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 (the "License").
6 * You may not use this file except in compliance with the License.
7 *
8 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
9 * or http://www.opensolaris.org/os/licensing.
10 * See the License for the specific language governing permissions
11 * and limitations under the License.
12 *
13 * When distributing Covered Code, include this CDDL HEADER in each
14 * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
15 * If applicable, add the following below this CDDL HEADER, with the
16 * fields enclosed by brackets "[]" replaced with your own identifying
17 * information: Portions Copyright [yyyy] [name of copyright owner]
18 *
19 * CDDL HEADER END
20 */
21 /*
22 * Copyright 2009 Sun Microsystems, Inc. All rights reserved.
23 * Use is subject to license terms.
24 */
25
26 /* Copyright (c) 1983, 1984, 1985, 1986, 1987, 1988, 1989 AT&T */
27 /* All Rights Reserved */
28
29 #ifndef _NFS_RNODE4_H
30 #define _NFS_RNODE4_H
31
32 #ifdef __cplusplus
33 extern "C" {
34 #endif
35
36 #include <nfs/rnode.h> /* for symlink_cache, nfs_rwlock_t, etc. */
37 #include <nfs/nfs4.h>
38 #include <nfs/nfs4_clnt.h>
39 #include <sys/thread.h>
40 #include <sys/sysmacros.h> /* for offsetof */
41
42 typedef enum nfs4_stub_type {
43 NFS4_STUB_NONE,
44 NFS4_STUB_MIRRORMOUNT,
45 NFS4_STUB_REFERRAL
46 } nfs4_stub_type_t;
47
48 typedef enum nfs4_access_type {
49 NFS4_ACCESS_UNKNOWN,
50 NFS4_ACCESS_ALLOWED,
51 NFS4_ACCESS_DENIED
52 } nfs4_access_type_t;
53
54 /*
55 * Access cache
56 */
57 typedef struct acache4_hash {
58 struct acache4 *next;
59 struct acache4 *prev;
60 krwlock_t lock;
61 } acache4_hash_t;
62
63 typedef struct acache4 {
64 struct acache4 *next; /* next and prev must be first */
65 struct acache4 *prev;
66 uint32_t known;
67 uint32_t allowed;
68 struct rnode4 *rnode;
69 cred_t *cred;
70 struct acache4 *list;
71 struct acache4_hash *hashq;
72 } acache4_t;
73
74 /*
75 * Note on the different buffer sizes in rddir4_cache:
76 * There seems to be some discrepancy between the intended and actual
77 * use of entlen and buflen, which does not correspond to the comment below.
78 * entlen - nfsv2/3 used as both alloc'd size of entries buffer and
79 * as the actual size of the entries (XXX is this correct?).
80 * nfsv4 will use it only as the alloc'd size.
81 * buflen - used for calculations of readahead.
82 * actlen - added for nfsv4 to serve as the size of the useful
83 * portion of the entries buffer. That is because in
84 * nfsv4, the otw entries are converted to system entries,
85 * and may not be the same size - thus buffer may not be full.
86 */
87 typedef struct rddir4_cache {
88 lloff_t _cookie; /* cookie used to find this cache entry */
89 lloff_t _ncookie; /* cookie used to find the next cache entry */
90 char *entries; /* buffer containing dirent entries */
91 int eof; /* EOF reached after this request */
92 int entlen; /* size of dirent entries in buf */
93 int buflen; /* size of the buffer used to store entries */
94 int actlen; /* size of the actual entries (nfsv4 only) */
95 int flags; /* control flags, see below */
96 kcondvar_t cv; /* cv for blocking */
97 int error; /* error from RPC operation */
98 void *data; /* private data */
99 } rddir4_cache;
100
101 #define nfs4_cookie _cookie._f
102 #define nfs4_ncookie _ncookie._f
103
104 /*
105 * Shadow vnode, v4 only.
106 *
107 * A file's shadow vnode list is protected by its hash bucket lock,
108 * r_hashq->r_lock.
109 *
110 * sv_r_vnode is protected by the appropriate vnode locks.
111 *
112 * sv_dfh, sv_name, sv_dfileid, and sv_dfileid_valid are protected
113 * by rp->r_svlock.
114 */
115
116 typedef struct insq_link {
117 void *forw;
118 void *back;
119 } insq_link_t;
120
121 typedef struct svnode {
122 insq_link_t sv_link; /* must be first for insque */
123 vnode_t *sv_r_vnode; /* vnode for this shadow */
124 nfs4_fname_t *sv_name; /* component name */
125 nfs4_sharedfh_t *sv_dfh; /* directory file handle */
126 } svnode_t;
127
128 #define sv_forw sv_link.forw
129 #define sv_back sv_link.back
130 extern svnode_t *vtosv(vnode_t *);
131 #define VTOSV(vp) vtosv(vp)
132 #define SVTOV(svp) (((svp)->sv_r_vnode))
133 #define IS_SHADOW(vp, rp) ((vp) != (rp)->r_vnode)
134
135 /*
136 * The format of the hash bucket used to lookup rnodes from a file handle.
137 */
138 typedef struct r4hashq {
139 struct rnode4 *r_hashf;
140 struct rnode4 *r_hashb;
141 krwlock_t r_lock;
142 } r4hashq_t;
143
144 /*
145 * Remote file information structure.
146 *
147 * The rnode is the "inode" for remote files. It contains all the
148 * information necessary to handle remote file on the client side.
149 *
150 * Note on file sizes: we keep two file sizes in the rnode: the size
151 * according to the client (r_size) and the size according to the server
152 * (r_attr.va_size). They can differ because we modify r_size during a
153 * write system call (nfs_rdwr), before the write request goes over the
154 * wire (before the file is actually modified on the server). If an OTW
155 * request occurs before the cached data is written to the server the file
156 * size returned from the server (r_attr.va_size) may not match r_size.
157 * r_size is the one we use, in general. r_attr.va_size is only used to
158 * determine whether or not our cached data is valid.
159 *
160 * Each rnode has 5 locks associated with it (not including the rnode
161 * hash table and free list locks):
162 *
163 * r_rwlock: Serializes nfs_write and nfs_setattr requests
164 * and allows nfs_read requests to proceed in parallel.
165 * Serializes reads/updates to directories.
166 *
167 * r_lkserlock: Serializes lock requests with map, write, and
168 * readahead operations.
169 *
170 * r_statelock: Protects all fields in the rnode except for
171 * those listed below. This lock is intented
172 * to be held for relatively short periods of
173 * time (not accross entire putpage operations,
174 * for example).
175 *
176 * r_statev4_lock: Protects the created_v4 flag, the lock_owners list,
177 * and all the delegation fields except r_deleg_list.
178 *
179 * r_os_lock: Protects r_open_streams.
180 *
181 *
182 * The following members are protected by the mutex rp4freelist_lock:
183 * r_freef
184 * r_freeb
185 *
186 * The following members are protected by the hash bucket rwlock:
187 * r_hashf
188 * r_hashb
189 *
190 * r_fh is read-only except when an rnode is created (or recycled from the
191 * free list).
192 *
193 * The following members are protected by nfs4_server_t::s_lock:
194 * r_deleg_list
195 *
196 * Note: r_modaddr is only accessed when the r_statelock mutex is held.
197 * Its value is also controlled via r_rwlock. It is assumed that
198 * there will be only 1 writer active at a time, so it safe to
199 * set r_modaddr and release r_statelock as long as the r_rwlock
200 * writer lock is held.
201 *
202 * r_inmap informs nfs4_read()/write() that there is a call to nfs4_map()
203 * in progress. nfs4_read()/write() check r_inmap to decide whether
204 * to perform directio on the file or not. r_inmap is atomically
205 * incremented in nfs4_map() before the address space routines are
206 * called and atomically decremented just before nfs4_map() exits.
207 * r_inmap is not protected by any lock.
208 *
209 * r_mapcnt tells that the rnode has mapped pages. r_inmap can be 0
210 * while the rnode has mapped pages.
211 *
212 * 64-bit offsets: the code formerly assumed that atomic reads of
213 * r_size were safe and reliable; on 32-bit architectures, this is
214 * not true since an intervening bus cycle from another processor
215 * could update half of the size field. The r_statelock must now
216 * be held whenever any kind of access of r_size is made.
217 *
218 * Lock ordering:
219 * r_rwlock > r_lkserlock > r_os_lock > r_statelock > r_statev4_lock
220 * vnode_t::v_lock > r_os_lock
221 */
222 struct exportinfo; /* defined in nfs/export.h */
223 struct servinfo4; /* defined in nfs/nfs4_clnt.h */
224 struct failinfo; /* defined in nfs/nfs_clnt.h */
225 struct mntinfo4; /* defined in nfs/nfs4_clnt.h */
226
227 typedef struct rnode4 {
228 /* the hash fields must be first to match the rhashq_t */
229 struct rnode4 *r_hashf; /* hash queue forward pointer */
230 struct rnode4 *r_hashb; /* hash queue back pointer */
231 struct rnode4 *r_freef; /* free list forward pointer */
232 struct rnode4 *r_freeb; /* free list back pointer */
233 r4hashq_t *r_hashq; /* pointer to the hash bucket */
234
235 svnode_t r_svnode; /* "master" shadow vnode for file */
236 kmutex_t r_svlock; /* serializes access to svnode list */
237 nfs_rwlock_t r_rwlock; /* serializes write/setattr requests */
238 nfs_rwlock_t r_lkserlock; /* serialize lock with other ops */
239 kmutex_t r_statelock; /* protects (most of) rnode contents */
240 nfs4_sharedfh_t *r_fh; /* file handle */
241 struct servinfo4
242 *r_server; /* current server */
243 u_offset_t r_nextr; /* next byte read offset (read-ahead) */
244 uint_t r_flags; /* flags, see below */
245 short r_error; /* async write error */
246 cred_t *r_unlcred; /* unlinked credentials */
247 char *r_unlname; /* unlinked file name */
248 vnode_t *r_unldvp; /* parent dir of unlinked file */
249 vnode_t *r_xattr_dir; /* cached xattr dir vnode */
250 len_t r_size; /* client's view of file size */
251 vattr_t r_attr; /* cached vnode attributes */
252 hrtime_t r_time_attr_saved; /* time attributes were cached */
253 hrtime_t r_time_attr_inval; /* time attributes become invalid */
254 hrtime_t r_time_cache_inval; /* time caches become invalid */
255 time_t r_delay_wait; /* future time for DELAY handling */
256 int r_delay_interval; /* Number of Secs of last DELAY */
257 time_t r_last_recov; /* time of last recovery operation */
258 nfs4_recov_t r_recov_act; /* action from last recovery op */
259 long r_mapcnt; /* count of mmapped pages */
260 uint_t r_count; /* # of refs not reflect in v_count */
261 uint_t r_awcount; /* # of outstanding async write */
262 uint_t r_gcount; /* getattrs waiting to flush pages */
263 kcondvar_t r_cv; /* condvar for blocked threads */
264 int (*r_putapage) /* address of putapage routine */
265 (vnode_t *, page_t *, u_offset_t *, size_t *, int, cred_t *);
266 void *r_dir; /* cache of readdir responses */
267 rddir4_cache *r_direof; /* pointer to the EOF entry */
268 symlink_cache r_symlink; /* cached readlink response */
269 verifier4 r_writeverf; /* file data write verifier */
270 u_offset_t r_modaddr; /* address for page in writerp */
271 commit_t r_commit; /* commit information */
272 u_offset_t r_truncaddr; /* base for truncate operation */
273 vsecattr_t *r_secattr; /* cached security attributes (acls) */
274 verifier4 r_cookieverf4; /* version 4 readdir cookie verifier */
275 nfs4_pathconf_info_t r_pathconf; /* cached pathconf info */
276 acache4_t *r_acache; /* list of access cache entries */
277 list_t r_open_streams; /* open streams list */
278 kmutex_t r_os_lock; /* protects r_open_streams */
279 nfs4_lock_owner_t
280 r_lo_head; /* lock owners list head */
281 int created_v4; /* 1 if file has been created in v4 */
282 kmutex_t r_statev4_lock; /* protects created_v4, state4ptr */
283
284 list_node_t r_deleg_link; /* linkage into list of */
285 /* delegated rnodes for this server */
286 open_delegation_type4
287 r_deleg_type; /* type of delegation granted */
288 stateid4 r_deleg_stateid;
289 /* delegation state id */
290 nfs_space_limit4
291 r_deleg_limit; /* file limits returned from */
292 /* server on delegated open */
293 nfsace4 r_deleg_perms; /* file permissions returned from */
294 /* server on delegated open */
295 fattr4_change r_deleg_change; /* current deleg change attr */
296 fattr4_change r_deleg_change_grant;
297 /* change @ write deleg grant */
298 cred_t *r_deleg_cred; /* credential in force when the */
299 /* delegation was granted */
300 open_delegation_type4
301 r_deleg_needs_recovery;
302 /* delegation needs recovery */
303 /* This contains the delegation type */
304 /* for use with CLAIM_PREVIOUS. */
305 /* OPEN_DELEGATE_NONE means recovery */
306 /* is not needed. */
307 unsigned r_deleg_needs_recall:1;
308 /* delegation has been recalled by */
309 /* the server during open with */
310 /* CLAIM_PREVIOUS */
311 unsigned r_deleg_return_pending:1;
312 /* delegreturn is pending, don't use */
313 /* the delegation stateid, set in */
314 /* nfs4_dlistadd */
315 unsigned r_deleg_return_inprog:1;
316 /* delegreturn is in progress, may */
317 /* only be set by nfs4delegreturn. */
318 nfs_rwlock_t r_deleg_recall_lock;
319 /* lock for synchronizing delegreturn */
320 /* with in other operations, acquired */
321 /* in read mode by nfs4_start_fop, */
322 /* acquired in write mode in */
323 /* nfs4delegreturn */
324 fattr4_change r_change; /* GETATTR4 change attr; client */
325 /* should always request change */
326 /* when c/mtime requested to keep */
327 /* change and c/mtime in sync */
328 fattr4_fileid r_mntd_fid; /* mounted on fileid attr */
329 kthread_t *r_serial; /* attrcache validation thread */
330 kthread_t *r_pgflush; /* thread flushing page cache */
331 list_t r_indelmap; /* list of delmap callers */
332 fattr4_fsid r_srv_fsid; /* fsid of srv fs containing object */
333 /* when rnode created; compare with */
334 /* sv_fsid (servinfo4_t) to see why */
335 /* stub type was set */
336 nfs4_stub_type_t r_stub_type;
337 /* e.g. mirror-mount or referral */
338 uint_t r_inmap; /* to serialize read/write and mmap */
339 } rnode4_t;
340
341 #define r_vnode r_svnode.sv_r_vnode
342
343 /*
344 * Flags
345 */
346 #define R4READDIRWATTR 0x1 /* Use READDIR with attributes */
347 #define R4DIRTY 0x2 /* dirty pages from write operation */
348 #define R4STALE 0x4 /* stale, don't even attempt to write */
349 #define R4MODINPROGRESS 0x8 /* page modification happening */
350 #define R4TRUNCATE 0x10 /* truncating, don't commit */
351 #define R4HAVEVERF 0x20 /* have a write verifier to compare against */
352 #define R4COMMIT 0x40 /* commit in progress */
353 #define R4COMMITWAIT 0x80 /* someone is waiting to do a commit */
354 #define R4HASHED 0x100 /* rnode is in hash queues */
355 #define R4OUTOFSPACE 0x200 /* an out of space error has happened */
356 #define R4LODANGLERS 0x400 /* rnode has dangling lock_owners to cleanup */
357 #define R4WRITEMODIFIED 0x800 /* file data has been modified by write */
358 #define R4DIRECTIO 0x1000 /* bypass the buffer cache */
359 #define R4RECOVERR 0x2000 /* couldn't recover */
360 #define R4RECEXPFH 0x4000 /* recovering expired filehandle */
361 #define R4RECOVERRP 0x8000 /* R4RECOVERR pending, but not set (yet) */
362 #define R4ISXATTR 0x20000 /* rnode is a named attribute */
363 #define R4DELMAPLIST 0x40000 /* delmap callers tracked for as callback */
364 #define R4PGFLUSH 0x80000 /* page flush thread active */
365 #define R4INCACHEPURGE 0x100000 /* purging caches due to file size change */
366 #define R4LOOKUP 0x200000 /* a lookup has been done in the directory */
367 /*
368 * Convert between vnode and rnode
369 */
370 #define RTOV4(rp) ((rp)->r_vnode)
371 #define VTOR4(vp) ((rnode4_t *)((vp)->v_data))
372
373 #define RP_ISSTUB(rp) (((rp)->r_stub_type != NFS4_STUB_NONE))
374 #define RP_ISSTUB_MIRRORMOUNT(rp) ((rp)->r_stub_type == NFS4_STUB_MIRRORMOUNT)
375 #define RP_ISSTUB_REFERRAL(rp) ((rp)->r_stub_type == NFS4_STUB_REFERRAL)
376
377 /*
378 * Open file instances.
379 */
380
381 typedef struct nfs4_opinst {
382 struct nfs4_opinst *re_next; /* next in list */
383 vnode_t *re_vp; /* held reference */
384 uint32_t re_numosp; /* number of valid open streams */
385 nfs4_open_stream_t **re_osp; /* held reference */
386 } nfs4_opinst_t;
387
388 #ifdef _KERNEL
389
390 extern long nrnode;
391
392 /* Used for r_delay_interval */
393 #define NFS4_INITIAL_DELAY_INTERVAL 1
394 #define NFS4_MAX_DELAY_INTERVAL 20
395
396 /* Used for check_rtable4 */
397 #define NFSV4_RTABLE4_OK 0
398 #define NFSV4_RTABLE4_NOT_FREE_LIST 1
399 #define NFSV4_RTABLE4_DIRTY_PAGES 2
400 #define NFSV4_RTABLE4_POS_R_COUNT 3
401
402 extern rnode4_t *r4find(r4hashq_t *, nfs4_sharedfh_t *, struct vfs *);
403 extern rnode4_t *r4find_unlocked(nfs4_sharedfh_t *, struct vfs *);
404 extern void r4flush(struct vfs *, cred_t *);
405 extern void destroy_rtable4(struct vfs *, cred_t *);
406 extern int check_rtable4(struct vfs *);
407 extern void rp4_addfree(rnode4_t *, cred_t *);
408 extern void rp4_addhash(rnode4_t *);
409 extern void rp4_rmhash(rnode4_t *);
410 extern void rp4_rmhash_locked(rnode4_t *);
411 extern int rtable4hash(nfs4_sharedfh_t *);
412
413 extern vnode_t *makenfs4node(nfs4_sharedfh_t *, nfs4_ga_res_t *, struct vfs *,
414 hrtime_t, cred_t *, vnode_t *, nfs4_fname_t *);
415 extern vnode_t *makenfs4node_by_fh(nfs4_sharedfh_t *, nfs4_sharedfh_t *,
416 nfs4_fname_t **, nfs4_ga_res_t *, mntinfo4_t *, cred_t *, hrtime_t);
417
418 extern nfs4_opinst_t *r4mkopenlist(struct mntinfo4 *);
419 extern void r4releopenlist(nfs4_opinst_t *);
420 extern int r4find_by_fsid(mntinfo4_t *, fattr4_fsid *);
421
422 /* Access cache calls */
423 extern nfs4_access_type_t nfs4_access_check(rnode4_t *, uint32_t, cred_t *);
424 extern void nfs4_access_cache(rnode4_t *rp, uint32_t, uint32_t, cred_t *);
425 extern int nfs4_access_purge_rp(rnode4_t *);
426
427 extern int nfs4_free_data_reclaim(rnode4_t *);
428 extern void nfs4_rnode_invalidate(struct vfs *);
429
430 extern int nfs4_directio(vnode_t *, int, cred_t *);
431
432 /* shadow vnode functions */
433 extern void sv_activate(vnode_t **, vnode_t *, nfs4_fname_t **, int);
434 extern vnode_t *sv_find(vnode_t *, vnode_t *, nfs4_fname_t **);
435 extern void sv_update_path(vnode_t *, char *, char *);
436 extern void sv_inactive(vnode_t *);
437 extern void sv_exchange(vnode_t **);
438 extern void sv_uninit(svnode_t *);
439 extern void nfs4_clear_open_streams(rnode4_t *);
440
441 /*
442 * Mark cached attributes as timed out
443 *
444 * The caller must not be holding the rnode r_statelock mutex.
445 */
446 #define PURGE_ATTRCACHE4_LOCKED(rp) \
447 rp->r_time_attr_inval = gethrtime(); \
448 rp->r_time_attr_saved = rp->r_time_attr_inval; \
449 rp->r_pathconf.pc4_xattr_valid = 0; \
450 rp->r_pathconf.pc4_cache_valid = 0;
451
452 #define PURGE_ATTRCACHE4(vp) { \
453 rnode4_t *rp = VTOR4(vp); \
454 mutex_enter(&rp->r_statelock); \
455 PURGE_ATTRCACHE4_LOCKED(rp); \
456 mutex_exit(&rp->r_statelock); \
457 }
458
459
460 extern void nfs4_async_readdir(vnode_t *, rddir4_cache *,
461 cred_t *, int (*)(vnode_t *, rddir4_cache *, cred_t *));
462 extern char *rnode4info(rnode4_t *rp);
463
464 extern int writerp4(rnode4_t *, caddr_t, int, struct uio *, int);
465 extern void nfs4_set_nonvattrs(rnode4_t *, struct nfs4attr_to_vattr *);
466 extern void nfs4delegabandon(rnode4_t *);
467 extern stateid4 nfs4_get_w_stateid(cred_t *, rnode4_t *, pid_t, mntinfo4_t *,
468 nfs_opnum4, nfs4_stateid_types_t *);
469 extern stateid4 nfs4_get_stateid(cred_t *, rnode4_t *, pid_t, mntinfo4_t *,
470 nfs_opnum4, nfs4_stateid_types_t *, bool_t);
471 extern nfsstat4 nfs4_find_or_create_lock_owner(pid_t, rnode4_t *, cred_t *,
472 nfs4_open_owner_t **, nfs4_open_stream_t **,
473 nfs4_lock_owner_t **);
474 extern cred_t *nfs4_get_otw_cred_by_osp(rnode4_t *, cred_t *,
475 nfs4_open_stream_t **, bool_t *, bool_t *);
476
477
478 /*
479 * Defines for the flag argument of nfs4delegreturn
480 */
481 #define NFS4_DR_FORCE 0x1 /* discard even if start_op fails */
482 #define NFS4_DR_PUSH 0x2 /* push modified data back to the server */
483 #define NFS4_DR_DISCARD 0x4 /* discard the delegation w/o delegreturn */
484 #define NFS4_DR_DID_OP 0x8 /* calling function did nfs4_start_op */
485 #define NFS4_DR_RECALL 0x10 /* delegreturn done in response to CB_RECALL */
486 #define NFS4_DR_REOPEN 0x20 /* perform file reopens, if applicable */
487
488 extern int nfs4delegreturn(rnode4_t *, int);
489 extern void nfs4_delegreturn_all(nfs4_server_t *);
490 extern void nfs4delegreturn_cleanup(rnode4_t *, nfs4_server_t *);
491 extern void nfs4_delegation_accept(rnode4_t *, open_claim_type4, OPEN4res *,
492 nfs4_ga_res_t *, cred_t *);
493
494 extern void nfs4_dlistclean(void);
495 extern void nfs4_deleg_discard(mntinfo4_t *, nfs4_server_t *);
496
497 extern void rddir4_cache_create(rnode4_t *);
498 extern void rddir4_cache_purge(rnode4_t *);
499 extern void rddir4_cache_destroy(rnode4_t *);
500 extern rddir4_cache *rddir4_cache_lookup(rnode4_t *, offset_t, int);
501 extern void rddir4_cache_rele(rnode4_t *, rddir4_cache *);
502
503 extern void r4_stub_mirrormount(rnode4_t *);
504 extern void r4_stub_referral(rnode4_t *);
505 extern void r4_stub_none(rnode4_t *);
506
507 #ifdef DEBUG
508 extern char *rddir4_cache_buf_alloc(size_t, int);
509 extern void rddir4_cache_buf_free(void *, size_t);
510 #endif
511
512
513
514 #endif /* _KERNEL */
515
516 #ifdef __cplusplus
517 }
518 #endif
519
520 #endif /* _NFS_RNODE4_H */