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 super_block *, 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 struct super_block *nfs_get_sb(struct file_system_type *, int, const char *, void *);
114 static struct super_block *nfs_clone_nfs_sb(struct file_system_type *fs_type,
115 int flags, const char *dev_name, void *raw_data);
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 struct super_block *nfs4_get_sb(struct file_system_type *fs_type,
147 int flags, const char *dev_name, void *raw_data);
148 static struct super_block *nfs_clone_nfs4_sb(struct file_system_type *fs_type,
149 int flags, const char *dev_name, void *raw_data);
150 static struct super_block *nfs_referral_nfs4_sb(struct file_system_type *fs_type,
151 int flags, const char *dev_name, void *raw_data);
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 const int nfs_set_port_min = 0;
191 static const int nfs_set_port_max = 65535;
193 static int param_set_port(const char *val, struct kernel_param *kp)
196 int num = simple_strtol(val, &endp, 0);
197 if (endp == val || *endp || num < nfs_set_port_min || num > nfs_set_port_max)
199 *((int *)kp->arg) = num;
203 module_param_call(callback_tcpport, param_set_port, param_get_int,
204 &nfs_callback_set_tcpport, 0644);
206 static int param_set_idmap_timeout(const char *val, struct kernel_param *kp)
209 int num = simple_strtol(val, &endp, 0);
211 if (endp == val || *endp || num < 0 || jif < num)
213 *((int *)kp->arg) = jif;
217 module_param_call(idmap_cache_timeout, param_set_idmap_timeout, param_get_int,
218 &nfs_idmap_cache_timeout, 0644);
221 * Register the NFS filesystems
223 int __init register_nfs_fs(void)
227 ret = register_filesystem(&nfs_fs_type);
232 ret = nfs_register_sysctl();
235 ret = register_filesystem(&nfs4_fs_type);
243 nfs_unregister_sysctl();
245 unregister_filesystem(&nfs_fs_type);
252 * Unregister the NFS filesystems
254 void __exit unregister_nfs_fs(void)
257 unregister_filesystem(&nfs4_fs_type);
258 nfs_unregister_sysctl();
260 unregister_filesystem(&nfs_fs_type);
264 * Deliver file system statistics to userspace
266 static int nfs_statfs(struct super_block *sb, struct kstatfs *buf)
268 struct nfs_server *server = NFS_SB(sb);
269 unsigned char blockbits;
270 unsigned long blockres;
271 struct nfs_fh *rootfh = NFS_FH(sb->s_root->d_inode);
272 struct nfs_fattr fattr;
273 struct nfs_fsstat res = {
280 error = server->rpc_ops->statfs(server, rootfh, &res);
281 buf->f_type = NFS_SUPER_MAGIC;
286 * Current versions of glibc do not correctly handle the
287 * case where f_frsize != f_bsize. Eventually we want to
288 * report the value of wtmult in this field.
290 buf->f_frsize = sb->s_blocksize;
293 * On most *nix systems, f_blocks, f_bfree, and f_bavail
294 * are reported in units of f_frsize. Linux hasn't had
295 * an f_frsize field in its statfs struct until recently,
296 * thus historically Linux's sys_statfs reports these
297 * fields in units of f_bsize.
299 buf->f_bsize = sb->s_blocksize;
300 blockbits = sb->s_blocksize_bits;
301 blockres = (1 << blockbits) - 1;
302 buf->f_blocks = (res.tbytes + blockres) >> blockbits;
303 buf->f_bfree = (res.fbytes + blockres) >> blockbits;
304 buf->f_bavail = (res.abytes + blockres) >> blockbits;
306 buf->f_files = res.tfiles;
307 buf->f_ffree = res.afiles;
309 buf->f_namelen = server->namelen;
315 dprintk("%s: statfs error = %d\n", __FUNCTION__, -error);
316 buf->f_bsize = buf->f_blocks = buf->f_bfree = buf->f_bavail = -1;
322 * Describe the mount options in force on this server representation
324 static void nfs_show_mount_options(struct seq_file *m, struct nfs_server *nfss, int showdefaults)
326 static struct proc_nfs_info {
331 { NFS_MOUNT_SOFT, ",soft", ",hard" },
332 { NFS_MOUNT_INTR, ",intr", "" },
333 { NFS_MOUNT_NOCTO, ",nocto", "" },
334 { NFS_MOUNT_NOAC, ",noac", "" },
335 { NFS_MOUNT_NONLM, ",nolock", "" },
336 { NFS_MOUNT_NOACL, ",noacl", "" },
339 struct proc_nfs_info *nfs_infop;
343 seq_printf(m, ",vers=%d", nfss->rpc_ops->version);
344 seq_printf(m, ",rsize=%d", nfss->rsize);
345 seq_printf(m, ",wsize=%d", nfss->wsize);
346 if (nfss->acregmin != 3*HZ || showdefaults)
347 seq_printf(m, ",acregmin=%d", nfss->acregmin/HZ);
348 if (nfss->acregmax != 60*HZ || showdefaults)
349 seq_printf(m, ",acregmax=%d", nfss->acregmax/HZ);
350 if (nfss->acdirmin != 30*HZ || showdefaults)
351 seq_printf(m, ",acdirmin=%d", nfss->acdirmin/HZ);
352 if (nfss->acdirmax != 60*HZ || showdefaults)
353 seq_printf(m, ",acdirmax=%d", nfss->acdirmax/HZ);
354 for (nfs_infop = nfs_info; nfs_infop->flag; nfs_infop++) {
355 if (nfss->flags & nfs_infop->flag)
356 seq_puts(m, nfs_infop->str);
358 seq_puts(m, nfs_infop->nostr);
360 switch (nfss->client->cl_xprt->prot) {
368 snprintf(buf, sizeof(buf), "%u", nfss->client->cl_xprt->prot);
371 seq_printf(m, ",proto=%s", proto);
372 seq_printf(m, ",timeo=%lu", 10U * nfss->retrans_timeo / HZ);
373 seq_printf(m, ",retrans=%u", nfss->retrans_count);
377 * Describe the mount options on this VFS mountpoint
379 static int nfs_show_options(struct seq_file *m, struct vfsmount *mnt)
381 struct nfs_server *nfss = NFS_SB(mnt->mnt_sb);
383 nfs_show_mount_options(m, nfss, 0);
385 seq_puts(m, ",addr=");
386 seq_escape(m, nfss->hostname, " \t\n\\");
392 * Present statistical information for this VFS mountpoint
394 static int nfs_show_stats(struct seq_file *m, struct vfsmount *mnt)
397 struct nfs_server *nfss = NFS_SB(mnt->mnt_sb);
398 struct rpc_auth *auth = nfss->client->cl_auth;
399 struct nfs_iostats totals = { };
401 seq_printf(m, "statvers=%s", NFS_IOSTAT_VERS);
404 * Display all mount option settings
406 seq_printf(m, "\n\topts:\t");
407 seq_puts(m, mnt->mnt_sb->s_flags & MS_RDONLY ? "ro" : "rw");
408 seq_puts(m, mnt->mnt_sb->s_flags & MS_SYNCHRONOUS ? ",sync" : "");
409 seq_puts(m, mnt->mnt_sb->s_flags & MS_NOATIME ? ",noatime" : "");
410 seq_puts(m, mnt->mnt_sb->s_flags & MS_NODIRATIME ? ",nodiratime" : "");
411 nfs_show_mount_options(m, nfss, 1);
413 seq_printf(m, "\n\tage:\t%lu", (jiffies - nfss->mount_time) / HZ);
415 seq_printf(m, "\n\tcaps:\t");
416 seq_printf(m, "caps=0x%x", nfss->caps);
417 seq_printf(m, ",wtmult=%d", nfss->wtmult);
418 seq_printf(m, ",dtsize=%d", nfss->dtsize);
419 seq_printf(m, ",bsize=%d", nfss->bsize);
420 seq_printf(m, ",namelen=%d", nfss->namelen);
423 if (nfss->rpc_ops->version == 4) {
424 seq_printf(m, "\n\tnfsv4:\t");
425 seq_printf(m, "bm0=0x%x", nfss->attr_bitmask[0]);
426 seq_printf(m, ",bm1=0x%x", nfss->attr_bitmask[1]);
427 seq_printf(m, ",acl=0x%x", nfss->acl_bitmask);
432 * Display security flavor in effect for this mount
434 seq_printf(m, "\n\tsec:\tflavor=%d", auth->au_ops->au_flavor);
436 seq_printf(m, ",pseudoflavor=%d", auth->au_flavor);
439 * Display superblock I/O counters
441 for_each_possible_cpu(cpu) {
442 struct nfs_iostats *stats;
445 stats = per_cpu_ptr(nfss->io_stats, cpu);
447 for (i = 0; i < __NFSIOS_COUNTSMAX; i++)
448 totals.events[i] += stats->events[i];
449 for (i = 0; i < __NFSIOS_BYTESMAX; i++)
450 totals.bytes[i] += stats->bytes[i];
455 seq_printf(m, "\n\tevents:\t");
456 for (i = 0; i < __NFSIOS_COUNTSMAX; i++)
457 seq_printf(m, "%lu ", totals.events[i]);
458 seq_printf(m, "\n\tbytes:\t");
459 for (i = 0; i < __NFSIOS_BYTESMAX; i++)
460 seq_printf(m, "%Lu ", totals.bytes[i]);
463 rpc_print_iostats(m, nfss->client);
469 * Begin unmount by attempting to remove all automounted mountpoints we added
470 * in response to traversals
472 static void nfs_umount_begin(struct vfsmount *vfsmnt, int flags)
474 struct nfs_server *server;
475 struct rpc_clnt *rpc;
477 shrink_submounts(vfsmnt, &nfs_automount_list);
478 if (!(flags & MNT_FORCE))
480 /* -EIO all pending I/O */
481 server = NFS_SB(vfsmnt->mnt_sb);
482 rpc = server->client;
484 rpc_killall_tasks(rpc);
485 rpc = server->client_acl;
487 rpc_killall_tasks(rpc);
491 * Obtain the root inode of the file system.
493 static struct inode *
494 nfs_get_root(struct super_block *sb, struct nfs_fh *rootfh, struct nfs_fsinfo *fsinfo)
496 struct nfs_server *server = NFS_SB(sb);
499 error = server->rpc_ops->getroot(server, rootfh, fsinfo);
501 dprintk("nfs_get_root: getattr error = %d\n", -error);
502 return ERR_PTR(error);
505 server->fsid = fsinfo->fattr->fsid;
506 return nfs_fhget(sb, rootfh, fsinfo->fattr);
510 * Do NFS version-independent mount processing, and sanity checking
513 nfs_sb_init(struct super_block *sb, rpc_authflavor_t authflavor)
515 struct nfs_server *server;
516 struct inode *root_inode;
517 struct nfs_fattr fattr;
518 struct nfs_fsinfo fsinfo = {
521 struct nfs_pathconf pathinfo = {
524 int no_root_error = 0;
525 unsigned long max_rpc_payload;
527 /* We probably want something more informative here */
528 snprintf(sb->s_id, sizeof(sb->s_id), "%x:%x", MAJOR(sb->s_dev), MINOR(sb->s_dev));
532 sb->s_magic = NFS_SUPER_MAGIC;
534 server->io_stats = nfs_alloc_iostats();
535 if (server->io_stats == NULL)
538 root_inode = nfs_get_root(sb, &server->fh, &fsinfo);
539 /* Did getting the root inode fail? */
540 if (IS_ERR(root_inode)) {
541 no_root_error = PTR_ERR(root_inode);
544 sb->s_root = d_alloc_root(root_inode);
546 no_root_error = -ENOMEM;
549 sb->s_root->d_op = server->rpc_ops->dentry_ops;
551 /* mount time stamp, in seconds */
552 server->mount_time = jiffies;
554 /* Get some general file system info */
555 if (server->namelen == 0 &&
556 server->rpc_ops->pathconf(server, &server->fh, &pathinfo) >= 0)
557 server->namelen = pathinfo.max_namelen;
558 /* Work out a lot of parameters */
559 if (server->rsize == 0)
560 server->rsize = nfs_block_size(fsinfo.rtpref, NULL);
561 if (server->wsize == 0)
562 server->wsize = nfs_block_size(fsinfo.wtpref, NULL);
564 if (fsinfo.rtmax >= 512 && server->rsize > fsinfo.rtmax)
565 server->rsize = nfs_block_size(fsinfo.rtmax, NULL);
566 if (fsinfo.wtmax >= 512 && server->wsize > fsinfo.wtmax)
567 server->wsize = nfs_block_size(fsinfo.wtmax, NULL);
569 max_rpc_payload = nfs_block_size(rpc_max_payload(server->client), NULL);
570 if (server->rsize > max_rpc_payload)
571 server->rsize = max_rpc_payload;
572 if (server->rsize > NFS_MAX_FILE_IO_SIZE)
573 server->rsize = NFS_MAX_FILE_IO_SIZE;
574 server->rpages = (server->rsize + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT;
576 if (server->wsize > max_rpc_payload)
577 server->wsize = max_rpc_payload;
578 if (server->wsize > NFS_MAX_FILE_IO_SIZE)
579 server->wsize = NFS_MAX_FILE_IO_SIZE;
580 server->wpages = (server->wsize + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT;
582 if (sb->s_blocksize == 0)
583 sb->s_blocksize = nfs_block_bits(server->wsize,
584 &sb->s_blocksize_bits);
585 server->wtmult = nfs_block_bits(fsinfo.wtmult, NULL);
587 server->dtsize = nfs_block_size(fsinfo.dtpref, NULL);
588 if (server->dtsize > PAGE_CACHE_SIZE)
589 server->dtsize = PAGE_CACHE_SIZE;
590 if (server->dtsize > server->rsize)
591 server->dtsize = server->rsize;
593 if (server->flags & NFS_MOUNT_NOAC) {
594 server->acregmin = server->acregmax = 0;
595 server->acdirmin = server->acdirmax = 0;
596 sb->s_flags |= MS_SYNCHRONOUS;
598 server->backing_dev_info.ra_pages = server->rpages * NFS_MAX_READAHEAD;
600 nfs_super_set_maxbytes(sb, fsinfo.maxfilesize);
602 server->client->cl_intr = (server->flags & NFS_MOUNT_INTR) ? 1 : 0;
603 server->client->cl_softrtry = (server->flags & NFS_MOUNT_SOFT) ? 1 : 0;
605 /* We're airborne Set socket buffersize */
606 rpc_setbufsize(server->client, server->wsize + 100, server->rsize + 100);
608 /* Yargs. It didn't work out. */
610 dprintk("nfs_sb_init: get root inode failed: errno %d\n", -no_root_error);
611 if (!IS_ERR(root_inode))
613 return no_root_error;
617 * Initialise the timeout values for a connection
619 static void nfs_init_timeout_values(struct rpc_timeout *to, int proto, unsigned int timeo, unsigned int retrans)
621 to->to_initval = timeo * HZ / 10;
622 to->to_retries = retrans;
629 to->to_initval = 60 * HZ;
630 if (to->to_initval > NFS_MAX_TCP_TIMEOUT)
631 to->to_initval = NFS_MAX_TCP_TIMEOUT;
632 to->to_increment = to->to_initval;
633 to->to_maxval = to->to_initval + (to->to_increment * to->to_retries);
634 to->to_exponential = 0;
639 to->to_initval = 11 * HZ / 10;
640 if (to->to_initval > NFS_MAX_UDP_TIMEOUT)
641 to->to_initval = NFS_MAX_UDP_TIMEOUT;
642 to->to_maxval = NFS_MAX_UDP_TIMEOUT;
643 to->to_exponential = 1;
649 * Create an RPC client handle.
651 static struct rpc_clnt *
652 nfs_create_client(struct nfs_server *server, const struct nfs_mount_data *data)
654 struct rpc_timeout timeparms;
655 struct rpc_xprt *xprt = NULL;
656 struct rpc_clnt *clnt = NULL;
657 int proto = (data->flags & NFS_MOUNT_TCP) ? IPPROTO_TCP : IPPROTO_UDP;
659 nfs_init_timeout_values(&timeparms, proto, data->timeo, data->retrans);
661 server->retrans_timeo = timeparms.to_initval;
662 server->retrans_count = timeparms.to_retries;
664 /* create transport and client */
665 xprt = xprt_create_proto(proto, &server->addr, &timeparms);
667 dprintk("%s: cannot create RPC transport. Error = %ld\n",
668 __FUNCTION__, PTR_ERR(xprt));
669 return (struct rpc_clnt *)xprt;
671 clnt = rpc_create_client(xprt, server->hostname, &nfs_program,
672 server->rpc_ops->version, data->pseudoflavor);
674 dprintk("%s: cannot create RPC client. Error = %ld\n",
675 __FUNCTION__, PTR_ERR(xprt));
680 clnt->cl_softrtry = 1;
689 * Clone a server record
691 static struct nfs_server *nfs_clone_server(struct super_block *sb, struct nfs_clone_mount *data)
693 struct nfs_server *server = NFS_SB(sb);
694 struct nfs_server *parent = NFS_SB(data->sb);
695 struct inode *root_inode;
696 struct nfs_fsinfo fsinfo;
697 void *err = ERR_PTR(-ENOMEM);
699 sb->s_op = data->sb->s_op;
700 sb->s_blocksize = data->sb->s_blocksize;
701 sb->s_blocksize_bits = data->sb->s_blocksize_bits;
702 sb->s_maxbytes = data->sb->s_maxbytes;
704 server->client_sys = server->client_acl = ERR_PTR(-EINVAL);
705 server->io_stats = nfs_alloc_iostats();
706 if (server->io_stats == NULL)
709 server->client = rpc_clone_client(parent->client);
710 if (IS_ERR((err = server->client)))
713 if (!IS_ERR(parent->client_sys)) {
714 server->client_sys = rpc_clone_client(parent->client_sys);
715 if (IS_ERR((err = server->client_sys)))
718 if (!IS_ERR(parent->client_acl)) {
719 server->client_acl = rpc_clone_client(parent->client_acl);
720 if (IS_ERR((err = server->client_acl)))
723 root_inode = nfs_fhget(sb, data->fh, data->fattr);
726 sb->s_root = d_alloc_root(root_inode);
729 fsinfo.fattr = data->fattr;
730 if (NFS_PROTO(root_inode)->fsinfo(server, data->fh, &fsinfo) == 0)
731 nfs_super_set_maxbytes(sb, fsinfo.maxfilesize);
732 sb->s_root->d_op = server->rpc_ops->dentry_ops;
733 sb->s_flags |= MS_ACTIVE;
742 * Copy an existing superblock and attach revised data
744 static struct super_block *nfs_clone_generic_sb(struct nfs_clone_mount *data,
745 struct super_block *(*fill_sb)(struct nfs_server *, struct nfs_clone_mount *),
746 struct nfs_server *(*fill_server)(struct super_block *, struct nfs_clone_mount *))
748 struct nfs_server *server;
749 struct nfs_server *parent = NFS_SB(data->sb);
750 struct super_block *sb = ERR_PTR(-EINVAL);
751 void *err = ERR_PTR(-ENOMEM);
755 server = kmalloc(sizeof(struct nfs_server), GFP_KERNEL);
758 memcpy(server, parent, sizeof(*server));
759 hostname = (data->hostname != NULL) ? data->hostname : parent->hostname;
760 len = strlen(hostname) + 1;
761 server->hostname = kmalloc(len, GFP_KERNEL);
762 if (server->hostname == NULL)
764 memcpy(server->hostname, hostname, len);
765 if (rpciod_up() != 0)
768 sb = fill_sb(server, data);
769 if (IS_ERR((err = sb)) || sb->s_root)
772 server = fill_server(sb, data);
773 if (IS_ERR((err = server)))
777 up_write(&sb->s_umount);
778 deactivate_super(sb);
779 return (struct super_block *)err;
783 kfree(server->hostname);
787 return (struct super_block *)err;
791 * Set up an NFS2/3 superblock
793 * The way this works is that the mount process passes a structure
794 * in the data argument which contains the server's IP address
795 * and the root file handle obtained from the server's mount
796 * daemon. We stash these away in the private superblock fields.
799 nfs_fill_super(struct super_block *sb, struct nfs_mount_data *data, int silent)
801 struct nfs_server *server;
802 rpc_authflavor_t authflavor;
805 sb->s_blocksize_bits = 0;
808 sb->s_blocksize = nfs_block_size(data->bsize, &sb->s_blocksize_bits);
810 server->rsize = nfs_block_size(data->rsize, NULL);
812 server->wsize = nfs_block_size(data->wsize, NULL);
813 server->flags = data->flags & NFS_MOUNT_FLAGMASK;
815 server->acregmin = data->acregmin*HZ;
816 server->acregmax = data->acregmax*HZ;
817 server->acdirmin = data->acdirmin*HZ;
818 server->acdirmax = data->acdirmax*HZ;
820 /* Start lockd here, before we might error out */
821 if (!(server->flags & NFS_MOUNT_NONLM))
824 server->namelen = data->namlen;
825 server->hostname = kmalloc(strlen(data->hostname) + 1, GFP_KERNEL);
826 if (!server->hostname)
828 strcpy(server->hostname, data->hostname);
830 /* Check NFS protocol revision and initialize RPC op vector
831 * and file handle pool. */
833 if (server->flags & NFS_MOUNT_VER3) {
834 server->rpc_ops = &nfs_v3_clientops;
835 server->caps |= NFS_CAP_READDIRPLUS;
837 server->rpc_ops = &nfs_v2_clientops;
840 server->rpc_ops = &nfs_v2_clientops;
843 /* Fill in pseudoflavor for mount version < 5 */
844 if (!(data->flags & NFS_MOUNT_SECFLAVOUR))
845 data->pseudoflavor = RPC_AUTH_UNIX;
846 authflavor = data->pseudoflavor; /* save for sb_init() */
847 /* XXX maybe we want to add a server->pseudoflavor field */
849 /* Create RPC client handles */
850 server->client = nfs_create_client(server, data);
851 if (IS_ERR(server->client))
852 return PTR_ERR(server->client);
853 /* RFC 2623, sec 2.3.2 */
854 if (authflavor != RPC_AUTH_UNIX) {
855 struct rpc_auth *auth;
857 server->client_sys = rpc_clone_client(server->client);
858 if (IS_ERR(server->client_sys))
859 return PTR_ERR(server->client_sys);
860 auth = rpcauth_create(RPC_AUTH_UNIX, server->client_sys);
862 return PTR_ERR(auth);
864 atomic_inc(&server->client->cl_count);
865 server->client_sys = server->client;
867 if (server->flags & NFS_MOUNT_VER3) {
868 #ifdef CONFIG_NFS_V3_ACL
869 if (!(server->flags & NFS_MOUNT_NOACL)) {
870 server->client_acl = rpc_bind_new_program(server->client, &nfsacl_program, 3);
871 /* No errors! Assume that Sun nfsacls are supported */
872 if (!IS_ERR(server->client_acl))
873 server->caps |= NFS_CAP_ACLS;
876 server->flags &= ~NFS_MOUNT_NOACL;
877 #endif /* CONFIG_NFS_V3_ACL */
879 * The VFS shouldn't apply the umask to mode bits. We will
880 * do so ourselves when necessary.
882 sb->s_flags |= MS_POSIXACL;
883 if (server->namelen == 0 || server->namelen > NFS3_MAXNAMLEN)
884 server->namelen = NFS3_MAXNAMLEN;
887 if (server->namelen == 0 || server->namelen > NFS2_MAXNAMLEN)
888 server->namelen = NFS2_MAXNAMLEN;
891 sb->s_op = &nfs_sops;
892 return nfs_sb_init(sb, authflavor);
895 static int nfs_set_super(struct super_block *s, void *data)
898 return set_anon_super(s, data);
901 static int nfs_compare_super(struct super_block *sb, void *data)
903 struct nfs_server *server = data;
904 struct nfs_server *old = NFS_SB(sb);
906 if (old->addr.sin_addr.s_addr != server->addr.sin_addr.s_addr)
908 if (old->addr.sin_port != server->addr.sin_port)
910 return !nfs_compare_fh(&old->fh, &server->fh);
913 static struct super_block *nfs_get_sb(struct file_system_type *fs_type,
914 int flags, const char *dev_name, void *raw_data)
917 struct nfs_server *server = NULL;
918 struct super_block *s;
920 struct nfs_mount_data *data = raw_data;
922 s = ERR_PTR(-EINVAL);
924 dprintk("%s: missing data argument\n", __FUNCTION__);
927 if (data->version <= 0 || data->version > NFS_MOUNT_VERSION) {
928 dprintk("%s: bad mount version\n", __FUNCTION__);
931 switch (data->version) {
937 if (data->flags & NFS_MOUNT_VER3) {
938 dprintk("%s: mount structure version %d does not support NFSv3\n",
943 data->root.size = NFS2_FHSIZE;
944 memcpy(data->root.data, data->old_root.data, NFS2_FHSIZE);
946 if (data->flags & NFS_MOUNT_SECFLAVOUR) {
947 dprintk("%s: mount structure version %d does not support strong security\n",
953 memset(data->context, 0, sizeof(data->context));
955 #ifndef CONFIG_NFS_V3
956 /* If NFSv3 is not compiled in, return -EPROTONOSUPPORT */
957 s = ERR_PTR(-EPROTONOSUPPORT);
958 if (data->flags & NFS_MOUNT_VER3) {
959 dprintk("%s: NFSv3 not compiled into kernel\n", __FUNCTION__);
962 #endif /* CONFIG_NFS_V3 */
964 s = ERR_PTR(-ENOMEM);
965 server = kzalloc(sizeof(struct nfs_server), GFP_KERNEL);
968 /* Zero out the NFS state stuff */
969 init_nfsv4_state(server);
970 server->client = server->client_sys = server->client_acl = ERR_PTR(-EINVAL);
973 if (data->flags & NFS_MOUNT_VER3)
974 root->size = data->root.size;
976 root->size = NFS2_FHSIZE;
977 s = ERR_PTR(-EINVAL);
978 if (root->size > sizeof(root->data)) {
979 dprintk("%s: invalid root filehandle\n", __FUNCTION__);
982 memcpy(root->data, data->root.data, root->size);
984 /* We now require that the mount process passes the remote address */
985 memcpy(&server->addr, &data->addr, sizeof(server->addr));
986 if (server->addr.sin_addr.s_addr == INADDR_ANY) {
987 dprintk("%s: mount program didn't pass remote address!\n",
992 /* Fire up rpciod if not yet running */
993 s = ERR_PTR(rpciod_up());
995 dprintk("%s: couldn't start rpciod! Error = %ld\n",
996 __FUNCTION__, PTR_ERR(s));
1000 s = sget(fs_type, nfs_compare_super, nfs_set_super, server);
1001 if (IS_ERR(s) || s->s_root)
1002 goto out_rpciod_down;
1006 error = nfs_fill_super(s, data, flags & MS_SILENT ? 1 : 0);
1008 up_write(&s->s_umount);
1009 deactivate_super(s);
1010 return ERR_PTR(error);
1012 s->s_flags |= MS_ACTIVE;
1021 static void nfs_kill_super(struct super_block *s)
1023 struct nfs_server *server = NFS_SB(s);
1027 if (!IS_ERR(server->client))
1028 rpc_shutdown_client(server->client);
1029 if (!IS_ERR(server->client_sys))
1030 rpc_shutdown_client(server->client_sys);
1031 if (!IS_ERR(server->client_acl))
1032 rpc_shutdown_client(server->client_acl);
1034 if (!(server->flags & NFS_MOUNT_NONLM))
1035 lockd_down(); /* release rpc.lockd */
1037 rpciod_down(); /* release rpciod */
1039 nfs_free_iostats(server->io_stats);
1040 kfree(server->hostname);
1042 nfs_release_automount_timer();
1045 static struct super_block *nfs_clone_sb(struct nfs_server *server, struct nfs_clone_mount *data)
1047 struct super_block *sb;
1049 server->fsid = data->fattr->fsid;
1050 nfs_copy_fh(&server->fh, data->fh);
1051 sb = sget(&nfs_fs_type, nfs_compare_super, nfs_set_super, server);
1052 if (!IS_ERR(sb) && sb->s_root == NULL && !(server->flags & NFS_MOUNT_NONLM))
1057 static struct super_block *nfs_clone_nfs_sb(struct file_system_type *fs_type,
1058 int flags, const char *dev_name, void *raw_data)
1060 struct nfs_clone_mount *data = raw_data;
1061 return nfs_clone_generic_sb(data, nfs_clone_sb, nfs_clone_server);
1064 #ifdef CONFIG_NFS_V4
1065 static struct rpc_clnt *nfs4_create_client(struct nfs_server *server,
1066 struct rpc_timeout *timeparms, int proto, rpc_authflavor_t flavor)
1068 struct nfs4_client *clp;
1069 struct rpc_xprt *xprt = NULL;
1070 struct rpc_clnt *clnt = NULL;
1073 clp = nfs4_get_client(&server->addr.sin_addr);
1075 dprintk("%s: failed to create NFS4 client.\n", __FUNCTION__);
1076 return ERR_PTR(err);
1079 /* Now create transport and client */
1080 down_write(&clp->cl_sem);
1081 if (IS_ERR(clp->cl_rpcclient)) {
1082 xprt = xprt_create_proto(proto, &server->addr, timeparms);
1084 up_write(&clp->cl_sem);
1085 err = PTR_ERR(xprt);
1086 dprintk("%s: cannot create RPC transport. Error = %d\n",
1090 /* Bind to a reserved port! */
1092 clnt = rpc_create_client(xprt, server->hostname, &nfs_program,
1093 server->rpc_ops->version, flavor);
1095 up_write(&clp->cl_sem);
1096 err = PTR_ERR(clnt);
1097 dprintk("%s: cannot create RPC client. Error = %d\n",
1102 clnt->cl_softrtry = 1;
1103 clp->cl_rpcclient = clnt;
1104 memcpy(clp->cl_ipaddr, server->ip_addr, sizeof(clp->cl_ipaddr));
1107 list_add_tail(&server->nfs4_siblings, &clp->cl_superblocks);
1108 clnt = rpc_clone_client(clp->cl_rpcclient);
1110 server->nfs4_state = clp;
1111 up_write(&clp->cl_sem);
1115 dprintk("%s: cannot create RPC client. Error = %d\n",
1120 if (server->nfs4_state->cl_idmap == NULL) {
1121 dprintk("%s: failed to create idmapper.\n", __FUNCTION__);
1122 return ERR_PTR(-ENOMEM);
1125 if (clnt->cl_auth->au_flavor != flavor) {
1126 struct rpc_auth *auth;
1128 auth = rpcauth_create(flavor, clnt);
1130 dprintk("%s: couldn't create credcache!\n", __FUNCTION__);
1131 return (struct rpc_clnt *)auth;
1138 nfs4_put_client(clp);
1139 return ERR_PTR(err);
1143 * Set up an NFS4 superblock
1145 static int nfs4_fill_super(struct super_block *sb, struct nfs4_mount_data *data, int silent)
1147 struct nfs_server *server;
1148 struct rpc_timeout timeparms;
1149 rpc_authflavor_t authflavour;
1152 sb->s_blocksize_bits = 0;
1153 sb->s_blocksize = 0;
1154 server = NFS_SB(sb);
1155 if (data->rsize != 0)
1156 server->rsize = nfs_block_size(data->rsize, NULL);
1157 if (data->wsize != 0)
1158 server->wsize = nfs_block_size(data->wsize, NULL);
1159 server->flags = data->flags & NFS_MOUNT_FLAGMASK;
1160 server->caps = NFS_CAP_ATOMIC_OPEN;
1162 server->acregmin = data->acregmin*HZ;
1163 server->acregmax = data->acregmax*HZ;
1164 server->acdirmin = data->acdirmin*HZ;
1165 server->acdirmax = data->acdirmax*HZ;
1167 server->rpc_ops = &nfs_v4_clientops;
1169 nfs_init_timeout_values(&timeparms, data->proto, data->timeo, data->retrans);
1171 server->retrans_timeo = timeparms.to_initval;
1172 server->retrans_count = timeparms.to_retries;
1174 /* Now create transport and client */
1175 authflavour = RPC_AUTH_UNIX;
1176 if (data->auth_flavourlen != 0) {
1177 if (data->auth_flavourlen != 1) {
1178 dprintk("%s: Invalid number of RPC auth flavours %d.\n",
1179 __FUNCTION__, data->auth_flavourlen);
1183 if (copy_from_user(&authflavour, data->auth_flavours, sizeof(authflavour))) {
1189 server->client = nfs4_create_client(server, &timeparms, data->proto, authflavour);
1190 if (IS_ERR(server->client)) {
1191 err = PTR_ERR(server->client);
1192 dprintk("%s: cannot create RPC client. Error = %d\n",
1197 sb->s_time_gran = 1;
1199 sb->s_op = &nfs4_sops;
1200 err = nfs_sb_init(sb, authflavour);
1206 static int nfs4_compare_super(struct super_block *sb, void *data)
1208 struct nfs_server *server = data;
1209 struct nfs_server *old = NFS_SB(sb);
1211 if (strcmp(server->hostname, old->hostname) != 0)
1213 if (strcmp(server->mnt_path, old->mnt_path) != 0)
1219 nfs_copy_user_string(char *dst, struct nfs_string *src, int maxlen)
1224 return ERR_PTR(-EINVAL);
1225 if (src->len < maxlen)
1228 p = dst = kmalloc(maxlen + 1, GFP_KERNEL);
1230 return ERR_PTR(-ENOMEM);
1232 if (copy_from_user(dst, src->data, maxlen)) {
1234 return ERR_PTR(-EFAULT);
1240 static struct super_block *nfs4_get_sb(struct file_system_type *fs_type,
1241 int flags, const char *dev_name, void *raw_data)
1244 struct nfs_server *server;
1245 struct super_block *s;
1246 struct nfs4_mount_data *data = raw_data;
1250 dprintk("%s: missing data argument\n", __FUNCTION__);
1251 return ERR_PTR(-EINVAL);
1253 if (data->version <= 0 || data->version > NFS4_MOUNT_VERSION) {
1254 dprintk("%s: bad mount version\n", __FUNCTION__);
1255 return ERR_PTR(-EINVAL);
1258 server = kzalloc(sizeof(struct nfs_server), GFP_KERNEL);
1260 return ERR_PTR(-ENOMEM);
1261 /* Zero out the NFS state stuff */
1262 init_nfsv4_state(server);
1263 server->client = server->client_sys = server->client_acl = ERR_PTR(-EINVAL);
1265 p = nfs_copy_user_string(NULL, &data->hostname, 256);
1268 server->hostname = p;
1270 p = nfs_copy_user_string(NULL, &data->mnt_path, 1024);
1273 server->mnt_path = p;
1275 p = nfs_copy_user_string(server->ip_addr, &data->client_addr,
1276 sizeof(server->ip_addr) - 1);
1280 /* We now require that the mount process passes the remote address */
1281 if (data->host_addrlen != sizeof(server->addr)) {
1282 s = ERR_PTR(-EINVAL);
1285 if (copy_from_user(&server->addr, data->host_addr, sizeof(server->addr))) {
1286 s = ERR_PTR(-EFAULT);
1289 if (server->addr.sin_family != AF_INET ||
1290 server->addr.sin_addr.s_addr == INADDR_ANY) {
1291 dprintk("%s: mount program didn't pass remote IP address!\n",
1293 s = ERR_PTR(-EINVAL);
1297 /* Fire up rpciod if not yet running */
1298 s = ERR_PTR(rpciod_up());
1300 dprintk("%s: couldn't start rpciod! Error = %ld\n",
1301 __FUNCTION__, PTR_ERR(s));
1305 s = sget(fs_type, nfs4_compare_super, nfs_set_super, server);
1307 if (IS_ERR(s) || s->s_root)
1312 error = nfs4_fill_super(s, data, flags & MS_SILENT ? 1 : 0);
1314 up_write(&s->s_umount);
1315 deactivate_super(s);
1316 return ERR_PTR(error);
1318 s->s_flags |= MS_ACTIVE;
1321 s = (struct super_block *)p;
1323 kfree(server->mnt_path);
1324 kfree(server->hostname);
1329 static void nfs4_kill_super(struct super_block *sb)
1331 struct nfs_server *server = NFS_SB(sb);
1333 nfs_return_all_delegations(sb);
1334 kill_anon_super(sb);
1336 nfs4_renewd_prepare_shutdown(server);
1338 if (server->client != NULL && !IS_ERR(server->client))
1339 rpc_shutdown_client(server->client);
1341 destroy_nfsv4_state(server);
1345 nfs_free_iostats(server->io_stats);
1346 kfree(server->hostname);
1348 nfs_release_automount_timer();
1352 * Constructs the SERVER-side path
1354 static inline char *nfs4_dup_path(const struct dentry *dentry)
1356 char *page = (char *) __get_free_page(GFP_USER);
1359 path = nfs4_path(dentry, page, PAGE_SIZE);
1360 if (!IS_ERR(path)) {
1361 int len = PAGE_SIZE + page - path;
1364 path = kmalloc(len, GFP_KERNEL);
1366 memcpy(path, tmp, len);
1368 path = ERR_PTR(-ENOMEM);
1370 free_page((unsigned long)page);
1374 static struct super_block *nfs4_clone_sb(struct nfs_server *server, struct nfs_clone_mount *data)
1376 const struct dentry *dentry = data->dentry;
1377 struct nfs4_client *clp = server->nfs4_state;
1378 struct super_block *sb;
1380 server->fsid = data->fattr->fsid;
1381 nfs_copy_fh(&server->fh, data->fh);
1382 server->mnt_path = nfs4_dup_path(dentry);
1383 if (IS_ERR(server->mnt_path)) {
1384 sb = (struct super_block *)server->mnt_path;
1387 sb = sget(&nfs4_fs_type, nfs4_compare_super, nfs_set_super, server);
1388 if (IS_ERR(sb) || sb->s_root)
1390 nfs4_server_capabilities(server, &server->fh);
1392 down_write(&clp->cl_sem);
1393 atomic_inc(&clp->cl_count);
1394 list_add_tail(&server->nfs4_siblings, &clp->cl_superblocks);
1395 up_write(&clp->cl_sem);
1398 kfree(server->mnt_path);
1400 server->mnt_path = NULL;
1404 static struct super_block *nfs_clone_nfs4_sb(struct file_system_type *fs_type,
1405 int flags, const char *dev_name, void *raw_data)
1407 struct nfs_clone_mount *data = raw_data;
1408 return nfs_clone_generic_sb(data, nfs4_clone_sb, nfs_clone_server);
1411 static struct super_block *nfs4_referral_sb(struct nfs_server *server, struct nfs_clone_mount *data)
1413 struct super_block *sb = ERR_PTR(-ENOMEM);
1416 len = strlen(data->mnt_path) + 1;
1417 server->mnt_path = kmalloc(len, GFP_KERNEL);
1418 if (server->mnt_path == NULL)
1420 memcpy(server->mnt_path, data->mnt_path, len);
1421 memcpy(&server->addr, data->addr, sizeof(struct sockaddr_in));
1423 sb = sget(&nfs4_fs_type, nfs4_compare_super, nfs_set_super, server);
1424 if (IS_ERR(sb) || sb->s_root)
1428 kfree(server->mnt_path);
1430 server->mnt_path = NULL;
1434 static struct nfs_server *nfs4_referral_server(struct super_block *sb, struct nfs_clone_mount *data)
1436 struct nfs_server *server = NFS_SB(sb);
1437 struct rpc_timeout timeparms;
1438 int proto, timeo, retrans;
1441 proto = IPPROTO_TCP;
1442 /* Since we are following a referral and there may be alternatives,
1443 set the timeouts and retries to low values */
1446 nfs_init_timeout_values(&timeparms, proto, timeo, retrans);
1448 server->client = nfs4_create_client(server, &timeparms, proto, data->authflavor);
1449 if (IS_ERR((err = server->client)))
1452 sb->s_time_gran = 1;
1453 sb->s_op = &nfs4_sops;
1454 err = ERR_PTR(nfs_sb_init(sb, data->authflavor));
1458 return (struct nfs_server *)err;
1461 static struct super_block *nfs_referral_nfs4_sb(struct file_system_type *fs_type,
1462 int flags, const char *dev_name, void *raw_data)
1464 struct nfs_clone_mount *data = raw_data;
1465 return nfs_clone_generic_sb(data, nfs4_referral_sb, nfs4_referral_server);