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_SUN5_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) && !(defined(AFS_LINUX26_ENV) && defined(AFS_LINUX_64BIT_KERNEL)))
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 CBRSIZE 512 /* call back returns hash table size */
75 #define PIGGYSIZE 1350 /* max piggyback size */
76 #define MAXVOLS 128 /* max vols we can store */
77 #define MAXSYSNAME 128 /* max sysname (i.e. @sys) size */
78 #define MAXNUMSYSNAMES 16 /* max that current constants allow */
79 #define NOTOKTIMEOUT (2*3600) /* time after which to timeout conns sans tokens */
80 #define NOPAG 0xffffffff
81 #define AFS_NCBRS 300 /* max # of call back return entries */
82 #define AFS_MAXCBRSCALL 16 /* max to return in a given call */
83 #define AFS_SALLOC_LOW_WATER 250 /* Min free blocks before allocating more */
84 #define AFS_LRALLOCSIZ 4096 /* "Large" allocated size */
86 #define AFS_NRXPACKETS 80
87 #define AFS_RXDEADTIME 50
88 #define AFS_HARDDEADTIME 120
90 extern afs_int32 afs_rx_deadtime;
91 extern afs_int32 afs_rx_harddead;
99 /* flags to use with AFSOP_CACHEINIT */
100 #define AFSCALL_INIT_MEMCACHE 0x1 /* use a memory-based cache */
102 /* below here used only for kernel procedures */
104 /* Store synchrony flags - SYNC means that data should be forced to server's
105 * disk immediately upon completion. */
108 #define AFS_VMSYNC_INVAL 2 /* sync and invalidate pages */
109 #define AFS_LASTSTORE 4
112 /* background request structure */
115 #define BOP_NOOP 0 /* leave 0 unused */
116 #define BOP_FETCH 1 /* parm1 is chunk to get */
117 #define BOP_STORE 2 /* parm1 is chunk to store */
118 #define BOP_PATH 3 /* parm1 is path, parm2 is chunk to fetch */
120 #define B_DONTWAIT 1 /* On failure return; don't wait */
122 /* protocol is: refCount is incremented by user to take block out of free pool.
123 Next, BSTARTED is set when daemon finds request. This prevents
124 other daemons from picking up the same request. Finally, when
125 request is done, refCount is zeroed. BDONE and BWAIT are used by
126 dudes waiting for operation to proceed to a certain point before returning.
128 #define BSTARTED 1 /* request picked up by a daemon */
129 #define BUVALID 2 /* code is valid (store) */
130 #define BUWAIT 4 /* someone is waiting for BUVALID */
132 struct vcache *vc; /* vnode to use, with vrefcount bumped */
133 struct AFS_UCRED *cred; /* credentials to use for operation */
134 afs_size_t size_parm[BPARMS]; /* random parameters */
135 void *ptr_parm[BPARMS]; /* pointer parameters */
136 afs_int32 code; /* return code */
137 short refCount; /* use counter for this structure */
138 char opcode; /* what to do (store, fetch, etc) */
139 char flags; /* free, etc */
140 afs_int32 ts; /* counter "timestamp" */
148 afs_int32 AuthHandle;
149 char HandShakeKey[8];
151 afs_int32 BeginTimestamp;
152 afs_int32 EndTimestamp;
156 afs_int32 Cell; /* better sun packing if at end of structure */
160 /* 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!) */
163 afs_int32 CellAndUnique;
166 /* The actual number of bytes in the SmallFid, not the sizeof struct. */
167 #define SIZEOF_SMALLFID 10
171 * Queues implemented with both pointers and short offsets into a disk file.
179 afs_int32 uid; /* user id making the request */
180 afs_int32 busyCount; /* how many busies we've seen so far */
181 afs_int32 flags; /* things like O_SYNC, O_NONBLOCK go here */
182 char initd; /* if non-zero, non-uid fields meaningful */
183 char accessError; /* flags for overriding error return code */
184 char volumeError; /* encountered a missing or busy volume */
185 char networkError; /* encountered network problems */
186 char permWriteError; /* fileserver returns permenent error. */
191 /* structure linked off of a server to keep track of queued returned
192 * callbacks. Sent asynchronously when we run a little low on free dudes.
195 struct afs_cbr **pprev;
196 struct afs_cbr *next;
198 struct afs_cbr **hash_pprev;
199 struct afs_cbr *hash_next;
204 /* cellinfo file magic number */
205 #define AFS_CELLINFO_MAGIC 0xf32817cd
208 #define CNoSUID 0x02 /* disable suid bit for this cell */
209 #define CLinkedCell4 0x04 /* reserved for ADDCELL2 pioctl */
210 #define CNoAFSDB 0x08 /* never bother trying AFSDB */
211 #define CHasVolRef 0x10 /* volumes were referenced */
212 #define CLinkedCell 0x20 /* has a linked cell in lcellp */
215 struct afs_q lruq; /* lru q next and prev */
216 char *cellName; /* char string name of cell */
217 afs_int32 cellIndex; /* sequence number */
218 afs_int32 cellNum; /* semi-permanent cell number */
219 struct server *cellHosts[MAXCELLHOSTS]; /* volume *location* hosts */
220 struct cell *lcellp; /* Associated linked cell */
221 u_short fsport; /* file server port */
222 u_short vlport; /* volume server port */
223 short states; /* state flags */
224 time_t timeout; /* data expire time, if non-zero */
225 struct cell_name *cnamep; /* pointer to our cell_name */
226 afs_rwlock_t lock; /* protects cell data */
230 struct cell_name *next;
237 struct cell_alias *next;
243 #define afs_PutCell(cellp, locktype)
245 /* the unixuser flag bit definitions */
246 #define UHasTokens 1 /* are the st and ct fields valid (ever set)? */
247 #define UTokensBad 2 /* are tokens bad? */
248 #define UPrimary 4 /* on iff primary identity */
249 #define UNeedsReset 8 /* needs afs_ResetAccessCache call done */
250 #define UPAGCounted 16 /* entry seen during PAG search (for stats) */
251 /* A flag used by afs_GCPAGs to keep track of
252 * which entries in afs_users need to be deleted.
253 * The lifetime of its presence in the table is the
254 * lifetime of the afs_GCPAGs function.
256 #define TMP_UPAGNotReferenced 128
258 /* values for afs_gcpags */
259 enum { AFS_GCPAGS_NOTCOMPILED = 0, AFS_GCPAGS_OK =
260 1, AFS_GCPAGS_USERDISABLED, AFS_GCPAGS_EPROC0, AFS_GCPAGS_EPROCN,
261 AFS_GCPAGS_EEQPID, AFS_GCPAGS_EINEXACT, AFS_GCPAGS_EPROCEND,
262 AFS_GCPAGS_EPROCWALK, AFS_GCPAGS_ECREDWALK, AFS_GCPAGS_EPIDCHECK,
263 AFS_GCPAGS_ENICECHECK
266 extern afs_int32 afs_gcpags;
267 extern afs_int32 afs_gcpags_procsize;
268 extern afs_int32 afs_bkvolpref;
271 struct unixuser *next; /* next hash pointer */
272 afs_int32 uid; /* search based on uid and cell */
274 afs_int32 vid; /* corresponding vice id in specified cell */
275 short refCount; /* reference count for allocation */
276 char states; /* flag info */
277 afs_int32 tokenTime; /* last time tokens were set, used for timing out conn data */
278 afs_int32 stLen; /* ticket length (if kerberos, includes kvno at head) */
279 char *stp; /* pointer to ticket itself */
280 struct ClearToken ct;
281 struct afs_exporter *exporter; /* more info about the exporter for the remote user */
285 /* Per-connection block. */
286 struct conn *next; /* Next dude same server. */
287 struct unixuser *user; /* user validated with respect to. */
288 struct rx_connection *id; /* RPC connid. */
289 struct srvAddr *srvr; /* server associated with this conn */
290 short refCount; /* reference count for allocation */
291 unsigned short port; /* port associated with this connection */
292 char forceConnectFS; /* Should we try again with these tokens? */
298 /* Fid comparison routines */
299 #define FidCmp(a,b) ((a)->Fid.Unique != (b)->Fid.Unique \
300 || (a)->Fid.Vnode != (b)->Fid.Vnode \
301 || (a)->Fid.Volume != (b)->Fid.Volume \
302 || (a)->Cell != (b)->Cell)
304 #define FidMatches(afid,tvc) ((tvc)->fid.Fid.Vnode == (afid)->Fid.Vnode && \
305 (tvc)->fid.Fid.Volume == (afid)->Fid.Volume && \
306 (tvc)->fid.Cell == (afid)->Cell && \
307 ( (tvc)->fid.Fid.Unique == (afid)->Fid.Unique || \
308 (!(afid)->Fid.Unique && ((tvc)->states & CUnique))))
312 * Operations on circular queues implemented with pointers. Note: these queue
313 * objects are always located at the beginning of the structures they are linking.
315 #define QInit(q) ((q)->prev = (q)->next = (q))
316 #define QAdd(q,e) ((e)->next = (q)->next, (e)->prev = (q), \
317 (q)->next->prev = (e), (q)->next = (e))
318 #define QRemove(e) ((e)->next->prev = (e)->prev, (e)->prev->next = (e)->next)
319 #define QNext(e) ((e)->next)
320 #define QPrev(e) ((e)->prev)
321 #define QEmpty(q) ((q)->prev == (q))
322 /* this one takes q1 and sticks it on the end of q2 - that is, the other end, not the end
323 * that things are added onto. q1 shouldn't be empty, it's silly */
324 #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))
326 * Do lots of address arithmetic to go from vlruq to the base of the vcache
327 * structure. Don't move struct vnode, since we think of a struct vcache as
328 * a specialization of a struct vnode
330 #define QTOV(e) ((struct vcache *)(((char *) (e)) - (((char *)(&(((struct vcache *)(e))->vlruq))) - ((char *)(e)))))
331 #define QTOC(e) ((struct cell *)((char *) (e)))
334 #define SRVADDR_ISDOWN 0x20 /* same as SRVR_ISDOWN */
335 #define SRVADDR_NOUSE 0x40 /* Don't use this srvAddr */
337 struct srvAddr *next_bkt; /* next item in hash bucket */
338 struct srvAddr *next_sa; /* another interface on same host */
339 struct server *server; /* back to parent */
340 struct conn *conns; /* All user connections to this server */
341 afs_int32 sa_ip; /* Host addr in network byte order */
342 u_short sa_iprank; /* indiv ip address priority */
343 u_short sa_portal; /* port addr in network byte order */
348 * Values used in the flags field of the server structure below.
350 * AFS_SERVER_FLAG_ACTIVATED Has the server ever had a user connection
351 * associated with it?
353 #define AFS_SERVER_FLAG_ACTIVATED 0x01
354 #define SNO_LHOSTS 0x04
355 #define SYES_LHOSTS 0x08
356 #define SVLSRV_UUID 0x10
357 #define SRVR_ISDOWN 0x20
358 #define SRVR_MULTIHOMED 0x40
359 #define SRVR_ISGONE 0x80
360 #define SNO_INLINEBULK 0x100
361 #define SNO_64BIT 0x200
363 #define afs_serverSetNo64Bit(s) ((s)->srvr->server->flags |= SNO_64BIT)
364 #define afs_serverHasNo64Bit(s) ((s)->srvr->server->flags & SNO_64BIT)
370 afs_int32 addr_uniquifier;
374 struct srvAddr haddr;
377 #define sr_uuid _suid._srvUuid.suuid
378 #define sr_addr_uniquifier _suid._srvUuid.addr_uniquifier
379 #define sr_host _suid._srvId.haddr.ip
380 #define sr_portal _suid._srvId.haddr.portal
381 #define sr_rank _suid._srvId.haddr.ip_rank
382 #define sr_flags _suid._srvId.haddr.flags
383 #define sr_conns _suid._srvId.haddr.conns
384 struct server *next; /* Ptr to next server in hash chain */
385 struct cell *cell; /* Cell in which this host resides */
386 struct afs_cbr *cbrs; /* Return list of callbacks */
387 afs_int32 activationTime; /* Time when this record was first activated */
388 afs_int32 lastDowntimeStart; /* Time when last downtime incident began */
389 afs_int32 numDowntimeIncidents; /* # (completed) downtime incidents */
390 afs_int32 sumOfDowntimes; /* Total downtime experienced, in seconds */
391 struct srvAddr *addr;
392 afs_uint32 flags; /* Misc flags */
395 #define afs_PutServer(servp, locktype)
397 /* structs for some pioctls - these are (or should be)
405 struct sprefrequest_33 {
406 unsigned short offset;
407 unsigned short num_servers;
411 struct sprefrequest { /* new struct for 3.4 */
412 unsigned short offset;
413 unsigned short num_servers;
414 unsigned short flags;
419 unsigned short next_offset;
420 unsigned short num_servers;
421 struct spref servers[1]; /* we overrun this array intentionally... */
425 unsigned short flags;
426 unsigned short num_servers;
427 struct spref servers[1]; /* we overrun this array intentionally... */
429 /* struct for GAG pioctl
432 afs_uint32 showflags, logflags, logwritethruflag, spare[3];
433 unsigned char spare2[128];
437 #define logwritethruON 1
440 afs_int32 rx_initReceiveWindow, rx_maxReceiveWindow, rx_initSendWindow,
441 rx_maxSendWindow, rxi_nSendFrags, rxi_nRecvFrags, rxi_OrphanFragSize;
442 afs_int32 rx_maxReceiveSize, rx_MyMaxSendSize;
443 afs_uint32 spare[21];
446 /* struct for checkservers */
457 /* state bits for volume */
458 #define VRO 1 /* volume is readonly */
459 #define VRecheck 2 /* recheck volume info with server */
460 #define VBackup 4 /* is this a backup volume? */
461 #define VForeign 8 /* this is a non-afs volume */
462 #define VResort 16 /* server order was rearranged, sort when able */
463 #define VMoreReps 32 /* This volume has more replicas than we are */
464 /* keeping track of now -- check with VLDB */
466 enum repstate { not_busy, end_not_busy = 6, rd_busy, rdwr_busy, offline };
469 /* One structure per volume, describing where the volume is located
470 * and where its mount points are. */
471 struct volume *next; /* Next volume in hash list. */
472 afs_int32 cell; /* the cell in which the volume resides */
473 afs_rwlock_t lock; /* the lock for this structure */
474 afs_int32 volume; /* This volume's ID number. */
475 char *name; /* This volume's name, or 0 if unknown */
476 struct server *serverHost[MAXHOSTS]; /* servers serving this volume */
477 enum repstate status[MAXHOSTS]; /* busy, offline, etc */
478 struct VenusFid dotdot; /* dir to access as .. */
479 struct VenusFid mtpoint; /* The mount point for this volume. */
480 afs_int32 rootVnode, rootUnique; /* Volume's root fid */
483 afs_int32 rwVol; /* For r/o vols, original read/write volume. */
484 afs_int32 accessTime; /* last time we used it */
485 afs_int32 vtix; /* volume table index */
486 afs_int32 copyDate; /* copyDate field, for tracking vol releases */
487 afs_int32 expireTime; /* for per-volume callbacks... */
488 short refCount; /* reference count for allocation */
489 char states; /* here for alignment reasons */
492 #define afs_PutVolume(av, locktype) ((av)->refCount--)
494 /* format of an entry in volume info file */
496 afs_int32 cell; /* cell for this entry */
497 afs_int32 volume; /* volume */
498 afs_int32 next; /* has index */
499 struct VenusFid dotdot; /* .. value */
500 struct VenusFid mtpoint; /* mt point's fid */
501 afs_int32 rootVnode, rootUnique; /* Volume's root fid */
505 struct SimpleLocks *next;
507 afs_int32 boff, eoff;
509 #if defined(AFS_AIX32_ENV) || defined(AFS_SUN5_ENV) || defined(AFS_SGI_ENV)
514 /* vcache state bits */
515 #define CStatd 0x00000001 /* has this file ever been stat'd? */
516 #define CBackup 0x00000002 /* file is on a backup volume */
517 #define CRO 0x00000004 /* is it on a read-only volume */
518 #define CMValid 0x00000008 /* is the mount point info valid? */
519 #define CCore 0x00000010 /* storing a core file, needed since we don't get an open */
520 #define CDirty 0x00000020 /* file has been modified since first open (... O_RDWR) */
521 #define CSafeStore 0x00000040 /* close must wait for store to finish (should be in fd) */
522 #define CMAPPED 0x00000080 /* Mapped files; primarily used by SunOS 4.0.x */
523 #define CNSHARE 0x00000100 /* support O_NSHARE semantics */
524 #define CLied 0x00000200
525 #define CTruth 0x00000400
527 #define CWired 0x00000800 /* OSF hack only */
529 #ifdef AFS_DARWIN_ENV
530 #define CUBCinit 0x00000800
532 #define CWRITE_IGN 0x00000800 /* Next OS hack only */
535 #define CUnique 0x00001000 /* vc's uniquifier - latest unifiquier for fid */
536 #define CForeign 0x00002000 /* this is a non-afs vcache */
537 #define CUnlinked 0x00010000
538 #define CBulkStat 0x00020000 /* loaded by a bulk stat, and not ref'd since */
539 #define CUnlinkedDel 0x00040000
540 #define CVFlushed 0x00080000
541 #define CCore1 0x00100000 /* osf1 core file; not same as CCore above */
542 #define CWritingUFS 0x00200000 /* to detect vm deadlock - used by sgi */
543 #define CCreating 0x00400000 /* avoid needless store after open truncate */
544 #define CPageHog 0x00800000 /* AIX - dumping large cores is a page hog. */
545 #define CDCLock 0x02000000 /* Vnode lock held over call to GetDownD */
546 #define CBulkFetching 0x04000000 /* stats are being fetched by bulk stat */
547 #define CExtendedFile 0x08000000 /* extended file via ftruncate call. */
549 /* vcache vstate bits */
550 #define VRevokeWait 0x1
551 #define VPageCleaning 0x2 /* Solaris - Cache Trunc Daemon sez keep out */
554 #if defined(AFS_XBSD_ENV) || defined(AFS_DARWIN_ENV)
555 #define vrefCount v->v_usecount
557 #define vrefCount v.v_count
558 #endif /* AFS_XBSD_ENV */
560 #if defined(AFS_DARWIN80_ENV)
561 #define VREFCOUNT_GT(v, y) vnode_isinuse(AFSTOV(v), (y))
562 #elif defined(AFS_XBSD_ENV) || defined(AFS_DARWIN_ENV)
563 #define VREFCOUNT(v) ((v)->vrefCount)
564 #define VREFCOUNT_GT(v, y) (AFSTOV(v)->v_usecount > (y))
565 #elif defined(AFS_LINUX24_ENV)
566 #define VREFCOUNT(v) atomic_read(&((vnode_t *) v)->v_count)
567 #define VREFCOUNT_GT(v, y) ((atomic_read(&((vnode_t *) v)->v_count)>y)?1:0)
568 #define VREFCOUNT_SET(v, c) atomic_set(&((vnode_t *) v)->v_count, c)
569 #define VREFCOUNT_DEC(v) atomic_dec(&((vnode_t *) v)->v_count)
570 #define VREFCOUNT_INC(v) atomic_inc(&((vnode_t *) v)->v_count)
572 #define VREFCOUNT(v) ((v)->vrefCount)
573 #define VREFCOUNT_GT(v,y) ((v)->vrefCount > (y))
574 #define VREFCOUNT_SET(v, c) (v)->vrefCount = c;
575 #define VREFCOUNT_DEC(v) (v)->vrefCount--;
576 #define VREFCOUNT_INC(v) (v)->vrefCount++;
577 #define d_unhash(d) list_empty(&(d)->d_hash)
578 #define dget_locked(d) dget(d)
581 #define AFS_MAXDV 0x7fffffff /* largest dataversion number */
582 #ifdef AFS_64BIT_CLIENT
583 #define AFS_NOTRUNC 0x7fffffffffffffffLL /* largest positive int64 number */
584 #else /* AFS_64BIT_CLIENT */
585 #define AFS_NOTRUNC 0x7fffffff /* largest dataversion number */
586 #endif /* AFS_64BIT_CLIENT */
588 extern afs_int32 vmPageHog; /* counter for # of vnodes which are page hogs. */
591 * Fast map from vcache to dcache
597 afs_offs_t minLoc; /* smallest offset into dc. */
598 afs_offs_t len; /* largest offset into dc. */
601 extern afs_uint32 afs_stampValue; /* stamp for pair's usage */
602 #define MakeStamp() (++afs_stampValue)
604 #if defined(AFS_DARWIN_ENV)
605 #define VTOAFS(v) ((struct vcache *)vnode_fsnode((v)))
606 #define AFSTOV(vc) ((vc)->v)
607 #elif defined(AFS_XBSD_ENV)
608 #define VTOAFS(v) ((struct vcache *)(v)->v_data)
609 #define AFSTOV(vc) ((vc)->v)
611 #define VTOAFS(V) ((struct vcache *)(V))
612 #define AFSTOV(V) (&(V)->v)
615 #ifdef AFS_LINUX22_ENV
616 #define ITOAFS(V) ((struct vcache*)(V))
617 #define AFSTOI(V) (struct inode *)(&(V)->v)
620 /* INVARIANTs: (vlruq.next != NULL) == (vlruq.prev != NULL)
621 * nextfree => !vlruq.next && ! vlruq.prev
622 * !(avc->nextfree) && !avc->vlruq.next => (FreeVCList == avc->nextfree)
625 #if defined(AFS_XBSD_ENV)||defined(AFS_DARWIN_ENV)
628 struct vnode v; /* Has reference count in v.v_count */
630 struct afs_q vlruq; /* lru q next and prev */
631 struct vcache *nextfree; /* next on free list (if free) */
632 struct vcache *hnext; /* Hash next */
636 afs_hyper_t DataVersion;
640 afs_uint16 Mode; /* XXXX Should be afs_int32 XXXX */
641 afs_uint16 LinkCount;
642 /* vnode type is in v.v_type */
644 afs_rwlock_t lock; /* The lock on the vcache contents. */
645 #if defined(AFS_SUN5_ENV)
646 /* Lock used to protect the activeV, multipage, and vstates fields.
647 * Do not try to get the vcache lock when the vlock is held */
649 #endif /* defined(AFS_SUN5_ENV) */
650 #if defined(AFS_SUN5_ENV)
654 #ifdef AFS_BOZONLOCK_ENV
655 afs_bozoLock_t pvnLock; /* see locks.x */
660 #if defined(AFS_AIX51_ENV)
668 int ownslock; /* pid of owner of excl lock, else 0 - defect 3083 */
670 #ifdef AFS_DARWIN_ENV
671 #ifdef AFS_DARWIN80_ENV
674 struct lock__bsd__ rwlock;
680 afs_int32 parentVnode; /* Parent dir, if a file. */
681 afs_int32 parentUnique;
682 struct VenusFid *mvid; /* Either parent dir (if root) or root (if mt pt) */
683 char *linkData; /* Link data if a symlink. */
684 afs_hyper_t flushDV; /* data version last flushed from text */
685 afs_hyper_t mapDV; /* data version last flushed from map */
686 afs_size_t truncPos; /* truncate file to this position at next store */
687 struct server *callback; /* The callback host, if any */
688 afs_uint32 cbExpires; /* time the callback expires */
689 struct afs_q callsort; /* queue in expiry order, sort of */
690 struct axscache *Access; /* a list of cached access bits */
691 afs_int32 anyAccess; /* System:AnyUser's access to this. */
692 afs_int32 last_looker; /* pag/uid from last lookup here */
693 #if defined(AFS_SUN5_ENV)
695 #endif /* defined(AFS_SUN5_ENV) */
696 struct SimpleLocks *slocks;
697 short opens; /* The numbers of opens, read or write, on this file. */
698 short execsOrWriters; /* The number of execs (if < 0) or writers (if > 0) of
700 short flockCount; /* count of flock readers, or -1 if writer */
701 char mvstat; /* 0->normal, 1->mt pt, 2->root. */
702 afs_uint32 states; /* state bits */
703 #if defined(AFS_SUN5_ENV)
704 afs_uint32 vstates; /* vstate bits */
705 #endif /* defined(AFS_SUN5_ENV) */
707 afs_uint32 symhintstamp;
709 struct vcache *symhint;
710 struct dcache *dchint;
712 #ifdef AFS_LINUX22_ENV
713 u_short mapcnt; /* Number of mappings of this file. */
715 #if defined(AFS_SGI_ENV)
716 daddr_t lastr; /* for read-ahead */
718 uint64_t vc_rwlockid; /* kthread owning rwlock */
720 short vc_rwlockid; /* pid of process owning rwlock */
722 short vc_locktrips; /* # of rwlock reacquisitions */
723 sema_t vc_rwlock; /* vop_rwlock for afs */
724 pgno_t mapcnt; /* # of pages mapped */
725 struct cred *cred; /* last writer's cred */
727 struct bhv_desc vc_bhv_desc; /* vnode's behavior data. */
729 #endif /* AFS_SGI_ENV */
730 afs_int32 vc_error; /* stash write error for this vnode. */
731 int xlatordv; /* Used by nfs xlator */
732 struct AFS_UCRED *uncred;
733 int asynchrony; /* num kbytes to store behind */
735 short multiPage; /* count of multi-page getpages in progress */
739 #define afs_symhint_inval(avc)
742 #define DONT_CHECK_MODE_BITS 0
743 #define CHECK_MODE_BITS 1
744 #define CMB_ALLOW_EXEC_AS_READ 2 /* For the NFS xlator */
746 #if defined(AFS_SGI_ENV)
747 #define AVCRWLOCK(avc) (valusema(&(avc)->vc_rwlock) <= 0)
749 /* SGI vnode rwlock macros and flags. */
750 #ifndef AFS_SGI62_ENV
751 /* The following are defined here. SGI 6.2 declares them in vnode.h */
752 #define VRWLOCK_READ 0
753 #define VRWLOCK_WRITE 1
754 #define VRWLOCK_WRITE_DIRECT 2
759 #define AFS_RWLOCK_T vrwlock_t
761 #define AFS_RWLOCK_T int
762 #endif /* AFS_SGI62_ENV */
764 #include <ksys/behavior.h>
765 #define AFS_RWLOCK(V,F) \
766 afs_rwlock(&VTOAFS(V)->vc_bhv_desc, (F));
767 #define AFS_RWUNLOCK(V,F) \
768 afs_rwunlock(&VTOAFS(V)->vc_bhv_desc, (F));
771 #define AFS_RWLOCK(V,F) afs_rwlock((vnode_t *)(V), (F) )
772 #define AFS_RWUNLOCK(V,F) afs_rwunlock((vnode_t *)(V), (F) )
774 #else /* AFS_SGI53_ENV */
775 #define AFS_RWLOCK(V,F) afs_rwlock((V))
776 #define AFS_RWUNLOCK(V,F) afs_rwunlock((V))
777 #endif /* AFS_SGI53_ENV */
778 #endif /* AFS_SGI_ENV */
782 afs_hyper_t DataVersion;
784 afs_int32 parentVnode;
785 afs_int32 parentUnique;
789 afs_int32 randomUid[CPSIZE];
790 afs_int32 callback; /* Now a pointer to 'server' struct */
792 afs_int32 randomAccess[CPSIZE];
795 short execsOrWriters;
802 afs_int32 callerAccess;
813 /* CM inititialization parameters. What CM actually used after calculations
814 * based on passed in arguments.
816 #define CMI_VERSION 1 /* increment when adding new fields. */
817 struct cm_initparams {
822 int cmi_nVolumeCaches;
823 int cmi_firstChunkSize;
824 int cmi_otherChunkSize;
825 int cmi_cacheSize; /* The original cache size, in 1K blocks. */
826 unsigned cmi_setTime:1;
827 unsigned cmi_memCache:1;
828 int spare[16 - 9]; /* size of struct is 16 * 4 = 64 bytes */
832 /*----------------------------------------------------------------------
833 * AFS Data cache definitions
835 * Each entry describes a Unix file on the local disk that is
836 * is serving as a cached copy of all or part of a Vice file.
837 * Entries live in circular queues for each hash table slot
839 * Which queue is this thing in? Good question.
840 * A struct dcache entry is in the freeDSlot queue when not associated with a cache slot (file).
841 * Otherwise, it is in the DLRU queue. The freeDSlot queue uses the lruq.next field as
842 * its "next" pointer.
844 * Cache entries in the DLRU queue are either associated with vice files, in which case
845 * they are hashed by afs_dvnextTbl and afs_dcnextTbl pointers, or they are in the freeDCList
846 * and are not associated with any vice file. This last list uses the afs_dvnextTbl pointer for
847 * its "next" pointer.
848 *----------------------------------------------------------------------*/
850 #define NULLIDX (-1) /* null index definition */
851 /* struct dcache states bits */
855 #define DWriting 8 /* file being written (used for cache validation) */
857 /* dcache data flags */
858 #define DFEntryMod 0x02 /* has entry itself been modified? */
859 #define DFFetching 0x04 /* file is currently being fetched */
861 /* dcache meta flags */
862 #define DFNextStarted 0x01 /* next chunk has been prefetched already */
863 #define DFFetchReq 0x10 /* someone is waiting for DFFetching to go on */
866 /* flags in afs_indexFlags array */
867 #define IFEverUsed 1 /* index entry has >= 1 byte of data */
868 #define IFFree 2 /* index entry in freeDCList */
869 #define IFDataMod 4 /* file needs to be written out */
870 #define IFFlag 8 /* utility flag */
871 #define IFDirtyPages 16
872 #define IFAnyPages 32
873 #define IFDiscarded 64 /* index entry in discardDCList */
876 char *in; /* input buffer */
877 char *out; /* output buffer */
878 short in_size; /* Size of input buffer <= 2K */
879 short out_size; /* Maximum size of output buffer, <= 2K */
883 * This version of afs_ioctl is required to pass in 32 bit user space
884 * pointers into a 64 bit kernel.
895 /* CacheItems file has a header of type struct afs_fheader
896 * (keep aligned properly). Since we already have sgi_62 clients running
897 * with a 32 bit inode, a change is required to the header so that
898 * they can distinguish the old 32 bit inode CacheItems file and zap it
899 * instead of using it.
902 #define AFS_FHMAGIC 0x7635abaf /* uses version number */
904 #if defined(AFS_SUN57_64BIT_ENV)
905 #define AFS_CI_VERSION 3
907 #define AFS_CI_VERSION 2
910 afs_int32 firstCSize;
911 afs_int32 otherCSize;
914 #if defined(AFS_SGI61_ENV) || defined(AFS_SUN57_64BIT_ENV)
915 /* Using ino64_t here so that user level debugging programs compile
916 * the size correctly.
918 #define afs_inode_t ino64_t
920 #if defined(AFS_LINUX_64BIT_KERNEL) && !defined(AFS_S390X_LINUX24_ENV)
921 #define afs_inode_t long
923 #if defined(AFS_AIX51_ENV) || defined(AFS_HPUX1123_ENV)
924 #define afs_inode_t ino_t
926 #define afs_inode_t afs_int32
932 afs_int32 fid; /* is adc->index, the cache file number */
933 afs_inode_t inode; /* is adc->f.inode, the inode number of the cac\
936 afs_int32 accesstime;
937 struct buffer *hashNext;
945 afs_rwlock_t lock; /* the lock for this structure */
948 /* kept on disk and in dcache entries */
950 struct VenusFid fid; /* Fid for this file */
951 afs_int32 modTime; /* last time this entry was modified */
952 afs_hyper_t versionNo; /* Associated data version number */
953 afs_int32 chunk; /* Relative chunk number */
954 afs_inode_t inode; /* Unix inode for this chunk */
955 afs_int32 chunkBytes; /* Num bytes in this chunk */
956 char states; /* Has this chunk been modified? */
959 /* magic numbers to specify the cache type */
961 #define AFS_FCACHE_TYPE_UFS 0x0
962 #define AFS_FCACHE_TYPE_MEM 0x1
963 #define AFS_FCACHE_TYPE_NFS 0x2
964 #define AFS_FCACHE_TYPE_EPI 0x3
968 struct afs_q lruq; /* Free queue for in-memory images */
969 struct afs_q dirty; /* Queue of dirty entries that need written */
970 afs_rwlock_t lock; /* Protects validPos, some f */
971 afs_rwlock_t tlock; /* Atomizes updates to refCount */
972 afs_rwlock_t mflock; /* Atomizes accesses/updates to mflags */
973 afs_size_t validPos; /* number of valid bytes during fetch */
974 afs_int32 index; /* The index in the CacheInfo file */
975 short refCount; /* Associated reference count. */
976 char dflags; /* Data flags */
977 char mflags; /* Meta flags */
978 struct fcache f; /* disk image */
979 afs_int32 stamp; /* used with vtodc struct for hints */
980 afs_int32 bucket; /* which bucket these dcache entries are in */
984 * dcache.lock protects the actual contents of the cache file (in
985 * f.inode), subfields of f except those noted below, dflags and
988 * dcache.tlock is used to make atomic updates to refCount. Zero
989 * refCount dcache entries are protected by afs_xdcache instead of
992 * dcache.mflock is used to access and update mflags. It cannot be
993 * held without holding the corresponding dcache.lock. Updating
994 * mflags requires holding dcache.lock(R) and dcache.mflock(W), and
995 * checking for mflags requires dcache.lock(R) and dcache.mflock(R).
996 * Note that dcache.lock(W) gives you the right to update mflags,
997 * as dcache.mflock(W) can only be held with dcache.lock(R).
999 * dcache.stamp is protected by the associated vcache lock, because
1000 * it's only purpose is to establish correspondence between vcache
1001 * and dcache entries.
1003 * dcache.index, dcache.f.fid, dcache.f.chunk and dcache.f.inode are
1004 * write-protected by afs_xdcache and read-protected by refCount.
1005 * Once an entry is referenced, these values cannot change, and if
1006 * it's on the free list (with refCount=0), it can be reused for a
1007 * different file/chunk. These values can only be written while
1008 * holding afs_xdcache(W) and allocating this dcache entry (thereby
1009 * ensuring noone else has a refCount on it).
1012 /* this is obsolete and should be removed */
1015 /* afs_memcache.c */
1016 struct memCacheEntry {
1017 int size; /* # of valid bytes in this entry */
1018 int dataSize; /* size of allocated data area */
1019 afs_lock_t afs_memLock;
1020 char *data; /* bytes */
1023 /* macro to mark a dcache entry as bad */
1026 (x)->f.fid.Fid.Unique = 0; \
1027 afs_indexUnique[(x)->index] = 0; \
1028 (x)->dflags |= DFEntryMod; \
1031 /* FakeOpen and Fake Close used to be real subroutines. They're only used in
1032 * sun_subr and afs_vnodeops, and they're very frequently called, so I made
1033 * them into macros. They do:
1034 * FakeOpen: fake the file being open for writing. avc->lock must be held
1035 * in write mode. Having the file open for writing is like having a DFS
1036 * write-token: you're known to have the best version of the data around,
1037 * and so the CM won't let it be overwritten by random server info.
1038 * FakeClose: undo the effects of FakeOpen, noting that we want to ensure
1039 * that a real close eventually gets done. We use CCore to achieve this if
1040 * we would end up closing the file. avc->lock must be held in write mode */
1042 #ifdef AFS_AIX_IAUTH_ENV
1043 #define CRKEEP(V, C) (V)->linkData = (char*)crdup((C))
1045 #define CRKEEP(V, C) crhold((C)); (V)->linkData = (char*)(C)
1048 #define afs_FakeOpen(avc) { avc->opens++; avc->execsOrWriters++; }
1049 #define afs_FakeClose(avc, acred) \
1050 { if (avc->execsOrWriters == 1) { \
1051 /* we're the last writer, just use CCore flag */ \
1052 avc->states |= CCore; /* causes close to be called later */ \
1054 /* The cred and vnode holds will be released in afs_FlushActiveVcaches */ \
1055 VN_HOLD(AFSTOV(avc)); /* So it won't disappear */ \
1056 CRKEEP(avc, acred); /* Should use a better place for the creds */ \
1059 /* we're not the last writer, let the last one do the store-back for us */ \
1061 avc->execsOrWriters--; \
1065 #define AFS_ZEROS 64 /* zero buffer */
1067 /*#define afs_DirtyPages(avc) (((avc)->states & CDirty) || osi_VMDirty_p((avc)))*/
1068 #define afs_DirtyPages(avc) ((avc)->states & CDirty)
1070 /* The PFlush algorithm makes use of the fact that Fid.Unique is not used in
1071 below hash algorithms. Change it if need be so that flushing algorithm
1072 doesn't move things from one hash chain to another
1074 /* extern int afs_dhashsize; */
1075 #define DCHash(v, c) ((((v)->Fid.Vnode + (v)->Fid.Volume + (c))) & (afs_dhashsize-1))
1076 /*Vnode, Chunk -> Hash table index */
1077 #define DVHash(v) ((((v)->Fid.Vnode + (v)->Fid.Volume )) & (afs_dhashsize-1))
1078 /*Vnode -> Other hash table index */
1079 /* don't hash on the cell, our callback-breaking code sometimes fails to compute
1080 the cell correctly, and only scans one hash bucket */
1081 #define VCHash(fid) (((fid)->Fid.Volume + (fid)->Fid.Vnode) & (VCSIZE-1))
1083 extern struct dcache **afs_indexTable; /*Pointers to in-memory dcache entries */
1084 extern afs_int32 *afs_indexUnique; /*dcache entry Fid.Unique */
1085 extern afs_int32 *afs_dvnextTbl; /*Dcache hash table links */
1086 extern afs_int32 *afs_dcnextTbl; /*Dcache hash table links */
1087 extern afs_int32 afs_cacheFiles; /*Size of afs_indexTable */
1088 extern afs_int32 afs_cacheBlocks; /*1K blocks in cache */
1089 extern afs_int32 afs_cacheStats; /*Stat entries in cache */
1090 extern struct vcache *afs_vhashT[VCSIZE]; /*Stat cache hash table */
1091 extern afs_int32 afs_initState; /*Initialization state */
1092 extern afs_int32 afs_termState; /* Termination state */
1093 extern struct VenusFid afs_rootFid; /*Root for whole file system */
1094 extern afs_int32 afs_allCBs; /* Count of callbacks */
1095 extern afs_int32 afs_oddCBs; /* Count of odd callbacks */
1096 extern afs_int32 afs_evenCBs; /* Count of even callbacks */
1097 extern afs_int32 afs_allZaps; /* Count of fid deletes */
1098 extern afs_int32 afs_oddZaps; /* Count of odd fid deletes */
1099 extern afs_int32 afs_evenZaps; /* Count of even fid deletes */
1100 extern struct brequest afs_brs[NBRS]; /* request structures */
1102 #define UHash(auid) ((auid) & (NUSERS-1))
1103 #define VHash(avol) ((avol)&(NVOLS-1))
1104 #define SHash(aserv) ((ntohl(aserv)) & (NSERVERS-1))
1105 #define FVHash(acell,avol) (((avol)+(acell)) & (NFENTRIES-1))
1107 /* Performance hack - we could replace VerifyVCache2 with the appropriate
1108 * GetVCache incantation, and could eliminate even this code from afs_UFSRead
1109 * by making intentionally invalidating quick.stamp in the various callbacks
1110 * expiration/breaking code */
1111 #ifdef AFS_LINUX20_ENV
1112 #define afs_VerifyVCache(avc, areq) \
1113 (((avc)->states & CStatd) ? (vcache2inode(avc), 0) : \
1114 afs_VerifyVCache2((avc),areq))
1116 #ifdef AFS_DARWIN_ENV
1117 #define afs_VerifyVCache(avc, areq) \
1118 (((avc)->states & CStatd) ? (osi_VM_Setup(avc, 0), 0) : \
1119 afs_VerifyVCache2((avc),areq))
1121 #define afs_VerifyVCache(avc, areq) \
1122 (((avc)->states & CStatd) ? 0 : afs_VerifyVCache2((avc),areq))
1126 #define DO_STATS 1 /* bits used by FindVCache */
1129 /* values for flag param of afs_CheckVolumeNames */
1130 #define AFS_VOLCHECK_EXPIRED 0x1 /* volumes whose callbacks have expired */
1131 #define AFS_VOLCHECK_BUSY 0x2 /* volumes which were marked busy */
1132 #define AFS_VOLCHECK_MTPTS 0x4 /* mount point invalidation also */
1133 #define AFS_VOLCHECK_FORCE 0x8 /* do all forcibly */
1137 #define AFS_FSPORT ((unsigned short) htons(7000))
1138 #define AFS_VLPORT ((unsigned short) htons(7003))
1140 #define afs_read(avc, uio, acred, albn, abpp, nolock) \
1141 (*(afs_cacheType->vread))(avc, uio, acred, albn, abpp, nolock)
1142 #define afs_write(avc, uio, aio, acred, nolock) \
1143 (*(afs_cacheType->vwrite))(avc, uio, aio, acred, nolock)
1145 #define afs_rdwr(avc, uio, rw, io, cred) \
1146 (((rw) == UIO_WRITE) ? afs_write(avc, uio, io, cred, 0) : afs_read(avc, uio, cred, 0, 0, 0))
1147 #define afs_nlrdwr(avc, uio, rw, io, cred) \
1148 (((rw) == UIO_WRITE) ? afs_write(avc, uio, io, cred, 1) : afs_read(avc, uio, cred, 0, 0, 1))
1150 /* Cache size truncation uses the following low and high water marks:
1151 * If the cache is more than 95% full (CM_DCACHECOUNTFREEPCT), the cache
1152 * truncation daemon is awakened and will free up space until the cache is 85%
1153 * (CM_DCACHESPACEFREEPCT - CM_DCACHEEXTRAPCT) full.
1154 * afs_UFSWrite and afs_GetDCache (when it needs to fetch data) will wait on
1155 * afs_WaitForCacheDrain if the cache is 98% (CM_WAITFORDRAINPCT) full.
1156 * afs_GetDownD wakes those processes once the cache is 95% full
1157 * (CM_CACHESIZEDRAINEDPCT).
1159 #define CM_MAXDISCARDEDCHUNKS 16 /* # of chunks */
1160 #define CM_DCACHECOUNTFREEPCT 95 /* max pct of chunks in use */
1161 #define CM_DCACHESPACEFREEPCT 90 /* max pct of space in use */
1162 #define CM_DCACHEEXTRAPCT 5 /* extra to get when freeing */
1163 #define CM_CACHESIZEDRAINEDPCT 95 /* wakeup processes when down to here. */
1164 #define CM_WAITFORDRAINPCT 98 /* sleep if cache is this full. */
1166 #define afs_CacheIsTooFull() \
1167 (afs_blocksUsed - afs_blocksDiscarded > \
1168 (CM_DCACHECOUNTFREEPCT*afs_cacheBlocks)/100 || \
1169 afs_freeDCCount - afs_discardDCCount < \
1170 ((100-CM_DCACHECOUNTFREEPCT)*afs_cacheFiles)/100)
1172 /* Handy max length of a numeric string. */
1173 #define CVBS 12 /* max afs_int32 is 2^32 ~ 4*10^9, +1 for NULL, +luck */
1175 #define refpanic(foo) if (afs_norefpanic) \
1176 { printf( foo ); afs_norefpanic++;} else osi_Panic( foo )
1179 ** these are defined in the AIX source code sys/fs_locks.h but are not
1180 ** defined anywhere in the /usr/include directory
1182 #if defined(AFS_AIX41_ENV)
1183 #define VN_LOCK(vp) simple_lock(&(vp)->v_lock)
1184 #define VN_UNLOCK(vp) simple_unlock(&(vp)->v_lock)
1187 /* get a file's serial number from a vnode */
1188 #ifndef afs_vnodeToInumber
1189 #if defined(AFS_SGI62_ENV) || defined(AFS_HAVE_VXFS) || defined(AFS_DARWIN_ENV)
1190 #define afs_vnodeToInumber(V) VnodeToIno(V)
1192 #ifdef AFS_DECOSF_ENV
1193 #define afs_vnodeToInumber(V) osi_vnodeToInumber(V)
1195 #define afs_vnodeToInumber(V) (VTOI(V)->i_number)
1196 #endif /* AFS_DECOSF_ENV */
1197 #endif /* AFS_SGI62_ENV */
1200 /* get a file's device number from a vnode */
1201 #ifndef afs_vnodeToDev
1202 #if defined(AFS_SGI62_ENV) || defined(AFS_HAVE_VXFS) || defined(AFS_DARWIN_ENV)
1203 #define afs_vnodeToDev(V) VnodeToDev(V)
1204 #elif defined(AFS_DECOSF_ENV)
1205 #define afs_vnodeToDev(V) osi_vnodeToDev(V)
1207 #define afs_vnodeToDev(V) (VTOI(V)->i_dev)
1212 /* Note: this should agree with the definition in kdump.c */
1213 #if defined(AFS_OSF_ENV)
1214 #if !defined(UKERNEL)
1215 #define AFS_USEBUFFERS 1
1219 #if !defined(UKERNEL) && !defined(HAVE_STRUCT_BUF)
1220 /* declare something so that prototypes don't flip out */
1221 /* appears struct buf stuff is only actually passed around as a pointer,
1222 except with libuafs, in which case it is actually defined */
1227 /* fakestat support: opaque storage for afs_EvalFakeStat to remember
1228 * what vcache should be released.
1230 struct afs_fakestat_state {
1234 struct vcache *root_vp;
1237 extern int afs_fakestat_enable;
1239 /* First 32 bits of capabilities */
1240 #define CAPABILITY_ERRORTRANS (1<<0)
1242 #define CAPABILITY_BITS 1
1244 #endif /* _AFS_H_ */