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 /* The following are various levels of afs debugging */
43 #define AFSDEB_GENERAL 1 /* Standard debugging */
44 #define AFSDEB_NETWORK 2 /* low level afs networking */
45 #define AFSDEB_RX 4 /* RX debugging */
46 #define AFSDEB_VNLAYER 8 /* interface layer to AFS (aixops, gfsops, etc) */
48 /* generic undefined vice id */
51 /* The basic defines for the Andrew file system
52 better keep things powers of two so "& (foo-1)" hack works for masking bits */
53 #define MAXHOSTS 13 /* max hosts per single volume */
54 #define OMAXHOSTS 8 /* backwards compatibility */
55 #define MAXCELLHOSTS 8 /* max vldb servers per cell */
56 #define NBRS 15 /* max number of queued daemon requests */
57 #define NUSERS 16 /* hash table size for unixuser table */
58 #define NSERVERS 16 /* hash table size for server table */
59 #define NVOLS 64 /* hash table size for volume table */
60 #define NFENTRIES 256 /* hash table size for disk volume table */
61 #define VCSIZE 1024 /* stat cache hash table size */
62 #define DCSIZE 512 /* disk cache hash table size */
63 #define PIGGYSIZE 1350 /* max piggyback size */
64 #define MAXVOLS 128 /* max vols we can store */
65 #define MAXSYSNAME 128 /* max sysname (i.e. @sys) size */
66 #define MAXNUMSYSNAMES 16 /* max that current constants allow */
67 #define NOTOKTIMEOUT (2*3600) /* time after which to timeout conns sans tokens */
68 #define NOPAG 0xffffffff
69 #define AFS_NCBRS 300 /* max # of call back return entries */
70 #define AFS_MAXCBRSCALL 16 /* max to return in a given call */
71 #define AFS_SALLOC_LOW_WATER 250 /* Min free blocks before allocating more */
72 #define AFS_LRALLOCSIZ 4096 /* "Large" allocated size */
74 #define AFS_NRXPACKETS 80
75 #define AFS_RXDEADTIME 50
76 #define AFS_HARDDEADTIME 120
84 /* flags to use with AFSOP_CACHEINIT */
85 #define AFSCALL_INIT_MEMCACHE 0x1 /* use a memory-based cache */
87 /* below here used only for kernel procedures */
89 /* Store synchrony flags - SYNC means that data should be forced to server's
90 * disk immediately upon completion. */
93 #define AFS_VMSYNC_INVAL 2 /* sync and invalidate pages */
94 #define AFS_LASTSTORE 4
97 /* background request structure */
100 #define BOP_NOOP 0 /* leave 0 unused */
101 #define BOP_FETCH 1 /* parm1 is chunk to get */
102 #define BOP_STORE 2 /* parm1 is chunk to store */
103 #define BOP_PATH 3 /* parm1 is path, parm2 is chunk to fetch */
105 #define B_DONTWAIT 1 /* On failure return; don't wait */
107 /* protocol is: refCount is incremented by user to take block out of free pool.
108 Next, BSTARTED is set when daemon finds request. This prevents
109 other daemons from picking up the same request. Finally, when
110 request is done, refCount is zeroed. BDONE and BWAIT are used by
111 dudes waiting for operation to proceed to a certain point before returning.
113 #define BSTARTED 1 /* request picked up by a daemon */
114 #define BUVALID 2 /* code is valid (store) */
115 #define BUWAIT 4 /* someone is waiting for BUVALID */
117 struct vcache *vnode; /* vnode to use, with vrefcount bumped */
118 struct AFS_UCRED *cred; /* credentials to use for operation */
119 afs_size_t size_parm[BPARMS]; /* random parameters */
120 void *ptr_parm[BPARMS]; /* pointer parameters */
121 afs_int32 code; /* return code */
122 short refCount; /* use counter for this structure */
123 char opcode; /* what to do (store, fetch, etc) */
124 char flags; /* free, etc */
125 afs_int32 ts; /* counter "timestamp" */
133 afs_int32 AuthHandle;
134 char HandShakeKey[8];
136 afs_int32 BeginTimestamp;
137 afs_int32 EndTimestamp;
141 afs_int32 Cell; /* better sun packing if at end of structure */
145 /* 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!) */
148 afs_int32 CellAndUnique;
151 /* The actual number of bytes in the SmallFid, not the sizeof struct. */
152 #define SIZEOF_SMALLFID 10
156 * Queues implemented with both pointers and short offsets into a disk file.
164 afs_int32 uid; /* user id making the request */
165 afs_int32 busyCount; /* how many busies we've seen so far */
166 afs_int32 flags; /* things like O_SYNC, O_NONBLOCK go here */
167 char initd; /* if non-zero, non-uid fields meaningful */
168 char accessError; /* flags for overriding error return code */
169 char volumeError; /* encountered a missing or busy volume */
170 char networkError; /* encountered network problems */
171 char permWriteError; /* fileserver returns permenent error. */
176 /* structure linked off of a server to keep track of queued returned
177 * callbacks. Sent asynchronously when we run a little low on free dudes.
180 struct afs_cbr *next;
185 #define LOCALCELL 1 /* local cell's number is always 1 */
188 #define CPrimary 1 /* on if this is the primary cell */
189 #define CNoSUID 2 /* 1 if no suid progs can run from this cell */
190 #define CHasVolRef 16 /* Volumes were referenced in this cell*/
191 #define CLinkedCell 32
192 #define CAlias 64 /* This cell entry is an alias */
195 struct afs_q lruq; /* lru q next and prev */
196 afs_int32 cell; /* unique id assigned by venus */
197 char *cellName; /* char string name of cell */
198 struct server *cellHosts[MAXCELLHOSTS]; /* volume *location* hosts for this cell */
199 struct cell *lcellp; /* Associated linked cell */
200 u_short fsport; /* file server port */
201 u_short vlport; /* volume server port */
202 short states; /* state flags */
203 short cellIndex; /* relative index number per cell */
204 time_t timeout; /* data expire time, if non-zero */
205 char *realName; /* who this cell is an alias for */
208 #define afs_PutCell(cellp, locktype)
210 /* the unixuser flag bit definitions */
211 #define UHasTokens 1 /* are the st and ct fields valid (ever set)? */
212 #define UTokensBad 2 /* are tokens bad? */
213 #define UPrimary 4 /* on iff primary identity */
214 #define UNeedsReset 8 /* needs afs_ResetAccessCache call done */
215 #define UPAGCounted 16 /* entry seen during PAG search (for stats) */
216 /* A flag used by afs_GCPAGs to keep track of
217 * which entries in afs_users need to be deleted.
218 * The lifetime of its presence in the table is the
219 * lifetime of the afs_GCPAGs function.
221 #define TMP_UPAGNotReferenced 128
223 /* values for afs_gcpags */
225 { AFS_GCPAGS_NOTCOMPILED = 0
227 , AFS_GCPAGS_USERDISABLED
231 , AFS_GCPAGS_EINEXACT
232 , AFS_GCPAGS_EPROCEND
233 , AFS_GCPAGS_EPROCWALK
234 , AFS_GCPAGS_ECREDWALK
235 , AFS_GCPAGS_EPIDCHECK
236 , AFS_GCPAGS_ENICECHECK
239 extern afs_int32 afs_gcpags;
240 extern afs_int32 afs_gcpags_procsize;
243 struct unixuser *next; /* next hash pointer */
244 afs_int32 uid; /* search based on uid and cell */
246 afs_int32 vid; /* corresponding vice id in specified cell */
247 short refCount; /* reference count for allocation */
248 char states; /* flag info */
249 afs_int32 tokenTime; /* last time tokens were set, used for timing out conn data */
250 afs_int32 stLen; /* ticket length (if kerberos, includes kvno at head) */
251 char *stp; /* pointer to ticket itself */
252 struct ClearToken ct;
253 struct afs_exporter *exporter; /* more info about the exporter for the remote user */
257 /* Per-connection block. */
258 struct conn *next; /* Next dude same server. */
259 struct unixuser *user; /* user validated with respect to. */
260 struct rx_connection *id; /* RPC connid. */
261 struct srvAddr *srvr; /* server associated with this conn */
262 short refCount; /* reference count for allocation */
263 unsigned short port; /* port associated with this connection */
264 char forceConnectFS; /* Should we try again with these tokens? */
270 /* Fid comparison routines */
271 #define FidCmp(a,b) ((a)->Fid.Unique != (b)->Fid.Unique \
272 || (a)->Fid.Vnode != (b)->Fid.Vnode \
273 || (a)->Fid.Volume != (b)->Fid.Volume \
274 || (a)->Cell != (b)->Cell)
276 #define FidMatches(afid,tvc) ((tvc)->fid.Fid.Vnode == (afid)->Fid.Vnode && \
277 (tvc)->fid.Fid.Volume == (afid)->Fid.Volume && \
278 (tvc)->fid.Cell == (afid)->Cell && \
279 ( (tvc)->fid.Fid.Unique == (afid)->Fid.Unique || \
280 (!(afid)->Fid.Unique && ((tvc)->states & CUnique))))
284 * Operations on circular queues implemented with pointers. Note: these queue
285 * objects are always located at the beginning of the structures they are linking.
287 #define QInit(q) ((q)->prev = (q)->next = (q))
288 #define QAdd(q,e) ((e)->next = (q)->next, (e)->prev = (q), \
289 (q)->next->prev = (e), (q)->next = (e))
290 #define QRemove(e) ((e)->next->prev = (e)->prev, (e)->prev->next = (e)->next)
291 #define QNext(e) ((e)->next)
292 #define QPrev(e) ((e)->prev)
293 #define QEmpty(q) ((q)->prev == (q))
294 /* this one takes q1 and sticks it on the end of q2 - that is, the other end, not the end
295 * that things are added onto. q1 shouldn't be empty, it's silly */
296 #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))
298 * Do lots of address arithmetic to go from vlruq to the base of the vcache
299 * structure. Don't move struct vnode, since we think of a struct vcache as
300 * a specialization of a struct vnode
302 #define QTOV(e) ((struct vcache *)(((char *) (e)) - (((char *)(&(((struct vcache *)(e))->vlruq))) - ((char *)(e)))))
303 #define QTOC(e) ((struct cell *)((char *) (e)))
306 #define SRVADDR_ISDOWN 0x20 /* same as SRVR_ISDOWN */
307 #define SRVADDR_NOUSE 0x40 /* Don't use this srvAddr */
309 struct srvAddr *next_bkt; /* next item in hash bucket */
310 struct srvAddr *next_sa; /* another interface on same host */
311 struct server *server; /* back to parent */
312 struct conn *conns; /* All user connections to this server */
313 afs_int32 sa_ip; /* Host addr in network byte order */
314 u_short sa_iprank; /* indiv ip address priority */
315 u_short sa_portal; /* port addr in network byte order */
320 * Values used in the flags field of the server structure below.
322 * AFS_SERVER_FLAG_ACTIVATED Has the server ever had a user connection
323 * associated with it?
325 #define AFS_SERVER_FLAG_ACTIVATED 0x01
326 #define SNO_LHOSTS 0x04
327 #define SYES_LHOSTS 0x08
328 #define SVLSRV_UUID 0x10
329 #define SRVR_ISDOWN 0x20
330 #define SRVR_MULTIHOMED 0x40
331 #define SRVR_ISGONE 0x80
332 #define SNO_INLINEBULK 0x100
333 #define SNO_64BIT 0x200
335 #define afs_serverSetNo64Bit(s) ((s)->srvr->server->flags |= SNO_64BIT)
336 #define afs_serverHasNo64Bit(s) ((s)->srvr->server->flags & SNO_64BIT)
342 afs_int32 addr_uniquifier;
346 struct srvAddr haddr;
349 #define sr_uuid _suid._srvUuid.suuid
350 #define sr_addr_uniquifier _suid._srvUuid.addr_uniquifier
351 #define sr_host _suid._srvId.haddr.ip
352 #define sr_portal _suid._srvId.haddr.portal
353 #define sr_rank _suid._srvId.haddr.ip_rank
354 #define sr_flags _suid._srvId.haddr.flags
355 #define sr_conns _suid._srvId.haddr.conns
356 struct server *next; /* Ptr to next server in hash chain */
357 struct cell *cell; /* Cell in which this host resides */
358 struct afs_cbr *cbrs; /* Return list of callbacks */
359 afs_int32 activationTime; /* Time when this record was first activated */
360 afs_int32 lastDowntimeStart; /* Time when last downtime incident began */
361 afs_int32 numDowntimeIncidents; /* # (completed) downtime incidents */
362 afs_int32 sumOfDowntimes; /* Total downtime experienced, in seconds */
363 struct srvAddr *addr;
364 afs_uint32 flags; /* Misc flags*/
367 #define afs_PutServer(servp, locktype)
369 /* structs for some pioctls - these are (or should be)
377 struct sprefrequest_33 {
378 unsigned short offset;
379 unsigned short num_servers;
383 struct sprefrequest { /* new struct for 3.4 */
384 unsigned short offset;
385 unsigned short num_servers;
386 unsigned short flags;
391 unsigned short next_offset;
392 unsigned short num_servers;
393 struct spref servers[1]; /* we overrun this array intentionally...*/
397 unsigned short flags;
398 unsigned short num_servers;
399 struct spref servers[1]; /* we overrun this array intentionally...*/
401 /* struct for GAG pioctl
404 afs_uint32 showflags, logflags, logwritethruflag, spare[3];
405 unsigned char spare2[128];
409 #define logwritethruON 1
412 afs_int32 rx_initReceiveWindow, rx_maxReceiveWindow,
413 rx_initSendWindow, rx_maxSendWindow, rxi_nSendFrags,
414 rxi_nRecvFrags, rxi_OrphanFragSize;
415 afs_int32 rx_maxReceiveSize, rx_MyMaxSendSize;
416 afs_uint32 spare[21];
419 /* struct for checkservers */
432 /* state bits for volume */
433 #define VRO 1 /* volume is readonly */
434 #define VRecheck 2 /* recheck volume info with server */
435 #define VBackup 4 /* is this a backup volume? */
436 #define VForeign 8 /* this is a non-afs volume */
437 #define VResort 16 /* server order was rearranged, sort when able */
438 #define VMoreReps 32 /* This volume has more replicas than we are */
439 /* keeping track of now -- check with VLDB */
441 enum repstate { not_busy, end_not_busy = 6, rd_busy, rdwr_busy, offline };
444 /* One structure per volume, describing where the volume is located
445 and where its mount points are. */
446 struct volume *next; /* Next volume in hash list. */
447 afs_int32 cell; /* the cell in which the volume resides */
448 afs_rwlock_t lock; /* the lock for this structure */
449 afs_int32 volume; /* This volume's ID number. */
450 char *name; /* This volume's name, or 0 if unknown */
451 struct server *serverHost[MAXHOSTS]; /* servers serving this volume */
452 enum repstate status[MAXHOSTS]; /* busy, offline, etc */
453 struct VenusFid dotdot; /* dir to access as .. */
454 struct VenusFid mtpoint; /* The mount point for this volume. */
455 afs_int32 rootVnode, rootUnique; /* Volume's root fid */
458 afs_int32 rwVol; /* For r/o vols, original read/write volume. */
459 afs_int32 accessTime; /* last time we used it */
460 afs_int32 vtix; /* volume table index */
461 afs_int32 copyDate; /* copyDate field, for tracking vol releases */
462 afs_int32 expireTime; /* for per-volume callbacks... */
463 short refCount; /* reference count for allocation */
464 char states; /* here for alignment reasons */
467 #define afs_PutVolume(av, locktype) ((av)->refCount--)
469 /* format of an entry in volume info file */
471 afs_int32 cell; /* cell for this entry */
472 afs_int32 volume; /* volume */
473 afs_int32 next; /* has index */
474 struct VenusFid dotdot; /* .. value */
475 struct VenusFid mtpoint; /* mt point's fid */
476 afs_int32 rootVnode, rootUnique; /* Volume's root fid */
480 struct SimpleLocks *next;
482 afs_int32 boff, eoff;
484 #if defined(AFS_AIX32_ENV) || defined(AFS_SUN5_ENV) || defined(AFS_SGI_ENV)
489 /* vcache state bits */
490 #define CStatd 0x00000001 /* has this file ever been stat'd? */
491 #define CBackup 0x00000002 /* file is on a backup volume */
492 #define CRO 0x00000004 /* is it on a read-only volume */
493 #define CMValid 0x00000008 /* is the mount point info valid? */
494 #define CCore 0x00000010 /* storing a core file, needed since we don't get an open */
495 #define CDirty 0x00000020 /* file has been modified since first open (... O_RDWR) */
496 #define CSafeStore 0x00000040 /* close must wait for store to finish (should be in fd) */
497 #define CMAPPED 0x00000080 /* Mapped files; primarily used by SunOS 4.0.x */
498 #define CNSHARE 0x00000100 /* support O_NSHARE semantics */
499 #define CLied 0x00000200
500 #define CTruth 0x00000400
502 #define CWired 0x00000800 /* OSF hack only */
504 #define CWRITE_IGN 0x00000800 /* Next OS hack only */
506 #define CUnique 0x00001000 /* vc's uniquifier - latest unifiquier for fid */
507 #define CForeign 0x00002000 /* this is a non-afs vcache */
508 #define CUnlinked 0x00010000
509 #define CBulkStat 0x00020000 /* loaded by a bulk stat, and not ref'd since */
510 #define CUnlinkedDel 0x00040000
511 #define CVFlushed 0x00080000
512 #define CCore1 0x00100000 /* osf1 core file; not same as CCore above */
513 #define CWritingUFS 0x00200000 /* to detect vm deadlock - used by sgi */
514 #define CCreating 0x00400000 /* avoid needless store after open truncate */
515 #define CPageHog 0x00800000 /* AIX - dumping large cores is a page hog. */
516 #define CDCLock 0x02000000 /* Vnode lock held over call to GetDownD */
517 #define CBulkFetching 0x04000000 /* stats are being fetched by bulk stat */
518 #define CExtendedFile 0x08000000 /* extended file via ftruncate call. */
520 /* vcache vstate bits */
521 #define VRevokeWait 0x1
522 #define VPageCleaning 0x2 /* Solaris - Cache Trunc Daemon sez keep out */
525 #if !defined(AFS_FBSD_ENV)
526 #define vrefCount v.v_count
528 #define vrefCount v.v_usecount
529 #endif /* AFS_FBSD_ENV */
531 #ifdef AFS_LINUX24_ENV
532 #define VREFCOUNT(v) atomic_read(&((vnode_t *) v)->v_count)
533 #define VREFCOUNT_SET(v, c) atomic_set(&((vnode_t *) v)->v_count, c)
534 #define VREFCOUNT_DEC(v) atomic_dec(&((vnode_t *) v)->v_count)
535 #define VREFCOUNT_INC(v) atomic_inc(&((vnode_t *) v)->v_count)
536 #define DLOCK() spin_lock(&dcache_lock)
537 #define DUNLOCK() spin_unlock(&dcache_lock)
538 #define DGET(d) dget_locked(d)
539 #define DCOUNT(d) atomic_read(&(d)->d_count)
541 #define VREFCOUNT(v) ((v)->vrefCount)
542 #define VREFCOUNT_SET(v, c) (v)->vrefCount = c;
543 #define VREFCOUNT_DEC(v) (v)->vrefCount--;
544 #define VREFCOUNT_INC(v) (v)->vrefCount++;
547 #define DGET(d) dget(d)
548 #define DCOUNT(d) ((d)->d_count)
551 #define AFS_MAXDV 0x7fffffff /* largest dataversion number */
552 #ifdef AFS_64BIT_CLIENT
553 #define AFS_NOTRUNC 0x7fffffffffffffffLL /* largest positive int64 number */
554 #else /* AFS_64BIT_CLIENT */
555 #define AFS_NOTRUNC 0x7fffffff /* largest dataversion number */
556 #endif /* AFS_64BIT_CLIENT */
558 extern afs_int32 vmPageHog; /* counter for # of vnodes which are page hogs. */
561 * Fast map from vcache to dcache
568 afs_offs_t minLoc; /* smallest offset into dc. */
569 afs_offs_t len; /* largest offset into dc. */
572 extern afs_uint32 afs_stampValue; /* stamp for pair's usage */
573 #define MakeStamp() (++afs_stampValue)
575 #define VTOAFS(V) ((struct vcache*)(V))
576 #define AFSTOV(V) (&(V)->v)
577 #ifdef AFS_LINUX22_ENV
578 #define ITOAFS(V) ((struct vcache*)(V))
579 #define AFSTOI(V) (struct inode *)(&(V)->v)
582 /* INVARIANTs: (vlruq.next != NULL) == (vlruq.prev != NULL)
583 * nextfree => !vlruq.next && ! vlruq.prev
584 * !(avc->nextfree) && !avc->vlruq.next => (FreeVCList == avc->nextfree)
587 struct vnode v; /* Has reference count in v.v_count */
588 struct afs_q vlruq; /* lru q next and prev */
589 struct vcache *nextfree; /* next on free list (if free) */
590 struct vcache *hnext; /* Hash next */
594 afs_hyper_t DataVersion;
598 ushort Mode; /* XXXX Should be afs_int32 XXXX */
600 /* vnode type is in v.v_type */
602 afs_rwlock_t lock; /* The lock on the vcache contents. */
603 #if defined(AFS_SUN5_ENV)
604 /* Lock used to protect the activeV, multipage, and vstates fields.
605 * Do not try to get the vcache lock when the vlock is held */
607 #endif /* defined(AFS_SUN5_ENV) */
608 #if defined(AFS_SUN5_ENV)
612 #if defined(AFS_SUN_ENV) || defined(AFS_ALPHA_ENV) || defined(AFS_DARWIN_ENV) || defined(AFS_FBSD_ENV)
613 afs_bozoLock_t pvnLock; /* see locks.x */
622 int ownslock; /* pid of owner of excl lock, else 0 - defect 3083 */
624 #ifdef AFS_DARWIN_ENV
625 struct lock__bsd__ rwlock;
630 afs_int32 parentVnode; /* Parent dir, if a file. */
631 afs_int32 parentUnique;
632 struct VenusFid *mvid; /* Either parent dir (if root) or root (if mt pt) */
633 char *linkData; /* Link data if a symlink. */
634 afs_hyper_t flushDV; /* data version last flushed from text */
635 afs_hyper_t mapDV; /* data version last flushed from map */
636 afs_size_t truncPos; /* truncate file to this position at next store */
637 struct server *callback; /* The callback host, if any */
638 afs_uint32 cbExpires; /* time the callback expires */
639 struct afs_q callsort; /* queue in expiry order, sort of */
640 struct axscache *Access; /* a list of cached access bits */
641 afs_int32 anyAccess; /* System:AnyUser's access to this. */
642 afs_int32 last_looker; /* pag/uid from last lookup here */
643 #if defined(AFS_SUN5_ENV)
645 #endif /* defined(AFS_SUN5_ENV) */
646 struct SimpleLocks *slocks;
647 short opens; /* The numbers of opens, read or write, on this file. */
648 short execsOrWriters; /* The number of execs (if < 0) or writers (if > 0) of
650 short flockCount; /* count of flock readers, or -1 if writer */
651 char mvstat; /* 0->normal, 1->mt pt, 2->root. */
652 afs_uint32 states; /* state bits */
653 #if defined(AFS_SUN5_ENV)
654 afs_uint32 vstates; /* vstate bits */
655 #endif /* defined(AFS_SUN5_ENV) */
657 afs_uint32 symhintstamp;
659 struct vcache *symhint;
660 struct dcache *dchint;
662 #ifdef AFS_LINUX22_ENV
663 u_short flushcnt; /* Number of flushes which haven't released yet. */
664 u_short mapcnt; /* Number of mappings of this file. */
666 #if defined(AFS_SGI_ENV)
667 daddr_t lastr; /* for read-ahead */
669 uint64_t vc_rwlockid; /* kthread owning rwlock */
671 short vc_rwlockid; /* pid of process owning rwlock */
673 short vc_locktrips; /* # of rwlock reacquisitions */
674 sema_t vc_rwlock; /* vop_rwlock for afs */
675 pgno_t mapcnt; /* # of pages mapped */
676 struct cred *cred; /* last writer's cred */
678 struct bhv_desc vc_bhv_desc; /* vnode's behavior data. */
680 #endif /* AFS_SGI_ENV */
681 afs_int32 vc_error; /* stash write error for this vnode. */
682 int xlatordv; /* Used by nfs xlator */
683 struct AFS_UCRED *uncred;
684 int asynchrony; /* num kbytes to store behind */
686 short multiPage; /* count of multi-page getpages in progress */
690 #define afs_symhint_inval(avc)
693 #define DONT_CHECK_MODE_BITS 0
694 #define CHECK_MODE_BITS 1
695 #define CMB_ALLOW_EXEC_AS_READ 2 /* For the NFS xlator */
697 #if defined(AFS_SGI_ENV)
698 #define AVCRWLOCK(avc) (valusema(&(avc)->vc_rwlock) <= 0)
700 /* SGI vnode rwlock macros and flags. */
701 #ifndef AFS_SGI62_ENV
702 /* The following are defined here. SGI 6.2 declares them in vnode.h */
703 #define VRWLOCK_READ 0
704 #define VRWLOCK_WRITE 1
705 #define VRWLOCK_WRITE_DIRECT 2
710 #define AFS_RWLOCK_T vrwlock_t
712 #define AFS_RWLOCK_T int
713 #endif /* AFS_SGI62_ENV */
715 #include <ksys/behavior.h>
716 #define AFS_RWLOCK(V,F) \
717 afs_rwlock(&VTOAFS(V)->vc_bhv_desc, (F));
718 #define AFS_RWUNLOCK(V,F) \
719 afs_rwunlock(&VTOAFS(V)->vc_bhv_desc, (F));
722 #define AFS_RWLOCK(V,F) afs_rwlock((vnode_t *)(V), (F) )
723 #define AFS_RWUNLOCK(V,F) afs_rwunlock((vnode_t *)(V), (F) )
725 #else /* AFS_SGI53_ENV */
726 #define AFS_RWLOCK(V,F) afs_rwlock((V))
727 #define AFS_RWUNLOCK(V,F) afs_rwunlock((V))
728 #endif /* AFS_SGI53_ENV */
729 #endif /* AFS_SGI_ENV */
733 afs_hyper_t DataVersion;
735 afs_int32 parentVnode;
736 afs_int32 parentUnique;
740 afs_int32 randomUid[CPSIZE];
741 afs_int32 callback; /* Now a pointer to 'server' struct */
743 afs_int32 randomAccess[CPSIZE];
746 short execsOrWriters;
757 /* CM inititialization parameters. What CM actually used after calculations
758 * based on passed in arguments.
760 #define CMI_VERSION 1 /* increment when adding new fields. */
761 struct cm_initparams {
766 int cmi_nVolumeCaches;
767 int cmi_firstChunkSize;
768 int cmi_otherChunkSize;
769 int cmi_cacheSize; /* The original cache size, in 1K blocks. */
770 unsigned cmi_setTime:1;
771 unsigned cmi_memCache:1;
772 int spare[16-9]; /* size of struct is 16 * 4 = 64 bytes */
776 /*----------------------------------------------------------------------
777 * AFS Data cache definitions
779 * Each entry describes a Unix file on the local disk that is
780 * is serving as a cached copy of all or part of a Vice file.
781 * Entries live in circular queues for each hash table slot
783 * Which queue is this thing in? Good question.
784 * A struct dcache entry is in the freeDSlot queue when not associated with a cache slot (file).
785 * Otherwise, it is in the DLRU queue. The freeDSlot queue uses the lruq.next field as
786 * its "next" pointer.
788 * Cache entries in the DLRU queue are either associated with vice files, in which case
789 * they are hashed by afs_dvnextTbl and afs_dcnextTbl pointers, or they are in the freeDCList
790 * and are not associated with any vice file. This last list uses the afs_dvnextTbl pointer for
791 * its "next" pointer.
792 *----------------------------------------------------------------------*/
794 #define NULLIDX (-1) /* null index definition */
795 /* struct dcache states bits */
796 #define DWriting 8 /* file being written (used for cache validation) */
798 /* dcache data flags */
799 #define DFEntryMod 0x02 /* has entry itself been modified? */
800 #define DFFetching 0x04 /* file is currently being fetched */
802 /* dcache meta flags */
803 #define DFNextStarted 0x01 /* next chunk has been prefetched already */
804 #define DFFetchReq 0x10 /* someone is waiting for DFFetching to go on */
807 /* flags in afs_indexFlags array */
808 #define IFEverUsed 1 /* index entry has >= 1 byte of data */
809 #define IFFree 2 /* index entry in freeDCList */
810 #define IFDataMod 4 /* file needs to be written out */
811 #define IFFlag 8 /* utility flag */
812 #define IFDirtyPages 16
813 #define IFAnyPages 32
814 #define IFDiscarded 64 /* index entry in discardDCList */
817 char *in; /* input buffer */
818 char *out; /* output buffer */
819 short in_size; /* Size of input buffer <= 2K */
820 short out_size; /* Maximum size of output buffer, <= 2K */
824 * This version of afs_ioctl is required to pass in 32 bit user space
825 * pointers into a 64 bit kernel.
836 /* CacheItems file has a header of type struct afs_fheader
837 * (keep aligned properly). Since we already have sgi_62 clients running
838 * with a 32 bit inode, a change is required to the header so that
839 * they can distinguish the old 32 bit inode CacheItems file and zap it
840 * instead of using it.
843 #define AFS_FHMAGIC 0x7635abaf /* uses version number */
845 #if defined(AFS_SUN57_64BIT_ENV)
846 #define AFS_CI_VERSION 3
848 #define AFS_CI_VERSION 2
851 afs_int32 firstCSize;
852 afs_int32 otherCSize;
855 /* kept on disk and in dcache entries */
857 struct VenusFid fid; /* Fid for this file */
858 afs_int32 modTime; /* last time this entry was modified */
859 afs_hyper_t versionNo; /* Associated data version number */
860 afs_int32 chunk; /* Relative chunk number */
861 #if defined(AFS_SGI61_ENV) || defined(AFS_SUN57_64BIT_ENV)
862 /* Using ino64_t here so that user level debugging programs compile
863 * the size correctly.
865 ino64_t inode; /* Unix inode for this chunk */
867 #if defined(AFS_LINUX_64BIT_KERNEL)
868 long inode; /* Unix inode for this chunk */
870 afs_int32 inode; /* Unix inode for this chunk */
873 afs_int32 chunkBytes; /* Num bytes in this chunk */
874 char states; /* Has this chunk been modified? */
877 /* magic numbers to specify the cache type */
879 #define AFS_FCACHE_TYPE_UFS 0x0
880 #define AFS_FCACHE_TYPE_MEM 0x1
881 #define AFS_FCACHE_TYPE_NFS 0x2
882 #define AFS_FCACHE_TYPE_EPI 0x3
886 struct afs_q lruq; /* Free queue for in-memory images */
887 afs_rwlock_t lock; /* Protects validPos, some f */
888 afs_rwlock_t tlock; /* Atomizes updates to refCount */
889 afs_rwlock_t mflock; /* Atomizes accesses/updates to mflags */
890 afs_size_t validPos; /* number of valid bytes during fetch */
891 afs_int32 index; /* The index in the CacheInfo file*/
892 short refCount; /* Associated reference count. */
893 char dflags; /* Data flags */
894 char mflags; /* Meta flags */
895 struct fcache f; /* disk image */
896 afs_int32 stamp; /* used with vtodc struct for hints */
901 * dcache.lock protects the actual contents of the cache file (in
902 * f.inode), subfields of f except those noted below, dflags and
905 * dcache.tlock is used to make atomic updates to refCount. Zero
906 * refCount dcache entries are protected by afs_xdcache instead of
909 * dcache.mflock is used to access and update mflags. It cannot be
910 * held without holding the corresponding dcache.lock. Updating
911 * mflags requires holding dcache.lock(R) and dcache.mflock(W), and
912 * checking for mflags requires dcache.lock(R) and dcache.mflock(R).
913 * Note that dcache.lock(W) gives you the right to update mflags,
914 * as dcache.mflock(W) can only be held with dcache.lock(R).
916 * dcache.stamp is protected by the associated vcache lock, because
917 * it's only purpose is to establish correspondence between vcache
918 * and dcache entries.
920 * dcache.index, dcache.f.fid, dcache.f.chunk and dcache.f.inode are
921 * write-protected by afs_xdcache and read-protected by refCount.
922 * Once an entry is referenced, these values cannot change, and if
923 * it's on the free list (with refCount=0), it can be reused for a
924 * different file/chunk. These values can only be written while
925 * holding afs_xdcache(W) and allocating this dcache entry (thereby
926 * ensuring noone else has a refCount on it).
929 /* this is obsolete and should be removed */
932 /* macro to mark a dcache entry as bad */
935 (x)->f.fid.Fid.Unique = 0; \
936 afs_indexUnique[(x)->index] = 0; \
937 (x)->dflags |= DFEntryMod; \
940 /* FakeOpen and Fake Close used to be real subroutines. They're only used in
941 * sun_subr and afs_vnodeops, and they're very frequently called, so I made
942 * them into macros. They do:
943 * FakeOpen: fake the file being open for writing. avc->lock must be held
944 * in write mode. Having the file open for writing is like having a DFS
945 * write-token: you're known to have the best version of the data around,
946 * and so the CM won't let it be overwritten by random server info.
947 * FakeClose: undo the effects of FakeOpen, noting that we want to ensure
948 * that a real close eventually gets done. We use CCore to achieve this if
949 * we would end up closing the file. avc->lock must be held in write mode */
951 #ifdef AFS_AIX_IAUTH_ENV
952 #define CRKEEP(V, C) (V)->linkData = (char*)crdup((C))
954 #define CRKEEP(V, C) crhold((C)); (V)->linkData = (char*)(C)
957 #define afs_FakeOpen(avc) { avc->opens++; avc->execsOrWriters++; }
958 #define afs_FakeClose(avc, acred) \
959 { if (avc->execsOrWriters == 1) { \
960 /* we're the last writer, just use CCore flag */ \
961 avc->states |= CCore; /* causes close to be called later */ \
963 /* The cred and vnode holds will be released in afs_FlushActiveVcaches */ \
964 VN_HOLD(AFSTOV(avc)); /* So it won't disappear */ \
965 CRKEEP(avc, acred); /* Should use a better place for the creds */ \
968 /* we're not the last writer, let the last one do the store-back for us */ \
970 avc->execsOrWriters--; \
974 #define AFS_ZEROS 64 /* zero buffer */
976 /*#define afs_DirtyPages(avc) (((avc)->states & CDirty) || osi_VMDirty_p((avc)))*/
977 #define afs_DirtyPages(avc) ((avc)->states & CDirty)
979 /* The PFlush algorithm makes use of the fact that Fid.Unique is not used in
980 below hash algorithms. Change it if need be so that flushing algorithm
981 doesn't move things from one hash chain to another
983 /* extern int afs_dhashsize; */
984 #define DCHash(v, c) ((((v)->Fid.Vnode + (v)->Fid.Volume + (c))) & (afs_dhashsize-1))
985 /*Vnode, Chunk -> Hash table index*/
986 #define DVHash(v) ((((v)->Fid.Vnode + (v)->Fid.Volume )) & (afs_dhashsize-1))
987 /*Vnode -> Other hash table index*/
988 /* don't hash on the cell, our callback-breaking code sometimes fails to compute
989 the cell correctly, and only scans one hash bucket */
990 #define VCHash(fid) (((fid)->Fid.Volume + (fid)->Fid.Vnode) & (VCSIZE-1))
992 extern struct dcache **afs_indexTable; /*Pointers to in-memory dcache entries*/
993 extern afs_int32 *afs_indexUnique; /*dcache entry Fid.Unique */
994 extern afs_int32 *afs_dvnextTbl; /*Dcache hash table links */
995 extern afs_int32 *afs_dcnextTbl; /*Dcache hash table links */
996 extern afs_int32 afs_cacheFiles; /*Size of afs_indexTable*/
997 extern afs_int32 afs_cacheBlocks; /*1K blocks in cache*/
998 extern afs_int32 afs_cacheStats; /*Stat entries in cache*/
999 extern struct vcache *afs_vhashT[VCSIZE]; /*Stat cache hash table*/
1000 extern afs_int32 afs_initState; /*Initialization state*/
1001 extern afs_int32 afs_termState; /* Termination state */
1002 extern struct VenusFid afs_rootFid; /*Root for whole file system*/
1003 extern afs_int32 afs_allCBs; /* Count of callbacks*/
1004 extern afs_int32 afs_oddCBs; /* Count of odd callbacks*/
1005 extern afs_int32 afs_evenCBs; /* Count of even callbacks*/
1006 extern afs_int32 afs_allZaps; /* Count of fid deletes */
1007 extern afs_int32 afs_oddZaps; /* Count of odd fid deletes */
1008 extern afs_int32 afs_evenZaps; /* Count of even fid deletes */
1009 extern struct brequest afs_brs[NBRS]; /* request structures */
1011 #define UHash(auid) ((auid) & (NUSERS-1))
1012 #define VHash(avol) ((avol)&(NVOLS-1))
1013 #define SHash(aserv) ((ntohl(aserv)) & (NSERVERS-1))
1014 #define FVHash(acell,avol) (((avol)+(acell)) & (NFENTRIES-1))
1016 extern struct cell *afs_GetCell();
1017 extern struct cell *afs_GetCellNoLock();
1018 extern struct cell *afs_GetCellByName();
1019 extern struct cell *afs_GetCellByName2();
1020 extern struct cell *afs_GetCellByIndex();
1021 extern struct unixuser *afs_GetUser();
1022 extern struct volume *afs_GetVolume();
1023 extern struct volume *afs_GetVolumeByName();
1024 extern struct conn *afs_Conn();
1025 extern struct conn *afs_ConnByHost();
1026 extern struct conn *afs_ConnByMHosts();
1027 extern afs_int32 afs_NewCell();
1028 extern struct dcache *afs_GetDCache();
1029 extern struct dcache *afs_FindDCache();
1030 extern struct dcache *afs_NewDCache();
1031 extern struct dcache *afs_GetDSlot();
1032 extern struct vcache *afs_GetVCache();
1033 extern struct brequest *afs_BQueue();
1036 extern int afs_CacheInit();
1037 extern void afs_StoreWarn();
1038 extern void afs_AdjustSize();
1039 extern void afs_ComputeCacheParms();
1040 extern void afs_FlushDCache();
1041 extern void afs_FlushActiveVcaches();
1042 extern void afs_StuffVcache();
1043 extern void afs_PutVCache();
1044 extern void afs_TryToSmush();
1045 extern void afs_ProcessFS();
1046 extern void afs_WriteThroughDSlots();
1047 extern void shutdown_cache();
1049 extern void afs_shutdown();
1051 extern void shutdown_osifile();
1054 extern int afs_IsDynrootFid();
1055 extern void afs_GetDynrootFid();
1056 extern int afs_IsDynroot();
1057 extern void afs_RefreshDynroot();
1058 extern void afs_GetDynroot();
1059 extern void afs_PutDynroot();
1060 extern int afs_DynrootNewVnode();
1061 extern int afs_SetDynrootEnable();
1062 extern int afs_GetDynrootEnable();
1063 extern int afs_DynrootVOPSymlink();
1064 extern int afs_DynrootVOPRemove();
1067 /* Performance hack - we could replace VerifyVCache2 with the appropriate
1068 * GetVCache incantation, and could eliminate even this code from afs_UFSRead
1069 * by making intentionally invalidating quick.stamp in the various callbacks
1070 * expiration/breaking code */
1071 #ifdef AFS_LINUX20_ENV
1072 #define afs_VerifyVCache(avc, areq) \
1073 (((avc)->states & CStatd) ? (vcache2inode(avc), 0) : \
1074 afs_VerifyVCache2((avc),areq))
1076 #ifdef AFS_DARWIN_ENV
1077 #define afs_VerifyVCache(avc, areq) \
1078 (((avc)->states & CStatd) ? (osi_VM_Setup(avc), 0) : \
1079 afs_VerifyVCache2((avc),areq))
1081 #define afs_VerifyVCache(avc, areq) \
1082 (((avc)->states & CStatd) ? 0 : afs_VerifyVCache2((avc),areq))
1086 #define DO_STATS 1 /* bits used by FindVCache */
1089 /* values for flag param of afs_CheckVolumeNames */
1090 #define AFS_VOLCHECK_EXPIRED 0x1 /* volumes whose callbacks have expired */
1091 #define AFS_VOLCHECK_BUSY 0x2 /* volumes which were marked busy */
1092 #define AFS_VOLCHECK_MTPTS 0x4 /* mount point invalidation also */
1093 #define AFS_VOLCHECK_FORCE 0x8 /* do all forcibly */
1097 #define AFS_FSPORT ((unsigned short) htons(7000))
1098 #define AFS_VLPORT ((unsigned short) htons(7003))
1100 #define afs_read(avc, uio, acred, albn, abpp, nolock) \
1101 (*(afs_cacheType->vread))(avc, uio, acred, albn, abpp, nolock)
1102 #define afs_write(avc, uio, aio, acred, nolock) \
1103 (*(afs_cacheType->vwrite))(avc, uio, aio, acred, nolock)
1105 #define afs_rdwr(avc, uio, rw, io, cred) \
1106 (((rw) == UIO_WRITE) ? afs_write(avc, uio, io, cred, 0) : afs_read(avc, uio, cred, 0, 0, 0))
1107 #define afs_nlrdwr(avc, uio, rw, io, cred) \
1108 (((rw) == UIO_WRITE) ? afs_write(avc, uio, io, cred, 1) : afs_read(avc, uio, cred, 0, 0, 1))
1110 extern afs_int32 afs_blocksUsed, afs_blocksDiscarded;
1111 extern afs_int32 afs_discardDCCount, afs_freeDCCount;
1112 extern afs_int32 afs_bulkStatsDone, afs_bulkStatsLost;
1113 extern int afs_TruncateDaemonRunning;
1114 extern int afs_CacheTooFull;
1115 /* Cache size truncation uses the following low and high water marks:
1116 * If the cache is more than 95% full (CM_DCACHECOUNTFREEPCT), the cache
1117 * truncation daemon is awakened and will free up space until the cache is 85%
1118 * (CM_DCACHESPACEFREEPCT - CM_DCACHEEXTRAPCT) full.
1119 * afs_UFSWrite and afs_GetDCache (when it needs to fetch data) will wait on
1120 * afs_WaitForCacheDrain if the cache is 98% (CM_WAITFORDRAINPCT) full.
1121 * afs_GetDownD wakes those processes once the cache is 95% full
1122 * (CM_CACHESIZEDRAINEDPCT).
1124 extern void afs_MaybeWakeupTruncateDaemon();
1125 extern void afs_CacheTruncateDaemon();
1126 extern int afs_WaitForCacheDrain;
1127 #define CM_MAXDISCARDEDCHUNKS 16 /* # of chunks */
1128 #define CM_DCACHECOUNTFREEPCT 95 /* max pct of chunks in use */
1129 #define CM_DCACHESPACEFREEPCT 90 /* max pct of space in use */
1130 #define CM_DCACHEEXTRAPCT 5 /* extra to get when freeing */
1131 #define CM_CACHESIZEDRAINEDPCT 95 /* wakeup processes when down to here.*/
1132 #define CM_WAITFORDRAINPCT 98 /* sleep if cache is this full. */
1134 #define afs_CacheIsTooFull() \
1135 (afs_blocksUsed - afs_blocksDiscarded > \
1136 (CM_DCACHECOUNTFREEPCT*afs_cacheBlocks)/100 || \
1137 afs_freeDCCount - afs_discardDCCount < \
1138 ((100-CM_DCACHECOUNTFREEPCT)*afs_cacheFiles)/100)
1140 /* Handy max length of a numeric string. */
1141 #define CVBS 12 /* max afs_int32 is 2^32 ~ 4*10^9, +1 for NULL, +luck */
1143 extern int afs_norefpanic;
1144 #define refpanic(foo) if (afs_norefpanic) \
1145 { printf( foo ); afs_norefpanic++;} else osi_Panic( foo )
1150 ** these are defined in the AIX source code sys/fs_locks.h but are not
1151 ** defined anywhere in the /usr/include directory
1153 #if defined(AFS_AIX41_ENV)
1154 #define VN_LOCK(vp) simple_lock(&(vp)->v_lock)
1155 #define VN_UNLOCK(vp) simple_unlock(&(vp)->v_lock)
1158 /* get a file's serial number from a vnode */
1159 #ifndef afs_vnodeToInumber
1160 #if defined(AFS_SGI62_ENV) || defined(AFS_HAVE_VXFS) || defined(AFS_DARWIN_ENV)
1161 #define afs_vnodeToInumber(V) VnodeToIno(V)
1163 #ifdef AFS_DECOSF_ENV
1164 #define afs_vnodeToInumber(V) osi_vnodeToInumber(V)
1166 #define afs_vnodeToInumber(V) (VTOI(V)->i_number)
1167 #endif /* AFS_DECOSF_ENV */
1168 #endif /* AFS_SGI62_ENV */
1171 /* get a file's device number from a vnode */
1172 #ifndef afs_vnodeToDev
1173 #if defined(AFS_SGI62_ENV) || defined(AFS_HAVE_VXFS) || defined(AFS_DARWIN_ENV)
1174 #define afs_vnodeToDev(V) VnodeToDev(V)
1176 #ifdef AFS_DECOSF_ENV
1177 #define afs_vnodeToDev(V) osi_vnodeToDev(V)
1179 #define afs_vnodeToDev(V) (VTOI(V)->i_dev)
1180 #endif /* AFS_DECOSF_ENV */
1181 #endif /* AFS_SGI62_ENV */
1184 /* fakestat support: opaque storage for afs_EvalFakeStat to remember
1185 * what vcache should be released.
1187 struct afs_fakestat_state {
1191 struct vcache *root_vp;
1194 extern int afs_fakestat_enable;
1196 #endif /* _AFS_H_ */