2 * Copyright (c) 2004 The Regents of the University of Michigan.
3 * Copyright (c) 2012 Jeff Layton <jlayton@redhat.com>
6 * Andy Adamson <andros@citi.umich.edu>
8 * Redistribution and use in source and binary forms, with or without
9 * modification, are permitted provided that the following conditions
12 * 1. Redistributions of source code must retain the above copyright
13 * notice, this list of conditions and the following disclaimer.
14 * 2. Redistributions in binary form must reproduce the above copyright
15 * notice, this list of conditions and the following disclaimer in the
16 * documentation and/or other materials provided with the distribution.
17 * 3. Neither the name of the University nor the names of its
18 * contributors may be used to endorse or promote products derived
19 * from this software without specific prior written permission.
21 * THIS SOFTWARE IS PROVIDED ``AS IS'' AND ANY EXPRESS OR IMPLIED
22 * WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
23 * MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
24 * DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
25 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
26 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
27 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR
28 * BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
29 * LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
30 * NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
31 * SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
35 #include <linux/file.h>
36 #include <linux/slab.h>
37 #include <linux/namei.h>
38 #include <linux/crypto.h>
39 #include <linux/sched.h>
41 #include <linux/module.h>
42 #include <net/net_namespace.h>
43 #include <linux/sunrpc/rpc_pipe_fs.h>
44 #include <linux/sunrpc/clnt.h>
45 #include <linux/nfsd/cld.h>
52 #define NFSDDBG_FACILITY NFSDDBG_PROC
55 struct nfsd4_client_tracking_ops {
56 int (*init)(struct net *);
57 void (*exit)(struct net *);
58 void (*create)(struct nfs4_client *);
59 void (*remove)(struct nfs4_client *);
60 int (*check)(struct nfs4_client *);
61 void (*grace_done)(struct nfsd_net *);
65 static char user_recovery_dirname[PATH_MAX] = "/var/lib/nfs/v4recovery";
68 nfs4_save_creds(const struct cred **original_creds)
72 new = prepare_creds();
76 new->fsuid = GLOBAL_ROOT_UID;
77 new->fsgid = GLOBAL_ROOT_GID;
78 *original_creds = override_creds(new);
84 nfs4_reset_creds(const struct cred *original)
86 revert_creds(original);
90 md5_to_hex(char *out, char *md5)
94 for (i=0; i<16; i++) {
95 unsigned char c = md5[i];
97 *out++ = '0' + ((c&0xf0)>>4) + (c>=0xa0)*('a'-'9'-1);
98 *out++ = '0' + (c&0x0f) + ((c&0x0f)>=0x0a)*('a'-'9'-1);
104 nfs4_make_rec_clidname(char *dname, const struct xdr_netobj *clname)
106 struct xdr_netobj cksum;
107 struct hash_desc desc;
108 struct scatterlist sg;
111 dprintk("NFSD: nfs4_make_rec_clidname for %.*s\n",
112 clname->len, clname->data);
113 desc.flags = CRYPTO_TFM_REQ_MAY_SLEEP;
114 desc.tfm = crypto_alloc_hash("md5", 0, CRYPTO_ALG_ASYNC);
115 if (IS_ERR(desc.tfm)) {
116 status = PTR_ERR(desc.tfm);
120 cksum.len = crypto_hash_digestsize(desc.tfm);
121 cksum.data = kmalloc(cksum.len, GFP_KERNEL);
122 if (cksum.data == NULL) {
127 sg_init_one(&sg, clname->data, clname->len);
129 status = crypto_hash_digest(&desc, &sg, sg.length, cksum.data);
133 md5_to_hex(dname, cksum.data);
138 crypto_free_hash(desc.tfm);
144 * If we had an error generating the recdir name for the legacy tracker
145 * then warn the admin. If the error doesn't appear to be transient,
146 * then disable recovery tracking.
149 legacy_recdir_name_error(struct nfs4_client *clp, int error)
151 printk(KERN_ERR "NFSD: unable to generate recoverydir "
152 "name (%d).\n", error);
155 * if the algorithm just doesn't exist, then disable the recovery
156 * tracker altogether. The crypto libs will generally return this if
157 * FIPS is enabled as well.
159 if (error == -ENOENT) {
160 printk(KERN_ERR "NFSD: disabling legacy clientid tracking. "
161 "Reboot recovery will not function correctly!\n");
162 nfsd4_client_tracking_exit(clp->net);
167 nfsd4_create_clid_dir(struct nfs4_client *clp)
169 const struct cred *original_cred;
170 char dname[HEXDIR_LEN];
171 struct dentry *dir, *dentry;
172 struct nfs4_client_reclaim *crp;
174 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
176 if (test_and_set_bit(NFSD4_CLIENT_STABLE, &clp->cl_flags))
181 status = nfs4_make_rec_clidname(dname, &clp->cl_name);
183 return legacy_recdir_name_error(clp, status);
185 status = nfs4_save_creds(&original_cred);
189 status = mnt_want_write_file(nn->rec_file);
193 dir = nn->rec_file->f_path.dentry;
194 /* lock the parent */
195 mutex_lock(&dir->d_inode->i_mutex);
197 dentry = lookup_one_len(dname, dir, HEXDIR_LEN-1);
198 if (IS_ERR(dentry)) {
199 status = PTR_ERR(dentry);
204 * In the 4.1 case, where we're called from
205 * reclaim_complete(), records from the previous reboot
206 * may still be left, so this is OK.
208 * In the 4.0 case, we should never get here; but we may
209 * as well be forgiving and just succeed silently.
212 status = vfs_mkdir(dir->d_inode, dentry, S_IRWXU);
216 mutex_unlock(&dir->d_inode->i_mutex);
219 crp = nfs4_client_to_reclaim(dname, nn);
223 vfs_fsync(nn->rec_file, 0);
225 printk(KERN_ERR "NFSD: failed to write recovery record"
226 " (err %d); please check that %s exists"
227 " and is writeable", status,
228 user_recovery_dirname);
230 mnt_drop_write_file(nn->rec_file);
232 nfs4_reset_creds(original_cred);
235 typedef int (recdir_func)(struct dentry *, struct dentry *, struct nfsd_net *);
238 char name[HEXDIR_LEN];
239 struct list_head list;
242 struct nfs4_dir_ctx {
243 struct dir_context ctx;
244 struct list_head names;
248 nfsd4_build_namelist(void *arg, const char *name, int namlen,
249 loff_t offset, u64 ino, unsigned int d_type)
251 struct nfs4_dir_ctx *ctx = arg;
252 struct name_list *entry;
254 if (namlen != HEXDIR_LEN - 1)
256 entry = kmalloc(sizeof(struct name_list), GFP_KERNEL);
259 memcpy(entry->name, name, HEXDIR_LEN - 1);
260 entry->name[HEXDIR_LEN - 1] = '\0';
261 list_add(&entry->list, &ctx->names);
266 nfsd4_list_rec_dir(recdir_func *f, struct nfsd_net *nn)
268 const struct cred *original_cred;
269 struct dentry *dir = nn->rec_file->f_path.dentry;
270 struct nfs4_dir_ctx ctx = {
271 .ctx.actor = nfsd4_build_namelist,
272 .names = LIST_HEAD_INIT(ctx.names)
276 status = nfs4_save_creds(&original_cred);
280 status = vfs_llseek(nn->rec_file, 0, SEEK_SET);
282 nfs4_reset_creds(original_cred);
286 status = iterate_dir(nn->rec_file, &ctx.ctx);
287 mutex_lock_nested(&dir->d_inode->i_mutex, I_MUTEX_PARENT);
288 while (!list_empty(&ctx.names)) {
289 struct name_list *entry;
290 entry = list_entry(ctx.names.next, struct name_list, list);
292 struct dentry *dentry;
293 dentry = lookup_one_len(entry->name, dir, HEXDIR_LEN-1);
294 if (IS_ERR(dentry)) {
295 status = PTR_ERR(dentry);
298 status = f(dir, dentry, nn);
301 list_del(&entry->list);
304 mutex_unlock(&dir->d_inode->i_mutex);
305 nfs4_reset_creds(original_cred);
310 nfsd4_unlink_clid_dir(char *name, int namlen, struct nfsd_net *nn)
312 struct dentry *dir, *dentry;
315 dprintk("NFSD: nfsd4_unlink_clid_dir. name %.*s\n", namlen, name);
317 dir = nn->rec_file->f_path.dentry;
318 mutex_lock_nested(&dir->d_inode->i_mutex, I_MUTEX_PARENT);
319 dentry = lookup_one_len(name, dir, namlen);
320 if (IS_ERR(dentry)) {
321 status = PTR_ERR(dentry);
325 if (!dentry->d_inode)
327 status = vfs_rmdir(dir->d_inode, dentry);
331 mutex_unlock(&dir->d_inode->i_mutex);
336 nfsd4_remove_clid_dir(struct nfs4_client *clp)
338 const struct cred *original_cred;
339 struct nfs4_client_reclaim *crp;
340 char dname[HEXDIR_LEN];
342 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
344 if (!nn->rec_file || !test_bit(NFSD4_CLIENT_STABLE, &clp->cl_flags))
347 status = nfs4_make_rec_clidname(dname, &clp->cl_name);
349 return legacy_recdir_name_error(clp, status);
351 status = mnt_want_write_file(nn->rec_file);
354 clear_bit(NFSD4_CLIENT_STABLE, &clp->cl_flags);
356 status = nfs4_save_creds(&original_cred);
360 status = nfsd4_unlink_clid_dir(dname, HEXDIR_LEN-1, nn);
361 nfs4_reset_creds(original_cred);
363 vfs_fsync(nn->rec_file, 0);
365 /* remove reclaim record */
366 crp = nfsd4_find_reclaim_client(dname, nn);
368 nfs4_remove_reclaim_record(crp, nn);
372 mnt_drop_write_file(nn->rec_file);
375 printk("NFSD: Failed to remove expired client state directory"
376 " %.*s\n", HEXDIR_LEN, dname);
380 purge_old(struct dentry *parent, struct dentry *child, struct nfsd_net *nn)
384 if (nfs4_has_reclaimed_state(child->d_name.name, nn))
387 status = vfs_rmdir(parent->d_inode, child);
389 printk("failed to remove client recovery directory %pd\n",
391 /* Keep trying, success or failure: */
396 nfsd4_recdir_purge_old(struct nfsd_net *nn)
400 nn->in_grace = false;
403 status = mnt_want_write_file(nn->rec_file);
406 status = nfsd4_list_rec_dir(purge_old, nn);
408 vfs_fsync(nn->rec_file, 0);
409 mnt_drop_write_file(nn->rec_file);
411 nfs4_release_reclaim(nn);
413 printk("nfsd4: failed to purge old clients from recovery"
414 " directory %pD\n", nn->rec_file);
418 load_recdir(struct dentry *parent, struct dentry *child, struct nfsd_net *nn)
420 if (child->d_name.len != HEXDIR_LEN - 1) {
421 printk("nfsd4: illegal name %pd in recovery directory\n",
423 /* Keep trying; maybe the others are OK: */
426 nfs4_client_to_reclaim(child->d_name.name, nn);
431 nfsd4_recdir_load(struct net *net) {
433 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
438 status = nfsd4_list_rec_dir(load_recdir, nn);
440 printk("nfsd4: failed loading clients from recovery"
441 " directory %pD\n", nn->rec_file);
446 * Hold reference to the recovery directory.
450 nfsd4_init_recdir(struct net *net)
452 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
453 const struct cred *original_cred;
456 printk("NFSD: Using %s as the NFSv4 state recovery directory\n",
457 user_recovery_dirname);
459 BUG_ON(nn->rec_file);
461 status = nfs4_save_creds(&original_cred);
463 printk("NFSD: Unable to change credentials to find recovery"
464 " directory: error %d\n",
469 nn->rec_file = filp_open(user_recovery_dirname, O_RDONLY | O_DIRECTORY, 0);
470 if (IS_ERR(nn->rec_file)) {
471 printk("NFSD: unable to find recovery directory %s\n",
472 user_recovery_dirname);
473 status = PTR_ERR(nn->rec_file);
477 nfs4_reset_creds(original_cred);
484 nfsd4_shutdown_recdir(struct net *net)
486 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
495 nfs4_legacy_state_init(struct net *net)
497 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
500 nn->reclaim_str_hashtbl = kmalloc(sizeof(struct list_head) *
501 CLIENT_HASH_SIZE, GFP_KERNEL);
502 if (!nn->reclaim_str_hashtbl)
505 for (i = 0; i < CLIENT_HASH_SIZE; i++)
506 INIT_LIST_HEAD(&nn->reclaim_str_hashtbl[i]);
507 nn->reclaim_str_hashtbl_size = 0;
513 nfs4_legacy_state_shutdown(struct net *net)
515 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
517 kfree(nn->reclaim_str_hashtbl);
521 nfsd4_load_reboot_recovery_data(struct net *net)
525 status = nfsd4_init_recdir(net);
529 status = nfsd4_recdir_load(net);
531 nfsd4_shutdown_recdir(net);
537 nfsd4_legacy_tracking_init(struct net *net)
541 /* XXX: The legacy code won't work in a container */
542 if (net != &init_net) {
543 WARN(1, KERN_ERR "NFSD: attempt to initialize legacy client "
544 "tracking in a container!\n");
548 status = nfs4_legacy_state_init(net);
552 status = nfsd4_load_reboot_recovery_data(net);
558 nfs4_legacy_state_shutdown(net);
563 nfsd4_legacy_tracking_exit(struct net *net)
565 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
567 nfs4_release_reclaim(nn);
568 nfsd4_shutdown_recdir(net);
569 nfs4_legacy_state_shutdown(net);
573 * Change the NFSv4 recovery directory to recdir.
576 nfs4_reset_recoverydir(char *recdir)
581 status = kern_path(recdir, LOOKUP_FOLLOW, &path);
585 if (S_ISDIR(path.dentry->d_inode->i_mode)) {
586 strcpy(user_recovery_dirname, recdir);
594 nfs4_recoverydir(void)
596 return user_recovery_dirname;
600 nfsd4_check_legacy_client(struct nfs4_client *clp)
603 char dname[HEXDIR_LEN];
604 struct nfs4_client_reclaim *crp;
605 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
607 /* did we already find that this client is stable? */
608 if (test_bit(NFSD4_CLIENT_STABLE, &clp->cl_flags))
611 status = nfs4_make_rec_clidname(dname, &clp->cl_name);
613 legacy_recdir_name_error(clp, status);
617 /* look for it in the reclaim hashtable otherwise */
618 crp = nfsd4_find_reclaim_client(dname, nn);
620 set_bit(NFSD4_CLIENT_STABLE, &clp->cl_flags);
628 static struct nfsd4_client_tracking_ops nfsd4_legacy_tracking_ops = {
629 .init = nfsd4_legacy_tracking_init,
630 .exit = nfsd4_legacy_tracking_exit,
631 .create = nfsd4_create_clid_dir,
632 .remove = nfsd4_remove_clid_dir,
633 .check = nfsd4_check_legacy_client,
634 .grace_done = nfsd4_recdir_purge_old,
638 #define NFSD_PIPE_DIR "nfsd"
639 #define NFSD_CLD_PIPE "cld"
641 /* per-net-ns structure for holding cld upcall info */
643 struct rpc_pipe *cn_pipe;
645 struct list_head cn_list;
650 struct list_head cu_list;
651 struct cld_net *cu_net;
652 struct task_struct *cu_task;
653 struct cld_msg cu_msg;
657 __cld_pipe_upcall(struct rpc_pipe *pipe, struct cld_msg *cmsg)
660 struct rpc_pipe_msg msg;
662 memset(&msg, 0, sizeof(msg));
664 msg.len = sizeof(*cmsg);
667 * Set task state before we queue the upcall. That prevents
668 * wake_up_process in the downcall from racing with schedule.
670 set_current_state(TASK_UNINTERRUPTIBLE);
671 ret = rpc_queue_upcall(pipe, &msg);
673 set_current_state(TASK_RUNNING);
686 cld_pipe_upcall(struct rpc_pipe *pipe, struct cld_msg *cmsg)
691 * -EAGAIN occurs when pipe is closed and reopened while there are
695 ret = __cld_pipe_upcall(pipe, cmsg);
696 } while (ret == -EAGAIN);
702 cld_pipe_downcall(struct file *filp, const char __user *src, size_t mlen)
704 struct cld_upcall *tmp, *cup;
705 struct cld_msg __user *cmsg = (struct cld_msg __user *)src;
707 struct nfsd_net *nn = net_generic(filp->f_dentry->d_sb->s_fs_info,
709 struct cld_net *cn = nn->cld_net;
711 if (mlen != sizeof(*cmsg)) {
712 dprintk("%s: got %zu bytes, expected %zu\n", __func__, mlen,
717 /* copy just the xid so we can try to find that */
718 if (copy_from_user(&xid, &cmsg->cm_xid, sizeof(xid)) != 0) {
719 dprintk("%s: error when copying xid from userspace", __func__);
723 /* walk the list and find corresponding xid */
725 spin_lock(&cn->cn_lock);
726 list_for_each_entry(tmp, &cn->cn_list, cu_list) {
727 if (get_unaligned(&tmp->cu_msg.cm_xid) == xid) {
729 list_del_init(&cup->cu_list);
733 spin_unlock(&cn->cn_lock);
735 /* couldn't find upcall? */
737 dprintk("%s: couldn't find upcall -- xid=%u\n", __func__, xid);
741 if (copy_from_user(&cup->cu_msg, src, mlen) != 0)
744 wake_up_process(cup->cu_task);
749 cld_pipe_destroy_msg(struct rpc_pipe_msg *msg)
751 struct cld_msg *cmsg = msg->data;
752 struct cld_upcall *cup = container_of(cmsg, struct cld_upcall,
755 /* errno >= 0 means we got a downcall */
759 wake_up_process(cup->cu_task);
762 static const struct rpc_pipe_ops cld_upcall_ops = {
763 .upcall = rpc_pipe_generic_upcall,
764 .downcall = cld_pipe_downcall,
765 .destroy_msg = cld_pipe_destroy_msg,
768 static struct dentry *
769 nfsd4_cld_register_sb(struct super_block *sb, struct rpc_pipe *pipe)
771 struct dentry *dir, *dentry;
773 dir = rpc_d_lookup_sb(sb, NFSD_PIPE_DIR);
775 return ERR_PTR(-ENOENT);
776 dentry = rpc_mkpipe_dentry(dir, NFSD_CLD_PIPE, NULL, pipe);
782 nfsd4_cld_unregister_sb(struct rpc_pipe *pipe)
785 rpc_unlink(pipe->dentry);
788 static struct dentry *
789 nfsd4_cld_register_net(struct net *net, struct rpc_pipe *pipe)
791 struct super_block *sb;
792 struct dentry *dentry;
794 sb = rpc_get_sb_net(net);
797 dentry = nfsd4_cld_register_sb(sb, pipe);
803 nfsd4_cld_unregister_net(struct net *net, struct rpc_pipe *pipe)
805 struct super_block *sb;
807 sb = rpc_get_sb_net(net);
809 nfsd4_cld_unregister_sb(pipe);
814 /* Initialize rpc_pipefs pipe for communication with client tracking daemon */
816 nfsd4_init_cld_pipe(struct net *net)
819 struct dentry *dentry;
820 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
826 cn = kzalloc(sizeof(*cn), GFP_KERNEL);
832 cn->cn_pipe = rpc_mkpipe_data(&cld_upcall_ops, RPC_PIPE_WAIT_FOR_OPEN);
833 if (IS_ERR(cn->cn_pipe)) {
834 ret = PTR_ERR(cn->cn_pipe);
837 spin_lock_init(&cn->cn_lock);
838 INIT_LIST_HEAD(&cn->cn_list);
840 dentry = nfsd4_cld_register_net(net, cn->cn_pipe);
841 if (IS_ERR(dentry)) {
842 ret = PTR_ERR(dentry);
843 goto err_destroy_data;
846 cn->cn_pipe->dentry = dentry;
851 rpc_destroy_pipe_data(cn->cn_pipe);
854 printk(KERN_ERR "NFSD: unable to create nfsdcld upcall pipe (%d)\n",
860 nfsd4_remove_cld_pipe(struct net *net)
862 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
863 struct cld_net *cn = nn->cld_net;
865 nfsd4_cld_unregister_net(net, cn->cn_pipe);
866 rpc_destroy_pipe_data(cn->cn_pipe);
871 static struct cld_upcall *
872 alloc_cld_upcall(struct cld_net *cn)
874 struct cld_upcall *new, *tmp;
876 new = kzalloc(sizeof(*new), GFP_KERNEL);
880 /* FIXME: hard cap on number in flight? */
882 spin_lock(&cn->cn_lock);
883 list_for_each_entry(tmp, &cn->cn_list, cu_list) {
884 if (tmp->cu_msg.cm_xid == cn->cn_xid) {
886 spin_unlock(&cn->cn_lock);
890 new->cu_task = current;
891 new->cu_msg.cm_vers = CLD_UPCALL_VERSION;
892 put_unaligned(cn->cn_xid++, &new->cu_msg.cm_xid);
894 list_add(&new->cu_list, &cn->cn_list);
895 spin_unlock(&cn->cn_lock);
897 dprintk("%s: allocated xid %u\n", __func__, new->cu_msg.cm_xid);
903 free_cld_upcall(struct cld_upcall *victim)
905 struct cld_net *cn = victim->cu_net;
907 spin_lock(&cn->cn_lock);
908 list_del(&victim->cu_list);
909 spin_unlock(&cn->cn_lock);
913 /* Ask daemon to create a new record */
915 nfsd4_cld_create(struct nfs4_client *clp)
918 struct cld_upcall *cup;
919 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
920 struct cld_net *cn = nn->cld_net;
922 /* Don't upcall if it's already stored */
923 if (test_bit(NFSD4_CLIENT_STABLE, &clp->cl_flags))
926 cup = alloc_cld_upcall(cn);
932 cup->cu_msg.cm_cmd = Cld_Create;
933 cup->cu_msg.cm_u.cm_name.cn_len = clp->cl_name.len;
934 memcpy(cup->cu_msg.cm_u.cm_name.cn_id, clp->cl_name.data,
937 ret = cld_pipe_upcall(cn->cn_pipe, &cup->cu_msg);
939 ret = cup->cu_msg.cm_status;
940 set_bit(NFSD4_CLIENT_STABLE, &clp->cl_flags);
943 free_cld_upcall(cup);
946 printk(KERN_ERR "NFSD: Unable to create client "
947 "record on stable storage: %d\n", ret);
950 /* Ask daemon to create a new record */
952 nfsd4_cld_remove(struct nfs4_client *clp)
955 struct cld_upcall *cup;
956 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
957 struct cld_net *cn = nn->cld_net;
959 /* Don't upcall if it's already removed */
960 if (!test_bit(NFSD4_CLIENT_STABLE, &clp->cl_flags))
963 cup = alloc_cld_upcall(cn);
969 cup->cu_msg.cm_cmd = Cld_Remove;
970 cup->cu_msg.cm_u.cm_name.cn_len = clp->cl_name.len;
971 memcpy(cup->cu_msg.cm_u.cm_name.cn_id, clp->cl_name.data,
974 ret = cld_pipe_upcall(cn->cn_pipe, &cup->cu_msg);
976 ret = cup->cu_msg.cm_status;
977 clear_bit(NFSD4_CLIENT_STABLE, &clp->cl_flags);
980 free_cld_upcall(cup);
983 printk(KERN_ERR "NFSD: Unable to remove client "
984 "record from stable storage: %d\n", ret);
987 /* Check for presence of a record, and update its timestamp */
989 nfsd4_cld_check(struct nfs4_client *clp)
992 struct cld_upcall *cup;
993 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
994 struct cld_net *cn = nn->cld_net;
996 /* Don't upcall if one was already stored during this grace pd */
997 if (test_bit(NFSD4_CLIENT_STABLE, &clp->cl_flags))
1000 cup = alloc_cld_upcall(cn);
1002 printk(KERN_ERR "NFSD: Unable to check client record on "
1003 "stable storage: %d\n", -ENOMEM);
1007 cup->cu_msg.cm_cmd = Cld_Check;
1008 cup->cu_msg.cm_u.cm_name.cn_len = clp->cl_name.len;
1009 memcpy(cup->cu_msg.cm_u.cm_name.cn_id, clp->cl_name.data,
1012 ret = cld_pipe_upcall(cn->cn_pipe, &cup->cu_msg);
1014 ret = cup->cu_msg.cm_status;
1015 set_bit(NFSD4_CLIENT_STABLE, &clp->cl_flags);
1018 free_cld_upcall(cup);
1023 nfsd4_cld_grace_done(struct nfsd_net *nn)
1026 struct cld_upcall *cup;
1027 struct cld_net *cn = nn->cld_net;
1029 cup = alloc_cld_upcall(cn);
1035 cup->cu_msg.cm_cmd = Cld_GraceDone;
1036 cup->cu_msg.cm_u.cm_gracetime = (int64_t)nn->boot_time;
1037 ret = cld_pipe_upcall(cn->cn_pipe, &cup->cu_msg);
1039 ret = cup->cu_msg.cm_status;
1041 free_cld_upcall(cup);
1044 printk(KERN_ERR "NFSD: Unable to end grace period: %d\n", ret);
1047 static struct nfsd4_client_tracking_ops nfsd4_cld_tracking_ops = {
1048 .init = nfsd4_init_cld_pipe,
1049 .exit = nfsd4_remove_cld_pipe,
1050 .create = nfsd4_cld_create,
1051 .remove = nfsd4_cld_remove,
1052 .check = nfsd4_cld_check,
1053 .grace_done = nfsd4_cld_grace_done,
1056 /* upcall via usermodehelper */
1057 static char cltrack_prog[PATH_MAX] = "/sbin/nfsdcltrack";
1058 module_param_string(cltrack_prog, cltrack_prog, sizeof(cltrack_prog),
1060 MODULE_PARM_DESC(cltrack_prog, "Path to the nfsdcltrack upcall program");
1062 static bool cltrack_legacy_disable;
1063 module_param(cltrack_legacy_disable, bool, S_IRUGO|S_IWUSR);
1064 MODULE_PARM_DESC(cltrack_legacy_disable,
1065 "Disable legacy recoverydir conversion. Default: false");
1067 #define LEGACY_TOPDIR_ENV_PREFIX "NFSDCLTRACK_LEGACY_TOPDIR="
1068 #define LEGACY_RECDIR_ENV_PREFIX "NFSDCLTRACK_LEGACY_RECDIR="
1069 #define HAS_SESSION_ENV_PREFIX "NFSDCLTRACK_CLIENT_HAS_SESSION="
1070 #define GRACE_START_ENV_PREFIX "NFSDCLTRACK_GRACE_START="
1073 nfsd4_cltrack_legacy_topdir(void)
1079 if (cltrack_legacy_disable)
1082 len = strlen(LEGACY_TOPDIR_ENV_PREFIX) +
1083 strlen(nfs4_recoverydir()) + 1;
1085 result = kmalloc(len, GFP_KERNEL);
1089 copied = snprintf(result, len, LEGACY_TOPDIR_ENV_PREFIX "%s",
1090 nfs4_recoverydir());
1091 if (copied >= len) {
1092 /* just return nothing if output was truncated */
1101 nfsd4_cltrack_legacy_recdir(const struct xdr_netobj *name)
1107 if (cltrack_legacy_disable)
1110 /* +1 is for '/' between "topdir" and "recdir" */
1111 len = strlen(LEGACY_RECDIR_ENV_PREFIX) +
1112 strlen(nfs4_recoverydir()) + 1 + HEXDIR_LEN;
1114 result = kmalloc(len, GFP_KERNEL);
1118 copied = snprintf(result, len, LEGACY_RECDIR_ENV_PREFIX "%s/",
1119 nfs4_recoverydir());
1120 if (copied > (len - HEXDIR_LEN)) {
1121 /* just return nothing if output will be truncated */
1126 copied = nfs4_make_rec_clidname(result + copied, name);
1136 nfsd4_cltrack_client_has_session(struct nfs4_client *clp)
1142 /* prefix + Y/N character + terminating NULL */
1143 len = strlen(HAS_SESSION_ENV_PREFIX) + 1 + 1;
1145 result = kmalloc(len, GFP_KERNEL);
1149 copied = snprintf(result, len, HAS_SESSION_ENV_PREFIX "%c",
1150 clp->cl_minorversion ? 'Y' : 'N');
1151 if (copied >= len) {
1152 /* just return nothing if output was truncated */
1161 nfsd4_cltrack_grace_start(time_t grace_start)
1167 /* prefix + max width of int64_t string + terminating NULL */
1168 len = strlen(GRACE_START_ENV_PREFIX) + 22 + 1;
1170 result = kmalloc(len, GFP_KERNEL);
1174 copied = snprintf(result, len, GRACE_START_ENV_PREFIX "%ld",
1176 if (copied >= len) {
1177 /* just return nothing if output was truncated */
1186 nfsd4_umh_cltrack_upcall(char *cmd, char *arg, char *env0, char *env1)
1192 if (unlikely(!cltrack_prog[0])) {
1193 dprintk("%s: cltrack_prog is disabled\n", __func__);
1197 dprintk("%s: cmd: %s\n", __func__, cmd);
1198 dprintk("%s: arg: %s\n", __func__, arg ? arg : "(null)");
1199 dprintk("%s: env0: %s\n", __func__, env0 ? env0 : "(null)");
1200 dprintk("%s: env1: %s\n", __func__, env1 ? env1 : "(null)");
1206 argv[0] = (char *)cltrack_prog;
1211 ret = call_usermodehelper(argv[0], argv, envp, UMH_WAIT_PROC);
1213 * Disable the upcall mechanism if we're getting an ENOENT or EACCES
1214 * error. The admin can re-enable it on the fly by using sysfs
1215 * once the problem has been fixed.
1217 if (ret == -ENOENT || ret == -EACCES) {
1218 dprintk("NFSD: %s was not found or isn't executable (%d). "
1219 "Setting cltrack_prog to blank string!",
1221 cltrack_prog[0] = '\0';
1223 dprintk("%s: %s return value: %d\n", __func__, cltrack_prog, ret);
1229 bin_to_hex_dup(const unsigned char *src, int srclen)
1234 /* +1 for terminating NULL */
1235 buf = kmalloc((srclen * 2) + 1, GFP_KERNEL);
1240 for (i = 0; i < srclen; i++) {
1241 sprintf(hex, "%2.2x", *src++);
1248 nfsd4_umh_cltrack_init(struct net *net)
1251 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
1252 char *grace_start = nfsd4_cltrack_grace_start(nn->boot_time);
1254 /* XXX: The usermode helper s not working in container yet. */
1255 if (net != &init_net) {
1256 WARN(1, KERN_ERR "NFSD: attempt to initialize umh client "
1257 "tracking in a container!\n");
1261 ret = nfsd4_umh_cltrack_upcall("init", NULL, grace_start, NULL);
1267 nfsd4_cltrack_upcall_lock(struct nfs4_client *clp)
1269 wait_on_bit_lock(&clp->cl_flags, NFSD4_CLIENT_UPCALL_LOCK,
1270 TASK_UNINTERRUPTIBLE);
1274 nfsd4_cltrack_upcall_unlock(struct nfs4_client *clp)
1276 smp_mb__before_atomic();
1277 clear_bit(NFSD4_CLIENT_UPCALL_LOCK, &clp->cl_flags);
1278 smp_mb__after_atomic();
1279 wake_up_bit(&clp->cl_flags, NFSD4_CLIENT_UPCALL_LOCK);
1283 nfsd4_umh_cltrack_create(struct nfs4_client *clp)
1285 char *hexid, *has_session, *grace_start;
1286 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
1289 * With v4.0 clients, there's little difference in outcome between a
1290 * create and check operation, and we can end up calling into this
1291 * function multiple times per client (once for each openowner). So,
1292 * for v4.0 clients skip upcalling once the client has been recorded
1293 * on stable storage.
1295 * For v4.1+ clients, the outcome of the two operations is different,
1296 * so we must ensure that we upcall for the create operation. v4.1+
1297 * clients call this on RECLAIM_COMPLETE though, so we should only end
1298 * up doing a single create upcall per client.
1300 if (clp->cl_minorversion == 0 &&
1301 test_bit(NFSD4_CLIENT_STABLE, &clp->cl_flags))
1304 hexid = bin_to_hex_dup(clp->cl_name.data, clp->cl_name.len);
1306 dprintk("%s: can't allocate memory for upcall!\n", __func__);
1310 has_session = nfsd4_cltrack_client_has_session(clp);
1311 grace_start = nfsd4_cltrack_grace_start(nn->boot_time);
1313 nfsd4_cltrack_upcall_lock(clp);
1314 if (!nfsd4_umh_cltrack_upcall("create", hexid, has_session, grace_start))
1315 set_bit(NFSD4_CLIENT_STABLE, &clp->cl_flags);
1316 nfsd4_cltrack_upcall_unlock(clp);
1324 nfsd4_umh_cltrack_remove(struct nfs4_client *clp)
1328 if (!test_bit(NFSD4_CLIENT_STABLE, &clp->cl_flags))
1331 hexid = bin_to_hex_dup(clp->cl_name.data, clp->cl_name.len);
1333 dprintk("%s: can't allocate memory for upcall!\n", __func__);
1337 nfsd4_cltrack_upcall_lock(clp);
1338 if (test_bit(NFSD4_CLIENT_STABLE, &clp->cl_flags) &&
1339 nfsd4_umh_cltrack_upcall("remove", hexid, NULL, NULL) == 0)
1340 clear_bit(NFSD4_CLIENT_STABLE, &clp->cl_flags);
1341 nfsd4_cltrack_upcall_unlock(clp);
1347 nfsd4_umh_cltrack_check(struct nfs4_client *clp)
1350 char *hexid, *has_session, *legacy;
1352 if (test_bit(NFSD4_CLIENT_STABLE, &clp->cl_flags))
1355 hexid = bin_to_hex_dup(clp->cl_name.data, clp->cl_name.len);
1357 dprintk("%s: can't allocate memory for upcall!\n", __func__);
1361 has_session = nfsd4_cltrack_client_has_session(clp);
1362 legacy = nfsd4_cltrack_legacy_recdir(&clp->cl_name);
1364 nfsd4_cltrack_upcall_lock(clp);
1365 if (test_bit(NFSD4_CLIENT_STABLE, &clp->cl_flags)) {
1368 ret = nfsd4_umh_cltrack_upcall("check", hexid, has_session, legacy);
1370 set_bit(NFSD4_CLIENT_STABLE, &clp->cl_flags);
1372 nfsd4_cltrack_upcall_unlock(clp);
1381 nfsd4_umh_cltrack_grace_done(struct nfsd_net *nn)
1384 char timestr[22]; /* FIXME: better way to determine max size? */
1386 sprintf(timestr, "%ld", nn->boot_time);
1387 legacy = nfsd4_cltrack_legacy_topdir();
1388 nfsd4_umh_cltrack_upcall("gracedone", timestr, legacy, NULL);
1392 static struct nfsd4_client_tracking_ops nfsd4_umh_tracking_ops = {
1393 .init = nfsd4_umh_cltrack_init,
1395 .create = nfsd4_umh_cltrack_create,
1396 .remove = nfsd4_umh_cltrack_remove,
1397 .check = nfsd4_umh_cltrack_check,
1398 .grace_done = nfsd4_umh_cltrack_grace_done,
1402 nfsd4_client_tracking_init(struct net *net)
1406 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
1408 /* just run the init if it the method is already decided */
1409 if (nn->client_tracking_ops)
1413 * First, try a UMH upcall. It should succeed or fail quickly, so
1414 * there's little harm in trying that first.
1416 nn->client_tracking_ops = &nfsd4_umh_tracking_ops;
1417 status = nn->client_tracking_ops->init(net);
1422 * See if the recoverydir exists and is a directory. If it is,
1423 * then use the legacy ops.
1425 nn->client_tracking_ops = &nfsd4_legacy_tracking_ops;
1426 status = kern_path(nfs4_recoverydir(), LOOKUP_FOLLOW, &path);
1428 status = S_ISDIR(path.dentry->d_inode->i_mode);
1434 /* Finally, try to use nfsdcld */
1435 nn->client_tracking_ops = &nfsd4_cld_tracking_ops;
1436 printk(KERN_WARNING "NFSD: the nfsdcld client tracking upcall will be "
1437 "removed in 3.10. Please transition to using "
1440 status = nn->client_tracking_ops->init(net);
1442 printk(KERN_WARNING "NFSD: Unable to initialize client "
1443 "recovery tracking! (%d)\n", status);
1444 nn->client_tracking_ops = NULL;
1450 nfsd4_client_tracking_exit(struct net *net)
1452 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
1454 if (nn->client_tracking_ops) {
1455 if (nn->client_tracking_ops->exit)
1456 nn->client_tracking_ops->exit(net);
1457 nn->client_tracking_ops = NULL;
1462 nfsd4_client_record_create(struct nfs4_client *clp)
1464 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
1466 if (nn->client_tracking_ops)
1467 nn->client_tracking_ops->create(clp);
1471 nfsd4_client_record_remove(struct nfs4_client *clp)
1473 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
1475 if (nn->client_tracking_ops)
1476 nn->client_tracking_ops->remove(clp);
1480 nfsd4_client_record_check(struct nfs4_client *clp)
1482 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
1484 if (nn->client_tracking_ops)
1485 return nn->client_tracking_ops->check(clp);
1491 nfsd4_record_grace_done(struct nfsd_net *nn)
1493 if (nn->client_tracking_ops)
1494 nn->client_tracking_ops->grace_done(nn);
1498 rpc_pipefs_event(struct notifier_block *nb, unsigned long event, void *ptr)
1500 struct super_block *sb = ptr;
1501 struct net *net = sb->s_fs_info;
1502 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
1503 struct cld_net *cn = nn->cld_net;
1504 struct dentry *dentry;
1507 if (!try_module_get(THIS_MODULE))
1511 module_put(THIS_MODULE);
1516 case RPC_PIPEFS_MOUNT:
1517 dentry = nfsd4_cld_register_sb(sb, cn->cn_pipe);
1518 if (IS_ERR(dentry)) {
1519 ret = PTR_ERR(dentry);
1522 cn->cn_pipe->dentry = dentry;
1524 case RPC_PIPEFS_UMOUNT:
1525 if (cn->cn_pipe->dentry)
1526 nfsd4_cld_unregister_sb(cn->cn_pipe);
1532 module_put(THIS_MODULE);
1536 static struct notifier_block nfsd4_cld_block = {
1537 .notifier_call = rpc_pipefs_event,
1541 register_cld_notifier(void)
1543 return rpc_pipefs_notifier_register(&nfsd4_cld_block);
1547 unregister_cld_notifier(void)
1549 rpc_pipefs_notifier_unregister(&nfsd4_cld_block);