1/*
2 * Copyright (c) 2001 The Regents of the University of Michigan.
3 * All rights reserved.
4 *
5 * Kendrick Smith <kmsmith@umich.edu>
6 * Andy Adamson <andros@umich.edu>
7 *
8 * Redistribution and use in source and binary forms, with or without
9 * modification, are permitted provided that the following conditions
10 * are met:
11 *
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.
20 *
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.
32 */
33
34#include <linux/nfs4.h>
35#include <linux/sunrpc/clnt.h>
36#include <linux/sunrpc/xprt.h>
37#include <linux/sunrpc/svc_xprt.h>
38#include <linux/slab.h>
39#include "nfsd.h"
40#include "state.h"
41#include "netns.h"
42#include "trace.h"
43#include "xdr4cb.h"
44#include "xdr4.h"
45#include "nfs4xdr_gen.h"
46
47#define NFSDDBG_FACILITY NFSDDBG_PROC
48
49#define NFSPROC4_CB_NULL 0
50#define NFSPROC4_CB_COMPOUND 1
51
52/* Index of predefined Linux callback client operations */
53
54struct nfs4_cb_compound_hdr {
55 /* args */
56 u32 ident; /* minorversion 0 only */
57 u32 nops;
58 __be32 *nops_p;
59 u32 minorversion;
60 /* res */
61 int status;
62};
63
64static __be32 *xdr_encode_empty_array(__be32 *p)
65{
66 *p++ = xdr_zero;
67 return p;
68}
69
70/*
71 * Encode/decode NFSv4 CB basic data types
72 *
73 * Basic NFSv4 callback data types are defined in section 15 of RFC
74 * 3530: "Network File System (NFS) version 4 Protocol" and section
75 * 20 of RFC 5661: "Network File System (NFS) Version 4 Minor Version
76 * 1 Protocol"
77 */
78
79static void encode_uint32(struct xdr_stream *xdr, u32 n)
80{
81 WARN_ON_ONCE(xdr_stream_encode_u32(xdr, n) < 0);
82}
83
84static void encode_bitmap4(struct xdr_stream *xdr, const __u32 *bitmap,
85 size_t len)
86{
87 xdr_stream_encode_uint32_array(xdr, array: bitmap, array_size: len);
88}
89
90static int decode_cb_fattr4(struct xdr_stream *xdr, uint32_t *bitmap,
91 struct nfs4_cb_fattr *fattr)
92{
93 fattr->ncf_cb_change = 0;
94 fattr->ncf_cb_fsize = 0;
95 fattr->ncf_cb_atime.tv_sec = 0;
96 fattr->ncf_cb_atime.tv_nsec = 0;
97 fattr->ncf_cb_mtime.tv_sec = 0;
98 fattr->ncf_cb_mtime.tv_nsec = 0;
99
100 if (bitmap[0] & FATTR4_WORD0_CHANGE)
101 if (xdr_stream_decode_u64(xdr, ptr: &fattr->ncf_cb_change) < 0)
102 return -EIO;
103 if (bitmap[0] & FATTR4_WORD0_SIZE)
104 if (xdr_stream_decode_u64(xdr, ptr: &fattr->ncf_cb_fsize) < 0)
105 return -EIO;
106 if (bitmap[2] & FATTR4_WORD2_TIME_DELEG_ACCESS) {
107 fattr4_time_deleg_access access;
108
109 if (!xdrgen_decode_fattr4_time_deleg_access(xdr, ptr: &access))
110 return -EIO;
111 fattr->ncf_cb_atime.tv_sec = access.seconds;
112 fattr->ncf_cb_atime.tv_nsec = access.nseconds;
113
114 }
115 if (bitmap[2] & FATTR4_WORD2_TIME_DELEG_MODIFY) {
116 fattr4_time_deleg_modify modify;
117
118 if (!xdrgen_decode_fattr4_time_deleg_modify(xdr, ptr: &modify))
119 return -EIO;
120 fattr->ncf_cb_mtime.tv_sec = modify.seconds;
121 fattr->ncf_cb_mtime.tv_nsec = modify.nseconds;
122
123 }
124 return 0;
125}
126
127static void encode_nfs_cb_opnum4(struct xdr_stream *xdr, enum nfs_cb_opnum4 op)
128{
129 __be32 *p;
130
131 p = xdr_reserve_space(xdr, nbytes: 4);
132 *p = cpu_to_be32(op);
133}
134
135/*
136 * nfs_fh4
137 *
138 * typedef opaque nfs_fh4<NFS4_FHSIZE>;
139 */
140static void encode_nfs_fh4(struct xdr_stream *xdr, const struct knfsd_fh *fh)
141{
142 u32 length = fh->fh_size;
143 __be32 *p;
144
145 BUG_ON(length > NFS4_FHSIZE);
146 p = xdr_reserve_space(xdr, nbytes: 4 + length);
147 xdr_encode_opaque(p, ptr: &fh->fh_raw, len: length);
148}
149
150/*
151 * stateid4
152 *
153 * struct stateid4 {
154 * uint32_t seqid;
155 * opaque other[12];
156 * };
157 */
158static void encode_stateid4(struct xdr_stream *xdr, const stateid_t *sid)
159{
160 __be32 *p;
161
162 p = xdr_reserve_space(xdr, NFS4_STATEID_SIZE);
163 *p++ = cpu_to_be32(sid->si_generation);
164 xdr_encode_opaque_fixed(p, ptr: &sid->si_opaque, NFS4_STATEID_OTHER_SIZE);
165}
166
167/*
168 * sessionid4
169 *
170 * typedef opaque sessionid4[NFS4_SESSIONID_SIZE];
171 */
172static void encode_sessionid4(struct xdr_stream *xdr,
173 const struct nfsd4_session *session)
174{
175 __be32 *p;
176
177 p = xdr_reserve_space(xdr, NFS4_MAX_SESSIONID_LEN);
178 xdr_encode_opaque_fixed(p, ptr: session->se_sessionid.data,
179 NFS4_MAX_SESSIONID_LEN);
180}
181
182/*
183 * nfsstat4
184 */
185static const struct {
186 int stat;
187 int errno;
188} nfs_cb_errtbl[] = {
189 { NFS4_OK, 0 },
190 { NFS4ERR_PERM, -EPERM },
191 { NFS4ERR_NOENT, -ENOENT },
192 { NFS4ERR_IO, -EIO },
193 { NFS4ERR_NXIO, -ENXIO },
194 { NFS4ERR_ACCESS, -EACCES },
195 { NFS4ERR_EXIST, -EEXIST },
196 { NFS4ERR_XDEV, -EXDEV },
197 { NFS4ERR_NOTDIR, -ENOTDIR },
198 { NFS4ERR_ISDIR, -EISDIR },
199 { NFS4ERR_INVAL, -EINVAL },
200 { NFS4ERR_FBIG, -EFBIG },
201 { NFS4ERR_NOSPC, -ENOSPC },
202 { NFS4ERR_ROFS, -EROFS },
203 { NFS4ERR_MLINK, -EMLINK },
204 { NFS4ERR_NAMETOOLONG, -ENAMETOOLONG },
205 { NFS4ERR_NOTEMPTY, -ENOTEMPTY },
206 { NFS4ERR_DQUOT, -EDQUOT },
207 { NFS4ERR_STALE, -ESTALE },
208 { NFS4ERR_BADHANDLE, -EBADHANDLE },
209 { NFS4ERR_BAD_COOKIE, -EBADCOOKIE },
210 { NFS4ERR_NOTSUPP, -ENOTSUPP },
211 { NFS4ERR_TOOSMALL, -ETOOSMALL },
212 { NFS4ERR_SERVERFAULT, -ESERVERFAULT },
213 { NFS4ERR_BADTYPE, -EBADTYPE },
214 { NFS4ERR_LOCKED, -EAGAIN },
215 { NFS4ERR_RESOURCE, -EREMOTEIO },
216 { NFS4ERR_SYMLINK, -ELOOP },
217 { NFS4ERR_OP_ILLEGAL, -EOPNOTSUPP },
218 { NFS4ERR_DEADLOCK, -EDEADLK },
219 { -1, -EIO }
220};
221
222/*
223 * If we cannot translate the error, the recovery routines should
224 * handle it.
225 *
226 * Note: remaining NFSv4 error codes have values > 10000, so should
227 * not conflict with native Linux error codes.
228 */
229static int nfs_cb_stat_to_errno(int status)
230{
231 int i;
232
233 for (i = 0; nfs_cb_errtbl[i].stat != -1; i++) {
234 if (nfs_cb_errtbl[i].stat == status)
235 return nfs_cb_errtbl[i].errno;
236 }
237
238 dprintk("NFSD: Unrecognized NFS CB status value: %u\n", status);
239 return -status;
240}
241
242static int decode_cb_op_status(struct xdr_stream *xdr,
243 enum nfs_cb_opnum4 expected, int *status)
244{
245 __be32 *p;
246 u32 op;
247
248 p = xdr_inline_decode(xdr, nbytes: 4 + 4);
249 if (unlikely(p == NULL))
250 goto out_overflow;
251 op = be32_to_cpup(p: p++);
252 if (unlikely(op != expected))
253 goto out_unexpected;
254 *status = nfs_cb_stat_to_errno(be32_to_cpup(p));
255 return 0;
256out_overflow:
257 return -EIO;
258out_unexpected:
259 dprintk("NFSD: Callback server returned operation %d but "
260 "we issued a request for %d\n", op, expected);
261 return -EIO;
262}
263
264/*
265 * CB_COMPOUND4args
266 *
267 * struct CB_COMPOUND4args {
268 * utf8str_cs tag;
269 * uint32_t minorversion;
270 * uint32_t callback_ident;
271 * nfs_cb_argop4 argarray<>;
272 * };
273*/
274static void encode_cb_compound4args(struct xdr_stream *xdr,
275 struct nfs4_cb_compound_hdr *hdr)
276{
277 __be32 * p;
278
279 p = xdr_reserve_space(xdr, nbytes: 4 + 4 + 4 + 4);
280 p = xdr_encode_empty_array(p); /* empty tag */
281 *p++ = cpu_to_be32(hdr->minorversion);
282 *p++ = cpu_to_be32(hdr->ident);
283
284 hdr->nops_p = p;
285 *p = cpu_to_be32(hdr->nops); /* argarray element count */
286}
287
288/*
289 * Update argarray element count
290 */
291static void encode_cb_nops(struct nfs4_cb_compound_hdr *hdr)
292{
293 BUG_ON(hdr->nops > NFS4_MAX_BACK_CHANNEL_OPS);
294 *hdr->nops_p = cpu_to_be32(hdr->nops);
295}
296
297/*
298 * CB_COMPOUND4res
299 *
300 * struct CB_COMPOUND4res {
301 * nfsstat4 status;
302 * utf8str_cs tag;
303 * nfs_cb_resop4 resarray<>;
304 * };
305 */
306static int decode_cb_compound4res(struct xdr_stream *xdr,
307 struct nfs4_cb_compound_hdr *hdr)
308{
309 u32 length;
310 __be32 *p;
311
312 p = xdr_inline_decode(xdr, XDR_UNIT);
313 if (unlikely(p == NULL))
314 goto out_overflow;
315 hdr->status = be32_to_cpup(p);
316 /* Ignore the tag */
317 if (xdr_stream_decode_u32(xdr, ptr: &length) < 0)
318 goto out_overflow;
319 if (xdr_inline_decode(xdr, nbytes: length) == NULL)
320 goto out_overflow;
321 if (xdr_stream_decode_u32(xdr, ptr: &hdr->nops) < 0)
322 goto out_overflow;
323 return 0;
324out_overflow:
325 return -EIO;
326}
327
328/*
329 * CB_RECALL4args
330 *
331 * struct CB_RECALL4args {
332 * stateid4 stateid;
333 * bool truncate;
334 * nfs_fh4 fh;
335 * };
336 */
337static void encode_cb_recall4args(struct xdr_stream *xdr,
338 const struct nfs4_delegation *dp,
339 struct nfs4_cb_compound_hdr *hdr)
340{
341 __be32 *p;
342
343 encode_nfs_cb_opnum4(xdr, op: OP_CB_RECALL);
344 encode_stateid4(xdr, sid: &dp->dl_stid.sc_stateid);
345
346 p = xdr_reserve_space(xdr, nbytes: 4);
347 *p++ = xdr_zero; /* truncate */
348
349 encode_nfs_fh4(xdr, fh: &dp->dl_stid.sc_file->fi_fhandle);
350
351 hdr->nops++;
352}
353
354/*
355 * CB_RECALLANY4args
356 *
357 * struct CB_RECALLANY4args {
358 * uint32_t craa_objects_to_keep;
359 * bitmap4 craa_type_mask;
360 * };
361 */
362static void
363encode_cb_recallany4args(struct xdr_stream *xdr,
364 struct nfs4_cb_compound_hdr *hdr, struct nfsd4_cb_recall_any *ra)
365{
366 encode_nfs_cb_opnum4(xdr, op: OP_CB_RECALL_ANY);
367 encode_uint32(xdr, n: ra->ra_keep);
368 encode_bitmap4(xdr, bitmap: ra->ra_bmval, ARRAY_SIZE(ra->ra_bmval));
369 hdr->nops++;
370}
371
372/*
373 * CB_GETATTR4args
374 * struct CB_GETATTR4args {
375 * nfs_fh4 fh;
376 * bitmap4 attr_request;
377 * };
378 *
379 * The size and change attributes are the only one
380 * guaranteed to be serviced by the client.
381 */
382static void
383encode_cb_getattr4args(struct xdr_stream *xdr, struct nfs4_cb_compound_hdr *hdr,
384 struct nfs4_cb_fattr *fattr)
385{
386 struct nfs4_delegation *dp = container_of(fattr, struct nfs4_delegation, dl_cb_fattr);
387 struct knfsd_fh *fh = &dp->dl_stid.sc_file->fi_fhandle;
388 struct nfs4_cb_fattr *ncf = &dp->dl_cb_fattr;
389 u32 bmap_size = 1;
390 u32 bmap[3];
391
392 bmap[0] = FATTR4_WORD0_SIZE;
393 if (!ncf->ncf_file_modified)
394 bmap[0] |= FATTR4_WORD0_CHANGE;
395
396 if (deleg_attrs_deleg(dl_type: dp->dl_type)) {
397 bmap[1] = 0;
398 bmap[2] = FATTR4_WORD2_TIME_DELEG_ACCESS | FATTR4_WORD2_TIME_DELEG_MODIFY;
399 bmap_size = 3;
400 }
401 encode_nfs_cb_opnum4(xdr, op: OP_CB_GETATTR);
402 encode_nfs_fh4(xdr, fh);
403 encode_bitmap4(xdr, bitmap: bmap, len: bmap_size);
404 hdr->nops++;
405}
406
407static u32 highest_slotid(struct nfsd4_session *ses)
408{
409 u32 idx;
410
411 spin_lock(lock: &ses->se_lock);
412 idx = fls(x: ~ses->se_cb_slot_avail);
413 if (idx > 0)
414 --idx;
415 idx = max(idx, ses->se_cb_highest_slot);
416 spin_unlock(lock: &ses->se_lock);
417 return idx;
418}
419
420static void
421encode_referring_call4(struct xdr_stream *xdr,
422 const struct nfsd4_referring_call *rc)
423{
424 encode_uint32(xdr, n: rc->rc_sequenceid);
425 encode_uint32(xdr, n: rc->rc_slotid);
426}
427
428static void
429encode_referring_call_list4(struct xdr_stream *xdr,
430 const struct nfsd4_referring_call_list *rcl)
431{
432 struct nfsd4_referring_call *rc;
433 __be32 *p;
434
435 p = xdr_reserve_space(xdr, NFS4_MAX_SESSIONID_LEN);
436 xdr_encode_opaque_fixed(p, ptr: rcl->rcl_sessionid.data,
437 NFS4_MAX_SESSIONID_LEN);
438 encode_uint32(xdr, n: rcl->__nr_referring_calls);
439 list_for_each_entry(rc, &rcl->rcl_referring_calls, __list)
440 encode_referring_call4(xdr, rc);
441}
442
443/*
444 * CB_SEQUENCE4args
445 *
446 * struct CB_SEQUENCE4args {
447 * sessionid4 csa_sessionid;
448 * sequenceid4 csa_sequenceid;
449 * slotid4 csa_slotid;
450 * slotid4 csa_highest_slotid;
451 * bool csa_cachethis;
452 * referring_call_list4 csa_referring_call_lists<>;
453 * };
454 */
455static void encode_cb_sequence4args(struct xdr_stream *xdr,
456 const struct nfsd4_callback *cb,
457 struct nfs4_cb_compound_hdr *hdr)
458{
459 struct nfsd4_session *session = cb->cb_clp->cl_cb_session;
460 struct nfsd4_referring_call_list *rcl;
461 __be32 *p;
462
463 if (hdr->minorversion == 0)
464 return;
465
466 encode_nfs_cb_opnum4(xdr, op: OP_CB_SEQUENCE);
467 encode_sessionid4(xdr, session);
468
469 p = xdr_reserve_space(xdr, XDR_UNIT * 4);
470 *p++ = cpu_to_be32(session->se_cb_seq_nr[cb->cb_held_slot]); /* csa_sequenceid */
471 *p++ = cpu_to_be32(cb->cb_held_slot); /* csa_slotid */
472 *p++ = cpu_to_be32(highest_slotid(session)); /* csa_highest_slotid */
473 *p++ = xdr_zero; /* csa_cachethis */
474
475 /* csa_referring_call_lists */
476 encode_uint32(xdr, n: cb->cb_nr_referring_call_list);
477 list_for_each_entry(rcl, &cb->cb_referring_call_list, __list)
478 encode_referring_call_list4(xdr, rcl);
479
480 hdr->nops++;
481}
482
483static void update_cb_slot_table(struct nfsd4_session *ses, u32 target)
484{
485 /* No need to do anything if nothing changed */
486 if (likely(target == READ_ONCE(ses->se_cb_highest_slot)))
487 return;
488
489 spin_lock(lock: &ses->se_lock);
490 if (target > ses->se_cb_highest_slot) {
491 int i;
492
493 target = min(target, NFSD_BC_SLOT_TABLE_SIZE - 1);
494
495 /*
496 * Growing the slot table. Reset any new sequences to 1.
497 *
498 * NB: There is some debate about whether the RFC requires this,
499 * but the Linux client expects it.
500 */
501 for (i = ses->se_cb_highest_slot + 1; i <= target; ++i)
502 ses->se_cb_seq_nr[i] = 1;
503 }
504 ses->se_cb_highest_slot = target;
505 spin_unlock(lock: &ses->se_lock);
506}
507
508/*
509 * CB_SEQUENCE4resok
510 *
511 * struct CB_SEQUENCE4resok {
512 * sessionid4 csr_sessionid;
513 * sequenceid4 csr_sequenceid;
514 * slotid4 csr_slotid;
515 * slotid4 csr_highest_slotid;
516 * slotid4 csr_target_highest_slotid;
517 * };
518 *
519 * union CB_SEQUENCE4res switch (nfsstat4 csr_status) {
520 * case NFS4_OK:
521 * CB_SEQUENCE4resok csr_resok4;
522 * default:
523 * void;
524 * };
525 *
526 * Our current back channel implmentation supports a single backchannel
527 * with a single slot.
528 */
529static int decode_cb_sequence4resok(struct xdr_stream *xdr,
530 struct nfsd4_callback *cb)
531{
532 struct nfsd4_session *session = cb->cb_clp->cl_cb_session;
533 int status = -ESERVERFAULT;
534 __be32 *p;
535 u32 seqid, slotid, target;
536
537 /*
538 * If the server returns different values for sessionID, slotID or
539 * sequence number, the server is looney tunes.
540 */
541 p = xdr_inline_decode(xdr, NFS4_MAX_SESSIONID_LEN + 4 + 4 + 4 + 4);
542 if (unlikely(p == NULL))
543 goto out_overflow;
544
545 if (memcmp(p, q: session->se_sessionid.data, NFS4_MAX_SESSIONID_LEN)) {
546 dprintk("NFS: %s Invalid session id\n", __func__);
547 goto out;
548 }
549 p += XDR_QUADLEN(NFS4_MAX_SESSIONID_LEN);
550
551 seqid = be32_to_cpup(p: p++);
552 if (seqid != session->se_cb_seq_nr[cb->cb_held_slot]) {
553 dprintk("NFS: %s Invalid sequence number\n", __func__);
554 goto out;
555 }
556
557 slotid = be32_to_cpup(p: p++);
558 if (slotid != cb->cb_held_slot) {
559 dprintk("NFS: %s Invalid slotid\n", __func__);
560 goto out;
561 }
562
563 p++; // ignore current highest slot value
564
565 target = be32_to_cpup(p: p++);
566 update_cb_slot_table(ses: session, target);
567 status = 0;
568out:
569 cb->cb_seq_status = status;
570 return status;
571out_overflow:
572 status = -EIO;
573 goto out;
574}
575
576static int decode_cb_sequence4res(struct xdr_stream *xdr,
577 struct nfsd4_callback *cb)
578{
579 int status;
580
581 if (cb->cb_clp->cl_minorversion == 0)
582 return 0;
583
584 status = decode_cb_op_status(xdr, expected: OP_CB_SEQUENCE, status: &cb->cb_seq_status);
585 if (unlikely(status || cb->cb_seq_status))
586 return status;
587
588 return decode_cb_sequence4resok(xdr, cb);
589}
590
591/*
592 * NFSv4.0 and NFSv4.1 XDR encode functions
593 *
594 * NFSv4.0 callback argument types are defined in section 15 of RFC
595 * 3530: "Network File System (NFS) version 4 Protocol" and section 20
596 * of RFC 5661: "Network File System (NFS) Version 4 Minor Version 1
597 * Protocol".
598 */
599
600/*
601 * NB: Without this zero space reservation, callbacks over krb5p fail
602 */
603static void nfs4_xdr_enc_cb_null(struct rpc_rqst *req, struct xdr_stream *xdr,
604 const void *__unused)
605{
606 xdr_reserve_space(xdr, nbytes: 0);
607}
608
609/*
610 * 20.1. Operation 3: CB_GETATTR - Get Attributes
611 */
612static void nfs4_xdr_enc_cb_getattr(struct rpc_rqst *req,
613 struct xdr_stream *xdr, const void *data)
614{
615 const struct nfsd4_callback *cb = data;
616 struct nfs4_cb_fattr *ncf =
617 container_of(cb, struct nfs4_cb_fattr, ncf_getattr);
618 struct nfs4_cb_compound_hdr hdr = {
619 .ident = cb->cb_clp->cl_cb_ident,
620 .minorversion = cb->cb_clp->cl_minorversion,
621 };
622
623 encode_cb_compound4args(xdr, hdr: &hdr);
624 encode_cb_sequence4args(xdr, cb, hdr: &hdr);
625 encode_cb_getattr4args(xdr, hdr: &hdr, fattr: ncf);
626 encode_cb_nops(hdr: &hdr);
627}
628
629/*
630 * 20.2. Operation 4: CB_RECALL - Recall a Delegation
631 */
632static void nfs4_xdr_enc_cb_recall(struct rpc_rqst *req, struct xdr_stream *xdr,
633 const void *data)
634{
635 const struct nfsd4_callback *cb = data;
636 const struct nfs4_delegation *dp = cb_to_delegation(cb);
637 struct nfs4_cb_compound_hdr hdr = {
638 .ident = cb->cb_clp->cl_cb_ident,
639 .minorversion = cb->cb_clp->cl_minorversion,
640 };
641
642 encode_cb_compound4args(xdr, hdr: &hdr);
643 encode_cb_sequence4args(xdr, cb, hdr: &hdr);
644 encode_cb_recall4args(xdr, dp, hdr: &hdr);
645 encode_cb_nops(hdr: &hdr);
646}
647
648/*
649 * 20.6. Operation 8: CB_RECALL_ANY - Keep Any N Recallable Objects
650 */
651static void
652nfs4_xdr_enc_cb_recall_any(struct rpc_rqst *req,
653 struct xdr_stream *xdr, const void *data)
654{
655 const struct nfsd4_callback *cb = data;
656 struct nfsd4_cb_recall_any *ra;
657 struct nfs4_cb_compound_hdr hdr = {
658 .ident = cb->cb_clp->cl_cb_ident,
659 .minorversion = cb->cb_clp->cl_minorversion,
660 };
661
662 ra = container_of(cb, struct nfsd4_cb_recall_any, ra_cb);
663 encode_cb_compound4args(xdr, hdr: &hdr);
664 encode_cb_sequence4args(xdr, cb, hdr: &hdr);
665 encode_cb_recallany4args(xdr, hdr: &hdr, ra);
666 encode_cb_nops(hdr: &hdr);
667}
668
669/*
670 * NFSv4.0 and NFSv4.1 XDR decode functions
671 *
672 * NFSv4.0 callback result types are defined in section 15 of RFC
673 * 3530: "Network File System (NFS) version 4 Protocol" and section 20
674 * of RFC 5661: "Network File System (NFS) Version 4 Minor Version 1
675 * Protocol".
676 */
677
678static int nfs4_xdr_dec_cb_null(struct rpc_rqst *req, struct xdr_stream *xdr,
679 void *__unused)
680{
681 return 0;
682}
683
684/*
685 * 20.1. Operation 3: CB_GETATTR - Get Attributes
686 */
687static int nfs4_xdr_dec_cb_getattr(struct rpc_rqst *rqstp,
688 struct xdr_stream *xdr,
689 void *data)
690{
691 struct nfsd4_callback *cb = data;
692 struct nfs4_cb_compound_hdr hdr;
693 int status;
694 u32 bitmap[3] = {0};
695 u32 attrlen, maxlen;
696 struct nfs4_cb_fattr *ncf =
697 container_of(cb, struct nfs4_cb_fattr, ncf_getattr);
698
699 status = decode_cb_compound4res(xdr, hdr: &hdr);
700 if (unlikely(status))
701 return status;
702
703 status = decode_cb_sequence4res(xdr, cb);
704 if (unlikely(status || cb->cb_seq_status))
705 return status;
706
707 status = decode_cb_op_status(xdr, expected: OP_CB_GETATTR, status: &cb->cb_status);
708 if (unlikely(status || cb->cb_status))
709 return status;
710 if (xdr_stream_decode_uint32_array(xdr, array: bitmap, array_size: 3) < 0)
711 return -EIO;
712 if (xdr_stream_decode_u32(xdr, ptr: &attrlen) < 0)
713 return -EIO;
714 maxlen = sizeof(ncf->ncf_cb_change) + sizeof(ncf->ncf_cb_fsize);
715 if (bitmap[2] != 0)
716 maxlen += (sizeof(ncf->ncf_cb_mtime.tv_sec) +
717 sizeof(ncf->ncf_cb_mtime.tv_nsec)) * 2;
718 if (attrlen > maxlen)
719 return -EIO;
720 status = decode_cb_fattr4(xdr, bitmap, fattr: ncf);
721 return status;
722}
723
724/*
725 * 20.2. Operation 4: CB_RECALL - Recall a Delegation
726 */
727static int nfs4_xdr_dec_cb_recall(struct rpc_rqst *rqstp,
728 struct xdr_stream *xdr,
729 void *data)
730{
731 struct nfsd4_callback *cb = data;
732 struct nfs4_cb_compound_hdr hdr;
733 int status;
734
735 status = decode_cb_compound4res(xdr, hdr: &hdr);
736 if (unlikely(status))
737 return status;
738
739 status = decode_cb_sequence4res(xdr, cb);
740 if (unlikely(status || cb->cb_seq_status))
741 return status;
742
743 return decode_cb_op_status(xdr, expected: OP_CB_RECALL, status: &cb->cb_status);
744}
745
746/*
747 * 20.6. Operation 8: CB_RECALL_ANY - Keep Any N Recallable Objects
748 */
749static int
750nfs4_xdr_dec_cb_recall_any(struct rpc_rqst *rqstp,
751 struct xdr_stream *xdr,
752 void *data)
753{
754 struct nfsd4_callback *cb = data;
755 struct nfs4_cb_compound_hdr hdr;
756 int status;
757
758 status = decode_cb_compound4res(xdr, hdr: &hdr);
759 if (unlikely(status))
760 return status;
761 status = decode_cb_sequence4res(xdr, cb);
762 if (unlikely(status || cb->cb_seq_status))
763 return status;
764 status = decode_cb_op_status(xdr, expected: OP_CB_RECALL_ANY, status: &cb->cb_status);
765 return status;
766}
767
768#ifdef CONFIG_NFSD_PNFS
769/*
770 * CB_LAYOUTRECALL4args
771 *
772 * struct layoutrecall_file4 {
773 * nfs_fh4 lor_fh;
774 * offset4 lor_offset;
775 * length4 lor_length;
776 * stateid4 lor_stateid;
777 * };
778 *
779 * union layoutrecall4 switch(layoutrecall_type4 lor_recalltype) {
780 * case LAYOUTRECALL4_FILE:
781 * layoutrecall_file4 lor_layout;
782 * case LAYOUTRECALL4_FSID:
783 * fsid4 lor_fsid;
784 * case LAYOUTRECALL4_ALL:
785 * void;
786 * };
787 *
788 * struct CB_LAYOUTRECALL4args {
789 * layouttype4 clora_type;
790 * layoutiomode4 clora_iomode;
791 * bool clora_changed;
792 * layoutrecall4 clora_recall;
793 * };
794 */
795static void encode_cb_layout4args(struct xdr_stream *xdr,
796 const struct nfs4_layout_stateid *ls,
797 struct nfs4_cb_compound_hdr *hdr)
798{
799 __be32 *p;
800
801 BUG_ON(hdr->minorversion == 0);
802
803 p = xdr_reserve_space(xdr, nbytes: 5 * 4);
804 *p++ = cpu_to_be32(OP_CB_LAYOUTRECALL);
805 *p++ = cpu_to_be32(ls->ls_layout_type);
806 *p++ = cpu_to_be32(IOMODE_ANY);
807 *p++ = cpu_to_be32(1);
808 *p = cpu_to_be32(RETURN_FILE);
809
810 encode_nfs_fh4(xdr, fh: &ls->ls_stid.sc_file->fi_fhandle);
811
812 p = xdr_reserve_space(xdr, nbytes: 2 * 8);
813 p = xdr_encode_hyper(p, val: 0);
814 xdr_encode_hyper(p, NFS4_MAX_UINT64);
815
816 encode_stateid4(xdr, sid: &ls->ls_recall_sid);
817
818 hdr->nops++;
819}
820
821static void nfs4_xdr_enc_cb_layout(struct rpc_rqst *req,
822 struct xdr_stream *xdr,
823 const void *data)
824{
825 const struct nfsd4_callback *cb = data;
826 const struct nfs4_layout_stateid *ls =
827 container_of(cb, struct nfs4_layout_stateid, ls_recall);
828 struct nfs4_cb_compound_hdr hdr = {
829 .ident = 0,
830 .minorversion = cb->cb_clp->cl_minorversion,
831 };
832
833 encode_cb_compound4args(xdr, hdr: &hdr);
834 encode_cb_sequence4args(xdr, cb, hdr: &hdr);
835 encode_cb_layout4args(xdr, ls, hdr: &hdr);
836 encode_cb_nops(hdr: &hdr);
837}
838
839static int nfs4_xdr_dec_cb_layout(struct rpc_rqst *rqstp,
840 struct xdr_stream *xdr,
841 void *data)
842{
843 struct nfsd4_callback *cb = data;
844 struct nfs4_cb_compound_hdr hdr;
845 int status;
846
847 status = decode_cb_compound4res(xdr, hdr: &hdr);
848 if (unlikely(status))
849 return status;
850
851 status = decode_cb_sequence4res(xdr, cb);
852 if (unlikely(status || cb->cb_seq_status))
853 return status;
854
855 return decode_cb_op_status(xdr, expected: OP_CB_LAYOUTRECALL, status: &cb->cb_status);
856}
857#endif /* CONFIG_NFSD_PNFS */
858
859static void encode_stateowner(struct xdr_stream *xdr, struct nfs4_stateowner *so)
860{
861 __be32 *p;
862
863 p = xdr_reserve_space(xdr, nbytes: 8 + 4 + so->so_owner.len);
864 p = xdr_encode_opaque_fixed(p, ptr: &so->so_client->cl_clientid, len: 8);
865 xdr_encode_opaque(p, ptr: so->so_owner.data, len: so->so_owner.len);
866}
867
868static void nfs4_xdr_enc_cb_notify_lock(struct rpc_rqst *req,
869 struct xdr_stream *xdr,
870 const void *data)
871{
872 const struct nfsd4_callback *cb = data;
873 const struct nfsd4_blocked_lock *nbl =
874 container_of(cb, struct nfsd4_blocked_lock, nbl_cb);
875 struct nfs4_lockowner *lo = (struct nfs4_lockowner *)nbl->nbl_lock.c.flc_owner;
876 struct nfs4_cb_compound_hdr hdr = {
877 .ident = 0,
878 .minorversion = cb->cb_clp->cl_minorversion,
879 };
880
881 __be32 *p;
882
883 BUG_ON(hdr.minorversion == 0);
884
885 encode_cb_compound4args(xdr, hdr: &hdr);
886 encode_cb_sequence4args(xdr, cb, hdr: &hdr);
887
888 p = xdr_reserve_space(xdr, nbytes: 4);
889 *p = cpu_to_be32(OP_CB_NOTIFY_LOCK);
890 encode_nfs_fh4(xdr, fh: &nbl->nbl_fh);
891 encode_stateowner(xdr, so: &lo->lo_owner);
892 hdr.nops++;
893
894 encode_cb_nops(hdr: &hdr);
895}
896
897static int nfs4_xdr_dec_cb_notify_lock(struct rpc_rqst *rqstp,
898 struct xdr_stream *xdr,
899 void *data)
900{
901 struct nfsd4_callback *cb = data;
902 struct nfs4_cb_compound_hdr hdr;
903 int status;
904
905 status = decode_cb_compound4res(xdr, hdr: &hdr);
906 if (unlikely(status))
907 return status;
908
909 status = decode_cb_sequence4res(xdr, cb);
910 if (unlikely(status || cb->cb_seq_status))
911 return status;
912
913 return decode_cb_op_status(xdr, expected: OP_CB_NOTIFY_LOCK, status: &cb->cb_status);
914}
915
916/*
917 * struct write_response4 {
918 * stateid4 wr_callback_id<1>;
919 * length4 wr_count;
920 * stable_how4 wr_committed;
921 * verifier4 wr_writeverf;
922 * };
923 * union offload_info4 switch (nfsstat4 coa_status) {
924 * case NFS4_OK:
925 * write_response4 coa_resok4;
926 * default:
927 * length4 coa_bytes_copied;
928 * };
929 * struct CB_OFFLOAD4args {
930 * nfs_fh4 coa_fh;
931 * stateid4 coa_stateid;
932 * offload_info4 coa_offload_info;
933 * };
934 */
935static void encode_offload_info4(struct xdr_stream *xdr,
936 const struct nfsd4_cb_offload *cbo)
937{
938 __be32 *p;
939
940 p = xdr_reserve_space(xdr, nbytes: 4);
941 *p = cbo->co_nfserr;
942 switch (cbo->co_nfserr) {
943 case nfs_ok:
944 p = xdr_reserve_space(xdr, nbytes: 4 + 8 + 4 + NFS4_VERIFIER_SIZE);
945 p = xdr_encode_empty_array(p);
946 p = xdr_encode_hyper(p, val: cbo->co_res.wr_bytes_written);
947 *p++ = cpu_to_be32(cbo->co_res.wr_stable_how);
948 p = xdr_encode_opaque_fixed(p, ptr: cbo->co_res.wr_verifier.data,
949 NFS4_VERIFIER_SIZE);
950 break;
951 default:
952 p = xdr_reserve_space(xdr, nbytes: 8);
953 /* We always return success if bytes were written */
954 p = xdr_encode_hyper(p, val: 0);
955 }
956}
957
958static void encode_cb_offload4args(struct xdr_stream *xdr,
959 const struct nfsd4_cb_offload *cbo,
960 struct nfs4_cb_compound_hdr *hdr)
961{
962 __be32 *p;
963
964 p = xdr_reserve_space(xdr, nbytes: 4);
965 *p = cpu_to_be32(OP_CB_OFFLOAD);
966 encode_nfs_fh4(xdr, fh: &cbo->co_fh);
967 encode_stateid4(xdr, sid: &cbo->co_res.cb_stateid);
968 encode_offload_info4(xdr, cbo);
969
970 hdr->nops++;
971}
972
973static void nfs4_xdr_enc_cb_offload(struct rpc_rqst *req,
974 struct xdr_stream *xdr,
975 const void *data)
976{
977 const struct nfsd4_callback *cb = data;
978 const struct nfsd4_cb_offload *cbo =
979 container_of(cb, struct nfsd4_cb_offload, co_cb);
980 struct nfs4_cb_compound_hdr hdr = {
981 .ident = 0,
982 .minorversion = cb->cb_clp->cl_minorversion,
983 };
984
985 encode_cb_compound4args(xdr, hdr: &hdr);
986 encode_cb_sequence4args(xdr, cb, hdr: &hdr);
987 encode_cb_offload4args(xdr, cbo, hdr: &hdr);
988 encode_cb_nops(hdr: &hdr);
989}
990
991static int nfs4_xdr_dec_cb_offload(struct rpc_rqst *rqstp,
992 struct xdr_stream *xdr,
993 void *data)
994{
995 struct nfsd4_callback *cb = data;
996 struct nfs4_cb_compound_hdr hdr;
997 int status;
998
999 status = decode_cb_compound4res(xdr, hdr: &hdr);
1000 if (unlikely(status))
1001 return status;
1002
1003 status = decode_cb_sequence4res(xdr, cb);
1004 if (unlikely(status || cb->cb_seq_status))
1005 return status;
1006
1007 return decode_cb_op_status(xdr, expected: OP_CB_OFFLOAD, status: &cb->cb_status);
1008}
1009/*
1010 * RPC procedure tables
1011 */
1012#define PROC(proc, call, argtype, restype) \
1013[NFSPROC4_CLNT_##proc] = { \
1014 .p_proc = NFSPROC4_CB_##call, \
1015 .p_encode = nfs4_xdr_enc_##argtype, \
1016 .p_decode = nfs4_xdr_dec_##restype, \
1017 .p_arglen = NFS4_enc_##argtype##_sz, \
1018 .p_replen = NFS4_dec_##restype##_sz, \
1019 .p_statidx = NFSPROC4_CB_##call, \
1020 .p_name = #proc, \
1021}
1022
1023static const struct rpc_procinfo nfs4_cb_procedures[] = {
1024 PROC(CB_NULL, NULL, cb_null, cb_null),
1025 PROC(CB_RECALL, COMPOUND, cb_recall, cb_recall),
1026#ifdef CONFIG_NFSD_PNFS
1027 PROC(CB_LAYOUT, COMPOUND, cb_layout, cb_layout),
1028#endif
1029 PROC(CB_NOTIFY_LOCK, COMPOUND, cb_notify_lock, cb_notify_lock),
1030 PROC(CB_OFFLOAD, COMPOUND, cb_offload, cb_offload),
1031 PROC(CB_RECALL_ANY, COMPOUND, cb_recall_any, cb_recall_any),
1032 PROC(CB_GETATTR, COMPOUND, cb_getattr, cb_getattr),
1033};
1034
1035static unsigned int nfs4_cb_counts[ARRAY_SIZE(nfs4_cb_procedures)];
1036static const struct rpc_version nfs_cb_version4 = {
1037/*
1038 * Note on the callback rpc program version number: despite language in rfc
1039 * 5661 section 18.36.3 requiring servers to use 4 in this field, the
1040 * official xdr descriptions for both 4.0 and 4.1 specify version 1, and
1041 * in practice that appears to be what implementations use. The section
1042 * 18.36.3 language is expected to be fixed in an erratum.
1043 */
1044 .number = 1,
1045 .nrprocs = ARRAY_SIZE(nfs4_cb_procedures),
1046 .procs = nfs4_cb_procedures,
1047 .counts = nfs4_cb_counts,
1048};
1049
1050static const struct rpc_version *nfs_cb_version[2] = {
1051 [1] = &nfs_cb_version4,
1052};
1053
1054static const struct rpc_program cb_program;
1055
1056static struct rpc_stat cb_stats = {
1057 .program = &cb_program
1058};
1059
1060#define NFS4_CALLBACK 0x40000000
1061static const struct rpc_program cb_program = {
1062 .name = "nfs4_cb",
1063 .number = NFS4_CALLBACK,
1064 .nrvers = ARRAY_SIZE(nfs_cb_version),
1065 .version = nfs_cb_version,
1066 .stats = &cb_stats,
1067 .pipe_dir_name = "nfsd4_cb",
1068};
1069
1070static int max_cb_time(struct net *net)
1071{
1072 struct nfsd_net *nn = net_generic(net, id: nfsd_net_id);
1073
1074 /*
1075 * nfsd4_lease is set to at most one hour in __nfsd4_write_time,
1076 * so we can use 32-bit math on it. Warn if that assumption
1077 * ever stops being true.
1078 */
1079 if (WARN_ON_ONCE(nn->nfsd4_lease > 3600))
1080 return 360 * HZ;
1081
1082 return max(((u32)nn->nfsd4_lease)/10, 1u) * HZ;
1083}
1084
1085static bool nfsd4_queue_cb(struct nfsd4_callback *cb)
1086{
1087 struct nfs4_client *clp = cb->cb_clp;
1088
1089 trace_nfsd_cb_queue(clp, cb);
1090 return queue_work(wq: clp->cl_callback_wq, work: &cb->cb_work);
1091}
1092
1093static void nfsd4_requeue_cb(struct rpc_task *task, struct nfsd4_callback *cb)
1094{
1095 struct nfs4_client *clp = cb->cb_clp;
1096
1097 if (!test_bit(NFSD4_CLIENT_CB_KILL, &clp->cl_flags)) {
1098 trace_nfsd_cb_restart(clp, cb);
1099 task->tk_status = 0;
1100 set_bit(NFSD4_CALLBACK_REQUEUE, addr: &cb->cb_flags);
1101 }
1102}
1103
1104static void nfsd41_cb_inflight_begin(struct nfs4_client *clp)
1105{
1106 atomic_inc(v: &clp->cl_cb_inflight);
1107}
1108
1109static void nfsd41_cb_inflight_end(struct nfs4_client *clp)
1110{
1111
1112 atomic_dec_and_wake_up(var: &clp->cl_cb_inflight);
1113}
1114
1115static void nfsd41_cb_inflight_wait_complete(struct nfs4_client *clp)
1116{
1117 wait_var_event(&clp->cl_cb_inflight,
1118 !atomic_read(&clp->cl_cb_inflight));
1119}
1120
1121static const struct cred *get_backchannel_cred(struct nfs4_client *clp, struct rpc_clnt *client, struct nfsd4_session *ses)
1122{
1123 if (clp->cl_minorversion == 0) {
1124 client->cl_principal = clp->cl_cred.cr_targ_princ ?
1125 clp->cl_cred.cr_targ_princ : "nfs";
1126
1127 return get_cred(cred: rpc_machine_cred());
1128 } else {
1129 struct cred *kcred;
1130
1131 kcred = prepare_kernel_cred(&init_task);
1132 if (!kcred)
1133 return NULL;
1134
1135 kcred->fsuid = ses->se_cb_sec.uid;
1136 kcred->fsgid = ses->se_cb_sec.gid;
1137 return kcred;
1138 }
1139}
1140
1141static int setup_callback_client(struct nfs4_client *clp, struct nfs4_cb_conn *conn, struct nfsd4_session *ses)
1142{
1143 int maxtime = max_cb_time(net: clp->net);
1144 struct rpc_timeout timeparms = {
1145 .to_initval = maxtime,
1146 .to_retries = 0,
1147 .to_maxval = maxtime,
1148 };
1149 struct rpc_create_args args = {
1150 .net = clp->net,
1151 .address = (struct sockaddr *) &conn->cb_addr,
1152 .addrsize = conn->cb_addrlen,
1153 .saddress = (struct sockaddr *) &conn->cb_saddr,
1154 .timeout = &timeparms,
1155 .program = &cb_program,
1156 .version = 1,
1157 .flags = (RPC_CLNT_CREATE_NOPING | RPC_CLNT_CREATE_QUIET),
1158 .cred = current_cred(),
1159 };
1160 struct rpc_clnt *client;
1161 const struct cred *cred;
1162
1163 if (clp->cl_minorversion == 0) {
1164 if (!clp->cl_cred.cr_principal &&
1165 (clp->cl_cred.cr_flavor >= RPC_AUTH_GSS_KRB5)) {
1166 trace_nfsd_cb_setup_err(clp, error: -EINVAL);
1167 return -EINVAL;
1168 }
1169 args.client_name = clp->cl_cred.cr_principal;
1170 args.prognumber = conn->cb_prog;
1171 args.protocol = XPRT_TRANSPORT_TCP;
1172 args.authflavor = clp->cl_cred.cr_flavor;
1173 clp->cl_cb_ident = conn->cb_ident;
1174 } else {
1175 if (!conn->cb_xprt || !ses)
1176 return -EINVAL;
1177 clp->cl_cb_session = ses;
1178 args.bc_xprt = conn->cb_xprt;
1179 args.prognumber = clp->cl_cb_session->se_cb_prog;
1180 args.protocol = conn->cb_xprt->xpt_class->xcl_ident |
1181 XPRT_TRANSPORT_BC;
1182 args.authflavor = ses->se_cb_sec.flavor;
1183 }
1184 /* Create RPC client */
1185 client = rpc_create(args: &args);
1186 if (IS_ERR(ptr: client)) {
1187 trace_nfsd_cb_setup_err(clp, error: PTR_ERR(ptr: client));
1188 return PTR_ERR(ptr: client);
1189 }
1190 cred = get_backchannel_cred(clp, client, ses);
1191 if (!cred) {
1192 trace_nfsd_cb_setup_err(clp, error: -ENOMEM);
1193 rpc_shutdown_client(client);
1194 return -ENOMEM;
1195 }
1196
1197 if (clp->cl_minorversion != 0)
1198 clp->cl_cb_conn.cb_xprt = conn->cb_xprt;
1199 clp->cl_cb_client = client;
1200 clp->cl_cb_cred = cred;
1201 rcu_read_lock();
1202 trace_nfsd_cb_setup(clp, netid: rpc_peeraddr2str(client, RPC_DISPLAY_NETID),
1203 authflavor: args.authflavor);
1204 rcu_read_unlock();
1205 return 0;
1206}
1207
1208static void nfsd4_mark_cb_state(struct nfs4_client *clp, int newstate)
1209{
1210 if (clp->cl_cb_state != newstate) {
1211 clp->cl_cb_state = newstate;
1212 trace_nfsd_cb_new_state(clp);
1213 }
1214}
1215
1216static void nfsd4_mark_cb_down(struct nfs4_client *clp)
1217{
1218 if (test_bit(NFSD4_CLIENT_CB_UPDATE, &clp->cl_flags))
1219 return;
1220 nfsd4_mark_cb_state(clp, NFSD4_CB_DOWN);
1221}
1222
1223static void nfsd4_mark_cb_fault(struct nfs4_client *clp)
1224{
1225 if (test_bit(NFSD4_CLIENT_CB_UPDATE, &clp->cl_flags))
1226 return;
1227 nfsd4_mark_cb_state(clp, NFSD4_CB_FAULT);
1228}
1229
1230static void nfsd4_cb_probe_done(struct rpc_task *task, void *calldata)
1231{
1232 struct nfs4_client *clp = container_of(calldata, struct nfs4_client, cl_cb_null);
1233
1234 if (task->tk_status)
1235 nfsd4_mark_cb_down(clp);
1236 else
1237 nfsd4_mark_cb_state(clp, NFSD4_CB_UP);
1238}
1239
1240static void nfsd4_cb_probe_release(void *calldata)
1241{
1242 struct nfs4_client *clp = container_of(calldata, struct nfs4_client, cl_cb_null);
1243
1244 nfsd41_cb_inflight_end(clp);
1245
1246}
1247
1248static const struct rpc_call_ops nfsd4_cb_probe_ops = {
1249 /* XXX: release method to ensure we set the cb channel down if
1250 * necessary on early failure? */
1251 .rpc_call_done = nfsd4_cb_probe_done,
1252 .rpc_release = nfsd4_cb_probe_release,
1253};
1254
1255/*
1256 * Poke the callback thread to process any updates to the callback
1257 * parameters, and send a null probe.
1258 */
1259void nfsd4_probe_callback(struct nfs4_client *clp)
1260{
1261 trace_nfsd_cb_probe(clp);
1262 nfsd4_mark_cb_state(clp, NFSD4_CB_UNKNOWN);
1263 set_bit(NFSD4_CLIENT_CB_UPDATE, addr: &clp->cl_flags);
1264 nfsd4_run_cb(cb: &clp->cl_cb_null);
1265}
1266
1267void nfsd4_probe_callback_sync(struct nfs4_client *clp)
1268{
1269 nfsd4_probe_callback(clp);
1270 flush_workqueue(clp->cl_callback_wq);
1271}
1272
1273void nfsd4_change_callback(struct nfs4_client *clp, struct nfs4_cb_conn *conn)
1274{
1275 nfsd4_mark_cb_state(clp, NFSD4_CB_UNKNOWN);
1276 spin_lock(lock: &clp->cl_lock);
1277 memcpy(&clp->cl_cb_conn, conn, sizeof(struct nfs4_cb_conn));
1278 spin_unlock(lock: &clp->cl_lock);
1279}
1280
1281static int grab_slot(struct nfsd4_session *ses)
1282{
1283 int idx;
1284
1285 spin_lock(lock: &ses->se_lock);
1286 idx = ffs(ses->se_cb_slot_avail) - 1;
1287 if (idx < 0 || idx > ses->se_cb_highest_slot) {
1288 spin_unlock(lock: &ses->se_lock);
1289 return -1;
1290 }
1291 /* clear the bit for the slot */
1292 ses->se_cb_slot_avail &= ~BIT(idx);
1293 spin_unlock(lock: &ses->se_lock);
1294 return idx;
1295}
1296
1297/*
1298 * There's currently a single callback channel slot.
1299 * If the slot is available, then mark it busy. Otherwise, set the
1300 * thread for sleeping on the callback RPC wait queue.
1301 */
1302static bool nfsd41_cb_get_slot(struct nfsd4_callback *cb, struct rpc_task *task)
1303{
1304 struct nfs4_client *clp = cb->cb_clp;
1305 struct nfsd4_session *ses = clp->cl_cb_session;
1306
1307 if (cb->cb_held_slot >= 0)
1308 return true;
1309 cb->cb_held_slot = grab_slot(ses);
1310 if (cb->cb_held_slot < 0) {
1311 rpc_sleep_on(&clp->cl_cb_waitq, task, NULL);
1312 /* Race breaker */
1313 cb->cb_held_slot = grab_slot(ses);
1314 if (cb->cb_held_slot < 0)
1315 return false;
1316 rpc_wake_up_queued_task(&clp->cl_cb_waitq, task);
1317 }
1318 return true;
1319}
1320
1321static void nfsd41_cb_release_slot(struct nfsd4_callback *cb)
1322{
1323 struct nfs4_client *clp = cb->cb_clp;
1324 struct nfsd4_session *ses = clp->cl_cb_session;
1325
1326 if (cb->cb_held_slot >= 0) {
1327 spin_lock(lock: &ses->se_lock);
1328 ses->se_cb_slot_avail |= BIT(cb->cb_held_slot);
1329 spin_unlock(lock: &ses->se_lock);
1330 cb->cb_held_slot = -1;
1331 rpc_wake_up_next(&clp->cl_cb_waitq);
1332 }
1333}
1334
1335static void nfsd41_destroy_cb(struct nfsd4_callback *cb)
1336{
1337 struct nfs4_client *clp = cb->cb_clp;
1338
1339 trace_nfsd_cb_destroy(clp, cb);
1340 nfsd41_cb_release_slot(cb);
1341 if (test_bit(NFSD4_CALLBACK_WAKE, &cb->cb_flags))
1342 clear_and_wake_up_bit(NFSD4_CALLBACK_RUNNING, word: &cb->cb_flags);
1343 else
1344 clear_bit(NFSD4_CALLBACK_RUNNING, addr: &cb->cb_flags);
1345
1346 if (cb->cb_ops && cb->cb_ops->release)
1347 cb->cb_ops->release(cb);
1348 nfsd41_cb_inflight_end(clp);
1349}
1350
1351/**
1352 * nfsd41_cb_referring_call - add a referring call to a callback operation
1353 * @cb: context of callback to add the rc to
1354 * @sessionid: referring call's session ID
1355 * @slotid: referring call's session slot index
1356 * @seqno: referring call's slot sequence number
1357 *
1358 * Caller serializes access to @cb.
1359 *
1360 * NB: If memory allocation fails, the referring call is not added.
1361 */
1362void nfsd41_cb_referring_call(struct nfsd4_callback *cb,
1363 struct nfs4_sessionid *sessionid,
1364 u32 slotid, u32 seqno)
1365{
1366 struct nfsd4_referring_call_list *rcl;
1367 struct nfsd4_referring_call *rc;
1368 bool found;
1369
1370 might_sleep();
1371
1372 found = false;
1373 list_for_each_entry(rcl, &cb->cb_referring_call_list, __list) {
1374 if (!memcmp(p: rcl->rcl_sessionid.data, q: sessionid->data,
1375 NFS4_MAX_SESSIONID_LEN)) {
1376 found = true;
1377 break;
1378 }
1379 }
1380 if (!found) {
1381 rcl = kmalloc(sizeof(*rcl), GFP_KERNEL);
1382 if (!rcl)
1383 return;
1384 memcpy(rcl->rcl_sessionid.data, sessionid->data,
1385 NFS4_MAX_SESSIONID_LEN);
1386 rcl->__nr_referring_calls = 0;
1387 INIT_LIST_HEAD(list: &rcl->rcl_referring_calls);
1388 list_add(new: &rcl->__list, head: &cb->cb_referring_call_list);
1389 cb->cb_nr_referring_call_list++;
1390 }
1391
1392 found = false;
1393 list_for_each_entry(rc, &rcl->rcl_referring_calls, __list) {
1394 if (rc->rc_sequenceid == seqno && rc->rc_slotid == slotid) {
1395 found = true;
1396 break;
1397 }
1398 }
1399 if (!found) {
1400 rc = kmalloc(sizeof(*rc), GFP_KERNEL);
1401 if (!rc)
1402 goto out;
1403 rc->rc_sequenceid = seqno;
1404 rc->rc_slotid = slotid;
1405 rcl->__nr_referring_calls++;
1406 list_add(new: &rc->__list, head: &rcl->rcl_referring_calls);
1407 }
1408
1409out:
1410 if (!rcl->__nr_referring_calls) {
1411 cb->cb_nr_referring_call_list--;
1412 list_del(entry: &rcl->__list);
1413 kfree(objp: rcl);
1414 }
1415}
1416
1417/**
1418 * nfsd41_cb_destroy_referring_call_list - release referring call info
1419 * @cb: context of a callback that has completed
1420 *
1421 * Callers who allocate referring calls using nfsd41_cb_referring_call() must
1422 * release those resources by calling nfsd41_cb_destroy_referring_call_list.
1423 *
1424 * Caller serializes access to @cb.
1425 */
1426void nfsd41_cb_destroy_referring_call_list(struct nfsd4_callback *cb)
1427{
1428 struct nfsd4_referring_call_list *rcl;
1429 struct nfsd4_referring_call *rc;
1430
1431 while (!list_empty(head: &cb->cb_referring_call_list)) {
1432 rcl = list_first_entry(&cb->cb_referring_call_list,
1433 struct nfsd4_referring_call_list,
1434 __list);
1435
1436 while (!list_empty(head: &rcl->rcl_referring_calls)) {
1437 rc = list_first_entry(&rcl->rcl_referring_calls,
1438 struct nfsd4_referring_call,
1439 __list);
1440 list_del(entry: &rc->__list);
1441 kfree(objp: rc);
1442 }
1443 list_del(entry: &rcl->__list);
1444 kfree(objp: rcl);
1445 }
1446}
1447
1448static void nfsd4_cb_prepare(struct rpc_task *task, void *calldata)
1449{
1450 struct nfsd4_callback *cb = calldata;
1451 struct nfs4_client *clp = cb->cb_clp;
1452 u32 minorversion = clp->cl_minorversion;
1453
1454 /*
1455 * cb_seq_status is only set in decode_cb_sequence4res,
1456 * and so will remain 1 if an rpc level failure occurs.
1457 */
1458 trace_nfsd_cb_rpc_prepare(clp);
1459 cb->cb_seq_status = 1;
1460 cb->cb_status = 0;
1461 if (minorversion && !nfsd41_cb_get_slot(cb, task))
1462 return;
1463 rpc_call_start(task);
1464}
1465
1466/* Returns true if CB_COMPOUND processing should continue */
1467static bool nfsd4_cb_sequence_done(struct rpc_task *task, struct nfsd4_callback *cb)
1468{
1469 struct nfsd4_session *session = cb->cb_clp->cl_cb_session;
1470 bool ret = false;
1471
1472 if (cb->cb_held_slot < 0)
1473 goto requeue;
1474
1475 /* This is the operation status code for CB_SEQUENCE */
1476 trace_nfsd_cb_seq_status(task, cb);
1477 switch (cb->cb_seq_status) {
1478 case 0:
1479 /*
1480 * No need for lock, access serialized in nfsd4_cb_prepare
1481 *
1482 * RFC5661 20.9.3
1483 * If CB_SEQUENCE returns an error, then the state of the slot
1484 * (sequence ID, cached reply) MUST NOT change.
1485 */
1486 ++session->se_cb_seq_nr[cb->cb_held_slot];
1487 ret = true;
1488 break;
1489 case -ESERVERFAULT:
1490 /*
1491 * Call succeeded, but the session, slot index, or slot
1492 * sequence number in the response do not match the same
1493 * in the server's call. The sequence information is thus
1494 * untrustworthy.
1495 */
1496 nfsd4_mark_cb_fault(clp: cb->cb_clp);
1497 break;
1498 case 1:
1499 /*
1500 * cb_seq_status remains 1 if an RPC Reply was never
1501 * received. NFSD can't know if the client processed
1502 * the CB_SEQUENCE operation. Ask the client to send a
1503 * DESTROY_SESSION to recover.
1504 */
1505 fallthrough;
1506 case -NFS4ERR_BADSESSION:
1507 nfsd4_mark_cb_fault(clp: cb->cb_clp);
1508 goto requeue;
1509 case -NFS4ERR_DELAY:
1510 cb->cb_seq_status = 1;
1511 if (RPC_SIGNALLED(task) || !rpc_restart_call(task))
1512 goto requeue;
1513 rpc_delay(task, 2 * HZ);
1514 return false;
1515 case -NFS4ERR_SEQ_MISORDERED:
1516 case -NFS4ERR_BADSLOT:
1517 /*
1518 * A SEQ_MISORDERED or BADSLOT error means that the client and
1519 * server are out of sync as to the backchannel parameters. Mark
1520 * the backchannel faulty and restart the RPC, but leak the slot
1521 * so that it's no longer used.
1522 */
1523 nfsd4_mark_cb_fault(clp: cb->cb_clp);
1524 cb->cb_held_slot = -1;
1525 goto retry_nowait;
1526 default:
1527 nfsd4_mark_cb_fault(clp: cb->cb_clp);
1528 }
1529 trace_nfsd_cb_free_slot(task, cb);
1530 nfsd41_cb_release_slot(cb);
1531 return ret;
1532retry_nowait:
1533 /*
1534 * RPC_SIGNALLED() means that the rpc_client is being torn down and
1535 * (possibly) recreated. Requeue the call in that case.
1536 */
1537 if (!RPC_SIGNALLED(task)) {
1538 if (rpc_restart_call_prepare(task))
1539 return false;
1540 }
1541requeue:
1542 nfsd41_cb_release_slot(cb);
1543 nfsd4_requeue_cb(task, cb);
1544 return false;
1545}
1546
1547static void nfsd4_cb_done(struct rpc_task *task, void *calldata)
1548{
1549 struct nfsd4_callback *cb = calldata;
1550 struct nfs4_client *clp = cb->cb_clp;
1551
1552 trace_nfsd_cb_rpc_done(clp);
1553
1554 if (!clp->cl_minorversion) {
1555 /*
1556 * If the backchannel connection was shut down while this
1557 * task was queued, we need to resubmit it after setting up
1558 * a new backchannel connection.
1559 *
1560 * Note that if we lost our callback connection permanently
1561 * the submission code will error out, so we don't need to
1562 * handle that case here.
1563 */
1564 if (RPC_SIGNALLED(task))
1565 nfsd4_requeue_cb(task, cb);
1566 } else if (!nfsd4_cb_sequence_done(task, cb)) {
1567 return;
1568 }
1569
1570 if (cb->cb_status) {
1571 WARN_ONCE(task->tk_status,
1572 "cb_status=%d tk_status=%d cb_opcode=%d",
1573 cb->cb_status, task->tk_status, cb->cb_ops->opcode);
1574 task->tk_status = cb->cb_status;
1575 }
1576
1577 switch (cb->cb_ops->done(cb, task)) {
1578 case 0:
1579 task->tk_status = 0;
1580 rpc_restart_call_prepare(task);
1581 return;
1582 case 1:
1583 switch (task->tk_status) {
1584 case -EIO:
1585 case -ETIMEDOUT:
1586 case -EACCES:
1587 nfsd4_mark_cb_down(clp);
1588 }
1589 break;
1590 default:
1591 BUG();
1592 }
1593}
1594
1595static void nfsd4_cb_release(void *calldata)
1596{
1597 struct nfsd4_callback *cb = calldata;
1598
1599 trace_nfsd_cb_rpc_release(clp: cb->cb_clp);
1600
1601 if (test_bit(NFSD4_CALLBACK_REQUEUE, &cb->cb_flags))
1602 nfsd4_queue_cb(cb);
1603 else
1604 nfsd41_destroy_cb(cb);
1605
1606}
1607
1608static const struct rpc_call_ops nfsd4_cb_ops = {
1609 .rpc_call_prepare = nfsd4_cb_prepare,
1610 .rpc_call_done = nfsd4_cb_done,
1611 .rpc_release = nfsd4_cb_release,
1612};
1613
1614/* must be called under the state lock */
1615void nfsd4_shutdown_callback(struct nfs4_client *clp)
1616{
1617 if (clp->cl_cb_state != NFSD4_CB_UNKNOWN)
1618 trace_nfsd_cb_shutdown(clp);
1619
1620 set_bit(NFSD4_CLIENT_CB_KILL, addr: &clp->cl_flags);
1621 /*
1622 * Note this won't actually result in a null callback;
1623 * instead, nfsd4_run_cb_null() will detect the killed
1624 * client, destroy the rpc client, and stop:
1625 */
1626 nfsd4_run_cb(cb: &clp->cl_cb_null);
1627 flush_workqueue(clp->cl_callback_wq);
1628 nfsd41_cb_inflight_wait_complete(clp);
1629}
1630
1631static struct nfsd4_conn * __nfsd4_find_backchannel(struct nfs4_client *clp)
1632{
1633 struct nfsd4_session *s;
1634 struct nfsd4_conn *c;
1635
1636 lockdep_assert_held(&clp->cl_lock);
1637
1638 list_for_each_entry(s, &clp->cl_sessions, se_perclnt) {
1639 list_for_each_entry(c, &s->se_conns, cn_persession) {
1640 if (c->cn_flags & NFS4_CDFC4_BACK)
1641 return c;
1642 }
1643 }
1644 return NULL;
1645}
1646
1647/*
1648 * Note there isn't a lot of locking in this code; instead we depend on
1649 * the fact that it is run from clp->cl_callback_wq, which won't run two
1650 * work items at once. So, for example, clp->cl_callback_wq handles all
1651 * access of cl_cb_client and all calls to rpc_create or rpc_shutdown_client.
1652 */
1653static void nfsd4_process_cb_update(struct nfsd4_callback *cb)
1654{
1655 struct nfs4_cb_conn conn;
1656 struct nfs4_client *clp = cb->cb_clp;
1657 struct nfsd4_session *ses = NULL;
1658 struct nfsd4_conn *c;
1659 int err;
1660
1661 trace_nfsd_cb_bc_update(clp, cb);
1662
1663 /*
1664 * This is either an update, or the client dying; in either case,
1665 * kill the old client:
1666 */
1667 if (clp->cl_cb_client) {
1668 trace_nfsd_cb_bc_shutdown(clp, cb);
1669 rpc_shutdown_client(clp->cl_cb_client);
1670 clp->cl_cb_client = NULL;
1671 put_cred(cred: clp->cl_cb_cred);
1672 clp->cl_cb_cred = NULL;
1673 }
1674 if (clp->cl_cb_conn.cb_xprt) {
1675 svc_xprt_put(xprt: clp->cl_cb_conn.cb_xprt);
1676 clp->cl_cb_conn.cb_xprt = NULL;
1677 }
1678 if (test_bit(NFSD4_CLIENT_CB_KILL, &clp->cl_flags))
1679 return;
1680
1681 spin_lock(lock: &clp->cl_lock);
1682 /*
1683 * Only serialized callback code is allowed to clear these
1684 * flags; main nfsd code can only set them:
1685 */
1686 WARN_ON(!(clp->cl_flags & NFSD4_CLIENT_CB_FLAG_MASK));
1687 clear_bit(NFSD4_CLIENT_CB_UPDATE, addr: &clp->cl_flags);
1688
1689 memcpy(&conn, &cb->cb_clp->cl_cb_conn, sizeof(struct nfs4_cb_conn));
1690 c = __nfsd4_find_backchannel(clp);
1691 if (c) {
1692 svc_xprt_get(xprt: c->cn_xprt);
1693 conn.cb_xprt = c->cn_xprt;
1694 ses = c->cn_session;
1695 }
1696 spin_unlock(lock: &clp->cl_lock);
1697
1698 err = setup_callback_client(clp, conn: &conn, ses);
1699 if (err) {
1700 nfsd4_mark_cb_down(clp);
1701 if (c)
1702 svc_xprt_put(xprt: c->cn_xprt);
1703 return;
1704 }
1705}
1706
1707static void
1708nfsd4_run_cb_work(struct work_struct *work)
1709{
1710 struct nfsd4_callback *cb =
1711 container_of(work, struct nfsd4_callback, cb_work);
1712 struct nfs4_client *clp = cb->cb_clp;
1713 struct rpc_clnt *clnt;
1714 int flags, ret;
1715
1716 trace_nfsd_cb_start(clp);
1717
1718 if (clp->cl_flags & NFSD4_CLIENT_CB_FLAG_MASK)
1719 nfsd4_process_cb_update(cb);
1720
1721 clnt = clp->cl_cb_client;
1722 if (!clnt || clp->cl_state == NFSD4_COURTESY) {
1723 /*
1724 * Callback channel broken, client killed or
1725 * nfs4_client in courtesy state; give up.
1726 */
1727 nfsd41_destroy_cb(cb);
1728 return;
1729 }
1730
1731 /*
1732 * Don't send probe messages for 4.1 or later.
1733 */
1734 if (!cb->cb_ops && clp->cl_minorversion) {
1735 nfsd4_mark_cb_state(clp, NFSD4_CB_UP);
1736 nfsd41_destroy_cb(cb);
1737 return;
1738 }
1739
1740 if (!test_and_clear_bit(NFSD4_CALLBACK_REQUEUE, addr: &cb->cb_flags)) {
1741 if (cb->cb_ops && cb->cb_ops->prepare)
1742 cb->cb_ops->prepare(cb);
1743 }
1744
1745 cb->cb_msg.rpc_cred = clp->cl_cb_cred;
1746 flags = clp->cl_minorversion ? RPC_TASK_NOCONNECT : RPC_TASK_SOFTCONN;
1747 ret = rpc_call_async(clnt, msg: &cb->cb_msg, RPC_TASK_SOFT | flags,
1748 tk_ops: cb->cb_ops ? &nfsd4_cb_ops : &nfsd4_cb_probe_ops, calldata: cb);
1749 if (ret != 0) {
1750 set_bit(NFSD4_CALLBACK_REQUEUE, addr: &cb->cb_flags);
1751 nfsd4_queue_cb(cb);
1752 }
1753}
1754
1755void nfsd4_init_cb(struct nfsd4_callback *cb, struct nfs4_client *clp,
1756 const struct nfsd4_callback_ops *ops, enum nfsd4_cb_op op)
1757{
1758 cb->cb_clp = clp;
1759 cb->cb_msg.rpc_proc = &nfs4_cb_procedures[op];
1760 cb->cb_msg.rpc_argp = cb;
1761 cb->cb_msg.rpc_resp = cb;
1762 cb->cb_flags = 0;
1763 cb->cb_ops = ops;
1764 INIT_WORK(&cb->cb_work, nfsd4_run_cb_work);
1765 cb->cb_status = 0;
1766 cb->cb_held_slot = -1;
1767 cb->cb_nr_referring_call_list = 0;
1768 INIT_LIST_HEAD(list: &cb->cb_referring_call_list);
1769}
1770
1771/**
1772 * nfsd4_run_cb - queue up a callback job to run
1773 * @cb: callback to queue
1774 *
1775 * Kick off a callback to do its thing. Returns false if it was already
1776 * on a queue, true otherwise.
1777 */
1778bool nfsd4_run_cb(struct nfsd4_callback *cb)
1779{
1780 struct nfs4_client *clp = cb->cb_clp;
1781 bool queued;
1782
1783 nfsd41_cb_inflight_begin(clp);
1784 queued = nfsd4_queue_cb(cb);
1785 if (!queued)
1786 nfsd41_cb_inflight_end(clp);
1787 return queued;
1788}
1789

source code of linux/fs/nfsd/nfs4callback.c