1 /* 2 * Copyright (c) 2001 The Regents of the University of Michigan. 3 * All rights reserved. 4 * 5 * Kendrick Smith <kmsmith@umich.edu> 6 * Andy Adamson <andros@umich.edu> 7 * 8 * Redistribution and use in source and binary forms, with or without 9 * modification, are permitted provided that the following conditions 10 * are met: 11 * 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. 20 * 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. 32 */ 33 34 #include <linux/sunrpc/clnt.h> 35 #include <linux/sunrpc/xprt.h> 36 #include <linux/sunrpc/svc_xprt.h> 37 #include <linux/slab.h> 38 #include "nfsd.h" 39 #include "state.h" 40 #include "netns.h" 41 #include "xdr4cb.h" 42 #include "xdr4.h" 43 44 #define NFSDDBG_FACILITY NFSDDBG_PROC 45 46 static void nfsd4_mark_cb_fault(struct nfs4_client *, int reason); 47 48 #define NFSPROC4_CB_NULL 0 49 #define NFSPROC4_CB_COMPOUND 1 50 51 /* Index of predefined Linux callback client operations */ 52 53 struct nfs4_cb_compound_hdr { 54 /* args */ 55 u32 ident; /* minorversion 0 only */ 56 u32 nops; 57 __be32 *nops_p; 58 u32 minorversion; 59 /* res */ 60 int status; 61 }; 62 63 static __be32 *xdr_encode_empty_array(__be32 *p) 64 { 65 *p++ = xdr_zero; 66 return p; 67 } 68 69 /* 70 * Encode/decode NFSv4 CB basic data types 71 * 72 * Basic NFSv4 callback data types are defined in section 15 of RFC 73 * 3530: "Network File System (NFS) version 4 Protocol" and section 74 * 20 of RFC 5661: "Network File System (NFS) Version 4 Minor Version 75 * 1 Protocol" 76 */ 77 78 /* 79 * nfs_cb_opnum4 80 * 81 * enum nfs_cb_opnum4 { 82 * OP_CB_GETATTR = 3, 83 * ... 84 * }; 85 */ 86 enum nfs_cb_opnum4 { 87 OP_CB_GETATTR = 3, 88 OP_CB_RECALL = 4, 89 OP_CB_LAYOUTRECALL = 5, 90 OP_CB_NOTIFY = 6, 91 OP_CB_PUSH_DELEG = 7, 92 OP_CB_RECALL_ANY = 8, 93 OP_CB_RECALLABLE_OBJ_AVAIL = 9, 94 OP_CB_RECALL_SLOT = 10, 95 OP_CB_SEQUENCE = 11, 96 OP_CB_WANTS_CANCELLED = 12, 97 OP_CB_NOTIFY_LOCK = 13, 98 OP_CB_NOTIFY_DEVICEID = 14, 99 OP_CB_OFFLOAD = 15, 100 OP_CB_ILLEGAL = 10044 101 }; 102 103 static void encode_nfs_cb_opnum4(struct xdr_stream *xdr, enum nfs_cb_opnum4 op) 104 { 105 __be32 *p; 106 107 p = xdr_reserve_space(xdr, 4); 108 *p = cpu_to_be32(op); 109 } 110 111 /* 112 * nfs_fh4 113 * 114 * typedef opaque nfs_fh4<NFS4_FHSIZE>; 115 */ 116 static void encode_nfs_fh4(struct xdr_stream *xdr, const struct knfsd_fh *fh) 117 { 118 u32 length = fh->fh_size; 119 __be32 *p; 120 121 BUG_ON(length > NFS4_FHSIZE); 122 p = xdr_reserve_space(xdr, 4 + length); 123 xdr_encode_opaque(p, &fh->fh_base, length); 124 } 125 126 /* 127 * stateid4 128 * 129 * struct stateid4 { 130 * uint32_t seqid; 131 * opaque other[12]; 132 * }; 133 */ 134 static void encode_stateid4(struct xdr_stream *xdr, const stateid_t *sid) 135 { 136 __be32 *p; 137 138 p = xdr_reserve_space(xdr, NFS4_STATEID_SIZE); 139 *p++ = cpu_to_be32(sid->si_generation); 140 xdr_encode_opaque_fixed(p, &sid->si_opaque, NFS4_STATEID_OTHER_SIZE); 141 } 142 143 /* 144 * sessionid4 145 * 146 * typedef opaque sessionid4[NFS4_SESSIONID_SIZE]; 147 */ 148 static void encode_sessionid4(struct xdr_stream *xdr, 149 const struct nfsd4_session *session) 150 { 151 __be32 *p; 152 153 p = xdr_reserve_space(xdr, NFS4_MAX_SESSIONID_LEN); 154 xdr_encode_opaque_fixed(p, session->se_sessionid.data, 155 NFS4_MAX_SESSIONID_LEN); 156 } 157 158 /* 159 * nfsstat4 160 */ 161 static const struct { 162 int stat; 163 int errno; 164 } nfs_cb_errtbl[] = { 165 { NFS4_OK, 0 }, 166 { NFS4ERR_PERM, -EPERM }, 167 { NFS4ERR_NOENT, -ENOENT }, 168 { NFS4ERR_IO, -EIO }, 169 { NFS4ERR_NXIO, -ENXIO }, 170 { NFS4ERR_ACCESS, -EACCES }, 171 { NFS4ERR_EXIST, -EEXIST }, 172 { NFS4ERR_XDEV, -EXDEV }, 173 { NFS4ERR_NOTDIR, -ENOTDIR }, 174 { NFS4ERR_ISDIR, -EISDIR }, 175 { NFS4ERR_INVAL, -EINVAL }, 176 { NFS4ERR_FBIG, -EFBIG }, 177 { NFS4ERR_NOSPC, -ENOSPC }, 178 { NFS4ERR_ROFS, -EROFS }, 179 { NFS4ERR_MLINK, -EMLINK }, 180 { NFS4ERR_NAMETOOLONG, -ENAMETOOLONG }, 181 { NFS4ERR_NOTEMPTY, -ENOTEMPTY }, 182 { NFS4ERR_DQUOT, -EDQUOT }, 183 { NFS4ERR_STALE, -ESTALE }, 184 { NFS4ERR_BADHANDLE, -EBADHANDLE }, 185 { NFS4ERR_BAD_COOKIE, -EBADCOOKIE }, 186 { NFS4ERR_NOTSUPP, -ENOTSUPP }, 187 { NFS4ERR_TOOSMALL, -ETOOSMALL }, 188 { NFS4ERR_SERVERFAULT, -ESERVERFAULT }, 189 { NFS4ERR_BADTYPE, -EBADTYPE }, 190 { NFS4ERR_LOCKED, -EAGAIN }, 191 { NFS4ERR_RESOURCE, -EREMOTEIO }, 192 { NFS4ERR_SYMLINK, -ELOOP }, 193 { NFS4ERR_OP_ILLEGAL, -EOPNOTSUPP }, 194 { NFS4ERR_DEADLOCK, -EDEADLK }, 195 { -1, -EIO } 196 }; 197 198 /* 199 * If we cannot translate the error, the recovery routines should 200 * handle it. 201 * 202 * Note: remaining NFSv4 error codes have values > 10000, so should 203 * not conflict with native Linux error codes. 204 */ 205 static int nfs_cb_stat_to_errno(int status) 206 { 207 int i; 208 209 for (i = 0; nfs_cb_errtbl[i].stat != -1; i++) { 210 if (nfs_cb_errtbl[i].stat == status) 211 return nfs_cb_errtbl[i].errno; 212 } 213 214 dprintk("NFSD: Unrecognized NFS CB status value: %u\n", status); 215 return -status; 216 } 217 218 static int decode_cb_op_status(struct xdr_stream *xdr, 219 enum nfs_cb_opnum4 expected, int *status) 220 { 221 __be32 *p; 222 u32 op; 223 224 p = xdr_inline_decode(xdr, 4 + 4); 225 if (unlikely(p == NULL)) 226 goto out_overflow; 227 op = be32_to_cpup(p++); 228 if (unlikely(op != expected)) 229 goto out_unexpected; 230 *status = nfs_cb_stat_to_errno(be32_to_cpup(p)); 231 return 0; 232 out_overflow: 233 return -EIO; 234 out_unexpected: 235 dprintk("NFSD: Callback server returned operation %d but " 236 "we issued a request for %d\n", op, expected); 237 return -EIO; 238 } 239 240 /* 241 * CB_COMPOUND4args 242 * 243 * struct CB_COMPOUND4args { 244 * utf8str_cs tag; 245 * uint32_t minorversion; 246 * uint32_t callback_ident; 247 * nfs_cb_argop4 argarray<>; 248 * }; 249 */ 250 static void encode_cb_compound4args(struct xdr_stream *xdr, 251 struct nfs4_cb_compound_hdr *hdr) 252 { 253 __be32 * p; 254 255 p = xdr_reserve_space(xdr, 4 + 4 + 4 + 4); 256 p = xdr_encode_empty_array(p); /* empty tag */ 257 *p++ = cpu_to_be32(hdr->minorversion); 258 *p++ = cpu_to_be32(hdr->ident); 259 260 hdr->nops_p = p; 261 *p = cpu_to_be32(hdr->nops); /* argarray element count */ 262 } 263 264 /* 265 * Update argarray element count 266 */ 267 static void encode_cb_nops(struct nfs4_cb_compound_hdr *hdr) 268 { 269 BUG_ON(hdr->nops > NFS4_MAX_BACK_CHANNEL_OPS); 270 *hdr->nops_p = cpu_to_be32(hdr->nops); 271 } 272 273 /* 274 * CB_COMPOUND4res 275 * 276 * struct CB_COMPOUND4res { 277 * nfsstat4 status; 278 * utf8str_cs tag; 279 * nfs_cb_resop4 resarray<>; 280 * }; 281 */ 282 static int decode_cb_compound4res(struct xdr_stream *xdr, 283 struct nfs4_cb_compound_hdr *hdr) 284 { 285 u32 length; 286 __be32 *p; 287 288 p = xdr_inline_decode(xdr, 4 + 4); 289 if (unlikely(p == NULL)) 290 goto out_overflow; 291 hdr->status = be32_to_cpup(p++); 292 /* Ignore the tag */ 293 length = be32_to_cpup(p++); 294 p = xdr_inline_decode(xdr, length + 4); 295 if (unlikely(p == NULL)) 296 goto out_overflow; 297 p += XDR_QUADLEN(length); 298 hdr->nops = be32_to_cpup(p); 299 return 0; 300 out_overflow: 301 return -EIO; 302 } 303 304 /* 305 * CB_RECALL4args 306 * 307 * struct CB_RECALL4args { 308 * stateid4 stateid; 309 * bool truncate; 310 * nfs_fh4 fh; 311 * }; 312 */ 313 static void encode_cb_recall4args(struct xdr_stream *xdr, 314 const struct nfs4_delegation *dp, 315 struct nfs4_cb_compound_hdr *hdr) 316 { 317 __be32 *p; 318 319 encode_nfs_cb_opnum4(xdr, OP_CB_RECALL); 320 encode_stateid4(xdr, &dp->dl_stid.sc_stateid); 321 322 p = xdr_reserve_space(xdr, 4); 323 *p++ = xdr_zero; /* truncate */ 324 325 encode_nfs_fh4(xdr, &dp->dl_stid.sc_file->fi_fhandle); 326 327 hdr->nops++; 328 } 329 330 /* 331 * CB_SEQUENCE4args 332 * 333 * struct CB_SEQUENCE4args { 334 * sessionid4 csa_sessionid; 335 * sequenceid4 csa_sequenceid; 336 * slotid4 csa_slotid; 337 * slotid4 csa_highest_slotid; 338 * bool csa_cachethis; 339 * referring_call_list4 csa_referring_call_lists<>; 340 * }; 341 */ 342 static void encode_cb_sequence4args(struct xdr_stream *xdr, 343 const struct nfsd4_callback *cb, 344 struct nfs4_cb_compound_hdr *hdr) 345 { 346 struct nfsd4_session *session = cb->cb_clp->cl_cb_session; 347 __be32 *p; 348 349 if (hdr->minorversion == 0) 350 return; 351 352 encode_nfs_cb_opnum4(xdr, OP_CB_SEQUENCE); 353 encode_sessionid4(xdr, session); 354 355 p = xdr_reserve_space(xdr, 4 + 4 + 4 + 4 + 4); 356 *p++ = cpu_to_be32(session->se_cb_seq_nr); /* csa_sequenceid */ 357 *p++ = xdr_zero; /* csa_slotid */ 358 *p++ = xdr_zero; /* csa_highest_slotid */ 359 *p++ = xdr_zero; /* csa_cachethis */ 360 xdr_encode_empty_array(p); /* csa_referring_call_lists */ 361 362 hdr->nops++; 363 } 364 365 /* 366 * CB_SEQUENCE4resok 367 * 368 * struct CB_SEQUENCE4resok { 369 * sessionid4 csr_sessionid; 370 * sequenceid4 csr_sequenceid; 371 * slotid4 csr_slotid; 372 * slotid4 csr_highest_slotid; 373 * slotid4 csr_target_highest_slotid; 374 * }; 375 * 376 * union CB_SEQUENCE4res switch (nfsstat4 csr_status) { 377 * case NFS4_OK: 378 * CB_SEQUENCE4resok csr_resok4; 379 * default: 380 * void; 381 * }; 382 * 383 * Our current back channel implmentation supports a single backchannel 384 * with a single slot. 385 */ 386 static int decode_cb_sequence4resok(struct xdr_stream *xdr, 387 struct nfsd4_callback *cb) 388 { 389 struct nfsd4_session *session = cb->cb_clp->cl_cb_session; 390 int status = -ESERVERFAULT; 391 __be32 *p; 392 u32 dummy; 393 394 /* 395 * If the server returns different values for sessionID, slotID or 396 * sequence number, the server is looney tunes. 397 */ 398 p = xdr_inline_decode(xdr, NFS4_MAX_SESSIONID_LEN + 4 + 4 + 4 + 4); 399 if (unlikely(p == NULL)) 400 goto out_overflow; 401 402 if (memcmp(p, session->se_sessionid.data, NFS4_MAX_SESSIONID_LEN)) { 403 dprintk("NFS: %s Invalid session id\n", __func__); 404 goto out; 405 } 406 p += XDR_QUADLEN(NFS4_MAX_SESSIONID_LEN); 407 408 dummy = be32_to_cpup(p++); 409 if (dummy != session->se_cb_seq_nr) { 410 dprintk("NFS: %s Invalid sequence number\n", __func__); 411 goto out; 412 } 413 414 dummy = be32_to_cpup(p++); 415 if (dummy != 0) { 416 dprintk("NFS: %s Invalid slotid\n", __func__); 417 goto out; 418 } 419 420 /* 421 * FIXME: process highest slotid and target highest slotid 422 */ 423 status = 0; 424 out: 425 cb->cb_seq_status = status; 426 return status; 427 out_overflow: 428 status = -EIO; 429 goto out; 430 } 431 432 static int decode_cb_sequence4res(struct xdr_stream *xdr, 433 struct nfsd4_callback *cb) 434 { 435 int status; 436 437 if (cb->cb_clp->cl_minorversion == 0) 438 return 0; 439 440 status = decode_cb_op_status(xdr, OP_CB_SEQUENCE, &cb->cb_seq_status); 441 if (unlikely(status || cb->cb_seq_status)) 442 return status; 443 444 return decode_cb_sequence4resok(xdr, cb); 445 } 446 447 /* 448 * NFSv4.0 and NFSv4.1 XDR encode functions 449 * 450 * NFSv4.0 callback argument types are defined in section 15 of RFC 451 * 3530: "Network File System (NFS) version 4 Protocol" and section 20 452 * of RFC 5661: "Network File System (NFS) Version 4 Minor Version 1 453 * Protocol". 454 */ 455 456 /* 457 * NB: Without this zero space reservation, callbacks over krb5p fail 458 */ 459 static void nfs4_xdr_enc_cb_null(struct rpc_rqst *req, struct xdr_stream *xdr, 460 const void *__unused) 461 { 462 xdr_reserve_space(xdr, 0); 463 } 464 465 /* 466 * 20.2. Operation 4: CB_RECALL - Recall a Delegation 467 */ 468 static void nfs4_xdr_enc_cb_recall(struct rpc_rqst *req, struct xdr_stream *xdr, 469 const void *data) 470 { 471 const struct nfsd4_callback *cb = data; 472 const struct nfs4_delegation *dp = cb_to_delegation(cb); 473 struct nfs4_cb_compound_hdr hdr = { 474 .ident = cb->cb_clp->cl_cb_ident, 475 .minorversion = cb->cb_clp->cl_minorversion, 476 }; 477 478 encode_cb_compound4args(xdr, &hdr); 479 encode_cb_sequence4args(xdr, cb, &hdr); 480 encode_cb_recall4args(xdr, dp, &hdr); 481 encode_cb_nops(&hdr); 482 } 483 484 485 /* 486 * NFSv4.0 and NFSv4.1 XDR decode functions 487 * 488 * NFSv4.0 callback result types are defined in section 15 of RFC 489 * 3530: "Network File System (NFS) version 4 Protocol" and section 20 490 * of RFC 5661: "Network File System (NFS) Version 4 Minor Version 1 491 * Protocol". 492 */ 493 494 static int nfs4_xdr_dec_cb_null(struct rpc_rqst *req, struct xdr_stream *xdr, 495 void *__unused) 496 { 497 return 0; 498 } 499 500 /* 501 * 20.2. Operation 4: CB_RECALL - Recall a Delegation 502 */ 503 static int nfs4_xdr_dec_cb_recall(struct rpc_rqst *rqstp, 504 struct xdr_stream *xdr, 505 void *data) 506 { 507 struct nfsd4_callback *cb = data; 508 struct nfs4_cb_compound_hdr hdr; 509 int status; 510 511 status = decode_cb_compound4res(xdr, &hdr); 512 if (unlikely(status)) 513 return status; 514 515 if (cb != NULL) { 516 status = decode_cb_sequence4res(xdr, cb); 517 if (unlikely(status || cb->cb_seq_status)) 518 return status; 519 } 520 521 return decode_cb_op_status(xdr, OP_CB_RECALL, &cb->cb_status); 522 } 523 524 #ifdef CONFIG_NFSD_PNFS 525 /* 526 * CB_LAYOUTRECALL4args 527 * 528 * struct layoutrecall_file4 { 529 * nfs_fh4 lor_fh; 530 * offset4 lor_offset; 531 * length4 lor_length; 532 * stateid4 lor_stateid; 533 * }; 534 * 535 * union layoutrecall4 switch(layoutrecall_type4 lor_recalltype) { 536 * case LAYOUTRECALL4_FILE: 537 * layoutrecall_file4 lor_layout; 538 * case LAYOUTRECALL4_FSID: 539 * fsid4 lor_fsid; 540 * case LAYOUTRECALL4_ALL: 541 * void; 542 * }; 543 * 544 * struct CB_LAYOUTRECALL4args { 545 * layouttype4 clora_type; 546 * layoutiomode4 clora_iomode; 547 * bool clora_changed; 548 * layoutrecall4 clora_recall; 549 * }; 550 */ 551 static void encode_cb_layout4args(struct xdr_stream *xdr, 552 const struct nfs4_layout_stateid *ls, 553 struct nfs4_cb_compound_hdr *hdr) 554 { 555 __be32 *p; 556 557 BUG_ON(hdr->minorversion == 0); 558 559 p = xdr_reserve_space(xdr, 5 * 4); 560 *p++ = cpu_to_be32(OP_CB_LAYOUTRECALL); 561 *p++ = cpu_to_be32(ls->ls_layout_type); 562 *p++ = cpu_to_be32(IOMODE_ANY); 563 *p++ = cpu_to_be32(1); 564 *p = cpu_to_be32(RETURN_FILE); 565 566 encode_nfs_fh4(xdr, &ls->ls_stid.sc_file->fi_fhandle); 567 568 p = xdr_reserve_space(xdr, 2 * 8); 569 p = xdr_encode_hyper(p, 0); 570 xdr_encode_hyper(p, NFS4_MAX_UINT64); 571 572 encode_stateid4(xdr, &ls->ls_recall_sid); 573 574 hdr->nops++; 575 } 576 577 static void nfs4_xdr_enc_cb_layout(struct rpc_rqst *req, 578 struct xdr_stream *xdr, 579 const void *data) 580 { 581 const struct nfsd4_callback *cb = data; 582 const struct nfs4_layout_stateid *ls = 583 container_of(cb, struct nfs4_layout_stateid, ls_recall); 584 struct nfs4_cb_compound_hdr hdr = { 585 .ident = 0, 586 .minorversion = cb->cb_clp->cl_minorversion, 587 }; 588 589 encode_cb_compound4args(xdr, &hdr); 590 encode_cb_sequence4args(xdr, cb, &hdr); 591 encode_cb_layout4args(xdr, ls, &hdr); 592 encode_cb_nops(&hdr); 593 } 594 595 static int nfs4_xdr_dec_cb_layout(struct rpc_rqst *rqstp, 596 struct xdr_stream *xdr, 597 void *data) 598 { 599 struct nfsd4_callback *cb = data; 600 struct nfs4_cb_compound_hdr hdr; 601 int status; 602 603 status = decode_cb_compound4res(xdr, &hdr); 604 if (unlikely(status)) 605 return status; 606 607 if (cb) { 608 status = decode_cb_sequence4res(xdr, cb); 609 if (unlikely(status || cb->cb_seq_status)) 610 return status; 611 } 612 return decode_cb_op_status(xdr, OP_CB_LAYOUTRECALL, &cb->cb_status); 613 } 614 #endif /* CONFIG_NFSD_PNFS */ 615 616 static void encode_stateowner(struct xdr_stream *xdr, struct nfs4_stateowner *so) 617 { 618 __be32 *p; 619 620 p = xdr_reserve_space(xdr, 8 + 4 + so->so_owner.len); 621 p = xdr_encode_opaque_fixed(p, &so->so_client->cl_clientid, 8); 622 xdr_encode_opaque(p, so->so_owner.data, so->so_owner.len); 623 } 624 625 static void nfs4_xdr_enc_cb_notify_lock(struct rpc_rqst *req, 626 struct xdr_stream *xdr, 627 const void *data) 628 { 629 const struct nfsd4_callback *cb = data; 630 const struct nfsd4_blocked_lock *nbl = 631 container_of(cb, struct nfsd4_blocked_lock, nbl_cb); 632 struct nfs4_lockowner *lo = (struct nfs4_lockowner *)nbl->nbl_lock.fl_owner; 633 struct nfs4_cb_compound_hdr hdr = { 634 .ident = 0, 635 .minorversion = cb->cb_clp->cl_minorversion, 636 }; 637 638 __be32 *p; 639 640 BUG_ON(hdr.minorversion == 0); 641 642 encode_cb_compound4args(xdr, &hdr); 643 encode_cb_sequence4args(xdr, cb, &hdr); 644 645 p = xdr_reserve_space(xdr, 4); 646 *p = cpu_to_be32(OP_CB_NOTIFY_LOCK); 647 encode_nfs_fh4(xdr, &nbl->nbl_fh); 648 encode_stateowner(xdr, &lo->lo_owner); 649 hdr.nops++; 650 651 encode_cb_nops(&hdr); 652 } 653 654 static int nfs4_xdr_dec_cb_notify_lock(struct rpc_rqst *rqstp, 655 struct xdr_stream *xdr, 656 void *data) 657 { 658 struct nfsd4_callback *cb = data; 659 struct nfs4_cb_compound_hdr hdr; 660 int status; 661 662 status = decode_cb_compound4res(xdr, &hdr); 663 if (unlikely(status)) 664 return status; 665 666 if (cb) { 667 status = decode_cb_sequence4res(xdr, cb); 668 if (unlikely(status || cb->cb_seq_status)) 669 return status; 670 } 671 return decode_cb_op_status(xdr, OP_CB_NOTIFY_LOCK, &cb->cb_status); 672 } 673 674 /* 675 * struct write_response4 { 676 * stateid4 wr_callback_id<1>; 677 * length4 wr_count; 678 * stable_how4 wr_committed; 679 * verifier4 wr_writeverf; 680 * }; 681 * union offload_info4 switch (nfsstat4 coa_status) { 682 * case NFS4_OK: 683 * write_response4 coa_resok4; 684 * default: 685 * length4 coa_bytes_copied; 686 * }; 687 * struct CB_OFFLOAD4args { 688 * nfs_fh4 coa_fh; 689 * stateid4 coa_stateid; 690 * offload_info4 coa_offload_info; 691 * }; 692 */ 693 static void encode_offload_info4(struct xdr_stream *xdr, 694 __be32 nfserr, 695 const struct nfsd4_copy *cp) 696 { 697 __be32 *p; 698 699 p = xdr_reserve_space(xdr, 4); 700 *p++ = nfserr; 701 if (!nfserr) { 702 p = xdr_reserve_space(xdr, 4 + 8 + 4 + NFS4_VERIFIER_SIZE); 703 p = xdr_encode_empty_array(p); 704 p = xdr_encode_hyper(p, cp->cp_res.wr_bytes_written); 705 *p++ = cpu_to_be32(cp->cp_res.wr_stable_how); 706 p = xdr_encode_opaque_fixed(p, cp->cp_res.wr_verifier.data, 707 NFS4_VERIFIER_SIZE); 708 } else { 709 p = xdr_reserve_space(xdr, 8); 710 /* We always return success if bytes were written */ 711 p = xdr_encode_hyper(p, 0); 712 } 713 } 714 715 static void encode_cb_offload4args(struct xdr_stream *xdr, 716 __be32 nfserr, 717 const struct knfsd_fh *fh, 718 const struct nfsd4_copy *cp, 719 struct nfs4_cb_compound_hdr *hdr) 720 { 721 __be32 *p; 722 723 p = xdr_reserve_space(xdr, 4); 724 *p++ = cpu_to_be32(OP_CB_OFFLOAD); 725 encode_nfs_fh4(xdr, fh); 726 encode_stateid4(xdr, &cp->cp_res.cb_stateid); 727 encode_offload_info4(xdr, nfserr, cp); 728 729 hdr->nops++; 730 } 731 732 static void nfs4_xdr_enc_cb_offload(struct rpc_rqst *req, 733 struct xdr_stream *xdr, 734 const void *data) 735 { 736 const struct nfsd4_callback *cb = data; 737 const struct nfsd4_copy *cp = 738 container_of(cb, struct nfsd4_copy, cp_cb); 739 struct nfs4_cb_compound_hdr hdr = { 740 .ident = 0, 741 .minorversion = cb->cb_clp->cl_minorversion, 742 }; 743 744 encode_cb_compound4args(xdr, &hdr); 745 encode_cb_sequence4args(xdr, cb, &hdr); 746 encode_cb_offload4args(xdr, cp->nfserr, &cp->fh, cp, &hdr); 747 encode_cb_nops(&hdr); 748 } 749 750 static int nfs4_xdr_dec_cb_offload(struct rpc_rqst *rqstp, 751 struct xdr_stream *xdr, 752 void *data) 753 { 754 struct nfsd4_callback *cb = data; 755 struct nfs4_cb_compound_hdr hdr; 756 int status; 757 758 status = decode_cb_compound4res(xdr, &hdr); 759 if (unlikely(status)) 760 return status; 761 762 if (cb) { 763 status = decode_cb_sequence4res(xdr, cb); 764 if (unlikely(status || cb->cb_seq_status)) 765 return status; 766 } 767 return decode_cb_op_status(xdr, OP_CB_OFFLOAD, &cb->cb_status); 768 } 769 /* 770 * RPC procedure tables 771 */ 772 #define PROC(proc, call, argtype, restype) \ 773 [NFSPROC4_CLNT_##proc] = { \ 774 .p_proc = NFSPROC4_CB_##call, \ 775 .p_encode = nfs4_xdr_enc_##argtype, \ 776 .p_decode = nfs4_xdr_dec_##restype, \ 777 .p_arglen = NFS4_enc_##argtype##_sz, \ 778 .p_replen = NFS4_dec_##restype##_sz, \ 779 .p_statidx = NFSPROC4_CB_##call, \ 780 .p_name = #proc, \ 781 } 782 783 static const struct rpc_procinfo nfs4_cb_procedures[] = { 784 PROC(CB_NULL, NULL, cb_null, cb_null), 785 PROC(CB_RECALL, COMPOUND, cb_recall, cb_recall), 786 #ifdef CONFIG_NFSD_PNFS 787 PROC(CB_LAYOUT, COMPOUND, cb_layout, cb_layout), 788 #endif 789 PROC(CB_NOTIFY_LOCK, COMPOUND, cb_notify_lock, cb_notify_lock), 790 PROC(CB_OFFLOAD, COMPOUND, cb_offload, cb_offload), 791 }; 792 793 static unsigned int nfs4_cb_counts[ARRAY_SIZE(nfs4_cb_procedures)]; 794 static const struct rpc_version nfs_cb_version4 = { 795 /* 796 * Note on the callback rpc program version number: despite language in rfc 797 * 5661 section 18.36.3 requiring servers to use 4 in this field, the 798 * official xdr descriptions for both 4.0 and 4.1 specify version 1, and 799 * in practice that appears to be what implementations use. The section 800 * 18.36.3 language is expected to be fixed in an erratum. 801 */ 802 .number = 1, 803 .nrprocs = ARRAY_SIZE(nfs4_cb_procedures), 804 .procs = nfs4_cb_procedures, 805 .counts = nfs4_cb_counts, 806 }; 807 808 static const struct rpc_version *nfs_cb_version[2] = { 809 [1] = &nfs_cb_version4, 810 }; 811 812 static const struct rpc_program cb_program; 813 814 static struct rpc_stat cb_stats = { 815 .program = &cb_program 816 }; 817 818 #define NFS4_CALLBACK 0x40000000 819 static const struct rpc_program cb_program = { 820 .name = "nfs4_cb", 821 .number = NFS4_CALLBACK, 822 .nrvers = ARRAY_SIZE(nfs_cb_version), 823 .version = nfs_cb_version, 824 .stats = &cb_stats, 825 .pipe_dir_name = "nfsd4_cb", 826 }; 827 828 static int max_cb_time(struct net *net) 829 { 830 struct nfsd_net *nn = net_generic(net, nfsd_net_id); 831 return max(nn->nfsd4_lease/10, (time_t)1) * HZ; 832 } 833 834 static const struct cred *get_backchannel_cred(struct nfs4_client *clp, struct rpc_clnt *client, struct nfsd4_session *ses) 835 { 836 if (clp->cl_minorversion == 0) { 837 client->cl_principal = clp->cl_cred.cr_targ_princ ? 838 clp->cl_cred.cr_targ_princ : "nfs"; 839 840 return get_cred(rpc_machine_cred()); 841 } else { 842 struct cred *kcred; 843 844 kcred = prepare_kernel_cred(NULL); 845 if (!kcred) 846 return NULL; 847 848 kcred->uid = ses->se_cb_sec.uid; 849 kcred->gid = ses->se_cb_sec.gid; 850 return kcred; 851 } 852 } 853 854 static int setup_callback_client(struct nfs4_client *clp, struct nfs4_cb_conn *conn, struct nfsd4_session *ses) 855 { 856 int maxtime = max_cb_time(clp->net); 857 struct rpc_timeout timeparms = { 858 .to_initval = maxtime, 859 .to_retries = 0, 860 .to_maxval = maxtime, 861 }; 862 struct rpc_create_args args = { 863 .net = clp->net, 864 .address = (struct sockaddr *) &conn->cb_addr, 865 .addrsize = conn->cb_addrlen, 866 .saddress = (struct sockaddr *) &conn->cb_saddr, 867 .timeout = &timeparms, 868 .program = &cb_program, 869 .version = 1, 870 .flags = (RPC_CLNT_CREATE_NOPING | RPC_CLNT_CREATE_QUIET), 871 }; 872 struct rpc_clnt *client; 873 const struct cred *cred; 874 875 if (clp->cl_minorversion == 0) { 876 if (!clp->cl_cred.cr_principal && 877 (clp->cl_cred.cr_flavor >= RPC_AUTH_GSS_KRB5)) 878 return -EINVAL; 879 args.client_name = clp->cl_cred.cr_principal; 880 args.prognumber = conn->cb_prog; 881 args.protocol = XPRT_TRANSPORT_TCP; 882 args.authflavor = clp->cl_cred.cr_flavor; 883 clp->cl_cb_ident = conn->cb_ident; 884 } else { 885 if (!conn->cb_xprt) 886 return -EINVAL; 887 clp->cl_cb_conn.cb_xprt = conn->cb_xprt; 888 clp->cl_cb_session = ses; 889 args.bc_xprt = conn->cb_xprt; 890 args.prognumber = clp->cl_cb_session->se_cb_prog; 891 args.protocol = conn->cb_xprt->xpt_class->xcl_ident | 892 XPRT_TRANSPORT_BC; 893 args.authflavor = ses->se_cb_sec.flavor; 894 } 895 /* Create RPC client */ 896 client = rpc_create(&args); 897 if (IS_ERR(client)) { 898 dprintk("NFSD: couldn't create callback client: %ld\n", 899 PTR_ERR(client)); 900 return PTR_ERR(client); 901 } 902 cred = get_backchannel_cred(clp, client, ses); 903 if (!cred) { 904 rpc_shutdown_client(client); 905 return -ENOMEM; 906 } 907 clp->cl_cb_client = client; 908 clp->cl_cb_cred = cred; 909 return 0; 910 } 911 912 static void warn_no_callback_path(struct nfs4_client *clp, int reason) 913 { 914 dprintk("NFSD: warning: no callback path to client %.*s: error %d\n", 915 (int)clp->cl_name.len, clp->cl_name.data, reason); 916 } 917 918 static void nfsd4_mark_cb_down(struct nfs4_client *clp, int reason) 919 { 920 if (test_bit(NFSD4_CLIENT_CB_UPDATE, &clp->cl_flags)) 921 return; 922 clp->cl_cb_state = NFSD4_CB_DOWN; 923 warn_no_callback_path(clp, reason); 924 } 925 926 static void nfsd4_mark_cb_fault(struct nfs4_client *clp, int reason) 927 { 928 if (test_bit(NFSD4_CLIENT_CB_UPDATE, &clp->cl_flags)) 929 return; 930 clp->cl_cb_state = NFSD4_CB_FAULT; 931 warn_no_callback_path(clp, reason); 932 } 933 934 static void nfsd4_cb_probe_done(struct rpc_task *task, void *calldata) 935 { 936 struct nfs4_client *clp = container_of(calldata, struct nfs4_client, cl_cb_null); 937 938 if (task->tk_status) 939 nfsd4_mark_cb_down(clp, task->tk_status); 940 else 941 clp->cl_cb_state = NFSD4_CB_UP; 942 } 943 944 static const struct rpc_call_ops nfsd4_cb_probe_ops = { 945 /* XXX: release method to ensure we set the cb channel down if 946 * necessary on early failure? */ 947 .rpc_call_done = nfsd4_cb_probe_done, 948 }; 949 950 static struct workqueue_struct *callback_wq; 951 952 /* 953 * Poke the callback thread to process any updates to the callback 954 * parameters, and send a null probe. 955 */ 956 void nfsd4_probe_callback(struct nfs4_client *clp) 957 { 958 clp->cl_cb_state = NFSD4_CB_UNKNOWN; 959 set_bit(NFSD4_CLIENT_CB_UPDATE, &clp->cl_flags); 960 nfsd4_run_cb(&clp->cl_cb_null); 961 } 962 963 void nfsd4_probe_callback_sync(struct nfs4_client *clp) 964 { 965 nfsd4_probe_callback(clp); 966 flush_workqueue(callback_wq); 967 } 968 969 void nfsd4_change_callback(struct nfs4_client *clp, struct nfs4_cb_conn *conn) 970 { 971 clp->cl_cb_state = NFSD4_CB_UNKNOWN; 972 spin_lock(&clp->cl_lock); 973 memcpy(&clp->cl_cb_conn, conn, sizeof(struct nfs4_cb_conn)); 974 spin_unlock(&clp->cl_lock); 975 } 976 977 /* 978 * There's currently a single callback channel slot. 979 * If the slot is available, then mark it busy. Otherwise, set the 980 * thread for sleeping on the callback RPC wait queue. 981 */ 982 static bool nfsd41_cb_get_slot(struct nfs4_client *clp, struct rpc_task *task) 983 { 984 if (test_and_set_bit(0, &clp->cl_cb_slot_busy) != 0) { 985 rpc_sleep_on(&clp->cl_cb_waitq, task, NULL); 986 /* Race breaker */ 987 if (test_and_set_bit(0, &clp->cl_cb_slot_busy) != 0) { 988 dprintk("%s slot is busy\n", __func__); 989 return false; 990 } 991 rpc_wake_up_queued_task(&clp->cl_cb_waitq, task); 992 } 993 return true; 994 } 995 996 /* 997 * TODO: cb_sequence should support referring call lists, cachethis, multiple 998 * slots, and mark callback channel down on communication errors. 999 */ 1000 static void nfsd4_cb_prepare(struct rpc_task *task, void *calldata) 1001 { 1002 struct nfsd4_callback *cb = calldata; 1003 struct nfs4_client *clp = cb->cb_clp; 1004 u32 minorversion = clp->cl_minorversion; 1005 1006 /* 1007 * cb_seq_status is only set in decode_cb_sequence4res, 1008 * and so will remain 1 if an rpc level failure occurs. 1009 */ 1010 cb->cb_seq_status = 1; 1011 cb->cb_status = 0; 1012 if (minorversion) { 1013 if (!nfsd41_cb_get_slot(clp, task)) 1014 return; 1015 } 1016 rpc_call_start(task); 1017 } 1018 1019 static bool nfsd4_cb_sequence_done(struct rpc_task *task, struct nfsd4_callback *cb) 1020 { 1021 struct nfs4_client *clp = cb->cb_clp; 1022 struct nfsd4_session *session = clp->cl_cb_session; 1023 bool ret = true; 1024 1025 if (!clp->cl_minorversion) { 1026 /* 1027 * If the backchannel connection was shut down while this 1028 * task was queued, we need to resubmit it after setting up 1029 * a new backchannel connection. 1030 * 1031 * Note that if we lost our callback connection permanently 1032 * the submission code will error out, so we don't need to 1033 * handle that case here. 1034 */ 1035 if (task->tk_flags & RPC_TASK_KILLED) 1036 goto need_restart; 1037 1038 return true; 1039 } 1040 1041 switch (cb->cb_seq_status) { 1042 case 0: 1043 /* 1044 * No need for lock, access serialized in nfsd4_cb_prepare 1045 * 1046 * RFC5661 20.9.3 1047 * If CB_SEQUENCE returns an error, then the state of the slot 1048 * (sequence ID, cached reply) MUST NOT change. 1049 */ 1050 ++session->se_cb_seq_nr; 1051 break; 1052 case -ESERVERFAULT: 1053 ++session->se_cb_seq_nr; 1054 /* Fall through */ 1055 case 1: 1056 case -NFS4ERR_BADSESSION: 1057 nfsd4_mark_cb_fault(cb->cb_clp, cb->cb_seq_status); 1058 ret = false; 1059 break; 1060 case -NFS4ERR_DELAY: 1061 if (!rpc_restart_call(task)) 1062 goto out; 1063 1064 rpc_delay(task, 2 * HZ); 1065 return false; 1066 case -NFS4ERR_BADSLOT: 1067 goto retry_nowait; 1068 case -NFS4ERR_SEQ_MISORDERED: 1069 if (session->se_cb_seq_nr != 1) { 1070 session->se_cb_seq_nr = 1; 1071 goto retry_nowait; 1072 } 1073 break; 1074 default: 1075 dprintk("%s: unprocessed error %d\n", __func__, 1076 cb->cb_seq_status); 1077 } 1078 1079 clear_bit(0, &clp->cl_cb_slot_busy); 1080 rpc_wake_up_next(&clp->cl_cb_waitq); 1081 dprintk("%s: freed slot, new seqid=%d\n", __func__, 1082 clp->cl_cb_session->se_cb_seq_nr); 1083 1084 if (task->tk_flags & RPC_TASK_KILLED) 1085 goto need_restart; 1086 out: 1087 return ret; 1088 retry_nowait: 1089 if (rpc_restart_call_prepare(task)) 1090 ret = false; 1091 goto out; 1092 need_restart: 1093 task->tk_status = 0; 1094 cb->cb_need_restart = true; 1095 return false; 1096 } 1097 1098 static void nfsd4_cb_done(struct rpc_task *task, void *calldata) 1099 { 1100 struct nfsd4_callback *cb = calldata; 1101 struct nfs4_client *clp = cb->cb_clp; 1102 1103 dprintk("%s: minorversion=%d\n", __func__, 1104 clp->cl_minorversion); 1105 1106 if (!nfsd4_cb_sequence_done(task, cb)) 1107 return; 1108 1109 if (cb->cb_status) { 1110 WARN_ON_ONCE(task->tk_status); 1111 task->tk_status = cb->cb_status; 1112 } 1113 1114 switch (cb->cb_ops->done(cb, task)) { 1115 case 0: 1116 task->tk_status = 0; 1117 rpc_restart_call_prepare(task); 1118 return; 1119 case 1: 1120 break; 1121 case -1: 1122 /* Network partition? */ 1123 nfsd4_mark_cb_down(clp, task->tk_status); 1124 break; 1125 default: 1126 BUG(); 1127 } 1128 } 1129 1130 static void nfsd4_cb_release(void *calldata) 1131 { 1132 struct nfsd4_callback *cb = calldata; 1133 1134 if (cb->cb_need_restart) 1135 nfsd4_run_cb(cb); 1136 else 1137 cb->cb_ops->release(cb); 1138 1139 } 1140 1141 static const struct rpc_call_ops nfsd4_cb_ops = { 1142 .rpc_call_prepare = nfsd4_cb_prepare, 1143 .rpc_call_done = nfsd4_cb_done, 1144 .rpc_release = nfsd4_cb_release, 1145 }; 1146 1147 int nfsd4_create_callback_queue(void) 1148 { 1149 callback_wq = alloc_ordered_workqueue("nfsd4_callbacks", 0); 1150 if (!callback_wq) 1151 return -ENOMEM; 1152 return 0; 1153 } 1154 1155 void nfsd4_destroy_callback_queue(void) 1156 { 1157 destroy_workqueue(callback_wq); 1158 } 1159 1160 /* must be called under the state lock */ 1161 void nfsd4_shutdown_callback(struct nfs4_client *clp) 1162 { 1163 set_bit(NFSD4_CLIENT_CB_KILL, &clp->cl_flags); 1164 /* 1165 * Note this won't actually result in a null callback; 1166 * instead, nfsd4_run_cb_null() will detect the killed 1167 * client, destroy the rpc client, and stop: 1168 */ 1169 nfsd4_run_cb(&clp->cl_cb_null); 1170 flush_workqueue(callback_wq); 1171 } 1172 1173 /* requires cl_lock: */ 1174 static struct nfsd4_conn * __nfsd4_find_backchannel(struct nfs4_client *clp) 1175 { 1176 struct nfsd4_session *s; 1177 struct nfsd4_conn *c; 1178 1179 list_for_each_entry(s, &clp->cl_sessions, se_perclnt) { 1180 list_for_each_entry(c, &s->se_conns, cn_persession) { 1181 if (c->cn_flags & NFS4_CDFC4_BACK) 1182 return c; 1183 } 1184 } 1185 return NULL; 1186 } 1187 1188 static void nfsd4_process_cb_update(struct nfsd4_callback *cb) 1189 { 1190 struct nfs4_cb_conn conn; 1191 struct nfs4_client *clp = cb->cb_clp; 1192 struct nfsd4_session *ses = NULL; 1193 struct nfsd4_conn *c; 1194 int err; 1195 1196 /* 1197 * This is either an update, or the client dying; in either case, 1198 * kill the old client: 1199 */ 1200 if (clp->cl_cb_client) { 1201 rpc_shutdown_client(clp->cl_cb_client); 1202 clp->cl_cb_client = NULL; 1203 put_cred(clp->cl_cb_cred); 1204 clp->cl_cb_cred = NULL; 1205 } 1206 if (clp->cl_cb_conn.cb_xprt) { 1207 svc_xprt_put(clp->cl_cb_conn.cb_xprt); 1208 clp->cl_cb_conn.cb_xprt = NULL; 1209 } 1210 if (test_bit(NFSD4_CLIENT_CB_KILL, &clp->cl_flags)) 1211 return; 1212 spin_lock(&clp->cl_lock); 1213 /* 1214 * Only serialized callback code is allowed to clear these 1215 * flags; main nfsd code can only set them: 1216 */ 1217 BUG_ON(!(clp->cl_flags & NFSD4_CLIENT_CB_FLAG_MASK)); 1218 clear_bit(NFSD4_CLIENT_CB_UPDATE, &clp->cl_flags); 1219 memcpy(&conn, &cb->cb_clp->cl_cb_conn, sizeof(struct nfs4_cb_conn)); 1220 c = __nfsd4_find_backchannel(clp); 1221 if (c) { 1222 svc_xprt_get(c->cn_xprt); 1223 conn.cb_xprt = c->cn_xprt; 1224 ses = c->cn_session; 1225 } 1226 spin_unlock(&clp->cl_lock); 1227 1228 err = setup_callback_client(clp, &conn, ses); 1229 if (err) { 1230 nfsd4_mark_cb_down(clp, err); 1231 return; 1232 } 1233 } 1234 1235 static void 1236 nfsd4_run_cb_work(struct work_struct *work) 1237 { 1238 struct nfsd4_callback *cb = 1239 container_of(work, struct nfsd4_callback, cb_work); 1240 struct nfs4_client *clp = cb->cb_clp; 1241 struct rpc_clnt *clnt; 1242 1243 if (cb->cb_need_restart) { 1244 cb->cb_need_restart = false; 1245 } else { 1246 if (cb->cb_ops && cb->cb_ops->prepare) 1247 cb->cb_ops->prepare(cb); 1248 } 1249 1250 if (clp->cl_flags & NFSD4_CLIENT_CB_FLAG_MASK) 1251 nfsd4_process_cb_update(cb); 1252 1253 clnt = clp->cl_cb_client; 1254 if (!clnt) { 1255 /* Callback channel broken, or client killed; give up: */ 1256 if (cb->cb_ops && cb->cb_ops->release) 1257 cb->cb_ops->release(cb); 1258 return; 1259 } 1260 1261 /* 1262 * Don't send probe messages for 4.1 or later. 1263 */ 1264 if (!cb->cb_ops && clp->cl_minorversion) { 1265 clp->cl_cb_state = NFSD4_CB_UP; 1266 return; 1267 } 1268 1269 cb->cb_msg.rpc_cred = clp->cl_cb_cred; 1270 rpc_call_async(clnt, &cb->cb_msg, RPC_TASK_SOFT | RPC_TASK_SOFTCONN, 1271 cb->cb_ops ? &nfsd4_cb_ops : &nfsd4_cb_probe_ops, cb); 1272 } 1273 1274 void nfsd4_init_cb(struct nfsd4_callback *cb, struct nfs4_client *clp, 1275 const struct nfsd4_callback_ops *ops, enum nfsd4_cb_op op) 1276 { 1277 cb->cb_clp = clp; 1278 cb->cb_msg.rpc_proc = &nfs4_cb_procedures[op]; 1279 cb->cb_msg.rpc_argp = cb; 1280 cb->cb_msg.rpc_resp = cb; 1281 cb->cb_ops = ops; 1282 INIT_WORK(&cb->cb_work, nfsd4_run_cb_work); 1283 cb->cb_seq_status = 1; 1284 cb->cb_status = 0; 1285 cb->cb_need_restart = false; 1286 } 1287 1288 void nfsd4_run_cb(struct nfsd4_callback *cb) 1289 { 1290 queue_work(callback_wq, &cb->cb_work); 1291 } 1292