* Author: Alex Zhuravlev <alexey.zhuravlev@intel.com>
* Author: Mikhail Pershin <mike.pershin@intel.com>
*/
+/**
+ * The Logical Object Device (LOD) layer manages access to striped
+ * objects (both regular files and directories). It implements the DT
+ * device and object APIs and is responsible for creating, storing,
+ * and loading striping information as an extended attribute of the
+ * underlying OSD object. LOD is the server side analog of the LOV and
+ * LMV layers on the client side.
+ *
+ * Metadata LU object stack (layers of the same compound LU object,
+ * all have the same FID):
+ *
+ * MDT
+ * | MD API
+ * MDD
+ * | DT API
+ * LOD
+ * / \ DT API
+ * OSD OSP
+ *
+ * During LOD object initialization the localness or remoteness of the
+ * object FID dictates the choice between OSD and OSP.
+ *
+ * An LOD object (file or directory) with N stripes (each has a
+ * different FID):
+ *
+ * LOD
+ * |
+ * +---+---+---+...+
+ * | | | | |
+ * S0 S1 S2 S3 S(N-1) OS[DP] objects, seen as DT objects by LOD
+ *
+ * When upper layers must access an object's stripes (which are
+ * themselves OST or MDT LU objects) LOD finds these objects by their
+ * FIDs and stores them as an array of DT object pointers on the
+ * object. Declarations and operations on LOD objects are received by
+ * LOD (as DT object operations) and performed on the underlying
+ * OS[DP] object and (as needed) on the stripes. From the perspective
+ * of LOD, a stripe-less file (created by mknod() or open with
+ * O_LOV_DELAY_CREATE) is an object which does not yet have stripes,
+ * while a non-striped directory (created by mkdir()) is an object
+ * which will never have stripes.
+ *
+ * The LOD layer also implements a small subset of the OBD device API
+ * to support MDT stack initialization and finalization (an MDD device
+ * connects and disconnects itself to and from the underlying LOD
+ * device), and pool management. In turn LOD uses the OBD device API
+ * to connect it self to the underlying OSD, and to connect itself to
+ * OSP devices representing the MDTs and OSTs that bear the stripes of
+ * its objects.
+ */
#define DEBUG_SUBSYSTEM S_MDS
* \param type indidcate the FID is on MDS or OST.
**/
int lod_fld_lookup(const struct lu_env *env, struct lod_device *lod,
- const struct lu_fid *fid, __u32 *tgt, int type)
+ const struct lu_fid *fid, __u32 *tgt, int *type)
{
struct lu_seq_range range = { 0 };
struct lu_server_fld *server_fld;
ENTRY;
LASSERTF(fid_is_sane(fid), "Invalid FID "DFID"\n", PFID(fid));
+
if (fid_is_idif(fid)) {
*tgt = fid_idif_ost_idx(fid);
+ *type = LU_SEQ_RANGE_OST;
RETURN(rc);
}
if (!lod->lod_initialized || (!fid_seq_in_fldb(fid_seq(fid)))) {
LASSERT(lu_site2seq(lod2lu_dev(lod)->ld_site) != NULL);
+
*tgt = lu_site2seq(lod2lu_dev(lod)->ld_site)->ss_node_id;
+ *type = LU_SEQ_RANGE_MDT;
RETURN(rc);
}
server_fld = lu_site2seq(lod2lu_dev(lod)->ld_site)->ss_server_fld;
- fld_range_set_type(&range, type);
+ fld_range_set_type(&range, *type);
rc = fld_server_lookup(env, server_fld, fid_seq(fid), &range);
if (rc)
RETURN(rc);
*tgt = range.lsr_index;
+ *type = range.lsr_flags;
CDEBUG(D_INFO, "LOD: got tgt %x for sequence: "
LPX64"\n", *tgt, fid_seq(fid));
}
extern struct lu_object_operations lod_lu_obj_ops;
-extern struct lu_object_operations lod_lu_robj_ops;
extern struct dt_object_operations lod_obj_ops;
/* Slab for OSD object allocation */
{
struct lod_object *lod_obj;
struct lu_object *lu_obj;
- const struct lu_fid *fid = &hdr->loh_fid;
- mdsno_t mds;
- int rc = 0;
ENTRY;
- OBD_SLAB_ALLOC_PTR_GFP(lod_obj, lod_object_kmem, __GFP_IO);
+ OBD_SLAB_ALLOC_PTR_GFP(lod_obj, lod_object_kmem, GFP_NOFS);
if (lod_obj == NULL)
RETURN(ERR_PTR(-ENOMEM));
- rc = lod_fld_lookup(env, lu2lod_dev(dev), fid, &mds, LU_SEQ_RANGE_MDT);
- if (rc) {
- OBD_SLAB_FREE_PTR(lod_obj, lod_object_kmem);
- RETURN(ERR_PTR(rc));
- }
-
- lod_obj->ldo_mds_num = mds;
lu_obj = lod2lu_obj(lod_obj);
dt_object_init(&lod_obj->ldo_obj, NULL, dev);
lod_obj->ldo_obj.do_ops = &lod_obj_ops;
- if (likely(mds == lu_site2seq(dev->ld_site)->ss_node_id))
- lu_obj->lo_ops = &lod_lu_obj_ops;
- else
- lu_obj->lo_ops = &lod_lu_robj_ops;
+ lu_obj->lo_ops = &lod_lu_obj_ops;
+
RETURN(lu_obj);
}
{
struct lu_device *next;
int rc = 0;
- int i;
+ unsigned int i;
lod_getref(ltd);
if (ltd->ltd_tgts_size <= 0) {
}
case LCFG_PARAM: {
- struct lprocfs_static_vars v = { 0 };
- struct obd_device *obd = lod2obd(lod);
+ struct obd_device *obd = lod2obd(lod);
- lprocfs_lod_init_vars(&v);
-
- rc = class_process_proc_param(PARAM_LOV, v.obd_vars, lcfg, obd);
+ rc = class_process_proc_seq_param(PARAM_LOV, obd->obd_vars,
+ lcfg, obd);
if (rc > 0)
rc = 0;
GOTO(out, rc);
{
struct lod_device *lod = lu2lod_dev(dev);
struct lu_device *next = &lod->lod_child->dd_lu_dev;
- int i, rc;
+ unsigned int i;
+ int rc;
ENTRY;
LASSERT(lod->lod_recovery_completed == 0);
if (unlikely(th->th_update != NULL)) {
struct thandle_update *tu = th->th_update;
- struct update_request *update;
+ struct dt_update_request *update;
list_for_each_entry(update, &tu->tu_remote_update_list,
- ur_list) {
- LASSERT(update->ur_dt != NULL);
- rc = dt_trans_start(env, update->ur_dt, th);
+ dur_list) {
+ LASSERT(update->dur_dt != NULL);
+ rc = dt_trans_start(env, update->dur_dt, th);
if (rc != 0)
return rc;
}
static int lod_trans_stop(const struct lu_env *env, struct dt_device *dt,
struct thandle *th)
{
- struct thandle_update *tu = th->th_update;
- struct update_request *update;
- struct update_request *tmp;
- int rc2 = 0;
- int rc;
+ struct thandle_update *tu = th->th_update;
+ struct dt_update_request *update;
+ struct dt_update_request *tmp;
+ int rc2 = 0;
+ int rc;
ENTRY;
rc = dt_trans_stop(env, th->th_dev, th);
if (likely(tu == NULL))
RETURN(rc);
- list_for_each_entry_safe(update, tmp, &tu->tu_remote_update_list,
- ur_list) {
+ list_for_each_entry_safe(update, tmp,
+ &tu->tu_remote_update_list,
+ dur_list) {
/* update will be freed inside dt_trans_stop */
- rc2 = dt_trans_stop(env, update->ur_dt, th);
+ rc2 = dt_trans_stop(env, update->dur_dt, th);
if (unlikely(rc2 != 0 && rc == 0))
rc = rc2;
}
{
struct lod_device *lod = dt2lod_dev(dev);
struct lod_ost_desc *ost;
- int rc = 0, i;
+ unsigned int i;
+ int rc = 0;
ENTRY;
lod_getref(&lod->lod_ost_descs);
struct lu_device *next = &lod->lod_child->dd_lu_dev;
ENTRY;
- LASSERT(cfs_atomic_read(&lu->ld_ref) == 0);
+ LASSERT(atomic_read(&lu->ld_ref) == 0);
dt_device_fini(&lod->lod_dt_dev);
OBD_FREE_PTR(lod);
RETURN(next);
info->lti_ea_store = NULL;
info->lti_ea_store_size = 0;
}
+ lu_buf_free(&info->lti_linkea_buf);
OBD_FREE_PTR(info);
}
struct obd_device *obd = exp->exp_obd;
struct lod_device *d;
struct lod_ost_desc *ost;
- int i, rc = 1;
+ unsigned int i;
+ int rc = 1;
if (!obd->obd_set_up || obd->obd_stopping)
RETURN(-EAGAIN);
static int __init lod_mod_init(void)
{
- struct lprocfs_static_vars lvars = { 0 };
- cfs_proc_dir_entry_t *lov_proc_dir;
- int rc;
+ struct obd_type *type;
+ int rc;
rc = lu_kmem_init(lod_caches);
if (rc)
return rc;
- lprocfs_lod_init_vars(&lvars);
-
- rc = class_register_type(&lod_obd_device_ops, NULL, NULL,
+ rc = class_register_type(&lod_obd_device_ops, NULL, true, NULL,
#ifndef HAVE_ONLY_PROCFS_SEQ
- lvars.module_vars,
+ NULL,
#endif
- LUSTRE_LOD_NAME, &lod_device_type);
+ LUSTRE_LOD_NAME, &lod_device_type);
if (rc) {
lu_kmem_fini(lod_caches);
return rc;
}
/* create "lov" entry in procfs for compatibility purposes */
- lov_proc_dir = lprocfs_srch(proc_lustre_root, "lov");
- if (lov_proc_dir == NULL) {
- lov_proc_dir = lprocfs_register("lov", proc_lustre_root,
- NULL, NULL);
- if (IS_ERR(lov_proc_dir))
- CERROR("lod: can't create compat entry \"lov\": %d\n",
- (int)PTR_ERR(lov_proc_dir));
- }
+ type = class_search_type(LUSTRE_LOV_NAME);
+ if (type != NULL && type->typ_procroot != NULL)
+ return rc;
+ type = class_search_type(LUSTRE_LOD_NAME);
+ type->typ_procsym = lprocfs_seq_register("lov", proc_lustre_root,
+ NULL, NULL);
+ if (IS_ERR(type->typ_procsym)) {
+ CERROR("lod: can't create compat entry \"lov\": %d\n",
+ (int)PTR_ERR(type->typ_procsym));
+ type->typ_procsym = NULL;
+ }
return rc;
}
static void __exit lod_mod_exit(void)
{
-
- lprocfs_try_remove_proc_entry("lov", proc_lustre_root);
-
class_unregister_type(LUSTRE_LOD_NAME);
lu_kmem_fini(lod_caches);
}
git://git.whamcloud.com / fs / lustre-release.git / blobdiff