xref: /openbmc/linux/fs/nfsd/nfs4callback.c (revision f125e2d4)
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/sunrpc/clnt.h>
35 #include <linux/sunrpc/xprt.h>
36 #include <linux/sunrpc/svc_xprt.h>
37 #include <linux/slab.h>
38 #include "nfsd.h"
39 #include "state.h"
40 #include "netns.h"
41 #include "xdr4cb.h"
42 #include "xdr4.h"
43 
44 #define NFSDDBG_FACILITY                NFSDDBG_PROC
45 
46 static void nfsd4_mark_cb_fault(struct nfs4_client *, int reason);
47 
48 #define NFSPROC4_CB_NULL 0
49 #define NFSPROC4_CB_COMPOUND 1
50 
51 /* Index of predefined Linux callback client operations */
52 
53 struct nfs4_cb_compound_hdr {
54 	/* args */
55 	u32		ident;	/* minorversion 0 only */
56 	u32		nops;
57 	__be32		*nops_p;
58 	u32		minorversion;
59 	/* res */
60 	int		status;
61 };
62 
63 static __be32 *xdr_encode_empty_array(__be32 *p)
64 {
65 	*p++ = xdr_zero;
66 	return p;
67 }
68 
69 /*
70  * Encode/decode NFSv4 CB basic data types
71  *
72  * Basic NFSv4 callback data types are defined in section 15 of RFC
73  * 3530: "Network File System (NFS) version 4 Protocol" and section
74  * 20 of RFC 5661: "Network File System (NFS) Version 4 Minor Version
75  * 1 Protocol"
76  */
77 
78 /*
79  *	nfs_cb_opnum4
80  *
81  *	enum nfs_cb_opnum4 {
82  *		OP_CB_GETATTR		= 3,
83  *		  ...
84  *	};
85  */
86 enum nfs_cb_opnum4 {
87 	OP_CB_GETATTR			= 3,
88 	OP_CB_RECALL			= 4,
89 	OP_CB_LAYOUTRECALL		= 5,
90 	OP_CB_NOTIFY			= 6,
91 	OP_CB_PUSH_DELEG		= 7,
92 	OP_CB_RECALL_ANY		= 8,
93 	OP_CB_RECALLABLE_OBJ_AVAIL	= 9,
94 	OP_CB_RECALL_SLOT		= 10,
95 	OP_CB_SEQUENCE			= 11,
96 	OP_CB_WANTS_CANCELLED		= 12,
97 	OP_CB_NOTIFY_LOCK		= 13,
98 	OP_CB_NOTIFY_DEVICEID		= 14,
99 	OP_CB_OFFLOAD			= 15,
100 	OP_CB_ILLEGAL			= 10044
101 };
102 
103 static void encode_nfs_cb_opnum4(struct xdr_stream *xdr, enum nfs_cb_opnum4 op)
104 {
105 	__be32 *p;
106 
107 	p = xdr_reserve_space(xdr, 4);
108 	*p = cpu_to_be32(op);
109 }
110 
111 /*
112  * nfs_fh4
113  *
114  *	typedef opaque nfs_fh4<NFS4_FHSIZE>;
115  */
116 static void encode_nfs_fh4(struct xdr_stream *xdr, const struct knfsd_fh *fh)
117 {
118 	u32 length = fh->fh_size;
119 	__be32 *p;
120 
121 	BUG_ON(length > NFS4_FHSIZE);
122 	p = xdr_reserve_space(xdr, 4 + length);
123 	xdr_encode_opaque(p, &fh->fh_base, length);
124 }
125 
126 /*
127  * stateid4
128  *
129  *	struct stateid4 {
130  *		uint32_t	seqid;
131  *		opaque		other[12];
132  *	};
133  */
134 static void encode_stateid4(struct xdr_stream *xdr, const stateid_t *sid)
135 {
136 	__be32 *p;
137 
138 	p = xdr_reserve_space(xdr, NFS4_STATEID_SIZE);
139 	*p++ = cpu_to_be32(sid->si_generation);
140 	xdr_encode_opaque_fixed(p, &sid->si_opaque, NFS4_STATEID_OTHER_SIZE);
141 }
142 
143 /*
144  * sessionid4
145  *
146  *	typedef opaque sessionid4[NFS4_SESSIONID_SIZE];
147  */
148 static void encode_sessionid4(struct xdr_stream *xdr,
149 			      const struct nfsd4_session *session)
150 {
151 	__be32 *p;
152 
153 	p = xdr_reserve_space(xdr, NFS4_MAX_SESSIONID_LEN);
154 	xdr_encode_opaque_fixed(p, session->se_sessionid.data,
155 					NFS4_MAX_SESSIONID_LEN);
156 }
157 
158 /*
159  * nfsstat4
160  */
161 static const struct {
162 	int stat;
163 	int errno;
164 } nfs_cb_errtbl[] = {
165 	{ NFS4_OK,		0		},
166 	{ NFS4ERR_PERM,		-EPERM		},
167 	{ NFS4ERR_NOENT,	-ENOENT		},
168 	{ NFS4ERR_IO,		-EIO		},
169 	{ NFS4ERR_NXIO,		-ENXIO		},
170 	{ NFS4ERR_ACCESS,	-EACCES		},
171 	{ NFS4ERR_EXIST,	-EEXIST		},
172 	{ NFS4ERR_XDEV,		-EXDEV		},
173 	{ NFS4ERR_NOTDIR,	-ENOTDIR	},
174 	{ NFS4ERR_ISDIR,	-EISDIR		},
175 	{ NFS4ERR_INVAL,	-EINVAL		},
176 	{ NFS4ERR_FBIG,		-EFBIG		},
177 	{ NFS4ERR_NOSPC,	-ENOSPC		},
178 	{ NFS4ERR_ROFS,		-EROFS		},
179 	{ NFS4ERR_MLINK,	-EMLINK		},
180 	{ NFS4ERR_NAMETOOLONG,	-ENAMETOOLONG	},
181 	{ NFS4ERR_NOTEMPTY,	-ENOTEMPTY	},
182 	{ NFS4ERR_DQUOT,	-EDQUOT		},
183 	{ NFS4ERR_STALE,	-ESTALE		},
184 	{ NFS4ERR_BADHANDLE,	-EBADHANDLE	},
185 	{ NFS4ERR_BAD_COOKIE,	-EBADCOOKIE	},
186 	{ NFS4ERR_NOTSUPP,	-ENOTSUPP	},
187 	{ NFS4ERR_TOOSMALL,	-ETOOSMALL	},
188 	{ NFS4ERR_SERVERFAULT,	-ESERVERFAULT	},
189 	{ NFS4ERR_BADTYPE,	-EBADTYPE	},
190 	{ NFS4ERR_LOCKED,	-EAGAIN		},
191 	{ NFS4ERR_RESOURCE,	-EREMOTEIO	},
192 	{ NFS4ERR_SYMLINK,	-ELOOP		},
193 	{ NFS4ERR_OP_ILLEGAL,	-EOPNOTSUPP	},
194 	{ NFS4ERR_DEADLOCK,	-EDEADLK	},
195 	{ -1,			-EIO		}
196 };
197 
198 /*
199  * If we cannot translate the error, the recovery routines should
200  * handle it.
201  *
202  * Note: remaining NFSv4 error codes have values > 10000, so should
203  * not conflict with native Linux error codes.
204  */
205 static int nfs_cb_stat_to_errno(int status)
206 {
207 	int i;
208 
209 	for (i = 0; nfs_cb_errtbl[i].stat != -1; i++) {
210 		if (nfs_cb_errtbl[i].stat == status)
211 			return nfs_cb_errtbl[i].errno;
212 	}
213 
214 	dprintk("NFSD: Unrecognized NFS CB status value: %u\n", status);
215 	return -status;
216 }
217 
218 static int decode_cb_op_status(struct xdr_stream *xdr,
219 			       enum nfs_cb_opnum4 expected, int *status)
220 {
221 	__be32 *p;
222 	u32 op;
223 
224 	p = xdr_inline_decode(xdr, 4 + 4);
225 	if (unlikely(p == NULL))
226 		goto out_overflow;
227 	op = be32_to_cpup(p++);
228 	if (unlikely(op != expected))
229 		goto out_unexpected;
230 	*status = nfs_cb_stat_to_errno(be32_to_cpup(p));
231 	return 0;
232 out_overflow:
233 	return -EIO;
234 out_unexpected:
235 	dprintk("NFSD: Callback server returned operation %d but "
236 		"we issued a request for %d\n", op, expected);
237 	return -EIO;
238 }
239 
240 /*
241  * CB_COMPOUND4args
242  *
243  *	struct CB_COMPOUND4args {
244  *		utf8str_cs	tag;
245  *		uint32_t	minorversion;
246  *		uint32_t	callback_ident;
247  *		nfs_cb_argop4	argarray<>;
248  *	};
249 */
250 static void encode_cb_compound4args(struct xdr_stream *xdr,
251 				    struct nfs4_cb_compound_hdr *hdr)
252 {
253 	__be32 * p;
254 
255 	p = xdr_reserve_space(xdr, 4 + 4 + 4 + 4);
256 	p = xdr_encode_empty_array(p);		/* empty tag */
257 	*p++ = cpu_to_be32(hdr->minorversion);
258 	*p++ = cpu_to_be32(hdr->ident);
259 
260 	hdr->nops_p = p;
261 	*p = cpu_to_be32(hdr->nops);		/* argarray element count */
262 }
263 
264 /*
265  * Update argarray element count
266  */
267 static void encode_cb_nops(struct nfs4_cb_compound_hdr *hdr)
268 {
269 	BUG_ON(hdr->nops > NFS4_MAX_BACK_CHANNEL_OPS);
270 	*hdr->nops_p = cpu_to_be32(hdr->nops);
271 }
272 
273 /*
274  * CB_COMPOUND4res
275  *
276  *	struct CB_COMPOUND4res {
277  *		nfsstat4	status;
278  *		utf8str_cs	tag;
279  *		nfs_cb_resop4	resarray<>;
280  *	};
281  */
282 static int decode_cb_compound4res(struct xdr_stream *xdr,
283 				  struct nfs4_cb_compound_hdr *hdr)
284 {
285 	u32 length;
286 	__be32 *p;
287 
288 	p = xdr_inline_decode(xdr, 4 + 4);
289 	if (unlikely(p == NULL))
290 		goto out_overflow;
291 	hdr->status = be32_to_cpup(p++);
292 	/* Ignore the tag */
293 	length = be32_to_cpup(p++);
294 	p = xdr_inline_decode(xdr, length + 4);
295 	if (unlikely(p == NULL))
296 		goto out_overflow;
297 	p += XDR_QUADLEN(length);
298 	hdr->nops = be32_to_cpup(p);
299 	return 0;
300 out_overflow:
301 	return -EIO;
302 }
303 
304 /*
305  * CB_RECALL4args
306  *
307  *	struct CB_RECALL4args {
308  *		stateid4	stateid;
309  *		bool		truncate;
310  *		nfs_fh4		fh;
311  *	};
312  */
313 static void encode_cb_recall4args(struct xdr_stream *xdr,
314 				  const struct nfs4_delegation *dp,
315 				  struct nfs4_cb_compound_hdr *hdr)
316 {
317 	__be32 *p;
318 
319 	encode_nfs_cb_opnum4(xdr, OP_CB_RECALL);
320 	encode_stateid4(xdr, &dp->dl_stid.sc_stateid);
321 
322 	p = xdr_reserve_space(xdr, 4);
323 	*p++ = xdr_zero;			/* truncate */
324 
325 	encode_nfs_fh4(xdr, &dp->dl_stid.sc_file->fi_fhandle);
326 
327 	hdr->nops++;
328 }
329 
330 /*
331  * CB_SEQUENCE4args
332  *
333  *	struct CB_SEQUENCE4args {
334  *		sessionid4		csa_sessionid;
335  *		sequenceid4		csa_sequenceid;
336  *		slotid4			csa_slotid;
337  *		slotid4			csa_highest_slotid;
338  *		bool			csa_cachethis;
339  *		referring_call_list4	csa_referring_call_lists<>;
340  *	};
341  */
342 static void encode_cb_sequence4args(struct xdr_stream *xdr,
343 				    const struct nfsd4_callback *cb,
344 				    struct nfs4_cb_compound_hdr *hdr)
345 {
346 	struct nfsd4_session *session = cb->cb_clp->cl_cb_session;
347 	__be32 *p;
348 
349 	if (hdr->minorversion == 0)
350 		return;
351 
352 	encode_nfs_cb_opnum4(xdr, OP_CB_SEQUENCE);
353 	encode_sessionid4(xdr, session);
354 
355 	p = xdr_reserve_space(xdr, 4 + 4 + 4 + 4 + 4);
356 	*p++ = cpu_to_be32(session->se_cb_seq_nr);	/* csa_sequenceid */
357 	*p++ = xdr_zero;			/* csa_slotid */
358 	*p++ = xdr_zero;			/* csa_highest_slotid */
359 	*p++ = xdr_zero;			/* csa_cachethis */
360 	xdr_encode_empty_array(p);		/* csa_referring_call_lists */
361 
362 	hdr->nops++;
363 }
364 
365 /*
366  * CB_SEQUENCE4resok
367  *
368  *	struct CB_SEQUENCE4resok {
369  *		sessionid4	csr_sessionid;
370  *		sequenceid4	csr_sequenceid;
371  *		slotid4		csr_slotid;
372  *		slotid4		csr_highest_slotid;
373  *		slotid4		csr_target_highest_slotid;
374  *	};
375  *
376  *	union CB_SEQUENCE4res switch (nfsstat4 csr_status) {
377  *	case NFS4_OK:
378  *		CB_SEQUENCE4resok	csr_resok4;
379  *	default:
380  *		void;
381  *	};
382  *
383  * Our current back channel implmentation supports a single backchannel
384  * with a single slot.
385  */
386 static int decode_cb_sequence4resok(struct xdr_stream *xdr,
387 				    struct nfsd4_callback *cb)
388 {
389 	struct nfsd4_session *session = cb->cb_clp->cl_cb_session;
390 	int status = -ESERVERFAULT;
391 	__be32 *p;
392 	u32 dummy;
393 
394 	/*
395 	 * If the server returns different values for sessionID, slotID or
396 	 * sequence number, the server is looney tunes.
397 	 */
398 	p = xdr_inline_decode(xdr, NFS4_MAX_SESSIONID_LEN + 4 + 4 + 4 + 4);
399 	if (unlikely(p == NULL))
400 		goto out_overflow;
401 
402 	if (memcmp(p, session->se_sessionid.data, NFS4_MAX_SESSIONID_LEN)) {
403 		dprintk("NFS: %s Invalid session id\n", __func__);
404 		goto out;
405 	}
406 	p += XDR_QUADLEN(NFS4_MAX_SESSIONID_LEN);
407 
408 	dummy = be32_to_cpup(p++);
409 	if (dummy != session->se_cb_seq_nr) {
410 		dprintk("NFS: %s Invalid sequence number\n", __func__);
411 		goto out;
412 	}
413 
414 	dummy = be32_to_cpup(p++);
415 	if (dummy != 0) {
416 		dprintk("NFS: %s Invalid slotid\n", __func__);
417 		goto out;
418 	}
419 
420 	/*
421 	 * FIXME: process highest slotid and target highest slotid
422 	 */
423 	status = 0;
424 out:
425 	cb->cb_seq_status = status;
426 	return status;
427 out_overflow:
428 	status = -EIO;
429 	goto out;
430 }
431 
432 static int decode_cb_sequence4res(struct xdr_stream *xdr,
433 				  struct nfsd4_callback *cb)
434 {
435 	int status;
436 
437 	if (cb->cb_clp->cl_minorversion == 0)
438 		return 0;
439 
440 	status = decode_cb_op_status(xdr, OP_CB_SEQUENCE, &cb->cb_seq_status);
441 	if (unlikely(status || cb->cb_seq_status))
442 		return status;
443 
444 	return decode_cb_sequence4resok(xdr, cb);
445 }
446 
447 /*
448  * NFSv4.0 and NFSv4.1 XDR encode functions
449  *
450  * NFSv4.0 callback argument types are defined in section 15 of RFC
451  * 3530: "Network File System (NFS) version 4 Protocol" and section 20
452  * of RFC 5661:  "Network File System (NFS) Version 4 Minor Version 1
453  * Protocol".
454  */
455 
456 /*
457  * NB: Without this zero space reservation, callbacks over krb5p fail
458  */
459 static void nfs4_xdr_enc_cb_null(struct rpc_rqst *req, struct xdr_stream *xdr,
460 				 const void *__unused)
461 {
462 	xdr_reserve_space(xdr, 0);
463 }
464 
465 /*
466  * 20.2. Operation 4: CB_RECALL - Recall a Delegation
467  */
468 static void nfs4_xdr_enc_cb_recall(struct rpc_rqst *req, struct xdr_stream *xdr,
469 				   const void *data)
470 {
471 	const struct nfsd4_callback *cb = data;
472 	const struct nfs4_delegation *dp = cb_to_delegation(cb);
473 	struct nfs4_cb_compound_hdr hdr = {
474 		.ident = cb->cb_clp->cl_cb_ident,
475 		.minorversion = cb->cb_clp->cl_minorversion,
476 	};
477 
478 	encode_cb_compound4args(xdr, &hdr);
479 	encode_cb_sequence4args(xdr, cb, &hdr);
480 	encode_cb_recall4args(xdr, dp, &hdr);
481 	encode_cb_nops(&hdr);
482 }
483 
484 
485 /*
486  * NFSv4.0 and NFSv4.1 XDR decode functions
487  *
488  * NFSv4.0 callback result types are defined in section 15 of RFC
489  * 3530: "Network File System (NFS) version 4 Protocol" and section 20
490  * of RFC 5661:  "Network File System (NFS) Version 4 Minor Version 1
491  * Protocol".
492  */
493 
494 static int nfs4_xdr_dec_cb_null(struct rpc_rqst *req, struct xdr_stream *xdr,
495 				void *__unused)
496 {
497 	return 0;
498 }
499 
500 /*
501  * 20.2. Operation 4: CB_RECALL - Recall a Delegation
502  */
503 static int nfs4_xdr_dec_cb_recall(struct rpc_rqst *rqstp,
504 				  struct xdr_stream *xdr,
505 				  void *data)
506 {
507 	struct nfsd4_callback *cb = data;
508 	struct nfs4_cb_compound_hdr hdr;
509 	int status;
510 
511 	status = decode_cb_compound4res(xdr, &hdr);
512 	if (unlikely(status))
513 		return status;
514 
515 	status = decode_cb_sequence4res(xdr, cb);
516 	if (unlikely(status || cb->cb_seq_status))
517 		return status;
518 
519 	return decode_cb_op_status(xdr, OP_CB_RECALL, &cb->cb_status);
520 }
521 
522 #ifdef CONFIG_NFSD_PNFS
523 /*
524  * CB_LAYOUTRECALL4args
525  *
526  *	struct layoutrecall_file4 {
527  *		nfs_fh4         lor_fh;
528  *		offset4         lor_offset;
529  *		length4         lor_length;
530  *		stateid4        lor_stateid;
531  *	};
532  *
533  *	union layoutrecall4 switch(layoutrecall_type4 lor_recalltype) {
534  *	case LAYOUTRECALL4_FILE:
535  *		layoutrecall_file4 lor_layout;
536  *	case LAYOUTRECALL4_FSID:
537  *		fsid4              lor_fsid;
538  *	case LAYOUTRECALL4_ALL:
539  *		void;
540  *	};
541  *
542  *	struct CB_LAYOUTRECALL4args {
543  *		layouttype4             clora_type;
544  *		layoutiomode4           clora_iomode;
545  *		bool                    clora_changed;
546  *		layoutrecall4           clora_recall;
547  *	};
548  */
549 static void encode_cb_layout4args(struct xdr_stream *xdr,
550 				  const struct nfs4_layout_stateid *ls,
551 				  struct nfs4_cb_compound_hdr *hdr)
552 {
553 	__be32 *p;
554 
555 	BUG_ON(hdr->minorversion == 0);
556 
557 	p = xdr_reserve_space(xdr, 5 * 4);
558 	*p++ = cpu_to_be32(OP_CB_LAYOUTRECALL);
559 	*p++ = cpu_to_be32(ls->ls_layout_type);
560 	*p++ = cpu_to_be32(IOMODE_ANY);
561 	*p++ = cpu_to_be32(1);
562 	*p = cpu_to_be32(RETURN_FILE);
563 
564 	encode_nfs_fh4(xdr, &ls->ls_stid.sc_file->fi_fhandle);
565 
566 	p = xdr_reserve_space(xdr, 2 * 8);
567 	p = xdr_encode_hyper(p, 0);
568 	xdr_encode_hyper(p, NFS4_MAX_UINT64);
569 
570 	encode_stateid4(xdr, &ls->ls_recall_sid);
571 
572 	hdr->nops++;
573 }
574 
575 static void nfs4_xdr_enc_cb_layout(struct rpc_rqst *req,
576 				   struct xdr_stream *xdr,
577 				   const void *data)
578 {
579 	const struct nfsd4_callback *cb = data;
580 	const struct nfs4_layout_stateid *ls =
581 		container_of(cb, struct nfs4_layout_stateid, ls_recall);
582 	struct nfs4_cb_compound_hdr hdr = {
583 		.ident = 0,
584 		.minorversion = cb->cb_clp->cl_minorversion,
585 	};
586 
587 	encode_cb_compound4args(xdr, &hdr);
588 	encode_cb_sequence4args(xdr, cb, &hdr);
589 	encode_cb_layout4args(xdr, ls, &hdr);
590 	encode_cb_nops(&hdr);
591 }
592 
593 static int nfs4_xdr_dec_cb_layout(struct rpc_rqst *rqstp,
594 				  struct xdr_stream *xdr,
595 				  void *data)
596 {
597 	struct nfsd4_callback *cb = data;
598 	struct nfs4_cb_compound_hdr hdr;
599 	int status;
600 
601 	status = decode_cb_compound4res(xdr, &hdr);
602 	if (unlikely(status))
603 		return status;
604 
605 	status = decode_cb_sequence4res(xdr, cb);
606 	if (unlikely(status || cb->cb_seq_status))
607 		return status;
608 
609 	return decode_cb_op_status(xdr, OP_CB_LAYOUTRECALL, &cb->cb_status);
610 }
611 #endif /* CONFIG_NFSD_PNFS */
612 
613 static void encode_stateowner(struct xdr_stream *xdr, struct nfs4_stateowner *so)
614 {
615 	__be32	*p;
616 
617 	p = xdr_reserve_space(xdr, 8 + 4 + so->so_owner.len);
618 	p = xdr_encode_opaque_fixed(p, &so->so_client->cl_clientid, 8);
619 	xdr_encode_opaque(p, so->so_owner.data, so->so_owner.len);
620 }
621 
622 static void nfs4_xdr_enc_cb_notify_lock(struct rpc_rqst *req,
623 					struct xdr_stream *xdr,
624 					const void *data)
625 {
626 	const struct nfsd4_callback *cb = data;
627 	const struct nfsd4_blocked_lock *nbl =
628 		container_of(cb, struct nfsd4_blocked_lock, nbl_cb);
629 	struct nfs4_lockowner *lo = (struct nfs4_lockowner *)nbl->nbl_lock.fl_owner;
630 	struct nfs4_cb_compound_hdr hdr = {
631 		.ident = 0,
632 		.minorversion = cb->cb_clp->cl_minorversion,
633 	};
634 
635 	__be32 *p;
636 
637 	BUG_ON(hdr.minorversion == 0);
638 
639 	encode_cb_compound4args(xdr, &hdr);
640 	encode_cb_sequence4args(xdr, cb, &hdr);
641 
642 	p = xdr_reserve_space(xdr, 4);
643 	*p = cpu_to_be32(OP_CB_NOTIFY_LOCK);
644 	encode_nfs_fh4(xdr, &nbl->nbl_fh);
645 	encode_stateowner(xdr, &lo->lo_owner);
646 	hdr.nops++;
647 
648 	encode_cb_nops(&hdr);
649 }
650 
651 static int nfs4_xdr_dec_cb_notify_lock(struct rpc_rqst *rqstp,
652 					struct xdr_stream *xdr,
653 					void *data)
654 {
655 	struct nfsd4_callback *cb = data;
656 	struct nfs4_cb_compound_hdr hdr;
657 	int status;
658 
659 	status = decode_cb_compound4res(xdr, &hdr);
660 	if (unlikely(status))
661 		return status;
662 
663 	status = decode_cb_sequence4res(xdr, cb);
664 	if (unlikely(status || cb->cb_seq_status))
665 		return status;
666 
667 	return decode_cb_op_status(xdr, OP_CB_NOTIFY_LOCK, &cb->cb_status);
668 }
669 
670 /*
671  * struct write_response4 {
672  *	stateid4	wr_callback_id<1>;
673  *	length4		wr_count;
674  *	stable_how4	wr_committed;
675  *	verifier4	wr_writeverf;
676  * };
677  * union offload_info4 switch (nfsstat4 coa_status) {
678  *	case NFS4_OK:
679  *		write_response4	coa_resok4;
680  *	default:
681  *	length4		coa_bytes_copied;
682  * };
683  * struct CB_OFFLOAD4args {
684  *	nfs_fh4		coa_fh;
685  *	stateid4	coa_stateid;
686  *	offload_info4	coa_offload_info;
687  * };
688  */
689 static void encode_offload_info4(struct xdr_stream *xdr,
690 				 __be32 nfserr,
691 				 const struct nfsd4_copy *cp)
692 {
693 	__be32 *p;
694 
695 	p = xdr_reserve_space(xdr, 4);
696 	*p++ = nfserr;
697 	if (!nfserr) {
698 		p = xdr_reserve_space(xdr, 4 + 8 + 4 + NFS4_VERIFIER_SIZE);
699 		p = xdr_encode_empty_array(p);
700 		p = xdr_encode_hyper(p, cp->cp_res.wr_bytes_written);
701 		*p++ = cpu_to_be32(cp->cp_res.wr_stable_how);
702 		p = xdr_encode_opaque_fixed(p, cp->cp_res.wr_verifier.data,
703 					    NFS4_VERIFIER_SIZE);
704 	} else {
705 		p = xdr_reserve_space(xdr, 8);
706 		/* We always return success if bytes were written */
707 		p = xdr_encode_hyper(p, 0);
708 	}
709 }
710 
711 static void encode_cb_offload4args(struct xdr_stream *xdr,
712 				   __be32 nfserr,
713 				   const struct knfsd_fh *fh,
714 				   const struct nfsd4_copy *cp,
715 				   struct nfs4_cb_compound_hdr *hdr)
716 {
717 	__be32 *p;
718 
719 	p = xdr_reserve_space(xdr, 4);
720 	*p++ = cpu_to_be32(OP_CB_OFFLOAD);
721 	encode_nfs_fh4(xdr, fh);
722 	encode_stateid4(xdr, &cp->cp_res.cb_stateid);
723 	encode_offload_info4(xdr, nfserr, cp);
724 
725 	hdr->nops++;
726 }
727 
728 static void nfs4_xdr_enc_cb_offload(struct rpc_rqst *req,
729 				    struct xdr_stream *xdr,
730 				    const void *data)
731 {
732 	const struct nfsd4_callback *cb = data;
733 	const struct nfsd4_copy *cp =
734 		container_of(cb, struct nfsd4_copy, cp_cb);
735 	struct nfs4_cb_compound_hdr hdr = {
736 		.ident = 0,
737 		.minorversion = cb->cb_clp->cl_minorversion,
738 	};
739 
740 	encode_cb_compound4args(xdr, &hdr);
741 	encode_cb_sequence4args(xdr, cb, &hdr);
742 	encode_cb_offload4args(xdr, cp->nfserr, &cp->fh, cp, &hdr);
743 	encode_cb_nops(&hdr);
744 }
745 
746 static int nfs4_xdr_dec_cb_offload(struct rpc_rqst *rqstp,
747 				   struct xdr_stream *xdr,
748 				   void *data)
749 {
750 	struct nfsd4_callback *cb = data;
751 	struct nfs4_cb_compound_hdr hdr;
752 	int status;
753 
754 	status = decode_cb_compound4res(xdr, &hdr);
755 	if (unlikely(status))
756 		return status;
757 
758 	status = decode_cb_sequence4res(xdr, cb);
759 	if (unlikely(status || cb->cb_seq_status))
760 		return status;
761 
762 	return decode_cb_op_status(xdr, OP_CB_OFFLOAD, &cb->cb_status);
763 }
764 /*
765  * RPC procedure tables
766  */
767 #define PROC(proc, call, argtype, restype)				\
768 [NFSPROC4_CLNT_##proc] = {						\
769 	.p_proc    = NFSPROC4_CB_##call,				\
770 	.p_encode  = nfs4_xdr_enc_##argtype,		\
771 	.p_decode  = nfs4_xdr_dec_##restype,				\
772 	.p_arglen  = NFS4_enc_##argtype##_sz,				\
773 	.p_replen  = NFS4_dec_##restype##_sz,				\
774 	.p_statidx = NFSPROC4_CB_##call,				\
775 	.p_name    = #proc,						\
776 }
777 
778 static const struct rpc_procinfo nfs4_cb_procedures[] = {
779 	PROC(CB_NULL,	NULL,		cb_null,	cb_null),
780 	PROC(CB_RECALL,	COMPOUND,	cb_recall,	cb_recall),
781 #ifdef CONFIG_NFSD_PNFS
782 	PROC(CB_LAYOUT,	COMPOUND,	cb_layout,	cb_layout),
783 #endif
784 	PROC(CB_NOTIFY_LOCK,	COMPOUND,	cb_notify_lock,	cb_notify_lock),
785 	PROC(CB_OFFLOAD,	COMPOUND,	cb_offload,	cb_offload),
786 };
787 
788 static unsigned int nfs4_cb_counts[ARRAY_SIZE(nfs4_cb_procedures)];
789 static const struct rpc_version nfs_cb_version4 = {
790 /*
791  * Note on the callback rpc program version number: despite language in rfc
792  * 5661 section 18.36.3 requiring servers to use 4 in this field, the
793  * official xdr descriptions for both 4.0 and 4.1 specify version 1, and
794  * in practice that appears to be what implementations use.  The section
795  * 18.36.3 language is expected to be fixed in an erratum.
796  */
797 	.number			= 1,
798 	.nrprocs		= ARRAY_SIZE(nfs4_cb_procedures),
799 	.procs			= nfs4_cb_procedures,
800 	.counts			= nfs4_cb_counts,
801 };
802 
803 static const struct rpc_version *nfs_cb_version[2] = {
804 	[1] = &nfs_cb_version4,
805 };
806 
807 static const struct rpc_program cb_program;
808 
809 static struct rpc_stat cb_stats = {
810 	.program		= &cb_program
811 };
812 
813 #define NFS4_CALLBACK 0x40000000
814 static const struct rpc_program cb_program = {
815 	.name			= "nfs4_cb",
816 	.number			= NFS4_CALLBACK,
817 	.nrvers			= ARRAY_SIZE(nfs_cb_version),
818 	.version		= nfs_cb_version,
819 	.stats			= &cb_stats,
820 	.pipe_dir_name		= "nfsd4_cb",
821 };
822 
823 static int max_cb_time(struct net *net)
824 {
825 	struct nfsd_net *nn = net_generic(net, nfsd_net_id);
826 
827 	/*
828 	 * nfsd4_lease is set to at most one hour in __nfsd4_write_time,
829 	 * so we can use 32-bit math on it. Warn if that assumption
830 	 * ever stops being true.
831 	 */
832 	if (WARN_ON_ONCE(nn->nfsd4_lease > 3600))
833 		return 360 * HZ;
834 
835 	return max(((u32)nn->nfsd4_lease)/10, 1u) * HZ;
836 }
837 
838 static struct workqueue_struct *callback_wq;
839 
840 static bool nfsd4_queue_cb(struct nfsd4_callback *cb)
841 {
842 	return queue_work(callback_wq, &cb->cb_work);
843 }
844 
845 static void nfsd41_cb_inflight_begin(struct nfs4_client *clp)
846 {
847 	atomic_inc(&clp->cl_cb_inflight);
848 }
849 
850 static void nfsd41_cb_inflight_end(struct nfs4_client *clp)
851 {
852 
853 	if (atomic_dec_and_test(&clp->cl_cb_inflight))
854 		wake_up_var(&clp->cl_cb_inflight);
855 }
856 
857 static void nfsd41_cb_inflight_wait_complete(struct nfs4_client *clp)
858 {
859 	wait_var_event(&clp->cl_cb_inflight,
860 			!atomic_read(&clp->cl_cb_inflight));
861 }
862 
863 static const struct cred *get_backchannel_cred(struct nfs4_client *clp, struct rpc_clnt *client, struct nfsd4_session *ses)
864 {
865 	if (clp->cl_minorversion == 0) {
866 		client->cl_principal = clp->cl_cred.cr_targ_princ ?
867 			clp->cl_cred.cr_targ_princ : "nfs";
868 
869 		return get_cred(rpc_machine_cred());
870 	} else {
871 		struct cred *kcred;
872 
873 		kcred = prepare_kernel_cred(NULL);
874 		if (!kcred)
875 			return NULL;
876 
877 		kcred->uid = ses->se_cb_sec.uid;
878 		kcred->gid = ses->se_cb_sec.gid;
879 		return kcred;
880 	}
881 }
882 
883 static int setup_callback_client(struct nfs4_client *clp, struct nfs4_cb_conn *conn, struct nfsd4_session *ses)
884 {
885 	int maxtime = max_cb_time(clp->net);
886 	struct rpc_timeout	timeparms = {
887 		.to_initval	= maxtime,
888 		.to_retries	= 0,
889 		.to_maxval	= maxtime,
890 	};
891 	struct rpc_create_args args = {
892 		.net		= clp->net,
893 		.address	= (struct sockaddr *) &conn->cb_addr,
894 		.addrsize	= conn->cb_addrlen,
895 		.saddress	= (struct sockaddr *) &conn->cb_saddr,
896 		.timeout	= &timeparms,
897 		.program	= &cb_program,
898 		.version	= 1,
899 		.flags		= (RPC_CLNT_CREATE_NOPING | RPC_CLNT_CREATE_QUIET),
900 		.cred		= current_cred(),
901 	};
902 	struct rpc_clnt *client;
903 	const struct cred *cred;
904 
905 	if (clp->cl_minorversion == 0) {
906 		if (!clp->cl_cred.cr_principal &&
907 				(clp->cl_cred.cr_flavor >= RPC_AUTH_GSS_KRB5))
908 			return -EINVAL;
909 		args.client_name = clp->cl_cred.cr_principal;
910 		args.prognumber	= conn->cb_prog;
911 		args.protocol = XPRT_TRANSPORT_TCP;
912 		args.authflavor = clp->cl_cred.cr_flavor;
913 		clp->cl_cb_ident = conn->cb_ident;
914 	} else {
915 		if (!conn->cb_xprt)
916 			return -EINVAL;
917 		clp->cl_cb_conn.cb_xprt = conn->cb_xprt;
918 		clp->cl_cb_session = ses;
919 		args.bc_xprt = conn->cb_xprt;
920 		args.prognumber = clp->cl_cb_session->se_cb_prog;
921 		args.protocol = conn->cb_xprt->xpt_class->xcl_ident |
922 				XPRT_TRANSPORT_BC;
923 		args.authflavor = ses->se_cb_sec.flavor;
924 	}
925 	/* Create RPC client */
926 	client = rpc_create(&args);
927 	if (IS_ERR(client)) {
928 		dprintk("NFSD: couldn't create callback client: %ld\n",
929 			PTR_ERR(client));
930 		return PTR_ERR(client);
931 	}
932 	cred = get_backchannel_cred(clp, client, ses);
933 	if (!cred) {
934 		rpc_shutdown_client(client);
935 		return -ENOMEM;
936 	}
937 	clp->cl_cb_client = client;
938 	clp->cl_cb_cred = cred;
939 	return 0;
940 }
941 
942 static void warn_no_callback_path(struct nfs4_client *clp, int reason)
943 {
944 	dprintk("NFSD: warning: no callback path to client %.*s: error %d\n",
945 		(int)clp->cl_name.len, clp->cl_name.data, reason);
946 }
947 
948 static void nfsd4_mark_cb_down(struct nfs4_client *clp, int reason)
949 {
950 	if (test_bit(NFSD4_CLIENT_CB_UPDATE, &clp->cl_flags))
951 		return;
952 	clp->cl_cb_state = NFSD4_CB_DOWN;
953 	warn_no_callback_path(clp, reason);
954 }
955 
956 static void nfsd4_mark_cb_fault(struct nfs4_client *clp, int reason)
957 {
958 	if (test_bit(NFSD4_CLIENT_CB_UPDATE, &clp->cl_flags))
959 		return;
960 	clp->cl_cb_state = NFSD4_CB_FAULT;
961 	warn_no_callback_path(clp, reason);
962 }
963 
964 static void nfsd4_cb_probe_done(struct rpc_task *task, void *calldata)
965 {
966 	struct nfs4_client *clp = container_of(calldata, struct nfs4_client, cl_cb_null);
967 
968 	if (task->tk_status)
969 		nfsd4_mark_cb_down(clp, task->tk_status);
970 	else
971 		clp->cl_cb_state = NFSD4_CB_UP;
972 }
973 
974 static void nfsd4_cb_probe_release(void *calldata)
975 {
976 	struct nfs4_client *clp = container_of(calldata, struct nfs4_client, cl_cb_null);
977 
978 	nfsd41_cb_inflight_end(clp);
979 
980 }
981 
982 static const struct rpc_call_ops nfsd4_cb_probe_ops = {
983 	/* XXX: release method to ensure we set the cb channel down if
984 	 * necessary on early failure? */
985 	.rpc_call_done = nfsd4_cb_probe_done,
986 	.rpc_release = nfsd4_cb_probe_release,
987 };
988 
989 /*
990  * Poke the callback thread to process any updates to the callback
991  * parameters, and send a null probe.
992  */
993 void nfsd4_probe_callback(struct nfs4_client *clp)
994 {
995 	clp->cl_cb_state = NFSD4_CB_UNKNOWN;
996 	set_bit(NFSD4_CLIENT_CB_UPDATE, &clp->cl_flags);
997 	nfsd4_run_cb(&clp->cl_cb_null);
998 }
999 
1000 void nfsd4_probe_callback_sync(struct nfs4_client *clp)
1001 {
1002 	nfsd4_probe_callback(clp);
1003 	flush_workqueue(callback_wq);
1004 }
1005 
1006 void nfsd4_change_callback(struct nfs4_client *clp, struct nfs4_cb_conn *conn)
1007 {
1008 	clp->cl_cb_state = NFSD4_CB_UNKNOWN;
1009 	spin_lock(&clp->cl_lock);
1010 	memcpy(&clp->cl_cb_conn, conn, sizeof(struct nfs4_cb_conn));
1011 	spin_unlock(&clp->cl_lock);
1012 }
1013 
1014 /*
1015  * There's currently a single callback channel slot.
1016  * If the slot is available, then mark it busy.  Otherwise, set the
1017  * thread for sleeping on the callback RPC wait queue.
1018  */
1019 static bool nfsd41_cb_get_slot(struct nfsd4_callback *cb, struct rpc_task *task)
1020 {
1021 	struct nfs4_client *clp = cb->cb_clp;
1022 
1023 	if (!cb->cb_holds_slot &&
1024 	    test_and_set_bit(0, &clp->cl_cb_slot_busy) != 0) {
1025 		rpc_sleep_on(&clp->cl_cb_waitq, task, NULL);
1026 		/* Race breaker */
1027 		if (test_and_set_bit(0, &clp->cl_cb_slot_busy) != 0) {
1028 			dprintk("%s slot is busy\n", __func__);
1029 			return false;
1030 		}
1031 		rpc_wake_up_queued_task(&clp->cl_cb_waitq, task);
1032 	}
1033 	cb->cb_holds_slot = true;
1034 	return true;
1035 }
1036 
1037 static void nfsd41_cb_release_slot(struct nfsd4_callback *cb)
1038 {
1039 	struct nfs4_client *clp = cb->cb_clp;
1040 
1041 	if (cb->cb_holds_slot) {
1042 		cb->cb_holds_slot = false;
1043 		clear_bit(0, &clp->cl_cb_slot_busy);
1044 		rpc_wake_up_next(&clp->cl_cb_waitq);
1045 	}
1046 }
1047 
1048 static void nfsd41_destroy_cb(struct nfsd4_callback *cb)
1049 {
1050 	struct nfs4_client *clp = cb->cb_clp;
1051 
1052 	nfsd41_cb_release_slot(cb);
1053 	if (cb->cb_ops && cb->cb_ops->release)
1054 		cb->cb_ops->release(cb);
1055 	nfsd41_cb_inflight_end(clp);
1056 }
1057 
1058 /*
1059  * TODO: cb_sequence should support referring call lists, cachethis, multiple
1060  * slots, and mark callback channel down on communication errors.
1061  */
1062 static void nfsd4_cb_prepare(struct rpc_task *task, void *calldata)
1063 {
1064 	struct nfsd4_callback *cb = calldata;
1065 	struct nfs4_client *clp = cb->cb_clp;
1066 	u32 minorversion = clp->cl_minorversion;
1067 
1068 	/*
1069 	 * cb_seq_status is only set in decode_cb_sequence4res,
1070 	 * and so will remain 1 if an rpc level failure occurs.
1071 	 */
1072 	cb->cb_seq_status = 1;
1073 	cb->cb_status = 0;
1074 	if (minorversion && !nfsd41_cb_get_slot(cb, task))
1075 		return;
1076 	rpc_call_start(task);
1077 }
1078 
1079 static bool nfsd4_cb_sequence_done(struct rpc_task *task, struct nfsd4_callback *cb)
1080 {
1081 	struct nfs4_client *clp = cb->cb_clp;
1082 	struct nfsd4_session *session = clp->cl_cb_session;
1083 	bool ret = true;
1084 
1085 	if (!clp->cl_minorversion) {
1086 		/*
1087 		 * If the backchannel connection was shut down while this
1088 		 * task was queued, we need to resubmit it after setting up
1089 		 * a new backchannel connection.
1090 		 *
1091 		 * Note that if we lost our callback connection permanently
1092 		 * the submission code will error out, so we don't need to
1093 		 * handle that case here.
1094 		 */
1095 		if (RPC_SIGNALLED(task))
1096 			goto need_restart;
1097 
1098 		return true;
1099 	}
1100 
1101 	if (!cb->cb_holds_slot)
1102 		goto need_restart;
1103 
1104 	switch (cb->cb_seq_status) {
1105 	case 0:
1106 		/*
1107 		 * No need for lock, access serialized in nfsd4_cb_prepare
1108 		 *
1109 		 * RFC5661 20.9.3
1110 		 * If CB_SEQUENCE returns an error, then the state of the slot
1111 		 * (sequence ID, cached reply) MUST NOT change.
1112 		 */
1113 		++session->se_cb_seq_nr;
1114 		break;
1115 	case -ESERVERFAULT:
1116 		++session->se_cb_seq_nr;
1117 		/* Fall through */
1118 	case 1:
1119 	case -NFS4ERR_BADSESSION:
1120 		nfsd4_mark_cb_fault(cb->cb_clp, cb->cb_seq_status);
1121 		ret = false;
1122 		break;
1123 	case -NFS4ERR_DELAY:
1124 		if (!rpc_restart_call(task))
1125 			goto out;
1126 
1127 		rpc_delay(task, 2 * HZ);
1128 		return false;
1129 	case -NFS4ERR_BADSLOT:
1130 		goto retry_nowait;
1131 	case -NFS4ERR_SEQ_MISORDERED:
1132 		if (session->se_cb_seq_nr != 1) {
1133 			session->se_cb_seq_nr = 1;
1134 			goto retry_nowait;
1135 		}
1136 		break;
1137 	default:
1138 		nfsd4_mark_cb_fault(cb->cb_clp, cb->cb_seq_status);
1139 		dprintk("%s: unprocessed error %d\n", __func__,
1140 			cb->cb_seq_status);
1141 	}
1142 
1143 	nfsd41_cb_release_slot(cb);
1144 	dprintk("%s: freed slot, new seqid=%d\n", __func__,
1145 		clp->cl_cb_session->se_cb_seq_nr);
1146 
1147 	if (RPC_SIGNALLED(task))
1148 		goto need_restart;
1149 out:
1150 	return ret;
1151 retry_nowait:
1152 	if (rpc_restart_call_prepare(task))
1153 		ret = false;
1154 	goto out;
1155 need_restart:
1156 	if (!test_bit(NFSD4_CLIENT_CB_KILL, &clp->cl_flags)) {
1157 		task->tk_status = 0;
1158 		cb->cb_need_restart = true;
1159 	}
1160 	return false;
1161 }
1162 
1163 static void nfsd4_cb_done(struct rpc_task *task, void *calldata)
1164 {
1165 	struct nfsd4_callback *cb = calldata;
1166 	struct nfs4_client *clp = cb->cb_clp;
1167 
1168 	dprintk("%s: minorversion=%d\n", __func__,
1169 		clp->cl_minorversion);
1170 
1171 	if (!nfsd4_cb_sequence_done(task, cb))
1172 		return;
1173 
1174 	if (cb->cb_status) {
1175 		WARN_ON_ONCE(task->tk_status);
1176 		task->tk_status = cb->cb_status;
1177 	}
1178 
1179 	switch (cb->cb_ops->done(cb, task)) {
1180 	case 0:
1181 		task->tk_status = 0;
1182 		rpc_restart_call_prepare(task);
1183 		return;
1184 	case 1:
1185 		switch (task->tk_status) {
1186 		case -EIO:
1187 		case -ETIMEDOUT:
1188 			nfsd4_mark_cb_down(clp, task->tk_status);
1189 		}
1190 		break;
1191 	default:
1192 		BUG();
1193 	}
1194 }
1195 
1196 static void nfsd4_cb_release(void *calldata)
1197 {
1198 	struct nfsd4_callback *cb = calldata;
1199 
1200 	if (cb->cb_need_restart)
1201 		nfsd4_queue_cb(cb);
1202 	else
1203 		nfsd41_destroy_cb(cb);
1204 
1205 }
1206 
1207 static const struct rpc_call_ops nfsd4_cb_ops = {
1208 	.rpc_call_prepare = nfsd4_cb_prepare,
1209 	.rpc_call_done = nfsd4_cb_done,
1210 	.rpc_release = nfsd4_cb_release,
1211 };
1212 
1213 int nfsd4_create_callback_queue(void)
1214 {
1215 	callback_wq = alloc_ordered_workqueue("nfsd4_callbacks", 0);
1216 	if (!callback_wq)
1217 		return -ENOMEM;
1218 	return 0;
1219 }
1220 
1221 void nfsd4_destroy_callback_queue(void)
1222 {
1223 	destroy_workqueue(callback_wq);
1224 }
1225 
1226 /* must be called under the state lock */
1227 void nfsd4_shutdown_callback(struct nfs4_client *clp)
1228 {
1229 	set_bit(NFSD4_CLIENT_CB_KILL, &clp->cl_flags);
1230 	/*
1231 	 * Note this won't actually result in a null callback;
1232 	 * instead, nfsd4_run_cb_null() will detect the killed
1233 	 * client, destroy the rpc client, and stop:
1234 	 */
1235 	nfsd4_run_cb(&clp->cl_cb_null);
1236 	flush_workqueue(callback_wq);
1237 	nfsd41_cb_inflight_wait_complete(clp);
1238 }
1239 
1240 /* requires cl_lock: */
1241 static struct nfsd4_conn * __nfsd4_find_backchannel(struct nfs4_client *clp)
1242 {
1243 	struct nfsd4_session *s;
1244 	struct nfsd4_conn *c;
1245 
1246 	list_for_each_entry(s, &clp->cl_sessions, se_perclnt) {
1247 		list_for_each_entry(c, &s->se_conns, cn_persession) {
1248 			if (c->cn_flags & NFS4_CDFC4_BACK)
1249 				return c;
1250 		}
1251 	}
1252 	return NULL;
1253 }
1254 
1255 /*
1256  * Note there isn't a lot of locking in this code; instead we depend on
1257  * the fact that it is run from the callback_wq, which won't run two
1258  * work items at once.  So, for example, callback_wq handles all access
1259  * of cl_cb_client and all calls to rpc_create or rpc_shutdown_client.
1260  */
1261 static void nfsd4_process_cb_update(struct nfsd4_callback *cb)
1262 {
1263 	struct nfs4_cb_conn conn;
1264 	struct nfs4_client *clp = cb->cb_clp;
1265 	struct nfsd4_session *ses = NULL;
1266 	struct nfsd4_conn *c;
1267 	int err;
1268 
1269 	/*
1270 	 * This is either an update, or the client dying; in either case,
1271 	 * kill the old client:
1272 	 */
1273 	if (clp->cl_cb_client) {
1274 		rpc_shutdown_client(clp->cl_cb_client);
1275 		clp->cl_cb_client = NULL;
1276 		put_cred(clp->cl_cb_cred);
1277 		clp->cl_cb_cred = NULL;
1278 	}
1279 	if (clp->cl_cb_conn.cb_xprt) {
1280 		svc_xprt_put(clp->cl_cb_conn.cb_xprt);
1281 		clp->cl_cb_conn.cb_xprt = NULL;
1282 	}
1283 	if (test_bit(NFSD4_CLIENT_CB_KILL, &clp->cl_flags))
1284 		return;
1285 	spin_lock(&clp->cl_lock);
1286 	/*
1287 	 * Only serialized callback code is allowed to clear these
1288 	 * flags; main nfsd code can only set them:
1289 	 */
1290 	BUG_ON(!(clp->cl_flags & NFSD4_CLIENT_CB_FLAG_MASK));
1291 	clear_bit(NFSD4_CLIENT_CB_UPDATE, &clp->cl_flags);
1292 	memcpy(&conn, &cb->cb_clp->cl_cb_conn, sizeof(struct nfs4_cb_conn));
1293 	c = __nfsd4_find_backchannel(clp);
1294 	if (c) {
1295 		svc_xprt_get(c->cn_xprt);
1296 		conn.cb_xprt = c->cn_xprt;
1297 		ses = c->cn_session;
1298 	}
1299 	spin_unlock(&clp->cl_lock);
1300 
1301 	err = setup_callback_client(clp, &conn, ses);
1302 	if (err) {
1303 		nfsd4_mark_cb_down(clp, err);
1304 		return;
1305 	}
1306 }
1307 
1308 static void
1309 nfsd4_run_cb_work(struct work_struct *work)
1310 {
1311 	struct nfsd4_callback *cb =
1312 		container_of(work, struct nfsd4_callback, cb_work);
1313 	struct nfs4_client *clp = cb->cb_clp;
1314 	struct rpc_clnt *clnt;
1315 
1316 	if (cb->cb_need_restart) {
1317 		cb->cb_need_restart = false;
1318 	} else {
1319 		if (cb->cb_ops && cb->cb_ops->prepare)
1320 			cb->cb_ops->prepare(cb);
1321 	}
1322 
1323 	if (clp->cl_flags & NFSD4_CLIENT_CB_FLAG_MASK)
1324 		nfsd4_process_cb_update(cb);
1325 
1326 	clnt = clp->cl_cb_client;
1327 	if (!clnt) {
1328 		/* Callback channel broken, or client killed; give up: */
1329 		nfsd41_destroy_cb(cb);
1330 		return;
1331 	}
1332 
1333 	/*
1334 	 * Don't send probe messages for 4.1 or later.
1335 	 */
1336 	if (!cb->cb_ops && clp->cl_minorversion) {
1337 		clp->cl_cb_state = NFSD4_CB_UP;
1338 		nfsd41_destroy_cb(cb);
1339 		return;
1340 	}
1341 
1342 	cb->cb_msg.rpc_cred = clp->cl_cb_cred;
1343 	rpc_call_async(clnt, &cb->cb_msg, RPC_TASK_SOFT | RPC_TASK_SOFTCONN,
1344 			cb->cb_ops ? &nfsd4_cb_ops : &nfsd4_cb_probe_ops, cb);
1345 }
1346 
1347 void nfsd4_init_cb(struct nfsd4_callback *cb, struct nfs4_client *clp,
1348 		const struct nfsd4_callback_ops *ops, enum nfsd4_cb_op op)
1349 {
1350 	cb->cb_clp = clp;
1351 	cb->cb_msg.rpc_proc = &nfs4_cb_procedures[op];
1352 	cb->cb_msg.rpc_argp = cb;
1353 	cb->cb_msg.rpc_resp = cb;
1354 	cb->cb_ops = ops;
1355 	INIT_WORK(&cb->cb_work, nfsd4_run_cb_work);
1356 	cb->cb_seq_status = 1;
1357 	cb->cb_status = 0;
1358 	cb->cb_need_restart = false;
1359 	cb->cb_holds_slot = false;
1360 }
1361 
1362 void nfsd4_run_cb(struct nfsd4_callback *cb)
1363 {
1364 	struct nfs4_client *clp = cb->cb_clp;
1365 
1366 	nfsd41_cb_inflight_begin(clp);
1367 	if (!nfsd4_queue_cb(cb))
1368 		nfsd41_cb_inflight_end(clp);
1369 }
1370