1 files changed, 268 insertions, 85 deletions
diff --git a/fs/exofs/super.c b/fs/exofs/super.c
index 9f500dec3b59..a1d1e77b12eb 100644
--- a/fs/exofs/super.c
+++ b/fs/exofs/super.c
@@ -203,49 +203,45 @@ int exofs_sync_fs(struct super_block *sb, int wait)
 {
        struct exofs_sb_info *sbi;
        struct exofs_fscb *fscb;
-        struct osd_request *or;
+        struct exofs_io_state *ios;
-        struct osd_obj_id obj;
        int ret = -ENOMEM;
-        fscb = kzalloc(sizeof(struct exofs_fscb), GFP_KERNEL);
-        if (!fscb) {
-                EXOFS_ERR("exofs_write_super: memory allocation failed.\n");
-                return -ENOMEM;
-        }
        lock_super(sb);
        sbi = sb->s_fs_info;
+        fscb = &sbi->s_fscb;
+        ret = exofs_get_io_state(sbi, &ios);
+        if (ret)
+                goto out;
+        /* Note: We only write the changing part of the fscb. .i.e upto the
+         *       the fscb->s_dev_table_oid member. There is no read-modify-write
+         *       here.
+         */
+        ios->length = offsetof(struct exofs_fscb, s_dev_table_oid);
+        memset(fscb, 0, ios->length);
        fscb->s_nextid = cpu_to_le64(sbi->s_nextid);
        fscb->s_numfiles = cpu_to_le32(sbi->s_numfiles);
        fscb->s_magic = cpu_to_le16(sb->s_magic);
        fscb->s_newfs = 0;
+        fscb->s_version = EXOFS_FSCB_VER;
-        or = osd_start_request(sbi->s_dev, GFP_KERNEL);
+        ios->obj.id = EXOFS_SUPER_ID;
-        if (unlikely(!or)) {
+        ios->offset = 0;
-                EXOFS_ERR("exofs_write_super: osd_start_request failed.\n");
+        ios->kern_buff = fscb;
-                goto out;
+        ios->cred = sbi->s_cred;
-        }
-        obj.partition = sbi->s_pid;
+        ret = exofs_sbi_write(ios);
-        obj.id = EXOFS_SUPER_ID;
-        ret = osd_req_write_kern(or, &obj, 0, fscb, sizeof(*fscb));
        if (unlikely(ret)) {
-                EXOFS_ERR("exofs_write_super: osd_req_write_kern failed.\n");
+                EXOFS_ERR("%s: exofs_sbi_write failed.\n", __func__);
-                goto out;
-        }
-        ret = exofs_sync_op(or, sbi->s_timeout, sbi->s_cred);
-        if (unlikely(ret)) {
-                EXOFS_ERR("exofs_write_super: exofs_sync_op failed.\n");
                goto out;
        }
        sb->s_dirt = 0;
 out:
-        if (or)
+        EXOFS_DBGMSG("s_nextid=0x%llx ret=%d\n", _LLU(sbi->s_nextid), ret);
-                osd_end_request(or);
+        exofs_put_io_state(ios);
        unlock_super(sb);
-        kfree(fscb);
        return ret;
 }
@@ -257,6 +253,29 @@ static void exofs_write_super(struct super_block *sb)
                sb->s_dirt = 0;
 }
+static void _exofs_print_device(const char *msg, const char *dev_path,
+                                struct osd_dev *od, u64 pid)
+{
+        const struct osd_dev_info *odi = osduld_device_info(od);
+        printk(KERN_NOTICE "exofs: %s %s osd_name-%s pid-0x%llx\n",
+                msg, dev_path ?: "", odi->osdname, _LLU(pid));
+}
+void exofs_free_sbi(struct exofs_sb_info *sbi)
+{
+        while (sbi->s_numdevs) {
+                int i = --sbi->s_numdevs;
+                struct osd_dev *od = sbi->s_ods[i];
+                if (od) {
+                        sbi->s_ods[i] = NULL;
+                        osduld_put_device(od);
+                }
+        }
+        kfree(sbi);
+}
 /*
 * This function is called when the vfs is freeing the superblock.  We just
 * need to free our own part.
@@ -279,11 +298,182 @@ static void exofs_put_super(struct super_block *sb)
                                  msecs_to_jiffies(100));
        }
-        osduld_put_device(sbi->s_dev);
+        _exofs_print_device("Unmounting", NULL, sbi->s_ods[0], sbi->s_pid);
-        kfree(sb->s_fs_info);
+        exofs_free_sbi(sbi);
        sb->s_fs_info = NULL;
 }
+static int _read_and_match_data_map(struct exofs_sb_info *sbi, unsigned numdevs,
+                                    struct exofs_device_table *dt)
+{
+        sbi->data_map.odm_num_comps   =
+                                le32_to_cpu(dt->dt_data_map.cb_num_comps);
+        sbi->data_map.odm_stripe_unit =
+                                le64_to_cpu(dt->dt_data_map.cb_stripe_unit);
+        sbi->data_map.odm_group_width =
+                                le32_to_cpu(dt->dt_data_map.cb_group_width);
+        sbi->data_map.odm_group_depth =
+                                le32_to_cpu(dt->dt_data_map.cb_group_depth);
+        sbi->data_map.odm_mirror_cnt  =
+                                le32_to_cpu(dt->dt_data_map.cb_mirror_cnt);
+        sbi->data_map.odm_raid_algorithm  =
+                                le32_to_cpu(dt->dt_data_map.cb_raid_algorithm);
+/* FIXME: Hard coded mirror only for now. if not so do not mount */
+        if ((sbi->data_map.odm_num_comps != numdevs) ||
+            (sbi->data_map.odm_stripe_unit != EXOFS_BLKSIZE) ||
+            (sbi->data_map.odm_raid_algorithm != PNFS_OSD_RAID_0) ||
+            (sbi->data_map.odm_mirror_cnt != (numdevs - 1)))
+                return -EINVAL;
+        else
+                return 0;
+}
+/* @odi is valid only as long as @fscb_dev is valid */
+static int exofs_devs_2_odi(struct exofs_dt_device_info *dt_dev,
+                             struct osd_dev_info *odi)
+{
+        odi->systemid_len = le32_to_cpu(dt_dev->systemid_len);
+        memcpy(odi->systemid, dt_dev->systemid, odi->systemid_len);
+        odi->osdname_len = le32_to_cpu(dt_dev->osdname_len);
+        odi->osdname = dt_dev->osdname;
+        /* FIXME support long names. Will need a _put function */
+        if (dt_dev->long_name_offset)
+                return -EINVAL;
+        /* Make sure osdname is printable!
+         * mkexofs should give us space for a null-terminator else the
+         * device-table is invalid.
+         */
+        if (unlikely(odi->osdname_len >= sizeof(dt_dev->osdname)))
+                odi->osdname_len = sizeof(dt_dev->osdname) - 1;
+        dt_dev->osdname[odi->osdname_len] = 0;
+        /* If it's all zeros something is bad we read past end-of-obj */
+        return !(odi->systemid_len || odi->osdname_len);
+}
+static int exofs_read_lookup_dev_table(struct exofs_sb_info **psbi,
+                                       unsigned table_count)
+{
+        struct exofs_sb_info *sbi = *psbi;
+        struct osd_dev *fscb_od;
+        struct osd_obj_id obj = {.partition = sbi->s_pid,
+                                 .id = EXOFS_DEVTABLE_ID};
+        struct exofs_device_table *dt;
+        unsigned table_bytes = table_count * sizeof(dt->dt_dev_table[0]) +
+                                             sizeof(*dt);
+        unsigned numdevs, i;
+        int ret;
+        dt = kmalloc(table_bytes, GFP_KERNEL);
+        if (unlikely(!dt)) {
+                EXOFS_ERR("ERROR: allocating %x bytes for device table\n",
+                          table_bytes);
+                return -ENOMEM;
+        }
+        fscb_od = sbi->s_ods[0];
+        sbi->s_ods[0] = NULL;
+        sbi->s_numdevs = 0;
+        ret = exofs_read_kern(fscb_od, sbi->s_cred, &obj, 0, dt, table_bytes);
+        if (unlikely(ret)) {
+                EXOFS_ERR("ERROR: reading device table\n");
+                goto out;
+        }
+        numdevs = le64_to_cpu(dt->dt_num_devices);
+        if (unlikely(!numdevs)) {
+                ret = -EINVAL;
+                goto out;
+        }
+        WARN_ON(table_count != numdevs);
+        ret = _read_and_match_data_map(sbi, numdevs, dt);
+        if (unlikely(ret))
+                goto out;
+        if (likely(numdevs > 1)) {
+                unsigned size = numdevs * sizeof(sbi->s_ods[0]);
+                sbi = krealloc(sbi, sizeof(*sbi) + size, GFP_KERNEL);
+                if (unlikely(!sbi)) {
+                        ret = -ENOMEM;
+                        goto out;
+                }
+                memset(&sbi->s_ods[1], 0, size - sizeof(sbi->s_ods[0]));
+                *psbi = sbi;
+        }
+        for (i = 0; i < numdevs; i++) {
+                struct exofs_fscb fscb;
+                struct osd_dev_info odi;
+                struct osd_dev *od;
+                if (exofs_devs_2_odi(&dt->dt_dev_table[i], &odi)) {
+                        EXOFS_ERR("ERROR: Read all-zeros device entry\n");
+                        ret = -EINVAL;
+                        goto out;
+                }
+                printk(KERN_NOTICE "Add device[%d]: osd_name-%s\n",
+                       i, odi.osdname);
+                /* On all devices the device table is identical. The user can
+                 * specify any one of the participating devices on the command
+                 * line. We always keep them in device-table order.
+                 */
+                if (fscb_od && osduld_device_same(fscb_od, &odi)) {
+                        sbi->s_ods[i] = fscb_od;
+                        ++sbi->s_numdevs;
+                        fscb_od = NULL;
+                        continue;
+                }
+                od = osduld_info_lookup(&odi);
+                if (unlikely(IS_ERR(od))) {
+                        ret = PTR_ERR(od);
+                        EXOFS_ERR("ERROR: device requested is not found "
+                                  "osd_name-%s =>%d\n", odi.osdname, ret);
+                        goto out;
+                }
+                sbi->s_ods[i] = od;
+                ++sbi->s_numdevs;
+                /* Read the fscb of the other devices to make sure the FS
+                 * partition is there.
+                 */
+                ret = exofs_read_kern(od, sbi->s_cred, &obj, 0, &fscb,
+                                      sizeof(fscb));
+                if (unlikely(ret)) {
+                        EXOFS_ERR("ERROR: Malformed participating device "
+                                  "error reading fscb osd_name-%s\n",
+                                  odi.osdname);
+                        goto out;
+                }
+                /* TODO: verify other information is correct and FS-uuid
+                 *       matches. Benny what did you say about device table
+                 *       generation and old devices?
+                 */
+        }
+out:
+        kfree(dt);
+        if (unlikely(!ret && fscb_od)) {
+                EXOFS_ERR(
+                      "ERROR: Bad device-table container device not present\n");
+                osduld_put_device(fscb_od);
+                ret = -EINVAL;
+        }
+        return ret;
+}
 /*
 * Read the superblock from the OSD and fill in the fields
 */
@@ -292,24 +482,25 @@ static int exofs_fill_super(struct super_block *sb, void *data, int silent)
        struct inode *root;
        struct exofs_mountopt *opts = data;
        struct exofs_sb_info *sbi;      /*extended info                  */
+        struct osd_dev *od;             /* Master device                 */
        struct exofs_fscb fscb;         /*on-disk superblock info        */
-        struct osd_request *or = NULL;
        struct osd_obj_id obj;
+        unsigned table_count;
        int ret;
        sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
        if (!sbi)
                return -ENOMEM;
-        sb->s_fs_info = sbi;
        /* use mount options to fill superblock */
-        sbi->s_dev = osduld_path_lookup(opts->dev_name);
+        od = osduld_path_lookup(opts->dev_name);
-        if (IS_ERR(sbi->s_dev)) {
+        if (IS_ERR(od)) {
-                ret = PTR_ERR(sbi->s_dev);
+                ret = PTR_ERR(od);
-                sbi->s_dev = NULL;
                goto free_sbi;
        }
+        sbi->s_ods[0] = od;
+        sbi->s_numdevs = 1;
        sbi->s_pid = opts->pid;
        sbi->s_timeout = opts->timeout;
@@ -323,35 +514,13 @@ static int exofs_fill_super(struct super_block *sb, void *data, int silent)
        sb->s_bdev = NULL;
        sb->s_dev = 0;
-        /* read data from on-disk superblock object */
        obj.partition = sbi->s_pid;
        obj.id = EXOFS_SUPER_ID;
        exofs_make_credential(sbi->s_cred, &obj);
-        or = osd_start_request(sbi->s_dev, GFP_KERNEL);
+        ret = exofs_read_kern(od, sbi->s_cred, &obj, 0, &fscb, sizeof(fscb));
-        if (unlikely(!or)) {
+        if (unlikely(ret))
-                if (!silent)
-                        EXOFS_ERR(
-                               "exofs_fill_super: osd_start_request failed.\n");
-                ret = -ENOMEM;
-                goto free_sbi;
-        }
-        ret = osd_req_read_kern(or, &obj, 0, &fscb, sizeof(fscb));
-        if (unlikely(ret)) {
-                if (!silent)
-                        EXOFS_ERR(
-                               "exofs_fill_super: osd_req_read_kern failed.\n");
-                ret = -ENOMEM;
-                goto free_sbi;
-        }
-        ret = exofs_sync_op(or, sbi->s_timeout, sbi->s_cred);
-        if (unlikely(ret)) {
-                if (!silent)
-                        EXOFS_ERR("exofs_fill_super: exofs_sync_op failed.\n");
-                ret = -EIO;
                goto free_sbi;
-        }
        sb->s_magic = le16_to_cpu(fscb.s_magic);
        sbi->s_nextid = le64_to_cpu(fscb.s_nextid);
@@ -364,12 +533,26 @@ static int exofs_fill_super(struct super_block *sb, void *data, int silent)
                ret = -EINVAL;
                goto free_sbi;
        }
+        if (le32_to_cpu(fscb.s_version) != EXOFS_FSCB_VER) {
+                EXOFS_ERR("ERROR: Bad FSCB version expected-%d got-%d\n",
+                          EXOFS_FSCB_VER, le32_to_cpu(fscb.s_version));
+                ret = -EINVAL;
+                goto free_sbi;
+        }
        /* start generation numbers from a random point */
        get_random_bytes(&sbi->s_next_generation, sizeof(u32));
        spin_lock_init(&sbi->s_next_gen_lock);
+        table_count = le64_to_cpu(fscb.s_dev_table_count);
+        if (table_count) {
+                ret = exofs_read_lookup_dev_table(&sbi, table_count);
+                if (unlikely(ret))
+                        goto free_sbi;
+        }
        /* set up operation vectors */
+        sb->s_fs_info = sbi;
        sb->s_op = &exofs_sops;
        sb->s_export_op = &exofs_export_ops;
        root = exofs_iget(sb, EXOFS_ROOT_ID - EXOFS_OBJ_OFF);
@@ -395,16 +578,15 @@ static int exofs_fill_super(struct super_block *sb, void *data, int silent)
                goto free_sbi;
        }
-        ret = 0;
+        _exofs_print_device("Mounting", opts->dev_name, sbi->s_ods[0],
-out:
+                            sbi->s_pid);
-        if (or)
+        return 0;
-                osd_end_request(or);
-        return ret;
 free_sbi:
-        osduld_put_device(sbi->s_dev); /* NULL safe */
+        EXOFS_ERR("Unable to mount exofs on %s pid=0x%llx err=%d\n",
-        kfree(sbi);
+                  opts->dev_name, sbi->s_pid, ret);
-        goto out;
+        exofs_free_sbi(sbi);
+        return ret;
 }
 /*
@@ -433,7 +615,7 @@ static int exofs_statfs(struct dentry *dentry, struct kstatfs *buf)
 {
        struct super_block *sb = dentry->d_sb;
        struct exofs_sb_info *sbi = sb->s_fs_info;
-        struct osd_obj_id obj = {sbi->s_pid, 0};
+        struct exofs_io_state *ios;
        struct osd_attr attrs[] = {
                ATTR_DEF(OSD_APAGE_PARTITION_QUOTAS,
                        OSD_ATTR_PQ_CAPACITY_QUOTA, sizeof(__be64)),
@@ -442,32 +624,33 @@ static int exofs_statfs(struct dentry *dentry, struct kstatfs *buf)
        };
        uint64_t capacity = ULLONG_MAX;
        uint64_t used = ULLONG_MAX;
-        struct osd_request *or;
        uint8_t cred_a[OSD_CAP_LEN];
        int ret;
-        /* get used/capacity attributes */
+        ret = exofs_get_io_state(sbi, &ios);
-        exofs_make_credential(cred_a, &obj);
+        if (ret) {
+                EXOFS_DBGMSG("exofs_get_io_state failed.\n");
-        or = osd_start_request(sbi->s_dev, GFP_KERNEL);
+                return ret;
-        if (unlikely(!or)) {
-                EXOFS_DBGMSG("exofs_statfs: osd_start_request failed.\n");
-                return -ENOMEM;
        }
-        osd_req_get_attributes(or, &obj);
+        exofs_make_credential(cred_a, &ios->obj);
-        osd_req_add_get_attr_list(or, attrs, ARRAY_SIZE(attrs));
+        ios->cred = sbi->s_cred;
-        ret = exofs_sync_op(or, sbi->s_timeout, cred_a);
+        ios->in_attr = attrs;
+        ios->in_attr_len = ARRAY_SIZE(attrs);
+        ret = exofs_sbi_read(ios);
        if (unlikely(ret))
                goto out;
-        ret = extract_attr_from_req(or, &attrs[0]);
+        ret = extract_attr_from_ios(ios, &attrs[0]);
-        if (likely(!ret))
+        if (likely(!ret)) {
                capacity = get_unaligned_be64(attrs[0].val_ptr);
-        else
+                if (unlikely(!capacity))
+                        capacity = ULLONG_MAX;
+        } else
                EXOFS_DBGMSG("exofs_statfs: get capacity failed.\n");
-        ret = extract_attr_from_req(or, &attrs[1]);
+        ret = extract_attr_from_ios(ios, &attrs[1]);
        if (likely(!ret))
                used = get_unaligned_be64(attrs[1].val_ptr);
        else
@@ -476,15 +659,15 @@ static int exofs_statfs(struct dentry *dentry, struct kstatfs *buf)
        /* fill in the stats buffer */
        buf->f_type = EXOFS_SUPER_MAGIC;
        buf->f_bsize = EXOFS_BLKSIZE;
-        buf->f_blocks = (capacity >> EXOFS_BLKSHIFT);
+        buf->f_blocks = capacity >> 9;
-        buf->f_bfree = ((capacity - used) >> EXOFS_BLKSHIFT);
+        buf->f_bfree = (capacity - used) >> 9;
        buf->f_bavail = buf->f_bfree;
        buf->f_files = sbi->s_numfiles;
        buf->f_ffree = EXOFS_MAX_ID - sbi->s_numfiles;
        buf->f_namelen = EXOFS_NAME_LEN;
 out:
-        osd_end_request(or);
+        exofs_put_io_state(ios);
        return ret;
 }

diff --git a/fs/exofs/super.c b/fs/exofs/super.c index 9f500dec3b59..a1d1e77b12eb 100644 --- a/fs/exofs/super.c +++ b/fs/exofs/super.c
@@ -203,49 +203,45 @@ int exofs_sync_fs(struct super_block *sb, int wait)
203	{	203	{
204	struct exofs_sb_info *sbi;	204	struct exofs_sb_info *sbi;
205	struct exofs_fscb *fscb;	205	struct exofs_fscb *fscb;
206	struct osd_request *or;	206	struct exofs_io_state *ios;
207	struct osd_obj_id obj;
208	int ret = -ENOMEM;	207	int ret = -ENOMEM;
209		208
210	fscb = kzalloc(sizeof(struct exofs_fscb), GFP_KERNEL);
211	if (!fscb) {
212	EXOFS_ERR("exofs_write_super: memory allocation failed.\n");
213	return -ENOMEM;
214	}
215
216	lock_super(sb);	209	lock_super(sb);
217	sbi = sb->s_fs_info;	210	sbi = sb->s_fs_info;
		211	fscb = &sbi->s_fscb;
		212
		213	ret = exofs_get_io_state(sbi, &ios);
		214	if (ret)
		215	goto out;
		216
		217	/* Note: We only write the changing part of the fscb. .i.e upto the
		218	* the fscb->s_dev_table_oid member. There is no read-modify-write
		219	* here.
		220	*/
		221	ios->length = offsetof(struct exofs_fscb, s_dev_table_oid);
		222	memset(fscb, 0, ios->length);
218	fscb->s_nextid = cpu_to_le64(sbi->s_nextid);	223	fscb->s_nextid = cpu_to_le64(sbi->s_nextid);
219	fscb->s_numfiles = cpu_to_le32(sbi->s_numfiles);	224	fscb->s_numfiles = cpu_to_le32(sbi->s_numfiles);
220	fscb->s_magic = cpu_to_le16(sb->s_magic);	225	fscb->s_magic = cpu_to_le16(sb->s_magic);
221	fscb->s_newfs = 0;	226	fscb->s_newfs = 0;
		227	fscb->s_version = EXOFS_FSCB_VER;
222		228
223	or = osd_start_request(sbi->s_dev, GFP_KERNEL);	229	ios->obj.id = EXOFS_SUPER_ID;
224	if (unlikely(!or)) {	230	ios->offset = 0;
225	EXOFS_ERR("exofs_write_super: osd_start_request failed.\n");	231	ios->kern_buff = fscb;
226	goto out;	232	ios->cred = sbi->s_cred;
227	}
228		233
229	obj.partition = sbi->s_pid;	234	ret = exofs_sbi_write(ios);
230	obj.id = EXOFS_SUPER_ID;
231	ret = osd_req_write_kern(or, &obj, 0, fscb, sizeof(*fscb));
232	if (unlikely(ret)) {	235	if (unlikely(ret)) {
233	EXOFS_ERR("exofs_write_super: osd_req_write_kern failed.\n");	236	EXOFS_ERR("%s: exofs_sbi_write failed.\n", __func__);
234	goto out;
235	}
236
237	ret = exofs_sync_op(or, sbi->s_timeout, sbi->s_cred);
238	if (unlikely(ret)) {
239	EXOFS_ERR("exofs_write_super: exofs_sync_op failed.\n");
240	goto out;	237	goto out;
241	}	238	}
242	sb->s_dirt = 0;	239	sb->s_dirt = 0;
243		240
244	out:	241	out:
245	if (or)	242	EXOFS_DBGMSG("s_nextid=0x%llx ret=%d\n", _LLU(sbi->s_nextid), ret);
246	osd_end_request(or);	243	exofs_put_io_state(ios);
247	unlock_super(sb);	244	unlock_super(sb);
248	kfree(fscb);
249	return ret;	245	return ret;
250	}	246	}
251		247
@@ -257,6 +253,29 @@ static void exofs_write_super(struct super_block *sb)
257	sb->s_dirt = 0;	253	sb->s_dirt = 0;
258	}	254	}
259		255
		256	static void _exofs_print_device(const char msg, const char dev_path,
		257	struct osd_dev *od, u64 pid)
		258	{
		259	const struct osd_dev_info *odi = osduld_device_info(od);
		260
		261	printk(KERN_NOTICE "exofs: %s %s osd_name-%s pid-0x%llx\n",
		262	msg, dev_path ?: "", odi->osdname, _LLU(pid));
		263	}
		264
		265	void exofs_free_sbi(struct exofs_sb_info *sbi)
		266	{
		267	while (sbi->s_numdevs) {
		268	int i = --sbi->s_numdevs;
		269	struct osd_dev *od = sbi->s_ods[i];
		270
		271	if (od) {
		272	sbi->s_ods[i] = NULL;
		273	osduld_put_device(od);
		274	}
		275	}
		276	kfree(sbi);
		277	}
		278
260	/*	279	/*
261	* This function is called when the vfs is freeing the superblock. We just	280	* This function is called when the vfs is freeing the superblock. We just
262	* need to free our own part.	281	* need to free our own part.
@@ -279,11 +298,182 @@ static void exofs_put_super(struct super_block *sb)
279	msecs_to_jiffies(100));	298	msecs_to_jiffies(100));
280	}	299	}
281		300
282	osduld_put_device(sbi->s_dev);	301	_exofs_print_device("Unmounting", NULL, sbi->s_ods[0], sbi->s_pid);
283	kfree(sb->s_fs_info);	302
		303	exofs_free_sbi(sbi);
284	sb->s_fs_info = NULL;	304	sb->s_fs_info = NULL;
285	}	305	}
286		306
		307	static int _read_and_match_data_map(struct exofs_sb_info *sbi, unsigned numdevs,
		308	struct exofs_device_table *dt)
		309	{
		310	sbi->data_map.odm_num_comps =
		311	le32_to_cpu(dt->dt_data_map.cb_num_comps);
		312	sbi->data_map.odm_stripe_unit =
		313	le64_to_cpu(dt->dt_data_map.cb_stripe_unit);
		314	sbi->data_map.odm_group_width =
		315	le32_to_cpu(dt->dt_data_map.cb_group_width);
		316	sbi->data_map.odm_group_depth =
		317	le32_to_cpu(dt->dt_data_map.cb_group_depth);
		318	sbi->data_map.odm_mirror_cnt =
		319	le32_to_cpu(dt->dt_data_map.cb_mirror_cnt);
		320	sbi->data_map.odm_raid_algorithm =
		321	le32_to_cpu(dt->dt_data_map.cb_raid_algorithm);
		322
		323	/* FIXME: Hard coded mirror only for now. if not so do not mount */
		324	if ((sbi->data_map.odm_num_comps != numdevs) \|\|
		325	(sbi->data_map.odm_stripe_unit != EXOFS_BLKSIZE) \|\|
		326	(sbi->data_map.odm_raid_algorithm != PNFS_OSD_RAID_0) \|\|
		327	(sbi->data_map.odm_mirror_cnt != (numdevs - 1)))
		328	return -EINVAL;
		329	else
		330	return 0;
		331	}
		332
		333	/* @odi is valid only as long as @fscb_dev is valid */
		334	static int exofs_devs_2_odi(struct exofs_dt_device_info *dt_dev,
		335	struct osd_dev_info *odi)
		336	{
		337	odi->systemid_len = le32_to_cpu(dt_dev->systemid_len);
		338	memcpy(odi->systemid, dt_dev->systemid, odi->systemid_len);
		339
		340	odi->osdname_len = le32_to_cpu(dt_dev->osdname_len);
		341	odi->osdname = dt_dev->osdname;
		342
		343	/* FIXME support long names. Will need a _put function */
		344	if (dt_dev->long_name_offset)
		345	return -EINVAL;
		346
		347	/* Make sure osdname is printable!
		348	* mkexofs should give us space for a null-terminator else the
		349	* device-table is invalid.
		350	*/
		351	if (unlikely(odi->osdname_len >= sizeof(dt_dev->osdname)))
		352	odi->osdname_len = sizeof(dt_dev->osdname) - 1;
		353	dt_dev->osdname[odi->osdname_len] = 0;
		354
		355	/* If it's all zeros something is bad we read past end-of-obj */
		356	return !(odi->systemid_len \|\| odi->osdname_len);
		357	}
		358
		359	static int exofs_read_lookup_dev_table(struct exofs_sb_info **psbi,
		360	unsigned table_count)
		361	{
		362	struct exofs_sb_info sbi = psbi;
		363	struct osd_dev *fscb_od;
		364	struct osd_obj_id obj = {.partition = sbi->s_pid,
		365	.id = EXOFS_DEVTABLE_ID};
		366	struct exofs_device_table *dt;
		367	unsigned table_bytes = table_count * sizeof(dt->dt_dev_table[0]) +
		368	sizeof(*dt);
		369	unsigned numdevs, i;
		370	int ret;
		371
		372	dt = kmalloc(table_bytes, GFP_KERNEL);
		373	if (unlikely(!dt)) {
		374	EXOFS_ERR("ERROR: allocating %x bytes for device table\n",
		375	table_bytes);
		376	return -ENOMEM;
		377	}
		378
		379	fscb_od = sbi->s_ods[0];
		380	sbi->s_ods[0] = NULL;
		381	sbi->s_numdevs = 0;
		382	ret = exofs_read_kern(fscb_od, sbi->s_cred, &obj, 0, dt, table_bytes);
		383	if (unlikely(ret)) {
		384	EXOFS_ERR("ERROR: reading device table\n");
		385	goto out;
		386	}
		387
		388	numdevs = le64_to_cpu(dt->dt_num_devices);
		389	if (unlikely(!numdevs)) {
		390	ret = -EINVAL;
		391	goto out;
		392	}
		393	WARN_ON(table_count != numdevs);
		394
		395	ret = _read_and_match_data_map(sbi, numdevs, dt);
		396	if (unlikely(ret))
		397	goto out;
		398
		399	if (likely(numdevs > 1)) {
		400	unsigned size = numdevs * sizeof(sbi->s_ods[0]);
		401
		402	sbi = krealloc(sbi, sizeof(*sbi) + size, GFP_KERNEL);
		403	if (unlikely(!sbi)) {
		404	ret = -ENOMEM;
		405	goto out;
		406	}
		407	memset(&sbi->s_ods[1], 0, size - sizeof(sbi->s_ods[0]));
		408	*psbi = sbi;
		409	}
		410
		411	for (i = 0; i < numdevs; i++) {
		412	struct exofs_fscb fscb;
		413	struct osd_dev_info odi;
		414	struct osd_dev *od;
		415
		416	if (exofs_devs_2_odi(&dt->dt_dev_table[i], &odi)) {
		417	EXOFS_ERR("ERROR: Read all-zeros device entry\n");
		418	ret = -EINVAL;
		419	goto out;
		420	}
		421
		422	printk(KERN_NOTICE "Add device[%d]: osd_name-%s\n",
		423	i, odi.osdname);
		424
		425	/* On all devices the device table is identical. The user can
		426	* specify any one of the participating devices on the command
		427	* line. We always keep them in device-table order.
		428	*/
		429	if (fscb_od && osduld_device_same(fscb_od, &odi)) {
		430	sbi->s_ods[i] = fscb_od;
		431	++sbi->s_numdevs;
		432	fscb_od = NULL;
		433	continue;
		434	}
		435
		436	od = osduld_info_lookup(&odi);
		437	if (unlikely(IS_ERR(od))) {
		438	ret = PTR_ERR(od);
		439	EXOFS_ERR("ERROR: device requested is not found "
		440	"osd_name-%s =>%d\n", odi.osdname, ret);
		441	goto out;
		442	}
		443
		444	sbi->s_ods[i] = od;
		445	++sbi->s_numdevs;
		446
		447	/* Read the fscb of the other devices to make sure the FS
		448	* partition is there.
		449	*/
		450	ret = exofs_read_kern(od, sbi->s_cred, &obj, 0, &fscb,
		451	sizeof(fscb));
		452	if (unlikely(ret)) {
		453	EXOFS_ERR("ERROR: Malformed participating device "
		454	"error reading fscb osd_name-%s\n",
		455	odi.osdname);
		456	goto out;
		457	}
		458
		459	/* TODO: verify other information is correct and FS-uuid
		460	* matches. Benny what did you say about device table
		461	* generation and old devices?
		462	*/
		463	}
		464
		465	out:
		466	kfree(dt);
		467	if (unlikely(!ret && fscb_od)) {
		468	EXOFS_ERR(
		469	"ERROR: Bad device-table container device not present\n");
		470	osduld_put_device(fscb_od);
		471	ret = -EINVAL;
		472	}
		473
		474	return ret;
		475	}
		476
287	/*	477	/*
288	* Read the superblock from the OSD and fill in the fields	478	* Read the superblock from the OSD and fill in the fields
289	*/	479	*/
@@ -292,24 +482,25 @@ static int exofs_fill_super(struct super_block sb, void data, int silent)
292	struct inode *root;	482	struct inode *root;
293	struct exofs_mountopt *opts = data;	483	struct exofs_mountopt *opts = data;
294	struct exofs_sb_info sbi; /extended info */	484	struct exofs_sb_info sbi; /extended info */
		485	struct osd_dev od; / Master device */
295	struct exofs_fscb fscb; /on-disk superblock info /	486	struct exofs_fscb fscb; /on-disk superblock info /
296	struct osd_request *or = NULL;
297	struct osd_obj_id obj;	487	struct osd_obj_id obj;
		488	unsigned table_count;
298	int ret;	489	int ret;
299		490
300	sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);	491	sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
301	if (!sbi)	492	if (!sbi)
302	return -ENOMEM;	493	return -ENOMEM;
303	sb->s_fs_info = sbi;
304		494
305	/* use mount options to fill superblock */	495	/* use mount options to fill superblock */
306	sbi->s_dev = osduld_path_lookup(opts->dev_name);	496	od = osduld_path_lookup(opts->dev_name);
307	if (IS_ERR(sbi->s_dev)) {	497	if (IS_ERR(od)) {
308	ret = PTR_ERR(sbi->s_dev);	498	ret = PTR_ERR(od);
309	sbi->s_dev = NULL;
310	goto free_sbi;	499	goto free_sbi;
311	}	500	}
312		501
		502	sbi->s_ods[0] = od;
		503	sbi->s_numdevs = 1;
313	sbi->s_pid = opts->pid;	504	sbi->s_pid = opts->pid;
314	sbi->s_timeout = opts->timeout;	505	sbi->s_timeout = opts->timeout;
315		506
@@ -323,35 +514,13 @@ static int exofs_fill_super(struct super_block sb, void data, int silent)
323	sb->s_bdev = NULL;	514	sb->s_bdev = NULL;
324	sb->s_dev = 0;	515	sb->s_dev = 0;
325		516
326	/* read data from on-disk superblock object */
327	obj.partition = sbi->s_pid;	517	obj.partition = sbi->s_pid;
328	obj.id = EXOFS_SUPER_ID;	518	obj.id = EXOFS_SUPER_ID;
329	exofs_make_credential(sbi->s_cred, &obj);	519	exofs_make_credential(sbi->s_cred, &obj);
330		520
331	or = osd_start_request(sbi->s_dev, GFP_KERNEL);	521	ret = exofs_read_kern(od, sbi->s_cred, &obj, 0, &fscb, sizeof(fscb));
332	if (unlikely(!or)) {	522	if (unlikely(ret))
333	if (!silent)
334	EXOFS_ERR(
335	"exofs_fill_super: osd_start_request failed.\n");
336	ret = -ENOMEM;
337	goto free_sbi;
338	}
339	ret = osd_req_read_kern(or, &obj, 0, &fscb, sizeof(fscb));
340	if (unlikely(ret)) {
341	if (!silent)
342	EXOFS_ERR(
343	"exofs_fill_super: osd_req_read_kern failed.\n");
344	ret = -ENOMEM;
345	goto free_sbi;
346	}
347
348	ret = exofs_sync_op(or, sbi->s_timeout, sbi->s_cred);
349	if (unlikely(ret)) {
350	if (!silent)
351	EXOFS_ERR("exofs_fill_super: exofs_sync_op failed.\n");
352	ret = -EIO;
353	goto free_sbi;	523	goto free_sbi;
354	}
355		524
356	sb->s_magic = le16_to_cpu(fscb.s_magic);	525	sb->s_magic = le16_to_cpu(fscb.s_magic);
357	sbi->s_nextid = le64_to_cpu(fscb.s_nextid);	526	sbi->s_nextid = le64_to_cpu(fscb.s_nextid);
@@ -364,12 +533,26 @@ static int exofs_fill_super(struct super_block sb, void data, int silent)
364	ret = -EINVAL;	533	ret = -EINVAL;
365	goto free_sbi;	534	goto free_sbi;
366	}	535	}
		536	if (le32_to_cpu(fscb.s_version) != EXOFS_FSCB_VER) {
		537	EXOFS_ERR("ERROR: Bad FSCB version expected-%d got-%d\n",
		538	EXOFS_FSCB_VER, le32_to_cpu(fscb.s_version));
		539	ret = -EINVAL;
		540	goto free_sbi;
		541	}
367		542
368	/* start generation numbers from a random point */	543	/* start generation numbers from a random point */
369	get_random_bytes(&sbi->s_next_generation, sizeof(u32));	544	get_random_bytes(&sbi->s_next_generation, sizeof(u32));
370	spin_lock_init(&sbi->s_next_gen_lock);	545	spin_lock_init(&sbi->s_next_gen_lock);
371		546
		547	table_count = le64_to_cpu(fscb.s_dev_table_count);
		548	if (table_count) {
		549	ret = exofs_read_lookup_dev_table(&sbi, table_count);
		550	if (unlikely(ret))
		551	goto free_sbi;
		552	}
		553
372	/* set up operation vectors */	554	/* set up operation vectors */
		555	sb->s_fs_info = sbi;
373	sb->s_op = &exofs_sops;	556	sb->s_op = &exofs_sops;
374	sb->s_export_op = &exofs_export_ops;	557	sb->s_export_op = &exofs_export_ops;
375	root = exofs_iget(sb, EXOFS_ROOT_ID - EXOFS_OBJ_OFF);	558	root = exofs_iget(sb, EXOFS_ROOT_ID - EXOFS_OBJ_OFF);
@@ -395,16 +578,15 @@ static int exofs_fill_super(struct super_block sb, void data, int silent)
395	goto free_sbi;	578	goto free_sbi;
396	}	579	}
397		580
398	ret = 0;	581	_exofs_print_device("Mounting", opts->dev_name, sbi->s_ods[0],
399	out:	582	sbi->s_pid);
400	if (or)	583	return 0;
401	osd_end_request(or);
402	return ret;
403		584
404	free_sbi:	585	free_sbi:
405	osduld_put_device(sbi->s_dev); /* NULL safe */	586	EXOFS_ERR("Unable to mount exofs on %s pid=0x%llx err=%d\n",
406	kfree(sbi);	587	opts->dev_name, sbi->s_pid, ret);
407	goto out;	588	exofs_free_sbi(sbi);
		589	return ret;
408	}	590	}
409		591
410	/*	592	/*
@@ -433,7 +615,7 @@ static int exofs_statfs(struct dentry dentry, struct kstatfs buf)
433	{	615	{
434	struct super_block *sb = dentry->d_sb;	616	struct super_block *sb = dentry->d_sb;
435	struct exofs_sb_info *sbi = sb->s_fs_info;	617	struct exofs_sb_info *sbi = sb->s_fs_info;
436	struct osd_obj_id obj = {sbi->s_pid, 0};	618	struct exofs_io_state *ios;
437	struct osd_attr attrs[] = {	619	struct osd_attr attrs[] = {
438	ATTR_DEF(OSD_APAGE_PARTITION_QUOTAS,	620	ATTR_DEF(OSD_APAGE_PARTITION_QUOTAS,
439	OSD_ATTR_PQ_CAPACITY_QUOTA, sizeof(__be64)),	621	OSD_ATTR_PQ_CAPACITY_QUOTA, sizeof(__be64)),
@@ -442,32 +624,33 @@ static int exofs_statfs(struct dentry dentry, struct kstatfs buf)
442	};	624	};
443	uint64_t capacity = ULLONG_MAX;	625	uint64_t capacity = ULLONG_MAX;
444	uint64_t used = ULLONG_MAX;	626	uint64_t used = ULLONG_MAX;
445	struct osd_request *or;
446	uint8_t cred_a[OSD_CAP_LEN];	627	uint8_t cred_a[OSD_CAP_LEN];
447	int ret;	628	int ret;
448		629
449	/* get used/capacity attributes */	630	ret = exofs_get_io_state(sbi, &ios);
450	exofs_make_credential(cred_a, &obj);	631	if (ret) {
451		632	EXOFS_DBGMSG("exofs_get_io_state failed.\n");
452	or = osd_start_request(sbi->s_dev, GFP_KERNEL);	633	return ret;
453	if (unlikely(!or)) {
454	EXOFS_DBGMSG("exofs_statfs: osd_start_request failed.\n");
455	return -ENOMEM;
456	}	634	}
457		635
458	osd_req_get_attributes(or, &obj);	636	exofs_make_credential(cred_a, &ios->obj);
459	osd_req_add_get_attr_list(or, attrs, ARRAY_SIZE(attrs));	637	ios->cred = sbi->s_cred;
460	ret = exofs_sync_op(or, sbi->s_timeout, cred_a);	638	ios->in_attr = attrs;
		639	ios->in_attr_len = ARRAY_SIZE(attrs);
		640
		641	ret = exofs_sbi_read(ios);
461	if (unlikely(ret))	642	if (unlikely(ret))
462	goto out;	643	goto out;
463		644
464	ret = extract_attr_from_req(or, &attrs[0]);	645	ret = extract_attr_from_ios(ios, &attrs[0]);
465	if (likely(!ret))	646	if (likely(!ret)) {
466	capacity = get_unaligned_be64(attrs[0].val_ptr);	647	capacity = get_unaligned_be64(attrs[0].val_ptr);
467	else	648	if (unlikely(!capacity))
		649	capacity = ULLONG_MAX;
		650	} else
468	EXOFS_DBGMSG("exofs_statfs: get capacity failed.\n");	651	EXOFS_DBGMSG("exofs_statfs: get capacity failed.\n");
469		652
470	ret = extract_attr_from_req(or, &attrs[1]);	653	ret = extract_attr_from_ios(ios, &attrs[1]);
471	if (likely(!ret))	654	if (likely(!ret))
472	used = get_unaligned_be64(attrs[1].val_ptr);	655	used = get_unaligned_be64(attrs[1].val_ptr);
473	else	656	else
@@ -476,15 +659,15 @@ static int exofs_statfs(struct dentry dentry, struct kstatfs buf)
476	/* fill in the stats buffer */	659	/* fill in the stats buffer */
477	buf->f_type = EXOFS_SUPER_MAGIC;	660	buf->f_type = EXOFS_SUPER_MAGIC;
478	buf->f_bsize = EXOFS_BLKSIZE;	661	buf->f_bsize = EXOFS_BLKSIZE;
479	buf->f_blocks = (capacity >> EXOFS_BLKSHIFT);	662	buf->f_blocks = capacity >> 9;
480	buf->f_bfree = ((capacity - used) >> EXOFS_BLKSHIFT);	663	buf->f_bfree = (capacity - used) >> 9;
481	buf->f_bavail = buf->f_bfree;	664	buf->f_bavail = buf->f_bfree;
482	buf->f_files = sbi->s_numfiles;	665	buf->f_files = sbi->s_numfiles;
483	buf->f_ffree = EXOFS_MAX_ID - sbi->s_numfiles;	666	buf->f_ffree = EXOFS_MAX_ID - sbi->s_numfiles;
484	buf->f_namelen = EXOFS_NAME_LEN;	667	buf->f_namelen = EXOFS_NAME_LEN;
485		668
486	out:	669	out:
487	osd_end_request(or);	670	exofs_put_io_state(ios);
488	return ret;	671	return ret;
489	}	672	}
490		673