Current Path : /sys/amd64/compile/hs32/modules/usr/src/sys/modules/s3/@/amd64/compile/hs32/modules/usr/src/sys/modules/cbb/@/fs/nfs/ |
FreeBSD hs32.drive.ne.jp 9.1-RELEASE FreeBSD 9.1-RELEASE #1: Wed Jan 14 12:18:08 JST 2015 root@hs32.drive.ne.jp:/sys/amd64/compile/hs32 amd64 |
Current File : //sys/amd64/compile/hs32/modules/usr/src/sys/modules/s3/@/amd64/compile/hs32/modules/usr/src/sys/modules/cbb/@/fs/nfs/nfsrvstate.h |
/*- * Copyright (c) 2009 Rick Macklem, University of Guelph * All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * * $FreeBSD: release/9.1.0/sys/fs/nfs/nfsrvstate.h 205941 2010-03-30 23:11:50Z rmacklem $ */ #ifndef _NFS_NFSRVSTATE_H_ #define _NFS_NFSRVSTATE_H_ /* * Definitions for NFS V4 server state handling. */ /* * List heads for nfsclient, nfsstate and nfslockfile. * (Some systems seem to like to dynamically size these things, but I * don't see any point in doing so for these ones.) */ LIST_HEAD(nfsclienthashhead, nfsclient); LIST_HEAD(nfsstatehead, nfsstate); LIST_HEAD(nfslockhead, nfslock); LIST_HEAD(nfslockhashhead, nfslockfile); /* * List head for nfsusrgrp. */ LIST_HEAD(nfsuserhashhead, nfsusrgrp); TAILQ_HEAD(nfsuserlruhead, nfsusrgrp); #define NFSCLIENTHASH(id) \ (&nfsclienthash[(id).lval[1] % NFSCLIENTHASHSIZE]) #define NFSSTATEHASH(clp, id) \ (&((clp)->lc_stateid[(id).other[2] % NFSSTATEHASHSIZE])) #define NFSUSERHASH(id) \ (&nfsuserhash[(id) % NFSUSERHASHSIZE]) #define NFSUSERNAMEHASH(p, l) \ (&nfsusernamehash[((l)>=4?(*(p)+*((p)+1)+*((p)+2)+*((p)+3)):*(p)) \ % NFSUSERHASHSIZE]) #define NFSGROUPHASH(id) \ (&nfsgrouphash[(id) % NFSGROUPHASHSIZE]) #define NFSGROUPNAMEHASH(p, l) \ (&nfsgroupnamehash[((l)>=4?(*(p)+*((p)+1)+*((p)+2)+*((p)+3)):*(p)) \ % NFSGROUPHASHSIZE]) /* * Client server structure for V4. It is doubly linked into two lists. * The first is a hash table based on the clientid and the second is a * list of all clients maintained in LRU order. * The actual size malloc'd is large enough to accomodate the id string. */ struct nfsclient { LIST_ENTRY(nfsclient) lc_hash; /* Clientid hash list */ struct nfsstatehead lc_stateid[NFSSTATEHASHSIZE]; /* stateid hash */ struct nfsstatehead lc_open; /* Open owner list */ struct nfsstatehead lc_deleg; /* Delegations */ struct nfsstatehead lc_olddeleg; /* and old delegations */ time_t lc_expiry; /* Expiry time (sec) */ time_t lc_delegtime; /* Old deleg expiry (sec) */ nfsquad_t lc_clientid; /* 64 bit clientid */ nfsquad_t lc_confirm; /* 64 bit confirm value */ u_int32_t lc_program; /* RPC Program # */ u_int32_t lc_callback; /* Callback id */ u_int32_t lc_stateindex; /* Current state index# */ u_int32_t lc_statemaxindex; /* Max state index# */ u_int32_t lc_cbref; /* Cnt of callbacks */ uid_t lc_uid; /* User credential */ gid_t lc_gid; u_int16_t lc_namelen; u_char *lc_name; struct nfssockreq lc_req; /* Callback info */ u_short lc_idlen; /* Length of id string */ u_int32_t lc_flags; /* LCL_ flag bits */ u_char lc_verf[NFSX_VERF]; /* client verifier */ u_char lc_id[1]; /* Malloc'd correct size */ }; #define CLOPS_CONFIRM 0x0001 #define CLOPS_RENEW 0x0002 #define CLOPS_RENEWOP 0x0004 /* * Nfs state structure. I couldn't resist overloading this one, since * it makes cleanup, etc. simpler. These structures are used in four ways: * - open_owner structures chained off of nfsclient * - open file structures chained off an open_owner structure * - lock_owner structures chained off an open file structure * - delegated file structures chained off of nfsclient and nfslockfile * - the ls_list field is used for the chain it is in * - the ls_head structure is used to chain off the sibling structure * (it is a union between an nfsstate and nfslock structure head) * If it is a lockowner stateid, nfslock structures hang off it. * For the open file and lockowner cases, it is in the hash table in * nfsclient for stateid. */ struct nfsstate { LIST_ENTRY(nfsstate) ls_hash; /* Hash list entry */ LIST_ENTRY(nfsstate) ls_list; /* List of opens/delegs */ LIST_ENTRY(nfsstate) ls_file; /* Opens/Delegs for a file */ union { struct nfsstatehead open; /* Opens list */ struct nfslockhead lock; /* Locks list */ } ls_head; nfsv4stateid_t ls_stateid; /* The state id */ u_int32_t ls_seq; /* seq id */ uid_t ls_uid; /* uid of locker */ u_int32_t ls_flags; /* Type of lock, etc. */ union { struct nfsstate *openowner; /* Open only */ u_int32_t opentolockseq; /* Lock call only */ u_int32_t noopens; /* Openowner only */ struct { u_quad_t filerev; /* Delegations only */ time_t expiry; time_t limit; u_int64_t compref; } deleg; } ls_un; struct nfslockfile *ls_lfp; /* Back pointer */ struct nfsrvcache *ls_op; /* Op cache reference */ struct nfsclient *ls_clp; /* Back pointer */ u_short ls_ownerlen; /* Length of ls_owner */ u_char ls_owner[1]; /* malloc'd the correct size */ }; #define ls_lock ls_head.lock #define ls_open ls_head.open #define ls_opentolockseq ls_un.opentolockseq #define ls_openowner ls_un.openowner #define ls_openstp ls_un.openowner #define ls_noopens ls_un.noopens #define ls_filerev ls_un.deleg.filerev #define ls_delegtime ls_un.deleg.expiry #define ls_delegtimelimit ls_un.deleg.limit #define ls_compref ls_un.deleg.compref /* * Nfs lock structure. * This structure is chained off of the nfsstate (the lockowner) and * nfslockfile (the file) structures, for the file and owner it * refers to. It holds flags and a byte range. * It also has back pointers to the associated lock_owner and lockfile. */ struct nfslock { LIST_ENTRY(nfslock) lo_lckowner; LIST_ENTRY(nfslock) lo_lckfile; struct nfsstate *lo_stp; struct nfslockfile *lo_lfp; u_int64_t lo_first; u_int64_t lo_end; u_int32_t lo_flags; }; /* * Structure used to return a conflicting lock. (Must be large * enough for the largest lock owner we can have.) */ struct nfslockconflict { nfsquad_t cl_clientid; u_int64_t cl_first; u_int64_t cl_end; u_int32_t cl_flags; u_short cl_ownerlen; u_char cl_owner[NFSV4_OPAQUELIMIT]; }; /* * This structure is used to keep track of local locks that might need * to be rolled back. */ struct nfsrollback { LIST_ENTRY(nfsrollback) rlck_list; uint64_t rlck_first; uint64_t rlck_end; int rlck_type; }; /* * This structure refers to a file for which lock(s) and/or open(s) exist. * Searched via hash table on file handle or found via the back pointer from an * open or lock owner. */ struct nfslockfile { LIST_HEAD(, nfsstate) lf_open; /* Open list */ LIST_HEAD(, nfsstate) lf_deleg; /* Delegation list */ LIST_HEAD(, nfslock) lf_lock; /* Lock list */ LIST_HEAD(, nfslock) lf_locallock; /* Local lock list */ LIST_HEAD(, nfsrollback) lf_rollback; /* Local lock rollback list */ LIST_ENTRY(nfslockfile) lf_hash; /* Hash list entry */ fhandle_t lf_fh; /* The file handle */ struct nfsv4lock lf_locallock_lck; /* serialize local locking */ int lf_usecount; /* Ref count for locking */ }; /* * This structure is malloc'd an chained off hash lists for user/group * names. */ struct nfsusrgrp { TAILQ_ENTRY(nfsusrgrp) lug_lru; /* LRU list */ LIST_ENTRY(nfsusrgrp) lug_numhash; /* Hash by id# */ LIST_ENTRY(nfsusrgrp) lug_namehash; /* and by name */ time_t lug_expiry; /* Expiry time in sec */ union { uid_t un_uid; /* id# */ gid_t un_gid; } lug_un; int lug_namelen; /* Name length */ u_char lug_name[1]; /* malloc'd correct length */ }; #define lug_uid lug_un.un_uid #define lug_gid lug_un.un_gid /* * These structures are used for the stable storage restart stuff. */ /* * Record at beginning of file. */ struct nfsf_rec { u_int32_t lease; /* Lease duration */ u_int32_t numboots; /* Number of boottimes */ }; #if defined(_KERNEL) || defined(KERNEL) void nfsrv_cleanclient(struct nfsclient *, NFSPROC_T *); void nfsrv_freedeleglist(struct nfsstatehead *); #endif #endif /* _NFS_NFSRVSTATE_H_ */