X-Git-Url: https://git.whamcloud.com/?p=fs%2Flustre-release.git;a=blobdiff_plain;f=lustre%2Fosd-zfs%2Fosd_internal.h;h=c860762ad30c908628e4c2cab47c88e0446d7cfa;hp=f94e6aaa5541659e04603fd72c45b697f3fd0d52;hb=1a7720934dfb3105afd2f025c953bea2167d4e5d;hpb=e95e4d3227fdf9777b58600b39173d1c6bde61bb diff --git a/lustre/osd-zfs/osd_internal.h b/lustre/osd-zfs/osd_internal.h index f94e6aa..c860762 100644 --- a/lustre/osd-zfs/osd_internal.h +++ b/lustre/osd-zfs/osd_internal.h @@ -23,7 +23,7 @@ * Copyright (c) 2009, 2010, Oracle and/or its affiliates. All rights reserved. * Use is subject to license terms. * - * Copyright (c) 2012, 2016, Intel Corporation. + * Copyright (c) 2012, 2017, Intel Corporation. */ /* * This file is part of Lustre, http://www.lustre.org/ @@ -43,6 +43,8 @@ #include #include #include +#include +#include #ifdef SHRINK_STOP #undef SHRINK_STOP #endif @@ -52,6 +54,7 @@ #include #include #include +#include /** * By design including kmem.h overrides the Linux slab interfaces to provide @@ -94,6 +97,16 @@ #define OSD_GRANT_FOR_LOCAL_OIDS (2ULL << 20) /* 2MB for last_rcvd, ... */ +#define OSD_MAX_CACHE_SIZE OBD_OBJECT_EOF + +#ifndef HAVE_ZFS_REFCOUNT_ADD +#define zfs_refcount_add refcount_add +#endif + +extern struct dt_body_operations osd_body_scrub_ops; +extern struct dt_body_operations osd_body_ops; +extern struct kmem_cache *osd_object_kmem; + /** * Iterator's in-memory data structure for quota file. */ @@ -108,6 +121,13 @@ struct osd_it_quota { unsigned oiq_reset:1; /* 1 -- no need to advance */ }; +enum osd_zap_pos { + OZI_POS_INIT = 0, + OZI_POS_DOT = 1, /* cursor at . */ + OZI_POS_DOTDOT = 2, /* cursor at .. */ + OZI_POS_REAL = 3, /* cursor at real entries */ +}; + /** * Iterator's in-memory data structure for ZAPs * @@ -119,12 +139,8 @@ struct osd_zap_it { zap_cursor_t *ozi_zc; struct osd_object *ozi_obj; unsigned ozi_reset:1; /* 1 -- no need to advance */ - /* ozi_pos - position of the cursor: - * 0 - before any record - * 1 - "." - * 2 - ".." - * 3 - real records */ - unsigned ozi_pos:3; + /* ozi_pos - position of the cursor */ + enum osd_zap_pos ozi_pos; union { char ozi_name[MAXNAMELEN]; /* file name for dir */ __u64 ozi_key; /* binary key for index files */ @@ -179,6 +195,38 @@ struct osd_idmap_cache { oic_remote:1; /* FID isn't local */ }; +struct osd_inconsistent_item { + /* link into lustre_scrub::os_inconsistent_items, + * protected by lustr_scrub::os_lock. */ + struct list_head oii_list; + + /* The right FID <=> oid mapping. */ + struct osd_idmap_cache oii_cache; + + unsigned int oii_insert:1; /* insert or update mapping. */ +}; + +struct osd_otable_it { + struct osd_device *ooi_dev; + struct lu_fid ooi_fid; + __u64 ooi_pos; + __u64 ooi_prefetched_dnode; + int ooi_prefetched; + + /* The following bits can be updated/checked w/o lock protection. + * If more bits will be introduced in the future and need lock to + * protect, please add comment. */ + unsigned int ooi_used_outside:1, /* Some user out of OSD + * uses the iteration. */ + ooi_all_cached:1, /* No more entries can be + * filled into cache. */ + ooi_user_ready:1, /* The user out of OSD is + * ready to iterate. */ + ooi_waiting:1; /* it::next is waiting. */ +}; + +extern const struct dt_index_operations osd_otable_ops; + /* max.number of regular attributes the callers may ask for */ # define OSD_MAX_IN_BULK (sizeof(struct osa_attr)/sizeof(uint64_t)) @@ -186,10 +234,6 @@ struct osd_thread_info { const struct lu_env *oti_env; struct lu_fid oti_fid; - /* - * XXX temporary: for ->i_op calls. - */ - struct timespec oti_time; struct ost_id oti_ostid; @@ -207,6 +251,7 @@ struct osd_thread_info { struct lu_attr oti_la; struct osa_attr oti_osa; zap_attribute_t oti_za; + zap_attribute_t oti_za2; dmu_object_info_t oti_doi; struct luz_direntry oti_zde; @@ -217,7 +262,11 @@ struct osd_thread_info { struct osd_idmap_cache *oti_ins_cache; int oti_ins_cache_size; int oti_ins_cache_used; + /* inc by osd_trans_create and dec by osd_trans_stop */ + int oti_ins_cache_depth; struct lu_buf oti_xattr_lbuf; + zap_cursor_t oti_zc; + zap_cursor_t oti_zc2; }; extern struct lu_context_key osd_key; @@ -235,17 +284,16 @@ struct osd_thandle { struct list_head ot_sa_list; dmu_tx_t *ot_tx; struct lquota_trans ot_quota_trans; - __u32 ot_write_commit:1, - ot_assigned:1; + __u32 ot_assigned:1; }; -#define OSD_OI_NAME_SIZE 16 +#define OSD_OI_NAME_SIZE 24 /* * Object Index (OI) instance. */ struct osd_oi { - char oi_name[OSD_OI_NAME_SIZE]; /* unused */ + char oi_name[OSD_OI_NAME_SIZE]; uint64_t oi_zapid; dnode_t *oi_dn; }; @@ -283,6 +331,8 @@ struct osd_device { struct proc_dir_entry *od_proc_entry; struct lprocfs_stats *od_stats; + uint64_t od_remote_parent_dir; + uint64_t od_index_backup_id; uint64_t od_max_blksz; uint64_t od_root; uint64_t od_O_id; @@ -294,13 +344,20 @@ struct osd_device { od_prop_rdonly:1, /**< ZFS property readonly */ od_xattr_in_sa:1, od_is_ost:1, - od_posix_acl:1; + od_in_init:1, + od_posix_acl:1, + od_nonrotational:1; unsigned int od_dnsize; + int od_index_backup_stop; + enum lustre_index_backup_policy od_index_backup_policy; char od_mntdev[128]; char od_svname[128]; + uuid_t od_uuid; int od_connects; + int od_index; + __s64 od_auto_scrub_interval; struct lu_site od_site; dnode_t *od_groupused_dn; @@ -309,8 +366,11 @@ struct osd_device { dnode_t *od_projectused_dn; #endif - /* quota slave instance */ - struct qsd_instance *od_quota_slave; + /* quota slave instance for inode */ + struct qsd_instance *od_quota_slave_md; + + /* quota slave instance for block */ + struct qsd_instance *od_quota_slave_dt; struct brw_stats od_brw_stats; atomic_t od_r_in_flight; @@ -327,8 +387,25 @@ struct osd_device { /* osd seq instance */ struct lu_client_seq *od_cl_seq; + + struct semaphore od_otable_sem; + struct osd_otable_it *od_otable_it; + struct lustre_scrub od_scrub; + struct list_head od_ios_list; + struct list_head od_index_backup_list; + struct list_head od_index_restore_list; + spinlock_t od_lock; + unsigned long long od_readcache_max_filesize; }; +static inline struct qsd_instance *osd_def_qsd(struct osd_device *osd) +{ + if (osd->od_is_ost) + return osd->od_quota_slave_dt; + else + return osd->od_quota_slave_md; +} + enum osd_destroy_type { OSD_DESTROY_NONE = 0, OSD_DESTROY_SYNC = 1, @@ -372,7 +449,8 @@ struct osd_object { #ifdef ZFS_PROJINHERIT oo_with_projid:1, #endif - oo_late_attr_set:1; + oo_late_attr_set:1, + oo_pfid_in_lma:1; /* the i_flags in LMA */ __u32 oo_lma_flags; @@ -386,9 +464,11 @@ struct osd_object { }; uint64_t oo_parent; /* used only at object creation */ }; + struct lu_object_header *oo_header; }; -int osd_statfs(const struct lu_env *, struct dt_device *, struct obd_statfs *); +int osd_statfs(const struct lu_env *, struct dt_device *, struct obd_statfs *, + struct obd_statfs_info *); extern const struct dt_index_operations osd_acct_index_ops; extern struct lu_device_operations osd_lu_ops; extern struct dt_index_operations osd_dir_ops; @@ -475,7 +555,33 @@ static inline struct seq_server_site *osd_seq_site(struct osd_device *osd) static inline char *osd_name(struct osd_device *osd) { - return osd->od_dt_dev.dd_lu_dev.ld_obd->obd_name; + return osd->od_svname; +} + +static inline void zfs_set_bit(int nr, __u8 *addr) +{ + set_bit(nr, (unsigned long *)addr); +} + +static inline int zfs_test_bit(int nr, __u8 *addr) +{ + return test_bit(nr, (const unsigned long *)addr); +} + +static inline int osd_oi_fid2idx(struct osd_device *dev, + const struct lu_fid *fid) +{ + return fid->f_seq & (dev->od_oi_count - 1); +} + +static inline struct osd_oi *osd_fid2oi(struct osd_device *osd, + const struct lu_fid *fid) +{ + LASSERTF(osd->od_oi_table && osd->od_oi_count >= 1, + "%s: "DFID", oi_count %d\n", + osd_name(osd), PFID(fid), osd->od_oi_count); + + return osd->od_oi_table[osd_oi_fid2idx(osd, fid)]; } #ifdef CONFIG_PROC_FS @@ -515,11 +621,14 @@ int osd_object_sa_update(struct osd_object *obj, sa_attr_type_t type, int __osd_zap_create(const struct lu_env *env, struct osd_device *osd, dnode_t **zap_dnp, dmu_tx_t *tx, struct lu_attr *la, unsigned dnsize, zap_flags_t flags); -int __osd_object_create(const struct lu_env *env, struct osd_object *obj, +int __osd_object_create(const struct lu_env *env, struct osd_device *osd, + struct osd_object *obj, const struct lu_fid *fid, dnode_t **dnp, dmu_tx_t *tx, struct lu_attr *la); int __osd_attr_init(const struct lu_env *env, struct osd_device *osd, struct osd_object *obj, sa_handle_t *sa_hdl, dmu_tx_t *tx, struct lu_attr *la, uint64_t parent, nvlist_t *); +int osd_find_new_dnode(const struct lu_env *env, dmu_tx_t *tx, + uint64_t oid, dnode_t **dnp); /* osd_oi.c */ int osd_oi_init(const struct lu_env *env, struct osd_device *o); @@ -540,8 +649,21 @@ struct osd_idmap_cache *osd_idc_find_or_init(const struct lu_env *env, struct osd_idmap_cache *osd_idc_find(const struct lu_env *env, struct osd_device *osd, const struct lu_fid *fid); +int osd_idc_find_and_init_with_oid(const struct lu_env *env, + struct osd_device *osd, + const struct lu_fid *fid, + uint64_t oid); +int fid_is_on_ost(const struct lu_env *env, struct osd_device *osd, + const struct lu_fid *fid); +int osd_obj_find_or_create(const struct lu_env *env, struct osd_device *o, + uint64_t parent, const char *name, uint64_t *child, + const struct lu_fid *fid, bool isdir); + +extern unsigned int osd_oi_count; /* osd_index.c */ +int osd_get_fid_by_oid(const struct lu_env *env, struct osd_device *osd, + uint64_t oid, struct lu_fid *fid); int osd_index_try(const struct lu_env *env, struct dt_object *dt, const struct dt_index_features *feat); int osd_fld_lookup(const struct lu_env *env, struct osd_device *osd, @@ -554,6 +676,27 @@ void osd_zap_cursor_fini(zap_cursor_t *zc); uint64_t osd_zap_cursor_serialize(zap_cursor_t *zc); int osd_remote_fid(const struct lu_env *env, struct osd_device *osd, const struct lu_fid *fid); +int osd_add_to_remote_parent(const struct lu_env *env, + struct osd_device *osd, + struct osd_object *obj, + struct osd_thandle *oh); +int osd_delete_from_remote_parent(const struct lu_env *env, + struct osd_device *osd, + struct osd_object *obj, + struct osd_thandle *oh, bool destroy); +int __osd_xattr_load_by_oid(struct osd_device *osd, uint64_t oid, + nvlist_t **sa); + +/* osd_scrub.c */ +int osd_scrub_setup(const struct lu_env *env, struct osd_device *dev); +void osd_scrub_cleanup(const struct lu_env *env, struct osd_device *dev); +int osd_scrub_start(const struct lu_env *env, struct osd_device *dev, + __u32 flags); +void osd_scrub_stop(struct osd_device *dev); +int osd_oii_insert(const struct lu_env *env, struct osd_device *dev, + const struct lu_fid *fid, uint64_t oid, bool insert); +int osd_oii_lookup(struct osd_device *dev, const struct lu_fid *fid, + uint64_t *oid); /* osd_xattr.c */ int __osd_sa_xattr_schedule_update(const struct lu_env *env, @@ -568,6 +711,8 @@ int __osd_xattr_load(struct osd_device *osd, sa_handle_t *hdl, int __osd_xattr_get_large(const struct lu_env *env, struct osd_device *osd, uint64_t xattr, struct lu_buf *buf, const char *name, int *sizep); +int osd_xattr_get_internal(const struct lu_env *env, struct osd_object *obj, + struct lu_buf *buf, const char *name, int *sizep); int osd_xattr_get(const struct lu_env *env, struct dt_object *dt, struct lu_buf *buf, const char *name); int osd_declare_xattr_set(const struct lu_env *env, struct dt_object *dt, @@ -597,6 +742,33 @@ int __osd_xattr_set(const struct lu_env *env, struct osd_object *obj, struct osd_thandle *oh); int __osd_sa_xattr_update(const struct lu_env *env, struct osd_object *obj, struct osd_thandle *oh); + +#define OSD_BASE_EA_IN_BONUS (ZFS_SA_BASE_ATTR_SIZE + \ + sizeof(__u64) /* VBR VERSION */ + \ + sizeof(struct lustre_mdt_attrs) /* LMA */) + +#ifdef HAVE_DMU_OBJECT_ALLOC_DNSIZE +int osd_find_dnsize(struct osd_device *osd, int ea_in_bonus); +#else +static inline int +osd_find_dnsize(struct osd_device *osd, int ea_in_bonus) +{ + return DN_MAX_BONUSLEN; +} +#endif + +static inline int osd_object_is_zap(dnode_t *dn) +{ + return (dn->dn_type == DMU_OT_DIRECTORY_CONTENTS || + dn->dn_type == DMU_OT_USERGROUP_USED); +} + +/* XXX: f_ver is not counted, but may differ too */ +static inline void osd_fid2str(char *buf, const struct lu_fid *fid, int len) +{ + snprintf(buf, len, DFID_NOBRACE, PFID(fid)); +} + static inline int osd_xattr_set_internal(const struct lu_env *env, struct osd_object *obj, const struct lu_buf *buf, const char *name, int fl, @@ -642,11 +814,11 @@ static inline uint32_t attrs_zfs2fs(const uint64_t flags) #endif #ifndef HAVE_DSL_POOL_CONFIG -static inline void dsl_pool_config_enter(dsl_pool_t *dp, char *name) +static inline void dsl_pool_config_enter(dsl_pool_t *dp, void *name) { } -static inline void dsl_pool_config_exit(dsl_pool_t *dp, char *name) +static inline void dsl_pool_config_exit(dsl_pool_t *dp, void *name) { } #endif @@ -771,7 +943,7 @@ static inline int osd_sa_handle_get(struct osd_object *obj) SA_HDL_PRIVATE, &obj->oo_sa_hdl); if (rc) return rc; - refcount_add(&dn->dn_bonus->db_holds, osd_obj_tag); + zfs_refcount_add(&dn->dn_bonus->db_holds, osd_obj_tag); return 0; } @@ -782,7 +954,6 @@ static inline void osd_dnode_rele(dnode_t *dn) LASSERT(dn->dn_bonus); db = dn->dn_bonus; - DB_DNODE_EXIT(db); dmu_buf_rele(&db->db, osd_obj_tag); } @@ -911,4 +1082,39 @@ static inline int osd_dmu_read(struct osd_device *osd, dnode_t *dn, dmu_objset_disown((os), (tag)) #endif +static inline int +osd_index_register(struct osd_device *osd, const struct lu_fid *fid, + __u32 keysize, __u32 recsize) +{ + return lustre_index_register(&osd->od_dt_dev, osd_name(osd), + &osd->od_index_backup_list, &osd->od_lock, + &osd->od_index_backup_stop, + fid, keysize, recsize); +} + +static inline void +osd_index_backup(const struct lu_env *env, struct osd_device *osd, bool backup) +{ + struct lu_fid *fid = &osd_oti_get(env)->oti_fid; + int rc; + + lu_local_obj_fid(fid, INDEX_BACKUP_OID); + rc = osd_idc_find_and_init_with_oid(env, osd, fid, + osd->od_index_backup_id); + if (rc) + backup = false; + + lustre_index_backup(env, &osd->od_dt_dev, osd_name(osd), + &osd->od_index_backup_list, &osd->od_lock, + &osd->od_index_backup_stop, backup); +} + +#ifndef HAVE_DMU_TX_MARK_NETFREE +#define dmu_tx_mark_netfree(tx) +#endif + +#ifndef HAVE_ZFS_INODE_TIMESPEC +#define inode_timespec_t timestruc_t +#endif + #endif /* _OSD_INTERNAL_H */