1 files changed, 113 insertions, 138 deletions

diff --git a/fs/exofs/super.c b/fs/exofs/super.c
index 274894053b0..ed73ed8ebbe 100644
--- a/fs/exofs/super.c
+++ b/fs/exofs/super.c
@@ -35,6 +35,7 @@
 #include <linux/parser.h>
 #include <linux/vfs.h>
 #include <linux/random.h>
+#include <linux/module.h>
 #include <linux/exportfs.h>
 #include <linux/slab.h>
 
@@ -165,7 +166,6 @@ static struct inode *exofs_alloc_inode(struct super_block *sb)
 static void exofs_i_callback(struct rcu_head *head)
 {
 	struct inode *inode = container_of(head, struct inode, i_rcu);
-	INIT_LIST_HEAD(&inode->i_dentry);
 	kmem_cache_free(exofs_inode_cachep, exofs_i(inode));
 }
 
@@ -206,6 +206,11 @@ 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);
 }
 
@@ -266,7 +271,7 @@ static int __sbi_read_stats(struct exofs_sb_info *sbi)
 	struct ore_io_state *ios;
 	int ret;
 
-	ret = ore_get_io_state(&sbi->layout, &sbi->comps, &ios);
+	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;
@@ -321,7 +326,7 @@ int exofs_sbi_write_stats(struct exofs_sb_info *sbi)
 	struct ore_io_state *ios;
 	int ret;
 
-	ret = ore_get_io_state(&sbi->layout, &sbi->comps, &ios);
+	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;
@@ -355,12 +360,12 @@ 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 ore_comp one_comp;
-	struct ore_components comps;
+	struct ore_components oc;
 	struct ore_io_state *ios;
 	int ret = -ENOMEM;
 
@@ -378,18 +383,16 @@ int exofs_sync_fs(struct super_block *sb, int wait)
 	 * the writeable info is set in exofs_sbi_write_stats() above.
 	 */
 
-	exofs_init_comps(&comps, &one_comp, sbi, EXOFS_SUPER_ID);
+	exofs_init_comps(&oc, &one_comp, sbi, EXOFS_SUPER_ID);
 
-	ret = ore_get_io_state(&sbi->layout, &comps, &ios);
+	ret = ore_get_io_state(&sbi->layout, &oc, &ios);
 	if (unlikely(ret))
 		goto out;
 
-	lock_super(sb);
-
 	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;
@@ -400,11 +403,7 @@ int exofs_sync_fs(struct super_block *sb, int wait)
 	ret = ore_write(ios);
 	if (unlikely(ret))
 		EXOFS_ERR("%s: ore_write failed.\n", __func__);
-	else
-		sb->s_dirt = 0;
-
 
-	unlock_super(sb);
 out:
 	EXOFS_DBGMSG("s_nextid=0x%llx ret=%d\n", _LLU(sbi->s_nextid), ret);
 	ore_put_io_state(ios);
@@ -412,14 +411,6 @@ out:
 	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)
 {
@@ -429,19 +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->comps.numdevs) {
-		int i = --sbi->comps.numdevs;
-		struct osd_dev *od = sbi->comps.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->comps.ods[i] = NULL;
+			ore_comp_set_dev(&sbi->oc, i, NULL);
 			osduld_put_device(od);
 		}
 	}
-	if (sbi->comps.ods != sbi->_min_one_dev)
-		kfree(sbi->comps.ods);
+	kfree(sbi->oc.ods);
 	kfree(sbi);
 }
 
@@ -468,9 +460,10 @@ static void exofs_put_super(struct super_block *sb)
 				  msecs_to_jiffies(100));
 	}
 
-	_exofs_print_device("Unmounting", NULL, sbi->comps.ods[0],
+	_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;
@@ -479,76 +472,20 @@ 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;
-	}
-
-	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;
-	}
-
-	sbi->layout.stripe_unit = sbi->data_map.odm_stripe_unit;
-	sbi->layout.mirrors_p1 = sbi->data_map.odm_mirror_cnt + 1;
-
-	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;
-	}
+	ret = ore_verify_layout(numdevs, &sbi->layout);
 
 	EXOFS_DBGMSG("exofs: layout: "
 		"num_comps=%u stripe_unit=0x%x group_width=%u "
@@ -558,8 +495,8 @@ static int _read_and_match_data_map(struct exofs_sb_info *sbi, unsigned numdevs,
 		sbi->layout.group_width,
 		_LLU(sbi->layout.group_depth),
 		sbi->layout.mirrors_p1,
-		sbi->data_map.odm_raid_algorithm);
-	return 0;
+		sbi->layout.raid_algorithm);
+	return ret;
 }
 
 static unsigned __ra_pages(struct ore_layout *layout)
@@ -584,7 +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;
@@ -605,12 +543,40 @@ static int exofs_devs_2_odi(struct exofs_dt_device_info *dt_dev,
 	return !(odi->systemid_len || odi->osdname_len);
 }
 
+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 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;
@@ -623,7 +589,7 @@ static int exofs_read_lookup_dev_table(struct exofs_sb_info *sbi,
 		return -ENOMEM;
 	}
 
-	sbi->comps.numdevs = 0;
+	sbi->oc.numdevs = 0;
 
 	comp.obj.partition = sbi->one_comp.obj.partition;
 	comp.obj.id = EXOFS_DEVTABLE_ID;
@@ -647,20 +613,22 @@ static int exofs_read_lookup_dev_table(struct exofs_sb_info *sbi,
 	if (unlikely(ret))
 		goto out;
 
-	if (likely(numdevs > 1)) {
-		unsigned size = numdevs * sizeof(sbi->comps.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]));
 
-		/* Twice bigger table: See exofs_init_comps() and below
-		 * comment
-		 */
-		sbi->comps.ods = kzalloc(size + size - 1, GFP_KERNEL);
-		if (unlikely(!sbi->comps.ods)) {
-			EXOFS_ERR("ERROR: faild allocating Device array[%d]\n",
-				  numdevs);
-			ret = -ENOMEM;
-			goto out;
-		}
-	}
+	/* 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;
@@ -676,14 +644,18 @@ static int exofs_read_lookup_dev_table(struct exofs_sb_info *sbi,
 		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->comps.ods[i] = fscb_od;
-			++sbi->comps.numdevs;
+			eds[i].ored.od = fscb_od;
+			++sbi->oc.numdevs;
 			fscb_od = NULL;
+			exofs_sysfs_odev_add(&eds[i], sbi);
 			continue;
 		}
 
@@ -695,8 +667,8 @@ static int exofs_read_lookup_dev_table(struct exofs_sb_info *sbi,
 			goto out;
 		}
 
-		sbi->comps.ods[i] = od;
-		++sbi->comps.numdevs;
+		eds[i].ored.od = od;
+		++sbi->oc.numdevs;
 
 		/* Read the fscb of the other devices to make sure the FS
 		 * partition is there.
@@ -709,6 +681,7 @@ static int exofs_read_lookup_dev_table(struct exofs_sb_info *sbi,
 				  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
@@ -718,21 +691,10 @@ static int exofs_read_lookup_dev_table(struct exofs_sb_info *sbi,
 
 out:
 	kfree(dt);
-	if (likely(!ret)) {
-		unsigned numdevs = sbi->comps.numdevs;
-
-		if (unlikely(fscb_od)) {
+	if (unlikely(fscb_od && !ret)) {
 			EXOFS_ERR("ERROR: Bad device-table container device not present\n");
 			osduld_put_device(fscb_od);
 			return -EINVAL;
-		}
-		/* 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()
-		 */
-		for (i = 0; i < numdevs - 1; ++i)
-			sbi->comps.ods[i + numdevs] = sbi->comps.ods[i];
 	}
 	return ret;
 }
@@ -783,10 +745,8 @@ static int exofs_fill_super(struct super_block *sb, void *data, int silent)
 	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->comps.numdevs = 1;
-	sbi->comps.single_comp = EC_SINGLE_COMP;
-	sbi->comps.comps = &sbi->one_comp;
-	sbi->comps.ods = sbi->_min_one_dev;
+	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));
@@ -794,6 +754,7 @@ static int exofs_fill_super(struct super_block *sb, void *data, int silent)
 	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;
@@ -835,7 +796,14 @@ static int exofs_fill_super(struct super_block *sb, void *data, int silent)
 		if (unlikely(ret))
 			goto free_sbi;
 	} else {
-		sbi->comps.ods[0] = od;
+		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);
@@ -852,9 +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;
@@ -872,10 +839,14 @@ static int exofs_fill_super(struct super_block *sb, void *data, int silent)
 	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_print_device("Mounting", opts->dev_name, sbi->comps.ods[0],
+	exofs_sysfs_dbg_print();
+	_exofs_print_device("Mounting", opts->dev_name,
+			    ore_comp_dev(&sbi->oc, 0),
 			    sbi->one_comp.obj.partition);
 	return 0;
 
@@ -924,7 +895,7 @@ static int exofs_statfs(struct dentry *dentry, struct kstatfs *buf)
 	uint64_t used = ULLONG_MAX;
 	int ret;
 
-	ret = ore_get_io_state(&sbi->layout, &sbi->comps, &ios);
+	ret = ore_get_io_state(&sbi->layout, &sbi->oc, &ios);
 	if (ret) {
 		EXOFS_DBGMSG("ore_get_io_state failed.\n");
 		return ret;
@@ -972,7 +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,
 };
@@ -981,7 +951,7 @@ 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);
 
@@ -1040,6 +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)
 {
@@ -1053,6 +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();
@@ -1062,6 +1036,7 @@ out:
 
 static void __exit exit_exofs(void)
 {
+	exofs_sysfs_uninit();
 	unregister_filesystem(&exofs_type);
 	destroy_inodecache();
 }