xref: /freebsd/sys/fs/nfs/nfsrvstate.h (revision af23369a6deaaeb612ab266eb88b8bb8d560c322)
1 /*-
2  * SPDX-License-Identifier: BSD-2-Clause-FreeBSD
3  *
4  * Copyright (c) 2009 Rick Macklem, University of Guelph
5  * All rights reserved.
6  *
7  * Redistribution and use in source and binary forms, with or without
8  * modification, are permitted provided that the following conditions
9  * are met:
10  * 1. Redistributions of source code must retain the above copyright
11  *    notice, this list of conditions and the following disclaimer.
12  * 2. Redistributions in binary form must reproduce the above copyright
13  *    notice, this list of conditions and the following disclaimer in the
14  *    documentation and/or other materials provided with the distribution.
15  *
16  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
17  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
18  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
19  * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
20  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
21  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
22  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
23  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
24  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
25  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
26  * SUCH DAMAGE.
27  *
28  * $FreeBSD$
29  */
30 
31 #ifndef _NFS_NFSRVSTATE_H_
32 #define	_NFS_NFSRVSTATE_H_
33 
34 #if defined(_KERNEL) || defined(KERNEL)
35 /*
36  * Definitions for NFS V4 server state handling.
37  */
38 
39 /*
40  * List heads for nfsclient, nfsstate and nfslockfile.
41  * (Some systems seem to like to dynamically size these things, but I
42  *  don't see any point in doing so for these ones.)
43  */
44 LIST_HEAD(nfsclienthashhead, nfsclient);
45 LIST_HEAD(nfsstatehead, nfsstate);
46 LIST_HEAD(nfslockhead, nfslock);
47 LIST_HEAD(nfslockhashhead, nfslockfile);
48 LIST_HEAD(nfssessionhead, nfsdsession);
49 LIST_HEAD(nfssessionhashhead, nfsdsession);
50 TAILQ_HEAD(nfslayouthead, nfslayout);
51 SLIST_HEAD(nfsdsdirhead, nfsdsdir);
52 TAILQ_HEAD(nfsdevicehead, nfsdevice);
53 LIST_HEAD(nfsdontlisthead, nfsdontlist);
54 
55 /*
56  * List head for nfsusrgrp.
57  */
58 TAILQ_HEAD(nfsuserhashhead, nfsusrgrp);
59 
60 #define	NFSCLIENTHASH(id)						\
61 	(&NFSD_VNET(nfsclienthash)[(id).lval[1] % nfsrv_clienthashsize])
62 #define	NFSSTATEHASH(clp, id)						\
63 	(&((clp)->lc_stateid[(id).other[2] % nfsrv_statehashsize]))
64 #define	NFSUSERHASH(id)							\
65 	(&NFSD_VNET(nfsuserhash)[(id) % nfsrv_lughashsize])
66 #define	NFSUSERNAMEHASH(p, l)						\
67 	(&NFSD_VNET(nfsusernamehash)[((l)>=4?(*(p)+*((p)+1)+*((p)+2)+*((p)+3)):*(p)) \
68 		% nfsrv_lughashsize])
69 #define	NFSGROUPHASH(id)						\
70 	(&NFSD_VNET(nfsgrouphash)[(id) % nfsrv_lughashsize])
71 #define	NFSGROUPNAMEHASH(p, l)						\
72 	(&NFSD_VNET(nfsgroupnamehash)[((l)>=4?(*(p)+*((p)+1)+*((p)+2)+*((p)+3)):*(p)) \
73 		% nfsrv_lughashsize])
74 
75 struct nfssessionhash {
76 	struct mtx			mtx;
77 	struct nfssessionhashhead	list;
78 };
79 #define	NFSSESSIONHASH(f) 						\
80 	(&NFSD_VNET(nfssessionhash)[nfsrv_hashsessionid(f) %		\
81 	 nfsrv_sessionhashsize])
82 
83 struct nfslayouthash {
84 	struct mtx		mtx;
85 	struct nfslayouthead	list;
86 };
87 #define	NFSLAYOUTHASH(f) 						\
88 	(&nfslayouthash[nfsrv_hashfh(f) % nfsrv_layouthashsize])
89 
90 /*
91  * Client server structure for V4. It is doubly linked into two lists.
92  * The first is a hash table based on the clientid and the second is a
93  * list of all clients maintained in LRU order.
94  * The actual size malloc'd is large enough to accommodate the id string.
95  */
96 struct nfsclient {
97 	LIST_ENTRY(nfsclient) lc_hash;		/* Clientid hash list */
98 	struct nfsstatehead *lc_stateid;	/* Stateid hash */
99 	struct nfsstatehead lc_open;		/* Open owner list */
100 	struct nfsstatehead lc_deleg;		/* Delegations */
101 	struct nfsstatehead lc_olddeleg;	/* and old delegations */
102 	struct nfssessionhead lc_session;	/* List of NFSv4.1 sessions */
103 	uint64_t	lc_prevsess;		/* CreateSession cache */
104 	time_t		lc_expiry;		/* Expiry time (sec) */
105 	time_t		lc_delegtime;		/* Old deleg expiry (sec) */
106 	nfsquad_t	lc_clientid;		/* 64 bit clientid */
107 	nfsquad_t	lc_confirm;		/* 64 bit confirm value */
108 	u_int32_t	lc_program;		/* RPC Program # */
109 	u_int32_t	lc_callback;		/* Callback id */
110 	u_int32_t	lc_stateindex;		/* Current state index# */
111 	u_int32_t	lc_statemaxindex;	/* Max state index# */
112 	u_int32_t	lc_cbref;		/* Cnt of callbacks */
113 	uid_t		lc_uid;			/* User credential */
114 	gid_t		lc_gid;
115 	u_int16_t	lc_idlen;		/* Client ID and len */
116 	u_int16_t	lc_namelen;		/* plus GSS principal and len */
117 	u_char		*lc_name;
118 	struct nfssockreq lc_req;		/* Callback info */
119 	u_int32_t	lc_flags;		/* LCL_ flag bits */
120 	u_char		lc_verf[NFSX_VERF];	 /* client verifier */
121 	u_char		lc_id[1];		/* Malloc'd correct size */
122 };
123 
124 #define	CLOPS_CONFIRM		0x0001
125 #define	CLOPS_RENEW		0x0002
126 #define	CLOPS_RENEWOP		0x0004
127 
128 /*
129  * Structure for NFSv4.1 Layouts.
130  * Malloc'd to correct size for the lay_xdr.
131  */
132 struct nfslayout {
133 	TAILQ_ENTRY(nfslayout)	lay_list;
134 	nfsv4stateid_t		lay_stateid;
135 	nfsquad_t		lay_clientid;
136 	fhandle_t		lay_fh;
137 	char			lay_deviceid[NFSX_V4DEVICEID];
138 	fsid_t			lay_fsid;
139 	uint32_t		lay_layoutlen;
140 	uint16_t		lay_mirrorcnt;
141 	uint16_t		lay_trycnt;
142 	uint16_t		lay_type;
143 	uint16_t		lay_flags;
144 	uint32_t		lay_xdr[0];
145 };
146 
147 /* Flags for lay_flags. */
148 #define	NFSLAY_READ	0x0001
149 #define	NFSLAY_RW	0x0002
150 #define	NFSLAY_RECALL	0x0004
151 #define	NFSLAY_RETURNED	0x0008
152 #define	NFSLAY_CALLB	0x0010
153 #define	NFSLAY_NOSPC	0x0020
154 
155 /*
156  * Structure for an NFSv4.1 session.
157  * Locking rules for this structure.
158  * To add/delete one of these structures from the lists, you must lock
159  * both: NFSLOCKSTATE() and NFSLOCKSESSION(session hashhead) in that order.
160  * To traverse the lists looking for one of these, you must hold one
161  * of these two locks.
162  * The exception is if the thread holds the exclusive root sleep lock.
163  * In this case, all other nfsd threads are blocked, so locking the
164  * mutexes isn't required.
165  * When manipulating sess_refcnt, NFSLOCKSTATE() must be locked.
166  * When manipulating the fields withinsess_cbsess except nfsess_xprt,
167  * sess_cbsess.nfsess_mtx must be locked.
168  * When manipulating sess_slots and sess_cbsess.nfsess_xprt,
169  * NFSLOCKSESSION(session hashhead) must be locked.
170  */
171 struct nfsdsession {
172 	uint64_t		sess_refcnt;	/* Reference count. */
173 	LIST_ENTRY(nfsdsession)	sess_hash;	/* Hash list of sessions. */
174 	LIST_ENTRY(nfsdsession)	sess_list;	/* List of client sessions. */
175 	struct nfsslot		sess_slots[NFSV4_SLOTS];
176 	struct nfsclient	*sess_clp;	/* Associated clientid. */
177 	uint32_t		sess_crflags;
178 	uint32_t		sess_cbprogram;
179 	uint32_t		sess_maxreq;
180 	uint32_t		sess_maxresp;
181 	uint32_t		sess_maxrespcached;
182 	uint32_t		sess_maxops;
183 	uint32_t		sess_maxslots;
184 	uint32_t		sess_cbmaxreq;
185 	uint32_t		sess_cbmaxresp;
186 	uint32_t		sess_cbmaxrespcached;
187 	uint32_t		sess_cbmaxops;
188 	uint8_t			sess_sessionid[NFSX_V4SESSIONID];
189 	struct nfsclsession	sess_cbsess;	/* Callback session. */
190 };
191 
192 /*
193  * Nfs state structure. I couldn't resist overloading this one, since
194  * it makes cleanup, etc. simpler. These structures are used in four ways:
195  * - open_owner structures chained off of nfsclient
196  * - open file structures chained off an open_owner structure
197  * - lock_owner structures chained off an open file structure
198  * - delegated file structures chained off of nfsclient and nfslockfile
199  * - the ls_list field is used for the chain it is in
200  * - the ls_head structure is used to chain off the sibling structure
201  *   (it is a union between an nfsstate and nfslock structure head)
202  *    If it is a lockowner stateid, nfslock structures hang off it.
203  * For the open file and lockowner cases, it is in the hash table in
204  * nfsclient for stateid.
205  */
206 struct nfsstate {
207 	LIST_ENTRY(nfsstate)	ls_hash;	/* Hash list entry */
208 	LIST_ENTRY(nfsstate)	ls_list;	/* List of opens/delegs */
209 	LIST_ENTRY(nfsstate)	ls_file;	/* Opens/Delegs for a file */
210 	union {
211 		struct nfsstatehead	open; /* Opens list */
212 		struct nfslockhead	lock; /* Locks list */
213 	} ls_head;
214 	nfsv4stateid_t		ls_stateid;	/* The state id */
215 	u_int32_t		ls_seq;		/* seq id */
216 	uid_t			ls_uid;		/* uid of locker */
217 	u_int32_t		ls_flags;	/* Type of lock, etc. */
218 	union {
219 		struct nfsstate	*openowner;	/* Open only */
220 		u_int32_t	opentolockseq;	/* Lock call only */
221 		u_int32_t	noopens;	/* Openowner only */
222 		struct {
223 			u_quad_t	filerev; /* Delegations only */
224 			time_t		expiry;
225 			time_t		limit;
226 			u_int64_t	compref;
227 			time_t		last;
228 		} deleg;
229 	} ls_un;
230 	struct nfslockfile	*ls_lfp;	/* Back pointer */
231 	struct nfsrvcache	*ls_op;		/* Op cache reference */
232 	struct nfsclient	*ls_clp;	/* Back pointer */
233 	u_short			ls_ownerlen;	/* Length of ls_owner */
234 	u_char			ls_owner[1];	/* malloc'd the correct size */
235 };
236 #define	ls_lock			ls_head.lock
237 #define	ls_open			ls_head.open
238 #define	ls_opentolockseq	ls_un.opentolockseq
239 #define	ls_openowner		ls_un.openowner
240 #define	ls_openstp		ls_un.openowner
241 #define	ls_noopens		ls_un.noopens
242 #define	ls_filerev		ls_un.deleg.filerev
243 #define	ls_delegtime		ls_un.deleg.expiry
244 #define	ls_delegtimelimit	ls_un.deleg.limit
245 #define	ls_compref		ls_un.deleg.compref
246 #define	ls_lastrecall		ls_un.deleg.last
247 
248 /*
249  * Nfs lock structure.
250  * This structure is chained off of the nfsstate (the lockowner) and
251  * nfslockfile (the file) structures, for the file and owner it
252  * refers to. It holds flags and a byte range.
253  * It also has back pointers to the associated lock_owner and lockfile.
254  */
255 struct nfslock {
256 	LIST_ENTRY(nfslock)	lo_lckowner;
257 	LIST_ENTRY(nfslock)	lo_lckfile;
258 	struct nfsstate		*lo_stp;
259 	struct nfslockfile	*lo_lfp;
260 	u_int64_t		lo_first;
261 	u_int64_t		lo_end;
262 	u_int32_t		lo_flags;
263 };
264 
265 /*
266  * Structure used to return a conflicting lock. (Must be large
267  * enough for the largest lock owner we can have.)
268  */
269 struct nfslockconflict {
270 	nfsquad_t		cl_clientid;
271 	u_int64_t		cl_first;
272 	u_int64_t		cl_end;
273 	u_int32_t		cl_flags;
274 	u_short			cl_ownerlen;
275 	u_char			cl_owner[NFSV4_OPAQUELIMIT];
276 };
277 
278 /*
279  * This structure is used to keep track of local locks that might need
280  * to be rolled back.
281  */
282 struct nfsrollback {
283 	LIST_ENTRY(nfsrollback)	rlck_list;
284 	uint64_t		rlck_first;
285 	uint64_t		rlck_end;
286 	int			rlck_type;
287 };
288 
289 /*
290  * This structure refers to a file for which lock(s) and/or open(s) exist.
291  * Searched via hash table on file handle or found via the back pointer from an
292  * open or lock owner.
293  */
294 struct nfslockfile {
295 	LIST_HEAD(, nfsstate)	lf_open;	/* Open list */
296 	LIST_HEAD(, nfsstate)	lf_deleg;	/* Delegation list */
297 	LIST_HEAD(, nfslock)	lf_lock;	/* Lock list */
298 	LIST_HEAD(, nfslock)	lf_locallock;	/* Local lock list */
299 	LIST_HEAD(, nfsrollback) lf_rollback;	/* Local lock rollback list */
300 	LIST_ENTRY(nfslockfile)	lf_hash;	/* Hash list entry */
301 	fhandle_t		lf_fh;		/* The file handle */
302 	struct nfsv4lock	lf_locallock_lck; /* serialize local locking */
303 	int			lf_usecount;	/* Ref count for locking */
304 };
305 
306 /*
307  * This structure is malloc'd an chained off hash lists for user/group
308  * names.
309  */
310 struct nfsusrgrp {
311 	TAILQ_ENTRY(nfsusrgrp)	lug_numhash;	/* Hash by id# */
312 	TAILQ_ENTRY(nfsusrgrp)	lug_namehash;	/* and by name */
313 	time_t			lug_expiry;	/* Expiry time in sec */
314 	union {
315 		uid_t		un_uid;		/* id# */
316 		gid_t		un_gid;
317 	} lug_un;
318 	struct ucred		*lug_cred;	/* Cred. with groups list */
319 	int			lug_namelen;	/* Name length */
320 	u_char			lug_name[1];	/* malloc'd correct length */
321 };
322 #define	lug_uid		lug_un.un_uid
323 #define	lug_gid		lug_un.un_gid
324 
325 /*
326  * These structures are used for the stable storage restart stuff.
327  */
328 /*
329  * Record at beginning of file.
330  */
331 struct nfsf_rec {
332 	u_int32_t	lease;			/* Lease duration */
333 	u_int32_t	numboots;		/* Number of boottimes */
334 };
335 
336 void nfsrv_cleanclient(struct nfsclient *, NFSPROC_T *);
337 void nfsrv_freedeleglist(struct nfsstatehead *);
338 
339 /*
340  * This structure is used to create the list of device info entries for
341  * a GetDeviceInfo operation and stores the DS server info.
342  * The nfsdev_addrandhost field has the fully qualified host domain name
343  * followed by the network address in XDR.
344  * It is allocated with nfsrv_dsdirsize nfsdev_dsdir[] entries.
345  */
346 struct nfsdevice {
347 	TAILQ_ENTRY(nfsdevice)	nfsdev_list;
348 	vnode_t			nfsdev_dvp;
349 	struct nfsmount		*nfsdev_nmp;
350 	char			nfsdev_deviceid[NFSX_V4DEVICEID];
351 	uint16_t		nfsdev_hostnamelen;
352 	uint16_t		nfsdev_fileaddrlen;
353 	uint16_t		nfsdev_flexaddrlen;
354 	uint16_t		nfsdev_mdsisset;
355 	char			*nfsdev_fileaddr;
356 	char			*nfsdev_flexaddr;
357 	char			*nfsdev_host;
358 	fsid_t			nfsdev_mdsfsid;
359 	uint32_t		nfsdev_nextdir;
360 	bool			nfsdev_nospc;
361 	vnode_t			nfsdev_dsdir[0];
362 };
363 
364 /*
365  * This structure holds the va_size, va_filerev, va_atime, va_mtime and
366  * va_bytes for the DS file and is stored in the metadata file's extended
367  * attribute pnfsd.dsattr.
368  * opnfsdsattr was missing the va_bytes field and, as such, it was updated.
369  */
370 struct opnfsdsattr {
371 	uint64_t	dsa_filerev;
372 	uint64_t	dsa_size;
373 	struct timespec	dsa_atime;
374 	struct timespec	dsa_mtime;
375 };
376 
377 struct pnfsdsattr {
378 	uint64_t	dsa_filerev;
379 	uint64_t	dsa_size;
380 	struct timespec	dsa_atime;
381 	struct timespec	dsa_mtime;
382 	uint64_t	dsa_bytes;
383 };
384 
385 /*
386  * This structure is a list element for a list the pNFS server uses to
387  * mark that the recovery of a mirror file is in progress.
388  */
389 struct nfsdontlist {
390 	LIST_ENTRY(nfsdontlist)	nfsmr_list;
391 	uint32_t		nfsmr_flags;
392 	fhandle_t		nfsmr_fh;
393 };
394 
395 /* nfsmr_flags bits. */
396 #define	NFSMR_DONTLAYOUT	0x00000001
397 
398 #endif	/* defined(_KERNEL) || defined(KERNEL) */
399 
400 /*
401  * This structure holds the information about the DS file and is stored
402  * in the metadata file's extended attribute called pnfsd.dsfile.
403  */
404 #define	PNFS_FILENAME_LEN	(2 * sizeof(fhandle_t))
405 struct pnfsdsfile {
406 	fhandle_t	dsf_fh;
407 	uint32_t	dsf_dir;
408 	union {
409 		struct sockaddr_in	sin;
410 		struct sockaddr_in6	sin6;
411 	} dsf_nam;
412 	char		dsf_filename[PNFS_FILENAME_LEN + 1];
413 };
414 #define	dsf_sin		dsf_nam.sin
415 #define	dsf_sin6	dsf_nam.sin6
416 
417 #endif	/* _NFS_NFSRVSTATE_H_ */
418