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>
17 #include <afs/afs_consts.h>
19 #include "afs/afs_args.h"
20 #include "afs/afs_consts.h"
24 * afs_fsfragsize cannot be less than 1023, or some cache-tracking
25 * calculations will be incorrect (since we track cache usage in kb).
26 * Some filesystems have fundamental blocksizes less than 1k, and
27 * normally we would tune afs_fsfragsize to be fragsize-1, but we must
28 * make sure to check that afs_fsfragsize does not go below this value.
30 #define AFS_MIN_FRAGSIZE 1023
32 /* Upper bound on number of iovecs out uio routines will deal with. */
33 #define AFS_MAXIOVCNT 16
36 extern int afs_shuttingdown;
39 * Macros to uniquely identify the AFS vfs struct
41 #define AFS_VFSMAGIC 0x1234
42 #if defined(AFS_SUN5_ENV) || defined(AFS_HPUX90_ENV) || defined(AFS_LINUX20_ENV)
43 #define AFS_VFSFSID 99
45 #if defined(AFS_SGI_ENV)
46 #define AFS_VFSFSID afs_fstype
48 #define AFS_VFSFSID AFS_MOUNT_AFS
51 /* use this value for reporting total space, free space, etc.
52 * fake a high number to satisfy programs that use the statfs call to make sure
53 * that there's enough space in the device partition before storing something
54 * there. keep it just under 2^31, to try and be safe about not breaking
55 * stuff that looks at the amount of free space. */
56 #define AFS_VFS_FAKEFREE (2147483647)
58 /* Moved from VNOPS/afs_vnop_flocks so can be used in prototypes */
59 #if defined(AFS_HPUX102_ENV)
60 #define AFS_FLOCK k_flock
62 #if defined(AFS_SUN5_ENV) || (defined(AFS_LINUX24_ENV) && !(defined(AFS_LINUX26_ENV) && defined(AFS_LINUX_64BIT_KERNEL)))
63 #define AFS_FLOCK flock64
65 #define AFS_FLOCK flock
66 #endif /* AFS_SUN65_ENV */
67 #endif /* AFS_HPUX102_ENV */
69 /* The following are various levels of afs debugging */
70 #define AFSDEB_GENERAL 1 /* Standard debugging */
71 #define AFSDEB_NETWORK 2 /* low level afs networking */
72 #define AFSDEB_RX 4 /* RX debugging */
73 #define AFSDEB_VNLAYER 8 /* interface layer to AFS (aixops, gfsops, etc) */
75 /* generic undefined vice id */
78 /* The basic defines for the Andrew file system
79 better keep things powers of two so "& (foo-1)" hack works for masking bits */
80 #define NBRS 15 /* max number of queued daemon requests */
81 #define NUSERS 16 /* hash table size for unixuser table */
82 #define NSERVERS 16 /* hash table size for server table */
83 #define NVOLS 64 /* hash table size for volume table */
84 #define NFENTRIES 256 /* hash table size for disk volume table */
85 #define VCSIZE 1024 /* stat cache hash table size */
86 #define CBRSIZE 512 /* call back returns hash table size */
87 #define PIGGYSIZE 1350 /* max piggyback size */
88 #define MAXVOLS 128 /* max vols we can store */
89 #define MAXSYSNAME 128 /* max sysname (i.e. @sys) size */
90 #define MAXNUMSYSNAMES 32 /* max that current constants allow */
91 #define NOTOKTIMEOUT (2*3600) /* time after which to timeout conns sans tokens */
92 #define NOPAG 0xffffffff
96 #define AFS_MAXCBRSCALL 32 /* max to return in a given call (must be <= AFSCBMAX) */
98 #define AFS_NRXPACKETS 80
99 #define AFS_RXDEADTIME 50
100 #define AFS_HARDDEADTIME 120
101 #define AFS_IDLEDEADTIME 1200
102 #define AFS_IDLEDEADTIME_REP 180 /* more than fs's cb dead time */
103 #define AFS_BLKBITS 12
104 #define AFS_BLKSIZE (1 << AFS_BLKBITS)
106 extern afs_int32 afs_rx_deadtime;
107 extern afs_int32 afs_rx_harddead;
108 extern afs_int32 afs_rx_idledead;
109 extern afs_int32 afs_rx_idledead_rep;
111 struct sysname_info {
114 signed char index, allocked;
117 /* flags to use with AFSOP_CACHEINIT */
118 #define AFSCALL_INIT_MEMCACHE 0x1 /* use a memory-based cache */
120 /* below here used only for kernel procedures */
122 /* Store synchrony flags - SYNC means that data should be forced to server's
123 * disk immediately upon completion. */
126 #define AFS_VMSYNC_INVAL 2 /* sync and invalidate pages */
127 #define AFS_LASTSTORE 4
128 #define AFS_VMSYNC 8 /* sync pages but do not invalidate */
130 /* background request structure */
133 #define BOP_NOOP 0 /* leave 0 unused */
134 #define BOP_FETCH 1 /* parm1 is chunk to get */
135 #define BOP_STORE 2 /* parm1 is chunk to store */
136 #define BOP_PATH 3 /* parm1 is path, parm2 is chunk to fetch */
137 #define BOP_FETCH_NOCACHE 4 /* parms are: vnode ptr, offset, segment ptr, addr, cred ptr */
138 #ifdef AFS_DARWIN_ENV
139 #define BOP_MOVE 5 /* ptr1 afs_uspc_param ptr2 sname ptr3 dname */
141 #define BOP_PARTIAL_STORE 6 /* parm1 is chunk to store */
143 #define B_DONTWAIT 1 /* On failure return; don't wait */
145 /* protocol is: refCount is incremented by user to take block out of free pool.
146 Next, BSTARTED is set when daemon finds request. This prevents
147 other daemons from picking up the same request. Finally, when
148 request is done, refCount is zeroed. BDONE and BWAIT are used by
149 dudes waiting for operation to proceed to a certain point before returning.
151 #define BSTARTED 1 /* request picked up by a daemon */
152 #define BUVALID 2 /* code is valid (store) */
153 #define BUWAIT 4 /* someone is waiting for BUVALID */
155 struct vcache *vc; /* vnode to use, with vrefcount bumped */
156 afs_ucred_t *cred; /* credentials to use for operation */
157 afs_size_t size_parm[BPARMS]; /* random parameters */
158 void *ptr_parm[BPARMS]; /* pointer parameters */
159 afs_int32 code_raw; /* return code from AFS routines */
160 afs_int32 code_checkcode; /* the afs_CheckCode-translated code */
161 short refCount; /* use counter for this structure */
162 char opcode; /* what to do (store, fetch, etc) */
163 char flags; /* free, etc */
164 afs_int32 ts; /* counter "timestamp" */
172 afs_int32 AuthHandle;
173 char HandShakeKey[8];
175 afs_int32 BeginTimestamp;
176 afs_int32 EndTimestamp;
180 afs_int32 Cell; /* better sun packing if at end of structure */
184 /* 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!) */
187 afs_int32 CellAndUnique;
190 /* The actual number of bytes in the SmallFid, not the sizeof struct. */
191 #define SIZEOF_SMALLFID 10
196 * Circular queues, implemented with pointers. Structures may contain as many
197 * queues as required, which may be located at any point within the structure,
198 * providing the QEntry macro is used to translate between a queue pointer, and
199 * the address of its containing structure
207 #define QZero(e) ((e)->prev = (e)->next = NULL)
208 #define QInit(q) ((q)->prev = (q)->next = (q))
209 #define QAdd(q,e) ((e)->next = (q)->next, (e)->prev = (q), \
210 (q)->next->prev = (e), (q)->next = (e))
211 #define QRemove(e) ((e)->next->prev = (e)->prev, (e)->prev->next = (e)->next, (e)->prev = NULL, (e)->next = NULL)
212 #define QNext(e) ((e)->next)
213 #define QPrev(e) ((e)->prev)
214 #define QEmpty(q) ((q)->prev == (q))
215 /* this one takes q1 and sticks it on the end of q2 - that is, the other end, not the end
216 * that things are added onto. q1 shouldn't be empty, it's silly */
217 #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))
219 /* Given a pointer to an afs_q within a structure, go back to the address of
220 * the parent structure
223 #define QEntry(queue, structure, member) \
224 ((structure *)((char *)(queue)-(char *)(&((structure *)NULL)->member)))
226 /* And implement operations for individual lists in terms of the above macro */
228 #define QTOV(e) QEntry(e, struct vcache, vlruq)
229 #define QTOC(e) QEntry(e, struct cell, lruq)
230 #define QTOVH(e) QEntry(e, struct vcache, vhashq)
233 * List of free slot numbers
235 struct afs_slotlist {
237 struct afs_slotlist *next;
241 afs_int32 uid; /* user id making the request */
242 afs_int32 busyCount; /* how many busies we've seen so far */
243 afs_int32 flags; /* things like O_SYNC, O_NONBLOCK go here */
244 char initd; /* if non-zero, Error fields meaningful */
245 char accessError; /* flags for overriding error return code */
246 char volumeError; /* encountered a missing or busy volume */
247 char networkError; /* encountered network problems */
248 char permWriteError; /* fileserver returns permenent error. */
249 char tokenError; /* a token error other than expired. */
250 char idleError; /* the server idled too long */
251 char skipserver[AFS_MAXHOSTS];
252 afs_int32 lasterror[AFS_MAXHOSTS];
257 /* structure linked off of a server to keep track of queued returned
258 * callbacks. Sent asynchronously when we run a little low on free dudes.
261 struct afs_cbr **pprev;
262 struct afs_cbr *next;
264 struct afs_cbr **hash_pprev;
265 struct afs_cbr *hash_next;
270 #ifdef AFS_LINUX22_ENV
271 /* On Linux, we have to be able to allocate the storage for this using
272 * kmalloc, as otherwise we may deadlock. So, it needs to be able to fit
275 # define AFS_NCBRS PAGE_SIZE/sizeof(struct afs_cbr)
277 # define AFS_NCBRS 1024 /* max # of call back return entries */
280 /* cellinfo file magic number */
281 #define AFS_CELLINFO_MAGIC 0xf32817cd
284 #define CNoSUID 0x02 /* disable suid bit for this cell */
285 #define CLinkedCell4 0x04 /* reserved for ADDCELL2 pioctl */
286 #define CNoAFSDB 0x08 /* never bother trying AFSDB */
287 #define CHasVolRef 0x10 /* volumes were referenced */
288 #define CLinkedCell 0x20 /* has a linked cell in lcellp */
289 #define CHush 0x40 /* don't display until referenced */
292 struct afs_q lruq; /* lru q next and prev */
293 char *cellName; /* char string name of cell */
294 afs_int32 cellIndex; /* sequence number */
295 afs_int32 cellNum; /* semi-permanent cell number */
296 struct server *cellHosts[AFS_MAXCELLHOSTS]; /* volume *location* hosts */
297 struct cell *lcellp; /* Associated linked cell */
298 u_short fsport; /* file server port */
299 u_short vlport; /* volume server port */
300 short states; /* state flags */
301 time_t timeout; /* data expire time, if non-zero */
302 struct cell_name *cnamep; /* pointer to our cell_name */
303 afs_rwlock_t lock; /* protects cell data */
304 unsigned char cellHandle[16]; /* deterministic handle for this cell */
308 struct cell_name *next;
315 struct cell_alias *next;
321 #define afs_PutCell(cellp, locktype)
323 /* the unixuser flag bit definitions */
324 #define UHasTokens 1 /* are the st and ct fields valid (ever set)? */
325 #define UTokensBad 2 /* are tokens bad? */
326 #define UPrimary 4 /* on iff primary identity */
327 #define UNeedsReset 8 /* needs afs_ResetAccessCache call done */
328 #define UPAGCounted 16 /* entry seen during PAG search (for stats) */
329 #define UNFSGetCreds 32 /* getting creds for NFS client */
330 /* A flag used by afs_GCPAGs to keep track of
331 * which entries in afs_users need to be deleted.
332 * The lifetime of its presence in the table is the
333 * lifetime of the afs_GCPAGs function.
335 #define TMP_UPAGNotReferenced 128
337 /* unixuser notify events */
338 #define UTokensObtained 1
339 #define UTokensDropped 2
341 /* values for afs_gcpags */
342 enum { AFS_GCPAGS_NOTCOMPILED = 0, AFS_GCPAGS_OK =
343 1, AFS_GCPAGS_USERDISABLED, AFS_GCPAGS_EPROC0, AFS_GCPAGS_EPROCN,
344 AFS_GCPAGS_EEQPID, AFS_GCPAGS_EINEXACT, AFS_GCPAGS_EPROCEND,
345 AFS_GCPAGS_EPROCWALK, AFS_GCPAGS_ECREDWALK, AFS_GCPAGS_EPIDCHECK,
346 AFS_GCPAGS_ENICECHECK
349 extern afs_int32 afs_gcpags;
350 extern afs_int32 afs_gcpags_procsize;
351 extern afs_int32 afs_bkvolpref;
352 extern char afs_cachebasedir[1024];
353 extern afs_int32 afs_numcachefiles;
354 extern afs_int32 afs_numfilesperdir;
359 struct ClearToken clearToken;
363 struct rxkadToken rxkad;
367 struct tokenJar *next;
369 union tokenUnion content;
373 struct unixuser *next; /* next hash pointer */
374 afs_int32 uid; /* search based on uid and cell */
376 afs_int32 viceId; /* Optional viced ID corresponding to current tokens */
377 short refCount; /* reference count for allocation */
378 char states; /* flag info */
379 afs_int32 tokenTime; /* last time tokens were set, used for timing out conn data */
380 struct tokenJar *tokens;
381 struct afs_exporter *exporter; /* more info about the exporter for the remote user */
382 void *cellinfo; /* pointer to cell info (PAG manager only) */
386 #define CVEC_LEN 3 /* per-user connection pool */
388 struct sa_conn_vector;
389 typedef struct sa_conn_vector * p_sa_conn_vector; /* forward decl */
391 #define CONN_REPLICATED 0x1
396 char forceConnectFS; /* Should we try again with these tokens? */
397 struct rx_connection *id; /* RPC connid */
398 struct sa_conn_vector *parent; /* the con_vector which contains us */
401 /* An sa_conn_vector replaces the erstwhile list of conn
402 structures maintained by the cache manager. The sa_conn_vector
403 contains a C array of connections which, if non-zero, represent
404 connections to AFS servers.
407 struct sa_conn_vector {
408 /* linked-list machinery */
409 struct sa_conn_vector *next;
411 /* AFS conn-identifying info */
412 struct unixuser *user; /* user validated with respect to. */
413 struct srvAddr *srvr; /* server associated with this conn */
414 short refCount; /* reference count for allocation */
415 unsigned short port; /* port associated with this connection */
418 /* next connection to return when all in cvec are fully utilized */
421 /* connections vector */
422 struct afs_conn cvec[CVEC_LEN];
427 /* Fid comparison routines */
428 #define FidCmp(a,b) ((a)->Fid.Unique != (b)->Fid.Unique \
429 || (a)->Fid.Vnode != (b)->Fid.Vnode \
430 || (a)->Fid.Volume != (b)->Fid.Volume \
431 || (a)->Cell != (b)->Cell)
433 #define FidMatches(afid,tvc) ((tvc)->f.fid.Fid.Vnode == (afid)->Fid.Vnode && \
434 (tvc)->f.fid.Fid.Volume == (afid)->Fid.Volume && \
435 (tvc)->f.fid.Cell == (afid)->Cell && \
436 ( (tvc)->f.fid.Fid.Unique == (afid)->Fid.Unique || \
437 (!(afid)->Fid.Unique && ((tvc)->f.states & CUnique))))
441 #define SRVADDR_ISDOWN 0x20 /* same as SRVR_ISDOWN */
442 #define SRVADDR_NOUSE 0x40 /* Don't use this srvAddr */
445 struct srvAddr *next_bkt; /* next item in hash bucket */
446 struct srvAddr *next_sa; /* another interface on same host */
447 struct server *server; /* back to parent */
448 struct sa_conn_vector *conns; /* All user connections to this server */
449 struct afs_conn *natping;
450 afs_int32 sa_ip; /* Host addr in network byte order */
451 u_short sa_iprank; /* indiv ip address priority */
452 u_short sa_portal; /* port addr in network byte order */
457 * Values used in the flags field of the server structure below.
459 * AFS_SERVER_FLAG_ACTIVATED Has the server ever had a user connection
460 * associated with it?
462 #define AFS_SERVER_FLAG_ACTIVATED 0x01
463 #define SNO_LHOSTS 0x04
464 #define SYES_LHOSTS 0x08
465 #define SVLSRV_UUID 0x10
466 #define SRVR_ISDOWN 0x20
467 #define SRVR_MULTIHOMED 0x40
468 #define SRVR_ISGONE 0x80
469 #define SNO_INLINEBULK 0x100
470 #define SNO_64BIT 0x200
471 #define SCAPS_KNOWN 0x400
473 #define SRV_CAPABILITIES(ts) \
474 { if ( !(ts->flags & SCAPS_KNOWN)) afs_GetCapabilities(ts); ts->capabilities; }
476 #define afs_serverSetNo64Bit(s) (((struct sa_conn_vector*)(s)->parent)->srvr->server->flags |= SNO_64BIT)
477 #define afs_serverHasNo64Bit(s) (((struct sa_conn_vector*)(s)->parent)->srvr->server->flags & SNO_64BIT)
483 afs_int32 addr_uniquifier;
487 struct srvAddr haddr;
490 #define sr_uuid _suid._srvUuid.suuid
491 #define sr_addr_uniquifier _suid._srvUuid.addr_uniquifier
492 #define sr_host _suid._srvId.haddr.ip
493 #define sr_portal _suid._srvId.haddr.portal
494 #define sr_rank _suid._srvId.haddr.ip_rank
495 #define sr_flags _suid._srvId.haddr.flags
496 #define sr_conns _suid._srvId.haddr.conns
497 struct server *next; /* Ptr to next server in hash chain */
498 struct cell *cell; /* Cell in which this host resides */
499 struct afs_cbr *cbrs; /* Return list of callbacks */
500 afs_int32 activationTime; /* Time when this record was first activated */
501 afs_int32 lastDowntimeStart; /* Time when last downtime incident began */
502 afs_int32 numDowntimeIncidents; /* # (completed) downtime incidents */
503 afs_int32 sumOfDowntimes; /* Total downtime experienced, in seconds */
504 struct srvAddr *addr;
505 afs_uint32 flags; /* Misc flags */
506 afs_int32 capabilities;
509 #define afs_PutServer(servp, locktype)
511 /* structs for some pioctls - these are (or should be)
519 struct sprefrequest_33 {
520 unsigned short offset;
521 unsigned short num_servers;
525 struct sprefrequest { /* new struct for 3.4 */
526 unsigned short offset;
527 unsigned short num_servers;
528 unsigned short flags;
533 unsigned short next_offset;
534 unsigned short num_servers;
535 struct spref servers[1]; /* we overrun this array intentionally... */
539 unsigned short flags;
540 unsigned short num_servers;
541 struct spref servers[1]; /* we overrun this array intentionally... */
543 /* struct for GAG pioctl
546 afs_uint32 showflags, logflags, logwritethruflag, spare[3];
547 unsigned char spare2[128];
551 #define logwritethruON 1
554 afs_int32 rx_initReceiveWindow, rx_maxReceiveWindow, rx_initSendWindow,
555 rx_maxSendWindow, rxi_nSendFrags, rxi_nRecvFrags, rxi_OrphanFragSize;
556 afs_int32 rx_maxReceiveSize, rx_MyMaxSendSize;
557 afs_uint32 spare[21];
560 /* struct for checkservers */
571 /* state bits for volume */
572 #define VRO 1 /* volume is readonly */
573 #define VRecheck 2 /* recheck volume info with server */
574 #define VBackup 4 /* is this a backup volume? */
575 #define VForeign 8 /* this is a non-afs volume */
576 #define VPartVisible 16 /* Volume's partition is visible on the client */
577 #define VHardMount 32 /* we are hard-mount waiting for the vol */
579 enum repstate { not_busy, end_not_busy = 6, rd_busy, rdwr_busy, offline };
582 /* One structure per volume, describing where the volume is located
583 * and where its mount points are. */
584 struct volume *next; /* Next volume in hash list. */
585 afs_int32 cell; /* the cell in which the volume resides */
586 afs_rwlock_t lock; /* the lock for this structure */
587 afs_int32 volume; /* This volume's ID number. */
588 char *name; /* This volume's name, or 0 if unknown */
589 struct server *serverHost[AFS_MAXHOSTS]; /* servers serving this volume */
590 enum repstate status[AFS_MAXHOSTS]; /* busy, offline, etc */
591 struct VenusFid dotdot; /* dir to access as .. */
592 struct VenusFid mtpoint; /* The mount point for this volume. */
593 afs_int32 rootVnode, rootUnique; /* Volume's root fid */
596 afs_int32 rwVol; /* For r/o vols, original read/write volume. */
597 afs_int32 accessTime; /* last time we used it */
598 afs_int32 vtix; /* volume table index */
599 afs_int32 copyDate; /* copyDate field, for tracking vol releases */
600 afs_int32 expireTime; /* for per-volume callbacks... */
601 short refCount; /* reference count for allocation */
602 char states; /* here for alignment reasons */
605 #define afs_PutVolume(av, locktype) ((av)->refCount--)
607 /* format of an entry in volume info file */
609 afs_int32 cell; /* cell for this entry */
610 afs_int32 volume; /* volume */
611 afs_int32 next; /* has index */
612 struct VenusFid dotdot; /* .. value */
613 struct VenusFid mtpoint; /* mt point's fid */
614 afs_int32 rootVnode, rootUnique; /* Volume's root fid */
618 struct SimpleLocks *next;
620 afs_int32 boff, eoff;
622 #if defined(AFS_AIX32_ENV) || defined(AFS_SUN5_ENV) || defined(AFS_SGI_ENV)
627 /* vcache state bits */
628 #define CStatd 0x00000001 /* has this file ever been stat'd? */
629 #define CBackup 0x00000002 /* file is on a backup volume */
630 #define CRO 0x00000004 /* is it on a read-only volume */
631 #define CMValid 0x00000008 /* is the mount point info valid? */
632 #define CCore 0x00000010 /* storing a core file, needed since we don't get an open */
633 #define CDirty 0x00000020 /* file has been modified since first open (... O_RDWR) */
634 #define CSafeStore 0x00000040 /* close must wait for store to finish (should be in fd) */
635 #define CMAPPED 0x00000080 /* Mapped files; primarily used by SunOS 4.0.x */
636 #define CNSHARE 0x00000100 /* support O_NSHARE semantics */
637 #define CLied 0x00000200
638 #define CTruth 0x00000400
640 #if defined(AFS_DARWIN80_ENV)
641 #define CDeadVnode 0x00000800
642 #elif defined(AFS_DARWIN_ENV)
643 #define CUBCinit 0x00000800
645 #define CWRITE_IGN 0x00000800 /* Next OS hack only */
648 #define CUnique 0x00001000 /* vc's uniquifier - latest unifiquier for fid */
649 #define CForeign 0x00002000 /* this is a non-afs vcache */
650 #define CReadDir 0x00004000 /* readdir in progress */
651 #define CUnlinked 0x00010000
652 #define CBulkStat 0x00020000 /* loaded by a bulk stat, and not ref'd since */
653 #define CUnlinkedDel 0x00040000
654 #define CVFlushed 0x00080000
655 #if defined(AFS_SGI_ENV)
656 #define CWritingUFS 0x00200000 /* to detect vm deadlock - used by sgi */
657 #elif defined(AFS_DARWIN80_ENV)
658 #define CEvent 0x00200000 /* to preclude deadlock when sending events */
660 #define CCreating 0x00400000 /* avoid needless store after open truncate */
661 #define CPageHog 0x00800000 /* AIX - dumping large cores is a page hog. */
662 #define CCorrupt 0x01000000 /* corrupt dir already reported */
663 #define CDCLock 0x02000000 /* Vnode lock held over call to GetDownD */
664 #define CBulkFetching 0x04000000 /* stats are being fetched by bulk stat */
665 #define CExtendedFile 0x08000000 /* extended file via ftruncate call. */
666 #define CVInit 0x10000000 /* being initialized */
667 #define CMetaDirty 0x20000000 /* vnode meta-data needs to be flushed */
668 #define CPartVisible 0x40000000 /* fileserver partition visible on client */
670 /* vcache vstate bits */
671 #define VRevokeWait 0x1
672 #define VPageCleaning 0x2 /* Solaris - Cache Trunc Daemon sez keep out */
674 /* Dirty disconnected vcache flags. */
675 #define VDisconSetTime 0x00000001 /* set time. */
676 #define VDisconSetMode 0x00000002 /* set mode. */
677 /* XXX: to be continued ? */
678 #define VDisconTrunc 0x00000020 /* truncate file. */
679 #define VDisconSetAttrMask 0x0000003F /* Masks for setattr ops. */
680 #define VDisconWriteClose 0x00000400 /* Write op on file close. */
681 #define VDisconWriteFlush 0x00000800 /* Write op on normal fsync/flush. */
682 #define VDisconWriteOsiFlush 0x00001000 /* Write op on osi flush. */
684 #define VDisconRemove 0x00002000 /* Remove vnop. */
685 #define VDisconCreate 0x00004000 /* Create vnop. */
686 #define VDisconCreated 0x00008000 /* A file that was created during
687 this resync operation */
688 #define VDisconRename 0x00010000 /* Rename vnop. */
689 #define VDisconRenameSameDir 0x00020000 /* Rename in same dir. */
691 /*... to be continued ... */
693 /* vcache (file) cachingStates bits */
694 #define FCSDesireBypass 0x1 /* This file should bypass the cache */
695 #define FCSBypass 0x2 /* This file is currently NOT being cached */
696 #define FCSManuallySet 0x4 /* The bypass flags were set, or reset, manually (via pioctl)
697 and should not be overridden by the file's name */
699 /* Flag values used by the Transition routines */
700 #define TRANSChangeDesiredBit 0x1 /* The Transition routine should set or
701 * reset the FCSDesireBypass bit */
702 #define TRANSVcacheIsLocked 0x2 /* The Transition routine does not need to
703 * lock vcache (it's already locked) */
704 #define TRANSSetManualBit 0x4 /* The Transition routine should set FCSManuallySet so that
705 * filename checking does not override pioctl requests */
708 #if defined(AFS_XBSD_ENV) || defined(AFS_DARWIN_ENV)
709 #define vrefCount v->v_usecount
711 #define vrefCount v.v_count
712 #endif /* AFS_XBSD_ENV */
714 #if defined(AFS_DARWIN80_ENV)
715 #define VREFCOUNT_GT(v, y) vnode_isinuse(AFSTOV(v), (y))
716 #elif defined(AFS_FBSD_ENV)
717 #define VREFCOUNT(v) (vrefcnt(AFSTOV(v)))
718 #define VREFCOUNT_GT(v, y) (vrefcnt(AFSTOV(v)) > y)
719 #elif defined(AFS_XBSD_ENV) || defined(AFS_DARWIN_ENV)
720 #define VREFCOUNT(v) ((v)->vrefCount)
721 #define VREFCOUNT_GT(v, y) (AFSTOV(v)->v_usecount > (y))
722 #elif defined(AFS_LINUX24_ENV)
723 #define VREFCOUNT(v) atomic_read(&(AFSTOV(v)->v_count))
724 #define VREFCOUNT_GT(v, y) (VREFCOUNT(v)>y)
725 #define VREFCOUNT_SET(v, c) atomic_set(&(AFSTOV(v)->v_count), c)
726 #define VREFCOUNT_DEC(v) atomic_dec(&(AFSTOV(v)->v_count))
727 #define VREFCOUNT_INC(v) atomic_inc(&(AFSTOV(v)->v_count))
729 #define VREFCOUNT(v) ((v)->vrefCount)
730 #define VREFCOUNT_GT(v,y) ((v)->vrefCount > (y))
731 #define VREFCOUNT_SET(v, c) (v)->vrefCount = c
732 #define VREFCOUNT_DEC(v) (v)->vrefCount--
733 #define VREFCOUNT_INC(v) (v)->vrefCount++
734 #define d_unhash(d) list_empty(&(d)->d_hash)
735 #define dget_locked(d) dget(d)
738 #define AFS_MAXDV 0x7fffffff /* largest dataversion number */
739 #ifdef AFS_64BIT_CLIENT
740 #define AFS_NOTRUNC 0x7fffffffffffffffLL /* largest positive int64 number */
741 #else /* AFS_64BIT_CLIENT */
742 #define AFS_NOTRUNC 0x7fffffff /* largest dataversion number */
743 #endif /* AFS_64BIT_CLIENT */
745 extern afs_int32 vmPageHog; /* counter for # of vnodes which are page hogs. */
747 #if defined(AFS_DARWIN80_ENV)
748 #define VTOAFS(v) ((struct vcache *)vnode_fsnode((v)))
749 #define AFSTOV(vc) ((vc)->v)
750 #elif defined(AFS_NBSD40_ENV)
752 struct genfs_node gfsn;
753 struct vcache *afsvc;
755 #define VTOAFS(v) ((((struct nbvdata *)((v)->v_data)))->afsvc)
756 #define AFSTOV(vc) ((vc)->v)
757 #elif defined(AFS_XBSD_ENV) || defined(AFS_DARWIN_ENV) || (defined(AFS_LINUX22_ENV) && !defined(STRUCT_SUPER_OPERATIONS_HAS_ALLOC_INODE))
758 #define VTOAFS(v) ((struct vcache *)(v)->v_data)
759 #define AFSTOV(vc) ((vc)->v)
761 #define VTOAFS(V) ((struct vcache *)(V))
762 #define AFSTOV(V) (&(V)->v)
770 /* VCache elements which are kept on disk, and in the kernel */
775 afs_hyper_t DataVersion;
779 afs_uint16 Mode; /* XXXX Should be afs_int32 XXXX */
780 afs_uint16 LinkCount;
781 #ifdef AFS_DARWIN80_ENV
784 /* vnode type is in v.v_type */
787 struct afs_vnuniq parent;
789 /*! Truncate file to this position at the next store */
792 /*! System:AnyUser's access to this. */
798 /*! Disconnected flags for this vcache element. */
799 afs_uint32 ddirty_flags;
800 /*! Shadow vnode + unique keep the shadow dir location. */
801 struct afs_vnuniq shadow;
802 /*! The old parent FID for renamed vnodes */
803 struct afs_vnuniq oldParent;
806 /* Values for 'mvstat' in struct vcache */
807 #define AFS_MVSTAT_FILE (0x0) /* regular file or directory */
808 #define AFS_MVSTAT_MTPT (0x1) /* mountpoint */
809 #define AFS_MVSTAT_ROOT (0x2) /* volume root dir */
813 * This is for the multiPage field in struct vcache. Each one of these
814 * represents an outstanding getpage request that is larger than a single page.
815 * Recording these is necessary to prevent afs_GetOnePage from trying to evict
816 * a dcache entry that an earlier afs_GetOnePage call got in the same getpage
817 * request. See osi_VM_MultiPageConflict and afs_getpage.
819 struct multiPage_range {
821 offset_t off; /**< offset of getpage request */
822 u_int len; /**< length of getpage request */
826 /* INVARIANTs: (vlruq.next != NULL) == (vlruq.prev != NULL)
827 * nextfree => !vlruq.next && ! vlruq.prev
828 * !(avc->nextfree) && !avc->vlruq.next => (FreeVCList == avc->nextfree)
831 #if defined(AFS_XBSD_ENV) || defined(AFS_DARWIN_ENV) || (defined(AFS_LINUX22_ENV) && !defined(STRUCT_SUPER_OPERATIONS_HAS_ALLOC_INODE))
834 struct vnode v; /* Has reference count in v.v_count */
836 struct afs_q vlruq; /* lru q next and prev */
837 #if !defined(AFS_LINUX22_ENV)
838 struct vcache *nextfree; /* next on free list (if free) */
840 struct vcache *hnext; /* Hash next */
841 struct afs_q vhashq; /* Hashed per-volume list */
842 /*! Queue of dirty vcaches. Lock with afs_disconDirtyLock */
844 /*! Queue of vcaches with shadow entries. Lock with afs_disconDirtyLock */
845 struct afs_q shadowq;
846 /*! Queue of vcaches with dirty metadata. Locked by afs_xvcdirty */
847 struct afs_q metadirty;
848 /*! Vcaches slot number in the disk backup. Protected by tvc->lock */
851 afs_rwlock_t lock; /* The lock on the vcache contents. */
852 #if defined(AFS_SUN5_ENV)
853 /* Lock used to protect the activeV, multipage, and vstates fields.
854 * Do not try to get the vcache lock when the vlock is held */
856 #endif /* defined(AFS_SUN5_ENV) */
857 #if defined(AFS_SUN5_ENV)
864 #if defined(AFS_AIX51_ENV)
872 int ownslock; /* pid of owner of excl lock, else 0 - defect 3083 */
874 #ifdef AFS_DARWIN80_ENV
876 #elif defined(AFS_DARWIN_ENV)
877 struct lock__bsd__ rwlock;
880 #if !defined(AFS_DFBSD_ENV) && !defined(AFS_NBSD_ENV)
886 char *silly_name; /* For sillyrenamed regular files, the silly
887 * name the file was renamed to. */
888 struct VenusFid *target_root; /* For mountpoints, the fid of the root dir
889 * in the target volume. */
890 struct VenusFid *parent; /* For root dir vcaches, the fid of the
894 char *linkData; /* Link data if a symlink. */
895 afs_hyper_t flushDV; /* data version last flushed from text */
896 afs_hyper_t mapDV; /* data version last flushed from map */
897 struct server *callback; /* The callback host, if any */
898 afs_uint32 cbExpires; /* time the callback expires */
899 struct afs_q callsort; /* queue in expiry order, sort of */
900 struct axscache *Access; /* a list of cached access bits */
901 afs_int32 last_looker; /* pag/uid from last lookup here */
902 #if defined(AFS_SUN5_ENV)
904 #endif /* defined(AFS_SUN5_ENV) */
905 struct SimpleLocks *slocks;
906 short opens; /* The numbers of opens, read or write, on this file. */
907 short execsOrWriters; /* The number of execs (if < 0) or writers (if > 0) of
909 short flockCount; /* count of flock readers, or -1 if writer */
910 char mvstat; /* see the AFS_MVSTAT_* constants */
912 char cachingStates; /* Caching policies for this file */
913 afs_uint32 cachingTransitions; /* # of times file has flopped between caching and not */
915 #if defined(AFS_LINUX24_ENV)
916 off_t next_seq_offset; /* Next sequential offset (used by prefetch/readahead) */
917 #elif defined(AFS_SUN5_ENV) || defined(AFS_SGI65_ENV)
918 off_t next_seq_blk_offset; /* accounted in blocks for Solaris & IRIX */
921 #if defined(AFS_SUN5_ENV)
922 afs_uint32 vstates; /* vstate bits */
923 #endif /* defined(AFS_SUN5_ENV) */
924 struct dcache *dchint;
925 struct dcache *dcreaddir; /* dcache for in-progress readdir */
926 unsigned int readdir_pid; /* pid of the thread in readdir */
927 #if defined(AFS_SGI_ENV)
928 daddr_t lastr; /* for read-ahead */
930 uint64_t vc_rwlockid; /* kthread owning rwlock */
932 short vc_rwlockid; /* pid of process owning rwlock */
934 short vc_locktrips; /* # of rwlock reacquisitions */
935 sema_t vc_rwlock; /* vop_rwlock for afs */
936 pgno_t mapcnt; /* # of pages mapped */
937 struct cred *cred; /* last writer's cred */
939 struct bhv_desc vc_bhv_desc; /* vnode's behavior data. */
941 #endif /* AFS_SGI_ENV */
942 #if defined(AFS_LINUX26_ENV)
943 cred_t *cred; /* last writer's cred */
945 #ifdef AFS_LINUX24_ENV
946 struct dentry *target_link; /* dentry we prefer, when we are redirecting
947 * all requests due to duplicate dentry aliases.
948 * See LINUX/osi_vnodeops.c. Note that this is
949 * NOT an actual reference to a dentry, so this
950 * pointer MUST NOT be dereferenced on its own. */
952 afs_int32 vc_error; /* stash write error for this vnode. */
953 int xlatordv; /* Used by nfs xlator */
955 int asynchrony; /* num kbytes to store behind */
957 struct afs_q multiPage; /* list of multiPage_range structs */
959 int protocol; /* RX_FILESERVER, RX_OSD, ... defined in afsint.xg */
960 #if !defined(UKERNEL)
961 void *vpacRock; /* used to read or write in visible partitions */
963 afs_uint32 lastBRLWarnTime; /* last time we warned about byte-range locks */
964 #ifdef AFS_LINUX26_ENV
965 spinlock_t pagewriter_lock;
966 struct list_head pagewriters; /* threads that are writing vm pages */
970 #ifdef AFS_LINUX26_ENV
972 struct list_head link;
977 #define DONT_CHECK_MODE_BITS 0
978 #define CHECK_MODE_BITS 1
979 #define CMB_ALLOW_EXEC_AS_READ 2 /* For the NFS xlator */
981 #if defined(AFS_SGI_ENV)
982 #define AVCRWLOCK(avc) (valusema(&(avc)->vc_rwlock) <= 0)
984 /* SGI vnode rwlock macros and flags. */
985 #ifndef AFS_SGI62_ENV
986 /* The following are defined here. SGI 6.2 declares them in vnode.h */
987 #define VRWLOCK_READ 0
988 #define VRWLOCK_WRITE 1
989 #define VRWLOCK_WRITE_DIRECT 2
994 #define AFS_RWLOCK_T vrwlock_t
996 #define AFS_RWLOCK_T int
997 #endif /* AFS_SGI62_ENV */
999 #include <ksys/behavior.h>
1000 #define AFS_RWLOCK(V,F) \
1001 afs_rwlock(&VTOAFS(V)->vc_bhv_desc, (F));
1002 #define AFS_RWUNLOCK(V,F) \
1003 afs_rwunlock(&VTOAFS(V)->vc_bhv_desc, (F));
1006 #define AFS_RWLOCK(V,F) afs_rwlock((vnode_t *)(V), (F) )
1007 #define AFS_RWUNLOCK(V,F) afs_rwunlock((vnode_t *)(V), (F) )
1009 #else /* AFS_SGI53_ENV */
1010 #define AFS_RWLOCK(V,F) afs_rwlock((V))
1011 #define AFS_RWUNLOCK(V,F) afs_rwunlock((V))
1012 #endif /* AFS_SGI53_ENV */
1013 #endif /* AFS_SGI_ENV */
1016 struct VenusFid fid;
1017 afs_hyper_t DataVersion;
1019 afs_int32 parentVnode;
1020 afs_int32 parentUnique;
1021 afs_hyper_t flushDV;
1024 afs_int32 randomUid[CPSIZE];
1025 afs_int32 callback; /* Now a pointer to 'server' struct */
1026 afs_int32 cbExpires;
1027 afs_int32 randomAccess[CPSIZE];
1028 afs_int32 anyAccess;
1030 short execsOrWriters;
1037 afs_int32 callerAccess;
1038 afs_int32 cbExpires;
1039 afs_int32 anyAccess;
1048 /* CM inititialization parameters. What CM actually used after calculations
1049 * based on passed in arguments.
1051 #define CMI_VERSION 1 /* increment when adding new fields. */
1052 struct cm_initparams {
1054 int cmi_nChunkFiles;
1055 int cmi_nStatCaches;
1056 int cmi_nDataCaches;
1057 int cmi_nVolumeCaches;
1058 int cmi_firstChunkSize;
1059 int cmi_otherChunkSize;
1060 int cmi_cacheSize; /* The original cache size, in 1K blocks. */
1061 unsigned cmi_setTime:1;
1062 unsigned cmi_memCache:1;
1063 int spare[16 - 9]; /* size of struct is 16 * 4 = 64 bytes */
1067 /*----------------------------------------------------------------------
1068 * AFS Data cache definitions
1070 * Each entry describes a Unix file on the local disk that is
1071 * is serving as a cached copy of all or part of a Vice file.
1072 * Entries live in circular queues for each hash table slot
1074 * Which queue is this thing in? Good question.
1075 * A struct dcache entry is in the freeDSlot queue when not associated with a cache slot (file).
1076 * Otherwise, it is in the DLRU queue. The freeDSlot queue uses the lruq.next field as
1077 * its "next" pointer.
1079 * Cache entries in the DLRU queue are either associated with vice files, in which case
1080 * they are hashed by afs_dvnextTbl and afs_dcnextTbl pointers, or they are in the freeDCList
1081 * and are not associated with any vice file. This last list uses the afs_dvnextTbl pointer for
1082 * its "next" pointer.
1083 *----------------------------------------------------------------------*/
1085 #define NULLIDX (-1) /* null index definition */
1086 /* struct dcache states bits */
1090 #define DWriting 8 /* file being written (used for cache validation) */
1092 /* dcache data flags */
1093 #define DFEntryMod 0x02 /* has entry itself been modified? */
1094 #define DFFetching 0x04 /* file is currently being fetched */
1096 /* dcache meta flags */
1097 #define DFNextStarted 0x01 /* next chunk has been prefetched already */
1098 #define DFFetchReq 0x10 /* someone is waiting for DFFetching to go on */
1101 /* flags in afs_indexFlags array */
1102 #define IFEverUsed 1 /* index entry has >= 1 byte of data */
1103 #define IFFree 2 /* index entry in freeDCList */
1104 #define IFDataMod 4 /* file needs to be written out */
1105 #define IFFlag 8 /* utility flag */
1106 #define IFDirtyPages 16 /* Solaris-only. contains dirty pages */
1107 #define IFAnyPages 32
1108 #define IFDiscarded 64 /* index entry in discardDCList */
1110 #ifdef AFS_DARWIN100_ENV
1111 typedef user_addr_t iparmtype; /* 64 bit */
1112 typedef user_addr_t uparmtype; /* 64 bit */
1114 typedef char * uparmtype;
1115 #ifdef AFS_SGI65_ENV
1116 typedef afs_uint32 iparmtype;
1118 typedef long iparmtype;
1122 #if SIZEOF_VOID_P == SIZEOF_UNSIGNED_INT
1123 # define uintptrsz unsigned int
1124 #elif SIZEOF_VOID_P == SIZEOF_UNSIGNED_LONG
1125 # define uintptrsz unsigned long
1126 #elif SIZEOF_VOID_P == SIZEOF_UNSIGNED_LONG_LONG
1127 # define uintptrsz unsigned long long
1129 # error "Unable to determine casting for pointers"
1133 uparmtype in; /* input buffer */
1134 uparmtype out; /* output buffer */
1135 short in_size; /* Size of input buffer <= 2K */
1136 short out_size; /* Maximum size of output buffer, <= 2K */
1140 * This version of afs_ioctl is required to pass in 32 bit user space
1141 * pointers into a 64 bit kernel.
1144 struct afs_ioctl32 {
1152 /* CacheItems file has a header of type struct afs_fheader
1153 * (keep aligned properly). Since we already have sgi_62 clients running
1154 * with a 32 bit inode, a change is required to the header so that
1155 * they can distinguish the old 32 bit inode CacheItems file and zap it
1156 * instead of using it.
1158 struct afs_fheader {
1159 #define AFS_FHMAGIC 0x7635abaf /* uses version number */
1161 #define AFS_CI_VERSION 4
1163 afs_uint32 dataSize;
1164 afs_int32 firstCSize;
1165 afs_int32 otherCSize;
1168 #if defined(AFS_CACHE_VNODE_PATH)
1169 typedef char *afs_ufs_dcache_id_t;
1170 #elif defined(AFS_SGI61_ENV) || defined(AFS_SUN5_64BIT_ENV)
1171 /* Using ino64_t here so that user level debugging programs compile
1172 * the size correctly.
1174 typedef ino64_t afs_ufs_dcache_id_t;
1175 #elif defined(AFS_LINUX26_ENV)
1176 #define MAX_FH_LEN 10
1178 #if defined(NEW_EXPORT_OPS)
1181 __u32 raw[MAX_FH_LEN];
1182 } afs_ufs_dcache_id_t;
1183 extern int cache_fh_type;
1184 extern int cache_fh_len;
1185 #elif defined(AFS_LINUX_64BIT_KERNEL) && !defined(AFS_S390X_LINUX24_ENV)
1186 typedef long afs_ufs_dcache_id_t;
1187 #elif defined(AFS_AIX51_ENV) || defined(AFS_HPUX1123_ENV)
1188 typedef ino_t afs_ufs_dcache_id_t;
1190 typedef afs_int32 afs_ufs_dcache_id_t;
1193 typedef afs_int32 afs_mem_dcache_id_t;
1196 afs_ufs_dcache_id_t ufs;
1197 afs_mem_dcache_id_t mem;
1200 /* it does not compile outside kernel */
1202 afs_int32 fid; /* is adc->index, the cache file number */
1203 afs_dcache_id_t inode; /* is adc->f.inode, the inode number of the cac\
1206 afs_int32 accesstime;
1207 struct buffer *hashNext;
1212 afs_rwlock_t lock; /* the lock for this structure */
1215 /* kept on disk and in dcache entries */
1217 struct VenusFid fid; /* Fid for this file */
1218 afs_int32 modTime; /* last time this entry was modified */
1219 afs_hyper_t versionNo; /* Associated data version number */
1220 afs_int32 chunk; /* Relative chunk number */
1221 afs_dcache_id_t inode; /* Unix inode for this chunk */
1222 afs_int32 chunkBytes; /* Num bytes in this chunk */
1223 char states; /* Has this chunk been modified? */
1226 /* magic numbers to specify the cache type */
1228 #define AFS_FCACHE_TYPE_UFS 0x0
1229 #define AFS_FCACHE_TYPE_MEM 0x1
1230 #define AFS_FCACHE_TYPE_NFS 0x2
1231 #define AFS_FCACHE_TYPE_EPI 0x3
1233 /* kept in memory */
1235 struct afs_q lruq; /* Free queue for in-memory images */
1236 struct afs_q dirty; /* Queue of dirty entries that need written */
1237 afs_rwlock_t lock; /* Protects validPos, some f */
1238 afs_rwlock_t tlock; /* Atomizes updates to refCount */
1239 afs_rwlock_t mflock; /* Atomizes accesses/updates to mflags */
1240 afs_size_t validPos; /* number of valid bytes during fetch */
1241 afs_int32 index; /* The index in the CacheInfo file */
1242 short refCount; /* Associated reference count. */
1243 char dflags; /* Data flags */
1244 char mflags; /* Meta flags */
1245 struct fcache f; /* disk image */
1246 afs_int32 bucket; /* which bucket these dcache entries are in */
1250 * dcache.lock protects the actual contents of the cache file (in
1251 * f.inode), subfields of f except those noted below, dflags and
1254 * dcache.tlock is used to make atomic updates to refCount. Zero
1255 * refCount dcache entries are protected by afs_xdcache instead of
1258 * dcache.mflock is used to access and update mflags. It cannot be
1259 * held without holding the corresponding dcache.lock. Updating
1260 * mflags requires holding dcache.lock(R) and dcache.mflock(W), and
1261 * checking for mflags requires dcache.lock(R) and dcache.mflock(R).
1262 * Note that dcache.lock(W) gives you the right to update mflags,
1263 * as dcache.mflock(W) can only be held with dcache.lock(R).
1265 * dcache.index, dcache.f.fid, dcache.f.chunk and dcache.f.inode are
1266 * write-protected by afs_xdcache and read-protected by refCount.
1267 * Once an entry is referenced, these values cannot change, and if
1268 * it's on the free list (with refCount=0), it can be reused for a
1269 * different file/chunk. These values can only be written while
1270 * holding afs_xdcache(W) and allocating this dcache entry (thereby
1271 * ensuring noone else has a refCount on it).
1275 /* afs_memcache.c */
1276 struct memCacheEntry {
1277 int size; /* # of valid bytes in this entry */
1278 int dataSize; /* size of allocated data area */
1279 afs_lock_t afs_memLock;
1280 char *data; /* bytes */
1283 struct afs_FetchOutput {
1284 struct AFSVolSync tsync;
1285 struct AFSFetchStatus OutStatus;
1286 struct AFSCallBack CallBack;
1289 /* macro to mark a dcache entry as bad */
1292 (x)->f.fid.Fid.Unique = 0; \
1293 afs_indexUnique[(x)->index] = 0; \
1294 (x)->dflags |= DFEntryMod; \
1297 /* FakeOpen and Fake Close used to be real subroutines. They're only used in
1298 * sun_subr and afs_vnodeops, and they're very frequently called, so I made
1299 * them into macros. They do:
1300 * FakeOpen: fake the file being open for writing. avc->lock must be held
1301 * in write mode. Having the file open for writing is like having a DFS
1302 * write-token: you're known to have the best version of the data around,
1303 * and so the CM won't let it be overwritten by random server info.
1304 * FakeClose: undo the effects of FakeOpen, noting that we want to ensure
1305 * that a real close eventually gets done. We use CCore to achieve this if
1306 * we would end up closing the file. avc->lock must be held in write mode */
1308 #ifdef AFS_AIX_IAUTH_ENV
1309 #define CRKEEP(V, C) (V)->linkData = (char*)crdup((C))
1311 #define CRKEEP(V, C) crhold((C)); (V)->linkData = (char*)(C)
1314 #define afs_FakeOpen(avc) { avc->opens++; avc->execsOrWriters++; }
1315 #define afs_FakeClose(avc, acred) \
1316 { if (avc->execsOrWriters == 1) { \
1317 /* we're the last writer, just use CCore flag */ \
1318 avc->f.states |= CCore; /* causes close to be called later */ \
1320 /* The cred and vnode holds will be released in afs_FlushActiveVcaches */ \
1321 AFS_FAST_HOLD(avc); /* So it won't disappear */ \
1322 CRKEEP(avc, acred); /* Should use a better place for the creds */ \
1325 /* we're not the last writer, let the last one do the store-back for us */ \
1327 avc->execsOrWriters--; \
1331 #define AFS_ZEROS 64 /* zero buffer */
1333 /*#define afs_DirtyPages(avc) (((avc)->f.states & CDirty) || osi_VMDirty_p((avc)))*/
1334 #define afs_DirtyPages(avc) ((avc)->f.states & CDirty)
1336 #define afs_InReadDir(avc) (((avc)->f.states & CReadDir) && (avc)->readdir_pid == MyPidxx2Pid(MyPidxx))
1338 /* The PFlush algorithm makes use of the fact that Fid.Unique is not used in
1339 below hash algorithms. Change it if need be so that flushing algorithm
1340 doesn't move things from one hash chain to another
1342 /* extern int afs_dhashsize; */
1343 #define DCHash(v, c) ((((v)->Fid.Vnode + (v)->Fid.Volume + (c))) & (afs_dhashsize-1))
1344 /*Vnode, Chunk -> Hash table index */
1345 #define DVHash(v) ((((v)->Fid.Vnode + (v)->Fid.Volume )) & (afs_dhashsize-1))
1346 /*Vnode -> Other hash table index */
1347 /* don't hash on the cell, our callback-breaking code sometimes fails to compute
1348 the cell correctly, and only scans one hash bucket */
1349 #define VCHash(fid) (((fid)->Fid.Volume + (fid)->Fid.Vnode) & (VCSIZE-1))
1350 /* Hash only on volume to speed up volume callbacks. */
1351 #define VCHashV(fid) ((fid)->Fid.Volume & (VCSIZE-1))
1353 extern struct dcache **afs_indexTable; /*Pointers to in-memory dcache entries */
1354 extern afs_int32 *afs_indexUnique; /*dcache entry Fid.Unique */
1355 extern afs_int32 *afs_dvnextTbl; /*Dcache hash table links */
1356 extern afs_int32 *afs_dcnextTbl; /*Dcache hash table links */
1357 extern afs_int32 afs_cacheFiles; /*Size of afs_indexTable */
1358 extern afs_int32 afs_cacheBlocks; /*1K blocks in cache */
1359 extern afs_int32 afs_cacheStats; /*Stat entries in cache */
1360 extern struct vcache *afs_vhashT[VCSIZE]; /*Stat cache hash table */
1361 extern struct afs_q afs_vhashTV[VCSIZE]; /* cache hash table on volume */
1362 extern afs_int32 afs_initState; /*Initialization state */
1363 extern afs_int32 afs_termState; /* Termination state */
1364 extern struct VenusFid afs_rootFid; /*Root for whole file system */
1365 extern afs_int32 afs_allCBs; /* Count of callbacks */
1366 extern afs_int32 afs_oddCBs; /* Count of odd callbacks */
1367 extern afs_int32 afs_evenCBs; /* Count of even callbacks */
1368 extern afs_int32 afs_allZaps; /* Count of fid deletes */
1369 extern afs_int32 afs_oddZaps; /* Count of odd fid deletes */
1370 extern afs_int32 afs_evenZaps; /* Count of even fid deletes */
1371 extern struct brequest afs_brs[NBRS]; /* request structures */
1373 #define UHash(auid) ((auid) & (NUSERS-1))
1374 #define VHash(avol) ((avol)&(NVOLS-1))
1375 #define SHash(aserv) ((ntohl(aserv)) & (NSERVERS-1))
1376 #define FVHash(acell,avol) (((avol)+(acell)) & (NFENTRIES-1))
1378 /* Performance hack - we could replace VerifyVCache2 with the appropriate
1379 * GetVCache incantation, and could eliminate even this code from afs_UFSRead
1380 * by making intentionally invalidating quick.stamp in the various callbacks
1381 * expiration/breaking code */
1382 #ifdef AFS_DARWIN_ENV
1383 #define afs_VerifyVCache(avc, areq) \
1384 (((avc)->f.states & CStatd) ? (osi_VM_Setup(avc, 0), 0) : \
1385 afs_VerifyVCache2((avc),areq))
1387 #define afs_VerifyVCache(avc, areq) \
1388 (((avc)->f.states & CStatd) ? 0 : afs_VerifyVCache2((avc),areq))
1391 #define DO_STATS 1 /* bits used by FindVCache */
1396 /* values for adown value of afs_LoopServers */
1398 #define AFS_LS_DOWN 1
1399 #define AFS_LS_ALL 2
1401 /* values for flag param of afs_CheckVolumeNames */
1402 #define AFS_VOLCHECK_EXPIRED 0x1 /* volumes whose callbacks have expired */
1403 #define AFS_VOLCHECK_BUSY 0x2 /* volumes which were marked busy */
1404 #define AFS_VOLCHECK_MTPTS 0x4 /* mount point invalidation also */
1405 #define AFS_VOLCHECK_FORCE 0x8 /* do all forcibly */
1409 #define AFS_FSPORT ((unsigned short) htons(7000))
1410 #define AFS_VLPORT ((unsigned short) htons(7003))
1411 #define AFS_RXOSDPORT ((unsigned short) htons(7011))
1413 #define afs_rdwr(avc, uio, rw, io, cred) \
1414 (((rw) == UIO_WRITE) ? afs_write(avc, uio, io, cred, 0) : afs_read(avc, uio, cred, 0))
1415 #define afs_nlrdwr(avc, uio, rw, io, cred) \
1416 (((rw) == UIO_WRITE) ? afs_write(avc, uio, io, cred, 1) : afs_read(avc, uio, cred, 1))
1418 /* Cache size truncation uses the following low and high water marks:
1419 * If the cache is more than 90% full by space (CM_DCACHESPACEFREEPCT) or
1420 * 95% full by # of chunks (CM_DCACHECOUNTFREEPCT), the cache truncation
1421 * daemon is awakened and will start freeing space. The cache truncation
1422 * daemon will continue to free space until the cache is at most 85% full
1423 * by space (CM_DCACHESPACEFREEPCT - CM_DCACHEEXTRAPCT). The cache truncation
1424 * daemon will also try to free space until the cache is at most 90% full by
1425 * chunks (CM_DCACHECOUNTFREEPCT - CM_DCACHEEXTRAPCT), but the 85% space limit
1426 * is the only limit that we must hit.
1427 * afs_UFSWrite and afs_GetDCache (when it needs to fetch data) will wait on
1428 * afs_WaitForCacheDrain if the cache is 98% (CM_WAITFORDRAINPCT) full.
1429 * afs_GetDownD wakes those processes once the cache is 95% full
1430 * (CM_CACHESIZEDRAINEDPCT).
1432 #define CM_MAXDISCARDEDCHUNKS 16 /* # of chunks */
1433 #define CM_DCACHECOUNTFREEPCT 95 /* max pct of chunks in use */
1434 #define CM_DCACHESPACEFREEPCT 90 /* max pct of space in use */
1435 #define CM_DCACHEEXTRAPCT 5 /* extra to get when freeing */
1436 #define CM_CACHESIZEDRAINEDPCT 95 /* wakeup processes when down to here. */
1437 #define CM_WAITFORDRAINPCT 98 /* sleep if cache is this full. */
1439 /* when afs_cacheBlocks is large, settle for slightly decreased precision */
1440 #define PERCENT(p, v) \
1441 ((afs_cacheBlocks & 0xffe00000) ? ((v) / 100 * (p)) : ((p) * (v) / 100))
1443 #define afs_CacheIsTooFull() \
1444 (afs_blocksUsed - afs_blocksDiscarded > \
1445 PERCENT(CM_DCACHESPACEFREEPCT, afs_cacheBlocks) || \
1446 afs_freeDCCount - afs_discardDCCount < \
1447 PERCENT(100 - CM_DCACHECOUNTFREEPCT, afs_cacheFiles))
1449 /* Handy max length of a numeric string. */
1450 #define CVBS 12 /* max afs_int32 is 2^32 ~ 4*10^9, +1 for NULL, +luck */
1452 #define refpanic(foo) if (afs_norefpanic) \
1453 { printf( foo ); afs_norefpanic++;} else osi_Panic( foo )
1456 ** these are defined in the AIX source code sys/fs_locks.h but are not
1457 ** defined anywhere in the /usr/include directory
1459 #if defined(AFS_AIX41_ENV)
1460 #define VN_LOCK(vp) simple_lock(&(vp)->v_lock)
1461 #define VN_UNLOCK(vp) simple_unlock(&(vp)->v_lock)
1464 /* get a file's serial number from a vnode */
1465 #ifndef afs_vnodeToInumber
1466 #if defined(AFS_SGI62_ENV) || defined(AFS_HAVE_VXFS) || defined(AFS_DARWIN_ENV)
1467 #define afs_vnodeToInumber(V) VnodeToIno(V)
1469 #define afs_vnodeToInumber(V) (VTOI(V)->i_number)
1470 #endif /* AFS_SGI62_ENV */
1473 /* get a file's device number from a vnode */
1474 #ifndef afs_vnodeToDev
1475 #if defined(AFS_SGI62_ENV) || defined(AFS_HAVE_VXFS) || defined(AFS_DARWIN_ENV)
1476 #define afs_vnodeToDev(V) VnodeToDev(V)
1477 #elif defined(UKERNEL)
1478 #define afs_vnodeToDev(V) (VTOI(V) ? (VTOI(V)->i_dev) : (-1))
1480 #define afs_vnodeToDev(V) (VTOI(V)->i_dev)
1484 /* declare something so that prototypes don't flip out */
1485 /* appears struct buf stuff is only actually passed around as a pointer,
1486 except with libuafs, in which case it is actually defined */
1490 struct rxfs_storeVariables {
1491 struct rx_call *call;
1492 struct vcache *vcache;
1496 afs_int32 hasNo64bit;
1497 struct AFSStoreStatus InStatus;
1501 int (*prepare)(void *rock, afs_uint32 size, afs_uint32 *bytestoxfer);
1502 int (*read)(void *rock, struct osi_file *tfile, afs_uint32 offset,
1503 afs_uint32 tlen, afs_uint32 *bytesread);
1504 int (*write)(void *rock, afs_uint32 tlen, afs_uint32 *byteswritten);
1505 int (*status)(void *rock);
1506 int (*padd)(void *rock, afs_uint32 tlen);
1507 int (*close)(void *rock, struct AFSFetchStatus *OutStatus,
1508 afs_int32 *doProcessFS);
1509 int (*destroy)(void **rock, afs_int32 error);
1510 int (*storeproc)(struct storeOps *, void *, struct dcache *, int *,
1515 int (*more)(void *rock, afs_int32 *length, afs_uint32 *moredata);
1516 int (*read)(void *rock, afs_uint32 tlen, afs_uint32 *bytesread);
1517 int (*write)(void *rock, struct osi_file *fp, afs_uint32 offset,
1518 afs_uint32 tlen, afs_uint32 *byteswritten);
1519 int (*close)(void *rock, struct vcache *avc, struct dcache *adc,
1520 struct afs_FetchOutput *Outputs);
1521 int (*destroy)(void **rock, afs_int32 error);
1524 /* fakestat support: opaque storage for afs_EvalFakeStat to remember
1525 * what vcache should be released.
1527 struct afs_fakestat_state {
1531 struct vcache *root_vp;
1534 extern int afs_fakestat_enable;
1535 extern int afs_rmtsys_enable;
1536 extern int afsd_dynamic_vcaches;
1539 * Wrappers for access to credentials structure members
1540 * Linux uses the kernel cred structure if available, with the
1541 * wrappers defined in LINUX/osi_machdep.h
1543 #if defined(AFS_NBSD40_ENV)
1544 /* in osi_machdep.h as expected */
1545 #elif defined (AFS_DARWIN110_ENV)
1546 #define afs_cr_uid(cred) kauth_cred_getuid((kauth_cred_t)(cred))
1547 #define afs_cr_gid(cred) kauth_cred_getgid((kauth_cred_t)(cred))
1548 #elif !(defined(AFS_LINUX26_ENV) && defined(STRUCT_TASK_STRUCT_HAS_CRED))
1549 #define afs_cr_uid(cred) ((cred)->cr_uid)
1550 #define afs_cr_gid(cred) ((cred)->cr_gid)
1551 #if !defined(AFS_OBSD_ENV)
1552 #define afs_cr_ruid(cred) ((cred)->cr_ruid)
1553 #define afs_cr_rgid(cred) ((cred)->cr_rgid)
1556 #if !defined(AFS_DARWIN110_ENV)
1558 afs_set_cr_uid(afs_ucred_t *cred, uid_t uid) {
1562 afs_set_cr_gid(afs_ucred_t *cred, gid_t gid) {
1565 #if !defined(AFS_OBSD_ENV)
1567 afs_set_cr_ruid(afs_ucred_t *cred, uid_t uid) {
1568 cred->cr_ruid = uid;
1571 afs_set_cr_rgid(afs_ucred_t *cred, gid_t gid) {
1572 cred->cr_rgid = gid;
1574 #endif /* ! AFS_OBSD_ENV */
1575 #endif /* ! AFS_DARWIN110_ENV */
1580 /** The 32 bit OS expects the members of this structure to be 32 bit
1581 * quantities and the 64 bit OS expects them as 64 bit quanties. Hence
1582 * to accomodate both, *long* is used instead of afs_int32
1594 extern int Afs_syscall(struct afssysa *uap, rval_t *rvp);
1595 #endif /* AFS_SUN5_ENV */
1598 * Various definitions for osi_sleep and its event hash table
1599 * DFBSD and SUNOS have no osi_sleep, and HPUX has its own hack for this stuff
1601 #define AFS_EVHASHSIZE 128 /* size of afs_evhasht, must be power of 2 */
1603 typedef struct afs_event {
1604 struct afs_event *next; /* next in hash chain */
1605 char *event; /* lwp event: an address */
1606 int refcount; /* Is it in use? */
1607 int seq; /* Sequence number: this is incremented
1608 * by wakeup calls; wait will not return until
1610 #if defined(AFS_AIX_ENV)
1612 #elif defined(AFS_DARWIN_ENV)
1613 # ifdef AFS_DARWIN80_ENV
1617 /* no cond member */
1618 #elif defined(AFS_FBSD_ENV) || defined(AFS_OBSD_ENV)
1619 int cond; /* "all this gluck should probably be replaced by CVs" */
1620 #elif defined(AFS_LINUX24_ENV)
1621 wait_queue_head_t cond;
1622 #elif defined(AFS_NBSD_ENV) || defined(AFS_SUN5_ENV) || defined(AFS_SGI_ENV)
1623 kcondvar_t cond; /* Currently associated condition variable */
1627 extern afs_event_t *afs_evhasht[AFS_EVHASHSIZE]; /* Hash table for events */
1629 #endif /* _AFS_H_ */