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/svc_xprt.h>
36 #include <linux/slab.h>
40 #define NFSDDBG_FACILITY NFSDDBG_PROC
42 #define NFSPROC4_CB_NULL 0
43 #define NFSPROC4_CB_COMPOUND 1
45 /* Index of predefined Linux callback client operations */
48 NFSPROC4_CLNT_CB_NULL = 0,
49 NFSPROC4_CLNT_CB_RECALL,
50 NFSPROC4_CLNT_CB_SEQUENCE,
58 #define NFS4_MAXTAGLEN 20
60 #define NFS4_enc_cb_null_sz 0
61 #define NFS4_dec_cb_null_sz 0
62 #define cb_compound_enc_hdr_sz 4
63 #define cb_compound_dec_hdr_sz (3 + (NFS4_MAXTAGLEN >> 2))
64 #define sessionid_sz (NFS4_MAX_SESSIONID_LEN >> 2)
65 #define cb_sequence_enc_sz (sessionid_sz + 4 + \
66 1 /* no referring calls list yet */)
67 #define cb_sequence_dec_sz (op_dec_sz + sessionid_sz + 4)
71 #define enc_nfs4_fh_sz (1 + (NFS4_FHSIZE >> 2))
72 #define enc_stateid_sz (NFS4_STATEID_SIZE >> 2)
73 #define NFS4_enc_cb_recall_sz (cb_compound_enc_hdr_sz + \
74 cb_sequence_enc_sz + \
75 1 + enc_stateid_sz + \
78 #define NFS4_dec_cb_recall_sz (cb_compound_dec_hdr_sz + \
79 cb_sequence_dec_sz + \
83 * Generic encode routines from fs/nfs/nfs4xdr.c
85 static inline __be32 *
86 xdr_writemem(__be32 *p, const void *ptr, int nbytes)
88 int tmp = XDR_QUADLEN(nbytes);
92 memcpy(p, ptr, nbytes);
96 #define WRITE32(n) *p++ = htonl(n)
97 #define WRITEMEM(ptr,nbytes) do { \
98 p = xdr_writemem(p, ptr, nbytes); \
100 #define RESERVE_SPACE(nbytes) do { \
101 p = xdr_reserve_space(xdr, nbytes); \
102 if (!p) dprintk("NFSD: RESERVE_SPACE(%d) failed in function %s\n", (int) (nbytes), __func__); \
107 * Generic decode routines from fs/nfs/nfs4xdr.c
109 #define DECODE_TAIL \
114 dprintk("NFSD: xdr error! (%s:%d)\n", __FILE__, __LINE__); \
118 #define READ32(x) (x) = ntohl(*p++)
119 #define READ64(x) do { \
120 (x) = (u64)ntohl(*p++) << 32; \
121 (x) |= ntohl(*p++); \
123 #define READTIME(x) do { \
125 (x.tv_sec) = ntohl(*p++); \
126 (x.tv_nsec) = ntohl(*p++); \
128 #define READ_BUF(nbytes) do { \
129 p = xdr_inline_decode(xdr, nbytes); \
131 dprintk("NFSD: %s: reply buffer overflowed in line %d.\n", \
132 __func__, __LINE__); \
137 struct nfs4_cb_compound_hdr {
139 u32 ident; /* minorversion 0 only */
150 } nfs_cb_errtbl[] = {
152 { NFS4ERR_PERM, EPERM },
153 { NFS4ERR_NOENT, ENOENT },
155 { NFS4ERR_NXIO, ENXIO },
156 { NFS4ERR_ACCESS, EACCES },
157 { NFS4ERR_EXIST, EEXIST },
158 { NFS4ERR_XDEV, EXDEV },
159 { NFS4ERR_NOTDIR, ENOTDIR },
160 { NFS4ERR_ISDIR, EISDIR },
161 { NFS4ERR_INVAL, EINVAL },
162 { NFS4ERR_FBIG, EFBIG },
163 { NFS4ERR_NOSPC, ENOSPC },
164 { NFS4ERR_ROFS, EROFS },
165 { NFS4ERR_MLINK, EMLINK },
166 { NFS4ERR_NAMETOOLONG, ENAMETOOLONG },
167 { NFS4ERR_NOTEMPTY, ENOTEMPTY },
168 { NFS4ERR_DQUOT, EDQUOT },
169 { NFS4ERR_STALE, ESTALE },
170 { NFS4ERR_BADHANDLE, EBADHANDLE },
171 { NFS4ERR_BAD_COOKIE, EBADCOOKIE },
172 { NFS4ERR_NOTSUPP, ENOTSUPP },
173 { NFS4ERR_TOOSMALL, ETOOSMALL },
174 { NFS4ERR_SERVERFAULT, ESERVERFAULT },
175 { NFS4ERR_BADTYPE, EBADTYPE },
176 { NFS4ERR_LOCKED, EAGAIN },
177 { NFS4ERR_RESOURCE, EREMOTEIO },
178 { NFS4ERR_SYMLINK, ELOOP },
179 { NFS4ERR_OP_ILLEGAL, EOPNOTSUPP },
180 { NFS4ERR_DEADLOCK, EDEADLK },
185 nfs_cb_stat_to_errno(int stat)
188 for (i = 0; nfs_cb_errtbl[i].stat != -1; i++) {
189 if (nfs_cb_errtbl[i].stat == stat)
190 return nfs_cb_errtbl[i].errno;
192 /* If we cannot translate the error, the recovery routines should
194 * Note: remaining NFSv4 error codes have values > 10000, so should
195 * not conflict with native Linux error codes.
205 encode_stateid(struct xdr_stream *xdr, stateid_t *sid)
209 RESERVE_SPACE(sizeof(stateid_t));
210 WRITE32(sid->si_generation);
211 WRITEMEM(&sid->si_opaque, sizeof(stateid_opaque_t));
215 encode_cb_compound_hdr(struct xdr_stream *xdr, struct nfs4_cb_compound_hdr *hdr)
220 WRITE32(0); /* tag length is always 0 */
221 WRITE32(hdr->minorversion);
227 static void encode_cb_nops(struct nfs4_cb_compound_hdr *hdr)
229 *hdr->nops_p = htonl(hdr->nops);
233 encode_cb_recall(struct xdr_stream *xdr, struct nfs4_delegation *dp,
234 struct nfs4_cb_compound_hdr *hdr)
237 int len = dp->dl_fh.fh_size;
240 WRITE32(OP_CB_RECALL);
241 encode_stateid(xdr, &dp->dl_stateid);
242 RESERVE_SPACE(8 + (XDR_QUADLEN(len) << 2));
243 WRITE32(0); /* truncate optimization not implemented */
245 WRITEMEM(&dp->dl_fh.fh_base, len);
250 encode_cb_sequence(struct xdr_stream *xdr, struct nfsd4_callback *cb,
251 struct nfs4_cb_compound_hdr *hdr)
255 if (hdr->minorversion == 0)
258 RESERVE_SPACE(1 + NFS4_MAX_SESSIONID_LEN + 20);
260 WRITE32(OP_CB_SEQUENCE);
261 WRITEMEM(cb->cb_clp->cl_sessionid.data, NFS4_MAX_SESSIONID_LEN);
262 WRITE32(cb->cb_clp->cl_cb_seq_nr);
263 WRITE32(0); /* slotid, always 0 */
264 WRITE32(0); /* highest slotid always 0 */
265 WRITE32(0); /* cachethis always 0 */
266 WRITE32(0); /* FIXME: support referring_call_lists */
271 nfs4_xdr_enc_cb_null(struct rpc_rqst *req, __be32 *p)
273 struct xdr_stream xdrs, *xdr = &xdrs;
275 xdr_init_encode(&xdrs, &req->rq_snd_buf, p);
281 nfs4_xdr_enc_cb_recall(struct rpc_rqst *req, __be32 *p,
282 struct nfsd4_callback *cb)
284 struct xdr_stream xdr;
285 struct nfs4_delegation *args = cb->cb_op;
286 struct nfs4_cb_compound_hdr hdr = {
287 .ident = args->dl_ident,
288 .minorversion = cb->cb_minorversion,
291 xdr_init_encode(&xdr, &req->rq_snd_buf, p);
292 encode_cb_compound_hdr(&xdr, &hdr);
293 encode_cb_sequence(&xdr, cb, &hdr);
294 encode_cb_recall(&xdr, args, &hdr);
295 encode_cb_nops(&hdr);
301 decode_cb_compound_hdr(struct xdr_stream *xdr, struct nfs4_cb_compound_hdr *hdr){
307 /* We've got no use for the tag; ignore it: */
309 READ_BUF(taglen + 4);
310 p += XDR_QUADLEN(taglen);
316 decode_cb_op_hdr(struct xdr_stream *xdr, enum nfs_opnum4 expected)
324 if (op != expected) {
325 dprintk("NFSD: decode_cb_op_hdr: Callback server returned "
326 " operation %d but we issued a request for %d\n",
331 if (nfserr != NFS_OK)
332 return -nfs_cb_stat_to_errno(nfserr);
337 * Our current back channel implmentation supports a single backchannel
338 * with a single slot.
341 decode_cb_sequence(struct xdr_stream *xdr, struct nfsd4_callback *cb,
342 struct rpc_rqst *rqstp)
344 struct nfs4_sessionid id;
349 if (cb->cb_minorversion == 0)
352 status = decode_cb_op_hdr(xdr, OP_CB_SEQUENCE);
357 * If the server returns different values for sessionID, slotID or
358 * sequence number, the server is looney tunes.
360 status = -ESERVERFAULT;
362 READ_BUF(NFS4_MAX_SESSIONID_LEN + 16);
363 memcpy(id.data, p, NFS4_MAX_SESSIONID_LEN);
364 p += XDR_QUADLEN(NFS4_MAX_SESSIONID_LEN);
365 if (memcmp(id.data, cb->cb_clp->cl_sessionid.data,
366 NFS4_MAX_SESSIONID_LEN)) {
367 dprintk("%s Invalid session id\n", __func__);
371 if (dummy != cb->cb_clp->cl_cb_seq_nr) {
372 dprintk("%s Invalid sequence number\n", __func__);
375 READ32(dummy); /* slotid must be 0 */
377 dprintk("%s Invalid slotid\n", __func__);
380 /* FIXME: process highest slotid and target highest slotid */
388 nfs4_xdr_dec_cb_null(struct rpc_rqst *req, __be32 *p)
394 nfs4_xdr_dec_cb_recall(struct rpc_rqst *rqstp, __be32 *p,
395 struct nfsd4_callback *cb)
397 struct xdr_stream xdr;
398 struct nfs4_cb_compound_hdr hdr;
401 xdr_init_decode(&xdr, &rqstp->rq_rcv_buf, p);
402 status = decode_cb_compound_hdr(&xdr, &hdr);
406 status = decode_cb_sequence(&xdr, cb, rqstp);
410 status = decode_cb_op_hdr(&xdr, OP_CB_RECALL);
416 * RPC procedure tables
418 #define PROC(proc, call, argtype, restype) \
419 [NFSPROC4_CLNT_##proc] = { \
420 .p_proc = NFSPROC4_CB_##call, \
421 .p_encode = (kxdrproc_t) nfs4_xdr_##argtype, \
422 .p_decode = (kxdrproc_t) nfs4_xdr_##restype, \
423 .p_arglen = NFS4_##argtype##_sz, \
424 .p_replen = NFS4_##restype##_sz, \
425 .p_statidx = NFSPROC4_CB_##call, \
429 static struct rpc_procinfo nfs4_cb_procedures[] = {
430 PROC(CB_NULL, NULL, enc_cb_null, dec_cb_null),
431 PROC(CB_RECALL, COMPOUND, enc_cb_recall, dec_cb_recall),
434 static struct rpc_version nfs_cb_version4 = {
436 * Note on the callback rpc program version number: despite language in rfc
437 * 5661 section 18.36.3 requiring servers to use 4 in this field, the
438 * official xdr descriptions for both 4.0 and 4.1 specify version 1, and
439 * in practice that appears to be what implementations use. The section
440 * 18.36.3 language is expected to be fixed in an erratum.
443 .nrprocs = ARRAY_SIZE(nfs4_cb_procedures),
444 .procs = nfs4_cb_procedures
447 static struct rpc_version * nfs_cb_version[] = {
451 static struct rpc_program cb_program;
453 static struct rpc_stat cb_stats = {
454 .program = &cb_program
457 #define NFS4_CALLBACK 0x40000000
458 static struct rpc_program cb_program = {
460 .number = NFS4_CALLBACK,
461 .nrvers = ARRAY_SIZE(nfs_cb_version),
462 .version = nfs_cb_version,
464 .pipe_dir_name = "/nfsd4_cb",
467 static int max_cb_time(void)
469 return max(nfsd4_lease/10, (time_t)1) * HZ;
472 /* Reference counting, callback cleanup, etc., all look racy as heck.
473 * And why is cl_cb_set an atomic? */
475 int setup_callback_client(struct nfs4_client *clp, struct nfs4_cb_conn *conn)
477 struct rpc_timeout timeparms = {
478 .to_initval = max_cb_time(),
481 struct rpc_create_args args = {
483 .protocol = XPRT_TRANSPORT_TCP,
484 .address = (struct sockaddr *) &conn->cb_addr,
485 .addrsize = conn->cb_addrlen,
486 .timeout = &timeparms,
487 .program = &cb_program,
488 .prognumber = conn->cb_prog,
490 .authflavor = clp->cl_flavor,
491 .flags = (RPC_CLNT_CREATE_NOPING | RPC_CLNT_CREATE_QUIET),
492 .client_name = clp->cl_principal,
494 struct rpc_clnt *client;
496 if (!clp->cl_principal && (clp->cl_flavor >= RPC_AUTH_GSS_KRB5))
498 if (conn->cb_minorversion) {
499 args.bc_xprt = conn->cb_xprt;
500 args.protocol = XPRT_TRANSPORT_BC_TCP;
502 /* Create RPC client */
503 client = rpc_create(&args);
504 if (IS_ERR(client)) {
505 dprintk("NFSD: couldn't create callback client: %ld\n",
507 return PTR_ERR(client);
509 nfsd4_set_callback_client(clp, client);
514 static void warn_no_callback_path(struct nfs4_client *clp, int reason)
516 dprintk("NFSD: warning: no callback path to client %.*s: error %d\n",
517 (int)clp->cl_name.len, clp->cl_name.data, reason);
520 static void nfsd4_cb_probe_done(struct rpc_task *task, void *calldata)
522 struct nfs4_client *clp = container_of(calldata, struct nfs4_client, cl_cb_null);
525 warn_no_callback_path(clp, task->tk_status);
527 atomic_set(&clp->cl_cb_set, 1);
530 static const struct rpc_call_ops nfsd4_cb_probe_ops = {
531 /* XXX: release method to ensure we set the cb channel down if
532 * necessary on early failure? */
533 .rpc_call_done = nfsd4_cb_probe_done,
536 static struct rpc_cred *callback_cred;
538 int set_callback_cred(void)
542 callback_cred = rpc_lookup_machine_cred();
548 static struct workqueue_struct *callback_wq;
550 void do_probe_callback(struct nfs4_client *clp)
552 struct nfsd4_callback *cb = &clp->cl_cb_null;
557 cb->cb_msg.rpc_proc = &nfs4_cb_procedures[NFSPROC4_CLNT_CB_NULL];
558 cb->cb_msg.rpc_argp = NULL;
559 cb->cb_msg.rpc_resp = NULL;
560 cb->cb_msg.rpc_cred = callback_cred;
562 cb->cb_ops = &nfsd4_cb_probe_ops;
564 queue_work(callback_wq, &cb->cb_work);
568 * Set up the callback client and put a NFSPROC4_CB_NULL on the wire...
570 void nfsd4_probe_callback(struct nfs4_client *clp, struct nfs4_cb_conn *conn)
574 BUG_ON(atomic_read(&clp->cl_cb_set));
576 status = setup_callback_client(clp, conn);
578 warn_no_callback_path(clp, status);
581 do_probe_callback(clp);
585 * There's currently a single callback channel slot.
586 * If the slot is available, then mark it busy. Otherwise, set the
587 * thread for sleeping on the callback RPC wait queue.
589 static int nfsd41_cb_setup_sequence(struct nfs4_client *clp,
590 struct rpc_task *task)
592 u32 *ptr = (u32 *)clp->cl_sessionid.data;
595 dprintk("%s: %u:%u:%u:%u\n", __func__,
596 ptr[0], ptr[1], ptr[2], ptr[3]);
598 if (test_and_set_bit(0, &clp->cl_cb_slot_busy) != 0) {
599 rpc_sleep_on(&clp->cl_cb_waitq, task, NULL);
600 dprintk("%s slot is busy\n", __func__);
605 dprintk("%s status=%d\n", __func__, status);
610 * TODO: cb_sequence should support referring call lists, cachethis, multiple
611 * slots, and mark callback channel down on communication errors.
613 static void nfsd4_cb_prepare(struct rpc_task *task, void *calldata)
615 struct nfsd4_callback *cb = calldata;
616 struct nfs4_delegation *dp = container_of(cb, struct nfs4_delegation, dl_recall);
617 struct nfs4_client *clp = dp->dl_client;
618 u32 minorversion = clp->cl_cb_conn.cb_minorversion;
621 cb->cb_minorversion = minorversion;
623 status = nfsd41_cb_setup_sequence(clp, task);
625 if (status != -EAGAIN) {
626 /* terminate rpc task */
627 task->tk_status = status;
628 task->tk_action = NULL;
633 rpc_call_start(task);
636 static void nfsd4_cb_done(struct rpc_task *task, void *calldata)
638 struct nfsd4_callback *cb = calldata;
639 struct nfs4_delegation *dp = container_of(cb, struct nfs4_delegation, dl_recall);
640 struct nfs4_client *clp = dp->dl_client;
642 dprintk("%s: minorversion=%d\n", __func__,
643 clp->cl_cb_conn.cb_minorversion);
645 if (clp->cl_cb_conn.cb_minorversion) {
646 /* No need for lock, access serialized in nfsd4_cb_prepare */
648 clear_bit(0, &clp->cl_cb_slot_busy);
649 rpc_wake_up_next(&clp->cl_cb_waitq);
650 dprintk("%s: freed slot, new seqid=%d\n", __func__,
653 /* We're done looking into the sequence information */
654 task->tk_msg.rpc_resp = NULL;
659 static void nfsd4_cb_recall_done(struct rpc_task *task, void *calldata)
661 struct nfsd4_callback *cb = calldata;
662 struct nfs4_delegation *dp = container_of(cb, struct nfs4_delegation, dl_recall);
663 struct nfs4_client *clp = dp->dl_client;
664 struct rpc_clnt *current_rpc_client = clp->cl_cb_client;
666 nfsd4_cb_done(task, calldata);
668 if (current_rpc_client == NULL) {
669 /* We're shutting down; give up. */
670 /* XXX: err, or is it ok just to fall through
671 * and rpc_restart_call? */
675 switch (task->tk_status) {
679 case -NFS4ERR_BAD_STATEID:
680 /* Race: client probably got cb_recall
681 * before open reply granting delegation */
684 /* Network partition? */
685 atomic_set(&clp->cl_cb_set, 0);
686 warn_no_callback_path(clp, task->tk_status);
687 if (current_rpc_client != task->tk_client) {
688 /* queue a callback on the new connection: */
689 atomic_inc(&dp->dl_count);
694 if (dp->dl_retries--) {
695 rpc_delay(task, 2*HZ);
697 rpc_restart_call_prepare(task);
700 atomic_set(&clp->cl_cb_set, 0);
701 warn_no_callback_path(clp, task->tk_status);
705 static void nfsd4_cb_recall_release(void *calldata)
707 struct nfsd4_callback *cb = calldata;
708 struct nfs4_delegation *dp = container_of(cb, struct nfs4_delegation, dl_recall);
710 nfs4_put_delegation(dp);
713 static const struct rpc_call_ops nfsd4_cb_recall_ops = {
714 .rpc_call_prepare = nfsd4_cb_prepare,
715 .rpc_call_done = nfsd4_cb_recall_done,
716 .rpc_release = nfsd4_cb_recall_release,
719 int nfsd4_create_callback_queue(void)
721 callback_wq = create_singlethread_workqueue("nfsd4_callbacks");
727 void nfsd4_destroy_callback_queue(void)
729 destroy_workqueue(callback_wq);
732 /* must be called under the state lock */
733 void nfsd4_set_callback_client(struct nfs4_client *clp, struct rpc_clnt *new)
735 struct rpc_clnt *old = clp->cl_cb_client;
737 clp->cl_cb_client = new;
739 * After this, any work that saw the old value of cl_cb_client will
742 flush_workqueue(callback_wq);
743 /* So we can safely shut it down: */
745 rpc_shutdown_client(old);
748 void nfsd4_release_cb(struct nfsd4_callback *cb)
750 if (cb->cb_ops->rpc_release)
751 cb->cb_ops->rpc_release(cb);
754 void nfsd4_do_callback_rpc(struct work_struct *w)
756 struct nfsd4_callback *cb = container_of(w, struct nfsd4_callback, cb_work);
757 struct nfs4_client *clp = cb->cb_clp;
758 struct rpc_clnt *clnt = clp->cl_cb_client;
761 nfsd4_release_cb(cb);
762 return; /* Client is shutting down; give up. */
764 rpc_call_async(clnt, &cb->cb_msg, RPC_TASK_SOFT | RPC_TASK_SOFTCONN,
768 void nfsd4_cb_recall(struct nfs4_delegation *dp)
770 struct nfsd4_callback *cb = &dp->dl_recall;
774 cb->cb_clp = dp->dl_client;
775 cb->cb_msg.rpc_proc = &nfs4_cb_procedures[NFSPROC4_CLNT_CB_RECALL];
776 cb->cb_msg.rpc_argp = cb;
777 cb->cb_msg.rpc_resp = cb;
778 cb->cb_msg.rpc_cred = callback_cred;
780 cb->cb_ops = &nfsd4_cb_recall_ops;
783 queue_work(callback_wq, &dp->dl_recall.cb_work);