1 /* -*- mode: c; c-basic-offset: 8; indent-tabs-mode: nil; -*-
2 * vim:expandtab:shiftwidth=8:tabstop=8:
6 * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
8 * This program is free software; you can redistribute it and/or modify
9 * it under the terms of the GNU General Public License version 2 only,
10 * as published by the Free Software Foundation.
12 * This program is distributed in the hope that it will be useful, but
13 * WITHOUT ANY WARRANTY; without even the implied warranty of
14 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
15 * General Public License version 2 for more details (a copy is included
16 * in the LICENSE file that accompanied this code).
18 * You should have received a copy of the GNU General Public License
19 * version 2 along with this program; If not, see
20 * http://www.sun.com/software/products/lustre/docs/GPLv2.pdf
22 * Please contact Sun Microsystems, Inc., 4150 Network Circle, Santa Clara,
23 * CA 95054 USA or visit www.sun.com if you need additional information or
29 * Copyright 2008 Sun Microsystems, Inc. All rights reserved
30 * Use is subject to license terms.
33 * This file is part of Lustre, http://www.lustre.org/
34 * Lustre is a trademark of Sun Microsystems, Inc.
40 #if defined(__linux__)
41 #include <linux/obd.h>
42 #elif defined(__APPLE__)
43 #include <darwin/obd.h>
44 #elif defined(__WINNT__)
45 #include <winnt/obd.h>
47 #error Unsupported operating system.
50 #define IOC_OSC_TYPE 'h'
51 #define IOC_OSC_MIN_NR 20
52 #define IOC_OSC_SET_ACTIVE _IOWR(IOC_OSC_TYPE, 21, struct obd_device *)
53 #define IOC_OSC_MAX_NR 50
55 #define IOC_MDC_TYPE 'i'
56 #define IOC_MDC_MIN_NR 20
57 /* Moved to lustre_user.h
58 #define IOC_MDC_LOOKUP _IOWR(IOC_MDC_TYPE, 20, struct obd_ioctl_data *)
59 #define IOC_MDC_GETSTRIPE _IOWR(IOC_MDC_TYPE, 21, struct lov_mds_md *) */
60 #define IOC_MDC_MAX_NR 50
62 #include <lustre/lustre_idl.h>
63 #include <lu_object.h>
65 #include <lustre_lib.h>
66 #include <lustre_export.h>
67 #include <lustre_quota.h>
68 #include <lustre_fld.h>
69 #include <lustre_capa.h>
70 #include <class_hash.h>
72 #include <libcfs/bitmap.h>
75 #define MAX_OBD_DEVICES 8192
77 /* this is really local to the OSC */
78 struct loi_oap_pages {
79 struct list_head lop_pending;
80 struct list_head lop_urgent;
81 struct list_head lop_pending_group;
91 struct lov_oinfo { /* per-stripe data structure */
92 __u64 loi_id; /* object ID on the target OST */
93 __u64 loi_gr; /* object group on the target OST */
94 int loi_ost_idx; /* OST stripe index in lov_tgt_desc->tgts */
95 int loi_ost_gen; /* generation of this loi_ost_idx */
97 /* used by the osc to keep track of what objects to build into rpcs */
98 struct loi_oap_pages loi_read_lop;
99 struct loi_oap_pages loi_write_lop;
100 /* _cli_ is poorly named, it should be _ready_ */
101 struct list_head loi_cli_item;
102 struct list_head loi_write_item;
103 struct list_head loi_read_item;
105 unsigned long loi_kms_valid:1;
106 __u64 loi_kms; /* known minimum size */
107 struct ost_lvb loi_lvb;
108 struct osc_async_rc loi_ar;
111 static inline void loi_kms_set(struct lov_oinfo *oinfo, __u64 kms)
113 oinfo->loi_kms = kms;
114 oinfo->loi_kms_valid = 1;
117 static inline void loi_init(struct lov_oinfo *loi)
119 CFS_INIT_LIST_HEAD(&loi->loi_read_lop.lop_pending);
120 CFS_INIT_LIST_HEAD(&loi->loi_read_lop.lop_urgent);
121 CFS_INIT_LIST_HEAD(&loi->loi_read_lop.lop_pending_group);
122 CFS_INIT_LIST_HEAD(&loi->loi_write_lop.lop_pending);
123 CFS_INIT_LIST_HEAD(&loi->loi_write_lop.lop_urgent);
124 CFS_INIT_LIST_HEAD(&loi->loi_write_lop.lop_pending_group);
125 CFS_INIT_LIST_HEAD(&loi->loi_cli_item);
126 CFS_INIT_LIST_HEAD(&loi->loi_write_item);
127 CFS_INIT_LIST_HEAD(&loi->loi_read_item);
130 /*extent array item for describing the joined file extent info*/
132 __u64 le_start; /* extent start */
133 __u64 le_len; /* extent length */
134 int le_loi_idx; /* extent #1 loi's index in lsm loi array */
135 int le_stripe_count; /* extent stripe count*/
138 /*Lov array info for describing joined file array EA info*/
139 struct lov_array_info {
140 struct llog_logid lai_array_id; /* MDS med llog object id */
141 unsigned lai_ext_count; /* number of extent count */
142 struct lov_extent *lai_ext_array; /* extent desc array */
145 struct lov_stripe_md {
147 pid_t lsm_lock_owner; /* debugging */
150 /* Public members. */
151 __u64 lw_object_id; /* lov object id */
152 __u64 lw_object_gr; /* lov object group */
153 __u64 lw_maxbytes; /* maximum possible file size */
155 /* LOV-private members start here -- only for use in lov/. */
157 __u32 lw_stripe_size; /* size of the stripe */
158 __u32 lw_pattern; /* striping pattern (RAID0, RAID1) */
159 unsigned lw_stripe_count; /* number of objects being striped over */
160 char lw_pool_name[LOV_MAXPOOLNAME]; /* pool name */
163 struct lov_array_info *lsm_array; /*Only for joined file array info*/
164 struct lov_oinfo *lsm_oinfo[0];
167 #define lsm_object_id lsm_wire.lw_object_id
168 #define lsm_object_gr lsm_wire.lw_object_gr
169 #define lsm_maxbytes lsm_wire.lw_maxbytes
170 #define lsm_magic lsm_wire.lw_magic
171 #define lsm_stripe_size lsm_wire.lw_stripe_size
172 #define lsm_pattern lsm_wire.lw_pattern
173 #define lsm_stripe_count lsm_wire.lw_stripe_count
174 #define lsm_pool_name lsm_wire.lw_pool_name
178 typedef int (*obd_enqueue_update_f)(void *cookie, int rc);
180 /* obd info for a particular level (lov, osc). */
182 /* Lock policy. It keeps an extent which is specific for a particular
183 * OSC. (e.g. lov_prep_enqueue_set initialises extent of the policy,
184 * and osc_enqueue passes it into ldlm_lock_match & ldlm_cli_enqueue. */
185 ldlm_policy_data_t oi_policy;
186 /* Flags used for set request specific flags:
187 - while lock handling, the flags obtained on the enqueue
188 request are set here.
189 - while stats, the flags used for control delay/resend.
192 /* Lock handle specific for every OSC lock. */
193 struct lustre_handle *oi_lockh;
194 /* lsm data specific for every OSC. */
195 struct lov_stripe_md *oi_md;
196 /* obdo data specific for every OSC, if needed at all. */
198 /* statfs data specific for every OSC, if needed at all. */
199 struct obd_statfs *oi_osfs;
200 /* An update callback which is called to update some data on upper
201 * level. E.g. it is used for update lsm->lsm_oinfo at every recieved
202 * request in osc level for enqueue requests. It is also possible to
203 * update some caller data from LOV layer if needed. */
204 obd_enqueue_update_f oi_cb_up;
205 /* oss capability, its type is obd_capa in client to avoid copy.
206 * in contrary its type is lustre_capa in OSS. */
210 /* compare all relevant fields. */
211 static inline int lov_stripe_md_cmp(struct lov_stripe_md *m1,
212 struct lov_stripe_md *m2)
215 * ->lsm_wire contains padding, but it should be zeroed out during
218 return memcmp(&m1->lsm_wire, &m2->lsm_wire, sizeof m1->lsm_wire);
221 void lov_stripe_lock(struct lov_stripe_md *md);
222 void lov_stripe_unlock(struct lov_stripe_md *md);
225 struct list_head typ_chain;
226 struct obd_ops *typ_dt_ops;
227 struct md_ops *typ_md_ops;
228 cfs_proc_dir_entry_t *typ_procroot;
231 struct lu_device_type *typ_lu;
232 spinlock_t obd_type_lock;
242 /* Individual type definitions */
244 struct ost_server_data;
246 /* hold common fields for "target" device */
247 struct obd_device_target {
248 struct super_block *obt_sb;
249 atomic_t obt_quotachecking;
250 struct lustre_quota_ctxt obt_qctxt;
251 lustre_quota_version_t obt_qfmt;
252 struct rw_semaphore obt_rwsem;
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 context */
273 #define FILTER_SUBDIR_COUNT 32 /* set to zero for no subdirs */
275 struct filter_subdirs {
276 cfs_dentry_t *dentry[FILTER_SUBDIR_COUNT];
286 /* NB this field MUST be first */
287 struct obd_device_target fo_obt;
288 const char *fo_fstype;
289 struct vfsmount *fo_vfsmnt;
292 cfs_dentry_t *fo_dentry_O;
293 cfs_dentry_t **fo_dentry_O_groups;
294 struct filter_subdirs *fo_dentry_O_sub;
295 struct semaphore fo_init_lock; /* group initialization lock */
296 int fo_committed_group;
299 spinlock_t fo_objidlock; /* protect fo_lastobjid */
300 spinlock_t fo_translock; /* protect fsd_last_transno */
301 struct file *fo_rcvd_filp;
302 struct file *fo_health_check_filp;
303 struct lr_server_data *fo_fsd;
304 unsigned long *fo_last_rcvd_slots;
305 __u64 fo_mount_count;
307 unsigned long fo_destroys_in_progress;
308 struct semaphore fo_create_locks[FILTER_SUBDIR_COUNT];
310 struct list_head fo_export_list;
313 obd_size fo_tot_dirty; /* protected by obd_osfs_lock */
314 obd_size fo_tot_granted; /* all values in bytes */
315 obd_size fo_tot_pending;
317 obd_size fo_readcache_max_filesize;
319 int fo_writethrough_cache;
321 struct obd_import *fo_mdc_imp;
322 struct obd_uuid fo_mdc_uuid;
323 struct lustre_handle fo_mdc_conn;
324 struct file **fo_last_objid_files;
325 __u64 *fo_last_objids; /* last created objid for groups,
326 * protected by fo_objidlock */
328 struct semaphore fo_alloc_lock;
330 atomic_t fo_r_in_flight;
331 atomic_t fo_w_in_flight;
334 * per-filter pool of kiobuf's allocated by filter_common_setup() and
335 * torn down by filter_cleanup(). Contains OST_NUM_THREADS elements of
336 * which ->fo_iobuf_count were allocated.
338 * This pool contains kiobuf used by
339 * filter_{prep,commit}rw_{read,write}() and is shared by all OST
342 * Locking: none, each OST thread uses only one element, determined by
343 * its "ordinal number", ->t_id.
345 struct filter_iobuf **fo_iobuf_pool;
348 struct list_head fo_llog_list;
349 spinlock_t fo_llog_list_lock;
351 struct brw_stats fo_filter_stats;
352 struct lustre_quota_ctxt fo_quota_ctxt;
353 spinlock_t fo_quotacheck_lock;
354 atomic_t fo_quotachecking;
356 int fo_fmd_max_num; /* per exp filter_mod_data */
357 int fo_fmd_max_age; /* jiffies to fmd expiry */
360 rwlock_t fo_sptlrpc_lock;
361 struct sptlrpc_rule_set fo_sptlrpc_rset;
363 /* capability related */
364 unsigned int fo_fl_oss_capa;
365 struct list_head fo_capa_keys;
366 struct hlist_head *fo_capa_hash;
367 struct llog_commit_master *fo_lcm;
371 #define OSC_MAX_RIF_DEFAULT 8
372 #define OSC_MAX_RIF_MAX 256
373 #define OSC_MAX_DIRTY_DEFAULT (OSC_MAX_RIF_DEFAULT * 4)
374 #define OSC_MAX_DIRTY_MB_MAX 2048 /* arbitrary, but < MAX_LONG bytes */
375 #define OSC_DEFAULT_RESENDS 10
377 #define MDC_MAX_RIF_DEFAULT 8
378 #define MDC_MAX_RIF_MAX 512
383 struct rw_semaphore cl_sem;
384 struct obd_uuid cl_target_uuid;
385 struct obd_import *cl_import; /* ptlrpc connection state */
387 /* max_mds_easize is purely a performance thing so we don't have to
388 * call obd_size_diskmd() all the time. */
389 int cl_default_mds_easize;
390 int cl_max_mds_easize;
391 int cl_max_mds_cookiesize;
393 enum lustre_sec_part cl_sp_me;
394 enum lustre_sec_part cl_sp_to;
395 struct sptlrpc_flavor cl_flvr_mgc; /* fixed flavor of mgc->mgs */
397 //struct llog_canceld_ctxt *cl_llcd; /* it's included by obd_llog_ctxt */
398 void *cl_llcd_offset;
400 /* the grant values are protected by loi_list_lock below */
401 long cl_dirty; /* all _dirty_ in bytes */
402 long cl_dirty_max; /* allowed w/o rpc */
403 long cl_dirty_transit; /* dirty synchronous */
404 long cl_avail_grant; /* bytes of credit for ost */
405 long cl_lost_grant; /* lost credits (trunc) */
406 struct list_head cl_cache_waiters; /* waiting for cache/grant */
408 /* keep track of objects that have lois that contain pages which
409 * have been queued for async brw. this lock also protects the
410 * lists of osc_client_pages that hang off of the loi */
412 * ->cl_loi_list_lock protects consistency of
413 * ->cl_loi_{ready,read,write}_list. ->ap_make_ready() and
414 * ->ap_completion() call-backs are executed under this lock. As we
415 * cannot guarantee that these call-backs never block on all platforms
416 * (as a matter of fact they do block on Mac OS X), type of
417 * ->cl_loi_list_lock is platform dependent: it's a spin-lock on Linux
418 * and blocking mutex on Mac OS X. (Alternative is to make this lock
419 * blocking everywhere, but we don't want to slow down fast-path of
420 * our main platform.)
422 * Exact type of ->cl_loi_list_lock is defined in arch/obd.h together
423 * with client_obd_list_{un,}lock() and
424 * client_obd_list_lock_{init,done}() functions.
426 client_obd_lock_t cl_loi_list_lock;
427 struct list_head cl_loi_ready_list;
428 struct list_head cl_loi_write_list;
429 struct list_head cl_loi_read_list;
432 /* just a sum of the loi/lop pending numbers to be exported by /proc */
433 int cl_pending_w_pages;
434 int cl_pending_r_pages;
435 int cl_max_pages_per_rpc;
436 int cl_max_rpcs_in_flight;
437 struct obd_histogram cl_read_rpc_hist;
438 struct obd_histogram cl_write_rpc_hist;
439 struct obd_histogram cl_read_page_hist;
440 struct obd_histogram cl_write_page_hist;
441 struct obd_histogram cl_read_offset_hist;
442 struct obd_histogram cl_write_offset_hist;
444 /* number of in flight destroy rpcs is limited to max_rpcs_in_flight */
445 atomic_t cl_destroy_in_flight;
446 cfs_waitq_t cl_destroy_waitq;
448 struct mdc_rpc_lock *cl_rpc_lock;
449 struct mdc_rpc_lock *cl_setattr_lock;
450 struct mdc_rpc_lock *cl_close_lock;
451 struct osc_creator cl_oscc;
454 struct semaphore cl_mgc_sem;
455 struct vfsmount *cl_mgc_vfsmnt;
456 struct dentry *cl_mgc_configs_dir;
457 atomic_t cl_mgc_refcount;
458 struct obd_export *cl_mgc_mgsexp;
460 /* checksumming for data sent over the network */
461 unsigned int cl_checksum:1; /* 0 = disabled, 1 = enabled */
462 /* supported checksum types that are worked out at connect time */
463 __u32 cl_supp_cksum_types;
464 /* checksum algorithm to be used */
465 cksum_type_t cl_cksum_type;
467 /* also protected by the poorly named _loi_list_lock lock above */
468 struct osc_async_rc cl_ar;
470 /* used by quotacheck */
471 int cl_qchk_stat; /* quotacheck stat of the peer */
473 /* sequence manager */
474 struct lu_client_seq *cl_seq;
476 atomic_t cl_resends; /* resend count */
478 #define obd2cli_tgt(obd) ((char *)(obd)->u.cli.cl_target_uuid.uuid)
480 #define CL_NOT_QUOTACHECKED 1 /* client->cl_qchk_stat init value */
483 struct ptlrpc_service *mgs_service;
484 struct vfsmount *mgs_vfsmnt;
485 struct super_block *mgs_sb;
486 struct dentry *mgs_configs_dir;
487 struct dentry *mgs_fid_de;
488 struct list_head mgs_fs_db_list;
489 struct semaphore mgs_sem;
490 cfs_proc_dir_entry_t *mgs_proc_live;
494 /* NB this field MUST be first */
495 struct obd_device_target mds_obt;
496 struct ptlrpc_service *mds_service;
497 struct ptlrpc_service *mds_setattr_service;
498 struct ptlrpc_service *mds_readpage_service;
499 struct vfsmount *mds_vfsmnt;
500 cfs_dentry_t *mds_fid_de;
502 int mds_max_cookiesize;
503 struct file *mds_rcvd_filp;
504 spinlock_t mds_transno_lock;
505 __u64 mds_last_transno;
506 __u64 mds_mount_count;
508 unsigned long mds_atime_diff;
509 struct semaphore mds_epoch_sem;
510 struct ll_fid mds_rootfid;
511 struct lr_server_data *mds_server_data;
512 cfs_dentry_t *mds_pending_dir;
513 cfs_dentry_t *mds_logs_dir;
514 cfs_dentry_t *mds_objects_dir;
515 struct llog_handle *mds_cfg_llh;
516 struct obd_device *mds_osc_obd; /* XXX lov_obd */
517 struct obd_uuid mds_lov_uuid;
519 struct obd_export *mds_osc_exp; /* XXX lov_exp */
520 struct lov_desc mds_lov_desc;
523 /* mark pages dirty for write. */
524 bitmap_t *mds_lov_page_dirty;
525 /* array for store pages with obd_id */
526 void **mds_lov_page_array;
527 /* file for store objid */
528 struct file *mds_lov_objid_filp;
529 __u32 mds_lov_objid_count;
530 __u32 mds_lov_objid_max_index;
531 __u32 mds_lov_objid_lastpage;
532 __u32 mds_lov_objid_lastidx;
534 struct file *mds_health_check_filp;
535 unsigned long *mds_client_bitmap;
536 // struct upcall_cache *mds_group_hash;
538 struct lustre_quota_info mds_quota_info;
539 struct semaphore mds_qonoff_sem;
540 struct semaphore mds_health_sem;
541 unsigned long mds_fl_user_xattr:1,
543 mds_evict_ost_nids:1,
547 mds_fl_target:1; /* mds have one or
550 struct upcall_cache *mds_identity_cache;
552 /* for capability keys update */
553 struct lustre_capa_key *mds_capa_keys;
554 struct rw_semaphore mds_notify_lock;
558 extern __u32 mds_max_ost_index;
560 #define MDS_LOV_ALLOC_SIZE (CFS_PAGE_SIZE)
562 #define OBJID_PER_PAGE() (MDS_LOV_ALLOC_SIZE / sizeof(obd_id))
564 #define MDS_LOV_OBJID_PAGES_COUNT (mds_max_ost_index/OBJID_PER_PAGE())
566 extern int mds_lov_init_objids(struct obd_device *obd);
567 extern void mds_lov_destroy_objids(struct obd_device *obd);
580 struct lustre_handle eo_nl_lock;
585 struct ptlrpc_service *ost_service;
586 struct ptlrpc_service *ost_create_service;
587 struct ptlrpc_service *ost_io_service;
588 struct semaphore ost_health_sem;
591 struct echo_client_obd {
592 struct obd_export *ec_exp; /* the local connection to osc/lov */
594 struct list_head ec_objects;
595 struct list_head ec_locks;
601 struct obd_uuid lqo_uuid; /* ptlrpc's c_remote_uuid */
602 struct list_head lqo_oss_list; /* link to lov_qos */
603 __u64 lqo_bavail; /* total bytes avail on OSS */
604 __u64 lqo_penalty; /* current penalty */
605 __u64 lqo_penalty_per_obj; /* penalty decrease every obj*/
606 time_t lqo_used; /* last used time, seconds */
607 __u32 lqo_ost_count; /* number of osts on this oss */
611 struct lov_qos_oss *ltq_oss; /* oss info */
612 __u64 ltq_penalty; /* current penalty */
613 __u64 ltq_penalty_per_obj; /* penalty decrease every obj*/
614 __u64 ltq_weight; /* net weighting */
615 time_t ltq_used; /* last used time, seconds */
616 unsigned int ltq_usable:1; /* usable for striping */
619 /* Generic subset of OSTs */
621 __u32 *op_array; /* array of index of
623 unsigned int op_count; /* number of OSTs in the array */
624 unsigned int op_size; /* allocated size of lp_array */
625 struct rw_semaphore op_rw_sem; /* to protect ost_pool use */
628 /* Round-robin allocator data */
630 __u32 lqr_start_idx; /* start index of new inode */
631 __u32 lqr_offset_idx; /* aliasing for start_idx */
632 int lqr_start_count; /* reseed counter */
633 struct ost_pool lqr_pool; /* round-robin optimized list */
634 unsigned long lqr_dirty:1; /* recalc round-robin list */
637 /* Stripe placement optimization */
639 struct list_head lq_oss_list; /* list of OSSs that targets use */
640 struct rw_semaphore lq_rw_sem;
641 __u32 lq_active_oss_count;
642 unsigned int lq_prio_free; /* priority for free space */
643 unsigned int lq_threshold_rr;/* priority for rr */
644 struct lov_qos_rr lq_rr; /* round robin qos data */
645 unsigned long lq_dirty:1, /* recalc qos data */
646 lq_same_space:1,/* the ost's all have approx.
647 the same space avail */
648 lq_reset:1; /* zero current penalties */
651 struct lov_tgt_desc {
652 struct list_head ltd_kill;
653 struct obd_uuid ltd_uuid;
654 struct obd_export *ltd_exp;
655 struct ltd_qos ltd_qos; /* qos info per target */
657 __u32 ltd_index; /* index in lov_obd->tgts */
658 unsigned long ltd_active:1,/* is this target up for requests */
659 ltd_activate:1,/* should this target be activated */
660 ltd_reap:1; /* should this target be deleted */
664 #define pool_tgt_size(_p) _p->pool_obds.op_size
665 #define pool_tgt_count(_p) _p->pool_obds.op_count
666 #define pool_tgt_array(_p) _p->pool_obds.op_array
667 #define pool_tgt_rw_sem(_p) _p->pool_obds.op_rw_sem
668 #define pool_tgt(_p, _i) _p->pool_lov->lov_tgts[_p->pool_obds.op_array[_i]]
671 char pool_name[LOV_MAXPOOLNAME + 1]; /* name of pool */
672 struct ost_pool pool_obds; /* pool members */
673 atomic_t pool_refcount; /* pool ref. counter */
674 struct lov_qos_rr pool_rr; /* round robin qos */
675 struct hlist_node pool_hash; /* access by poolname */
676 struct list_head pool_list; /* serial access */
677 cfs_proc_dir_entry_t *pool_proc_entry; /* file in /proc */
678 struct lov_obd *pool_lov; /* lov obd to which this
683 struct lov_desc desc;
684 struct lov_tgt_desc **lov_tgts; /* sparse array */
685 struct ost_pool lov_packed; /* all OSTs in a packed
687 struct semaphore lov_lock;
688 struct obd_connect_data lov_ocd;
689 struct lov_qos lov_qos; /* qos info per lov */
690 atomic_t lov_refcount;
691 __u32 lov_tgt_count; /* how many OBD's */
692 __u32 lov_active_tgt_count; /* how many active */
693 __u32 lov_death_row;/* tgts scheduled to be deleted */
694 __u32 lov_tgt_size; /* size of tgts array */
697 lustre_hash_t *lov_pools_hash_body; /* used for key access */
698 struct list_head lov_pool_list; /* used for sequential access */
699 cfs_proc_dir_entry_t *lov_pool_proc_entry;
700 enum lustre_sec_part lov_sp_me;
703 struct lmv_tgt_desc {
704 struct obd_uuid ltd_uuid;
705 struct obd_export *ltd_exp;
706 int ltd_active; /* is this target up for requests */
708 struct semaphore ltd_fid_sem;
711 enum placement_policy {
712 PLACEMENT_CHAR_POLICY = 0,
713 PLACEMENT_NID_POLICY = 1,
714 PLACEMENT_INVAL_POLICY = 2,
718 typedef enum placement_policy placement_policy_t;
722 struct lu_client_fld lmv_fld;
724 placement_policy_t lmv_placement;
725 struct lmv_desc desc;
726 struct obd_uuid cluuid;
727 struct obd_export *exp;
734 struct semaphore init_sem;
736 struct lmv_tgt_desc *tgts;
739 struct obd_connect_data *datas;
742 struct obd_connect_data conn_data;
745 struct niobuf_local {
750 cfs_dentry_t *dentry;
755 #define LUSTRE_FLD_NAME "fld"
756 #define LUSTRE_SEQ_NAME "seq"
758 #define LUSTRE_CMM_NAME "cmm"
759 #define LUSTRE_MDD_NAME "mdd"
760 #define LUSTRE_OSD_NAME "osd"
761 #define LUSTRE_VVP_NAME "vvp"
762 #define LUSTRE_LMV_NAME "lmv"
763 #define LUSTRE_CMM_MDC_NAME "cmm-mdc"
764 #define LUSTRE_SLP_NAME "slp"
766 /* obd device type names */
767 /* FIXME all the references to LUSTRE_MDS_NAME should be swapped with LUSTRE_MDT_NAME */
768 #define LUSTRE_MDS_NAME "mds"
769 #define LUSTRE_MDT_NAME "mdt"
770 #define LUSTRE_MDC_NAME "mdc"
771 #define LUSTRE_OSS_NAME "ost" /* FIXME change name to oss */
772 #define LUSTRE_OST_NAME "obdfilter" /* FIXME change name to ost */
773 #define LUSTRE_OSC_NAME "osc"
774 #define LUSTRE_LOV_NAME "lov"
775 #define LUSTRE_MGS_NAME "mgs"
776 #define LUSTRE_MGC_NAME "mgc"
778 #define LUSTRE_CACHEOBD_NAME "cobd"
779 #define LUSTRE_ECHO_NAME "obdecho"
780 #define LUSTRE_ECHO_CLIENT_NAME "echo_client"
782 /* Constant obd names (post-rename) */
783 #define LUSTRE_MDS_OBDNAME "MDS"
784 #define LUSTRE_OSS_OBDNAME "OSS"
785 #define LUSTRE_MGS_OBDNAME "MGS"
786 #define LUSTRE_MGC_OBDNAME "MGC"
788 /* Don't conflict with on-wire flags OBD_BRW_WRITE, etc */
789 #define N_LOCAL_TEMP_PAGE 0x10000000
791 struct obd_trans_info {
794 /* Only used on the server side for tracking acks. */
795 struct oti_req_ack_lock {
796 struct lustre_handle lock;
800 struct llog_cookie oti_onecookie;
801 struct llog_cookie *oti_logcookies;
804 /* initial thread handling transaction */
805 struct ptlrpc_thread * oti_thread;
808 struct obd_uuid *oti_ost_uuid;
811 static inline void oti_init(struct obd_trans_info *oti,
812 struct ptlrpc_request *req)
816 memset(oti, 0, sizeof(*oti));
821 oti->oti_xid = req->rq_xid;
823 if (req->rq_repmsg != NULL)
824 oti->oti_transno = lustre_msg_get_transno(req->rq_repmsg);
825 oti->oti_thread = req->rq_svc_thread;
826 if (req->rq_reqmsg != NULL)
827 oti->oti_conn_cnt = lustre_msg_get_conn_cnt(req->rq_reqmsg);
830 static inline void oti_alloc_cookies(struct obd_trans_info *oti,int num_cookies)
835 if (num_cookies == 1)
836 oti->oti_logcookies = &oti->oti_onecookie;
838 OBD_ALLOC(oti->oti_logcookies,
839 num_cookies * sizeof(oti->oti_onecookie));
841 oti->oti_numcookies = num_cookies;
844 static inline void oti_free_cookies(struct obd_trans_info *oti)
846 if (!oti || !oti->oti_logcookies)
849 if (oti->oti_logcookies == &oti->oti_onecookie)
850 LASSERT(oti->oti_numcookies == 1);
852 OBD_FREE(oti->oti_logcookies,
853 oti->oti_numcookies * sizeof(oti->oti_onecookie));
854 oti->oti_logcookies = NULL;
855 oti->oti_numcookies = 0;
859 * Events signalled through obd_notify() upcall-chain.
861 enum obd_notify_event {
862 /* Device connect start */
864 /* Device activated */
866 /* Device deactivated */
868 /* Device disconnected */
870 /* Connect data for import were changed */
873 OBD_NOTIFY_SYNC_NONBLOCK,
875 /* Configuration event */
879 /* bit-mask flags for config events */
881 CONFIG_LOG = 0x1, /* finished processing config log */
882 CONFIG_SYNC = 0x2, /* mdt synced 1 ost */
883 CONFIG_TARGET = 0x4 /* one target is added */
887 * Data structure used to pass obd_notify()-event to non-obd listeners (llite
888 * and liblustre being main examples).
890 struct obd_notify_upcall {
891 int (*onu_upcall)(struct obd_device *host, struct obd_device *watched,
892 enum obd_notify_event ev, void *owner);
893 /* Opaque datum supplied by upper layer listener */
897 struct target_recovery_data {
898 svc_handler_t trd_recovery_handler;
899 pid_t trd_processing_task;
900 struct completion trd_starting;
901 struct completion trd_finishing;
905 FILTER_GROUP_MDS0 = 0,
906 FILTER_GROUP_LLOG = 1,
907 FILTER_GROUP_ECHO = 2 ,
908 FILTER_GROUP_MDS1_N_BASE = 3
911 static inline __u64 obdo_mdsno(struct obdo *oa)
914 return oa->o_gr - FILTER_GROUP_MDS1_N_BASE;
918 static inline int mdt_to_obd_objgrp(int mdtid)
921 return FILTER_GROUP_MDS1_N_BASE + mdtid;
926 * In HEAD for CMD, the object is created in group number which is 3>=
927 * or indexing starts from 3. To test this assertions are added to disallow
928 * group 0. But to run 2.0 mds server on 1.8.x disk format (i.e. interop_mode)
929 * object in group 0 needs to be allowed.
930 * So for interop mode following changes needs to be done:
931 * 1. No need to assert on group 0 or allow group 0
932 * 2. The group number indexing starts from 0 instead of 3
935 #define CHECK_MDS_GROUP(group) (group == FILTER_GROUP_MDS0 || \
936 group > FILTER_GROUP_MDS1_N_BASE)
937 #define LASSERT_MDS_GROUP(group) LASSERT(CHECK_MDS_GROUP(group))
939 struct obd_llog_group {
940 struct list_head olg_list;
942 struct llog_ctxt *olg_ctxts[LLOG_MAX_CTXTS];
943 cfs_waitq_t olg_waitq;
945 struct obd_export *olg_exp;
946 int olg_initializing;
947 struct semaphore olg_cat_processing;
950 /* corresponds to one of the obd's */
951 #define MAX_OBD_NAME 128
952 #define OBD_DEVICE_MAGIC 0XAB5CD6EF
953 #define OBD_DEV_BY_DEVNAME 0xffffd0de
955 struct obd_type *obd_type;
958 /* common and UUID name of this device */
959 char obd_name[MAX_OBD_NAME];
960 struct obd_uuid obd_uuid;
962 struct lu_device *obd_lu_dev;
965 unsigned long obd_attached:1, /* finished attach */
966 obd_set_up:1, /* finished setup */
967 obd_recovering:1, /* there are recoverable clients */
968 obd_abort_recovery:1,/* somebody ioctl'ed us to abort */
969 obd_replayable:1, /* recovery is enabled; inform clients */
970 obd_no_transno:1, /* no committed-transno notification */
971 obd_no_recov:1, /* fail instead of retry messages */
972 obd_req_replaying:1, /* replaying requests */
973 obd_stopping:1, /* started cleanup */
974 obd_starting:1, /* started setup */
975 obd_force:1, /* cleanup with > 0 obd refcount */
976 obd_fail:1, /* cleanup with failover */
977 obd_async_recov:1, /* allow asyncronous orphan cleanup */
978 obd_no_conn:1, /* deny new connections */
979 obd_inactive:1; /* device active/inactive
980 * (for /proc/status only!!) */
981 /* uuid-export hash body */
982 struct lustre_hash *obd_uuid_hash;
983 /* nid-export hash body */
984 struct lustre_hash *obd_nid_hash;
986 struct lustre_hash *obd_nid_stats_hash;
987 struct list_head obd_nid_stats;
988 atomic_t obd_refcount;
989 cfs_waitq_t obd_refcount_waitq;
990 struct list_head obd_exports;
992 spinlock_t obd_nid_lock;
993 struct ldlm_namespace *obd_namespace;
994 struct ptlrpc_client obd_ldlm_client; /* XXX OST/MDS only */
995 /* a spinlock is OK for what we do now, may need a semaphore later */
996 spinlock_t obd_dev_lock;
997 struct semaphore obd_dev_sem;
998 __u64 obd_last_committed;
999 struct fsfilt_operations *obd_fsops;
1000 spinlock_t obd_osfs_lock;
1001 struct obd_statfs obd_osfs; /* locked by obd_osfs_lock */
1003 struct lvfs_run_ctxt obd_lvfs_ctxt;
1004 struct obd_llog_group obd_olg; /* default llog group */
1005 struct obd_device *obd_observer;
1006 struct obd_notify_upcall obd_upcall;
1007 struct obd_export *obd_self_export;
1008 /* list of exports in LRU order, for ping evictor, with obd_dev_lock */
1009 struct list_head obd_exports_timed;
1010 time_t obd_eviction_timer; /* for ping evictor */
1012 /* XXX encapsulate all this recovery data into one struct */
1013 svc_handler_t obd_recovery_handler;
1014 pid_t obd_processing_task;
1016 int obd_max_recoverable_clients;
1017 int obd_connected_clients;
1018 int obd_recoverable_clients;
1019 spinlock_t obd_processing_task_lock; /* BH lock (timer) */
1020 __u64 obd_next_recovery_transno;
1021 int obd_replayed_requests;
1022 int obd_requests_queued_for_recovery;
1023 cfs_waitq_t obd_next_transno_waitq;
1024 struct list_head obd_uncommitted_replies;
1025 spinlock_t obd_uncommitted_replies_lock;
1026 cfs_timer_t obd_recovery_timer;
1027 time_t obd_recovery_start; /* seconds */
1028 time_t obd_recovery_end; /* seconds, for lprocfs_status */
1029 time_t obd_recovery_max_time; /* seconds, bz13079 */
1030 int obd_recovery_timeout;
1032 /* new recovery stuff from CMD2 */
1033 struct target_recovery_data obd_recovery_data;
1034 int obd_replayed_locks;
1035 atomic_t obd_req_replay_clients;
1036 atomic_t obd_lock_replay_clients;
1037 struct list_head obd_req_replay_queue;
1038 struct list_head obd_lock_replay_queue;
1039 struct list_head obd_final_req_queue;
1040 int obd_recovery_stage;
1043 struct obd_device_target obt;
1044 struct filter_obd filter;
1046 struct client_obd cli;
1048 struct echo_client_obd echo_client;
1049 struct echo_obd echo;
1054 /* Fields used by LProcFS */
1055 unsigned int obd_cntr_base;
1056 struct lprocfs_stats *obd_stats;
1058 unsigned int md_cntr_base;
1059 struct lprocfs_stats *md_stats;
1061 cfs_proc_dir_entry_t *obd_proc_entry;
1062 cfs_proc_dir_entry_t *obd_proc_exports_entry;
1063 cfs_proc_dir_entry_t *obd_svc_procroot;
1064 struct lprocfs_stats *obd_svc_stats;
1065 atomic_t obd_evict_inprogress;
1066 cfs_waitq_t obd_evict_inprogress_waitq;
1069 * Ldlm pool part. Save last calculated SLV and Limit.
1071 rwlock_t obd_pool_lock;
1076 * A list of outstanding class_incref()'s against this obd. For
1079 struct lu_ref obd_reference;
1082 #define OBD_LLOG_FL_SENDNOW 0x0001
1084 enum obd_cleanup_stage {
1085 /* Special case hack for MDS LOVs */
1087 /* can be directly mapped to .ldto_device_fini() */
1088 OBD_CLEANUP_EXPORTS,
1091 /* get/set_info keys */
1092 #define KEY_READ_ONLY "read-only"
1093 #define KEY_MDS_CONN "mds_conn"
1094 #define KEY_NEXT_ID "next_id"
1095 #define KEY_LOVDESC "lovdesc"
1096 #define KEY_INIT_RECOV "initial_recov"
1097 #define KEY_INIT_RECOV_BACKUP "init_recov_bk"
1098 #define KEY_FLUSH_CTX "flush_ctx"
1099 #define KEY_CAPA_KEY "capa_key"
1100 #define KEY_CONN_DATA "conn_data"
1101 #define KEY_MAX_EASIZE "max_easize"
1102 #define KEY_REVIMP_UPD "revimp_update"
1103 #define KEY_LOV_IDX "lov_idx"
1104 #define KEY_LAST_ID "last_id"
1105 #define KEY_READONLY "read-only"
1106 #define KEY_LOCK_TO_STRIPE "lock_to_stripe"
1107 #define KEY_CHECKSUM "checksum"
1108 #define KEY_UNLINKED "unlinked"
1109 #define KEY_EVICT_BY_NID "evict_by_nid"
1110 #define KEY_REGISTER_TARGET "register_target"
1111 #define KEY_SET_FS "set_fs"
1112 #define KEY_CLEAR_FS "clear_fs"
1113 #define KEY_BLOCKSIZE "blocksize"
1114 #define KEY_BLOCKSIZE_BITS "blocksize_bits"
1115 #define KEY_FIEMAP "fiemap"
1116 #define KEY_SPTLRPC_CONF "sptlrpc_conf"
1117 #define KEY_MGSSEC "mgssec"
1119 #define KEY_INTERMDS "inter_mds"
1120 #define KEY_ASYNC "async"
1124 static inline int it_to_lock_mode(struct lookup_intent *it)
1126 /* CREAT needs to be tested before open (both could be set) */
1127 if (it->it_op & IT_CREAT)
1129 else if (it->it_op & (IT_READDIR | IT_GETATTR | IT_OPEN | IT_LOOKUP))
1132 LASSERTF(0, "Invalid it_op: %d\n", it->it_op);
1137 struct lu_fid op_fid1; /* operation fid1 (usualy parent) */
1138 struct lu_fid op_fid2; /* operation fid2 (usualy child) */
1139 struct lu_fid op_fid3; /* 2 extra fids to find conflicting */
1140 struct lu_fid op_fid4; /* to the operation locks. */
1141 mdsno_t op_mds; /* what mds server open will go to */
1142 struct lustre_handle op_handle;
1144 const char *op_name;
1147 struct lmv_stripe_md *op_mea1;
1148 struct lmv_stripe_md *op_mea2;
1149 __u32 op_suppgids[2];
1155 /* iattr fields and blocks. */
1156 struct iattr op_attr;
1158 #if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,14)
1159 unsigned int op_attr_flags;
1162 loff_t op_attr_blocks;
1164 /* Size-on-MDS epoch and flags. */
1169 struct obd_capa *op_capa1;
1170 struct obd_capa *op_capa2;
1172 /* Various operation flags. */
1175 /* Operation type */
1179 struct md_enqueue_info;
1180 /* metadata stat-ahead */
1181 typedef int (* md_enqueue_cb_t)(struct ptlrpc_request *req,
1182 struct md_enqueue_info *minfo,
1185 struct md_enqueue_info {
1186 struct md_op_data mi_data;
1187 struct lookup_intent mi_it;
1188 struct lustre_handle mi_lockh;
1189 struct dentry *mi_dentry;
1190 md_enqueue_cb_t mi_cb;
1191 unsigned int mi_generation;
1196 struct module *o_owner;
1197 int (*o_iocontrol)(unsigned int cmd, struct obd_export *exp, int len,
1198 void *karg, void *uarg);
1199 int (*o_get_info)(struct obd_export *, __u32 keylen, void *key,
1200 __u32 *vallen, void *val, struct lov_stripe_md *lsm);
1201 int (*o_set_info_async)(struct obd_export *, __u32 keylen, void *key,
1202 __u32 vallen, void *val,
1203 struct ptlrpc_request_set *set);
1204 int (*o_attach)(struct obd_device *dev, obd_count len, void *data);
1205 int (*o_detach)(struct obd_device *dev);
1206 int (*o_setup) (struct obd_device *dev, struct lustre_cfg *cfg);
1207 int (*o_precleanup)(struct obd_device *dev,
1208 enum obd_cleanup_stage cleanup_stage);
1209 int (*o_cleanup)(struct obd_device *dev);
1210 int (*o_process_config)(struct obd_device *dev, obd_count len,
1212 int (*o_postrecov)(struct obd_device *dev);
1213 int (*o_add_conn)(struct obd_import *imp, struct obd_uuid *uuid,
1215 int (*o_del_conn)(struct obd_import *imp, struct obd_uuid *uuid);
1216 /* connect to the target device with given connection
1217 * data. @ocd->ocd_connect_flags is modified to reflect flags actually
1218 * granted by the target, which are guaranteed to be a subset of flags
1219 * asked for. If @ocd == NULL, use default parameters. */
1220 int (*o_connect)(const struct lu_env *env,
1221 struct obd_export **exp, struct obd_device *src,
1222 struct obd_uuid *cluuid, struct obd_connect_data *ocd,
1224 int (*o_reconnect)(const struct lu_env *env,
1225 struct obd_export *exp, struct obd_device *src,
1226 struct obd_uuid *cluuid,
1227 struct obd_connect_data *ocd,
1229 int (*o_disconnect)(struct obd_export *exp);
1231 /* Initialize/finalize fids infrastructure. */
1232 int (*o_fid_init)(struct obd_export *exp);
1233 int (*o_fid_fini)(struct obd_export *exp);
1235 /* Allocate new fid according to passed @hint. */
1236 int (*o_fid_alloc)(struct obd_export *exp, struct lu_fid *fid,
1237 struct md_op_data *op_data);
1240 * Object with @fid is getting deleted, we may want to do something
1243 int (*o_fid_delete)(struct obd_export *exp, const struct lu_fid *fid);
1245 int (*o_statfs)(struct obd_device *obd, struct obd_statfs *osfs,
1246 __u64 max_age, __u32 flags);
1247 int (*o_statfs_async)(struct obd_device *obd, struct obd_info *oinfo,
1248 __u64 max_age, struct ptlrpc_request_set *set);
1249 int (*o_packmd)(struct obd_export *exp, struct lov_mds_md **disk_tgt,
1250 struct lov_stripe_md *mem_src);
1251 int (*o_unpackmd)(struct obd_export *exp,struct lov_stripe_md **mem_tgt,
1252 struct lov_mds_md *disk_src, int disk_len);
1253 int (*o_checkmd)(struct obd_export *exp, struct obd_export *md_exp,
1254 struct lov_stripe_md *mem_tgt);
1255 int (*o_preallocate)(struct lustre_handle *, obd_count *req,
1257 /* FIXME: add fid capability support for create & destroy! */
1258 int (*o_precreate)(struct obd_export *exp);
1259 int (*o_create)(struct obd_export *exp, struct obdo *oa,
1260 struct lov_stripe_md **ea, struct obd_trans_info *oti);
1261 int (*o_destroy)(struct obd_export *exp, struct obdo *oa,
1262 struct lov_stripe_md *ea, struct obd_trans_info *oti,
1263 struct obd_export *md_exp, void *capa);
1264 int (*o_setattr)(struct obd_export *exp, struct obd_info *oinfo,
1265 struct obd_trans_info *oti);
1266 int (*o_setattr_async)(struct obd_export *exp, struct obd_info *oinfo,
1267 struct obd_trans_info *oti,
1268 struct ptlrpc_request_set *rqset);
1269 int (*o_getattr)(struct obd_export *exp, struct obd_info *oinfo);
1270 int (*o_getattr_async)(struct obd_export *exp, struct obd_info *oinfo,
1271 struct ptlrpc_request_set *set);
1272 int (*o_brw)(int rw, struct obd_export *exp, struct obd_info *oinfo,
1273 obd_count oa_bufs, struct brw_page *pgarr,
1274 struct obd_trans_info *oti);
1275 int (*o_merge_lvb)(struct obd_export *exp, struct lov_stripe_md *lsm,
1276 struct ost_lvb *lvb, int kms_only);
1277 int (*o_adjust_kms)(struct obd_export *exp, struct lov_stripe_md *lsm,
1278 obd_off size, int shrink);
1279 int (*o_punch)(struct obd_export *exp, struct obd_info *oinfo,
1280 struct obd_trans_info *oti,
1281 struct ptlrpc_request_set *rqset);
1282 int (*o_sync)(struct obd_export *exp, struct obdo *oa,
1283 struct lov_stripe_md *ea, obd_size start, obd_size end,
1285 int (*o_migrate)(struct lustre_handle *conn, struct lov_stripe_md *dst,
1286 struct lov_stripe_md *src, obd_size start,
1287 obd_size end, struct obd_trans_info *oti);
1288 int (*o_copy)(struct lustre_handle *dstconn, struct lov_stripe_md *dst,
1289 struct lustre_handle *srconn, struct lov_stripe_md *src,
1290 obd_size start, obd_size end, struct obd_trans_info *);
1291 int (*o_iterate)(struct lustre_handle *conn,
1292 int (*)(obd_id, obd_gr, void *),
1293 obd_id *startid, obd_gr group, void *data);
1294 int (*o_preprw)(int cmd, struct obd_export *exp, struct obdo *oa,
1295 int objcount, struct obd_ioobj *obj,
1296 struct niobuf_remote *remote, int *nr_pages,
1297 struct niobuf_local *local,
1298 struct obd_trans_info *oti,
1299 struct lustre_capa *capa);
1300 int (*o_commitrw)(int cmd, struct obd_export *exp, struct obdo *oa,
1301 int objcount, struct obd_ioobj *obj,
1302 struct niobuf_remote *remote, int pages,
1303 struct niobuf_local *local,
1304 struct obd_trans_info *oti, int rc);
1305 int (*o_enqueue)(struct obd_export *, struct obd_info *oinfo,
1306 struct ldlm_enqueue_info *einfo,
1307 struct ptlrpc_request_set *rqset);
1308 int (*o_change_cbdata)(struct obd_export *, struct lov_stripe_md *,
1309 ldlm_iterator_t it, void *data);
1310 int (*o_cancel)(struct obd_export *, struct lov_stripe_md *md,
1311 __u32 mode, struct lustre_handle *);
1312 int (*o_cancel_unused)(struct obd_export *, struct lov_stripe_md *,
1313 int flags, void *opaque);
1314 int (*o_init_export)(struct obd_export *exp);
1315 int (*o_destroy_export)(struct obd_export *exp);
1316 int (*o_extent_calc)(struct obd_export *, struct lov_stripe_md *,
1317 int cmd, obd_off *);
1319 /* llog related obd_methods */
1320 int (*o_llog_init)(struct obd_device *obd, struct obd_llog_group *grp,
1321 struct obd_device *disk_obd, int count,
1322 struct llog_catid *logid, struct obd_uuid *uuid);
1323 int (*o_llog_finish)(struct obd_device *obd, int count);
1324 int (*o_llog_connect)(struct obd_export *, struct llogd_conn_body *);
1326 /* metadata-only methods */
1327 int (*o_pin)(struct obd_export *, const struct lu_fid *fid,
1328 struct obd_capa *, struct obd_client_handle *, int flag);
1329 int (*o_unpin)(struct obd_export *, struct obd_client_handle *, int);
1331 int (*o_import_event)(struct obd_device *, struct obd_import *,
1332 enum obd_import_event);
1334 int (*o_notify)(struct obd_device *obd, struct obd_device *watched,
1335 enum obd_notify_event ev, void *data);
1337 int (*o_health_check)(struct obd_device *);
1338 struct obd_uuid *(*o_get_uuid) (struct obd_export *exp);
1341 int (*o_quotacheck)(struct obd_device *, struct obd_export *,
1342 struct obd_quotactl *);
1343 int (*o_quotactl)(struct obd_device *, struct obd_export *,
1344 struct obd_quotactl *);
1345 int (*o_quota_adjust_qunit)(struct obd_export *exp,
1346 struct quota_adjust_qunit *oqaq,
1347 struct lustre_quota_ctxt *qctxt);
1350 int (*o_ping)(struct obd_export *exp);
1353 int (*o_pool_new)(struct obd_device *obd, char *poolname);
1354 int (*o_pool_del)(struct obd_device *obd, char *poolname);
1355 int (*o_pool_add)(struct obd_device *obd, char *poolname,
1357 int (*o_pool_rem)(struct obd_device *obd, char *poolname,
1360 * NOTE: If adding ops, add another LPROCFS_OBD_OP_INIT() line
1361 * to lprocfs_alloc_obd_stats() in obdclass/lprocfs_status.c.
1362 * Also, add a wrapper function in include/linux/obd_class.h. */
1366 LUSTRE_OPC_MKDIR = (1 << 0),
1367 LUSTRE_OPC_SYMLINK = (1 << 1),
1368 LUSTRE_OPC_MKNOD = (1 << 2),
1369 LUSTRE_OPC_CREATE = (1 << 3),
1370 LUSTRE_OPC_ANY = (1 << 4)
1373 /* lmv structures */
1374 #define MEA_MAGIC_LAST_CHAR 0xb2221ca1
1375 #define MEA_MAGIC_ALL_CHARS 0xb222a11c
1376 #define MEA_MAGIC_HASH_SEGMENT 0xb222a11b
1378 #define MAX_HASH_SIZE_32 0x7fffffffUL
1379 #define MAX_HASH_SIZE 0x7fffffffffffffffULL
1380 #define MAX_HASH_HIGHEST_BIT 0x1000000000000000ULL
1383 struct mdt_body *body;
1384 struct lov_stripe_md *lsm;
1385 struct lmv_stripe_md *mea;
1386 #ifdef CONFIG_FS_POSIX_ACL
1387 struct posix_acl *posix_acl;
1389 struct mdt_remote_perm *remote_perm;
1390 struct obd_capa *mds_capa;
1391 struct obd_capa *oss_capa;
1394 struct md_open_data {
1395 struct obd_client_handle *mod_och;
1396 struct list_head mod_replay_list;
1399 struct lookup_intent;
1402 int (*m_getstatus)(struct obd_export *, struct lu_fid *,
1403 struct obd_capa **);
1404 int (*m_change_cbdata)(struct obd_export *, const struct lu_fid *,
1405 ldlm_iterator_t, void *);
1406 int (*m_close)(struct obd_export *, struct md_op_data *,
1407 struct md_open_data *, struct ptlrpc_request **);
1408 int (*m_create)(struct obd_export *, struct md_op_data *,
1409 const void *, int, int, __u32, __u32, cfs_cap_t,
1410 __u64, struct ptlrpc_request **);
1411 int (*m_done_writing)(struct obd_export *, struct md_op_data *,
1412 struct md_open_data *);
1413 int (*m_enqueue)(struct obd_export *, struct ldlm_enqueue_info *,
1414 struct lookup_intent *, struct md_op_data *,
1415 struct lustre_handle *, void *, int,
1416 struct ptlrpc_request **, int);
1417 int (*m_getattr)(struct obd_export *, const struct lu_fid *,
1418 struct obd_capa *, obd_valid, int,
1419 struct ptlrpc_request **);
1420 int (*m_getattr_name)(struct obd_export *, const struct lu_fid *,
1421 struct obd_capa *, const char *, int, obd_valid,
1422 int, __u32, struct ptlrpc_request **);
1423 int (*m_intent_lock)(struct obd_export *, struct md_op_data *,
1424 void *, int, struct lookup_intent *, int,
1425 struct ptlrpc_request **,
1426 ldlm_blocking_callback, int);
1427 int (*m_link)(struct obd_export *, struct md_op_data *,
1428 struct ptlrpc_request **);
1429 int (*m_rename)(struct obd_export *, struct md_op_data *,
1430 const char *, int, const char *, int,
1431 struct ptlrpc_request **);
1432 int (*m_is_subdir)(struct obd_export *, const struct lu_fid *,
1433 const struct lu_fid *,
1434 struct ptlrpc_request **);
1435 int (*m_setattr)(struct obd_export *, struct md_op_data *, void *,
1436 int , void *, int, struct ptlrpc_request **,
1437 struct md_open_data **mod);
1438 int (*m_sync)(struct obd_export *, const struct lu_fid *,
1439 struct obd_capa *, struct ptlrpc_request **);
1440 int (*m_readpage)(struct obd_export *, const struct lu_fid *,
1441 struct obd_capa *, __u64, struct page *,
1442 struct ptlrpc_request **);
1444 int (*m_unlink)(struct obd_export *, struct md_op_data *,
1445 struct ptlrpc_request **);
1447 int (*m_setxattr)(struct obd_export *, const struct lu_fid *,
1448 struct obd_capa *, obd_valid, const char *,
1449 const char *, int, int, int, __u32,
1450 struct ptlrpc_request **);
1452 int (*m_getxattr)(struct obd_export *, const struct lu_fid *,
1453 struct obd_capa *, obd_valid, const char *,
1454 const char *, int, int, int,
1455 struct ptlrpc_request **);
1457 int (*m_init_ea_size)(struct obd_export *, int, int, int);
1459 int (*m_get_lustre_md)(struct obd_export *, struct ptlrpc_request *,
1460 struct obd_export *, struct obd_export *,
1461 struct lustre_md *);
1463 int (*m_free_lustre_md)(struct obd_export *, struct lustre_md *);
1465 int (*m_set_open_replay_data)(struct obd_export *,
1466 struct obd_client_handle *,
1467 struct ptlrpc_request *);
1468 int (*m_clear_open_replay_data)(struct obd_export *,
1469 struct obd_client_handle *);
1470 int (*m_set_lock_data)(struct obd_export *, __u64 *, void *);
1472 ldlm_mode_t (*m_lock_match)(struct obd_export *, int,
1473 const struct lu_fid *, ldlm_type_t,
1474 ldlm_policy_data_t *, ldlm_mode_t,
1475 struct lustre_handle *);
1477 int (*m_cancel_unused)(struct obd_export *, const struct lu_fid *,
1478 ldlm_policy_data_t *, ldlm_mode_t, int flags,
1480 int (*m_renew_capa)(struct obd_export *, struct obd_capa *oc,
1481 renew_capa_cb_t cb);
1482 int (*m_unpack_capa)(struct obd_export *, struct ptlrpc_request *,
1483 const struct req_msg_field *, struct obd_capa **);
1485 int (*m_get_remote_perm)(struct obd_export *, const struct lu_fid *,
1486 struct obd_capa *, __u32,
1487 struct ptlrpc_request **);
1489 int (*m_intent_getattr_async)(struct obd_export *,
1490 struct md_enqueue_info *,
1491 struct ldlm_enqueue_info *);
1493 int (*m_revalidate_lock)(struct obd_export *,
1494 struct lookup_intent *,
1498 * NOTE: If adding ops, add another LPROCFS_MD_OP_INIT() line to
1499 * lprocfs_alloc_md_stats() in obdclass/lprocfs_status.c. Also, add a
1500 * wrapper function in include/linux/obd_class.h.
1504 struct lsm_operations {
1505 void (*lsm_free)(struct lov_stripe_md *);
1506 int (*lsm_destroy)(struct lov_stripe_md *, struct obdo *oa,
1507 struct obd_export *md_exp);
1508 void (*lsm_stripe_by_index)(struct lov_stripe_md *, int *, obd_off *,
1510 void (*lsm_stripe_by_offset)(struct lov_stripe_md *, int *, obd_off *,
1512 obd_off (*lsm_stripe_offset_by_index)(struct lov_stripe_md *, int);
1513 obd_off (*lsm_stripe_offset_by_offset)(struct lov_stripe_md *, obd_off);
1514 int (*lsm_stripe_index_by_offset)(struct lov_stripe_md *, obd_off);
1515 int (*lsm_revalidate) (struct lov_stripe_md *, struct obd_device *obd);
1516 int (*lsm_lmm_verify) (struct lov_mds_md *lmm, int lmm_bytes,
1518 int (*lsm_unpackmd) (struct lov_obd *lov, struct lov_stripe_md *lsm,
1519 struct lov_mds_md *lmm);
1522 extern const struct lsm_operations lsm_v1_ops;
1523 extern const struct lsm_operations lsm_join_ops;
1524 extern const struct lsm_operations lsm_v3_ops;
1525 static inline const struct lsm_operations *lsm_op_find(int magic)
1530 case LOV_MAGIC_JOIN:
1531 return &lsm_join_ops;
1535 CERROR("Cannot recognize lsm_magic %d\n", magic);
1540 int lvfs_check_io_health(struct obd_device *obd, struct file *file);
1542 /* Requests for obd_extent_calc() */
1543 #define OBD_CALC_STRIPE_START 1
1544 #define OBD_CALC_STRIPE_END 2
1546 static inline void obd_transno_commit_cb(struct obd_device *obd, __u64 transno,
1550 CERROR("%s: transno "LPU64" commit error: %d\n",
1551 obd->obd_name, transno, error);
1554 if (transno > obd->obd_last_committed) {
1555 CDEBUG(D_HA, "%s: transno "LPD64" committed\n",
1556 obd->obd_name, transno);
1557 obd->obd_last_committed = transno;
1558 ptlrpc_commit_replies (obd);
1560 CDEBUG(D_INFO, "%s: transno "LPD64" committed\n",
1561 obd->obd_name, transno);
1565 static inline void init_obd_quota_ops(quota_interface_t *interface,
1566 struct obd_ops *obd_ops)
1572 obd_ops->o_quotacheck = QUOTA_OP(interface, check);
1573 obd_ops->o_quotactl = QUOTA_OP(interface, ctl);
1574 obd_ops->o_quota_adjust_qunit = QUOTA_OP(interface, adjust_qunit);
1577 static inline __u64 oinfo_mdsno(struct obd_info *oinfo)
1579 return obdo_mdsno(oinfo->oi_oa);
1582 static inline struct lustre_capa *oinfo_capa(struct obd_info *oinfo)
1584 return oinfo->oi_capa;
1587 #endif /* __OBD_H */