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
16 #include <afs/afs_args.h>
18 #include "afs/afs_args.h"
22 /* Upper bound on number of iovecs out uio routines will deal with. */
23 #define AFS_MAXIOVCNT 16
26 extern int afs_shuttingdown;
29 * Macros to uniquely identify the AFS vfs struct
31 #define AFS_VFSMAGIC 0x1234
32 #if defined(AFS_SUN_ENV) || defined(AFS_HPUX90_ENV) || defined(AFS_LINUX20_ENV)
33 #define AFS_VFSFSID 99
35 #if defined(AFS_SGI_ENV)
36 #define AFS_VFSFSID afs_fstype
38 #define AFS_VFSFSID AFS_MOUNT_AFS
42 /* Moved from VNOPS/afs_vnop_flocks so can be used in prototypes */
43 #if defined(AFS_HPUX102_ENV)
44 #define AFS_FLOCK k_flock
46 #if defined(AFS_SUN56_ENV) || defined(AFS_LINUX24_ENV)
47 #define AFS_FLOCK flock64
49 #define AFS_FLOCK flock
50 #endif /* AFS_SUN65_ENV */
51 #endif /* AFS_HPUX102_ENV */
53 /* The following are various levels of afs debugging */
54 #define AFSDEB_GENERAL 1 /* Standard debugging */
55 #define AFSDEB_NETWORK 2 /* low level afs networking */
56 #define AFSDEB_RX 4 /* RX debugging */
57 #define AFSDEB_VNLAYER 8 /* interface layer to AFS (aixops, gfsops, etc) */
59 /* generic undefined vice id */
62 /* The basic defines for the Andrew file system
63 better keep things powers of two so "& (foo-1)" hack works for masking bits */
64 #define MAXHOSTS 13 /* max hosts per single volume */
65 #define OMAXHOSTS 8 /* backwards compatibility */
66 #define MAXCELLHOSTS 8 /* max vldb servers per cell */
67 #define NBRS 15 /* max number of queued daemon requests */
68 #define NUSERS 16 /* hash table size for unixuser table */
69 #define NSERVERS 16 /* hash table size for server table */
70 #define NVOLS 64 /* hash table size for volume table */
71 #define NFENTRIES 256 /* hash table size for disk volume table */
72 #define VCSIZE 1024 /* stat cache hash table size */
73 #define DCSIZE 512 /* disk cache hash table size */
74 #define PIGGYSIZE 1350 /* max piggyback size */
75 #define MAXVOLS 128 /* max vols we can store */
76 #define MAXSYSNAME 128 /* max sysname (i.e. @sys) size */
77 #define MAXNUMSYSNAMES 16 /* max that current constants allow */
78 #define NOTOKTIMEOUT (2*3600) /* time after which to timeout conns sans tokens */
79 #define NOPAG 0xffffffff
80 #define AFS_NCBRS 300 /* max # of call back return entries */
81 #define AFS_MAXCBRSCALL 16 /* max to return in a given call */
82 #define AFS_SALLOC_LOW_WATER 250 /* Min free blocks before allocating more */
83 #define AFS_LRALLOCSIZ 4096 /* "Large" allocated size */
85 #define AFS_NRXPACKETS 80
86 #define AFS_RXDEADTIME 50
87 #define AFS_HARDDEADTIME 120
89 extern afs_int32 afs_rx_deadtime;
90 extern afs_int32 afs_rx_harddead;
98 /* flags to use with AFSOP_CACHEINIT */
99 #define AFSCALL_INIT_MEMCACHE 0x1 /* use a memory-based cache */
101 /* below here used only for kernel procedures */
103 /* Store synchrony flags - SYNC means that data should be forced to server's
104 * disk immediately upon completion. */
107 #define AFS_VMSYNC_INVAL 2 /* sync and invalidate pages */
108 #define AFS_LASTSTORE 4
111 /* background request structure */
114 #define BOP_NOOP 0 /* leave 0 unused */
115 #define BOP_FETCH 1 /* parm1 is chunk to get */
116 #define BOP_STORE 2 /* parm1 is chunk to store */
117 #define BOP_PATH 3 /* parm1 is path, parm2 is chunk to fetch */
119 #define B_DONTWAIT 1 /* On failure return; don't wait */
121 /* protocol is: refCount is incremented by user to take block out of free pool.
122 Next, BSTARTED is set when daemon finds request. This prevents
123 other daemons from picking up the same request. Finally, when
124 request is done, refCount is zeroed. BDONE and BWAIT are used by
125 dudes waiting for operation to proceed to a certain point before returning.
127 #define BSTARTED 1 /* request picked up by a daemon */
128 #define BUVALID 2 /* code is valid (store) */
129 #define BUWAIT 4 /* someone is waiting for BUVALID */
131 struct vcache *vc; /* vnode to use, with vrefcount bumped */
132 struct AFS_UCRED *cred; /* credentials to use for operation */
133 afs_size_t size_parm[BPARMS]; /* random parameters */
134 void *ptr_parm[BPARMS]; /* pointer parameters */
135 afs_int32 code; /* return code */
136 short refCount; /* use counter for this structure */
137 char opcode; /* what to do (store, fetch, etc) */
138 char flags; /* free, etc */
139 afs_int32 ts; /* counter "timestamp" */
147 afs_int32 AuthHandle;
148 char HandShakeKey[8];
150 afs_int32 BeginTimestamp;
151 afs_int32 EndTimestamp;
155 afs_int32 Cell; /* better sun packing if at end of structure */
159 /* Temporary struct to be passed between afs_fid and afs_vget; in SunOS4.x we can only pass a maximum of 10 bytes for a handle (we ideally need 16!) */
162 afs_int32 CellAndUnique;
165 /* The actual number of bytes in the SmallFid, not the sizeof struct. */
166 #define SIZEOF_SMALLFID 10
170 * Queues implemented with both pointers and short offsets into a disk file.
178 afs_int32 uid; /* user id making the request */
179 afs_int32 busyCount; /* how many busies we've seen so far */
180 afs_int32 flags; /* things like O_SYNC, O_NONBLOCK go here */
181 char initd; /* if non-zero, non-uid fields meaningful */
182 char accessError; /* flags for overriding error return code */
183 char volumeError; /* encountered a missing or busy volume */
184 char networkError; /* encountered network problems */
185 char permWriteError; /* fileserver returns permenent error. */
190 /* structure linked off of a server to keep track of queued returned
191 * callbacks. Sent asynchronously when we run a little low on free dudes.
194 struct afs_cbr *next;
198 /* cellinfo file magic number */
199 #define AFS_CELLINFO_MAGIC 0xf32817cd
202 #define CNoSUID 0x02 /* disable suid bit for this cell */
203 #define CLinkedCell4 0x04 /* reserved for ADDCELL2 pioctl */
204 #define CNoAFSDB 0x08 /* never bother trying AFSDB */
205 #define CHasVolRef 0x10 /* volumes were referenced */
206 #define CLinkedCell 0x20 /* has a linked cell in lcellp */
209 struct afs_q lruq; /* lru q next and prev */
210 char *cellName; /* char string name of cell */
211 afs_int32 cellIndex; /* sequence number */
212 afs_int32 cellNum; /* semi-permanent cell number */
213 struct server *cellHosts[MAXCELLHOSTS]; /* volume *location* hosts */
214 struct cell *lcellp; /* Associated linked cell */
215 u_short fsport; /* file server port */
216 u_short vlport; /* volume server port */
217 short states; /* state flags */
218 time_t timeout; /* data expire time, if non-zero */
219 struct cell_name *cnamep; /* pointer to our cell_name */
220 afs_rwlock_t lock; /* protects cell data */
224 struct cell_name *next;
231 struct cell_alias *next;
237 #define afs_PutCell(cellp, locktype)
239 /* the unixuser flag bit definitions */
240 #define UHasTokens 1 /* are the st and ct fields valid (ever set)? */
241 #define UTokensBad 2 /* are tokens bad? */
242 #define UPrimary 4 /* on iff primary identity */
243 #define UNeedsReset 8 /* needs afs_ResetAccessCache call done */
244 #define UPAGCounted 16 /* entry seen during PAG search (for stats) */
245 /* A flag used by afs_GCPAGs to keep track of
246 * which entries in afs_users need to be deleted.
247 * The lifetime of its presence in the table is the
248 * lifetime of the afs_GCPAGs function.
250 #define TMP_UPAGNotReferenced 128
252 /* values for afs_gcpags */
253 enum { AFS_GCPAGS_NOTCOMPILED = 0, AFS_GCPAGS_OK =
254 1, AFS_GCPAGS_USERDISABLED, AFS_GCPAGS_EPROC0, AFS_GCPAGS_EPROCN,
255 AFS_GCPAGS_EEQPID, AFS_GCPAGS_EINEXACT, AFS_GCPAGS_EPROCEND,
256 AFS_GCPAGS_EPROCWALK, AFS_GCPAGS_ECREDWALK, AFS_GCPAGS_EPIDCHECK,
257 AFS_GCPAGS_ENICECHECK
260 extern afs_int32 afs_gcpags;
261 extern afs_int32 afs_gcpags_procsize;
262 extern afs_int32 afs_bkvolpref;
265 struct unixuser *next; /* next hash pointer */
266 afs_int32 uid; /* search based on uid and cell */
268 afs_int32 vid; /* corresponding vice id in specified cell */
269 short refCount; /* reference count for allocation */
270 char states; /* flag info */
271 afs_int32 tokenTime; /* last time tokens were set, used for timing out conn data */
272 afs_int32 stLen; /* ticket length (if kerberos, includes kvno at head) */
273 char *stp; /* pointer to ticket itself */
274 struct ClearToken ct;
275 struct afs_exporter *exporter; /* more info about the exporter for the remote user */
279 /* Per-connection block. */
280 struct conn *next; /* Next dude same server. */
281 struct unixuser *user; /* user validated with respect to. */
282 struct rx_connection *id; /* RPC connid. */
283 struct srvAddr *srvr; /* server associated with this conn */
284 short refCount; /* reference count for allocation */
285 unsigned short port; /* port associated with this connection */
286 char forceConnectFS; /* Should we try again with these tokens? */
292 /* Fid comparison routines */
293 #define FidCmp(a,b) ((a)->Fid.Unique != (b)->Fid.Unique \
294 || (a)->Fid.Vnode != (b)->Fid.Vnode \
295 || (a)->Fid.Volume != (b)->Fid.Volume \
296 || (a)->Cell != (b)->Cell)
298 #define FidMatches(afid,tvc) ((tvc)->fid.Fid.Vnode == (afid)->Fid.Vnode && \
299 (tvc)->fid.Fid.Volume == (afid)->Fid.Volume && \
300 (tvc)->fid.Cell == (afid)->Cell && \
301 ( (tvc)->fid.Fid.Unique == (afid)->Fid.Unique || \
302 (!(afid)->Fid.Unique && ((tvc)->states & CUnique))))
306 * Operations on circular queues implemented with pointers. Note: these queue
307 * objects are always located at the beginning of the structures they are linking.
309 #define QInit(q) ((q)->prev = (q)->next = (q))
310 #define QAdd(q,e) ((e)->next = (q)->next, (e)->prev = (q), \
311 (q)->next->prev = (e), (q)->next = (e))
312 #define QRemove(e) ((e)->next->prev = (e)->prev, (e)->prev->next = (e)->next)
313 #define QNext(e) ((e)->next)
314 #define QPrev(e) ((e)->prev)
315 #define QEmpty(q) ((q)->prev == (q))
316 /* this one takes q1 and sticks it on the end of q2 - that is, the other end, not the end
317 * that things are added onto. q1 shouldn't be empty, it's silly */
318 #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))
320 * Do lots of address arithmetic to go from vlruq to the base of the vcache
321 * structure. Don't move struct vnode, since we think of a struct vcache as
322 * a specialization of a struct vnode
324 #define QTOV(e) ((struct vcache *)(((char *) (e)) - (((char *)(&(((struct vcache *)(e))->vlruq))) - ((char *)(e)))))
325 #define QTOC(e) ((struct cell *)((char *) (e)))
328 #define SRVADDR_ISDOWN 0x20 /* same as SRVR_ISDOWN */
329 #define SRVADDR_NOUSE 0x40 /* Don't use this srvAddr */
331 struct srvAddr *next_bkt; /* next item in hash bucket */
332 struct srvAddr *next_sa; /* another interface on same host */
333 struct server *server; /* back to parent */
334 struct conn *conns; /* All user connections to this server */
335 afs_int32 sa_ip; /* Host addr in network byte order */
336 u_short sa_iprank; /* indiv ip address priority */
337 u_short sa_portal; /* port addr in network byte order */
342 * Values used in the flags field of the server structure below.
344 * AFS_SERVER_FLAG_ACTIVATED Has the server ever had a user connection
345 * associated with it?
347 #define AFS_SERVER_FLAG_ACTIVATED 0x01
348 #define SNO_LHOSTS 0x04
349 #define SYES_LHOSTS 0x08
350 #define SVLSRV_UUID 0x10
351 #define SRVR_ISDOWN 0x20
352 #define SRVR_MULTIHOMED 0x40
353 #define SRVR_ISGONE 0x80
354 #define SNO_INLINEBULK 0x100
355 #define SNO_64BIT 0x200
357 #define afs_serverSetNo64Bit(s) ((s)->srvr->server->flags |= SNO_64BIT)
358 #define afs_serverHasNo64Bit(s) ((s)->srvr->server->flags & SNO_64BIT)
364 afs_int32 addr_uniquifier;
368 struct srvAddr haddr;
371 #define sr_uuid _suid._srvUuid.suuid
372 #define sr_addr_uniquifier _suid._srvUuid.addr_uniquifier
373 #define sr_host _suid._srvId.haddr.ip
374 #define sr_portal _suid._srvId.haddr.portal
375 #define sr_rank _suid._srvId.haddr.ip_rank
376 #define sr_flags _suid._srvId.haddr.flags
377 #define sr_conns _suid._srvId.haddr.conns
378 struct server *next; /* Ptr to next server in hash chain */
379 struct cell *cell; /* Cell in which this host resides */
380 struct afs_cbr *cbrs; /* Return list of callbacks */
381 afs_int32 activationTime; /* Time when this record was first activated */
382 afs_int32 lastDowntimeStart; /* Time when last downtime incident began */
383 afs_int32 numDowntimeIncidents; /* # (completed) downtime incidents */
384 afs_int32 sumOfDowntimes; /* Total downtime experienced, in seconds */
385 struct srvAddr *addr;
386 afs_uint32 flags; /* Misc flags */
389 #define afs_PutServer(servp, locktype)
391 /* structs for some pioctls - these are (or should be)
399 struct sprefrequest_33 {
400 unsigned short offset;
401 unsigned short num_servers;
405 struct sprefrequest { /* new struct for 3.4 */
406 unsigned short offset;
407 unsigned short num_servers;
408 unsigned short flags;
413 unsigned short next_offset;
414 unsigned short num_servers;
415 struct spref servers[1]; /* we overrun this array intentionally... */
419 unsigned short flags;
420 unsigned short num_servers;
421 struct spref servers[1]; /* we overrun this array intentionally... */
423 /* struct for GAG pioctl
426 afs_uint32 showflags, logflags, logwritethruflag, spare[3];
427 unsigned char spare2[128];
431 #define logwritethruON 1
434 afs_int32 rx_initReceiveWindow, rx_maxReceiveWindow, rx_initSendWindow,
435 rx_maxSendWindow, rxi_nSendFrags, rxi_nRecvFrags, rxi_OrphanFragSize;
436 afs_int32 rx_maxReceiveSize, rx_MyMaxSendSize;
437 afs_uint32 spare[21];
440 /* struct for checkservers */
451 /* state bits for volume */
452 #define VRO 1 /* volume is readonly */
453 #define VRecheck 2 /* recheck volume info with server */
454 #define VBackup 4 /* is this a backup volume? */
455 #define VForeign 8 /* this is a non-afs volume */
456 #define VResort 16 /* server order was rearranged, sort when able */
457 #define VMoreReps 32 /* This volume has more replicas than we are */
458 /* keeping track of now -- check with VLDB */
460 enum repstate { not_busy, end_not_busy = 6, rd_busy, rdwr_busy, offline };
463 /* One structure per volume, describing where the volume is located
464 * and where its mount points are. */
465 struct volume *next; /* Next volume in hash list. */
466 afs_int32 cell; /* the cell in which the volume resides */
467 afs_rwlock_t lock; /* the lock for this structure */
468 afs_int32 volume; /* This volume's ID number. */
469 char *name; /* This volume's name, or 0 if unknown */
470 struct server *serverHost[MAXHOSTS]; /* servers serving this volume */
471 enum repstate status[MAXHOSTS]; /* busy, offline, etc */
472 struct VenusFid dotdot; /* dir to access as .. */
473 struct VenusFid mtpoint; /* The mount point for this volume. */
474 afs_int32 rootVnode, rootUnique; /* Volume's root fid */
477 afs_int32 rwVol; /* For r/o vols, original read/write volume. */
478 afs_int32 accessTime; /* last time we used it */
479 afs_int32 vtix; /* volume table index */
480 afs_int32 copyDate; /* copyDate field, for tracking vol releases */
481 afs_int32 expireTime; /* for per-volume callbacks... */
482 short refCount; /* reference count for allocation */
483 char states; /* here for alignment reasons */
486 #define afs_PutVolume(av, locktype) ((av)->refCount--)
488 /* format of an entry in volume info file */
490 afs_int32 cell; /* cell for this entry */
491 afs_int32 volume; /* volume */
492 afs_int32 next; /* has index */
493 struct VenusFid dotdot; /* .. value */
494 struct VenusFid mtpoint; /* mt point's fid */
495 afs_int32 rootVnode, rootUnique; /* Volume's root fid */
499 struct SimpleLocks *next;
501 afs_int32 boff, eoff;
503 #if defined(AFS_AIX32_ENV) || defined(AFS_SUN5_ENV) || defined(AFS_SGI_ENV)
508 /* vcache state bits */
509 #define CStatd 0x00000001 /* has this file ever been stat'd? */
510 #define CBackup 0x00000002 /* file is on a backup volume */
511 #define CRO 0x00000004 /* is it on a read-only volume */
512 #define CMValid 0x00000008 /* is the mount point info valid? */
513 #define CCore 0x00000010 /* storing a core file, needed since we don't get an open */
514 #define CDirty 0x00000020 /* file has been modified since first open (... O_RDWR) */
515 #define CSafeStore 0x00000040 /* close must wait for store to finish (should be in fd) */
516 #define CMAPPED 0x00000080 /* Mapped files; primarily used by SunOS 4.0.x */
517 #define CNSHARE 0x00000100 /* support O_NSHARE semantics */
518 #define CLied 0x00000200
519 #define CTruth 0x00000400
521 #define CWired 0x00000800 /* OSF hack only */
523 #ifdef AFS_DARWIN_ENV
524 #define CUBCinit 0x00000800
526 #define CWRITE_IGN 0x00000800 /* Next OS hack only */
529 #define CUnique 0x00001000 /* vc's uniquifier - latest unifiquier for fid */
530 #define CForeign 0x00002000 /* this is a non-afs vcache */
531 #define CUnlinked 0x00010000
532 #define CBulkStat 0x00020000 /* loaded by a bulk stat, and not ref'd since */
533 #define CUnlinkedDel 0x00040000
534 #define CVFlushed 0x00080000
535 #define CCore1 0x00100000 /* osf1 core file; not same as CCore above */
536 #define CWritingUFS 0x00200000 /* to detect vm deadlock - used by sgi */
537 #define CCreating 0x00400000 /* avoid needless store after open truncate */
538 #define CPageHog 0x00800000 /* AIX - dumping large cores is a page hog. */
539 #define CDCLock 0x02000000 /* Vnode lock held over call to GetDownD */
540 #define CBulkFetching 0x04000000 /* stats are being fetched by bulk stat */
541 #define CExtendedFile 0x08000000 /* extended file via ftruncate call. */
543 /* vcache vstate bits */
544 #define VRevokeWait 0x1
545 #define VPageCleaning 0x2 /* Solaris - Cache Trunc Daemon sez keep out */
548 #if defined(AFS_FBSD_ENV)
549 #define vrefCount v.v_usecount
551 #if defined(AFS_OBSD_ENV)
552 #define vrefCount v->v_usecount
554 #define vrefCount v.v_count
555 #endif /* AFS_OBSD_ENV */
556 #endif /* AFS_FBSD_ENV */
558 #ifdef AFS_LINUX24_ENV
559 #define VREFCOUNT(v) atomic_read(&((vnode_t *) v)->v_count)
560 #define VREFCOUNT_SET(v, c) atomic_set(&((vnode_t *) v)->v_count, c)
561 #define VREFCOUNT_DEC(v) atomic_dec(&((vnode_t *) v)->v_count)
562 #define VREFCOUNT_INC(v) atomic_inc(&((vnode_t *) v)->v_count)
563 #define DLOCK() spin_lock(&dcache_lock)
564 #define DUNLOCK() spin_unlock(&dcache_lock)
565 #define DGET(d) dget_locked(d)
566 #define DCOUNT(d) atomic_read(&(d)->d_count)
568 #define VREFCOUNT(v) ((v)->vrefCount)
569 #define VREFCOUNT_SET(v, c) (v)->vrefCount = c;
570 #define VREFCOUNT_DEC(v) (v)->vrefCount--;
571 #define VREFCOUNT_INC(v) (v)->vrefCount++;
574 #define DGET(d) dget(d)
575 #define DCOUNT(d) ((d)->d_count)
578 #define AFS_MAXDV 0x7fffffff /* largest dataversion number */
579 #ifdef AFS_64BIT_CLIENT
580 #define AFS_NOTRUNC 0x7fffffffffffffffLL /* largest positive int64 number */
581 #else /* AFS_64BIT_CLIENT */
582 #define AFS_NOTRUNC 0x7fffffff /* largest dataversion number */
583 #endif /* AFS_64BIT_CLIENT */
585 extern afs_int32 vmPageHog; /* counter for # of vnodes which are page hogs. */
588 * Fast map from vcache to dcache
594 afs_offs_t minLoc; /* smallest offset into dc. */
595 afs_offs_t len; /* largest offset into dc. */
598 extern afs_uint32 afs_stampValue; /* stamp for pair's usage */
599 #define MakeStamp() (++afs_stampValue)
601 #if defined(AFS_OBSD_ENV)
602 #define VTOAFS(v) ((struct vcache *)(v)->v_data)
603 #define AFSTOV(vc) ((vc)->v)
605 #define VTOAFS(V) ((struct vcache *)(V))
606 #define AFSTOV(V) (&(V)->v)
609 #ifdef AFS_LINUX22_ENV
610 #define ITOAFS(V) ((struct vcache*)(V))
611 #define AFSTOI(V) (struct inode *)(&(V)->v)
614 /* INVARIANTs: (vlruq.next != NULL) == (vlruq.prev != NULL)
615 * nextfree => !vlruq.next && ! vlruq.prev
616 * !(avc->nextfree) && !avc->vlruq.next => (FreeVCList == avc->nextfree)
619 #if defined(AFS_OBSD_ENV)
622 struct vnode v; /* Has reference count in v.v_count */
624 struct afs_q vlruq; /* lru q next and prev */
625 struct vcache *nextfree; /* next on free list (if free) */
626 struct vcache *hnext; /* Hash next */
630 afs_hyper_t DataVersion;
634 ushort Mode; /* XXXX Should be afs_int32 XXXX */
636 /* vnode type is in v.v_type */
638 afs_rwlock_t lock; /* The lock on the vcache contents. */
639 #if defined(AFS_SUN5_ENV)
640 /* Lock used to protect the activeV, multipage, and vstates fields.
641 * Do not try to get the vcache lock when the vlock is held */
643 #endif /* defined(AFS_SUN5_ENV) */
644 #if defined(AFS_SUN5_ENV)
648 #if defined(AFS_SUN_ENV) || defined(AFS_ALPHA_ENV) || defined(AFS_DARWIN_ENV) || defined(AFS_FBSD_ENV)
649 afs_bozoLock_t pvnLock; /* see locks.x */
654 #if defined(AFS_AIX51_ENV)
662 int ownslock; /* pid of owner of excl lock, else 0 - defect 3083 */
664 #ifdef AFS_DARWIN_ENV
665 struct lock__bsd__ rwlock;
670 afs_int32 parentVnode; /* Parent dir, if a file. */
671 afs_int32 parentUnique;
672 struct VenusFid *mvid; /* Either parent dir (if root) or root (if mt pt) */
673 char *linkData; /* Link data if a symlink. */
674 afs_hyper_t flushDV; /* data version last flushed from text */
675 afs_hyper_t mapDV; /* data version last flushed from map */
676 afs_size_t truncPos; /* truncate file to this position at next store */
677 struct server *callback; /* The callback host, if any */
678 afs_uint32 cbExpires; /* time the callback expires */
679 struct afs_q callsort; /* queue in expiry order, sort of */
680 struct axscache *Access; /* a list of cached access bits */
681 afs_int32 anyAccess; /* System:AnyUser's access to this. */
682 afs_int32 last_looker; /* pag/uid from last lookup here */
683 #if defined(AFS_SUN5_ENV)
685 #endif /* defined(AFS_SUN5_ENV) */
686 struct SimpleLocks *slocks;
687 short opens; /* The numbers of opens, read or write, on this file. */
688 short execsOrWriters; /* The number of execs (if < 0) or writers (if > 0) of
690 short flockCount; /* count of flock readers, or -1 if writer */
691 char mvstat; /* 0->normal, 1->mt pt, 2->root. */
692 afs_uint32 states; /* state bits */
693 #if defined(AFS_SUN5_ENV)
694 afs_uint32 vstates; /* vstate bits */
695 #endif /* defined(AFS_SUN5_ENV) */
697 afs_uint32 symhintstamp;
699 struct vcache *symhint;
700 struct dcache *dchint;
702 #ifdef AFS_LINUX22_ENV
703 u_short flushcnt; /* Number of flushes which haven't released yet. */
704 u_short mapcnt; /* Number of mappings of this file. */
706 #if defined(AFS_SGI_ENV)
707 daddr_t lastr; /* for read-ahead */
709 uint64_t vc_rwlockid; /* kthread owning rwlock */
711 short vc_rwlockid; /* pid of process owning rwlock */
713 short vc_locktrips; /* # of rwlock reacquisitions */
714 sema_t vc_rwlock; /* vop_rwlock for afs */
715 pgno_t mapcnt; /* # of pages mapped */
716 struct cred *cred; /* last writer's cred */
718 struct bhv_desc vc_bhv_desc; /* vnode's behavior data. */
720 #endif /* AFS_SGI_ENV */
721 afs_int32 vc_error; /* stash write error for this vnode. */
722 int xlatordv; /* Used by nfs xlator */
723 struct AFS_UCRED *uncred;
724 int asynchrony; /* num kbytes to store behind */
726 short multiPage; /* count of multi-page getpages in progress */
730 #define afs_symhint_inval(avc)
733 #define DONT_CHECK_MODE_BITS 0
734 #define CHECK_MODE_BITS 1
735 #define CMB_ALLOW_EXEC_AS_READ 2 /* For the NFS xlator */
737 #if defined(AFS_SGI_ENV)
738 #define AVCRWLOCK(avc) (valusema(&(avc)->vc_rwlock) <= 0)
740 /* SGI vnode rwlock macros and flags. */
741 #ifndef AFS_SGI62_ENV
742 /* The following are defined here. SGI 6.2 declares them in vnode.h */
743 #define VRWLOCK_READ 0
744 #define VRWLOCK_WRITE 1
745 #define VRWLOCK_WRITE_DIRECT 2
750 #define AFS_RWLOCK_T vrwlock_t
752 #define AFS_RWLOCK_T int
753 #endif /* AFS_SGI62_ENV */
755 #include <ksys/behavior.h>
756 #define AFS_RWLOCK(V,F) \
757 afs_rwlock(&VTOAFS(V)->vc_bhv_desc, (F));
758 #define AFS_RWUNLOCK(V,F) \
759 afs_rwunlock(&VTOAFS(V)->vc_bhv_desc, (F));
762 #define AFS_RWLOCK(V,F) afs_rwlock((vnode_t *)(V), (F) )
763 #define AFS_RWUNLOCK(V,F) afs_rwunlock((vnode_t *)(V), (F) )
765 #else /* AFS_SGI53_ENV */
766 #define AFS_RWLOCK(V,F) afs_rwlock((V))
767 #define AFS_RWUNLOCK(V,F) afs_rwunlock((V))
768 #endif /* AFS_SGI53_ENV */
769 #endif /* AFS_SGI_ENV */
773 afs_hyper_t DataVersion;
775 afs_int32 parentVnode;
776 afs_int32 parentUnique;
780 afs_int32 randomUid[CPSIZE];
781 afs_int32 callback; /* Now a pointer to 'server' struct */
783 afs_int32 randomAccess[CPSIZE];
786 short execsOrWriters;
797 /* CM inititialization parameters. What CM actually used after calculations
798 * based on passed in arguments.
800 #define CMI_VERSION 1 /* increment when adding new fields. */
801 struct cm_initparams {
806 int cmi_nVolumeCaches;
807 int cmi_firstChunkSize;
808 int cmi_otherChunkSize;
809 int cmi_cacheSize; /* The original cache size, in 1K blocks. */
810 unsigned cmi_setTime:1;
811 unsigned cmi_memCache:1;
812 int spare[16 - 9]; /* size of struct is 16 * 4 = 64 bytes */
816 /*----------------------------------------------------------------------
817 * AFS Data cache definitions
819 * Each entry describes a Unix file on the local disk that is
820 * is serving as a cached copy of all or part of a Vice file.
821 * Entries live in circular queues for each hash table slot
823 * Which queue is this thing in? Good question.
824 * A struct dcache entry is in the freeDSlot queue when not associated with a cache slot (file).
825 * Otherwise, it is in the DLRU queue. The freeDSlot queue uses the lruq.next field as
826 * its "next" pointer.
828 * Cache entries in the DLRU queue are either associated with vice files, in which case
829 * they are hashed by afs_dvnextTbl and afs_dcnextTbl pointers, or they are in the freeDCList
830 * and are not associated with any vice file. This last list uses the afs_dvnextTbl pointer for
831 * its "next" pointer.
832 *----------------------------------------------------------------------*/
834 #define NULLIDX (-1) /* null index definition */
835 /* struct dcache states bits */
836 #define DWriting 8 /* file being written (used for cache validation) */
838 /* dcache data flags */
839 #define DFEntryMod 0x02 /* has entry itself been modified? */
840 #define DFFetching 0x04 /* file is currently being fetched */
842 /* dcache meta flags */
843 #define DFNextStarted 0x01 /* next chunk has been prefetched already */
844 #define DFFetchReq 0x10 /* someone is waiting for DFFetching to go on */
847 /* flags in afs_indexFlags array */
848 #define IFEverUsed 1 /* index entry has >= 1 byte of data */
849 #define IFFree 2 /* index entry in freeDCList */
850 #define IFDataMod 4 /* file needs to be written out */
851 #define IFFlag 8 /* utility flag */
852 #define IFDirtyPages 16
853 #define IFAnyPages 32
854 #define IFDiscarded 64 /* index entry in discardDCList */
857 char *in; /* input buffer */
858 char *out; /* output buffer */
859 short in_size; /* Size of input buffer <= 2K */
860 short out_size; /* Maximum size of output buffer, <= 2K */
864 * This version of afs_ioctl is required to pass in 32 bit user space
865 * pointers into a 64 bit kernel.
876 /* CacheItems file has a header of type struct afs_fheader
877 * (keep aligned properly). Since we already have sgi_62 clients running
878 * with a 32 bit inode, a change is required to the header so that
879 * they can distinguish the old 32 bit inode CacheItems file and zap it
880 * instead of using it.
883 #define AFS_FHMAGIC 0x7635abaf /* uses version number */
885 #if defined(AFS_SUN57_64BIT_ENV)
886 #define AFS_CI_VERSION 3
888 #define AFS_CI_VERSION 2
891 afs_int32 firstCSize;
892 afs_int32 otherCSize;
895 #if defined(AFS_SGI61_ENV) || defined(AFS_SUN57_64BIT_ENV)
896 /* Using ino64_t here so that user level debugging programs compile
897 * the size correctly.
899 #define afs_inode_t ino64_t
901 #if defined(AFS_LINUX_64BIT_KERNEL)
902 #define afs_inode_t long
904 #if defined(AFS_AIX51_ENV) || defined(AFS_HPUX1123_ENV)
905 #define afs_inode_t ino_t
907 #define afs_inode_t afs_int32
912 /* kept on disk and in dcache entries */
914 struct VenusFid fid; /* Fid for this file */
915 afs_int32 modTime; /* last time this entry was modified */
916 afs_hyper_t versionNo; /* Associated data version number */
917 afs_int32 chunk; /* Relative chunk number */
918 afs_inode_t inode; /* Unix inode for this chunk */
919 afs_int32 chunkBytes; /* Num bytes in this chunk */
920 char states; /* Has this chunk been modified? */
923 /* magic numbers to specify the cache type */
925 #define AFS_FCACHE_TYPE_UFS 0x0
926 #define AFS_FCACHE_TYPE_MEM 0x1
927 #define AFS_FCACHE_TYPE_NFS 0x2
928 #define AFS_FCACHE_TYPE_EPI 0x3
932 struct afs_q lruq; /* Free queue for in-memory images */
933 afs_rwlock_t lock; /* Protects validPos, some f */
934 afs_rwlock_t tlock; /* Atomizes updates to refCount */
935 afs_rwlock_t mflock; /* Atomizes accesses/updates to mflags */
936 afs_size_t validPos; /* number of valid bytes during fetch */
937 afs_int32 index; /* The index in the CacheInfo file */
938 short refCount; /* Associated reference count. */
939 char dflags; /* Data flags */
940 char mflags; /* Meta flags */
941 struct fcache f; /* disk image */
942 afs_int32 stamp; /* used with vtodc struct for hints */
947 * dcache.lock protects the actual contents of the cache file (in
948 * f.inode), subfields of f except those noted below, dflags and
951 * dcache.tlock is used to make atomic updates to refCount. Zero
952 * refCount dcache entries are protected by afs_xdcache instead of
955 * dcache.mflock is used to access and update mflags. It cannot be
956 * held without holding the corresponding dcache.lock. Updating
957 * mflags requires holding dcache.lock(R) and dcache.mflock(W), and
958 * checking for mflags requires dcache.lock(R) and dcache.mflock(R).
959 * Note that dcache.lock(W) gives you the right to update mflags,
960 * as dcache.mflock(W) can only be held with dcache.lock(R).
962 * dcache.stamp is protected by the associated vcache lock, because
963 * it's only purpose is to establish correspondence between vcache
964 * and dcache entries.
966 * dcache.index, dcache.f.fid, dcache.f.chunk and dcache.f.inode are
967 * write-protected by afs_xdcache and read-protected by refCount.
968 * Once an entry is referenced, these values cannot change, and if
969 * it's on the free list (with refCount=0), it can be reused for a
970 * different file/chunk. These values can only be written while
971 * holding afs_xdcache(W) and allocating this dcache entry (thereby
972 * ensuring noone else has a refCount on it).
975 /* this is obsolete and should be removed */
978 /* macro to mark a dcache entry as bad */
981 (x)->f.fid.Fid.Unique = 0; \
982 afs_indexUnique[(x)->index] = 0; \
983 (x)->dflags |= DFEntryMod; \
986 /* FakeOpen and Fake Close used to be real subroutines. They're only used in
987 * sun_subr and afs_vnodeops, and they're very frequently called, so I made
988 * them into macros. They do:
989 * FakeOpen: fake the file being open for writing. avc->lock must be held
990 * in write mode. Having the file open for writing is like having a DFS
991 * write-token: you're known to have the best version of the data around,
992 * and so the CM won't let it be overwritten by random server info.
993 * FakeClose: undo the effects of FakeOpen, noting that we want to ensure
994 * that a real close eventually gets done. We use CCore to achieve this if
995 * we would end up closing the file. avc->lock must be held in write mode */
997 #ifdef AFS_AIX_IAUTH_ENV
998 #define CRKEEP(V, C) (V)->linkData = (char*)crdup((C))
1000 #define CRKEEP(V, C) crhold((C)); (V)->linkData = (char*)(C)
1003 #define afs_FakeOpen(avc) { avc->opens++; avc->execsOrWriters++; }
1004 #define afs_FakeClose(avc, acred) \
1005 { if (avc->execsOrWriters == 1) { \
1006 /* we're the last writer, just use CCore flag */ \
1007 avc->states |= CCore; /* causes close to be called later */ \
1009 /* The cred and vnode holds will be released in afs_FlushActiveVcaches */ \
1010 VN_HOLD(AFSTOV(avc)); /* So it won't disappear */ \
1011 CRKEEP(avc, acred); /* Should use a better place for the creds */ \
1014 /* we're not the last writer, let the last one do the store-back for us */ \
1016 avc->execsOrWriters--; \
1020 #define AFS_ZEROS 64 /* zero buffer */
1022 /*#define afs_DirtyPages(avc) (((avc)->states & CDirty) || osi_VMDirty_p((avc)))*/
1023 #define afs_DirtyPages(avc) ((avc)->states & CDirty)
1025 /* The PFlush algorithm makes use of the fact that Fid.Unique is not used in
1026 below hash algorithms. Change it if need be so that flushing algorithm
1027 doesn't move things from one hash chain to another
1029 /* extern int afs_dhashsize; */
1030 #define DCHash(v, c) ((((v)->Fid.Vnode + (v)->Fid.Volume + (c))) & (afs_dhashsize-1))
1031 /*Vnode, Chunk -> Hash table index */
1032 #define DVHash(v) ((((v)->Fid.Vnode + (v)->Fid.Volume )) & (afs_dhashsize-1))
1033 /*Vnode -> Other hash table index */
1034 /* don't hash on the cell, our callback-breaking code sometimes fails to compute
1035 the cell correctly, and only scans one hash bucket */
1036 #define VCHash(fid) (((fid)->Fid.Volume + (fid)->Fid.Vnode) & (VCSIZE-1))
1038 extern struct dcache **afs_indexTable; /*Pointers to in-memory dcache entries */
1039 extern afs_int32 *afs_indexUnique; /*dcache entry Fid.Unique */
1040 extern afs_int32 *afs_dvnextTbl; /*Dcache hash table links */
1041 extern afs_int32 *afs_dcnextTbl; /*Dcache hash table links */
1042 extern afs_int32 afs_cacheFiles; /*Size of afs_indexTable */
1043 extern afs_int32 afs_cacheBlocks; /*1K blocks in cache */
1044 extern afs_int32 afs_cacheStats; /*Stat entries in cache */
1045 extern struct vcache *afs_vhashT[VCSIZE]; /*Stat cache hash table */
1046 extern afs_int32 afs_initState; /*Initialization state */
1047 extern afs_int32 afs_termState; /* Termination state */
1048 extern struct VenusFid afs_rootFid; /*Root for whole file system */
1049 extern afs_int32 afs_allCBs; /* Count of callbacks */
1050 extern afs_int32 afs_oddCBs; /* Count of odd callbacks */
1051 extern afs_int32 afs_evenCBs; /* Count of even callbacks */
1052 extern afs_int32 afs_allZaps; /* Count of fid deletes */
1053 extern afs_int32 afs_oddZaps; /* Count of odd fid deletes */
1054 extern afs_int32 afs_evenZaps; /* Count of even fid deletes */
1055 extern struct brequest afs_brs[NBRS]; /* request structures */
1057 #define UHash(auid) ((auid) & (NUSERS-1))
1058 #define VHash(avol) ((avol)&(NVOLS-1))
1059 #define SHash(aserv) ((ntohl(aserv)) & (NSERVERS-1))
1060 #define FVHash(acell,avol) (((avol)+(acell)) & (NFENTRIES-1))
1062 /* Performance hack - we could replace VerifyVCache2 with the appropriate
1063 * GetVCache incantation, and could eliminate even this code from afs_UFSRead
1064 * by making intentionally invalidating quick.stamp in the various callbacks
1065 * expiration/breaking code */
1066 #ifdef AFS_LINUX20_ENV
1067 #define afs_VerifyVCache(avc, areq) \
1068 (((avc)->states & CStatd) ? (vcache2inode(avc), 0) : \
1069 afs_VerifyVCache2((avc),areq))
1071 #ifdef AFS_DARWIN_ENV
1072 #define afs_VerifyVCache(avc, areq) \
1073 (((avc)->states & CStatd) ? (osi_VM_Setup(avc, 0), 0) : \
1074 afs_VerifyVCache2((avc),areq))
1076 #define afs_VerifyVCache(avc, areq) \
1077 (((avc)->states & CStatd) ? 0 : afs_VerifyVCache2((avc),areq))
1081 #define DO_STATS 1 /* bits used by FindVCache */
1084 /* values for flag param of afs_CheckVolumeNames */
1085 #define AFS_VOLCHECK_EXPIRED 0x1 /* volumes whose callbacks have expired */
1086 #define AFS_VOLCHECK_BUSY 0x2 /* volumes which were marked busy */
1087 #define AFS_VOLCHECK_MTPTS 0x4 /* mount point invalidation also */
1088 #define AFS_VOLCHECK_FORCE 0x8 /* do all forcibly */
1092 #define AFS_FSPORT ((unsigned short) htons(7000))
1093 #define AFS_VLPORT ((unsigned short) htons(7003))
1095 #define afs_read(avc, uio, acred, albn, abpp, nolock) \
1096 (*(afs_cacheType->vread))(avc, uio, acred, albn, abpp, nolock)
1097 #define afs_write(avc, uio, aio, acred, nolock) \
1098 (*(afs_cacheType->vwrite))(avc, uio, aio, acred, nolock)
1100 #define afs_rdwr(avc, uio, rw, io, cred) \
1101 (((rw) == UIO_WRITE) ? afs_write(avc, uio, io, cred, 0) : afs_read(avc, uio, cred, 0, 0, 0))
1102 #define afs_nlrdwr(avc, uio, rw, io, cred) \
1103 (((rw) == UIO_WRITE) ? afs_write(avc, uio, io, cred, 1) : afs_read(avc, uio, cred, 0, 0, 1))
1105 /* Cache size truncation uses the following low and high water marks:
1106 * If the cache is more than 95% full (CM_DCACHECOUNTFREEPCT), the cache
1107 * truncation daemon is awakened and will free up space until the cache is 85%
1108 * (CM_DCACHESPACEFREEPCT - CM_DCACHEEXTRAPCT) full.
1109 * afs_UFSWrite and afs_GetDCache (when it needs to fetch data) will wait on
1110 * afs_WaitForCacheDrain if the cache is 98% (CM_WAITFORDRAINPCT) full.
1111 * afs_GetDownD wakes those processes once the cache is 95% full
1112 * (CM_CACHESIZEDRAINEDPCT).
1114 #define CM_MAXDISCARDEDCHUNKS 16 /* # of chunks */
1115 #define CM_DCACHECOUNTFREEPCT 95 /* max pct of chunks in use */
1116 #define CM_DCACHESPACEFREEPCT 90 /* max pct of space in use */
1117 #define CM_DCACHEEXTRAPCT 5 /* extra to get when freeing */
1118 #define CM_CACHESIZEDRAINEDPCT 95 /* wakeup processes when down to here. */
1119 #define CM_WAITFORDRAINPCT 98 /* sleep if cache is this full. */
1121 #define afs_CacheIsTooFull() \
1122 (afs_blocksUsed - afs_blocksDiscarded > \
1123 (CM_DCACHECOUNTFREEPCT*afs_cacheBlocks)/100 || \
1124 afs_freeDCCount - afs_discardDCCount < \
1125 ((100-CM_DCACHECOUNTFREEPCT)*afs_cacheFiles)/100)
1127 /* Handy max length of a numeric string. */
1128 #define CVBS 12 /* max afs_int32 is 2^32 ~ 4*10^9, +1 for NULL, +luck */
1130 #define refpanic(foo) if (afs_norefpanic) \
1131 { printf( foo ); afs_norefpanic++;} else osi_Panic( foo )
1134 ** these are defined in the AIX source code sys/fs_locks.h but are not
1135 ** defined anywhere in the /usr/include directory
1137 #if defined(AFS_AIX41_ENV)
1138 #define VN_LOCK(vp) simple_lock(&(vp)->v_lock)
1139 #define VN_UNLOCK(vp) simple_unlock(&(vp)->v_lock)
1142 /* get a file's serial number from a vnode */
1143 #ifndef afs_vnodeToInumber
1144 #if defined(AFS_SGI62_ENV) || defined(AFS_HAVE_VXFS) || defined(AFS_DARWIN_ENV)
1145 #define afs_vnodeToInumber(V) VnodeToIno(V)
1147 #ifdef AFS_DECOSF_ENV
1148 #define afs_vnodeToInumber(V) osi_vnodeToInumber(V)
1150 #define afs_vnodeToInumber(V) (VTOI(V)->i_number)
1151 #endif /* AFS_DECOSF_ENV */
1152 #endif /* AFS_SGI62_ENV */
1155 /* get a file's device number from a vnode */
1156 #ifndef afs_vnodeToDev
1157 #if defined(AFS_SGI62_ENV) || defined(AFS_HAVE_VXFS) || defined(AFS_DARWIN_ENV)
1158 #define afs_vnodeToDev(V) VnodeToDev(V)
1159 #elif defined(AFS_DECOSF_ENV)
1160 #define afs_vnodeToDev(V) osi_vnodeToDev(V)
1162 #define afs_vnodeToDev(V) (VTOI(V)->i_dev)
1167 /* Note: this should agree with the definition in kdump.c */
1168 #if defined(AFS_OSF_ENV)
1169 #if !defined(UKERNEL)
1170 #define AFS_USEBUFFERS 1
1174 #if !defined(UKERNEL) && !defined(HAVE_STRUCT_BUF)
1175 /* declare something so that prototypes don't flip out */
1176 /* appears struct buf stuff is only actually passed around as a pointer,
1177 except with libuafs, in which case it is actually defined */
1182 /* fakestat support: opaque storage for afs_EvalFakeStat to remember
1183 * what vcache should be released.
1185 struct afs_fakestat_state {
1189 struct vcache *root_vp;
1192 extern int afs_fakestat_enable;
1195 ino_t fid[1]; /* Unique cache key + i/o addressing */
1197 afs_int32 accesstime;
1198 struct buffer *hashNext;
1206 afs_rwlock_t lock; /* the lock for this structure */
1209 /* afs_memcache.c */
1210 struct memCacheEntry {
1211 int size; /* # of valid bytes in this entry */
1212 int dataSize; /* size of allocated data area */
1213 afs_lock_t afs_memLock;
1214 char *data; /* bytes */
1217 /* First 32 bits of capabilities */
1218 #define CAPABILITY_ERRORTRANS (1<<0)
1220 #define CAPABILITY_BITS 1
1222 #endif /* _AFS_H_ */