1 /* -*- mode: c; c-basic-offset: 8; indent-tabs-mode: nil; -*-
2 * vim:expandtab:shiftwidth=8:tabstop=8:
6 * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
8 * This program is free software; you can redistribute it and/or modify
9 * it under the terms of the GNU General Public License version 2 only,
10 * as published by the Free Software Foundation.
12 * This program is distributed in the hope that it will be useful, but
13 * WITHOUT ANY WARRANTY; without even the implied warranty of
14 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
15 * General Public License version 2 for more details (a copy is included
16 * in the LICENSE file that accompanied this code).
18 * You should have received a copy of the GNU General Public License
19 * version 2 along with this program; If not, see
20 * http://www.sun.com/software/products/lustre/docs/GPLv2.pdf
22 * Please contact Sun Microsystems, Inc., 4150 Network Circle, Santa Clara,
23 * CA 95054 USA or visit www.sun.com if you need additional information or
29 * Copyright 2008 Sun Microsystems, Inc. All rights reserved
30 * Use is subject to license terms.
33 * This file is part of Lustre, http://www.lustre.org/
34 * Lustre is a trademark of Sun Microsystems, Inc.
40 #if defined(__linux__)
41 #include <linux/obd.h>
42 #elif defined(__APPLE__)
43 #include <darwin/obd.h>
44 #elif defined(__WINNT__)
45 #include <winnt/obd.h>
47 #error Unsupported operating system.
50 #define IOC_OSC_TYPE 'h'
51 #define IOC_OSC_MIN_NR 20
52 #define IOC_OSC_SET_ACTIVE _IOWR(IOC_OSC_TYPE, 21, struct obd_device *)
53 #define IOC_OSC_MAX_NR 50
55 #define IOC_MDC_TYPE 'i'
56 #define IOC_MDC_MIN_NR 20
57 /* Moved to lustre_user.h
58 #define IOC_MDC_LOOKUP _IOWR(IOC_MDC_TYPE, 20, struct obd_ioctl_data *)
59 #define IOC_MDC_GETSTRIPE _IOWR(IOC_MDC_TYPE, 21, struct lov_mds_md *) */
60 #define IOC_MDC_MAX_NR 50
62 #include <lustre/lustre_idl.h>
63 #include <lu_object.h>
64 #include <lustre_lib.h>
65 #include <lustre_export.h>
66 #include <lustre_quota.h>
67 #include <lustre_fld.h>
68 #include <lustre_capa.h>
69 #include <class_hash.h>
71 #include <libcfs/bitmap.h>
74 #define MAX_OBD_DEVICES 8192
76 /* this is really local to the OSC */
77 struct loi_oap_pages {
78 struct list_head lop_pending;
79 struct list_head lop_urgent;
80 struct list_head lop_pending_group;
90 struct lov_oinfo { /* per-stripe data structure */
91 __u64 loi_id; /* object ID on the target OST */
92 __u64 loi_gr; /* object group on the target OST */
93 int loi_ost_idx; /* OST stripe index in lov_tgt_desc->tgts */
94 int loi_ost_gen; /* generation of this loi_ost_idx */
96 /* used by the osc to keep track of what objects to build into rpcs */
97 struct loi_oap_pages loi_read_lop;
98 struct loi_oap_pages loi_write_lop;
99 /* _cli_ is poorly named, it should be _ready_ */
100 struct list_head loi_cli_item;
101 struct list_head loi_write_item;
102 struct list_head loi_read_item;
104 unsigned long loi_kms_valid:1;
105 __u64 loi_kms; /* known minimum size */
106 struct ost_lvb loi_lvb;
107 struct osc_async_rc loi_ar;
110 static inline void loi_init(struct lov_oinfo *loi)
112 CFS_INIT_LIST_HEAD(&loi->loi_read_lop.lop_pending);
113 CFS_INIT_LIST_HEAD(&loi->loi_read_lop.lop_urgent);
114 CFS_INIT_LIST_HEAD(&loi->loi_read_lop.lop_pending_group);
115 CFS_INIT_LIST_HEAD(&loi->loi_write_lop.lop_pending);
116 CFS_INIT_LIST_HEAD(&loi->loi_write_lop.lop_urgent);
117 CFS_INIT_LIST_HEAD(&loi->loi_write_lop.lop_pending_group);
118 CFS_INIT_LIST_HEAD(&loi->loi_cli_item);
119 CFS_INIT_LIST_HEAD(&loi->loi_write_item);
120 CFS_INIT_LIST_HEAD(&loi->loi_read_item);
123 /*extent array item for describing the joined file extent info*/
125 __u64 le_start; /* extent start */
126 __u64 le_len; /* extent length */
127 int le_loi_idx; /* extent #1 loi's index in lsm loi array */
128 int le_stripe_count; /* extent stripe count*/
131 /*Lov array info for describing joined file array EA info*/
132 struct lov_array_info {
133 struct llog_logid lai_array_id; /* MDS med llog object id */
134 unsigned lai_ext_count; /* number of extent count */
135 struct lov_extent *lai_ext_array; /* extent desc array */
138 struct lov_stripe_md {
140 pid_t lsm_lock_owner; /* debugging */
143 /* Public members. */
144 __u64 lw_object_id; /* lov object id */
145 __u64 lw_object_gr; /* lov object group */
146 __u64 lw_maxbytes; /* maximum possible file size */
148 /* LOV-private members start here -- only for use in lov/. */
150 __u32 lw_stripe_size; /* size of the stripe */
151 __u32 lw_pattern; /* striping pattern (RAID0, RAID1) */
152 unsigned lw_stripe_count; /* number of objects being striped over */
153 char lw_pool_name[MAXPOOLNAME]; /* pool name */
156 struct lov_array_info *lsm_array; /*Only for joined file array info*/
157 struct lov_oinfo *lsm_oinfo[0];
160 #define lsm_object_id lsm_wire.lw_object_id
161 #define lsm_object_gr lsm_wire.lw_object_gr
162 #define lsm_maxbytes lsm_wire.lw_maxbytes
163 #define lsm_magic lsm_wire.lw_magic
164 #define lsm_stripe_size lsm_wire.lw_stripe_size
165 #define lsm_pattern lsm_wire.lw_pattern
166 #define lsm_stripe_count lsm_wire.lw_stripe_count
167 #define lsm_pool_name lsm_wire.lw_pool_name
171 typedef int (*obd_enqueue_update_f)(struct obd_info *oinfo, int rc);
173 /* obd info for a particular level (lov, osc). */
175 /* Lock policy. It keeps an extent which is specific for a particular
176 * OSC. (e.g. lov_prep_enqueue_set initialises extent of the policy,
177 * and osc_enqueue passes it into ldlm_lock_match & ldlm_cli_enqueue. */
178 ldlm_policy_data_t oi_policy;
179 /* Flags used for set request specific flags:
180 - while lock handling, the flags obtained on the enqueue
181 request are set here.
182 - while stats, the flags used for control delay/resend.
185 /* Lock handle specific for every OSC lock. */
186 struct lustre_handle *oi_lockh;
187 /* lsm data specific for every OSC. */
188 struct lov_stripe_md *oi_md;
189 /* obdo data specific for every OSC, if needed at all. */
191 /* statfs data specific for every OSC, if needed at all. */
192 struct obd_statfs *oi_osfs;
193 /* An update callback which is called to update some data on upper
194 * level. E.g. it is used for update lsm->lsm_oinfo at every recieved
195 * request in osc level for enqueue requests. It is also possible to
196 * update some caller data from LOV layer if needed. */
197 obd_enqueue_update_f oi_cb_up;
198 /* oss capability, its type is obd_capa in client to avoid copy.
199 * in contrary its type is lustre_capa in OSS. */
203 /* compare all relevant fields. */
204 static inline int lov_stripe_md_cmp(struct lov_stripe_md *m1,
205 struct lov_stripe_md *m2)
208 * ->lsm_wire contains padding, but it should be zeroed out during
211 return memcmp(&m1->lsm_wire, &m2->lsm_wire, sizeof m1->lsm_wire);
214 void lov_stripe_lock(struct lov_stripe_md *md);
215 void lov_stripe_unlock(struct lov_stripe_md *md);
218 struct list_head typ_chain;
219 struct obd_ops *typ_dt_ops;
220 struct md_ops *typ_md_ops;
221 cfs_proc_dir_entry_t *typ_procroot;
224 struct lu_device_type *typ_lu;
225 spinlock_t obd_type_lock;
236 ASYNC_READY = 0x1, /* ap_make_ready will not be called before this
237 page is added to an rpc */
238 ASYNC_URGENT = 0x2, /* page must be put into an RPC before return */
239 ASYNC_COUNT_STABLE = 0x4, /* ap_refresh_count will not be called
240 to give the caller a chance to update
241 or cancel the size of the io */
242 ASYNC_GROUP_SYNC = 0x8, /* ap_completion will not be called, instead
243 the page is accounted for in the
244 obd_io_group given to
245 obd_queue_group_io */
248 struct obd_async_page_ops {
249 int (*ap_make_ready)(void *data, int cmd);
250 int (*ap_refresh_count)(void *data, int cmd);
251 void (*ap_fill_obdo)(void *data, int cmd, struct obdo *oa);
252 void (*ap_update_obdo)(void *data, int cmd, struct obdo *oa,
254 int (*ap_completion)(void *data, int cmd, struct obdo *oa, int rc);
255 struct obd_capa *(*ap_lookup_capa)(void *data, int cmd);
258 /* the `oig' is passed down from a caller of obd rw methods. the callee
259 * records enough state such that the caller can sleep on the oig and
260 * be woken when all the callees have finished their work */
261 struct obd_io_group {
263 atomic_t oig_refcount;
266 struct list_head oig_occ_list;
267 cfs_waitq_t oig_waitq;
270 /* the oig callback context lets the callee of obd rw methods register
271 * for callbacks from the caller. */
272 struct oig_callback_context {
273 struct list_head occ_oig_item;
274 /* called when the caller has received a signal while sleeping.
275 * callees of this method are encouraged to abort their state
276 * in the oig. This may be called multiple times. */
277 void (*occ_interrupted)(struct oig_callback_context *occ);
278 unsigned long interrupted:1;
281 /* Individual type definitions */
283 struct ost_server_data;
285 /* hold common fields for "target" device */
286 struct obd_device_target {
287 struct super_block *obt_sb;
288 atomic_t obt_quotachecking;
289 struct lustre_quota_ctxt obt_qctxt;
292 typedef void (*obd_pin_extent_cb)(void *data);
293 typedef int (*obd_page_removal_cb_t)(void *data, int discard);
294 typedef int (*obd_lock_cancel_cb)(struct ldlm_lock *,struct ldlm_lock_desc *,
299 LLOG_CONFIG_ORIG_CTXT = 0,
300 LLOG_CONFIG_REPL_CTXT,
301 LLOG_MDS_OST_ORIG_CTXT,
302 LLOG_MDS_OST_REPL_CTXT,
309 LLOG_LOVEA_ORIG_CTXT,
310 LLOG_LOVEA_REPL_CTXT,
314 #define FILTER_SUBDIR_COUNT 32 /* set to zero for no subdirs */
316 struct filter_subdirs {
317 cfs_dentry_t *dentry[FILTER_SUBDIR_COUNT];
327 /* NB this field MUST be first */
328 struct obd_device_target fo_obt;
329 const char *fo_fstype;
330 struct vfsmount *fo_vfsmnt;
333 cfs_dentry_t *fo_dentry_O;
334 cfs_dentry_t **fo_dentry_O_groups;
335 struct filter_subdirs *fo_dentry_O_sub;
336 struct semaphore fo_init_lock; /* group initialization lock */
337 int fo_committed_group;
340 spinlock_t fo_objidlock; /* protect fo_lastobjid */
341 spinlock_t fo_translock; /* protect fsd_last_transno */
342 struct file *fo_rcvd_filp;
343 struct file *fo_health_check_filp;
344 struct lr_server_data *fo_fsd;
345 unsigned long *fo_last_rcvd_slots;
346 __u64 fo_mount_count;
348 unsigned long fo_destroys_in_progress;
349 struct semaphore fo_create_locks[FILTER_SUBDIR_COUNT];
351 struct list_head fo_export_list;
354 obd_size fo_tot_dirty; /* protected by obd_osfs_lock */
355 obd_size fo_tot_granted; /* all values in bytes */
356 obd_size fo_tot_pending;
358 obd_size fo_readcache_max_filesize;
360 struct obd_import *fo_mdc_imp;
361 struct obd_uuid fo_mdc_uuid;
362 struct lustre_handle fo_mdc_conn;
363 struct file **fo_last_objid_files;
364 __u64 *fo_last_objids; /* last created objid for groups,
365 * protected by fo_objidlock */
367 struct semaphore fo_alloc_lock;
369 atomic_t fo_r_in_flight;
370 atomic_t fo_w_in_flight;
373 * per-filter pool of kiobuf's allocated by filter_common_setup() and
374 * torn down by filter_cleanup(). Contains OST_NUM_THREADS elements of
375 * which ->fo_iobuf_count were allocated.
377 * This pool contains kiobuf used by
378 * filter_{prep,commit}rw_{read,write}() and is shared by all OST
381 * Locking: none, each OST thread uses only one element, determined by
382 * its "ordinal number", ->t_id.
384 struct filter_iobuf **fo_iobuf_pool;
387 struct list_head fo_llog_list;
388 spinlock_t fo_llog_list_lock;
390 struct brw_stats fo_filter_stats;
391 struct lustre_quota_ctxt fo_quota_ctxt;
392 spinlock_t fo_quotacheck_lock;
393 atomic_t fo_quotachecking;
395 int fo_fmd_max_num; /* per exp filter_mod_data */
396 int fo_fmd_max_age; /* jiffies to fmd expiry */
399 rwlock_t fo_sptlrpc_lock;
400 struct sptlrpc_rule_set fo_sptlrpc_rset;
402 /* capability related */
403 unsigned int fo_fl_oss_capa;
404 struct list_head fo_capa_keys;
405 struct hlist_head *fo_capa_hash;
406 struct llog_commit_master *fo_lcm;
409 #define OSC_MAX_RIF_DEFAULT 8
410 #define OSC_MAX_RIF_MAX 256
411 #define OSC_MAX_DIRTY_DEFAULT (OSC_MAX_RIF_DEFAULT * 4)
412 #define OSC_MAX_DIRTY_MB_MAX 2048 /* arbitrary, but < MAX_LONG bytes */
413 #define OSC_DEFAULT_RESENDS 10
415 #define MDC_MAX_RIF_DEFAULT 8
416 #define MDC_MAX_RIF_MAX 512
422 struct rw_semaphore cl_sem;
423 struct obd_uuid cl_target_uuid;
424 struct obd_import *cl_import; /* ptlrpc connection state */
426 /* max_mds_easize is purely a performance thing so we don't have to
427 * call obd_size_diskmd() all the time. */
428 int cl_default_mds_easize;
429 int cl_max_mds_easize;
430 int cl_max_mds_cookiesize;
432 /* security configuration */
433 struct sptlrpc_rule_set cl_sptlrpc_rset;
434 enum lustre_sec_part cl_sec_part;
436 //struct llog_canceld_ctxt *cl_llcd; /* it's included by obd_llog_ctxt */
437 void *cl_llcd_offset;
439 /* the grant values are protected by loi_list_lock below */
440 long cl_dirty; /* all _dirty_ in bytes */
441 long cl_dirty_max; /* allowed w/o rpc */
442 long cl_avail_grant; /* bytes of credit for ost */
443 long cl_lost_grant; /* lost credits (trunc) */
444 struct list_head cl_cache_waiters; /* waiting for cache/grant */
446 /* keep track of objects that have lois that contain pages which
447 * have been queued for async brw. this lock also protects the
448 * lists of osc_client_pages that hang off of the loi */
450 * ->cl_loi_list_lock protects consistency of
451 * ->cl_loi_{ready,read,write}_list. ->ap_make_ready() and
452 * ->ap_completion() call-backs are executed under this lock. As we
453 * cannot guarantee that these call-backs never block on all platforms
454 * (as a matter of fact they do block on Mac OS X), type of
455 * ->cl_loi_list_lock is platform dependent: it's a spin-lock on Linux
456 * and blocking mutex on Mac OS X. (Alternative is to make this lock
457 * blocking everywhere, but we don't want to slow down fast-path of
458 * our main platform.)
460 * Exact type of ->cl_loi_list_lock is defined in arch/obd.h together
461 * with client_obd_list_{un,}lock() and
462 * client_obd_list_lock_{init,done}() functions.
464 client_obd_lock_t cl_loi_list_lock;
465 struct list_head cl_loi_ready_list;
466 struct list_head cl_loi_write_list;
467 struct list_head cl_loi_read_list;
470 /* just a sum of the loi/lop pending numbers to be exported by /proc */
471 int cl_pending_w_pages;
472 int cl_pending_r_pages;
473 int cl_max_pages_per_rpc;
474 int cl_max_rpcs_in_flight;
475 struct obd_histogram cl_read_rpc_hist;
476 struct obd_histogram cl_write_rpc_hist;
477 struct obd_histogram cl_read_page_hist;
478 struct obd_histogram cl_write_page_hist;
479 struct obd_histogram cl_read_offset_hist;
480 struct obd_histogram cl_write_offset_hist;
482 /* number of in flight destroy rpcs is limited to max_rpcs_in_flight */
483 atomic_t cl_destroy_in_flight;
484 cfs_waitq_t cl_destroy_waitq;
486 struct mdc_rpc_lock *cl_rpc_lock;
487 struct mdc_rpc_lock *cl_setattr_lock;
488 struct mdc_rpc_lock *cl_close_lock;
489 struct osc_creator cl_oscc;
492 struct semaphore cl_mgc_sem;
493 struct vfsmount *cl_mgc_vfsmnt;
494 struct dentry *cl_mgc_configs_dir;
495 atomic_t cl_mgc_refcount;
496 struct obd_export *cl_mgc_mgsexp;
498 /* checksumming for data sent over the network */
499 unsigned int cl_checksum:1; /* 0 = disabled, 1 = enabled */
500 /* supported checksum types that are worked out at connect time */
501 __u32 cl_supp_cksum_types;
502 /* checksum algorithm to be used */
503 cksum_type_t cl_cksum_type;
505 /* also protected by the poorly named _loi_list_lock lock above */
506 struct osc_async_rc cl_ar;
508 /* used by quotacheck */
509 int cl_qchk_stat; /* quotacheck stat of the peer */
511 /* sequence manager */
512 struct lu_client_seq *cl_seq;
514 atomic_t cl_resends; /* resend count */
516 /* Cache of triples */
517 struct lustre_cache *cl_cache;
518 obd_lock_cancel_cb cl_ext_lock_cancel_cb;
520 #define obd2cli_tgt(obd) ((char *)(obd)->u.cli.cl_target_uuid.uuid)
522 #define CL_NOT_QUOTACHECKED 1 /* client->cl_qchk_stat init value */
525 struct ptlrpc_service *mgs_service;
526 struct vfsmount *mgs_vfsmnt;
527 struct super_block *mgs_sb;
528 struct dentry *mgs_configs_dir;
529 struct dentry *mgs_fid_de;
530 struct list_head mgs_fs_db_list;
531 struct semaphore mgs_sem;
532 cfs_proc_dir_entry_t *mgs_proc_live;
536 /* NB this field MUST be first */
537 struct obd_device_target mds_obt;
538 struct ptlrpc_service *mds_service;
539 struct ptlrpc_service *mds_setattr_service;
540 struct ptlrpc_service *mds_readpage_service;
541 struct vfsmount *mds_vfsmnt;
542 cfs_dentry_t *mds_fid_de;
544 int mds_max_cookiesize;
545 struct file *mds_rcvd_filp;
546 spinlock_t mds_transno_lock;
547 __u64 mds_last_transno;
548 __u64 mds_mount_count;
550 unsigned long mds_atime_diff;
551 struct semaphore mds_epoch_sem;
552 struct ll_fid mds_rootfid;
553 struct lr_server_data *mds_server_data;
554 cfs_dentry_t *mds_pending_dir;
555 cfs_dentry_t *mds_logs_dir;
556 cfs_dentry_t *mds_objects_dir;
557 struct llog_handle *mds_cfg_llh;
558 struct obd_device *mds_osc_obd; /* XXX lov_obd */
559 struct obd_uuid mds_lov_uuid;
561 struct obd_export *mds_osc_exp; /* XXX lov_exp */
562 struct lov_desc mds_lov_desc;
565 /* mark pages dirty for write. */
566 bitmap_t *mds_lov_page_dirty;
567 /* array for store pages with obd_id */
568 void **mds_lov_page_array;
569 /* file for store objid */
570 struct file *mds_lov_objid_filp;
571 __u32 mds_lov_objid_count;
572 __u32 mds_lov_objid_lastpage;
573 __u32 mds_lov_objid_lastidx;
575 struct file *mds_health_check_filp;
576 unsigned long *mds_client_bitmap;
577 // struct upcall_cache *mds_group_hash;
579 struct lustre_quota_info mds_quota_info;
580 struct semaphore mds_qonoff_sem;
581 struct semaphore mds_health_sem;
582 unsigned long mds_fl_user_xattr:1,
584 mds_evict_ost_nids:1,
588 struct upcall_cache *mds_identity_cache;
590 /* for capability keys update */
591 struct lustre_capa_key *mds_capa_keys;
592 struct rw_semaphore mds_notify_lock;
596 extern __u32 mds_max_ost_index;
598 #define MDS_LOV_ALLOC_SIZE (CFS_PAGE_SIZE)
600 #define OBJID_PER_PAGE() (MDS_LOV_ALLOC_SIZE / sizeof(obd_id))
602 #define MDS_LOV_OBJID_PAGES_COUNT (mds_max_ost_index/OBJID_PER_PAGE())
604 extern int mds_lov_init_objids(struct obd_device *obd);
605 extern void mds_lov_destroy_objids(struct obd_device *obd);
618 struct lustre_handle eo_nl_lock;
623 struct ptlrpc_service *ost_service;
624 struct ptlrpc_service *ost_create_service;
625 struct ptlrpc_service *ost_io_service;
626 struct semaphore ost_health_sem;
629 struct echo_client_obd {
630 struct obd_export *ec_exp; /* the local connection to osc/lov */
632 struct list_head ec_objects;
638 struct obd_uuid lqo_uuid; /* ptlrpc's c_remote_uuid */
639 struct list_head lqo_oss_list; /* link to lov_qos */
640 __u32 lqo_ost_count; /* number of osts on this oss */
641 __u64 lqo_bavail; /* total bytes avail on OSS */
642 __u64 lqo_penalty; /* current penalty */
643 __u64 lqo_penalty_per_obj; /* penalty decrease every obj*/
647 struct lov_qos_oss *ltq_oss; /* oss info */
648 __u64 ltq_penalty; /* current penalty */
649 __u64 ltq_penalty_per_obj; /* penalty decrease every obj*/
650 __u64 ltq_weight; /* net weighting */
651 unsigned int ltq_usable:1; /* usable for striping */
654 /* Generic subset of OSTs */
656 __u32 *op_array; /* array of index of
658 unsigned int op_count; /* number of OSTs in the array */
659 unsigned int op_size; /* allocated size of lp_array */
660 rwlock_t op_rwlock; /* to protect lov_pool use */
663 /* Round-robin allocator data */
665 __u32 lqr_start_idx; /* start index of new inode */
666 __u32 lqr_offset_idx; /* aliasing for start_idx */
667 int lqr_start_count; /* reseed counter */
668 struct ost_pool lqr_pool; /* round-robin optimized list */
669 unsigned long lqr_dirty:1; /* recalc round-robin list */
672 /* Stripe placement optimization */
674 struct list_head lq_oss_list; /* list of OSSs that targets use */
675 struct rw_semaphore lq_rw_sem;
676 __u32 lq_active_oss_count;
677 unsigned int lq_prio_free; /* priority for free space */
678 struct lov_qos_rr lq_rr; /* round robin qos data */
679 unsigned long lq_dirty:1, /* recalc qos data */
680 lq_same_space:1,/* the ost's all have approx.
681 the same space avail */
682 lq_reset:1; /* zero current penalties */
685 struct lov_tgt_desc {
686 struct obd_uuid ltd_uuid;
687 struct obd_export *ltd_exp;
688 struct ltd_qos ltd_qos; /* qos info per target */
690 __u32 ltd_index; /* index in lov_obd->tgts */
691 unsigned long ltd_active:1,/* is this target up for requests */
692 ltd_activate:1,/* should this target be activated */
693 ltd_reap:1; /* should this target be deleted */
697 #define pool_tgt_size(_p) _p->pool_obds.op_size
698 #define pool_tgt_count(_p) _p->pool_obds.op_count
699 #define pool_tgt_array(_p) _p->pool_obds.op_array
700 #define pool_tgt_rwlock(_p) _p->pool_obds.op_rwlock
701 #define pool_tgt(_p, _i) _p->pool_lov->lov_tgts[_p->pool_obds.op_array[_i]]
704 char pool_name[MAXPOOLNAME + 1]; /* name of pool */
705 struct ost_pool pool_obds; /* pool members */
706 struct lov_qos_rr pool_rr; /* round robin qos */
707 struct hlist_node pool_hash; /* access by poolname */
708 struct list_head pool_list; /* serial access */
709 cfs_proc_dir_entry_t *pool_proc_entry; /* file in /proc */
710 struct lov_obd *pool_lov; /* lov obd to which this
715 struct lov_desc desc;
716 struct lov_tgt_desc **lov_tgts; /* sparse array */
717 struct ost_pool lov_packed; /* all OSTs in a packed
719 struct semaphore lov_lock;
720 struct obd_connect_data lov_ocd;
721 struct lov_qos lov_qos; /* qos info per lov */
722 atomic_t lov_refcount;
723 __u32 lov_tgt_count; /* how many OBD's */
724 __u32 lov_active_tgt_count; /* how many active */
725 __u32 lov_death_row;/* tgts scheduled to be deleted */
726 __u32 lov_tgt_size; /* size of tgts array */
728 obd_page_removal_cb_t lov_page_removal_cb;
729 obd_pin_extent_cb lov_page_pin_cb;
730 obd_lock_cancel_cb lov_lock_cancel_cb;
732 lustre_hash_t *lov_pools_hash_body; /* used for key access */
733 struct list_head lov_pool_list; /* used for sequential access */
734 cfs_proc_dir_entry_t *lov_pool_proc_entry;
737 struct lmv_tgt_desc {
738 struct obd_uuid ltd_uuid;
739 struct obd_export *ltd_exp;
740 int ltd_active; /* is this target up for requests */
742 struct semaphore ltd_fid_sem;
747 struct lu_client_fld lmv_fld;
749 struct lmv_desc desc;
750 struct obd_uuid cluuid;
751 struct obd_export *exp;
758 struct semaphore init_sem;
760 struct lmv_tgt_desc *tgts;
763 struct obd_connect_data *datas;
766 struct obd_connect_data conn_data;
769 struct niobuf_local {
774 cfs_dentry_t *dentry;
779 #define LUSTRE_FLD_NAME "fld"
780 #define LUSTRE_SEQ_NAME "seq"
782 #define LUSTRE_CMM_NAME "cmm"
783 #define LUSTRE_MDD_NAME "mdd"
784 #define LUSTRE_OSD_NAME "osd"
785 #define LUSTRE_LMV_NAME "lmv"
786 #define LUSTRE_CMM_MDC_NAME "cmm-mdc"
788 /* obd device type names */
789 /* FIXME all the references to LUSTRE_MDS_NAME should be swapped with LUSTRE_MDT_NAME */
790 #define LUSTRE_MDS_NAME "mds"
791 #define LUSTRE_MDT_NAME "mdt"
792 #define LUSTRE_MDC_NAME "mdc"
793 #define LUSTRE_OSS_NAME "ost" /* FIXME change name to oss */
794 #define LUSTRE_OST_NAME "obdfilter" /* FIXME change name to ost */
795 #define LUSTRE_OSC_NAME "osc"
796 #define LUSTRE_LOV_NAME "lov"
797 #define LUSTRE_MGS_NAME "mgs"
798 #define LUSTRE_MGC_NAME "mgc"
800 #define LUSTRE_CACHEOBD_NAME "cobd"
801 #define LUSTRE_ECHO_NAME "obdecho"
802 #define LUSTRE_ECHO_CLIENT_NAME "echo_client"
804 /* Constant obd names (post-rename) */
805 #define LUSTRE_MDS_OBDNAME "MDS"
806 #define LUSTRE_OSS_OBDNAME "OSS"
807 #define LUSTRE_MGS_OBDNAME "MGS"
808 #define LUSTRE_MGC_OBDNAME "MGC"
810 /* Don't conflict with on-wire flags OBD_BRW_WRITE, etc */
811 #define N_LOCAL_TEMP_PAGE 0x10000000
813 struct obd_trans_info {
816 /* Only used on the server side for tracking acks. */
817 struct oti_req_ack_lock {
818 struct lustre_handle lock;
822 struct llog_cookie oti_onecookie;
823 struct llog_cookie *oti_logcookies;
826 /* initial thread handling transaction */
830 struct obd_uuid *oti_ost_uuid;
833 static inline void oti_init(struct obd_trans_info *oti,
834 struct ptlrpc_request *req)
838 memset(oti, 0, sizeof(*oti));
843 oti->oti_xid = req->rq_xid;
845 if (req->rq_repmsg != NULL)
846 oti->oti_transno = lustre_msg_get_transno(req->rq_repmsg);
847 oti->oti_thread_id = req->rq_svc_thread ? req->rq_svc_thread->t_id : -1;
848 if (req->rq_reqmsg != NULL)
849 oti->oti_conn_cnt = lustre_msg_get_conn_cnt(req->rq_reqmsg);
852 static inline void oti_alloc_cookies(struct obd_trans_info *oti,int num_cookies)
857 if (num_cookies == 1)
858 oti->oti_logcookies = &oti->oti_onecookie;
860 OBD_ALLOC(oti->oti_logcookies,
861 num_cookies * sizeof(oti->oti_onecookie));
863 oti->oti_numcookies = num_cookies;
866 static inline void oti_free_cookies(struct obd_trans_info *oti)
868 if (!oti || !oti->oti_logcookies)
871 if (oti->oti_logcookies == &oti->oti_onecookie)
872 LASSERT(oti->oti_numcookies == 1);
874 OBD_FREE(oti->oti_logcookies,
875 oti->oti_numcookies * sizeof(oti->oti_onecookie));
876 oti->oti_logcookies = NULL;
877 oti->oti_numcookies = 0;
881 * Events signalled through obd_notify() upcall-chain.
883 enum obd_notify_event {
884 /* Device activated */
886 /* Device deactivated */
888 /* Device disconnected */
890 /* Connect data for import were changed */
893 OBD_NOTIFY_SYNC_NONBLOCK,
895 /* Configuration event */
899 /* bit-mask flags for config events */
901 CONFIG_LOG = 0x1, /* finished processing config log */
902 CONFIG_SYNC = 0x2 /* mdt synced 1 ost */
906 * Data structure used to pass obd_notify()-event to non-obd listeners (llite
907 * and liblustre being main examples).
909 struct obd_notify_upcall {
910 int (*onu_upcall)(struct obd_device *host, struct obd_device *watched,
911 enum obd_notify_event ev, void *owner);
912 /* Opaque datum supplied by upper layer listener */
916 struct target_recovery_data {
917 svc_handler_t trd_recovery_handler;
918 pid_t trd_processing_task;
919 struct completion trd_starting;
920 struct completion trd_finishing;
923 #define OBD_LLOG_GROUP 0
926 FILTER_GROUP_LLOG = 1,
931 struct obd_llog_group {
932 struct list_head olg_list;
934 struct llog_ctxt *olg_ctxts[LLOG_MAX_CTXTS];
935 cfs_waitq_t olg_waitq;
937 struct obd_export *olg_exp;
938 int olg_initializing;
941 /* corresponds to one of the obd's */
942 #define MAX_OBD_NAME 128
943 #define OBD_DEVICE_MAGIC 0XAB5CD6EF
945 struct obd_type *obd_type;
948 /* common and UUID name of this device */
949 char obd_name[MAX_OBD_NAME];
950 struct obd_uuid obd_uuid;
952 struct lu_device *obd_lu_dev;
955 unsigned long obd_attached:1, /* finished attach */
956 obd_set_up:1, /* finished setup */
957 obd_recovering:1, /* there are recoverable clients */
958 obd_abort_recovery:1,/* somebody ioctl'ed us to abort */
959 obd_replayable:1, /* recovery is enabled; inform clients */
960 obd_no_transno:1, /* no committed-transno notification */
961 obd_no_recov:1, /* fail instead of retry messages */
962 obd_req_replaying:1, /* replaying requests */
963 obd_stopping:1, /* started cleanup */
964 obd_starting:1, /* started setup */
965 obd_force:1, /* cleanup with > 0 obd refcount */
966 obd_fail:1, /* cleanup with failover */
967 obd_async_recov:1, /* allow asyncronous orphan cleanup */
968 obd_no_conn:1, /* deny new connections */
969 obd_inactive:1; /* device active/inactive
970 * (for /proc/status only!!) */
971 /* uuid-export hash body */
972 struct lustre_hash *obd_uuid_hash;
973 /* nid-export hash body */
974 struct lustre_hash *obd_nid_hash;
976 struct lustre_hash *obd_nid_stats_hash;
977 struct list_head obd_nid_stats;
978 atomic_t obd_refcount;
979 cfs_waitq_t obd_refcount_waitq;
980 struct list_head obd_exports;
982 spinlock_t obd_nid_lock;
983 struct ldlm_namespace *obd_namespace;
984 struct ptlrpc_client obd_ldlm_client; /* XXX OST/MDS only */
985 /* a spinlock is OK for what we do now, may need a semaphore later */
986 spinlock_t obd_dev_lock;
987 struct semaphore obd_dev_sem;
988 __u64 obd_last_committed;
989 struct fsfilt_operations *obd_fsops;
990 spinlock_t obd_osfs_lock;
991 struct obd_statfs obd_osfs; /* locked by obd_osfs_lock */
993 struct lvfs_run_ctxt obd_lvfs_ctxt;
994 struct obd_llog_group obd_olg; /* default llog group */
995 struct obd_device *obd_observer;
996 struct obd_notify_upcall obd_upcall;
997 struct obd_export *obd_self_export;
998 /* list of exports in LRU order, for ping evictor, with obd_dev_lock */
999 struct list_head obd_exports_timed;
1000 time_t obd_eviction_timer; /* for ping evictor */
1002 /* XXX encapsulate all this recovery data into one struct */
1003 svc_handler_t obd_recovery_handler;
1004 pid_t obd_processing_task;
1006 int obd_max_recoverable_clients;
1007 int obd_connected_clients;
1008 int obd_recoverable_clients;
1009 spinlock_t obd_processing_task_lock; /* BH lock (timer) */
1010 __u64 obd_next_recovery_transno;
1011 int obd_replayed_requests;
1012 int obd_requests_queued_for_recovery;
1013 cfs_waitq_t obd_next_transno_waitq;
1014 struct list_head obd_uncommitted_replies;
1015 spinlock_t obd_uncommitted_replies_lock;
1016 cfs_timer_t obd_recovery_timer;
1017 time_t obd_recovery_start; /* seconds */
1018 time_t obd_recovery_end; /* seconds, for lprocfs_status */
1019 time_t obd_recovery_max_time; /* seconds, bz13079 */
1020 int obd_recovery_timeout;
1022 /* new recovery stuff from CMD2 */
1023 struct target_recovery_data obd_recovery_data;
1024 int obd_replayed_locks;
1025 atomic_t obd_req_replay_clients;
1026 atomic_t obd_lock_replay_clients;
1027 struct list_head obd_req_replay_queue;
1028 struct list_head obd_lock_replay_queue;
1029 struct list_head obd_final_req_queue;
1030 int obd_recovery_stage;
1033 struct obd_device_target obt;
1034 struct filter_obd filter;
1036 struct client_obd cli;
1038 struct echo_client_obd echo_client;
1039 struct echo_obd echo;
1044 /* Fields used by LProcFS */
1045 unsigned int obd_cntr_base;
1046 struct lprocfs_stats *obd_stats;
1048 unsigned int md_cntr_base;
1049 struct lprocfs_stats *md_stats;
1051 cfs_proc_dir_entry_t *obd_proc_entry;
1052 cfs_proc_dir_entry_t *obd_proc_exports_entry;
1053 cfs_proc_dir_entry_t *obd_svc_procroot;
1054 struct lprocfs_stats *obd_svc_stats;
1055 atomic_t obd_evict_inprogress;
1056 cfs_waitq_t obd_evict_inprogress_waitq;
1059 * Ldlm pool part. Save last calculated SLV and Limit.
1061 rwlock_t obd_pool_lock;
1066 #define OBD_OPT_FORCE 0x0001
1067 #define OBD_OPT_FAILOVER 0x0002
1069 #define OBD_LLOG_FL_SENDNOW 0x0001
1071 enum obd_cleanup_stage {
1072 /* Special case hack for MDS LOVs */
1074 /* can be directly mapped to .ldto_device_fini() */
1075 OBD_CLEANUP_EXPORTS,
1078 /* get/set_info keys */
1079 #define KEY_READ_ONLY "read-only"
1080 #define KEY_MDS_CONN "mds_conn"
1081 #define KEY_NEXT_ID "next_id"
1082 #define KEY_LOVDESC "lovdesc"
1083 #define KEY_INIT_RECOV "initial_recov"
1084 #define KEY_INIT_RECOV_BACKUP "init_recov_bk"
1085 #define KEY_FLUSH_CTX "flush_ctx"
1086 #define KEY_CAPA_KEY "capa_key"
1087 #define KEY_CONN_DATA "conn_data"
1088 #define KEY_MAX_EASIZE "max_easize"
1089 #define KEY_REVIMP_UPD "revimp_update"
1090 #define KEY_LOV_IDX "lov_idx"
1091 #define KEY_LAST_ID "last_id"
1092 #define KEY_READONLY "read-only"
1093 #define KEY_LOCK_TO_STRIPE "lock_to_stripe"
1094 #define KEY_CHECKSUM "checksum"
1095 #define KEY_UNLINKED "unlinked"
1096 #define KEY_EVICT_BY_NID "evict_by_nid"
1097 #define KEY_REGISTER_TARGET "register_target"
1098 #define KEY_SET_FS "set_fs"
1099 #define KEY_CLEAR_FS "clear_fs"
1100 #define KEY_BLOCKSIZE "blocksize"
1101 #define KEY_BLOCKSIZE_BITS "blocksize_bits"
1102 #define KEY_FIEMAP "FIEMAP"
1104 #define KEY_INTERMDS "inter_mds"
1105 #define KEY_ASYNC "async"
1109 static inline int it_to_lock_mode(struct lookup_intent *it)
1111 /* CREAT needs to be tested before open (both could be set) */
1112 if (it->it_op & IT_CREAT)
1114 else if (it->it_op & (IT_READDIR | IT_GETATTR | IT_OPEN | IT_LOOKUP))
1117 LASSERTF(0, "Invalid it_op: %d\n", it->it_op);
1122 struct lu_fid op_fid1; /* operation fid1 (usualy parent) */
1123 struct lu_fid op_fid2; /* operation fid2 (usualy child) */
1124 struct lu_fid op_fid3; /* 2 extra fids to find conflicting */
1125 struct lu_fid op_fid4; /* to the operation locks. */
1126 mdsno_t op_mds; /* what mds server open will go to */
1127 struct lustre_handle op_handle;
1129 const char *op_name;
1132 struct lmv_stripe_md *op_mea1;
1133 struct lmv_stripe_md *op_mea2;
1134 __u32 op_suppgids[2];
1140 /* iattr fields and blocks. */
1141 struct iattr op_attr;
1143 #if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,14)
1144 unsigned int op_attr_flags;
1147 loff_t op_attr_blocks;
1149 /* Size-on-MDS epoch and flags. */
1154 struct obd_capa *op_capa1;
1155 struct obd_capa *op_capa2;
1157 /* Various operation flags. */
1160 /* Operation type */
1164 struct md_enqueue_info;
1165 /* metadata stat-ahead */
1166 typedef int (* md_enqueue_cb_t)(struct ptlrpc_request *req,
1167 struct md_enqueue_info *minfo,
1170 struct md_enqueue_info {
1171 struct md_op_data mi_data;
1172 struct lookup_intent mi_it;
1173 struct lustre_handle mi_lockh;
1174 struct dentry *mi_dentry;
1175 md_enqueue_cb_t mi_cb;
1176 unsigned int mi_generation;
1181 struct module *o_owner;
1182 int (*o_iocontrol)(unsigned int cmd, struct obd_export *exp, int len,
1183 void *karg, void *uarg);
1184 int (*o_get_info)(struct obd_export *, __u32 keylen, void *key,
1185 __u32 *vallen, void *val, struct lov_stripe_md *lsm);
1186 int (*o_set_info_async)(struct obd_export *, __u32 keylen, void *key,
1187 __u32 vallen, void *val,
1188 struct ptlrpc_request_set *set);
1189 int (*o_attach)(struct obd_device *dev, obd_count len, void *data);
1190 int (*o_detach)(struct obd_device *dev);
1191 int (*o_setup) (struct obd_device *dev, struct lustre_cfg *cfg);
1192 int (*o_precleanup)(struct obd_device *dev,
1193 enum obd_cleanup_stage cleanup_stage);
1194 int (*o_cleanup)(struct obd_device *dev);
1195 int (*o_process_config)(struct obd_device *dev, obd_count len,
1197 int (*o_postrecov)(struct obd_device *dev);
1198 int (*o_add_conn)(struct obd_import *imp, struct obd_uuid *uuid,
1200 int (*o_del_conn)(struct obd_import *imp, struct obd_uuid *uuid);
1201 /* connect to the target device with given connection
1202 * data. @ocd->ocd_connect_flags is modified to reflect flags actually
1203 * granted by the target, which are guaranteed to be a subset of flags
1204 * asked for. If @ocd == NULL, use default parameters. */
1205 int (*o_connect)(const struct lu_env *env,
1206 struct lustre_handle *conn, struct obd_device *src,
1207 struct obd_uuid *cluuid, struct obd_connect_data *ocd,
1209 int (*o_reconnect)(const struct lu_env *env,
1210 struct obd_export *exp, struct obd_device *src,
1211 struct obd_uuid *cluuid,
1212 struct obd_connect_data *ocd);
1213 int (*o_disconnect)(struct obd_export *exp);
1215 /* Initialize/finalize fids infrastructure. */
1216 int (*o_fid_init)(struct obd_export *exp);
1217 int (*o_fid_fini)(struct obd_export *exp);
1219 /* Allocate new fid according to passed @hint. */
1220 int (*o_fid_alloc)(struct obd_export *exp, struct lu_fid *fid,
1221 struct md_op_data *op_data);
1224 * Object with @fid is getting deleted, we may want to do something
1227 int (*o_fid_delete)(struct obd_export *exp, const struct lu_fid *fid);
1229 int (*o_statfs)(struct obd_device *obd, struct obd_statfs *osfs,
1230 __u64 max_age, __u32 flags);
1231 int (*o_statfs_async)(struct obd_device *obd, struct obd_info *oinfo,
1232 __u64 max_age, struct ptlrpc_request_set *set);
1233 int (*o_packmd)(struct obd_export *exp, struct lov_mds_md **disk_tgt,
1234 struct lov_stripe_md *mem_src);
1235 int (*o_unpackmd)(struct obd_export *exp,struct lov_stripe_md **mem_tgt,
1236 struct lov_mds_md *disk_src, int disk_len);
1237 int (*o_checkmd)(struct obd_export *exp, struct obd_export *md_exp,
1238 struct lov_stripe_md *mem_tgt);
1239 int (*o_preallocate)(struct lustre_handle *, obd_count *req,
1241 /* FIXME: add fid capability support for create & destroy! */
1242 int (*o_precreate)(struct obd_export *exp);
1243 int (*o_create)(struct obd_export *exp, struct obdo *oa,
1244 struct lov_stripe_md **ea, struct obd_trans_info *oti);
1245 int (*o_destroy)(struct obd_export *exp, struct obdo *oa,
1246 struct lov_stripe_md *ea, struct obd_trans_info *oti,
1247 struct obd_export *md_exp);
1248 int (*o_setattr)(struct obd_export *exp, struct obd_info *oinfo,
1249 struct obd_trans_info *oti);
1250 int (*o_setattr_async)(struct obd_export *exp, struct obd_info *oinfo,
1251 struct obd_trans_info *oti,
1252 struct ptlrpc_request_set *rqset);
1253 int (*o_getattr)(struct obd_export *exp, struct obd_info *oinfo);
1254 int (*o_getattr_async)(struct obd_export *exp, struct obd_info *oinfo,
1255 struct ptlrpc_request_set *set);
1256 int (*o_brw)(int rw, struct obd_export *exp, struct obd_info *oinfo,
1257 obd_count oa_bufs, struct brw_page *pgarr,
1258 struct obd_trans_info *oti);
1259 int (*o_brw_async)(int rw, struct obd_export *exp,
1260 struct obd_info *oinfo, obd_count oa_bufs,
1261 struct brw_page *pgarr, struct obd_trans_info *oti,
1262 struct ptlrpc_request_set *);
1263 int (*o_prep_async_page)(struct obd_export *exp,
1264 struct lov_stripe_md *lsm,
1265 struct lov_oinfo *loi,
1266 cfs_page_t *page, obd_off offset,
1267 struct obd_async_page_ops *ops, void *data,
1268 void **res, int nocache,
1269 struct lustre_handle *lockh);
1270 int (*o_reget_short_lock)(struct obd_export *exp,
1271 struct lov_stripe_md *lsm,
1273 obd_off start, obd_off end,
1275 int (*o_release_short_lock)(struct obd_export *exp,
1276 struct lov_stripe_md *lsm, obd_off end,
1277 void *cookie, int rw);
1278 int (*o_queue_async_io)(struct obd_export *exp,
1279 struct lov_stripe_md *lsm,
1280 struct lov_oinfo *loi, void *cookie,
1281 int cmd, obd_off off, int count,
1282 obd_flag brw_flags, obd_flag async_flags);
1283 int (*o_queue_group_io)(struct obd_export *exp,
1284 struct lov_stripe_md *lsm,
1285 struct lov_oinfo *loi,
1286 struct obd_io_group *oig,
1287 void *cookie, int cmd, obd_off off, int count,
1288 obd_flag brw_flags, obd_flag async_flags);
1289 int (*o_trigger_group_io)(struct obd_export *exp,
1290 struct lov_stripe_md *lsm,
1291 struct lov_oinfo *loi,
1292 struct obd_io_group *oig);
1293 int (*o_set_async_flags)(struct obd_export *exp,
1294 struct lov_stripe_md *lsm,
1295 struct lov_oinfo *loi, void *cookie,
1296 obd_flag async_flags);
1297 int (*o_teardown_async_page)(struct obd_export *exp,
1298 struct lov_stripe_md *lsm,
1299 struct lov_oinfo *loi, void *cookie);
1300 int (*o_merge_lvb)(struct obd_export *exp, struct lov_stripe_md *lsm,
1301 struct ost_lvb *lvb, int kms_only);
1302 int (*o_adjust_kms)(struct obd_export *exp, struct lov_stripe_md *lsm,
1303 obd_off size, int shrink);
1304 int (*o_punch)(struct obd_export *exp, struct obd_info *oinfo,
1305 struct obd_trans_info *oti,
1306 struct ptlrpc_request_set *rqset);
1307 int (*o_sync)(struct obd_export *exp, struct obdo *oa,
1308 struct lov_stripe_md *ea, obd_size start, obd_size end,
1310 int (*o_migrate)(struct lustre_handle *conn, struct lov_stripe_md *dst,
1311 struct lov_stripe_md *src, obd_size start,
1312 obd_size end, struct obd_trans_info *oti);
1313 int (*o_copy)(struct lustre_handle *dstconn, struct lov_stripe_md *dst,
1314 struct lustre_handle *srconn, struct lov_stripe_md *src,
1315 obd_size start, obd_size end, struct obd_trans_info *);
1316 int (*o_iterate)(struct lustre_handle *conn,
1317 int (*)(obd_id, obd_gr, void *),
1318 obd_id *startid, obd_gr group, void *data);
1319 int (*o_preprw)(int cmd, struct obd_export *exp, struct obdo *oa,
1320 int objcount, struct obd_ioobj *obj,
1321 int niocount, struct niobuf_remote *remote,
1322 struct niobuf_local *local, 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 int niocount, struct niobuf_local *local,
1327 struct obd_trans_info *oti, int rc);
1328 int (*o_enqueue)(struct obd_export *, struct obd_info *oinfo,
1329 struct ldlm_enqueue_info *einfo,
1330 struct ptlrpc_request_set *rqset);
1331 int (*o_match)(struct obd_export *, struct lov_stripe_md *, __u32 type,
1332 ldlm_policy_data_t *, __u32 mode, int *flags, void *data,
1333 struct lustre_handle *lockh);
1334 int (*o_change_cbdata)(struct obd_export *, struct lov_stripe_md *,
1335 ldlm_iterator_t it, void *data);
1336 int (*o_cancel)(struct obd_export *, struct lov_stripe_md *md,
1337 __u32 mode, struct lustre_handle *);
1338 int (*o_cancel_unused)(struct obd_export *, struct lov_stripe_md *,
1339 int flags, void *opaque);
1340 int (*o_join_lru)(struct obd_export *, struct lov_stripe_md *,
1342 int (*o_init_export)(struct obd_export *exp);
1343 int (*o_destroy_export)(struct obd_export *exp);
1344 int (*o_extent_calc)(struct obd_export *, struct lov_stripe_md *,
1345 int cmd, obd_off *);
1347 /* llog related obd_methods */
1348 int (*o_llog_init)(struct obd_device *obd, struct obd_llog_group *grp,
1349 struct obd_device *disk_obd, int count,
1350 struct llog_catid *logid, struct obd_uuid *uuid);
1351 int (*o_llog_finish)(struct obd_device *obd, int count);
1352 int (*o_llog_connect)(struct obd_export *, struct llogd_conn_body *);
1354 /* metadata-only methods */
1355 int (*o_pin)(struct obd_export *, const struct lu_fid *fid,
1356 struct obd_capa *, struct obd_client_handle *, int flag);
1357 int (*o_unpin)(struct obd_export *, struct obd_client_handle *, int);
1359 int (*o_import_event)(struct obd_device *, struct obd_import *,
1360 enum obd_import_event);
1362 int (*o_notify)(struct obd_device *obd, struct obd_device *watched,
1363 enum obd_notify_event ev, void *data);
1365 int (*o_health_check)(struct obd_device *);
1366 struct obd_uuid *(*o_get_uuid) (struct obd_export *exp);
1369 int (*o_quotacheck)(struct obd_export *, struct obd_quotactl *);
1370 int (*o_quotactl)(struct obd_export *, struct obd_quotactl *);
1372 int (*o_ping)(struct obd_export *exp);
1374 int (*o_register_page_removal_cb)(struct obd_export *exp,
1375 obd_page_removal_cb_t cb,
1376 obd_pin_extent_cb pin_cb);
1377 int (*o_unregister_page_removal_cb)(struct obd_export *exp,
1378 obd_page_removal_cb_t cb);
1379 int (*o_register_lock_cancel_cb)(struct obd_export *exp,
1380 obd_lock_cancel_cb cb);
1381 int (*o_unregister_lock_cancel_cb)(struct obd_export *exp,
1382 obd_lock_cancel_cb cb);
1384 int (*o_pool_new)(struct obd_device *obd, char *poolname);
1385 int (*o_pool_del)(struct obd_device *obd, char *poolname);
1386 int (*o_pool_add)(struct obd_device *obd, char *poolname,
1388 int (*o_pool_rem)(struct obd_device *obd, char *poolname,
1391 * NOTE: If adding ops, add another LPROCFS_OBD_OP_INIT() line
1392 * to lprocfs_alloc_obd_stats() in obdclass/lprocfs_status.c.
1393 * Also, add a wrapper function in include/linux/obd_class.h. */
1396 /* TODO: lmv_stripe_md should contain mds capabilities for all slave fids */
1397 struct lmv_stripe_md {
1402 struct lu_fid mea_ids[0];
1406 LUSTRE_OPC_MKDIR = (1 << 0),
1407 LUSTRE_OPC_SYMLINK = (1 << 1),
1408 LUSTRE_OPC_MKNOD = (1 << 2),
1409 LUSTRE_OPC_CREATE = (1 << 3),
1410 LUSTRE_OPC_ANY = (1 << 4)
1413 /* lmv structures */
1414 #define MEA_MAGIC_LAST_CHAR 0xb2221ca1
1415 #define MEA_MAGIC_ALL_CHARS 0xb222a11c
1416 #define MEA_MAGIC_HASH_SEGMENT 0xb222a11b
1418 #define MAX_HASH_SIZE_32 0x7fffffffUL
1419 #define MAX_HASH_SIZE 0x7fffffffffffffffULL
1420 #define MAX_HASH_HIGHEST_BIT 0x1000000000000000
1423 struct mdt_body *body;
1424 struct lov_stripe_md *lsm;
1425 struct lmv_stripe_md *mea;
1426 #ifdef CONFIG_FS_POSIX_ACL
1427 struct posix_acl *posix_acl;
1429 struct mdt_remote_perm *remote_perm;
1430 struct obd_capa *mds_capa;
1431 struct obd_capa *oss_capa;
1434 struct md_open_data {
1435 struct obd_client_handle *mod_och;
1436 struct list_head mod_replay_list;
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_close)(struct obd_export *, struct md_op_data *,
1447 struct md_open_data *, struct ptlrpc_request **);
1448 int (*m_create)(struct obd_export *, struct md_op_data *,
1449 const void *, int, int, __u32, __u32, cfs_cap_t,
1450 __u64, struct ptlrpc_request **);
1451 int (*m_done_writing)(struct obd_export *, struct md_op_data *,
1452 struct md_open_data *);
1453 int (*m_enqueue)(struct obd_export *, struct ldlm_enqueue_info *,
1454 struct lookup_intent *, struct md_op_data *,
1455 struct lustre_handle *, void *, int,
1456 struct ptlrpc_request **, int);
1457 int (*m_getattr)(struct obd_export *, const struct lu_fid *,
1458 struct obd_capa *, obd_valid, int,
1459 struct ptlrpc_request **);
1460 int (*m_getattr_name)(struct obd_export *, const struct lu_fid *,
1461 struct obd_capa *, const char *, int, obd_valid,
1462 int, __u32, 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 *);
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);
1523 int (*m_get_remote_perm)(struct obd_export *, const struct lu_fid *,
1524 struct obd_capa *, __u32,
1525 struct ptlrpc_request **);
1527 int (*m_intent_getattr_async)(struct obd_export *,
1528 struct md_enqueue_info *,
1529 struct ldlm_enqueue_info *);
1531 int (*m_revalidate_lock)(struct obd_export *,
1532 struct lookup_intent *,
1536 * NOTE: If adding ops, add another LPROCFS_MD_OP_INIT() line to
1537 * lprocfs_alloc_md_stats() in obdclass/lprocfs_status.c. Also, add a
1538 * wrapper function in include/linux/obd_class.h.
1542 struct lsm_operations {
1543 void (*lsm_free)(struct lov_stripe_md *);
1544 int (*lsm_destroy)(struct lov_stripe_md *, struct obdo *oa,
1545 struct obd_export *md_exp);
1546 void (*lsm_stripe_by_index)(struct lov_stripe_md *, int *, obd_off *,
1548 void (*lsm_stripe_by_offset)(struct lov_stripe_md *, int *, obd_off *,
1550 obd_off (*lsm_stripe_offset_by_index)(struct lov_stripe_md *, int);
1551 obd_off (*lsm_stripe_offset_by_offset)(struct lov_stripe_md *, obd_off);
1552 int (*lsm_stripe_index_by_offset)(struct lov_stripe_md *, obd_off);
1553 int (*lsm_revalidate) (struct lov_stripe_md *, struct obd_device *obd);
1554 int (*lsm_lmm_verify) (struct lov_mds_md *lmm, int lmm_bytes,
1556 int (*lsm_unpackmd) (struct lov_obd *lov, struct lov_stripe_md *lsm,
1557 struct lov_mds_md *lmm);
1560 extern struct lsm_operations lsm_v1_ops;
1561 extern struct lsm_operations lsm_join_ops;
1562 extern struct lsm_operations lsm_v3_ops;
1563 static inline struct lsm_operations *lsm_op_find(int magic)
1568 case LOV_MAGIC_JOIN:
1569 return &lsm_join_ops;
1573 CERROR("Cannot recognize lsm_magic %d\n", magic);
1578 int lvfs_check_io_health(struct obd_device *obd, struct file *file);
1580 /* Requests for obd_extent_calc() */
1581 #define OBD_CALC_STRIPE_START 1
1582 #define OBD_CALC_STRIPE_END 2
1584 static inline void obd_transno_commit_cb(struct obd_device *obd, __u64 transno,
1588 CERROR("%s: transno "LPU64" commit error: %d\n",
1589 obd->obd_name, transno, error);
1592 if (transno > obd->obd_last_committed) {
1593 CDEBUG(D_HA, "%s: transno "LPD64" committed\n",
1594 obd->obd_name, transno);
1595 obd->obd_last_committed = transno;
1596 ptlrpc_commit_replies (obd);
1598 CDEBUG(D_INFO, "%s: transno "LPD64" committed\n",
1599 obd->obd_name, transno);
1603 static inline void init_obd_quota_ops(quota_interface_t *interface,
1604 struct obd_ops *obd_ops)
1610 obd_ops->o_quotacheck = QUOTA_OP(interface, check);
1611 obd_ops->o_quotactl = QUOTA_OP(interface, ctl);
1614 static inline __u64 oinfo_mdsno(struct obd_info *oinfo)
1616 return oinfo->oi_oa->o_gr - FILTER_GROUP_MDS0;
1619 static inline struct lustre_capa *oinfo_capa(struct obd_info *oinfo)
1621 return oinfo->oi_capa;
1624 #endif /* __OBD_H */