1 files changed, 374 insertions, 171 deletions

diff --git a/fs/exofs/super.c b/fs/exofs/super.c
index 79c3ae6e045..ed73ed8ebbe 100644
--- a/fs/exofs/super.c
+++ b/fs/exofs/super.c
@@ -35,11 +35,14 @@
 #include <linux/parser.h>
 #include <linux/vfs.h>
 #include <linux/random.h>
+#include <linux/module.h>
 #include <linux/exportfs.h>
 #include <linux/slab.h>
 
 #include "exofs.h"
 
+#define EXOFS_DBGMSG2(M...) do {} while (0)
+
 /******************************************************************************
  * MOUNT OPTIONS
  *****************************************************************************/
@@ -48,6 +51,7 @@
  * struct to hold what we get from mount options
  */
 struct exofs_mountopt {
+	bool is_osdname;
 	const char *dev_name;
 	uint64_t pid;
 	int timeout;
@@ -56,7 +60,7 @@ struct exofs_mountopt {
 /*
  * exofs-specific mount-time options.
  */
-enum { Opt_pid, Opt_to, Opt_mkfs, Opt_format, Opt_err };
+enum { Opt_name, Opt_pid, Opt_to, Opt_err };
 
 /*
  * Our mount-time options.  These should ideally be 64-bit unsigned, but the
@@ -64,6 +68,7 @@ enum { Opt_pid, Opt_to, Opt_mkfs, Opt_format, Opt_err };
  * sufficient for most applications now.
  */
 static match_table_t tokens = {
+	{Opt_name, "osdname=%s"},
 	{Opt_pid, "pid=%u"},
 	{Opt_to, "to=%u"},
 	{Opt_err, NULL}
@@ -94,6 +99,14 @@ static int parse_options(char *options, struct exofs_mountopt *opts)
 
 		token = match_token(p, tokens, args);
 		switch (token) {
+		case Opt_name:
+			opts->dev_name = match_strdup(&args[0]);
+			if (unlikely(!opts->dev_name)) {
+				EXOFS_ERR("Error allocating dev_name");
+				return -ENOMEM;
+			}
+			opts->is_osdname = true;
+			break;
 		case Opt_pid:
 			if (0 == match_strlcpy(str, &args[0], sizeof(str)))
 				return -EINVAL;
@@ -150,12 +163,18 @@ static struct inode *exofs_alloc_inode(struct super_block *sb)
 	return &oi->vfs_inode;
 }
 
+static void exofs_i_callback(struct rcu_head *head)
+{
+	struct inode *inode = container_of(head, struct inode, i_rcu);
+	kmem_cache_free(exofs_inode_cachep, exofs_i(inode));
+}
+
 /*
  * Remove an inode from the cache
  */
 static void exofs_destroy_inode(struct inode *inode)
 {
-	kmem_cache_free(exofs_inode_cachep, exofs_i(inode));
+	call_rcu(&inode->i_rcu, exofs_i_callback);
 }
 
 /*
@@ -187,10 +206,152 @@ static int init_inodecache(void)
  */
 static void destroy_inodecache(void)
 {
+	/*
+	 * Make sure all delayed rcu free inodes are flushed before we
+	 * destroy cache.
+	 */
+	rcu_barrier();
 	kmem_cache_destroy(exofs_inode_cachep);
 }
 
 /******************************************************************************
+ * Some osd helpers
+ *****************************************************************************/
+void exofs_make_credential(u8 cred_a[OSD_CAP_LEN], const struct osd_obj_id *obj)
+{
+	osd_sec_init_nosec_doall_caps(cred_a, obj, false, true);
+}
+
+static int exofs_read_kern(struct osd_dev *od, u8 *cred, struct osd_obj_id *obj,
+		    u64 offset, void *p, unsigned length)
+{
+	struct osd_request *or = osd_start_request(od, GFP_KERNEL);
+/*	struct osd_sense_info osi = {.key = 0};*/
+	int ret;
+
+	if (unlikely(!or)) {
+		EXOFS_DBGMSG("%s: osd_start_request failed.\n", __func__);
+		return -ENOMEM;
+	}
+	ret = osd_req_read_kern(or, obj, offset, p, length);
+	if (unlikely(ret)) {
+		EXOFS_DBGMSG("%s: osd_req_read_kern failed.\n", __func__);
+		goto out;
+	}
+
+	ret = osd_finalize_request(or, 0, cred, NULL);
+	if (unlikely(ret)) {
+		EXOFS_DBGMSG("Failed to osd_finalize_request() => %d\n", ret);
+		goto out;
+	}
+
+	ret = osd_execute_request(or);
+	if (unlikely(ret))
+		EXOFS_DBGMSG("osd_execute_request() => %d\n", ret);
+	/* osd_req_decode_sense(or, ret); */
+
+out:
+	osd_end_request(or);
+	EXOFS_DBGMSG2("read_kern(0x%llx) offset=0x%llx "
+		      "length=0x%llx dev=%p ret=>%d\n",
+		      _LLU(obj->id), _LLU(offset), _LLU(length), od, ret);
+	return ret;
+}
+
+static const struct osd_attr g_attr_sb_stats = ATTR_DEF(
+	EXOFS_APAGE_SB_DATA,
+	EXOFS_ATTR_SB_STATS,
+	sizeof(struct exofs_sb_stats));
+
+static int __sbi_read_stats(struct exofs_sb_info *sbi)
+{
+	struct osd_attr attrs[] = {
+		[0] = g_attr_sb_stats,
+	};
+	struct ore_io_state *ios;
+	int ret;
+
+	ret = ore_get_io_state(&sbi->layout, &sbi->oc, &ios);
+	if (unlikely(ret)) {
+		EXOFS_ERR("%s: ore_get_io_state failed.\n", __func__);
+		return ret;
+	}
+
+	ios->in_attr = attrs;
+	ios->in_attr_len = ARRAY_SIZE(attrs);
+
+	ret = ore_read(ios);
+	if (unlikely(ret)) {
+		EXOFS_ERR("Error reading super_block stats => %d\n", ret);
+		goto out;
+	}
+
+	ret = extract_attr_from_ios(ios, &attrs[0]);
+	if (ret) {
+		EXOFS_ERR("%s: extract_attr of sb_stats failed\n", __func__);
+		goto out;
+	}
+	if (attrs[0].len) {
+		struct exofs_sb_stats *ess;
+
+		if (unlikely(attrs[0].len != sizeof(*ess))) {
+			EXOFS_ERR("%s: Wrong version of exofs_sb_stats "
+				  "size(%d) != expected(%zd)\n",
+				  __func__, attrs[0].len, sizeof(*ess));
+			goto out;
+		}
+
+		ess = attrs[0].val_ptr;
+		sbi->s_nextid = le64_to_cpu(ess->s_nextid);
+		sbi->s_numfiles = le32_to_cpu(ess->s_numfiles);
+	}
+
+out:
+	ore_put_io_state(ios);
+	return ret;
+}
+
+static void stats_done(struct ore_io_state *ios, void *p)
+{
+	ore_put_io_state(ios);
+	/* Good thanks nothing to do anymore */
+}
+
+/* Asynchronously write the stats attribute */
+int exofs_sbi_write_stats(struct exofs_sb_info *sbi)
+{
+	struct osd_attr attrs[] = {
+		[0] = g_attr_sb_stats,
+	};
+	struct ore_io_state *ios;
+	int ret;
+
+	ret = ore_get_io_state(&sbi->layout, &sbi->oc, &ios);
+	if (unlikely(ret)) {
+		EXOFS_ERR("%s: ore_get_io_state failed.\n", __func__);
+		return ret;
+	}
+
+	sbi->s_ess.s_nextid   = cpu_to_le64(sbi->s_nextid);
+	sbi->s_ess.s_numfiles = cpu_to_le64(sbi->s_numfiles);
+	attrs[0].val_ptr = &sbi->s_ess;
+
+
+	ios->done = stats_done;
+	ios->private = sbi;
+	ios->out_attr = attrs;
+	ios->out_attr_len = ARRAY_SIZE(attrs);
+
+	ret = ore_write(ios);
+	if (unlikely(ret)) {
+		EXOFS_ERR("%s: ore_write failed.\n", __func__);
+		ore_put_io_state(ios);
+	}
+
+	return ret;
+}
+
+/******************************************************************************
  * SUPERBLOCK FUNCTIONS
  *****************************************************************************/
 static const struct super_operations exofs_sops;
@@ -199,60 +360,57 @@ static const struct export_operations exofs_export_ops;
 /*
  * Write the superblock to the OSD
  */
-int exofs_sync_fs(struct super_block *sb, int wait)
+static int exofs_sync_fs(struct super_block *sb, int wait)
 {
 	struct exofs_sb_info *sbi;
 	struct exofs_fscb *fscb;
-	struct exofs_io_state *ios;
+	struct ore_comp one_comp;
+	struct ore_components oc;
+	struct ore_io_state *ios;
 	int ret = -ENOMEM;
 
-	lock_super(sb);
+	fscb = kmalloc(sizeof(*fscb), GFP_KERNEL);
+	if (unlikely(!fscb))
+		return -ENOMEM;
+
 	sbi = sb->s_fs_info;
-	fscb = &sbi->s_fscb;
 
-	ret = exofs_get_io_state(&sbi->layout, &ios);
-	if (ret)
+	/* NOTE: We no longer dirty the super_block anywhere in exofs. The
+	 * reason we write the fscb here on unmount is so we can stay backwards
+	 * compatible with fscb->s_version == 1. (What we are not compatible
+	 * with is if a new version FS crashed and then we try to mount an old
+	 * version). Otherwise the exofs_fscb is read-only from mkfs time. All
+	 * the writeable info is set in exofs_sbi_write_stats() above.
+	 */
+
+	exofs_init_comps(&oc, &one_comp, sbi, EXOFS_SUPER_ID);
+
+	ret = ore_get_io_state(&sbi->layout, &oc, &ios);
+	if (unlikely(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_numfiles = cpu_to_le64(sbi->s_numfiles);
 	fscb->s_magic = cpu_to_le16(sb->s_magic);
 	fscb->s_newfs = 0;
 	fscb->s_version = EXOFS_FSCB_VER;
 
-	ios->obj.id = EXOFS_SUPER_ID;
 	ios->offset = 0;
 	ios->kern_buff = fscb;
-	ios->cred = sbi->s_cred;
 
-	ret = exofs_sbi_write(ios);
-	if (unlikely(ret)) {
-		EXOFS_ERR("%s: exofs_sbi_write failed.\n", __func__);
-		goto out;
-	}
-	sb->s_dirt = 0;
+	ret = ore_write(ios);
+	if (unlikely(ret))
+		EXOFS_ERR("%s: ore_write failed.\n", __func__);
 
 out:
 	EXOFS_DBGMSG("s_nextid=0x%llx ret=%d\n", _LLU(sbi->s_nextid), ret);
-	exofs_put_io_state(ios);
-	unlock_super(sb);
+	ore_put_io_state(ios);
+	kfree(fscb);
 	return ret;
 }
 
-static void exofs_write_super(struct super_block *sb)
-{
-	if (!(sb->s_flags & MS_RDONLY))
-		exofs_sync_fs(sb, 1);
-	else
-		sb->s_dirt = 0;
-}
-
 static void _exofs_print_device(const char *msg, const char *dev_path,
 				struct osd_dev *od, u64 pid)
 {
@@ -262,17 +420,20 @@ static void _exofs_print_device(const char *msg, const char *dev_path,
 		msg, dev_path ?: "", odi->osdname, _LLU(pid));
 }
 
-void exofs_free_sbi(struct exofs_sb_info *sbi)
+static void exofs_free_sbi(struct exofs_sb_info *sbi)
 {
-	while (sbi->layout.s_numdevs) {
-		int i = --sbi->layout.s_numdevs;
-		struct osd_dev *od = sbi->layout.s_ods[i];
+	unsigned numdevs = sbi->oc.numdevs;
+
+	while (numdevs) {
+		unsigned i = --numdevs;
+		struct osd_dev *od = ore_comp_dev(&sbi->oc, i);
 
 		if (od) {
-			sbi->layout.s_ods[i] = NULL;
+			ore_comp_set_dev(&sbi->oc, i, NULL);
 			osduld_put_device(od);
 		}
 	}
+	kfree(sbi->oc.ods);
 	kfree(sbi);
 }
 
@@ -285,22 +446,24 @@ static void exofs_put_super(struct super_block *sb)
 	int num_pend;
 	struct exofs_sb_info *sbi = sb->s_fs_info;
 
-	if (sb->s_dirt)
-		exofs_write_super(sb);
-
 	/* make sure there are no pending commands */
 	for (num_pend = atomic_read(&sbi->s_curr_pending); num_pend > 0;
 	     num_pend = atomic_read(&sbi->s_curr_pending)) {
 		wait_queue_head_t wq;
+
+		printk(KERN_NOTICE "%s: !!Pending operations in flight. "
+		       "This is a BUG. please report to osd-dev@open-osd.org\n",
+		       __func__);
 		init_waitqueue_head(&wq);
 		wait_event_timeout(wq,
 				  (atomic_read(&sbi->s_curr_pending) == 0),
 				  msecs_to_jiffies(100));
 	}
 
-	_exofs_print_device("Unmounting", NULL, sbi->layout.s_ods[0],
-			    sbi->layout.s_pid);
+	_exofs_print_device("Unmounting", NULL, ore_comp_dev(&sbi->oc, 0),
+			    sbi->one_comp.obj.partition);
 
+	exofs_sysfs_sb_del(sbi);
 	bdi_destroy(&sbi->bdi);
 	exofs_free_sbi(sbi);
 	sb->s_fs_info = NULL;
@@ -309,78 +472,48 @@ static void exofs_put_super(struct super_block *sb)
 static int _read_and_match_data_map(struct exofs_sb_info *sbi, unsigned numdevs,
 				    struct exofs_device_table *dt)
 {
-	u64 stripe_length;
+	int ret;
 
-	sbi->data_map.odm_num_comps   =
-				le32_to_cpu(dt->dt_data_map.cb_num_comps);
-	sbi->data_map.odm_stripe_unit =
+	sbi->layout.stripe_unit =
 				le64_to_cpu(dt->dt_data_map.cb_stripe_unit);
-	sbi->data_map.odm_group_width =
+	sbi->layout.group_width =
 				le32_to_cpu(dt->dt_data_map.cb_group_width);
-	sbi->data_map.odm_group_depth =
+	sbi->layout.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  =
+	sbi->layout.mirrors_p1  =
+				le32_to_cpu(dt->dt_data_map.cb_mirror_cnt) + 1;
+	sbi->layout.raid_algorithm  =
 				le32_to_cpu(dt->dt_data_map.cb_raid_algorithm);
 
-/* FIXME: Only raid0 for now. if not so, do not mount */
-	if (sbi->data_map.odm_num_comps != numdevs) {
-		EXOFS_ERR("odm_num_comps(%u) != numdevs(%u)\n",
-			  sbi->data_map.odm_num_comps, numdevs);
-		return -EINVAL;
-	}
-	if (sbi->data_map.odm_raid_algorithm != PNFS_OSD_RAID_0) {
-		EXOFS_ERR("Only RAID_0 for now\n");
-		return -EINVAL;
-	}
-	if (0 != (numdevs % (sbi->data_map.odm_mirror_cnt + 1))) {
-		EXOFS_ERR("Data Map wrong, numdevs=%d mirrors=%d\n",
-			  numdevs, sbi->data_map.odm_mirror_cnt);
-		return -EINVAL;
-	}
+	ret = ore_verify_layout(numdevs, &sbi->layout);
+
+	EXOFS_DBGMSG("exofs: layout: "
+		"num_comps=%u stripe_unit=0x%x group_width=%u "
+		"group_depth=0x%llx mirrors_p1=%u raid_algorithm=%u\n",
+		numdevs,
+		sbi->layout.stripe_unit,
+		sbi->layout.group_width,
+		_LLU(sbi->layout.group_depth),
+		sbi->layout.mirrors_p1,
+		sbi->layout.raid_algorithm);
+	return ret;
+}
 
-	if (0 != (sbi->data_map.odm_stripe_unit & ~PAGE_MASK)) {
-		EXOFS_ERR("Stripe Unit(0x%llx)"
-			  " must be Multples of PAGE_SIZE(0x%lx)\n",
-			  _LLU(sbi->data_map.odm_stripe_unit), PAGE_SIZE);
-		return -EINVAL;
-	}
+static unsigned __ra_pages(struct ore_layout *layout)
+{
+	const unsigned _MIN_RA = 32; /* min 128K read-ahead */
+	unsigned ra_pages = layout->group_width * layout->stripe_unit /
+				PAGE_SIZE;
+	unsigned max_io_pages = exofs_max_io_pages(layout, ~0);
 
-	sbi->layout.stripe_unit = sbi->data_map.odm_stripe_unit;
-	sbi->layout.mirrors_p1 = sbi->data_map.odm_mirror_cnt + 1;
+	ra_pages *= 2; /* two stripes */
+	if (ra_pages < _MIN_RA)
+		ra_pages = roundup(_MIN_RA, ra_pages / 2);
 
-	if (sbi->data_map.odm_group_width) {
-		sbi->layout.group_width = sbi->data_map.odm_group_width;
-		sbi->layout.group_depth = sbi->data_map.odm_group_depth;
-		if (!sbi->layout.group_depth) {
-			EXOFS_ERR("group_depth == 0 && group_width != 0\n");
-			return -EINVAL;
-		}
-		sbi->layout.group_count = sbi->data_map.odm_num_comps /
-						sbi->layout.mirrors_p1 /
-						sbi->data_map.odm_group_width;
-	} else {
-		if (sbi->data_map.odm_group_depth) {
-			printk(KERN_NOTICE "Warning: group_depth ignored "
-				"group_width == 0 && group_depth == %d\n",
-				sbi->data_map.odm_group_depth);
-			sbi->data_map.odm_group_depth = 0;
-		}
-		sbi->layout.group_width = sbi->data_map.odm_num_comps /
-							sbi->layout.mirrors_p1;
-		sbi->layout.group_depth = -1;
-		sbi->layout.group_count = 1;
-	}
-
-	stripe_length = (u64)sbi->layout.group_width * sbi->layout.stripe_unit;
-	if (stripe_length >= (1ULL << 32)) {
-		EXOFS_ERR("Total Stripe length(0x%llx)"
-			  " >= 32bit is not supported\n", _LLU(stripe_length));
-		return -EINVAL;
-	}
+	if (ra_pages > max_io_pages)
+		ra_pages = max_io_pages;
 
-	return 0;
+	return ra_pages;
 }
 
 /* @odi is valid only as long as @fscb_dev is valid */
@@ -388,7 +521,8 @@ 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);
+	if (likely(odi->systemid_len))
+		memcpy(odi->systemid, dt_dev->systemid, OSD_SYSTEMID_LEN);
 
 	odi->osdname_len = le32_to_cpu(dt_dev->osdname_len);
 	odi->osdname = dt_dev->osdname;
@@ -409,14 +543,40 @@ static int exofs_devs_2_odi(struct exofs_dt_device_info *dt_dev,
 	return !(odi->systemid_len || odi->osdname_len);
 }
 
-static int exofs_read_lookup_dev_table(struct exofs_sb_info **psbi,
+static int __alloc_dev_table(struct exofs_sb_info *sbi, unsigned numdevs,
+		      struct exofs_dev **peds)
+{
+	struct __alloc_ore_devs_and_exofs_devs {
+		/* Twice bigger table: See exofs_init_comps() and comment at
+		 * exofs_read_lookup_dev_table()
+		 */
+		struct ore_dev *oreds[numdevs * 2 - 1];
+		struct exofs_dev eds[numdevs];
+	} *aoded;
+	struct exofs_dev *eds;
+	unsigned i;
+
+	aoded = kzalloc(sizeof(*aoded), GFP_KERNEL);
+	if (unlikely(!aoded)) {
+		EXOFS_ERR("ERROR: failed allocating Device array[%d]\n",
+			  numdevs);
+		return -ENOMEM;
+	}
+
+	sbi->oc.ods = aoded->oreds;
+	*peds = eds = aoded->eds;
+	for (i = 0; i < numdevs; ++i)
+		aoded->oreds[i] = &eds[i].ored;
+	return 0;
+}
+
+static int exofs_read_lookup_dev_table(struct exofs_sb_info *sbi,
+				       struct osd_dev *fscb_od,
 				       unsigned table_count)
 {
-	struct exofs_sb_info *sbi = *psbi;
-	struct osd_dev *fscb_od;
-	struct osd_obj_id obj = {.partition = sbi->layout.s_pid,
-				 .id = EXOFS_DEVTABLE_ID};
+	struct ore_comp comp;
 	struct exofs_device_table *dt;
+	struct exofs_dev *eds;
 	unsigned table_bytes = table_count * sizeof(dt->dt_dev_table[0]) +
 					     sizeof(*dt);
 	unsigned numdevs, i;
@@ -429,10 +589,14 @@ static int exofs_read_lookup_dev_table(struct exofs_sb_info **psbi,
 		return -ENOMEM;
 	}
 
-	fscb_od = sbi->layout.s_ods[0];
-	sbi->layout.s_ods[0] = NULL;
-	sbi->layout.s_numdevs = 0;
-	ret = exofs_read_kern(fscb_od, sbi->s_cred, &obj, 0, dt, table_bytes);
+	sbi->oc.numdevs = 0;
+
+	comp.obj.partition = sbi->one_comp.obj.partition;
+	comp.obj.id = EXOFS_DEVTABLE_ID;
+	exofs_make_credential(comp.cred, &comp.obj);
+
+	ret = exofs_read_kern(fscb_od, comp.cred, &comp.obj, 0, dt,
+			      table_bytes);
 	if (unlikely(ret)) {
 		EXOFS_ERR("ERROR: reading device table\n");
 		goto out;
@@ -449,18 +613,22 @@ static int exofs_read_lookup_dev_table(struct exofs_sb_info **psbi,
 	if (unlikely(ret))
 		goto out;
 
-	if (likely(numdevs > 1)) {
-		unsigned size = numdevs * sizeof(sbi->layout.s_ods[0]);
+	ret = __alloc_dev_table(sbi, numdevs, &eds);
+	if (unlikely(ret))
+		goto out;
+	/* exofs round-robins the device table view according to inode
+	 * number. We hold a: twice bigger table hence inodes can point
+	 * to any device and have a sequential view of the table
+	 * starting at this device. See exofs_init_comps()
+	 */
+	memcpy(&sbi->oc.ods[numdevs], &sbi->oc.ods[0],
+		(numdevs - 1) * sizeof(sbi->oc.ods[0]));
 
-		sbi = krealloc(sbi, sizeof(*sbi) + size, GFP_KERNEL);
-		if (unlikely(!sbi)) {
-			ret = -ENOMEM;
-			goto out;
-		}
-		memset(&sbi->layout.s_ods[1], 0,
-		       size - sizeof(sbi->layout.s_ods[0]));
-		*psbi = sbi;
-	}
+	/* create sysfs subdir under which we put the device table
+	 * And cluster layout. A Superblock is identified by the string:
+	 *	"dev[0].osdname"_"pid"
+	 */
+	exofs_sysfs_sb_add(sbi, &dt->dt_dev_table[0]);
 
 	for (i = 0; i < numdevs; i++) {
 		struct exofs_fscb fscb;
@@ -476,32 +644,36 @@ static int exofs_read_lookup_dev_table(struct exofs_sb_info **psbi,
 		printk(KERN_NOTICE "Add device[%d]: osd_name-%s\n",
 		       i, odi.osdname);
 
+		/* the exofs id is currently the table index */
+		eds[i].did = i;
+
 		/* 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->layout.s_ods[i] = fscb_od;
-			++sbi->layout.s_numdevs;
+			eds[i].ored.od = fscb_od;
+			++sbi->oc.numdevs;
 			fscb_od = NULL;
+			exofs_sysfs_odev_add(&eds[i], sbi);
 			continue;
 		}
 
 		od = osduld_info_lookup(&odi);
-		if (unlikely(IS_ERR(od))) {
+		if (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->layout.s_ods[i] = od;
-		++sbi->layout.s_numdevs;
+		eds[i].ored.od = od;
+		++sbi->oc.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,
+		ret = exofs_read_kern(od, comp.cred, &comp.obj, 0, &fscb,
 				      sizeof(fscb));
 		if (unlikely(ret)) {
 			EXOFS_ERR("ERROR: Malformed participating device "
@@ -509,6 +681,7 @@ static int exofs_read_lookup_dev_table(struct exofs_sb_info **psbi,
 				  odi.osdname);
 			goto out;
 		}
+		exofs_sysfs_odev_add(&eds[i], sbi);
 
 		/* TODO: verify other information is correct and FS-uuid
 		 *	 matches. Benny what did you say about device table
@@ -518,13 +691,11 @@ static int exofs_read_lookup_dev_table(struct exofs_sb_info **psbi,
 
 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;
+	if (unlikely(fscb_od && !ret)) {
+			EXOFS_ERR("ERROR: Bad device-table container device not present\n");
+			osduld_put_device(fscb_od);
+			return -EINVAL;
 	}
-
 	return ret;
 }
 
@@ -538,7 +709,7 @@ static int exofs_fill_super(struct super_block *sb, void *data, int silent)
 	struct exofs_sb_info *sbi;	/*extended info                  */
 	struct osd_dev *od;		/* Master device                 */
 	struct exofs_fscb fscb;		/*on-disk superblock info        */
-	struct osd_obj_id obj;
+	struct ore_comp comp;
 	unsigned table_count;
 	int ret;
 
@@ -546,14 +717,20 @@ static int exofs_fill_super(struct super_block *sb, void *data, int silent)
 	if (!sbi)
 		return -ENOMEM;
 
-	ret = bdi_setup_and_register(&sbi->bdi, "exofs", BDI_CAP_MAP_COPY);
-	if (ret)
-		goto free_bdi;
-
 	/* use mount options to fill superblock */
-	od = osduld_path_lookup(opts->dev_name);
+	if (opts->is_osdname) {
+		struct osd_dev_info odi = {.systemid_len = 0};
+
+		odi.osdname_len = strlen(opts->dev_name);
+		odi.osdname = (u8 *)opts->dev_name;
+		od = osduld_info_lookup(&odi);
+		kfree(opts->dev_name);
+		opts->dev_name = NULL;
+	} else {
+		od = osduld_path_lookup(opts->dev_name);
+	}
 	if (IS_ERR(od)) {
-		ret = PTR_ERR(od);
+		ret = -EINVAL;
 		goto free_sbi;
 	}
 
@@ -563,30 +740,35 @@ static int exofs_fill_super(struct super_block *sb, void *data, int silent)
 	sbi->layout.group_width = 1;
 	sbi->layout.group_depth = -1;
 	sbi->layout.group_count = 1;
-	sbi->layout.s_ods[0] = od;
-	sbi->layout.s_numdevs = 1;
-	sbi->layout.s_pid = opts->pid;
 	sbi->s_timeout = opts->timeout;
 
+	sbi->one_comp.obj.partition = opts->pid;
+	sbi->one_comp.obj.id = 0;
+	exofs_make_credential(sbi->one_comp.cred, &sbi->one_comp.obj);
+	sbi->oc.single_comp = EC_SINGLE_COMP;
+	sbi->oc.comps = &sbi->one_comp;
+
 	/* fill in some other data by hand */
 	memset(sb->s_id, 0, sizeof(sb->s_id));
 	strcpy(sb->s_id, "exofs");
 	sb->s_blocksize = EXOFS_BLKSIZE;
 	sb->s_blocksize_bits = EXOFS_BLKSHIFT;
 	sb->s_maxbytes = MAX_LFS_FILESIZE;
+	sb->s_max_links = EXOFS_LINK_MAX;
 	atomic_set(&sbi->s_curr_pending, 0);
 	sb->s_bdev = NULL;
 	sb->s_dev = 0;
 
-	obj.partition = sbi->layout.s_pid;
-	obj.id = EXOFS_SUPER_ID;
-	exofs_make_credential(sbi->s_cred, &obj);
+	comp.obj.partition = sbi->one_comp.obj.partition;
+	comp.obj.id = EXOFS_SUPER_ID;
+	exofs_make_credential(comp.cred, &comp.obj);
 
-	ret = exofs_read_kern(od, sbi->s_cred, &obj, 0, &fscb, sizeof(fscb));
+	ret = exofs_read_kern(od, comp.cred, &comp.obj, 0, &fscb, sizeof(fscb));
 	if (unlikely(ret))
 		goto free_sbi;
 
 	sb->s_magic = le16_to_cpu(fscb.s_magic);
+	/* NOTE: we read below to be backward compatible with old versions */
 	sbi->s_nextid = le64_to_cpu(fscb.s_nextid);
 	sbi->s_numfiles = le32_to_cpu(fscb.s_numfiles);
 
@@ -597,7 +779,7 @@ 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) {
+	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;
@@ -610,12 +792,24 @@ static int exofs_fill_super(struct super_block *sb, void *data, int silent)
 
 	table_count = le64_to_cpu(fscb.s_dev_table_count);
 	if (table_count) {
-		ret = exofs_read_lookup_dev_table(&sbi, table_count);
+		ret = exofs_read_lookup_dev_table(sbi, od, table_count);
 		if (unlikely(ret))
 			goto free_sbi;
+	} else {
+		struct exofs_dev *eds;
+
+		ret = __alloc_dev_table(sbi, 1, &eds);
+		if (unlikely(ret))
+			goto free_sbi;
+
+		ore_comp_set_dev(&sbi->oc, 0, od);
+		sbi->oc.numdevs = 1;
 	}
 
+	__sbi_read_stats(sbi);
+
 	/* set up operation vectors */
+	sbi->bdi.ra_pages = __ra_pages(&sbi->layout);
 	sb->s_bdi = &sbi->bdi;
 	sb->s_fs_info = sbi;
 	sb->s_op = &exofs_sops;
@@ -626,9 +820,8 @@ static int exofs_fill_super(struct super_block *sb, void *data, int silent)
 		ret = PTR_ERR(root);
 		goto free_sbi;
 	}
-	sb->s_root = d_alloc_root(root);
+	sb->s_root = d_make_root(root);
 	if (!sb->s_root) {
-		iput(root);
 		EXOFS_ERR("ERROR: get root inode failed\n");
 		ret = -ENOMEM;
 		goto free_sbi;
@@ -643,15 +836,23 @@ static int exofs_fill_super(struct super_block *sb, void *data, int silent)
 		goto free_sbi;
 	}
 
-	_exofs_print_device("Mounting", opts->dev_name, sbi->layout.s_ods[0],
-			    sbi->layout.s_pid);
+	ret = bdi_setup_and_register(&sbi->bdi, "exofs", BDI_CAP_MAP_COPY);
+	if (ret) {
+		EXOFS_DBGMSG("Failed to bdi_setup_and_register\n");
+		dput(sb->s_root);
+		sb->s_root = NULL;
+		goto free_sbi;
+	}
+
+	exofs_sysfs_dbg_print();
+	_exofs_print_device("Mounting", opts->dev_name,
+			    ore_comp_dev(&sbi->oc, 0),
+			    sbi->one_comp.obj.partition);
 	return 0;
 
 free_sbi:
-	bdi_destroy(&sbi->bdi);
-free_bdi:
 	EXOFS_ERR("Unable to mount exofs on %s pid=0x%llx err=%d\n",
-		  opts->dev_name, sbi->layout.s_pid, ret);
+		  opts->dev_name, sbi->one_comp.obj.partition, ret);
 	exofs_free_sbi(sbi);
 	return ret;
 }
@@ -670,7 +871,8 @@ static struct dentry *exofs_mount(struct file_system_type *type,
 	if (ret)
 		return ERR_PTR(ret);
 
-	opts.dev_name = dev_name;
+	if (!opts.dev_name)
+		opts.dev_name = dev_name;
 	return mount_nodev(type, flags, &opts, exofs_fill_super);
 }
 
@@ -682,7 +884,7 @@ 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 exofs_io_state *ios;
+	struct ore_io_state *ios;
 	struct osd_attr attrs[] = {
 		ATTR_DEF(OSD_APAGE_PARTITION_QUOTAS,
 			OSD_ATTR_PQ_CAPACITY_QUOTA, sizeof(__be64)),
@@ -691,21 +893,18 @@ static int exofs_statfs(struct dentry *dentry, struct kstatfs *buf)
 	};
 	uint64_t capacity = ULLONG_MAX;
 	uint64_t used = ULLONG_MAX;
-	uint8_t cred_a[OSD_CAP_LEN];
 	int ret;
 
-	ret = exofs_get_io_state(&sbi->layout, &ios);
+	ret = ore_get_io_state(&sbi->layout, &sbi->oc, &ios);
 	if (ret) {
-		EXOFS_DBGMSG("exofs_get_io_state failed.\n");
+		EXOFS_DBGMSG("ore_get_io_state failed.\n");
 		return ret;
 	}
 
-	exofs_make_credential(cred_a, &ios->obj);
-	ios->cred = sbi->s_cred;
 	ios->in_attr = attrs;
 	ios->in_attr_len = ARRAY_SIZE(attrs);
 
-	ret = exofs_sbi_read(ios);
+	ret = ore_read(ios);
 	if (unlikely(ret))
 		goto out;
 
@@ -734,7 +933,7 @@ static int exofs_statfs(struct dentry *dentry, struct kstatfs *buf)
 	buf->f_namelen = EXOFS_NAME_LEN;
 
 out:
-	exofs_put_io_state(ios);
+	ore_put_io_state(ios);
 	return ret;
 }
 
@@ -744,7 +943,6 @@ static const struct super_operations exofs_sops = {
 	.write_inode    = exofs_write_inode,
 	.evict_inode    = exofs_evict_inode,
 	.put_super      = exofs_put_super,
-	.write_super    = exofs_write_super,
 	.sync_fs	= exofs_sync_fs,
 	.statfs         = exofs_statfs,
 };
@@ -753,12 +951,12 @@ static const struct super_operations exofs_sops = {
  * EXPORT OPERATIONS
  *****************************************************************************/
 
-struct dentry *exofs_get_parent(struct dentry *child)
+static struct dentry *exofs_get_parent(struct dentry *child)
 {
 	unsigned long ino = exofs_parent_ino(child);
 
 	if (!ino)
-		return NULL;
+		return ERR_PTR(-ESTALE);
 
 	return d_obtain_alias(exofs_iget(child->d_inode->i_sb, ino));
 }
@@ -812,6 +1010,7 @@ static struct file_system_type exofs_type = {
 	.mount          = exofs_mount,
 	.kill_sb        = generic_shutdown_super,
 };
+MODULE_ALIAS_FS("exofs");
 
 static int __init init_exofs(void)
 {
@@ -825,6 +1024,9 @@ static int __init init_exofs(void)
 	if (err)
 		goto out_d;
 
+	/* We don't fail if sysfs creation failed */
+	exofs_sysfs_init();
+
 	return 0;
 out_d:
 	destroy_inodecache();
@@ -834,6 +1036,7 @@ out:
 
 static void __exit exit_exofs(void)
 {
+	exofs_sysfs_uninit();
 	unregister_filesystem(&exofs_type);
 	destroy_inodecache();
 }