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
89 #define AFS_BLKBITS 12
90 #define AFS_BLKSIZE (1 << AFS_BLKBITS)
92 extern afs_int32 afs_rx_deadtime;
93 extern afs_int32 afs_rx_harddead;
101 /* flags to use with AFSOP_CACHEINIT */
102 #define AFSCALL_INIT_MEMCACHE 0x1 /* use a memory-based cache */
104 /* below here used only for kernel procedures */
106 /* Store synchrony flags - SYNC means that data should be forced to server's
107 * disk immediately upon completion. */
110 #define AFS_VMSYNC_INVAL 2 /* sync and invalidate pages */
111 #define AFS_LASTSTORE 4
114 /* background request structure */
117 #define BOP_NOOP 0 /* leave 0 unused */
118 #define BOP_FETCH 1 /* parm1 is chunk to get */
119 #define BOP_STORE 2 /* parm1 is chunk to store */
120 #define BOP_PATH 3 /* parm1 is path, parm2 is chunk to fetch */
122 #define B_DONTWAIT 1 /* On failure return; don't wait */
124 /* protocol is: refCount is incremented by user to take block out of free pool.
125 Next, BSTARTED is set when daemon finds request. This prevents
126 other daemons from picking up the same request. Finally, when
127 request is done, refCount is zeroed. BDONE and BWAIT are used by
128 dudes waiting for operation to proceed to a certain point before returning.
130 #define BSTARTED 1 /* request picked up by a daemon */
131 #define BUVALID 2 /* code is valid (store) */
132 #define BUWAIT 4 /* someone is waiting for BUVALID */
134 struct vcache *vc; /* vnode to use, with vrefcount bumped */
135 struct AFS_UCRED *cred; /* credentials to use for operation */
136 afs_size_t size_parm[BPARMS]; /* random parameters */
137 void *ptr_parm[BPARMS]; /* pointer parameters */
138 afs_int32 code; /* return code */
139 short refCount; /* use counter for this structure */
140 char opcode; /* what to do (store, fetch, etc) */
141 char flags; /* free, etc */
142 afs_int32 ts; /* counter "timestamp" */
150 afs_int32 AuthHandle;
151 char HandShakeKey[8];
153 afs_int32 BeginTimestamp;
154 afs_int32 EndTimestamp;
158 afs_int32 Cell; /* better sun packing if at end of structure */
162 /* 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!) */
165 afs_int32 CellAndUnique;
168 /* The actual number of bytes in the SmallFid, not the sizeof struct. */
169 #define SIZEOF_SMALLFID 10
173 * Queues implemented with both pointers and short offsets into a disk file.
181 afs_int32 uid; /* user id making the request */
182 afs_int32 busyCount; /* how many busies we've seen so far */
183 afs_int32 flags; /* things like O_SYNC, O_NONBLOCK go here */
184 char initd; /* if non-zero, non-uid fields meaningful */
185 char accessError; /* flags for overriding error return code */
186 char volumeError; /* encountered a missing or busy volume */
187 char networkError; /* encountered network problems */
188 char permWriteError; /* fileserver returns permenent error. */
193 /* structure linked off of a server to keep track of queued returned
194 * callbacks. Sent asynchronously when we run a little low on free dudes.
197 struct afs_cbr **pprev;
198 struct afs_cbr *next;
200 struct afs_cbr **hash_pprev;
201 struct afs_cbr *hash_next;
206 /* cellinfo file magic number */
207 #define AFS_CELLINFO_MAGIC 0xf32817cd
210 #define CNoSUID 0x02 /* disable suid bit for this cell */
211 #define CLinkedCell4 0x04 /* reserved for ADDCELL2 pioctl */
212 #define CNoAFSDB 0x08 /* never bother trying AFSDB */
213 #define CHasVolRef 0x10 /* volumes were referenced */
214 #define CLinkedCell 0x20 /* has a linked cell in lcellp */
217 struct afs_q lruq; /* lru q next and prev */
218 char *cellName; /* char string name of cell */
219 afs_int32 cellIndex; /* sequence number */
220 afs_int32 cellNum; /* semi-permanent cell number */
221 struct server *cellHosts[MAXCELLHOSTS]; /* volume *location* hosts */
222 struct cell *lcellp; /* Associated linked cell */
223 u_short fsport; /* file server port */
224 u_short vlport; /* volume server port */
225 short states; /* state flags */
226 time_t timeout; /* data expire time, if non-zero */
227 struct cell_name *cnamep; /* pointer to our cell_name */
228 afs_rwlock_t lock; /* protects cell data */
229 unsigned char cellHandle[16]; /* deterministic handle for this cell */
233 struct cell_name *next;
240 struct cell_alias *next;
246 #define afs_PutCell(cellp, locktype)
248 /* the unixuser flag bit definitions */
249 #define UHasTokens 1 /* are the st and ct fields valid (ever set)? */
250 #define UTokensBad 2 /* are tokens bad? */
251 #define UPrimary 4 /* on iff primary identity */
252 #define UNeedsReset 8 /* needs afs_ResetAccessCache call done */
253 #define UPAGCounted 16 /* entry seen during PAG search (for stats) */
254 #define UNFSGetCreds 32 /* getting creds for NFS client */
255 /* A flag used by afs_GCPAGs to keep track of
256 * which entries in afs_users need to be deleted.
257 * The lifetime of its presence in the table is the
258 * lifetime of the afs_GCPAGs function.
260 #define TMP_UPAGNotReferenced 128
262 /* values for afs_gcpags */
263 enum { AFS_GCPAGS_NOTCOMPILED = 0, AFS_GCPAGS_OK =
264 1, AFS_GCPAGS_USERDISABLED, AFS_GCPAGS_EPROC0, AFS_GCPAGS_EPROCN,
265 AFS_GCPAGS_EEQPID, AFS_GCPAGS_EINEXACT, AFS_GCPAGS_EPROCEND,
266 AFS_GCPAGS_EPROCWALK, AFS_GCPAGS_ECREDWALK, AFS_GCPAGS_EPIDCHECK,
267 AFS_GCPAGS_ENICECHECK
270 extern afs_int32 afs_gcpags;
271 extern afs_int32 afs_gcpags_procsize;
272 extern afs_int32 afs_bkvolpref;
273 extern char afs_cachebasedir[1024];
274 extern afs_int32 afs_numcachefiles;
275 extern afs_int32 afs_numfilesperdir;
278 struct unixuser *next; /* next hash pointer */
279 afs_int32 uid; /* search based on uid and cell */
281 afs_int32 vid; /* corresponding vice id in specified cell */
282 short refCount; /* reference count for allocation */
283 char states; /* flag info */
284 afs_int32 tokenTime; /* last time tokens were set, used for timing out conn data */
285 afs_int32 stLen; /* ticket length (if kerberos, includes kvno at head) */
286 char *stp; /* pointer to ticket itself */
287 struct ClearToken ct;
288 struct afs_exporter *exporter; /* more info about the exporter for the remote user */
289 void *cellinfo; /* pointer to cell info (PAG manager only) */
293 /* Per-connection block. */
294 struct conn *next; /* Next dude same server. */
295 struct unixuser *user; /* user validated with respect to. */
296 struct rx_connection *id; /* RPC connid. */
297 struct srvAddr *srvr; /* server associated with this conn */
298 short refCount; /* reference count for allocation */
299 unsigned short port; /* port associated with this connection */
300 char forceConnectFS; /* Should we try again with these tokens? */
306 /* Fid comparison routines */
307 #define FidCmp(a,b) ((a)->Fid.Unique != (b)->Fid.Unique \
308 || (a)->Fid.Vnode != (b)->Fid.Vnode \
309 || (a)->Fid.Volume != (b)->Fid.Volume \
310 || (a)->Cell != (b)->Cell)
312 #define FidMatches(afid,tvc) ((tvc)->fid.Fid.Vnode == (afid)->Fid.Vnode && \
313 (tvc)->fid.Fid.Volume == (afid)->Fid.Volume && \
314 (tvc)->fid.Cell == (afid)->Cell && \
315 ( (tvc)->fid.Fid.Unique == (afid)->Fid.Unique || \
316 (!(afid)->Fid.Unique && ((tvc)->states & CUnique))))
320 * Operations on circular queues implemented with pointers. Note: these queue
321 * objects are always located at the beginning of the structures they are linking.
323 #define QInit(q) ((q)->prev = (q)->next = (q))
324 #define QAdd(q,e) ((e)->next = (q)->next, (e)->prev = (q), \
325 (q)->next->prev = (e), (q)->next = (e))
326 #define QRemove(e) ((e)->next->prev = (e)->prev, (e)->prev->next = (e)->next, (e)->prev = NULL, (e)->next = NULL)
327 #define QNext(e) ((e)->next)
328 #define QPrev(e) ((e)->prev)
329 #define QEmpty(q) ((q)->prev == (q))
330 /* this one takes q1 and sticks it on the end of q2 - that is, the other end, not the end
331 * that things are added onto. q1 shouldn't be empty, it's silly */
332 #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))
334 * Do lots of address arithmetic to go from vlruq to the base of the vcache
335 * structure. Don't move struct vnode, since we think of a struct vcache as
336 * a specialization of a struct vnode
338 #define QTOV(e) ((struct vcache *)(((char *) (e)) - (((char *)(&(((struct vcache *)(e))->vlruq))) - ((char *)(e)))))
339 #define QTOC(e) ((struct cell *)((char *) (e)))
340 #define QTOVH(e) ((struct vcache *)(((char *) (e)) - (((char *)(&(((struct vcache *)(e))->vhashq))) - ((char *)(e)))))
343 #define SRVADDR_ISDOWN 0x20 /* same as SRVR_ISDOWN */
344 #define SRVADDR_NOUSE 0x40 /* Don't use this srvAddr */
346 struct srvAddr *next_bkt; /* next item in hash bucket */
347 struct srvAddr *next_sa; /* another interface on same host */
348 struct server *server; /* back to parent */
349 struct conn *conns; /* All user connections to this server */
350 afs_int32 sa_ip; /* Host addr in network byte order */
351 u_short sa_iprank; /* indiv ip address priority */
352 u_short sa_portal; /* port addr in network byte order */
357 * Values used in the flags field of the server structure below.
359 * AFS_SERVER_FLAG_ACTIVATED Has the server ever had a user connection
360 * associated with it?
362 #define AFS_SERVER_FLAG_ACTIVATED 0x01
363 #define SNO_LHOSTS 0x04
364 #define SYES_LHOSTS 0x08
365 #define SVLSRV_UUID 0x10
366 #define SRVR_ISDOWN 0x20
367 #define SRVR_MULTIHOMED 0x40
368 #define SRVR_ISGONE 0x80
369 #define SNO_INLINEBULK 0x100
370 #define SNO_64BIT 0x200
372 #define afs_serverSetNo64Bit(s) ((s)->srvr->server->flags |= SNO_64BIT)
373 #define afs_serverHasNo64Bit(s) ((s)->srvr->server->flags & SNO_64BIT)
379 afs_int32 addr_uniquifier;
383 struct srvAddr haddr;
386 #define sr_uuid _suid._srvUuid.suuid
387 #define sr_addr_uniquifier _suid._srvUuid.addr_uniquifier
388 #define sr_host _suid._srvId.haddr.ip
389 #define sr_portal _suid._srvId.haddr.portal
390 #define sr_rank _suid._srvId.haddr.ip_rank
391 #define sr_flags _suid._srvId.haddr.flags
392 #define sr_conns _suid._srvId.haddr.conns
393 struct server *next; /* Ptr to next server in hash chain */
394 struct cell *cell; /* Cell in which this host resides */
395 struct afs_cbr *cbrs; /* Return list of callbacks */
396 afs_int32 activationTime; /* Time when this record was first activated */
397 afs_int32 lastDowntimeStart; /* Time when last downtime incident began */
398 afs_int32 numDowntimeIncidents; /* # (completed) downtime incidents */
399 afs_int32 sumOfDowntimes; /* Total downtime experienced, in seconds */
400 struct srvAddr *addr;
401 afs_uint32 flags; /* Misc flags */
404 #define afs_PutServer(servp, locktype)
406 /* structs for some pioctls - these are (or should be)
414 struct sprefrequest_33 {
415 unsigned short offset;
416 unsigned short num_servers;
420 struct sprefrequest { /* new struct for 3.4 */
421 unsigned short offset;
422 unsigned short num_servers;
423 unsigned short flags;
428 unsigned short next_offset;
429 unsigned short num_servers;
430 struct spref servers[1]; /* we overrun this array intentionally... */
434 unsigned short flags;
435 unsigned short num_servers;
436 struct spref servers[1]; /* we overrun this array intentionally... */
438 /* struct for GAG pioctl
441 afs_uint32 showflags, logflags, logwritethruflag, spare[3];
442 unsigned char spare2[128];
446 #define logwritethruON 1
449 afs_int32 rx_initReceiveWindow, rx_maxReceiveWindow, rx_initSendWindow,
450 rx_maxSendWindow, rxi_nSendFrags, rxi_nRecvFrags, rxi_OrphanFragSize;
451 afs_int32 rx_maxReceiveSize, rx_MyMaxSendSize;
452 afs_uint32 spare[21];
455 /* struct for checkservers */
466 /* state bits for volume */
467 #define VRO 1 /* volume is readonly */
468 #define VRecheck 2 /* recheck volume info with server */
469 #define VBackup 4 /* is this a backup volume? */
470 #define VForeign 8 /* this is a non-afs volume */
471 #define VResort 16 /* server order was rearranged, sort when able */
472 #define VMoreReps 32 /* This volume has more replicas than we are */
473 /* keeping track of now -- check with VLDB */
475 enum repstate { not_busy, end_not_busy = 6, rd_busy, rdwr_busy, offline };
478 /* One structure per volume, describing where the volume is located
479 * and where its mount points are. */
480 struct volume *next; /* Next volume in hash list. */
481 afs_int32 cell; /* the cell in which the volume resides */
482 afs_rwlock_t lock; /* the lock for this structure */
483 afs_int32 volume; /* This volume's ID number. */
484 char *name; /* This volume's name, or 0 if unknown */
485 struct server *serverHost[MAXHOSTS]; /* servers serving this volume */
486 enum repstate status[MAXHOSTS]; /* busy, offline, etc */
487 struct VenusFid dotdot; /* dir to access as .. */
488 struct VenusFid mtpoint; /* The mount point for this volume. */
489 afs_int32 rootVnode, rootUnique; /* Volume's root fid */
492 afs_int32 rwVol; /* For r/o vols, original read/write volume. */
493 afs_int32 accessTime; /* last time we used it */
494 afs_int32 vtix; /* volume table index */
495 afs_int32 copyDate; /* copyDate field, for tracking vol releases */
496 afs_int32 expireTime; /* for per-volume callbacks... */
497 short refCount; /* reference count for allocation */
498 char states; /* here for alignment reasons */
501 #define afs_PutVolume(av, locktype) ((av)->refCount--)
503 /* format of an entry in volume info file */
505 afs_int32 cell; /* cell for this entry */
506 afs_int32 volume; /* volume */
507 afs_int32 next; /* has index */
508 struct VenusFid dotdot; /* .. value */
509 struct VenusFid mtpoint; /* mt point's fid */
510 afs_int32 rootVnode, rootUnique; /* Volume's root fid */
514 struct SimpleLocks *next;
516 afs_int32 boff, eoff;
518 #if defined(AFS_AIX32_ENV) || defined(AFS_SUN5_ENV) || defined(AFS_SGI_ENV)
523 /* vcache state bits */
524 #define CStatd 0x00000001 /* has this file ever been stat'd? */
525 #define CBackup 0x00000002 /* file is on a backup volume */
526 #define CRO 0x00000004 /* is it on a read-only volume */
527 #define CMValid 0x00000008 /* is the mount point info valid? */
528 #define CCore 0x00000010 /* storing a core file, needed since we don't get an open */
529 #define CDirty 0x00000020 /* file has been modified since first open (... O_RDWR) */
530 #define CSafeStore 0x00000040 /* close must wait for store to finish (should be in fd) */
531 #define CMAPPED 0x00000080 /* Mapped files; primarily used by SunOS 4.0.x */
532 #define CNSHARE 0x00000100 /* support O_NSHARE semantics */
533 #define CLied 0x00000200
534 #define CTruth 0x00000400
536 #define CWired 0x00000800 /* OSF hack only */
538 #ifdef AFS_DARWIN80_ENV
539 #define CDeadVnode 0x00000800
541 #ifdef AFS_DARWIN_ENV
542 #define CUBCinit 0x00000800
544 #define CWRITE_IGN 0x00000800 /* Next OS hack only */
548 #define CUnique 0x00001000 /* vc's uniquifier - latest unifiquier for fid */
549 #define CForeign 0x00002000 /* this is a non-afs vcache */
550 #define CReadDir 0x00004000 /* readdir in progress */
551 #define CUnlinked 0x00010000
552 #define CBulkStat 0x00020000 /* loaded by a bulk stat, and not ref'd since */
553 #define CUnlinkedDel 0x00040000
554 #define CVFlushed 0x00080000
555 #define CCore1 0x00100000 /* osf1 core file; not same as CCore above */
556 #define CWritingUFS 0x00200000 /* to detect vm deadlock - used by sgi */
557 #define CCreating 0x00400000 /* avoid needless store after open truncate */
558 #define CPageHog 0x00800000 /* AIX - dumping large cores is a page hog. */
559 #define CDCLock 0x02000000 /* Vnode lock held over call to GetDownD */
560 #define CBulkFetching 0x04000000 /* stats are being fetched by bulk stat */
561 #define CExtendedFile 0x08000000 /* extended file via ftruncate call. */
562 #define CVInit 0x10000000 /* being initialized */
564 /* vcache vstate bits */
565 #define VRevokeWait 0x1
566 #define VPageCleaning 0x2 /* Solaris - Cache Trunc Daemon sez keep out */
569 #if defined(AFS_XBSD_ENV) || defined(AFS_DARWIN_ENV)
570 #define vrefCount v->v_usecount
572 #define vrefCount v.v_count
573 #endif /* AFS_XBSD_ENV */
575 #if defined(AFS_DARWIN80_ENV)
576 #define VREFCOUNT_GT(v, y) vnode_isinuse(AFSTOV(v), (y))
577 #elif defined(AFS_XBSD_ENV) || defined(AFS_DARWIN_ENV)
578 #define VREFCOUNT(v) ((v)->vrefCount)
579 #define VREFCOUNT_GT(v, y) (AFSTOV(v)->v_usecount > (y))
580 #elif defined(AFS_LINUX24_ENV)
581 #define VREFCOUNT(v) atomic_read(&(AFSTOV(v)->v_count))
582 #define VREFCOUNT_GT(v, y) (VREFCOUNT(v)>y)
583 #define VREFCOUNT_SET(v, c) atomic_set(&(AFSTOV(v)->v_count), c)
584 #define VREFCOUNT_DEC(v) atomic_dec(&(AFSTOV(v)->v_count))
585 #define VREFCOUNT_INC(v) atomic_inc(&(AFSTOV(v)->v_count))
587 #define VREFCOUNT(v) ((v)->vrefCount)
588 #define VREFCOUNT_GT(v,y) ((v)->vrefCount > (y))
589 #define VREFCOUNT_SET(v, c) (v)->vrefCount = c;
590 #define VREFCOUNT_DEC(v) (v)->vrefCount--;
591 #define VREFCOUNT_INC(v) (v)->vrefCount++;
592 #define d_unhash(d) list_empty(&(d)->d_hash)
593 #define dget_locked(d) dget(d)
596 #define AFS_MAXDV 0x7fffffff /* largest dataversion number */
597 #ifdef AFS_64BIT_CLIENT
598 #define AFS_NOTRUNC 0x7fffffffffffffffLL /* largest positive int64 number */
599 #else /* AFS_64BIT_CLIENT */
600 #define AFS_NOTRUNC 0x7fffffff /* largest dataversion number */
601 #endif /* AFS_64BIT_CLIENT */
603 extern afs_int32 vmPageHog; /* counter for # of vnodes which are page hogs. */
605 #if defined(AFS_DARWIN80_ENV)
606 #define VTOAFS(v) ((struct vcache *)vnode_fsnode((v)))
607 #define AFSTOV(vc) ((vc)->v)
608 #elif defined(AFS_XBSD_ENV) || defined(AFS_DARWIN_ENV) || (defined(AFS_LINUX22_ENV) && !defined(STRUCT_SUPER_HAS_ALLOC_INODE))
609 #define VTOAFS(v) ((struct vcache *)(v)->v_data)
610 #define AFSTOV(vc) ((vc)->v)
612 #define VTOAFS(V) ((struct vcache *)(V))
613 #define AFSTOV(V) (&(V)->v)
616 /* INVARIANTs: (vlruq.next != NULL) == (vlruq.prev != NULL)
617 * nextfree => !vlruq.next && ! vlruq.prev
618 * !(avc->nextfree) && !avc->vlruq.next => (FreeVCList == avc->nextfree)
621 #if defined(AFS_XBSD_ENV) || defined(AFS_DARWIN_ENV) || (defined(AFS_LINUX22_ENV) && !defined(STRUCT_SUPER_HAS_ALLOC_INODE))
624 struct vnode v; /* Has reference count in v.v_count */
626 struct afs_q vlruq; /* lru q next and prev */
627 #if !defined(AFS_LINUX22_ENV)
628 struct vcache *nextfree; /* next on free list (if free) */
630 struct vcache *hnext; /* Hash next */
631 struct afs_q vhashq; /* Hashed per-volume list */
635 afs_hyper_t DataVersion;
639 afs_uint16 Mode; /* XXXX Should be afs_int32 XXXX */
640 afs_uint16 LinkCount;
641 #ifdef AFS_DARWIN80_ENV
644 /* vnode type is in v.v_type */
647 afs_rwlock_t lock; /* The lock on the vcache contents. */
648 #if defined(AFS_SUN5_ENV)
649 /* Lock used to protect the activeV, multipage, and vstates fields.
650 * Do not try to get the vcache lock when the vlock is held */
652 #endif /* defined(AFS_SUN5_ENV) */
653 #if defined(AFS_SUN5_ENV)
657 #ifdef AFS_BOZONLOCK_ENV
658 afs_bozoLock_t pvnLock; /* see locks.x */
663 #if defined(AFS_AIX51_ENV)
671 int ownslock; /* pid of owner of excl lock, else 0 - defect 3083 */
673 #ifdef AFS_DARWIN80_ENV
675 #elif defined(AFS_DARWIN_ENV)
676 struct lock__bsd__ rwlock;
681 afs_int32 parentVnode; /* Parent dir, if a file. */
682 afs_int32 parentUnique;
683 struct VenusFid *mvid; /* Either parent dir (if root) or root (if mt pt) */
684 char *linkData; /* Link data if a symlink. */
685 afs_hyper_t flushDV; /* data version last flushed from text */
686 afs_hyper_t mapDV; /* data version last flushed from map */
687 afs_size_t truncPos; /* truncate file to this position at next store */
688 struct server *callback; /* The callback host, if any */
689 afs_uint32 cbExpires; /* time the callback expires */
690 struct afs_q callsort; /* queue in expiry order, sort of */
691 struct axscache *Access; /* a list of cached access bits */
692 afs_int32 anyAccess; /* System:AnyUser's access to this. */
693 afs_int32 last_looker; /* pag/uid from last lookup here */
694 #if defined(AFS_SUN5_ENV)
696 #endif /* defined(AFS_SUN5_ENV) */
697 struct SimpleLocks *slocks;
698 short opens; /* The numbers of opens, read or write, on this file. */
699 short execsOrWriters; /* The number of execs (if < 0) or writers (if > 0) of
701 short flockCount; /* count of flock readers, or -1 if writer */
702 char mvstat; /* 0->normal, 1->mt pt, 2->root. */
703 afs_uint32 states; /* state bits */
704 #if defined(AFS_SUN5_ENV)
705 afs_uint32 vstates; /* vstate bits */
706 #endif /* defined(AFS_SUN5_ENV) */
707 struct dcache *dchint;
708 struct dcache *dcreaddir; /* dcache for in-progress readdir */
709 unsigned int readdir_pid; /* pid of the thread in readdir */
710 #ifdef AFS_LINUX22_ENV
711 u_short mapcnt; /* Number of mappings of this file. */
713 #if defined(AFS_SGI_ENV)
714 daddr_t lastr; /* for read-ahead */
716 uint64_t vc_rwlockid; /* kthread owning rwlock */
718 short vc_rwlockid; /* pid of process owning rwlock */
720 short vc_locktrips; /* # of rwlock reacquisitions */
721 sema_t vc_rwlock; /* vop_rwlock for afs */
722 pgno_t mapcnt; /* # of pages mapped */
723 struct cred *cred; /* last writer's cred */
725 struct bhv_desc vc_bhv_desc; /* vnode's behavior data. */
727 #endif /* AFS_SGI_ENV */
728 afs_int32 vc_error; /* stash write error for this vnode. */
729 int xlatordv; /* Used by nfs xlator */
730 struct AFS_UCRED *uncred;
731 int asynchrony; /* num kbytes to store behind */
733 short multiPage; /* count of multi-page getpages in progress */
737 #define DONT_CHECK_MODE_BITS 0
738 #define CHECK_MODE_BITS 1
739 #define CMB_ALLOW_EXEC_AS_READ 2 /* For the NFS xlator */
741 #if defined(AFS_SGI_ENV)
742 #define AVCRWLOCK(avc) (valusema(&(avc)->vc_rwlock) <= 0)
744 /* SGI vnode rwlock macros and flags. */
745 #ifndef AFS_SGI62_ENV
746 /* The following are defined here. SGI 6.2 declares them in vnode.h */
747 #define VRWLOCK_READ 0
748 #define VRWLOCK_WRITE 1
749 #define VRWLOCK_WRITE_DIRECT 2
754 #define AFS_RWLOCK_T vrwlock_t
756 #define AFS_RWLOCK_T int
757 #endif /* AFS_SGI62_ENV */
759 #include <ksys/behavior.h>
760 #define AFS_RWLOCK(V,F) \
761 afs_rwlock(&VTOAFS(V)->vc_bhv_desc, (F));
762 #define AFS_RWUNLOCK(V,F) \
763 afs_rwunlock(&VTOAFS(V)->vc_bhv_desc, (F));
766 #define AFS_RWLOCK(V,F) afs_rwlock((vnode_t *)(V), (F) )
767 #define AFS_RWUNLOCK(V,F) afs_rwunlock((vnode_t *)(V), (F) )
769 #else /* AFS_SGI53_ENV */
770 #define AFS_RWLOCK(V,F) afs_rwlock((V))
771 #define AFS_RWUNLOCK(V,F) afs_rwunlock((V))
772 #endif /* AFS_SGI53_ENV */
773 #endif /* AFS_SGI_ENV */
777 afs_hyper_t DataVersion;
779 afs_int32 parentVnode;
780 afs_int32 parentUnique;
784 afs_int32 randomUid[CPSIZE];
785 afs_int32 callback; /* Now a pointer to 'server' struct */
787 afs_int32 randomAccess[CPSIZE];
790 short execsOrWriters;
797 afs_int32 callerAccess;
808 /* CM inititialization parameters. What CM actually used after calculations
809 * based on passed in arguments.
811 #define CMI_VERSION 1 /* increment when adding new fields. */
812 struct cm_initparams {
817 int cmi_nVolumeCaches;
818 int cmi_firstChunkSize;
819 int cmi_otherChunkSize;
820 int cmi_cacheSize; /* The original cache size, in 1K blocks. */
821 unsigned cmi_setTime:1;
822 unsigned cmi_memCache:1;
823 int spare[16 - 9]; /* size of struct is 16 * 4 = 64 bytes */
827 /*----------------------------------------------------------------------
828 * AFS Data cache definitions
830 * Each entry describes a Unix file on the local disk that is
831 * is serving as a cached copy of all or part of a Vice file.
832 * Entries live in circular queues for each hash table slot
834 * Which queue is this thing in? Good question.
835 * A struct dcache entry is in the freeDSlot queue when not associated with a cache slot (file).
836 * Otherwise, it is in the DLRU queue. The freeDSlot queue uses the lruq.next field as
837 * its "next" pointer.
839 * Cache entries in the DLRU queue are either associated with vice files, in which case
840 * they are hashed by afs_dvnextTbl and afs_dcnextTbl pointers, or they are in the freeDCList
841 * and are not associated with any vice file. This last list uses the afs_dvnextTbl pointer for
842 * its "next" pointer.
843 *----------------------------------------------------------------------*/
845 #define NULLIDX (-1) /* null index definition */
846 /* struct dcache states bits */
850 #define DWriting 8 /* file being written (used for cache validation) */
852 /* dcache data flags */
853 #define DFEntryMod 0x02 /* has entry itself been modified? */
854 #define DFFetching 0x04 /* file is currently being fetched */
856 /* dcache meta flags */
857 #define DFNextStarted 0x01 /* next chunk has been prefetched already */
858 #define DFFetchReq 0x10 /* someone is waiting for DFFetching to go on */
861 /* flags in afs_indexFlags array */
862 #define IFEverUsed 1 /* index entry has >= 1 byte of data */
863 #define IFFree 2 /* index entry in freeDCList */
864 #define IFDataMod 4 /* file needs to be written out */
865 #define IFFlag 8 /* utility flag */
866 #define IFDirtyPages 16
867 #define IFAnyPages 32
868 #define IFDiscarded 64 /* index entry in discardDCList */
871 char *in; /* input buffer */
872 char *out; /* output buffer */
873 short in_size; /* Size of input buffer <= 2K */
874 short out_size; /* Maximum size of output buffer, <= 2K */
878 * This version of afs_ioctl is required to pass in 32 bit user space
879 * pointers into a 64 bit kernel.
890 /* CacheItems file has a header of type struct afs_fheader
891 * (keep aligned properly). Since we already have sgi_62 clients running
892 * with a 32 bit inode, a change is required to the header so that
893 * they can distinguish the old 32 bit inode CacheItems file and zap it
894 * instead of using it.
897 #define AFS_FHMAGIC 0x7635abaf /* uses version number */
899 #if defined(AFS_SUN57_64BIT_ENV)
900 #define AFS_CI_VERSION 3
902 #define AFS_CI_VERSION 2
905 afs_int32 firstCSize;
906 afs_int32 otherCSize;
909 #if defined(AFS_SGI61_ENV) || defined(AFS_SUN57_64BIT_ENV)
910 /* Using ino64_t here so that user level debugging programs compile
911 * the size correctly.
913 #define afs_inode_t ino64_t
915 #if defined(AFS_LINUX_64BIT_KERNEL) && !defined(AFS_S390X_LINUX24_ENV)
916 #define afs_inode_t long
918 #if defined(AFS_AIX51_ENV) || defined(AFS_HPUX1123_ENV)
919 #define afs_inode_t ino_t
921 #define afs_inode_t afs_int32
928 /* it does not compile outside kernel */
930 afs_int32 fid; /* is adc->index, the cache file number */
931 afs_inode_t inode; /* is adc->f.inode, the inode number of the cac\
934 afs_int32 accesstime;
935 struct buffer *hashNext;
940 #if defined(AFS_USEBUFFERS)
943 afs_rwlock_t lock; /* the lock for this structure */
946 /* kept on disk and in dcache entries */
948 struct VenusFid fid; /* Fid for this file */
949 afs_int32 modTime; /* last time this entry was modified */
950 afs_hyper_t versionNo; /* Associated data version number */
951 afs_int32 chunk; /* Relative chunk number */
952 afs_inode_t inode; /* Unix inode for this chunk */
953 afs_int32 chunkBytes; /* Num bytes in this chunk */
954 char states; /* Has this chunk been modified? */
958 /* magic numbers to specify the cache type */
960 #define AFS_FCACHE_TYPE_UFS 0x0
961 #define AFS_FCACHE_TYPE_MEM 0x1
962 #define AFS_FCACHE_TYPE_NFS 0x2
963 #define AFS_FCACHE_TYPE_EPI 0x3
967 struct afs_q lruq; /* Free queue for in-memory images */
968 struct afs_q dirty; /* Queue of dirty entries that need written */
969 afs_rwlock_t lock; /* Protects validPos, some f */
970 afs_rwlock_t tlock; /* Atomizes updates to refCount */
971 afs_rwlock_t mflock; /* Atomizes accesses/updates to mflags */
972 afs_size_t validPos; /* number of valid bytes during fetch */
973 afs_int32 index; /* The index in the CacheInfo file */
974 short refCount; /* Associated reference count. */
975 char dflags; /* Data flags */
976 char mflags; /* Meta flags */
977 struct fcache f; /* disk image */
978 afs_int32 bucket; /* which bucket these dcache entries are in */
982 * dcache.lock protects the actual contents of the cache file (in
983 * f.inode), subfields of f except those noted below, dflags and
986 * dcache.tlock is used to make atomic updates to refCount. Zero
987 * refCount dcache entries are protected by afs_xdcache instead of
990 * dcache.mflock is used to access and update mflags. It cannot be
991 * held without holding the corresponding dcache.lock. Updating
992 * mflags requires holding dcache.lock(R) and dcache.mflock(W), and
993 * checking for mflags requires dcache.lock(R) and dcache.mflock(R).
994 * Note that dcache.lock(W) gives you the right to update mflags,
995 * as dcache.mflock(W) can only be held with dcache.lock(R).
997 * dcache.index, dcache.f.fid, dcache.f.chunk and dcache.f.inode are
998 * write-protected by afs_xdcache and read-protected by refCount.
999 * Once an entry is referenced, these values cannot change, and if
1000 * it's on the free list (with refCount=0), it can be reused for a
1001 * different file/chunk. These values can only be written while
1002 * holding afs_xdcache(W) and allocating this dcache entry (thereby
1003 * ensuring noone else has a refCount on it).
1007 /* afs_memcache.c */
1008 struct memCacheEntry {
1009 int size; /* # of valid bytes in this entry */
1010 int dataSize; /* size of allocated data area */
1011 afs_lock_t afs_memLock;
1012 char *data; /* bytes */
1015 /* macro to mark a dcache entry as bad */
1018 (x)->f.fid.Fid.Unique = 0; \
1019 afs_indexUnique[(x)->index] = 0; \
1020 (x)->dflags |= DFEntryMod; \
1023 /* FakeOpen and Fake Close used to be real subroutines. They're only used in
1024 * sun_subr and afs_vnodeops, and they're very frequently called, so I made
1025 * them into macros. They do:
1026 * FakeOpen: fake the file being open for writing. avc->lock must be held
1027 * in write mode. Having the file open for writing is like having a DFS
1028 * write-token: you're known to have the best version of the data around,
1029 * and so the CM won't let it be overwritten by random server info.
1030 * FakeClose: undo the effects of FakeOpen, noting that we want to ensure
1031 * that a real close eventually gets done. We use CCore to achieve this if
1032 * we would end up closing the file. avc->lock must be held in write mode */
1034 #ifdef AFS_AIX_IAUTH_ENV
1035 #define CRKEEP(V, C) (V)->linkData = (char*)crdup((C))
1037 #define CRKEEP(V, C) crhold((C)); (V)->linkData = (char*)(C)
1040 #define afs_FakeOpen(avc) { avc->opens++; avc->execsOrWriters++; }
1041 #define afs_FakeClose(avc, acred) \
1042 { if (avc->execsOrWriters == 1) { \
1043 /* we're the last writer, just use CCore flag */ \
1044 avc->states |= CCore; /* causes close to be called later */ \
1046 /* The cred and vnode holds will be released in afs_FlushActiveVcaches */ \
1047 VN_HOLD(AFSTOV(avc)); /* So it won't disappear */ \
1048 CRKEEP(avc, acred); /* Should use a better place for the creds */ \
1051 /* we're not the last writer, let the last one do the store-back for us */ \
1053 avc->execsOrWriters--; \
1057 #define AFS_ZEROS 64 /* zero buffer */
1059 /*#define afs_DirtyPages(avc) (((avc)->states & CDirty) || osi_VMDirty_p((avc)))*/
1060 #define afs_DirtyPages(avc) ((avc)->states & CDirty)
1062 #define afs_InReadDir(avc) (((avc)->states & CReadDir) && (avc)->readdir_pid == MyPidxx)
1064 /* The PFlush algorithm makes use of the fact that Fid.Unique is not used in
1065 below hash algorithms. Change it if need be so that flushing algorithm
1066 doesn't move things from one hash chain to another
1068 /* extern int afs_dhashsize; */
1069 #define DCHash(v, c) ((((v)->Fid.Vnode + (v)->Fid.Volume + (c))) & (afs_dhashsize-1))
1070 /*Vnode, Chunk -> Hash table index */
1071 #define DVHash(v) ((((v)->Fid.Vnode + (v)->Fid.Volume )) & (afs_dhashsize-1))
1072 /*Vnode -> Other hash table index */
1073 /* don't hash on the cell, our callback-breaking code sometimes fails to compute
1074 the cell correctly, and only scans one hash bucket */
1075 #define VCHash(fid) (((fid)->Fid.Volume + (fid)->Fid.Vnode) & (VCSIZE-1))
1076 /* Hash only on volume to speed up volume callbacks. */
1077 #define VCHashV(fid) ((fid)->Fid.Volume & (VCSIZE-1))
1079 extern struct dcache **afs_indexTable; /*Pointers to in-memory dcache entries */
1080 extern afs_int32 *afs_indexUnique; /*dcache entry Fid.Unique */
1081 extern afs_int32 *afs_dvnextTbl; /*Dcache hash table links */
1082 extern afs_int32 *afs_dcnextTbl; /*Dcache hash table links */
1083 extern afs_int32 afs_cacheFiles; /*Size of afs_indexTable */
1084 extern afs_int32 afs_cacheBlocks; /*1K blocks in cache */
1085 extern afs_int32 afs_cacheStats; /*Stat entries in cache */
1086 extern struct vcache *afs_vhashT[VCSIZE]; /*Stat cache hash table */
1087 extern struct afs_q afs_vhashTV[VCSIZE]; /* cache hash table on volume */
1088 extern afs_int32 afs_initState; /*Initialization state */
1089 extern afs_int32 afs_termState; /* Termination state */
1090 extern struct VenusFid afs_rootFid; /*Root for whole file system */
1091 extern afs_int32 afs_allCBs; /* Count of callbacks */
1092 extern afs_int32 afs_oddCBs; /* Count of odd callbacks */
1093 extern afs_int32 afs_evenCBs; /* Count of even callbacks */
1094 extern afs_int32 afs_allZaps; /* Count of fid deletes */
1095 extern afs_int32 afs_oddZaps; /* Count of odd fid deletes */
1096 extern afs_int32 afs_evenZaps; /* Count of even fid deletes */
1097 extern struct brequest afs_brs[NBRS]; /* request structures */
1099 #define UHash(auid) ((auid) & (NUSERS-1))
1100 #define VHash(avol) ((avol)&(NVOLS-1))
1101 #define SHash(aserv) ((ntohl(aserv)) & (NSERVERS-1))
1102 #define FVHash(acell,avol) (((avol)+(acell)) & (NFENTRIES-1))
1104 /* Performance hack - we could replace VerifyVCache2 with the appropriate
1105 * GetVCache incantation, and could eliminate even this code from afs_UFSRead
1106 * by making intentionally invalidating quick.stamp in the various callbacks
1107 * expiration/breaking code */
1108 #ifdef AFS_DARWIN_ENV
1109 #define afs_VerifyVCache(avc, areq) \
1110 (((avc)->states & CStatd) ? (osi_VM_Setup(avc, 0), 0) : \
1111 afs_VerifyVCache2((avc),areq))
1113 #define afs_VerifyVCache(avc, areq) \
1114 (((avc)->states & CStatd) ? 0 : afs_VerifyVCache2((avc),areq))
1117 #define DO_STATS 1 /* bits used by FindVCache */
1122 /* values for flag param of afs_CheckVolumeNames */
1123 #define AFS_VOLCHECK_EXPIRED 0x1 /* volumes whose callbacks have expired */
1124 #define AFS_VOLCHECK_BUSY 0x2 /* volumes which were marked busy */
1125 #define AFS_VOLCHECK_MTPTS 0x4 /* mount point invalidation also */
1126 #define AFS_VOLCHECK_FORCE 0x8 /* do all forcibly */
1130 #define AFS_FSPORT ((unsigned short) htons(7000))
1131 #define AFS_VLPORT ((unsigned short) htons(7003))
1133 #define afs_read(avc, uio, acred, albn, abpp, nolock) \
1134 (*(afs_cacheType->vread))(avc, uio, acred, albn, abpp, nolock)
1135 #define afs_write(avc, uio, aio, acred, nolock) \
1136 (*(afs_cacheType->vwrite))(avc, uio, aio, acred, nolock)
1138 #define afs_rdwr(avc, uio, rw, io, cred) \
1139 (((rw) == UIO_WRITE) ? afs_write(avc, uio, io, cred, 0) : afs_read(avc, uio, cred, 0, 0, 0))
1140 #define afs_nlrdwr(avc, uio, rw, io, cred) \
1141 (((rw) == UIO_WRITE) ? afs_write(avc, uio, io, cred, 1) : afs_read(avc, uio, cred, 0, 0, 1))
1143 /* Cache size truncation uses the following low and high water marks:
1144 * If the cache is more than 95% full (CM_DCACHECOUNTFREEPCT), the cache
1145 * truncation daemon is awakened and will free up space until the cache is 85%
1146 * (CM_DCACHESPACEFREEPCT - CM_DCACHEEXTRAPCT) full.
1147 * afs_UFSWrite and afs_GetDCache (when it needs to fetch data) will wait on
1148 * afs_WaitForCacheDrain if the cache is 98% (CM_WAITFORDRAINPCT) full.
1149 * afs_GetDownD wakes those processes once the cache is 95% full
1150 * (CM_CACHESIZEDRAINEDPCT).
1152 #define CM_MAXDISCARDEDCHUNKS 16 /* # of chunks */
1153 #define CM_DCACHECOUNTFREEPCT 95 /* max pct of chunks in use */
1154 #define CM_DCACHESPACEFREEPCT 90 /* max pct of space in use */
1155 #define CM_DCACHEEXTRAPCT 5 /* extra to get when freeing */
1156 #define CM_CACHESIZEDRAINEDPCT 95 /* wakeup processes when down to here. */
1157 #define CM_WAITFORDRAINPCT 98 /* sleep if cache is this full. */
1159 /* when afs_cacheBlocks is large, settle for slightly decreased precision */
1160 #define PERCENT(p, v) \
1161 ((afs_cacheBlocks & 0xffe00000) ? ((v) / 100 * (p)) : ((p) * (v) / 100))
1163 #define afs_CacheIsTooFull() \
1164 (afs_blocksUsed - afs_blocksDiscarded > \
1165 PERCENT(CM_DCACHECOUNTFREEPCT, afs_cacheBlocks) || \
1166 afs_freeDCCount - afs_discardDCCount < \
1167 PERCENT(100 - CM_DCACHECOUNTFREEPCT, afs_cacheFiles))
1169 /* Handy max length of a numeric string. */
1170 #define CVBS 12 /* max afs_int32 is 2^32 ~ 4*10^9, +1 for NULL, +luck */
1172 #define refpanic(foo) if (afs_norefpanic) \
1173 { printf( foo ); afs_norefpanic++;} else osi_Panic( foo )
1176 ** these are defined in the AIX source code sys/fs_locks.h but are not
1177 ** defined anywhere in the /usr/include directory
1179 #if defined(AFS_AIX41_ENV)
1180 #define VN_LOCK(vp) simple_lock(&(vp)->v_lock)
1181 #define VN_UNLOCK(vp) simple_unlock(&(vp)->v_lock)
1184 /* get a file's serial number from a vnode */
1185 #ifndef afs_vnodeToInumber
1186 #if defined(AFS_SGI62_ENV) || defined(AFS_HAVE_VXFS) || defined(AFS_DARWIN_ENV)
1187 #define afs_vnodeToInumber(V) VnodeToIno(V)
1189 #ifdef AFS_DECOSF_ENV
1190 #define afs_vnodeToInumber(V) osi_vnodeToInumber(V)
1192 #define afs_vnodeToInumber(V) (VTOI(V)->i_number)
1193 #endif /* AFS_DECOSF_ENV */
1194 #endif /* AFS_SGI62_ENV */
1197 /* get a file's device number from a vnode */
1198 #ifndef afs_vnodeToDev
1199 #if defined(AFS_SGI62_ENV) || defined(AFS_HAVE_VXFS) || defined(AFS_DARWIN_ENV)
1200 #define afs_vnodeToDev(V) VnodeToDev(V)
1201 #elif defined(AFS_DECOSF_ENV)
1202 #define afs_vnodeToDev(V) osi_vnodeToDev(V)
1204 #define afs_vnodeToDev(V) (VTOI(V)->i_dev)
1209 /* Note: this should agree with the definition in kdump.c */
1210 #if defined(AFS_OSF_ENV)
1211 #if !defined(UKERNEL)
1212 #define AFS_USEBUFFERS 1
1216 #if !defined(UKERNEL) && !defined(HAVE_STRUCT_BUF)
1217 /* declare something so that prototypes don't flip out */
1218 /* appears struct buf stuff is only actually passed around as a pointer,
1219 except with libuafs, in which case it is actually defined */
1224 /* fakestat support: opaque storage for afs_EvalFakeStat to remember
1225 * what vcache should be released.
1227 struct afs_fakestat_state {
1231 struct vcache *root_vp;
1234 extern int afs_fakestat_enable;
1235 #endif /* _AFS_H_ */