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 */
227 unsigned char cellHandle[16]; /* deterministic handle for this cell */
231 struct cell_name *next;
238 struct cell_alias *next;
244 #define afs_PutCell(cellp, locktype)
246 /* the unixuser flag bit definitions */
247 #define UHasTokens 1 /* are the st and ct fields valid (ever set)? */
248 #define UTokensBad 2 /* are tokens bad? */
249 #define UPrimary 4 /* on iff primary identity */
250 #define UNeedsReset 8 /* needs afs_ResetAccessCache call done */
251 #define UPAGCounted 16 /* entry seen during PAG search (for stats) */
252 #define UNFSGetCreds 32 /* getting creds for NFS client */
253 /* A flag used by afs_GCPAGs to keep track of
254 * which entries in afs_users need to be deleted.
255 * The lifetime of its presence in the table is the
256 * lifetime of the afs_GCPAGs function.
258 #define TMP_UPAGNotReferenced 128
260 /* values for afs_gcpags */
261 enum { AFS_GCPAGS_NOTCOMPILED = 0, AFS_GCPAGS_OK =
262 1, AFS_GCPAGS_USERDISABLED, AFS_GCPAGS_EPROC0, AFS_GCPAGS_EPROCN,
263 AFS_GCPAGS_EEQPID, AFS_GCPAGS_EINEXACT, AFS_GCPAGS_EPROCEND,
264 AFS_GCPAGS_EPROCWALK, AFS_GCPAGS_ECREDWALK, AFS_GCPAGS_EPIDCHECK,
265 AFS_GCPAGS_ENICECHECK
268 extern afs_int32 afs_gcpags;
269 extern afs_int32 afs_gcpags_procsize;
270 extern afs_int32 afs_bkvolpref;
273 struct unixuser *next; /* next hash pointer */
274 afs_int32 uid; /* search based on uid and cell */
276 afs_int32 vid; /* corresponding vice id in specified cell */
277 short refCount; /* reference count for allocation */
278 char states; /* flag info */
279 afs_int32 tokenTime; /* last time tokens were set, used for timing out conn data */
280 afs_int32 stLen; /* ticket length (if kerberos, includes kvno at head) */
281 char *stp; /* pointer to ticket itself */
282 struct ClearToken ct;
283 struct afs_exporter *exporter; /* more info about the exporter for the remote user */
284 void *cellinfo; /* pointer to cell info (PAG manager only) */
288 /* Per-connection block. */
289 struct conn *next; /* Next dude same server. */
290 struct unixuser *user; /* user validated with respect to. */
291 struct rx_connection *id; /* RPC connid. */
292 struct srvAddr *srvr; /* server associated with this conn */
293 short refCount; /* reference count for allocation */
294 unsigned short port; /* port associated with this connection */
295 char forceConnectFS; /* Should we try again with these tokens? */
301 /* Fid comparison routines */
302 #define FidCmp(a,b) ((a)->Fid.Unique != (b)->Fid.Unique \
303 || (a)->Fid.Vnode != (b)->Fid.Vnode \
304 || (a)->Fid.Volume != (b)->Fid.Volume \
305 || (a)->Cell != (b)->Cell)
307 #define FidMatches(afid,tvc) ((tvc)->fid.Fid.Vnode == (afid)->Fid.Vnode && \
308 (tvc)->fid.Fid.Volume == (afid)->Fid.Volume && \
309 (tvc)->fid.Cell == (afid)->Cell && \
310 ( (tvc)->fid.Fid.Unique == (afid)->Fid.Unique || \
311 (!(afid)->Fid.Unique && ((tvc)->states & CUnique))))
315 * Operations on circular queues implemented with pointers. Note: these queue
316 * objects are always located at the beginning of the structures they are linking.
318 #define QInit(q) ((q)->prev = (q)->next = (q))
319 #define QAdd(q,e) ((e)->next = (q)->next, (e)->prev = (q), \
320 (q)->next->prev = (e), (q)->next = (e))
321 #define QRemove(e) ((e)->next->prev = (e)->prev, (e)->prev->next = (e)->next, (e)->prev = NULL, (e)->next = NULL)
322 #define QNext(e) ((e)->next)
323 #define QPrev(e) ((e)->prev)
324 #define QEmpty(q) ((q)->prev == (q))
325 /* this one takes q1 and sticks it on the end of q2 - that is, the other end, not the end
326 * that things are added onto. q1 shouldn't be empty, it's silly */
327 #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))
329 * Do lots of address arithmetic to go from vlruq to the base of the vcache
330 * structure. Don't move struct vnode, since we think of a struct vcache as
331 * a specialization of a struct vnode
333 #define QTOV(e) ((struct vcache *)(((char *) (e)) - (((char *)(&(((struct vcache *)(e))->vlruq))) - ((char *)(e)))))
334 #define QTOC(e) ((struct cell *)((char *) (e)))
335 #define QTOVH(e) ((struct vcache *)(((char *) (e)) - (((char *)(&(((struct vcache *)(e))->vhashq))) - ((char *)(e)))))
338 #define SRVADDR_ISDOWN 0x20 /* same as SRVR_ISDOWN */
339 #define SRVADDR_NOUSE 0x40 /* Don't use this srvAddr */
341 struct srvAddr *next_bkt; /* next item in hash bucket */
342 struct srvAddr *next_sa; /* another interface on same host */
343 struct server *server; /* back to parent */
344 struct conn *conns; /* All user connections to this server */
345 afs_int32 sa_ip; /* Host addr in network byte order */
346 u_short sa_iprank; /* indiv ip address priority */
347 u_short sa_portal; /* port addr in network byte order */
352 * Values used in the flags field of the server structure below.
354 * AFS_SERVER_FLAG_ACTIVATED Has the server ever had a user connection
355 * associated with it?
357 #define AFS_SERVER_FLAG_ACTIVATED 0x01
358 #define SNO_LHOSTS 0x04
359 #define SYES_LHOSTS 0x08
360 #define SVLSRV_UUID 0x10
361 #define SRVR_ISDOWN 0x20
362 #define SRVR_MULTIHOMED 0x40
363 #define SRVR_ISGONE 0x80
364 #define SNO_INLINEBULK 0x100
365 #define SNO_64BIT 0x200
367 #define afs_serverSetNo64Bit(s) ((s)->srvr->server->flags |= SNO_64BIT)
368 #define afs_serverHasNo64Bit(s) ((s)->srvr->server->flags & SNO_64BIT)
374 afs_int32 addr_uniquifier;
378 struct srvAddr haddr;
381 #define sr_uuid _suid._srvUuid.suuid
382 #define sr_addr_uniquifier _suid._srvUuid.addr_uniquifier
383 #define sr_host _suid._srvId.haddr.ip
384 #define sr_portal _suid._srvId.haddr.portal
385 #define sr_rank _suid._srvId.haddr.ip_rank
386 #define sr_flags _suid._srvId.haddr.flags
387 #define sr_conns _suid._srvId.haddr.conns
388 struct server *next; /* Ptr to next server in hash chain */
389 struct cell *cell; /* Cell in which this host resides */
390 struct afs_cbr *cbrs; /* Return list of callbacks */
391 afs_int32 activationTime; /* Time when this record was first activated */
392 afs_int32 lastDowntimeStart; /* Time when last downtime incident began */
393 afs_int32 numDowntimeIncidents; /* # (completed) downtime incidents */
394 afs_int32 sumOfDowntimes; /* Total downtime experienced, in seconds */
395 struct srvAddr *addr;
396 afs_uint32 flags; /* Misc flags */
399 #define afs_PutServer(servp, locktype)
401 /* structs for some pioctls - these are (or should be)
409 struct sprefrequest_33 {
410 unsigned short offset;
411 unsigned short num_servers;
415 struct sprefrequest { /* new struct for 3.4 */
416 unsigned short offset;
417 unsigned short num_servers;
418 unsigned short flags;
423 unsigned short next_offset;
424 unsigned short num_servers;
425 struct spref servers[1]; /* we overrun this array intentionally... */
429 unsigned short flags;
430 unsigned short num_servers;
431 struct spref servers[1]; /* we overrun this array intentionally... */
433 /* struct for GAG pioctl
436 afs_uint32 showflags, logflags, logwritethruflag, spare[3];
437 unsigned char spare2[128];
441 #define logwritethruON 1
444 afs_int32 rx_initReceiveWindow, rx_maxReceiveWindow, rx_initSendWindow,
445 rx_maxSendWindow, rxi_nSendFrags, rxi_nRecvFrags, rxi_OrphanFragSize;
446 afs_int32 rx_maxReceiveSize, rx_MyMaxSendSize;
447 afs_uint32 spare[21];
450 /* struct for checkservers */
461 /* state bits for volume */
462 #define VRO 1 /* volume is readonly */
463 #define VRecheck 2 /* recheck volume info with server */
464 #define VBackup 4 /* is this a backup volume? */
465 #define VForeign 8 /* this is a non-afs volume */
466 #define VResort 16 /* server order was rearranged, sort when able */
467 #define VMoreReps 32 /* This volume has more replicas than we are */
468 /* keeping track of now -- check with VLDB */
470 enum repstate { not_busy, end_not_busy = 6, rd_busy, rdwr_busy, offline };
473 /* One structure per volume, describing where the volume is located
474 * and where its mount points are. */
475 struct volume *next; /* Next volume in hash list. */
476 afs_int32 cell; /* the cell in which the volume resides */
477 afs_rwlock_t lock; /* the lock for this structure */
478 afs_int32 volume; /* This volume's ID number. */
479 char *name; /* This volume's name, or 0 if unknown */
480 struct server *serverHost[MAXHOSTS]; /* servers serving this volume */
481 enum repstate status[MAXHOSTS]; /* busy, offline, etc */
482 struct VenusFid dotdot; /* dir to access as .. */
483 struct VenusFid mtpoint; /* The mount point for this volume. */
484 afs_int32 rootVnode, rootUnique; /* Volume's root fid */
487 afs_int32 rwVol; /* For r/o vols, original read/write volume. */
488 afs_int32 accessTime; /* last time we used it */
489 afs_int32 vtix; /* volume table index */
490 afs_int32 copyDate; /* copyDate field, for tracking vol releases */
491 afs_int32 expireTime; /* for per-volume callbacks... */
492 short refCount; /* reference count for allocation */
493 char states; /* here for alignment reasons */
496 #define afs_PutVolume(av, locktype) ((av)->refCount--)
498 /* format of an entry in volume info file */
500 afs_int32 cell; /* cell for this entry */
501 afs_int32 volume; /* volume */
502 afs_int32 next; /* has index */
503 struct VenusFid dotdot; /* .. value */
504 struct VenusFid mtpoint; /* mt point's fid */
505 afs_int32 rootVnode, rootUnique; /* Volume's root fid */
509 struct SimpleLocks *next;
511 afs_int32 boff, eoff;
513 #if defined(AFS_AIX32_ENV) || defined(AFS_SUN5_ENV) || defined(AFS_SGI_ENV)
518 /* vcache state bits */
519 #define CStatd 0x00000001 /* has this file ever been stat'd? */
520 #define CBackup 0x00000002 /* file is on a backup volume */
521 #define CRO 0x00000004 /* is it on a read-only volume */
522 #define CMValid 0x00000008 /* is the mount point info valid? */
523 #define CCore 0x00000010 /* storing a core file, needed since we don't get an open */
524 #define CDirty 0x00000020 /* file has been modified since first open (... O_RDWR) */
525 #define CSafeStore 0x00000040 /* close must wait for store to finish (should be in fd) */
526 #define CMAPPED 0x00000080 /* Mapped files; primarily used by SunOS 4.0.x */
527 #define CNSHARE 0x00000100 /* support O_NSHARE semantics */
528 #define CLied 0x00000200
529 #define CTruth 0x00000400
531 #define CWired 0x00000800 /* OSF hack only */
533 #ifdef AFS_DARWIN80_ENV
534 #define CDeadVnode 0x00000800
536 #ifdef AFS_DARWIN_ENV
537 #define CUBCinit 0x00000800
539 #define CWRITE_IGN 0x00000800 /* Next OS hack only */
543 #define CUnique 0x00001000 /* vc's uniquifier - latest unifiquier for fid */
544 #define CForeign 0x00002000 /* this is a non-afs vcache */
545 #define CReadDir 0x00004000 /* readdir in progress */
546 #define CUnlinked 0x00010000
547 #define CBulkStat 0x00020000 /* loaded by a bulk stat, and not ref'd since */
548 #define CUnlinkedDel 0x00040000
549 #define CVFlushed 0x00080000
550 #define CCore1 0x00100000 /* osf1 core file; not same as CCore above */
551 #define CWritingUFS 0x00200000 /* to detect vm deadlock - used by sgi */
552 #define CCreating 0x00400000 /* avoid needless store after open truncate */
553 #define CPageHog 0x00800000 /* AIX - dumping large cores is a page hog. */
554 #define CDCLock 0x02000000 /* Vnode lock held over call to GetDownD */
555 #define CBulkFetching 0x04000000 /* stats are being fetched by bulk stat */
556 #define CExtendedFile 0x08000000 /* extended file via ftruncate call. */
557 #define CVInit 0x10000000 /* being initialized */
559 /* vcache vstate bits */
560 #define VRevokeWait 0x1
561 #define VPageCleaning 0x2 /* Solaris - Cache Trunc Daemon sez keep out */
564 #if defined(AFS_XBSD_ENV) || defined(AFS_DARWIN_ENV)
565 #define vrefCount v->v_usecount
567 #define vrefCount v.v_count
568 #endif /* AFS_XBSD_ENV */
570 #if defined(AFS_DARWIN80_ENV)
571 #define VREFCOUNT_GT(v, y) vnode_isinuse(AFSTOV(v), (y))
572 #elif defined(AFS_XBSD_ENV) || defined(AFS_DARWIN_ENV)
573 #define VREFCOUNT(v) ((v)->vrefCount)
574 #define VREFCOUNT_GT(v, y) (AFSTOV(v)->v_usecount > (y))
575 #elif defined(AFS_LINUX24_ENV)
576 #define VREFCOUNT(v) atomic_read(&(AFSTOV(v)->v_count))
577 #define VREFCOUNT_GT(v, y) (VREFCOUNT(v)>y)
578 #define VREFCOUNT_SET(v, c) atomic_set(&(AFSTOV(v)->v_count), c)
579 #define VREFCOUNT_DEC(v) atomic_dec(&(AFSTOV(v)->v_count))
580 #define VREFCOUNT_INC(v) atomic_inc(&(AFSTOV(v)->v_count))
582 #define VREFCOUNT(v) ((v)->vrefCount)
583 #define VREFCOUNT_GT(v,y) ((v)->vrefCount > (y))
584 #define VREFCOUNT_SET(v, c) (v)->vrefCount = c;
585 #define VREFCOUNT_DEC(v) (v)->vrefCount--;
586 #define VREFCOUNT_INC(v) (v)->vrefCount++;
587 #define d_unhash(d) list_empty(&(d)->d_hash)
588 #define dget_locked(d) dget(d)
591 #define AFS_MAXDV 0x7fffffff /* largest dataversion number */
592 #ifdef AFS_64BIT_CLIENT
593 #define AFS_NOTRUNC 0x7fffffffffffffffLL /* largest positive int64 number */
594 #else /* AFS_64BIT_CLIENT */
595 #define AFS_NOTRUNC 0x7fffffff /* largest dataversion number */
596 #endif /* AFS_64BIT_CLIENT */
598 extern afs_int32 vmPageHog; /* counter for # of vnodes which are page hogs. */
600 #if defined(AFS_DARWIN80_ENV)
601 #define VTOAFS(v) ((struct vcache *)vnode_fsnode((v)))
602 #define AFSTOV(vc) ((vc)->v)
603 #elif defined(AFS_XBSD_ENV) || defined(AFS_DARWIN_ENV) || (defined(AFS_LINUX22_ENV) && !defined(STRUCT_SUPER_HAS_ALLOC_INODE))
604 #define VTOAFS(v) ((struct vcache *)(v)->v_data)
605 #define AFSTOV(vc) ((vc)->v)
607 #define VTOAFS(V) ((struct vcache *)(V))
608 #define AFSTOV(V) (&(V)->v)
611 /* INVARIANTs: (vlruq.next != NULL) == (vlruq.prev != NULL)
612 * nextfree => !vlruq.next && ! vlruq.prev
613 * !(avc->nextfree) && !avc->vlruq.next => (FreeVCList == avc->nextfree)
616 #if defined(AFS_XBSD_ENV) || defined(AFS_DARWIN_ENV) || (defined(AFS_LINUX22_ENV) && !defined(STRUCT_SUPER_HAS_ALLOC_INODE))
619 struct vnode v; /* Has reference count in v.v_count */
621 struct afs_q vlruq; /* lru q next and prev */
622 #if !defined(AFS_LINUX22_ENV)
623 struct vcache *nextfree; /* next on free list (if free) */
625 struct vcache *hnext; /* Hash next */
626 struct afs_q vhashq; /* Hashed per-volume list */
630 afs_hyper_t DataVersion;
634 afs_uint16 Mode; /* XXXX Should be afs_int32 XXXX */
635 afs_uint16 LinkCount;
636 #ifdef AFS_DARWIN80_ENV
639 /* vnode type is in v.v_type */
642 afs_rwlock_t lock; /* The lock on the vcache contents. */
643 #if defined(AFS_SUN5_ENV)
644 /* Lock used to protect the activeV, multipage, and vstates fields.
645 * Do not try to get the vcache lock when the vlock is held */
647 #endif /* defined(AFS_SUN5_ENV) */
648 #if defined(AFS_SUN5_ENV)
652 #ifdef AFS_BOZONLOCK_ENV
653 afs_bozoLock_t pvnLock; /* see locks.x */
658 #if defined(AFS_AIX51_ENV)
666 int ownslock; /* pid of owner of excl lock, else 0 - defect 3083 */
668 #ifdef AFS_DARWIN80_ENV
670 #elif defined(AFS_DARWIN_ENV)
671 struct lock__bsd__ rwlock;
676 afs_int32 parentVnode; /* Parent dir, if a file. */
677 afs_int32 parentUnique;
678 struct VenusFid *mvid; /* Either parent dir (if root) or root (if mt pt) */
679 char *linkData; /* Link data if a symlink. */
680 afs_hyper_t flushDV; /* data version last flushed from text */
681 afs_hyper_t mapDV; /* data version last flushed from map */
682 afs_size_t truncPos; /* truncate file to this position at next store */
683 struct server *callback; /* The callback host, if any */
684 afs_uint32 cbExpires; /* time the callback expires */
685 struct afs_q callsort; /* queue in expiry order, sort of */
686 struct axscache *Access; /* a list of cached access bits */
687 afs_int32 anyAccess; /* System:AnyUser's access to this. */
688 afs_int32 last_looker; /* pag/uid from last lookup here */
689 #if defined(AFS_SUN5_ENV)
691 #endif /* defined(AFS_SUN5_ENV) */
692 struct SimpleLocks *slocks;
693 short opens; /* The numbers of opens, read or write, on this file. */
694 short execsOrWriters; /* The number of execs (if < 0) or writers (if > 0) of
696 short flockCount; /* count of flock readers, or -1 if writer */
697 char mvstat; /* 0->normal, 1->mt pt, 2->root. */
698 afs_uint32 states; /* state bits */
699 #if defined(AFS_SUN5_ENV)
700 afs_uint32 vstates; /* vstate bits */
701 #endif /* defined(AFS_SUN5_ENV) */
702 struct dcache *dchint;
703 struct dcache *dcreaddir; /* dcache for in-progress readdir */
704 unsigned int readdir_pid; /* pid of the thread in readdir */
705 #ifdef AFS_LINUX22_ENV
706 u_short mapcnt; /* Number of mappings of this file. */
708 #if defined(AFS_SGI_ENV)
709 daddr_t lastr; /* for read-ahead */
711 uint64_t vc_rwlockid; /* kthread owning rwlock */
713 short vc_rwlockid; /* pid of process owning rwlock */
715 short vc_locktrips; /* # of rwlock reacquisitions */
716 sema_t vc_rwlock; /* vop_rwlock for afs */
717 pgno_t mapcnt; /* # of pages mapped */
718 struct cred *cred; /* last writer's cred */
720 struct bhv_desc vc_bhv_desc; /* vnode's behavior data. */
722 #endif /* AFS_SGI_ENV */
723 afs_int32 vc_error; /* stash write error for this vnode. */
724 int xlatordv; /* Used by nfs xlator */
725 struct AFS_UCRED *uncred;
726 int asynchrony; /* num kbytes to store behind */
728 short multiPage; /* count of multi-page getpages in progress */
732 #define DONT_CHECK_MODE_BITS 0
733 #define CHECK_MODE_BITS 1
734 #define CMB_ALLOW_EXEC_AS_READ 2 /* For the NFS xlator */
736 #if defined(AFS_SGI_ENV)
737 #define AVCRWLOCK(avc) (valusema(&(avc)->vc_rwlock) <= 0)
739 /* SGI vnode rwlock macros and flags. */
740 #ifndef AFS_SGI62_ENV
741 /* The following are defined here. SGI 6.2 declares them in vnode.h */
742 #define VRWLOCK_READ 0
743 #define VRWLOCK_WRITE 1
744 #define VRWLOCK_WRITE_DIRECT 2
749 #define AFS_RWLOCK_T vrwlock_t
751 #define AFS_RWLOCK_T int
752 #endif /* AFS_SGI62_ENV */
754 #include <ksys/behavior.h>
755 #define AFS_RWLOCK(V,F) \
756 afs_rwlock(&VTOAFS(V)->vc_bhv_desc, (F));
757 #define AFS_RWUNLOCK(V,F) \
758 afs_rwunlock(&VTOAFS(V)->vc_bhv_desc, (F));
761 #define AFS_RWLOCK(V,F) afs_rwlock((vnode_t *)(V), (F) )
762 #define AFS_RWUNLOCK(V,F) afs_rwunlock((vnode_t *)(V), (F) )
764 #else /* AFS_SGI53_ENV */
765 #define AFS_RWLOCK(V,F) afs_rwlock((V))
766 #define AFS_RWUNLOCK(V,F) afs_rwunlock((V))
767 #endif /* AFS_SGI53_ENV */
768 #endif /* AFS_SGI_ENV */
772 afs_hyper_t DataVersion;
774 afs_int32 parentVnode;
775 afs_int32 parentUnique;
779 afs_int32 randomUid[CPSIZE];
780 afs_int32 callback; /* Now a pointer to 'server' struct */
782 afs_int32 randomAccess[CPSIZE];
785 short execsOrWriters;
792 afs_int32 callerAccess;
803 /* CM inititialization parameters. What CM actually used after calculations
804 * based on passed in arguments.
806 #define CMI_VERSION 1 /* increment when adding new fields. */
807 struct cm_initparams {
812 int cmi_nVolumeCaches;
813 int cmi_firstChunkSize;
814 int cmi_otherChunkSize;
815 int cmi_cacheSize; /* The original cache size, in 1K blocks. */
816 unsigned cmi_setTime:1;
817 unsigned cmi_memCache:1;
818 int spare[16 - 9]; /* size of struct is 16 * 4 = 64 bytes */
822 /*----------------------------------------------------------------------
823 * AFS Data cache definitions
825 * Each entry describes a Unix file on the local disk that is
826 * is serving as a cached copy of all or part of a Vice file.
827 * Entries live in circular queues for each hash table slot
829 * Which queue is this thing in? Good question.
830 * A struct dcache entry is in the freeDSlot queue when not associated with a cache slot (file).
831 * Otherwise, it is in the DLRU queue. The freeDSlot queue uses the lruq.next field as
832 * its "next" pointer.
834 * Cache entries in the DLRU queue are either associated with vice files, in which case
835 * they are hashed by afs_dvnextTbl and afs_dcnextTbl pointers, or they are in the freeDCList
836 * and are not associated with any vice file. This last list uses the afs_dvnextTbl pointer for
837 * its "next" pointer.
838 *----------------------------------------------------------------------*/
840 #define NULLIDX (-1) /* null index definition */
841 /* struct dcache states bits */
845 #define DWriting 8 /* file being written (used for cache validation) */
847 /* dcache data flags */
848 #define DFEntryMod 0x02 /* has entry itself been modified? */
849 #define DFFetching 0x04 /* file is currently being fetched */
851 /* dcache meta flags */
852 #define DFNextStarted 0x01 /* next chunk has been prefetched already */
853 #define DFFetchReq 0x10 /* someone is waiting for DFFetching to go on */
856 /* flags in afs_indexFlags array */
857 #define IFEverUsed 1 /* index entry has >= 1 byte of data */
858 #define IFFree 2 /* index entry in freeDCList */
859 #define IFDataMod 4 /* file needs to be written out */
860 #define IFFlag 8 /* utility flag */
861 #define IFDirtyPages 16
862 #define IFAnyPages 32
863 #define IFDiscarded 64 /* index entry in discardDCList */
866 char *in; /* input buffer */
867 char *out; /* output buffer */
868 short in_size; /* Size of input buffer <= 2K */
869 short out_size; /* Maximum size of output buffer, <= 2K */
873 * This version of afs_ioctl is required to pass in 32 bit user space
874 * pointers into a 64 bit kernel.
885 /* CacheItems file has a header of type struct afs_fheader
886 * (keep aligned properly). Since we already have sgi_62 clients running
887 * with a 32 bit inode, a change is required to the header so that
888 * they can distinguish the old 32 bit inode CacheItems file and zap it
889 * instead of using it.
892 #define AFS_FHMAGIC 0x7635abaf /* uses version number */
894 #if defined(AFS_SUN57_64BIT_ENV)
895 #define AFS_CI_VERSION 3
897 #define AFS_CI_VERSION 2
900 afs_int32 firstCSize;
901 afs_int32 otherCSize;
904 #if defined(AFS_SGI61_ENV) || defined(AFS_SUN57_64BIT_ENV)
905 /* Using ino64_t here so that user level debugging programs compile
906 * the size correctly.
908 #define afs_inode_t ino64_t
910 #if defined(AFS_LINUX_64BIT_KERNEL) && !defined(AFS_S390X_LINUX24_ENV)
911 #define afs_inode_t long
913 #if defined(AFS_AIX51_ENV) || defined(AFS_HPUX1123_ENV)
914 #define afs_inode_t ino_t
916 #define afs_inode_t afs_int32
922 afs_int32 fid; /* is adc->index, the cache file number */
923 afs_inode_t inode; /* is adc->f.inode, the inode number of the cac\
926 afs_int32 accesstime;
927 struct buffer *hashNext;
932 #if defined(AFS_USEBUFFERS)
935 afs_rwlock_t lock; /* the lock for this structure */
938 /* kept on disk and in dcache entries */
940 struct VenusFid fid; /* Fid for this file */
941 afs_int32 modTime; /* last time this entry was modified */
942 afs_hyper_t versionNo; /* Associated data version number */
943 afs_int32 chunk; /* Relative chunk number */
944 afs_inode_t inode; /* Unix inode for this chunk */
945 afs_int32 chunkBytes; /* Num bytes in this chunk */
946 char states; /* Has this chunk been modified? */
949 /* magic numbers to specify the cache type */
951 #define AFS_FCACHE_TYPE_UFS 0x0
952 #define AFS_FCACHE_TYPE_MEM 0x1
953 #define AFS_FCACHE_TYPE_NFS 0x2
954 #define AFS_FCACHE_TYPE_EPI 0x3
958 struct afs_q lruq; /* Free queue for in-memory images */
959 struct afs_q dirty; /* Queue of dirty entries that need written */
960 afs_rwlock_t lock; /* Protects validPos, some f */
961 afs_rwlock_t tlock; /* Atomizes updates to refCount */
962 afs_rwlock_t mflock; /* Atomizes accesses/updates to mflags */
963 afs_size_t validPos; /* number of valid bytes during fetch */
964 afs_int32 index; /* The index in the CacheInfo file */
965 short refCount; /* Associated reference count. */
966 char dflags; /* Data flags */
967 char mflags; /* Meta flags */
968 struct fcache f; /* disk image */
969 afs_int32 bucket; /* which bucket these dcache entries are in */
973 * dcache.lock protects the actual contents of the cache file (in
974 * f.inode), subfields of f except those noted below, dflags and
977 * dcache.tlock is used to make atomic updates to refCount. Zero
978 * refCount dcache entries are protected by afs_xdcache instead of
981 * dcache.mflock is used to access and update mflags. It cannot be
982 * held without holding the corresponding dcache.lock. Updating
983 * mflags requires holding dcache.lock(R) and dcache.mflock(W), and
984 * checking for mflags requires dcache.lock(R) and dcache.mflock(R).
985 * Note that dcache.lock(W) gives you the right to update mflags,
986 * as dcache.mflock(W) can only be held with dcache.lock(R).
988 * dcache.index, dcache.f.fid, dcache.f.chunk and dcache.f.inode are
989 * write-protected by afs_xdcache and read-protected by refCount.
990 * Once an entry is referenced, these values cannot change, and if
991 * it's on the free list (with refCount=0), it can be reused for a
992 * different file/chunk. These values can only be written while
993 * holding afs_xdcache(W) and allocating this dcache entry (thereby
994 * ensuring noone else has a refCount on it).
999 struct memCacheEntry {
1000 int size; /* # of valid bytes in this entry */
1001 int dataSize; /* size of allocated data area */
1002 afs_lock_t afs_memLock;
1003 char *data; /* bytes */
1006 /* macro to mark a dcache entry as bad */
1009 (x)->f.fid.Fid.Unique = 0; \
1010 afs_indexUnique[(x)->index] = 0; \
1011 (x)->dflags |= DFEntryMod; \
1014 /* FakeOpen and Fake Close used to be real subroutines. They're only used in
1015 * sun_subr and afs_vnodeops, and they're very frequently called, so I made
1016 * them into macros. They do:
1017 * FakeOpen: fake the file being open for writing. avc->lock must be held
1018 * in write mode. Having the file open for writing is like having a DFS
1019 * write-token: you're known to have the best version of the data around,
1020 * and so the CM won't let it be overwritten by random server info.
1021 * FakeClose: undo the effects of FakeOpen, noting that we want to ensure
1022 * that a real close eventually gets done. We use CCore to achieve this if
1023 * we would end up closing the file. avc->lock must be held in write mode */
1025 #ifdef AFS_AIX_IAUTH_ENV
1026 #define CRKEEP(V, C) (V)->linkData = (char*)crdup((C))
1028 #define CRKEEP(V, C) crhold((C)); (V)->linkData = (char*)(C)
1031 #define afs_FakeOpen(avc) { avc->opens++; avc->execsOrWriters++; }
1032 #define afs_FakeClose(avc, acred) \
1033 { if (avc->execsOrWriters == 1) { \
1034 /* we're the last writer, just use CCore flag */ \
1035 avc->states |= CCore; /* causes close to be called later */ \
1037 /* The cred and vnode holds will be released in afs_FlushActiveVcaches */ \
1038 VN_HOLD(AFSTOV(avc)); /* So it won't disappear */ \
1039 CRKEEP(avc, acred); /* Should use a better place for the creds */ \
1042 /* we're not the last writer, let the last one do the store-back for us */ \
1044 avc->execsOrWriters--; \
1048 #define AFS_ZEROS 64 /* zero buffer */
1050 /*#define afs_DirtyPages(avc) (((avc)->states & CDirty) || osi_VMDirty_p((avc)))*/
1051 #define afs_DirtyPages(avc) ((avc)->states & CDirty)
1053 #define afs_InReadDir(avc) (((avc)->states & CReadDir) && (avc)->readdir_pid == MyPidxx)
1055 /* The PFlush algorithm makes use of the fact that Fid.Unique is not used in
1056 below hash algorithms. Change it if need be so that flushing algorithm
1057 doesn't move things from one hash chain to another
1059 /* extern int afs_dhashsize; */
1060 #define DCHash(v, c) ((((v)->Fid.Vnode + (v)->Fid.Volume + (c))) & (afs_dhashsize-1))
1061 /*Vnode, Chunk -> Hash table index */
1062 #define DVHash(v) ((((v)->Fid.Vnode + (v)->Fid.Volume )) & (afs_dhashsize-1))
1063 /*Vnode -> Other hash table index */
1064 /* don't hash on the cell, our callback-breaking code sometimes fails to compute
1065 the cell correctly, and only scans one hash bucket */
1066 #define VCHash(fid) (((fid)->Fid.Volume + (fid)->Fid.Vnode) & (VCSIZE-1))
1067 /* Hash only on volume to speed up volume callbacks. */
1068 #define VCHashV(fid) ((fid)->Fid.Volume & (VCSIZE-1))
1070 extern struct dcache **afs_indexTable; /*Pointers to in-memory dcache entries */
1071 extern afs_int32 *afs_indexUnique; /*dcache entry Fid.Unique */
1072 extern afs_int32 *afs_dvnextTbl; /*Dcache hash table links */
1073 extern afs_int32 *afs_dcnextTbl; /*Dcache hash table links */
1074 extern afs_int32 afs_cacheFiles; /*Size of afs_indexTable */
1075 extern afs_int32 afs_cacheBlocks; /*1K blocks in cache */
1076 extern afs_int32 afs_cacheStats; /*Stat entries in cache */
1077 extern struct vcache *afs_vhashT[VCSIZE]; /*Stat cache hash table */
1078 extern struct afs_q afs_vhashTV[VCSIZE]; /* cache hash table on volume */
1079 extern afs_int32 afs_initState; /*Initialization state */
1080 extern afs_int32 afs_termState; /* Termination state */
1081 extern struct VenusFid afs_rootFid; /*Root for whole file system */
1082 extern afs_int32 afs_allCBs; /* Count of callbacks */
1083 extern afs_int32 afs_oddCBs; /* Count of odd callbacks */
1084 extern afs_int32 afs_evenCBs; /* Count of even callbacks */
1085 extern afs_int32 afs_allZaps; /* Count of fid deletes */
1086 extern afs_int32 afs_oddZaps; /* Count of odd fid deletes */
1087 extern afs_int32 afs_evenZaps; /* Count of even fid deletes */
1088 extern struct brequest afs_brs[NBRS]; /* request structures */
1090 #define UHash(auid) ((auid) & (NUSERS-1))
1091 #define VHash(avol) ((avol)&(NVOLS-1))
1092 #define SHash(aserv) ((ntohl(aserv)) & (NSERVERS-1))
1093 #define FVHash(acell,avol) (((avol)+(acell)) & (NFENTRIES-1))
1095 /* Performance hack - we could replace VerifyVCache2 with the appropriate
1096 * GetVCache incantation, and could eliminate even this code from afs_UFSRead
1097 * by making intentionally invalidating quick.stamp in the various callbacks
1098 * expiration/breaking code */
1099 #ifdef AFS_DARWIN_ENV
1100 #define afs_VerifyVCache(avc, areq) \
1101 (((avc)->states & CStatd) ? (osi_VM_Setup(avc, 0), 0) : \
1102 afs_VerifyVCache2((avc),areq))
1104 #define afs_VerifyVCache(avc, areq) \
1105 (((avc)->states & CStatd) ? 0 : afs_VerifyVCache2((avc),areq))
1108 #define DO_STATS 1 /* bits used by FindVCache */
1113 /* values for flag param of afs_CheckVolumeNames */
1114 #define AFS_VOLCHECK_EXPIRED 0x1 /* volumes whose callbacks have expired */
1115 #define AFS_VOLCHECK_BUSY 0x2 /* volumes which were marked busy */
1116 #define AFS_VOLCHECK_MTPTS 0x4 /* mount point invalidation also */
1117 #define AFS_VOLCHECK_FORCE 0x8 /* do all forcibly */
1121 #define AFS_FSPORT ((unsigned short) htons(7000))
1122 #define AFS_VLPORT ((unsigned short) htons(7003))
1124 #define afs_read(avc, uio, acred, albn, abpp, nolock) \
1125 (*(afs_cacheType->vread))(avc, uio, acred, albn, abpp, nolock)
1126 #define afs_write(avc, uio, aio, acred, nolock) \
1127 (*(afs_cacheType->vwrite))(avc, uio, aio, acred, nolock)
1129 #define afs_rdwr(avc, uio, rw, io, cred) \
1130 (((rw) == UIO_WRITE) ? afs_write(avc, uio, io, cred, 0) : afs_read(avc, uio, cred, 0, 0, 0))
1131 #define afs_nlrdwr(avc, uio, rw, io, cred) \
1132 (((rw) == UIO_WRITE) ? afs_write(avc, uio, io, cred, 1) : afs_read(avc, uio, cred, 0, 0, 1))
1134 /* Cache size truncation uses the following low and high water marks:
1135 * If the cache is more than 95% full (CM_DCACHECOUNTFREEPCT), the cache
1136 * truncation daemon is awakened and will free up space until the cache is 85%
1137 * (CM_DCACHESPACEFREEPCT - CM_DCACHEEXTRAPCT) full.
1138 * afs_UFSWrite and afs_GetDCache (when it needs to fetch data) will wait on
1139 * afs_WaitForCacheDrain if the cache is 98% (CM_WAITFORDRAINPCT) full.
1140 * afs_GetDownD wakes those processes once the cache is 95% full
1141 * (CM_CACHESIZEDRAINEDPCT).
1143 #define CM_MAXDISCARDEDCHUNKS 16 /* # of chunks */
1144 #define CM_DCACHECOUNTFREEPCT 95 /* max pct of chunks in use */
1145 #define CM_DCACHESPACEFREEPCT 90 /* max pct of space in use */
1146 #define CM_DCACHEEXTRAPCT 5 /* extra to get when freeing */
1147 #define CM_CACHESIZEDRAINEDPCT 95 /* wakeup processes when down to here. */
1148 #define CM_WAITFORDRAINPCT 98 /* sleep if cache is this full. */
1150 /* when afs_cacheBlocks is large, settle for slightly decreased precision */
1151 #define PERCENT(p, v) \
1152 ((afs_cacheBlocks & 0xffe00000) ? ((v) / 100 * (p)) : ((p) * (v) / 100))
1154 #define afs_CacheIsTooFull() \
1155 (afs_blocksUsed - afs_blocksDiscarded > \
1156 PERCENT(CM_DCACHECOUNTFREEPCT, afs_cacheBlocks) || \
1157 afs_freeDCCount - afs_discardDCCount < \
1158 PERCENT(100 - CM_DCACHECOUNTFREEPCT, afs_cacheFiles))
1160 /* Handy max length of a numeric string. */
1161 #define CVBS 12 /* max afs_int32 is 2^32 ~ 4*10^9, +1 for NULL, +luck */
1163 #define refpanic(foo) if (afs_norefpanic) \
1164 { printf( foo ); afs_norefpanic++;} else osi_Panic( foo )
1167 ** these are defined in the AIX source code sys/fs_locks.h but are not
1168 ** defined anywhere in the /usr/include directory
1170 #if defined(AFS_AIX41_ENV)
1171 #define VN_LOCK(vp) simple_lock(&(vp)->v_lock)
1172 #define VN_UNLOCK(vp) simple_unlock(&(vp)->v_lock)
1175 /* get a file's serial number from a vnode */
1176 #ifndef afs_vnodeToInumber
1177 #if defined(AFS_SGI62_ENV) || defined(AFS_HAVE_VXFS) || defined(AFS_DARWIN_ENV)
1178 #define afs_vnodeToInumber(V) VnodeToIno(V)
1180 #ifdef AFS_DECOSF_ENV
1181 #define afs_vnodeToInumber(V) osi_vnodeToInumber(V)
1183 #define afs_vnodeToInumber(V) (VTOI(V)->i_number)
1184 #endif /* AFS_DECOSF_ENV */
1185 #endif /* AFS_SGI62_ENV */
1188 /* get a file's device number from a vnode */
1189 #ifndef afs_vnodeToDev
1190 #if defined(AFS_SGI62_ENV) || defined(AFS_HAVE_VXFS) || defined(AFS_DARWIN_ENV)
1191 #define afs_vnodeToDev(V) VnodeToDev(V)
1192 #elif defined(AFS_DECOSF_ENV)
1193 #define afs_vnodeToDev(V) osi_vnodeToDev(V)
1195 #define afs_vnodeToDev(V) (VTOI(V)->i_dev)
1200 /* Note: this should agree with the definition in kdump.c */
1201 #if defined(AFS_OSF_ENV)
1202 #if !defined(UKERNEL)
1203 #define AFS_USEBUFFERS 1
1207 #if !defined(UKERNEL) && !defined(HAVE_STRUCT_BUF)
1208 /* declare something so that prototypes don't flip out */
1209 /* appears struct buf stuff is only actually passed around as a pointer,
1210 except with libuafs, in which case it is actually defined */
1215 /* fakestat support: opaque storage for afs_EvalFakeStat to remember
1216 * what vcache should be released.
1218 struct afs_fakestat_state {
1222 struct vcache *root_vp;
1225 extern int afs_fakestat_enable;
1227 #endif /* _AFS_H_ */