* 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/
#include <lustre_acl.h>
#include <lustre_eacl.h>
-#include <obd.h>
#include <md_object.h>
#include <dt_object.h>
-#include <linux/sched.h>
-#include <linux/capability.h>
-#ifdef HAVE_QUOTA_SUPPORT
-# include <lustre_quota.h>
-#endif
-#include <lustre_fsfilt.h>
-#include <lustre/lustre_lfsck_user.h>
-
-#ifdef HAVE_QUOTA_SUPPORT
-/* quota stuff */
-extern quota_interface_t *mds_quota_interface_ref;
-
-static inline void mdd_quota_wrapper(struct lu_attr *la, unsigned int *qids)
-{
- qids[USRQUOTA] = la->la_uid;
- qids[GRPQUOTA] = la->la_gid;
-}
-#endif
+#include <lustre_lfsck.h>
+#include <lustre_fid.h>
+#include <lustre_capa.h>
+#include <lprocfs_status.h>
+#include <lustre_log.h>
+#include <lustre_linkea.h>
/* PDO lock is unnecessary for current MDT stack because operations
* are already protected by ldlm lock */
#define CLM_PURGE 0x40000
struct mdd_changelog {
- cfs_spinlock_t mc_lock; /* for index */
- int mc_flags;
- int mc_mask;
- __u64 mc_index;
- __u64 mc_starttime;
- cfs_spinlock_t mc_user_lock;
- int mc_lastuser;
+ spinlock_t mc_lock; /* for index */
+ int mc_flags;
+ int mc_mask;
+ __u64 mc_index;
+ __u64 mc_starttime;
+ spinlock_t mc_user_lock;
+ int mc_lastuser;
};
+static inline __u64 cl_time(void) {
+ cfs_fs_time_t time;
+
+ cfs_fs_time_current(&time);
+ return (((__u64)time.tv_sec) << 30) + time.tv_nsec;
+}
+
/** Objects in .lustre dir */
struct mdd_dot_lustre_objs {
struct mdd_object *mdd_obf;
};
-extern const char lfsck_bookmark_name[];
-
-struct md_lfsck {
- cfs_mutex_t ml_mutex;
- cfs_spinlock_t ml_lock;
- struct ptlrpc_thread ml_thread;
- struct dt_object *ml_bookmark_obj;
- struct dt_object *ml_it_obj;
- __u32 ml_new_scanned;
- /* Arguments for low layer iteration. */
- __u32 ml_args;
-
- /* Raw value for LFSCK speed limit. */
- __u32 ml_speed_limit;
-
- /* Schedule for every N objects. */
- __u32 ml_sleep_rate;
-
- /* Sleep N jiffies for each schedule. */
- __u32 ml_sleep_jif;
- __u16 ml_version;
- unsigned int ml_paused:1; /* The lfsck is paused. */
-};
-
struct mdd_device {
struct md_device mdd_md_dev;
+ struct obd_export *mdd_child_exp;
struct dt_device *mdd_child;
- struct obd_device *mdd_obd_dev;
- struct lu_fid mdd_root_fid;
+ struct dt_device *mdd_bottom;
+ struct lu_fid mdd_root_fid; /* /ROOT */
+ struct lu_fid mdd_local_root_fid;
struct dt_device_param mdd_dt_conf;
struct dt_object *mdd_orphans; /* PENDING directory */
- struct dt_object *mdd_capa;
- struct dt_txn_callback mdd_txn_cb;
cfs_proc_dir_entry_t *mdd_proc_entry;
- struct lprocfs_stats *mdd_stats;
struct mdd_changelog mdd_cl;
unsigned long mdd_atime_diff;
struct mdd_object *mdd_dot_lustre;
struct mdd_dot_lustre_objs mdd_dot_lustre_objs;
- struct md_lfsck mdd_lfsck;
unsigned int mdd_sync_permission;
+ int mdd_connects;
+ struct local_oid_storage *mdd_los;
};
enum mod_flags {
ORPHAN_OBJ = 1 << 3,
};
-enum mdd_object_role {
- MOR_SRC_PARENT,
- MOR_SRC_CHILD,
- MOR_TGT_PARENT,
- MOR_TGT_CHILD,
- MOR_TGT_ORPHAN
-};
-
struct mdd_object {
struct md_object mod_obj;
/* open count */
struct lu_fid mti_fid2; /* used for be & cpu converting */
struct lu_attr mti_la;
struct lu_attr mti_la_for_fix;
+ struct lu_attr mti_pattr;
+ struct lu_attr mti_cattr;
struct md_attr mti_ma;
struct obd_info mti_oi;
- /* mti_orph_ent and mti_orph_key must be conjoint,
- * then mti_orph_ent::lde_name will be mti_orph_key. */
- struct lu_dirent mti_orph_ent;
- char mti_orph_key[NAME_MAX + 1];
+ /* mti_ent and mti_key must be conjoint,
+ * then mti_ent::lde_name will be mti_key. */
+ struct lu_dirent mti_ent;
+ char mti_key[NAME_MAX + 16];
struct obd_trans_info mti_oti;
- struct lu_buf mti_buf;
+ struct lu_buf mti_buf[4];
struct lu_buf mti_big_buf; /* biggish persistent buf */
+ struct lu_buf mti_link_buf; /* buf for link ea */
struct lu_name mti_name;
+ struct lu_name mti_name2;
struct obdo mti_oa;
char mti_xattr_buf[LUSTRE_POSIX_ACL_MAX_SIZE];
struct dt_allocation_hint mti_hint;
int mti_max_cookie_size;
struct dt_object_format mti_dof;
struct obd_quotactl mti_oqctl;
+ struct linkea_data mti_link_data;
};
extern const char orph_index_name[];
int mdd_attr_set(const struct lu_env *env, struct md_object *obj,
const struct md_attr *ma);
int mdd_attr_set_internal(const struct lu_env *env,
- struct mdd_object *obj,
- struct lu_attr *attr,
- struct thandle *handle,
- int needacl);
+ struct mdd_object *obj,
+ const struct lu_attr *attr,
+ struct thandle *handle,
+ int needacl);
int mdd_attr_check_set_internal(const struct lu_env *env,
struct mdd_object *obj,
struct lu_attr *attr,
struct md_attr *ma, struct thandle *handle);
int mdd_iattr_get(const struct lu_env *env, struct mdd_object *mdd_obj,
struct md_attr *ma);
-int mdd_attr_get_internal(const struct lu_env *env, struct mdd_object *mdd_obj,
- struct md_attr *ma);
-int mdd_attr_get_internal_locked(const struct lu_env *env,
- struct mdd_object *mdd_obj,
- struct md_attr *ma);
int mdd_object_create_internal(const struct lu_env *env, struct mdd_object *p,
- struct mdd_object *c, struct md_attr *ma,
- struct thandle *handle,
- const struct md_op_spec *spec);
+ struct mdd_object *c, struct lu_attr *attr,
+ struct thandle *handle,
+ const struct md_op_spec *spec);
int mdd_attr_check_set_internal_locked(const struct lu_env *env,
struct mdd_object *obj,
struct lu_attr *attr,
void mdd_pdo_read_unlock(const struct lu_env *env, struct mdd_object *obj,
void *dlh);
/* mdd_dir.c */
+int mdd_parent_fid(const struct lu_env *env, struct mdd_object *obj,
+ struct lu_fid *fid);
int mdd_is_subdir(const struct lu_env *env, struct md_object *mo,
const struct lu_fid *fid, struct lu_fid *sfid);
int mdd_may_create(const struct lu_env *env, struct mdd_object *pobj,
struct mdd_object *cobj, int check_perm, int check_nlink);
int mdd_may_unlink(const struct lu_env *env, struct mdd_object *pobj,
- const struct md_attr *ma);
+ const struct lu_attr *attr);
int mdd_may_delete(const struct lu_env *env, struct mdd_object *pobj,
- struct mdd_object *cobj, struct md_attr *ma,
- int check_perm, int check_empty);
+ struct mdd_object *cobj, struct lu_attr *cattr,
+ struct lu_attr *src_attr, int check_perm, int check_empty);
int mdd_unlink_sanity_check(const struct lu_env *env, struct mdd_object *pobj,
- struct mdd_object *cobj, struct md_attr *ma);
+ struct mdd_object *cobj, struct lu_attr *cattr);
int mdd_finish_unlink(const struct lu_env *env, struct mdd_object *obj,
struct md_attr *ma, struct thandle *th);
-int mdd_object_initialize(const struct lu_env *env, const struct lu_fid *pfid,
- const struct lu_name *lname, struct mdd_object *child,
- struct md_attr *ma, struct thandle *handle,
- const struct md_op_spec *spec);
+
int mdd_link_sanity_check(const struct lu_env *env, struct mdd_object *tgt_obj,
const struct lu_name *lname, struct mdd_object *src_obj);
int mdd_is_root(struct mdd_device *mdd, const struct lu_fid *fid);
int mdd_lookup(const struct lu_env *env,
struct md_object *pobj, const struct lu_name *lname,
struct lu_fid* fid, struct md_op_spec *spec);
+int mdd_links_read(const struct lu_env *env, struct mdd_object *mdd_obj,
+ struct linkea_data *ldata);
+int mdd_declare_links_add(const struct lu_env *env, struct mdd_object *mdd_obj,
+ struct thandle *handle, struct linkea_data *ldata);
+int mdd_links_write(const struct lu_env *env, struct mdd_object *mdd_obj,
+ struct linkea_data *ldata, struct thandle *handle);
struct lu_buf *mdd_links_get(const struct lu_env *env,
struct mdd_object *mdd_obj);
-void mdd_lee_unpack(const struct link_ea_entry *lee, int *reclen,
- struct lu_name *lname, struct lu_fid *pfid);
+int mdd_links_rename(const struct lu_env *env,
+ struct mdd_object *mdd_obj,
+ const struct lu_fid *oldpfid,
+ const struct lu_name *oldlname,
+ const struct lu_fid *newpfid,
+ const struct lu_name *newlname,
+ struct thandle *handle,
+ struct linkea_data *ldata,
+ int first, int check);
+int mdd_declare_links_add(const struct lu_env *env, struct mdd_object *mdd_obj,
+ struct thandle *handle, struct linkea_data *ldata);
/* mdd_lov.c */
int mdd_declare_unlink_log(const struct lu_env *env, struct mdd_object *obj,
struct mdd_thread_info *mdd_env_info(const struct lu_env *env);
+const struct lu_name *mdd_name_get_const(const struct lu_env *env,
+ const void *area, ssize_t len);
struct lu_buf *mdd_buf_get(const struct lu_env *env, void *area, ssize_t len);
const struct lu_buf *mdd_buf_get_const(const struct lu_env *env,
const void *area, ssize_t len);
int orph_index_init(const struct lu_env *env, struct mdd_device *mdd);
void orph_index_fini(const struct lu_env *env, struct mdd_device *mdd);
int orph_declare_index_insert(const struct lu_env *, struct mdd_object *,
- struct thandle *);
+ cfs_umode_t mode, struct thandle *);
int orph_declare_index_delete(const struct lu_env *, struct mdd_object *,
struct thandle *);
void lprocfs_mdd_init_vars(struct lprocfs_static_vars *lvars);
int mdd_procfs_init(struct mdd_device *mdd, const char *name);
int mdd_procfs_fini(struct mdd_device *mdd);
-void mdd_lprocfs_time_start(const struct lu_env *env);
-void mdd_lprocfs_time_end(const struct lu_env *env,
- struct mdd_device *mdd, int op);
/* mdd_object.c */
int mdd_get_flags(const struct lu_env *env, struct mdd_object *obj);
int mdd_buf_grow(const struct lu_env *env, ssize_t len);
void mdd_buf_put(struct lu_buf *buf);
+struct lu_buf *mdd_link_buf_alloc(const struct lu_env *env, ssize_t len);
+int mdd_link_buf_grow(const struct lu_env *env, ssize_t len);
extern const struct md_dir_operations mdd_dir_ops;
extern const struct md_object_operations mdd_obj_ops;
int mdd_get_default_md(struct mdd_object *mdd_obj, struct lov_mds_md *lmm);
int mdd_readpage(const struct lu_env *env, struct md_object *obj,
const struct lu_rdpg *rdpg);
-int mdd_declare_llog_record(const struct lu_env *env, struct mdd_device *mdd,
- int reclen, struct thandle *handle);
int mdd_declare_changelog_store(const struct lu_env *env,
- struct mdd_device *mdd,
- const struct lu_name *fname,
- struct thandle *handle);
-int mdd_changelog(const struct lu_env *env, enum changelog_rec_type type,
- int flags, struct md_object *obj);
+ struct mdd_device *mdd,
+ const struct lu_name *fname,
+ struct thandle *handle);
+int mdd_changelog_store(const struct lu_env *env, struct mdd_device *mdd,
+ struct llog_changelog_rec *rec, struct thandle *th);
+int mdd_changelog_data_store(const struct lu_env *env, struct mdd_device *mdd,
+ enum changelog_rec_type type, int flags,
+ struct mdd_object *mdd_obj,
+ struct thandle *handle);
+int mdd_changelog_ns_store(const struct lu_env *env, struct mdd_device *mdd,
+ enum changelog_rec_type type, unsigned flags,
+ struct mdd_object *target, struct mdd_object *parent,
+ const struct lu_name *tname, struct thandle *handle);
int mdd_declare_object_create_internal(const struct lu_env *env,
- struct mdd_object *p,
- struct mdd_object *c,
- struct md_attr *ma,
- struct thandle *handle,
- const struct md_op_spec *spec);
-/* mdd_quota.c*/
-#ifdef HAVE_QUOTA_SUPPORT
-int mdd_quota_notify(const struct lu_env *env, struct md_device *m);
-int mdd_quota_setup(const struct lu_env *env, struct md_device *m,
- void *data);
-int mdd_quota_cleanup(const struct lu_env *env, struct md_device *m);
-int mdd_quota_recovery(const struct lu_env *env, struct md_device *m);
-int mdd_quota_check(const struct lu_env *env, struct md_device *m,
- __u32 type);
-int mdd_quota_on(const struct lu_env *env, struct md_device *m,
- __u32 type);
-int mdd_quota_off(const struct lu_env *env, struct md_device *m,
- __u32 type);
-int mdd_quota_setinfo(const struct lu_env *env, struct md_device *m,
- __u32 type, __u32 id, struct obd_dqinfo *dqinfo);
-int mdd_quota_getinfo(const struct lu_env *env, const struct md_device *m,
- __u32 type, __u32 id, struct obd_dqinfo *dqinfo);
-int mdd_quota_setquota(const struct lu_env *env, struct md_device *m,
- __u32 type, __u32 id, struct obd_dqblk *dqblk);
-int mdd_quota_getquota(const struct lu_env *env, const struct md_device *m,
- __u32 type, __u32 id, struct obd_dqblk *dqblk);
-int mdd_quota_getoinfo(const struct lu_env *env, const struct md_device *m,
- __u32 type, __u32 id, struct obd_dqinfo *dqinfo);
-int mdd_quota_getoquota(const struct lu_env *env, const struct md_device *m,
- __u32 type, __u32 id, struct obd_dqblk *dqblk);
-int mdd_quota_invalidate(const struct lu_env *env, struct md_device *m,
- __u32 type);
-int mdd_quota_finvalidate(const struct lu_env *env, struct md_device *m,
- __u32 type);
-#endif
+ struct mdd_object *p,
+ struct mdd_object *c,
+ struct lu_attr *attr,
+ struct thandle *handle,
+ const struct md_op_spec *spec);
/* mdd_trans.c */
int mdd_lov_destroy(const struct lu_env *env, struct mdd_device *mdd,
struct mdd_object *obj, struct lu_attr *la);
+void mdd_object_make_hint(const struct lu_env *env, struct mdd_object *parent,
+ struct mdd_object *child, struct lu_attr *attr);
+
+static inline void mdd_object_get(struct mdd_object *o)
+{
+ lu_object_get(&o->mod_obj.mo_lu);
+}
+
static inline void mdd_object_put(const struct lu_env *env,
struct mdd_object *o)
{
int mdd_txn_start_cb(const struct lu_env *env, struct thandle *,
void *cookie);
-/* mdd_lfsck.c */
-void mdd_lfsck_set_speed(struct md_lfsck *lfsck, __u32 limit);
-int mdd_lfsck_start(const struct lu_env *env, struct md_lfsck *lfsck,
- struct lfsck_start *start);
-int mdd_lfsck_stop(const struct lu_env *env, struct md_lfsck *lfsck);
-int mdd_lfsck_setup(const struct lu_env *env, struct mdd_device *mdd);
-void mdd_lfsck_cleanup(const struct lu_env *env, struct mdd_device *mdd);
-
/* mdd_device.c */
struct lu_object *mdd_object_alloc(const struct lu_env *env,
const struct lu_object_header *hdr,
struct lu_device *d);
-struct llog_changelog_rec;
-int mdd_changelog_llog_write(struct mdd_device *mdd,
- struct llog_changelog_rec *rec,
- struct thandle *handle);
-int mdd_changelog_ext_llog_write(struct mdd_device *mdd,
- struct llog_changelog_ext_rec *rec,
- struct thandle *handle);
-int mdd_changelog_llog_cancel(struct mdd_device *mdd, long long endrec);
-int mdd_changelog_write_header(struct mdd_device *mdd, int markerflags);
-int mdd_changelog_on(struct mdd_device *mdd, int on);
-
-/* mdd_permission.c */
-#define mdd_cap_t(x) (x)
-
-#define MDD_CAP_TO_MASK(x) (1 << (x))
-
-#define mdd_cap_raised(c, flag) (mdd_cap_t(c) & MDD_CAP_TO_MASK(flag))
+int mdd_local_file_create(const struct lu_env *env, struct mdd_device *mdd,
+ const struct lu_fid *pfid, const char *name,
+ __u32 mode, struct lu_fid *fid);
-/* capable() is copied from linux kernel! */
-static inline int mdd_capable(struct md_ucred *uc, cfs_cap_t cap)
-{
- if (mdd_cap_raised(uc->mu_cap, cap))
- return 1;
- return 0;
-}
-
-int mdd_def_acl_get(const struct lu_env *env, struct mdd_object *mdd_obj,
- struct md_attr *ma);
int mdd_acl_chmod(const struct lu_env *env, struct mdd_object *o, __u32 mode,
struct thandle *handle);
int __mdd_declare_acl_init(const struct lu_env *env, struct mdd_object *obj,
int is_dir, struct thandle *handle);
-int __mdd_acl_init(const struct lu_env *env, struct mdd_object *obj,
- struct lu_buf *buf, __u32 *mode, struct thandle *handle);
+int mdd_acl_set(const struct lu_env *env, struct mdd_object *obj,
+ const struct lu_buf *buf, int fl);
+int __mdd_fix_mode_acl(const struct lu_env *env, struct lu_buf *buf,
+ __u32 *mode);
int __mdd_permission_internal(const struct lu_env *env, struct mdd_object *obj,
struct lu_attr *la, int mask, int role);
int mdd_permission(const struct lu_env *env,
int mdd_capa_get(const struct lu_env *env, struct md_object *obj,
struct lustre_capa *capa, int renewal);
+/* mdd_prepare.c */
+int mdd_compat_fixes(const struct lu_env *env, struct mdd_device *mdd);
+
+/* inline functions */
static inline int lu_device_is_mdd(struct lu_device *d)
{
return ergo(d != NULL && d->ld_ops != NULL, d->ld_ops == &mdd_lu_ops);
static inline struct obd_device *mdd2obd_dev(struct mdd_device *mdd)
{
- return mdd->mdd_obd_dev;
+ return (mdd->mdd_md_dev.md_lu_dev.ld_obd);
}
static inline struct mdd_device *mdd_obj2mdd_dev(struct mdd_object *obj)
return lu_object_fid(&obj->mod_obj.mo_lu);
}
-static inline cfs_umode_t mdd_object_type(const struct mdd_object *obj)
-{
- return lu_object_attr(&obj->mod_obj.mo_lu);
-}
-
-static inline int mdd_lov_mdsize(const struct lu_env *env,
- struct mdd_device *mdd)
+static inline int mdd_object_obf(const struct mdd_object *obj)
{
- struct obd_device *obd = mdd2obd_dev(mdd);
- return obd->u.mds.mds_max_mdsize;
+ return lu_fid_eq(mdo2fid(obj), &LU_OBF_FID);
}
-static inline int mdd_lov_cookiesize(const struct lu_env *env,
- struct mdd_device *mdd)
+static inline cfs_umode_t mdd_object_type(const struct mdd_object *obj)
{
- struct obd_device *obd = mdd2obd_dev(mdd);
- return obd->u.mds.mds_max_cookiesize;
+ return lu_object_attr(&obj->mod_obj.mo_lu);
}
static inline int mdd_is_immutable(struct mdd_object *obj)
return lu_object_exists(mdd2lu_obj(obj));
}
+static inline int mdd_object_remote(struct mdd_object *obj)
+{
+ return lu_object_remote(mdd2lu_obj(obj));
+}
+
static inline const struct lu_fid *mdd_object_fid(struct mdd_object *obj)
{
return lu_object_fid(mdd2lu_obj(obj));
}
+static inline struct seq_server_site *mdd_seq_site(struct mdd_device *mdd)
+{
+ return mdd2lu_dev(mdd)->ld_site->ld_seq_site;
+}
+
static inline struct lustre_capa *mdd_object_capa(const struct lu_env *env,
- const struct mdd_object *obj)
+ const struct mdd_object *obj)
{
- struct md_capainfo *ci = md_capainfo(env);
- const struct lu_fid *fid = mdo2fid(obj);
- int i;
+ struct lu_capainfo *lci = lu_capainfo_get(env);
+ const struct lu_fid *fid = mdo2fid(obj);
+ int i;
+
+ /* NB: in mdt_init0 */
+ if (lci == NULL)
+ return BYPASS_CAPA;
- /* NB: in mdt_init0 */
- if (!ci)
- return BYPASS_CAPA;
- for (i = 0; i < MD_CAPAINFO_MAX; i++)
- if (lu_fid_eq(&ci->mc_fid[i], fid))
- return ci->mc_capa[i];
- return NULL;
+ for (i = 0; i < LU_CAPAINFO_MAX; i++)
+ if (lu_fid_eq(&lci->lci_fid[i], fid))
+ return lci->lci_capa[i];
+ return NULL;
}
static inline void mdd_set_capainfo(const struct lu_env *env, int offset,
- const struct mdd_object *obj,
- struct lustre_capa *capa)
+ const struct mdd_object *obj,
+ struct lustre_capa *capa)
{
- struct md_capainfo *ci = md_capainfo(env);
- const struct lu_fid *fid = mdo2fid(obj);
+ struct lu_capainfo *lci = lu_capainfo_get(env);
+ const struct lu_fid *fid = mdo2fid(obj);
- LASSERT(offset >= 0 && offset <= MD_CAPAINFO_MAX);
- /* NB: in mdt_init0 */
- if (!ci)
- return;
- ci->mc_fid[offset] = *fid;
- ci->mc_capa[offset] = capa;
+ LASSERT(offset >= 0 && offset < LU_CAPAINFO_MAX);
+ /* NB: in mdt_init0 */
+ if (lci == NULL)
+ return;
+
+ lci->lci_fid[offset] = *fid;
+ lci->lci_capa[offset] = capa;
}
static inline const char *mdd_obj_dev_name(const struct mdd_object *obj)
* if the object doesn't exist yet, then it's supposed to be created
* and declaration of the creation should be enough to insert ./..
*/
- if (mdd_object_exists(obj)) {
+ /* FIXME: remote object should not be awared by MDD layer, but local
+ * creation does not declare insert ./.. (comments above), which
+ * is required by remote directory creation.
+ * This remote check should be removed when mdd_object_exists check is
+ * removed.
+ */
+ if (mdd_object_exists(obj) || mdd_object_remote(obj)) {
rc = -ENOTDIR;
if (dt_try_as_dir(env, next))
rc = dt_declare_insert(env, next,
struct thandle *handle)
{
struct dt_object *next = mdd_object_child(o);
- return next->do_ops->do_create(env, next, attr, hint, dof, handle);
+ int rc;
+
+ rc = next->do_ops->do_create(env, next, attr, hint, dof, handle);
+
+ return rc;
}
static inline