argument\n", __FUNCTION__);
970 goto out_err_noserver;
972 if (data->version <= 0 || data->version > NFS_MOUNT_VERSION) {
973 dprintk("%s: bad mount version\n", __FUNCTION__);
974 goto out_err_noserver;
976 switch (data->version) {
982 if (data->flags & NFS_MOUNT_VER3) {
983 dprintk("%s: mount structure version %d does not support NFSv3\n",
986 goto out_err_noserver;
988 data->root.size = NFS2_FHSIZE;
989 memcpy(data->root.data, data->old_root.data, NFS2_FHSIZE);
991 if (data->flags & NFS_MOUNT_SECFLAVOUR) {
992 dprintk("%s: mount structure version %d does not support strong security\n",
995 goto out_err_noserver;
998 memset(data->context, 0, sizeof(data->context));
1000 #ifndef CONFIG_NFS_V3
1001 /* If NFSv3 is not compiled in, return -EPROTONOSUPPORT */
1002 error = -EPROTONOSUPPORT;
1003 if (data->flags & NFS_MOUNT_VER3) {
1004 dprintk("%s: NFSv3 not compiled into kernel\n", __FUNCTION__);
1005 goto out_err_noserver;
1007 #endif /* CONFIG_NFS_V3 */
1010 server = kzalloc(sizeof(struct nfs_server), GFP_KERNEL);
1012 goto out_err_noserver;
1013 /* Zero out the NFS state stuff */
1014 init_nfsv4_state(server);
1015 server->client = server->client_sys = server->client_acl = ERR_PTR(-EINVAL);
1018 if (data->flags & NFS_MOUNT_VER3)
1019 root->size = data->root.size;
1021 root->size = NFS2_FHSIZE;
1023 if (root->size > sizeof(root->data)) {
1024 dprintk("%s: invalid root filehandle\n", __FUNCTION__);
1027 memcpy(root->data, data->root.data, root->size);
1029 /* We now require that the mount process passes the remote address */
1030 memcpy(&server->addr, &data->addr, sizeof(server->addr));
1031 if (server->addr.sin_addr.s_addr == INADDR_ANY) {
1032 dprintk("%s: mount program didn't pass remote address!\n",
1037 /* Fire up rpciod if not yet running */
1038 error = rpciod_up();
1040 dprintk("%s: couldn't start rpciod! Error = %d\n",
1041 __FUNCTION__, error);
1045 s = sget(fs_type, nfs_compare_super, nfs_set_super, server);
1048 goto out_err_rpciod;
1052 goto out_rpciod_down;
1056 error = nfs_fill_super(s, data, flags & MS_SILENT ? 1 : 0);
1058 up_write(&s->s_umount);
1059 deactivate_super(s);
1062 s->s_flags |= MS_ACTIVE;
1063 return simple_set_mnt(mnt, s);
1068 return simple_set_mnt(mnt, s);
1078 static void nfs_kill_super(struct super_block *s)
1080 struct nfs_server *server = NFS_SB(s);
1084 if (!IS_ERR(server->client))
1085 rpc_shutdown_client(server->client);
1086 if (!IS_ERR(server->client_sys))
1087 rpc_shutdown_client(server->client_sys);
1088 if (!IS_ERR(server->client_acl))
1089 rpc_shutdown_client(server->client_acl);
1091 if (!(server->flags & NFS_MOUNT_NONLM))
1092 lockd_down(); /* release rpc.lockd */
1094 rpciod_down(); /* release rpciod */
1096 nfs_free_iostats(server->io_stats);
1097 kfree(server->hostname);
1098 nfs_put_client(server->nfs_client);
1100 nfs_release_automount_timer();
1103 static struct super_block *nfs_clone_sb(struct nfs_server *server, struct nfs_clone_mount *data)
1105 struct super_block *sb;
1107 server->fsid = data->fattr->fsid;
1108 nfs_copy_fh(&server->fh, data->fh);
1109 sb = sget(&nfs_fs_type, nfs_compare_super, nfs_set_super, server);
1110 if (!IS_ERR(sb) && sb->s_root == NULL && !(server->flags & NFS_MOUNT_NONLM))
1115 static int nfs_clone_nfs_sb(struct file_system_type *fs_type,
1116 int flags, const char *dev_name, void *raw_data, struct vfsmount *mnt)
1118 struct nfs_clone_mount *data = raw_data;
1119 return nfs_clone_generic_sb(data, nfs_clone_sb, nfs_clone_server, mnt);
1122 #ifdef CONFIG_NFS_V4
1123 static struct rpc_clnt *nfs4_create_client(struct nfs_server *server,
1124 struct rpc_timeout *timeparms, int proto, rpc_authflavor_t flavor)
1126 struct nfs_client *clp;
1127 struct rpc_xprt *xprt = NULL;
1128 struct rpc_clnt *clnt = NULL;
1131 clp = nfs_get_client(server->hostname, &server->addr, 4);
1133 dprintk("%s: failed to create NFS4 client.\n", __FUNCTION__);
1134 return ERR_PTR(err);
1137 /* Now create transport and client */
1138 if (clp->cl_cons_state == NFS_CS_INITING) {
1139 clp->rpc_ops = &nfs_v4_clientops;
1141 xprt = xprt_create_proto(proto, &server->addr, timeparms);
1143 err = PTR_ERR(xprt);
1144 dprintk("%s: cannot create RPC transport. Error = %d\n",
1146 goto client_init_error;
1148 /* Bind to a reserved port! */
1150 clnt = rpc_create_client(xprt, server->hostname, &nfs_program,
1151 clp->cl_nfsversion, flavor);
1153 err = PTR_ERR(clnt);
1154 dprintk("%s: cannot create RPC client. Error = %d\n",
1156 goto client_init_error;
1159 clnt->cl_softrtry = 1;
1160 clp->cl_rpcclient = clnt;
1161 memcpy(clp->cl_ipaddr, server->ip_addr, sizeof(clp->cl_ipaddr));
1162 err = nfs_idmap_new(clp);
1164 dprintk("%s: failed to create idmapper.\n",
1166 goto client_init_error;
1168 __set_bit(NFS_CS_IDMAP, &clp->cl_res_state);
1169 nfs_mark_client_ready(clp, 0);
1172 clnt = rpc_clone_client(clp->cl_rpcclient);
1175 dprintk("%s: cannot create RPC client. Error = %d\n",
1180 if (clnt->cl_auth->au_flavor != flavor) {
1181 struct rpc_auth *auth;
1183 auth = rpcauth_create(flavor, clnt);
1185 dprintk("%s: couldn't create credcache!\n", __FUNCTION__);
1186 return (struct rpc_clnt *)auth;
1190 server->nfs_client = clp;
1191 down_write(&clp->cl_sem);
1192 list_add_tail(&server->nfs4_siblings, &clp->cl_superblocks);
1193 up_write(&clp->cl_sem);
1197 nfs_mark_client_ready(clp, err);
1198 nfs_put_client(clp);
1199 return ERR_PTR(err);
1203 * Set up an NFS4 superblock
1205 static int nfs4_fill_super(struct super_block *sb, struct nfs4_mount_data *data, int silent)
1207 struct nfs_server *server;
1208 struct rpc_timeout timeparms;
1209 rpc_authflavor_t authflavour;
1212 sb->s_blocksize_bits = 0;
1213 sb->s_blocksize = 0;
1214 server = NFS_SB(sb);
1215 if (data->rsize != 0)
1216 server->rsize = nfs_block_size(data->rsize, NULL);
1217 if (data->wsize != 0)
1218 server->wsize = nfs_block_size(data->wsize, NULL);
1219 server->flags = data->flags & NFS_MOUNT_FLAGMASK;
1220 server->caps = NFS_CAP_ATOMIC_OPEN;
1222 server->acregmin = data->acregmin*HZ;
1223 server->acregmax = data->acregmax*HZ;
1224 server->acdirmin = data->acdirmin*HZ;
1225 server->acdirmax = data->acdirmax*HZ;
1227 nfs_init_timeout_values(&timeparms, data->proto, data->timeo, data->retrans);
1229 server->retrans_timeo = timeparms.to_initval;
1230 server->retrans_count = timeparms.to_retries;
1232 /* Now create transport and client */
1233 authflavour = RPC_AUTH_UNIX;
1234 if (data->auth_flavourlen != 0) {
1235 if (data->auth_flavourlen != 1) {
1236 dprintk("%s: Invalid number of RPC auth flavours %d.\n",
1237 __FUNCTION__, data->auth_flavourlen);
1241 if (copy_from_user(&authflavour, data->auth_flavours, sizeof(authflavour))) {
1247 server->client = nfs4_create_client(server, &timeparms, data->proto, authflavour);
1248 if (IS_ERR(server->client)) {
1249 err = PTR_ERR(server->client);
1250 dprintk("%s: cannot create RPC client. Error = %d\n",
1255 sb->s_time_gran = 1;
1257 sb->s_op = &nfs4_sops;
1258 err = nfs_sb_init(sb, authflavour);
1264 static int nfs4_compare_super(struct super_block *sb, void *data)
1266 struct nfs_server *server = data;
1267 struct nfs_server *old = NFS_SB(sb);
1269 if (strcmp(server->hostname, old->hostname) != 0)
1271 if (strcmp(server->mnt_path, old->mnt_path) != 0)
1277 nfs_copy_user_string(char *dst, struct nfs_string *src, int maxlen)
1282 return ERR_PTR(-EINVAL);
1283 if (src->len < maxlen)
1286 p = dst = kmalloc(maxlen + 1, GFP_KERNEL);
1288 return ERR_PTR(-ENOMEM);
1290 if (copy_from_user(dst, src->data, maxlen)) {
1292 return ERR_PTR(-EFAULT);
1298 static int nfs4_get_sb(struct file_system_type *fs_type,
1299 int flags, const char *dev_name, void *raw_data, struct vfsmount *mnt)
1302 struct nfs_server *server;
1303 struct super_block *s;
1304 struct nfs4_mount_data *data = raw_data;
1308 dprintk("%s: missing data argument\n", __FUNCTION__);
1311 if (data->version <= 0 || data->version > NFS4_MOUNT_VERSION) {
1312 dprintk("%s: bad mount version\n", __FUNCTION__);
1316 server = kzalloc(sizeof(struct nfs_server), GFP_KERNEL);
1319 /* Zero out the NFS state stuff */
1320 init_nfsv4_state(server);
1321 server->client = server->client_sys = server->client_acl = ERR_PTR(-EINVAL);
1323 p = nfs_copy_user_string(NULL, &data->hostname, 256);
1326 server->hostname = p;
1328 p = nfs_copy_user_string(NULL, &data->mnt_path, 1024);
1331 server->mnt_path = p;
1333 p = nfs_copy_user_string(server->ip_addr, &data->client_addr,
1334 sizeof(server->ip_addr) - 1);
1338 /* We now require that the mount process passes the remote address */
1339 if (data->host_addrlen != sizeof(server->addr)) {
1343 if (copy_from_user(&server->addr, data->host_addr, sizeof(server->addr))) {
1347 if (server->addr.sin_family != AF_INET ||
1348 server->addr.sin_addr.s_addr == INADDR_ANY) {
1349 dprintk("%s: mount program didn't pass remote IP address!\n",
1355 s = sget(fs_type, nfs4_compare_super, nfs_set_super, server);
1362 kfree(server->mnt_path);
1363 kfree(server->hostname);
1365 return simple_set_mnt(mnt, s);
1370 error = nfs4_fill_super(s, data, flags & MS_SILENT ? 1 : 0);
1372 up_write(&s->s_umount);
1373 deactivate_super(s);
1376 s->s_flags |= MS_ACTIVE;
1377 return simple_set_mnt(mnt, s);
1381 kfree(server->mnt_path);
1382 kfree(server->hostname);
1387 static void nfs4_kill_super(struct super_block *sb)
1389 struct nfs_server *server = NFS_SB(sb);
1391 nfs_return_all_delegations(sb);
1392 kill_anon_super(sb);
1394 nfs4_renewd_prepare_shutdown(server);
1396 if (server->client != NULL && !IS_ERR(server->client))
1397 rpc_shutdown_client(server->client);
1399 destroy_nfsv4_state(server);
1401 nfs_free_iostats(server->io_stats);
1402 kfree(server->hostname);
1404 nfs_release_automount_timer();
1408 * Constructs the SERVER-side path
1410 static inline char *nfs4_dup_path(const struct dentry *dentry)
1412 char *page = (char *) __get_free_page(GFP_USER);
1415 path = nfs4_path(dentry, page, PAGE_SIZE);
1416 if (!IS_ERR(path)) {
1417 int len = PAGE_SIZE + page - path;
1420 path = kmalloc(len, GFP_KERNEL);
1422 memcpy(path, tmp, len);
1424 path = ERR_PTR(-ENOMEM);
1426 free_page((unsigned long)page);
1430 static struct super_block *nfs4_clone_sb(struct nfs_server *server, struct nfs_clone_mount *data)
1432 const struct dentry *dentry = data->dentry;
1433 struct nfs_client *clp = server->nfs_client;
1434 struct super_block *sb;
1436 server->fsid = data->fattr->fsid;
1437 nfs_copy_fh(&server->fh, data->fh);
1438 server->mnt_path = nfs4_dup_path(dentry);
1439 if (IS_ERR(server->mnt_path)) {
1440 sb = (struct super_block *)server->mnt_path;
1443 sb = sget(&nfs4_fs_type, nfs4_compare_super, nfs_set_super, server);
1444 if (IS_ERR(sb) || sb->s_root)
1446 nfs4_server_capabilities(server, &server->fh);
1448 down_write(&clp->cl_sem);
1449 list_add_tail(&server->nfs4_siblings, &clp->cl_superblocks);
1450 up_write(&clp->cl_sem);
1453 kfree(server->mnt_path);
1455 server->mnt_path = NULL;
1459 static int nfs_clone_nfs4_sb(struct file_system_type *fs_type,
1460 int flags, const char *dev_name, void *raw_data, struct vfsmount *mnt)
1462 struct nfs_clone_mount *data = raw_data;
1463 return nfs_clone_generic_sb(data, nfs4_clone_sb, nfs_clone_server, mnt);
1466 static struct super_block *nfs4_referral_sb(struct nfs_server *server, struct nfs_clone_mount *data)
1468 struct super_block *sb = ERR_PTR(-ENOMEM);
1471 len = strlen(data->mnt_path) + 1;
1472 server->mnt_path = kmalloc(len, GFP_KERNEL);
1473 if (server->mnt_path == NULL)
1475 memcpy(server->mnt_path, data->mnt_path, len);
1476 memcpy(&server->addr, data->addr, sizeof(struct sockaddr_in));
1478 sb = sget(&nfs4_fs_type, nfs4_compare_super, nfs_set_super, server);
1479 if (IS_ERR(sb) || sb->s_root)
1483 kfree(server->mnt_path);
1485 server->mnt_path = NULL;
1489 static struct nfs_server *nfs4_referral_server(struct super_block *sb, struct nfs_clone_mount *data)
1491 struct nfs_server *server = NFS_SB(sb);
1492 struct rpc_timeout timeparms;
1493 int proto, timeo, retrans;
1496 proto = IPPROTO_TCP;
1497 /* Since we are following a referral and there may be alternatives,
1498 set the timeouts and retries to low values */
1501 nfs_init_timeout_values(&timeparms, proto, timeo, retrans);
1503 nfs_put_client(server->nfs_client);
1504 server->nfs_client = NULL;
1505 server->client = nfs4_create_client(server, &timeparms, proto, data->authflavor);
1506 if (IS_ERR((err = server->client)))
1509 sb->s_time_gran = 1;
1510 sb->s_op = &nfs4_sops;
1511 err = ERR_PTR(nfs_sb_init(sb, data->authflavor));
1515 return (struct nfs_server *)err;
1518 static int nfs_referral_nfs4_sb(struct file_system_type *fs_type,
1519 int flags, const char *dev_name, void *raw_data, struct vfsmount *mnt)
1521 struct nfs_clone_mount *data = raw_data;
1522 return nfs_clone_generic_sb(data, nfs4_referral_sb, nfs4_referral_server, mnt);