1 /* -*- mode: c; c-basic-offset: 8; indent-tabs-mode: nil; -*-
2 * vim:expandtab:shiftwidth=8:tabstop=8:
6 * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
8 * This program is free software; you can redistribute it and/or modify
9 * it under the terms of the GNU General Public License version 2 only,
10 * as published by the Free Software Foundation.
12 * This program is distributed in the hope that it will be useful, but
13 * WITHOUT ANY WARRANTY; without even the implied warranty of
14 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
15 * General Public License version 2 for more details (a copy is included
16 * in the LICENSE file that accompanied this code).
18 * You should have received a copy of the GNU General Public License
19 * version 2 along with this program; If not, see
20 * http://www.sun.com/software/products/lustre/docs/GPLv2.pdf
22 * Please contact Sun Microsystems, Inc., 4150 Network Circle, Santa Clara,
23 * CA 95054 USA or visit www.sun.com if you need additional information or
29 * Copyright 2008 Sun Microsystems, Inc. All rights reserved
30 * Use is subject to license terms.
33 * This file is part of Lustre, http://www.lustre.org/
34 * Lustre is a trademark of Sun Microsystems, Inc.
40 #if defined(__linux__)
41 #include <linux/obd.h>
42 #elif defined(__APPLE__)
43 #include <darwin/obd.h>
44 #elif defined(__WINNT__)
45 #include <winnt/obd.h>
47 #error Unsupported operating system.
50 #define IOC_OSC_TYPE 'h'
51 #define IOC_OSC_MIN_NR 20
52 #define IOC_OSC_SET_ACTIVE _IOWR(IOC_OSC_TYPE, 21, struct obd_device *)
53 #define IOC_OSC_MAX_NR 50
55 #define IOC_MDC_TYPE 'i'
56 #define IOC_MDC_MIN_NR 20
57 /* Moved to lustre_user.h
58 #define IOC_MDC_LOOKUP _IOWR(IOC_MDC_TYPE, 20, struct obd_ioctl_data *)
59 #define IOC_MDC_GETSTRIPE _IOWR(IOC_MDC_TYPE, 21, struct lov_mds_md *) */
60 #define IOC_MDC_MAX_NR 50
62 #include <lustre/lustre_idl.h>
63 #include <lu_object.h>
65 #include <lustre_lib.h>
66 #include <lustre_export.h>
67 #include <lustre_quota.h>
68 #include <lustre_fld.h>
69 #include <lustre_capa.h>
70 #include <class_hash.h>
72 #include <libcfs/bitmap.h>
75 #define MAX_OBD_DEVICES 8192
77 /* this is really local to the OSC */
78 struct loi_oap_pages {
79 struct list_head lop_pending;
80 struct list_head lop_urgent;
81 struct list_head lop_pending_group;
91 struct lov_oinfo { /* per-stripe data structure */
92 __u64 loi_id; /* object ID on the target OST */
93 __u64 loi_gr; /* object group on the target OST */
94 int loi_ost_idx; /* OST stripe index in lov_tgt_desc->tgts */
95 int loi_ost_gen; /* generation of this loi_ost_idx */
97 /* used by the osc to keep track of what objects to build into rpcs */
98 struct loi_oap_pages loi_read_lop;
99 struct loi_oap_pages loi_write_lop;
100 /* _cli_ is poorly named, it should be _ready_ */
101 struct list_head loi_cli_item;
102 struct list_head loi_write_item;
103 struct list_head loi_read_item;
105 unsigned long loi_kms_valid:1;
106 __u64 loi_kms; /* known minimum size */
107 struct ost_lvb loi_lvb;
108 struct osc_async_rc loi_ar;
111 static inline void loi_kms_set(struct lov_oinfo *oinfo, __u64 kms)
113 oinfo->loi_kms = kms;
114 oinfo->loi_kms_valid = 1;
117 static inline void loi_init(struct lov_oinfo *loi)
119 CFS_INIT_LIST_HEAD(&loi->loi_read_lop.lop_pending);
120 CFS_INIT_LIST_HEAD(&loi->loi_read_lop.lop_urgent);
121 CFS_INIT_LIST_HEAD(&loi->loi_read_lop.lop_pending_group);
122 CFS_INIT_LIST_HEAD(&loi->loi_write_lop.lop_pending);
123 CFS_INIT_LIST_HEAD(&loi->loi_write_lop.lop_urgent);
124 CFS_INIT_LIST_HEAD(&loi->loi_write_lop.lop_pending_group);
125 CFS_INIT_LIST_HEAD(&loi->loi_cli_item);
126 CFS_INIT_LIST_HEAD(&loi->loi_write_item);
127 CFS_INIT_LIST_HEAD(&loi->loi_read_item);
130 /*extent array item for describing the joined file extent info*/
132 __u64 le_start; /* extent start */
133 __u64 le_len; /* extent length */
134 int le_loi_idx; /* extent #1 loi's index in lsm loi array */
135 int le_stripe_count; /* extent stripe count*/
138 /*Lov array info for describing joined file array EA info*/
139 struct lov_array_info {
140 struct llog_logid lai_array_id; /* MDS med llog object id */
141 unsigned lai_ext_count; /* number of extent count */
142 struct lov_extent *lai_ext_array; /* extent desc array */
145 struct lov_stripe_md {
147 pid_t lsm_lock_owner; /* debugging */
150 /* Public members. */
151 __u64 lw_object_id; /* lov object id */
152 __u64 lw_object_gr; /* lov object group */
153 __u64 lw_maxbytes; /* maximum possible file size */
155 /* LOV-private members start here -- only for use in lov/. */
157 __u32 lw_stripe_size; /* size of the stripe */
158 __u32 lw_pattern; /* striping pattern (RAID0, RAID1) */
159 unsigned lw_stripe_count; /* number of objects being striped over */
160 char lw_pool_name[LOV_MAXPOOLNAME]; /* pool name */
163 struct lov_array_info *lsm_array; /*Only for joined file array info*/
164 struct lov_oinfo *lsm_oinfo[0];
167 #define lsm_object_id lsm_wire.lw_object_id
168 #define lsm_object_gr lsm_wire.lw_object_gr
169 #define lsm_maxbytes lsm_wire.lw_maxbytes
170 #define lsm_magic lsm_wire.lw_magic
171 #define lsm_stripe_size lsm_wire.lw_stripe_size
172 #define lsm_pattern lsm_wire.lw_pattern
173 #define lsm_stripe_count lsm_wire.lw_stripe_count
174 #define lsm_pool_name lsm_wire.lw_pool_name
178 typedef int (*obd_enqueue_update_f)(struct obd_info *oinfo, int rc);
180 /* obd info for a particular level (lov, osc). */
182 /* Lock policy. It keeps an extent which is specific for a particular
183 * OSC. (e.g. lov_prep_enqueue_set initialises extent of the policy,
184 * and osc_enqueue passes it into ldlm_lock_match & ldlm_cli_enqueue. */
185 ldlm_policy_data_t oi_policy;
186 /* Flags used for set request specific flags:
187 - while lock handling, the flags obtained on the enqueue
188 request are set here.
189 - while stats, the flags used for control delay/resend.
192 /* Lock handle specific for every OSC lock. */
193 struct lustre_handle *oi_lockh;
194 /* lsm data specific for every OSC. */
195 struct lov_stripe_md *oi_md;
196 /* obdo data specific for every OSC, if needed at all. */
198 /* statfs data specific for every OSC, if needed at all. */
199 struct obd_statfs *oi_osfs;
200 /* An update callback which is called to update some data on upper
201 * level. E.g. it is used for update lsm->lsm_oinfo at every recieved
202 * request in osc level for enqueue requests. It is also possible to
203 * update some caller data from LOV layer if needed. */
204 obd_enqueue_update_f oi_cb_up;
205 /* oss capability, its type is obd_capa in client to avoid copy.
206 * in contrary its type is lustre_capa in OSS. */
210 /* compare all relevant fields. */
211 static inline int lov_stripe_md_cmp(struct lov_stripe_md *m1,
212 struct lov_stripe_md *m2)
215 * ->lsm_wire contains padding, but it should be zeroed out during
218 return memcmp(&m1->lsm_wire, &m2->lsm_wire, sizeof m1->lsm_wire);
221 void lov_stripe_lock(struct lov_stripe_md *md);
222 void lov_stripe_unlock(struct lov_stripe_md *md);
225 struct list_head typ_chain;
226 struct obd_ops *typ_dt_ops;
227 struct md_ops *typ_md_ops;
228 cfs_proc_dir_entry_t *typ_procroot;
231 struct lu_device_type *typ_lu;
232 spinlock_t obd_type_lock;
243 ASYNC_READY = 0x1, /* ap_make_ready will not be called before this
244 page is added to an rpc */
245 ASYNC_URGENT = 0x2, /* page must be put into an RPC before return */
246 ASYNC_COUNT_STABLE = 0x4, /* ap_refresh_count will not be called
247 to give the caller a chance to update
248 or cancel the size of the io */
249 ASYNC_GROUP_SYNC = 0x8, /* ap_completion will not be called, instead
250 the page is accounted for in the
251 obd_io_group given to
252 obd_queue_group_io */
255 struct obd_async_page_ops {
256 int (*ap_make_ready)(void *data, int cmd);
257 int (*ap_refresh_count)(void *data, int cmd);
258 void (*ap_fill_obdo)(void *data, int cmd, struct obdo *oa);
259 void (*ap_update_obdo)(void *data, int cmd, struct obdo *oa,
261 int (*ap_completion)(void *data, int cmd, struct obdo *oa, int rc);
262 struct obd_capa *(*ap_lookup_capa)(void *data, int cmd);
265 /* the `oig' is passed down from a caller of obd rw methods. the callee
266 * records enough state such that the caller can sleep on the oig and
267 * be woken when all the callees have finished their work */
268 struct obd_io_group {
270 atomic_t oig_refcount;
273 struct list_head oig_occ_list;
274 cfs_waitq_t oig_waitq;
277 /* the oig callback context lets the callee of obd rw methods register
278 * for callbacks from the caller. */
279 struct oig_callback_context {
280 struct list_head occ_oig_item;
281 /* called when the caller has received a signal while sleeping.
282 * callees of this method are encouraged to abort their state
283 * in the oig. This may be called multiple times. */
284 void (*occ_interrupted)(struct oig_callback_context *occ);
285 unsigned long interrupted:1;
288 /* Individual type definitions */
290 struct ost_server_data;
292 /* hold common fields for "target" device */
293 struct obd_device_target {
294 struct super_block *obt_sb;
295 atomic_t obt_quotachecking;
296 struct lustre_quota_ctxt obt_qctxt;
299 typedef void (*obd_pin_extent_cb)(void *data);
300 typedef int (*obd_page_removal_cb_t)(void *data, int discard);
301 typedef int (*obd_lock_cancel_cb)(struct ldlm_lock *,struct ldlm_lock_desc *,
306 LLOG_CONFIG_ORIG_CTXT = 0,
307 LLOG_CONFIG_REPL_CTXT,
308 LLOG_MDS_OST_ORIG_CTXT,
309 LLOG_MDS_OST_REPL_CTXT,
316 LLOG_LOVEA_ORIG_CTXT,
317 LLOG_LOVEA_REPL_CTXT,
321 #define FILTER_SUBDIR_COUNT 32 /* set to zero for no subdirs */
323 struct filter_subdirs {
324 cfs_dentry_t *dentry[FILTER_SUBDIR_COUNT];
334 /* NB this field MUST be first */
335 struct obd_device_target fo_obt;
336 const char *fo_fstype;
337 struct vfsmount *fo_vfsmnt;
340 cfs_dentry_t *fo_dentry_O;
341 cfs_dentry_t **fo_dentry_O_groups;
342 struct filter_subdirs *fo_dentry_O_sub;
343 struct semaphore fo_init_lock; /* group initialization lock */
344 int fo_committed_group;
347 spinlock_t fo_objidlock; /* protect fo_lastobjid */
348 spinlock_t fo_translock; /* protect fsd_last_transno */
349 struct file *fo_rcvd_filp;
350 struct file *fo_health_check_filp;
351 struct lr_server_data *fo_fsd;
352 unsigned long *fo_last_rcvd_slots;
353 __u64 fo_mount_count;
355 unsigned long fo_destroys_in_progress;
356 struct semaphore fo_create_locks[FILTER_SUBDIR_COUNT];
358 struct list_head fo_export_list;
361 obd_size fo_tot_dirty; /* protected by obd_osfs_lock */
362 obd_size fo_tot_granted; /* all values in bytes */
363 obd_size fo_tot_pending;
365 obd_size fo_readcache_max_filesize;
367 int fo_writethrough_cache;
369 struct obd_import *fo_mdc_imp;
370 struct obd_uuid fo_mdc_uuid;
371 struct lustre_handle fo_mdc_conn;
372 struct file **fo_last_objid_files;
373 __u64 *fo_last_objids; /* last created objid for groups,
374 * protected by fo_objidlock */
376 struct semaphore fo_alloc_lock;
378 atomic_t fo_r_in_flight;
379 atomic_t fo_w_in_flight;
382 * per-filter pool of kiobuf's allocated by filter_common_setup() and
383 * torn down by filter_cleanup(). Contains OST_NUM_THREADS elements of
384 * which ->fo_iobuf_count were allocated.
386 * This pool contains kiobuf used by
387 * filter_{prep,commit}rw_{read,write}() and is shared by all OST
390 * Locking: none, each OST thread uses only one element, determined by
391 * its "ordinal number", ->t_id.
393 struct filter_iobuf **fo_iobuf_pool;
396 struct list_head fo_llog_list;
397 spinlock_t fo_llog_list_lock;
399 struct brw_stats fo_filter_stats;
400 struct lustre_quota_ctxt fo_quota_ctxt;
401 spinlock_t fo_quotacheck_lock;
402 atomic_t fo_quotachecking;
404 int fo_fmd_max_num; /* per exp filter_mod_data */
405 int fo_fmd_max_age; /* jiffies to fmd expiry */
408 rwlock_t fo_sptlrpc_lock;
409 struct sptlrpc_rule_set fo_sptlrpc_rset;
411 /* capability related */
412 unsigned int fo_fl_oss_capa;
413 struct list_head fo_capa_keys;
414 struct hlist_head *fo_capa_hash;
415 struct llog_commit_master *fo_lcm;
418 #define OSC_MAX_RIF_DEFAULT 8
419 #define OSC_MAX_RIF_MAX 256
420 #define OSC_MAX_DIRTY_DEFAULT (OSC_MAX_RIF_DEFAULT * 4)
421 #define OSC_MAX_DIRTY_MB_MAX 2048 /* arbitrary, but < MAX_LONG bytes */
422 #define OSC_DEFAULT_RESENDS 10
424 #define MDC_MAX_RIF_DEFAULT 8
425 #define MDC_MAX_RIF_MAX 512
431 struct rw_semaphore cl_sem;
432 struct obd_uuid cl_target_uuid;
433 struct obd_import *cl_import; /* ptlrpc connection state */
435 /* max_mds_easize is purely a performance thing so we don't have to
436 * call obd_size_diskmd() all the time. */
437 int cl_default_mds_easize;
438 int cl_max_mds_easize;
439 int cl_max_mds_cookiesize;
441 /* security configuration */
442 struct sptlrpc_rule_set cl_sptlrpc_rset;
443 enum lustre_sec_part cl_sec_part;
445 //struct llog_canceld_ctxt *cl_llcd; /* it's included by obd_llog_ctxt */
446 void *cl_llcd_offset;
448 /* the grant values are protected by loi_list_lock below */
449 long cl_dirty; /* all _dirty_ in bytes */
450 long cl_dirty_max; /* allowed w/o rpc */
451 long cl_avail_grant; /* bytes of credit for ost */
452 long cl_lost_grant; /* lost credits (trunc) */
453 struct list_head cl_cache_waiters; /* waiting for cache/grant */
455 /* keep track of objects that have lois that contain pages which
456 * have been queued for async brw. this lock also protects the
457 * lists of osc_client_pages that hang off of the loi */
459 * ->cl_loi_list_lock protects consistency of
460 * ->cl_loi_{ready,read,write}_list. ->ap_make_ready() and
461 * ->ap_completion() call-backs are executed under this lock. As we
462 * cannot guarantee that these call-backs never block on all platforms
463 * (as a matter of fact they do block on Mac OS X), type of
464 * ->cl_loi_list_lock is platform dependent: it's a spin-lock on Linux
465 * and blocking mutex on Mac OS X. (Alternative is to make this lock
466 * blocking everywhere, but we don't want to slow down fast-path of
467 * our main platform.)
469 * Exact type of ->cl_loi_list_lock is defined in arch/obd.h together
470 * with client_obd_list_{un,}lock() and
471 * client_obd_list_lock_{init,done}() functions.
473 client_obd_lock_t cl_loi_list_lock;
474 struct list_head cl_loi_ready_list;
475 struct list_head cl_loi_write_list;
476 struct list_head cl_loi_read_list;
479 /* just a sum of the loi/lop pending numbers to be exported by /proc */
480 int cl_pending_w_pages;
481 int cl_pending_r_pages;
482 int cl_max_pages_per_rpc;
483 int cl_max_rpcs_in_flight;
484 struct obd_histogram cl_read_rpc_hist;
485 struct obd_histogram cl_write_rpc_hist;
486 struct obd_histogram cl_read_page_hist;
487 struct obd_histogram cl_write_page_hist;
488 struct obd_histogram cl_read_offset_hist;
489 struct obd_histogram cl_write_offset_hist;
491 /* number of in flight destroy rpcs is limited to max_rpcs_in_flight */
492 atomic_t cl_destroy_in_flight;
493 cfs_waitq_t cl_destroy_waitq;
495 struct mdc_rpc_lock *cl_rpc_lock;
496 struct mdc_rpc_lock *cl_setattr_lock;
497 struct mdc_rpc_lock *cl_close_lock;
498 struct osc_creator cl_oscc;
501 struct semaphore cl_mgc_sem;
502 struct vfsmount *cl_mgc_vfsmnt;
503 struct dentry *cl_mgc_configs_dir;
504 atomic_t cl_mgc_refcount;
505 struct obd_export *cl_mgc_mgsexp;
507 /* checksumming for data sent over the network */
508 unsigned int cl_checksum:1; /* 0 = disabled, 1 = enabled */
509 /* supported checksum types that are worked out at connect time */
510 __u32 cl_supp_cksum_types;
511 /* checksum algorithm to be used */
512 cksum_type_t cl_cksum_type;
514 /* also protected by the poorly named _loi_list_lock lock above */
515 struct osc_async_rc cl_ar;
517 /* used by quotacheck */
518 int cl_qchk_stat; /* quotacheck stat of the peer */
520 /* sequence manager */
521 struct lu_client_seq *cl_seq;
523 atomic_t cl_resends; /* resend count */
525 /* Cache of triples */
526 struct lustre_cache *cl_cache;
527 obd_lock_cancel_cb cl_ext_lock_cancel_cb;
529 #define obd2cli_tgt(obd) ((char *)(obd)->u.cli.cl_target_uuid.uuid)
531 #define CL_NOT_QUOTACHECKED 1 /* client->cl_qchk_stat init value */
534 struct ptlrpc_service *mgs_service;
535 struct vfsmount *mgs_vfsmnt;
536 struct super_block *mgs_sb;
537 struct dentry *mgs_configs_dir;
538 struct dentry *mgs_fid_de;
539 struct list_head mgs_fs_db_list;
540 struct semaphore mgs_sem;
541 cfs_proc_dir_entry_t *mgs_proc_live;
545 /* NB this field MUST be first */
546 struct obd_device_target mds_obt;
547 struct ptlrpc_service *mds_service;
548 struct ptlrpc_service *mds_setattr_service;
549 struct ptlrpc_service *mds_readpage_service;
550 struct vfsmount *mds_vfsmnt;
551 cfs_dentry_t *mds_fid_de;
553 int mds_max_cookiesize;
554 struct file *mds_rcvd_filp;
555 spinlock_t mds_transno_lock;
556 __u64 mds_last_transno;
557 __u64 mds_mount_count;
559 unsigned long mds_atime_diff;
560 struct semaphore mds_epoch_sem;
561 struct ll_fid mds_rootfid;
562 struct lr_server_data *mds_server_data;
563 cfs_dentry_t *mds_pending_dir;
564 cfs_dentry_t *mds_logs_dir;
565 cfs_dentry_t *mds_objects_dir;
566 struct llog_handle *mds_cfg_llh;
567 struct obd_device *mds_osc_obd; /* XXX lov_obd */
568 struct obd_uuid mds_lov_uuid;
570 struct obd_export *mds_osc_exp; /* XXX lov_exp */
571 struct lov_desc mds_lov_desc;
574 /* mark pages dirty for write. */
575 bitmap_t *mds_lov_page_dirty;
576 /* array for store pages with obd_id */
577 void **mds_lov_page_array;
578 /* file for store objid */
579 struct file *mds_lov_objid_filp;
580 __u32 mds_lov_objid_count;
581 __u32 mds_lov_objid_lastpage;
582 __u32 mds_lov_objid_lastidx;
584 struct file *mds_health_check_filp;
585 unsigned long *mds_client_bitmap;
586 // struct upcall_cache *mds_group_hash;
588 struct lustre_quota_info mds_quota_info;
589 struct semaphore mds_qonoff_sem;
590 struct semaphore mds_health_sem;
591 unsigned long mds_fl_user_xattr:1,
593 mds_evict_ost_nids:1,
597 struct upcall_cache *mds_identity_cache;
599 /* for capability keys update */
600 struct lustre_capa_key *mds_capa_keys;
601 struct rw_semaphore mds_notify_lock;
605 extern __u32 mds_max_ost_index;
607 #define MDS_LOV_ALLOC_SIZE (CFS_PAGE_SIZE)
609 #define OBJID_PER_PAGE() (MDS_LOV_ALLOC_SIZE / sizeof(obd_id))
611 #define MDS_LOV_OBJID_PAGES_COUNT (mds_max_ost_index/OBJID_PER_PAGE())
613 extern int mds_lov_init_objids(struct obd_device *obd);
614 extern void mds_lov_destroy_objids(struct obd_device *obd);
627 struct lustre_handle eo_nl_lock;
632 struct ptlrpc_service *ost_service;
633 struct ptlrpc_service *ost_create_service;
634 struct ptlrpc_service *ost_io_service;
635 struct semaphore ost_health_sem;
638 struct echo_client_obd {
639 struct obd_export *ec_exp; /* the local connection to osc/lov */
641 struct list_head ec_objects;
647 struct obd_uuid lqo_uuid; /* ptlrpc's c_remote_uuid */
648 struct list_head lqo_oss_list; /* link to lov_qos */
649 __u32 lqo_ost_count; /* number of osts on this oss */
650 __u64 lqo_bavail; /* total bytes avail on OSS */
651 __u64 lqo_penalty; /* current penalty */
652 __u64 lqo_penalty_per_obj; /* penalty decrease every obj*/
656 struct lov_qos_oss *ltq_oss; /* oss info */
657 __u64 ltq_penalty; /* current penalty */
658 __u64 ltq_penalty_per_obj; /* penalty decrease every obj*/
659 __u64 ltq_weight; /* net weighting */
660 unsigned int ltq_usable:1; /* usable for striping */
663 /* Generic subset of OSTs */
665 __u32 *op_array; /* array of index of
667 unsigned int op_count; /* number of OSTs in the array */
668 unsigned int op_size; /* allocated size of lp_array */
669 rwlock_t op_rwlock; /* to protect lov_pool use */
672 /* Round-robin allocator data */
674 __u32 lqr_start_idx; /* start index of new inode */
675 __u32 lqr_offset_idx; /* aliasing for start_idx */
676 int lqr_start_count; /* reseed counter */
677 struct ost_pool lqr_pool; /* round-robin optimized list */
678 unsigned long lqr_dirty:1; /* recalc round-robin list */
681 /* Stripe placement optimization */
683 struct list_head lq_oss_list; /* list of OSSs that targets use */
684 struct rw_semaphore lq_rw_sem;
685 __u32 lq_active_oss_count;
686 unsigned int lq_prio_free; /* priority for free space */
687 struct lov_qos_rr lq_rr; /* round robin qos data */
688 unsigned long lq_dirty:1, /* recalc qos data */
689 lq_same_space:1,/* the ost's all have approx.
690 the same space avail */
691 lq_reset:1; /* zero current penalties */
694 struct lov_tgt_desc {
695 struct obd_uuid ltd_uuid;
696 struct obd_export *ltd_exp;
697 struct ltd_qos ltd_qos; /* qos info per target */
699 __u32 ltd_index; /* index in lov_obd->tgts */
700 unsigned long ltd_active:1,/* is this target up for requests */
701 ltd_activate:1,/* should this target be activated */
702 ltd_reap:1; /* should this target be deleted */
706 #define pool_tgt_size(_p) _p->pool_obds.op_size
707 #define pool_tgt_count(_p) _p->pool_obds.op_count
708 #define pool_tgt_array(_p) _p->pool_obds.op_array
709 #define pool_tgt_rwlock(_p) _p->pool_obds.op_rwlock
710 #define pool_tgt(_p, _i) _p->pool_lov->lov_tgts[_p->pool_obds.op_array[_i]]
713 char pool_name[LOV_MAXPOOLNAME + 1]; /* name of pool */
714 struct ost_pool pool_obds; /* pool members */
715 struct lov_qos_rr pool_rr; /* round robin qos */
716 struct hlist_node pool_hash; /* access by poolname */
717 struct list_head pool_list; /* serial access */
718 cfs_proc_dir_entry_t *pool_proc_entry; /* file in /proc */
719 struct lov_obd *pool_lov; /* lov obd to which this
724 struct lov_desc desc;
725 struct lov_tgt_desc **lov_tgts; /* sparse array */
726 struct ost_pool lov_packed; /* all OSTs in a packed
728 struct semaphore lov_lock;
729 struct obd_connect_data lov_ocd;
730 struct lov_qos lov_qos; /* qos info per lov */
731 atomic_t lov_refcount;
732 __u32 lov_tgt_count; /* how many OBD's */
733 __u32 lov_active_tgt_count; /* how many active */
734 __u32 lov_death_row;/* tgts scheduled to be deleted */
735 __u32 lov_tgt_size; /* size of tgts array */
737 obd_page_removal_cb_t lov_page_removal_cb;
738 obd_pin_extent_cb lov_page_pin_cb;
739 obd_lock_cancel_cb lov_lock_cancel_cb;
741 lustre_hash_t *lov_pools_hash_body; /* used for key access */
742 struct list_head lov_pool_list; /* used for sequential access */
743 cfs_proc_dir_entry_t *lov_pool_proc_entry;
746 struct lmv_tgt_desc {
747 struct obd_uuid ltd_uuid;
748 struct obd_export *ltd_exp;
749 int ltd_active; /* is this target up for requests */
751 struct semaphore ltd_fid_sem;
754 enum placement_policy {
755 PLACEMENT_CHAR_POLICY = 0,
756 PLACEMENT_NID_POLICY = 1,
757 PLACEMENT_INVAL_POLICY = 2,
761 typedef enum placement_policy placement_policy_t;
765 struct lu_client_fld lmv_fld;
767 placement_policy_t lmv_placement;
768 struct lmv_desc desc;
769 struct obd_uuid cluuid;
770 struct obd_export *exp;
777 struct semaphore init_sem;
779 struct lmv_tgt_desc *tgts;
782 struct obd_connect_data *datas;
785 struct obd_connect_data conn_data;
788 struct niobuf_local {
793 cfs_dentry_t *dentry;
798 #define LUSTRE_FLD_NAME "fld"
799 #define LUSTRE_SEQ_NAME "seq"
801 #define LUSTRE_CMM_NAME "cmm"
802 #define LUSTRE_MDD_NAME "mdd"
803 #define LUSTRE_OSD_NAME "osd"
804 #define LUSTRE_LMV_NAME "lmv"
805 #define LUSTRE_CMM_MDC_NAME "cmm-mdc"
807 /* obd device type names */
808 /* FIXME all the references to LUSTRE_MDS_NAME should be swapped with LUSTRE_MDT_NAME */
809 #define LUSTRE_MDS_NAME "mds"
810 #define LUSTRE_MDT_NAME "mdt"
811 #define LUSTRE_MDC_NAME "mdc"
812 #define LUSTRE_OSS_NAME "ost" /* FIXME change name to oss */
813 #define LUSTRE_OST_NAME "obdfilter" /* FIXME change name to ost */
814 #define LUSTRE_OSC_NAME "osc"
815 #define LUSTRE_LOV_NAME "lov"
816 #define LUSTRE_MGS_NAME "mgs"
817 #define LUSTRE_MGC_NAME "mgc"
819 #define LUSTRE_CACHEOBD_NAME "cobd"
820 #define LUSTRE_ECHO_NAME "obdecho"
821 #define LUSTRE_ECHO_CLIENT_NAME "echo_client"
823 /* Constant obd names (post-rename) */
824 #define LUSTRE_MDS_OBDNAME "MDS"
825 #define LUSTRE_OSS_OBDNAME "OSS"
826 #define LUSTRE_MGS_OBDNAME "MGS"
827 #define LUSTRE_MGC_OBDNAME "MGC"
829 /* Don't conflict with on-wire flags OBD_BRW_WRITE, etc */
830 #define N_LOCAL_TEMP_PAGE 0x10000000
832 struct obd_trans_info {
835 /* Only used on the server side for tracking acks. */
836 struct oti_req_ack_lock {
837 struct lustre_handle lock;
841 struct llog_cookie oti_onecookie;
842 struct llog_cookie *oti_logcookies;
845 /* initial thread handling transaction */
849 struct obd_uuid *oti_ost_uuid;
852 static inline void oti_init(struct obd_trans_info *oti,
853 struct ptlrpc_request *req)
857 memset(oti, 0, sizeof(*oti));
862 oti->oti_xid = req->rq_xid;
864 if (req->rq_repmsg != NULL)
865 oti->oti_transno = lustre_msg_get_transno(req->rq_repmsg);
866 oti->oti_thread_id = req->rq_svc_thread ? req->rq_svc_thread->t_id : -1;
867 if (req->rq_reqmsg != NULL)
868 oti->oti_conn_cnt = lustre_msg_get_conn_cnt(req->rq_reqmsg);
871 static inline void oti_alloc_cookies(struct obd_trans_info *oti,int num_cookies)
876 if (num_cookies == 1)
877 oti->oti_logcookies = &oti->oti_onecookie;
879 OBD_ALLOC(oti->oti_logcookies,
880 num_cookies * sizeof(oti->oti_onecookie));
882 oti->oti_numcookies = num_cookies;
885 static inline void oti_free_cookies(struct obd_trans_info *oti)
887 if (!oti || !oti->oti_logcookies)
890 if (oti->oti_logcookies == &oti->oti_onecookie)
891 LASSERT(oti->oti_numcookies == 1);
893 OBD_FREE(oti->oti_logcookies,
894 oti->oti_numcookies * sizeof(oti->oti_onecookie));
895 oti->oti_logcookies = NULL;
896 oti->oti_numcookies = 0;
900 * Events signalled through obd_notify() upcall-chain.
902 enum obd_notify_event {
903 /* Device activated */
905 /* Device deactivated */
907 /* Device disconnected */
909 /* Connect data for import were changed */
912 OBD_NOTIFY_SYNC_NONBLOCK,
914 /* Configuration event */
918 /* bit-mask flags for config events */
920 CONFIG_LOG = 0x1, /* finished processing config log */
921 CONFIG_SYNC = 0x2 /* mdt synced 1 ost */
925 * Data structure used to pass obd_notify()-event to non-obd listeners (llite
926 * and liblustre being main examples).
928 struct obd_notify_upcall {
929 int (*onu_upcall)(struct obd_device *host, struct obd_device *watched,
930 enum obd_notify_event ev, void *owner);
931 /* Opaque datum supplied by upper layer listener */
935 struct target_recovery_data {
936 svc_handler_t trd_recovery_handler;
937 pid_t trd_processing_task;
938 struct completion trd_starting;
939 struct completion trd_finishing;
943 FILTER_GROUP_LLOG = 1,
948 struct obd_llog_group {
949 struct list_head olg_list;
951 struct llog_ctxt *olg_ctxts[LLOG_MAX_CTXTS];
952 cfs_waitq_t olg_waitq;
954 struct obd_export *olg_exp;
955 int olg_initializing;
956 struct semaphore olg_cat_processing;
959 /* corresponds to one of the obd's */
960 #define MAX_OBD_NAME 128
961 #define OBD_DEVICE_MAGIC 0XAB5CD6EF
963 struct obd_type *obd_type;
966 /* common and UUID name of this device */
967 char obd_name[MAX_OBD_NAME];
968 struct obd_uuid obd_uuid;
970 struct lu_device *obd_lu_dev;
973 unsigned long obd_attached:1, /* finished attach */
974 obd_set_up:1, /* finished setup */
975 obd_recovering:1, /* there are recoverable clients */
976 obd_abort_recovery:1,/* somebody ioctl'ed us to abort */
977 obd_replayable:1, /* recovery is enabled; inform clients */
978 obd_no_transno:1, /* no committed-transno notification */
979 obd_no_recov:1, /* fail instead of retry messages */
980 obd_req_replaying:1, /* replaying requests */
981 obd_stopping:1, /* started cleanup */
982 obd_starting:1, /* started setup */
983 obd_force:1, /* cleanup with > 0 obd refcount */
984 obd_fail:1, /* cleanup with failover */
985 obd_async_recov:1, /* allow asyncronous orphan cleanup */
986 obd_no_conn:1, /* deny new connections */
987 obd_inactive:1; /* device active/inactive
988 * (for /proc/status only!!) */
989 /* uuid-export hash body */
990 struct lustre_hash *obd_uuid_hash;
991 /* nid-export hash body */
992 struct lustre_hash *obd_nid_hash;
994 struct lustre_hash *obd_nid_stats_hash;
995 struct list_head obd_nid_stats;
996 atomic_t obd_refcount;
997 cfs_waitq_t obd_refcount_waitq;
998 struct list_head obd_exports;
1000 spinlock_t obd_nid_lock;
1001 struct ldlm_namespace *obd_namespace;
1002 struct ptlrpc_client obd_ldlm_client; /* XXX OST/MDS only */
1003 /* a spinlock is OK for what we do now, may need a semaphore later */
1004 spinlock_t obd_dev_lock;
1005 struct semaphore obd_dev_sem;
1006 __u64 obd_last_committed;
1007 struct fsfilt_operations *obd_fsops;
1008 spinlock_t obd_osfs_lock;
1009 struct obd_statfs obd_osfs; /* locked by obd_osfs_lock */
1011 struct lvfs_run_ctxt obd_lvfs_ctxt;
1012 struct obd_llog_group obd_olg; /* default llog group */
1013 struct obd_device *obd_observer;
1014 struct obd_notify_upcall obd_upcall;
1015 struct obd_export *obd_self_export;
1016 /* list of exports in LRU order, for ping evictor, with obd_dev_lock */
1017 struct list_head obd_exports_timed;
1018 time_t obd_eviction_timer; /* for ping evictor */
1020 /* XXX encapsulate all this recovery data into one struct */
1021 svc_handler_t obd_recovery_handler;
1022 pid_t obd_processing_task;
1024 int obd_max_recoverable_clients;
1025 int obd_connected_clients;
1026 int obd_recoverable_clients;
1027 spinlock_t obd_processing_task_lock; /* BH lock (timer) */
1028 __u64 obd_next_recovery_transno;
1029 int obd_replayed_requests;
1030 int obd_requests_queued_for_recovery;
1031 cfs_waitq_t obd_next_transno_waitq;
1032 struct list_head obd_uncommitted_replies;
1033 spinlock_t obd_uncommitted_replies_lock;
1034 cfs_timer_t obd_recovery_timer;
1035 time_t obd_recovery_start; /* seconds */
1036 time_t obd_recovery_end; /* seconds, for lprocfs_status */
1037 time_t obd_recovery_max_time; /* seconds, bz13079 */
1038 int obd_recovery_timeout;
1040 /* new recovery stuff from CMD2 */
1041 struct target_recovery_data obd_recovery_data;
1042 int obd_replayed_locks;
1043 atomic_t obd_req_replay_clients;
1044 atomic_t obd_lock_replay_clients;
1045 struct list_head obd_req_replay_queue;
1046 struct list_head obd_lock_replay_queue;
1047 struct list_head obd_final_req_queue;
1048 int obd_recovery_stage;
1051 struct obd_device_target obt;
1052 struct filter_obd filter;
1054 struct client_obd cli;
1056 struct echo_client_obd echo_client;
1057 struct echo_obd echo;
1062 /* Fields used by LProcFS */
1063 unsigned int obd_cntr_base;
1064 struct lprocfs_stats *obd_stats;
1066 unsigned int md_cntr_base;
1067 struct lprocfs_stats *md_stats;
1069 cfs_proc_dir_entry_t *obd_proc_entry;
1070 cfs_proc_dir_entry_t *obd_proc_exports_entry;
1071 cfs_proc_dir_entry_t *obd_svc_procroot;
1072 struct lprocfs_stats *obd_svc_stats;
1073 atomic_t obd_evict_inprogress;
1074 cfs_waitq_t obd_evict_inprogress_waitq;
1077 * Ldlm pool part. Save last calculated SLV and Limit.
1079 rwlock_t obd_pool_lock;
1084 * A list of outstanding class_incref()'s against this obd. For
1087 struct lu_ref obd_reference;
1090 #define OBD_OPT_FORCE 0x0001
1091 #define OBD_OPT_FAILOVER 0x0002
1093 #define OBD_LLOG_FL_SENDNOW 0x0001
1095 enum obd_cleanup_stage {
1096 /* Special case hack for MDS LOVs */
1098 /* can be directly mapped to .ldto_device_fini() */
1099 OBD_CLEANUP_EXPORTS,
1102 /* get/set_info keys */
1103 #define KEY_READ_ONLY "read-only"
1104 #define KEY_MDS_CONN "mds_conn"
1105 #define KEY_NEXT_ID "next_id"
1106 #define KEY_LOVDESC "lovdesc"
1107 #define KEY_INIT_RECOV "initial_recov"
1108 #define KEY_INIT_RECOV_BACKUP "init_recov_bk"
1109 #define KEY_FLUSH_CTX "flush_ctx"
1110 #define KEY_CAPA_KEY "capa_key"
1111 #define KEY_CONN_DATA "conn_data"
1112 #define KEY_MAX_EASIZE "max_easize"
1113 #define KEY_REVIMP_UPD "revimp_update"
1114 #define KEY_LOV_IDX "lov_idx"
1115 #define KEY_LAST_ID "last_id"
1116 #define KEY_READONLY "read-only"
1117 #define KEY_LOCK_TO_STRIPE "lock_to_stripe"
1118 #define KEY_CHECKSUM "checksum"
1119 #define KEY_UNLINKED "unlinked"
1120 #define KEY_EVICT_BY_NID "evict_by_nid"
1121 #define KEY_REGISTER_TARGET "register_target"
1122 #define KEY_SET_FS "set_fs"
1123 #define KEY_CLEAR_FS "clear_fs"
1124 #define KEY_BLOCKSIZE "blocksize"
1125 #define KEY_BLOCKSIZE_BITS "blocksize_bits"
1126 #define KEY_FIEMAP "FIEMAP"
1128 #define KEY_INTERMDS "inter_mds"
1129 #define KEY_ASYNC "async"
1133 static inline int it_to_lock_mode(struct lookup_intent *it)
1135 /* CREAT needs to be tested before open (both could be set) */
1136 if (it->it_op & IT_CREAT)
1138 else if (it->it_op & (IT_READDIR | IT_GETATTR | IT_OPEN | IT_LOOKUP))
1141 LASSERTF(0, "Invalid it_op: %d\n", it->it_op);
1146 struct lu_fid op_fid1; /* operation fid1 (usualy parent) */
1147 struct lu_fid op_fid2; /* operation fid2 (usualy child) */
1148 struct lu_fid op_fid3; /* 2 extra fids to find conflicting */
1149 struct lu_fid op_fid4; /* to the operation locks. */
1150 mdsno_t op_mds; /* what mds server open will go to */
1151 struct lustre_handle op_handle;
1153 const char *op_name;
1156 struct lmv_stripe_md *op_mea1;
1157 struct lmv_stripe_md *op_mea2;
1158 __u32 op_suppgids[2];
1164 /* iattr fields and blocks. */
1165 struct iattr op_attr;
1167 #if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,14)
1168 unsigned int op_attr_flags;
1171 loff_t op_attr_blocks;
1173 /* Size-on-MDS epoch and flags. */
1178 struct obd_capa *op_capa1;
1179 struct obd_capa *op_capa2;
1181 /* Various operation flags. */
1184 /* Operation type */
1188 struct md_enqueue_info;
1189 /* metadata stat-ahead */
1190 typedef int (* md_enqueue_cb_t)(struct ptlrpc_request *req,
1191 struct md_enqueue_info *minfo,
1194 struct md_enqueue_info {
1195 struct md_op_data mi_data;
1196 struct lookup_intent mi_it;
1197 struct lustre_handle mi_lockh;
1198 struct dentry *mi_dentry;
1199 md_enqueue_cb_t mi_cb;
1200 unsigned int mi_generation;
1205 struct module *o_owner;
1206 int (*o_iocontrol)(unsigned int cmd, struct obd_export *exp, int len,
1207 void *karg, void *uarg);
1208 int (*o_get_info)(struct obd_export *, __u32 keylen, void *key,
1209 __u32 *vallen, void *val, struct lov_stripe_md *lsm);
1210 int (*o_set_info_async)(struct obd_export *, __u32 keylen, void *key,
1211 __u32 vallen, void *val,
1212 struct ptlrpc_request_set *set);
1213 int (*o_attach)(struct obd_device *dev, obd_count len, void *data);
1214 int (*o_detach)(struct obd_device *dev);
1215 int (*o_setup) (struct obd_device *dev, struct lustre_cfg *cfg);
1216 int (*o_precleanup)(struct obd_device *dev,
1217 enum obd_cleanup_stage cleanup_stage);
1218 int (*o_cleanup)(struct obd_device *dev);
1219 int (*o_process_config)(struct obd_device *dev, obd_count len,
1221 int (*o_postrecov)(struct obd_device *dev);
1222 int (*o_add_conn)(struct obd_import *imp, struct obd_uuid *uuid,
1224 int (*o_del_conn)(struct obd_import *imp, struct obd_uuid *uuid);
1225 /* connect to the target device with given connection
1226 * data. @ocd->ocd_connect_flags is modified to reflect flags actually
1227 * granted by the target, which are guaranteed to be a subset of flags
1228 * asked for. If @ocd == NULL, use default parameters. */
1229 int (*o_connect)(const struct lu_env *env,
1230 struct lustre_handle *conn, struct obd_device *src,
1231 struct obd_uuid *cluuid, struct obd_connect_data *ocd,
1233 int (*o_reconnect)(const struct lu_env *env,
1234 struct obd_export *exp, struct obd_device *src,
1235 struct obd_uuid *cluuid,
1236 struct obd_connect_data *ocd,
1238 int (*o_disconnect)(struct obd_export *exp);
1240 /* Initialize/finalize fids infrastructure. */
1241 int (*o_fid_init)(struct obd_export *exp);
1242 int (*o_fid_fini)(struct obd_export *exp);
1244 /* Allocate new fid according to passed @hint. */
1245 int (*o_fid_alloc)(struct obd_export *exp, struct lu_fid *fid,
1246 struct md_op_data *op_data);
1249 * Object with @fid is getting deleted, we may want to do something
1252 int (*o_fid_delete)(struct obd_export *exp, const struct lu_fid *fid);
1254 int (*o_statfs)(struct obd_device *obd, struct obd_statfs *osfs,
1255 __u64 max_age, __u32 flags);
1256 int (*o_statfs_async)(struct obd_device *obd, struct obd_info *oinfo,
1257 __u64 max_age, struct ptlrpc_request_set *set);
1258 int (*o_packmd)(struct obd_export *exp, struct lov_mds_md **disk_tgt,
1259 struct lov_stripe_md *mem_src);
1260 int (*o_unpackmd)(struct obd_export *exp,struct lov_stripe_md **mem_tgt,
1261 struct lov_mds_md *disk_src, int disk_len);
1262 int (*o_checkmd)(struct obd_export *exp, struct obd_export *md_exp,
1263 struct lov_stripe_md *mem_tgt);
1264 int (*o_preallocate)(struct lustre_handle *, obd_count *req,
1266 /* FIXME: add fid capability support for create & destroy! */
1267 int (*o_precreate)(struct obd_export *exp);
1268 int (*o_create)(struct obd_export *exp, struct obdo *oa,
1269 struct lov_stripe_md **ea, struct obd_trans_info *oti);
1270 int (*o_destroy)(struct obd_export *exp, struct obdo *oa,
1271 struct lov_stripe_md *ea, struct obd_trans_info *oti,
1272 struct obd_export *md_exp);
1273 int (*o_setattr)(struct obd_export *exp, struct obd_info *oinfo,
1274 struct obd_trans_info *oti);
1275 int (*o_setattr_async)(struct obd_export *exp, struct obd_info *oinfo,
1276 struct obd_trans_info *oti,
1277 struct ptlrpc_request_set *rqset);
1278 int (*o_getattr)(struct obd_export *exp, struct obd_info *oinfo);
1279 int (*o_getattr_async)(struct obd_export *exp, struct obd_info *oinfo,
1280 struct ptlrpc_request_set *set);
1281 int (*o_brw)(int rw, struct obd_export *exp, struct obd_info *oinfo,
1282 obd_count oa_bufs, struct brw_page *pgarr,
1283 struct obd_trans_info *oti);
1284 int (*o_brw_async)(int rw, struct obd_export *exp,
1285 struct obd_info *oinfo, obd_count oa_bufs,
1286 struct brw_page *pgarr, struct obd_trans_info *oti,
1287 struct ptlrpc_request_set *);
1288 int (*o_prep_async_page)(struct obd_export *exp,
1289 struct lov_stripe_md *lsm,
1290 struct lov_oinfo *loi,
1291 cfs_page_t *page, obd_off offset,
1292 struct obd_async_page_ops *ops, void *data,
1293 void **res, int nocache,
1294 struct lustre_handle *lockh);
1295 int (*o_reget_short_lock)(struct obd_export *exp,
1296 struct lov_stripe_md *lsm,
1298 obd_off start, obd_off end,
1300 int (*o_release_short_lock)(struct obd_export *exp,
1301 struct lov_stripe_md *lsm, obd_off end,
1302 void *cookie, int rw);
1303 int (*o_queue_async_io)(struct obd_export *exp,
1304 struct lov_stripe_md *lsm,
1305 struct lov_oinfo *loi, void *cookie,
1306 int cmd, obd_off off, int count,
1307 obd_flag brw_flags, obd_flag async_flags);
1308 int (*o_queue_group_io)(struct obd_export *exp,
1309 struct lov_stripe_md *lsm,
1310 struct lov_oinfo *loi,
1311 struct obd_io_group *oig,
1312 void *cookie, int cmd, obd_off off, int count,
1313 obd_flag brw_flags, obd_flag async_flags);
1314 int (*o_trigger_group_io)(struct obd_export *exp,
1315 struct lov_stripe_md *lsm,
1316 struct lov_oinfo *loi,
1317 struct obd_io_group *oig);
1318 int (*o_set_async_flags)(struct obd_export *exp,
1319 struct lov_stripe_md *lsm,
1320 struct lov_oinfo *loi, void *cookie,
1321 obd_flag async_flags);
1322 int (*o_teardown_async_page)(struct obd_export *exp,
1323 struct lov_stripe_md *lsm,
1324 struct lov_oinfo *loi, void *cookie);
1325 int (*o_merge_lvb)(struct obd_export *exp, struct lov_stripe_md *lsm,
1326 struct ost_lvb *lvb, int kms_only);
1327 int (*o_adjust_kms)(struct obd_export *exp, struct lov_stripe_md *lsm,
1328 obd_off size, int shrink);
1329 int (*o_punch)(struct obd_export *exp, struct obd_info *oinfo,
1330 struct obd_trans_info *oti,
1331 struct ptlrpc_request_set *rqset);
1332 int (*o_sync)(struct obd_export *exp, struct obdo *oa,
1333 struct lov_stripe_md *ea, obd_size start, obd_size end,
1335 int (*o_migrate)(struct lustre_handle *conn, struct lov_stripe_md *dst,
1336 struct lov_stripe_md *src, obd_size start,
1337 obd_size end, struct obd_trans_info *oti);
1338 int (*o_copy)(struct lustre_handle *dstconn, struct lov_stripe_md *dst,
1339 struct lustre_handle *srconn, struct lov_stripe_md *src,
1340 obd_size start, obd_size end, struct obd_trans_info *);
1341 int (*o_iterate)(struct lustre_handle *conn,
1342 int (*)(obd_id, obd_gr, void *),
1343 obd_id *startid, obd_gr group, void *data);
1344 int (*o_preprw)(int cmd, struct obd_export *exp, struct obdo *oa,
1345 int objcount, struct obd_ioobj *obj,
1346 struct niobuf_remote *remote, int *nr_pages,
1347 struct niobuf_local *local,
1348 struct obd_trans_info *oti,
1349 struct lustre_capa *capa);
1350 int (*o_commitrw)(int cmd, struct obd_export *exp, struct obdo *oa,
1351 int objcount, struct obd_ioobj *obj,
1352 struct niobuf_remote *remote, int pages,
1353 struct niobuf_local *local,
1354 struct obd_trans_info *oti, int rc);
1355 int (*o_enqueue)(struct obd_export *, struct obd_info *oinfo,
1356 struct ldlm_enqueue_info *einfo,
1357 struct ptlrpc_request_set *rqset);
1358 int (*o_match)(struct obd_export *, struct lov_stripe_md *, __u32 type,
1359 ldlm_policy_data_t *, __u32 mode, int *flags, void *data,
1360 struct lustre_handle *lockh);
1361 int (*o_change_cbdata)(struct obd_export *, struct lov_stripe_md *,
1362 ldlm_iterator_t it, void *data);
1363 int (*o_cancel)(struct obd_export *, struct lov_stripe_md *md,
1364 __u32 mode, struct lustre_handle *);
1365 int (*o_cancel_unused)(struct obd_export *, struct lov_stripe_md *,
1366 int flags, void *opaque);
1367 int (*o_init_export)(struct obd_export *exp);
1368 int (*o_destroy_export)(struct obd_export *exp);
1369 int (*o_extent_calc)(struct obd_export *, struct lov_stripe_md *,
1370 int cmd, obd_off *);
1372 /* llog related obd_methods */
1373 int (*o_llog_init)(struct obd_device *obd, struct obd_llog_group *grp,
1374 struct obd_device *disk_obd, int count,
1375 struct llog_catid *logid, struct obd_uuid *uuid);
1376 int (*o_llog_finish)(struct obd_device *obd, int count);
1377 int (*o_llog_connect)(struct obd_export *, struct llogd_conn_body *);
1379 /* metadata-only methods */
1380 int (*o_pin)(struct obd_export *, const struct lu_fid *fid,
1381 struct obd_capa *, struct obd_client_handle *, int flag);
1382 int (*o_unpin)(struct obd_export *, struct obd_client_handle *, int);
1384 int (*o_import_event)(struct obd_device *, struct obd_import *,
1385 enum obd_import_event);
1387 int (*o_notify)(struct obd_device *obd, struct obd_device *watched,
1388 enum obd_notify_event ev, void *data);
1390 int (*o_health_check)(struct obd_device *);
1391 struct obd_uuid *(*o_get_uuid) (struct obd_export *exp);
1394 int (*o_quotacheck)(struct obd_export *, struct obd_quotactl *);
1395 int (*o_quotactl)(struct obd_export *, struct obd_quotactl *);
1397 int (*o_ping)(struct obd_export *exp);
1399 int (*o_register_page_removal_cb)(struct obd_export *exp,
1400 obd_page_removal_cb_t cb,
1401 obd_pin_extent_cb pin_cb);
1402 int (*o_unregister_page_removal_cb)(struct obd_export *exp,
1403 obd_page_removal_cb_t cb);
1404 int (*o_register_lock_cancel_cb)(struct obd_export *exp,
1405 obd_lock_cancel_cb cb);
1406 int (*o_unregister_lock_cancel_cb)(struct obd_export *exp,
1407 obd_lock_cancel_cb cb);
1409 int (*o_pool_new)(struct obd_device *obd, char *poolname);
1410 int (*o_pool_del)(struct obd_device *obd, char *poolname);
1411 int (*o_pool_add)(struct obd_device *obd, char *poolname,
1413 int (*o_pool_rem)(struct obd_device *obd, char *poolname,
1416 * NOTE: If adding ops, add another LPROCFS_OBD_OP_INIT() line
1417 * to lprocfs_alloc_obd_stats() in obdclass/lprocfs_status.c.
1418 * Also, add a wrapper function in include/linux/obd_class.h. */
1421 /* TODO: lmv_stripe_md should contain mds capabilities for all slave fids */
1422 struct lmv_stripe_md {
1427 struct lu_fid mea_ids[0];
1431 LUSTRE_OPC_MKDIR = (1 << 0),
1432 LUSTRE_OPC_SYMLINK = (1 << 1),
1433 LUSTRE_OPC_MKNOD = (1 << 2),
1434 LUSTRE_OPC_CREATE = (1 << 3),
1435 LUSTRE_OPC_ANY = (1 << 4)
1438 /* lmv structures */
1439 #define MEA_MAGIC_LAST_CHAR 0xb2221ca1
1440 #define MEA_MAGIC_ALL_CHARS 0xb222a11c
1441 #define MEA_MAGIC_HASH_SEGMENT 0xb222a11b
1443 #define MAX_HASH_SIZE_32 0x7fffffffUL
1444 #define MAX_HASH_SIZE 0x7fffffffffffffffULL
1445 #define MAX_HASH_HIGHEST_BIT 0x1000000000000000ULL
1448 struct mdt_body *body;
1449 struct lov_stripe_md *lsm;
1450 struct lmv_stripe_md *mea;
1451 #ifdef CONFIG_FS_POSIX_ACL
1452 struct posix_acl *posix_acl;
1454 struct mdt_remote_perm *remote_perm;
1455 struct obd_capa *mds_capa;
1456 struct obd_capa *oss_capa;
1459 struct md_open_data {
1460 struct obd_client_handle *mod_och;
1461 struct list_head mod_replay_list;
1464 struct lookup_intent;
1467 int (*m_getstatus)(struct obd_export *, struct lu_fid *,
1468 struct obd_capa **);
1469 int (*m_change_cbdata)(struct obd_export *, const struct lu_fid *,
1470 ldlm_iterator_t, void *);
1471 int (*m_close)(struct obd_export *, struct md_op_data *,
1472 struct md_open_data *, struct ptlrpc_request **);
1473 int (*m_create)(struct obd_export *, struct md_op_data *,
1474 const void *, int, int, __u32, __u32, cfs_cap_t,
1475 __u64, struct ptlrpc_request **);
1476 int (*m_done_writing)(struct obd_export *, struct md_op_data *,
1477 struct md_open_data *);
1478 int (*m_enqueue)(struct obd_export *, struct ldlm_enqueue_info *,
1479 struct lookup_intent *, struct md_op_data *,
1480 struct lustre_handle *, void *, int,
1481 struct ptlrpc_request **, int);
1482 int (*m_getattr)(struct obd_export *, const struct lu_fid *,
1483 struct obd_capa *, obd_valid, int,
1484 struct ptlrpc_request **);
1485 int (*m_getattr_name)(struct obd_export *, const struct lu_fid *,
1486 struct obd_capa *, const char *, int, obd_valid,
1487 int, __u32, struct ptlrpc_request **);
1488 int (*m_intent_lock)(struct obd_export *, struct md_op_data *,
1489 void *, int, struct lookup_intent *, int,
1490 struct ptlrpc_request **,
1491 ldlm_blocking_callback, int);
1492 int (*m_link)(struct obd_export *, struct md_op_data *,
1493 struct ptlrpc_request **);
1494 int (*m_rename)(struct obd_export *, struct md_op_data *,
1495 const char *, int, const char *, int,
1496 struct ptlrpc_request **);
1497 int (*m_is_subdir)(struct obd_export *, const struct lu_fid *,
1498 const struct lu_fid *,
1499 struct ptlrpc_request **);
1500 int (*m_setattr)(struct obd_export *, struct md_op_data *, void *,
1501 int , void *, int, struct ptlrpc_request **,
1502 struct md_open_data **mod);
1503 int (*m_sync)(struct obd_export *, const struct lu_fid *,
1504 struct obd_capa *, struct ptlrpc_request **);
1505 int (*m_readpage)(struct obd_export *, const struct lu_fid *,
1506 struct obd_capa *, __u64, struct page *,
1507 struct ptlrpc_request **);
1509 int (*m_unlink)(struct obd_export *, struct md_op_data *,
1510 struct ptlrpc_request **);
1512 int (*m_setxattr)(struct obd_export *, const struct lu_fid *,
1513 struct obd_capa *, obd_valid, const char *,
1514 const char *, int, int, int, __u32,
1515 struct ptlrpc_request **);
1517 int (*m_getxattr)(struct obd_export *, const struct lu_fid *,
1518 struct obd_capa *, obd_valid, const char *,
1519 const char *, int, int, int,
1520 struct ptlrpc_request **);
1522 int (*m_init_ea_size)(struct obd_export *, int, int, int);
1524 int (*m_get_lustre_md)(struct obd_export *, struct ptlrpc_request *,
1525 struct obd_export *, struct obd_export *,
1526 struct lustre_md *);
1528 int (*m_free_lustre_md)(struct obd_export *, struct lustre_md *);
1530 int (*m_set_open_replay_data)(struct obd_export *,
1531 struct obd_client_handle *,
1532 struct ptlrpc_request *);
1533 int (*m_clear_open_replay_data)(struct obd_export *,
1534 struct obd_client_handle *);
1535 int (*m_set_lock_data)(struct obd_export *, __u64 *, void *);
1537 ldlm_mode_t (*m_lock_match)(struct obd_export *, int,
1538 const struct lu_fid *, ldlm_type_t,
1539 ldlm_policy_data_t *, ldlm_mode_t,
1540 struct lustre_handle *);
1542 int (*m_cancel_unused)(struct obd_export *, const struct lu_fid *,
1543 ldlm_policy_data_t *, ldlm_mode_t, int flags,
1545 int (*m_renew_capa)(struct obd_export *, struct obd_capa *oc,
1546 renew_capa_cb_t cb);
1548 int (*m_get_remote_perm)(struct obd_export *, const struct lu_fid *,
1549 struct obd_capa *, __u32,
1550 struct ptlrpc_request **);
1552 int (*m_intent_getattr_async)(struct obd_export *,
1553 struct md_enqueue_info *,
1554 struct ldlm_enqueue_info *);
1556 int (*m_revalidate_lock)(struct obd_export *,
1557 struct lookup_intent *,
1561 * NOTE: If adding ops, add another LPROCFS_MD_OP_INIT() line to
1562 * lprocfs_alloc_md_stats() in obdclass/lprocfs_status.c. Also, add a
1563 * wrapper function in include/linux/obd_class.h.
1567 struct lsm_operations {
1568 void (*lsm_free)(struct lov_stripe_md *);
1569 int (*lsm_destroy)(struct lov_stripe_md *, struct obdo *oa,
1570 struct obd_export *md_exp);
1571 void (*lsm_stripe_by_index)(struct lov_stripe_md *, int *, obd_off *,
1573 void (*lsm_stripe_by_offset)(struct lov_stripe_md *, int *, obd_off *,
1575 obd_off (*lsm_stripe_offset_by_index)(struct lov_stripe_md *, int);
1576 obd_off (*lsm_stripe_offset_by_offset)(struct lov_stripe_md *, obd_off);
1577 int (*lsm_stripe_index_by_offset)(struct lov_stripe_md *, obd_off);
1578 int (*lsm_revalidate) (struct lov_stripe_md *, struct obd_device *obd);
1579 int (*lsm_lmm_verify) (struct lov_mds_md *lmm, int lmm_bytes,
1581 int (*lsm_unpackmd) (struct lov_obd *lov, struct lov_stripe_md *lsm,
1582 struct lov_mds_md *lmm);
1585 extern const struct lsm_operations lsm_v1_ops;
1586 extern const struct lsm_operations lsm_join_ops;
1587 extern const struct lsm_operations lsm_v3_ops;
1588 static inline const struct lsm_operations *lsm_op_find(int magic)
1593 case LOV_MAGIC_JOIN:
1594 return &lsm_join_ops;
1598 CERROR("Cannot recognize lsm_magic %d\n", magic);
1603 int lvfs_check_io_health(struct obd_device *obd, struct file *file);
1605 /* Requests for obd_extent_calc() */
1606 #define OBD_CALC_STRIPE_START 1
1607 #define OBD_CALC_STRIPE_END 2
1609 static inline void obd_transno_commit_cb(struct obd_device *obd, __u64 transno,
1613 CERROR("%s: transno "LPU64" commit error: %d\n",
1614 obd->obd_name, transno, error);
1617 if (transno > obd->obd_last_committed) {
1618 CDEBUG(D_HA, "%s: transno "LPD64" committed\n",
1619 obd->obd_name, transno);
1620 obd->obd_last_committed = transno;
1621 ptlrpc_commit_replies (obd);
1623 CDEBUG(D_INFO, "%s: transno "LPD64" committed\n",
1624 obd->obd_name, transno);
1628 static inline void init_obd_quota_ops(quota_interface_t *interface,
1629 struct obd_ops *obd_ops)
1635 obd_ops->o_quotacheck = QUOTA_OP(interface, check);
1636 obd_ops->o_quotactl = QUOTA_OP(interface, ctl);
1639 static inline __u64 oinfo_mdsno(struct obd_info *oinfo)
1641 return oinfo->oi_oa->o_gr - FILTER_GROUP_MDS0;
1644 static inline struct lustre_capa *oinfo_capa(struct obd_info *oinfo)
1646 return oinfo->oi_capa;
1649 #endif /* __OBD_H */