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 * The string that we use for the "device" we mount /afs on. That is, we're not
47 * mounted on a drive like /dev/sda, but instead mounted on "afs".
49 #define AFS_MOUNT_STR "afs"
51 # define MOUNT_AFS AFS_MOUNT_STR
55 * Macros to uniquely identify the AFS vfs struct
57 #define AFS_VFSMAGIC 0x1234
60 * Define AFS_VFSFSID, an id number corresponding to the AFS filesystem. On
61 * Irix, this is stored in a global var (afs_fstype); everywhere else, we have
62 * a constant. If we have a platform-defined value for this, use it (AFS_FSNO);
63 * otherwise, we just make up an arbitrary value here (99).
66 # define AFS_VFSFSID afs_fstype
67 #elif defined(AFS_FSNO)
68 # define AFS_VFSFSID AFS_FSNO
70 # define AFS_VFSFSID 99
73 /* use this value for reporting total space, free space, etc.
74 * fake a high number to satisfy programs that use the statfs call to make sure
75 * that there's enough space in the device partition before storing something
76 * there. keep it just under 2^31, to try and be safe about not breaking
77 * stuff that looks at the amount of free space. */
78 #define AFS_VFS_FAKEFREE (2147483647)
80 /* Moved from VNOPS/afs_vnop_flocks so can be used in prototypes */
81 #if defined(AFS_HPUX102_ENV)
82 #define AFS_FLOCK k_flock
84 #if defined(AFS_SUN5_ENV) || (defined(AFS_LINUX24_ENV) && !(defined(AFS_LINUX26_ENV) && defined(AFS_LINUX_64BIT_KERNEL)))
85 #define AFS_FLOCK flock64
87 #define AFS_FLOCK flock
88 #endif /* AFS_SUN65_ENV */
89 #endif /* AFS_HPUX102_ENV */
91 /* The following are various levels of afs debugging */
92 #define AFSDEB_GENERAL 1 /* Standard debugging */
93 #define AFSDEB_NETWORK 2 /* low level afs networking */
94 #define AFSDEB_RX 4 /* RX debugging */
95 #define AFSDEB_VNLAYER 8 /* interface layer to AFS (aixops, gfsops, etc) */
97 /* generic undefined vice id */
100 /* The basic defines for the Andrew file system
101 better keep things powers of two so "& (foo-1)" hack works for masking bits */
102 #define NBRS 15 /* max number of queued daemon requests */
103 #define NUSERS 16 /* hash table size for unixuser table */
104 #define NSERVERS 16 /* hash table size for server table */
105 #define NVOLS 64 /* hash table size for volume table */
106 #define NFENTRIES 256 /* hash table size for disk volume table */
107 #define VCSIZEBITS 16 /* log of stat cache hash table size */
108 #define VCSIZE (opr_jhash_size(VCSIZEBITS))
109 #define CBRSIZE 512 /* call back returns hash table size */
110 #define PIGGYSIZE 1350 /* max piggyback size */
111 #define MAXVOLS 128 /* max vols we can store */
112 #define MAXSYSNAME 128 /* max sysname (i.e. @sys) size */
113 #define MAXNUMSYSNAMES 32 /* max that current constants allow */
114 #define MAXROOTVOLNAMELEN 64 /* max length of root volume name */
115 #define NOTOKTIMEOUT (2*3600) /* time after which to timeout conns sans tokens */
116 #define NOPAG 0xffffffff
120 #define AFS_MAXCBRSCALL 32 /* max to return in a given call (must be <= AFSCBMAX) */
121 #define VCACHE_FREE 5
122 #define VCACHE_DYNAMIC_STRESSED 10000
123 #define VCACHE_STRESS_LOGINTERVAL (60*60*4) /* 4 hours */
124 #define AFS_NRXPACKETS 80
125 #define AFS_RXDEADTIME 50
126 #define AFS_HARDDEADTIME 120
127 #define AFS_IDLEDEADTIME 1200
128 #define AFS_IDLEDEADTIME_REP 180 /* more than fs's cb dead time */
129 #define AFS_BLKBITS 12
130 #define AFS_BLKSIZE (1 << AFS_BLKBITS)
132 extern afs_int32 afs_rx_deadtime;
133 extern afs_int32 afs_rx_harddead;
134 extern afs_int32 afs_rx_idledead;
135 extern afs_int32 afs_rx_idledead_rep;
137 struct sysname_info {
140 signed char index, allocked;
143 /* flags to use with AFSOP_CACHEINIT */
144 #define AFSCALL_INIT_MEMCACHE 0x1 /* use a memory-based cache */
146 /* below here used only for kernel procedures */
148 /* Store synchrony flags - SYNC means that data should be forced to server's
149 * disk immediately upon completion. */
152 #define AFS_VMSYNC_INVAL 2 /* sync and invalidate pages */
153 #define AFS_LASTSTORE 4
154 #define AFS_VMSYNC 8 /* sync pages but do not invalidate */
156 /* background request structure */
159 #define BOP_NOOP 0 /* leave 0 unused */
160 #define BOP_FETCH 1 /* parm1 is chunk to get */
161 #define BOP_STORE 2 /* parm1 is chunk to store */
162 #define BOP_PATH 3 /* parm1 is path, parm2 is chunk to fetch */
163 #define BOP_FETCH_NOCACHE 4 /* parms are: vnode ptr, offset, segment ptr, addr, cred ptr */
164 #ifdef AFS_DARWIN_ENV
165 #define BOP_MOVE 5 /* ptr1 afs_uspc_param ptr2 sname ptr3 dname */
167 #define BOP_PARTIAL_STORE 6 /* parm1 is chunk to store */
168 #define BOP_INVALIDATE_SEGMENTS 7 /* no parms: just uses the 'bp->vc' vcache */
170 #define B_DONTWAIT 1 /* On failure return; don't wait */
172 /* protocol is: refCount is incremented by user to take block out of free pool.
173 Next, BSTARTED is set when daemon finds request. This prevents
174 other daemons from picking up the same request. Finally, when
175 request is done, refCount is zeroed. BDONE and BWAIT are used by
176 dudes waiting for operation to proceed to a certain point before returning.
178 #define BSTARTED 1 /* request picked up by a daemon */
179 #define BUVALID 2 /* code is valid (store) */
180 #define BUWAIT 4 /* someone is waiting for BUVALID */
182 struct vcache *vc; /* vnode to use, with vrefcount bumped */
183 afs_ucred_t *cred; /* credentials to use for operation */
184 afs_size_t size_parm[BPARMS]; /* random parameters */
185 void *ptr_parm[BPARMS]; /* pointer parameters */
186 afs_int32 code_raw; /* return code from AFS routines */
187 afs_int32 code_checkcode; /* the afs_CheckCode-translated code */
188 short refCount; /* use counter for this structure */
189 char opcode; /* what to do (store, fetch, etc) */
190 char flags; /* free, etc */
191 afs_int32 ts; /* counter "timestamp" */
199 afs_int32 AuthHandle;
200 char HandShakeKey[8];
202 afs_int32 BeginTimestamp;
203 afs_int32 EndTimestamp;
207 afs_int32 Cell; /* better sun packing if at end of structure */
211 /* Temporary struct to be passed between afs_fid and afs_vget */
214 afs_int32 CellAndUnique;
217 /* The actual number of bytes in the SmallFid, not the sizeof struct. */
218 #define SIZEOF_SMALLFID 10
223 * Circular queues, implemented with pointers. Structures may contain as many
224 * queues as required, which may be located at any point within the structure,
225 * providing the QEntry macro is used to translate between a queue pointer, and
226 * the address of its containing structure
234 #define QZero(e) ((e)->prev = (e)->next = NULL)
235 #define QInit(q) ((q)->prev = (q)->next = (q))
236 #define QAdd(q,e) ((e)->next = (q)->next, (e)->prev = (q), \
237 (q)->next->prev = (e), (q)->next = (e))
238 #define QRemove(e) ((e)->next->prev = (e)->prev, (e)->prev->next = (e)->next, (e)->prev = NULL, (e)->next = NULL)
239 #define QNext(e) ((e)->next)
240 #define QPrev(e) ((e)->prev)
241 #define QEmpty(q) ((q)->prev == (q))
242 /* this one takes q1 and sticks it on the end of q2 - that is, the other end, not the end
243 * that things are added onto. q1 shouldn't be empty, it's silly */
244 #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))
246 /* Given a pointer to an afs_q within a structure, go back to the address of
247 * the parent structure
250 #define QEntry(queue, structure, member) \
251 ((structure *)((char *)(queue)-(char *)(&((structure *)NULL)->member)))
253 /* And implement operations for individual lists in terms of the above macro */
255 #define QTOV(e) QEntry(e, struct vcache, vlruq)
256 #define QTOC(e) QEntry(e, struct cell, lruq)
257 #define QTOVH(e) QEntry(e, struct vcache, vhashq)
260 * List of free slot numbers
262 struct afs_slotlist {
264 struct afs_slotlist *next;
268 afs_int32 uid; /* user id making the request */
269 afs_int32 busyCount; /* how many busies we've seen so far */
270 afs_int32 flags; /* things like O_SYNC, O_NONBLOCK go here */
271 char initd; /* if non-zero, Error fields meaningful */
272 char accessError; /* flags for overriding error return code */
273 char volumeError; /* encountered a missing or busy volume */
274 char networkError; /* encountered network problems */
275 char permWriteError; /* fileserver returns permenent error. */
276 char tokenError; /* a token error other than expired. */
277 char idleError; /* the server idled too long */
278 char skipserver[AFS_MAXHOSTS];
279 afs_int32 lasterror[AFS_MAXHOSTS];
284 /* structure linked off of a server to keep track of queued returned
285 * callbacks. Sent asynchronously when we run a little low on free dudes.
288 struct afs_cbr **pprev;
289 struct afs_cbr *next;
291 struct afs_cbr **hash_pprev;
292 struct afs_cbr *hash_next;
297 #ifdef AFS_LINUX22_ENV
298 /* On Linux, we have to be able to allocate the storage for this using
299 * kmalloc, as otherwise we may deadlock. So, it needs to be able to fit
302 # define AFS_NCBRS PAGE_SIZE/sizeof(struct afs_cbr)
304 # define AFS_NCBRS 1024 /* max # of call back return entries */
307 /* cellinfo file magic number */
308 #define AFS_CELLINFO_MAGIC 0xf32817cd
311 #define CNoSUID 0x02 /* disable suid bit for this cell */
312 #define CLinkedCell4 0x04 /* reserved for ADDCELL2 pioctl */
313 #define CNoAFSDB 0x08 /* never bother trying AFSDB */
314 #define CHasVolRef 0x10 /* volumes were referenced */
315 #define CLinkedCell 0x20 /* has a linked cell in lcellp */
316 #define CHush 0x40 /* don't display until referenced */
319 struct afs_q lruq; /* lru q next and prev */
320 char *cellName; /* char string name of cell */
321 afs_int32 cellIndex; /* sequence number */
322 afs_int32 cellNum; /* semi-permanent cell number */
323 struct server *cellHosts[AFS_MAXCELLHOSTS]; /* volume *location* hosts */
324 struct cell *lcellp; /* Associated linked cell */
325 u_short fsport; /* file server port */
326 u_short vlport; /* volume server port */
327 short states; /* state flags */
328 time_t timeout; /* data expire time, if non-zero */
329 struct cell_name *cnamep; /* pointer to our cell_name */
330 afs_rwlock_t lock; /* protects cell data */
331 unsigned char cellHandle[16]; /* deterministic handle for this cell */
335 struct cell_name *next;
342 struct cell_alias *next;
348 #define afs_PutCell(cellp, locktype)
350 /* the unixuser flag bit definitions */
351 #define UHasTokens 1 /* are the st and ct fields valid (ever set)? */
352 #define UTokensBad 2 /* are tokens bad? */
353 #define UPrimary 4 /* on iff primary identity */
354 #define UNeedsReset 8 /* needs afs_ResetAccessCache call done */
355 #define UPAGCounted 16 /* entry seen during PAG search (for stats) */
356 #define UNFSGetCreds 32 /* getting creds for NFS client */
357 /* A flag used by afs_GCPAGs to keep track of
358 * which entries in afs_users need to be deleted.
359 * The lifetime of its presence in the table is the
360 * lifetime of the afs_GCPAGs function.
362 #define TMP_UPAGNotReferenced 128
364 /* unixuser notify events */
365 #define UTokensObtained 1
366 #define UTokensDropped 2
368 /* values for afs_gcpags */
369 enum { AFS_GCPAGS_NOTCOMPILED = 0, AFS_GCPAGS_OK =
370 1, AFS_GCPAGS_USERDISABLED, AFS_GCPAGS_EPROC0, AFS_GCPAGS_EPROCN,
371 AFS_GCPAGS_EEQPID, AFS_GCPAGS_EINEXACT, AFS_GCPAGS_EPROCEND,
372 AFS_GCPAGS_EPROCWALK, AFS_GCPAGS_ECREDWALK, AFS_GCPAGS_EPIDCHECK,
373 AFS_GCPAGS_ENICECHECK
376 extern afs_int32 afs_gcpags;
377 extern afs_int32 afs_gcpags_procsize;
378 extern afs_int32 afs_bkvolpref;
379 extern char afs_cachebasedir[1024];
380 extern afs_int32 afs_numcachefiles;
381 extern afs_int32 afs_numfilesperdir;
386 struct ClearToken clearToken;
390 struct rxkadToken rxkad;
394 struct tokenJar *next;
396 union tokenUnion content;
400 struct unixuser *next; /* next hash pointer */
401 afs_int32 uid; /* search based on uid and cell */
403 afs_int32 viceId; /* Optional viced ID corresponding to current tokens */
404 short refCount; /* reference count for allocation */
405 char states; /* flag info */
406 afs_int32 tokenTime; /* last time tokens were set, used for timing out conn data */
407 struct tokenJar *tokens;
408 struct afs_exporter *exporter; /* more info about the exporter for the remote user */
409 void *cellinfo; /* pointer to cell info (PAG manager only) */
413 #define CVEC_LEN 3 /* per-user connection pool */
415 struct sa_conn_vector;
416 typedef struct sa_conn_vector * p_sa_conn_vector; /* forward decl */
418 #define CONN_REPLICATED 0x1
423 char forceConnectFS; /* Should we try again with these tokens? */
424 struct rx_connection *id; /* RPC connid */
425 struct sa_conn_vector *parent; /* the con_vector which contains us */
428 /* An sa_conn_vector replaces the erstwhile list of conn
429 structures maintained by the cache manager. The sa_conn_vector
430 contains a C array of connections which, if non-zero, represent
431 connections to AFS servers.
434 struct sa_conn_vector {
435 /* linked-list machinery */
436 struct sa_conn_vector *next;
438 /* AFS conn-identifying info */
439 struct unixuser *user; /* user validated with respect to. */
440 struct srvAddr *srvr; /* server associated with this conn */
441 short refCount; /* reference count for allocation */
442 unsigned short port; /* port associated with this connection */
445 /* next connection to return when all in cvec are fully utilized */
448 /* connections vector */
449 struct afs_conn cvec[CVEC_LEN];
454 /* Fid comparison routines */
455 #define FidCmp(a,b) ((a)->Fid.Unique != (b)->Fid.Unique \
456 || (a)->Fid.Vnode != (b)->Fid.Vnode \
457 || (a)->Fid.Volume != (b)->Fid.Volume \
458 || (a)->Cell != (b)->Cell)
460 #define FidMatches(afid,tvc) ((tvc)->f.fid.Fid.Vnode == (afid)->Fid.Vnode && \
461 (tvc)->f.fid.Fid.Volume == (afid)->Fid.Volume && \
462 (tvc)->f.fid.Cell == (afid)->Cell && \
463 ( (tvc)->f.fid.Fid.Unique == (afid)->Fid.Unique || \
464 (!(afid)->Fid.Unique && ((tvc)->f.states & CUnique))))
468 #define SRVADDR_ISDOWN 0x20 /* same as SRVR_ISDOWN */
469 #define SRVADDR_NOUSE 0x40 /* Don't use this srvAddr */
472 struct srvAddr *next_bkt; /* next item in hash bucket */
473 struct srvAddr *next_sa; /* another interface on same host */
474 struct server *server; /* back to parent */
475 struct sa_conn_vector *conns; /* All user connections to this server */
476 struct afs_conn *natping;
477 afs_int32 sa_ip; /* Host addr in network byte order */
478 u_short sa_iprank; /* indiv ip address priority */
479 u_short sa_portal; /* port addr in network byte order */
484 * Values used in the flags field of the server structure below.
486 * AFS_SERVER_FLAG_ACTIVATED Has the server ever had a user connection
487 * associated with it?
489 #define AFS_SERVER_FLAG_ACTIVATED 0x01
490 #define SNO_LHOSTS 0x04
491 #define SYES_LHOSTS 0x08
492 #define SVLSRV_UUID 0x10
493 #define SRVR_ISDOWN 0x20
494 #define SRVR_MULTIHOMED 0x40
495 #define SRVR_ISGONE 0x80
496 #define SNO_INLINEBULK 0x100
497 #define SNO_64BIT 0x200
498 #define SCAPS_KNOWN 0x400
500 #define SRV_CAPABILITIES(ts) \
501 { if ( !(ts->flags & SCAPS_KNOWN)) afs_GetCapabilities(ts); ts->capabilities; }
503 #define afs_serverSetNo64Bit(s) (((struct sa_conn_vector*)(s)->parent)->srvr->server->flags |= SNO_64BIT)
504 #define afs_serverHasNo64Bit(s) (((struct sa_conn_vector*)(s)->parent)->srvr->server->flags & SNO_64BIT)
510 afs_int32 addr_uniquifier;
514 struct srvAddr haddr;
517 #define sr_uuid _suid._srvUuid.suuid
518 #define sr_addr_uniquifier _suid._srvUuid.addr_uniquifier
519 #define sr_host _suid._srvId.haddr.ip
520 #define sr_portal _suid._srvId.haddr.portal
521 #define sr_rank _suid._srvId.haddr.ip_rank
522 #define sr_flags _suid._srvId.haddr.flags
523 #define sr_conns _suid._srvId.haddr.conns
524 struct server *next; /* Ptr to next server in hash chain */
525 struct cell *cell; /* Cell in which this host resides */
526 struct afs_cbr *cbrs; /* Return list of callbacks */
527 afs_int32 activationTime; /* Time when this record was first activated */
528 afs_int32 lastDowntimeStart; /* Time when last downtime incident began */
529 afs_int32 numDowntimeIncidents; /* # (completed) downtime incidents */
530 afs_int32 sumOfDowntimes; /* Total downtime experienced, in seconds */
531 struct srvAddr *addr;
532 afs_uint32 flags; /* Misc flags */
533 afs_int32 capabilities;
536 #define afs_PutServer(servp, locktype)
538 /* structs for some pioctls - these are (or should be)
546 struct sprefrequest_33 {
547 unsigned short offset;
548 unsigned short num_servers;
552 struct sprefrequest { /* new struct for 3.4 */
553 unsigned short offset;
554 unsigned short num_servers;
555 unsigned short flags;
560 unsigned short next_offset;
561 unsigned short num_servers;
562 struct spref servers[1]; /* we overrun this array intentionally... */
566 unsigned short flags;
567 unsigned short num_servers;
568 struct spref servers[1]; /* we overrun this array intentionally... */
570 /* struct for GAG pioctl
573 afs_uint32 showflags, logflags, logwritethruflag, spare[3];
574 unsigned char spare2[128];
578 #define logwritethruON 1
581 afs_int32 rx_initReceiveWindow, rx_maxReceiveWindow, rx_initSendWindow,
582 rx_maxSendWindow, rxi_nSendFrags, rxi_nRecvFrags, rxi_OrphanFragSize;
583 afs_int32 rx_maxReceiveSize, rx_MyMaxSendSize;
584 afs_uint32 spare[21];
587 /* struct for checkservers */
598 /* state bits for volume */
599 #define VRO 1 /* volume is readonly */
600 #define VRecheck 2 /* recheck volume info with server */
601 #define VBackup 4 /* is this a backup volume? */
602 #define VForeign 8 /* this is a non-afs volume */
603 #define VPartVisible 16 /* Volume's partition is visible on the client */
604 #define VHardMount 32 /* we are hard-mount waiting for the vol */
606 enum repstate { not_busy, end_not_busy = 6, rd_busy, rdwr_busy, offline };
609 /* One structure per volume, describing where the volume is located
610 * and where its mount points are. */
611 struct volume *next; /* Next volume in hash list. */
612 afs_int32 cell; /* the cell in which the volume resides */
613 afs_rwlock_t lock; /* the lock for this structure */
614 afs_int32 volume; /* This volume's ID number. */
615 char *name; /* This volume's name, or 0 if unknown */
616 struct server *serverHost[AFS_MAXHOSTS]; /* servers serving this volume */
617 enum repstate status[AFS_MAXHOSTS]; /* busy, offline, etc */
618 struct VenusFid dotdot; /* dir to access as .. */
619 struct VenusFid mtpoint; /* The mount point for this volume. */
620 afs_int32 rootVnode, rootUnique; /* Volume's root fid */
623 afs_int32 rwVol; /* For r/o vols, original read/write volume. */
624 afs_int32 setupTime; /* time volume was setup from vldb info */
625 afs_int32 vtix; /* volume table index */
626 afs_int32 copyDate; /* copyDate field, for tracking vol releases */
627 afs_int32 expireTime; /* for per-volume callbacks... */
628 short refCount; /* reference count for allocation */
629 char states; /* here for alignment reasons */
632 #define afs_PutVolume(av, locktype) ((av)->refCount--)
634 /* format of an entry in volume info file */
636 afs_int32 cell; /* cell for this entry */
637 afs_int32 volume; /* volume */
638 afs_int32 next; /* has index */
639 struct VenusFid dotdot; /* .. value */
640 struct VenusFid mtpoint; /* mt point's fid */
641 afs_int32 rootVnode, rootUnique; /* Volume's root fid */
645 struct SimpleLocks *next;
647 afs_int32 boff, eoff;
649 #if defined(AFS_AIX32_ENV) || defined(AFS_SUN5_ENV) || defined(AFS_SGI_ENV)
654 /* vcache state bits */
655 #define CStatd 0x00000001 /* has this file ever been stat'd? */
656 #define CBackup 0x00000002 /* file is on a backup volume */
657 #define CRO 0x00000004 /* is it on a read-only volume */
658 #define CMValid 0x00000008 /* is the mount point info valid? */
659 #define CCore 0x00000010 /* storing a core file, needed since we don't get an open */
660 #define CDirty 0x00000020 /* file has been modified since first open (... O_RDWR) */
661 #define CSafeStore 0x00000040 /* close must wait for store to finish (should be in fd) */
662 #define CMAPPED 0x00000080 /* Mapped files */
663 #define CNSHARE 0x00000100 /* support O_NSHARE semantics */
664 #define CLied 0x00000200
665 #define CTruth 0x00000400
667 #if defined(AFS_DARWIN80_ENV)
668 #define CDeadVnode 0x00000800
669 #elif defined(AFS_DARWIN_ENV)
670 #define CUBCinit 0x00000800
672 #define CWRITE_IGN 0x00000800 /* Next OS hack only */
675 #define CUnique 0x00001000 /* vc's uniquifier - latest unifiquier for fid */
676 #define CForeign 0x00002000 /* this is a non-afs vcache */
677 #define CReadDir 0x00004000 /* readdir in progress */
678 #define CUnlinked 0x00010000
679 #define CBulkStat 0x00020000 /* loaded by a bulk stat, and not ref'd since */
680 #define CUnlinkedDel 0x00040000
681 #define CVFlushed 0x00080000
682 #if defined(AFS_SGI_ENV)
683 #define CWritingUFS 0x00200000 /* to detect vm deadlock - used by sgi */
684 #elif defined(AFS_DARWIN80_ENV)
685 #define CEvent 0x00200000 /* to preclude deadlock when sending events */
687 #define CCreating 0x00400000 /* avoid needless store after open truncate */
688 #define CPageHog 0x00800000 /* AIX - dumping large cores is a page hog. */
689 #define CCorrupt 0x01000000 /* corrupt dir already reported */
690 #define CDCLock 0x02000000 /* Vnode lock held over call to GetDownD */
691 #define CBulkFetching 0x04000000 /* stats are being fetched by bulk stat */
692 #define CExtendedFile 0x08000000 /* extended file via ftruncate call. */
693 #define CVInit 0x10000000 /* being initialized */
694 #define CMetaDirty 0x20000000 /* vnode meta-data needs to be flushed */
695 #define CPartVisible 0x40000000 /* fileserver partition visible on client */
697 /* vcache vstate bits */
698 #define VRevokeWait 0x1
699 #define VPageCleaning 0x2 /* Solaris - Cache Trunc Daemon sez keep out */
701 /* Dirty disconnected vcache flags. */
702 #define VDisconSetTime 0x00000001 /* set time. */
703 #define VDisconSetMode 0x00000002 /* set mode. */
704 /* XXX: to be continued ? */
705 #define VDisconTrunc 0x00000020 /* truncate file. */
706 #define VDisconSetAttrMask 0x0000003F /* Masks for setattr ops. */
707 #define VDisconWriteClose 0x00000400 /* Write op on file close. */
708 #define VDisconWriteFlush 0x00000800 /* Write op on normal fsync/flush. */
709 #define VDisconWriteOsiFlush 0x00001000 /* Write op on osi flush. */
711 #define VDisconRemove 0x00002000 /* Remove vnop. */
712 #define VDisconCreate 0x00004000 /* Create vnop. */
713 #define VDisconCreated 0x00008000 /* A file that was created during
714 this resync operation */
715 #define VDisconRename 0x00010000 /* Rename vnop. */
716 #define VDisconRenameSameDir 0x00020000 /* Rename in same dir. */
718 /*... to be continued ... */
720 /* vcache (file) cachingStates bits */
721 #define FCSDesireBypass 0x1 /* This file should bypass the cache */
722 #define FCSBypass 0x2 /* This file is currently NOT being cached */
723 #define FCSManuallySet 0x4 /* The bypass flags were set, or reset, manually (via pioctl)
724 and should not be overridden by the file's name */
726 /* Flag values used by the Transition routines */
727 #define TRANSChangeDesiredBit 0x1 /* The Transition routine should set or
728 * reset the FCSDesireBypass bit */
729 #define TRANSVcacheIsLocked 0x2 /* The Transition routine does not need to
730 * lock vcache (it's already locked) */
731 #define TRANSSetManualBit 0x4 /* The Transition routine should set FCSManuallySet so that
732 * filename checking does not override pioctl requests */
735 #if defined(AFS_XBSD_ENV) || defined(AFS_DARWIN_ENV)
736 #define vrefCount v->v_usecount
737 #elif defined(AFS_SUN511_ENV)
738 # define vrefCount v->v_count
740 #define vrefCount v.v_count
741 #endif /* AFS_XBSD_ENV */
743 #if defined(AFS_DARWIN80_ENV)
744 #define VREFCOUNT_GT(v, y) vnode_isinuse(AFSTOV(v), (y))
745 #elif defined(AFS_FBSD_ENV)
746 #define VREFCOUNT(v) (vrefcnt(AFSTOV(v)))
747 #define VREFCOUNT_GT(v, y) (vrefcnt(AFSTOV(v)) > y)
748 #elif defined(AFS_XBSD_ENV) || defined(AFS_DARWIN_ENV)
749 #define VREFCOUNT(v) ((v)->vrefCount)
750 #define VREFCOUNT_GT(v, y) (AFSTOV(v)->v_usecount > (y))
751 #elif defined(AFS_LINUX24_ENV)
752 #define VREFCOUNT(v) atomic_read(&(AFSTOV(v)->v_count))
753 #define VREFCOUNT_GT(v, y) (VREFCOUNT(v)>y)
754 #define VREFCOUNT_SET(v, c) atomic_set(&(AFSTOV(v)->v_count), c)
755 #define VREFCOUNT_DEC(v) atomic_dec(&(AFSTOV(v)->v_count))
756 #define VREFCOUNT_INC(v) atomic_inc(&(AFSTOV(v)->v_count))
758 #define VREFCOUNT(v) ((v)->vrefCount)
759 #define VREFCOUNT_GT(v,y) ((v)->vrefCount > (y))
760 #define VREFCOUNT_SET(v, c) (v)->vrefCount = c
761 #define VREFCOUNT_DEC(v) (v)->vrefCount--
762 #define VREFCOUNT_INC(v) (v)->vrefCount++
763 #define d_unhash(d) list_empty(&(d)->d_hash)
764 #define dget_locked(d) dget(d)
767 #define AFS_MAXDV 0x7fffffff /* largest dataversion number */
768 #ifdef AFS_64BIT_CLIENT
769 #define AFS_NOTRUNC 0x7fffffffffffffffLL /* largest positive int64 number */
770 #else /* AFS_64BIT_CLIENT */
771 #define AFS_NOTRUNC 0x7fffffff /* largest dataversion number */
772 #endif /* AFS_64BIT_CLIENT */
774 extern afs_int32 vmPageHog; /* counter for # of vnodes which are page hogs. */
776 #if defined(AFS_DARWIN80_ENV)
777 #define VTOAFS(v) ((struct vcache *)vnode_fsnode((v)))
778 #define AFSTOV(vc) ((vc)->v)
779 #elif defined(AFS_NBSD40_ENV)
781 struct genfs_node gfsn;
782 struct vcache *afsvc;
784 #define VTOAFS(v) ((((struct nbvdata *)((v)->v_data)))->afsvc)
785 #define AFSTOV(vc) ((vc)->v)
786 #elif defined(AFS_XBSD_ENV) || defined(AFS_DARWIN_ENV) || defined(AFS_SUN511_ENV) || (defined(AFS_LINUX22_ENV) && !defined(STRUCT_SUPER_OPERATIONS_HAS_ALLOC_INODE))
787 #define VTOAFS(v) ((struct vcache *)(v)->v_data)
788 #define AFSTOV(vc) ((vc)->v)
790 #define VTOAFS(V) ((struct vcache *)(V))
791 #define AFSTOV(V) (&(V)->v)
799 /* VCache elements which are kept on disk, and in the kernel */
804 afs_hyper_t DataVersion;
808 afs_uint16 Mode; /* XXXX Should be afs_int32 XXXX */
809 afs_uint16 LinkCount;
810 #ifdef AFS_DARWIN80_ENV
813 /* vnode type is in v.v_type */
816 struct afs_vnuniq parent;
818 /*! Truncate file to this position at the next store */
821 /*! System:AnyUser's access to this. */
827 /*! Disconnected flags for this vcache element. */
828 afs_uint32 ddirty_flags;
829 /*! Shadow vnode + unique keep the shadow dir location. */
830 struct afs_vnuniq shadow;
831 /*! The old parent FID for renamed vnodes */
832 struct afs_vnuniq oldParent;
835 /* Values for 'mvstat' in struct vcache */
836 #define AFS_MVSTAT_FILE (0x0) /* regular file or directory */
837 #define AFS_MVSTAT_MTPT (0x1) /* mountpoint */
838 #define AFS_MVSTAT_ROOT (0x2) /* volume root dir */
842 * This is for the multiPage field in struct vcache. Each one of these
843 * represents an outstanding getpage request that is larger than a single page.
844 * Recording these is necessary to prevent afs_GetOnePage from trying to evict
845 * a dcache entry that an earlier afs_GetOnePage call got in the same getpage
846 * request. See osi_VM_MultiPageConflict and afs_getpage.
848 struct multiPage_range {
850 offset_t off; /**< offset of getpage request */
851 u_int len; /**< length of getpage request */
855 /* INVARIANTs: (vlruq.next != NULL) == (vlruq.prev != NULL)
856 * nextfree => !vlruq.next && ! vlruq.prev
857 * !(avc->nextfree) && !avc->vlruq.next => (FreeVCList == avc->nextfree)
860 #if defined(AFS_XBSD_ENV) || defined(AFS_DARWIN_ENV) || defined(AFS_SUN511_ENV) || (defined(AFS_LINUX22_ENV) && !defined(STRUCT_SUPER_OPERATIONS_HAS_ALLOC_INODE))
863 struct vnode v; /* Has reference count in v.v_count */
865 struct afs_q vlruq; /* lru q next and prev */
866 #if !defined(AFS_LINUX22_ENV)
867 struct vcache *nextfree; /* next on free list (if free) */
869 struct vcache *hnext; /* Hash next */
870 struct afs_q vhashq; /* Hashed per-volume list */
871 /*! Queue of dirty vcaches. Lock with afs_disconDirtyLock */
873 /*! Queue of vcaches with shadow entries. Lock with afs_disconDirtyLock */
874 struct afs_q shadowq;
875 /*! Queue of vcaches with dirty metadata. Locked by afs_xvcdirty */
876 struct afs_q metadirty;
877 /*! Vcaches slot number in the disk backup. Protected by tvc->lock */
880 afs_rwlock_t lock; /* The lock on the vcache contents. */
881 #if defined(AFS_SUN5_ENV)
882 /* Lock used to protect the activeV, multipage, and vstates fields.
883 * Do not try to get the vcache lock when the vlock is held */
885 #endif /* defined(AFS_SUN5_ENV) */
886 #if defined(AFS_SUN5_ENV)
893 #if defined(AFS_AIX51_ENV)
901 int ownslock; /* pid of owner of excl lock, else 0 - defect 3083 */
903 #ifdef AFS_DARWIN80_ENV
905 #elif defined(AFS_DARWIN_ENV)
906 struct lock__bsd__ rwlock;
909 #if !defined(AFS_DFBSD_ENV) && !defined(AFS_NBSD_ENV)
915 char *silly_name; /* For sillyrenamed regular files, the silly
916 * name the file was renamed to. */
917 struct VenusFid *target_root; /* For mountpoints, the fid of the root dir
918 * in the target volume. */
919 struct VenusFid *parent; /* For root dir vcaches, the fid of the
923 char *linkData; /* Link data if a symlink. */
924 afs_hyper_t flushDV; /* data version last flushed from text */
925 afs_hyper_t mapDV; /* data version last flushed from map */
926 struct server *callback; /* The callback host, if any */
927 afs_uint32 cbExpires; /* time the callback expires */
928 struct afs_q callsort; /* queue in expiry order, sort of */
929 struct axscache *Access; /* a list of cached access bits */
930 afs_int32 last_looker; /* pag/uid from last lookup here */
931 #if defined(AFS_SUN5_ENV)
933 #endif /* defined(AFS_SUN5_ENV) */
934 struct SimpleLocks *slocks;
935 short opens; /* The numbers of opens, read or write, on this file. */
936 short execsOrWriters; /* The number of execs (if < 0) or writers (if > 0) of
938 short flockCount; /* count of flock readers, or -1 if writer */
939 char mvstat; /* see the AFS_MVSTAT_* constants */
941 char cachingStates; /* Caching policies for this file */
942 afs_uint32 cachingTransitions; /* # of times file has flopped between caching and not */
944 #if defined(AFS_LINUX24_ENV)
945 off_t next_seq_offset; /* Next sequential offset (used by prefetch/readahead) */
946 #elif defined(AFS_SUN5_ENV) || defined(AFS_SGI65_ENV)
947 off_t next_seq_blk_offset; /* accounted in blocks for Solaris & IRIX */
950 #if defined(AFS_SUN5_ENV)
951 afs_uint32 vstates; /* vstate bits */
952 #endif /* defined(AFS_SUN5_ENV) */
953 struct dcache *dchint;
954 struct dcache *dcreaddir; /* dcache for in-progress readdir */
955 unsigned int readdir_pid; /* pid of the thread in readdir */
956 #if defined(AFS_SGI_ENV)
957 daddr_t lastr; /* for read-ahead */
959 uint64_t vc_rwlockid; /* kthread owning rwlock */
961 short vc_rwlockid; /* pid of process owning rwlock */
963 short vc_locktrips; /* # of rwlock reacquisitions */
964 sema_t vc_rwlock; /* vop_rwlock for afs */
965 pgno_t mapcnt; /* # of pages mapped */
966 struct cred *cred; /* last writer's cred */
968 struct bhv_desc vc_bhv_desc; /* vnode's behavior data. */
970 #endif /* AFS_SGI_ENV */
971 #if defined(AFS_LINUX26_ENV)
972 cred_t *cred; /* last writer's cred */
974 #ifdef AFS_LINUX24_ENV
975 struct dentry *target_link; /* dentry we prefer, when we are redirecting
976 * all requests due to duplicate dentry aliases.
977 * See LINUX/osi_vnodeops.c. Note that this is
978 * NOT an actual reference to a dentry, so this
979 * pointer MUST NOT be dereferenced on its own. */
981 afs_int32 vc_error; /* stash write error for this vnode. */
982 int xlatordv; /* Used by nfs xlator */
984 int asynchrony; /* num kbytes to store behind */
986 struct afs_q multiPage; /* list of multiPage_range structs */
988 int protocol; /* RX_FILESERVER, RX_OSD, ... defined in afsint.xg */
989 #if !defined(UKERNEL)
990 void *vpacRock; /* used to read or write in visible partitions */
992 afs_uint32 lastBRLWarnTime; /* last time we warned about byte-range locks */
993 #ifdef AFS_LINUX26_ENV
994 spinlock_t pagewriter_lock;
995 struct list_head pagewriters; /* threads that are writing vm pages */
999 #ifdef AFS_LINUX26_ENV
1001 struct list_head link;
1006 #define DONT_CHECK_MODE_BITS 0
1007 #define CHECK_MODE_BITS 1
1008 #define CMB_ALLOW_EXEC_AS_READ 2 /* For the NFS xlator */
1010 #if defined(AFS_SGI_ENV)
1011 #define AVCRWLOCK(avc) (valusema(&(avc)->vc_rwlock) <= 0)
1013 /* SGI vnode rwlock macros and flags. */
1014 #ifndef AFS_SGI62_ENV
1015 /* The following are defined here. SGI 6.2 declares them in vnode.h */
1016 #define VRWLOCK_READ 0
1017 #define VRWLOCK_WRITE 1
1018 #define VRWLOCK_WRITE_DIRECT 2
1021 #ifdef AFS_SGI53_ENV
1022 #ifdef AFS_SGI62_ENV
1023 #define AFS_RWLOCK_T vrwlock_t
1025 #define AFS_RWLOCK_T int
1026 #endif /* AFS_SGI62_ENV */
1027 #ifdef AFS_SGI64_ENV
1028 #include <ksys/behavior.h>
1029 #define AFS_RWLOCK(V,F) \
1030 afs_rwlock(&VTOAFS(V)->vc_bhv_desc, (F));
1031 #define AFS_RWUNLOCK(V,F) \
1032 afs_rwunlock(&VTOAFS(V)->vc_bhv_desc, (F));
1035 #define AFS_RWLOCK(V,F) afs_rwlock((vnode_t *)(V), (F) )
1036 #define AFS_RWUNLOCK(V,F) afs_rwunlock((vnode_t *)(V), (F) )
1038 #else /* AFS_SGI53_ENV */
1039 #define AFS_RWLOCK(V,F) afs_rwlock((V))
1040 #define AFS_RWUNLOCK(V,F) afs_rwunlock((V))
1041 #endif /* AFS_SGI53_ENV */
1042 #endif /* AFS_SGI_ENV */
1045 struct VenusFid fid;
1046 afs_hyper_t DataVersion;
1048 afs_int32 parentVnode;
1049 afs_int32 parentUnique;
1050 afs_hyper_t flushDV;
1053 afs_int32 randomUid[CPSIZE];
1054 afs_int32 callback; /* Now a pointer to 'server' struct */
1055 afs_int32 cbExpires;
1056 afs_int32 randomAccess[CPSIZE];
1057 afs_int32 anyAccess;
1059 short execsOrWriters;
1066 afs_int32 callerAccess;
1067 afs_int32 cbExpires;
1068 afs_int32 anyAccess;
1077 /* CM inititialization parameters. What CM actually used after calculations
1078 * based on passed in arguments.
1080 #define CMI_VERSION 1 /* increment when adding new fields. */
1081 struct cm_initparams {
1083 int cmi_nChunkFiles;
1084 int cmi_nStatCaches;
1085 int cmi_nDataCaches;
1086 int cmi_nVolumeCaches;
1087 int cmi_firstChunkSize;
1088 int cmi_otherChunkSize;
1089 int cmi_cacheSize; /* The original cache size, in 1K blocks. */
1090 unsigned cmi_setTime:1;
1091 unsigned cmi_memCache:1;
1092 int spare[16 - 9]; /* size of struct is 16 * 4 = 64 bytes */
1096 /*----------------------------------------------------------------------
1097 * AFS Data cache definitions
1099 * Each entry describes a Unix file on the local disk that is
1100 * is serving as a cached copy of all or part of a Vice file.
1101 * Entries live in circular queues for each hash table slot
1103 * Which queue is this thing in? Good question.
1104 * A struct dcache entry is in the freeDSlot queue when not associated with a cache slot (file).
1105 * Otherwise, it is in the DLRU queue. The freeDSlot queue uses the lruq.next field as
1106 * its "next" pointer.
1108 * Cache entries in the DLRU queue are either associated with vice files, in which case
1109 * they are hashed by afs_dvnextTbl and afs_dcnextTbl pointers, or they are in the freeDCList
1110 * and are not associated with any vice file. This last list uses the afs_dvnextTbl pointer for
1111 * its "next" pointer.
1112 *----------------------------------------------------------------------*/
1114 #define NULLIDX (-1) /* null index definition */
1115 /* struct dcache states bits */
1119 #define DWriting 8 /* file being written (used for cache validation) */
1121 /* dcache data flags */
1122 #define DFEntryMod 0x02 /* has entry itself been modified? */
1123 #define DFFetching 0x04 /* file is currently being fetched */
1125 /* dcache meta flags */
1126 #define DFNextStarted 0x01 /* next chunk has been prefetched already */
1127 #define DFFetchReq 0x10 /* someone is waiting for DFFetching to go on */
1130 /* flags in afs_indexFlags array */
1131 #define IFEverUsed 1 /* index entry has >= 1 byte of data */
1132 #define IFFree 2 /* index entry in freeDCList */
1133 #define IFDataMod 4 /* file needs to be written out */
1134 #define IFFlag 8 /* utility flag */
1135 #define IFDirtyPages 16 /* Solaris-only. contains dirty pages */
1136 #define IFAnyPages 32
1137 #define IFDiscarded 64 /* index entry in discardDCList */
1139 #ifdef AFS_DARWIN100_ENV
1140 typedef user_addr_t iparmtype; /* 64 bit */
1141 typedef user_addr_t uparmtype; /* 64 bit */
1143 typedef char * uparmtype;
1144 #ifdef AFS_SGI65_ENV
1145 typedef afs_uint32 iparmtype;
1147 typedef long iparmtype;
1151 #if SIZEOF_VOID_P == SIZEOF_UNSIGNED_INT
1152 # define uintptrsz unsigned int
1153 #elif SIZEOF_VOID_P == SIZEOF_UNSIGNED_LONG
1154 # define uintptrsz unsigned long
1155 #elif SIZEOF_VOID_P == SIZEOF_UNSIGNED_LONG_LONG
1156 # define uintptrsz unsigned long long
1158 # error "Unable to determine casting for pointers"
1162 uparmtype in; /* input buffer */
1163 uparmtype out; /* output buffer */
1164 short in_size; /* Size of input buffer <= 2K */
1165 short out_size; /* Maximum size of output buffer, <= 2K */
1169 * This version of afs_ioctl is required to pass in 32 bit user space
1170 * pointers into a 64 bit kernel.
1173 struct afs_ioctl32 {
1181 /* CacheItems file has a header of type struct afs_fheader
1182 * (keep aligned properly). Since we already have sgi_62 clients running
1183 * with a 32 bit inode, a change is required to the header so that
1184 * they can distinguish the old 32 bit inode CacheItems file and zap it
1185 * instead of using it.
1187 struct afs_fheader {
1188 #define AFS_FHMAGIC 0x7635abaf /* uses version number */
1190 #define AFS_CI_VERSION 4
1192 afs_uint32 dataSize;
1193 afs_int32 firstCSize;
1194 afs_int32 otherCSize;
1197 #if defined(AFS_CACHE_VNODE_PATH)
1198 typedef char *afs_ufs_dcache_id_t;
1199 #elif defined(AFS_SGI61_ENV) || defined(AFS_SUN5_64BIT_ENV)
1200 /* Using ino64_t here so that user level debugging programs compile
1201 * the size correctly.
1203 typedef ino64_t afs_ufs_dcache_id_t;
1204 #elif defined(AFS_LINUX26_ENV)
1205 #define MAX_FH_LEN 10
1207 #if defined(NEW_EXPORT_OPS)
1210 __u32 raw[MAX_FH_LEN];
1211 } afs_ufs_dcache_id_t;
1212 extern int cache_fh_type;
1213 extern int cache_fh_len;
1214 #elif defined(AFS_LINUX_64BIT_KERNEL) && !defined(AFS_S390X_LINUX24_ENV)
1215 typedef long afs_ufs_dcache_id_t;
1216 #elif defined(AFS_AIX51_ENV) || defined(AFS_HPUX1123_ENV)
1217 typedef ino_t afs_ufs_dcache_id_t;
1219 typedef afs_int32 afs_ufs_dcache_id_t;
1222 typedef afs_int32 afs_mem_dcache_id_t;
1225 afs_ufs_dcache_id_t ufs;
1226 afs_mem_dcache_id_t mem;
1229 /* it does not compile outside kernel */
1231 afs_int32 fid; /* is adc->index, the cache file number */
1232 afs_dcache_id_t inode; /* is adc->f.inode, the inode number of the cac\
1235 afs_int32 accesstime;
1236 struct buffer *hashNext;
1241 afs_rwlock_t lock; /* the lock for this structure */
1244 /* kept on disk and in dcache entries */
1246 struct VenusFid fid; /* Fid for this file */
1247 afs_int32 modTime; /* last time this entry was modified */
1248 afs_hyper_t versionNo; /* Associated data version number */
1249 afs_int32 chunk; /* Relative chunk number */
1250 afs_dcache_id_t inode; /* Unix inode for this chunk */
1251 afs_int32 chunkBytes; /* Num bytes in this chunk */
1252 char states; /* Has this chunk been modified? */
1255 /* magic numbers to specify the cache type */
1257 #define AFS_FCACHE_TYPE_UFS 0x0
1258 #define AFS_FCACHE_TYPE_MEM 0x1
1259 #define AFS_FCACHE_TYPE_NFS 0x2
1260 #define AFS_FCACHE_TYPE_EPI 0x3
1262 /* kept in memory */
1264 struct afs_q lruq; /* Free queue for in-memory images */
1265 struct afs_q dirty; /* Queue of dirty entries that need written */
1266 afs_rwlock_t lock; /* Protects validPos, some f */
1267 afs_rwlock_t tlock; /* Atomizes updates to refCount */
1268 afs_rwlock_t mflock; /* Atomizes accesses/updates to mflags */
1269 afs_size_t validPos; /* number of valid bytes during fetch */
1270 afs_int32 index; /* The index in the CacheInfo file */
1271 short refCount; /* Associated reference count. */
1272 char dflags; /* Data flags */
1273 char mflags; /* Meta flags */
1274 struct fcache f; /* disk image */
1275 afs_int32 bucket; /* which bucket these dcache entries are in */
1279 * dcache.lock protects the actual contents of the cache file (in
1280 * f.inode), subfields of f except those noted below, dflags and
1283 * dcache.tlock is used to make atomic updates to refCount. Zero
1284 * refCount dcache entries are protected by afs_xdcache instead of
1287 * dcache.mflock is used to access and update mflags. It cannot be
1288 * held without holding the corresponding dcache.lock. Updating
1289 * mflags requires holding dcache.lock(R) and dcache.mflock(W), and
1290 * checking for mflags requires dcache.lock(R) and dcache.mflock(R).
1291 * Note that dcache.lock(W) gives you the right to update mflags,
1292 * as dcache.mflock(W) can only be held with dcache.lock(R).
1294 * dcache.index, dcache.f.fid, dcache.f.chunk and dcache.f.inode are
1295 * write-protected by afs_xdcache and read-protected by refCount.
1296 * Once an entry is referenced, these values cannot change, and if
1297 * it's on the free list (with refCount=0), it can be reused for a
1298 * different file/chunk. These values can only be written while
1299 * holding afs_xdcache(W) and allocating this dcache entry (thereby
1300 * ensuring noone else has a refCount on it).
1304 /* afs_memcache.c */
1305 struct memCacheEntry {
1306 int size; /* # of valid bytes in this entry */
1307 int dataSize; /* size of allocated data area */
1308 afs_lock_t afs_memLock;
1309 char *data; /* bytes */
1312 struct afs_FetchOutput {
1313 struct AFSVolSync tsync;
1314 struct AFSFetchStatus OutStatus;
1315 struct AFSCallBack CallBack;
1318 /* macro to mark a dcache entry as bad */
1321 (x)->f.fid.Fid.Unique = 0; \
1322 afs_indexUnique[(x)->index] = 0; \
1323 (x)->dflags |= DFEntryMod; \
1326 /* FakeOpen and Fake Close used to be real subroutines. They're only used in
1327 * sun_subr and afs_vnodeops, and they're very frequently called, so I made
1328 * them into macros. They do:
1329 * FakeOpen: fake the file being open for writing. avc->lock must be held
1330 * in write mode. Having the file open for writing is like having a DFS
1331 * write-token: you're known to have the best version of the data around,
1332 * and so the CM won't let it be overwritten by random server info.
1333 * FakeClose: undo the effects of FakeOpen, noting that we want to ensure
1334 * that a real close eventually gets done. We use CCore to achieve this if
1335 * we would end up closing the file. avc->lock must be held in write mode */
1337 #ifdef AFS_AIX_IAUTH_ENV
1338 #define CRKEEP(V, C) (V)->linkData = (char*)crdup((C))
1340 #define CRKEEP(V, C) crhold((C)); (V)->linkData = (char*)(C)
1343 #define afs_FakeOpen(avc) { avc->opens++; avc->execsOrWriters++; }
1344 #define afs_FakeClose(avc, acred) \
1345 { if (avc->execsOrWriters == 1) { \
1346 /* we're the last writer, just use CCore flag */ \
1347 avc->f.states |= CCore; /* causes close to be called later */ \
1349 /* The cred and vnode holds will be released in afs_FlushActiveVcaches */ \
1350 osi_Assert(osi_vnhold(avc) == 0); /* So it won't disappear */ \
1351 CRKEEP(avc, acred); /* Should use a better place for the creds */ \
1354 /* we're not the last writer, let the last one do the store-back for us */ \
1356 avc->execsOrWriters--; \
1360 #define AFS_ZEROS 64 /* zero buffer */
1362 #define afs_DirtyPages(avc) ((avc)->f.states & CDirty)
1364 #define afs_InReadDir(avc) (((avc)->f.states & CReadDir) && (avc)->readdir_pid == MyPidxx2Pid(MyPidxx))
1366 extern struct dcache **afs_indexTable; /*Pointers to in-memory dcache entries */
1367 extern afs_int32 *afs_indexUnique; /*dcache entry Fid.Unique */
1368 extern afs_int32 *afs_dvnextTbl; /*Dcache hash table links */
1369 extern afs_int32 *afs_dcnextTbl; /*Dcache hash table links */
1370 extern afs_int32 afs_cacheFiles; /*Size of afs_indexTable */
1371 extern afs_int32 afs_cacheBlocks; /*1K blocks in cache */
1372 extern afs_int32 afs_cacheStats; /*Stat entries in cache */
1373 extern afs_int32 afs_freeDCCount; /*Count of elts in freeDCList */
1374 extern afs_uint32 afs_CacheTooFullCount;
1375 extern afs_uint32 afs_WaitForCacheDrainCount;
1376 extern struct vcache *afs_vhashT[VCSIZE]; /*Stat cache hash table */
1377 extern struct afs_q afs_vhashTV[VCSIZE]; /* cache hash table on volume */
1378 extern afs_int32 afs_initState; /*Initialization state */
1379 extern afs_int32 afs_termState; /* Termination state */
1380 extern struct VenusFid afs_rootFid; /*Root for whole file system */
1381 extern afs_int32 afs_allCBs; /* Count of callbacks */
1382 extern afs_int32 afs_oddCBs; /* Count of odd callbacks */
1383 extern afs_int32 afs_evenCBs; /* Count of even callbacks */
1384 extern afs_int32 afs_allZaps; /* Count of fid deletes */
1385 extern afs_int32 afs_oddZaps; /* Count of odd fid deletes */
1386 extern afs_int32 afs_evenZaps; /* Count of even fid deletes */
1387 extern struct brequest afs_brs[NBRS]; /* request structures */
1389 #define UHash(auid) ((auid) & (NUSERS-1))
1390 #define VHash(avol) ((avol)&(NVOLS-1))
1391 #define SHash(aserv) ((ntohl(aserv)) & (NSERVERS-1))
1392 #define FVHash(acell,avol) (((avol)+(acell)) & (NFENTRIES-1))
1394 /* Performance hack - we could replace VerifyVCache2 with the appropriate
1395 * GetVCache incantation, and could eliminate even this code from afs_UFSRead
1396 * by making intentionally invalidating quick.stamp in the various callbacks
1397 * expiration/breaking code */
1398 #ifdef AFS_DARWIN_ENV
1399 #define afs_VerifyVCache(avc, areq) \
1400 (((avc)->f.states & CStatd) ? (osi_VM_Setup(avc, 0), 0) : \
1401 afs_VerifyVCache2((avc),areq))
1403 #define afs_VerifyVCache(avc, areq) \
1404 (((avc)->f.states & CStatd) ? 0 : afs_VerifyVCache2((avc),areq))
1407 #define DO_STATS 1 /* bits used by FindVCache */
1412 /* values for adown value of afs_LoopServers */
1414 #define AFS_LS_DOWN 1
1415 #define AFS_LS_ALL 2
1417 /* values for flag param of afs_CheckVolumeNames */
1418 #define AFS_VOLCHECK_EXPIRED 0x1 /* volumes whose callbacks have expired */
1419 #define AFS_VOLCHECK_BUSY 0x2 /* volumes which were marked busy */
1420 #define AFS_VOLCHECK_MTPTS 0x4 /* mount point invalidation also */
1421 #define AFS_VOLCHECK_FORCE 0x8 /* do all forcibly */
1425 #define AFS_FSPORT ((unsigned short) htons(7000))
1426 #define AFS_VLPORT ((unsigned short) htons(7003))
1427 #define AFS_RXOSDPORT ((unsigned short) htons(7011))
1429 #define afs_rdwr(avc, uio, rw, io, cred) \
1430 (((rw) == UIO_WRITE) ? afs_write(avc, uio, io, cred, 0) : afs_read(avc, uio, cred, 0))
1431 #define afs_nlrdwr(avc, uio, rw, io, cred) \
1432 (((rw) == UIO_WRITE) ? afs_write(avc, uio, io, cred, 1) : afs_read(avc, uio, cred, 1))
1434 /* Cache size truncation uses the following low and high water marks:
1435 * If the cache is more than 90% full by space (CM_DCACHESPACEFREEPCT) or
1436 * 95% full by # of chunks (CM_DCACHECOUNTFREEPCT), the cache truncation
1437 * daemon is awakened and will start freeing space. The cache truncation
1438 * daemon will continue to free space until the cache is at most 85% full
1439 * by space (CM_DCACHESPACEFREEPCT - CM_DCACHEEXTRAPCT). The cache truncation
1440 * daemon will also try to free space until the cache is at most 90% full by
1441 * chunks (CM_DCACHECOUNTFREEPCT - CM_DCACHEEXTRAPCT), but the 85% space limit
1442 * is the only limit that we must hit.
1443 * afs_UFSWrite and afs_GetDCache (when it needs to fetch data) will wait on
1444 * afs_WaitForCacheDrain if the cache is 98% (CM_WAITFORDRAINPCT) full.
1445 * afs_GetDownD wakes those processes once the cache is 95% full
1446 * (CM_CACHESIZEDRAINEDPCT).
1448 #define CM_MAXDISCARDEDCHUNKS 16 /* # of chunks */
1449 #define CM_DCACHECOUNTFREEPCT 95 /* max pct of chunks in use */
1450 #define CM_DCACHESPACEFREEPCT 90 /* max pct of space in use */
1451 #define CM_DCACHEEXTRAPCT 5 /* extra to get when freeing */
1452 #define CM_CACHESIZEDRAINEDPCT 95 /* wakeup processes when down to here. */
1453 #define CM_WAITFORDRAINPCT 98 /* sleep if cache is this full. */
1455 /* when afs_cacheBlocks is large, settle for slightly decreased precision */
1456 #define PERCENT(p, v) \
1457 ((afs_cacheBlocks & 0xffe00000) ? ((v) / 100 * (p)) : ((p) * (v) / 100))
1459 #define afs_CacheIsTooFull() \
1460 (afs_blocksUsed - afs_blocksDiscarded > \
1461 PERCENT(CM_DCACHESPACEFREEPCT, afs_cacheBlocks) || \
1462 afs_freeDCCount - afs_discardDCCount < \
1463 PERCENT(100 - CM_DCACHECOUNTFREEPCT, afs_cacheFiles))
1465 /* Handy max length of a numeric string. */
1466 #define CVBS 12 /* max afs_int32 is 2^32 ~ 4*10^9, +1 for NULL, +luck */
1468 #define refpanic(foo) if (afs_norefpanic) \
1469 { printf( foo ); afs_norefpanic++;} else osi_Panic( foo )
1472 ** these are defined in the AIX source code sys/fs_locks.h but are not
1473 ** defined anywhere in the /usr/include directory
1475 #if defined(AFS_AIX41_ENV)
1476 #define VN_LOCK(vp) simple_lock(&(vp)->v_lock)
1477 #define VN_UNLOCK(vp) simple_unlock(&(vp)->v_lock)
1480 /* get a file's serial number from a vnode */
1481 #ifndef afs_vnodeToInumber
1482 #if defined(AFS_SGI62_ENV) || defined(AFS_HAVE_VXFS) || defined(AFS_DARWIN_ENV)
1483 #define afs_vnodeToInumber(V) VnodeToIno(V)
1485 #define afs_vnodeToInumber(V) (VTOI(V)->i_number)
1486 #endif /* AFS_SGI62_ENV */
1489 /* get a file's device number from a vnode */
1490 #ifndef afs_vnodeToDev
1491 #if defined(AFS_SGI62_ENV) || defined(AFS_HAVE_VXFS) || defined(AFS_DARWIN_ENV)
1492 #define afs_vnodeToDev(V) VnodeToDev(V)
1493 #elif defined(UKERNEL)
1494 #define afs_vnodeToDev(V) (VTOI(V) ? (VTOI(V)->i_dev) : (-1))
1496 #define afs_vnodeToDev(V) (VTOI(V)->i_dev)
1500 /* declare something so that prototypes don't flip out */
1501 /* appears struct buf stuff is only actually passed around as a pointer,
1502 except with libuafs, in which case it is actually defined */
1506 struct rxfs_storeVariables {
1507 struct rx_call *call;
1508 struct vcache *vcache;
1512 afs_int32 hasNo64bit;
1513 struct AFSStoreStatus InStatus;
1517 int (*prepare)(void *rock, afs_uint32 size, afs_uint32 *bytestoxfer);
1518 int (*read)(void *rock, struct osi_file *tfile, afs_uint32 offset,
1519 afs_uint32 tlen, afs_uint32 *bytesread);
1520 int (*write)(void *rock, afs_uint32 tlen, afs_uint32 *byteswritten);
1521 int (*status)(void *rock);
1522 int (*padd)(void *rock, afs_uint32 tlen);
1523 int (*close)(void *rock, struct AFSFetchStatus *OutStatus,
1524 afs_int32 *doProcessFS);
1525 int (*destroy)(void **rock, afs_int32 error);
1526 int (*storeproc)(struct storeOps *, void *, struct dcache *, int *,
1531 int (*more)(void *rock, afs_int32 *length, afs_uint32 *moredata);
1532 int (*read)(void *rock, afs_uint32 tlen, afs_uint32 *bytesread);
1533 int (*write)(void *rock, struct osi_file *fp, afs_uint32 offset,
1534 afs_uint32 tlen, afs_uint32 *byteswritten);
1535 int (*close)(void *rock, struct vcache *avc, struct dcache *adc,
1536 struct afs_FetchOutput *Outputs);
1537 int (*destroy)(void **rock, afs_int32 error);
1540 /* fakestat support: opaque storage for afs_EvalFakeStat to remember
1541 * what vcache should be released.
1543 struct afs_fakestat_state {
1547 struct vcache *root_vp;
1550 extern int afs_fakestat_enable;
1551 extern int afs_rmtsys_enable;
1552 extern int afsd_dynamic_vcaches;
1555 * Wrappers for access to credentials structure members
1556 * Linux uses the kernel cred structure if available, with the
1557 * wrappers defined in LINUX/osi_machdep.h
1559 #if defined(AFS_NBSD40_ENV)
1560 /* in osi_machdep.h as expected */
1561 #elif defined (AFS_DARWIN110_ENV)
1562 #define afs_cr_uid(cred) kauth_cred_getuid((kauth_cred_t)(cred))
1563 #define afs_cr_gid(cred) kauth_cred_getgid((kauth_cred_t)(cred))
1564 #elif !(defined(AFS_LINUX26_ENV) && defined(STRUCT_TASK_STRUCT_HAS_CRED))
1565 #define afs_cr_uid(cred) ((cred)->cr_uid)
1566 #define afs_cr_gid(cred) ((cred)->cr_gid)
1567 #if !defined(AFS_OBSD_ENV)
1568 #define afs_cr_ruid(cred) ((cred)->cr_ruid)
1569 #define afs_cr_rgid(cred) ((cred)->cr_rgid)
1572 #if !defined(AFS_DARWIN110_ENV)
1574 afs_set_cr_uid(afs_ucred_t *cred, uid_t uid) {
1578 afs_set_cr_gid(afs_ucred_t *cred, gid_t gid) {
1581 #if !defined(AFS_OBSD_ENV)
1583 afs_set_cr_ruid(afs_ucred_t *cred, uid_t uid) {
1584 cred->cr_ruid = uid;
1587 afs_set_cr_rgid(afs_ucred_t *cred, gid_t gid) {
1588 cred->cr_rgid = gid;
1590 #endif /* ! AFS_OBSD_ENV */
1591 #endif /* ! AFS_DARWIN110_ENV */
1596 /** The 32 bit OS expects the members of this structure to be 32 bit
1597 * quantities and the 64 bit OS expects them as 64 bit quanties. Hence
1598 * to accomodate both, *long* is used instead of afs_int32
1610 extern int Afs_syscall(struct afssysa *uap, rval_t *rvp);
1611 #endif /* AFS_SUN5_ENV */
1614 * Various definitions for osi_sleep and its event hash table
1615 * DFBSD has no osi_sleep, and HPUX has its own hack for this stuff
1617 #define AFS_EVHASHSIZE 128 /* size of afs_evhasht, must be power of 2 */
1619 typedef struct afs_event {
1620 struct afs_event *next; /* next in hash chain */
1621 char *event; /* lwp event: an address */
1622 int refcount; /* Is it in use? */
1623 int seq; /* Sequence number: this is incremented
1624 * by wakeup calls; wait will not return until
1626 #if defined(AFS_AIX_ENV)
1628 #elif defined(AFS_DARWIN_ENV)
1629 # ifdef AFS_DARWIN80_ENV
1633 /* no cond member */
1634 #elif defined(AFS_FBSD_ENV) || defined(AFS_OBSD_ENV)
1635 int cond; /* "all this gluck should probably be replaced by CVs" */
1636 #elif defined(AFS_LINUX24_ENV)
1637 wait_queue_head_t cond;
1638 #elif defined(AFS_NBSD_ENV) || defined(AFS_SUN5_ENV) || defined(AFS_SGI_ENV)
1639 kcondvar_t cond; /* Currently associated condition variable */
1643 extern afs_event_t *afs_evhasht[AFS_EVHASHSIZE]; /* Hash table for events */
1645 #endif /* _AFS_H_ */