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
79 static void encode_uint32(struct xdr_stream *xdr, u32 n)
81 WARN_ON_ONCE(xdr_stream_encode_u32(xdr, n) < 0);
84 static void encode_bitmap4(struct xdr_stream *xdr, const __u32 *bitmap,
87 WARN_ON_ONCE(xdr_stream_encode_uint32_array(xdr, bitmap, len) < 0);
93 * enum nfs_cb_opnum4 {
101 OP_CB_LAYOUTRECALL = 5,
103 OP_CB_PUSH_DELEG = 7,
104 OP_CB_RECALL_ANY = 8,
105 OP_CB_RECALLABLE_OBJ_AVAIL = 9,
106 OP_CB_RECALL_SLOT = 10,
108 OP_CB_WANTS_CANCELLED = 12,
109 OP_CB_NOTIFY_LOCK = 13,
110 OP_CB_NOTIFY_DEVICEID = 14,
112 OP_CB_ILLEGAL = 10044
115 static void encode_nfs_cb_opnum4(struct xdr_stream *xdr, enum nfs_cb_opnum4 op)
119 p = xdr_reserve_space(xdr, 4);
120 *p = cpu_to_be32(op);
126 * typedef opaque nfs_fh4<NFS4_FHSIZE>;
128 static void encode_nfs_fh4(struct xdr_stream *xdr, const struct knfsd_fh *fh)
130 u32 length = fh->fh_size;
133 BUG_ON(length > NFS4_FHSIZE);
134 p = xdr_reserve_space(xdr, 4 + length);
135 xdr_encode_opaque(p, &fh->fh_raw, length);
146 static void encode_stateid4(struct xdr_stream *xdr, const stateid_t *sid)
150 p = xdr_reserve_space(xdr, NFS4_STATEID_SIZE);
151 *p++ = cpu_to_be32(sid->si_generation);
152 xdr_encode_opaque_fixed(p, &sid->si_opaque, NFS4_STATEID_OTHER_SIZE);
158 * typedef opaque sessionid4[NFS4_SESSIONID_SIZE];
160 static void encode_sessionid4(struct xdr_stream *xdr,
161 const struct nfsd4_session *session)
165 p = xdr_reserve_space(xdr, NFS4_MAX_SESSIONID_LEN);
166 xdr_encode_opaque_fixed(p, session->se_sessionid.data,
167 NFS4_MAX_SESSIONID_LEN);
173 static const struct {
176 } nfs_cb_errtbl[] = {
178 { NFS4ERR_PERM, -EPERM },
179 { NFS4ERR_NOENT, -ENOENT },
180 { NFS4ERR_IO, -EIO },
181 { NFS4ERR_NXIO, -ENXIO },
182 { NFS4ERR_ACCESS, -EACCES },
183 { NFS4ERR_EXIST, -EEXIST },
184 { NFS4ERR_XDEV, -EXDEV },
185 { NFS4ERR_NOTDIR, -ENOTDIR },
186 { NFS4ERR_ISDIR, -EISDIR },
187 { NFS4ERR_INVAL, -EINVAL },
188 { NFS4ERR_FBIG, -EFBIG },
189 { NFS4ERR_NOSPC, -ENOSPC },
190 { NFS4ERR_ROFS, -EROFS },
191 { NFS4ERR_MLINK, -EMLINK },
192 { NFS4ERR_NAMETOOLONG, -ENAMETOOLONG },
193 { NFS4ERR_NOTEMPTY, -ENOTEMPTY },
194 { NFS4ERR_DQUOT, -EDQUOT },
195 { NFS4ERR_STALE, -ESTALE },
196 { NFS4ERR_BADHANDLE, -EBADHANDLE },
197 { NFS4ERR_BAD_COOKIE, -EBADCOOKIE },
198 { NFS4ERR_NOTSUPP, -ENOTSUPP },
199 { NFS4ERR_TOOSMALL, -ETOOSMALL },
200 { NFS4ERR_SERVERFAULT, -ESERVERFAULT },
201 { NFS4ERR_BADTYPE, -EBADTYPE },
202 { NFS4ERR_LOCKED, -EAGAIN },
203 { NFS4ERR_RESOURCE, -EREMOTEIO },
204 { NFS4ERR_SYMLINK, -ELOOP },
205 { NFS4ERR_OP_ILLEGAL, -EOPNOTSUPP },
206 { NFS4ERR_DEADLOCK, -EDEADLK },
211 * If we cannot translate the error, the recovery routines should
214 * Note: remaining NFSv4 error codes have values > 10000, so should
215 * not conflict with native Linux error codes.
217 static int nfs_cb_stat_to_errno(int status)
221 for (i = 0; nfs_cb_errtbl[i].stat != -1; i++) {
222 if (nfs_cb_errtbl[i].stat == status)
223 return nfs_cb_errtbl[i].errno;
226 dprintk("NFSD: Unrecognized NFS CB status value: %u\n", status);
230 static int decode_cb_op_status(struct xdr_stream *xdr,
231 enum nfs_cb_opnum4 expected, int *status)
236 p = xdr_inline_decode(xdr, 4 + 4);
237 if (unlikely(p == NULL))
239 op = be32_to_cpup(p++);
240 if (unlikely(op != expected))
242 *status = nfs_cb_stat_to_errno(be32_to_cpup(p));
247 dprintk("NFSD: Callback server returned operation %d but "
248 "we issued a request for %d\n", op, expected);
255 * struct CB_COMPOUND4args {
257 * uint32_t minorversion;
258 * uint32_t callback_ident;
259 * nfs_cb_argop4 argarray<>;
262 static void encode_cb_compound4args(struct xdr_stream *xdr,
263 struct nfs4_cb_compound_hdr *hdr)
267 p = xdr_reserve_space(xdr, 4 + 4 + 4 + 4);
268 p = xdr_encode_empty_array(p); /* empty tag */
269 *p++ = cpu_to_be32(hdr->minorversion);
270 *p++ = cpu_to_be32(hdr->ident);
273 *p = cpu_to_be32(hdr->nops); /* argarray element count */
277 * Update argarray element count
279 static void encode_cb_nops(struct nfs4_cb_compound_hdr *hdr)
281 BUG_ON(hdr->nops > NFS4_MAX_BACK_CHANNEL_OPS);
282 *hdr->nops_p = cpu_to_be32(hdr->nops);
288 * struct CB_COMPOUND4res {
291 * nfs_cb_resop4 resarray<>;
294 static int decode_cb_compound4res(struct xdr_stream *xdr,
295 struct nfs4_cb_compound_hdr *hdr)
300 p = xdr_inline_decode(xdr, 4 + 4);
301 if (unlikely(p == NULL))
303 hdr->status = be32_to_cpup(p++);
305 length = be32_to_cpup(p++);
306 p = xdr_inline_decode(xdr, length + 4);
307 if (unlikely(p == NULL))
309 p += XDR_QUADLEN(length);
310 hdr->nops = be32_to_cpup(p);
319 * struct CB_RECALL4args {
325 static void encode_cb_recall4args(struct xdr_stream *xdr,
326 const struct nfs4_delegation *dp,
327 struct nfs4_cb_compound_hdr *hdr)
331 encode_nfs_cb_opnum4(xdr, OP_CB_RECALL);
332 encode_stateid4(xdr, &dp->dl_stid.sc_stateid);
334 p = xdr_reserve_space(xdr, 4);
335 *p++ = xdr_zero; /* truncate */
337 encode_nfs_fh4(xdr, &dp->dl_stid.sc_file->fi_fhandle);
345 * struct CB_RECALLANY4args {
346 * uint32_t craa_objects_to_keep;
347 * bitmap4 craa_type_mask;
351 encode_cb_recallany4args(struct xdr_stream *xdr,
352 struct nfs4_cb_compound_hdr *hdr, struct nfsd4_cb_recall_any *ra)
354 encode_nfs_cb_opnum4(xdr, OP_CB_RECALL_ANY);
355 encode_uint32(xdr, ra->ra_keep);
356 encode_bitmap4(xdr, ra->ra_bmval, ARRAY_SIZE(ra->ra_bmval));
363 * struct CB_SEQUENCE4args {
364 * sessionid4 csa_sessionid;
365 * sequenceid4 csa_sequenceid;
366 * slotid4 csa_slotid;
367 * slotid4 csa_highest_slotid;
368 * bool csa_cachethis;
369 * referring_call_list4 csa_referring_call_lists<>;
372 static void encode_cb_sequence4args(struct xdr_stream *xdr,
373 const struct nfsd4_callback *cb,
374 struct nfs4_cb_compound_hdr *hdr)
376 struct nfsd4_session *session = cb->cb_clp->cl_cb_session;
379 if (hdr->minorversion == 0)
382 encode_nfs_cb_opnum4(xdr, OP_CB_SEQUENCE);
383 encode_sessionid4(xdr, session);
385 p = xdr_reserve_space(xdr, 4 + 4 + 4 + 4 + 4);
386 *p++ = cpu_to_be32(session->se_cb_seq_nr); /* csa_sequenceid */
387 *p++ = xdr_zero; /* csa_slotid */
388 *p++ = xdr_zero; /* csa_highest_slotid */
389 *p++ = xdr_zero; /* csa_cachethis */
390 xdr_encode_empty_array(p); /* csa_referring_call_lists */
398 * struct CB_SEQUENCE4resok {
399 * sessionid4 csr_sessionid;
400 * sequenceid4 csr_sequenceid;
401 * slotid4 csr_slotid;
402 * slotid4 csr_highest_slotid;
403 * slotid4 csr_target_highest_slotid;
406 * union CB_SEQUENCE4res switch (nfsstat4 csr_status) {
408 * CB_SEQUENCE4resok csr_resok4;
413 * Our current back channel implmentation supports a single backchannel
414 * with a single slot.
416 static int decode_cb_sequence4resok(struct xdr_stream *xdr,
417 struct nfsd4_callback *cb)
419 struct nfsd4_session *session = cb->cb_clp->cl_cb_session;
420 int status = -ESERVERFAULT;
425 * If the server returns different values for sessionID, slotID or
426 * sequence number, the server is looney tunes.
428 p = xdr_inline_decode(xdr, NFS4_MAX_SESSIONID_LEN + 4 + 4 + 4 + 4);
429 if (unlikely(p == NULL))
432 if (memcmp(p, session->se_sessionid.data, NFS4_MAX_SESSIONID_LEN)) {
433 dprintk("NFS: %s Invalid session id\n", __func__);
436 p += XDR_QUADLEN(NFS4_MAX_SESSIONID_LEN);
438 dummy = be32_to_cpup(p++);
439 if (dummy != session->se_cb_seq_nr) {
440 dprintk("NFS: %s Invalid sequence number\n", __func__);
444 dummy = be32_to_cpup(p++);
446 dprintk("NFS: %s Invalid slotid\n", __func__);
451 * FIXME: process highest slotid and target highest slotid
455 cb->cb_seq_status = status;
462 static int decode_cb_sequence4res(struct xdr_stream *xdr,
463 struct nfsd4_callback *cb)
467 if (cb->cb_clp->cl_minorversion == 0)
470 status = decode_cb_op_status(xdr, OP_CB_SEQUENCE, &cb->cb_seq_status);
471 if (unlikely(status || cb->cb_seq_status))
474 return decode_cb_sequence4resok(xdr, cb);
478 * NFSv4.0 and NFSv4.1 XDR encode functions
480 * NFSv4.0 callback argument types are defined in section 15 of RFC
481 * 3530: "Network File System (NFS) version 4 Protocol" and section 20
482 * of RFC 5661: "Network File System (NFS) Version 4 Minor Version 1
487 * NB: Without this zero space reservation, callbacks over krb5p fail
489 static void nfs4_xdr_enc_cb_null(struct rpc_rqst *req, struct xdr_stream *xdr,
490 const void *__unused)
492 xdr_reserve_space(xdr, 0);
496 * 20.2. Operation 4: CB_RECALL - Recall a Delegation
498 static void nfs4_xdr_enc_cb_recall(struct rpc_rqst *req, struct xdr_stream *xdr,
501 const struct nfsd4_callback *cb = data;
502 const struct nfs4_delegation *dp = cb_to_delegation(cb);
503 struct nfs4_cb_compound_hdr hdr = {
504 .ident = cb->cb_clp->cl_cb_ident,
505 .minorversion = cb->cb_clp->cl_minorversion,
508 encode_cb_compound4args(xdr, &hdr);
509 encode_cb_sequence4args(xdr, cb, &hdr);
510 encode_cb_recall4args(xdr, dp, &hdr);
511 encode_cb_nops(&hdr);
515 * 20.6. Operation 8: CB_RECALL_ANY - Keep Any N Recallable Objects
518 nfs4_xdr_enc_cb_recall_any(struct rpc_rqst *req,
519 struct xdr_stream *xdr, const void *data)
521 const struct nfsd4_callback *cb = data;
522 struct nfsd4_cb_recall_any *ra;
523 struct nfs4_cb_compound_hdr hdr = {
524 .ident = cb->cb_clp->cl_cb_ident,
525 .minorversion = cb->cb_clp->cl_minorversion,
528 ra = container_of(cb, struct nfsd4_cb_recall_any, ra_cb);
529 encode_cb_compound4args(xdr, &hdr);
530 encode_cb_sequence4args(xdr, cb, &hdr);
531 encode_cb_recallany4args(xdr, &hdr, ra);
532 encode_cb_nops(&hdr);
536 * NFSv4.0 and NFSv4.1 XDR decode functions
538 * NFSv4.0 callback result types are defined in section 15 of RFC
539 * 3530: "Network File System (NFS) version 4 Protocol" and section 20
540 * of RFC 5661: "Network File System (NFS) Version 4 Minor Version 1
544 static int nfs4_xdr_dec_cb_null(struct rpc_rqst *req, struct xdr_stream *xdr,
551 * 20.2. Operation 4: CB_RECALL - Recall a Delegation
553 static int nfs4_xdr_dec_cb_recall(struct rpc_rqst *rqstp,
554 struct xdr_stream *xdr,
557 struct nfsd4_callback *cb = data;
558 struct nfs4_cb_compound_hdr hdr;
561 status = decode_cb_compound4res(xdr, &hdr);
562 if (unlikely(status))
565 status = decode_cb_sequence4res(xdr, cb);
566 if (unlikely(status || cb->cb_seq_status))
569 return decode_cb_op_status(xdr, OP_CB_RECALL, &cb->cb_status);
573 * 20.6. Operation 8: CB_RECALL_ANY - Keep Any N Recallable Objects
576 nfs4_xdr_dec_cb_recall_any(struct rpc_rqst *rqstp,
577 struct xdr_stream *xdr,
580 struct nfsd4_callback *cb = data;
581 struct nfs4_cb_compound_hdr hdr;
584 status = decode_cb_compound4res(xdr, &hdr);
585 if (unlikely(status))
587 status = decode_cb_sequence4res(xdr, cb);
588 if (unlikely(status || cb->cb_seq_status))
590 status = decode_cb_op_status(xdr, OP_CB_RECALL_ANY, &cb->cb_status);
594 #ifdef CONFIG_NFSD_PNFS
596 * CB_LAYOUTRECALL4args
598 * struct layoutrecall_file4 {
600 * offset4 lor_offset;
601 * length4 lor_length;
602 * stateid4 lor_stateid;
605 * union layoutrecall4 switch(layoutrecall_type4 lor_recalltype) {
606 * case LAYOUTRECALL4_FILE:
607 * layoutrecall_file4 lor_layout;
608 * case LAYOUTRECALL4_FSID:
610 * case LAYOUTRECALL4_ALL:
614 * struct CB_LAYOUTRECALL4args {
615 * layouttype4 clora_type;
616 * layoutiomode4 clora_iomode;
617 * bool clora_changed;
618 * layoutrecall4 clora_recall;
621 static void encode_cb_layout4args(struct xdr_stream *xdr,
622 const struct nfs4_layout_stateid *ls,
623 struct nfs4_cb_compound_hdr *hdr)
627 BUG_ON(hdr->minorversion == 0);
629 p = xdr_reserve_space(xdr, 5 * 4);
630 *p++ = cpu_to_be32(OP_CB_LAYOUTRECALL);
631 *p++ = cpu_to_be32(ls->ls_layout_type);
632 *p++ = cpu_to_be32(IOMODE_ANY);
633 *p++ = cpu_to_be32(1);
634 *p = cpu_to_be32(RETURN_FILE);
636 encode_nfs_fh4(xdr, &ls->ls_stid.sc_file->fi_fhandle);
638 p = xdr_reserve_space(xdr, 2 * 8);
639 p = xdr_encode_hyper(p, 0);
640 xdr_encode_hyper(p, NFS4_MAX_UINT64);
642 encode_stateid4(xdr, &ls->ls_recall_sid);
647 static void nfs4_xdr_enc_cb_layout(struct rpc_rqst *req,
648 struct xdr_stream *xdr,
651 const struct nfsd4_callback *cb = data;
652 const struct nfs4_layout_stateid *ls =
653 container_of(cb, struct nfs4_layout_stateid, ls_recall);
654 struct nfs4_cb_compound_hdr hdr = {
656 .minorversion = cb->cb_clp->cl_minorversion,
659 encode_cb_compound4args(xdr, &hdr);
660 encode_cb_sequence4args(xdr, cb, &hdr);
661 encode_cb_layout4args(xdr, ls, &hdr);
662 encode_cb_nops(&hdr);
665 static int nfs4_xdr_dec_cb_layout(struct rpc_rqst *rqstp,
666 struct xdr_stream *xdr,
669 struct nfsd4_callback *cb = data;
670 struct nfs4_cb_compound_hdr hdr;
673 status = decode_cb_compound4res(xdr, &hdr);
674 if (unlikely(status))
677 status = decode_cb_sequence4res(xdr, cb);
678 if (unlikely(status || cb->cb_seq_status))
681 return decode_cb_op_status(xdr, OP_CB_LAYOUTRECALL, &cb->cb_status);
683 #endif /* CONFIG_NFSD_PNFS */
685 static void encode_stateowner(struct xdr_stream *xdr, struct nfs4_stateowner *so)
689 p = xdr_reserve_space(xdr, 8 + 4 + so->so_owner.len);
690 p = xdr_encode_opaque_fixed(p, &so->so_client->cl_clientid, 8);
691 xdr_encode_opaque(p, so->so_owner.data, so->so_owner.len);
694 static void nfs4_xdr_enc_cb_notify_lock(struct rpc_rqst *req,
695 struct xdr_stream *xdr,
698 const struct nfsd4_callback *cb = data;
699 const struct nfsd4_blocked_lock *nbl =
700 container_of(cb, struct nfsd4_blocked_lock, nbl_cb);
701 struct nfs4_lockowner *lo = (struct nfs4_lockowner *)nbl->nbl_lock.fl_owner;
702 struct nfs4_cb_compound_hdr hdr = {
704 .minorversion = cb->cb_clp->cl_minorversion,
709 BUG_ON(hdr.minorversion == 0);
711 encode_cb_compound4args(xdr, &hdr);
712 encode_cb_sequence4args(xdr, cb, &hdr);
714 p = xdr_reserve_space(xdr, 4);
715 *p = cpu_to_be32(OP_CB_NOTIFY_LOCK);
716 encode_nfs_fh4(xdr, &nbl->nbl_fh);
717 encode_stateowner(xdr, &lo->lo_owner);
720 encode_cb_nops(&hdr);
723 static int nfs4_xdr_dec_cb_notify_lock(struct rpc_rqst *rqstp,
724 struct xdr_stream *xdr,
727 struct nfsd4_callback *cb = data;
728 struct nfs4_cb_compound_hdr hdr;
731 status = decode_cb_compound4res(xdr, &hdr);
732 if (unlikely(status))
735 status = decode_cb_sequence4res(xdr, cb);
736 if (unlikely(status || cb->cb_seq_status))
739 return decode_cb_op_status(xdr, OP_CB_NOTIFY_LOCK, &cb->cb_status);
743 * struct write_response4 {
744 * stateid4 wr_callback_id<1>;
746 * stable_how4 wr_committed;
747 * verifier4 wr_writeverf;
749 * union offload_info4 switch (nfsstat4 coa_status) {
751 * write_response4 coa_resok4;
753 * length4 coa_bytes_copied;
755 * struct CB_OFFLOAD4args {
757 * stateid4 coa_stateid;
758 * offload_info4 coa_offload_info;
761 static void encode_offload_info4(struct xdr_stream *xdr,
762 const struct nfsd4_cb_offload *cbo)
766 p = xdr_reserve_space(xdr, 4);
768 switch (cbo->co_nfserr) {
770 p = xdr_reserve_space(xdr, 4 + 8 + 4 + NFS4_VERIFIER_SIZE);
771 p = xdr_encode_empty_array(p);
772 p = xdr_encode_hyper(p, cbo->co_res.wr_bytes_written);
773 *p++ = cpu_to_be32(cbo->co_res.wr_stable_how);
774 p = xdr_encode_opaque_fixed(p, cbo->co_res.wr_verifier.data,
778 p = xdr_reserve_space(xdr, 8);
779 /* We always return success if bytes were written */
780 p = xdr_encode_hyper(p, 0);
784 static void encode_cb_offload4args(struct xdr_stream *xdr,
785 const struct nfsd4_cb_offload *cbo,
786 struct nfs4_cb_compound_hdr *hdr)
790 p = xdr_reserve_space(xdr, 4);
791 *p = cpu_to_be32(OP_CB_OFFLOAD);
792 encode_nfs_fh4(xdr, &cbo->co_fh);
793 encode_stateid4(xdr, &cbo->co_res.cb_stateid);
794 encode_offload_info4(xdr, cbo);
799 static void nfs4_xdr_enc_cb_offload(struct rpc_rqst *req,
800 struct xdr_stream *xdr,
803 const struct nfsd4_callback *cb = data;
804 const struct nfsd4_cb_offload *cbo =
805 container_of(cb, struct nfsd4_cb_offload, co_cb);
806 struct nfs4_cb_compound_hdr hdr = {
808 .minorversion = cb->cb_clp->cl_minorversion,
811 encode_cb_compound4args(xdr, &hdr);
812 encode_cb_sequence4args(xdr, cb, &hdr);
813 encode_cb_offload4args(xdr, cbo, &hdr);
814 encode_cb_nops(&hdr);
817 static int nfs4_xdr_dec_cb_offload(struct rpc_rqst *rqstp,
818 struct xdr_stream *xdr,
821 struct nfsd4_callback *cb = data;
822 struct nfs4_cb_compound_hdr hdr;
825 status = decode_cb_compound4res(xdr, &hdr);
826 if (unlikely(status))
829 status = decode_cb_sequence4res(xdr, cb);
830 if (unlikely(status || cb->cb_seq_status))
833 return decode_cb_op_status(xdr, OP_CB_OFFLOAD, &cb->cb_status);
836 * RPC procedure tables
838 #define PROC(proc, call, argtype, restype) \
839 [NFSPROC4_CLNT_##proc] = { \
840 .p_proc = NFSPROC4_CB_##call, \
841 .p_encode = nfs4_xdr_enc_##argtype, \
842 .p_decode = nfs4_xdr_dec_##restype, \
843 .p_arglen = NFS4_enc_##argtype##_sz, \
844 .p_replen = NFS4_dec_##restype##_sz, \
845 .p_statidx = NFSPROC4_CB_##call, \
849 static const struct rpc_procinfo nfs4_cb_procedures[] = {
850 PROC(CB_NULL, NULL, cb_null, cb_null),
851 PROC(CB_RECALL, COMPOUND, cb_recall, cb_recall),
852 #ifdef CONFIG_NFSD_PNFS
853 PROC(CB_LAYOUT, COMPOUND, cb_layout, cb_layout),
855 PROC(CB_NOTIFY_LOCK, COMPOUND, cb_notify_lock, cb_notify_lock),
856 PROC(CB_OFFLOAD, COMPOUND, cb_offload, cb_offload),
857 PROC(CB_RECALL_ANY, COMPOUND, cb_recall_any, cb_recall_any),
860 static unsigned int nfs4_cb_counts[ARRAY_SIZE(nfs4_cb_procedures)];
861 static const struct rpc_version nfs_cb_version4 = {
863 * Note on the callback rpc program version number: despite language in rfc
864 * 5661 section 18.36.3 requiring servers to use 4 in this field, the
865 * official xdr descriptions for both 4.0 and 4.1 specify version 1, and
866 * in practice that appears to be what implementations use. The section
867 * 18.36.3 language is expected to be fixed in an erratum.
870 .nrprocs = ARRAY_SIZE(nfs4_cb_procedures),
871 .procs = nfs4_cb_procedures,
872 .counts = nfs4_cb_counts,
875 static const struct rpc_version *nfs_cb_version[2] = {
876 [1] = &nfs_cb_version4,
879 static const struct rpc_program cb_program;
881 static struct rpc_stat cb_stats = {
882 .program = &cb_program
885 #define NFS4_CALLBACK 0x40000000
886 static const struct rpc_program cb_program = {
888 .number = NFS4_CALLBACK,
889 .nrvers = ARRAY_SIZE(nfs_cb_version),
890 .version = nfs_cb_version,
892 .pipe_dir_name = "nfsd4_cb",
895 static int max_cb_time(struct net *net)
897 struct nfsd_net *nn = net_generic(net, nfsd_net_id);
900 * nfsd4_lease is set to at most one hour in __nfsd4_write_time,
901 * so we can use 32-bit math on it. Warn if that assumption
902 * ever stops being true.
904 if (WARN_ON_ONCE(nn->nfsd4_lease > 3600))
907 return max(((u32)nn->nfsd4_lease)/10, 1u) * HZ;
910 static struct workqueue_struct *callback_wq;
912 static bool nfsd4_queue_cb(struct nfsd4_callback *cb)
914 return queue_work(callback_wq, &cb->cb_work);
917 static void nfsd41_cb_inflight_begin(struct nfs4_client *clp)
919 atomic_inc(&clp->cl_cb_inflight);
922 static void nfsd41_cb_inflight_end(struct nfs4_client *clp)
925 if (atomic_dec_and_test(&clp->cl_cb_inflight))
926 wake_up_var(&clp->cl_cb_inflight);
929 static void nfsd41_cb_inflight_wait_complete(struct nfs4_client *clp)
931 wait_var_event(&clp->cl_cb_inflight,
932 !atomic_read(&clp->cl_cb_inflight));
935 static const struct cred *get_backchannel_cred(struct nfs4_client *clp, struct rpc_clnt *client, struct nfsd4_session *ses)
937 if (clp->cl_minorversion == 0) {
938 client->cl_principal = clp->cl_cred.cr_targ_princ ?
939 clp->cl_cred.cr_targ_princ : "nfs";
941 return get_cred(rpc_machine_cred());
945 kcred = prepare_kernel_cred(&init_task);
949 kcred->uid = ses->se_cb_sec.uid;
950 kcred->gid = ses->se_cb_sec.gid;
955 static int setup_callback_client(struct nfs4_client *clp, struct nfs4_cb_conn *conn, struct nfsd4_session *ses)
957 int maxtime = max_cb_time(clp->net);
958 struct rpc_timeout timeparms = {
959 .to_initval = maxtime,
961 .to_maxval = maxtime,
963 struct rpc_create_args args = {
965 .address = (struct sockaddr *) &conn->cb_addr,
966 .addrsize = conn->cb_addrlen,
967 .saddress = (struct sockaddr *) &conn->cb_saddr,
968 .timeout = &timeparms,
969 .program = &cb_program,
971 .flags = (RPC_CLNT_CREATE_NOPING | RPC_CLNT_CREATE_QUIET),
972 .cred = current_cred(),
974 struct rpc_clnt *client;
975 const struct cred *cred;
977 if (clp->cl_minorversion == 0) {
978 if (!clp->cl_cred.cr_principal &&
979 (clp->cl_cred.cr_flavor >= RPC_AUTH_GSS_KRB5)) {
980 trace_nfsd_cb_setup_err(clp, -EINVAL);
983 args.client_name = clp->cl_cred.cr_principal;
984 args.prognumber = conn->cb_prog;
985 args.protocol = XPRT_TRANSPORT_TCP;
986 args.authflavor = clp->cl_cred.cr_flavor;
987 clp->cl_cb_ident = conn->cb_ident;
991 clp->cl_cb_conn.cb_xprt = conn->cb_xprt;
992 clp->cl_cb_session = ses;
993 args.bc_xprt = conn->cb_xprt;
994 args.prognumber = clp->cl_cb_session->se_cb_prog;
995 args.protocol = conn->cb_xprt->xpt_class->xcl_ident |
997 args.authflavor = ses->se_cb_sec.flavor;
999 /* Create RPC client */
1000 client = rpc_create(&args);
1001 if (IS_ERR(client)) {
1002 trace_nfsd_cb_setup_err(clp, PTR_ERR(client));
1003 return PTR_ERR(client);
1005 cred = get_backchannel_cred(clp, client, ses);
1007 trace_nfsd_cb_setup_err(clp, -ENOMEM);
1008 rpc_shutdown_client(client);
1011 clp->cl_cb_client = client;
1012 clp->cl_cb_cred = cred;
1014 trace_nfsd_cb_setup(clp, rpc_peeraddr2str(client, RPC_DISPLAY_NETID),
1020 static void nfsd4_mark_cb_state(struct nfs4_client *clp, int newstate)
1022 if (clp->cl_cb_state != newstate) {
1023 clp->cl_cb_state = newstate;
1024 trace_nfsd_cb_state(clp);
1028 static void nfsd4_mark_cb_down(struct nfs4_client *clp, int reason)
1030 if (test_bit(NFSD4_CLIENT_CB_UPDATE, &clp->cl_flags))
1032 nfsd4_mark_cb_state(clp, NFSD4_CB_DOWN);
1035 static void nfsd4_mark_cb_fault(struct nfs4_client *clp, int reason)
1037 if (test_bit(NFSD4_CLIENT_CB_UPDATE, &clp->cl_flags))
1039 nfsd4_mark_cb_state(clp, NFSD4_CB_FAULT);
1042 static void nfsd4_cb_probe_done(struct rpc_task *task, void *calldata)
1044 struct nfs4_client *clp = container_of(calldata, struct nfs4_client, cl_cb_null);
1046 if (task->tk_status)
1047 nfsd4_mark_cb_down(clp, task->tk_status);
1049 nfsd4_mark_cb_state(clp, NFSD4_CB_UP);
1052 static void nfsd4_cb_probe_release(void *calldata)
1054 struct nfs4_client *clp = container_of(calldata, struct nfs4_client, cl_cb_null);
1056 nfsd41_cb_inflight_end(clp);
1060 static const struct rpc_call_ops nfsd4_cb_probe_ops = {
1061 /* XXX: release method to ensure we set the cb channel down if
1062 * necessary on early failure? */
1063 .rpc_call_done = nfsd4_cb_probe_done,
1064 .rpc_release = nfsd4_cb_probe_release,
1068 * Poke the callback thread to process any updates to the callback
1069 * parameters, and send a null probe.
1071 void nfsd4_probe_callback(struct nfs4_client *clp)
1073 trace_nfsd_cb_probe(clp);
1074 nfsd4_mark_cb_state(clp, NFSD4_CB_UNKNOWN);
1075 set_bit(NFSD4_CLIENT_CB_UPDATE, &clp->cl_flags);
1076 nfsd4_run_cb(&clp->cl_cb_null);
1079 void nfsd4_probe_callback_sync(struct nfs4_client *clp)
1081 nfsd4_probe_callback(clp);
1082 flush_workqueue(callback_wq);
1085 void nfsd4_change_callback(struct nfs4_client *clp, struct nfs4_cb_conn *conn)
1087 nfsd4_mark_cb_state(clp, NFSD4_CB_UNKNOWN);
1088 spin_lock(&clp->cl_lock);
1089 memcpy(&clp->cl_cb_conn, conn, sizeof(struct nfs4_cb_conn));
1090 spin_unlock(&clp->cl_lock);
1094 * There's currently a single callback channel slot.
1095 * If the slot is available, then mark it busy. Otherwise, set the
1096 * thread for sleeping on the callback RPC wait queue.
1098 static bool nfsd41_cb_get_slot(struct nfsd4_callback *cb, struct rpc_task *task)
1100 struct nfs4_client *clp = cb->cb_clp;
1102 if (!cb->cb_holds_slot &&
1103 test_and_set_bit(0, &clp->cl_cb_slot_busy) != 0) {
1104 rpc_sleep_on(&clp->cl_cb_waitq, task, NULL);
1106 if (test_and_set_bit(0, &clp->cl_cb_slot_busy) != 0) {
1107 dprintk("%s slot is busy\n", __func__);
1110 rpc_wake_up_queued_task(&clp->cl_cb_waitq, task);
1112 cb->cb_holds_slot = true;
1116 static void nfsd41_cb_release_slot(struct nfsd4_callback *cb)
1118 struct nfs4_client *clp = cb->cb_clp;
1120 if (cb->cb_holds_slot) {
1121 cb->cb_holds_slot = false;
1122 clear_bit(0, &clp->cl_cb_slot_busy);
1123 rpc_wake_up_next(&clp->cl_cb_waitq);
1127 static void nfsd41_destroy_cb(struct nfsd4_callback *cb)
1129 struct nfs4_client *clp = cb->cb_clp;
1131 nfsd41_cb_release_slot(cb);
1132 if (cb->cb_ops && cb->cb_ops->release)
1133 cb->cb_ops->release(cb);
1134 nfsd41_cb_inflight_end(clp);
1138 * TODO: cb_sequence should support referring call lists, cachethis, multiple
1139 * slots, and mark callback channel down on communication errors.
1141 static void nfsd4_cb_prepare(struct rpc_task *task, void *calldata)
1143 struct nfsd4_callback *cb = calldata;
1144 struct nfs4_client *clp = cb->cb_clp;
1145 u32 minorversion = clp->cl_minorversion;
1148 * cb_seq_status is only set in decode_cb_sequence4res,
1149 * and so will remain 1 if an rpc level failure occurs.
1151 cb->cb_seq_status = 1;
1153 if (minorversion && !nfsd41_cb_get_slot(cb, task))
1155 rpc_call_start(task);
1158 static bool nfsd4_cb_sequence_done(struct rpc_task *task, struct nfsd4_callback *cb)
1160 struct nfs4_client *clp = cb->cb_clp;
1161 struct nfsd4_session *session = clp->cl_cb_session;
1164 if (!clp->cl_minorversion) {
1166 * If the backchannel connection was shut down while this
1167 * task was queued, we need to resubmit it after setting up
1168 * a new backchannel connection.
1170 * Note that if we lost our callback connection permanently
1171 * the submission code will error out, so we don't need to
1172 * handle that case here.
1174 if (RPC_SIGNALLED(task))
1180 if (!cb->cb_holds_slot)
1183 switch (cb->cb_seq_status) {
1186 * No need for lock, access serialized in nfsd4_cb_prepare
1189 * If CB_SEQUENCE returns an error, then the state of the slot
1190 * (sequence ID, cached reply) MUST NOT change.
1192 ++session->se_cb_seq_nr;
1195 ++session->se_cb_seq_nr;
1198 case -NFS4ERR_BADSESSION:
1199 nfsd4_mark_cb_fault(cb->cb_clp, cb->cb_seq_status);
1202 case -NFS4ERR_DELAY:
1203 if (!rpc_restart_call(task))
1206 rpc_delay(task, 2 * HZ);
1208 case -NFS4ERR_BADSLOT:
1210 case -NFS4ERR_SEQ_MISORDERED:
1211 if (session->se_cb_seq_nr != 1) {
1212 session->se_cb_seq_nr = 1;
1217 nfsd4_mark_cb_fault(cb->cb_clp, cb->cb_seq_status);
1218 dprintk("%s: unprocessed error %d\n", __func__,
1222 nfsd41_cb_release_slot(cb);
1223 dprintk("%s: freed slot, new seqid=%d\n", __func__,
1224 clp->cl_cb_session->se_cb_seq_nr);
1226 if (RPC_SIGNALLED(task))
1231 if (rpc_restart_call_prepare(task))
1235 if (!test_bit(NFSD4_CLIENT_CB_KILL, &clp->cl_flags)) {
1236 task->tk_status = 0;
1237 cb->cb_need_restart = true;
1242 static void nfsd4_cb_done(struct rpc_task *task, void *calldata)
1244 struct nfsd4_callback *cb = calldata;
1245 struct nfs4_client *clp = cb->cb_clp;
1247 if (!nfsd4_cb_sequence_done(task, cb))
1250 if (cb->cb_status) {
1251 WARN_ON_ONCE(task->tk_status);
1252 task->tk_status = cb->cb_status;
1255 switch (cb->cb_ops->done(cb, task)) {
1257 task->tk_status = 0;
1258 rpc_restart_call_prepare(task);
1261 switch (task->tk_status) {
1265 nfsd4_mark_cb_down(clp, task->tk_status);
1273 static void nfsd4_cb_release(void *calldata)
1275 struct nfsd4_callback *cb = calldata;
1277 if (cb->cb_need_restart)
1280 nfsd41_destroy_cb(cb);
1284 static const struct rpc_call_ops nfsd4_cb_ops = {
1285 .rpc_call_prepare = nfsd4_cb_prepare,
1286 .rpc_call_done = nfsd4_cb_done,
1287 .rpc_release = nfsd4_cb_release,
1290 int nfsd4_create_callback_queue(void)
1292 callback_wq = alloc_ordered_workqueue("nfsd4_callbacks", 0);
1298 void nfsd4_destroy_callback_queue(void)
1300 destroy_workqueue(callback_wq);
1303 /* must be called under the state lock */
1304 void nfsd4_shutdown_callback(struct nfs4_client *clp)
1306 if (clp->cl_cb_state != NFSD4_CB_UNKNOWN)
1307 trace_nfsd_cb_shutdown(clp);
1309 set_bit(NFSD4_CLIENT_CB_KILL, &clp->cl_flags);
1311 * Note this won't actually result in a null callback;
1312 * instead, nfsd4_run_cb_null() will detect the killed
1313 * client, destroy the rpc client, and stop:
1315 nfsd4_run_cb(&clp->cl_cb_null);
1316 flush_workqueue(callback_wq);
1317 nfsd41_cb_inflight_wait_complete(clp);
1320 /* requires cl_lock: */
1321 static struct nfsd4_conn * __nfsd4_find_backchannel(struct nfs4_client *clp)
1323 struct nfsd4_session *s;
1324 struct nfsd4_conn *c;
1326 list_for_each_entry(s, &clp->cl_sessions, se_perclnt) {
1327 list_for_each_entry(c, &s->se_conns, cn_persession) {
1328 if (c->cn_flags & NFS4_CDFC4_BACK)
1336 * Note there isn't a lot of locking in this code; instead we depend on
1337 * the fact that it is run from the callback_wq, which won't run two
1338 * work items at once. So, for example, callback_wq handles all access
1339 * of cl_cb_client and all calls to rpc_create or rpc_shutdown_client.
1341 static void nfsd4_process_cb_update(struct nfsd4_callback *cb)
1343 struct nfs4_cb_conn conn;
1344 struct nfs4_client *clp = cb->cb_clp;
1345 struct nfsd4_session *ses = NULL;
1346 struct nfsd4_conn *c;
1350 * This is either an update, or the client dying; in either case,
1351 * kill the old client:
1353 if (clp->cl_cb_client) {
1354 rpc_shutdown_client(clp->cl_cb_client);
1355 clp->cl_cb_client = NULL;
1356 put_cred(clp->cl_cb_cred);
1357 clp->cl_cb_cred = NULL;
1359 if (clp->cl_cb_conn.cb_xprt) {
1360 svc_xprt_put(clp->cl_cb_conn.cb_xprt);
1361 clp->cl_cb_conn.cb_xprt = NULL;
1363 if (test_bit(NFSD4_CLIENT_CB_KILL, &clp->cl_flags))
1365 spin_lock(&clp->cl_lock);
1367 * Only serialized callback code is allowed to clear these
1368 * flags; main nfsd code can only set them:
1370 BUG_ON(!(clp->cl_flags & NFSD4_CLIENT_CB_FLAG_MASK));
1371 clear_bit(NFSD4_CLIENT_CB_UPDATE, &clp->cl_flags);
1372 memcpy(&conn, &cb->cb_clp->cl_cb_conn, sizeof(struct nfs4_cb_conn));
1373 c = __nfsd4_find_backchannel(clp);
1375 svc_xprt_get(c->cn_xprt);
1376 conn.cb_xprt = c->cn_xprt;
1377 ses = c->cn_session;
1379 spin_unlock(&clp->cl_lock);
1381 err = setup_callback_client(clp, &conn, ses);
1383 nfsd4_mark_cb_down(clp, err);
1385 svc_xprt_put(c->cn_xprt);
1391 nfsd4_run_cb_work(struct work_struct *work)
1393 struct nfsd4_callback *cb =
1394 container_of(work, struct nfsd4_callback, cb_work);
1395 struct nfs4_client *clp = cb->cb_clp;
1396 struct rpc_clnt *clnt;
1399 if (cb->cb_need_restart) {
1400 cb->cb_need_restart = false;
1402 if (cb->cb_ops && cb->cb_ops->prepare)
1403 cb->cb_ops->prepare(cb);
1406 if (clp->cl_flags & NFSD4_CLIENT_CB_FLAG_MASK)
1407 nfsd4_process_cb_update(cb);
1409 clnt = clp->cl_cb_client;
1411 /* Callback channel broken, or client killed; give up: */
1412 nfsd41_destroy_cb(cb);
1417 * Don't send probe messages for 4.1 or later.
1419 if (!cb->cb_ops && clp->cl_minorversion) {
1420 nfsd4_mark_cb_state(clp, NFSD4_CB_UP);
1421 nfsd41_destroy_cb(cb);
1425 cb->cb_msg.rpc_cred = clp->cl_cb_cred;
1426 flags = clp->cl_minorversion ? RPC_TASK_NOCONNECT : RPC_TASK_SOFTCONN;
1427 rpc_call_async(clnt, &cb->cb_msg, RPC_TASK_SOFT | flags,
1428 cb->cb_ops ? &nfsd4_cb_ops : &nfsd4_cb_probe_ops, cb);
1431 void nfsd4_init_cb(struct nfsd4_callback *cb, struct nfs4_client *clp,
1432 const struct nfsd4_callback_ops *ops, enum nfsd4_cb_op op)
1435 cb->cb_msg.rpc_proc = &nfs4_cb_procedures[op];
1436 cb->cb_msg.rpc_argp = cb;
1437 cb->cb_msg.rpc_resp = cb;
1439 INIT_WORK(&cb->cb_work, nfsd4_run_cb_work);
1440 cb->cb_seq_status = 1;
1442 cb->cb_need_restart = false;
1443 cb->cb_holds_slot = false;
1447 * nfsd4_run_cb - queue up a callback job to run
1448 * @cb: callback to queue
1450 * Kick off a callback to do its thing. Returns false if it was already
1451 * on a queue, true otherwise.
1453 bool nfsd4_run_cb(struct nfsd4_callback *cb)
1455 struct nfs4_client *clp = cb->cb_clp;
1458 nfsd41_cb_inflight_begin(clp);
1459 queued = nfsd4_queue_cb(cb);
1461 nfsd41_cb_inflight_end(clp);