2 * Copyright 2000, International Business Machines Corporation and others.
5 * This software has been released under the terms of the IBM Public
6 * License. For details, see the LICENSE file in the top-level source
7 * directory or online at http://www.openafs.org/dl/license10.html
13 * AFS system call opcodes
16 #include <afs/afs_args.h>
18 #include "../afs/afs_args.h"
22 /* Upper bound on number of iovecs out uio routines will deal with. */
23 #define AFS_MAXIOVCNT 16
26 extern int afs_shuttingdown;
29 * Macros to uniquely identify the AFS vfs struct
31 #define AFS_VFSMAGIC 0x1234
32 #if defined(AFS_SUN_ENV) || defined(AFS_HPUX90_ENV) || defined(AFS_LINUX20_ENV)
33 #define AFS_VFSFSID 99
35 #if defined(AFS_SGI_ENV)
36 #define AFS_VFSFSID afs_fstype
38 #define AFS_VFSFSID AFS_MOUNT_AFS
42 /* The following are various levels of afs debugging */
43 #define AFSDEB_GENERAL 1 /* Standard debugging */
44 #define AFSDEB_NETWORK 2 /* low level afs networking */
45 #define AFSDEB_RX 4 /* RX debugging */
46 #define AFSDEB_VNLAYER 8 /* interface layer to AFS (aixops, gfsops, etc) */
48 /* generic undefined vice id */
51 /* The basic defines for the Andrew file system
52 better keep things powers of two so "& (foo-1)" hack works for masking bits */
53 #define MAXHOSTS 13 /* max hosts per single volume */
54 #define OMAXHOSTS 8 /* backwards compatibility */
55 #define MAXCELLHOSTS 8 /* max vldb servers per cell */
56 #define NBRS 15 /* max number of queued daemon requests */
57 #define NUSERS 16 /* hash table size for unixuser table */
58 #define NSERVERS 16 /* hash table size for server table */
59 #define NVOLS 64 /* hash table size for volume table */
60 #define NFENTRIES 256 /* hash table size for disk volume table */
61 #define VCSIZE 1024 /* stat cache hash table size */
62 #define DCSIZE 512 /* disk cache hash table size */
63 #define PIGGYSIZE 1350 /* max piggyback size */
64 #define MAXVOLS 128 /* max vols we can store */
65 #define MAXSYSNAME 128 /* max sysname (i.e. @sys) size */
66 #define MAXNUMSYSNAMES 16 /* max that current constants allow */
67 #define NOTOKTIMEOUT (2*3600) /* time after which to timeout conns sans tokens */
68 #define NOPAG 0xffffffff
69 #define AFS_NCBRS 300 /* max # of call back return entries */
70 #define AFS_MAXCBRSCALL 16 /* max to return in a given call */
71 #define AFS_SALLOC_LOW_WATER 250 /* Min free blocks before allocating more */
72 #define AFS_LRALLOCSIZ 4096 /* "Large" allocated size */
74 #define AFS_NRXPACKETS 80
75 #define AFS_RXDEADTIME 50
76 #define AFS_HARDDEADTIME 120
84 /* flags to use with AFSOP_CACHEINIT */
85 #define AFSCALL_INIT_MEMCACHE 0x1 /* use a memory-based cache */
87 /* below here used only for kernel procedures */
89 /* Store synchrony flags - SYNC means that data should be forced to server's
90 * disk immediately upon completion. */
93 #define AFS_VMSYNC_INVAL 2 /* sync and invalidate pages */
94 #define AFS_LASTSTORE 4
97 /* background request structure */
100 #define BOP_NOOP 0 /* leave 0 unused */
101 #define BOP_FETCH 1 /* parm1 is chunk to get */
102 #define BOP_STORE 2 /* parm1 is chunk to store */
103 #define BOP_PATH 3 /* parm1 is path, parm2 is chunk to fetch */
105 #define B_DONTWAIT 1 /* On failure return; don't wait */
107 /* protocol is: refCount is incremented by user to take block out of free pool.
108 Next, BSTARTED is set when daemon finds request. This prevents
109 other daemons from picking up the same request. Finally, when
110 request is done, refCount is zeroed. BDONE and BWAIT are used by
111 dudes waiting for operation to proceed to a certain point before returning.
113 #define BSTARTED 1 /* request picked up by a daemon */
114 #define BUVALID 2 /* code is valid (store) */
115 #define BUWAIT 4 /* someone is waiting for BUVALID */
117 struct vcache *vnode; /* vnode to use, with vrefcount bumped */
118 struct AFS_UCRED *cred; /* credentials to use for operation */
119 long parm[BPARMS]; /* random parameters - long's work for
120 * both 32 and 64 bit platforms.
122 afs_int32 code; /* return code */
123 short refCount; /* use counter for this structure */
124 char opcode; /* what to do (store, fetch, etc) */
125 char flags; /* free, etc */
133 afs_int32 AuthHandle;
134 char HandShakeKey[8];
136 afs_int32 BeginTimestamp;
137 afs_int32 EndTimestamp;
141 afs_int32 Cell; /* better sun packing if at end of structure */
145 /* Temporary struct to be passed between afs_fid and afs_vget; in SunOS4.x we can only pass a maximum of 10 bytes for a handle (we ideally need 16!) */
148 afs_int32 CellAndUnique;
151 /* The actual number of bytes in the SmallFid, not the sizeof struct. */
152 #define SIZEOF_SMALLFID 10
156 * Queues implemented with both pointers and short offsets into a disk file.
164 afs_int32 uid; /* user id making the request */
165 afs_int32 busyCount; /* how many busies we've seen so far */
166 afs_int32 flags; /* things like O_SYNC, O_NONBLOCK go here */
167 char initd; /* if non-zero, non-uid fields meaningful */
168 char accessError; /* flags for overriding error return code */
169 char volumeError; /* encountered a missing or busy volume */
170 char networkError; /* encountered network problems */
171 char permWriteError; /* fileserver returns permenent error. */
176 /* structure linked off of a server to keep track of queued returned
177 * callbacks. Sent asynchronously when we run a little low on free dudes.
180 struct afs_cbr *next;
185 #define LOCALCELL 1 /* local cell's number is always 1 */
188 #define CPrimary 1 /* on if this is the primary cell */
189 #define CNoSUID 2 /* 1 if no suid progs can run from this cell */
190 #define CHasVolRef 16 /* Volumes were referenced in this cell*/
191 #define CLinkedCell 32
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 time_t timeout; /* data expire time, if non-zero */
205 struct cell *alias; /* what this cell is an alias for */
208 #define afs_PutCell(cellp, locktype)
210 /* the unixuser flag bit definitions */
211 #define UHasTokens 1 /* are the st and ct fields valid (ever set)? */
212 #define UTokensBad 2 /* are tokens bad? */
213 #define UPrimary 4 /* on iff primary identity */
214 #define UNeedsReset 8 /* needs afs_ResetAccessCache call done */
215 #define UPAGCounted 16 /* entry seen during PAG search (for stats) */
216 /* A flag used by afs_GCPAGs to keep track of
217 * which entries in afs_users need to be deleted.
218 * The lifetime of its presence in the table is the
219 * lifetime of the afs_GCPAGs function.
221 #define TMP_UPAGNotReferenced 128
223 /* values for afs_gcpags */
225 { AFS_GCPAGS_NOTCOMPILED = 0
227 , AFS_GCPAGS_USERDISABLED
231 , AFS_GCPAGS_EINEXACT
232 , AFS_GCPAGS_EPROCEND
233 , AFS_GCPAGS_EPROCWALK
234 , AFS_GCPAGS_ECREDWALK
235 , AFS_GCPAGS_EPIDCHECK
236 , AFS_GCPAGS_ENICECHECK
239 extern afs_int32 afs_gcpags;
240 extern afs_int32 afs_gcpags_procsize;
243 struct unixuser *next; /* next hash pointer */
244 afs_int32 uid; /* search based on uid and cell */
246 afs_int32 vid; /* corresponding vice id in specified cell */
247 short refCount; /* reference count for allocation */
248 char states; /* flag info */
249 afs_int32 tokenTime; /* last time tokens were set, used for timing out conn data */
250 afs_int32 stLen; /* ticket length (if kerberos, includes kvno at head) */
251 char *stp; /* pointer to ticket itself */
252 struct ClearToken ct;
253 struct afs_exporter *exporter; /* more info about the exporter for the remote user */
257 /* Per-connection block. */
258 struct conn *next; /* Next dude same server. */
259 struct unixuser *user; /* user validated with respect to. */
260 struct rx_connection *id; /* RPC connid. */
261 struct srvAddr *srvr; /* server associated with this conn */
262 short refCount; /* reference count for allocation */
263 unsigned short port; /* port associated with this connection */
264 char forceConnectFS; /* Should we try again with these tokens? */
270 /* Fid comparison routines */
271 #define FidCmp(a,b) ((a)->Fid.Unique != (b)->Fid.Unique \
272 || (a)->Fid.Vnode != (b)->Fid.Vnode \
273 || (a)->Fid.Volume != (b)->Fid.Volume \
274 || (a)->Cell != (b)->Cell)
276 #define FidMatches(afid,tvc) ((tvc)->fid.Fid.Vnode == (afid)->Fid.Vnode && \
277 (tvc)->fid.Fid.Volume == (afid)->Fid.Volume && \
278 (tvc)->fid.Cell == (afid)->Cell && \
279 ( (tvc)->fid.Fid.Unique == (afid)->Fid.Unique || \
280 (!(afid)->Fid.Unique && ((tvc)->states & CUnique))))
284 * Operations on circular queues implemented with pointers. Note: these queue
285 * objects are always located at the beginning of the structures they are linking.
287 #define QInit(q) ((q)->prev = (q)->next = (q))
288 #define QAdd(q,e) ((e)->next = (q)->next, (e)->prev = (q), \
289 (q)->next->prev = (e), (q)->next = (e))
290 #define QRemove(e) ((e)->next->prev = (e)->prev, (e)->prev->next = (e)->next)
291 #define QNext(e) ((e)->next)
292 #define QPrev(e) ((e)->prev)
293 #define QEmpty(q) ((q)->prev == (q))
294 /* this one takes q1 and sticks it on the end of q2 - that is, the other end, not the end
295 * that things are added onto. q1 shouldn't be empty, it's silly */
296 #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))
298 * Do lots of address arithmetic to go from vlruq to the base of the vcache
299 * structure. Don't move struct vnode, since we think of a struct vcache as
300 * a specialization of a struct vnode
302 #define QTOV(e) ((struct vcache *)(((char *) (e)) - (((char *)(&(((struct vcache *)(e))->vlruq))) - ((char *)(e)))))
303 #define QTOC(e) ((struct cell *)((char *) (e)))
306 #define SRVADDR_ISDOWN 0x20 /* same as SRVR_ISDOWN */
307 #define SRVADDR_NOUSE 0x40 /* Don't use this srvAddr */
309 struct srvAddr *next_bkt; /* next item in hash bucket */
310 struct srvAddr *next_sa; /* another interface on same host */
311 struct server *server; /* back to parent */
312 struct conn *conns; /* All user connections to this server */
313 afs_int32 sa_ip; /* Host addr in network byte order */
314 u_short sa_iprank; /* indiv ip address priority */
315 u_short sa_portal; /* port addr in network byte order */
320 * Values used in the flags field of the server structure below.
322 * AFS_SERVER_FLAG_ACTIVATED Has the server ever had a user connection
323 * associated with it?
325 #define AFS_SERVER_FLAG_ACTIVATED 0x01
326 #define SNO_LHOSTS 0x04
327 #define SYES_LHOSTS 0x08
328 #define SVLSRV_UUID 0x10
329 #define SRVR_ISDOWN 0x20
330 #define SRVR_MULTIHOMED 0x40
331 #define SRVR_ISGONE 0x80
337 afs_int32 addr_uniquifier;
341 struct srvAddr haddr;
344 #define sr_uuid _suid._srvUuid.suuid
345 #define sr_addr_uniquifier _suid._srvUuid.addr_uniquifier
346 #define sr_host _suid._srvId.haddr.ip
347 #define sr_portal _suid._srvId.haddr.portal
348 #define sr_rank _suid._srvId.haddr.ip_rank
349 #define sr_flags _suid._srvId.haddr.flags
350 #define sr_conns _suid._srvId.haddr.conns
351 struct server *next; /* Ptr to next server in hash chain */
352 struct cell *cell; /* Cell in which this host resides */
353 struct afs_cbr *cbrs; /* Return list of callbacks */
354 afs_int32 activationTime; /* Time when this record was first activated */
355 afs_int32 lastDowntimeStart; /* Time when last downtime incident began */
356 afs_int32 numDowntimeIncidents; /* # (completed) downtime incidents */
357 afs_int32 sumOfDowntimes; /* Total downtime experienced, in seconds */
358 struct srvAddr *addr;
359 char flags; /* Misc flags*/
362 #define afs_PutServer(servp, locktype)
364 /* structs for some pioctls - these are (or should be)
372 struct sprefrequest_33 {
373 unsigned short offset;
374 unsigned short num_servers;
378 struct sprefrequest { /* new struct for 3.4 */
379 unsigned short offset;
380 unsigned short num_servers;
381 unsigned short flags;
386 unsigned short next_offset;
387 unsigned short num_servers;
388 struct spref servers[1]; /* we overrun this array intentionally...*/
392 unsigned short flags;
393 unsigned short num_servers;
394 struct spref servers[1]; /* we overrun this array intentionally...*/
396 /* struct for GAG pioctl
399 afs_uint32 showflags, logflags, logwritethruflag, spare[3];
400 unsigned char spare2[128];
404 #define logwritethruON 1
407 afs_int32 rx_initReceiveWindow, rx_maxReceiveWindow,
408 rx_initSendWindow, rx_maxSendWindow, rxi_nSendFrags,
409 rxi_nRecvFrags, rxi_OrphanFragSize;
410 afs_int32 rx_maxReceiveSize, rx_MyMaxSendSize;
411 afs_uint32 spare[21];
414 /* struct for checkservers */
427 /* state bits for volume */
428 #define VRO 1 /* volume is readonly */
429 #define VRecheck 2 /* recheck volume info with server */
430 #define VBackup 4 /* is this a backup volume? */
431 #define VForeign 8 /* this is a non-afs volume */
432 #define VResort 16 /* server order was rearranged, sort when able */
433 #define VMoreReps 32 /* This volume has more replicas than we are */
434 /* keeping track of now -- check with VLDB */
436 enum repstate { not_busy, end_not_busy = 6, rd_busy, rdwr_busy, offline };
439 /* One structure per volume, describing where the volume is located
440 and where its mount points are. */
441 struct volume *next; /* Next volume in hash list. */
442 afs_int32 cell; /* the cell in which the volume resides */
443 afs_rwlock_t lock; /* the lock for this structure */
444 afs_int32 volume; /* This volume's ID number. */
445 char *name; /* This volume's name, or 0 if unknown */
446 struct server *serverHost[MAXHOSTS]; /* servers serving this volume */
447 enum repstate status[MAXHOSTS]; /* busy, offline, etc */
448 struct VenusFid dotdot; /* dir to access as .. */
449 struct VenusFid mtpoint; /* The mount point for this volume. */
450 afs_int32 rootVnode, rootUnique; /* Volume's root fid */
453 afs_int32 rwVol; /* For r/o vols, original read/write volume. */
454 afs_int32 accessTime; /* last time we used it */
455 afs_int32 vtix; /* volume table index */
456 afs_int32 copyDate; /* copyDate field, for tracking vol releases */
457 afs_int32 expireTime; /* for per-volume callbacks... */
458 short refCount; /* reference count for allocation */
459 char states; /* here for alignment reasons */
462 #define afs_PutVolume(av, locktype) ((av)->refCount--)
464 /* format of an entry in volume info file */
466 afs_int32 cell; /* cell for this entry */
467 afs_int32 volume; /* volume */
468 afs_int32 next; /* has index */
469 struct VenusFid dotdot; /* .. value */
470 struct VenusFid mtpoint; /* mt point's fid */
471 afs_int32 rootVnode, rootUnique; /* Volume's root fid */
475 struct SimpleLocks *next;
477 afs_int32 boff, eoff;
479 #if defined(AFS_AIX32_ENV) || defined(AFS_SUN5_ENV) || defined(AFS_SGI_ENV)
484 /* vcache state bits */
485 #define CStatd 0x00000001 /* has this file ever been stat'd? */
486 #define CBackup 0x00000002 /* file is on a backup volume */
487 #define CRO 0x00000004 /* is it on a read-only volume */
488 #define CMValid 0x00000008 /* is the mount point info valid? */
489 #define CCore 0x00000010 /* storing a core file, needed since we don't get an open */
490 #define CDirty 0x00000020 /* file has been modified since first open (... O_RDWR) */
491 #define CSafeStore 0x00000040 /* close must wait for store to finish (should be in fd) */
492 #define CMAPPED 0x00000080 /* Mapped files; primarily used by SunOS 4.0.x */
493 #define CNSHARE 0x00000100 /* support O_NSHARE semantics */
494 #define CLied 0x00000200
495 #define CTruth 0x00000400
497 #define CWired 0x00000800 /* OSF hack only */
499 #define CWRITE_IGN 0x00000800 /* Next OS hack only */
501 #define CUnique 0x00001000 /* vc's uniquifier - latest unifiquier for fid */
502 #define CForeign 0x00002000 /* this is a non-afs vcache */
503 #define CHasPages 0x00004000
504 #define CUnlinked 0x00010000
505 #define CBulkStat 0x00020000 /* loaded by a bulk stat, and not ref'd since */
506 #define CUnlinkedDel 0x00040000
507 #define CVFlushed 0x00080000
508 #define CCore1 0x00100000 /* osf1 core file; not same as CCore above */
509 #define CWritingUFS 0x00200000 /* to detect vm deadlock - used by sgi */
510 #define CCreating 0x00400000 /* avoid needless store after open truncate */
511 #define CPageHog 0x00800000 /* AIX - dumping large cores is a page hog. */
512 #define CDCLock 0x02000000 /* Vnode lock held over call to GetDownD */
513 #define CBulkFetching 0x04000000 /* stats are being fetched by bulk stat */
514 #define CExtendedFile 0x08000000 /* extended file via ftruncate call. */
516 /* vcache vstate bits */
517 #define VRevokeWait 0x1
518 #define VPageCleaning 0x2 /* Solaris - Cache Trunc Daemon sez keep out */
521 #if !defined(AFS_FBSD_ENV)
522 #define vrefCount v.v_count
524 #define vrefCount v.v_usecount
525 #endif /* AFS_FBSD_ENV */
527 #ifdef AFS_LINUX24_ENV
528 #define VREFCOUNT(v) atomic_read(&((vnode_t *) v)->v_count)
529 #define VREFCOUNT_SET(v, c) atomic_set(&((vnode_t *) v)->v_count, c)
530 #define VREFCOUNT_DEC(v) atomic_dec(&((vnode_t *) v)->v_count)
531 #define VREFCOUNT_INC(v) atomic_inc(&((vnode_t *) v)->v_count)
533 #define VREFCOUNT(v) ((v)->vrefCount)
534 #define VREFCOUNT_SET(v, c) (v)->vrefCount = c;
535 #define VREFCOUNT_DEC(v) (v)->vrefCount--;
536 #define VREFCOUNT_INC(v) (v)->vrefCount++;
539 #define AFS_MAXDV 0x7fffffff /* largest dataversion number */
540 #define AFS_NOTRUNC 0x7fffffff /* largest dataversion number */
542 extern afs_int32 vmPageHog; /* counter for # of vnodes which are page hogs. */
545 * Fast map from vcache to dcache
552 afs_uint32 minLoc; /* smallest offset into dc. */
553 afs_uint32 len; /* largest offset into dc. */
556 extern afs_uint32 afs_stampValue; /* stamp for pair's usage */
557 #define MakeStamp() (++afs_stampValue)
559 /* INVARIANTs: (vlruq.next != NULL) == (vlruq.prev != NULL)
560 * nextfree => !vlruq.next && ! vlruq.prev
561 * !(avc->nextfree) && !avc->vlruq.next => (FreeVCList == avc->nextfree)
564 struct vnode v; /* Has reference count in v.v_count */
565 struct afs_q vlruq; /* lru q next and prev */
566 struct vcache *nextfree; /* next on free list (if free) */
567 struct vcache *hnext; /* Hash next */
571 afs_hyper_t DataVersion;
575 ushort Mode; /* XXXX Should be afs_int32 XXXX */
577 /* vnode type is in v.v_type */
579 afs_rwlock_t lock; /* The lock on the vcache contents. */
580 #if defined(AFS_SUN5_ENV)
581 /* Lock used to protect the activeV, multipage, and vstates fields.
582 * Do not try to get the vcache lock when the vlock is held */
584 #endif /* defined(AFS_SUN5_ENV) */
585 #if defined(AFS_SUN_ENV) || defined(AFS_ALPHA_ENV) || defined(AFS_DARWIN_ENV)
586 #if defined(AFS_SUN5_ENV)
590 afs_bozoLock_t pvnLock; /* see locks.x */
599 int ownslock; /* pid of owner of excl lock, else 0 - defect 3083 */
601 #ifdef AFS_DARWIN_ENV
602 struct lock__bsd__ rwlock;
604 afs_int32 parentVnode; /* Parent dir, if a file. */
605 afs_int32 parentUnique;
606 struct VenusFid *mvid; /* Either parent dir (if root) or root (if mt pt) */
607 char *linkData; /* Link data if a symlink. */
608 afs_hyper_t flushDV; /* data version last flushed from text */
609 afs_hyper_t mapDV; /* data version last flushed from map */
610 afs_uint32 truncPos; /* truncate file to this position at next store */
611 struct server *callback; /* The callback host, if any */
612 afs_uint32 cbExpires; /* time the callback expires */
613 struct afs_q callsort; /* queue in expiry order, sort of */
614 struct axscache *Access; /* a list of cached access bits */
615 afs_int32 anyAccess; /* System:AnyUser's access to this. */
616 afs_int32 last_looker; /* pag/uid from last lookup here */
617 #if defined(AFS_SUN5_ENV)
619 #endif /* defined(AFS_SUN5_ENV) */
620 struct SimpleLocks *slocks;
621 short opens; /* The numbers of opens, read or write, on this file. */
622 short execsOrWriters; /* The number of execs (if < 0) or writers (if > 0) of
624 short flockCount; /* count of flock readers, or -1 if writer */
625 char mvstat; /* 0->normal, 1->mt pt, 2->root. */
626 afs_uint32 states; /* state bits */
627 #if defined(AFS_SUN5_ENV)
628 afs_uint32 vstates; /* vstate bits */
629 #endif /* defined(AFS_SUN5_ENV) */
631 afs_uint32 symhintstamp;
633 struct vcache *symhint;
634 struct dcache *dchint;
636 #ifdef AFS_LINUX22_ENV
637 u_short flushcnt; /* Number of flushes which haven't released yet. */
638 u_short mapcnt; /* Number of mappings of this file. */
640 #if defined(AFS_SGI_ENV)
641 daddr_t lastr; /* for read-ahead */
643 uint64_t vc_rwlockid; /* kthread owning rwlock */
645 short vc_rwlockid; /* pid of process owning rwlock */
647 short vc_locktrips; /* # of rwlock reacquisitions */
648 sema_t vc_rwlock; /* vop_rwlock for afs */
649 pgno_t mapcnt; /* # of pages mapped */
650 struct cred *cred; /* last writer's cred */
652 struct bhv_desc vc_bhv_desc; /* vnode's behavior data. */
654 #endif /* AFS_SGI_ENV */
655 afs_int32 vc_error; /* stash write error for this vnode. */
656 int xlatordv; /* Used by nfs xlator */
657 struct AFS_UCRED *uncred;
658 int asynchrony; /* num kbytes to store behind */
660 short multiPage; /* count of multi-page getpages in progress */
664 #define afs_symhint_inval(avc)
667 #define DONT_CHECK_MODE_BITS 0
668 #define CHECK_MODE_BITS 1
669 #define CMB_ALLOW_EXEC_AS_READ 2 /* For the NFS xlator */
671 #if defined(AFS_SGI_ENV)
672 #define AVCRWLOCK(avc) (valusema(&(avc)->vc_rwlock) <= 0)
674 /* SGI vnode rwlock macros and flags. */
675 #ifndef AFS_SGI62_ENV
676 /* The following are defined here. SGI 6.2 declares them in vnode.h */
677 #define VRWLOCK_READ 0
678 #define VRWLOCK_WRITE 1
679 #define VRWLOCK_WRITE_DIRECT 2
684 #define AFS_RWLOCK_T vrwlock_t
686 #define AFS_RWLOCK_T int
687 #endif /* AFS_SGI62_ENV */
689 #include <ksys/behavior.h>
690 #define AFS_RWLOCK(V,F) \
691 afs_rwlock(&(((struct vcache *)(V))->vc_bhv_desc), (F));
692 #define AFS_RWUNLOCK(V,F) \
693 afs_rwunlock(&(((struct vcache *)(V))->vc_bhv_desc), (F));
696 #define AFS_RWLOCK(V,F) afs_rwlock((vnode_t *)(V), (F) )
697 #define AFS_RWUNLOCK(V,F) afs_rwunlock((vnode_t *)(V), (F) )
699 #else /* AFS_SGI53_ENV */
700 #define AFS_RWLOCK(V,F) afs_rwlock((V))
701 #define AFS_RWUNLOCK(V,F) afs_rwunlock((V))
702 #endif /* AFS_SGI53_ENV */
703 #endif /* AFS_SGI_ENV */
707 afs_hyper_t DataVersion;
709 afs_int32 parentVnode;
710 afs_int32 parentUnique;
714 afs_int32 randomUid[CPSIZE];
715 afs_int32 callback; /* Now a pointer to 'server' struct */
717 afs_int32 randomAccess[CPSIZE];
720 short execsOrWriters;
731 /* CM inititialization parameters. What CM actually used after calculations
732 * based on passed in arguments.
734 #define CMI_VERSION 1 /* increment when adding new fields. */
735 struct cm_initparams {
740 int cmi_nVolumeCaches;
741 int cmi_firstChunkSize;
742 int cmi_otherChunkSize;
743 int cmi_cacheSize; /* The original cache size, in 1K blocks. */
744 unsigned cmi_setTime:1;
745 unsigned cmi_memCache:1;
746 int spare[16-9]; /* size of struct is 16 * 4 = 64 bytes */
750 /*----------------------------------------------------------------------
751 * AFS Data cache definitions
753 * Each entry describes a Unix file on the local disk that is
754 * is serving as a cached copy of all or part of a Vice file.
755 * Entries live in circular queues for each hash table slot
757 * Which queue is this thing in? Good question.
758 * A struct dcache entry is in the freeDSlot queue when not associated with a cache slot (file).
759 * Otherwise, it is in the DLRU queue. The freeDSlot queue uses the lruq.next field as
760 * its "next" pointer.
762 * Cache entries in the DLRU queue are either associated with vice files, in which case
763 * they are hashed by afs_dvnextTbl and afs_dcnextTbl pointers, or they are in the freeDCList
764 * and are not associated with any vice file. This last list uses the afs_dvnextTbl pointer for
765 * its "next" pointer.
766 *----------------------------------------------------------------------*/
768 #define NULLIDX (-1) /* null index definition */
769 /* struct dcache states bits */
770 #define DWriting 8 /* file being written (used for cache validation) */
773 #define DFNextStarted 1 /* next chunk has been prefetched already */
774 #define DFEntryMod 2 /* has entry itself been modified? */
775 #define DFFetching 4 /* file is currently being fetched */
776 #define DFWaiting 8 /* someone waiting for file */
777 #define DFFetchReq 0x10 /* someone is waiting for DFFetching to go on */
780 /* flags in afs_indexFlags array */
781 #define IFEverUsed 1 /* index entry has >= 1 byte of data */
782 #define IFFree 2 /* index entry in freeDCList */
783 #define IFDataMod 4 /* file needs to be written out */
784 #define IFFlag 8 /* utility flag */
785 #define IFDirtyPages 16
786 #define IFAnyPages 32
787 #define IFDiscarded 64 /* index entry in discardDCList */
790 char *in; /* input buffer */
791 char *out; /* output buffer */
792 short in_size; /* Size of input buffer <= 2K */
793 short out_size; /* Maximum size of output buffer, <= 2K */
797 * This version of afs_ioctl is required to pass in 32 bit user space
798 * pointers into a 64 bit kernel.
809 /* CacheItems file has a header of type struct afs_fheader
810 * (keep aligned properly). Since we already have sgi_62 clients running
811 * with a 32 bit inode, a change is required to the header so that
812 * they can distinguish the old 32 bit inode CacheItems file and zap it
813 * instead of using it.
816 #define AFS_FHMAGIC 0x7635abaf /* uses version number */
818 #if defined(AFS_SUN57_64BIT_ENV)
819 #define AFS_CI_VERSION 3
821 #define AFS_CI_VERSION 2
824 afs_int32 firstCSize;
825 afs_int32 otherCSize;
828 /* kept on disk and in dcache entries */
830 struct VenusFid fid; /* Fid for this file */
831 afs_int32 modTime; /* last time this entry was modified */
832 afs_hyper_t versionNo; /* Associated data version number */
833 afs_int32 chunk; /* Relative chunk number */
834 #if defined(AFS_SGI61_ENV) || defined(AFS_SUN57_64BIT_ENV)
835 /* Using ino64_t here so that user level debugging programs compile
836 * the size correctly.
838 ino64_t inode; /* Unix inode for this chunk */
840 #if defined(AFS_LINUX_64BIT_KERNEL)
841 long inode; /* Unix inode for this chunk */
843 afs_int32 inode; /* Unix inode for this chunk */
846 afs_int32 chunkBytes; /* Num bytes in this chunk */
847 char states; /* Has this chunk been modified? */
850 /* magic numbers to specify the cache type */
852 #define AFS_FCACHE_TYPE_UFS 0x0
853 #define AFS_FCACHE_TYPE_MEM 0x1
854 #define AFS_FCACHE_TYPE_NFS 0x2
855 #define AFS_FCACHE_TYPE_EPI 0x3
859 struct afs_q lruq; /* Free queue for in-memory images */
860 afs_rwlock_t lock; /* XXX */
861 short refCount; /* Associated reference count. */
862 afs_int32 index; /* The index in the CacheInfo file*/
863 short flags; /* more flags bits */
864 afs_int32 validPos; /* number of valid bytes during fetch */
865 struct fcache f; /* disk image */
866 afs_int32 stamp; /* used with vtodc struct for hints */
868 /* this is obsolete and should be removed */
871 /* macro to mark a dcache entry as bad */
874 (x)->f.fid.Fid.Unique = 0; \
875 afs_indexUnique[(x)->index] = 0; \
876 (x)->flags |= DFEntryMod; \
880 * Convenient release macro for use when afs_PutDCache would cause
881 * deadlock on afs_xdcache lock
883 #define lockedPutDCache(ad) ((ad)->refCount--)
885 /* FakeOpen and Fake Close used to be real subroutines. They're only used in
886 * sun_subr and afs_vnodeops, and they're very frequently called, so I made
887 * them into macros. They do:
888 * FakeOpen: fake the file being open for writing. avc->lock must be held
889 * in write mode. Having the file open for writing is like having a DFS
890 * write-token: you're known to have the best version of the data around,
891 * and so the CM won't let it be overwritten by random server info.
892 * FakeClose: undo the effects of FakeOpen, noting that we want to ensure
893 * that a real close eventually gets done. We use CCore to achieve this if
894 * we would end up closing the file. avc->lock must be held in write mode */
896 #ifdef AFS_AIX_IAUTH_ENV
897 #define CRKEEP(V, C) (V)->linkData = (char*)crdup((C))
899 #define CRKEEP(V, C) crhold((C)); (V)->linkData = (char*)(C)
902 #define afs_FakeOpen(avc) { avc->opens++; avc->execsOrWriters++; }
903 #define afs_FakeClose(avc, acred) \
904 { if (avc->execsOrWriters == 1) { \
905 /* we're the last writer, just use CCore flag */ \
906 avc->states |= CCore; /* causes close to be called later */ \
908 /* The cred and vnode holds will be released in afs_FlushActiveVcaches */ \
909 VN_HOLD((struct vnode *)avc); /* So it won't disappear */ \
910 CRKEEP(avc, acred); /* Should use a better place for the creds */ \
913 /* we're not the last writer, let the last one do the store-back for us */ \
915 avc->execsOrWriters--; \
919 #define AFS_ZEROS 64 /* zero buffer */
921 /*#define afs_DirtyPages(avc) (((avc)->states & CDirty) || osi_VMDirty_p((avc)))*/
922 #define afs_DirtyPages(avc) ((avc)->states & CDirty)
924 /* The PFlush algorithm makes use of the fact that Fid.Unique is not used in
925 below hash algorithms. Change it if need be so that flushing algorithm
926 doesn't move things from one hash chain to another
928 /* extern int afs_dhashsize; */
929 #define DCHash(v, c) ((((v)->Fid.Vnode + (v)->Fid.Volume + (c))) & (afs_dhashsize-1))
930 /*Vnode, Chunk -> Hash table index*/
931 #define DVHash(v) ((((v)->Fid.Vnode + (v)->Fid.Volume )) & (afs_dhashsize-1))
932 /*Vnode -> Other hash table index*/
933 /* don't hash on the cell, our callback-breaking code sometimes fails to compute
934 the cell correctly, and only scans one hash bucket */
935 #define VCHash(fid) (((fid)->Fid.Volume + (fid)->Fid.Vnode) & (VCSIZE-1))
937 extern struct dcache **afs_indexTable; /*Pointers to in-memory dcache entries*/
938 extern afs_int32 *afs_indexUnique; /*dcache entry Fid.Unique */
939 extern afs_int32 *afs_dvnextTbl; /*Dcache hash table links */
940 extern afs_int32 *afs_dcnextTbl; /*Dcache hash table links */
941 extern afs_int32 afs_cacheFiles; /*Size of afs_indexTable*/
942 extern afs_int32 afs_cacheBlocks; /*1K blocks in cache*/
943 extern afs_int32 afs_cacheStats; /*Stat entries in cache*/
944 extern struct vcache *afs_vhashT[VCSIZE]; /*Stat cache hash table*/
945 extern afs_int32 afs_initState; /*Initialization state*/
946 extern afs_int32 afs_termState; /* Termination state */
947 extern struct VenusFid afs_rootFid; /*Root for whole file system*/
948 extern afs_int32 afs_allCBs; /* Count of callbacks*/
949 extern afs_int32 afs_oddCBs; /* Count of odd callbacks*/
950 extern afs_int32 afs_evenCBs; /* Count of even callbacks*/
951 extern afs_int32 afs_allZaps; /* Count of fid deletes */
952 extern afs_int32 afs_oddZaps; /* Count of odd fid deletes */
953 extern afs_int32 afs_evenZaps; /* Count of even fid deletes */
954 extern struct brequest afs_brs[NBRS]; /* request structures */
956 #define UHash(auid) ((auid) & (NUSERS-1))
957 #define VHash(avol) ((avol)&(NVOLS-1))
958 #define SHash(aserv) ((ntohl(aserv)) & (NSERVERS-1))
959 #define FVHash(acell,avol) (((avol)+(acell)) & (NFENTRIES-1))
961 extern struct cell *afs_GetCell();
962 extern struct cell *afs_GetCellByName();
963 extern struct cell *afs_GetCellByIndex();
964 extern struct unixuser *afs_GetUser();
965 extern struct volume *afs_GetVolume();
966 extern struct volume *afs_GetVolumeByName();
967 extern struct conn *afs_Conn();
968 extern struct conn *afs_ConnByHost();
969 extern struct conn *afs_ConnByMHosts();
970 extern afs_int32 afs_NewCell();
971 extern struct dcache *afs_GetDCache();
972 extern struct dcache *afs_FindDCache();
973 extern struct dcache *afs_NewDCache();
974 extern struct dcache *afs_GetDSlot();
975 extern struct vcache *afs_GetVCache();
976 extern struct brequest *afs_BQueue();
979 extern int afs_CacheInit();
980 extern void afs_StoreWarn();
981 extern void afs_AdjustSize();
982 extern void afs_ComputeCacheParms();
983 extern void afs_FlushDCache();
984 extern void afs_FlushActiveVcaches();
985 extern void afs_StuffVcache();
986 extern void afs_PutVCache();
987 extern void afs_TryToSmush();
988 extern void afs_ProcessFS();
989 extern void afs_WriteThroughDSlots();
990 extern void shutdown_cache();
992 extern void afs_shutdown();
994 extern void shutdown_osifile();
997 extern int afs_IsDynrootFid();
998 extern void afs_GetDynrootFid();
999 extern int afs_IsDynroot();
1000 extern void afs_RefreshDynroot();
1001 extern void afs_GetDynroot();
1002 extern void afs_PutDynroot();
1003 extern int afs_DynrootNewVnode();
1004 extern int afs_SetDynrootEnable();
1005 extern int afs_GetDynrootEnable();
1008 /* Performance hack - we could replace VerifyVCache2 with the appropriate
1009 * GetVCache incantation, and could eliminate even this code from afs_UFSRead
1010 * by making intentionally invalidating quick.stamp in the various callbacks
1011 * expiration/breaking code */
1012 #ifdef AFS_LINUX20_ENV
1013 #define afs_VerifyVCache(avc, areq) \
1014 (((avc)->states & CStatd) ? (vcache2inode(avc), 0) : \
1015 afs_VerifyVCache2((avc),areq))
1017 #ifdef AFS_DARWIN_ENV
1018 #define afs_VerifyVCache(avc, areq) \
1019 (((avc)->states & CStatd) ? (osi_VM_Setup(avc), 0) : \
1020 afs_VerifyVCache2((avc),areq))
1022 #define afs_VerifyVCache(avc, areq) \
1023 (((avc)->states & CStatd) ? 0 : afs_VerifyVCache2((avc),areq))
1027 #define DO_STATS 1 /* bits used by FindVCache */
1030 /* values for flag param of afs_CheckVolumeNames */
1031 #define AFS_VOLCHECK_EXPIRED 0x1 /* volumes whose callbacks have expired */
1032 #define AFS_VOLCHECK_BUSY 0x2 /* volumes which were marked busy */
1033 #define AFS_VOLCHECK_MTPTS 0x4 /* mount point invalidation also */
1034 #define AFS_VOLCHECK_FORCE 0x8 /* do all forcibly */
1038 #define AFS_FSPORT ((unsigned short) htons(7000))
1039 #define AFS_VLPORT ((unsigned short) htons(7003))
1041 #define afs_read(avc, uio, acred, albn, abpp, nolock) \
1042 (*(afs_cacheType->vread))(avc, uio, acred, albn, abpp, nolock)
1043 #define afs_write(avc, uio, aio, acred, nolock) \
1044 (*(afs_cacheType->vwrite))(avc, uio, aio, acred, nolock)
1046 #define afs_rdwr(avc, uio, rw, io, cred) \
1047 (((rw) == UIO_WRITE) ? afs_write(avc, uio, io, cred, 0) : afs_read(avc, uio, cred, 0, 0, 0))
1048 #define afs_nlrdwr(avc, uio, rw, io, cred) \
1049 (((rw) == UIO_WRITE) ? afs_write(avc, uio, io, cred, 1) : afs_read(avc, uio, cred, 0, 0, 1))
1051 extern afs_int32 afs_blocksUsed, afs_blocksDiscarded;
1052 extern afs_int32 afs_discardDCCount, afs_freeDCCount;
1053 extern afs_int32 afs_bulkStatsDone, afs_bulkStatsLost;
1054 extern int afs_TruncateDaemonRunning;
1055 extern int afs_CacheTooFull;
1056 /* Cache size truncation uses the following low and high water marks:
1057 * If the cache is more than 95% full (CM_DCACHECOUNTFREEPCT), the cache
1058 * truncation daemon is awakened and will free up space until the cache is 85%
1059 * (CM_DCACHESPACEFREEPCT - CM_DCACHEEXTRAPCT) full.
1060 * afs_UFSWrite and afs_GetDCache (when it needs to fetch data) will wait on
1061 * afs_WaitForCacheDrain if the cache is 98% (CM_WAITFORDRAINPCT) full.
1062 * afs_GetDownD wakes those processes once the cache is 95% full
1063 * (CM_CACHESIZEDRAINEDPCT).
1065 extern void afs_MaybeWakeupTruncateDaemon();
1066 extern void afs_CacheTruncateDaemon();
1067 extern int afs_WaitForCacheDrain;
1068 #define CM_MAXDISCARDEDCHUNKS 16 /* # of chunks */
1069 #define CM_DCACHECOUNTFREEPCT 95 /* max pct of chunks in use */
1070 #define CM_DCACHESPACEFREEPCT 90 /* max pct of space in use */
1071 #define CM_DCACHEEXTRAPCT 5 /* extra to get when freeing */
1072 #define CM_CACHESIZEDRAINEDPCT 95 /* wakeup processes when down to here.*/
1073 #define CM_WAITFORDRAINPCT 98 /* sleep if cache is this full. */
1075 #define afs_CacheIsTooFull() \
1076 (afs_blocksUsed - afs_blocksDiscarded > \
1077 (CM_DCACHECOUNTFREEPCT*afs_cacheBlocks)/100 || \
1078 afs_freeDCCount - afs_discardDCCount < \
1079 ((100-CM_DCACHECOUNTFREEPCT)*afs_cacheFiles)/100)
1081 /* Handy max length of a numeric string. */
1082 #define CVBS 12 /* max afs_int32 is 2^32 ~ 4*10^9, +1 for NULL, +luck */
1084 extern int afs_norefpanic;
1085 #define refpanic(foo) if (afs_norefpanic) \
1086 { printf( foo ); afs_norefpanic++;} else osi_Panic( foo )
1091 ** these are defined in the AIX source code sys/fs_locks.h but are not
1092 ** defined anywhere in the /usr/include directory
1094 #if defined(AFS_AIX41_ENV)
1095 #define VN_LOCK(vp) simple_lock(&(vp)->v_lock)
1096 #define VN_UNLOCK(vp) simple_unlock(&(vp)->v_lock)
1099 /* get a file's serial number from a vnode */
1100 #ifndef afs_vnodeToInumber
1101 #if defined(AFS_SGI62_ENV) || defined(AFS_HAVE_VXFS) || defined(AFS_DARWIN_ENV)
1102 #define afs_vnodeToInumber(V) VnodeToIno(V)
1104 #ifdef AFS_DECOSF_ENV
1105 #define afs_vnodeToInumber(V) osi_vnodeToInumber(V)
1107 #define afs_vnodeToInumber(V) (VTOI(V)->i_number)
1108 #endif /* AFS_DECOSF_ENV */
1109 #endif /* AFS_SGI62_ENV */
1112 /* get a file's device number from a vnode */
1113 #ifndef afs_vnodeToDev
1114 #if defined(AFS_SGI62_ENV) || defined(AFS_HAVE_VXFS) || defined(AFS_DARWIN_ENV)
1115 #define afs_vnodeToDev(V) VnodeToDev(V)
1117 #ifdef AFS_DECOSF_ENV
1118 #define afs_vnodeToDev(V) osi_vnodeToDev(V)
1120 #define afs_vnodeToDev(V) (VTOI(V)->i_dev)
1121 #endif /* AFS_DECOSF_ENV */
1122 #endif /* AFS_SGI62_ENV */
1125 #endif /* _AFS_H_ */