1 // SPDX-License-Identifier: GPL-2.0-only
3 * Syscall interface to knfsd.
5 * Copyright (C) 1995, 1996 Olaf Kirch <okir@monad.swb.de>
8 #include <linux/slab.h>
9 #include <linux/namei.h>
10 #include <linux/ctype.h>
11 #include <linux/fs_context.h>
13 #include <linux/sunrpc/svcsock.h>
14 #include <linux/lockd/lockd.h>
15 #include <linux/sunrpc/addr.h>
16 #include <linux/sunrpc/gss_api.h>
17 #include <linux/sunrpc/rpc_pipe_fs.h>
18 #include <linux/module.h>
19 #include <linux/fsnotify.h>
27 #include "filecache.h"
32 * We have a single directory with several nodes in it.
45 NFSD_Reply_Cache_Stats,
61 * write() for these nodes.
63 static ssize_t write_filehandle(struct file *file, char *buf, size_t size);
64 static ssize_t write_unlock_ip(struct file *file, char *buf, size_t size);
65 static ssize_t write_unlock_fs(struct file *file, char *buf, size_t size);
66 static ssize_t write_threads(struct file *file, char *buf, size_t size);
67 static ssize_t write_pool_threads(struct file *file, char *buf, size_t size);
68 static ssize_t write_versions(struct file *file, char *buf, size_t size);
69 static ssize_t write_ports(struct file *file, char *buf, size_t size);
70 static ssize_t write_maxblksize(struct file *file, char *buf, size_t size);
71 static ssize_t write_maxconn(struct file *file, char *buf, size_t size);
73 static ssize_t write_leasetime(struct file *file, char *buf, size_t size);
74 static ssize_t write_gracetime(struct file *file, char *buf, size_t size);
75 #ifdef CONFIG_NFSD_LEGACY_CLIENT_TRACKING
76 static ssize_t write_recoverydir(struct file *file, char *buf, size_t size);
78 static ssize_t write_v4_end_grace(struct file *file, char *buf, size_t size);
81 static ssize_t (*const write_op[])(struct file *, char *, size_t) = {
82 [NFSD_Fh] = write_filehandle,
83 [NFSD_FO_UnlockIP] = write_unlock_ip,
84 [NFSD_FO_UnlockFS] = write_unlock_fs,
85 [NFSD_Threads] = write_threads,
86 [NFSD_Pool_Threads] = write_pool_threads,
87 [NFSD_Versions] = write_versions,
88 [NFSD_Ports] = write_ports,
89 [NFSD_MaxBlkSize] = write_maxblksize,
90 [NFSD_MaxConnections] = write_maxconn,
92 [NFSD_Leasetime] = write_leasetime,
93 [NFSD_Gracetime] = write_gracetime,
94 #ifdef CONFIG_NFSD_LEGACY_CLIENT_TRACKING
95 [NFSD_RecoveryDir] = write_recoverydir,
97 [NFSD_V4EndGrace] = write_v4_end_grace,
101 static ssize_t nfsctl_transaction_write(struct file *file, const char __user *buf, size_t size, loff_t *pos)
103 ino_t ino = file_inode(file)->i_ino;
107 if (ino >= ARRAY_SIZE(write_op) || !write_op[ino])
110 data = simple_transaction_get(file, buf, size);
112 return PTR_ERR(data);
114 rv = write_op[ino](file, data, size);
118 simple_transaction_set(file, rv);
122 static ssize_t nfsctl_transaction_read(struct file *file, char __user *buf, size_t size, loff_t *pos)
124 if (! file->private_data) {
125 /* An attempt to read a transaction file without writing
126 * causes a 0-byte write so that the file can return
129 ssize_t rv = nfsctl_transaction_write(file, buf, 0, pos);
133 return simple_transaction_read(file, buf, size, pos);
136 static const struct file_operations transaction_ops = {
137 .write = nfsctl_transaction_write,
138 .read = nfsctl_transaction_read,
139 .release = simple_transaction_release,
140 .llseek = default_llseek,
143 static int exports_net_open(struct net *net, struct file *file)
146 struct seq_file *seq;
147 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
149 err = seq_open(file, &nfs_exports_op);
153 seq = file->private_data;
154 seq->private = nn->svc_export_cache;
158 static int exports_nfsd_open(struct inode *inode, struct file *file)
160 return exports_net_open(inode->i_sb->s_fs_info, file);
163 static const struct file_operations exports_nfsd_operations = {
164 .open = exports_nfsd_open,
167 .release = seq_release,
170 static int export_features_show(struct seq_file *m, void *v)
172 seq_printf(m, "0x%x 0x%x\n", NFSEXP_ALLFLAGS, NFSEXP_SECINFO_FLAGS);
176 DEFINE_SHOW_ATTRIBUTE(export_features);
178 static const struct file_operations pool_stats_operations = {
179 .open = nfsd_pool_stats_open,
182 .release = seq_release,
185 DEFINE_SHOW_ATTRIBUTE(nfsd_reply_cache_stats);
187 DEFINE_SHOW_ATTRIBUTE(nfsd_file_cache_stats);
189 /*----------------------------------------------------------------------------*/
191 * payload - write methods
194 static inline struct net *netns(struct file *file)
196 return file_inode(file)->i_sb->s_fs_info;
200 * write_unlock_ip - Release all locks used by a client
205 * buf: '\n'-terminated C string containing a
206 * presentation format IP address
207 * size: length of C string in @buf
209 * On success: returns zero if all specified locks were released;
210 * returns one if one or more locks were not released
211 * On error: return code is negative errno value
213 static ssize_t write_unlock_ip(struct file *file, char *buf, size_t size)
215 struct sockaddr_storage address;
216 struct sockaddr *sap = (struct sockaddr *)&address;
217 size_t salen = sizeof(address);
219 struct net *net = netns(file);
225 if (buf[size-1] != '\n')
229 if (qword_get(&buf, fo_path, size) < 0)
232 if (rpc_pton(net, fo_path, size, sap, salen) == 0)
235 trace_nfsd_ctl_unlock_ip(net, buf);
236 return nlmsvc_unlock_all_by_ip(sap);
240 * write_unlock_fs - Release all locks on a local file system
245 * buf: '\n'-terminated C string containing the
246 * absolute pathname of a local file system
247 * size: length of C string in @buf
249 * On success: returns zero if all specified locks were released;
250 * returns one if one or more locks were not released
251 * On error: return code is negative errno value
253 static ssize_t write_unlock_fs(struct file *file, char *buf, size_t size)
263 if (buf[size-1] != '\n')
267 if (qword_get(&buf, fo_path, size) < 0)
269 trace_nfsd_ctl_unlock_fs(netns(file), fo_path);
270 error = kern_path(fo_path, 0, &path);
275 * XXX: Needs better sanity checking. Otherwise we could end up
276 * releasing locks on the wrong file system.
279 * 1. Does the path refer to a directory?
280 * 2. Is that directory a mount point, or
281 * 3. Is that directory the root of an exported file system?
283 error = nlmsvc_unlock_all_by_sb(path.dentry->d_sb);
290 * write_filehandle - Get a variable-length NFS file handle by path
292 * On input, the buffer contains a '\n'-terminated C string comprised of
293 * three alphanumeric words separated by whitespace. The string may
294 * contain escape sequences.
298 * domain: client domain name
299 * path: export pathname
300 * maxsize: numeric maximum size of
302 * size: length of C string in @buf
304 * On success: passed-in buffer filled with '\n'-terminated C
305 * string containing a ASCII hex text version
306 * of the NFS file handle;
307 * return code is the size in bytes of the string
308 * On error: return code is negative errno value
310 static ssize_t write_filehandle(struct file *file, char *buf, size_t size)
316 struct auth_domain *dom;
322 if (buf[size-1] != '\n')
327 len = qword_get(&mesg, dname, size);
332 len = qword_get(&mesg, path, size);
336 len = get_int(&mesg, &maxsize);
340 if (maxsize < NFS_FHSIZE)
342 maxsize = min(maxsize, NFS3_FHSIZE);
344 if (qword_get(&mesg, mesg, size) > 0)
347 trace_nfsd_ctl_filehandle(netns(file), dname, path, maxsize);
349 /* we have all the words, they are in buf.. */
350 dom = unix_domain_find(dname);
354 len = exp_rootfh(netns(file), dom, path, &fh, maxsize);
355 auth_domain_put(dom);
360 len = SIMPLE_TRANSACTION_LIMIT;
361 qword_addhex(&mesg, &len, fh.fh_raw, fh.fh_size);
367 * write_threads - Start NFSD, or report the current number of running threads
373 * On success: passed-in buffer filled with '\n'-terminated C
374 * string numeric value representing the number of
375 * running NFSD threads;
376 * return code is the size in bytes of the string
377 * On error: return code is zero
382 * buf: C string containing an unsigned
383 * integer value representing the
384 * number of NFSD threads to start
385 * size: non-zero length of C string in @buf
387 * On success: NFS service is started;
388 * passed-in buffer filled with '\n'-terminated C
389 * string numeric value representing the number of
390 * running NFSD threads;
391 * return code is the size in bytes of the string
392 * On error: return code is zero or a negative errno value
394 static ssize_t write_threads(struct file *file, char *buf, size_t size)
398 struct net *net = netns(file);
402 rv = get_int(&mesg, &newthreads);
407 trace_nfsd_ctl_threads(net, newthreads);
408 rv = nfsd_svc(newthreads, net, file->f_cred);
412 rv = nfsd_nrthreads(net);
414 return scnprintf(buf, SIMPLE_TRANSACTION_LIMIT, "%d\n", rv);
418 * write_pool_threads - Set or report the current number of threads per pool
427 * buf: C string containing whitespace-
428 * separated unsigned integer values
429 * representing the number of NFSD
430 * threads to start in each pool
431 * size: non-zero length of C string in @buf
433 * On success: passed-in buffer filled with '\n'-terminated C
434 * string containing integer values representing the
435 * number of NFSD threads in each pool;
436 * return code is the size in bytes of the string
437 * On error: return code is zero or a negative errno value
439 static ssize_t write_pool_threads(struct file *file, char *buf, size_t size)
441 /* if size > 0, look for an array of number of threads per node
442 * and apply them then write out number of threads per node as reply
450 struct net *net = netns(file);
452 mutex_lock(&nfsd_mutex);
453 npools = nfsd_nrpools(net);
456 * NFS is shut down. The admin can start it by
457 * writing to the threads file but NOT the pool_threads
458 * file, sorry. Report zero threads.
460 mutex_unlock(&nfsd_mutex);
465 nthreads = kcalloc(npools, sizeof(int), GFP_KERNEL);
467 if (nthreads == NULL)
471 for (i = 0; i < npools; i++) {
472 rv = get_int(&mesg, &nthreads[i]);
474 break; /* fewer numbers than pools */
476 goto out_free; /* syntax error */
480 trace_nfsd_ctl_pool_threads(net, i, nthreads[i]);
482 rv = nfsd_set_nrthreads(i, nthreads, net);
487 rv = nfsd_get_nrthreads(npools, nthreads, net);
492 size = SIMPLE_TRANSACTION_LIMIT;
493 for (i = 0; i < npools && size > 0; i++) {
494 snprintf(mesg, size, "%d%c", nthreads[i], (i == npools-1 ? '\n' : ' '));
502 mutex_unlock(&nfsd_mutex);
507 nfsd_print_version_support(struct nfsd_net *nn, char *buf, int remaining,
508 const char *sep, unsigned vers, int minor)
510 const char *format = minor < 0 ? "%s%c%u" : "%s%c%u.%u";
511 bool supported = !!nfsd_vers(nn, vers, NFSD_TEST);
513 if (vers == 4 && minor >= 0 &&
514 !nfsd_minorversion(nn, minor, NFSD_TEST))
516 if (minor == 0 && supported)
518 * special case for backward compatability.
519 * +4.0 is never reported, it is implied by
520 * +4, unless -4.0 is present.
523 return snprintf(buf, remaining, format, sep,
524 supported ? '+' : '-', vers, minor);
527 static ssize_t __write_versions(struct file *file, char *buf, size_t size)
530 char *vers, *minorp, sign;
531 int len, num, remaining;
534 struct nfsd_net *nn = net_generic(netns(file), nfsd_net_id);
538 /* Cannot change versions without updating
539 * nn->nfsd_serv->sv_xdrsize, and reallocing
540 * rq_argp and rq_resp
543 if (buf[size-1] != '\n')
546 trace_nfsd_ctl_version(netns(file), buf);
549 len = qword_get(&mesg, vers, size);
550 if (len <= 0) return -EINVAL;
555 if (sign == '+' || sign == '-')
556 num = simple_strtol((vers+1), &minorp, 0);
558 num = simple_strtol(vers, &minorp, 0);
559 if (*minorp == '.') {
562 if (kstrtouint(minorp+1, 0, &minor) < 0)
566 cmd = sign == '-' ? NFSD_CLEAR : NFSD_SET;
568 #ifdef CONFIG_NFSD_V2
572 nfsd_vers(nn, num, cmd);
575 if (*minorp == '.') {
576 if (nfsd_minorversion(nn, minor, cmd) < 0)
578 } else if ((cmd == NFSD_SET) != nfsd_vers(nn, num, NFSD_TEST)) {
580 * Either we have +4 and no minors are enabled,
581 * or we have -4 and at least one minor is enabled.
582 * In either case, propagate 'cmd' to all minors.
585 while (nfsd_minorversion(nn, minor, cmd) >= 0)
590 /* Ignore requests to disable non-existent versions */
595 } while ((len = qword_get(&mesg, vers, size)) > 0);
596 /* If all get turned off, turn them back on, as
597 * having no versions is BAD
599 nfsd_reset_versions(nn);
602 /* Now write current state into reply buffer */
604 remaining = SIMPLE_TRANSACTION_LIMIT;
605 for (num=2 ; num <= 4 ; num++) {
607 if (!nfsd_vers(nn, num, NFSD_AVAIL))
612 len = nfsd_print_version_support(nn, buf, remaining,
614 if (len >= remaining)
622 } while (num == 4 && minor <= NFSD_SUPPORTED_MINOR_VERSION);
625 len = snprintf(buf, remaining, "\n");
626 if (len >= remaining)
632 * write_versions - Set or report the available NFS protocol versions
638 * On success: passed-in buffer filled with '\n'-terminated C
639 * string containing positive or negative integer
640 * values representing the current status of each
642 * return code is the size in bytes of the string
643 * On error: return code is zero or a negative errno value
648 * buf: C string containing whitespace-
649 * separated positive or negative
650 * integer values representing NFS
651 * protocol versions to enable ("+n")
653 * size: non-zero length of C string in @buf
655 * On success: status of zero or more protocol versions has
656 * been updated; passed-in buffer filled with
657 * '\n'-terminated C string containing positive
658 * or negative integer values representing the
659 * current status of each protocol version;
660 * return code is the size in bytes of the string
661 * On error: return code is zero or a negative errno value
663 static ssize_t write_versions(struct file *file, char *buf, size_t size)
667 mutex_lock(&nfsd_mutex);
668 rv = __write_versions(file, buf, size);
669 mutex_unlock(&nfsd_mutex);
674 * Zero-length write. Return a list of NFSD's current listener
677 static ssize_t __write_ports_names(char *buf, struct net *net)
679 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
681 if (nn->nfsd_serv == NULL)
683 return svc_xprt_names(nn->nfsd_serv, buf, SIMPLE_TRANSACTION_LIMIT);
687 * A single 'fd' number was written, in which case it must be for
688 * a socket of a supported family/protocol, and we use it as an
691 static ssize_t __write_ports_addfd(char *buf, struct net *net, const struct cred *cred)
695 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
696 struct svc_serv *serv;
698 err = get_int(&mesg, &fd);
699 if (err != 0 || fd < 0)
701 trace_nfsd_ctl_ports_addfd(net, fd);
703 err = nfsd_create_serv(net);
707 serv = nn->nfsd_serv;
708 err = svc_addsock(serv, net, fd, buf, SIMPLE_TRANSACTION_LIMIT, cred);
710 if (!serv->sv_nrthreads && list_empty(&nn->nfsd_serv->sv_permsocks))
711 nfsd_destroy_serv(net);
717 * A transport listener is added by writing its transport name and
720 static ssize_t __write_ports_addxprt(char *buf, struct net *net, const struct cred *cred)
723 struct svc_xprt *xprt;
725 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
726 struct svc_serv *serv;
728 if (sscanf(buf, "%15s %5u", transport, &port) != 2)
731 if (port < 1 || port > USHRT_MAX)
733 trace_nfsd_ctl_ports_addxprt(net, transport, port);
735 err = nfsd_create_serv(net);
739 serv = nn->nfsd_serv;
740 err = svc_xprt_create(serv, transport, net,
741 PF_INET, port, SVC_SOCK_ANONYMOUS, cred);
745 err = svc_xprt_create(serv, transport, net,
746 PF_INET6, port, SVC_SOCK_ANONYMOUS, cred);
747 if (err < 0 && err != -EAFNOSUPPORT)
752 xprt = svc_find_xprt(serv, transport, net, PF_INET, port);
754 svc_xprt_close(xprt);
758 if (!serv->sv_nrthreads && list_empty(&nn->nfsd_serv->sv_permsocks))
759 nfsd_destroy_serv(net);
764 static ssize_t __write_ports(struct file *file, char *buf, size_t size,
768 return __write_ports_names(buf, net);
771 return __write_ports_addfd(buf, net, file->f_cred);
774 return __write_ports_addxprt(buf, net, file->f_cred);
780 * write_ports - Pass a socket file descriptor or transport name to listen on
786 * On success: passed-in buffer filled with a '\n'-terminated C
787 * string containing a whitespace-separated list of
788 * named NFSD listeners;
789 * return code is the size in bytes of the string
790 * On error: return code is zero or a negative errno value
795 * buf: C string containing an unsigned
796 * integer value representing a bound
797 * but unconnected socket that is to be
798 * used as an NFSD listener; listen(3)
799 * must be called for a SOCK_STREAM
800 * socket, otherwise it is ignored
801 * size: non-zero length of C string in @buf
803 * On success: NFS service is started;
804 * passed-in buffer filled with a '\n'-terminated C
805 * string containing a unique alphanumeric name of
807 * return code is the size in bytes of the string
808 * On error: return code is a negative errno value
813 * buf: C string containing a transport
814 * name and an unsigned integer value
815 * representing the port to listen on,
816 * separated by whitespace
817 * size: non-zero length of C string in @buf
819 * On success: returns zero; NFS service is started
820 * On error: return code is a negative errno value
822 static ssize_t write_ports(struct file *file, char *buf, size_t size)
826 mutex_lock(&nfsd_mutex);
827 rv = __write_ports(file, buf, size, netns(file));
828 mutex_unlock(&nfsd_mutex);
833 int nfsd_max_blksize;
836 * write_maxblksize - Set or report the current NFS blksize
845 * buf: C string containing an unsigned
846 * integer value representing the new
848 * size: non-zero length of C string in @buf
850 * On success: passed-in buffer filled with '\n'-terminated C string
851 * containing numeric value of the current NFS blksize
853 * return code is the size in bytes of the string
854 * On error: return code is zero or a negative errno value
856 static ssize_t write_maxblksize(struct file *file, char *buf, size_t size)
859 struct nfsd_net *nn = net_generic(netns(file), nfsd_net_id);
863 int rv = get_int(&mesg, &bsize);
866 trace_nfsd_ctl_maxblksize(netns(file), bsize);
868 /* force bsize into allowed range and
869 * required alignment.
871 bsize = max_t(int, bsize, 1024);
872 bsize = min_t(int, bsize, NFSSVC_MAXBLKSIZE);
874 mutex_lock(&nfsd_mutex);
876 mutex_unlock(&nfsd_mutex);
879 nfsd_max_blksize = bsize;
880 mutex_unlock(&nfsd_mutex);
883 return scnprintf(buf, SIMPLE_TRANSACTION_LIMIT, "%d\n",
888 * write_maxconn - Set or report the current max number of connections
896 * buf: C string containing an unsigned
897 * integer value representing the new
898 * number of max connections
899 * size: non-zero length of C string in @buf
901 * On success: passed-in buffer filled with '\n'-terminated C string
902 * containing numeric value of max_connections setting
903 * for this net namespace;
904 * return code is the size in bytes of the string
905 * On error: return code is zero or a negative errno value
907 static ssize_t write_maxconn(struct file *file, char *buf, size_t size)
910 struct nfsd_net *nn = net_generic(netns(file), nfsd_net_id);
911 unsigned int maxconn = nn->max_connections;
914 int rv = get_uint(&mesg, &maxconn);
918 trace_nfsd_ctl_maxconn(netns(file), maxconn);
919 nn->max_connections = maxconn;
922 return scnprintf(buf, SIMPLE_TRANSACTION_LIMIT, "%u\n", maxconn);
925 #ifdef CONFIG_NFSD_V4
926 static ssize_t __nfsd4_write_time(struct file *file, char *buf, size_t size,
927 time64_t *time, struct nfsd_net *nn)
929 struct dentry *dentry = file_dentry(file);
936 rv = get_int(&mesg, &i);
939 trace_nfsd_ctl_time(netns(file), dentry->d_name.name,
940 dentry->d_name.len, i);
943 * Some sanity checking. We don't have a reason for
944 * these particular numbers, but problems with the
946 * - Too short: the briefest network outage may
947 * cause clients to lose all their locks. Also,
948 * the frequent polling may be wasteful.
949 * - Too long: do you really want reboot recovery
950 * to take more than an hour? Or to make other
951 * clients wait an hour before being able to
952 * revoke a dead client's locks?
954 if (i < 10 || i > 3600)
959 return scnprintf(buf, SIMPLE_TRANSACTION_LIMIT, "%lld\n", *time);
962 static ssize_t nfsd4_write_time(struct file *file, char *buf, size_t size,
963 time64_t *time, struct nfsd_net *nn)
967 mutex_lock(&nfsd_mutex);
968 rv = __nfsd4_write_time(file, buf, size, time, nn);
969 mutex_unlock(&nfsd_mutex);
974 * write_leasetime - Set or report the current NFSv4 lease time
983 * buf: C string containing an unsigned
984 * integer value representing the new
985 * NFSv4 lease expiry time
986 * size: non-zero length of C string in @buf
988 * On success: passed-in buffer filled with '\n'-terminated C
989 * string containing unsigned integer value of the
990 * current lease expiry time;
991 * return code is the size in bytes of the string
992 * On error: return code is zero or a negative errno value
994 static ssize_t write_leasetime(struct file *file, char *buf, size_t size)
996 struct nfsd_net *nn = net_generic(netns(file), nfsd_net_id);
997 return nfsd4_write_time(file, buf, size, &nn->nfsd4_lease, nn);
1001 * write_gracetime - Set or report current NFSv4 grace period time
1003 * As above, but sets the time of the NFSv4 grace period.
1005 * Note this should never be set to less than the *previous*
1006 * lease-period time, but we don't try to enforce this. (In the common
1007 * case (a new boot), we don't know what the previous lease time was
1010 static ssize_t write_gracetime(struct file *file, char *buf, size_t size)
1012 struct nfsd_net *nn = net_generic(netns(file), nfsd_net_id);
1013 return nfsd4_write_time(file, buf, size, &nn->nfsd4_grace, nn);
1016 #ifdef CONFIG_NFSD_LEGACY_CLIENT_TRACKING
1017 static ssize_t __write_recoverydir(struct file *file, char *buf, size_t size,
1018 struct nfsd_net *nn)
1027 if (size > PATH_MAX || buf[size-1] != '\n')
1032 len = qword_get(&mesg, recdir, size);
1035 trace_nfsd_ctl_recoverydir(netns(file), recdir);
1037 status = nfs4_reset_recoverydir(recdir);
1042 return scnprintf(buf, SIMPLE_TRANSACTION_LIMIT, "%s\n",
1043 nfs4_recoverydir());
1047 * write_recoverydir - Set or report the pathname of the recovery directory
1056 * buf: C string containing the pathname
1057 * of the directory on a local file
1058 * system containing permanent NFSv4
1060 * size: non-zero length of C string in @buf
1062 * On success: passed-in buffer filled with '\n'-terminated C string
1063 * containing the current recovery pathname setting;
1064 * return code is the size in bytes of the string
1065 * On error: return code is zero or a negative errno value
1067 static ssize_t write_recoverydir(struct file *file, char *buf, size_t size)
1070 struct nfsd_net *nn = net_generic(netns(file), nfsd_net_id);
1072 mutex_lock(&nfsd_mutex);
1073 rv = __write_recoverydir(file, buf, size, nn);
1074 mutex_unlock(&nfsd_mutex);
1080 * write_v4_end_grace - release grace period for nfsd's v4.x lock manager
1089 * size: non-zero length of C string in @buf
1091 * passed-in buffer filled with "Y" or "N" with a newline
1092 * and NULL-terminated C string. This indicates whether
1093 * the grace period has ended in the current net
1094 * namespace. Return code is the size in bytes of the
1095 * string. Writing a string that starts with 'Y', 'y', or
1096 * '1' to the file will end the grace period for nfsd's v4
1099 static ssize_t write_v4_end_grace(struct file *file, char *buf, size_t size)
1101 struct nfsd_net *nn = net_generic(netns(file), nfsd_net_id);
1110 trace_nfsd_end_grace(netns(file));
1111 nfsd4_end_grace(nn);
1118 return scnprintf(buf, SIMPLE_TRANSACTION_LIMIT, "%c\n",
1119 nn->grace_ended ? 'Y' : 'N');
1124 /*----------------------------------------------------------------------------*/
1126 * populating the filesystem.
1129 /* Basically copying rpc_get_inode. */
1130 static struct inode *nfsd_get_inode(struct super_block *sb, umode_t mode)
1132 struct inode *inode = new_inode(sb);
1135 /* Following advice from simple_fill_super documentation: */
1136 inode->i_ino = iunique(sb, NFSD_MaxReserved);
1137 inode->i_mode = mode;
1138 simple_inode_init_ts(inode);
1139 switch (mode & S_IFMT) {
1141 inode->i_fop = &simple_dir_operations;
1142 inode->i_op = &simple_dir_inode_operations;
1146 inode->i_op = &simple_symlink_inode_operations;
1154 static int __nfsd_mkdir(struct inode *dir, struct dentry *dentry, umode_t mode, struct nfsdfs_client *ncl)
1156 struct inode *inode;
1158 inode = nfsd_get_inode(dir->i_sb, mode);
1162 inode->i_private = ncl;
1163 kref_get(&ncl->cl_ref);
1165 d_add(dentry, inode);
1167 fsnotify_mkdir(dir, dentry);
1171 static struct dentry *nfsd_mkdir(struct dentry *parent, struct nfsdfs_client *ncl, char *name)
1173 struct inode *dir = parent->d_inode;
1174 struct dentry *dentry;
1178 dentry = d_alloc_name(parent, name);
1181 ret = __nfsd_mkdir(d_inode(parent), dentry, S_IFDIR | 0600, ncl);
1189 dentry = ERR_PTR(ret);
1193 #if IS_ENABLED(CONFIG_SUNRPC_GSS)
1194 static int __nfsd_symlink(struct inode *dir, struct dentry *dentry,
1195 umode_t mode, const char *content)
1197 struct inode *inode;
1199 inode = nfsd_get_inode(dir->i_sb, mode);
1203 inode->i_link = (char *)content;
1204 inode->i_size = strlen(content);
1206 d_add(dentry, inode);
1208 fsnotify_create(dir, dentry);
1213 * @content is assumed to be a NUL-terminated string that lives
1214 * longer than the symlink itself.
1216 static void _nfsd_symlink(struct dentry *parent, const char *name,
1217 const char *content)
1219 struct inode *dir = parent->d_inode;
1220 struct dentry *dentry;
1224 dentry = d_alloc_name(parent, name);
1227 ret = __nfsd_symlink(d_inode(parent), dentry, S_IFLNK | 0777, content);
1234 static inline void _nfsd_symlink(struct dentry *parent, const char *name,
1235 const char *content)
1241 static void clear_ncl(struct dentry *dentry)
1243 struct inode *inode = d_inode(dentry);
1244 struct nfsdfs_client *ncl = inode->i_private;
1246 spin_lock(&inode->i_lock);
1247 inode->i_private = NULL;
1248 spin_unlock(&inode->i_lock);
1249 kref_put(&ncl->cl_ref, ncl->cl_release);
1252 struct nfsdfs_client *get_nfsdfs_client(struct inode *inode)
1254 struct nfsdfs_client *nc;
1256 spin_lock(&inode->i_lock);
1257 nc = inode->i_private;
1259 kref_get(&nc->cl_ref);
1260 spin_unlock(&inode->i_lock);
1264 /* XXX: cut'n'paste from simple_fill_super; figure out if we could share
1266 static int nfsdfs_create_files(struct dentry *root,
1267 const struct tree_descr *files,
1268 struct nfsdfs_client *ncl,
1269 struct dentry **fdentries)
1271 struct inode *dir = d_inode(root);
1272 struct inode *inode;
1273 struct dentry *dentry;
1277 for (i = 0; files->name && files->name[0]; i++, files++) {
1278 dentry = d_alloc_name(root, files->name);
1281 inode = nfsd_get_inode(d_inode(root)->i_sb,
1282 S_IFREG | files->mode);
1287 kref_get(&ncl->cl_ref);
1288 inode->i_fop = files->ops;
1289 inode->i_private = ncl;
1290 d_add(dentry, inode);
1291 fsnotify_create(dir, dentry);
1293 fdentries[i] = dentry;
1302 /* on success, returns positive number unique to that client. */
1303 struct dentry *nfsd_client_mkdir(struct nfsd_net *nn,
1304 struct nfsdfs_client *ncl, u32 id,
1305 const struct tree_descr *files,
1306 struct dentry **fdentries)
1308 struct dentry *dentry;
1312 sprintf(name, "%u", id);
1314 dentry = nfsd_mkdir(nn->nfsd_client_dir, ncl, name);
1315 if (IS_ERR(dentry)) /* XXX: tossing errors? */
1317 ret = nfsdfs_create_files(dentry, files, ncl, fdentries);
1319 nfsd_client_rmdir(dentry);
1325 /* Taken from __rpc_rmdir: */
1326 void nfsd_client_rmdir(struct dentry *dentry)
1328 simple_recursive_removal(dentry, clear_ncl);
1331 static int nfsd_fill_super(struct super_block *sb, struct fs_context *fc)
1333 struct nfsd_net *nn = net_generic(current->nsproxy->net_ns,
1335 struct dentry *dentry;
1338 static const struct tree_descr nfsd_files[] = {
1339 [NFSD_List] = {"exports", &exports_nfsd_operations, S_IRUGO},
1340 /* Per-export io stats use same ops as exports file */
1341 [NFSD_Export_Stats] = {"export_stats", &exports_nfsd_operations, S_IRUGO},
1342 [NFSD_Export_features] = {"export_features",
1343 &export_features_fops, S_IRUGO},
1344 [NFSD_FO_UnlockIP] = {"unlock_ip",
1345 &transaction_ops, S_IWUSR|S_IRUSR},
1346 [NFSD_FO_UnlockFS] = {"unlock_filesystem",
1347 &transaction_ops, S_IWUSR|S_IRUSR},
1348 [NFSD_Fh] = {"filehandle", &transaction_ops, S_IWUSR|S_IRUSR},
1349 [NFSD_Threads] = {"threads", &transaction_ops, S_IWUSR|S_IRUSR},
1350 [NFSD_Pool_Threads] = {"pool_threads", &transaction_ops, S_IWUSR|S_IRUSR},
1351 [NFSD_Pool_Stats] = {"pool_stats", &pool_stats_operations, S_IRUGO},
1352 [NFSD_Reply_Cache_Stats] = {"reply_cache_stats",
1353 &nfsd_reply_cache_stats_fops, S_IRUGO},
1354 [NFSD_Versions] = {"versions", &transaction_ops, S_IWUSR|S_IRUSR},
1355 [NFSD_Ports] = {"portlist", &transaction_ops, S_IWUSR|S_IRUGO},
1356 [NFSD_MaxBlkSize] = {"max_block_size", &transaction_ops, S_IWUSR|S_IRUGO},
1357 [NFSD_MaxConnections] = {"max_connections", &transaction_ops, S_IWUSR|S_IRUGO},
1358 [NFSD_Filecache] = {"filecache", &nfsd_file_cache_stats_fops, S_IRUGO},
1359 #ifdef CONFIG_NFSD_V4
1360 [NFSD_Leasetime] = {"nfsv4leasetime", &transaction_ops, S_IWUSR|S_IRUSR},
1361 [NFSD_Gracetime] = {"nfsv4gracetime", &transaction_ops, S_IWUSR|S_IRUSR},
1362 [NFSD_RecoveryDir] = {"nfsv4recoverydir", &transaction_ops, S_IWUSR|S_IRUSR},
1363 [NFSD_V4EndGrace] = {"v4_end_grace", &transaction_ops, S_IWUSR|S_IRUGO},
1368 ret = simple_fill_super(sb, 0x6e667364, nfsd_files);
1371 _nfsd_symlink(sb->s_root, "supported_krb5_enctypes",
1372 "/proc/net/rpc/gss_krb5_enctypes");
1373 dentry = nfsd_mkdir(sb->s_root, NULL, "clients");
1375 return PTR_ERR(dentry);
1376 nn->nfsd_client_dir = dentry;
1380 static int nfsd_fs_get_tree(struct fs_context *fc)
1382 return get_tree_keyed(fc, nfsd_fill_super, get_net(fc->net_ns));
1385 static void nfsd_fs_free_fc(struct fs_context *fc)
1388 put_net(fc->s_fs_info);
1391 static const struct fs_context_operations nfsd_fs_context_ops = {
1392 .free = nfsd_fs_free_fc,
1393 .get_tree = nfsd_fs_get_tree,
1396 static int nfsd_init_fs_context(struct fs_context *fc)
1398 put_user_ns(fc->user_ns);
1399 fc->user_ns = get_user_ns(fc->net_ns->user_ns);
1400 fc->ops = &nfsd_fs_context_ops;
1404 static void nfsd_umount(struct super_block *sb)
1406 struct net *net = sb->s_fs_info;
1408 nfsd_shutdown_threads(net);
1410 kill_litter_super(sb);
1414 static struct file_system_type nfsd_fs_type = {
1415 .owner = THIS_MODULE,
1417 .init_fs_context = nfsd_init_fs_context,
1418 .kill_sb = nfsd_umount,
1420 MODULE_ALIAS_FS("nfsd");
1422 #ifdef CONFIG_PROC_FS
1424 static int exports_proc_open(struct inode *inode, struct file *file)
1426 return exports_net_open(current->nsproxy->net_ns, file);
1429 static const struct proc_ops exports_proc_ops = {
1430 .proc_open = exports_proc_open,
1431 .proc_read = seq_read,
1432 .proc_lseek = seq_lseek,
1433 .proc_release = seq_release,
1436 static int create_proc_exports_entry(void)
1438 struct proc_dir_entry *entry;
1440 entry = proc_mkdir("fs/nfs", NULL);
1443 entry = proc_create("exports", 0, entry, &exports_proc_ops);
1445 remove_proc_entry("fs/nfs", NULL);
1450 #else /* CONFIG_PROC_FS */
1451 static int create_proc_exports_entry(void)
1457 unsigned int nfsd_net_id;
1460 * nfsd_nl_rpc_status_get_start - Prepare rpc_status_get dumpit
1461 * @cb: netlink metadata and command arguments
1464 * %0: The rpc_status_get command may proceed
1465 * %-ENODEV: There is no NFSD running in this namespace
1467 int nfsd_nl_rpc_status_get_start(struct netlink_callback *cb)
1469 struct nfsd_net *nn = net_generic(sock_net(cb->skb->sk), nfsd_net_id);
1472 mutex_lock(&nfsd_mutex);
1476 mutex_unlock(&nfsd_mutex);
1481 static int nfsd_genl_rpc_status_compose_msg(struct sk_buff *skb,
1482 struct netlink_callback *cb,
1483 struct nfsd_genl_rqstp *rqstp)
1488 hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
1489 &nfsd_nl_family, 0, NFSD_CMD_RPC_STATUS_GET);
1493 if (nla_put_be32(skb, NFSD_A_RPC_STATUS_XID, rqstp->rq_xid) ||
1494 nla_put_u32(skb, NFSD_A_RPC_STATUS_FLAGS, rqstp->rq_flags) ||
1495 nla_put_u32(skb, NFSD_A_RPC_STATUS_PROG, rqstp->rq_prog) ||
1496 nla_put_u32(skb, NFSD_A_RPC_STATUS_PROC, rqstp->rq_proc) ||
1497 nla_put_u8(skb, NFSD_A_RPC_STATUS_VERSION, rqstp->rq_vers) ||
1498 nla_put_s64(skb, NFSD_A_RPC_STATUS_SERVICE_TIME,
1499 ktime_to_us(rqstp->rq_stime),
1500 NFSD_A_RPC_STATUS_PAD))
1503 switch (rqstp->rq_saddr.sa_family) {
1505 const struct sockaddr_in *s_in, *d_in;
1507 s_in = (const struct sockaddr_in *)&rqstp->rq_saddr;
1508 d_in = (const struct sockaddr_in *)&rqstp->rq_daddr;
1509 if (nla_put_in_addr(skb, NFSD_A_RPC_STATUS_SADDR4,
1510 s_in->sin_addr.s_addr) ||
1511 nla_put_in_addr(skb, NFSD_A_RPC_STATUS_DADDR4,
1512 d_in->sin_addr.s_addr) ||
1513 nla_put_be16(skb, NFSD_A_RPC_STATUS_SPORT,
1515 nla_put_be16(skb, NFSD_A_RPC_STATUS_DPORT,
1521 const struct sockaddr_in6 *s_in, *d_in;
1523 s_in = (const struct sockaddr_in6 *)&rqstp->rq_saddr;
1524 d_in = (const struct sockaddr_in6 *)&rqstp->rq_daddr;
1525 if (nla_put_in6_addr(skb, NFSD_A_RPC_STATUS_SADDR6,
1526 &s_in->sin6_addr) ||
1527 nla_put_in6_addr(skb, NFSD_A_RPC_STATUS_DADDR6,
1528 &d_in->sin6_addr) ||
1529 nla_put_be16(skb, NFSD_A_RPC_STATUS_SPORT,
1531 nla_put_be16(skb, NFSD_A_RPC_STATUS_DPORT,
1538 for (i = 0; i < rqstp->rq_opcnt; i++)
1539 if (nla_put_u32(skb, NFSD_A_RPC_STATUS_COMPOUND_OPS,
1540 rqstp->rq_opnum[i]))
1543 genlmsg_end(skb, hdr);
1548 * nfsd_nl_rpc_status_get_dumpit - Handle rpc_status_get dumpit
1549 * @skb: reply buffer
1550 * @cb: netlink metadata and command arguments
1552 * Returns the size of the reply or a negative errno.
1554 int nfsd_nl_rpc_status_get_dumpit(struct sk_buff *skb,
1555 struct netlink_callback *cb)
1557 struct nfsd_net *nn = net_generic(sock_net(skb->sk), nfsd_net_id);
1558 int i, ret, rqstp_index = 0;
1562 for (i = 0; i < nn->nfsd_serv->sv_nrpools; i++) {
1563 struct svc_rqst *rqstp;
1565 if (i < cb->args[0]) /* already consumed */
1569 list_for_each_entry_rcu(rqstp,
1570 &nn->nfsd_serv->sv_pools[i].sp_all_threads,
1572 struct nfsd_genl_rqstp genl_rqstp;
1573 unsigned int status_counter;
1575 if (rqstp_index++ < cb->args[1]) /* already consumed */
1578 * Acquire rq_status_counter before parsing the rqst
1579 * fields. rq_status_counter is set to an odd value in
1580 * order to notify the consumers the rqstp fields are
1584 smp_load_acquire(&rqstp->rq_status_counter);
1585 if (!(status_counter & 1))
1588 genl_rqstp.rq_xid = rqstp->rq_xid;
1589 genl_rqstp.rq_flags = rqstp->rq_flags;
1590 genl_rqstp.rq_vers = rqstp->rq_vers;
1591 genl_rqstp.rq_prog = rqstp->rq_prog;
1592 genl_rqstp.rq_proc = rqstp->rq_proc;
1593 genl_rqstp.rq_stime = rqstp->rq_stime;
1594 genl_rqstp.rq_opcnt = 0;
1595 memcpy(&genl_rqstp.rq_daddr, svc_daddr(rqstp),
1596 sizeof(struct sockaddr));
1597 memcpy(&genl_rqstp.rq_saddr, svc_addr(rqstp),
1598 sizeof(struct sockaddr));
1600 #ifdef CONFIG_NFSD_V4
1601 if (rqstp->rq_vers == NFS4_VERSION &&
1602 rqstp->rq_proc == NFSPROC4_COMPOUND) {
1603 /* NFSv4 compound */
1604 struct nfsd4_compoundargs *args;
1607 args = rqstp->rq_argp;
1608 genl_rqstp.rq_opcnt = args->opcnt;
1609 for (j = 0; j < genl_rqstp.rq_opcnt; j++)
1610 genl_rqstp.rq_opnum[j] =
1613 #endif /* CONFIG_NFSD_V4 */
1616 * Acquire rq_status_counter before reporting the rqst
1617 * fields to the user.
1619 if (smp_load_acquire(&rqstp->rq_status_counter) !=
1623 ret = nfsd_genl_rpc_status_compose_msg(skb, cb,
1631 cb->args[1] = rqstp_index;
1640 * nfsd_nl_rpc_status_get_done - rpc_status_get dumpit post-processing
1641 * @cb: netlink metadata and command arguments
1646 int nfsd_nl_rpc_status_get_done(struct netlink_callback *cb)
1648 mutex_unlock(&nfsd_mutex);
1654 * nfsd_net_init - Prepare the nfsd_net portion of a new net namespace
1655 * @net: a freshly-created network namespace
1657 * This information stays around as long as the network namespace is
1658 * alive whether or not there is an NFSD instance running in the
1661 * Returns zero on success, or a negative errno otherwise.
1663 static __net_init int nfsd_net_init(struct net *net)
1666 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
1668 retval = nfsd_export_init(net);
1670 goto out_export_error;
1671 retval = nfsd_idmap_init(net);
1673 goto out_idmap_error;
1674 retval = nfsd_net_reply_cache_init(nn);
1676 goto out_repcache_error;
1677 nn->nfsd_versions = NULL;
1678 nn->nfsd4_minorversions = NULL;
1679 nfsd4_init_leases_net(nn);
1680 get_random_bytes(&nn->siphash_key, sizeof(nn->siphash_key));
1681 seqlock_init(&nn->writeverf_lock);
1686 nfsd_idmap_shutdown(net);
1688 nfsd_export_shutdown(net);
1694 * nfsd_net_exit - Release the nfsd_net portion of a net namespace
1695 * @net: a network namespace that is about to be destroyed
1698 static __net_exit void nfsd_net_exit(struct net *net)
1700 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
1702 nfsd_net_reply_cache_destroy(nn);
1703 nfsd_idmap_shutdown(net);
1704 nfsd_export_shutdown(net);
1705 nfsd_netns_free_versions(nn);
1708 static struct pernet_operations nfsd_net_ops = {
1709 .init = nfsd_net_init,
1710 .exit = nfsd_net_exit,
1712 .size = sizeof(struct nfsd_net),
1715 static int __init init_nfsd(void)
1719 retval = nfsd4_init_slabs();
1722 retval = nfsd4_init_pnfs();
1724 goto out_free_slabs;
1725 retval = nfsd_stat_init(); /* Statistics */
1728 retval = nfsd_drc_slab_create();
1731 nfsd_lockd_init(); /* lockd->nfsd callbacks */
1732 retval = create_proc_exports_entry();
1734 goto out_free_lockd;
1735 retval = register_pernet_subsys(&nfsd_net_ops);
1737 goto out_free_exports;
1738 retval = register_cld_notifier();
1740 goto out_free_subsys;
1741 retval = nfsd4_create_laundry_wq();
1744 retval = register_filesystem(&nfsd_fs_type);
1747 retval = genl_register_family(&nfsd_nl_family);
1753 nfsd4_destroy_laundry_wq();
1755 unregister_cld_notifier();
1757 unregister_pernet_subsys(&nfsd_net_ops);
1759 remove_proc_entry("fs/nfs/exports", NULL);
1760 remove_proc_entry("fs/nfs", NULL);
1762 nfsd_lockd_shutdown();
1763 nfsd_drc_slab_free();
1765 nfsd_stat_shutdown();
1773 static void __exit exit_nfsd(void)
1775 genl_unregister_family(&nfsd_nl_family);
1776 unregister_filesystem(&nfsd_fs_type);
1777 nfsd4_destroy_laundry_wq();
1778 unregister_cld_notifier();
1779 unregister_pernet_subsys(&nfsd_net_ops);
1780 nfsd_drc_slab_free();
1781 remove_proc_entry("fs/nfs/exports", NULL);
1782 remove_proc_entry("fs/nfs", NULL);
1783 nfsd_stat_shutdown();
1784 nfsd_lockd_shutdown();
1789 MODULE_AUTHOR("Olaf Kirch <okir@monad.swb.de>");
1790 MODULE_DESCRIPTION("In-kernel NFS server");
1791 MODULE_LICENSE("GPL");
1792 module_init(init_nfsd)
1793 module_exit(exit_nfsd)