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 long parm[BPARMS]; /* random parameters - long's work for
120 * both 32 and 64 bit platforms.
122 afs_int32 code; /* return code */
123 short refCount; /* use counter for this structure */
124 char opcode; /* what to do (store, fetch, etc) */
125 char flags; /* free, etc */
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
194 struct afs_q lruq; /* lru q next and prev */
195 afs_int32 cell; /* unique id assigned by venus */
196 char *cellName; /* char string name of cell */
197 struct server *cellHosts[MAXCELLHOSTS]; /* volume *location* hosts for this cell */
198 struct cell *lcellp; /* Associated linked cell */
199 u_short fsport; /* file server port */
200 u_short vlport; /* volume server port */
201 short states; /* state flags */
202 short cellIndex; /* relative index number per cell */
203 time_t timeout; /* data expire time, if non-zero */
206 #define afs_PutCell(cellp, locktype)
208 /* the unixuser flag bit definitions */
209 #define UHasTokens 1 /* are the st and ct fields valid (ever set)? */
210 #define UTokensBad 2 /* are tokens bad? */
211 #define UPrimary 4 /* on iff primary identity */
212 #define UNeedsReset 8 /* needs afs_ResetAccessCache call done */
213 #define UPAGCounted 16 /* entry seen during PAG search (for stats) */
214 /* A flag used by afs_GCPAGs to keep track of
215 * which entries in afs_users need to be deleted.
216 * The lifetime of its presence in the table is the
217 * lifetime of the afs_GCPAGs function.
219 #define TMP_UPAGNotReferenced 128
221 /* values for afs_gcpags */
223 { AFS_GCPAGS_NOTCOMPILED = 0
225 , AFS_GCPAGS_USERDISABLED
229 , AFS_GCPAGS_EINEXACT
230 , AFS_GCPAGS_EPROCEND
231 , AFS_GCPAGS_EPROCWALK
232 , AFS_GCPAGS_ECREDWALK
233 , AFS_GCPAGS_EPIDCHECK
234 , AFS_GCPAGS_ENICECHECK
237 extern afs_int32 afs_gcpags;
238 extern afs_int32 afs_gcpags_procsize;
241 struct unixuser *next; /* next hash pointer */
242 afs_int32 uid; /* search based on uid and cell */
244 afs_int32 vid; /* corresponding vice id in specified cell */
245 short refCount; /* reference count for allocation */
246 char states; /* flag info */
247 afs_int32 tokenTime; /* last time tokens were set, used for timing out conn data */
248 afs_int32 stLen; /* ticket length (if kerberos, includes kvno at head) */
249 char *stp; /* pointer to ticket itself */
250 struct ClearToken ct;
251 struct afs_exporter *exporter; /* more info about the exporter for the remote user */
255 /* Per-connection block. */
256 struct conn *next; /* Next dude same server. */
257 struct unixuser *user; /* user validated with respect to. */
258 struct rx_connection *id; /* RPC connid. */
259 struct srvAddr *srvr; /* server associated with this conn */
260 short refCount; /* reference count for allocation */
261 unsigned short port; /* port associated with this connection */
262 char forceConnectFS; /* Should we try again with these tokens? */
268 /* Fid comparison routines */
269 #define FidCmp(a,b) ((a)->Fid.Unique != (b)->Fid.Unique \
270 || (a)->Fid.Vnode != (b)->Fid.Vnode \
271 || (a)->Fid.Volume != (b)->Fid.Volume \
272 || (a)->Cell != (b)->Cell)
274 #define FidMatches(afid,tvc) ((tvc)->fid.Fid.Vnode == (afid)->Fid.Vnode && \
275 (tvc)->fid.Fid.Volume == (afid)->Fid.Volume && \
276 (tvc)->fid.Cell == (afid)->Cell && \
277 ( (tvc)->fid.Fid.Unique == (afid)->Fid.Unique || \
278 (!(afid)->Fid.Unique && ((tvc)->states & CUnique))))
282 * Operations on circular queues implemented with pointers. Note: these queue
283 * objects are always located at the beginning of the structures they are linking.
285 #define QInit(q) ((q)->prev = (q)->next = (q))
286 #define QAdd(q,e) ((e)->next = (q)->next, (e)->prev = (q), \
287 (q)->next->prev = (e), (q)->next = (e))
288 #define QRemove(e) ((e)->next->prev = (e)->prev, (e)->prev->next = (e)->next)
289 #define QNext(e) ((e)->next)
290 #define QPrev(e) ((e)->prev)
291 #define QEmpty(q) ((q)->prev == (q))
292 /* this one takes q1 and sticks it on the end of q2 - that is, the other end, not the end
293 * that things are added onto. q1 shouldn't be empty, it's silly */
294 #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))
296 * Do lots of address arithmetic to go from vlruq to the base of the vcache
297 * structure. Don't move struct vnode, since we think of a struct vcache as
298 * a specialization of a struct vnode
300 #define QTOV(e) ((struct vcache *)(((char *) (e)) - (((char *)(&(((struct vcache *)(e))->vlruq))) - ((char *)(e)))))
301 #define QTOC(e) ((struct cell *)((char *) (e)))
304 #define SRVADDR_ISDOWN 0x20 /* same as SRVR_ISDOWN */
305 #define SRVADDR_NOUSE 0x40 /* Don't use this srvAddr */
307 struct srvAddr *next_bkt; /* next item in hash bucket */
308 struct srvAddr *next_sa; /* another interface on same host */
309 struct server *server; /* back to parent */
310 struct conn *conns; /* All user connections to this server */
311 afs_int32 sa_ip; /* Host addr in network byte order */
312 u_short sa_iprank; /* indiv ip address priority */
313 u_short sa_portal; /* port addr in network byte order */
318 * Values used in the flags field of the server structure below.
320 * AFS_SERVER_FLAG_ACTIVATED Has the server ever had a user connection
321 * associated with it?
323 #define AFS_SERVER_FLAG_ACTIVATED 0x01
324 #define SNO_LHOSTS 0x04
325 #define SYES_LHOSTS 0x08
326 #define SVLSRV_UUID 0x10
327 #define SRVR_ISDOWN 0x20
328 #define SRVR_MULTIHOMED 0x40
329 #define SRVR_ISGONE 0x80
335 afs_int32 addr_uniquifier;
339 struct srvAddr haddr;
342 #define sr_uuid _suid._srvUuid.suuid
343 #define sr_addr_uniquifier _suid._srvUuid.addr_uniquifier
344 #define sr_host _suid._srvId.haddr.ip
345 #define sr_portal _suid._srvId.haddr.portal
346 #define sr_rank _suid._srvId.haddr.ip_rank
347 #define sr_flags _suid._srvId.haddr.flags
348 #define sr_conns _suid._srvId.haddr.conns
349 struct server *next; /* Ptr to next server in hash chain */
350 struct cell *cell; /* Cell in which this host resides */
351 struct afs_cbr *cbrs; /* Return list of callbacks */
352 afs_int32 activationTime; /* Time when this record was first activated */
353 afs_int32 lastDowntimeStart; /* Time when last downtime incident began */
354 afs_int32 numDowntimeIncidents; /* # (completed) downtime incidents */
355 afs_int32 sumOfDowntimes; /* Total downtime experienced, in seconds */
356 struct srvAddr *addr;
357 char flags; /* Misc flags*/
360 #define afs_PutServer(servp, locktype)
362 /* structs for some pioctls - these are (or should be)
370 struct sprefrequest_33 {
371 unsigned short offset;
372 unsigned short num_servers;
376 struct sprefrequest { /* new struct for 3.4 */
377 unsigned short offset;
378 unsigned short num_servers;
379 unsigned short flags;
384 unsigned short next_offset;
385 unsigned short num_servers;
386 struct spref servers[1]; /* we overrun this array intentionally...*/
390 unsigned short flags;
391 unsigned short num_servers;
392 struct spref servers[1]; /* we overrun this array intentionally...*/
394 /* struct for GAG pioctl
397 afs_uint32 showflags, logflags, logwritethruflag, spare[3];
398 unsigned char spare2[128];
402 #define logwritethruON 1
405 afs_int32 rx_initReceiveWindow, rx_maxReceiveWindow,
406 rx_initSendWindow, rx_maxSendWindow, rxi_nSendFrags,
407 rxi_nRecvFrags, rxi_OrphanFragSize;
408 afs_int32 rx_maxReceiveSize, rx_MyMaxSendSize;
409 afs_uint32 spare[21];
412 /* struct for checkservers */
425 /* state bits for volume */
426 #define VRO 1 /* volume is readonly */
427 #define VRecheck 2 /* recheck volume info with server */
428 #define VBackup 4 /* is this a backup volume? */
429 #define VForeign 8 /* this is a non-afs volume */
430 #define VResort 16 /* server order was rearranged, sort when able */
431 #define VMoreReps 32 /* This volume has more replicas than we are */
432 /* keeping track of now -- check with VLDB */
434 enum repstate { not_busy, end_not_busy = 6, rd_busy, rdwr_busy, offline };
437 /* One structure per volume, describing where the volume is located
438 and where its mount points are. */
439 struct volume *next; /* Next volume in hash list. */
440 afs_int32 cell; /* the cell in which the volume resides */
441 afs_rwlock_t lock; /* the lock for this structure */
442 afs_int32 volume; /* This volume's ID number. */
443 char *name; /* This volume's name, or 0 if unknown */
444 struct server *serverHost[MAXHOSTS]; /* servers serving this volume */
445 enum repstate status[MAXHOSTS]; /* busy, offline, etc */
446 struct VenusFid dotdot; /* dir to access as .. */
447 struct VenusFid mtpoint; /* The mount point for this volume. */
448 afs_int32 rootVnode, rootUnique; /* Volume's root fid */
451 afs_int32 rwVol; /* For r/o vols, original read/write volume. */
452 afs_int32 accessTime; /* last time we used it */
453 afs_int32 vtix; /* volume table index */
454 afs_int32 copyDate; /* copyDate field, for tracking vol releases */
455 afs_int32 expireTime; /* for per-volume callbacks... */
456 short refCount; /* reference count for allocation */
457 char states; /* here for alignment reasons */
460 #define afs_PutVolume(av, locktype) ((av)->refCount--)
462 /* format of an entry in volume info file */
464 afs_int32 cell; /* cell for this entry */
465 afs_int32 volume; /* volume */
466 afs_int32 next; /* has index */
467 struct VenusFid dotdot; /* .. value */
468 struct VenusFid mtpoint; /* mt point's fid */
469 afs_int32 rootVnode, rootUnique; /* Volume's root fid */
473 struct SimpleLocks *next;
475 afs_int32 boff, eoff;
477 #if defined(AFS_AIX32_ENV) || defined(AFS_SUN5_ENV) || defined(AFS_SGI_ENV)
482 /* vcache state bits */
483 #define CStatd 1 /* has this file ever been stat'd? */
484 #define CBackup 2 /* file is on a backup volume */
485 #define CRO 4 /* is it on a read-only volume */
486 #define CMValid 8 /* is the mount point info valid? */
487 #define CCore 0x10 /* storing a core file, needed since we don't get an open */
488 #define CDirty 0x20 /* file has been modified since first open (... O_RDWR) */
489 #define CSafeStore 0x40 /* close must wait for store to finish (should be in fd) */
490 #define CMAPPED 0x80 /* Mapped files; primarily used by SunOS 4.0.x */
491 #define CNSHARE 0x100 /* support O_NSHARE semantics */
495 #define CWired 0x800 /* OSF hack only */
497 #define CWRITE_IGN 0x800 /* Next OS hack only */
499 #define CUnique 0x1000 /* vc's uniquifier - latest unifiquier for fid */
500 #define CForeign 0x2000 /* this is a non-afs vcache */
501 #define CHasPages 0x4000
502 #define CUnlinked 0x10000
503 #define CBulkStat 0x20000 /* loaded by a bulk stat, and not ref'd since */
504 #define CUnlinkedDel 0x40000
505 #define CVFlushed 0x80000
506 #define CCore1 0x100000 /* osf1 core file; not same as CCore above */
507 #define CWritingUFS 0x200000 /* to detect vm deadlock - used by sgi */
508 #define CCreating 0x400000 /* avoid needless store after open truncate */
509 #define CPageHog 0x800000 /* AIX - dumping large cores is a page hog. */
510 #define CDCLock 0x200000 /* Vnode lock held over call to GetDownD */
511 #define CBulkFetching 0x4000000 /* stats are being fetched by bulk stat */
512 #define CExtendedFile 0x8000000 /* extended file via ftruncate call. */
514 /* vcache vstate bits */
515 #define VRevokeWait 0x1
516 #define VPageCleaning 0x2 /* Solaris - Cache Trunc Daemon sez keep out */
519 #if !defined(AFS_FBSD_ENV)
520 #define vrefCount v.v_count
522 #define vrefCount v.v_usecount
523 #endif /* AFS_FBSD_ENV */
525 #define AFS_MAXDV 0x7fffffff /* largest dataversion number */
526 #define AFS_NOTRUNC 0x7fffffff /* largest dataversion number */
528 extern afs_int32 vmPageHog; /* counter for # of vnodes which are page hogs. */
531 * Fast map from vcache to dcache
538 afs_uint32 minLoc; /* smallest offset into dc. */
539 afs_uint32 len; /* largest offset into dc. */
542 extern afs_uint32 afs_stampValue; /* stamp for pair's usage */
543 #define MakeStamp() (++afs_stampValue)
545 /* INVARIANTs: (vlruq.next != NULL) == (vlruq.prev != NULL)
546 * nextfree => !vlruq.next && ! vlruq.prev
547 * !(avc->nextfree) && !avc->vlruq.next => (FreeVCList == avc->nextfree)
550 struct vnode v; /* Has reference count in v.v_count */
551 struct afs_q vlruq; /* lru q next and prev */
552 struct vcache *nextfree; /* next on free list (if free) */
553 struct vcache *hnext; /* Hash next */
557 afs_hyper_t DataVersion;
561 ushort Mode; /* XXXX Should be afs_int32 XXXX */
563 /* vnode type is in v.v_type */
565 afs_rwlock_t lock; /* The lock on the vcache contents. */
566 #if defined(AFS_SUN5_ENV)
567 /* Lock used to protect the activeV, multipage, and vstates fields.
568 * Do not try to get the vcache lock when the vlock is held */
570 #endif /* defined(AFS_SUN5_ENV) */
571 #if defined(AFS_SUN_ENV) || defined(AFS_ALPHA_ENV) || defined(AFS_DARWIN_ENV)
572 #if defined(AFS_SUN5_ENV)
576 afs_bozoLock_t pvnLock; /* see locks.x */
585 int ownslock; /* pid of owner of excl lock, else 0 - defect 3083 */
587 #ifdef AFS_DARWIN_ENV
588 struct lock__bsd__ rwlock;
590 afs_int32 parentVnode; /* Parent dir, if a file. */
591 afs_int32 parentUnique;
592 struct VenusFid *mvid; /* Either parent dir (if root) or root (if mt pt) */
593 char *linkData; /* Link data if a symlink. */
594 afs_hyper_t flushDV; /* data version last flushed from text */
595 afs_hyper_t mapDV; /* data version last flushed from map */
596 afs_uint32 truncPos; /* truncate file to this position at next store */
597 struct server *callback; /* The callback host, if any */
598 afs_uint32 cbExpires; /* time the callback expires */
599 struct afs_q callsort; /* queue in expiry order, sort of */
600 struct axscache *Access; /* a list of cached access bits */
601 afs_int32 anyAccess; /* System:AnyUser's access to this. */
602 afs_int32 last_looker; /* pag/uid from last lookup here */
603 #if defined(AFS_SUN5_ENV)
605 #endif /* defined(AFS_SUN5_ENV) */
606 struct SimpleLocks *slocks;
607 short opens; /* The numbers of opens, read or write, on this file. */
608 short execsOrWriters; /* The number of execs (if < 0) or writers (if > 0) of
610 short flockCount; /* count of flock readers, or -1 if writer */
611 char mvstat; /* 0->normal, 1->mt pt, 2->root. */
612 afs_uint32 states; /* state bits */
613 #if defined(AFS_SUN5_ENV)
614 afs_uint32 vstates; /* vstate bits */
615 #endif /* defined(AFS_SUN5_ENV) */
617 afs_uint32 symhintstamp;
619 struct vcache *symhint;
620 struct dcache *dchint;
622 #ifdef AFS_LINUX22_ENV
623 u_short flushcnt; /* Number of flushes which haven't released yet. */
624 u_short mapcnt; /* Number of mappings of this file. */
626 #if defined(AFS_SGI_ENV)
627 daddr_t lastr; /* for read-ahead */
629 uint64_t vc_rwlockid; /* kthread owning rwlock */
631 short vc_rwlockid; /* pid of process owning rwlock */
633 short vc_locktrips; /* # of rwlock reacquisitions */
634 sema_t vc_rwlock; /* vop_rwlock for afs */
635 pgno_t mapcnt; /* # of pages mapped */
636 struct cred *cred; /* last writer's cred */
638 struct bhv_desc vc_bhv_desc; /* vnode's behavior data. */
640 #endif /* AFS_SGI_ENV */
641 afs_int32 vc_error; /* stash write error for this vnode. */
642 int xlatordv; /* Used by nfs xlator */
643 struct AFS_UCRED *uncred;
644 int asynchrony; /* num kbytes to store behind */
646 short multiPage; /* count of multi-page getpages in progress */
650 #define afs_symhint_inval(avc)
653 #define DONT_CHECK_MODE_BITS 0
654 #define CHECK_MODE_BITS 1
655 #define CMB_ALLOW_EXEC_AS_READ 2 /* For the NFS xlator */
657 #if defined(AFS_SGI_ENV)
658 #define AVCRWLOCK(avc) (valusema(&(avc)->vc_rwlock) <= 0)
660 /* SGI vnode rwlock macros and flags. */
661 #ifndef AFS_SGI62_ENV
662 /* The following are defined here. SGI 6.2 declares them in vnode.h */
663 #define VRWLOCK_READ 0
664 #define VRWLOCK_WRITE 1
665 #define VRWLOCK_WRITE_DIRECT 2
670 #define AFS_RWLOCK_T vrwlock_t
672 #define AFS_RWLOCK_T int
673 #endif /* AFS_SGI62_ENV */
675 #include <ksys/behavior.h>
676 #define AFS_RWLOCK(V,F) \
677 afs_rwlock(&(((struct vcache *)(V))->vc_bhv_desc), (F));
678 #define AFS_RWUNLOCK(V,F) \
679 afs_rwunlock(&(((struct vcache *)(V))->vc_bhv_desc), (F));
682 #define AFS_RWLOCK(V,F) afs_rwlock((vnode_t *)(V), (F) )
683 #define AFS_RWUNLOCK(V,F) afs_rwunlock((vnode_t *)(V), (F) )
685 #else /* AFS_SGI53_ENV */
686 #define AFS_RWLOCK(V,F) afs_rwlock((V))
687 #define AFS_RWUNLOCK(V,F) afs_rwunlock((V))
688 #endif /* AFS_SGI53_ENV */
689 #endif /* AFS_SGI_ENV */
693 afs_hyper_t DataVersion;
695 afs_int32 parentVnode;
696 afs_int32 parentUnique;
700 afs_int32 randomUid[CPSIZE];
701 afs_int32 callback; /* Now a pointer to 'server' struct */
703 afs_int32 randomAccess[CPSIZE];
706 short execsOrWriters;
717 /* CM inititialization parameters. What CM actually used after calculations
718 * based on passed in arguments.
720 #define CMI_VERSION 1 /* increment when adding new fields. */
721 struct cm_initparams {
726 int cmi_nVolumeCaches;
727 int cmi_firstChunkSize;
728 int cmi_otherChunkSize;
729 int cmi_cacheSize; /* The original cache size, in 1K blocks. */
730 unsigned cmi_setTime:1;
731 unsigned cmi_memCache:1;
732 int spare[16-9]; /* size of struct is 16 * 4 = 64 bytes */
736 /*----------------------------------------------------------------------
737 * AFS Data cache definitions
739 * Each entry describes a Unix file on the local disk that is
740 * is serving as a cached copy of all or part of a Vice file.
741 * Entries live in circular queues for each hash table slot
743 * Which queue is this thing in? Good question.
744 * A struct dcache entry is in the freeDSlot queue when not associated with a cache slot (file).
745 * Otherwise, it is in the DLRU queue. The freeDSlot queue uses the lruq.next field as
746 * its "next" pointer.
748 * Cache entries in the DLRU queue are either associated with vice files, in which case
749 * they are hashed by afs_dvnextTbl and afs_dcnextTbl pointers, or they are in the freeDCList
750 * and are not associated with any vice file. This last list uses the afs_dvnextTbl pointer for
751 * its "next" pointer.
752 *----------------------------------------------------------------------*/
754 #define NULLIDX (-1) /* null index definition */
755 /* struct dcache states bits */
756 #define DWriting 8 /* file being written (used for cache validation) */
759 #define DFNextStarted 1 /* next chunk has been prefetched already */
760 #define DFEntryMod 2 /* has entry itself been modified? */
761 #define DFFetching 4 /* file is currently being fetched */
762 #define DFWaiting 8 /* someone waiting for file */
763 #define DFFetchReq 0x10 /* someone is waiting for DFFetching to go on */
766 /* flags in afs_indexFlags array */
767 #define IFEverUsed 1 /* index entry has >= 1 byte of data */
768 #define IFFree 2 /* index entry in freeDCList */
769 #define IFDataMod 4 /* file needs to be written out */
770 #define IFFlag 8 /* utility flag */
771 #define IFDirtyPages 16
772 #define IFAnyPages 32
773 #define IFDiscarded 64 /* index entry in discardDCList */
776 char *in; /* input buffer */
777 char *out; /* output buffer */
778 short in_size; /* Size of input buffer <= 2K */
779 short out_size; /* Maximum size of output buffer, <= 2K */
783 * This version of afs_ioctl is required to pass in 32 bit user space
784 * pointers into a 64 bit kernel.
795 /* CacheItems file has a header of type struct afs_fheader
796 * (keep aligned properly). Since we already have sgi_62 clients running
797 * with a 32 bit inode, a change is required to the header so that
798 * they can distinguish the old 32 bit inode CacheItems file and zap it
799 * instead of using it.
802 #define AFS_FHMAGIC 0x7635abaf /* uses version number */
804 #if defined(AFS_SUN57_64BIT_ENV)
805 #define AFS_CI_VERSION 3
807 #define AFS_CI_VERSION 2
810 afs_int32 firstCSize;
811 afs_int32 otherCSize;
814 /* kept on disk and in dcache entries */
816 struct VenusFid fid; /* Fid for this file */
817 afs_int32 modTime; /* last time this entry was modified */
818 afs_hyper_t versionNo; /* Associated data version number */
819 afs_int32 chunk; /* Relative chunk number */
820 #if defined(AFS_SGI61_ENV) || defined(AFS_SUN57_64BIT_ENV)
821 /* Using ino64_t here so that user level debugging programs compile
822 * the size correctly.
824 ino64_t inode; /* Unix inode for this chunk */
826 #if defined(AFS_LINUX_64BIT_KERNEL)
827 long inode; /* Unix inode for this chunk */
829 afs_int32 inode; /* Unix inode for this chunk */
832 afs_int32 chunkBytes; /* Num bytes in this chunk */
833 char states; /* Has this chunk been modified? */
836 /* magic numbers to specify the cache type */
838 #define AFS_FCACHE_TYPE_UFS 0x0
839 #define AFS_FCACHE_TYPE_MEM 0x1
840 #define AFS_FCACHE_TYPE_NFS 0x2
841 #define AFS_FCACHE_TYPE_EPI 0x3
845 struct afs_q lruq; /* Free queue for in-memory images */
846 afs_rwlock_t lock; /* XXX */
847 short refCount; /* Associated reference count. */
848 afs_int32 index; /* The index in the CacheInfo file*/
849 short flags; /* more flags bits */
850 afs_int32 validPos; /* number of valid bytes during fetch */
851 struct fcache f; /* disk image */
852 afs_int32 stamp; /* used with vtodc struct for hints */
854 /* this is obsolete and should be removed */
857 /* macro to mark a dcache entry as bad */
860 (x)->f.fid.Fid.Unique = 0; \
861 afs_indexUnique[(x)->index] = 0; \
862 (x)->flags |= DFEntryMod; \
866 * Convenient release macro for use when afs_PutDCache would cause
867 * deadlock on afs_xdcache lock
869 #define lockedPutDCache(ad) ((ad)->refCount--)
871 /* FakeOpen and Fake Close used to be real subroutines. They're only used in
872 * sun_subr and afs_vnodeops, and they're very frequently called, so I made
873 * them into macros. They do:
874 * FakeOpen: fake the file being open for writing. avc->lock must be held
875 * in write mode. Having the file open for writing is like having a DFS
876 * write-token: you're known to have the best version of the data around,
877 * and so the CM won't let it be overwritten by random server info.
878 * FakeClose: undo the effects of FakeOpen, noting that we want to ensure
879 * that a real close eventually gets done. We use CCore to achieve this if
880 * we would end up closing the file. avc->lock must be held in write mode */
882 #ifdef AFS_AIX_IAUTH_ENV
883 #define CRKEEP(V, C) (V)->linkData = (char*)crdup((C))
885 #define CRKEEP(V, C) crhold((C)); (V)->linkData = (char*)(C)
888 #define afs_FakeOpen(avc) { avc->opens++; avc->execsOrWriters++; }
889 #define afs_FakeClose(avc, acred) \
890 { if (avc->execsOrWriters == 1) { \
891 /* we're the last writer, just use CCore flag */ \
892 avc->states |= CCore; /* causes close to be called later */ \
894 /* The cred and vnode holds will be released in afs_FlushActiveVcaches */ \
895 VN_HOLD((struct vnode *)avc); /* So it won't disappear */ \
896 CRKEEP(avc, acred); /* Should use a better place for the creds */ \
899 /* we're not the last writer, let the last one do the store-back for us */ \
901 avc->execsOrWriters--; \
905 #define AFS_ZEROS 64 /* zero buffer */
907 /*#define afs_DirtyPages(avc) (((avc)->states & CDirty) || osi_VMDirty_p((avc)))*/
908 #define afs_DirtyPages(avc) ((avc)->states & CDirty)
910 /* The PFlush algorithm makes use of the fact that Fid.Unique is not used in
911 below hash algorithms. Change it if need be so that flushing algorithm
912 doesn't move things from one hash chain to another
914 /* extern int afs_dhashsize; */
915 #define DCHash(v, c) ((((v)->Fid.Vnode + (v)->Fid.Volume + (c))) & (afs_dhashsize-1))
916 /*Vnode, Chunk -> Hash table index*/
917 #define DVHash(v) ((((v)->Fid.Vnode + (v)->Fid.Volume )) & (afs_dhashsize-1))
918 /*Vnode -> Other hash table index*/
919 /* don't hash on the cell, our callback-breaking code sometimes fails to compute
920 the cell correctly, and only scans one hash bucket */
921 #define VCHash(fid) (((fid)->Fid.Volume + (fid)->Fid.Vnode) & (VCSIZE-1))
923 extern struct dcache **afs_indexTable; /*Pointers to in-memory dcache entries*/
924 extern afs_int32 *afs_indexUnique; /*dcache entry Fid.Unique */
925 extern afs_int32 *afs_dvnextTbl; /*Dcache hash table links */
926 extern afs_int32 *afs_dcnextTbl; /*Dcache hash table links */
927 extern afs_int32 afs_cacheFiles; /*Size of afs_indexTable*/
928 extern afs_int32 afs_cacheBlocks; /*1K blocks in cache*/
929 extern afs_int32 afs_cacheStats; /*Stat entries in cache*/
930 extern struct vcache *afs_vhashT[VCSIZE]; /*Stat cache hash table*/
931 extern afs_int32 afs_initState; /*Initialization state*/
932 extern afs_int32 afs_termState; /* Termination state */
933 extern struct VenusFid afs_rootFid; /*Root for whole file system*/
934 extern afs_int32 afs_allCBs; /* Count of callbacks*/
935 extern afs_int32 afs_oddCBs; /* Count of odd callbacks*/
936 extern afs_int32 afs_evenCBs; /* Count of even callbacks*/
937 extern afs_int32 afs_allZaps; /* Count of fid deletes */
938 extern afs_int32 afs_oddZaps; /* Count of odd fid deletes */
939 extern afs_int32 afs_evenZaps; /* Count of even fid deletes */
940 extern struct brequest afs_brs[NBRS]; /* request structures */
942 #define UHash(auid) ((auid) & (NUSERS-1))
943 #define VHash(avol) ((avol)&(NVOLS-1))
944 #define SHash(aserv) ((ntohl(aserv)) & (NSERVERS-1))
945 #define FVHash(acell,avol) (((avol)+(acell)) & (NFENTRIES-1))
947 extern struct cell *afs_GetCell();
948 extern struct cell *afs_GetCellByName();
949 extern struct unixuser *afs_GetUser();
950 extern struct volume *afs_GetVolume();
951 extern struct volume *afs_GetVolumeByName();
952 extern struct conn *afs_Conn();
953 extern struct conn *afs_ConnByHost();
954 extern struct conn *afs_ConnByMHosts();
955 extern afs_int32 afs_NewCell();
956 extern struct dcache *afs_GetDCache();
957 extern struct dcache *afs_FindDCache();
958 extern struct dcache *afs_NewDCache();
959 extern struct dcache *afs_GetDSlot();
960 extern struct vcache *afs_GetVCache();
961 extern struct brequest *afs_BQueue();
964 extern int afs_CacheInit();
965 extern void afs_StoreWarn();
966 extern void afs_AdjustSize();
967 extern void afs_ComputeCacheParms();
968 extern void afs_FlushDCache();
969 extern void afs_FlushActiveVcaches();
970 extern void afs_StuffVcache();
971 extern void afs_PutVCache();
972 extern void afs_TryToSmush();
973 extern void afs_ProcessFS();
974 extern void afs_WriteThroughDSlots();
975 extern void shutdown_cache();
977 extern void afs_shutdown();
979 extern void shutdown_osifile();
982 /* Performance hack - we could replace VerifyVCache2 with the appropriate
983 * GetVCache incantation, and could eliminate even this code from afs_UFSRead
984 * by making intentionally invalidating quick.stamp in the various callbacks
985 * expiration/breaking code */
986 #ifdef AFS_LINUX20_ENV
987 #define afs_VerifyVCache(avc, areq) \
988 (((avc)->states & CStatd) ? (vcache2inode(avc), 0) : \
989 afs_VerifyVCache2((avc),areq))
991 #ifdef AFS_DARWIN_ENV
992 #define afs_VerifyVCache(avc, areq) \
993 (((avc)->states & CStatd) ? (osi_VM_Setup(avc), 0) : \
994 afs_VerifyVCache2((avc),areq))
996 #define afs_VerifyVCache(avc, areq) \
997 (((avc)->states & CStatd) ? 0 : afs_VerifyVCache2((avc),areq))
1001 #define DO_STATS 1 /* bits used by FindVCache */
1004 /* values for flag param of afs_CheckVolumeNames */
1005 #define AFS_VOLCHECK_EXPIRED 0x1 /* volumes whose callbacks have expired */
1006 #define AFS_VOLCHECK_BUSY 0x2 /* volumes which were marked busy */
1007 #define AFS_VOLCHECK_MTPTS 0x4 /* mount point invalidation also */
1008 #define AFS_VOLCHECK_FORCE 0x8 /* do all forcibly */
1012 #define AFS_FSPORT ((unsigned short) htons(7000))
1013 #define AFS_VLPORT ((unsigned short) htons(7003))
1015 #define afs_read(avc, uio, acred, albn, abpp, nolock) \
1016 (*(afs_cacheType->vread))(avc, uio, acred, albn, abpp, nolock)
1017 #define afs_write(avc, uio, aio, acred, nolock) \
1018 (*(afs_cacheType->vwrite))(avc, uio, aio, acred, nolock)
1020 #define afs_rdwr(avc, uio, rw, io, cred) \
1021 (((rw) == UIO_WRITE) ? afs_write(avc, uio, io, cred, 0) : afs_read(avc, uio, cred, 0, 0, 0))
1022 #define afs_nlrdwr(avc, uio, rw, io, cred) \
1023 (((rw) == UIO_WRITE) ? afs_write(avc, uio, io, cred, 1) : afs_read(avc, uio, cred, 0, 0, 1))
1025 extern afs_int32 afs_blocksUsed, afs_blocksDiscarded;
1026 extern afs_int32 afs_discardDCCount, afs_freeDCCount;
1027 extern afs_int32 afs_bulkStatsDone, afs_bulkStatsLost;
1028 extern int afs_TruncateDaemonRunning;
1029 extern int afs_CacheTooFull;
1030 /* Cache size truncation uses the following low and high water marks:
1031 * If the cache is more than 95% full (CM_DCACHECOUNTFREEPCT), the cache
1032 * truncation daemon is awakened and will free up space until the cache is 85%
1033 * (CM_DCACHESPACEFREEPCT - CM_DCACHEEXTRAPCT) full.
1034 * afs_UFSWrite and afs_GetDCache (when it needs to fetch data) will wait on
1035 * afs_WaitForCacheDrain if the cache is 98% (CM_WAITFORDRAINPCT) full.
1036 * afs_GetDownD wakes those processes once the cache is 95% full
1037 * (CM_CACHESIZEDRAINEDPCT).
1039 extern void afs_CacheTruncateDaemon();
1040 extern int afs_WaitForCacheDrain;
1041 #define CM_MAXDISCARDEDCHUNKS 16 /* # of chunks */
1042 #define CM_DCACHECOUNTFREEPCT 95 /* max pct of chunks in use */
1043 #define CM_DCACHESPACEFREEPCT 90 /* max pct of space in use */
1044 #define CM_DCACHEEXTRAPCT 5 /* extra to get when freeing */
1045 #define CM_CACHESIZEDRAINEDPCT 95 /* wakeup processes when down to here.*/
1046 #define CM_WAITFORDRAINPCT 98 /* sleep if cache is this full. */
1048 #define afs_CacheIsTooFull() \
1049 (afs_blocksUsed - afs_blocksDiscarded > \
1050 (CM_DCACHECOUNTFREEPCT*afs_cacheBlocks)/100 || \
1051 afs_freeDCCount - afs_discardDCCount < \
1052 ((100-CM_DCACHECOUNTFREEPCT)*afs_cacheFiles)/100)
1054 #define afs_MaybeWakeupTruncateDaemon() \
1056 if (!afs_CacheTooFull && afs_CacheIsTooFull()) { \
1057 afs_CacheTooFull = 1; \
1058 if (!afs_TruncateDaemonRunning) { \
1059 afs_osi_Wakeup((char *)afs_CacheTruncateDaemon); \
1061 } else if (!afs_TruncateDaemonRunning && \
1062 afs_blocksDiscarded > CM_MAXDISCARDEDCHUNKS) { \
1063 afs_osi_Wakeup((char *)afs_CacheTruncateDaemon); \
1067 /* Handy max length of a numeric string. */
1068 #define CVBS 12 /* max afs_int32 is 2^32 ~ 4*10^9, +1 for NULL, +luck */
1070 extern int afs_norefpanic;
1071 #define refpanic(foo) if (afs_norefpanic) \
1072 { printf( foo ); afs_norefpanic++;} else osi_Panic( foo )
1077 ** these are defined in the AIX source code sys/fs_locks.h but are not
1078 ** defined anywhere in the /usr/include directory
1080 #if defined(AFS_AIX41_ENV)
1081 #define VN_LOCK(vp) simple_lock(&(vp)->v_lock)
1082 #define VN_UNLOCK(vp) simple_unlock(&(vp)->v_lock)
1085 /* get a file's serial number from a vnode */
1086 #ifndef afs_vnodeToInumber
1087 #if defined(AFS_SGI62_ENV) || defined(AFS_HAVE_VXFS) || defined(AFS_DARWIN_ENV)
1088 #define afs_vnodeToInumber(V) VnodeToIno(V)
1090 #ifdef AFS_DECOSF_ENV
1091 #define afs_vnodeToInumber(V) osi_vnodeToInumber(V)
1093 #define afs_vnodeToInumber(V) (VTOI(V)->i_number)
1094 #endif /* AFS_DECOSF_ENV */
1095 #endif /* AFS_SGI62_ENV */
1098 /* get a file's device number from a vnode */
1099 #ifndef afs_vnodeToDev
1100 #if defined(AFS_SGI62_ENV) || defined(AFS_HAVE_VXFS) || defined(AFS_DARWIN_ENV)
1101 #define afs_vnodeToDev(V) VnodeToDev(V)
1103 #ifdef AFS_DECOSF_ENV
1104 #define afs_vnodeToDev(V) osi_vnodeToDev(V)
1106 #define afs_vnodeToDev(V) (VTOI(V)->i_dev)
1107 #endif /* AFS_DECOSF_ENV */
1108 #endif /* AFS_SGI62_ENV */
1111 #endif /* _AFS_H_ */