nfsd4: probe callback channel only once
[linux-2.6-block.git] / fs / nfsd / nfs4callback.c
CommitLineData
1da177e4
LT
1/*
2 * linux/fs/nfsd/nfs4callback.c
3 *
4 * Copyright (c) 2001 The Regents of the University of Michigan.
5 * All rights reserved.
6 *
7 * Kendrick Smith <kmsmith@umich.edu>
8 * Andy Adamson <andros@umich.edu>
9 *
10 * Redistribution and use in source and binary forms, with or without
11 * modification, are permitted provided that the following conditions
12 * are met:
13 *
14 * 1. Redistributions of source code must retain the above copyright
15 * notice, this list of conditions and the following disclaimer.
16 * 2. Redistributions in binary form must reproduce the above copyright
17 * notice, this list of conditions and the following disclaimer in the
18 * documentation and/or other materials provided with the distribution.
19 * 3. Neither the name of the University nor the names of its
20 * contributors may be used to endorse or promote products derived
21 * from this software without specific prior written permission.
22 *
23 * THIS SOFTWARE IS PROVIDED ``AS IS'' AND ANY EXPRESS OR IMPLIED
24 * WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
25 * MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
26 * DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
27 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
28 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
29 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR
30 * BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
31 * LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
32 * NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
33 * SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
34 */
35
1da177e4
LT
36#include <linux/module.h>
37#include <linux/list.h>
38#include <linux/inet.h>
39#include <linux/errno.h>
40#include <linux/delay.h>
e8edc6e0 41#include <linux/sched.h>
2b47eece 42#include <linux/kthread.h>
1da177e4
LT
43#include <linux/sunrpc/xdr.h>
44#include <linux/sunrpc/svc.h>
45#include <linux/sunrpc/clnt.h>
46#include <linux/nfsd/nfsd.h>
47#include <linux/nfsd/state.h>
48#include <linux/sunrpc/sched.h>
49#include <linux/nfs4.h>
50
51#define NFSDDBG_FACILITY NFSDDBG_PROC
52
53#define NFSPROC4_CB_NULL 0
54#define NFSPROC4_CB_COMPOUND 1
55
56/* declarations */
963d8fe5 57static const struct rpc_call_ops nfs4_cb_null_ops;
1da177e4
LT
58
59/* Index of predefined Linux callback client operations */
60
61enum {
62 NFSPROC4_CLNT_CB_NULL = 0,
63 NFSPROC4_CLNT_CB_RECALL,
64};
65
66enum nfs_cb_opnum4 {
67 OP_CB_RECALL = 4,
68};
69
70#define NFS4_MAXTAGLEN 20
71
72#define NFS4_enc_cb_null_sz 0
73#define NFS4_dec_cb_null_sz 0
74#define cb_compound_enc_hdr_sz 4
75#define cb_compound_dec_hdr_sz (3 + (NFS4_MAXTAGLEN >> 2))
76#define op_enc_sz 1
77#define op_dec_sz 2
78#define enc_nfs4_fh_sz (1 + (NFS4_FHSIZE >> 2))
0ac68d17 79#define enc_stateid_sz (NFS4_STATEID_SIZE >> 2)
1da177e4
LT
80#define NFS4_enc_cb_recall_sz (cb_compound_enc_hdr_sz + \
81 1 + enc_stateid_sz + \
82 enc_nfs4_fh_sz)
83
84#define NFS4_dec_cb_recall_sz (cb_compound_dec_hdr_sz + \
85 op_dec_sz)
86
87/*
88* Generic encode routines from fs/nfs/nfs4xdr.c
89*/
f00f328f
AV
90static inline __be32 *
91xdr_writemem(__be32 *p, const void *ptr, int nbytes)
1da177e4
LT
92{
93 int tmp = XDR_QUADLEN(nbytes);
94 if (!tmp)
95 return p;
96 p[tmp-1] = 0;
97 memcpy(p, ptr, nbytes);
98 return p + tmp;
99}
100
101#define WRITE32(n) *p++ = htonl(n)
102#define WRITEMEM(ptr,nbytes) do { \
103 p = xdr_writemem(p, ptr, nbytes); \
104} while (0)
105#define RESERVE_SPACE(nbytes) do { \
106 p = xdr_reserve_space(xdr, nbytes); \
107 if (!p) dprintk("NFSD: RESERVE_SPACE(%d) failed in function %s\n", (int) (nbytes), __FUNCTION__); \
108 BUG_ON(!p); \
109} while (0)
110
111/*
112 * Generic decode routines from fs/nfs/nfs4xdr.c
113 */
114#define DECODE_TAIL \
115 status = 0; \
116out: \
117 return status; \
118xdr_error: \
119 dprintk("NFSD: xdr error! (%s:%d)\n", __FILE__, __LINE__); \
120 status = -EIO; \
121 goto out
122
123#define READ32(x) (x) = ntohl(*p++)
124#define READ64(x) do { \
125 (x) = (u64)ntohl(*p++) << 32; \
126 (x) |= ntohl(*p++); \
127} while (0)
128#define READTIME(x) do { \
129 p++; \
130 (x.tv_sec) = ntohl(*p++); \
131 (x.tv_nsec) = ntohl(*p++); \
132} while (0)
133#define READ_BUF(nbytes) do { \
134 p = xdr_inline_decode(xdr, nbytes); \
135 if (!p) { \
3e3b4800 136 dprintk("NFSD: %s: reply buffer overflowed in line %d.\n", \
1da177e4
LT
137 __FUNCTION__, __LINE__); \
138 return -EIO; \
139 } \
140} while (0)
141
142struct nfs4_cb_compound_hdr {
143 int status;
144 u32 ident;
145 u32 nops;
146 u32 taglen;
147 char * tag;
148};
149
150static struct {
151int stat;
152int errno;
153} nfs_cb_errtbl[] = {
154 { NFS4_OK, 0 },
155 { NFS4ERR_PERM, EPERM },
156 { NFS4ERR_NOENT, ENOENT },
157 { NFS4ERR_IO, EIO },
158 { NFS4ERR_NXIO, ENXIO },
159 { NFS4ERR_ACCESS, EACCES },
160 { NFS4ERR_EXIST, EEXIST },
161 { NFS4ERR_XDEV, EXDEV },
162 { NFS4ERR_NOTDIR, ENOTDIR },
163 { NFS4ERR_ISDIR, EISDIR },
164 { NFS4ERR_INVAL, EINVAL },
165 { NFS4ERR_FBIG, EFBIG },
166 { NFS4ERR_NOSPC, ENOSPC },
167 { NFS4ERR_ROFS, EROFS },
168 { NFS4ERR_MLINK, EMLINK },
169 { NFS4ERR_NAMETOOLONG, ENAMETOOLONG },
170 { NFS4ERR_NOTEMPTY, ENOTEMPTY },
171 { NFS4ERR_DQUOT, EDQUOT },
172 { NFS4ERR_STALE, ESTALE },
173 { NFS4ERR_BADHANDLE, EBADHANDLE },
174 { NFS4ERR_BAD_COOKIE, EBADCOOKIE },
175 { NFS4ERR_NOTSUPP, ENOTSUPP },
176 { NFS4ERR_TOOSMALL, ETOOSMALL },
177 { NFS4ERR_SERVERFAULT, ESERVERFAULT },
178 { NFS4ERR_BADTYPE, EBADTYPE },
179 { NFS4ERR_LOCKED, EAGAIN },
180 { NFS4ERR_RESOURCE, EREMOTEIO },
181 { NFS4ERR_SYMLINK, ELOOP },
182 { NFS4ERR_OP_ILLEGAL, EOPNOTSUPP },
183 { NFS4ERR_DEADLOCK, EDEADLK },
184 { -1, EIO }
185};
186
187static int
188nfs_cb_stat_to_errno(int stat)
189{
190 int i;
191 for (i = 0; nfs_cb_errtbl[i].stat != -1; i++) {
192 if (nfs_cb_errtbl[i].stat == stat)
193 return nfs_cb_errtbl[i].errno;
194 }
195 /* If we cannot translate the error, the recovery routines should
196 * handle it.
197 * Note: remaining NFSv4 error codes have values > 10000, so should
198 * not conflict with native Linux error codes.
199 */
200 return stat;
201}
202
203/*
204 * XDR encode
205 */
206
207static int
208encode_cb_compound_hdr(struct xdr_stream *xdr, struct nfs4_cb_compound_hdr *hdr)
209{
f00f328f 210 __be32 * p;
1da177e4
LT
211
212 RESERVE_SPACE(16);
213 WRITE32(0); /* tag length is always 0 */
214 WRITE32(NFS4_MINOR_VERSION);
215 WRITE32(hdr->ident);
216 WRITE32(hdr->nops);
217 return 0;
218}
219
220static int
221encode_cb_recall(struct xdr_stream *xdr, struct nfs4_cb_recall *cb_rec)
222{
f00f328f 223 __be32 *p;
1da177e4
LT
224 int len = cb_rec->cbr_fhlen;
225
226 RESERVE_SPACE(12+sizeof(cb_rec->cbr_stateid) + len);
227 WRITE32(OP_CB_RECALL);
228 WRITEMEM(&cb_rec->cbr_stateid, sizeof(stateid_t));
229 WRITE32(cb_rec->cbr_trunc);
230 WRITE32(len);
231 WRITEMEM(cb_rec->cbr_fhval, len);
232 return 0;
233}
234
235static int
f00f328f 236nfs4_xdr_enc_cb_null(struct rpc_rqst *req, __be32 *p)
1da177e4
LT
237{
238 struct xdr_stream xdrs, *xdr = &xdrs;
239
240 xdr_init_encode(&xdrs, &req->rq_snd_buf, p);
241 RESERVE_SPACE(0);
242 return 0;
243}
244
245static int
f00f328f 246nfs4_xdr_enc_cb_recall(struct rpc_rqst *req, __be32 *p, struct nfs4_cb_recall *args)
1da177e4
LT
247{
248 struct xdr_stream xdr;
249 struct nfs4_cb_compound_hdr hdr = {
250 .ident = args->cbr_ident,
251 .nops = 1,
252 };
253
254 xdr_init_encode(&xdr, &req->rq_snd_buf, p);
255 encode_cb_compound_hdr(&xdr, &hdr);
256 return (encode_cb_recall(&xdr, args));
257}
258
259
260static int
261decode_cb_compound_hdr(struct xdr_stream *xdr, struct nfs4_cb_compound_hdr *hdr){
f00f328f 262 __be32 *p;
1da177e4
LT
263
264 READ_BUF(8);
265 READ32(hdr->status);
266 READ32(hdr->taglen);
267 READ_BUF(hdr->taglen + 4);
268 hdr->tag = (char *)p;
269 p += XDR_QUADLEN(hdr->taglen);
270 READ32(hdr->nops);
271 return 0;
272}
273
274static int
275decode_cb_op_hdr(struct xdr_stream *xdr, enum nfs_opnum4 expected)
276{
f00f328f 277 __be32 *p;
1da177e4
LT
278 u32 op;
279 int32_t nfserr;
280
281 READ_BUF(8);
282 READ32(op);
283 if (op != expected) {
284 dprintk("NFSD: decode_cb_op_hdr: Callback server returned "
285 " operation %d but we issued a request for %d\n",
286 op, expected);
287 return -EIO;
288 }
289 READ32(nfserr);
290 if (nfserr != NFS_OK)
291 return -nfs_cb_stat_to_errno(nfserr);
292 return 0;
293}
294
295static int
f00f328f 296nfs4_xdr_dec_cb_null(struct rpc_rqst *req, __be32 *p)
1da177e4
LT
297{
298 return 0;
299}
300
301static int
f00f328f 302nfs4_xdr_dec_cb_recall(struct rpc_rqst *rqstp, __be32 *p)
1da177e4
LT
303{
304 struct xdr_stream xdr;
305 struct nfs4_cb_compound_hdr hdr;
306 int status;
307
308 xdr_init_decode(&xdr, &rqstp->rq_rcv_buf, p);
309 status = decode_cb_compound_hdr(&xdr, &hdr);
310 if (status)
311 goto out;
312 status = decode_cb_op_hdr(&xdr, OP_CB_RECALL);
313out:
314 return status;
315}
316
317/*
318 * RPC procedure tables
319 */
1da177e4
LT
320#define PROC(proc, call, argtype, restype) \
321[NFSPROC4_CLNT_##proc] = { \
322 .p_proc = NFSPROC4_CB_##call, \
323 .p_encode = (kxdrproc_t) nfs4_xdr_##argtype, \
324 .p_decode = (kxdrproc_t) nfs4_xdr_##restype, \
2bea90d4
CL
325 .p_arglen = NFS4_##argtype##_sz, \
326 .p_replen = NFS4_##restype##_sz, \
cc0175c1
CL
327 .p_statidx = NFSPROC4_CB_##call, \
328 .p_name = #proc, \
1da177e4
LT
329}
330
fd39ca9a 331static struct rpc_procinfo nfs4_cb_procedures[] = {
1da177e4
LT
332 PROC(CB_NULL, NULL, enc_cb_null, dec_cb_null),
333 PROC(CB_RECALL, COMPOUND, enc_cb_recall, dec_cb_recall),
334};
335
fd39ca9a 336static struct rpc_version nfs_cb_version4 = {
1da177e4 337 .number = 1,
e8c96f8c 338 .nrprocs = ARRAY_SIZE(nfs4_cb_procedures),
1da177e4
LT
339 .procs = nfs4_cb_procedures
340};
341
342static struct rpc_version * nfs_cb_version[] = {
343 NULL,
344 &nfs_cb_version4,
345};
346
2b47eece
BF
347/* Reference counting, callback cleanup, etc., all look racy as heck.
348 * And why is cb_set an atomic? */
349
350static int do_probe_callback(void *data)
351{
352 struct nfs4_client *clp = data;
353 struct nfs4_callback *cb = &clp->cl_callback;
354 struct rpc_message msg = {
355 .rpc_proc = &nfs4_cb_procedures[NFSPROC4_CLNT_CB_NULL],
356 .rpc_argp = clp,
357 };
358 int status;
359
2b47eece 360 status = rpc_call_sync(cb->cb_client, &msg, RPC_TASK_SOFT);
2b47eece
BF
361
362 if (status) {
363 rpc_shutdown_client(cb->cb_client);
364 cb->cb_client = NULL;
365 } else
366 atomic_set(&cb->cb_set, 1);
2b47eece
BF
367 put_nfs4_client(clp);
368 return 0;
369}
370
1da177e4
LT
371/*
372 * Set up the callback client and put a NFSPROC4_CB_NULL on the wire...
373 */
374void
375nfsd4_probe_callback(struct nfs4_client *clp)
376{
377 struct sockaddr_in addr;
378 struct nfs4_callback *cb = &clp->cl_callback;
ae5c7947
CL
379 struct rpc_timeout timeparms = {
380 .to_initval = (NFSD_LEASE_TIME/4) * HZ,
381 .to_retries = 5,
382 .to_maxval = (NFSD_LEASE_TIME/2) * HZ,
383 .to_exponential = 1,
384 };
1da177e4 385 struct rpc_program * program = &cb->cb_program;
ae5c7947
CL
386 struct rpc_create_args args = {
387 .protocol = IPPROTO_TCP,
388 .address = (struct sockaddr *)&addr,
389 .addrsize = sizeof(addr),
390 .timeout = &timeparms,
ae5c7947
CL
391 .program = program,
392 .version = nfs_cb_version[1]->number,
393 .authflavor = RPC_AUTH_UNIX, /* XXX: need AUTH_GSS... */
394 .flags = (RPC_CLNT_CREATE_NOPING),
395 };
2b47eece 396 struct task_struct *t;
1da177e4 397
46f8a64b 398 BUG_ON(atomic_read(&clp->cl_callback.cb_set));
1da177e4
LT
399
400 /* Initialize address */
401 memset(&addr, 0, sizeof(addr));
402 addr.sin_family = AF_INET;
403 addr.sin_port = htons(cb->cb_port);
404 addr.sin_addr.s_addr = htonl(cb->cb_addr);
405
1da177e4
LT
406 /* Initialize rpc_program */
407 program->name = "nfs4_cb";
408 program->number = cb->cb_prog;
e8c96f8c 409 program->nrvers = ARRAY_SIZE(nfs_cb_version);
1da177e4 410 program->version = nfs_cb_version;
ae5c7947 411 program->stats = &cb->cb_stat;
1da177e4
LT
412
413 /* Initialize rpc_stat */
ae5c7947
CL
414 memset(program->stats, 0, sizeof(cb->cb_stat));
415 program->stats->program = program;
416
417 /* Create RPC client */
418 cb->cb_client = rpc_create(&args);
0942176f 419 if (IS_ERR(cb->cb_client)) {
1da177e4 420 dprintk("NFSD: couldn't create callback client\n");
5b616f5d 421 goto out_err;
1da177e4 422 }
1da177e4 423
541e0e09 424 /* the task holds a reference to the nfs4_client struct */
1da177e4
LT
425 atomic_inc(&clp->cl_count);
426
2b47eece 427 t = kthread_run(do_probe_callback, clp, "nfs4_cb_probe");
1da177e4 428
2b47eece 429 if (IS_ERR(t))
f61534df 430 goto out_release_clp;
2b47eece 431
1da177e4
LT
432 return;
433
f61534df 434out_release_clp:
1da177e4 435 atomic_dec(&clp->cl_count);
ae5c7947 436 rpc_shutdown_client(cb->cb_client);
1da177e4 437out_err:
0942176f 438 cb->cb_client = NULL;
1da177e4
LT
439 dprintk("NFSD: warning: no callback path to client %.*s\n",
440 (int)clp->cl_name.len, clp->cl_name.data);
1da177e4
LT
441}
442
1da177e4
LT
443/*
444 * called with dp->dl_count inc'ed.
445 * nfs4_lock_state() may or may not have been called.
446 */
447void
448nfsd4_cb_recall(struct nfs4_delegation *dp)
449{
450 struct nfs4_client *clp = dp->dl_client;
451 struct rpc_clnt *clnt = clp->cl_callback.cb_client;
452 struct nfs4_cb_recall *cbr = &dp->dl_recall;
453 struct rpc_message msg = {
454 .rpc_proc = &nfs4_cb_procedures[NFSPROC4_CLNT_CB_RECALL],
455 .rpc_argp = cbr,
456 };
457 int retries = 1;
458 int status = 0;
459
460 if ((!atomic_read(&clp->cl_callback.cb_set)) || !clnt)
461 return;
462
1da177e4
LT
463 cbr->cbr_trunc = 0; /* XXX need to implement truncate optimization */
464 cbr->cbr_dp = dp;
465
466 status = rpc_call_sync(clnt, &msg, RPC_TASK_SOFT);
467 while (retries--) {
468 switch (status) {
469 case -EIO:
470 /* Network partition? */
471 case -EBADHANDLE:
472 case -NFS4ERR_BAD_STATEID:
473 /* Race: client probably got cb_recall
474 * before open reply granting delegation */
475 break;
476 default:
477 goto out_put_cred;
478 }
479 ssleep(2);
480 status = rpc_call_sync(clnt, &msg, RPC_TASK_SOFT);
481 }
482out_put_cred:
1da177e4
LT
483 if (status == -EIO)
484 atomic_set(&clp->cl_callback.cb_set, 0);
485 /* Success or failure, now we're either waiting for lease expiration
486 * or deleg_return. */
487 dprintk("NFSD: nfs4_cb_recall: dp %p dl_flock %p dl_count %d\n",dp, dp->dl_flock, atomic_read(&dp->dl_count));
cfdcad4d 488 put_nfs4_client(clp);
1da177e4
LT
489 nfs4_put_delegation(dp);
490 return;
491}