1 /* 2 * CDDL HEADER START 3 * 4 * The contents of this file are subject to the terms of the 5 * Common Development and Distribution License, Version 1.0 only 6 * (the "License"). You may not use this file except in compliance 7 * with the License. 8 * 9 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE 10 * or http://www.opensolaris.org/os/licensing. 11 * See the License for the specific language governing permissions 12 * and limitations under the License. 13 * 14 * When distributing Covered Code, include this CDDL HEADER in each 15 * file and include the License file at usr/src/OPENSOLARIS.LICENSE. 16 * If applicable, add the following below this CDDL HEADER, with the 17 * fields enclosed by brackets "[]" replaced with your own identifying 18 * information: Portions Copyright [yyyy] [name of copyright owner] 19 * 20 * CDDL HEADER END 21 */ 22 /* 23 * Copyright 2005 Sun Microsystems, Inc. All rights reserved. 24 * Use is subject to license terms. 25 */ 26 27 /* Copyright (c) 1984, 1986, 1987, 1988, 1989 AT&T */ 28 /* All Rights Reserved */ 29 30 /* 31 * Portions of this source code were derived from Berkeley 4.3 BSD 32 * under license from the Regents of the University of California. 33 */ 34 35 #ifndef _NFS4_CLNT_H 36 #define _NFS4_CLNT_H 37 38 #pragma ident "%Z%%M% %I% %E% SMI" 39 40 #include <sys/errno.h> 41 #include <sys/types.h> 42 #include <sys/kstat.h> 43 #include <sys/time.h> 44 #include <sys/flock.h> 45 #include <vm/page.h> 46 #include <nfs/nfs4_kprot.h> 47 #include <nfs/nfs4.h> 48 #include <nfs/rnode.h> 49 #include <sys/avl.h> 50 #include <sys/list.h> 51 52 #ifdef __cplusplus 53 extern "C" { 54 #endif 55 56 #define NFS4_SIZE_OK(size) ((size) <= MAXOFFSET_T) 57 58 /* Four states of nfs4_server's lease_valid */ 59 #define NFS4_LEASE_INVALID 0 60 #define NFS4_LEASE_VALID 1 61 #define NFS4_LEASE_UNINITIALIZED 2 62 #define NFS4_LEASE_NOT_STARTED 3 63 64 /* flag to tell the renew thread it should exit */ 65 #define NFS4_THREAD_EXIT 1 66 67 /* Default number of seconds to wait on GRACE and DELAY errors */ 68 #define NFS4ERR_DELAY_TIME 10 69 70 /* Number of hash buckets for open owners for each nfs4_server */ 71 #define NFS4_NUM_OO_BUCKETS 53 72 73 /* Number of freed open owners (per mntinfo4_t) to keep around */ 74 #define NFS4_NUM_FREED_OPEN_OWNERS 8 75 76 /* Number of seconds to wait before retrying a SETCLIENTID(_CONFIRM) op */ 77 #define NFS4_RETRY_SCLID_DELAY 10 78 79 /* Number of times we should retry a SETCLIENTID(_CONFIRM) op */ 80 #define NFS4_NUM_SCLID_RETRIES 3 81 82 /* Number of times we should retry on open after getting NFS4ERR_BAD_SEQID */ 83 #define NFS4_NUM_RETRY_BAD_SEQID 3 84 85 /* 86 * Is the attribute cache valid? If client holds a delegation, then attrs 87 * are by definition valid. If not, then check to see if attrs have timed out. 88 */ 89 #define ATTRCACHE4_VALID(vp) (VTOR4(vp)->r_deleg_type != OPEN_DELEGATE_NONE || \ 90 gethrtime() < VTOR4(vp)->r_time_attr_inval) 91 92 /* 93 * Flags to indicate whether to purge the DNLC for non-directory vnodes 94 * in a call to nfs_purge_caches. 95 */ 96 #define NFS4_NOPURGE_DNLC 0 97 #define NFS4_PURGE_DNLC 1 98 99 /* 100 * Is cache valid? 101 * Swap is always valid, if no attributes (attrtime == 0) or 102 * if mtime matches cached mtime it is valid 103 * NOTE: mtime is now a timestruc_t. 104 * Caller should be holding the rnode r_statelock mutex. 105 */ 106 #define CACHE4_VALID(rp, mtime, fsize) \ 107 ((RTOV4(rp)->v_flag & VISSWAP) == VISSWAP || \ 108 (((mtime).tv_sec == (rp)->r_attr.va_mtime.tv_sec && \ 109 (mtime).tv_nsec == (rp)->r_attr.va_mtime.tv_nsec) && \ 110 ((fsize) == (rp)->r_attr.va_size))) 111 112 /* 113 * Macro to detect forced unmount or a zone shutdown. 114 */ 115 #define FS_OR_ZONE_GONE4(vfsp) \ 116 (((vfsp)->vfs_flag & VFS_UNMOUNTED) || \ 117 zone_status_get(curproc->p_zone) >= ZONE_IS_SHUTTING_DOWN) 118 119 /* 120 * Macro to help determine whether a request failed because the underlying 121 * filesystem has been forcibly unmounted or because of zone shutdown. 122 */ 123 #define NFS4_FRC_UNMT_ERR(err, vfsp) \ 124 ((err) == EIO && FS_OR_ZONE_GONE4((vfsp))) 125 126 /* 127 * Due to the way the address space callbacks are used to execute a delmap, 128 * we must keep track of how many times the same thread has called 129 * VOP_DELMAP()->nfs4_delmap(). This is done by having a list of 130 * nfs4_delmapcall_t's associated with each rnode4_t. This list is protected 131 * by the rnode4_t's r_statelock. The individual elements do not need to be 132 * protected as they will only ever be created, modified and destroyed by 133 * one thread (the call_id). 134 * See nfs4_delmap() for further explanation. 135 */ 136 typedef struct nfs4_delmapcall { 137 kthread_t *call_id; 138 int error; /* error from delmap */ 139 list_node_t call_node; 140 } nfs4_delmapcall_t; 141 142 /* 143 * delmap address space callback args 144 */ 145 typedef struct nfs4_delmap_args { 146 vnode_t *vp; 147 offset_t off; 148 caddr_t addr; 149 size_t len; 150 uint_t prot; 151 uint_t maxprot; 152 uint_t flags; 153 cred_t *cr; 154 nfs4_delmapcall_t *caller; /* to retrieve errors from the cb */ 155 } nfs4_delmap_args_t; 156 157 /* 158 * client side statistics 159 */ 160 /* 161 * Per-zone counters 162 */ 163 struct clstat4 { 164 kstat_named_t calls; /* client requests */ 165 kstat_named_t badcalls; /* rpc failures */ 166 kstat_named_t clgets; /* client handle gets */ 167 kstat_named_t cltoomany; /* client handle cache misses */ 168 #ifdef DEBUG 169 kstat_named_t clalloc; /* number of client handles */ 170 kstat_named_t noresponse; /* server not responding cnt */ 171 kstat_named_t failover; /* server failover count */ 172 kstat_named_t remap; /* server remap count */ 173 #endif 174 }; 175 176 #ifdef DEBUG 177 /* 178 * The following are statistics that describe the behavior of the system as a 179 * whole and don't correspond to any particular zone. 180 */ 181 struct clstat4_debug { 182 kstat_named_t nrnode; /* number of allocated rnodes */ 183 kstat_named_t access; /* size of access cache */ 184 kstat_named_t dirent; /* size of readdir cache */ 185 kstat_named_t dirents; /* size of readdir buf cache */ 186 kstat_named_t reclaim; /* number of reclaims */ 187 kstat_named_t clreclaim; /* number of cl reclaims */ 188 kstat_named_t f_reclaim; /* number of free reclaims */ 189 kstat_named_t a_reclaim; /* number of active reclaims */ 190 kstat_named_t r_reclaim; /* number of rnode reclaims */ 191 kstat_named_t rpath; /* bytes used to store rpaths */ 192 }; 193 extern struct clstat4_debug clstat4_debug; 194 195 #endif 196 197 /* 198 * The NFS specific async_reqs structure. 199 */ 200 201 enum iotype4 { 202 NFS4_READ_AHEAD, 203 NFS4_PUTAPAGE, 204 NFS4_PAGEIO, 205 NFS4_READDIR, 206 NFS4_INACTIVE, 207 NFS4_COMMIT 208 }; 209 #define NFS4_ASYNC_TYPES (NFS4_COMMIT + 1) 210 211 struct nfs4_async_read_req { 212 void (*readahead)(); /* pointer to readahead function */ 213 u_offset_t blkoff; /* offset in file */ 214 struct seg *seg; /* segment to do i/o to */ 215 caddr_t addr; /* address to do i/o to */ 216 }; 217 218 struct nfs4_pageio_req { 219 int (*pageio)(); /* pointer to pageio function */ 220 page_t *pp; /* page list */ 221 u_offset_t io_off; /* offset in file */ 222 uint_t io_len; /* size of request */ 223 int flags; 224 }; 225 226 struct nfs4_readdir_req { 227 int (*readdir)(); /* pointer to readdir function */ 228 struct rddir4_cache *rdc; /* pointer to cache entry to fill */ 229 }; 230 231 struct nfs4_commit_req { 232 void (*commit)(); /* pointer to commit function */ 233 page_t *plist; /* page list */ 234 offset4 offset; /* starting offset */ 235 count4 count; /* size of range to be commited */ 236 }; 237 238 struct nfs4_async_reqs { 239 struct nfs4_async_reqs *a_next; /* pointer to next arg struct */ 240 #ifdef DEBUG 241 kthread_t *a_queuer; /* thread id of queueing thread */ 242 #endif 243 struct vnode *a_vp; /* vnode pointer */ 244 struct cred *a_cred; /* cred pointer */ 245 enum iotype4 a_io; /* i/o type */ 246 union { 247 struct nfs4_async_read_req a_read_args; 248 struct nfs4_pageio_req a_pageio_args; 249 struct nfs4_readdir_req a_readdir_args; 250 struct nfs4_commit_req a_commit_args; 251 } a_args; 252 }; 253 254 #define a_nfs4_readahead a_args.a_read_args.readahead 255 #define a_nfs4_blkoff a_args.a_read_args.blkoff 256 #define a_nfs4_seg a_args.a_read_args.seg 257 #define a_nfs4_addr a_args.a_read_args.addr 258 259 #define a_nfs4_putapage a_args.a_pageio_args.pageio 260 #define a_nfs4_pageio a_args.a_pageio_args.pageio 261 #define a_nfs4_pp a_args.a_pageio_args.pp 262 #define a_nfs4_off a_args.a_pageio_args.io_off 263 #define a_nfs4_len a_args.a_pageio_args.io_len 264 #define a_nfs4_flags a_args.a_pageio_args.flags 265 266 #define a_nfs4_readdir a_args.a_readdir_args.readdir 267 #define a_nfs4_rdc a_args.a_readdir_args.rdc 268 269 #define a_nfs4_commit a_args.a_commit_args.commit 270 #define a_nfs4_plist a_args.a_commit_args.plist 271 #define a_nfs4_offset a_args.a_commit_args.offset 272 #define a_nfs4_count a_args.a_commit_args.count 273 274 /* 275 * Security information 276 */ 277 typedef struct sv_secinfo { 278 uint_t count; /* how many sdata there are */ 279 uint_t index; /* which sdata[index] */ 280 struct sec_data *sdata; 281 } sv_secinfo_t; 282 283 /* 284 * Hash bucket for the mi's open owner list (mi_oo_list). 285 */ 286 typedef struct nfs4_oo_hash_bucket { 287 list_t b_oo_hash_list; 288 kmutex_t b_lock; 289 } nfs4_oo_hash_bucket_t; 290 291 /* 292 * Global array of ctags. 293 */ 294 extern ctag_t nfs4_ctags[]; 295 296 typedef enum nfs4_tag_type { 297 TAG_NONE, 298 TAG_ACCESS, 299 TAG_CLOSE, 300 TAG_CLOSE_LOST, 301 TAG_CLOSE_UNDO, 302 TAG_COMMIT, 303 TAG_DELEGRETURN, 304 TAG_FSINFO, 305 TAG_GET_SYMLINK, 306 TAG_GETATTR, 307 TAG_INACTIVE, 308 TAG_LINK, 309 TAG_LOCK, 310 TAG_LOCK_RECLAIM, 311 TAG_LOCK_RESEND, 312 TAG_LOCK_REINSTATE, 313 TAG_LOCK_UNKNOWN, 314 TAG_LOCKT, 315 TAG_LOCKU, 316 TAG_LOCKU_RESEND, 317 TAG_LOCKU_REINSTATE, 318 TAG_LOOKUP, 319 TAG_LOOKUP_PARENT, 320 TAG_LOOKUP_VALID, 321 TAG_LOOKUP_VPARENT, 322 TAG_MKDIR, 323 TAG_MKNOD, 324 TAG_MOUNT, 325 TAG_OPEN, 326 TAG_OPEN_CONFIRM, 327 TAG_OPEN_CONFIRM_LOST, 328 TAG_OPEN_DG, 329 TAG_OPEN_DG_LOST, 330 TAG_OPEN_LOST, 331 TAG_OPENATTR, 332 TAG_PATHCONF, 333 TAG_PUTROOTFH, 334 TAG_READ, 335 TAG_READAHEAD, 336 TAG_READDIR, 337 TAG_READLINK, 338 TAG_RELOCK, 339 TAG_REMAP_LOOKUP, 340 TAG_REMAP_LOOKUP_AD, 341 TAG_REMAP_LOOKUP_NA, 342 TAG_REMAP_MOUNT, 343 TAG_RMDIR, 344 TAG_REMOVE, 345 TAG_RENAME, 346 TAG_RENAME_VFH, 347 TAG_RENEW, 348 TAG_REOPEN, 349 TAG_REOPEN_LOST, 350 TAG_SECINFO, 351 TAG_SETATTR, 352 TAG_SETCLIENTID, 353 TAG_SETCLIENTID_CF, 354 TAG_SYMLINK, 355 TAG_WRITE 356 } nfs4_tag_type_t; 357 358 #define NFS4_TAG_INITIALIZER { \ 359 {TAG_NONE, "", \ 360 {0x20202020, 0x20202020, 0x20202020}}, \ 361 {TAG_ACCESS, "access", \ 362 {0x61636365, 0x73732020, 0x20202020}}, \ 363 {TAG_CLOSE, "close", \ 364 {0x636c6f73, 0x65202020, 0x20202020}}, \ 365 {TAG_CLOSE_LOST, "lost close", \ 366 {0x6c6f7374, 0x20636c6f, 0x73652020}}, \ 367 {TAG_CLOSE_UNDO, "undo close", \ 368 {0x756e646f, 0x20636c6f, 0x73652020}}, \ 369 {TAG_COMMIT, "commit", \ 370 {0x636f6d6d, 0x69742020, 0x20202020}}, \ 371 {TAG_DELEGRETURN, "delegreturn", \ 372 {0x64656c65, 0x67726574, 0x75726e20}}, \ 373 {TAG_FSINFO, "fsinfo", \ 374 {0x6673696e, 0x666f2020, 0x20202020}}, \ 375 {TAG_GET_SYMLINK, "get symlink text", \ 376 {0x67657420, 0x736c6e6b, 0x20747874}}, \ 377 {TAG_GETATTR, "getattr", \ 378 {0x67657461, 0x74747220, 0x20202020}}, \ 379 {TAG_INACTIVE, "inactive", \ 380 {0x696e6163, 0x74697665, 0x20202020}}, \ 381 {TAG_LINK, "link", \ 382 {0x6c696e6b, 0x20202020, 0x20202020}}, \ 383 {TAG_LOCK, "lock", \ 384 {0x6c6f636b, 0x20202020, 0x20202020}}, \ 385 {TAG_LOCK_RECLAIM, "reclaim lock", \ 386 {0x7265636c, 0x61696d20, 0x6c6f636b}}, \ 387 {TAG_LOCK_RESEND, "resend lock", \ 388 {0x72657365, 0x6e64206c, 0x6f636b20}}, \ 389 {TAG_LOCK_REINSTATE, "reinstate lock", \ 390 {0x7265696e, 0x7374206c, 0x6f636b20}}, \ 391 {TAG_LOCK_UNKNOWN, "unknown lock", \ 392 {0x756e6b6e, 0x6f776e20, 0x6c6f636b}}, \ 393 {TAG_LOCKT, "lock test", \ 394 {0x6c6f636b, 0x5f746573, 0x74202020}}, \ 395 {TAG_LOCKU, "unlock", \ 396 {0x756e6c6f, 0x636b2020, 0x20202020}}, \ 397 {TAG_LOCKU_RESEND, "resend locku", \ 398 {0x72657365, 0x6e64206c, 0x6f636b75}}, \ 399 {TAG_LOCKU_REINSTATE, "reinstate unlock", \ 400 {0x7265696e, 0x73742075, 0x6e6c636b}}, \ 401 {TAG_LOOKUP, "lookup", \ 402 {0x6c6f6f6b, 0x75702020, 0x20202020}}, \ 403 {TAG_LOOKUP_PARENT, "lookup parent", \ 404 {0x6c6f6f6b, 0x75702070, 0x6172656e}}, \ 405 {TAG_LOOKUP_VALID, "lookup valid", \ 406 {0x6c6f6f6b, 0x75702076, 0x616c6964}}, \ 407 {TAG_LOOKUP_VPARENT, "lookup valid parent", \ 408 {0x6c6f6f6b, 0x766c6420, 0x7061726e}}, \ 409 {TAG_MKDIR, "mkdir", \ 410 {0x6d6b6469, 0x72202020, 0x20202020}}, \ 411 {TAG_MKNOD, "mknod", \ 412 {0x6d6b6e6f, 0x64202020, 0x20202020}}, \ 413 {TAG_MOUNT, "mount", \ 414 {0x6d6f756e, 0x74202020, 0x20202020}}, \ 415 {TAG_OPEN, "open", \ 416 {0x6f70656e, 0x20202020, 0x20202020}}, \ 417 {TAG_OPEN_CONFIRM, "open confirm", \ 418 {0x6f70656e, 0x5f636f6e, 0x6669726d}}, \ 419 {TAG_OPEN_CONFIRM_LOST, "lost open confirm", \ 420 {0x6c6f7374, 0x206f7065, 0x6e5f636f}}, \ 421 {TAG_OPEN_DG, "open downgrade", \ 422 {0x6f70656e, 0x20646772, 0x61646520}}, \ 423 {TAG_OPEN_DG_LOST, "lost open downgrade", \ 424 {0x6c737420, 0x6f70656e, 0x20646772}}, \ 425 {TAG_OPEN_LOST, "lost open", \ 426 {0x6c6f7374, 0x206f7065, 0x6e202020}}, \ 427 {TAG_OPENATTR, "openattr", \ 428 {0x6f70656e, 0x61747472, 0x20202020}}, \ 429 {TAG_PATHCONF, "pathhconf", \ 430 {0x70617468, 0x636f6e66, 0x20202020}}, \ 431 {TAG_PUTROOTFH, "putrootfh", \ 432 {0x70757472, 0x6f6f7466, 0x68202020}}, \ 433 {TAG_READ, "read", \ 434 {0x72656164, 0x20202020, 0x20202020}}, \ 435 {TAG_READAHEAD, "readahead", \ 436 {0x72656164, 0x61686561, 0x64202020}}, \ 437 {TAG_READDIR, "readdir", \ 438 {0x72656164, 0x64697220, 0x20202020}}, \ 439 {TAG_READLINK, "readlink", \ 440 {0x72656164, 0x6c696e6b, 0x20202020}}, \ 441 {TAG_RELOCK, "relock", \ 442 {0x72656c6f, 0x636b2020, 0x20202020}}, \ 443 {TAG_REMAP_LOOKUP, "remap lookup", \ 444 {0x72656d61, 0x70206c6f, 0x6f6b7570}}, \ 445 {TAG_REMAP_LOOKUP_AD, "remap lookup attr dir", \ 446 {0x72656d70, 0x206c6b75, 0x70206164}}, \ 447 {TAG_REMAP_LOOKUP_NA, "remap lookup named attrs", \ 448 {0x72656d70, 0x206c6b75, 0x70206e61}}, \ 449 {TAG_REMAP_MOUNT, "remap mount", \ 450 {0x72656d61, 0x70206d6f, 0x756e7420}}, \ 451 {TAG_RMDIR, "rmdir", \ 452 {0x726d6469, 0x72202020, 0x20202020}}, \ 453 {TAG_REMOVE, "remove", \ 454 {0x72656d6f, 0x76652020, 0x20202020}}, \ 455 {TAG_RENAME, "rename", \ 456 {0x72656e61, 0x6d652020, 0x20202020}}, \ 457 {TAG_RENAME_VFH, "rename volatile fh", \ 458 {0x72656e61, 0x6d652028, 0x76666829}}, \ 459 {TAG_RENEW, "renew", \ 460 {0x72656e65, 0x77202020, 0x20202020}}, \ 461 {TAG_REOPEN, "reopen", \ 462 {0x72656f70, 0x656e2020, 0x20202020}}, \ 463 {TAG_REOPEN_LOST, "lost reopen", \ 464 {0x6c6f7374, 0x2072656f, 0x70656e20}}, \ 465 {TAG_SECINFO, "secinfo", \ 466 {0x73656369, 0x6e666f20, 0x20202020}}, \ 467 {TAG_SETATTR, "setattr", \ 468 {0x73657461, 0x74747220, 0x20202020}}, \ 469 {TAG_SETCLIENTID, "setclientid", \ 470 {0x73657463, 0x6c69656e, 0x74696420}}, \ 471 {TAG_SETCLIENTID_CF, "setclientid_confirm", \ 472 {0x73636c6e, 0x7469645f, 0x636f6e66}}, \ 473 {TAG_SYMLINK, "symlink", \ 474 {0x73796d6c, 0x696e6b20, 0x20202020}}, \ 475 {TAG_WRITE, "write", \ 476 {0x77726974, 0x65202020, 0x20202020}} \ 477 } 478 479 /* 480 * These flags are for differentiating the search criterian for 481 * find_open_owner(). The comparison is done with the open_owners's 482 * 'oo_just_created' flag. 483 */ 484 #define NFS4_PERM_CREATED 0x0 485 #define NFS4_JUST_CREATED 0x1 486 487 /* 488 * Hashed by the cr_uid and cr_ruid of credential 'oo_cred'. 'oo_cred_otw' 489 * is stored upon a successful OPEN. This is needed when the user's effective 490 * and real uid's don't match. The 'oo_cred_otw' overrides the credential 491 * passed down by VFS for async read/write, commit, lock, and close operations. 492 * 493 * The oo_ref_count keeps track the number of active references on this 494 * data structure + number of nfs4_open_streams point to this structure. 495 * 496 * 'oo_valid' tells whether this stuct is about to be freed or not. 497 * 498 * 'oo_just_created' tells us whether this struct has just been created but 499 * not been fully finalized (that is created upon an OPEN request and 500 * finalized upon the OPEN success). 501 * 502 * The 'oo_seqid_inuse' is for the open seqid synchronization. If a thread 503 * is currently using the open owner and it's open_seqid, then it sets the 504 * oo_seqid_inuse to true if it currently is not set. If it is set then it 505 * does a cv_wait on the oo_cv_seqid_sync condition variable. When the thread 506 * is done it unsets the oo_seqid_inuse and does a cv_signal to wake a process 507 * waiting on the condition variable. 508 * 509 * 'oo_last_good_seqid' is the last valid seqid this open owner sent OTW, 510 * and 'oo_last_good_op' is the operation that issued the last valid seqid. 511 * 512 * Lock ordering: 513 * mntinfo4_t::mi_lock > oo_lock (for searching mi_oo_list) 514 * 515 * oo_seqid_inuse > mntinfo4_t::mi_lock 516 * oo_seqid_inuse > rnode4_t::r_statelock 517 * oo_seqid_inuse > rnode4_t::r_statev4_lock 518 * oo_seqid_inuse > nfs4_open_stream_t::os_sync_lock 519 * 520 * The 'oo_seqid_inuse'/'oo_cv_seqid_sync' protects: 521 * oo_last_good_op 522 * oo_last_good_seqid 523 * oo_name 524 * oo_seqid 525 * 526 * The 'oo_lock' protects: 527 * oo_cred 528 * oo_cred_otw 529 * oo_foo_node 530 * oo_hash_node 531 * oo_just_created 532 * oo_ref_count 533 * oo_valid 534 */ 535 536 typedef struct nfs4_open_owner { 537 cred_t *oo_cred; 538 int oo_ref_count; 539 int oo_valid; 540 int oo_just_created; 541 seqid4 oo_seqid; 542 seqid4 oo_last_good_seqid; 543 nfs4_tag_type_t oo_last_good_op; 544 unsigned oo_seqid_inuse:1; 545 cred_t *oo_cred_otw; 546 kcondvar_t oo_cv_seqid_sync; 547 /* 548 * Fix this to always be 8 bytes 549 */ 550 uint64_t oo_name; 551 list_node_t oo_hash_node; 552 list_node_t oo_foo_node; 553 kmutex_t oo_lock; 554 } nfs4_open_owner_t; 555 556 /* 557 * Static server information. 558 * These fields are read-only once they are initialized: 559 * sv_addr 560 * sv_dhsec 561 * sv_hostname 562 * sv_hostnamelen 563 * sv_knconf 564 * sv_next 565 * sv_origknconf 566 * 567 * These fields are protected by sv_lock: 568 * sv_currsec 569 * sv_fhandle 570 * sv_flags 571 * sv_fsid 572 * sv_path 573 * sv_pathlen 574 * sv_pfhandle 575 * sv_save_secinfo 576 * sv_savesec 577 * sv_secdata 578 * sv_secinfo 579 * sv_supp_attrs 580 * 581 * Lock ordering: 582 * nfs_rtable4_lock > sv_lock 583 * rnode4_t::r_statelock > sv_lock 584 */ 585 typedef struct servinfo4 { 586 struct knetconfig *sv_knconf; /* bound TLI fd */ 587 struct knetconfig *sv_origknconf; /* For RDMA save orig knconf */ 588 struct netbuf sv_addr; /* server's address */ 589 nfs4_fhandle_t sv_fhandle; /* this server's filehandle */ 590 nfs4_fhandle_t sv_pfhandle; /* parent dir filehandle */ 591 int sv_pathlen; /* Length of server path */ 592 char *sv_path; /* Path name on server */ 593 uint32_t sv_flags; /* flags for this server */ 594 sec_data_t *sv_secdata; /* client initiated security data */ 595 sv_secinfo_t *sv_secinfo; /* server security information */ 596 sec_data_t *sv_currsec; /* security data currently used; */ 597 /* points to one of the sec_data */ 598 /* entries in sv_secinfo */ 599 sv_secinfo_t *sv_save_secinfo; /* saved secinfo */ 600 sec_data_t *sv_savesec; /* saved security data */ 601 sec_data_t *sv_dhsec; /* AUTH_DH data from the user land */ 602 char *sv_hostname; /* server's hostname */ 603 int sv_hostnamelen; /* server's hostname length */ 604 fattr4_fsid sv_fsid; /* fsid of shared obj */ 605 fattr4_supported_attrs sv_supp_attrs; 606 struct servinfo4 *sv_next; /* next in list */ 607 nfs_rwlock_t sv_lock; 608 } servinfo4_t; 609 610 /* sv_flags fields */ 611 #define SV4_TRYSECINFO 0x001 /* try secinfo data from the server */ 612 #define SV4_TRYSECDEFAULT 0x002 /* try a default flavor */ 613 #define SV4_NOTINUSE 0x004 /* servinfo4_t had fatal errors */ 614 #define SV4_ROOT_STALE 0x008 /* root vnode got ESTALE */ 615 616 /* 617 * Lock call types. See nfs4frlock(). 618 */ 619 typedef enum nfs4_lock_call_type { 620 NFS4_LCK_CTYPE_NORM, 621 NFS4_LCK_CTYPE_RECLAIM, 622 NFS4_LCK_CTYPE_RESEND, 623 NFS4_LCK_CTYPE_REINSTATE 624 } nfs4_lock_call_type_t; 625 626 /* 627 * This structure holds the information for a lost open/close/open downgrade/ 628 * lock/locku request. It is also used for requests that are queued up so 629 * that the recovery thread can release server state after a forced 630 * unmount. 631 * "lr_op" is 0 if the struct is uninitialized. Otherwise, it is set to 632 * the proper OP_* nfs_opnum4 number. The other fields contain information 633 * to reconstruct the call. 634 * 635 * lr_dvp is used for OPENs with CREATE, so that we can do a PUTFH of the 636 * parent directroy without relying on vtodv (since we may not have a vp 637 * for the file we wish to create). 638 * 639 * lr_putfirst means that the request should go to the front of the resend 640 * queue, rather than the end. 641 */ 642 typedef struct nfs4_lost_rqst { 643 list_node_t lr_node; 644 nfs_opnum4 lr_op; 645 vnode_t *lr_vp; 646 vnode_t *lr_dvp; 647 nfs4_open_owner_t *lr_oop; 648 struct nfs4_open_stream *lr_osp; 649 struct nfs4_lock_owner *lr_lop; 650 cred_t *lr_cr; 651 flock64_t *lr_flk; 652 bool_t lr_putfirst; 653 union { 654 struct { 655 nfs4_lock_call_type_t lru_ctype; 656 nfs_lock_type4 lru_locktype; 657 } lru_lockargs; /* LOCK, LOCKU */ 658 struct { 659 uint32_t lru_oaccess; 660 uint32_t lru_odeny; 661 enum open_claim_type4 lru_oclaim; 662 stateid4 lru_ostateid; /* reopen only */ 663 component4 lru_ofile; 664 } lru_open_args; 665 struct { 666 uint32_t lru_dg_access; 667 uint32_t lru_dg_deny; 668 } lru_open_dg_args; 669 } nfs4_lr_u; 670 } nfs4_lost_rqst_t; 671 672 #define lr_oacc nfs4_lr_u.lru_open_args.lru_oaccess 673 #define lr_odeny nfs4_lr_u.lru_open_args.lru_odeny 674 #define lr_oclaim nfs4_lr_u.lru_open_args.lru_oclaim 675 #define lr_ostateid nfs4_lr_u.lru_open_args.lru_ostateid 676 #define lr_ofile nfs4_lr_u.lru_open_args.lru_ofile 677 #define lr_dg_acc nfs4_lr_u.lru_open_dg_args.lru_dg_access 678 #define lr_dg_deny nfs4_lr_u.lru_open_dg_args.lru_dg_deny 679 #define lr_ctype nfs4_lr_u.lru_lockargs.lru_ctype 680 #define lr_locktype nfs4_lr_u.lru_lockargs.lru_locktype 681 682 /* 683 * Recovery actions. Some actions can imply further recovery using a 684 * different recovery action (e.g., recovering the clientid leads to 685 * recovering open files and locks). 686 */ 687 688 typedef enum { 689 NR_UNUSED, 690 NR_CLIENTID, 691 NR_OPENFILES, 692 NR_FHEXPIRED, 693 NR_FAILOVER, 694 NR_WRONGSEC, 695 NR_EXPIRED, 696 NR_BAD_STATEID, 697 NR_BADHANDLE, 698 NR_BAD_SEQID, 699 NR_OLDSTATEID, 700 NR_GRACE, 701 NR_DELAY, 702 NR_LOST_LOCK, 703 NR_LOST_STATE_RQST, 704 NR_STALE 705 } nfs4_recov_t; 706 707 /* 708 * Administrative and debug message framework. 709 */ 710 711 #define NFS4_MSG_MAX 100 712 extern int nfs4_msg_max; 713 714 typedef enum { 715 RE_BAD_SEQID, 716 RE_BADHANDLE, 717 RE_CLIENTID, 718 RE_DEAD_FILE, 719 RE_END, 720 RE_FAIL_RELOCK, 721 RE_FAIL_REMAP_LEN, 722 RE_FAIL_REMAP_OP, 723 RE_FAILOVER, 724 RE_FILE_DIFF, 725 RE_LOST_STATE, 726 RE_OPENS_CHANGED, 727 RE_SIGLOST, 728 RE_SIGLOST_NO_DUMP, 729 RE_START, 730 RE_UNEXPECTED_ACTION, 731 RE_UNEXPECTED_ERRNO, 732 RE_UNEXPECTED_STATUS, 733 RE_WRONGSEC, 734 RE_LOST_STATE_BAD_OP 735 } nfs4_event_type_t; 736 737 typedef enum { 738 RFS_NO_INSPECT, 739 RFS_INSPECT 740 } nfs4_fact_status_t; 741 742 typedef enum { 743 RF_BADOWNER, 744 RF_ERR, 745 RF_RENEW_EXPIRED, 746 RF_SRV_NOT_RESPOND, 747 RF_SRV_OK, 748 RF_SRVS_NOT_RESPOND, 749 RF_SRVS_OK, 750 RF_DELMAP_CB_ERR 751 } nfs4_fact_type_t; 752 753 typedef enum { 754 NFS4_MS_DUMP, 755 NFS4_MS_NO_DUMP 756 } nfs4_msg_status_t; 757 758 typedef struct nfs4_rfact { 759 nfs4_fact_type_t rf_type; 760 nfs4_fact_status_t rf_status; 761 bool_t rf_reboot; 762 nfs4_recov_t rf_action; 763 nfs_opnum4 rf_op; 764 nfsstat4 rf_stat4; 765 timespec_t rf_time; 766 int rf_error; 767 struct rnode4 *rf_rp1; 768 char *rf_char1; 769 } nfs4_rfact_t; 770 771 typedef struct nfs4_revent { 772 nfs4_event_type_t re_type; 773 nfsstat4 re_stat4; 774 uint_t re_uint; 775 pid_t re_pid; 776 struct mntinfo4 *re_mi; 777 struct rnode4 *re_rp1; 778 struct rnode4 *re_rp2; 779 char *re_char1; 780 char *re_char2; 781 nfs4_tag_type_t re_tag1; 782 nfs4_tag_type_t re_tag2; 783 seqid4 re_seqid1; 784 seqid4 re_seqid2; 785 } nfs4_revent_t; 786 787 typedef enum { 788 RM_EVENT, 789 RM_FACT 790 } nfs4_msg_type_t; 791 792 typedef struct nfs4_debug_msg { 793 timespec_t msg_time; 794 nfs4_msg_type_t msg_type; 795 char *msg_srv; 796 char *msg_mntpt; 797 union { 798 nfs4_rfact_t msg_fact; 799 nfs4_revent_t msg_event; 800 } rmsg_u; 801 nfs4_msg_status_t msg_status; 802 list_node_t msg_node; 803 } nfs4_debug_msg_t; 804 805 /* 806 * NFS private data per mounted file system 807 * The mi_lock mutex protects the following fields: 808 * mi_flags 809 * mi_in_recovery 810 * mi_recovflags 811 * mi_recovthread 812 * mi_error 813 * mi_printed 814 * mi_down 815 * mi_stsize 816 * mi_curread 817 * mi_curwrite 818 * mi_timers 819 * mi_curr_serv 820 * mi_klmconfig 821 * mi_oo_list 822 * mi_foo_list 823 * mi_foo_num 824 * mi_foo_max 825 * mi_lost_state 826 * mi_bseqid_list 827 * 828 * Normally the netconfig information for the mount comes from 829 * mi_curr_serv and mi_klmconfig is NULL. If NLM calls need to use a 830 * different transport, mi_klmconfig contains the necessary netconfig 831 * information. 832 * 833 * The mi_async_lock mutex protects the following fields: 834 * mi_async_reqs 835 * mi_async_req_count 836 * mi_async_tail 837 * mi_async_curr 838 * mi_async_clusters 839 * mi_async_init_clusters 840 * mi_threads 841 * mi_inactive_thread 842 * mi_manager_thread 843 * 844 * The nfs4_server_t::s_lock protects the following fields: 845 * mi_clientid 846 * mi_clientid_next 847 * mi_clientid_prev 848 * mi_open_files 849 * mi_srvsettime 850 * 851 * Locking order: 852 * mi4_globals::mig_lock > mi_async_lock 853 * mi_async_lock > nfs4_server_t::s_lock > mi_lock 854 * mi_recovlock > mi_rename_lock > nfs_rtable4_lock 855 * nfs4_server_t::s_recovlock > mi_recovlock 856 * rnode4_t::r_rwlock > mi_rename_lock 857 * nfs_rtable4_lock > mi_lock 858 * nfs4_server_t::s_lock > mi_msg_list_lock 859 * 860 * The 'mi_oo_list' represents the hash buckets that contain the 861 * nfs4_open_owenrs for this particular mntinfo4. 862 * 863 * The 'mi_foo_list' represents the freed nfs4_open_owners for this mntinfo4. 864 * 'mi_foo_num' is the current number of freed open owners on the list, 865 * 'mi_foo_max' is the maximum number of freed open owners that are allowable 866 * on the list. 867 * 868 * mi_rootfh and mi_srvparentfh are read-only once created, but that just 869 * refers to the pointer. The contents must be updated to keep in sync 870 * with mi_curr_serv. 871 * 872 * The mi_msg_list_lock protects against adding/deleting entries to the 873 * mi_msg_list, and also the updating/retrieving of mi_lease_period; 874 * 875 * 'mi_zone' is initialized at structure creation time, and never 876 * changes; it may be read without a lock. 877 * 878 * mi_zone_node is linkage into the mi4_globals.mig_list, and is 879 * protected by mi4_globals.mig_list_lock. 880 */ 881 struct zone; 882 typedef struct mntinfo4 { 883 kmutex_t mi_lock; /* protects mntinfo4 fields */ 884 struct servinfo4 *mi_servers; /* server list */ 885 struct servinfo4 *mi_curr_serv; /* current server */ 886 struct nfs4_sharedfh *mi_rootfh; /* root filehandle */ 887 struct nfs4_sharedfh *mi_srvparentfh; /* root's parent on server */ 888 kcondvar_t mi_failover_cv; /* failover synchronization */ 889 struct vfs *mi_vfsp; /* back pointer to vfs */ 890 enum vtype mi_type; /* file type of the root vnode */ 891 uint_t mi_flags; /* see below */ 892 uint_t mi_recovflags; /* if recovery active; see below */ 893 kthread_t *mi_recovthread; /* active recov thread or NULL */ 894 uint_t mi_error; /* only set/valid when MI4_RECOV_FAIL */ 895 /* is set in mi_flags */ 896 int mi_tsize; /* transfer size (bytes) */ 897 /* really read size */ 898 int mi_stsize; /* server's max transfer size (bytes) */ 899 /* really write size */ 900 int mi_timeo; /* inital timeout in 10th sec */ 901 int mi_retrans; /* times to retry request */ 902 hrtime_t mi_acregmin; /* min time to hold cached file attr */ 903 hrtime_t mi_acregmax; /* max time to hold cached file attr */ 904 hrtime_t mi_acdirmin; /* min time to hold cached dir attr */ 905 hrtime_t mi_acdirmax; /* max time to hold cached dir attr */ 906 len_t mi_maxfilesize; /* for pathconf _PC_FILESIZEBITS */ 907 int mi_curread; /* current read size */ 908 int mi_curwrite; /* current write size */ 909 /* 910 * async I/O management. There may be a pool of threads to handle 911 * async I/O requests, etc., plus there is always one thread that 912 * handles over-the-wire requests for VOP_INACTIVE. The async pool 913 * can also help out with VOP_INACTIVE. 914 */ 915 struct nfs4_async_reqs *mi_async_reqs[NFS4_ASYNC_TYPES]; 916 struct nfs4_async_reqs *mi_async_tail[NFS4_ASYNC_TYPES]; 917 struct nfs4_async_reqs **mi_async_curr; /* current async queue */ 918 uint_t mi_async_clusters[NFS4_ASYNC_TYPES]; 919 uint_t mi_async_init_clusters; 920 uint_t mi_async_req_count; /* # outstanding work requests */ 921 kcondvar_t mi_async_reqs_cv; /* signaled when there's work */ 922 ushort_t mi_threads; /* number of active async threads */ 923 ushort_t mi_max_threads; /* max number of async threads */ 924 kthread_t *mi_manager_thread; /* async manager thread id */ 925 kthread_t *mi_inactive_thread; /* inactive thread id */ 926 kcondvar_t mi_inact_req_cv; /* notify VOP_INACTIVE thread */ 927 kcondvar_t mi_async_work_cv; /* tell workers to work */ 928 kcondvar_t mi_async_cv; /* all pool threads exited */ 929 kmutex_t mi_async_lock; 930 /* 931 * Other stuff 932 */ 933 struct pathcnf *mi_pathconf; /* static pathconf kludge */ 934 rpcprog_t mi_prog; /* RPC program number */ 935 rpcvers_t mi_vers; /* RPC program version number */ 936 char **mi_rfsnames; /* mapping to proc names */ 937 kstat_named_t *mi_reqs; /* count of requests */ 938 clock_t mi_printftime; /* last error printf time */ 939 nfs_rwlock_t mi_recovlock; /* separate ops from recovery (v4) */ 940 time_t mi_grace_wait; /* non-zero represents time to wait */ 941 time_t mi_srvsettime; /* when we switched nfs4_server_t */ 942 nfs_rwlock_t mi_rename_lock; /* atomic volfh rename */ 943 struct nfs4_fname *mi_fname; /* root fname */ 944 list_t mi_lost_state; /* resend list */ 945 list_t mi_bseqid_list; /* bad seqid list */ 946 /* 947 * Client Side Failover stats 948 */ 949 uint_t mi_noresponse; /* server not responding count */ 950 uint_t mi_failover; /* failover to new server count */ 951 uint_t mi_remap; /* remap to new server count */ 952 /* 953 * Kstat statistics 954 */ 955 struct kstat *mi_io_kstats; 956 struct kstat *mi_ro_kstats; 957 kstat_t *mi_recov_ksp; /* ptr to the recovery kstat */ 958 959 /* 960 * Volatile fh flags (nfsv4) 961 */ 962 uint32_t mi_fh_expire_type; 963 /* 964 * Lease Management 965 */ 966 struct mntinfo4 *mi_clientid_next; 967 struct mntinfo4 *mi_clientid_prev; 968 clientid4 mi_clientid; /* redundant info found in nfs4_server */ 969 int mi_open_files; /* count of open files */ 970 int mi_in_recovery; /* count of recovery instances */ 971 kcondvar_t mi_cv_in_recov; /* cv for recovery threads */ 972 /* 973 * Open owner stuff. 974 */ 975 struct nfs4_oo_hash_bucket mi_oo_list[NFS4_NUM_OO_BUCKETS]; 976 list_t mi_foo_list; 977 int mi_foo_num; 978 int mi_foo_max; 979 /* 980 * Shared filehandle pool. 981 */ 982 nfs_rwlock_t mi_fh_lock; 983 avl_tree_t mi_filehandles; 984 /* 985 * fileid map 986 * 987 * Used on servers with volatile filehandles, to check for 988 * pre-existing rnodes with a given fileid, before we create 989 * another rnode for the same file on the server. 990 */ 991 kmutex_t mi_fileid_lock; 992 avl_tree_t mi_fileid_map; 993 994 /* 995 * Debug message queue. 996 */ 997 list_t mi_msg_list; 998 int mi_msg_count; 999 time_t mi_lease_period; 1000 /* 1001 * not guaranteed to be accurate. 1002 * only should be used by debug queue. 1003 */ 1004 kmutex_t mi_msg_list_lock; 1005 /* 1006 * Zones support. 1007 */ 1008 struct zone *mi_zone; /* Zone mounted in */ 1009 list_node_t mi_zone_node; /* linkage into per-zone mi list */ 1010 } mntinfo4_t; 1011 1012 /* 1013 * The values for mi_flags. 1014 * 1015 * MI4_HARD hard or soft mount 1016 * MI4_PRINTED responding message printed 1017 * MI4_INT allow INTR on hard mount 1018 * MI4_DOWN server is down 1019 * MI4_NOAC don't cache attributes 1020 * MI4_NOCTO no close-to-open consistency 1021 * MI4_LLOCK local locking only (no lockmgr) 1022 * MI4_GRPID System V group id inheritance 1023 * MI4_SHUTDOWN System is rebooting or shutting down 1024 * MI4_LINK server supports link 1025 * MI4_SYMLINK server supports symlink 1026 * MI4_ACL server supports NFSv4 ACLs 1027 * MI4_NOPRINT don't print messages 1028 * MI4_DIRECTIO do direct I/O 1029 * MI4_RECOV_ACTIV filesystem has recovery a thread 1030 * MI4_REMOVE_ON_LAST_CLOSE remove from server's list 1031 * MI4_RECOV_FAIL client recovery failed 1032 * MI4_PUBLIC public/url option used 1033 * MI4_MOUNTING mount in progress, don't failover 1034 * MI4_POSIX_LOCK if server is using POSIX locking 1035 * MI4_LOCK_DEBUG cmn_err'd posix lock err msg 1036 * MI4_DEAD last VFS_RELE() called on mount 1037 * MI4_INACTIVE_IDLE inactive thread idle 1038 * MI4_BADOWNER_DEBUG badowner error msg per mount 1039 * MI4_ASYNC_MGR_STOP tell async manager to die 1040 * MI4_TIMEDOUT saw a timeout during zone shutdown 1041 */ 1042 #define MI4_HARD 0x1 1043 #define MI4_PRINTED 0x2 1044 #define MI4_INT 0x4 1045 #define MI4_DOWN 0x8 1046 #define MI4_NOAC 0x10 1047 #define MI4_NOCTO 0x20 1048 #define MI4_LLOCK 0x80 1049 #define MI4_GRPID 0x100 1050 #define MI4_SHUTDOWN 0x200 1051 #define MI4_LINK 0x400 1052 #define MI4_SYMLINK 0x800 1053 /* 0x1000 is available */ 1054 #define MI4_ACL 0x2000 1055 /* 0x4000 is available */ 1056 /* 0x8000 is available */ 1057 /* 0x10000 is available */ 1058 #define MI4_NOPRINT 0x20000 1059 #define MI4_DIRECTIO 0x40000 1060 /* 0x80000 is available */ 1061 #define MI4_RECOV_ACTIV 0x100000 1062 #define MI4_REMOVE_ON_LAST_CLOSE 0x200000 1063 #define MI4_RECOV_FAIL 0x400000 1064 #define MI4_PUBLIC 0x800000 1065 #define MI4_MOUNTING 0x1000000 1066 #define MI4_POSIX_LOCK 0x2000000 1067 #define MI4_LOCK_DEBUG 0x4000000 1068 #define MI4_DEAD 0x8000000 1069 #define MI4_INACTIVE_IDLE 0x10000000 1070 #define MI4_BADOWNER_DEBUG 0x20000000 1071 #define MI4_ASYNC_MGR_STOP 0x40000000 1072 #define MI4_TIMEDOUT 0x80000000 1073 1074 #define INTR4(vp) (VTOMI4(vp)->mi_flags & MI4_INT) 1075 1076 #define FAILOVER_MOUNT4(mi) (mi->mi_servers->sv_next) 1077 1078 /* 1079 * Recovery flags. 1080 * 1081 * MI4R_NEED_CLIENTID is sort of redundant (it's the nfs4_server_t flag 1082 * that's important), but some flag is needed to indicate that recovery is 1083 * going on for the filesystem. 1084 */ 1085 #define MI4R_NEED_CLIENTID 0x1 1086 #define MI4R_REOPEN_FILES 0x2 1087 #define MI4R_NEED_SECINFO 0x4 1088 #define MI4R_NEED_NEW_SERVER 0x8 1089 #define MI4R_REMAP_FILES 0x10 1090 #define MI4R_SRV_REBOOT 0x20 /* server has rebooted */ 1091 #define MI4R_LOST_STATE 0x40 1092 #define MI4R_BAD_SEQID 0x80 1093 1094 /* 1095 * vfs pointer to mount info 1096 */ 1097 #define VFTOMI4(vfsp) ((mntinfo4_t *)((vfsp)->vfs_data)) 1098 1099 /* 1100 * vnode pointer to mount info 1101 */ 1102 #define VTOMI4(vp) ((mntinfo4_t *)(((vp)->v_vfsp)->vfs_data)) 1103 1104 /* 1105 * Lease Management 1106 * 1107 * lease_valid is initially set to NFS4_LEASE_NOT_STARTED. This is when the 1108 * nfs4_server is first created. lease_valid is then set to 1109 * NFS4_LEASE_UNITIALIZED when the renew thread is started. The extra state of 1110 * NFS4_LEASE_NOT_STARTED is needed for client recovery (so we know if a thread 1111 * already exists when we do SETCLIENTID). lease_valid is then set to 1112 * NFS4_LEASE_VALID (if it is at NFS4_LEASE_UNITIALIZED) when a state creating 1113 * operation (OPEN) is done. lease_valid stays at NFS4_LEASE_VALID as long as 1114 * the lease is renewed. It is set to NFS4_LEASE_INVALID when the lease 1115 * expires. Client recovery is needed to set the lease back to 1116 * NFS4_LEASE_VALID from NFS4_LEASE_INVALID. 1117 * 1118 * The s_cred is the credential used to mount the first file system for this 1119 * server. It used as the credential for the renew thread's calls to the 1120 * server. 1121 * 1122 * The renew thread waits on the condition variable cv_thread_exit. If the cv 1123 * is signalled, then the thread knows it must check s_thread_exit to see if 1124 * it should exit. The cv is signaled when the last file system is unmounted 1125 * from a particular server. s_thread_exit is set to 0 upon thread startup, 1126 * and set to NFS4_THREAD_EXIT, when the last file system is unmounted thereby 1127 * telling the thread to exit. s_thread_exit is needed to avoid spurious 1128 * wakeups. 1129 * 1130 * state_ref_count is incremented every time a new file is opened and 1131 * decremented every time a file is closed otw. This keeps track of whether 1132 * the nfs4_server has state associated with it or not. 1133 * 1134 * s_refcnt is the reference count for storage management of the struct 1135 * itself. 1136 * 1137 * mntinfo4_list points to the doubly linked list of mntinfo4s that share 1138 * this nfs4_server (ie: <clientid, saddr> pair) in the current zone. This is 1139 * needed for a nfs4_server to get a mntinfo4 for use in rfs4call. 1140 * 1141 * s_recovlock is used to synchronize recovery operations. The thread 1142 * that is recovering the client must acquire it as a writer. If the 1143 * thread is using the clientid (including recovery operations on other 1144 * state), acquire it as a reader. 1145 * 1146 * The 's_otw_call_count' keeps track of the number of outstanding over the 1147 * wire requests for this structure. The struct will not go away as long 1148 * as this is non-zero (or s_refcnt is non-zero). 1149 * 1150 * The 's_cv_otw_count' is used in conjuntion with the 's_otw_call_count' 1151 * variable to let the renew thread when an outstanding otw request has 1152 * finished. 1153 * 1154 * 'zoneid' and 'zone_globals' are set at creation of this structure 1155 * and are read-only after that; no lock is required to read them. 1156 * 1157 * s_lock protects: everything except cv_thread_exit and s_recovlock. 1158 * 1159 * s_program is used as the index into the nfs4_callback_globals's 1160 * nfs4prog2server table. When a callback request comes in, we can 1161 * use that request's program number (minus NFS4_CALLBACK) as an index 1162 * into the nfs4prog2server. That entry will hold the nfs4_server_t ptr. 1163 * We can then access that nfs4_server_t and its 's_deleg_list' (its list of 1164 * delegated rnode4_ts). 1165 * 1166 * Lock order: 1167 * nfs4_server::s_lock > mntinfo4::mi_lock 1168 * nfs_rtable4_lock > s_lock 1169 * nfs4_server_lst_lock > s_lock 1170 * s_recovlock > s_lock 1171 */ 1172 struct nfs4_callback_globals; 1173 1174 typedef struct nfs4_server { 1175 struct nfs4_server *forw; 1176 struct nfs4_server *back; 1177 struct netbuf saddr; 1178 uint_t s_flags; /* see below */ 1179 uint_t s_refcnt; 1180 clientid4 clientid; /* what we get from server */ 1181 nfs_client_id4 clidtosend; /* what we send to server */ 1182 mntinfo4_t *mntinfo4_list; 1183 int lease_valid; 1184 time_t s_lease_time; 1185 time_t last_renewal_time; 1186 timespec_t propagation_delay; 1187 cred_t *s_cred; 1188 kcondvar_t cv_thread_exit; 1189 int s_thread_exit; 1190 int state_ref_count; 1191 int s_otw_call_count; 1192 kcondvar_t s_cv_otw_count; 1193 kmutex_t s_lock; 1194 list_t s_deleg_list; 1195 rpcprog_t s_program; 1196 nfs_rwlock_t s_recovlock; 1197 kcondvar_t wait_cb_null; /* used to wait for CB_NULL */ 1198 zoneid_t zoneid; /* zone using this nfs4_server_t */ 1199 struct nfs4_callback_globals *zone_globals; /* globals */ 1200 } nfs4_server_t; 1201 1202 /* nfs4_server flags */ 1203 #define N4S_CLIENTID_SET 1 /* server has our clientid */ 1204 #define N4S_INSERTED 0x2 /* server has been put in global list */ 1205 #define N4S_CB_PINGED 0x4 /* server has sent us a CB_NULL */ 1206 #define N4S_CB_WAITER 0x8 /* is/has wait{ing/ed} for cb_null */ 1207 #define N4S_BADOWNER_DEBUG 0x10 /* bad owner err msg per client */ 1208 1209 #define N4S_CB_PAUSE_TIME 10000 /* Amount of time to pause (10ms) */ 1210 1211 struct lease_time_arg { 1212 time_t lease_time; 1213 }; 1214 1215 enum nfs4_delegreturn_policy { 1216 IMMEDIATE, 1217 FIRSTCLOSE, 1218 LASTCLOSE, 1219 INACTIVE 1220 }; 1221 1222 /* 1223 * Operation hints for the recovery framework (mostly). 1224 * 1225 * EXCEPTIONS: 1226 * OH_ACCESS, OH_GETACL, OH_GETATTR, OH_LOOKUP, OH_READDIR 1227 * These hints exist to allow user visit/readdir a R4SRVSTUB dir. 1228 * (dir represents the root of a server fs that has not yet been 1229 * mounted at client) 1230 */ 1231 typedef enum { 1232 OH_OTHER, 1233 OH_READ, 1234 OH_WRITE, 1235 OH_COMMIT, 1236 OH_VFH_RENAME, 1237 OH_MOUNT, 1238 OH_CLOSE, 1239 OH_LOCKU, 1240 OH_DELEGRETURN, 1241 OH_ACCESS, 1242 OH_GETACL, 1243 OH_GETATTR, 1244 OH_LOOKUP, 1245 OH_READDIR 1246 } nfs4_op_hint_t; 1247 1248 /* 1249 * This macro evaluates to non-zero if the given op releases state at the 1250 * server. 1251 */ 1252 #define OH_IS_STATE_RELE(op) ((op) == OH_CLOSE || (op) == OH_LOCKU || \ 1253 (op) == OH_DELEGRETURN) 1254 1255 #ifdef _KERNEL 1256 1257 extern void nfs4_async_manager(struct vfs *); 1258 extern void nfs4_async_manager_stop(struct vfs *); 1259 extern void nfs4_async_stop(struct vfs *); 1260 extern int nfs4_async_stop_sig(struct vfs *); 1261 extern int nfs4_async_readahead(vnode_t *, u_offset_t, caddr_t, 1262 struct seg *, cred_t *, 1263 void (*)(vnode_t *, u_offset_t, 1264 caddr_t, struct seg *, cred_t *)); 1265 extern int nfs4_async_putapage(vnode_t *, page_t *, u_offset_t, size_t, 1266 int, cred_t *, int (*)(vnode_t *, page_t *, 1267 u_offset_t, size_t, int, cred_t *)); 1268 extern int nfs4_async_pageio(vnode_t *, page_t *, u_offset_t, size_t, 1269 int, cred_t *, int (*)(vnode_t *, page_t *, 1270 u_offset_t, size_t, int, cred_t *)); 1271 extern void nfs4_async_commit(vnode_t *, page_t *, offset3, count3, 1272 cred_t *, void (*)(vnode_t *, page_t *, 1273 offset3, count3, cred_t *)); 1274 extern void nfs4_async_inactive(vnode_t *, cred_t *); 1275 extern void nfs4_inactive_thread(mntinfo4_t *mi); 1276 extern void nfs4_inactive_otw(vnode_t *, cred_t *); 1277 extern int nfs4_putpages(vnode_t *, u_offset_t, size_t, int, cred_t *); 1278 1279 extern int nfs4_setopts(vnode_t *, model_t, struct nfs_args *); 1280 extern void nfs4_mnt_kstat_init(struct vfs *); 1281 1282 extern void rfs4call(struct mntinfo4 *, struct COMPOUND4args_clnt *, 1283 struct COMPOUND4res_clnt *, cred_t *, int *, int, 1284 nfs4_error_t *); 1285 extern void nfs4_acl_fill_cache(struct rnode4 *, vsecattr_t *); 1286 extern int nfs4_attr_otw(vnode_t *, nfs4_tag_type_t, 1287 nfs4_ga_res_t *, bitmap4, cred_t *); 1288 1289 extern void nfs4_attrcache_noinval(vnode_t *, nfs4_ga_res_t *, hrtime_t); 1290 extern void nfs4_attr_cache(vnode_t *, nfs4_ga_res_t *, 1291 hrtime_t, cred_t *, int, 1292 change_info4 *); 1293 extern void nfs4_purge_rddir_cache(vnode_t *); 1294 extern void nfs4_invalidate_pages(vnode_t *, u_offset_t, cred_t *); 1295 extern void nfs4_purge_caches(vnode_t *, int, cred_t *, int); 1296 extern void nfs4_purge_stale_fh(int, vnode_t *, cred_t *); 1297 1298 extern void nfs4rename_update(vnode_t *, vnode_t *, nfs_fh4 *, char *); 1299 extern void nfs4_update_paths(vnode_t *, char *, vnode_t *, char *, 1300 vnode_t *); 1301 1302 extern void nfs4args_lookup_free(nfs_argop4 *, int); 1303 extern void nfs4args_copen_free(OPEN4cargs *); 1304 1305 extern void nfs4_printfhandle(nfs4_fhandle_t *); 1306 1307 extern void nfs_free_mi4(mntinfo4_t *); 1308 extern void sv4_free(servinfo4_t *); 1309 extern void nfs4_mi_zonelist_add(mntinfo4_t *); 1310 extern int nfs4_secinfo_recov(mntinfo4_t *, vnode_t *, vnode_t *); 1311 extern void nfs4_secinfo_init(void); 1312 extern void nfs4_secinfo_fini(void); 1313 extern int nfs4_secinfo_path(mntinfo4_t *, cred_t *, int); 1314 extern int nfs4_secinfo_vnode_otw(vnode_t *, char *, cred_t *); 1315 extern void secinfo_free(sv_secinfo_t *); 1316 extern void save_mnt_secinfo(servinfo4_t *); 1317 extern void check_mnt_secinfo(servinfo4_t *, vnode_t *); 1318 extern int vattr_to_fattr4(vattr_t *, vsecattr_t *, fattr4 *, int, 1319 enum nfs_opnum4, bitmap4 supp_mask); 1320 extern int nfs4_putapage(vnode_t *, page_t *, u_offset_t *, size_t *, 1321 int, cred_t *); 1322 extern void nfs4_write_error(vnode_t *, int, cred_t *); 1323 extern void nfs4_lockcompletion(vnode_t *, int); 1324 extern bool_t nfs4_map_lost_lock_conflict(vnode_t *); 1325 extern int vtodv(vnode_t *, vnode_t **, cred_t *, bool_t); 1326 extern void nfs4open_confirm(vnode_t *, seqid4*, stateid4 *, cred_t *, 1327 bool_t, bool_t *, nfs4_open_owner_t *, bool_t, 1328 nfs4_error_t *, int *); 1329 extern void nfs4_error_zinit(nfs4_error_t *); 1330 extern void nfs4_error_init(nfs4_error_t *, int); 1331 #ifdef DEBUG 1332 extern int nfs4_consistent_type(vnode_t *); 1333 #endif 1334 1335 extern void nfs4_init_dot_entries(void); 1336 extern void nfs4_destroy_dot_entries(void); 1337 extern struct nfs4_callback_globals *nfs4_get_callback_globals(void); 1338 1339 extern struct nfs4_server nfs4_server_lst; 1340 1341 extern clock_t nfs_write_error_interval; 1342 1343 #endif /* _KERNEL */ 1344 1345 /* 1346 * Flags for nfs4getfh_otw. 1347 */ 1348 1349 #define NFS4_GETFH_PUBLIC 0x01 1350 #define NFS4_GETFH_NEEDSOP 0x02 1351 1352 /* 1353 * Found through rnodes. 1354 * 1355 * The os_open_ref_count keeps track the number of open file descriptor 1356 * refernces on this data structure. It will be bumped for any successful 1357 * OTW OPEN call and any OPEN call that determines the OTW call is not 1358 * necessary and the open stream hasn't just been created (see 1359 * nfs4_is_otw_open_necessary). 1360 * 1361 * os_mapcnt is a count of the number of mmapped pages for a particular 1362 * open stream; this in conjunction w/ os_open_ref_count is used to 1363 * determine when to do a close to the server. This is necessary because 1364 * of the semantics of doing open, mmap, close; the OTW close must be wait 1365 * until all open and mmap references have vanished. 1366 * 1367 * 'os_valid' tells us whether this structure is about to be freed or not, 1368 * if it is then don't return it in find_open_stream(). 1369 * 1370 * 'os_final_close' is set when a CLOSE OTW was attempted. This is needed 1371 * so we can properly count the os_open_ref_count in cases where we VOP_CLOSE 1372 * without a VOP_OPEN, and have nfs4_inactive() drive the OTW CLOSE. It 1373 * also helps differentiate the VOP_OPEN/VN_RELE case from the VOP_CLOSE 1374 * that tried to close OTW but failed, and left the state cleanup to 1375 * nfs4_inactive/CLOSE_FORCE. 1376 * 1377 * 'os_force_close' is used to let us know if an intervening thread came 1378 * and reopened the open stream after we decided to issue a CLOSE_FORCE, 1379 * but before we could actually process the CLOSE_FORCE. 1380 * 1381 * 'os_pending_close' is set when an over-the-wire CLOSE is deferred to the 1382 * lost state queue. 1383 * 1384 * 'open_stateid' is set the last open stateid returned by the server unless 1385 * 'os_delegation' is 1, in which case 'open_stateid' refers to the 1386 * delegation stateid returned by the server. This is used in cases where the 1387 * client tries to OPEN a file but already has a suitable delegation, so we 1388 * just stick the delegation stateid in the open stream. 1389 * 1390 * os_dc_openacc are open access bits which have been granted to the 1391 * open stream by virtue of a delegation, but which have not been seen 1392 * by the server. This applies even if the open stream does not have 1393 * os_delegation set. These bits are used when setting file locks to 1394 * determine whether an open with CLAIM_DELEGATE_CUR needs to be done 1395 * before the lock request can be sent to the server. See 1396 * nfs4frlock_check_deleg(). 1397 * 1398 * 'os_mmap_read/write' keep track of the read and write access our memory 1399 * maps require. We need to keep track of this so we can provide the proper 1400 * access bits in the open/mmap/close/reboot/reopen case. 1401 * 1402 * 'os_failed_reopen' tells us that we failed to successfully reopen this 1403 * open stream; therefore, we should not use this open stateid as it is 1404 * not valid anymore. This flag is also used to indicate an unsuccessful 1405 * attempt to reopen a delegation open stream with CLAIM_DELEGATE_CUR. 1406 * 1407 * If 'os_orig_oo_name' is different than os_open_owner's oo_name 1408 * then this tells us that this open stream's open owner used a 1409 * bad seqid (that is, got NFS4ERR_BAD_SEQID). If different, this open 1410 * stream will no longer be used for future OTW state releasing calls. 1411 * 1412 * Lock ordering: 1413 * rnode4_t::r_os_lock > os_sync_lock 1414 * os_sync_lock > rnode4_t::r_statelock 1415 * os_sync_lock > rnode4_t::r_statev4_lock 1416 * os_sync_lock > mntinfo4_t::mi_lock (via hold over rfs4call) 1417 * 1418 * The 'os_sync_lock' protects: 1419 * open_stateid 1420 * os_dc_openacc 1421 * os_delegation 1422 * os_failed_reopen 1423 * os_final_close 1424 * os_force_close 1425 * os_mapcnt 1426 * os_mmap_read 1427 * os_mmap_write 1428 * os_open_ref_count 1429 * os_pending_close 1430 * os_share_acc_read 1431 * os_share_acc_write 1432 * os_share_deny_none 1433 * os_share_deny_read 1434 * os_share_deny_write 1435 * os_ref_count 1436 * os_valid 1437 * 1438 * The rnode4_t::r_os_lock protects: 1439 * os_node 1440 * 1441 * These fields are set at creation time and 1442 * read only after that: 1443 * os_open_owner 1444 * os_orig_oo_name 1445 */ 1446 typedef struct nfs4_open_stream { 1447 uint64_t os_share_acc_read; 1448 uint64_t os_share_acc_write; 1449 uint64_t os_mmap_read; 1450 uint64_t os_mmap_write; 1451 uint32_t os_share_deny_none; 1452 uint32_t os_share_deny_read; 1453 uint32_t os_share_deny_write; 1454 stateid4 open_stateid; 1455 int os_dc_openacc; 1456 int os_ref_count; 1457 unsigned os_valid:1; 1458 unsigned os_delegation:1; 1459 unsigned os_final_close:1; 1460 unsigned os_pending_close:1; 1461 unsigned os_failed_reopen:1; 1462 unsigned os_force_close:1; 1463 int os_open_ref_count; 1464 long os_mapcnt; 1465 list_node_t os_node; 1466 struct nfs4_open_owner *os_open_owner; 1467 uint64_t os_orig_oo_name; 1468 kmutex_t os_sync_lock; 1469 } nfs4_open_stream_t; 1470 1471 /* 1472 * This structure describes the format of the lock_owner_name 1473 * field of the lock owner. 1474 */ 1475 1476 typedef struct nfs4_lo_name { 1477 uint64_t ln_seq_num; 1478 pid_t ln_pid; 1479 } nfs4_lo_name_t; 1480 1481 /* 1482 * Flags for lo_flags. 1483 */ 1484 #define NFS4_LOCK_SEQID_INUSE 0x1 1485 #define NFS4_BAD_SEQID_LOCK 0x2 1486 1487 /* 1488 * The lo_prev_rnode and lo_next_rnode are for a circular list that hangs 1489 * off the rnode. If the links are NULL it means this object is not on the 1490 * list. 1491 * 1492 * 'lo_pending_rqsts' is non-zero if we ever tried to send a request and 1493 * didn't get a response back. This is used to figure out if we have 1494 * possible remote v4 locks, so that we can clean up at process exit. In 1495 * theory, the client should be able to figure out if the server received 1496 * the request (based on what seqid works), so maybe we can get rid of this 1497 * flag someday. 1498 * 1499 * 'lo_ref_count' tells us how many processes/threads are using this data 1500 * structure. The rnode's list accounts for one reference. 1501 * 1502 * 'lo_just_created' is set to NFS4_JUST_CREATED when we first create the 1503 * data structure. It is then set to NFS4_PERM_CREATED when a lock request 1504 * is successful using this lock owner structure. We need to keep 'temporary' 1505 * lock owners around so we can properly keep the lock seqid synchronization 1506 * when multiple processes/threads are trying to create the lock owner for the 1507 * first time (especially with the DENIED error case). Once 1508 * 'lo_just_created' is set to NFS4_PERM_CREATED, it doesn't change. 1509 * 1510 * 'lo_valid' tells us whether this structure is about to be freed or not, 1511 * if it is then don't return it from find_lock_owner(). 1512 * 1513 * Retrieving and setting of 'lock_seqid' is protected by the 1514 * NFS4_LOCK_SEQID_INUSE flag. Waiters for NFS4_LOCK_SEQID_INUSE should 1515 * use 'lo_cv_seqid_sync'. 1516 * 1517 * The setting of 'lock_stateid' is protected by the 1518 * NFS4_LOCK_SEQID_INUSE flag and 'lo_lock'. The retrieving of the 1519 * 'lock_stateid' is protected by 'lo_lock', with the additional 1520 * requirement that the calling function can handle NFS4ERR_OLD_STATEID and 1521 * NFS4ERR_BAD_STATEID as appropiate. 1522 * 1523 * The setting of NFS4_BAD_SEQID_LOCK to lo_flags tells us whether this lock 1524 * owner used a bad seqid (that is, got NFS4ERR_BAD_SEQID). With this set, 1525 * this lock owner will no longer be used for future OTW calls. Once set, 1526 * it is never unset. 1527 * 1528 * Lock ordering: 1529 * rnode4_t::r_statev4_lock > lo_lock 1530 */ 1531 typedef struct nfs4_lock_owner { 1532 struct nfs4_lock_owner *lo_next_rnode; 1533 struct nfs4_lock_owner *lo_prev_rnode; 1534 int lo_pid; 1535 stateid4 lock_stateid; 1536 seqid4 lock_seqid; 1537 /* 1538 * Fix this to always be 12 bytes 1539 */ 1540 nfs4_lo_name_t lock_owner_name; 1541 int lo_ref_count; 1542 int lo_valid; 1543 int lo_pending_rqsts; 1544 int lo_just_created; 1545 int lo_flags; 1546 kcondvar_t lo_cv_seqid_sync; 1547 kmutex_t lo_lock; 1548 kthread_t *lo_seqid_holder; /* debugging aid */ 1549 } nfs4_lock_owner_t; 1550 1551 /* for nfs4_lock_owner_t lookups */ 1552 typedef enum {LOWN_ANY, LOWN_VALID_STATEID} lown_which_t; 1553 1554 /* Number of times to retry a call that fails with state independent error */ 1555 #define NFS4_NUM_RECOV_RETRIES 3 1556 1557 typedef enum { 1558 NO_SID, 1559 DEL_SID, 1560 LOCK_SID, 1561 OPEN_SID, 1562 SPEC_SID 1563 } nfs4_stateid_type_t; 1564 1565 typedef struct nfs4_stateid_types { 1566 stateid4 d_sid; 1567 stateid4 l_sid; 1568 stateid4 o_sid; 1569 nfs4_stateid_type_t cur_sid_type; 1570 } nfs4_stateid_types_t; 1571 1572 /* 1573 * Per-zone data for dealing with callbacks. Included here solely for the 1574 * benefit of MDB. 1575 */ 1576 struct nfs4_callback_stats { 1577 kstat_named_t delegations; 1578 kstat_named_t cb_getattr; 1579 kstat_named_t cb_recall; 1580 kstat_named_t cb_null; 1581 kstat_named_t cb_dispatch; 1582 kstat_named_t delegaccept_r; 1583 kstat_named_t delegaccept_rw; 1584 kstat_named_t delegreturn; 1585 kstat_named_t callbacks; 1586 kstat_named_t claim_cur; 1587 kstat_named_t claim_cur_ok; 1588 kstat_named_t recall_trunc; 1589 kstat_named_t recall_failed; 1590 kstat_named_t return_limit_write; 1591 kstat_named_t return_limit_addmap; 1592 kstat_named_t deleg_recover; 1593 kstat_named_t cb_illegal; 1594 }; 1595 1596 struct nfs4_callback_globals { 1597 kmutex_t nfs4_cb_lock; 1598 kmutex_t nfs4_dlist_lock; 1599 int nfs4_program_hint; 1600 /* this table maps the program number to the nfs4_server structure */ 1601 struct nfs4_server **nfs4prog2server; 1602 list_t nfs4_dlist; 1603 list_t nfs4_cb_ports; 1604 struct nfs4_callback_stats nfs4_callback_stats; 1605 #ifdef DEBUG 1606 int nfs4_dlistadd_c; 1607 int nfs4_dlistclean_c; 1608 #endif 1609 }; 1610 1611 typedef enum { 1612 CLOSE_NORM, 1613 CLOSE_DELMAP, 1614 CLOSE_FORCE, 1615 CLOSE_RESEND, 1616 CLOSE_AFTER_RESEND 1617 } nfs4_close_type_t; 1618 1619 /* 1620 * Structure to hold the bad seqid information that is passed 1621 * to the recovery framework. 1622 */ 1623 typedef struct nfs4_bseqid_entry { 1624 nfs4_open_owner_t *bs_oop; 1625 nfs4_lock_owner_t *bs_lop; 1626 vnode_t *bs_vp; 1627 pid_t bs_pid; 1628 nfs4_tag_type_t bs_tag; 1629 seqid4 bs_seqid; 1630 list_node_t bs_node; 1631 } nfs4_bseqid_entry_t; 1632 1633 #ifdef _KERNEL 1634 1635 extern void nfs4close_one(vnode_t *, nfs4_open_stream_t *, cred_t *, int, 1636 nfs4_lost_rqst_t *, nfs4_error_t *, nfs4_close_type_t, 1637 size_t, uint_t, uint_t); 1638 extern void nfs4close_notw(vnode_t *, nfs4_open_stream_t *, int *); 1639 extern void nfs4_set_lock_stateid(nfs4_lock_owner_t *, stateid4); 1640 extern void open_owner_hold(nfs4_open_owner_t *); 1641 extern void open_owner_rele(nfs4_open_owner_t *); 1642 extern nfs4_open_stream_t *find_or_create_open_stream(nfs4_open_owner_t *, 1643 struct rnode4 *, int *); 1644 extern nfs4_open_stream_t *find_open_stream(nfs4_open_owner_t *, 1645 struct rnode4 *); 1646 extern nfs4_open_stream_t *create_open_stream(nfs4_open_owner_t *oop, 1647 struct rnode4 *rp); 1648 extern void open_stream_hold(nfs4_open_stream_t *); 1649 extern void open_stream_rele(nfs4_open_stream_t *, struct rnode4 *); 1650 extern int nfs4close_all(vnode_t *, cred_t *); 1651 extern void lock_owner_hold(nfs4_lock_owner_t *); 1652 extern void lock_owner_rele(nfs4_lock_owner_t *); 1653 extern nfs4_lock_owner_t *create_lock_owner(struct rnode4 *, pid_t); 1654 extern nfs4_lock_owner_t *find_lock_owner(struct rnode4 *, pid_t, lown_which_t); 1655 extern void nfs4_rnode_remove_lock_owner(struct rnode4 *, 1656 nfs4_lock_owner_t *); 1657 extern void nfs4_flush_lock_owners(struct rnode4 *); 1658 extern void nfs4_setlockowner_args(lock_owner4 *, struct rnode4 *, pid_t); 1659 extern void nfs4_set_open_seqid(seqid4, nfs4_open_owner_t *, 1660 nfs4_tag_type_t); 1661 extern void nfs4_set_lock_seqid(seqid4, nfs4_lock_owner_t *); 1662 extern void nfs4_get_and_set_next_open_seqid(nfs4_open_owner_t *, 1663 nfs4_tag_type_t); 1664 extern void nfs4_end_open_seqid_sync(nfs4_open_owner_t *); 1665 extern int nfs4_start_open_seqid_sync(nfs4_open_owner_t *, mntinfo4_t *); 1666 extern void nfs4_end_lock_seqid_sync(nfs4_lock_owner_t *); 1667 extern int nfs4_start_lock_seqid_sync(nfs4_lock_owner_t *, mntinfo4_t *); 1668 extern void nfs4_setup_lock_args(nfs4_lock_owner_t *, nfs4_open_owner_t *, 1669 nfs4_open_stream_t *, clientid4, locker4 *); 1670 extern void nfs4_destroy_open_owner(nfs4_open_owner_t *); 1671 1672 extern void nfs4_renew_lease_thread(nfs4_server_t *); 1673 extern nfs4_server_t *find_nfs4_server(mntinfo4_t *); 1674 extern nfs4_server_t *find_nfs4_server_all(mntinfo4_t *, int all); 1675 extern nfs4_server_t *new_nfs4_server(servinfo4_t *, cred_t *); 1676 extern void nfs4_mark_srv_dead(nfs4_server_t *); 1677 extern nfs4_server_t *servinfo4_to_nfs4_server(servinfo4_t *); 1678 extern void nfs4_inc_state_ref_count(mntinfo4_t *); 1679 extern void nfs4_inc_state_ref_count_nolock(nfs4_server_t *, 1680 mntinfo4_t *); 1681 extern void nfs4_dec_state_ref_count(mntinfo4_t *); 1682 extern void nfs4_dec_state_ref_count_nolock(nfs4_server_t *, 1683 mntinfo4_t *); 1684 extern clientid4 mi2clientid(mntinfo4_t *); 1685 extern int nfs4_server_in_recovery(nfs4_server_t *); 1686 extern bool_t nfs4_server_vlock(nfs4_server_t *, int); 1687 extern nfs4_open_owner_t *create_open_owner(cred_t *, mntinfo4_t *); 1688 extern uint64_t nfs4_get_new_oo_name(void); 1689 extern nfs4_open_owner_t *find_open_owner(cred_t *, int, mntinfo4_t *); 1690 extern nfs4_open_owner_t *find_open_owner_nolock(cred_t *, int, mntinfo4_t *); 1691 extern void nfs4frlock(nfs4_lock_call_type_t, vnode_t *, int, flock64_t *, 1692 int, u_offset_t, cred_t *, nfs4_error_t *, 1693 nfs4_lost_rqst_t *, int *); 1694 extern void nfs4open_dg_save_lost_rqst(int, nfs4_lost_rqst_t *, 1695 nfs4_open_owner_t *, nfs4_open_stream_t *, cred_t *, 1696 vnode_t *, int, int); 1697 extern void nfs4_open_downgrade(int, int, nfs4_open_owner_t *, 1698 nfs4_open_stream_t *, vnode_t *, cred_t *, 1699 nfs4_lost_rqst_t *, nfs4_error_t *, cred_t **, seqid4 *); 1700 extern seqid4 nfs4_get_open_seqid(nfs4_open_owner_t *); 1701 extern cred_t *nfs4_get_otw_cred(cred_t *, mntinfo4_t *, nfs4_open_owner_t *); 1702 extern void nfs4_init_stateid_types(nfs4_stateid_types_t *); 1703 extern void nfs4_save_stateid(stateid4 *, nfs4_stateid_types_t *); 1704 1705 extern kmutex_t nfs4_server_lst_lock; 1706 1707 extern void nfs4callback_destroy(nfs4_server_t *); 1708 extern void nfs4_callback_init(void); 1709 extern void nfs4_callback_fini(void); 1710 extern void nfs4_cb_args(nfs4_server_t *, struct knetconfig *, 1711 SETCLIENTID4args *); 1712 extern void nfs4delegreturn_async(struct rnode4 *, int, bool_t); 1713 1714 extern enum nfs4_delegreturn_policy nfs4_delegreturn_policy; 1715 1716 extern void nfs4_add_mi_to_server(nfs4_server_t *, mntinfo4_t *); 1717 extern void nfs4_remove_mi_from_server(mntinfo4_t *, nfs4_server_t *); 1718 extern nfs4_server_t *nfs4_move_mi(mntinfo4_t *, servinfo4_t *, servinfo4_t *); 1719 extern bool_t nfs4_fs_active(nfs4_server_t *); 1720 extern void nfs4_server_rele(nfs4_server_t *); 1721 extern bool_t inlease(nfs4_server_t *); 1722 extern bool_t nfs4_has_pages(vnode_t *); 1723 extern void nfs4_log_badowner(mntinfo4_t *, nfs_opnum4); 1724 1725 #endif /* _KERNEL */ 1726 1727 /* 1728 * Client State Recovery 1729 */ 1730 1731 /* 1732 * The following defines are used for rs_flags in 1733 * a nfs4_recov_state_t structure. 1734 * 1735 * NFS4_RS_RENAME_HELD Indicates that the mi_rename_lock was held. 1736 * NFS4_RS_GRACE_MSG Set once we have uprintf'ed a grace message. 1737 * NFS4_RS_DELAY_MSG Set once we have uprintf'ed a delay message. 1738 * NFS4_RS_RECALL_HELD1 r_deleg_recall_lock for vp1 was held. 1739 * NFS4_RS_RECALL_HELD2 r_deleg_recall_lock for vp2 was held. 1740 */ 1741 #define NFS4_RS_RENAME_HELD 0x000000001 1742 #define NFS4_RS_GRACE_MSG 0x000000002 1743 #define NFS4_RS_DELAY_MSG 0x000000004 1744 #define NFS4_RS_RECALL_HELD1 0x000000008 1745 #define NFS4_RS_RECALL_HELD2 0x000000010 1746 1747 /* 1748 * Information that is retrieved from nfs4_start_op() and that is 1749 * passed into nfs4_end_op(). 1750 * 1751 * rs_sp is a reference to the nfs4_server that was found, or NULL. 1752 * 1753 * rs_num_retry_despite_err is the number times client retried an 1754 * OTW op despite a recovery error. It is only incremented for hints 1755 * exempt to normal R4RECOVERR processing 1756 * (OH_CLOSE/OH_LOCKU/OH_DELEGRETURN). (XXX this special-case code 1757 * needs review for possible removal.) 1758 * It is initialized wherever nfs4_recov_state_t is declared -- usually 1759 * very near initialization of rs_flags. 1760 */ 1761 typedef struct { 1762 nfs4_server_t *rs_sp; 1763 int rs_flags; 1764 int rs_num_retry_despite_err; 1765 } nfs4_recov_state_t; 1766 1767 /* 1768 * Flags for nfs4_check_remap, nfs4_remap_file and nfs4_remap_root. 1769 */ 1770 1771 #define NFS4_REMAP_CKATTRS 1 1772 #define NFS4_REMAP_NEEDSOP 2 1773 1774 #ifdef _KERNEL 1775 1776 extern int nfs4_is_otw_open_necessary(nfs4_open_owner_t *, int, 1777 vnode_t *, int, int *, int, nfs4_recov_state_t *); 1778 extern void nfs4setclientid(struct mntinfo4 *, struct cred *, bool_t, 1779 nfs4_error_t *); 1780 extern void nfs4_reopen(vnode_t *, nfs4_open_stream_t *, nfs4_error_t *, 1781 open_claim_type4, bool_t, bool_t); 1782 extern void nfs4_remap_root(struct mntinfo4 *, nfs4_error_t *, int); 1783 extern void nfs4_check_remap(mntinfo4_t *mi, vnode_t *vp, int, 1784 nfs4_error_t *); 1785 extern void nfs4_remap_file(mntinfo4_t *mi, vnode_t *vp, int, 1786 nfs4_error_t *); 1787 extern int nfs4_make_dotdot(struct nfs4_sharedfh *, hrtime_t, 1788 vnode_t *, cred_t *, vnode_t **, int); 1789 extern void nfs4_fail_recov(vnode_t *, char *, int, nfsstat4); 1790 1791 extern int nfs4_needs_recovery(nfs4_error_t *, bool_t, vfs_t *); 1792 extern int nfs4_recov_marks_dead(nfsstat4); 1793 extern bool_t nfs4_start_recovery(nfs4_error_t *, struct mntinfo4 *, 1794 vnode_t *, vnode_t *, stateid4 *, 1795 nfs4_lost_rqst_t *, nfs_opnum4, nfs4_bseqid_entry_t *); 1796 extern int nfs4_start_op(struct mntinfo4 *, vnode_t *, vnode_t *, 1797 nfs4_recov_state_t *); 1798 extern void nfs4_end_op(struct mntinfo4 *, vnode_t *, vnode_t *, 1799 nfs4_recov_state_t *, bool_t); 1800 extern int nfs4_start_fop(struct mntinfo4 *, vnode_t *, vnode_t *, 1801 nfs4_op_hint_t, nfs4_recov_state_t *, bool_t *); 1802 extern void nfs4_end_fop(struct mntinfo4 *, vnode_t *, vnode_t *, 1803 nfs4_op_hint_t, nfs4_recov_state_t *, bool_t); 1804 extern char *nfs4_recov_action_to_str(nfs4_recov_t); 1805 1806 extern int wait_for_recall(vnode_t *, vnode_t *, nfs4_op_hint_t, 1807 nfs4_recov_state_t *); 1808 extern void nfs4_end_op_recall(vnode_t *, vnode_t *, nfs4_recov_state_t *); 1809 extern void nfs4_send_siglost(pid_t, mntinfo4_t *mi, vnode_t *vp, bool_t, 1810 int, nfsstat4); 1811 extern time_t nfs4err_delay_time; 1812 extern void nfs4_set_grace_wait(mntinfo4_t *); 1813 extern void nfs4_set_delay_wait(vnode_t *); 1814 extern int nfs4_wait_for_grace(mntinfo4_t *, nfs4_recov_state_t *); 1815 extern int nfs4_wait_for_delay(vnode_t *, nfs4_recov_state_t *); 1816 extern nfs4_bseqid_entry_t *nfs4_create_bseqid_entry(nfs4_open_owner_t *, 1817 nfs4_lock_owner_t *, vnode_t *, pid_t, nfs4_tag_type_t, 1818 seqid4); 1819 1820 extern void nfs4_resend_open_otw(vnode_t **, nfs4_lost_rqst_t *, 1821 nfs4_error_t *); 1822 extern void nfs4_resend_delegreturn(nfs4_lost_rqst_t *, nfs4_error_t *, 1823 nfs4_server_t *); 1824 extern int nfs4_rpc_retry_error(int); 1825 extern int nfs4_try_failover(nfs4_error_t *); 1826 extern void nfs4_free_msg(nfs4_debug_msg_t *); 1827 extern void nfs4_mnt_recov_kstat_init(vfs_t *); 1828 extern void nfs4_mi_kstat_inc_delay(mntinfo4_t *); 1829 extern void nfs4_mi_kstat_inc_no_grace(mntinfo4_t *); 1830 extern char *nfs4_stat_to_str(nfsstat4); 1831 extern char *nfs4_op_to_str(nfs_opnum4); 1832 1833 extern void nfs4_queue_event(nfs4_event_type_t, mntinfo4_t *, char *, 1834 uint_t, vnode_t *, vnode_t *, nfsstat4, char *, pid_t, 1835 nfs4_tag_type_t, nfs4_tag_type_t, seqid4, seqid4); 1836 extern void nfs4_queue_fact(nfs4_fact_type_t, mntinfo4_t *, nfsstat4, 1837 nfs4_recov_t, nfs_opnum4, bool_t, char *, int, vnode_t *); 1838 #pragma rarely_called(nfs4_queue_event) 1839 #pragma rarely_called(nfs4_queue_fact) 1840 1841 /* Used for preformed "." and ".." dirents */ 1842 extern char *nfs4_dot_entries; 1843 extern char *nfs4_dot_dot_entry; 1844 1845 #ifdef DEBUG 1846 extern uint_t nfs4_tsd_key; 1847 #endif 1848 1849 #endif /* _KERNEL */ 1850 1851 /* 1852 * Filehandle management. 1853 * 1854 * Filehandles can change in v4, so rather than storing the filehandle 1855 * directly in the rnode, etc., we manage the filehandle through one of 1856 * these objects. 1857 * Locking: sfh_fh and sfh_tree is protected by the filesystem's 1858 * mi_fh_lock. The reference count and flags are protected by sfh_lock. 1859 * sfh_mi is read-only. 1860 * 1861 * mntinfo4_t::mi_fh_lock > sfh_lock. 1862 */ 1863 1864 typedef struct nfs4_sharedfh { 1865 nfs_fh4 sfh_fh; /* key and current filehandle */ 1866 kmutex_t sfh_lock; 1867 uint_t sfh_refcnt; /* reference count */ 1868 uint_t sfh_flags; 1869 mntinfo4_t *sfh_mi; /* backptr to filesystem */ 1870 avl_node_t sfh_tree; /* used by avl package */ 1871 } nfs4_sharedfh_t; 1872 1873 #define SFH4_SAME(sfh1, sfh2) ((sfh1) == (sfh2)) 1874 1875 /* 1876 * Flags. 1877 */ 1878 #define SFH4_IN_TREE 0x1 /* currently in an AVL tree */ 1879 1880 #ifdef _KERNEL 1881 1882 extern void sfh4_createtab(avl_tree_t *); 1883 extern nfs4_sharedfh_t *sfh4_get(const nfs_fh4 *, mntinfo4_t *); 1884 extern nfs4_sharedfh_t *sfh4_put(const nfs_fh4 *, mntinfo4_t *, 1885 nfs4_sharedfh_t *); 1886 extern void sfh4_update(nfs4_sharedfh_t *, const nfs_fh4 *); 1887 extern void sfh4_copyval(const nfs4_sharedfh_t *, nfs4_fhandle_t *); 1888 extern void sfh4_hold(nfs4_sharedfh_t *); 1889 extern void sfh4_rele(nfs4_sharedfh_t **); 1890 extern void sfh4_printfhandle(const nfs4_sharedfh_t *); 1891 1892 #endif 1893 1894 /* 1895 * Path and file name management. 1896 * 1897 * This type stores the name of an entry in the filesystem and keeps enough 1898 * information that it can provide a complete path. All fields are 1899 * protected by fn_lock, except for the reference count, which is managed 1900 * using atomic add/subtract. 1901 * 1902 * Lock order: child and then parent. 1903 */ 1904 1905 typedef struct nfs4_fname { 1906 struct nfs4_fname *fn_parent; /* parent name; null if fs root */ 1907 char *fn_name; /* the actual name */ 1908 ssize_t fn_len; /* strlen(fn_name) */ 1909 uint32_t fn_refcnt; /* reference count */ 1910 kmutex_t fn_lock; 1911 avl_node_t fn_tree; 1912 avl_tree_t fn_children; /* children, if any */ 1913 } nfs4_fname_t; 1914 1915 #ifdef _KERNEL 1916 1917 extern vnode_t nfs4_xattr_notsupp_vnode; 1918 #define NFS4_XATTR_DIR_NOTSUPP &nfs4_xattr_notsupp_vnode 1919 1920 extern nfs4_fname_t *fn_get(nfs4_fname_t *, char *); 1921 extern void fn_hold(nfs4_fname_t *); 1922 extern void fn_rele(nfs4_fname_t **); 1923 extern char *fn_name(nfs4_fname_t *); 1924 extern char *fn_path(nfs4_fname_t *); 1925 extern void fn_move(nfs4_fname_t *, nfs4_fname_t *, char *); 1926 extern nfs4_fname_t *fn_parent(nfs4_fname_t *); 1927 1928 #endif 1929 1930 /* 1931 * Per-zone data for managing client handles, included in this file for the 1932 * benefit of MDB. 1933 */ 1934 struct nfs4_clnt { 1935 struct chhead *nfscl_chtable4; 1936 kmutex_t nfscl_chtable4_lock; 1937 zoneid_t nfscl_zoneid; 1938 list_node_t nfscl_node; 1939 struct clstat4 nfscl_stat; 1940 }; 1941 1942 #ifdef __cplusplus 1943 } 1944 #endif 1945 1946 #endif /* _NFS4_CLNT_H */ 1947