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 <crypto/hash.h>
36 #include <linux/file.h>
37 #include <linux/slab.h>
38 #include <linux/namei.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 crypto_shash *tfm;
110 dprintk("NFSD: nfs4_make_rec_clidname for %.*s\n",
111 clname->len, clname->data);
112 tfm = crypto_alloc_shash("md5", 0, 0);
114 status = PTR_ERR(tfm);
118 cksum.len = crypto_shash_digestsize(tfm);
119 cksum.data = kmalloc(cksum.len, GFP_KERNEL);
120 if (cksum.data == NULL) {
126 SHASH_DESC_ON_STACK(desc, tfm);
129 desc->flags = CRYPTO_TFM_REQ_MAY_SLEEP;
131 status = crypto_shash_digest(desc, clname->data, clname->len,
133 shash_desc_zero(desc);
139 md5_to_hex(dname, cksum.data);
144 crypto_free_shash(tfm);
150 * If we had an error generating the recdir name for the legacy tracker
151 * then warn the admin. If the error doesn't appear to be transient,
152 * then disable recovery tracking.
155 legacy_recdir_name_error(struct nfs4_client *clp, int error)
157 printk(KERN_ERR "NFSD: unable to generate recoverydir "
158 "name (%d).\n", error);
161 * if the algorithm just doesn't exist, then disable the recovery
162 * tracker altogether. The crypto libs will generally return this if
163 * FIPS is enabled as well.
165 if (error == -ENOENT) {
166 printk(KERN_ERR "NFSD: disabling legacy clientid tracking. "
167 "Reboot recovery will not function correctly!\n");
168 nfsd4_client_tracking_exit(clp->net);
173 nfsd4_create_clid_dir(struct nfs4_client *clp)
175 const struct cred *original_cred;
176 char dname[HEXDIR_LEN];
177 struct dentry *dir, *dentry;
178 struct nfs4_client_reclaim *crp;
180 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
182 if (test_and_set_bit(NFSD4_CLIENT_STABLE, &clp->cl_flags))
187 status = nfs4_make_rec_clidname(dname, &clp->cl_name);
189 return legacy_recdir_name_error(clp, status);
191 status = nfs4_save_creds(&original_cred);
195 status = mnt_want_write_file(nn->rec_file);
199 dir = nn->rec_file->f_path.dentry;
200 /* lock the parent */
201 inode_lock(d_inode(dir));
203 dentry = lookup_one_len(dname, dir, HEXDIR_LEN-1);
204 if (IS_ERR(dentry)) {
205 status = PTR_ERR(dentry);
208 if (d_really_is_positive(dentry))
210 * In the 4.1 case, where we're called from
211 * reclaim_complete(), records from the previous reboot
212 * may still be left, so this is OK.
214 * In the 4.0 case, we should never get here; but we may
215 * as well be forgiving and just succeed silently.
218 status = vfs_mkdir(d_inode(dir), dentry, S_IRWXU);
222 inode_unlock(d_inode(dir));
225 crp = nfs4_client_to_reclaim(dname, nn);
229 vfs_fsync(nn->rec_file, 0);
231 printk(KERN_ERR "NFSD: failed to write recovery record"
232 " (err %d); please check that %s exists"
233 " and is writeable", status,
234 user_recovery_dirname);
236 mnt_drop_write_file(nn->rec_file);
238 nfs4_reset_creds(original_cred);
241 typedef int (recdir_func)(struct dentry *, struct dentry *, struct nfsd_net *);
244 char name[HEXDIR_LEN];
245 struct list_head list;
248 struct nfs4_dir_ctx {
249 struct dir_context ctx;
250 struct list_head names;
254 nfsd4_build_namelist(struct dir_context *__ctx, const char *name, int namlen,
255 loff_t offset, u64 ino, unsigned int d_type)
257 struct nfs4_dir_ctx *ctx =
258 container_of(__ctx, struct nfs4_dir_ctx, ctx);
259 struct name_list *entry;
261 if (namlen != HEXDIR_LEN - 1)
263 entry = kmalloc(sizeof(struct name_list), GFP_KERNEL);
266 memcpy(entry->name, name, HEXDIR_LEN - 1);
267 entry->name[HEXDIR_LEN - 1] = '\0';
268 list_add(&entry->list, &ctx->names);
273 nfsd4_list_rec_dir(recdir_func *f, struct nfsd_net *nn)
275 const struct cred *original_cred;
276 struct dentry *dir = nn->rec_file->f_path.dentry;
277 struct nfs4_dir_ctx ctx = {
278 .ctx.actor = nfsd4_build_namelist,
279 .names = LIST_HEAD_INIT(ctx.names)
281 struct name_list *entry, *tmp;
284 status = nfs4_save_creds(&original_cred);
288 status = vfs_llseek(nn->rec_file, 0, SEEK_SET);
290 nfs4_reset_creds(original_cred);
294 status = iterate_dir(nn->rec_file, &ctx.ctx);
295 inode_lock_nested(d_inode(dir), I_MUTEX_PARENT);
297 list_for_each_entry_safe(entry, tmp, &ctx.names, list) {
299 struct dentry *dentry;
300 dentry = lookup_one_len(entry->name, dir, HEXDIR_LEN-1);
301 if (IS_ERR(dentry)) {
302 status = PTR_ERR(dentry);
305 status = f(dir, dentry, nn);
308 list_del(&entry->list);
311 inode_unlock(d_inode(dir));
312 nfs4_reset_creds(original_cred);
314 list_for_each_entry_safe(entry, tmp, &ctx.names, list) {
315 dprintk("NFSD: %s. Left entry %s\n", __func__, entry->name);
316 list_del(&entry->list);
323 nfsd4_unlink_clid_dir(char *name, int namlen, struct nfsd_net *nn)
325 struct dentry *dir, *dentry;
328 dprintk("NFSD: nfsd4_unlink_clid_dir. name %.*s\n", namlen, name);
330 dir = nn->rec_file->f_path.dentry;
331 inode_lock_nested(d_inode(dir), I_MUTEX_PARENT);
332 dentry = lookup_one_len(name, dir, namlen);
333 if (IS_ERR(dentry)) {
334 status = PTR_ERR(dentry);
338 if (d_really_is_negative(dentry))
340 status = vfs_rmdir(d_inode(dir), dentry);
344 inode_unlock(d_inode(dir));
349 nfsd4_remove_clid_dir(struct nfs4_client *clp)
351 const struct cred *original_cred;
352 struct nfs4_client_reclaim *crp;
353 char dname[HEXDIR_LEN];
355 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
357 if (!nn->rec_file || !test_bit(NFSD4_CLIENT_STABLE, &clp->cl_flags))
360 status = nfs4_make_rec_clidname(dname, &clp->cl_name);
362 return legacy_recdir_name_error(clp, status);
364 status = mnt_want_write_file(nn->rec_file);
367 clear_bit(NFSD4_CLIENT_STABLE, &clp->cl_flags);
369 status = nfs4_save_creds(&original_cred);
373 status = nfsd4_unlink_clid_dir(dname, HEXDIR_LEN-1, nn);
374 nfs4_reset_creds(original_cred);
376 vfs_fsync(nn->rec_file, 0);
378 /* remove reclaim record */
379 crp = nfsd4_find_reclaim_client(dname, nn);
381 nfs4_remove_reclaim_record(crp, nn);
385 mnt_drop_write_file(nn->rec_file);
388 printk("NFSD: Failed to remove expired client state directory"
389 " %.*s\n", HEXDIR_LEN, dname);
393 purge_old(struct dentry *parent, struct dentry *child, struct nfsd_net *nn)
397 if (nfs4_has_reclaimed_state(child->d_name.name, nn))
400 status = vfs_rmdir(d_inode(parent), child);
402 printk("failed to remove client recovery directory %pd\n",
404 /* Keep trying, success or failure: */
409 nfsd4_recdir_purge_old(struct nfsd_net *nn)
413 nn->in_grace = false;
416 status = mnt_want_write_file(nn->rec_file);
419 status = nfsd4_list_rec_dir(purge_old, nn);
421 vfs_fsync(nn->rec_file, 0);
422 mnt_drop_write_file(nn->rec_file);
424 nfs4_release_reclaim(nn);
426 printk("nfsd4: failed to purge old clients from recovery"
427 " directory %pD\n", nn->rec_file);
431 load_recdir(struct dentry *parent, struct dentry *child, struct nfsd_net *nn)
433 if (child->d_name.len != HEXDIR_LEN - 1) {
434 printk("nfsd4: illegal name %pd in recovery directory\n",
436 /* Keep trying; maybe the others are OK: */
439 nfs4_client_to_reclaim(child->d_name.name, nn);
444 nfsd4_recdir_load(struct net *net) {
446 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
451 status = nfsd4_list_rec_dir(load_recdir, nn);
453 printk("nfsd4: failed loading clients from recovery"
454 " directory %pD\n", nn->rec_file);
459 * Hold reference to the recovery directory.
463 nfsd4_init_recdir(struct net *net)
465 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
466 const struct cred *original_cred;
469 printk("NFSD: Using %s as the NFSv4 state recovery directory\n",
470 user_recovery_dirname);
472 BUG_ON(nn->rec_file);
474 status = nfs4_save_creds(&original_cred);
476 printk("NFSD: Unable to change credentials to find recovery"
477 " directory: error %d\n",
482 nn->rec_file = filp_open(user_recovery_dirname, O_RDONLY | O_DIRECTORY, 0);
483 if (IS_ERR(nn->rec_file)) {
484 printk("NFSD: unable to find recovery directory %s\n",
485 user_recovery_dirname);
486 status = PTR_ERR(nn->rec_file);
490 nfs4_reset_creds(original_cred);
497 nfsd4_shutdown_recdir(struct net *net)
499 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
508 nfs4_legacy_state_init(struct net *net)
510 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
513 nn->reclaim_str_hashtbl = kmalloc(sizeof(struct list_head) *
514 CLIENT_HASH_SIZE, GFP_KERNEL);
515 if (!nn->reclaim_str_hashtbl)
518 for (i = 0; i < CLIENT_HASH_SIZE; i++)
519 INIT_LIST_HEAD(&nn->reclaim_str_hashtbl[i]);
520 nn->reclaim_str_hashtbl_size = 0;
526 nfs4_legacy_state_shutdown(struct net *net)
528 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
530 kfree(nn->reclaim_str_hashtbl);
534 nfsd4_load_reboot_recovery_data(struct net *net)
538 status = nfsd4_init_recdir(net);
542 status = nfsd4_recdir_load(net);
544 nfsd4_shutdown_recdir(net);
550 nfsd4_legacy_tracking_init(struct net *net)
554 /* XXX: The legacy code won't work in a container */
555 if (net != &init_net) {
556 pr_warn("NFSD: attempt to initialize legacy client tracking in a container ignored.\n");
560 status = nfs4_legacy_state_init(net);
564 status = nfsd4_load_reboot_recovery_data(net);
570 nfs4_legacy_state_shutdown(net);
575 nfsd4_legacy_tracking_exit(struct net *net)
577 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
579 nfs4_release_reclaim(nn);
580 nfsd4_shutdown_recdir(net);
581 nfs4_legacy_state_shutdown(net);
585 * Change the NFSv4 recovery directory to recdir.
588 nfs4_reset_recoverydir(char *recdir)
593 status = kern_path(recdir, LOOKUP_FOLLOW, &path);
597 if (d_is_dir(path.dentry)) {
598 strcpy(user_recovery_dirname, recdir);
606 nfs4_recoverydir(void)
608 return user_recovery_dirname;
612 nfsd4_check_legacy_client(struct nfs4_client *clp)
615 char dname[HEXDIR_LEN];
616 struct nfs4_client_reclaim *crp;
617 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
619 /* did we already find that this client is stable? */
620 if (test_bit(NFSD4_CLIENT_STABLE, &clp->cl_flags))
623 status = nfs4_make_rec_clidname(dname, &clp->cl_name);
625 legacy_recdir_name_error(clp, status);
629 /* look for it in the reclaim hashtable otherwise */
630 crp = nfsd4_find_reclaim_client(dname, nn);
632 set_bit(NFSD4_CLIENT_STABLE, &clp->cl_flags);
640 static const struct nfsd4_client_tracking_ops nfsd4_legacy_tracking_ops = {
641 .init = nfsd4_legacy_tracking_init,
642 .exit = nfsd4_legacy_tracking_exit,
643 .create = nfsd4_create_clid_dir,
644 .remove = nfsd4_remove_clid_dir,
645 .check = nfsd4_check_legacy_client,
646 .grace_done = nfsd4_recdir_purge_old,
650 #define NFSD_PIPE_DIR "nfsd"
651 #define NFSD_CLD_PIPE "cld"
653 /* per-net-ns structure for holding cld upcall info */
655 struct rpc_pipe *cn_pipe;
657 struct list_head cn_list;
662 struct list_head cu_list;
663 struct cld_net *cu_net;
664 struct completion cu_done;
665 struct cld_msg cu_msg;
669 __cld_pipe_upcall(struct rpc_pipe *pipe, struct cld_msg *cmsg)
672 struct rpc_pipe_msg msg;
673 struct cld_upcall *cup = container_of(cmsg, struct cld_upcall, cu_msg);
675 memset(&msg, 0, sizeof(msg));
677 msg.len = sizeof(*cmsg);
679 ret = rpc_queue_upcall(pipe, &msg);
684 wait_for_completion(&cup->cu_done);
693 cld_pipe_upcall(struct rpc_pipe *pipe, struct cld_msg *cmsg)
698 * -EAGAIN occurs when pipe is closed and reopened while there are
702 ret = __cld_pipe_upcall(pipe, cmsg);
703 } while (ret == -EAGAIN);
709 cld_pipe_downcall(struct file *filp, const char __user *src, size_t mlen)
711 struct cld_upcall *tmp, *cup;
712 struct cld_msg __user *cmsg = (struct cld_msg __user *)src;
714 struct nfsd_net *nn = net_generic(file_inode(filp)->i_sb->s_fs_info,
716 struct cld_net *cn = nn->cld_net;
718 if (mlen != sizeof(*cmsg)) {
719 dprintk("%s: got %zu bytes, expected %zu\n", __func__, mlen,
724 /* copy just the xid so we can try to find that */
725 if (copy_from_user(&xid, &cmsg->cm_xid, sizeof(xid)) != 0) {
726 dprintk("%s: error when copying xid from userspace", __func__);
730 /* walk the list and find corresponding xid */
732 spin_lock(&cn->cn_lock);
733 list_for_each_entry(tmp, &cn->cn_list, cu_list) {
734 if (get_unaligned(&tmp->cu_msg.cm_xid) == xid) {
736 list_del_init(&cup->cu_list);
740 spin_unlock(&cn->cn_lock);
742 /* couldn't find upcall? */
744 dprintk("%s: couldn't find upcall -- xid=%u\n", __func__, xid);
748 if (copy_from_user(&cup->cu_msg, src, mlen) != 0)
751 complete(&cup->cu_done);
756 cld_pipe_destroy_msg(struct rpc_pipe_msg *msg)
758 struct cld_msg *cmsg = msg->data;
759 struct cld_upcall *cup = container_of(cmsg, struct cld_upcall,
762 /* errno >= 0 means we got a downcall */
766 complete(&cup->cu_done);
769 static const struct rpc_pipe_ops cld_upcall_ops = {
770 .upcall = rpc_pipe_generic_upcall,
771 .downcall = cld_pipe_downcall,
772 .destroy_msg = cld_pipe_destroy_msg,
775 static struct dentry *
776 nfsd4_cld_register_sb(struct super_block *sb, struct rpc_pipe *pipe)
778 struct dentry *dir, *dentry;
780 dir = rpc_d_lookup_sb(sb, NFSD_PIPE_DIR);
782 return ERR_PTR(-ENOENT);
783 dentry = rpc_mkpipe_dentry(dir, NFSD_CLD_PIPE, NULL, pipe);
789 nfsd4_cld_unregister_sb(struct rpc_pipe *pipe)
792 rpc_unlink(pipe->dentry);
795 static struct dentry *
796 nfsd4_cld_register_net(struct net *net, struct rpc_pipe *pipe)
798 struct super_block *sb;
799 struct dentry *dentry;
801 sb = rpc_get_sb_net(net);
804 dentry = nfsd4_cld_register_sb(sb, pipe);
810 nfsd4_cld_unregister_net(struct net *net, struct rpc_pipe *pipe)
812 struct super_block *sb;
814 sb = rpc_get_sb_net(net);
816 nfsd4_cld_unregister_sb(pipe);
821 /* Initialize rpc_pipefs pipe for communication with client tracking daemon */
823 nfsd4_init_cld_pipe(struct net *net)
826 struct dentry *dentry;
827 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
833 cn = kzalloc(sizeof(*cn), GFP_KERNEL);
839 cn->cn_pipe = rpc_mkpipe_data(&cld_upcall_ops, RPC_PIPE_WAIT_FOR_OPEN);
840 if (IS_ERR(cn->cn_pipe)) {
841 ret = PTR_ERR(cn->cn_pipe);
844 spin_lock_init(&cn->cn_lock);
845 INIT_LIST_HEAD(&cn->cn_list);
847 dentry = nfsd4_cld_register_net(net, cn->cn_pipe);
848 if (IS_ERR(dentry)) {
849 ret = PTR_ERR(dentry);
850 goto err_destroy_data;
853 cn->cn_pipe->dentry = dentry;
858 rpc_destroy_pipe_data(cn->cn_pipe);
861 printk(KERN_ERR "NFSD: unable to create nfsdcld upcall pipe (%d)\n",
867 nfsd4_remove_cld_pipe(struct net *net)
869 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
870 struct cld_net *cn = nn->cld_net;
872 nfsd4_cld_unregister_net(net, cn->cn_pipe);
873 rpc_destroy_pipe_data(cn->cn_pipe);
878 static struct cld_upcall *
879 alloc_cld_upcall(struct cld_net *cn)
881 struct cld_upcall *new, *tmp;
883 new = kzalloc(sizeof(*new), GFP_KERNEL);
887 /* FIXME: hard cap on number in flight? */
889 spin_lock(&cn->cn_lock);
890 list_for_each_entry(tmp, &cn->cn_list, cu_list) {
891 if (tmp->cu_msg.cm_xid == cn->cn_xid) {
893 spin_unlock(&cn->cn_lock);
897 init_completion(&new->cu_done);
898 new->cu_msg.cm_vers = CLD_UPCALL_VERSION;
899 put_unaligned(cn->cn_xid++, &new->cu_msg.cm_xid);
901 list_add(&new->cu_list, &cn->cn_list);
902 spin_unlock(&cn->cn_lock);
904 dprintk("%s: allocated xid %u\n", __func__, new->cu_msg.cm_xid);
910 free_cld_upcall(struct cld_upcall *victim)
912 struct cld_net *cn = victim->cu_net;
914 spin_lock(&cn->cn_lock);
915 list_del(&victim->cu_list);
916 spin_unlock(&cn->cn_lock);
920 /* Ask daemon to create a new record */
922 nfsd4_cld_create(struct nfs4_client *clp)
925 struct cld_upcall *cup;
926 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
927 struct cld_net *cn = nn->cld_net;
929 /* Don't upcall if it's already stored */
930 if (test_bit(NFSD4_CLIENT_STABLE, &clp->cl_flags))
933 cup = alloc_cld_upcall(cn);
939 cup->cu_msg.cm_cmd = Cld_Create;
940 cup->cu_msg.cm_u.cm_name.cn_len = clp->cl_name.len;
941 memcpy(cup->cu_msg.cm_u.cm_name.cn_id, clp->cl_name.data,
944 ret = cld_pipe_upcall(cn->cn_pipe, &cup->cu_msg);
946 ret = cup->cu_msg.cm_status;
947 set_bit(NFSD4_CLIENT_STABLE, &clp->cl_flags);
950 free_cld_upcall(cup);
953 printk(KERN_ERR "NFSD: Unable to create client "
954 "record on stable storage: %d\n", ret);
957 /* Ask daemon to create a new record */
959 nfsd4_cld_remove(struct nfs4_client *clp)
962 struct cld_upcall *cup;
963 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
964 struct cld_net *cn = nn->cld_net;
966 /* Don't upcall if it's already removed */
967 if (!test_bit(NFSD4_CLIENT_STABLE, &clp->cl_flags))
970 cup = alloc_cld_upcall(cn);
976 cup->cu_msg.cm_cmd = Cld_Remove;
977 cup->cu_msg.cm_u.cm_name.cn_len = clp->cl_name.len;
978 memcpy(cup->cu_msg.cm_u.cm_name.cn_id, clp->cl_name.data,
981 ret = cld_pipe_upcall(cn->cn_pipe, &cup->cu_msg);
983 ret = cup->cu_msg.cm_status;
984 clear_bit(NFSD4_CLIENT_STABLE, &clp->cl_flags);
987 free_cld_upcall(cup);
990 printk(KERN_ERR "NFSD: Unable to remove client "
991 "record from stable storage: %d\n", ret);
994 /* Check for presence of a record, and update its timestamp */
996 nfsd4_cld_check(struct nfs4_client *clp)
999 struct cld_upcall *cup;
1000 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
1001 struct cld_net *cn = nn->cld_net;
1003 /* Don't upcall if one was already stored during this grace pd */
1004 if (test_bit(NFSD4_CLIENT_STABLE, &clp->cl_flags))
1007 cup = alloc_cld_upcall(cn);
1009 printk(KERN_ERR "NFSD: Unable to check client record on "
1010 "stable storage: %d\n", -ENOMEM);
1014 cup->cu_msg.cm_cmd = Cld_Check;
1015 cup->cu_msg.cm_u.cm_name.cn_len = clp->cl_name.len;
1016 memcpy(cup->cu_msg.cm_u.cm_name.cn_id, clp->cl_name.data,
1019 ret = cld_pipe_upcall(cn->cn_pipe, &cup->cu_msg);
1021 ret = cup->cu_msg.cm_status;
1022 set_bit(NFSD4_CLIENT_STABLE, &clp->cl_flags);
1025 free_cld_upcall(cup);
1030 nfsd4_cld_grace_done(struct nfsd_net *nn)
1033 struct cld_upcall *cup;
1034 struct cld_net *cn = nn->cld_net;
1036 cup = alloc_cld_upcall(cn);
1042 cup->cu_msg.cm_cmd = Cld_GraceDone;
1043 cup->cu_msg.cm_u.cm_gracetime = (int64_t)nn->boot_time;
1044 ret = cld_pipe_upcall(cn->cn_pipe, &cup->cu_msg);
1046 ret = cup->cu_msg.cm_status;
1048 free_cld_upcall(cup);
1051 printk(KERN_ERR "NFSD: Unable to end grace period: %d\n", ret);
1054 static const struct nfsd4_client_tracking_ops nfsd4_cld_tracking_ops = {
1055 .init = nfsd4_init_cld_pipe,
1056 .exit = nfsd4_remove_cld_pipe,
1057 .create = nfsd4_cld_create,
1058 .remove = nfsd4_cld_remove,
1059 .check = nfsd4_cld_check,
1060 .grace_done = nfsd4_cld_grace_done,
1063 /* upcall via usermodehelper */
1064 static char cltrack_prog[PATH_MAX] = "/sbin/nfsdcltrack";
1065 module_param_string(cltrack_prog, cltrack_prog, sizeof(cltrack_prog),
1067 MODULE_PARM_DESC(cltrack_prog, "Path to the nfsdcltrack upcall program");
1069 static bool cltrack_legacy_disable;
1070 module_param(cltrack_legacy_disable, bool, S_IRUGO|S_IWUSR);
1071 MODULE_PARM_DESC(cltrack_legacy_disable,
1072 "Disable legacy recoverydir conversion. Default: false");
1074 #define LEGACY_TOPDIR_ENV_PREFIX "NFSDCLTRACK_LEGACY_TOPDIR="
1075 #define LEGACY_RECDIR_ENV_PREFIX "NFSDCLTRACK_LEGACY_RECDIR="
1076 #define HAS_SESSION_ENV_PREFIX "NFSDCLTRACK_CLIENT_HAS_SESSION="
1077 #define GRACE_START_ENV_PREFIX "NFSDCLTRACK_GRACE_START="
1080 nfsd4_cltrack_legacy_topdir(void)
1086 if (cltrack_legacy_disable)
1089 len = strlen(LEGACY_TOPDIR_ENV_PREFIX) +
1090 strlen(nfs4_recoverydir()) + 1;
1092 result = kmalloc(len, GFP_KERNEL);
1096 copied = snprintf(result, len, LEGACY_TOPDIR_ENV_PREFIX "%s",
1097 nfs4_recoverydir());
1098 if (copied >= len) {
1099 /* just return nothing if output was truncated */
1108 nfsd4_cltrack_legacy_recdir(const struct xdr_netobj *name)
1114 if (cltrack_legacy_disable)
1117 /* +1 is for '/' between "topdir" and "recdir" */
1118 len = strlen(LEGACY_RECDIR_ENV_PREFIX) +
1119 strlen(nfs4_recoverydir()) + 1 + HEXDIR_LEN;
1121 result = kmalloc(len, GFP_KERNEL);
1125 copied = snprintf(result, len, LEGACY_RECDIR_ENV_PREFIX "%s/",
1126 nfs4_recoverydir());
1127 if (copied > (len - HEXDIR_LEN)) {
1128 /* just return nothing if output will be truncated */
1133 copied = nfs4_make_rec_clidname(result + copied, name);
1143 nfsd4_cltrack_client_has_session(struct nfs4_client *clp)
1149 /* prefix + Y/N character + terminating NULL */
1150 len = strlen(HAS_SESSION_ENV_PREFIX) + 1 + 1;
1152 result = kmalloc(len, GFP_KERNEL);
1156 copied = snprintf(result, len, HAS_SESSION_ENV_PREFIX "%c",
1157 clp->cl_minorversion ? 'Y' : 'N');
1158 if (copied >= len) {
1159 /* just return nothing if output was truncated */
1168 nfsd4_cltrack_grace_start(time_t grace_start)
1174 /* prefix + max width of int64_t string + terminating NULL */
1175 len = strlen(GRACE_START_ENV_PREFIX) + 22 + 1;
1177 result = kmalloc(len, GFP_KERNEL);
1181 copied = snprintf(result, len, GRACE_START_ENV_PREFIX "%ld",
1183 if (copied >= len) {
1184 /* just return nothing if output was truncated */
1193 nfsd4_umh_cltrack_upcall(char *cmd, char *arg, char *env0, char *env1)
1199 if (unlikely(!cltrack_prog[0])) {
1200 dprintk("%s: cltrack_prog is disabled\n", __func__);
1204 dprintk("%s: cmd: %s\n", __func__, cmd);
1205 dprintk("%s: arg: %s\n", __func__, arg ? arg : "(null)");
1206 dprintk("%s: env0: %s\n", __func__, env0 ? env0 : "(null)");
1207 dprintk("%s: env1: %s\n", __func__, env1 ? env1 : "(null)");
1213 argv[0] = (char *)cltrack_prog;
1218 ret = call_usermodehelper(argv[0], argv, envp, UMH_WAIT_PROC);
1220 * Disable the upcall mechanism if we're getting an ENOENT or EACCES
1221 * error. The admin can re-enable it on the fly by using sysfs
1222 * once the problem has been fixed.
1224 if (ret == -ENOENT || ret == -EACCES) {
1225 dprintk("NFSD: %s was not found or isn't executable (%d). "
1226 "Setting cltrack_prog to blank string!",
1228 cltrack_prog[0] = '\0';
1230 dprintk("%s: %s return value: %d\n", __func__, cltrack_prog, ret);
1236 bin_to_hex_dup(const unsigned char *src, int srclen)
1241 /* +1 for terminating NULL */
1242 buf = kmalloc((srclen * 2) + 1, GFP_KERNEL);
1247 for (i = 0; i < srclen; i++) {
1248 sprintf(hex, "%2.2x", *src++);
1255 nfsd4_umh_cltrack_init(struct net *net)
1258 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
1259 char *grace_start = nfsd4_cltrack_grace_start(nn->boot_time);
1261 /* XXX: The usermode helper s not working in container yet. */
1262 if (net != &init_net) {
1263 pr_warn("NFSD: attempt to initialize umh client tracking in a container ignored.\n");
1268 ret = nfsd4_umh_cltrack_upcall("init", NULL, grace_start, NULL);
1274 nfsd4_cltrack_upcall_lock(struct nfs4_client *clp)
1276 wait_on_bit_lock(&clp->cl_flags, NFSD4_CLIENT_UPCALL_LOCK,
1277 TASK_UNINTERRUPTIBLE);
1281 nfsd4_cltrack_upcall_unlock(struct nfs4_client *clp)
1283 smp_mb__before_atomic();
1284 clear_bit(NFSD4_CLIENT_UPCALL_LOCK, &clp->cl_flags);
1285 smp_mb__after_atomic();
1286 wake_up_bit(&clp->cl_flags, NFSD4_CLIENT_UPCALL_LOCK);
1290 nfsd4_umh_cltrack_create(struct nfs4_client *clp)
1292 char *hexid, *has_session, *grace_start;
1293 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
1296 * With v4.0 clients, there's little difference in outcome between a
1297 * create and check operation, and we can end up calling into this
1298 * function multiple times per client (once for each openowner). So,
1299 * for v4.0 clients skip upcalling once the client has been recorded
1300 * on stable storage.
1302 * For v4.1+ clients, the outcome of the two operations is different,
1303 * so we must ensure that we upcall for the create operation. v4.1+
1304 * clients call this on RECLAIM_COMPLETE though, so we should only end
1305 * up doing a single create upcall per client.
1307 if (clp->cl_minorversion == 0 &&
1308 test_bit(NFSD4_CLIENT_STABLE, &clp->cl_flags))
1311 hexid = bin_to_hex_dup(clp->cl_name.data, clp->cl_name.len);
1313 dprintk("%s: can't allocate memory for upcall!\n", __func__);
1317 has_session = nfsd4_cltrack_client_has_session(clp);
1318 grace_start = nfsd4_cltrack_grace_start(nn->boot_time);
1320 nfsd4_cltrack_upcall_lock(clp);
1321 if (!nfsd4_umh_cltrack_upcall("create", hexid, has_session, grace_start))
1322 set_bit(NFSD4_CLIENT_STABLE, &clp->cl_flags);
1323 nfsd4_cltrack_upcall_unlock(clp);
1331 nfsd4_umh_cltrack_remove(struct nfs4_client *clp)
1335 if (!test_bit(NFSD4_CLIENT_STABLE, &clp->cl_flags))
1338 hexid = bin_to_hex_dup(clp->cl_name.data, clp->cl_name.len);
1340 dprintk("%s: can't allocate memory for upcall!\n", __func__);
1344 nfsd4_cltrack_upcall_lock(clp);
1345 if (test_bit(NFSD4_CLIENT_STABLE, &clp->cl_flags) &&
1346 nfsd4_umh_cltrack_upcall("remove", hexid, NULL, NULL) == 0)
1347 clear_bit(NFSD4_CLIENT_STABLE, &clp->cl_flags);
1348 nfsd4_cltrack_upcall_unlock(clp);
1354 nfsd4_umh_cltrack_check(struct nfs4_client *clp)
1357 char *hexid, *has_session, *legacy;
1359 if (test_bit(NFSD4_CLIENT_STABLE, &clp->cl_flags))
1362 hexid = bin_to_hex_dup(clp->cl_name.data, clp->cl_name.len);
1364 dprintk("%s: can't allocate memory for upcall!\n", __func__);
1368 has_session = nfsd4_cltrack_client_has_session(clp);
1369 legacy = nfsd4_cltrack_legacy_recdir(&clp->cl_name);
1371 nfsd4_cltrack_upcall_lock(clp);
1372 if (test_bit(NFSD4_CLIENT_STABLE, &clp->cl_flags)) {
1375 ret = nfsd4_umh_cltrack_upcall("check", hexid, has_session, legacy);
1377 set_bit(NFSD4_CLIENT_STABLE, &clp->cl_flags);
1379 nfsd4_cltrack_upcall_unlock(clp);
1388 nfsd4_umh_cltrack_grace_done(struct nfsd_net *nn)
1391 char timestr[22]; /* FIXME: better way to determine max size? */
1393 sprintf(timestr, "%ld", nn->boot_time);
1394 legacy = nfsd4_cltrack_legacy_topdir();
1395 nfsd4_umh_cltrack_upcall("gracedone", timestr, legacy, NULL);
1399 static const struct nfsd4_client_tracking_ops nfsd4_umh_tracking_ops = {
1400 .init = nfsd4_umh_cltrack_init,
1402 .create = nfsd4_umh_cltrack_create,
1403 .remove = nfsd4_umh_cltrack_remove,
1404 .check = nfsd4_umh_cltrack_check,
1405 .grace_done = nfsd4_umh_cltrack_grace_done,
1409 nfsd4_client_tracking_init(struct net *net)
1413 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
1415 /* just run the init if it the method is already decided */
1416 if (nn->client_tracking_ops)
1420 * First, try a UMH upcall. It should succeed or fail quickly, so
1421 * there's little harm in trying that first.
1423 nn->client_tracking_ops = &nfsd4_umh_tracking_ops;
1424 status = nn->client_tracking_ops->init(net);
1429 * See if the recoverydir exists and is a directory. If it is,
1430 * then use the legacy ops.
1432 nn->client_tracking_ops = &nfsd4_legacy_tracking_ops;
1433 status = kern_path(nfs4_recoverydir(), LOOKUP_FOLLOW, &path);
1435 status = d_is_dir(path.dentry);
1441 /* Finally, try to use nfsdcld */
1442 nn->client_tracking_ops = &nfsd4_cld_tracking_ops;
1443 printk(KERN_WARNING "NFSD: the nfsdcld client tracking upcall will be "
1444 "removed in 3.10. Please transition to using "
1447 status = nn->client_tracking_ops->init(net);
1449 printk(KERN_WARNING "NFSD: Unable to initialize client "
1450 "recovery tracking! (%d)\n", status);
1451 nn->client_tracking_ops = NULL;
1457 nfsd4_client_tracking_exit(struct net *net)
1459 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
1461 if (nn->client_tracking_ops) {
1462 if (nn->client_tracking_ops->exit)
1463 nn->client_tracking_ops->exit(net);
1464 nn->client_tracking_ops = NULL;
1469 nfsd4_client_record_create(struct nfs4_client *clp)
1471 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
1473 if (nn->client_tracking_ops)
1474 nn->client_tracking_ops->create(clp);
1478 nfsd4_client_record_remove(struct nfs4_client *clp)
1480 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
1482 if (nn->client_tracking_ops)
1483 nn->client_tracking_ops->remove(clp);
1487 nfsd4_client_record_check(struct nfs4_client *clp)
1489 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
1491 if (nn->client_tracking_ops)
1492 return nn->client_tracking_ops->check(clp);
1498 nfsd4_record_grace_done(struct nfsd_net *nn)
1500 if (nn->client_tracking_ops)
1501 nn->client_tracking_ops->grace_done(nn);
1505 rpc_pipefs_event(struct notifier_block *nb, unsigned long event, void *ptr)
1507 struct super_block *sb = ptr;
1508 struct net *net = sb->s_fs_info;
1509 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
1510 struct cld_net *cn = nn->cld_net;
1511 struct dentry *dentry;
1514 if (!try_module_get(THIS_MODULE))
1518 module_put(THIS_MODULE);
1523 case RPC_PIPEFS_MOUNT:
1524 dentry = nfsd4_cld_register_sb(sb, cn->cn_pipe);
1525 if (IS_ERR(dentry)) {
1526 ret = PTR_ERR(dentry);
1529 cn->cn_pipe->dentry = dentry;
1531 case RPC_PIPEFS_UMOUNT:
1532 if (cn->cn_pipe->dentry)
1533 nfsd4_cld_unregister_sb(cn->cn_pipe);
1539 module_put(THIS_MODULE);
1543 static struct notifier_block nfsd4_cld_block = {
1544 .notifier_call = rpc_pipefs_event,
1548 register_cld_notifier(void)
1550 return rpc_pipefs_notifier_register(&nfsd4_cld_block);
1554 unregister_cld_notifier(void)
1556 rpc_pipefs_notifier_unregister(&nfsd4_cld_block);