1 /* -*- mode: c; c-basic-offset: 8; indent-tabs-mode: nil; -*-
2 * vim:expandtab:shiftwidth=8:tabstop=8:
6 * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
8 * This program is free software; you can redistribute it and/or modify
9 * it under the terms of the GNU General Public License version 2 only,
10 * as published by the Free Software Foundation.
12 * This program is distributed in the hope that it will be useful, but
13 * WITHOUT ANY WARRANTY; without even the implied warranty of
14 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
15 * General Public License version 2 for more details (a copy is included
16 * in the LICENSE file that accompanied this code).
18 * You should have received a copy of the GNU General Public License
19 * version 2 along with this program; If not, see
20 * http://www.sun.com/software/products/lustre/docs/GPLv2.pdf
22 * Please contact Sun Microsystems, Inc., 4150 Network Circle, Santa Clara,
23 * CA 95054 USA or visit www.sun.com if you need additional information or
29 * Copyright 2008 Sun Microsystems, Inc. All rights reserved
30 * Use is subject to license terms.
33 * This file is part of Lustre, http://www.lustre.org/
34 * Lustre is a trademark of Sun Microsystems, Inc.
40 #if defined(__linux__)
41 #include <linux/obd.h>
42 #elif defined(__APPLE__)
43 #include <darwin/obd.h>
44 #elif defined(__WINNT__)
45 #include <winnt/obd.h>
47 #error Unsupported operating system.
50 #define IOC_OSC_TYPE 'h'
51 #define IOC_OSC_MIN_NR 20
52 #define IOC_OSC_SET_ACTIVE _IOWR(IOC_OSC_TYPE, 21, struct obd_device *)
53 #define IOC_OSC_MAX_NR 50
55 #define IOC_MDC_TYPE 'i'
56 #define IOC_MDC_MIN_NR 20
57 /* Moved to lustre_user.h
58 #define IOC_MDC_LOOKUP _IOWR(IOC_MDC_TYPE, 20, struct obd_ioctl_data *)
59 #define IOC_MDC_GETSTRIPE _IOWR(IOC_MDC_TYPE, 21, struct lov_mds_md *) */
60 #define IOC_MDC_MAX_NR 50
62 #include <lustre/lustre_idl.h>
63 #include <lu_object.h>
65 #include <lustre_lib.h>
66 #include <lustre_export.h>
67 #include <lustre_quota.h>
68 #include <lustre_fld.h>
69 #include <lustre_capa.h>
70 #include <class_hash.h>
72 #include <libcfs/bitmap.h>
75 #define MAX_OBD_DEVICES 8192
77 /* this is really local to the OSC */
78 struct loi_oap_pages {
79 struct list_head lop_pending;
80 struct list_head lop_urgent;
81 struct list_head lop_pending_group;
91 struct lov_oinfo { /* per-stripe data structure */
92 __u64 loi_id; /* object ID on the target OST */
93 __u64 loi_gr; /* object group on the target OST */
94 int loi_ost_idx; /* OST stripe index in lov_tgt_desc->tgts */
95 int loi_ost_gen; /* generation of this loi_ost_idx */
97 /* used by the osc to keep track of what objects to build into rpcs */
98 struct loi_oap_pages loi_read_lop;
99 struct loi_oap_pages loi_write_lop;
100 /* _cli_ is poorly named, it should be _ready_ */
101 struct list_head loi_cli_item;
102 struct list_head loi_write_item;
103 struct list_head loi_read_item;
105 unsigned long loi_kms_valid:1;
106 __u64 loi_kms; /* known minimum size */
107 struct ost_lvb loi_lvb;
108 struct osc_async_rc loi_ar;
111 static inline void loi_kms_set(struct lov_oinfo *oinfo, __u64 kms)
113 oinfo->loi_kms = kms;
114 oinfo->loi_kms_valid = 1;
117 static inline void loi_init(struct lov_oinfo *loi)
119 CFS_INIT_LIST_HEAD(&loi->loi_read_lop.lop_pending);
120 CFS_INIT_LIST_HEAD(&loi->loi_read_lop.lop_urgent);
121 CFS_INIT_LIST_HEAD(&loi->loi_read_lop.lop_pending_group);
122 CFS_INIT_LIST_HEAD(&loi->loi_write_lop.lop_pending);
123 CFS_INIT_LIST_HEAD(&loi->loi_write_lop.lop_urgent);
124 CFS_INIT_LIST_HEAD(&loi->loi_write_lop.lop_pending_group);
125 CFS_INIT_LIST_HEAD(&loi->loi_cli_item);
126 CFS_INIT_LIST_HEAD(&loi->loi_write_item);
127 CFS_INIT_LIST_HEAD(&loi->loi_read_item);
130 /*extent array item for describing the joined file extent info*/
132 __u64 le_start; /* extent start */
133 __u64 le_len; /* extent length */
134 int le_loi_idx; /* extent #1 loi's index in lsm loi array */
135 int le_stripe_count; /* extent stripe count*/
138 /*Lov array info for describing joined file array EA info*/
139 struct lov_array_info {
140 struct llog_logid lai_array_id; /* MDS med llog object id */
141 unsigned lai_ext_count; /* number of extent count */
142 struct lov_extent *lai_ext_array; /* extent desc array */
145 struct lov_stripe_md {
147 pid_t lsm_lock_owner; /* debugging */
150 /* Public members. */
151 __u64 lw_object_id; /* lov object id */
152 __u64 lw_object_gr; /* lov object group */
153 __u64 lw_maxbytes; /* maximum possible file size */
155 /* LOV-private members start here -- only for use in lov/. */
157 __u32 lw_stripe_size; /* size of the stripe */
158 __u32 lw_pattern; /* striping pattern (RAID0, RAID1) */
159 unsigned lw_stripe_count; /* number of objects being striped over */
160 char lw_pool_name[LOV_MAXPOOLNAME]; /* pool name */
163 struct lov_array_info *lsm_array; /*Only for joined file array info*/
164 struct lov_oinfo *lsm_oinfo[0];
167 #define lsm_object_id lsm_wire.lw_object_id
168 #define lsm_object_gr lsm_wire.lw_object_gr
169 #define lsm_maxbytes lsm_wire.lw_maxbytes
170 #define lsm_magic lsm_wire.lw_magic
171 #define lsm_stripe_size lsm_wire.lw_stripe_size
172 #define lsm_pattern lsm_wire.lw_pattern
173 #define lsm_stripe_count lsm_wire.lw_stripe_count
174 #define lsm_pool_name lsm_wire.lw_pool_name
178 typedef int (*obd_enqueue_update_f)(void *cookie, int rc);
180 /* obd info for a particular level (lov, osc). */
182 /* Lock policy. It keeps an extent which is specific for a particular
183 * OSC. (e.g. lov_prep_enqueue_set initialises extent of the policy,
184 * and osc_enqueue passes it into ldlm_lock_match & ldlm_cli_enqueue. */
185 ldlm_policy_data_t oi_policy;
186 /* Flags used for set request specific flags:
187 - while lock handling, the flags obtained on the enqueue
188 request are set here.
189 - while stats, the flags used for control delay/resend.
192 /* Lock handle specific for every OSC lock. */
193 struct lustre_handle *oi_lockh;
194 /* lsm data specific for every OSC. */
195 struct lov_stripe_md *oi_md;
196 /* obdo data specific for every OSC, if needed at all. */
198 /* statfs data specific for every OSC, if needed at all. */
199 struct obd_statfs *oi_osfs;
200 /* An update callback which is called to update some data on upper
201 * level. E.g. it is used for update lsm->lsm_oinfo at every recieved
202 * request in osc level for enqueue requests. It is also possible to
203 * update some caller data from LOV layer if needed. */
204 obd_enqueue_update_f oi_cb_up;
205 /* oss capability, its type is obd_capa in client to avoid copy.
206 * in contrary its type is lustre_capa in OSS. */
210 /* compare all relevant fields. */
211 static inline int lov_stripe_md_cmp(struct lov_stripe_md *m1,
212 struct lov_stripe_md *m2)
215 * ->lsm_wire contains padding, but it should be zeroed out during
218 return memcmp(&m1->lsm_wire, &m2->lsm_wire, sizeof m1->lsm_wire);
221 void lov_stripe_lock(struct lov_stripe_md *md);
222 void lov_stripe_unlock(struct lov_stripe_md *md);
225 struct list_head typ_chain;
226 struct obd_ops *typ_dt_ops;
227 struct md_ops *typ_md_ops;
228 cfs_proc_dir_entry_t *typ_procroot;
231 struct lu_device_type *typ_lu;
232 spinlock_t obd_type_lock;
242 /* Individual type definitions */
244 struct ost_server_data;
246 /* hold common fields for "target" device */
247 struct obd_device_target {
248 struct super_block *obt_sb;
249 atomic_t obt_quotachecking;
250 struct lustre_quota_ctxt obt_qctxt;
251 lustre_quota_version_t obt_qfmt;
252 struct rw_semaphore obt_rwsem;
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 const char *fo_fstype;
291 struct vfsmount *fo_vfsmnt;
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;
301 spinlock_t fo_objidlock; /* protect fo_lastobjid */
302 spinlock_t fo_translock; /* protect fsd_last_transno */
303 struct file *fo_rcvd_filp;
304 struct file *fo_health_check_filp;
305 struct lr_server_data *fo_fsd;
306 unsigned long *fo_last_rcvd_slots;
307 __u64 fo_mount_count;
309 unsigned long fo_destroys_in_progress;
310 struct semaphore fo_create_locks[FILTER_SUBDIR_COUNT];
312 struct list_head fo_export_list;
315 obd_size fo_tot_dirty; /* protected by obd_osfs_lock */
316 obd_size fo_tot_granted; /* all values in bytes */
317 obd_size fo_tot_pending;
318 int fo_tot_granted_clients;
320 obd_size fo_readcache_max_filesize;
322 int fo_writethrough_cache;
324 struct obd_import *fo_mdc_imp;
325 struct obd_uuid fo_mdc_uuid;
326 struct lustre_handle fo_mdc_conn;
327 struct file **fo_last_objid_files;
328 __u64 *fo_last_objids; /* last created objid for groups,
329 * protected by fo_objidlock */
331 struct semaphore fo_alloc_lock;
333 atomic_t fo_r_in_flight;
334 atomic_t fo_w_in_flight;
337 * per-filter pool of kiobuf's allocated by filter_common_setup() and
338 * torn down by filter_cleanup(). Contains OST_NUM_THREADS elements of
339 * which ->fo_iobuf_count were allocated.
341 * This pool contains kiobuf used by
342 * filter_{prep,commit}rw_{read,write}() and is shared by all OST
345 * Locking: none, each OST thread uses only one element, determined by
346 * its "ordinal number", ->t_id.
348 struct filter_iobuf **fo_iobuf_pool;
351 struct list_head fo_llog_list;
352 spinlock_t fo_llog_list_lock;
354 struct brw_stats fo_filter_stats;
355 struct lustre_quota_ctxt fo_quota_ctxt;
356 spinlock_t fo_quotacheck_lock;
357 atomic_t fo_quotachecking;
359 int fo_fmd_max_num; /* per exp filter_mod_data */
360 int fo_fmd_max_age; /* jiffies to fmd expiry */
363 rwlock_t fo_sptlrpc_lock;
364 struct sptlrpc_rule_set fo_sptlrpc_rset;
366 /* capability related */
367 unsigned int fo_fl_oss_capa;
368 struct list_head fo_capa_keys;
369 struct hlist_head *fo_capa_hash;
370 struct llog_commit_master *fo_lcm;
374 struct timeout_item {
375 enum timeout_event ti_event;
376 cfs_time_t ti_timeout;
379 struct list_head ti_obd_list;
380 struct list_head ti_chain;
382 #define OSC_MAX_RIF_DEFAULT 8
383 #define OSC_MAX_RIF_MAX 256
384 #define OSC_MAX_DIRTY_DEFAULT (OSC_MAX_RIF_DEFAULT * 4)
385 #define OSC_MAX_DIRTY_MB_MAX 2048 /* arbitrary, but < MAX_LONG bytes */
386 #define OSC_DEFAULT_RESENDS 10
388 #define MDC_MAX_RIF_DEFAULT 8
389 #define MDC_MAX_RIF_MAX 512
394 struct rw_semaphore cl_sem;
395 struct obd_uuid cl_target_uuid;
396 struct obd_import *cl_import; /* ptlrpc connection state */
398 /* max_mds_easize is purely a performance thing so we don't have to
399 * call obd_size_diskmd() all the time. */
400 int cl_default_mds_easize;
401 int cl_max_mds_easize;
402 int cl_max_mds_cookiesize;
404 enum lustre_sec_part cl_sp_me;
405 enum lustre_sec_part cl_sp_to;
406 struct sptlrpc_flavor cl_flvr_mgc; /* fixed flavor of mgc->mgs */
408 //struct llog_canceld_ctxt *cl_llcd; /* it's included by obd_llog_ctxt */
409 void *cl_llcd_offset;
411 /* the grant values are protected by loi_list_lock below */
412 long cl_dirty; /* all _dirty_ in bytes */
413 long cl_dirty_max; /* allowed w/o rpc */
414 long cl_dirty_transit; /* dirty synchronous */
415 long cl_avail_grant; /* bytes of credit for ost */
416 long cl_lost_grant; /* lost credits (trunc) */
417 struct list_head cl_cache_waiters; /* waiting for cache/grant */
418 cfs_time_t cl_next_shrink_grant; /* jiffies */
419 struct list_head cl_grant_shrink_list; /* Timeout event list */
420 struct semaphore cl_grant_sem; /*grant shrink list semaphore*/
422 /* keep track of objects that have lois that contain pages which
423 * have been queued for async brw. this lock also protects the
424 * lists of osc_client_pages that hang off of the loi */
426 * ->cl_loi_list_lock protects consistency of
427 * ->cl_loi_{ready,read,write}_list. ->ap_make_ready() and
428 * ->ap_completion() call-backs are executed under this lock. As we
429 * cannot guarantee that these call-backs never block on all platforms
430 * (as a matter of fact they do block on Mac OS X), type of
431 * ->cl_loi_list_lock is platform dependent: it's a spin-lock on Linux
432 * and blocking mutex on Mac OS X. (Alternative is to make this lock
433 * blocking everywhere, but we don't want to slow down fast-path of
434 * our main platform.)
436 * Exact type of ->cl_loi_list_lock is defined in arch/obd.h together
437 * with client_obd_list_{un,}lock() and
438 * client_obd_list_lock_{init,done}() functions.
440 client_obd_lock_t cl_loi_list_lock;
441 struct list_head cl_loi_ready_list;
442 struct list_head cl_loi_write_list;
443 struct list_head cl_loi_read_list;
446 /* just a sum of the loi/lop pending numbers to be exported by /proc */
447 int cl_pending_w_pages;
448 int cl_pending_r_pages;
449 int cl_max_pages_per_rpc;
450 int cl_max_rpcs_in_flight;
451 struct obd_histogram cl_read_rpc_hist;
452 struct obd_histogram cl_write_rpc_hist;
453 struct obd_histogram cl_read_page_hist;
454 struct obd_histogram cl_write_page_hist;
455 struct obd_histogram cl_read_offset_hist;
456 struct obd_histogram cl_write_offset_hist;
458 /* number of in flight destroy rpcs is limited to max_rpcs_in_flight */
459 atomic_t cl_destroy_in_flight;
460 cfs_waitq_t cl_destroy_waitq;
462 struct mdc_rpc_lock *cl_rpc_lock;
463 struct mdc_rpc_lock *cl_setattr_lock;
464 struct mdc_rpc_lock *cl_close_lock;
465 struct osc_creator cl_oscc;
468 struct semaphore cl_mgc_sem;
469 struct vfsmount *cl_mgc_vfsmnt;
470 struct dentry *cl_mgc_configs_dir;
471 atomic_t cl_mgc_refcount;
472 struct obd_export *cl_mgc_mgsexp;
474 /* checksumming for data sent over the network */
475 unsigned int cl_checksum:1; /* 0 = disabled, 1 = enabled */
476 /* supported checksum types that are worked out at connect time */
477 __u32 cl_supp_cksum_types;
478 /* checksum algorithm to be used */
479 cksum_type_t cl_cksum_type;
481 /* also protected by the poorly named _loi_list_lock lock above */
482 struct osc_async_rc cl_ar;
484 /* used by quotacheck */
485 int cl_qchk_stat; /* quotacheck stat of the peer */
487 /* sequence manager */
488 struct lu_client_seq *cl_seq;
490 atomic_t cl_resends; /* resend count */
492 #define obd2cli_tgt(obd) ((char *)(obd)->u.cli.cl_target_uuid.uuid)
494 #define CL_NOT_QUOTACHECKED 1 /* client->cl_qchk_stat init value */
497 struct ptlrpc_service *mgs_service;
498 struct vfsmount *mgs_vfsmnt;
499 struct super_block *mgs_sb;
500 struct dentry *mgs_configs_dir;
501 struct dentry *mgs_fid_de;
502 struct list_head mgs_fs_db_list;
503 struct semaphore mgs_sem;
504 cfs_proc_dir_entry_t *mgs_proc_live;
508 /* NB this field MUST be first */
509 struct obd_device_target mds_obt;
510 struct ptlrpc_service *mds_service;
511 struct ptlrpc_service *mds_setattr_service;
512 struct ptlrpc_service *mds_readpage_service;
513 struct vfsmount *mds_vfsmnt;
514 cfs_dentry_t *mds_fid_de;
516 int mds_max_cookiesize;
517 struct file *mds_rcvd_filp;
518 spinlock_t mds_transno_lock;
519 __u64 mds_last_transno;
520 __u64 mds_mount_count;
522 unsigned long mds_atime_diff;
523 struct semaphore mds_epoch_sem;
524 struct ll_fid mds_rootfid;
525 struct lr_server_data *mds_server_data;
526 cfs_dentry_t *mds_pending_dir;
527 cfs_dentry_t *mds_logs_dir;
528 cfs_dentry_t *mds_objects_dir;
529 struct llog_handle *mds_cfg_llh;
530 struct obd_device *mds_osc_obd; /* XXX lov_obd */
531 struct obd_uuid mds_lov_uuid;
533 struct obd_export *mds_osc_exp; /* XXX lov_exp */
534 struct lov_desc mds_lov_desc;
537 /* mark pages dirty for write. */
538 bitmap_t *mds_lov_page_dirty;
539 /* array for store pages with obd_id */
540 void **mds_lov_page_array;
541 /* file for store objid */
542 struct file *mds_lov_objid_filp;
543 __u32 mds_lov_objid_count;
544 __u32 mds_lov_objid_max_index;
545 __u32 mds_lov_objid_lastpage;
546 __u32 mds_lov_objid_lastidx;
548 struct file *mds_health_check_filp;
549 unsigned long *mds_client_bitmap;
550 // struct upcall_cache *mds_group_hash;
552 struct lustre_quota_info mds_quota_info;
553 struct semaphore mds_qonoff_sem;
554 struct semaphore mds_health_sem;
555 unsigned long mds_fl_user_xattr:1,
557 mds_evict_ost_nids:1,
561 mds_fl_target:1; /* mds have one or
564 struct upcall_cache *mds_identity_cache;
566 /* for capability keys update */
567 struct lustre_capa_key *mds_capa_keys;
568 struct rw_semaphore mds_notify_lock;
572 extern __u32 mds_max_ost_index;
574 #define MDS_LOV_ALLOC_SIZE (CFS_PAGE_SIZE)
576 #define OBJID_PER_PAGE() (MDS_LOV_ALLOC_SIZE / sizeof(obd_id))
578 #define MDS_LOV_OBJID_PAGES_COUNT (mds_max_ost_index/OBJID_PER_PAGE())
580 extern int mds_lov_init_objids(struct obd_device *obd);
581 extern void mds_lov_destroy_objids(struct obd_device *obd);
594 struct lustre_handle eo_nl_lock;
599 struct ptlrpc_service *ost_service;
600 struct ptlrpc_service *ost_create_service;
601 struct ptlrpc_service *ost_io_service;
602 struct semaphore ost_health_sem;
605 struct echo_client_obd {
606 struct obd_export *ec_exp; /* the local connection to osc/lov */
608 struct list_head ec_objects;
609 struct list_head ec_locks;
615 struct obd_uuid lqo_uuid; /* ptlrpc's c_remote_uuid */
616 struct list_head lqo_oss_list; /* link to lov_qos */
617 __u64 lqo_bavail; /* total bytes avail on OSS */
618 __u64 lqo_penalty; /* current penalty */
619 __u64 lqo_penalty_per_obj; /* penalty decrease every obj*/
620 time_t lqo_used; /* last used time, seconds */
621 __u32 lqo_ost_count; /* number of osts on this oss */
625 struct lov_qos_oss *ltq_oss; /* oss info */
626 __u64 ltq_penalty; /* current penalty */
627 __u64 ltq_penalty_per_obj; /* penalty decrease every obj*/
628 __u64 ltq_weight; /* net weighting */
629 time_t ltq_used; /* last used time, seconds */
630 unsigned int ltq_usable:1; /* usable for striping */
633 /* Generic subset of OSTs */
635 __u32 *op_array; /* array of index of
637 unsigned int op_count; /* number of OSTs in the array */
638 unsigned int op_size; /* allocated size of lp_array */
639 struct rw_semaphore op_rw_sem; /* to protect ost_pool use */
642 /* Round-robin allocator data */
644 __u32 lqr_start_idx; /* start index of new inode */
645 __u32 lqr_offset_idx; /* aliasing for start_idx */
646 int lqr_start_count; /* reseed counter */
647 struct ost_pool lqr_pool; /* round-robin optimized list */
648 unsigned long lqr_dirty:1; /* recalc round-robin list */
651 struct lov_statfs_data {
652 struct obd_info lsd_oi;
653 struct obd_statfs lsd_statfs;
655 /* Stripe placement optimization */
657 struct list_head lq_oss_list; /* list of OSSs that targets use */
658 struct rw_semaphore lq_rw_sem;
659 __u32 lq_active_oss_count;
660 unsigned int lq_prio_free; /* priority for free space */
661 unsigned int lq_threshold_rr;/* priority for rr */
662 struct lov_qos_rr lq_rr; /* round robin qos data */
663 unsigned long lq_dirty:1, /* recalc qos data */
664 lq_same_space:1,/* the ost's all have approx.
665 the same space avail */
666 lq_reset:1, /* zero current penalties */
667 lq_statfs_in_progress:1; /* statfs op in progress */
668 /* qos statfs data */
669 struct lov_statfs_data *lq_statfs_data;
670 cfs_waitq_t lq_statfs_waitq; /* waitqueue to notify statfs
671 * requests completion */
674 struct lov_tgt_desc {
675 struct list_head ltd_kill;
676 struct obd_uuid ltd_uuid;
677 struct obd_export *ltd_exp;
678 struct ltd_qos ltd_qos; /* qos info per target */
680 __u32 ltd_index; /* index in lov_obd->tgts */
681 unsigned long ltd_active:1,/* is this target up for requests */
682 ltd_activate:1,/* should this target be activated */
683 ltd_reap:1; /* should this target be deleted */
687 #define pool_tgt_size(_p) _p->pool_obds.op_size
688 #define pool_tgt_count(_p) _p->pool_obds.op_count
689 #define pool_tgt_array(_p) _p->pool_obds.op_array
690 #define pool_tgt_rw_sem(_p) _p->pool_obds.op_rw_sem
691 #define pool_tgt(_p, _i) _p->pool_lov->lov_tgts[_p->pool_obds.op_array[_i]]
694 char pool_name[LOV_MAXPOOLNAME + 1]; /* name of pool */
695 struct ost_pool pool_obds; /* pool members */
696 atomic_t pool_refcount; /* pool ref. counter */
697 struct lov_qos_rr pool_rr; /* round robin qos */
698 struct hlist_node pool_hash; /* access by poolname */
699 struct list_head pool_list; /* serial access */
700 cfs_proc_dir_entry_t *pool_proc_entry; /* file in /proc */
701 struct lov_obd *pool_lov; /* lov obd to which this
706 struct lov_desc desc;
707 struct lov_tgt_desc **lov_tgts; /* sparse array */
708 struct ost_pool lov_packed; /* all OSTs in a packed
710 struct semaphore lov_lock;
711 struct obd_connect_data lov_ocd;
712 struct lov_qos lov_qos; /* qos info per lov */
713 atomic_t lov_refcount;
714 __u32 lov_tgt_count; /* how many OBD's */
715 __u32 lov_active_tgt_count; /* how many active */
716 __u32 lov_death_row;/* tgts scheduled to be deleted */
717 __u32 lov_tgt_size; /* size of tgts array */
720 lustre_hash_t *lov_pools_hash_body; /* used for key access */
721 struct list_head lov_pool_list; /* used for sequential access */
722 cfs_proc_dir_entry_t *lov_pool_proc_entry;
723 enum lustre_sec_part lov_sp_me;
726 struct lmv_tgt_desc {
727 struct obd_uuid ltd_uuid;
728 struct obd_export *ltd_exp;
729 int ltd_active; /* is this target up for requests */
731 struct semaphore ltd_fid_sem;
734 enum placement_policy {
735 PLACEMENT_CHAR_POLICY = 0,
736 PLACEMENT_NID_POLICY = 1,
737 PLACEMENT_INVAL_POLICY = 2,
741 typedef enum placement_policy placement_policy_t;
745 struct lu_client_fld lmv_fld;
747 placement_policy_t lmv_placement;
748 struct lmv_desc desc;
749 struct obd_uuid cluuid;
750 struct obd_export *exp;
757 struct semaphore init_sem;
759 struct lmv_tgt_desc *tgts;
762 struct obd_connect_data *datas;
765 struct obd_connect_data conn_data;
768 struct niobuf_local {
773 cfs_dentry_t *dentry;
778 #define LUSTRE_FLD_NAME "fld"
779 #define LUSTRE_SEQ_NAME "seq"
781 #define LUSTRE_CMM_NAME "cmm"
782 #define LUSTRE_MDD_NAME "mdd"
783 #define LUSTRE_OSD_NAME "osd"
784 #define LUSTRE_VVP_NAME "vvp"
785 #define LUSTRE_LMV_NAME "lmv"
786 #define LUSTRE_CMM_MDC_NAME "cmm-mdc"
787 #define LUSTRE_SLP_NAME "slp"
789 /* obd device type names */
790 /* FIXME all the references to LUSTRE_MDS_NAME should be swapped with LUSTRE_MDT_NAME */
791 #define LUSTRE_MDS_NAME "mds"
792 #define LUSTRE_MDT_NAME "mdt"
793 #define LUSTRE_MDC_NAME "mdc"
794 #define LUSTRE_OSS_NAME "ost" /* FIXME change name to oss */
795 #define LUSTRE_OST_NAME "obdfilter" /* FIXME change name to ost */
796 #define LUSTRE_OSC_NAME "osc"
797 #define LUSTRE_LOV_NAME "lov"
798 #define LUSTRE_MGS_NAME "mgs"
799 #define LUSTRE_MGC_NAME "mgc"
801 #define LUSTRE_CACHEOBD_NAME "cobd"
802 #define LUSTRE_ECHO_NAME "obdecho"
803 #define LUSTRE_ECHO_CLIENT_NAME "echo_client"
805 /* Constant obd names (post-rename) */
806 #define LUSTRE_MDS_OBDNAME "MDS"
807 #define LUSTRE_OSS_OBDNAME "OSS"
808 #define LUSTRE_MGS_OBDNAME "MGS"
809 #define LUSTRE_MGC_OBDNAME "MGC"
811 /* Don't conflict with on-wire flags OBD_BRW_WRITE, etc */
812 #define N_LOCAL_TEMP_PAGE 0x10000000
814 struct obd_trans_info {
817 /* Only used on the server side for tracking acks. */
818 struct oti_req_ack_lock {
819 struct lustre_handle lock;
823 struct llog_cookie oti_onecookie;
824 struct llog_cookie *oti_logcookies;
827 /* initial thread handling transaction */
828 struct ptlrpc_thread * oti_thread;
831 struct obd_uuid *oti_ost_uuid;
834 static inline void oti_init(struct obd_trans_info *oti,
835 struct ptlrpc_request *req)
839 memset(oti, 0, sizeof(*oti));
844 oti->oti_xid = req->rq_xid;
846 if (req->rq_repmsg != NULL)
847 oti->oti_transno = lustre_msg_get_transno(req->rq_repmsg);
848 oti->oti_thread = req->rq_svc_thread;
849 if (req->rq_reqmsg != NULL)
850 oti->oti_conn_cnt = lustre_msg_get_conn_cnt(req->rq_reqmsg);
853 static inline void oti_alloc_cookies(struct obd_trans_info *oti,int num_cookies)
858 if (num_cookies == 1)
859 oti->oti_logcookies = &oti->oti_onecookie;
861 OBD_ALLOC(oti->oti_logcookies,
862 num_cookies * sizeof(oti->oti_onecookie));
864 oti->oti_numcookies = num_cookies;
867 static inline void oti_free_cookies(struct obd_trans_info *oti)
869 if (!oti || !oti->oti_logcookies)
872 if (oti->oti_logcookies == &oti->oti_onecookie)
873 LASSERT(oti->oti_numcookies == 1);
875 OBD_FREE(oti->oti_logcookies,
876 oti->oti_numcookies * sizeof(oti->oti_onecookie));
877 oti->oti_logcookies = NULL;
878 oti->oti_numcookies = 0;
882 * Events signalled through obd_notify() upcall-chain.
884 enum obd_notify_event {
885 /* Device connect start */
887 /* Device activated */
889 /* Device deactivated */
891 /* Device disconnected */
893 /* Connect data for import were changed */
896 OBD_NOTIFY_SYNC_NONBLOCK,
898 /* Configuration event */
902 /* bit-mask flags for config events */
904 CONFIG_LOG = 0x1, /* finished processing config log */
905 CONFIG_SYNC = 0x2, /* mdt synced 1 ost */
906 CONFIG_TARGET = 0x4 /* one target is added */
910 * Data structure used to pass obd_notify()-event to non-obd listeners (llite
911 * and liblustre being main examples).
913 struct obd_notify_upcall {
914 int (*onu_upcall)(struct obd_device *host, struct obd_device *watched,
915 enum obd_notify_event ev, void *owner);
916 /* Opaque datum supplied by upper layer listener */
920 struct target_recovery_data {
921 svc_handler_t trd_recovery_handler;
922 pid_t trd_processing_task;
923 struct completion trd_starting;
924 struct completion trd_finishing;
928 FILTER_GROUP_MDS0 = 0,
929 FILTER_GROUP_LLOG = 1,
930 FILTER_GROUP_ECHO = 2 ,
931 FILTER_GROUP_MDS1_N_BASE = 3
934 static inline __u64 obdo_mdsno(struct obdo *oa)
937 return oa->o_gr - FILTER_GROUP_MDS1_N_BASE;
941 static inline int mdt_to_obd_objgrp(int mdtid)
944 return FILTER_GROUP_MDS1_N_BASE + mdtid;
949 * In HEAD for CMD, the object is created in group number which is 3>=
950 * or indexing starts from 3. To test this assertions are added to disallow
951 * group 0. But to run 2.0 mds server on 1.8.x disk format (i.e. interop_mode)
952 * object in group 0 needs to be allowed.
953 * So for interop mode following changes needs to be done:
954 * 1. No need to assert on group 0 or allow group 0
955 * 2. The group number indexing starts from 0 instead of 3
958 #define CHECK_MDS_GROUP(group) (group == FILTER_GROUP_MDS0 || \
959 group > FILTER_GROUP_MDS1_N_BASE)
960 #define LASSERT_MDS_GROUP(group) LASSERT(CHECK_MDS_GROUP(group))
962 struct obd_llog_group {
963 struct list_head olg_list;
965 struct llog_ctxt *olg_ctxts[LLOG_MAX_CTXTS];
966 cfs_waitq_t olg_waitq;
968 struct obd_export *olg_exp;
969 int olg_initializing;
970 struct semaphore olg_cat_processing;
973 /* corresponds to one of the obd's */
974 #define MAX_OBD_NAME 128
975 #define OBD_DEVICE_MAGIC 0XAB5CD6EF
976 #define OBD_DEV_BY_DEVNAME 0xffffd0de
978 struct obd_type *obd_type;
981 /* common and UUID name of this device */
982 char obd_name[MAX_OBD_NAME];
983 struct obd_uuid obd_uuid;
985 struct lu_device *obd_lu_dev;
988 unsigned long obd_attached:1, /* finished attach */
989 obd_set_up:1, /* finished setup */
990 obd_recovering:1, /* there are recoverable clients */
991 obd_abort_recovery:1,/* somebody ioctl'ed us to abort */
992 obd_replayable:1, /* recovery is enabled; inform clients */
993 obd_no_transno:1, /* no committed-transno notification */
994 obd_no_recov:1, /* fail instead of retry messages */
995 obd_req_replaying:1, /* replaying requests */
996 obd_stopping:1, /* started cleanup */
997 obd_starting:1, /* started setup */
998 obd_force:1, /* cleanup with > 0 obd refcount */
999 obd_fail:1, /* cleanup with failover */
1000 obd_async_recov:1, /* allow asyncronous orphan cleanup */
1001 obd_no_conn:1, /* deny new connections */
1002 obd_inactive:1; /* device active/inactive
1003 * (for /proc/status only!!) */
1004 /* uuid-export hash body */
1005 struct lustre_hash *obd_uuid_hash;
1006 /* nid-export hash body */
1007 struct lustre_hash *obd_nid_hash;
1008 /* nid stats body */
1009 struct lustre_hash *obd_nid_stats_hash;
1010 struct list_head obd_nid_stats;
1011 atomic_t obd_refcount;
1012 cfs_waitq_t obd_refcount_waitq;
1013 struct list_head obd_exports;
1014 int obd_num_exports;
1015 spinlock_t obd_nid_lock;
1016 struct ldlm_namespace *obd_namespace;
1017 struct ptlrpc_client obd_ldlm_client; /* XXX OST/MDS only */
1018 /* a spinlock is OK for what we do now, may need a semaphore later */
1019 spinlock_t obd_dev_lock;
1020 struct semaphore obd_dev_sem;
1021 __u64 obd_last_committed;
1022 struct fsfilt_operations *obd_fsops;
1023 spinlock_t obd_osfs_lock;
1024 struct obd_statfs obd_osfs; /* locked by obd_osfs_lock */
1026 struct lvfs_run_ctxt obd_lvfs_ctxt;
1027 struct obd_llog_group obd_olg; /* default llog group */
1028 struct obd_device *obd_observer;
1029 struct obd_notify_upcall obd_upcall;
1030 struct obd_export *obd_self_export;
1031 /* list of exports in LRU order, for ping evictor, with obd_dev_lock */
1032 struct list_head obd_exports_timed;
1033 time_t obd_eviction_timer; /* for ping evictor */
1035 /* XXX encapsulate all this recovery data into one struct */
1036 svc_handler_t obd_recovery_handler;
1037 pid_t obd_processing_task;
1039 int obd_max_recoverable_clients;
1040 int obd_connected_clients;
1041 int obd_recoverable_clients;
1042 spinlock_t obd_processing_task_lock; /* BH lock (timer) */
1043 __u64 obd_next_recovery_transno;
1044 int obd_replayed_requests;
1045 int obd_requests_queued_for_recovery;
1046 cfs_waitq_t obd_next_transno_waitq;
1047 struct list_head obd_uncommitted_replies;
1048 spinlock_t obd_uncommitted_replies_lock;
1049 cfs_timer_t obd_recovery_timer;
1050 time_t obd_recovery_start; /* seconds */
1051 time_t obd_recovery_end; /* seconds, for lprocfs_status */
1052 time_t obd_recovery_max_time; /* seconds, bz13079 */
1053 int obd_recovery_timeout;
1055 /* new recovery stuff from CMD2 */
1056 struct target_recovery_data obd_recovery_data;
1057 int obd_replayed_locks;
1058 atomic_t obd_req_replay_clients;
1059 atomic_t obd_lock_replay_clients;
1060 struct list_head obd_req_replay_queue;
1061 struct list_head obd_lock_replay_queue;
1062 struct list_head obd_final_req_queue;
1063 int obd_recovery_stage;
1066 struct obd_device_target obt;
1067 struct filter_obd filter;
1069 struct client_obd cli;
1071 struct echo_client_obd echo_client;
1072 struct echo_obd echo;
1077 /* Fields used by LProcFS */
1078 unsigned int obd_cntr_base;
1079 struct lprocfs_stats *obd_stats;
1081 unsigned int md_cntr_base;
1082 struct lprocfs_stats *md_stats;
1084 cfs_proc_dir_entry_t *obd_proc_entry;
1085 cfs_proc_dir_entry_t *obd_proc_exports_entry;
1086 cfs_proc_dir_entry_t *obd_svc_procroot;
1087 struct lprocfs_stats *obd_svc_stats;
1088 atomic_t obd_evict_inprogress;
1089 cfs_waitq_t obd_evict_inprogress_waitq;
1092 * Ldlm pool part. Save last calculated SLV and Limit.
1094 rwlock_t obd_pool_lock;
1099 * A list of outstanding class_incref()'s against this obd. For
1102 struct lu_ref obd_reference;
1105 #define OBD_LLOG_FL_SENDNOW 0x0001
1107 enum obd_cleanup_stage {
1108 /* Special case hack for MDS LOVs */
1110 /* can be directly mapped to .ldto_device_fini() */
1111 OBD_CLEANUP_EXPORTS,
1114 /* get/set_info keys */
1115 #define KEY_BLOCKSIZE_BITS "blocksize_bits"
1116 #define KEY_BLOCKSIZE "blocksize"
1117 #define KEY_CAPA_KEY "capa_key"
1118 #define KEY_CHANGELOG_CLEAR "changelog_clear"
1119 #define KEY_CHECKSUM "checksum"
1120 #define KEY_CLEAR_FS "clear_fs"
1121 #define KEY_CONN_DATA "conn_data"
1122 #define KEY_EVICT_BY_NID "evict_by_nid"
1123 #define KEY_FIEMAP "fiemap"
1124 #define KEY_FLUSH_CTX "flush_ctx"
1125 #define KEY_INIT_RECOV_BACKUP "init_recov_bk"
1126 #define KEY_INIT_RECOV "initial_recov"
1127 #define KEY_LAST_ID "last_id"
1128 #define KEY_LOCK_TO_STRIPE "lock_to_stripe"
1129 #define KEY_LOVDESC "lovdesc"
1130 #define KEY_LOV_IDX "lov_idx"
1131 #define KEY_MAX_EASIZE "max_easize"
1132 #define KEY_MDS_CONN "mds_conn"
1133 #define KEY_MGSSEC "mgssec"
1134 #define KEY_NEXT_ID "next_id"
1135 #define KEY_READ_ONLY "read-only"
1136 #define KEY_REGISTER_TARGET "register_target"
1137 #define KEY_REVIMP_UPD "revimp_update"
1138 #define KEY_SET_FS "set_fs"
1139 #define KEY_SPTLRPC_CONF "sptlrpc_conf"
1140 #define KEY_UNLINKED "unlinked"
1142 #define KEY_INTERMDS "inter_mds"
1143 #define KEY_ASYNC "async"
1144 #define KEY_GRANT_SHRINK "grant_shrink"
1148 static inline int it_to_lock_mode(struct lookup_intent *it)
1150 /* CREAT needs to be tested before open (both could be set) */
1151 if (it->it_op & IT_CREAT)
1153 else if (it->it_op & (IT_READDIR | IT_GETATTR | IT_OPEN | IT_LOOKUP))
1156 LASSERTF(0, "Invalid it_op: %d\n", it->it_op);
1161 struct lu_fid op_fid1; /* operation fid1 (usualy parent) */
1162 struct lu_fid op_fid2; /* operation fid2 (usualy child) */
1163 struct lu_fid op_fid3; /* 2 extra fids to find conflicting */
1164 struct lu_fid op_fid4; /* to the operation locks. */
1165 mdsno_t op_mds; /* what mds server open will go to */
1166 struct lustre_handle op_handle;
1168 const char *op_name;
1171 struct lmv_stripe_md *op_mea1;
1172 struct lmv_stripe_md *op_mea2;
1173 __u32 op_suppgids[2];
1179 /* iattr fields and blocks. */
1180 struct iattr op_attr;
1182 #if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,14)
1183 unsigned int op_attr_flags;
1186 loff_t op_attr_blocks;
1188 /* Size-on-MDS epoch and flags. */
1193 struct obd_capa *op_capa1;
1194 struct obd_capa *op_capa2;
1196 /* Various operation flags. */
1199 /* Operation type */
1203 struct md_enqueue_info;
1204 /* metadata stat-ahead */
1205 typedef int (* md_enqueue_cb_t)(struct ptlrpc_request *req,
1206 struct md_enqueue_info *minfo,
1209 struct md_enqueue_info {
1210 struct md_op_data mi_data;
1211 struct lookup_intent mi_it;
1212 struct lustre_handle mi_lockh;
1213 struct dentry *mi_dentry;
1214 md_enqueue_cb_t mi_cb;
1215 unsigned int mi_generation;
1220 struct module *o_owner;
1221 int (*o_iocontrol)(unsigned int cmd, struct obd_export *exp, int len,
1222 void *karg, void *uarg);
1223 int (*o_get_info)(struct obd_export *, __u32 keylen, void *key,
1224 __u32 *vallen, void *val, struct lov_stripe_md *lsm);
1225 int (*o_set_info_async)(struct obd_export *, __u32 keylen, void *key,
1226 __u32 vallen, void *val,
1227 struct ptlrpc_request_set *set);
1228 int (*o_attach)(struct obd_device *dev, obd_count len, void *data);
1229 int (*o_detach)(struct obd_device *dev);
1230 int (*o_setup) (struct obd_device *dev, struct lustre_cfg *cfg);
1231 int (*o_precleanup)(struct obd_device *dev,
1232 enum obd_cleanup_stage cleanup_stage);
1233 int (*o_cleanup)(struct obd_device *dev);
1234 int (*o_process_config)(struct obd_device *dev, obd_count len,
1236 int (*o_postrecov)(struct obd_device *dev);
1237 int (*o_add_conn)(struct obd_import *imp, struct obd_uuid *uuid,
1239 int (*o_del_conn)(struct obd_import *imp, struct obd_uuid *uuid);
1240 /* connect to the target device with given connection
1241 * data. @ocd->ocd_connect_flags is modified to reflect flags actually
1242 * granted by the target, which are guaranteed to be a subset of flags
1243 * asked for. If @ocd == NULL, use default parameters. */
1244 int (*o_connect)(const struct lu_env *env,
1245 struct obd_export **exp, struct obd_device *src,
1246 struct obd_uuid *cluuid, struct obd_connect_data *ocd,
1248 int (*o_reconnect)(const struct lu_env *env,
1249 struct obd_export *exp, struct obd_device *src,
1250 struct obd_uuid *cluuid,
1251 struct obd_connect_data *ocd,
1253 int (*o_disconnect)(struct obd_export *exp);
1255 /* Initialize/finalize fids infrastructure. */
1256 int (*o_fid_init)(struct obd_export *exp);
1257 int (*o_fid_fini)(struct obd_export *exp);
1259 /* Allocate new fid according to passed @hint. */
1260 int (*o_fid_alloc)(struct obd_export *exp, struct lu_fid *fid,
1261 struct md_op_data *op_data);
1264 * Object with @fid is getting deleted, we may want to do something
1267 int (*o_fid_delete)(struct obd_export *exp, const struct lu_fid *fid);
1269 int (*o_statfs)(struct obd_device *obd, struct obd_statfs *osfs,
1270 __u64 max_age, __u32 flags);
1271 int (*o_statfs_async)(struct obd_device *obd, struct obd_info *oinfo,
1272 __u64 max_age, struct ptlrpc_request_set *set);
1273 int (*o_packmd)(struct obd_export *exp, struct lov_mds_md **disk_tgt,
1274 struct lov_stripe_md *mem_src);
1275 int (*o_unpackmd)(struct obd_export *exp,struct lov_stripe_md **mem_tgt,
1276 struct lov_mds_md *disk_src, int disk_len);
1277 int (*o_checkmd)(struct obd_export *exp, struct obd_export *md_exp,
1278 struct lov_stripe_md *mem_tgt);
1279 int (*o_preallocate)(struct lustre_handle *, obd_count *req,
1281 /* FIXME: add fid capability support for create & destroy! */
1282 int (*o_precreate)(struct obd_export *exp);
1283 int (*o_create)(struct obd_export *exp, struct obdo *oa,
1284 struct lov_stripe_md **ea, struct obd_trans_info *oti);
1285 int (*o_destroy)(struct obd_export *exp, struct obdo *oa,
1286 struct lov_stripe_md *ea, struct obd_trans_info *oti,
1287 struct obd_export *md_exp, void *capa);
1288 int (*o_setattr)(struct obd_export *exp, struct obd_info *oinfo,
1289 struct obd_trans_info *oti);
1290 int (*o_setattr_async)(struct obd_export *exp, struct obd_info *oinfo,
1291 struct obd_trans_info *oti,
1292 struct ptlrpc_request_set *rqset);
1293 int (*o_getattr)(struct obd_export *exp, struct obd_info *oinfo);
1294 int (*o_getattr_async)(struct obd_export *exp, struct obd_info *oinfo,
1295 struct ptlrpc_request_set *set);
1296 int (*o_brw)(int rw, struct obd_export *exp, struct obd_info *oinfo,
1297 obd_count oa_bufs, struct brw_page *pgarr,
1298 struct obd_trans_info *oti);
1299 int (*o_merge_lvb)(struct obd_export *exp, struct lov_stripe_md *lsm,
1300 struct ost_lvb *lvb, int kms_only);
1301 int (*o_adjust_kms)(struct obd_export *exp, struct lov_stripe_md *lsm,
1302 obd_off size, int shrink);
1303 int (*o_punch)(struct obd_export *exp, struct obd_info *oinfo,
1304 struct obd_trans_info *oti,
1305 struct ptlrpc_request_set *rqset);
1306 int (*o_sync)(struct obd_export *exp, struct obdo *oa,
1307 struct lov_stripe_md *ea, obd_size start, obd_size end,
1309 int (*o_migrate)(struct lustre_handle *conn, struct lov_stripe_md *dst,
1310 struct lov_stripe_md *src, obd_size start,
1311 obd_size end, struct obd_trans_info *oti);
1312 int (*o_copy)(struct lustre_handle *dstconn, struct lov_stripe_md *dst,
1313 struct lustre_handle *srconn, struct lov_stripe_md *src,
1314 obd_size start, obd_size end, struct obd_trans_info *);
1315 int (*o_iterate)(struct lustre_handle *conn,
1316 int (*)(obd_id, obd_gr, void *),
1317 obd_id *startid, obd_gr group, void *data);
1318 int (*o_preprw)(int cmd, struct obd_export *exp, struct obdo *oa,
1319 int objcount, struct obd_ioobj *obj,
1320 struct niobuf_remote *remote, int *nr_pages,
1321 struct niobuf_local *local,
1322 struct obd_trans_info *oti,
1323 struct lustre_capa *capa);
1324 int (*o_commitrw)(int cmd, struct obd_export *exp, struct obdo *oa,
1325 int objcount, struct obd_ioobj *obj,
1326 struct niobuf_remote *remote, int pages,
1327 struct niobuf_local *local,
1328 struct obd_trans_info *oti, int rc);
1329 int (*o_enqueue)(struct obd_export *, struct obd_info *oinfo,
1330 struct ldlm_enqueue_info *einfo,
1331 struct ptlrpc_request_set *rqset);
1332 int (*o_change_cbdata)(struct obd_export *, struct lov_stripe_md *,
1333 ldlm_iterator_t it, void *data);
1334 int (*o_cancel)(struct obd_export *, struct lov_stripe_md *md,
1335 __u32 mode, struct lustre_handle *);
1336 int (*o_cancel_unused)(struct obd_export *, struct lov_stripe_md *,
1337 int flags, void *opaque);
1338 int (*o_init_export)(struct obd_export *exp);
1339 int (*o_destroy_export)(struct obd_export *exp);
1340 int (*o_extent_calc)(struct obd_export *, struct lov_stripe_md *,
1341 int cmd, obd_off *);
1343 /* llog related obd_methods */
1344 int (*o_llog_init)(struct obd_device *obd, struct obd_llog_group *grp,
1345 struct obd_device *disk_obd, int count,
1346 struct llog_catid *logid, struct obd_uuid *uuid);
1347 int (*o_llog_finish)(struct obd_device *obd, int count);
1348 int (*o_llog_connect)(struct obd_export *, struct llogd_conn_body *);
1350 /* metadata-only methods */
1351 int (*o_pin)(struct obd_export *, const struct lu_fid *fid,
1352 struct obd_capa *, struct obd_client_handle *, int flag);
1353 int (*o_unpin)(struct obd_export *, struct obd_client_handle *, int);
1355 int (*o_import_event)(struct obd_device *, struct obd_import *,
1356 enum obd_import_event);
1358 int (*o_notify)(struct obd_device *obd, struct obd_device *watched,
1359 enum obd_notify_event ev, void *data);
1361 int (*o_health_check)(struct obd_device *);
1362 struct obd_uuid *(*o_get_uuid) (struct obd_export *exp);
1365 int (*o_quotacheck)(struct obd_device *, struct obd_export *,
1366 struct obd_quotactl *);
1367 int (*o_quotactl)(struct obd_device *, struct obd_export *,
1368 struct obd_quotactl *);
1369 int (*o_quota_adjust_qunit)(struct obd_export *exp,
1370 struct quota_adjust_qunit *oqaq,
1371 struct lustre_quota_ctxt *qctxt);
1374 int (*o_ping)(struct obd_export *exp);
1377 int (*o_pool_new)(struct obd_device *obd, char *poolname);
1378 int (*o_pool_del)(struct obd_device *obd, char *poolname);
1379 int (*o_pool_add)(struct obd_device *obd, char *poolname,
1381 int (*o_pool_rem)(struct obd_device *obd, char *poolname,
1384 * NOTE: If adding ops, add another LPROCFS_OBD_OP_INIT() line
1385 * to lprocfs_alloc_obd_stats() in obdclass/lprocfs_status.c.
1386 * Also, add a wrapper function in include/linux/obd_class.h. */
1390 LUSTRE_OPC_MKDIR = (1 << 0),
1391 LUSTRE_OPC_SYMLINK = (1 << 1),
1392 LUSTRE_OPC_MKNOD = (1 << 2),
1393 LUSTRE_OPC_CREATE = (1 << 3),
1394 LUSTRE_OPC_ANY = (1 << 4)
1397 /* lmv structures */
1398 #define MEA_MAGIC_LAST_CHAR 0xb2221ca1
1399 #define MEA_MAGIC_ALL_CHARS 0xb222a11c
1400 #define MEA_MAGIC_HASH_SEGMENT 0xb222a11b
1402 #define MAX_HASH_SIZE_32 0x7fffffffUL
1403 #define MAX_HASH_SIZE 0x7fffffffffffffffULL
1404 #define MAX_HASH_HIGHEST_BIT 0x1000000000000000ULL
1407 struct mdt_body *body;
1408 struct lov_stripe_md *lsm;
1409 struct lmv_stripe_md *mea;
1410 #ifdef CONFIG_FS_POSIX_ACL
1411 struct posix_acl *posix_acl;
1413 struct mdt_remote_perm *remote_perm;
1414 struct obd_capa *mds_capa;
1415 struct obd_capa *oss_capa;
1418 struct md_open_data {
1419 struct obd_client_handle *mod_och;
1420 struct ptlrpc_request *mod_open_req;
1421 struct ptlrpc_request *mod_close_req;
1424 struct lookup_intent;
1427 int (*m_getstatus)(struct obd_export *, struct lu_fid *,
1428 struct obd_capa **);
1429 int (*m_change_cbdata)(struct obd_export *, const struct lu_fid *,
1430 ldlm_iterator_t, void *);
1431 int (*m_close)(struct obd_export *, struct md_op_data *,
1432 struct md_open_data *, struct ptlrpc_request **);
1433 int (*m_create)(struct obd_export *, struct md_op_data *,
1434 const void *, int, int, __u32, __u32, cfs_cap_t,
1435 __u64, struct ptlrpc_request **);
1436 int (*m_done_writing)(struct obd_export *, struct md_op_data *,
1437 struct md_open_data *);
1438 int (*m_enqueue)(struct obd_export *, struct ldlm_enqueue_info *,
1439 struct lookup_intent *, struct md_op_data *,
1440 struct lustre_handle *, void *, int,
1441 struct ptlrpc_request **, int);
1442 int (*m_getattr)(struct obd_export *, const struct lu_fid *,
1443 struct obd_capa *, obd_valid, int,
1444 struct ptlrpc_request **);
1445 int (*m_getattr_name)(struct obd_export *, const struct lu_fid *,
1446 struct obd_capa *, const char *, int, obd_valid,
1447 int, __u32, struct ptlrpc_request **);
1448 int (*m_intent_lock)(struct obd_export *, struct md_op_data *,
1449 void *, int, struct lookup_intent *, int,
1450 struct ptlrpc_request **,
1451 ldlm_blocking_callback, int);
1452 int (*m_link)(struct obd_export *, struct md_op_data *,
1453 struct ptlrpc_request **);
1454 int (*m_rename)(struct obd_export *, struct md_op_data *,
1455 const char *, int, const char *, int,
1456 struct ptlrpc_request **);
1457 int (*m_is_subdir)(struct obd_export *, const struct lu_fid *,
1458 const struct lu_fid *,
1459 struct ptlrpc_request **);
1460 int (*m_setattr)(struct obd_export *, struct md_op_data *, void *,
1461 int , void *, int, struct ptlrpc_request **,
1462 struct md_open_data **mod);
1463 int (*m_sync)(struct obd_export *, const struct lu_fid *,
1464 struct obd_capa *, struct ptlrpc_request **);
1465 int (*m_readpage)(struct obd_export *, const struct lu_fid *,
1466 struct obd_capa *, __u64, struct page *,
1467 struct ptlrpc_request **);
1469 int (*m_unlink)(struct obd_export *, struct md_op_data *,
1470 struct ptlrpc_request **);
1472 int (*m_setxattr)(struct obd_export *, const struct lu_fid *,
1473 struct obd_capa *, obd_valid, const char *,
1474 const char *, int, int, int, __u32,
1475 struct ptlrpc_request **);
1477 int (*m_getxattr)(struct obd_export *, const struct lu_fid *,
1478 struct obd_capa *, obd_valid, const char *,
1479 const char *, int, int, int,
1480 struct ptlrpc_request **);
1482 int (*m_init_ea_size)(struct obd_export *, int, int, int);
1484 int (*m_get_lustre_md)(struct obd_export *, struct ptlrpc_request *,
1485 struct obd_export *, struct obd_export *,
1486 struct lustre_md *);
1488 int (*m_free_lustre_md)(struct obd_export *, struct lustre_md *);
1490 int (*m_set_open_replay_data)(struct obd_export *,
1491 struct obd_client_handle *,
1492 struct ptlrpc_request *);
1493 int (*m_clear_open_replay_data)(struct obd_export *,
1494 struct obd_client_handle *);
1495 int (*m_set_lock_data)(struct obd_export *, __u64 *, void *);
1497 ldlm_mode_t (*m_lock_match)(struct obd_export *, int,
1498 const struct lu_fid *, ldlm_type_t,
1499 ldlm_policy_data_t *, ldlm_mode_t,
1500 struct lustre_handle *);
1502 int (*m_cancel_unused)(struct obd_export *, const struct lu_fid *,
1503 ldlm_policy_data_t *, ldlm_mode_t, int flags,
1505 int (*m_renew_capa)(struct obd_export *, struct obd_capa *oc,
1506 renew_capa_cb_t cb);
1507 int (*m_unpack_capa)(struct obd_export *, struct ptlrpc_request *,
1508 const struct req_msg_field *, struct obd_capa **);
1510 int (*m_get_remote_perm)(struct obd_export *, const struct lu_fid *,
1511 struct obd_capa *, __u32,
1512 struct ptlrpc_request **);
1514 int (*m_intent_getattr_async)(struct obd_export *,
1515 struct md_enqueue_info *,
1516 struct ldlm_enqueue_info *);
1518 int (*m_revalidate_lock)(struct obd_export *,
1519 struct lookup_intent *,
1523 * NOTE: If adding ops, add another LPROCFS_MD_OP_INIT() line to
1524 * lprocfs_alloc_md_stats() in obdclass/lprocfs_status.c. Also, add a
1525 * wrapper function in include/linux/obd_class.h.
1529 struct lsm_operations {
1530 void (*lsm_free)(struct lov_stripe_md *);
1531 int (*lsm_destroy)(struct lov_stripe_md *, struct obdo *oa,
1532 struct obd_export *md_exp);
1533 void (*lsm_stripe_by_index)(struct lov_stripe_md *, int *, obd_off *,
1535 void (*lsm_stripe_by_offset)(struct lov_stripe_md *, int *, obd_off *,
1537 obd_off (*lsm_stripe_offset_by_index)(struct lov_stripe_md *, int);
1538 obd_off (*lsm_stripe_offset_by_offset)(struct lov_stripe_md *, obd_off);
1539 int (*lsm_stripe_index_by_offset)(struct lov_stripe_md *, obd_off);
1540 int (*lsm_revalidate) (struct lov_stripe_md *, struct obd_device *obd);
1541 int (*lsm_lmm_verify) (struct lov_mds_md *lmm, int lmm_bytes,
1543 int (*lsm_unpackmd) (struct lov_obd *lov, struct lov_stripe_md *lsm,
1544 struct lov_mds_md *lmm);
1547 extern const struct lsm_operations lsm_v1_ops;
1548 extern const struct lsm_operations lsm_join_ops;
1549 extern const struct lsm_operations lsm_v3_ops;
1550 static inline const struct lsm_operations *lsm_op_find(int magic)
1555 case LOV_MAGIC_JOIN:
1556 return &lsm_join_ops;
1560 CERROR("Cannot recognize lsm_magic %d\n", magic);
1565 int lvfs_check_io_health(struct obd_device *obd, struct file *file);
1567 /* Requests for obd_extent_calc() */
1568 #define OBD_CALC_STRIPE_START 1
1569 #define OBD_CALC_STRIPE_END 2
1571 static inline void obd_transno_commit_cb(struct obd_device *obd, __u64 transno,
1575 CERROR("%s: transno "LPU64" commit error: %d\n",
1576 obd->obd_name, transno, error);
1579 if (transno > obd->obd_last_committed) {
1580 CDEBUG(D_INFO, "%s: transno "LPD64" committed\n",
1581 obd->obd_name, transno);
1582 obd->obd_last_committed = transno;
1583 ptlrpc_commit_replies (obd);
1585 CDEBUG(D_INFO, "%s: transno "LPD64" committed\n",
1586 obd->obd_name, transno);
1590 static inline void init_obd_quota_ops(quota_interface_t *interface,
1591 struct obd_ops *obd_ops)
1597 obd_ops->o_quotacheck = QUOTA_OP(interface, check);
1598 obd_ops->o_quotactl = QUOTA_OP(interface, ctl);
1599 obd_ops->o_quota_adjust_qunit = QUOTA_OP(interface, adjust_qunit);
1602 static inline __u64 oinfo_mdsno(struct obd_info *oinfo)
1604 return obdo_mdsno(oinfo->oi_oa);
1607 static inline struct lustre_capa *oinfo_capa(struct obd_info *oinfo)
1609 return oinfo->oi_capa;
1612 #endif /* __OBD_H */