1 /* -*- mode: c; c-basic-offset: 8; indent-tabs-mode: nil; -*-
2 * vim:expandtab:shiftwidth=8:tabstop=8:
10 #elif defined(__APPLE__)
11 #include <darwin/obd.h>
12 #elif defined(__WINNT__)
13 #include <winnt/obd.h>
15 #error Unsupported operating system.
18 #define IOC_OSC_TYPE 'h'
19 #define IOC_OSC_MIN_NR 20
20 #define IOC_OSC_SET_ACTIVE _IOWR(IOC_OSC_TYPE, 21, struct obd_device *)
21 #define IOC_OSC_MAX_NR 50
23 #define IOC_MDC_TYPE 'i'
24 #define IOC_MDC_MIN_NR 20
25 /* Moved to lustre_user.h
26 #define IOC_MDC_LOOKUP _IOWR(IOC_MDC_TYPE, 20, struct obd_ioctl_data *)
27 #define IOC_MDC_GETSTRIPE _IOWR(IOC_MDC_TYPE, 21, struct lov_mds_md *) */
28 #define IOC_MDC_MAX_NR 50
30 #include <lustre/lustre_idl.h>
31 #include <lu_object.h>
32 #include <lustre_lib.h>
33 #include <lustre_export.h>
34 #include <lustre_quota.h>
35 #include <lustre_fld.h>
36 #include <lustre_capa.h>
37 #include <class_hash.h>
39 #include <libcfs/bitmap.h>
42 #define MAX_OBD_DEVICES 8192
44 /* this is really local to the OSC */
45 struct loi_oap_pages {
46 struct list_head lop_pending;
47 struct list_head lop_urgent;
48 struct list_head lop_pending_group;
58 struct lov_oinfo { /* per-stripe data structure */
59 __u64 loi_id; /* object ID on the target OST */
60 __u64 loi_gr; /* object group on the target OST */
61 int loi_ost_idx; /* OST stripe index in lov_tgt_desc->tgts */
62 int loi_ost_gen; /* generation of this loi_ost_idx */
64 /* used by the osc to keep track of what objects to build into rpcs */
65 struct loi_oap_pages loi_read_lop;
66 struct loi_oap_pages loi_write_lop;
67 /* _cli_ is poorly named, it should be _ready_ */
68 struct list_head loi_cli_item;
69 struct list_head loi_write_item;
70 struct list_head loi_read_item;
72 unsigned long loi_kms_valid:1;
73 __u64 loi_kms; /* known minimum size */
74 struct ost_lvb loi_lvb;
75 struct osc_async_rc loi_ar;
78 static inline void loi_init(struct lov_oinfo *loi)
80 CFS_INIT_LIST_HEAD(&loi->loi_read_lop.lop_pending);
81 CFS_INIT_LIST_HEAD(&loi->loi_read_lop.lop_urgent);
82 CFS_INIT_LIST_HEAD(&loi->loi_read_lop.lop_pending_group);
83 CFS_INIT_LIST_HEAD(&loi->loi_write_lop.lop_pending);
84 CFS_INIT_LIST_HEAD(&loi->loi_write_lop.lop_urgent);
85 CFS_INIT_LIST_HEAD(&loi->loi_write_lop.lop_pending_group);
86 CFS_INIT_LIST_HEAD(&loi->loi_cli_item);
87 CFS_INIT_LIST_HEAD(&loi->loi_write_item);
88 CFS_INIT_LIST_HEAD(&loi->loi_read_item);
91 /*extent array item for describing the joined file extent info*/
93 __u64 le_start; /* extent start */
94 __u64 le_len; /* extent length */
95 int le_loi_idx; /* extent #1 loi's index in lsm loi array */
96 int le_stripe_count; /* extent stripe count*/
99 /*Lov array info for describing joined file array EA info*/
100 struct lov_array_info {
101 struct llog_logid lai_array_id; /* MDS med llog object id */
102 unsigned lai_ext_count; /* number of extent count */
103 struct lov_extent *lai_ext_array; /* extent desc array */
106 struct lov_stripe_md {
108 pid_t lsm_lock_owner; /* debugging */
111 /* Public members. */
112 __u64 lw_object_id; /* lov object id */
113 __u64 lw_object_gr; /* lov object group */
114 __u64 lw_maxbytes; /* maximum possible file size */
116 /* LOV-private members start here -- only for use in lov/. */
118 __u32 lw_stripe_size; /* size of the stripe */
119 __u32 lw_pattern; /* striping pattern (RAID0, RAID1) */
120 unsigned lw_stripe_count; /* number of objects being striped over */
123 struct lov_array_info *lsm_array; /*Only for joined file array info*/
124 struct lov_oinfo *lsm_oinfo[0];
127 #define lsm_object_id lsm_wire.lw_object_id
128 #define lsm_object_gr lsm_wire.lw_object_gr
129 #define lsm_maxbytes lsm_wire.lw_maxbytes
130 #define lsm_magic lsm_wire.lw_magic
131 #define lsm_stripe_size lsm_wire.lw_stripe_size
132 #define lsm_pattern lsm_wire.lw_pattern
133 #define lsm_stripe_count lsm_wire.lw_stripe_count
137 typedef int (*obd_enqueue_update_f)(struct obd_info *oinfo, int rc);
139 /* obd info for a particular level (lov, osc). */
141 /* Lock policy. It keeps an extent which is specific for a particular
142 * OSC. (e.g. lov_prep_enqueue_set initialises extent of the policy,
143 * and osc_enqueue passes it into ldlm_lock_match & ldlm_cli_enqueue. */
144 ldlm_policy_data_t oi_policy;
145 /* Flags used for set request specific flags:
146 - while lock handling, the flags obtained on the enqueue
147 request are set here.
148 - while stats, the flags used for control delay/resend.
151 /* Lock handle specific for every OSC lock. */
152 struct lustre_handle *oi_lockh;
153 /* lsm data specific for every OSC. */
154 struct lov_stripe_md *oi_md;
155 /* obdo data specific for every OSC, if needed at all. */
157 /* statfs data specific for every OSC, if needed at all. */
158 struct obd_statfs *oi_osfs;
159 /* An update callback which is called to update some data on upper
160 * level. E.g. it is used for update lsm->lsm_oinfo at every recieved
161 * request in osc level for enqueue requests. It is also possible to
162 * update some caller data from LOV layer if needed. */
163 obd_enqueue_update_f oi_cb_up;
164 /* oss capability, its type is obd_capa in client to avoid copy.
165 * in contrary its type is lustre_capa in OSS. */
169 /* compare all relevant fields. */
170 static inline int lov_stripe_md_cmp(struct lov_stripe_md *m1,
171 struct lov_stripe_md *m2)
174 * ->lsm_wire contains padding, but it should be zeroed out during
177 return memcmp(&m1->lsm_wire, &m2->lsm_wire, sizeof m1->lsm_wire);
180 void lov_stripe_lock(struct lov_stripe_md *md);
181 void lov_stripe_unlock(struct lov_stripe_md *md);
184 struct list_head typ_chain;
185 struct obd_ops *typ_dt_ops;
186 struct md_ops *typ_md_ops;
187 cfs_proc_dir_entry_t *typ_procroot;
190 struct lu_device_type *typ_lu;
191 spinlock_t obd_type_lock;
202 ASYNC_READY = 0x1, /* ap_make_ready will not be called before this
203 page is added to an rpc */
204 ASYNC_URGENT = 0x2, /* page must be put into an RPC before return */
205 ASYNC_COUNT_STABLE = 0x4, /* ap_refresh_count will not be called
206 to give the caller a chance to update
207 or cancel the size of the io */
208 ASYNC_GROUP_SYNC = 0x8, /* ap_completion will not be called, instead
209 the page is accounted for in the
210 obd_io_group given to
211 obd_queue_group_io */
214 struct obd_async_page_ops {
215 int (*ap_make_ready)(void *data, int cmd);
216 int (*ap_refresh_count)(void *data, int cmd);
217 void (*ap_fill_obdo)(void *data, int cmd, struct obdo *oa);
218 void (*ap_update_obdo)(void *data, int cmd, struct obdo *oa,
220 int (*ap_completion)(void *data, int cmd, struct obdo *oa, int rc);
221 struct obd_capa *(*ap_lookup_capa)(void *data, int cmd);
224 /* the `oig' is passed down from a caller of obd rw methods. the callee
225 * records enough state such that the caller can sleep on the oig and
226 * be woken when all the callees have finished their work */
227 struct obd_io_group {
229 atomic_t oig_refcount;
232 struct list_head oig_occ_list;
233 cfs_waitq_t oig_waitq;
236 /* the oig callback context lets the callee of obd rw methods register
237 * for callbacks from the caller. */
238 struct oig_callback_context {
239 struct list_head occ_oig_item;
240 /* called when the caller has received a signal while sleeping.
241 * callees of this method are encouraged to abort their state
242 * in the oig. This may be called multiple times. */
243 void (*occ_interrupted)(struct oig_callback_context *occ);
244 unsigned long interrupted:1;
247 /* Individual type definitions */
249 struct ost_server_data;
251 /* hold common fields for "target" device */
252 struct obd_device_target {
253 struct super_block *obt_sb;
254 atomic_t obt_quotachecking;
255 struct lustre_quota_ctxt obt_qctxt;
258 typedef void (*obd_pin_extent_cb)(void *data);
259 typedef int (*obd_page_removal_cb_t)(void *data, int discard);
260 typedef int (*obd_lock_cancel_cb)(struct ldlm_lock *,struct ldlm_lock_desc *,
265 LLOG_CONFIG_ORIG_CTXT = 0,
266 LLOG_CONFIG_REPL_CTXT,
267 LLOG_MDS_OST_ORIG_CTXT,
268 LLOG_MDS_OST_REPL_CTXT,
275 LLOG_LOVEA_ORIG_CTXT,
276 LLOG_LOVEA_REPL_CTXT,
280 #define FILTER_SUBDIR_COUNT 32 /* set to zero for no subdirs */
282 struct filter_subdirs {
283 cfs_dentry_t *dentry[FILTER_SUBDIR_COUNT];
293 /* NB this field MUST be first */
294 struct obd_device_target fo_obt;
295 const char *fo_fstype;
296 struct vfsmount *fo_vfsmnt;
299 cfs_dentry_t *fo_dentry_O;
300 cfs_dentry_t **fo_dentry_O_groups;
301 struct filter_subdirs *fo_dentry_O_sub;
302 struct semaphore fo_init_lock; /* group initialization lock */
303 int fo_committed_group;
306 spinlock_t fo_objidlock; /* protect fo_lastobjid */
307 spinlock_t fo_translock; /* protect fsd_last_transno */
308 struct file *fo_rcvd_filp;
309 struct file *fo_health_check_filp;
310 struct lr_server_data *fo_fsd;
311 unsigned long *fo_last_rcvd_slots;
312 __u64 fo_mount_count;
314 unsigned long fo_destroys_in_progress;
315 struct semaphore fo_create_locks[FILTER_SUBDIR_COUNT];
317 struct list_head fo_export_list;
320 obd_size fo_tot_dirty; /* protected by obd_osfs_lock */
321 obd_size fo_tot_granted; /* all values in bytes */
322 obd_size fo_tot_pending;
324 obd_size fo_readcache_max_filesize;
326 struct obd_import *fo_mdc_imp;
327 struct obd_uuid fo_mdc_uuid;
328 struct lustre_handle fo_mdc_conn;
329 struct file **fo_last_objid_files;
330 __u64 *fo_last_objids; /* last created objid for groups,
331 * protected by fo_objidlock */
333 struct semaphore fo_alloc_lock;
335 atomic_t fo_r_in_flight;
336 atomic_t fo_w_in_flight;
339 * per-filter pool of kiobuf's allocated by filter_common_setup() and
340 * torn down by filter_cleanup(). Contains OST_NUM_THREADS elements of
341 * which ->fo_iobuf_count were allocated.
343 * This pool contains kiobuf used by
344 * filter_{prep,commit}rw_{read,write}() and is shared by all OST
347 * Locking: none, each OST thread uses only one element, determined by
348 * its "ordinal number", ->t_id.
350 struct filter_iobuf **fo_iobuf_pool;
353 struct list_head fo_llog_list;
354 spinlock_t fo_llog_list_lock;
356 struct brw_stats fo_filter_stats;
357 struct lustre_quota_ctxt fo_quota_ctxt;
358 spinlock_t fo_quotacheck_lock;
359 atomic_t fo_quotachecking;
361 int fo_fmd_max_num; /* per exp filter_mod_data */
362 int fo_fmd_max_age; /* jiffies to fmd expiry */
365 rwlock_t fo_sptlrpc_lock;
366 struct sptlrpc_rule_set fo_sptlrpc_rset;
368 /* capability related */
369 unsigned int fo_fl_oss_capa;
370 struct list_head fo_capa_keys;
371 struct hlist_head *fo_capa_hash;
376 #define OSC_MAX_RIF_DEFAULT 8
377 #define OSC_MAX_RIF_MAX 256
378 #define OSC_MAX_DIRTY_DEFAULT (OSC_MAX_RIF_DEFAULT * 4)
379 #define OSC_MAX_DIRTY_MB_MAX 2048 /* arbitrary, but < MAX_LONG bytes */
380 #define OSC_DEFAULT_RESENDS 10
382 #define MDC_MAX_RIF_DEFAULT 8
383 #define MDC_MAX_RIF_MAX 512
389 struct rw_semaphore cl_sem;
390 struct obd_uuid cl_target_uuid;
391 struct obd_import *cl_import; /* ptlrpc connection state */
393 /* max_mds_easize is purely a performance thing so we don't have to
394 * call obd_size_diskmd() all the time. */
395 int cl_default_mds_easize;
396 int cl_max_mds_easize;
397 int cl_max_mds_cookiesize;
399 /* security configuration */
400 struct sptlrpc_rule_set cl_sptlrpc_rset;
401 enum lustre_sec_part cl_sec_part;
403 //struct llog_canceld_ctxt *cl_llcd; /* it's included by obd_llog_ctxt */
404 void *cl_llcd_offset;
406 /* the grant values are protected by loi_list_lock below */
407 long cl_dirty; /* all _dirty_ in bytes */
408 long cl_dirty_max; /* allowed w/o rpc */
409 long cl_avail_grant; /* bytes of credit for ost */
410 long cl_lost_grant; /* lost credits (trunc) */
411 struct list_head cl_cache_waiters; /* waiting for cache/grant */
413 /* keep track of objects that have lois that contain pages which
414 * have been queued for async brw. this lock also protects the
415 * lists of osc_client_pages that hang off of the loi */
417 * ->cl_loi_list_lock protects consistency of
418 * ->cl_loi_{ready,read,write}_list. ->ap_make_ready() and
419 * ->ap_completion() call-backs are executed under this lock. As we
420 * cannot guarantee that these call-backs never block on all platforms
421 * (as a matter of fact they do block on Mac OS X), type of
422 * ->cl_loi_list_lock is platform dependent: it's a spin-lock on Linux
423 * and blocking mutex on Mac OS X. (Alternative is to make this lock
424 * blocking everywhere, but we don't want to slow down fast-path of
425 * our main platform.)
427 * Exact type of ->cl_loi_list_lock is defined in arch/obd.h together
428 * with client_obd_list_{un,}lock() and
429 * client_obd_list_lock_{init,done}() functions.
431 client_obd_lock_t cl_loi_list_lock;
432 struct list_head cl_loi_ready_list;
433 struct list_head cl_loi_write_list;
434 struct list_head cl_loi_read_list;
437 /* just a sum of the loi/lop pending numbers to be exported by /proc */
438 int cl_pending_w_pages;
439 int cl_pending_r_pages;
440 int cl_max_pages_per_rpc;
441 int cl_max_rpcs_in_flight;
442 struct obd_histogram cl_read_rpc_hist;
443 struct obd_histogram cl_write_rpc_hist;
444 struct obd_histogram cl_read_page_hist;
445 struct obd_histogram cl_write_page_hist;
446 struct obd_histogram cl_read_offset_hist;
447 struct obd_histogram cl_write_offset_hist;
449 /* number of in flight destroy rpcs is limited to max_rpcs_in_flight */
450 atomic_t cl_destroy_in_flight;
451 cfs_waitq_t cl_destroy_waitq;
453 struct mdc_rpc_lock *cl_rpc_lock;
454 struct mdc_rpc_lock *cl_setattr_lock;
455 struct mdc_rpc_lock *cl_close_lock;
456 struct osc_creator cl_oscc;
459 struct semaphore cl_mgc_sem;
460 struct vfsmount *cl_mgc_vfsmnt;
461 struct dentry *cl_mgc_configs_dir;
462 atomic_t cl_mgc_refcount;
463 struct obd_export *cl_mgc_mgsexp;
465 /* checksumming for data sent over the network */
466 unsigned int cl_checksum:1; /* 0 = disabled, 1 = enabled */
467 /* supported checksum types that are worked out at connect time */
468 __u32 cl_supp_cksum_types;
469 /* checksum algorithm to be used */
470 cksum_type_t cl_cksum_type;
472 /* also protected by the poorly named _loi_list_lock lock above */
473 struct osc_async_rc cl_ar;
475 /* used by quotacheck */
476 int cl_qchk_stat; /* quotacheck stat of the peer */
478 /* sequence manager */
479 struct lu_client_seq *cl_seq;
481 atomic_t cl_resends; /* resend count */
483 /* Cache of triples */
484 struct lustre_cache *cl_cache;
485 obd_lock_cancel_cb cl_ext_lock_cancel_cb;
487 #define obd2cli_tgt(obd) ((char *)(obd)->u.cli.cl_target_uuid.uuid)
489 #define CL_NOT_QUOTACHECKED 1 /* client->cl_qchk_stat init value */
492 struct ptlrpc_service *mgs_service;
493 struct vfsmount *mgs_vfsmnt;
494 struct super_block *mgs_sb;
495 struct dentry *mgs_configs_dir;
496 struct dentry *mgs_fid_de;
497 struct list_head mgs_fs_db_list;
498 struct semaphore mgs_sem;
499 cfs_proc_dir_entry_t *mgs_proc_live;
503 /* NB this field MUST be first */
504 struct obd_device_target mds_obt;
505 struct ptlrpc_service *mds_service;
506 struct ptlrpc_service *mds_setattr_service;
507 struct ptlrpc_service *mds_readpage_service;
508 struct vfsmount *mds_vfsmnt;
509 cfs_dentry_t *mds_fid_de;
511 int mds_max_cookiesize;
512 struct file *mds_rcvd_filp;
513 spinlock_t mds_transno_lock;
514 __u64 mds_last_transno;
515 __u64 mds_mount_count;
517 unsigned long mds_atime_diff;
518 struct semaphore mds_epoch_sem;
519 struct ll_fid mds_rootfid;
520 struct lr_server_data *mds_server_data;
521 cfs_dentry_t *mds_pending_dir;
522 cfs_dentry_t *mds_logs_dir;
523 cfs_dentry_t *mds_objects_dir;
524 struct llog_handle *mds_cfg_llh;
525 struct obd_device *mds_osc_obd; /* XXX lov_obd */
526 struct obd_uuid mds_lov_uuid;
528 struct obd_export *mds_osc_exp; /* XXX lov_exp */
529 struct lov_desc mds_lov_desc;
532 /* mark pages dirty for write. */
533 bitmap_t *mds_lov_page_dirty;
534 /* array for store pages with obd_id */
535 void **mds_lov_page_array;
536 /* file for store objid */
537 struct file *mds_lov_objid_filp;
538 __u32 mds_lov_objid_count;
539 __u32 mds_lov_objid_lastpage;
540 __u32 mds_lov_objid_lastidx;
542 struct file *mds_health_check_filp;
543 unsigned long *mds_client_bitmap;
544 // struct upcall_cache *mds_group_hash;
546 struct lustre_quota_info mds_quota_info;
547 struct semaphore mds_qonoff_sem;
548 struct semaphore mds_health_sem;
549 unsigned long mds_fl_user_xattr:1,
551 mds_evict_ost_nids:1,
555 struct upcall_cache *mds_identity_cache;
557 /* for capability keys update */
558 struct lustre_capa_key *mds_capa_keys;
559 struct rw_semaphore mds_notify_lock;
563 extern __u32 mds_max_ost_index;
565 #define MDS_LOV_ALLOC_SIZE (CFS_PAGE_SIZE)
567 #define OBJID_PER_PAGE() (MDS_LOV_ALLOC_SIZE / sizeof(obd_id))
569 #define MDS_LOV_OBJID_PAGES_COUNT (mds_max_ost_index/OBJID_PER_PAGE())
571 extern int mds_lov_init_objids(struct obd_device *obd);
572 extern void mds_lov_destroy_objids(struct obd_device *obd);
585 struct lustre_handle eo_nl_lock;
590 struct ptlrpc_service *ost_service;
591 struct ptlrpc_service *ost_create_service;
592 struct ptlrpc_service *ost_io_service;
593 struct semaphore ost_health_sem;
596 struct echo_client_obd {
597 struct obd_export *ec_exp; /* the local connection to osc/lov */
599 struct list_head ec_objects;
605 struct obd_uuid lqo_uuid; /* ptlrpc's c_remote_uuid */
606 struct list_head lqo_oss_list; /* link to lov_qos */
607 __u32 lqo_ost_count; /* number of osts on this oss */
608 __u64 lqo_bavail; /* total bytes avail on OSS */
609 __u64 lqo_penalty; /* current penalty */
610 __u64 lqo_penalty_per_obj; /* penalty decrease every obj*/
614 struct lov_qos_oss *ltq_oss; /* oss info */
615 __u64 ltq_penalty; /* current penalty */
616 __u64 ltq_penalty_per_obj; /* penalty decrease every obj*/
617 __u64 ltq_weight; /* net weighting */
618 unsigned int ltq_usable:1; /* usable for striping */
622 struct list_head lq_oss_list; /* list of OSSs that targets use */
623 struct rw_semaphore lq_rw_sem;
624 __u32 lq_active_oss_count;
625 __u32 *lq_rr_array; /* round-robin optimized list */
626 unsigned int lq_rr_size; /* rr array size */
627 unsigned int lq_prio_free; /* priority for free space */
628 unsigned long lq_dirty:1, /* recalc qos data */
629 lq_dirty_rr:1, /* recalc round-robin list */
630 lq_same_space:1,/* the ost's all have approx.
631 the same space avail */
632 lq_reset:1; /* zero current penalties */
635 struct lov_tgt_desc {
636 struct obd_uuid ltd_uuid;
637 struct obd_export *ltd_exp;
638 struct ltd_qos ltd_qos; /* qos info per target */
640 __u32 ltd_index; /* index in lov_obd->tgts */
641 unsigned long ltd_active:1,/* is this target up for requests */
642 ltd_activate:1,/* should this target be activated */
643 ltd_reap:1; /* should this target be deleted */
647 struct lov_desc desc;
648 struct lov_tgt_desc **lov_tgts;
649 struct semaphore lov_lock;
650 struct obd_connect_data lov_ocd;
651 struct lov_qos lov_qos; /* qos info per lov */
652 atomic_t lov_refcount;
653 __u32 lov_tgt_count; /* how many OBD's */
654 __u32 lov_active_tgt_count; /* how many active */
655 __u32 lov_death_row;/* tgts scheduled to be deleted */
656 __u32 lov_tgt_size; /* size of tgts array */
657 __u32 lov_start_idx; /* start index of new inode */
658 __u32 lov_offset_idx; /* aliasing for start_idx */
659 int lov_start_count;/* reseed counter */
661 obd_page_removal_cb_t lov_page_removal_cb;
662 obd_pin_extent_cb lov_page_pin_cb;
663 obd_lock_cancel_cb lov_lock_cancel_cb;
666 struct lmv_tgt_desc {
667 struct obd_uuid ltd_uuid;
668 struct obd_export *ltd_exp;
669 int ltd_active; /* is this target up for requests */
671 struct semaphore ltd_fid_sem;
676 struct lu_client_fld lmv_fld;
678 struct lmv_desc desc;
679 struct obd_uuid cluuid;
680 struct obd_export *exp;
687 struct semaphore init_sem;
689 struct lmv_tgt_desc *tgts;
692 struct obd_connect_data *datas;
695 struct obd_connect_data conn_data;
698 struct niobuf_local {
703 cfs_dentry_t *dentry;
708 #define LUSTRE_FLD_NAME "fld"
709 #define LUSTRE_SEQ_NAME "seq"
711 #define LUSTRE_CMM_NAME "cmm"
712 #define LUSTRE_MDD_NAME "mdd"
713 #define LUSTRE_OSD_NAME "osd"
714 #define LUSTRE_LMV_NAME "lmv"
715 #define LUSTRE_CMM_MDC_NAME "cmm-mdc"
717 /* obd device type names */
718 /* FIXME all the references to LUSTRE_MDS_NAME should be swapped with LUSTRE_MDT_NAME */
719 #define LUSTRE_MDS_NAME "mds"
720 #define LUSTRE_MDT_NAME "mdt"
721 #define LUSTRE_MDC_NAME "mdc"
722 #define LUSTRE_OSS_NAME "ost" /* FIXME change name to oss */
723 #define LUSTRE_OST_NAME "obdfilter" /* FIXME change name to ost */
724 #define LUSTRE_OSC_NAME "osc"
725 #define LUSTRE_LOV_NAME "lov"
726 #define LUSTRE_MGS_NAME "mgs"
727 #define LUSTRE_MGC_NAME "mgc"
729 #define LUSTRE_CACHEOBD_NAME "cobd"
730 #define LUSTRE_ECHO_NAME "obdecho"
731 #define LUSTRE_ECHO_CLIENT_NAME "echo_client"
733 /* Constant obd names (post-rename) */
734 #define LUSTRE_MDS_OBDNAME "MDS"
735 #define LUSTRE_OSS_OBDNAME "OSS"
736 #define LUSTRE_MGS_OBDNAME "MGS"
737 #define LUSTRE_MGC_OBDNAME "MGC"
739 /* Don't conflict with on-wire flags OBD_BRW_WRITE, etc */
740 #define N_LOCAL_TEMP_PAGE 0x10000000
742 struct obd_trans_info {
745 /* Only used on the server side for tracking acks. */
746 struct oti_req_ack_lock {
747 struct lustre_handle lock;
751 struct llog_cookie oti_onecookie;
752 struct llog_cookie *oti_logcookies;
755 /* initial thread handling transaction */
760 static inline void oti_init(struct obd_trans_info *oti,
761 struct ptlrpc_request *req)
765 memset(oti, 0, sizeof(*oti));
770 oti->oti_xid = req->rq_xid;
772 if (req->rq_repmsg != NULL)
773 oti->oti_transno = lustre_msg_get_transno(req->rq_repmsg);
774 oti->oti_thread_id = req->rq_svc_thread ? req->rq_svc_thread->t_id : -1;
775 if (req->rq_reqmsg != NULL)
776 oti->oti_conn_cnt = lustre_msg_get_conn_cnt(req->rq_reqmsg);
779 static inline void oti_alloc_cookies(struct obd_trans_info *oti,int num_cookies)
784 if (num_cookies == 1)
785 oti->oti_logcookies = &oti->oti_onecookie;
787 OBD_ALLOC(oti->oti_logcookies,
788 num_cookies * sizeof(oti->oti_onecookie));
790 oti->oti_numcookies = num_cookies;
793 static inline void oti_free_cookies(struct obd_trans_info *oti)
795 if (!oti || !oti->oti_logcookies)
798 if (oti->oti_logcookies == &oti->oti_onecookie)
799 LASSERT(oti->oti_numcookies == 1);
801 OBD_FREE(oti->oti_logcookies,
802 oti->oti_numcookies * sizeof(oti->oti_onecookie));
803 oti->oti_logcookies = NULL;
804 oti->oti_numcookies = 0;
808 * Events signalled through obd_notify() upcall-chain.
810 enum obd_notify_event {
811 /* Device activated */
813 /* Device deactivated */
815 /* Device disconnected */
817 /* Connect data for import were changed */
820 OBD_NOTIFY_SYNC_NONBLOCK,
822 /* Configuration event */
826 /* bit-mask flags for config events */
828 CONFIG_LOG = 0x1, /* finished processing config log */
829 CONFIG_SYNC = 0x2 /* mdt synced 1 ost */
833 * Data structure used to pass obd_notify()-event to non-obd listeners (llite
834 * and liblustre being main examples).
836 struct obd_notify_upcall {
837 int (*onu_upcall)(struct obd_device *host, struct obd_device *watched,
838 enum obd_notify_event ev, void *owner);
839 /* Opaque datum supplied by upper layer listener */
843 struct target_recovery_data {
844 svc_handler_t trd_recovery_handler;
845 pid_t trd_processing_task;
846 struct completion trd_starting;
847 struct completion trd_finishing;
850 #define OBD_LLOG_GROUP 0
853 FILTER_GROUP_LLOG = 1,
858 struct obd_llog_group {
859 struct list_head olg_list;
861 struct llog_ctxt *olg_ctxts[LLOG_MAX_CTXTS];
862 cfs_waitq_t olg_waitq;
864 struct obd_export *olg_exp;
865 int olg_initializing;
868 /* corresponds to one of the obd's */
869 #define MAX_OBD_NAME 128
870 #define OBD_DEVICE_MAGIC 0XAB5CD6EF
872 struct obd_type *obd_type;
875 /* common and UUID name of this device */
876 char obd_name[MAX_OBD_NAME];
877 struct obd_uuid obd_uuid;
879 struct lu_device *obd_lu_dev;
882 unsigned long obd_attached:1, /* finished attach */
883 obd_set_up:1, /* finished setup */
884 obd_recovering:1, /* there are recoverable clients */
885 obd_abort_recovery:1,/* somebody ioctl'ed us to abort */
886 obd_replayable:1, /* recovery is enabled; inform clients */
887 obd_no_transno:1, /* no committed-transno notification */
888 obd_no_recov:1, /* fail instead of retry messages */
889 obd_req_replaying:1, /* replaying requests */
890 obd_stopping:1, /* started cleanup */
891 obd_starting:1, /* started setup */
892 obd_force:1, /* cleanup with > 0 obd refcount */
893 obd_fail:1, /* cleanup with failover */
894 obd_async_recov:1, /* allow asyncronous orphan cleanup */
895 obd_no_conn:1, /* deny new connections */
896 obd_inactive:1; /* device active/inactive
897 * (for /proc/status only!!) */
898 /* uuid-export hash body */
899 struct lustre_class_hash_body *obd_uuid_hash_body;
900 /* nid-export hash body */
901 struct lustre_class_hash_body *obd_nid_hash_body;
903 struct lustre_class_hash_body *obd_nid_stats_hash_body;
904 struct list_head obd_nid_stats;
905 atomic_t obd_refcount;
906 cfs_waitq_t obd_refcount_waitq;
907 struct list_head obd_exports;
909 spinlock_t obd_nid_lock;
910 struct ldlm_namespace *obd_namespace;
911 struct ptlrpc_client obd_ldlm_client; /* XXX OST/MDS only */
912 /* a spinlock is OK for what we do now, may need a semaphore later */
913 spinlock_t obd_dev_lock;
914 struct semaphore obd_dev_sem;
915 __u64 obd_last_committed;
916 struct fsfilt_operations *obd_fsops;
917 spinlock_t obd_osfs_lock;
918 struct obd_statfs obd_osfs; /* locked by obd_osfs_lock */
920 struct lvfs_run_ctxt obd_lvfs_ctxt;
921 struct obd_llog_group obd_olg; /* default llog group */
922 struct obd_device *obd_observer;
923 struct obd_notify_upcall obd_upcall;
924 struct obd_export *obd_self_export;
925 /* list of exports in LRU order, for ping evictor, with obd_dev_lock */
926 struct list_head obd_exports_timed;
927 time_t obd_eviction_timer; /* for ping evictor */
929 /* XXX encapsulate all this recovery data into one struct */
930 svc_handler_t obd_recovery_handler;
931 pid_t obd_processing_task;
933 int obd_max_recoverable_clients;
934 int obd_connected_clients;
935 int obd_recoverable_clients;
936 spinlock_t obd_processing_task_lock; /* BH lock (timer) */
937 __u64 obd_next_recovery_transno;
938 int obd_replayed_requests;
939 int obd_requests_queued_for_recovery;
940 cfs_waitq_t obd_next_transno_waitq;
941 struct list_head obd_uncommitted_replies;
942 spinlock_t obd_uncommitted_replies_lock;
943 cfs_timer_t obd_recovery_timer;
944 time_t obd_recovery_start; /* seconds */
945 time_t obd_recovery_end; /* seconds, for lprocfs_status */
946 time_t obd_recovery_max_time; /* seconds, bz13079 */
947 int obd_recovery_timeout;
949 /* new recovery stuff from CMD2 */
950 struct target_recovery_data obd_recovery_data;
951 int obd_replayed_locks;
952 atomic_t obd_req_replay_clients;
953 atomic_t obd_lock_replay_clients;
954 struct list_head obd_req_replay_queue;
955 struct list_head obd_lock_replay_queue;
956 struct list_head obd_final_req_queue;
957 int obd_recovery_stage;
960 struct obd_device_target obt;
961 struct filter_obd filter;
963 struct client_obd cli;
965 struct echo_client_obd echo_client;
966 struct echo_obd echo;
971 /* Fields used by LProcFS */
972 unsigned int obd_cntr_base;
973 struct lprocfs_stats *obd_stats;
975 unsigned int md_cntr_base;
976 struct lprocfs_stats *md_stats;
978 cfs_proc_dir_entry_t *obd_proc_entry;
979 cfs_proc_dir_entry_t *obd_proc_exports_entry;
980 cfs_proc_dir_entry_t *obd_svc_procroot;
981 struct lprocfs_stats *obd_svc_stats;
982 atomic_t obd_evict_inprogress;
983 cfs_waitq_t obd_evict_inprogress_waitq;
986 * Ldlm pool part. Save last calculated SLV and Limit.
988 rwlock_t obd_pool_lock;
993 #define OBD_OPT_FORCE 0x0001
994 #define OBD_OPT_FAILOVER 0x0002
996 #define OBD_LLOG_FL_SENDNOW 0x0001
998 enum obd_cleanup_stage {
999 /* Special case hack for MDS LOVs */
1001 /* can be directly mapped to .ldto_device_fini() */
1002 OBD_CLEANUP_EXPORTS,
1005 /* get/set_info keys */
1006 #define KEY_READ_ONLY "read-only"
1007 #define KEY_MDS_CONN "mds_conn"
1008 #define KEY_NEXT_ID "next_id"
1009 #define KEY_LOVDESC "lovdesc"
1010 #define KEY_INIT_RECOV "initial_recov"
1011 #define KEY_INIT_RECOV_BACKUP "init_recov_bk"
1012 #define KEY_FLUSH_CTX "flush_ctx"
1013 #define KEY_CAPA_KEY "capa_key"
1014 #define KEY_CONN_DATA "conn_data"
1015 #define KEY_MAX_EASIZE "max_easize"
1016 #define KEY_REVIMP_UPD "revimp_update"
1017 #define KEY_LOV_IDX "lov_idx"
1018 #define KEY_LAST_ID "last_id"
1019 #define KEY_READONLY "read-only"
1020 #define KEY_LOCK_TO_STRIPE "lock_to_stripe"
1021 #define KEY_CHECKSUM "checksum"
1022 #define KEY_UNLINKED "unlinked"
1023 #define KEY_EVICT_BY_NID "evict_by_nid"
1024 #define KEY_REGISTER_TARGET "register_target"
1025 #define KEY_SET_FS "set_fs"
1026 #define KEY_CLEAR_FS "clear_fs"
1027 #define KEY_BLOCKSIZE "blocksize"
1028 #define KEY_BLOCKSIZE_BITS "blocksize_bits"
1030 #define KEY_INTERMDS "inter_mds"
1031 #define KEY_ASYNC "async"
1035 static inline int it_to_lock_mode(struct lookup_intent *it)
1037 /* CREAT needs to be tested before open (both could be set) */
1038 if (it->it_op & IT_CREAT)
1040 else if (it->it_op & (IT_READDIR | IT_GETATTR | IT_OPEN | IT_LOOKUP))
1043 LASSERTF(0, "Invalid it_op: %d\n", it->it_op);
1048 struct lu_fid op_fid1; /* operation fid1 (usualy parent) */
1049 struct lu_fid op_fid2; /* operation fid2 (usualy child) */
1050 struct lu_fid op_fid3; /* 2 extra fids to find conflicting */
1051 struct lu_fid op_fid4; /* to the operation locks. */
1052 mdsno_t op_mds; /* what mds server open will go to */
1053 struct lustre_handle op_handle;
1055 const char *op_name;
1058 struct lmv_stripe_md *op_mea1;
1059 struct lmv_stripe_md *op_mea2;
1060 __u32 op_suppgids[2];
1066 /* iattr fields and blocks. */
1067 struct iattr op_attr;
1069 #if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,14)
1070 unsigned int op_attr_flags;
1073 loff_t op_attr_blocks;
1075 /* Size-on-MDS epoch and flags. */
1080 struct obd_capa *op_capa1;
1081 struct obd_capa *op_capa2;
1083 /* Various operation flags. */
1086 /* Operation type */
1090 struct md_enqueue_info;
1091 /* metadata stat-ahead */
1092 typedef int (* md_enqueue_cb_t)(struct ptlrpc_request *req,
1093 struct md_enqueue_info *minfo,
1096 struct md_enqueue_info {
1097 struct md_op_data mi_data;
1098 struct lookup_intent mi_it;
1099 struct lustre_handle mi_lockh;
1100 struct dentry *mi_dentry;
1101 md_enqueue_cb_t mi_cb;
1102 unsigned int mi_generation;
1107 struct module *o_owner;
1108 int (*o_iocontrol)(unsigned int cmd, struct obd_export *exp, int len,
1109 void *karg, void *uarg);
1110 int (*o_get_info)(struct obd_export *, __u32 keylen, void *key,
1111 __u32 *vallen, void *val);
1112 int (*o_set_info_async)(struct obd_export *, __u32 keylen, void *key,
1113 __u32 vallen, void *val,
1114 struct ptlrpc_request_set *set);
1115 int (*o_attach)(struct obd_device *dev, obd_count len, void *data);
1116 int (*o_detach)(struct obd_device *dev);
1117 int (*o_setup) (struct obd_device *dev, struct lustre_cfg *cfg);
1118 int (*o_precleanup)(struct obd_device *dev,
1119 enum obd_cleanup_stage cleanup_stage);
1120 int (*o_cleanup)(struct obd_device *dev);
1121 int (*o_process_config)(struct obd_device *dev, obd_count len,
1123 int (*o_postrecov)(struct obd_device *dev);
1124 int (*o_add_conn)(struct obd_import *imp, struct obd_uuid *uuid,
1126 int (*o_del_conn)(struct obd_import *imp, struct obd_uuid *uuid);
1127 /* connect to the target device with given connection
1128 * data. @ocd->ocd_connect_flags is modified to reflect flags actually
1129 * granted by the target, which are guaranteed to be a subset of flags
1130 * asked for. If @ocd == NULL, use default parameters. */
1131 int (*o_connect)(const struct lu_env *env,
1132 struct lustre_handle *conn, struct obd_device *src,
1133 struct obd_uuid *cluuid, struct obd_connect_data *ocd,
1135 int (*o_reconnect)(const struct lu_env *env,
1136 struct obd_export *exp, struct obd_device *src,
1137 struct obd_uuid *cluuid,
1138 struct obd_connect_data *ocd);
1139 int (*o_disconnect)(struct obd_export *exp);
1141 /* Initialize/finalize fids infrastructure. */
1142 int (*o_fid_init)(struct obd_export *exp);
1143 int (*o_fid_fini)(struct obd_export *exp);
1145 /* Allocate new fid according to passed @hint. */
1146 int (*o_fid_alloc)(struct obd_export *exp, struct lu_fid *fid,
1147 struct md_op_data *op_data);
1150 * Object with @fid is getting deleted, we may want to do something
1153 int (*o_fid_delete)(struct obd_export *exp, const struct lu_fid *fid);
1155 int (*o_statfs)(struct obd_device *obd, struct obd_statfs *osfs,
1156 __u64 max_age, __u32 flags);
1157 int (*o_statfs_async)(struct obd_device *obd, struct obd_info *oinfo,
1158 __u64 max_age, struct ptlrpc_request_set *set);
1159 int (*o_packmd)(struct obd_export *exp, struct lov_mds_md **disk_tgt,
1160 struct lov_stripe_md *mem_src);
1161 int (*o_unpackmd)(struct obd_export *exp,struct lov_stripe_md **mem_tgt,
1162 struct lov_mds_md *disk_src, int disk_len);
1163 int (*o_checkmd)(struct obd_export *exp, struct obd_export *md_exp,
1164 struct lov_stripe_md *mem_tgt);
1165 int (*o_preallocate)(struct lustre_handle *, obd_count *req,
1167 /* FIXME: add fid capability support for create & destroy! */
1168 int (*o_precreate)(struct obd_export *exp);
1169 int (*o_create)(struct obd_export *exp, struct obdo *oa,
1170 struct lov_stripe_md **ea, struct obd_trans_info *oti);
1171 int (*o_destroy)(struct obd_export *exp, struct obdo *oa,
1172 struct lov_stripe_md *ea, struct obd_trans_info *oti,
1173 struct obd_export *md_exp);
1174 int (*o_setattr)(struct obd_export *exp, struct obd_info *oinfo,
1175 struct obd_trans_info *oti);
1176 int (*o_setattr_async)(struct obd_export *exp, struct obd_info *oinfo,
1177 struct obd_trans_info *oti,
1178 struct ptlrpc_request_set *rqset);
1179 int (*o_getattr)(struct obd_export *exp, struct obd_info *oinfo);
1180 int (*o_getattr_async)(struct obd_export *exp, struct obd_info *oinfo,
1181 struct ptlrpc_request_set *set);
1182 int (*o_brw)(int rw, struct obd_export *exp, struct obd_info *oinfo,
1183 obd_count oa_bufs, struct brw_page *pgarr,
1184 struct obd_trans_info *oti);
1185 int (*o_brw_async)(int rw, struct obd_export *exp,
1186 struct obd_info *oinfo, obd_count oa_bufs,
1187 struct brw_page *pgarr, struct obd_trans_info *oti,
1188 struct ptlrpc_request_set *);
1189 int (*o_prep_async_page)(struct obd_export *exp,
1190 struct lov_stripe_md *lsm,
1191 struct lov_oinfo *loi,
1192 cfs_page_t *page, obd_off offset,
1193 struct obd_async_page_ops *ops, void *data,
1194 void **res, int nocache,
1195 struct lustre_handle *lockh);
1196 int (*o_reget_short_lock)(struct obd_export *exp,
1197 struct lov_stripe_md *lsm,
1199 obd_off start, obd_off end,
1201 int (*o_release_short_lock)(struct obd_export *exp,
1202 struct lov_stripe_md *lsm, obd_off end,
1203 void *cookie, int rw);
1204 int (*o_queue_async_io)(struct obd_export *exp,
1205 struct lov_stripe_md *lsm,
1206 struct lov_oinfo *loi, void *cookie,
1207 int cmd, obd_off off, int count,
1208 obd_flag brw_flags, obd_flag async_flags);
1209 int (*o_queue_group_io)(struct obd_export *exp,
1210 struct lov_stripe_md *lsm,
1211 struct lov_oinfo *loi,
1212 struct obd_io_group *oig,
1213 void *cookie, int cmd, obd_off off, int count,
1214 obd_flag brw_flags, obd_flag async_flags);
1215 int (*o_trigger_group_io)(struct obd_export *exp,
1216 struct lov_stripe_md *lsm,
1217 struct lov_oinfo *loi,
1218 struct obd_io_group *oig);
1219 int (*o_set_async_flags)(struct obd_export *exp,
1220 struct lov_stripe_md *lsm,
1221 struct lov_oinfo *loi, void *cookie,
1222 obd_flag async_flags);
1223 int (*o_teardown_async_page)(struct obd_export *exp,
1224 struct lov_stripe_md *lsm,
1225 struct lov_oinfo *loi, void *cookie);
1226 int (*o_merge_lvb)(struct obd_export *exp, struct lov_stripe_md *lsm,
1227 struct ost_lvb *lvb, int kms_only);
1228 int (*o_adjust_kms)(struct obd_export *exp, struct lov_stripe_md *lsm,
1229 obd_off size, int shrink);
1230 int (*o_punch)(struct obd_export *exp, struct obd_info *oinfo,
1231 struct obd_trans_info *oti,
1232 struct ptlrpc_request_set *rqset);
1233 int (*o_sync)(struct obd_export *exp, struct obdo *oa,
1234 struct lov_stripe_md *ea, obd_size start, obd_size end,
1236 int (*o_migrate)(struct lustre_handle *conn, struct lov_stripe_md *dst,
1237 struct lov_stripe_md *src, obd_size start,
1238 obd_size end, struct obd_trans_info *oti);
1239 int (*o_copy)(struct lustre_handle *dstconn, struct lov_stripe_md *dst,
1240 struct lustre_handle *srconn, struct lov_stripe_md *src,
1241 obd_size start, obd_size end, struct obd_trans_info *);
1242 int (*o_iterate)(struct lustre_handle *conn,
1243 int (*)(obd_id, obd_gr, void *),
1244 obd_id *startid, obd_gr group, void *data);
1245 int (*o_preprw)(int cmd, struct obd_export *exp, struct obdo *oa,
1246 int objcount, struct obd_ioobj *obj,
1247 int niocount, struct niobuf_remote *remote,
1248 struct niobuf_local *local, struct obd_trans_info *oti,
1249 struct lustre_capa *capa);
1250 int (*o_commitrw)(int cmd, struct obd_export *exp, struct obdo *oa,
1251 int objcount, struct obd_ioobj *obj,
1252 int niocount, struct niobuf_local *local,
1253 struct obd_trans_info *oti, int rc);
1254 int (*o_enqueue)(struct obd_export *, struct obd_info *oinfo,
1255 struct ldlm_enqueue_info *einfo,
1256 struct ptlrpc_request_set *rqset);
1257 int (*o_match)(struct obd_export *, struct lov_stripe_md *, __u32 type,
1258 ldlm_policy_data_t *, __u32 mode, int *flags, void *data,
1259 struct lustre_handle *lockh);
1260 int (*o_change_cbdata)(struct obd_export *, struct lov_stripe_md *,
1261 ldlm_iterator_t it, void *data);
1262 int (*o_cancel)(struct obd_export *, struct lov_stripe_md *md,
1263 __u32 mode, struct lustre_handle *);
1264 int (*o_cancel_unused)(struct obd_export *, struct lov_stripe_md *,
1265 int flags, void *opaque);
1266 int (*o_join_lru)(struct obd_export *, struct lov_stripe_md *,
1268 int (*o_init_export)(struct obd_export *exp);
1269 int (*o_destroy_export)(struct obd_export *exp);
1270 int (*o_extent_calc)(struct obd_export *, struct lov_stripe_md *,
1271 int cmd, obd_off *);
1273 /* llog related obd_methods */
1274 int (*o_llog_init)(struct obd_device *obd, struct obd_llog_group *grp,
1275 struct obd_device *disk_obd, int count,
1276 struct llog_catid *logid, struct obd_uuid *uuid);
1277 int (*o_llog_finish)(struct obd_device *obd, int count);
1278 int (*o_llog_connect)(struct obd_export *, struct llogd_conn_body *);
1280 /* metadata-only methods */
1281 int (*o_pin)(struct obd_export *, const struct lu_fid *fid,
1282 struct obd_capa *, struct obd_client_handle *, int flag);
1283 int (*o_unpin)(struct obd_export *, struct obd_client_handle *, int);
1285 int (*o_import_event)(struct obd_device *, struct obd_import *,
1286 enum obd_import_event);
1288 int (*o_notify)(struct obd_device *obd, struct obd_device *watched,
1289 enum obd_notify_event ev, void *data);
1291 int (*o_health_check)(struct obd_device *);
1294 int (*o_quotacheck)(struct obd_export *, struct obd_quotactl *);
1295 int (*o_quotactl)(struct obd_export *, struct obd_quotactl *);
1297 int (*o_ping)(struct obd_export *exp);
1299 int (*o_register_page_removal_cb)(struct obd_export *exp,
1300 obd_page_removal_cb_t cb,
1301 obd_pin_extent_cb pin_cb);
1302 int (*o_unregister_page_removal_cb)(struct obd_export *exp,
1303 obd_page_removal_cb_t cb);
1304 int (*o_register_lock_cancel_cb)(struct obd_export *exp,
1305 obd_lock_cancel_cb cb);
1306 int (*o_unregister_lock_cancel_cb)(struct obd_export *exp,
1307 obd_lock_cancel_cb cb);
1310 * NOTE: If adding ops, add another LPROCFS_OBD_OP_INIT() line
1311 * to lprocfs_alloc_obd_stats() in obdclass/lprocfs_status.c.
1312 * Also, add a wrapper function in include/linux/obd_class.h. */
1315 /* TODO: lmv_stripe_md should contain mds capabilities for all slave fids */
1316 struct lmv_stripe_md {
1321 struct lu_fid mea_ids[0];
1325 LUSTRE_OPC_MKDIR = (1 << 0),
1326 LUSTRE_OPC_SYMLINK = (1 << 1),
1327 LUSTRE_OPC_MKNOD = (1 << 2),
1328 LUSTRE_OPC_CREATE = (1 << 3),
1329 LUSTRE_OPC_ANY = (1 << 4)
1332 /* lmv structures */
1333 #define MEA_MAGIC_LAST_CHAR 0xb2221ca1
1334 #define MEA_MAGIC_ALL_CHARS 0xb222a11c
1335 #define MEA_MAGIC_HASH_SEGMENT 0xb222a11b
1337 #define MAX_HASH_SIZE_32 0x7fffffffUL
1338 #define MAX_HASH_SIZE 0x7fffffffffffffffULL
1339 #define MAX_HASH_HIGHEST_BIT 0x1000000000000000
1342 struct mdt_body *body;
1343 struct lov_stripe_md *lsm;
1344 struct lmv_stripe_md *mea;
1345 #ifdef CONFIG_FS_POSIX_ACL
1346 struct posix_acl *posix_acl;
1348 struct mdt_remote_perm *remote_perm;
1349 struct obd_capa *mds_capa;
1350 struct obd_capa *oss_capa;
1353 struct md_open_data {
1354 struct obd_client_handle *mod_och;
1355 struct list_head mod_replay_list;
1358 struct lookup_intent;
1361 int (*m_getstatus)(struct obd_export *, struct lu_fid *,
1362 struct obd_capa **);
1363 int (*m_change_cbdata)(struct obd_export *, const struct lu_fid *,
1364 ldlm_iterator_t, void *);
1365 int (*m_close)(struct obd_export *, struct md_op_data *,
1366 struct md_open_data *, struct ptlrpc_request **);
1367 int (*m_create)(struct obd_export *, struct md_op_data *,
1368 const void *, int, int, __u32, __u32, __u32,
1369 __u64, struct ptlrpc_request **);
1370 int (*m_done_writing)(struct obd_export *, struct md_op_data *,
1371 struct md_open_data *);
1372 int (*m_enqueue)(struct obd_export *, struct ldlm_enqueue_info *,
1373 struct lookup_intent *, struct md_op_data *,
1374 struct lustre_handle *, void *, int, int);
1375 int (*m_getattr)(struct obd_export *, const struct lu_fid *,
1376 struct obd_capa *, obd_valid, int,
1377 struct ptlrpc_request **);
1378 int (*m_getattr_name)(struct obd_export *, const struct lu_fid *,
1379 struct obd_capa *, const char *, int, obd_valid,
1380 int, __u32, struct ptlrpc_request **);
1381 int (*m_intent_lock)(struct obd_export *, struct md_op_data *,
1382 void *, int, struct lookup_intent *, int,
1383 struct ptlrpc_request **,
1384 ldlm_blocking_callback, int);
1385 int (*m_link)(struct obd_export *, struct md_op_data *,
1386 struct ptlrpc_request **);
1387 int (*m_rename)(struct obd_export *, struct md_op_data *,
1388 const char *, int, const char *, int,
1389 struct ptlrpc_request **);
1390 int (*m_is_subdir)(struct obd_export *, const struct lu_fid *,
1391 const struct lu_fid *,
1392 struct ptlrpc_request **);
1393 int (*m_setattr)(struct obd_export *, struct md_op_data *, void *,
1394 int , void *, int, struct ptlrpc_request **,
1395 struct md_open_data **mod);
1396 int (*m_sync)(struct obd_export *, const struct lu_fid *,
1397 struct obd_capa *, struct ptlrpc_request **);
1398 int (*m_readpage)(struct obd_export *, const struct lu_fid *,
1399 struct obd_capa *, __u64, struct page *,
1400 struct ptlrpc_request **);
1402 int (*m_unlink)(struct obd_export *, struct md_op_data *,
1403 struct ptlrpc_request **);
1405 int (*m_setxattr)(struct obd_export *, const struct lu_fid *,
1406 struct obd_capa *, obd_valid, const char *,
1407 const char *, int, int, int, __u32,
1408 struct ptlrpc_request **);
1410 int (*m_getxattr)(struct obd_export *, const struct lu_fid *,
1411 struct obd_capa *, obd_valid, const char *,
1412 const char *, int, int, int,
1413 struct ptlrpc_request **);
1415 int (*m_init_ea_size)(struct obd_export *, int, int, int);
1417 int (*m_get_lustre_md)(struct obd_export *, struct ptlrpc_request *,
1418 struct obd_export *, struct obd_export *,
1419 struct lustre_md *);
1421 int (*m_free_lustre_md)(struct obd_export *, struct lustre_md *);
1423 int (*m_set_open_replay_data)(struct obd_export *,
1424 struct obd_client_handle *,
1425 struct ptlrpc_request *);
1426 int (*m_clear_open_replay_data)(struct obd_export *,
1427 struct obd_client_handle *);
1428 int (*m_set_lock_data)(struct obd_export *, __u64 *, void *);
1430 ldlm_mode_t (*m_lock_match)(struct obd_export *, int,
1431 const struct lu_fid *, ldlm_type_t,
1432 ldlm_policy_data_t *, ldlm_mode_t,
1433 struct lustre_handle *);
1435 int (*m_cancel_unused)(struct obd_export *, const struct lu_fid *,
1436 ldlm_policy_data_t *, ldlm_mode_t, int flags,
1438 int (*m_renew_capa)(struct obd_export *, struct obd_capa *oc,
1439 renew_capa_cb_t cb);
1441 int (*m_get_remote_perm)(struct obd_export *, const struct lu_fid *,
1442 struct obd_capa *, __u32,
1443 struct ptlrpc_request **);
1445 int (*m_intent_getattr_async)(struct obd_export *,
1446 struct md_enqueue_info *,
1447 struct ldlm_enqueue_info *);
1449 int (*m_revalidate_lock)(struct obd_export *,
1450 struct lookup_intent *,
1454 * NOTE: If adding ops, add another LPROCFS_MD_OP_INIT() line to
1455 * lprocfs_alloc_md_stats() in obdclass/lprocfs_status.c. Also, add a
1456 * wrapper function in include/linux/obd_class.h.
1460 struct lsm_operations {
1461 void (*lsm_free)(struct lov_stripe_md *);
1462 int (*lsm_destroy)(struct lov_stripe_md *, struct obdo *oa,
1463 struct obd_export *md_exp);
1464 void (*lsm_stripe_by_index)(struct lov_stripe_md *, int *, obd_off *,
1466 void (*lsm_stripe_by_offset)(struct lov_stripe_md *, int *, obd_off *,
1468 obd_off (*lsm_stripe_offset_by_index)(struct lov_stripe_md *, int);
1469 obd_off (*lsm_stripe_offset_by_offset)(struct lov_stripe_md *, obd_off);
1470 int (*lsm_stripe_index_by_offset)(struct lov_stripe_md *, obd_off);
1471 int (*lsm_revalidate) (struct lov_stripe_md *, struct obd_device *obd);
1472 int (*lsm_lmm_verify) (struct lov_mds_md *lmm, int lmm_bytes,
1474 int (*lsm_unpackmd) (struct lov_obd *lov, struct lov_stripe_md *lsm,
1475 struct lov_mds_md *lmm);
1478 extern struct lsm_operations lsm_plain_ops;
1479 extern struct lsm_operations lsm_join_ops;
1480 static inline struct lsm_operations *lsm_op_find(int magic)
1484 return &lsm_plain_ops;
1485 case LOV_MAGIC_JOIN:
1486 return &lsm_join_ops;
1488 CERROR("Cannot recognize lsm_magic %d\n", magic);
1493 int lvfs_check_io_health(struct obd_device *obd, struct file *file);
1495 /* Requests for obd_extent_calc() */
1496 #define OBD_CALC_STRIPE_START 1
1497 #define OBD_CALC_STRIPE_END 2
1499 static inline void obd_transno_commit_cb(struct obd_device *obd, __u64 transno,
1503 CERROR("%s: transno "LPU64" commit error: %d\n",
1504 obd->obd_name, transno, error);
1507 if (transno > obd->obd_last_committed) {
1508 CDEBUG(D_HA, "%s: transno "LPD64" committed\n",
1509 obd->obd_name, transno);
1510 obd->obd_last_committed = transno;
1511 ptlrpc_commit_replies (obd);
1513 CDEBUG(D_INFO, "%s: transno "LPD64" committed\n",
1514 obd->obd_name, transno);
1518 static inline void init_obd_quota_ops(quota_interface_t *interface,
1519 struct obd_ops *obd_ops)
1525 obd_ops->o_quotacheck = QUOTA_OP(interface, check);
1526 obd_ops->o_quotactl = QUOTA_OP(interface, ctl);
1529 static inline __u64 oinfo_mdsno(struct obd_info *oinfo)
1531 return oinfo->oi_oa->o_gr - FILTER_GROUP_MDS0;
1534 static inline struct lustre_capa *oinfo_capa(struct obd_info *oinfo)
1536 return oinfo->oi_capa;
1539 #endif /* __OBD_H */