X-Git-Url: https://git.whamcloud.com/?p=fs%2Flustre-release.git;a=blobdiff_plain;f=lustre%2Fmdd%2Fmdd_device.c;h=6129e110e4586a99a915be08e6c161235299e400;hp=e4f5b09037e9ca10b9adbc19f7ab4733a0976b23;hb=36e5b7203d5f80b5ac1341a05a35c19915c05bd6;hpb=f81e1d87e2b8e3360c42d5c6477926367b906fe2

diff --git a/lustre/mdd/mdd_device.c b/lustre/mdd/mdd_device.c
index e4f5b09..6129e11 100644
--- a/lustre/mdd/mdd_device.c
+++ b/lustre/mdd/mdd_device.c
@@ -27,7 +27,7 @@
  * Copyright (c) 2007, 2010, Oracle and/or its affiliates. All rights reserved.
  * Use is subject to license terms.
  *
- * Copyright (c) 2011, 2012, Intel Corporation.
+ * Copyright (c) 2011, 2014, Intel Corporation.
  */
 /*
  * This file is part of Lustre, http://www.lustre.org/
@@ -43,7 +43,9 @@
 #define DEBUG_SUBSYSTEM S_MDS
 
 #include <linux/module.h>
+#include <linux/kthread.h>
 #include <obd_class.h>
+#include <lustre_ioctl.h>
 #include <lustre_mds.h>
 #include <obd_support.h>
 #include <lu_object.h>
@@ -52,14 +54,15 @@
 
 #include "mdd_internal.h"
 
-const struct md_device_operations mdd_ops;
+static const struct md_device_operations mdd_ops;
 static struct lu_device_type mdd_device_type;
 
 static const char mdd_root_dir_name[] = "ROOT";
 static const char mdd_obf_dir_name[] = "fid";
+static const char mdd_lpf_dir_name[] = "lost+found";
 
 /* Slab for MDD object allocation */
-cfs_mem_cache_t *mdd_object_kmem;
+struct kmem_cache *mdd_object_kmem;
 
 static struct lu_kmem_descr mdd_caches[] = {
 	{
@@ -117,14 +120,26 @@ out:
 static int mdd_init0(const struct lu_env *env, struct mdd_device *mdd,
 		struct lu_device_type *t, struct lustre_cfg *lcfg)
 {
-	int rc;
+	int rc = -EINVAL;
+	const char *dev;
 	ENTRY;
 
+	/* LU-8040 Set defaults here, before values configs */
+	mdd->mdd_cl.mc_flags = 0; /* off by default */
+	mdd->mdd_cl.mc_mask = CHANGELOG_DEFMASK;
+
+	dev = lustre_cfg_string(lcfg, 0);
+	if (dev == NULL)
+		RETURN(rc);
+
+	mdd->mdd_md_dev.md_lu_dev.ld_obd = class_name2obd(dev);
+	if (mdd->mdd_md_dev.md_lu_dev.ld_obd == NULL)
+		RETURN(rc);
 	mdd->mdd_md_dev.md_lu_dev.ld_ops = &mdd_lu_ops;
 	mdd->mdd_md_dev.md_ops = &mdd_ops;
 
 	rc = mdd_connect_to_next(env, mdd, lustre_cfg_string(lcfg, 3));
-	if (rc)
+	if (rc != 0)
 		RETURN(rc);
 
 	mdd->mdd_atime_diff = MAX_ATIME_DIFF;
@@ -146,17 +161,12 @@ static int mdd_init0(const struct lu_env *env, struct mdd_device *mdd,
 static struct lu_device *mdd_device_fini(const struct lu_env *env,
                                          struct lu_device *d)
 {
-        struct mdd_device *mdd = lu2mdd_dev(d);
-        int rc;
+	struct mdd_device *mdd = lu2mdd_dev(d);
 
 	if (d->ld_site)
 		lu_dev_del_linkage(d->ld_site, d);
 
-        rc = mdd_procfs_fini(mdd);
-        if (rc) {
-                CERROR("proc fini error %d \n", rc);
-                return ERR_PTR(rc);
-        }
+	mdd_procfs_fini(mdd);
 	return NULL;
 }
 
@@ -173,7 +183,7 @@ static int changelog_init_cb(const struct lu_env *env, struct llog_handle *llh,
 	       "seeing record at index %d/%d/"LPU64" t=%x %.*s in log"
 	       DOSTID"\n", hdr->lrh_index, rec->cr_hdr.lrh_index,
 	       rec->cr.cr_index, rec->cr.cr_type, rec->cr.cr_namelen,
-	       rec->cr.cr_name, POSTID(&llh->lgh_id.lgl_oi));
+	       changelog_rec_name(&rec->cr), POSTID(&llh->lgh_id.lgl_oi));
 
 	mdd->mdd_cl.mc_index = rec->cr.cr_index;
 	return LLOG_PROC_BREAK;
@@ -196,6 +206,8 @@ static int changelog_user_init_cb(const struct lu_env *env,
 
 	spin_lock(&mdd->mdd_cl.mc_user_lock);
 	mdd->mdd_cl.mc_lastuser = rec->cur_id;
+	if (rec->cur_endrec > mdd->mdd_cl.mc_index)
+		mdd->mdd_cl.mc_index = rec->cur_endrec;
 	spin_unlock(&mdd->mdd_cl.mc_user_lock);
 
 	return LLOG_PROC_BREAK;
@@ -232,7 +244,6 @@ static int llog_changelog_cancel_cb(const struct lu_env *env,
 
 static int llog_changelog_cancel(const struct lu_env *env,
 				 struct llog_ctxt *ctxt,
-				 struct lov_stripe_md *lsm, int count,
 				 struct llog_cookie *cookies, int flags)
 {
 	struct llog_handle	*cathandle = ctxt->loc_handle;
@@ -258,7 +269,49 @@ static int llog_changelog_cancel(const struct lu_env *env,
 
 static struct llog_operations changelog_orig_logops;
 
-int mdd_changelog_on(const struct lu_env *env, struct mdd_device *mdd, int on);
+static int
+mdd_changelog_write_header(const struct lu_env *env, struct mdd_device *mdd,
+			   int markerflags);
+
+static int
+mdd_changelog_on(const struct lu_env *env, struct mdd_device *mdd)
+{
+	int rc = 0;
+
+	if ((mdd->mdd_cl.mc_flags & CLM_ON) != 0)
+		return rc;
+
+	LCONSOLE_INFO("%s: changelog on\n", mdd2obd_dev(mdd)->obd_name);
+	if (mdd->mdd_cl.mc_flags & CLM_ERR) {
+		CERROR("Changelogs cannot be enabled due to error "
+		       "condition (see %s log).\n",
+		       mdd2obd_dev(mdd)->obd_name);
+		rc = -ESRCH;
+	} else {
+		spin_lock(&mdd->mdd_cl.mc_lock);
+		mdd->mdd_cl.mc_flags |= CLM_ON;
+		spin_unlock(&mdd->mdd_cl.mc_lock);
+		rc = mdd_changelog_write_header(env, mdd, CLM_START);
+	}
+	return rc;
+}
+
+static int
+mdd_changelog_off(const struct lu_env *env, struct mdd_device *mdd)
+{
+	int rc = 0;
+
+	if ((mdd->mdd_cl.mc_flags & CLM_ON) != CLM_ON)
+		return rc;
+
+	LCONSOLE_INFO("%s: changelog off\n", mdd2obd_dev(mdd)->obd_name);
+	rc = mdd_changelog_write_header(env, mdd, CLM_FINI);
+	spin_lock(&mdd->mdd_cl.mc_lock);
+	mdd->mdd_cl.mc_flags &= ~CLM_ON;
+	spin_unlock(&mdd->mdd_cl.mc_lock);
+
+	return rc;
+}
 
 static int mdd_changelog_llog_init(const struct lu_env *env,
 				   struct mdd_device *mdd)
@@ -291,7 +344,7 @@ static int mdd_changelog_llog_init(const struct lu_env *env,
 	if (rc)
 		GOTO(out_cleanup, rc);
 
-	rc = llog_cat_init_and_process(env, ctxt->loc_handle);
+	rc = llog_init_handle(env, ctxt->loc_handle, LLOG_F_IS_CAT, NULL);
 	if (rc)
 		GOTO(out_close, rc);
 
@@ -327,7 +380,7 @@ static int mdd_changelog_llog_init(const struct lu_env *env,
 	uctxt->loc_handle->lgh_logops->lop_add = llog_cat_add_rec;
 	uctxt->loc_handle->lgh_logops->lop_declare_add = llog_cat_declare_add_rec;
 
-	rc = llog_cat_init_and_process(env, uctxt->loc_handle);
+	rc = llog_init_handle(env, uctxt->loc_handle, LLOG_F_IS_CAT, NULL);
 	if (rc)
 		GOTO(out_uclose, rc);
 
@@ -341,7 +394,7 @@ static int mdd_changelog_llog_init(const struct lu_env *env,
 
 	/* If we have registered users, assume we want changelogs on */
 	if (mdd->mdd_cl.mc_lastuser > 0) {
-		rc = mdd_changelog_on(env, mdd, 1);
+		rc = mdd_changelog_on(env, mdd);
 		if (rc < 0)
 			GOTO(out_uclose, rc);
 	}
@@ -367,8 +420,6 @@ static int mdd_changelog_init(const struct lu_env *env, struct mdd_device *mdd)
 	mdd->mdd_cl.mc_index = 0;
 	spin_lock_init(&mdd->mdd_cl.mc_lock);
 	mdd->mdd_cl.mc_starttime = cfs_time_current_64();
-	mdd->mdd_cl.mc_flags = 0; /* off by default */
-	mdd->mdd_cl.mc_mask = CHANGELOG_DEFMASK;
 	spin_lock_init(&mdd->mdd_cl.mc_user_lock);
 	mdd->mdd_cl.mc_lastuser = 0;
 
@@ -402,45 +453,15 @@ static void mdd_changelog_fini(const struct lu_env *env,
 	}
 }
 
-int mdd_changelog_write_header(const struct lu_env *env,
-			       struct mdd_device *mdd, int markerflags);
-
-/* Start / stop recording */
-int mdd_changelog_on(const struct lu_env *env, struct mdd_device *mdd, int on)
-{
-        int rc = 0;
-
-        if ((on == 1) && ((mdd->mdd_cl.mc_flags & CLM_ON) == 0)) {
-                LCONSOLE_INFO("%s: changelog on\n", mdd2obd_dev(mdd)->obd_name);
-                if (mdd->mdd_cl.mc_flags & CLM_ERR) {
-                        CERROR("Changelogs cannot be enabled due to error "
-                               "condition (see %s log).\n",
-                               mdd2obd_dev(mdd)->obd_name);
-                        rc = -ESRCH;
-                } else {
-			spin_lock(&mdd->mdd_cl.mc_lock);
-			mdd->mdd_cl.mc_flags |= CLM_ON;
-			spin_unlock(&mdd->mdd_cl.mc_lock);
-			rc = mdd_changelog_write_header(env, mdd, CLM_START);
-		}
-	} else if ((on == 0) && ((mdd->mdd_cl.mc_flags & CLM_ON) == CLM_ON)) {
-		LCONSOLE_INFO("%s: changelog off\n",mdd2obd_dev(mdd)->obd_name);
-		rc = mdd_changelog_write_header(env, mdd, CLM_FINI);
-		spin_lock(&mdd->mdd_cl.mc_lock);
-		mdd->mdd_cl.mc_flags &= ~CLM_ON;
-		spin_unlock(&mdd->mdd_cl.mc_lock);
-	}
-	return rc;
-}
-
 /** Remove entries with indicies up to and including \a endrec from the
  *  changelog
  * \param mdd
  * \param endrec
  * \retval 0 ok
  */
-int mdd_changelog_llog_cancel(const struct lu_env *env,
-			      struct mdd_device *mdd, long long endrec)
+static int
+mdd_changelog_llog_cancel(const struct lu_env *env, struct mdd_device *mdd,
+			  long long endrec)
 {
         struct obd_device *obd = mdd2obd_dev(mdd);
         struct llog_ctxt *ctxt;
@@ -477,7 +498,7 @@ int mdd_changelog_llog_cancel(const struct lu_env *env,
            changed since the last purge) */
         mdd->mdd_cl.mc_starttime = cfs_time_current_64();
 
-	rc = llog_cancel(env, ctxt, NULL, 1, (struct llog_cookie *)&endrec, 0);
+	rc = llog_cancel(env, ctxt, (struct llog_cookie *)&endrec, 0);
 out:
         llog_ctxt_put(ctxt);
         return rc;
@@ -515,10 +536,11 @@ int mdd_changelog_write_header(const struct lu_env *env,
         rec->cr.cr_flags = CLF_VERSION;
         rec->cr.cr_type = CL_MARK;
         rec->cr.cr_namelen = len;
-        memcpy(rec->cr.cr_name, obd->obd_name, rec->cr.cr_namelen);
+	memcpy(changelog_rec_name(&rec->cr), obd->obd_name, rec->cr.cr_namelen);
         /* Status and action flags */
 	rec->cr.cr_markerflags = mdd->mdd_cl.mc_flags | markerflags;
-	rec->cr_hdr.lrh_len = llog_data_len(sizeof(*rec) + rec->cr.cr_namelen);
+	rec->cr_hdr.lrh_len = llog_data_len(changelog_rec_size(&rec->cr) +
+					    rec->cr.cr_namelen);
 	rec->cr_hdr.lrh_type = CHANGELOG_REC;
 	rec->cr.cr_time = cl_time();
 	spin_lock(&mdd->mdd_cl.mc_lock);
@@ -528,7 +550,7 @@ int mdd_changelog_write_header(const struct lu_env *env,
 	ctxt = llog_get_context(obd, LLOG_CHANGELOG_ORIG_CTXT);
 	LASSERT(ctxt);
 
-	rc = llog_cat_add(env, ctxt->loc_handle, &rec->cr_hdr, NULL, NULL);
+	rc = llog_cat_add(env, ctxt->loc_handle, &rec->cr_hdr, NULL);
 	if (rc > 0)
 		rc = 0;
 	llog_ctxt_put(ctxt);
@@ -556,25 +578,25 @@ static int obf_lookup(const struct lu_env *env, struct md_object *p,
 
         sscanf(name, SFID, RFID(f));
         if (!fid_is_sane(f)) {
-		CWARN("%s: bad FID format [%s], should be "DFID"\n",
-		      mdd2obd_dev(mdd)->obd_name, lname->ln_name,
+		CWARN("%s: Trying to lookup invalid FID [%s] in %s/%s, FID "
+		      "format should be "DFID"\n", mdd2obd_dev(mdd)->obd_name,
+		      lname->ln_name, dot_lustre_name, mdd_obf_dir_name,
 		      (__u64)FID_SEQ_NORMAL, 1, 0);
                 GOTO(out, rc = -EINVAL);
         }
 
-	if (!fid_is_norm(f) && !fid_is_igif(f)) {
-		CWARN("%s: "DFID" is invalid, sequence should be "
-		      ">= "LPX64" or within ["LPX64","LPX64"].\n",
-		      mdd2obd_dev(mdd)->obd_name, PFID(f),
-		      (__u64)FID_SEQ_NORMAL, (__u64)FID_SEQ_IGIF,
-		      (__u64)FID_SEQ_IGIF_MAX);
+	if (!fid_is_norm(f) && !fid_is_igif(f) && !fid_is_root(f) &&
+	    !fid_seq_is_dot(f->f_seq)) {
+		CWARN("%s: Trying to lookup invalid FID "DFID" in %s/%s, "
+		      "sequence should be >= "LPX64" or within ["LPX64","
+		      ""LPX64"].\n", mdd2obd_dev(mdd)->obd_name, PFID(f),
+		      dot_lustre_name, mdd_obf_dir_name, (__u64)FID_SEQ_NORMAL,
+		      (__u64)FID_SEQ_IGIF, (__u64)FID_SEQ_IGIF_MAX);
 		GOTO(out, rc = -EINVAL);
 	}
 
         /* Check if object with this fid exists */
         child = mdd_object_find(env, mdd, f);
-        if (child == NULL)
-                GOTO(out, rc = 0);
         if (IS_ERR(child))
                 GOTO(out, rc = PTR_ERR(child));
 
@@ -587,42 +609,61 @@ out:
         return rc;
 }
 
-static int obf_create(const struct lu_env *env, struct md_object *pobj,
-                      const struct lu_name *lname, struct md_object *child,
-                      struct md_op_spec *spec, struct md_attr* ma)
+static int mdd_dummy_create(const struct lu_env *env,
+			    struct md_object *pobj,
+			    const struct lu_name *lname,
+			    struct md_object *child,
+			    struct md_op_spec *spec,
+			    struct md_attr* ma)
 {
-        return -EPERM;
+	return -EPERM;
 }
 
-static int obf_rename(const struct lu_env *env,
-                      struct md_object *src_pobj, struct md_object *tgt_pobj,
-                      const struct lu_fid *lf, const struct lu_name *lsname,
-                      struct md_object *tobj, const struct lu_name *ltname,
-                      struct md_attr *ma)
+static int mdd_dummy_rename(const struct lu_env *env,
+			    struct md_object *src_pobj,
+			    struct md_object *tgt_pobj,
+			    const struct lu_fid *lf,
+			    const struct lu_name *lsname,
+			    struct md_object *tobj,
+			    const struct lu_name *ltname,
+			    struct md_attr *ma)
 {
-        return -EPERM;
+	return -EPERM;
 }
 
-static int obf_link(const struct lu_env *env, struct md_object *tgt_obj,
-                    struct md_object *src_obj, const struct lu_name *lname,
-                    struct md_attr *ma)
+static int mdd_dummy_link(const struct lu_env *env,
+			  struct md_object *tgt_obj,
+			  struct md_object *src_obj,
+			  const struct lu_name *lname,
+			  struct md_attr *ma)
 {
-        return -EPERM;
+	return -EPERM;
 }
 
-static int obf_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)
+static int mdd_dummy_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)
 {
 	return -EPERM;
 }
 
 static struct md_dir_operations mdd_obf_dir_ops = {
-        .mdo_lookup = obf_lookup,
-        .mdo_create = obf_create,
-        .mdo_rename = obf_rename,
-        .mdo_link   = obf_link,
-        .mdo_unlink = obf_unlink
+	.mdo_lookup = obf_lookup,
+	.mdo_create = mdd_dummy_create,
+	.mdo_rename = mdd_dummy_rename,
+	.mdo_link   = mdd_dummy_link,
+	.mdo_unlink = mdd_dummy_unlink
+};
+
+static struct md_dir_operations mdd_lpf_dir_ops = {
+	.mdo_lookup = mdd_lookup,
+	.mdo_create = mdd_dummy_create,
+	.mdo_rename = mdd_dummy_rename,
+	.mdo_link   = mdd_dummy_link,
+	.mdo_unlink = mdd_dummy_unlink
 };
 
 static struct md_object *mdo_locate(const struct lu_env *env,
@@ -643,6 +684,33 @@ static struct md_object *mdo_locate(const struct lu_env *env,
 	return mdo;
 }
 
+static int mdd_lpf_setup(const struct lu_env *env, struct mdd_device *m)
+{
+	struct md_object	*mdo;
+	struct mdd_object	*mdd_lpf;
+	struct lu_fid		 fid	= LU_LPF_FID;
+	int			 rc;
+	ENTRY;
+
+	rc = mdd_local_file_create(env, m, mdd_object_fid(m->mdd_dot_lustre),
+				   mdd_lpf_dir_name, S_IFDIR | S_IRUSR | S_IXUSR,
+				   &fid);
+	if (rc != 0)
+		RETURN(rc);
+
+	mdo = mdo_locate(env, &m->mdd_md_dev, &fid);
+	if (IS_ERR(mdo))
+		RETURN(PTR_ERR(mdo));
+
+	LASSERT(lu_object_exists(&mdo->mo_lu));
+
+	mdd_lpf = md2mdd_obj(mdo);
+	mdd_lpf->mod_obj.mo_dir_ops = &mdd_lpf_dir_ops;
+	m->mdd_dot_lustre_objs.mdd_lpf = mdd_lpf;
+
+	RETURN(0);
+}
+
 /**
  * Create special in-memory "fid" object for open-by-fid.
  */
@@ -671,6 +739,23 @@ static int mdd_obf_setup(const struct lu_env *env, struct mdd_device *m)
 	return 0;
 }
 
+static void mdd_dot_lustre_cleanup(const struct lu_env *env,
+				   struct mdd_device *m)
+{
+	if (m->mdd_dot_lustre_objs.mdd_lpf != NULL) {
+		mdd_object_put(env, m->mdd_dot_lustre_objs.mdd_lpf);
+		m->mdd_dot_lustre_objs.mdd_lpf = NULL;
+	}
+	if (m->mdd_dot_lustre_objs.mdd_obf != NULL) {
+		mdd_object_put(env, m->mdd_dot_lustre_objs.mdd_obf);
+		m->mdd_dot_lustre_objs.mdd_obf = NULL;
+	}
+	if (m->mdd_dot_lustre != NULL) {
+		mdd_object_put(env, m->mdd_dot_lustre);
+		m->mdd_dot_lustre = NULL;
+	}
+}
+
 /** Setup ".lustre" directory object */
 static int mdd_dot_lustre_setup(const struct lu_env *env, struct mdd_device *m)
 {
@@ -700,25 +785,108 @@ static int mdd_dot_lustre_setup(const struct lu_env *env, struct mdd_device *m)
 		       mdd2obd_dev(m)->obd_name, rc);
 		GOTO(out, rc);
 	}
+
+	rc = mdd_lpf_setup(env, m);
+	if (rc != 0) {
+		CERROR("%s: error initializing \"lost+found\": rc = %d.\n",
+		       mdd2obd_dev(m)->obd_name, rc);
+		GOTO(out, rc);
+	}
+
 	RETURN(0);
+
 out:
-	mdd_object_put(env, m->mdd_dot_lustre);
-	m->mdd_dot_lustre = NULL;
+	mdd_dot_lustre_cleanup(env, m);
+
 	return rc;
 }
 
+
+static struct llog_operations hsm_actions_logops;
+
+/**
+ * set llog methods and create LLOG_AGENT_ORIG_CTXT llog
+ * object in obd_device
+ */
+static int mdd_hsm_actions_llog_init(const struct lu_env *env,
+				     struct mdd_device *m)
+{
+	struct obd_device	*obd = mdd2obd_dev(m);
+	struct llog_ctxt	*ctxt = NULL;
+	int			 rc;
+	ENTRY;
+
+	OBD_SET_CTXT_MAGIC(&obd->obd_lvfs_ctxt);
+	obd->obd_lvfs_ctxt.dt = m->mdd_bottom;
+
+	rc = llog_setup(env, obd, &obd->obd_olg, LLOG_AGENT_ORIG_CTXT,
+			obd, &hsm_actions_logops);
+	if (rc) {
+		CERROR("%s: hsm actions llog setup failed: rc = %d\n",
+			obd->obd_name, rc);
+		RETURN(rc);
+	}
+
+	ctxt = llog_get_context(obd, LLOG_AGENT_ORIG_CTXT);
+	LASSERT(ctxt);
+
+	rc = llog_open_create(env, ctxt, &ctxt->loc_handle, NULL,
+			      HSM_ACTIONS);
+	if (rc) {
+		CERROR("%s: hsm actions llog open_create failed: rc = %d\n",
+			obd->obd_name, rc);
+		GOTO(out_cleanup, rc);
+	}
+
+	rc = llog_init_handle(env, ctxt->loc_handle, LLOG_F_IS_CAT, NULL);
+	if (rc)
+		GOTO(out_close, rc);
+
+	llog_ctxt_put(ctxt);
+	RETURN(0);
+
+out_close:
+	llog_cat_close(env, ctxt->loc_handle);
+	ctxt->loc_handle = NULL;
+out_cleanup:
+	llog_cleanup(env, ctxt);
+
+	return rc;
+}
+
+/**
+ * cleanup the context created by llog_setup_named()
+ */
+static int mdd_hsm_actions_llog_fini(const struct lu_env *env,
+				     struct mdd_device *m)
+{
+	struct obd_device	*obd = mdd2obd_dev(m);
+	struct llog_ctxt	*lctxt;
+	ENTRY;
+
+	lctxt = llog_get_context(obd, LLOG_AGENT_ORIG_CTXT);
+	if (lctxt) {
+		llog_cat_close(env, lctxt->loc_handle);
+		lctxt->loc_handle = NULL;
+		llog_cleanup(env, lctxt);
+	}
+
+	RETURN(0);
+}
+
 static void mdd_device_shutdown(const struct lu_env *env, struct mdd_device *m,
 				struct lustre_cfg *cfg)
 {
-	mdd_lfsck_cleanup(env, m);
+	mdd_generic_thread_stop(&m->mdd_orph_cleanup_thread);
+	lfsck_degister(env, m->mdd_bottom);
+	mdd_hsm_actions_llog_fini(env, m);
 	mdd_changelog_fini(env, m);
 	orph_index_fini(env, m);
-	if (m->mdd_dot_lustre_objs.mdd_obf)
-		mdd_object_put(env, m->mdd_dot_lustre_objs.mdd_obf);
-	if (m->mdd_dot_lustre)
-		mdd_object_put(env, m->mdd_dot_lustre);
-	if (m->mdd_los != NULL)
+	mdd_dot_lustre_cleanup(env, m);
+	if (m->mdd_los != NULL) {
 		local_oid_storage_fini(env, m->mdd_los);
+		m->mdd_los = NULL;
+	}
 	lu_site_purge(env, mdd2lu_dev(m)->ld_site, ~0);
 
 	if (m->mdd_child_exp)
@@ -735,16 +903,15 @@ static int mdd_process_config(const struct lu_env *env,
         ENTRY;
 
         switch (cfg->lcfg_command) {
-        case LCFG_PARAM: {
-                struct lprocfs_static_vars lvars;
-
-                lprocfs_mdd_init_vars(&lvars);
-                rc = class_process_proc_param(PARAM_MDD, lvars.obd_vars, cfg,m);
-                if (rc > 0 || rc == -ENOSYS)
-                        /* we don't understand; pass it on */
-                        rc = next->ld_ops->ldo_process_config(env, next, cfg);
-                break;
-        }
+	case LCFG_PARAM: {
+		struct obd_device *obd = mdd2obd_dev(m);
+
+		rc = class_process_proc_param(PARAM_MDD, obd->obd_vars, cfg, m);
+		if (rc > 0 || rc == -ENOSYS)
+			/* we don't understand; pass it on */
+			rc = next->ld_ops->ldo_process_config(env, next, cfg);
+		break;
+	}
         case LCFG_SETUP:
                 rc = next->ld_ops->ldo_process_config(env, next, cfg);
                 if (rc)
@@ -776,18 +943,18 @@ static int mdd_recovery_complete(const struct lu_env *env,
 	next = &mdd->mdd_child->dd_lu_dev;
 
         /* XXX: orphans handling. */
-        __mdd_orphan_cleanup(env, mdd);
+        mdd_orphan_cleanup(env, mdd);
         rc = next->ld_ops->ldo_recovery_complete(env, next);
 
         RETURN(rc);
 }
 
 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)
+			  const struct lu_fid *pfid, const char *name,
+			  __u32 mode, struct lu_fid *fid)
 {
-	struct dt_object	*parent, *dto;
-	int			 rc;
+	struct dt_object *parent, *dto;
+	int rc;
 
 	ENTRY;
 
@@ -816,6 +983,12 @@ out_put:
 	return 0;
 }
 
+static int mdd_lfsck_out_notify(const struct lu_env *env, void *data,
+				enum lfsck_events event)
+{
+	return 0;
+}
+
 static int mdd_prepare(const struct lu_env *env,
                        struct lu_device *pdev,
                        struct lu_device *cdev)
@@ -844,8 +1017,8 @@ static int mdd_prepare(const struct lu_env *env,
 	if (rc < 0)
 		GOTO(out_los, rc);
 
+	lu_root_fid(&fid);
 	if (mdd_seq_site(mdd)->ss_node_id == 0) {
-		lu_root_fid(&fid);
 		rc = mdd_local_file_create(env, mdd, &mdd->mdd_local_root_fid,
 					   mdd_root_dir_name, S_IFDIR |
 					   S_IRUGO | S_IWUSR | S_IXUGO, &fid);
@@ -854,19 +1027,18 @@ static int mdd_prepare(const struct lu_env *env,
 			       mdd2obd_dev(mdd)->obd_name, rc);
 			GOTO(out_los, rc);
 		}
-		mdd->mdd_root_fid = fid;
 
+		mdd->mdd_root_fid = fid;
 		rc = mdd_dot_lustre_setup(env, mdd);
 		if (rc != 0) {
 			CERROR("%s: initializing .lustre failed: rc = %d\n",
 			       mdd2obd_dev(mdd)->obd_name, rc);
 			GOTO(out_los, rc);
 		}
-
-		rc = mdd_compat_fixes(env, mdd);
-		if (rc)
-			GOTO(out_los, rc);
-
+	} else {
+		/* Normal client usually send root access to MDT0 directly,
+		 * the root FID on non-MDT0 will only be used by echo client. */
+		mdd->mdd_root_fid = fid;
 	}
 
 	rc = orph_index_init(env, mdd);
@@ -880,27 +1052,32 @@ static int mdd_prepare(const struct lu_env *env,
 		GOTO(out_orph, rc);
 	}
 
-	rc = mdd_lfsck_setup(env, mdd);
+	rc = mdd_hsm_actions_llog_init(env, mdd);
+	if (rc != 0)
+		GOTO(out_changelog, rc);
+
+	rc = lfsck_register(env, mdd->mdd_bottom, mdd->mdd_child,
+			    mdd2obd_dev(mdd), mdd_lfsck_out_notify,
+			    mdd, true);
 	if (rc != 0) {
 		CERROR("%s: failed to initialize lfsck: rc = %d\n",
 		       mdd2obd_dev(mdd)->obd_name, rc);
-		GOTO(out_changelog, rc);
+		GOTO(out_hsm, rc);
 	}
 	RETURN(0);
+out_hsm:
+	mdd_hsm_actions_llog_fini(env, mdd);
 out_changelog:
 	mdd_changelog_fini(env, mdd);
 out_orph:
 	orph_index_fini(env, mdd);
 out_dot:
-	if (mdd_seq_site(mdd)->ss_node_id == 0) {
-		mdd_object_put(env, mdd->mdd_dot_lustre);
-		mdd->mdd_dot_lustre = NULL;
-		mdd_object_put(env, mdd->mdd_dot_lustre_objs.mdd_obf);
-		mdd->mdd_dot_lustre_objs.mdd_obf = NULL;
-	}
+	if (mdd_seq_site(mdd)->ss_node_id == 0)
+		mdd_dot_lustre_cleanup(env, mdd);
 out_los:
 	local_oid_storage_fini(env, mdd->mdd_los);
 	mdd->mdd_los = NULL;
+
 	return rc;
 }
 
@@ -925,42 +1102,29 @@ static int mdd_root_get(const struct lu_env *env,
  * No permission check is needed.
  */
 static int mdd_statfs(const struct lu_env *env, struct md_device *m,
-                      struct obd_statfs *sfs)
+		      struct obd_statfs *sfs)
 {
-        struct mdd_device *mdd = lu2mdd_dev(&m->md_lu_dev);
-        int rc;
+	struct mdd_device *mdd = lu2mdd_dev(&m->md_lu_dev);
+	int rc;
 
-        ENTRY;
+	ENTRY;
 
-        rc = mdd_child_ops(mdd)->dt_statfs(env, mdd->mdd_child, sfs);
+	rc = mdd_child_ops(mdd)->dt_statfs(env, mdd->mdd_child, sfs);
 
-        RETURN(rc);
+	sfs->os_namelen = min_t(__u32, sfs->os_namelen, NAME_MAX);
+
+	RETURN(rc);
 }
 
-/*
- * No permission check is needed.
- */
-static int mdd_init_capa_ctxt(const struct lu_env *env, struct md_device *m,
-                              int mode, unsigned long timeout, __u32 alg,
-                              struct lustre_capa_key *keys)
+static int mdd_maxeasize_get(const struct lu_env *env, struct md_device *m,
+				int *easize)
 {
-        struct mdd_device *mdd = lu2mdd_dev(&m->md_lu_dev);
-        int rc;
-        ENTRY;
+	struct mdd_device *mdd = lu2mdd_dev(&m->md_lu_dev);
+	ENTRY;
 
-        /* need barrier for mds_capa_keys access. */
+	*easize = mdd->mdd_dt_conf.ddp_max_ea_size;
 
-        rc = mdd_child_ops(mdd)->dt_init_capa_ctxt(env, mdd->mdd_child, mode,
-                                                   timeout, alg, keys);
-        RETURN(rc);
-}
-
-static int mdd_update_capa_key(const struct lu_env *env,
-                               struct md_device *m,
-                               struct lustre_capa_key *key)
-{
-	/* we do not support capabilities ... */
-	return -EINVAL;
+	RETURN(0);
 }
 
 static int mdd_llog_ctxt_get(const struct lu_env *env, struct md_device *m,
@@ -978,7 +1142,7 @@ static struct lu_device *mdd_device_free(const struct lu_env *env,
 	struct mdd_device *m = lu2mdd_dev(lu);
 	ENTRY;
 
-	LASSERT(cfs_atomic_read(&lu->ld_ref) == 0);
+	LASSERT(atomic_read(&lu->ld_ref) == 0);
 	md_device_fini(&m->mdd_md_dev);
 	OBD_FREE_PTR(m);
 	RETURN(NULL);
@@ -1058,15 +1222,45 @@ static int mdd_obd_disconnect(struct obd_export *exp)
 	RETURN(rc);
 }
 
-static int mdd_obd_health_check(const struct lu_env *env,
-				struct obd_device *obd)
+static int mdd_obd_get_info(const struct lu_env *env, struct obd_export *exp,
+			    __u32 keylen, void *key, __u32 *vallen, void *val)
 {
-	struct mdd_device *mdd = lu2mdd_dev(obd->obd_lu_dev);
-	int		   rc;
-	ENTRY;
+	int rc = -EINVAL;
+
+	if (KEY_IS(KEY_OSP_CONNECTED)) {
+		struct obd_device	*obd = exp->exp_obd;
+		struct mdd_device	*mdd;
+
+		if (!obd->obd_set_up || obd->obd_stopping)
+			RETURN(-EAGAIN);
+
+		mdd = lu2mdd_dev(obd->obd_lu_dev);
+		LASSERT(mdd);
+		rc = obd_get_info(env, mdd->mdd_child_exp, keylen, key, vallen,
+				  val);
+		RETURN(rc);
+	}
+
+	RETURN(rc);
+}
+
+static int mdd_obd_set_info_async(const struct lu_env *env,
+				  struct obd_export *exp,
+				  __u32 keylen, void *key,
+				  __u32 vallen, void *val,
+				  struct ptlrpc_request_set *set)
+{
+	struct obd_device	*obd = exp->exp_obd;
+	struct mdd_device	*mdd;
+	int			 rc;
+
+	if (!obd->obd_set_up || obd->obd_stopping)
+		RETURN(-EAGAIN);
 
+	mdd = lu2mdd_dev(obd->obd_lu_dev);
 	LASSERT(mdd);
-	rc = obd_health_check(env, mdd->mdd_child_exp->exp_obd);
+	rc = obd_set_info_async(env, mdd->mdd_child_exp, keylen, key,
+				vallen, val, set);
 	RETURN(rc);
 }
 
@@ -1074,30 +1268,10 @@ static struct obd_ops mdd_obd_device_ops = {
 	.o_owner	= THIS_MODULE,
 	.o_connect	= mdd_obd_connect,
 	.o_disconnect	= mdd_obd_disconnect,
-	.o_health_check	= mdd_obd_health_check
-};
-
-/*
- * context key constructor/destructor:
- * mdd_capainfo_key_init, mdd_capainfo_key_fini
- */
-LU_KEY_INIT_FINI(mdd_capainfo, struct md_capainfo);
-
-struct lu_context_key mdd_capainfo_key = {
-        .lct_tags = LCT_SESSION,
-        .lct_init = mdd_capainfo_key_init,
-        .lct_fini = mdd_capainfo_key_fini
+	.o_get_info     = mdd_obd_get_info,
+	.o_set_info_async = mdd_obd_set_info_async,
 };
 
-struct md_capainfo *md_capainfo(const struct lu_env *env)
-{
-        /* NB, in mdt_init0 */
-        if (env->le_ses == NULL)
-                return NULL;
-        return lu_context_key_get(env->le_ses, &mdd_capainfo_key);
-}
-EXPORT_SYMBOL(md_capainfo);
-
 static int mdd_changelog_user_register(const struct lu_env *env,
 				       struct mdd_device *mdd, int *id)
 {
@@ -1117,10 +1291,10 @@ static int mdd_changelog_user_register(const struct lu_env *env,
                 RETURN(-ENOMEM);
         }
 
-        /* Assume we want it on since somebody registered */
-        rc = mdd_changelog_on(env, mdd, 1);
-        if (rc)
-                GOTO(out, rc);
+	/* Assume we want it on since somebody registered */
+	rc = mdd_changelog_on(env, mdd);
+	if (rc)
+		GOTO(out, rc);
 
         rec->cur_hdr.lrh_len = sizeof(*rec);
         rec->cur_hdr.lrh_type = CHANGELOG_USER_REC;
@@ -1134,7 +1308,7 @@ static int mdd_changelog_user_register(const struct lu_env *env,
 	rec->cur_endrec = mdd->mdd_cl.mc_index;
 	spin_unlock(&mdd->mdd_cl.mc_user_lock);
 
-	rc = llog_cat_add(env, ctxt->loc_handle, &rec->cur_hdr, NULL, NULL);
+	rc = llog_cat_add(env, ctxt->loc_handle, &rec->cur_hdr, NULL);
 
         CDEBUG(D_IOCTL, "Registered changelog user %d\n", *id);
 out:
@@ -1209,10 +1383,7 @@ static int mdd_changelog_user_purge_cb(const struct lu_env *env,
         CDEBUG(D_IOCTL, "Rewriting changelog user %d endrec to "LPU64"\n",
                mcud->mcud_id, rec->cur_endrec);
 
-        /* hdr+1 is loc of data */
-        hdr->lrh_len -= sizeof(*hdr) + sizeof(struct llog_rec_tail);
-	rc = llog_write(env, llh, hdr, NULL, 0, (void *)(hdr + 1),
-			hdr->lrh_index);
+	rc = llog_write(env, llh, hdr, hdr->lrh_index);
 
         RETURN(rc);
 }
@@ -1270,11 +1441,11 @@ static int mdd_changelog_user_purge(const struct lu_env *env,
                rc = -ENOENT;
         }
 
-        if (!rc && data.mcud_usercount == 0)
-                /* No more users; turn changelogs off */
-                rc = mdd_changelog_on(env, mdd, 0);
+	if (!rc && data.mcud_usercount == 0)
+		/* No more users; turn changelogs off */
+		rc = mdd_changelog_off(env, mdd);
 
-        RETURN (rc);
+	RETURN(rc);
 }
 
 /** mdd_iocontrol
@@ -1309,19 +1480,21 @@ static int mdd_iocontrol(const struct lu_env *env, struct md_device *m,
                 RETURN(0);
         }
 	case OBD_IOC_START_LFSCK: {
-		rc = mdd_lfsck_start(env, &mdd->mdd_lfsck,
-				     (struct lfsck_start *)karg);
+		rc = lfsck_start(env, mdd->mdd_bottom,
+				 (struct lfsck_start_param *)karg);
 		RETURN(rc);
 	}
 	case OBD_IOC_STOP_LFSCK: {
-		rc = mdd_lfsck_stop(env, &mdd->mdd_lfsck, false);
+		rc = lfsck_stop(env, mdd->mdd_bottom,
+				(struct lfsck_stop *)karg);
 		RETURN(rc);
 	}
-	case OBD_IOC_PAUSE_LFSCK: {
-		rc = mdd_lfsck_stop(env, &mdd->mdd_lfsck, true);
+	case OBD_IOC_QUERY_LFSCK: {
+		rc = lfsck_query(env, mdd->mdd_bottom, NULL, NULL,
+				 (struct lfsck_query *)karg);
 		RETURN(rc);
 	}
-        }
+	}
 
         /* Below ioctls use obd_ioctl_data */
         if (len != sizeof(*data)) {
@@ -1350,15 +1523,14 @@ static int mdd_iocontrol(const struct lu_env *env, struct md_device *m,
 }
 
 /* type constructor/destructor: mdd_type_init, mdd_type_fini */
-LU_TYPE_INIT_FINI(mdd, &mdd_thread_key, &mdd_capainfo_key);
+LU_TYPE_INIT_FINI(mdd, &mdd_thread_key);
 
-const struct md_device_operations mdd_ops = {
+static const struct md_device_operations mdd_ops = {
 	.mdo_statfs         = mdd_statfs,
 	.mdo_root_get	    = mdd_root_get,
-	.mdo_init_capa_ctxt = mdd_init_capa_ctxt,
-	.mdo_update_capa_key= mdd_update_capa_key,
 	.mdo_llog_ctxt_get  = mdd_llog_ctxt_get,
 	.mdo_iocontrol      = mdd_iocontrol,
+	.mdo_maxeasize_get  = mdd_maxeasize_get,
 };
 
 static struct lu_device_type_operations mdd_device_type_ops = {
@@ -1398,12 +1570,38 @@ static void mdd_key_fini(const struct lu_context *ctx,
 /* context key: mdd_thread_key */
 LU_CONTEXT_KEY_DEFINE(mdd, LCT_MD_THREAD);
 
-static int __init mdd_mod_init(void)
+int mdd_generic_thread_start(struct mdd_generic_thread *thread,
+			     int (*func)(void *), void *data, char *name)
 {
-	struct lprocfs_static_vars lvars;
-	int rc;
+	struct task_struct	*task;
+
+	LASSERT(thread->mgt_init == false);
+	init_completion(&thread->mgt_started);
+	init_completion(&thread->mgt_finished);
+	thread->mgt_data = data;
+	thread->mgt_abort = false;
+	thread->mgt_init = true;
+
+	task = kthread_run(func, thread, name);
+	if (IS_ERR(task)) {
+		complete(&thread->mgt_finished);
+		return PTR_ERR(task);
+	}
+	wait_for_completion(&thread->mgt_started);
+	return 0;
+}
 
-	lprocfs_mdd_init_vars(&lvars);
+void mdd_generic_thread_stop(struct mdd_generic_thread *thread)
+{
+	if (thread->mgt_init == true) {
+		thread->mgt_abort = true;
+		wait_for_completion(&thread->mgt_finished);
+	}
+}
+
+static int __init mdd_init(void)
+{
+	int rc;
 
 	rc = lu_kmem_init(mdd_caches);
 	if (rc)
@@ -1414,21 +1612,27 @@ static int __init mdd_mod_init(void)
 	changelog_orig_logops.lop_add = llog_cat_add_rec;
 	changelog_orig_logops.lop_declare_add = llog_cat_declare_add_rec;
 
-	rc = class_register_type(&mdd_obd_device_ops, NULL, lvars.module_vars,
+	hsm_actions_logops = llog_osd_ops;
+	hsm_actions_logops.lop_add = llog_cat_add_rec;
+	hsm_actions_logops.lop_declare_add = llog_cat_declare_add_rec;
+
+	rc = class_register_type(&mdd_obd_device_ops, NULL, true, NULL,
 				 LUSTRE_MDD_NAME, &mdd_device_type);
 	if (rc)
 		lu_kmem_fini(mdd_caches);
 	return rc;
 }
 
-static void __exit mdd_mod_exit(void)
+static void __exit mdd_exit(void)
 {
 	class_unregister_type(LUSTRE_MDD_NAME);
 	lu_kmem_fini(mdd_caches);
 }
 
-MODULE_AUTHOR("Sun Microsystems, Inc. <http://www.lustre.org/>");
-MODULE_DESCRIPTION("Lustre Meta-data Device Prototype ("LUSTRE_MDD_NAME")");
+MODULE_AUTHOR("OpenSFS, Inc. <http://www.lustre.org/>");
+MODULE_DESCRIPTION("Lustre Meta-data Device Driver ("LUSTRE_MDD_NAME")");
+MODULE_VERSION(LUSTRE_VERSION_STRING);
 MODULE_LICENSE("GPL");
 
-cfs_module(mdd, "0.1.0", mdd_mod_init, mdd_mod_exit);
+module_init(mdd_init);
+module_exit(mdd_exit);