*
* You should have received a copy of the GNU General Public License
* version 2 along with this program; If not, see
- * http://www.sun.com/software/products/lustre/docs/GPLv2.pdf
- *
- * Please contact Sun Microsystems, Inc., 4150 Network Circle, Santa Clara,
- * CA 95054 USA or visit www.sun.com if you need additional information or
- * have any questions.
+ * http://www.gnu.org/licenses/gpl-2.0.html
*
* GPL HEADER END
*/
* Copyright (c) 2009, 2010, Oracle and/or its affiliates. All rights reserved.
* Use is subject to license terms.
*
- * Copyright (c) 2012, 2015, Intel Corporation.
+ * Copyright (c) 2012, 2016, Intel Corporation.
*/
/*
* This file is part of Lustre, http://www.lustre.org/
#include <sys/zfs_znode.h>
#include <sys/zap.h>
#include <sys/dbuf.h>
+#include <sys/dmu_objset.h>
/**
* By design including kmem.h overrides the Linux slab interfaces to provide
/* Statfs {minimum, safe estimate, and maximum} dnodes per block */
#define OSD_DNODE_MIN_BLKSHIFT (DNODES_PER_BLOCK_SHIFT)
-#define OSD_DNODE_EST_BLKSHIFT (DNODES_PER_BLOCK_SHIFT >> 1)
-#define OSD_DNODE_EST_COUNT 1024
+#define OSD_DNODE_EST_BLKSHIFT (12) /* est 4KB/dnode */
+#define OSD_DNODE_EST_COUNT 4096
#define OSD_GRANT_FOR_LOCAL_OIDS (2ULL << 20) /* 2MB for last_rcvd, ... */
uint64_t ctime[2];
};
+
+#define OSD_INS_CACHE_SIZE 8
+
+/* OI cache entry */
+struct osd_idmap_cache {
+ struct osd_device *oic_dev;
+ struct lu_fid oic_fid;
+ /** max 2^48 dnodes per dataset, avoid spilling into another word */
+ uint64_t oic_dnode:DN_MAX_OBJECT_SHIFT,
+ oic_remote:1; /* FID isn't local */
+};
+
/* max.number of regular attrubites the callers may ask for */
#define OSD_MAX_IN_BULK 13
struct lquota_id_info oti_qi;
struct lu_seq_range oti_seq_range;
- struct lu_buf oti_xattr_lbuf;
+
+ /* dedicated OI cache for insert (which needs inum) */
+ struct osd_idmap_cache *oti_ins_cache;
+ int oti_ins_cache_size;
+ int oti_ins_cache_used;
};
extern struct lu_context_key osd_key;
struct list_head ot_stop_dcb_list;
struct list_head ot_unlinked_list;
struct list_head ot_sa_list;
+ struct semaphore ot_sa_lock;
dmu_tx_t *ot_tx;
struct lquota_trans ot_quota_trans;
__u32 ot_write_commit:1,
struct osd_oi {
char oi_name[OSD_OI_NAME_SIZE]; /* unused */
uint64_t oi_zapid;
- dmu_buf_t *oi_db;
+ dnode_t *oi_dn;
};
struct osd_seq {
unsigned int od_dev_set_rdonly:1, /**< osd_ro() called */
od_prop_rdonly:1, /**< ZFS property readonly */
od_xattr_in_sa:1,
- od_quota_iused_est:1,
od_is_ost:1,
od_posix_acl:1;
* Not modified concurrently (either setup early during object
* creation, or assigned by osd_object_create() under write lock).
*/
- dmu_buf_t *oo_db;
+ dnode_t *oo_dn;
sa_handle_t *oo_sa_hdl;
nvlist_t *oo_sa_xattr;
struct list_head oo_sa_linkage;
struct rw_semaphore oo_sem;
/* to serialize some updates: destroy vs. others,
- * xattr_set, etc */
+ * xattr_set, object block size change etc */
struct rw_semaphore oo_guard;
/* protected by oo_guard */
uint64_t oo_xattr;
enum osd_destroy_type oo_destroy;
- __u32 oo_destroyed:1,
- oo_late_xattr:1;
+ __u32 oo_destroyed:1;
/* the i_flags in LMA */
__u32 oo_lma_flags;
- /* record size for index file */
- unsigned char oo_keysize;
- unsigned char oo_recsize;
- unsigned char oo_recusize; /* unit size */
+ union {
+ int oo_ea_in_bonus; /* EA bytes we expect */
+ struct {
+ /* record size for index file */
+ unsigned char oo_keysize;
+ unsigned char oo_recsize;
+ unsigned char oo_recusize; /* unit size */
+ };
+ };
};
int osd_statfs(const struct lu_env *, struct dt_device *, struct obd_statfs *);
extern const struct dt_index_operations osd_acct_index_ops;
-uint64_t osd_quota_fid2dmu(const struct lu_fid *fid);
+int osd_quota_fid2dmu(const struct lu_fid *fid, uint64_t *oid);
extern struct lu_device_operations osd_lu_ops;
extern struct dt_index_operations osd_dir_ops;
int osd_declare_quota(const struct lu_env *env, struct osd_device *osd,
return 1;
}
+/**
+ * Put the osd object once done with it.
+ *
+ * \param obj osd object that needs to be put
+ */
+static inline void osd_object_put(const struct lu_env *env,
+ struct osd_object *obj)
+{
+ dt_object_put(env, &obj->oo_dt);
+}
+
static inline int osd_object_invariant(const struct lu_object *l)
{
return osd_invariant(osd_obj(l));
/* osd_object.c */
extern char *osd_obj_tag;
-void osd_object_sa_dirty_rele(const struct lu_env *env, struct osd_thandle *oh);
-void osd_object_sa_dirty_add(struct osd_object *obj, struct osd_thandle *oh);
-int __osd_obj2dbuf(const struct lu_env *env, objset_t *os,
- uint64_t oid, dmu_buf_t **dbp);
+void osd_object_sa_dirty_rele(struct osd_thandle *oh);
+int __osd_obj2dnode(const struct lu_env *env, objset_t *os,
+ uint64_t oid, dnode_t **dnp);
struct lu_object *osd_object_alloc(const struct lu_env *env,
const struct lu_object_header *hdr,
struct lu_device *d);
int osd_object_sa_update(struct osd_object *obj, sa_attr_type_t type,
void *buf, uint32_t buflen, struct osd_thandle *oh);
int __osd_zap_create(const struct lu_env *env, struct osd_device *osd,
- dmu_buf_t **zap_dbp, dmu_tx_t *tx, struct lu_attr *la,
- uint64_t parent, zap_flags_t flags);
+ dnode_t **zap_dnp, dmu_tx_t *tx, struct lu_attr *la,
+ zap_flags_t flags);
int __osd_object_create(const struct lu_env *env, struct osd_object *obj,
- dmu_buf_t **dbp, dmu_tx_t *tx, struct lu_attr *la,
- uint64_t parent);
+ dnode_t **dnp, dmu_tx_t *tx, struct lu_attr *la);
+int __osd_attr_init(const struct lu_env *env, struct osd_device *osd,
+ sa_handle_t *sa_hdl, dmu_tx_t *tx,
+ struct lu_attr *la, uint64_t parent);
/* osd_oi.c */
int osd_oi_init(const struct lu_env *env, struct osd_device *o);
int osd_fid_lookup(const struct lu_env *env,
struct osd_device *, const struct lu_fid *, uint64_t *);
uint64_t osd_get_name_n_idx(const struct lu_env *env, struct osd_device *osd,
- const struct lu_fid *fid, char *buf);
+ const struct lu_fid *fid, char *buf, int bufsize);
int osd_options_init(void);
int osd_ost_seq_exists(const struct lu_env *env, struct osd_device *osd,
__u64 seq);
+int osd_idc_find_and_init(const struct lu_env *env, struct osd_device *osd,
+ struct osd_object *obj);
+struct osd_idmap_cache *osd_idc_find_or_init(const struct lu_env *env,
+ struct osd_device *osd,
+ const struct lu_fid *fid);
+struct osd_idmap_cache *osd_idc_find(const struct lu_env *env,
+ struct osd_device *osd,
+ const struct lu_fid *fid);
+
/* osd_index.c */
int osd_index_try(const struct lu_env *env, struct dt_object *dt,
const struct dt_index_features *feat);
uint64_t id, uint64_t dirhash);
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);
/* osd_xattr.c */
-int __osd_sa_xattr_update(const struct lu_env *env, struct osd_object *obj,
- struct osd_thandle *oh);
int __osd_xattr_load(struct osd_device *osd, uint64_t dnode,
nvlist_t **sa_xattr);
int __osd_xattr_get_large(const struct lu_env *env, struct osd_device *osd,
int __osd_xattr_set(const struct lu_env *env, struct osd_object *obj,
const struct lu_buf *buf, const char *name, int fl,
struct osd_thandle *oh);
+int __osd_sa_xattr_update(const struct lu_env *env, struct osd_object *obj,
+ struct osd_thandle *oh);
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,
if (unlikely(!dt_object_exists(&obj->oo_dt) || obj->oo_destroyed))
return -ENOENT;
- LASSERT(obj->oo_db);
+ LASSERT(obj->oo_dn);
if (osd_obj2dev(obj)->od_xattr_in_sa) {
rc = __osd_sa_xattr_set(env, obj, buf, name, fl, oh);
if (rc == -EFBIG)
DMU_OT_SA, DN_BONUS_SIZE(dnodesize),
dnodesize, tx);
}
+
+static inline int
+osd_obj_bonuslen(struct osd_object *obj)
+{
+ int bonuslen = DN_BONUS_SIZE(DNODE_MIN_SIZE);
+
+ if (obj->oo_dn != NULL && obj->oo_dn->dn_num_slots != 0) {
+ bonuslen = DN_SLOTS_TO_BONUSLEN(obj->oo_dn->dn_num_slots);
+ } else {
+ objset_t *os = osd_dtobj2objset(&obj->oo_dt);
+ int dnodesize;
+
+ if (os != NULL) {
+ dnodesize = dmu_objset_dnodesize(os);
+ if (dnodesize != 0)
+ bonuslen = DN_BONUS_SIZE(dnodesize);
+ }
+ }
+
+ return bonuslen;
+}
#else
static inline uint64_t
osd_dmu_object_alloc(objset_t *os, dmu_object_type_t objtype, int blocksize,
indirect_blockshift, DMU_OT_SA,
DN_MAX_BONUSLEN, tx);
}
+
+static inline int
+osd_obj_bonuslen(struct osd_object *obj)
+{
+ return DN_MAX_BONUSLEN;
+}
#endif /* HAVE_DMU_OBJECT_ALLOC_DNSIZE */
+#ifdef HAVE_DMU_PREFETCH_6ARG
+#define osd_dmu_prefetch(os, obj, lvl, off, len, pri) \
+ dmu_prefetch((os), (obj), (lvl), (off), (len), (pri))
+#else
+#define osd_dmu_prefetch(os, obj, lvl, off, len, pri) \
+ dmu_prefetch((os), (obj), (lvl), (off))
+#endif
+
+static inline void osd_dnode_rele(dnode_t *dn)
+{
+ dmu_buf_impl_t *db;
+ LASSERT(dn);
+ LASSERT(dn->dn_bonus);
+ db = dn->dn_bonus;
+
+ DB_DNODE_EXIT(db);
+ dmu_buf_rele(&db->db, osd_obj_tag);
+}
+
+#ifdef HAVE_DMU_USEROBJ_ACCOUNTING
+
+#define OSD_DMU_USEROBJ_PREFIX DMU_OBJACCT_PREFIX
+
+static inline bool osd_dmu_userobj_accounting_available(struct osd_device *osd)
+{
+ if (unlikely(dmu_objset_userobjspace_upgradable(osd->od_os)))
+ dmu_objset_userobjspace_upgrade(osd->od_os);
+
+ return dmu_objset_userobjspace_present(osd->od_os);
+}
+#else
+
+#define OSD_DMU_USEROBJ_PREFIX "obj-"
+
+static inline bool osd_dmu_userobj_accounting_available(struct osd_device *osd)
+{
+ return false;
+}
+#endif /* #ifdef HAVE_DMU_USEROBJ_ACCOUNTING */
+
#endif /* _OSD_INTERNAL_H */