1 // SPDX-License-Identifier: GPL-2.0-only
3 * Syscall interface to knfsd.
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/gss_krb5_enctypes.h>
18 #include <linux/sunrpc/rpc_pipe_fs.h>
19 #include <linux/module.h>
20 #include <linux/fsnotify.h>
28 #include "filecache.h"
31 * We have a single directory with several nodes in it.
44 NFSD_Reply_Cache_Stats,
50 NFSD_SupportedEnctypes,
52 * The below MUST come last. Otherwise we leave a hole in nfsd_files[]
53 * with !CONFIG_NFSD_V4 and simple_fill_super() goes oops
65 * write() for these nodes.
67 static ssize_t write_filehandle(struct file *file, char *buf, size_t size);
68 static ssize_t write_unlock_ip(struct file *file, char *buf, size_t size);
69 static ssize_t write_unlock_fs(struct file *file, char *buf, size_t size);
70 static ssize_t write_threads(struct file *file, char *buf, size_t size);
71 static ssize_t write_pool_threads(struct file *file, char *buf, size_t size);
72 static ssize_t write_versions(struct file *file, char *buf, size_t size);
73 static ssize_t write_ports(struct file *file, char *buf, size_t size);
74 static ssize_t write_maxblksize(struct file *file, char *buf, size_t size);
75 static ssize_t write_maxconn(struct file *file, char *buf, size_t size);
77 static ssize_t write_leasetime(struct file *file, char *buf, size_t size);
78 static ssize_t write_gracetime(struct file *file, char *buf, size_t size);
79 static ssize_t write_recoverydir(struct file *file, char *buf, size_t size);
80 static ssize_t write_v4_end_grace(struct file *file, char *buf, size_t size);
83 static ssize_t (*const write_op[])(struct file *, char *, size_t) = {
84 [NFSD_Fh] = write_filehandle,
85 [NFSD_FO_UnlockIP] = write_unlock_ip,
86 [NFSD_FO_UnlockFS] = write_unlock_fs,
87 [NFSD_Threads] = write_threads,
88 [NFSD_Pool_Threads] = write_pool_threads,
89 [NFSD_Versions] = write_versions,
90 [NFSD_Ports] = write_ports,
91 [NFSD_MaxBlkSize] = write_maxblksize,
92 [NFSD_MaxConnections] = write_maxconn,
94 [NFSD_Leasetime] = write_leasetime,
95 [NFSD_Gracetime] = write_gracetime,
96 [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);
116 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_proc_open(struct inode *inode, struct file *file)
160 return exports_net_open(current->nsproxy->net_ns, file);
163 static const struct proc_ops exports_proc_ops = {
164 .proc_open = exports_proc_open,
165 .proc_read = seq_read,
166 .proc_lseek = seq_lseek,
167 .proc_release = seq_release,
170 static int exports_nfsd_open(struct inode *inode, struct file *file)
172 return exports_net_open(inode->i_sb->s_fs_info, file);
175 static const struct file_operations exports_nfsd_operations = {
176 .open = exports_nfsd_open,
179 .release = seq_release,
182 static int export_features_show(struct seq_file *m, void *v)
184 seq_printf(m, "0x%x 0x%x\n", NFSEXP_ALLFLAGS, NFSEXP_SECINFO_FLAGS);
188 DEFINE_SHOW_ATTRIBUTE(export_features);
190 #if defined(CONFIG_SUNRPC_GSS) || defined(CONFIG_SUNRPC_GSS_MODULE)
191 static int supported_enctypes_show(struct seq_file *m, void *v)
193 seq_printf(m, KRB5_SUPPORTED_ENCTYPES);
197 DEFINE_SHOW_ATTRIBUTE(supported_enctypes);
198 #endif /* CONFIG_SUNRPC_GSS or CONFIG_SUNRPC_GSS_MODULE */
200 static const struct file_operations pool_stats_operations = {
201 .open = nfsd_pool_stats_open,
204 .release = nfsd_pool_stats_release,
207 DEFINE_SHOW_ATTRIBUTE(nfsd_reply_cache_stats);
209 DEFINE_SHOW_ATTRIBUTE(nfsd_file_cache_stats);
211 /*----------------------------------------------------------------------------*/
213 * payload - write methods
216 static inline struct net *netns(struct file *file)
218 return file_inode(file)->i_sb->s_fs_info;
222 * write_unlock_ip - Release all locks used by a client
227 * buf: '\n'-terminated C string containing a
228 * presentation format IP address
229 * size: length of C string in @buf
231 * On success: returns zero if all specified locks were released;
232 * returns one if one or more locks were not released
233 * On error: return code is negative errno value
235 static ssize_t write_unlock_ip(struct file *file, char *buf, size_t size)
237 struct sockaddr_storage address;
238 struct sockaddr *sap = (struct sockaddr *)&address;
239 size_t salen = sizeof(address);
241 struct net *net = netns(file);
247 if (buf[size-1] != '\n')
251 if (qword_get(&buf, fo_path, size) < 0)
254 if (rpc_pton(net, fo_path, size, sap, salen) == 0)
257 return nlmsvc_unlock_all_by_ip(sap);
261 * write_unlock_fs - Release all locks on a local file system
266 * buf: '\n'-terminated C string containing the
267 * absolute pathname of a local file system
268 * size: length of C string in @buf
270 * On success: returns zero if all specified locks were released;
271 * returns one if one or more locks were not released
272 * On error: return code is negative errno value
274 static ssize_t write_unlock_fs(struct file *file, char *buf, size_t size)
284 if (buf[size-1] != '\n')
288 if (qword_get(&buf, fo_path, size) < 0)
291 error = kern_path(fo_path, 0, &path);
296 * XXX: Needs better sanity checking. Otherwise we could end up
297 * releasing locks on the wrong file system.
300 * 1. Does the path refer to a directory?
301 * 2. Is that directory a mount point, or
302 * 3. Is that directory the root of an exported file system?
304 error = nlmsvc_unlock_all_by_sb(path.dentry->d_sb);
311 * write_filehandle - Get a variable-length NFS file handle by path
313 * On input, the buffer contains a '\n'-terminated C string comprised of
314 * three alphanumeric words separated by whitespace. The string may
315 * contain escape sequences.
319 * domain: client domain name
320 * path: export pathname
321 * maxsize: numeric maximum size of
323 * size: length of C string in @buf
325 * On success: passed-in buffer filled with '\n'-terminated C
326 * string containing a ASCII hex text version
327 * of the NFS file handle;
328 * return code is the size in bytes of the string
329 * On error: return code is negative errno value
331 static ssize_t write_filehandle(struct file *file, char *buf, size_t size)
337 struct auth_domain *dom;
343 if (buf[size-1] != '\n')
348 len = qword_get(&mesg, dname, size);
353 len = qword_get(&mesg, path, size);
357 len = get_int(&mesg, &maxsize);
361 if (maxsize < NFS_FHSIZE)
363 maxsize = min(maxsize, NFS3_FHSIZE);
365 if (qword_get(&mesg, mesg, size)>0)
368 /* we have all the words, they are in buf.. */
369 dom = unix_domain_find(dname);
373 len = exp_rootfh(netns(file), dom, path, &fh, maxsize);
374 auth_domain_put(dom);
379 len = SIMPLE_TRANSACTION_LIMIT;
380 qword_addhex(&mesg, &len, fh.fh_raw, fh.fh_size);
386 * write_threads - Start NFSD, or report the current number of running threads
392 * On success: passed-in buffer filled with '\n'-terminated C
393 * string numeric value representing the number of
394 * running NFSD threads;
395 * return code is the size in bytes of the string
396 * On error: return code is zero
401 * buf: C string containing an unsigned
402 * integer value representing the
403 * number of NFSD threads to start
404 * size: non-zero length of C string in @buf
406 * On success: NFS service is started;
407 * passed-in buffer filled with '\n'-terminated C
408 * string numeric value representing the number of
409 * running NFSD threads;
410 * return code is the size in bytes of the string
411 * On error: return code is zero or a negative errno value
413 static ssize_t write_threads(struct file *file, char *buf, size_t size)
417 struct net *net = netns(file);
421 rv = get_int(&mesg, &newthreads);
426 rv = nfsd_svc(newthreads, net, file->f_cred);
430 rv = nfsd_nrthreads(net);
432 return scnprintf(buf, SIMPLE_TRANSACTION_LIMIT, "%d\n", rv);
436 * write_pool_threads - Set or report the current number of threads per pool
445 * buf: C string containing whitespace-
446 * separated unsigned integer values
447 * representing the number of NFSD
448 * threads to start in each pool
449 * size: non-zero length of C string in @buf
451 * On success: passed-in buffer filled with '\n'-terminated C
452 * string containing integer values representing the
453 * number of NFSD threads in each pool;
454 * return code is the size in bytes of the string
455 * On error: return code is zero or a negative errno value
457 static ssize_t write_pool_threads(struct file *file, char *buf, size_t size)
459 /* if size > 0, look for an array of number of threads per node
460 * and apply them then write out number of threads per node as reply
468 struct net *net = netns(file);
470 mutex_lock(&nfsd_mutex);
471 npools = nfsd_nrpools(net);
474 * NFS is shut down. The admin can start it by
475 * writing to the threads file but NOT the pool_threads
476 * file, sorry. Report zero threads.
478 mutex_unlock(&nfsd_mutex);
483 nthreads = kcalloc(npools, sizeof(int), GFP_KERNEL);
485 if (nthreads == NULL)
489 for (i = 0; i < npools; i++) {
490 rv = get_int(&mesg, &nthreads[i]);
492 break; /* fewer numbers than pools */
494 goto out_free; /* syntax error */
499 rv = nfsd_set_nrthreads(i, nthreads, net);
504 rv = nfsd_get_nrthreads(npools, nthreads, net);
509 size = SIMPLE_TRANSACTION_LIMIT;
510 for (i = 0; i < npools && size > 0; i++) {
511 snprintf(mesg, size, "%d%c", nthreads[i], (i == npools-1 ? '\n' : ' '));
519 mutex_unlock(&nfsd_mutex);
524 nfsd_print_version_support(struct nfsd_net *nn, char *buf, int remaining,
525 const char *sep, unsigned vers, int minor)
527 const char *format = minor < 0 ? "%s%c%u" : "%s%c%u.%u";
528 bool supported = !!nfsd_vers(nn, vers, NFSD_TEST);
530 if (vers == 4 && minor >= 0 &&
531 !nfsd_minorversion(nn, minor, NFSD_TEST))
533 if (minor == 0 && supported)
535 * special case for backward compatability.
536 * +4.0 is never reported, it is implied by
537 * +4, unless -4.0 is present.
540 return snprintf(buf, remaining, format, sep,
541 supported ? '+' : '-', vers, minor);
544 static ssize_t __write_versions(struct file *file, char *buf, size_t size)
547 char *vers, *minorp, sign;
548 int len, num, remaining;
551 struct nfsd_net *nn = net_generic(netns(file), nfsd_net_id);
555 /* Cannot change versions without updating
556 * nn->nfsd_serv->sv_xdrsize, and reallocing
557 * rq_argp and rq_resp
560 if (buf[size-1] != '\n')
565 len = qword_get(&mesg, vers, size);
566 if (len <= 0) return -EINVAL;
571 if (sign == '+' || sign == '-')
572 num = simple_strtol((vers+1), &minorp, 0);
574 num = simple_strtol(vers, &minorp, 0);
575 if (*minorp == '.') {
578 if (kstrtouint(minorp+1, 0, &minor) < 0)
582 cmd = sign == '-' ? NFSD_CLEAR : NFSD_SET;
584 #ifdef CONFIG_NFSD_V2
588 nfsd_vers(nn, num, cmd);
591 if (*minorp == '.') {
592 if (nfsd_minorversion(nn, minor, cmd) < 0)
594 } else if ((cmd == NFSD_SET) != nfsd_vers(nn, num, NFSD_TEST)) {
596 * Either we have +4 and no minors are enabled,
597 * or we have -4 and at least one minor is enabled.
598 * In either case, propagate 'cmd' to all minors.
601 while (nfsd_minorversion(nn, minor, cmd) >= 0)
606 /* Ignore requests to disable non-existent versions */
611 } while ((len = qword_get(&mesg, vers, size)) > 0);
612 /* If all get turned off, turn them back on, as
613 * having no versions is BAD
615 nfsd_reset_versions(nn);
618 /* Now write current state into reply buffer */
620 remaining = SIMPLE_TRANSACTION_LIMIT;
621 for (num=2 ; num <= 4 ; num++) {
623 if (!nfsd_vers(nn, num, NFSD_AVAIL))
628 len = nfsd_print_version_support(nn, buf, remaining,
630 if (len >= remaining)
638 } while (num == 4 && minor <= NFSD_SUPPORTED_MINOR_VERSION);
641 len = snprintf(buf, remaining, "\n");
642 if (len >= remaining)
648 * write_versions - Set or report the available NFS protocol versions
654 * On success: passed-in buffer filled with '\n'-terminated C
655 * string containing positive or negative integer
656 * values representing the current status of each
658 * return code is the size in bytes of the string
659 * On error: return code is zero or a negative errno value
664 * buf: C string containing whitespace-
665 * separated positive or negative
666 * integer values representing NFS
667 * protocol versions to enable ("+n")
669 * size: non-zero length of C string in @buf
671 * On success: status of zero or more protocol versions has
672 * been updated; passed-in buffer filled with
673 * '\n'-terminated C string containing positive
674 * or negative integer values representing the
675 * current status of each protocol version;
676 * return code is the size in bytes of the string
677 * On error: return code is zero or a negative errno value
679 static ssize_t write_versions(struct file *file, char *buf, size_t size)
683 mutex_lock(&nfsd_mutex);
684 rv = __write_versions(file, buf, size);
685 mutex_unlock(&nfsd_mutex);
690 * Zero-length write. Return a list of NFSD's current listener
693 static ssize_t __write_ports_names(char *buf, struct net *net)
695 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
697 if (nn->nfsd_serv == NULL)
699 return svc_xprt_names(nn->nfsd_serv, buf, SIMPLE_TRANSACTION_LIMIT);
703 * A single 'fd' number was written, in which case it must be for
704 * a socket of a supported family/protocol, and we use it as an
707 static ssize_t __write_ports_addfd(char *buf, struct net *net, const struct cred *cred)
711 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
713 err = get_int(&mesg, &fd);
714 if (err != 0 || fd < 0)
717 if (svc_alien_sock(net, fd)) {
718 printk(KERN_ERR "%s: socket net is different to NFSd's one\n", __func__);
722 err = nfsd_create_serv(net);
726 err = svc_addsock(nn->nfsd_serv, fd, buf, SIMPLE_TRANSACTION_LIMIT, cred);
729 !nn->nfsd_serv->sv_nrthreads && !xchg(&nn->keep_active, 1))
730 svc_get(nn->nfsd_serv);
737 * A transport listener is added by writing it's transport name and
740 static ssize_t __write_ports_addxprt(char *buf, struct net *net, const struct cred *cred)
743 struct svc_xprt *xprt;
745 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
747 if (sscanf(buf, "%15s %5u", transport, &port) != 2)
750 if (port < 1 || port > USHRT_MAX)
753 err = nfsd_create_serv(net);
757 err = svc_xprt_create(nn->nfsd_serv, transport, net,
758 PF_INET, port, SVC_SOCK_ANONYMOUS, cred);
762 err = svc_xprt_create(nn->nfsd_serv, transport, net,
763 PF_INET6, port, SVC_SOCK_ANONYMOUS, cred);
764 if (err < 0 && err != -EAFNOSUPPORT)
767 if (!nn->nfsd_serv->sv_nrthreads && !xchg(&nn->keep_active, 1))
768 svc_get(nn->nfsd_serv);
773 xprt = svc_find_xprt(nn->nfsd_serv, transport, net, PF_INET, port);
775 svc_xprt_close(xprt);
783 static ssize_t __write_ports(struct file *file, char *buf, size_t size,
787 return __write_ports_names(buf, net);
790 return __write_ports_addfd(buf, net, file->f_cred);
793 return __write_ports_addxprt(buf, net, file->f_cred);
799 * write_ports - Pass a socket file descriptor or transport name to listen on
805 * On success: passed-in buffer filled with a '\n'-terminated C
806 * string containing a whitespace-separated list of
807 * named NFSD listeners;
808 * return code is the size in bytes of the string
809 * On error: return code is zero or a negative errno value
814 * buf: C string containing an unsigned
815 * integer value representing a bound
816 * but unconnected socket that is to be
817 * used as an NFSD listener; listen(3)
818 * must be called for a SOCK_STREAM
819 * socket, otherwise it is ignored
820 * size: non-zero length of C string in @buf
822 * On success: NFS service is started;
823 * passed-in buffer filled with a '\n'-terminated C
824 * string containing a unique alphanumeric name of
826 * return code is the size in bytes of the string
827 * On error: return code is a negative errno value
832 * buf: C string containing a transport
833 * name and an unsigned integer value
834 * representing the port to listen on,
835 * separated by whitespace
836 * size: non-zero length of C string in @buf
838 * On success: returns zero; NFS service is started
839 * On error: return code is a negative errno value
841 static ssize_t write_ports(struct file *file, char *buf, size_t size)
845 mutex_lock(&nfsd_mutex);
846 rv = __write_ports(file, buf, size, netns(file));
847 mutex_unlock(&nfsd_mutex);
852 int nfsd_max_blksize;
855 * write_maxblksize - Set or report the current NFS blksize
864 * buf: C string containing an unsigned
865 * integer value representing the new
867 * size: non-zero length of C string in @buf
869 * On success: passed-in buffer filled with '\n'-terminated C string
870 * containing numeric value of the current NFS blksize
872 * return code is the size in bytes of the string
873 * On error: return code is zero or a negative errno value
875 static ssize_t write_maxblksize(struct file *file, char *buf, size_t size)
878 struct nfsd_net *nn = net_generic(netns(file), nfsd_net_id);
882 int rv = get_int(&mesg, &bsize);
885 /* force bsize into allowed range and
886 * required alignment.
888 bsize = max_t(int, bsize, 1024);
889 bsize = min_t(int, bsize, NFSSVC_MAXBLKSIZE);
891 mutex_lock(&nfsd_mutex);
893 mutex_unlock(&nfsd_mutex);
896 nfsd_max_blksize = bsize;
897 mutex_unlock(&nfsd_mutex);
900 return scnprintf(buf, SIMPLE_TRANSACTION_LIMIT, "%d\n",
905 * write_maxconn - Set or report the current max number of connections
913 * buf: C string containing an unsigned
914 * integer value representing the new
915 * number of max connections
916 * size: non-zero length of C string in @buf
918 * On success: passed-in buffer filled with '\n'-terminated C string
919 * containing numeric value of max_connections setting
920 * for this net namespace;
921 * return code is the size in bytes of the string
922 * On error: return code is zero or a negative errno value
924 static ssize_t write_maxconn(struct file *file, char *buf, size_t size)
927 struct nfsd_net *nn = net_generic(netns(file), nfsd_net_id);
928 unsigned int maxconn = nn->max_connections;
931 int rv = get_uint(&mesg, &maxconn);
935 nn->max_connections = maxconn;
938 return scnprintf(buf, SIMPLE_TRANSACTION_LIMIT, "%u\n", maxconn);
941 #ifdef CONFIG_NFSD_V4
942 static ssize_t __nfsd4_write_time(struct file *file, char *buf, size_t size,
943 time64_t *time, struct nfsd_net *nn)
951 rv = get_int(&mesg, &i);
955 * Some sanity checking. We don't have a reason for
956 * these particular numbers, but problems with the
958 * - Too short: the briefest network outage may
959 * cause clients to lose all their locks. Also,
960 * the frequent polling may be wasteful.
961 * - Too long: do you really want reboot recovery
962 * to take more than an hour? Or to make other
963 * clients wait an hour before being able to
964 * revoke a dead client's locks?
966 if (i < 10 || i > 3600)
971 return scnprintf(buf, SIMPLE_TRANSACTION_LIMIT, "%lld\n", *time);
974 static ssize_t nfsd4_write_time(struct file *file, char *buf, size_t size,
975 time64_t *time, struct nfsd_net *nn)
979 mutex_lock(&nfsd_mutex);
980 rv = __nfsd4_write_time(file, buf, size, time, nn);
981 mutex_unlock(&nfsd_mutex);
986 * write_leasetime - Set or report the current NFSv4 lease time
995 * buf: C string containing an unsigned
996 * integer value representing the new
997 * NFSv4 lease expiry time
998 * size: non-zero length of C string in @buf
1000 * On success: passed-in buffer filled with '\n'-terminated C
1001 * string containing unsigned integer value of the
1002 * current lease expiry time;
1003 * return code is the size in bytes of the string
1004 * On error: return code is zero or a negative errno value
1006 static ssize_t write_leasetime(struct file *file, char *buf, size_t size)
1008 struct nfsd_net *nn = net_generic(netns(file), nfsd_net_id);
1009 return nfsd4_write_time(file, buf, size, &nn->nfsd4_lease, nn);
1013 * write_gracetime - Set or report current NFSv4 grace period time
1015 * As above, but sets the time of the NFSv4 grace period.
1017 * Note this should never be set to less than the *previous*
1018 * lease-period time, but we don't try to enforce this. (In the common
1019 * case (a new boot), we don't know what the previous lease time was
1022 static ssize_t write_gracetime(struct file *file, char *buf, size_t size)
1024 struct nfsd_net *nn = net_generic(netns(file), nfsd_net_id);
1025 return nfsd4_write_time(file, buf, size, &nn->nfsd4_grace, nn);
1028 static ssize_t __write_recoverydir(struct file *file, char *buf, size_t size,
1029 struct nfsd_net *nn)
1038 if (size > PATH_MAX || buf[size-1] != '\n')
1043 len = qword_get(&mesg, recdir, size);
1047 status = nfs4_reset_recoverydir(recdir);
1052 return scnprintf(buf, SIMPLE_TRANSACTION_LIMIT, "%s\n",
1053 nfs4_recoverydir());
1057 * write_recoverydir - Set or report the pathname of the recovery directory
1066 * buf: C string containing the pathname
1067 * of the directory on a local file
1068 * system containing permanent NFSv4
1070 * size: non-zero length of C string in @buf
1072 * On success: passed-in buffer filled with '\n'-terminated C string
1073 * containing the current recovery pathname setting;
1074 * return code is the size in bytes of the string
1075 * On error: return code is zero or a negative errno value
1077 static ssize_t write_recoverydir(struct file *file, char *buf, size_t size)
1080 struct nfsd_net *nn = net_generic(netns(file), nfsd_net_id);
1082 mutex_lock(&nfsd_mutex);
1083 rv = __write_recoverydir(file, buf, size, nn);
1084 mutex_unlock(&nfsd_mutex);
1089 * write_v4_end_grace - release grace period for nfsd's v4.x lock manager
1098 * size: non-zero length of C string in @buf
1100 * passed-in buffer filled with "Y" or "N" with a newline
1101 * and NULL-terminated C string. This indicates whether
1102 * the grace period has ended in the current net
1103 * namespace. Return code is the size in bytes of the
1104 * string. Writing a string that starts with 'Y', 'y', or
1105 * '1' to the file will end the grace period for nfsd's v4
1108 static ssize_t write_v4_end_grace(struct file *file, char *buf, size_t size)
1110 struct nfsd_net *nn = net_generic(netns(file), nfsd_net_id);
1119 nfsd4_end_grace(nn);
1126 return scnprintf(buf, SIMPLE_TRANSACTION_LIMIT, "%c\n",
1127 nn->grace_ended ? 'Y' : 'N');
1132 /*----------------------------------------------------------------------------*/
1134 * populating the filesystem.
1137 /* Basically copying rpc_get_inode. */
1138 static struct inode *nfsd_get_inode(struct super_block *sb, umode_t mode)
1140 struct inode *inode = new_inode(sb);
1143 /* Following advice from simple_fill_super documentation: */
1144 inode->i_ino = iunique(sb, NFSD_MaxReserved);
1145 inode->i_mode = mode;
1146 inode->i_atime = inode->i_mtime = inode->i_ctime = current_time(inode);
1147 switch (mode & S_IFMT) {
1149 inode->i_fop = &simple_dir_operations;
1150 inode->i_op = &simple_dir_inode_operations;
1159 static int __nfsd_mkdir(struct inode *dir, struct dentry *dentry, umode_t mode, struct nfsdfs_client *ncl)
1161 struct inode *inode;
1163 inode = nfsd_get_inode(dir->i_sb, mode);
1167 inode->i_private = ncl;
1168 kref_get(&ncl->cl_ref);
1170 d_add(dentry, inode);
1172 fsnotify_mkdir(dir, dentry);
1176 static struct dentry *nfsd_mkdir(struct dentry *parent, struct nfsdfs_client *ncl, char *name)
1178 struct inode *dir = parent->d_inode;
1179 struct dentry *dentry;
1183 dentry = d_alloc_name(parent, name);
1186 ret = __nfsd_mkdir(d_inode(parent), dentry, S_IFDIR | 0600, ncl);
1194 dentry = ERR_PTR(ret);
1198 static void clear_ncl(struct inode *inode)
1200 struct nfsdfs_client *ncl = inode->i_private;
1202 inode->i_private = NULL;
1203 kref_put(&ncl->cl_ref, ncl->cl_release);
1206 static struct nfsdfs_client *__get_nfsdfs_client(struct inode *inode)
1208 struct nfsdfs_client *nc = inode->i_private;
1211 kref_get(&nc->cl_ref);
1215 struct nfsdfs_client *get_nfsdfs_client(struct inode *inode)
1217 struct nfsdfs_client *nc;
1219 inode_lock_shared(inode);
1220 nc = __get_nfsdfs_client(inode);
1221 inode_unlock_shared(inode);
1224 /* from __rpc_unlink */
1225 static void nfsdfs_remove_file(struct inode *dir, struct dentry *dentry)
1229 clear_ncl(d_inode(dentry));
1231 ret = simple_unlink(dir, dentry);
1233 fsnotify_unlink(dir, dentry);
1238 static void nfsdfs_remove_files(struct dentry *root)
1240 struct dentry *dentry, *tmp;
1242 list_for_each_entry_safe(dentry, tmp, &root->d_subdirs, d_child) {
1243 if (!simple_positive(dentry)) {
1244 WARN_ON_ONCE(1); /* I think this can't happen? */
1247 nfsdfs_remove_file(d_inode(root), dentry);
1251 /* XXX: cut'n'paste from simple_fill_super; figure out if we could share
1253 static int nfsdfs_create_files(struct dentry *root,
1254 const struct tree_descr *files,
1255 struct dentry **fdentries)
1257 struct inode *dir = d_inode(root);
1258 struct inode *inode;
1259 struct dentry *dentry;
1263 for (i = 0; files->name && files->name[0]; i++, files++) {
1264 dentry = d_alloc_name(root, files->name);
1267 inode = nfsd_get_inode(d_inode(root)->i_sb,
1268 S_IFREG | files->mode);
1273 inode->i_fop = files->ops;
1274 inode->i_private = __get_nfsdfs_client(dir);
1275 d_add(dentry, inode);
1276 fsnotify_create(dir, dentry);
1278 fdentries[i] = dentry;
1283 nfsdfs_remove_files(root);
1288 /* on success, returns positive number unique to that client. */
1289 struct dentry *nfsd_client_mkdir(struct nfsd_net *nn,
1290 struct nfsdfs_client *ncl, u32 id,
1291 const struct tree_descr *files,
1292 struct dentry **fdentries)
1294 struct dentry *dentry;
1298 sprintf(name, "%u", id);
1300 dentry = nfsd_mkdir(nn->nfsd_client_dir, ncl, name);
1301 if (IS_ERR(dentry)) /* XXX: tossing errors? */
1303 ret = nfsdfs_create_files(dentry, files, fdentries);
1305 nfsd_client_rmdir(dentry);
1311 /* Taken from __rpc_rmdir: */
1312 void nfsd_client_rmdir(struct dentry *dentry)
1314 struct inode *dir = d_inode(dentry->d_parent);
1315 struct inode *inode = d_inode(dentry);
1319 nfsdfs_remove_files(dentry);
1322 ret = simple_rmdir(dir, dentry);
1325 fsnotify_rmdir(dir, dentry);
1330 static int nfsd_fill_super(struct super_block *sb, struct fs_context *fc)
1332 struct nfsd_net *nn = net_generic(current->nsproxy->net_ns,
1334 struct dentry *dentry;
1337 static const struct tree_descr nfsd_files[] = {
1338 [NFSD_List] = {"exports", &exports_nfsd_operations, S_IRUGO},
1339 /* Per-export io stats use same ops as exports file */
1340 [NFSD_Export_Stats] = {"export_stats", &exports_nfsd_operations, S_IRUGO},
1341 [NFSD_Export_features] = {"export_features",
1342 &export_features_fops, S_IRUGO},
1343 [NFSD_FO_UnlockIP] = {"unlock_ip",
1344 &transaction_ops, S_IWUSR|S_IRUSR},
1345 [NFSD_FO_UnlockFS] = {"unlock_filesystem",
1346 &transaction_ops, S_IWUSR|S_IRUSR},
1347 [NFSD_Fh] = {"filehandle", &transaction_ops, S_IWUSR|S_IRUSR},
1348 [NFSD_Threads] = {"threads", &transaction_ops, S_IWUSR|S_IRUSR},
1349 [NFSD_Pool_Threads] = {"pool_threads", &transaction_ops, S_IWUSR|S_IRUSR},
1350 [NFSD_Pool_Stats] = {"pool_stats", &pool_stats_operations, S_IRUGO},
1351 [NFSD_Reply_Cache_Stats] = {"reply_cache_stats",
1352 &nfsd_reply_cache_stats_fops, S_IRUGO},
1353 [NFSD_Versions] = {"versions", &transaction_ops, S_IWUSR|S_IRUSR},
1354 [NFSD_Ports] = {"portlist", &transaction_ops, S_IWUSR|S_IRUGO},
1355 [NFSD_MaxBlkSize] = {"max_block_size", &transaction_ops, S_IWUSR|S_IRUGO},
1356 [NFSD_MaxConnections] = {"max_connections", &transaction_ops, S_IWUSR|S_IRUGO},
1357 [NFSD_Filecache] = {"filecache", &nfsd_file_cache_stats_fops, S_IRUGO},
1358 #if defined(CONFIG_SUNRPC_GSS) || defined(CONFIG_SUNRPC_GSS_MODULE)
1359 [NFSD_SupportedEnctypes] = {"supported_krb5_enctypes",
1360 &supported_enctypes_fops, S_IRUGO},
1361 #endif /* CONFIG_SUNRPC_GSS or CONFIG_SUNRPC_GSS_MODULE */
1362 #ifdef CONFIG_NFSD_V4
1363 [NFSD_Leasetime] = {"nfsv4leasetime", &transaction_ops, S_IWUSR|S_IRUSR},
1364 [NFSD_Gracetime] = {"nfsv4gracetime", &transaction_ops, S_IWUSR|S_IRUSR},
1365 [NFSD_RecoveryDir] = {"nfsv4recoverydir", &transaction_ops, S_IWUSR|S_IRUSR},
1366 [NFSD_V4EndGrace] = {"v4_end_grace", &transaction_ops, S_IWUSR|S_IRUGO},
1371 ret = simple_fill_super(sb, 0x6e667364, nfsd_files);
1374 dentry = nfsd_mkdir(sb->s_root, NULL, "clients");
1376 return PTR_ERR(dentry);
1377 nn->nfsd_client_dir = dentry;
1381 static int nfsd_fs_get_tree(struct fs_context *fc)
1383 return get_tree_keyed(fc, nfsd_fill_super, get_net(fc->net_ns));
1386 static void nfsd_fs_free_fc(struct fs_context *fc)
1389 put_net(fc->s_fs_info);
1392 static const struct fs_context_operations nfsd_fs_context_ops = {
1393 .free = nfsd_fs_free_fc,
1394 .get_tree = nfsd_fs_get_tree,
1397 static int nfsd_init_fs_context(struct fs_context *fc)
1399 put_user_ns(fc->user_ns);
1400 fc->user_ns = get_user_ns(fc->net_ns->user_ns);
1401 fc->ops = &nfsd_fs_context_ops;
1405 static void nfsd_umount(struct super_block *sb)
1407 struct net *net = sb->s_fs_info;
1409 nfsd_shutdown_threads(net);
1411 kill_litter_super(sb);
1415 static struct file_system_type nfsd_fs_type = {
1416 .owner = THIS_MODULE,
1418 .init_fs_context = nfsd_init_fs_context,
1419 .kill_sb = nfsd_umount,
1421 MODULE_ALIAS_FS("nfsd");
1423 #ifdef CONFIG_PROC_FS
1424 static int create_proc_exports_entry(void)
1426 struct proc_dir_entry *entry;
1428 entry = proc_mkdir("fs/nfs", NULL);
1431 entry = proc_create("exports", 0, entry, &exports_proc_ops);
1433 remove_proc_entry("fs/nfs", NULL);
1438 #else /* CONFIG_PROC_FS */
1439 static int create_proc_exports_entry(void)
1445 unsigned int nfsd_net_id;
1447 static __net_init int nfsd_init_net(struct net *net)
1450 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
1452 retval = nfsd_export_init(net);
1454 goto out_export_error;
1455 retval = nfsd_idmap_init(net);
1457 goto out_idmap_error;
1458 nn->nfsd_versions = NULL;
1459 nn->nfsd4_minorversions = NULL;
1460 nfsd4_init_leases_net(nn);
1461 retval = nfsd_reply_cache_init(nn);
1463 goto out_cache_error;
1464 get_random_bytes(&nn->siphash_key, sizeof(nn->siphash_key));
1465 seqlock_init(&nn->writeverf_lock);
1470 nfsd_idmap_shutdown(net);
1472 nfsd_export_shutdown(net);
1477 static __net_exit void nfsd_exit_net(struct net *net)
1479 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
1481 nfsd_reply_cache_shutdown(nn);
1482 nfsd_idmap_shutdown(net);
1483 nfsd_export_shutdown(net);
1484 nfsd_netns_free_versions(net_generic(net, nfsd_net_id));
1487 static struct pernet_operations nfsd_net_ops = {
1488 .init = nfsd_init_net,
1489 .exit = nfsd_exit_net,
1491 .size = sizeof(struct nfsd_net),
1494 static int __init init_nfsd(void)
1498 retval = nfsd4_init_slabs();
1501 retval = nfsd4_init_pnfs();
1503 goto out_free_slabs;
1504 retval = nfsd_stat_init(); /* Statistics */
1507 retval = nfsd_drc_slab_create();
1510 nfsd_lockd_init(); /* lockd->nfsd callbacks */
1511 retval = create_proc_exports_entry();
1513 goto out_free_lockd;
1514 retval = register_pernet_subsys(&nfsd_net_ops);
1516 goto out_free_exports;
1517 retval = register_cld_notifier();
1519 goto out_free_subsys;
1520 retval = nfsd4_create_laundry_wq();
1523 retval = register_filesystem(&nfsd_fs_type);
1528 nfsd4_destroy_laundry_wq();
1530 unregister_cld_notifier();
1532 unregister_pernet_subsys(&nfsd_net_ops);
1534 remove_proc_entry("fs/nfs/exports", NULL);
1535 remove_proc_entry("fs/nfs", NULL);
1537 nfsd_lockd_shutdown();
1538 nfsd_drc_slab_free();
1540 nfsd_stat_shutdown();
1548 static void __exit exit_nfsd(void)
1550 unregister_filesystem(&nfsd_fs_type);
1551 nfsd4_destroy_laundry_wq();
1552 unregister_cld_notifier();
1553 unregister_pernet_subsys(&nfsd_net_ops);
1554 nfsd_drc_slab_free();
1555 remove_proc_entry("fs/nfs/exports", NULL);
1556 remove_proc_entry("fs/nfs", NULL);
1557 nfsd_stat_shutdown();
1558 nfsd_lockd_shutdown();
1564 MODULE_LICENSE("GPL");
1565 module_init(init_nfsd)
1566 module_exit(exit_nfsd)