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>
64 #include <lustre_lib.h>
65 #include <lustre_export.h>
66 #include <lustre_quota.h>
67 #include <lustre_fld.h>
68 #include <lustre_capa.h>
69 #include <class_hash.h>
71 #include <libcfs/bitmap.h>
74 #define MAX_OBD_DEVICES 8192
76 /* this is really local to the OSC */
77 struct loi_oap_pages {
78 struct list_head lop_pending;
79 struct list_head lop_urgent;
80 struct list_head lop_pending_group;
90 struct lov_oinfo { /* per-stripe data structure */
91 __u64 loi_id; /* object ID on the target OST */
92 __u64 loi_gr; /* object group on the target OST */
93 int loi_ost_idx; /* OST stripe index in lov_tgt_desc->tgts */
94 int loi_ost_gen; /* generation of this loi_ost_idx */
96 /* used by the osc to keep track of what objects to build into rpcs */
97 struct loi_oap_pages loi_read_lop;
98 struct loi_oap_pages loi_write_lop;
99 /* _cli_ is poorly named, it should be _ready_ */
100 struct list_head loi_cli_item;
101 struct list_head loi_write_item;
102 struct list_head loi_read_item;
104 unsigned long loi_kms_valid:1;
105 __u64 loi_kms; /* known minimum size */
106 struct ost_lvb loi_lvb;
107 struct osc_async_rc loi_ar;
110 static inline void loi_kms_set(struct lov_oinfo *oinfo, __u64 kms)
112 oinfo->loi_kms = kms;
113 oinfo->loi_kms_valid = 1;
116 static inline void loi_init(struct lov_oinfo *loi)
118 CFS_INIT_LIST_HEAD(&loi->loi_read_lop.lop_pending);
119 CFS_INIT_LIST_HEAD(&loi->loi_read_lop.lop_urgent);
120 CFS_INIT_LIST_HEAD(&loi->loi_read_lop.lop_pending_group);
121 CFS_INIT_LIST_HEAD(&loi->loi_write_lop.lop_pending);
122 CFS_INIT_LIST_HEAD(&loi->loi_write_lop.lop_urgent);
123 CFS_INIT_LIST_HEAD(&loi->loi_write_lop.lop_pending_group);
124 CFS_INIT_LIST_HEAD(&loi->loi_cli_item);
125 CFS_INIT_LIST_HEAD(&loi->loi_write_item);
126 CFS_INIT_LIST_HEAD(&loi->loi_read_item);
129 /*extent array item for describing the joined file extent info*/
131 __u64 le_start; /* extent start */
132 __u64 le_len; /* extent length */
133 int le_loi_idx; /* extent #1 loi's index in lsm loi array */
134 int le_stripe_count; /* extent stripe count*/
137 /*Lov array info for describing joined file array EA info*/
138 struct lov_array_info {
139 struct llog_logid lai_array_id; /* MDS med llog object id */
140 unsigned lai_ext_count; /* number of extent count */
141 struct lov_extent *lai_ext_array; /* extent desc array */
144 struct lov_stripe_md {
146 pid_t lsm_lock_owner; /* debugging */
149 /* Public members. */
150 __u64 lw_object_id; /* lov object id */
151 __u64 lw_object_gr; /* lov object group */
152 __u64 lw_maxbytes; /* maximum possible file size */
154 /* LOV-private members start here -- only for use in lov/. */
156 __u32 lw_stripe_size; /* size of the stripe */
157 __u32 lw_pattern; /* striping pattern (RAID0, RAID1) */
158 unsigned lw_stripe_count; /* number of objects being striped over */
159 char lw_pool_name[LOV_MAXPOOLNAME]; /* pool name */
162 struct lov_array_info *lsm_array; /*Only for joined file array info*/
163 struct lov_oinfo *lsm_oinfo[0];
166 #define lsm_object_id lsm_wire.lw_object_id
167 #define lsm_object_gr lsm_wire.lw_object_gr
168 #define lsm_maxbytes lsm_wire.lw_maxbytes
169 #define lsm_magic lsm_wire.lw_magic
170 #define lsm_stripe_size lsm_wire.lw_stripe_size
171 #define lsm_pattern lsm_wire.lw_pattern
172 #define lsm_stripe_count lsm_wire.lw_stripe_count
173 #define lsm_pool_name lsm_wire.lw_pool_name
177 typedef int (*obd_enqueue_update_f)(struct obd_info *oinfo, int rc);
179 /* obd info for a particular level (lov, osc). */
181 /* Lock policy. It keeps an extent which is specific for a particular
182 * OSC. (e.g. lov_prep_enqueue_set initialises extent of the policy,
183 * and osc_enqueue passes it into ldlm_lock_match & ldlm_cli_enqueue. */
184 ldlm_policy_data_t oi_policy;
185 /* Flags used for set request specific flags:
186 - while lock handling, the flags obtained on the enqueue
187 request are set here.
188 - while stats, the flags used for control delay/resend.
191 /* Lock handle specific for every OSC lock. */
192 struct lustre_handle *oi_lockh;
193 /* lsm data specific for every OSC. */
194 struct lov_stripe_md *oi_md;
195 /* obdo data specific for every OSC, if needed at all. */
197 /* statfs data specific for every OSC, if needed at all. */
198 struct obd_statfs *oi_osfs;
199 /* An update callback which is called to update some data on upper
200 * level. E.g. it is used for update lsm->lsm_oinfo at every recieved
201 * request in osc level for enqueue requests. It is also possible to
202 * update some caller data from LOV layer if needed. */
203 obd_enqueue_update_f oi_cb_up;
204 /* oss capability, its type is obd_capa in client to avoid copy.
205 * in contrary its type is lustre_capa in OSS. */
209 /* compare all relevant fields. */
210 static inline int lov_stripe_md_cmp(struct lov_stripe_md *m1,
211 struct lov_stripe_md *m2)
214 * ->lsm_wire contains padding, but it should be zeroed out during
217 return memcmp(&m1->lsm_wire, &m2->lsm_wire, sizeof m1->lsm_wire);
220 void lov_stripe_lock(struct lov_stripe_md *md);
221 void lov_stripe_unlock(struct lov_stripe_md *md);
224 struct list_head typ_chain;
225 struct obd_ops *typ_dt_ops;
226 struct md_ops *typ_md_ops;
227 cfs_proc_dir_entry_t *typ_procroot;
230 struct lu_device_type *typ_lu;
231 spinlock_t obd_type_lock;
242 ASYNC_READY = 0x1, /* ap_make_ready will not be called before this
243 page is added to an rpc */
244 ASYNC_URGENT = 0x2, /* page must be put into an RPC before return */
245 ASYNC_COUNT_STABLE = 0x4, /* ap_refresh_count will not be called
246 to give the caller a chance to update
247 or cancel the size of the io */
248 ASYNC_GROUP_SYNC = 0x8, /* ap_completion will not be called, instead
249 the page is accounted for in the
250 obd_io_group given to
251 obd_queue_group_io */
254 struct obd_async_page_ops {
255 int (*ap_make_ready)(void *data, int cmd);
256 int (*ap_refresh_count)(void *data, int cmd);
257 void (*ap_fill_obdo)(void *data, int cmd, struct obdo *oa);
258 void (*ap_update_obdo)(void *data, int cmd, struct obdo *oa,
260 int (*ap_completion)(void *data, int cmd, struct obdo *oa, int rc);
261 struct obd_capa *(*ap_lookup_capa)(void *data, int cmd);
264 /* the `oig' is passed down from a caller of obd rw methods. the callee
265 * records enough state such that the caller can sleep on the oig and
266 * be woken when all the callees have finished their work */
267 struct obd_io_group {
269 atomic_t oig_refcount;
272 struct list_head oig_occ_list;
273 cfs_waitq_t oig_waitq;
276 /* the oig callback context lets the callee of obd rw methods register
277 * for callbacks from the caller. */
278 struct oig_callback_context {
279 struct list_head occ_oig_item;
280 /* called when the caller has received a signal while sleeping.
281 * callees of this method are encouraged to abort their state
282 * in the oig. This may be called multiple times. */
283 void (*occ_interrupted)(struct oig_callback_context *occ);
284 unsigned long interrupted:1;
287 /* Individual type definitions */
289 struct ost_server_data;
291 /* hold common fields for "target" device */
292 struct obd_device_target {
293 struct super_block *obt_sb;
294 atomic_t obt_quotachecking;
295 struct lustre_quota_ctxt obt_qctxt;
298 typedef void (*obd_pin_extent_cb)(void *data);
299 typedef int (*obd_page_removal_cb_t)(void *data, int discard);
300 typedef int (*obd_lock_cancel_cb)(struct ldlm_lock *,struct ldlm_lock_desc *,
305 LLOG_CONFIG_ORIG_CTXT = 0,
306 LLOG_CONFIG_REPL_CTXT,
307 LLOG_MDS_OST_ORIG_CTXT,
308 LLOG_MDS_OST_REPL_CTXT,
315 LLOG_LOVEA_ORIG_CTXT,
316 LLOG_LOVEA_REPL_CTXT,
320 #define FILTER_SUBDIR_COUNT 32 /* set to zero for no subdirs */
322 struct filter_subdirs {
323 cfs_dentry_t *dentry[FILTER_SUBDIR_COUNT];
333 /* NB this field MUST be first */
334 struct obd_device_target fo_obt;
335 const char *fo_fstype;
336 struct vfsmount *fo_vfsmnt;
339 cfs_dentry_t *fo_dentry_O;
340 cfs_dentry_t **fo_dentry_O_groups;
341 struct filter_subdirs *fo_dentry_O_sub;
342 struct semaphore fo_init_lock; /* group initialization lock */
343 int fo_committed_group;
346 spinlock_t fo_objidlock; /* protect fo_lastobjid */
347 spinlock_t fo_translock; /* protect fsd_last_transno */
348 struct file *fo_rcvd_filp;
349 struct file *fo_health_check_filp;
350 struct lr_server_data *fo_fsd;
351 unsigned long *fo_last_rcvd_slots;
352 __u64 fo_mount_count;
354 unsigned long fo_destroys_in_progress;
355 struct semaphore fo_create_locks[FILTER_SUBDIR_COUNT];
357 struct list_head fo_export_list;
360 obd_size fo_tot_dirty; /* protected by obd_osfs_lock */
361 obd_size fo_tot_granted; /* all values in bytes */
362 obd_size fo_tot_pending;
364 obd_size fo_readcache_max_filesize;
366 int fo_writethrough_cache;
368 struct obd_import *fo_mdc_imp;
369 struct obd_uuid fo_mdc_uuid;
370 struct lustre_handle fo_mdc_conn;
371 struct file **fo_last_objid_files;
372 __u64 *fo_last_objids; /* last created objid for groups,
373 * protected by fo_objidlock */
375 struct semaphore fo_alloc_lock;
377 atomic_t fo_r_in_flight;
378 atomic_t fo_w_in_flight;
381 * per-filter pool of kiobuf's allocated by filter_common_setup() and
382 * torn down by filter_cleanup(). Contains OST_NUM_THREADS elements of
383 * which ->fo_iobuf_count were allocated.
385 * This pool contains kiobuf used by
386 * filter_{prep,commit}rw_{read,write}() and is shared by all OST
389 * Locking: none, each OST thread uses only one element, determined by
390 * its "ordinal number", ->t_id.
392 struct filter_iobuf **fo_iobuf_pool;
395 struct list_head fo_llog_list;
396 spinlock_t fo_llog_list_lock;
398 struct brw_stats fo_filter_stats;
399 struct lustre_quota_ctxt fo_quota_ctxt;
400 spinlock_t fo_quotacheck_lock;
401 atomic_t fo_quotachecking;
403 int fo_fmd_max_num; /* per exp filter_mod_data */
404 int fo_fmd_max_age; /* jiffies to fmd expiry */
407 rwlock_t fo_sptlrpc_lock;
408 struct sptlrpc_rule_set fo_sptlrpc_rset;
410 /* capability related */
411 unsigned int fo_fl_oss_capa;
412 struct list_head fo_capa_keys;
413 struct hlist_head *fo_capa_hash;
414 struct llog_commit_master *fo_lcm;
417 #define OSC_MAX_RIF_DEFAULT 8
418 #define OSC_MAX_RIF_MAX 256
419 #define OSC_MAX_DIRTY_DEFAULT (OSC_MAX_RIF_DEFAULT * 4)
420 #define OSC_MAX_DIRTY_MB_MAX 2048 /* arbitrary, but < MAX_LONG bytes */
421 #define OSC_DEFAULT_RESENDS 10
423 #define MDC_MAX_RIF_DEFAULT 8
424 #define MDC_MAX_RIF_MAX 512
430 struct rw_semaphore cl_sem;
431 struct obd_uuid cl_target_uuid;
432 struct obd_import *cl_import; /* ptlrpc connection state */
434 /* max_mds_easize is purely a performance thing so we don't have to
435 * call obd_size_diskmd() all the time. */
436 int cl_default_mds_easize;
437 int cl_max_mds_easize;
438 int cl_max_mds_cookiesize;
440 /* security configuration */
441 struct sptlrpc_rule_set cl_sptlrpc_rset;
442 enum lustre_sec_part cl_sec_part;
444 //struct llog_canceld_ctxt *cl_llcd; /* it's included by obd_llog_ctxt */
445 void *cl_llcd_offset;
447 /* the grant values are protected by loi_list_lock below */
448 long cl_dirty; /* all _dirty_ in bytes */
449 long cl_dirty_max; /* allowed w/o rpc */
450 long cl_avail_grant; /* bytes of credit for ost */
451 long cl_lost_grant; /* lost credits (trunc) */
452 struct list_head cl_cache_waiters; /* waiting for cache/grant */
454 /* keep track of objects that have lois that contain pages which
455 * have been queued for async brw. this lock also protects the
456 * lists of osc_client_pages that hang off of the loi */
458 * ->cl_loi_list_lock protects consistency of
459 * ->cl_loi_{ready,read,write}_list. ->ap_make_ready() and
460 * ->ap_completion() call-backs are executed under this lock. As we
461 * cannot guarantee that these call-backs never block on all platforms
462 * (as a matter of fact they do block on Mac OS X), type of
463 * ->cl_loi_list_lock is platform dependent: it's a spin-lock on Linux
464 * and blocking mutex on Mac OS X. (Alternative is to make this lock
465 * blocking everywhere, but we don't want to slow down fast-path of
466 * our main platform.)
468 * Exact type of ->cl_loi_list_lock is defined in arch/obd.h together
469 * with client_obd_list_{un,}lock() and
470 * client_obd_list_lock_{init,done}() functions.
472 client_obd_lock_t cl_loi_list_lock;
473 struct list_head cl_loi_ready_list;
474 struct list_head cl_loi_write_list;
475 struct list_head cl_loi_read_list;
478 /* just a sum of the loi/lop pending numbers to be exported by /proc */
479 int cl_pending_w_pages;
480 int cl_pending_r_pages;
481 int cl_max_pages_per_rpc;
482 int cl_max_rpcs_in_flight;
483 struct obd_histogram cl_read_rpc_hist;
484 struct obd_histogram cl_write_rpc_hist;
485 struct obd_histogram cl_read_page_hist;
486 struct obd_histogram cl_write_page_hist;
487 struct obd_histogram cl_read_offset_hist;
488 struct obd_histogram cl_write_offset_hist;
490 /* number of in flight destroy rpcs is limited to max_rpcs_in_flight */
491 atomic_t cl_destroy_in_flight;
492 cfs_waitq_t cl_destroy_waitq;
494 struct mdc_rpc_lock *cl_rpc_lock;
495 struct mdc_rpc_lock *cl_setattr_lock;
496 struct mdc_rpc_lock *cl_close_lock;
497 struct osc_creator cl_oscc;
500 struct semaphore cl_mgc_sem;
501 struct vfsmount *cl_mgc_vfsmnt;
502 struct dentry *cl_mgc_configs_dir;
503 atomic_t cl_mgc_refcount;
504 struct obd_export *cl_mgc_mgsexp;
506 /* checksumming for data sent over the network */
507 unsigned int cl_checksum:1; /* 0 = disabled, 1 = enabled */
508 /* supported checksum types that are worked out at connect time */
509 __u32 cl_supp_cksum_types;
510 /* checksum algorithm to be used */
511 cksum_type_t cl_cksum_type;
513 /* also protected by the poorly named _loi_list_lock lock above */
514 struct osc_async_rc cl_ar;
516 /* used by quotacheck */
517 int cl_qchk_stat; /* quotacheck stat of the peer */
519 /* sequence manager */
520 struct lu_client_seq *cl_seq;
522 atomic_t cl_resends; /* resend count */
524 /* Cache of triples */
525 struct lustre_cache *cl_cache;
526 obd_lock_cancel_cb cl_ext_lock_cancel_cb;
528 #define obd2cli_tgt(obd) ((char *)(obd)->u.cli.cl_target_uuid.uuid)
530 #define CL_NOT_QUOTACHECKED 1 /* client->cl_qchk_stat init value */
533 struct ptlrpc_service *mgs_service;
534 struct vfsmount *mgs_vfsmnt;
535 struct super_block *mgs_sb;
536 struct dentry *mgs_configs_dir;
537 struct dentry *mgs_fid_de;
538 struct list_head mgs_fs_db_list;
539 struct semaphore mgs_sem;
540 cfs_proc_dir_entry_t *mgs_proc_live;
544 /* NB this field MUST be first */
545 struct obd_device_target mds_obt;
546 struct ptlrpc_service *mds_service;
547 struct ptlrpc_service *mds_setattr_service;
548 struct ptlrpc_service *mds_readpage_service;
549 struct vfsmount *mds_vfsmnt;
550 cfs_dentry_t *mds_fid_de;
552 int mds_max_cookiesize;
553 struct file *mds_rcvd_filp;
554 spinlock_t mds_transno_lock;
555 __u64 mds_last_transno;
556 __u64 mds_mount_count;
558 unsigned long mds_atime_diff;
559 struct semaphore mds_epoch_sem;
560 struct ll_fid mds_rootfid;
561 struct lr_server_data *mds_server_data;
562 cfs_dentry_t *mds_pending_dir;
563 cfs_dentry_t *mds_logs_dir;
564 cfs_dentry_t *mds_objects_dir;
565 struct llog_handle *mds_cfg_llh;
566 struct obd_device *mds_osc_obd; /* XXX lov_obd */
567 struct obd_uuid mds_lov_uuid;
569 struct obd_export *mds_osc_exp; /* XXX lov_exp */
570 struct lov_desc mds_lov_desc;
573 /* mark pages dirty for write. */
574 bitmap_t *mds_lov_page_dirty;
575 /* array for store pages with obd_id */
576 void **mds_lov_page_array;
577 /* file for store objid */
578 struct file *mds_lov_objid_filp;
579 __u32 mds_lov_objid_count;
580 __u32 mds_lov_objid_lastpage;
581 __u32 mds_lov_objid_lastidx;
583 struct file *mds_health_check_filp;
584 unsigned long *mds_client_bitmap;
585 // struct upcall_cache *mds_group_hash;
587 struct lustre_quota_info mds_quota_info;
588 struct semaphore mds_qonoff_sem;
589 struct semaphore mds_health_sem;
590 unsigned long mds_fl_user_xattr:1,
592 mds_evict_ost_nids:1,
596 struct upcall_cache *mds_identity_cache;
598 /* for capability keys update */
599 struct lustre_capa_key *mds_capa_keys;
600 struct rw_semaphore mds_notify_lock;
604 extern __u32 mds_max_ost_index;
606 #define MDS_LOV_ALLOC_SIZE (CFS_PAGE_SIZE)
608 #define OBJID_PER_PAGE() (MDS_LOV_ALLOC_SIZE / sizeof(obd_id))
610 #define MDS_LOV_OBJID_PAGES_COUNT (mds_max_ost_index/OBJID_PER_PAGE())
612 extern int mds_lov_init_objids(struct obd_device *obd);
613 extern void mds_lov_destroy_objids(struct obd_device *obd);
626 struct lustre_handle eo_nl_lock;
631 struct ptlrpc_service *ost_service;
632 struct ptlrpc_service *ost_create_service;
633 struct ptlrpc_service *ost_io_service;
634 struct semaphore ost_health_sem;
637 struct echo_client_obd {
638 struct obd_export *ec_exp; /* the local connection to osc/lov */
640 struct list_head ec_objects;
646 struct obd_uuid lqo_uuid; /* ptlrpc's c_remote_uuid */
647 struct list_head lqo_oss_list; /* link to lov_qos */
648 __u32 lqo_ost_count; /* number of osts on this oss */
649 __u64 lqo_bavail; /* total bytes avail on OSS */
650 __u64 lqo_penalty; /* current penalty */
651 __u64 lqo_penalty_per_obj; /* penalty decrease every obj*/
655 struct lov_qos_oss *ltq_oss; /* oss info */
656 __u64 ltq_penalty; /* current penalty */
657 __u64 ltq_penalty_per_obj; /* penalty decrease every obj*/
658 __u64 ltq_weight; /* net weighting */
659 unsigned int ltq_usable:1; /* usable for striping */
662 /* Generic subset of OSTs */
664 __u32 *op_array; /* array of index of
666 unsigned int op_count; /* number of OSTs in the array */
667 unsigned int op_size; /* allocated size of lp_array */
668 rwlock_t op_rwlock; /* to protect lov_pool use */
671 /* Round-robin allocator data */
673 __u32 lqr_start_idx; /* start index of new inode */
674 __u32 lqr_offset_idx; /* aliasing for start_idx */
675 int lqr_start_count; /* reseed counter */
676 struct ost_pool lqr_pool; /* round-robin optimized list */
677 unsigned long lqr_dirty:1; /* recalc round-robin list */
680 /* Stripe placement optimization */
682 struct list_head lq_oss_list; /* list of OSSs that targets use */
683 struct rw_semaphore lq_rw_sem;
684 __u32 lq_active_oss_count;
685 unsigned int lq_prio_free; /* priority for free space */
686 struct lov_qos_rr lq_rr; /* round robin qos data */
687 unsigned long lq_dirty:1, /* recalc qos data */
688 lq_same_space:1,/* the ost's all have approx.
689 the same space avail */
690 lq_reset:1; /* zero current penalties */
693 struct lov_tgt_desc {
694 struct obd_uuid ltd_uuid;
695 struct obd_export *ltd_exp;
696 struct ltd_qos ltd_qos; /* qos info per target */
698 __u32 ltd_index; /* index in lov_obd->tgts */
699 unsigned long ltd_active:1,/* is this target up for requests */
700 ltd_activate:1,/* should this target be activated */
701 ltd_reap:1; /* should this target be deleted */
705 #define pool_tgt_size(_p) _p->pool_obds.op_size
706 #define pool_tgt_count(_p) _p->pool_obds.op_count
707 #define pool_tgt_array(_p) _p->pool_obds.op_array
708 #define pool_tgt_rwlock(_p) _p->pool_obds.op_rwlock
709 #define pool_tgt(_p, _i) _p->pool_lov->lov_tgts[_p->pool_obds.op_array[_i]]
712 char pool_name[LOV_MAXPOOLNAME + 1]; /* name of pool */
713 struct ost_pool pool_obds; /* pool members */
714 struct lov_qos_rr pool_rr; /* round robin qos */
715 struct hlist_node pool_hash; /* access by poolname */
716 struct list_head pool_list; /* serial access */
717 cfs_proc_dir_entry_t *pool_proc_entry; /* file in /proc */
718 struct lov_obd *pool_lov; /* lov obd to which this
723 struct lov_desc desc;
724 struct lov_tgt_desc **lov_tgts; /* sparse array */
725 struct ost_pool lov_packed; /* all OSTs in a packed
727 struct semaphore lov_lock;
728 struct obd_connect_data lov_ocd;
729 struct lov_qos lov_qos; /* qos info per lov */
730 atomic_t lov_refcount;
731 __u32 lov_tgt_count; /* how many OBD's */
732 __u32 lov_active_tgt_count; /* how many active */
733 __u32 lov_death_row;/* tgts scheduled to be deleted */
734 __u32 lov_tgt_size; /* size of tgts array */
736 obd_page_removal_cb_t lov_page_removal_cb;
737 obd_pin_extent_cb lov_page_pin_cb;
738 obd_lock_cancel_cb lov_lock_cancel_cb;
740 lustre_hash_t *lov_pools_hash_body; /* used for key access */
741 struct list_head lov_pool_list; /* used for sequential access */
742 cfs_proc_dir_entry_t *lov_pool_proc_entry;
745 struct lmv_tgt_desc {
746 struct obd_uuid ltd_uuid;
747 struct obd_export *ltd_exp;
748 int ltd_active; /* is this target up for requests */
750 struct semaphore ltd_fid_sem;
753 enum placement_policy {
754 PLACEMENT_CHAR_POLICY = 0,
755 PLACEMENT_NID_POLICY = 1,
756 PLACEMENT_INVAL_POLICY = 2,
760 typedef enum placement_policy placement_policy_t;
764 struct lu_client_fld lmv_fld;
766 placement_policy_t lmv_placement;
767 struct lmv_desc desc;
768 struct obd_uuid cluuid;
769 struct obd_export *exp;
776 struct semaphore init_sem;
778 struct lmv_tgt_desc *tgts;
781 struct obd_connect_data *datas;
784 struct obd_connect_data conn_data;
787 struct niobuf_local {
792 cfs_dentry_t *dentry;
797 #define LUSTRE_FLD_NAME "fld"
798 #define LUSTRE_SEQ_NAME "seq"
800 #define LUSTRE_CMM_NAME "cmm"
801 #define LUSTRE_MDD_NAME "mdd"
802 #define LUSTRE_OSD_NAME "osd"
803 #define LUSTRE_LMV_NAME "lmv"
804 #define LUSTRE_CMM_MDC_NAME "cmm-mdc"
806 /* obd device type names */
807 /* FIXME all the references to LUSTRE_MDS_NAME should be swapped with LUSTRE_MDT_NAME */
808 #define LUSTRE_MDS_NAME "mds"
809 #define LUSTRE_MDT_NAME "mdt"
810 #define LUSTRE_MDC_NAME "mdc"
811 #define LUSTRE_OSS_NAME "ost" /* FIXME change name to oss */
812 #define LUSTRE_OST_NAME "obdfilter" /* FIXME change name to ost */
813 #define LUSTRE_OSC_NAME "osc"
814 #define LUSTRE_LOV_NAME "lov"
815 #define LUSTRE_MGS_NAME "mgs"
816 #define LUSTRE_MGC_NAME "mgc"
818 #define LUSTRE_CACHEOBD_NAME "cobd"
819 #define LUSTRE_ECHO_NAME "obdecho"
820 #define LUSTRE_ECHO_CLIENT_NAME "echo_client"
822 /* Constant obd names (post-rename) */
823 #define LUSTRE_MDS_OBDNAME "MDS"
824 #define LUSTRE_OSS_OBDNAME "OSS"
825 #define LUSTRE_MGS_OBDNAME "MGS"
826 #define LUSTRE_MGC_OBDNAME "MGC"
828 /* Don't conflict with on-wire flags OBD_BRW_WRITE, etc */
829 #define N_LOCAL_TEMP_PAGE 0x10000000
831 struct obd_trans_info {
834 /* Only used on the server side for tracking acks. */
835 struct oti_req_ack_lock {
836 struct lustre_handle lock;
840 struct llog_cookie oti_onecookie;
841 struct llog_cookie *oti_logcookies;
844 /* initial thread handling transaction */
848 struct obd_uuid *oti_ost_uuid;
851 static inline void oti_init(struct obd_trans_info *oti,
852 struct ptlrpc_request *req)
856 memset(oti, 0, sizeof(*oti));
861 oti->oti_xid = req->rq_xid;
863 if (req->rq_repmsg != NULL)
864 oti->oti_transno = lustre_msg_get_transno(req->rq_repmsg);
865 oti->oti_thread_id = req->rq_svc_thread ? req->rq_svc_thread->t_id : -1;
866 if (req->rq_reqmsg != NULL)
867 oti->oti_conn_cnt = lustre_msg_get_conn_cnt(req->rq_reqmsg);
870 static inline void oti_alloc_cookies(struct obd_trans_info *oti,int num_cookies)
875 if (num_cookies == 1)
876 oti->oti_logcookies = &oti->oti_onecookie;
878 OBD_ALLOC(oti->oti_logcookies,
879 num_cookies * sizeof(oti->oti_onecookie));
881 oti->oti_numcookies = num_cookies;
884 static inline void oti_free_cookies(struct obd_trans_info *oti)
886 if (!oti || !oti->oti_logcookies)
889 if (oti->oti_logcookies == &oti->oti_onecookie)
890 LASSERT(oti->oti_numcookies == 1);
892 OBD_FREE(oti->oti_logcookies,
893 oti->oti_numcookies * sizeof(oti->oti_onecookie));
894 oti->oti_logcookies = NULL;
895 oti->oti_numcookies = 0;
899 * Events signalled through obd_notify() upcall-chain.
901 enum obd_notify_event {
902 /* Device activated */
904 /* Device deactivated */
906 /* Device disconnected */
908 /* Connect data for import were changed */
911 OBD_NOTIFY_SYNC_NONBLOCK,
913 /* Configuration event */
917 /* bit-mask flags for config events */
919 CONFIG_LOG = 0x1, /* finished processing config log */
920 CONFIG_SYNC = 0x2 /* mdt synced 1 ost */
924 * Data structure used to pass obd_notify()-event to non-obd listeners (llite
925 * and liblustre being main examples).
927 struct obd_notify_upcall {
928 int (*onu_upcall)(struct obd_device *host, struct obd_device *watched,
929 enum obd_notify_event ev, void *owner);
930 /* Opaque datum supplied by upper layer listener */
934 struct target_recovery_data {
935 svc_handler_t trd_recovery_handler;
936 pid_t trd_processing_task;
937 struct completion trd_starting;
938 struct completion trd_finishing;
942 FILTER_GROUP_LLOG = 1,
947 struct obd_llog_group {
948 struct list_head olg_list;
950 struct llog_ctxt *olg_ctxts[LLOG_MAX_CTXTS];
951 cfs_waitq_t olg_waitq;
953 struct obd_export *olg_exp;
954 int olg_initializing;
955 struct semaphore olg_cat_processing;
958 /* corresponds to one of the obd's */
959 #define MAX_OBD_NAME 128
960 #define OBD_DEVICE_MAGIC 0XAB5CD6EF
962 struct obd_type *obd_type;
965 /* common and UUID name of this device */
966 char obd_name[MAX_OBD_NAME];
967 struct obd_uuid obd_uuid;
969 struct lu_device *obd_lu_dev;
972 unsigned long obd_attached:1, /* finished attach */
973 obd_set_up:1, /* finished setup */
974 obd_recovering:1, /* there are recoverable clients */
975 obd_abort_recovery:1,/* somebody ioctl'ed us to abort */
976 obd_replayable:1, /* recovery is enabled; inform clients */
977 obd_no_transno:1, /* no committed-transno notification */
978 obd_no_recov:1, /* fail instead of retry messages */
979 obd_req_replaying:1, /* replaying requests */
980 obd_stopping:1, /* started cleanup */
981 obd_starting:1, /* started setup */
982 obd_force:1, /* cleanup with > 0 obd refcount */
983 obd_fail:1, /* cleanup with failover */
984 obd_async_recov:1, /* allow asyncronous orphan cleanup */
985 obd_no_conn:1, /* deny new connections */
986 obd_inactive:1; /* device active/inactive
987 * (for /proc/status only!!) */
988 /* uuid-export hash body */
989 struct lustre_hash *obd_uuid_hash;
990 /* nid-export hash body */
991 struct lustre_hash *obd_nid_hash;
993 struct lustre_hash *obd_nid_stats_hash;
994 struct list_head obd_nid_stats;
995 atomic_t obd_refcount;
996 cfs_waitq_t obd_refcount_waitq;
997 struct list_head obd_exports;
999 spinlock_t obd_nid_lock;
1000 struct ldlm_namespace *obd_namespace;
1001 struct ptlrpc_client obd_ldlm_client; /* XXX OST/MDS only */
1002 /* a spinlock is OK for what we do now, may need a semaphore later */
1003 spinlock_t obd_dev_lock;
1004 struct semaphore obd_dev_sem;
1005 __u64 obd_last_committed;
1006 struct fsfilt_operations *obd_fsops;
1007 spinlock_t obd_osfs_lock;
1008 struct obd_statfs obd_osfs; /* locked by obd_osfs_lock */
1010 struct lvfs_run_ctxt obd_lvfs_ctxt;
1011 struct obd_llog_group obd_olg; /* default llog group */
1012 struct obd_device *obd_observer;
1013 struct obd_notify_upcall obd_upcall;
1014 struct obd_export *obd_self_export;
1015 /* list of exports in LRU order, for ping evictor, with obd_dev_lock */
1016 struct list_head obd_exports_timed;
1017 time_t obd_eviction_timer; /* for ping evictor */
1019 /* XXX encapsulate all this recovery data into one struct */
1020 svc_handler_t obd_recovery_handler;
1021 pid_t obd_processing_task;
1023 int obd_max_recoverable_clients;
1024 int obd_connected_clients;
1025 int obd_recoverable_clients;
1026 spinlock_t obd_processing_task_lock; /* BH lock (timer) */
1027 __u64 obd_next_recovery_transno;
1028 int obd_replayed_requests;
1029 int obd_requests_queued_for_recovery;
1030 cfs_waitq_t obd_next_transno_waitq;
1031 struct list_head obd_uncommitted_replies;
1032 spinlock_t obd_uncommitted_replies_lock;
1033 cfs_timer_t obd_recovery_timer;
1034 time_t obd_recovery_start; /* seconds */
1035 time_t obd_recovery_end; /* seconds, for lprocfs_status */
1036 time_t obd_recovery_max_time; /* seconds, bz13079 */
1037 int obd_recovery_timeout;
1039 /* new recovery stuff from CMD2 */
1040 struct target_recovery_data obd_recovery_data;
1041 int obd_replayed_locks;
1042 atomic_t obd_req_replay_clients;
1043 atomic_t obd_lock_replay_clients;
1044 struct list_head obd_req_replay_queue;
1045 struct list_head obd_lock_replay_queue;
1046 struct list_head obd_final_req_queue;
1047 int obd_recovery_stage;
1050 struct obd_device_target obt;
1051 struct filter_obd filter;
1053 struct client_obd cli;
1055 struct echo_client_obd echo_client;
1056 struct echo_obd echo;
1061 /* Fields used by LProcFS */
1062 unsigned int obd_cntr_base;
1063 struct lprocfs_stats *obd_stats;
1065 unsigned int md_cntr_base;
1066 struct lprocfs_stats *md_stats;
1068 cfs_proc_dir_entry_t *obd_proc_entry;
1069 cfs_proc_dir_entry_t *obd_proc_exports_entry;
1070 cfs_proc_dir_entry_t *obd_svc_procroot;
1071 struct lprocfs_stats *obd_svc_stats;
1072 atomic_t obd_evict_inprogress;
1073 cfs_waitq_t obd_evict_inprogress_waitq;
1076 * Ldlm pool part. Save last calculated SLV and Limit.
1078 rwlock_t obd_pool_lock;
1083 #define OBD_OPT_FORCE 0x0001
1084 #define OBD_OPT_FAILOVER 0x0002
1086 #define OBD_LLOG_FL_SENDNOW 0x0001
1088 enum obd_cleanup_stage {
1089 /* Special case hack for MDS LOVs */
1091 /* can be directly mapped to .ldto_device_fini() */
1092 OBD_CLEANUP_EXPORTS,
1095 /* get/set_info keys */
1096 #define KEY_READ_ONLY "read-only"
1097 #define KEY_MDS_CONN "mds_conn"
1098 #define KEY_NEXT_ID "next_id"
1099 #define KEY_LOVDESC "lovdesc"
1100 #define KEY_INIT_RECOV "initial_recov"
1101 #define KEY_INIT_RECOV_BACKUP "init_recov_bk"
1102 #define KEY_FLUSH_CTX "flush_ctx"
1103 #define KEY_CAPA_KEY "capa_key"
1104 #define KEY_CONN_DATA "conn_data"
1105 #define KEY_MAX_EASIZE "max_easize"
1106 #define KEY_REVIMP_UPD "revimp_update"
1107 #define KEY_LOV_IDX "lov_idx"
1108 #define KEY_LAST_ID "last_id"
1109 #define KEY_READONLY "read-only"
1110 #define KEY_LOCK_TO_STRIPE "lock_to_stripe"
1111 #define KEY_CHECKSUM "checksum"
1112 #define KEY_UNLINKED "unlinked"
1113 #define KEY_EVICT_BY_NID "evict_by_nid"
1114 #define KEY_REGISTER_TARGET "register_target"
1115 #define KEY_SET_FS "set_fs"
1116 #define KEY_CLEAR_FS "clear_fs"
1117 #define KEY_BLOCKSIZE "blocksize"
1118 #define KEY_BLOCKSIZE_BITS "blocksize_bits"
1119 #define KEY_FIEMAP "FIEMAP"
1121 #define KEY_INTERMDS "inter_mds"
1122 #define KEY_ASYNC "async"
1126 static inline int it_to_lock_mode(struct lookup_intent *it)
1128 /* CREAT needs to be tested before open (both could be set) */
1129 if (it->it_op & IT_CREAT)
1131 else if (it->it_op & (IT_READDIR | IT_GETATTR | IT_OPEN | IT_LOOKUP))
1134 LASSERTF(0, "Invalid it_op: %d\n", it->it_op);
1139 struct lu_fid op_fid1; /* operation fid1 (usualy parent) */
1140 struct lu_fid op_fid2; /* operation fid2 (usualy child) */
1141 struct lu_fid op_fid3; /* 2 extra fids to find conflicting */
1142 struct lu_fid op_fid4; /* to the operation locks. */
1143 mdsno_t op_mds; /* what mds server open will go to */
1144 struct lustre_handle op_handle;
1146 const char *op_name;
1149 struct lmv_stripe_md *op_mea1;
1150 struct lmv_stripe_md *op_mea2;
1151 __u32 op_suppgids[2];
1157 /* iattr fields and blocks. */
1158 struct iattr op_attr;
1160 #if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,14)
1161 unsigned int op_attr_flags;
1164 loff_t op_attr_blocks;
1166 /* Size-on-MDS epoch and flags. */
1171 struct obd_capa *op_capa1;
1172 struct obd_capa *op_capa2;
1174 /* Various operation flags. */
1177 /* Operation type */
1181 struct md_enqueue_info;
1182 /* metadata stat-ahead */
1183 typedef int (* md_enqueue_cb_t)(struct ptlrpc_request *req,
1184 struct md_enqueue_info *minfo,
1187 struct md_enqueue_info {
1188 struct md_op_data mi_data;
1189 struct lookup_intent mi_it;
1190 struct lustre_handle mi_lockh;
1191 struct dentry *mi_dentry;
1192 md_enqueue_cb_t mi_cb;
1193 unsigned int mi_generation;
1198 struct module *o_owner;
1199 int (*o_iocontrol)(unsigned int cmd, struct obd_export *exp, int len,
1200 void *karg, void *uarg);
1201 int (*o_get_info)(struct obd_export *, __u32 keylen, void *key,
1202 __u32 *vallen, void *val, struct lov_stripe_md *lsm);
1203 int (*o_set_info_async)(struct obd_export *, __u32 keylen, void *key,
1204 __u32 vallen, void *val,
1205 struct ptlrpc_request_set *set);
1206 int (*o_attach)(struct obd_device *dev, obd_count len, void *data);
1207 int (*o_detach)(struct obd_device *dev);
1208 int (*o_setup) (struct obd_device *dev, struct lustre_cfg *cfg);
1209 int (*o_precleanup)(struct obd_device *dev,
1210 enum obd_cleanup_stage cleanup_stage);
1211 int (*o_cleanup)(struct obd_device *dev);
1212 int (*o_process_config)(struct obd_device *dev, obd_count len,
1214 int (*o_postrecov)(struct obd_device *dev);
1215 int (*o_add_conn)(struct obd_import *imp, struct obd_uuid *uuid,
1217 int (*o_del_conn)(struct obd_import *imp, struct obd_uuid *uuid);
1218 /* connect to the target device with given connection
1219 * data. @ocd->ocd_connect_flags is modified to reflect flags actually
1220 * granted by the target, which are guaranteed to be a subset of flags
1221 * asked for. If @ocd == NULL, use default parameters. */
1222 int (*o_connect)(const struct lu_env *env,
1223 struct lustre_handle *conn, struct obd_device *src,
1224 struct obd_uuid *cluuid, struct obd_connect_data *ocd,
1226 int (*o_reconnect)(const struct lu_env *env,
1227 struct obd_export *exp, struct obd_device *src,
1228 struct obd_uuid *cluuid,
1229 struct obd_connect_data *ocd);
1230 int (*o_disconnect)(struct obd_export *exp);
1232 /* Initialize/finalize fids infrastructure. */
1233 int (*o_fid_init)(struct obd_export *exp);
1234 int (*o_fid_fini)(struct obd_export *exp);
1236 /* Allocate new fid according to passed @hint. */
1237 int (*o_fid_alloc)(struct obd_export *exp, struct lu_fid *fid,
1238 struct md_op_data *op_data);
1241 * Object with @fid is getting deleted, we may want to do something
1244 int (*o_fid_delete)(struct obd_export *exp, const struct lu_fid *fid);
1246 int (*o_statfs)(struct obd_device *obd, struct obd_statfs *osfs,
1247 __u64 max_age, __u32 flags);
1248 int (*o_statfs_async)(struct obd_device *obd, struct obd_info *oinfo,
1249 __u64 max_age, struct ptlrpc_request_set *set);
1250 int (*o_packmd)(struct obd_export *exp, struct lov_mds_md **disk_tgt,
1251 struct lov_stripe_md *mem_src);
1252 int (*o_unpackmd)(struct obd_export *exp,struct lov_stripe_md **mem_tgt,
1253 struct lov_mds_md *disk_src, int disk_len);
1254 int (*o_checkmd)(struct obd_export *exp, struct obd_export *md_exp,
1255 struct lov_stripe_md *mem_tgt);
1256 int (*o_preallocate)(struct lustre_handle *, obd_count *req,
1258 /* FIXME: add fid capability support for create & destroy! */
1259 int (*o_precreate)(struct obd_export *exp);
1260 int (*o_create)(struct obd_export *exp, struct obdo *oa,
1261 struct lov_stripe_md **ea, struct obd_trans_info *oti);
1262 int (*o_destroy)(struct obd_export *exp, struct obdo *oa,
1263 struct lov_stripe_md *ea, struct obd_trans_info *oti,
1264 struct obd_export *md_exp);
1265 int (*o_setattr)(struct obd_export *exp, struct obd_info *oinfo,
1266 struct obd_trans_info *oti);
1267 int (*o_setattr_async)(struct obd_export *exp, struct obd_info *oinfo,
1268 struct obd_trans_info *oti,
1269 struct ptlrpc_request_set *rqset);
1270 int (*o_getattr)(struct obd_export *exp, struct obd_info *oinfo);
1271 int (*o_getattr_async)(struct obd_export *exp, struct obd_info *oinfo,
1272 struct ptlrpc_request_set *set);
1273 int (*o_brw)(int rw, struct obd_export *exp, struct obd_info *oinfo,
1274 obd_count oa_bufs, struct brw_page *pgarr,
1275 struct obd_trans_info *oti);
1276 int (*o_brw_async)(int rw, struct obd_export *exp,
1277 struct obd_info *oinfo, obd_count oa_bufs,
1278 struct brw_page *pgarr, struct obd_trans_info *oti,
1279 struct ptlrpc_request_set *);
1280 int (*o_prep_async_page)(struct obd_export *exp,
1281 struct lov_stripe_md *lsm,
1282 struct lov_oinfo *loi,
1283 cfs_page_t *page, obd_off offset,
1284 struct obd_async_page_ops *ops, void *data,
1285 void **res, int nocache,
1286 struct lustre_handle *lockh);
1287 int (*o_reget_short_lock)(struct obd_export *exp,
1288 struct lov_stripe_md *lsm,
1290 obd_off start, obd_off end,
1292 int (*o_release_short_lock)(struct obd_export *exp,
1293 struct lov_stripe_md *lsm, obd_off end,
1294 void *cookie, int rw);
1295 int (*o_queue_async_io)(struct obd_export *exp,
1296 struct lov_stripe_md *lsm,
1297 struct lov_oinfo *loi, void *cookie,
1298 int cmd, obd_off off, int count,
1299 obd_flag brw_flags, obd_flag async_flags);
1300 int (*o_queue_group_io)(struct obd_export *exp,
1301 struct lov_stripe_md *lsm,
1302 struct lov_oinfo *loi,
1303 struct obd_io_group *oig,
1304 void *cookie, int cmd, obd_off off, int count,
1305 obd_flag brw_flags, obd_flag async_flags);
1306 int (*o_trigger_group_io)(struct obd_export *exp,
1307 struct lov_stripe_md *lsm,
1308 struct lov_oinfo *loi,
1309 struct obd_io_group *oig);
1310 int (*o_set_async_flags)(struct obd_export *exp,
1311 struct lov_stripe_md *lsm,
1312 struct lov_oinfo *loi, void *cookie,
1313 obd_flag async_flags);
1314 int (*o_teardown_async_page)(struct obd_export *exp,
1315 struct lov_stripe_md *lsm,
1316 struct lov_oinfo *loi, void *cookie);
1317 int (*o_merge_lvb)(struct obd_export *exp, struct lov_stripe_md *lsm,
1318 struct ost_lvb *lvb, int kms_only);
1319 int (*o_adjust_kms)(struct obd_export *exp, struct lov_stripe_md *lsm,
1320 obd_off size, int shrink);
1321 int (*o_punch)(struct obd_export *exp, struct obd_info *oinfo,
1322 struct obd_trans_info *oti,
1323 struct ptlrpc_request_set *rqset);
1324 int (*o_sync)(struct obd_export *exp, struct obdo *oa,
1325 struct lov_stripe_md *ea, obd_size start, obd_size end,
1327 int (*o_migrate)(struct lustre_handle *conn, struct lov_stripe_md *dst,
1328 struct lov_stripe_md *src, obd_size start,
1329 obd_size end, struct obd_trans_info *oti);
1330 int (*o_copy)(struct lustre_handle *dstconn, struct lov_stripe_md *dst,
1331 struct lustre_handle *srconn, struct lov_stripe_md *src,
1332 obd_size start, obd_size end, struct obd_trans_info *);
1333 int (*o_iterate)(struct lustre_handle *conn,
1334 int (*)(obd_id, obd_gr, void *),
1335 obd_id *startid, obd_gr group, void *data);
1336 int (*o_preprw)(int cmd, struct obd_export *exp, struct obdo *oa,
1337 int objcount, struct obd_ioobj *obj,
1338 struct niobuf_remote *remote, int *nr_pages,
1339 struct niobuf_local *local,
1340 struct obd_trans_info *oti,
1341 struct lustre_capa *capa);
1342 int (*o_commitrw)(int cmd, struct obd_export *exp, struct obdo *oa,
1343 int objcount, struct obd_ioobj *obj,
1344 struct niobuf_remote *remote, int pages,
1345 struct niobuf_local *local,
1346 struct obd_trans_info *oti, int rc);
1347 int (*o_enqueue)(struct obd_export *, struct obd_info *oinfo,
1348 struct ldlm_enqueue_info *einfo,
1349 struct ptlrpc_request_set *rqset);
1350 int (*o_match)(struct obd_export *, struct lov_stripe_md *, __u32 type,
1351 ldlm_policy_data_t *, __u32 mode, int *flags, void *data,
1352 struct lustre_handle *lockh);
1353 int (*o_change_cbdata)(struct obd_export *, struct lov_stripe_md *,
1354 ldlm_iterator_t it, void *data);
1355 int (*o_cancel)(struct obd_export *, struct lov_stripe_md *md,
1356 __u32 mode, struct lustre_handle *);
1357 int (*o_cancel_unused)(struct obd_export *, struct lov_stripe_md *,
1358 int flags, void *opaque);
1359 int (*o_init_export)(struct obd_export *exp);
1360 int (*o_destroy_export)(struct obd_export *exp);
1361 int (*o_extent_calc)(struct obd_export *, struct lov_stripe_md *,
1362 int cmd, obd_off *);
1364 /* llog related obd_methods */
1365 int (*o_llog_init)(struct obd_device *obd, struct obd_llog_group *grp,
1366 struct obd_device *disk_obd, int count,
1367 struct llog_catid *logid, struct obd_uuid *uuid);
1368 int (*o_llog_finish)(struct obd_device *obd, int count);
1369 int (*o_llog_connect)(struct obd_export *, struct llogd_conn_body *);
1371 /* metadata-only methods */
1372 int (*o_pin)(struct obd_export *, const struct lu_fid *fid,
1373 struct obd_capa *, struct obd_client_handle *, int flag);
1374 int (*o_unpin)(struct obd_export *, struct obd_client_handle *, int);
1376 int (*o_import_event)(struct obd_device *, struct obd_import *,
1377 enum obd_import_event);
1379 int (*o_notify)(struct obd_device *obd, struct obd_device *watched,
1380 enum obd_notify_event ev, void *data);
1382 int (*o_health_check)(struct obd_device *);
1383 struct obd_uuid *(*o_get_uuid) (struct obd_export *exp);
1386 int (*o_quotacheck)(struct obd_export *, struct obd_quotactl *);
1387 int (*o_quotactl)(struct obd_export *, struct obd_quotactl *);
1389 int (*o_ping)(struct obd_export *exp);
1391 int (*o_register_page_removal_cb)(struct obd_export *exp,
1392 obd_page_removal_cb_t cb,
1393 obd_pin_extent_cb pin_cb);
1394 int (*o_unregister_page_removal_cb)(struct obd_export *exp,
1395 obd_page_removal_cb_t cb);
1396 int (*o_register_lock_cancel_cb)(struct obd_export *exp,
1397 obd_lock_cancel_cb cb);
1398 int (*o_unregister_lock_cancel_cb)(struct obd_export *exp,
1399 obd_lock_cancel_cb cb);
1401 int (*o_pool_new)(struct obd_device *obd, char *poolname);
1402 int (*o_pool_del)(struct obd_device *obd, char *poolname);
1403 int (*o_pool_add)(struct obd_device *obd, char *poolname,
1405 int (*o_pool_rem)(struct obd_device *obd, char *poolname,
1408 * NOTE: If adding ops, add another LPROCFS_OBD_OP_INIT() line
1409 * to lprocfs_alloc_obd_stats() in obdclass/lprocfs_status.c.
1410 * Also, add a wrapper function in include/linux/obd_class.h. */
1413 /* TODO: lmv_stripe_md should contain mds capabilities for all slave fids */
1414 struct lmv_stripe_md {
1419 struct lu_fid mea_ids[0];
1423 LUSTRE_OPC_MKDIR = (1 << 0),
1424 LUSTRE_OPC_SYMLINK = (1 << 1),
1425 LUSTRE_OPC_MKNOD = (1 << 2),
1426 LUSTRE_OPC_CREATE = (1 << 3),
1427 LUSTRE_OPC_ANY = (1 << 4)
1430 /* lmv structures */
1431 #define MEA_MAGIC_LAST_CHAR 0xb2221ca1
1432 #define MEA_MAGIC_ALL_CHARS 0xb222a11c
1433 #define MEA_MAGIC_HASH_SEGMENT 0xb222a11b
1435 #define MAX_HASH_SIZE_32 0x7fffffffUL
1436 #define MAX_HASH_SIZE 0x7fffffffffffffffULL
1437 #define MAX_HASH_HIGHEST_BIT 0x1000000000000000ULL
1440 struct mdt_body *body;
1441 struct lov_stripe_md *lsm;
1442 struct lmv_stripe_md *mea;
1443 #ifdef CONFIG_FS_POSIX_ACL
1444 struct posix_acl *posix_acl;
1446 struct mdt_remote_perm *remote_perm;
1447 struct obd_capa *mds_capa;
1448 struct obd_capa *oss_capa;
1451 struct md_open_data {
1452 struct obd_client_handle *mod_och;
1453 struct list_head mod_replay_list;
1456 struct lookup_intent;
1459 int (*m_getstatus)(struct obd_export *, struct lu_fid *,
1460 struct obd_capa **);
1461 int (*m_change_cbdata)(struct obd_export *, const struct lu_fid *,
1462 ldlm_iterator_t, void *);
1463 int (*m_close)(struct obd_export *, struct md_op_data *,
1464 struct md_open_data *, struct ptlrpc_request **);
1465 int (*m_create)(struct obd_export *, struct md_op_data *,
1466 const void *, int, int, __u32, __u32, cfs_cap_t,
1467 __u64, struct ptlrpc_request **);
1468 int (*m_done_writing)(struct obd_export *, struct md_op_data *,
1469 struct md_open_data *);
1470 int (*m_enqueue)(struct obd_export *, struct ldlm_enqueue_info *,
1471 struct lookup_intent *, struct md_op_data *,
1472 struct lustre_handle *, void *, int,
1473 struct ptlrpc_request **, int);
1474 int (*m_getattr)(struct obd_export *, const struct lu_fid *,
1475 struct obd_capa *, obd_valid, int,
1476 struct ptlrpc_request **);
1477 int (*m_getattr_name)(struct obd_export *, const struct lu_fid *,
1478 struct obd_capa *, const char *, int, obd_valid,
1479 int, __u32, struct ptlrpc_request **);
1480 int (*m_intent_lock)(struct obd_export *, struct md_op_data *,
1481 void *, int, struct lookup_intent *, int,
1482 struct ptlrpc_request **,
1483 ldlm_blocking_callback, int);
1484 int (*m_link)(struct obd_export *, struct md_op_data *,
1485 struct ptlrpc_request **);
1486 int (*m_rename)(struct obd_export *, struct md_op_data *,
1487 const char *, int, const char *, int,
1488 struct ptlrpc_request **);
1489 int (*m_is_subdir)(struct obd_export *, const struct lu_fid *,
1490 const struct lu_fid *,
1491 struct ptlrpc_request **);
1492 int (*m_setattr)(struct obd_export *, struct md_op_data *, void *,
1493 int , void *, int, struct ptlrpc_request **,
1494 struct md_open_data **mod);
1495 int (*m_sync)(struct obd_export *, const struct lu_fid *,
1496 struct obd_capa *, struct ptlrpc_request **);
1497 int (*m_readpage)(struct obd_export *, const struct lu_fid *,
1498 struct obd_capa *, __u64, struct page *,
1499 struct ptlrpc_request **);
1501 int (*m_unlink)(struct obd_export *, struct md_op_data *,
1502 struct ptlrpc_request **);
1504 int (*m_setxattr)(struct obd_export *, const struct lu_fid *,
1505 struct obd_capa *, obd_valid, const char *,
1506 const char *, int, int, int, __u32,
1507 struct ptlrpc_request **);
1509 int (*m_getxattr)(struct obd_export *, const struct lu_fid *,
1510 struct obd_capa *, obd_valid, const char *,
1511 const char *, int, int, int,
1512 struct ptlrpc_request **);
1514 int (*m_init_ea_size)(struct obd_export *, int, int, int);
1516 int (*m_get_lustre_md)(struct obd_export *, struct ptlrpc_request *,
1517 struct obd_export *, struct obd_export *,
1518 struct lustre_md *);
1520 int (*m_free_lustre_md)(struct obd_export *, struct lustre_md *);
1522 int (*m_set_open_replay_data)(struct obd_export *,
1523 struct obd_client_handle *,
1524 struct ptlrpc_request *);
1525 int (*m_clear_open_replay_data)(struct obd_export *,
1526 struct obd_client_handle *);
1527 int (*m_set_lock_data)(struct obd_export *, __u64 *, void *);
1529 ldlm_mode_t (*m_lock_match)(struct obd_export *, int,
1530 const struct lu_fid *, ldlm_type_t,
1531 ldlm_policy_data_t *, ldlm_mode_t,
1532 struct lustre_handle *);
1534 int (*m_cancel_unused)(struct obd_export *, const struct lu_fid *,
1535 ldlm_policy_data_t *, ldlm_mode_t, int flags,
1537 int (*m_renew_capa)(struct obd_export *, struct obd_capa *oc,
1538 renew_capa_cb_t cb);
1540 int (*m_get_remote_perm)(struct obd_export *, const struct lu_fid *,
1541 struct obd_capa *, __u32,
1542 struct ptlrpc_request **);
1544 int (*m_intent_getattr_async)(struct obd_export *,
1545 struct md_enqueue_info *,
1546 struct ldlm_enqueue_info *);
1548 int (*m_revalidate_lock)(struct obd_export *,
1549 struct lookup_intent *,
1553 * NOTE: If adding ops, add another LPROCFS_MD_OP_INIT() line to
1554 * lprocfs_alloc_md_stats() in obdclass/lprocfs_status.c. Also, add a
1555 * wrapper function in include/linux/obd_class.h.
1559 struct lsm_operations {
1560 void (*lsm_free)(struct lov_stripe_md *);
1561 int (*lsm_destroy)(struct lov_stripe_md *, struct obdo *oa,
1562 struct obd_export *md_exp);
1563 void (*lsm_stripe_by_index)(struct lov_stripe_md *, int *, obd_off *,
1565 void (*lsm_stripe_by_offset)(struct lov_stripe_md *, int *, obd_off *,
1567 obd_off (*lsm_stripe_offset_by_index)(struct lov_stripe_md *, int);
1568 obd_off (*lsm_stripe_offset_by_offset)(struct lov_stripe_md *, obd_off);
1569 int (*lsm_stripe_index_by_offset)(struct lov_stripe_md *, obd_off);
1570 int (*lsm_revalidate) (struct lov_stripe_md *, struct obd_device *obd);
1571 int (*lsm_lmm_verify) (struct lov_mds_md *lmm, int lmm_bytes,
1573 int (*lsm_unpackmd) (struct lov_obd *lov, struct lov_stripe_md *lsm,
1574 struct lov_mds_md *lmm);
1577 extern struct lsm_operations lsm_v1_ops;
1578 extern struct lsm_operations lsm_join_ops;
1579 extern struct lsm_operations lsm_v3_ops;
1580 static inline struct lsm_operations *lsm_op_find(int magic)
1585 case LOV_MAGIC_JOIN:
1586 return &lsm_join_ops;
1590 CERROR("Cannot recognize lsm_magic %d\n", magic);
1595 int lvfs_check_io_health(struct obd_device *obd, struct file *file);
1597 /* Requests for obd_extent_calc() */
1598 #define OBD_CALC_STRIPE_START 1
1599 #define OBD_CALC_STRIPE_END 2
1601 static inline void obd_transno_commit_cb(struct obd_device *obd, __u64 transno,
1605 CERROR("%s: transno "LPU64" commit error: %d\n",
1606 obd->obd_name, transno, error);
1609 if (transno > obd->obd_last_committed) {
1610 CDEBUG(D_HA, "%s: transno "LPD64" committed\n",
1611 obd->obd_name, transno);
1612 obd->obd_last_committed = transno;
1613 ptlrpc_commit_replies (obd);
1615 CDEBUG(D_INFO, "%s: transno "LPD64" committed\n",
1616 obd->obd_name, transno);
1620 static inline void init_obd_quota_ops(quota_interface_t *interface,
1621 struct obd_ops *obd_ops)
1627 obd_ops->o_quotacheck = QUOTA_OP(interface, check);
1628 obd_ops->o_quotactl = QUOTA_OP(interface, ctl);
1631 static inline __u64 oinfo_mdsno(struct obd_info *oinfo)
1633 return oinfo->oi_oa->o_gr - FILTER_GROUP_MDS0;
1636 static inline struct lustre_capa *oinfo_capa(struct obd_info *oinfo)
1638 return oinfo->oi_capa;
1641 #endif /* __OBD_H */