1 /* -*- mode: c; c-basic-offset: 8; indent-tabs-mode: nil; -*-
2 * vim:expandtab:shiftwidth=8:tabstop=8:
6 * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
8 * This program is free software; you can redistribute it and/or modify
9 * it under the terms of the GNU General Public License version 2 only,
10 * as published by the Free Software Foundation.
12 * This program is distributed in the hope that it will be useful, but
13 * WITHOUT ANY WARRANTY; without even the implied warranty of
14 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
15 * General Public License version 2 for more details (a copy is included
16 * in the LICENSE file that accompanied this code).
18 * You should have received a copy of the GNU General Public License
19 * version 2 along with this program; If not, see
20 * http://www.sun.com/software/products/lustre/docs/GPLv2.pdf
22 * Please contact Sun Microsystems, Inc., 4150 Network Circle, Santa Clara,
23 * CA 95054 USA or visit www.sun.com if you need additional information or
29 * Copyright 2008 Sun Microsystems, Inc. All rights reserved
30 * Use is subject to license terms.
33 * This file is part of Lustre, http://www.lustre.org/
34 * Lustre is a trademark of Sun Microsystems, Inc.
40 #if defined(__linux__)
41 #include <linux/obd.h>
42 #elif defined(__APPLE__)
43 #include <darwin/obd.h>
44 #elif defined(__WINNT__)
45 #include <winnt/obd.h>
47 #error Unsupported operating system.
50 #define IOC_OSC_TYPE 'h'
51 #define IOC_OSC_MIN_NR 20
52 #define IOC_OSC_SET_ACTIVE _IOWR(IOC_OSC_TYPE, 21, struct obd_device *)
53 #define IOC_OSC_MAX_NR 50
55 #define IOC_MDC_TYPE 'i'
56 #define IOC_MDC_MIN_NR 20
57 /* Moved to lustre_user.h
58 #define IOC_MDC_LOOKUP _IOWR(IOC_MDC_TYPE, 20, struct obd_ioctl_data *)
59 #define IOC_MDC_GETSTRIPE _IOWR(IOC_MDC_TYPE, 21, struct lov_mds_md *) */
60 #define IOC_MDC_MAX_NR 50
62 #include <lustre/lustre_idl.h>
63 #include <lu_target.h>
65 #include <lustre_lib.h>
66 #include <lustre_export.h>
67 #include <lustre_quota.h>
68 #include <lustre_fld.h>
69 #include <lustre_capa.h>
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 struct list_head loi_ready_item;
100 struct list_head loi_hp_ready_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_ready_item);
125 CFS_INIT_LIST_HEAD(&loi->loi_hp_ready_item);
126 CFS_INIT_LIST_HEAD(&loi->loi_write_item);
127 CFS_INIT_LIST_HEAD(&loi->loi_read_item);
130 struct lov_stripe_md {
132 pid_t lsm_lock_owner; /* debugging */
135 /* Public members. */
136 __u64 lw_object_id; /* lov object id */
137 __u64 lw_object_gr; /* lov object group */
138 __u64 lw_maxbytes; /* maximum possible file size */
140 /* LOV-private members start here -- only for use in lov/. */
142 __u32 lw_stripe_size; /* size of the stripe */
143 __u32 lw_pattern; /* striping pattern (RAID0, RAID1) */
144 unsigned lw_stripe_count; /* number of objects being striped over */
145 char lw_pool_name[LOV_MAXPOOLNAME]; /* pool name */
148 struct lov_oinfo *lsm_oinfo[0];
151 #define lsm_object_id lsm_wire.lw_object_id
152 #define lsm_object_gr lsm_wire.lw_object_gr
153 #define lsm_maxbytes lsm_wire.lw_maxbytes
154 #define lsm_magic lsm_wire.lw_magic
155 #define lsm_stripe_size lsm_wire.lw_stripe_size
156 #define lsm_pattern lsm_wire.lw_pattern
157 #define lsm_stripe_count lsm_wire.lw_stripe_count
158 #define lsm_pool_name lsm_wire.lw_pool_name
162 typedef int (*obd_enqueue_update_f)(void *cookie, int rc);
164 /* obd info for a particular level (lov, osc). */
166 /* Lock policy. It keeps an extent which is specific for a particular
167 * OSC. (e.g. lov_prep_enqueue_set initialises extent of the policy,
168 * and osc_enqueue passes it into ldlm_lock_match & ldlm_cli_enqueue. */
169 ldlm_policy_data_t oi_policy;
170 /* Flags used for set request specific flags:
171 - while lock handling, the flags obtained on the enqueue
172 request are set here.
173 - while stats, the flags used for control delay/resend.
176 /* Lock handle specific for every OSC lock. */
177 struct lustre_handle *oi_lockh;
178 /* lsm data specific for every OSC. */
179 struct lov_stripe_md *oi_md;
180 /* obdo data specific for every OSC, if needed at all. */
182 /* statfs data specific for every OSC, if needed at all. */
183 struct obd_statfs *oi_osfs;
184 /* An update callback which is called to update some data on upper
185 * level. E.g. it is used for update lsm->lsm_oinfo at every recieved
186 * request in osc level for enqueue requests. It is also possible to
187 * update some caller data from LOV layer if needed. */
188 obd_enqueue_update_f oi_cb_up;
189 /* oss capability, its type is obd_capa in client to avoid copy.
190 * in contrary its type is lustre_capa in OSS. */
194 /* compare all relevant fields. */
195 static inline int lov_stripe_md_cmp(struct lov_stripe_md *m1,
196 struct lov_stripe_md *m2)
199 * ->lsm_wire contains padding, but it should be zeroed out during
202 return memcmp(&m1->lsm_wire, &m2->lsm_wire, sizeof m1->lsm_wire);
205 void lov_stripe_lock(struct lov_stripe_md *md);
206 void lov_stripe_unlock(struct lov_stripe_md *md);
209 struct list_head typ_chain;
210 struct obd_ops *typ_dt_ops;
211 struct md_ops *typ_md_ops;
212 cfs_proc_dir_entry_t *typ_procroot;
215 struct lu_device_type *typ_lu;
216 spinlock_t obd_type_lock;
226 /* Individual type definitions */
228 struct ost_server_data;
230 /* hold common fields for "target" device */
231 struct obd_device_target {
232 struct super_block *obt_sb;
233 /** last_rcvd file */
234 struct file *obt_rcvd_filp;
235 /** server data in last_rcvd file */
236 struct lr_server_data *obt_lsd;
237 /** Lock protecting client bitmap */
238 spinlock_t obt_client_bitmap_lock;
239 /** Bitmap of known clients */
240 unsigned long *obt_client_bitmap;
241 /** Server last transaction number */
242 __u64 obt_last_transno;
243 /** Lock protecting last transaction number */
244 spinlock_t obt_translock;
245 /** Number of mounts */
246 __u64 obt_mount_count;
247 struct semaphore obt_quotachecking;
248 struct lustre_quota_ctxt obt_qctxt;
249 lustre_quota_version_t obt_qfmt;
250 struct rw_semaphore obt_rwsem;
251 struct vfsmount *obt_vfsmnt;
252 struct file *obt_health_check_filp;
257 LLOG_CONFIG_ORIG_CTXT = 0,
258 LLOG_CONFIG_REPL_CTXT,
259 LLOG_MDS_OST_ORIG_CTXT,
260 LLOG_MDS_OST_REPL_CTXT,
267 LLOG_LOVEA_ORIG_CTXT,
268 LLOG_LOVEA_REPL_CTXT,
269 LLOG_CHANGELOG_ORIG_CTXT, /**< changelog generation on mdd */
270 LLOG_CHANGELOG_REPL_CTXT, /**< changelog access on clients */
271 LLOG_CHANGELOG_USER_ORIG_CTXT, /**< for multiple changelog consumers */
275 #define FILTER_SUBDIR_COUNT 32 /* set to zero for no subdirs */
277 struct filter_subdirs {
278 cfs_dentry_t *dentry[FILTER_SUBDIR_COUNT];
288 /* NB this field MUST be first */
289 struct obd_device_target fo_obt;
290 struct lu_target fo_lut;
291 const char *fo_fstype;
294 cfs_dentry_t *fo_dentry_O;
295 cfs_dentry_t **fo_dentry_O_groups;
296 struct filter_subdirs *fo_dentry_O_sub;
297 struct semaphore fo_init_lock; /* group initialization lock */
298 int fo_committed_group;
300 #define CLIENT_QUOTA_DEFAULT_RESENDS 10
302 spinlock_t fo_objidlock; /* protect fo_lastobjid */
304 unsigned long fo_destroys_in_progress;
305 struct semaphore fo_create_locks[FILTER_SUBDIR_COUNT];
307 struct list_head fo_export_list;
310 obd_size fo_tot_dirty; /* protected by obd_osfs_lock */
311 obd_size fo_tot_granted; /* all values in bytes */
312 obd_size fo_tot_pending;
313 int fo_tot_granted_clients;
315 obd_size fo_readcache_max_filesize;
316 int fo_read_cache:1, /**< enable read-only cache */
317 fo_writethrough_cache:1,/**< read cache writes */
318 fo_mds_ost_sync:1, /**< MDS-OST orphan recovery*/
319 fo_raid_degraded:1;/**< RAID device degraded */
321 struct obd_import *fo_mdc_imp;
322 struct obd_uuid fo_mdc_uuid;
323 struct lustre_handle fo_mdc_conn;
324 struct file **fo_last_objid_files;
325 __u64 *fo_last_objids; /* last created objid for groups,
326 * protected by fo_objidlock */
328 struct semaphore fo_alloc_lock;
330 atomic_t fo_r_in_flight;
331 atomic_t fo_w_in_flight;
334 * per-filter pool of kiobuf's allocated by filter_common_setup() and
335 * torn down by filter_cleanup(). Contains OST_NUM_THREADS elements of
336 * which ->fo_iobuf_count were allocated.
338 * This pool contains kiobuf used by
339 * filter_{prep,commit}rw_{read,write}() and is shared by all OST
342 * Locking: none, each OST thread uses only one element, determined by
343 * its "ordinal number", ->t_id.
345 struct filter_iobuf **fo_iobuf_pool;
348 struct list_head fo_llog_list;
349 spinlock_t fo_llog_list_lock;
351 struct brw_stats fo_filter_stats;
352 struct lustre_quota_ctxt fo_quota_ctxt;
353 spinlock_t fo_quotacheck_lock;
354 atomic_t fo_quotachecking;
356 int fo_fmd_max_num; /* per exp filter_mod_data */
357 int fo_fmd_max_age; /* jiffies to fmd expiry */
360 rwlock_t fo_sptlrpc_lock;
361 struct sptlrpc_rule_set fo_sptlrpc_rset;
363 /* capability related */
364 unsigned int fo_fl_oss_capa;
365 struct list_head fo_capa_keys;
366 struct hlist_head *fo_capa_hash;
367 struct llog_commit_master *fo_lcm;
371 #define fo_translock fo_obt.obt_translock
372 #define fo_rcvd_filp fo_obt.obt_rcvd_filp
373 #define fo_fsd fo_obt.obt_lsd
374 #define fo_last_rcvd_slots fo_obt.obt_client_bitmap
375 #define fo_mount_count fo_obt.obt_mount_count
376 #define fo_vfsmnt fo_obt.obt_vfsmnt
378 struct timeout_item {
379 enum timeout_event ti_event;
380 cfs_time_t ti_timeout;
383 struct list_head ti_obd_list;
384 struct list_head ti_chain;
387 #define OSC_MAX_RIF_DEFAULT 8
388 #define OSC_MAX_RIF_MAX 256
389 #define OSC_MAX_DIRTY_DEFAULT (OSC_MAX_RIF_DEFAULT * 4)
390 #define OSC_MAX_DIRTY_MB_MAX 2048 /* arbitrary, but < MAX_LONG bytes */
391 #define OSC_DEFAULT_RESENDS 10
393 #define MDC_MAX_RIF_DEFAULT 8
394 #define MDC_MAX_RIF_MAX 512
399 struct rw_semaphore cl_sem;
400 struct obd_uuid cl_target_uuid;
401 struct obd_import *cl_import; /* ptlrpc connection state */
403 /* max_mds_easize is purely a performance thing so we don't have to
404 * call obd_size_diskmd() all the time. */
405 int cl_default_mds_easize;
406 int cl_max_mds_easize;
407 int cl_max_mds_cookiesize;
409 enum lustre_sec_part cl_sp_me;
410 enum lustre_sec_part cl_sp_to;
411 struct sptlrpc_flavor cl_flvr_mgc; /* fixed flavor of mgc->mgs */
413 //struct llog_canceld_ctxt *cl_llcd; /* it's included by obd_llog_ctxt */
414 void *cl_llcd_offset;
416 /* the grant values are protected by loi_list_lock below */
417 long cl_dirty; /* all _dirty_ in bytes */
418 long cl_dirty_max; /* allowed w/o rpc */
419 long cl_dirty_transit; /* dirty synchronous */
420 long cl_avail_grant; /* bytes of credit for ost */
421 long cl_lost_grant; /* lost credits (trunc) */
422 struct list_head cl_cache_waiters; /* waiting for cache/grant */
423 cfs_time_t cl_next_shrink_grant; /* jiffies */
424 struct list_head cl_grant_shrink_list; /* Timeout event list */
425 struct semaphore cl_grant_sem; /*grant shrink list semaphore*/
426 int cl_grant_shrink_interval; /* seconds */
428 /* keep track of objects that have lois that contain pages which
429 * have been queued for async brw. this lock also protects the
430 * lists of osc_client_pages that hang off of the loi */
432 * ->cl_loi_list_lock protects consistency of
433 * ->cl_loi_{ready,read,write}_list. ->ap_make_ready() and
434 * ->ap_completion() call-backs are executed under this lock. As we
435 * cannot guarantee that these call-backs never block on all platforms
436 * (as a matter of fact they do block on Mac OS X), type of
437 * ->cl_loi_list_lock is platform dependent: it's a spin-lock on Linux
438 * and blocking mutex on Mac OS X. (Alternative is to make this lock
439 * blocking everywhere, but we don't want to slow down fast-path of
440 * our main platform.)
442 * Exact type of ->cl_loi_list_lock is defined in arch/obd.h together
443 * with client_obd_list_{un,}lock() and
444 * client_obd_list_lock_{init,done}() functions.
446 client_obd_lock_t cl_loi_list_lock;
447 struct list_head cl_loi_ready_list;
448 struct list_head cl_loi_hp_ready_list;
449 struct list_head cl_loi_write_list;
450 struct list_head cl_loi_read_list;
453 /* just a sum of the loi/lop pending numbers to be exported by /proc */
454 int cl_pending_w_pages;
455 int cl_pending_r_pages;
456 int cl_max_pages_per_rpc;
457 int cl_max_rpcs_in_flight;
458 struct obd_histogram cl_read_rpc_hist;
459 struct obd_histogram cl_write_rpc_hist;
460 struct obd_histogram cl_read_page_hist;
461 struct obd_histogram cl_write_page_hist;
462 struct obd_histogram cl_read_offset_hist;
463 struct obd_histogram cl_write_offset_hist;
465 /* number of in flight destroy rpcs is limited to max_rpcs_in_flight */
466 atomic_t cl_destroy_in_flight;
467 cfs_waitq_t cl_destroy_waitq;
469 struct mdc_rpc_lock *cl_rpc_lock;
470 struct mdc_rpc_lock *cl_setattr_lock;
471 struct mdc_rpc_lock *cl_close_lock;
472 struct osc_creator cl_oscc;
475 struct semaphore cl_mgc_sem;
476 struct vfsmount *cl_mgc_vfsmnt;
477 struct dentry *cl_mgc_configs_dir;
478 atomic_t cl_mgc_refcount;
479 struct obd_export *cl_mgc_mgsexp;
481 /* checksumming for data sent over the network */
482 unsigned int cl_checksum:1; /* 0 = disabled, 1 = enabled */
483 /* supported checksum types that are worked out at connect time */
484 __u32 cl_supp_cksum_types;
485 /* checksum algorithm to be used */
486 cksum_type_t cl_cksum_type;
488 /* also protected by the poorly named _loi_list_lock lock above */
489 struct osc_async_rc cl_ar;
491 /* used by quotacheck */
492 int cl_qchk_stat; /* quotacheck stat of the peer */
494 /* sequence manager */
495 struct lu_client_seq *cl_seq;
497 atomic_t cl_resends; /* resend count */
498 atomic_t cl_quota_resends; /* quota related resend count */
500 #define obd2cli_tgt(obd) ((char *)(obd)->u.cli.cl_target_uuid.uuid)
502 #define CL_NOT_QUOTACHECKED 1 /* client->cl_qchk_stat init value */
505 struct ptlrpc_service *mgs_service;
506 struct vfsmount *mgs_vfsmnt;
507 struct super_block *mgs_sb;
508 struct dentry *mgs_configs_dir;
509 struct dentry *mgs_fid_de;
510 struct list_head mgs_fs_db_list;
511 struct semaphore mgs_sem;
512 cfs_proc_dir_entry_t *mgs_proc_live;
516 /* NB this field MUST be first */
517 struct obd_device_target mds_obt;
518 struct ptlrpc_service *mds_service;
519 struct ptlrpc_service *mds_setattr_service;
520 struct ptlrpc_service *mds_readpage_service;
521 cfs_dentry_t *mds_fid_de;
523 int mds_max_cookiesize;
525 unsigned long mds_atime_diff;
526 struct semaphore mds_epoch_sem;
527 struct ll_fid mds_rootfid;
528 cfs_dentry_t *mds_pending_dir;
529 cfs_dentry_t *mds_logs_dir;
530 cfs_dentry_t *mds_objects_dir;
531 struct llog_handle *mds_cfg_llh;
532 struct obd_device *mds_osc_obd; /* XXX lov_obd */
533 struct obd_uuid mds_lov_uuid;
535 struct obd_export *mds_osc_exp; /* XXX lov_exp */
536 struct lov_desc mds_lov_desc;
539 /* mark pages dirty for write. */
540 bitmap_t *mds_lov_page_dirty;
541 /* array for store pages with obd_id */
542 void **mds_lov_page_array;
543 /* file for store objid */
544 struct file *mds_lov_objid_filp;
545 __u32 mds_lov_objid_count;
546 __u32 mds_lov_objid_max_index;
547 __u32 mds_lov_objid_lastpage;
548 __u32 mds_lov_objid_lastidx;
551 struct lustre_quota_info mds_quota_info;
552 struct semaphore mds_qonoff_sem;
553 struct semaphore mds_health_sem;
554 unsigned long mds_fl_user_xattr:1,
556 mds_evict_ost_nids:1,
560 mds_fl_target:1; /* mds have one or
563 struct upcall_cache *mds_identity_cache;
565 /* for capability keys update */
566 struct lustre_capa_key *mds_capa_keys;
567 struct rw_semaphore mds_notify_lock;
570 #define mds_transno_lock mds_obt.obt_translock
571 #define mds_rcvd_filp mds_obt.obt_rcvd_filp
572 #define mds_server_data mds_obt.obt_lsd
573 #define mds_client_bitmap mds_obt.obt_client_bitmap
574 #define mds_mount_count mds_obt.obt_mount_count
575 #define mds_last_transno mds_obt.obt_last_transno
576 #define mds_vfsmnt mds_obt.obt_vfsmnt
579 extern __u32 mds_max_ost_index;
581 #define MDS_LOV_ALLOC_SIZE (CFS_PAGE_SIZE)
583 #define OBJID_PER_PAGE() (MDS_LOV_ALLOC_SIZE / sizeof(obd_id))
585 #define MDS_LOV_OBJID_PAGES_COUNT (mds_max_ost_index/OBJID_PER_PAGE())
587 extern int mds_lov_init_objids(struct obd_device *obd);
588 extern void mds_lov_destroy_objids(struct obd_device *obd);
601 struct lustre_handle eo_nl_lock;
606 struct ptlrpc_service *ost_service;
607 struct ptlrpc_service *ost_create_service;
608 struct ptlrpc_service *ost_io_service;
609 struct semaphore ost_health_sem;
612 struct echo_client_obd {
613 struct obd_export *ec_exp; /* the local connection to osc/lov */
615 struct list_head ec_objects;
616 struct list_head ec_locks;
622 struct obd_uuid lqo_uuid; /* ptlrpc's c_remote_uuid */
623 struct list_head lqo_oss_list; /* link to lov_qos */
624 __u64 lqo_bavail; /* total bytes avail on OSS */
625 __u64 lqo_penalty; /* current penalty */
626 __u64 lqo_penalty_per_obj; /* penalty decrease every obj*/
627 time_t lqo_used; /* last used time, seconds */
628 __u32 lqo_ost_count; /* number of osts on this oss */
632 struct lov_qos_oss *ltq_oss; /* oss info */
633 __u64 ltq_penalty; /* current penalty */
634 __u64 ltq_penalty_per_obj; /* penalty decrease every obj*/
635 __u64 ltq_weight; /* net weighting */
636 time_t ltq_used; /* last used time, seconds */
637 unsigned int ltq_usable:1; /* usable for striping */
640 /* Generic subset of OSTs */
642 __u32 *op_array; /* array of index of
644 unsigned int op_count; /* number of OSTs in the array */
645 unsigned int op_size; /* allocated size of lp_array */
646 struct rw_semaphore op_rw_sem; /* to protect ost_pool use */
649 /* Round-robin allocator data */
651 __u32 lqr_start_idx; /* start index of new inode */
652 __u32 lqr_offset_idx; /* aliasing for start_idx */
653 int lqr_start_count; /* reseed counter */
654 struct ost_pool lqr_pool; /* round-robin optimized list */
655 unsigned long lqr_dirty:1; /* recalc round-robin list */
658 struct lov_statfs_data {
659 struct obd_info lsd_oi;
660 struct obd_statfs lsd_statfs;
662 /* Stripe placement optimization */
664 struct list_head lq_oss_list; /* list of OSSs that targets use */
665 struct rw_semaphore lq_rw_sem;
666 __u32 lq_active_oss_count;
667 unsigned int lq_prio_free; /* priority for free space */
668 unsigned int lq_threshold_rr;/* priority for rr */
669 struct lov_qos_rr lq_rr; /* round robin qos data */
670 unsigned long lq_dirty:1, /* recalc qos data */
671 lq_same_space:1,/* the ost's all have approx.
672 the same space avail */
673 lq_reset:1, /* zero current penalties */
674 lq_statfs_in_progress:1; /* statfs op in progress */
675 /* qos statfs data */
676 struct lov_statfs_data *lq_statfs_data;
677 cfs_waitq_t lq_statfs_waitq; /* waitqueue to notify statfs
678 * requests completion */
681 struct lov_tgt_desc {
682 struct list_head ltd_kill;
683 struct obd_uuid ltd_uuid;
684 struct obd_device *ltd_obd;
685 struct obd_export *ltd_exp;
686 struct ltd_qos ltd_qos; /* qos info per target */
688 __u32 ltd_index; /* index in lov_obd->tgts */
689 unsigned long ltd_active:1,/* is this target up for requests */
690 ltd_activate:1,/* should this target be activated */
691 ltd_reap:1; /* should this target be deleted */
695 #define pool_tgt_size(_p) _p->pool_obds.op_size
696 #define pool_tgt_count(_p) _p->pool_obds.op_count
697 #define pool_tgt_array(_p) _p->pool_obds.op_array
698 #define pool_tgt_rw_sem(_p) _p->pool_obds.op_rw_sem
699 #define pool_tgt(_p, _i) _p->pool_lov->lov_tgts[_p->pool_obds.op_array[_i]]
702 char pool_name[LOV_MAXPOOLNAME + 1]; /* name of pool */
703 struct ost_pool pool_obds; /* pool members */
704 atomic_t pool_refcount; /* pool ref. counter */
705 struct lov_qos_rr pool_rr; /* round robin qos */
706 struct hlist_node pool_hash; /* access by poolname */
707 struct list_head pool_list; /* serial access */
708 cfs_proc_dir_entry_t *pool_proc_entry; /* file in /proc */
709 struct lov_obd *pool_lov; /* lov obd to which this
714 struct lov_desc desc;
715 struct lov_tgt_desc **lov_tgts; /* sparse array */
716 struct ost_pool lov_packed; /* all OSTs in a packed
718 struct semaphore lov_lock;
719 struct obd_connect_data lov_ocd;
720 struct lov_qos lov_qos; /* qos info per lov */
721 atomic_t lov_refcount;
722 __u32 lov_tgt_count; /* how many OBD's */
723 __u32 lov_active_tgt_count; /* how many active */
724 __u32 lov_death_row;/* tgts scheduled to be deleted */
725 __u32 lov_tgt_size; /* size of tgts array */
728 cfs_hash_t *lov_pools_hash_body; /* used for key access */
729 struct list_head lov_pool_list; /* used for sequential access */
730 cfs_proc_dir_entry_t *lov_pool_proc_entry;
731 enum lustre_sec_part lov_sp_me;
734 struct lmv_tgt_desc {
735 struct obd_uuid ltd_uuid;
736 struct obd_export *ltd_exp;
737 int ltd_active; /* is this target up for requests */
739 struct semaphore ltd_fid_sem;
742 enum placement_policy {
743 PLACEMENT_CHAR_POLICY = 0,
744 PLACEMENT_NID_POLICY = 1,
745 PLACEMENT_INVAL_POLICY = 2,
749 typedef enum placement_policy placement_policy_t;
753 struct lu_client_fld lmv_fld;
755 placement_policy_t lmv_placement;
756 struct lmv_desc desc;
757 struct obd_uuid cluuid;
758 struct obd_export *exp;
765 struct semaphore init_sem;
767 struct lmv_tgt_desc *tgts;
770 struct obd_connect_data *datas;
773 struct obd_connect_data conn_data;
776 struct niobuf_local {
781 cfs_dentry_t *dentry;
786 #define LUSTRE_FLD_NAME "fld"
787 #define LUSTRE_SEQ_NAME "seq"
789 #define LUSTRE_CMM_NAME "cmm"
790 #define LUSTRE_MDD_NAME "mdd"
791 #define LUSTRE_OSD_NAME "osd"
792 #define LUSTRE_VVP_NAME "vvp"
793 #define LUSTRE_LMV_NAME "lmv"
794 #define LUSTRE_CMM_MDC_NAME "cmm-mdc"
795 #define LUSTRE_SLP_NAME "slp"
797 /* obd device type names */
798 /* FIXME all the references to LUSTRE_MDS_NAME should be swapped with LUSTRE_MDT_NAME */
799 #define LUSTRE_MDS_NAME "mds"
800 #define LUSTRE_MDT_NAME "mdt"
801 #define LUSTRE_MDC_NAME "mdc"
802 #define LUSTRE_OSS_NAME "ost" /* FIXME change name to oss */
803 #define LUSTRE_OST_NAME "obdfilter" /* FIXME change name to ost */
804 #define LUSTRE_OSC_NAME "osc"
805 #define LUSTRE_LOV_NAME "lov"
806 #define LUSTRE_MGS_NAME "mgs"
807 #define LUSTRE_MGC_NAME "mgc"
809 #define LUSTRE_CACHEOBD_NAME "cobd"
810 #define LUSTRE_ECHO_NAME "obdecho"
811 #define LUSTRE_ECHO_CLIENT_NAME "echo_client"
813 /* Constant obd names (post-rename) */
814 #define LUSTRE_MDS_OBDNAME "MDS"
815 #define LUSTRE_OSS_OBDNAME "OSS"
816 #define LUSTRE_MGS_OBDNAME "MGS"
817 #define LUSTRE_MGC_OBDNAME "MGC"
819 /* Don't conflict with on-wire flags OBD_BRW_WRITE, etc */
820 #define N_LOCAL_TEMP_PAGE 0x10000000
822 struct obd_trans_info {
825 /* Only used on the server side for tracking acks. */
826 struct oti_req_ack_lock {
827 struct lustre_handle lock;
831 struct llog_cookie oti_onecookie;
832 struct llog_cookie *oti_logcookies;
835 /* initial thread handling transaction */
836 struct ptlrpc_thread * oti_thread;
839 __u64 oti_pre_version;
841 struct obd_uuid *oti_ost_uuid;
844 static inline void oti_init(struct obd_trans_info *oti,
845 struct ptlrpc_request *req)
849 memset(oti, 0, sizeof(*oti));
854 oti->oti_xid = req->rq_xid;
855 /** VBR: take versions from request */
856 if (req->rq_reqmsg != NULL &&
857 lustre_msg_get_flags(req->rq_reqmsg) & MSG_REPLAY) {
858 __u64 *pre_version = lustre_msg_get_versions(req->rq_reqmsg);
859 oti->oti_pre_version = pre_version ? pre_version[0] : 0;
860 oti->oti_transno = lustre_msg_get_transno(req->rq_reqmsg);
863 /** called from mds_create_objects */
864 if (req->rq_repmsg != NULL)
865 oti->oti_transno = lustre_msg_get_transno(req->rq_repmsg);
866 oti->oti_thread = req->rq_svc_thread;
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 connect start */
905 /* Device activated */
907 /* Device deactivated */
909 /* Device disconnected */
911 /* Connect data for import were changed */
914 OBD_NOTIFY_SYNC_NONBLOCK,
916 /* Configuration event */
918 /* Trigger quota recovery */
922 /* bit-mask flags for config events */
924 CONFIG_LOG = 0x1, /* finished processing config log */
925 CONFIG_SYNC = 0x2, /* mdt synced 1 ost */
926 CONFIG_TARGET = 0x4 /* one target is added */
930 * Data structure used to pass obd_notify()-event to non-obd listeners (llite
931 * and liblustre being main examples).
933 struct obd_notify_upcall {
934 int (*onu_upcall)(struct obd_device *host, struct obd_device *watched,
935 enum obd_notify_event ev, void *owner, void *data);
936 /* Opaque datum supplied by upper layer listener */
940 struct target_recovery_data {
941 svc_handler_t trd_recovery_handler;
942 pid_t trd_processing_task;
943 struct completion trd_starting;
944 struct completion trd_finishing;
948 FILTER_GROUP_MDS0 = 0,
949 FILTER_GROUP_LLOG = 1,
950 FILTER_GROUP_ECHO = 2 ,
951 FILTER_GROUP_MDS1_N_BASE = 3
955 * In HEAD for CMD, the object is created in group number which is 3>=
956 * or indexing starts from 3. To test this assertions are added to disallow
957 * group 0. But to run 2.0 mds server on 1.8.x disk format (i.e. interop_mode)
958 * object in group 0 needs to be allowed.
959 * So for interop mode following changes needs to be done:
960 * 1. No need to assert on group 0 or allow group 0
961 * 2. The group number indexing starts from 0 instead of 3
964 static inline int filter_group_is_mds(obd_gr group)
966 return (group == FILTER_GROUP_MDS0 ||
967 group >= FILTER_GROUP_MDS1_N_BASE);
970 #define LASSERT_MDS_GROUP(group) LASSERT(filter_group_is_mds(group))
972 static inline __u64 objgrp_to_mdsno(obd_gr group)
974 LASSERT(filter_group_is_mds(group));
975 if (group == FILTER_GROUP_MDS0)
977 return group - FILTER_GROUP_MDS1_N_BASE + 1;
980 static inline int mdt_to_obd_objgrp(int mdtid)
983 * MDS0 uses group 0 always, other MDSes will use groups from
984 * FILTER_GROUP_MDS1_N_BASE
987 return FILTER_GROUP_MDS1_N_BASE + mdtid - 1;
991 static inline __u64 obdo_mdsno(struct obdo *oa)
993 LASSERT((oa->o_valid & OBD_MD_FLGROUP));
994 return objgrp_to_mdsno(oa->o_gr);
997 static inline int obdo_is_mds(struct obdo *oa)
999 LASSERT(oa->o_valid & OBD_MD_FLGROUP);
1000 return filter_group_is_mds(oa->o_gr);
1003 struct obd_llog_group {
1004 struct list_head olg_list;
1006 struct llog_ctxt *olg_ctxts[LLOG_MAX_CTXTS];
1007 cfs_waitq_t olg_waitq;
1008 spinlock_t olg_lock;
1009 struct obd_export *olg_exp;
1010 int olg_initializing;
1011 struct semaphore olg_cat_processing;
1014 /* corresponds to one of the obd's */
1015 #define MAX_OBD_NAME 128
1016 #define OBD_DEVICE_MAGIC 0XAB5CD6EF
1017 #define OBD_DEV_BY_DEVNAME 0xffffd0de
1019 struct obd_type *obd_type;
1022 /* common and UUID name of this device */
1023 char obd_name[MAX_OBD_NAME];
1024 struct obd_uuid obd_uuid;
1026 struct lu_device *obd_lu_dev;
1029 unsigned long obd_attached:1, /* finished attach */
1030 obd_set_up:1, /* finished setup */
1031 obd_recovering:1, /* there are recoverable clients */
1032 obd_abort_recovery:1,/* recovery expired */
1033 obd_version_recov:1, /* obd uses version checking */
1034 obd_recovery_expired:1,
1035 obd_replayable:1, /* recovery is enabled; inform clients */
1036 obd_no_transno:1, /* no committed-transno notification */
1037 obd_no_recov:1, /* fail instead of retry messages */
1038 obd_req_replaying:1, /* replaying requests */
1039 obd_stopping:1, /* started cleanup */
1040 obd_starting:1, /* started setup */
1041 obd_force:1, /* cleanup with > 0 obd refcount */
1042 obd_fail:1, /* cleanup with failover */
1043 obd_async_recov:1, /* allow asynchronous orphan cleanup */
1044 obd_no_conn:1, /* deny new connections */
1045 obd_inactive:1, /* device active/inactive
1046 * (for /proc/status only!!) */
1047 obd_process_conf:1; /* device is processing mgs config */
1048 /* uuid-export hash body */
1049 cfs_hash_t *obd_uuid_hash;
1050 /* nid-export hash body */
1051 cfs_hash_t *obd_nid_hash;
1052 /* nid stats body */
1053 cfs_hash_t *obd_nid_stats_hash;
1054 struct list_head obd_nid_stats;
1055 atomic_t obd_refcount;
1056 cfs_waitq_t obd_refcount_waitq;
1057 struct list_head obd_exports;
1058 struct list_head obd_unlinked_exports;
1059 struct list_head obd_delayed_exports;
1060 int obd_num_exports;
1061 spinlock_t obd_nid_lock;
1062 struct ldlm_namespace *obd_namespace;
1063 struct ptlrpc_client obd_ldlm_client; /* XXX OST/MDS only */
1064 /* a spinlock is OK for what we do now, may need a semaphore later */
1065 spinlock_t obd_dev_lock;
1066 struct semaphore obd_dev_sem;
1067 __u64 obd_last_committed;
1068 struct fsfilt_operations *obd_fsops;
1069 spinlock_t obd_osfs_lock;
1070 struct obd_statfs obd_osfs; /* locked by obd_osfs_lock */
1072 struct lvfs_run_ctxt obd_lvfs_ctxt;
1073 struct obd_llog_group obd_olg; /* default llog group */
1074 struct obd_device *obd_observer;
1075 struct rw_semaphore obd_observer_link_sem;
1076 struct obd_notify_upcall obd_upcall;
1077 struct obd_export *obd_self_export;
1078 /* list of exports in LRU order, for ping evictor, with obd_dev_lock */
1079 struct list_head obd_exports_timed;
1080 time_t obd_eviction_timer; /* for ping evictor */
1082 int obd_max_recoverable_clients;
1083 int obd_connected_clients;
1084 int obd_stale_clients;
1085 int obd_delayed_clients;
1086 spinlock_t obd_processing_task_lock; /* BH lock (timer) */
1087 __u64 obd_next_recovery_transno;
1088 int obd_replayed_requests;
1089 int obd_requests_queued_for_recovery;
1090 cfs_waitq_t obd_next_transno_waitq;
1091 cfs_timer_t obd_recovery_timer;
1092 time_t obd_recovery_start; /* seconds */
1093 time_t obd_recovery_end; /* seconds, for lprocfs_status */
1094 time_t obd_recovery_max_time; /* seconds, bz13079 */
1095 int obd_recovery_timeout;
1097 /* new recovery stuff from CMD2 */
1098 struct target_recovery_data obd_recovery_data;
1099 int obd_replayed_locks;
1100 atomic_t obd_req_replay_clients;
1101 atomic_t obd_lock_replay_clients;
1102 struct list_head obd_req_replay_queue;
1103 struct list_head obd_lock_replay_queue;
1104 struct list_head obd_final_req_queue;
1105 int obd_recovery_stage;
1108 struct obd_device_target obt;
1109 struct filter_obd filter;
1111 struct client_obd cli;
1113 struct echo_client_obd echo_client;
1114 struct echo_obd echo;
1119 /* Fields used by LProcFS */
1120 unsigned int obd_cntr_base;
1121 struct lprocfs_stats *obd_stats;
1123 unsigned int md_cntr_base;
1124 struct lprocfs_stats *md_stats;
1126 cfs_proc_dir_entry_t *obd_proc_entry;
1127 cfs_proc_dir_entry_t *obd_proc_exports_entry;
1128 cfs_proc_dir_entry_t *obd_svc_procroot;
1129 struct lprocfs_stats *obd_svc_stats;
1130 atomic_t obd_evict_inprogress;
1131 cfs_waitq_t obd_evict_inprogress_waitq;
1132 struct list_head obd_evict_list; /* protected with pet_lock */
1135 * Ldlm pool part. Save last calculated SLV and Limit.
1137 rwlock_t obd_pool_lock;
1142 * A list of outstanding class_incref()'s against this obd. For
1145 struct lu_ref obd_reference;
1148 #define OBD_LLOG_FL_SENDNOW 0x0001
1150 enum obd_cleanup_stage {
1151 /* Special case hack for MDS LOVs */
1153 /* can be directly mapped to .ldto_device_fini() */
1154 OBD_CLEANUP_EXPORTS,
1157 /* get/set_info keys */
1158 #define KEY_ASYNC "async"
1159 #define KEY_BLOCKSIZE_BITS "blocksize_bits"
1160 #define KEY_BLOCKSIZE "blocksize"
1161 #define KEY_CAPA_KEY "capa_key"
1162 #define KEY_CHANGELOG_CLEAR "changelog_clear"
1163 #define KEY_FID2PATH "fid2path"
1164 #define KEY_CHECKSUM "checksum"
1165 #define KEY_CLEAR_FS "clear_fs"
1166 #define KEY_CONN_DATA "conn_data"
1167 #define KEY_EVICT_BY_NID "evict_by_nid"
1168 #define KEY_FIEMAP "fiemap"
1169 #define KEY_FLUSH_CTX "flush_ctx"
1170 #define KEY_GRANT_SHRINK "grant_shrink"
1171 #define KEY_HSM_COPYTOOL_SEND "hsm_send"
1172 #define KEY_INIT_RECOV_BACKUP "init_recov_bk"
1173 #define KEY_INIT_RECOV "initial_recov"
1174 #define KEY_INTERMDS "inter_mds"
1175 #define KEY_LAST_ID "last_id"
1176 #define KEY_LOCK_TO_STRIPE "lock_to_stripe"
1177 #define KEY_LOVDESC "lovdesc"
1178 #define KEY_LOV_IDX "lov_idx"
1179 #define KEY_MAX_EASIZE "max_easize"
1180 #define KEY_MDS_CONN "mds_conn"
1181 #define KEY_MGSSEC "mgssec"
1182 #define KEY_NEXT_ID "next_id"
1183 #define KEY_READ_ONLY "read-only"
1184 #define KEY_REGISTER_TARGET "register_target"
1185 #define KEY_REVIMP_UPD "revimp_update"
1186 #define KEY_SET_FS "set_fs"
1187 /* KEY_SET_INFO in lustre_idl.h */
1188 #define KEY_SPTLRPC_CONF "sptlrpc_conf"
1193 static inline int it_to_lock_mode(struct lookup_intent *it)
1195 /* CREAT needs to be tested before open (both could be set) */
1196 if (it->it_op & IT_CREAT)
1198 else if (it->it_op & (IT_READDIR | IT_GETATTR | IT_OPEN | IT_LOOKUP))
1201 LASSERTF(0, "Invalid it_op: %d\n", it->it_op);
1206 struct lu_fid op_fid1; /* operation fid1 (usualy parent) */
1207 struct lu_fid op_fid2; /* operation fid2 (usualy child) */
1208 struct lu_fid op_fid3; /* 2 extra fids to find conflicting */
1209 struct lu_fid op_fid4; /* to the operation locks. */
1210 mdsno_t op_mds; /* what mds server open will go to */
1211 struct lustre_handle op_handle;
1213 const char *op_name;
1216 struct lmv_stripe_md *op_mea1;
1217 struct lmv_stripe_md *op_mea2;
1218 __u32 op_suppgids[2];
1224 /* iattr fields and blocks. */
1225 struct iattr op_attr;
1227 #if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,14)
1228 unsigned int op_attr_flags;
1231 loff_t op_attr_blocks;
1233 /* Size-on-MDS epoch and flags. */
1238 struct obd_capa *op_capa1;
1239 struct obd_capa *op_capa2;
1241 /* Various operation flags. */
1244 /* Operation type */
1248 struct md_enqueue_info;
1249 /* metadata stat-ahead */
1250 typedef int (* md_enqueue_cb_t)(struct ptlrpc_request *req,
1251 struct md_enqueue_info *minfo,
1254 struct md_enqueue_info {
1255 struct md_op_data mi_data;
1256 struct lookup_intent mi_it;
1257 struct lustre_handle mi_lockh;
1258 struct dentry *mi_dentry;
1259 struct inode *mi_dir;
1260 md_enqueue_cb_t mi_cb;
1261 unsigned int mi_generation;
1266 struct module *o_owner;
1267 int (*o_iocontrol)(unsigned int cmd, struct obd_export *exp, int len,
1268 void *karg, void *uarg);
1269 int (*o_get_info)(struct obd_export *, __u32 keylen, void *key,
1270 __u32 *vallen, void *val, struct lov_stripe_md *lsm);
1271 int (*o_set_info_async)(struct obd_export *, __u32 keylen, void *key,
1272 __u32 vallen, void *val,
1273 struct ptlrpc_request_set *set);
1274 int (*o_attach)(struct obd_device *dev, obd_count len, void *data);
1275 int (*o_detach)(struct obd_device *dev);
1276 int (*o_setup) (struct obd_device *dev, struct lustre_cfg *cfg);
1277 int (*o_precleanup)(struct obd_device *dev,
1278 enum obd_cleanup_stage cleanup_stage);
1279 int (*o_cleanup)(struct obd_device *dev);
1280 int (*o_process_config)(struct obd_device *dev, obd_count len,
1282 int (*o_postrecov)(struct obd_device *dev);
1283 int (*o_add_conn)(struct obd_import *imp, struct obd_uuid *uuid,
1285 int (*o_del_conn)(struct obd_import *imp, struct obd_uuid *uuid);
1286 /* connect to the target device with given connection
1287 * data. @ocd->ocd_connect_flags is modified to reflect flags actually
1288 * granted by the target, which are guaranteed to be a subset of flags
1289 * asked for. If @ocd == NULL, use default parameters. */
1290 int (*o_connect)(const struct lu_env *env,
1291 struct obd_export **exp, struct obd_device *src,
1292 struct obd_uuid *cluuid, struct obd_connect_data *ocd,
1294 int (*o_reconnect)(const struct lu_env *env,
1295 struct obd_export *exp, struct obd_device *src,
1296 struct obd_uuid *cluuid,
1297 struct obd_connect_data *ocd,
1299 int (*o_disconnect)(struct obd_export *exp);
1301 /* Initialize/finalize fids infrastructure. */
1302 int (*o_fid_init)(struct obd_export *exp);
1303 int (*o_fid_fini)(struct obd_export *exp);
1305 /* Allocate new fid according to passed @hint. */
1306 int (*o_fid_alloc)(struct obd_export *exp, struct lu_fid *fid,
1307 struct md_op_data *op_data);
1310 * Object with @fid is getting deleted, we may want to do something
1313 int (*o_fid_delete)(struct obd_export *exp, const struct lu_fid *fid);
1315 int (*o_statfs)(struct obd_device *obd, struct obd_statfs *osfs,
1316 __u64 max_age, __u32 flags);
1317 int (*o_statfs_async)(struct obd_device *obd, struct obd_info *oinfo,
1318 __u64 max_age, struct ptlrpc_request_set *set);
1319 int (*o_packmd)(struct obd_export *exp, struct lov_mds_md **disk_tgt,
1320 struct lov_stripe_md *mem_src);
1321 int (*o_unpackmd)(struct obd_export *exp,struct lov_stripe_md **mem_tgt,
1322 struct lov_mds_md *disk_src, int disk_len);
1323 int (*o_preallocate)(struct lustre_handle *, obd_count *req,
1325 /* FIXME: add fid capability support for create & destroy! */
1326 int (*o_precreate)(struct obd_export *exp);
1327 int (*o_create)(struct obd_export *exp, struct obdo *oa,
1328 struct lov_stripe_md **ea, struct obd_trans_info *oti);
1329 int (*o_create_async)(struct obd_export *exp, struct obd_info *oinfo,
1330 struct lov_stripe_md **ea,
1331 struct obd_trans_info *oti);
1332 int (*o_destroy)(struct obd_export *exp, struct obdo *oa,
1333 struct lov_stripe_md *ea, struct obd_trans_info *oti,
1334 struct obd_export *md_exp, void *capa);
1335 int (*o_setattr)(struct obd_export *exp, struct obd_info *oinfo,
1336 struct obd_trans_info *oti);
1337 int (*o_setattr_async)(struct obd_export *exp, struct obd_info *oinfo,
1338 struct obd_trans_info *oti,
1339 struct ptlrpc_request_set *rqset);
1340 int (*o_getattr)(struct obd_export *exp, struct obd_info *oinfo);
1341 int (*o_getattr_async)(struct obd_export *exp, struct obd_info *oinfo,
1342 struct ptlrpc_request_set *set);
1343 int (*o_brw)(int rw, struct obd_export *exp, struct obd_info *oinfo,
1344 obd_count oa_bufs, struct brw_page *pgarr,
1345 struct obd_trans_info *oti);
1346 int (*o_merge_lvb)(struct obd_export *exp, struct lov_stripe_md *lsm,
1347 struct ost_lvb *lvb, int kms_only);
1348 int (*o_adjust_kms)(struct obd_export *exp, struct lov_stripe_md *lsm,
1349 obd_off size, int shrink);
1350 int (*o_punch)(struct obd_export *exp, struct obd_info *oinfo,
1351 struct obd_trans_info *oti,
1352 struct ptlrpc_request_set *rqset);
1353 int (*o_sync)(struct obd_export *exp, struct obdo *oa,
1354 struct lov_stripe_md *ea, obd_size start, obd_size end,
1356 int (*o_migrate)(struct lustre_handle *conn, struct lov_stripe_md *dst,
1357 struct lov_stripe_md *src, obd_size start,
1358 obd_size end, struct obd_trans_info *oti);
1359 int (*o_copy)(struct lustre_handle *dstconn, struct lov_stripe_md *dst,
1360 struct lustre_handle *srconn, struct lov_stripe_md *src,
1361 obd_size start, obd_size end, struct obd_trans_info *);
1362 int (*o_iterate)(struct lustre_handle *conn,
1363 int (*)(obd_id, obd_gr, void *),
1364 obd_id *startid, obd_gr group, void *data);
1365 int (*o_preprw)(int cmd, struct obd_export *exp, struct obdo *oa,
1366 int objcount, struct obd_ioobj *obj,
1367 struct niobuf_remote *remote, int *nr_pages,
1368 struct niobuf_local *local,
1369 struct obd_trans_info *oti,
1370 struct lustre_capa *capa);
1371 int (*o_commitrw)(int cmd, struct obd_export *exp, struct obdo *oa,
1372 int objcount, struct obd_ioobj *obj,
1373 struct niobuf_remote *remote, int pages,
1374 struct niobuf_local *local,
1375 struct obd_trans_info *oti, int rc);
1376 int (*o_enqueue)(struct obd_export *, struct obd_info *oinfo,
1377 struct ldlm_enqueue_info *einfo,
1378 struct ptlrpc_request_set *rqset);
1379 int (*o_change_cbdata)(struct obd_export *, struct lov_stripe_md *,
1380 ldlm_iterator_t it, void *data);
1381 int (*o_cancel)(struct obd_export *, struct lov_stripe_md *md,
1382 __u32 mode, struct lustre_handle *);
1383 int (*o_cancel_unused)(struct obd_export *, struct lov_stripe_md *,
1384 int flags, void *opaque);
1385 int (*o_init_export)(struct obd_export *exp);
1386 int (*o_destroy_export)(struct obd_export *exp);
1387 int (*o_extent_calc)(struct obd_export *, struct lov_stripe_md *,
1388 int cmd, obd_off *);
1390 /* llog related obd_methods */
1391 int (*o_llog_init)(struct obd_device *obd, struct obd_llog_group *grp,
1392 struct obd_device *disk_obd, int *idx);
1393 int (*o_llog_finish)(struct obd_device *obd, int count);
1394 int (*o_llog_connect)(struct obd_export *, struct llogd_conn_body *);
1396 /* metadata-only methods */
1397 int (*o_pin)(struct obd_export *, const struct lu_fid *fid,
1398 struct obd_capa *, struct obd_client_handle *, int flag);
1399 int (*o_unpin)(struct obd_export *, struct obd_client_handle *, int);
1401 int (*o_import_event)(struct obd_device *, struct obd_import *,
1402 enum obd_import_event);
1404 int (*o_notify)(struct obd_device *obd, struct obd_device *watched,
1405 enum obd_notify_event ev, void *data);
1407 int (*o_health_check)(struct obd_device *);
1408 struct obd_uuid *(*o_get_uuid) (struct obd_export *exp);
1411 int (*o_quotacheck)(struct obd_device *, struct obd_export *,
1412 struct obd_quotactl *);
1413 int (*o_quotactl)(struct obd_device *, struct obd_export *,
1414 struct obd_quotactl *);
1415 int (*o_quota_adjust_qunit)(struct obd_export *exp,
1416 struct quota_adjust_qunit *oqaq,
1417 struct lustre_quota_ctxt *qctxt);
1420 int (*o_ping)(struct obd_export *exp);
1423 int (*o_pool_new)(struct obd_device *obd, char *poolname);
1424 int (*o_pool_del)(struct obd_device *obd, char *poolname);
1425 int (*o_pool_add)(struct obd_device *obd, char *poolname,
1427 int (*o_pool_rem)(struct obd_device *obd, char *poolname,
1429 void (*o_getref)(struct obd_device *obd);
1430 void (*o_putref)(struct obd_device *obd);
1432 * NOTE: If adding ops, add another LPROCFS_OBD_OP_INIT() line
1433 * to lprocfs_alloc_obd_stats() in obdclass/lprocfs_status.c.
1434 * Also, add a wrapper function in include/linux/obd_class.h. */
1438 LUSTRE_OPC_MKDIR = (1 << 0),
1439 LUSTRE_OPC_SYMLINK = (1 << 1),
1440 LUSTRE_OPC_MKNOD = (1 << 2),
1441 LUSTRE_OPC_CREATE = (1 << 3),
1442 LUSTRE_OPC_ANY = (1 << 4)
1445 /* lmv structures */
1446 #define MEA_MAGIC_LAST_CHAR 0xb2221ca1
1447 #define MEA_MAGIC_ALL_CHARS 0xb222a11c
1448 #define MEA_MAGIC_HASH_SEGMENT 0xb222a11b
1450 #define MAX_HASH_SIZE_32 0x7fffffffUL
1451 #define MAX_HASH_SIZE 0x7fffffffffffffffULL
1452 #define MAX_HASH_HIGHEST_BIT 0x1000000000000000ULL
1455 struct mdt_body *body;
1456 struct lov_stripe_md *lsm;
1457 struct lmv_stripe_md *mea;
1458 #ifdef CONFIG_FS_POSIX_ACL
1459 struct posix_acl *posix_acl;
1461 struct mdt_remote_perm *remote_perm;
1462 struct obd_capa *mds_capa;
1463 struct obd_capa *oss_capa;
1466 struct md_open_data {
1467 struct obd_client_handle *mod_och;
1468 struct ptlrpc_request *mod_open_req;
1469 struct ptlrpc_request *mod_close_req;
1470 atomic_t mod_refcount;
1473 struct lookup_intent;
1476 int (*m_getstatus)(struct obd_export *, struct lu_fid *,
1477 struct obd_capa **);
1478 int (*m_change_cbdata)(struct obd_export *, const struct lu_fid *,
1479 ldlm_iterator_t, void *);
1480 int (*m_close)(struct obd_export *, struct md_op_data *,
1481 struct md_open_data *, struct ptlrpc_request **);
1482 int (*m_create)(struct obd_export *, struct md_op_data *,
1483 const void *, int, int, __u32, __u32, cfs_cap_t,
1484 __u64, struct ptlrpc_request **);
1485 int (*m_done_writing)(struct obd_export *, struct md_op_data *,
1486 struct md_open_data *);
1487 int (*m_enqueue)(struct obd_export *, struct ldlm_enqueue_info *,
1488 struct lookup_intent *, struct md_op_data *,
1489 struct lustre_handle *, void *, int,
1490 struct ptlrpc_request **, int);
1491 int (*m_getattr)(struct obd_export *, const struct lu_fid *,
1492 struct obd_capa *, obd_valid, int,
1493 struct ptlrpc_request **);
1494 int (*m_getattr_name)(struct obd_export *, const struct lu_fid *,
1495 struct obd_capa *, const char *, int, obd_valid,
1496 int, __u32, struct ptlrpc_request **);
1497 int (*m_intent_lock)(struct obd_export *, struct md_op_data *,
1498 void *, int, struct lookup_intent *, int,
1499 struct ptlrpc_request **,
1500 ldlm_blocking_callback, int);
1501 int (*m_link)(struct obd_export *, struct md_op_data *,
1502 struct ptlrpc_request **);
1503 int (*m_rename)(struct obd_export *, struct md_op_data *,
1504 const char *, int, const char *, int,
1505 struct ptlrpc_request **);
1506 int (*m_is_subdir)(struct obd_export *, const struct lu_fid *,
1507 const struct lu_fid *,
1508 struct ptlrpc_request **);
1509 int (*m_setattr)(struct obd_export *, struct md_op_data *, void *,
1510 int , void *, int, struct ptlrpc_request **,
1511 struct md_open_data **mod);
1512 int (*m_sync)(struct obd_export *, const struct lu_fid *,
1513 struct obd_capa *, struct ptlrpc_request **);
1514 int (*m_readpage)(struct obd_export *, const struct lu_fid *,
1515 struct obd_capa *, __u64, struct page *,
1516 struct ptlrpc_request **);
1518 int (*m_unlink)(struct obd_export *, struct md_op_data *,
1519 struct ptlrpc_request **);
1521 int (*m_setxattr)(struct obd_export *, const struct lu_fid *,
1522 struct obd_capa *, obd_valid, const char *,
1523 const char *, int, int, int, __u32,
1524 struct ptlrpc_request **);
1526 int (*m_getxattr)(struct obd_export *, const struct lu_fid *,
1527 struct obd_capa *, obd_valid, const char *,
1528 const char *, int, int, int,
1529 struct ptlrpc_request **);
1531 int (*m_init_ea_size)(struct obd_export *, int, int, int);
1533 int (*m_get_lustre_md)(struct obd_export *, struct ptlrpc_request *,
1534 struct obd_export *, struct obd_export *,
1535 struct lustre_md *);
1537 int (*m_free_lustre_md)(struct obd_export *, struct lustre_md *);
1539 int (*m_set_open_replay_data)(struct obd_export *,
1540 struct obd_client_handle *,
1541 struct ptlrpc_request *);
1542 int (*m_clear_open_replay_data)(struct obd_export *,
1543 struct obd_client_handle *);
1544 int (*m_set_lock_data)(struct obd_export *, __u64 *, void *, __u32 *);
1546 ldlm_mode_t (*m_lock_match)(struct obd_export *, int,
1547 const struct lu_fid *, ldlm_type_t,
1548 ldlm_policy_data_t *, ldlm_mode_t,
1549 struct lustre_handle *);
1551 int (*m_cancel_unused)(struct obd_export *, const struct lu_fid *,
1552 ldlm_policy_data_t *, ldlm_mode_t, int flags,
1554 int (*m_renew_capa)(struct obd_export *, struct obd_capa *oc,
1555 renew_capa_cb_t cb);
1556 int (*m_unpack_capa)(struct obd_export *, struct ptlrpc_request *,
1557 const struct req_msg_field *, struct obd_capa **);
1559 int (*m_get_remote_perm)(struct obd_export *, const struct lu_fid *,
1560 struct obd_capa *, __u32,
1561 struct ptlrpc_request **);
1563 int (*m_intent_getattr_async)(struct obd_export *,
1564 struct md_enqueue_info *,
1565 struct ldlm_enqueue_info *);
1567 int (*m_revalidate_lock)(struct obd_export *,
1568 struct lookup_intent *,
1572 * NOTE: If adding ops, add another LPROCFS_MD_OP_INIT() line to
1573 * lprocfs_alloc_md_stats() in obdclass/lprocfs_status.c. Also, add a
1574 * wrapper function in include/linux/obd_class.h.
1578 struct lsm_operations {
1579 void (*lsm_free)(struct lov_stripe_md *);
1580 int (*lsm_destroy)(struct lov_stripe_md *, struct obdo *oa,
1581 struct obd_export *md_exp);
1582 void (*lsm_stripe_by_index)(struct lov_stripe_md *, int *, obd_off *,
1584 void (*lsm_stripe_by_offset)(struct lov_stripe_md *, int *, obd_off *,
1586 int (*lsm_lmm_verify) (struct lov_mds_md *lmm, int lmm_bytes,
1588 int (*lsm_unpackmd) (struct lov_obd *lov, struct lov_stripe_md *lsm,
1589 struct lov_mds_md *lmm);
1592 extern const struct lsm_operations lsm_v1_ops;
1593 extern const struct lsm_operations lsm_v3_ops;
1594 static inline const struct lsm_operations *lsm_op_find(int magic)
1602 CERROR("Cannot recognize lsm_magic %08x\n", magic);
1607 int lvfs_check_io_health(struct obd_device *obd, struct file *file);
1609 /* Requests for obd_extent_calc() */
1610 #define OBD_CALC_STRIPE_START 1
1611 #define OBD_CALC_STRIPE_END 2
1613 static inline void obd_transno_commit_cb(struct obd_device *obd, __u64 transno,
1614 struct obd_export *exp, int error)
1617 CERROR("%s: transno "LPU64" commit error: %d\n",
1618 obd->obd_name, transno, error);
1621 if (exp && transno > exp->exp_last_committed) {
1622 CDEBUG(D_HA, "%s: transno "LPU64" committed\n",
1623 obd->obd_name, transno);
1624 exp->exp_last_committed = transno;
1625 ptlrpc_commit_replies(exp);
1627 CDEBUG(D_INFO, "%s: transno "LPU64" committed\n",
1628 obd->obd_name, transno);
1630 if (transno > obd->obd_last_committed)
1631 obd->obd_last_committed = transno;
1634 static inline void init_obd_quota_ops(quota_interface_t *interface,
1635 struct obd_ops *obd_ops)
1641 obd_ops->o_quotacheck = QUOTA_OP(interface, check);
1642 obd_ops->o_quotactl = QUOTA_OP(interface, ctl);
1643 obd_ops->o_quota_adjust_qunit = QUOTA_OP(interface, adjust_qunit);
1646 static inline struct lustre_capa *oinfo_capa(struct obd_info *oinfo)
1648 return oinfo->oi_capa;
1651 static inline struct md_open_data *obd_mod_alloc(void)
1653 struct md_open_data *mod;
1657 atomic_set(&mod->mod_refcount, 1);
1661 #define obd_mod_get(mod) atomic_inc(&(mod)->mod_refcount)
1662 #define obd_mod_put(mod) \
1664 if (atomic_dec_and_test(&(mod)->mod_refcount)) { \
1665 if ((mod)->mod_open_req) \
1666 ptlrpc_req_finished((mod)->mod_open_req); \
1667 OBD_FREE_PTR(mod); \
1671 #endif /* __OBD_H */