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"
23 /* jhash.h is a standalone header and is fine to pull into kernel code. */
24 #include <opr/jhash.h>
27 * afs_fsfragsize cannot be less than 1023, or some cache-tracking
28 * calculations will be incorrect (since we track cache usage in kb).
29 * Some filesystems have fundamental blocksizes less than 1k, and
30 * normally we would tune afs_fsfragsize to be fragsize-1, but we must
31 * make sure to check that afs_fsfragsize does not go below this value.
33 #define AFS_MIN_FRAGSIZE 1023
35 /* Upper bound on number of iovecs out uio routines will deal with. */
36 #define AFS_MAXIOVCNT 16
38 enum afs_shutdown_state {
43 extern enum afs_shutdown_state afs_shuttingdown;
46 * Macros to uniquely identify the AFS vfs struct
48 #define AFS_VFSMAGIC 0x1234
51 # if defined(AFS_USR_AIX_ENV) || defined(AFS_USR_SGI_ENV)
52 # define AFS_VFSFSID AFS_MOUNT_AFS
54 # define AFS_VFSFSID 99
56 #elif defined(AFS_SUN5_ENV) || defined(AFS_HPUX90_ENV) || defined(AFS_LINUX20_ENV)
57 # define AFS_VFSFSID 99
58 #elif defined(AFS_SGI_ENV)
59 # define AFS_VFSFSID afs_fstype
61 # define AFS_VFSFSID AFS_MOUNT_AFS
63 /* use this value for reporting total space, free space, etc.
64 * fake a high number to satisfy programs that use the statfs call to make sure
65 * that there's enough space in the device partition before storing something
66 * there. keep it just under 2^31, to try and be safe about not breaking
67 * stuff that looks at the amount of free space. */
68 #define AFS_VFS_FAKEFREE (2147483647)
70 /* Moved from VNOPS/afs_vnop_flocks so can be used in prototypes */
71 #if defined(AFS_HPUX102_ENV)
72 #define AFS_FLOCK k_flock
74 #if defined(AFS_SUN5_ENV) || (defined(AFS_LINUX24_ENV) && !(defined(AFS_LINUX26_ENV) && defined(AFS_LINUX_64BIT_KERNEL)))
75 #define AFS_FLOCK flock64
77 #define AFS_FLOCK flock
78 #endif /* AFS_SUN65_ENV */
79 #endif /* AFS_HPUX102_ENV */
81 /* The following are various levels of afs debugging */
82 #define AFSDEB_GENERAL 1 /* Standard debugging */
83 #define AFSDEB_NETWORK 2 /* low level afs networking */
84 #define AFSDEB_RX 4 /* RX debugging */
85 #define AFSDEB_VNLAYER 8 /* interface layer to AFS (aixops, gfsops, etc) */
87 /* generic undefined vice id */
90 /* The basic defines for the Andrew file system
91 better keep things powers of two so "& (foo-1)" hack works for masking bits */
92 #define NBRS 15 /* max number of queued daemon requests */
93 #define NUSERS 16 /* hash table size for unixuser table */
94 #define NSERVERS 16 /* hash table size for server table */
95 #define NVOLS 64 /* hash table size for volume table */
96 #define NFENTRIES 256 /* hash table size for disk volume table */
97 #define VCSIZEBITS 16 /* log of stat cache hash table size */
98 #define VCSIZE (opr_jhash_size(VCSIZEBITS))
99 #define CBRSIZE 512 /* call back returns hash table size */
100 #define PIGGYSIZE 1350 /* max piggyback size */
101 #define MAXVOLS 128 /* max vols we can store */
102 #define MAXSYSNAME 128 /* max sysname (i.e. @sys) size */
103 #define MAXNUMSYSNAMES 32 /* max that current constants allow */
104 #define MAXROOTVOLNAMELEN 64 /* max length of root volume name */
105 #define NOTOKTIMEOUT (2*3600) /* time after which to timeout conns sans tokens */
106 #define NOPAG 0xffffffff
110 #define AFS_MAXCBRSCALL 32 /* max to return in a given call (must be <= AFSCBMAX) */
111 #define VCACHE_FREE 5
112 #define VCACHE_DYNAMIC_STRESSED 10000
113 #define VCACHE_STRESS_LOGINTERVAL (60*60*4) /* 4 hours */
114 #define AFS_NRXPACKETS 80
115 #define AFS_RXDEADTIME 50
116 #define AFS_HARDDEADTIME 120
117 #define AFS_IDLEDEADTIME 1200
118 #define AFS_IDLEDEADTIME_REP 180 /* more than fs's cb dead time */
119 #define AFS_BLKBITS 12
120 #define AFS_BLKSIZE (1 << AFS_BLKBITS)
122 extern afs_int32 afs_rx_deadtime;
123 extern afs_int32 afs_rx_harddead;
124 extern afs_int32 afs_rx_idledead;
125 extern afs_int32 afs_rx_idledead_rep;
127 struct sysname_info {
130 signed char index, allocked;
133 /* flags to use with AFSOP_CACHEINIT */
134 #define AFSCALL_INIT_MEMCACHE 0x1 /* use a memory-based cache */
136 /* below here used only for kernel procedures */
138 /* Store synchrony flags - SYNC means that data should be forced to server's
139 * disk immediately upon completion. */
142 #define AFS_VMSYNC_INVAL 2 /* sync and invalidate pages */
143 #define AFS_LASTSTORE 4
144 #define AFS_VMSYNC 8 /* sync pages but do not invalidate */
146 /* background request structure */
149 #define BOP_NOOP 0 /* leave 0 unused */
150 #define BOP_FETCH 1 /* parm1 is chunk to get */
151 #define BOP_STORE 2 /* parm1 is chunk to store */
152 #define BOP_PATH 3 /* parm1 is path, parm2 is chunk to fetch */
153 #define BOP_FETCH_NOCACHE 4 /* parms are: vnode ptr, offset, segment ptr, addr, cred ptr */
154 #ifdef AFS_DARWIN_ENV
155 #define BOP_MOVE 5 /* ptr1 afs_uspc_param ptr2 sname ptr3 dname */
157 #define BOP_PARTIAL_STORE 6 /* parm1 is chunk to store */
158 #define BOP_INVALIDATE_SEGMENTS 7 /* no parms: just uses the 'bp->vc' vcache */
160 #define B_DONTWAIT 1 /* On failure return; don't wait */
162 /* protocol is: refCount is incremented by user to take block out of free pool.
163 Next, BSTARTED is set when daemon finds request. This prevents
164 other daemons from picking up the same request. Finally, when
165 request is done, refCount is zeroed. BDONE and BWAIT are used by
166 dudes waiting for operation to proceed to a certain point before returning.
168 #define BSTARTED 1 /* request picked up by a daemon */
169 #define BUVALID 2 /* code is valid (store) */
170 #define BUWAIT 4 /* someone is waiting for BUVALID */
172 struct vcache *vc; /* vnode to use, with vrefcount bumped */
173 afs_ucred_t *cred; /* credentials to use for operation */
174 afs_size_t size_parm[BPARMS]; /* random parameters */
175 void *ptr_parm[BPARMS]; /* pointer parameters */
176 afs_int32 code_raw; /* return code from AFS routines */
177 afs_int32 code_checkcode; /* the afs_CheckCode-translated code */
178 short refCount; /* use counter for this structure */
179 char opcode; /* what to do (store, fetch, etc) */
180 char flags; /* free, etc */
181 afs_int32 ts; /* counter "timestamp" */
189 afs_int32 AuthHandle;
190 char HandShakeKey[8];
192 afs_int32 BeginTimestamp;
193 afs_int32 EndTimestamp;
197 afs_int32 Cell; /* better sun packing if at end of structure */
201 /* Temporary struct to be passed between afs_fid and afs_vget */
204 afs_int32 CellAndUnique;
207 /* The actual number of bytes in the SmallFid, not the sizeof struct. */
208 #define SIZEOF_SMALLFID 10
213 * Circular queues, implemented with pointers. Structures may contain as many
214 * queues as required, which may be located at any point within the structure,
215 * providing the QEntry macro is used to translate between a queue pointer, and
216 * the address of its containing structure
224 #define QZero(e) ((e)->prev = (e)->next = NULL)
225 #define QInit(q) ((q)->prev = (q)->next = (q))
226 #define QAdd(q,e) ((e)->next = (q)->next, (e)->prev = (q), \
227 (q)->next->prev = (e), (q)->next = (e))
228 #define QRemove(e) ((e)->next->prev = (e)->prev, (e)->prev->next = (e)->next, (e)->prev = NULL, (e)->next = NULL)
229 #define QNext(e) ((e)->next)
230 #define QPrev(e) ((e)->prev)
231 #define QEmpty(q) ((q)->prev == (q))
232 /* this one takes q1 and sticks it on the end of q2 - that is, the other end, not the end
233 * that things are added onto. q1 shouldn't be empty, it's silly */
234 #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))
236 /* Given a pointer to an afs_q within a structure, go back to the address of
237 * the parent structure
240 #define QEntry(queue, structure, member) \
241 ((structure *)((char *)(queue)-(char *)(&((structure *)NULL)->member)))
243 /* And implement operations for individual lists in terms of the above macro */
245 #define QTOV(e) QEntry(e, struct vcache, vlruq)
246 #define QTOC(e) QEntry(e, struct cell, lruq)
247 #define QTOVH(e) QEntry(e, struct vcache, vhashq)
250 * List of free slot numbers
252 struct afs_slotlist {
254 struct afs_slotlist *next;
258 afs_int32 uid; /* user id making the request */
259 afs_int32 busyCount; /* how many busies we've seen so far */
260 afs_int32 flags; /* things like O_SYNC, O_NONBLOCK go here */
261 char initd; /* if non-zero, Error fields meaningful */
262 char accessError; /* flags for overriding error return code */
263 char volumeError; /* encountered a missing or busy volume */
264 char networkError; /* encountered network problems */
265 char permWriteError; /* fileserver returns permenent error. */
266 char tokenError; /* a token error other than expired. */
267 char idleError; /* the server idled too long */
268 char skipserver[AFS_MAXHOSTS];
269 afs_int32 lasterror[AFS_MAXHOSTS];
274 /* structure linked off of a server to keep track of queued returned
275 * callbacks. Sent asynchronously when we run a little low on free dudes.
278 struct afs_cbr **pprev;
279 struct afs_cbr *next;
281 struct afs_cbr **hash_pprev;
282 struct afs_cbr *hash_next;
287 #ifdef AFS_LINUX22_ENV
288 /* On Linux, we have to be able to allocate the storage for this using
289 * kmalloc, as otherwise we may deadlock. So, it needs to be able to fit
292 # define AFS_NCBRS PAGE_SIZE/sizeof(struct afs_cbr)
294 # define AFS_NCBRS 1024 /* max # of call back return entries */
297 /* cellinfo file magic number */
298 #define AFS_CELLINFO_MAGIC 0xf32817cd
301 #define CNoSUID 0x02 /* disable suid bit for this cell */
302 #define CLinkedCell4 0x04 /* reserved for ADDCELL2 pioctl */
303 #define CNoAFSDB 0x08 /* never bother trying AFSDB */
304 #define CHasVolRef 0x10 /* volumes were referenced */
305 #define CLinkedCell 0x20 /* has a linked cell in lcellp */
306 #define CHush 0x40 /* don't display until referenced */
309 struct afs_q lruq; /* lru q next and prev */
310 char *cellName; /* char string name of cell */
311 afs_int32 cellIndex; /* sequence number */
312 afs_int32 cellNum; /* semi-permanent cell number */
313 struct server *cellHosts[AFS_MAXCELLHOSTS]; /* volume *location* hosts */
314 struct cell *lcellp; /* Associated linked cell */
315 u_short fsport; /* file server port */
316 u_short vlport; /* volume server port */
317 short states; /* state flags */
318 time_t timeout; /* data expire time, if non-zero */
319 struct cell_name *cnamep; /* pointer to our cell_name */
320 afs_rwlock_t lock; /* protects cell data */
321 unsigned char cellHandle[16]; /* deterministic handle for this cell */
325 struct cell_name *next;
332 struct cell_alias *next;
338 #define afs_PutCell(cellp, locktype)
340 /* the unixuser flag bit definitions */
341 #define UHasTokens 1 /* are the st and ct fields valid (ever set)? */
342 #define UTokensBad 2 /* are tokens bad? */
343 #define UPrimary 4 /* on iff primary identity */
344 #define UNeedsReset 8 /* needs afs_ResetAccessCache call done */
345 #define UPAGCounted 16 /* entry seen during PAG search (for stats) */
346 #define UNFSGetCreds 32 /* getting creds for NFS client */
347 /* A flag used by afs_GCPAGs to keep track of
348 * which entries in afs_users need to be deleted.
349 * The lifetime of its presence in the table is the
350 * lifetime of the afs_GCPAGs function.
352 #define TMP_UPAGNotReferenced 128
354 /* unixuser notify events */
355 #define UTokensObtained 1
356 #define UTokensDropped 2
358 /* values for afs_gcpags */
359 enum { AFS_GCPAGS_NOTCOMPILED = 0, AFS_GCPAGS_OK =
360 1, AFS_GCPAGS_USERDISABLED, AFS_GCPAGS_EPROC0, AFS_GCPAGS_EPROCN,
361 AFS_GCPAGS_EEQPID, AFS_GCPAGS_EINEXACT, AFS_GCPAGS_EPROCEND,
362 AFS_GCPAGS_EPROCWALK, AFS_GCPAGS_ECREDWALK, AFS_GCPAGS_EPIDCHECK,
363 AFS_GCPAGS_ENICECHECK
366 extern afs_int32 afs_gcpags;
367 extern afs_int32 afs_gcpags_procsize;
368 extern afs_int32 afs_bkvolpref;
369 extern char afs_cachebasedir[1024];
370 extern afs_int32 afs_numcachefiles;
371 extern afs_int32 afs_numfilesperdir;
376 struct ClearToken clearToken;
380 struct rxkadToken rxkad;
384 struct tokenJar *next;
386 union tokenUnion content;
390 struct unixuser *next; /* next hash pointer */
391 afs_int32 uid; /* search based on uid and cell */
393 afs_int32 viceId; /* Optional viced ID corresponding to current tokens */
394 short refCount; /* reference count for allocation */
395 char states; /* flag info */
396 afs_int32 tokenTime; /* last time tokens were set, used for timing out conn data */
397 struct tokenJar *tokens;
398 struct afs_exporter *exporter; /* more info about the exporter for the remote user */
399 void *cellinfo; /* pointer to cell info (PAG manager only) */
403 #define CVEC_LEN 3 /* per-user connection pool */
405 struct sa_conn_vector;
406 typedef struct sa_conn_vector * p_sa_conn_vector; /* forward decl */
408 #define CONN_REPLICATED 0x1
413 char forceConnectFS; /* Should we try again with these tokens? */
414 struct rx_connection *id; /* RPC connid */
415 struct sa_conn_vector *parent; /* the con_vector which contains us */
418 /* An sa_conn_vector replaces the erstwhile list of conn
419 structures maintained by the cache manager. The sa_conn_vector
420 contains a C array of connections which, if non-zero, represent
421 connections to AFS servers.
424 struct sa_conn_vector {
425 /* linked-list machinery */
426 struct sa_conn_vector *next;
428 /* AFS conn-identifying info */
429 struct unixuser *user; /* user validated with respect to. */
430 struct srvAddr *srvr; /* server associated with this conn */
431 short refCount; /* reference count for allocation */
432 unsigned short port; /* port associated with this connection */
435 /* next connection to return when all in cvec are fully utilized */
438 /* connections vector */
439 struct afs_conn cvec[CVEC_LEN];
444 /* Fid comparison routines */
445 #define FidCmp(a,b) ((a)->Fid.Unique != (b)->Fid.Unique \
446 || (a)->Fid.Vnode != (b)->Fid.Vnode \
447 || (a)->Fid.Volume != (b)->Fid.Volume \
448 || (a)->Cell != (b)->Cell)
450 #define FidMatches(afid,tvc) ((tvc)->f.fid.Fid.Vnode == (afid)->Fid.Vnode && \
451 (tvc)->f.fid.Fid.Volume == (afid)->Fid.Volume && \
452 (tvc)->f.fid.Cell == (afid)->Cell && \
453 ( (tvc)->f.fid.Fid.Unique == (afid)->Fid.Unique || \
454 (!(afid)->Fid.Unique && ((tvc)->f.states & CUnique))))
458 #define SRVADDR_ISDOWN 0x20 /* same as SRVR_ISDOWN */
459 #define SRVADDR_NOUSE 0x40 /* Don't use this srvAddr */
462 struct srvAddr *next_bkt; /* next item in hash bucket */
463 struct srvAddr *next_sa; /* another interface on same host */
464 struct server *server; /* back to parent */
465 struct sa_conn_vector *conns; /* All user connections to this server */
466 struct afs_conn *natping;
467 afs_int32 sa_ip; /* Host addr in network byte order */
468 u_short sa_iprank; /* indiv ip address priority */
469 u_short sa_portal; /* port addr in network byte order */
474 * Values used in the flags field of the server structure below.
476 * AFS_SERVER_FLAG_ACTIVATED Has the server ever had a user connection
477 * associated with it?
479 #define AFS_SERVER_FLAG_ACTIVATED 0x01
480 #define SNO_LHOSTS 0x04
481 #define SYES_LHOSTS 0x08
482 #define SVLSRV_UUID 0x10
483 #define SRVR_ISDOWN 0x20
484 #define SRVR_MULTIHOMED 0x40
485 #define SRVR_ISGONE 0x80
486 #define SNO_INLINEBULK 0x100
487 #define SNO_64BIT 0x200
488 #define SCAPS_KNOWN 0x400
490 #define SRV_CAPABILITIES(ts) \
491 { if ( !(ts->flags & SCAPS_KNOWN)) afs_GetCapabilities(ts); ts->capabilities; }
493 #define afs_serverSetNo64Bit(s) (((struct sa_conn_vector*)(s)->parent)->srvr->server->flags |= SNO_64BIT)
494 #define afs_serverHasNo64Bit(s) (((struct sa_conn_vector*)(s)->parent)->srvr->server->flags & SNO_64BIT)
500 afs_int32 addr_uniquifier;
504 struct srvAddr haddr;
507 #define sr_uuid _suid._srvUuid.suuid
508 #define sr_addr_uniquifier _suid._srvUuid.addr_uniquifier
509 #define sr_host _suid._srvId.haddr.ip
510 #define sr_portal _suid._srvId.haddr.portal
511 #define sr_rank _suid._srvId.haddr.ip_rank
512 #define sr_flags _suid._srvId.haddr.flags
513 #define sr_conns _suid._srvId.haddr.conns
514 struct server *next; /* Ptr to next server in hash chain */
515 struct cell *cell; /* Cell in which this host resides */
516 struct afs_cbr *cbrs; /* Return list of callbacks */
517 afs_int32 activationTime; /* Time when this record was first activated */
518 afs_int32 lastDowntimeStart; /* Time when last downtime incident began */
519 afs_int32 numDowntimeIncidents; /* # (completed) downtime incidents */
520 afs_int32 sumOfDowntimes; /* Total downtime experienced, in seconds */
521 struct srvAddr *addr;
522 afs_uint32 flags; /* Misc flags */
523 afs_int32 capabilities;
526 #define afs_PutServer(servp, locktype)
528 /* structs for some pioctls - these are (or should be)
536 struct sprefrequest_33 {
537 unsigned short offset;
538 unsigned short num_servers;
542 struct sprefrequest { /* new struct for 3.4 */
543 unsigned short offset;
544 unsigned short num_servers;
545 unsigned short flags;
550 unsigned short next_offset;
551 unsigned short num_servers;
552 struct spref servers[1]; /* we overrun this array intentionally... */
556 unsigned short flags;
557 unsigned short num_servers;
558 struct spref servers[1]; /* we overrun this array intentionally... */
560 /* struct for GAG pioctl
563 afs_uint32 showflags, logflags, logwritethruflag, spare[3];
564 unsigned char spare2[128];
568 #define logwritethruON 1
571 afs_int32 rx_initReceiveWindow, rx_maxReceiveWindow, rx_initSendWindow,
572 rx_maxSendWindow, rxi_nSendFrags, rxi_nRecvFrags, rxi_OrphanFragSize;
573 afs_int32 rx_maxReceiveSize, rx_MyMaxSendSize;
574 afs_uint32 spare[21];
577 /* struct for checkservers */
588 /* state bits for volume */
589 #define VRO 1 /* volume is readonly */
590 #define VRecheck 2 /* recheck volume info with server */
591 #define VBackup 4 /* is this a backup volume? */
592 #define VForeign 8 /* this is a non-afs volume */
593 #define VPartVisible 16 /* Volume's partition is visible on the client */
594 #define VHardMount 32 /* we are hard-mount waiting for the vol */
596 enum repstate { not_busy, end_not_busy = 6, rd_busy, rdwr_busy, offline };
599 /* One structure per volume, describing where the volume is located
600 * and where its mount points are. */
601 struct volume *next; /* Next volume in hash list. */
602 afs_int32 cell; /* the cell in which the volume resides */
603 afs_rwlock_t lock; /* the lock for this structure */
604 afs_int32 volume; /* This volume's ID number. */
605 char *name; /* This volume's name, or 0 if unknown */
606 struct server *serverHost[AFS_MAXHOSTS]; /* servers serving this volume */
607 enum repstate status[AFS_MAXHOSTS]; /* busy, offline, etc */
608 struct VenusFid dotdot; /* dir to access as .. */
609 struct VenusFid mtpoint; /* The mount point for this volume. */
610 afs_int32 rootVnode, rootUnique; /* Volume's root fid */
613 afs_int32 rwVol; /* For r/o vols, original read/write volume. */
614 afs_int32 setupTime; /* time volume was setup from vldb info */
615 afs_int32 vtix; /* volume table index */
616 afs_int32 copyDate; /* copyDate field, for tracking vol releases */
617 afs_int32 expireTime; /* for per-volume callbacks... */
618 short refCount; /* reference count for allocation */
619 char states; /* here for alignment reasons */
622 #define afs_PutVolume(av, locktype) ((av)->refCount--)
624 /* format of an entry in volume info file */
626 afs_int32 cell; /* cell for this entry */
627 afs_int32 volume; /* volume */
628 afs_int32 next; /* has index */
629 struct VenusFid dotdot; /* .. value */
630 struct VenusFid mtpoint; /* mt point's fid */
631 afs_int32 rootVnode, rootUnique; /* Volume's root fid */
635 struct SimpleLocks *next;
637 afs_int32 boff, eoff;
639 #if defined(AFS_AIX32_ENV) || defined(AFS_SUN5_ENV) || defined(AFS_SGI_ENV)
644 /* vcache state bits */
645 #define CStatd 0x00000001 /* has this file ever been stat'd? */
646 #define CBackup 0x00000002 /* file is on a backup volume */
647 #define CRO 0x00000004 /* is it on a read-only volume */
648 #define CMValid 0x00000008 /* is the mount point info valid? */
649 #define CCore 0x00000010 /* storing a core file, needed since we don't get an open */
650 #define CDirty 0x00000020 /* file has been modified since first open (... O_RDWR) */
651 #define CSafeStore 0x00000040 /* close must wait for store to finish (should be in fd) */
652 #define CMAPPED 0x00000080 /* Mapped files */
653 #define CNSHARE 0x00000100 /* support O_NSHARE semantics */
654 #define CLied 0x00000200
655 #define CTruth 0x00000400
657 #if defined(AFS_DARWIN80_ENV)
658 #define CDeadVnode 0x00000800
659 #elif defined(AFS_DARWIN_ENV)
660 #define CUBCinit 0x00000800
662 #define CWRITE_IGN 0x00000800 /* Next OS hack only */
665 #define CUnique 0x00001000 /* vc's uniquifier - latest unifiquier for fid */
666 #define CForeign 0x00002000 /* this is a non-afs vcache */
667 #define CReadDir 0x00004000 /* readdir in progress */
668 #define CUnlinked 0x00010000
669 #define CBulkStat 0x00020000 /* loaded by a bulk stat, and not ref'd since */
670 #define CUnlinkedDel 0x00040000
671 #define CVFlushed 0x00080000
672 #if defined(AFS_SGI_ENV)
673 #define CWritingUFS 0x00200000 /* to detect vm deadlock - used by sgi */
674 #elif defined(AFS_DARWIN80_ENV)
675 #define CEvent 0x00200000 /* to preclude deadlock when sending events */
677 #define CCreating 0x00400000 /* avoid needless store after open truncate */
678 #define CPageHog 0x00800000 /* AIX - dumping large cores is a page hog. */
679 #define CCorrupt 0x01000000 /* corrupt dir already reported */
680 #define CDCLock 0x02000000 /* Vnode lock held over call to GetDownD */
681 #define CBulkFetching 0x04000000 /* stats are being fetched by bulk stat */
682 #define CExtendedFile 0x08000000 /* extended file via ftruncate call. */
683 #define CVInit 0x10000000 /* being initialized */
684 #define CMetaDirty 0x20000000 /* vnode meta-data needs to be flushed */
685 #define CPartVisible 0x40000000 /* fileserver partition visible on client */
687 /* vcache vstate bits */
688 #define VRevokeWait 0x1
689 #define VPageCleaning 0x2 /* Solaris - Cache Trunc Daemon sez keep out */
691 /* Dirty disconnected vcache flags. */
692 #define VDisconSetTime 0x00000001 /* set time. */
693 #define VDisconSetMode 0x00000002 /* set mode. */
694 /* XXX: to be continued ? */
695 #define VDisconTrunc 0x00000020 /* truncate file. */
696 #define VDisconSetAttrMask 0x0000003F /* Masks for setattr ops. */
697 #define VDisconWriteClose 0x00000400 /* Write op on file close. */
698 #define VDisconWriteFlush 0x00000800 /* Write op on normal fsync/flush. */
699 #define VDisconWriteOsiFlush 0x00001000 /* Write op on osi flush. */
701 #define VDisconRemove 0x00002000 /* Remove vnop. */
702 #define VDisconCreate 0x00004000 /* Create vnop. */
703 #define VDisconCreated 0x00008000 /* A file that was created during
704 this resync operation */
705 #define VDisconRename 0x00010000 /* Rename vnop. */
706 #define VDisconRenameSameDir 0x00020000 /* Rename in same dir. */
708 /*... to be continued ... */
710 /* vcache (file) cachingStates bits */
711 #define FCSDesireBypass 0x1 /* This file should bypass the cache */
712 #define FCSBypass 0x2 /* This file is currently NOT being cached */
713 #define FCSManuallySet 0x4 /* The bypass flags were set, or reset, manually (via pioctl)
714 and should not be overridden by the file's name */
716 /* Flag values used by the Transition routines */
717 #define TRANSChangeDesiredBit 0x1 /* The Transition routine should set or
718 * reset the FCSDesireBypass bit */
719 #define TRANSVcacheIsLocked 0x2 /* The Transition routine does not need to
720 * lock vcache (it's already locked) */
721 #define TRANSSetManualBit 0x4 /* The Transition routine should set FCSManuallySet so that
722 * filename checking does not override pioctl requests */
725 #if defined(AFS_XBSD_ENV) || defined(AFS_DARWIN_ENV)
726 #define vrefCount v->v_usecount
727 #elif defined(AFS_SUN511_ENV)
728 # define vrefCount v->v_count
730 #define vrefCount v.v_count
731 #endif /* AFS_XBSD_ENV */
733 #if defined(AFS_DARWIN80_ENV)
734 #define VREFCOUNT_GT(v, y) vnode_isinuse(AFSTOV(v), (y))
735 #elif defined(AFS_FBSD_ENV)
736 #define VREFCOUNT(v) (vrefcnt(AFSTOV(v)))
737 #define VREFCOUNT_GT(v, y) (vrefcnt(AFSTOV(v)) > y)
738 #elif defined(AFS_XBSD_ENV) || defined(AFS_DARWIN_ENV)
739 #define VREFCOUNT(v) ((v)->vrefCount)
740 #define VREFCOUNT_GT(v, y) (AFSTOV(v)->v_usecount > (y))
741 #elif defined(AFS_LINUX24_ENV)
742 #define VREFCOUNT(v) atomic_read(&(AFSTOV(v)->v_count))
743 #define VREFCOUNT_GT(v, y) (VREFCOUNT(v)>y)
744 #define VREFCOUNT_SET(v, c) atomic_set(&(AFSTOV(v)->v_count), c)
745 #define VREFCOUNT_DEC(v) atomic_dec(&(AFSTOV(v)->v_count))
746 #define VREFCOUNT_INC(v) atomic_inc(&(AFSTOV(v)->v_count))
748 #define VREFCOUNT(v) ((v)->vrefCount)
749 #define VREFCOUNT_GT(v,y) ((v)->vrefCount > (y))
750 #define VREFCOUNT_SET(v, c) (v)->vrefCount = c
751 #define VREFCOUNT_DEC(v) (v)->vrefCount--
752 #define VREFCOUNT_INC(v) (v)->vrefCount++
753 #define d_unhash(d) list_empty(&(d)->d_hash)
754 #define dget_locked(d) dget(d)
757 #define AFS_MAXDV 0x7fffffff /* largest dataversion number */
758 #ifdef AFS_64BIT_CLIENT
759 #define AFS_NOTRUNC 0x7fffffffffffffffLL /* largest positive int64 number */
760 #else /* AFS_64BIT_CLIENT */
761 #define AFS_NOTRUNC 0x7fffffff /* largest dataversion number */
762 #endif /* AFS_64BIT_CLIENT */
764 extern afs_int32 vmPageHog; /* counter for # of vnodes which are page hogs. */
766 #if defined(AFS_DARWIN80_ENV)
767 #define VTOAFS(v) ((struct vcache *)vnode_fsnode((v)))
768 #define AFSTOV(vc) ((vc)->v)
769 #elif defined(AFS_NBSD40_ENV)
771 struct genfs_node gfsn;
772 struct vcache *afsvc;
774 #define VTOAFS(v) ((((struct nbvdata *)((v)->v_data)))->afsvc)
775 #define AFSTOV(vc) ((vc)->v)
776 #elif defined(AFS_XBSD_ENV) || defined(AFS_DARWIN_ENV) || defined(AFS_SUN511_ENV) || (defined(AFS_LINUX22_ENV) && !defined(STRUCT_SUPER_OPERATIONS_HAS_ALLOC_INODE))
777 #define VTOAFS(v) ((struct vcache *)(v)->v_data)
778 #define AFSTOV(vc) ((vc)->v)
780 #define VTOAFS(V) ((struct vcache *)(V))
781 #define AFSTOV(V) (&(V)->v)
789 /* VCache elements which are kept on disk, and in the kernel */
794 afs_hyper_t DataVersion;
798 afs_uint16 Mode; /* XXXX Should be afs_int32 XXXX */
799 afs_uint16 LinkCount;
800 #ifdef AFS_DARWIN80_ENV
803 /* vnode type is in v.v_type */
806 struct afs_vnuniq parent;
808 /*! Truncate file to this position at the next store */
811 /*! System:AnyUser's access to this. */
817 /*! Disconnected flags for this vcache element. */
818 afs_uint32 ddirty_flags;
819 /*! Shadow vnode + unique keep the shadow dir location. */
820 struct afs_vnuniq shadow;
821 /*! The old parent FID for renamed vnodes */
822 struct afs_vnuniq oldParent;
825 /* Values for 'mvstat' in struct vcache */
826 #define AFS_MVSTAT_FILE (0x0) /* regular file or directory */
827 #define AFS_MVSTAT_MTPT (0x1) /* mountpoint */
828 #define AFS_MVSTAT_ROOT (0x2) /* volume root dir */
832 * This is for the multiPage field in struct vcache. Each one of these
833 * represents an outstanding getpage request that is larger than a single page.
834 * Recording these is necessary to prevent afs_GetOnePage from trying to evict
835 * a dcache entry that an earlier afs_GetOnePage call got in the same getpage
836 * request. See osi_VM_MultiPageConflict and afs_getpage.
838 struct multiPage_range {
840 offset_t off; /**< offset of getpage request */
841 u_int len; /**< length of getpage request */
845 /* INVARIANTs: (vlruq.next != NULL) == (vlruq.prev != NULL)
846 * nextfree => !vlruq.next && ! vlruq.prev
847 * !(avc->nextfree) && !avc->vlruq.next => (FreeVCList == avc->nextfree)
850 #if defined(AFS_XBSD_ENV) || defined(AFS_DARWIN_ENV) || defined(AFS_SUN511_ENV) || (defined(AFS_LINUX22_ENV) && !defined(STRUCT_SUPER_OPERATIONS_HAS_ALLOC_INODE))
853 struct vnode v; /* Has reference count in v.v_count */
855 struct afs_q vlruq; /* lru q next and prev */
856 #if !defined(AFS_LINUX22_ENV)
857 struct vcache *nextfree; /* next on free list (if free) */
859 struct vcache *hnext; /* Hash next */
860 struct afs_q vhashq; /* Hashed per-volume list */
861 /*! Queue of dirty vcaches. Lock with afs_disconDirtyLock */
863 /*! Queue of vcaches with shadow entries. Lock with afs_disconDirtyLock */
864 struct afs_q shadowq;
865 /*! Queue of vcaches with dirty metadata. Locked by afs_xvcdirty */
866 struct afs_q metadirty;
867 /*! Vcaches slot number in the disk backup. Protected by tvc->lock */
870 afs_rwlock_t lock; /* The lock on the vcache contents. */
871 #if defined(AFS_SUN5_ENV)
872 /* Lock used to protect the activeV, multipage, and vstates fields.
873 * Do not try to get the vcache lock when the vlock is held */
875 #endif /* defined(AFS_SUN5_ENV) */
876 #if defined(AFS_SUN5_ENV)
883 #if defined(AFS_AIX51_ENV)
891 int ownslock; /* pid of owner of excl lock, else 0 - defect 3083 */
893 #ifdef AFS_DARWIN80_ENV
895 #elif defined(AFS_DARWIN_ENV)
896 struct lock__bsd__ rwlock;
899 #if !defined(AFS_DFBSD_ENV) && !defined(AFS_NBSD_ENV)
905 char *silly_name; /* For sillyrenamed regular files, the silly
906 * name the file was renamed to. */
907 struct VenusFid *target_root; /* For mountpoints, the fid of the root dir
908 * in the target volume. */
909 struct VenusFid *parent; /* For root dir vcaches, the fid of the
913 char *linkData; /* Link data if a symlink. */
914 afs_hyper_t flushDV; /* data version last flushed from text */
915 afs_hyper_t mapDV; /* data version last flushed from map */
916 struct server *callback; /* The callback host, if any */
917 afs_uint32 cbExpires; /* time the callback expires */
918 struct afs_q callsort; /* queue in expiry order, sort of */
919 struct axscache *Access; /* a list of cached access bits */
920 afs_int32 last_looker; /* pag/uid from last lookup here */
921 #if defined(AFS_SUN5_ENV)
923 #endif /* defined(AFS_SUN5_ENV) */
924 struct SimpleLocks *slocks;
925 short opens; /* The numbers of opens, read or write, on this file. */
926 short execsOrWriters; /* The number of execs (if < 0) or writers (if > 0) of
928 short flockCount; /* count of flock readers, or -1 if writer */
929 char mvstat; /* see the AFS_MVSTAT_* constants */
931 char cachingStates; /* Caching policies for this file */
932 afs_uint32 cachingTransitions; /* # of times file has flopped between caching and not */
934 #if defined(AFS_LINUX24_ENV)
935 off_t next_seq_offset; /* Next sequential offset (used by prefetch/readahead) */
936 #elif defined(AFS_SUN5_ENV) || defined(AFS_SGI65_ENV)
937 off_t next_seq_blk_offset; /* accounted in blocks for Solaris & IRIX */
940 #if defined(AFS_SUN5_ENV)
941 afs_uint32 vstates; /* vstate bits */
942 #endif /* defined(AFS_SUN5_ENV) */
943 struct dcache *dchint;
944 struct dcache *dcreaddir; /* dcache for in-progress readdir */
945 unsigned int readdir_pid; /* pid of the thread in readdir */
946 #if defined(AFS_SGI_ENV)
947 daddr_t lastr; /* for read-ahead */
949 uint64_t vc_rwlockid; /* kthread owning rwlock */
951 short vc_rwlockid; /* pid of process owning rwlock */
953 short vc_locktrips; /* # of rwlock reacquisitions */
954 sema_t vc_rwlock; /* vop_rwlock for afs */
955 pgno_t mapcnt; /* # of pages mapped */
956 struct cred *cred; /* last writer's cred */
958 struct bhv_desc vc_bhv_desc; /* vnode's behavior data. */
960 #endif /* AFS_SGI_ENV */
961 #if defined(AFS_LINUX26_ENV)
962 cred_t *cred; /* last writer's cred */
964 #ifdef AFS_LINUX24_ENV
965 struct dentry *target_link; /* dentry we prefer, when we are redirecting
966 * all requests due to duplicate dentry aliases.
967 * See LINUX/osi_vnodeops.c. Note that this is
968 * NOT an actual reference to a dentry, so this
969 * pointer MUST NOT be dereferenced on its own. */
971 afs_int32 vc_error; /* stash write error for this vnode. */
972 int xlatordv; /* Used by nfs xlator */
974 int asynchrony; /* num kbytes to store behind */
976 struct afs_q multiPage; /* list of multiPage_range structs */
978 int protocol; /* RX_FILESERVER, RX_OSD, ... defined in afsint.xg */
979 #if !defined(UKERNEL)
980 void *vpacRock; /* used to read or write in visible partitions */
982 afs_uint32 lastBRLWarnTime; /* last time we warned about byte-range locks */
983 #ifdef AFS_LINUX26_ENV
984 spinlock_t pagewriter_lock;
985 struct list_head pagewriters; /* threads that are writing vm pages */
989 #ifdef AFS_LINUX26_ENV
991 struct list_head link;
996 #define DONT_CHECK_MODE_BITS 0
997 #define CHECK_MODE_BITS 1
998 #define CMB_ALLOW_EXEC_AS_READ 2 /* For the NFS xlator */
1000 #if defined(AFS_SGI_ENV)
1001 #define AVCRWLOCK(avc) (valusema(&(avc)->vc_rwlock) <= 0)
1003 /* SGI vnode rwlock macros and flags. */
1004 #ifndef AFS_SGI62_ENV
1005 /* The following are defined here. SGI 6.2 declares them in vnode.h */
1006 #define VRWLOCK_READ 0
1007 #define VRWLOCK_WRITE 1
1008 #define VRWLOCK_WRITE_DIRECT 2
1011 #ifdef AFS_SGI53_ENV
1012 #ifdef AFS_SGI62_ENV
1013 #define AFS_RWLOCK_T vrwlock_t
1015 #define AFS_RWLOCK_T int
1016 #endif /* AFS_SGI62_ENV */
1017 #ifdef AFS_SGI64_ENV
1018 #include <ksys/behavior.h>
1019 #define AFS_RWLOCK(V,F) \
1020 afs_rwlock(&VTOAFS(V)->vc_bhv_desc, (F));
1021 #define AFS_RWUNLOCK(V,F) \
1022 afs_rwunlock(&VTOAFS(V)->vc_bhv_desc, (F));
1025 #define AFS_RWLOCK(V,F) afs_rwlock((vnode_t *)(V), (F) )
1026 #define AFS_RWUNLOCK(V,F) afs_rwunlock((vnode_t *)(V), (F) )
1028 #else /* AFS_SGI53_ENV */
1029 #define AFS_RWLOCK(V,F) afs_rwlock((V))
1030 #define AFS_RWUNLOCK(V,F) afs_rwunlock((V))
1031 #endif /* AFS_SGI53_ENV */
1032 #endif /* AFS_SGI_ENV */
1035 struct VenusFid fid;
1036 afs_hyper_t DataVersion;
1038 afs_int32 parentVnode;
1039 afs_int32 parentUnique;
1040 afs_hyper_t flushDV;
1043 afs_int32 randomUid[CPSIZE];
1044 afs_int32 callback; /* Now a pointer to 'server' struct */
1045 afs_int32 cbExpires;
1046 afs_int32 randomAccess[CPSIZE];
1047 afs_int32 anyAccess;
1049 short execsOrWriters;
1056 afs_int32 callerAccess;
1057 afs_int32 cbExpires;
1058 afs_int32 anyAccess;
1067 /* CM inititialization parameters. What CM actually used after calculations
1068 * based on passed in arguments.
1070 #define CMI_VERSION 1 /* increment when adding new fields. */
1071 struct cm_initparams {
1073 int cmi_nChunkFiles;
1074 int cmi_nStatCaches;
1075 int cmi_nDataCaches;
1076 int cmi_nVolumeCaches;
1077 int cmi_firstChunkSize;
1078 int cmi_otherChunkSize;
1079 int cmi_cacheSize; /* The original cache size, in 1K blocks. */
1080 unsigned cmi_setTime:1;
1081 unsigned cmi_memCache:1;
1082 int spare[16 - 9]; /* size of struct is 16 * 4 = 64 bytes */
1086 /*----------------------------------------------------------------------
1087 * AFS Data cache definitions
1089 * Each entry describes a Unix file on the local disk that is
1090 * is serving as a cached copy of all or part of a Vice file.
1091 * Entries live in circular queues for each hash table slot
1093 * Which queue is this thing in? Good question.
1094 * A struct dcache entry is in the freeDSlot queue when not associated with a cache slot (file).
1095 * Otherwise, it is in the DLRU queue. The freeDSlot queue uses the lruq.next field as
1096 * its "next" pointer.
1098 * Cache entries in the DLRU queue are either associated with vice files, in which case
1099 * they are hashed by afs_dvnextTbl and afs_dcnextTbl pointers, or they are in the freeDCList
1100 * and are not associated with any vice file. This last list uses the afs_dvnextTbl pointer for
1101 * its "next" pointer.
1102 *----------------------------------------------------------------------*/
1104 #define NULLIDX (-1) /* null index definition */
1105 /* struct dcache states bits */
1109 #define DWriting 8 /* file being written (used for cache validation) */
1111 /* dcache data flags */
1112 #define DFEntryMod 0x02 /* has entry itself been modified? */
1113 #define DFFetching 0x04 /* file is currently being fetched */
1115 /* dcache meta flags */
1116 #define DFNextStarted 0x01 /* next chunk has been prefetched already */
1117 #define DFFetchReq 0x10 /* someone is waiting for DFFetching to go on */
1120 /* flags in afs_indexFlags array */
1121 #define IFEverUsed 1 /* index entry has >= 1 byte of data */
1122 #define IFFree 2 /* index entry in freeDCList */
1123 #define IFDataMod 4 /* file needs to be written out */
1124 #define IFFlag 8 /* utility flag */
1125 #define IFDirtyPages 16 /* Solaris-only. contains dirty pages */
1126 #define IFAnyPages 32
1127 #define IFDiscarded 64 /* index entry in discardDCList */
1129 #ifdef AFS_DARWIN100_ENV
1130 typedef user_addr_t iparmtype; /* 64 bit */
1131 typedef user_addr_t uparmtype; /* 64 bit */
1133 typedef char * uparmtype;
1134 #ifdef AFS_SGI65_ENV
1135 typedef afs_uint32 iparmtype;
1137 typedef long iparmtype;
1141 #if SIZEOF_VOID_P == SIZEOF_UNSIGNED_INT
1142 # define uintptrsz unsigned int
1143 #elif SIZEOF_VOID_P == SIZEOF_UNSIGNED_LONG
1144 # define uintptrsz unsigned long
1145 #elif SIZEOF_VOID_P == SIZEOF_UNSIGNED_LONG_LONG
1146 # define uintptrsz unsigned long long
1148 # error "Unable to determine casting for pointers"
1152 uparmtype in; /* input buffer */
1153 uparmtype out; /* output buffer */
1154 short in_size; /* Size of input buffer <= 2K */
1155 short out_size; /* Maximum size of output buffer, <= 2K */
1159 * This version of afs_ioctl is required to pass in 32 bit user space
1160 * pointers into a 64 bit kernel.
1163 struct afs_ioctl32 {
1171 /* CacheItems file has a header of type struct afs_fheader
1172 * (keep aligned properly). Since we already have sgi_62 clients running
1173 * with a 32 bit inode, a change is required to the header so that
1174 * they can distinguish the old 32 bit inode CacheItems file and zap it
1175 * instead of using it.
1177 struct afs_fheader {
1178 #define AFS_FHMAGIC 0x7635abaf /* uses version number */
1180 #define AFS_CI_VERSION 4
1182 afs_uint32 dataSize;
1183 afs_int32 firstCSize;
1184 afs_int32 otherCSize;
1187 #if defined(AFS_CACHE_VNODE_PATH)
1188 typedef char *afs_ufs_dcache_id_t;
1189 #elif defined(AFS_SGI61_ENV) || defined(AFS_SUN5_64BIT_ENV)
1190 /* Using ino64_t here so that user level debugging programs compile
1191 * the size correctly.
1193 typedef ino64_t afs_ufs_dcache_id_t;
1194 #elif defined(AFS_LINUX26_ENV)
1195 #define MAX_FH_LEN 10
1197 #if defined(NEW_EXPORT_OPS)
1200 __u32 raw[MAX_FH_LEN];
1201 } afs_ufs_dcache_id_t;
1202 extern int cache_fh_type;
1203 extern int cache_fh_len;
1204 #elif defined(AFS_LINUX_64BIT_KERNEL) && !defined(AFS_S390X_LINUX24_ENV)
1205 typedef long afs_ufs_dcache_id_t;
1206 #elif defined(AFS_AIX51_ENV) || defined(AFS_HPUX1123_ENV)
1207 typedef ino_t afs_ufs_dcache_id_t;
1209 typedef afs_int32 afs_ufs_dcache_id_t;
1212 typedef afs_int32 afs_mem_dcache_id_t;
1215 afs_ufs_dcache_id_t ufs;
1216 afs_mem_dcache_id_t mem;
1219 /* it does not compile outside kernel */
1221 afs_int32 fid; /* is adc->index, the cache file number */
1222 afs_dcache_id_t inode; /* is adc->f.inode, the inode number of the cac\
1225 afs_int32 accesstime;
1226 struct buffer *hashNext;
1231 afs_rwlock_t lock; /* the lock for this structure */
1234 /* kept on disk and in dcache entries */
1236 struct VenusFid fid; /* Fid for this file */
1237 afs_int32 modTime; /* last time this entry was modified */
1238 afs_hyper_t versionNo; /* Associated data version number */
1239 afs_int32 chunk; /* Relative chunk number */
1240 afs_dcache_id_t inode; /* Unix inode for this chunk */
1241 afs_int32 chunkBytes; /* Num bytes in this chunk */
1242 char states; /* Has this chunk been modified? */
1245 /* magic numbers to specify the cache type */
1247 #define AFS_FCACHE_TYPE_UFS 0x0
1248 #define AFS_FCACHE_TYPE_MEM 0x1
1249 #define AFS_FCACHE_TYPE_NFS 0x2
1250 #define AFS_FCACHE_TYPE_EPI 0x3
1252 /* kept in memory */
1254 struct afs_q lruq; /* Free queue for in-memory images */
1255 struct afs_q dirty; /* Queue of dirty entries that need written */
1256 afs_rwlock_t lock; /* Protects validPos, some f */
1257 afs_rwlock_t tlock; /* Atomizes updates to refCount */
1258 afs_rwlock_t mflock; /* Atomizes accesses/updates to mflags */
1259 afs_size_t validPos; /* number of valid bytes during fetch */
1260 afs_int32 index; /* The index in the CacheInfo file */
1261 short refCount; /* Associated reference count. */
1262 char dflags; /* Data flags */
1263 char mflags; /* Meta flags */
1264 struct fcache f; /* disk image */
1265 afs_int32 bucket; /* which bucket these dcache entries are in */
1269 * dcache.lock protects the actual contents of the cache file (in
1270 * f.inode), subfields of f except those noted below, dflags and
1273 * dcache.tlock is used to make atomic updates to refCount. Zero
1274 * refCount dcache entries are protected by afs_xdcache instead of
1277 * dcache.mflock is used to access and update mflags. It cannot be
1278 * held without holding the corresponding dcache.lock. Updating
1279 * mflags requires holding dcache.lock(R) and dcache.mflock(W), and
1280 * checking for mflags requires dcache.lock(R) and dcache.mflock(R).
1281 * Note that dcache.lock(W) gives you the right to update mflags,
1282 * as dcache.mflock(W) can only be held with dcache.lock(R).
1284 * dcache.index, dcache.f.fid, dcache.f.chunk and dcache.f.inode are
1285 * write-protected by afs_xdcache and read-protected by refCount.
1286 * Once an entry is referenced, these values cannot change, and if
1287 * it's on the free list (with refCount=0), it can be reused for a
1288 * different file/chunk. These values can only be written while
1289 * holding afs_xdcache(W) and allocating this dcache entry (thereby
1290 * ensuring noone else has a refCount on it).
1294 /* afs_memcache.c */
1295 struct memCacheEntry {
1296 int size; /* # of valid bytes in this entry */
1297 int dataSize; /* size of allocated data area */
1298 afs_lock_t afs_memLock;
1299 char *data; /* bytes */
1302 struct afs_FetchOutput {
1303 struct AFSVolSync tsync;
1304 struct AFSFetchStatus OutStatus;
1305 struct AFSCallBack CallBack;
1308 /* macro to mark a dcache entry as bad */
1311 (x)->f.fid.Fid.Unique = 0; \
1312 afs_indexUnique[(x)->index] = 0; \
1313 (x)->dflags |= DFEntryMod; \
1316 /* FakeOpen and Fake Close used to be real subroutines. They're only used in
1317 * sun_subr and afs_vnodeops, and they're very frequently called, so I made
1318 * them into macros. They do:
1319 * FakeOpen: fake the file being open for writing. avc->lock must be held
1320 * in write mode. Having the file open for writing is like having a DFS
1321 * write-token: you're known to have the best version of the data around,
1322 * and so the CM won't let it be overwritten by random server info.
1323 * FakeClose: undo the effects of FakeOpen, noting that we want to ensure
1324 * that a real close eventually gets done. We use CCore to achieve this if
1325 * we would end up closing the file. avc->lock must be held in write mode */
1327 #ifdef AFS_AIX_IAUTH_ENV
1328 #define CRKEEP(V, C) (V)->linkData = (char*)crdup((C))
1330 #define CRKEEP(V, C) crhold((C)); (V)->linkData = (char*)(C)
1333 #define afs_FakeOpen(avc) { avc->opens++; avc->execsOrWriters++; }
1334 #define afs_FakeClose(avc, acred) \
1335 { if (avc->execsOrWriters == 1) { \
1336 /* we're the last writer, just use CCore flag */ \
1337 avc->f.states |= CCore; /* causes close to be called later */ \
1339 /* The cred and vnode holds will be released in afs_FlushActiveVcaches */ \
1340 osi_Assert(osi_vnhold(avc) == 0); /* So it won't disappear */ \
1341 CRKEEP(avc, acred); /* Should use a better place for the creds */ \
1344 /* we're not the last writer, let the last one do the store-back for us */ \
1346 avc->execsOrWriters--; \
1350 #define AFS_ZEROS 64 /* zero buffer */
1352 #define afs_DirtyPages(avc) ((avc)->f.states & CDirty)
1354 #define afs_InReadDir(avc) (((avc)->f.states & CReadDir) && (avc)->readdir_pid == MyPidxx2Pid(MyPidxx))
1356 extern struct dcache **afs_indexTable; /*Pointers to in-memory dcache entries */
1357 extern afs_int32 *afs_indexUnique; /*dcache entry Fid.Unique */
1358 extern afs_int32 *afs_dvnextTbl; /*Dcache hash table links */
1359 extern afs_int32 *afs_dcnextTbl; /*Dcache hash table links */
1360 extern afs_int32 afs_cacheFiles; /*Size of afs_indexTable */
1361 extern afs_int32 afs_cacheBlocks; /*1K blocks in cache */
1362 extern afs_int32 afs_cacheStats; /*Stat entries in cache */
1363 extern afs_uint32 afs_WaitForCacheDrainCount;
1364 extern struct vcache *afs_vhashT[VCSIZE]; /*Stat cache hash table */
1365 extern struct afs_q afs_vhashTV[VCSIZE]; /* cache hash table on volume */
1366 extern afs_int32 afs_initState; /*Initialization state */
1367 extern afs_int32 afs_termState; /* Termination state */
1368 extern struct VenusFid afs_rootFid; /*Root for whole file system */
1369 extern afs_int32 afs_allCBs; /* Count of callbacks */
1370 extern afs_int32 afs_oddCBs; /* Count of odd callbacks */
1371 extern afs_int32 afs_evenCBs; /* Count of even callbacks */
1372 extern afs_int32 afs_allZaps; /* Count of fid deletes */
1373 extern afs_int32 afs_oddZaps; /* Count of odd fid deletes */
1374 extern afs_int32 afs_evenZaps; /* Count of even fid deletes */
1375 extern struct brequest afs_brs[NBRS]; /* request structures */
1377 #define UHash(auid) ((auid) & (NUSERS-1))
1378 #define VHash(avol) ((avol)&(NVOLS-1))
1379 #define SHash(aserv) ((ntohl(aserv)) & (NSERVERS-1))
1380 #define FVHash(acell,avol) (((avol)+(acell)) & (NFENTRIES-1))
1382 /* Performance hack - we could replace VerifyVCache2 with the appropriate
1383 * GetVCache incantation, and could eliminate even this code from afs_UFSRead
1384 * by making intentionally invalidating quick.stamp in the various callbacks
1385 * expiration/breaking code */
1386 #ifdef AFS_DARWIN_ENV
1387 #define afs_VerifyVCache(avc, areq) \
1388 (((avc)->f.states & CStatd) ? (osi_VM_Setup(avc, 0), 0) : \
1389 afs_VerifyVCache2((avc),areq))
1391 #define afs_VerifyVCache(avc, areq) \
1392 (((avc)->f.states & CStatd) ? 0 : afs_VerifyVCache2((avc),areq))
1395 #define DO_STATS 1 /* bits used by FindVCache */
1400 /* values for adown value of afs_LoopServers */
1402 #define AFS_LS_DOWN 1
1403 #define AFS_LS_ALL 2
1405 /* values for flag param of afs_CheckVolumeNames */
1406 #define AFS_VOLCHECK_EXPIRED 0x1 /* volumes whose callbacks have expired */
1407 #define AFS_VOLCHECK_BUSY 0x2 /* volumes which were marked busy */
1408 #define AFS_VOLCHECK_MTPTS 0x4 /* mount point invalidation also */
1409 #define AFS_VOLCHECK_FORCE 0x8 /* do all forcibly */
1413 #define AFS_FSPORT ((unsigned short) htons(7000))
1414 #define AFS_VLPORT ((unsigned short) htons(7003))
1415 #define AFS_RXOSDPORT ((unsigned short) htons(7011))
1417 #define afs_rdwr(avc, uio, rw, io, cred) \
1418 (((rw) == UIO_WRITE) ? afs_write(avc, uio, io, cred, 0) : afs_read(avc, uio, cred, 0))
1419 #define afs_nlrdwr(avc, uio, rw, io, cred) \
1420 (((rw) == UIO_WRITE) ? afs_write(avc, uio, io, cred, 1) : afs_read(avc, uio, cred, 1))
1422 /* Cache size truncation uses the following low and high water marks:
1423 * If the cache is more than 90% full by space (CM_DCACHESPACEFREEPCT) or
1424 * 95% full by # of chunks (CM_DCACHECOUNTFREEPCT), the cache truncation
1425 * daemon is awakened and will start freeing space. The cache truncation
1426 * daemon will continue to free space until the cache is at most 85% full
1427 * by space (CM_DCACHESPACEFREEPCT - CM_DCACHEEXTRAPCT). The cache truncation
1428 * daemon will also try to free space until the cache is at most 90% full by
1429 * chunks (CM_DCACHECOUNTFREEPCT - CM_DCACHEEXTRAPCT), but the 85% space limit
1430 * is the only limit that we must hit.
1431 * afs_UFSWrite and afs_GetDCache (when it needs to fetch data) will wait on
1432 * afs_WaitForCacheDrain if the cache is 98% (CM_WAITFORDRAINPCT) full.
1433 * afs_GetDownD wakes those processes once the cache is 95% full
1434 * (CM_CACHESIZEDRAINEDPCT).
1436 #define CM_MAXDISCARDEDCHUNKS 16 /* # of chunks */
1437 #define CM_DCACHECOUNTFREEPCT 95 /* max pct of chunks in use */
1438 #define CM_DCACHESPACEFREEPCT 90 /* max pct of space in use */
1439 #define CM_DCACHEEXTRAPCT 5 /* extra to get when freeing */
1440 #define CM_CACHESIZEDRAINEDPCT 95 /* wakeup processes when down to here. */
1441 #define CM_WAITFORDRAINPCT 98 /* sleep if cache is this full. */
1443 /* when afs_cacheBlocks is large, settle for slightly decreased precision */
1444 #define PERCENT(p, v) \
1445 ((afs_cacheBlocks & 0xffe00000) ? ((v) / 100 * (p)) : ((p) * (v) / 100))
1447 #define afs_CacheIsTooFull() \
1448 (afs_blocksUsed - afs_blocksDiscarded > \
1449 PERCENT(CM_DCACHESPACEFREEPCT, afs_cacheBlocks) || \
1450 afs_freeDCCount - afs_discardDCCount < \
1451 PERCENT(100 - CM_DCACHECOUNTFREEPCT, afs_cacheFiles))
1453 /* Handy max length of a numeric string. */
1454 #define CVBS 12 /* max afs_int32 is 2^32 ~ 4*10^9, +1 for NULL, +luck */
1456 #define refpanic(foo) if (afs_norefpanic) \
1457 { printf( foo ); afs_norefpanic++;} else osi_Panic( foo )
1460 ** these are defined in the AIX source code sys/fs_locks.h but are not
1461 ** defined anywhere in the /usr/include directory
1463 #if defined(AFS_AIX41_ENV)
1464 #define VN_LOCK(vp) simple_lock(&(vp)->v_lock)
1465 #define VN_UNLOCK(vp) simple_unlock(&(vp)->v_lock)
1468 /* get a file's serial number from a vnode */
1469 #ifndef afs_vnodeToInumber
1470 #if defined(AFS_SGI62_ENV) || defined(AFS_HAVE_VXFS) || defined(AFS_DARWIN_ENV)
1471 #define afs_vnodeToInumber(V) VnodeToIno(V)
1473 #define afs_vnodeToInumber(V) (VTOI(V)->i_number)
1474 #endif /* AFS_SGI62_ENV */
1477 /* get a file's device number from a vnode */
1478 #ifndef afs_vnodeToDev
1479 #if defined(AFS_SGI62_ENV) || defined(AFS_HAVE_VXFS) || defined(AFS_DARWIN_ENV)
1480 #define afs_vnodeToDev(V) VnodeToDev(V)
1481 #elif defined(UKERNEL)
1482 #define afs_vnodeToDev(V) (VTOI(V) ? (VTOI(V)->i_dev) : (-1))
1484 #define afs_vnodeToDev(V) (VTOI(V)->i_dev)
1488 /* declare something so that prototypes don't flip out */
1489 /* appears struct buf stuff is only actually passed around as a pointer,
1490 except with libuafs, in which case it is actually defined */
1494 struct rxfs_storeVariables {
1495 struct rx_call *call;
1496 struct vcache *vcache;
1500 afs_int32 hasNo64bit;
1501 struct AFSStoreStatus InStatus;
1505 int (*prepare)(void *rock, afs_uint32 size, afs_uint32 *bytestoxfer);
1506 int (*read)(void *rock, struct osi_file *tfile, afs_uint32 offset,
1507 afs_uint32 tlen, afs_uint32 *bytesread);
1508 int (*write)(void *rock, afs_uint32 tlen, afs_uint32 *byteswritten);
1509 int (*status)(void *rock);
1510 int (*padd)(void *rock, afs_uint32 tlen);
1511 int (*close)(void *rock, struct AFSFetchStatus *OutStatus,
1512 afs_int32 *doProcessFS);
1513 int (*destroy)(void **rock, afs_int32 error);
1514 int (*storeproc)(struct storeOps *, void *, struct dcache *, int *,
1519 int (*more)(void *rock, afs_int32 *length, afs_uint32 *moredata);
1520 int (*read)(void *rock, afs_uint32 tlen, afs_uint32 *bytesread);
1521 int (*write)(void *rock, struct osi_file *fp, afs_uint32 offset,
1522 afs_uint32 tlen, afs_uint32 *byteswritten);
1523 int (*close)(void *rock, struct vcache *avc, struct dcache *adc,
1524 struct afs_FetchOutput *Outputs);
1525 int (*destroy)(void **rock, afs_int32 error);
1528 /* fakestat support: opaque storage for afs_EvalFakeStat to remember
1529 * what vcache should be released.
1531 struct afs_fakestat_state {
1535 struct vcache *root_vp;
1538 extern int afs_fakestat_enable;
1539 extern int afs_rmtsys_enable;
1540 extern int afsd_dynamic_vcaches;
1543 * Wrappers for access to credentials structure members
1544 * Linux uses the kernel cred structure if available, with the
1545 * wrappers defined in LINUX/osi_machdep.h
1547 #if defined(AFS_NBSD40_ENV)
1548 /* in osi_machdep.h as expected */
1549 #elif defined (AFS_DARWIN110_ENV)
1550 #define afs_cr_uid(cred) kauth_cred_getuid((kauth_cred_t)(cred))
1551 #define afs_cr_gid(cred) kauth_cred_getgid((kauth_cred_t)(cred))
1552 #elif !(defined(AFS_LINUX26_ENV) && defined(STRUCT_TASK_STRUCT_HAS_CRED))
1553 #define afs_cr_uid(cred) ((cred)->cr_uid)
1554 #define afs_cr_gid(cred) ((cred)->cr_gid)
1555 #if !defined(AFS_OBSD_ENV)
1556 #define afs_cr_ruid(cred) ((cred)->cr_ruid)
1557 #define afs_cr_rgid(cred) ((cred)->cr_rgid)
1560 #if !defined(AFS_DARWIN110_ENV)
1562 afs_set_cr_uid(afs_ucred_t *cred, uid_t uid) {
1566 afs_set_cr_gid(afs_ucred_t *cred, gid_t gid) {
1569 #if !defined(AFS_OBSD_ENV)
1571 afs_set_cr_ruid(afs_ucred_t *cred, uid_t uid) {
1572 cred->cr_ruid = uid;
1575 afs_set_cr_rgid(afs_ucred_t *cred, gid_t gid) {
1576 cred->cr_rgid = gid;
1578 #endif /* ! AFS_OBSD_ENV */
1579 #endif /* ! AFS_DARWIN110_ENV */
1584 /** The 32 bit OS expects the members of this structure to be 32 bit
1585 * quantities and the 64 bit OS expects them as 64 bit quanties. Hence
1586 * to accomodate both, *long* is used instead of afs_int32
1598 extern int Afs_syscall(struct afssysa *uap, rval_t *rvp);
1599 #endif /* AFS_SUN5_ENV */
1602 * Various definitions for osi_sleep and its event hash table
1603 * DFBSD has no osi_sleep, and HPUX has its own hack for this stuff
1605 #define AFS_EVHASHSIZE 128 /* size of afs_evhasht, must be power of 2 */
1607 typedef struct afs_event {
1608 struct afs_event *next; /* next in hash chain */
1609 char *event; /* lwp event: an address */
1610 int refcount; /* Is it in use? */
1611 int seq; /* Sequence number: this is incremented
1612 * by wakeup calls; wait will not return until
1614 #if defined(AFS_AIX_ENV)
1616 #elif defined(AFS_DARWIN_ENV)
1617 # ifdef AFS_DARWIN80_ENV
1621 /* no cond member */
1622 #elif defined(AFS_FBSD_ENV) || defined(AFS_OBSD_ENV)
1623 int cond; /* "all this gluck should probably be replaced by CVs" */
1624 #elif defined(AFS_LINUX24_ENV)
1625 wait_queue_head_t cond;
1626 #elif defined(AFS_NBSD_ENV) || defined(AFS_SUN5_ENV) || defined(AFS_SGI_ENV)
1627 kcondvar_t cond; /* Currently associated condition variable */
1631 extern afs_event_t *afs_evhasht[AFS_EVHASHSIZE]; /* Hash table for events */
1633 #endif /* _AFS_H_ */