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/nfs4.h> 35 #include <linux/sunrpc/clnt.h> 36 #include <linux/sunrpc/xprt.h> 37 #include <linux/sunrpc/svc_xprt.h> 38 #include <linux/slab.h> 39 #include "nfsd.h" 40 #include "state.h" 41 #include "netns.h" 42 #include "trace.h" 43 #include "xdr4cb.h" 44 #include "xdr4.h" 45 #include "nfs4xdr_gen.h" 46 47 #define NFSDDBG_FACILITY NFSDDBG_PROC 48 49 #define NFSPROC4_CB_NULL 0 50 #define NFSPROC4_CB_COMPOUND 1 51 52 /* Index of predefined Linux callback client operations */ 53 54 struct nfs4_cb_compound_hdr { 55 /* args */ 56 u32 ident; /* minorversion 0 only */ 57 u32 nops; 58 __be32 *nops_p; 59 u32 minorversion; 60 /* res */ 61 int status; 62 }; 63 64 static __be32 *xdr_encode_empty_array(__be32 *p) 65 { 66 *p++ = xdr_zero; 67 return p; 68 } 69 70 /* 71 * Encode/decode NFSv4 CB basic data types 72 * 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 76 * 1 Protocol" 77 */ 78 79 static void encode_uint32(struct xdr_stream *xdr, u32 n) 80 { 81 WARN_ON_ONCE(xdr_stream_encode_u32(xdr, n) < 0); 82 } 83 84 static void encode_bitmap4(struct xdr_stream *xdr, const __u32 *bitmap, 85 size_t len) 86 { 87 xdr_stream_encode_uint32_array(xdr, bitmap, len); 88 } 89 90 static int decode_cb_fattr4(struct xdr_stream *xdr, uint32_t *bitmap, 91 struct nfs4_cb_fattr *fattr) 92 { 93 fattr->ncf_cb_change = 0; 94 fattr->ncf_cb_fsize = 0; 95 fattr->ncf_cb_atime.tv_sec = 0; 96 fattr->ncf_cb_atime.tv_nsec = 0; 97 fattr->ncf_cb_mtime.tv_sec = 0; 98 fattr->ncf_cb_mtime.tv_nsec = 0; 99 100 if (bitmap[0] & FATTR4_WORD0_CHANGE) 101 if (xdr_stream_decode_u64(xdr, &fattr->ncf_cb_change) < 0) 102 return -EIO; 103 if (bitmap[0] & FATTR4_WORD0_SIZE) 104 if (xdr_stream_decode_u64(xdr, &fattr->ncf_cb_fsize) < 0) 105 return -EIO; 106 if (bitmap[2] & FATTR4_WORD2_TIME_DELEG_ACCESS) { 107 fattr4_time_deleg_access access; 108 109 if (!xdrgen_decode_fattr4_time_deleg_access(xdr, &access)) 110 return -EIO; 111 fattr->ncf_cb_atime.tv_sec = access.seconds; 112 fattr->ncf_cb_atime.tv_nsec = access.nseconds; 113 114 } 115 if (bitmap[2] & FATTR4_WORD2_TIME_DELEG_MODIFY) { 116 fattr4_time_deleg_modify modify; 117 118 if (!xdrgen_decode_fattr4_time_deleg_modify(xdr, &modify)) 119 return -EIO; 120 fattr->ncf_cb_mtime.tv_sec = modify.seconds; 121 fattr->ncf_cb_mtime.tv_nsec = modify.nseconds; 122 123 } 124 return 0; 125 } 126 127 static void encode_nfs_cb_opnum4(struct xdr_stream *xdr, enum nfs_cb_opnum4 op) 128 { 129 __be32 *p; 130 131 p = xdr_reserve_space(xdr, 4); 132 *p = cpu_to_be32(op); 133 } 134 135 /* 136 * nfs_fh4 137 * 138 * typedef opaque nfs_fh4<NFS4_FHSIZE>; 139 */ 140 static void encode_nfs_fh4(struct xdr_stream *xdr, const struct knfsd_fh *fh) 141 { 142 u32 length = fh->fh_size; 143 __be32 *p; 144 145 BUG_ON(length > NFS4_FHSIZE); 146 p = xdr_reserve_space(xdr, 4 + length); 147 xdr_encode_opaque(p, &fh->fh_raw, length); 148 } 149 150 /* 151 * stateid4 152 * 153 * struct stateid4 { 154 * uint32_t seqid; 155 * opaque other[12]; 156 * }; 157 */ 158 static void encode_stateid4(struct xdr_stream *xdr, const stateid_t *sid) 159 { 160 __be32 *p; 161 162 p = xdr_reserve_space(xdr, NFS4_STATEID_SIZE); 163 *p++ = cpu_to_be32(sid->si_generation); 164 xdr_encode_opaque_fixed(p, &sid->si_opaque, NFS4_STATEID_OTHER_SIZE); 165 } 166 167 /* 168 * sessionid4 169 * 170 * typedef opaque sessionid4[NFS4_SESSIONID_SIZE]; 171 */ 172 static void encode_sessionid4(struct xdr_stream *xdr, 173 const struct nfsd4_session *session) 174 { 175 __be32 *p; 176 177 p = xdr_reserve_space(xdr, NFS4_MAX_SESSIONID_LEN); 178 xdr_encode_opaque_fixed(p, session->se_sessionid.data, 179 NFS4_MAX_SESSIONID_LEN); 180 } 181 182 /* 183 * nfsstat4 184 */ 185 static const struct { 186 int stat; 187 int errno; 188 } nfs_cb_errtbl[] = { 189 { NFS4_OK, 0 }, 190 { NFS4ERR_PERM, -EPERM }, 191 { NFS4ERR_NOENT, -ENOENT }, 192 { NFS4ERR_IO, -EIO }, 193 { NFS4ERR_NXIO, -ENXIO }, 194 { NFS4ERR_ACCESS, -EACCES }, 195 { NFS4ERR_EXIST, -EEXIST }, 196 { NFS4ERR_XDEV, -EXDEV }, 197 { NFS4ERR_NOTDIR, -ENOTDIR }, 198 { NFS4ERR_ISDIR, -EISDIR }, 199 { NFS4ERR_INVAL, -EINVAL }, 200 { NFS4ERR_FBIG, -EFBIG }, 201 { NFS4ERR_NOSPC, -ENOSPC }, 202 { NFS4ERR_ROFS, -EROFS }, 203 { NFS4ERR_MLINK, -EMLINK }, 204 { NFS4ERR_NAMETOOLONG, -ENAMETOOLONG }, 205 { NFS4ERR_NOTEMPTY, -ENOTEMPTY }, 206 { NFS4ERR_DQUOT, -EDQUOT }, 207 { NFS4ERR_STALE, -ESTALE }, 208 { NFS4ERR_BADHANDLE, -EBADHANDLE }, 209 { NFS4ERR_BAD_COOKIE, -EBADCOOKIE }, 210 { NFS4ERR_NOTSUPP, -ENOTSUPP }, 211 { NFS4ERR_TOOSMALL, -ETOOSMALL }, 212 { NFS4ERR_SERVERFAULT, -ESERVERFAULT }, 213 { NFS4ERR_BADTYPE, -EBADTYPE }, 214 { NFS4ERR_LOCKED, -EAGAIN }, 215 { NFS4ERR_RESOURCE, -EREMOTEIO }, 216 { NFS4ERR_SYMLINK, -ELOOP }, 217 { NFS4ERR_OP_ILLEGAL, -EOPNOTSUPP }, 218 { NFS4ERR_DEADLOCK, -EDEADLK }, 219 { -1, -EIO } 220 }; 221 222 /* 223 * If we cannot translate the error, the recovery routines should 224 * handle it. 225 * 226 * Note: remaining NFSv4 error codes have values > 10000, so should 227 * not conflict with native Linux error codes. 228 */ 229 static int nfs_cb_stat_to_errno(int status) 230 { 231 int i; 232 233 for (i = 0; nfs_cb_errtbl[i].stat != -1; i++) { 234 if (nfs_cb_errtbl[i].stat == status) 235 return nfs_cb_errtbl[i].errno; 236 } 237 238 dprintk("NFSD: Unrecognized NFS CB status value: %u\n", status); 239 return -status; 240 } 241 242 static int decode_cb_op_status(struct xdr_stream *xdr, 243 enum nfs_cb_opnum4 expected, int *status) 244 { 245 __be32 *p; 246 u32 op; 247 248 p = xdr_inline_decode(xdr, 4 + 4); 249 if (unlikely(p == NULL)) 250 goto out_overflow; 251 op = be32_to_cpup(p++); 252 if (unlikely(op != expected)) 253 goto out_unexpected; 254 *status = nfs_cb_stat_to_errno(be32_to_cpup(p)); 255 return 0; 256 out_overflow: 257 return -EIO; 258 out_unexpected: 259 dprintk("NFSD: Callback server returned operation %d but " 260 "we issued a request for %d\n", op, expected); 261 return -EIO; 262 } 263 264 /* 265 * CB_COMPOUND4args 266 * 267 * struct CB_COMPOUND4args { 268 * utf8str_cs tag; 269 * uint32_t minorversion; 270 * uint32_t callback_ident; 271 * nfs_cb_argop4 argarray<>; 272 * }; 273 */ 274 static void encode_cb_compound4args(struct xdr_stream *xdr, 275 struct nfs4_cb_compound_hdr *hdr) 276 { 277 __be32 * p; 278 279 p = xdr_reserve_space(xdr, 4 + 4 + 4 + 4); 280 p = xdr_encode_empty_array(p); /* empty tag */ 281 *p++ = cpu_to_be32(hdr->minorversion); 282 *p++ = cpu_to_be32(hdr->ident); 283 284 hdr->nops_p = p; 285 *p = cpu_to_be32(hdr->nops); /* argarray element count */ 286 } 287 288 /* 289 * Update argarray element count 290 */ 291 static void encode_cb_nops(struct nfs4_cb_compound_hdr *hdr) 292 { 293 BUG_ON(hdr->nops > NFS4_MAX_BACK_CHANNEL_OPS); 294 *hdr->nops_p = cpu_to_be32(hdr->nops); 295 } 296 297 /* 298 * CB_COMPOUND4res 299 * 300 * struct CB_COMPOUND4res { 301 * nfsstat4 status; 302 * utf8str_cs tag; 303 * nfs_cb_resop4 resarray<>; 304 * }; 305 */ 306 static int decode_cb_compound4res(struct xdr_stream *xdr, 307 struct nfs4_cb_compound_hdr *hdr) 308 { 309 u32 length; 310 __be32 *p; 311 312 p = xdr_inline_decode(xdr, XDR_UNIT); 313 if (unlikely(p == NULL)) 314 goto out_overflow; 315 hdr->status = be32_to_cpup(p); 316 /* Ignore the tag */ 317 if (xdr_stream_decode_u32(xdr, &length) < 0) 318 goto out_overflow; 319 if (xdr_inline_decode(xdr, length) == NULL) 320 goto out_overflow; 321 if (xdr_stream_decode_u32(xdr, &hdr->nops) < 0) 322 goto out_overflow; 323 return 0; 324 out_overflow: 325 return -EIO; 326 } 327 328 /* 329 * CB_RECALL4args 330 * 331 * struct CB_RECALL4args { 332 * stateid4 stateid; 333 * bool truncate; 334 * nfs_fh4 fh; 335 * }; 336 */ 337 static void encode_cb_recall4args(struct xdr_stream *xdr, 338 const struct nfs4_delegation *dp, 339 struct nfs4_cb_compound_hdr *hdr) 340 { 341 __be32 *p; 342 343 encode_nfs_cb_opnum4(xdr, OP_CB_RECALL); 344 encode_stateid4(xdr, &dp->dl_stid.sc_stateid); 345 346 p = xdr_reserve_space(xdr, 4); 347 *p++ = xdr_zero; /* truncate */ 348 349 encode_nfs_fh4(xdr, &dp->dl_stid.sc_file->fi_fhandle); 350 351 hdr->nops++; 352 } 353 354 /* 355 * CB_RECALLANY4args 356 * 357 * struct CB_RECALLANY4args { 358 * uint32_t craa_objects_to_keep; 359 * bitmap4 craa_type_mask; 360 * }; 361 */ 362 static void 363 encode_cb_recallany4args(struct xdr_stream *xdr, 364 struct nfs4_cb_compound_hdr *hdr, struct nfsd4_cb_recall_any *ra) 365 { 366 encode_nfs_cb_opnum4(xdr, OP_CB_RECALL_ANY); 367 encode_uint32(xdr, ra->ra_keep); 368 encode_bitmap4(xdr, ra->ra_bmval, ARRAY_SIZE(ra->ra_bmval)); 369 hdr->nops++; 370 } 371 372 /* 373 * CB_GETATTR4args 374 * struct CB_GETATTR4args { 375 * nfs_fh4 fh; 376 * bitmap4 attr_request; 377 * }; 378 * 379 * The size and change attributes are the only one 380 * guaranteed to be serviced by the client. 381 */ 382 static void 383 encode_cb_getattr4args(struct xdr_stream *xdr, struct nfs4_cb_compound_hdr *hdr, 384 struct nfs4_cb_fattr *fattr) 385 { 386 struct nfs4_delegation *dp = container_of(fattr, struct nfs4_delegation, dl_cb_fattr); 387 struct knfsd_fh *fh = &dp->dl_stid.sc_file->fi_fhandle; 388 struct nfs4_cb_fattr *ncf = &dp->dl_cb_fattr; 389 u32 bmap_size = 1; 390 u32 bmap[3]; 391 392 bmap[0] = FATTR4_WORD0_SIZE; 393 if (!ncf->ncf_file_modified) 394 bmap[0] |= FATTR4_WORD0_CHANGE; 395 396 if (deleg_attrs_deleg(dp->dl_type)) { 397 bmap[1] = 0; 398 bmap[2] = FATTR4_WORD2_TIME_DELEG_ACCESS | FATTR4_WORD2_TIME_DELEG_MODIFY; 399 bmap_size = 3; 400 } 401 encode_nfs_cb_opnum4(xdr, OP_CB_GETATTR); 402 encode_nfs_fh4(xdr, fh); 403 encode_bitmap4(xdr, bmap, bmap_size); 404 hdr->nops++; 405 } 406 407 static u32 highest_slotid(struct nfsd4_session *ses) 408 { 409 u32 idx; 410 411 spin_lock(&ses->se_lock); 412 idx = fls(~ses->se_cb_slot_avail); 413 if (idx > 0) 414 --idx; 415 idx = max(idx, ses->se_cb_highest_slot); 416 spin_unlock(&ses->se_lock); 417 return idx; 418 } 419 420 /* 421 * CB_SEQUENCE4args 422 * 423 * struct CB_SEQUENCE4args { 424 * sessionid4 csa_sessionid; 425 * sequenceid4 csa_sequenceid; 426 * slotid4 csa_slotid; 427 * slotid4 csa_highest_slotid; 428 * bool csa_cachethis; 429 * referring_call_list4 csa_referring_call_lists<>; 430 * }; 431 */ 432 static void encode_cb_sequence4args(struct xdr_stream *xdr, 433 const struct nfsd4_callback *cb, 434 struct nfs4_cb_compound_hdr *hdr) 435 { 436 struct nfsd4_session *session = cb->cb_clp->cl_cb_session; 437 __be32 *p; 438 439 if (hdr->minorversion == 0) 440 return; 441 442 encode_nfs_cb_opnum4(xdr, OP_CB_SEQUENCE); 443 encode_sessionid4(xdr, session); 444 445 p = xdr_reserve_space(xdr, 4 + 4 + 4 + 4 + 4); 446 *p++ = cpu_to_be32(session->se_cb_seq_nr[cb->cb_held_slot]); /* csa_sequenceid */ 447 *p++ = cpu_to_be32(cb->cb_held_slot); /* csa_slotid */ 448 *p++ = cpu_to_be32(highest_slotid(session)); /* csa_highest_slotid */ 449 *p++ = xdr_zero; /* csa_cachethis */ 450 xdr_encode_empty_array(p); /* csa_referring_call_lists */ 451 452 hdr->nops++; 453 } 454 455 static void update_cb_slot_table(struct nfsd4_session *ses, u32 target) 456 { 457 /* No need to do anything if nothing changed */ 458 if (likely(target == READ_ONCE(ses->se_cb_highest_slot))) 459 return; 460 461 spin_lock(&ses->se_lock); 462 if (target > ses->se_cb_highest_slot) { 463 int i; 464 465 target = min(target, NFSD_BC_SLOT_TABLE_SIZE - 1); 466 467 /* 468 * Growing the slot table. Reset any new sequences to 1. 469 * 470 * NB: There is some debate about whether the RFC requires this, 471 * but the Linux client expects it. 472 */ 473 for (i = ses->se_cb_highest_slot + 1; i <= target; ++i) 474 ses->se_cb_seq_nr[i] = 1; 475 } 476 ses->se_cb_highest_slot = target; 477 spin_unlock(&ses->se_lock); 478 } 479 480 /* 481 * CB_SEQUENCE4resok 482 * 483 * struct CB_SEQUENCE4resok { 484 * sessionid4 csr_sessionid; 485 * sequenceid4 csr_sequenceid; 486 * slotid4 csr_slotid; 487 * slotid4 csr_highest_slotid; 488 * slotid4 csr_target_highest_slotid; 489 * }; 490 * 491 * union CB_SEQUENCE4res switch (nfsstat4 csr_status) { 492 * case NFS4_OK: 493 * CB_SEQUENCE4resok csr_resok4; 494 * default: 495 * void; 496 * }; 497 * 498 * Our current back channel implmentation supports a single backchannel 499 * with a single slot. 500 */ 501 static int decode_cb_sequence4resok(struct xdr_stream *xdr, 502 struct nfsd4_callback *cb) 503 { 504 struct nfsd4_session *session = cb->cb_clp->cl_cb_session; 505 int status = -ESERVERFAULT; 506 __be32 *p; 507 u32 seqid, slotid, target; 508 509 /* 510 * If the server returns different values for sessionID, slotID or 511 * sequence number, the server is looney tunes. 512 */ 513 p = xdr_inline_decode(xdr, NFS4_MAX_SESSIONID_LEN + 4 + 4 + 4 + 4); 514 if (unlikely(p == NULL)) 515 goto out_overflow; 516 517 if (memcmp(p, session->se_sessionid.data, NFS4_MAX_SESSIONID_LEN)) { 518 dprintk("NFS: %s Invalid session id\n", __func__); 519 goto out; 520 } 521 p += XDR_QUADLEN(NFS4_MAX_SESSIONID_LEN); 522 523 seqid = be32_to_cpup(p++); 524 if (seqid != session->se_cb_seq_nr[cb->cb_held_slot]) { 525 dprintk("NFS: %s Invalid sequence number\n", __func__); 526 goto out; 527 } 528 529 slotid = be32_to_cpup(p++); 530 if (slotid != cb->cb_held_slot) { 531 dprintk("NFS: %s Invalid slotid\n", __func__); 532 goto out; 533 } 534 535 p++; // ignore current highest slot value 536 537 target = be32_to_cpup(p++); 538 update_cb_slot_table(session, target); 539 status = 0; 540 out: 541 cb->cb_seq_status = status; 542 return status; 543 out_overflow: 544 status = -EIO; 545 goto out; 546 } 547 548 static int decode_cb_sequence4res(struct xdr_stream *xdr, 549 struct nfsd4_callback *cb) 550 { 551 int status; 552 553 if (cb->cb_clp->cl_minorversion == 0) 554 return 0; 555 556 status = decode_cb_op_status(xdr, OP_CB_SEQUENCE, &cb->cb_seq_status); 557 if (unlikely(status || cb->cb_seq_status)) 558 return status; 559 560 return decode_cb_sequence4resok(xdr, cb); 561 } 562 563 /* 564 * NFSv4.0 and NFSv4.1 XDR encode functions 565 * 566 * NFSv4.0 callback argument types are defined in section 15 of RFC 567 * 3530: "Network File System (NFS) version 4 Protocol" and section 20 568 * of RFC 5661: "Network File System (NFS) Version 4 Minor Version 1 569 * Protocol". 570 */ 571 572 /* 573 * NB: Without this zero space reservation, callbacks over krb5p fail 574 */ 575 static void nfs4_xdr_enc_cb_null(struct rpc_rqst *req, struct xdr_stream *xdr, 576 const void *__unused) 577 { 578 xdr_reserve_space(xdr, 0); 579 } 580 581 /* 582 * 20.1. Operation 3: CB_GETATTR - Get Attributes 583 */ 584 static void nfs4_xdr_enc_cb_getattr(struct rpc_rqst *req, 585 struct xdr_stream *xdr, const void *data) 586 { 587 const struct nfsd4_callback *cb = data; 588 struct nfs4_cb_fattr *ncf = 589 container_of(cb, struct nfs4_cb_fattr, ncf_getattr); 590 struct nfs4_cb_compound_hdr hdr = { 591 .ident = cb->cb_clp->cl_cb_ident, 592 .minorversion = cb->cb_clp->cl_minorversion, 593 }; 594 595 encode_cb_compound4args(xdr, &hdr); 596 encode_cb_sequence4args(xdr, cb, &hdr); 597 encode_cb_getattr4args(xdr, &hdr, ncf); 598 encode_cb_nops(&hdr); 599 } 600 601 /* 602 * 20.2. Operation 4: CB_RECALL - Recall a Delegation 603 */ 604 static void nfs4_xdr_enc_cb_recall(struct rpc_rqst *req, struct xdr_stream *xdr, 605 const void *data) 606 { 607 const struct nfsd4_callback *cb = data; 608 const struct nfs4_delegation *dp = cb_to_delegation(cb); 609 struct nfs4_cb_compound_hdr hdr = { 610 .ident = cb->cb_clp->cl_cb_ident, 611 .minorversion = cb->cb_clp->cl_minorversion, 612 }; 613 614 encode_cb_compound4args(xdr, &hdr); 615 encode_cb_sequence4args(xdr, cb, &hdr); 616 encode_cb_recall4args(xdr, dp, &hdr); 617 encode_cb_nops(&hdr); 618 } 619 620 /* 621 * 20.6. Operation 8: CB_RECALL_ANY - Keep Any N Recallable Objects 622 */ 623 static void 624 nfs4_xdr_enc_cb_recall_any(struct rpc_rqst *req, 625 struct xdr_stream *xdr, const void *data) 626 { 627 const struct nfsd4_callback *cb = data; 628 struct nfsd4_cb_recall_any *ra; 629 struct nfs4_cb_compound_hdr hdr = { 630 .ident = cb->cb_clp->cl_cb_ident, 631 .minorversion = cb->cb_clp->cl_minorversion, 632 }; 633 634 ra = container_of(cb, struct nfsd4_cb_recall_any, ra_cb); 635 encode_cb_compound4args(xdr, &hdr); 636 encode_cb_sequence4args(xdr, cb, &hdr); 637 encode_cb_recallany4args(xdr, &hdr, ra); 638 encode_cb_nops(&hdr); 639 } 640 641 /* 642 * NFSv4.0 and NFSv4.1 XDR decode functions 643 * 644 * NFSv4.0 callback result types are defined in section 15 of RFC 645 * 3530: "Network File System (NFS) version 4 Protocol" and section 20 646 * of RFC 5661: "Network File System (NFS) Version 4 Minor Version 1 647 * Protocol". 648 */ 649 650 static int nfs4_xdr_dec_cb_null(struct rpc_rqst *req, struct xdr_stream *xdr, 651 void *__unused) 652 { 653 return 0; 654 } 655 656 /* 657 * 20.1. Operation 3: CB_GETATTR - Get Attributes 658 */ 659 static int nfs4_xdr_dec_cb_getattr(struct rpc_rqst *rqstp, 660 struct xdr_stream *xdr, 661 void *data) 662 { 663 struct nfsd4_callback *cb = data; 664 struct nfs4_cb_compound_hdr hdr; 665 int status; 666 u32 bitmap[3] = {0}; 667 u32 attrlen, maxlen; 668 struct nfs4_cb_fattr *ncf = 669 container_of(cb, struct nfs4_cb_fattr, ncf_getattr); 670 671 status = decode_cb_compound4res(xdr, &hdr); 672 if (unlikely(status)) 673 return status; 674 675 status = decode_cb_sequence4res(xdr, cb); 676 if (unlikely(status || cb->cb_seq_status)) 677 return status; 678 679 status = decode_cb_op_status(xdr, OP_CB_GETATTR, &cb->cb_status); 680 if (unlikely(status || cb->cb_status)) 681 return status; 682 if (xdr_stream_decode_uint32_array(xdr, bitmap, 3) < 0) 683 return -EIO; 684 if (xdr_stream_decode_u32(xdr, &attrlen) < 0) 685 return -EIO; 686 maxlen = sizeof(ncf->ncf_cb_change) + sizeof(ncf->ncf_cb_fsize); 687 if (bitmap[2] != 0) 688 maxlen += (sizeof(ncf->ncf_cb_mtime.tv_sec) + 689 sizeof(ncf->ncf_cb_mtime.tv_nsec)) * 2; 690 if (attrlen > maxlen) 691 return -EIO; 692 status = decode_cb_fattr4(xdr, bitmap, ncf); 693 return status; 694 } 695 696 /* 697 * 20.2. Operation 4: CB_RECALL - Recall a Delegation 698 */ 699 static int nfs4_xdr_dec_cb_recall(struct rpc_rqst *rqstp, 700 struct xdr_stream *xdr, 701 void *data) 702 { 703 struct nfsd4_callback *cb = data; 704 struct nfs4_cb_compound_hdr hdr; 705 int status; 706 707 status = decode_cb_compound4res(xdr, &hdr); 708 if (unlikely(status)) 709 return status; 710 711 status = decode_cb_sequence4res(xdr, cb); 712 if (unlikely(status || cb->cb_seq_status)) 713 return status; 714 715 return decode_cb_op_status(xdr, OP_CB_RECALL, &cb->cb_status); 716 } 717 718 /* 719 * 20.6. Operation 8: CB_RECALL_ANY - Keep Any N Recallable Objects 720 */ 721 static int 722 nfs4_xdr_dec_cb_recall_any(struct rpc_rqst *rqstp, 723 struct xdr_stream *xdr, 724 void *data) 725 { 726 struct nfsd4_callback *cb = data; 727 struct nfs4_cb_compound_hdr hdr; 728 int status; 729 730 status = decode_cb_compound4res(xdr, &hdr); 731 if (unlikely(status)) 732 return status; 733 status = decode_cb_sequence4res(xdr, cb); 734 if (unlikely(status || cb->cb_seq_status)) 735 return status; 736 status = decode_cb_op_status(xdr, OP_CB_RECALL_ANY, &cb->cb_status); 737 return status; 738 } 739 740 #ifdef CONFIG_NFSD_PNFS 741 /* 742 * CB_LAYOUTRECALL4args 743 * 744 * struct layoutrecall_file4 { 745 * nfs_fh4 lor_fh; 746 * offset4 lor_offset; 747 * length4 lor_length; 748 * stateid4 lor_stateid; 749 * }; 750 * 751 * union layoutrecall4 switch(layoutrecall_type4 lor_recalltype) { 752 * case LAYOUTRECALL4_FILE: 753 * layoutrecall_file4 lor_layout; 754 * case LAYOUTRECALL4_FSID: 755 * fsid4 lor_fsid; 756 * case LAYOUTRECALL4_ALL: 757 * void; 758 * }; 759 * 760 * struct CB_LAYOUTRECALL4args { 761 * layouttype4 clora_type; 762 * layoutiomode4 clora_iomode; 763 * bool clora_changed; 764 * layoutrecall4 clora_recall; 765 * }; 766 */ 767 static void encode_cb_layout4args(struct xdr_stream *xdr, 768 const struct nfs4_layout_stateid *ls, 769 struct nfs4_cb_compound_hdr *hdr) 770 { 771 __be32 *p; 772 773 BUG_ON(hdr->minorversion == 0); 774 775 p = xdr_reserve_space(xdr, 5 * 4); 776 *p++ = cpu_to_be32(OP_CB_LAYOUTRECALL); 777 *p++ = cpu_to_be32(ls->ls_layout_type); 778 *p++ = cpu_to_be32(IOMODE_ANY); 779 *p++ = cpu_to_be32(1); 780 *p = cpu_to_be32(RETURN_FILE); 781 782 encode_nfs_fh4(xdr, &ls->ls_stid.sc_file->fi_fhandle); 783 784 p = xdr_reserve_space(xdr, 2 * 8); 785 p = xdr_encode_hyper(p, 0); 786 xdr_encode_hyper(p, NFS4_MAX_UINT64); 787 788 encode_stateid4(xdr, &ls->ls_recall_sid); 789 790 hdr->nops++; 791 } 792 793 static void nfs4_xdr_enc_cb_layout(struct rpc_rqst *req, 794 struct xdr_stream *xdr, 795 const void *data) 796 { 797 const struct nfsd4_callback *cb = data; 798 const struct nfs4_layout_stateid *ls = 799 container_of(cb, struct nfs4_layout_stateid, ls_recall); 800 struct nfs4_cb_compound_hdr hdr = { 801 .ident = 0, 802 .minorversion = cb->cb_clp->cl_minorversion, 803 }; 804 805 encode_cb_compound4args(xdr, &hdr); 806 encode_cb_sequence4args(xdr, cb, &hdr); 807 encode_cb_layout4args(xdr, ls, &hdr); 808 encode_cb_nops(&hdr); 809 } 810 811 static int nfs4_xdr_dec_cb_layout(struct rpc_rqst *rqstp, 812 struct xdr_stream *xdr, 813 void *data) 814 { 815 struct nfsd4_callback *cb = data; 816 struct nfs4_cb_compound_hdr hdr; 817 int status; 818 819 status = decode_cb_compound4res(xdr, &hdr); 820 if (unlikely(status)) 821 return status; 822 823 status = decode_cb_sequence4res(xdr, cb); 824 if (unlikely(status || cb->cb_seq_status)) 825 return status; 826 827 return decode_cb_op_status(xdr, OP_CB_LAYOUTRECALL, &cb->cb_status); 828 } 829 #endif /* CONFIG_NFSD_PNFS */ 830 831 static void encode_stateowner(struct xdr_stream *xdr, struct nfs4_stateowner *so) 832 { 833 __be32 *p; 834 835 p = xdr_reserve_space(xdr, 8 + 4 + so->so_owner.len); 836 p = xdr_encode_opaque_fixed(p, &so->so_client->cl_clientid, 8); 837 xdr_encode_opaque(p, so->so_owner.data, so->so_owner.len); 838 } 839 840 static void nfs4_xdr_enc_cb_notify_lock(struct rpc_rqst *req, 841 struct xdr_stream *xdr, 842 const void *data) 843 { 844 const struct nfsd4_callback *cb = data; 845 const struct nfsd4_blocked_lock *nbl = 846 container_of(cb, struct nfsd4_blocked_lock, nbl_cb); 847 struct nfs4_lockowner *lo = (struct nfs4_lockowner *)nbl->nbl_lock.c.flc_owner; 848 struct nfs4_cb_compound_hdr hdr = { 849 .ident = 0, 850 .minorversion = cb->cb_clp->cl_minorversion, 851 }; 852 853 __be32 *p; 854 855 BUG_ON(hdr.minorversion == 0); 856 857 encode_cb_compound4args(xdr, &hdr); 858 encode_cb_sequence4args(xdr, cb, &hdr); 859 860 p = xdr_reserve_space(xdr, 4); 861 *p = cpu_to_be32(OP_CB_NOTIFY_LOCK); 862 encode_nfs_fh4(xdr, &nbl->nbl_fh); 863 encode_stateowner(xdr, &lo->lo_owner); 864 hdr.nops++; 865 866 encode_cb_nops(&hdr); 867 } 868 869 static int nfs4_xdr_dec_cb_notify_lock(struct rpc_rqst *rqstp, 870 struct xdr_stream *xdr, 871 void *data) 872 { 873 struct nfsd4_callback *cb = data; 874 struct nfs4_cb_compound_hdr hdr; 875 int status; 876 877 status = decode_cb_compound4res(xdr, &hdr); 878 if (unlikely(status)) 879 return status; 880 881 status = decode_cb_sequence4res(xdr, cb); 882 if (unlikely(status || cb->cb_seq_status)) 883 return status; 884 885 return decode_cb_op_status(xdr, OP_CB_NOTIFY_LOCK, &cb->cb_status); 886 } 887 888 /* 889 * struct write_response4 { 890 * stateid4 wr_callback_id<1>; 891 * length4 wr_count; 892 * stable_how4 wr_committed; 893 * verifier4 wr_writeverf; 894 * }; 895 * union offload_info4 switch (nfsstat4 coa_status) { 896 * case NFS4_OK: 897 * write_response4 coa_resok4; 898 * default: 899 * length4 coa_bytes_copied; 900 * }; 901 * struct CB_OFFLOAD4args { 902 * nfs_fh4 coa_fh; 903 * stateid4 coa_stateid; 904 * offload_info4 coa_offload_info; 905 * }; 906 */ 907 static void encode_offload_info4(struct xdr_stream *xdr, 908 const struct nfsd4_cb_offload *cbo) 909 { 910 __be32 *p; 911 912 p = xdr_reserve_space(xdr, 4); 913 *p = cbo->co_nfserr; 914 switch (cbo->co_nfserr) { 915 case nfs_ok: 916 p = xdr_reserve_space(xdr, 4 + 8 + 4 + NFS4_VERIFIER_SIZE); 917 p = xdr_encode_empty_array(p); 918 p = xdr_encode_hyper(p, cbo->co_res.wr_bytes_written); 919 *p++ = cpu_to_be32(cbo->co_res.wr_stable_how); 920 p = xdr_encode_opaque_fixed(p, cbo->co_res.wr_verifier.data, 921 NFS4_VERIFIER_SIZE); 922 break; 923 default: 924 p = xdr_reserve_space(xdr, 8); 925 /* We always return success if bytes were written */ 926 p = xdr_encode_hyper(p, 0); 927 } 928 } 929 930 static void encode_cb_offload4args(struct xdr_stream *xdr, 931 const struct nfsd4_cb_offload *cbo, 932 struct nfs4_cb_compound_hdr *hdr) 933 { 934 __be32 *p; 935 936 p = xdr_reserve_space(xdr, 4); 937 *p = cpu_to_be32(OP_CB_OFFLOAD); 938 encode_nfs_fh4(xdr, &cbo->co_fh); 939 encode_stateid4(xdr, &cbo->co_res.cb_stateid); 940 encode_offload_info4(xdr, cbo); 941 942 hdr->nops++; 943 } 944 945 static void nfs4_xdr_enc_cb_offload(struct rpc_rqst *req, 946 struct xdr_stream *xdr, 947 const void *data) 948 { 949 const struct nfsd4_callback *cb = data; 950 const struct nfsd4_cb_offload *cbo = 951 container_of(cb, struct nfsd4_cb_offload, co_cb); 952 struct nfs4_cb_compound_hdr hdr = { 953 .ident = 0, 954 .minorversion = cb->cb_clp->cl_minorversion, 955 }; 956 957 encode_cb_compound4args(xdr, &hdr); 958 encode_cb_sequence4args(xdr, cb, &hdr); 959 encode_cb_offload4args(xdr, cbo, &hdr); 960 encode_cb_nops(&hdr); 961 } 962 963 static int nfs4_xdr_dec_cb_offload(struct rpc_rqst *rqstp, 964 struct xdr_stream *xdr, 965 void *data) 966 { 967 struct nfsd4_callback *cb = data; 968 struct nfs4_cb_compound_hdr hdr; 969 int status; 970 971 status = decode_cb_compound4res(xdr, &hdr); 972 if (unlikely(status)) 973 return status; 974 975 status = decode_cb_sequence4res(xdr, cb); 976 if (unlikely(status || cb->cb_seq_status)) 977 return status; 978 979 return decode_cb_op_status(xdr, OP_CB_OFFLOAD, &cb->cb_status); 980 } 981 /* 982 * RPC procedure tables 983 */ 984 #define PROC(proc, call, argtype, restype) \ 985 [NFSPROC4_CLNT_##proc] = { \ 986 .p_proc = NFSPROC4_CB_##call, \ 987 .p_encode = nfs4_xdr_enc_##argtype, \ 988 .p_decode = nfs4_xdr_dec_##restype, \ 989 .p_arglen = NFS4_enc_##argtype##_sz, \ 990 .p_replen = NFS4_dec_##restype##_sz, \ 991 .p_statidx = NFSPROC4_CB_##call, \ 992 .p_name = #proc, \ 993 } 994 995 static const struct rpc_procinfo nfs4_cb_procedures[] = { 996 PROC(CB_NULL, NULL, cb_null, cb_null), 997 PROC(CB_RECALL, COMPOUND, cb_recall, cb_recall), 998 #ifdef CONFIG_NFSD_PNFS 999 PROC(CB_LAYOUT, COMPOUND, cb_layout, cb_layout), 1000 #endif 1001 PROC(CB_NOTIFY_LOCK, COMPOUND, cb_notify_lock, cb_notify_lock), 1002 PROC(CB_OFFLOAD, COMPOUND, cb_offload, cb_offload), 1003 PROC(CB_RECALL_ANY, COMPOUND, cb_recall_any, cb_recall_any), 1004 PROC(CB_GETATTR, COMPOUND, cb_getattr, cb_getattr), 1005 }; 1006 1007 static unsigned int nfs4_cb_counts[ARRAY_SIZE(nfs4_cb_procedures)]; 1008 static const struct rpc_version nfs_cb_version4 = { 1009 /* 1010 * Note on the callback rpc program version number: despite language in rfc 1011 * 5661 section 18.36.3 requiring servers to use 4 in this field, the 1012 * official xdr descriptions for both 4.0 and 4.1 specify version 1, and 1013 * in practice that appears to be what implementations use. The section 1014 * 18.36.3 language is expected to be fixed in an erratum. 1015 */ 1016 .number = 1, 1017 .nrprocs = ARRAY_SIZE(nfs4_cb_procedures), 1018 .procs = nfs4_cb_procedures, 1019 .counts = nfs4_cb_counts, 1020 }; 1021 1022 static const struct rpc_version *nfs_cb_version[2] = { 1023 [1] = &nfs_cb_version4, 1024 }; 1025 1026 static const struct rpc_program cb_program; 1027 1028 static struct rpc_stat cb_stats = { 1029 .program = &cb_program 1030 }; 1031 1032 #define NFS4_CALLBACK 0x40000000 1033 static const struct rpc_program cb_program = { 1034 .name = "nfs4_cb", 1035 .number = NFS4_CALLBACK, 1036 .nrvers = ARRAY_SIZE(nfs_cb_version), 1037 .version = nfs_cb_version, 1038 .stats = &cb_stats, 1039 .pipe_dir_name = "nfsd4_cb", 1040 }; 1041 1042 static int max_cb_time(struct net *net) 1043 { 1044 struct nfsd_net *nn = net_generic(net, nfsd_net_id); 1045 1046 /* 1047 * nfsd4_lease is set to at most one hour in __nfsd4_write_time, 1048 * so we can use 32-bit math on it. Warn if that assumption 1049 * ever stops being true. 1050 */ 1051 if (WARN_ON_ONCE(nn->nfsd4_lease > 3600)) 1052 return 360 * HZ; 1053 1054 return max(((u32)nn->nfsd4_lease)/10, 1u) * HZ; 1055 } 1056 1057 static bool nfsd4_queue_cb(struct nfsd4_callback *cb) 1058 { 1059 struct nfs4_client *clp = cb->cb_clp; 1060 1061 trace_nfsd_cb_queue(clp, cb); 1062 return queue_work(clp->cl_callback_wq, &cb->cb_work); 1063 } 1064 1065 static void nfsd4_requeue_cb(struct rpc_task *task, struct nfsd4_callback *cb) 1066 { 1067 struct nfs4_client *clp = cb->cb_clp; 1068 1069 if (!test_bit(NFSD4_CLIENT_CB_KILL, &clp->cl_flags)) { 1070 trace_nfsd_cb_restart(clp, cb); 1071 task->tk_status = 0; 1072 set_bit(NFSD4_CALLBACK_REQUEUE, &cb->cb_flags); 1073 } 1074 } 1075 1076 static void nfsd41_cb_inflight_begin(struct nfs4_client *clp) 1077 { 1078 atomic_inc(&clp->cl_cb_inflight); 1079 } 1080 1081 static void nfsd41_cb_inflight_end(struct nfs4_client *clp) 1082 { 1083 1084 atomic_dec_and_wake_up(&clp->cl_cb_inflight); 1085 } 1086 1087 static void nfsd41_cb_inflight_wait_complete(struct nfs4_client *clp) 1088 { 1089 wait_var_event(&clp->cl_cb_inflight, 1090 !atomic_read(&clp->cl_cb_inflight)); 1091 } 1092 1093 static const struct cred *get_backchannel_cred(struct nfs4_client *clp, struct rpc_clnt *client, struct nfsd4_session *ses) 1094 { 1095 if (clp->cl_minorversion == 0) { 1096 client->cl_principal = clp->cl_cred.cr_targ_princ ? 1097 clp->cl_cred.cr_targ_princ : "nfs"; 1098 1099 return get_cred(rpc_machine_cred()); 1100 } else { 1101 struct cred *kcred; 1102 1103 kcred = prepare_kernel_cred(&init_task); 1104 if (!kcred) 1105 return NULL; 1106 1107 kcred->fsuid = ses->se_cb_sec.uid; 1108 kcred->fsgid = ses->se_cb_sec.gid; 1109 return kcred; 1110 } 1111 } 1112 1113 static int setup_callback_client(struct nfs4_client *clp, struct nfs4_cb_conn *conn, struct nfsd4_session *ses) 1114 { 1115 int maxtime = max_cb_time(clp->net); 1116 struct rpc_timeout timeparms = { 1117 .to_initval = maxtime, 1118 .to_retries = 0, 1119 .to_maxval = maxtime, 1120 }; 1121 struct rpc_create_args args = { 1122 .net = clp->net, 1123 .address = (struct sockaddr *) &conn->cb_addr, 1124 .addrsize = conn->cb_addrlen, 1125 .saddress = (struct sockaddr *) &conn->cb_saddr, 1126 .timeout = &timeparms, 1127 .program = &cb_program, 1128 .version = 1, 1129 .flags = (RPC_CLNT_CREATE_NOPING | RPC_CLNT_CREATE_QUIET), 1130 .cred = current_cred(), 1131 }; 1132 struct rpc_clnt *client; 1133 const struct cred *cred; 1134 1135 if (clp->cl_minorversion == 0) { 1136 if (!clp->cl_cred.cr_principal && 1137 (clp->cl_cred.cr_flavor >= RPC_AUTH_GSS_KRB5)) { 1138 trace_nfsd_cb_setup_err(clp, -EINVAL); 1139 return -EINVAL; 1140 } 1141 args.client_name = clp->cl_cred.cr_principal; 1142 args.prognumber = conn->cb_prog; 1143 args.protocol = XPRT_TRANSPORT_TCP; 1144 args.authflavor = clp->cl_cred.cr_flavor; 1145 clp->cl_cb_ident = conn->cb_ident; 1146 } else { 1147 if (!conn->cb_xprt || !ses) 1148 return -EINVAL; 1149 clp->cl_cb_session = ses; 1150 args.bc_xprt = conn->cb_xprt; 1151 args.prognumber = clp->cl_cb_session->se_cb_prog; 1152 args.protocol = conn->cb_xprt->xpt_class->xcl_ident | 1153 XPRT_TRANSPORT_BC; 1154 args.authflavor = ses->se_cb_sec.flavor; 1155 } 1156 /* Create RPC client */ 1157 client = rpc_create(&args); 1158 if (IS_ERR(client)) { 1159 trace_nfsd_cb_setup_err(clp, PTR_ERR(client)); 1160 return PTR_ERR(client); 1161 } 1162 cred = get_backchannel_cred(clp, client, ses); 1163 if (!cred) { 1164 trace_nfsd_cb_setup_err(clp, -ENOMEM); 1165 rpc_shutdown_client(client); 1166 return -ENOMEM; 1167 } 1168 1169 if (clp->cl_minorversion != 0) 1170 clp->cl_cb_conn.cb_xprt = conn->cb_xprt; 1171 clp->cl_cb_client = client; 1172 clp->cl_cb_cred = cred; 1173 rcu_read_lock(); 1174 trace_nfsd_cb_setup(clp, rpc_peeraddr2str(client, RPC_DISPLAY_NETID), 1175 args.authflavor); 1176 rcu_read_unlock(); 1177 return 0; 1178 } 1179 1180 static void nfsd4_mark_cb_state(struct nfs4_client *clp, int newstate) 1181 { 1182 if (clp->cl_cb_state != newstate) { 1183 clp->cl_cb_state = newstate; 1184 trace_nfsd_cb_new_state(clp); 1185 } 1186 } 1187 1188 static void nfsd4_mark_cb_down(struct nfs4_client *clp) 1189 { 1190 if (test_bit(NFSD4_CLIENT_CB_UPDATE, &clp->cl_flags)) 1191 return; 1192 nfsd4_mark_cb_state(clp, NFSD4_CB_DOWN); 1193 } 1194 1195 static void nfsd4_mark_cb_fault(struct nfs4_client *clp) 1196 { 1197 if (test_bit(NFSD4_CLIENT_CB_UPDATE, &clp->cl_flags)) 1198 return; 1199 nfsd4_mark_cb_state(clp, NFSD4_CB_FAULT); 1200 } 1201 1202 static void nfsd4_cb_probe_done(struct rpc_task *task, void *calldata) 1203 { 1204 struct nfs4_client *clp = container_of(calldata, struct nfs4_client, cl_cb_null); 1205 1206 if (task->tk_status) 1207 nfsd4_mark_cb_down(clp); 1208 else 1209 nfsd4_mark_cb_state(clp, NFSD4_CB_UP); 1210 } 1211 1212 static void nfsd4_cb_probe_release(void *calldata) 1213 { 1214 struct nfs4_client *clp = container_of(calldata, struct nfs4_client, cl_cb_null); 1215 1216 nfsd41_cb_inflight_end(clp); 1217 1218 } 1219 1220 static const struct rpc_call_ops nfsd4_cb_probe_ops = { 1221 /* XXX: release method to ensure we set the cb channel down if 1222 * necessary on early failure? */ 1223 .rpc_call_done = nfsd4_cb_probe_done, 1224 .rpc_release = nfsd4_cb_probe_release, 1225 }; 1226 1227 /* 1228 * Poke the callback thread to process any updates to the callback 1229 * parameters, and send a null probe. 1230 */ 1231 void nfsd4_probe_callback(struct nfs4_client *clp) 1232 { 1233 trace_nfsd_cb_probe(clp); 1234 nfsd4_mark_cb_state(clp, NFSD4_CB_UNKNOWN); 1235 set_bit(NFSD4_CLIENT_CB_UPDATE, &clp->cl_flags); 1236 nfsd4_run_cb(&clp->cl_cb_null); 1237 } 1238 1239 void nfsd4_probe_callback_sync(struct nfs4_client *clp) 1240 { 1241 nfsd4_probe_callback(clp); 1242 flush_workqueue(clp->cl_callback_wq); 1243 } 1244 1245 void nfsd4_change_callback(struct nfs4_client *clp, struct nfs4_cb_conn *conn) 1246 { 1247 nfsd4_mark_cb_state(clp, NFSD4_CB_UNKNOWN); 1248 spin_lock(&clp->cl_lock); 1249 memcpy(&clp->cl_cb_conn, conn, sizeof(struct nfs4_cb_conn)); 1250 spin_unlock(&clp->cl_lock); 1251 } 1252 1253 static int grab_slot(struct nfsd4_session *ses) 1254 { 1255 int idx; 1256 1257 spin_lock(&ses->se_lock); 1258 idx = ffs(ses->se_cb_slot_avail) - 1; 1259 if (idx < 0 || idx > ses->se_cb_highest_slot) { 1260 spin_unlock(&ses->se_lock); 1261 return -1; 1262 } 1263 /* clear the bit for the slot */ 1264 ses->se_cb_slot_avail &= ~BIT(idx); 1265 spin_unlock(&ses->se_lock); 1266 return idx; 1267 } 1268 1269 /* 1270 * There's currently a single callback channel slot. 1271 * If the slot is available, then mark it busy. Otherwise, set the 1272 * thread for sleeping on the callback RPC wait queue. 1273 */ 1274 static bool nfsd41_cb_get_slot(struct nfsd4_callback *cb, struct rpc_task *task) 1275 { 1276 struct nfs4_client *clp = cb->cb_clp; 1277 struct nfsd4_session *ses = clp->cl_cb_session; 1278 1279 if (cb->cb_held_slot >= 0) 1280 return true; 1281 cb->cb_held_slot = grab_slot(ses); 1282 if (cb->cb_held_slot < 0) { 1283 rpc_sleep_on(&clp->cl_cb_waitq, task, NULL); 1284 /* Race breaker */ 1285 cb->cb_held_slot = grab_slot(ses); 1286 if (cb->cb_held_slot < 0) 1287 return false; 1288 rpc_wake_up_queued_task(&clp->cl_cb_waitq, task); 1289 } 1290 return true; 1291 } 1292 1293 static void nfsd41_cb_release_slot(struct nfsd4_callback *cb) 1294 { 1295 struct nfs4_client *clp = cb->cb_clp; 1296 struct nfsd4_session *ses = clp->cl_cb_session; 1297 1298 if (cb->cb_held_slot >= 0) { 1299 spin_lock(&ses->se_lock); 1300 ses->se_cb_slot_avail |= BIT(cb->cb_held_slot); 1301 spin_unlock(&ses->se_lock); 1302 cb->cb_held_slot = -1; 1303 rpc_wake_up_next(&clp->cl_cb_waitq); 1304 } 1305 } 1306 1307 static void nfsd41_destroy_cb(struct nfsd4_callback *cb) 1308 { 1309 struct nfs4_client *clp = cb->cb_clp; 1310 1311 trace_nfsd_cb_destroy(clp, cb); 1312 nfsd41_cb_release_slot(cb); 1313 if (test_bit(NFSD4_CALLBACK_WAKE, &cb->cb_flags)) 1314 clear_and_wake_up_bit(NFSD4_CALLBACK_RUNNING, &cb->cb_flags); 1315 else 1316 clear_bit(NFSD4_CALLBACK_RUNNING, &cb->cb_flags); 1317 1318 if (cb->cb_ops && cb->cb_ops->release) 1319 cb->cb_ops->release(cb); 1320 nfsd41_cb_inflight_end(clp); 1321 } 1322 1323 /* 1324 * TODO: cb_sequence should support referring call lists, cachethis, 1325 * and mark callback channel down on communication errors. 1326 */ 1327 static void nfsd4_cb_prepare(struct rpc_task *task, void *calldata) 1328 { 1329 struct nfsd4_callback *cb = calldata; 1330 struct nfs4_client *clp = cb->cb_clp; 1331 u32 minorversion = clp->cl_minorversion; 1332 1333 /* 1334 * cb_seq_status is only set in decode_cb_sequence4res, 1335 * and so will remain 1 if an rpc level failure occurs. 1336 */ 1337 trace_nfsd_cb_rpc_prepare(clp); 1338 cb->cb_seq_status = 1; 1339 cb->cb_status = 0; 1340 if (minorversion && !nfsd41_cb_get_slot(cb, task)) 1341 return; 1342 rpc_call_start(task); 1343 } 1344 1345 /* Returns true if CB_COMPOUND processing should continue */ 1346 static bool nfsd4_cb_sequence_done(struct rpc_task *task, struct nfsd4_callback *cb) 1347 { 1348 struct nfsd4_session *session = cb->cb_clp->cl_cb_session; 1349 bool ret = false; 1350 1351 if (cb->cb_held_slot < 0) 1352 goto requeue; 1353 1354 /* This is the operation status code for CB_SEQUENCE */ 1355 trace_nfsd_cb_seq_status(task, cb); 1356 switch (cb->cb_seq_status) { 1357 case 0: 1358 /* 1359 * No need for lock, access serialized in nfsd4_cb_prepare 1360 * 1361 * RFC5661 20.9.3 1362 * If CB_SEQUENCE returns an error, then the state of the slot 1363 * (sequence ID, cached reply) MUST NOT change. 1364 */ 1365 ++session->se_cb_seq_nr[cb->cb_held_slot]; 1366 ret = true; 1367 break; 1368 case -ESERVERFAULT: 1369 /* 1370 * Call succeeded, but the session, slot index, or slot 1371 * sequence number in the response do not match the same 1372 * in the server's call. The sequence information is thus 1373 * untrustworthy. 1374 */ 1375 nfsd4_mark_cb_fault(cb->cb_clp); 1376 break; 1377 case 1: 1378 /* 1379 * cb_seq_status remains 1 if an RPC Reply was never 1380 * received. NFSD can't know if the client processed 1381 * the CB_SEQUENCE operation. Ask the client to send a 1382 * DESTROY_SESSION to recover. 1383 */ 1384 fallthrough; 1385 case -NFS4ERR_BADSESSION: 1386 nfsd4_mark_cb_fault(cb->cb_clp); 1387 goto requeue; 1388 case -NFS4ERR_DELAY: 1389 cb->cb_seq_status = 1; 1390 if (RPC_SIGNALLED(task) || !rpc_restart_call(task)) 1391 goto requeue; 1392 rpc_delay(task, 2 * HZ); 1393 return false; 1394 case -NFS4ERR_SEQ_MISORDERED: 1395 case -NFS4ERR_BADSLOT: 1396 /* 1397 * A SEQ_MISORDERED or BADSLOT error means that the client and 1398 * server are out of sync as to the backchannel parameters. Mark 1399 * the backchannel faulty and restart the RPC, but leak the slot 1400 * so that it's no longer used. 1401 */ 1402 nfsd4_mark_cb_fault(cb->cb_clp); 1403 cb->cb_held_slot = -1; 1404 goto retry_nowait; 1405 default: 1406 nfsd4_mark_cb_fault(cb->cb_clp); 1407 } 1408 trace_nfsd_cb_free_slot(task, cb); 1409 nfsd41_cb_release_slot(cb); 1410 return ret; 1411 retry_nowait: 1412 /* 1413 * RPC_SIGNALLED() means that the rpc_client is being torn down and 1414 * (possibly) recreated. Requeue the call in that case. 1415 */ 1416 if (!RPC_SIGNALLED(task)) { 1417 if (rpc_restart_call_prepare(task)) 1418 return false; 1419 } 1420 requeue: 1421 nfsd41_cb_release_slot(cb); 1422 nfsd4_requeue_cb(task, cb); 1423 return false; 1424 } 1425 1426 static void nfsd4_cb_done(struct rpc_task *task, void *calldata) 1427 { 1428 struct nfsd4_callback *cb = calldata; 1429 struct nfs4_client *clp = cb->cb_clp; 1430 1431 trace_nfsd_cb_rpc_done(clp); 1432 1433 if (!clp->cl_minorversion) { 1434 /* 1435 * If the backchannel connection was shut down while this 1436 * task was queued, we need to resubmit it after setting up 1437 * a new backchannel connection. 1438 * 1439 * Note that if we lost our callback connection permanently 1440 * the submission code will error out, so we don't need to 1441 * handle that case here. 1442 */ 1443 if (RPC_SIGNALLED(task)) 1444 nfsd4_requeue_cb(task, cb); 1445 } else if (!nfsd4_cb_sequence_done(task, cb)) { 1446 return; 1447 } 1448 1449 if (cb->cb_status) { 1450 WARN_ONCE(task->tk_status, 1451 "cb_status=%d tk_status=%d cb_opcode=%d", 1452 cb->cb_status, task->tk_status, cb->cb_ops->opcode); 1453 task->tk_status = cb->cb_status; 1454 } 1455 1456 switch (cb->cb_ops->done(cb, task)) { 1457 case 0: 1458 task->tk_status = 0; 1459 rpc_restart_call_prepare(task); 1460 return; 1461 case 1: 1462 switch (task->tk_status) { 1463 case -EIO: 1464 case -ETIMEDOUT: 1465 case -EACCES: 1466 nfsd4_mark_cb_down(clp); 1467 } 1468 break; 1469 default: 1470 BUG(); 1471 } 1472 } 1473 1474 static void nfsd4_cb_release(void *calldata) 1475 { 1476 struct nfsd4_callback *cb = calldata; 1477 1478 trace_nfsd_cb_rpc_release(cb->cb_clp); 1479 1480 if (test_bit(NFSD4_CALLBACK_REQUEUE, &cb->cb_flags)) 1481 nfsd4_queue_cb(cb); 1482 else 1483 nfsd41_destroy_cb(cb); 1484 1485 } 1486 1487 static const struct rpc_call_ops nfsd4_cb_ops = { 1488 .rpc_call_prepare = nfsd4_cb_prepare, 1489 .rpc_call_done = nfsd4_cb_done, 1490 .rpc_release = nfsd4_cb_release, 1491 }; 1492 1493 /* must be called under the state lock */ 1494 void nfsd4_shutdown_callback(struct nfs4_client *clp) 1495 { 1496 if (clp->cl_cb_state != NFSD4_CB_UNKNOWN) 1497 trace_nfsd_cb_shutdown(clp); 1498 1499 set_bit(NFSD4_CLIENT_CB_KILL, &clp->cl_flags); 1500 /* 1501 * Note this won't actually result in a null callback; 1502 * instead, nfsd4_run_cb_null() will detect the killed 1503 * client, destroy the rpc client, and stop: 1504 */ 1505 nfsd4_run_cb(&clp->cl_cb_null); 1506 flush_workqueue(clp->cl_callback_wq); 1507 nfsd41_cb_inflight_wait_complete(clp); 1508 } 1509 1510 static struct nfsd4_conn * __nfsd4_find_backchannel(struct nfs4_client *clp) 1511 { 1512 struct nfsd4_session *s; 1513 struct nfsd4_conn *c; 1514 1515 lockdep_assert_held(&clp->cl_lock); 1516 1517 list_for_each_entry(s, &clp->cl_sessions, se_perclnt) { 1518 list_for_each_entry(c, &s->se_conns, cn_persession) { 1519 if (c->cn_flags & NFS4_CDFC4_BACK) 1520 return c; 1521 } 1522 } 1523 return NULL; 1524 } 1525 1526 /* 1527 * Note there isn't a lot of locking in this code; instead we depend on 1528 * the fact that it is run from clp->cl_callback_wq, which won't run two 1529 * work items at once. So, for example, clp->cl_callback_wq handles all 1530 * access of cl_cb_client and all calls to rpc_create or rpc_shutdown_client. 1531 */ 1532 static void nfsd4_process_cb_update(struct nfsd4_callback *cb) 1533 { 1534 struct nfs4_cb_conn conn; 1535 struct nfs4_client *clp = cb->cb_clp; 1536 struct nfsd4_session *ses = NULL; 1537 struct nfsd4_conn *c; 1538 int err; 1539 1540 trace_nfsd_cb_bc_update(clp, cb); 1541 1542 /* 1543 * This is either an update, or the client dying; in either case, 1544 * kill the old client: 1545 */ 1546 if (clp->cl_cb_client) { 1547 trace_nfsd_cb_bc_shutdown(clp, cb); 1548 rpc_shutdown_client(clp->cl_cb_client); 1549 clp->cl_cb_client = NULL; 1550 put_cred(clp->cl_cb_cred); 1551 clp->cl_cb_cred = NULL; 1552 } 1553 if (clp->cl_cb_conn.cb_xprt) { 1554 svc_xprt_put(clp->cl_cb_conn.cb_xprt); 1555 clp->cl_cb_conn.cb_xprt = NULL; 1556 } 1557 if (test_bit(NFSD4_CLIENT_CB_KILL, &clp->cl_flags)) 1558 return; 1559 1560 spin_lock(&clp->cl_lock); 1561 /* 1562 * Only serialized callback code is allowed to clear these 1563 * flags; main nfsd code can only set them: 1564 */ 1565 WARN_ON(!(clp->cl_flags & NFSD4_CLIENT_CB_FLAG_MASK)); 1566 clear_bit(NFSD4_CLIENT_CB_UPDATE, &clp->cl_flags); 1567 1568 memcpy(&conn, &cb->cb_clp->cl_cb_conn, sizeof(struct nfs4_cb_conn)); 1569 c = __nfsd4_find_backchannel(clp); 1570 if (c) { 1571 svc_xprt_get(c->cn_xprt); 1572 conn.cb_xprt = c->cn_xprt; 1573 ses = c->cn_session; 1574 } 1575 spin_unlock(&clp->cl_lock); 1576 1577 err = setup_callback_client(clp, &conn, ses); 1578 if (err) { 1579 nfsd4_mark_cb_down(clp); 1580 if (c) 1581 svc_xprt_put(c->cn_xprt); 1582 return; 1583 } 1584 } 1585 1586 static void 1587 nfsd4_run_cb_work(struct work_struct *work) 1588 { 1589 struct nfsd4_callback *cb = 1590 container_of(work, struct nfsd4_callback, cb_work); 1591 struct nfs4_client *clp = cb->cb_clp; 1592 struct rpc_clnt *clnt; 1593 int flags, ret; 1594 1595 trace_nfsd_cb_start(clp); 1596 1597 if (clp->cl_flags & NFSD4_CLIENT_CB_FLAG_MASK) 1598 nfsd4_process_cb_update(cb); 1599 1600 clnt = clp->cl_cb_client; 1601 if (!clnt || clp->cl_state == NFSD4_COURTESY) { 1602 /* 1603 * Callback channel broken, client killed or 1604 * nfs4_client in courtesy state; give up. 1605 */ 1606 nfsd41_destroy_cb(cb); 1607 return; 1608 } 1609 1610 /* 1611 * Don't send probe messages for 4.1 or later. 1612 */ 1613 if (!cb->cb_ops && clp->cl_minorversion) { 1614 nfsd4_mark_cb_state(clp, NFSD4_CB_UP); 1615 nfsd41_destroy_cb(cb); 1616 return; 1617 } 1618 1619 if (!test_and_clear_bit(NFSD4_CALLBACK_REQUEUE, &cb->cb_flags)) { 1620 if (cb->cb_ops && cb->cb_ops->prepare) 1621 cb->cb_ops->prepare(cb); 1622 } 1623 1624 cb->cb_msg.rpc_cred = clp->cl_cb_cred; 1625 flags = clp->cl_minorversion ? RPC_TASK_NOCONNECT : RPC_TASK_SOFTCONN; 1626 ret = rpc_call_async(clnt, &cb->cb_msg, RPC_TASK_SOFT | flags, 1627 cb->cb_ops ? &nfsd4_cb_ops : &nfsd4_cb_probe_ops, cb); 1628 if (ret != 0) { 1629 set_bit(NFSD4_CALLBACK_REQUEUE, &cb->cb_flags); 1630 nfsd4_queue_cb(cb); 1631 } 1632 } 1633 1634 void nfsd4_init_cb(struct nfsd4_callback *cb, struct nfs4_client *clp, 1635 const struct nfsd4_callback_ops *ops, enum nfsd4_cb_op op) 1636 { 1637 cb->cb_clp = clp; 1638 cb->cb_msg.rpc_proc = &nfs4_cb_procedures[op]; 1639 cb->cb_msg.rpc_argp = cb; 1640 cb->cb_msg.rpc_resp = cb; 1641 cb->cb_flags = 0; 1642 cb->cb_ops = ops; 1643 INIT_WORK(&cb->cb_work, nfsd4_run_cb_work); 1644 cb->cb_status = 0; 1645 cb->cb_held_slot = -1; 1646 } 1647 1648 /** 1649 * nfsd4_run_cb - queue up a callback job to run 1650 * @cb: callback to queue 1651 * 1652 * Kick off a callback to do its thing. Returns false if it was already 1653 * on a queue, true otherwise. 1654 */ 1655 bool nfsd4_run_cb(struct nfsd4_callback *cb) 1656 { 1657 struct nfs4_client *clp = cb->cb_clp; 1658 bool queued; 1659 1660 nfsd41_cb_inflight_begin(clp); 1661 queued = nfsd4_queue_cb(cb); 1662 if (!queued) 1663 nfsd41_cb_inflight_end(clp); 1664 return queued; 1665 } 1666