xref: /openbmc/linux/fs/nfsd/nfs4callback.c (revision 2455f0e1)
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 "trace.h"
42 #include "xdr4cb.h"
43 #include "xdr4.h"
44 
45 #define NFSDDBG_FACILITY                NFSDDBG_PROC
46 
47 static void nfsd4_mark_cb_fault(struct nfs4_client *, int reason);
48 
49 #define NFSPROC4_CB_NULL 0
50 #define NFSPROC4_CB_COMPOUND 1
51 
52 /* Index of predefined Linux callback client operations */
53 
54 struct 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 
64 static __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 
79 static void encode_uint32(struct xdr_stream *xdr, u32 n)
80 {
81 	WARN_ON_ONCE(xdr_stream_encode_u32(xdr, n) < 0);
82 }
83 
84 static void encode_bitmap4(struct xdr_stream *xdr, const __u32 *bitmap,
85 			   size_t len)
86 {
87 	WARN_ON_ONCE(xdr_stream_encode_uint32_array(xdr, bitmap, len) < 0);
88 }
89 
90 /*
91  *	nfs_cb_opnum4
92  *
93  *	enum nfs_cb_opnum4 {
94  *		OP_CB_GETATTR		= 3,
95  *		  ...
96  *	};
97  */
98 enum nfs_cb_opnum4 {
99 	OP_CB_GETATTR			= 3,
100 	OP_CB_RECALL			= 4,
101 	OP_CB_LAYOUTRECALL		= 5,
102 	OP_CB_NOTIFY			= 6,
103 	OP_CB_PUSH_DELEG		= 7,
104 	OP_CB_RECALL_ANY		= 8,
105 	OP_CB_RECALLABLE_OBJ_AVAIL	= 9,
106 	OP_CB_RECALL_SLOT		= 10,
107 	OP_CB_SEQUENCE			= 11,
108 	OP_CB_WANTS_CANCELLED		= 12,
109 	OP_CB_NOTIFY_LOCK		= 13,
110 	OP_CB_NOTIFY_DEVICEID		= 14,
111 	OP_CB_OFFLOAD			= 15,
112 	OP_CB_ILLEGAL			= 10044
113 };
114 
115 static void encode_nfs_cb_opnum4(struct xdr_stream *xdr, enum nfs_cb_opnum4 op)
116 {
117 	__be32 *p;
118 
119 	p = xdr_reserve_space(xdr, 4);
120 	*p = cpu_to_be32(op);
121 }
122 
123 /*
124  * nfs_fh4
125  *
126  *	typedef opaque nfs_fh4<NFS4_FHSIZE>;
127  */
128 static void encode_nfs_fh4(struct xdr_stream *xdr, const struct knfsd_fh *fh)
129 {
130 	u32 length = fh->fh_size;
131 	__be32 *p;
132 
133 	BUG_ON(length > NFS4_FHSIZE);
134 	p = xdr_reserve_space(xdr, 4 + length);
135 	xdr_encode_opaque(p, &fh->fh_raw, length);
136 }
137 
138 /*
139  * stateid4
140  *
141  *	struct stateid4 {
142  *		uint32_t	seqid;
143  *		opaque		other[12];
144  *	};
145  */
146 static void encode_stateid4(struct xdr_stream *xdr, const stateid_t *sid)
147 {
148 	__be32 *p;
149 
150 	p = xdr_reserve_space(xdr, NFS4_STATEID_SIZE);
151 	*p++ = cpu_to_be32(sid->si_generation);
152 	xdr_encode_opaque_fixed(p, &sid->si_opaque, NFS4_STATEID_OTHER_SIZE);
153 }
154 
155 /*
156  * sessionid4
157  *
158  *	typedef opaque sessionid4[NFS4_SESSIONID_SIZE];
159  */
160 static void encode_sessionid4(struct xdr_stream *xdr,
161 			      const struct nfsd4_session *session)
162 {
163 	__be32 *p;
164 
165 	p = xdr_reserve_space(xdr, NFS4_MAX_SESSIONID_LEN);
166 	xdr_encode_opaque_fixed(p, session->se_sessionid.data,
167 					NFS4_MAX_SESSIONID_LEN);
168 }
169 
170 /*
171  * nfsstat4
172  */
173 static const struct {
174 	int stat;
175 	int errno;
176 } nfs_cb_errtbl[] = {
177 	{ NFS4_OK,		0		},
178 	{ NFS4ERR_PERM,		-EPERM		},
179 	{ NFS4ERR_NOENT,	-ENOENT		},
180 	{ NFS4ERR_IO,		-EIO		},
181 	{ NFS4ERR_NXIO,		-ENXIO		},
182 	{ NFS4ERR_ACCESS,	-EACCES		},
183 	{ NFS4ERR_EXIST,	-EEXIST		},
184 	{ NFS4ERR_XDEV,		-EXDEV		},
185 	{ NFS4ERR_NOTDIR,	-ENOTDIR	},
186 	{ NFS4ERR_ISDIR,	-EISDIR		},
187 	{ NFS4ERR_INVAL,	-EINVAL		},
188 	{ NFS4ERR_FBIG,		-EFBIG		},
189 	{ NFS4ERR_NOSPC,	-ENOSPC		},
190 	{ NFS4ERR_ROFS,		-EROFS		},
191 	{ NFS4ERR_MLINK,	-EMLINK		},
192 	{ NFS4ERR_NAMETOOLONG,	-ENAMETOOLONG	},
193 	{ NFS4ERR_NOTEMPTY,	-ENOTEMPTY	},
194 	{ NFS4ERR_DQUOT,	-EDQUOT		},
195 	{ NFS4ERR_STALE,	-ESTALE		},
196 	{ NFS4ERR_BADHANDLE,	-EBADHANDLE	},
197 	{ NFS4ERR_BAD_COOKIE,	-EBADCOOKIE	},
198 	{ NFS4ERR_NOTSUPP,	-ENOTSUPP	},
199 	{ NFS4ERR_TOOSMALL,	-ETOOSMALL	},
200 	{ NFS4ERR_SERVERFAULT,	-ESERVERFAULT	},
201 	{ NFS4ERR_BADTYPE,	-EBADTYPE	},
202 	{ NFS4ERR_LOCKED,	-EAGAIN		},
203 	{ NFS4ERR_RESOURCE,	-EREMOTEIO	},
204 	{ NFS4ERR_SYMLINK,	-ELOOP		},
205 	{ NFS4ERR_OP_ILLEGAL,	-EOPNOTSUPP	},
206 	{ NFS4ERR_DEADLOCK,	-EDEADLK	},
207 	{ -1,			-EIO		}
208 };
209 
210 /*
211  * If we cannot translate the error, the recovery routines should
212  * handle it.
213  *
214  * Note: remaining NFSv4 error codes have values > 10000, so should
215  * not conflict with native Linux error codes.
216  */
217 static int nfs_cb_stat_to_errno(int status)
218 {
219 	int i;
220 
221 	for (i = 0; nfs_cb_errtbl[i].stat != -1; i++) {
222 		if (nfs_cb_errtbl[i].stat == status)
223 			return nfs_cb_errtbl[i].errno;
224 	}
225 
226 	dprintk("NFSD: Unrecognized NFS CB status value: %u\n", status);
227 	return -status;
228 }
229 
230 static int decode_cb_op_status(struct xdr_stream *xdr,
231 			       enum nfs_cb_opnum4 expected, int *status)
232 {
233 	__be32 *p;
234 	u32 op;
235 
236 	p = xdr_inline_decode(xdr, 4 + 4);
237 	if (unlikely(p == NULL))
238 		goto out_overflow;
239 	op = be32_to_cpup(p++);
240 	if (unlikely(op != expected))
241 		goto out_unexpected;
242 	*status = nfs_cb_stat_to_errno(be32_to_cpup(p));
243 	return 0;
244 out_overflow:
245 	return -EIO;
246 out_unexpected:
247 	dprintk("NFSD: Callback server returned operation %d but "
248 		"we issued a request for %d\n", op, expected);
249 	return -EIO;
250 }
251 
252 /*
253  * CB_COMPOUND4args
254  *
255  *	struct CB_COMPOUND4args {
256  *		utf8str_cs	tag;
257  *		uint32_t	minorversion;
258  *		uint32_t	callback_ident;
259  *		nfs_cb_argop4	argarray<>;
260  *	};
261 */
262 static void encode_cb_compound4args(struct xdr_stream *xdr,
263 				    struct nfs4_cb_compound_hdr *hdr)
264 {
265 	__be32 * p;
266 
267 	p = xdr_reserve_space(xdr, 4 + 4 + 4 + 4);
268 	p = xdr_encode_empty_array(p);		/* empty tag */
269 	*p++ = cpu_to_be32(hdr->minorversion);
270 	*p++ = cpu_to_be32(hdr->ident);
271 
272 	hdr->nops_p = p;
273 	*p = cpu_to_be32(hdr->nops);		/* argarray element count */
274 }
275 
276 /*
277  * Update argarray element count
278  */
279 static void encode_cb_nops(struct nfs4_cb_compound_hdr *hdr)
280 {
281 	BUG_ON(hdr->nops > NFS4_MAX_BACK_CHANNEL_OPS);
282 	*hdr->nops_p = cpu_to_be32(hdr->nops);
283 }
284 
285 /*
286  * CB_COMPOUND4res
287  *
288  *	struct CB_COMPOUND4res {
289  *		nfsstat4	status;
290  *		utf8str_cs	tag;
291  *		nfs_cb_resop4	resarray<>;
292  *	};
293  */
294 static int decode_cb_compound4res(struct xdr_stream *xdr,
295 				  struct nfs4_cb_compound_hdr *hdr)
296 {
297 	u32 length;
298 	__be32 *p;
299 
300 	p = xdr_inline_decode(xdr, 4 + 4);
301 	if (unlikely(p == NULL))
302 		goto out_overflow;
303 	hdr->status = be32_to_cpup(p++);
304 	/* Ignore the tag */
305 	length = be32_to_cpup(p++);
306 	p = xdr_inline_decode(xdr, length + 4);
307 	if (unlikely(p == NULL))
308 		goto out_overflow;
309 	p += XDR_QUADLEN(length);
310 	hdr->nops = be32_to_cpup(p);
311 	return 0;
312 out_overflow:
313 	return -EIO;
314 }
315 
316 /*
317  * CB_RECALL4args
318  *
319  *	struct CB_RECALL4args {
320  *		stateid4	stateid;
321  *		bool		truncate;
322  *		nfs_fh4		fh;
323  *	};
324  */
325 static void encode_cb_recall4args(struct xdr_stream *xdr,
326 				  const struct nfs4_delegation *dp,
327 				  struct nfs4_cb_compound_hdr *hdr)
328 {
329 	__be32 *p;
330 
331 	encode_nfs_cb_opnum4(xdr, OP_CB_RECALL);
332 	encode_stateid4(xdr, &dp->dl_stid.sc_stateid);
333 
334 	p = xdr_reserve_space(xdr, 4);
335 	*p++ = xdr_zero;			/* truncate */
336 
337 	encode_nfs_fh4(xdr, &dp->dl_stid.sc_file->fi_fhandle);
338 
339 	hdr->nops++;
340 }
341 
342 /*
343  * CB_RECALLANY4args
344  *
345  *	struct CB_RECALLANY4args {
346  *		uint32_t	craa_objects_to_keep;
347  *		bitmap4		craa_type_mask;
348  *	};
349  */
350 static void
351 encode_cb_recallany4args(struct xdr_stream *xdr,
352 	struct nfs4_cb_compound_hdr *hdr, struct nfsd4_cb_recall_any *ra)
353 {
354 	encode_nfs_cb_opnum4(xdr, OP_CB_RECALL_ANY);
355 	encode_uint32(xdr, ra->ra_keep);
356 	encode_bitmap4(xdr, ra->ra_bmval, ARRAY_SIZE(ra->ra_bmval));
357 	hdr->nops++;
358 }
359 
360 /*
361  * CB_SEQUENCE4args
362  *
363  *	struct CB_SEQUENCE4args {
364  *		sessionid4		csa_sessionid;
365  *		sequenceid4		csa_sequenceid;
366  *		slotid4			csa_slotid;
367  *		slotid4			csa_highest_slotid;
368  *		bool			csa_cachethis;
369  *		referring_call_list4	csa_referring_call_lists<>;
370  *	};
371  */
372 static void encode_cb_sequence4args(struct xdr_stream *xdr,
373 				    const struct nfsd4_callback *cb,
374 				    struct nfs4_cb_compound_hdr *hdr)
375 {
376 	struct nfsd4_session *session = cb->cb_clp->cl_cb_session;
377 	__be32 *p;
378 
379 	if (hdr->minorversion == 0)
380 		return;
381 
382 	encode_nfs_cb_opnum4(xdr, OP_CB_SEQUENCE);
383 	encode_sessionid4(xdr, session);
384 
385 	p = xdr_reserve_space(xdr, 4 + 4 + 4 + 4 + 4);
386 	*p++ = cpu_to_be32(session->se_cb_seq_nr);	/* csa_sequenceid */
387 	*p++ = xdr_zero;			/* csa_slotid */
388 	*p++ = xdr_zero;			/* csa_highest_slotid */
389 	*p++ = xdr_zero;			/* csa_cachethis */
390 	xdr_encode_empty_array(p);		/* csa_referring_call_lists */
391 
392 	hdr->nops++;
393 }
394 
395 /*
396  * CB_SEQUENCE4resok
397  *
398  *	struct CB_SEQUENCE4resok {
399  *		sessionid4	csr_sessionid;
400  *		sequenceid4	csr_sequenceid;
401  *		slotid4		csr_slotid;
402  *		slotid4		csr_highest_slotid;
403  *		slotid4		csr_target_highest_slotid;
404  *	};
405  *
406  *	union CB_SEQUENCE4res switch (nfsstat4 csr_status) {
407  *	case NFS4_OK:
408  *		CB_SEQUENCE4resok	csr_resok4;
409  *	default:
410  *		void;
411  *	};
412  *
413  * Our current back channel implmentation supports a single backchannel
414  * with a single slot.
415  */
416 static int decode_cb_sequence4resok(struct xdr_stream *xdr,
417 				    struct nfsd4_callback *cb)
418 {
419 	struct nfsd4_session *session = cb->cb_clp->cl_cb_session;
420 	int status = -ESERVERFAULT;
421 	__be32 *p;
422 	u32 dummy;
423 
424 	/*
425 	 * If the server returns different values for sessionID, slotID or
426 	 * sequence number, the server is looney tunes.
427 	 */
428 	p = xdr_inline_decode(xdr, NFS4_MAX_SESSIONID_LEN + 4 + 4 + 4 + 4);
429 	if (unlikely(p == NULL))
430 		goto out_overflow;
431 
432 	if (memcmp(p, session->se_sessionid.data, NFS4_MAX_SESSIONID_LEN)) {
433 		dprintk("NFS: %s Invalid session id\n", __func__);
434 		goto out;
435 	}
436 	p += XDR_QUADLEN(NFS4_MAX_SESSIONID_LEN);
437 
438 	dummy = be32_to_cpup(p++);
439 	if (dummy != session->se_cb_seq_nr) {
440 		dprintk("NFS: %s Invalid sequence number\n", __func__);
441 		goto out;
442 	}
443 
444 	dummy = be32_to_cpup(p++);
445 	if (dummy != 0) {
446 		dprintk("NFS: %s Invalid slotid\n", __func__);
447 		goto out;
448 	}
449 
450 	/*
451 	 * FIXME: process highest slotid and target highest slotid
452 	 */
453 	status = 0;
454 out:
455 	cb->cb_seq_status = status;
456 	return status;
457 out_overflow:
458 	status = -EIO;
459 	goto out;
460 }
461 
462 static int decode_cb_sequence4res(struct xdr_stream *xdr,
463 				  struct nfsd4_callback *cb)
464 {
465 	int status;
466 
467 	if (cb->cb_clp->cl_minorversion == 0)
468 		return 0;
469 
470 	status = decode_cb_op_status(xdr, OP_CB_SEQUENCE, &cb->cb_seq_status);
471 	if (unlikely(status || cb->cb_seq_status))
472 		return status;
473 
474 	return decode_cb_sequence4resok(xdr, cb);
475 }
476 
477 /*
478  * NFSv4.0 and NFSv4.1 XDR encode functions
479  *
480  * NFSv4.0 callback argument types are defined in section 15 of RFC
481  * 3530: "Network File System (NFS) version 4 Protocol" and section 20
482  * of RFC 5661:  "Network File System (NFS) Version 4 Minor Version 1
483  * Protocol".
484  */
485 
486 /*
487  * NB: Without this zero space reservation, callbacks over krb5p fail
488  */
489 static void nfs4_xdr_enc_cb_null(struct rpc_rqst *req, struct xdr_stream *xdr,
490 				 const void *__unused)
491 {
492 	xdr_reserve_space(xdr, 0);
493 }
494 
495 /*
496  * 20.2. Operation 4: CB_RECALL - Recall a Delegation
497  */
498 static void nfs4_xdr_enc_cb_recall(struct rpc_rqst *req, struct xdr_stream *xdr,
499 				   const void *data)
500 {
501 	const struct nfsd4_callback *cb = data;
502 	const struct nfs4_delegation *dp = cb_to_delegation(cb);
503 	struct nfs4_cb_compound_hdr hdr = {
504 		.ident = cb->cb_clp->cl_cb_ident,
505 		.minorversion = cb->cb_clp->cl_minorversion,
506 	};
507 
508 	encode_cb_compound4args(xdr, &hdr);
509 	encode_cb_sequence4args(xdr, cb, &hdr);
510 	encode_cb_recall4args(xdr, dp, &hdr);
511 	encode_cb_nops(&hdr);
512 }
513 
514 /*
515  * 20.6. Operation 8: CB_RECALL_ANY - Keep Any N Recallable Objects
516  */
517 static void
518 nfs4_xdr_enc_cb_recall_any(struct rpc_rqst *req,
519 		struct xdr_stream *xdr, const void *data)
520 {
521 	const struct nfsd4_callback *cb = data;
522 	struct nfsd4_cb_recall_any *ra;
523 	struct nfs4_cb_compound_hdr hdr = {
524 		.ident = cb->cb_clp->cl_cb_ident,
525 		.minorversion = cb->cb_clp->cl_minorversion,
526 	};
527 
528 	ra = container_of(cb, struct nfsd4_cb_recall_any, ra_cb);
529 	encode_cb_compound4args(xdr, &hdr);
530 	encode_cb_sequence4args(xdr, cb, &hdr);
531 	encode_cb_recallany4args(xdr, &hdr, ra);
532 	encode_cb_nops(&hdr);
533 }
534 
535 /*
536  * NFSv4.0 and NFSv4.1 XDR decode functions
537  *
538  * NFSv4.0 callback result types are defined in section 15 of RFC
539  * 3530: "Network File System (NFS) version 4 Protocol" and section 20
540  * of RFC 5661:  "Network File System (NFS) Version 4 Minor Version 1
541  * Protocol".
542  */
543 
544 static int nfs4_xdr_dec_cb_null(struct rpc_rqst *req, struct xdr_stream *xdr,
545 				void *__unused)
546 {
547 	return 0;
548 }
549 
550 /*
551  * 20.2. Operation 4: CB_RECALL - Recall a Delegation
552  */
553 static int nfs4_xdr_dec_cb_recall(struct rpc_rqst *rqstp,
554 				  struct xdr_stream *xdr,
555 				  void *data)
556 {
557 	struct nfsd4_callback *cb = data;
558 	struct nfs4_cb_compound_hdr hdr;
559 	int status;
560 
561 	status = decode_cb_compound4res(xdr, &hdr);
562 	if (unlikely(status))
563 		return status;
564 
565 	status = decode_cb_sequence4res(xdr, cb);
566 	if (unlikely(status || cb->cb_seq_status))
567 		return status;
568 
569 	return decode_cb_op_status(xdr, OP_CB_RECALL, &cb->cb_status);
570 }
571 
572 /*
573  * 20.6. Operation 8: CB_RECALL_ANY - Keep Any N Recallable Objects
574  */
575 static int
576 nfs4_xdr_dec_cb_recall_any(struct rpc_rqst *rqstp,
577 				  struct xdr_stream *xdr,
578 				  void *data)
579 {
580 	struct nfsd4_callback *cb = data;
581 	struct nfs4_cb_compound_hdr hdr;
582 	int status;
583 
584 	status = decode_cb_compound4res(xdr, &hdr);
585 	if (unlikely(status))
586 		return status;
587 	status = decode_cb_sequence4res(xdr, cb);
588 	if (unlikely(status || cb->cb_seq_status))
589 		return status;
590 	status =  decode_cb_op_status(xdr, OP_CB_RECALL_ANY, &cb->cb_status);
591 	return status;
592 }
593 
594 #ifdef CONFIG_NFSD_PNFS
595 /*
596  * CB_LAYOUTRECALL4args
597  *
598  *	struct layoutrecall_file4 {
599  *		nfs_fh4         lor_fh;
600  *		offset4         lor_offset;
601  *		length4         lor_length;
602  *		stateid4        lor_stateid;
603  *	};
604  *
605  *	union layoutrecall4 switch(layoutrecall_type4 lor_recalltype) {
606  *	case LAYOUTRECALL4_FILE:
607  *		layoutrecall_file4 lor_layout;
608  *	case LAYOUTRECALL4_FSID:
609  *		fsid4              lor_fsid;
610  *	case LAYOUTRECALL4_ALL:
611  *		void;
612  *	};
613  *
614  *	struct CB_LAYOUTRECALL4args {
615  *		layouttype4             clora_type;
616  *		layoutiomode4           clora_iomode;
617  *		bool                    clora_changed;
618  *		layoutrecall4           clora_recall;
619  *	};
620  */
621 static void encode_cb_layout4args(struct xdr_stream *xdr,
622 				  const struct nfs4_layout_stateid *ls,
623 				  struct nfs4_cb_compound_hdr *hdr)
624 {
625 	__be32 *p;
626 
627 	BUG_ON(hdr->minorversion == 0);
628 
629 	p = xdr_reserve_space(xdr, 5 * 4);
630 	*p++ = cpu_to_be32(OP_CB_LAYOUTRECALL);
631 	*p++ = cpu_to_be32(ls->ls_layout_type);
632 	*p++ = cpu_to_be32(IOMODE_ANY);
633 	*p++ = cpu_to_be32(1);
634 	*p = cpu_to_be32(RETURN_FILE);
635 
636 	encode_nfs_fh4(xdr, &ls->ls_stid.sc_file->fi_fhandle);
637 
638 	p = xdr_reserve_space(xdr, 2 * 8);
639 	p = xdr_encode_hyper(p, 0);
640 	xdr_encode_hyper(p, NFS4_MAX_UINT64);
641 
642 	encode_stateid4(xdr, &ls->ls_recall_sid);
643 
644 	hdr->nops++;
645 }
646 
647 static void nfs4_xdr_enc_cb_layout(struct rpc_rqst *req,
648 				   struct xdr_stream *xdr,
649 				   const void *data)
650 {
651 	const struct nfsd4_callback *cb = data;
652 	const struct nfs4_layout_stateid *ls =
653 		container_of(cb, struct nfs4_layout_stateid, ls_recall);
654 	struct nfs4_cb_compound_hdr hdr = {
655 		.ident = 0,
656 		.minorversion = cb->cb_clp->cl_minorversion,
657 	};
658 
659 	encode_cb_compound4args(xdr, &hdr);
660 	encode_cb_sequence4args(xdr, cb, &hdr);
661 	encode_cb_layout4args(xdr, ls, &hdr);
662 	encode_cb_nops(&hdr);
663 }
664 
665 static int nfs4_xdr_dec_cb_layout(struct rpc_rqst *rqstp,
666 				  struct xdr_stream *xdr,
667 				  void *data)
668 {
669 	struct nfsd4_callback *cb = data;
670 	struct nfs4_cb_compound_hdr hdr;
671 	int status;
672 
673 	status = decode_cb_compound4res(xdr, &hdr);
674 	if (unlikely(status))
675 		return status;
676 
677 	status = decode_cb_sequence4res(xdr, cb);
678 	if (unlikely(status || cb->cb_seq_status))
679 		return status;
680 
681 	return decode_cb_op_status(xdr, OP_CB_LAYOUTRECALL, &cb->cb_status);
682 }
683 #endif /* CONFIG_NFSD_PNFS */
684 
685 static void encode_stateowner(struct xdr_stream *xdr, struct nfs4_stateowner *so)
686 {
687 	__be32	*p;
688 
689 	p = xdr_reserve_space(xdr, 8 + 4 + so->so_owner.len);
690 	p = xdr_encode_opaque_fixed(p, &so->so_client->cl_clientid, 8);
691 	xdr_encode_opaque(p, so->so_owner.data, so->so_owner.len);
692 }
693 
694 static void nfs4_xdr_enc_cb_notify_lock(struct rpc_rqst *req,
695 					struct xdr_stream *xdr,
696 					const void *data)
697 {
698 	const struct nfsd4_callback *cb = data;
699 	const struct nfsd4_blocked_lock *nbl =
700 		container_of(cb, struct nfsd4_blocked_lock, nbl_cb);
701 	struct nfs4_lockowner *lo = (struct nfs4_lockowner *)nbl->nbl_lock.fl_owner;
702 	struct nfs4_cb_compound_hdr hdr = {
703 		.ident = 0,
704 		.minorversion = cb->cb_clp->cl_minorversion,
705 	};
706 
707 	__be32 *p;
708 
709 	BUG_ON(hdr.minorversion == 0);
710 
711 	encode_cb_compound4args(xdr, &hdr);
712 	encode_cb_sequence4args(xdr, cb, &hdr);
713 
714 	p = xdr_reserve_space(xdr, 4);
715 	*p = cpu_to_be32(OP_CB_NOTIFY_LOCK);
716 	encode_nfs_fh4(xdr, &nbl->nbl_fh);
717 	encode_stateowner(xdr, &lo->lo_owner);
718 	hdr.nops++;
719 
720 	encode_cb_nops(&hdr);
721 }
722 
723 static int nfs4_xdr_dec_cb_notify_lock(struct rpc_rqst *rqstp,
724 					struct xdr_stream *xdr,
725 					void *data)
726 {
727 	struct nfsd4_callback *cb = data;
728 	struct nfs4_cb_compound_hdr hdr;
729 	int status;
730 
731 	status = decode_cb_compound4res(xdr, &hdr);
732 	if (unlikely(status))
733 		return status;
734 
735 	status = decode_cb_sequence4res(xdr, cb);
736 	if (unlikely(status || cb->cb_seq_status))
737 		return status;
738 
739 	return decode_cb_op_status(xdr, OP_CB_NOTIFY_LOCK, &cb->cb_status);
740 }
741 
742 /*
743  * struct write_response4 {
744  *	stateid4	wr_callback_id<1>;
745  *	length4		wr_count;
746  *	stable_how4	wr_committed;
747  *	verifier4	wr_writeverf;
748  * };
749  * union offload_info4 switch (nfsstat4 coa_status) {
750  *	case NFS4_OK:
751  *		write_response4	coa_resok4;
752  *	default:
753  *		length4		coa_bytes_copied;
754  * };
755  * struct CB_OFFLOAD4args {
756  *	nfs_fh4		coa_fh;
757  *	stateid4	coa_stateid;
758  *	offload_info4	coa_offload_info;
759  * };
760  */
761 static void encode_offload_info4(struct xdr_stream *xdr,
762 				 const struct nfsd4_cb_offload *cbo)
763 {
764 	__be32 *p;
765 
766 	p = xdr_reserve_space(xdr, 4);
767 	*p = cbo->co_nfserr;
768 	switch (cbo->co_nfserr) {
769 	case nfs_ok:
770 		p = xdr_reserve_space(xdr, 4 + 8 + 4 + NFS4_VERIFIER_SIZE);
771 		p = xdr_encode_empty_array(p);
772 		p = xdr_encode_hyper(p, cbo->co_res.wr_bytes_written);
773 		*p++ = cpu_to_be32(cbo->co_res.wr_stable_how);
774 		p = xdr_encode_opaque_fixed(p, cbo->co_res.wr_verifier.data,
775 					    NFS4_VERIFIER_SIZE);
776 		break;
777 	default:
778 		p = xdr_reserve_space(xdr, 8);
779 		/* We always return success if bytes were written */
780 		p = xdr_encode_hyper(p, 0);
781 	}
782 }
783 
784 static void encode_cb_offload4args(struct xdr_stream *xdr,
785 				   const struct nfsd4_cb_offload *cbo,
786 				   struct nfs4_cb_compound_hdr *hdr)
787 {
788 	__be32 *p;
789 
790 	p = xdr_reserve_space(xdr, 4);
791 	*p = cpu_to_be32(OP_CB_OFFLOAD);
792 	encode_nfs_fh4(xdr, &cbo->co_fh);
793 	encode_stateid4(xdr, &cbo->co_res.cb_stateid);
794 	encode_offload_info4(xdr, cbo);
795 
796 	hdr->nops++;
797 }
798 
799 static void nfs4_xdr_enc_cb_offload(struct rpc_rqst *req,
800 				    struct xdr_stream *xdr,
801 				    const void *data)
802 {
803 	const struct nfsd4_callback *cb = data;
804 	const struct nfsd4_cb_offload *cbo =
805 		container_of(cb, struct nfsd4_cb_offload, co_cb);
806 	struct nfs4_cb_compound_hdr hdr = {
807 		.ident = 0,
808 		.minorversion = cb->cb_clp->cl_minorversion,
809 	};
810 
811 	encode_cb_compound4args(xdr, &hdr);
812 	encode_cb_sequence4args(xdr, cb, &hdr);
813 	encode_cb_offload4args(xdr, cbo, &hdr);
814 	encode_cb_nops(&hdr);
815 }
816 
817 static int nfs4_xdr_dec_cb_offload(struct rpc_rqst *rqstp,
818 				   struct xdr_stream *xdr,
819 				   void *data)
820 {
821 	struct nfsd4_callback *cb = data;
822 	struct nfs4_cb_compound_hdr hdr;
823 	int status;
824 
825 	status = decode_cb_compound4res(xdr, &hdr);
826 	if (unlikely(status))
827 		return status;
828 
829 	status = decode_cb_sequence4res(xdr, cb);
830 	if (unlikely(status || cb->cb_seq_status))
831 		return status;
832 
833 	return decode_cb_op_status(xdr, OP_CB_OFFLOAD, &cb->cb_status);
834 }
835 /*
836  * RPC procedure tables
837  */
838 #define PROC(proc, call, argtype, restype)				\
839 [NFSPROC4_CLNT_##proc] = {						\
840 	.p_proc    = NFSPROC4_CB_##call,				\
841 	.p_encode  = nfs4_xdr_enc_##argtype,		\
842 	.p_decode  = nfs4_xdr_dec_##restype,				\
843 	.p_arglen  = NFS4_enc_##argtype##_sz,				\
844 	.p_replen  = NFS4_dec_##restype##_sz,				\
845 	.p_statidx = NFSPROC4_CB_##call,				\
846 	.p_name    = #proc,						\
847 }
848 
849 static const struct rpc_procinfo nfs4_cb_procedures[] = {
850 	PROC(CB_NULL,	NULL,		cb_null,	cb_null),
851 	PROC(CB_RECALL,	COMPOUND,	cb_recall,	cb_recall),
852 #ifdef CONFIG_NFSD_PNFS
853 	PROC(CB_LAYOUT,	COMPOUND,	cb_layout,	cb_layout),
854 #endif
855 	PROC(CB_NOTIFY_LOCK,	COMPOUND,	cb_notify_lock,	cb_notify_lock),
856 	PROC(CB_OFFLOAD,	COMPOUND,	cb_offload,	cb_offload),
857 	PROC(CB_RECALL_ANY,	COMPOUND,	cb_recall_any,	cb_recall_any),
858 };
859 
860 static unsigned int nfs4_cb_counts[ARRAY_SIZE(nfs4_cb_procedures)];
861 static const struct rpc_version nfs_cb_version4 = {
862 /*
863  * Note on the callback rpc program version number: despite language in rfc
864  * 5661 section 18.36.3 requiring servers to use 4 in this field, the
865  * official xdr descriptions for both 4.0 and 4.1 specify version 1, and
866  * in practice that appears to be what implementations use.  The section
867  * 18.36.3 language is expected to be fixed in an erratum.
868  */
869 	.number			= 1,
870 	.nrprocs		= ARRAY_SIZE(nfs4_cb_procedures),
871 	.procs			= nfs4_cb_procedures,
872 	.counts			= nfs4_cb_counts,
873 };
874 
875 static const struct rpc_version *nfs_cb_version[2] = {
876 	[1] = &nfs_cb_version4,
877 };
878 
879 static const struct rpc_program cb_program;
880 
881 static struct rpc_stat cb_stats = {
882 	.program		= &cb_program
883 };
884 
885 #define NFS4_CALLBACK 0x40000000
886 static const struct rpc_program cb_program = {
887 	.name			= "nfs4_cb",
888 	.number			= NFS4_CALLBACK,
889 	.nrvers			= ARRAY_SIZE(nfs_cb_version),
890 	.version		= nfs_cb_version,
891 	.stats			= &cb_stats,
892 	.pipe_dir_name		= "nfsd4_cb",
893 };
894 
895 static int max_cb_time(struct net *net)
896 {
897 	struct nfsd_net *nn = net_generic(net, nfsd_net_id);
898 
899 	/*
900 	 * nfsd4_lease is set to at most one hour in __nfsd4_write_time,
901 	 * so we can use 32-bit math on it. Warn if that assumption
902 	 * ever stops being true.
903 	 */
904 	if (WARN_ON_ONCE(nn->nfsd4_lease > 3600))
905 		return 360 * HZ;
906 
907 	return max(((u32)nn->nfsd4_lease)/10, 1u) * HZ;
908 }
909 
910 static struct workqueue_struct *callback_wq;
911 
912 static bool nfsd4_queue_cb(struct nfsd4_callback *cb)
913 {
914 	return queue_work(callback_wq, &cb->cb_work);
915 }
916 
917 static void nfsd41_cb_inflight_begin(struct nfs4_client *clp)
918 {
919 	atomic_inc(&clp->cl_cb_inflight);
920 }
921 
922 static void nfsd41_cb_inflight_end(struct nfs4_client *clp)
923 {
924 
925 	if (atomic_dec_and_test(&clp->cl_cb_inflight))
926 		wake_up_var(&clp->cl_cb_inflight);
927 }
928 
929 static void nfsd41_cb_inflight_wait_complete(struct nfs4_client *clp)
930 {
931 	wait_var_event(&clp->cl_cb_inflight,
932 			!atomic_read(&clp->cl_cb_inflight));
933 }
934 
935 static const struct cred *get_backchannel_cred(struct nfs4_client *clp, struct rpc_clnt *client, struct nfsd4_session *ses)
936 {
937 	if (clp->cl_minorversion == 0) {
938 		client->cl_principal = clp->cl_cred.cr_targ_princ ?
939 			clp->cl_cred.cr_targ_princ : "nfs";
940 
941 		return get_cred(rpc_machine_cred());
942 	} else {
943 		struct cred *kcred;
944 
945 		kcred = prepare_kernel_cred(&init_task);
946 		if (!kcred)
947 			return NULL;
948 
949 		kcred->uid = ses->se_cb_sec.uid;
950 		kcred->gid = ses->se_cb_sec.gid;
951 		return kcred;
952 	}
953 }
954 
955 static int setup_callback_client(struct nfs4_client *clp, struct nfs4_cb_conn *conn, struct nfsd4_session *ses)
956 {
957 	int maxtime = max_cb_time(clp->net);
958 	struct rpc_timeout	timeparms = {
959 		.to_initval	= maxtime,
960 		.to_retries	= 0,
961 		.to_maxval	= maxtime,
962 	};
963 	struct rpc_create_args args = {
964 		.net		= clp->net,
965 		.address	= (struct sockaddr *) &conn->cb_addr,
966 		.addrsize	= conn->cb_addrlen,
967 		.saddress	= (struct sockaddr *) &conn->cb_saddr,
968 		.timeout	= &timeparms,
969 		.program	= &cb_program,
970 		.version	= 1,
971 		.flags		= (RPC_CLNT_CREATE_NOPING | RPC_CLNT_CREATE_QUIET),
972 		.cred		= current_cred(),
973 	};
974 	struct rpc_clnt *client;
975 	const struct cred *cred;
976 
977 	if (clp->cl_minorversion == 0) {
978 		if (!clp->cl_cred.cr_principal &&
979 		    (clp->cl_cred.cr_flavor >= RPC_AUTH_GSS_KRB5)) {
980 			trace_nfsd_cb_setup_err(clp, -EINVAL);
981 			return -EINVAL;
982 		}
983 		args.client_name = clp->cl_cred.cr_principal;
984 		args.prognumber	= conn->cb_prog;
985 		args.protocol = XPRT_TRANSPORT_TCP;
986 		args.authflavor = clp->cl_cred.cr_flavor;
987 		clp->cl_cb_ident = conn->cb_ident;
988 	} else {
989 		if (!conn->cb_xprt)
990 			return -EINVAL;
991 		clp->cl_cb_session = ses;
992 		args.bc_xprt = conn->cb_xprt;
993 		args.prognumber = clp->cl_cb_session->se_cb_prog;
994 		args.protocol = conn->cb_xprt->xpt_class->xcl_ident |
995 				XPRT_TRANSPORT_BC;
996 		args.authflavor = ses->se_cb_sec.flavor;
997 	}
998 	/* Create RPC client */
999 	client = rpc_create(&args);
1000 	if (IS_ERR(client)) {
1001 		trace_nfsd_cb_setup_err(clp, PTR_ERR(client));
1002 		return PTR_ERR(client);
1003 	}
1004 	cred = get_backchannel_cred(clp, client, ses);
1005 	if (!cred) {
1006 		trace_nfsd_cb_setup_err(clp, -ENOMEM);
1007 		rpc_shutdown_client(client);
1008 		return -ENOMEM;
1009 	}
1010 
1011 	if (clp->cl_minorversion != 0)
1012 		clp->cl_cb_conn.cb_xprt = conn->cb_xprt;
1013 	clp->cl_cb_client = client;
1014 	clp->cl_cb_cred = cred;
1015 	rcu_read_lock();
1016 	trace_nfsd_cb_setup(clp, rpc_peeraddr2str(client, RPC_DISPLAY_NETID),
1017 			    args.authflavor);
1018 	rcu_read_unlock();
1019 	return 0;
1020 }
1021 
1022 static void nfsd4_mark_cb_state(struct nfs4_client *clp, int newstate)
1023 {
1024 	if (clp->cl_cb_state != newstate) {
1025 		clp->cl_cb_state = newstate;
1026 		trace_nfsd_cb_state(clp);
1027 	}
1028 }
1029 
1030 static void nfsd4_mark_cb_down(struct nfs4_client *clp, int reason)
1031 {
1032 	if (test_bit(NFSD4_CLIENT_CB_UPDATE, &clp->cl_flags))
1033 		return;
1034 	nfsd4_mark_cb_state(clp, NFSD4_CB_DOWN);
1035 }
1036 
1037 static void nfsd4_mark_cb_fault(struct nfs4_client *clp, int reason)
1038 {
1039 	if (test_bit(NFSD4_CLIENT_CB_UPDATE, &clp->cl_flags))
1040 		return;
1041 	nfsd4_mark_cb_state(clp, NFSD4_CB_FAULT);
1042 }
1043 
1044 static void nfsd4_cb_probe_done(struct rpc_task *task, void *calldata)
1045 {
1046 	struct nfs4_client *clp = container_of(calldata, struct nfs4_client, cl_cb_null);
1047 
1048 	if (task->tk_status)
1049 		nfsd4_mark_cb_down(clp, task->tk_status);
1050 	else
1051 		nfsd4_mark_cb_state(clp, NFSD4_CB_UP);
1052 }
1053 
1054 static void nfsd4_cb_probe_release(void *calldata)
1055 {
1056 	struct nfs4_client *clp = container_of(calldata, struct nfs4_client, cl_cb_null);
1057 
1058 	nfsd41_cb_inflight_end(clp);
1059 
1060 }
1061 
1062 static const struct rpc_call_ops nfsd4_cb_probe_ops = {
1063 	/* XXX: release method to ensure we set the cb channel down if
1064 	 * necessary on early failure? */
1065 	.rpc_call_done = nfsd4_cb_probe_done,
1066 	.rpc_release = nfsd4_cb_probe_release,
1067 };
1068 
1069 /*
1070  * Poke the callback thread to process any updates to the callback
1071  * parameters, and send a null probe.
1072  */
1073 void nfsd4_probe_callback(struct nfs4_client *clp)
1074 {
1075 	trace_nfsd_cb_probe(clp);
1076 	nfsd4_mark_cb_state(clp, NFSD4_CB_UNKNOWN);
1077 	set_bit(NFSD4_CLIENT_CB_UPDATE, &clp->cl_flags);
1078 	nfsd4_run_cb(&clp->cl_cb_null);
1079 }
1080 
1081 void nfsd4_probe_callback_sync(struct nfs4_client *clp)
1082 {
1083 	nfsd4_probe_callback(clp);
1084 	flush_workqueue(callback_wq);
1085 }
1086 
1087 void nfsd4_change_callback(struct nfs4_client *clp, struct nfs4_cb_conn *conn)
1088 {
1089 	nfsd4_mark_cb_state(clp, NFSD4_CB_UNKNOWN);
1090 	spin_lock(&clp->cl_lock);
1091 	memcpy(&clp->cl_cb_conn, conn, sizeof(struct nfs4_cb_conn));
1092 	spin_unlock(&clp->cl_lock);
1093 }
1094 
1095 /*
1096  * There's currently a single callback channel slot.
1097  * If the slot is available, then mark it busy.  Otherwise, set the
1098  * thread for sleeping on the callback RPC wait queue.
1099  */
1100 static bool nfsd41_cb_get_slot(struct nfsd4_callback *cb, struct rpc_task *task)
1101 {
1102 	struct nfs4_client *clp = cb->cb_clp;
1103 
1104 	if (!cb->cb_holds_slot &&
1105 	    test_and_set_bit(0, &clp->cl_cb_slot_busy) != 0) {
1106 		rpc_sleep_on(&clp->cl_cb_waitq, task, NULL);
1107 		/* Race breaker */
1108 		if (test_and_set_bit(0, &clp->cl_cb_slot_busy) != 0) {
1109 			dprintk("%s slot is busy\n", __func__);
1110 			return false;
1111 		}
1112 		rpc_wake_up_queued_task(&clp->cl_cb_waitq, task);
1113 	}
1114 	cb->cb_holds_slot = true;
1115 	return true;
1116 }
1117 
1118 static void nfsd41_cb_release_slot(struct nfsd4_callback *cb)
1119 {
1120 	struct nfs4_client *clp = cb->cb_clp;
1121 
1122 	if (cb->cb_holds_slot) {
1123 		cb->cb_holds_slot = false;
1124 		clear_bit(0, &clp->cl_cb_slot_busy);
1125 		rpc_wake_up_next(&clp->cl_cb_waitq);
1126 	}
1127 }
1128 
1129 static void nfsd41_destroy_cb(struct nfsd4_callback *cb)
1130 {
1131 	struct nfs4_client *clp = cb->cb_clp;
1132 
1133 	nfsd41_cb_release_slot(cb);
1134 	if (cb->cb_ops && cb->cb_ops->release)
1135 		cb->cb_ops->release(cb);
1136 	nfsd41_cb_inflight_end(clp);
1137 }
1138 
1139 /*
1140  * TODO: cb_sequence should support referring call lists, cachethis, multiple
1141  * slots, and mark callback channel down on communication errors.
1142  */
1143 static void nfsd4_cb_prepare(struct rpc_task *task, void *calldata)
1144 {
1145 	struct nfsd4_callback *cb = calldata;
1146 	struct nfs4_client *clp = cb->cb_clp;
1147 	u32 minorversion = clp->cl_minorversion;
1148 
1149 	/*
1150 	 * cb_seq_status is only set in decode_cb_sequence4res,
1151 	 * and so will remain 1 if an rpc level failure occurs.
1152 	 */
1153 	cb->cb_seq_status = 1;
1154 	cb->cb_status = 0;
1155 	if (minorversion && !nfsd41_cb_get_slot(cb, task))
1156 		return;
1157 	rpc_call_start(task);
1158 }
1159 
1160 static bool nfsd4_cb_sequence_done(struct rpc_task *task, struct nfsd4_callback *cb)
1161 {
1162 	struct nfs4_client *clp = cb->cb_clp;
1163 	struct nfsd4_session *session = clp->cl_cb_session;
1164 	bool ret = true;
1165 
1166 	if (!clp->cl_minorversion) {
1167 		/*
1168 		 * If the backchannel connection was shut down while this
1169 		 * task was queued, we need to resubmit it after setting up
1170 		 * a new backchannel connection.
1171 		 *
1172 		 * Note that if we lost our callback connection permanently
1173 		 * the submission code will error out, so we don't need to
1174 		 * handle that case here.
1175 		 */
1176 		if (RPC_SIGNALLED(task))
1177 			goto need_restart;
1178 
1179 		return true;
1180 	}
1181 
1182 	if (!cb->cb_holds_slot)
1183 		goto need_restart;
1184 
1185 	switch (cb->cb_seq_status) {
1186 	case 0:
1187 		/*
1188 		 * No need for lock, access serialized in nfsd4_cb_prepare
1189 		 *
1190 		 * RFC5661 20.9.3
1191 		 * If CB_SEQUENCE returns an error, then the state of the slot
1192 		 * (sequence ID, cached reply) MUST NOT change.
1193 		 */
1194 		++session->se_cb_seq_nr;
1195 		break;
1196 	case -ESERVERFAULT:
1197 		++session->se_cb_seq_nr;
1198 		fallthrough;
1199 	case 1:
1200 	case -NFS4ERR_BADSESSION:
1201 		nfsd4_mark_cb_fault(cb->cb_clp, cb->cb_seq_status);
1202 		ret = false;
1203 		break;
1204 	case -NFS4ERR_DELAY:
1205 		if (!rpc_restart_call(task))
1206 			goto out;
1207 
1208 		rpc_delay(task, 2 * HZ);
1209 		return false;
1210 	case -NFS4ERR_BADSLOT:
1211 		goto retry_nowait;
1212 	case -NFS4ERR_SEQ_MISORDERED:
1213 		if (session->se_cb_seq_nr != 1) {
1214 			session->se_cb_seq_nr = 1;
1215 			goto retry_nowait;
1216 		}
1217 		break;
1218 	default:
1219 		nfsd4_mark_cb_fault(cb->cb_clp, cb->cb_seq_status);
1220 		dprintk("%s: unprocessed error %d\n", __func__,
1221 			cb->cb_seq_status);
1222 	}
1223 
1224 	nfsd41_cb_release_slot(cb);
1225 	dprintk("%s: freed slot, new seqid=%d\n", __func__,
1226 		clp->cl_cb_session->se_cb_seq_nr);
1227 
1228 	if (RPC_SIGNALLED(task))
1229 		goto need_restart;
1230 out:
1231 	return ret;
1232 retry_nowait:
1233 	if (rpc_restart_call_prepare(task))
1234 		ret = false;
1235 	goto out;
1236 need_restart:
1237 	if (!test_bit(NFSD4_CLIENT_CB_KILL, &clp->cl_flags)) {
1238 		task->tk_status = 0;
1239 		cb->cb_need_restart = true;
1240 	}
1241 	return false;
1242 }
1243 
1244 static void nfsd4_cb_done(struct rpc_task *task, void *calldata)
1245 {
1246 	struct nfsd4_callback *cb = calldata;
1247 	struct nfs4_client *clp = cb->cb_clp;
1248 
1249 	if (!nfsd4_cb_sequence_done(task, cb))
1250 		return;
1251 
1252 	if (cb->cb_status) {
1253 		WARN_ON_ONCE(task->tk_status);
1254 		task->tk_status = cb->cb_status;
1255 	}
1256 
1257 	switch (cb->cb_ops->done(cb, task)) {
1258 	case 0:
1259 		task->tk_status = 0;
1260 		rpc_restart_call_prepare(task);
1261 		return;
1262 	case 1:
1263 		switch (task->tk_status) {
1264 		case -EIO:
1265 		case -ETIMEDOUT:
1266 		case -EACCES:
1267 			nfsd4_mark_cb_down(clp, task->tk_status);
1268 		}
1269 		break;
1270 	default:
1271 		BUG();
1272 	}
1273 }
1274 
1275 static void nfsd4_cb_release(void *calldata)
1276 {
1277 	struct nfsd4_callback *cb = calldata;
1278 
1279 	if (cb->cb_need_restart)
1280 		nfsd4_queue_cb(cb);
1281 	else
1282 		nfsd41_destroy_cb(cb);
1283 
1284 }
1285 
1286 static const struct rpc_call_ops nfsd4_cb_ops = {
1287 	.rpc_call_prepare = nfsd4_cb_prepare,
1288 	.rpc_call_done = nfsd4_cb_done,
1289 	.rpc_release = nfsd4_cb_release,
1290 };
1291 
1292 int nfsd4_create_callback_queue(void)
1293 {
1294 	callback_wq = alloc_ordered_workqueue("nfsd4_callbacks", 0);
1295 	if (!callback_wq)
1296 		return -ENOMEM;
1297 	return 0;
1298 }
1299 
1300 void nfsd4_destroy_callback_queue(void)
1301 {
1302 	destroy_workqueue(callback_wq);
1303 }
1304 
1305 /* must be called under the state lock */
1306 void nfsd4_shutdown_callback(struct nfs4_client *clp)
1307 {
1308 	if (clp->cl_cb_state != NFSD4_CB_UNKNOWN)
1309 		trace_nfsd_cb_shutdown(clp);
1310 
1311 	set_bit(NFSD4_CLIENT_CB_KILL, &clp->cl_flags);
1312 	/*
1313 	 * Note this won't actually result in a null callback;
1314 	 * instead, nfsd4_run_cb_null() will detect the killed
1315 	 * client, destroy the rpc client, and stop:
1316 	 */
1317 	nfsd4_run_cb(&clp->cl_cb_null);
1318 	flush_workqueue(callback_wq);
1319 	nfsd41_cb_inflight_wait_complete(clp);
1320 }
1321 
1322 /* requires cl_lock: */
1323 static struct nfsd4_conn * __nfsd4_find_backchannel(struct nfs4_client *clp)
1324 {
1325 	struct nfsd4_session *s;
1326 	struct nfsd4_conn *c;
1327 
1328 	list_for_each_entry(s, &clp->cl_sessions, se_perclnt) {
1329 		list_for_each_entry(c, &s->se_conns, cn_persession) {
1330 			if (c->cn_flags & NFS4_CDFC4_BACK)
1331 				return c;
1332 		}
1333 	}
1334 	return NULL;
1335 }
1336 
1337 /*
1338  * Note there isn't a lot of locking in this code; instead we depend on
1339  * the fact that it is run from the callback_wq, which won't run two
1340  * work items at once.  So, for example, callback_wq handles all access
1341  * of cl_cb_client and all calls to rpc_create or rpc_shutdown_client.
1342  */
1343 static void nfsd4_process_cb_update(struct nfsd4_callback *cb)
1344 {
1345 	struct nfs4_cb_conn conn;
1346 	struct nfs4_client *clp = cb->cb_clp;
1347 	struct nfsd4_session *ses = NULL;
1348 	struct nfsd4_conn *c;
1349 	int err;
1350 
1351 	/*
1352 	 * This is either an update, or the client dying; in either case,
1353 	 * kill the old client:
1354 	 */
1355 	if (clp->cl_cb_client) {
1356 		rpc_shutdown_client(clp->cl_cb_client);
1357 		clp->cl_cb_client = NULL;
1358 		put_cred(clp->cl_cb_cred);
1359 		clp->cl_cb_cred = NULL;
1360 	}
1361 	if (clp->cl_cb_conn.cb_xprt) {
1362 		svc_xprt_put(clp->cl_cb_conn.cb_xprt);
1363 		clp->cl_cb_conn.cb_xprt = NULL;
1364 	}
1365 	if (test_bit(NFSD4_CLIENT_CB_KILL, &clp->cl_flags))
1366 		return;
1367 	spin_lock(&clp->cl_lock);
1368 	/*
1369 	 * Only serialized callback code is allowed to clear these
1370 	 * flags; main nfsd code can only set them:
1371 	 */
1372 	BUG_ON(!(clp->cl_flags & NFSD4_CLIENT_CB_FLAG_MASK));
1373 	clear_bit(NFSD4_CLIENT_CB_UPDATE, &clp->cl_flags);
1374 	memcpy(&conn, &cb->cb_clp->cl_cb_conn, sizeof(struct nfs4_cb_conn));
1375 	c = __nfsd4_find_backchannel(clp);
1376 	if (c) {
1377 		svc_xprt_get(c->cn_xprt);
1378 		conn.cb_xprt = c->cn_xprt;
1379 		ses = c->cn_session;
1380 	}
1381 	spin_unlock(&clp->cl_lock);
1382 
1383 	err = setup_callback_client(clp, &conn, ses);
1384 	if (err) {
1385 		nfsd4_mark_cb_down(clp, err);
1386 		if (c)
1387 			svc_xprt_put(c->cn_xprt);
1388 		return;
1389 	}
1390 }
1391 
1392 static void
1393 nfsd4_run_cb_work(struct work_struct *work)
1394 {
1395 	struct nfsd4_callback *cb =
1396 		container_of(work, struct nfsd4_callback, cb_work);
1397 	struct nfs4_client *clp = cb->cb_clp;
1398 	struct rpc_clnt *clnt;
1399 	int flags;
1400 
1401 	if (cb->cb_need_restart) {
1402 		cb->cb_need_restart = false;
1403 	} else {
1404 		if (cb->cb_ops && cb->cb_ops->prepare)
1405 			cb->cb_ops->prepare(cb);
1406 	}
1407 
1408 	if (clp->cl_flags & NFSD4_CLIENT_CB_FLAG_MASK)
1409 		nfsd4_process_cb_update(cb);
1410 
1411 	clnt = clp->cl_cb_client;
1412 	if (!clnt) {
1413 		/* Callback channel broken, or client killed; give up: */
1414 		nfsd41_destroy_cb(cb);
1415 		return;
1416 	}
1417 
1418 	/*
1419 	 * Don't send probe messages for 4.1 or later.
1420 	 */
1421 	if (!cb->cb_ops && clp->cl_minorversion) {
1422 		nfsd4_mark_cb_state(clp, NFSD4_CB_UP);
1423 		nfsd41_destroy_cb(cb);
1424 		return;
1425 	}
1426 
1427 	cb->cb_msg.rpc_cred = clp->cl_cb_cred;
1428 	flags = clp->cl_minorversion ? RPC_TASK_NOCONNECT : RPC_TASK_SOFTCONN;
1429 	rpc_call_async(clnt, &cb->cb_msg, RPC_TASK_SOFT | flags,
1430 			cb->cb_ops ? &nfsd4_cb_ops : &nfsd4_cb_probe_ops, cb);
1431 }
1432 
1433 void nfsd4_init_cb(struct nfsd4_callback *cb, struct nfs4_client *clp,
1434 		const struct nfsd4_callback_ops *ops, enum nfsd4_cb_op op)
1435 {
1436 	cb->cb_clp = clp;
1437 	cb->cb_msg.rpc_proc = &nfs4_cb_procedures[op];
1438 	cb->cb_msg.rpc_argp = cb;
1439 	cb->cb_msg.rpc_resp = cb;
1440 	cb->cb_ops = ops;
1441 	INIT_WORK(&cb->cb_work, nfsd4_run_cb_work);
1442 	cb->cb_seq_status = 1;
1443 	cb->cb_status = 0;
1444 	cb->cb_need_restart = false;
1445 	cb->cb_holds_slot = false;
1446 }
1447 
1448 /**
1449  * nfsd4_run_cb - queue up a callback job to run
1450  * @cb: callback to queue
1451  *
1452  * Kick off a callback to do its thing. Returns false if it was already
1453  * on a queue, true otherwise.
1454  */
1455 bool nfsd4_run_cb(struct nfsd4_callback *cb)
1456 {
1457 	struct nfs4_client *clp = cb->cb_clp;
1458 	bool queued;
1459 
1460 	nfsd41_cb_inflight_begin(clp);
1461 	queued = nfsd4_queue_cb(cb);
1462 	if (!queued)
1463 		nfsd41_cb_inflight_end(clp);
1464 	return queued;
1465 }
1466