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>
30 * We have a single directory with several nodes in it.
43 NFSD_Reply_Cache_Stats,
48 NFSD_SupportedEnctypes,
50 * The below MUST come last. Otherwise we leave a hole in nfsd_files[]
51 * with !CONFIG_NFSD_V4 and simple_fill_super() goes oops
63 * write() for these nodes.
65 static ssize_t write_filehandle(struct file *file, char *buf, size_t size);
66 static ssize_t write_unlock_ip(struct file *file, char *buf, size_t size);
67 static ssize_t write_unlock_fs(struct file *file, char *buf, size_t size);
68 static ssize_t write_threads(struct file *file, char *buf, size_t size);
69 static ssize_t write_pool_threads(struct file *file, char *buf, size_t size);
70 static ssize_t write_versions(struct file *file, char *buf, size_t size);
71 static ssize_t write_ports(struct file *file, char *buf, size_t size);
72 static ssize_t write_maxblksize(struct file *file, char *buf, size_t size);
73 static ssize_t write_maxconn(struct file *file, char *buf, size_t size);
75 static ssize_t write_leasetime(struct file *file, char *buf, size_t size);
76 static ssize_t write_gracetime(struct file *file, char *buf, size_t size);
77 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 [NFSD_RecoveryDir] = write_recoverydir,
95 [NFSD_V4EndGrace] = write_v4_end_grace,
99 static ssize_t nfsctl_transaction_write(struct file *file, const char __user *buf, size_t size, loff_t *pos)
101 ino_t ino = file_inode(file)->i_ino;
105 if (ino >= ARRAY_SIZE(write_op) || !write_op[ino])
108 data = simple_transaction_get(file, buf, size);
110 return PTR_ERR(data);
112 rv = write_op[ino](file, data, size);
114 simple_transaction_set(file, rv);
120 static ssize_t nfsctl_transaction_read(struct file *file, char __user *buf, size_t size, loff_t *pos)
122 if (! file->private_data) {
123 /* An attempt to read a transaction file without writing
124 * causes a 0-byte write so that the file can return
127 ssize_t rv = nfsctl_transaction_write(file, buf, 0, pos);
131 return simple_transaction_read(file, buf, size, pos);
134 static const struct file_operations transaction_ops = {
135 .write = nfsctl_transaction_write,
136 .read = nfsctl_transaction_read,
137 .release = simple_transaction_release,
138 .llseek = default_llseek,
141 static int exports_net_open(struct net *net, struct file *file)
144 struct seq_file *seq;
145 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
147 err = seq_open(file, &nfs_exports_op);
151 seq = file->private_data;
152 seq->private = nn->svc_export_cache;
156 static int exports_proc_open(struct inode *inode, struct file *file)
158 return exports_net_open(current->nsproxy->net_ns, file);
161 static const struct proc_ops exports_proc_ops = {
162 .proc_open = exports_proc_open,
163 .proc_read = seq_read,
164 .proc_lseek = seq_lseek,
165 .proc_release = seq_release,
168 static int exports_nfsd_open(struct inode *inode, struct file *file)
170 return exports_net_open(inode->i_sb->s_fs_info, file);
173 static const struct file_operations exports_nfsd_operations = {
174 .open = exports_nfsd_open,
177 .release = seq_release,
180 static int export_features_show(struct seq_file *m, void *v)
182 seq_printf(m, "0x%x 0x%x\n", NFSEXP_ALLFLAGS, NFSEXP_SECINFO_FLAGS);
186 static int export_features_open(struct inode *inode, struct file *file)
188 return single_open(file, export_features_show, NULL);
191 static const struct file_operations export_features_operations = {
192 .open = export_features_open,
195 .release = single_release,
198 #if defined(CONFIG_SUNRPC_GSS) || defined(CONFIG_SUNRPC_GSS_MODULE)
199 static int supported_enctypes_show(struct seq_file *m, void *v)
201 seq_printf(m, KRB5_SUPPORTED_ENCTYPES);
205 static int supported_enctypes_open(struct inode *inode, struct file *file)
207 return single_open(file, supported_enctypes_show, NULL);
210 static const struct file_operations supported_enctypes_ops = {
211 .open = supported_enctypes_open,
214 .release = single_release,
216 #endif /* CONFIG_SUNRPC_GSS or CONFIG_SUNRPC_GSS_MODULE */
218 static const struct file_operations pool_stats_operations = {
219 .open = nfsd_pool_stats_open,
222 .release = nfsd_pool_stats_release,
225 static const struct file_operations reply_cache_stats_operations = {
226 .open = nfsd_reply_cache_stats_open,
229 .release = single_release,
232 /*----------------------------------------------------------------------------*/
234 * payload - write methods
237 static inline struct net *netns(struct file *file)
239 return file_inode(file)->i_sb->s_fs_info;
243 * write_unlock_ip - Release all locks used by a client
248 * buf: '\n'-terminated C string containing a
249 * presentation format IP address
250 * size: length of C string in @buf
252 * On success: returns zero if all specified locks were released;
253 * returns one if one or more locks were not released
254 * On error: return code is negative errno value
256 static ssize_t write_unlock_ip(struct file *file, char *buf, size_t size)
258 struct sockaddr_storage address;
259 struct sockaddr *sap = (struct sockaddr *)&address;
260 size_t salen = sizeof(address);
262 struct net *net = netns(file);
268 if (buf[size-1] != '\n')
272 if (qword_get(&buf, fo_path, size) < 0)
275 if (rpc_pton(net, fo_path, size, sap, salen) == 0)
278 return nlmsvc_unlock_all_by_ip(sap);
282 * write_unlock_fs - Release all locks on a local file system
287 * buf: '\n'-terminated C string containing the
288 * absolute pathname of a local file system
289 * size: length of C string in @buf
291 * On success: returns zero if all specified locks were released;
292 * returns one if one or more locks were not released
293 * On error: return code is negative errno value
295 static ssize_t write_unlock_fs(struct file *file, char *buf, size_t size)
305 if (buf[size-1] != '\n')
309 if (qword_get(&buf, fo_path, size) < 0)
312 error = kern_path(fo_path, 0, &path);
317 * XXX: Needs better sanity checking. Otherwise we could end up
318 * releasing locks on the wrong file system.
321 * 1. Does the path refer to a directory?
322 * 2. Is that directory a mount point, or
323 * 3. Is that directory the root of an exported file system?
325 error = nlmsvc_unlock_all_by_sb(path.dentry->d_sb);
332 * write_filehandle - Get a variable-length NFS file handle by path
334 * On input, the buffer contains a '\n'-terminated C string comprised of
335 * three alphanumeric words separated by whitespace. The string may
336 * contain escape sequences.
340 * domain: client domain name
341 * path: export pathname
342 * maxsize: numeric maximum size of
344 * size: length of C string in @buf
346 * On success: passed-in buffer filled with '\n'-terminated C
347 * string containing a ASCII hex text version
348 * of the NFS file handle;
349 * return code is the size in bytes of the string
350 * On error: return code is negative errno value
352 static ssize_t write_filehandle(struct file *file, char *buf, size_t size)
358 struct auth_domain *dom;
364 if (buf[size-1] != '\n')
369 len = qword_get(&mesg, dname, size);
374 len = qword_get(&mesg, path, size);
378 len = get_int(&mesg, &maxsize);
382 if (maxsize < NFS_FHSIZE)
384 maxsize = min(maxsize, NFS3_FHSIZE);
386 if (qword_get(&mesg, mesg, size)>0)
389 /* we have all the words, they are in buf.. */
390 dom = unix_domain_find(dname);
394 len = exp_rootfh(netns(file), dom, path, &fh, maxsize);
395 auth_domain_put(dom);
400 len = SIMPLE_TRANSACTION_LIMIT;
401 qword_addhex(&mesg, &len, fh.fh_raw, fh.fh_size);
407 * write_threads - Start NFSD, or report the current number of running threads
413 * On success: passed-in buffer filled with '\n'-terminated C
414 * string numeric value representing the number of
415 * running NFSD threads;
416 * return code is the size in bytes of the string
417 * On error: return code is zero
422 * buf: C string containing an unsigned
423 * integer value representing the
424 * number of NFSD threads to start
425 * size: non-zero length of C string in @buf
427 * On success: NFS service is started;
428 * passed-in buffer filled with '\n'-terminated C
429 * string numeric value representing the number of
430 * running NFSD threads;
431 * return code is the size in bytes of the string
432 * On error: return code is zero or a negative errno value
434 static ssize_t write_threads(struct file *file, char *buf, size_t size)
438 struct net *net = netns(file);
442 rv = get_int(&mesg, &newthreads);
447 rv = nfsd_svc(newthreads, net, file->f_cred);
451 rv = nfsd_nrthreads(net);
453 return scnprintf(buf, SIMPLE_TRANSACTION_LIMIT, "%d\n", rv);
457 * write_pool_threads - Set or report the current number of threads per pool
466 * buf: C string containing whitespace-
467 * separated unsigned integer values
468 * representing the number of NFSD
469 * threads to start in each pool
470 * size: non-zero length of C string in @buf
472 * On success: passed-in buffer filled with '\n'-terminated C
473 * string containing integer values representing the
474 * number of NFSD threads in each pool;
475 * return code is the size in bytes of the string
476 * On error: return code is zero or a negative errno value
478 static ssize_t write_pool_threads(struct file *file, char *buf, size_t size)
480 /* if size > 0, look for an array of number of threads per node
481 * and apply them then write out number of threads per node as reply
489 struct net *net = netns(file);
491 mutex_lock(&nfsd_mutex);
492 npools = nfsd_nrpools(net);
495 * NFS is shut down. The admin can start it by
496 * writing to the threads file but NOT the pool_threads
497 * file, sorry. Report zero threads.
499 mutex_unlock(&nfsd_mutex);
504 nthreads = kcalloc(npools, sizeof(int), GFP_KERNEL);
506 if (nthreads == NULL)
510 for (i = 0; i < npools; i++) {
511 rv = get_int(&mesg, &nthreads[i]);
513 break; /* fewer numbers than pools */
515 goto out_free; /* syntax error */
520 rv = nfsd_set_nrthreads(i, nthreads, net);
525 rv = nfsd_get_nrthreads(npools, nthreads, net);
530 size = SIMPLE_TRANSACTION_LIMIT;
531 for (i = 0; i < npools && size > 0; i++) {
532 snprintf(mesg, size, "%d%c", nthreads[i], (i == npools-1 ? '\n' : ' '));
540 mutex_unlock(&nfsd_mutex);
545 nfsd_print_version_support(struct nfsd_net *nn, char *buf, int remaining,
546 const char *sep, unsigned vers, int minor)
548 const char *format = minor < 0 ? "%s%c%u" : "%s%c%u.%u";
549 bool supported = !!nfsd_vers(nn, vers, NFSD_TEST);
551 if (vers == 4 && minor >= 0 &&
552 !nfsd_minorversion(nn, minor, NFSD_TEST))
554 if (minor == 0 && supported)
556 * special case for backward compatability.
557 * +4.0 is never reported, it is implied by
558 * +4, unless -4.0 is present.
561 return snprintf(buf, remaining, format, sep,
562 supported ? '+' : '-', vers, minor);
565 static ssize_t __write_versions(struct file *file, char *buf, size_t size)
568 char *vers, *minorp, sign;
569 int len, num, remaining;
572 struct nfsd_net *nn = net_generic(netns(file), nfsd_net_id);
576 /* Cannot change versions without updating
577 * nn->nfsd_serv->sv_xdrsize, and reallocing
578 * rq_argp and rq_resp
581 if (buf[size-1] != '\n')
586 len = qword_get(&mesg, vers, size);
587 if (len <= 0) return -EINVAL;
592 if (sign == '+' || sign == '-')
593 num = simple_strtol((vers+1), &minorp, 0);
595 num = simple_strtol(vers, &minorp, 0);
596 if (*minorp == '.') {
599 if (kstrtouint(minorp+1, 0, &minor) < 0)
603 cmd = sign == '-' ? NFSD_CLEAR : NFSD_SET;
607 nfsd_vers(nn, num, cmd);
610 if (*minorp == '.') {
611 if (nfsd_minorversion(nn, minor, cmd) < 0)
613 } else if ((cmd == NFSD_SET) != nfsd_vers(nn, num, NFSD_TEST)) {
615 * Either we have +4 and no minors are enabled,
616 * or we have -4 and at least one minor is enabled.
617 * In either case, propagate 'cmd' to all minors.
620 while (nfsd_minorversion(nn, minor, cmd) >= 0)
628 } while ((len = qword_get(&mesg, vers, size)) > 0);
629 /* If all get turned off, turn them back on, as
630 * having no versions is BAD
632 nfsd_reset_versions(nn);
635 /* Now write current state into reply buffer */
638 remaining = SIMPLE_TRANSACTION_LIMIT;
639 for (num=2 ; num <= 4 ; num++) {
641 if (!nfsd_vers(nn, num, NFSD_AVAIL))
646 len = nfsd_print_version_support(nn, buf, remaining,
648 if (len >= remaining)
656 } while (num == 4 && minor <= NFSD_SUPPORTED_MINOR_VERSION);
659 len = snprintf(buf, remaining, "\n");
660 if (len >= remaining)
666 * write_versions - Set or report the available NFS protocol versions
672 * On success: passed-in buffer filled with '\n'-terminated C
673 * string containing positive or negative integer
674 * values representing the current status of each
676 * return code is the size in bytes of the string
677 * On error: return code is zero or a negative errno value
682 * buf: C string containing whitespace-
683 * separated positive or negative
684 * integer values representing NFS
685 * protocol versions to enable ("+n")
687 * size: non-zero length of C string in @buf
689 * On success: status of zero or more protocol versions has
690 * been updated; passed-in buffer filled with
691 * '\n'-terminated C string containing positive
692 * or negative integer values representing the
693 * current status of each protocol version;
694 * return code is the size in bytes of the string
695 * On error: return code is zero or a negative errno value
697 static ssize_t write_versions(struct file *file, char *buf, size_t size)
701 mutex_lock(&nfsd_mutex);
702 rv = __write_versions(file, buf, size);
703 mutex_unlock(&nfsd_mutex);
708 * Zero-length write. Return a list of NFSD's current listener
711 static ssize_t __write_ports_names(char *buf, struct net *net)
713 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
715 if (nn->nfsd_serv == NULL)
717 return svc_xprt_names(nn->nfsd_serv, buf, SIMPLE_TRANSACTION_LIMIT);
721 * A single 'fd' number was written, in which case it must be for
722 * a socket of a supported family/protocol, and we use it as an
725 static ssize_t __write_ports_addfd(char *buf, struct net *net, const struct cred *cred)
729 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
731 err = get_int(&mesg, &fd);
732 if (err != 0 || fd < 0)
735 if (svc_alien_sock(net, fd)) {
736 printk(KERN_ERR "%s: socket net is different to NFSd's one\n", __func__);
740 err = nfsd_create_serv(net);
744 err = svc_addsock(nn->nfsd_serv, fd, buf, SIMPLE_TRANSACTION_LIMIT, cred);
747 !nn->nfsd_serv->sv_nrthreads && !xchg(&nn->keep_active, 1))
748 svc_get(nn->nfsd_serv);
755 * A transport listener is added by writing it's transport name and
758 static ssize_t __write_ports_addxprt(char *buf, struct net *net, const struct cred *cred)
761 struct svc_xprt *xprt;
763 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
765 if (sscanf(buf, "%15s %5u", transport, &port) != 2)
768 if (port < 1 || port > USHRT_MAX)
771 err = nfsd_create_serv(net);
775 err = svc_xprt_create(nn->nfsd_serv, transport, net,
776 PF_INET, port, SVC_SOCK_ANONYMOUS, cred);
780 err = svc_xprt_create(nn->nfsd_serv, transport, net,
781 PF_INET6, port, SVC_SOCK_ANONYMOUS, cred);
782 if (err < 0 && err != -EAFNOSUPPORT)
785 if (!nn->nfsd_serv->sv_nrthreads && !xchg(&nn->keep_active, 1))
786 svc_get(nn->nfsd_serv);
791 xprt = svc_find_xprt(nn->nfsd_serv, transport, net, PF_INET, port);
793 svc_xprt_close(xprt);
801 static ssize_t __write_ports(struct file *file, char *buf, size_t size,
805 return __write_ports_names(buf, net);
808 return __write_ports_addfd(buf, net, file->f_cred);
811 return __write_ports_addxprt(buf, net, file->f_cred);
817 * write_ports - Pass a socket file descriptor or transport name to listen on
823 * On success: passed-in buffer filled with a '\n'-terminated C
824 * string containing a whitespace-separated list of
825 * named NFSD listeners;
826 * return code is the size in bytes of the string
827 * On error: return code is zero or a negative errno value
832 * buf: C string containing an unsigned
833 * integer value representing a bound
834 * but unconnected socket that is to be
835 * used as an NFSD listener; listen(3)
836 * must be called for a SOCK_STREAM
837 * socket, otherwise it is ignored
838 * size: non-zero length of C string in @buf
840 * On success: NFS service is started;
841 * passed-in buffer filled with a '\n'-terminated C
842 * string containing a unique alphanumeric name of
844 * return code is the size in bytes of the string
845 * On error: return code is a negative errno value
850 * buf: C string containing a transport
851 * name and an unsigned integer value
852 * representing the port to listen on,
853 * separated by whitespace
854 * size: non-zero length of C string in @buf
856 * On success: returns zero; NFS service is started
857 * On error: return code is a negative errno value
859 static ssize_t write_ports(struct file *file, char *buf, size_t size)
863 mutex_lock(&nfsd_mutex);
864 rv = __write_ports(file, buf, size, netns(file));
865 mutex_unlock(&nfsd_mutex);
870 int nfsd_max_blksize;
873 * write_maxblksize - Set or report the current NFS blksize
882 * buf: C string containing an unsigned
883 * integer value representing the new
885 * size: non-zero length of C string in @buf
887 * On success: passed-in buffer filled with '\n'-terminated C string
888 * containing numeric value of the current NFS blksize
890 * return code is the size in bytes of the string
891 * On error: return code is zero or a negative errno value
893 static ssize_t write_maxblksize(struct file *file, char *buf, size_t size)
896 struct nfsd_net *nn = net_generic(netns(file), nfsd_net_id);
900 int rv = get_int(&mesg, &bsize);
903 /* force bsize into allowed range and
904 * required alignment.
906 bsize = max_t(int, bsize, 1024);
907 bsize = min_t(int, bsize, NFSSVC_MAXBLKSIZE);
909 mutex_lock(&nfsd_mutex);
911 mutex_unlock(&nfsd_mutex);
914 nfsd_max_blksize = bsize;
915 mutex_unlock(&nfsd_mutex);
918 return scnprintf(buf, SIMPLE_TRANSACTION_LIMIT, "%d\n",
923 * write_maxconn - Set or report the current max number of connections
931 * buf: C string containing an unsigned
932 * integer value representing the new
933 * number of max connections
934 * size: non-zero length of C string in @buf
936 * On success: passed-in buffer filled with '\n'-terminated C string
937 * containing numeric value of max_connections setting
938 * for this net namespace;
939 * return code is the size in bytes of the string
940 * On error: return code is zero or a negative errno value
942 static ssize_t write_maxconn(struct file *file, char *buf, size_t size)
945 struct nfsd_net *nn = net_generic(netns(file), nfsd_net_id);
946 unsigned int maxconn = nn->max_connections;
949 int rv = get_uint(&mesg, &maxconn);
953 nn->max_connections = maxconn;
956 return scnprintf(buf, SIMPLE_TRANSACTION_LIMIT, "%u\n", maxconn);
959 #ifdef CONFIG_NFSD_V4
960 static ssize_t __nfsd4_write_time(struct file *file, char *buf, size_t size,
961 time64_t *time, struct nfsd_net *nn)
969 rv = get_int(&mesg, &i);
973 * Some sanity checking. We don't have a reason for
974 * these particular numbers, but problems with the
976 * - Too short: the briefest network outage may
977 * cause clients to lose all their locks. Also,
978 * the frequent polling may be wasteful.
979 * - Too long: do you really want reboot recovery
980 * to take more than an hour? Or to make other
981 * clients wait an hour before being able to
982 * revoke a dead client's locks?
984 if (i < 10 || i > 3600)
989 return scnprintf(buf, SIMPLE_TRANSACTION_LIMIT, "%lld\n", *time);
992 static ssize_t nfsd4_write_time(struct file *file, char *buf, size_t size,
993 time64_t *time, struct nfsd_net *nn)
997 mutex_lock(&nfsd_mutex);
998 rv = __nfsd4_write_time(file, buf, size, time, nn);
999 mutex_unlock(&nfsd_mutex);
1004 * write_leasetime - Set or report the current NFSv4 lease time
1013 * buf: C string containing an unsigned
1014 * integer value representing the new
1015 * NFSv4 lease expiry time
1016 * size: non-zero length of C string in @buf
1018 * On success: passed-in buffer filled with '\n'-terminated C
1019 * string containing unsigned integer value of the
1020 * current lease expiry time;
1021 * return code is the size in bytes of the string
1022 * On error: return code is zero or a negative errno value
1024 static ssize_t write_leasetime(struct file *file, char *buf, size_t size)
1026 struct nfsd_net *nn = net_generic(netns(file), nfsd_net_id);
1027 return nfsd4_write_time(file, buf, size, &nn->nfsd4_lease, nn);
1031 * write_gracetime - Set or report current NFSv4 grace period time
1033 * As above, but sets the time of the NFSv4 grace period.
1035 * Note this should never be set to less than the *previous*
1036 * lease-period time, but we don't try to enforce this. (In the common
1037 * case (a new boot), we don't know what the previous lease time was
1040 static ssize_t write_gracetime(struct file *file, char *buf, size_t size)
1042 struct nfsd_net *nn = net_generic(netns(file), nfsd_net_id);
1043 return nfsd4_write_time(file, buf, size, &nn->nfsd4_grace, nn);
1046 static ssize_t __write_recoverydir(struct file *file, char *buf, size_t size,
1047 struct nfsd_net *nn)
1056 if (size > PATH_MAX || buf[size-1] != '\n')
1061 len = qword_get(&mesg, recdir, size);
1065 status = nfs4_reset_recoverydir(recdir);
1070 return scnprintf(buf, SIMPLE_TRANSACTION_LIMIT, "%s\n",
1071 nfs4_recoverydir());
1075 * write_recoverydir - Set or report the pathname of the recovery directory
1084 * buf: C string containing the pathname
1085 * of the directory on a local file
1086 * system containing permanent NFSv4
1088 * size: non-zero length of C string in @buf
1090 * On success: passed-in buffer filled with '\n'-terminated C string
1091 * containing the current recovery pathname setting;
1092 * return code is the size in bytes of the string
1093 * On error: return code is zero or a negative errno value
1095 static ssize_t write_recoverydir(struct file *file, char *buf, size_t size)
1098 struct nfsd_net *nn = net_generic(netns(file), nfsd_net_id);
1100 mutex_lock(&nfsd_mutex);
1101 rv = __write_recoverydir(file, buf, size, nn);
1102 mutex_unlock(&nfsd_mutex);
1107 * write_v4_end_grace - release grace period for nfsd's v4.x lock manager
1116 * size: non-zero length of C string in @buf
1118 * passed-in buffer filled with "Y" or "N" with a newline
1119 * and NULL-terminated C string. This indicates whether
1120 * the grace period has ended in the current net
1121 * namespace. Return code is the size in bytes of the
1122 * string. Writing a string that starts with 'Y', 'y', or
1123 * '1' to the file will end the grace period for nfsd's v4
1126 static ssize_t write_v4_end_grace(struct file *file, char *buf, size_t size)
1128 struct nfsd_net *nn = net_generic(netns(file), nfsd_net_id);
1137 nfsd4_end_grace(nn);
1144 return scnprintf(buf, SIMPLE_TRANSACTION_LIMIT, "%c\n",
1145 nn->grace_ended ? 'Y' : 'N');
1150 /*----------------------------------------------------------------------------*/
1152 * populating the filesystem.
1155 /* Basically copying rpc_get_inode. */
1156 static struct inode *nfsd_get_inode(struct super_block *sb, umode_t mode)
1158 struct inode *inode = new_inode(sb);
1161 /* Following advice from simple_fill_super documentation: */
1162 inode->i_ino = iunique(sb, NFSD_MaxReserved);
1163 inode->i_mode = mode;
1164 inode->i_atime = inode->i_mtime = inode->i_ctime = current_time(inode);
1165 switch (mode & S_IFMT) {
1167 inode->i_fop = &simple_dir_operations;
1168 inode->i_op = &simple_dir_inode_operations;
1177 static int __nfsd_mkdir(struct inode *dir, struct dentry *dentry, umode_t mode, struct nfsdfs_client *ncl)
1179 struct inode *inode;
1181 inode = nfsd_get_inode(dir->i_sb, mode);
1185 inode->i_private = ncl;
1186 kref_get(&ncl->cl_ref);
1188 d_add(dentry, inode);
1190 fsnotify_mkdir(dir, dentry);
1194 static struct dentry *nfsd_mkdir(struct dentry *parent, struct nfsdfs_client *ncl, char *name)
1196 struct inode *dir = parent->d_inode;
1197 struct dentry *dentry;
1201 dentry = d_alloc_name(parent, name);
1204 ret = __nfsd_mkdir(d_inode(parent), dentry, S_IFDIR | 0600, ncl);
1212 dentry = ERR_PTR(ret);
1216 static void clear_ncl(struct inode *inode)
1218 struct nfsdfs_client *ncl = inode->i_private;
1220 inode->i_private = NULL;
1221 kref_put(&ncl->cl_ref, ncl->cl_release);
1224 static struct nfsdfs_client *__get_nfsdfs_client(struct inode *inode)
1226 struct nfsdfs_client *nc = inode->i_private;
1229 kref_get(&nc->cl_ref);
1233 struct nfsdfs_client *get_nfsdfs_client(struct inode *inode)
1235 struct nfsdfs_client *nc;
1237 inode_lock_shared(inode);
1238 nc = __get_nfsdfs_client(inode);
1239 inode_unlock_shared(inode);
1242 /* from __rpc_unlink */
1243 static void nfsdfs_remove_file(struct inode *dir, struct dentry *dentry)
1247 clear_ncl(d_inode(dentry));
1249 ret = simple_unlink(dir, dentry);
1251 fsnotify_unlink(dir, dentry);
1256 static void nfsdfs_remove_files(struct dentry *root)
1258 struct dentry *dentry, *tmp;
1260 list_for_each_entry_safe(dentry, tmp, &root->d_subdirs, d_child) {
1261 if (!simple_positive(dentry)) {
1262 WARN_ON_ONCE(1); /* I think this can't happen? */
1265 nfsdfs_remove_file(d_inode(root), dentry);
1269 /* XXX: cut'n'paste from simple_fill_super; figure out if we could share
1271 static int nfsdfs_create_files(struct dentry *root,
1272 const struct tree_descr *files,
1273 struct dentry **fdentries)
1275 struct inode *dir = d_inode(root);
1276 struct inode *inode;
1277 struct dentry *dentry;
1281 for (i = 0; files->name && files->name[0]; i++, files++) {
1282 dentry = d_alloc_name(root, files->name);
1285 inode = nfsd_get_inode(d_inode(root)->i_sb,
1286 S_IFREG | files->mode);
1291 inode->i_fop = files->ops;
1292 inode->i_private = __get_nfsdfs_client(dir);
1293 d_add(dentry, inode);
1294 fsnotify_create(dir, dentry);
1296 fdentries[i] = dentry;
1301 nfsdfs_remove_files(root);
1306 /* on success, returns positive number unique to that client. */
1307 struct dentry *nfsd_client_mkdir(struct nfsd_net *nn,
1308 struct nfsdfs_client *ncl, u32 id,
1309 const struct tree_descr *files,
1310 struct dentry **fdentries)
1312 struct dentry *dentry;
1316 sprintf(name, "%u", id);
1318 dentry = nfsd_mkdir(nn->nfsd_client_dir, ncl, name);
1319 if (IS_ERR(dentry)) /* XXX: tossing errors? */
1321 ret = nfsdfs_create_files(dentry, files, fdentries);
1323 nfsd_client_rmdir(dentry);
1329 /* Taken from __rpc_rmdir: */
1330 void nfsd_client_rmdir(struct dentry *dentry)
1332 struct inode *dir = d_inode(dentry->d_parent);
1333 struct inode *inode = d_inode(dentry);
1337 nfsdfs_remove_files(dentry);
1340 ret = simple_rmdir(dir, dentry);
1343 fsnotify_rmdir(dir, dentry);
1348 static int nfsd_fill_super(struct super_block *sb, struct fs_context *fc)
1350 struct nfsd_net *nn = net_generic(current->nsproxy->net_ns,
1352 struct dentry *dentry;
1355 static const struct tree_descr nfsd_files[] = {
1356 [NFSD_List] = {"exports", &exports_nfsd_operations, S_IRUGO},
1357 /* Per-export io stats use same ops as exports file */
1358 [NFSD_Export_Stats] = {"export_stats", &exports_nfsd_operations, S_IRUGO},
1359 [NFSD_Export_features] = {"export_features",
1360 &export_features_operations, S_IRUGO},
1361 [NFSD_FO_UnlockIP] = {"unlock_ip",
1362 &transaction_ops, S_IWUSR|S_IRUSR},
1363 [NFSD_FO_UnlockFS] = {"unlock_filesystem",
1364 &transaction_ops, S_IWUSR|S_IRUSR},
1365 [NFSD_Fh] = {"filehandle", &transaction_ops, S_IWUSR|S_IRUSR},
1366 [NFSD_Threads] = {"threads", &transaction_ops, S_IWUSR|S_IRUSR},
1367 [NFSD_Pool_Threads] = {"pool_threads", &transaction_ops, S_IWUSR|S_IRUSR},
1368 [NFSD_Pool_Stats] = {"pool_stats", &pool_stats_operations, S_IRUGO},
1369 [NFSD_Reply_Cache_Stats] = {"reply_cache_stats", &reply_cache_stats_operations, S_IRUGO},
1370 [NFSD_Versions] = {"versions", &transaction_ops, S_IWUSR|S_IRUSR},
1371 [NFSD_Ports] = {"portlist", &transaction_ops, S_IWUSR|S_IRUGO},
1372 [NFSD_MaxBlkSize] = {"max_block_size", &transaction_ops, S_IWUSR|S_IRUGO},
1373 [NFSD_MaxConnections] = {"max_connections", &transaction_ops, S_IWUSR|S_IRUGO},
1374 #if defined(CONFIG_SUNRPC_GSS) || defined(CONFIG_SUNRPC_GSS_MODULE)
1375 [NFSD_SupportedEnctypes] = {"supported_krb5_enctypes", &supported_enctypes_ops, S_IRUGO},
1376 #endif /* CONFIG_SUNRPC_GSS or CONFIG_SUNRPC_GSS_MODULE */
1377 #ifdef CONFIG_NFSD_V4
1378 [NFSD_Leasetime] = {"nfsv4leasetime", &transaction_ops, S_IWUSR|S_IRUSR},
1379 [NFSD_Gracetime] = {"nfsv4gracetime", &transaction_ops, S_IWUSR|S_IRUSR},
1380 [NFSD_RecoveryDir] = {"nfsv4recoverydir", &transaction_ops, S_IWUSR|S_IRUSR},
1381 [NFSD_V4EndGrace] = {"v4_end_grace", &transaction_ops, S_IWUSR|S_IRUGO},
1386 ret = simple_fill_super(sb, 0x6e667364, nfsd_files);
1389 dentry = nfsd_mkdir(sb->s_root, NULL, "clients");
1391 return PTR_ERR(dentry);
1392 nn->nfsd_client_dir = dentry;
1396 static int nfsd_fs_get_tree(struct fs_context *fc)
1398 return get_tree_keyed(fc, nfsd_fill_super, get_net(fc->net_ns));
1401 static void nfsd_fs_free_fc(struct fs_context *fc)
1404 put_net(fc->s_fs_info);
1407 static const struct fs_context_operations nfsd_fs_context_ops = {
1408 .free = nfsd_fs_free_fc,
1409 .get_tree = nfsd_fs_get_tree,
1412 static int nfsd_init_fs_context(struct fs_context *fc)
1414 put_user_ns(fc->user_ns);
1415 fc->user_ns = get_user_ns(fc->net_ns->user_ns);
1416 fc->ops = &nfsd_fs_context_ops;
1420 static void nfsd_umount(struct super_block *sb)
1422 struct net *net = sb->s_fs_info;
1424 nfsd_shutdown_threads(net);
1426 kill_litter_super(sb);
1430 static struct file_system_type nfsd_fs_type = {
1431 .owner = THIS_MODULE,
1433 .init_fs_context = nfsd_init_fs_context,
1434 .kill_sb = nfsd_umount,
1436 MODULE_ALIAS_FS("nfsd");
1438 #ifdef CONFIG_PROC_FS
1439 static int create_proc_exports_entry(void)
1441 struct proc_dir_entry *entry;
1443 entry = proc_mkdir("fs/nfs", NULL);
1446 entry = proc_create("exports", 0, entry, &exports_proc_ops);
1448 remove_proc_entry("fs/nfs", NULL);
1453 #else /* CONFIG_PROC_FS */
1454 static int create_proc_exports_entry(void)
1460 unsigned int nfsd_net_id;
1462 static __net_init int nfsd_init_net(struct net *net)
1465 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
1467 retval = nfsd_export_init(net);
1469 goto out_export_error;
1470 retval = nfsd_idmap_init(net);
1472 goto out_idmap_error;
1473 nn->nfsd_versions = NULL;
1474 nn->nfsd4_minorversions = NULL;
1475 retval = nfsd_reply_cache_init(nn);
1478 nn->nfsd4_lease = 90; /* default lease time */
1479 nn->nfsd4_grace = 90;
1480 nn->somebody_reclaimed = false;
1481 nn->track_reclaim_completes = false;
1482 nn->clverifier_counter = prandom_u32();
1483 nn->clientid_base = prandom_u32();
1484 nn->clientid_counter = nn->clientid_base + 1;
1485 nn->s2s_cp_cl_id = nn->clientid_counter++;
1487 get_random_bytes(&nn->siphash_key, sizeof(nn->siphash_key));
1488 seqlock_init(&nn->writeverf_lock);
1493 nfsd_idmap_shutdown(net);
1495 nfsd_export_shutdown(net);
1500 static __net_exit void nfsd_exit_net(struct net *net)
1502 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
1504 nfsd_reply_cache_shutdown(nn);
1505 nfsd_idmap_shutdown(net);
1506 nfsd_export_shutdown(net);
1507 nfsd_netns_free_versions(net_generic(net, nfsd_net_id));
1510 static struct pernet_operations nfsd_net_ops = {
1511 .init = nfsd_init_net,
1512 .exit = nfsd_exit_net,
1514 .size = sizeof(struct nfsd_net),
1517 static int __init init_nfsd(void)
1522 retval = nfsd4_init_slabs();
1525 retval = nfsd4_init_pnfs();
1527 goto out_free_slabs;
1528 retval = nfsd_stat_init(); /* Statistics */
1531 retval = nfsd_drc_slab_create();
1534 nfsd_lockd_init(); /* lockd->nfsd callbacks */
1535 retval = create_proc_exports_entry();
1537 goto out_free_lockd;
1538 retval = register_filesystem(&nfsd_fs_type);
1540 goto out_free_exports;
1541 retval = register_pernet_subsys(&nfsd_net_ops);
1543 goto out_free_filesystem;
1544 retval = register_cld_notifier();
1549 unregister_pernet_subsys(&nfsd_net_ops);
1550 out_free_filesystem:
1551 unregister_filesystem(&nfsd_fs_type);
1553 remove_proc_entry("fs/nfs/exports", NULL);
1554 remove_proc_entry("fs/nfs", NULL);
1556 nfsd_lockd_shutdown();
1557 nfsd_drc_slab_free();
1559 nfsd_stat_shutdown();
1567 static void __exit exit_nfsd(void)
1569 unregister_cld_notifier();
1570 unregister_pernet_subsys(&nfsd_net_ops);
1571 nfsd_drc_slab_free();
1572 remove_proc_entry("fs/nfs/exports", NULL);
1573 remove_proc_entry("fs/nfs", NULL);
1574 nfsd_stat_shutdown();
1575 nfsd_lockd_shutdown();
1578 unregister_filesystem(&nfsd_fs_type);
1582 MODULE_LICENSE("GPL");
1583 module_init(init_nfsd)
1584 module_exit(exit_nfsd)