1 // SPDX-License-Identifier: GPL-2.0-only
5 * Copyright (C) 1992 Rick Sladkey
7 * nfs superblock handling functions
9 * Modularised by Alan Cox <alan@lxorguk.ukuu.org.uk>, while hacking some
10 * experimental NFS changes. Modularisation taken straight from SYS5 fs.
12 * Change to nfs_read_super() to permit NFS mounts to multi-homed hosts.
13 * J.S.Peatfield@damtp.cam.ac.uk
15 * Split from inode.c by David Howells <dhowells@redhat.com>
17 * - superblocks are indexed on server only - all inodes, dentries, etc. associated with a
18 * particular server are held in the same superblock
19 * - NFS superblocks can have several effective roots to the dentry tree
20 * - directory type roots are spliced into the tree when a path from one root reaches the root
21 * of another (see nfs_lookup())
24 #include <linux/module.h>
25 #include <linux/init.h>
27 #include <linux/time.h>
28 #include <linux/kernel.h>
30 #include <linux/string.h>
31 #include <linux/stat.h>
32 #include <linux/errno.h>
33 #include <linux/unistd.h>
34 #include <linux/sunrpc/clnt.h>
35 #include <linux/sunrpc/addr.h>
36 #include <linux/sunrpc/stats.h>
37 #include <linux/sunrpc/metrics.h>
38 #include <linux/sunrpc/xprtsock.h>
39 #include <linux/sunrpc/xprtrdma.h>
40 #include <linux/nfs_fs.h>
41 #include <linux/nfs_mount.h>
42 #include <linux/nfs4_mount.h>
43 #include <linux/lockd/bind.h>
44 #include <linux/seq_file.h>
45 #include <linux/mount.h>
46 #include <linux/namei.h>
47 #include <linux/vfs.h>
48 #include <linux/inet.h>
49 #include <linux/in6.h>
50 #include <linux/sched.h>
51 #include <linux/slab.h>
53 #include <linux/netdevice.h>
54 #include <linux/nfs_xdr.h>
55 #include <linux/magic.h>
56 #include <linux/parser.h>
57 #include <linux/nsproxy.h>
58 #include <linux/rcupdate.h>
60 #include <linux/uaccess.h>
61 #include <linux/nfs_ssc.h>
63 #include <uapi/linux/tls.h>
67 #include "delegation.h"
71 #include "nfs4session.h"
76 #include "nfs4idmap.h"
78 #define NFSDBG_FACILITY NFSDBG_VFS
80 const struct super_operations nfs_sops = {
81 .alloc_inode = nfs_alloc_inode,
82 .free_inode = nfs_free_inode,
83 .write_inode = nfs_write_inode,
84 .drop_inode = nfs_drop_inode,
86 .evict_inode = nfs_evict_inode,
87 .umount_begin = nfs_umount_begin,
88 .show_options = nfs_show_options,
89 .show_devname = nfs_show_devname,
90 .show_path = nfs_show_path,
91 .show_stats = nfs_show_stats,
93 EXPORT_SYMBOL_GPL(nfs_sops);
95 #ifdef CONFIG_NFS_V4_2
96 static const struct nfs_ssc_client_ops nfs_ssc_clnt_ops_tbl = {
97 .sco_sb_deactive = nfs_sb_deactive,
101 #if IS_ENABLED(CONFIG_NFS_V4)
102 static int __init register_nfs4_fs(void)
104 return register_filesystem(&nfs4_fs_type);
107 static void unregister_nfs4_fs(void)
109 unregister_filesystem(&nfs4_fs_type);
112 static int __init register_nfs4_fs(void)
117 static void unregister_nfs4_fs(void)
122 #ifdef CONFIG_NFS_V4_2
123 static void nfs_ssc_register_ops(void)
125 nfs_ssc_register(&nfs_ssc_clnt_ops_tbl);
128 static void nfs_ssc_unregister_ops(void)
130 nfs_ssc_unregister(&nfs_ssc_clnt_ops_tbl);
132 #endif /* CONFIG_NFS_V4_2 */
134 static struct shrinker *acl_shrinker;
137 * Register the NFS filesystems
139 int __init register_nfs_fs(void)
143 ret = register_filesystem(&nfs_fs_type);
147 ret = register_nfs4_fs();
151 ret = nfs_register_sysctl();
155 acl_shrinker = shrinker_alloc(0, "nfs-acl");
161 acl_shrinker->count_objects = nfs_access_cache_count;
162 acl_shrinker->scan_objects = nfs_access_cache_scan;
164 shrinker_register(acl_shrinker);
166 #ifdef CONFIG_NFS_V4_2
167 nfs_ssc_register_ops();
171 nfs_unregister_sysctl();
173 unregister_nfs4_fs();
175 unregister_filesystem(&nfs_fs_type);
181 * Unregister the NFS filesystems
183 void __exit unregister_nfs_fs(void)
185 shrinker_free(acl_shrinker);
186 nfs_unregister_sysctl();
187 unregister_nfs4_fs();
188 #ifdef CONFIG_NFS_V4_2
189 nfs_ssc_unregister_ops();
191 unregister_filesystem(&nfs_fs_type);
194 bool nfs_sb_active(struct super_block *sb)
196 struct nfs_server *server = NFS_SB(sb);
198 if (!atomic_inc_not_zero(&sb->s_active))
200 if (atomic_inc_return(&server->active) != 1)
201 atomic_dec(&sb->s_active);
204 EXPORT_SYMBOL_GPL(nfs_sb_active);
206 void nfs_sb_deactive(struct super_block *sb)
208 struct nfs_server *server = NFS_SB(sb);
210 if (atomic_dec_and_test(&server->active))
211 deactivate_super(sb);
213 EXPORT_SYMBOL_GPL(nfs_sb_deactive);
215 static int __nfs_list_for_each_server(struct list_head *head,
216 int (*fn)(struct nfs_server *, void *),
219 struct nfs_server *server, *last = NULL;
223 list_for_each_entry_rcu(server, head, client_link) {
224 if (!(server->super && nfs_sb_active(server->super)))
228 nfs_sb_deactive(last->super);
230 ret = fn(server, data);
239 nfs_sb_deactive(last->super);
243 int nfs_client_for_each_server(struct nfs_client *clp,
244 int (*fn)(struct nfs_server *, void *),
247 return __nfs_list_for_each_server(&clp->cl_superblocks, fn, data);
249 EXPORT_SYMBOL_GPL(nfs_client_for_each_server);
252 * Deliver file system statistics to userspace
254 int nfs_statfs(struct dentry *dentry, struct kstatfs *buf)
256 struct nfs_server *server = NFS_SB(dentry->d_sb);
257 unsigned char blockbits;
258 unsigned long blockres;
259 struct nfs_fh *fh = NFS_FH(d_inode(dentry));
260 struct nfs_fsstat res;
263 res.fattr = nfs_alloc_fattr();
264 if (res.fattr == NULL)
267 error = server->nfs_client->rpc_ops->statfs(server, fh, &res);
268 if (unlikely(error == -ESTALE)) {
269 struct dentry *pd_dentry;
271 pd_dentry = dget_parent(dentry);
272 nfs_zap_caches(d_inode(pd_dentry));
275 nfs_free_fattr(res.fattr);
279 buf->f_type = NFS_SUPER_MAGIC;
282 * Current versions of glibc do not correctly handle the
283 * case where f_frsize != f_bsize. Eventually we want to
284 * report the value of wtmult in this field.
286 buf->f_frsize = dentry->d_sb->s_blocksize;
289 * On most *nix systems, f_blocks, f_bfree, and f_bavail
290 * are reported in units of f_frsize. Linux hasn't had
291 * an f_frsize field in its statfs struct until recently,
292 * thus historically Linux's sys_statfs reports these
293 * fields in units of f_bsize.
295 buf->f_bsize = dentry->d_sb->s_blocksize;
296 blockbits = dentry->d_sb->s_blocksize_bits;
297 blockres = (1 << blockbits) - 1;
298 buf->f_blocks = (res.tbytes + blockres) >> blockbits;
299 buf->f_bfree = (res.fbytes + blockres) >> blockbits;
300 buf->f_bavail = (res.abytes + blockres) >> blockbits;
302 buf->f_files = res.tfiles;
303 buf->f_ffree = res.afiles;
305 buf->f_namelen = server->namelen;
310 dprintk("%s: statfs error = %d\n", __func__, -error);
313 EXPORT_SYMBOL_GPL(nfs_statfs);
316 * Map the security flavour number to a name
318 static const char *nfs_pseudoflavour_to_name(rpc_authflavor_t flavour)
320 static const struct {
321 rpc_authflavor_t flavour;
323 } sec_flavours[NFS_AUTH_INFO_MAX_FLAVORS] = {
324 /* update NFS_AUTH_INFO_MAX_FLAVORS when this list changes! */
325 { RPC_AUTH_NULL, "null" },
326 { RPC_AUTH_UNIX, "sys" },
327 { RPC_AUTH_GSS_KRB5, "krb5" },
328 { RPC_AUTH_GSS_KRB5I, "krb5i" },
329 { RPC_AUTH_GSS_KRB5P, "krb5p" },
330 { RPC_AUTH_GSS_LKEY, "lkey" },
331 { RPC_AUTH_GSS_LKEYI, "lkeyi" },
332 { RPC_AUTH_GSS_LKEYP, "lkeyp" },
333 { RPC_AUTH_GSS_SPKM, "spkm" },
334 { RPC_AUTH_GSS_SPKMI, "spkmi" },
335 { RPC_AUTH_GSS_SPKMP, "spkmp" },
336 { UINT_MAX, "unknown" }
340 for (i = 0; sec_flavours[i].flavour != UINT_MAX; i++) {
341 if (sec_flavours[i].flavour == flavour)
344 return sec_flavours[i].str;
347 static void nfs_show_mountd_netid(struct seq_file *m, struct nfs_server *nfss,
350 struct sockaddr *sap = (struct sockaddr *) &nfss->mountd_address;
353 switch (sap->sa_family) {
355 switch (nfss->mountd_protocol) {
357 proto = RPCBIND_NETID_UDP;
360 proto = RPCBIND_NETID_TCP;
365 switch (nfss->mountd_protocol) {
367 proto = RPCBIND_NETID_UDP6;
370 proto = RPCBIND_NETID_TCP6;
375 if (proto || showdefaults)
376 seq_printf(m, ",mountproto=%s", proto ?: "auto");
379 static void nfs_show_mountd_options(struct seq_file *m, struct nfs_server *nfss,
382 struct sockaddr *sap = (struct sockaddr *)&nfss->mountd_address;
384 if (nfss->flags & NFS_MOUNT_LEGACY_INTERFACE)
387 switch (sap->sa_family) {
389 struct sockaddr_in *sin = (struct sockaddr_in *)sap;
390 seq_printf(m, ",mountaddr=%pI4", &sin->sin_addr.s_addr);
394 struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)sap;
395 seq_printf(m, ",mountaddr=%pI6c", &sin6->sin6_addr);
400 seq_puts(m, ",mountaddr=unspecified");
403 if (nfss->mountd_version || showdefaults)
404 seq_printf(m, ",mountvers=%u", nfss->mountd_version);
405 if ((nfss->mountd_port &&
406 nfss->mountd_port != (unsigned short)NFS_UNSPEC_PORT) ||
408 seq_printf(m, ",mountport=%u", nfss->mountd_port);
410 nfs_show_mountd_netid(m, nfss, showdefaults);
413 #if IS_ENABLED(CONFIG_NFS_V4)
414 static void nfs_show_nfsv4_options(struct seq_file *m, struct nfs_server *nfss,
417 struct nfs_client *clp = nfss->nfs_client;
419 seq_printf(m, ",clientaddr=%s", clp->cl_ipaddr);
422 static void nfs_show_nfsv4_options(struct seq_file *m, struct nfs_server *nfss,
428 static void nfs_show_nfs_version(struct seq_file *m,
429 unsigned int version,
430 unsigned int minorversion)
432 seq_printf(m, ",vers=%u", version);
434 seq_printf(m, ".%u", minorversion);
438 * Describe the mount options in force on this server representation
440 static void nfs_show_mount_options(struct seq_file *m, struct nfs_server *nfss,
443 static const struct proc_nfs_info {
448 { NFS_MOUNT_SOFT, ",soft", "" },
449 { NFS_MOUNT_SOFTERR, ",softerr", "" },
450 { NFS_MOUNT_SOFTREVAL, ",softreval", "" },
451 { NFS_MOUNT_POSIX, ",posix", "" },
452 { NFS_MOUNT_NOCTO, ",nocto", "" },
453 { NFS_MOUNT_NOAC, ",noac", "" },
454 { NFS_MOUNT_NONLM, ",nolock", "" },
455 { NFS_MOUNT_NOACL, ",noacl", "" },
456 { NFS_MOUNT_NORDIRPLUS, ",nordirplus", "" },
457 { NFS_MOUNT_FORCE_RDIRPLUS, ",rdirplus=force", "" },
458 { NFS_MOUNT_UNSHARED, ",nosharecache", "" },
459 { NFS_MOUNT_NORESVPORT, ",noresvport", "" },
460 { NFS_MOUNT_NETUNREACH_FATAL,
461 ",fatal_neterrors=ENETDOWN:ENETUNREACH",
462 ",fatal_neterrors=none" },
465 const struct proc_nfs_info *nfs_infop;
466 struct nfs_client *clp = nfss->nfs_client;
467 u32 version = clp->rpc_ops->version;
468 int local_flock, local_fcntl;
470 nfs_show_nfs_version(m, version, clp->cl_minorversion);
471 seq_printf(m, ",rsize=%u", nfss->rsize);
472 seq_printf(m, ",wsize=%u", nfss->wsize);
473 if (nfss->bsize != 0)
474 seq_printf(m, ",bsize=%u", nfss->bsize);
475 seq_printf(m, ",namlen=%u", nfss->namelen);
476 if (nfss->acregmin != NFS_DEF_ACREGMIN*HZ || showdefaults)
477 seq_printf(m, ",acregmin=%u", nfss->acregmin/HZ);
478 if (nfss->acregmax != NFS_DEF_ACREGMAX*HZ || showdefaults)
479 seq_printf(m, ",acregmax=%u", nfss->acregmax/HZ);
480 if (nfss->acdirmin != NFS_DEF_ACDIRMIN*HZ || showdefaults)
481 seq_printf(m, ",acdirmin=%u", nfss->acdirmin/HZ);
482 if (nfss->acdirmax != NFS_DEF_ACDIRMAX*HZ || showdefaults)
483 seq_printf(m, ",acdirmax=%u", nfss->acdirmax/HZ);
484 if (!(nfss->flags & (NFS_MOUNT_SOFT|NFS_MOUNT_SOFTERR)))
485 seq_puts(m, ",hard");
486 for (nfs_infop = nfs_info; nfs_infop->flag; nfs_infop++) {
487 if (nfss->flags & nfs_infop->flag)
488 seq_puts(m, nfs_infop->str);
490 seq_puts(m, nfs_infop->nostr);
493 seq_printf(m, ",proto=%s",
494 rpc_peeraddr2str(nfss->client, RPC_DISPLAY_NETID));
496 if (clp->cl_nconnect > 0)
497 seq_printf(m, ",nconnect=%u", clp->cl_nconnect);
499 if (clp->cl_max_connect > 1)
500 seq_printf(m, ",max_connect=%u", clp->cl_max_connect);
501 if (nfss->port != NFS_PORT)
502 seq_printf(m, ",port=%u", nfss->port);
505 seq_printf(m, ",port=%u", nfss->port);
507 seq_printf(m, ",timeo=%lu", 10U * nfss->client->cl_timeout->to_initval / HZ);
508 seq_printf(m, ",retrans=%u", nfss->client->cl_timeout->to_retries);
509 seq_printf(m, ",sec=%s", nfs_pseudoflavour_to_name(nfss->client->cl_auth->au_flavor));
510 switch (clp->cl_xprtsec.policy) {
511 case RPC_XPRTSEC_TLS_ANON:
512 seq_puts(m, ",xprtsec=tls");
514 case RPC_XPRTSEC_TLS_X509:
515 seq_puts(m, ",xprtsec=mtls");
522 nfs_show_mountd_options(m, nfss, showdefaults);
524 nfs_show_nfsv4_options(m, nfss, showdefaults);
526 if (nfss->options & NFS_OPTION_FSCACHE) {
527 #ifdef CONFIG_NFS_FSCACHE
528 if (nfss->fscache_uniq)
529 seq_printf(m, ",fsc=%s", nfss->fscache_uniq);
537 if (nfss->options & NFS_OPTION_MIGRATION)
538 seq_puts(m, ",migration");
540 if (nfss->flags & NFS_MOUNT_LOOKUP_CACHE_NONEG) {
541 if (nfss->flags & NFS_MOUNT_LOOKUP_CACHE_NONE)
542 seq_puts(m, ",lookupcache=none");
544 seq_puts(m, ",lookupcache=pos");
547 local_flock = nfss->flags & NFS_MOUNT_LOCAL_FLOCK;
548 local_fcntl = nfss->flags & NFS_MOUNT_LOCAL_FCNTL;
550 if (!local_flock && !local_fcntl)
551 seq_puts(m, ",local_lock=none");
552 else if (local_flock && local_fcntl)
553 seq_puts(m, ",local_lock=all");
554 else if (local_flock)
555 seq_puts(m, ",local_lock=flock");
557 seq_puts(m, ",local_lock=posix");
559 if (nfss->flags & NFS_MOUNT_NO_ALIGNWRITE)
560 seq_puts(m, ",noalignwrite");
562 if (nfss->flags & NFS_MOUNT_WRITE_EAGER) {
563 if (nfss->flags & NFS_MOUNT_WRITE_WAIT)
564 seq_puts(m, ",write=wait");
566 seq_puts(m, ",write=eager");
571 * Describe the mount options on this VFS mountpoint
573 int nfs_show_options(struct seq_file *m, struct dentry *root)
575 struct nfs_server *nfss = NFS_SB(root->d_sb);
577 nfs_show_mount_options(m, nfss, 0);
580 seq_printf(m, ",addr=%s",
581 rpc_peeraddr2str(nfss->nfs_client->cl_rpcclient,
587 EXPORT_SYMBOL_GPL(nfs_show_options);
589 #if IS_ENABLED(CONFIG_NFS_V4)
590 static void show_lease(struct seq_file *m, struct nfs_server *server)
592 struct nfs_client *clp = server->nfs_client;
593 unsigned long expire;
595 seq_printf(m, ",lease_time=%ld", clp->cl_lease_time / HZ);
596 expire = clp->cl_last_renewal + clp->cl_lease_time;
597 seq_printf(m, ",lease_expired=%ld",
598 time_after(expire, jiffies) ? 0 : (jiffies - expire) / HZ);
600 #ifdef CONFIG_NFS_V4_1
601 static void show_sessions(struct seq_file *m, struct nfs_server *server)
603 if (nfs4_has_session(server->nfs_client))
604 seq_puts(m, ",sessions");
607 static void show_sessions(struct seq_file *m, struct nfs_server *server) {}
611 #ifdef CONFIG_NFS_V4_1
612 static void show_pnfs(struct seq_file *m, struct nfs_server *server)
614 seq_printf(m, ",pnfs=");
615 if (server->pnfs_curr_ld)
616 seq_printf(m, "%s", server->pnfs_curr_ld->name);
618 seq_printf(m, "not configured");
621 static void show_implementation_id(struct seq_file *m, struct nfs_server *nfss)
623 if (nfss->nfs_client && nfss->nfs_client->cl_implid) {
624 struct nfs41_impl_id *impl_id = nfss->nfs_client->cl_implid;
625 seq_printf(m, "\n\timpl_id:\tname='%s',domain='%s',"
627 impl_id->name, impl_id->domain,
628 impl_id->date.seconds, impl_id->date.nseconds);
632 #if IS_ENABLED(CONFIG_NFS_V4)
633 static void show_pnfs(struct seq_file *m, struct nfs_server *server)
637 static void show_implementation_id(struct seq_file *m, struct nfs_server *nfss)
642 int nfs_show_devname(struct seq_file *m, struct dentry *root)
644 char *page = (char *) __get_free_page(GFP_KERNEL);
645 char *devname, *dummy;
649 devname = nfs_path(&dummy, root, page, PAGE_SIZE, 0);
651 err = PTR_ERR(devname);
653 seq_escape(m, devname, " \t\n\\");
654 free_page((unsigned long)page);
657 EXPORT_SYMBOL_GPL(nfs_show_devname);
659 int nfs_show_path(struct seq_file *m, struct dentry *dentry)
664 EXPORT_SYMBOL_GPL(nfs_show_path);
667 * Present statistical information for this VFS mountpoint
669 int nfs_show_stats(struct seq_file *m, struct dentry *root)
672 struct nfs_server *nfss = NFS_SB(root->d_sb);
673 struct rpc_auth *auth = nfss->client->cl_auth;
674 struct nfs_iostats totals = { };
676 seq_printf(m, "statvers=%s", NFS_IOSTAT_VERS);
679 * Display all mount option settings
681 seq_puts(m, "\n\topts:\t");
682 seq_puts(m, sb_rdonly(root->d_sb) ? "ro" : "rw");
683 seq_puts(m, root->d_sb->s_flags & SB_SYNCHRONOUS ? ",sync" : "");
684 seq_puts(m, root->d_sb->s_flags & SB_NOATIME ? ",noatime" : "");
685 seq_puts(m, root->d_sb->s_flags & SB_NODIRATIME ? ",nodiratime" : "");
686 nfs_show_mount_options(m, nfss, 1);
688 seq_printf(m, "\n\tage:\t%lu", (jiffies - nfss->mount_time) / HZ);
690 show_implementation_id(m, nfss);
692 seq_puts(m, "\n\tcaps:\t");
693 seq_printf(m, "caps=0x%x", nfss->caps);
694 seq_printf(m, ",wtmult=%u", nfss->wtmult);
695 seq_printf(m, ",dtsize=%u", nfss->dtsize);
696 seq_printf(m, ",bsize=%u", nfss->bsize);
697 seq_printf(m, ",namlen=%u", nfss->namelen);
699 #if IS_ENABLED(CONFIG_NFS_V4)
700 if (nfss->nfs_client->rpc_ops->version == 4) {
701 seq_puts(m, "\n\tnfsv4:\t");
702 seq_printf(m, "bm0=0x%x", nfss->attr_bitmask[0]);
703 seq_printf(m, ",bm1=0x%x", nfss->attr_bitmask[1]);
704 seq_printf(m, ",bm2=0x%x", nfss->attr_bitmask[2]);
705 seq_printf(m, ",acl=0x%x", nfss->acl_bitmask);
706 show_sessions(m, nfss);
713 * Display security flavor in effect for this mount
715 seq_printf(m, "\n\tsec:\tflavor=%u", auth->au_ops->au_flavor);
717 seq_printf(m, ",pseudoflavor=%u", auth->au_flavor);
720 * Display superblock I/O counters
722 for_each_possible_cpu(cpu) {
723 struct nfs_iostats *stats;
726 stats = per_cpu_ptr(nfss->io_stats, cpu);
728 for (i = 0; i < __NFSIOS_COUNTSMAX; i++)
729 totals.events[i] += stats->events[i];
730 for (i = 0; i < __NFSIOS_BYTESMAX; i++)
731 totals.bytes[i] += stats->bytes[i];
736 seq_puts(m, "\n\tevents:\t");
737 for (i = 0; i < __NFSIOS_COUNTSMAX; i++)
738 seq_printf(m, "%lu ", totals.events[i]);
739 seq_puts(m, "\n\tbytes:\t");
740 for (i = 0; i < __NFSIOS_BYTESMAX; i++)
741 seq_printf(m, "%Lu ", totals.bytes[i]);
744 rpc_clnt_show_stats(m, nfss->client);
748 EXPORT_SYMBOL_GPL(nfs_show_stats);
751 * Begin unmount by attempting to remove all automounted mountpoints we added
752 * in response to xdev traversals and referrals
754 void nfs_umount_begin(struct super_block *sb)
756 struct nfs_server *server;
757 struct rpc_clnt *rpc;
760 /* -EIO all pending I/O */
761 rpc = server->client_acl;
763 rpc_killall_tasks(rpc);
764 rpc = server->client;
766 rpc_killall_tasks(rpc);
768 EXPORT_SYMBOL_GPL(nfs_umount_begin);
771 * Return true if 'match' is in auth_info or auth_info is empty.
772 * Return false otherwise.
774 bool nfs_auth_info_match(const struct nfs_auth_info *auth_info,
775 rpc_authflavor_t match)
779 if (!auth_info->flavor_len)
782 for (i = 0; i < auth_info->flavor_len; i++) {
783 if (auth_info->flavors[i] == match)
788 EXPORT_SYMBOL_GPL(nfs_auth_info_match);
791 * Ensure that a specified authtype in ctx->auth_info is supported by
792 * the server. Returns 0 and sets ctx->selected_flavor if it's ok, and
795 static int nfs_verify_authflavors(struct nfs_fs_context *ctx,
796 rpc_authflavor_t *server_authlist,
799 rpc_authflavor_t flavor = RPC_AUTH_MAXFLAVOR;
800 bool found_auth_null = false;
804 * If the sec= mount option is used, the specified flavor or AUTH_NULL
805 * must be in the list returned by the server.
807 * AUTH_NULL has a special meaning when it's in the server list - it
808 * means that the server will ignore the rpc creds, so any flavor
809 * can be used but still use the sec= that was specified.
811 * Note also that the MNT procedure in MNTv1 does not return a list
812 * of supported security flavors. In this case, nfs_mount() fabricates
813 * a security flavor list containing just AUTH_NULL.
815 for (i = 0; i < count; i++) {
816 flavor = server_authlist[i];
818 if (nfs_auth_info_match(&ctx->auth_info, flavor))
821 if (flavor == RPC_AUTH_NULL)
822 found_auth_null = true;
825 if (found_auth_null) {
826 flavor = ctx->auth_info.flavors[0];
831 "NFS: specified auth flavors not supported by server\n");
835 ctx->selected_flavor = flavor;
836 dfprintk(MOUNT, "NFS: using auth flavor %u\n", ctx->selected_flavor);
841 * Use the remote server's MOUNT service to request the NFS file handle
842 * corresponding to the provided path.
844 static int nfs_request_mount(struct fs_context *fc,
845 struct nfs_fh *root_fh,
846 rpc_authflavor_t *server_authlist,
847 unsigned int *server_authlist_len)
849 struct nfs_fs_context *ctx = nfs_fc2context(fc);
850 struct nfs_mount_request request = {
851 .sap = &ctx->mount_server._address,
852 .dirpath = ctx->nfs_server.export_path,
853 .protocol = ctx->mount_server.protocol,
855 .noresvport = ctx->flags & NFS_MOUNT_NORESVPORT,
856 .auth_flav_len = server_authlist_len,
857 .auth_flavs = server_authlist,
862 if (ctx->mount_server.version == 0) {
863 switch (ctx->version) {
865 ctx->mount_server.version = NFS_MNT3_VERSION;
868 ctx->mount_server.version = NFS_MNT_VERSION;
871 request.version = ctx->mount_server.version;
873 if (ctx->mount_server.hostname)
874 request.hostname = ctx->mount_server.hostname;
876 request.hostname = ctx->nfs_server.hostname;
879 * Construct the mount server's address.
881 if (ctx->mount_server.address.sa_family == AF_UNSPEC) {
882 memcpy(request.sap, &ctx->nfs_server._address,
883 ctx->nfs_server.addrlen);
884 ctx->mount_server.addrlen = ctx->nfs_server.addrlen;
886 request.salen = ctx->mount_server.addrlen;
887 nfs_set_port(request.sap, &ctx->mount_server.port, 0);
890 * Now ask the mount server to map our export path
893 if ((request.protocol == XPRT_TRANSPORT_UDP) ==
894 !(ctx->flags & NFS_MOUNT_TCP))
896 * NFS protocol and mount protocol are both UDP or neither UDP
897 * so timeouts are compatible. Use NFS timeouts for MOUNT
899 status = nfs_mount(&request, ctx->timeo, ctx->retrans);
901 status = nfs_mount(&request, NFS_UNSPEC_TIMEO, NFS_UNSPEC_RETRANS);
903 dfprintk(MOUNT, "NFS: unable to mount server %s, error %d\n",
904 request.hostname, status);
911 static struct nfs_server *nfs_try_mount_request(struct fs_context *fc)
913 struct nfs_fs_context *ctx = nfs_fc2context(fc);
916 bool tried_auth_unix = false;
917 bool auth_null_in_list = false;
918 struct nfs_server *server = ERR_PTR(-EACCES);
919 rpc_authflavor_t authlist[NFS_MAX_SECFLAVORS];
920 unsigned int authlist_len = ARRAY_SIZE(authlist);
922 /* make sure 'nolock'/'lock' override the 'local_lock' mount option */
923 if (ctx->lock_status) {
924 if (ctx->lock_status == NFS_LOCK_NOLOCK) {
925 ctx->flags |= NFS_MOUNT_NONLM;
926 ctx->flags |= (NFS_MOUNT_LOCAL_FLOCK | NFS_MOUNT_LOCAL_FCNTL);
928 ctx->flags &= ~NFS_MOUNT_NONLM;
929 ctx->flags &= ~(NFS_MOUNT_LOCAL_FLOCK | NFS_MOUNT_LOCAL_FCNTL);
932 status = nfs_request_mount(fc, ctx->mntfh, authlist, &authlist_len);
934 return ERR_PTR(status);
937 * Was a sec= authflavor specified in the options? First, verify
938 * whether the server supports it, and then just try to use it if so.
940 if (ctx->auth_info.flavor_len > 0) {
941 status = nfs_verify_authflavors(ctx, authlist, authlist_len);
942 dfprintk(MOUNT, "NFS: using auth flavor %u\n",
943 ctx->selected_flavor);
945 return ERR_PTR(status);
946 return ctx->nfs_mod->rpc_ops->create_server(fc);
950 * No sec= option was provided. RFC 2623, section 2.7 suggests we
951 * SHOULD prefer the flavor listed first. However, some servers list
952 * AUTH_NULL first. Avoid ever choosing AUTH_NULL.
954 for (i = 0; i < authlist_len; ++i) {
955 rpc_authflavor_t flavor;
956 struct rpcsec_gss_info info;
958 flavor = authlist[i];
961 tried_auth_unix = true;
964 auth_null_in_list = true;
967 if (rpcauth_get_gssinfo(flavor, &info) != 0)
971 dfprintk(MOUNT, "NFS: attempting to use auth flavor %u\n", flavor);
972 ctx->selected_flavor = flavor;
973 server = ctx->nfs_mod->rpc_ops->create_server(fc);
979 * Nothing we tried so far worked. At this point, give up if we've
980 * already tried AUTH_UNIX or if the server's list doesn't contain
983 if (tried_auth_unix || !auth_null_in_list)
986 /* Last chance! Try AUTH_UNIX */
987 dfprintk(MOUNT, "NFS: attempting to use auth flavor %u\n", RPC_AUTH_UNIX);
988 ctx->selected_flavor = RPC_AUTH_UNIX;
989 return ctx->nfs_mod->rpc_ops->create_server(fc);
992 int nfs_try_get_tree(struct fs_context *fc)
994 struct nfs_fs_context *ctx = nfs_fc2context(fc);
997 ctx->server = nfs_try_mount_request(fc);
999 ctx->server = ctx->nfs_mod->rpc_ops->create_server(fc);
1001 return nfs_get_tree_common(fc);
1003 EXPORT_SYMBOL_GPL(nfs_try_get_tree);
1006 #define NFS_REMOUNT_CMP_FLAGMASK ~(NFS_MOUNT_INTR \
1007 | NFS_MOUNT_SECURE \
1010 | NFS_MOUNT_KERBEROS \
1012 | NFS_MOUNT_BROKEN_SUID \
1013 | NFS_MOUNT_STRICTLOCK \
1014 | NFS_MOUNT_LEGACY_INTERFACE)
1016 #define NFS_MOUNT_CMP_FLAGMASK (NFS_REMOUNT_CMP_FLAGMASK & \
1017 ~(NFS_MOUNT_UNSHARED | NFS_MOUNT_NORESVPORT))
1020 nfs_compare_remount_data(struct nfs_server *nfss,
1021 struct nfs_fs_context *ctx)
1023 if ((ctx->flags ^ nfss->flags) & NFS_REMOUNT_CMP_FLAGMASK ||
1024 ctx->rsize != nfss->rsize ||
1025 ctx->wsize != nfss->wsize ||
1026 ctx->version != nfss->nfs_client->rpc_ops->version ||
1027 ctx->minorversion != nfss->nfs_client->cl_minorversion ||
1028 ctx->retrans != nfss->client->cl_timeout->to_retries ||
1029 !nfs_auth_info_match(&ctx->auth_info, nfss->client->cl_auth->au_flavor) ||
1030 ctx->acregmin != nfss->acregmin / HZ ||
1031 ctx->acregmax != nfss->acregmax / HZ ||
1032 ctx->acdirmin != nfss->acdirmin / HZ ||
1033 ctx->acdirmax != nfss->acdirmax / HZ ||
1034 ctx->timeo != (10U * nfss->client->cl_timeout->to_initval / HZ) ||
1035 (ctx->options & NFS_OPTION_FSCACHE) != (nfss->options & NFS_OPTION_FSCACHE) ||
1036 ctx->nfs_server.port != nfss->port ||
1037 ctx->nfs_server.addrlen != nfss->nfs_client->cl_addrlen ||
1038 !rpc_cmp_addr((struct sockaddr *)&ctx->nfs_server.address,
1039 (struct sockaddr *)&nfss->nfs_client->cl_addr))
1045 int nfs_reconfigure(struct fs_context *fc)
1047 struct nfs_fs_context *ctx = nfs_fc2context(fc);
1048 struct super_block *sb = fc->root->d_sb;
1049 struct nfs_server *nfss = sb->s_fs_info;
1052 sync_filesystem(sb);
1055 * The SB_RDONLY flag has been removed from the superblock during
1056 * mounts to prevent interference between different filesystems.
1057 * Similarly, it is also necessary to ignore the SB_RDONLY flag
1058 * during reconfiguration; otherwise, it may also result in the
1059 * creation of redundant superblocks when mounting a directory with
1060 * different rw and ro flags multiple times.
1062 fc->sb_flags_mask &= ~SB_RDONLY;
1065 * Userspace mount programs that send binary options generally send
1066 * them populated with default values. We have no way to know which
1067 * ones were explicitly specified. Fall back to legacy behavior and
1068 * just return success.
1070 if (ctx->skip_reconfig_option_check)
1074 * noac is a special case. It implies -o sync, but that's not
1075 * necessarily reflected in the mtab options. reconfigure_super
1076 * will clear SB_SYNCHRONOUS if -o sync wasn't specified in the
1077 * remount options, so we have to explicitly reset it.
1079 if (ctx->flags & NFS_MOUNT_NOAC) {
1080 fc->sb_flags |= SB_SYNCHRONOUS;
1081 fc->sb_flags_mask |= SB_SYNCHRONOUS;
1084 /* compare new mount options with old ones */
1085 ret = nfs_compare_remount_data(nfss, ctx);
1089 return nfs_probe_server(nfss, NFS_FH(d_inode(fc->root)));
1091 EXPORT_SYMBOL_GPL(nfs_reconfigure);
1094 * Finish setting up an NFS superblock
1096 static void nfs_fill_super(struct super_block *sb, struct nfs_fs_context *ctx)
1098 struct nfs_server *server = NFS_SB(sb);
1100 sb->s_blocksize_bits = 0;
1101 sb->s_blocksize = 0;
1102 sb->s_xattr = server->nfs_client->cl_nfs_mod->xattr;
1103 sb->s_op = server->nfs_client->cl_nfs_mod->sops;
1105 sb->s_blocksize = nfs_block_size(ctx->bsize, &sb->s_blocksize_bits);
1107 switch (server->nfs_client->rpc_ops->version) {
1109 sb->s_time_gran = 1000;
1111 sb->s_time_max = U32_MAX;
1115 * The VFS shouldn't apply the umask to mode bits.
1116 * We will do so ourselves when necessary.
1118 sb->s_flags |= SB_POSIXACL;
1119 sb->s_time_gran = 1;
1121 sb->s_time_max = U32_MAX;
1122 sb->s_export_op = &nfs_export_ops;
1125 sb->s_iflags |= SB_I_NOUMASK;
1126 sb->s_time_gran = 1;
1127 sb->s_time_min = S64_MIN;
1128 sb->s_time_max = S64_MAX;
1129 if (server->caps & NFS_CAP_ATOMIC_OPEN_V1)
1130 sb->s_export_op = &nfs_export_ops;
1134 sb->s_magic = NFS_SUPER_MAGIC;
1136 /* We probably want something more informative here */
1137 snprintf(sb->s_id, sizeof(sb->s_id),
1138 "%u:%u", MAJOR(sb->s_dev), MINOR(sb->s_dev));
1140 if (sb->s_blocksize == 0)
1141 sb->s_blocksize = nfs_block_bits(server->wsize,
1142 &sb->s_blocksize_bits);
1144 nfs_super_set_maxbytes(sb, server->maxfilesize);
1145 nfs_sysfs_move_server_to_sb(sb);
1146 server->has_sec_mnt_opts = ctx->has_sec_mnt_opts;
1149 static int nfs_compare_mount_options(const struct super_block *s, const struct nfs_server *b,
1150 const struct fs_context *fc)
1152 const struct nfs_server *a = s->s_fs_info;
1153 const struct rpc_clnt *clnt_a = a->client;
1154 const struct rpc_clnt *clnt_b = b->client;
1156 if ((s->s_flags & NFS_SB_MASK) != (fc->sb_flags & NFS_SB_MASK))
1158 if (a->nfs_client != b->nfs_client)
1160 if ((a->flags ^ b->flags) & NFS_MOUNT_CMP_FLAGMASK)
1162 if (a->wsize != b->wsize)
1164 if (a->rsize != b->rsize)
1166 if (a->acregmin != b->acregmin)
1168 if (a->acregmax != b->acregmax)
1170 if (a->acdirmin != b->acdirmin)
1172 if (a->acdirmax != b->acdirmax)
1174 if (clnt_a->cl_auth->au_flavor != clnt_b->cl_auth->au_flavor)
1181 static int nfs_set_super(struct super_block *s, struct fs_context *fc)
1183 struct nfs_server *server = fc->s_fs_info;
1186 s->s_d_op = server->nfs_client->rpc_ops->dentry_ops;
1187 ret = set_anon_super(s, server);
1189 server->s_dev = s->s_dev;
1193 static int nfs_compare_super_address(struct nfs_server *server1,
1194 struct nfs_server *server2)
1196 struct sockaddr *sap1, *sap2;
1197 struct rpc_xprt *xprt1 = server1->client->cl_xprt;
1198 struct rpc_xprt *xprt2 = server2->client->cl_xprt;
1200 if (!net_eq(xprt1->xprt_net, xprt2->xprt_net))
1203 sap1 = (struct sockaddr *)&server1->nfs_client->cl_addr;
1204 sap2 = (struct sockaddr *)&server2->nfs_client->cl_addr;
1206 if (sap1->sa_family != sap2->sa_family)
1209 switch (sap1->sa_family) {
1211 struct sockaddr_in *sin1 = (struct sockaddr_in *)sap1;
1212 struct sockaddr_in *sin2 = (struct sockaddr_in *)sap2;
1213 if (sin1->sin_addr.s_addr != sin2->sin_addr.s_addr)
1215 if (sin1->sin_port != sin2->sin_port)
1220 struct sockaddr_in6 *sin1 = (struct sockaddr_in6 *)sap1;
1221 struct sockaddr_in6 *sin2 = (struct sockaddr_in6 *)sap2;
1222 if (!ipv6_addr_equal(&sin1->sin6_addr, &sin2->sin6_addr))
1224 if (sin1->sin6_port != sin2->sin6_port)
1235 static int nfs_compare_userns(const struct nfs_server *old,
1236 const struct nfs_server *new)
1238 const struct user_namespace *oldns = &init_user_ns;
1239 const struct user_namespace *newns = &init_user_ns;
1241 if (old->client && old->client->cl_cred)
1242 oldns = old->client->cl_cred->user_ns;
1243 if (new->client && new->client->cl_cred)
1244 newns = new->client->cl_cred->user_ns;
1250 static int nfs_compare_super(struct super_block *sb, struct fs_context *fc)
1252 struct nfs_server *server = fc->s_fs_info, *old = NFS_SB(sb);
1254 if (!nfs_compare_super_address(old, server))
1256 /* Note: NFS_MOUNT_UNSHARED == NFS4_MOUNT_UNSHARED */
1257 if (old->flags & NFS_MOUNT_UNSHARED)
1259 if (memcmp(&old->fsid, &server->fsid, sizeof(old->fsid)) != 0)
1261 if (!nfs_compare_userns(old, server))
1263 if ((old->has_sec_mnt_opts || fc->security) &&
1264 security_sb_mnt_opts_compat(sb, fc->security))
1266 return nfs_compare_mount_options(sb, server, fc);
1269 #ifdef CONFIG_NFS_FSCACHE
1270 static int nfs_get_cache_cookie(struct super_block *sb,
1271 struct nfs_fs_context *ctx)
1273 struct nfs_server *nfss = NFS_SB(sb);
1277 nfss->fscache = NULL;
1282 if (ctx->clone_data.sb) {
1283 struct nfs_server *mnt_s = NFS_SB(ctx->clone_data.sb);
1284 if (!(mnt_s->options & NFS_OPTION_FSCACHE))
1286 if (mnt_s->fscache_uniq) {
1287 uniq = mnt_s->fscache_uniq;
1288 ulen = strlen(uniq);
1291 if (!(ctx->options & NFS_OPTION_FSCACHE))
1293 if (ctx->fscache_uniq) {
1294 uniq = ctx->fscache_uniq;
1295 ulen = strlen(ctx->fscache_uniq);
1299 return nfs_fscache_get_super_cookie(sb, uniq, ulen);
1302 static int nfs_get_cache_cookie(struct super_block *sb,
1303 struct nfs_fs_context *ctx)
1309 int nfs_get_tree_common(struct fs_context *fc)
1311 struct nfs_fs_context *ctx = nfs_fc2context(fc);
1312 struct super_block *s;
1313 int (*compare_super)(struct super_block *, struct fs_context *) = nfs_compare_super;
1314 struct nfs_server *server = ctx->server;
1319 return PTR_ERR(server);
1322 * When NFS_MOUNT_UNSHARED is not set, NFS forces the sharing of a
1323 * superblock among each filesystem that mounts sub-directories
1324 * belonging to a single exported root path.
1325 * To prevent interference between different filesystems, the
1326 * SB_RDONLY flag should be removed from the superblock.
1328 if (server->flags & NFS_MOUNT_UNSHARED)
1329 compare_super = NULL;
1331 fc->sb_flags &= ~SB_RDONLY;
1333 /* -o noac implies -o sync */
1334 if (server->flags & NFS_MOUNT_NOAC)
1335 fc->sb_flags |= SB_SYNCHRONOUS;
1337 if (ctx->clone_data.sb)
1338 if (ctx->clone_data.sb->s_flags & SB_SYNCHRONOUS)
1339 fc->sb_flags |= SB_SYNCHRONOUS;
1341 /* Get a superblock - note that we may end up sharing one that already exists */
1342 fc->s_fs_info = server;
1343 s = sget_fc(fc, compare_super, nfs_set_super);
1344 fc->s_fs_info = NULL;
1347 nfs_errorf(fc, "NFS: Couldn't get superblock");
1351 if (s->s_fs_info != server) {
1352 nfs_free_server(server);
1355 error = super_setup_bdi_name(s, "%u:%u", MAJOR(server->s_dev),
1356 MINOR(server->s_dev));
1358 goto error_splat_super;
1359 s->s_bdi->io_pages = server->rpages;
1364 unsigned bsize = ctx->clone_data.inherited_bsize;
1365 /* initial superblock/root creation */
1366 nfs_fill_super(s, ctx);
1368 s->s_blocksize_bits = bsize;
1369 s->s_blocksize = 1U << bsize;
1371 error = nfs_get_cache_cookie(s, ctx);
1373 goto error_splat_super;
1376 error = nfs_get_root(s, fc);
1378 nfs_errorf(fc, "NFS: Couldn't get root dentry");
1379 goto error_splat_super;
1382 s->s_flags |= SB_ACTIVE;
1389 nfs_free_server(server);
1392 deactivate_locked_super(s);
1397 * Destroy an NFS superblock
1399 void nfs_kill_super(struct super_block *s)
1401 struct nfs_server *server = NFS_SB(s);
1403 nfs_sysfs_move_sb_to_server(server);
1406 nfs_fscache_release_super_cookie(s);
1408 nfs_free_server(server);
1410 EXPORT_SYMBOL_GPL(nfs_kill_super);
1412 #if IS_ENABLED(CONFIG_NFS_V4)
1415 * NFS v4 module parameters need to stay in the
1416 * NFS client for backwards compatibility
1418 unsigned int nfs_callback_set_tcpport;
1419 unsigned short nfs_callback_nr_threads;
1420 /* Default cache timeout is 10 minutes */
1421 unsigned int nfs_idmap_cache_timeout = 600;
1422 /* Turn off NFSv4 uid/gid mapping when using AUTH_SYS */
1423 bool nfs4_disable_idmapping = true;
1424 unsigned short max_session_slots = NFS4_DEF_SLOT_TABLE_SIZE;
1425 unsigned short max_session_cb_slots = NFS4_DEF_CB_SLOT_TABLE_SIZE;
1426 unsigned short send_implementation_id = 1;
1427 char nfs4_client_id_uniquifier[NFS4_CLIENT_ID_UNIQ_LEN] = "";
1428 bool recover_lost_locks = false;
1429 short nfs_delay_retrans = -1;
1431 EXPORT_SYMBOL_GPL(nfs_callback_nr_threads);
1432 EXPORT_SYMBOL_GPL(nfs_callback_set_tcpport);
1433 EXPORT_SYMBOL_GPL(nfs_idmap_cache_timeout);
1434 EXPORT_SYMBOL_GPL(nfs4_disable_idmapping);
1435 EXPORT_SYMBOL_GPL(max_session_slots);
1436 EXPORT_SYMBOL_GPL(max_session_cb_slots);
1437 EXPORT_SYMBOL_GPL(send_implementation_id);
1438 EXPORT_SYMBOL_GPL(nfs4_client_id_uniquifier);
1439 EXPORT_SYMBOL_GPL(recover_lost_locks);
1440 EXPORT_SYMBOL_GPL(nfs_delay_retrans);
1442 #define NFS_CALLBACK_MAXPORTNR (65535U)
1444 static int param_set_portnr(const char *val, const struct kernel_param *kp)
1451 ret = kstrtoul(val, 0, &num);
1452 if (ret || num > NFS_CALLBACK_MAXPORTNR)
1454 *((unsigned int *)kp->arg) = num;
1457 static const struct kernel_param_ops param_ops_portnr = {
1458 .set = param_set_portnr,
1459 .get = param_get_uint,
1461 #define param_check_portnr(name, p) __param_check(name, p, unsigned int)
1463 module_param_named(callback_tcpport, nfs_callback_set_tcpport, portnr, 0644);
1464 module_param_named(callback_nr_threads, nfs_callback_nr_threads, ushort, 0644);
1465 MODULE_PARM_DESC(callback_nr_threads, "Number of threads that will be "
1466 "assigned to the NFSv4 callback channels.");
1467 module_param(nfs_idmap_cache_timeout, int, 0644);
1468 module_param(nfs4_disable_idmapping, bool, 0644);
1469 module_param_string(nfs4_unique_id, nfs4_client_id_uniquifier,
1470 NFS4_CLIENT_ID_UNIQ_LEN, 0600);
1471 MODULE_PARM_DESC(nfs4_disable_idmapping,
1472 "Turn off NFSv4 idmapping when using 'sec=sys'");
1473 module_param(max_session_slots, ushort, 0644);
1474 MODULE_PARM_DESC(max_session_slots, "Maximum number of outstanding NFSv4.1 "
1475 "requests the client will negotiate");
1476 module_param(max_session_cb_slots, ushort, 0644);
1477 MODULE_PARM_DESC(max_session_cb_slots, "Maximum number of parallel NFSv4.1 "
1478 "callbacks the client will process for a given server");
1479 module_param(send_implementation_id, ushort, 0644);
1480 MODULE_PARM_DESC(send_implementation_id,
1481 "Send implementation ID with NFSv4.1 exchange_id");
1482 MODULE_PARM_DESC(nfs4_unique_id, "nfs_client_id4 uniquifier string");
1484 module_param(recover_lost_locks, bool, 0644);
1485 MODULE_PARM_DESC(recover_lost_locks,
1486 "If the server reports that a lock might be lost, "
1487 "try to recover it risking data corruption.");
1489 module_param_named(delay_retrans, nfs_delay_retrans, short, 0644);
1490 MODULE_PARM_DESC(delay_retrans,
1491 "Unless negative, specifies the number of times the NFSv4 "
1492 "client retries a request before returning an EAGAIN error, "
1493 "after a reply of NFS4ERR_DELAY from the server.");
1494 #endif /* CONFIG_NFS_V4 */