Merge tag 'regmap-fix-v5.11-rc2' of git://git.kernel.org/pub/scm/linux/kernel/git...
[linux/fpc-iii.git] / fs / nfsd / nfs4recover.c
blob186fa2c2c6ba6a14d26f9d31bf6002793484f783
1 /*
2 * Copyright (c) 2004 The Regents of the University of Michigan.
3 * Copyright (c) 2012 Jeff Layton <jlayton@redhat.com>
4 * All rights reserved.
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
10 * are met:
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>
40 #include <linux/fs.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>
47 #include "nfsd.h"
48 #include "state.h"
49 #include "vfs.h"
50 #include "netns.h"
52 #define NFSDDBG_FACILITY NFSDDBG_PROC
54 /* Declarations */
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 *);
62 uint8_t version;
63 size_t msglen;
66 static const struct nfsd4_client_tracking_ops nfsd4_cld_tracking_ops;
67 static const struct nfsd4_client_tracking_ops nfsd4_cld_tracking_ops_v2;
69 /* Globals */
70 static char user_recovery_dirname[PATH_MAX] = "/var/lib/nfs/v4recovery";
72 static int
73 nfs4_save_creds(const struct cred **original_creds)
75 struct cred *new;
77 new = prepare_creds();
78 if (!new)
79 return -ENOMEM;
81 new->fsuid = GLOBAL_ROOT_UID;
82 new->fsgid = GLOBAL_ROOT_GID;
83 *original_creds = override_creds(new);
84 put_cred(new);
85 return 0;
88 static void
89 nfs4_reset_creds(const struct cred *original)
91 revert_creds(original);
94 static void
95 md5_to_hex(char *out, char *md5)
97 int i;
99 for (i=0; i<16; i++) {
100 unsigned char c = md5[i];
102 *out++ = '0' + ((c&0xf0)>>4) + (c>=0xa0)*('a'-'9'-1);
103 *out++ = '0' + (c&0x0f) + ((c&0x0f)>=0x0a)*('a'-'9'-1);
105 *out = '\0';
108 static int
109 nfs4_make_rec_clidname(char *dname, const struct xdr_netobj *clname)
111 struct xdr_netobj cksum;
112 struct crypto_shash *tfm;
113 int status;
115 dprintk("NFSD: nfs4_make_rec_clidname for %.*s\n",
116 clname->len, clname->data);
117 tfm = crypto_alloc_shash("md5", 0, 0);
118 if (IS_ERR(tfm)) {
119 status = PTR_ERR(tfm);
120 goto out_no_tfm;
123 cksum.len = crypto_shash_digestsize(tfm);
124 cksum.data = kmalloc(cksum.len, GFP_KERNEL);
125 if (cksum.data == NULL) {
126 status = -ENOMEM;
127 goto out;
130 status = crypto_shash_tfm_digest(tfm, clname->data, clname->len,
131 cksum.data);
132 if (status)
133 goto out;
135 md5_to_hex(dname, cksum.data);
137 status = 0;
138 out:
139 kfree(cksum.data);
140 crypto_free_shash(tfm);
141 out_no_tfm:
142 return status;
146 * If we had an error generating the recdir name for the legacy tracker
147 * then warn the admin. If the error doesn't appear to be transient,
148 * then disable recovery tracking.
150 static void
151 legacy_recdir_name_error(struct nfs4_client *clp, int error)
153 printk(KERN_ERR "NFSD: unable to generate recoverydir "
154 "name (%d).\n", error);
157 * if the algorithm just doesn't exist, then disable the recovery
158 * tracker altogether. The crypto libs will generally return this if
159 * FIPS is enabled as well.
161 if (error == -ENOENT) {
162 printk(KERN_ERR "NFSD: disabling legacy clientid tracking. "
163 "Reboot recovery will not function correctly!\n");
164 nfsd4_client_tracking_exit(clp->net);
168 static void
169 __nfsd4_create_reclaim_record_grace(struct nfs4_client *clp,
170 const char *dname, int len, struct nfsd_net *nn)
172 struct xdr_netobj name;
173 struct xdr_netobj princhash = { .len = 0, .data = NULL };
174 struct nfs4_client_reclaim *crp;
176 name.data = kmemdup(dname, len, GFP_KERNEL);
177 if (!name.data) {
178 dprintk("%s: failed to allocate memory for name.data!\n",
179 __func__);
180 return;
182 name.len = len;
183 crp = nfs4_client_to_reclaim(name, princhash, nn);
184 if (!crp) {
185 kfree(name.data);
186 return;
188 crp->cr_clp = clp;
191 static void
192 nfsd4_create_clid_dir(struct nfs4_client *clp)
194 const struct cred *original_cred;
195 char dname[HEXDIR_LEN];
196 struct dentry *dir, *dentry;
197 int status;
198 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
200 if (test_and_set_bit(NFSD4_CLIENT_STABLE, &clp->cl_flags))
201 return;
202 if (!nn->rec_file)
203 return;
205 status = nfs4_make_rec_clidname(dname, &clp->cl_name);
206 if (status)
207 return legacy_recdir_name_error(clp, status);
209 status = nfs4_save_creds(&original_cred);
210 if (status < 0)
211 return;
213 status = mnt_want_write_file(nn->rec_file);
214 if (status)
215 goto out_creds;
217 dir = nn->rec_file->f_path.dentry;
218 /* lock the parent */
219 inode_lock(d_inode(dir));
221 dentry = lookup_one_len(dname, dir, HEXDIR_LEN-1);
222 if (IS_ERR(dentry)) {
223 status = PTR_ERR(dentry);
224 goto out_unlock;
226 if (d_really_is_positive(dentry))
228 * In the 4.1 case, where we're called from
229 * reclaim_complete(), records from the previous reboot
230 * may still be left, so this is OK.
232 * In the 4.0 case, we should never get here; but we may
233 * as well be forgiving and just succeed silently.
235 goto out_put;
236 status = vfs_mkdir(d_inode(dir), dentry, S_IRWXU);
237 out_put:
238 dput(dentry);
239 out_unlock:
240 inode_unlock(d_inode(dir));
241 if (status == 0) {
242 if (nn->in_grace)
243 __nfsd4_create_reclaim_record_grace(clp, dname,
244 HEXDIR_LEN, nn);
245 vfs_fsync(nn->rec_file, 0);
246 } else {
247 printk(KERN_ERR "NFSD: failed to write recovery record"
248 " (err %d); please check that %s exists"
249 " and is writeable", status,
250 user_recovery_dirname);
252 mnt_drop_write_file(nn->rec_file);
253 out_creds:
254 nfs4_reset_creds(original_cred);
257 typedef int (recdir_func)(struct dentry *, struct dentry *, struct nfsd_net *);
259 struct name_list {
260 char name[HEXDIR_LEN];
261 struct list_head list;
264 struct nfs4_dir_ctx {
265 struct dir_context ctx;
266 struct list_head names;
269 static int
270 nfsd4_build_namelist(struct dir_context *__ctx, const char *name, int namlen,
271 loff_t offset, u64 ino, unsigned int d_type)
273 struct nfs4_dir_ctx *ctx =
274 container_of(__ctx, struct nfs4_dir_ctx, ctx);
275 struct name_list *entry;
277 if (namlen != HEXDIR_LEN - 1)
278 return 0;
279 entry = kmalloc(sizeof(struct name_list), GFP_KERNEL);
280 if (entry == NULL)
281 return -ENOMEM;
282 memcpy(entry->name, name, HEXDIR_LEN - 1);
283 entry->name[HEXDIR_LEN - 1] = '\0';
284 list_add(&entry->list, &ctx->names);
285 return 0;
288 static int
289 nfsd4_list_rec_dir(recdir_func *f, struct nfsd_net *nn)
291 const struct cred *original_cred;
292 struct dentry *dir = nn->rec_file->f_path.dentry;
293 struct nfs4_dir_ctx ctx = {
294 .ctx.actor = nfsd4_build_namelist,
295 .names = LIST_HEAD_INIT(ctx.names)
297 struct name_list *entry, *tmp;
298 int status;
300 status = nfs4_save_creds(&original_cred);
301 if (status < 0)
302 return status;
304 status = vfs_llseek(nn->rec_file, 0, SEEK_SET);
305 if (status < 0) {
306 nfs4_reset_creds(original_cred);
307 return status;
310 status = iterate_dir(nn->rec_file, &ctx.ctx);
311 inode_lock_nested(d_inode(dir), I_MUTEX_PARENT);
313 list_for_each_entry_safe(entry, tmp, &ctx.names, list) {
314 if (!status) {
315 struct dentry *dentry;
316 dentry = lookup_one_len(entry->name, dir, HEXDIR_LEN-1);
317 if (IS_ERR(dentry)) {
318 status = PTR_ERR(dentry);
319 break;
321 status = f(dir, dentry, nn);
322 dput(dentry);
324 list_del(&entry->list);
325 kfree(entry);
327 inode_unlock(d_inode(dir));
328 nfs4_reset_creds(original_cred);
330 list_for_each_entry_safe(entry, tmp, &ctx.names, list) {
331 dprintk("NFSD: %s. Left entry %s\n", __func__, entry->name);
332 list_del(&entry->list);
333 kfree(entry);
335 return status;
338 static int
339 nfsd4_unlink_clid_dir(char *name, int namlen, struct nfsd_net *nn)
341 struct dentry *dir, *dentry;
342 int status;
344 dprintk("NFSD: nfsd4_unlink_clid_dir. name %.*s\n", namlen, name);
346 dir = nn->rec_file->f_path.dentry;
347 inode_lock_nested(d_inode(dir), I_MUTEX_PARENT);
348 dentry = lookup_one_len(name, dir, namlen);
349 if (IS_ERR(dentry)) {
350 status = PTR_ERR(dentry);
351 goto out_unlock;
353 status = -ENOENT;
354 if (d_really_is_negative(dentry))
355 goto out;
356 status = vfs_rmdir(d_inode(dir), dentry);
357 out:
358 dput(dentry);
359 out_unlock:
360 inode_unlock(d_inode(dir));
361 return status;
364 static void
365 __nfsd4_remove_reclaim_record_grace(const char *dname, int len,
366 struct nfsd_net *nn)
368 struct xdr_netobj name;
369 struct nfs4_client_reclaim *crp;
371 name.data = kmemdup(dname, len, GFP_KERNEL);
372 if (!name.data) {
373 dprintk("%s: failed to allocate memory for name.data!\n",
374 __func__);
375 return;
377 name.len = len;
378 crp = nfsd4_find_reclaim_client(name, nn);
379 kfree(name.data);
380 if (crp)
381 nfs4_remove_reclaim_record(crp, nn);
384 static void
385 nfsd4_remove_clid_dir(struct nfs4_client *clp)
387 const struct cred *original_cred;
388 char dname[HEXDIR_LEN];
389 int status;
390 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
392 if (!nn->rec_file || !test_bit(NFSD4_CLIENT_STABLE, &clp->cl_flags))
393 return;
395 status = nfs4_make_rec_clidname(dname, &clp->cl_name);
396 if (status)
397 return legacy_recdir_name_error(clp, status);
399 status = mnt_want_write_file(nn->rec_file);
400 if (status)
401 goto out;
402 clear_bit(NFSD4_CLIENT_STABLE, &clp->cl_flags);
404 status = nfs4_save_creds(&original_cred);
405 if (status < 0)
406 goto out_drop_write;
408 status = nfsd4_unlink_clid_dir(dname, HEXDIR_LEN-1, nn);
409 nfs4_reset_creds(original_cred);
410 if (status == 0) {
411 vfs_fsync(nn->rec_file, 0);
412 if (nn->in_grace)
413 __nfsd4_remove_reclaim_record_grace(dname,
414 HEXDIR_LEN, nn);
416 out_drop_write:
417 mnt_drop_write_file(nn->rec_file);
418 out:
419 if (status)
420 printk("NFSD: Failed to remove expired client state directory"
421 " %.*s\n", HEXDIR_LEN, dname);
424 static int
425 purge_old(struct dentry *parent, struct dentry *child, struct nfsd_net *nn)
427 int status;
428 struct xdr_netobj name;
430 if (child->d_name.len != HEXDIR_LEN - 1) {
431 printk("%s: illegal name %pd in recovery directory\n",
432 __func__, child);
433 /* Keep trying; maybe the others are OK: */
434 return 0;
436 name.data = kmemdup_nul(child->d_name.name, child->d_name.len, GFP_KERNEL);
437 if (!name.data) {
438 dprintk("%s: failed to allocate memory for name.data!\n",
439 __func__);
440 goto out;
442 name.len = HEXDIR_LEN;
443 if (nfs4_has_reclaimed_state(name, nn))
444 goto out_free;
446 status = vfs_rmdir(d_inode(parent), child);
447 if (status)
448 printk("failed to remove client recovery directory %pd\n",
449 child);
450 out_free:
451 kfree(name.data);
452 out:
453 /* Keep trying, success or failure: */
454 return 0;
457 static void
458 nfsd4_recdir_purge_old(struct nfsd_net *nn)
460 int status;
462 nn->in_grace = false;
463 if (!nn->rec_file)
464 return;
465 status = mnt_want_write_file(nn->rec_file);
466 if (status)
467 goto out;
468 status = nfsd4_list_rec_dir(purge_old, nn);
469 if (status == 0)
470 vfs_fsync(nn->rec_file, 0);
471 mnt_drop_write_file(nn->rec_file);
472 out:
473 nfs4_release_reclaim(nn);
474 if (status)
475 printk("nfsd4: failed to purge old clients from recovery"
476 " directory %pD\n", nn->rec_file);
479 static int
480 load_recdir(struct dentry *parent, struct dentry *child, struct nfsd_net *nn)
482 struct xdr_netobj name;
483 struct xdr_netobj princhash = { .len = 0, .data = NULL };
485 if (child->d_name.len != HEXDIR_LEN - 1) {
486 printk("%s: illegal name %pd in recovery directory\n",
487 __func__, child);
488 /* Keep trying; maybe the others are OK: */
489 return 0;
491 name.data = kmemdup_nul(child->d_name.name, child->d_name.len, GFP_KERNEL);
492 if (!name.data) {
493 dprintk("%s: failed to allocate memory for name.data!\n",
494 __func__);
495 goto out;
497 name.len = HEXDIR_LEN;
498 if (!nfs4_client_to_reclaim(name, princhash, nn))
499 kfree(name.data);
500 out:
501 return 0;
504 static int
505 nfsd4_recdir_load(struct net *net) {
506 int status;
507 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
509 if (!nn->rec_file)
510 return 0;
512 status = nfsd4_list_rec_dir(load_recdir, nn);
513 if (status)
514 printk("nfsd4: failed loading clients from recovery"
515 " directory %pD\n", nn->rec_file);
516 return status;
520 * Hold reference to the recovery directory.
523 static int
524 nfsd4_init_recdir(struct net *net)
526 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
527 const struct cred *original_cred;
528 int status;
530 printk("NFSD: Using %s as the NFSv4 state recovery directory\n",
531 user_recovery_dirname);
533 BUG_ON(nn->rec_file);
535 status = nfs4_save_creds(&original_cred);
536 if (status < 0) {
537 printk("NFSD: Unable to change credentials to find recovery"
538 " directory: error %d\n",
539 status);
540 return status;
543 nn->rec_file = filp_open(user_recovery_dirname, O_RDONLY | O_DIRECTORY, 0);
544 if (IS_ERR(nn->rec_file)) {
545 printk("NFSD: unable to find recovery directory %s\n",
546 user_recovery_dirname);
547 status = PTR_ERR(nn->rec_file);
548 nn->rec_file = NULL;
551 nfs4_reset_creds(original_cred);
552 if (!status)
553 nn->in_grace = true;
554 return status;
557 static void
558 nfsd4_shutdown_recdir(struct net *net)
560 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
562 if (!nn->rec_file)
563 return;
564 fput(nn->rec_file);
565 nn->rec_file = NULL;
568 static int
569 nfs4_legacy_state_init(struct net *net)
571 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
572 int i;
574 nn->reclaim_str_hashtbl = kmalloc_array(CLIENT_HASH_SIZE,
575 sizeof(struct list_head),
576 GFP_KERNEL);
577 if (!nn->reclaim_str_hashtbl)
578 return -ENOMEM;
580 for (i = 0; i < CLIENT_HASH_SIZE; i++)
581 INIT_LIST_HEAD(&nn->reclaim_str_hashtbl[i]);
582 nn->reclaim_str_hashtbl_size = 0;
584 return 0;
587 static void
588 nfs4_legacy_state_shutdown(struct net *net)
590 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
592 kfree(nn->reclaim_str_hashtbl);
595 static int
596 nfsd4_load_reboot_recovery_data(struct net *net)
598 int status;
600 status = nfsd4_init_recdir(net);
601 if (status)
602 return status;
604 status = nfsd4_recdir_load(net);
605 if (status)
606 nfsd4_shutdown_recdir(net);
608 return status;
611 static int
612 nfsd4_legacy_tracking_init(struct net *net)
614 int status;
616 /* XXX: The legacy code won't work in a container */
617 if (net != &init_net) {
618 pr_warn("NFSD: attempt to initialize legacy client tracking in a container ignored.\n");
619 return -EINVAL;
622 status = nfs4_legacy_state_init(net);
623 if (status)
624 return status;
626 status = nfsd4_load_reboot_recovery_data(net);
627 if (status)
628 goto err;
629 printk("NFSD: Using legacy client tracking operations.\n");
630 return 0;
632 err:
633 nfs4_legacy_state_shutdown(net);
634 return status;
637 static void
638 nfsd4_legacy_tracking_exit(struct net *net)
640 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
642 nfs4_release_reclaim(nn);
643 nfsd4_shutdown_recdir(net);
644 nfs4_legacy_state_shutdown(net);
648 * Change the NFSv4 recovery directory to recdir.
651 nfs4_reset_recoverydir(char *recdir)
653 int status;
654 struct path path;
656 status = kern_path(recdir, LOOKUP_FOLLOW, &path);
657 if (status)
658 return status;
659 status = -ENOTDIR;
660 if (d_is_dir(path.dentry)) {
661 strcpy(user_recovery_dirname, recdir);
662 status = 0;
664 path_put(&path);
665 return status;
668 char *
669 nfs4_recoverydir(void)
671 return user_recovery_dirname;
674 static int
675 nfsd4_check_legacy_client(struct nfs4_client *clp)
677 int status;
678 char dname[HEXDIR_LEN];
679 struct nfs4_client_reclaim *crp;
680 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
681 struct xdr_netobj name;
683 /* did we already find that this client is stable? */
684 if (test_bit(NFSD4_CLIENT_STABLE, &clp->cl_flags))
685 return 0;
687 status = nfs4_make_rec_clidname(dname, &clp->cl_name);
688 if (status) {
689 legacy_recdir_name_error(clp, status);
690 return status;
693 /* look for it in the reclaim hashtable otherwise */
694 name.data = kmemdup(dname, HEXDIR_LEN, GFP_KERNEL);
695 if (!name.data) {
696 dprintk("%s: failed to allocate memory for name.data!\n",
697 __func__);
698 goto out_enoent;
700 name.len = HEXDIR_LEN;
701 crp = nfsd4_find_reclaim_client(name, nn);
702 kfree(name.data);
703 if (crp) {
704 set_bit(NFSD4_CLIENT_STABLE, &clp->cl_flags);
705 crp->cr_clp = clp;
706 return 0;
709 out_enoent:
710 return -ENOENT;
713 static const struct nfsd4_client_tracking_ops nfsd4_legacy_tracking_ops = {
714 .init = nfsd4_legacy_tracking_init,
715 .exit = nfsd4_legacy_tracking_exit,
716 .create = nfsd4_create_clid_dir,
717 .remove = nfsd4_remove_clid_dir,
718 .check = nfsd4_check_legacy_client,
719 .grace_done = nfsd4_recdir_purge_old,
720 .version = 1,
721 .msglen = 0,
724 /* Globals */
725 #define NFSD_PIPE_DIR "nfsd"
726 #define NFSD_CLD_PIPE "cld"
728 /* per-net-ns structure for holding cld upcall info */
729 struct cld_net {
730 struct rpc_pipe *cn_pipe;
731 spinlock_t cn_lock;
732 struct list_head cn_list;
733 unsigned int cn_xid;
734 bool cn_has_legacy;
735 struct crypto_shash *cn_tfm;
738 struct cld_upcall {
739 struct list_head cu_list;
740 struct cld_net *cu_net;
741 struct completion cu_done;
742 union {
743 struct cld_msg_hdr cu_hdr;
744 struct cld_msg cu_msg;
745 struct cld_msg_v2 cu_msg_v2;
746 } cu_u;
749 static int
750 __cld_pipe_upcall(struct rpc_pipe *pipe, void *cmsg, struct nfsd_net *nn)
752 int ret;
753 struct rpc_pipe_msg msg;
754 struct cld_upcall *cup = container_of(cmsg, struct cld_upcall, cu_u);
756 memset(&msg, 0, sizeof(msg));
757 msg.data = cmsg;
758 msg.len = nn->client_tracking_ops->msglen;
760 ret = rpc_queue_upcall(pipe, &msg);
761 if (ret < 0) {
762 goto out;
765 wait_for_completion(&cup->cu_done);
767 if (msg.errno < 0)
768 ret = msg.errno;
769 out:
770 return ret;
773 static int
774 cld_pipe_upcall(struct rpc_pipe *pipe, void *cmsg, struct nfsd_net *nn)
776 int ret;
779 * -EAGAIN occurs when pipe is closed and reopened while there are
780 * upcalls queued.
782 do {
783 ret = __cld_pipe_upcall(pipe, cmsg, nn);
784 } while (ret == -EAGAIN);
786 return ret;
789 static ssize_t
790 __cld_pipe_inprogress_downcall(const struct cld_msg_v2 __user *cmsg,
791 struct nfsd_net *nn)
793 uint8_t cmd, princhashlen;
794 struct xdr_netobj name, princhash = { .len = 0, .data = NULL };
795 uint16_t namelen;
796 struct cld_net *cn = nn->cld_net;
798 if (get_user(cmd, &cmsg->cm_cmd)) {
799 dprintk("%s: error when copying cmd from userspace", __func__);
800 return -EFAULT;
802 if (cmd == Cld_GraceStart) {
803 if (nn->client_tracking_ops->version >= 2) {
804 const struct cld_clntinfo __user *ci;
806 ci = &cmsg->cm_u.cm_clntinfo;
807 if (get_user(namelen, &ci->cc_name.cn_len))
808 return -EFAULT;
809 name.data = memdup_user(&ci->cc_name.cn_id, namelen);
810 if (IS_ERR_OR_NULL(name.data))
811 return -EFAULT;
812 name.len = namelen;
813 get_user(princhashlen, &ci->cc_princhash.cp_len);
814 if (princhashlen > 0) {
815 princhash.data = memdup_user(
816 &ci->cc_princhash.cp_data,
817 princhashlen);
818 if (IS_ERR_OR_NULL(princhash.data))
819 return -EFAULT;
820 princhash.len = princhashlen;
821 } else
822 princhash.len = 0;
823 } else {
824 const struct cld_name __user *cnm;
826 cnm = &cmsg->cm_u.cm_name;
827 if (get_user(namelen, &cnm->cn_len))
828 return -EFAULT;
829 name.data = memdup_user(&cnm->cn_id, namelen);
830 if (IS_ERR_OR_NULL(name.data))
831 return -EFAULT;
832 name.len = namelen;
834 if (name.len > 5 && memcmp(name.data, "hash:", 5) == 0) {
835 name.len = name.len - 5;
836 memmove(name.data, name.data + 5, name.len);
837 cn->cn_has_legacy = true;
839 if (!nfs4_client_to_reclaim(name, princhash, nn)) {
840 kfree(name.data);
841 kfree(princhash.data);
842 return -EFAULT;
844 return nn->client_tracking_ops->msglen;
846 return -EFAULT;
849 static ssize_t
850 cld_pipe_downcall(struct file *filp, const char __user *src, size_t mlen)
852 struct cld_upcall *tmp, *cup;
853 struct cld_msg_hdr __user *hdr = (struct cld_msg_hdr __user *)src;
854 struct cld_msg_v2 __user *cmsg = (struct cld_msg_v2 __user *)src;
855 uint32_t xid;
856 struct nfsd_net *nn = net_generic(file_inode(filp)->i_sb->s_fs_info,
857 nfsd_net_id);
858 struct cld_net *cn = nn->cld_net;
859 int16_t status;
861 if (mlen != nn->client_tracking_ops->msglen) {
862 dprintk("%s: got %zu bytes, expected %zu\n", __func__, mlen,
863 nn->client_tracking_ops->msglen);
864 return -EINVAL;
867 /* copy just the xid so we can try to find that */
868 if (copy_from_user(&xid, &hdr->cm_xid, sizeof(xid)) != 0) {
869 dprintk("%s: error when copying xid from userspace", __func__);
870 return -EFAULT;
874 * copy the status so we know whether to remove the upcall from the
875 * list (for -EINPROGRESS, we just want to make sure the xid is
876 * valid, not remove the upcall from the list)
878 if (get_user(status, &hdr->cm_status)) {
879 dprintk("%s: error when copying status from userspace", __func__);
880 return -EFAULT;
883 /* walk the list and find corresponding xid */
884 cup = NULL;
885 spin_lock(&cn->cn_lock);
886 list_for_each_entry(tmp, &cn->cn_list, cu_list) {
887 if (get_unaligned(&tmp->cu_u.cu_hdr.cm_xid) == xid) {
888 cup = tmp;
889 if (status != -EINPROGRESS)
890 list_del_init(&cup->cu_list);
891 break;
894 spin_unlock(&cn->cn_lock);
896 /* couldn't find upcall? */
897 if (!cup) {
898 dprintk("%s: couldn't find upcall -- xid=%u\n", __func__, xid);
899 return -EINVAL;
902 if (status == -EINPROGRESS)
903 return __cld_pipe_inprogress_downcall(cmsg, nn);
905 if (copy_from_user(&cup->cu_u.cu_msg_v2, src, mlen) != 0)
906 return -EFAULT;
908 complete(&cup->cu_done);
909 return mlen;
912 static void
913 cld_pipe_destroy_msg(struct rpc_pipe_msg *msg)
915 struct cld_msg *cmsg = msg->data;
916 struct cld_upcall *cup = container_of(cmsg, struct cld_upcall,
917 cu_u.cu_msg);
919 /* errno >= 0 means we got a downcall */
920 if (msg->errno >= 0)
921 return;
923 complete(&cup->cu_done);
926 static const struct rpc_pipe_ops cld_upcall_ops = {
927 .upcall = rpc_pipe_generic_upcall,
928 .downcall = cld_pipe_downcall,
929 .destroy_msg = cld_pipe_destroy_msg,
932 static struct dentry *
933 nfsd4_cld_register_sb(struct super_block *sb, struct rpc_pipe *pipe)
935 struct dentry *dir, *dentry;
937 dir = rpc_d_lookup_sb(sb, NFSD_PIPE_DIR);
938 if (dir == NULL)
939 return ERR_PTR(-ENOENT);
940 dentry = rpc_mkpipe_dentry(dir, NFSD_CLD_PIPE, NULL, pipe);
941 dput(dir);
942 return dentry;
945 static void
946 nfsd4_cld_unregister_sb(struct rpc_pipe *pipe)
948 if (pipe->dentry)
949 rpc_unlink(pipe->dentry);
952 static struct dentry *
953 nfsd4_cld_register_net(struct net *net, struct rpc_pipe *pipe)
955 struct super_block *sb;
956 struct dentry *dentry;
958 sb = rpc_get_sb_net(net);
959 if (!sb)
960 return NULL;
961 dentry = nfsd4_cld_register_sb(sb, pipe);
962 rpc_put_sb_net(net);
963 return dentry;
966 static void
967 nfsd4_cld_unregister_net(struct net *net, struct rpc_pipe *pipe)
969 struct super_block *sb;
971 sb = rpc_get_sb_net(net);
972 if (sb) {
973 nfsd4_cld_unregister_sb(pipe);
974 rpc_put_sb_net(net);
978 /* Initialize rpc_pipefs pipe for communication with client tracking daemon */
979 static int
980 __nfsd4_init_cld_pipe(struct net *net)
982 int ret;
983 struct dentry *dentry;
984 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
985 struct cld_net *cn;
987 if (nn->cld_net)
988 return 0;
990 cn = kzalloc(sizeof(*cn), GFP_KERNEL);
991 if (!cn) {
992 ret = -ENOMEM;
993 goto err;
996 cn->cn_pipe = rpc_mkpipe_data(&cld_upcall_ops, RPC_PIPE_WAIT_FOR_OPEN);
997 if (IS_ERR(cn->cn_pipe)) {
998 ret = PTR_ERR(cn->cn_pipe);
999 goto err;
1001 spin_lock_init(&cn->cn_lock);
1002 INIT_LIST_HEAD(&cn->cn_list);
1004 dentry = nfsd4_cld_register_net(net, cn->cn_pipe);
1005 if (IS_ERR(dentry)) {
1006 ret = PTR_ERR(dentry);
1007 goto err_destroy_data;
1010 cn->cn_pipe->dentry = dentry;
1011 cn->cn_has_legacy = false;
1012 nn->cld_net = cn;
1013 return 0;
1015 err_destroy_data:
1016 rpc_destroy_pipe_data(cn->cn_pipe);
1017 err:
1018 kfree(cn);
1019 printk(KERN_ERR "NFSD: unable to create nfsdcld upcall pipe (%d)\n",
1020 ret);
1021 return ret;
1024 static int
1025 nfsd4_init_cld_pipe(struct net *net)
1027 int status;
1029 status = __nfsd4_init_cld_pipe(net);
1030 if (!status)
1031 printk("NFSD: Using old nfsdcld client tracking operations.\n");
1032 return status;
1035 static void
1036 nfsd4_remove_cld_pipe(struct net *net)
1038 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
1039 struct cld_net *cn = nn->cld_net;
1041 nfsd4_cld_unregister_net(net, cn->cn_pipe);
1042 rpc_destroy_pipe_data(cn->cn_pipe);
1043 if (cn->cn_tfm)
1044 crypto_free_shash(cn->cn_tfm);
1045 kfree(nn->cld_net);
1046 nn->cld_net = NULL;
1049 static struct cld_upcall *
1050 alloc_cld_upcall(struct nfsd_net *nn)
1052 struct cld_upcall *new, *tmp;
1053 struct cld_net *cn = nn->cld_net;
1055 new = kzalloc(sizeof(*new), GFP_KERNEL);
1056 if (!new)
1057 return new;
1059 /* FIXME: hard cap on number in flight? */
1060 restart_search:
1061 spin_lock(&cn->cn_lock);
1062 list_for_each_entry(tmp, &cn->cn_list, cu_list) {
1063 if (tmp->cu_u.cu_msg.cm_xid == cn->cn_xid) {
1064 cn->cn_xid++;
1065 spin_unlock(&cn->cn_lock);
1066 goto restart_search;
1069 init_completion(&new->cu_done);
1070 new->cu_u.cu_msg.cm_vers = nn->client_tracking_ops->version;
1071 put_unaligned(cn->cn_xid++, &new->cu_u.cu_msg.cm_xid);
1072 new->cu_net = cn;
1073 list_add(&new->cu_list, &cn->cn_list);
1074 spin_unlock(&cn->cn_lock);
1076 dprintk("%s: allocated xid %u\n", __func__, new->cu_u.cu_msg.cm_xid);
1078 return new;
1081 static void
1082 free_cld_upcall(struct cld_upcall *victim)
1084 struct cld_net *cn = victim->cu_net;
1086 spin_lock(&cn->cn_lock);
1087 list_del(&victim->cu_list);
1088 spin_unlock(&cn->cn_lock);
1089 kfree(victim);
1092 /* Ask daemon to create a new record */
1093 static void
1094 nfsd4_cld_create(struct nfs4_client *clp)
1096 int ret;
1097 struct cld_upcall *cup;
1098 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
1099 struct cld_net *cn = nn->cld_net;
1101 /* Don't upcall if it's already stored */
1102 if (test_bit(NFSD4_CLIENT_STABLE, &clp->cl_flags))
1103 return;
1105 cup = alloc_cld_upcall(nn);
1106 if (!cup) {
1107 ret = -ENOMEM;
1108 goto out_err;
1111 cup->cu_u.cu_msg.cm_cmd = Cld_Create;
1112 cup->cu_u.cu_msg.cm_u.cm_name.cn_len = clp->cl_name.len;
1113 memcpy(cup->cu_u.cu_msg.cm_u.cm_name.cn_id, clp->cl_name.data,
1114 clp->cl_name.len);
1116 ret = cld_pipe_upcall(cn->cn_pipe, &cup->cu_u.cu_msg, nn);
1117 if (!ret) {
1118 ret = cup->cu_u.cu_msg.cm_status;
1119 set_bit(NFSD4_CLIENT_STABLE, &clp->cl_flags);
1122 free_cld_upcall(cup);
1123 out_err:
1124 if (ret)
1125 printk(KERN_ERR "NFSD: Unable to create client "
1126 "record on stable storage: %d\n", ret);
1129 /* Ask daemon to create a new record */
1130 static void
1131 nfsd4_cld_create_v2(struct nfs4_client *clp)
1133 int ret;
1134 struct cld_upcall *cup;
1135 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
1136 struct cld_net *cn = nn->cld_net;
1137 struct cld_msg_v2 *cmsg;
1138 struct crypto_shash *tfm = cn->cn_tfm;
1139 struct xdr_netobj cksum;
1140 char *principal = NULL;
1142 /* Don't upcall if it's already stored */
1143 if (test_bit(NFSD4_CLIENT_STABLE, &clp->cl_flags))
1144 return;
1146 cup = alloc_cld_upcall(nn);
1147 if (!cup) {
1148 ret = -ENOMEM;
1149 goto out_err;
1152 cmsg = &cup->cu_u.cu_msg_v2;
1153 cmsg->cm_cmd = Cld_Create;
1154 cmsg->cm_u.cm_clntinfo.cc_name.cn_len = clp->cl_name.len;
1155 memcpy(cmsg->cm_u.cm_clntinfo.cc_name.cn_id, clp->cl_name.data,
1156 clp->cl_name.len);
1157 if (clp->cl_cred.cr_raw_principal)
1158 principal = clp->cl_cred.cr_raw_principal;
1159 else if (clp->cl_cred.cr_principal)
1160 principal = clp->cl_cred.cr_principal;
1161 if (principal) {
1162 cksum.len = crypto_shash_digestsize(tfm);
1163 cksum.data = kmalloc(cksum.len, GFP_KERNEL);
1164 if (cksum.data == NULL) {
1165 ret = -ENOMEM;
1166 goto out;
1168 ret = crypto_shash_tfm_digest(tfm, principal, strlen(principal),
1169 cksum.data);
1170 if (ret) {
1171 kfree(cksum.data);
1172 goto out;
1174 cmsg->cm_u.cm_clntinfo.cc_princhash.cp_len = cksum.len;
1175 memcpy(cmsg->cm_u.cm_clntinfo.cc_princhash.cp_data,
1176 cksum.data, cksum.len);
1177 kfree(cksum.data);
1178 } else
1179 cmsg->cm_u.cm_clntinfo.cc_princhash.cp_len = 0;
1181 ret = cld_pipe_upcall(cn->cn_pipe, cmsg, nn);
1182 if (!ret) {
1183 ret = cmsg->cm_status;
1184 set_bit(NFSD4_CLIENT_STABLE, &clp->cl_flags);
1187 out:
1188 free_cld_upcall(cup);
1189 out_err:
1190 if (ret)
1191 pr_err("NFSD: Unable to create client record on stable storage: %d\n",
1192 ret);
1195 /* Ask daemon to create a new record */
1196 static void
1197 nfsd4_cld_remove(struct nfs4_client *clp)
1199 int ret;
1200 struct cld_upcall *cup;
1201 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
1202 struct cld_net *cn = nn->cld_net;
1204 /* Don't upcall if it's already removed */
1205 if (!test_bit(NFSD4_CLIENT_STABLE, &clp->cl_flags))
1206 return;
1208 cup = alloc_cld_upcall(nn);
1209 if (!cup) {
1210 ret = -ENOMEM;
1211 goto out_err;
1214 cup->cu_u.cu_msg.cm_cmd = Cld_Remove;
1215 cup->cu_u.cu_msg.cm_u.cm_name.cn_len = clp->cl_name.len;
1216 memcpy(cup->cu_u.cu_msg.cm_u.cm_name.cn_id, clp->cl_name.data,
1217 clp->cl_name.len);
1219 ret = cld_pipe_upcall(cn->cn_pipe, &cup->cu_u.cu_msg, nn);
1220 if (!ret) {
1221 ret = cup->cu_u.cu_msg.cm_status;
1222 clear_bit(NFSD4_CLIENT_STABLE, &clp->cl_flags);
1225 free_cld_upcall(cup);
1226 out_err:
1227 if (ret)
1228 printk(KERN_ERR "NFSD: Unable to remove client "
1229 "record from stable storage: %d\n", ret);
1233 * For older nfsdcld's that do not allow us to "slurp" the clients
1234 * from the tracking database during startup.
1236 * Check for presence of a record, and update its timestamp
1238 static int
1239 nfsd4_cld_check_v0(struct nfs4_client *clp)
1241 int ret;
1242 struct cld_upcall *cup;
1243 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
1244 struct cld_net *cn = nn->cld_net;
1246 /* Don't upcall if one was already stored during this grace pd */
1247 if (test_bit(NFSD4_CLIENT_STABLE, &clp->cl_flags))
1248 return 0;
1250 cup = alloc_cld_upcall(nn);
1251 if (!cup) {
1252 printk(KERN_ERR "NFSD: Unable to check client record on "
1253 "stable storage: %d\n", -ENOMEM);
1254 return -ENOMEM;
1257 cup->cu_u.cu_msg.cm_cmd = Cld_Check;
1258 cup->cu_u.cu_msg.cm_u.cm_name.cn_len = clp->cl_name.len;
1259 memcpy(cup->cu_u.cu_msg.cm_u.cm_name.cn_id, clp->cl_name.data,
1260 clp->cl_name.len);
1262 ret = cld_pipe_upcall(cn->cn_pipe, &cup->cu_u.cu_msg, nn);
1263 if (!ret) {
1264 ret = cup->cu_u.cu_msg.cm_status;
1265 set_bit(NFSD4_CLIENT_STABLE, &clp->cl_flags);
1268 free_cld_upcall(cup);
1269 return ret;
1273 * For newer nfsdcld's that allow us to "slurp" the clients
1274 * from the tracking database during startup.
1276 * Check for presence of a record in the reclaim_str_hashtbl
1278 static int
1279 nfsd4_cld_check(struct nfs4_client *clp)
1281 struct nfs4_client_reclaim *crp;
1282 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
1283 struct cld_net *cn = nn->cld_net;
1284 int status;
1285 char dname[HEXDIR_LEN];
1286 struct xdr_netobj name;
1288 /* did we already find that this client is stable? */
1289 if (test_bit(NFSD4_CLIENT_STABLE, &clp->cl_flags))
1290 return 0;
1292 /* look for it in the reclaim hashtable otherwise */
1293 crp = nfsd4_find_reclaim_client(clp->cl_name, nn);
1294 if (crp)
1295 goto found;
1297 if (cn->cn_has_legacy) {
1298 status = nfs4_make_rec_clidname(dname, &clp->cl_name);
1299 if (status)
1300 return -ENOENT;
1302 name.data = kmemdup(dname, HEXDIR_LEN, GFP_KERNEL);
1303 if (!name.data) {
1304 dprintk("%s: failed to allocate memory for name.data!\n",
1305 __func__);
1306 return -ENOENT;
1308 name.len = HEXDIR_LEN;
1309 crp = nfsd4_find_reclaim_client(name, nn);
1310 kfree(name.data);
1311 if (crp)
1312 goto found;
1315 return -ENOENT;
1316 found:
1317 crp->cr_clp = clp;
1318 return 0;
1321 static int
1322 nfsd4_cld_check_v2(struct nfs4_client *clp)
1324 struct nfs4_client_reclaim *crp;
1325 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
1326 struct cld_net *cn = nn->cld_net;
1327 int status;
1328 char dname[HEXDIR_LEN];
1329 struct xdr_netobj name;
1330 struct crypto_shash *tfm = cn->cn_tfm;
1331 struct xdr_netobj cksum;
1332 char *principal = NULL;
1334 /* did we already find that this client is stable? */
1335 if (test_bit(NFSD4_CLIENT_STABLE, &clp->cl_flags))
1336 return 0;
1338 /* look for it in the reclaim hashtable otherwise */
1339 crp = nfsd4_find_reclaim_client(clp->cl_name, nn);
1340 if (crp)
1341 goto found;
1343 if (cn->cn_has_legacy) {
1344 status = nfs4_make_rec_clidname(dname, &clp->cl_name);
1345 if (status)
1346 return -ENOENT;
1348 name.data = kmemdup(dname, HEXDIR_LEN, GFP_KERNEL);
1349 if (!name.data) {
1350 dprintk("%s: failed to allocate memory for name.data\n",
1351 __func__);
1352 return -ENOENT;
1354 name.len = HEXDIR_LEN;
1355 crp = nfsd4_find_reclaim_client(name, nn);
1356 kfree(name.data);
1357 if (crp)
1358 goto found;
1361 return -ENOENT;
1362 found:
1363 if (crp->cr_princhash.len) {
1364 if (clp->cl_cred.cr_raw_principal)
1365 principal = clp->cl_cred.cr_raw_principal;
1366 else if (clp->cl_cred.cr_principal)
1367 principal = clp->cl_cred.cr_principal;
1368 if (principal == NULL)
1369 return -ENOENT;
1370 cksum.len = crypto_shash_digestsize(tfm);
1371 cksum.data = kmalloc(cksum.len, GFP_KERNEL);
1372 if (cksum.data == NULL)
1373 return -ENOENT;
1374 status = crypto_shash_tfm_digest(tfm, principal,
1375 strlen(principal), cksum.data);
1376 if (status) {
1377 kfree(cksum.data);
1378 return -ENOENT;
1380 if (memcmp(crp->cr_princhash.data, cksum.data,
1381 crp->cr_princhash.len)) {
1382 kfree(cksum.data);
1383 return -ENOENT;
1385 kfree(cksum.data);
1387 crp->cr_clp = clp;
1388 return 0;
1391 static int
1392 nfsd4_cld_grace_start(struct nfsd_net *nn)
1394 int ret;
1395 struct cld_upcall *cup;
1396 struct cld_net *cn = nn->cld_net;
1398 cup = alloc_cld_upcall(nn);
1399 if (!cup) {
1400 ret = -ENOMEM;
1401 goto out_err;
1404 cup->cu_u.cu_msg.cm_cmd = Cld_GraceStart;
1405 ret = cld_pipe_upcall(cn->cn_pipe, &cup->cu_u.cu_msg, nn);
1406 if (!ret)
1407 ret = cup->cu_u.cu_msg.cm_status;
1409 free_cld_upcall(cup);
1410 out_err:
1411 if (ret)
1412 dprintk("%s: Unable to get clients from userspace: %d\n",
1413 __func__, ret);
1414 return ret;
1417 /* For older nfsdcld's that need cm_gracetime */
1418 static void
1419 nfsd4_cld_grace_done_v0(struct nfsd_net *nn)
1421 int ret;
1422 struct cld_upcall *cup;
1423 struct cld_net *cn = nn->cld_net;
1425 cup = alloc_cld_upcall(nn);
1426 if (!cup) {
1427 ret = -ENOMEM;
1428 goto out_err;
1431 cup->cu_u.cu_msg.cm_cmd = Cld_GraceDone;
1432 cup->cu_u.cu_msg.cm_u.cm_gracetime = nn->boot_time;
1433 ret = cld_pipe_upcall(cn->cn_pipe, &cup->cu_u.cu_msg, nn);
1434 if (!ret)
1435 ret = cup->cu_u.cu_msg.cm_status;
1437 free_cld_upcall(cup);
1438 out_err:
1439 if (ret)
1440 printk(KERN_ERR "NFSD: Unable to end grace period: %d\n", ret);
1444 * For newer nfsdcld's that do not need cm_gracetime. We also need to call
1445 * nfs4_release_reclaim() to clear out the reclaim_str_hashtbl.
1447 static void
1448 nfsd4_cld_grace_done(struct nfsd_net *nn)
1450 int ret;
1451 struct cld_upcall *cup;
1452 struct cld_net *cn = nn->cld_net;
1454 cup = alloc_cld_upcall(nn);
1455 if (!cup) {
1456 ret = -ENOMEM;
1457 goto out_err;
1460 cup->cu_u.cu_msg.cm_cmd = Cld_GraceDone;
1461 ret = cld_pipe_upcall(cn->cn_pipe, &cup->cu_u.cu_msg, nn);
1462 if (!ret)
1463 ret = cup->cu_u.cu_msg.cm_status;
1465 free_cld_upcall(cup);
1466 out_err:
1467 nfs4_release_reclaim(nn);
1468 if (ret)
1469 printk(KERN_ERR "NFSD: Unable to end grace period: %d\n", ret);
1472 static int
1473 nfs4_cld_state_init(struct net *net)
1475 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
1476 int i;
1478 nn->reclaim_str_hashtbl = kmalloc_array(CLIENT_HASH_SIZE,
1479 sizeof(struct list_head),
1480 GFP_KERNEL);
1481 if (!nn->reclaim_str_hashtbl)
1482 return -ENOMEM;
1484 for (i = 0; i < CLIENT_HASH_SIZE; i++)
1485 INIT_LIST_HEAD(&nn->reclaim_str_hashtbl[i]);
1486 nn->reclaim_str_hashtbl_size = 0;
1487 nn->track_reclaim_completes = true;
1488 atomic_set(&nn->nr_reclaim_complete, 0);
1490 return 0;
1493 static void
1494 nfs4_cld_state_shutdown(struct net *net)
1496 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
1498 nn->track_reclaim_completes = false;
1499 kfree(nn->reclaim_str_hashtbl);
1502 static bool
1503 cld_running(struct nfsd_net *nn)
1505 struct cld_net *cn = nn->cld_net;
1506 struct rpc_pipe *pipe = cn->cn_pipe;
1508 return pipe->nreaders || pipe->nwriters;
1511 static int
1512 nfsd4_cld_get_version(struct nfsd_net *nn)
1514 int ret = 0;
1515 struct cld_upcall *cup;
1516 struct cld_net *cn = nn->cld_net;
1517 uint8_t version;
1519 cup = alloc_cld_upcall(nn);
1520 if (!cup) {
1521 ret = -ENOMEM;
1522 goto out_err;
1524 cup->cu_u.cu_msg.cm_cmd = Cld_GetVersion;
1525 ret = cld_pipe_upcall(cn->cn_pipe, &cup->cu_u.cu_msg, nn);
1526 if (!ret) {
1527 ret = cup->cu_u.cu_msg.cm_status;
1528 if (ret)
1529 goto out_free;
1530 version = cup->cu_u.cu_msg.cm_u.cm_version;
1531 dprintk("%s: userspace returned version %u\n",
1532 __func__, version);
1533 if (version < 1)
1534 version = 1;
1535 else if (version > CLD_UPCALL_VERSION)
1536 version = CLD_UPCALL_VERSION;
1538 switch (version) {
1539 case 1:
1540 nn->client_tracking_ops = &nfsd4_cld_tracking_ops;
1541 break;
1542 case 2:
1543 nn->client_tracking_ops = &nfsd4_cld_tracking_ops_v2;
1544 break;
1545 default:
1546 break;
1549 out_free:
1550 free_cld_upcall(cup);
1551 out_err:
1552 if (ret)
1553 dprintk("%s: Unable to get version from userspace: %d\n",
1554 __func__, ret);
1555 return ret;
1558 static int
1559 nfsd4_cld_tracking_init(struct net *net)
1561 int status;
1562 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
1563 bool running;
1564 int retries = 10;
1565 struct crypto_shash *tfm;
1567 status = nfs4_cld_state_init(net);
1568 if (status)
1569 return status;
1571 status = __nfsd4_init_cld_pipe(net);
1572 if (status)
1573 goto err_shutdown;
1576 * rpc pipe upcalls take 30 seconds to time out, so we don't want to
1577 * queue an upcall unless we know that nfsdcld is running (because we
1578 * want this to fail fast so that nfsd4_client_tracking_init() can try
1579 * the next client tracking method). nfsdcld should already be running
1580 * before nfsd is started, so the wait here is for nfsdcld to open the
1581 * pipefs file we just created.
1583 while (!(running = cld_running(nn)) && retries--)
1584 msleep(100);
1586 if (!running) {
1587 status = -ETIMEDOUT;
1588 goto err_remove;
1590 tfm = crypto_alloc_shash("sha256", 0, 0);
1591 if (IS_ERR(tfm)) {
1592 status = PTR_ERR(tfm);
1593 goto err_remove;
1595 nn->cld_net->cn_tfm = tfm;
1597 status = nfsd4_cld_get_version(nn);
1598 if (status == -EOPNOTSUPP)
1599 pr_warn("NFSD: nfsdcld GetVersion upcall failed. Please upgrade nfsdcld.\n");
1601 status = nfsd4_cld_grace_start(nn);
1602 if (status) {
1603 if (status == -EOPNOTSUPP)
1604 pr_warn("NFSD: nfsdcld GraceStart upcall failed. Please upgrade nfsdcld.\n");
1605 nfs4_release_reclaim(nn);
1606 goto err_remove;
1607 } else
1608 printk("NFSD: Using nfsdcld client tracking operations.\n");
1609 return 0;
1611 err_remove:
1612 nfsd4_remove_cld_pipe(net);
1613 err_shutdown:
1614 nfs4_cld_state_shutdown(net);
1615 return status;
1618 static void
1619 nfsd4_cld_tracking_exit(struct net *net)
1621 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
1623 nfs4_release_reclaim(nn);
1624 nfsd4_remove_cld_pipe(net);
1625 nfs4_cld_state_shutdown(net);
1628 /* For older nfsdcld's */
1629 static const struct nfsd4_client_tracking_ops nfsd4_cld_tracking_ops_v0 = {
1630 .init = nfsd4_init_cld_pipe,
1631 .exit = nfsd4_remove_cld_pipe,
1632 .create = nfsd4_cld_create,
1633 .remove = nfsd4_cld_remove,
1634 .check = nfsd4_cld_check_v0,
1635 .grace_done = nfsd4_cld_grace_done_v0,
1636 .version = 1,
1637 .msglen = sizeof(struct cld_msg),
1640 /* For newer nfsdcld's */
1641 static const struct nfsd4_client_tracking_ops nfsd4_cld_tracking_ops = {
1642 .init = nfsd4_cld_tracking_init,
1643 .exit = nfsd4_cld_tracking_exit,
1644 .create = nfsd4_cld_create,
1645 .remove = nfsd4_cld_remove,
1646 .check = nfsd4_cld_check,
1647 .grace_done = nfsd4_cld_grace_done,
1648 .version = 1,
1649 .msglen = sizeof(struct cld_msg),
1652 /* v2 create/check ops include the principal, if available */
1653 static const struct nfsd4_client_tracking_ops nfsd4_cld_tracking_ops_v2 = {
1654 .init = nfsd4_cld_tracking_init,
1655 .exit = nfsd4_cld_tracking_exit,
1656 .create = nfsd4_cld_create_v2,
1657 .remove = nfsd4_cld_remove,
1658 .check = nfsd4_cld_check_v2,
1659 .grace_done = nfsd4_cld_grace_done,
1660 .version = 2,
1661 .msglen = sizeof(struct cld_msg_v2),
1664 /* upcall via usermodehelper */
1665 static char cltrack_prog[PATH_MAX] = "/sbin/nfsdcltrack";
1666 module_param_string(cltrack_prog, cltrack_prog, sizeof(cltrack_prog),
1667 S_IRUGO|S_IWUSR);
1668 MODULE_PARM_DESC(cltrack_prog, "Path to the nfsdcltrack upcall program");
1670 static bool cltrack_legacy_disable;
1671 module_param(cltrack_legacy_disable, bool, S_IRUGO|S_IWUSR);
1672 MODULE_PARM_DESC(cltrack_legacy_disable,
1673 "Disable legacy recoverydir conversion. Default: false");
1675 #define LEGACY_TOPDIR_ENV_PREFIX "NFSDCLTRACK_LEGACY_TOPDIR="
1676 #define LEGACY_RECDIR_ENV_PREFIX "NFSDCLTRACK_LEGACY_RECDIR="
1677 #define HAS_SESSION_ENV_PREFIX "NFSDCLTRACK_CLIENT_HAS_SESSION="
1678 #define GRACE_START_ENV_PREFIX "NFSDCLTRACK_GRACE_START="
1680 static char *
1681 nfsd4_cltrack_legacy_topdir(void)
1683 int copied;
1684 size_t len;
1685 char *result;
1687 if (cltrack_legacy_disable)
1688 return NULL;
1690 len = strlen(LEGACY_TOPDIR_ENV_PREFIX) +
1691 strlen(nfs4_recoverydir()) + 1;
1693 result = kmalloc(len, GFP_KERNEL);
1694 if (!result)
1695 return result;
1697 copied = snprintf(result, len, LEGACY_TOPDIR_ENV_PREFIX "%s",
1698 nfs4_recoverydir());
1699 if (copied >= len) {
1700 /* just return nothing if output was truncated */
1701 kfree(result);
1702 return NULL;
1705 return result;
1708 static char *
1709 nfsd4_cltrack_legacy_recdir(const struct xdr_netobj *name)
1711 int copied;
1712 size_t len;
1713 char *result;
1715 if (cltrack_legacy_disable)
1716 return NULL;
1718 /* +1 is for '/' between "topdir" and "recdir" */
1719 len = strlen(LEGACY_RECDIR_ENV_PREFIX) +
1720 strlen(nfs4_recoverydir()) + 1 + HEXDIR_LEN;
1722 result = kmalloc(len, GFP_KERNEL);
1723 if (!result)
1724 return result;
1726 copied = snprintf(result, len, LEGACY_RECDIR_ENV_PREFIX "%s/",
1727 nfs4_recoverydir());
1728 if (copied > (len - HEXDIR_LEN)) {
1729 /* just return nothing if output will be truncated */
1730 kfree(result);
1731 return NULL;
1734 copied = nfs4_make_rec_clidname(result + copied, name);
1735 if (copied) {
1736 kfree(result);
1737 return NULL;
1740 return result;
1743 static char *
1744 nfsd4_cltrack_client_has_session(struct nfs4_client *clp)
1746 int copied;
1747 size_t len;
1748 char *result;
1750 /* prefix + Y/N character + terminating NULL */
1751 len = strlen(HAS_SESSION_ENV_PREFIX) + 1 + 1;
1753 result = kmalloc(len, GFP_KERNEL);
1754 if (!result)
1755 return result;
1757 copied = snprintf(result, len, HAS_SESSION_ENV_PREFIX "%c",
1758 clp->cl_minorversion ? 'Y' : 'N');
1759 if (copied >= len) {
1760 /* just return nothing if output was truncated */
1761 kfree(result);
1762 return NULL;
1765 return result;
1768 static char *
1769 nfsd4_cltrack_grace_start(time64_t grace_start)
1771 int copied;
1772 size_t len;
1773 char *result;
1775 /* prefix + max width of int64_t string + terminating NULL */
1776 len = strlen(GRACE_START_ENV_PREFIX) + 22 + 1;
1778 result = kmalloc(len, GFP_KERNEL);
1779 if (!result)
1780 return result;
1782 copied = snprintf(result, len, GRACE_START_ENV_PREFIX "%lld",
1783 grace_start);
1784 if (copied >= len) {
1785 /* just return nothing if output was truncated */
1786 kfree(result);
1787 return NULL;
1790 return result;
1793 static int
1794 nfsd4_umh_cltrack_upcall(char *cmd, char *arg, char *env0, char *env1)
1796 char *envp[3];
1797 char *argv[4];
1798 int ret;
1800 if (unlikely(!cltrack_prog[0])) {
1801 dprintk("%s: cltrack_prog is disabled\n", __func__);
1802 return -EACCES;
1805 dprintk("%s: cmd: %s\n", __func__, cmd);
1806 dprintk("%s: arg: %s\n", __func__, arg ? arg : "(null)");
1807 dprintk("%s: env0: %s\n", __func__, env0 ? env0 : "(null)");
1808 dprintk("%s: env1: %s\n", __func__, env1 ? env1 : "(null)");
1810 envp[0] = env0;
1811 envp[1] = env1;
1812 envp[2] = NULL;
1814 argv[0] = (char *)cltrack_prog;
1815 argv[1] = cmd;
1816 argv[2] = arg;
1817 argv[3] = NULL;
1819 ret = call_usermodehelper(argv[0], argv, envp, UMH_WAIT_PROC);
1821 * Disable the upcall mechanism if we're getting an ENOENT or EACCES
1822 * error. The admin can re-enable it on the fly by using sysfs
1823 * once the problem has been fixed.
1825 if (ret == -ENOENT || ret == -EACCES) {
1826 dprintk("NFSD: %s was not found or isn't executable (%d). "
1827 "Setting cltrack_prog to blank string!",
1828 cltrack_prog, ret);
1829 cltrack_prog[0] = '\0';
1831 dprintk("%s: %s return value: %d\n", __func__, cltrack_prog, ret);
1833 return ret;
1836 static char *
1837 bin_to_hex_dup(const unsigned char *src, int srclen)
1839 char *buf;
1841 /* +1 for terminating NULL */
1842 buf = kzalloc((srclen * 2) + 1, GFP_KERNEL);
1843 if (!buf)
1844 return buf;
1846 bin2hex(buf, src, srclen);
1847 return buf;
1850 static int
1851 nfsd4_umh_cltrack_init(struct net *net)
1853 int ret;
1854 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
1855 char *grace_start = nfsd4_cltrack_grace_start(nn->boot_time);
1857 /* XXX: The usermode helper s not working in container yet. */
1858 if (net != &init_net) {
1859 pr_warn("NFSD: attempt to initialize umh client tracking in a container ignored.\n");
1860 kfree(grace_start);
1861 return -EINVAL;
1864 ret = nfsd4_umh_cltrack_upcall("init", NULL, grace_start, NULL);
1865 kfree(grace_start);
1866 if (!ret)
1867 printk("NFSD: Using UMH upcall client tracking operations.\n");
1868 return ret;
1871 static void
1872 nfsd4_cltrack_upcall_lock(struct nfs4_client *clp)
1874 wait_on_bit_lock(&clp->cl_flags, NFSD4_CLIENT_UPCALL_LOCK,
1875 TASK_UNINTERRUPTIBLE);
1878 static void
1879 nfsd4_cltrack_upcall_unlock(struct nfs4_client *clp)
1881 smp_mb__before_atomic();
1882 clear_bit(NFSD4_CLIENT_UPCALL_LOCK, &clp->cl_flags);
1883 smp_mb__after_atomic();
1884 wake_up_bit(&clp->cl_flags, NFSD4_CLIENT_UPCALL_LOCK);
1887 static void
1888 nfsd4_umh_cltrack_create(struct nfs4_client *clp)
1890 char *hexid, *has_session, *grace_start;
1891 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
1894 * With v4.0 clients, there's little difference in outcome between a
1895 * create and check operation, and we can end up calling into this
1896 * function multiple times per client (once for each openowner). So,
1897 * for v4.0 clients skip upcalling once the client has been recorded
1898 * on stable storage.
1900 * For v4.1+ clients, the outcome of the two operations is different,
1901 * so we must ensure that we upcall for the create operation. v4.1+
1902 * clients call this on RECLAIM_COMPLETE though, so we should only end
1903 * up doing a single create upcall per client.
1905 if (clp->cl_minorversion == 0 &&
1906 test_bit(NFSD4_CLIENT_STABLE, &clp->cl_flags))
1907 return;
1909 hexid = bin_to_hex_dup(clp->cl_name.data, clp->cl_name.len);
1910 if (!hexid) {
1911 dprintk("%s: can't allocate memory for upcall!\n", __func__);
1912 return;
1915 has_session = nfsd4_cltrack_client_has_session(clp);
1916 grace_start = nfsd4_cltrack_grace_start(nn->boot_time);
1918 nfsd4_cltrack_upcall_lock(clp);
1919 if (!nfsd4_umh_cltrack_upcall("create", hexid, has_session, grace_start))
1920 set_bit(NFSD4_CLIENT_STABLE, &clp->cl_flags);
1921 nfsd4_cltrack_upcall_unlock(clp);
1923 kfree(has_session);
1924 kfree(grace_start);
1925 kfree(hexid);
1928 static void
1929 nfsd4_umh_cltrack_remove(struct nfs4_client *clp)
1931 char *hexid;
1933 if (!test_bit(NFSD4_CLIENT_STABLE, &clp->cl_flags))
1934 return;
1936 hexid = bin_to_hex_dup(clp->cl_name.data, clp->cl_name.len);
1937 if (!hexid) {
1938 dprintk("%s: can't allocate memory for upcall!\n", __func__);
1939 return;
1942 nfsd4_cltrack_upcall_lock(clp);
1943 if (test_bit(NFSD4_CLIENT_STABLE, &clp->cl_flags) &&
1944 nfsd4_umh_cltrack_upcall("remove", hexid, NULL, NULL) == 0)
1945 clear_bit(NFSD4_CLIENT_STABLE, &clp->cl_flags);
1946 nfsd4_cltrack_upcall_unlock(clp);
1948 kfree(hexid);
1951 static int
1952 nfsd4_umh_cltrack_check(struct nfs4_client *clp)
1954 int ret;
1955 char *hexid, *has_session, *legacy;
1957 if (test_bit(NFSD4_CLIENT_STABLE, &clp->cl_flags))
1958 return 0;
1960 hexid = bin_to_hex_dup(clp->cl_name.data, clp->cl_name.len);
1961 if (!hexid) {
1962 dprintk("%s: can't allocate memory for upcall!\n", __func__);
1963 return -ENOMEM;
1966 has_session = nfsd4_cltrack_client_has_session(clp);
1967 legacy = nfsd4_cltrack_legacy_recdir(&clp->cl_name);
1969 nfsd4_cltrack_upcall_lock(clp);
1970 if (test_bit(NFSD4_CLIENT_STABLE, &clp->cl_flags)) {
1971 ret = 0;
1972 } else {
1973 ret = nfsd4_umh_cltrack_upcall("check", hexid, has_session, legacy);
1974 if (ret == 0)
1975 set_bit(NFSD4_CLIENT_STABLE, &clp->cl_flags);
1977 nfsd4_cltrack_upcall_unlock(clp);
1978 kfree(has_session);
1979 kfree(legacy);
1980 kfree(hexid);
1982 return ret;
1985 static void
1986 nfsd4_umh_cltrack_grace_done(struct nfsd_net *nn)
1988 char *legacy;
1989 char timestr[22]; /* FIXME: better way to determine max size? */
1991 sprintf(timestr, "%lld", nn->boot_time);
1992 legacy = nfsd4_cltrack_legacy_topdir();
1993 nfsd4_umh_cltrack_upcall("gracedone", timestr, legacy, NULL);
1994 kfree(legacy);
1997 static const struct nfsd4_client_tracking_ops nfsd4_umh_tracking_ops = {
1998 .init = nfsd4_umh_cltrack_init,
1999 .exit = NULL,
2000 .create = nfsd4_umh_cltrack_create,
2001 .remove = nfsd4_umh_cltrack_remove,
2002 .check = nfsd4_umh_cltrack_check,
2003 .grace_done = nfsd4_umh_cltrack_grace_done,
2004 .version = 1,
2005 .msglen = 0,
2009 nfsd4_client_tracking_init(struct net *net)
2011 int status;
2012 struct path path;
2013 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
2015 /* just run the init if it the method is already decided */
2016 if (nn->client_tracking_ops)
2017 goto do_init;
2019 /* First, try to use nfsdcld */
2020 nn->client_tracking_ops = &nfsd4_cld_tracking_ops;
2021 status = nn->client_tracking_ops->init(net);
2022 if (!status)
2023 return status;
2024 if (status != -ETIMEDOUT) {
2025 nn->client_tracking_ops = &nfsd4_cld_tracking_ops_v0;
2026 status = nn->client_tracking_ops->init(net);
2027 if (!status)
2028 return status;
2032 * Next, try the UMH upcall.
2034 nn->client_tracking_ops = &nfsd4_umh_tracking_ops;
2035 status = nn->client_tracking_ops->init(net);
2036 if (!status)
2037 return status;
2040 * Finally, See if the recoverydir exists and is a directory.
2041 * If it is, then use the legacy ops.
2043 nn->client_tracking_ops = &nfsd4_legacy_tracking_ops;
2044 status = kern_path(nfs4_recoverydir(), LOOKUP_FOLLOW, &path);
2045 if (!status) {
2046 status = d_is_dir(path.dentry);
2047 path_put(&path);
2048 if (!status) {
2049 status = -EINVAL;
2050 goto out;
2054 do_init:
2055 status = nn->client_tracking_ops->init(net);
2056 out:
2057 if (status) {
2058 printk(KERN_WARNING "NFSD: Unable to initialize client "
2059 "recovery tracking! (%d)\n", status);
2060 nn->client_tracking_ops = NULL;
2062 return status;
2065 void
2066 nfsd4_client_tracking_exit(struct net *net)
2068 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
2070 if (nn->client_tracking_ops) {
2071 if (nn->client_tracking_ops->exit)
2072 nn->client_tracking_ops->exit(net);
2073 nn->client_tracking_ops = NULL;
2077 void
2078 nfsd4_client_record_create(struct nfs4_client *clp)
2080 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
2082 if (nn->client_tracking_ops)
2083 nn->client_tracking_ops->create(clp);
2086 void
2087 nfsd4_client_record_remove(struct nfs4_client *clp)
2089 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
2091 if (nn->client_tracking_ops)
2092 nn->client_tracking_ops->remove(clp);
2096 nfsd4_client_record_check(struct nfs4_client *clp)
2098 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
2100 if (nn->client_tracking_ops)
2101 return nn->client_tracking_ops->check(clp);
2103 return -EOPNOTSUPP;
2106 void
2107 nfsd4_record_grace_done(struct nfsd_net *nn)
2109 if (nn->client_tracking_ops)
2110 nn->client_tracking_ops->grace_done(nn);
2113 static int
2114 rpc_pipefs_event(struct notifier_block *nb, unsigned long event, void *ptr)
2116 struct super_block *sb = ptr;
2117 struct net *net = sb->s_fs_info;
2118 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
2119 struct cld_net *cn = nn->cld_net;
2120 struct dentry *dentry;
2121 int ret = 0;
2123 if (!try_module_get(THIS_MODULE))
2124 return 0;
2126 if (!cn) {
2127 module_put(THIS_MODULE);
2128 return 0;
2131 switch (event) {
2132 case RPC_PIPEFS_MOUNT:
2133 dentry = nfsd4_cld_register_sb(sb, cn->cn_pipe);
2134 if (IS_ERR(dentry)) {
2135 ret = PTR_ERR(dentry);
2136 break;
2138 cn->cn_pipe->dentry = dentry;
2139 break;
2140 case RPC_PIPEFS_UMOUNT:
2141 if (cn->cn_pipe->dentry)
2142 nfsd4_cld_unregister_sb(cn->cn_pipe);
2143 break;
2144 default:
2145 ret = -ENOTSUPP;
2146 break;
2148 module_put(THIS_MODULE);
2149 return ret;
2152 static struct notifier_block nfsd4_cld_block = {
2153 .notifier_call = rpc_pipefs_event,
2157 register_cld_notifier(void)
2159 return rpc_pipefs_notifier_register(&nfsd4_cld_block);
2162 void
2163 unregister_cld_notifier(void)
2165 rpc_pipefs_notifier_unregister(&nfsd4_cld_block);