2 * Copyright (c) 2001 The Regents of the University of Michigan.
5 * Kendrick Smith <kmsmith@umich.edu>
6 * Andy Adamson <andros@umich.edu>
8 * Redistribution and use in source and binary forms, with or without
9 * modification, are permitted provided that the following conditions
12 * 1. Redistributions of source code must retain the above copyright
13 * notice, this list of conditions and the following disclaimer.
14 * 2. Redistributions in binary form must reproduce the above copyright
15 * notice, this list of conditions and the following disclaimer in the
16 * documentation and/or other materials provided with the distribution.
17 * 3. Neither the name of the University nor the names of its
18 * contributors may be used to endorse or promote products derived
19 * from this software without specific prior written permission.
21 * THIS SOFTWARE IS PROVIDED ``AS IS'' AND ANY EXPRESS OR IMPLIED
22 * WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
23 * MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
24 * DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
25 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
26 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
27 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR
28 * BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
29 * LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
30 * NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
31 * SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
34 #include <linux/nfs4.h>
35 #include <linux/sunrpc/clnt.h>
36 #include <linux/sunrpc/xprt.h>
37 #include <linux/sunrpc/svc_xprt.h>
38 #include <linux/slab.h>
46 #define NFSDDBG_FACILITY NFSDDBG_PROC
48 static void nfsd4_mark_cb_fault(struct nfs4_client *, int reason);
50 #define NFSPROC4_CB_NULL 0
51 #define NFSPROC4_CB_COMPOUND 1
53 /* Index of predefined Linux callback client operations */
55 struct nfs4_cb_compound_hdr {
57 u32 ident; /* minorversion 0 only */
65 static __be32 *xdr_encode_empty_array(__be32 *p)
72 * Encode/decode NFSv4 CB basic data types
74 * Basic NFSv4 callback data types are defined in section 15 of RFC
75 * 3530: "Network File System (NFS) version 4 Protocol" and section
76 * 20 of RFC 5661: "Network File System (NFS) Version 4 Minor Version
80 static void encode_uint32(struct xdr_stream *xdr, u32 n)
82 WARN_ON_ONCE(xdr_stream_encode_u32(xdr, n) < 0);
85 static void encode_bitmap4(struct xdr_stream *xdr, const __u32 *bitmap,
88 WARN_ON_ONCE(xdr_stream_encode_uint32_array(xdr, bitmap, len) < 0);
91 static void encode_nfs_cb_opnum4(struct xdr_stream *xdr, enum nfs_cb_opnum4 op)
95 p = xdr_reserve_space(xdr, 4);
102 * typedef opaque nfs_fh4<NFS4_FHSIZE>;
104 static void encode_nfs_fh4(struct xdr_stream *xdr, const struct knfsd_fh *fh)
106 u32 length = fh->fh_size;
109 BUG_ON(length > NFS4_FHSIZE);
110 p = xdr_reserve_space(xdr, 4 + length);
111 xdr_encode_opaque(p, &fh->fh_raw, length);
122 static void encode_stateid4(struct xdr_stream *xdr, const stateid_t *sid)
126 p = xdr_reserve_space(xdr, NFS4_STATEID_SIZE);
127 *p++ = cpu_to_be32(sid->si_generation);
128 xdr_encode_opaque_fixed(p, &sid->si_opaque, NFS4_STATEID_OTHER_SIZE);
134 * typedef opaque sessionid4[NFS4_SESSIONID_SIZE];
136 static void encode_sessionid4(struct xdr_stream *xdr,
137 const struct nfsd4_session *session)
141 p = xdr_reserve_space(xdr, NFS4_MAX_SESSIONID_LEN);
142 xdr_encode_opaque_fixed(p, session->se_sessionid.data,
143 NFS4_MAX_SESSIONID_LEN);
149 static const struct {
152 } nfs_cb_errtbl[] = {
154 { NFS4ERR_PERM, -EPERM },
155 { NFS4ERR_NOENT, -ENOENT },
156 { NFS4ERR_IO, -EIO },
157 { NFS4ERR_NXIO, -ENXIO },
158 { NFS4ERR_ACCESS, -EACCES },
159 { NFS4ERR_EXIST, -EEXIST },
160 { NFS4ERR_XDEV, -EXDEV },
161 { NFS4ERR_NOTDIR, -ENOTDIR },
162 { NFS4ERR_ISDIR, -EISDIR },
163 { NFS4ERR_INVAL, -EINVAL },
164 { NFS4ERR_FBIG, -EFBIG },
165 { NFS4ERR_NOSPC, -ENOSPC },
166 { NFS4ERR_ROFS, -EROFS },
167 { NFS4ERR_MLINK, -EMLINK },
168 { NFS4ERR_NAMETOOLONG, -ENAMETOOLONG },
169 { NFS4ERR_NOTEMPTY, -ENOTEMPTY },
170 { NFS4ERR_DQUOT, -EDQUOT },
171 { NFS4ERR_STALE, -ESTALE },
172 { NFS4ERR_BADHANDLE, -EBADHANDLE },
173 { NFS4ERR_BAD_COOKIE, -EBADCOOKIE },
174 { NFS4ERR_NOTSUPP, -ENOTSUPP },
175 { NFS4ERR_TOOSMALL, -ETOOSMALL },
176 { NFS4ERR_SERVERFAULT, -ESERVERFAULT },
177 { NFS4ERR_BADTYPE, -EBADTYPE },
178 { NFS4ERR_LOCKED, -EAGAIN },
179 { NFS4ERR_RESOURCE, -EREMOTEIO },
180 { NFS4ERR_SYMLINK, -ELOOP },
181 { NFS4ERR_OP_ILLEGAL, -EOPNOTSUPP },
182 { NFS4ERR_DEADLOCK, -EDEADLK },
187 * If we cannot translate the error, the recovery routines should
190 * Note: remaining NFSv4 error codes have values > 10000, so should
191 * not conflict with native Linux error codes.
193 static int nfs_cb_stat_to_errno(int status)
197 for (i = 0; nfs_cb_errtbl[i].stat != -1; i++) {
198 if (nfs_cb_errtbl[i].stat == status)
199 return nfs_cb_errtbl[i].errno;
202 dprintk("NFSD: Unrecognized NFS CB status value: %u\n", status);
206 static int decode_cb_op_status(struct xdr_stream *xdr,
207 enum nfs_cb_opnum4 expected, int *status)
212 p = xdr_inline_decode(xdr, 4 + 4);
213 if (unlikely(p == NULL))
215 op = be32_to_cpup(p++);
216 if (unlikely(op != expected))
218 *status = nfs_cb_stat_to_errno(be32_to_cpup(p));
223 dprintk("NFSD: Callback server returned operation %d but "
224 "we issued a request for %d\n", op, expected);
231 * struct CB_COMPOUND4args {
233 * uint32_t minorversion;
234 * uint32_t callback_ident;
235 * nfs_cb_argop4 argarray<>;
238 static void encode_cb_compound4args(struct xdr_stream *xdr,
239 struct nfs4_cb_compound_hdr *hdr)
243 p = xdr_reserve_space(xdr, 4 + 4 + 4 + 4);
244 p = xdr_encode_empty_array(p); /* empty tag */
245 *p++ = cpu_to_be32(hdr->minorversion);
246 *p++ = cpu_to_be32(hdr->ident);
249 *p = cpu_to_be32(hdr->nops); /* argarray element count */
253 * Update argarray element count
255 static void encode_cb_nops(struct nfs4_cb_compound_hdr *hdr)
257 BUG_ON(hdr->nops > NFS4_MAX_BACK_CHANNEL_OPS);
258 *hdr->nops_p = cpu_to_be32(hdr->nops);
264 * struct CB_COMPOUND4res {
267 * nfs_cb_resop4 resarray<>;
270 static int decode_cb_compound4res(struct xdr_stream *xdr,
271 struct nfs4_cb_compound_hdr *hdr)
276 p = xdr_inline_decode(xdr, 4 + 4);
277 if (unlikely(p == NULL))
279 hdr->status = be32_to_cpup(p++);
281 length = be32_to_cpup(p++);
282 p = xdr_inline_decode(xdr, length + 4);
283 if (unlikely(p == NULL))
285 p += XDR_QUADLEN(length);
286 hdr->nops = be32_to_cpup(p);
295 * struct CB_RECALL4args {
301 static void encode_cb_recall4args(struct xdr_stream *xdr,
302 const struct nfs4_delegation *dp,
303 struct nfs4_cb_compound_hdr *hdr)
307 encode_nfs_cb_opnum4(xdr, OP_CB_RECALL);
308 encode_stateid4(xdr, &dp->dl_stid.sc_stateid);
310 p = xdr_reserve_space(xdr, 4);
311 *p++ = xdr_zero; /* truncate */
313 encode_nfs_fh4(xdr, &dp->dl_stid.sc_file->fi_fhandle);
321 * struct CB_RECALLANY4args {
322 * uint32_t craa_objects_to_keep;
323 * bitmap4 craa_type_mask;
327 encode_cb_recallany4args(struct xdr_stream *xdr,
328 struct nfs4_cb_compound_hdr *hdr, struct nfsd4_cb_recall_any *ra)
330 encode_nfs_cb_opnum4(xdr, OP_CB_RECALL_ANY);
331 encode_uint32(xdr, ra->ra_keep);
332 encode_bitmap4(xdr, ra->ra_bmval, ARRAY_SIZE(ra->ra_bmval));
339 * struct CB_SEQUENCE4args {
340 * sessionid4 csa_sessionid;
341 * sequenceid4 csa_sequenceid;
342 * slotid4 csa_slotid;
343 * slotid4 csa_highest_slotid;
344 * bool csa_cachethis;
345 * referring_call_list4 csa_referring_call_lists<>;
348 static void encode_cb_sequence4args(struct xdr_stream *xdr,
349 const struct nfsd4_callback *cb,
350 struct nfs4_cb_compound_hdr *hdr)
352 struct nfsd4_session *session = cb->cb_clp->cl_cb_session;
355 if (hdr->minorversion == 0)
358 encode_nfs_cb_opnum4(xdr, OP_CB_SEQUENCE);
359 encode_sessionid4(xdr, session);
361 p = xdr_reserve_space(xdr, 4 + 4 + 4 + 4 + 4);
362 *p++ = cpu_to_be32(session->se_cb_seq_nr); /* csa_sequenceid */
363 *p++ = xdr_zero; /* csa_slotid */
364 *p++ = xdr_zero; /* csa_highest_slotid */
365 *p++ = xdr_zero; /* csa_cachethis */
366 xdr_encode_empty_array(p); /* csa_referring_call_lists */
374 * struct CB_SEQUENCE4resok {
375 * sessionid4 csr_sessionid;
376 * sequenceid4 csr_sequenceid;
377 * slotid4 csr_slotid;
378 * slotid4 csr_highest_slotid;
379 * slotid4 csr_target_highest_slotid;
382 * union CB_SEQUENCE4res switch (nfsstat4 csr_status) {
384 * CB_SEQUENCE4resok csr_resok4;
389 * Our current back channel implmentation supports a single backchannel
390 * with a single slot.
392 static int decode_cb_sequence4resok(struct xdr_stream *xdr,
393 struct nfsd4_callback *cb)
395 struct nfsd4_session *session = cb->cb_clp->cl_cb_session;
396 int status = -ESERVERFAULT;
401 * If the server returns different values for sessionID, slotID or
402 * sequence number, the server is looney tunes.
404 p = xdr_inline_decode(xdr, NFS4_MAX_SESSIONID_LEN + 4 + 4 + 4 + 4);
405 if (unlikely(p == NULL))
408 if (memcmp(p, session->se_sessionid.data, NFS4_MAX_SESSIONID_LEN)) {
409 dprintk("NFS: %s Invalid session id\n", __func__);
412 p += XDR_QUADLEN(NFS4_MAX_SESSIONID_LEN);
414 dummy = be32_to_cpup(p++);
415 if (dummy != session->se_cb_seq_nr) {
416 dprintk("NFS: %s Invalid sequence number\n", __func__);
420 dummy = be32_to_cpup(p++);
422 dprintk("NFS: %s Invalid slotid\n", __func__);
427 * FIXME: process highest slotid and target highest slotid
431 cb->cb_seq_status = status;
438 static int decode_cb_sequence4res(struct xdr_stream *xdr,
439 struct nfsd4_callback *cb)
443 if (cb->cb_clp->cl_minorversion == 0)
446 status = decode_cb_op_status(xdr, OP_CB_SEQUENCE, &cb->cb_seq_status);
447 if (unlikely(status || cb->cb_seq_status))
450 return decode_cb_sequence4resok(xdr, cb);
454 * NFSv4.0 and NFSv4.1 XDR encode functions
456 * NFSv4.0 callback argument types are defined in section 15 of RFC
457 * 3530: "Network File System (NFS) version 4 Protocol" and section 20
458 * of RFC 5661: "Network File System (NFS) Version 4 Minor Version 1
463 * NB: Without this zero space reservation, callbacks over krb5p fail
465 static void nfs4_xdr_enc_cb_null(struct rpc_rqst *req, struct xdr_stream *xdr,
466 const void *__unused)
468 xdr_reserve_space(xdr, 0);
472 * 20.2. Operation 4: CB_RECALL - Recall a Delegation
474 static void nfs4_xdr_enc_cb_recall(struct rpc_rqst *req, struct xdr_stream *xdr,
477 const struct nfsd4_callback *cb = data;
478 const struct nfs4_delegation *dp = cb_to_delegation(cb);
479 struct nfs4_cb_compound_hdr hdr = {
480 .ident = cb->cb_clp->cl_cb_ident,
481 .minorversion = cb->cb_clp->cl_minorversion,
484 encode_cb_compound4args(xdr, &hdr);
485 encode_cb_sequence4args(xdr, cb, &hdr);
486 encode_cb_recall4args(xdr, dp, &hdr);
487 encode_cb_nops(&hdr);
491 * 20.6. Operation 8: CB_RECALL_ANY - Keep Any N Recallable Objects
494 nfs4_xdr_enc_cb_recall_any(struct rpc_rqst *req,
495 struct xdr_stream *xdr, const void *data)
497 const struct nfsd4_callback *cb = data;
498 struct nfsd4_cb_recall_any *ra;
499 struct nfs4_cb_compound_hdr hdr = {
500 .ident = cb->cb_clp->cl_cb_ident,
501 .minorversion = cb->cb_clp->cl_minorversion,
504 ra = container_of(cb, struct nfsd4_cb_recall_any, ra_cb);
505 encode_cb_compound4args(xdr, &hdr);
506 encode_cb_sequence4args(xdr, cb, &hdr);
507 encode_cb_recallany4args(xdr, &hdr, ra);
508 encode_cb_nops(&hdr);
512 * NFSv4.0 and NFSv4.1 XDR decode functions
514 * NFSv4.0 callback result types are defined in section 15 of RFC
515 * 3530: "Network File System (NFS) version 4 Protocol" and section 20
516 * of RFC 5661: "Network File System (NFS) Version 4 Minor Version 1
520 static int nfs4_xdr_dec_cb_null(struct rpc_rqst *req, struct xdr_stream *xdr,
527 * 20.2. Operation 4: CB_RECALL - Recall a Delegation
529 static int nfs4_xdr_dec_cb_recall(struct rpc_rqst *rqstp,
530 struct xdr_stream *xdr,
533 struct nfsd4_callback *cb = data;
534 struct nfs4_cb_compound_hdr hdr;
537 status = decode_cb_compound4res(xdr, &hdr);
538 if (unlikely(status))
541 status = decode_cb_sequence4res(xdr, cb);
542 if (unlikely(status || cb->cb_seq_status))
545 return decode_cb_op_status(xdr, OP_CB_RECALL, &cb->cb_status);
549 * 20.6. Operation 8: CB_RECALL_ANY - Keep Any N Recallable Objects
552 nfs4_xdr_dec_cb_recall_any(struct rpc_rqst *rqstp,
553 struct xdr_stream *xdr,
556 struct nfsd4_callback *cb = data;
557 struct nfs4_cb_compound_hdr hdr;
560 status = decode_cb_compound4res(xdr, &hdr);
561 if (unlikely(status))
563 status = decode_cb_sequence4res(xdr, cb);
564 if (unlikely(status || cb->cb_seq_status))
566 status = decode_cb_op_status(xdr, OP_CB_RECALL_ANY, &cb->cb_status);
570 #ifdef CONFIG_NFSD_PNFS
572 * CB_LAYOUTRECALL4args
574 * struct layoutrecall_file4 {
576 * offset4 lor_offset;
577 * length4 lor_length;
578 * stateid4 lor_stateid;
581 * union layoutrecall4 switch(layoutrecall_type4 lor_recalltype) {
582 * case LAYOUTRECALL4_FILE:
583 * layoutrecall_file4 lor_layout;
584 * case LAYOUTRECALL4_FSID:
586 * case LAYOUTRECALL4_ALL:
590 * struct CB_LAYOUTRECALL4args {
591 * layouttype4 clora_type;
592 * layoutiomode4 clora_iomode;
593 * bool clora_changed;
594 * layoutrecall4 clora_recall;
597 static void encode_cb_layout4args(struct xdr_stream *xdr,
598 const struct nfs4_layout_stateid *ls,
599 struct nfs4_cb_compound_hdr *hdr)
603 BUG_ON(hdr->minorversion == 0);
605 p = xdr_reserve_space(xdr, 5 * 4);
606 *p++ = cpu_to_be32(OP_CB_LAYOUTRECALL);
607 *p++ = cpu_to_be32(ls->ls_layout_type);
608 *p++ = cpu_to_be32(IOMODE_ANY);
609 *p++ = cpu_to_be32(1);
610 *p = cpu_to_be32(RETURN_FILE);
612 encode_nfs_fh4(xdr, &ls->ls_stid.sc_file->fi_fhandle);
614 p = xdr_reserve_space(xdr, 2 * 8);
615 p = xdr_encode_hyper(p, 0);
616 xdr_encode_hyper(p, NFS4_MAX_UINT64);
618 encode_stateid4(xdr, &ls->ls_recall_sid);
623 static void nfs4_xdr_enc_cb_layout(struct rpc_rqst *req,
624 struct xdr_stream *xdr,
627 const struct nfsd4_callback *cb = data;
628 const struct nfs4_layout_stateid *ls =
629 container_of(cb, struct nfs4_layout_stateid, ls_recall);
630 struct nfs4_cb_compound_hdr hdr = {
632 .minorversion = cb->cb_clp->cl_minorversion,
635 encode_cb_compound4args(xdr, &hdr);
636 encode_cb_sequence4args(xdr, cb, &hdr);
637 encode_cb_layout4args(xdr, ls, &hdr);
638 encode_cb_nops(&hdr);
641 static int nfs4_xdr_dec_cb_layout(struct rpc_rqst *rqstp,
642 struct xdr_stream *xdr,
645 struct nfsd4_callback *cb = data;
646 struct nfs4_cb_compound_hdr hdr;
649 status = decode_cb_compound4res(xdr, &hdr);
650 if (unlikely(status))
653 status = decode_cb_sequence4res(xdr, cb);
654 if (unlikely(status || cb->cb_seq_status))
657 return decode_cb_op_status(xdr, OP_CB_LAYOUTRECALL, &cb->cb_status);
659 #endif /* CONFIG_NFSD_PNFS */
661 static void encode_stateowner(struct xdr_stream *xdr, struct nfs4_stateowner *so)
665 p = xdr_reserve_space(xdr, 8 + 4 + so->so_owner.len);
666 p = xdr_encode_opaque_fixed(p, &so->so_client->cl_clientid, 8);
667 xdr_encode_opaque(p, so->so_owner.data, so->so_owner.len);
670 static void nfs4_xdr_enc_cb_notify_lock(struct rpc_rqst *req,
671 struct xdr_stream *xdr,
674 const struct nfsd4_callback *cb = data;
675 const struct nfsd4_blocked_lock *nbl =
676 container_of(cb, struct nfsd4_blocked_lock, nbl_cb);
677 struct nfs4_lockowner *lo = (struct nfs4_lockowner *)nbl->nbl_lock.c.flc_owner;
678 struct nfs4_cb_compound_hdr hdr = {
680 .minorversion = cb->cb_clp->cl_minorversion,
685 BUG_ON(hdr.minorversion == 0);
687 encode_cb_compound4args(xdr, &hdr);
688 encode_cb_sequence4args(xdr, cb, &hdr);
690 p = xdr_reserve_space(xdr, 4);
691 *p = cpu_to_be32(OP_CB_NOTIFY_LOCK);
692 encode_nfs_fh4(xdr, &nbl->nbl_fh);
693 encode_stateowner(xdr, &lo->lo_owner);
696 encode_cb_nops(&hdr);
699 static int nfs4_xdr_dec_cb_notify_lock(struct rpc_rqst *rqstp,
700 struct xdr_stream *xdr,
703 struct nfsd4_callback *cb = data;
704 struct nfs4_cb_compound_hdr hdr;
707 status = decode_cb_compound4res(xdr, &hdr);
708 if (unlikely(status))
711 status = decode_cb_sequence4res(xdr, cb);
712 if (unlikely(status || cb->cb_seq_status))
715 return decode_cb_op_status(xdr, OP_CB_NOTIFY_LOCK, &cb->cb_status);
719 * struct write_response4 {
720 * stateid4 wr_callback_id<1>;
722 * stable_how4 wr_committed;
723 * verifier4 wr_writeverf;
725 * union offload_info4 switch (nfsstat4 coa_status) {
727 * write_response4 coa_resok4;
729 * length4 coa_bytes_copied;
731 * struct CB_OFFLOAD4args {
733 * stateid4 coa_stateid;
734 * offload_info4 coa_offload_info;
737 static void encode_offload_info4(struct xdr_stream *xdr,
738 const struct nfsd4_cb_offload *cbo)
742 p = xdr_reserve_space(xdr, 4);
744 switch (cbo->co_nfserr) {
746 p = xdr_reserve_space(xdr, 4 + 8 + 4 + NFS4_VERIFIER_SIZE);
747 p = xdr_encode_empty_array(p);
748 p = xdr_encode_hyper(p, cbo->co_res.wr_bytes_written);
749 *p++ = cpu_to_be32(cbo->co_res.wr_stable_how);
750 p = xdr_encode_opaque_fixed(p, cbo->co_res.wr_verifier.data,
754 p = xdr_reserve_space(xdr, 8);
755 /* We always return success if bytes were written */
756 p = xdr_encode_hyper(p, 0);
760 static void encode_cb_offload4args(struct xdr_stream *xdr,
761 const struct nfsd4_cb_offload *cbo,
762 struct nfs4_cb_compound_hdr *hdr)
766 p = xdr_reserve_space(xdr, 4);
767 *p = cpu_to_be32(OP_CB_OFFLOAD);
768 encode_nfs_fh4(xdr, &cbo->co_fh);
769 encode_stateid4(xdr, &cbo->co_res.cb_stateid);
770 encode_offload_info4(xdr, cbo);
775 static void nfs4_xdr_enc_cb_offload(struct rpc_rqst *req,
776 struct xdr_stream *xdr,
779 const struct nfsd4_callback *cb = data;
780 const struct nfsd4_cb_offload *cbo =
781 container_of(cb, struct nfsd4_cb_offload, co_cb);
782 struct nfs4_cb_compound_hdr hdr = {
784 .minorversion = cb->cb_clp->cl_minorversion,
787 encode_cb_compound4args(xdr, &hdr);
788 encode_cb_sequence4args(xdr, cb, &hdr);
789 encode_cb_offload4args(xdr, cbo, &hdr);
790 encode_cb_nops(&hdr);
793 static int nfs4_xdr_dec_cb_offload(struct rpc_rqst *rqstp,
794 struct xdr_stream *xdr,
797 struct nfsd4_callback *cb = data;
798 struct nfs4_cb_compound_hdr hdr;
801 status = decode_cb_compound4res(xdr, &hdr);
802 if (unlikely(status))
805 status = decode_cb_sequence4res(xdr, cb);
806 if (unlikely(status || cb->cb_seq_status))
809 return decode_cb_op_status(xdr, OP_CB_OFFLOAD, &cb->cb_status);
812 * RPC procedure tables
814 #define PROC(proc, call, argtype, restype) \
815 [NFSPROC4_CLNT_##proc] = { \
816 .p_proc = NFSPROC4_CB_##call, \
817 .p_encode = nfs4_xdr_enc_##argtype, \
818 .p_decode = nfs4_xdr_dec_##restype, \
819 .p_arglen = NFS4_enc_##argtype##_sz, \
820 .p_replen = NFS4_dec_##restype##_sz, \
821 .p_statidx = NFSPROC4_CB_##call, \
825 static const struct rpc_procinfo nfs4_cb_procedures[] = {
826 PROC(CB_NULL, NULL, cb_null, cb_null),
827 PROC(CB_RECALL, COMPOUND, cb_recall, cb_recall),
828 #ifdef CONFIG_NFSD_PNFS
829 PROC(CB_LAYOUT, COMPOUND, cb_layout, cb_layout),
831 PROC(CB_NOTIFY_LOCK, COMPOUND, cb_notify_lock, cb_notify_lock),
832 PROC(CB_OFFLOAD, COMPOUND, cb_offload, cb_offload),
833 PROC(CB_RECALL_ANY, COMPOUND, cb_recall_any, cb_recall_any),
836 static unsigned int nfs4_cb_counts[ARRAY_SIZE(nfs4_cb_procedures)];
837 static const struct rpc_version nfs_cb_version4 = {
839 * Note on the callback rpc program version number: despite language in rfc
840 * 5661 section 18.36.3 requiring servers to use 4 in this field, the
841 * official xdr descriptions for both 4.0 and 4.1 specify version 1, and
842 * in practice that appears to be what implementations use. The section
843 * 18.36.3 language is expected to be fixed in an erratum.
846 .nrprocs = ARRAY_SIZE(nfs4_cb_procedures),
847 .procs = nfs4_cb_procedures,
848 .counts = nfs4_cb_counts,
851 static const struct rpc_version *nfs_cb_version[2] = {
852 [1] = &nfs_cb_version4,
855 static const struct rpc_program cb_program;
857 static struct rpc_stat cb_stats = {
858 .program = &cb_program
861 #define NFS4_CALLBACK 0x40000000
862 static const struct rpc_program cb_program = {
864 .number = NFS4_CALLBACK,
865 .nrvers = ARRAY_SIZE(nfs_cb_version),
866 .version = nfs_cb_version,
868 .pipe_dir_name = "nfsd4_cb",
871 static int max_cb_time(struct net *net)
873 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
876 * nfsd4_lease is set to at most one hour in __nfsd4_write_time,
877 * so we can use 32-bit math on it. Warn if that assumption
878 * ever stops being true.
880 if (WARN_ON_ONCE(nn->nfsd4_lease > 3600))
883 return max(((u32)nn->nfsd4_lease)/10, 1u) * HZ;
886 static struct workqueue_struct *callback_wq;
888 static bool nfsd4_queue_cb(struct nfsd4_callback *cb)
890 return queue_work(callback_wq, &cb->cb_work);
893 static void nfsd41_cb_inflight_begin(struct nfs4_client *clp)
895 atomic_inc(&clp->cl_cb_inflight);
898 static void nfsd41_cb_inflight_end(struct nfs4_client *clp)
901 if (atomic_dec_and_test(&clp->cl_cb_inflight))
902 wake_up_var(&clp->cl_cb_inflight);
905 static void nfsd41_cb_inflight_wait_complete(struct nfs4_client *clp)
907 wait_var_event(&clp->cl_cb_inflight,
908 !atomic_read(&clp->cl_cb_inflight));
911 static const struct cred *get_backchannel_cred(struct nfs4_client *clp, struct rpc_clnt *client, struct nfsd4_session *ses)
913 if (clp->cl_minorversion == 0) {
914 client->cl_principal = clp->cl_cred.cr_targ_princ ?
915 clp->cl_cred.cr_targ_princ : "nfs";
917 return get_cred(rpc_machine_cred());
921 kcred = prepare_kernel_cred(&init_task);
925 kcred->fsuid = ses->se_cb_sec.uid;
926 kcred->fsgid = ses->se_cb_sec.gid;
931 static int setup_callback_client(struct nfs4_client *clp, struct nfs4_cb_conn *conn, struct nfsd4_session *ses)
933 int maxtime = max_cb_time(clp->net);
934 struct rpc_timeout timeparms = {
935 .to_initval = maxtime,
937 .to_maxval = maxtime,
939 struct rpc_create_args args = {
941 .address = (struct sockaddr *) &conn->cb_addr,
942 .addrsize = conn->cb_addrlen,
943 .saddress = (struct sockaddr *) &conn->cb_saddr,
944 .timeout = &timeparms,
945 .program = &cb_program,
947 .flags = (RPC_CLNT_CREATE_NOPING | RPC_CLNT_CREATE_QUIET),
948 .cred = current_cred(),
950 struct rpc_clnt *client;
951 const struct cred *cred;
953 if (clp->cl_minorversion == 0) {
954 if (!clp->cl_cred.cr_principal &&
955 (clp->cl_cred.cr_flavor >= RPC_AUTH_GSS_KRB5)) {
956 trace_nfsd_cb_setup_err(clp, -EINVAL);
959 args.client_name = clp->cl_cred.cr_principal;
960 args.prognumber = conn->cb_prog;
961 args.protocol = XPRT_TRANSPORT_TCP;
962 args.authflavor = clp->cl_cred.cr_flavor;
963 clp->cl_cb_ident = conn->cb_ident;
967 clp->cl_cb_session = ses;
968 args.bc_xprt = conn->cb_xprt;
969 args.prognumber = clp->cl_cb_session->se_cb_prog;
970 args.protocol = conn->cb_xprt->xpt_class->xcl_ident |
972 args.authflavor = ses->se_cb_sec.flavor;
974 /* Create RPC client */
975 client = rpc_create(&args);
976 if (IS_ERR(client)) {
977 trace_nfsd_cb_setup_err(clp, PTR_ERR(client));
978 return PTR_ERR(client);
980 cred = get_backchannel_cred(clp, client, ses);
982 trace_nfsd_cb_setup_err(clp, -ENOMEM);
983 rpc_shutdown_client(client);
987 if (clp->cl_minorversion != 0)
988 clp->cl_cb_conn.cb_xprt = conn->cb_xprt;
989 clp->cl_cb_client = client;
990 clp->cl_cb_cred = cred;
992 trace_nfsd_cb_setup(clp, rpc_peeraddr2str(client, RPC_DISPLAY_NETID),
998 static void nfsd4_mark_cb_state(struct nfs4_client *clp, int newstate)
1000 if (clp->cl_cb_state != newstate) {
1001 clp->cl_cb_state = newstate;
1002 trace_nfsd_cb_state(clp);
1006 static void nfsd4_mark_cb_down(struct nfs4_client *clp, int reason)
1008 if (test_bit(NFSD4_CLIENT_CB_UPDATE, &clp->cl_flags))
1010 nfsd4_mark_cb_state(clp, NFSD4_CB_DOWN);
1013 static void nfsd4_mark_cb_fault(struct nfs4_client *clp, int reason)
1015 if (test_bit(NFSD4_CLIENT_CB_UPDATE, &clp->cl_flags))
1017 nfsd4_mark_cb_state(clp, NFSD4_CB_FAULT);
1020 static void nfsd4_cb_probe_done(struct rpc_task *task, void *calldata)
1022 struct nfs4_client *clp = container_of(calldata, struct nfs4_client, cl_cb_null);
1024 if (task->tk_status)
1025 nfsd4_mark_cb_down(clp, task->tk_status);
1027 nfsd4_mark_cb_state(clp, NFSD4_CB_UP);
1030 static void nfsd4_cb_probe_release(void *calldata)
1032 struct nfs4_client *clp = container_of(calldata, struct nfs4_client, cl_cb_null);
1034 nfsd41_cb_inflight_end(clp);
1038 static const struct rpc_call_ops nfsd4_cb_probe_ops = {
1039 /* XXX: release method to ensure we set the cb channel down if
1040 * necessary on early failure? */
1041 .rpc_call_done = nfsd4_cb_probe_done,
1042 .rpc_release = nfsd4_cb_probe_release,
1046 * Poke the callback thread to process any updates to the callback
1047 * parameters, and send a null probe.
1049 void nfsd4_probe_callback(struct nfs4_client *clp)
1051 trace_nfsd_cb_probe(clp);
1052 nfsd4_mark_cb_state(clp, NFSD4_CB_UNKNOWN);
1053 set_bit(NFSD4_CLIENT_CB_UPDATE, &clp->cl_flags);
1054 nfsd4_run_cb(&clp->cl_cb_null);
1057 void nfsd4_probe_callback_sync(struct nfs4_client *clp)
1059 nfsd4_probe_callback(clp);
1060 flush_workqueue(callback_wq);
1063 void nfsd4_change_callback(struct nfs4_client *clp, struct nfs4_cb_conn *conn)
1065 nfsd4_mark_cb_state(clp, NFSD4_CB_UNKNOWN);
1066 spin_lock(&clp->cl_lock);
1067 memcpy(&clp->cl_cb_conn, conn, sizeof(struct nfs4_cb_conn));
1068 spin_unlock(&clp->cl_lock);
1072 * There's currently a single callback channel slot.
1073 * If the slot is available, then mark it busy. Otherwise, set the
1074 * thread for sleeping on the callback RPC wait queue.
1076 static bool nfsd41_cb_get_slot(struct nfsd4_callback *cb, struct rpc_task *task)
1078 struct nfs4_client *clp = cb->cb_clp;
1080 if (!cb->cb_holds_slot &&
1081 test_and_set_bit(0, &clp->cl_cb_slot_busy) != 0) {
1082 rpc_sleep_on(&clp->cl_cb_waitq, task, NULL);
1084 if (test_and_set_bit(0, &clp->cl_cb_slot_busy) != 0) {
1085 dprintk("%s slot is busy\n", __func__);
1088 rpc_wake_up_queued_task(&clp->cl_cb_waitq, task);
1090 cb->cb_holds_slot = true;
1094 static void nfsd41_cb_release_slot(struct nfsd4_callback *cb)
1096 struct nfs4_client *clp = cb->cb_clp;
1098 if (cb->cb_holds_slot) {
1099 cb->cb_holds_slot = false;
1100 clear_bit(0, &clp->cl_cb_slot_busy);
1101 rpc_wake_up_next(&clp->cl_cb_waitq);
1105 static void nfsd41_destroy_cb(struct nfsd4_callback *cb)
1107 struct nfs4_client *clp = cb->cb_clp;
1109 nfsd41_cb_release_slot(cb);
1110 if (cb->cb_ops && cb->cb_ops->release)
1111 cb->cb_ops->release(cb);
1112 nfsd41_cb_inflight_end(clp);
1116 * TODO: cb_sequence should support referring call lists, cachethis, multiple
1117 * slots, and mark callback channel down on communication errors.
1119 static void nfsd4_cb_prepare(struct rpc_task *task, void *calldata)
1121 struct nfsd4_callback *cb = calldata;
1122 struct nfs4_client *clp = cb->cb_clp;
1123 u32 minorversion = clp->cl_minorversion;
1126 * cb_seq_status is only set in decode_cb_sequence4res,
1127 * and so will remain 1 if an rpc level failure occurs.
1129 cb->cb_seq_status = 1;
1131 if (minorversion && !nfsd41_cb_get_slot(cb, task))
1133 rpc_call_start(task);
1136 static bool nfsd4_cb_sequence_done(struct rpc_task *task, struct nfsd4_callback *cb)
1138 struct nfs4_client *clp = cb->cb_clp;
1139 struct nfsd4_session *session = clp->cl_cb_session;
1142 if (!clp->cl_minorversion) {
1144 * If the backchannel connection was shut down while this
1145 * task was queued, we need to resubmit it after setting up
1146 * a new backchannel connection.
1148 * Note that if we lost our callback connection permanently
1149 * the submission code will error out, so we don't need to
1150 * handle that case here.
1152 if (RPC_SIGNALLED(task))
1158 if (!cb->cb_holds_slot)
1161 switch (cb->cb_seq_status) {
1164 * No need for lock, access serialized in nfsd4_cb_prepare
1167 * If CB_SEQUENCE returns an error, then the state of the slot
1168 * (sequence ID, cached reply) MUST NOT change.
1170 ++session->se_cb_seq_nr;
1173 ++session->se_cb_seq_nr;
1176 case -NFS4ERR_BADSESSION:
1177 nfsd4_mark_cb_fault(cb->cb_clp, cb->cb_seq_status);
1180 case -NFS4ERR_DELAY:
1181 if (!rpc_restart_call(task))
1184 rpc_delay(task, 2 * HZ);
1186 case -NFS4ERR_BADSLOT:
1188 case -NFS4ERR_SEQ_MISORDERED:
1189 if (session->se_cb_seq_nr != 1) {
1190 session->se_cb_seq_nr = 1;
1195 nfsd4_mark_cb_fault(cb->cb_clp, cb->cb_seq_status);
1196 dprintk("%s: unprocessed error %d\n", __func__,
1200 nfsd41_cb_release_slot(cb);
1201 dprintk("%s: freed slot, new seqid=%d\n", __func__,
1202 clp->cl_cb_session->se_cb_seq_nr);
1204 if (RPC_SIGNALLED(task))
1209 if (rpc_restart_call_prepare(task))
1213 if (!test_bit(NFSD4_CLIENT_CB_KILL, &clp->cl_flags)) {
1214 task->tk_status = 0;
1215 cb->cb_need_restart = true;
1220 static void nfsd4_cb_done(struct rpc_task *task, void *calldata)
1222 struct nfsd4_callback *cb = calldata;
1223 struct nfs4_client *clp = cb->cb_clp;
1225 if (!nfsd4_cb_sequence_done(task, cb))
1228 if (cb->cb_status) {
1229 WARN_ON_ONCE(task->tk_status);
1230 task->tk_status = cb->cb_status;
1233 switch (cb->cb_ops->done(cb, task)) {
1235 task->tk_status = 0;
1236 rpc_restart_call_prepare(task);
1239 switch (task->tk_status) {
1243 nfsd4_mark_cb_down(clp, task->tk_status);
1251 static void nfsd4_cb_release(void *calldata)
1253 struct nfsd4_callback *cb = calldata;
1255 if (cb->cb_need_restart)
1258 nfsd41_destroy_cb(cb);
1262 static const struct rpc_call_ops nfsd4_cb_ops = {
1263 .rpc_call_prepare = nfsd4_cb_prepare,
1264 .rpc_call_done = nfsd4_cb_done,
1265 .rpc_release = nfsd4_cb_release,
1268 int nfsd4_create_callback_queue(void)
1270 callback_wq = alloc_ordered_workqueue("nfsd4_callbacks", 0);
1276 void nfsd4_destroy_callback_queue(void)
1278 destroy_workqueue(callback_wq);
1281 /* must be called under the state lock */
1282 void nfsd4_shutdown_callback(struct nfs4_client *clp)
1284 if (clp->cl_cb_state != NFSD4_CB_UNKNOWN)
1285 trace_nfsd_cb_shutdown(clp);
1287 set_bit(NFSD4_CLIENT_CB_KILL, &clp->cl_flags);
1289 * Note this won't actually result in a null callback;
1290 * instead, nfsd4_run_cb_null() will detect the killed
1291 * client, destroy the rpc client, and stop:
1293 nfsd4_run_cb(&clp->cl_cb_null);
1294 flush_workqueue(callback_wq);
1295 nfsd41_cb_inflight_wait_complete(clp);
1298 /* requires cl_lock: */
1299 static struct nfsd4_conn * __nfsd4_find_backchannel(struct nfs4_client *clp)
1301 struct nfsd4_session *s;
1302 struct nfsd4_conn *c;
1304 list_for_each_entry(s, &clp->cl_sessions, se_perclnt) {
1305 list_for_each_entry(c, &s->se_conns, cn_persession) {
1306 if (c->cn_flags & NFS4_CDFC4_BACK)
1314 * Note there isn't a lot of locking in this code; instead we depend on
1315 * the fact that it is run from the callback_wq, which won't run two
1316 * work items at once. So, for example, callback_wq handles all access
1317 * of cl_cb_client and all calls to rpc_create or rpc_shutdown_client.
1319 static void nfsd4_process_cb_update(struct nfsd4_callback *cb)
1321 struct nfs4_cb_conn conn;
1322 struct nfs4_client *clp = cb->cb_clp;
1323 struct nfsd4_session *ses = NULL;
1324 struct nfsd4_conn *c;
1328 * This is either an update, or the client dying; in either case,
1329 * kill the old client:
1331 if (clp->cl_cb_client) {
1332 rpc_shutdown_client(clp->cl_cb_client);
1333 clp->cl_cb_client = NULL;
1334 put_cred(clp->cl_cb_cred);
1335 clp->cl_cb_cred = NULL;
1337 if (clp->cl_cb_conn.cb_xprt) {
1338 svc_xprt_put(clp->cl_cb_conn.cb_xprt);
1339 clp->cl_cb_conn.cb_xprt = NULL;
1341 if (test_bit(NFSD4_CLIENT_CB_KILL, &clp->cl_flags))
1343 spin_lock(&clp->cl_lock);
1345 * Only serialized callback code is allowed to clear these
1346 * flags; main nfsd code can only set them:
1348 BUG_ON(!(clp->cl_flags & NFSD4_CLIENT_CB_FLAG_MASK));
1349 clear_bit(NFSD4_CLIENT_CB_UPDATE, &clp->cl_flags);
1350 memcpy(&conn, &cb->cb_clp->cl_cb_conn, sizeof(struct nfs4_cb_conn));
1351 c = __nfsd4_find_backchannel(clp);
1353 svc_xprt_get(c->cn_xprt);
1354 conn.cb_xprt = c->cn_xprt;
1355 ses = c->cn_session;
1357 spin_unlock(&clp->cl_lock);
1359 err = setup_callback_client(clp, &conn, ses);
1361 nfsd4_mark_cb_down(clp, err);
1363 svc_xprt_put(c->cn_xprt);
1369 nfsd4_run_cb_work(struct work_struct *work)
1371 struct nfsd4_callback *cb =
1372 container_of(work, struct nfsd4_callback, cb_work);
1373 struct nfs4_client *clp = cb->cb_clp;
1374 struct rpc_clnt *clnt;
1377 if (cb->cb_need_restart) {
1378 cb->cb_need_restart = false;
1380 if (cb->cb_ops && cb->cb_ops->prepare)
1381 cb->cb_ops->prepare(cb);
1384 if (clp->cl_flags & NFSD4_CLIENT_CB_FLAG_MASK)
1385 nfsd4_process_cb_update(cb);
1387 clnt = clp->cl_cb_client;
1389 /* Callback channel broken, or client killed; give up: */
1390 nfsd41_destroy_cb(cb);
1395 * Don't send probe messages for 4.1 or later.
1397 if (!cb->cb_ops && clp->cl_minorversion) {
1398 nfsd4_mark_cb_state(clp, NFSD4_CB_UP);
1399 nfsd41_destroy_cb(cb);
1403 cb->cb_msg.rpc_cred = clp->cl_cb_cred;
1404 flags = clp->cl_minorversion ? RPC_TASK_NOCONNECT : RPC_TASK_SOFTCONN;
1405 rpc_call_async(clnt, &cb->cb_msg, RPC_TASK_SOFT | flags,
1406 cb->cb_ops ? &nfsd4_cb_ops : &nfsd4_cb_probe_ops, cb);
1409 void nfsd4_init_cb(struct nfsd4_callback *cb, struct nfs4_client *clp,
1410 const struct nfsd4_callback_ops *ops, enum nfsd4_cb_op op)
1413 cb->cb_msg.rpc_proc = &nfs4_cb_procedures[op];
1414 cb->cb_msg.rpc_argp = cb;
1415 cb->cb_msg.rpc_resp = cb;
1417 INIT_WORK(&cb->cb_work, nfsd4_run_cb_work);
1418 cb->cb_seq_status = 1;
1420 cb->cb_need_restart = false;
1421 cb->cb_holds_slot = false;
1425 * nfsd4_run_cb - queue up a callback job to run
1426 * @cb: callback to queue
1428 * Kick off a callback to do its thing. Returns false if it was already
1429 * on a queue, true otherwise.
1431 bool nfsd4_run_cb(struct nfsd4_callback *cb)
1433 struct nfs4_client *clp = cb->cb_clp;
1436 nfsd41_cb_inflight_begin(clp);
1437 queued = nfsd4_queue_cb(cb);
1439 nfsd41_cb_inflight_end(clp);