4 * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
6 * This program is free software; you can redistribute it and/or modify
7 * it under the terms of the GNU General Public License version 2 only,
8 * as published by the Free Software Foundation.
10 * This program is distributed in the hope that it will be useful, but
11 * WITHOUT ANY WARRANTY; without even the implied warranty of
12 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
13 * General Public License version 2 for more details (a copy is included
14 * in the LICENSE file that accompanied this code).
16 * You should have received a copy of the GNU General Public License
17 * version 2 along with this program; If not, see
18 * http://www.sun.com/software/products/lustre/docs/GPLv2.pdf
20 * Please contact Sun Microsystems, Inc., 4150 Network Circle, Santa Clara,
21 * CA 95054 USA or visit www.sun.com if you need additional information or
27 * Copyright (c) 2007, 2010, Oracle and/or its affiliates. All rights reserved.
28 * Use is subject to license terms.
30 * Copyright (c) 2011, 2012, Intel Corporation.
33 * This file is part of Lustre, http://www.lustre.org/
34 * Lustre is a trademark of Sun Microsystems, Inc.
40 #include <linux/obd.h>
42 #define IOC_OSC_TYPE 'h'
43 #define IOC_OSC_MIN_NR 20
44 #define IOC_OSC_SET_ACTIVE _IOWR(IOC_OSC_TYPE, 21, struct obd_device *)
45 #define IOC_OSC_MAX_NR 50
47 #define IOC_MDC_TYPE 'i'
48 #define IOC_MDC_MIN_NR 20
49 #define IOC_MDC_MAX_NR 50
51 #include <lustre/lustre_idl.h>
52 #include <lustre_lib.h>
53 #include <linux/libcfs/bitmap.h>
55 #include <lustre_export.h>
56 #include <lustre_fid.h>
57 #include <lustre_fld.h>
58 #include <lustre_capa.h>
60 #define MAX_OBD_DEVICES 8192
68 struct lov_oinfo { /* per-stripe data structure */
69 struct ost_id loi_oi; /* object ID/Sequence on the target OST */
70 int loi_ost_idx; /* OST stripe index in lov_tgt_desc->tgts */
71 int loi_ost_gen; /* generation of this loi_ost_idx */
73 unsigned long loi_kms_valid:1;
74 __u64 loi_kms; /* known minimum size */
75 struct ost_lvb loi_lvb;
76 struct osc_async_rc loi_ar;
79 static inline void loi_kms_set(struct lov_oinfo *oinfo, __u64 kms)
82 oinfo->loi_kms_valid = 1;
85 static inline void loi_init(struct lov_oinfo *loi)
89 struct lov_stripe_md {
92 pid_t lsm_lock_owner; /* debugging */
94 /* maximum possible file size, might change as OSTs status changes,
95 * e.g. disconnected, deactivated */
99 struct ost_id lw_object_oi; /* lov object id/seq */
101 /* LOV-private members start here -- only for use in lov/. */
103 __u32 lw_stripe_size; /* size of the stripe */
104 __u32 lw_pattern; /* striping pattern (RAID0, RAID1) */
105 __u16 lw_stripe_count; /* number of objects being striped over */
106 __u16 lw_layout_gen; /* generation of the layout */
107 char lw_pool_name[LOV_MAXPOOLNAME]; /* pool name */
110 struct lov_oinfo *lsm_oinfo[0];
113 #define lsm_oi lsm_wire.lw_object_oi
114 #define lsm_magic lsm_wire.lw_magic
115 #define lsm_layout_gen lsm_wire.lw_layout_gen
116 #define lsm_stripe_size lsm_wire.lw_stripe_size
117 #define lsm_pattern lsm_wire.lw_pattern
118 #define lsm_stripe_count lsm_wire.lw_stripe_count
119 #define lsm_pool_name lsm_wire.lw_pool_name
121 static inline bool lsm_is_released(struct lov_stripe_md *lsm)
123 return !!(lsm->lsm_pattern & LOV_PATTERN_F_RELEASED);
126 static inline bool lsm_has_objects(struct lov_stripe_md *lsm)
130 if (lsm_is_released(lsm))
137 typedef int (*obd_enqueue_update_f)(void *cookie, int rc);
139 /* obd info for a particular level (lov, osc). */
141 /* Lock policy. It keeps an extent which is specific for a particular
142 * OSC. (e.g. lov_prep_enqueue_set initialises extent of the policy,
143 * and osc_enqueue passes it into ldlm_lock_match & ldlm_cli_enqueue. */
144 ldlm_policy_data_t oi_policy;
145 /* Flags used for set request specific flags:
146 - while lock handling, the flags obtained on the enqueue
147 request are set here.
148 - while stats, the flags used for control delay/resend.
149 - while setattr, the flags used for distinguish punch operation
152 /* Lock handle specific for every OSC lock. */
153 struct lustre_handle *oi_lockh;
154 /* lsm data specific for every OSC. */
155 struct lov_stripe_md *oi_md;
156 /* obdo data specific for every OSC, if needed at all. */
158 /* statfs data specific for every OSC, if needed at all. */
159 struct obd_statfs *oi_osfs;
160 /* An update callback which is called to update some data on upper
161 * level. E.g. it is used for update lsm->lsm_oinfo at every recieved
162 * request in osc level for enqueue requests. It is also possible to
163 * update some caller data from LOV layer if needed. */
164 obd_enqueue_update_f oi_cb_up;
165 /* oss capability, its type is obd_capa in client to avoid copy.
166 * in contrary its type is lustre_capa in OSS. */
168 /* transfer jobid from ost_sync() to filter_sync()... */
172 /* compare all relevant fields. */
173 static inline int lov_stripe_md_cmp(struct lov_stripe_md *m1,
174 struct lov_stripe_md *m2)
177 * ->lsm_wire contains padding, but it should be zeroed out during
180 return memcmp(&m1->lsm_wire, &m2->lsm_wire, sizeof(m1->lsm_wire));
183 static inline int lov_lum_lsm_cmp(struct lov_user_md *lum,
184 struct lov_stripe_md *lsm)
186 if (lsm->lsm_magic != lum->lmm_magic)
188 if ((lsm->lsm_stripe_count != 0) && (lum->lmm_stripe_count != 0) &&
189 (lsm->lsm_stripe_count != lum->lmm_stripe_count))
191 if ((lsm->lsm_stripe_size != 0) && (lum->lmm_stripe_size != 0) &&
192 (lsm->lsm_stripe_size != lum->lmm_stripe_size))
194 if ((lsm->lsm_pattern != 0) && (lum->lmm_pattern != 0) &&
195 (lsm->lsm_pattern != lum->lmm_pattern))
197 if ((lsm->lsm_magic == LOV_MAGIC_V3) &&
198 (strncmp(lsm->lsm_pool_name,
199 ((struct lov_user_md_v3 *)lum)->lmm_pool_name,
200 LOV_MAXPOOLNAME) != 0))
205 static inline int lov_lum_swab_if_needed(struct lov_user_md_v3 *lumv3,
207 struct lov_user_md *lum)
209 if (lum && copy_from_user(lumv3, lum,sizeof(struct lov_user_md_v1)))
212 *lmm_magic = lumv3->lmm_magic;
214 if (*lmm_magic == __swab32(LOV_USER_MAGIC_V1)) {
215 lustre_swab_lov_user_md_v1((struct lov_user_md_v1 *)lumv3);
216 *lmm_magic = LOV_USER_MAGIC_V1;
217 } else if (*lmm_magic == LOV_USER_MAGIC_V3) {
218 if (lum && copy_from_user(lumv3, lum, sizeof(*lumv3)))
220 } else if (*lmm_magic == __swab32(LOV_USER_MAGIC_V3)) {
221 if (lum && copy_from_user(lumv3, lum, sizeof(*lumv3)))
223 lustre_swab_lov_user_md_v3(lumv3);
224 *lmm_magic = LOV_USER_MAGIC_V3;
225 } else if (*lmm_magic != LOV_USER_MAGIC_V1) {
227 "bad userland LOV MAGIC: %#08x != %#08x nor %#08x\n",
228 *lmm_magic, LOV_USER_MAGIC_V1, LOV_USER_MAGIC_V3);
234 void lov_stripe_lock(struct lov_stripe_md *md);
235 void lov_stripe_unlock(struct lov_stripe_md *md);
238 struct list_head typ_chain;
239 struct obd_ops *typ_dt_ops;
240 struct md_ops *typ_md_ops;
241 struct proc_dir_entry *typ_procroot;
244 struct lu_device_type *typ_lu;
245 spinlock_t obd_type_lock;
257 LLOG_CONFIG_ORIG_CTXT = 0,
258 LLOG_CONFIG_REPL_CTXT,
259 LLOG_MDS_OST_ORIG_CTXT,
260 LLOG_MDS_OST_REPL_CTXT,
267 LLOG_LOVEA_ORIG_CTXT,
268 LLOG_LOVEA_REPL_CTXT,
269 LLOG_CHANGELOG_ORIG_CTXT, /**< changelog generation on mdd */
270 LLOG_CHANGELOG_REPL_CTXT, /**< changelog access on clients */
271 LLOG_CHANGELOG_USER_ORIG_CTXT, /**< for multiple changelog consumers */
272 LLOG_AGENT_ORIG_CTXT, /**< agent requests generation on cdt */
276 struct timeout_item {
277 enum timeout_event ti_event;
278 cfs_time_t ti_timeout;
281 struct list_head ti_obd_list;
282 struct list_head ti_chain;
285 #define OSC_MAX_RIF_DEFAULT 8
286 #define MDS_OSC_MAX_RIF_DEFAULT 50
287 #define OSC_MAX_RIF_MAX 256
288 #define OSC_MAX_DIRTY_DEFAULT (OSC_MAX_RIF_DEFAULT * 4)
289 #define OSC_MAX_DIRTY_MB_MAX 2048 /* arbitrary, but < MAX_LONG bytes */
290 #define OSC_DEFAULT_RESENDS 10
292 /* possible values for fo_sync_lock_cancel */
294 NEVER_SYNC_ON_CANCEL = 0,
295 BLOCKING_SYNC_ON_CANCEL = 1,
296 ALWAYS_SYNC_ON_CANCEL = 2,
297 NUM_SYNC_ON_CANCEL_STATES
300 #define MDC_MAX_RIF_DEFAULT 8
301 #define MDC_MAX_RIF_MAX 512
306 struct rw_semaphore cl_sem;
307 struct obd_uuid cl_target_uuid;
308 struct obd_import *cl_import; /* ptlrpc connection state */
310 /* max_mds_easize is purely a performance thing so we don't have to
311 * call obd_size_diskmd() all the time. */
312 int cl_default_mds_easize;
313 int cl_max_mds_easize;
314 int cl_max_mds_cookiesize;
316 enum lustre_sec_part cl_sp_me;
317 enum lustre_sec_part cl_sp_to;
318 struct sptlrpc_flavor cl_flvr_mgc; /* fixed flavor of mgc->mgs */
320 /* the grant values are protected by loi_list_lock below */
321 long cl_dirty; /* all _dirty_ in bytes */
322 long cl_dirty_max; /* allowed w/o rpc */
323 long cl_dirty_transit; /* dirty synchronous */
324 long cl_avail_grant; /* bytes of credit for ost */
325 long cl_lost_grant; /* lost credits (trunc) */
327 /* since we allocate grant by blocks, we don't know how many grant will
328 * be used to add a page into cache. As a solution, we reserve maximum
329 * grant before trying to dirty a page and unreserve the rest.
330 * See osc_{reserve|unreserve}_grant for details. */
331 long cl_reserved_grant;
332 struct list_head cl_cache_waiters; /* waiting for cache/grant */
333 cfs_time_t cl_next_shrink_grant; /* jiffies */
334 struct list_head cl_grant_shrink_list; /* Timeout event list */
335 int cl_grant_shrink_interval; /* seconds */
337 /* A chunk is an optimal size used by osc_extent to determine
338 * the extent size. A chunk is max(PAGE_CACHE_SIZE, OST block size) */
341 int cl_extent_tax; /* extent overhead, by bytes */
343 /* keep track of objects that have lois that contain pages which
344 * have been queued for async brw. this lock also protects the
345 * lists of osc_client_pages that hang off of the loi */
347 * ->cl_loi_list_lock protects consistency of
348 * ->cl_loi_{ready,read,write}_list. ->ap_make_ready() and
349 * ->ap_completion() call-backs are executed under this lock. As we
350 * cannot guarantee that these call-backs never block on all platforms
351 * (as a matter of fact they do block on Mac OS X), type of
352 * ->cl_loi_list_lock is platform dependent: it's a spin-lock on Linux
353 * and blocking mutex on Mac OS X. (Alternative is to make this lock
354 * blocking everywhere, but we don't want to slow down fast-path of
355 * our main platform.)
357 * Exact type of ->cl_loi_list_lock is defined in arch/obd.h together
358 * with client_obd_list_{un,}lock() and
359 * client_obd_list_lock_{init,done}() functions.
361 * NB by Jinshan: though field names are still _loi_, but actually
362 * osc_object{}s are in the list.
364 client_obd_lock_t cl_loi_list_lock;
365 struct list_head cl_loi_ready_list;
366 struct list_head cl_loi_hp_ready_list;
367 struct list_head cl_loi_write_list;
368 struct list_head cl_loi_read_list;
371 /* just a sum of the loi/lop pending numbers to be exported by /proc */
372 atomic_t cl_pending_w_pages;
373 atomic_t cl_pending_r_pages;
374 __u32 cl_max_pages_per_rpc;
375 int cl_max_rpcs_in_flight;
376 struct obd_histogram cl_read_rpc_hist;
377 struct obd_histogram cl_write_rpc_hist;
378 struct obd_histogram cl_read_page_hist;
379 struct obd_histogram cl_write_page_hist;
380 struct obd_histogram cl_read_offset_hist;
381 struct obd_histogram cl_write_offset_hist;
383 /* lru for osc caching pages */
384 struct cl_client_cache *cl_cache;
385 struct list_head cl_lru_osc; /* member of cl_cache->ccc_lru */
386 atomic_t *cl_lru_left;
387 atomic_t cl_lru_busy;
388 atomic_t cl_lru_shrinkers;
389 atomic_t cl_lru_in_list;
390 struct list_head cl_lru_list; /* lru page list */
391 client_obd_lock_t cl_lru_list_lock; /* page list protector */
393 /* number of in flight destroy rpcs is limited to max_rpcs_in_flight */
394 atomic_t cl_destroy_in_flight;
395 wait_queue_head_t cl_destroy_waitq;
397 struct mdc_rpc_lock *cl_rpc_lock;
398 struct mdc_rpc_lock *cl_close_lock;
401 struct semaphore cl_mgc_sem;
402 struct local_oid_storage *cl_mgc_los;
403 struct dt_object *cl_mgc_configs_dir;
404 atomic_t cl_mgc_refcount;
405 struct obd_export *cl_mgc_mgsexp;
407 /* checksumming for data sent over the network */
408 unsigned int cl_checksum:1; /* 0 = disabled, 1 = enabled */
409 /* supported checksum types that are worked out at connect time */
410 __u32 cl_supp_cksum_types;
411 /* checksum algorithm to be used */
412 cksum_type_t cl_cksum_type;
414 /* also protected by the poorly named _loi_list_lock lock above */
415 struct osc_async_rc cl_ar;
417 /* used by quotacheck when the servers are older than 2.4 */
418 int cl_qchk_stat; /* quotacheck stat of the peer */
419 #define CL_NOT_QUOTACHECKED 1 /* client->cl_qchk_stat init value */
420 #if LUSTRE_VERSION_CODE >= OBD_OCD_VERSION(2, 7, 50, 0)
421 #warning "please consider removing quotacheck compatibility code"
424 /* sequence manager */
425 struct lu_client_seq *cl_seq;
427 atomic_t cl_resends; /* resend count */
429 /* ptlrpc work for writeback in ptlrpcd context */
430 void *cl_writeback_work;
431 /* hash tables for osc_quota_info */
432 struct cfs_hash *cl_quota_hash[MAXQUOTAS];
434 #define obd2cli_tgt(obd) ((char *)(obd)->u.cli.cl_target_uuid.uuid)
441 struct echo_client_obd {
442 struct obd_export *ec_exp; /* the local connection to osc/lov */
444 struct list_head ec_objects;
445 struct list_head ec_locks;
451 struct obd_uuid lqo_uuid; /* ptlrpc's c_remote_uuid */
452 struct list_head lqo_oss_list; /* link to lov_qos */
453 __u64 lqo_bavail; /* total bytes avail on OSS */
454 __u64 lqo_penalty; /* current penalty */
455 __u64 lqo_penalty_per_obj;/* penalty decrease every obj*/
456 time_t lqo_used; /* last used time, seconds */
457 __u32 lqo_ost_count; /* number of osts on this oss */
461 struct lov_qos_oss *ltq_oss; /* oss info */
462 __u64 ltq_penalty; /* current penalty */
463 __u64 ltq_penalty_per_obj; /* penalty decrease every obj*/
464 __u64 ltq_weight; /* net weighting */
465 time_t ltq_used; /* last used time, seconds */
466 unsigned int ltq_usable:1; /* usable for striping */
469 /* Generic subset of OSTs */
471 __u32 *op_array; /* array of index of
473 unsigned int op_count; /* number of OSTs in the array */
474 unsigned int op_size; /* allocated size of lp_array */
475 struct rw_semaphore op_rw_sem; /* to protect ost_pool use */
478 /* Round-robin allocator data */
480 __u32 lqr_start_idx; /* start index of new inode */
481 __u32 lqr_offset_idx; /* aliasing for start_idx */
482 int lqr_start_count; /* reseed counter */
483 struct ost_pool lqr_pool; /* round-robin optimized list */
484 unsigned long lqr_dirty:1; /* recalc round-robin list */
487 /* allow statfs data caching for 1 second */
488 #define OBD_STATFS_CACHE_SECONDS 1
490 struct lov_statfs_data {
491 struct obd_info lsd_oi;
492 struct obd_statfs lsd_statfs;
494 /* Stripe placement optimization */
496 struct list_head lq_oss_list; /* list of OSSs that targets use */
497 struct rw_semaphore lq_rw_sem;
498 __u32 lq_active_oss_count;
499 unsigned int lq_prio_free; /* priority for free space */
500 unsigned int lq_threshold_rr;/* priority for rr */
501 struct lov_qos_rr lq_rr; /* round robin qos data */
502 unsigned long lq_dirty:1, /* recalc qos data */
503 lq_same_space:1,/* the ost's all have approx.
504 the same space avail */
505 lq_reset:1, /* zero current penalties */
506 lq_statfs_in_progress:1; /* statfs op in
508 /* qos statfs data */
509 struct lov_statfs_data *lq_statfs_data;
510 wait_queue_head_t lq_statfs_waitq; /* waitqueue to notify statfs
511 * requests completion */
514 struct lov_tgt_desc {
515 struct list_head ltd_kill;
516 struct obd_uuid ltd_uuid;
517 struct obd_device *ltd_obd;
518 struct obd_export *ltd_exp;
519 struct ltd_qos ltd_qos; /* qos info per target */
521 __u32 ltd_index; /* index in lov_obd->tgts */
522 unsigned long ltd_active:1,/* is this target up for requests */
523 ltd_activate:1,/* should target be activated */
524 ltd_reap:1; /* should this target be deleted */
528 #define pool_tgt_size(_p) _p->pool_obds.op_size
529 #define pool_tgt_count(_p) _p->pool_obds.op_count
530 #define pool_tgt_array(_p) _p->pool_obds.op_array
531 #define pool_tgt_rw_sem(_p) _p->pool_obds.op_rw_sem
534 char pool_name[LOV_MAXPOOLNAME + 1]; /* name of pool */
535 struct ost_pool pool_obds; /* pool members */
536 atomic_t pool_refcount; /* pool ref. counter */
537 struct lov_qos_rr pool_rr; /* round robin qos */
538 struct hlist_node pool_hash; /* access by poolname */
539 struct list_head pool_list; /* serial access */
540 struct proc_dir_entry *pool_proc_entry; /* file in /proc */
541 struct obd_device *pool_lobd; /* obd of the lov/lod to which
542 * this pool belongs */
546 struct lov_desc desc;
547 struct lov_tgt_desc **lov_tgts; /* sparse array */
548 struct ost_pool lov_packed; /* all OSTs in a packed
550 struct mutex lov_lock;
551 struct obd_connect_data lov_ocd;
552 atomic_t lov_refcount;
553 __u32 lov_tgt_count; /* how many OBD's */
554 __u32 lov_active_tgt_count; /* how many active */
555 __u32 lov_death_row;/* tgts scheduled to be deleted */
556 __u32 lov_tgt_size; /* size of tgts array */
559 struct cfs_hash *lov_pools_hash_body; /* used for key access */
560 struct list_head lov_pool_list; /* used for sequential access */
561 struct proc_dir_entry *lov_pool_proc_entry;
562 enum lustre_sec_part lov_sp_me;
564 /* Cached LRU pages from upper layer */
567 struct rw_semaphore lov_notify_lock;
570 struct lmv_tgt_desc {
571 struct obd_uuid ltd_uuid;
572 struct obd_export *ltd_exp;
574 struct mutex ltd_fid_mutex;
575 unsigned long ltd_active:1; /* target up for requests */
578 enum placement_policy {
579 PLACEMENT_CHAR_POLICY = 0,
580 PLACEMENT_NID_POLICY = 1,
581 PLACEMENT_INVAL_POLICY = 2,
585 typedef enum placement_policy placement_policy_t;
589 struct lu_client_fld lmv_fld;
591 placement_policy_t lmv_placement;
592 struct lmv_desc desc;
593 struct obd_uuid cluuid;
594 struct obd_export *exp;
596 struct mutex init_mutex;
603 int tgts_size; /* size of tgts array */
604 struct lmv_tgt_desc **tgts;
606 struct obd_connect_data conn_data;
609 struct niobuf_local {
610 __u64 lnb_file_offset;
611 __u32 lnb_page_offset;
615 struct dentry *dentry;
620 #define LUSTRE_FLD_NAME "fld"
621 #define LUSTRE_SEQ_NAME "seq"
623 #define LUSTRE_MDD_NAME "mdd"
624 #define LUSTRE_OSD_LDISKFS_NAME "osd-ldiskfs"
625 #define LUSTRE_OSD_ZFS_NAME "osd-zfs"
626 #define LUSTRE_VVP_NAME "vvp"
627 #define LUSTRE_LMV_NAME "lmv"
628 #define LUSTRE_SLP_NAME "slp"
629 #define LUSTRE_LOD_NAME "lod"
630 #define LUSTRE_OSP_NAME "osp"
631 #define LUSTRE_LWP_NAME "lwp"
633 /* obd device type names */
634 /* FIXME all the references to LUSTRE_MDS_NAME should be swapped with LUSTRE_MDT_NAME */
635 #define LUSTRE_MDS_NAME "mds"
636 #define LUSTRE_MDT_NAME "mdt"
637 #define LUSTRE_MDC_NAME "mdc"
638 #define LUSTRE_OSS_NAME "ost" /* FIXME change name to oss */
639 #define LUSTRE_OST_NAME "obdfilter" /* FIXME change name to ost */
640 #define LUSTRE_OSC_NAME "osc"
641 #define LUSTRE_LOV_NAME "lov"
642 #define LUSTRE_MGS_NAME "mgs"
643 #define LUSTRE_MGC_NAME "mgc"
645 #define LUSTRE_ECHO_NAME "obdecho"
646 #define LUSTRE_ECHO_CLIENT_NAME "echo_client"
647 #define LUSTRE_QMT_NAME "qmt"
649 /* Constant obd names (post-rename) */
650 #define LUSTRE_MDS_OBDNAME "MDS"
651 #define LUSTRE_OSS_OBDNAME "OSS"
652 #define LUSTRE_MGS_OBDNAME "MGS"
653 #define LUSTRE_MGC_OBDNAME "MGC"
655 static inline int is_osp_on_mdt(char *name)
659 ptr = strrchr(name, '-');
661 CERROR("%s is not a obdname\n", name);
665 /* 1.8 OSC/OSP name on MDT is fsname-OSTxxxx-osc */
666 if (strncmp(ptr + 1, "osc", 3) == 0)
669 if (strncmp(ptr + 1, "MDT", 3) != 0)
672 while (*(--ptr) != '-' && ptr != name);
677 if (strncmp(ptr + 1, LUSTRE_OSP_NAME, strlen(LUSTRE_OSP_NAME)) != 0 &&
678 strncmp(ptr + 1, LUSTRE_OSC_NAME, strlen(LUSTRE_OSC_NAME)) != 0)
684 /* Don't conflict with on-wire flags OBD_BRW_WRITE, etc */
685 #define N_LOCAL_TEMP_PAGE 0x10000000
687 struct obd_trans_info {
690 /* Only used on the server side for tracking acks. */
691 struct oti_req_ack_lock {
692 struct lustre_handle lock;
696 struct llog_cookie oti_onecookie;
697 struct llog_cookie *oti_logcookies;
699 /** synchronous write is needed */
700 unsigned long oti_sync_write:1;
702 /* initial thread handling transaction */
703 struct ptlrpc_thread * oti_thread;
706 __u64 oti_pre_version;
710 struct obd_uuid *oti_ost_uuid;
713 static inline void oti_init(struct obd_trans_info *oti,
714 struct ptlrpc_request *req)
718 memset(oti, 0, sizeof(*oti));
723 oti->oti_xid = req->rq_xid;
724 /** VBR: take versions from request */
725 if (req->rq_reqmsg != NULL &&
726 lustre_msg_get_flags(req->rq_reqmsg) & MSG_REPLAY) {
727 __u64 *pre_version = lustre_msg_get_versions(req->rq_reqmsg);
728 oti->oti_pre_version = pre_version ? pre_version[0] : 0;
729 oti->oti_transno = lustre_msg_get_transno(req->rq_reqmsg);
732 /** called from mds_create_objects */
733 if (req->rq_repmsg != NULL)
734 oti->oti_transno = lustre_msg_get_transno(req->rq_repmsg);
735 oti->oti_thread = req->rq_svc_thread;
736 if (req->rq_reqmsg != NULL)
737 oti->oti_conn_cnt = lustre_msg_get_conn_cnt(req->rq_reqmsg);
740 static inline void oti_alloc_cookies(struct obd_trans_info *oti,int num_cookies)
745 if (num_cookies == 1)
746 oti->oti_logcookies = &oti->oti_onecookie;
748 OBD_ALLOC_LARGE(oti->oti_logcookies,
749 num_cookies * sizeof(oti->oti_onecookie));
751 oti->oti_numcookies = num_cookies;
754 static inline void oti_free_cookies(struct obd_trans_info *oti)
756 if (!oti || !oti->oti_logcookies)
759 if (oti->oti_logcookies == &oti->oti_onecookie)
760 LASSERT(oti->oti_numcookies == 1);
762 OBD_FREE_LARGE(oti->oti_logcookies,
763 oti->oti_numcookies*sizeof(oti->oti_onecookie));
764 oti->oti_logcookies = NULL;
765 oti->oti_numcookies = 0;
769 * Events signalled through obd_notify() upcall-chain.
771 enum obd_notify_event {
774 /* Device connect start */
776 /* Device activated */
778 /* Device deactivated */
780 /* Device disconnected */
782 /* Connect data for import were changed */
785 OBD_NOTIFY_SYNC_NONBLOCK,
787 /* Configuration event */
789 /* Administratively deactivate/activate event */
790 OBD_NOTIFY_DEACTIVATE,
795 * Data structure used to pass obd_notify()-event to non-obd listeners (llite
796 * and liblustre being main examples).
798 struct obd_notify_upcall {
799 int (*onu_upcall)(struct obd_device *host, struct obd_device *watched,
800 enum obd_notify_event ev, void *owner, void *data);
801 /* Opaque datum supplied by upper layer listener */
805 struct target_recovery_data {
806 svc_handler_t trd_recovery_handler;
807 pid_t trd_processing_task;
808 struct completion trd_starting;
809 struct completion trd_finishing;
812 struct obd_llog_group {
814 struct llog_ctxt *olg_ctxts[LLOG_MAX_CTXTS];
815 wait_queue_head_t olg_waitq;
817 struct mutex olg_cat_processing;
820 /* corresponds to one of the obd's */
821 #define OBD_DEVICE_MAGIC 0XAB5CD6EF
822 #define OBD_DEV_BY_DEVNAME 0xffffd0de
825 struct obd_type *obd_type;
828 /* common and UUID name of this device */
829 char obd_name[MAX_OBD_NAME];
830 struct obd_uuid obd_uuid;
832 struct lu_device *obd_lu_dev;
835 /* bitfield modification is protected by obd_dev_lock */
836 unsigned long obd_attached:1, /* finished attach */
837 obd_set_up:1, /* finished setup */
838 obd_recovering:1, /* there are recoverable clients */
839 obd_abort_recovery:1,/* recovery expired */
840 obd_version_recov:1, /* obd uses version checking */
841 obd_replayable:1, /* recovery is enabled; inform clients */
842 obd_no_transno:1, /* no committed-transno notification */
843 obd_no_recov:1, /* fail instead of retry messages */
844 obd_stopping:1, /* started cleanup */
845 obd_starting:1, /* started setup */
846 obd_force:1, /* cleanup with > 0 obd refcount */
847 obd_fail:1, /* cleanup with failover */
848 obd_async_recov:1, /* allow asynchronous orphan cleanup */
849 obd_no_conn:1, /* deny new connections */
850 obd_inactive:1, /* device active/inactive
851 * (for /proc/status only!!) */
852 obd_no_ir:1, /* no imperative recovery. */
853 obd_process_conf:1; /* device is processing mgs config */
854 /* use separate field as it is set in interrupt to don't mess with
855 * protection of other bits using _bh lock */
856 unsigned long obd_recovery_expired:1;
857 /* uuid-export hash body */
858 struct cfs_hash *obd_uuid_hash;
859 /* nid-export hash body */
860 struct cfs_hash *obd_nid_hash;
862 struct cfs_hash *obd_nid_stats_hash;
863 struct list_head obd_nid_stats;
864 atomic_t obd_refcount;
865 wait_queue_head_t obd_refcount_waitq;
866 struct list_head obd_exports;
867 struct list_head obd_unlinked_exports;
868 struct list_head obd_delayed_exports;
870 spinlock_t obd_nid_lock;
871 struct ldlm_namespace *obd_namespace;
872 struct ptlrpc_client obd_ldlm_client; /* XXX OST/MDS only */
873 /* a spinlock is OK for what we do now, may need a semaphore later */
874 spinlock_t obd_dev_lock; /* protect OBD bitfield above */
875 struct mutex obd_dev_mutex;
876 __u64 obd_last_committed;
877 struct fsfilt_operations *obd_fsops;
878 spinlock_t obd_osfs_lock;
879 struct obd_statfs obd_osfs; /* locked by obd_osfs_lock */
881 struct lvfs_run_ctxt obd_lvfs_ctxt;
882 struct obd_llog_group obd_olg; /* default llog group */
883 struct obd_device *obd_observer;
884 struct rw_semaphore obd_observer_link_sem;
885 struct obd_notify_upcall obd_upcall;
886 struct obd_export *obd_self_export;
887 /* list of exports in LRU order, for ping evictor, with obd_dev_lock */
888 struct list_head obd_exports_timed;
889 time_t obd_eviction_timer; /* for ping evictor */
891 int obd_max_recoverable_clients;
892 atomic_t obd_connected_clients;
893 int obd_stale_clients;
894 int obd_delayed_clients;
895 /* this lock protects all recovery list_heads, timer and
896 * obd_next_recovery_transno value */
897 spinlock_t obd_recovery_task_lock;
898 __u64 obd_next_recovery_transno;
899 int obd_replayed_requests;
900 int obd_requests_queued_for_recovery;
901 wait_queue_head_t obd_next_transno_waitq;
902 /* protected by obd_recovery_task_lock */
903 struct timer_list obd_recovery_timer;
904 time_t obd_recovery_start; /* seconds */
905 time_t obd_recovery_end; /* seconds, for lprocfs_status */
906 int obd_recovery_time_hard;
907 int obd_recovery_timeout;
908 int obd_recovery_ir_factor;
910 /* new recovery stuff from CMD2 */
911 struct target_recovery_data obd_recovery_data;
912 int obd_replayed_locks;
913 atomic_t obd_req_replay_clients;
914 atomic_t obd_lock_replay_clients;
915 /* all lists are protected by obd_recovery_task_lock */
916 struct list_head obd_req_replay_queue;
917 struct list_head obd_lock_replay_queue;
918 struct list_head obd_final_req_queue;
919 int obd_recovery_stage;
922 struct client_obd cli;
923 struct echo_client_obd echo_client;
927 /* Fields used by LProcFS */
928 unsigned int obd_cntr_base;
929 struct lprocfs_stats *obd_stats;
931 unsigned int md_cntr_base;
932 struct lprocfs_stats *md_stats;
934 struct proc_dir_entry *obd_proc_entry;
935 void *obd_proc_private; /* type private PDEs */
936 struct proc_dir_entry *obd_proc_exports_entry;
937 struct proc_dir_entry *obd_svc_procroot;
938 struct lprocfs_stats *obd_svc_stats;
939 atomic_t obd_evict_inprogress;
940 wait_queue_head_t obd_evict_inprogress_waitq;
941 struct list_head obd_evict_list; /* protected with pet_lock */
944 * Ldlm pool part. Save last calculated SLV and Limit.
946 rwlock_t obd_pool_lock;
951 * A list of outstanding class_incref()'s against this obd. For
954 struct lu_ref obd_reference;
956 int obd_conn_inprogress;
959 #define OBD_LLOG_FL_SENDNOW 0x0001
960 #define OBD_LLOG_FL_EXIT 0x0002
962 enum obd_cleanup_stage {
963 /* Special case hack for MDS LOVs */
965 /* can be directly mapped to .ldto_device_fini() */
969 /* get/set_info keys */
970 #define KEY_ASYNC "async"
971 #define KEY_BLOCKSIZE_BITS "blocksize_bits"
972 #define KEY_BLOCKSIZE "blocksize"
973 #define KEY_CAPA_KEY "capa_key"
974 #define KEY_CHANGELOG_CLEAR "changelog_clear"
975 #define KEY_FID2PATH "fid2path"
976 #define KEY_CHECKSUM "checksum"
977 #define KEY_CLEAR_FS "clear_fs"
978 #define KEY_CONN_DATA "conn_data"
979 #define KEY_EVICT_BY_NID "evict_by_nid"
980 #define KEY_FIEMAP "fiemap"
981 #define KEY_FLUSH_CTX "flush_ctx"
982 #define KEY_GRANT_SHRINK "grant_shrink"
983 #define KEY_HSM_COPYTOOL_SEND "hsm_send"
984 #define KEY_INIT_RECOV_BACKUP "init_recov_bk"
985 #define KEY_INIT_RECOV "initial_recov"
986 #define KEY_INTERMDS "inter_mds"
987 #define KEY_LAST_ID "last_id"
988 #define KEY_LAST_FID "last_fid"
989 #define KEY_LOCK_TO_STRIPE "lock_to_stripe"
990 #define KEY_LOVDESC "lovdesc"
991 #define KEY_LOV_IDX "lov_idx"
992 #define KEY_MAX_EASIZE "max_easize"
993 #define KEY_MDS_CONN "mds_conn"
994 #define KEY_MGSSEC "mgssec"
995 #define KEY_NEXT_ID "next_id"
996 #define KEY_READ_ONLY "read-only"
997 #define KEY_REGISTER_TARGET "register_target"
998 #define KEY_SET_FS "set_fs"
999 #define KEY_TGT_COUNT "tgt_count"
1000 /* KEY_SET_INFO in lustre_idl.h */
1001 #define KEY_SPTLRPC_CONF "sptlrpc_conf"
1002 #define KEY_CONNECT_FLAG "connect_flags"
1003 #define KEY_SYNC_LOCK_CANCEL "sync_lock_cancel"
1005 #define KEY_CACHE_SET "cache_set"
1006 #define KEY_CACHE_LRU_SHRINK "cache_lru_shrink"
1007 #define KEY_CHANGELOG_INDEX "changelog_index"
1011 /* /!\ must be coherent with include/linux/namei.h on patched kernel */
1012 #define IT_OPEN (1 << 0)
1013 #define IT_CREAT (1 << 1)
1014 #define IT_READDIR (1 << 2)
1015 #define IT_GETATTR (1 << 3)
1016 #define IT_LOOKUP (1 << 4)
1017 #define IT_UNLINK (1 << 5)
1018 #define IT_TRUNC (1 << 6)
1019 #define IT_GETXATTR (1 << 7)
1020 #define IT_EXEC (1 << 8)
1021 #define IT_PIN (1 << 9)
1022 #define IT_LAYOUT (1 << 10)
1023 #define IT_QUOTA_DQACQ (1 << 11)
1024 #define IT_QUOTA_CONN (1 << 12)
1026 static inline int it_to_lock_mode(struct lookup_intent *it)
1028 /* CREAT needs to be tested before open (both could be set) */
1029 if (it->it_op & IT_CREAT)
1031 else if (it->it_op & (IT_READDIR | IT_GETATTR | IT_OPEN | IT_LOOKUP |
1035 LASSERTF(0, "Invalid it_op: %d\n", it->it_op);
1040 struct lu_fid op_fid1; /* operation fid1 (usualy parent) */
1041 struct lu_fid op_fid2; /* operation fid2 (usualy child) */
1042 struct lu_fid op_fid3; /* 2 extra fids to find conflicting */
1043 struct lu_fid op_fid4; /* to the operation locks. */
1044 mdsno_t op_mds; /* what mds server open will go to */
1045 struct lustre_handle op_handle;
1046 obd_time op_mod_time;
1047 const char *op_name;
1050 struct lmv_stripe_md *op_mea1;
1051 struct lmv_stripe_md *op_mea2;
1052 __u32 op_suppgids[2];
1058 /* iattr fields and blocks. */
1059 struct iattr op_attr;
1060 unsigned int op_attr_flags;
1062 loff_t op_attr_blocks;
1064 /* Size-on-MDS epoch and flags. */
1069 struct obd_capa *op_capa1;
1070 struct obd_capa *op_capa2;
1072 /* Various operation flags. */
1073 enum mds_op_bias op_bias;
1075 /* Operation type */
1078 /* Used by readdir */
1081 /* Used by readdir */
1084 /* used to transfer info between the stacks of MD client
1085 * see enum op_cli_flags */
1088 /* File object data version for HSM release, on client */
1089 __u64 op_data_version;
1090 struct lustre_handle op_lease_handle;
1094 CLI_SET_MEA = 1 << 0,
1095 CLI_RM_ENTRY = 1 << 1,
1098 struct md_enqueue_info;
1099 /* metadata stat-ahead */
1100 typedef int (* md_enqueue_cb_t)(struct ptlrpc_request *req,
1101 struct md_enqueue_info *minfo,
1104 struct md_enqueue_info {
1105 struct md_op_data mi_data;
1106 struct lookup_intent mi_it;
1107 struct lustre_handle mi_lockh;
1108 struct inode *mi_dir;
1109 md_enqueue_cb_t mi_cb;
1111 unsigned int mi_generation;
1115 struct module *o_owner;
1116 int (*o_iocontrol)(unsigned int cmd, struct obd_export *exp, int len,
1117 void *karg, void *uarg);
1118 int (*o_get_info)(const struct lu_env *env, struct obd_export *,
1119 __u32 keylen, void *key, __u32 *vallen, void *val,
1120 struct lov_stripe_md *lsm);
1121 int (*o_set_info_async)(const struct lu_env *, struct obd_export *,
1122 __u32 keylen, void *key,
1123 __u32 vallen, void *val,
1124 struct ptlrpc_request_set *set);
1125 int (*o_attach)(struct obd_device *dev, obd_count len, void *data);
1126 int (*o_detach)(struct obd_device *dev);
1127 int (*o_setup) (struct obd_device *dev, struct lustre_cfg *cfg);
1128 int (*o_precleanup)(struct obd_device *dev,
1129 enum obd_cleanup_stage cleanup_stage);
1130 int (*o_cleanup)(struct obd_device *dev);
1131 int (*o_process_config)(struct obd_device *dev, obd_count len,
1133 int (*o_postrecov)(struct obd_device *dev);
1134 int (*o_add_conn)(struct obd_import *imp, struct obd_uuid *uuid,
1136 int (*o_del_conn)(struct obd_import *imp, struct obd_uuid *uuid);
1137 /* connect to the target device with given connection
1138 * data. @ocd->ocd_connect_flags is modified to reflect flags actually
1139 * granted by the target, which are guaranteed to be a subset of flags
1140 * asked for. If @ocd == NULL, use default parameters. */
1141 int (*o_connect)(const struct lu_env *env,
1142 struct obd_export **exp, struct obd_device *src,
1143 struct obd_uuid *cluuid, struct obd_connect_data *ocd,
1145 int (*o_reconnect)(const struct lu_env *env,
1146 struct obd_export *exp, struct obd_device *src,
1147 struct obd_uuid *cluuid,
1148 struct obd_connect_data *ocd,
1150 int (*o_disconnect)(struct obd_export *exp);
1152 /* Initialize/finalize fids infrastructure. */
1153 int (*o_fid_init)(struct obd_device *obd,
1154 struct obd_export *exp, enum lu_cli_type type);
1155 int (*o_fid_fini)(struct obd_device *obd);
1157 /* Allocate new fid according to passed @hint. */
1158 int (*o_fid_alloc)(struct obd_export *exp, struct lu_fid *fid,
1159 struct md_op_data *op_data);
1162 * Object with @fid is getting deleted, we may want to do something
1165 int (*o_statfs)(const struct lu_env *, struct obd_export *exp,
1166 struct obd_statfs *osfs, __u64 max_age, __u32 flags);
1167 int (*o_statfs_async)(struct obd_export *exp, struct obd_info *oinfo,
1168 __u64 max_age, struct ptlrpc_request_set *set);
1169 int (*o_packmd)(struct obd_export *exp, struct lov_mds_md **disk_tgt,
1170 struct lov_stripe_md *mem_src);
1171 int (*o_unpackmd)(struct obd_export *exp,struct lov_stripe_md **mem_tgt,
1172 struct lov_mds_md *disk_src, int disk_len);
1173 int (*o_preallocate)(struct lustre_handle *, obd_count *req,
1175 /* FIXME: add fid capability support for create & destroy! */
1176 int (*o_precreate)(struct obd_export *exp);
1177 int (*o_create)(const struct lu_env *env, struct obd_export *exp,
1178 struct obdo *oa, struct lov_stripe_md **ea,
1179 struct obd_trans_info *oti);
1180 int (*o_create_async)(struct obd_export *exp, struct obd_info *oinfo,
1181 struct lov_stripe_md **ea,
1182 struct obd_trans_info *oti);
1183 int (*o_destroy)(const struct lu_env *env, struct obd_export *exp,
1184 struct obdo *oa, struct lov_stripe_md *ea,
1185 struct obd_trans_info *oti, struct obd_export *md_exp,
1187 int (*o_setattr)(const struct lu_env *, struct obd_export *exp,
1188 struct obd_info *oinfo, struct obd_trans_info *oti);
1189 int (*o_setattr_async)(struct obd_export *exp, struct obd_info *oinfo,
1190 struct obd_trans_info *oti,
1191 struct ptlrpc_request_set *rqset);
1192 int (*o_getattr)(const struct lu_env *env, struct obd_export *exp,
1193 struct obd_info *oinfo);
1194 int (*o_getattr_async)(struct obd_export *exp, struct obd_info *oinfo,
1195 struct ptlrpc_request_set *set);
1196 int (*o_brw)(int rw, struct obd_export *exp, struct obd_info *oinfo,
1197 obd_count oa_bufs, struct brw_page *pgarr,
1198 struct obd_trans_info *oti);
1199 int (*o_merge_lvb)(struct obd_export *exp, struct lov_stripe_md *lsm,
1200 struct ost_lvb *lvb, int kms_only);
1201 int (*o_adjust_kms)(struct obd_export *exp, struct lov_stripe_md *lsm,
1202 obd_off size, int shrink);
1203 int (*o_punch)(const struct lu_env *, struct obd_export *exp,
1204 struct obd_info *oinfo, struct obd_trans_info *oti,
1205 struct ptlrpc_request_set *rqset);
1206 int (*o_sync)(const struct lu_env *env, struct obd_export *exp,
1207 struct obd_info *oinfo, obd_size start, obd_size end,
1208 struct ptlrpc_request_set *set);
1209 int (*o_migrate)(struct lustre_handle *conn, struct lov_stripe_md *dst,
1210 struct lov_stripe_md *src, obd_size start,
1211 obd_size end, struct obd_trans_info *oti);
1212 int (*o_copy)(struct lustre_handle *dstconn, struct lov_stripe_md *dst,
1213 struct lustre_handle *srconn, struct lov_stripe_md *src,
1214 obd_size start, obd_size end, struct obd_trans_info *);
1215 int (*o_iterate)(struct lustre_handle *conn,
1216 int (*)(obd_id, obd_seq, void *),
1217 obd_id *startid, obd_seq seq, void *data);
1218 int (*o_preprw)(const struct lu_env *env, int cmd,
1219 struct obd_export *exp, struct obdo *oa, int objcount,
1220 struct obd_ioobj *obj, struct niobuf_remote *remote,
1221 int *nr_pages, struct niobuf_local *local,
1222 struct obd_trans_info *oti, struct lustre_capa *capa);
1223 int (*o_commitrw)(const struct lu_env *env, int cmd,
1224 struct obd_export *exp, struct obdo *oa,
1225 int objcount, struct obd_ioobj *obj,
1226 struct niobuf_remote *remote, int pages,
1227 struct niobuf_local *local,
1228 struct obd_trans_info *oti, int rc);
1229 int (*o_enqueue)(struct obd_export *, struct obd_info *oinfo,
1230 struct ldlm_enqueue_info *einfo,
1231 struct ptlrpc_request_set *rqset);
1232 int (*o_change_cbdata)(struct obd_export *, struct lov_stripe_md *,
1233 ldlm_iterator_t it, void *data);
1234 int (*o_find_cbdata)(struct obd_export *, struct lov_stripe_md *,
1235 ldlm_iterator_t it, void *data);
1236 int (*o_cancel)(struct obd_export *, struct lov_stripe_md *md,
1237 __u32 mode, struct lustre_handle *);
1238 int (*o_cancel_unused)(struct obd_export *, struct lov_stripe_md *,
1239 ldlm_cancel_flags_t flags, void *opaque);
1240 int (*o_init_export)(struct obd_export *exp);
1241 int (*o_destroy_export)(struct obd_export *exp);
1242 int (*o_extent_calc)(struct obd_export *, struct lov_stripe_md *,
1243 int cmd, obd_off *);
1245 /* llog related obd_methods */
1246 int (*o_llog_init)(struct obd_device *obd, struct obd_llog_group *grp,
1247 struct obd_device *disk_obd, int *idx);
1248 int (*o_llog_finish)(struct obd_device *obd, int count);
1249 int (*o_llog_connect)(struct obd_export *, struct llogd_conn_body *);
1251 /* metadata-only methods */
1252 int (*o_pin)(struct obd_export *, const struct lu_fid *fid,
1253 struct obd_capa *, struct obd_client_handle *, int flag);
1254 int (*o_unpin)(struct obd_export *, struct obd_client_handle *, int);
1256 int (*o_import_event)(struct obd_device *, struct obd_import *,
1257 enum obd_import_event);
1259 int (*o_notify)(struct obd_device *obd, struct obd_device *watched,
1260 enum obd_notify_event ev, void *data);
1262 int (*o_health_check)(const struct lu_env *env, struct obd_device *);
1263 struct obd_uuid *(*o_get_uuid) (struct obd_export *exp);
1266 int (*o_quotacheck)(struct obd_device *, struct obd_export *,
1267 struct obd_quotactl *);
1268 int (*o_quotactl)(struct obd_device *, struct obd_export *,
1269 struct obd_quotactl *);
1271 int (*o_ping)(const struct lu_env *, struct obd_export *exp);
1274 int (*o_pool_new)(struct obd_device *obd, char *poolname);
1275 int (*o_pool_del)(struct obd_device *obd, char *poolname);
1276 int (*o_pool_add)(struct obd_device *obd, char *poolname,
1278 int (*o_pool_rem)(struct obd_device *obd, char *poolname,
1280 void (*o_getref)(struct obd_device *obd);
1281 void (*o_putref)(struct obd_device *obd);
1283 * NOTE: If adding ops, add another LPROCFS_OBD_OP_INIT() line
1284 * to lprocfs_alloc_obd_stats() in obdclass/lprocfs_status.c.
1285 * Also, add a wrapper function in include/linux/obd_class.h. */
1289 LUSTRE_OPC_MKDIR = (1 << 0),
1290 LUSTRE_OPC_SYMLINK = (1 << 1),
1291 LUSTRE_OPC_MKNOD = (1 << 2),
1292 LUSTRE_OPC_CREATE = (1 << 3),
1293 LUSTRE_OPC_ANY = (1 << 4)
1296 /* lmv structures */
1297 #define MEA_MAGIC_LAST_CHAR 0xb2221ca1
1298 #define MEA_MAGIC_ALL_CHARS 0xb222a11c
1299 #define MEA_MAGIC_HASH_SEGMENT 0xb222a11b
1301 #define MAX_HASH_SIZE_32 0x7fffffffUL
1302 #define MAX_HASH_SIZE 0x7fffffffffffffffULL
1303 #define MAX_HASH_HIGHEST_BIT 0x1000000000000000ULL
1306 struct mdt_body *body;
1307 struct lov_stripe_md *lsm;
1308 struct lmv_stripe_md *mea;
1309 #ifdef CONFIG_FS_POSIX_ACL
1310 struct posix_acl *posix_acl;
1312 struct mdt_remote_perm *remote_perm;
1313 struct obd_capa *mds_capa;
1314 struct obd_capa *oss_capa;
1317 struct md_open_data {
1318 struct obd_client_handle *mod_och;
1319 struct ptlrpc_request *mod_open_req;
1320 struct ptlrpc_request *mod_close_req;
1321 atomic_t mod_refcount;
1324 struct lookup_intent;
1327 int (*m_getstatus)(struct obd_export *, struct lu_fid *,
1328 struct obd_capa **);
1329 int (*m_null_inode)(struct obd_export *, const struct lu_fid *);
1330 int (*m_find_cbdata)(struct obd_export *, const struct lu_fid *,
1331 ldlm_iterator_t, void *);
1332 int (*m_close)(struct obd_export *, struct md_op_data *,
1333 struct md_open_data *, struct ptlrpc_request **);
1334 int (*m_create)(struct obd_export *, struct md_op_data *,
1335 const void *, int, int, __u32, __u32, cfs_cap_t,
1336 __u64, struct ptlrpc_request **);
1337 int (*m_done_writing)(struct obd_export *, struct md_op_data *,
1338 struct md_open_data *);
1339 int (*m_enqueue)(struct obd_export *, struct ldlm_enqueue_info *,
1340 struct lookup_intent *, struct md_op_data *,
1341 struct lustre_handle *, void *, int,
1342 struct ptlrpc_request **, __u64);
1343 int (*m_getattr)(struct obd_export *, struct md_op_data *,
1344 struct ptlrpc_request **);
1345 int (*m_getattr_name)(struct obd_export *, struct md_op_data *,
1346 struct ptlrpc_request **);
1347 int (*m_intent_lock)(struct obd_export *, struct md_op_data *,
1348 void *, int, struct lookup_intent *, int,
1349 struct ptlrpc_request **,
1350 ldlm_blocking_callback, __u64);
1351 int (*m_link)(struct obd_export *, struct md_op_data *,
1352 struct ptlrpc_request **);
1353 int (*m_rename)(struct obd_export *, struct md_op_data *,
1354 const char *, int, const char *, int,
1355 struct ptlrpc_request **);
1356 int (*m_is_subdir)(struct obd_export *, const struct lu_fid *,
1357 const struct lu_fid *,
1358 struct ptlrpc_request **);
1359 int (*m_setattr)(struct obd_export *, struct md_op_data *, void *,
1360 int , void *, int, struct ptlrpc_request **,
1361 struct md_open_data **mod);
1362 int (*m_sync)(struct obd_export *, const struct lu_fid *,
1363 struct obd_capa *, struct ptlrpc_request **);
1364 int (*m_readpage)(struct obd_export *, struct md_op_data *,
1365 struct page **, struct ptlrpc_request **);
1367 int (*m_unlink)(struct obd_export *, struct md_op_data *,
1368 struct ptlrpc_request **);
1370 int (*m_setxattr)(struct obd_export *, const struct lu_fid *,
1371 struct obd_capa *, obd_valid, const char *,
1372 const char *, int, int, int, __u32,
1373 struct ptlrpc_request **);
1375 int (*m_getxattr)(struct obd_export *, const struct lu_fid *,
1376 struct obd_capa *, obd_valid, const char *,
1377 const char *, int, int, int,
1378 struct ptlrpc_request **);
1380 int (*m_init_ea_size)(struct obd_export *, int, int, int);
1382 int (*m_get_lustre_md)(struct obd_export *, struct ptlrpc_request *,
1383 struct obd_export *, struct obd_export *,
1384 struct lustre_md *);
1386 int (*m_free_lustre_md)(struct obd_export *, struct lustre_md *);
1388 int (*m_set_open_replay_data)(struct obd_export *,
1389 struct obd_client_handle *,
1390 struct ptlrpc_request *);
1391 int (*m_clear_open_replay_data)(struct obd_export *,
1392 struct obd_client_handle *);
1393 int (*m_set_lock_data)(struct obd_export *, __u64 *, void *, __u64 *);
1395 ldlm_mode_t (*m_lock_match)(struct obd_export *, __u64,
1396 const struct lu_fid *, ldlm_type_t,
1397 ldlm_policy_data_t *, ldlm_mode_t,
1398 struct lustre_handle *);
1400 int (*m_cancel_unused)(struct obd_export *, const struct lu_fid *,
1401 ldlm_policy_data_t *, ldlm_mode_t,
1402 ldlm_cancel_flags_t flags, void *opaque);
1403 int (*m_renew_capa)(struct obd_export *, struct obd_capa *oc,
1404 renew_capa_cb_t cb);
1405 int (*m_unpack_capa)(struct obd_export *, struct ptlrpc_request *,
1406 const struct req_msg_field *, struct obd_capa **);
1408 int (*m_get_remote_perm)(struct obd_export *, const struct lu_fid *,
1409 struct obd_capa *, __u32,
1410 struct ptlrpc_request **);
1412 int (*m_intent_getattr_async)(struct obd_export *,
1413 struct md_enqueue_info *,
1414 struct ldlm_enqueue_info *);
1416 int (*m_revalidate_lock)(struct obd_export *, struct lookup_intent *,
1417 struct lu_fid *, __u64 *bits);
1420 * NOTE: If adding ops, add another LPROCFS_MD_OP_INIT() line to
1421 * lprocfs_alloc_md_stats() in obdclass/lprocfs_status.c. Also, add a
1422 * wrapper function in include/linux/obd_class.h.
1426 struct lsm_operations {
1427 void (*lsm_free)(struct lov_stripe_md *);
1428 int (*lsm_destroy)(struct lov_stripe_md *, struct obdo *oa,
1429 struct obd_export *md_exp);
1430 void (*lsm_stripe_by_index)(struct lov_stripe_md *, int *, obd_off *,
1432 void (*lsm_stripe_by_offset)(struct lov_stripe_md *, int *, obd_off *,
1434 int (*lsm_lmm_verify) (struct lov_mds_md *lmm, int lmm_bytes,
1435 __u16 *stripe_count);
1436 int (*lsm_unpackmd) (struct lov_obd *lov, struct lov_stripe_md *lsm,
1437 struct lov_mds_md *lmm);
1440 extern const struct lsm_operations lsm_v1_ops;
1441 extern const struct lsm_operations lsm_v3_ops;
1442 static inline const struct lsm_operations *lsm_op_find(int magic)
1450 CERROR("Cannot recognize lsm_magic %08x\n", magic);
1455 /* Requests for obd_extent_calc() */
1456 #define OBD_CALC_STRIPE_START 1
1457 #define OBD_CALC_STRIPE_END 2
1459 static inline struct lustre_capa *oinfo_capa(struct obd_info *oinfo)
1461 return oinfo->oi_capa;
1464 static inline struct md_open_data *obd_mod_alloc(void)
1466 struct md_open_data *mod;
1470 atomic_set(&mod->mod_refcount, 1);
1474 #define obd_mod_get(mod) atomic_inc(&(mod)->mod_refcount)
1475 #define obd_mod_put(mod) \
1477 if (atomic_dec_and_test(&(mod)->mod_refcount)) { \
1478 if ((mod)->mod_open_req) \
1479 ptlrpc_req_finished((mod)->mod_open_req); \
1480 OBD_FREE_PTR(mod); \
1484 void obdo_from_inode(struct obdo *dst, struct inode *src, obd_flag valid);
1485 void obdo_set_parent_fid(struct obdo *dst, const struct lu_fid *parent);
1487 /* return 1 if client should be resend request */
1488 static inline int client_should_resend(int resend, struct client_obd *cli)
1490 return atomic_read(&cli->cl_resends) ?
1491 atomic_read(&cli->cl_resends) > resend : 1;
1495 * Return device name for this device
1497 * XXX: lu_device is declared before obd_device, while a pointer pointing
1498 * back to obd_device in lu_device, so this helper function defines here
1499 * instead of in lu_object.h
1501 static inline const char *lu_dev_name(const struct lu_device *lu_dev)
1503 return lu_dev->ld_obd->obd_name;
1506 static inline bool filename_is_volatile(const char *name, int namelen, int *idx)
1511 if (strncmp(name, LUSTRE_VOLATILE_HDR, LUSTRE_VOLATILE_HDR_LEN) != 0)
1514 /* caller does not care of idx */
1518 /* volatile file, the MDT can be set from name */
1519 /* name format is LUSTRE_VOLATILE_HDR:[idx]: */
1520 /* if no MDT is specified, use std way */
1521 if (namelen < LUSTRE_VOLATILE_HDR_LEN + 2)
1523 /* test for no MDT idx case */
1524 if ((*(name + LUSTRE_VOLATILE_HDR_LEN) == ':') &&
1525 (*(name + LUSTRE_VOLATILE_HDR_LEN + 1) == ':')) {
1529 /* we have an idx, read it */
1530 start = name + LUSTRE_VOLATILE_HDR_LEN + 1;
1531 *idx = strtoul(start, &end, 0);
1533 * no digit, no trailing :, negative value
1535 if (((*idx == 0) && (end == start)) ||
1536 (*end != ':') || (*idx < 0))
1541 /* bad format of mdt idx, we cannot return an error
1542 * to caller so we use hash algo */
1543 CERROR("Bad volatile file name format: %s\n",
1544 name + LUSTRE_VOLATILE_HDR_LEN);
1548 static inline int cli_brw_size(struct obd_device *obd)
1550 LASSERT(obd != NULL);
1551 return obd->u.cli.cl_max_pages_per_rpc << PAGE_CACHE_SHIFT;
1554 #endif /* __OBD_H */