xref: /openbmc/linux/fs/nfsd/nfs4callback.c (revision 0da85d1e)
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 
43 #define NFSDDBG_FACILITY                NFSDDBG_PROC
44 
45 static void nfsd4_mark_cb_fault(struct nfs4_client *, int reason);
46 
47 #define NFSPROC4_CB_NULL 0
48 #define NFSPROC4_CB_COMPOUND 1
49 
50 /* Index of predefined Linux callback client operations */
51 
52 struct nfs4_cb_compound_hdr {
53 	/* args */
54 	u32		ident;	/* minorversion 0 only */
55 	u32		nops;
56 	__be32		*nops_p;
57 	u32		minorversion;
58 	/* res */
59 	int		status;
60 };
61 
62 /*
63  * Handle decode buffer overflows out-of-line.
64  */
65 static void print_overflow_msg(const char *func, const struct xdr_stream *xdr)
66 {
67 	dprintk("NFS: %s prematurely hit the end of our receive buffer. "
68 		"Remaining buffer length is %tu words.\n",
69 		func, xdr->end - xdr->p);
70 }
71 
72 static __be32 *xdr_encode_empty_array(__be32 *p)
73 {
74 	*p++ = xdr_zero;
75 	return p;
76 }
77 
78 /*
79  * Encode/decode NFSv4 CB basic data types
80  *
81  * Basic NFSv4 callback data types are defined in section 15 of RFC
82  * 3530: "Network File System (NFS) version 4 Protocol" and section
83  * 20 of RFC 5661: "Network File System (NFS) Version 4 Minor Version
84  * 1 Protocol"
85  */
86 
87 /*
88  *	nfs_cb_opnum4
89  *
90  *	enum nfs_cb_opnum4 {
91  *		OP_CB_GETATTR		= 3,
92  *		  ...
93  *	};
94  */
95 enum nfs_cb_opnum4 {
96 	OP_CB_GETATTR			= 3,
97 	OP_CB_RECALL			= 4,
98 	OP_CB_LAYOUTRECALL		= 5,
99 	OP_CB_NOTIFY			= 6,
100 	OP_CB_PUSH_DELEG		= 7,
101 	OP_CB_RECALL_ANY		= 8,
102 	OP_CB_RECALLABLE_OBJ_AVAIL	= 9,
103 	OP_CB_RECALL_SLOT		= 10,
104 	OP_CB_SEQUENCE			= 11,
105 	OP_CB_WANTS_CANCELLED		= 12,
106 	OP_CB_NOTIFY_LOCK		= 13,
107 	OP_CB_NOTIFY_DEVICEID		= 14,
108 	OP_CB_ILLEGAL			= 10044
109 };
110 
111 static void encode_nfs_cb_opnum4(struct xdr_stream *xdr, enum nfs_cb_opnum4 op)
112 {
113 	__be32 *p;
114 
115 	p = xdr_reserve_space(xdr, 4);
116 	*p = cpu_to_be32(op);
117 }
118 
119 /*
120  * nfs_fh4
121  *
122  *	typedef opaque nfs_fh4<NFS4_FHSIZE>;
123  */
124 static void encode_nfs_fh4(struct xdr_stream *xdr, const struct knfsd_fh *fh)
125 {
126 	u32 length = fh->fh_size;
127 	__be32 *p;
128 
129 	BUG_ON(length > NFS4_FHSIZE);
130 	p = xdr_reserve_space(xdr, 4 + length);
131 	xdr_encode_opaque(p, &fh->fh_base, length);
132 }
133 
134 /*
135  * stateid4
136  *
137  *	struct stateid4 {
138  *		uint32_t	seqid;
139  *		opaque		other[12];
140  *	};
141  */
142 static void encode_stateid4(struct xdr_stream *xdr, const stateid_t *sid)
143 {
144 	__be32 *p;
145 
146 	p = xdr_reserve_space(xdr, NFS4_STATEID_SIZE);
147 	*p++ = cpu_to_be32(sid->si_generation);
148 	xdr_encode_opaque_fixed(p, &sid->si_opaque, NFS4_STATEID_OTHER_SIZE);
149 }
150 
151 /*
152  * sessionid4
153  *
154  *	typedef opaque sessionid4[NFS4_SESSIONID_SIZE];
155  */
156 static void encode_sessionid4(struct xdr_stream *xdr,
157 			      const struct nfsd4_session *session)
158 {
159 	__be32 *p;
160 
161 	p = xdr_reserve_space(xdr, NFS4_MAX_SESSIONID_LEN);
162 	xdr_encode_opaque_fixed(p, session->se_sessionid.data,
163 					NFS4_MAX_SESSIONID_LEN);
164 }
165 
166 /*
167  * nfsstat4
168  */
169 static const struct {
170 	int stat;
171 	int errno;
172 } nfs_cb_errtbl[] = {
173 	{ NFS4_OK,		0		},
174 	{ NFS4ERR_PERM,		-EPERM		},
175 	{ NFS4ERR_NOENT,	-ENOENT		},
176 	{ NFS4ERR_IO,		-EIO		},
177 	{ NFS4ERR_NXIO,		-ENXIO		},
178 	{ NFS4ERR_ACCESS,	-EACCES		},
179 	{ NFS4ERR_EXIST,	-EEXIST		},
180 	{ NFS4ERR_XDEV,		-EXDEV		},
181 	{ NFS4ERR_NOTDIR,	-ENOTDIR	},
182 	{ NFS4ERR_ISDIR,	-EISDIR		},
183 	{ NFS4ERR_INVAL,	-EINVAL		},
184 	{ NFS4ERR_FBIG,		-EFBIG		},
185 	{ NFS4ERR_NOSPC,	-ENOSPC		},
186 	{ NFS4ERR_ROFS,		-EROFS		},
187 	{ NFS4ERR_MLINK,	-EMLINK		},
188 	{ NFS4ERR_NAMETOOLONG,	-ENAMETOOLONG	},
189 	{ NFS4ERR_NOTEMPTY,	-ENOTEMPTY	},
190 	{ NFS4ERR_DQUOT,	-EDQUOT		},
191 	{ NFS4ERR_STALE,	-ESTALE		},
192 	{ NFS4ERR_BADHANDLE,	-EBADHANDLE	},
193 	{ NFS4ERR_BAD_COOKIE,	-EBADCOOKIE	},
194 	{ NFS4ERR_NOTSUPP,	-ENOTSUPP	},
195 	{ NFS4ERR_TOOSMALL,	-ETOOSMALL	},
196 	{ NFS4ERR_SERVERFAULT,	-ESERVERFAULT	},
197 	{ NFS4ERR_BADTYPE,	-EBADTYPE	},
198 	{ NFS4ERR_LOCKED,	-EAGAIN		},
199 	{ NFS4ERR_RESOURCE,	-EREMOTEIO	},
200 	{ NFS4ERR_SYMLINK,	-ELOOP		},
201 	{ NFS4ERR_OP_ILLEGAL,	-EOPNOTSUPP	},
202 	{ NFS4ERR_DEADLOCK,	-EDEADLK	},
203 	{ -1,			-EIO		}
204 };
205 
206 /*
207  * If we cannot translate the error, the recovery routines should
208  * handle it.
209  *
210  * Note: remaining NFSv4 error codes have values > 10000, so should
211  * not conflict with native Linux error codes.
212  */
213 static int nfs_cb_stat_to_errno(int status)
214 {
215 	int i;
216 
217 	for (i = 0; nfs_cb_errtbl[i].stat != -1; i++) {
218 		if (nfs_cb_errtbl[i].stat == status)
219 			return nfs_cb_errtbl[i].errno;
220 	}
221 
222 	dprintk("NFSD: Unrecognized NFS CB status value: %u\n", status);
223 	return -status;
224 }
225 
226 static int decode_cb_op_status(struct xdr_stream *xdr, enum nfs_opnum4 expected,
227 			       enum nfsstat4 *status)
228 {
229 	__be32 *p;
230 	u32 op;
231 
232 	p = xdr_inline_decode(xdr, 4 + 4);
233 	if (unlikely(p == NULL))
234 		goto out_overflow;
235 	op = be32_to_cpup(p++);
236 	if (unlikely(op != expected))
237 		goto out_unexpected;
238 	*status = be32_to_cpup(p);
239 	return 0;
240 out_overflow:
241 	print_overflow_msg(__func__, xdr);
242 	return -EIO;
243 out_unexpected:
244 	dprintk("NFSD: Callback server returned operation %d but "
245 		"we issued a request for %d\n", op, expected);
246 	return -EIO;
247 }
248 
249 /*
250  * CB_COMPOUND4args
251  *
252  *	struct CB_COMPOUND4args {
253  *		utf8str_cs	tag;
254  *		uint32_t	minorversion;
255  *		uint32_t	callback_ident;
256  *		nfs_cb_argop4	argarray<>;
257  *	};
258 */
259 static void encode_cb_compound4args(struct xdr_stream *xdr,
260 				    struct nfs4_cb_compound_hdr *hdr)
261 {
262 	__be32 * p;
263 
264 	p = xdr_reserve_space(xdr, 4 + 4 + 4 + 4);
265 	p = xdr_encode_empty_array(p);		/* empty tag */
266 	*p++ = cpu_to_be32(hdr->minorversion);
267 	*p++ = cpu_to_be32(hdr->ident);
268 
269 	hdr->nops_p = p;
270 	*p = cpu_to_be32(hdr->nops);		/* argarray element count */
271 }
272 
273 /*
274  * Update argarray element count
275  */
276 static void encode_cb_nops(struct nfs4_cb_compound_hdr *hdr)
277 {
278 	BUG_ON(hdr->nops > NFS4_MAX_BACK_CHANNEL_OPS);
279 	*hdr->nops_p = cpu_to_be32(hdr->nops);
280 }
281 
282 /*
283  * CB_COMPOUND4res
284  *
285  *	struct CB_COMPOUND4res {
286  *		nfsstat4	status;
287  *		utf8str_cs	tag;
288  *		nfs_cb_resop4	resarray<>;
289  *	};
290  */
291 static int decode_cb_compound4res(struct xdr_stream *xdr,
292 				  struct nfs4_cb_compound_hdr *hdr)
293 {
294 	u32 length;
295 	__be32 *p;
296 
297 	p = xdr_inline_decode(xdr, 4 + 4);
298 	if (unlikely(p == NULL))
299 		goto out_overflow;
300 	hdr->status = be32_to_cpup(p++);
301 	/* Ignore the tag */
302 	length = be32_to_cpup(p++);
303 	p = xdr_inline_decode(xdr, length + 4);
304 	if (unlikely(p == NULL))
305 		goto out_overflow;
306 	hdr->nops = be32_to_cpup(p);
307 	return 0;
308 out_overflow:
309 	print_overflow_msg(__func__, xdr);
310 	return -EIO;
311 }
312 
313 /*
314  * CB_RECALL4args
315  *
316  *	struct CB_RECALL4args {
317  *		stateid4	stateid;
318  *		bool		truncate;
319  *		nfs_fh4		fh;
320  *	};
321  */
322 static void encode_cb_recall4args(struct xdr_stream *xdr,
323 				  const struct nfs4_delegation *dp,
324 				  struct nfs4_cb_compound_hdr *hdr)
325 {
326 	__be32 *p;
327 
328 	encode_nfs_cb_opnum4(xdr, OP_CB_RECALL);
329 	encode_stateid4(xdr, &dp->dl_stid.sc_stateid);
330 
331 	p = xdr_reserve_space(xdr, 4);
332 	*p++ = xdr_zero;			/* truncate */
333 
334 	encode_nfs_fh4(xdr, &dp->dl_stid.sc_file->fi_fhandle);
335 
336 	hdr->nops++;
337 }
338 
339 /*
340  * CB_SEQUENCE4args
341  *
342  *	struct CB_SEQUENCE4args {
343  *		sessionid4		csa_sessionid;
344  *		sequenceid4		csa_sequenceid;
345  *		slotid4			csa_slotid;
346  *		slotid4			csa_highest_slotid;
347  *		bool			csa_cachethis;
348  *		referring_call_list4	csa_referring_call_lists<>;
349  *	};
350  */
351 static void encode_cb_sequence4args(struct xdr_stream *xdr,
352 				    const struct nfsd4_callback *cb,
353 				    struct nfs4_cb_compound_hdr *hdr)
354 {
355 	struct nfsd4_session *session = cb->cb_clp->cl_cb_session;
356 	__be32 *p;
357 
358 	if (hdr->minorversion == 0)
359 		return;
360 
361 	encode_nfs_cb_opnum4(xdr, OP_CB_SEQUENCE);
362 	encode_sessionid4(xdr, session);
363 
364 	p = xdr_reserve_space(xdr, 4 + 4 + 4 + 4 + 4);
365 	*p++ = cpu_to_be32(session->se_cb_seq_nr);	/* csa_sequenceid */
366 	*p++ = xdr_zero;			/* csa_slotid */
367 	*p++ = xdr_zero;			/* csa_highest_slotid */
368 	*p++ = xdr_zero;			/* csa_cachethis */
369 	xdr_encode_empty_array(p);		/* csa_referring_call_lists */
370 
371 	hdr->nops++;
372 }
373 
374 /*
375  * CB_SEQUENCE4resok
376  *
377  *	struct CB_SEQUENCE4resok {
378  *		sessionid4	csr_sessionid;
379  *		sequenceid4	csr_sequenceid;
380  *		slotid4		csr_slotid;
381  *		slotid4		csr_highest_slotid;
382  *		slotid4		csr_target_highest_slotid;
383  *	};
384  *
385  *	union CB_SEQUENCE4res switch (nfsstat4 csr_status) {
386  *	case NFS4_OK:
387  *		CB_SEQUENCE4resok	csr_resok4;
388  *	default:
389  *		void;
390  *	};
391  *
392  * Our current back channel implmentation supports a single backchannel
393  * with a single slot.
394  */
395 static int decode_cb_sequence4resok(struct xdr_stream *xdr,
396 				    struct nfsd4_callback *cb)
397 {
398 	struct nfsd4_session *session = cb->cb_clp->cl_cb_session;
399 	struct nfs4_sessionid id;
400 	int status;
401 	__be32 *p;
402 	u32 dummy;
403 
404 	status = -ESERVERFAULT;
405 
406 	/*
407 	 * If the server returns different values for sessionID, slotID or
408 	 * sequence number, the server is looney tunes.
409 	 */
410 	p = xdr_inline_decode(xdr, NFS4_MAX_SESSIONID_LEN + 4 + 4 + 4 + 4);
411 	if (unlikely(p == NULL))
412 		goto out_overflow;
413 	memcpy(id.data, p, NFS4_MAX_SESSIONID_LEN);
414 	if (memcmp(id.data, session->se_sessionid.data,
415 					NFS4_MAX_SESSIONID_LEN) != 0) {
416 		dprintk("NFS: %s Invalid session id\n", __func__);
417 		goto out;
418 	}
419 	p += XDR_QUADLEN(NFS4_MAX_SESSIONID_LEN);
420 
421 	dummy = be32_to_cpup(p++);
422 	if (dummy != session->se_cb_seq_nr) {
423 		dprintk("NFS: %s Invalid sequence number\n", __func__);
424 		goto out;
425 	}
426 
427 	dummy = be32_to_cpup(p++);
428 	if (dummy != 0) {
429 		dprintk("NFS: %s Invalid slotid\n", __func__);
430 		goto out;
431 	}
432 
433 	/*
434 	 * FIXME: process highest slotid and target highest slotid
435 	 */
436 	status = 0;
437 out:
438 	if (status)
439 		nfsd4_mark_cb_fault(cb->cb_clp, status);
440 	return status;
441 out_overflow:
442 	print_overflow_msg(__func__, xdr);
443 	return -EIO;
444 }
445 
446 static int decode_cb_sequence4res(struct xdr_stream *xdr,
447 				  struct nfsd4_callback *cb)
448 {
449 	enum nfsstat4 nfserr;
450 	int status;
451 
452 	if (cb->cb_minorversion == 0)
453 		return 0;
454 
455 	status = decode_cb_op_status(xdr, OP_CB_SEQUENCE, &nfserr);
456 	if (unlikely(status))
457 		goto out;
458 	if (unlikely(nfserr != NFS4_OK))
459 		goto out_default;
460 	status = decode_cb_sequence4resok(xdr, cb);
461 out:
462 	return status;
463 out_default:
464 	return nfs_cb_stat_to_errno(nfserr);
465 }
466 
467 /*
468  * NFSv4.0 and NFSv4.1 XDR encode functions
469  *
470  * NFSv4.0 callback argument types are defined in section 15 of RFC
471  * 3530: "Network File System (NFS) version 4 Protocol" and section 20
472  * of RFC 5661:  "Network File System (NFS) Version 4 Minor Version 1
473  * Protocol".
474  */
475 
476 /*
477  * NB: Without this zero space reservation, callbacks over krb5p fail
478  */
479 static void nfs4_xdr_enc_cb_null(struct rpc_rqst *req, struct xdr_stream *xdr,
480 				 void *__unused)
481 {
482 	xdr_reserve_space(xdr, 0);
483 }
484 
485 /*
486  * 20.2. Operation 4: CB_RECALL - Recall a Delegation
487  */
488 static void nfs4_xdr_enc_cb_recall(struct rpc_rqst *req, struct xdr_stream *xdr,
489 				   const struct nfsd4_callback *cb)
490 {
491 	const struct nfs4_delegation *dp = cb_to_delegation(cb);
492 	struct nfs4_cb_compound_hdr hdr = {
493 		.ident = cb->cb_clp->cl_cb_ident,
494 		.minorversion = cb->cb_minorversion,
495 	};
496 
497 	encode_cb_compound4args(xdr, &hdr);
498 	encode_cb_sequence4args(xdr, cb, &hdr);
499 	encode_cb_recall4args(xdr, dp, &hdr);
500 	encode_cb_nops(&hdr);
501 }
502 
503 
504 /*
505  * NFSv4.0 and NFSv4.1 XDR decode functions
506  *
507  * NFSv4.0 callback result types are defined in section 15 of RFC
508  * 3530: "Network File System (NFS) version 4 Protocol" and section 20
509  * of RFC 5661:  "Network File System (NFS) Version 4 Minor Version 1
510  * Protocol".
511  */
512 
513 static int nfs4_xdr_dec_cb_null(struct rpc_rqst *req, struct xdr_stream *xdr,
514 				void *__unused)
515 {
516 	return 0;
517 }
518 
519 /*
520  * 20.2. Operation 4: CB_RECALL - Recall a Delegation
521  */
522 static int nfs4_xdr_dec_cb_recall(struct rpc_rqst *rqstp,
523 				  struct xdr_stream *xdr,
524 				  struct nfsd4_callback *cb)
525 {
526 	struct nfs4_cb_compound_hdr hdr;
527 	enum nfsstat4 nfserr;
528 	int status;
529 
530 	status = decode_cb_compound4res(xdr, &hdr);
531 	if (unlikely(status))
532 		goto out;
533 
534 	if (cb != NULL) {
535 		status = decode_cb_sequence4res(xdr, cb);
536 		if (unlikely(status))
537 			goto out;
538 	}
539 
540 	status = decode_cb_op_status(xdr, OP_CB_RECALL, &nfserr);
541 	if (unlikely(status))
542 		goto out;
543 	if (unlikely(nfserr != NFS4_OK))
544 		status = nfs_cb_stat_to_errno(nfserr);
545 out:
546 	return status;
547 }
548 
549 #ifdef CONFIG_NFSD_PNFS
550 /*
551  * CB_LAYOUTRECALL4args
552  *
553  *	struct layoutrecall_file4 {
554  *		nfs_fh4         lor_fh;
555  *		offset4         lor_offset;
556  *		length4         lor_length;
557  *		stateid4        lor_stateid;
558  *	};
559  *
560  *	union layoutrecall4 switch(layoutrecall_type4 lor_recalltype) {
561  *	case LAYOUTRECALL4_FILE:
562  *		layoutrecall_file4 lor_layout;
563  *	case LAYOUTRECALL4_FSID:
564  *		fsid4              lor_fsid;
565  *	case LAYOUTRECALL4_ALL:
566  *		void;
567  *	};
568  *
569  *	struct CB_LAYOUTRECALL4args {
570  *		layouttype4             clora_type;
571  *		layoutiomode4           clora_iomode;
572  *		bool                    clora_changed;
573  *		layoutrecall4           clora_recall;
574  *	};
575  */
576 static void encode_cb_layout4args(struct xdr_stream *xdr,
577 				  const struct nfs4_layout_stateid *ls,
578 				  struct nfs4_cb_compound_hdr *hdr)
579 {
580 	__be32 *p;
581 
582 	BUG_ON(hdr->minorversion == 0);
583 
584 	p = xdr_reserve_space(xdr, 5 * 4);
585 	*p++ = cpu_to_be32(OP_CB_LAYOUTRECALL);
586 	*p++ = cpu_to_be32(ls->ls_layout_type);
587 	*p++ = cpu_to_be32(IOMODE_ANY);
588 	*p++ = cpu_to_be32(1);
589 	*p = cpu_to_be32(RETURN_FILE);
590 
591 	encode_nfs_fh4(xdr, &ls->ls_stid.sc_file->fi_fhandle);
592 
593 	p = xdr_reserve_space(xdr, 2 * 8);
594 	p = xdr_encode_hyper(p, 0);
595 	xdr_encode_hyper(p, NFS4_MAX_UINT64);
596 
597 	encode_stateid4(xdr, &ls->ls_recall_sid);
598 
599 	hdr->nops++;
600 }
601 
602 static void nfs4_xdr_enc_cb_layout(struct rpc_rqst *req,
603 				   struct xdr_stream *xdr,
604 				   const struct nfsd4_callback *cb)
605 {
606 	const struct nfs4_layout_stateid *ls =
607 		container_of(cb, struct nfs4_layout_stateid, ls_recall);
608 	struct nfs4_cb_compound_hdr hdr = {
609 		.ident = 0,
610 		.minorversion = cb->cb_minorversion,
611 	};
612 
613 	encode_cb_compound4args(xdr, &hdr);
614 	encode_cb_sequence4args(xdr, cb, &hdr);
615 	encode_cb_layout4args(xdr, ls, &hdr);
616 	encode_cb_nops(&hdr);
617 }
618 
619 static int nfs4_xdr_dec_cb_layout(struct rpc_rqst *rqstp,
620 				  struct xdr_stream *xdr,
621 				  struct nfsd4_callback *cb)
622 {
623 	struct nfs4_cb_compound_hdr hdr;
624 	enum nfsstat4 nfserr;
625 	int status;
626 
627 	status = decode_cb_compound4res(xdr, &hdr);
628 	if (unlikely(status))
629 		goto out;
630 	if (cb) {
631 		status = decode_cb_sequence4res(xdr, cb);
632 		if (unlikely(status))
633 			goto out;
634 	}
635 	status = decode_cb_op_status(xdr, OP_CB_LAYOUTRECALL, &nfserr);
636 	if (unlikely(status))
637 		goto out;
638 	if (unlikely(nfserr != NFS4_OK))
639 		status = nfs_cb_stat_to_errno(nfserr);
640 out:
641 	return status;
642 }
643 #endif /* CONFIG_NFSD_PNFS */
644 
645 /*
646  * RPC procedure tables
647  */
648 #define PROC(proc, call, argtype, restype)				\
649 [NFSPROC4_CLNT_##proc] = {						\
650 	.p_proc    = NFSPROC4_CB_##call,				\
651 	.p_encode  = (kxdreproc_t)nfs4_xdr_enc_##argtype,		\
652 	.p_decode  = (kxdrdproc_t)nfs4_xdr_dec_##restype,		\
653 	.p_arglen  = NFS4_enc_##argtype##_sz,				\
654 	.p_replen  = NFS4_dec_##restype##_sz,				\
655 	.p_statidx = NFSPROC4_CB_##call,				\
656 	.p_name    = #proc,						\
657 }
658 
659 static struct rpc_procinfo nfs4_cb_procedures[] = {
660 	PROC(CB_NULL,	NULL,		cb_null,	cb_null),
661 	PROC(CB_RECALL,	COMPOUND,	cb_recall,	cb_recall),
662 #ifdef CONFIG_NFSD_PNFS
663 	PROC(CB_LAYOUT,	COMPOUND,	cb_layout,	cb_layout),
664 #endif
665 };
666 
667 static struct rpc_version nfs_cb_version4 = {
668 /*
669  * Note on the callback rpc program version number: despite language in rfc
670  * 5661 section 18.36.3 requiring servers to use 4 in this field, the
671  * official xdr descriptions for both 4.0 and 4.1 specify version 1, and
672  * in practice that appears to be what implementations use.  The section
673  * 18.36.3 language is expected to be fixed in an erratum.
674  */
675 	.number			= 1,
676 	.nrprocs		= ARRAY_SIZE(nfs4_cb_procedures),
677 	.procs			= nfs4_cb_procedures
678 };
679 
680 static const struct rpc_version *nfs_cb_version[] = {
681 	&nfs_cb_version4,
682 };
683 
684 static const struct rpc_program cb_program;
685 
686 static struct rpc_stat cb_stats = {
687 	.program		= &cb_program
688 };
689 
690 #define NFS4_CALLBACK 0x40000000
691 static const struct rpc_program cb_program = {
692 	.name			= "nfs4_cb",
693 	.number			= NFS4_CALLBACK,
694 	.nrvers			= ARRAY_SIZE(nfs_cb_version),
695 	.version		= nfs_cb_version,
696 	.stats			= &cb_stats,
697 	.pipe_dir_name		= "nfsd4_cb",
698 };
699 
700 static int max_cb_time(struct net *net)
701 {
702 	struct nfsd_net *nn = net_generic(net, nfsd_net_id);
703 	return max(nn->nfsd4_lease/10, (time_t)1) * HZ;
704 }
705 
706 static struct rpc_cred *callback_cred;
707 
708 int set_callback_cred(void)
709 {
710 	if (callback_cred)
711 		return 0;
712 	callback_cred = rpc_lookup_machine_cred("nfs");
713 	if (!callback_cred)
714 		return -ENOMEM;
715 	return 0;
716 }
717 
718 static struct rpc_cred *get_backchannel_cred(struct nfs4_client *clp, struct rpc_clnt *client, struct nfsd4_session *ses)
719 {
720 	if (clp->cl_minorversion == 0) {
721 		return get_rpccred(callback_cred);
722 	} else {
723 		struct rpc_auth *auth = client->cl_auth;
724 		struct auth_cred acred = {};
725 
726 		acred.uid = ses->se_cb_sec.uid;
727 		acred.gid = ses->se_cb_sec.gid;
728 		return auth->au_ops->lookup_cred(client->cl_auth, &acred, 0);
729 	}
730 }
731 
732 static struct rpc_clnt *create_backchannel_client(struct rpc_create_args *args)
733 {
734 	struct rpc_xprt *xprt;
735 
736 	if (args->protocol != XPRT_TRANSPORT_BC_TCP)
737 		return rpc_create(args);
738 
739 	xprt = args->bc_xprt->xpt_bc_xprt;
740 	if (xprt) {
741 		xprt_get(xprt);
742 		return rpc_create_xprt(args, xprt);
743 	}
744 
745 	return rpc_create(args);
746 }
747 
748 static int setup_callback_client(struct nfs4_client *clp, struct nfs4_cb_conn *conn, struct nfsd4_session *ses)
749 {
750 	int maxtime = max_cb_time(clp->net);
751 	struct rpc_timeout	timeparms = {
752 		.to_initval	= maxtime,
753 		.to_retries	= 0,
754 		.to_maxval	= maxtime,
755 	};
756 	struct rpc_create_args args = {
757 		.net		= clp->net,
758 		.address	= (struct sockaddr *) &conn->cb_addr,
759 		.addrsize	= conn->cb_addrlen,
760 		.saddress	= (struct sockaddr *) &conn->cb_saddr,
761 		.timeout	= &timeparms,
762 		.program	= &cb_program,
763 		.version	= 0,
764 		.flags		= (RPC_CLNT_CREATE_NOPING | RPC_CLNT_CREATE_QUIET),
765 	};
766 	struct rpc_clnt *client;
767 	struct rpc_cred *cred;
768 
769 	if (clp->cl_minorversion == 0) {
770 		if (!clp->cl_cred.cr_principal &&
771 				(clp->cl_cred.cr_flavor >= RPC_AUTH_GSS_KRB5))
772 			return -EINVAL;
773 		args.client_name = clp->cl_cred.cr_principal;
774 		args.prognumber	= conn->cb_prog;
775 		args.protocol = XPRT_TRANSPORT_TCP;
776 		args.authflavor = clp->cl_cred.cr_flavor;
777 		clp->cl_cb_ident = conn->cb_ident;
778 	} else {
779 		if (!conn->cb_xprt)
780 			return -EINVAL;
781 		clp->cl_cb_conn.cb_xprt = conn->cb_xprt;
782 		clp->cl_cb_session = ses;
783 		args.bc_xprt = conn->cb_xprt;
784 		args.prognumber = clp->cl_cb_session->se_cb_prog;
785 		args.protocol = conn->cb_xprt->xpt_class->xcl_ident |
786 				XPRT_TRANSPORT_BC;
787 		args.authflavor = ses->se_cb_sec.flavor;
788 	}
789 	/* Create RPC client */
790 	client = create_backchannel_client(&args);
791 	if (IS_ERR(client)) {
792 		dprintk("NFSD: couldn't create callback client: %ld\n",
793 			PTR_ERR(client));
794 		return PTR_ERR(client);
795 	}
796 	cred = get_backchannel_cred(clp, client, ses);
797 	if (IS_ERR(cred)) {
798 		rpc_shutdown_client(client);
799 		return PTR_ERR(cred);
800 	}
801 	clp->cl_cb_client = client;
802 	clp->cl_cb_cred = cred;
803 	return 0;
804 }
805 
806 static void warn_no_callback_path(struct nfs4_client *clp, int reason)
807 {
808 	dprintk("NFSD: warning: no callback path to client %.*s: error %d\n",
809 		(int)clp->cl_name.len, clp->cl_name.data, reason);
810 }
811 
812 static void nfsd4_mark_cb_down(struct nfs4_client *clp, int reason)
813 {
814 	clp->cl_cb_state = NFSD4_CB_DOWN;
815 	warn_no_callback_path(clp, reason);
816 }
817 
818 static void nfsd4_mark_cb_fault(struct nfs4_client *clp, int reason)
819 {
820 	clp->cl_cb_state = NFSD4_CB_FAULT;
821 	warn_no_callback_path(clp, reason);
822 }
823 
824 static void nfsd4_cb_probe_done(struct rpc_task *task, void *calldata)
825 {
826 	struct nfs4_client *clp = container_of(calldata, struct nfs4_client, cl_cb_null);
827 
828 	if (task->tk_status)
829 		nfsd4_mark_cb_down(clp, task->tk_status);
830 	else
831 		clp->cl_cb_state = NFSD4_CB_UP;
832 }
833 
834 static const struct rpc_call_ops nfsd4_cb_probe_ops = {
835 	/* XXX: release method to ensure we set the cb channel down if
836 	 * necessary on early failure? */
837 	.rpc_call_done = nfsd4_cb_probe_done,
838 };
839 
840 static struct workqueue_struct *callback_wq;
841 
842 /*
843  * Poke the callback thread to process any updates to the callback
844  * parameters, and send a null probe.
845  */
846 void nfsd4_probe_callback(struct nfs4_client *clp)
847 {
848 	clp->cl_cb_state = NFSD4_CB_UNKNOWN;
849 	set_bit(NFSD4_CLIENT_CB_UPDATE, &clp->cl_flags);
850 	nfsd4_run_cb(&clp->cl_cb_null);
851 }
852 
853 void nfsd4_probe_callback_sync(struct nfs4_client *clp)
854 {
855 	nfsd4_probe_callback(clp);
856 	flush_workqueue(callback_wq);
857 }
858 
859 void nfsd4_change_callback(struct nfs4_client *clp, struct nfs4_cb_conn *conn)
860 {
861 	clp->cl_cb_state = NFSD4_CB_UNKNOWN;
862 	spin_lock(&clp->cl_lock);
863 	memcpy(&clp->cl_cb_conn, conn, sizeof(struct nfs4_cb_conn));
864 	spin_unlock(&clp->cl_lock);
865 }
866 
867 /*
868  * There's currently a single callback channel slot.
869  * If the slot is available, then mark it busy.  Otherwise, set the
870  * thread for sleeping on the callback RPC wait queue.
871  */
872 static bool nfsd41_cb_get_slot(struct nfs4_client *clp, struct rpc_task *task)
873 {
874 	if (test_and_set_bit(0, &clp->cl_cb_slot_busy) != 0) {
875 		rpc_sleep_on(&clp->cl_cb_waitq, task, NULL);
876 		/* Race breaker */
877 		if (test_and_set_bit(0, &clp->cl_cb_slot_busy) != 0) {
878 			dprintk("%s slot is busy\n", __func__);
879 			return false;
880 		}
881 		rpc_wake_up_queued_task(&clp->cl_cb_waitq, task);
882 	}
883 	return true;
884 }
885 
886 /*
887  * TODO: cb_sequence should support referring call lists, cachethis, multiple
888  * slots, and mark callback channel down on communication errors.
889  */
890 static void nfsd4_cb_prepare(struct rpc_task *task, void *calldata)
891 {
892 	struct nfsd4_callback *cb = calldata;
893 	struct nfs4_client *clp = cb->cb_clp;
894 	u32 minorversion = clp->cl_minorversion;
895 
896 	cb->cb_minorversion = minorversion;
897 	if (minorversion) {
898 		if (!nfsd41_cb_get_slot(clp, task))
899 			return;
900 	}
901 	spin_lock(&clp->cl_lock);
902 	if (list_empty(&cb->cb_per_client)) {
903 		/* This is the first call, not a restart */
904 		cb->cb_done = false;
905 		list_add(&cb->cb_per_client, &clp->cl_callbacks);
906 	}
907 	spin_unlock(&clp->cl_lock);
908 	rpc_call_start(task);
909 }
910 
911 static void nfsd4_cb_done(struct rpc_task *task, void *calldata)
912 {
913 	struct nfsd4_callback *cb = calldata;
914 	struct nfs4_client *clp = cb->cb_clp;
915 
916 	dprintk("%s: minorversion=%d\n", __func__,
917 		clp->cl_minorversion);
918 
919 	if (clp->cl_minorversion) {
920 		/* No need for lock, access serialized in nfsd4_cb_prepare */
921 		++clp->cl_cb_session->se_cb_seq_nr;
922 		clear_bit(0, &clp->cl_cb_slot_busy);
923 		rpc_wake_up_next(&clp->cl_cb_waitq);
924 		dprintk("%s: freed slot, new seqid=%d\n", __func__,
925 			clp->cl_cb_session->se_cb_seq_nr);
926 	}
927 
928 	if (clp->cl_cb_client != task->tk_client) {
929 		/* We're shutting down or changing cl_cb_client; leave
930 		 * it to nfsd4_process_cb_update to restart the call if
931 		 * necessary. */
932 		return;
933 	}
934 
935 	if (cb->cb_done)
936 		return;
937 
938 	switch (cb->cb_ops->done(cb, task)) {
939 	case 0:
940 		task->tk_status = 0;
941 		rpc_restart_call_prepare(task);
942 		return;
943 	case 1:
944 		break;
945 	case -1:
946 		/* Network partition? */
947 		nfsd4_mark_cb_down(clp, task->tk_status);
948 		break;
949 	default:
950 		BUG();
951 	}
952 	cb->cb_done = true;
953 }
954 
955 static void nfsd4_cb_release(void *calldata)
956 {
957 	struct nfsd4_callback *cb = calldata;
958 	struct nfs4_client *clp = cb->cb_clp;
959 
960 	if (cb->cb_done) {
961 		spin_lock(&clp->cl_lock);
962 		list_del(&cb->cb_per_client);
963 		spin_unlock(&clp->cl_lock);
964 
965 		cb->cb_ops->release(cb);
966 	}
967 }
968 
969 static const struct rpc_call_ops nfsd4_cb_ops = {
970 	.rpc_call_prepare = nfsd4_cb_prepare,
971 	.rpc_call_done = nfsd4_cb_done,
972 	.rpc_release = nfsd4_cb_release,
973 };
974 
975 int nfsd4_create_callback_queue(void)
976 {
977 	callback_wq = create_singlethread_workqueue("nfsd4_callbacks");
978 	if (!callback_wq)
979 		return -ENOMEM;
980 	return 0;
981 }
982 
983 void nfsd4_destroy_callback_queue(void)
984 {
985 	destroy_workqueue(callback_wq);
986 }
987 
988 /* must be called under the state lock */
989 void nfsd4_shutdown_callback(struct nfs4_client *clp)
990 {
991 	set_bit(NFSD4_CLIENT_CB_KILL, &clp->cl_flags);
992 	/*
993 	 * Note this won't actually result in a null callback;
994 	 * instead, nfsd4_run_cb_null() will detect the killed
995 	 * client, destroy the rpc client, and stop:
996 	 */
997 	nfsd4_run_cb(&clp->cl_cb_null);
998 	flush_workqueue(callback_wq);
999 }
1000 
1001 /* requires cl_lock: */
1002 static struct nfsd4_conn * __nfsd4_find_backchannel(struct nfs4_client *clp)
1003 {
1004 	struct nfsd4_session *s;
1005 	struct nfsd4_conn *c;
1006 
1007 	list_for_each_entry(s, &clp->cl_sessions, se_perclnt) {
1008 		list_for_each_entry(c, &s->se_conns, cn_persession) {
1009 			if (c->cn_flags & NFS4_CDFC4_BACK)
1010 				return c;
1011 		}
1012 	}
1013 	return NULL;
1014 }
1015 
1016 static void nfsd4_process_cb_update(struct nfsd4_callback *cb)
1017 {
1018 	struct nfs4_cb_conn conn;
1019 	struct nfs4_client *clp = cb->cb_clp;
1020 	struct nfsd4_session *ses = NULL;
1021 	struct nfsd4_conn *c;
1022 	int err;
1023 
1024 	/*
1025 	 * This is either an update, or the client dying; in either case,
1026 	 * kill the old client:
1027 	 */
1028 	if (clp->cl_cb_client) {
1029 		rpc_shutdown_client(clp->cl_cb_client);
1030 		clp->cl_cb_client = NULL;
1031 		put_rpccred(clp->cl_cb_cred);
1032 		clp->cl_cb_cred = NULL;
1033 	}
1034 	if (clp->cl_cb_conn.cb_xprt) {
1035 		svc_xprt_put(clp->cl_cb_conn.cb_xprt);
1036 		clp->cl_cb_conn.cb_xprt = NULL;
1037 	}
1038 	if (test_bit(NFSD4_CLIENT_CB_KILL, &clp->cl_flags))
1039 		return;
1040 	spin_lock(&clp->cl_lock);
1041 	/*
1042 	 * Only serialized callback code is allowed to clear these
1043 	 * flags; main nfsd code can only set them:
1044 	 */
1045 	BUG_ON(!(clp->cl_flags & NFSD4_CLIENT_CB_FLAG_MASK));
1046 	clear_bit(NFSD4_CLIENT_CB_UPDATE, &clp->cl_flags);
1047 	memcpy(&conn, &cb->cb_clp->cl_cb_conn, sizeof(struct nfs4_cb_conn));
1048 	c = __nfsd4_find_backchannel(clp);
1049 	if (c) {
1050 		svc_xprt_get(c->cn_xprt);
1051 		conn.cb_xprt = c->cn_xprt;
1052 		ses = c->cn_session;
1053 	}
1054 	spin_unlock(&clp->cl_lock);
1055 
1056 	err = setup_callback_client(clp, &conn, ses);
1057 	if (err) {
1058 		nfsd4_mark_cb_down(clp, err);
1059 		return;
1060 	}
1061 	/* Yay, the callback channel's back! Restart any callbacks: */
1062 	list_for_each_entry(cb, &clp->cl_callbacks, cb_per_client)
1063 		queue_work(callback_wq, &cb->cb_work);
1064 }
1065 
1066 static void
1067 nfsd4_run_cb_work(struct work_struct *work)
1068 {
1069 	struct nfsd4_callback *cb =
1070 		container_of(work, struct nfsd4_callback, cb_work);
1071 	struct nfs4_client *clp = cb->cb_clp;
1072 	struct rpc_clnt *clnt;
1073 
1074 	if (cb->cb_ops && cb->cb_ops->prepare)
1075 		cb->cb_ops->prepare(cb);
1076 
1077 	if (clp->cl_flags & NFSD4_CLIENT_CB_FLAG_MASK)
1078 		nfsd4_process_cb_update(cb);
1079 
1080 	clnt = clp->cl_cb_client;
1081 	if (!clnt) {
1082 		/* Callback channel broken, or client killed; give up: */
1083 		if (cb->cb_ops && cb->cb_ops->release)
1084 			cb->cb_ops->release(cb);
1085 		return;
1086 	}
1087 	cb->cb_msg.rpc_cred = clp->cl_cb_cred;
1088 	rpc_call_async(clnt, &cb->cb_msg, RPC_TASK_SOFT | RPC_TASK_SOFTCONN,
1089 			cb->cb_ops ? &nfsd4_cb_ops : &nfsd4_cb_probe_ops, cb);
1090 }
1091 
1092 void nfsd4_init_cb(struct nfsd4_callback *cb, struct nfs4_client *clp,
1093 		struct nfsd4_callback_ops *ops, enum nfsd4_cb_op op)
1094 {
1095 	cb->cb_clp = clp;
1096 	cb->cb_msg.rpc_proc = &nfs4_cb_procedures[op];
1097 	cb->cb_msg.rpc_argp = cb;
1098 	cb->cb_msg.rpc_resp = cb;
1099 	cb->cb_ops = ops;
1100 	INIT_WORK(&cb->cb_work, nfsd4_run_cb_work);
1101 	INIT_LIST_HEAD(&cb->cb_per_client);
1102 	cb->cb_done = true;
1103 }
1104 
1105 void nfsd4_run_cb(struct nfsd4_callback *cb)
1106 {
1107 	queue_work(callback_wq, &cb->cb_work);
1108 }
1109