X-Git-Url: https://git.whamcloud.com/?p=fs%2Flustre-release.git;a=blobdiff_plain;f=lustre%2Finclude%2Flinux%2Fobd.h;h=104acc0590c81945f37e656c4103643f7f0c5ad3;hp=b3f5d8cc8aeccfe1ab69c0459e231b12f6796275;hb=87c86d444e61e38d6454bba5700ba966dc1ac83d;hpb=994f154b10b2975ba3f30dc797edfa9d08e159a6 diff --git a/lustre/include/linux/obd.h b/lustre/include/linux/obd.h index b3f5d8c..104acc0 100644 --- a/lustre/include/linux/obd.h +++ b/lustre/include/linux/obd.h @@ -13,12 +13,14 @@ #define IOC_OSC_TYPE 'h' #define IOC_OSC_MIN_NR 20 #define IOC_OSC_SET_ACTIVE _IOWR(IOC_OSC_TYPE, 21, struct obd_device *) +#define IOC_OSC_CTL_RECOVERY _IOWR(IOC_OSC_TYPE, 22, struct obd_device *) #define IOC_OSC_MAX_NR 50 #define IOC_MDC_TYPE 'i' #define IOC_MDC_MIN_NR 20 #define IOC_MDC_LOOKUP _IOWR(IOC_MDC_TYPE, 20, struct obd_device *) -#define IOC_MDC_GETSTRIPE _IOWR(IOC_MDC_TYPE, 21, struct lov_mds_md *) +/* Moved to lustre_user.h +#define IOC_MDC_GETSTRIPE _IOWR(IOC_MDC_TYPE, 21, struct lov_mds_md *) */ #define IOC_MDC_MAX_NR 50 #ifdef __KERNEL__ @@ -31,6 +33,7 @@ # include #endif +#include #include #include #include @@ -40,38 +43,49 @@ struct loi_oap_pages { struct list_head lop_pending; int lop_num_pending; struct list_head lop_urgent; - struct list_head lop_pending_sync; + struct list_head lop_pending_group; }; struct lov_oinfo { /* per-stripe data structure */ __u64 loi_id; /* object ID on the target OST */ __u64 loi_gr; /* object group on the target OST */ - struct lustre_handle *loi_handle; /* open file handle for obj on OST */ int loi_ost_idx; /* OST stripe index in lov_tgt_desc->tgts */ int loi_ost_gen; /* generation of this loi_ost_idx */ - /* tracking offsets per file, per stripe.. */ + /* used by the osc to keep track of what objects to build into rpcs */ struct loi_oap_pages loi_read_lop; struct loi_oap_pages loi_write_lop; + /* _cli_ is poorly named, it should be _ready_ */ struct list_head loi_cli_item; + struct list_head loi_write_item; + struct list_head loi_read_item; + + unsigned loi_kms_valid:1; + __u64 loi_kms; /* known minimum size */ + __u64 loi_rss; /* recently seen size */ + __u64 loi_mtime; /* recently seen mtime */ + __u64 loi_blocks; /* recently seen blocks */ }; static inline void loi_init(struct lov_oinfo *loi) { INIT_LIST_HEAD(&loi->loi_read_lop.lop_pending); INIT_LIST_HEAD(&loi->loi_read_lop.lop_urgent); - INIT_LIST_HEAD(&loi->loi_read_lop.lop_pending_sync); + INIT_LIST_HEAD(&loi->loi_read_lop.lop_pending_group); INIT_LIST_HEAD(&loi->loi_write_lop.lop_pending); INIT_LIST_HEAD(&loi->loi_write_lop.lop_urgent); - INIT_LIST_HEAD(&loi->loi_write_lop.lop_pending_sync); + INIT_LIST_HEAD(&loi->loi_write_lop.lop_pending_group); INIT_LIST_HEAD(&loi->loi_cli_item); + INIT_LIST_HEAD(&loi->loi_write_item); + INIT_LIST_HEAD(&loi->loi_read_item); } struct lov_stripe_md { /* Public members. */ __u64 lsm_object_id; /* lov object id */ __u64 lsm_object_gr; /* lov object id */ - __u64 lsm_maxbytes; + __u64 lsm_maxbytes; /* maximum possible file size */ + unsigned long lsm_xfersize; /* optimal transfer size */ /* LOV-private members start here -- only for use in lov/. */ __u32 lsm_magic; @@ -81,40 +95,78 @@ struct lov_stripe_md { struct lov_oinfo lsm_oinfo[0]; }; +static inline void dump_lsm(int level, struct lov_stripe_md *lsm) +{ + int i; + CDEBUG(level, "objid "LPX64"/"LPU64", maxbytes "LPX64", magic 0x%08X, " + "stripe_size %u, stripe_count %u\n", + lsm->lsm_object_id, lsm->lsm_object_gr, lsm->lsm_maxbytes, + lsm->lsm_magic, lsm->lsm_stripe_size, lsm->lsm_stripe_count); + for (i = 0; i < lsm->lsm_stripe_count; i++) + CDEBUG(level, "idx %u ostidx %u/%u object "LPU64"/"LPU64"\n", + i, lsm->lsm_oinfo[i].loi_ost_idx, + lsm->lsm_oinfo[i].loi_ost_gen, + lsm->lsm_oinfo[i].loi_id, lsm->lsm_oinfo[i].loi_gr); +} + struct obd_type { struct list_head typ_chain; struct obd_ops *typ_ops; + struct md_ops *typ_md_ops; struct proc_dir_entry *typ_procroot; char *typ_name; int typ_refcnt; }; struct brw_page { - obd_off off; + obd_off disk_offset; /* modulo PAGE_SIZE */ + obd_off page_offset; /* modulo PAGE_SIZE (obviously) */ struct page *pg; int count; - obd_flag flag; + obd_flags flag; }; enum async_flags { - ASYNC_READY = 0x1, + ASYNC_READY = 0x1, /* ap_make_ready will not be called before this + page is added to an rpc */ ASYNC_URGENT = 0x2, - ASYNC_COUNT_STABLE = 0x4, + ASYNC_COUNT_STABLE = 0x4, /* ap_refresh_count will not be called + to give the caller a chance to update + or cancel the size of the io */ + ASYNC_GROUP_SYNC = 0x8, /* ap_completion will not be called, instead + the page is accounted for in the + obd_io_group given to + obd_queue_group_io */ }; struct obd_async_page_ops { int (*ap_make_ready)(void *data, int cmd); int (*ap_refresh_count)(void *data, int cmd); void (*ap_fill_obdo)(void *data, int cmd, struct obdo *oa); - void (*ap_completion)(void *data, int cmd, int rc); + void (*ap_completion)(void *data, int cmd, struct obdo *oa, int rc); +}; + +/* the `oig' is passed down from a caller of obd rw methods. the callee + * records enough state such that the caller can sleep on the oig and + * be woken when all the callees have finished their work */ +struct obd_io_group { + spinlock_t oig_lock; + atomic_t oig_refcount; + int oig_pending; + int oig_rc; + struct list_head oig_occ_list; + wait_queue_head_t oig_waitq; }; -struct obd_sync_io_container { - spinlock_t osic_lock; - atomic_t osic_refcount; - int osic_pending; - int osic_rc; - wait_queue_head_t osic_waitq; +/* the oig callback context lets the callee of obd rw methods register + * for callbacks from the caller. */ +struct oig_callback_context { + struct list_head occ_oig_item; + /* called when the caller has received a signal while sleeping. + * callees of this method are encouraged to abort their state + * in the oig. This may be called multiple times. */ + void (*occ_interrupted)(struct oig_callback_context *occ); + int interrupted; }; /* if we find more consumers this could be generalized */ @@ -124,42 +176,73 @@ struct obd_histogram { unsigned long oh_buckets[OBD_HIST_MAX]; }; -/* Individual type definitions */ +/* reports average service time with the help of lprocfs_status.c */ +struct obd_service_time { + __u32 st_num; + __u64 st_total_us; +}; + struct ost_server_data; +#define FILTER_SUBDIR_COUNT 32 /* set to zero for no subdirs */ + +#define FILTER_GROUP_LLOG 1 +#define FILTER_GROUP_ECHO 2 +#define FILTER_GROUP_FIRST_MDS 3 + +struct filter_subdirs { + struct dentry *dentry[FILTER_SUBDIR_COUNT]; +}; + +struct filter_group_llog { + struct list_head list; + int group; + struct obd_llogs *llogs; + struct obd_export *exp; +}; + struct filter_obd { const char *fo_fstype; struct super_block *fo_sb; struct vfsmount *fo_vfsmnt; - struct dentry *fo_dentry_O; - struct dentry **fo_dentry_O_groups; - struct dentry **fo_dentry_O_sub; - spinlock_t fo_objidlock; /* protect fo_lastobjid increment */ - spinlock_t fo_translock; /* protect fsd_last_rcvd increment */ + struct lvfs_obd_ctxt *fo_lvfs_ctxt; + + int fo_group_count; + struct dentry *fo_dentry_O; /* the "O"bject directory dentry */ + struct dentry **fo_groups; /* dentries for each group dir */ + struct filter_subdirs *fo_subdirs; /* subdir array per group */ + __u64 *fo_last_objids; /* per-group last created objid */ + struct file **fo_last_objid_files; + struct semaphore fo_init_lock; /* group initialization lock */ + int fo_committed_group; + + spinlock_t fo_objidlock; /* protect fo_lastobjid increment */ + spinlock_t fo_translock; /* protect fsd_last_rcvd increment */ struct file *fo_rcvd_filp; struct filter_server_data *fo_fsd; unsigned long *fo_last_rcvd_slots; __u64 fo_mount_count; + unsigned long fo_destroys_in_progress; + struct semaphore fo_create_locks[32]; + struct file_operations *fo_fop; struct inode_operations *fo_iop; struct address_space_operations *fo_aops; struct list_head fo_export_list; int fo_subdir_count; - spinlock_t fo_grant_lock; /* protects tot_granted */ - obd_size fo_tot_granted; - obd_size fo_tot_cached; + + obd_size fo_tot_dirty; /* protected by obd_osfs_lock */ + obd_size fo_tot_granted; /* all values in bytes */ + obd_size fo_tot_pending; + + obd_size fo_readcache_max_filesize; struct obd_import *fo_mdc_imp; struct obd_uuid fo_mdc_uuid; struct lustre_handle fo_mdc_conn; -#if 0 - struct ptlrpc_client fo_mdc_client; -#endif - struct file **fo_last_objid_files; - __u64 *fo_last_objids; //last created object ID for groups struct semaphore fo_alloc_lock; @@ -169,10 +252,24 @@ struct filter_obd { struct obd_histogram fo_w_discont_pages; struct obd_histogram fo_r_discont_blocks; struct obd_histogram fo_w_discont_blocks; + + struct list_head fo_llog_list; + spinlock_t fo_llog_list_lock; + + /* which secure flavor from remote is denied */ + spinlock_t fo_denylist_lock; + struct list_head fo_denylist; + }; struct mds_server_data; +#define OSC_MAX_RIF_DEFAULT 8 +#define OSC_MAX_RIF_MAX 64 +#define OSC_MAX_DIRTY_DEFAULT (4*OSC_MAX_RIF_DEFAULT*PTLRPC_MAX_BRW_SIZE>>20) +#define OSC_MAX_DIRTY_MB_MAX 512 /* totally arbitrary */ + + struct mdc_rpc_lock; struct client_obd { struct obd_import *cl_import; @@ -184,28 +281,31 @@ struct client_obd { int cl_max_mds_cookiesize; kdev_t cl_sandev; + /* security flavors */ + __u32 cl_sec_flavor; + __u32 cl_sec_subflavor; + //struct llog_canceld_ctxt *cl_llcd; /* it's included by obd_llog_ctxt */ void *cl_llcd_offset; - struct semaphore cl_dirty_sem; - obd_size cl_dirty; /* all _dirty_ in bytes */ - obd_size cl_dirty_granted; /* from ost */ - obd_size cl_dirty_max; /* allowed w/o rpc */ - struct list_head cl_cache_waiters; - struct obd_device *cl_mgmtcli_obd; - /* this is just to keep existing infinitely caching behaviour between - * clients and OSTs that don't have the grant code in yet.. it can - * be yanked once everything speaks grants */ - char cl_ost_can_grant; + /* the grant values are protected by loi_list_lock below */ + long cl_dirty; /* all _dirty_ in bytes */ + long cl_dirty_max; /* allowed w/o rpc */ + long cl_avail_grant; /* bytes of credit for ost */ + long cl_lost_grant; /* lost credits (trunc) */ + struct list_head cl_cache_waiters; /* waiting for cache/grant */ /* keep track of objects that have lois that contain pages which * have been queued for async brw. this lock also protects the * lists of osc_client_pages that hang off of the loi */ spinlock_t cl_loi_list_lock; struct list_head cl_loi_ready_list; - int cl_brw_in_flight; + struct list_head cl_loi_write_list; + struct list_head cl_loi_read_list; + int cl_r_in_flight; + int cl_w_in_flight; /* just a sum of the loi/lop pending numbers to be exported by /proc */ int cl_pending_w_pages; int cl_pending_r_pages; @@ -215,9 +315,15 @@ struct client_obd { struct obd_histogram cl_write_rpc_hist; struct obd_histogram cl_read_page_hist; struct obd_histogram cl_write_page_hist; + struct obd_service_time cl_read_stime; + struct obd_service_time cl_write_stime; + struct obd_service_time cl_enter_stime; struct mdc_rpc_lock *cl_rpc_lock; - struct mdc_rpc_lock *cl_setattr_lock; + struct mdc_rpc_lock *cl_setattr_lock; + struct mdc_rpc_lock *cl_close_lock; + struct osc_creator cl_oscc; + int cl_async:1; }; /* Like a client, with some hangers-on. Keep mc_client_obd first so that we @@ -238,49 +344,79 @@ struct mds_obd { struct ptlrpc_service *mds_readpage_service; struct super_block *mds_sb; struct vfsmount *mds_vfsmnt; - struct dentry *mds_fid_de; + struct dentry *mds_id_de; + struct lvfs_obd_ctxt *mds_lvfs_ctxt; int mds_max_mdsize; int mds_max_cookiesize; struct file *mds_rcvd_filp; + struct file *mds_fid_filp; + struct file *mds_virtid_filp; spinlock_t mds_transno_lock; __u64 mds_last_transno; __u64 mds_mount_count; __u64 mds_io_epoch; + + __u64 mds_last_fid; + __u64 mds_virtid_fid; + spinlock_t mds_last_fid_lock; + struct semaphore mds_epoch_sem; - struct ll_fid mds_rootfid; + struct lustre_id mds_rootid; struct mds_server_data *mds_server_data; struct dentry *mds_pending_dir; struct dentry *mds_logs_dir; struct dentry *mds_objects_dir; struct llog_handle *mds_cfg_llh; -// struct llog_handle *mds_catalog; - struct obd_device *mds_osc_obd; /* XXX lov_obd */ - struct obd_uuid mds_lov_uuid; char *mds_profile; - struct obd_export *mds_osc_exp; /* XXX lov_exp */ - int mds_has_lov_desc; - struct lov_desc mds_lov_desc; - obd_id *mds_lov_objids; - int mds_lov_objids_valid; - int mds_lov_nextid_set; - struct file *mds_lov_objid_filp; + struct obd_device *mds_dt_obd; + struct obd_uuid mds_dt_uuid; + struct obd_export *mds_dt_exp; + int mds_has_dt_desc; + struct lov_desc mds_dt_desc; + + spinlock_t mds_dt_lock; + obd_id *mds_dt_objids; + struct file *mds_dt_objid_filp; + int mds_dt_objids_valid; + unsigned long *mds_client_bitmap; struct semaphore mds_orphan_recovery_sem; - + + int mds_num; + int mds_config_version; + + char *mds_md_name; + struct obd_device *mds_md_obd; + struct obd_export *mds_md_exp; + struct semaphore mds_md_sem; + struct obd_uuid mds_md_uuid; + int mds_md_connected; + + struct ptlrpc_service *mds_create_service; + uid_t mds_squash_uid; + gid_t mds_squash_gid; + ptl_nid_t mds_nosquash_nid; + atomic_t mds_real_clients; atomic_t mds_open_count; + struct dentry *mds_id_dir; + int mds_obd_type; + struct dentry *mds_unnamed_dir; /* for mdt_obd_create only */ + + /* security related */ + char *mds_mds_sec; + char *mds_ost_sec; + /* which secure flavor from remote to this mds is denied */ + spinlock_t mds_denylist_lock; + struct list_head mds_denylist; + struct semaphore mds_create_sem; }; struct echo_obd { - struct obdo oa; - spinlock_t eo_lock; - __u64 eo_lastino; - atomic_t eo_getattr; - atomic_t eo_setattr; - atomic_t eo_create; - atomic_t eo_destroy; - atomic_t eo_prep; - atomic_t eo_read; - atomic_t eo_write; + struct obdo eo_oa; + spinlock_t eo_lock; + __u64 eo_lastino; + struct lustre_handle eo_nl_lock; + atomic_t eo_prep; }; /* @@ -312,48 +448,107 @@ struct recovd_obd { }; struct ost_obd { + spinlock_t ost_lock; struct ptlrpc_service *ost_service; struct ptlrpc_service *ost_create_service; + struct obd_service_time ost_stimes[6]; }; struct echo_client_obd { - struct obd_export *ec_exp; /* the local connection to osc/lov */ - spinlock_t ec_lock; - struct list_head ec_objects; - int ec_nstripes; - __u64 ec_unique; + struct obd_export *ec_exp; /* the local connection to osc/lov */ + spinlock_t ec_lock; + struct list_head ec_objects; + int ec_nstripes; + __u64 ec_unique; }; struct cache_obd { - struct obd_export *cobd_target_exp;/* local connection to target obd */ - struct obd_export *cobd_cache_exp; /* local connection to cache obd */ + struct obd_export *master_exp; /* local connection to master obd */ + struct obd_export *cache_exp; /* local connection to cache obd */ + struct obd_export *cache_real_exp; + struct obd_export *master_real_exp; + struct obd_device *master; + struct obd_device *cache; + char *master_name; + char *cache_name; + int refcount; + int cache_on; + struct semaphore sem; +}; + +struct cm_obd { + struct obd_export *cache_exp; /* local connection to cache obd */ + struct obd_export *master_exp; + struct obd_device *cache_obd; + struct obd_device *master_obd; + int master_group; + struct cmobd_write_service *write_srv; +}; + +struct conf_obd { + struct super_block *cfobd_sb; + struct vfsmount *cfobd_vfsmnt; + struct dentry *cfobd_logs_dir; + struct dentry *cfobd_objects_dir; + struct dentry *cfobd_pending_dir; + struct llog_handle *cfobd_cfg_llh; + struct lvfs_obd_ctxt *cfobd_lvfs_ctxt; }; struct lov_tgt_desc { - struct obd_uuid uuid; - struct obd_export *ltd_exp; - int active; /* is this target up for requests */ + struct obd_uuid uuid; + __u32 ltd_gen; + struct obd_export *ltd_exp; + int active; /* is this target up for requests */ }; struct lov_obd { - spinlock_t lov_lock; - struct lov_desc desc; - int bufsize; - int refcount; - int lo_catalog_loaded:1; - struct lov_tgt_desc *tgts; + spinlock_t lov_lock; + struct lov_desc desc; + int bufsize; + int refcount; + int lo_catalog_loaded:1, async:1; + struct semaphore lov_llog_sem; + unsigned long lov_connect_flags; + struct lov_tgt_desc *tgts; +}; + +struct lmv_tgt_desc { + struct obd_uuid uuid; + struct obd_export *ltd_exp; + int active; /* is this target up for requests */ +}; + +struct lmv_obd { + int refcount; + spinlock_t lmv_lock; + struct lmv_desc desc; + struct lmv_tgt_desc *tgts; + struct obd_uuid cluuid; + struct obd_export *exp; + + int tgts_size; + int connected; + int max_easize; + int max_cookiesize; + int server_timeout; + int connect_flags; + struct semaphore init_sem; + struct obd_connect_data conn_data; }; struct niobuf_local { __u64 offset; __u32 len; __u32 flags; - int rc; struct page *page; struct dentry *dentry; - unsigned long start; + int lnb_grant_used; + int rc; }; +#define OBD_MODE_ASYNC (1 << 0) +#define OBD_MODE_CROW (1 << 1) /* Don't conflict with on-wire flags OBD_BRW_WRITE, etc */ #define N_LOCAL_TEMP_PAGE 0x10000000 @@ -361,7 +556,8 @@ struct niobuf_local { struct obd_trans_info { __u64 oti_transno; __u64 *oti_objid; - /* Only used on the server side for tracking acks. */ + + /* only used on the server side for tracking acks. */ struct oti_req_ack_lock { struct lustre_handle lock; __u32 mode; @@ -370,6 +566,7 @@ struct obd_trans_info { struct llog_cookie oti_onecookie; struct llog_cookie *oti_logcookies; int oti_numcookies; + int oti_flags; }; static inline void oti_alloc_cookies(struct obd_trans_info *oti,int num_cookies) @@ -414,9 +611,20 @@ enum llog_ctxt_id { LLOG_RD1_REPL_CTXT = 9, LLOG_TEST_ORIG_CTXT = 10, LLOG_TEST_REPL_CTXT = 11, + LLOG_REINT_ORIG_CTXT = 12, LLOG_MAX_CTXTS }; +struct obd_llogs { + struct llog_ctxt *llog_ctxt[LLOG_MAX_CTXTS]; +}; + +struct target_recovery_data { + svc_handler_t trd_recovery_handler; + pid_t trd_processing_task; + struct completion trd_starting; + struct completion trd_finishing; +}; /* corresponds to one of the obd's */ struct obd_device { @@ -427,9 +635,9 @@ struct obd_device { struct obd_uuid obd_uuid; int obd_minor; - int obd_attached:1, obd_set_up:1, obd_recovering:1, - obd_abort_recovery:1, obd_replayable:1, obd_no_transno:1, - obd_no_recov:1, obd_stopping:1; + unsigned int obd_attached:1, obd_set_up:1, obd_recovering:1, + obd_abort_recovery:1, obd_replayable:1, obd_no_transno:1, + obd_no_recov:1, obd_stopping:1; atomic_t obd_refcount; wait_queue_head_t obd_refcount_waitq; struct proc_dir_entry *obd_proc_entry; @@ -441,53 +649,76 @@ struct obd_device { spinlock_t obd_dev_lock; __u64 obd_last_committed; struct fsfilt_operations *obd_fsops; - struct llog_ctxt *obd_llog_ctxt[LLOG_MAX_CTXTS]; + spinlock_t obd_osfs_lock; struct obd_statfs obd_osfs; unsigned long obd_osfs_age; - struct obd_run_ctxt obd_ctxt; + struct lvfs_run_ctxt obd_lvfs_ctxt; + struct obd_llogs obd_llogs; + struct llog_ctxt *obd_llog_ctxt[LLOG_MAX_CTXTS]; struct obd_device *obd_observer; struct obd_export *obd_self_export; - /* XXX encapsulate all this recovery data into one struct */ - svc_handler_t obd_recovery_handler; + struct target_recovery_data obd_recovery_data; + /* XXX encapsulate all this recovery data into target_recovery_data */ int obd_max_recoverable_clients; int obd_connected_clients; int obd_recoverable_clients; spinlock_t obd_processing_task_lock; - pid_t obd_processing_task; __u64 obd_next_recovery_transno; int obd_replayed_requests; + int obd_replayed_locks; int obd_requests_queued_for_recovery; wait_queue_head_t obd_next_transno_waitq; - wait_queue_head_t obd_commit_waitq; + struct list_head obd_uncommitted_replies; + spinlock_t obd_uncommitted_replies_lock; struct timer_list obd_recovery_timer; - struct list_head obd_recovery_queue; - struct list_head obd_delayed_reply_queue; + time_t obd_recovery_start; + time_t obd_recovery_end; + + atomic_t obd_req_replay_clients; + atomic_t obd_lock_replay_clients; + + struct list_head obd_req_replay_queue; + struct list_head obd_lock_replay_queue; + struct list_head obd_final_req_queue; + int obd_recovery_stage; union { - struct filter_obd filter; - struct mds_obd mds; - struct client_obd cli; - struct ost_obd ost; - struct echo_client_obd echo_client; - struct echo_obd echo; - struct recovd_obd recovd; - struct lov_obd lov; - struct cache_obd cobd; - struct ptlbd_obd ptlbd; - struct mgmtcli_obd mgmtcli; + struct filter_obd filter; + struct mds_obd mds; + struct client_obd cli; + struct ost_obd ost; + struct echo_client_obd echocli; + struct echo_obd echo; + struct recovd_obd recovd; + struct lov_obd lov; + struct cache_obd cobd; + struct ptlbd_obd ptlbd; + struct mgmtcli_obd mgmtcli; + struct lmv_obd lmv; + struct cm_obd cm; + struct conf_obd conf; } u; - /* Fields used by LProcFS */ + + /* fields used by LProcFS */ unsigned int obd_cntr_base; struct lprocfs_stats *obd_stats; + unsigned int md_cntr_base; + struct lprocfs_stats *md_stats; + struct proc_dir_entry *obd_svc_procroot; struct lprocfs_stats *obd_svc_stats; }; -#define OBD_OPT_FORCE 0x0001 -#define OBD_OPT_FAILOVER 0x0002 +#define OBD_OPT_FORCE (1 << 0) +#define OBD_OPT_FAILOVER (1 << 1) +#define OBD_OPT_REAL_CLIENT (1 << 2) +#define OBD_OPT_MDS_CONNECTION (1 << 3) + +#define OBD_LLOG_FL_SENDNOW (1 << 0) +#define OBD_LLOG_FL_CREATE (1 << 1) -#define OBD_LLOG_FL_SENDNOW 0x0001 +struct mdc_op_data; struct obd_ops { struct module *o_owner; @@ -500,13 +731,19 @@ struct obd_ops { int (*o_attach)(struct obd_device *dev, obd_count len, void *data); int (*o_detach)(struct obd_device *dev); int (*o_setup) (struct obd_device *dev, obd_count len, void *data); - int (*o_postsetup) (struct obd_device *dev); int (*o_precleanup)(struct obd_device *dev, int flags); int (*o_cleanup)(struct obd_device *dev, int flags); + int (*o_process_config)(struct obd_device *dev, obd_count len, + void *data); int (*o_postrecov)(struct obd_device *dev); + int (*o_add_conn)(struct obd_import *imp, struct obd_uuid *uuid, + int priority); + int (*o_del_conn)(struct obd_import *imp, struct obd_uuid *uuid); int (*o_connect)(struct lustre_handle *conn, struct obd_device *src, - struct obd_uuid *cluuid); - int (*o_disconnect)(struct obd_export *exp, int flags); + struct obd_uuid *cluuid, struct obd_connect_data *data, + unsigned long flags); + int (*o_connect_post)(struct obd_export *exp, unsigned, unsigned long); + int (*o_disconnect)(struct obd_export *exp, unsigned long flags); int (*o_statfs)(struct obd_device *obd, struct obd_statfs *osfs, unsigned long max_age); @@ -514,9 +751,13 @@ struct obd_ops { struct lov_stripe_md *mem_src); int (*o_unpackmd)(struct obd_export *exp,struct lov_stripe_md **mem_tgt, struct lov_mds_md *disk_src, int disk_len); + int (*o_revalidate_md)(struct obd_export *exp, struct obdo *oa, + struct lov_stripe_md *ea, + struct obd_trans_info *oti); int (*o_preallocate)(struct lustre_handle *, obd_count *req, obd_id *ids); int (*o_create)(struct obd_export *exp, struct obdo *oa, + void *acl, int acl_size, struct lov_stripe_md **ea, struct obd_trans_info *oti); int (*o_destroy)(struct obd_export *exp, struct obdo *oa, struct lov_stripe_md *ea, struct obd_trans_info *oti); @@ -544,24 +785,26 @@ struct obd_ops { struct lov_stripe_md *lsm, struct lov_oinfo *loi, void *cookie, int cmd, obd_off off, int count, - obd_flag brw_flags, obd_flag async_flags); - int (*o_queue_sync_io)(struct obd_export *exp, - struct lov_stripe_md *lsm, - struct lov_oinfo *loi, - struct obd_sync_io_container *osic, - void *cookie, int cmd, obd_off off, int count, - obd_flag brw_flags); - int (*o_trigger_sync_io)(struct obd_export *exp, - struct lov_stripe_md *lsm, - struct lov_oinfo *loi, - struct obd_sync_io_container *osic); + obd_flags brw_flags, obd_flags async_flags); + int (*o_queue_group_io)(struct obd_export *exp, + struct lov_stripe_md *lsm, + struct lov_oinfo *loi, + struct obd_io_group *oig, + void *cookie, int cmd, obd_off off, int count, + obd_flags brw_flags, obd_flags async_flags); + int (*o_trigger_group_io)(struct obd_export *exp, + struct lov_stripe_md *lsm, + struct lov_oinfo *loi, + struct obd_io_group *oig); int (*o_set_async_flags)(struct obd_export *exp, struct lov_stripe_md *lsm, struct lov_oinfo *loi, void *cookie, - obd_flag async_flags); + obd_flags async_flags); int (*o_teardown_async_page)(struct obd_export *exp, struct lov_stripe_md *lsm, struct lov_oinfo *loi, void *cookie); + int (*o_adjust_kms)(struct obd_export *exp, struct lov_stripe_md *lsm, + obd_off size, int shrink); int (*o_punch)(struct obd_export *exp, struct obdo *oa, struct lov_stripe_md *ea, obd_size start, obd_size end, struct obd_trans_info *oti); @@ -583,18 +826,22 @@ struct obd_ops { int (*o_commitrw)(int cmd, struct obd_export *exp, struct obdo *oa, int objcount, struct obd_ioobj *obj, int niocount, struct niobuf_local *local, - struct obd_trans_info *oti); - int (*o_enqueue)(struct obd_export *exp, struct lov_stripe_md *md, - struct lustre_handle *parent_lock, - __u32 type, void *cookie, int cookielen, __u32 mode, - int *flags, void *cb, void *data, + struct obd_trans_info *oti, int rc); + int (*o_do_cow)(struct obd_export *exp, struct obd_ioobj *obj, + int objcount, struct niobuf_remote *rnb); + int (*o_write_extents)(struct obd_export *exp, struct obd_ioobj *obj, + int objcount, int niocount, + struct niobuf_local *local,int rc); + int (*o_enqueue)(struct obd_export *, struct lov_stripe_md *, + __u32 type, ldlm_policy_data_t *, __u32 mode, + int *flags, void *bl_cb, void *cp_cb, void *gl_cb, + void *data, __u32 lvb_len, void *lvb_swabber, struct lustre_handle *lockh); - int (*o_match)(struct obd_export *exp, struct lov_stripe_md *md, - __u32 type, void *cookie, int cookielen, __u32 mode, - int *flags, void *data, struct lustre_handle *lockh); - int (*o_change_cbdata)(struct obd_export *exp, - struct lov_stripe_md *lsm, ldlm_iterator_t it, - void *data); + int (*o_match)(struct obd_export *, struct lov_stripe_md *, __u32 type, + ldlm_policy_data_t *, __u32 mode, int *flags, void *data, + struct lustre_handle *lockh); + int (*o_change_cbdata)(struct obd_export *, struct lov_stripe_md *, + ldlm_iterator_t it, void *data); int (*o_cancel)(struct obd_export *, struct lov_stripe_md *md, __u32 mode, struct lustre_handle *); int (*o_cancel_unused)(struct obd_export *, struct lov_stripe_md *, @@ -607,46 +854,121 @@ struct obd_ops { int (*o_destroy_export)(struct obd_export *exp); /* llog related obd_methods */ - int (*o_llog_init)(struct obd_device *obd, struct obd_device *disk_obd, - int count, struct llog_logid *logid); - int (*o_llog_finish)(struct obd_device *obd, int count); - - /* only until proper file size mechanics arrive */ - int (*o_lock_contains)(struct obd_export *exp, - struct lov_stripe_md *lsm, - struct ldlm_lock *lock, obd_off offset); + int (*o_llog_init)(struct obd_device *, struct obd_llogs *, + struct obd_device *, int, struct llog_catid *); + int (*o_llog_finish)(struct obd_device *, struct obd_llogs *, int); + int (*o_llog_connect)(struct obd_export *, struct llogd_conn_body *); + /* metadata-only methods */ int (*o_pin)(struct obd_export *, obd_id ino, __u32 gen, int type, struct obd_client_handle *, int flag); int (*o_unpin)(struct obd_export *, struct obd_client_handle *, int); - int (*o_invalidate_import)(struct obd_device *, struct obd_import *); + int (*o_import_event)(struct obd_device *, struct obd_import *, + enum obd_import_event); int (*o_notify)(struct obd_device *obd, struct obd_device *watched, - int active); + int active, void *data); + + /* this method is needed for cmobd to make say to LMV "get ready" when + * master obd is LMV. This is needed, because LMV is initialized in + * "defered" manner to let all MDSs to be set up first. */ + int (*o_getready)(struct obd_export *exp); + int (*o_init_ea_size)(struct obd_export *, int, int); + /* * NOTE: If adding ops, add another LPROCFS_OBD_OP_INIT() line * to lprocfs_alloc_obd_stats() in obdclass/lprocfs_status.c. * Also, add a wrapper function in include/linux/obd_class.h. */ - }; +struct md_ops { + int (*m_getstatus)(struct obd_export *, struct lustre_id *); + int (*m_change_cbdata)(struct obd_export *, struct lustre_id *, + ldlm_iterator_t, void *); + int (*m_change_cbdata_name)(struct obd_export *, struct lustre_id *, + char *, int, struct lustre_id *, + ldlm_iterator_t, void *); + int (*m_close)(struct obd_export *, struct obdo *, + struct obd_client_handle *, + struct ptlrpc_request **); + int (*m_create)(struct obd_export *, struct mdc_op_data *, + const void *, int, int, __u32, __u32, + __u64, struct ptlrpc_request **); + int (*m_done_writing)(struct obd_export *, struct obdo *); + int (*m_enqueue)(struct obd_export *, int, struct lookup_intent *, + int, struct mdc_op_data *, struct lustre_handle *, + void *, int, ldlm_completion_callback, + ldlm_blocking_callback, void *); + int (*m_getattr)(struct obd_export *, struct lustre_id *, + __u64, const char *, int, + unsigned int, struct ptlrpc_request **); + int (*m_getattr_lock)(struct obd_export *, struct lustre_id *, + char *, int, __u64, + unsigned int, struct ptlrpc_request **); + int (*m_intent_lock)(struct obd_export *, + struct lustre_id *, const char *, int, + void *, int, struct lustre_id *, + struct lookup_intent *, int, + struct ptlrpc_request **, + ldlm_blocking_callback); + int (*m_link)(struct obd_export *, struct mdc_op_data *, + struct ptlrpc_request **); + int (*m_rename)(struct obd_export *, struct mdc_op_data *, + const char *, int, const char *, int, + struct ptlrpc_request **); + int (*m_setattr)(struct obd_export *, struct mdc_op_data *, + struct iattr *, void *, int , void *, int, + struct ptlrpc_request **); + int (*m_sync)(struct obd_export *, struct lustre_id *, + struct ptlrpc_request **); + int (*m_readpage)(struct obd_export *, struct lustre_id *, + __u64, struct page *, struct ptlrpc_request **); + int (*m_unlink)(struct obd_export *, struct mdc_op_data *, + struct ptlrpc_request **); + int (*m_valid_attrs)(struct obd_export *, struct lustre_id *); + + struct obd_device *(*m_get_real_obd)(struct obd_export *, struct lustre_id *); + + int (*m_req2lustre_md)(struct obd_export *exp, + struct ptlrpc_request *req, unsigned int offset, + struct obd_export *osc_exp, struct lustre_md *md); + int (*m_set_open_replay_data)(struct obd_export *exp, + struct obd_client_handle *och, + struct ptlrpc_request *open_req); + int (*m_clear_open_replay_data)(struct obd_export *exp, + struct obd_client_handle *och); + int (*m_store_inode_generation)(struct obd_export *exp, + struct ptlrpc_request *req, int reqoff, + int repoff); + int (*m_set_lock_data)(struct obd_export *exp, __u64 *l, void *data); + + int (*m_delete_inode)(struct obd_export *, struct lustre_id *); + + /* + * NOTE: If adding ops, add another LPROCFS_MD_OP_INIT() line to + * lprocfs_alloc_md_stats() in obdclass/lprocfs_status.c. Also, add a + * wrapper function in include/linux/obd_class.h. + */ +}; -static inline void obd_transno_commit_cb(struct obd_device *obd, __u64 transno, - int error) +static inline void obd_transno_commit_cb(struct obd_device *obd, + __u64 transno, int error) { if (error) { CERROR("%s: transno "LPD64" commit error: %d\n", obd->obd_name, transno, error); return; } + CDEBUG(D_HA, "%s: transno "LPD64" committed\n", obd->obd_name, transno); + if (transno > obd->obd_last_committed) { obd->obd_last_committed = transno; - wake_up(&obd->obd_commit_waitq); + ptlrpc_commit_replies (obd); } }