Whamcloud - gitweb
LU-911 osd: OI is implemented internally within OSD
[fs/lustre-release.git] / lustre / osd-ldiskfs / osd_oi.c
index ac6744f..c1ce0a6 100644 (file)
@@ -96,6 +96,131 @@ static struct dt_index_features oi_feat = {
 
 #define OSD_OI_NAME_BASE        "oi.16"
 
+static void osd_oi_table_put(struct osd_thread_info *info,
+                             struct osd_oi **oi_table, unsigned oi_count)
+{
+        struct iam_container *bag;
+        int                   i;
+
+        for (i = 0; i < oi_count; i++) {
+                LASSERT(oi_table[i] != NULL);
+                LASSERT(oi_table[i]->oi_inode != NULL);
+
+                bag = &(oi_table[i]->oi_dir.od_container);
+                if (bag->ic_object == oi_table[i]->oi_inode)
+                        iam_container_fini(bag);
+                iput(oi_table[i]->oi_inode);
+                oi_table[i]->oi_inode = NULL;
+                OBD_FREE_PTR(oi_table[i]);
+        }
+}
+
+static int osd_oi_index_create_one(struct osd_thread_info *info,
+                                   struct osd_device *osd, const char *name,
+                                   struct dt_index_features *feat)
+{
+        const struct lu_env             *env = info->oti_env;
+        struct osd_inode_id             *id  = &info->oti_id;
+        struct buffer_head              *bh;
+        struct inode                    *inode;
+        struct ldiskfs_dir_entry_2      *de;
+        struct dentry                   *dentry;
+        struct inode                    *dir;
+        handle_t                        *jh;
+        int                              rc;
+
+        dentry = osd_child_dentry_by_inode(env, osd_sb(osd)->s_root->d_inode,
+                                           name, strlen(name));
+        dir = osd_sb(osd)->s_root->d_inode;
+        bh = osd_ldiskfs_find_entry(dir, dentry, &de, NULL);
+        if (bh) {
+                brelse(bh);
+
+                id->oii_ino = le32_to_cpu(de->inode);
+                id->oii_gen = OSD_OII_NOGEN;
+
+                inode = osd_iget(info, osd, id);
+                if (!IS_ERR(inode)) {
+                        iput(inode);
+                        RETURN(-EEXIST);
+                }
+                RETURN(PTR_ERR(inode));
+        }
+
+        jh = ldiskfs_journal_start_sb(osd_sb(osd), 100);
+        LASSERT(!IS_ERR(jh));
+
+        inode = ldiskfs_create_inode(jh, osd_sb(osd)->s_root->d_inode,
+                                     (S_IFREG | S_IRUGO | S_IWUSR));
+        LASSERT(!IS_ERR(inode));
+
+        if (feat->dif_flags & DT_IND_VARKEY)
+                rc = iam_lvar_create(inode, feat->dif_keysize_max,
+                                     feat->dif_ptrsize, feat->dif_recsize_max,
+                                     jh);
+        else
+                rc = iam_lfix_create(inode, feat->dif_keysize_max,
+                                     feat->dif_ptrsize, feat->dif_recsize_max,
+                                     jh);
+
+        dentry = osd_child_dentry_by_inode(env, osd_sb(osd)->s_root->d_inode,
+                                           name, strlen(name));
+        rc = osd_ldiskfs_add_entry(jh, dentry, inode, NULL);
+        LASSERT(rc == 0);
+
+        ldiskfs_journal_stop(jh);
+        iput(inode);
+
+        return rc;
+}
+
+static struct inode *osd_oi_index_open(struct osd_thread_info *info,
+                                       struct osd_device *osd,
+                                       const char *name,
+                                       struct dt_index_features *f,
+                                       bool create)
+{
+        struct dentry *dentry;
+        struct inode  *inode;
+        int            rc;
+
+        dentry = ll_lookup_one_len(name, osd_sb(osd)->s_root, strlen(name));
+        if (IS_ERR(dentry))
+                return (void *) dentry;
+
+        if (dentry->d_inode) {
+                LASSERT(!is_bad_inode(dentry->d_inode));
+                inode = dentry->d_inode;
+                atomic_inc(&inode->i_count);
+                dput(dentry);
+                return inode;
+        }
+
+        /* create */
+        dput(dentry);
+        shrink_dcache_parent(osd_sb(osd)->s_root);
+        if (!create)
+                return ERR_PTR(-ENOENT);
+
+        rc = osd_oi_index_create_one(info, osd, name, f);
+        if (rc)
+                RETURN(ERR_PTR(rc));
+
+        dentry = ll_lookup_one_len(name, osd_sb(osd)->s_root, strlen(name));
+        if (IS_ERR(dentry))
+                return (void *) dentry;
+
+        if (dentry->d_inode) {
+                LASSERT(!is_bad_inode(dentry->d_inode));
+                inode = dentry->d_inode;
+                atomic_inc(&inode->i_count);
+                dput(dentry);
+                return inode;
+        }
+
+        return ERR_PTR(-ENOENT);
+}
+
 /**
  * Open an OI(Ojbect Index) container.
  *
@@ -105,46 +230,50 @@ static struct dt_index_features oi_feat = {
  * \retval      0       success
  * \retval      -ve     failure
  */
-static int
-osd_oi_open(struct osd_thread_info *info,
-            struct dt_device *dev, char *name, struct dt_object **objp)
+static int osd_oi_open(struct osd_thread_info *info, struct osd_device *osd,
+                       char *name, struct osd_oi **oi_slot, bool create)
 {
-        const struct lu_env *env = info->oti_env;
-        struct dt_object    *obj;
-        int                  rc;
+        struct osd_directory *dir;
+        struct iam_container *bag;
+        struct inode         *inode;
+        struct osd_oi        *oi;
+        int                   rc;
 
-        obj = dt_store_open(env, dev, "", name, &info->oti_fid);
-        if (IS_ERR(obj))
-                return PTR_ERR(obj);
+        ENTRY;
 
-        oi_feat.dif_keysize_min = sizeof(info->oti_fid);
-        oi_feat.dif_keysize_max = sizeof(info->oti_fid);
+        oi_feat.dif_keysize_min = sizeof(struct lu_fid);
+        oi_feat.dif_keysize_max = sizeof(struct lu_fid);
 
-        rc = obj->do_ops->do_index_try(env, obj, &oi_feat);
-        if (rc != 0) {
-                lu_object_put(info->oti_env, &obj->do_lu);
-                CERROR("%s: wrong index %s: rc = %d\n",
-                       dev->dd_lu_dev.ld_obd->obd_name, name, rc);
-                return rc;
-        }
+        inode = osd_oi_index_open(info, osd, name, &oi_feat, create);
+        if (IS_ERR(inode))
+                RETURN(PTR_ERR(inode));
 
-        *objp = obj;
-        return 0;
-}
+        OBD_ALLOC_PTR(oi);
+        if (oi == NULL)
+                GOTO(out_inode, rc = -ENOMEM);
 
+        oi->oi_inode = inode;
+        dir = &oi->oi_dir;
 
-static void
-osd_oi_table_put(struct osd_thread_info *info,
-                 struct osd_oi *oi_table, unsigned oi_count)
-{
-        int     i;
+        bag = &dir->od_container;
+        rc = iam_container_init(bag, &dir->od_descr, inode);
+        if (rc < 0)
+                GOTO(out_free, rc);
 
-        for (i = 0; i < oi_count; i++) {
-                LASSERT(oi_table[i].oi_dir != NULL);
+        rc = iam_container_setup(bag);
+        if (rc < 0)
+                GOTO(out_container, rc);
 
-                lu_object_put(info->oti_env, &oi_table[i].oi_dir->do_lu);
-                oi_table[i].oi_dir = NULL;
-        }
+        *oi_slot = oi;
+        RETURN(0);
+
+out_container:
+        iam_container_fini(bag);
+out_free:
+        OBD_FREE_PTR(oi);
+out_inode:
+        iput(inode);
+        return rc;
 }
 
 /**
@@ -158,19 +287,20 @@ osd_oi_table_put(struct osd_thread_info *info,
  * success, or error code in failure.
  *
  * \param     oi_count  Number of expected OI containers
- * \param     try_all   Try to open all OIs even see failures
+ * \param     create    Create OIs if doesn't exist
  *
  * \retval    +ve       number of opened OI containers
  * \retval      0       no OI containers found
  * \retval    -ve       failure
  */
 static int
-osd_oi_table_open(struct osd_thread_info *info, struct dt_device *dev,
-                  struct osd_oi *oi_table, unsigned oi_count, int try_all)
+osd_oi_table_open(struct osd_thread_info *info, struct osd_device *osd,
+                  struct osd_oi **oi_table, unsigned oi_count, bool create)
 {
-        int     count = 0;
-        int     rc = 0;
-        int     i;
+        struct dt_device *dev = &osd->od_dt_dev;
+        int               count = 0;
+        int               rc = 0;
+        int               i;
 
         /* NB: oi_count != 0 means that we have already created/known all OIs
          * and have known exact number of OIs. */
@@ -180,32 +310,24 @@ osd_oi_table_open(struct osd_thread_info *info, struct dt_device *dev,
                 char name[12];
 
                 sprintf(name, "%s.%d", OSD_OI_NAME_BASE, i);
-                rc = osd_oi_open(info, dev, name, &oi_table[i].oi_dir);
+                rc = osd_oi_open(info, osd, name, &oi_table[i], create);
                 if (rc == 0) {
                         count++;
                         continue;
                 }
 
-                if (try_all)
-                        continue;
-
                 if (rc == -ENOENT && oi_count == 0)
                         return count;
 
                 CERROR("%s: can't open %s: rc = %d\n",
                        dev->dd_lu_dev.ld_obd->obd_name, name, rc);
-
                 if (oi_count > 0) {
                         CERROR("%s: expect to open total %d OI files.\n",
                                dev->dd_lu_dev.ld_obd->obd_name, oi_count);
                 }
-
                 break;
         }
 
-        if (try_all)
-                return count;
-
         if (rc < 0) {
                 osd_oi_table_put(info, oi_table, count);
                 return rc;
@@ -214,172 +336,248 @@ osd_oi_table_open(struct osd_thread_info *info, struct dt_device *dev,
         return count;
 }
 
-static int osd_oi_table_create(struct osd_thread_info *info,
-                               struct dt_device *dev,
-                               struct md_device *mdev, int oi_count)
-{
-        const struct lu_env *env;
-        struct md_object *mdo;
-        int i;
-
-        env = info->oti_env;
-        for (i = 0; i < oi_count; ++i) {
-                char name[12];
-
-                sprintf(name, "%s.%d", OSD_OI_NAME_BASE, i);
-
-                lu_local_obj_fid(&info->oti_fid, OSD_OI_FID_OID_FIRST + i);
-                oi_feat.dif_keysize_min = sizeof(info->oti_fid);
-                oi_feat.dif_keysize_max = sizeof(info->oti_fid);
-
-                mdo = llo_store_create_index(env, mdev, dev, "", name,
-                                             &info->oti_fid, &oi_feat);
-                if (IS_ERR(mdo)) {
-                        CERROR("Failed to create OI[%d] on %s: %d\n",
-                               i, dev->dd_lu_dev.ld_obd->obd_name,
-                               (int)PTR_ERR(mdo));
-                        RETURN(PTR_ERR(mdo));
-                }
-
-                lu_object_put(env, &mdo->mo_lu);
-        }
-        return 0;
-}
-
-int osd_oi_init(struct osd_thread_info *info,
-                struct osd_oi **oi_table,
-                struct dt_device *dev,
-                struct md_device *mdev)
+int osd_oi_init(struct osd_thread_info *info, struct osd_device *osd)
 {
-        struct osd_oi *oi;
-        int rc;
+        struct dt_device *dev = &osd->od_dt_dev;
+        struct osd_oi   **oi;
+        int               rc;
 
         OBD_ALLOC(oi, sizeof(*oi) * OSD_OI_FID_NR_MAX);
         if (oi == NULL)
                 return -ENOMEM;
 
         cfs_mutex_lock(&oi_init_lock);
-
-        rc = osd_oi_table_open(info, dev, oi, 0, 0);
+        /* try to open existing multiple OIs first */
+        rc = osd_oi_table_open(info, osd, oi, 0, false);
         if (rc != 0)
                 goto out;
 
-        rc = osd_oi_open(info, dev, OSD_OI_NAME_BASE, &oi[0].oi_dir);
+        /* if previous failed then try found single OI from old filesystem */
+        rc = osd_oi_open(info, osd, OSD_OI_NAME_BASE, &oi[0], false);
         if (rc == 0) { /* found single OI from old filesystem */
                 rc = 1;
                 goto out;
-        }
-
-        if (rc != -ENOENT) {
+        } else if (rc != -ENOENT) {
                 CERROR("%s: can't open %s: rc = %d\n",
                        dev->dd_lu_dev.ld_obd->obd_name, OSD_OI_NAME_BASE, rc);
                 goto out;
         }
 
-        /* create OI objects */
-        rc = osd_oi_table_create(info, dev, mdev, osd_oi_count);
-        if (rc != 0)
-                goto out;
-
-        rc = osd_oi_table_open(info, dev, oi, osd_oi_count, 0);
-        LASSERT(rc == osd_oi_count || rc < 0);
-
- out:
+        /* No OIs exist, new filesystem, create OI objects */
+        rc = osd_oi_table_open(info, osd, oi, osd_oi_count, true);
+        LASSERT(ergo(rc >= 0, rc == osd_oi_count));
+out:
         if (rc < 0) {
                 OBD_FREE(oi, sizeof(*oi) * OSD_OI_FID_NR_MAX);
         } else {
                 LASSERT((rc & (rc - 1)) == 0);
-                *oi_table = oi;
+                osd->od_oi_table = oi;
+                osd->od_oi_count = rc;
+                rc = 0;
         }
 
         cfs_mutex_unlock(&oi_init_lock);
         return rc;
 }
 
-void osd_oi_fini(struct osd_thread_info *info,
-                 struct osd_oi **oi_table, unsigned oi_count)
+void osd_oi_fini(struct osd_thread_info *info, struct osd_device *osd)
 {
-        struct osd_oi *oi = *oi_table;
+        osd_oi_table_put(info, osd->od_oi_table, osd->od_oi_count);
 
-        osd_oi_table_put(info, oi, oi_count);
+        OBD_FREE(osd->od_oi_table,
+                 sizeof(*(osd->od_oi_table)) * OSD_OI_FID_NR_MAX);
+        osd->od_oi_table = NULL;
+}
 
-        OBD_FREE(oi, sizeof(*oi) * OSD_OI_FID_NR_MAX);
-        *oi_table = NULL;
+static inline int fid_is_fs_root(const struct lu_fid *fid)
+{
+        /* Map root inode to special local object FID */
+        return (unlikely(fid_seq(fid) == FID_SEQ_LOCAL_FILE &&
+                         fid_oid(fid) == OSD_FS_ROOT_OID));
 }
 
-int osd_oi_lookup(struct osd_thread_info *info, struct osd_oi *oi,
+static int osd_oi_iam_lookup(struct osd_thread_info *oti,
+                             struct osd_oi *oi, struct dt_rec *rec,
+                             const struct dt_key *key)
+{
+        struct iam_container  *bag;
+        struct iam_iterator   *it = &oti->oti_idx_it;
+        struct iam_rec        *iam_rec;
+        struct iam_path_descr *ipd;
+        int                    rc;
+        ENTRY;
+
+        LASSERT(oi);
+        LASSERT(oi->oi_inode);
+
+        bag = &oi->oi_dir.od_container;
+        ipd = osd_idx_ipd_get(oti->oti_env, bag);
+        if (IS_ERR(ipd))
+                RETURN(-ENOMEM);
+
+        /* got ipd now we can start iterator. */
+        iam_it_init(it, bag, 0, ipd);
+
+        rc = iam_it_get(it, (struct iam_key *)key);
+        if (rc >= 0) {
+                if (S_ISDIR(oi->oi_inode->i_mode))
+                        iam_rec = (struct iam_rec *)oti->oti_ldp;
+                else
+                        iam_rec = (struct iam_rec *)rec;
+
+                iam_reccpy(&it->ii_path.ip_leaf, (struct iam_rec *)iam_rec);
+                if (S_ISDIR(oi->oi_inode->i_mode))
+                        osd_fid_unpack((struct lu_fid *)rec,
+                                       (struct osd_fid_pack *)iam_rec);
+        }
+        iam_it_put(it);
+        iam_it_fini(it);
+        osd_ipd_put(oti->oti_env, bag, ipd);
+
+        LINVRNT(osd_invariant(obj));
+
+        RETURN(rc);
+}
+
+int osd_oi_lookup(struct osd_thread_info *info, struct osd_device *osd,
                   const struct lu_fid *fid, struct osd_inode_id *id)
 {
-        struct lu_fid *oi_fid = &info->oti_fid;
-        int rc;
+        struct lu_fid       *oi_fid = &info->oti_fid;
+        const struct dt_key *key;
+        int                  rc     = 0;
 
         if (osd_fid_is_igif(fid)) {
                 lu_igif_to_id(fid, id);
                 rc = 0;
         } else {
-                struct dt_object    *idx;
-                const struct dt_key *key;
-
                 if (!fid_is_norm(fid))
                         return -ENOENT;
 
-                idx = oi->oi_dir;
                 fid_cpu_to_be(oi_fid, fid);
                 key = (struct dt_key *) oi_fid;
-                rc = idx->do_index_ops->dio_lookup(info->oti_env, idx,
-                                                   (struct dt_rec *)id, key,
-                                                   BYPASS_CAPA);
+
+                rc = osd_oi_iam_lookup(info, osd_fid2oi(osd, fid),
+                                       (struct dt_rec *)id, key);
+
                 if (rc > 0) {
                         id->oii_ino = be32_to_cpu(id->oii_ino);
                         id->oii_gen = be32_to_cpu(id->oii_gen);
                         rc = 0;
-                } else if (rc == 0)
+                } else if (rc == 0) {
                         rc = -ENOENT;
+                }
         }
         return rc;
 }
 
-int osd_oi_insert(struct osd_thread_info *info, struct osd_oi *oi,
+static int osd_oi_iam_insert(struct osd_thread_info *oti, struct osd_oi *oi,
+                             const struct dt_rec *rec, const struct dt_key *key,
+                             struct thandle *th, int ignore_quota)
+{
+        struct iam_container  *bag;
+        struct iam_rec        *iam_rec = (struct iam_rec *)oti->oti_ldp;
+        struct iam_path_descr *ipd;
+        struct osd_thandle    *oh;
+        int                    rc;
+#ifdef HAVE_QUOTA_SUPPORT
+        cfs_cap_t              save    = cfs_curproc_cap_pack();
+#endif
+        ENTRY;
+
+        LASSERT(oi);
+        LASSERT(oi->oi_inode);
+
+        bag = &oi->oi_dir.od_container;
+        ipd = osd_idx_ipd_get(oti->oti_env, bag);
+        if (unlikely(ipd == NULL))
+                RETURN(-ENOMEM);
+
+        oh = container_of0(th, struct osd_thandle, ot_super);
+        LASSERT(oh->ot_handle != NULL);
+        LASSERT(oh->ot_handle->h_transaction != NULL);
+#ifdef HAVE_QUOTA_SUPPORT
+        if (ignore_quota)
+                cfs_cap_raise(CFS_CAP_SYS_RESOURCE);
+        else
+                cfs_cap_lower(CFS_CAP_SYS_RESOURCE);
+#endif
+        if (S_ISDIR(oi->oi_inode->i_mode))
+                osd_fid_pack((struct osd_fid_pack *)iam_rec, rec,
+                             &oti->oti_fid);
+        else
+                iam_rec = (struct iam_rec *) rec;
+        rc = iam_insert(oh->ot_handle, bag, (const struct iam_key *)key,
+                        iam_rec, ipd);
+#ifdef HAVE_QUOTA_SUPPORT
+        cfs_curproc_cap_unpack(save);
+#endif
+        osd_ipd_put(oti->oti_env, bag, ipd);
+        LINVRNT(osd_invariant(obj));
+        RETURN(rc);
+}
+
+int osd_oi_insert(struct osd_thread_info *info, struct osd_device *osd,
                   const struct lu_fid *fid, const struct osd_inode_id *id0,
                   struct thandle *th, int ignore_quota)
 {
-        struct lu_fid *oi_fid = &info->oti_fid;
-        struct dt_object    *idx;
+        struct lu_fid       *oi_fid = &info->oti_fid;
         struct osd_inode_id *id;
         const struct dt_key *key;
 
         if (!fid_is_norm(fid))
                 return 0;
 
-        idx = oi->oi_dir;
         fid_cpu_to_be(oi_fid, fid);
-        key = (struct dt_key *) oi_fid;
+        key = (struct dt_key *)oi_fid;
 
         id  = &info->oti_id;
         id->oii_ino = cpu_to_be32(id0->oii_ino);
         id->oii_gen = cpu_to_be32(id0->oii_gen);
-        return idx->do_index_ops->dio_insert(info->oti_env, idx,
-                                             (struct dt_rec *)id,
-                                             key, th, BYPASS_CAPA,
-                                             ignore_quota);
+
+        return osd_oi_iam_insert(info, osd_fid2oi(osd, fid),
+                                 (struct dt_rec *)id, key, th, ignore_quota);
+}
+
+static int osd_oi_iam_delete(struct osd_thread_info *oti, struct osd_oi *oi,
+                             const struct dt_key *key, struct thandle *handle)
+{
+        struct iam_container  *bag;
+        struct iam_path_descr *ipd;
+        struct osd_thandle    *oh;
+        int                    rc;
+        ENTRY;
+
+        LASSERT(oi);
+
+        bag = &oi->oi_dir.od_container;
+        ipd = osd_idx_ipd_get(oti->oti_env, bag);
+        if (unlikely(ipd == NULL))
+                RETURN(-ENOMEM);
+
+        oh = container_of0(handle, struct osd_thandle, ot_super);
+        LASSERT(oh->ot_handle != NULL);
+        LASSERT(oh->ot_handle->h_transaction != NULL);
+
+        rc = iam_delete(oh->ot_handle, bag, (const struct iam_key *)key, ipd);
+        osd_ipd_put(oti->oti_env, bag, ipd);
+        LINVRNT(osd_invariant(obj));
+        RETURN(rc);
 }
 
 int osd_oi_delete(struct osd_thread_info *info,
-                  struct osd_oi *oi, const struct lu_fid *fid,
+                  struct osd_device *osd, const struct lu_fid *fid,
                   struct thandle *th)
 {
-        struct lu_fid *oi_fid = &info->oti_fid;
-        struct dt_object    *idx;
+        struct lu_fid       *oi_fid = &info->oti_fid;
         const struct dt_key *key;
 
         if (!fid_is_norm(fid))
                 return 0;
 
-        idx = oi->oi_dir;
         fid_cpu_to_be(oi_fid, fid);
-        key = (struct dt_key *) oi_fid;
-        return idx->do_index_ops->dio_delete(info->oti_env, idx,
-                                             key, th, BYPASS_CAPA);
+        key = (struct dt_key *)oi_fid;
+
+        return osd_oi_iam_delete(info, osd_fid2oi(osd, fid), key, th);
 }
 
 int osd_oi_mod_init()