*
* 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) 2007, 2010, Oracle and/or its affiliates. All rights reserved.
* Use is subject to license terms.
*
- * Copyright (c) 2011, 2015, Intel Corporation.
+ * Copyright (c) 2011, 2017, Intel Corporation.
*/
/*
* This file is part of Lustre, http://www.lustre.org/
unsigned ddp_max_extent_blks;
/* per-extent insertion overhead to be used by client for grant
* calculation */
- unsigned ddp_extent_tax;
+ unsigned int ddp_extent_tax;
+ unsigned int ddp_brw_size; /* optimal RPC size */
+ /* T10PI checksum type, zero if not supported */
+ enum cksum_types ddp_t10_cksum_type;
+ bool ddp_has_lseek_data_hole;
};
/**
*/
int (*dt_statfs)(const struct lu_env *env,
struct dt_device *dev,
- struct obd_statfs *osfs);
+ struct obd_statfs *osfs,
+ struct obd_statfs_info *info);
/**
* Create transaction.
struct dt_device_param *param);
/**
+ * Return device's super block.
+ *
+ * \param[in] dev dt device
+ */
+ struct super_block *(*dt_mnt_sb_get)(const struct dt_device *dev);
+
+ /**
* Sync the device.
*
* Sync all the cached state (dirty buffers, pages, etc) to the
struct dt_device *dev);
/**
+ * Wait pending quota update finish
+ *
+ * There might be a window that quota usage has been updated,
+ * but commit callback to reduce pending write have not been
+ * finished, this is used to wait all pending update done.
+ *
+ * \param[in] dev dt device
+ */
+ void (*dt_wait_quota_pending)(struct dt_device *dev);
+
+ /**
* Start transaction commit asynchronously.
*
};
enum dt_index_flags {
- /** index supports variable sized keys */
- DT_IND_VARKEY = 1 << 0,
- /** index supports variable sized records */
- DT_IND_VARREC = 1 << 1,
- /** index can be modified */
- DT_IND_UPDATE = 1 << 2,
- /** index supports records with non-unique (duplicate) keys */
- DT_IND_NONUNQ = 1 << 3,
- /**
- * index support fixed-size keys sorted with natural numerical way
- * and is able to return left-side value if no exact value found
- */
- DT_IND_RANGE = 1 << 4,
+ /** index supports variable sized keys */
+ DT_IND_VARKEY = BIT(0),
+ /** index supports variable sized records */
+ DT_IND_VARREC = BIT(1),
+ /** index can be modified */
+ DT_IND_UPDATE = BIT(2),
+ /** index supports records with non-unique (duplicate) keys */
+ DT_IND_NONUNQ = BIT(3),
+ /**
+ * index support fixed-size keys sorted with natural numerical way
+ * and is able to return left-side value if no exact value found
+ */
+ DT_IND_RANGE = BIT(4),
+};
+
+/* for dt_read_lock() and dt_write_lock() object lock rule */
+enum dt_object_role {
+ DT_SRC_PARENT,
+ DT_SRC_CHILD,
+ DT_TGT_PARENT,
+ DT_TGT_CHILD,
+ DT_TGT_ORPHAN,
+ DT_LASTID,
};
/**
*/
extern const struct dt_index_features dt_directory_features;
extern const struct dt_index_features dt_otable_features;
-extern const struct dt_index_features dt_lfsck_orphan_features;
-extern const struct dt_index_features dt_lfsck_features;
+extern const struct dt_index_features dt_lfsck_layout_orphan_features;
+extern const struct dt_index_features dt_lfsck_layout_dangling_features;
+extern const struct dt_index_features dt_lfsck_namespace_features;
/* index features supported by the accounting objects */
extern const struct dt_index_features dt_acct_features;
const void *dah_eadata;
int dah_eadata_len;
__u32 dah_mode;
+ int dah_append_stripes;
+ char *dah_append_pool;
};
/**
union ldlm_policy_data;
+struct md_layout_change;
+
/**
* A dt_object provides common operations to create and destroy
* objects and to manage regular and extended attributes.
* \retval 0 on success
* \retval negative negated errno on error
*/
- int (*do_object_sync)(const struct lu_env *env,
- struct dt_object *obj,
- __u64 start,
- __u64 end);
+ int (*do_object_sync)(const struct lu_env *env, struct dt_object *obj,
+ __u64 start, __u64 end);
/**
* Lock object.
struct dt_object *dt,
struct ldlm_enqueue_info *einfo,
union ldlm_policy_data *policy);
+
+ /**
+ * Invalidate attribute cache.
+ *
+ * This method invalidate attribute cache of the object, which is on OSP
+ * only.
+ *
+ * \param[in] env execution envionment for this thread
+ * \param[in] dt object
+ *
+ * \retval 0 on success
+ * \retval negative negated errno on error
+ */
+ int (*do_invalidate)(const struct lu_env *env, struct dt_object *dt);
+
+ /**
+ * Declare intention to instaintiate extended layout component.
+ *
+ * \param[in] env execution environment
+ * \param[in] dt DT object
+ * \param[in] layout data structure to describe the changes to
+ * the DT object's layout
+ * \param[in] buf buffer containing client's lovea or empty
+ *
+ * \retval 0 success
+ * \retval -ne error code
+ */
+ int (*do_declare_layout_change)(const struct lu_env *env,
+ struct dt_object *dt,
+ struct md_layout_change *mlc,
+ struct thandle *th);
+
+ /**
+ * Client is trying to write to un-instantiated layout component.
+ *
+ * \param[in] env execution environment
+ * \param[in] dt DT object
+ * \param[in] layout data structure to describe the changes to
+ * the DT object's layout
+ * \param[in] buf buffer containing client's lovea or empty
+ *
+ * \retval 0 success
+ * \retval -ne error code
+ */
+ int (*do_layout_change)(const struct lu_env *env, struct dt_object *dt,
+ struct md_layout_change *mlc,
+ struct thandle *th);
+};
+
+enum dt_bufs_type {
+ DT_BUFS_TYPE_READ = 0x0000,
+ DT_BUFS_TYPE_WRITE = 0x0001,
+ DT_BUFS_TYPE_READAHEAD = 0x0002,
+ DT_BUFS_TYPE_LOCAL = 0x0004,
};
/**
* \param[in] pos position in the object to start
* \param[out] pos \a pos + bytes written
* \param[in] th transaction handle
- * \param[in] ignore unused (was used to request quota ignorance)
*
* \retval positive bytes written on success
* \retval negative negated errno on error
struct dt_object *dt,
const struct lu_buf *buf,
loff_t *pos,
- struct thandle *th,
- int ignore);
+ struct thandle *th);
/**
* Return buffers for data.
* \param[in] pos position in the object to start
* \param[in] len size of region in bytes
* \param[out] lb array of descriptors to fill
+ * \param[in] maxlnb max slots in @lnb array
* \param[in] rw 0 if used to read, 1 if used for write
*
* \retval positive number of descriptors on success
loff_t pos,
ssize_t len,
struct niobuf_local *lb,
- int rw);
+ int maxlnb,
+ enum dt_bufs_type rw);
/**
* Release reference granted by ->dbo_bufs_get().
* then the method should maintain space accounting for the given
* credentials.
*
+ * user_size parameter is the apparent size of the file, ie the size
+ * of the clear text version of the file. It can differ from the actual
+ * amount of valuable data received when a file is encrypted,
+ * because encrypted pages always contain PAGE_SIZE bytes of data,
+ * even if clear text data is only a few bytes.
+ * In case of encrypted file, apparent size will be stored as the inode
+ * size, so that servers return to clients an object size they can use
+ * to determine clear text size.
+ *
* \param[in] env execution environment for this thread
* \param[in] dt object
* \param[in] lb array of descriptors for the buffers
* \param[in] nr size of the array
* \param[in] th transaction handle
+ * \param[in] user_size apparent size
*
* \retval 0 on success
* \retval negative negated errno on error
struct dt_object *dt,
struct niobuf_local *lb,
int nr,
- struct thandle *th);
+ struct thandle *th,
+ __u64 user_size);
/**
* Return logical to physical block mapping for a given extent
__u64 start,
__u64 end,
enum lu_ladvise_type advice);
+
+ /**
+ * Declare intention to preallocate space for an object
+ *
+ * \param[in] env execution environment for this thread
+ * \param[in] dt object
+ * \param[in] th transaction handle
+ *
+ * \retval 0 on success
+ * \retval negative negated errno on error
+ */
+ int (*dbo_declare_fallocate)(const struct lu_env *env,
+ struct dt_object *dt, __u64 start,
+ __u64 end, int mode, struct thandle *th);
+ /**
+ * Allocate specified region for an object
+ *
+ * \param[in] env execution environment for this thread
+ * \param[in] dt object
+ * \param[in] start the start of the region to allocate
+ * \param[in] end the end of the region to allocate
+ * \param[in] mode fallocate mode
+ * \param[in] th transaction handle
+ *
+ * \retval 0 on success
+ * \retval negative negated errno on error
+ */
+ int (*dbo_fallocate)(const struct lu_env *env,
+ struct dt_object *dt,
+ __u64 start,
+ __u64 end,
+ int mode,
+ struct thandle *th);
+ /**
+ * Do SEEK_HOLE/SEEK_DATA request on object
+ *
+ * \param[in] env execution environment for this thread
+ * \param[in] dt object
+ * \param[in] offset the offset to start seek from
+ * \param[in] whence seek mode, SEEK_HOLE or SEEK_DATA
+ *
+ * \retval hole/data offset on success
+ * \retval negative negated errno on error
+ */
+ loff_t (*dbo_lseek)(const struct lu_env *env, struct dt_object *dt,
+ loff_t offset, int whence);
};
/**
* \param[in] rec buffer storing value
* \param[in] key key
* \param[in] th transaction handle
- * \param[in] ignore unused (was used to request quota ignorance)
*
* \retval 0 on success
* \retval negative negated errno on error
struct dt_object *dt,
const struct dt_rec *rec,
const struct dt_key *key,
- struct thandle *th,
- int ignore);
+ struct thandle *th);
/**
* Declare intention to delete a key/value from an index.
* single-threaded start-up shut-down procedures.
*/
struct list_head dd_txn_callbacks;
- unsigned int dd_record_fid_accessed:1;
+ unsigned int dd_record_fid_accessed:1,
+ dd_rdonly:1;
+
+ /* sysfs and debugfs handling */
+ struct dentry *dd_debugfs_entry;
+
+ const struct attribute **dd_def_attrs;
+ struct kobject dd_kobj;
+ struct kobj_type dd_ktype;
+ struct completion dd_kobj_unregister;
};
int dt_device_init(struct dt_device *dev, struct lu_device_type *t);
static inline struct dt_device * lu2dt_dev(struct lu_device *l)
{
- LASSERT(lu_device_is_dt(l));
- return container_of0(l, struct dt_device, dd_lu_dev);
+ LASSERT(lu_device_is_dt(l));
+ return container_of_safe(l, struct dt_device, dd_lu_dev);
}
struct dt_object {
static inline struct dt_object *lu2dt(struct lu_object *l)
{
- LASSERT(l == NULL || IS_ERR(l) || lu_device_is_dt(l->lo_dev));
- return container_of0(l, struct dt_object, do_lu);
+ LASSERT(l == NULL || IS_ERR(l) || lu_device_is_dt(l->lo_dev));
+ return container_of_safe(l, struct dt_object, do_lu);
}
int dt_object_init(struct dt_object *obj,
static inline struct dt_object *lu2dt_obj(struct lu_object *o)
{
LASSERT(ergo(o != NULL, lu_device_is_dt(o->lo_dev)));
- return container_of0(o, struct dt_object, do_lu);
+ return container_of_safe(o, struct dt_object, do_lu);
}
static inline struct dt_object *dt_object_child(struct dt_object *o)
{
- return container_of0(lu_object_next(&(o)->do_lu),
- struct dt_object, do_lu);
+ return container_of(lu_object_next(&(o)->do_lu),
+ struct dt_object, do_lu);
}
/**
* top thandle here for now, will fix it when we have better
* callback mechanism */
struct thandle *th_top;
- /** context for this transaction, tag is LCT_TX_HANDLE */
- struct lu_context th_ctx;
-
- /** additional tags (layers can add in declare) */
- __u32 th_tags;
/** the last operation result in this transaction.
* this value is used in recovery */
unsigned int th_sync:1,
/* local transation, no need to inform other layers */
th_local:1,
- /* Whether we need wait the transaction to be submitted */
+ /* Whether we need wait the transaction to be submitted
+ * (send to remote target) */
th_wait_submit:1,
- /* complex transaction which will track updates on all targets */
- th_complex:1;
+ /* complex transaction which will track updates on all targets,
+ * including OSTs */
+ th_complex:1,
+ /* whether ignore quota */
+ th_ignore_quota:1;
};
/**
struct thandle *txn, void *cookie);
int (*dtc_txn_stop)(const struct lu_env *env,
struct thandle *txn, void *cookie);
- void (*dtc_txn_commit)(struct thandle *txn, void *cookie);
void *dtc_cookie;
__u32 dtc_tag;
struct list_head dtc_linkage;
int dt_txn_hook_start(const struct lu_env *env,
struct dt_device *dev, struct thandle *txn);
int dt_txn_hook_stop(const struct lu_env *env, struct thandle *txn);
-void dt_txn_hook_commit(struct thandle *txn);
int dt_try_as_dir(const struct lu_env *env, struct dt_object *obj);
return NULL;
}
+static inline void dt_object_put(const struct lu_env *env,
+ struct dt_object *dto)
+{
+ lu_object_put(env, &dto->do_lu);
+}
+
+static inline void dt_object_put_nocache(const struct lu_env *env,
+ struct dt_object *dto)
+{
+ lu_object_put_nocache(env, &dto->do_lu);
+}
+
int local_oid_storage_init(const struct lu_env *env, struct dt_device *dev,
const struct lu_fid *first_fid,
struct local_oid_storage **los);
struct dt_object *o,
struct lu_attr *attr, struct dt_object_format *dof,
struct thandle *th);
+struct dt_object *local_file_find(const struct lu_env *env,
+ struct local_oid_storage *los,
+ struct dt_object *parent,
+ const char *name);
struct dt_object *local_file_find_or_create(const struct lu_env *env,
struct local_oid_storage *los,
struct dt_object *parent,
return o->do_ops->do_object_sync(env, o, start, end);
}
+static inline int dt_fid_alloc(const struct lu_env *env,
+ struct dt_device *d,
+ struct lu_fid *fid,
+ struct lu_object *parent,
+ const struct lu_name *name)
+{
+ struct lu_device *l = dt2lu_dev(d);
+
+ return l->ld_ops->ldo_fid_alloc(env, l, fid, parent, name);
+}
+
int dt_declare_version_set(const struct lu_env *env, struct dt_object *o,
struct thandle *th);
void dt_version_set(const struct lu_env *env, struct dt_object *o,
LASSERTF(dt != NULL, "dt is NULL when we want to write record\n");
LASSERT(th != NULL);
- LASSERT(dt->do_body_ops);
+ LASSERTF(dt->do_body_ops, DFID" doesn't exit\n",
+ PFID(lu_object_fid(&dt->do_lu)));
LASSERT(dt->do_body_ops->dbo_declare_write);
rc = dt->do_body_ops->dbo_declare_write(env, dt, buf, pos, th);
return rc;
static inline int dt_bufs_get(const struct lu_env *env, struct dt_object *d,
struct niobuf_remote *rnb,
- struct niobuf_local *lnb, int rw)
+ struct niobuf_local *lnb, int maxlnb,
+ enum dt_bufs_type rw)
{
LASSERT(d);
LASSERT(d->do_body_ops);
LASSERT(d->do_body_ops->dbo_bufs_get);
return d->do_body_ops->dbo_bufs_get(env, d, rnb->rnb_offset,
- rnb->rnb_len, lnb, rw);
+ rnb->rnb_len, lnb, maxlnb, rw);
}
static inline int dt_bufs_put(const struct lu_env *env, struct dt_object *d,
static inline int dt_write_commit(const struct lu_env *env,
- struct dt_object *d, struct niobuf_local *lnb,
- int n, struct thandle *th)
+ struct dt_object *d, struct niobuf_local *lnb,
+ int n, struct thandle *th, __u64 size)
{
- LASSERT(d);
- LASSERT(d->do_body_ops);
- LASSERT(d->do_body_ops->dbo_write_commit);
- return d->do_body_ops->dbo_write_commit(env, d, lnb, n, th);
+ LASSERT(d);
+ LASSERT(d->do_body_ops);
+ LASSERT(d->do_body_ops->dbo_write_commit);
+ return d->do_body_ops->dbo_write_commit(env, d, lnb, n, th, size);
}
static inline int dt_read_prep(const struct lu_env *env, struct dt_object *d,
static inline ssize_t dt_write(const struct lu_env *env, struct dt_object *dt,
const struct lu_buf *buf, loff_t *pos,
- struct thandle *th, int rq)
+ struct thandle *th)
{
LASSERT(dt);
LASSERT(dt->do_body_ops);
LASSERT(dt->do_body_ops->dbo_write);
- return dt->do_body_ops->dbo_write(env, dt, buf, pos, th, rq);
+ return dt->do_body_ops->dbo_write(env, dt, buf, pos, th);
}
static inline int dt_declare_punch(const struct lu_env *env,
return dt->do_body_ops->dbo_ladvise(env, dt, start, end, advice);
}
+static inline int dt_declare_fallocate(const struct lu_env *env,
+ struct dt_object *dt, __u64 start,
+ __u64 end, int mode, struct thandle *th)
+{
+ LASSERT(dt);
+ if (!dt->do_body_ops)
+ return -EOPNOTSUPP;
+ LASSERT(dt->do_body_ops);
+ LASSERT(dt->do_body_ops->dbo_declare_fallocate);
+ return dt->do_body_ops->dbo_declare_fallocate(env, dt, start, end,
+ mode, th);
+}
+
+static inline int dt_falloc(const struct lu_env *env, struct dt_object *dt,
+ __u64 start, __u64 end, int mode,
+ struct thandle *th)
+{
+ LASSERT(dt);
+ if (!dt->do_body_ops)
+ return -EOPNOTSUPP;
+ LASSERT(dt->do_body_ops);
+ LASSERT(dt->do_body_ops->dbo_fallocate);
+ return dt->do_body_ops->dbo_fallocate(env, dt, start, end, mode, th);
+}
+
static inline int dt_fiemap_get(const struct lu_env *env, struct dt_object *d,
struct fiemap *fm)
{
return d->do_body_ops->dbo_fiemap_get(env, d, fm);
}
+static inline loff_t dt_lseek(const struct lu_env *env, struct dt_object *d,
+ loff_t offset, int whence)
+{
+ LASSERT(d);
+ if (d->do_body_ops == NULL)
+ return -EPROTO;
+ if (d->do_body_ops->dbo_lseek == NULL)
+ return -EOPNOTSUPP;
+ return d->do_body_ops->dbo_lseek(env, d, offset, whence);
+}
+
+static inline int dt_statfs_info(const struct lu_env *env,
+ struct dt_device *dev,
+ struct obd_statfs *osfs,
+ struct obd_statfs_info *info)
+{
+ LASSERT(dev);
+ LASSERT(dev->dd_ops);
+ LASSERT(dev->dd_ops->dt_statfs);
+ return dev->dd_ops->dt_statfs(env, dev, osfs, info);
+}
+
static inline int dt_statfs(const struct lu_env *env, struct dt_device *dev,
- struct obd_statfs *osfs)
+ struct obd_statfs *osfs)
{
- LASSERT(dev);
- LASSERT(dev->dd_ops);
- LASSERT(dev->dd_ops->dt_statfs);
- return dev->dd_ops->dt_statfs(env, dev, osfs);
+ return dt_statfs_info(env, dev, osfs, NULL);
}
static inline int dt_root_get(const struct lu_env *env, struct dt_device *dev,
return dev->dd_ops->dt_conf_get(env, dev, param);
}
+static inline struct super_block *dt_mnt_sb_get(const struct dt_device *dev)
+{
+ LASSERT(dev);
+ LASSERT(dev->dd_ops);
+ if (dev->dd_ops->dt_mnt_sb_get)
+ return dev->dd_ops->dt_mnt_sb_get(dev);
+
+ return ERR_PTR(-EOPNOTSUPP);
+}
+
static inline int dt_sync(const struct lu_env *env, struct dt_device *dev)
{
LASSERT(dev);
return dev->dd_ops->dt_ro(env, dev);
}
+static inline void dt_wait_quota_pending(struct dt_device *dev)
+{
+ LASSERT(dev);
+ LASSERT(dev->dd_ops);
+ if (dev->dd_ops->dt_wait_quota_pending)
+ dev->dd_ops->dt_wait_quota_pending(dev);
+}
+
static inline int dt_declare_insert(const struct lu_env *env,
struct dt_object *dt,
const struct dt_rec *rec,
}
static inline int dt_insert(const struct lu_env *env,
- struct dt_object *dt,
- const struct dt_rec *rec,
- const struct dt_key *key,
- struct thandle *th,
- int noquota)
+ struct dt_object *dt,
+ const struct dt_rec *rec,
+ const struct dt_key *key,
+ struct thandle *th)
{
LASSERT(dt);
LASSERT(dt->do_index_ops);
if (CFS_FAULT_CHECK(OBD_FAIL_DT_INSERT))
return cfs_fail_err;
- return dt->do_index_ops->dio_insert(env, dt, rec, key, th, noquota);
+ return dt->do_index_ops->dio_insert(env, dt, rec, key, th);
}
static inline int dt_declare_xattr_del(const struct lu_env *env,
return dt->do_ops->do_xattr_list(env, dt, buf);
}
+static inline int dt_invalidate(const struct lu_env *env, struct dt_object *dt)
+{
+ LASSERT(dt);
+ LASSERT(dt->do_ops);
+ LASSERT(dt->do_ops->do_invalidate);
+
+ return dt->do_ops->do_invalidate(env, dt);
+}
+
static inline int dt_declare_delete(const struct lu_env *env,
struct dt_object *dt,
const struct dt_key *key,
return ret;
}
+static inline int dt_declare_layout_change(const struct lu_env *env,
+ struct dt_object *o,
+ struct md_layout_change *mlc,
+ struct thandle *th)
+{
+ LASSERT(o);
+ LASSERT(o->do_ops);
+ LASSERT(o->do_ops->do_declare_layout_change);
+ return o->do_ops->do_declare_layout_change(env, o, mlc, th);
+}
+
+static inline int dt_layout_change(const struct lu_env *env,
+ struct dt_object *o,
+ struct md_layout_change *mlc,
+ struct thandle *th)
+{
+ LASSERT(o);
+ LASSERT(o->do_ops);
+ LASSERT(o->do_ops->do_layout_change);
+ return o->do_ops->do_layout_change(env, o, mlc, th);
+}
+
struct dt_find_hint {
struct lu_fid *dfh_fid;
struct dt_device *dfh_dt;
int dt_global_init(void);
void dt_global_fini(void);
+int dt_tunables_init(struct dt_device *dt, struct obd_type *type,
+ const char *name, struct ldebugfs_vars *list);
+int dt_tunables_fini(struct dt_device *dt);
# ifdef CONFIG_PROC_FS
int lprocfs_dt_blksize_seq_show(struct seq_file *m, void *v);