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 cfs_list_t lop_pending;
79 cfs_list_t lop_urgent;
80 cfs_list_t lop_pending_group;
90 struct lov_oinfo { /* per-stripe data structure */
91 struct ost_id loi_oi; /* object ID/Sequence on the target OST */
92 int loi_ost_idx; /* OST stripe index in lov_tgt_desc->tgts */
93 int loi_ost_gen; /* generation of this loi_ost_idx */
95 /* used by the osc to keep track of what objects to build into rpcs */
96 struct loi_oap_pages loi_read_lop;
97 struct loi_oap_pages loi_write_lop;
98 cfs_list_t loi_ready_item;
99 cfs_list_t loi_hp_ready_item;
100 cfs_list_t loi_write_item;
101 cfs_list_t loi_read_item;
103 unsigned long loi_kms_valid:1;
104 __u64 loi_kms; /* known minimum size */
105 struct ost_lvb loi_lvb;
106 struct osc_async_rc loi_ar;
108 #define loi_id loi_oi.oi_id
109 #define loi_seq loi_oi.oi_seq
111 static inline void loi_kms_set(struct lov_oinfo *oinfo, __u64 kms)
113 oinfo->loi_kms = kms;
114 oinfo->loi_kms_valid = 1;
117 static inline void loi_init(struct lov_oinfo *loi)
119 CFS_INIT_LIST_HEAD(&loi->loi_read_lop.lop_pending);
120 CFS_INIT_LIST_HEAD(&loi->loi_read_lop.lop_urgent);
121 CFS_INIT_LIST_HEAD(&loi->loi_read_lop.lop_pending_group);
122 CFS_INIT_LIST_HEAD(&loi->loi_write_lop.lop_pending);
123 CFS_INIT_LIST_HEAD(&loi->loi_write_lop.lop_urgent);
124 CFS_INIT_LIST_HEAD(&loi->loi_write_lop.lop_pending_group);
125 CFS_INIT_LIST_HEAD(&loi->loi_ready_item);
126 CFS_INIT_LIST_HEAD(&loi->loi_hp_ready_item);
127 CFS_INIT_LIST_HEAD(&loi->loi_write_item);
128 CFS_INIT_LIST_HEAD(&loi->loi_read_item);
131 struct lov_stripe_md {
132 cfs_spinlock_t lsm_lock;
133 pid_t lsm_lock_owner; /* debugging */
136 /* Public members. */
137 __u64 lw_object_id; /* lov object id */
138 __u64 lw_object_seq; /* lov object seq */
139 __u64 lw_maxbytes; /* maximum possible file size */
141 /* LOV-private members start here -- only for use in lov/. */
143 __u32 lw_stripe_size; /* size of the stripe */
144 __u32 lw_pattern; /* striping pattern (RAID0, RAID1) */
145 unsigned lw_stripe_count; /* number of objects being striped over */
146 char lw_pool_name[LOV_MAXPOOLNAME]; /* pool name */
149 struct lov_oinfo *lsm_oinfo[0];
152 #define lsm_object_id lsm_wire.lw_object_id
153 #define lsm_object_seq lsm_wire.lw_object_seq
154 #define lsm_maxbytes lsm_wire.lw_maxbytes
155 #define lsm_magic lsm_wire.lw_magic
156 #define lsm_stripe_size lsm_wire.lw_stripe_size
157 #define lsm_pattern lsm_wire.lw_pattern
158 #define lsm_stripe_count lsm_wire.lw_stripe_count
159 #define lsm_pool_name lsm_wire.lw_pool_name
163 typedef int (*obd_enqueue_update_f)(void *cookie, int rc);
165 /* obd info for a particular level (lov, osc). */
167 /* Lock policy. It keeps an extent which is specific for a particular
168 * OSC. (e.g. lov_prep_enqueue_set initialises extent of the policy,
169 * and osc_enqueue passes it into ldlm_lock_match & ldlm_cli_enqueue. */
170 ldlm_policy_data_t oi_policy;
171 /* Flags used for set request specific flags:
172 - while lock handling, the flags obtained on the enqueue
173 request are set here.
174 - while stats, the flags used for control delay/resend.
177 /* Lock handle specific for every OSC lock. */
178 struct lustre_handle *oi_lockh;
179 /* lsm data specific for every OSC. */
180 struct lov_stripe_md *oi_md;
181 /* obdo data specific for every OSC, if needed at all. */
183 /* statfs data specific for every OSC, if needed at all. */
184 struct obd_statfs *oi_osfs;
185 /* An update callback which is called to update some data on upper
186 * level. E.g. it is used for update lsm->lsm_oinfo at every recieved
187 * request in osc level for enqueue requests. It is also possible to
188 * update some caller data from LOV layer if needed. */
189 obd_enqueue_update_f oi_cb_up;
190 /* oss capability, its type is obd_capa in client to avoid copy.
191 * in contrary its type is lustre_capa in OSS. */
195 /* compare all relevant fields. */
196 static inline int lov_stripe_md_cmp(struct lov_stripe_md *m1,
197 struct lov_stripe_md *m2)
200 * ->lsm_wire contains padding, but it should be zeroed out during
203 return memcmp(&m1->lsm_wire, &m2->lsm_wire, sizeof m1->lsm_wire);
206 void lov_stripe_lock(struct lov_stripe_md *md);
207 void lov_stripe_unlock(struct lov_stripe_md *md);
210 cfs_list_t typ_chain;
211 struct obd_ops *typ_dt_ops;
212 struct md_ops *typ_md_ops;
213 cfs_proc_dir_entry_t *typ_procroot;
216 struct lu_device_type *typ_lu;
217 cfs_spinlock_t obd_type_lock;
227 /* Individual type definitions */
229 struct ost_server_data;
231 /* hold common fields for "target" device */
232 struct obd_device_target {
233 struct super_block *obt_sb;
234 /** last_rcvd file */
235 struct file *obt_rcvd_filp;
236 struct lu_target *obt_lut;
237 __u64 obt_mount_count;
238 cfs_semaphore_t obt_quotachecking;
239 struct lustre_quota_ctxt obt_qctxt;
240 lustre_quota_version_t obt_qfmt;
241 cfs_rw_semaphore_t obt_rwsem;
242 struct vfsmount *obt_vfsmnt;
243 struct file *obt_health_check_filp;
248 LLOG_CONFIG_ORIG_CTXT = 0,
249 LLOG_CONFIG_REPL_CTXT,
250 LLOG_MDS_OST_ORIG_CTXT,
251 LLOG_MDS_OST_REPL_CTXT,
258 LLOG_LOVEA_ORIG_CTXT,
259 LLOG_LOVEA_REPL_CTXT,
260 LLOG_CHANGELOG_ORIG_CTXT, /**< changelog generation on mdd */
261 LLOG_CHANGELOG_REPL_CTXT, /**< changelog access on clients */
262 LLOG_CHANGELOG_USER_ORIG_CTXT, /**< for multiple changelog consumers */
266 #define FILTER_SUBDIR_COUNT 32 /* set to zero for no subdirs */
268 struct filter_subdirs {
269 cfs_dentry_t *dentry[FILTER_SUBDIR_COUNT];
279 /* NB this field MUST be first */
280 struct obd_device_target fo_obt;
281 const char *fo_fstype;
284 cfs_dentry_t *fo_dentry_O;
285 cfs_dentry_t **fo_dentry_O_groups;
286 struct filter_subdirs *fo_dentry_O_sub;
287 cfs_semaphore_t fo_init_lock; /* group initialization lock */
288 int fo_committed_group;
290 cfs_spinlock_t fo_objidlock; /* protect fo_lastobjid */
292 unsigned long fo_destroys_in_progress;
293 cfs_semaphore_t fo_create_locks[FILTER_SUBDIR_COUNT];
295 cfs_list_t fo_export_list;
298 obd_size fo_tot_dirty; /* protected by obd_osfs_lock */
299 obd_size fo_tot_granted; /* all values in bytes */
300 obd_size fo_tot_pending;
301 int fo_tot_granted_clients;
303 obd_size fo_readcache_max_filesize;
304 int fo_read_cache:1, /**< enable read-only cache */
305 fo_writethrough_cache:1,/**< read cache writes */
306 fo_mds_ost_sync:1, /**< MDS-OST orphan recovery*/
307 fo_raid_degraded:1;/**< RAID device degraded */
309 struct obd_import *fo_mdc_imp;
310 struct obd_uuid fo_mdc_uuid;
311 struct lustre_handle fo_mdc_conn;
312 struct file **fo_last_objid_files;
313 __u64 *fo_last_objids; /* last created objid for groups,
314 * protected by fo_objidlock */
316 cfs_semaphore_t fo_alloc_lock;
318 cfs_atomic_t fo_r_in_flight;
319 cfs_atomic_t fo_w_in_flight;
322 * per-filter pool of kiobuf's allocated by filter_common_setup() and
323 * torn down by filter_cleanup(). Contains OST_NUM_THREADS elements of
324 * which ->fo_iobuf_count were allocated.
326 * This pool contains kiobuf used by
327 * filter_{prep,commit}rw_{read,write}() and is shared by all OST
330 * Locking: none, each OST thread uses only one element, determined by
331 * its "ordinal number", ->t_id.
333 struct filter_iobuf **fo_iobuf_pool;
336 cfs_list_t fo_llog_list;
337 cfs_spinlock_t fo_llog_list_lock;
339 struct brw_stats fo_filter_stats;
340 struct lustre_quota_ctxt fo_quota_ctxt;
341 cfs_spinlock_t fo_quotacheck_lock;
342 cfs_atomic_t fo_quotachecking;
344 int fo_fmd_max_num; /* per exp filter_mod_data */
345 int fo_fmd_max_age; /* jiffies to fmd expiry */
348 cfs_rwlock_t fo_sptlrpc_lock;
349 struct sptlrpc_rule_set fo_sptlrpc_rset;
351 /* capability related */
352 unsigned int fo_fl_oss_capa;
353 cfs_list_t fo_capa_keys;
354 cfs_hlist_head_t *fo_capa_hash;
355 struct llog_commit_master *fo_lcm;
359 #define fo_translock fo_obt.obt_lut->lut_translock
360 #define fo_last_rcvd_slots fo_obt.obt_lut->lut_client_bitmap
361 #define fo_mount_count fo_obt.obt_lut->lut_mount_count
362 #define fo_rcvd_filp fo_obt.obt_rcvd_filp
363 #define fo_vfsmnt fo_obt.obt_vfsmnt
365 struct timeout_item {
366 enum timeout_event ti_event;
367 cfs_time_t ti_timeout;
370 cfs_list_t ti_obd_list;
374 #define OSC_MAX_RIF_DEFAULT 8
375 #define OSC_MAX_RIF_MAX 256
376 #define OSC_MAX_DIRTY_DEFAULT (OSC_MAX_RIF_DEFAULT * 4)
377 #define OSC_MAX_DIRTY_MB_MAX 2048 /* arbitrary, but < MAX_LONG bytes */
378 #define OSC_DEFAULT_RESENDS 10
380 #define MDC_MAX_RIF_DEFAULT 8
381 #define MDC_MAX_RIF_MAX 512
386 cfs_rw_semaphore_t cl_sem;
387 struct obd_uuid cl_target_uuid;
388 struct obd_import *cl_import; /* ptlrpc connection state */
390 /* max_mds_easize is purely a performance thing so we don't have to
391 * call obd_size_diskmd() all the time. */
392 int cl_default_mds_easize;
393 int cl_max_mds_easize;
394 int cl_max_mds_cookiesize;
396 enum lustre_sec_part cl_sp_me;
397 enum lustre_sec_part cl_sp_to;
398 struct sptlrpc_flavor cl_flvr_mgc; /* fixed flavor of mgc->mgs */
400 //struct llog_canceld_ctxt *cl_llcd; /* it's included by obd_llog_ctxt */
401 void *cl_llcd_offset;
403 /* the grant values are protected by loi_list_lock below */
404 long cl_dirty; /* all _dirty_ in bytes */
405 long cl_dirty_max; /* allowed w/o rpc */
406 long cl_dirty_transit; /* dirty synchronous */
407 long cl_avail_grant; /* bytes of credit for ost */
408 long cl_lost_grant; /* lost credits (trunc) */
409 cfs_list_t cl_cache_waiters; /* waiting for cache/grant */
410 cfs_time_t cl_next_shrink_grant; /* jiffies */
411 cfs_list_t cl_grant_shrink_list; /* Timeout event list */
412 cfs_semaphore_t cl_grant_sem; /*grant shrink list cfs_semaphore*/
413 int cl_grant_shrink_interval; /* seconds */
415 /* keep track of objects that have lois that contain pages which
416 * have been queued for async brw. this lock also protects the
417 * lists of osc_client_pages that hang off of the loi */
419 * ->cl_loi_list_lock protects consistency of
420 * ->cl_loi_{ready,read,write}_list. ->ap_make_ready() and
421 * ->ap_completion() call-backs are executed under this lock. As we
422 * cannot guarantee that these call-backs never block on all platforms
423 * (as a matter of fact they do block on Mac OS X), type of
424 * ->cl_loi_list_lock is platform dependent: it's a spin-lock on Linux
425 * and blocking mutex on Mac OS X. (Alternative is to make this lock
426 * blocking everywhere, but we don't want to slow down fast-path of
427 * our main platform.)
429 * Exact type of ->cl_loi_list_lock is defined in arch/obd.h together
430 * with client_obd_list_{un,}lock() and
431 * client_obd_list_lock_{init,done}() functions.
433 client_obd_lock_t cl_loi_list_lock;
434 cfs_list_t cl_loi_ready_list;
435 cfs_list_t cl_loi_hp_ready_list;
436 cfs_list_t cl_loi_write_list;
437 cfs_list_t cl_loi_read_list;
440 /* just a sum of the loi/lop pending numbers to be exported by /proc */
441 int cl_pending_w_pages;
442 int cl_pending_r_pages;
443 int cl_max_pages_per_rpc;
444 int cl_max_rpcs_in_flight;
445 struct obd_histogram cl_read_rpc_hist;
446 struct obd_histogram cl_write_rpc_hist;
447 struct obd_histogram cl_read_page_hist;
448 struct obd_histogram cl_write_page_hist;
449 struct obd_histogram cl_read_offset_hist;
450 struct obd_histogram cl_write_offset_hist;
452 /* number of in flight destroy rpcs is limited to max_rpcs_in_flight */
453 cfs_atomic_t cl_destroy_in_flight;
454 cfs_waitq_t cl_destroy_waitq;
456 struct mdc_rpc_lock *cl_rpc_lock;
457 struct mdc_rpc_lock *cl_setattr_lock;
458 struct mdc_rpc_lock *cl_close_lock;
459 struct osc_creator cl_oscc;
462 cfs_semaphore_t cl_mgc_sem;
463 struct vfsmount *cl_mgc_vfsmnt;
464 struct dentry *cl_mgc_configs_dir;
465 cfs_atomic_t cl_mgc_refcount;
466 struct obd_export *cl_mgc_mgsexp;
468 /* checksumming for data sent over the network */
469 unsigned int cl_checksum:1; /* 0 = disabled, 1 = enabled */
470 /* supported checksum types that are worked out at connect time */
471 __u32 cl_supp_cksum_types;
472 /* checksum algorithm to be used */
473 cksum_type_t cl_cksum_type;
475 /* also protected by the poorly named _loi_list_lock lock above */
476 struct osc_async_rc cl_ar;
478 /* used by quotacheck */
479 int cl_qchk_stat; /* quotacheck stat of the peer */
481 /* sequence manager */
482 struct lu_client_seq *cl_seq;
484 cfs_atomic_t cl_resends; /* resend count */
486 #define obd2cli_tgt(obd) ((char *)(obd)->u.cli.cl_target_uuid.uuid)
488 #define CL_NOT_QUOTACHECKED 1 /* client->cl_qchk_stat init value */
491 struct ptlrpc_service *mgs_service;
492 struct vfsmount *mgs_vfsmnt;
493 struct super_block *mgs_sb;
494 struct dentry *mgs_configs_dir;
495 struct dentry *mgs_fid_de;
496 cfs_list_t mgs_fs_db_list;
497 cfs_semaphore_t mgs_sem;
498 cfs_proc_dir_entry_t *mgs_proc_live;
502 /* NB this field MUST be first */
503 struct obd_device_target mds_obt;
504 struct ptlrpc_service *mds_service;
505 struct ptlrpc_service *mds_setattr_service;
506 struct ptlrpc_service *mds_readpage_service;
507 cfs_dentry_t *mds_fid_de;
509 int mds_max_cookiesize;
511 unsigned long mds_atime_diff;
512 cfs_semaphore_t mds_epoch_sem;
513 struct ll_fid mds_rootfid;
514 cfs_dentry_t *mds_pending_dir;
515 cfs_dentry_t *mds_logs_dir;
516 cfs_dentry_t *mds_objects_dir;
517 struct llog_handle *mds_cfg_llh;
518 struct obd_device *mds_lov_obd;
519 struct obd_uuid mds_lov_uuid;
521 struct obd_export *mds_lov_exp;
522 struct lov_desc mds_lov_desc;
525 /* mark pages dirty for write. */
526 cfs_bitmap_t *mds_lov_page_dirty;
527 /* array for store pages with obd_id */
528 void **mds_lov_page_array;
529 /* file for store objid */
530 struct file *mds_lov_objid_filp;
531 __u32 mds_lov_objid_count;
532 __u32 mds_lov_objid_max_index;
533 __u32 mds_lov_objid_lastpage;
534 __u32 mds_lov_objid_lastidx;
537 struct lustre_quota_info mds_quota_info;
538 cfs_semaphore_t mds_qonoff_sem;
539 cfs_semaphore_t mds_health_sem;
540 unsigned long mds_fl_user_xattr:1,
542 mds_evict_ost_nids:1,
546 mds_fl_target:1; /* mds have one or
549 struct upcall_cache *mds_identity_cache;
551 /* for capability keys update */
552 struct lustre_capa_key *mds_capa_keys;
553 cfs_rw_semaphore_t mds_notify_lock;
556 #define mds_transno_lock mds_obt.obt_translock
557 #define mds_rcvd_filp mds_obt.obt_rcvd_filp
558 #define mds_server_data mds_obt.obt_lsd
559 #define mds_client_bitmap mds_obt.obt_client_bitmap
560 #define mds_mount_count mds_obt.obt_mount_count
561 #define mds_last_transno mds_obt.obt_last_transno
562 #define mds_vfsmnt mds_obt.obt_vfsmnt
565 extern __u32 mds_max_ost_index;
567 #define MDS_LOV_ALLOC_SIZE (CFS_PAGE_SIZE)
569 #define OBJID_PER_PAGE() (MDS_LOV_ALLOC_SIZE / sizeof(obd_id))
571 #define MDS_LOV_OBJID_PAGES_COUNT (mds_max_ost_index/OBJID_PER_PAGE())
573 extern int mds_lov_init_objids(struct obd_device *obd);
574 extern void mds_lov_destroy_objids(struct obd_device *obd);
585 cfs_spinlock_t eo_lock;
587 struct lustre_handle eo_nl_lock;
588 cfs_atomic_t eo_prep;
592 struct ptlrpc_service *ost_service;
593 struct ptlrpc_service *ost_create_service;
594 struct ptlrpc_service *ost_io_service;
595 cfs_semaphore_t ost_health_sem;
598 struct echo_client_obd {
599 struct obd_export *ec_exp; /* the local connection to osc/lov */
600 cfs_spinlock_t ec_lock;
601 cfs_list_t ec_objects;
608 struct obd_uuid lqo_uuid; /* ptlrpc's c_remote_uuid */
609 cfs_list_t lqo_oss_list; /* link to lov_qos */
610 __u64 lqo_bavail; /* total bytes avail on OSS */
611 __u64 lqo_penalty; /* current penalty */
612 __u64 lqo_penalty_per_obj;/* penalty decrease every obj*/
613 time_t lqo_used; /* last used time, seconds */
614 __u32 lqo_ost_count; /* number of osts on this oss */
618 struct lov_qos_oss *ltq_oss; /* oss info */
619 __u64 ltq_penalty; /* current penalty */
620 __u64 ltq_penalty_per_obj; /* penalty decrease every obj*/
621 __u64 ltq_weight; /* net weighting */
622 time_t ltq_used; /* last used time, seconds */
623 unsigned int ltq_usable:1; /* usable for striping */
626 /* Generic subset of OSTs */
628 __u32 *op_array; /* array of index of
630 unsigned int op_count; /* number of OSTs in the array */
631 unsigned int op_size; /* allocated size of lp_array */
632 cfs_rw_semaphore_t op_rw_sem; /* to protect ost_pool use */
635 /* Round-robin allocator data */
637 __u32 lqr_start_idx; /* start index of new inode */
638 __u32 lqr_offset_idx; /* aliasing for start_idx */
639 int lqr_start_count; /* reseed counter */
640 struct ost_pool lqr_pool; /* round-robin optimized list */
641 unsigned long lqr_dirty:1; /* recalc round-robin list */
644 struct lov_statfs_data {
645 struct obd_info lsd_oi;
646 struct obd_statfs lsd_statfs;
648 /* Stripe placement optimization */
650 cfs_list_t lq_oss_list; /* list of OSSs that targets use */
651 cfs_rw_semaphore_t lq_rw_sem;
652 __u32 lq_active_oss_count;
653 unsigned int lq_prio_free; /* priority for free space */
654 unsigned int lq_threshold_rr;/* priority for rr */
655 struct lov_qos_rr lq_rr; /* round robin qos data */
656 unsigned long lq_dirty:1, /* recalc qos data */
657 lq_same_space:1,/* the ost's all have approx.
658 the same space avail */
659 lq_reset:1, /* zero current penalties */
660 lq_statfs_in_progress:1; /* statfs op in
662 /* qos statfs data */
663 struct lov_statfs_data *lq_statfs_data;
664 cfs_waitq_t lq_statfs_waitq; /* waitqueue to notify statfs
665 * requests completion */
668 struct lov_tgt_desc {
670 struct obd_uuid ltd_uuid;
671 struct obd_device *ltd_obd;
672 struct obd_export *ltd_exp;
673 struct ltd_qos ltd_qos; /* qos info per target */
675 __u32 ltd_index; /* index in lov_obd->tgts */
676 unsigned long ltd_active:1,/* is this target up for requests */
677 ltd_activate:1,/* should target be activated */
678 ltd_reap:1; /* should this target be deleted */
682 #define pool_tgt_size(_p) _p->pool_obds.op_size
683 #define pool_tgt_count(_p) _p->pool_obds.op_count
684 #define pool_tgt_array(_p) _p->pool_obds.op_array
685 #define pool_tgt_rw_sem(_p) _p->pool_obds.op_rw_sem
686 #define pool_tgt(_p, _i) _p->pool_lov->lov_tgts[_p->pool_obds.op_array[_i]]
689 char pool_name[LOV_MAXPOOLNAME + 1]; /* name of pool */
690 struct ost_pool pool_obds; /* pool members */
691 cfs_atomic_t pool_refcount; /* pool ref. counter */
692 struct lov_qos_rr pool_rr; /* round robin qos */
693 cfs_hlist_node_t pool_hash; /* access by poolname */
694 cfs_list_t pool_list; /* serial access */
695 cfs_proc_dir_entry_t *pool_proc_entry; /* file in /proc */
696 struct lov_obd *pool_lov; /* lov obd to which this
701 struct lov_desc desc;
702 struct lov_tgt_desc **lov_tgts; /* sparse array */
703 struct ost_pool lov_packed; /* all OSTs in a packed
705 cfs_semaphore_t lov_lock;
706 struct obd_connect_data lov_ocd;
707 struct lov_qos lov_qos; /* qos info per lov */
708 cfs_atomic_t lov_refcount;
709 __u32 lov_tgt_count; /* how many OBD's */
710 __u32 lov_active_tgt_count; /* how many active */
711 __u32 lov_death_row;/* tgts scheduled to be deleted */
712 __u32 lov_tgt_size; /* size of tgts array */
715 cfs_hash_t *lov_pools_hash_body; /* used for key access */
716 cfs_list_t lov_pool_list; /* used for sequential access */
717 cfs_proc_dir_entry_t *lov_pool_proc_entry;
718 enum lustre_sec_part lov_sp_me;
721 struct lmv_tgt_desc {
722 struct obd_uuid ltd_uuid;
723 struct obd_export *ltd_exp;
724 int ltd_active; /* is this target up for requests */
726 cfs_semaphore_t ltd_fid_sem;
729 enum placement_policy {
730 PLACEMENT_CHAR_POLICY = 0,
731 PLACEMENT_NID_POLICY = 1,
732 PLACEMENT_INVAL_POLICY = 2,
736 typedef enum placement_policy placement_policy_t;
740 struct lu_client_fld lmv_fld;
741 cfs_spinlock_t lmv_lock;
742 placement_policy_t lmv_placement;
743 struct lmv_desc desc;
744 struct obd_uuid cluuid;
745 struct obd_export *exp;
752 cfs_semaphore_t init_sem;
754 struct lmv_tgt_desc *tgts;
757 struct obd_connect_data *datas;
760 struct obd_connect_data conn_data;
763 struct niobuf_local {
768 cfs_dentry_t *dentry;
773 #define LUSTRE_FLD_NAME "fld"
774 #define LUSTRE_SEQ_NAME "seq"
776 #define LUSTRE_CMM_NAME "cmm"
777 #define LUSTRE_MDD_NAME "mdd"
778 #define LUSTRE_OSD_NAME "osd"
779 #define LUSTRE_VVP_NAME "vvp"
780 #define LUSTRE_LMV_NAME "lmv"
781 #define LUSTRE_CMM_MDC_NAME "cmm-mdc"
782 #define LUSTRE_SLP_NAME "slp"
784 /* obd device type names */
785 /* FIXME all the references to LUSTRE_MDS_NAME should be swapped with LUSTRE_MDT_NAME */
786 #define LUSTRE_MDS_NAME "mds"
787 #define LUSTRE_MDT_NAME "mdt"
788 #define LUSTRE_MDC_NAME "mdc"
789 #define LUSTRE_OSS_NAME "ost" /* FIXME change name to oss */
790 #define LUSTRE_OST_NAME "obdfilter" /* FIXME change name to ost */
791 #define LUSTRE_OSC_NAME "osc"
792 #define LUSTRE_LOV_NAME "lov"
793 #define LUSTRE_MGS_NAME "mgs"
794 #define LUSTRE_MGC_NAME "mgc"
796 #define LUSTRE_CACHEOBD_NAME "cobd"
797 #define LUSTRE_ECHO_NAME "obdecho"
798 #define LUSTRE_ECHO_CLIENT_NAME "echo_client"
800 /* Constant obd names (post-rename) */
801 #define LUSTRE_MDS_OBDNAME "MDS"
802 #define LUSTRE_OSS_OBDNAME "OSS"
803 #define LUSTRE_MGS_OBDNAME "MGS"
804 #define LUSTRE_MGC_OBDNAME "MGC"
806 /* Don't conflict with on-wire flags OBD_BRW_WRITE, etc */
807 #define N_LOCAL_TEMP_PAGE 0x10000000
809 struct obd_trans_info {
812 /* Only used on the server side for tracking acks. */
813 struct oti_req_ack_lock {
814 struct lustre_handle lock;
818 struct llog_cookie oti_onecookie;
819 struct llog_cookie *oti_logcookies;
822 /* initial thread handling transaction */
823 struct ptlrpc_thread * oti_thread;
826 __u64 oti_pre_version;
828 struct obd_uuid *oti_ost_uuid;
831 static inline void oti_init(struct obd_trans_info *oti,
832 struct ptlrpc_request *req)
836 memset(oti, 0, sizeof(*oti));
841 oti->oti_xid = req->rq_xid;
842 /** VBR: take versions from request */
843 if (req->rq_reqmsg != NULL &&
844 lustre_msg_get_flags(req->rq_reqmsg) & MSG_REPLAY) {
845 __u64 *pre_version = lustre_msg_get_versions(req->rq_reqmsg);
846 oti->oti_pre_version = pre_version ? pre_version[0] : 0;
847 oti->oti_transno = lustre_msg_get_transno(req->rq_reqmsg);
850 /** called from mds_create_objects */
851 if (req->rq_repmsg != NULL)
852 oti->oti_transno = lustre_msg_get_transno(req->rq_repmsg);
853 oti->oti_thread = req->rq_svc_thread;
854 if (req->rq_reqmsg != NULL)
855 oti->oti_conn_cnt = lustre_msg_get_conn_cnt(req->rq_reqmsg);
858 static inline void oti_alloc_cookies(struct obd_trans_info *oti,int num_cookies)
863 if (num_cookies == 1)
864 oti->oti_logcookies = &oti->oti_onecookie;
866 OBD_ALLOC(oti->oti_logcookies,
867 num_cookies * sizeof(oti->oti_onecookie));
869 oti->oti_numcookies = num_cookies;
872 static inline void oti_free_cookies(struct obd_trans_info *oti)
874 if (!oti || !oti->oti_logcookies)
877 if (oti->oti_logcookies == &oti->oti_onecookie)
878 LASSERT(oti->oti_numcookies == 1);
880 OBD_FREE(oti->oti_logcookies,
881 oti->oti_numcookies * sizeof(oti->oti_onecookie));
882 oti->oti_logcookies = NULL;
883 oti->oti_numcookies = 0;
887 * Events signalled through obd_notify() upcall-chain.
889 enum obd_notify_event {
890 /* Device connect start */
892 /* Device activated */
894 /* Device deactivated */
896 /* Device disconnected */
898 /* Connect data for import were changed */
901 OBD_NOTIFY_SYNC_NONBLOCK,
903 /* Configuration event */
905 /* Trigger quota recovery */
909 /* bit-mask flags for config events */
911 CONFIG_LOG = 0x1, /* finished processing config log */
912 CONFIG_SYNC = 0x2, /* mdt synced 1 ost */
913 CONFIG_TARGET = 0x4 /* one target is added */
917 * Data structure used to pass obd_notify()-event to non-obd listeners (llite
918 * and liblustre being main examples).
920 struct obd_notify_upcall {
921 int (*onu_upcall)(struct obd_device *host, struct obd_device *watched,
922 enum obd_notify_event ev, void *owner, void *data);
923 /* Opaque datum supplied by upper layer listener */
927 struct target_recovery_data {
928 svc_handler_t trd_recovery_handler;
929 pid_t trd_processing_task;
930 cfs_completion_t trd_starting;
931 cfs_completion_t trd_finishing;
935 * In HEAD for CMD, the object is created in group number which is 3>=
936 * or indexing starts from 3. To test this assertions are added to disallow
937 * group 0. But to run 2.0 mds server on 1.8.x disk format (i.e. interop_mode)
938 * object in group 0 needs to be allowed.
939 * So for interop mode following changes needs to be done:
940 * 1. No need to assert on group 0 or allow group 0
941 * 2. The group number indexing starts from 0 instead of 3
944 #define LASSERT_SEQ_IS_MDT(seq) LASSERT(fid_seq_is_mdt(seq))
946 static inline __u64 objseq_to_mdsno(obd_seq seq)
948 LASSERT_SEQ_IS_MDT(seq);
949 if (seq == FID_SEQ_OST_MDT0)
951 return seq - FID_SEQ_OST_MDT1 + 1;
954 static inline int mdt_to_obd_objseq(int mdtid)
957 * MDS0 uses seq 0 pre FID-on-OST, other MDSes will use seq from
961 return FID_SEQ_OST_MDT1 + mdtid - 1;
965 struct obd_llog_group {
968 struct llog_ctxt *olg_ctxts[LLOG_MAX_CTXTS];
969 cfs_waitq_t olg_waitq;
970 cfs_spinlock_t olg_lock;
971 struct obd_export *olg_exp;
972 int olg_initializing;
973 cfs_semaphore_t olg_cat_processing;
976 /* corresponds to one of the obd's */
977 #define MAX_OBD_NAME 128
978 #define OBD_DEVICE_MAGIC 0XAB5CD6EF
979 #define OBD_DEV_BY_DEVNAME 0xffffd0de
981 struct obd_type *obd_type;
984 /* common and UUID name of this device */
985 char obd_name[MAX_OBD_NAME];
986 struct obd_uuid obd_uuid;
988 struct lu_device *obd_lu_dev;
991 unsigned long obd_attached:1, /* finished attach */
992 obd_set_up:1, /* finished setup */
993 obd_recovering:1, /* there are recoverable clients */
994 obd_abort_recovery:1,/* recovery expired */
995 obd_version_recov:1, /* obd uses version checking */
996 obd_recovery_expired:1,
997 obd_replayable:1, /* recovery is enabled; inform clients */
998 obd_no_transno:1, /* no committed-transno notification */
999 obd_no_recov:1, /* fail instead of retry messages */
1000 obd_req_replaying:1, /* replaying requests */
1001 obd_stopping:1, /* started cleanup */
1002 obd_starting:1, /* started setup */
1003 obd_force:1, /* cleanup with > 0 obd refcount */
1004 obd_fail:1, /* cleanup with failover */
1005 obd_async_recov:1, /* allow asynchronous orphan cleanup */
1006 obd_no_conn:1, /* deny new connections */
1007 obd_inactive:1, /* device active/inactive
1008 * (for /proc/status only!!) */
1009 obd_process_conf:1; /* device is processing mgs config */
1010 /* uuid-export hash body */
1011 cfs_hash_t *obd_uuid_hash;
1012 /* nid-export hash body */
1013 cfs_hash_t *obd_nid_hash;
1014 /* nid stats body */
1015 cfs_hash_t *obd_nid_stats_hash;
1016 cfs_list_t obd_nid_stats;
1017 cfs_atomic_t obd_refcount;
1018 cfs_waitq_t obd_refcount_waitq;
1019 cfs_list_t obd_exports;
1020 cfs_list_t obd_unlinked_exports;
1021 cfs_list_t obd_delayed_exports;
1022 int obd_num_exports;
1023 cfs_spinlock_t obd_nid_lock;
1024 struct ldlm_namespace *obd_namespace;
1025 struct ptlrpc_client obd_ldlm_client; /* XXX OST/MDS only */
1026 /* a spinlock is OK for what we do now, may need a semaphore later */
1027 cfs_spinlock_t obd_dev_lock;
1028 cfs_semaphore_t obd_dev_sem;
1029 __u64 obd_last_committed;
1030 struct fsfilt_operations *obd_fsops;
1031 cfs_spinlock_t obd_osfs_lock;
1032 struct obd_statfs obd_osfs; /* locked by obd_osfs_lock */
1034 struct lvfs_run_ctxt obd_lvfs_ctxt;
1035 struct obd_llog_group obd_olg; /* default llog group */
1036 struct obd_device *obd_observer;
1037 cfs_rw_semaphore_t obd_observer_link_sem;
1038 struct obd_notify_upcall obd_upcall;
1039 struct obd_export *obd_self_export;
1040 /* list of exports in LRU order, for ping evictor, with obd_dev_lock */
1041 cfs_list_t obd_exports_timed;
1042 time_t obd_eviction_timer; /* for ping evictor */
1044 int obd_max_recoverable_clients;
1045 int obd_connected_clients;
1046 int obd_stale_clients;
1047 int obd_delayed_clients;
1048 cfs_spinlock_t obd_processing_task_lock; /* BH lock (timer) */
1049 __u64 obd_next_recovery_transno;
1050 int obd_replayed_requests;
1051 int obd_requests_queued_for_recovery;
1052 cfs_waitq_t obd_next_transno_waitq;
1053 cfs_timer_t obd_recovery_timer;
1054 time_t obd_recovery_start; /* seconds */
1055 time_t obd_recovery_end; /* seconds, for lprocfs_status */
1056 time_t obd_recovery_time_hard;
1057 int obd_recovery_timeout;
1059 /* new recovery stuff from CMD2 */
1060 struct target_recovery_data obd_recovery_data;
1061 int obd_replayed_locks;
1062 cfs_atomic_t obd_req_replay_clients;
1063 cfs_atomic_t obd_lock_replay_clients;
1064 cfs_list_t obd_req_replay_queue;
1065 cfs_list_t obd_lock_replay_queue;
1066 cfs_list_t obd_final_req_queue;
1067 int obd_recovery_stage;
1070 struct obd_device_target obt;
1071 struct filter_obd filter;
1073 struct client_obd cli;
1075 struct echo_client_obd echo_client;
1076 struct echo_obd echo;
1081 /* Fields used by LProcFS */
1082 unsigned int obd_cntr_base;
1083 struct lprocfs_stats *obd_stats;
1085 unsigned int md_cntr_base;
1086 struct lprocfs_stats *md_stats;
1088 cfs_proc_dir_entry_t *obd_proc_entry;
1089 cfs_proc_dir_entry_t *obd_proc_exports_entry;
1090 cfs_proc_dir_entry_t *obd_svc_procroot;
1091 struct lprocfs_stats *obd_svc_stats;
1092 cfs_atomic_t obd_evict_inprogress;
1093 cfs_waitq_t obd_evict_inprogress_waitq;
1094 cfs_list_t obd_evict_list; /* protected with pet_lock */
1097 * Ldlm pool part. Save last calculated SLV and Limit.
1099 cfs_rwlock_t obd_pool_lock;
1104 * A list of outstanding class_incref()'s against this obd. For
1107 struct lu_ref obd_reference;
1110 #define OBD_LLOG_FL_SENDNOW 0x0001
1112 enum obd_cleanup_stage {
1113 /* Special case hack for MDS LOVs */
1115 /* can be directly mapped to .ldto_device_fini() */
1116 OBD_CLEANUP_EXPORTS,
1119 /* get/set_info keys */
1120 #define KEY_ASYNC "async"
1121 #define KEY_BLOCKSIZE_BITS "blocksize_bits"
1122 #define KEY_BLOCKSIZE "blocksize"
1123 #define KEY_CAPA_KEY "capa_key"
1124 #define KEY_CHANGELOG_CLEAR "changelog_clear"
1125 #define KEY_FID2PATH "fid2path"
1126 #define KEY_CHECKSUM "checksum"
1127 #define KEY_CLEAR_FS "clear_fs"
1128 #define KEY_CONN_DATA "conn_data"
1129 #define KEY_EVICT_BY_NID "evict_by_nid"
1130 #define KEY_FIEMAP "fiemap"
1131 #define KEY_FLUSH_CTX "flush_ctx"
1132 #define KEY_GRANT_SHRINK "grant_shrink"
1133 #define KEY_HSM_COPYTOOL_SEND "hsm_send"
1134 #define KEY_INIT_RECOV_BACKUP "init_recov_bk"
1135 #define KEY_INIT_RECOV "initial_recov"
1136 #define KEY_INTERMDS "inter_mds"
1137 #define KEY_LAST_ID "last_id"
1138 #define KEY_LOCK_TO_STRIPE "lock_to_stripe"
1139 #define KEY_LOVDESC "lovdesc"
1140 #define KEY_LOV_IDX "lov_idx"
1141 #define KEY_MAX_EASIZE "max_easize"
1142 #define KEY_MDS_CONN "mds_conn"
1143 #define KEY_MGSSEC "mgssec"
1144 #define KEY_NEXT_ID "next_id"
1145 #define KEY_READ_ONLY "read-only"
1146 #define KEY_REGISTER_TARGET "register_target"
1147 #define KEY_REVIMP_UPD "revimp_update"
1148 #define KEY_SET_FS "set_fs"
1149 /* KEY_SET_INFO in lustre_idl.h */
1150 #define KEY_SPTLRPC_CONF "sptlrpc_conf"
1151 #define KEY_CONNECT_FLAG "connect_flags"
1156 static inline int it_to_lock_mode(struct lookup_intent *it)
1158 /* CREAT needs to be tested before open (both could be set) */
1159 if (it->it_op & IT_CREAT)
1161 else if (it->it_op & (IT_READDIR | IT_GETATTR | IT_OPEN | IT_LOOKUP))
1164 LASSERTF(0, "Invalid it_op: %d\n", it->it_op);
1169 struct lu_fid op_fid1; /* operation fid1 (usualy parent) */
1170 struct lu_fid op_fid2; /* operation fid2 (usualy child) */
1171 struct lu_fid op_fid3; /* 2 extra fids to find conflicting */
1172 struct lu_fid op_fid4; /* to the operation locks. */
1173 mdsno_t op_mds; /* what mds server open will go to */
1174 struct lustre_handle op_handle;
1176 const char *op_name;
1179 struct lmv_stripe_md *op_mea1;
1180 struct lmv_stripe_md *op_mea2;
1181 __u32 op_suppgids[2];
1187 /* iattr fields and blocks. */
1188 struct iattr op_attr;
1190 #if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,14)
1191 unsigned int op_attr_flags;
1195 loff_t op_attr_blocks;
1197 /* Size-on-MDS epoch and flags. */
1202 struct obd_capa *op_capa1;
1203 struct obd_capa *op_capa2;
1205 /* Various operation flags. */
1208 /* Operation type */
1212 struct md_enqueue_info;
1213 /* metadata stat-ahead */
1214 typedef int (* md_enqueue_cb_t)(struct ptlrpc_request *req,
1215 struct md_enqueue_info *minfo,
1218 struct md_enqueue_info {
1219 struct md_op_data mi_data;
1220 struct lookup_intent mi_it;
1221 struct lustre_handle mi_lockh;
1222 struct dentry *mi_dentry;
1223 struct inode *mi_dir;
1224 md_enqueue_cb_t mi_cb;
1225 unsigned int mi_generation;
1230 cfs_module_t *o_owner;
1231 int (*o_iocontrol)(unsigned int cmd, struct obd_export *exp, int len,
1232 void *karg, void *uarg);
1233 int (*o_get_info)(struct obd_export *, __u32 keylen, void *key,
1234 __u32 *vallen, void *val, struct lov_stripe_md *lsm);
1235 int (*o_set_info_async)(struct obd_export *, __u32 keylen, void *key,
1236 __u32 vallen, void *val,
1237 struct ptlrpc_request_set *set);
1238 int (*o_attach)(struct obd_device *dev, obd_count len, void *data);
1239 int (*o_detach)(struct obd_device *dev);
1240 int (*o_setup) (struct obd_device *dev, struct lustre_cfg *cfg);
1241 int (*o_precleanup)(struct obd_device *dev,
1242 enum obd_cleanup_stage cleanup_stage);
1243 int (*o_cleanup)(struct obd_device *dev);
1244 int (*o_process_config)(struct obd_device *dev, obd_count len,
1246 int (*o_postrecov)(struct obd_device *dev);
1247 int (*o_add_conn)(struct obd_import *imp, struct obd_uuid *uuid,
1249 int (*o_del_conn)(struct obd_import *imp, struct obd_uuid *uuid);
1250 /* connect to the target device with given connection
1251 * data. @ocd->ocd_connect_flags is modified to reflect flags actually
1252 * granted by the target, which are guaranteed to be a subset of flags
1253 * asked for. If @ocd == NULL, use default parameters. */
1254 int (*o_connect)(const struct lu_env *env,
1255 struct obd_export **exp, struct obd_device *src,
1256 struct obd_uuid *cluuid, struct obd_connect_data *ocd,
1258 int (*o_reconnect)(const struct lu_env *env,
1259 struct obd_export *exp, struct obd_device *src,
1260 struct obd_uuid *cluuid,
1261 struct obd_connect_data *ocd,
1263 int (*o_disconnect)(struct obd_export *exp);
1265 /* Initialize/finalize fids infrastructure. */
1266 int (*o_fid_init)(struct obd_export *exp);
1267 int (*o_fid_fini)(struct obd_export *exp);
1269 /* Allocate new fid according to passed @hint. */
1270 int (*o_fid_alloc)(struct obd_export *exp, struct lu_fid *fid,
1271 struct md_op_data *op_data);
1274 * Object with @fid is getting deleted, we may want to do something
1277 int (*o_fid_delete)(struct obd_export *exp, const struct lu_fid *fid);
1279 int (*o_statfs)(struct obd_device *obd, struct obd_statfs *osfs,
1280 __u64 max_age, __u32 flags);
1281 int (*o_statfs_async)(struct obd_device *obd, struct obd_info *oinfo,
1282 __u64 max_age, struct ptlrpc_request_set *set);
1283 int (*o_packmd)(struct obd_export *exp, struct lov_mds_md **disk_tgt,
1284 struct lov_stripe_md *mem_src);
1285 int (*o_unpackmd)(struct obd_export *exp,struct lov_stripe_md **mem_tgt,
1286 struct lov_mds_md *disk_src, int disk_len);
1287 int (*o_preallocate)(struct lustre_handle *, obd_count *req,
1289 /* FIXME: add fid capability support for create & destroy! */
1290 int (*o_precreate)(struct obd_export *exp);
1291 int (*o_create)(struct obd_export *exp, struct obdo *oa,
1292 struct lov_stripe_md **ea, struct obd_trans_info *oti);
1293 int (*o_create_async)(struct obd_export *exp, struct obd_info *oinfo,
1294 struct lov_stripe_md **ea,
1295 struct obd_trans_info *oti);
1296 int (*o_destroy)(struct obd_export *exp, struct obdo *oa,
1297 struct lov_stripe_md *ea, struct obd_trans_info *oti,
1298 struct obd_export *md_exp, void *capa);
1299 int (*o_setattr)(struct obd_export *exp, struct obd_info *oinfo,
1300 struct obd_trans_info *oti);
1301 int (*o_setattr_async)(struct obd_export *exp, struct obd_info *oinfo,
1302 struct obd_trans_info *oti,
1303 struct ptlrpc_request_set *rqset);
1304 int (*o_getattr)(struct obd_export *exp, struct obd_info *oinfo);
1305 int (*o_getattr_async)(struct obd_export *exp, struct obd_info *oinfo,
1306 struct ptlrpc_request_set *set);
1307 int (*o_brw)(int rw, struct obd_export *exp, struct obd_info *oinfo,
1308 obd_count oa_bufs, struct brw_page *pgarr,
1309 struct obd_trans_info *oti);
1310 int (*o_merge_lvb)(struct obd_export *exp, struct lov_stripe_md *lsm,
1311 struct ost_lvb *lvb, int kms_only);
1312 int (*o_adjust_kms)(struct obd_export *exp, struct lov_stripe_md *lsm,
1313 obd_off size, int shrink);
1314 int (*o_punch)(struct obd_export *exp, struct obd_info *oinfo,
1315 struct obd_trans_info *oti,
1316 struct ptlrpc_request_set *rqset);
1317 int (*o_sync)(struct obd_export *exp, struct obdo *oa,
1318 struct lov_stripe_md *ea, obd_size start, obd_size end,
1320 int (*o_migrate)(struct lustre_handle *conn, struct lov_stripe_md *dst,
1321 struct lov_stripe_md *src, obd_size start,
1322 obd_size end, struct obd_trans_info *oti);
1323 int (*o_copy)(struct lustre_handle *dstconn, struct lov_stripe_md *dst,
1324 struct lustre_handle *srconn, struct lov_stripe_md *src,
1325 obd_size start, obd_size end, struct obd_trans_info *);
1326 int (*o_iterate)(struct lustre_handle *conn,
1327 int (*)(obd_id, obd_seq, void *),
1328 obd_id *startid, obd_seq seq, void *data);
1329 int (*o_preprw)(int cmd, struct obd_export *exp, struct obdo *oa,
1330 int objcount, struct obd_ioobj *obj,
1331 struct niobuf_remote *remote, int *nr_pages,
1332 struct niobuf_local *local,
1333 struct obd_trans_info *oti,
1334 struct lustre_capa *capa);
1335 int (*o_commitrw)(int cmd, struct obd_export *exp, struct obdo *oa,
1336 int objcount, struct obd_ioobj *obj,
1337 struct niobuf_remote *remote, int pages,
1338 struct niobuf_local *local,
1339 struct obd_trans_info *oti, int rc);
1340 int (*o_enqueue)(struct obd_export *, struct obd_info *oinfo,
1341 struct ldlm_enqueue_info *einfo,
1342 struct ptlrpc_request_set *rqset);
1343 int (*o_change_cbdata)(struct obd_export *, struct lov_stripe_md *,
1344 ldlm_iterator_t it, void *data);
1345 int (*o_find_cbdata)(struct obd_export *, struct lov_stripe_md *,
1346 ldlm_iterator_t it, void *data);
1347 int (*o_cancel)(struct obd_export *, struct lov_stripe_md *md,
1348 __u32 mode, struct lustre_handle *);
1349 int (*o_cancel_unused)(struct obd_export *, struct lov_stripe_md *,
1350 int flags, void *opaque);
1351 int (*o_init_export)(struct obd_export *exp);
1352 int (*o_destroy_export)(struct obd_export *exp);
1353 int (*o_extent_calc)(struct obd_export *, struct lov_stripe_md *,
1354 int cmd, obd_off *);
1356 /* llog related obd_methods */
1357 int (*o_llog_init)(struct obd_device *obd, struct obd_llog_group *grp,
1358 struct obd_device *disk_obd, int *idx);
1359 int (*o_llog_finish)(struct obd_device *obd, int count);
1360 int (*o_llog_connect)(struct obd_export *, struct llogd_conn_body *);
1362 /* metadata-only methods */
1363 int (*o_pin)(struct obd_export *, const struct lu_fid *fid,
1364 struct obd_capa *, struct obd_client_handle *, int flag);
1365 int (*o_unpin)(struct obd_export *, struct obd_client_handle *, int);
1367 int (*o_import_event)(struct obd_device *, struct obd_import *,
1368 enum obd_import_event);
1370 int (*o_notify)(struct obd_device *obd, struct obd_device *watched,
1371 enum obd_notify_event ev, void *data);
1373 int (*o_health_check)(struct obd_device *);
1374 struct obd_uuid *(*o_get_uuid) (struct obd_export *exp);
1377 int (*o_quotacheck)(struct obd_device *, struct obd_export *,
1378 struct obd_quotactl *);
1379 int (*o_quotactl)(struct obd_device *, struct obd_export *,
1380 struct obd_quotactl *);
1381 int (*o_quota_adjust_qunit)(struct obd_export *exp,
1382 struct quota_adjust_qunit *oqaq,
1383 struct lustre_quota_ctxt *qctxt);
1386 int (*o_ping)(struct obd_export *exp);
1389 int (*o_pool_new)(struct obd_device *obd, char *poolname);
1390 int (*o_pool_del)(struct obd_device *obd, char *poolname);
1391 int (*o_pool_add)(struct obd_device *obd, char *poolname,
1393 int (*o_pool_rem)(struct obd_device *obd, char *poolname,
1395 void (*o_getref)(struct obd_device *obd);
1396 void (*o_putref)(struct obd_device *obd);
1398 * NOTE: If adding ops, add another LPROCFS_OBD_OP_INIT() line
1399 * to lprocfs_alloc_obd_stats() in obdclass/lprocfs_status.c.
1400 * Also, add a wrapper function in include/linux/obd_class.h. */
1404 LUSTRE_OPC_MKDIR = (1 << 0),
1405 LUSTRE_OPC_SYMLINK = (1 << 1),
1406 LUSTRE_OPC_MKNOD = (1 << 2),
1407 LUSTRE_OPC_CREATE = (1 << 3),
1408 LUSTRE_OPC_ANY = (1 << 4)
1411 /* lmv structures */
1412 #define MEA_MAGIC_LAST_CHAR 0xb2221ca1
1413 #define MEA_MAGIC_ALL_CHARS 0xb222a11c
1414 #define MEA_MAGIC_HASH_SEGMENT 0xb222a11b
1416 #define MAX_HASH_SIZE_32 0x7fffffffUL
1417 #define MAX_HASH_SIZE 0x7fffffffffffffffULL
1418 #define MAX_HASH_HIGHEST_BIT 0x1000000000000000ULL
1421 struct mdt_body *body;
1422 struct lov_stripe_md *lsm;
1423 struct lmv_stripe_md *mea;
1424 #ifdef CONFIG_FS_POSIX_ACL
1425 struct posix_acl *posix_acl;
1427 struct mdt_remote_perm *remote_perm;
1428 struct obd_capa *mds_capa;
1429 struct obd_capa *oss_capa;
1432 struct md_open_data {
1433 struct obd_client_handle *mod_och;
1434 struct ptlrpc_request *mod_open_req;
1435 struct ptlrpc_request *mod_close_req;
1436 cfs_atomic_t mod_refcount;
1439 struct lookup_intent;
1442 int (*m_getstatus)(struct obd_export *, struct lu_fid *,
1443 struct obd_capa **);
1444 int (*m_change_cbdata)(struct obd_export *, const struct lu_fid *,
1445 ldlm_iterator_t, void *);
1446 int (*m_find_cbdata)(struct obd_export *, const struct lu_fid *,
1447 ldlm_iterator_t, void *);
1448 int (*m_close)(struct obd_export *, struct md_op_data *,
1449 struct md_open_data *, struct ptlrpc_request **);
1450 int (*m_create)(struct obd_export *, struct md_op_data *,
1451 const void *, int, int, __u32, __u32, cfs_cap_t,
1452 __u64, struct ptlrpc_request **);
1453 int (*m_done_writing)(struct obd_export *, struct md_op_data *,
1454 struct md_open_data *);
1455 int (*m_enqueue)(struct obd_export *, struct ldlm_enqueue_info *,
1456 struct lookup_intent *, struct md_op_data *,
1457 struct lustre_handle *, void *, int,
1458 struct ptlrpc_request **, int);
1459 int (*m_getattr)(struct obd_export *, struct md_op_data *,
1460 struct ptlrpc_request **);
1461 int (*m_getattr_name)(struct obd_export *, struct md_op_data *,
1462 struct ptlrpc_request **);
1463 int (*m_intent_lock)(struct obd_export *, struct md_op_data *,
1464 void *, int, struct lookup_intent *, int,
1465 struct ptlrpc_request **,
1466 ldlm_blocking_callback, int);
1467 int (*m_link)(struct obd_export *, struct md_op_data *,
1468 struct ptlrpc_request **);
1469 int (*m_rename)(struct obd_export *, struct md_op_data *,
1470 const char *, int, const char *, int,
1471 struct ptlrpc_request **);
1472 int (*m_is_subdir)(struct obd_export *, const struct lu_fid *,
1473 const struct lu_fid *,
1474 struct ptlrpc_request **);
1475 int (*m_setattr)(struct obd_export *, struct md_op_data *, void *,
1476 int , void *, int, struct ptlrpc_request **,
1477 struct md_open_data **mod);
1478 int (*m_sync)(struct obd_export *, const struct lu_fid *,
1479 struct obd_capa *, struct ptlrpc_request **);
1480 int (*m_readpage)(struct obd_export *, const struct lu_fid *,
1481 struct obd_capa *, __u64, struct page *,
1482 struct ptlrpc_request **);
1484 int (*m_unlink)(struct obd_export *, struct md_op_data *,
1485 struct ptlrpc_request **);
1487 int (*m_setxattr)(struct obd_export *, const struct lu_fid *,
1488 struct obd_capa *, obd_valid, const char *,
1489 const char *, int, int, int, __u32,
1490 struct ptlrpc_request **);
1492 int (*m_getxattr)(struct obd_export *, const struct lu_fid *,
1493 struct obd_capa *, obd_valid, const char *,
1494 const char *, int, int, int,
1495 struct ptlrpc_request **);
1497 int (*m_init_ea_size)(struct obd_export *, int, int, int);
1499 int (*m_get_lustre_md)(struct obd_export *, struct ptlrpc_request *,
1500 struct obd_export *, struct obd_export *,
1501 struct lustre_md *);
1503 int (*m_free_lustre_md)(struct obd_export *, struct lustre_md *);
1505 int (*m_set_open_replay_data)(struct obd_export *,
1506 struct obd_client_handle *,
1507 struct ptlrpc_request *);
1508 int (*m_clear_open_replay_data)(struct obd_export *,
1509 struct obd_client_handle *);
1510 int (*m_set_lock_data)(struct obd_export *, __u64 *, void *, __u32 *);
1512 ldlm_mode_t (*m_lock_match)(struct obd_export *, int,
1513 const struct lu_fid *, ldlm_type_t,
1514 ldlm_policy_data_t *, ldlm_mode_t,
1515 struct lustre_handle *);
1517 int (*m_cancel_unused)(struct obd_export *, const struct lu_fid *,
1518 ldlm_policy_data_t *, ldlm_mode_t, int flags,
1520 int (*m_renew_capa)(struct obd_export *, struct obd_capa *oc,
1521 renew_capa_cb_t cb);
1522 int (*m_unpack_capa)(struct obd_export *, struct ptlrpc_request *,
1523 const struct req_msg_field *, struct obd_capa **);
1525 int (*m_get_remote_perm)(struct obd_export *, const struct lu_fid *,
1526 struct obd_capa *, __u32,
1527 struct ptlrpc_request **);
1529 int (*m_intent_getattr_async)(struct obd_export *,
1530 struct md_enqueue_info *,
1531 struct ldlm_enqueue_info *);
1533 int (*m_revalidate_lock)(struct obd_export *, struct lookup_intent *,
1534 struct lu_fid *, __u32 *);
1537 * NOTE: If adding ops, add another LPROCFS_MD_OP_INIT() line to
1538 * lprocfs_alloc_md_stats() in obdclass/lprocfs_status.c. Also, add a
1539 * wrapper function in include/linux/obd_class.h.
1543 struct lsm_operations {
1544 void (*lsm_free)(struct lov_stripe_md *);
1545 int (*lsm_destroy)(struct lov_stripe_md *, struct obdo *oa,
1546 struct obd_export *md_exp);
1547 void (*lsm_stripe_by_index)(struct lov_stripe_md *, int *, obd_off *,
1549 void (*lsm_stripe_by_offset)(struct lov_stripe_md *, int *, obd_off *,
1551 int (*lsm_lmm_verify) (struct lov_mds_md *lmm, int lmm_bytes,
1553 int (*lsm_unpackmd) (struct lov_obd *lov, struct lov_stripe_md *lsm,
1554 struct lov_mds_md *lmm);
1557 extern const struct lsm_operations lsm_v1_ops;
1558 extern const struct lsm_operations lsm_v3_ops;
1559 static inline const struct lsm_operations *lsm_op_find(int magic)
1567 CERROR("Cannot recognize lsm_magic %08x\n", magic);
1572 int lvfs_check_io_health(struct obd_device *obd, struct file *file);
1574 /* Requests for obd_extent_calc() */
1575 #define OBD_CALC_STRIPE_START 1
1576 #define OBD_CALC_STRIPE_END 2
1578 static inline void obd_transno_commit_cb(struct obd_device *obd, __u64 transno,
1579 struct obd_export *exp, int error)
1582 CERROR("%s: transno "LPU64" commit error: %d\n",
1583 obd->obd_name, transno, error);
1586 if (exp && transno > exp->exp_last_committed) {
1587 CDEBUG(D_HA, "%s: transno "LPU64" committed\n",
1588 obd->obd_name, transno);
1589 exp->exp_last_committed = transno;
1590 ptlrpc_commit_replies(exp);
1592 CDEBUG(D_INFO, "%s: transno "LPU64" committed\n",
1593 obd->obd_name, transno);
1595 if (transno > obd->obd_last_committed)
1596 obd->obd_last_committed = transno;
1599 static inline void init_obd_quota_ops(quota_interface_t *interface,
1600 struct obd_ops *obd_ops)
1606 obd_ops->o_quotacheck = QUOTA_OP(interface, check);
1607 obd_ops->o_quotactl = QUOTA_OP(interface, ctl);
1608 obd_ops->o_quota_adjust_qunit = QUOTA_OP(interface, adjust_qunit);
1611 static inline struct lustre_capa *oinfo_capa(struct obd_info *oinfo)
1613 return oinfo->oi_capa;
1616 static inline struct md_open_data *obd_mod_alloc(void)
1618 struct md_open_data *mod;
1622 cfs_atomic_set(&mod->mod_refcount, 1);
1626 #define obd_mod_get(mod) cfs_atomic_inc(&(mod)->mod_refcount)
1627 #define obd_mod_put(mod) \
1629 if (cfs_atomic_dec_and_test(&(mod)->mod_refcount)) { \
1630 if ((mod)->mod_open_req) \
1631 ptlrpc_req_finished((mod)->mod_open_req); \
1632 OBD_FREE_PTR(mod); \
1636 extern void obdo_from_inode(struct obdo *dst, struct inode *src,
1637 struct lu_fid *parent, obd_flag valid);
1639 #endif /* __OBD_H */