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
95 /* flags to use with AFSOP_CACHEINIT */
96 #define AFSCALL_INIT_MEMCACHE 0x1 /* use a memory-based cache */
98 /* below here used only for kernel procedures */
100 /* Store synchrony flags - SYNC means that data should be forced to server's
101 * disk immediately upon completion. */
104 #define AFS_VMSYNC_INVAL 2 /* sync and invalidate pages */
105 #define AFS_LASTSTORE 4
108 /* background request structure */
111 #define BOP_NOOP 0 /* leave 0 unused */
112 #define BOP_FETCH 1 /* parm1 is chunk to get */
113 #define BOP_STORE 2 /* parm1 is chunk to store */
114 #define BOP_PATH 3 /* parm1 is path, parm2 is chunk to fetch */
116 #define B_DONTWAIT 1 /* On failure return; don't wait */
118 /* protocol is: refCount is incremented by user to take block out of free pool.
119 Next, BSTARTED is set when daemon finds request. This prevents
120 other daemons from picking up the same request. Finally, when
121 request is done, refCount is zeroed. BDONE and BWAIT are used by
122 dudes waiting for operation to proceed to a certain point before returning.
124 #define BSTARTED 1 /* request picked up by a daemon */
125 #define BUVALID 2 /* code is valid (store) */
126 #define BUWAIT 4 /* someone is waiting for BUVALID */
128 struct vcache *vnode; /* vnode to use, with vrefcount bumped */
129 struct AFS_UCRED *cred; /* credentials to use for operation */
130 afs_size_t size_parm[BPARMS]; /* random parameters */
131 void *ptr_parm[BPARMS]; /* pointer parameters */
132 afs_int32 code; /* return code */
133 short refCount; /* use counter for this structure */
134 char opcode; /* what to do (store, fetch, etc) */
135 char flags; /* free, etc */
136 afs_int32 ts; /* counter "timestamp" */
144 afs_int32 AuthHandle;
145 char HandShakeKey[8];
147 afs_int32 BeginTimestamp;
148 afs_int32 EndTimestamp;
152 afs_int32 Cell; /* better sun packing if at end of structure */
156 /* 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!) */
159 afs_int32 CellAndUnique;
162 /* The actual number of bytes in the SmallFid, not the sizeof struct. */
163 #define SIZEOF_SMALLFID 10
167 * Queues implemented with both pointers and short offsets into a disk file.
175 afs_int32 uid; /* user id making the request */
176 afs_int32 busyCount; /* how many busies we've seen so far */
177 afs_int32 flags; /* things like O_SYNC, O_NONBLOCK go here */
178 char initd; /* if non-zero, non-uid fields meaningful */
179 char accessError; /* flags for overriding error return code */
180 char volumeError; /* encountered a missing or busy volume */
181 char networkError; /* encountered network problems */
182 char permWriteError; /* fileserver returns permenent error. */
187 /* structure linked off of a server to keep track of queued returned
188 * callbacks. Sent asynchronously when we run a little low on free dudes.
191 struct afs_cbr *next;
196 #define LOCALCELL 1 /* local cell's number is always 1 */
199 #define CPrimary 1 /* on if this is the primary cell */
200 #define CNoSUID 2 /* 1 if no suid progs can run from this cell */
201 #define CHasVolRef 16 /* Volumes were referenced in this cell*/
202 #define CLinkedCell 32
203 #define CAlias 64 /* This cell entry is an alias */
206 struct afs_q lruq; /* lru q next and prev */
207 afs_int32 cell; /* unique id assigned by venus */
208 char *cellName; /* char string name of cell */
209 struct server *cellHosts[MAXCELLHOSTS]; /* volume *location* hosts for this cell */
210 struct cell *lcellp; /* Associated linked cell */
211 u_short fsport; /* file server port */
212 u_short vlport; /* volume server port */
213 short states; /* state flags */
214 short cellIndex; /* relative index number per cell */
215 short realcellIndex; /* as above but ignoring aliases */
216 time_t timeout; /* data expire time, if non-zero */
217 char *realName; /* who this cell is an alias for */
220 #define afs_PutCell(cellp, locktype)
222 /* the unixuser flag bit definitions */
223 #define UHasTokens 1 /* are the st and ct fields valid (ever set)? */
224 #define UTokensBad 2 /* are tokens bad? */
225 #define UPrimary 4 /* on iff primary identity */
226 #define UNeedsReset 8 /* needs afs_ResetAccessCache call done */
227 #define UPAGCounted 16 /* entry seen during PAG search (for stats) */
228 /* A flag used by afs_GCPAGs to keep track of
229 * which entries in afs_users need to be deleted.
230 * The lifetime of its presence in the table is the
231 * lifetime of the afs_GCPAGs function.
233 #define TMP_UPAGNotReferenced 128
235 /* values for afs_gcpags */
237 { AFS_GCPAGS_NOTCOMPILED = 0
239 , AFS_GCPAGS_USERDISABLED
243 , AFS_GCPAGS_EINEXACT
244 , AFS_GCPAGS_EPROCEND
245 , AFS_GCPAGS_EPROCWALK
246 , AFS_GCPAGS_ECREDWALK
247 , AFS_GCPAGS_EPIDCHECK
248 , AFS_GCPAGS_ENICECHECK
251 extern afs_int32 afs_gcpags;
252 extern afs_int32 afs_gcpags_procsize;
255 struct unixuser *next; /* next hash pointer */
256 afs_int32 uid; /* search based on uid and cell */
258 afs_int32 vid; /* corresponding vice id in specified cell */
259 short refCount; /* reference count for allocation */
260 char states; /* flag info */
261 afs_int32 tokenTime; /* last time tokens were set, used for timing out conn data */
262 afs_int32 stLen; /* ticket length (if kerberos, includes kvno at head) */
263 char *stp; /* pointer to ticket itself */
264 struct ClearToken ct;
265 struct afs_exporter *exporter; /* more info about the exporter for the remote user */
269 /* Per-connection block. */
270 struct conn *next; /* Next dude same server. */
271 struct unixuser *user; /* user validated with respect to. */
272 struct rx_connection *id; /* RPC connid. */
273 struct srvAddr *srvr; /* server associated with this conn */
274 short refCount; /* reference count for allocation */
275 unsigned short port; /* port associated with this connection */
276 char forceConnectFS; /* Should we try again with these tokens? */
282 /* Fid comparison routines */
283 #define FidCmp(a,b) ((a)->Fid.Unique != (b)->Fid.Unique \
284 || (a)->Fid.Vnode != (b)->Fid.Vnode \
285 || (a)->Fid.Volume != (b)->Fid.Volume \
286 || (a)->Cell != (b)->Cell)
288 #define FidMatches(afid,tvc) ((tvc)->fid.Fid.Vnode == (afid)->Fid.Vnode && \
289 (tvc)->fid.Fid.Volume == (afid)->Fid.Volume && \
290 (tvc)->fid.Cell == (afid)->Cell && \
291 ( (tvc)->fid.Fid.Unique == (afid)->Fid.Unique || \
292 (!(afid)->Fid.Unique && ((tvc)->states & CUnique))))
296 * Operations on circular queues implemented with pointers. Note: these queue
297 * objects are always located at the beginning of the structures they are linking.
299 #define QInit(q) ((q)->prev = (q)->next = (q))
300 #define QAdd(q,e) ((e)->next = (q)->next, (e)->prev = (q), \
301 (q)->next->prev = (e), (q)->next = (e))
302 #define QRemove(e) ((e)->next->prev = (e)->prev, (e)->prev->next = (e)->next)
303 #define QNext(e) ((e)->next)
304 #define QPrev(e) ((e)->prev)
305 #define QEmpty(q) ((q)->prev == (q))
306 /* this one takes q1 and sticks it on the end of q2 - that is, the other end, not the end
307 * that things are added onto. q1 shouldn't be empty, it's silly */
308 #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))
310 * Do lots of address arithmetic to go from vlruq to the base of the vcache
311 * structure. Don't move struct vnode, since we think of a struct vcache as
312 * a specialization of a struct vnode
314 #define QTOV(e) ((struct vcache *)(((char *) (e)) - (((char *)(&(((struct vcache *)(e))->vlruq))) - ((char *)(e)))))
315 #define QTOC(e) ((struct cell *)((char *) (e)))
318 #define SRVADDR_ISDOWN 0x20 /* same as SRVR_ISDOWN */
319 #define SRVADDR_NOUSE 0x40 /* Don't use this srvAddr */
321 struct srvAddr *next_bkt; /* next item in hash bucket */
322 struct srvAddr *next_sa; /* another interface on same host */
323 struct server *server; /* back to parent */
324 struct conn *conns; /* All user connections to this server */
325 afs_int32 sa_ip; /* Host addr in network byte order */
326 u_short sa_iprank; /* indiv ip address priority */
327 u_short sa_portal; /* port addr in network byte order */
332 * Values used in the flags field of the server structure below.
334 * AFS_SERVER_FLAG_ACTIVATED Has the server ever had a user connection
335 * associated with it?
337 #define AFS_SERVER_FLAG_ACTIVATED 0x01
338 #define SNO_LHOSTS 0x04
339 #define SYES_LHOSTS 0x08
340 #define SVLSRV_UUID 0x10
341 #define SRVR_ISDOWN 0x20
342 #define SRVR_MULTIHOMED 0x40
343 #define SRVR_ISGONE 0x80
344 #define SNO_INLINEBULK 0x100
345 #define SNO_64BIT 0x200
347 #define afs_serverSetNo64Bit(s) ((s)->srvr->server->flags |= SNO_64BIT)
348 #define afs_serverHasNo64Bit(s) ((s)->srvr->server->flags & SNO_64BIT)
354 afs_int32 addr_uniquifier;
358 struct srvAddr haddr;
361 #define sr_uuid _suid._srvUuid.suuid
362 #define sr_addr_uniquifier _suid._srvUuid.addr_uniquifier
363 #define sr_host _suid._srvId.haddr.ip
364 #define sr_portal _suid._srvId.haddr.portal
365 #define sr_rank _suid._srvId.haddr.ip_rank
366 #define sr_flags _suid._srvId.haddr.flags
367 #define sr_conns _suid._srvId.haddr.conns
368 struct server *next; /* Ptr to next server in hash chain */
369 struct cell *cell; /* Cell in which this host resides */
370 struct afs_cbr *cbrs; /* Return list of callbacks */
371 afs_int32 activationTime; /* Time when this record was first activated */
372 afs_int32 lastDowntimeStart; /* Time when last downtime incident began */
373 afs_int32 numDowntimeIncidents; /* # (completed) downtime incidents */
374 afs_int32 sumOfDowntimes; /* Total downtime experienced, in seconds */
375 struct srvAddr *addr;
376 afs_uint32 flags; /* Misc flags*/
379 #define afs_PutServer(servp, locktype)
381 /* structs for some pioctls - these are (or should be)
389 struct sprefrequest_33 {
390 unsigned short offset;
391 unsigned short num_servers;
395 struct sprefrequest { /* new struct for 3.4 */
396 unsigned short offset;
397 unsigned short num_servers;
398 unsigned short flags;
403 unsigned short next_offset;
404 unsigned short num_servers;
405 struct spref servers[1]; /* we overrun this array intentionally...*/
409 unsigned short flags;
410 unsigned short num_servers;
411 struct spref servers[1]; /* we overrun this array intentionally...*/
413 /* struct for GAG pioctl
416 afs_uint32 showflags, logflags, logwritethruflag, spare[3];
417 unsigned char spare2[128];
421 #define logwritethruON 1
424 afs_int32 rx_initReceiveWindow, rx_maxReceiveWindow,
425 rx_initSendWindow, rx_maxSendWindow, rxi_nSendFrags,
426 rxi_nRecvFrags, rxi_OrphanFragSize;
427 afs_int32 rx_maxReceiveSize, rx_MyMaxSendSize;
428 afs_uint32 spare[21];
431 /* struct for checkservers */
444 /* state bits for volume */
445 #define VRO 1 /* volume is readonly */
446 #define VRecheck 2 /* recheck volume info with server */
447 #define VBackup 4 /* is this a backup volume? */
448 #define VForeign 8 /* this is a non-afs volume */
449 #define VResort 16 /* server order was rearranged, sort when able */
450 #define VMoreReps 32 /* This volume has more replicas than we are */
451 /* keeping track of now -- check with VLDB */
453 enum repstate { not_busy, end_not_busy = 6, rd_busy, rdwr_busy, offline };
456 /* One structure per volume, describing where the volume is located
457 and where its mount points are. */
458 struct volume *next; /* Next volume in hash list. */
459 afs_int32 cell; /* the cell in which the volume resides */
460 afs_rwlock_t lock; /* the lock for this structure */
461 afs_int32 volume; /* This volume's ID number. */
462 char *name; /* This volume's name, or 0 if unknown */
463 struct server *serverHost[MAXHOSTS]; /* servers serving this volume */
464 enum repstate status[MAXHOSTS]; /* busy, offline, etc */
465 struct VenusFid dotdot; /* dir to access as .. */
466 struct VenusFid mtpoint; /* The mount point for this volume. */
467 afs_int32 rootVnode, rootUnique; /* Volume's root fid */
470 afs_int32 rwVol; /* For r/o vols, original read/write volume. */
471 afs_int32 accessTime; /* last time we used it */
472 afs_int32 vtix; /* volume table index */
473 afs_int32 copyDate; /* copyDate field, for tracking vol releases */
474 afs_int32 expireTime; /* for per-volume callbacks... */
475 short refCount; /* reference count for allocation */
476 char states; /* here for alignment reasons */
479 #define afs_PutVolume(av, locktype) ((av)->refCount--)
481 /* format of an entry in volume info file */
483 afs_int32 cell; /* cell for this entry */
484 afs_int32 volume; /* volume */
485 afs_int32 next; /* has index */
486 struct VenusFid dotdot; /* .. value */
487 struct VenusFid mtpoint; /* mt point's fid */
488 afs_int32 rootVnode, rootUnique; /* Volume's root fid */
492 struct SimpleLocks *next;
494 afs_int32 boff, eoff;
496 #if defined(AFS_AIX32_ENV) || defined(AFS_SUN5_ENV) || defined(AFS_SGI_ENV)
501 /* vcache state bits */
502 #define CStatd 0x00000001 /* has this file ever been stat'd? */
503 #define CBackup 0x00000002 /* file is on a backup volume */
504 #define CRO 0x00000004 /* is it on a read-only volume */
505 #define CMValid 0x00000008 /* is the mount point info valid? */
506 #define CCore 0x00000010 /* storing a core file, needed since we don't get an open */
507 #define CDirty 0x00000020 /* file has been modified since first open (... O_RDWR) */
508 #define CSafeStore 0x00000040 /* close must wait for store to finish (should be in fd) */
509 #define CMAPPED 0x00000080 /* Mapped files; primarily used by SunOS 4.0.x */
510 #define CNSHARE 0x00000100 /* support O_NSHARE semantics */
511 #define CLied 0x00000200
512 #define CTruth 0x00000400
514 #define CWired 0x00000800 /* OSF hack only */
516 #ifdef AFS_DARWIN_ENV
517 #define CUBCinit 0x00000800
519 #define CWRITE_IGN 0x00000800 /* Next OS hack only */
522 #define CUnique 0x00001000 /* vc's uniquifier - latest unifiquier for fid */
523 #define CForeign 0x00002000 /* this is a non-afs vcache */
524 #define CUnlinked 0x00010000
525 #define CBulkStat 0x00020000 /* loaded by a bulk stat, and not ref'd since */
526 #define CUnlinkedDel 0x00040000
527 #define CVFlushed 0x00080000
528 #define CCore1 0x00100000 /* osf1 core file; not same as CCore above */
529 #define CWritingUFS 0x00200000 /* to detect vm deadlock - used by sgi */
530 #define CCreating 0x00400000 /* avoid needless store after open truncate */
531 #define CPageHog 0x00800000 /* AIX - dumping large cores is a page hog. */
532 #define CDCLock 0x02000000 /* Vnode lock held over call to GetDownD */
533 #define CBulkFetching 0x04000000 /* stats are being fetched by bulk stat */
534 #define CExtendedFile 0x08000000 /* extended file via ftruncate call. */
536 /* vcache vstate bits */
537 #define VRevokeWait 0x1
538 #define VPageCleaning 0x2 /* Solaris - Cache Trunc Daemon sez keep out */
541 #if !defined(AFS_FBSD_ENV)
542 #define vrefCount v.v_count
544 #define vrefCount v.v_usecount
545 #endif /* AFS_FBSD_ENV */
547 #ifdef AFS_LINUX24_ENV
548 #define VREFCOUNT(v) atomic_read(&((vnode_t *) v)->v_count)
549 #define VREFCOUNT_SET(v, c) atomic_set(&((vnode_t *) v)->v_count, c)
550 #define VREFCOUNT_DEC(v) atomic_dec(&((vnode_t *) v)->v_count)
551 #define VREFCOUNT_INC(v) atomic_inc(&((vnode_t *) v)->v_count)
552 #define DLOCK() spin_lock(&dcache_lock)
553 #define DUNLOCK() spin_unlock(&dcache_lock)
554 #define DGET(d) dget_locked(d)
555 #define DCOUNT(d) atomic_read(&(d)->d_count)
557 #define VREFCOUNT(v) ((v)->vrefCount)
558 #define VREFCOUNT_SET(v, c) (v)->vrefCount = c;
559 #define VREFCOUNT_DEC(v) (v)->vrefCount--;
560 #define VREFCOUNT_INC(v) (v)->vrefCount++;
563 #define DGET(d) dget(d)
564 #define DCOUNT(d) ((d)->d_count)
567 #define AFS_MAXDV 0x7fffffff /* largest dataversion number */
568 #ifdef AFS_64BIT_CLIENT
569 #define AFS_NOTRUNC 0x7fffffffffffffffLL /* largest positive int64 number */
570 #else /* AFS_64BIT_CLIENT */
571 #define AFS_NOTRUNC 0x7fffffff /* largest dataversion number */
572 #endif /* AFS_64BIT_CLIENT */
574 extern afs_int32 vmPageHog; /* counter for # of vnodes which are page hogs. */
577 * Fast map from vcache to dcache
584 afs_offs_t minLoc; /* smallest offset into dc. */
585 afs_offs_t len; /* largest offset into dc. */
588 extern afs_uint32 afs_stampValue; /* stamp for pair's usage */
589 #define MakeStamp() (++afs_stampValue)
591 #define VTOAFS(V) ((struct vcache*)(V))
592 #define AFSTOV(V) (&(V)->v)
593 #ifdef AFS_LINUX22_ENV
594 #define ITOAFS(V) ((struct vcache*)(V))
595 #define AFSTOI(V) (struct inode *)(&(V)->v)
598 /* INVARIANTs: (vlruq.next != NULL) == (vlruq.prev != NULL)
599 * nextfree => !vlruq.next && ! vlruq.prev
600 * !(avc->nextfree) && !avc->vlruq.next => (FreeVCList == avc->nextfree)
603 struct vnode v; /* Has reference count in v.v_count */
604 struct afs_q vlruq; /* lru q next and prev */
605 struct vcache *nextfree; /* next on free list (if free) */
606 struct vcache *hnext; /* Hash next */
610 afs_hyper_t DataVersion;
614 ushort Mode; /* XXXX Should be afs_int32 XXXX */
616 /* vnode type is in v.v_type */
618 afs_rwlock_t lock; /* The lock on the vcache contents. */
619 #if defined(AFS_SUN5_ENV)
620 /* Lock used to protect the activeV, multipage, and vstates fields.
621 * Do not try to get the vcache lock when the vlock is held */
623 #endif /* defined(AFS_SUN5_ENV) */
624 #if defined(AFS_SUN5_ENV)
628 #if defined(AFS_SUN_ENV) || defined(AFS_ALPHA_ENV) || defined(AFS_DARWIN_ENV) || defined(AFS_FBSD_ENV)
629 afs_bozoLock_t pvnLock; /* see locks.x */
638 int ownslock; /* pid of owner of excl lock, else 0 - defect 3083 */
640 #ifdef AFS_DARWIN_ENV
641 struct lock__bsd__ rwlock;
646 afs_int32 parentVnode; /* Parent dir, if a file. */
647 afs_int32 parentUnique;
648 struct VenusFid *mvid; /* Either parent dir (if root) or root (if mt pt) */
649 char *linkData; /* Link data if a symlink. */
650 afs_hyper_t flushDV; /* data version last flushed from text */
651 afs_hyper_t mapDV; /* data version last flushed from map */
652 afs_size_t truncPos; /* truncate file to this position at next store */
653 struct server *callback; /* The callback host, if any */
654 afs_uint32 cbExpires; /* time the callback expires */
655 struct afs_q callsort; /* queue in expiry order, sort of */
656 struct axscache *Access; /* a list of cached access bits */
657 afs_int32 anyAccess; /* System:AnyUser's access to this. */
658 afs_int32 last_looker; /* pag/uid from last lookup here */
659 #if defined(AFS_SUN5_ENV)
661 #endif /* defined(AFS_SUN5_ENV) */
662 struct SimpleLocks *slocks;
663 short opens; /* The numbers of opens, read or write, on this file. */
664 short execsOrWriters; /* The number of execs (if < 0) or writers (if > 0) of
666 short flockCount; /* count of flock readers, or -1 if writer */
667 char mvstat; /* 0->normal, 1->mt pt, 2->root. */
668 afs_uint32 states; /* state bits */
669 #if defined(AFS_SUN5_ENV)
670 afs_uint32 vstates; /* vstate bits */
671 #endif /* defined(AFS_SUN5_ENV) */
673 afs_uint32 symhintstamp;
675 struct vcache *symhint;
676 struct dcache *dchint;
678 #ifdef AFS_LINUX22_ENV
679 u_short flushcnt; /* Number of flushes which haven't released yet. */
680 u_short mapcnt; /* Number of mappings of this file. */
682 #if defined(AFS_SGI_ENV)
683 daddr_t lastr; /* for read-ahead */
685 uint64_t vc_rwlockid; /* kthread owning rwlock */
687 short vc_rwlockid; /* pid of process owning rwlock */
689 short vc_locktrips; /* # of rwlock reacquisitions */
690 sema_t vc_rwlock; /* vop_rwlock for afs */
691 pgno_t mapcnt; /* # of pages mapped */
692 struct cred *cred; /* last writer's cred */
694 struct bhv_desc vc_bhv_desc; /* vnode's behavior data. */
696 #endif /* AFS_SGI_ENV */
697 afs_int32 vc_error; /* stash write error for this vnode. */
698 int xlatordv; /* Used by nfs xlator */
699 struct AFS_UCRED *uncred;
700 int asynchrony; /* num kbytes to store behind */
702 short multiPage; /* count of multi-page getpages in progress */
706 #define afs_symhint_inval(avc)
709 #define DONT_CHECK_MODE_BITS 0
710 #define CHECK_MODE_BITS 1
711 #define CMB_ALLOW_EXEC_AS_READ 2 /* For the NFS xlator */
713 #if defined(AFS_SGI_ENV)
714 #define AVCRWLOCK(avc) (valusema(&(avc)->vc_rwlock) <= 0)
716 /* SGI vnode rwlock macros and flags. */
717 #ifndef AFS_SGI62_ENV
718 /* The following are defined here. SGI 6.2 declares them in vnode.h */
719 #define VRWLOCK_READ 0
720 #define VRWLOCK_WRITE 1
721 #define VRWLOCK_WRITE_DIRECT 2
726 #define AFS_RWLOCK_T vrwlock_t
728 #define AFS_RWLOCK_T int
729 #endif /* AFS_SGI62_ENV */
731 #include <ksys/behavior.h>
732 #define AFS_RWLOCK(V,F) \
733 afs_rwlock(&VTOAFS(V)->vc_bhv_desc, (F));
734 #define AFS_RWUNLOCK(V,F) \
735 afs_rwunlock(&VTOAFS(V)->vc_bhv_desc, (F));
738 #define AFS_RWLOCK(V,F) afs_rwlock((vnode_t *)(V), (F) )
739 #define AFS_RWUNLOCK(V,F) afs_rwunlock((vnode_t *)(V), (F) )
741 #else /* AFS_SGI53_ENV */
742 #define AFS_RWLOCK(V,F) afs_rwlock((V))
743 #define AFS_RWUNLOCK(V,F) afs_rwunlock((V))
744 #endif /* AFS_SGI53_ENV */
745 #endif /* AFS_SGI_ENV */
749 afs_hyper_t DataVersion;
751 afs_int32 parentVnode;
752 afs_int32 parentUnique;
756 afs_int32 randomUid[CPSIZE];
757 afs_int32 callback; /* Now a pointer to 'server' struct */
759 afs_int32 randomAccess[CPSIZE];
762 short execsOrWriters;
773 /* CM inititialization parameters. What CM actually used after calculations
774 * based on passed in arguments.
776 #define CMI_VERSION 1 /* increment when adding new fields. */
777 struct cm_initparams {
782 int cmi_nVolumeCaches;
783 int cmi_firstChunkSize;
784 int cmi_otherChunkSize;
785 int cmi_cacheSize; /* The original cache size, in 1K blocks. */
786 unsigned cmi_setTime:1;
787 unsigned cmi_memCache:1;
788 int spare[16-9]; /* size of struct is 16 * 4 = 64 bytes */
792 /*----------------------------------------------------------------------
793 * AFS Data cache definitions
795 * Each entry describes a Unix file on the local disk that is
796 * is serving as a cached copy of all or part of a Vice file.
797 * Entries live in circular queues for each hash table slot
799 * Which queue is this thing in? Good question.
800 * A struct dcache entry is in the freeDSlot queue when not associated with a cache slot (file).
801 * Otherwise, it is in the DLRU queue. The freeDSlot queue uses the lruq.next field as
802 * its "next" pointer.
804 * Cache entries in the DLRU queue are either associated with vice files, in which case
805 * they are hashed by afs_dvnextTbl and afs_dcnextTbl pointers, or they are in the freeDCList
806 * and are not associated with any vice file. This last list uses the afs_dvnextTbl pointer for
807 * its "next" pointer.
808 *----------------------------------------------------------------------*/
810 #define NULLIDX (-1) /* null index definition */
811 /* struct dcache states bits */
812 #define DWriting 8 /* file being written (used for cache validation) */
814 /* dcache data flags */
815 #define DFEntryMod 0x02 /* has entry itself been modified? */
816 #define DFFetching 0x04 /* file is currently being fetched */
818 /* dcache meta flags */
819 #define DFNextStarted 0x01 /* next chunk has been prefetched already */
820 #define DFFetchReq 0x10 /* someone is waiting for DFFetching to go on */
823 /* flags in afs_indexFlags array */
824 #define IFEverUsed 1 /* index entry has >= 1 byte of data */
825 #define IFFree 2 /* index entry in freeDCList */
826 #define IFDataMod 4 /* file needs to be written out */
827 #define IFFlag 8 /* utility flag */
828 #define IFDirtyPages 16
829 #define IFAnyPages 32
830 #define IFDiscarded 64 /* index entry in discardDCList */
833 char *in; /* input buffer */
834 char *out; /* output buffer */
835 short in_size; /* Size of input buffer <= 2K */
836 short out_size; /* Maximum size of output buffer, <= 2K */
840 * This version of afs_ioctl is required to pass in 32 bit user space
841 * pointers into a 64 bit kernel.
852 /* CacheItems file has a header of type struct afs_fheader
853 * (keep aligned properly). Since we already have sgi_62 clients running
854 * with a 32 bit inode, a change is required to the header so that
855 * they can distinguish the old 32 bit inode CacheItems file and zap it
856 * instead of using it.
859 #define AFS_FHMAGIC 0x7635abaf /* uses version number */
861 #if defined(AFS_SUN57_64BIT_ENV)
862 #define AFS_CI_VERSION 3
864 #define AFS_CI_VERSION 2
867 afs_int32 firstCSize;
868 afs_int32 otherCSize;
871 /* kept on disk and in dcache entries */
873 struct VenusFid fid; /* Fid for this file */
874 afs_int32 modTime; /* last time this entry was modified */
875 afs_hyper_t versionNo; /* Associated data version number */
876 afs_int32 chunk; /* Relative chunk number */
877 #if defined(AFS_SGI61_ENV) || defined(AFS_SUN57_64BIT_ENV)
878 /* Using ino64_t here so that user level debugging programs compile
879 * the size correctly.
881 ino64_t inode; /* Unix inode for this chunk */
883 #if defined(AFS_LINUX_64BIT_KERNEL)
884 long inode; /* Unix inode for this chunk */
886 afs_int32 inode; /* Unix inode for this chunk */
889 afs_int32 chunkBytes; /* Num bytes in this chunk */
890 char states; /* Has this chunk been modified? */
893 /* magic numbers to specify the cache type */
895 #define AFS_FCACHE_TYPE_UFS 0x0
896 #define AFS_FCACHE_TYPE_MEM 0x1
897 #define AFS_FCACHE_TYPE_NFS 0x2
898 #define AFS_FCACHE_TYPE_EPI 0x3
902 struct afs_q lruq; /* Free queue for in-memory images */
903 afs_rwlock_t lock; /* Protects validPos, some f */
904 afs_rwlock_t tlock; /* Atomizes updates to refCount */
905 afs_rwlock_t mflock; /* Atomizes accesses/updates to mflags */
906 afs_size_t validPos; /* number of valid bytes during fetch */
907 afs_int32 index; /* The index in the CacheInfo file*/
908 short refCount; /* Associated reference count. */
909 char dflags; /* Data flags */
910 char mflags; /* Meta flags */
911 struct fcache f; /* disk image */
912 afs_int32 stamp; /* used with vtodc struct for hints */
917 * dcache.lock protects the actual contents of the cache file (in
918 * f.inode), subfields of f except those noted below, dflags and
921 * dcache.tlock is used to make atomic updates to refCount. Zero
922 * refCount dcache entries are protected by afs_xdcache instead of
925 * dcache.mflock is used to access and update mflags. It cannot be
926 * held without holding the corresponding dcache.lock. Updating
927 * mflags requires holding dcache.lock(R) and dcache.mflock(W), and
928 * checking for mflags requires dcache.lock(R) and dcache.mflock(R).
929 * Note that dcache.lock(W) gives you the right to update mflags,
930 * as dcache.mflock(W) can only be held with dcache.lock(R).
932 * dcache.stamp is protected by the associated vcache lock, because
933 * it's only purpose is to establish correspondence between vcache
934 * and dcache entries.
936 * dcache.index, dcache.f.fid, dcache.f.chunk and dcache.f.inode are
937 * write-protected by afs_xdcache and read-protected by refCount.
938 * Once an entry is referenced, these values cannot change, and if
939 * it's on the free list (with refCount=0), it can be reused for a
940 * different file/chunk. These values can only be written while
941 * holding afs_xdcache(W) and allocating this dcache entry (thereby
942 * ensuring noone else has a refCount on it).
945 /* this is obsolete and should be removed */
948 /* macro to mark a dcache entry as bad */
951 (x)->f.fid.Fid.Unique = 0; \
952 afs_indexUnique[(x)->index] = 0; \
953 (x)->dflags |= DFEntryMod; \
956 /* FakeOpen and Fake Close used to be real subroutines. They're only used in
957 * sun_subr and afs_vnodeops, and they're very frequently called, so I made
958 * them into macros. They do:
959 * FakeOpen: fake the file being open for writing. avc->lock must be held
960 * in write mode. Having the file open for writing is like having a DFS
961 * write-token: you're known to have the best version of the data around,
962 * and so the CM won't let it be overwritten by random server info.
963 * FakeClose: undo the effects of FakeOpen, noting that we want to ensure
964 * that a real close eventually gets done. We use CCore to achieve this if
965 * we would end up closing the file. avc->lock must be held in write mode */
967 #ifdef AFS_AIX_IAUTH_ENV
968 #define CRKEEP(V, C) (V)->linkData = (char*)crdup((C))
970 #define CRKEEP(V, C) crhold((C)); (V)->linkData = (char*)(C)
973 #define afs_FakeOpen(avc) { avc->opens++; avc->execsOrWriters++; }
974 #define afs_FakeClose(avc, acred) \
975 { if (avc->execsOrWriters == 1) { \
976 /* we're the last writer, just use CCore flag */ \
977 avc->states |= CCore; /* causes close to be called later */ \
979 /* The cred and vnode holds will be released in afs_FlushActiveVcaches */ \
980 VN_HOLD(AFSTOV(avc)); /* So it won't disappear */ \
981 CRKEEP(avc, acred); /* Should use a better place for the creds */ \
984 /* we're not the last writer, let the last one do the store-back for us */ \
986 avc->execsOrWriters--; \
990 #define AFS_ZEROS 64 /* zero buffer */
992 /*#define afs_DirtyPages(avc) (((avc)->states & CDirty) || osi_VMDirty_p((avc)))*/
993 #define afs_DirtyPages(avc) ((avc)->states & CDirty)
995 /* The PFlush algorithm makes use of the fact that Fid.Unique is not used in
996 below hash algorithms. Change it if need be so that flushing algorithm
997 doesn't move things from one hash chain to another
999 /* extern int afs_dhashsize; */
1000 #define DCHash(v, c) ((((v)->Fid.Vnode + (v)->Fid.Volume + (c))) & (afs_dhashsize-1))
1001 /*Vnode, Chunk -> Hash table index*/
1002 #define DVHash(v) ((((v)->Fid.Vnode + (v)->Fid.Volume )) & (afs_dhashsize-1))
1003 /*Vnode -> Other hash table index*/
1004 /* don't hash on the cell, our callback-breaking code sometimes fails to compute
1005 the cell correctly, and only scans one hash bucket */
1006 #define VCHash(fid) (((fid)->Fid.Volume + (fid)->Fid.Vnode) & (VCSIZE-1))
1008 extern struct dcache **afs_indexTable; /*Pointers to in-memory dcache entries*/
1009 extern afs_int32 *afs_indexUnique; /*dcache entry Fid.Unique */
1010 extern afs_int32 *afs_dvnextTbl; /*Dcache hash table links */
1011 extern afs_int32 *afs_dcnextTbl; /*Dcache hash table links */
1012 extern afs_int32 afs_cacheFiles; /*Size of afs_indexTable*/
1013 extern afs_int32 afs_cacheBlocks; /*1K blocks in cache*/
1014 extern afs_int32 afs_cacheStats; /*Stat entries in cache*/
1015 extern struct vcache *afs_vhashT[VCSIZE]; /*Stat cache hash table*/
1016 extern afs_int32 afs_initState; /*Initialization state*/
1017 extern afs_int32 afs_termState; /* Termination state */
1018 extern struct VenusFid afs_rootFid; /*Root for whole file system*/
1019 extern afs_int32 afs_allCBs; /* Count of callbacks*/
1020 extern afs_int32 afs_oddCBs; /* Count of odd callbacks*/
1021 extern afs_int32 afs_evenCBs; /* Count of even callbacks*/
1022 extern afs_int32 afs_allZaps; /* Count of fid deletes */
1023 extern afs_int32 afs_oddZaps; /* Count of odd fid deletes */
1024 extern afs_int32 afs_evenZaps; /* Count of even fid deletes */
1025 extern struct brequest afs_brs[NBRS]; /* request structures */
1027 #define UHash(auid) ((auid) & (NUSERS-1))
1028 #define VHash(avol) ((avol)&(NVOLS-1))
1029 #define SHash(aserv) ((ntohl(aserv)) & (NSERVERS-1))
1030 #define FVHash(acell,avol) (((avol)+(acell)) & (NFENTRIES-1))
1032 /* Performance hack - we could replace VerifyVCache2 with the appropriate
1033 * GetVCache incantation, and could eliminate even this code from afs_UFSRead
1034 * by making intentionally invalidating quick.stamp in the various callbacks
1035 * expiration/breaking code */
1036 #ifdef AFS_LINUX20_ENV
1037 #define afs_VerifyVCache(avc, areq) \
1038 (((avc)->states & CStatd) ? (vcache2inode(avc), 0) : \
1039 afs_VerifyVCache2((avc),areq))
1041 #ifdef AFS_DARWIN_ENV
1042 #define afs_VerifyVCache(avc, areq) \
1043 (((avc)->states & CStatd) ? (osi_VM_Setup(avc, 0), 0) : \
1044 afs_VerifyVCache2((avc),areq))
1046 #define afs_VerifyVCache(avc, areq) \
1047 (((avc)->states & CStatd) ? 0 : afs_VerifyVCache2((avc),areq))
1051 #define DO_STATS 1 /* bits used by FindVCache */
1054 /* values for flag param of afs_CheckVolumeNames */
1055 #define AFS_VOLCHECK_EXPIRED 0x1 /* volumes whose callbacks have expired */
1056 #define AFS_VOLCHECK_BUSY 0x2 /* volumes which were marked busy */
1057 #define AFS_VOLCHECK_MTPTS 0x4 /* mount point invalidation also */
1058 #define AFS_VOLCHECK_FORCE 0x8 /* do all forcibly */
1062 #define AFS_FSPORT ((unsigned short) htons(7000))
1063 #define AFS_VLPORT ((unsigned short) htons(7003))
1065 #define afs_read(avc, uio, acred, albn, abpp, nolock) \
1066 (*(afs_cacheType->vread))(avc, uio, acred, albn, abpp, nolock)
1067 #define afs_write(avc, uio, aio, acred, nolock) \
1068 (*(afs_cacheType->vwrite))(avc, uio, aio, acred, nolock)
1070 #define afs_rdwr(avc, uio, rw, io, cred) \
1071 (((rw) == UIO_WRITE) ? afs_write(avc, uio, io, cred, 0) : afs_read(avc, uio, cred, 0, 0, 0))
1072 #define afs_nlrdwr(avc, uio, rw, io, cred) \
1073 (((rw) == UIO_WRITE) ? afs_write(avc, uio, io, cred, 1) : afs_read(avc, uio, cred, 0, 0, 1))
1075 /* Cache size truncation uses the following low and high water marks:
1076 * If the cache is more than 95% full (CM_DCACHECOUNTFREEPCT), the cache
1077 * truncation daemon is awakened and will free up space until the cache is 85%
1078 * (CM_DCACHESPACEFREEPCT - CM_DCACHEEXTRAPCT) full.
1079 * afs_UFSWrite and afs_GetDCache (when it needs to fetch data) will wait on
1080 * afs_WaitForCacheDrain if the cache is 98% (CM_WAITFORDRAINPCT) full.
1081 * afs_GetDownD wakes those processes once the cache is 95% full
1082 * (CM_CACHESIZEDRAINEDPCT).
1084 #define CM_MAXDISCARDEDCHUNKS 16 /* # of chunks */
1085 #define CM_DCACHECOUNTFREEPCT 95 /* max pct of chunks in use */
1086 #define CM_DCACHESPACEFREEPCT 90 /* max pct of space in use */
1087 #define CM_DCACHEEXTRAPCT 5 /* extra to get when freeing */
1088 #define CM_CACHESIZEDRAINEDPCT 95 /* wakeup processes when down to here.*/
1089 #define CM_WAITFORDRAINPCT 98 /* sleep if cache is this full. */
1091 #define afs_CacheIsTooFull() \
1092 (afs_blocksUsed - afs_blocksDiscarded > \
1093 (CM_DCACHECOUNTFREEPCT*afs_cacheBlocks)/100 || \
1094 afs_freeDCCount - afs_discardDCCount < \
1095 ((100-CM_DCACHECOUNTFREEPCT)*afs_cacheFiles)/100)
1097 /* Handy max length of a numeric string. */
1098 #define CVBS 12 /* max afs_int32 is 2^32 ~ 4*10^9, +1 for NULL, +luck */
1100 #define refpanic(foo) if (afs_norefpanic) \
1101 { printf( foo ); afs_norefpanic++;} else osi_Panic( foo )
1104 ** these are defined in the AIX source code sys/fs_locks.h but are not
1105 ** defined anywhere in the /usr/include directory
1107 #if defined(AFS_AIX41_ENV)
1108 #define VN_LOCK(vp) simple_lock(&(vp)->v_lock)
1109 #define VN_UNLOCK(vp) simple_unlock(&(vp)->v_lock)
1112 /* get a file's serial number from a vnode */
1113 #ifndef afs_vnodeToInumber
1114 #if defined(AFS_SGI62_ENV) || defined(AFS_HAVE_VXFS) || defined(AFS_DARWIN_ENV)
1115 #define afs_vnodeToInumber(V) VnodeToIno(V)
1117 #ifdef AFS_DECOSF_ENV
1118 #define afs_vnodeToInumber(V) osi_vnodeToInumber(V)
1120 #define afs_vnodeToInumber(V) (VTOI(V)->i_number)
1121 #endif /* AFS_DECOSF_ENV */
1122 #endif /* AFS_SGI62_ENV */
1125 /* get a file's device number from a vnode */
1126 #ifndef afs_vnodeToDev
1127 #if defined(AFS_SGI62_ENV) || defined(AFS_HAVE_VXFS) || defined(AFS_DARWIN_ENV)
1128 #define afs_vnodeToDev(V) VnodeToDev(V)
1130 #ifdef AFS_DECOSF_ENV
1131 #define afs_vnodeToDev(V) osi_vnodeToDev(V)
1133 #define afs_vnodeToDev(V) (VTOI(V)->i_dev)
1134 #endif /* AFS_DECOSF_ENV */
1135 #endif /* AFS_SGI62_ENV */
1139 /* Note: this should agree with the definition in kdump.c */
1140 #if defined(AFS_OSF_ENV)
1141 #if !defined(UKERNEL)
1142 #define AFS_USEBUFFERS 1
1146 #if !defined(UKERNEL) && !defined(HAVE_STRUCT_BUF)
1147 /* declare something so that prototypes don't flip out */
1148 /* appears struct buf stuff is only actually passed around as a pointer,
1149 except with libuafs, in which case it is actually defined */
1154 /* fakestat support: opaque storage for afs_EvalFakeStat to remember
1155 * what vcache should be released.
1157 struct afs_fakestat_state {
1161 struct vcache *root_vp;
1164 extern int afs_fakestat_enable;
1167 ino_t fid[1]; /* Unique cache key + i/o addressing */
1169 afs_int32 accesstime;
1170 struct buffer *hashNext;
1178 afs_rwlock_t lock; /* the lock for this structure */
1181 /* afs_memcache.c */
1182 struct memCacheEntry {
1183 int size; /* # of valid bytes in this entry */
1184 int dataSize; /* size of allocated data area */
1185 afs_lock_t afs_memLock;
1186 char *data; /* bytes */
1189 #endif /* _AFS_H_ */