Whamcloud - gitweb
LU-5518 lfsck: recover orphans from backend lost+found
[fs/lustre-release.git] / lustre / osd-ldiskfs / osd_scrub.c
index e558163..d7e5d57 100644 (file)
@@ -35,7 +35,7 @@
  * Author: Fan Yong <yong.fan@whamcloud.com>
  */
 
-#define DEBUG_SUBSYSTEM S_MDS
+#define DEBUG_SUBSYSTEM S_LFSCK
 
 #include <lustre/lustre_idl.h>
 #include <lustre_disk.h>
@@ -80,6 +80,11 @@ static inline int osd_scrub_has_window(struct osd_scrub *scrub,
        return scrub->os_pos_current < ooc->ooc_pos_preload + SCRUB_WINDOW_SIZE;
 }
 
+static inline const char *osd_scrub2name(struct osd_scrub *scrub)
+{
+       return LDISKFS_SB(osd_scrub2sb(scrub))->s_es->s_volume_name;
+}
+
 /**
  * update/insert/delete the specified OI mapping (@fid @id) according to the ops
  *
@@ -103,12 +108,13 @@ static int osd_scrub_refresh_mapping(struct osd_thread_info *info,
 
        /* DTO_INDEX_INSERT is enough for other two ops:
         * delete/update, but save stack. */
-       th = ldiskfs_journal_start_sb(osd_sb(dev),
+       th = osd_journal_start_sb(osd_sb(dev), LDISKFS_HT_MISC,
                                osd_dto_credits_noquota[DTO_INDEX_INSERT]);
        if (IS_ERR(th)) {
                rc = PTR_ERR(th);
-               CERROR("%s: fail to start trans for scrub %d: rc = %d\n",
-                      osd_name(dev), ops, rc);
+               CDEBUG(D_LFSCK, "%s: fail to start trans for scrub op %d "
+                      DFID" => %u/%u: rc = %d\n", osd_name(dev), ops,
+                      PFID(fid), id->oii_ino, id->oii_gen, rc);
                RETURN(rc);
        }
 
@@ -171,7 +177,12 @@ static int osd_scrub_refresh_mapping(struct osd_thread_info *info,
        }
 
        ldiskfs_journal_stop(th);
-       return rc;
+       if (rc < 0)
+               CDEBUG(D_LFSCK, "%s: fail to refresh OI map for scrub op %d "
+                      DFID" => %u/%u: rc = %d\n", osd_name(dev), ops,
+                      PFID(fid), id->oii_ino, id->oii_gen, rc);
+
+       RETURN(rc);
 }
 
 /* OI_scrub file ops */
@@ -260,7 +271,8 @@ void osd_scrub_file_reset(struct osd_scrub *scrub, __u8 *uuid, __u64 flags)
 {
        struct scrub_file *sf = &scrub->os_file;
 
-       CDEBUG(D_LFSCK, "Reset OI scrub file, flags = "LPX64"\n", flags);
+       CDEBUG(D_LFSCK, "%.16s: reset OI scrub file, flags = "LPX64"\n",
+              osd_scrub2name(scrub), flags);
        memcpy(sf->sf_uuid, uuid, 16);
        sf->sf_status = SS_INIT;
        sf->sf_flags |= flags;
@@ -281,7 +293,6 @@ void osd_scrub_file_reset(struct osd_scrub *scrub, __u8 *uuid, __u64 flags)
 static int osd_scrub_file_load(struct osd_scrub *scrub)
 {
        loff_t  pos  = 0;
-       char   *name = LDISKFS_SB(osd_scrub2sb(scrub))->s_es->s_volume_name;
        int     len  = sizeof(scrub->os_file_disk);
        int     rc;
 
@@ -291,16 +302,18 @@ static int osd_scrub_file_load(struct osd_scrub *scrub)
 
                osd_scrub_file_to_cpu(sf, &scrub->os_file_disk);
                if (sf->sf_magic != SCRUB_MAGIC_V1) {
-                       CWARN("%.16s: invalid scrub magic 0x%x != 0x%x\n,",
-                             name, sf->sf_magic, SCRUB_MAGIC_V1);
+                       CDEBUG(D_LFSCK, "%.16s: invalid scrub magic "
+                              "0x%x != 0x%x\n", osd_scrub2name(scrub),
+                              sf->sf_magic, SCRUB_MAGIC_V1);
                        /* Process it as new scrub file. */
                        rc = -ENOENT;
                } else {
                        rc = 0;
                }
        } else if (rc != 0) {
-               CERROR("%.16s: fail to load scrub file, expected = %d, "
-                      "rc = %d\n", name, len, rc);
+               CDEBUG(D_LFSCK, "%.16s: fail to load scrub file, "
+                      "expected = %d: rc = %d\n",
+                      osd_scrub2name(scrub), len, rc);
                if (rc > 0)
                        rc = -EFAULT;
        } else {
@@ -326,8 +339,8 @@ int osd_scrub_file_store(struct osd_scrub *scrub)
        jh = osd_journal_start_sb(osd_sb(dev), LDISKFS_HT_MISC, credits);
        if (IS_ERR(jh)) {
                rc = PTR_ERR(jh);
-               CERROR("%.16s: fail to start trans for scrub store, rc = %d\n",
-                      LDISKFS_SB(osd_scrub2sb(scrub))->s_es->s_volume_name,rc);
+               CDEBUG(D_LFSCK, "%.16s: fail to start trans for scrub store: "
+                      "rc = %d\n", osd_scrub2name(scrub), rc);
                return rc;
        }
 
@@ -336,10 +349,10 @@ int osd_scrub_file_store(struct osd_scrub *scrub)
                                      len, 0, &pos, jh);
        ldiskfs_journal_stop(jh);
        if (rc != 0)
-               CERROR("%.16s: fail to store scrub file, expected = %d, "
-                      "rc = %d\n",
-                      LDISKFS_SB(osd_scrub2sb(scrub))->s_es->s_volume_name,
-                      len, rc);
+               CDEBUG(D_LFSCK, "%.16s: fail to store scrub file, "
+                      "expected = %d: rc = %d\n",
+                      osd_scrub2name(scrub), len, rc);
+
        scrub->os_time_last_checkpoint = cfs_time_current();
        scrub->os_time_next_checkpoint = scrub->os_time_last_checkpoint +
                                cfs_time_seconds(SCRUB_CHECKPOINT_INTERVAL);
@@ -376,9 +389,8 @@ osd_scrub_convert_ff(struct osd_thread_info *info, struct osd_device *dev,
                                osd_dto_credits_noquota[DTO_XATTR_SET] * 3);
        if (IS_ERR(jh)) {
                rc = PTR_ERR(jh);
-               CERROR("%s: fail to start trans for convert ff: "DFID
-                      ": rc = %d\n",
-                      osd_name(dev), PFID(fid), rc);
+               CDEBUG(D_LFSCK, "%s: fail to start trans for convert ff "
+                      DFID": rc = %d\n", osd_name(dev), PFID(fid), rc);
                RETURN(rc);
        }
 
@@ -420,6 +432,9 @@ osd_scrub_convert_ff(struct osd_thread_info *info, struct osd_device *dev,
 
 stop:
        ldiskfs_journal_stop(jh);
+       if (rc < 0)
+               CDEBUG(D_LFSCK, "%s: fail to convert ff "DFID": rc = %d\n",
+                      osd_name(dev), PFID(fid), rc);
        return rc;
 }
 
@@ -446,8 +461,8 @@ osd_scrub_check_update(struct osd_thread_info *info, struct osd_device *dev,
                GOTO(out, rc = val);
 
        if (scrub->os_in_prior)
-               oii = cfs_list_entry(oic, struct osd_inconsistent_item,
-                                    oii_cache);
+               oii = list_entry(oic, struct osd_inconsistent_item,
+                                oii_cache);
 
        if (lid->oii_ino < sf->sf_pos_latest_start && oii == NULL)
                GOTO(out, rc = 0);
@@ -486,7 +501,7 @@ osd_scrub_check_update(struct osd_thread_info *info, struct osd_device *dev,
                (val == SCRUB_NEXT_OSTOBJ ||
                 val == SCRUB_NEXT_OSTOBJ_OLD) ? OI_KNOWN_ON_OST : 0);
        if (rc != 0) {
-               if (rc != -ENOENT)
+               if (rc != -ENOENT && rc != -ESTALE)
                        GOTO(out, rc);
 
 iget:
@@ -590,10 +605,10 @@ out:
                iput(inode);
 
        if (oii != NULL) {
-               LASSERT(!cfs_list_empty(&oii->oii_list));
+               LASSERT(!list_empty(&oii->oii_list));
 
                spin_lock(&scrub->os_lock);
-               cfs_list_del_init(&oii->oii_list);
+               list_del_init(&oii->oii_list);
                spin_unlock(&scrub->os_lock);
                OBD_FREE_PTR(oii);
        }
@@ -735,10 +750,6 @@ static void osd_scrub_post(struct osd_scrub *scrub, int result)
        sf->sf_run_time += cfs_duration_sec(cfs_time_current() + HALF_SEC -
                                            scrub->os_time_last_checkpoint);
        result = osd_scrub_file_store(scrub);
-       if (result < 0)
-               CERROR("%.16s: fail to osd_scrub_post, rc = %d\n",
-                      LDISKFS_SB(osd_scrub2sb(scrub))->s_es->s_volume_name,
-                      result);
        up_write(&scrub->os_rwsem);
 
        EXIT;
@@ -892,6 +903,12 @@ static int osd_iit_iget(struct osd_thread_info *info, struct osd_device *dev,
        int           rc;
        ENTRY;
 
+       /* Not handle the backend root object and agent parent object.
+        * They are neither visible to namespace nor have OI mappings. */
+       if (unlikely(pos == osd_sb(dev)->s_root->d_inode->i_ino ||
+                    pos == osd_remote_parent_ino(dev)))
+               RETURN(SCRUB_NEXT_CONTINUE);
+
        osd_id_gen(lid, pos, OSD_OII_NOGEN);
        inode = osd_iget(info, dev, lid);
        if (IS_ERR(inode)) {
@@ -901,8 +918,9 @@ static int osd_iit_iget(struct osd_thread_info *info, struct osd_device *dev,
                if (rc == -ENOENT || rc == -ESTALE)
                        RETURN(SCRUB_NEXT_CONTINUE);
 
-               CERROR("%.16s: fail to read inode, ino# = %u, rc = %d\n",
-                      LDISKFS_SB(sb)->s_es->s_volume_name, pos, rc);
+               CDEBUG(D_LFSCK, "%.16s: fail to read inode, ino# = %u: "
+                      "rc = %d\n", LDISKFS_SB(sb)->s_es->s_volume_name,
+                      pos, rc);
                RETURN(rc);
        }
 
@@ -937,7 +955,7 @@ static int osd_scrub_next(struct osd_thread_info *info, struct osd_device *dev,
 
                lwi = LWI_TIMEOUT(cfs_time_seconds(cfs_fail_val), NULL, NULL);
                l_wait_event(thread->t_ctl_waitq,
-                            !cfs_list_empty(&scrub->os_inconsistent_items) ||
+                            !list_empty(&scrub->os_inconsistent_items) ||
                             !thread_is_running(thread),
                             &lwi);
        }
@@ -955,11 +973,11 @@ static int osd_scrub_next(struct osd_thread_info *info, struct osd_device *dev,
        if (unlikely(!thread_is_running(thread)))
                return SCRUB_NEXT_EXIT;
 
-       if (!cfs_list_empty(&scrub->os_inconsistent_items)) {
+       if (!list_empty(&scrub->os_inconsistent_items)) {
                struct osd_inconsistent_item *oii;
 
-               oii = cfs_list_entry(scrub->os_inconsistent_items.next,
-                                    struct osd_inconsistent_item, oii_list);
+               oii = list_entry(scrub->os_inconsistent_items.next,
+                                struct osd_inconsistent_item, oii_list);
                *oic = &oii->oii_cache;
                scrub->os_in_prior = 1;
                return 0;
@@ -1014,7 +1032,7 @@ osd_scrub_wakeup(struct osd_scrub *scrub, struct osd_otable_it *it)
 {
        spin_lock(&scrub->os_lock);
        if (osd_scrub_has_window(scrub, &it->ooi_cache) ||
-           !cfs_list_empty(&scrub->os_inconsistent_items) ||
+           !list_empty(&scrub->os_inconsistent_items) ||
            it->ooi_waiting || !thread_is_running(&scrub->os_thread))
                scrub->os_waiting = 0;
        else
@@ -1054,8 +1072,8 @@ static int osd_scrub_exec(struct osd_thread_info *info, struct osd_device *dev,
 
        rc = osd_scrub_checkpoint(scrub);
        if (rc != 0) {
-               CERROR("%.16s: fail to checkpoint, pos = %u, rc = %d\n",
-                      LDISKFS_SB(param->sb)->s_es->s_volume_name,
+               CDEBUG(D_LFSCK, "%.16s: fail to checkpoint, pos = %u: "
+                      "rc = %d\n", osd_scrub2name(scrub),
                       scrub->os_pos_current, rc);
                /* Continue, as long as the scrub itself can go ahead. */
        }
@@ -1080,16 +1098,16 @@ wait:
        if (scrub->os_full_speed || rc == SCRUB_NEXT_CONTINUE)
                return 0;
 
-       if (osd_scrub_has_window(scrub, ooc)) {
+       if (ooc != NULL && osd_scrub_has_window(scrub, ooc)) {
                *noslot = 0;
                return 0;
        }
 
-       l_wait_event(thread->t_ctl_waitq,
-                    osd_scrub_wakeup(scrub, it),
-                    &lwi);
+       if (it != NULL)
+               l_wait_event(thread->t_ctl_waitq, osd_scrub_wakeup(scrub, it),
+                            &lwi);
 
-       if (osd_scrub_has_window(scrub, ooc))
+       if (ooc != NULL && osd_scrub_has_window(scrub, ooc))
                *noslot = 0;
        else
                *noslot = 1;
@@ -1152,7 +1170,7 @@ static int osd_inode_iteration(struct osd_thread_info *info,
                param.gbase = 1 + param.bg * LDISKFS_INODES_PER_GROUP(param.sb);
                param.bitmap = ldiskfs_read_inode_bitmap(param.sb, param.bg);
                if (param.bitmap == NULL) {
-                       CERROR("%.16s: fail to read bitmap for %u, "
+                       CDEBUG(D_LFSCK, "%.16s: fail to read bitmap for %u, "
                               "scrub will stop, urgent mode\n",
                               LDISKFS_SB(param.sb)->s_es->s_volume_name,
                               (__u32)param.bg);
@@ -1206,10 +1224,6 @@ static int osd_otable_it_preload(const struct lu_env *env,
        if (rc == SCRUB_IT_ALL)
                it->ooi_all_cached = 1;
 
-       CDEBUG(D_LFSCK, "OSD pre-loaded: max = %u, preload = %u, rc = %d\n",
-              le32_to_cpu(LDISKFS_SB(osd_sb(dev))->s_es->s_inodes_count),
-              ooc->ooc_pos_preload, rc);
-
        if (scrub->os_waiting && osd_scrub_has_window(scrub, ooc)) {
                spin_lock(&scrub->os_lock);
                scrub->os_waiting = 0;
@@ -1226,21 +1240,20 @@ static int osd_scrub_main(void *args)
        struct osd_device    *dev    = (struct osd_device *)args;
        struct osd_scrub     *scrub  = &dev->od_scrub;
        struct ptlrpc_thread *thread = &scrub->os_thread;
-       struct super_block   *sb     = osd_sb(dev);
        int                   rc;
        ENTRY;
 
        rc = lu_env_init(&env, LCT_LOCAL);
        if (rc != 0) {
-               CERROR("%.16s: OI scrub, fail to init env, rc = %d\n",
-                      LDISKFS_SB(sb)->s_es->s_volume_name, rc);
+               CDEBUG(D_LFSCK, "%.16s: OI scrub fail to init env: rc = %d\n",
+                      osd_scrub2name(scrub), rc);
                GOTO(noenv, rc);
        }
 
        rc = osd_scrub_prep(dev);
        if (rc != 0) {
-               CERROR("%.16s: OI scrub, fail to scrub prep, rc = %d\n",
-                      LDISKFS_SB(sb)->s_es->s_volume_name, rc);
+               CDEBUG(D_LFSCK, "%.16s: OI scrub fail to scrub prep: rc = %d\n",
+                      osd_scrub2name(scrub), rc);
                GOTO(out, rc);
        }
 
@@ -1258,8 +1271,9 @@ static int osd_scrub_main(void *args)
                scrub->os_pos_current = ooc->ooc_pos_preload;
        }
 
-       CDEBUG(D_LFSCK, "OI scrub: flags = 0x%x, pos = %u\n",
-              scrub->os_start_flags, scrub->os_pos_current);
+       CDEBUG(D_LFSCK, "%.16s: OI scrub start, flags = 0x%x, pos = %u\n",
+              osd_scrub2name(scrub), scrub->os_start_flags,
+              scrub->os_pos_current);
 
        rc = osd_inode_iteration(osd_oti_get(&env), dev, ~0U, false);
        if (unlikely(rc == SCRUB_IT_CRASH))
@@ -1268,16 +1282,16 @@ static int osd_scrub_main(void *args)
 
 post:
        osd_scrub_post(scrub, rc);
-       CDEBUG(D_LFSCK, "OI scrub: stop, rc = %d, pos = %u\n",
-              rc, scrub->os_pos_current);
+       CDEBUG(D_LFSCK, "%.16s: OI scrub: stop, pos = %u: rc = %d\n",
+              osd_scrub2name(scrub), scrub->os_pos_current, rc);
 
 out:
-       while (!cfs_list_empty(&scrub->os_inconsistent_items)) {
+       while (!list_empty(&scrub->os_inconsistent_items)) {
                struct osd_inconsistent_item *oii;
 
-               oii = cfs_list_entry(scrub->os_inconsistent_items.next,
+               oii = list_entry(scrub->os_inconsistent_items.next,
                                     struct osd_inconsistent_item, oii_list);
-               cfs_list_del_init(&oii->oii_list);
+               list_del_init(&oii->oii_list);
                OBD_FREE_PTR(oii);
        }
        lu_env_fini(&env);
@@ -1416,8 +1430,8 @@ static const struct osd_lf_map osd_lf_maps[] = {
               OLF_SHOW_NAME, NULL, NULL },
 
        /* lost+found */
-       { "lost+found", { 0, 0, 0 }, OLF_SCAN_SUBITEMS | OLF_NO_OI,
-               osd_ios_general_scan, osd_ios_lf_fill },
+       { "lost+found", { FID_SEQ_LOCAL_FILE, OSD_LPF_OID, 0 },
+               OLF_SCAN_SUBITEMS, osd_ios_general_scan, osd_ios_lf_fill },
 
        { NULL, { 0, 0, 0 }, 0, NULL, NULL }
 };
@@ -1428,17 +1442,25 @@ static const struct osd_lf_map osd_dl_maps[] = {
        { "fid", { FID_SEQ_DOT_LUSTRE, FID_OID_DOT_LUSTRE_OBF, 0 }, 0,
                NULL, NULL },
 
+       /* .lustre/lost+found */
+       { "lost+found", { FID_SEQ_DOT_LUSTRE, FID_OID_DOT_LUSTRE_LPF, 0 }, 0,
+               NULL, NULL },
+
        { NULL, { 0, 0, 0 }, 0, NULL, NULL }
 };
 
 struct osd_ios_item {
-       cfs_list_t       oii_list;
+       struct list_head oii_list;
        struct dentry   *oii_dentry;
        scandir_t        oii_scandir;
        filldir_t        oii_filldir;
 };
 
 struct osd_ios_filldir_buf {
+#ifdef HAVE_DIR_CONTEXT
+       /* please keep it as first member */
+       struct dir_context       ctx;
+#endif
        struct osd_thread_info  *oifb_info;
        struct osd_device       *oifb_dev;
        struct dentry           *oifb_dentry;
@@ -1449,9 +1471,6 @@ osd_ios_lookup_one_len(const char *name, struct dentry *parent, int namelen)
 {
        struct dentry *dentry;
 
-       CDEBUG(D_LFSCK, "init lookup one: parent = %.*s, name = %.*s\n",
-              parent->d_name.len, parent->d_name.name, namelen, name);
-
        dentry = ll_lookup_one_len(name, parent, namelen);
        if (!IS_ERR(dentry) && dentry->d_inode == NULL) {
                dput(dentry);
@@ -1472,11 +1491,11 @@ osd_ios_new_item(struct osd_device *dev, struct dentry *dentry,
        if (item == NULL)
                RETURN(-ENOMEM);
 
-       CFS_INIT_LIST_HEAD(&item->oii_list);
+       INIT_LIST_HEAD(&item->oii_list);
        item->oii_dentry = dget(dentry);
        item->oii_scandir = scandir;
        item->oii_filldir = filldir;
-       cfs_list_add_tail(&item->oii_list, &dev->od_ios_list);
+       list_add_tail(&item->oii_list, &dev->od_ios_list);
 
        RETURN(0);
 }
@@ -1501,11 +1520,13 @@ osd_ios_scan_one(struct osd_thread_info *info, struct osd_device *dev,
        int                      rc;
        ENTRY;
 
-       CDEBUG(D_LFSCK, "init scan one: ino = %ld\n", inode->i_ino);
-
        rc = osd_get_lma(info, inode, &info->oti_obj_dentry, lma);
-       if (rc != 0 && rc != -ENODATA)
+       if (rc != 0 && rc != -ENODATA) {
+               CDEBUG(D_LFSCK, "%s: fail to get lma for init OI scrub: "
+                      "rc = %d\n", osd_name(dev), rc);
+
                RETURN(rc);
+       }
 
        osd_id_gen(id, inode->i_ino, inode->i_generation);
        if (rc == -ENODATA) {
@@ -1514,8 +1535,12 @@ osd_ios_scan_one(struct osd_thread_info *info, struct osd_device *dev,
                else
                        tfid = *fid;
                rc = osd_ea_fid_set(info, inode, &tfid, 0, 0);
-               if (rc != 0)
+               if (rc != 0) {
+                       CDEBUG(D_LFSCK, "%s: fail to set LMA for init OI "
+                             "scrub: rc = %d\n", osd_name(dev), rc);
+
                        RETURN(rc);
+               }
        } else {
                if (lma->lma_compat & LMAC_NOT_IN_OI)
                        RETURN(0);
@@ -1582,7 +1607,7 @@ static int osd_ios_lf_fill(void *buf, const char *name, int namelen,
        scrub->os_lf_scanned++;
        child = osd_ios_lookup_one_len(name, parent, namelen);
        if (IS_ERR(child)) {
-               CWARN("%s: cannot lookup child '%.*s': rc = %d\n",
+               CDEBUG(D_LFSCK, "%s: cannot lookup child '%.*s': rc = %d\n",
                      osd_name(dev), namelen, name, (int)PTR_ERR(child));
                RETURN(0);
        }
@@ -1592,8 +1617,8 @@ static int osd_ios_lf_fill(void *buf, const char *name, int namelen,
                rc = osd_ios_new_item(dev, child, osd_ios_general_scan,
                                      osd_ios_lf_fill);
                if (rc != 0)
-                       CWARN("%s: cannot add child '%.*s': rc = %d\n",
-                             osd_name(dev), namelen, name, rc);
+                       CDEBUG(D_LFSCK, "%s: cannot add child '%.*s': "
+                             "rc = %d\n", osd_name(dev), namelen, name, rc);
                GOTO(put, rc);
        }
 
@@ -1608,8 +1633,9 @@ static int osd_ios_lf_fill(void *buf, const char *name, int namelen,
                               "/lost+found.\n", namelen, name, PFID(fid));
                        scrub->os_lf_repaired++;
                } else {
-                       CWARN("%s: cannot rename for '%.*s' "DFID": rc = %d\n",
-                             osd_name(dev), namelen, name, PFID(fid), rc);
+                       CDEBUG(D_LFSCK, "%s: cannot rename for '%.*s' "
+                              DFID": rc = %d\n",
+                              osd_name(dev), namelen, name, PFID(fid), rc);
                }
        }
 
@@ -1734,7 +1760,13 @@ static int
 osd_ios_general_scan(struct osd_thread_info *info, struct osd_device *dev,
                     struct dentry *dentry, filldir_t filldir)
 {
-       struct osd_ios_filldir_buf    buf   = { info, dev, dentry };
+       struct osd_ios_filldir_buf    buf   = {
+#ifdef HAVE_DIR_CONTEXT
+                                               .ctx.actor = filldir,
+#endif
+                                               .oifb_info = info,
+                                               .oifb_dev = dev,
+                                               .oifb_dentry = dentry };
        struct file                  *filp  = &info->oti_it_ea.oie_file;
        struct inode                 *inode = dentry->d_inode;
        const struct file_operations *fops  = inode->i_fop;
@@ -1751,7 +1783,13 @@ osd_ios_general_scan(struct osd_thread_info *info, struct osd_device *dev,
        filp->private_data = NULL;
        set_file_inode(filp, inode);
 
+#ifdef HAVE_DIR_CONTEXT
+       buf.ctx.pos = filp->f_pos;
+       rc = fops->iterate(filp, &buf.ctx);
+       filp->f_pos = buf.ctx.pos;
+#else
        rc = fops->readdir(filp, &buf, filldir);
+#endif
        fops->release(inode, filp);
 
        RETURN(rc);
@@ -1898,12 +1936,12 @@ static int osd_initial_OI_scrub(struct osd_thread_info *info,
                if (rc != 0)
                        break;
 
-               if (cfs_list_empty(&dev->od_ios_list))
+               if (list_empty(&dev->od_ios_list))
                        break;
 
-               item = cfs_list_entry(dev->od_ios_list.next,
-                                     struct osd_ios_item, oii_list);
-               cfs_list_del_init(&item->oii_list);
+               item = list_entry(dev->od_ios_list.next,
+                                 struct osd_ios_item, oii_list);
+               list_del_init(&item->oii_list);
 
                LASSERT(item->oii_scandir != NULL);
                scandir = item->oii_scandir;
@@ -1911,10 +1949,10 @@ static int osd_initial_OI_scrub(struct osd_thread_info *info,
                dentry = item->oii_dentry;
        }
 
-       while (!cfs_list_empty(&dev->od_ios_list)) {
-               item = cfs_list_entry(dev->od_ios_list.next,
-                                     struct osd_ios_item, oii_list);
-               cfs_list_del_init(&item->oii_list);
+       while (!list_empty(&dev->od_ios_list)) {
+               item = list_entry(dev->od_ios_list.next,
+                                 struct osd_ios_item, oii_list);
+               list_del_init(&item->oii_list);
                dput(item->oii_dentry);
                OBD_FREE_PTR(item);
        }
@@ -1974,6 +2012,7 @@ static int do_osd_scrub_start(struct osd_device *dev, __u32 flags)
        struct osd_scrub     *scrub  = &dev->od_scrub;
        struct ptlrpc_thread *thread = &scrub->os_thread;
        struct l_wait_info    lwi    = { 0 };
+       struct task_struct   *task;
        int                   rc;
        ENTRY;
 
@@ -1992,15 +2031,23 @@ again:
        }
        spin_unlock(&scrub->os_lock);
 
-       if (scrub->os_file.sf_status == SS_COMPLETED)
+       if (scrub->os_file.sf_status == SS_COMPLETED) {
+               if (!(flags & SS_SET_FAILOUT))
+                       flags |= SS_CLEAR_FAILOUT;
+
+               if (!(flags & SS_SET_DRYRUN))
+                       flags |= SS_CLEAR_DRYRUN;
+
                flags |= SS_RESET;
+       }
 
        scrub->os_start_flags = flags;
        thread_set_flags(thread, 0);
-       rc = PTR_ERR(kthread_run(osd_scrub_main, dev, "OI_scrub"));
-       if (IS_ERR_VALUE(rc)) {
-               CERROR("%.16s: cannot start iteration thread, rc = %d\n",
-                      LDISKFS_SB(osd_sb(dev))->s_es->s_volume_name, rc);
+       task = kthread_run(osd_scrub_main, dev, "OI_scrub");
+       if (IS_ERR(task)) {
+               rc = PTR_ERR(task);
+               CERROR("%.16s: cannot start iteration thread: rc = %d\n",
+                      osd_scrub2name(scrub), rc);
                RETURN(rc);
        }
 
@@ -2083,7 +2130,7 @@ int osd_scrub_setup(const struct lu_env *env, struct osd_device *dev)
        init_waitqueue_head(&scrub->os_thread.t_ctl_waitq);
        init_rwsem(&scrub->os_rwsem);
        spin_lock_init(&scrub->os_lock);
-       CFS_INIT_LIST_HEAD(&scrub->os_inconsistent_items);
+       INIT_LIST_HEAD(&scrub->os_inconsistent_items);
 
        push_ctxt(&saved, ctxt);
        filp = filp_open(osd_scrub_name, O_RDWR | O_CREAT, 0644);
@@ -2482,7 +2529,7 @@ int osd_oii_insert(struct osd_device *dev, struct osd_idmap_cache *oic,
        if (unlikely(oii == NULL))
                RETURN(-ENOMEM);
 
-       CFS_INIT_LIST_HEAD(&oii->oii_list);
+       INIT_LIST_HEAD(&oii->oii_list);
        oii->oii_cache = *oic;
        oii->oii_insert = insert;
 
@@ -2493,9 +2540,9 @@ int osd_oii_insert(struct osd_device *dev, struct osd_idmap_cache *oic,
                RETURN(-EAGAIN);
        }
 
-       if (cfs_list_empty(&scrub->os_inconsistent_items))
+       if (list_empty(&scrub->os_inconsistent_items))
                wakeup = 1;
-       cfs_list_add_tail(&oii->oii_list, &scrub->os_inconsistent_items);
+       list_add_tail(&oii->oii_list, &scrub->os_inconsistent_items);
        spin_unlock(&scrub->os_lock);
 
        if (wakeup != 0)
@@ -2512,7 +2559,7 @@ int osd_oii_lookup(struct osd_device *dev, const struct lu_fid *fid,
        ENTRY;
 
        spin_lock(&scrub->os_lock);
-       cfs_list_for_each_entry(oii, &scrub->os_inconsistent_items, oii_list) {
+       list_for_each_entry(oii, &scrub->os_inconsistent_items, oii_list) {
                if (lu_fid_eq(fid, &oii->oii_cache.oic_fid)) {
                        *id = oii->oii_cache.oic_lid;
                        spin_unlock(&scrub->os_lock);
@@ -2551,134 +2598,112 @@ static const char *scrub_param_names[] = {
        NULL
 };
 
-static int scrub_bits_dump(char **buf, int *len, int bits, const char *names[],
+static int scrub_bits_dump(struct seq_file *m, int bits, const char *names[],
                           const char *prefix)
 {
-       int save = *len;
        int flag;
        int rc;
        int i;
 
-       rc = snprintf(*buf, *len, "%s:%c", prefix, bits != 0 ? ' ' : '\n');
-       if (rc <= 0)
-               return -ENOSPC;
+       rc = seq_printf(m, "%s:%c", prefix, bits != 0 ? ' ' : '\n');
+       if (rc < 0)
+               return rc;
 
-       *buf += rc;
-       *len -= rc;
        for (i = 0, flag = 1; bits != 0; i++, flag = 1 << i) {
                if (flag & bits) {
                        bits &= ~flag;
-                       rc = snprintf(*buf, *len, "%s%c", names[i],
-                                     bits != 0 ? ',' : '\n');
-                       if (rc <= 0)
-                               return -ENOSPC;
-
-                       *buf += rc;
-                       *len -= rc;
+                       rc = seq_printf(m, "%s%c", names[i],
+                                       bits != 0 ? ',' : '\n');
+                       if (rc < 0)
+                               return rc;
                }
        }
-       return save - *len;
+       return 0;
 }
 
-static int scrub_time_dump(char **buf, int *len, __u64 time, const char *prefix)
+static int scrub_time_dump(struct seq_file *m, __u64 time, const char *prefix)
 {
        int rc;
 
        if (time != 0)
-               rc = snprintf(*buf, *len, "%s: "LPU64" seconds\n", prefix,
+               rc = seq_printf(m, "%s: "LPU64" seconds\n", prefix,
                              cfs_time_current_sec() - time);
        else
-               rc = snprintf(*buf, *len, "%s: N/A\n", prefix);
-       if (rc <= 0)
-               return -ENOSPC;
-
-       *buf += rc;
-       *len -= rc;
+               rc = seq_printf(m, "%s: N/A\n", prefix);
        return rc;
 }
 
-static int scrub_pos_dump(char **buf, int *len, __u64 pos, const char *prefix)
+static int scrub_pos_dump(struct seq_file *m, __u64 pos, const char *prefix)
 {
        int rc;
 
        if (pos != 0)
-               rc = snprintf(*buf, *len, "%s: "LPU64"\n", prefix, pos);
+               rc = seq_printf(m, "%s: "LPU64"\n", prefix, pos);
        else
-               rc = snprintf(*buf, *len, "%s: N/A\n", prefix);
-       if (rc <= 0)
-               return -ENOSPC;
-
-       *buf += rc;
-       *len -= rc;
+               rc = seq_printf(m, "%s: N/A\n", prefix);
        return rc;
 }
 
-int osd_scrub_dump(struct osd_device *dev, char *buf, int len)
+int osd_scrub_dump(struct seq_file *m, struct osd_device *dev)
 {
        struct osd_scrub  *scrub   = &dev->od_scrub;
        struct scrub_file *sf      = &scrub->os_file;
        __u64              checked;
        __u64              speed;
-       int                save    = len;
-       int                ret     = -ENOSPC;
        int                rc;
 
        down_read(&scrub->os_rwsem);
-       rc = snprintf(buf, len,
-                     "name: OI_scrub\n"
-                     "magic: 0x%x\n"
-                     "oi_files: %d\n"
-                     "status: %s\n",
-                     sf->sf_magic, (int)sf->sf_oi_count,
-                     scrub_status_names[sf->sf_status]);
-       if (rc <= 0)
+       rc = seq_printf(m, "name: OI_scrub\n"
+                       "magic: 0x%x\n"
+                       "oi_files: %d\n"
+                       "status: %s\n",
+                       sf->sf_magic, (int)sf->sf_oi_count,
+                       scrub_status_names[sf->sf_status]);
+       if (rc < 0)
                goto out;
 
-       buf += rc;
-       len -= rc;
-       rc = scrub_bits_dump(&buf, &len, sf->sf_flags, scrub_flags_names,
+       rc = scrub_bits_dump(m, sf->sf_flags, scrub_flags_names,
                             "flags");
        if (rc < 0)
                goto out;
 
-       rc = scrub_bits_dump(&buf, &len, sf->sf_param, scrub_param_names,
+       rc = scrub_bits_dump(m, sf->sf_param, scrub_param_names,
                             "param");
        if (rc < 0)
                goto out;
 
-       rc = scrub_time_dump(&buf, &len, sf->sf_time_last_complete,
+       rc = scrub_time_dump(m, sf->sf_time_last_complete,
                             "time_since_last_completed");
        if (rc < 0)
                goto out;
 
-       rc = scrub_time_dump(&buf, &len, sf->sf_time_latest_start,
+       rc = scrub_time_dump(m, sf->sf_time_latest_start,
                             "time_since_latest_start");
        if (rc < 0)
                goto out;
 
-       rc = scrub_time_dump(&buf, &len, sf->sf_time_last_checkpoint,
+       rc = scrub_time_dump(m, sf->sf_time_last_checkpoint,
                             "time_since_last_checkpoint");
        if (rc < 0)
                goto out;
 
-       rc = scrub_pos_dump(&buf, &len, sf->sf_pos_latest_start,
+       rc = scrub_pos_dump(m, sf->sf_pos_latest_start,
                            "latest_start_position");
        if (rc < 0)
                goto out;
 
-       rc = scrub_pos_dump(&buf, &len, sf->sf_pos_last_checkpoint,
+       rc = scrub_pos_dump(m, sf->sf_pos_last_checkpoint,
                            "last_checkpoint_position");
        if (rc < 0)
                goto out;
 
-       rc = scrub_pos_dump(&buf, &len, sf->sf_pos_first_inconsistent,
+       rc = scrub_pos_dump(m, sf->sf_pos_first_inconsistent,
                            "first_failure_position");
        if (rc < 0)
                goto out;
 
        checked = sf->sf_items_checked + scrub->os_new_checked;
-       rc = snprintf(buf, len,
-                     "checked: "LPU64"\n"
+       rc = seq_printf(m, "checked: "LPU64"\n"
                      "updated: "LPU64"\n"
                      "failed: "LPU64"\n"
                      "prior_updated: "LPU64"\n"
@@ -2688,11 +2713,9 @@ int osd_scrub_dump(struct osd_device *dev, char *buf, int len)
                      checked, sf->sf_items_updated, sf->sf_items_failed,
                      sf->sf_items_updated_prior, sf->sf_items_noscrub,
                      sf->sf_items_igif, sf->sf_success_count);
-       if (rc <= 0)
+       if (rc < 0)
                goto out;
 
-       buf += rc;
-       len -= rc;
        speed = checked;
        if (thread_is_running(&scrub->os_thread)) {
                cfs_duration_t duration = cfs_time_current() -
@@ -2705,8 +2728,7 @@ int osd_scrub_dump(struct osd_device *dev, char *buf, int len)
                        do_div(new_checked, duration);
                if (rtime != 0)
                        do_div(speed, rtime);
-               rc = snprintf(buf, len,
-                             "run_time: %u seconds\n"
+               rc = seq_printf(m, "run_time: %u seconds\n"
                              "average_speed: "LPU64" objects/sec\n"
                              "real-time_speed: "LPU64" objects/sec\n"
                              "current_position: %u\n"
@@ -2719,8 +2741,7 @@ int osd_scrub_dump(struct osd_device *dev, char *buf, int len)
        } else {
                if (sf->sf_run_time != 0)
                        do_div(speed, sf->sf_run_time);
-               rc = snprintf(buf, len,
-                             "run_time: %u seconds\n"
+               rc = seq_printf(m, "run_time: %u seconds\n"
                              "average_speed: "LPU64" objects/sec\n"
                              "real-time_speed: N/A\n"
                              "current_position: N/A\n"
@@ -2730,14 +2751,8 @@ int osd_scrub_dump(struct osd_device *dev, char *buf, int len)
                              sf->sf_run_time, speed, scrub->os_lf_scanned,
                              scrub->os_lf_repaired, scrub->os_lf_failed);
        }
-       if (rc <= 0)
-               goto out;
-
-       buf += rc;
-       len -= rc;
-       ret = save - len;
 
 out:
        up_read(&scrub->os_rwsem);
-       return ret;
+       return (rc < 0 ? -ENOSPC : 0);
 }