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_array(CLIENT_HASH_SIZE
,
514 sizeof(struct list_head
),
516 if (!nn
->reclaim_str_hashtbl
)
519 for (i
= 0; i
< CLIENT_HASH_SIZE
; i
++)
520 INIT_LIST_HEAD(&nn
->reclaim_str_hashtbl
[i
]);
521 nn
->reclaim_str_hashtbl_size
= 0;
527 nfs4_legacy_state_shutdown(struct net
*net
)
529 struct nfsd_net
*nn
= net_generic(net
, nfsd_net_id
);
531 kfree(nn
->reclaim_str_hashtbl
);
535 nfsd4_load_reboot_recovery_data(struct net
*net
)
539 status
= nfsd4_init_recdir(net
);
543 status
= nfsd4_recdir_load(net
);
545 nfsd4_shutdown_recdir(net
);
551 nfsd4_legacy_tracking_init(struct net
*net
)
555 /* XXX: The legacy code won't work in a container */
556 if (net
!= &init_net
) {
557 pr_warn("NFSD: attempt to initialize legacy client tracking in a container ignored.\n");
561 status
= nfs4_legacy_state_init(net
);
565 status
= nfsd4_load_reboot_recovery_data(net
);
571 nfs4_legacy_state_shutdown(net
);
576 nfsd4_legacy_tracking_exit(struct net
*net
)
578 struct nfsd_net
*nn
= net_generic(net
, nfsd_net_id
);
580 nfs4_release_reclaim(nn
);
581 nfsd4_shutdown_recdir(net
);
582 nfs4_legacy_state_shutdown(net
);
586 * Change the NFSv4 recovery directory to recdir.
589 nfs4_reset_recoverydir(char *recdir
)
594 status
= kern_path(recdir
, LOOKUP_FOLLOW
, &path
);
598 if (d_is_dir(path
.dentry
)) {
599 strcpy(user_recovery_dirname
, recdir
);
607 nfs4_recoverydir(void)
609 return user_recovery_dirname
;
613 nfsd4_check_legacy_client(struct nfs4_client
*clp
)
616 char dname
[HEXDIR_LEN
];
617 struct nfs4_client_reclaim
*crp
;
618 struct nfsd_net
*nn
= net_generic(clp
->net
, nfsd_net_id
);
620 /* did we already find that this client is stable? */
621 if (test_bit(NFSD4_CLIENT_STABLE
, &clp
->cl_flags
))
624 status
= nfs4_make_rec_clidname(dname
, &clp
->cl_name
);
626 legacy_recdir_name_error(clp
, status
);
630 /* look for it in the reclaim hashtable otherwise */
631 crp
= nfsd4_find_reclaim_client(dname
, nn
);
633 set_bit(NFSD4_CLIENT_STABLE
, &clp
->cl_flags
);
641 static const struct nfsd4_client_tracking_ops nfsd4_legacy_tracking_ops
= {
642 .init
= nfsd4_legacy_tracking_init
,
643 .exit
= nfsd4_legacy_tracking_exit
,
644 .create
= nfsd4_create_clid_dir
,
645 .remove
= nfsd4_remove_clid_dir
,
646 .check
= nfsd4_check_legacy_client
,
647 .grace_done
= nfsd4_recdir_purge_old
,
651 #define NFSD_PIPE_DIR "nfsd"
652 #define NFSD_CLD_PIPE "cld"
654 /* per-net-ns structure for holding cld upcall info */
656 struct rpc_pipe
*cn_pipe
;
658 struct list_head cn_list
;
663 struct list_head cu_list
;
664 struct cld_net
*cu_net
;
665 struct completion cu_done
;
666 struct cld_msg cu_msg
;
670 __cld_pipe_upcall(struct rpc_pipe
*pipe
, struct cld_msg
*cmsg
)
673 struct rpc_pipe_msg msg
;
674 struct cld_upcall
*cup
= container_of(cmsg
, struct cld_upcall
, cu_msg
);
676 memset(&msg
, 0, sizeof(msg
));
678 msg
.len
= sizeof(*cmsg
);
680 ret
= rpc_queue_upcall(pipe
, &msg
);
685 wait_for_completion(&cup
->cu_done
);
694 cld_pipe_upcall(struct rpc_pipe
*pipe
, struct cld_msg
*cmsg
)
699 * -EAGAIN occurs when pipe is closed and reopened while there are
703 ret
= __cld_pipe_upcall(pipe
, cmsg
);
704 } while (ret
== -EAGAIN
);
710 cld_pipe_downcall(struct file
*filp
, const char __user
*src
, size_t mlen
)
712 struct cld_upcall
*tmp
, *cup
;
713 struct cld_msg __user
*cmsg
= (struct cld_msg __user
*)src
;
715 struct nfsd_net
*nn
= net_generic(file_inode(filp
)->i_sb
->s_fs_info
,
717 struct cld_net
*cn
= nn
->cld_net
;
719 if (mlen
!= sizeof(*cmsg
)) {
720 dprintk("%s: got %zu bytes, expected %zu\n", __func__
, mlen
,
725 /* copy just the xid so we can try to find that */
726 if (copy_from_user(&xid
, &cmsg
->cm_xid
, sizeof(xid
)) != 0) {
727 dprintk("%s: error when copying xid from userspace", __func__
);
731 /* walk the list and find corresponding xid */
733 spin_lock(&cn
->cn_lock
);
734 list_for_each_entry(tmp
, &cn
->cn_list
, cu_list
) {
735 if (get_unaligned(&tmp
->cu_msg
.cm_xid
) == xid
) {
737 list_del_init(&cup
->cu_list
);
741 spin_unlock(&cn
->cn_lock
);
743 /* couldn't find upcall? */
745 dprintk("%s: couldn't find upcall -- xid=%u\n", __func__
, xid
);
749 if (copy_from_user(&cup
->cu_msg
, src
, mlen
) != 0)
752 complete(&cup
->cu_done
);
757 cld_pipe_destroy_msg(struct rpc_pipe_msg
*msg
)
759 struct cld_msg
*cmsg
= msg
->data
;
760 struct cld_upcall
*cup
= container_of(cmsg
, struct cld_upcall
,
763 /* errno >= 0 means we got a downcall */
767 complete(&cup
->cu_done
);
770 static const struct rpc_pipe_ops cld_upcall_ops
= {
771 .upcall
= rpc_pipe_generic_upcall
,
772 .downcall
= cld_pipe_downcall
,
773 .destroy_msg
= cld_pipe_destroy_msg
,
776 static struct dentry
*
777 nfsd4_cld_register_sb(struct super_block
*sb
, struct rpc_pipe
*pipe
)
779 struct dentry
*dir
, *dentry
;
781 dir
= rpc_d_lookup_sb(sb
, NFSD_PIPE_DIR
);
783 return ERR_PTR(-ENOENT
);
784 dentry
= rpc_mkpipe_dentry(dir
, NFSD_CLD_PIPE
, NULL
, pipe
);
790 nfsd4_cld_unregister_sb(struct rpc_pipe
*pipe
)
793 rpc_unlink(pipe
->dentry
);
796 static struct dentry
*
797 nfsd4_cld_register_net(struct net
*net
, struct rpc_pipe
*pipe
)
799 struct super_block
*sb
;
800 struct dentry
*dentry
;
802 sb
= rpc_get_sb_net(net
);
805 dentry
= nfsd4_cld_register_sb(sb
, pipe
);
811 nfsd4_cld_unregister_net(struct net
*net
, struct rpc_pipe
*pipe
)
813 struct super_block
*sb
;
815 sb
= rpc_get_sb_net(net
);
817 nfsd4_cld_unregister_sb(pipe
);
822 /* Initialize rpc_pipefs pipe for communication with client tracking daemon */
824 nfsd4_init_cld_pipe(struct net
*net
)
827 struct dentry
*dentry
;
828 struct nfsd_net
*nn
= net_generic(net
, nfsd_net_id
);
834 cn
= kzalloc(sizeof(*cn
), GFP_KERNEL
);
840 cn
->cn_pipe
= rpc_mkpipe_data(&cld_upcall_ops
, RPC_PIPE_WAIT_FOR_OPEN
);
841 if (IS_ERR(cn
->cn_pipe
)) {
842 ret
= PTR_ERR(cn
->cn_pipe
);
845 spin_lock_init(&cn
->cn_lock
);
846 INIT_LIST_HEAD(&cn
->cn_list
);
848 dentry
= nfsd4_cld_register_net(net
, cn
->cn_pipe
);
849 if (IS_ERR(dentry
)) {
850 ret
= PTR_ERR(dentry
);
851 goto err_destroy_data
;
854 cn
->cn_pipe
->dentry
= dentry
;
859 rpc_destroy_pipe_data(cn
->cn_pipe
);
862 printk(KERN_ERR
"NFSD: unable to create nfsdcld upcall pipe (%d)\n",
868 nfsd4_remove_cld_pipe(struct net
*net
)
870 struct nfsd_net
*nn
= net_generic(net
, nfsd_net_id
);
871 struct cld_net
*cn
= nn
->cld_net
;
873 nfsd4_cld_unregister_net(net
, cn
->cn_pipe
);
874 rpc_destroy_pipe_data(cn
->cn_pipe
);
879 static struct cld_upcall
*
880 alloc_cld_upcall(struct cld_net
*cn
)
882 struct cld_upcall
*new, *tmp
;
884 new = kzalloc(sizeof(*new), GFP_KERNEL
);
888 /* FIXME: hard cap on number in flight? */
890 spin_lock(&cn
->cn_lock
);
891 list_for_each_entry(tmp
, &cn
->cn_list
, cu_list
) {
892 if (tmp
->cu_msg
.cm_xid
== cn
->cn_xid
) {
894 spin_unlock(&cn
->cn_lock
);
898 init_completion(&new->cu_done
);
899 new->cu_msg
.cm_vers
= CLD_UPCALL_VERSION
;
900 put_unaligned(cn
->cn_xid
++, &new->cu_msg
.cm_xid
);
902 list_add(&new->cu_list
, &cn
->cn_list
);
903 spin_unlock(&cn
->cn_lock
);
905 dprintk("%s: allocated xid %u\n", __func__
, new->cu_msg
.cm_xid
);
911 free_cld_upcall(struct cld_upcall
*victim
)
913 struct cld_net
*cn
= victim
->cu_net
;
915 spin_lock(&cn
->cn_lock
);
916 list_del(&victim
->cu_list
);
917 spin_unlock(&cn
->cn_lock
);
921 /* Ask daemon to create a new record */
923 nfsd4_cld_create(struct nfs4_client
*clp
)
926 struct cld_upcall
*cup
;
927 struct nfsd_net
*nn
= net_generic(clp
->net
, nfsd_net_id
);
928 struct cld_net
*cn
= nn
->cld_net
;
930 /* Don't upcall if it's already stored */
931 if (test_bit(NFSD4_CLIENT_STABLE
, &clp
->cl_flags
))
934 cup
= alloc_cld_upcall(cn
);
940 cup
->cu_msg
.cm_cmd
= Cld_Create
;
941 cup
->cu_msg
.cm_u
.cm_name
.cn_len
= clp
->cl_name
.len
;
942 memcpy(cup
->cu_msg
.cm_u
.cm_name
.cn_id
, clp
->cl_name
.data
,
945 ret
= cld_pipe_upcall(cn
->cn_pipe
, &cup
->cu_msg
);
947 ret
= cup
->cu_msg
.cm_status
;
948 set_bit(NFSD4_CLIENT_STABLE
, &clp
->cl_flags
);
951 free_cld_upcall(cup
);
954 printk(KERN_ERR
"NFSD: Unable to create client "
955 "record on stable storage: %d\n", ret
);
958 /* Ask daemon to create a new record */
960 nfsd4_cld_remove(struct nfs4_client
*clp
)
963 struct cld_upcall
*cup
;
964 struct nfsd_net
*nn
= net_generic(clp
->net
, nfsd_net_id
);
965 struct cld_net
*cn
= nn
->cld_net
;
967 /* Don't upcall if it's already removed */
968 if (!test_bit(NFSD4_CLIENT_STABLE
, &clp
->cl_flags
))
971 cup
= alloc_cld_upcall(cn
);
977 cup
->cu_msg
.cm_cmd
= Cld_Remove
;
978 cup
->cu_msg
.cm_u
.cm_name
.cn_len
= clp
->cl_name
.len
;
979 memcpy(cup
->cu_msg
.cm_u
.cm_name
.cn_id
, clp
->cl_name
.data
,
982 ret
= cld_pipe_upcall(cn
->cn_pipe
, &cup
->cu_msg
);
984 ret
= cup
->cu_msg
.cm_status
;
985 clear_bit(NFSD4_CLIENT_STABLE
, &clp
->cl_flags
);
988 free_cld_upcall(cup
);
991 printk(KERN_ERR
"NFSD: Unable to remove client "
992 "record from stable storage: %d\n", ret
);
995 /* Check for presence of a record, and update its timestamp */
997 nfsd4_cld_check(struct nfs4_client
*clp
)
1000 struct cld_upcall
*cup
;
1001 struct nfsd_net
*nn
= net_generic(clp
->net
, nfsd_net_id
);
1002 struct cld_net
*cn
= nn
->cld_net
;
1004 /* Don't upcall if one was already stored during this grace pd */
1005 if (test_bit(NFSD4_CLIENT_STABLE
, &clp
->cl_flags
))
1008 cup
= alloc_cld_upcall(cn
);
1010 printk(KERN_ERR
"NFSD: Unable to check client record on "
1011 "stable storage: %d\n", -ENOMEM
);
1015 cup
->cu_msg
.cm_cmd
= Cld_Check
;
1016 cup
->cu_msg
.cm_u
.cm_name
.cn_len
= clp
->cl_name
.len
;
1017 memcpy(cup
->cu_msg
.cm_u
.cm_name
.cn_id
, clp
->cl_name
.data
,
1020 ret
= cld_pipe_upcall(cn
->cn_pipe
, &cup
->cu_msg
);
1022 ret
= cup
->cu_msg
.cm_status
;
1023 set_bit(NFSD4_CLIENT_STABLE
, &clp
->cl_flags
);
1026 free_cld_upcall(cup
);
1031 nfsd4_cld_grace_done(struct nfsd_net
*nn
)
1034 struct cld_upcall
*cup
;
1035 struct cld_net
*cn
= nn
->cld_net
;
1037 cup
= alloc_cld_upcall(cn
);
1043 cup
->cu_msg
.cm_cmd
= Cld_GraceDone
;
1044 cup
->cu_msg
.cm_u
.cm_gracetime
= (int64_t)nn
->boot_time
;
1045 ret
= cld_pipe_upcall(cn
->cn_pipe
, &cup
->cu_msg
);
1047 ret
= cup
->cu_msg
.cm_status
;
1049 free_cld_upcall(cup
);
1052 printk(KERN_ERR
"NFSD: Unable to end grace period: %d\n", ret
);
1055 static const struct nfsd4_client_tracking_ops nfsd4_cld_tracking_ops
= {
1056 .init
= nfsd4_init_cld_pipe
,
1057 .exit
= nfsd4_remove_cld_pipe
,
1058 .create
= nfsd4_cld_create
,
1059 .remove
= nfsd4_cld_remove
,
1060 .check
= nfsd4_cld_check
,
1061 .grace_done
= nfsd4_cld_grace_done
,
1064 /* upcall via usermodehelper */
1065 static char cltrack_prog
[PATH_MAX
] = "/sbin/nfsdcltrack";
1066 module_param_string(cltrack_prog
, cltrack_prog
, sizeof(cltrack_prog
),
1068 MODULE_PARM_DESC(cltrack_prog
, "Path to the nfsdcltrack upcall program");
1070 static bool cltrack_legacy_disable
;
1071 module_param(cltrack_legacy_disable
, bool, S_IRUGO
|S_IWUSR
);
1072 MODULE_PARM_DESC(cltrack_legacy_disable
,
1073 "Disable legacy recoverydir conversion. Default: false");
1075 #define LEGACY_TOPDIR_ENV_PREFIX "NFSDCLTRACK_LEGACY_TOPDIR="
1076 #define LEGACY_RECDIR_ENV_PREFIX "NFSDCLTRACK_LEGACY_RECDIR="
1077 #define HAS_SESSION_ENV_PREFIX "NFSDCLTRACK_CLIENT_HAS_SESSION="
1078 #define GRACE_START_ENV_PREFIX "NFSDCLTRACK_GRACE_START="
1081 nfsd4_cltrack_legacy_topdir(void)
1087 if (cltrack_legacy_disable
)
1090 len
= strlen(LEGACY_TOPDIR_ENV_PREFIX
) +
1091 strlen(nfs4_recoverydir()) + 1;
1093 result
= kmalloc(len
, GFP_KERNEL
);
1097 copied
= snprintf(result
, len
, LEGACY_TOPDIR_ENV_PREFIX
"%s",
1098 nfs4_recoverydir());
1099 if (copied
>= len
) {
1100 /* just return nothing if output was truncated */
1109 nfsd4_cltrack_legacy_recdir(const struct xdr_netobj
*name
)
1115 if (cltrack_legacy_disable
)
1118 /* +1 is for '/' between "topdir" and "recdir" */
1119 len
= strlen(LEGACY_RECDIR_ENV_PREFIX
) +
1120 strlen(nfs4_recoverydir()) + 1 + HEXDIR_LEN
;
1122 result
= kmalloc(len
, GFP_KERNEL
);
1126 copied
= snprintf(result
, len
, LEGACY_RECDIR_ENV_PREFIX
"%s/",
1127 nfs4_recoverydir());
1128 if (copied
> (len
- HEXDIR_LEN
)) {
1129 /* just return nothing if output will be truncated */
1134 copied
= nfs4_make_rec_clidname(result
+ copied
, name
);
1144 nfsd4_cltrack_client_has_session(struct nfs4_client
*clp
)
1150 /* prefix + Y/N character + terminating NULL */
1151 len
= strlen(HAS_SESSION_ENV_PREFIX
) + 1 + 1;
1153 result
= kmalloc(len
, GFP_KERNEL
);
1157 copied
= snprintf(result
, len
, HAS_SESSION_ENV_PREFIX
"%c",
1158 clp
->cl_minorversion
? 'Y' : 'N');
1159 if (copied
>= len
) {
1160 /* just return nothing if output was truncated */
1169 nfsd4_cltrack_grace_start(time_t grace_start
)
1175 /* prefix + max width of int64_t string + terminating NULL */
1176 len
= strlen(GRACE_START_ENV_PREFIX
) + 22 + 1;
1178 result
= kmalloc(len
, GFP_KERNEL
);
1182 copied
= snprintf(result
, len
, GRACE_START_ENV_PREFIX
"%ld",
1184 if (copied
>= len
) {
1185 /* just return nothing if output was truncated */
1194 nfsd4_umh_cltrack_upcall(char *cmd
, char *arg
, char *env0
, char *env1
)
1200 if (unlikely(!cltrack_prog
[0])) {
1201 dprintk("%s: cltrack_prog is disabled\n", __func__
);
1205 dprintk("%s: cmd: %s\n", __func__
, cmd
);
1206 dprintk("%s: arg: %s\n", __func__
, arg
? arg
: "(null)");
1207 dprintk("%s: env0: %s\n", __func__
, env0
? env0
: "(null)");
1208 dprintk("%s: env1: %s\n", __func__
, env1
? env1
: "(null)");
1214 argv
[0] = (char *)cltrack_prog
;
1219 ret
= call_usermodehelper(argv
[0], argv
, envp
, UMH_WAIT_PROC
);
1221 * Disable the upcall mechanism if we're getting an ENOENT or EACCES
1222 * error. The admin can re-enable it on the fly by using sysfs
1223 * once the problem has been fixed.
1225 if (ret
== -ENOENT
|| ret
== -EACCES
) {
1226 dprintk("NFSD: %s was not found or isn't executable (%d). "
1227 "Setting cltrack_prog to blank string!",
1229 cltrack_prog
[0] = '\0';
1231 dprintk("%s: %s return value: %d\n", __func__
, cltrack_prog
, ret
);
1237 bin_to_hex_dup(const unsigned char *src
, int srclen
)
1242 /* +1 for terminating NULL */
1243 buf
= kmalloc((srclen
* 2) + 1, GFP_KERNEL
);
1248 for (i
= 0; i
< srclen
; i
++) {
1249 sprintf(hex
, "%2.2x", *src
++);
1256 nfsd4_umh_cltrack_init(struct net
*net
)
1259 struct nfsd_net
*nn
= net_generic(net
, nfsd_net_id
);
1260 char *grace_start
= nfsd4_cltrack_grace_start(nn
->boot_time
);
1262 /* XXX: The usermode helper s not working in container yet. */
1263 if (net
!= &init_net
) {
1264 pr_warn("NFSD: attempt to initialize umh client tracking in a container ignored.\n");
1269 ret
= nfsd4_umh_cltrack_upcall("init", NULL
, grace_start
, NULL
);
1275 nfsd4_cltrack_upcall_lock(struct nfs4_client
*clp
)
1277 wait_on_bit_lock(&clp
->cl_flags
, NFSD4_CLIENT_UPCALL_LOCK
,
1278 TASK_UNINTERRUPTIBLE
);
1282 nfsd4_cltrack_upcall_unlock(struct nfs4_client
*clp
)
1284 smp_mb__before_atomic();
1285 clear_bit(NFSD4_CLIENT_UPCALL_LOCK
, &clp
->cl_flags
);
1286 smp_mb__after_atomic();
1287 wake_up_bit(&clp
->cl_flags
, NFSD4_CLIENT_UPCALL_LOCK
);
1291 nfsd4_umh_cltrack_create(struct nfs4_client
*clp
)
1293 char *hexid
, *has_session
, *grace_start
;
1294 struct nfsd_net
*nn
= net_generic(clp
->net
, nfsd_net_id
);
1297 * With v4.0 clients, there's little difference in outcome between a
1298 * create and check operation, and we can end up calling into this
1299 * function multiple times per client (once for each openowner). So,
1300 * for v4.0 clients skip upcalling once the client has been recorded
1301 * on stable storage.
1303 * For v4.1+ clients, the outcome of the two operations is different,
1304 * so we must ensure that we upcall for the create operation. v4.1+
1305 * clients call this on RECLAIM_COMPLETE though, so we should only end
1306 * up doing a single create upcall per client.
1308 if (clp
->cl_minorversion
== 0 &&
1309 test_bit(NFSD4_CLIENT_STABLE
, &clp
->cl_flags
))
1312 hexid
= bin_to_hex_dup(clp
->cl_name
.data
, clp
->cl_name
.len
);
1314 dprintk("%s: can't allocate memory for upcall!\n", __func__
);
1318 has_session
= nfsd4_cltrack_client_has_session(clp
);
1319 grace_start
= nfsd4_cltrack_grace_start(nn
->boot_time
);
1321 nfsd4_cltrack_upcall_lock(clp
);
1322 if (!nfsd4_umh_cltrack_upcall("create", hexid
, has_session
, grace_start
))
1323 set_bit(NFSD4_CLIENT_STABLE
, &clp
->cl_flags
);
1324 nfsd4_cltrack_upcall_unlock(clp
);
1332 nfsd4_umh_cltrack_remove(struct nfs4_client
*clp
)
1336 if (!test_bit(NFSD4_CLIENT_STABLE
, &clp
->cl_flags
))
1339 hexid
= bin_to_hex_dup(clp
->cl_name
.data
, clp
->cl_name
.len
);
1341 dprintk("%s: can't allocate memory for upcall!\n", __func__
);
1345 nfsd4_cltrack_upcall_lock(clp
);
1346 if (test_bit(NFSD4_CLIENT_STABLE
, &clp
->cl_flags
) &&
1347 nfsd4_umh_cltrack_upcall("remove", hexid
, NULL
, NULL
) == 0)
1348 clear_bit(NFSD4_CLIENT_STABLE
, &clp
->cl_flags
);
1349 nfsd4_cltrack_upcall_unlock(clp
);
1355 nfsd4_umh_cltrack_check(struct nfs4_client
*clp
)
1358 char *hexid
, *has_session
, *legacy
;
1360 if (test_bit(NFSD4_CLIENT_STABLE
, &clp
->cl_flags
))
1363 hexid
= bin_to_hex_dup(clp
->cl_name
.data
, clp
->cl_name
.len
);
1365 dprintk("%s: can't allocate memory for upcall!\n", __func__
);
1369 has_session
= nfsd4_cltrack_client_has_session(clp
);
1370 legacy
= nfsd4_cltrack_legacy_recdir(&clp
->cl_name
);
1372 nfsd4_cltrack_upcall_lock(clp
);
1373 if (test_bit(NFSD4_CLIENT_STABLE
, &clp
->cl_flags
)) {
1376 ret
= nfsd4_umh_cltrack_upcall("check", hexid
, has_session
, legacy
);
1378 set_bit(NFSD4_CLIENT_STABLE
, &clp
->cl_flags
);
1380 nfsd4_cltrack_upcall_unlock(clp
);
1389 nfsd4_umh_cltrack_grace_done(struct nfsd_net
*nn
)
1392 char timestr
[22]; /* FIXME: better way to determine max size? */
1394 sprintf(timestr
, "%ld", nn
->boot_time
);
1395 legacy
= nfsd4_cltrack_legacy_topdir();
1396 nfsd4_umh_cltrack_upcall("gracedone", timestr
, legacy
, NULL
);
1400 static const struct nfsd4_client_tracking_ops nfsd4_umh_tracking_ops
= {
1401 .init
= nfsd4_umh_cltrack_init
,
1403 .create
= nfsd4_umh_cltrack_create
,
1404 .remove
= nfsd4_umh_cltrack_remove
,
1405 .check
= nfsd4_umh_cltrack_check
,
1406 .grace_done
= nfsd4_umh_cltrack_grace_done
,
1410 nfsd4_client_tracking_init(struct net
*net
)
1414 struct nfsd_net
*nn
= net_generic(net
, nfsd_net_id
);
1416 /* just run the init if it the method is already decided */
1417 if (nn
->client_tracking_ops
)
1421 * First, try a UMH upcall. It should succeed or fail quickly, so
1422 * there's little harm in trying that first.
1424 nn
->client_tracking_ops
= &nfsd4_umh_tracking_ops
;
1425 status
= nn
->client_tracking_ops
->init(net
);
1430 * See if the recoverydir exists and is a directory. If it is,
1431 * then use the legacy ops.
1433 nn
->client_tracking_ops
= &nfsd4_legacy_tracking_ops
;
1434 status
= kern_path(nfs4_recoverydir(), LOOKUP_FOLLOW
, &path
);
1436 status
= d_is_dir(path
.dentry
);
1442 /* Finally, try to use nfsdcld */
1443 nn
->client_tracking_ops
= &nfsd4_cld_tracking_ops
;
1444 printk(KERN_WARNING
"NFSD: the nfsdcld client tracking upcall will be "
1445 "removed in 3.10. Please transition to using "
1448 status
= nn
->client_tracking_ops
->init(net
);
1450 printk(KERN_WARNING
"NFSD: Unable to initialize client "
1451 "recovery tracking! (%d)\n", status
);
1452 nn
->client_tracking_ops
= NULL
;
1458 nfsd4_client_tracking_exit(struct net
*net
)
1460 struct nfsd_net
*nn
= net_generic(net
, nfsd_net_id
);
1462 if (nn
->client_tracking_ops
) {
1463 if (nn
->client_tracking_ops
->exit
)
1464 nn
->client_tracking_ops
->exit(net
);
1465 nn
->client_tracking_ops
= NULL
;
1470 nfsd4_client_record_create(struct nfs4_client
*clp
)
1472 struct nfsd_net
*nn
= net_generic(clp
->net
, nfsd_net_id
);
1474 if (nn
->client_tracking_ops
)
1475 nn
->client_tracking_ops
->create(clp
);
1479 nfsd4_client_record_remove(struct nfs4_client
*clp
)
1481 struct nfsd_net
*nn
= net_generic(clp
->net
, nfsd_net_id
);
1483 if (nn
->client_tracking_ops
)
1484 nn
->client_tracking_ops
->remove(clp
);
1488 nfsd4_client_record_check(struct nfs4_client
*clp
)
1490 struct nfsd_net
*nn
= net_generic(clp
->net
, nfsd_net_id
);
1492 if (nn
->client_tracking_ops
)
1493 return nn
->client_tracking_ops
->check(clp
);
1499 nfsd4_record_grace_done(struct nfsd_net
*nn
)
1501 if (nn
->client_tracking_ops
)
1502 nn
->client_tracking_ops
->grace_done(nn
);
1506 rpc_pipefs_event(struct notifier_block
*nb
, unsigned long event
, void *ptr
)
1508 struct super_block
*sb
= ptr
;
1509 struct net
*net
= sb
->s_fs_info
;
1510 struct nfsd_net
*nn
= net_generic(net
, nfsd_net_id
);
1511 struct cld_net
*cn
= nn
->cld_net
;
1512 struct dentry
*dentry
;
1515 if (!try_module_get(THIS_MODULE
))
1519 module_put(THIS_MODULE
);
1524 case RPC_PIPEFS_MOUNT
:
1525 dentry
= nfsd4_cld_register_sb(sb
, cn
->cn_pipe
);
1526 if (IS_ERR(dentry
)) {
1527 ret
= PTR_ERR(dentry
);
1530 cn
->cn_pipe
->dentry
= dentry
;
1532 case RPC_PIPEFS_UMOUNT
:
1533 if (cn
->cn_pipe
->dentry
)
1534 nfsd4_cld_unregister_sb(cn
->cn_pipe
);
1540 module_put(THIS_MODULE
);
1544 static struct notifier_block nfsd4_cld_block
= {
1545 .notifier_call
= rpc_pipefs_event
,
1549 register_cld_notifier(void)
1551 return rpc_pipefs_notifier_register(&nfsd4_cld_block
);
1555 unregister_cld_notifier(void)
1557 rpc_pipefs_notifier_unregister(&nfsd4_cld_block
);