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/slab.h>
52 #include <linux/netdevice.h>
53 #include <linux/nfs_xdr.h>
54 #include <linux/magic.h>
55 #include <linux/parser.h>
56 #include <linux/nsproxy.h>
57 #include <linux/rcupdate.h>
59 #include <linux/uaccess.h>
60 #include <linux/nfs_ssc.h>
64 #include "delegation.h"
68 #include "nfs4session.h"
72 #define NFSDBG_FACILITY NFSDBG_VFS
74 const struct super_operations nfs_sops = {
75 .alloc_inode = nfs_alloc_inode,
76 .free_inode = nfs_free_inode,
77 .write_inode = nfs_write_inode,
78 .drop_inode = nfs_drop_inode,
80 .evict_inode = nfs_evict_inode,
81 .umount_begin = nfs_umount_begin,
82 .show_options = nfs_show_options,
83 .show_devname = nfs_show_devname,
84 .show_path = nfs_show_path,
85 .show_stats = nfs_show_stats,
87 EXPORT_SYMBOL_GPL(nfs_sops);
89 #ifdef CONFIG_NFS_V4_2
90 static const struct nfs_ssc_client_ops nfs_ssc_clnt_ops_tbl = {
91 .sco_sb_deactive = nfs_sb_deactive,
95 #if IS_ENABLED(CONFIG_NFS_V4)
96 static int __init register_nfs4_fs(void)
98 return register_filesystem(&nfs4_fs_type);
101 static void unregister_nfs4_fs(void)
103 unregister_filesystem(&nfs4_fs_type);
106 static int __init register_nfs4_fs(void)
111 static void unregister_nfs4_fs(void)
116 #ifdef CONFIG_NFS_V4_2
117 static void nfs_ssc_register_ops(void)
119 nfs_ssc_register(&nfs_ssc_clnt_ops_tbl);
122 static void nfs_ssc_unregister_ops(void)
124 nfs_ssc_unregister(&nfs_ssc_clnt_ops_tbl);
126 #endif /* CONFIG_NFS_V4_2 */
128 static struct shrinker acl_shrinker = {
129 .count_objects = nfs_access_cache_count,
130 .scan_objects = nfs_access_cache_scan,
131 .seeks = DEFAULT_SEEKS,
135 * Register the NFS filesystems
137 int __init register_nfs_fs(void)
141 ret = register_filesystem(&nfs_fs_type);
145 ret = register_nfs4_fs();
149 ret = nfs_register_sysctl();
152 ret = register_shrinker(&acl_shrinker, "nfs-acl");
155 #ifdef CONFIG_NFS_V4_2
156 nfs_ssc_register_ops();
160 nfs_unregister_sysctl();
162 unregister_nfs4_fs();
164 unregister_filesystem(&nfs_fs_type);
170 * Unregister the NFS filesystems
172 void __exit unregister_nfs_fs(void)
174 unregister_shrinker(&acl_shrinker);
175 nfs_unregister_sysctl();
176 unregister_nfs4_fs();
177 #ifdef CONFIG_NFS_V4_2
178 nfs_ssc_unregister_ops();
180 unregister_filesystem(&nfs_fs_type);
183 bool nfs_sb_active(struct super_block *sb)
185 struct nfs_server *server = NFS_SB(sb);
187 if (!atomic_inc_not_zero(&sb->s_active))
189 if (atomic_inc_return(&server->active) != 1)
190 atomic_dec(&sb->s_active);
193 EXPORT_SYMBOL_GPL(nfs_sb_active);
195 void nfs_sb_deactive(struct super_block *sb)
197 struct nfs_server *server = NFS_SB(sb);
199 if (atomic_dec_and_test(&server->active))
200 deactivate_super(sb);
202 EXPORT_SYMBOL_GPL(nfs_sb_deactive);
204 static int __nfs_list_for_each_server(struct list_head *head,
205 int (*fn)(struct nfs_server *, void *),
208 struct nfs_server *server, *last = NULL;
212 list_for_each_entry_rcu(server, head, client_link) {
213 if (!(server->super && nfs_sb_active(server->super)))
217 nfs_sb_deactive(last->super);
219 ret = fn(server, data);
227 nfs_sb_deactive(last->super);
231 int nfs_client_for_each_server(struct nfs_client *clp,
232 int (*fn)(struct nfs_server *, void *),
235 return __nfs_list_for_each_server(&clp->cl_superblocks, fn, data);
237 EXPORT_SYMBOL_GPL(nfs_client_for_each_server);
240 * Deliver file system statistics to userspace
242 int nfs_statfs(struct dentry *dentry, struct kstatfs *buf)
244 struct nfs_server *server = NFS_SB(dentry->d_sb);
245 unsigned char blockbits;
246 unsigned long blockres;
247 struct nfs_fh *fh = NFS_FH(d_inode(dentry));
248 struct nfs_fsstat res;
251 res.fattr = nfs_alloc_fattr();
252 if (res.fattr == NULL)
255 error = server->nfs_client->rpc_ops->statfs(server, fh, &res);
256 if (unlikely(error == -ESTALE)) {
257 struct dentry *pd_dentry;
259 pd_dentry = dget_parent(dentry);
260 nfs_zap_caches(d_inode(pd_dentry));
263 nfs_free_fattr(res.fattr);
267 buf->f_type = NFS_SUPER_MAGIC;
270 * Current versions of glibc do not correctly handle the
271 * case where f_frsize != f_bsize. Eventually we want to
272 * report the value of wtmult in this field.
274 buf->f_frsize = dentry->d_sb->s_blocksize;
277 * On most *nix systems, f_blocks, f_bfree, and f_bavail
278 * are reported in units of f_frsize. Linux hasn't had
279 * an f_frsize field in its statfs struct until recently,
280 * thus historically Linux's sys_statfs reports these
281 * fields in units of f_bsize.
283 buf->f_bsize = dentry->d_sb->s_blocksize;
284 blockbits = dentry->d_sb->s_blocksize_bits;
285 blockres = (1 << blockbits) - 1;
286 buf->f_blocks = (res.tbytes + blockres) >> blockbits;
287 buf->f_bfree = (res.fbytes + blockres) >> blockbits;
288 buf->f_bavail = (res.abytes + blockres) >> blockbits;
290 buf->f_files = res.tfiles;
291 buf->f_ffree = res.afiles;
293 buf->f_namelen = server->namelen;
298 dprintk("%s: statfs error = %d\n", __func__, -error);
301 EXPORT_SYMBOL_GPL(nfs_statfs);
304 * Map the security flavour number to a name
306 static const char *nfs_pseudoflavour_to_name(rpc_authflavor_t flavour)
308 static const struct {
309 rpc_authflavor_t flavour;
311 } sec_flavours[NFS_AUTH_INFO_MAX_FLAVORS] = {
312 /* update NFS_AUTH_INFO_MAX_FLAVORS when this list changes! */
313 { RPC_AUTH_NULL, "null" },
314 { RPC_AUTH_UNIX, "sys" },
315 { RPC_AUTH_GSS_KRB5, "krb5" },
316 { RPC_AUTH_GSS_KRB5I, "krb5i" },
317 { RPC_AUTH_GSS_KRB5P, "krb5p" },
318 { RPC_AUTH_GSS_LKEY, "lkey" },
319 { RPC_AUTH_GSS_LKEYI, "lkeyi" },
320 { RPC_AUTH_GSS_LKEYP, "lkeyp" },
321 { RPC_AUTH_GSS_SPKM, "spkm" },
322 { RPC_AUTH_GSS_SPKMI, "spkmi" },
323 { RPC_AUTH_GSS_SPKMP, "spkmp" },
324 { UINT_MAX, "unknown" }
328 for (i = 0; sec_flavours[i].flavour != UINT_MAX; i++) {
329 if (sec_flavours[i].flavour == flavour)
332 return sec_flavours[i].str;
335 static void nfs_show_mountd_netid(struct seq_file *m, struct nfs_server *nfss,
338 struct sockaddr *sap = (struct sockaddr *) &nfss->mountd_address;
341 switch (sap->sa_family) {
343 switch (nfss->mountd_protocol) {
345 proto = RPCBIND_NETID_UDP;
348 proto = RPCBIND_NETID_TCP;
353 switch (nfss->mountd_protocol) {
355 proto = RPCBIND_NETID_UDP6;
358 proto = RPCBIND_NETID_TCP6;
363 if (proto || showdefaults)
364 seq_printf(m, ",mountproto=%s", proto ?: "auto");
367 static void nfs_show_mountd_options(struct seq_file *m, struct nfs_server *nfss,
370 struct sockaddr *sap = (struct sockaddr *)&nfss->mountd_address;
372 if (nfss->flags & NFS_MOUNT_LEGACY_INTERFACE)
375 switch (sap->sa_family) {
377 struct sockaddr_in *sin = (struct sockaddr_in *)sap;
378 seq_printf(m, ",mountaddr=%pI4", &sin->sin_addr.s_addr);
382 struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)sap;
383 seq_printf(m, ",mountaddr=%pI6c", &sin6->sin6_addr);
388 seq_puts(m, ",mountaddr=unspecified");
391 if (nfss->mountd_version || showdefaults)
392 seq_printf(m, ",mountvers=%u", nfss->mountd_version);
393 if ((nfss->mountd_port &&
394 nfss->mountd_port != (unsigned short)NFS_UNSPEC_PORT) ||
396 seq_printf(m, ",mountport=%u", nfss->mountd_port);
398 nfs_show_mountd_netid(m, nfss, showdefaults);
401 #if IS_ENABLED(CONFIG_NFS_V4)
402 static void nfs_show_nfsv4_options(struct seq_file *m, struct nfs_server *nfss,
405 struct nfs_client *clp = nfss->nfs_client;
407 seq_printf(m, ",clientaddr=%s", clp->cl_ipaddr);
410 static void nfs_show_nfsv4_options(struct seq_file *m, struct nfs_server *nfss,
416 static void nfs_show_nfs_version(struct seq_file *m,
417 unsigned int version,
418 unsigned int minorversion)
420 seq_printf(m, ",vers=%u", version);
422 seq_printf(m, ".%u", minorversion);
426 * Describe the mount options in force on this server representation
428 static void nfs_show_mount_options(struct seq_file *m, struct nfs_server *nfss,
431 static const struct proc_nfs_info {
436 { NFS_MOUNT_SOFT, ",soft", "" },
437 { NFS_MOUNT_SOFTERR, ",softerr", "" },
438 { NFS_MOUNT_SOFTREVAL, ",softreval", "" },
439 { NFS_MOUNT_POSIX, ",posix", "" },
440 { NFS_MOUNT_NOCTO, ",nocto", "" },
441 { NFS_MOUNT_NOAC, ",noac", "" },
442 { NFS_MOUNT_NONLM, ",nolock", "" },
443 { NFS_MOUNT_NOACL, ",noacl", "" },
444 { NFS_MOUNT_NORDIRPLUS, ",nordirplus", "" },
445 { NFS_MOUNT_UNSHARED, ",nosharecache", "" },
446 { NFS_MOUNT_NORESVPORT, ",noresvport", "" },
449 const struct proc_nfs_info *nfs_infop;
450 struct nfs_client *clp = nfss->nfs_client;
451 u32 version = clp->rpc_ops->version;
452 int local_flock, local_fcntl;
454 nfs_show_nfs_version(m, version, clp->cl_minorversion);
455 seq_printf(m, ",rsize=%u", nfss->rsize);
456 seq_printf(m, ",wsize=%u", nfss->wsize);
457 if (nfss->bsize != 0)
458 seq_printf(m, ",bsize=%u", nfss->bsize);
459 seq_printf(m, ",namlen=%u", nfss->namelen);
460 if (nfss->acregmin != NFS_DEF_ACREGMIN*HZ || showdefaults)
461 seq_printf(m, ",acregmin=%u", nfss->acregmin/HZ);
462 if (nfss->acregmax != NFS_DEF_ACREGMAX*HZ || showdefaults)
463 seq_printf(m, ",acregmax=%u", nfss->acregmax/HZ);
464 if (nfss->acdirmin != NFS_DEF_ACDIRMIN*HZ || showdefaults)
465 seq_printf(m, ",acdirmin=%u", nfss->acdirmin/HZ);
466 if (nfss->acdirmax != NFS_DEF_ACDIRMAX*HZ || showdefaults)
467 seq_printf(m, ",acdirmax=%u", nfss->acdirmax/HZ);
468 if (!(nfss->flags & (NFS_MOUNT_SOFT|NFS_MOUNT_SOFTERR)))
469 seq_puts(m, ",hard");
470 for (nfs_infop = nfs_info; nfs_infop->flag; nfs_infop++) {
471 if (nfss->flags & nfs_infop->flag)
472 seq_puts(m, nfs_infop->str);
474 seq_puts(m, nfs_infop->nostr);
477 seq_printf(m, ",proto=%s",
478 rpc_peeraddr2str(nfss->client, RPC_DISPLAY_NETID));
480 if (clp->cl_nconnect > 0)
481 seq_printf(m, ",nconnect=%u", clp->cl_nconnect);
483 if (clp->cl_max_connect > 1)
484 seq_printf(m, ",max_connect=%u", clp->cl_max_connect);
485 if (nfss->port != NFS_PORT)
486 seq_printf(m, ",port=%u", nfss->port);
489 seq_printf(m, ",port=%u", nfss->port);
491 seq_printf(m, ",timeo=%lu", 10U * nfss->client->cl_timeout->to_initval / HZ);
492 seq_printf(m, ",retrans=%u", nfss->client->cl_timeout->to_retries);
493 seq_printf(m, ",sec=%s", nfs_pseudoflavour_to_name(nfss->client->cl_auth->au_flavor));
496 nfs_show_mountd_options(m, nfss, showdefaults);
498 nfs_show_nfsv4_options(m, nfss, showdefaults);
500 if (nfss->options & NFS_OPTION_FSCACHE)
503 if (nfss->options & NFS_OPTION_MIGRATION)
504 seq_puts(m, ",migration");
506 if (nfss->flags & NFS_MOUNT_LOOKUP_CACHE_NONEG) {
507 if (nfss->flags & NFS_MOUNT_LOOKUP_CACHE_NONE)
508 seq_puts(m, ",lookupcache=none");
510 seq_puts(m, ",lookupcache=pos");
513 local_flock = nfss->flags & NFS_MOUNT_LOCAL_FLOCK;
514 local_fcntl = nfss->flags & NFS_MOUNT_LOCAL_FCNTL;
516 if (!local_flock && !local_fcntl)
517 seq_puts(m, ",local_lock=none");
518 else if (local_flock && local_fcntl)
519 seq_puts(m, ",local_lock=all");
520 else if (local_flock)
521 seq_puts(m, ",local_lock=flock");
523 seq_puts(m, ",local_lock=posix");
525 if (nfss->flags & NFS_MOUNT_WRITE_EAGER) {
526 if (nfss->flags & NFS_MOUNT_WRITE_WAIT)
527 seq_puts(m, ",write=wait");
529 seq_puts(m, ",write=eager");
534 * Describe the mount options on this VFS mountpoint
536 int nfs_show_options(struct seq_file *m, struct dentry *root)
538 struct nfs_server *nfss = NFS_SB(root->d_sb);
540 nfs_show_mount_options(m, nfss, 0);
543 seq_printf(m, ",addr=%s",
544 rpc_peeraddr2str(nfss->nfs_client->cl_rpcclient,
550 EXPORT_SYMBOL_GPL(nfs_show_options);
552 #if IS_ENABLED(CONFIG_NFS_V4)
553 static void show_lease(struct seq_file *m, struct nfs_server *server)
555 struct nfs_client *clp = server->nfs_client;
556 unsigned long expire;
558 seq_printf(m, ",lease_time=%ld", clp->cl_lease_time / HZ);
559 expire = clp->cl_last_renewal + clp->cl_lease_time;
560 seq_printf(m, ",lease_expired=%ld",
561 time_after(expire, jiffies) ? 0 : (jiffies - expire) / HZ);
563 #ifdef CONFIG_NFS_V4_1
564 static void show_sessions(struct seq_file *m, struct nfs_server *server)
566 if (nfs4_has_session(server->nfs_client))
567 seq_puts(m, ",sessions");
570 static void show_sessions(struct seq_file *m, struct nfs_server *server) {}
574 #ifdef CONFIG_NFS_V4_1
575 static void show_pnfs(struct seq_file *m, struct nfs_server *server)
577 seq_printf(m, ",pnfs=");
578 if (server->pnfs_curr_ld)
579 seq_printf(m, "%s", server->pnfs_curr_ld->name);
581 seq_printf(m, "not configured");
584 static void show_implementation_id(struct seq_file *m, struct nfs_server *nfss)
586 if (nfss->nfs_client && nfss->nfs_client->cl_implid) {
587 struct nfs41_impl_id *impl_id = nfss->nfs_client->cl_implid;
588 seq_printf(m, "\n\timpl_id:\tname='%s',domain='%s',"
590 impl_id->name, impl_id->domain,
591 impl_id->date.seconds, impl_id->date.nseconds);
595 #if IS_ENABLED(CONFIG_NFS_V4)
596 static void show_pnfs(struct seq_file *m, struct nfs_server *server)
600 static void show_implementation_id(struct seq_file *m, struct nfs_server *nfss)
605 int nfs_show_devname(struct seq_file *m, struct dentry *root)
607 char *page = (char *) __get_free_page(GFP_KERNEL);
608 char *devname, *dummy;
612 devname = nfs_path(&dummy, root, page, PAGE_SIZE, 0);
614 err = PTR_ERR(devname);
616 seq_escape(m, devname, " \t\n\\");
617 free_page((unsigned long)page);
620 EXPORT_SYMBOL_GPL(nfs_show_devname);
622 int nfs_show_path(struct seq_file *m, struct dentry *dentry)
627 EXPORT_SYMBOL_GPL(nfs_show_path);
630 * Present statistical information for this VFS mountpoint
632 int nfs_show_stats(struct seq_file *m, struct dentry *root)
635 struct nfs_server *nfss = NFS_SB(root->d_sb);
636 struct rpc_auth *auth = nfss->client->cl_auth;
637 struct nfs_iostats totals = { };
639 seq_printf(m, "statvers=%s", NFS_IOSTAT_VERS);
642 * Display all mount option settings
644 seq_puts(m, "\n\topts:\t");
645 seq_puts(m, sb_rdonly(root->d_sb) ? "ro" : "rw");
646 seq_puts(m, root->d_sb->s_flags & SB_SYNCHRONOUS ? ",sync" : "");
647 seq_puts(m, root->d_sb->s_flags & SB_NOATIME ? ",noatime" : "");
648 seq_puts(m, root->d_sb->s_flags & SB_NODIRATIME ? ",nodiratime" : "");
649 nfs_show_mount_options(m, nfss, 1);
651 seq_printf(m, "\n\tage:\t%lu", (jiffies - nfss->mount_time) / HZ);
653 show_implementation_id(m, nfss);
655 seq_puts(m, "\n\tcaps:\t");
656 seq_printf(m, "caps=0x%x", nfss->caps);
657 seq_printf(m, ",wtmult=%u", nfss->wtmult);
658 seq_printf(m, ",dtsize=%u", nfss->dtsize);
659 seq_printf(m, ",bsize=%u", nfss->bsize);
660 seq_printf(m, ",namlen=%u", nfss->namelen);
662 #if IS_ENABLED(CONFIG_NFS_V4)
663 if (nfss->nfs_client->rpc_ops->version == 4) {
664 seq_puts(m, "\n\tnfsv4:\t");
665 seq_printf(m, "bm0=0x%x", nfss->attr_bitmask[0]);
666 seq_printf(m, ",bm1=0x%x", nfss->attr_bitmask[1]);
667 seq_printf(m, ",bm2=0x%x", nfss->attr_bitmask[2]);
668 seq_printf(m, ",acl=0x%x", nfss->acl_bitmask);
669 show_sessions(m, nfss);
676 * Display security flavor in effect for this mount
678 seq_printf(m, "\n\tsec:\tflavor=%u", auth->au_ops->au_flavor);
680 seq_printf(m, ",pseudoflavor=%u", auth->au_flavor);
683 * Display superblock I/O counters
685 for_each_possible_cpu(cpu) {
686 struct nfs_iostats *stats;
689 stats = per_cpu_ptr(nfss->io_stats, cpu);
691 for (i = 0; i < __NFSIOS_COUNTSMAX; i++)
692 totals.events[i] += stats->events[i];
693 for (i = 0; i < __NFSIOS_BYTESMAX; i++)
694 totals.bytes[i] += stats->bytes[i];
699 seq_puts(m, "\n\tevents:\t");
700 for (i = 0; i < __NFSIOS_COUNTSMAX; i++)
701 seq_printf(m, "%lu ", totals.events[i]);
702 seq_puts(m, "\n\tbytes:\t");
703 for (i = 0; i < __NFSIOS_BYTESMAX; i++)
704 seq_printf(m, "%Lu ", totals.bytes[i]);
707 rpc_clnt_show_stats(m, nfss->client);
711 EXPORT_SYMBOL_GPL(nfs_show_stats);
714 * Begin unmount by attempting to remove all automounted mountpoints we added
715 * in response to xdev traversals and referrals
717 void nfs_umount_begin(struct super_block *sb)
719 struct nfs_server *server;
720 struct rpc_clnt *rpc;
723 /* -EIO all pending I/O */
724 rpc = server->client_acl;
726 rpc_killall_tasks(rpc);
727 rpc = server->client;
729 rpc_killall_tasks(rpc);
731 EXPORT_SYMBOL_GPL(nfs_umount_begin);
734 * Return true if 'match' is in auth_info or auth_info is empty.
735 * Return false otherwise.
737 bool nfs_auth_info_match(const struct nfs_auth_info *auth_info,
738 rpc_authflavor_t match)
742 if (!auth_info->flavor_len)
745 for (i = 0; i < auth_info->flavor_len; i++) {
746 if (auth_info->flavors[i] == match)
751 EXPORT_SYMBOL_GPL(nfs_auth_info_match);
754 * Ensure that a specified authtype in ctx->auth_info is supported by
755 * the server. Returns 0 and sets ctx->selected_flavor if it's ok, and
758 static int nfs_verify_authflavors(struct nfs_fs_context *ctx,
759 rpc_authflavor_t *server_authlist,
762 rpc_authflavor_t flavor = RPC_AUTH_MAXFLAVOR;
763 bool found_auth_null = false;
767 * If the sec= mount option is used, the specified flavor or AUTH_NULL
768 * must be in the list returned by the server.
770 * AUTH_NULL has a special meaning when it's in the server list - it
771 * means that the server will ignore the rpc creds, so any flavor
772 * can be used but still use the sec= that was specified.
774 * Note also that the MNT procedure in MNTv1 does not return a list
775 * of supported security flavors. In this case, nfs_mount() fabricates
776 * a security flavor list containing just AUTH_NULL.
778 for (i = 0; i < count; i++) {
779 flavor = server_authlist[i];
781 if (nfs_auth_info_match(&ctx->auth_info, flavor))
784 if (flavor == RPC_AUTH_NULL)
785 found_auth_null = true;
788 if (found_auth_null) {
789 flavor = ctx->auth_info.flavors[0];
794 "NFS: specified auth flavors not supported by server\n");
798 ctx->selected_flavor = flavor;
799 dfprintk(MOUNT, "NFS: using auth flavor %u\n", ctx->selected_flavor);
804 * Use the remote server's MOUNT service to request the NFS file handle
805 * corresponding to the provided path.
807 static int nfs_request_mount(struct fs_context *fc,
808 struct nfs_fh *root_fh,
809 rpc_authflavor_t *server_authlist,
810 unsigned int *server_authlist_len)
812 struct nfs_fs_context *ctx = nfs_fc2context(fc);
813 struct nfs_mount_request request = {
814 .sap = &ctx->mount_server._address,
815 .dirpath = ctx->nfs_server.export_path,
816 .protocol = ctx->mount_server.protocol,
818 .noresvport = ctx->flags & NFS_MOUNT_NORESVPORT,
819 .auth_flav_len = server_authlist_len,
820 .auth_flavs = server_authlist,
825 if (ctx->mount_server.version == 0) {
826 switch (ctx->version) {
828 ctx->mount_server.version = NFS_MNT3_VERSION;
831 ctx->mount_server.version = NFS_MNT_VERSION;
834 request.version = ctx->mount_server.version;
836 if (ctx->mount_server.hostname)
837 request.hostname = ctx->mount_server.hostname;
839 request.hostname = ctx->nfs_server.hostname;
842 * Construct the mount server's address.
844 if (ctx->mount_server.address.sa_family == AF_UNSPEC) {
845 memcpy(request.sap, &ctx->nfs_server._address,
846 ctx->nfs_server.addrlen);
847 ctx->mount_server.addrlen = ctx->nfs_server.addrlen;
849 request.salen = ctx->mount_server.addrlen;
850 nfs_set_port(request.sap, &ctx->mount_server.port, 0);
853 * Now ask the mount server to map our export path
856 status = nfs_mount(&request, ctx->timeo, ctx->retrans);
858 dfprintk(MOUNT, "NFS: unable to mount server %s, error %d\n",
859 request.hostname, status);
866 static struct nfs_server *nfs_try_mount_request(struct fs_context *fc)
868 struct nfs_fs_context *ctx = nfs_fc2context(fc);
871 bool tried_auth_unix = false;
872 bool auth_null_in_list = false;
873 struct nfs_server *server = ERR_PTR(-EACCES);
874 rpc_authflavor_t authlist[NFS_MAX_SECFLAVORS];
875 unsigned int authlist_len = ARRAY_SIZE(authlist);
877 status = nfs_request_mount(fc, ctx->mntfh, authlist, &authlist_len);
879 return ERR_PTR(status);
882 * Was a sec= authflavor specified in the options? First, verify
883 * whether the server supports it, and then just try to use it if so.
885 if (ctx->auth_info.flavor_len > 0) {
886 status = nfs_verify_authflavors(ctx, authlist, authlist_len);
887 dfprintk(MOUNT, "NFS: using auth flavor %u\n",
888 ctx->selected_flavor);
890 return ERR_PTR(status);
891 return ctx->nfs_mod->rpc_ops->create_server(fc);
895 * No sec= option was provided. RFC 2623, section 2.7 suggests we
896 * SHOULD prefer the flavor listed first. However, some servers list
897 * AUTH_NULL first. Avoid ever choosing AUTH_NULL.
899 for (i = 0; i < authlist_len; ++i) {
900 rpc_authflavor_t flavor;
901 struct rpcsec_gss_info info;
903 flavor = authlist[i];
906 tried_auth_unix = true;
909 auth_null_in_list = true;
912 if (rpcauth_get_gssinfo(flavor, &info) != 0)
916 dfprintk(MOUNT, "NFS: attempting to use auth flavor %u\n", flavor);
917 ctx->selected_flavor = flavor;
918 server = ctx->nfs_mod->rpc_ops->create_server(fc);
924 * Nothing we tried so far worked. At this point, give up if we've
925 * already tried AUTH_UNIX or if the server's list doesn't contain
928 if (tried_auth_unix || !auth_null_in_list)
931 /* Last chance! Try AUTH_UNIX */
932 dfprintk(MOUNT, "NFS: attempting to use auth flavor %u\n", RPC_AUTH_UNIX);
933 ctx->selected_flavor = RPC_AUTH_UNIX;
934 return ctx->nfs_mod->rpc_ops->create_server(fc);
937 int nfs_try_get_tree(struct fs_context *fc)
939 struct nfs_fs_context *ctx = nfs_fc2context(fc);
942 ctx->server = nfs_try_mount_request(fc);
944 ctx->server = ctx->nfs_mod->rpc_ops->create_server(fc);
946 return nfs_get_tree_common(fc);
948 EXPORT_SYMBOL_GPL(nfs_try_get_tree);
951 #define NFS_REMOUNT_CMP_FLAGMASK ~(NFS_MOUNT_INTR \
955 | NFS_MOUNT_KERBEROS \
957 | NFS_MOUNT_BROKEN_SUID \
958 | NFS_MOUNT_STRICTLOCK \
959 | NFS_MOUNT_LEGACY_INTERFACE)
961 #define NFS_MOUNT_CMP_FLAGMASK (NFS_REMOUNT_CMP_FLAGMASK & \
962 ~(NFS_MOUNT_UNSHARED | NFS_MOUNT_NORESVPORT))
965 nfs_compare_remount_data(struct nfs_server *nfss,
966 struct nfs_fs_context *ctx)
968 if ((ctx->flags ^ nfss->flags) & NFS_REMOUNT_CMP_FLAGMASK ||
969 ctx->rsize != nfss->rsize ||
970 ctx->wsize != nfss->wsize ||
971 ctx->version != nfss->nfs_client->rpc_ops->version ||
972 ctx->minorversion != nfss->nfs_client->cl_minorversion ||
973 ctx->retrans != nfss->client->cl_timeout->to_retries ||
974 !nfs_auth_info_match(&ctx->auth_info, nfss->client->cl_auth->au_flavor) ||
975 ctx->acregmin != nfss->acregmin / HZ ||
976 ctx->acregmax != nfss->acregmax / HZ ||
977 ctx->acdirmin != nfss->acdirmin / HZ ||
978 ctx->acdirmax != nfss->acdirmax / HZ ||
979 ctx->timeo != (10U * nfss->client->cl_timeout->to_initval / HZ) ||
980 (ctx->options & NFS_OPTION_FSCACHE) != (nfss->options & NFS_OPTION_FSCACHE) ||
981 ctx->nfs_server.port != nfss->port ||
982 ctx->nfs_server.addrlen != nfss->nfs_client->cl_addrlen ||
983 !rpc_cmp_addr((struct sockaddr *)&ctx->nfs_server.address,
984 (struct sockaddr *)&nfss->nfs_client->cl_addr))
990 int nfs_reconfigure(struct fs_context *fc)
992 struct nfs_fs_context *ctx = nfs_fc2context(fc);
993 struct super_block *sb = fc->root->d_sb;
994 struct nfs_server *nfss = sb->s_fs_info;
1000 * Userspace mount programs that send binary options generally send
1001 * them populated with default values. We have no way to know which
1002 * ones were explicitly specified. Fall back to legacy behavior and
1003 * just return success.
1005 if (ctx->skip_reconfig_option_check)
1009 * noac is a special case. It implies -o sync, but that's not
1010 * necessarily reflected in the mtab options. reconfigure_super
1011 * will clear SB_SYNCHRONOUS if -o sync wasn't specified in the
1012 * remount options, so we have to explicitly reset it.
1014 if (ctx->flags & NFS_MOUNT_NOAC) {
1015 fc->sb_flags |= SB_SYNCHRONOUS;
1016 fc->sb_flags_mask |= SB_SYNCHRONOUS;
1019 /* compare new mount options with old ones */
1020 ret = nfs_compare_remount_data(nfss, ctx);
1024 return nfs_probe_server(nfss, NFS_FH(d_inode(fc->root)));
1026 EXPORT_SYMBOL_GPL(nfs_reconfigure);
1029 * Finish setting up an NFS superblock
1031 static void nfs_fill_super(struct super_block *sb, struct nfs_fs_context *ctx)
1033 struct nfs_server *server = NFS_SB(sb);
1035 sb->s_blocksize_bits = 0;
1036 sb->s_blocksize = 0;
1037 sb->s_xattr = server->nfs_client->cl_nfs_mod->xattr;
1038 sb->s_op = server->nfs_client->cl_nfs_mod->sops;
1040 sb->s_blocksize = nfs_block_size(ctx->bsize, &sb->s_blocksize_bits);
1042 switch (server->nfs_client->rpc_ops->version) {
1044 sb->s_time_gran = 1000;
1046 sb->s_time_max = U32_MAX;
1050 * The VFS shouldn't apply the umask to mode bits.
1051 * We will do so ourselves when necessary.
1053 sb->s_flags |= SB_POSIXACL;
1054 sb->s_time_gran = 1;
1056 sb->s_time_max = U32_MAX;
1057 sb->s_export_op = &nfs_export_ops;
1060 sb->s_flags |= SB_POSIXACL;
1061 sb->s_time_gran = 1;
1062 sb->s_time_min = S64_MIN;
1063 sb->s_time_max = S64_MAX;
1064 if (server->caps & NFS_CAP_ATOMIC_OPEN_V1)
1065 sb->s_export_op = &nfs_export_ops;
1069 sb->s_magic = NFS_SUPER_MAGIC;
1071 /* We probably want something more informative here */
1072 snprintf(sb->s_id, sizeof(sb->s_id),
1073 "%u:%u", MAJOR(sb->s_dev), MINOR(sb->s_dev));
1075 if (sb->s_blocksize == 0)
1076 sb->s_blocksize = nfs_block_bits(server->wsize,
1077 &sb->s_blocksize_bits);
1079 nfs_super_set_maxbytes(sb, server->maxfilesize);
1080 server->has_sec_mnt_opts = ctx->has_sec_mnt_opts;
1083 static int nfs_compare_mount_options(const struct super_block *s, const struct nfs_server *b,
1084 const struct fs_context *fc)
1086 const struct nfs_server *a = s->s_fs_info;
1087 const struct rpc_clnt *clnt_a = a->client;
1088 const struct rpc_clnt *clnt_b = b->client;
1090 if ((s->s_flags & NFS_SB_MASK) != (fc->sb_flags & NFS_SB_MASK))
1092 if (a->nfs_client != b->nfs_client)
1094 if ((a->flags ^ b->flags) & NFS_MOUNT_CMP_FLAGMASK)
1096 if (a->wsize != b->wsize)
1098 if (a->rsize != b->rsize)
1100 if (a->acregmin != b->acregmin)
1102 if (a->acregmax != b->acregmax)
1104 if (a->acdirmin != b->acdirmin)
1106 if (a->acdirmax != b->acdirmax)
1108 if (clnt_a->cl_auth->au_flavor != clnt_b->cl_auth->au_flavor)
1115 static int nfs_set_super(struct super_block *s, struct fs_context *fc)
1117 struct nfs_server *server = fc->s_fs_info;
1120 s->s_d_op = server->nfs_client->rpc_ops->dentry_ops;
1121 ret = set_anon_super(s, server);
1123 server->s_dev = s->s_dev;
1127 static int nfs_compare_super_address(struct nfs_server *server1,
1128 struct nfs_server *server2)
1130 struct sockaddr *sap1, *sap2;
1131 struct rpc_xprt *xprt1 = server1->client->cl_xprt;
1132 struct rpc_xprt *xprt2 = server2->client->cl_xprt;
1134 if (!net_eq(xprt1->xprt_net, xprt2->xprt_net))
1137 sap1 = (struct sockaddr *)&server1->nfs_client->cl_addr;
1138 sap2 = (struct sockaddr *)&server2->nfs_client->cl_addr;
1140 if (sap1->sa_family != sap2->sa_family)
1143 switch (sap1->sa_family) {
1145 struct sockaddr_in *sin1 = (struct sockaddr_in *)sap1;
1146 struct sockaddr_in *sin2 = (struct sockaddr_in *)sap2;
1147 if (sin1->sin_addr.s_addr != sin2->sin_addr.s_addr)
1149 if (sin1->sin_port != sin2->sin_port)
1154 struct sockaddr_in6 *sin1 = (struct sockaddr_in6 *)sap1;
1155 struct sockaddr_in6 *sin2 = (struct sockaddr_in6 *)sap2;
1156 if (!ipv6_addr_equal(&sin1->sin6_addr, &sin2->sin6_addr))
1158 if (sin1->sin6_port != sin2->sin6_port)
1169 static int nfs_compare_userns(const struct nfs_server *old,
1170 const struct nfs_server *new)
1172 const struct user_namespace *oldns = &init_user_ns;
1173 const struct user_namespace *newns = &init_user_ns;
1175 if (old->client && old->client->cl_cred)
1176 oldns = old->client->cl_cred->user_ns;
1177 if (new->client && new->client->cl_cred)
1178 newns = new->client->cl_cred->user_ns;
1184 static int nfs_compare_super(struct super_block *sb, struct fs_context *fc)
1186 struct nfs_server *server = fc->s_fs_info, *old = NFS_SB(sb);
1188 if (!nfs_compare_super_address(old, server))
1190 /* Note: NFS_MOUNT_UNSHARED == NFS4_MOUNT_UNSHARED */
1191 if (old->flags & NFS_MOUNT_UNSHARED)
1193 if (memcmp(&old->fsid, &server->fsid, sizeof(old->fsid)) != 0)
1195 if (!nfs_compare_userns(old, server))
1197 if ((old->has_sec_mnt_opts || fc->security) &&
1198 security_sb_mnt_opts_compat(sb, fc->security))
1200 return nfs_compare_mount_options(sb, server, fc);
1203 #ifdef CONFIG_NFS_FSCACHE
1204 static int nfs_get_cache_cookie(struct super_block *sb,
1205 struct nfs_fs_context *ctx)
1207 struct nfs_server *nfss = NFS_SB(sb);
1211 nfss->fscache = NULL;
1216 if (ctx->clone_data.sb) {
1217 struct nfs_server *mnt_s = NFS_SB(ctx->clone_data.sb);
1218 if (!(mnt_s->options & NFS_OPTION_FSCACHE))
1220 if (mnt_s->fscache_uniq) {
1221 uniq = mnt_s->fscache_uniq;
1222 ulen = strlen(uniq);
1225 if (!(ctx->options & NFS_OPTION_FSCACHE))
1227 if (ctx->fscache_uniq) {
1228 uniq = ctx->fscache_uniq;
1229 ulen = strlen(ctx->fscache_uniq);
1233 return nfs_fscache_get_super_cookie(sb, uniq, ulen);
1236 static int nfs_get_cache_cookie(struct super_block *sb,
1237 struct nfs_fs_context *ctx)
1243 int nfs_get_tree_common(struct fs_context *fc)
1245 struct nfs_fs_context *ctx = nfs_fc2context(fc);
1246 struct super_block *s;
1247 int (*compare_super)(struct super_block *, struct fs_context *) = nfs_compare_super;
1248 struct nfs_server *server = ctx->server;
1253 return PTR_ERR(server);
1255 if (server->flags & NFS_MOUNT_UNSHARED)
1256 compare_super = NULL;
1258 /* -o noac implies -o sync */
1259 if (server->flags & NFS_MOUNT_NOAC)
1260 fc->sb_flags |= SB_SYNCHRONOUS;
1262 if (ctx->clone_data.sb)
1263 if (ctx->clone_data.sb->s_flags & SB_SYNCHRONOUS)
1264 fc->sb_flags |= SB_SYNCHRONOUS;
1266 /* Get a superblock - note that we may end up sharing one that already exists */
1267 fc->s_fs_info = server;
1268 s = sget_fc(fc, compare_super, nfs_set_super);
1269 fc->s_fs_info = NULL;
1272 nfs_errorf(fc, "NFS: Couldn't get superblock");
1276 if (s->s_fs_info != server) {
1277 nfs_free_server(server);
1280 error = super_setup_bdi_name(s, "%u:%u", MAJOR(server->s_dev),
1281 MINOR(server->s_dev));
1283 goto error_splat_super;
1284 s->s_bdi->io_pages = server->rpages;
1289 unsigned bsize = ctx->clone_data.inherited_bsize;
1290 /* initial superblock/root creation */
1291 nfs_fill_super(s, ctx);
1293 s->s_blocksize_bits = bsize;
1294 s->s_blocksize = 1U << bsize;
1296 error = nfs_get_cache_cookie(s, ctx);
1298 goto error_splat_super;
1301 error = nfs_get_root(s, fc);
1303 nfs_errorf(fc, "NFS: Couldn't get root dentry");
1304 goto error_splat_super;
1307 s->s_flags |= SB_ACTIVE;
1314 nfs_free_server(server);
1317 deactivate_locked_super(s);
1322 * Destroy an NFS2/3 superblock
1324 void nfs_kill_super(struct super_block *s)
1326 struct nfs_server *server = NFS_SB(s);
1327 dev_t dev = s->s_dev;
1329 generic_shutdown_super(s);
1331 nfs_fscache_release_super_cookie(s);
1333 nfs_free_server(server);
1334 free_anon_bdev(dev);
1336 EXPORT_SYMBOL_GPL(nfs_kill_super);
1338 #if IS_ENABLED(CONFIG_NFS_V4)
1341 * NFS v4 module parameters need to stay in the
1342 * NFS client for backwards compatibility
1344 unsigned int nfs_callback_set_tcpport;
1345 unsigned short nfs_callback_nr_threads;
1346 /* Default cache timeout is 10 minutes */
1347 unsigned int nfs_idmap_cache_timeout = 600;
1348 /* Turn off NFSv4 uid/gid mapping when using AUTH_SYS */
1349 bool nfs4_disable_idmapping = true;
1350 unsigned short max_session_slots = NFS4_DEF_SLOT_TABLE_SIZE;
1351 unsigned short max_session_cb_slots = NFS4_DEF_CB_SLOT_TABLE_SIZE;
1352 unsigned short send_implementation_id = 1;
1353 char nfs4_client_id_uniquifier[NFS4_CLIENT_ID_UNIQ_LEN] = "";
1354 bool recover_lost_locks = false;
1356 EXPORT_SYMBOL_GPL(nfs_callback_nr_threads);
1357 EXPORT_SYMBOL_GPL(nfs_callback_set_tcpport);
1358 EXPORT_SYMBOL_GPL(nfs_idmap_cache_timeout);
1359 EXPORT_SYMBOL_GPL(nfs4_disable_idmapping);
1360 EXPORT_SYMBOL_GPL(max_session_slots);
1361 EXPORT_SYMBOL_GPL(max_session_cb_slots);
1362 EXPORT_SYMBOL_GPL(send_implementation_id);
1363 EXPORT_SYMBOL_GPL(nfs4_client_id_uniquifier);
1364 EXPORT_SYMBOL_GPL(recover_lost_locks);
1366 #define NFS_CALLBACK_MAXPORTNR (65535U)
1368 static int param_set_portnr(const char *val, const struct kernel_param *kp)
1375 ret = kstrtoul(val, 0, &num);
1376 if (ret || num > NFS_CALLBACK_MAXPORTNR)
1378 *((unsigned int *)kp->arg) = num;
1381 static const struct kernel_param_ops param_ops_portnr = {
1382 .set = param_set_portnr,
1383 .get = param_get_uint,
1385 #define param_check_portnr(name, p) __param_check(name, p, unsigned int)
1387 module_param_named(callback_tcpport, nfs_callback_set_tcpport, portnr, 0644);
1388 module_param_named(callback_nr_threads, nfs_callback_nr_threads, ushort, 0644);
1389 MODULE_PARM_DESC(callback_nr_threads, "Number of threads that will be "
1390 "assigned to the NFSv4 callback channels.");
1391 module_param(nfs_idmap_cache_timeout, int, 0644);
1392 module_param(nfs4_disable_idmapping, bool, 0644);
1393 module_param_string(nfs4_unique_id, nfs4_client_id_uniquifier,
1394 NFS4_CLIENT_ID_UNIQ_LEN, 0600);
1395 MODULE_PARM_DESC(nfs4_disable_idmapping,
1396 "Turn off NFSv4 idmapping when using 'sec=sys'");
1397 module_param(max_session_slots, ushort, 0644);
1398 MODULE_PARM_DESC(max_session_slots, "Maximum number of outstanding NFSv4.1 "
1399 "requests the client will negotiate");
1400 module_param(max_session_cb_slots, ushort, 0644);
1401 MODULE_PARM_DESC(max_session_cb_slots, "Maximum number of parallel NFSv4.1 "
1402 "callbacks the client will process for a given server");
1403 module_param(send_implementation_id, ushort, 0644);
1404 MODULE_PARM_DESC(send_implementation_id,
1405 "Send implementation ID with NFSv4.1 exchange_id");
1406 MODULE_PARM_DESC(nfs4_unique_id, "nfs_client_id4 uniquifier string");
1408 module_param(recover_lost_locks, bool, 0644);
1409 MODULE_PARM_DESC(recover_lost_locks,
1410 "If the server reports that a lock might be lost, "
1411 "try to recover it risking data corruption.");
1414 #endif /* CONFIG_NFS_V4 */