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;
321 static const char *nfs_pseudoflavour_to_name(rpc_authflavor_t flavour)
324 rpc_authflavor_t flavour;
327 { RPC_AUTH_NULL, "null" },
328 { RPC_AUTH_UNIX, "sys" },
329 { RPC_AUTH_GSS_KRB5, "krb5" },
330 { RPC_AUTH_GSS_KRB5I, "krb5i" },
331 { RPC_AUTH_GSS_KRB5P, "krb5p" },
332 { RPC_AUTH_GSS_LKEY, "lkey" },
333 { RPC_AUTH_GSS_LKEYI, "lkeyi" },
334 { RPC_AUTH_GSS_LKEYP, "lkeyp" },
335 { RPC_AUTH_GSS_SPKM, "spkm" },
336 { RPC_AUTH_GSS_SPKMI, "spkmi" },
337 { RPC_AUTH_GSS_SPKMP, "spkmp" },
342 for (i=0; sec_flavours[i].flavour != -1; i++) {
343 if (sec_flavours[i].flavour == flavour)
346 return sec_flavours[i].str;
350 * Describe the mount options in force on this server representation
352 static void nfs_show_mount_options(struct seq_file *m, struct nfs_server *nfss, int showdefaults)
354 static struct proc_nfs_info {
359 { NFS_MOUNT_SOFT, ",soft", ",hard" },
360 { NFS_MOUNT_INTR, ",intr", "" },
361 { NFS_MOUNT_NOCTO, ",nocto", "" },
362 { NFS_MOUNT_NOAC, ",noac", "" },
363 { NFS_MOUNT_NONLM, ",nolock", "" },
364 { NFS_MOUNT_NOACL, ",noacl", "" },
367 struct proc_nfs_info *nfs_infop;
371 seq_printf(m, ",vers=%d", nfss->rpc_ops->version);
372 seq_printf(m, ",rsize=%d", nfss->rsize);
373 seq_printf(m, ",wsize=%d", nfss->wsize);
374 if (nfss->acregmin != 3*HZ || showdefaults)
375 seq_printf(m, ",acregmin=%d", nfss->acregmin/HZ);
376 if (nfss->acregmax != 60*HZ || showdefaults)
377 seq_printf(m, ",acregmax=%d", nfss->acregmax/HZ);
378 if (nfss->acdirmin != 30*HZ || showdefaults)
379 seq_printf(m, ",acdirmin=%d", nfss->acdirmin/HZ);
380 if (nfss->acdirmax != 60*HZ || showdefaults)
381 seq_printf(m, ",acdirmax=%d", nfss->acdirmax/HZ);
382 for (nfs_infop = nfs_info; nfs_infop->flag; nfs_infop++) {
383 if (nfss->flags & nfs_infop->flag)
384 seq_puts(m, nfs_infop->str);
386 seq_puts(m, nfs_infop->nostr);
388 switch (nfss->client->cl_xprt->prot) {
396 snprintf(buf, sizeof(buf), "%u", nfss->client->cl_xprt->prot);
399 seq_printf(m, ",proto=%s", proto);
400 seq_printf(m, ",timeo=%lu", 10U * nfss->retrans_timeo / HZ);
401 seq_printf(m, ",retrans=%u", nfss->retrans_count);
402 seq_printf(m, ",sec=%s", nfs_pseudoflavour_to_name(nfss->client->cl_auth->au_flavor));
406 * Describe the mount options on this VFS mountpoint
408 static int nfs_show_options(struct seq_file *m, struct vfsmount *mnt)
410 struct nfs_server *nfss = NFS_SB(mnt->mnt_sb);
412 nfs_show_mount_options(m, nfss, 0);
414 seq_puts(m, ",addr=");
415 seq_escape(m, nfss->hostname, " \t\n\\");
421 * Present statistical information for this VFS mountpoint
423 static int nfs_show_stats(struct seq_file *m, struct vfsmount *mnt)
426 struct nfs_server *nfss = NFS_SB(mnt->mnt_sb);
427 struct rpc_auth *auth = nfss->client->cl_auth;
428 struct nfs_iostats totals = { };
430 seq_printf(m, "statvers=%s", NFS_IOSTAT_VERS);
433 * Display all mount option settings
435 seq_printf(m, "\n\topts:\t");
436 seq_puts(m, mnt->mnt_sb->s_flags & MS_RDONLY ? "ro" : "rw");
437 seq_puts(m, mnt->mnt_sb->s_flags & MS_SYNCHRONOUS ? ",sync" : "");
438 seq_puts(m, mnt->mnt_sb->s_flags & MS_NOATIME ? ",noatime" : "");
439 seq_puts(m, mnt->mnt_sb->s_flags & MS_NODIRATIME ? ",nodiratime" : "");
440 nfs_show_mount_options(m, nfss, 1);
442 seq_printf(m, "\n\tage:\t%lu", (jiffies - nfss->mount_time) / HZ);
444 seq_printf(m, "\n\tcaps:\t");
445 seq_printf(m, "caps=0x%x", nfss->caps);
446 seq_printf(m, ",wtmult=%d", nfss->wtmult);
447 seq_printf(m, ",dtsize=%d", nfss->dtsize);
448 seq_printf(m, ",bsize=%d", nfss->bsize);
449 seq_printf(m, ",namelen=%d", nfss->namelen);
452 if (nfss->rpc_ops->version == 4) {
453 seq_printf(m, "\n\tnfsv4:\t");
454 seq_printf(m, "bm0=0x%x", nfss->attr_bitmask[0]);
455 seq_printf(m, ",bm1=0x%x", nfss->attr_bitmask[1]);
456 seq_printf(m, ",acl=0x%x", nfss->acl_bitmask);
461 * Display security flavor in effect for this mount
463 seq_printf(m, "\n\tsec:\tflavor=%d", auth->au_ops->au_flavor);
465 seq_printf(m, ",pseudoflavor=%d", auth->au_flavor);
468 * Display superblock I/O counters
470 for_each_possible_cpu(cpu) {
471 struct nfs_iostats *stats;
474 stats = per_cpu_ptr(nfss->io_stats, cpu);
476 for (i = 0; i < __NFSIOS_COUNTSMAX; i++)
477 totals.events[i] += stats->events[i];
478 for (i = 0; i < __NFSIOS_BYTESMAX; i++)
479 totals.bytes[i] += stats->bytes[i];
484 seq_printf(m, "\n\tevents:\t");
485 for (i = 0; i < __NFSIOS_COUNTSMAX; i++)
486 seq_printf(m, "%lu ", totals.events[i]);
487 seq_printf(m, "\n\tbytes:\t");
488 for (i = 0; i < __NFSIOS_BYTESMAX; i++)
489 seq_printf(m, "%Lu ", totals.bytes[i]);
492 rpc_print_iostats(m, nfss->client);
498 * Begin unmount by attempting to remove all automounted mountpoints we added
499 * in response to traversals
501 static void nfs_umount_begin(struct vfsmount *vfsmnt, int flags)
503 struct nfs_server *server;
504 struct rpc_clnt *rpc;
506 shrink_submounts(vfsmnt, &nfs_automount_list);
507 if (!(flags & MNT_FORCE))
509 /* -EIO all pending I/O */
510 server = NFS_SB(vfsmnt->mnt_sb);
511 rpc = server->client;
513 rpc_killall_tasks(rpc);
514 rpc = server->client_acl;
516 rpc_killall_tasks(rpc);
520 * Obtain the root inode of the file system.
522 static struct inode *
523 nfs_get_root(struct super_block *sb, struct nfs_fh *rootfh, struct nfs_fsinfo *fsinfo)
525 struct nfs_server *server = NFS_SB(sb);
528 error = server->rpc_ops->getroot(server, rootfh, fsinfo);
530 dprintk("nfs_get_root: getattr error = %d\n", -error);
531 return ERR_PTR(error);
534 server->fsid = fsinfo->fattr->fsid;
535 return nfs_fhget(sb, rootfh, fsinfo->fattr);
539 * Do NFS version-independent mount processing, and sanity checking
542 nfs_sb_init(struct super_block *sb, rpc_authflavor_t authflavor)
544 struct nfs_server *server;
545 struct inode *root_inode;
546 struct nfs_fattr fattr;
547 struct nfs_fsinfo fsinfo = {
550 struct nfs_pathconf pathinfo = {
553 int no_root_error = 0;
554 unsigned long max_rpc_payload;
556 /* We probably want something more informative here */
557 snprintf(sb->s_id, sizeof(sb->s_id), "%x:%x", MAJOR(sb->s_dev), MINOR(sb->s_dev));
561 sb->s_magic = NFS_SUPER_MAGIC;
563 server->io_stats = nfs_alloc_iostats();
564 if (server->io_stats == NULL)
567 root_inode = nfs_get_root(sb, &server->fh, &fsinfo);
568 /* Did getting the root inode fail? */
569 if (IS_ERR(root_inode)) {
570 no_root_error = PTR_ERR(root_inode);
573 sb->s_root = d_alloc_root(root_inode);
575 no_root_error = -ENOMEM;
578 sb->s_root->d_op = server->rpc_ops->dentry_ops;
580 /* mount time stamp, in seconds */
581 server->mount_time = jiffies;
583 /* Get some general file system info */
584 if (server->namelen == 0 &&
585 server->rpc_ops->pathconf(server, &server->fh, &pathinfo) >= 0)
586 server->namelen = pathinfo.max_namelen;
587 /* Work out a lot of parameters */
588 if (server->rsize == 0)
589 server->rsize = nfs_block_size(fsinfo.rtpref, NULL);
590 if (server->wsize == 0)
591 server->wsize = nfs_block_size(fsinfo.wtpref, NULL);
593 if (fsinfo.rtmax >= 512 && server->rsize > fsinfo.rtmax)
594 server->rsize = nfs_block_size(fsinfo.rtmax, NULL);
595 if (fsinfo.wtmax >= 512 && server->wsize > fsinfo.wtmax)
596 server->wsize = nfs_block_size(fsinfo.wtmax, NULL);
598 max_rpc_payload = nfs_block_size(rpc_max_payload(server->client), NULL);
599 if (server->rsize > max_rpc_payload)
600 server->rsize = max_rpc_payload;
601 if (server->rsize > NFS_MAX_FILE_IO_SIZE)
602 server->rsize = NFS_MAX_FILE_IO_SIZE;
603 server->rpages = (server->rsize + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT;
605 if (server->wsize > max_rpc_payload)
606 server->wsize = max_rpc_payload;
607 if (server->wsize > NFS_MAX_FILE_IO_SIZE)
608 server->wsize = NFS_MAX_FILE_IO_SIZE;
609 server->wpages = (server->wsize + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT;
611 if (sb->s_blocksize == 0)
612 sb->s_blocksize = nfs_block_bits(server->wsize,
613 &sb->s_blocksize_bits);
614 server->wtmult = nfs_block_bits(fsinfo.wtmult, NULL);
616 server->dtsize = nfs_block_size(fsinfo.dtpref, NULL);
617 if (server->dtsize > PAGE_CACHE_SIZE)
618 server->dtsize = PAGE_CACHE_SIZE;
619 if (server->dtsize > server->rsize)
620 server->dtsize = server->rsize;
622 if (server->flags & NFS_MOUNT_NOAC) {
623 server->acregmin = server->acregmax = 0;
624 server->acdirmin = server->acdirmax = 0;
625 sb->s_flags |= MS_SYNCHRONOUS;
627 server->backing_dev_info.ra_pages = server->rpages * NFS_MAX_READAHEAD;
629 nfs_super_set_maxbytes(sb, fsinfo.maxfilesize);
631 server->client->cl_intr = (server->flags & NFS_MOUNT_INTR) ? 1 : 0;
632 server->client->cl_softrtry = (server->flags & NFS_MOUNT_SOFT) ? 1 : 0;
634 /* We're airborne Set socket buffersize */
635 rpc_setbufsize(server->client, server->wsize + 100, server->rsize + 100);
637 /* Yargs. It didn't work out. */
639 dprintk("nfs_sb_init: get root inode failed: errno %d\n", -no_root_error);
640 if (!IS_ERR(root_inode))
642 return no_root_error;
646 * Initialise the timeout values for a connection
648 static void nfs_init_timeout_values(struct rpc_timeout *to, int proto, unsigned int timeo, unsigned int retrans)
650 to->to_initval = timeo * HZ / 10;
651 to->to_retries = retrans;
658 to->to_initval = 60 * HZ;
659 if (to->to_initval > NFS_MAX_TCP_TIMEOUT)
660 to->to_initval = NFS_MAX_TCP_TIMEOUT;
661 to->to_increment = to->to_initval;
662 to->to_maxval = to->to_initval + (to->to_increment * to->to_retries);
663 to->to_exponential = 0;
668 to->to_initval = 11 * HZ / 10;
669 if (to->to_initval > NFS_MAX_UDP_TIMEOUT)
670 to->to_initval = NFS_MAX_UDP_TIMEOUT;
671 to->to_maxval = NFS_MAX_UDP_TIMEOUT;
672 to->to_exponential = 1;
678 * Create an RPC client handle.
680 static struct rpc_clnt *
681 nfs_create_client(struct nfs_server *server, const struct nfs_mount_data *data)
683 struct rpc_timeout timeparms;
684 struct rpc_xprt *xprt = NULL;
685 struct rpc_clnt *clnt = NULL;
686 int proto = (data->flags & NFS_MOUNT_TCP) ? IPPROTO_TCP : IPPROTO_UDP;
688 nfs_init_timeout_values(&timeparms, proto, data->timeo, data->retrans);
690 server->retrans_timeo = timeparms.to_initval;
691 server->retrans_count = timeparms.to_retries;
693 /* create transport and client */
694 xprt = xprt_create_proto(proto, &server->addr, &timeparms);
696 dprintk("%s: cannot create RPC transport. Error = %ld\n",
697 __FUNCTION__, PTR_ERR(xprt));
698 return (struct rpc_clnt *)xprt;
700 clnt = rpc_create_client(xprt, server->hostname, &nfs_program,
701 server->rpc_ops->version, data->pseudoflavor);
703 dprintk("%s: cannot create RPC client. Error = %ld\n",
704 __FUNCTION__, PTR_ERR(xprt));
709 clnt->cl_softrtry = 1;
718 * Clone a server record
720 static struct nfs_server *nfs_clone_server(struct super_block *sb, struct nfs_clone_mount *data)
722 struct nfs_server *server = NFS_SB(sb);
723 struct nfs_server *parent = NFS_SB(data->sb);
724 struct inode *root_inode;
725 struct nfs_fsinfo fsinfo;
726 void *err = ERR_PTR(-ENOMEM);
728 sb->s_op = data->sb->s_op;
729 sb->s_blocksize = data->sb->s_blocksize;
730 sb->s_blocksize_bits = data->sb->s_blocksize_bits;
731 sb->s_maxbytes = data->sb->s_maxbytes;
733 server->client_sys = server->client_acl = ERR_PTR(-EINVAL);
734 server->io_stats = nfs_alloc_iostats();
735 if (server->io_stats == NULL)
738 server->client = rpc_clone_client(parent->client);
739 if (IS_ERR((err = server->client)))
742 if (!IS_ERR(parent->client_sys)) {
743 server->client_sys = rpc_clone_client(parent->client_sys);
744 if (IS_ERR((err = server->client_sys)))
747 if (!IS_ERR(parent->client_acl)) {
748 server->client_acl = rpc_clone_client(parent->client_acl);
749 if (IS_ERR((err = server->client_acl)))
752 root_inode = nfs_fhget(sb, data->fh, data->fattr);
755 sb->s_root = d_alloc_root(root_inode);
758 fsinfo.fattr = data->fattr;
759 if (NFS_PROTO(root_inode)->fsinfo(server, data->fh, &fsinfo) == 0)
760 nfs_super_set_maxbytes(sb, fsinfo.maxfilesize);
761 sb->s_root->d_op = server->rpc_ops->dentry_ops;
762 sb->s_flags |= MS_ACTIVE;
771 * Copy an existing superblock and attach revised data
773 static struct super_block *nfs_clone_generic_sb(struct nfs_clone_mount *data,
774 struct super_block *(*fill_sb)(struct nfs_server *, struct nfs_clone_mount *),
775 struct nfs_server *(*fill_server)(struct super_block *, struct nfs_clone_mount *))
777 struct nfs_server *server;
778 struct nfs_server *parent = NFS_SB(data->sb);
779 struct super_block *sb = ERR_PTR(-EINVAL);
780 void *err = ERR_PTR(-ENOMEM);
784 server = kmalloc(sizeof(struct nfs_server), GFP_KERNEL);
787 memcpy(server, parent, sizeof(*server));
788 hostname = (data->hostname != NULL) ? data->hostname : parent->hostname;
789 len = strlen(hostname) + 1;
790 server->hostname = kmalloc(len, GFP_KERNEL);
791 if (server->hostname == NULL)
793 memcpy(server->hostname, hostname, len);
794 if (rpciod_up() != 0)
797 sb = fill_sb(server, data);
798 if (IS_ERR((err = sb)) || sb->s_root)
801 server = fill_server(sb, data);
802 if (IS_ERR((err = server)))
806 up_write(&sb->s_umount);
807 deactivate_super(sb);
808 return (struct super_block *)err;
812 kfree(server->hostname);
816 return (struct super_block *)err;
820 * Set up an NFS2/3 superblock
822 * The way this works is that the mount process passes a structure
823 * in the data argument which contains the server's IP address
824 * and the root file handle obtained from the server's mount
825 * daemon. We stash these away in the private superblock fields.
828 nfs_fill_super(struct super_block *sb, struct nfs_mount_data *data, int silent)
830 struct nfs_server *server;
831 rpc_authflavor_t authflavor;
834 sb->s_blocksize_bits = 0;
837 sb->s_blocksize = nfs_block_size(data->bsize, &sb->s_blocksize_bits);
839 server->rsize = nfs_block_size(data->rsize, NULL);
841 server->wsize = nfs_block_size(data->wsize, NULL);
842 server->flags = data->flags & NFS_MOUNT_FLAGMASK;
844 server->acregmin = data->acregmin*HZ;
845 server->acregmax = data->acregmax*HZ;
846 server->acdirmin = data->acdirmin*HZ;
847 server->acdirmax = data->acdirmax*HZ;
849 /* Start lockd here, before we might error out */
850 if (!(server->flags & NFS_MOUNT_NONLM))
853 server->namelen = data->namlen;
854 server->hostname = kmalloc(strlen(data->hostname) + 1, GFP_KERNEL);
855 if (!server->hostname)
857 strcpy(server->hostname, data->hostname);
859 /* Check NFS protocol revision and initialize RPC op vector
860 * and file handle pool. */
862 if (server->flags & NFS_MOUNT_VER3) {
863 server->rpc_ops = &nfs_v3_clientops;
864 server->caps |= NFS_CAP_READDIRPLUS;
866 server->rpc_ops = &nfs_v2_clientops;
869 server->rpc_ops = &nfs_v2_clientops;
872 /* Fill in pseudoflavor for mount version < 5 */
873 if (!(data->flags & NFS_MOUNT_SECFLAVOUR))
874 data->pseudoflavor = RPC_AUTH_UNIX;
875 authflavor = data->pseudoflavor; /* save for sb_init() */
876 /* XXX maybe we want to add a server->pseudoflavor field */
878 /* Create RPC client handles */
879 server->client = nfs_create_client(server, data);
880 if (IS_ERR(server->client))
881 return PTR_ERR(server->client);
882 /* RFC 2623, sec 2.3.2 */
883 if (authflavor != RPC_AUTH_UNIX) {
884 struct rpc_auth *auth;
886 server->client_sys = rpc_clone_client(server->client);
887 if (IS_ERR(server->client_sys))
888 return PTR_ERR(server->client_sys);
889 auth = rpcauth_create(RPC_AUTH_UNIX, server->client_sys);
891 return PTR_ERR(auth);
893 atomic_inc(&server->client->cl_count);
894 server->client_sys = server->client;
896 if (server->flags & NFS_MOUNT_VER3) {
897 #ifdef CONFIG_NFS_V3_ACL
898 if (!(server->flags & NFS_MOUNT_NOACL)) {
899 server->client_acl = rpc_bind_new_program(server->client, &nfsacl_program, 3);
900 /* No errors! Assume that Sun nfsacls are supported */
901 if (!IS_ERR(server->client_acl))
902 server->caps |= NFS_CAP_ACLS;
905 server->flags &= ~NFS_MOUNT_NOACL;
906 #endif /* CONFIG_NFS_V3_ACL */
908 * The VFS shouldn't apply the umask to mode bits. We will
909 * do so ourselves when necessary.
911 sb->s_flags |= MS_POSIXACL;
912 if (server->namelen == 0 || server->namelen > NFS3_MAXNAMLEN)
913 server->namelen = NFS3_MAXNAMLEN;
916 if (server->namelen == 0 || server->namelen > NFS2_MAXNAMLEN)
917 server->namelen = NFS2_MAXNAMLEN;
920 sb->s_op = &nfs_sops;
921 return nfs_sb_init(sb, authflavor);
924 static int nfs_set_super(struct super_block *s, void *data)
927 return set_anon_super(s, data);
930 static int nfs_compare_super(struct super_block *sb, void *data)
932 struct nfs_server *server = data;
933 struct nfs_server *old = NFS_SB(sb);
935 if (old->addr.sin_addr.s_addr != server->addr.sin_addr.s_addr)
937 if (old->addr.sin_port != server->addr.sin_port)
939 return !nfs_compare_fh(&old->fh, &server->fh);
942 static struct super_block *nfs_get_sb(struct file_system_type *fs_type,
943 int flags, const char *dev_name, void *raw_data)
946 struct nfs_server *server = NULL;
947 struct super_block *s;
949 struct nfs_mount_data *data = raw_data;
951 s = ERR_PTR(-EINVAL);
953 dprintk("%s: missing data argument\n", __FUNCTION__);
956 if (data->version <= 0 || data->version > NFS_MOUNT_VERSION) {
957 dprintk("%s: bad mount version\n", __FUNCTION__);
960 switch (data->version) {
966 if (data->flags & NFS_MOUNT_VER3) {
967 dprintk("%s: mount structure version %d does not support NFSv3\n",
972 data->root.size = NFS2_FHSIZE;
973 memcpy(data->root.data, data->old_root.data, NFS2_FHSIZE);
975 if (data->flags & NFS_MOUNT_SECFLAVOUR) {
976 dprintk("%s: mount structure version %d does not support strong security\n",
982 memset(data->context, 0, sizeof(data->context));
984 #ifndef CONFIG_NFS_V3
985 /* If NFSv3 is not compiled in, return -EPROTONOSUPPORT */
986 s = ERR_PTR(-EPROTONOSUPPORT);
987 if (data->flags & NFS_MOUNT_VER3) {
988 dprintk("%s: NFSv3 not compiled into kernel\n", __FUNCTION__);
991 #endif /* CONFIG_NFS_V3 */
993 s = ERR_PTR(-ENOMEM);
994 server = kzalloc(sizeof(struct nfs_server), GFP_KERNEL);
997 /* Zero out the NFS state stuff */
998 init_nfsv4_state(server);
999 server->client = server->client_sys = server->client_acl = ERR_PTR(-EINVAL);
1002 if (data->flags & NFS_MOUNT_VER3)
1003 root->size = data->root.size;
1005 root->size = NFS2_FHSIZE;
1006 s = ERR_PTR(-EINVAL);
1007 if (root->size > sizeof(root->data)) {
1008 dprintk("%s: invalid root filehandle\n", __FUNCTION__);
1011 memcpy(root->data, data->root.data, root->size);
1013 /* We now require that the mount process passes the remote address */
1014 memcpy(&server->addr, &data->addr, sizeof(server->addr));
1015 if (server->addr.sin_addr.s_addr == INADDR_ANY) {
1016 dprintk("%s: mount program didn't pass remote address!\n",
1021 /* Fire up rpciod if not yet running */
1022 s = ERR_PTR(rpciod_up());
1024 dprintk("%s: couldn't start rpciod! Error = %ld\n",
1025 __FUNCTION__, PTR_ERR(s));
1029 s = sget(fs_type, nfs_compare_super, nfs_set_super, server);
1030 if (IS_ERR(s) || s->s_root)
1031 goto out_rpciod_down;
1035 error = nfs_fill_super(s, data, flags & MS_SILENT ? 1 : 0);
1037 up_write(&s->s_umount);
1038 deactivate_super(s);
1039 return ERR_PTR(error);
1041 s->s_flags |= MS_ACTIVE;
1050 static void nfs_kill_super(struct super_block *s)
1052 struct nfs_server *server = NFS_SB(s);
1056 if (!IS_ERR(server->client))
1057 rpc_shutdown_client(server->client);
1058 if (!IS_ERR(server->client_sys))
1059 rpc_shutdown_client(server->client_sys);
1060 if (!IS_ERR(server->client_acl))
1061 rpc_shutdown_client(server->client_acl);
1063 if (!(server->flags & NFS_MOUNT_NONLM))
1064 lockd_down(); /* release rpc.lockd */
1066 rpciod_down(); /* release rpciod */
1068 nfs_free_iostats(server->io_stats);
1069 kfree(server->hostname);
1071 nfs_release_automount_timer();
1074 static struct super_block *nfs_clone_sb(struct nfs_server *server, struct nfs_clone_mount *data)
1076 struct super_block *sb;
1078 server->fsid = data->fattr->fsid;
1079 nfs_copy_fh(&server->fh, data->fh);
1080 sb = sget(&nfs_fs_type, nfs_compare_super, nfs_set_super, server);
1081 if (!IS_ERR(sb) && sb->s_root == NULL && !(server->flags & NFS_MOUNT_NONLM))
1086 static struct super_block *nfs_clone_nfs_sb(struct file_system_type *fs_type,
1087 int flags, const char *dev_name, void *raw_data)
1089 struct nfs_clone_mount *data = raw_data;
1090 return nfs_clone_generic_sb(data, nfs_clone_sb, nfs_clone_server);
1093 #ifdef CONFIG_NFS_V4
1094 static struct rpc_clnt *nfs4_create_client(struct nfs_server *server,
1095 struct rpc_timeout *timeparms, int proto, rpc_authflavor_t flavor)
1097 struct nfs4_client *clp;
1098 struct rpc_xprt *xprt = NULL;
1099 struct rpc_clnt *clnt = NULL;
1102 clp = nfs4_get_client(&server->addr.sin_addr);
1104 dprintk("%s: failed to create NFS4 client.\n", __FUNCTION__);
1105 return ERR_PTR(err);
1108 /* Now create transport and client */
1109 down_write(&clp->cl_sem);
1110 if (IS_ERR(clp->cl_rpcclient)) {
1111 xprt = xprt_create_proto(proto, &server->addr, timeparms);
1113 up_write(&clp->cl_sem);
1114 err = PTR_ERR(xprt);
1115 dprintk("%s: cannot create RPC transport. Error = %d\n",
1119 /* Bind to a reserved port! */
1121 clnt = rpc_create_client(xprt, server->hostname, &nfs_program,
1122 server->rpc_ops->version, flavor);
1124 up_write(&clp->cl_sem);
1125 err = PTR_ERR(clnt);
1126 dprintk("%s: cannot create RPC client. Error = %d\n",
1131 clnt->cl_softrtry = 1;
1132 clp->cl_rpcclient = clnt;
1133 memcpy(clp->cl_ipaddr, server->ip_addr, sizeof(clp->cl_ipaddr));
1136 list_add_tail(&server->nfs4_siblings, &clp->cl_superblocks);
1137 clnt = rpc_clone_client(clp->cl_rpcclient);
1139 server->nfs4_state = clp;
1140 up_write(&clp->cl_sem);
1144 dprintk("%s: cannot create RPC client. Error = %d\n",
1149 if (server->nfs4_state->cl_idmap == NULL) {
1150 dprintk("%s: failed to create idmapper.\n", __FUNCTION__);
1151 return ERR_PTR(-ENOMEM);
1154 if (clnt->cl_auth->au_flavor != flavor) {
1155 struct rpc_auth *auth;
1157 auth = rpcauth_create(flavor, clnt);
1159 dprintk("%s: couldn't create credcache!\n", __FUNCTION__);
1160 return (struct rpc_clnt *)auth;
1167 nfs4_put_client(clp);
1168 return ERR_PTR(err);
1172 * Set up an NFS4 superblock
1174 static int nfs4_fill_super(struct super_block *sb, struct nfs4_mount_data *data, int silent)
1176 struct nfs_server *server;
1177 struct rpc_timeout timeparms;
1178 rpc_authflavor_t authflavour;
1181 sb->s_blocksize_bits = 0;
1182 sb->s_blocksize = 0;
1183 server = NFS_SB(sb);
1184 if (data->rsize != 0)
1185 server->rsize = nfs_block_size(data->rsize, NULL);
1186 if (data->wsize != 0)
1187 server->wsize = nfs_block_size(data->wsize, NULL);
1188 server->flags = data->flags & NFS_MOUNT_FLAGMASK;
1189 server->caps = NFS_CAP_ATOMIC_OPEN;
1191 server->acregmin = data->acregmin*HZ;
1192 server->acregmax = data->acregmax*HZ;
1193 server->acdirmin = data->acdirmin*HZ;
1194 server->acdirmax = data->acdirmax*HZ;
1196 server->rpc_ops = &nfs_v4_clientops;
1198 nfs_init_timeout_values(&timeparms, data->proto, data->timeo, data->retrans);
1200 server->retrans_timeo = timeparms.to_initval;
1201 server->retrans_count = timeparms.to_retries;
1203 /* Now create transport and client */
1204 authflavour = RPC_AUTH_UNIX;
1205 if (data->auth_flavourlen != 0) {
1206 if (data->auth_flavourlen != 1) {
1207 dprintk("%s: Invalid number of RPC auth flavours %d.\n",
1208 __FUNCTION__, data->auth_flavourlen);
1212 if (copy_from_user(&authflavour, data->auth_flavours, sizeof(authflavour))) {
1218 server->client = nfs4_create_client(server, &timeparms, data->proto, authflavour);
1219 if (IS_ERR(server->client)) {
1220 err = PTR_ERR(server->client);
1221 dprintk("%s: cannot create RPC client. Error = %d\n",
1226 sb->s_time_gran = 1;
1228 sb->s_op = &nfs4_sops;
1229 err = nfs_sb_init(sb, authflavour);
1235 static int nfs4_compare_super(struct super_block *sb, void *data)
1237 struct nfs_server *server = data;
1238 struct nfs_server *old = NFS_SB(sb);
1240 if (strcmp(server->hostname, old->hostname) != 0)
1242 if (strcmp(server->mnt_path, old->mnt_path) != 0)
1248 nfs_copy_user_string(char *dst, struct nfs_string *src, int maxlen)
1253 return ERR_PTR(-EINVAL);
1254 if (src->len < maxlen)
1257 p = dst = kmalloc(maxlen + 1, GFP_KERNEL);
1259 return ERR_PTR(-ENOMEM);
1261 if (copy_from_user(dst, src->data, maxlen)) {
1263 return ERR_PTR(-EFAULT);
1269 static struct super_block *nfs4_get_sb(struct file_system_type *fs_type,
1270 int flags, const char *dev_name, void *raw_data)
1273 struct nfs_server *server;
1274 struct super_block *s;
1275 struct nfs4_mount_data *data = raw_data;
1279 dprintk("%s: missing data argument\n", __FUNCTION__);
1280 return ERR_PTR(-EINVAL);
1282 if (data->version <= 0 || data->version > NFS4_MOUNT_VERSION) {
1283 dprintk("%s: bad mount version\n", __FUNCTION__);
1284 return ERR_PTR(-EINVAL);
1287 server = kzalloc(sizeof(struct nfs_server), GFP_KERNEL);
1289 return ERR_PTR(-ENOMEM);
1290 /* Zero out the NFS state stuff */
1291 init_nfsv4_state(server);
1292 server->client = server->client_sys = server->client_acl = ERR_PTR(-EINVAL);
1294 p = nfs_copy_user_string(NULL, &data->hostname, 256);
1297 server->hostname = p;
1299 p = nfs_copy_user_string(NULL, &data->mnt_path, 1024);
1302 server->mnt_path = p;
1304 p = nfs_copy_user_string(server->ip_addr, &data->client_addr,
1305 sizeof(server->ip_addr) - 1);
1309 /* We now require that the mount process passes the remote address */
1310 if (data->host_addrlen != sizeof(server->addr)) {
1311 s = ERR_PTR(-EINVAL);
1314 if (copy_from_user(&server->addr, data->host_addr, sizeof(server->addr))) {
1315 s = ERR_PTR(-EFAULT);
1318 if (server->addr.sin_family != AF_INET ||
1319 server->addr.sin_addr.s_addr == INADDR_ANY) {
1320 dprintk("%s: mount program didn't pass remote IP address!\n",
1322 s = ERR_PTR(-EINVAL);
1326 /* Fire up rpciod if not yet running */
1327 s = ERR_PTR(rpciod_up());
1329 dprintk("%s: couldn't start rpciod! Error = %ld\n",
1330 __FUNCTION__, PTR_ERR(s));
1334 s = sget(fs_type, nfs4_compare_super, nfs_set_super, server);
1336 if (IS_ERR(s) || s->s_root)
1341 error = nfs4_fill_super(s, data, flags & MS_SILENT ? 1 : 0);
1343 up_write(&s->s_umount);
1344 deactivate_super(s);
1345 return ERR_PTR(error);
1347 s->s_flags |= MS_ACTIVE;
1350 s = (struct super_block *)p;
1352 kfree(server->mnt_path);
1353 kfree(server->hostname);
1358 static void nfs4_kill_super(struct super_block *sb)
1360 struct nfs_server *server = NFS_SB(sb);
1362 nfs_return_all_delegations(sb);
1363 kill_anon_super(sb);
1365 nfs4_renewd_prepare_shutdown(server);
1367 if (server->client != NULL && !IS_ERR(server->client))
1368 rpc_shutdown_client(server->client);
1370 destroy_nfsv4_state(server);
1374 nfs_free_iostats(server->io_stats);
1375 kfree(server->hostname);
1377 nfs_release_automount_timer();
1381 * Constructs the SERVER-side path
1383 static inline char *nfs4_dup_path(const struct dentry *dentry)
1385 char *page = (char *) __get_free_page(GFP_USER);
1388 path = nfs4_path(dentry, page, PAGE_SIZE);
1389 if (!IS_ERR(path)) {
1390 int len = PAGE_SIZE + page - path;
1393 path = kmalloc(len, GFP_KERNEL);
1395 memcpy(path, tmp, len);
1397 path = ERR_PTR(-ENOMEM);
1399 free_page((unsigned long)page);
1403 static struct super_block *nfs4_clone_sb(struct nfs_server *server, struct nfs_clone_mount *data)
1405 const struct dentry *dentry = data->dentry;
1406 struct nfs4_client *clp = server->nfs4_state;
1407 struct super_block *sb;
1409 server->fsid = data->fattr->fsid;
1410 nfs_copy_fh(&server->fh, data->fh);
1411 server->mnt_path = nfs4_dup_path(dentry);
1412 if (IS_ERR(server->mnt_path)) {
1413 sb = (struct super_block *)server->mnt_path;
1416 sb = sget(&nfs4_fs_type, nfs4_compare_super, nfs_set_super, server);
1417 if (IS_ERR(sb) || sb->s_root)
1419 nfs4_server_capabilities(server, &server->fh);
1421 down_write(&clp->cl_sem);
1422 atomic_inc(&clp->cl_count);
1423 list_add_tail(&server->nfs4_siblings, &clp->cl_superblocks);
1424 up_write(&clp->cl_sem);
1427 kfree(server->mnt_path);
1429 server->mnt_path = NULL;
1433 static struct super_block *nfs_clone_nfs4_sb(struct file_system_type *fs_type,
1434 int flags, const char *dev_name, void *raw_data)
1436 struct nfs_clone_mount *data = raw_data;
1437 return nfs_clone_generic_sb(data, nfs4_clone_sb, nfs_clone_server);
1440 static struct super_block *nfs4_referral_sb(struct nfs_server *server, struct nfs_clone_mount *data)
1442 struct super_block *sb = ERR_PTR(-ENOMEM);
1445 len = strlen(data->mnt_path) + 1;
1446 server->mnt_path = kmalloc(len, GFP_KERNEL);
1447 if (server->mnt_path == NULL)
1449 memcpy(server->mnt_path, data->mnt_path, len);
1450 memcpy(&server->addr, data->addr, sizeof(struct sockaddr_in));
1452 sb = sget(&nfs4_fs_type, nfs4_compare_super, nfs_set_super, server);
1453 if (IS_ERR(sb) || sb->s_root)
1457 kfree(server->mnt_path);
1459 server->mnt_path = NULL;
1463 static struct nfs_server *nfs4_referral_server(struct super_block *sb, struct nfs_clone_mount *data)
1465 struct nfs_server *server = NFS_SB(sb);
1466 struct rpc_timeout timeparms;
1467 int proto, timeo, retrans;
1470 proto = IPPROTO_TCP;
1471 /* Since we are following a referral and there may be alternatives,
1472 set the timeouts and retries to low values */
1475 nfs_init_timeout_values(&timeparms, proto, timeo, retrans);
1477 server->client = nfs4_create_client(server, &timeparms, proto, data->authflavor);
1478 if (IS_ERR((err = server->client)))
1481 sb->s_time_gran = 1;
1482 sb->s_op = &nfs4_sops;
1483 err = ERR_PTR(nfs_sb_init(sb, data->authflavor));
1487 return (struct nfs_server *)err;
1490 static struct super_block *nfs_referral_nfs4_sb(struct file_system_type *fs_type,
1491 int flags, const char *dev_name, void *raw_data)
1493 struct nfs_clone_mount *data = raw_data;
1494 return nfs_clone_generic_sb(data, nfs4_referral_sb, nfs4_referral_server);