X-Git-Url: https://git.whamcloud.com/?p=fs%2Flustre-release.git;a=blobdiff_plain;f=lustre%2Finclude%2Flustre_fid.h;h=0deb5dbec1f9ba67a0013e22212c9720fdde3863;hp=e4c8324331f55e72f7bc4040e0114d89aa00419b;hb=79d06b6fac3d18f2e5755a940e8afa42e70f3ba2;hpb=7f72fee8c31abd37823386a4587fb5de937b7d3f diff --git a/lustre/include/lustre_fid.h b/lustre/include/lustre_fid.h index e4c8324..0deb5db 100644 --- a/lustre/include/lustre_fid.h +++ b/lustre/include/lustre_fid.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, 2012, Intel Corporation. */ /* * This file is part of Lustre, http://www.lustre.org/ @@ -44,15 +44,119 @@ /** \defgroup fid fid * * @{ + * + * http://wiki.lustre.org/index.php/Architecture_-_Interoperability_fids_zfs + * describes the FID namespace and interoperability requirements for FIDs. + * The important parts of that document are included here for reference. + * + * FID + * File IDentifier generated by client from range allocated by the SEQuence + * service and stored in struct lu_fid. The FID is composed of three parts: + * SEQuence, ObjectID, and VERsion. The SEQ component is a filesystem + * unique 64-bit integer, and only one client is ever assigned any SEQ value. + * The first 0x400 FID_SEQ_NORMAL [2^33, 2^33 + 0x400] values are reserved + * for system use. The OID component is a 32-bit value generated by the + * client on a per-SEQ basis to allow creating many unique FIDs without + * communication with the server. The VER component is a 32-bit value that + * distinguishes between different FID instantiations, such as snapshots or + * separate subtrees within the filesystem. FIDs with the same VER field + * are considered part of the same namespace. + * + * OLD filesystems are those upgraded from Lustre 1.x that predate FIDs, and + * MDTs use 32-bit ldiskfs internal inode/generation numbers (IGIFs), while + * OSTs use 64-bit Lustre object IDs and generation numbers. + * + * NEW filesystems are those formatted since the introduction of FIDs. + * + * IGIF + * Inode and Generation In FID, a surrogate FID used to globally identify + * an existing object on OLD formatted MDT file system. This would only be + * used on MDT0 in a DNE filesystem, because there cannot be more than one + * MDT in an OLD formatted filesystem. Belongs to sequence in [12, 2^32 - 1] + * range, where inode number is stored in SEQ, and inode generation is in OID. + * NOTE: This assumes no more than 2^32-1 inodes exist in the MDT filesystem, + * which is the maximum possible for an ldiskfs backend. It also assumes + * that the reserved ext3/ext4/ldiskfs inode numbers [0-11] are never visible + * to clients, which has always been true. + * + * IDIF + * object ID In FID, a surrogate FID used to globally identify an existing + * OST object on OLD formatted OST file system. Belongs to a sequence in + * [2^32, 2^33 - 1]. Sequence number is calculated as: + * + * 1 << 32 | (ost_index << 16) | ((objid >> 32) & 0xffff) + * + * that is, SEQ consists of 16-bit OST index, and higher 16 bits of object + * ID. The generation of unique SEQ values per OST allows the IDIF FIDs to + * be identified in the FLD correctly. The OID field is calculated as: + * + * objid & 0xffffffff + * + * that is, it consists of lower 32 bits of object ID. For objects within + * the IDIF range, object ID extraction will be: + * + * o_id = (fid->f_seq & 0x7fff) << 16 | fid->f_oid; + * o_seq = 0; // formerly group number + * + * NOTE: This assumes that no more than 2^48-1 objects have ever been created + * on any OST, and that no more than 65535 OSTs are in use. Both are very + * reasonable assumptions, i.e. an IDIF can uniquely map all objects assuming + * a maximum creation rate of 1M objects per second for a maximum of 9 years, + * or combinations thereof. + * + * OST_MDT0 + * Surrogate FID used to identify an existing object on OLD formatted OST + * filesystem. Belongs to the reserved SEQuence 0, and is used prior to + * the introduction of FID-on-OST, at which point IDIF will be used to + * identify objects as residing on a specific OST. + * + * LLOG + * For Lustre Log objects the object sequence 1 is used. This is compatible + * with both OLD and NEW namespaces, as this SEQ number is in the + * ext3/ldiskfs reserved inode range and does not conflict with IGIF + * sequence numbers. + * + * ECHO + * For testing OST IO performance the object sequence 2 is used. This is + * compatible with both OLD and NEW namespaces, as this SEQ number is in + * the ext3/ldiskfs reserved inode range and does not conflict with IGIF + * sequence numbers. + * + * OST_MDT1 .. OST_MAX + * For testing with multiple MDTs the object sequence 3 through 9 is used, + * allowing direct mapping of MDTs 1 through 7 respectively, for a total + * of 8 MDTs including OST_MDT0. This matches the legacy CMD project "group" + * mappings. However, this SEQ range is only for testing prior to any + * production DNE release, as the objects in this range conflict across all + * OSTs, as the OST index is not part of the FID. For production DNE usage, + * OST objects created by MDT1+ will use FID_SEQ_NORMAL FIDs. + * + * DLM OST objid to IDIF mapping + * For compatibility with existing OLD OST network protocol structures, the + * FID must map onto the o_id and o_seq in a manner that ensures existing + * objects are identified consistently for IO, as well as onto the LDLM + * namespace to ensure IDIFs there is only a single resource name for any + * object in the DLM. The OLD OST object DLM resource mapping is: + * + * resource[] = {o_id, o_seq, 0, 0}; // o_seq == 0 for production releases + * + * The NEW OST object DLM resource mapping is the same for both MDT and OST: + * + * resource[] = {SEQ, OID, VER, HASH}; + * + * NOTE: for mapping IDIF values to DLM resource names the o_id may be + * larger than the 2^33 reserved sequence numbers for IDIF, so it is possible + * for the o_id numbers to overlap FID SEQ numbers in the resource. However, + * in all production releases the OLD o_seq field is always zero, and all + * valid FID OID values are non-zero, so the lock resources will not collide. + * Even so, the MDT and OST resources are also in different LDLM namespaces. */ -/* - * struct lu_fid - */ #include #include #include #include +#include struct lu_site; @@ -66,26 +170,31 @@ extern const struct lu_fid LU_OBF_FID; extern const struct lu_fid LU_DOT_LUSTRE_FID; enum { - /* - * This is how may FIDs may be allocated in one sequence(128k) - */ - LUSTRE_SEQ_MAX_WIDTH = 0x0000000000020000ULL, - - /* - * How many sequences to allocate to a client at once. - */ - LUSTRE_SEQ_META_WIDTH = 0x0000000000000001ULL, - - /* - * seq allocation pool size. - */ - LUSTRE_SEQ_BATCH_WIDTH = LUSTRE_SEQ_META_WIDTH * 1000, - - /* - * This is how many sequences may be in one super-sequence allocated to - * MDTs. - */ - LUSTRE_SEQ_SUPER_WIDTH = ((1ULL << 30ULL) * LUSTRE_SEQ_META_WIDTH) + /* + * This is how may metadata FIDs may be allocated in one sequence(128k) + */ + LUSTRE_METADATA_SEQ_MAX_WIDTH = 0x0000000000020000ULL, + + /* + * This is how many data FIDs could be allocated in one sequence(4B - 1) + */ + LUSTRE_DATA_SEQ_MAX_WIDTH = 0x00000000FFFFFFFFULL, + + /* + * How many sequences to allocate to a client at once. + */ + LUSTRE_SEQ_META_WIDTH = 0x0000000000000001ULL, + + /* + * seq allocation pool size. + */ + LUSTRE_SEQ_BATCH_WIDTH = LUSTRE_SEQ_META_WIDTH * 1000, + + /* + * This is how many sequences may be in one super-sequence allocated to + * MDTs. + */ + LUSTRE_SEQ_SUPER_WIDTH = ((1ULL << 30ULL) * LUSTRE_SEQ_META_WIDTH) }; enum { @@ -97,28 +206,33 @@ enum { /** special OID for local objects */ enum local_oid { - /** \see fld_mod_init */ - FLD_INDEX_OID = 3UL, - /** \see fid_mod_init */ - FID_SEQ_CTL_OID = 4UL, - FID_SEQ_SRV_OID = 5UL, - /** \see mdd_mod_init */ - MDD_ROOT_INDEX_OID = 6UL, - MDD_ORPHAN_OID = 7UL, - MDD_LOV_OBJ_OID = 8UL, - MDD_CAPA_KEYS_OID = 9UL, - /** \see mdt_mod_init */ - MDT_LAST_RECV_OID = 11UL, - OSD_FS_ROOT_OID = 13UL, - ACCT_USER_OID = 15UL, - ACCT_GROUP_OID = 16UL, - OFD_LAST_RECV_OID = 19UL, - OFD_GROUP0_LAST_OID = 20UL, - OFD_GROUP4K_LAST_OID = 20UL+4096, - OFD_LAST_GROUP_OID = 4117UL, - LLOG_CATALOGS_OID = 4118UL, - MGS_CONFIGS_OID = 4119UL, - OFD_HEALTH_CHECK_OID = 4120UL, + /** \see fld_mod_init */ + FLD_INDEX_OID = 3UL, + /** \see fid_mod_init */ + FID_SEQ_CTL_OID = 4UL, + FID_SEQ_SRV_OID = 5UL, + /** \see mdd_mod_init */ + MDD_ROOT_INDEX_OID = 6UL, + MDD_ORPHAN_OID = 7UL, + MDD_LOV_OBJ_OID = 8UL, + MDD_CAPA_KEYS_OID = 9UL, + /** \see mdt_mod_init */ + LAST_RECV_OID = 11UL, + OSD_FS_ROOT_OID = 13UL, + ACCT_USER_OID = 15UL, + ACCT_GROUP_OID = 16UL, + LFSCK_BOOKMARK_OID = 17UL, + OTABLE_IT_OID = 18UL, + /* These two definitions are obsolete + * OFD_GROUP0_LAST_OID = 20UL, + * OFD_GROUP4K_LAST_OID = 20UL+4096, + */ + OFD_LAST_GROUP_OID = 4117UL, + LLOG_CATALOGS_OID = 4118UL, + MGS_CONFIGS_OID = 4119UL, + OFD_HEALTH_CHECK_OID = 4120UL, + MDD_LOV_OBJ_OSEQ = 4121UL, + LFSCK_NAMESPACE_OID = 4122UL, }; static inline void lu_local_obj_fid(struct lu_fid *fid, __u32 oid) @@ -128,6 +242,34 @@ static inline void lu_local_obj_fid(struct lu_fid *fid, __u32 oid) fid->f_ver = 0; } +static inline void lu_local_name_obj_fid(struct lu_fid *fid, __u32 oid) +{ + fid->f_seq = FID_SEQ_LOCAL_NAME; + fid->f_oid = oid; + fid->f_ver = 0; +} + +/* For new FS (>= 2.4), the root FID will be changed to + * [FID_SEQ_ROOT:1:0], for existing FS, (upgraded to 2.4), + * the root FID will still be IGIF */ +static inline int fid_is_root(const struct lu_fid *fid) +{ + return unlikely((fid_seq(fid) == FID_SEQ_ROOT && + fid_oid(fid) == 1)); +} + +static inline int fid_is_dot_lustre(const struct lu_fid *fid) +{ + return unlikely(fid_seq(fid) == FID_SEQ_DOT_LUSTRE && + fid_oid(fid) == FID_OID_DOT_LUSTRE); +} + +static inline int fid_is_otable_it(const struct lu_fid *fid) +{ + return unlikely(fid_seq(fid) == FID_SEQ_LOCAL_FILE && + fid_oid(fid) == OTABLE_IT_OID); +} + static inline int fid_is_acct(const struct lu_fid *fid) { return fid_seq(fid) == FID_SEQ_LOCAL_FILE && @@ -135,23 +277,59 @@ static inline int fid_is_acct(const struct lu_fid *fid) fid_oid(fid) == ACCT_GROUP_OID); } +static inline int fid_is_quota(const struct lu_fid *fid) +{ + return fid_seq(fid) == FID_SEQ_QUOTA || + fid_seq(fid) == FID_SEQ_QUOTA_GLB; +} + +static inline int fid_is_client_mdt_visible(const struct lu_fid *fid) +{ + const __u64 seq = fid_seq(fid); + + /* Here, we cannot distinguish whether the normal FID is for OST + * object or not. It is caller's duty to check more if needed. */ + return (!fid_is_last_id(fid) && + (fid_seq_is_norm(seq) || fid_seq_is_igif(seq))) || + fid_is_root(fid) || fid_is_dot_lustre(fid); +} + +static inline int fid_is_client_visible(const struct lu_fid *fid) +{ + return fid_is_client_mdt_visible(fid) || fid_is_idif(fid); +} + +static inline int fid_seq_in_fldb(__u64 seq) +{ + return fid_seq_is_igif(seq) || fid_seq_is_norm(seq) || + seq == FID_SEQ_ROOT; +} + +static inline void lu_last_id_fid(struct lu_fid *fid, __u64 seq) +{ + if (fid_seq_is_mdt0(seq)) { + fid->f_seq = fid_idif_seq(0, 0); + } else { + LASSERTF(fid_seq_is_norm(seq) || fid_seq_is_echo(seq) || + fid_seq_is_idif(seq), LPX64"\n", seq); + fid->f_seq = seq; + } + fid->f_oid = 0; + fid->f_ver = 0; +} + enum lu_mgr_type { LUSTRE_SEQ_SERVER, LUSTRE_SEQ_CONTROLLER }; -enum lu_cli_type { - LUSTRE_SEQ_METADATA, - LUSTRE_SEQ_DATA -}; - struct lu_server_seq; /* Client sequence manager interface. */ struct lu_client_seq { /* Sequence-controller export. */ struct obd_export *lcs_exp; - cfs_mutex_t lcs_mutex; + struct mutex lcs_mutex; /* * Range of allowed for allocation sequeces. When using lu_client_seq on @@ -217,7 +395,7 @@ struct lu_server_seq { struct lu_client_seq *lss_cli; /* Mutex for protecting allocation */ - cfs_mutex_t lss_mutex; + struct mutex lss_mutex; /* * Service uuid, passed from MDT + seq name to form unique seq name to @@ -239,21 +417,23 @@ struct lu_server_seq { /* sync is needed for update operation */ __u32 lss_need_sync; - /** - * Pointer to site object, required to access site fld. - */ - struct md_site *lss_site; + + /** + * Pointer to site object, required to access site fld. + */ + struct seq_server_site *lss_site; }; int seq_query(struct com_thread_info *info); +int seq_handle(struct ptlrpc_request *req); /* Server methods */ int seq_server_init(struct lu_server_seq *seq, - struct dt_device *dev, - const char *prefix, - enum lu_mgr_type type, - struct md_site *ls, - const struct lu_env *env); + struct dt_device *dev, + const char *prefix, + enum lu_mgr_type type, + struct seq_server_site *ss, + const struct lu_env *env); void seq_server_fini(struct lu_server_seq *seq, const struct lu_env *env); @@ -285,11 +465,15 @@ int seq_client_alloc_fid(const struct lu_env *env, struct lu_client_seq *seq, struct lu_fid *fid); int seq_client_get_seq(const struct lu_env *env, struct lu_client_seq *seq, seqno_t *seqnr); - +int seq_site_fini(const struct lu_env *env, struct seq_server_site *ss); /* Fids common stuff */ int fid_is_local(const struct lu_env *env, struct lu_site *site, const struct lu_fid *fid); +int client_fid_init(struct obd_device *obd, struct obd_export *exp, + enum lu_cli_type type); +int client_fid_fini(struct obd_device *obd); + /* fid locking */ struct ldlm_namespace; @@ -312,6 +496,36 @@ fid_build_reg_res_name(const struct lu_fid *f, } /* + * Build (DLM) resource identifier from global quota FID and quota ID. + */ +static inline struct ldlm_res_id * +fid_build_quota_resid(const struct lu_fid *glb_fid, union lquota_id *qid, + struct ldlm_res_id *res) +{ + fid_build_reg_res_name(glb_fid, res); + res->name[LUSTRE_RES_ID_QUOTA_SEQ_OFF] = fid_seq(&qid->qid_fid); + res->name[LUSTRE_RES_ID_QUOTA_VER_OID_OFF] = fid_ver_oid(&qid->qid_fid); + return res; +} + +/* + * Extract global FID and quota ID from resource name + */ +static inline void fid_extract_quota_resid(struct ldlm_res_id *res, + struct lu_fid *glb_fid, + union lquota_id *qid) +{ + glb_fid->f_seq = res->name[LUSTRE_RES_ID_SEQ_OFF]; + glb_fid->f_oid = (__u32)res->name[LUSTRE_RES_ID_VER_OID_OFF]; + glb_fid->f_ver = (__u32)(res->name[LUSTRE_RES_ID_VER_OID_OFF] >> 32); + + qid->qid_fid.f_seq = res->name[LUSTRE_RES_ID_QUOTA_SEQ_OFF]; + qid->qid_fid.f_oid = (__u32)res->name[LUSTRE_RES_ID_QUOTA_VER_OID_OFF]; + qid->qid_fid.f_ver = + (__u32)(res->name[LUSTRE_RES_ID_QUOTA_VER_OID_OFF] >> 32); +} + +/* * Return true if resource is for object identified by fid. */ static inline int fid_res_name_eq(const struct lu_fid *f, @@ -321,6 +535,16 @@ static inline int fid_res_name_eq(const struct lu_fid *f, name->name[LUSTRE_RES_ID_VER_OID_OFF] == fid_ver_oid(f); } +/* reverse function of fid_build_reg_res_name() */ +static inline void fid_build_from_res_name(struct lu_fid *f, + const struct ldlm_res_id *name) +{ + fid_zero(f); + f->f_seq = name->name[LUSTRE_RES_ID_SEQ_OFF]; + f->f_oid = name->name[LUSTRE_RES_ID_VER_OID_OFF] & 0xffffffff; + f->f_ver = name->name[LUSTRE_RES_ID_VER_OID_OFF] >> 32; + LASSERT(fid_res_name_eq(f, name)); +} static inline struct ldlm_res_id * fid_build_pdo_res_name(const struct lu_fid *f, @@ -332,6 +556,36 @@ fid_build_pdo_res_name(const struct lu_fid *f, return name; } +/** + * Build DLM resource name from object id & seq, which will be removed + * finally, when we replace ost_id with FID in data stack. + * + * To keep the compatibility, res[0] = oid, res[1] = seq + */ +static inline void ostid_build_res_name(struct ost_id *oid, + struct ldlm_res_id *name) +{ + memset(name, 0, sizeof *name); + name->name[LUSTRE_RES_ID_SEQ_OFF] = oid->oi_id; + name->name[LUSTRE_RES_ID_VER_OID_OFF] = oid->oi_seq; +} + +static inline void ostid_res_name_to_id(struct ost_id *oid, + struct ldlm_res_id *name) +{ + oid->oi_id = name->name[LUSTRE_RES_ID_SEQ_OFF]; + oid->oi_seq = name->name[LUSTRE_RES_ID_VER_OID_OFF]; +} + +/** + * Return true if the resource is for the object identified by this id & group. + */ +static inline int ostid_res_name_eq(struct ost_id *oid, + struct ldlm_res_id *name) +{ + return name->name[LUSTRE_RES_ID_SEQ_OFF] == oid->oi_id && + name->name[LUSTRE_RES_ID_VER_OID_OFF] == oid->oi_seq; +} /** * Flatten 128-bit FID values into a 64-bit value for use as an inode number. @@ -394,6 +648,18 @@ static inline __u32 fid_flatten32(const struct lu_fid *fid) RETURN(ino ? ino : fid_oid(fid)); } +static inline int lu_fid_diff(struct lu_fid *fid1, struct lu_fid *fid2) +{ + LASSERTF(fid_seq(fid1) == fid_seq(fid2), "fid1:"DFID", fid2:"DFID"\n", + PFID(fid1), PFID(fid2)); + + if (fid_is_idif(fid1) && fid_is_idif(fid2)) + return fid_idif_id(fid1->f_seq, fid1->f_oid, fid1->f_ver) - + fid_idif_id(fid2->f_seq, fid2->f_oid, fid2->f_ver); + + return fid_oid(fid1) - fid_oid(fid2); +} + #define LUSTRE_SEQ_SRV_NAME "seq_srv" #define LUSTRE_SEQ_CTL_NAME "seq_ctl"