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_target.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 struct list_head loi_ready_item;
101 struct list_head loi_hp_ready_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_ready_item);
126 CFS_INIT_LIST_HEAD(&loi->loi_hp_ready_item);
127 CFS_INIT_LIST_HEAD(&loi->loi_write_item);
128 CFS_INIT_LIST_HEAD(&loi->loi_read_item);
131 /*extent array item for describing the joined file extent info*/
133 __u64 le_start; /* extent start */
134 __u64 le_len; /* extent length */
135 int le_loi_idx; /* extent #1 loi's index in lsm loi array */
136 int le_stripe_count; /* extent stripe count*/
139 /*Lov array info for describing joined file array EA info*/
140 struct lov_array_info {
141 struct llog_logid lai_array_id; /* MDS med llog object id */
142 unsigned lai_ext_count; /* number of extent count */
143 struct lov_extent *lai_ext_array; /* extent desc array */
146 struct lov_stripe_md {
148 pid_t lsm_lock_owner; /* debugging */
151 /* Public members. */
152 __u64 lw_object_id; /* lov object id */
153 __u64 lw_object_gr; /* lov object group */
154 __u64 lw_maxbytes; /* maximum possible file size */
156 /* LOV-private members start here -- only for use in lov/. */
158 __u32 lw_stripe_size; /* size of the stripe */
159 __u32 lw_pattern; /* striping pattern (RAID0, RAID1) */
160 unsigned lw_stripe_count; /* number of objects being striped over */
161 char lw_pool_name[LOV_MAXPOOLNAME]; /* pool name */
164 struct lov_array_info *lsm_array; /*Only for joined file array info*/
165 struct lov_oinfo *lsm_oinfo[0];
168 #define lsm_object_id lsm_wire.lw_object_id
169 #define lsm_object_gr lsm_wire.lw_object_gr
170 #define lsm_maxbytes lsm_wire.lw_maxbytes
171 #define lsm_magic lsm_wire.lw_magic
172 #define lsm_stripe_size lsm_wire.lw_stripe_size
173 #define lsm_pattern lsm_wire.lw_pattern
174 #define lsm_stripe_count lsm_wire.lw_stripe_count
175 #define lsm_pool_name lsm_wire.lw_pool_name
179 typedef int (*obd_enqueue_update_f)(void *cookie, int rc);
181 /* obd info for a particular level (lov, osc). */
183 /* Lock policy. It keeps an extent which is specific for a particular
184 * OSC. (e.g. lov_prep_enqueue_set initialises extent of the policy,
185 * and osc_enqueue passes it into ldlm_lock_match & ldlm_cli_enqueue. */
186 ldlm_policy_data_t oi_policy;
187 /* Flags used for set request specific flags:
188 - while lock handling, the flags obtained on the enqueue
189 request are set here.
190 - while stats, the flags used for control delay/resend.
193 /* Lock handle specific for every OSC lock. */
194 struct lustre_handle *oi_lockh;
195 /* lsm data specific for every OSC. */
196 struct lov_stripe_md *oi_md;
197 /* obdo data specific for every OSC, if needed at all. */
199 /* statfs data specific for every OSC, if needed at all. */
200 struct obd_statfs *oi_osfs;
201 /* An update callback which is called to update some data on upper
202 * level. E.g. it is used for update lsm->lsm_oinfo at every recieved
203 * request in osc level for enqueue requests. It is also possible to
204 * update some caller data from LOV layer if needed. */
205 obd_enqueue_update_f oi_cb_up;
206 /* oss capability, its type is obd_capa in client to avoid copy.
207 * in contrary its type is lustre_capa in OSS. */
211 /* compare all relevant fields. */
212 static inline int lov_stripe_md_cmp(struct lov_stripe_md *m1,
213 struct lov_stripe_md *m2)
216 * ->lsm_wire contains padding, but it should be zeroed out during
219 return memcmp(&m1->lsm_wire, &m2->lsm_wire, sizeof m1->lsm_wire);
222 void lov_stripe_lock(struct lov_stripe_md *md);
223 void lov_stripe_unlock(struct lov_stripe_md *md);
226 struct list_head typ_chain;
227 struct obd_ops *typ_dt_ops;
228 struct md_ops *typ_md_ops;
229 cfs_proc_dir_entry_t *typ_procroot;
232 struct lu_device_type *typ_lu;
233 spinlock_t obd_type_lock;
243 /* Individual type definitions */
245 struct ost_server_data;
247 /* hold common fields for "target" device */
248 struct obd_device_target {
249 struct super_block *obt_sb;
250 /** last_rcvd file */
251 struct file *obt_rcvd_filp;
252 /** server data in last_rcvd file */
253 struct lr_server_data *obt_lsd;
254 /** Lock protecting client bitmap */
255 spinlock_t obt_client_bitmap_lock;
256 /** Bitmap of known clients */
257 unsigned long *obt_client_bitmap;
258 /** Server last transaction number */
259 __u64 obt_last_transno;
260 /** Lock protecting last transaction number */
261 spinlock_t obt_translock;
262 /** Number of mounts */
263 __u64 obt_mount_count;
264 struct semaphore obt_quotachecking;
265 struct lustre_quota_ctxt obt_qctxt;
266 lustre_quota_version_t obt_qfmt;
267 struct rw_semaphore obt_rwsem;
268 struct vfsmount *obt_vfsmnt;
269 struct file *obt_health_check_filp;
274 LLOG_CONFIG_ORIG_CTXT = 0,
275 LLOG_CONFIG_REPL_CTXT,
276 LLOG_MDS_OST_ORIG_CTXT,
277 LLOG_MDS_OST_REPL_CTXT,
284 LLOG_LOVEA_ORIG_CTXT,
285 LLOG_LOVEA_REPL_CTXT,
286 LLOG_CHANGELOG_ORIG_CTXT, /**< changelog generation on mdd */
287 LLOG_CHANGELOG_REPL_CTXT, /**< changelog access on clients */
288 LLOG_CHANGELOG_USER_ORIG_CTXT, /**< for multiple changelog consumers */
292 #define FILTER_SUBDIR_COUNT 32 /* set to zero for no subdirs */
294 struct filter_subdirs {
295 cfs_dentry_t *dentry[FILTER_SUBDIR_COUNT];
305 /* NB this field MUST be first */
306 struct obd_device_target fo_obt;
307 struct lu_target fo_lut;
308 const char *fo_fstype;
311 cfs_dentry_t *fo_dentry_O;
312 cfs_dentry_t **fo_dentry_O_groups;
313 struct filter_subdirs *fo_dentry_O_sub;
314 struct semaphore fo_init_lock; /* group initialization lock */
315 int fo_committed_group;
318 spinlock_t fo_objidlock; /* protect fo_lastobjid */
320 unsigned long fo_destroys_in_progress;
321 struct semaphore fo_create_locks[FILTER_SUBDIR_COUNT];
323 struct list_head fo_export_list;
326 obd_size fo_tot_dirty; /* protected by obd_osfs_lock */
327 obd_size fo_tot_granted; /* all values in bytes */
328 obd_size fo_tot_pending;
329 int fo_tot_granted_clients;
331 obd_size fo_readcache_max_filesize;
332 int fo_read_cache:1, /**< enable read-only cache */
333 fo_writethrough_cache:1,/**< read cache writes */
334 fo_mds_ost_sync:1, /**< MDS-OST orphan recovery*/
335 fo_raid_degraded:1;/**< RAID device degraded */
337 struct obd_import *fo_mdc_imp;
338 struct obd_uuid fo_mdc_uuid;
339 struct lustre_handle fo_mdc_conn;
340 struct file **fo_last_objid_files;
341 __u64 *fo_last_objids; /* last created objid for groups,
342 * protected by fo_objidlock */
344 struct semaphore fo_alloc_lock;
346 atomic_t fo_r_in_flight;
347 atomic_t fo_w_in_flight;
350 * per-filter pool of kiobuf's allocated by filter_common_setup() and
351 * torn down by filter_cleanup(). Contains OST_NUM_THREADS elements of
352 * which ->fo_iobuf_count were allocated.
354 * This pool contains kiobuf used by
355 * filter_{prep,commit}rw_{read,write}() and is shared by all OST
358 * Locking: none, each OST thread uses only one element, determined by
359 * its "ordinal number", ->t_id.
361 struct filter_iobuf **fo_iobuf_pool;
364 struct list_head fo_llog_list;
365 spinlock_t fo_llog_list_lock;
367 struct brw_stats fo_filter_stats;
368 struct lustre_quota_ctxt fo_quota_ctxt;
369 spinlock_t fo_quotacheck_lock;
370 atomic_t fo_quotachecking;
372 int fo_fmd_max_num; /* per exp filter_mod_data */
373 int fo_fmd_max_age; /* jiffies to fmd expiry */
376 rwlock_t fo_sptlrpc_lock;
377 struct sptlrpc_rule_set fo_sptlrpc_rset;
379 /* capability related */
380 unsigned int fo_fl_oss_capa;
381 struct list_head fo_capa_keys;
382 struct hlist_head *fo_capa_hash;
383 struct llog_commit_master *fo_lcm;
387 #define fo_translock fo_obt.obt_translock
388 #define fo_rcvd_filp fo_obt.obt_rcvd_filp
389 #define fo_fsd fo_obt.obt_lsd
390 #define fo_last_rcvd_slots fo_obt.obt_client_bitmap
391 #define fo_mount_count fo_obt.obt_mount_count
392 #define fo_vfsmnt fo_obt.obt_vfsmnt
394 struct timeout_item {
395 enum timeout_event ti_event;
396 cfs_time_t ti_timeout;
399 struct list_head ti_obd_list;
400 struct list_head ti_chain;
403 #define OSC_MAX_RIF_DEFAULT 8
404 #define OSC_MAX_RIF_MAX 256
405 #define OSC_MAX_DIRTY_DEFAULT (OSC_MAX_RIF_DEFAULT * 4)
406 #define OSC_MAX_DIRTY_MB_MAX 2048 /* arbitrary, but < MAX_LONG bytes */
407 #define OSC_DEFAULT_RESENDS 10
409 #define MDC_MAX_RIF_DEFAULT 8
410 #define MDC_MAX_RIF_MAX 512
415 struct rw_semaphore cl_sem;
416 struct obd_uuid cl_target_uuid;
417 struct obd_import *cl_import; /* ptlrpc connection state */
419 /* max_mds_easize is purely a performance thing so we don't have to
420 * call obd_size_diskmd() all the time. */
421 int cl_default_mds_easize;
422 int cl_max_mds_easize;
423 int cl_max_mds_cookiesize;
425 enum lustre_sec_part cl_sp_me;
426 enum lustre_sec_part cl_sp_to;
427 struct sptlrpc_flavor cl_flvr_mgc; /* fixed flavor of mgc->mgs */
429 //struct llog_canceld_ctxt *cl_llcd; /* it's included by obd_llog_ctxt */
430 void *cl_llcd_offset;
432 /* the grant values are protected by loi_list_lock below */
433 long cl_dirty; /* all _dirty_ in bytes */
434 long cl_dirty_max; /* allowed w/o rpc */
435 long cl_dirty_transit; /* dirty synchronous */
436 long cl_avail_grant; /* bytes of credit for ost */
437 long cl_lost_grant; /* lost credits (trunc) */
438 struct list_head cl_cache_waiters; /* waiting for cache/grant */
439 cfs_time_t cl_next_shrink_grant; /* jiffies */
440 struct list_head cl_grant_shrink_list; /* Timeout event list */
441 struct semaphore cl_grant_sem; /*grant shrink list semaphore*/
442 int cl_grant_shrink_interval; /* seconds */
444 /* keep track of objects that have lois that contain pages which
445 * have been queued for async brw. this lock also protects the
446 * lists of osc_client_pages that hang off of the loi */
448 * ->cl_loi_list_lock protects consistency of
449 * ->cl_loi_{ready,read,write}_list. ->ap_make_ready() and
450 * ->ap_completion() call-backs are executed under this lock. As we
451 * cannot guarantee that these call-backs never block on all platforms
452 * (as a matter of fact they do block on Mac OS X), type of
453 * ->cl_loi_list_lock is platform dependent: it's a spin-lock on Linux
454 * and blocking mutex on Mac OS X. (Alternative is to make this lock
455 * blocking everywhere, but we don't want to slow down fast-path of
456 * our main platform.)
458 * Exact type of ->cl_loi_list_lock is defined in arch/obd.h together
459 * with client_obd_list_{un,}lock() and
460 * client_obd_list_lock_{init,done}() functions.
462 client_obd_lock_t cl_loi_list_lock;
463 struct list_head cl_loi_ready_list;
464 struct list_head cl_loi_hp_ready_list;
465 struct list_head cl_loi_write_list;
466 struct list_head cl_loi_read_list;
469 /* just a sum of the loi/lop pending numbers to be exported by /proc */
470 int cl_pending_w_pages;
471 int cl_pending_r_pages;
472 int cl_max_pages_per_rpc;
473 int cl_max_rpcs_in_flight;
474 struct obd_histogram cl_read_rpc_hist;
475 struct obd_histogram cl_write_rpc_hist;
476 struct obd_histogram cl_read_page_hist;
477 struct obd_histogram cl_write_page_hist;
478 struct obd_histogram cl_read_offset_hist;
479 struct obd_histogram cl_write_offset_hist;
481 /* number of in flight destroy rpcs is limited to max_rpcs_in_flight */
482 atomic_t cl_destroy_in_flight;
483 cfs_waitq_t cl_destroy_waitq;
485 struct mdc_rpc_lock *cl_rpc_lock;
486 struct mdc_rpc_lock *cl_setattr_lock;
487 struct mdc_rpc_lock *cl_close_lock;
488 struct osc_creator cl_oscc;
491 struct semaphore cl_mgc_sem;
492 struct vfsmount *cl_mgc_vfsmnt;
493 struct dentry *cl_mgc_configs_dir;
494 atomic_t cl_mgc_refcount;
495 struct obd_export *cl_mgc_mgsexp;
497 /* checksumming for data sent over the network */
498 unsigned int cl_checksum:1; /* 0 = disabled, 1 = enabled */
499 /* supported checksum types that are worked out at connect time */
500 __u32 cl_supp_cksum_types;
501 /* checksum algorithm to be used */
502 cksum_type_t cl_cksum_type;
504 /* also protected by the poorly named _loi_list_lock lock above */
505 struct osc_async_rc cl_ar;
507 /* used by quotacheck */
508 int cl_qchk_stat; /* quotacheck stat of the peer */
510 /* sequence manager */
511 struct lu_client_seq *cl_seq;
513 atomic_t cl_resends; /* resend count */
515 #define obd2cli_tgt(obd) ((char *)(obd)->u.cli.cl_target_uuid.uuid)
517 #define CL_NOT_QUOTACHECKED 1 /* client->cl_qchk_stat init value */
520 struct ptlrpc_service *mgs_service;
521 struct vfsmount *mgs_vfsmnt;
522 struct super_block *mgs_sb;
523 struct dentry *mgs_configs_dir;
524 struct dentry *mgs_fid_de;
525 struct list_head mgs_fs_db_list;
526 struct semaphore mgs_sem;
527 cfs_proc_dir_entry_t *mgs_proc_live;
531 /* NB this field MUST be first */
532 struct obd_device_target mds_obt;
533 struct ptlrpc_service *mds_service;
534 struct ptlrpc_service *mds_setattr_service;
535 struct ptlrpc_service *mds_readpage_service;
536 cfs_dentry_t *mds_fid_de;
538 int mds_max_cookiesize;
540 unsigned long mds_atime_diff;
541 struct semaphore mds_epoch_sem;
542 struct ll_fid mds_rootfid;
543 cfs_dentry_t *mds_pending_dir;
544 cfs_dentry_t *mds_logs_dir;
545 cfs_dentry_t *mds_objects_dir;
546 struct llog_handle *mds_cfg_llh;
547 struct obd_device *mds_osc_obd; /* XXX lov_obd */
548 struct obd_uuid mds_lov_uuid;
550 struct obd_export *mds_osc_exp; /* XXX lov_exp */
551 struct lov_desc mds_lov_desc;
554 /* mark pages dirty for write. */
555 bitmap_t *mds_lov_page_dirty;
556 /* array for store pages with obd_id */
557 void **mds_lov_page_array;
558 /* file for store objid */
559 struct file *mds_lov_objid_filp;
560 __u32 mds_lov_objid_count;
561 __u32 mds_lov_objid_max_index;
562 __u32 mds_lov_objid_lastpage;
563 __u32 mds_lov_objid_lastidx;
566 struct lustre_quota_info mds_quota_info;
567 struct semaphore mds_qonoff_sem;
568 struct semaphore mds_health_sem;
569 unsigned long mds_fl_user_xattr:1,
571 mds_evict_ost_nids:1,
575 mds_fl_target:1; /* mds have one or
578 struct upcall_cache *mds_identity_cache;
580 /* for capability keys update */
581 struct lustre_capa_key *mds_capa_keys;
582 struct rw_semaphore mds_notify_lock;
585 #define mds_transno_lock mds_obt.obt_translock
586 #define mds_rcvd_filp mds_obt.obt_rcvd_filp
587 #define mds_server_data mds_obt.obt_lsd
588 #define mds_client_bitmap mds_obt.obt_client_bitmap
589 #define mds_mount_count mds_obt.obt_mount_count
590 #define mds_last_transno mds_obt.obt_last_transno
591 #define mds_vfsmnt mds_obt.obt_vfsmnt
594 extern __u32 mds_max_ost_index;
596 #define MDS_LOV_ALLOC_SIZE (CFS_PAGE_SIZE)
598 #define OBJID_PER_PAGE() (MDS_LOV_ALLOC_SIZE / sizeof(obd_id))
600 #define MDS_LOV_OBJID_PAGES_COUNT (mds_max_ost_index/OBJID_PER_PAGE())
602 extern int mds_lov_init_objids(struct obd_device *obd);
603 extern void mds_lov_destroy_objids(struct obd_device *obd);
616 struct lustre_handle eo_nl_lock;
621 struct ptlrpc_service *ost_service;
622 struct ptlrpc_service *ost_create_service;
623 struct ptlrpc_service *ost_io_service;
624 struct semaphore ost_health_sem;
627 struct echo_client_obd {
628 struct obd_export *ec_exp; /* the local connection to osc/lov */
630 struct list_head ec_objects;
631 struct list_head ec_locks;
637 struct obd_uuid lqo_uuid; /* ptlrpc's c_remote_uuid */
638 struct list_head lqo_oss_list; /* link to lov_qos */
639 __u64 lqo_bavail; /* total bytes avail on OSS */
640 __u64 lqo_penalty; /* current penalty */
641 __u64 lqo_penalty_per_obj; /* penalty decrease every obj*/
642 time_t lqo_used; /* last used time, seconds */
643 __u32 lqo_ost_count; /* number of osts on this oss */
647 struct lov_qos_oss *ltq_oss; /* oss info */
648 __u64 ltq_penalty; /* current penalty */
649 __u64 ltq_penalty_per_obj; /* penalty decrease every obj*/
650 __u64 ltq_weight; /* net weighting */
651 time_t ltq_used; /* last used time, seconds */
652 unsigned int ltq_usable:1; /* usable for striping */
655 /* Generic subset of OSTs */
657 __u32 *op_array; /* array of index of
659 unsigned int op_count; /* number of OSTs in the array */
660 unsigned int op_size; /* allocated size of lp_array */
661 struct rw_semaphore op_rw_sem; /* to protect ost_pool use */
664 /* Round-robin allocator data */
666 __u32 lqr_start_idx; /* start index of new inode */
667 __u32 lqr_offset_idx; /* aliasing for start_idx */
668 int lqr_start_count; /* reseed counter */
669 struct ost_pool lqr_pool; /* round-robin optimized list */
670 unsigned long lqr_dirty:1; /* recalc round-robin list */
673 struct lov_statfs_data {
674 struct obd_info lsd_oi;
675 struct obd_statfs lsd_statfs;
677 /* Stripe placement optimization */
679 struct list_head lq_oss_list; /* list of OSSs that targets use */
680 struct rw_semaphore lq_rw_sem;
681 __u32 lq_active_oss_count;
682 unsigned int lq_prio_free; /* priority for free space */
683 unsigned int lq_threshold_rr;/* priority for rr */
684 struct lov_qos_rr lq_rr; /* round robin qos data */
685 unsigned long lq_dirty:1, /* recalc qos data */
686 lq_same_space:1,/* the ost's all have approx.
687 the same space avail */
688 lq_reset:1, /* zero current penalties */
689 lq_statfs_in_progress:1; /* statfs op in progress */
690 /* qos statfs data */
691 struct lov_statfs_data *lq_statfs_data;
692 cfs_waitq_t lq_statfs_waitq; /* waitqueue to notify statfs
693 * requests completion */
696 struct lov_tgt_desc {
697 struct list_head ltd_kill;
698 struct obd_uuid ltd_uuid;
699 struct obd_device *ltd_obd;
700 struct obd_export *ltd_exp;
701 struct ltd_qos ltd_qos; /* qos info per target */
703 __u32 ltd_index; /* index in lov_obd->tgts */
704 unsigned long ltd_active:1,/* is this target up for requests */
705 ltd_activate:1,/* should this target be activated */
706 ltd_reap:1; /* should this target be deleted */
710 #define pool_tgt_size(_p) _p->pool_obds.op_size
711 #define pool_tgt_count(_p) _p->pool_obds.op_count
712 #define pool_tgt_array(_p) _p->pool_obds.op_array
713 #define pool_tgt_rw_sem(_p) _p->pool_obds.op_rw_sem
714 #define pool_tgt(_p, _i) _p->pool_lov->lov_tgts[_p->pool_obds.op_array[_i]]
717 char pool_name[LOV_MAXPOOLNAME + 1]; /* name of pool */
718 struct ost_pool pool_obds; /* pool members */
719 atomic_t pool_refcount; /* pool ref. counter */
720 struct lov_qos_rr pool_rr; /* round robin qos */
721 struct hlist_node pool_hash; /* access by poolname */
722 struct list_head pool_list; /* serial access */
723 cfs_proc_dir_entry_t *pool_proc_entry; /* file in /proc */
724 struct lov_obd *pool_lov; /* lov obd to which this
729 struct lov_desc desc;
730 struct lov_tgt_desc **lov_tgts; /* sparse array */
731 struct ost_pool lov_packed; /* all OSTs in a packed
733 struct semaphore lov_lock;
734 struct obd_connect_data lov_ocd;
735 struct lov_qos lov_qos; /* qos info per lov */
736 atomic_t lov_refcount;
737 __u32 lov_tgt_count; /* how many OBD's */
738 __u32 lov_active_tgt_count; /* how many active */
739 __u32 lov_death_row;/* tgts scheduled to be deleted */
740 __u32 lov_tgt_size; /* size of tgts array */
743 lustre_hash_t *lov_pools_hash_body; /* used for key access */
744 struct list_head lov_pool_list; /* used for sequential access */
745 cfs_proc_dir_entry_t *lov_pool_proc_entry;
746 enum lustre_sec_part lov_sp_me;
749 struct lmv_tgt_desc {
750 struct obd_uuid ltd_uuid;
751 struct obd_export *ltd_exp;
752 int ltd_active; /* is this target up for requests */
754 struct semaphore ltd_fid_sem;
757 enum placement_policy {
758 PLACEMENT_CHAR_POLICY = 0,
759 PLACEMENT_NID_POLICY = 1,
760 PLACEMENT_INVAL_POLICY = 2,
764 typedef enum placement_policy placement_policy_t;
768 struct lu_client_fld lmv_fld;
770 placement_policy_t lmv_placement;
771 struct lmv_desc desc;
772 struct obd_uuid cluuid;
773 struct obd_export *exp;
780 struct semaphore init_sem;
782 struct lmv_tgt_desc *tgts;
785 struct obd_connect_data *datas;
788 struct obd_connect_data conn_data;
791 struct niobuf_local {
796 cfs_dentry_t *dentry;
801 #define LUSTRE_FLD_NAME "fld"
802 #define LUSTRE_SEQ_NAME "seq"
804 #define LUSTRE_CMM_NAME "cmm"
805 #define LUSTRE_MDD_NAME "mdd"
806 #define LUSTRE_OSD_NAME "osd"
807 #define LUSTRE_VVP_NAME "vvp"
808 #define LUSTRE_LMV_NAME "lmv"
809 #define LUSTRE_CMM_MDC_NAME "cmm-mdc"
810 #define LUSTRE_SLP_NAME "slp"
812 /* obd device type names */
813 /* FIXME all the references to LUSTRE_MDS_NAME should be swapped with LUSTRE_MDT_NAME */
814 #define LUSTRE_MDS_NAME "mds"
815 #define LUSTRE_MDT_NAME "mdt"
816 #define LUSTRE_MDC_NAME "mdc"
817 #define LUSTRE_OSS_NAME "ost" /* FIXME change name to oss */
818 #define LUSTRE_OST_NAME "obdfilter" /* FIXME change name to ost */
819 #define LUSTRE_OSC_NAME "osc"
820 #define LUSTRE_LOV_NAME "lov"
821 #define LUSTRE_MGS_NAME "mgs"
822 #define LUSTRE_MGC_NAME "mgc"
824 #define LUSTRE_CACHEOBD_NAME "cobd"
825 #define LUSTRE_ECHO_NAME "obdecho"
826 #define LUSTRE_ECHO_CLIENT_NAME "echo_client"
828 /* Constant obd names (post-rename) */
829 #define LUSTRE_MDS_OBDNAME "MDS"
830 #define LUSTRE_OSS_OBDNAME "OSS"
831 #define LUSTRE_MGS_OBDNAME "MGS"
832 #define LUSTRE_MGC_OBDNAME "MGC"
834 /* Don't conflict with on-wire flags OBD_BRW_WRITE, etc */
835 #define N_LOCAL_TEMP_PAGE 0x10000000
837 struct obd_trans_info {
840 /* Only used on the server side for tracking acks. */
841 struct oti_req_ack_lock {
842 struct lustre_handle lock;
846 struct llog_cookie oti_onecookie;
847 struct llog_cookie *oti_logcookies;
850 /* initial thread handling transaction */
851 struct ptlrpc_thread * oti_thread;
854 __u64 oti_pre_version;
856 struct obd_uuid *oti_ost_uuid;
859 static inline void oti_init(struct obd_trans_info *oti,
860 struct ptlrpc_request *req)
864 memset(oti, 0, sizeof(*oti));
869 oti->oti_xid = req->rq_xid;
870 /** VBR: take versions from request */
871 if (req->rq_reqmsg != NULL &&
872 lustre_msg_get_flags(req->rq_reqmsg) & MSG_REPLAY) {
873 __u64 *pre_version = lustre_msg_get_versions(req->rq_reqmsg);
874 oti->oti_pre_version = pre_version ? pre_version[0] : 0;
875 oti->oti_transno = lustre_msg_get_transno(req->rq_reqmsg);
878 /** called from mds_create_objects */
879 if (req->rq_repmsg != NULL)
880 oti->oti_transno = lustre_msg_get_transno(req->rq_repmsg);
881 oti->oti_thread = req->rq_svc_thread;
882 if (req->rq_reqmsg != NULL)
883 oti->oti_conn_cnt = lustre_msg_get_conn_cnt(req->rq_reqmsg);
886 static inline void oti_alloc_cookies(struct obd_trans_info *oti,int num_cookies)
891 if (num_cookies == 1)
892 oti->oti_logcookies = &oti->oti_onecookie;
894 OBD_ALLOC(oti->oti_logcookies,
895 num_cookies * sizeof(oti->oti_onecookie));
897 oti->oti_numcookies = num_cookies;
900 static inline void oti_free_cookies(struct obd_trans_info *oti)
902 if (!oti || !oti->oti_logcookies)
905 if (oti->oti_logcookies == &oti->oti_onecookie)
906 LASSERT(oti->oti_numcookies == 1);
908 OBD_FREE(oti->oti_logcookies,
909 oti->oti_numcookies * sizeof(oti->oti_onecookie));
910 oti->oti_logcookies = NULL;
911 oti->oti_numcookies = 0;
915 * Events signalled through obd_notify() upcall-chain.
917 enum obd_notify_event {
918 /* Device connect start */
920 /* Device activated */
922 /* Device deactivated */
924 /* Device disconnected */
926 /* Connect data for import were changed */
929 OBD_NOTIFY_SYNC_NONBLOCK,
931 /* Configuration event */
933 /* Trigger quota recovery */
937 /* bit-mask flags for config events */
939 CONFIG_LOG = 0x1, /* finished processing config log */
940 CONFIG_SYNC = 0x2, /* mdt synced 1 ost */
941 CONFIG_TARGET = 0x4 /* one target is added */
945 * Data structure used to pass obd_notify()-event to non-obd listeners (llite
946 * and liblustre being main examples).
948 struct obd_notify_upcall {
949 int (*onu_upcall)(struct obd_device *host, struct obd_device *watched,
950 enum obd_notify_event ev, void *owner, void *data);
951 /* Opaque datum supplied by upper layer listener */
955 struct target_recovery_data {
956 svc_handler_t trd_recovery_handler;
957 pid_t trd_processing_task;
958 struct completion trd_starting;
959 struct completion trd_finishing;
963 FILTER_GROUP_MDS0 = 0,
964 FILTER_GROUP_LLOG = 1,
965 FILTER_GROUP_ECHO = 2 ,
966 FILTER_GROUP_MDS1_N_BASE = 3
969 static inline __u64 obdo_mdsno(struct obdo *oa)
972 LASSERT(oa->o_gr >= FILTER_GROUP_MDS1_N_BASE);
973 return oa->o_gr - FILTER_GROUP_MDS1_N_BASE + 1;
978 static inline int mdt_to_obd_objgrp(int mdtid)
981 * MDS0 uses group 0 always, other MDSes will use groups from
982 * FILTER_GROUP_MDS1_N_BASE
985 return FILTER_GROUP_MDS1_N_BASE + mdtid - 1;
990 * In HEAD for CMD, the object is created in group number which is 3>=
991 * or indexing starts from 3. To test this assertions are added to disallow
992 * group 0. But to run 2.0 mds server on 1.8.x disk format (i.e. interop_mode)
993 * object in group 0 needs to be allowed.
994 * So for interop mode following changes needs to be done:
995 * 1. No need to assert on group 0 or allow group 0
996 * 2. The group number indexing starts from 0 instead of 3
999 #define CHECK_MDS_GROUP(group) (group == FILTER_GROUP_MDS0 || \
1000 group > FILTER_GROUP_MDS1_N_BASE)
1001 #define LASSERT_MDS_GROUP(group) LASSERT(CHECK_MDS_GROUP(group))
1003 struct obd_llog_group {
1004 struct list_head olg_list;
1006 struct llog_ctxt *olg_ctxts[LLOG_MAX_CTXTS];
1007 cfs_waitq_t olg_waitq;
1008 spinlock_t olg_lock;
1009 struct obd_export *olg_exp;
1010 int olg_initializing;
1011 struct semaphore olg_cat_processing;
1014 /* corresponds to one of the obd's */
1015 #define MAX_OBD_NAME 128
1016 #define OBD_DEVICE_MAGIC 0XAB5CD6EF
1017 #define OBD_DEV_BY_DEVNAME 0xffffd0de
1019 struct obd_type *obd_type;
1022 /* common and UUID name of this device */
1023 char obd_name[MAX_OBD_NAME];
1024 struct obd_uuid obd_uuid;
1026 struct lu_device *obd_lu_dev;
1029 unsigned long obd_attached:1, /* finished attach */
1030 obd_set_up:1, /* finished setup */
1031 obd_recovering:1, /* there are recoverable clients */
1032 obd_abort_recovery:1,/* recovery expired */
1033 obd_version_recov:1, /* obd uses version checking */
1034 obd_recovery_expired:1,
1035 obd_replayable:1, /* recovery is enabled; inform clients */
1036 obd_no_transno:1, /* no committed-transno notification */
1037 obd_no_recov:1, /* fail instead of retry messages */
1038 obd_req_replaying:1, /* replaying requests */
1039 obd_stopping:1, /* started cleanup */
1040 obd_starting:1, /* started setup */
1041 obd_force:1, /* cleanup with > 0 obd refcount */
1042 obd_fail:1, /* cleanup with failover */
1043 obd_async_recov:1, /* allow asyncronous orphan cleanup */
1044 obd_no_conn:1, /* deny new connections */
1045 obd_inactive:1, /* device active/inactive
1046 * (for /proc/status only!!) */
1047 obd_process_conf:1; /* device is processing mgs config */
1048 /* uuid-export hash body */
1049 struct lustre_hash *obd_uuid_hash;
1050 /* nid-export hash body */
1051 struct lustre_hash *obd_nid_hash;
1052 /* nid stats body */
1053 struct lustre_hash *obd_nid_stats_hash;
1054 struct list_head obd_nid_stats;
1055 atomic_t obd_refcount;
1056 cfs_waitq_t obd_refcount_waitq;
1057 struct list_head obd_exports;
1058 struct list_head obd_unlinked_exports;
1059 struct list_head obd_delayed_exports;
1060 int obd_num_exports;
1061 spinlock_t obd_nid_lock;
1062 struct ldlm_namespace *obd_namespace;
1063 struct ptlrpc_client obd_ldlm_client; /* XXX OST/MDS only */
1064 /* a spinlock is OK for what we do now, may need a semaphore later */
1065 spinlock_t obd_dev_lock;
1066 struct semaphore obd_dev_sem;
1067 __u64 obd_last_committed;
1068 struct fsfilt_operations *obd_fsops;
1069 spinlock_t obd_osfs_lock;
1070 struct obd_statfs obd_osfs; /* locked by obd_osfs_lock */
1072 struct lvfs_run_ctxt obd_lvfs_ctxt;
1073 struct obd_llog_group obd_olg; /* default llog group */
1074 struct obd_device *obd_observer;
1075 struct obd_notify_upcall obd_upcall;
1076 struct obd_export *obd_self_export;
1077 /* list of exports in LRU order, for ping evictor, with obd_dev_lock */
1078 struct list_head obd_exports_timed;
1079 time_t obd_eviction_timer; /* for ping evictor */
1081 /* XXX encapsulate all this recovery data into one struct */
1082 svc_handler_t obd_recovery_handler;
1083 pid_t obd_processing_task;
1085 int obd_max_recoverable_clients;
1086 int obd_connected_clients;
1087 int obd_stale_clients;
1088 int obd_delayed_clients;
1089 spinlock_t obd_processing_task_lock; /* BH lock (timer) */
1090 __u64 obd_next_recovery_transno;
1091 int obd_replayed_requests;
1092 int obd_requests_queued_for_recovery;
1093 cfs_waitq_t obd_next_transno_waitq;
1094 cfs_timer_t obd_recovery_timer;
1095 time_t obd_recovery_start; /* seconds */
1096 time_t obd_recovery_end; /* seconds, for lprocfs_status */
1097 time_t obd_recovery_max_time; /* seconds, bz13079 */
1098 int obd_recovery_timeout;
1100 /* new recovery stuff from CMD2 */
1101 struct target_recovery_data obd_recovery_data;
1102 int obd_replayed_locks;
1103 atomic_t obd_req_replay_clients;
1104 atomic_t obd_lock_replay_clients;
1105 struct list_head obd_req_replay_queue;
1106 struct list_head obd_lock_replay_queue;
1107 struct list_head obd_final_req_queue;
1108 int obd_recovery_stage;
1111 struct obd_device_target obt;
1112 struct filter_obd filter;
1114 struct client_obd cli;
1116 struct echo_client_obd echo_client;
1117 struct echo_obd echo;
1122 /* Fields used by LProcFS */
1123 unsigned int obd_cntr_base;
1124 struct lprocfs_stats *obd_stats;
1126 unsigned int md_cntr_base;
1127 struct lprocfs_stats *md_stats;
1129 cfs_proc_dir_entry_t *obd_proc_entry;
1130 cfs_proc_dir_entry_t *obd_proc_exports_entry;
1131 cfs_proc_dir_entry_t *obd_svc_procroot;
1132 struct lprocfs_stats *obd_svc_stats;
1133 atomic_t obd_evict_inprogress;
1134 cfs_waitq_t obd_evict_inprogress_waitq;
1137 * Ldlm pool part. Save last calculated SLV and Limit.
1139 rwlock_t obd_pool_lock;
1144 * A list of outstanding class_incref()'s against this obd. For
1147 struct lu_ref obd_reference;
1150 #define OBD_LLOG_FL_SENDNOW 0x0001
1152 enum obd_cleanup_stage {
1153 /* Special case hack for MDS LOVs */
1155 /* can be directly mapped to .ldto_device_fini() */
1156 OBD_CLEANUP_EXPORTS,
1159 /* get/set_info keys */
1160 #define KEY_BLOCKSIZE_BITS "blocksize_bits"
1161 #define KEY_BLOCKSIZE "blocksize"
1162 #define KEY_CAPA_KEY "capa_key"
1163 #define KEY_CHANGELOG_CLEAR "changelog_clear"
1164 #define KEY_FID2PATH "fid2path"
1165 #define KEY_CHECKSUM "checksum"
1166 #define KEY_CLEAR_FS "clear_fs"
1167 #define KEY_CONN_DATA "conn_data"
1168 #define KEY_EVICT_BY_NID "evict_by_nid"
1169 #define KEY_FIEMAP "fiemap"
1170 #define KEY_FLUSH_CTX "flush_ctx"
1171 #define KEY_INIT_RECOV_BACKUP "init_recov_bk"
1172 #define KEY_INIT_RECOV "initial_recov"
1173 #define KEY_LAST_ID "last_id"
1174 #define KEY_LOCK_TO_STRIPE "lock_to_stripe"
1175 #define KEY_LOVDESC "lovdesc"
1176 #define KEY_LOV_IDX "lov_idx"
1177 #define KEY_MAX_EASIZE "max_easize"
1178 #define KEY_MDS_CONN "mds_conn"
1179 #define KEY_MGSSEC "mgssec"
1180 #define KEY_NEXT_ID "next_id"
1181 #define KEY_READ_ONLY "read-only"
1182 #define KEY_REGISTER_TARGET "register_target"
1183 #define KEY_REVIMP_UPD "revimp_update"
1184 #define KEY_SET_FS "set_fs"
1185 #define KEY_SPTLRPC_CONF "sptlrpc_conf"
1186 #define KEY_UNLINKED "unlinked"
1188 #define KEY_INTERMDS "inter_mds"
1189 #define KEY_ASYNC "async"
1190 #define KEY_GRANT_SHRINK "grant_shrink"
1194 static inline int it_to_lock_mode(struct lookup_intent *it)
1196 /* CREAT needs to be tested before open (both could be set) */
1197 if (it->it_op & IT_CREAT)
1199 else if (it->it_op & (IT_READDIR | IT_GETATTR | IT_OPEN | IT_LOOKUP))
1202 LASSERTF(0, "Invalid it_op: %d\n", it->it_op);
1207 struct lu_fid op_fid1; /* operation fid1 (usualy parent) */
1208 struct lu_fid op_fid2; /* operation fid2 (usualy child) */
1209 struct lu_fid op_fid3; /* 2 extra fids to find conflicting */
1210 struct lu_fid op_fid4; /* to the operation locks. */
1211 mdsno_t op_mds; /* what mds server open will go to */
1212 struct lustre_handle op_handle;
1214 const char *op_name;
1217 struct lmv_stripe_md *op_mea1;
1218 struct lmv_stripe_md *op_mea2;
1219 __u32 op_suppgids[2];
1225 /* iattr fields and blocks. */
1226 struct iattr op_attr;
1228 #if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,14)
1229 unsigned int op_attr_flags;
1232 loff_t op_attr_blocks;
1234 /* Size-on-MDS epoch and flags. */
1239 struct obd_capa *op_capa1;
1240 struct obd_capa *op_capa2;
1242 /* Various operation flags. */
1245 /* Operation type */
1249 struct md_enqueue_info;
1250 /* metadata stat-ahead */
1251 typedef int (* md_enqueue_cb_t)(struct ptlrpc_request *req,
1252 struct md_enqueue_info *minfo,
1255 struct md_enqueue_info {
1256 struct md_op_data mi_data;
1257 struct lookup_intent mi_it;
1258 struct lustre_handle mi_lockh;
1259 struct dentry *mi_dentry;
1260 md_enqueue_cb_t mi_cb;
1261 unsigned int mi_generation;
1266 struct module *o_owner;
1267 int (*o_iocontrol)(unsigned int cmd, struct obd_export *exp, int len,
1268 void *karg, void *uarg);
1269 int (*o_get_info)(struct obd_export *, __u32 keylen, void *key,
1270 __u32 *vallen, void *val, struct lov_stripe_md *lsm);
1271 int (*o_set_info_async)(struct obd_export *, __u32 keylen, void *key,
1272 __u32 vallen, void *val,
1273 struct ptlrpc_request_set *set);
1274 int (*o_attach)(struct obd_device *dev, obd_count len, void *data);
1275 int (*o_detach)(struct obd_device *dev);
1276 int (*o_setup) (struct obd_device *dev, struct lustre_cfg *cfg);
1277 int (*o_precleanup)(struct obd_device *dev,
1278 enum obd_cleanup_stage cleanup_stage);
1279 int (*o_cleanup)(struct obd_device *dev);
1280 int (*o_process_config)(struct obd_device *dev, obd_count len,
1282 int (*o_postrecov)(struct obd_device *dev);
1283 int (*o_add_conn)(struct obd_import *imp, struct obd_uuid *uuid,
1285 int (*o_del_conn)(struct obd_import *imp, struct obd_uuid *uuid);
1286 /* connect to the target device with given connection
1287 * data. @ocd->ocd_connect_flags is modified to reflect flags actually
1288 * granted by the target, which are guaranteed to be a subset of flags
1289 * asked for. If @ocd == NULL, use default parameters. */
1290 int (*o_connect)(const struct lu_env *env,
1291 struct obd_export **exp, struct obd_device *src,
1292 struct obd_uuid *cluuid, struct obd_connect_data *ocd,
1294 int (*o_reconnect)(const struct lu_env *env,
1295 struct obd_export *exp, struct obd_device *src,
1296 struct obd_uuid *cluuid,
1297 struct obd_connect_data *ocd,
1299 int (*o_disconnect)(struct obd_export *exp);
1301 /* Initialize/finalize fids infrastructure. */
1302 int (*o_fid_init)(struct obd_export *exp);
1303 int (*o_fid_fini)(struct obd_export *exp);
1305 /* Allocate new fid according to passed @hint. */
1306 int (*o_fid_alloc)(struct obd_export *exp, struct lu_fid *fid,
1307 struct md_op_data *op_data);
1310 * Object with @fid is getting deleted, we may want to do something
1313 int (*o_fid_delete)(struct obd_export *exp, const struct lu_fid *fid);
1315 int (*o_statfs)(struct obd_device *obd, struct obd_statfs *osfs,
1316 __u64 max_age, __u32 flags);
1317 int (*o_statfs_async)(struct obd_device *obd, struct obd_info *oinfo,
1318 __u64 max_age, struct ptlrpc_request_set *set);
1319 int (*o_packmd)(struct obd_export *exp, struct lov_mds_md **disk_tgt,
1320 struct lov_stripe_md *mem_src);
1321 int (*o_unpackmd)(struct obd_export *exp,struct lov_stripe_md **mem_tgt,
1322 struct lov_mds_md *disk_src, int disk_len);
1323 int (*o_checkmd)(struct obd_export *exp, struct obd_export *md_exp,
1324 struct lov_stripe_md *mem_tgt);
1325 int (*o_preallocate)(struct lustre_handle *, obd_count *req,
1327 /* FIXME: add fid capability support for create & destroy! */
1328 int (*o_precreate)(struct obd_export *exp);
1329 int (*o_create)(struct obd_export *exp, struct obdo *oa,
1330 struct lov_stripe_md **ea, struct obd_trans_info *oti);
1331 int (*o_create_async)(struct obd_export *exp, struct obd_info *oinfo,
1332 struct lov_stripe_md **ea,
1333 struct obd_trans_info *oti);
1334 int (*o_destroy)(struct obd_export *exp, struct obdo *oa,
1335 struct lov_stripe_md *ea, struct obd_trans_info *oti,
1336 struct obd_export *md_exp, void *capa);
1337 int (*o_setattr)(struct obd_export *exp, struct obd_info *oinfo,
1338 struct obd_trans_info *oti);
1339 int (*o_setattr_async)(struct obd_export *exp, struct obd_info *oinfo,
1340 struct obd_trans_info *oti,
1341 struct ptlrpc_request_set *rqset);
1342 int (*o_getattr)(struct obd_export *exp, struct obd_info *oinfo);
1343 int (*o_getattr_async)(struct obd_export *exp, struct obd_info *oinfo,
1344 struct ptlrpc_request_set *set);
1345 int (*o_brw)(int rw, struct obd_export *exp, struct obd_info *oinfo,
1346 obd_count oa_bufs, struct brw_page *pgarr,
1347 struct obd_trans_info *oti);
1348 int (*o_merge_lvb)(struct obd_export *exp, struct lov_stripe_md *lsm,
1349 struct ost_lvb *lvb, int kms_only);
1350 int (*o_adjust_kms)(struct obd_export *exp, struct lov_stripe_md *lsm,
1351 obd_off size, int shrink);
1352 int (*o_punch)(struct obd_export *exp, struct obd_info *oinfo,
1353 struct obd_trans_info *oti,
1354 struct ptlrpc_request_set *rqset);
1355 int (*o_sync)(struct obd_export *exp, struct obdo *oa,
1356 struct lov_stripe_md *ea, obd_size start, obd_size end,
1358 int (*o_migrate)(struct lustre_handle *conn, struct lov_stripe_md *dst,
1359 struct lov_stripe_md *src, obd_size start,
1360 obd_size end, struct obd_trans_info *oti);
1361 int (*o_copy)(struct lustre_handle *dstconn, struct lov_stripe_md *dst,
1362 struct lustre_handle *srconn, struct lov_stripe_md *src,
1363 obd_size start, obd_size end, struct obd_trans_info *);
1364 int (*o_iterate)(struct lustre_handle *conn,
1365 int (*)(obd_id, obd_gr, void *),
1366 obd_id *startid, obd_gr group, void *data);
1367 int (*o_preprw)(int cmd, struct obd_export *exp, struct obdo *oa,
1368 int objcount, struct obd_ioobj *obj,
1369 struct niobuf_remote *remote, int *nr_pages,
1370 struct niobuf_local *local,
1371 struct obd_trans_info *oti,
1372 struct lustre_capa *capa);
1373 int (*o_commitrw)(int cmd, struct obd_export *exp, struct obdo *oa,
1374 int objcount, struct obd_ioobj *obj,
1375 struct niobuf_remote *remote, int pages,
1376 struct niobuf_local *local,
1377 struct obd_trans_info *oti, int rc);
1378 int (*o_enqueue)(struct obd_export *, struct obd_info *oinfo,
1379 struct ldlm_enqueue_info *einfo,
1380 struct ptlrpc_request_set *rqset);
1381 int (*o_change_cbdata)(struct obd_export *, struct lov_stripe_md *,
1382 ldlm_iterator_t it, void *data);
1383 int (*o_cancel)(struct obd_export *, struct lov_stripe_md *md,
1384 __u32 mode, struct lustre_handle *);
1385 int (*o_cancel_unused)(struct obd_export *, struct lov_stripe_md *,
1386 int flags, void *opaque);
1387 int (*o_init_export)(struct obd_export *exp);
1388 int (*o_destroy_export)(struct obd_export *exp);
1389 int (*o_extent_calc)(struct obd_export *, struct lov_stripe_md *,
1390 int cmd, obd_off *);
1392 /* llog related obd_methods */
1393 int (*o_llog_init)(struct obd_device *obd, struct obd_llog_group *grp,
1394 struct obd_device *disk_obd, int *idx);
1395 int (*o_llog_finish)(struct obd_device *obd, int count);
1396 int (*o_llog_connect)(struct obd_export *, struct llogd_conn_body *);
1398 /* metadata-only methods */
1399 int (*o_pin)(struct obd_export *, const struct lu_fid *fid,
1400 struct obd_capa *, struct obd_client_handle *, int flag);
1401 int (*o_unpin)(struct obd_export *, struct obd_client_handle *, int);
1403 int (*o_import_event)(struct obd_device *, struct obd_import *,
1404 enum obd_import_event);
1406 int (*o_notify)(struct obd_device *obd, struct obd_device *watched,
1407 enum obd_notify_event ev, void *data);
1409 int (*o_health_check)(struct obd_device *);
1410 struct obd_uuid *(*o_get_uuid) (struct obd_export *exp);
1413 int (*o_quotacheck)(struct obd_device *, struct obd_export *,
1414 struct obd_quotactl *);
1415 int (*o_quotactl)(struct obd_device *, struct obd_export *,
1416 struct obd_quotactl *);
1417 int (*o_quota_adjust_qunit)(struct obd_export *exp,
1418 struct quota_adjust_qunit *oqaq,
1419 struct lustre_quota_ctxt *qctxt);
1422 int (*o_ping)(struct obd_export *exp);
1425 int (*o_pool_new)(struct obd_device *obd, char *poolname);
1426 int (*o_pool_del)(struct obd_device *obd, char *poolname);
1427 int (*o_pool_add)(struct obd_device *obd, char *poolname,
1429 int (*o_pool_rem)(struct obd_device *obd, char *poolname,
1431 void (*o_getref)(struct obd_device *obd);
1432 void (*o_putref)(struct obd_device *obd);
1434 * NOTE: If adding ops, add another LPROCFS_OBD_OP_INIT() line
1435 * to lprocfs_alloc_obd_stats() in obdclass/lprocfs_status.c.
1436 * Also, add a wrapper function in include/linux/obd_class.h. */
1440 LUSTRE_OPC_MKDIR = (1 << 0),
1441 LUSTRE_OPC_SYMLINK = (1 << 1),
1442 LUSTRE_OPC_MKNOD = (1 << 2),
1443 LUSTRE_OPC_CREATE = (1 << 3),
1444 LUSTRE_OPC_ANY = (1 << 4)
1447 /* lmv structures */
1448 #define MEA_MAGIC_LAST_CHAR 0xb2221ca1
1449 #define MEA_MAGIC_ALL_CHARS 0xb222a11c
1450 #define MEA_MAGIC_HASH_SEGMENT 0xb222a11b
1452 #define MAX_HASH_SIZE_32 0x7fffffffUL
1453 #define MAX_HASH_SIZE 0x7fffffffffffffffULL
1454 #define MAX_HASH_HIGHEST_BIT 0x1000000000000000ULL
1457 struct mdt_body *body;
1458 struct lov_stripe_md *lsm;
1459 struct lmv_stripe_md *mea;
1460 #ifdef CONFIG_FS_POSIX_ACL
1461 struct posix_acl *posix_acl;
1463 struct mdt_remote_perm *remote_perm;
1464 struct obd_capa *mds_capa;
1465 struct obd_capa *oss_capa;
1468 struct md_open_data {
1469 struct obd_client_handle *mod_och;
1470 struct ptlrpc_request *mod_open_req;
1471 struct ptlrpc_request *mod_close_req;
1474 struct lookup_intent;
1477 int (*m_getstatus)(struct obd_export *, struct lu_fid *,
1478 struct obd_capa **);
1479 int (*m_change_cbdata)(struct obd_export *, const struct lu_fid *,
1480 ldlm_iterator_t, void *);
1481 int (*m_close)(struct obd_export *, struct md_op_data *,
1482 struct md_open_data *, struct ptlrpc_request **);
1483 int (*m_create)(struct obd_export *, struct md_op_data *,
1484 const void *, int, int, __u32, __u32, cfs_cap_t,
1485 __u64, struct ptlrpc_request **);
1486 int (*m_done_writing)(struct obd_export *, struct md_op_data *,
1487 struct md_open_data *);
1488 int (*m_enqueue)(struct obd_export *, struct ldlm_enqueue_info *,
1489 struct lookup_intent *, struct md_op_data *,
1490 struct lustre_handle *, void *, int,
1491 struct ptlrpc_request **, int);
1492 int (*m_getattr)(struct obd_export *, const struct lu_fid *,
1493 struct obd_capa *, obd_valid, int,
1494 struct ptlrpc_request **);
1495 int (*m_getattr_name)(struct obd_export *, const struct lu_fid *,
1496 struct obd_capa *, const char *, int, obd_valid,
1497 int, __u32, struct ptlrpc_request **);
1498 int (*m_intent_lock)(struct obd_export *, struct md_op_data *,
1499 void *, int, struct lookup_intent *, int,
1500 struct ptlrpc_request **,
1501 ldlm_blocking_callback, int);
1502 int (*m_link)(struct obd_export *, struct md_op_data *,
1503 struct ptlrpc_request **);
1504 int (*m_rename)(struct obd_export *, struct md_op_data *,
1505 const char *, int, const char *, int,
1506 struct ptlrpc_request **);
1507 int (*m_is_subdir)(struct obd_export *, const struct lu_fid *,
1508 const struct lu_fid *,
1509 struct ptlrpc_request **);
1510 int (*m_setattr)(struct obd_export *, struct md_op_data *, void *,
1511 int , void *, int, struct ptlrpc_request **,
1512 struct md_open_data **mod);
1513 int (*m_sync)(struct obd_export *, const struct lu_fid *,
1514 struct obd_capa *, struct ptlrpc_request **);
1515 int (*m_readpage)(struct obd_export *, const struct lu_fid *,
1516 struct obd_capa *, __u64, struct page *,
1517 struct ptlrpc_request **);
1519 int (*m_unlink)(struct obd_export *, struct md_op_data *,
1520 struct ptlrpc_request **);
1522 int (*m_setxattr)(struct obd_export *, const struct lu_fid *,
1523 struct obd_capa *, obd_valid, const char *,
1524 const char *, int, int, int, __u32,
1525 struct ptlrpc_request **);
1527 int (*m_getxattr)(struct obd_export *, const struct lu_fid *,
1528 struct obd_capa *, obd_valid, const char *,
1529 const char *, int, int, int,
1530 struct ptlrpc_request **);
1532 int (*m_init_ea_size)(struct obd_export *, int, int, int);
1534 int (*m_get_lustre_md)(struct obd_export *, struct ptlrpc_request *,
1535 struct obd_export *, struct obd_export *,
1536 struct lustre_md *);
1538 int (*m_free_lustre_md)(struct obd_export *, struct lustre_md *);
1540 int (*m_set_open_replay_data)(struct obd_export *,
1541 struct obd_client_handle *,
1542 struct ptlrpc_request *);
1543 int (*m_clear_open_replay_data)(struct obd_export *,
1544 struct obd_client_handle *);
1545 int (*m_set_lock_data)(struct obd_export *, __u64 *, void *, __u32 *);
1547 ldlm_mode_t (*m_lock_match)(struct obd_export *, int,
1548 const struct lu_fid *, ldlm_type_t,
1549 ldlm_policy_data_t *, ldlm_mode_t,
1550 struct lustre_handle *);
1552 int (*m_cancel_unused)(struct obd_export *, const struct lu_fid *,
1553 ldlm_policy_data_t *, ldlm_mode_t, int flags,
1555 int (*m_renew_capa)(struct obd_export *, struct obd_capa *oc,
1556 renew_capa_cb_t cb);
1557 int (*m_unpack_capa)(struct obd_export *, struct ptlrpc_request *,
1558 const struct req_msg_field *, struct obd_capa **);
1560 int (*m_get_remote_perm)(struct obd_export *, const struct lu_fid *,
1561 struct obd_capa *, __u32,
1562 struct ptlrpc_request **);
1564 int (*m_intent_getattr_async)(struct obd_export *,
1565 struct md_enqueue_info *,
1566 struct ldlm_enqueue_info *);
1568 int (*m_revalidate_lock)(struct obd_export *,
1569 struct lookup_intent *,
1573 * NOTE: If adding ops, add another LPROCFS_MD_OP_INIT() line to
1574 * lprocfs_alloc_md_stats() in obdclass/lprocfs_status.c. Also, add a
1575 * wrapper function in include/linux/obd_class.h.
1579 struct lsm_operations {
1580 void (*lsm_free)(struct lov_stripe_md *);
1581 int (*lsm_destroy)(struct lov_stripe_md *, struct obdo *oa,
1582 struct obd_export *md_exp);
1583 void (*lsm_stripe_by_index)(struct lov_stripe_md *, int *, obd_off *,
1585 void (*lsm_stripe_by_offset)(struct lov_stripe_md *, int *, obd_off *,
1587 obd_off (*lsm_stripe_offset_by_index)(struct lov_stripe_md *, int);
1588 obd_off (*lsm_stripe_offset_by_offset)(struct lov_stripe_md *, obd_off);
1589 int (*lsm_stripe_index_by_offset)(struct lov_stripe_md *, obd_off);
1590 int (*lsm_revalidate) (struct lov_stripe_md *, struct obd_device *obd);
1591 int (*lsm_lmm_verify) (struct lov_mds_md *lmm, int lmm_bytes,
1593 int (*lsm_unpackmd) (struct lov_obd *lov, struct lov_stripe_md *lsm,
1594 struct lov_mds_md *lmm);
1597 extern const struct lsm_operations lsm_v1_ops;
1598 extern const struct lsm_operations lsm_join_ops;
1599 extern const struct lsm_operations lsm_v3_ops;
1600 static inline const struct lsm_operations *lsm_op_find(int magic)
1605 case LOV_MAGIC_JOIN:
1606 return &lsm_join_ops;
1610 CERROR("Cannot recognize lsm_magic %d\n", magic);
1615 int lvfs_check_io_health(struct obd_device *obd, struct file *file);
1617 /* Requests for obd_extent_calc() */
1618 #define OBD_CALC_STRIPE_START 1
1619 #define OBD_CALC_STRIPE_END 2
1621 static inline void obd_transno_commit_cb(struct obd_device *obd, __u64 transno,
1622 struct obd_export *exp, int error)
1625 CERROR("%s: transno "LPU64" commit error: %d\n",
1626 obd->obd_name, transno, error);
1629 if (exp && transno > exp->exp_last_committed) {
1630 CDEBUG(D_HA, "%s: transno "LPU64" committed\n",
1631 obd->obd_name, transno);
1632 exp->exp_last_committed = transno;
1633 ptlrpc_commit_replies(exp);
1635 CDEBUG(D_INFO, "%s: transno "LPU64" committed\n",
1636 obd->obd_name, transno);
1638 if (transno > obd->obd_last_committed)
1639 obd->obd_last_committed = transno;
1642 static inline void init_obd_quota_ops(quota_interface_t *interface,
1643 struct obd_ops *obd_ops)
1649 obd_ops->o_quotacheck = QUOTA_OP(interface, check);
1650 obd_ops->o_quotactl = QUOTA_OP(interface, ctl);
1651 obd_ops->o_quota_adjust_qunit = QUOTA_OP(interface, adjust_qunit);
1654 static inline __u64 oinfo_mdsno(struct obd_info *oinfo)
1656 return obdo_mdsno(oinfo->oi_oa);
1659 static inline struct lustre_capa *oinfo_capa(struct obd_info *oinfo)
1661 return oinfo->oi_capa;
1664 #endif /* __OBD_H */