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/sunrpc/clnt.h>
35 #include <linux/sunrpc/xprt.h>
36 #include <linux/sunrpc/svc_xprt.h>
37 #include <linux/slab.h>
45 #define NFSDDBG_FACILITY NFSDDBG_PROC
47 static void nfsd4_mark_cb_fault(struct nfs4_client *, int reason);
49 #define NFSPROC4_CB_NULL 0
50 #define NFSPROC4_CB_COMPOUND 1
52 /* Index of predefined Linux callback client operations */
54 struct nfs4_cb_compound_hdr {
56 u32 ident; /* minorversion 0 only */
64 static __be32 *xdr_encode_empty_array(__be32 *p)
71 * Encode/decode NFSv4 CB basic data types
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
82 * enum nfs_cb_opnum4 {
90 OP_CB_LAYOUTRECALL = 5,
94 OP_CB_RECALLABLE_OBJ_AVAIL = 9,
95 OP_CB_RECALL_SLOT = 10,
97 OP_CB_WANTS_CANCELLED = 12,
98 OP_CB_NOTIFY_LOCK = 13,
99 OP_CB_NOTIFY_DEVICEID = 14,
101 OP_CB_ILLEGAL = 10044
104 static void encode_nfs_cb_opnum4(struct xdr_stream *xdr, enum nfs_cb_opnum4 op)
108 p = xdr_reserve_space(xdr, 4);
109 *p = cpu_to_be32(op);
115 * typedef opaque nfs_fh4<NFS4_FHSIZE>;
117 static void encode_nfs_fh4(struct xdr_stream *xdr, const struct knfsd_fh *fh)
119 u32 length = fh->fh_size;
122 BUG_ON(length > NFS4_FHSIZE);
123 p = xdr_reserve_space(xdr, 4 + length);
124 xdr_encode_opaque(p, &fh->fh_raw, length);
135 static void encode_stateid4(struct xdr_stream *xdr, const stateid_t *sid)
139 p = xdr_reserve_space(xdr, NFS4_STATEID_SIZE);
140 *p++ = cpu_to_be32(sid->si_generation);
141 xdr_encode_opaque_fixed(p, &sid->si_opaque, NFS4_STATEID_OTHER_SIZE);
147 * typedef opaque sessionid4[NFS4_SESSIONID_SIZE];
149 static void encode_sessionid4(struct xdr_stream *xdr,
150 const struct nfsd4_session *session)
154 p = xdr_reserve_space(xdr, NFS4_MAX_SESSIONID_LEN);
155 xdr_encode_opaque_fixed(p, session->se_sessionid.data,
156 NFS4_MAX_SESSIONID_LEN);
162 static const struct {
165 } nfs_cb_errtbl[] = {
167 { NFS4ERR_PERM, -EPERM },
168 { NFS4ERR_NOENT, -ENOENT },
169 { NFS4ERR_IO, -EIO },
170 { NFS4ERR_NXIO, -ENXIO },
171 { NFS4ERR_ACCESS, -EACCES },
172 { NFS4ERR_EXIST, -EEXIST },
173 { NFS4ERR_XDEV, -EXDEV },
174 { NFS4ERR_NOTDIR, -ENOTDIR },
175 { NFS4ERR_ISDIR, -EISDIR },
176 { NFS4ERR_INVAL, -EINVAL },
177 { NFS4ERR_FBIG, -EFBIG },
178 { NFS4ERR_NOSPC, -ENOSPC },
179 { NFS4ERR_ROFS, -EROFS },
180 { NFS4ERR_MLINK, -EMLINK },
181 { NFS4ERR_NAMETOOLONG, -ENAMETOOLONG },
182 { NFS4ERR_NOTEMPTY, -ENOTEMPTY },
183 { NFS4ERR_DQUOT, -EDQUOT },
184 { NFS4ERR_STALE, -ESTALE },
185 { NFS4ERR_BADHANDLE, -EBADHANDLE },
186 { NFS4ERR_BAD_COOKIE, -EBADCOOKIE },
187 { NFS4ERR_NOTSUPP, -ENOTSUPP },
188 { NFS4ERR_TOOSMALL, -ETOOSMALL },
189 { NFS4ERR_SERVERFAULT, -ESERVERFAULT },
190 { NFS4ERR_BADTYPE, -EBADTYPE },
191 { NFS4ERR_LOCKED, -EAGAIN },
192 { NFS4ERR_RESOURCE, -EREMOTEIO },
193 { NFS4ERR_SYMLINK, -ELOOP },
194 { NFS4ERR_OP_ILLEGAL, -EOPNOTSUPP },
195 { NFS4ERR_DEADLOCK, -EDEADLK },
200 * If we cannot translate the error, the recovery routines should
203 * Note: remaining NFSv4 error codes have values > 10000, so should
204 * not conflict with native Linux error codes.
206 static int nfs_cb_stat_to_errno(int status)
210 for (i = 0; nfs_cb_errtbl[i].stat != -1; i++) {
211 if (nfs_cb_errtbl[i].stat == status)
212 return nfs_cb_errtbl[i].errno;
215 dprintk("NFSD: Unrecognized NFS CB status value: %u\n", status);
219 static int decode_cb_op_status(struct xdr_stream *xdr,
220 enum nfs_cb_opnum4 expected, int *status)
225 p = xdr_inline_decode(xdr, 4 + 4);
226 if (unlikely(p == NULL))
228 op = be32_to_cpup(p++);
229 if (unlikely(op != expected))
231 *status = nfs_cb_stat_to_errno(be32_to_cpup(p));
236 dprintk("NFSD: Callback server returned operation %d but "
237 "we issued a request for %d\n", op, expected);
244 * struct CB_COMPOUND4args {
246 * uint32_t minorversion;
247 * uint32_t callback_ident;
248 * nfs_cb_argop4 argarray<>;
251 static void encode_cb_compound4args(struct xdr_stream *xdr,
252 struct nfs4_cb_compound_hdr *hdr)
256 p = xdr_reserve_space(xdr, 4 + 4 + 4 + 4);
257 p = xdr_encode_empty_array(p); /* empty tag */
258 *p++ = cpu_to_be32(hdr->minorversion);
259 *p++ = cpu_to_be32(hdr->ident);
262 *p = cpu_to_be32(hdr->nops); /* argarray element count */
266 * Update argarray element count
268 static void encode_cb_nops(struct nfs4_cb_compound_hdr *hdr)
270 BUG_ON(hdr->nops > NFS4_MAX_BACK_CHANNEL_OPS);
271 *hdr->nops_p = cpu_to_be32(hdr->nops);
277 * struct CB_COMPOUND4res {
280 * nfs_cb_resop4 resarray<>;
283 static int decode_cb_compound4res(struct xdr_stream *xdr,
284 struct nfs4_cb_compound_hdr *hdr)
289 p = xdr_inline_decode(xdr, 4 + 4);
290 if (unlikely(p == NULL))
292 hdr->status = be32_to_cpup(p++);
294 length = be32_to_cpup(p++);
295 p = xdr_inline_decode(xdr, length + 4);
296 if (unlikely(p == NULL))
298 p += XDR_QUADLEN(length);
299 hdr->nops = be32_to_cpup(p);
308 * struct CB_RECALL4args {
314 static void encode_cb_recall4args(struct xdr_stream *xdr,
315 const struct nfs4_delegation *dp,
316 struct nfs4_cb_compound_hdr *hdr)
320 encode_nfs_cb_opnum4(xdr, OP_CB_RECALL);
321 encode_stateid4(xdr, &dp->dl_stid.sc_stateid);
323 p = xdr_reserve_space(xdr, 4);
324 *p++ = xdr_zero; /* truncate */
326 encode_nfs_fh4(xdr, &dp->dl_stid.sc_file->fi_fhandle);
334 * struct CB_SEQUENCE4args {
335 * sessionid4 csa_sessionid;
336 * sequenceid4 csa_sequenceid;
337 * slotid4 csa_slotid;
338 * slotid4 csa_highest_slotid;
339 * bool csa_cachethis;
340 * referring_call_list4 csa_referring_call_lists<>;
343 static void encode_cb_sequence4args(struct xdr_stream *xdr,
344 const struct nfsd4_callback *cb,
345 struct nfs4_cb_compound_hdr *hdr)
347 struct nfsd4_session *session = cb->cb_clp->cl_cb_session;
350 if (hdr->minorversion == 0)
353 encode_nfs_cb_opnum4(xdr, OP_CB_SEQUENCE);
354 encode_sessionid4(xdr, session);
356 p = xdr_reserve_space(xdr, 4 + 4 + 4 + 4 + 4);
357 *p++ = cpu_to_be32(session->se_cb_seq_nr); /* csa_sequenceid */
358 *p++ = xdr_zero; /* csa_slotid */
359 *p++ = xdr_zero; /* csa_highest_slotid */
360 *p++ = xdr_zero; /* csa_cachethis */
361 xdr_encode_empty_array(p); /* csa_referring_call_lists */
369 * struct CB_SEQUENCE4resok {
370 * sessionid4 csr_sessionid;
371 * sequenceid4 csr_sequenceid;
372 * slotid4 csr_slotid;
373 * slotid4 csr_highest_slotid;
374 * slotid4 csr_target_highest_slotid;
377 * union CB_SEQUENCE4res switch (nfsstat4 csr_status) {
379 * CB_SEQUENCE4resok csr_resok4;
384 * Our current back channel implmentation supports a single backchannel
385 * with a single slot.
387 static int decode_cb_sequence4resok(struct xdr_stream *xdr,
388 struct nfsd4_callback *cb)
390 struct nfsd4_session *session = cb->cb_clp->cl_cb_session;
391 int status = -ESERVERFAULT;
396 * If the server returns different values for sessionID, slotID or
397 * sequence number, the server is looney tunes.
399 p = xdr_inline_decode(xdr, NFS4_MAX_SESSIONID_LEN + 4 + 4 + 4 + 4);
400 if (unlikely(p == NULL))
403 if (memcmp(p, session->se_sessionid.data, NFS4_MAX_SESSIONID_LEN)) {
404 dprintk("NFS: %s Invalid session id\n", __func__);
407 p += XDR_QUADLEN(NFS4_MAX_SESSIONID_LEN);
409 dummy = be32_to_cpup(p++);
410 if (dummy != session->se_cb_seq_nr) {
411 dprintk("NFS: %s Invalid sequence number\n", __func__);
415 dummy = be32_to_cpup(p++);
417 dprintk("NFS: %s Invalid slotid\n", __func__);
422 * FIXME: process highest slotid and target highest slotid
426 cb->cb_seq_status = status;
433 static int decode_cb_sequence4res(struct xdr_stream *xdr,
434 struct nfsd4_callback *cb)
438 if (cb->cb_clp->cl_minorversion == 0)
441 status = decode_cb_op_status(xdr, OP_CB_SEQUENCE, &cb->cb_seq_status);
442 if (unlikely(status || cb->cb_seq_status))
445 return decode_cb_sequence4resok(xdr, cb);
449 * NFSv4.0 and NFSv4.1 XDR encode functions
451 * NFSv4.0 callback argument types are defined in section 15 of RFC
452 * 3530: "Network File System (NFS) version 4 Protocol" and section 20
453 * of RFC 5661: "Network File System (NFS) Version 4 Minor Version 1
458 * NB: Without this zero space reservation, callbacks over krb5p fail
460 static void nfs4_xdr_enc_cb_null(struct rpc_rqst *req, struct xdr_stream *xdr,
461 const void *__unused)
463 xdr_reserve_space(xdr, 0);
467 * 20.2. Operation 4: CB_RECALL - Recall a Delegation
469 static void nfs4_xdr_enc_cb_recall(struct rpc_rqst *req, struct xdr_stream *xdr,
472 const struct nfsd4_callback *cb = data;
473 const struct nfs4_delegation *dp = cb_to_delegation(cb);
474 struct nfs4_cb_compound_hdr hdr = {
475 .ident = cb->cb_clp->cl_cb_ident,
476 .minorversion = cb->cb_clp->cl_minorversion,
479 encode_cb_compound4args(xdr, &hdr);
480 encode_cb_sequence4args(xdr, cb, &hdr);
481 encode_cb_recall4args(xdr, dp, &hdr);
482 encode_cb_nops(&hdr);
487 * NFSv4.0 and NFSv4.1 XDR decode functions
489 * NFSv4.0 callback result types are defined in section 15 of RFC
490 * 3530: "Network File System (NFS) version 4 Protocol" and section 20
491 * of RFC 5661: "Network File System (NFS) Version 4 Minor Version 1
495 static int nfs4_xdr_dec_cb_null(struct rpc_rqst *req, struct xdr_stream *xdr,
502 * 20.2. Operation 4: CB_RECALL - Recall a Delegation
504 static int nfs4_xdr_dec_cb_recall(struct rpc_rqst *rqstp,
505 struct xdr_stream *xdr,
508 struct nfsd4_callback *cb = data;
509 struct nfs4_cb_compound_hdr hdr;
512 status = decode_cb_compound4res(xdr, &hdr);
513 if (unlikely(status))
516 status = decode_cb_sequence4res(xdr, cb);
517 if (unlikely(status || cb->cb_seq_status))
520 return decode_cb_op_status(xdr, OP_CB_RECALL, &cb->cb_status);
523 #ifdef CONFIG_NFSD_PNFS
525 * CB_LAYOUTRECALL4args
527 * struct layoutrecall_file4 {
529 * offset4 lor_offset;
530 * length4 lor_length;
531 * stateid4 lor_stateid;
534 * union layoutrecall4 switch(layoutrecall_type4 lor_recalltype) {
535 * case LAYOUTRECALL4_FILE:
536 * layoutrecall_file4 lor_layout;
537 * case LAYOUTRECALL4_FSID:
539 * case LAYOUTRECALL4_ALL:
543 * struct CB_LAYOUTRECALL4args {
544 * layouttype4 clora_type;
545 * layoutiomode4 clora_iomode;
546 * bool clora_changed;
547 * layoutrecall4 clora_recall;
550 static void encode_cb_layout4args(struct xdr_stream *xdr,
551 const struct nfs4_layout_stateid *ls,
552 struct nfs4_cb_compound_hdr *hdr)
556 BUG_ON(hdr->minorversion == 0);
558 p = xdr_reserve_space(xdr, 5 * 4);
559 *p++ = cpu_to_be32(OP_CB_LAYOUTRECALL);
560 *p++ = cpu_to_be32(ls->ls_layout_type);
561 *p++ = cpu_to_be32(IOMODE_ANY);
562 *p++ = cpu_to_be32(1);
563 *p = cpu_to_be32(RETURN_FILE);
565 encode_nfs_fh4(xdr, &ls->ls_stid.sc_file->fi_fhandle);
567 p = xdr_reserve_space(xdr, 2 * 8);
568 p = xdr_encode_hyper(p, 0);
569 xdr_encode_hyper(p, NFS4_MAX_UINT64);
571 encode_stateid4(xdr, &ls->ls_recall_sid);
576 static void nfs4_xdr_enc_cb_layout(struct rpc_rqst *req,
577 struct xdr_stream *xdr,
580 const struct nfsd4_callback *cb = data;
581 const struct nfs4_layout_stateid *ls =
582 container_of(cb, struct nfs4_layout_stateid, ls_recall);
583 struct nfs4_cb_compound_hdr hdr = {
585 .minorversion = cb->cb_clp->cl_minorversion,
588 encode_cb_compound4args(xdr, &hdr);
589 encode_cb_sequence4args(xdr, cb, &hdr);
590 encode_cb_layout4args(xdr, ls, &hdr);
591 encode_cb_nops(&hdr);
594 static int nfs4_xdr_dec_cb_layout(struct rpc_rqst *rqstp,
595 struct xdr_stream *xdr,
598 struct nfsd4_callback *cb = data;
599 struct nfs4_cb_compound_hdr hdr;
602 status = decode_cb_compound4res(xdr, &hdr);
603 if (unlikely(status))
606 status = decode_cb_sequence4res(xdr, cb);
607 if (unlikely(status || cb->cb_seq_status))
610 return decode_cb_op_status(xdr, OP_CB_LAYOUTRECALL, &cb->cb_status);
612 #endif /* CONFIG_NFSD_PNFS */
614 static void encode_stateowner(struct xdr_stream *xdr, struct nfs4_stateowner *so)
618 p = xdr_reserve_space(xdr, 8 + 4 + so->so_owner.len);
619 p = xdr_encode_opaque_fixed(p, &so->so_client->cl_clientid, 8);
620 xdr_encode_opaque(p, so->so_owner.data, so->so_owner.len);
623 static void nfs4_xdr_enc_cb_notify_lock(struct rpc_rqst *req,
624 struct xdr_stream *xdr,
627 const struct nfsd4_callback *cb = data;
628 const struct nfsd4_blocked_lock *nbl =
629 container_of(cb, struct nfsd4_blocked_lock, nbl_cb);
630 struct nfs4_lockowner *lo = (struct nfs4_lockowner *)nbl->nbl_lock.fl_owner;
631 struct nfs4_cb_compound_hdr hdr = {
633 .minorversion = cb->cb_clp->cl_minorversion,
638 BUG_ON(hdr.minorversion == 0);
640 encode_cb_compound4args(xdr, &hdr);
641 encode_cb_sequence4args(xdr, cb, &hdr);
643 p = xdr_reserve_space(xdr, 4);
644 *p = cpu_to_be32(OP_CB_NOTIFY_LOCK);
645 encode_nfs_fh4(xdr, &nbl->nbl_fh);
646 encode_stateowner(xdr, &lo->lo_owner);
649 encode_cb_nops(&hdr);
652 static int nfs4_xdr_dec_cb_notify_lock(struct rpc_rqst *rqstp,
653 struct xdr_stream *xdr,
656 struct nfsd4_callback *cb = data;
657 struct nfs4_cb_compound_hdr hdr;
660 status = decode_cb_compound4res(xdr, &hdr);
661 if (unlikely(status))
664 status = decode_cb_sequence4res(xdr, cb);
665 if (unlikely(status || cb->cb_seq_status))
668 return decode_cb_op_status(xdr, OP_CB_NOTIFY_LOCK, &cb->cb_status);
672 * struct write_response4 {
673 * stateid4 wr_callback_id<1>;
675 * stable_how4 wr_committed;
676 * verifier4 wr_writeverf;
678 * union offload_info4 switch (nfsstat4 coa_status) {
680 * write_response4 coa_resok4;
682 * length4 coa_bytes_copied;
684 * struct CB_OFFLOAD4args {
686 * stateid4 coa_stateid;
687 * offload_info4 coa_offload_info;
690 static void encode_offload_info4(struct xdr_stream *xdr,
691 const struct nfsd4_cb_offload *cbo)
695 p = xdr_reserve_space(xdr, 4);
697 switch (cbo->co_nfserr) {
699 p = xdr_reserve_space(xdr, 4 + 8 + 4 + NFS4_VERIFIER_SIZE);
700 p = xdr_encode_empty_array(p);
701 p = xdr_encode_hyper(p, cbo->co_res.wr_bytes_written);
702 *p++ = cpu_to_be32(cbo->co_res.wr_stable_how);
703 p = xdr_encode_opaque_fixed(p, cbo->co_res.wr_verifier.data,
707 p = xdr_reserve_space(xdr, 8);
708 /* We always return success if bytes were written */
709 p = xdr_encode_hyper(p, 0);
713 static void encode_cb_offload4args(struct xdr_stream *xdr,
714 const struct nfsd4_cb_offload *cbo,
715 struct nfs4_cb_compound_hdr *hdr)
719 p = xdr_reserve_space(xdr, 4);
720 *p = cpu_to_be32(OP_CB_OFFLOAD);
721 encode_nfs_fh4(xdr, &cbo->co_fh);
722 encode_stateid4(xdr, &cbo->co_res.cb_stateid);
723 encode_offload_info4(xdr, cbo);
728 static void nfs4_xdr_enc_cb_offload(struct rpc_rqst *req,
729 struct xdr_stream *xdr,
732 const struct nfsd4_callback *cb = data;
733 const struct nfsd4_cb_offload *cbo =
734 container_of(cb, struct nfsd4_cb_offload, co_cb);
735 struct nfs4_cb_compound_hdr hdr = {
737 .minorversion = cb->cb_clp->cl_minorversion,
740 encode_cb_compound4args(xdr, &hdr);
741 encode_cb_sequence4args(xdr, cb, &hdr);
742 encode_cb_offload4args(xdr, cbo, &hdr);
743 encode_cb_nops(&hdr);
746 static int nfs4_xdr_dec_cb_offload(struct rpc_rqst *rqstp,
747 struct xdr_stream *xdr,
750 struct nfsd4_callback *cb = data;
751 struct nfs4_cb_compound_hdr hdr;
754 status = decode_cb_compound4res(xdr, &hdr);
755 if (unlikely(status))
758 status = decode_cb_sequence4res(xdr, cb);
759 if (unlikely(status || cb->cb_seq_status))
762 return decode_cb_op_status(xdr, OP_CB_OFFLOAD, &cb->cb_status);
765 * RPC procedure tables
767 #define PROC(proc, call, argtype, restype) \
768 [NFSPROC4_CLNT_##proc] = { \
769 .p_proc = NFSPROC4_CB_##call, \
770 .p_encode = nfs4_xdr_enc_##argtype, \
771 .p_decode = nfs4_xdr_dec_##restype, \
772 .p_arglen = NFS4_enc_##argtype##_sz, \
773 .p_replen = NFS4_dec_##restype##_sz, \
774 .p_statidx = NFSPROC4_CB_##call, \
778 static const struct rpc_procinfo nfs4_cb_procedures[] = {
779 PROC(CB_NULL, NULL, cb_null, cb_null),
780 PROC(CB_RECALL, COMPOUND, cb_recall, cb_recall),
781 #ifdef CONFIG_NFSD_PNFS
782 PROC(CB_LAYOUT, COMPOUND, cb_layout, cb_layout),
784 PROC(CB_NOTIFY_LOCK, COMPOUND, cb_notify_lock, cb_notify_lock),
785 PROC(CB_OFFLOAD, COMPOUND, cb_offload, cb_offload),
788 static unsigned int nfs4_cb_counts[ARRAY_SIZE(nfs4_cb_procedures)];
789 static const struct rpc_version nfs_cb_version4 = {
791 * Note on the callback rpc program version number: despite language in rfc
792 * 5661 section 18.36.3 requiring servers to use 4 in this field, the
793 * official xdr descriptions for both 4.0 and 4.1 specify version 1, and
794 * in practice that appears to be what implementations use. The section
795 * 18.36.3 language is expected to be fixed in an erratum.
798 .nrprocs = ARRAY_SIZE(nfs4_cb_procedures),
799 .procs = nfs4_cb_procedures,
800 .counts = nfs4_cb_counts,
803 static const struct rpc_version *nfs_cb_version[2] = {
804 [1] = &nfs_cb_version4,
807 static const struct rpc_program cb_program;
809 static struct rpc_stat cb_stats = {
810 .program = &cb_program
813 #define NFS4_CALLBACK 0x40000000
814 static const struct rpc_program cb_program = {
816 .number = NFS4_CALLBACK,
817 .nrvers = ARRAY_SIZE(nfs_cb_version),
818 .version = nfs_cb_version,
820 .pipe_dir_name = "nfsd4_cb",
823 static int max_cb_time(struct net *net)
825 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
828 * nfsd4_lease is set to at most one hour in __nfsd4_write_time,
829 * so we can use 32-bit math on it. Warn if that assumption
830 * ever stops being true.
832 if (WARN_ON_ONCE(nn->nfsd4_lease > 3600))
835 return max(((u32)nn->nfsd4_lease)/10, 1u) * HZ;
838 static struct workqueue_struct *callback_wq;
840 static bool nfsd4_queue_cb(struct nfsd4_callback *cb)
842 return queue_work(callback_wq, &cb->cb_work);
845 static void nfsd41_cb_inflight_begin(struct nfs4_client *clp)
847 atomic_inc(&clp->cl_cb_inflight);
850 static void nfsd41_cb_inflight_end(struct nfs4_client *clp)
853 if (atomic_dec_and_test(&clp->cl_cb_inflight))
854 wake_up_var(&clp->cl_cb_inflight);
857 static void nfsd41_cb_inflight_wait_complete(struct nfs4_client *clp)
859 wait_var_event(&clp->cl_cb_inflight,
860 !atomic_read(&clp->cl_cb_inflight));
863 static const struct cred *get_backchannel_cred(struct nfs4_client *clp, struct rpc_clnt *client, struct nfsd4_session *ses)
865 if (clp->cl_minorversion == 0) {
866 client->cl_principal = clp->cl_cred.cr_targ_princ ?
867 clp->cl_cred.cr_targ_princ : "nfs";
869 return get_cred(rpc_machine_cred());
873 kcred = prepare_kernel_cred(&init_task);
877 kcred->uid = ses->se_cb_sec.uid;
878 kcred->gid = ses->se_cb_sec.gid;
883 static int setup_callback_client(struct nfs4_client *clp, struct nfs4_cb_conn *conn, struct nfsd4_session *ses)
885 int maxtime = max_cb_time(clp->net);
886 struct rpc_timeout timeparms = {
887 .to_initval = maxtime,
889 .to_maxval = maxtime,
891 struct rpc_create_args args = {
893 .address = (struct sockaddr *) &conn->cb_addr,
894 .addrsize = conn->cb_addrlen,
895 .saddress = (struct sockaddr *) &conn->cb_saddr,
896 .timeout = &timeparms,
897 .program = &cb_program,
899 .flags = (RPC_CLNT_CREATE_NOPING | RPC_CLNT_CREATE_QUIET),
900 .cred = current_cred(),
902 struct rpc_clnt *client;
903 const struct cred *cred;
905 if (clp->cl_minorversion == 0) {
906 if (!clp->cl_cred.cr_principal &&
907 (clp->cl_cred.cr_flavor >= RPC_AUTH_GSS_KRB5)) {
908 trace_nfsd_cb_setup_err(clp, -EINVAL);
911 args.client_name = clp->cl_cred.cr_principal;
912 args.prognumber = conn->cb_prog;
913 args.protocol = XPRT_TRANSPORT_TCP;
914 args.authflavor = clp->cl_cred.cr_flavor;
915 clp->cl_cb_ident = conn->cb_ident;
919 clp->cl_cb_conn.cb_xprt = conn->cb_xprt;
920 clp->cl_cb_session = ses;
921 args.bc_xprt = conn->cb_xprt;
922 args.prognumber = clp->cl_cb_session->se_cb_prog;
923 args.protocol = conn->cb_xprt->xpt_class->xcl_ident |
925 args.authflavor = ses->se_cb_sec.flavor;
927 /* Create RPC client */
928 client = rpc_create(&args);
929 if (IS_ERR(client)) {
930 trace_nfsd_cb_setup_err(clp, PTR_ERR(client));
931 return PTR_ERR(client);
933 cred = get_backchannel_cred(clp, client, ses);
935 trace_nfsd_cb_setup_err(clp, -ENOMEM);
936 rpc_shutdown_client(client);
939 clp->cl_cb_client = client;
940 clp->cl_cb_cred = cred;
942 trace_nfsd_cb_setup(clp, rpc_peeraddr2str(client, RPC_DISPLAY_NETID),
948 static void nfsd4_mark_cb_state(struct nfs4_client *clp, int newstate)
950 if (clp->cl_cb_state != newstate) {
951 clp->cl_cb_state = newstate;
952 trace_nfsd_cb_state(clp);
956 static void nfsd4_mark_cb_down(struct nfs4_client *clp, int reason)
958 if (test_bit(NFSD4_CLIENT_CB_UPDATE, &clp->cl_flags))
960 nfsd4_mark_cb_state(clp, NFSD4_CB_DOWN);
963 static void nfsd4_mark_cb_fault(struct nfs4_client *clp, int reason)
965 if (test_bit(NFSD4_CLIENT_CB_UPDATE, &clp->cl_flags))
967 nfsd4_mark_cb_state(clp, NFSD4_CB_FAULT);
970 static void nfsd4_cb_probe_done(struct rpc_task *task, void *calldata)
972 struct nfs4_client *clp = container_of(calldata, struct nfs4_client, cl_cb_null);
975 nfsd4_mark_cb_down(clp, task->tk_status);
977 nfsd4_mark_cb_state(clp, NFSD4_CB_UP);
980 static void nfsd4_cb_probe_release(void *calldata)
982 struct nfs4_client *clp = container_of(calldata, struct nfs4_client, cl_cb_null);
984 nfsd41_cb_inflight_end(clp);
988 static const struct rpc_call_ops nfsd4_cb_probe_ops = {
989 /* XXX: release method to ensure we set the cb channel down if
990 * necessary on early failure? */
991 .rpc_call_done = nfsd4_cb_probe_done,
992 .rpc_release = nfsd4_cb_probe_release,
996 * Poke the callback thread to process any updates to the callback
997 * parameters, and send a null probe.
999 void nfsd4_probe_callback(struct nfs4_client *clp)
1001 trace_nfsd_cb_probe(clp);
1002 nfsd4_mark_cb_state(clp, NFSD4_CB_UNKNOWN);
1003 set_bit(NFSD4_CLIENT_CB_UPDATE, &clp->cl_flags);
1004 nfsd4_run_cb(&clp->cl_cb_null);
1007 void nfsd4_probe_callback_sync(struct nfs4_client *clp)
1009 nfsd4_probe_callback(clp);
1010 flush_workqueue(callback_wq);
1013 void nfsd4_change_callback(struct nfs4_client *clp, struct nfs4_cb_conn *conn)
1015 nfsd4_mark_cb_state(clp, NFSD4_CB_UNKNOWN);
1016 spin_lock(&clp->cl_lock);
1017 memcpy(&clp->cl_cb_conn, conn, sizeof(struct nfs4_cb_conn));
1018 spin_unlock(&clp->cl_lock);
1022 * There's currently a single callback channel slot.
1023 * If the slot is available, then mark it busy. Otherwise, set the
1024 * thread for sleeping on the callback RPC wait queue.
1026 static bool nfsd41_cb_get_slot(struct nfsd4_callback *cb, struct rpc_task *task)
1028 struct nfs4_client *clp = cb->cb_clp;
1030 if (!cb->cb_holds_slot &&
1031 test_and_set_bit(0, &clp->cl_cb_slot_busy) != 0) {
1032 rpc_sleep_on(&clp->cl_cb_waitq, task, NULL);
1034 if (test_and_set_bit(0, &clp->cl_cb_slot_busy) != 0) {
1035 dprintk("%s slot is busy\n", __func__);
1038 rpc_wake_up_queued_task(&clp->cl_cb_waitq, task);
1040 cb->cb_holds_slot = true;
1044 static void nfsd41_cb_release_slot(struct nfsd4_callback *cb)
1046 struct nfs4_client *clp = cb->cb_clp;
1048 if (cb->cb_holds_slot) {
1049 cb->cb_holds_slot = false;
1050 clear_bit(0, &clp->cl_cb_slot_busy);
1051 rpc_wake_up_next(&clp->cl_cb_waitq);
1055 static void nfsd41_destroy_cb(struct nfsd4_callback *cb)
1057 struct nfs4_client *clp = cb->cb_clp;
1059 nfsd41_cb_release_slot(cb);
1060 if (cb->cb_ops && cb->cb_ops->release)
1061 cb->cb_ops->release(cb);
1062 nfsd41_cb_inflight_end(clp);
1066 * TODO: cb_sequence should support referring call lists, cachethis, multiple
1067 * slots, and mark callback channel down on communication errors.
1069 static void nfsd4_cb_prepare(struct rpc_task *task, void *calldata)
1071 struct nfsd4_callback *cb = calldata;
1072 struct nfs4_client *clp = cb->cb_clp;
1073 u32 minorversion = clp->cl_minorversion;
1076 * cb_seq_status is only set in decode_cb_sequence4res,
1077 * and so will remain 1 if an rpc level failure occurs.
1079 cb->cb_seq_status = 1;
1081 if (minorversion && !nfsd41_cb_get_slot(cb, task))
1083 rpc_call_start(task);
1086 static bool nfsd4_cb_sequence_done(struct rpc_task *task, struct nfsd4_callback *cb)
1088 struct nfs4_client *clp = cb->cb_clp;
1089 struct nfsd4_session *session = clp->cl_cb_session;
1092 if (!clp->cl_minorversion) {
1094 * If the backchannel connection was shut down while this
1095 * task was queued, we need to resubmit it after setting up
1096 * a new backchannel connection.
1098 * Note that if we lost our callback connection permanently
1099 * the submission code will error out, so we don't need to
1100 * handle that case here.
1102 if (RPC_SIGNALLED(task))
1108 if (!cb->cb_holds_slot)
1111 switch (cb->cb_seq_status) {
1114 * No need for lock, access serialized in nfsd4_cb_prepare
1117 * If CB_SEQUENCE returns an error, then the state of the slot
1118 * (sequence ID, cached reply) MUST NOT change.
1120 ++session->se_cb_seq_nr;
1123 ++session->se_cb_seq_nr;
1126 case -NFS4ERR_BADSESSION:
1127 nfsd4_mark_cb_fault(cb->cb_clp, cb->cb_seq_status);
1130 case -NFS4ERR_DELAY:
1131 if (!rpc_restart_call(task))
1134 rpc_delay(task, 2 * HZ);
1136 case -NFS4ERR_BADSLOT:
1138 case -NFS4ERR_SEQ_MISORDERED:
1139 if (session->se_cb_seq_nr != 1) {
1140 session->se_cb_seq_nr = 1;
1145 nfsd4_mark_cb_fault(cb->cb_clp, cb->cb_seq_status);
1146 dprintk("%s: unprocessed error %d\n", __func__,
1150 nfsd41_cb_release_slot(cb);
1151 dprintk("%s: freed slot, new seqid=%d\n", __func__,
1152 clp->cl_cb_session->se_cb_seq_nr);
1154 if (RPC_SIGNALLED(task))
1159 if (rpc_restart_call_prepare(task))
1163 if (!test_bit(NFSD4_CLIENT_CB_KILL, &clp->cl_flags)) {
1164 task->tk_status = 0;
1165 cb->cb_need_restart = true;
1170 static void nfsd4_cb_done(struct rpc_task *task, void *calldata)
1172 struct nfsd4_callback *cb = calldata;
1173 struct nfs4_client *clp = cb->cb_clp;
1175 if (!nfsd4_cb_sequence_done(task, cb))
1178 if (cb->cb_status) {
1179 WARN_ON_ONCE(task->tk_status);
1180 task->tk_status = cb->cb_status;
1183 switch (cb->cb_ops->done(cb, task)) {
1185 task->tk_status = 0;
1186 rpc_restart_call_prepare(task);
1189 switch (task->tk_status) {
1193 nfsd4_mark_cb_down(clp, task->tk_status);
1201 static void nfsd4_cb_release(void *calldata)
1203 struct nfsd4_callback *cb = calldata;
1205 if (cb->cb_need_restart)
1208 nfsd41_destroy_cb(cb);
1212 static const struct rpc_call_ops nfsd4_cb_ops = {
1213 .rpc_call_prepare = nfsd4_cb_prepare,
1214 .rpc_call_done = nfsd4_cb_done,
1215 .rpc_release = nfsd4_cb_release,
1218 int nfsd4_create_callback_queue(void)
1220 callback_wq = alloc_ordered_workqueue("nfsd4_callbacks", 0);
1226 void nfsd4_destroy_callback_queue(void)
1228 destroy_workqueue(callback_wq);
1231 /* must be called under the state lock */
1232 void nfsd4_shutdown_callback(struct nfs4_client *clp)
1234 if (clp->cl_cb_state != NFSD4_CB_UNKNOWN)
1235 trace_nfsd_cb_shutdown(clp);
1237 set_bit(NFSD4_CLIENT_CB_KILL, &clp->cl_flags);
1239 * Note this won't actually result in a null callback;
1240 * instead, nfsd4_run_cb_null() will detect the killed
1241 * client, destroy the rpc client, and stop:
1243 nfsd4_run_cb(&clp->cl_cb_null);
1244 flush_workqueue(callback_wq);
1245 nfsd41_cb_inflight_wait_complete(clp);
1248 /* requires cl_lock: */
1249 static struct nfsd4_conn * __nfsd4_find_backchannel(struct nfs4_client *clp)
1251 struct nfsd4_session *s;
1252 struct nfsd4_conn *c;
1254 list_for_each_entry(s, &clp->cl_sessions, se_perclnt) {
1255 list_for_each_entry(c, &s->se_conns, cn_persession) {
1256 if (c->cn_flags & NFS4_CDFC4_BACK)
1264 * Note there isn't a lot of locking in this code; instead we depend on
1265 * the fact that it is run from the callback_wq, which won't run two
1266 * work items at once. So, for example, callback_wq handles all access
1267 * of cl_cb_client and all calls to rpc_create or rpc_shutdown_client.
1269 static void nfsd4_process_cb_update(struct nfsd4_callback *cb)
1271 struct nfs4_cb_conn conn;
1272 struct nfs4_client *clp = cb->cb_clp;
1273 struct nfsd4_session *ses = NULL;
1274 struct nfsd4_conn *c;
1278 * This is either an update, or the client dying; in either case,
1279 * kill the old client:
1281 if (clp->cl_cb_client) {
1282 rpc_shutdown_client(clp->cl_cb_client);
1283 clp->cl_cb_client = NULL;
1284 put_cred(clp->cl_cb_cred);
1285 clp->cl_cb_cred = NULL;
1287 if (clp->cl_cb_conn.cb_xprt) {
1288 svc_xprt_put(clp->cl_cb_conn.cb_xprt);
1289 clp->cl_cb_conn.cb_xprt = NULL;
1291 if (test_bit(NFSD4_CLIENT_CB_KILL, &clp->cl_flags))
1293 spin_lock(&clp->cl_lock);
1295 * Only serialized callback code is allowed to clear these
1296 * flags; main nfsd code can only set them:
1298 BUG_ON(!(clp->cl_flags & NFSD4_CLIENT_CB_FLAG_MASK));
1299 clear_bit(NFSD4_CLIENT_CB_UPDATE, &clp->cl_flags);
1300 memcpy(&conn, &cb->cb_clp->cl_cb_conn, sizeof(struct nfs4_cb_conn));
1301 c = __nfsd4_find_backchannel(clp);
1303 svc_xprt_get(c->cn_xprt);
1304 conn.cb_xprt = c->cn_xprt;
1305 ses = c->cn_session;
1307 spin_unlock(&clp->cl_lock);
1309 err = setup_callback_client(clp, &conn, ses);
1311 nfsd4_mark_cb_down(clp, err);
1313 svc_xprt_put(c->cn_xprt);
1319 nfsd4_run_cb_work(struct work_struct *work)
1321 struct nfsd4_callback *cb =
1322 container_of(work, struct nfsd4_callback, cb_work);
1323 struct nfs4_client *clp = cb->cb_clp;
1324 struct rpc_clnt *clnt;
1327 if (cb->cb_need_restart) {
1328 cb->cb_need_restart = false;
1330 if (cb->cb_ops && cb->cb_ops->prepare)
1331 cb->cb_ops->prepare(cb);
1334 if (clp->cl_flags & NFSD4_CLIENT_CB_FLAG_MASK)
1335 nfsd4_process_cb_update(cb);
1337 clnt = clp->cl_cb_client;
1339 /* Callback channel broken, or client killed; give up: */
1340 nfsd41_destroy_cb(cb);
1345 * Don't send probe messages for 4.1 or later.
1347 if (!cb->cb_ops && clp->cl_minorversion) {
1348 nfsd4_mark_cb_state(clp, NFSD4_CB_UP);
1349 nfsd41_destroy_cb(cb);
1353 cb->cb_msg.rpc_cred = clp->cl_cb_cred;
1354 flags = clp->cl_minorversion ? RPC_TASK_NOCONNECT : RPC_TASK_SOFTCONN;
1355 rpc_call_async(clnt, &cb->cb_msg, RPC_TASK_SOFT | flags,
1356 cb->cb_ops ? &nfsd4_cb_ops : &nfsd4_cb_probe_ops, cb);
1359 void nfsd4_init_cb(struct nfsd4_callback *cb, struct nfs4_client *clp,
1360 const struct nfsd4_callback_ops *ops, enum nfsd4_cb_op op)
1363 cb->cb_msg.rpc_proc = &nfs4_cb_procedures[op];
1364 cb->cb_msg.rpc_argp = cb;
1365 cb->cb_msg.rpc_resp = cb;
1367 INIT_WORK(&cb->cb_work, nfsd4_run_cb_work);
1368 cb->cb_seq_status = 1;
1370 cb->cb_need_restart = false;
1371 cb->cb_holds_slot = false;
1375 * nfsd4_run_cb - queue up a callback job to run
1376 * @cb: callback to queue
1378 * Kick off a callback to do its thing. Returns false if it was already
1379 * on a queue, true otherwise.
1381 bool nfsd4_run_cb(struct nfsd4_callback *cb)
1383 struct nfs4_client *clp = cb->cb_clp;
1386 nfsd41_cb_inflight_begin(clp);
1387 queued = nfsd4_queue_cb(cb);
1389 nfsd41_cb_inflight_end(clp);