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 short realcellIndex; /* as above but ignoring aliases */
205 time_t timeout; /* data expire time, if non-zero */
206 char *realName; /* who this cell is an alias for */
209 #define afs_PutCell(cellp, locktype)
211 /* the unixuser flag bit definitions */
212 #define UHasTokens 1 /* are the st and ct fields valid (ever set)? */
213 #define UTokensBad 2 /* are tokens bad? */
214 #define UPrimary 4 /* on iff primary identity */
215 #define UNeedsReset 8 /* needs afs_ResetAccessCache call done */
216 #define UPAGCounted 16 /* entry seen during PAG search (for stats) */
217 /* A flag used by afs_GCPAGs to keep track of
218 * which entries in afs_users need to be deleted.
219 * The lifetime of its presence in the table is the
220 * lifetime of the afs_GCPAGs function.
222 #define TMP_UPAGNotReferenced 128
224 /* values for afs_gcpags */
226 { AFS_GCPAGS_NOTCOMPILED = 0
228 , AFS_GCPAGS_USERDISABLED
232 , AFS_GCPAGS_EINEXACT
233 , AFS_GCPAGS_EPROCEND
234 , AFS_GCPAGS_EPROCWALK
235 , AFS_GCPAGS_ECREDWALK
236 , AFS_GCPAGS_EPIDCHECK
237 , AFS_GCPAGS_ENICECHECK
240 extern afs_int32 afs_gcpags;
241 extern afs_int32 afs_gcpags_procsize;
244 struct unixuser *next; /* next hash pointer */
245 afs_int32 uid; /* search based on uid and cell */
247 afs_int32 vid; /* corresponding vice id in specified cell */
248 short refCount; /* reference count for allocation */
249 char states; /* flag info */
250 afs_int32 tokenTime; /* last time tokens were set, used for timing out conn data */
251 afs_int32 stLen; /* ticket length (if kerberos, includes kvno at head) */
252 char *stp; /* pointer to ticket itself */
253 struct ClearToken ct;
254 struct afs_exporter *exporter; /* more info about the exporter for the remote user */
258 /* Per-connection block. */
259 struct conn *next; /* Next dude same server. */
260 struct unixuser *user; /* user validated with respect to. */
261 struct rx_connection *id; /* RPC connid. */
262 struct srvAddr *srvr; /* server associated with this conn */
263 short refCount; /* reference count for allocation */
264 unsigned short port; /* port associated with this connection */
265 char forceConnectFS; /* Should we try again with these tokens? */
271 /* Fid comparison routines */
272 #define FidCmp(a,b) ((a)->Fid.Unique != (b)->Fid.Unique \
273 || (a)->Fid.Vnode != (b)->Fid.Vnode \
274 || (a)->Fid.Volume != (b)->Fid.Volume \
275 || (a)->Cell != (b)->Cell)
277 #define FidMatches(afid,tvc) ((tvc)->fid.Fid.Vnode == (afid)->Fid.Vnode && \
278 (tvc)->fid.Fid.Volume == (afid)->Fid.Volume && \
279 (tvc)->fid.Cell == (afid)->Cell && \
280 ( (tvc)->fid.Fid.Unique == (afid)->Fid.Unique || \
281 (!(afid)->Fid.Unique && ((tvc)->states & CUnique))))
285 * Operations on circular queues implemented with pointers. Note: these queue
286 * objects are always located at the beginning of the structures they are linking.
288 #define QInit(q) ((q)->prev = (q)->next = (q))
289 #define QAdd(q,e) ((e)->next = (q)->next, (e)->prev = (q), \
290 (q)->next->prev = (e), (q)->next = (e))
291 #define QRemove(e) ((e)->next->prev = (e)->prev, (e)->prev->next = (e)->next)
292 #define QNext(e) ((e)->next)
293 #define QPrev(e) ((e)->prev)
294 #define QEmpty(q) ((q)->prev == (q))
295 /* this one takes q1 and sticks it on the end of q2 - that is, the other end, not the end
296 * that things are added onto. q1 shouldn't be empty, it's silly */
297 #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))
299 * Do lots of address arithmetic to go from vlruq to the base of the vcache
300 * structure. Don't move struct vnode, since we think of a struct vcache as
301 * a specialization of a struct vnode
303 #define QTOV(e) ((struct vcache *)(((char *) (e)) - (((char *)(&(((struct vcache *)(e))->vlruq))) - ((char *)(e)))))
304 #define QTOC(e) ((struct cell *)((char *) (e)))
307 #define SRVADDR_ISDOWN 0x20 /* same as SRVR_ISDOWN */
308 #define SRVADDR_NOUSE 0x40 /* Don't use this srvAddr */
310 struct srvAddr *next_bkt; /* next item in hash bucket */
311 struct srvAddr *next_sa; /* another interface on same host */
312 struct server *server; /* back to parent */
313 struct conn *conns; /* All user connections to this server */
314 afs_int32 sa_ip; /* Host addr in network byte order */
315 u_short sa_iprank; /* indiv ip address priority */
316 u_short sa_portal; /* port addr in network byte order */
321 * Values used in the flags field of the server structure below.
323 * AFS_SERVER_FLAG_ACTIVATED Has the server ever had a user connection
324 * associated with it?
326 #define AFS_SERVER_FLAG_ACTIVATED 0x01
327 #define SNO_LHOSTS 0x04
328 #define SYES_LHOSTS 0x08
329 #define SVLSRV_UUID 0x10
330 #define SRVR_ISDOWN 0x20
331 #define SRVR_MULTIHOMED 0x40
332 #define SRVR_ISGONE 0x80
333 #define SNO_INLINEBULK 0x100
334 #define SNO_64BIT 0x200
336 #define afs_serverSetNo64Bit(s) ((s)->srvr->server->flags |= SNO_64BIT)
337 #define afs_serverHasNo64Bit(s) ((s)->srvr->server->flags & SNO_64BIT)
343 afs_int32 addr_uniquifier;
347 struct srvAddr haddr;
350 #define sr_uuid _suid._srvUuid.suuid
351 #define sr_addr_uniquifier _suid._srvUuid.addr_uniquifier
352 #define sr_host _suid._srvId.haddr.ip
353 #define sr_portal _suid._srvId.haddr.portal
354 #define sr_rank _suid._srvId.haddr.ip_rank
355 #define sr_flags _suid._srvId.haddr.flags
356 #define sr_conns _suid._srvId.haddr.conns
357 struct server *next; /* Ptr to next server in hash chain */
358 struct cell *cell; /* Cell in which this host resides */
359 struct afs_cbr *cbrs; /* Return list of callbacks */
360 afs_int32 activationTime; /* Time when this record was first activated */
361 afs_int32 lastDowntimeStart; /* Time when last downtime incident began */
362 afs_int32 numDowntimeIncidents; /* # (completed) downtime incidents */
363 afs_int32 sumOfDowntimes; /* Total downtime experienced, in seconds */
364 struct srvAddr *addr;
365 afs_uint32 flags; /* Misc flags*/
368 #define afs_PutServer(servp, locktype)
370 /* structs for some pioctls - these are (or should be)
378 struct sprefrequest_33 {
379 unsigned short offset;
380 unsigned short num_servers;
384 struct sprefrequest { /* new struct for 3.4 */
385 unsigned short offset;
386 unsigned short num_servers;
387 unsigned short flags;
392 unsigned short next_offset;
393 unsigned short num_servers;
394 struct spref servers[1]; /* we overrun this array intentionally...*/
398 unsigned short flags;
399 unsigned short num_servers;
400 struct spref servers[1]; /* we overrun this array intentionally...*/
402 /* struct for GAG pioctl
405 afs_uint32 showflags, logflags, logwritethruflag, spare[3];
406 unsigned char spare2[128];
410 #define logwritethruON 1
413 afs_int32 rx_initReceiveWindow, rx_maxReceiveWindow,
414 rx_initSendWindow, rx_maxSendWindow, rxi_nSendFrags,
415 rxi_nRecvFrags, rxi_OrphanFragSize;
416 afs_int32 rx_maxReceiveSize, rx_MyMaxSendSize;
417 afs_uint32 spare[21];
420 /* struct for checkservers */
433 /* state bits for volume */
434 #define VRO 1 /* volume is readonly */
435 #define VRecheck 2 /* recheck volume info with server */
436 #define VBackup 4 /* is this a backup volume? */
437 #define VForeign 8 /* this is a non-afs volume */
438 #define VResort 16 /* server order was rearranged, sort when able */
439 #define VMoreReps 32 /* This volume has more replicas than we are */
440 /* keeping track of now -- check with VLDB */
442 enum repstate { not_busy, end_not_busy = 6, rd_busy, rdwr_busy, offline };
445 /* One structure per volume, describing where the volume is located
446 and where its mount points are. */
447 struct volume *next; /* Next volume in hash list. */
448 afs_int32 cell; /* the cell in which the volume resides */
449 afs_rwlock_t lock; /* the lock for this structure */
450 afs_int32 volume; /* This volume's ID number. */
451 char *name; /* This volume's name, or 0 if unknown */
452 struct server *serverHost[MAXHOSTS]; /* servers serving this volume */
453 enum repstate status[MAXHOSTS]; /* busy, offline, etc */
454 struct VenusFid dotdot; /* dir to access as .. */
455 struct VenusFid mtpoint; /* The mount point for this volume. */
456 afs_int32 rootVnode, rootUnique; /* Volume's root fid */
459 afs_int32 rwVol; /* For r/o vols, original read/write volume. */
460 afs_int32 accessTime; /* last time we used it */
461 afs_int32 vtix; /* volume table index */
462 afs_int32 copyDate; /* copyDate field, for tracking vol releases */
463 afs_int32 expireTime; /* for per-volume callbacks... */
464 short refCount; /* reference count for allocation */
465 char states; /* here for alignment reasons */
468 #define afs_PutVolume(av, locktype) ((av)->refCount--)
470 /* format of an entry in volume info file */
472 afs_int32 cell; /* cell for this entry */
473 afs_int32 volume; /* volume */
474 afs_int32 next; /* has index */
475 struct VenusFid dotdot; /* .. value */
476 struct VenusFid mtpoint; /* mt point's fid */
477 afs_int32 rootVnode, rootUnique; /* Volume's root fid */
481 struct SimpleLocks *next;
483 afs_int32 boff, eoff;
485 #if defined(AFS_AIX32_ENV) || defined(AFS_SUN5_ENV) || defined(AFS_SGI_ENV)
490 /* vcache state bits */
491 #define CStatd 0x00000001 /* has this file ever been stat'd? */
492 #define CBackup 0x00000002 /* file is on a backup volume */
493 #define CRO 0x00000004 /* is it on a read-only volume */
494 #define CMValid 0x00000008 /* is the mount point info valid? */
495 #define CCore 0x00000010 /* storing a core file, needed since we don't get an open */
496 #define CDirty 0x00000020 /* file has been modified since first open (... O_RDWR) */
497 #define CSafeStore 0x00000040 /* close must wait for store to finish (should be in fd) */
498 #define CMAPPED 0x00000080 /* Mapped files; primarily used by SunOS 4.0.x */
499 #define CNSHARE 0x00000100 /* support O_NSHARE semantics */
500 #define CLied 0x00000200
501 #define CTruth 0x00000400
503 #define CWired 0x00000800 /* OSF hack only */
505 #define CWRITE_IGN 0x00000800 /* Next OS hack only */
507 #define CUnique 0x00001000 /* vc's uniquifier - latest unifiquier for fid */
508 #define CForeign 0x00002000 /* this is a non-afs vcache */
509 #define CUnlinked 0x00010000
510 #define CBulkStat 0x00020000 /* loaded by a bulk stat, and not ref'd since */
511 #define CUnlinkedDel 0x00040000
512 #define CVFlushed 0x00080000
513 #define CCore1 0x00100000 /* osf1 core file; not same as CCore above */
514 #define CWritingUFS 0x00200000 /* to detect vm deadlock - used by sgi */
515 #define CCreating 0x00400000 /* avoid needless store after open truncate */
516 #define CPageHog 0x00800000 /* AIX - dumping large cores is a page hog. */
517 #define CDCLock 0x02000000 /* Vnode lock held over call to GetDownD */
518 #define CBulkFetching 0x04000000 /* stats are being fetched by bulk stat */
519 #define CExtendedFile 0x08000000 /* extended file via ftruncate call. */
521 /* vcache vstate bits */
522 #define VRevokeWait 0x1
523 #define VPageCleaning 0x2 /* Solaris - Cache Trunc Daemon sez keep out */
526 #if !defined(AFS_FBSD_ENV)
527 #define vrefCount v.v_count
529 #define vrefCount v.v_usecount
530 #endif /* AFS_FBSD_ENV */
532 #ifdef AFS_LINUX24_ENV
533 #define VREFCOUNT(v) atomic_read(&((vnode_t *) v)->v_count)
534 #define VREFCOUNT_SET(v, c) atomic_set(&((vnode_t *) v)->v_count, c)
535 #define VREFCOUNT_DEC(v) atomic_dec(&((vnode_t *) v)->v_count)
536 #define VREFCOUNT_INC(v) atomic_inc(&((vnode_t *) v)->v_count)
537 #define DLOCK() spin_lock(&dcache_lock)
538 #define DUNLOCK() spin_unlock(&dcache_lock)
539 #define DGET(d) dget_locked(d)
540 #define DCOUNT(d) atomic_read(&(d)->d_count)
542 #define VREFCOUNT(v) ((v)->vrefCount)
543 #define VREFCOUNT_SET(v, c) (v)->vrefCount = c;
544 #define VREFCOUNT_DEC(v) (v)->vrefCount--;
545 #define VREFCOUNT_INC(v) (v)->vrefCount++;
548 #define DGET(d) dget(d)
549 #define DCOUNT(d) ((d)->d_count)
552 #define AFS_MAXDV 0x7fffffff /* largest dataversion number */
553 #ifdef AFS_64BIT_CLIENT
554 #define AFS_NOTRUNC 0x7fffffffffffffffLL /* largest positive int64 number */
555 #else /* AFS_64BIT_CLIENT */
556 #define AFS_NOTRUNC 0x7fffffff /* largest dataversion number */
557 #endif /* AFS_64BIT_CLIENT */
559 extern afs_int32 vmPageHog; /* counter for # of vnodes which are page hogs. */
562 * Fast map from vcache to dcache
569 afs_offs_t minLoc; /* smallest offset into dc. */
570 afs_offs_t len; /* largest offset into dc. */
573 extern afs_uint32 afs_stampValue; /* stamp for pair's usage */
574 #define MakeStamp() (++afs_stampValue)
576 #define VTOAFS(V) ((struct vcache*)(V))
577 #define AFSTOV(V) (&(V)->v)
578 #ifdef AFS_LINUX22_ENV
579 #define ITOAFS(V) ((struct vcache*)(V))
580 #define AFSTOI(V) (struct inode *)(&(V)->v)
583 /* INVARIANTs: (vlruq.next != NULL) == (vlruq.prev != NULL)
584 * nextfree => !vlruq.next && ! vlruq.prev
585 * !(avc->nextfree) && !avc->vlruq.next => (FreeVCList == avc->nextfree)
588 struct vnode v; /* Has reference count in v.v_count */
589 struct afs_q vlruq; /* lru q next and prev */
590 struct vcache *nextfree; /* next on free list (if free) */
591 struct vcache *hnext; /* Hash next */
595 afs_hyper_t DataVersion;
599 ushort Mode; /* XXXX Should be afs_int32 XXXX */
601 /* vnode type is in v.v_type */
603 afs_rwlock_t lock; /* The lock on the vcache contents. */
604 #if defined(AFS_SUN5_ENV)
605 /* Lock used to protect the activeV, multipage, and vstates fields.
606 * Do not try to get the vcache lock when the vlock is held */
608 #endif /* defined(AFS_SUN5_ENV) */
609 #if defined(AFS_SUN5_ENV)
613 #if defined(AFS_SUN_ENV) || defined(AFS_ALPHA_ENV) || defined(AFS_DARWIN_ENV) || defined(AFS_FBSD_ENV)
614 afs_bozoLock_t pvnLock; /* see locks.x */
623 int ownslock; /* pid of owner of excl lock, else 0 - defect 3083 */
625 #ifdef AFS_DARWIN_ENV
626 struct lock__bsd__ rwlock;
631 afs_int32 parentVnode; /* Parent dir, if a file. */
632 afs_int32 parentUnique;
633 struct VenusFid *mvid; /* Either parent dir (if root) or root (if mt pt) */
634 char *linkData; /* Link data if a symlink. */
635 afs_hyper_t flushDV; /* data version last flushed from text */
636 afs_hyper_t mapDV; /* data version last flushed from map */
637 afs_size_t truncPos; /* truncate file to this position at next store */
638 struct server *callback; /* The callback host, if any */
639 afs_uint32 cbExpires; /* time the callback expires */
640 struct afs_q callsort; /* queue in expiry order, sort of */
641 struct axscache *Access; /* a list of cached access bits */
642 afs_int32 anyAccess; /* System:AnyUser's access to this. */
643 afs_int32 last_looker; /* pag/uid from last lookup here */
644 #if defined(AFS_SUN5_ENV)
646 #endif /* defined(AFS_SUN5_ENV) */
647 struct SimpleLocks *slocks;
648 short opens; /* The numbers of opens, read or write, on this file. */
649 short execsOrWriters; /* The number of execs (if < 0) or writers (if > 0) of
651 short flockCount; /* count of flock readers, or -1 if writer */
652 char mvstat; /* 0->normal, 1->mt pt, 2->root. */
653 afs_uint32 states; /* state bits */
654 #if defined(AFS_SUN5_ENV)
655 afs_uint32 vstates; /* vstate bits */
656 #endif /* defined(AFS_SUN5_ENV) */
658 afs_uint32 symhintstamp;
660 struct vcache *symhint;
661 struct dcache *dchint;
663 #ifdef AFS_LINUX22_ENV
664 u_short flushcnt; /* Number of flushes which haven't released yet. */
665 u_short mapcnt; /* Number of mappings of this file. */
667 #if defined(AFS_SGI_ENV)
668 daddr_t lastr; /* for read-ahead */
670 uint64_t vc_rwlockid; /* kthread owning rwlock */
672 short vc_rwlockid; /* pid of process owning rwlock */
674 short vc_locktrips; /* # of rwlock reacquisitions */
675 sema_t vc_rwlock; /* vop_rwlock for afs */
676 pgno_t mapcnt; /* # of pages mapped */
677 struct cred *cred; /* last writer's cred */
679 struct bhv_desc vc_bhv_desc; /* vnode's behavior data. */
681 #endif /* AFS_SGI_ENV */
682 afs_int32 vc_error; /* stash write error for this vnode. */
683 int xlatordv; /* Used by nfs xlator */
684 struct AFS_UCRED *uncred;
685 int asynchrony; /* num kbytes to store behind */
687 short multiPage; /* count of multi-page getpages in progress */
691 #define afs_symhint_inval(avc)
694 #define DONT_CHECK_MODE_BITS 0
695 #define CHECK_MODE_BITS 1
696 #define CMB_ALLOW_EXEC_AS_READ 2 /* For the NFS xlator */
698 #if defined(AFS_SGI_ENV)
699 #define AVCRWLOCK(avc) (valusema(&(avc)->vc_rwlock) <= 0)
701 /* SGI vnode rwlock macros and flags. */
702 #ifndef AFS_SGI62_ENV
703 /* The following are defined here. SGI 6.2 declares them in vnode.h */
704 #define VRWLOCK_READ 0
705 #define VRWLOCK_WRITE 1
706 #define VRWLOCK_WRITE_DIRECT 2
711 #define AFS_RWLOCK_T vrwlock_t
713 #define AFS_RWLOCK_T int
714 #endif /* AFS_SGI62_ENV */
716 #include <ksys/behavior.h>
717 #define AFS_RWLOCK(V,F) \
718 afs_rwlock(&VTOAFS(V)->vc_bhv_desc, (F));
719 #define AFS_RWUNLOCK(V,F) \
720 afs_rwunlock(&VTOAFS(V)->vc_bhv_desc, (F));
723 #define AFS_RWLOCK(V,F) afs_rwlock((vnode_t *)(V), (F) )
724 #define AFS_RWUNLOCK(V,F) afs_rwunlock((vnode_t *)(V), (F) )
726 #else /* AFS_SGI53_ENV */
727 #define AFS_RWLOCK(V,F) afs_rwlock((V))
728 #define AFS_RWUNLOCK(V,F) afs_rwunlock((V))
729 #endif /* AFS_SGI53_ENV */
730 #endif /* AFS_SGI_ENV */
734 afs_hyper_t DataVersion;
736 afs_int32 parentVnode;
737 afs_int32 parentUnique;
741 afs_int32 randomUid[CPSIZE];
742 afs_int32 callback; /* Now a pointer to 'server' struct */
744 afs_int32 randomAccess[CPSIZE];
747 short execsOrWriters;
758 /* CM inititialization parameters. What CM actually used after calculations
759 * based on passed in arguments.
761 #define CMI_VERSION 1 /* increment when adding new fields. */
762 struct cm_initparams {
767 int cmi_nVolumeCaches;
768 int cmi_firstChunkSize;
769 int cmi_otherChunkSize;
770 int cmi_cacheSize; /* The original cache size, in 1K blocks. */
771 unsigned cmi_setTime:1;
772 unsigned cmi_memCache:1;
773 int spare[16-9]; /* size of struct is 16 * 4 = 64 bytes */
777 /*----------------------------------------------------------------------
778 * AFS Data cache definitions
780 * Each entry describes a Unix file on the local disk that is
781 * is serving as a cached copy of all or part of a Vice file.
782 * Entries live in circular queues for each hash table slot
784 * Which queue is this thing in? Good question.
785 * A struct dcache entry is in the freeDSlot queue when not associated with a cache slot (file).
786 * Otherwise, it is in the DLRU queue. The freeDSlot queue uses the lruq.next field as
787 * its "next" pointer.
789 * Cache entries in the DLRU queue are either associated with vice files, in which case
790 * they are hashed by afs_dvnextTbl and afs_dcnextTbl pointers, or they are in the freeDCList
791 * and are not associated with any vice file. This last list uses the afs_dvnextTbl pointer for
792 * its "next" pointer.
793 *----------------------------------------------------------------------*/
795 #define NULLIDX (-1) /* null index definition */
796 /* struct dcache states bits */
797 #define DWriting 8 /* file being written (used for cache validation) */
799 /* dcache data flags */
800 #define DFEntryMod 0x02 /* has entry itself been modified? */
801 #define DFFetching 0x04 /* file is currently being fetched */
803 /* dcache meta flags */
804 #define DFNextStarted 0x01 /* next chunk has been prefetched already */
805 #define DFFetchReq 0x10 /* someone is waiting for DFFetching to go on */
808 /* flags in afs_indexFlags array */
809 #define IFEverUsed 1 /* index entry has >= 1 byte of data */
810 #define IFFree 2 /* index entry in freeDCList */
811 #define IFDataMod 4 /* file needs to be written out */
812 #define IFFlag 8 /* utility flag */
813 #define IFDirtyPages 16
814 #define IFAnyPages 32
815 #define IFDiscarded 64 /* index entry in discardDCList */
818 char *in; /* input buffer */
819 char *out; /* output buffer */
820 short in_size; /* Size of input buffer <= 2K */
821 short out_size; /* Maximum size of output buffer, <= 2K */
825 * This version of afs_ioctl is required to pass in 32 bit user space
826 * pointers into a 64 bit kernel.
837 /* CacheItems file has a header of type struct afs_fheader
838 * (keep aligned properly). Since we already have sgi_62 clients running
839 * with a 32 bit inode, a change is required to the header so that
840 * they can distinguish the old 32 bit inode CacheItems file and zap it
841 * instead of using it.
844 #define AFS_FHMAGIC 0x7635abaf /* uses version number */
846 #if defined(AFS_SUN57_64BIT_ENV)
847 #define AFS_CI_VERSION 3
849 #define AFS_CI_VERSION 2
852 afs_int32 firstCSize;
853 afs_int32 otherCSize;
856 /* kept on disk and in dcache entries */
858 struct VenusFid fid; /* Fid for this file */
859 afs_int32 modTime; /* last time this entry was modified */
860 afs_hyper_t versionNo; /* Associated data version number */
861 afs_int32 chunk; /* Relative chunk number */
862 #if defined(AFS_SGI61_ENV) || defined(AFS_SUN57_64BIT_ENV)
863 /* Using ino64_t here so that user level debugging programs compile
864 * the size correctly.
866 ino64_t inode; /* Unix inode for this chunk */
868 #if defined(AFS_LINUX_64BIT_KERNEL)
869 long inode; /* Unix inode for this chunk */
871 afs_int32 inode; /* Unix inode for this chunk */
874 afs_int32 chunkBytes; /* Num bytes in this chunk */
875 char states; /* Has this chunk been modified? */
878 /* magic numbers to specify the cache type */
880 #define AFS_FCACHE_TYPE_UFS 0x0
881 #define AFS_FCACHE_TYPE_MEM 0x1
882 #define AFS_FCACHE_TYPE_NFS 0x2
883 #define AFS_FCACHE_TYPE_EPI 0x3
887 struct afs_q lruq; /* Free queue for in-memory images */
888 afs_rwlock_t lock; /* Protects validPos, some f */
889 afs_rwlock_t tlock; /* Atomizes updates to refCount */
890 afs_rwlock_t mflock; /* Atomizes accesses/updates to mflags */
891 afs_size_t validPos; /* number of valid bytes during fetch */
892 afs_int32 index; /* The index in the CacheInfo file*/
893 short refCount; /* Associated reference count. */
894 char dflags; /* Data flags */
895 char mflags; /* Meta flags */
896 struct fcache f; /* disk image */
897 afs_int32 stamp; /* used with vtodc struct for hints */
902 * dcache.lock protects the actual contents of the cache file (in
903 * f.inode), subfields of f except those noted below, dflags and
906 * dcache.tlock is used to make atomic updates to refCount. Zero
907 * refCount dcache entries are protected by afs_xdcache instead of
910 * dcache.mflock is used to access and update mflags. It cannot be
911 * held without holding the corresponding dcache.lock. Updating
912 * mflags requires holding dcache.lock(R) and dcache.mflock(W), and
913 * checking for mflags requires dcache.lock(R) and dcache.mflock(R).
914 * Note that dcache.lock(W) gives you the right to update mflags,
915 * as dcache.mflock(W) can only be held with dcache.lock(R).
917 * dcache.stamp is protected by the associated vcache lock, because
918 * it's only purpose is to establish correspondence between vcache
919 * and dcache entries.
921 * dcache.index, dcache.f.fid, dcache.f.chunk and dcache.f.inode are
922 * write-protected by afs_xdcache and read-protected by refCount.
923 * Once an entry is referenced, these values cannot change, and if
924 * it's on the free list (with refCount=0), it can be reused for a
925 * different file/chunk. These values can only be written while
926 * holding afs_xdcache(W) and allocating this dcache entry (thereby
927 * ensuring noone else has a refCount on it).
930 /* this is obsolete and should be removed */
933 /* macro to mark a dcache entry as bad */
936 (x)->f.fid.Fid.Unique = 0; \
937 afs_indexUnique[(x)->index] = 0; \
938 (x)->dflags |= DFEntryMod; \
941 /* FakeOpen and Fake Close used to be real subroutines. They're only used in
942 * sun_subr and afs_vnodeops, and they're very frequently called, so I made
943 * them into macros. They do:
944 * FakeOpen: fake the file being open for writing. avc->lock must be held
945 * in write mode. Having the file open for writing is like having a DFS
946 * write-token: you're known to have the best version of the data around,
947 * and so the CM won't let it be overwritten by random server info.
948 * FakeClose: undo the effects of FakeOpen, noting that we want to ensure
949 * that a real close eventually gets done. We use CCore to achieve this if
950 * we would end up closing the file. avc->lock must be held in write mode */
952 #ifdef AFS_AIX_IAUTH_ENV
953 #define CRKEEP(V, C) (V)->linkData = (char*)crdup((C))
955 #define CRKEEP(V, C) crhold((C)); (V)->linkData = (char*)(C)
958 #define afs_FakeOpen(avc) { avc->opens++; avc->execsOrWriters++; }
959 #define afs_FakeClose(avc, acred) \
960 { if (avc->execsOrWriters == 1) { \
961 /* we're the last writer, just use CCore flag */ \
962 avc->states |= CCore; /* causes close to be called later */ \
964 /* The cred and vnode holds will be released in afs_FlushActiveVcaches */ \
965 VN_HOLD(AFSTOV(avc)); /* So it won't disappear */ \
966 CRKEEP(avc, acred); /* Should use a better place for the creds */ \
969 /* we're not the last writer, let the last one do the store-back for us */ \
971 avc->execsOrWriters--; \
975 #define AFS_ZEROS 64 /* zero buffer */
977 /*#define afs_DirtyPages(avc) (((avc)->states & CDirty) || osi_VMDirty_p((avc)))*/
978 #define afs_DirtyPages(avc) ((avc)->states & CDirty)
980 /* The PFlush algorithm makes use of the fact that Fid.Unique is not used in
981 below hash algorithms. Change it if need be so that flushing algorithm
982 doesn't move things from one hash chain to another
984 /* extern int afs_dhashsize; */
985 #define DCHash(v, c) ((((v)->Fid.Vnode + (v)->Fid.Volume + (c))) & (afs_dhashsize-1))
986 /*Vnode, Chunk -> Hash table index*/
987 #define DVHash(v) ((((v)->Fid.Vnode + (v)->Fid.Volume )) & (afs_dhashsize-1))
988 /*Vnode -> Other hash table index*/
989 /* don't hash on the cell, our callback-breaking code sometimes fails to compute
990 the cell correctly, and only scans one hash bucket */
991 #define VCHash(fid) (((fid)->Fid.Volume + (fid)->Fid.Vnode) & (VCSIZE-1))
993 extern struct dcache **afs_indexTable; /*Pointers to in-memory dcache entries*/
994 extern afs_int32 *afs_indexUnique; /*dcache entry Fid.Unique */
995 extern afs_int32 *afs_dvnextTbl; /*Dcache hash table links */
996 extern afs_int32 *afs_dcnextTbl; /*Dcache hash table links */
997 extern afs_int32 afs_cacheFiles; /*Size of afs_indexTable*/
998 extern afs_int32 afs_cacheBlocks; /*1K blocks in cache*/
999 extern afs_int32 afs_cacheStats; /*Stat entries in cache*/
1000 extern struct vcache *afs_vhashT[VCSIZE]; /*Stat cache hash table*/
1001 extern afs_int32 afs_initState; /*Initialization state*/
1002 extern afs_int32 afs_termState; /* Termination state */
1003 extern struct VenusFid afs_rootFid; /*Root for whole file system*/
1004 extern afs_int32 afs_allCBs; /* Count of callbacks*/
1005 extern afs_int32 afs_oddCBs; /* Count of odd callbacks*/
1006 extern afs_int32 afs_evenCBs; /* Count of even callbacks*/
1007 extern afs_int32 afs_allZaps; /* Count of fid deletes */
1008 extern afs_int32 afs_oddZaps; /* Count of odd fid deletes */
1009 extern afs_int32 afs_evenZaps; /* Count of even fid deletes */
1010 extern struct brequest afs_brs[NBRS]; /* request structures */
1012 #define UHash(auid) ((auid) & (NUSERS-1))
1013 #define VHash(avol) ((avol)&(NVOLS-1))
1014 #define SHash(aserv) ((ntohl(aserv)) & (NSERVERS-1))
1015 #define FVHash(acell,avol) (((avol)+(acell)) & (NFENTRIES-1))
1017 extern struct cell *afs_GetCell();
1018 extern struct cell *afs_GetCellNoLock();
1019 extern struct cell *afs_GetCellByName();
1020 extern struct cell *afs_FindCellByName();
1021 extern struct cell *afs_GetCellByIndex();
1022 extern struct unixuser *afs_GetUser();
1023 extern struct volume *afs_GetVolume();
1024 extern struct volume *afs_GetVolumeByName();
1025 extern struct conn *afs_Conn();
1026 extern struct conn *afs_ConnByHost();
1027 extern struct conn *afs_ConnByMHosts();
1028 extern afs_int32 afs_NewCell();
1029 extern struct dcache *afs_GetDCache();
1030 extern struct dcache *afs_FindDCache();
1031 extern struct dcache *afs_NewDCache();
1032 extern struct dcache *afs_GetDSlot();
1033 extern struct vcache *afs_GetVCache();
1034 extern struct brequest *afs_BQueue();
1037 extern int afs_CacheInit();
1038 extern void afs_StoreWarn();
1039 extern void afs_AdjustSize();
1040 extern void afs_ComputeCacheParms();
1041 extern void afs_FlushDCache();
1042 extern void afs_FlushActiveVcaches();
1043 extern void afs_StuffVcache();
1044 extern void afs_PutVCache();
1045 extern void afs_TryToSmush();
1046 extern void afs_ProcessFS();
1047 extern void afs_WriteThroughDSlots();
1048 extern void shutdown_cache();
1050 extern void afs_shutdown();
1052 extern void shutdown_osifile();
1055 extern int afs_IsDynrootFid();
1056 extern void afs_GetDynrootFid();
1057 extern int afs_IsDynroot();
1058 extern void afs_RefreshDynroot();
1059 extern void afs_GetDynroot();
1060 extern void afs_PutDynroot();
1061 extern int afs_DynrootNewVnode();
1062 extern int afs_SetDynrootEnable();
1063 extern int afs_GetDynrootEnable();
1064 extern int afs_DynrootVOPSymlink();
1065 extern int afs_DynrootVOPRemove();
1068 /* Performance hack - we could replace VerifyVCache2 with the appropriate
1069 * GetVCache incantation, and could eliminate even this code from afs_UFSRead
1070 * by making intentionally invalidating quick.stamp in the various callbacks
1071 * expiration/breaking code */
1072 #ifdef AFS_LINUX20_ENV
1073 #define afs_VerifyVCache(avc, areq) \
1074 (((avc)->states & CStatd) ? (vcache2inode(avc), 0) : \
1075 afs_VerifyVCache2((avc),areq))
1077 #ifdef AFS_DARWIN_ENV
1078 #define afs_VerifyVCache(avc, areq) \
1079 (((avc)->states & CStatd) ? (osi_VM_Setup(avc), 0) : \
1080 afs_VerifyVCache2((avc),areq))
1082 #define afs_VerifyVCache(avc, areq) \
1083 (((avc)->states & CStatd) ? 0 : afs_VerifyVCache2((avc),areq))
1087 #define DO_STATS 1 /* bits used by FindVCache */
1090 /* values for flag param of afs_CheckVolumeNames */
1091 #define AFS_VOLCHECK_EXPIRED 0x1 /* volumes whose callbacks have expired */
1092 #define AFS_VOLCHECK_BUSY 0x2 /* volumes which were marked busy */
1093 #define AFS_VOLCHECK_MTPTS 0x4 /* mount point invalidation also */
1094 #define AFS_VOLCHECK_FORCE 0x8 /* do all forcibly */
1098 #define AFS_FSPORT ((unsigned short) htons(7000))
1099 #define AFS_VLPORT ((unsigned short) htons(7003))
1101 #define afs_read(avc, uio, acred, albn, abpp, nolock) \
1102 (*(afs_cacheType->vread))(avc, uio, acred, albn, abpp, nolock)
1103 #define afs_write(avc, uio, aio, acred, nolock) \
1104 (*(afs_cacheType->vwrite))(avc, uio, aio, acred, nolock)
1106 #define afs_rdwr(avc, uio, rw, io, cred) \
1107 (((rw) == UIO_WRITE) ? afs_write(avc, uio, io, cred, 0) : afs_read(avc, uio, cred, 0, 0, 0))
1108 #define afs_nlrdwr(avc, uio, rw, io, cred) \
1109 (((rw) == UIO_WRITE) ? afs_write(avc, uio, io, cred, 1) : afs_read(avc, uio, cred, 0, 0, 1))
1111 extern afs_int32 afs_blocksUsed, afs_blocksDiscarded;
1112 extern afs_int32 afs_discardDCCount, afs_freeDCCount;
1113 extern afs_int32 afs_bulkStatsDone, afs_bulkStatsLost;
1114 extern int afs_TruncateDaemonRunning;
1115 extern int afs_CacheTooFull;
1116 /* Cache size truncation uses the following low and high water marks:
1117 * If the cache is more than 95% full (CM_DCACHECOUNTFREEPCT), the cache
1118 * truncation daemon is awakened and will free up space until the cache is 85%
1119 * (CM_DCACHESPACEFREEPCT - CM_DCACHEEXTRAPCT) full.
1120 * afs_UFSWrite and afs_GetDCache (when it needs to fetch data) will wait on
1121 * afs_WaitForCacheDrain if the cache is 98% (CM_WAITFORDRAINPCT) full.
1122 * afs_GetDownD wakes those processes once the cache is 95% full
1123 * (CM_CACHESIZEDRAINEDPCT).
1125 extern void afs_MaybeWakeupTruncateDaemon();
1126 extern void afs_CacheTruncateDaemon();
1127 extern int afs_WaitForCacheDrain;
1128 #define CM_MAXDISCARDEDCHUNKS 16 /* # of chunks */
1129 #define CM_DCACHECOUNTFREEPCT 95 /* max pct of chunks in use */
1130 #define CM_DCACHESPACEFREEPCT 90 /* max pct of space in use */
1131 #define CM_DCACHEEXTRAPCT 5 /* extra to get when freeing */
1132 #define CM_CACHESIZEDRAINEDPCT 95 /* wakeup processes when down to here.*/
1133 #define CM_WAITFORDRAINPCT 98 /* sleep if cache is this full. */
1135 #define afs_CacheIsTooFull() \
1136 (afs_blocksUsed - afs_blocksDiscarded > \
1137 (CM_DCACHECOUNTFREEPCT*afs_cacheBlocks)/100 || \
1138 afs_freeDCCount - afs_discardDCCount < \
1139 ((100-CM_DCACHECOUNTFREEPCT)*afs_cacheFiles)/100)
1141 /* Handy max length of a numeric string. */
1142 #define CVBS 12 /* max afs_int32 is 2^32 ~ 4*10^9, +1 for NULL, +luck */
1144 extern int afs_norefpanic;
1145 #define refpanic(foo) if (afs_norefpanic) \
1146 { printf( foo ); afs_norefpanic++;} else osi_Panic( foo )
1151 ** these are defined in the AIX source code sys/fs_locks.h but are not
1152 ** defined anywhere in the /usr/include directory
1154 #if defined(AFS_AIX41_ENV)
1155 #define VN_LOCK(vp) simple_lock(&(vp)->v_lock)
1156 #define VN_UNLOCK(vp) simple_unlock(&(vp)->v_lock)
1159 /* get a file's serial number from a vnode */
1160 #ifndef afs_vnodeToInumber
1161 #if defined(AFS_SGI62_ENV) || defined(AFS_HAVE_VXFS) || defined(AFS_DARWIN_ENV)
1162 #define afs_vnodeToInumber(V) VnodeToIno(V)
1164 #ifdef AFS_DECOSF_ENV
1165 #define afs_vnodeToInumber(V) osi_vnodeToInumber(V)
1167 #define afs_vnodeToInumber(V) (VTOI(V)->i_number)
1168 #endif /* AFS_DECOSF_ENV */
1169 #endif /* AFS_SGI62_ENV */
1172 /* get a file's device number from a vnode */
1173 #ifndef afs_vnodeToDev
1174 #if defined(AFS_SGI62_ENV) || defined(AFS_HAVE_VXFS) || defined(AFS_DARWIN_ENV)
1175 #define afs_vnodeToDev(V) VnodeToDev(V)
1177 #ifdef AFS_DECOSF_ENV
1178 #define afs_vnodeToDev(V) osi_vnodeToDev(V)
1180 #define afs_vnodeToDev(V) (VTOI(V)->i_dev)
1181 #endif /* AFS_DECOSF_ENV */
1182 #endif /* AFS_SGI62_ENV */
1185 /* fakestat support: opaque storage for afs_EvalFakeStat to remember
1186 * what vcache should be released.
1188 struct afs_fakestat_state {
1192 struct vcache *root_vp;
1195 extern int afs_fakestat_enable;
1197 #endif /* _AFS_H_ */