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 static void 421 encode_referring_call4(struct xdr_stream *xdr, 422 const struct nfsd4_referring_call *rc) 423 { 424 encode_uint32(xdr, rc->rc_sequenceid); 425 encode_uint32(xdr, rc->rc_slotid); 426 } 427 428 static void 429 encode_referring_call_list4(struct xdr_stream *xdr, 430 const struct nfsd4_referring_call_list *rcl) 431 { 432 struct nfsd4_referring_call *rc; 433 __be32 *p; 434 435 p = xdr_reserve_space(xdr, NFS4_MAX_SESSIONID_LEN); 436 xdr_encode_opaque_fixed(p, rcl->rcl_sessionid.data, 437 NFS4_MAX_SESSIONID_LEN); 438 encode_uint32(xdr, rcl->__nr_referring_calls); 439 list_for_each_entry(rc, &rcl->rcl_referring_calls, __list) 440 encode_referring_call4(xdr, rc); 441 } 442 443 /* 444 * CB_SEQUENCE4args 445 * 446 * struct CB_SEQUENCE4args { 447 * sessionid4 csa_sessionid; 448 * sequenceid4 csa_sequenceid; 449 * slotid4 csa_slotid; 450 * slotid4 csa_highest_slotid; 451 * bool csa_cachethis; 452 * referring_call_list4 csa_referring_call_lists<>; 453 * }; 454 */ 455 static void encode_cb_sequence4args(struct xdr_stream *xdr, 456 const struct nfsd4_callback *cb, 457 struct nfs4_cb_compound_hdr *hdr) 458 { 459 struct nfsd4_session *session = cb->cb_clp->cl_cb_session; 460 struct nfsd4_referring_call_list *rcl; 461 __be32 *p; 462 463 if (hdr->minorversion == 0) 464 return; 465 466 encode_nfs_cb_opnum4(xdr, OP_CB_SEQUENCE); 467 encode_sessionid4(xdr, session); 468 469 p = xdr_reserve_space(xdr, XDR_UNIT * 4); 470 *p++ = cpu_to_be32(session->se_cb_seq_nr[cb->cb_held_slot]); /* csa_sequenceid */ 471 *p++ = cpu_to_be32(cb->cb_held_slot); /* csa_slotid */ 472 *p++ = cpu_to_be32(highest_slotid(session)); /* csa_highest_slotid */ 473 *p++ = xdr_zero; /* csa_cachethis */ 474 475 /* csa_referring_call_lists */ 476 encode_uint32(xdr, cb->cb_nr_referring_call_list); 477 list_for_each_entry(rcl, &cb->cb_referring_call_list, __list) 478 encode_referring_call_list4(xdr, rcl); 479 480 hdr->nops++; 481 } 482 483 static void update_cb_slot_table(struct nfsd4_session *ses, u32 target) 484 { 485 /* No need to do anything if nothing changed */ 486 if (likely(target == READ_ONCE(ses->se_cb_highest_slot))) 487 return; 488 489 spin_lock(&ses->se_lock); 490 if (target > ses->se_cb_highest_slot) { 491 int i; 492 493 target = min(target, NFSD_BC_SLOT_TABLE_SIZE - 1); 494 495 /* 496 * Growing the slot table. Reset any new sequences to 1. 497 * 498 * NB: There is some debate about whether the RFC requires this, 499 * but the Linux client expects it. 500 */ 501 for (i = ses->se_cb_highest_slot + 1; i <= target; ++i) 502 ses->se_cb_seq_nr[i] = 1; 503 } 504 ses->se_cb_highest_slot = target; 505 spin_unlock(&ses->se_lock); 506 } 507 508 /* 509 * CB_SEQUENCE4resok 510 * 511 * struct CB_SEQUENCE4resok { 512 * sessionid4 csr_sessionid; 513 * sequenceid4 csr_sequenceid; 514 * slotid4 csr_slotid; 515 * slotid4 csr_highest_slotid; 516 * slotid4 csr_target_highest_slotid; 517 * }; 518 * 519 * union CB_SEQUENCE4res switch (nfsstat4 csr_status) { 520 * case NFS4_OK: 521 * CB_SEQUENCE4resok csr_resok4; 522 * default: 523 * void; 524 * }; 525 * 526 * Our current back channel implmentation supports a single backchannel 527 * with a single slot. 528 */ 529 static int decode_cb_sequence4resok(struct xdr_stream *xdr, 530 struct nfsd4_callback *cb) 531 { 532 struct nfsd4_session *session = cb->cb_clp->cl_cb_session; 533 int status = -ESERVERFAULT; 534 __be32 *p; 535 u32 seqid, slotid, target; 536 537 /* 538 * If the server returns different values for sessionID, slotID or 539 * sequence number, the server is looney tunes. 540 */ 541 p = xdr_inline_decode(xdr, NFS4_MAX_SESSIONID_LEN + 4 + 4 + 4 + 4); 542 if (unlikely(p == NULL)) 543 goto out_overflow; 544 545 if (memcmp(p, session->se_sessionid.data, NFS4_MAX_SESSIONID_LEN)) { 546 dprintk("NFS: %s Invalid session id\n", __func__); 547 goto out; 548 } 549 p += XDR_QUADLEN(NFS4_MAX_SESSIONID_LEN); 550 551 seqid = be32_to_cpup(p++); 552 if (seqid != session->se_cb_seq_nr[cb->cb_held_slot]) { 553 dprintk("NFS: %s Invalid sequence number\n", __func__); 554 goto out; 555 } 556 557 slotid = be32_to_cpup(p++); 558 if (slotid != cb->cb_held_slot) { 559 dprintk("NFS: %s Invalid slotid\n", __func__); 560 goto out; 561 } 562 563 p++; // ignore current highest slot value 564 565 target = be32_to_cpup(p++); 566 update_cb_slot_table(session, target); 567 status = 0; 568 out: 569 cb->cb_seq_status = status; 570 return status; 571 out_overflow: 572 status = -EIO; 573 goto out; 574 } 575 576 static int decode_cb_sequence4res(struct xdr_stream *xdr, 577 struct nfsd4_callback *cb) 578 { 579 int status; 580 581 if (cb->cb_clp->cl_minorversion == 0) 582 return 0; 583 584 status = decode_cb_op_status(xdr, OP_CB_SEQUENCE, &cb->cb_seq_status); 585 if (unlikely(status || cb->cb_seq_status)) 586 return status; 587 588 return decode_cb_sequence4resok(xdr, cb); 589 } 590 591 /* 592 * NFSv4.0 and NFSv4.1 XDR encode functions 593 * 594 * NFSv4.0 callback argument types are defined in section 15 of RFC 595 * 3530: "Network File System (NFS) version 4 Protocol" and section 20 596 * of RFC 5661: "Network File System (NFS) Version 4 Minor Version 1 597 * Protocol". 598 */ 599 600 /* 601 * NB: Without this zero space reservation, callbacks over krb5p fail 602 */ 603 static void nfs4_xdr_enc_cb_null(struct rpc_rqst *req, struct xdr_stream *xdr, 604 const void *__unused) 605 { 606 xdr_reserve_space(xdr, 0); 607 } 608 609 /* 610 * 20.1. Operation 3: CB_GETATTR - Get Attributes 611 */ 612 static void nfs4_xdr_enc_cb_getattr(struct rpc_rqst *req, 613 struct xdr_stream *xdr, const void *data) 614 { 615 const struct nfsd4_callback *cb = data; 616 struct nfs4_cb_fattr *ncf = 617 container_of(cb, struct nfs4_cb_fattr, ncf_getattr); 618 struct nfs4_cb_compound_hdr hdr = { 619 .ident = cb->cb_clp->cl_cb_ident, 620 .minorversion = cb->cb_clp->cl_minorversion, 621 }; 622 623 encode_cb_compound4args(xdr, &hdr); 624 encode_cb_sequence4args(xdr, cb, &hdr); 625 encode_cb_getattr4args(xdr, &hdr, ncf); 626 encode_cb_nops(&hdr); 627 } 628 629 /* 630 * 20.2. Operation 4: CB_RECALL - Recall a Delegation 631 */ 632 static void nfs4_xdr_enc_cb_recall(struct rpc_rqst *req, struct xdr_stream *xdr, 633 const void *data) 634 { 635 const struct nfsd4_callback *cb = data; 636 const struct nfs4_delegation *dp = cb_to_delegation(cb); 637 struct nfs4_cb_compound_hdr hdr = { 638 .ident = cb->cb_clp->cl_cb_ident, 639 .minorversion = cb->cb_clp->cl_minorversion, 640 }; 641 642 encode_cb_compound4args(xdr, &hdr); 643 encode_cb_sequence4args(xdr, cb, &hdr); 644 encode_cb_recall4args(xdr, dp, &hdr); 645 encode_cb_nops(&hdr); 646 } 647 648 /* 649 * 20.6. Operation 8: CB_RECALL_ANY - Keep Any N Recallable Objects 650 */ 651 static void 652 nfs4_xdr_enc_cb_recall_any(struct rpc_rqst *req, 653 struct xdr_stream *xdr, const void *data) 654 { 655 const struct nfsd4_callback *cb = data; 656 struct nfsd4_cb_recall_any *ra; 657 struct nfs4_cb_compound_hdr hdr = { 658 .ident = cb->cb_clp->cl_cb_ident, 659 .minorversion = cb->cb_clp->cl_minorversion, 660 }; 661 662 ra = container_of(cb, struct nfsd4_cb_recall_any, ra_cb); 663 encode_cb_compound4args(xdr, &hdr); 664 encode_cb_sequence4args(xdr, cb, &hdr); 665 encode_cb_recallany4args(xdr, &hdr, ra); 666 encode_cb_nops(&hdr); 667 } 668 669 /* 670 * NFSv4.0 and NFSv4.1 XDR decode functions 671 * 672 * NFSv4.0 callback result types are defined in section 15 of RFC 673 * 3530: "Network File System (NFS) version 4 Protocol" and section 20 674 * of RFC 5661: "Network File System (NFS) Version 4 Minor Version 1 675 * Protocol". 676 */ 677 678 static int nfs4_xdr_dec_cb_null(struct rpc_rqst *req, struct xdr_stream *xdr, 679 void *__unused) 680 { 681 return 0; 682 } 683 684 /* 685 * 20.1. Operation 3: CB_GETATTR - Get Attributes 686 */ 687 static int nfs4_xdr_dec_cb_getattr(struct rpc_rqst *rqstp, 688 struct xdr_stream *xdr, 689 void *data) 690 { 691 struct nfsd4_callback *cb = data; 692 struct nfs4_cb_compound_hdr hdr; 693 int status; 694 u32 bitmap[3] = {0}; 695 u32 attrlen, maxlen; 696 struct nfs4_cb_fattr *ncf = 697 container_of(cb, struct nfs4_cb_fattr, ncf_getattr); 698 699 status = decode_cb_compound4res(xdr, &hdr); 700 if (unlikely(status)) 701 return status; 702 703 status = decode_cb_sequence4res(xdr, cb); 704 if (unlikely(status || cb->cb_seq_status)) 705 return status; 706 707 status = decode_cb_op_status(xdr, OP_CB_GETATTR, &cb->cb_status); 708 if (unlikely(status || cb->cb_status)) 709 return status; 710 if (xdr_stream_decode_uint32_array(xdr, bitmap, 3) < 0) 711 return -EIO; 712 if (xdr_stream_decode_u32(xdr, &attrlen) < 0) 713 return -EIO; 714 maxlen = sizeof(ncf->ncf_cb_change) + sizeof(ncf->ncf_cb_fsize); 715 if (bitmap[2] != 0) 716 maxlen += (sizeof(ncf->ncf_cb_mtime.tv_sec) + 717 sizeof(ncf->ncf_cb_mtime.tv_nsec)) * 2; 718 if (attrlen > maxlen) 719 return -EIO; 720 status = decode_cb_fattr4(xdr, bitmap, ncf); 721 return status; 722 } 723 724 /* 725 * 20.2. Operation 4: CB_RECALL - Recall a Delegation 726 */ 727 static int nfs4_xdr_dec_cb_recall(struct rpc_rqst *rqstp, 728 struct xdr_stream *xdr, 729 void *data) 730 { 731 struct nfsd4_callback *cb = data; 732 struct nfs4_cb_compound_hdr hdr; 733 int status; 734 735 status = decode_cb_compound4res(xdr, &hdr); 736 if (unlikely(status)) 737 return status; 738 739 status = decode_cb_sequence4res(xdr, cb); 740 if (unlikely(status || cb->cb_seq_status)) 741 return status; 742 743 return decode_cb_op_status(xdr, OP_CB_RECALL, &cb->cb_status); 744 } 745 746 /* 747 * 20.6. Operation 8: CB_RECALL_ANY - Keep Any N Recallable Objects 748 */ 749 static int 750 nfs4_xdr_dec_cb_recall_any(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 status = decode_cb_sequence4res(xdr, cb); 762 if (unlikely(status || cb->cb_seq_status)) 763 return status; 764 status = decode_cb_op_status(xdr, OP_CB_RECALL_ANY, &cb->cb_status); 765 return status; 766 } 767 768 #ifdef CONFIG_NFSD_PNFS 769 /* 770 * CB_LAYOUTRECALL4args 771 * 772 * struct layoutrecall_file4 { 773 * nfs_fh4 lor_fh; 774 * offset4 lor_offset; 775 * length4 lor_length; 776 * stateid4 lor_stateid; 777 * }; 778 * 779 * union layoutrecall4 switch(layoutrecall_type4 lor_recalltype) { 780 * case LAYOUTRECALL4_FILE: 781 * layoutrecall_file4 lor_layout; 782 * case LAYOUTRECALL4_FSID: 783 * fsid4 lor_fsid; 784 * case LAYOUTRECALL4_ALL: 785 * void; 786 * }; 787 * 788 * struct CB_LAYOUTRECALL4args { 789 * layouttype4 clora_type; 790 * layoutiomode4 clora_iomode; 791 * bool clora_changed; 792 * layoutrecall4 clora_recall; 793 * }; 794 */ 795 static void encode_cb_layout4args(struct xdr_stream *xdr, 796 const struct nfs4_layout_stateid *ls, 797 struct nfs4_cb_compound_hdr *hdr) 798 { 799 __be32 *p; 800 801 BUG_ON(hdr->minorversion == 0); 802 803 p = xdr_reserve_space(xdr, 5 * 4); 804 *p++ = cpu_to_be32(OP_CB_LAYOUTRECALL); 805 *p++ = cpu_to_be32(ls->ls_layout_type); 806 *p++ = cpu_to_be32(IOMODE_ANY); 807 *p++ = cpu_to_be32(1); 808 *p = cpu_to_be32(RETURN_FILE); 809 810 encode_nfs_fh4(xdr, &ls->ls_stid.sc_file->fi_fhandle); 811 812 p = xdr_reserve_space(xdr, 2 * 8); 813 p = xdr_encode_hyper(p, 0); 814 xdr_encode_hyper(p, NFS4_MAX_UINT64); 815 816 encode_stateid4(xdr, &ls->ls_recall_sid); 817 818 hdr->nops++; 819 } 820 821 static void nfs4_xdr_enc_cb_layout(struct rpc_rqst *req, 822 struct xdr_stream *xdr, 823 const void *data) 824 { 825 const struct nfsd4_callback *cb = data; 826 const struct nfs4_layout_stateid *ls = 827 container_of(cb, struct nfs4_layout_stateid, ls_recall); 828 struct nfs4_cb_compound_hdr hdr = { 829 .ident = 0, 830 .minorversion = cb->cb_clp->cl_minorversion, 831 }; 832 833 encode_cb_compound4args(xdr, &hdr); 834 encode_cb_sequence4args(xdr, cb, &hdr); 835 encode_cb_layout4args(xdr, ls, &hdr); 836 encode_cb_nops(&hdr); 837 } 838 839 static int nfs4_xdr_dec_cb_layout(struct rpc_rqst *rqstp, 840 struct xdr_stream *xdr, 841 void *data) 842 { 843 struct nfsd4_callback *cb = data; 844 struct nfs4_cb_compound_hdr hdr; 845 int status; 846 847 status = decode_cb_compound4res(xdr, &hdr); 848 if (unlikely(status)) 849 return status; 850 851 status = decode_cb_sequence4res(xdr, cb); 852 if (unlikely(status || cb->cb_seq_status)) 853 return status; 854 855 return decode_cb_op_status(xdr, OP_CB_LAYOUTRECALL, &cb->cb_status); 856 } 857 #endif /* CONFIG_NFSD_PNFS */ 858 859 static void encode_stateowner(struct xdr_stream *xdr, struct nfs4_stateowner *so) 860 { 861 __be32 *p; 862 863 p = xdr_reserve_space(xdr, 8 + 4 + so->so_owner.len); 864 p = xdr_encode_opaque_fixed(p, &so->so_client->cl_clientid, 8); 865 xdr_encode_opaque(p, so->so_owner.data, so->so_owner.len); 866 } 867 868 static void nfs4_xdr_enc_cb_notify_lock(struct rpc_rqst *req, 869 struct xdr_stream *xdr, 870 const void *data) 871 { 872 const struct nfsd4_callback *cb = data; 873 const struct nfsd4_blocked_lock *nbl = 874 container_of(cb, struct nfsd4_blocked_lock, nbl_cb); 875 struct nfs4_lockowner *lo = (struct nfs4_lockowner *)nbl->nbl_lock.c.flc_owner; 876 struct nfs4_cb_compound_hdr hdr = { 877 .ident = 0, 878 .minorversion = cb->cb_clp->cl_minorversion, 879 }; 880 881 __be32 *p; 882 883 BUG_ON(hdr.minorversion == 0); 884 885 encode_cb_compound4args(xdr, &hdr); 886 encode_cb_sequence4args(xdr, cb, &hdr); 887 888 p = xdr_reserve_space(xdr, 4); 889 *p = cpu_to_be32(OP_CB_NOTIFY_LOCK); 890 encode_nfs_fh4(xdr, &nbl->nbl_fh); 891 encode_stateowner(xdr, &lo->lo_owner); 892 hdr.nops++; 893 894 encode_cb_nops(&hdr); 895 } 896 897 static int nfs4_xdr_dec_cb_notify_lock(struct rpc_rqst *rqstp, 898 struct xdr_stream *xdr, 899 void *data) 900 { 901 struct nfsd4_callback *cb = data; 902 struct nfs4_cb_compound_hdr hdr; 903 int status; 904 905 status = decode_cb_compound4res(xdr, &hdr); 906 if (unlikely(status)) 907 return status; 908 909 status = decode_cb_sequence4res(xdr, cb); 910 if (unlikely(status || cb->cb_seq_status)) 911 return status; 912 913 return decode_cb_op_status(xdr, OP_CB_NOTIFY_LOCK, &cb->cb_status); 914 } 915 916 /* 917 * struct write_response4 { 918 * stateid4 wr_callback_id<1>; 919 * length4 wr_count; 920 * stable_how4 wr_committed; 921 * verifier4 wr_writeverf; 922 * }; 923 * union offload_info4 switch (nfsstat4 coa_status) { 924 * case NFS4_OK: 925 * write_response4 coa_resok4; 926 * default: 927 * length4 coa_bytes_copied; 928 * }; 929 * struct CB_OFFLOAD4args { 930 * nfs_fh4 coa_fh; 931 * stateid4 coa_stateid; 932 * offload_info4 coa_offload_info; 933 * }; 934 */ 935 static void encode_offload_info4(struct xdr_stream *xdr, 936 const struct nfsd4_cb_offload *cbo) 937 { 938 __be32 *p; 939 940 p = xdr_reserve_space(xdr, 4); 941 *p = cbo->co_nfserr; 942 switch (cbo->co_nfserr) { 943 case nfs_ok: 944 p = xdr_reserve_space(xdr, 4 + 8 + 4 + NFS4_VERIFIER_SIZE); 945 p = xdr_encode_empty_array(p); 946 p = xdr_encode_hyper(p, cbo->co_res.wr_bytes_written); 947 *p++ = cpu_to_be32(cbo->co_res.wr_stable_how); 948 p = xdr_encode_opaque_fixed(p, cbo->co_res.wr_verifier.data, 949 NFS4_VERIFIER_SIZE); 950 break; 951 default: 952 p = xdr_reserve_space(xdr, 8); 953 /* We always return success if bytes were written */ 954 p = xdr_encode_hyper(p, 0); 955 } 956 } 957 958 static void encode_cb_offload4args(struct xdr_stream *xdr, 959 const struct nfsd4_cb_offload *cbo, 960 struct nfs4_cb_compound_hdr *hdr) 961 { 962 __be32 *p; 963 964 p = xdr_reserve_space(xdr, 4); 965 *p = cpu_to_be32(OP_CB_OFFLOAD); 966 encode_nfs_fh4(xdr, &cbo->co_fh); 967 encode_stateid4(xdr, &cbo->co_res.cb_stateid); 968 encode_offload_info4(xdr, cbo); 969 970 hdr->nops++; 971 } 972 973 static void nfs4_xdr_enc_cb_offload(struct rpc_rqst *req, 974 struct xdr_stream *xdr, 975 const void *data) 976 { 977 const struct nfsd4_callback *cb = data; 978 const struct nfsd4_cb_offload *cbo = 979 container_of(cb, struct nfsd4_cb_offload, co_cb); 980 struct nfs4_cb_compound_hdr hdr = { 981 .ident = 0, 982 .minorversion = cb->cb_clp->cl_minorversion, 983 }; 984 985 encode_cb_compound4args(xdr, &hdr); 986 encode_cb_sequence4args(xdr, cb, &hdr); 987 encode_cb_offload4args(xdr, cbo, &hdr); 988 encode_cb_nops(&hdr); 989 } 990 991 static int nfs4_xdr_dec_cb_offload(struct rpc_rqst *rqstp, 992 struct xdr_stream *xdr, 993 void *data) 994 { 995 struct nfsd4_callback *cb = data; 996 struct nfs4_cb_compound_hdr hdr; 997 int status; 998 999 status = decode_cb_compound4res(xdr, &hdr); 1000 if (unlikely(status)) 1001 return status; 1002 1003 status = decode_cb_sequence4res(xdr, cb); 1004 if (unlikely(status || cb->cb_seq_status)) 1005 return status; 1006 1007 return decode_cb_op_status(xdr, OP_CB_OFFLOAD, &cb->cb_status); 1008 } 1009 /* 1010 * RPC procedure tables 1011 */ 1012 #define PROC(proc, call, argtype, restype) \ 1013 [NFSPROC4_CLNT_##proc] = { \ 1014 .p_proc = NFSPROC4_CB_##call, \ 1015 .p_encode = nfs4_xdr_enc_##argtype, \ 1016 .p_decode = nfs4_xdr_dec_##restype, \ 1017 .p_arglen = NFS4_enc_##argtype##_sz, \ 1018 .p_replen = NFS4_dec_##restype##_sz, \ 1019 .p_statidx = NFSPROC4_CLNT_##proc, \ 1020 .p_name = #proc, \ 1021 } 1022 1023 static const struct rpc_procinfo nfs4_cb_procedures[] = { 1024 PROC(CB_NULL, NULL, cb_null, cb_null), 1025 PROC(CB_RECALL, COMPOUND, cb_recall, cb_recall), 1026 #ifdef CONFIG_NFSD_PNFS 1027 PROC(CB_LAYOUT, COMPOUND, cb_layout, cb_layout), 1028 #endif 1029 PROC(CB_NOTIFY_LOCK, COMPOUND, cb_notify_lock, cb_notify_lock), 1030 PROC(CB_OFFLOAD, COMPOUND, cb_offload, cb_offload), 1031 PROC(CB_RECALL_ANY, COMPOUND, cb_recall_any, cb_recall_any), 1032 PROC(CB_GETATTR, COMPOUND, cb_getattr, cb_getattr), 1033 }; 1034 1035 #define NFS4_CB_PROGRAM 0x40000000 1036 #define NFS4_CB_VERSION 1 1037 1038 struct nfsd_net_cb { 1039 struct rpc_version version4; 1040 const struct rpc_version *versions[NFS4_CB_VERSION + 1]; 1041 struct rpc_program program; 1042 struct rpc_stat stat; 1043 }; 1044 1045 static int max_cb_time(struct net *net) 1046 { 1047 struct nfsd_net *nn = net_generic(net, nfsd_net_id); 1048 1049 /* 1050 * nfsd4_lease is set to at most one hour in __nfsd4_write_time, 1051 * so we can use 32-bit math on it. Warn if that assumption 1052 * ever stops being true. 1053 */ 1054 if (WARN_ON_ONCE(nn->nfsd4_lease > 3600)) 1055 return 360 * HZ; 1056 1057 return max(((u32)nn->nfsd4_lease)/10, 1u) * HZ; 1058 } 1059 1060 static bool nfsd4_queue_cb(struct nfsd4_callback *cb) 1061 { 1062 struct nfs4_client *clp = cb->cb_clp; 1063 1064 trace_nfsd_cb_queue(clp, cb); 1065 return queue_work(clp->cl_callback_wq, &cb->cb_work); 1066 } 1067 1068 static void nfsd4_requeue_cb(struct rpc_task *task, struct nfsd4_callback *cb) 1069 { 1070 struct nfs4_client *clp = cb->cb_clp; 1071 1072 if (!test_bit(NFSD4_CLIENT_CB_KILL, &clp->cl_flags)) { 1073 trace_nfsd_cb_restart(clp, cb); 1074 task->tk_status = 0; 1075 set_bit(NFSD4_CALLBACK_REQUEUE, &cb->cb_flags); 1076 } 1077 } 1078 1079 static void nfsd41_cb_inflight_begin(struct nfs4_client *clp) 1080 { 1081 atomic_inc(&clp->cl_cb_inflight); 1082 } 1083 1084 static void nfsd41_cb_inflight_end(struct nfs4_client *clp) 1085 { 1086 1087 atomic_dec_and_wake_up(&clp->cl_cb_inflight); 1088 } 1089 1090 static void nfsd41_cb_inflight_wait_complete(struct nfs4_client *clp) 1091 { 1092 wait_var_event(&clp->cl_cb_inflight, 1093 !atomic_read(&clp->cl_cb_inflight)); 1094 } 1095 1096 static const struct cred *get_backchannel_cred(struct nfs4_client *clp, struct rpc_clnt *client, struct nfsd4_session *ses) 1097 { 1098 if (clp->cl_minorversion == 0) { 1099 client->cl_principal = clp->cl_cred.cr_targ_princ ? 1100 clp->cl_cred.cr_targ_princ : "nfs"; 1101 1102 return get_cred(rpc_machine_cred()); 1103 } else { 1104 struct cred *kcred; 1105 1106 kcred = prepare_kernel_cred(&init_task); 1107 if (!kcred) 1108 return NULL; 1109 1110 kcred->fsuid = ses->se_cb_sec.uid; 1111 kcred->fsgid = ses->se_cb_sec.gid; 1112 return kcred; 1113 } 1114 } 1115 1116 static int setup_callback_client(struct nfs4_client *clp, struct nfs4_cb_conn *conn, struct nfsd4_session *ses) 1117 { 1118 struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id); 1119 int maxtime = max_cb_time(clp->net); 1120 struct rpc_timeout timeparms = { 1121 .to_initval = maxtime, 1122 .to_retries = 0, 1123 .to_maxval = maxtime, 1124 }; 1125 struct rpc_create_args args = { 1126 .net = clp->net, 1127 .address = (struct sockaddr *) &conn->cb_addr, 1128 .addrsize = conn->cb_addrlen, 1129 .saddress = (struct sockaddr *) &conn->cb_saddr, 1130 .timeout = &timeparms, 1131 .version = NFS4_CB_VERSION, 1132 .flags = (RPC_CLNT_CREATE_NOPING | RPC_CLNT_CREATE_QUIET), 1133 .cred = current_cred(), 1134 }; 1135 struct rpc_clnt *client; 1136 const struct cred *cred; 1137 1138 args.program = &nn->nfsd_cb->program; 1139 if (clp->cl_minorversion == 0) { 1140 if (!clp->cl_cred.cr_principal && 1141 (clp->cl_cred.cr_flavor >= RPC_AUTH_GSS_KRB5)) { 1142 trace_nfsd_cb_setup_err(clp, -EINVAL); 1143 return -EINVAL; 1144 } 1145 args.client_name = clp->cl_cred.cr_principal; 1146 args.prognumber = conn->cb_prog; 1147 args.protocol = XPRT_TRANSPORT_TCP; 1148 args.authflavor = clp->cl_cred.cr_flavor; 1149 clp->cl_cb_ident = conn->cb_ident; 1150 } else { 1151 if (!conn->cb_xprt || !ses) 1152 return -EINVAL; 1153 clp->cl_cb_session = ses; 1154 args.bc_xprt = conn->cb_xprt; 1155 args.prognumber = clp->cl_cb_session->se_cb_prog; 1156 args.protocol = conn->cb_xprt->xpt_class->xcl_ident | 1157 XPRT_TRANSPORT_BC; 1158 args.authflavor = ses->se_cb_sec.flavor; 1159 } 1160 /* Create RPC client */ 1161 client = rpc_create(&args); 1162 if (IS_ERR(client)) { 1163 trace_nfsd_cb_setup_err(clp, PTR_ERR(client)); 1164 return PTR_ERR(client); 1165 } 1166 cred = get_backchannel_cred(clp, client, ses); 1167 if (!cred) { 1168 trace_nfsd_cb_setup_err(clp, -ENOMEM); 1169 rpc_shutdown_client(client); 1170 return -ENOMEM; 1171 } 1172 1173 if (clp->cl_minorversion != 0) 1174 clp->cl_cb_conn.cb_xprt = conn->cb_xprt; 1175 clp->cl_cb_client = client; 1176 clp->cl_cb_cred = cred; 1177 rcu_read_lock(); 1178 trace_nfsd_cb_setup(clp, rpc_peeraddr2str(client, RPC_DISPLAY_NETID), 1179 args.authflavor); 1180 rcu_read_unlock(); 1181 return 0; 1182 } 1183 1184 static void nfsd4_mark_cb_state(struct nfs4_client *clp, int newstate) 1185 { 1186 if (clp->cl_cb_state != newstate) { 1187 clp->cl_cb_state = newstate; 1188 trace_nfsd_cb_new_state(clp); 1189 } 1190 } 1191 1192 static void nfsd4_mark_cb_down(struct nfs4_client *clp) 1193 { 1194 if (test_bit(NFSD4_CLIENT_CB_UPDATE, &clp->cl_flags)) 1195 return; 1196 nfsd4_mark_cb_state(clp, NFSD4_CB_DOWN); 1197 } 1198 1199 static void nfsd4_mark_cb_fault(struct nfs4_client *clp) 1200 { 1201 if (test_bit(NFSD4_CLIENT_CB_UPDATE, &clp->cl_flags)) 1202 return; 1203 nfsd4_mark_cb_state(clp, NFSD4_CB_FAULT); 1204 } 1205 1206 static void nfsd4_cb_probe_done(struct rpc_task *task, void *calldata) 1207 { 1208 struct nfs4_client *clp = container_of(calldata, struct nfs4_client, cl_cb_null); 1209 1210 if (task->tk_status) 1211 nfsd4_mark_cb_down(clp); 1212 else 1213 nfsd4_mark_cb_state(clp, NFSD4_CB_UP); 1214 } 1215 1216 static void nfsd4_cb_probe_release(void *calldata) 1217 { 1218 struct nfs4_client *clp = container_of(calldata, struct nfs4_client, cl_cb_null); 1219 1220 nfsd41_cb_inflight_end(clp); 1221 1222 } 1223 1224 static const struct rpc_call_ops nfsd4_cb_probe_ops = { 1225 /* XXX: release method to ensure we set the cb channel down if 1226 * necessary on early failure? */ 1227 .rpc_call_done = nfsd4_cb_probe_done, 1228 .rpc_release = nfsd4_cb_probe_release, 1229 }; 1230 1231 /* 1232 * Poke the callback thread to process any updates to the callback 1233 * parameters, and send a null probe. 1234 */ 1235 void nfsd4_probe_callback(struct nfs4_client *clp) 1236 { 1237 trace_nfsd_cb_probe(clp); 1238 nfsd4_mark_cb_state(clp, NFSD4_CB_UNKNOWN); 1239 set_bit(NFSD4_CLIENT_CB_UPDATE, &clp->cl_flags); 1240 nfsd4_run_cb(&clp->cl_cb_null); 1241 } 1242 1243 void nfsd4_probe_callback_sync(struct nfs4_client *clp) 1244 { 1245 nfsd4_probe_callback(clp); 1246 flush_workqueue(clp->cl_callback_wq); 1247 } 1248 1249 void nfsd4_change_callback(struct nfs4_client *clp, struct nfs4_cb_conn *conn) 1250 { 1251 nfsd4_mark_cb_state(clp, NFSD4_CB_UNKNOWN); 1252 spin_lock(&clp->cl_lock); 1253 memcpy(&clp->cl_cb_conn, conn, sizeof(struct nfs4_cb_conn)); 1254 spin_unlock(&clp->cl_lock); 1255 } 1256 1257 static int grab_slot(struct nfsd4_session *ses) 1258 { 1259 int idx; 1260 1261 spin_lock(&ses->se_lock); 1262 idx = ffs(ses->se_cb_slot_avail) - 1; 1263 if (idx < 0 || idx > ses->se_cb_highest_slot) { 1264 spin_unlock(&ses->se_lock); 1265 return -1; 1266 } 1267 /* clear the bit for the slot */ 1268 ses->se_cb_slot_avail &= ~BIT(idx); 1269 spin_unlock(&ses->se_lock); 1270 return idx; 1271 } 1272 1273 /* 1274 * There's currently a single callback channel slot. 1275 * If the slot is available, then mark it busy. Otherwise, set the 1276 * thread for sleeping on the callback RPC wait queue. 1277 */ 1278 static bool nfsd41_cb_get_slot(struct nfsd4_callback *cb, struct rpc_task *task) 1279 { 1280 struct nfs4_client *clp = cb->cb_clp; 1281 struct nfsd4_session *ses = clp->cl_cb_session; 1282 1283 if (cb->cb_held_slot >= 0) 1284 return true; 1285 cb->cb_held_slot = grab_slot(ses); 1286 if (cb->cb_held_slot < 0) { 1287 rpc_sleep_on(&clp->cl_cb_waitq, task, NULL); 1288 /* Race breaker */ 1289 cb->cb_held_slot = grab_slot(ses); 1290 if (cb->cb_held_slot < 0) 1291 return false; 1292 rpc_wake_up_queued_task(&clp->cl_cb_waitq, task); 1293 } 1294 return true; 1295 } 1296 1297 static void nfsd41_cb_release_slot(struct nfsd4_callback *cb) 1298 { 1299 struct nfs4_client *clp = cb->cb_clp; 1300 struct nfsd4_session *ses = clp->cl_cb_session; 1301 1302 if (cb->cb_held_slot >= 0) { 1303 spin_lock(&ses->se_lock); 1304 ses->se_cb_slot_avail |= BIT(cb->cb_held_slot); 1305 spin_unlock(&ses->se_lock); 1306 cb->cb_held_slot = -1; 1307 rpc_wake_up_next(&clp->cl_cb_waitq); 1308 } 1309 } 1310 1311 static void nfsd41_destroy_cb(struct nfsd4_callback *cb) 1312 { 1313 struct nfs4_client *clp = cb->cb_clp; 1314 1315 trace_nfsd_cb_destroy(clp, cb); 1316 nfsd41_cb_release_slot(cb); 1317 if (test_bit(NFSD4_CALLBACK_WAKE, &cb->cb_flags)) 1318 clear_and_wake_up_bit(NFSD4_CALLBACK_RUNNING, &cb->cb_flags); 1319 else 1320 clear_bit(NFSD4_CALLBACK_RUNNING, &cb->cb_flags); 1321 1322 if (cb->cb_ops && cb->cb_ops->release) 1323 cb->cb_ops->release(cb); 1324 nfsd41_cb_inflight_end(clp); 1325 } 1326 1327 /** 1328 * nfsd41_cb_referring_call - add a referring call to a callback operation 1329 * @cb: context of callback to add the rc to 1330 * @sessionid: referring call's session ID 1331 * @slotid: referring call's session slot index 1332 * @seqno: referring call's slot sequence number 1333 * 1334 * Caller serializes access to @cb. 1335 * 1336 * NB: If memory allocation fails, the referring call is not added. 1337 */ 1338 void nfsd41_cb_referring_call(struct nfsd4_callback *cb, 1339 struct nfs4_sessionid *sessionid, 1340 u32 slotid, u32 seqno) 1341 { 1342 struct nfsd4_referring_call_list *rcl; 1343 struct nfsd4_referring_call *rc; 1344 bool found; 1345 1346 might_sleep(); 1347 1348 found = false; 1349 list_for_each_entry(rcl, &cb->cb_referring_call_list, __list) { 1350 if (!memcmp(rcl->rcl_sessionid.data, sessionid->data, 1351 NFS4_MAX_SESSIONID_LEN)) { 1352 found = true; 1353 break; 1354 } 1355 } 1356 if (!found) { 1357 rcl = kmalloc_obj(*rcl); 1358 if (!rcl) 1359 return; 1360 memcpy(rcl->rcl_sessionid.data, sessionid->data, 1361 NFS4_MAX_SESSIONID_LEN); 1362 rcl->__nr_referring_calls = 0; 1363 INIT_LIST_HEAD(&rcl->rcl_referring_calls); 1364 list_add(&rcl->__list, &cb->cb_referring_call_list); 1365 cb->cb_nr_referring_call_list++; 1366 } 1367 1368 found = false; 1369 list_for_each_entry(rc, &rcl->rcl_referring_calls, __list) { 1370 if (rc->rc_sequenceid == seqno && rc->rc_slotid == slotid) { 1371 found = true; 1372 break; 1373 } 1374 } 1375 if (!found) { 1376 rc = kmalloc_obj(*rc); 1377 if (!rc) 1378 goto out; 1379 rc->rc_sequenceid = seqno; 1380 rc->rc_slotid = slotid; 1381 rcl->__nr_referring_calls++; 1382 list_add(&rc->__list, &rcl->rcl_referring_calls); 1383 } 1384 1385 out: 1386 if (!rcl->__nr_referring_calls) { 1387 cb->cb_nr_referring_call_list--; 1388 list_del(&rcl->__list); 1389 kfree(rcl); 1390 } 1391 } 1392 1393 /** 1394 * nfsd41_cb_destroy_referring_call_list - release referring call info 1395 * @cb: context of a callback that has completed 1396 * 1397 * Callers who allocate referring calls using nfsd41_cb_referring_call() must 1398 * release those resources by calling nfsd41_cb_destroy_referring_call_list. 1399 * 1400 * Caller serializes access to @cb. 1401 */ 1402 void nfsd41_cb_destroy_referring_call_list(struct nfsd4_callback *cb) 1403 { 1404 struct nfsd4_referring_call_list *rcl; 1405 struct nfsd4_referring_call *rc; 1406 1407 while (!list_empty(&cb->cb_referring_call_list)) { 1408 rcl = list_first_entry(&cb->cb_referring_call_list, 1409 struct nfsd4_referring_call_list, 1410 __list); 1411 1412 while (!list_empty(&rcl->rcl_referring_calls)) { 1413 rc = list_first_entry(&rcl->rcl_referring_calls, 1414 struct nfsd4_referring_call, 1415 __list); 1416 list_del(&rc->__list); 1417 kfree(rc); 1418 } 1419 list_del(&rcl->__list); 1420 kfree(rcl); 1421 } 1422 } 1423 1424 static void nfsd4_cb_prepare(struct rpc_task *task, void *calldata) 1425 { 1426 struct nfsd4_callback *cb = calldata; 1427 struct nfs4_client *clp = cb->cb_clp; 1428 u32 minorversion = clp->cl_minorversion; 1429 1430 /* 1431 * cb_seq_status is only set in decode_cb_sequence4res, 1432 * and so will remain 1 if an rpc level failure occurs. 1433 */ 1434 trace_nfsd_cb_rpc_prepare(clp); 1435 cb->cb_seq_status = 1; 1436 cb->cb_status = 0; 1437 if (minorversion && !nfsd41_cb_get_slot(cb, task)) 1438 return; 1439 rpc_call_start(task); 1440 } 1441 1442 /* Returns true if CB_COMPOUND processing should continue */ 1443 static bool nfsd4_cb_sequence_done(struct rpc_task *task, struct nfsd4_callback *cb) 1444 { 1445 struct nfsd4_session *session = cb->cb_clp->cl_cb_session; 1446 bool ret = false; 1447 1448 if (cb->cb_held_slot < 0) 1449 goto requeue; 1450 1451 /* This is the operation status code for CB_SEQUENCE */ 1452 trace_nfsd_cb_seq_status(task, cb); 1453 switch (cb->cb_seq_status) { 1454 case 0: 1455 /* 1456 * No need for lock, access serialized in nfsd4_cb_prepare 1457 * 1458 * RFC5661 20.9.3 1459 * If CB_SEQUENCE returns an error, then the state of the slot 1460 * (sequence ID, cached reply) MUST NOT change. 1461 */ 1462 ++session->se_cb_seq_nr[cb->cb_held_slot]; 1463 ret = true; 1464 break; 1465 case -ESERVERFAULT: 1466 /* 1467 * Call succeeded, but the session, slot index, or slot 1468 * sequence number in the response do not match the same 1469 * in the server's call. The sequence information is thus 1470 * untrustworthy. 1471 */ 1472 nfsd4_mark_cb_fault(cb->cb_clp); 1473 break; 1474 case 1: 1475 /* 1476 * cb_seq_status remains 1 if an RPC Reply was never 1477 * received. NFSD can't know if the client processed 1478 * the CB_SEQUENCE operation. Ask the client to send a 1479 * DESTROY_SESSION to recover. 1480 */ 1481 fallthrough; 1482 case -NFS4ERR_BADSESSION: 1483 nfsd4_mark_cb_fault(cb->cb_clp); 1484 goto requeue; 1485 case -NFS4ERR_DELAY: 1486 cb->cb_seq_status = 1; 1487 if (RPC_SIGNALLED(task) || !rpc_restart_call(task)) 1488 goto requeue; 1489 rpc_delay(task, 2 * HZ); 1490 return false; 1491 case -NFS4ERR_SEQ_MISORDERED: 1492 case -NFS4ERR_BADSLOT: 1493 /* 1494 * A SEQ_MISORDERED or BADSLOT error means that the client and 1495 * server are out of sync as to the backchannel parameters. Mark 1496 * the backchannel faulty and restart the RPC, but leak the slot 1497 * so that it's no longer used. 1498 */ 1499 nfsd4_mark_cb_fault(cb->cb_clp); 1500 cb->cb_held_slot = -1; 1501 goto retry_nowait; 1502 default: 1503 nfsd4_mark_cb_fault(cb->cb_clp); 1504 } 1505 trace_nfsd_cb_free_slot(task, cb); 1506 nfsd41_cb_release_slot(cb); 1507 return ret; 1508 retry_nowait: 1509 /* 1510 * RPC_SIGNALLED() means that the rpc_client is being torn down and 1511 * (possibly) recreated. Requeue the call in that case. 1512 */ 1513 if (!RPC_SIGNALLED(task)) { 1514 if (rpc_restart_call_prepare(task)) 1515 return false; 1516 } 1517 requeue: 1518 nfsd41_cb_release_slot(cb); 1519 nfsd4_requeue_cb(task, cb); 1520 return false; 1521 } 1522 1523 static void nfsd4_cb_done(struct rpc_task *task, void *calldata) 1524 { 1525 struct nfsd4_callback *cb = calldata; 1526 struct nfs4_client *clp = cb->cb_clp; 1527 1528 trace_nfsd_cb_rpc_done(clp); 1529 1530 if (!clp->cl_minorversion) { 1531 /* 1532 * If the backchannel connection was shut down while this 1533 * task was queued, we need to resubmit it after setting up 1534 * a new backchannel connection. 1535 * 1536 * Note that if we lost our callback connection permanently 1537 * the submission code will error out, so we don't need to 1538 * handle that case here. 1539 */ 1540 if (RPC_SIGNALLED(task)) 1541 nfsd4_requeue_cb(task, cb); 1542 } else if (!nfsd4_cb_sequence_done(task, cb)) { 1543 return; 1544 } 1545 1546 if (cb->cb_status) { 1547 WARN_ONCE(task->tk_status, 1548 "cb_status=%d tk_status=%d cb_opcode=%d", 1549 cb->cb_status, task->tk_status, cb->cb_ops->opcode); 1550 task->tk_status = cb->cb_status; 1551 } 1552 1553 switch (cb->cb_ops->done(cb, task)) { 1554 case 0: 1555 task->tk_status = 0; 1556 rpc_restart_call_prepare(task); 1557 return; 1558 case 1: 1559 switch (task->tk_status) { 1560 case -EIO: 1561 case -ETIMEDOUT: 1562 case -EACCES: 1563 nfsd4_mark_cb_down(clp); 1564 } 1565 break; 1566 default: 1567 BUG(); 1568 } 1569 } 1570 1571 static void nfsd4_cb_release(void *calldata) 1572 { 1573 struct nfsd4_callback *cb = calldata; 1574 1575 trace_nfsd_cb_rpc_release(cb->cb_clp); 1576 1577 if (test_bit(NFSD4_CALLBACK_REQUEUE, &cb->cb_flags)) 1578 nfsd4_queue_cb(cb); 1579 else 1580 nfsd41_destroy_cb(cb); 1581 1582 } 1583 1584 static const struct rpc_call_ops nfsd4_cb_ops = { 1585 .rpc_call_prepare = nfsd4_cb_prepare, 1586 .rpc_call_done = nfsd4_cb_done, 1587 .rpc_release = nfsd4_cb_release, 1588 }; 1589 1590 /* must be called under the state lock */ 1591 void nfsd4_shutdown_callback(struct nfs4_client *clp) 1592 { 1593 if (clp->cl_cb_state != NFSD4_CB_UNKNOWN) 1594 trace_nfsd_cb_shutdown(clp); 1595 1596 set_bit(NFSD4_CLIENT_CB_KILL, &clp->cl_flags); 1597 /* 1598 * Note this won't actually result in a null callback; 1599 * instead, nfsd4_run_cb_null() will detect the killed 1600 * client, destroy the rpc client, and stop: 1601 */ 1602 nfsd4_run_cb(&clp->cl_cb_null); 1603 flush_workqueue(clp->cl_callback_wq); 1604 nfsd41_cb_inflight_wait_complete(clp); 1605 } 1606 1607 static struct nfsd4_conn * __nfsd4_find_backchannel(struct nfs4_client *clp) 1608 { 1609 struct nfsd4_session *s; 1610 struct nfsd4_conn *c; 1611 1612 lockdep_assert_held(&clp->cl_lock); 1613 1614 list_for_each_entry(s, &clp->cl_sessions, se_perclnt) { 1615 list_for_each_entry(c, &s->se_conns, cn_persession) { 1616 if (c->cn_flags & NFS4_CDFC4_BACK) 1617 return c; 1618 } 1619 } 1620 return NULL; 1621 } 1622 1623 /* 1624 * Note there isn't a lot of locking in this code; instead we depend on 1625 * the fact that it is run from clp->cl_callback_wq, which won't run two 1626 * work items at once. So, for example, clp->cl_callback_wq handles all 1627 * access of cl_cb_client and all calls to rpc_create or rpc_shutdown_client. 1628 */ 1629 static void nfsd4_process_cb_update(struct nfsd4_callback *cb) 1630 { 1631 struct nfs4_cb_conn conn; 1632 struct nfs4_client *clp = cb->cb_clp; 1633 struct nfsd4_session *ses = NULL; 1634 struct nfsd4_conn *c; 1635 int err; 1636 1637 trace_nfsd_cb_bc_update(clp, cb); 1638 1639 /* 1640 * This is either an update, or the client dying; in either case, 1641 * kill the old client: 1642 */ 1643 if (clp->cl_cb_client) { 1644 trace_nfsd_cb_bc_shutdown(clp, cb); 1645 rpc_shutdown_client(clp->cl_cb_client); 1646 clp->cl_cb_client = NULL; 1647 put_cred(clp->cl_cb_cred); 1648 clp->cl_cb_cred = NULL; 1649 } 1650 if (clp->cl_cb_conn.cb_xprt) { 1651 svc_xprt_put(clp->cl_cb_conn.cb_xprt); 1652 clp->cl_cb_conn.cb_xprt = NULL; 1653 } 1654 if (test_bit(NFSD4_CLIENT_CB_KILL, &clp->cl_flags)) 1655 return; 1656 1657 spin_lock(&clp->cl_lock); 1658 /* 1659 * Only serialized callback code is allowed to clear these 1660 * flags; main nfsd code can only set them: 1661 */ 1662 WARN_ON(!(clp->cl_flags & NFSD4_CLIENT_CB_FLAG_MASK)); 1663 clear_bit(NFSD4_CLIENT_CB_UPDATE, &clp->cl_flags); 1664 1665 memcpy(&conn, &cb->cb_clp->cl_cb_conn, sizeof(struct nfs4_cb_conn)); 1666 c = __nfsd4_find_backchannel(clp); 1667 if (c) { 1668 svc_xprt_get(c->cn_xprt); 1669 conn.cb_xprt = c->cn_xprt; 1670 ses = c->cn_session; 1671 } 1672 spin_unlock(&clp->cl_lock); 1673 1674 err = setup_callback_client(clp, &conn, ses); 1675 if (err) { 1676 nfsd4_mark_cb_down(clp); 1677 if (c) 1678 svc_xprt_put(c->cn_xprt); 1679 return; 1680 } 1681 } 1682 1683 static void 1684 nfsd4_run_cb_work(struct work_struct *work) 1685 { 1686 struct nfsd4_callback *cb = 1687 container_of(work, struct nfsd4_callback, cb_work); 1688 struct nfs4_client *clp = cb->cb_clp; 1689 struct rpc_clnt *clnt; 1690 int flags, ret; 1691 1692 trace_nfsd_cb_start(clp); 1693 1694 if (clp->cl_flags & NFSD4_CLIENT_CB_FLAG_MASK) 1695 nfsd4_process_cb_update(cb); 1696 1697 clnt = clp->cl_cb_client; 1698 if (!clnt || clp->cl_state == NFSD4_COURTESY) { 1699 /* 1700 * Callback channel broken, client killed or 1701 * nfs4_client in courtesy state; give up. 1702 */ 1703 nfsd41_destroy_cb(cb); 1704 return; 1705 } 1706 1707 /* 1708 * Don't send probe messages for 4.1 or later. 1709 */ 1710 if (!cb->cb_ops && clp->cl_minorversion) { 1711 nfsd4_mark_cb_state(clp, NFSD4_CB_UP); 1712 nfsd41_destroy_cb(cb); 1713 return; 1714 } 1715 1716 if (!test_and_clear_bit(NFSD4_CALLBACK_REQUEUE, &cb->cb_flags)) { 1717 if (cb->cb_ops && cb->cb_ops->prepare) 1718 cb->cb_ops->prepare(cb); 1719 } 1720 1721 cb->cb_msg.rpc_cred = clp->cl_cb_cred; 1722 flags = clp->cl_minorversion ? RPC_TASK_NOCONNECT : RPC_TASK_SOFTCONN; 1723 ret = rpc_call_async(clnt, &cb->cb_msg, RPC_TASK_SOFT | flags, 1724 cb->cb_ops ? &nfsd4_cb_ops : &nfsd4_cb_probe_ops, cb); 1725 if (ret != 0) { 1726 set_bit(NFSD4_CALLBACK_REQUEUE, &cb->cb_flags); 1727 nfsd4_queue_cb(cb); 1728 } 1729 } 1730 1731 void nfsd4_init_cb(struct nfsd4_callback *cb, struct nfs4_client *clp, 1732 const struct nfsd4_callback_ops *ops, enum nfsd4_cb_op op) 1733 { 1734 cb->cb_clp = clp; 1735 cb->cb_msg.rpc_proc = &nfs4_cb_procedures[op]; 1736 cb->cb_msg.rpc_argp = cb; 1737 cb->cb_msg.rpc_resp = cb; 1738 cb->cb_flags = 0; 1739 cb->cb_ops = ops; 1740 INIT_WORK(&cb->cb_work, nfsd4_run_cb_work); 1741 cb->cb_status = 0; 1742 cb->cb_held_slot = -1; 1743 cb->cb_nr_referring_call_list = 0; 1744 INIT_LIST_HEAD(&cb->cb_referring_call_list); 1745 } 1746 1747 /** 1748 * nfsd4_run_cb - queue up a callback job to run 1749 * @cb: callback to queue 1750 * 1751 * Kick off a callback to do its thing. Returns false if it was already 1752 * on a queue, true otherwise. 1753 */ 1754 bool nfsd4_run_cb(struct nfsd4_callback *cb) 1755 { 1756 struct nfs4_client *clp = cb->cb_clp; 1757 bool queued; 1758 1759 nfsd41_cb_inflight_begin(clp); 1760 queued = nfsd4_queue_cb(cb); 1761 if (!queued) 1762 nfsd41_cb_inflight_end(clp); 1763 return queued; 1764 } 1765 1766 /** 1767 * nfsd_net_cb_shutdown - release per-netns callback RPC program resources 1768 * @nn: NFS server network namespace 1769 * 1770 * Frees resources allocated by nfsd_net_cb_init(). 1771 */ 1772 void nfsd_net_cb_shutdown(struct nfsd_net *nn) 1773 { 1774 struct nfsd_net_cb *cb = nn->nfsd_cb; 1775 1776 if (cb) { 1777 kfree(cb->version4.counts); 1778 kfree(cb); 1779 nn->nfsd_cb = NULL; 1780 } 1781 } 1782 1783 /** 1784 * nfsd_net_cb_init - initialize per-netns callback RPC program 1785 * @nn: NFS server network namespace 1786 * 1787 * Sets up the callback RPC program, version table, procedure 1788 * counts, and statistics structure for @nn. Caller must release 1789 * these resources using nfsd_net_cb_shutdown(). 1790 * 1791 * Return: 0 on success, or -ENOMEM if allocation fails. 1792 */ 1793 int nfsd_net_cb_init(struct nfsd_net *nn) 1794 { 1795 struct nfsd_net_cb *cb; 1796 1797 cb = kzalloc(sizeof(*cb), GFP_KERNEL); 1798 if (!cb) 1799 return -ENOMEM; 1800 1801 cb->version4.counts = kzalloc_objs(unsigned int, 1802 ARRAY_SIZE(nfs4_cb_procedures), GFP_KERNEL); 1803 if (!cb->version4.counts) { 1804 kfree(cb); 1805 return -ENOMEM; 1806 } 1807 /* 1808 * Note on the callback rpc program version number: despite language 1809 * in rfc 5661 section 18.36.3 requiring servers to use 4 in this 1810 * field, the official xdr descriptions for both 4.0 and 4.1 specify 1811 * version 1, and in practice that appears to be what implementations 1812 * use. The section 18.36.3 language is expected to be fixed in an 1813 * erratum. 1814 */ 1815 cb->version4.number = NFS4_CB_VERSION; 1816 cb->version4.nrprocs = ARRAY_SIZE(nfs4_cb_procedures); 1817 cb->version4.procs = nfs4_cb_procedures; 1818 cb->versions[NFS4_CB_VERSION] = &cb->version4; 1819 1820 cb->program.name = "nfs4_cb"; 1821 cb->program.number = NFS4_CB_PROGRAM; 1822 cb->program.nrvers = ARRAY_SIZE(cb->versions); 1823 cb->program.version = &cb->versions[0]; 1824 cb->program.pipe_dir_name = "nfsd4_cb"; 1825 cb->program.stats = &cb->stat; 1826 cb->stat.program = &cb->program; 1827 1828 nn->nfsd_cb = cb; 1829 1830 return 0; 1831 } 1832