2 * Copyright 2000, International Business Machines Corporation and others.
5 * This software has been released under the terms of the IBM Public
6 * License. For details, see the LICENSE file in the top-level source
7 * directory or online at http://www.openafs.org/dl/license10.html
13 * AFS system call opcodes
15 #include "afs/afs_args.h"
16 #include "afs/afs_consts.h"
18 /* jhash.h is a standalone header and is fine to pull into kernel code. */
19 #include <opr/jhash.h>
22 * afs_fsfragsize cannot be less than 1023, or some cache-tracking
23 * calculations will be incorrect (since we track cache usage in kb).
24 * Some filesystems have fundamental blocksizes less than 1k, and
25 * normally we would tune afs_fsfragsize to be fragsize-1, but we must
26 * make sure to check that afs_fsfragsize does not go below this value.
28 #define AFS_MIN_FRAGSIZE 1023
30 /* Upper bound on number of iovecs out uio routines will deal with. */
31 #define AFS_MAXIOVCNT 16
33 enum afs_shutdown_state {
38 extern enum afs_shutdown_state afs_shuttingdown;
41 * The string that we use for the "device" we mount /afs on. That is, we're not
42 * mounted on a drive like /dev/sda, but instead mounted on "afs".
44 #define AFS_MOUNT_STR "afs"
46 # define MOUNT_AFS AFS_MOUNT_STR
50 * Macros to uniquely identify the AFS vfs struct
52 #define AFS_VFSMAGIC 0x1234
55 * Define AFS_VFSFSID, an id number corresponding to the AFS filesystem. On
56 * Irix, this is stored in a global var (afs_fstype); everywhere else, we have
57 * a constant. If we have a platform-defined value for this, use it (AFS_FSNO);
58 * otherwise, we just make up an arbitrary value here (99).
61 # define AFS_VFSFSID afs_fstype
62 #elif defined(AFS_FSNO)
63 # define AFS_VFSFSID AFS_FSNO
65 # define AFS_VFSFSID 99
68 /* use this value for reporting total space, free space, etc.
69 * fake a high number to satisfy programs that use the statfs call to make sure
70 * that there's enough space in the device partition before storing something
71 * there. keep it just under 2^31, to try and be safe about not breaking
72 * stuff that looks at the amount of free space. */
73 #define AFS_VFS_FAKEFREE (2147483647)
75 /* Moved from VNOPS/afs_vnop_flocks so can be used in prototypes */
76 #if defined(AFS_HPUX102_ENV)
77 #define AFS_FLOCK k_flock
79 #if defined(AFS_SUN5_ENV) || (defined(AFS_LINUX_ENV) && !(defined(AFS_LINUX_ENV) && defined(AFS_LINUX_64BIT_KERNEL)))
80 #define AFS_FLOCK flock64
82 #define AFS_FLOCK flock
83 #endif /* AFS_SUN65_ENV */
84 #endif /* AFS_HPUX102_ENV */
86 /* The following are various levels of afs debugging */
87 #define AFSDEB_GENERAL 1 /* Standard debugging */
88 #define AFSDEB_NETWORK 2 /* low level afs networking */
89 #define AFSDEB_RX 4 /* RX debugging */
90 #define AFSDEB_VNLAYER 8 /* interface layer to AFS (aixops, gfsops, etc) */
92 /* generic undefined vice id */
95 /* The basic defines for the Andrew file system
96 better keep things powers of two so "& (foo-1)" hack works for masking bits */
97 #define NBRS 15 /* max number of queued daemon requests */
98 #define NUSERS 16 /* hash table size for unixuser table */
99 #define NSERVERS 16 /* hash table size for server table */
100 #define NVOLS 64 /* hash table size for volume table */
101 #define NFENTRIES 256 /* hash table size for disk volume table */
102 #define VCSIZEBITS 16 /* log of stat cache hash table size */
103 #define VCSIZE (opr_jhash_size(VCSIZEBITS))
104 #define CBRSIZE 512 /* call back returns hash table size */
105 #define PIGGYSIZE 1350 /* max piggyback size */
106 #define MAXVOLS 128 /* max vols we can store */
107 #define MAXSYSNAME 128 /* max sysname (i.e. @sys) size */
108 #define MAXNUMSYSNAMES 32 /* max that current constants allow */
109 #define MAXROOTVOLNAMELEN 64 /* max length of root volume name */
110 #define NOTOKTIMEOUT (2*3600) /* time after which to timeout conns sans tokens */
111 #define NOPAG 0xffffffff
115 #define AFS_MAXCBRSCALL 32 /* max to return in a given call (must be <= AFSCBMAX) */
116 #define VCACHE_FREE 5
117 #define VCACHE_DYNAMIC_STRESSED 10000
118 #define VCACHE_STRESS_LOGINTERVAL (60*60*4) /* 4 hours */
119 #define AFS_NRXPACKETS 80
120 #define AFS_RXDEADTIME 50
121 #define AFS_HARDDEADTIME 120
122 #define AFS_IDLEDEADTIME 1200
123 #define AFS_IDLEDEADTIME_REP 180 /* more than fs's cb dead time */
124 #define AFS_BLKBITS 12
125 #define AFS_BLKSIZE (1 << AFS_BLKBITS)
127 extern afs_int32 afs_rx_deadtime;
128 extern afs_int32 afs_rx_harddead;
129 extern afs_int32 afs_rx_idledead;
130 extern afs_int32 afs_rx_idledead_rep;
132 struct sysname_info {
135 signed char index, allocked;
138 /* flags to use with AFSOP_CACHEINIT */
139 #define AFSCALL_INIT_MEMCACHE 0x1 /* use a memory-based cache */
141 /* below here used only for kernel procedures */
143 /* Store synchrony flags - SYNC means that data should be forced to server's
144 * disk immediately upon completion. */
147 #define AFS_VMSYNC_INVAL 2 /* sync and invalidate pages */
148 #define AFS_LASTSTORE 4
149 #define AFS_VMSYNC 8 /* sync pages but do not invalidate */
150 #define AFS_NOVMSYNC 16 /* force skipping syncing vm pages; just write afs dcache data */
152 /* background request structure */
155 #define BOP_NOOP 0 /* leave 0 unused */
156 #define BOP_FETCH 1 /* parm1 is chunk to get */
157 #define BOP_STORE 2 /* parm1 is chunk to store */
158 #define BOP_PATH 3 /* parm1 is path, parm2 is chunk to fetch */
159 #define BOP_FETCH_NOCACHE 4 /* parms are: vnode ptr, offset, segment ptr, addr, cred ptr */
160 #ifdef AFS_DARWIN_ENV
161 #define BOP_MOVE 5 /* ptr1 afs_uspc_param ptr2 sname ptr3 dname */
163 #define BOP_PARTIAL_STORE 6 /* parm1 is chunk to store */
164 #define BOP_INVALIDATE_SEGMENTS 7 /* no parms: just uses the 'bp->vc' vcache */
166 #define B_DONTWAIT 1 /* On failure return; don't wait */
168 /* protocol is: refCount is incremented by user to take block out of free pool.
169 Next, BSTARTED is set when daemon finds request. This prevents
170 other daemons from picking up the same request. Finally, when
171 request is done, refCount is zeroed. BDONE and BWAIT are used by
172 dudes waiting for operation to proceed to a certain point before returning.
174 #define BSTARTED 1 /* request picked up by a daemon */
175 #define BUVALID 2 /* code is valid (store) */
176 #define BUWAIT 4 /* someone is waiting for BUVALID */
178 struct vcache *vc; /* vnode to use, with vrefcount bumped */
179 afs_ucred_t *cred; /* credentials to use for operation */
180 afs_size_t size_parm[BPARMS]; /* random parameters */
181 void *ptr_parm[BPARMS]; /* pointer parameters */
182 afs_int32 code_raw; /* return code from AFS routines */
183 afs_int32 code_checkcode; /* the afs_CheckCode-translated code */
184 short refCount; /* use counter for this structure */
185 char opcode; /* what to do (store, fetch, etc) */
186 char flags; /* free, etc */
187 afs_int32 ts; /* counter "timestamp" */
195 afs_int32 AuthHandle;
196 char HandShakeKey[8];
198 afs_int32 BeginTimestamp;
199 afs_int32 EndTimestamp;
203 afs_int32 Cell; /* better sun packing if at end of structure */
207 /* Temporary struct to be passed between afs_fid and afs_vget */
210 afs_int32 CellAndUnique;
213 /* The actual number of bytes in the SmallFid, not the sizeof struct. */
214 #define SIZEOF_SMALLFID 10
219 * Circular queues, implemented with pointers. Structures may contain as many
220 * queues as required, which may be located at any point within the structure,
221 * providing the QEntry macro is used to translate between a queue pointer, and
222 * the address of its containing structure
230 #define QZero(e) ((e)->prev = (e)->next = NULL)
231 #define QInit(q) ((q)->prev = (q)->next = (q))
232 #define QAdd(q,e) ((e)->next = (q)->next, (e)->prev = (q), \
233 (q)->next->prev = (e), (q)->next = (e))
234 #define QRemove(e) ((e)->next->prev = (e)->prev, (e)->prev->next = (e)->next, (e)->prev = NULL, (e)->next = NULL)
235 #define QNext(e) ((e)->next)
236 #define QPrev(e) ((e)->prev)
237 #define QEmpty(q) ((q)->prev == (q))
238 /* this one takes q1 and sticks it on the end of q2 - that is, the other end, not the end
239 * that things are added onto. q1 shouldn't be empty, it's silly */
240 #define QCat(q1,q2) ((q2)->prev->next = (q1)->next, (q1)->next->prev=(q2)->prev, (q1)->prev->next=(q2), (q2)->prev=(q1)->prev, (q1)->prev=(q1)->next=(q1))
242 /* Given a pointer to an afs_q within a structure, go back to the address of
243 * the parent structure
246 #define QEntry(queue, structure, member) \
247 ((structure *)((char *)(queue)-(char *)(&((structure *)NULL)->member)))
249 /* And implement operations for individual lists in terms of the above macro */
251 #define QTOV(e) QEntry(e, struct vcache, vlruq)
252 #define QTOC(e) QEntry(e, struct cell, lruq)
253 #define QTOVH(e) QEntry(e, struct vcache, vhashq)
256 * List of free slot numbers
258 struct afs_slotlist {
260 struct afs_slotlist *next;
264 afs_int32 uid; /* user id making the request */
265 afs_int32 busyCount; /* how many busies we've seen so far */
266 afs_int32 flags; /* things like O_SYNC, O_NONBLOCK go here */
267 char initd; /* if non-zero, Error fields meaningful */
268 char accessError; /* flags for overriding error return code */
269 char volumeError; /* encountered a missing or busy volume */
270 char networkError; /* encountered network problems */
271 char permWriteError; /* fileserver returns permenent error. */
272 char tokenError; /* a token error other than expired. */
273 char idleError; /* the server idled too long */
274 char skipserver[AFS_MAXHOSTS];
275 afs_int32 lasterror[AFS_MAXHOSTS];
280 /* structure linked off of a server to keep track of queued returned
281 * callbacks. Sent asynchronously when we run a little low on free dudes.
284 struct afs_cbr **pprev;
285 struct afs_cbr *next;
287 struct afs_cbr **hash_pprev;
288 struct afs_cbr *hash_next;
294 /* On Linux, we have to be able to allocate the storage for this using
295 * kmalloc, as otherwise we may deadlock. So, it needs to be able to fit
298 # define AFS_NCBRS PAGE_SIZE/sizeof(struct afs_cbr)
300 # define AFS_NCBRS 1024 /* max # of call back return entries */
303 /* cellinfo file magic number */
304 #define AFS_CELLINFO_MAGIC 0xf32817cd
307 #define CNoSUID 0x02 /* disable suid bit for this cell */
308 #define CLinkedCell4 0x04 /* reserved for ADDCELL2 pioctl */
309 #define CNoAFSDB 0x08 /* never bother trying AFSDB */
310 #define CHasVolRef 0x10 /* volumes were referenced */
311 #define CLinkedCell 0x20 /* has a linked cell in lcellp */
312 #define CHush 0x40 /* don't display until referenced */
315 struct afs_q lruq; /* lru q next and prev */
316 char *cellName; /* char string name of cell */
317 afs_int32 cellIndex; /* sequence number */
318 afs_int32 cellNum; /* semi-permanent cell number */
319 struct server *cellHosts[AFS_MAXCELLHOSTS]; /* volume *location* hosts */
320 struct cell *lcellp; /* Associated linked cell */
321 u_short fsport; /* file server port */
322 u_short vlport; /* volume server port */
323 short states; /* state flags */
324 time_t timeout; /* data expire time, if non-zero */
325 struct cell_name *cnamep; /* pointer to our cell_name */
326 afs_rwlock_t lock; /* protects cell data */
327 unsigned char cellHandle[16]; /* deterministic handle for this cell */
331 struct cell_name *next;
338 struct cell_alias *next;
344 #define afs_PutCell(cellp, locktype)
346 /* the unixuser flag bit definitions */
347 #define UHasTokens 1 /* are the st and ct fields valid (ever set)? */
348 #define UTokensBad 2 /* are tokens bad? */
349 #define UPrimary 4 /* on iff primary identity */
350 #define UNeedsReset 8 /* needs afs_ResetAccessCache call done */
351 #define UPAGCounted 16 /* entry seen during PAG search (for stats) */
352 #define UNFSGetCreds 32 /* getting creds for NFS client */
353 /* A flag used by afs_GCPAGs to keep track of
354 * which entries in afs_users need to be deleted.
355 * The lifetime of its presence in the table is the
356 * lifetime of the afs_GCPAGs function.
358 #define TMP_UPAGNotReferenced 128
360 /* unixuser notify events */
361 #define UTokensObtained 1
362 #define UTokensDropped 2
364 /* values for afs_gcpags */
365 enum { AFS_GCPAGS_NOTCOMPILED = 0, AFS_GCPAGS_OK =
366 1, AFS_GCPAGS_USERDISABLED, AFS_GCPAGS_EPROC0, AFS_GCPAGS_EPROCN,
367 AFS_GCPAGS_EEQPID, AFS_GCPAGS_EINEXACT, AFS_GCPAGS_EPROCEND,
368 AFS_GCPAGS_EPROCWALK, AFS_GCPAGS_ECREDWALK, AFS_GCPAGS_EPIDCHECK,
369 AFS_GCPAGS_ENICECHECK
372 extern afs_int32 afs_gcpags;
373 extern afs_int32 afs_gcpags_procsize;
374 extern afs_int32 afs_bkvolpref;
375 extern char afs_cachebasedir[1024];
376 extern afs_int32 afs_numcachefiles;
377 extern afs_int32 afs_numfilesperdir;
382 struct ClearToken clearToken;
386 struct rxkadToken rxkad;
390 struct tokenJar *next;
392 union tokenUnion content;
396 struct unixuser *next; /* next hash pointer */
397 afs_int32 uid; /* search based on uid and cell */
399 afs_int32 viceId; /* Optional viced ID corresponding to current tokens */
400 short refCount; /* reference count for allocation */
401 char states; /* flag info */
402 afs_int32 tokenTime; /* last time tokens were set, used for timing out conn data */
403 struct tokenJar *tokens;
404 struct afs_exporter *exporter; /* more info about the exporter for the remote user */
405 void *cellinfo; /* pointer to cell info (PAG manager only) */
409 #define CVEC_LEN 3 /* per-user connection pool */
411 struct sa_conn_vector;
412 typedef struct sa_conn_vector * p_sa_conn_vector; /* forward decl */
414 #define CONN_REPLICATED 0x1
419 char forceConnectFS; /* Should we try again with these tokens? */
420 struct rx_connection *id; /* RPC connid */
421 struct sa_conn_vector *parent; /* the con_vector which contains us */
424 /* An sa_conn_vector replaces the erstwhile list of conn
425 structures maintained by the cache manager. The sa_conn_vector
426 contains a C array of connections which, if non-zero, represent
427 connections to AFS servers.
430 struct sa_conn_vector {
431 /* linked-list machinery */
432 struct sa_conn_vector *next;
434 /* AFS conn-identifying info */
435 struct unixuser *user; /* user validated with respect to. */
436 struct srvAddr *srvr; /* server associated with this conn */
437 short refCount; /* reference count for allocation */
438 unsigned short port; /* port associated with this connection */
441 /* next connection to return when all in cvec are fully utilized */
444 /* connections vector */
445 struct afs_conn cvec[CVEC_LEN];
450 /* Fid comparison routines */
451 #define FidCmp(a,b) ((a)->Fid.Unique != (b)->Fid.Unique \
452 || (a)->Fid.Vnode != (b)->Fid.Vnode \
453 || (a)->Fid.Volume != (b)->Fid.Volume \
454 || (a)->Cell != (b)->Cell)
456 #define FidMatches(afid,tvc) ((tvc)->f.fid.Fid.Vnode == (afid)->Fid.Vnode && \
457 (tvc)->f.fid.Fid.Volume == (afid)->Fid.Volume && \
458 (tvc)->f.fid.Cell == (afid)->Cell && \
459 ( (tvc)->f.fid.Fid.Unique == (afid)->Fid.Unique || \
460 (!(afid)->Fid.Unique && ((tvc)->f.states & CUnique))))
464 #define SRVADDR_ISDOWN 0x20 /* same as SRVR_ISDOWN */
465 #define SRVADDR_NOUSE 0x40 /* Don't use this srvAddr */
468 struct srvAddr *next_bkt; /* next item in hash bucket */
469 struct srvAddr *next_sa; /* another interface on same host */
470 struct server *server; /* back to parent */
471 struct sa_conn_vector *conns; /* All user connections to this server */
472 struct afs_conn *natping;
473 afs_int32 sa_ip; /* Host addr in network byte order */
474 u_short sa_iprank; /* indiv ip address priority */
475 u_short sa_portal; /* port addr in network byte order */
480 * Values used in the flags field of the server structure below.
482 * AFS_SERVER_FLAG_ACTIVATED Has the server ever had a user connection
483 * associated with it?
485 #define AFS_SERVER_FLAG_ACTIVATED 0x01
486 #define SNO_LHOSTS 0x04
487 #define SYES_LHOSTS 0x08
488 #define SVLSRV_UUID 0x10
489 #define SRVR_ISDOWN 0x20
490 #define SRVR_MULTIHOMED 0x40
491 #define SRVR_ISGONE 0x80
492 #define SNO_INLINEBULK 0x100
493 #define SNO_64BIT 0x200
494 #define SCAPS_KNOWN 0x400
496 #define SRV_CAPABILITIES(ts) \
497 { if ( !(ts->flags & SCAPS_KNOWN)) afs_GetCapabilities(ts); ts->capabilities; }
499 #define afs_serverSetNo64Bit(s) (((struct sa_conn_vector*)(s)->parent)->srvr->server->flags |= SNO_64BIT)
500 #define afs_serverHasNo64Bit(s) (((struct sa_conn_vector*)(s)->parent)->srvr->server->flags & SNO_64BIT)
506 afs_int32 addr_uniquifier;
510 struct srvAddr haddr;
513 #define sr_uuid _suid._srvUuid.suuid
514 #define sr_addr_uniquifier _suid._srvUuid.addr_uniquifier
515 #define sr_host _suid._srvId.haddr.ip
516 #define sr_portal _suid._srvId.haddr.portal
517 #define sr_rank _suid._srvId.haddr.ip_rank
518 #define sr_flags _suid._srvId.haddr.flags
519 #define sr_conns _suid._srvId.haddr.conns
520 struct server *next; /* Ptr to next server in hash chain */
521 struct cell *cell; /* Cell in which this host resides */
522 struct afs_cbr *cbrs; /* Return list of callbacks */
523 afs_int32 activationTime; /* Time when this record was first activated */
524 afs_int32 lastDowntimeStart; /* Time when last downtime incident began */
525 afs_int32 numDowntimeIncidents; /* # (completed) downtime incidents */
526 afs_int32 sumOfDowntimes; /* Total downtime experienced, in seconds */
527 struct srvAddr *addr;
528 afs_uint32 flags; /* Misc flags */
529 afs_int32 capabilities;
532 #define afs_PutServer(servp, locktype)
534 /* structs for some pioctls - these are (or should be)
542 struct sprefrequest_33 {
543 unsigned short offset;
544 unsigned short num_servers;
548 struct sprefrequest { /* new struct for 3.4 */
549 unsigned short offset;
550 unsigned short num_servers;
551 unsigned short flags;
556 unsigned short next_offset;
557 unsigned short num_servers;
558 struct spref servers[1]; /* we overrun this array intentionally... */
562 unsigned short flags;
563 unsigned short num_servers;
564 struct spref servers[1]; /* we overrun this array intentionally... */
566 /* struct for GAG pioctl
569 afs_uint32 showflags, logflags, logwritethruflag, spare[3];
570 unsigned char spare2[128];
574 #define logwritethruON 1
577 afs_int32 rx_initReceiveWindow, rx_maxReceiveWindow, rx_initSendWindow,
578 rx_maxSendWindow, rxi_nSendFrags, rxi_nRecvFrags, rxi_OrphanFragSize;
579 afs_int32 rx_maxReceiveSize, rx_MyMaxSendSize;
580 afs_uint32 spare[21];
583 /* struct for checkservers */
594 /* state bits for volume */
595 #define VRO 1 /* volume is readonly */
596 #define VRecheck 2 /* recheck volume info with server */
597 #define VBackup 4 /* is this a backup volume? */
598 #define VForeign 8 /* this is a non-afs volume */
599 #define VPartVisible 16 /* Volume's partition is visible on the client */
600 #define VHardMount 32 /* we are hard-mount waiting for the vol */
602 enum repstate { not_busy, end_not_busy = 6, rd_busy, rdwr_busy, offline };
605 /* One structure per volume, describing where the volume is located
606 * and where its mount points are. */
607 struct volume *next; /* Next volume in hash list. */
608 afs_int32 cell; /* the cell in which the volume resides */
609 afs_rwlock_t lock; /* the lock for this structure */
610 afs_int32 volume; /* This volume's ID number. */
611 char *name; /* This volume's name, or 0 if unknown */
612 struct server *serverHost[AFS_MAXHOSTS]; /* servers serving this volume */
613 enum repstate status[AFS_MAXHOSTS]; /* busy, offline, etc */
614 struct VenusFid dotdot; /* dir to access as .. */
615 struct VenusFid mtpoint; /* The mount point for this volume. */
616 afs_int32 rootVnode, rootUnique; /* Volume's root fid */
619 afs_int32 rwVol; /* For r/o vols, original read/write volume. */
620 afs_int32 setupTime; /* time volume was setup from vldb info */
621 afs_int32 vtix; /* volume table index */
622 afs_int32 copyDate; /* copyDate field, for tracking vol releases */
623 afs_int32 expireTime; /* for per-volume callbacks... */
624 short refCount; /* reference count for allocation */
625 char states; /* here for alignment reasons */
628 #define afs_PutVolume(av, locktype) ((av)->refCount--)
630 /* format of an entry in volume info file */
632 afs_int32 cell; /* cell for this entry */
633 afs_int32 volume; /* volume */
634 afs_int32 next; /* has index */
635 struct VenusFid dotdot; /* .. value */
636 struct VenusFid mtpoint; /* mt point's fid */
637 afs_int32 rootVnode, rootUnique; /* Volume's root fid */
641 struct SimpleLocks *next;
643 afs_int32 boff, eoff;
645 #if defined(AFS_AIX32_ENV) || defined(AFS_SUN5_ENV) || defined(AFS_SGI_ENV)
650 /* vcache state bits */
651 #define CStatd 0x00000001 /* has this file ever been stat'd? */
652 #define CBackup 0x00000002 /* file is on a backup volume */
653 #define CRO 0x00000004 /* is it on a read-only volume */
654 #define CMValid 0x00000008 /* is the mount point info valid? */
655 #define CCore 0x00000010 /* storing a core file, needed since we don't get an open */
656 #define CDirty 0x00000020 /* file has been modified since first open (... O_RDWR) */
657 #define CSafeStore 0x00000040 /* close must wait for store to finish (should be in fd) */
658 #define CMAPPED 0x00000080 /* Mapped files */
659 #define CNSHARE 0x00000100 /* support O_NSHARE semantics */
660 #define CLied 0x00000200
661 #define CTruth 0x00000400
663 #if defined(AFS_DARWIN80_ENV)
664 #define CDeadVnode 0x00000800
665 #elif defined(AFS_DARWIN_ENV)
666 #define CUBCinit 0x00000800
668 #define CWRITE_IGN 0x00000800 /* Next OS hack only */
671 #define CUnique 0x00001000 /* vc's uniquifier - latest unifiquier for fid */
672 #define CForeign 0x00002000 /* this is a non-afs vcache */
673 #define CReadDir 0x00004000 /* readdir in progress */
674 #define CUnlinked 0x00010000
675 #define CBulkStat 0x00020000 /* loaded by a bulk stat, and not ref'd since */
676 #define CUnlinkedDel 0x00040000
677 #define CVFlushed 0x00080000
678 #if defined(AFS_SGI_ENV)
679 #define CWritingUFS 0x00200000 /* to detect vm deadlock - used by sgi */
680 #elif defined(AFS_DARWIN80_ENV)
681 #define CEvent 0x00200000 /* to preclude deadlock when sending events */
683 #define CCreating 0x00400000 /* avoid needless store after open truncate */
684 #define CPageHog 0x00800000 /* AIX - dumping large cores is a page hog. */
685 #define CCorrupt 0x01000000 /* corrupt dir already reported */
686 #define CDCLock 0x02000000 /* Vnode lock held over call to GetDownD */
687 #define CBulkFetching 0x04000000 /* stats are being fetched by bulk stat */
688 #define CExtendedFile 0x08000000 /* extended file via ftruncate call. */
689 #define CVInit 0x10000000 /* being initialized */
690 #define CMetaDirty 0x20000000 /* vnode meta-data needs to be flushed */
691 #define CPartVisible 0x40000000 /* fileserver partition visible on client */
693 /* vcache vstate bits */
694 #define VRevokeWait 0x1
695 #define VPageCleaning 0x2 /* Solaris - Cache Trunc Daemon sez keep out */
697 /* Dirty disconnected vcache flags. */
698 #define VDisconSetTime 0x00000001 /* set time. */
699 #define VDisconSetMode 0x00000002 /* set mode. */
700 /* XXX: to be continued ? */
701 #define VDisconTrunc 0x00000020 /* truncate file. */
702 #define VDisconSetAttrMask 0x0000003F /* Masks for setattr ops. */
703 #define VDisconWriteClose 0x00000400 /* Write op on file close. */
704 #define VDisconWriteFlush 0x00000800 /* Write op on normal fsync/flush. */
705 #define VDisconWriteOsiFlush 0x00001000 /* Write op on osi flush. */
707 #define VDisconRemove 0x00002000 /* Remove vnop. */
708 #define VDisconCreate 0x00004000 /* Create vnop. */
709 #define VDisconCreated 0x00008000 /* A file that was created during
710 this resync operation */
711 #define VDisconRename 0x00010000 /* Rename vnop. */
712 #define VDisconRenameSameDir 0x00020000 /* Rename in same dir. */
714 /*... to be continued ... */
716 /* vcache (file) cachingStates bits */
717 #define FCSDesireBypass 0x1 /* This file should bypass the cache */
718 #define FCSBypass 0x2 /* This file is currently NOT being cached */
719 #define FCSManuallySet 0x4 /* The bypass flags were set, or reset, manually (via pioctl)
720 and should not be overridden by the file's name */
722 /* Flag values used by the Transition routines */
723 #define TRANSChangeDesiredBit 0x1 /* The Transition routine should set or
724 * reset the FCSDesireBypass bit */
725 #define TRANSVcacheIsLocked 0x2 /* The Transition routine does not need to
726 * lock vcache (it's already locked) */
727 #define TRANSSetManualBit 0x4 /* The Transition routine should set FCSManuallySet so that
728 * filename checking does not override pioctl requests */
731 #if defined(AFS_XBSD_ENV) || defined(AFS_DARWIN_ENV)
732 #define vrefCount v->v_usecount
733 #elif defined(AFS_SUN511_ENV)
734 # define vrefCount v->v_count
736 #define vrefCount v.v_count
737 #endif /* AFS_XBSD_ENV */
739 #if defined(AFS_DARWIN80_ENV)
740 #define VREFCOUNT_GT(v, y) vnode_isinuse(AFSTOV(v), (y))
741 #elif defined(AFS_FBSD_ENV)
742 #define VREFCOUNT(v) (vrefcnt(AFSTOV(v)))
743 #define VREFCOUNT_GT(v, y) (vrefcnt(AFSTOV(v)) > y)
744 #elif defined(AFS_XBSD_ENV) || defined(AFS_DARWIN_ENV)
745 #define VREFCOUNT(v) ((v)->vrefCount)
746 #define VREFCOUNT_GT(v, y) (AFSTOV(v)->v_usecount > (y))
747 #elif defined(AFS_LINUX_ENV)
748 #define VREFCOUNT(v) atomic_read(&(AFSTOV(v)->v_count))
749 #define VREFCOUNT_GT(v, y) (VREFCOUNT(v)>y)
750 #define VREFCOUNT_SET(v, c) atomic_set(&(AFSTOV(v)->v_count), c)
751 #define VREFCOUNT_DEC(v) atomic_dec(&(AFSTOV(v)->v_count))
752 #define VREFCOUNT_INC(v) atomic_inc(&(AFSTOV(v)->v_count))
754 #define VREFCOUNT(v) ((v)->vrefCount)
755 #define VREFCOUNT_GT(v,y) ((v)->vrefCount > (y))
756 #define VREFCOUNT_SET(v, c) (v)->vrefCount = c
757 #define VREFCOUNT_DEC(v) (v)->vrefCount--
758 #define VREFCOUNT_INC(v) (v)->vrefCount++
759 #define d_unhash(d) list_empty(&(d)->d_hash)
760 #define dget_locked(d) dget(d)
763 #define AFS_MAXDV 0x7fffffff /* largest dataversion number */
764 #ifdef AFS_64BIT_CLIENT
765 #define AFS_NOTRUNC 0x7fffffffffffffffLL /* largest positive int64 number */
766 #else /* AFS_64BIT_CLIENT */
767 #define AFS_NOTRUNC 0x7fffffff /* largest dataversion number */
768 #endif /* AFS_64BIT_CLIENT */
770 extern afs_int32 vmPageHog; /* counter for # of vnodes which are page hogs. */
772 #if defined(AFS_DARWIN80_ENV)
773 #define VTOAFS(v) ((struct vcache *)vnode_fsnode((v)))
774 #define AFSTOV(vc) ((vc)->v)
775 #elif defined(AFS_NBSD40_ENV)
777 struct genfs_node gfsn;
778 struct vcache *afsvc;
780 #define VTOAFS(v) ((((struct nbvdata *)((v)->v_data)))->afsvc)
781 #define AFSTOV(vc) ((vc)->v)
782 #elif defined(AFS_XBSD_ENV) || defined(AFS_DARWIN_ENV) || defined(AFS_SUN511_ENV) || (defined(AFS_LINUX_ENV) && !defined(STRUCT_SUPER_OPERATIONS_HAS_ALLOC_INODE))
783 #define VTOAFS(v) ((struct vcache *)(v)->v_data)
784 #define AFSTOV(vc) ((vc)->v)
786 #define VTOAFS(V) ((struct vcache *)(V))
787 #define AFSTOV(V) (&(V)->v)
795 /* VCache elements which are kept on disk, and in the kernel */
800 afs_hyper_t DataVersion;
804 afs_uint16 Mode; /* XXXX Should be afs_int32 XXXX */
805 afs_uint16 LinkCount;
806 #ifdef AFS_DARWIN80_ENV
809 /* vnode type is in v.v_type */
812 struct afs_vnuniq parent;
814 /*! Truncate file to this position at the next store */
817 /*! System:AnyUser's access to this. */
823 /*! Disconnected flags for this vcache element. */
824 afs_uint32 ddirty_flags;
825 /*! Shadow vnode + unique keep the shadow dir location. */
826 struct afs_vnuniq shadow;
827 /*! The old parent FID for renamed vnodes */
828 struct afs_vnuniq oldParent;
831 /* Values for 'mvstat' in struct vcache */
832 #define AFS_MVSTAT_FILE (0x0) /* regular file or directory */
833 #define AFS_MVSTAT_MTPT (0x1) /* mountpoint */
834 #define AFS_MVSTAT_ROOT (0x2) /* volume root dir */
838 * This is for the multiPage field in struct vcache. Each one of these
839 * represents an outstanding getpage request that is larger than a single page.
840 * Recording these is necessary to prevent afs_GetOnePage from trying to evict
841 * a dcache entry that an earlier afs_GetOnePage call got in the same getpage
842 * request. See osi_VM_MultiPageConflict and afs_getpage.
844 struct multiPage_range {
846 offset_t off; /**< offset of getpage request */
847 u_int len; /**< length of getpage request */
851 /* INVARIANTs: (vlruq.next != NULL) == (vlruq.prev != NULL)
852 * nextfree => !vlruq.next && ! vlruq.prev
853 * !(avc->nextfree) && !avc->vlruq.next => (FreeVCList == avc->nextfree)
856 #if defined(AFS_XBSD_ENV) || defined(AFS_DARWIN_ENV) || defined(AFS_SUN511_ENV) || (defined(AFS_LINUX_ENV) && !defined(STRUCT_SUPER_OPERATIONS_HAS_ALLOC_INODE))
859 struct vnode v; /* Has reference count in v.v_count */
861 struct afs_q vlruq; /* lru q next and prev */
862 #if !defined(AFS_LINUX_ENV)
863 struct vcache *nextfree; /* next on free list (if free) */
865 struct vcache *hnext; /* Hash next */
866 struct afs_q vhashq; /* Hashed per-volume list */
867 /*! Queue of dirty vcaches. Lock with afs_disconDirtyLock */
869 /*! Queue of vcaches with shadow entries. Lock with afs_disconDirtyLock */
870 struct afs_q shadowq;
871 /*! Queue of vcaches with dirty metadata. Locked by afs_xvcdirty */
872 struct afs_q metadirty;
873 /*! Vcaches slot number in the disk backup. Protected by tvc->lock */
876 afs_rwlock_t lock; /* The lock on the vcache contents. */
877 #if defined(AFS_SUN5_ENV)
878 /* Lock used to protect the activeV, multipage, and vstates fields.
879 * Do not try to get the vcache lock when the vlock is held */
881 #endif /* defined(AFS_SUN5_ENV) */
882 #if defined(AFS_SUN5_ENV)
889 #if defined(AFS_AIX51_ENV)
897 int ownslock; /* pid of owner of excl lock, else 0 - defect 3083 */
899 #ifdef AFS_DARWIN80_ENV
901 #elif defined(AFS_DARWIN_ENV)
902 struct lock__bsd__ rwlock;
905 #if !defined(AFS_DFBSD_ENV) && !defined(AFS_NBSD_ENV)
911 char *silly_name; /* For sillyrenamed regular files, the silly
912 * name the file was renamed to. */
913 struct VenusFid *target_root; /* For mountpoints, the fid of the root dir
914 * in the target volume. */
915 struct VenusFid *parent; /* For root dir vcaches, the fid of the
919 char *linkData; /* Link data if a symlink. */
920 afs_hyper_t flushDV; /* data version last flushed from text */
921 afs_hyper_t mapDV; /* data version last flushed from map */
922 struct server *callback; /* The callback host, if any */
923 afs_uint32 cbExpires; /* time the callback expires */
924 struct afs_q callsort; /* queue in expiry order, sort of */
925 struct axscache *Access; /* a list of cached access bits */
926 afs_int32 last_looker; /* pag/uid from last lookup here */
927 #if defined(AFS_SUN5_ENV)
929 #endif /* defined(AFS_SUN5_ENV) */
930 struct SimpleLocks *slocks;
931 short opens; /* The numbers of opens, read or write, on this file. */
932 short execsOrWriters; /* The number of execs (if < 0) or writers (if > 0) of
934 short flockCount; /* count of flock readers, or -1 if writer */
935 char mvstat; /* see the AFS_MVSTAT_* constants */
937 char cachingStates; /* Caching policies for this file */
938 afs_uint32 cachingTransitions; /* # of times file has flopped between caching and not */
940 #if defined(AFS_LINUX_ENV)
941 off_t next_seq_offset; /* Next sequential offset (used by prefetch/readahead) */
942 #elif defined(AFS_SUN5_ENV) || defined(AFS_SGI_ENV)
943 off_t next_seq_blk_offset; /* accounted in blocks for Solaris & IRIX */
946 #if defined(AFS_SUN5_ENV)
947 afs_uint32 vstates; /* vstate bits */
948 #endif /* defined(AFS_SUN5_ENV) */
949 struct dcache *dchint;
950 struct dcache *dcreaddir; /* dcache for in-progress readdir */
951 unsigned int readdir_pid; /* pid of the thread in readdir */
952 #if defined(AFS_SGI_ENV)
953 daddr_t lastr; /* for read-ahead */
954 uint64_t vc_rwlockid; /* kthread owning rwlock */
955 short vc_locktrips; /* # of rwlock reacquisitions */
956 sema_t vc_rwlock; /* vop_rwlock for afs */
957 pgno_t mapcnt; /* # of pages mapped */
958 struct cred *cred; /* last writer's cred */
959 struct bhv_desc vc_bhv_desc; /* vnode's behavior data. */
960 #endif /* AFS_SGI_ENV */
961 #if defined(AFS_LINUX_ENV) || defined(AFS_FBSD_ENV)
962 afs_ucred_t *cred; /* last writer's cred */
965 struct dentry *target_link; /* dentry we prefer, when we are redirecting
966 * all requests due to duplicate dentry aliases.
967 * See LINUX/osi_vnodeops.c. Note that this is
968 * NOT an actual reference to a dentry, so this
969 * pointer MUST NOT be dereferenced on its own. */
971 afs_int32 vc_error; /* stash write error for this vnode. */
972 int xlatordv; /* Used by nfs xlator */
974 int asynchrony; /* num kbytes to store behind */
976 struct afs_q multiPage; /* list of multiPage_range structs */
978 int protocol; /* RX_FILESERVER, RX_OSD, ... defined in afsint.xg */
979 #if !defined(UKERNEL)
980 void *vpacRock; /* used to read or write in visible partitions */
982 afs_uint32 lastBRLWarnTime; /* last time we warned about byte-range locks */
984 spinlock_t pagewriter_lock;
985 struct list_head pagewriters; /* threads that are writing vm pages */
991 struct list_head link;
996 #define DONT_CHECK_MODE_BITS 0
997 #define CHECK_MODE_BITS 1
998 #define CMB_ALLOW_EXEC_AS_READ 2 /* For the NFS xlator */
1000 #if defined(AFS_SGI_ENV)
1001 #define AVCRWLOCK(avc) (valusema(&(avc)->vc_rwlock) <= 0)
1003 #define AFS_RWLOCK_T vrwlock_t
1004 #include <ksys/behavior.h>
1005 #define AFS_RWLOCK(V,F) \
1006 afs_rwlock(&VTOAFS(V)->vc_bhv_desc, (F));
1007 #define AFS_RWUNLOCK(V,F) \
1008 afs_rwunlock(&VTOAFS(V)->vc_bhv_desc, (F));
1010 #endif /* AFS_SGI_ENV */
1013 struct VenusFid fid;
1014 afs_hyper_t DataVersion;
1016 afs_int32 parentVnode;
1017 afs_int32 parentUnique;
1018 afs_hyper_t flushDV;
1021 afs_int32 randomUid[CPSIZE];
1022 afs_int32 callback; /* Now a pointer to 'server' struct */
1023 afs_int32 cbExpires;
1024 afs_int32 randomAccess[CPSIZE];
1025 afs_int32 anyAccess;
1027 short execsOrWriters;
1034 afs_int32 callerAccess;
1035 afs_int32 cbExpires;
1036 afs_int32 anyAccess;
1045 /* CM inititialization parameters. What CM actually used after calculations
1046 * based on passed in arguments.
1048 #define CMI_VERSION 1 /* increment when adding new fields. */
1049 struct cm_initparams {
1051 int cmi_nChunkFiles;
1052 int cmi_nStatCaches;
1053 int cmi_nDataCaches;
1054 int cmi_nVolumeCaches;
1055 int cmi_firstChunkSize;
1056 int cmi_otherChunkSize;
1057 int cmi_cacheSize; /* The original cache size, in 1K blocks. */
1058 unsigned cmi_setTime:1;
1059 unsigned cmi_memCache:1;
1060 int spare[16 - 9]; /* size of struct is 16 * 4 = 64 bytes */
1064 /*----------------------------------------------------------------------
1065 * AFS Data cache definitions
1067 * Each entry describes a Unix file on the local disk that is
1068 * is serving as a cached copy of all or part of a Vice file.
1069 * Entries live in circular queues for each hash table slot
1071 * Which queue is this thing in? Good question.
1072 * A struct dcache entry is in the freeDSlot queue when not associated with a cache slot (file).
1073 * Otherwise, it is in the DLRU queue. The freeDSlot queue uses the lruq.next field as
1074 * its "next" pointer.
1076 * Cache entries in the DLRU queue are either associated with vice files, in which case
1077 * they are hashed by afs_dvnextTbl and afs_dcnextTbl pointers, or they are in the freeDCList
1078 * and are not associated with any vice file. This last list uses the afs_dvnextTbl pointer for
1079 * its "next" pointer.
1080 *----------------------------------------------------------------------*/
1082 #define NULLIDX (-1) /* null index definition */
1083 /* struct dcache states bits */
1087 #define DWriting 8 /* file being written (used for cache validation) */
1089 /* dcache data flags */
1090 #define DFEntryMod 0x02 /* has entry itself been modified? */
1091 #define DFFetching 0x04 /* file is currently being fetched */
1093 /* dcache meta flags */
1094 #define DFNextStarted 0x01 /* next chunk has been prefetched already */
1095 #define DFFetchReq 0x10 /* someone is waiting for DFFetching to go on */
1098 /* flags in afs_indexFlags array */
1099 #define IFEverUsed 1 /* index entry has >= 1 byte of data */
1100 #define IFFree 2 /* index entry in freeDCList */
1101 #define IFDataMod 4 /* file needs to be written out */
1102 #define IFFlag 8 /* utility flag */
1103 #define IFDirtyPages 16 /* Solaris-only. contains dirty pages */
1104 #define IFAnyPages 32
1105 #define IFDiscarded 64 /* index entry in discardDCList */
1107 #ifdef AFS_DARWIN100_ENV
1108 typedef user_addr_t iparmtype; /* 64 bit */
1109 typedef user_addr_t uparmtype; /* 64 bit */
1111 typedef char * uparmtype;
1113 typedef afs_uint32 iparmtype;
1115 typedef long iparmtype;
1119 #if SIZEOF_VOID_P == SIZEOF_UNSIGNED_INT
1120 # define uintptrsz unsigned int
1121 #elif SIZEOF_VOID_P == SIZEOF_UNSIGNED_LONG
1122 # define uintptrsz unsigned long
1123 #elif SIZEOF_VOID_P == SIZEOF_UNSIGNED_LONG_LONG
1124 # define uintptrsz unsigned long long
1126 # error "Unable to determine casting for pointers"
1130 uparmtype in; /* input buffer */
1131 uparmtype out; /* output buffer */
1132 short in_size; /* Size of input buffer <= 2K */
1133 short out_size; /* Maximum size of output buffer, <= 2K */
1137 * This version of afs_ioctl is required to pass in 32 bit user space
1138 * pointers into a 64 bit kernel.
1141 struct afs_ioctl32 {
1149 /* CacheItems file has a header of type struct afs_fheader
1150 * (keep aligned properly). Since we already have sgi_62 clients running
1151 * with a 32 bit inode, a change is required to the header so that
1152 * they can distinguish the old 32 bit inode CacheItems file and zap it
1153 * instead of using it.
1155 struct afs_fheader {
1156 #define AFS_FHMAGIC 0x7635abaf /* uses version number */
1158 #define AFS_CI_VERSION 4
1160 afs_uint32 dataSize;
1161 afs_int32 firstCSize;
1162 afs_int32 otherCSize;
1165 #if defined(AFS_CACHE_VNODE_PATH)
1166 typedef char *afs_ufs_dcache_id_t;
1167 #elif defined(AFS_SGI_ENV) || defined(AFS_SUN5_64BIT_ENV)
1168 /* Using ino64_t here so that user level debugging programs compile
1169 * the size correctly.
1171 typedef ino64_t afs_ufs_dcache_id_t;
1172 #elif defined(AFS_LINUX_ENV)
1173 #define MAX_FH_LEN 10
1175 #if defined(NEW_EXPORT_OPS)
1178 __u32 raw[MAX_FH_LEN];
1179 } afs_ufs_dcache_id_t;
1180 extern int cache_fh_type;
1181 extern int cache_fh_len;
1182 #elif defined(AFS_LINUX_64BIT_KERNEL) && !defined(AFS_S390X_LINUX_ENV)
1183 typedef long afs_ufs_dcache_id_t;
1184 #elif defined(AFS_AIX51_ENV) || defined(AFS_HPUX1123_ENV)
1185 typedef ino_t afs_ufs_dcache_id_t;
1187 typedef afs_int32 afs_ufs_dcache_id_t;
1190 typedef afs_int32 afs_mem_dcache_id_t;
1193 afs_ufs_dcache_id_t ufs;
1194 afs_mem_dcache_id_t mem;
1197 /* it does not compile outside kernel */
1199 afs_int32 fid; /* is adc->index, the cache file number */
1200 afs_dcache_id_t inode; /* is adc->f.inode, the inode number of the cac\
1203 afs_int32 accesstime;
1204 struct buffer *hashNext;
1209 afs_rwlock_t lock; /* the lock for this structure */
1212 /* kept on disk and in dcache entries */
1214 struct VenusFid fid; /* Fid for this file */
1215 afs_int32 modTime; /* last time this entry was modified */
1216 afs_hyper_t versionNo; /* Associated data version number */
1217 afs_int32 chunk; /* Relative chunk number */
1218 afs_dcache_id_t inode; /* Unix inode for this chunk */
1219 afs_int32 chunkBytes; /* Num bytes in this chunk */
1220 char states; /* Has this chunk been modified? */
1223 /* magic numbers to specify the cache type */
1225 #define AFS_FCACHE_TYPE_UFS 0x0
1226 #define AFS_FCACHE_TYPE_MEM 0x1
1227 #define AFS_FCACHE_TYPE_NFS 0x2
1228 #define AFS_FCACHE_TYPE_EPI 0x3
1230 /* kept in memory */
1232 struct afs_q lruq; /* Free queue for in-memory images */
1233 struct afs_q dirty; /* Queue of dirty entries that need written */
1234 afs_rwlock_t lock; /* Protects validPos, some f */
1235 afs_rwlock_t tlock; /* Atomizes updates to refCount */
1236 afs_rwlock_t mflock; /* Atomizes accesses/updates to mflags */
1237 afs_size_t validPos; /* number of valid bytes during fetch */
1238 afs_int32 index; /* The index in the CacheInfo file */
1239 short refCount; /* Associated reference count. */
1240 char dflags; /* Data flags */
1241 char mflags; /* Meta flags */
1242 struct fcache f; /* disk image */
1243 afs_int32 bucket; /* which bucket these dcache entries are in */
1247 * dcache.lock protects the actual contents of the cache file (in
1248 * f.inode), subfields of f except those noted below, dflags and
1251 * dcache.tlock is used to make atomic updates to refCount. Zero
1252 * refCount dcache entries are protected by afs_xdcache instead of
1255 * dcache.mflock is used to access and update mflags. It cannot be
1256 * held without holding the corresponding dcache.lock. Updating
1257 * mflags requires holding dcache.lock(R) and dcache.mflock(W), and
1258 * checking for mflags requires dcache.lock(R) and dcache.mflock(R).
1259 * Note that dcache.lock(W) gives you the right to update mflags,
1260 * as dcache.mflock(W) can only be held with dcache.lock(R).
1262 * dcache.index, dcache.f.fid, dcache.f.chunk and dcache.f.inode are
1263 * write-protected by afs_xdcache and read-protected by refCount.
1264 * Once an entry is referenced, these values cannot change, and if
1265 * it's on the free list (with refCount=0), it can be reused for a
1266 * different file/chunk. These values can only be written while
1267 * holding afs_xdcache(W) and allocating this dcache entry (thereby
1268 * ensuring noone else has a refCount on it).
1272 /* afs_memcache.c */
1273 struct memCacheEntry {
1274 int size; /* # of valid bytes in this entry */
1275 int dataSize; /* size of allocated data area */
1276 afs_lock_t afs_memLock;
1277 char *data; /* bytes */
1280 struct afs_FetchOutput {
1281 struct AFSVolSync tsync;
1282 struct AFSFetchStatus OutStatus;
1283 struct AFSCallBack CallBack;
1286 /* macro to mark a dcache entry as bad */
1289 (x)->f.fid.Fid.Unique = 0; \
1290 afs_indexUnique[(x)->index] = 0; \
1291 (x)->dflags |= DFEntryMod; \
1294 /* FakeOpen and Fake Close used to be real subroutines. They're only used in
1295 * sun_subr and afs_vnodeops, and they're very frequently called, so I made
1296 * them into macros. They do:
1297 * FakeOpen: fake the file being open for writing. avc->lock must be held
1298 * in write mode. Having the file open for writing is like having a DFS
1299 * write-token: you're known to have the best version of the data around,
1300 * and so the CM won't let it be overwritten by random server info.
1301 * FakeClose: undo the effects of FakeOpen, noting that we want to ensure
1302 * that a real close eventually gets done. We use CCore to achieve this if
1303 * we would end up closing the file. avc->lock must be held in write mode */
1305 #ifdef AFS_AIX_IAUTH_ENV
1306 #define CRKEEP(V, C) (V)->linkData = (char*)crdup((C))
1308 #define CRKEEP(V, C) crhold((C)); (V)->linkData = (char*)(C)
1311 #define afs_FakeOpen(avc) { avc->opens++; avc->execsOrWriters++; }
1312 #define afs_FakeClose(avc, acred) \
1313 { if (avc->execsOrWriters == 1) { \
1314 /* we're the last writer, just use CCore flag */ \
1315 avc->f.states |= CCore; /* causes close to be called later */ \
1317 /* The cred and vnode holds will be released in afs_FlushActiveVcaches */ \
1318 osi_Assert(osi_vnhold(avc) == 0); /* So it won't disappear */ \
1319 CRKEEP(avc, acred); /* Should use a better place for the creds */ \
1322 /* we're not the last writer, let the last one do the store-back for us */ \
1324 avc->execsOrWriters--; \
1328 #define AFS_ZEROS 64 /* zero buffer */
1330 #define afs_DirtyPages(avc) ((avc)->f.states & CDirty)
1332 #define afs_InReadDir(avc) (((avc)->f.states & CReadDir) && (avc)->readdir_pid == MyPidxx2Pid(MyPidxx))
1334 extern struct dcache **afs_indexTable; /*Pointers to in-memory dcache entries */
1335 extern afs_int32 *afs_indexUnique; /*dcache entry Fid.Unique */
1336 extern afs_int32 *afs_dvnextTbl; /*Dcache hash table links */
1337 extern afs_int32 *afs_dcnextTbl; /*Dcache hash table links */
1338 extern afs_int32 afs_cacheFiles; /*Size of afs_indexTable */
1339 extern afs_int32 afs_cacheBlocks; /*1K blocks in cache */
1340 extern afs_int32 afs_cacheStats; /*Stat entries in cache */
1341 extern afs_int32 afs_freeDCCount; /*Count of elts in freeDCList */
1342 extern afs_uint32 afs_CacheTooFullCount;
1343 extern afs_uint32 afs_WaitForCacheDrainCount;
1344 extern struct vcache *afs_vhashT[VCSIZE]; /*Stat cache hash table */
1345 extern struct afs_q afs_vhashTV[VCSIZE]; /* cache hash table on volume */
1346 extern afs_int32 afs_initState; /*Initialization state */
1347 extern afs_int32 afs_termState; /* Termination state */
1348 extern struct VenusFid afs_rootFid; /*Root for whole file system */
1349 extern afs_int32 afs_allCBs; /* Count of callbacks */
1350 extern afs_int32 afs_oddCBs; /* Count of odd callbacks */
1351 extern afs_int32 afs_evenCBs; /* Count of even callbacks */
1352 extern afs_int32 afs_allZaps; /* Count of fid deletes */
1353 extern afs_int32 afs_oddZaps; /* Count of odd fid deletes */
1354 extern afs_int32 afs_evenZaps; /* Count of even fid deletes */
1355 extern struct brequest afs_brs[NBRS]; /* request structures */
1357 #define UHash(auid) ((auid) & (NUSERS-1))
1358 #define VHash(avol) ((avol)&(NVOLS-1))
1359 #define SHash(aserv) ((ntohl(aserv)) & (NSERVERS-1))
1360 #define FVHash(acell,avol) (((avol)+(acell)) & (NFENTRIES-1))
1362 /* Performance hack - we could replace VerifyVCache2 with the appropriate
1363 * GetVCache incantation, and could eliminate even this code from afs_UFSRead
1364 * by making intentionally invalidating quick.stamp in the various callbacks
1365 * expiration/breaking code */
1366 #ifdef AFS_DARWIN_ENV
1367 #define afs_VerifyVCache(avc, areq) \
1368 (((avc)->f.states & CStatd) ? (osi_VM_Setup(avc, 0), 0) : \
1369 afs_VerifyVCache2((avc),areq))
1371 #define afs_VerifyVCache(avc, areq) \
1372 (((avc)->f.states & CStatd) ? 0 : afs_VerifyVCache2((avc),areq))
1375 #define DO_STATS 1 /* bits used by FindVCache */
1380 /* values for adown value of afs_LoopServers */
1382 #define AFS_LS_DOWN 1
1383 #define AFS_LS_ALL 2
1385 /* values for flag param of afs_CheckVolumeNames */
1386 #define AFS_VOLCHECK_EXPIRED 0x1 /* volumes whose callbacks have expired */
1387 #define AFS_VOLCHECK_BUSY 0x2 /* volumes which were marked busy */
1388 #define AFS_VOLCHECK_MTPTS 0x4 /* mount point invalidation also */
1389 #define AFS_VOLCHECK_FORCE 0x8 /* do all forcibly */
1393 #define AFS_FSPORT ((unsigned short) htons(7000))
1394 #define AFS_VLPORT ((unsigned short) htons(7003))
1395 #define AFS_RXOSDPORT ((unsigned short) htons(7011))
1397 #define afs_rdwr(avc, uio, rw, io, cred) \
1398 (((rw) == UIO_WRITE) ? afs_write(avc, uio, io, cred, 0) : afs_read(avc, uio, cred, 0))
1399 #define afs_nlrdwr(avc, uio, rw, io, cred) \
1400 (((rw) == UIO_WRITE) ? afs_write(avc, uio, io, cred, 1) : afs_read(avc, uio, cred, 1))
1402 /* Cache size truncation uses the following low and high water marks:
1403 * If the cache is more than 90% full by space (CM_DCACHESPACEFREEPCT) or
1404 * 95% full by # of chunks (CM_DCACHECOUNTFREEPCT), the cache truncation
1405 * daemon is awakened and will start freeing space. The cache truncation
1406 * daemon will continue to free space until the cache is at most 85% full
1407 * by space (CM_DCACHESPACEFREEPCT - CM_DCACHEEXTRAPCT). The cache truncation
1408 * daemon will also try to free space until the cache is at most 90% full by
1409 * chunks (CM_DCACHECOUNTFREEPCT - CM_DCACHEEXTRAPCT), but the 85% space limit
1410 * is the only limit that we must hit.
1411 * afs_UFSWrite and afs_GetDCache (when it needs to fetch data) will wait on
1412 * afs_WaitForCacheDrain if the cache is 98% (CM_WAITFORDRAINPCT) full.
1413 * afs_GetDownD wakes those processes once the cache is 95% full
1414 * (CM_CACHESIZEDRAINEDPCT).
1416 #define CM_MAXDISCARDEDCHUNKS 16 /* # of chunks */
1417 #define CM_DCACHECOUNTFREEPCT 95 /* max pct of chunks in use */
1418 #define CM_DCACHESPACEFREEPCT 90 /* max pct of space in use */
1419 #define CM_DCACHEEXTRAPCT 5 /* extra to get when freeing */
1420 #define CM_CACHESIZEDRAINEDPCT 95 /* wakeup processes when down to here. */
1421 #define CM_WAITFORDRAINPCT 98 /* sleep if cache is this full. */
1423 /* when afs_cacheBlocks is large, settle for slightly decreased precision */
1424 #define PERCENT(p, v) \
1425 ((afs_cacheBlocks & 0xffe00000) ? ((v) / 100 * (p)) : ((p) * (v) / 100))
1427 #define afs_CacheIsTooFull() \
1428 (afs_blocksUsed - afs_blocksDiscarded > \
1429 PERCENT(CM_DCACHESPACEFREEPCT, afs_cacheBlocks) || \
1430 afs_freeDCCount - afs_discardDCCount < \
1431 PERCENT(100 - CM_DCACHECOUNTFREEPCT, afs_cacheFiles))
1433 /* Handy max length of a numeric string. */
1434 #define CVBS 12 /* max afs_int32 is 2^32 ~ 4*10^9, +1 for NULL, +luck */
1436 #define refpanic(foo) if (afs_norefpanic) \
1437 { printf( foo ); afs_norefpanic++;} else osi_Panic( foo )
1440 ** these are defined in the AIX source code sys/fs_locks.h but are not
1441 ** defined anywhere in the /usr/include directory
1443 #if defined(AFS_AIX41_ENV)
1444 #define VN_LOCK(vp) simple_lock(&(vp)->v_lock)
1445 #define VN_UNLOCK(vp) simple_unlock(&(vp)->v_lock)
1448 /* get a file's serial number from a vnode */
1449 #ifndef afs_vnodeToInumber
1450 #if defined(AFS_SGI_ENV) || defined(AFS_HAVE_VXFS) || defined(AFS_DARWIN_ENV)
1451 #define afs_vnodeToInumber(V) VnodeToIno(V)
1453 #define afs_vnodeToInumber(V) (VTOI(V)->i_number)
1454 #endif /* AFS_SGI_ENV */
1457 /* get a file's device number from a vnode */
1458 #ifndef afs_vnodeToDev
1459 #if defined(AFS_SGI_ENV) || defined(AFS_HAVE_VXFS) || defined(AFS_DARWIN_ENV)
1460 #define afs_vnodeToDev(V) VnodeToDev(V)
1461 #elif defined(UKERNEL)
1462 #define afs_vnodeToDev(V) (VTOI(V) ? (VTOI(V)->i_dev) : (-1))
1464 #define afs_vnodeToDev(V) (VTOI(V)->i_dev)
1468 /* declare something so that prototypes don't flip out */
1469 /* appears struct buf stuff is only actually passed around as a pointer,
1470 except with libuafs, in which case it is actually defined */
1474 struct rxfs_storeVariables {
1475 struct rx_call *call;
1476 struct vcache *vcache;
1480 afs_int32 hasNo64bit;
1481 struct AFSStoreStatus InStatus;
1485 int (*prepare)(void *rock, afs_uint32 size, afs_uint32 *bytestoxfer);
1486 int (*read)(void *rock, struct osi_file *tfile, afs_uint32 offset,
1487 afs_uint32 tlen, afs_uint32 *bytesread);
1488 int (*write)(void *rock, afs_uint32 tlen, afs_uint32 *byteswritten);
1489 int (*status)(void *rock);
1490 int (*padd)(void *rock, afs_uint32 tlen);
1491 int (*close)(void *rock, struct AFSFetchStatus *OutStatus,
1492 afs_int32 *doProcessFS);
1493 int (*destroy)(void **rock, afs_int32 error);
1494 int (*storeproc)(struct storeOps *, void *, struct dcache *, int *,
1499 int (*more)(void *rock, afs_int32 *length, afs_uint32 *moredata);
1500 int (*read)(void *rock, afs_uint32 tlen, afs_uint32 *bytesread);
1501 int (*write)(void *rock, struct osi_file *fp, afs_uint32 offset,
1502 afs_uint32 tlen, afs_uint32 *byteswritten);
1503 int (*close)(void *rock, struct vcache *avc, struct dcache *adc,
1504 struct afs_FetchOutput *Outputs);
1505 int (*destroy)(void **rock, afs_int32 error);
1508 /* fakestat support: opaque storage for afs_EvalFakeStat to remember
1509 * what vcache should be released.
1511 struct afs_fakestat_state {
1515 struct vcache *root_vp;
1518 extern int afs_fakestat_enable;
1519 extern int afs_rmtsys_enable;
1520 extern int afsd_dynamic_vcaches;
1523 * Wrappers for access to credentials structure members
1524 * Linux uses the kernel cred structure if available, with the
1525 * wrappers defined in LINUX/osi_machdep.h
1527 #if defined(AFS_NBSD40_ENV)
1528 /* in osi_machdep.h as expected */
1529 #elif defined (AFS_DARWIN110_ENV)
1530 #define afs_cr_uid(cred) kauth_cred_getuid((kauth_cred_t)(cred))
1531 #define afs_cr_gid(cred) kauth_cred_getgid((kauth_cred_t)(cred))
1532 #elif !(defined(AFS_LINUX_ENV) && defined(STRUCT_TASK_STRUCT_HAS_CRED))
1533 #define afs_cr_uid(cred) ((cred)->cr_uid)
1534 #define afs_cr_gid(cred) ((cred)->cr_gid)
1535 #if !defined(AFS_OBSD_ENV)
1536 #define afs_cr_ruid(cred) ((cred)->cr_ruid)
1537 #define afs_cr_rgid(cred) ((cred)->cr_rgid)
1540 #if !defined(AFS_DARWIN110_ENV)
1542 afs_set_cr_uid(afs_ucred_t *cred, uid_t uid) {
1546 afs_set_cr_gid(afs_ucred_t *cred, gid_t gid) {
1549 #if !defined(AFS_OBSD_ENV)
1551 afs_set_cr_ruid(afs_ucred_t *cred, uid_t uid) {
1552 cred->cr_ruid = uid;
1555 afs_set_cr_rgid(afs_ucred_t *cred, gid_t gid) {
1556 cred->cr_rgid = gid;
1558 #endif /* ! AFS_OBSD_ENV */
1559 #endif /* ! AFS_DARWIN110_ENV */
1564 /** The 32 bit OS expects the members of this structure to be 32 bit
1565 * quantities and the 64 bit OS expects them as 64 bit quanties. Hence
1566 * to accomodate both, *long* is used instead of afs_int32
1578 extern int Afs_syscall(struct afssysa *uap, rval_t *rvp);
1579 #endif /* AFS_SUN5_ENV */
1582 * Various definitions for osi_sleep and its event hash table
1583 * DFBSD has no osi_sleep, and HPUX has its own hack for this stuff
1585 #define AFS_EVHASHSIZE 128 /* size of afs_evhasht, must be power of 2 */
1587 typedef struct afs_event {
1588 struct afs_event *next; /* next in hash chain */
1589 char *event; /* lwp event: an address */
1590 int refcount; /* Is it in use? */
1591 int seq; /* Sequence number: this is incremented
1592 * by wakeup calls; wait will not return until
1594 #if defined(AFS_AIX_ENV)
1596 #elif defined(AFS_DARWIN_ENV)
1597 # ifdef AFS_DARWIN80_ENV
1601 /* no cond member */
1602 #elif defined(AFS_FBSD_ENV) || defined(AFS_OBSD_ENV)
1603 int cond; /* "all this gluck should probably be replaced by CVs" */
1604 #elif defined(AFS_LINUX_ENV)
1605 wait_queue_head_t cond;
1606 #elif defined(AFS_NBSD_ENV) || defined(AFS_SUN5_ENV) || defined(AFS_SGI_ENV)
1607 kcondvar_t cond; /* Currently associated condition variable */
1611 extern afs_event_t *afs_evhasht[AFS_EVHASHSIZE]; /* Hash table for events */
1613 #endif /* _AFS_H_ */