X-Git-Url: https://git.whamcloud.com/?p=fs%2Flustre-release.git;a=blobdiff_plain;f=lustre%2Finclude%2Fobd.h;h=6afec97acd30dee967f62ba3f0e133cac04364f6;hp=7c20ac2fc9260df455ae1f9d11faf69d2d00f0bc;hb=a067251099b6b225f2409f680d9e4423253d0730;hpb=6d0c7ccce95ebbbb84bf08c32c7d00b33591e9cf diff --git a/lustre/include/obd.h b/lustre/include/obd.h index 7c20ac2..6afec97 100644 --- a/lustre/include/obd.h +++ b/lustre/include/obd.h @@ -27,7 +27,7 @@ * Copyright (c) 2007, 2010, Oracle and/or its affiliates. All rights reserved. * Use is subject to license terms. * - * Copyright (c) 2011, 2012, Whamcloud, Inc. + * Copyright (c) 2011, 2013, Intel Corporation. */ /* * This file is part of Lustre, http://www.lustre.org/ @@ -57,19 +57,18 @@ #define IOC_MDC_MAX_NR 50 #include +#include +#include #ifdef HAVE_SERVER_SUPPORT # include +# include #endif #include -#include #include -#include +#include #include #include -#include - - #define MAX_OBD_DEVICES 8192 struct osc_async_rc { @@ -79,17 +78,15 @@ struct osc_async_rc { }; struct lov_oinfo { /* per-stripe data structure */ - struct ost_id loi_oi; /* object ID/Sequence on the target OST */ - int loi_ost_idx; /* OST stripe index in lov_tgt_desc->tgts */ - int loi_ost_gen; /* generation of this loi_ost_idx */ - - unsigned long loi_kms_valid:1; - __u64 loi_kms; /* known minimum size */ - struct ost_lvb loi_lvb; - struct osc_async_rc loi_ar; + struct ost_id loi_oi; /* object ID/Sequence on the target OST */ + int loi_ost_idx; /* OST stripe index in lov_tgt_desc->tgts */ + int loi_ost_gen; /* generation of this loi_ost_idx */ + + unsigned long loi_kms_valid:1; + __u64 loi_kms; /* known minimum size */ + struct ost_lvb loi_lvb; + struct osc_async_rc loi_ar; }; -#define loi_id loi_oi.oi_id -#define loi_seq loi_oi.oi_seq static inline void loi_kms_set(struct lov_oinfo *oinfo, __u64 kms) { @@ -102,7 +99,8 @@ static inline void loi_init(struct lov_oinfo *loi) } struct lov_stripe_md { - cfs_spinlock_t lsm_lock; + cfs_atomic_t lsm_refc; + spinlock_t lsm_lock; pid_t lsm_lock_owner; /* debugging */ /* maximum possible file size, might change as OSTs status changes, @@ -110,8 +108,7 @@ struct lov_stripe_md { __u64 lsm_maxbytes; struct { /* Public members. */ - __u64 lw_object_id; /* lov object id */ - __u64 lw_object_seq; /* lov object seq */ + struct ost_id lw_object_oi; /* lov object id/seq */ /* LOV-private members start here -- only for use in lov/. */ __u32 lw_magic; @@ -125,8 +122,7 @@ struct lov_stripe_md { struct lov_oinfo *lsm_oinfo[0]; }; -#define lsm_object_id lsm_wire.lw_object_id -#define lsm_object_seq lsm_wire.lw_object_seq +#define lsm_oi lsm_wire.lw_object_oi #define lsm_magic lsm_wire.lw_magic #define lsm_layout_gen lsm_wire.lw_layout_gen #define lsm_stripe_size lsm_wire.lw_stripe_size @@ -134,6 +130,20 @@ struct lov_stripe_md { #define lsm_stripe_count lsm_wire.lw_stripe_count #define lsm_pool_name lsm_wire.lw_pool_name +static inline bool lsm_is_released(struct lov_stripe_md *lsm) +{ + return !!(lsm->lsm_pattern & LOV_PATTERN_F_RELEASED); +} + +static inline bool lsm_has_objects(struct lov_stripe_md *lsm) +{ + if (lsm == NULL) + return false; + if (lsm_is_released(lsm)) + return false; + return true; +} + struct obd_info; typedef int (*obd_enqueue_update_f)(void *cookie, int rc); @@ -150,7 +160,7 @@ struct obd_info { - while stats, the flags used for control delay/resend. - while setattr, the flags used for distinguish punch operation */ - int oi_flags; + __u64 oi_flags; /* Lock handle specific for every OSC lock. */ struct lustre_handle *oi_lockh; /* lsm data specific for every OSC. */ @@ -205,32 +215,32 @@ static inline int lov_lum_lsm_cmp(struct lov_user_md *lum, } static inline int lov_lum_swab_if_needed(struct lov_user_md_v3 *lumv3, - int *lmm_magic, - struct lov_user_md *lum) + int *lmm_magic, + struct lov_user_md *lum) { - if (lum && cfs_copy_from_user(lumv3, lum,sizeof(struct lov_user_md_v1))) - return -EFAULT; - - *lmm_magic = lumv3->lmm_magic; - - if (*lmm_magic == __swab32(LOV_USER_MAGIC_V1)) { - lustre_swab_lov_user_md_v1((struct lov_user_md_v1 *)lumv3); - *lmm_magic = LOV_USER_MAGIC_V1; - } else if (*lmm_magic == LOV_USER_MAGIC_V3) { - if (lum && cfs_copy_from_user(lumv3, lum, sizeof(*lumv3))) - return -EFAULT; - } else if (*lmm_magic == __swab32(LOV_USER_MAGIC_V3)) { - if (lum && cfs_copy_from_user(lumv3, lum, sizeof(*lumv3))) - return -EFAULT; - lustre_swab_lov_user_md_v3(lumv3); - *lmm_magic = LOV_USER_MAGIC_V3; - } else if (*lmm_magic != LOV_USER_MAGIC_V1) { - CDEBUG(D_IOCTL, - "bad userland LOV MAGIC: %#08x != %#08x nor %#08x\n", - *lmm_magic, LOV_USER_MAGIC_V1, LOV_USER_MAGIC_V3); - return -EINVAL; - } - return 0; + if (lum && copy_from_user(lumv3, lum, sizeof(struct lov_user_md_v1))) + return -EFAULT; + + *lmm_magic = lumv3->lmm_magic; + + if (*lmm_magic == __swab32(LOV_USER_MAGIC_V1)) { + lustre_swab_lov_user_md_v1((struct lov_user_md_v1 *)lumv3); + *lmm_magic = LOV_USER_MAGIC_V1; + } else if (*lmm_magic == LOV_USER_MAGIC_V3) { + if (lum && copy_from_user(lumv3, lum, sizeof(*lumv3))) + return -EFAULT; + } else if (*lmm_magic == __swab32(LOV_USER_MAGIC_V3)) { + if (lum && copy_from_user(lumv3, lum, sizeof(*lumv3))) + return -EFAULT; + lustre_swab_lov_user_md_v3(lumv3); + *lmm_magic = LOV_USER_MAGIC_V3; + } else if (*lmm_magic != LOV_USER_MAGIC_V1) { + CDEBUG(D_IOCTL, + "bad userland LOV MAGIC: %#08x != %#08x nor %#08x\n", + *lmm_magic, LOV_USER_MAGIC_V1, LOV_USER_MAGIC_V3); + return -EINVAL; + } + return 0; } void lov_stripe_lock(struct lov_stripe_md *md); @@ -244,44 +254,14 @@ struct obd_type { char *typ_name; int typ_refcnt; struct lu_device_type *typ_lu; - cfs_spinlock_t obd_type_lock; + spinlock_t obd_type_lock; }; struct brw_page { - obd_off off; - cfs_page_t *pg; - int count; - obd_flag flag; -}; - -/* Individual type definitions */ - -struct ost_server_data; - -struct osd_properties { - size_t osd_max_ea_size; -}; - -#define OBT_MAGIC 0xBDDECEAE -/* hold common fields for "target" device */ -struct obd_device_target { - __u32 obt_magic; - __u32 obt_instance; - struct super_block *obt_sb; - /** last_rcvd file */ - struct file *obt_rcvd_filp; -#ifdef HAVE_SERVER_SUPPORT - struct lu_target *obt_lut; -#endif - __u64 obt_mount_count; - cfs_semaphore_t obt_quotachecking; - struct lustre_quota_ctxt obt_qctxt; - lustre_quota_version_t obt_qfmt; - cfs_rw_semaphore_t obt_rwsem; - struct vfsmount *obt_vfsmnt; - struct file *obt_health_check_filp; - struct osd_properties obt_osd_properties; - struct obd_job_stats obt_jobstats; + obd_off off; + struct page *pg; + int count; + obd_flag flag; }; /* llog contexts */ @@ -304,103 +284,6 @@ enum llog_ctxt_id { LLOG_MAX_CTXTS }; -#define FILTER_SUBDIR_COUNT 32 /* set to zero for no subdirs */ - -struct filter_subdirs { - cfs_dentry_t *dentry[FILTER_SUBDIR_COUNT]; -}; - - -struct filter_ext { - __u64 fe_start; - __u64 fe_end; -}; - -struct filter_obd { - /* NB this field MUST be first */ - struct obd_device_target fo_obt; - const char *fo_fstype; - - int fo_group_count; - cfs_dentry_t *fo_dentry_O; - cfs_dentry_t **fo_dentry_O_groups; - struct filter_subdirs *fo_dentry_O_sub; - cfs_mutex_t fo_init_lock; /* group initialization lock */ - int fo_committed_group; - - cfs_spinlock_t fo_objidlock; /* protect fo_lastobjid */ - - unsigned long fo_destroys_in_progress; - cfs_mutex_t fo_create_locks[FILTER_SUBDIR_COUNT]; - - cfs_list_t fo_export_list; - int fo_subdir_count; - - obd_size fo_tot_dirty; /* protected by obd_osfs_lock */ - obd_size fo_tot_granted; /* all values in bytes */ - obd_size fo_tot_pending; - int fo_tot_granted_clients; - - obd_size fo_readcache_max_filesize; - cfs_spinlock_t fo_flags_lock; - unsigned int fo_read_cache:1, /**< enable read-only cache */ - fo_writethrough_cache:1,/**< read cache writes */ - fo_mds_ost_sync:1, /**< MDS-OST orphan recovery*/ - fo_raid_degraded:1;/**< RAID device degraded */ - - struct obd_import *fo_mdc_imp; - struct obd_uuid fo_mdc_uuid; - struct lustre_handle fo_mdc_conn; - struct file **fo_last_objid_files; - __u64 *fo_last_objids; /* last created objid for groups, - * protected by fo_objidlock */ - - cfs_mutex_t fo_alloc_lock; - - cfs_atomic_t fo_r_in_flight; - cfs_atomic_t fo_w_in_flight; - - /* - * per-filter pool of kiobuf's allocated by filter_common_setup() and - * torn down by filter_cleanup(). Contains OST_NUM_THREADS elements of - * which ->fo_iobuf_count were allocated. - * - * This pool contains kiobuf used by - * filter_{prep,commit}rw_{read,write}() and is shared by all OST - * threads. - * - * Locking: none, each OST thread uses only one element, determined by - * its "ordinal number", ->t_id. - */ - struct filter_iobuf **fo_iobuf_pool; - int fo_iobuf_count; - - cfs_list_t fo_llog_list; - cfs_spinlock_t fo_llog_list_lock; - - struct brw_stats fo_filter_stats; - struct lustre_quota_ctxt fo_quota_ctxt; - cfs_spinlock_t fo_quotacheck_lock; - cfs_atomic_t fo_quotachecking; - - int fo_fmd_max_num; /* per exp filter_mod_data */ - int fo_fmd_max_age; /* jiffies to fmd expiry */ - unsigned long fo_syncjournal:1, /* sync journal on writes */ - fo_sync_lock_cancel:2;/* sync on lock cancel */ - - - /* sptlrpc stuff */ - cfs_rwlock_t fo_sptlrpc_lock; - struct sptlrpc_rule_set fo_sptlrpc_rset; - - /* capability related */ - unsigned int fo_fl_oss_capa; - cfs_list_t fo_capa_keys; - cfs_hlist_head_t *fo_capa_hash; - struct llog_commit_master *fo_lcm; - int fo_sec_level; -}; - struct timeout_item { enum timeout_event ti_event; cfs_time_t ti_timeout; @@ -411,6 +294,7 @@ struct timeout_item { }; #define OSC_MAX_RIF_DEFAULT 8 +#define MDS_OSC_MAX_RIF_DEFAULT 50 #define OSC_MAX_RIF_MAX 256 #define OSC_MAX_DIRTY_DEFAULT (OSC_MAX_RIF_DEFAULT * 4) #define OSC_MAX_DIRTY_MB_MAX 2048 /* arbitrary, but < MAX_LONG bytes */ @@ -430,7 +314,7 @@ enum { struct mdc_rpc_lock; struct obd_import; struct client_obd { - cfs_rw_semaphore_t cl_sem; + struct rw_semaphore cl_sem; struct obd_uuid cl_target_uuid; struct obd_import *cl_import; /* ptlrpc connection state */ int cl_conn_count; @@ -444,19 +328,28 @@ struct client_obd { enum lustre_sec_part cl_sp_to; struct sptlrpc_flavor cl_flvr_mgc; /* fixed flavor of mgc->mgs */ - //struct llog_canceld_ctxt *cl_llcd; /* it's included by obd_llog_ctxt */ - void *cl_llcd_offset; - /* 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_dirty_transit; /* dirty synchronous */ long cl_avail_grant; /* bytes of credit for ost */ long cl_lost_grant; /* lost credits (trunc) */ - cfs_list_t cl_cache_waiters; /* waiting for cache/grant */ - cfs_time_t cl_next_shrink_grant; /* jiffies */ - cfs_list_t cl_grant_shrink_list; /* Timeout event list */ - int cl_grant_shrink_interval; /* seconds */ + + /* since we allocate grant by blocks, we don't know how many grant will + * be used to add a page into cache. As a solution, we reserve maximum + * grant before trying to dirty a page and unreserve the rest. + * See osc_{reserve|unreserve}_grant for details. */ + long cl_reserved_grant; + cfs_list_t cl_cache_waiters; /* waiting for cache/grant */ + cfs_time_t cl_next_shrink_grant; /* jiffies */ + cfs_list_t cl_grant_shrink_list; /* Timeout event list */ + int cl_grant_shrink_interval; /* seconds */ + + /* A chunk is an optimal size used by osc_extent to determine + * the extent size. A chunk is max(PAGE_CACHE_SIZE, OST block size) */ + int cl_chunkbits; + int cl_chunk; + int cl_extent_tax; /* extent overhead, by bytes */ /* keep track of objects that have lois that contain pages which * have been queued for async brw. this lock also protects the @@ -478,7 +371,7 @@ struct client_obd { * * NB by Jinshan: though field names are still _loi_, but actually * osc_object{}s are in the list. - */ + */ client_obd_lock_t cl_loi_list_lock; cfs_list_t cl_loi_ready_list; cfs_list_t cl_loi_hp_ready_list; @@ -487,9 +380,9 @@ struct client_obd { 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; - int cl_max_pages_per_rpc; + cfs_atomic_t cl_pending_w_pages; + cfs_atomic_t cl_pending_r_pages; + __u32 cl_max_pages_per_rpc; int cl_max_rpcs_in_flight; struct obd_histogram cl_read_rpc_hist; struct obd_histogram cl_write_rpc_hist; @@ -498,16 +391,25 @@ struct client_obd { struct obd_histogram cl_read_offset_hist; struct obd_histogram cl_write_offset_hist; + /* lru for osc caching pages */ + struct cl_client_cache *cl_cache; + cfs_list_t cl_lru_osc; /* member of cl_cache->ccc_lru */ + cfs_atomic_t *cl_lru_left; + cfs_atomic_t cl_lru_busy; + cfs_atomic_t cl_lru_shrinkers; + cfs_atomic_t cl_lru_in_list; + cfs_list_t cl_lru_list; /* lru page list */ + client_obd_lock_t cl_lru_list_lock; /* page list protector */ + /* number of in flight destroy rpcs is limited to max_rpcs_in_flight */ cfs_atomic_t cl_destroy_in_flight; cfs_waitq_t cl_destroy_waitq; struct mdc_rpc_lock *cl_rpc_lock; struct mdc_rpc_lock *cl_close_lock; - struct osc_creator cl_oscc; /* mgc datastruct */ - cfs_semaphore_t cl_mgc_sem; + struct semaphore cl_mgc_sem; struct vfsmount *cl_mgc_vfsmnt; struct dentry *cl_mgc_configs_dir; cfs_atomic_t cl_mgc_refcount; @@ -523,8 +425,12 @@ struct client_obd { /* also protected by the poorly named _loi_list_lock lock above */ struct osc_async_rc cl_ar; - /* used by quotacheck */ - int cl_qchk_stat; /* quotacheck stat of the peer */ + /* used by quotacheck when the servers are older than 2.4 */ + int cl_qchk_stat; /* quotacheck stat of the peer */ +#define CL_NOT_QUOTACHECKED 1 /* client->cl_qchk_stat init value */ +#if LUSTRE_VERSION_CODE >= OBD_OCD_VERSION(2, 7, 50, 0) +#warning "please consider removing quotacheck compatibility code" +#endif /* sequence manager */ struct lu_client_seq *cl_seq; @@ -533,113 +439,19 @@ struct client_obd { /* ptlrpc work for writeback in ptlrpcd context */ void *cl_writeback_work; + /* hash tables for osc_quota_info */ + cfs_hash_t *cl_quota_hash[MAXQUOTAS]; }; #define obd2cli_tgt(obd) ((char *)(obd)->u.cli.cl_target_uuid.uuid) -#define CL_NOT_QUOTACHECKED 1 /* client->cl_qchk_stat init value */ - -struct mgs_obd { - struct obd_device_target mgs_obt; - struct ptlrpc_service *mgs_service; - struct vfsmount *mgs_vfsmnt; - struct super_block *mgs_sb; - struct dentry *mgs_configs_dir; - cfs_list_t mgs_fs_db_list; - cfs_mutex_t mgs_mutex; - cfs_proc_dir_entry_t *mgs_proc_live; - cfs_time_t mgs_start_time; -}; - -struct mds_obd { - /* NB this field MUST be first */ - struct obd_device_target mds_obt; - struct ptlrpc_service *mds_service; - struct ptlrpc_service *mds_setattr_service; - struct ptlrpc_service *mds_readpage_service; - int mds_max_mdsize; - int mds_max_cookiesize; - __u64 mds_io_epoch; - unsigned long mds_atime_diff; - struct ll_fid mds_rootfid; - cfs_dentry_t *mds_pending_dir; - cfs_dentry_t *mds_logs_dir; - cfs_dentry_t *mds_objects_dir; - struct llog_handle *mds_cfg_llh; - struct obd_device *mds_lov_obd; - struct obd_uuid mds_lov_uuid; - char *mds_profile; - struct obd_export *mds_lov_exp; - struct lov_desc mds_lov_desc; - __u32 mds_id; - - /* mark pages dirty for write. */ - cfs_bitmap_t *mds_lov_page_dirty; - /* array for store pages with obd_id */ - void **mds_lov_page_array; - /* file for store objid */ - struct file *mds_lov_objid_filp; - __u32 mds_lov_objid_count; - __u32 mds_lov_objid_max_index; - __u32 mds_lov_objid_lastpage; - __u32 mds_lov_objid_lastidx; - - - struct lustre_quota_info mds_quota_info; - cfs_rw_semaphore_t mds_qonoff_sem; - unsigned long mds_fl_user_xattr:1, - mds_fl_acl:1, - mds_evict_ost_nids:1, - mds_fl_cfglog:1, - mds_fl_synced:1, - mds_quota:1, - mds_fl_target:1; /* mds have one or - * more targets */ - - struct upcall_cache *mds_identity_cache; - - /* for capability keys update */ - struct lustre_capa_key *mds_capa_keys; - cfs_rw_semaphore_t mds_notify_lock; -}; - -/* lov objid */ -extern __u32 mds_max_ost_index; - -#define MDS_LOV_ALLOC_SIZE (CFS_PAGE_SIZE) - -#define OBJID_PER_PAGE() (MDS_LOV_ALLOC_SIZE / sizeof(obd_id)) - -#define MDS_LOV_OBJID_PAGES_COUNT (mds_max_ost_index/OBJID_PER_PAGE()) - -extern int mds_lov_init_objids(struct obd_device *obd); -extern void mds_lov_destroy_objids(struct obd_device *obd); - struct obd_id_info { __u32 idx; obd_id *data; }; -/* */ - -struct echo_obd { - struct obd_device_target eo_obt; - struct obdo eo_oa; - cfs_spinlock_t eo_lock; - __u64 eo_lastino; - struct lustre_handle eo_nl_lock; - cfs_atomic_t eo_prep; -}; - -struct ost_obd { - struct ptlrpc_service *ost_service; - struct ptlrpc_service *ost_create_service; - struct ptlrpc_service *ost_io_service; - cfs_mutex_t ost_health_mutex; -}; - struct echo_client_obd { - struct obd_export *ec_exp; /* the local connection to osc/lov */ - cfs_spinlock_t ec_lock; + struct obd_export *ec_exp; /* the local connection to osc/lov */ + spinlock_t ec_lock; cfs_list_t ec_objects; cfs_list_t ec_locks; int ec_nstripes; @@ -671,7 +483,7 @@ struct ost_pool { lov_obd->lov_tgts */ unsigned int op_count; /* number of OSTs in the array */ unsigned int op_size; /* allocated size of lp_array */ - cfs_rw_semaphore_t op_rw_sem; /* to protect ost_pool use */ + struct rw_semaphore op_rw_sem; /* to protect ost_pool use */ }; /* Round-robin allocator data */ @@ -693,7 +505,7 @@ struct lov_statfs_data { /* Stripe placement optimization */ struct lov_qos { cfs_list_t lq_oss_list; /* list of OSSs that targets use */ - cfs_rw_semaphore_t lq_rw_sem; + struct rw_semaphore lq_rw_sem; __u32 lq_active_oss_count; unsigned int lq_prio_free; /* priority for free space */ unsigned int lq_threshold_rr;/* priority for rr */ @@ -728,7 +540,6 @@ struct lov_tgt_desc { #define pool_tgt_count(_p) _p->pool_obds.op_count #define pool_tgt_array(_p) _p->pool_obds.op_array #define pool_tgt_rw_sem(_p) _p->pool_obds.op_rw_sem -#define pool_tgt(_p, _i) _p->pool_lov->lov_tgts[_p->pool_obds.op_array[_i]] struct pool_desc { char pool_name[LOV_MAXPOOLNAME + 1]; /* name of pool */ @@ -738,8 +549,8 @@ struct pool_desc { cfs_hlist_node_t pool_hash; /* access by poolname */ cfs_list_t pool_list; /* serial access */ cfs_proc_dir_entry_t *pool_proc_entry; /* file in /proc */ - struct lov_obd *pool_lov; /* lov obd to which this - pool belong */ + struct obd_device *pool_lobd; /* obd of the lov/lod to which + * this pool belongs */ }; struct lov_obd { @@ -747,9 +558,8 @@ struct lov_obd { struct lov_tgt_desc **lov_tgts; /* sparse array */ struct ost_pool lov_packed; /* all OSTs in a packed array */ - cfs_mutex_t lov_lock; + struct mutex lov_lock; struct obd_connect_data lov_ocd; - struct lov_qos lov_qos; /* qos info per lov */ cfs_atomic_t lov_refcount; __u32 lov_tgt_count; /* how many OBD's */ __u32 lov_active_tgt_count; /* how many active */ @@ -761,14 +571,19 @@ struct lov_obd { cfs_list_t lov_pool_list; /* used for sequential access */ cfs_proc_dir_entry_t *lov_pool_proc_entry; enum lustre_sec_part lov_sp_me; + + /* Cached LRU pages from upper layer */ + void *lov_cache; + + struct rw_semaphore lov_notify_lock; }; struct lmv_tgt_desc { - struct obd_uuid ltd_uuid; - struct obd_export *ltd_exp; - int ltd_active; /* is this target up for requests */ - int ltd_idx; - cfs_mutex_t ltd_fid_mutex; + struct obd_uuid ltd_uuid; + struct obd_export *ltd_exp; + int ltd_idx; + struct mutex ltd_fid_mutex; + unsigned long ltd_active:1; /* target up for requests */ }; enum placement_policy { @@ -781,51 +596,50 @@ enum placement_policy { typedef enum placement_policy placement_policy_t; struct lmv_obd { - int refcount; - struct lu_client_fld lmv_fld; - cfs_spinlock_t lmv_lock; - placement_policy_t lmv_placement; - struct lmv_desc desc; - struct obd_uuid cluuid; - struct obd_export *exp; - - int connected; - int max_easize; - int max_def_easize; - int max_cookiesize; - int server_timeout; - cfs_mutex_t init_mutex; - - struct lmv_tgt_desc *tgts; - int tgts_size; - - struct obd_connect_data *datas; - int datas_size; - - struct obd_connect_data conn_data; + int refcount; + struct lu_client_fld lmv_fld; + spinlock_t lmv_lock; + placement_policy_t lmv_placement; + struct lmv_desc desc; + struct obd_uuid cluuid; + struct obd_export *exp; + + struct mutex init_mutex; + int connected; + int max_easize; + int max_def_easize; + int max_cookiesize; + int server_timeout; + + int tgts_size; /* size of tgts array */ + struct lmv_tgt_desc **tgts; + + struct obd_connect_data conn_data; }; struct niobuf_local { - __u64 lnb_file_offset; - __u64 offset; - __u32 len; - __u32 flags; - cfs_page_t *page; - cfs_dentry_t *dentry; - int lnb_grant_used; - int rc; + __u64 lnb_file_offset; + __u32 lnb_page_offset; + __u32 len; + __u32 flags; + struct page *page; + struct dentry *dentry; + int lnb_grant_used; + int rc; }; #define LUSTRE_FLD_NAME "fld" #define LUSTRE_SEQ_NAME "seq" -#define LUSTRE_CMM_NAME "cmm" #define LUSTRE_MDD_NAME "mdd" -#define LUSTRE_OSD_NAME "osd-ldiskfs" +#define LUSTRE_OSD_LDISKFS_NAME "osd-ldiskfs" +#define LUSTRE_OSD_ZFS_NAME "osd-zfs" #define LUSTRE_VVP_NAME "vvp" #define LUSTRE_LMV_NAME "lmv" -#define LUSTRE_CMM_MDC_NAME "cmm-mdc" #define LUSTRE_SLP_NAME "slp" +#define LUSTRE_LOD_NAME "lod" +#define LUSTRE_OSP_NAME "osp" +#define LUSTRE_LWP_NAME "lwp" /* obd device type names */ /* FIXME all the references to LUSTRE_MDS_NAME should be swapped with LUSTRE_MDT_NAME */ @@ -839,9 +653,9 @@ struct niobuf_local { #define LUSTRE_MGS_NAME "mgs" #define LUSTRE_MGC_NAME "mgc" -#define LUSTRE_CACHEOBD_NAME "cobd" #define LUSTRE_ECHO_NAME "obdecho" #define LUSTRE_ECHO_CLIENT_NAME "echo_client" +#define LUSTRE_QMT_NAME "qmt" /* Constant obd names (post-rename) */ #define LUSTRE_MDS_OBDNAME "MDS" @@ -849,6 +663,35 @@ struct niobuf_local { #define LUSTRE_MGS_OBDNAME "MGS" #define LUSTRE_MGC_OBDNAME "MGC" +static inline int is_osp_on_mdt(char *name) +{ + char *ptr; + + ptr = strrchr(name, '-'); + if (ptr == NULL) { + CERROR("%s is not a obdname\n", name); + return 0; + } + + /* 1.8 OSC/OSP name on MDT is fsname-OSTxxxx-osc */ + if (strncmp(ptr + 1, "osc", 3) == 0) + return 1; + + if (strncmp(ptr + 1, "MDT", 3) != 0) + return 0; + + while (*(--ptr) != '-' && ptr != name); + + if (ptr == name) + return 0; + + if (strncmp(ptr + 1, LUSTRE_OSP_NAME, strlen(LUSTRE_OSP_NAME)) != 0 && + strncmp(ptr + 1, LUSTRE_OSC_NAME, strlen(LUSTRE_OSC_NAME)) != 0) + return 0; + + return 1; +} + /* Don't conflict with on-wire flags OBD_BRW_WRITE, etc */ #define N_LOCAL_TEMP_PAGE 0x10000000 @@ -864,6 +707,8 @@ struct obd_trans_info { struct llog_cookie oti_onecookie; struct llog_cookie *oti_logcookies; int oti_numcookies; + /** synchronous write is needed */ + unsigned long oti_sync_write:1; /* initial thread handling transaction */ struct ptlrpc_thread * oti_thread; @@ -952,8 +797,6 @@ enum obd_notify_event { OBD_NOTIFY_SYNC, /* Configuration event */ OBD_NOTIFY_CONFIG, - /* Trigger quota recovery */ - OBD_NOTIFY_QUOTA, /* Administratively deactivate/activate event */ OBD_NOTIFY_DEACTIVATE, OBD_NOTIFY_ACTIVATE @@ -978,52 +821,18 @@ struct obd_notify_upcall { }; struct target_recovery_data { - svc_handler_t trd_recovery_handler; - pid_t trd_processing_task; - cfs_completion_t trd_starting; - cfs_completion_t trd_finishing; + svc_handler_t trd_recovery_handler; + pid_t trd_processing_task; + struct completion trd_starting; + struct completion trd_finishing; }; -/** - * In HEAD for CMD, the object is created in group number which is 3>= - * or indexing starts from 3. To test this assertions are added to disallow - * group 0. But to run 2.0 mds server on 1.8.x disk format (i.e. interop_mode) - * object in group 0 needs to be allowed. - * So for interop mode following changes needs to be done: - * 1. No need to assert on group 0 or allow group 0 - * 2. The group number indexing starts from 0 instead of 3 - */ - -#define LASSERT_SEQ_IS_MDT(seq) LASSERT(fid_seq_is_mdt(seq)) - -static inline __u64 objseq_to_mdsno(obd_seq seq) -{ - LASSERT_SEQ_IS_MDT(seq); - if (seq == FID_SEQ_OST_MDT0) - return 0; - return seq - FID_SEQ_OST_MDT1 + 1; -} - -static inline int mdt_to_obd_objseq(int mdtid) -{ - /** - * MDS0 uses seq 0 pre FID-on-OST, other MDSes will use seq from - * FID_SEQ_OST_MDT1 - */ - if (mdtid) - return FID_SEQ_OST_MDT1 + mdtid - 1; - return 0; -} - struct obd_llog_group { - cfs_list_t olg_list; int olg_seq; struct llog_ctxt *olg_ctxts[LLOG_MAX_CTXTS]; cfs_waitq_t olg_waitq; - cfs_spinlock_t olg_lock; - struct obd_export *olg_exp; - int olg_initializing; - cfs_mutex_t olg_cat_processing; + spinlock_t olg_lock; + struct mutex olg_cat_processing; }; /* corresponds to one of the obd's */ @@ -1076,21 +885,21 @@ struct obd_device { cfs_list_t obd_unlinked_exports; cfs_list_t obd_delayed_exports; int obd_num_exports; - cfs_spinlock_t obd_nid_lock; - struct ldlm_namespace *obd_namespace; - struct ptlrpc_client obd_ldlm_client; /* XXX OST/MDS only */ - /* a spinlock is OK for what we do now, may need a semaphore later */ - cfs_spinlock_t obd_dev_lock; /* protects obd bitfield above */ - cfs_mutex_t obd_dev_mutex; - __u64 obd_last_committed; - struct fsfilt_operations *obd_fsops; - cfs_spinlock_t obd_osfs_lock; - struct obd_statfs obd_osfs; /* locked by obd_osfs_lock */ - __u64 obd_osfs_age; - struct lvfs_run_ctxt obd_lvfs_ctxt; - struct obd_llog_group obd_olg; /* default llog group */ - struct obd_device *obd_observer; - cfs_rw_semaphore_t obd_observer_link_sem; + spinlock_t obd_nid_lock; + struct ldlm_namespace *obd_namespace; + struct ptlrpc_client obd_ldlm_client; /* XXX OST/MDS only */ + /* a spinlock is OK for what we do now, may need a semaphore later */ + spinlock_t obd_dev_lock; /* protect OBD bitfield above */ + struct mutex obd_dev_mutex; + __u64 obd_last_committed; + struct fsfilt_operations *obd_fsops; + spinlock_t obd_osfs_lock; + struct obd_statfs obd_osfs; /* locked by obd_osfs_lock */ + __u64 obd_osfs_age; + struct lvfs_run_ctxt obd_lvfs_ctxt; + struct obd_llog_group obd_olg; /* default llog group */ + struct obd_device *obd_observer; + struct rw_semaphore obd_observer_link_sem; struct obd_notify_upcall obd_upcall; struct obd_export *obd_self_export; /* list of exports in LRU order, for ping evictor, with obd_dev_lock */ @@ -1103,7 +912,7 @@ struct obd_device { int obd_delayed_clients; /* this lock protects all recovery list_heads, timer and * obd_next_recovery_transno value */ - cfs_spinlock_t obd_recovery_task_lock; + spinlock_t obd_recovery_task_lock; __u64 obd_next_recovery_transno; int obd_replayed_requests; int obd_requests_queued_for_recovery; @@ -1127,21 +936,21 @@ struct obd_device { cfs_list_t obd_final_req_queue; int obd_recovery_stage; - union { - struct obd_device_target obt; - 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 lov_obd lov; - struct lmv_obd lmv; - struct mgs_obd mgs; - } u; - /* Fields used by LProcFS */ - unsigned int obd_cntr_base; - struct lprocfs_stats *obd_stats; + union { +#ifdef HAVE_SERVER_SUPPORT + struct obd_device_target obt; + struct filter_obd filter; + struct ost_obd ost; + struct echo_obd echo; +#endif + struct client_obd cli; + struct echo_client_obd echo_client; + struct lov_obd lov; + struct lmv_obd lmv; + } u; + /* Fields used by LProcFS */ + unsigned int obd_cntr_base; + struct lprocfs_stats *obd_stats; unsigned int md_cntr_base; struct lprocfs_stats *md_stats; @@ -1157,7 +966,7 @@ struct obd_device { /** * Ldlm pool part. Save last calculated SLV and Limit. */ - cfs_rwlock_t obd_pool_lock; + rwlock_t obd_pool_lock; int obd_pool_limit; __u64 obd_pool_slv; @@ -1166,9 +975,12 @@ struct obd_device { * debugging. */ struct lu_ref obd_reference; + + int obd_conn_inprogress; }; #define OBD_LLOG_FL_SENDNOW 0x0001 +#define OBD_LLOG_FL_EXIT 0x0002 enum obd_cleanup_stage { /* Special case hack for MDS LOVs */ @@ -1196,6 +1008,7 @@ enum obd_cleanup_stage { #define KEY_INIT_RECOV "initial_recov" #define KEY_INTERMDS "inter_mds" #define KEY_LAST_ID "last_id" +#define KEY_LAST_FID "last_fid" #define KEY_LOCK_TO_STRIPE "lock_to_stripe" #define KEY_LOVDESC "lovdesc" #define KEY_LOV_IDX "lov_idx" @@ -1205,7 +1018,6 @@ enum obd_cleanup_stage { #define KEY_NEXT_ID "next_id" #define KEY_READ_ONLY "read-only" #define KEY_REGISTER_TARGET "register_target" -#define KEY_REVIMP_UPD "revimp_update" #define KEY_SET_FS "set_fs" #define KEY_TGT_COUNT "tgt_count" /* KEY_SET_INFO in lustre_idl.h */ @@ -1213,6 +1025,10 @@ enum obd_cleanup_stage { #define KEY_CONNECT_FLAG "connect_flags" #define KEY_SYNC_LOCK_CANCEL "sync_lock_cancel" +#define KEY_CACHE_SET "cache_set" +#define KEY_CACHE_LRU_SHRINK "cache_lru_shrink" +#define KEY_CHANGELOG_INDEX "changelog_index" + struct lu_context; /* /!\ must be coherent with include/linux/namei.h on patched kernel */ @@ -1227,6 +1043,8 @@ struct lu_context; #define IT_EXEC (1 << 8) #define IT_PIN (1 << 9) #define IT_LAYOUT (1 << 10) +#define IT_QUOTA_DQACQ (1 << 11) +#define IT_QUOTA_CONN (1 << 12) static inline int it_to_lock_mode(struct lookup_intent *it) { @@ -1263,30 +1081,39 @@ struct md_op_data { /* iattr fields and blocks. */ struct iattr op_attr; #ifdef __KERNEL__ -#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,14) - unsigned int op_attr_flags; -#endif + unsigned int op_attr_flags; #endif __u64 op_valid; loff_t op_attr_blocks; /* Size-on-MDS epoch and flags. */ __u64 op_ioepoch; - __u32 op_flags; + __u32 op_flags; /* Capa fields */ struct obd_capa *op_capa1; struct obd_capa *op_capa2; /* Various operation flags. */ - __u32 op_bias; + __u32 op_bias; /* Operation type */ - __u32 op_opc; + __u32 op_opc; /* Used by readdir */ - __u32 op_npages; - __u64 op_offset; + __u64 op_offset; + + /* Used by readdir */ + __u32 op_npages; + + /* used to transfer info between the stacks of MD client + * see enum op_cli_flags */ + __u32 op_cli_flags; +}; + +enum op_cli_flags { + CLI_SET_MEA = 1 << 0, + CLI_RM_ENTRY = 1 << 1, }; struct md_enqueue_info; @@ -1343,9 +1170,10 @@ struct obd_ops { void *localdata); int (*o_disconnect)(struct obd_export *exp); - /* Initialize/finalize fids infrastructure. */ - int (*o_fid_init)(struct obd_export *exp); - int (*o_fid_fini)(struct obd_export *exp); + /* Initialize/finalize fids infrastructure. */ + int (*o_fid_init)(struct obd_device *obd, + struct obd_export *exp, enum lu_cli_type type); + int (*o_fid_fini)(struct obd_device *obd); /* Allocate new fid according to passed @hint. */ int (*o_fid_alloc)(struct obd_export *exp, struct lu_fid *fid, @@ -1355,8 +1183,6 @@ struct obd_ops { * Object with @fid is getting deleted, we may want to do something * about this. */ - int (*o_fid_delete)(struct obd_export *exp, const struct lu_fid *fid); - int (*o_statfs)(const struct lu_env *, struct obd_export *exp, struct obd_statfs *osfs, __u64 max_age, __u32 flags); int (*o_statfs_async)(struct obd_export *exp, struct obd_info *oinfo, @@ -1462,11 +1288,6 @@ struct obd_ops { struct obd_quotactl *); int (*o_quotactl)(struct obd_device *, struct obd_export *, struct obd_quotactl *); - int (*o_quota_adjust_qunit)(struct obd_export *exp, - struct quota_adjust_qunit *oqaq, - struct lustre_quota_ctxt *qctxt, - struct ptlrpc_request_set *rqset); - int (*o_ping)(const struct lu_env *, struct obd_export *exp); @@ -1526,8 +1347,7 @@ struct lookup_intent; struct md_ops { int (*m_getstatus)(struct obd_export *, struct lu_fid *, struct obd_capa **); - int (*m_change_cbdata)(struct obd_export *, const struct lu_fid *, - ldlm_iterator_t, void *); + int (*m_null_inode)(struct obd_export *, const struct lu_fid *); int (*m_find_cbdata)(struct obd_export *, const struct lu_fid *, ldlm_iterator_t, void *); int (*m_close)(struct obd_export *, struct md_op_data *, @@ -1540,7 +1360,7 @@ struct md_ops { int (*m_enqueue)(struct obd_export *, struct ldlm_enqueue_info *, struct lookup_intent *, struct md_op_data *, struct lustre_handle *, void *, int, - struct ptlrpc_request **, int); + struct ptlrpc_request **, __u64); int (*m_getattr)(struct obd_export *, struct md_op_data *, struct ptlrpc_request **); int (*m_getattr_name)(struct obd_export *, struct md_op_data *, @@ -1548,7 +1368,7 @@ struct md_ops { int (*m_intent_lock)(struct obd_export *, struct md_op_data *, void *, int, struct lookup_intent *, int, struct ptlrpc_request **, - ldlm_blocking_callback, int); + ldlm_blocking_callback, __u64); int (*m_link)(struct obd_export *, struct md_op_data *, struct ptlrpc_request **); int (*m_rename)(struct obd_export *, struct md_op_data *, @@ -1593,7 +1413,7 @@ struct md_ops { struct obd_client_handle *); int (*m_set_lock_data)(struct obd_export *, __u64 *, void *, __u64 *); - ldlm_mode_t (*m_lock_match)(struct obd_export *, int, + ldlm_mode_t (*m_lock_match)(struct obd_export *, __u64, const struct lu_fid *, ldlm_type_t, ldlm_policy_data_t *, ldlm_mode_t, struct lustre_handle *); @@ -1653,45 +1473,10 @@ static inline const struct lsm_operations *lsm_op_find(int magic) } } -int lvfs_check_io_health(struct obd_device *obd, struct file *file); - /* Requests for obd_extent_calc() */ #define OBD_CALC_STRIPE_START 1 #define OBD_CALC_STRIPE_END 2 -static inline void obd_transno_commit_cb(struct obd_device *obd, __u64 transno, - struct obd_export *exp, int error) -{ - if (error) { - CERROR("%s: transno "LPU64" commit error: %d\n", - obd->obd_name, transno, error); - return; - } - if (exp && transno > exp->exp_last_committed) { - CDEBUG(D_HA, "%s: transno "LPU64" committed\n", - obd->obd_name, transno); - exp->exp_last_committed = transno; - ptlrpc_commit_replies(exp); - } else { - CDEBUG(D_INFO, "%s: transno "LPU64" committed\n", - obd->obd_name, transno); - } - if (transno > obd->obd_last_committed) - obd->obd_last_committed = transno; -} - -static inline void init_obd_quota_ops(quota_interface_t *interface, - struct obd_ops *obd_ops) -{ - if (!interface) - return; - - LASSERT(obd_ops); - obd_ops->o_quotacheck = QUOTA_OP(interface, check); - obd_ops->o_quotactl = QUOTA_OP(interface, ctl); - obd_ops->o_quota_adjust_qunit = QUOTA_OP(interface, adjust_qunit); -} - static inline struct lustre_capa *oinfo_capa(struct obd_info *oinfo) { return oinfo->oi_capa; @@ -1708,13 +1493,13 @@ static inline struct md_open_data *obd_mod_alloc(void) } #define obd_mod_get(mod) cfs_atomic_inc(&(mod)->mod_refcount) -#define obd_mod_put(mod) \ -({ \ - if (cfs_atomic_dec_and_test(&(mod)->mod_refcount)) { \ - if ((mod)->mod_open_req) \ - ptlrpc_req_finished((mod)->mod_open_req); \ - OBD_FREE_PTR(mod); \ - } \ +#define obd_mod_put(mod) \ +({ \ + if (cfs_atomic_dec_and_test(&(mod)->mod_refcount)) { \ + if ((mod)->mod_open_req) \ + ptlrpc_req_finished((mod)->mod_open_req); \ + OBD_FREE_PTR(mod); \ + } \ }) void obdo_from_inode(struct obdo *dst, struct inode *src, obd_flag valid); @@ -1739,4 +1524,52 @@ static inline const char *lu_dev_name(const struct lu_device *lu_dev) return lu_dev->ld_obd->obd_name; } +static inline bool filename_is_volatile(const char *name, int namelen, int *idx) +{ + const char *start; + char *end; + + if (strncmp(name, LUSTRE_VOLATILE_HDR, LUSTRE_VOLATILE_HDR_LEN) != 0) + return false; + + /* caller does not care of idx */ + if (idx == NULL) + return true; + + /* volatile file, the MDT can be set from name */ + /* name format is LUSTRE_VOLATILE_HDR:[idx]: */ + /* if no MDT is specified, use std way */ + if (namelen < LUSTRE_VOLATILE_HDR_LEN + 2) + goto bad_format; + /* test for no MDT idx case */ + if ((*(name + LUSTRE_VOLATILE_HDR_LEN) == ':') && + (*(name + LUSTRE_VOLATILE_HDR_LEN + 1) == ':')) { + *idx = -1; + return true; + } + /* we have an idx, read it */ + start = name + LUSTRE_VOLATILE_HDR_LEN + 1; + *idx = strtoul(start, &end, 0); + /* error cases: + * no digit, no trailing :, negative value + */ + if (((*idx == 0) && (end == start)) || + (*end != ':') || (*idx < 0)) + goto bad_format; + + return true; +bad_format: + /* bad format of mdt idx, we cannot return an error + * to caller so we use hash algo */ + CERROR("Bad volatile file name format: %s\n", + name + LUSTRE_VOLATILE_HDR_LEN); + return false; +} + +static inline int cli_brw_size(struct obd_device *obd) +{ + LASSERT(obd != NULL); + return obd->u.cli.cl_max_pages_per_rpc << PAGE_CACHE_SHIFT; +} + #endif /* __OBD_H */