X-Git-Url: https://git.whamcloud.com/?p=fs%2Flustre-release.git;a=blobdiff_plain;f=lustre%2Finclude%2Fmd_object.h;h=99998e2c71ffe49481e59e2d77e997e0029e7af6;hp=5b11fb77379aa08bfefe1447018cea3f9517b62e;hb=a0b959c53d10bf3f0fd6b22de46397d0c7e5f667;hpb=a1efc1411e2e7b67102d132e656d2a11a99636b3;ds=sidebyside diff --git a/lustre/include/md_object.h b/lustre/include/md_object.h index 5b11fb7..99998e2 100644 --- a/lustre/include/md_object.h +++ b/lustre/include/md_object.h @@ -1,6 +1,4 @@ -/* -*- mode: c; c-basic-offset: 8; indent-tabs-mode: nil; -*- - * vim:expandtab:shiftwidth=8:tabstop=8: - * +/* * GPL HEADER START * * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER. @@ -26,8 +24,10 @@ * GPL HEADER END */ /* - * Copyright 2008 Sun Microsystems, Inc. All rights reserved + * Copyright (c) 2007, 2010, Oracle and/or its affiliates. All rights reserved. * Use is subject to license terms. + * + * Copyright (c) 2011, 2013, Intel Corporation. */ /* * This file is part of Lustre, http://www.lustre.org/ @@ -41,6 +41,10 @@ #ifndef _LUSTRE_MD_OBJECT_H #define _LUSTRE_MD_OBJECT_H +#ifndef HAVE_SERVER_SUPPORT +# error "client code should not depend on md_object.h" +#endif /* !HAVE_SERVER_SUPPORT */ + /** \defgroup md md * Sub-class of lu_object with methods common for "meta-data" objects in MDT * stack. @@ -56,65 +60,26 @@ /* * super-class definitions. */ -#include -#include +#include struct md_device; struct md_device_operations; struct md_object; - - -enum { - UCRED_INVALID = -1, - UCRED_INIT = 0, - UCRED_OLD = 1, - UCRED_NEW = 2 -}; - -struct md_ucred { - __u32 mu_valid; - __u32 mu_o_uid; - __u32 mu_o_gid; - __u32 mu_o_fsuid; - __u32 mu_o_fsgid; - __u32 mu_uid; - __u32 mu_gid; - __u32 mu_fsuid; - __u32 mu_fsgid; - __u32 mu_suppgids[2]; - cfs_cap_t mu_cap; - __u32 mu_umask; - struct group_info *mu_ginfo; - struct md_identity *mu_identity; -}; - -#define MD_CAPAINFO_MAX 5 - -/** there are at most 5 fids in one operation, see rename, NOTE the last one - * is a temporary one used for is_subdir() */ -struct md_capainfo { - const struct lu_fid *mc_fid[MD_CAPAINFO_MAX]; - struct lustre_capa *mc_capa[MD_CAPAINFO_MAX]; -}; - -/** - * Implemented in mdd/mdd_handler.c. - * - * XXX should be moved into separate .h/.c together with all md security - * related definitions. - */ -struct md_ucred *md_ucred(const struct lu_env *env); -struct md_capainfo *md_capainfo(const struct lu_env *env); +struct obd_export; /** metadata attributes */ enum ma_valid { - MA_INODE = (1 << 0), - MA_LOV = (1 << 1), - MA_COOKIE = (1 << 2), - MA_FLAGS = (1 << 3), - MA_LMV = (1 << 4), - MA_ACL_DEF = (1 << 5), - MA_LOV_DEF = (1 << 6) + MA_INODE = (1 << 0), + MA_LOV = (1 << 1), + MA_COOKIE = (1 << 2), + MA_FLAGS = (1 << 3), + MA_LMV = (1 << 4), + MA_ACL_DEF = (1 << 5), + MA_LOV_DEF = (1 << 6), + MA_HSM = (1 << 7), + MA_SOM = (1 << 8), + MA_PFID = (1 << 9), + MA_LMV_DEF = (1 << 10) }; typedef enum { @@ -135,20 +100,47 @@ typedef enum { MDT_PDO_LOCK = (1 << 1) } mdl_type_t; +/* memory structure for hsm attributes + * for fields description see the on disk structure hsm_attrs + * which is defined in lustre_idl.h + */ +struct md_hsm { + __u32 mh_compat; + __u32 mh_flags; + __u64 mh_arch_id; + __u64 mh_arch_ver; +}; + +#define IOEPOCH_INVAL 0 + +/* memory structure for som attributes + * for fields description see the on disk structure som_attrs + * which is defined in lustre_idl.h + */ +struct md_som_data { + __u32 msd_compat; + __u32 msd_incompat; + __u64 msd_ioepoch; + __u64 msd_size; + __u64 msd_blocks; + __u64 msd_mountid; +}; + struct md_attr { __u64 ma_valid; __u64 ma_need; __u64 ma_attr_flags; struct lu_attr ma_attr; + struct lu_fid ma_pfid; + struct md_hsm ma_hsm; struct lov_mds_md *ma_lmm; + union lmv_mds_md *ma_lmv; + void *ma_acl; + struct lustre_capa *ma_capa; + struct md_som_data *ma_som; int ma_lmm_size; - struct lmv_stripe_md *ma_lmv; int ma_lmv_size; - void *ma_acl; int ma_acl_size; - struct llog_cookie *ma_cookie; - int ma_cookie_size; - struct lustre_capa *ma_capa; }; /** Additional parameters for create */ @@ -156,31 +148,30 @@ struct md_op_spec { union { /** symlink target */ const char *sp_symname; - /** parent FID for cross-ref mkdir */ - const struct lu_fid *sp_pfid; /** eadata for regular files */ struct md_spec_reg { - /** lov objs exist already */ - const struct lu_fid *fid; - int no_lov_create; const void *eadata; int eadatalen; } sp_ea; } u; - + /** Create flag from client: such as MDS_OPEN_CREAT, and others. */ - __u32 sp_cr_flags; + __u64 sp_cr_flags; - /** Should mdd do lookup sanity check or not. */ - int sp_cr_lookup; + /** don't create lov objects or llog cookie - this replay */ + unsigned int no_create:1, + sp_cr_lookup:1, /* do lookup sanity check or not. */ + sp_rm_entry:1, /* only remove name entry */ + sp_permitted:1; /* do not check permission */ - /** Current lock mode for parent dir where create is performing. */ + /** Current lock mode for parent dir where create is performing. */ mdl_mode_t sp_cr_mode; - /** Check for split */ - int sp_ck_split; + /** to create directory */ + const struct dt_index_features *sp_feat; }; +union ldlm_policy_data; /** * Operations implemented for each md object (both directory and leaf). */ @@ -208,35 +199,40 @@ struct md_object_operations { int (*moo_xattr_del)(const struct lu_env *env, struct md_object *obj, const char *name); + /** This method is used to swap the layouts between 2 objects */ + int (*moo_swap_layouts)(const struct lu_env *env, + struct md_object *obj1, struct md_object *obj2, + __u64 flags); + + /** \retval number of bytes actually read upon success */ int (*moo_readpage)(const struct lu_env *env, struct md_object *obj, const struct lu_rdpg *rdpg); int (*moo_readlink)(const struct lu_env *env, struct md_object *obj, struct lu_buf *buf); - - /** part of cross-ref operation */ - int (*moo_object_create)(const struct lu_env *env, - struct md_object *obj, - const struct md_op_spec *spec, - struct md_attr *ma); - - int (*moo_ref_add)(const struct lu_env *env, - struct md_object *obj, - const struct md_attr *ma); - - int (*moo_ref_del)(const struct lu_env *env, - struct md_object *obj, - struct md_attr *ma); + int (*moo_changelog)(const struct lu_env *env, + enum changelog_rec_type type, int flags, + struct md_object *obj); int (*moo_open)(const struct lu_env *env, struct md_object *obj, int flag); int (*moo_close)(const struct lu_env *env, struct md_object *obj, - struct md_attr *ma); - + struct md_attr *ma, int mode); + int (*moo_capa_get)(const struct lu_env *, struct md_object *, struct lustre_capa *, int renewal); + int (*moo_object_sync)(const struct lu_env *, struct md_object *); + + int (*moo_object_lock)(const struct lu_env *env, struct md_object *obj, + struct lustre_handle *lh, + struct ldlm_enqueue_info *einfo, + union ldlm_policy_data *policy); + int (*moo_object_unlock)(const struct lu_env *env, + struct md_object *obj, + struct ldlm_enqueue_info *einfo, + union ldlm_policy_data *policy); }; /** @@ -274,37 +270,25 @@ struct md_dir_operations { struct md_object *src_obj, const struct lu_name *lname, struct md_attr *ma); - int (*mdo_unlink)(const struct lu_env *env, struct md_object *pobj, - struct md_object *cobj, const struct lu_name *lname, - struct md_attr *ma); + int (*mdo_unlink)(const struct lu_env *env, struct md_object *pobj, + struct md_object *cobj, const struct lu_name *lname, + struct md_attr *ma, int no_name); - /** partial ops for cross-ref case */ - int (*mdo_name_insert)(const struct lu_env *env, - struct md_object *obj, - const struct lu_name *lname, - const struct lu_fid *fid, - const struct md_attr *ma); - - int (*mdo_name_remove)(const struct lu_env *env, - struct md_object *obj, - const struct lu_name *lname, - const struct md_attr *ma); - - int (*mdo_rename_tgt)(const struct lu_env *env, struct md_object *pobj, - struct md_object *tobj, const struct lu_fid *fid, - const struct lu_name *lname, struct md_attr *ma); + int (*mdo_migrate)(const struct lu_env *env, struct md_object *pobj, + struct md_object *sobj, const struct lu_name *lname, + struct md_object *tobj, struct md_attr *ma); }; struct md_device_operations { /** meta-data device related handlers. */ - int (*mdo_root_get)(const struct lu_env *env, struct md_device *m, - struct lu_fid *f); + int (*mdo_root_get)(const struct lu_env *env, struct md_device *m, + struct lu_fid *f); - int (*mdo_maxsize_get)(const struct lu_env *env, struct md_device *m, - int *md_size, int *cookie_size); + int (*mdo_maxeasize_get)(const struct lu_env *env, struct md_device *m, + int *easize); int (*mdo_statfs)(const struct lu_env *env, struct md_device *m, - struct kstatfs *sfs); + struct obd_statfs *sfs); int (*mdo_init_capa_ctxt)(const struct lu_env *env, struct md_device *m, int mode, unsigned long timeout, __u32 alg, @@ -313,115 +297,36 @@ struct md_device_operations { int (*mdo_update_capa_key)(const struct lu_env *env, struct md_device *m, struct lustre_capa_key *key); -}; -enum md_upcall_event { - /**sync the md layer*/ - MD_LOV_SYNC = (1 << 0), - /** Just for split, no need trans, for replay */ - MD_NO_TRANS = (1 << 1), - MD_LOV_CONFIG = (1 << 2) -}; + int (*mdo_llog_ctxt_get)(const struct lu_env *env, + struct md_device *m, int idx, void **h); -struct md_upcall { - /** this lock protects upcall using against its removal - * read lock is for usage the upcall, write - for init/fini */ - struct rw_semaphore mu_upcall_sem; - /** device to call, upper layer normally */ - struct md_device *mu_upcall_dev; - /** upcall function */ - int (*mu_upcall)(const struct lu_env *env, struct md_device *md, - enum md_upcall_event ev); + int (*mdo_iocontrol)(const struct lu_env *env, struct md_device *m, + unsigned int cmd, int len, void *data); }; struct md_device { - struct lu_device md_lu_dev; + struct lu_device md_lu_dev; const struct md_device_operations *md_ops; - struct md_upcall md_upcall; }; -static inline void md_upcall_init(struct md_device *m, void *upcl) -{ - init_rwsem(&m->md_upcall.mu_upcall_sem); - m->md_upcall.mu_upcall_dev = NULL; - m->md_upcall.mu_upcall = upcl; -} - -static inline void md_upcall_dev_set(struct md_device *m, struct md_device *up) -{ - down_write(&m->md_upcall.mu_upcall_sem); - m->md_upcall.mu_upcall_dev = up; - up_write(&m->md_upcall.mu_upcall_sem); -} - -static inline void md_upcall_fini(struct md_device *m) -{ - down_write(&m->md_upcall.mu_upcall_sem); - m->md_upcall.mu_upcall_dev = NULL; - m->md_upcall.mu_upcall = NULL; - up_write(&m->md_upcall.mu_upcall_sem); -} - -static inline int md_do_upcall(const struct lu_env *env, struct md_device *m, - enum md_upcall_event ev) -{ - int rc = 0; - down_read(&m->md_upcall.mu_upcall_sem); - if (m->md_upcall.mu_upcall_dev != NULL && - m->md_upcall.mu_upcall_dev->md_upcall.mu_upcall != NULL) { - rc = m->md_upcall.mu_upcall_dev->md_upcall.mu_upcall(env, - m->md_upcall.mu_upcall_dev, ev); - } - up_read(&m->md_upcall.mu_upcall_sem); - return rc; -} - struct md_object { - struct lu_object mo_lu; + struct lu_object mo_lu; const struct md_object_operations *mo_ops; const struct md_dir_operations *mo_dir_ops; }; -/** - * md-server site. - */ -struct md_site { - struct lu_site ms_lu; - /** - * mds number of this site. - */ - mdsno_t ms_node_id; - /** - * Fid location database - */ - struct lu_server_fld *ms_server_fld; - struct lu_client_fld *ms_client_fld; - - /** - * Server Seq Manager - */ - struct lu_server_seq *ms_server_seq; - - /** - * Controller Seq Manager - */ - struct lu_server_seq *ms_control_seq; - struct obd_export *ms_control_exp; - - /** - * Client Seq Manager - */ - struct lu_client_seq *ms_client_seq; -}; - -static inline int lu_device_is_md(const struct lu_device *d) -{ - return ergo(d != NULL, d->ld_type->ldt_tags & LU_DEVICE_MD); -} +/* Mark the object to be dead, and can not be accessed anymore. + * XXX, right now, it will only be used for striped directory to + * mark the slave stripes dead, when deleting master object. It will be + * stored in slave LMV EA (see lod_mark_dead_object), which is only + * temporary, and will be removed later when we have proper way to mark + * the dead object. */ +#define LUSTRE_SLAVE_DEAD_FL 0x80000000 static inline struct md_device *lu2md_dev(const struct lu_device *d) { - LASSERT(lu_device_is_md(d)); + LASSERT(IS_ERR(d) || lu_device_is_md(d)); return container_of0(d, struct md_device, md_lu_dev); } @@ -432,34 +337,18 @@ static inline struct lu_device *md2lu_dev(struct md_device *d) static inline struct md_object *lu2md(const struct lu_object *o) { - LASSERT(lu_device_is_md(o->lo_dev)); + LASSERT(o == NULL || IS_ERR(o) || lu_device_is_md(o->lo_dev)); return container_of0(o, struct md_object, mo_lu); } -static inline struct md_object *md_object_next(const struct md_object *obj) -{ - return (obj ? lu2md(lu_object_next(&obj->mo_lu)) : NULL); -} - -static inline struct md_device *md_obj2dev(const struct md_object *o) -{ - LASSERT(lu_device_is_md(o->mo_lu.lo_dev)); - return container_of0(o->mo_lu.lo_dev, struct md_device, md_lu_dev); -} - -static inline struct md_site *lu_site2md(const struct lu_site *s) -{ - return container_of0(s, struct md_site, ms_lu); -} - static inline int md_device_init(struct md_device *md, struct lu_device_type *t) { - return lu_device_init(&md->md_lu_dev, t); + return lu_device_init(&md->md_lu_dev, t); } static inline void md_device_fini(struct md_device *md) { - lu_device_fini(&md->md_lu_dev); + lu_device_fini(&md->md_lu_dev); } static inline struct md_object *md_object_find_slice(const struct lu_env *env, @@ -497,6 +386,14 @@ static inline int mo_readlink(const struct lu_env *env, return m->mo_ops->moo_readlink(env, m, buf); } +static inline int mo_changelog(const struct lu_env *env, + enum changelog_rec_type type, + int flags, struct md_object *m) +{ + LASSERT(m->mo_ops->moo_changelog); + return m->mo_ops->moo_changelog(env, type, flags, m); +} + static inline int mo_attr_set(const struct lu_env *env, struct md_object *m, const struct md_attr *at) @@ -540,6 +437,17 @@ static inline int mo_xattr_list(const struct lu_env *env, return m->mo_ops->moo_xattr_list(env, m, buf); } +static inline int mo_swap_layouts(const struct lu_env *env, + struct md_object *o1, + struct md_object *o2, __u64 flags) +{ + LASSERT(o1->mo_ops->moo_swap_layouts); + LASSERT(o2->mo_ops->moo_swap_layouts); + if (o1->mo_ops->moo_swap_layouts != o2->mo_ops->moo_swap_layouts) + return -EPERM; + return o1->mo_ops->moo_swap_layouts(env, o1, o2, flags); +} + static inline int mo_open(const struct lu_env *env, struct md_object *m, int flags) @@ -550,10 +458,11 @@ static inline int mo_open(const struct lu_env *env, static inline int mo_close(const struct lu_env *env, struct md_object *m, - struct md_attr *ma) + struct md_attr *ma, + int mode) { LASSERT(m->mo_ops->moo_close); - return m->mo_ops->moo_close(env, m, ma); + return m->mo_ops->moo_close(env, m, ma, mode); } static inline int mo_readpage(const struct lu_env *env, @@ -564,31 +473,6 @@ static inline int mo_readpage(const struct lu_env *env, return m->mo_ops->moo_readpage(env, m, rdpg); } -static inline int mo_object_create(const struct lu_env *env, - struct md_object *m, - const struct md_op_spec *spc, - struct md_attr *at) -{ - LASSERT(m->mo_ops->moo_object_create); - return m->mo_ops->moo_object_create(env, m, spc, at); -} - -static inline int mo_ref_add(const struct lu_env *env, - struct md_object *m, - const struct md_attr *ma) -{ - LASSERT(m->mo_ops->moo_ref_add); - return m->mo_ops->moo_ref_add(env, m, ma); -} - -static inline int mo_ref_del(const struct lu_env *env, - struct md_object *m, - struct md_attr *ma) -{ - LASSERT(m->mo_ops->moo_ref_del); - return m->mo_ops->moo_ref_del(env, m, ma); -} - static inline int mo_capa_get(const struct lu_env *env, struct md_object *m, struct lustre_capa *c, @@ -604,6 +488,25 @@ static inline int mo_object_sync(const struct lu_env *env, struct md_object *m) return m->mo_ops->moo_object_sync(env, m); } +static inline int mo_object_lock(const struct lu_env *env, + struct md_object *m, + struct lustre_handle *lh, + struct ldlm_enqueue_info *einfo, + union ldlm_policy_data *policy) +{ + LASSERT(m->mo_ops->moo_object_lock); + return m->mo_ops->moo_object_lock(env, m, lh, einfo, policy); +} + +static inline int mo_object_unlock(const struct lu_env *env, + struct md_object *m, + struct ldlm_enqueue_info *einfo, + union ldlm_policy_data *policy) +{ + LASSERT(m->mo_ops->moo_object_unlock); + return m->mo_ops->moo_object_unlock(env, m, einfo, policy); +} + static inline int mdo_lookup(const struct lu_env *env, struct md_object *p, const struct lu_name *lname, @@ -630,8 +533,8 @@ static inline int mdo_create(const struct lu_env *env, struct md_op_spec *spc, struct md_attr *at) { - LASSERT(c->mo_dir_ops->mdo_create); - return c->mo_dir_ops->mdo_create(env, p, lchild_name, c, spc, at); + LASSERT(p->mo_dir_ops->mdo_create); + return p->mo_dir_ops->mdo_create(env, p, lchild_name, c, spc, at); } static inline int mdo_create_data(const struct lu_env *env, @@ -658,6 +561,17 @@ static inline int mdo_rename(const struct lu_env *env, ma); } +static inline int mdo_migrate(const struct lu_env *env, + struct md_object *pobj, + struct md_object *sobj, + const struct lu_name *lname, + struct md_object *tobj, + struct md_attr *ma) +{ + LASSERT(pobj->mo_dir_ops->mdo_migrate); + return pobj->mo_dir_ops->mdo_migrate(env, pobj, sobj, lname, tobj, ma); +} + static inline int mdo_is_subdir(const struct lu_env *env, struct md_object *mo, const struct lu_fid *fid, @@ -678,50 +592,80 @@ static inline int mdo_link(const struct lu_env *env, } static inline int mdo_unlink(const struct lu_env *env, - struct md_object *p, - struct md_object *c, - const struct lu_name *lname, - struct md_attr *ma) + struct md_object *p, + struct md_object *c, + const struct lu_name *lname, + struct md_attr *ma, int no_name) { - LASSERT(c->mo_dir_ops->mdo_unlink); - return c->mo_dir_ops->mdo_unlink(env, p, c, lname, ma); + LASSERT(p->mo_dir_ops->mdo_unlink); + return p->mo_dir_ops->mdo_unlink(env, p, c, lname, ma, no_name); } -static inline int mdo_name_insert(const struct lu_env *env, - struct md_object *p, - const struct lu_name *lname, - const struct lu_fid *f, - const struct md_attr *ma) -{ - LASSERT(p->mo_dir_ops->mdo_name_insert); - return p->mo_dir_ops->mdo_name_insert(env, p, lname, f, ma); -} +/** + * Used in MDD/OUT layer for object lock rule + **/ +enum mdd_object_role { + MOR_SRC_PARENT, + MOR_SRC_CHILD, + MOR_TGT_PARENT, + MOR_TGT_CHILD, + MOR_TGT_ORPHAN +}; -static inline int mdo_name_remove(const struct lu_env *env, - struct md_object *p, - const struct lu_name *lname, - const struct md_attr *ma) -{ - LASSERT(p->mo_dir_ops->mdo_name_remove); - return p->mo_dir_ops->mdo_name_remove(env, p, lname, ma); -} +struct dt_device; + +int lustre_buf2som(void *buf, int rc, struct md_som_data *msd); +int lustre_buf2hsm(void *buf, int rc, struct md_hsm *mh); +void lustre_hsm2buf(void *buf, const struct md_hsm *mh); -static inline int mdo_rename_tgt(const struct lu_env *env, - struct md_object *p, - struct md_object *t, - const struct lu_fid *lf, - const struct lu_name *lname, - struct md_attr *ma) +enum { + UCRED_INVALID = -1, + UCRED_INIT = 0, + UCRED_OLD = 1, + UCRED_NEW = 2, +}; + +struct lu_ucred { + __u32 uc_valid; + __u32 uc_o_uid; + __u32 uc_o_gid; + __u32 uc_o_fsuid; + __u32 uc_o_fsgid; + __u32 uc_uid; + __u32 uc_gid; + __u32 uc_fsuid; + __u32 uc_fsgid; + __u32 uc_suppgids[2]; + cfs_cap_t uc_cap; + __u32 uc_umask; + struct group_info *uc_ginfo; + struct md_identity *uc_identity; + char uc_jobid[LUSTRE_JOBID_SIZE]; +}; + +struct lu_ucred *lu_ucred(const struct lu_env *env); + +struct lu_ucred *lu_ucred_check(const struct lu_env *env); + +struct lu_ucred *lu_ucred_assert(const struct lu_env *env); + +int lu_ucred_global_init(void); + +void lu_ucred_global_fini(void); + +#define md_cap_t(x) (x) + +#define MD_CAP_TO_MASK(x) (1 << (x)) + +#define md_cap_raised(c, flag) (md_cap_t(c) & MD_CAP_TO_MASK(flag)) + +/* capable() is copied from linux kernel! */ +static inline int md_capable(struct lu_ucred *uc, cfs_cap_t cap) { - if (t) { - LASSERT(t->mo_dir_ops->mdo_rename_tgt); - return t->mo_dir_ops->mdo_rename_tgt(env, p, t, lf, lname, ma); - } else { - LASSERT(p->mo_dir_ops->mdo_rename_tgt); - return p->mo_dir_ops->mdo_rename_tgt(env, p, t, lf, lname, ma); - } + if (md_cap_raised(uc->uc_cap, cap)) + return 1; + return 0; } /** @} md */ - #endif /* _LINUX_MD_OBJECT_H */