quota: Split off quota tree handling into a separate file

There is going to be a new version of quota format having 64-bit
quota limits and a new quota format for OCFS2. They are both
going to use the same tree structure as VFSv0 quota format. So
split out tree handling into a separate file and make size of
leaf blocks, amount of space usable in each block (needed for
checksumming) and structures contained in them configurable
so that the code can be shared.

Signed-off-by: Jan Kara <jack@suse.cz>
Signed-off-by: Mark Fasheh <mfasheh@suse.com>

fs/Kconfig

@@ -302,6 +302,10 @@ config PRINT_QUOTA_WARNING
 	  Note that this behavior is currently deprecated and may go away in
 	  future. Please use notification via netlink socket instead.
 
+# Generic support for tree structured quota files. Seleted when needed.
+config QUOTA_TREE
+	 tristate
+
 config QFMT_V1
 	tristate "Old quota format support"
 	depends on QUOTA
@@ -313,6 +317,7 @@ config QFMT_V1
 config QFMT_V2
 	tristate "Quota format v2 support"
 	depends on QUOTA
+	select QUOTA_TREE
 	help
 	  This quota format allows using quotas with 32-bit UIDs/GIDs. If you
 	  need this functionality say Y here.

fs/Makefile

@@ -54,6 +54,7 @@ obj-$(CONFIG_GENERIC_ACL)	+= generic_acl.o
 obj-$(CONFIG_QUOTA)		+= dquot.o
 obj-$(CONFIG_QFMT_V1)		+= quota_v1.o
 obj-$(CONFIG_QFMT_V2)		+= quota_v2.o
+obj-$(CONFIG_QUOTA_TREE)	+= quota_tree.o
 obj-$(CONFIG_QUOTACTL)		+= quota.o
 
 obj-$(CONFIG_PROC_FS)		+= proc/

fs/quota_tree.c

@@ -0,0 +1,645 @@
+/*
+ *	vfsv0 quota IO operations on file
+ */
+
+#include <linux/errno.h>
+#include <linux/fs.h>
+#include <linux/mount.h>
+#include <linux/dqblk_v2.h>
+#include <linux/kernel.h>
+#include <linux/init.h>
+#include <linux/module.h>
+#include <linux/slab.h>
+#include <linux/quotaops.h>
+
+#include <asm/byteorder.h>
+
+#include "quota_tree.h"
+
+MODULE_AUTHOR("Jan Kara");
+MODULE_DESCRIPTION("Quota trie support");
+MODULE_LICENSE("GPL");
+
+#define __QUOTA_QT_PARANOIA
+
+typedef char *dqbuf_t;
+
+static int get_index(struct qtree_mem_dqinfo *info, qid_t id, int depth)
+{
+	unsigned int epb = info->dqi_usable_bs >> 2;
+
+	depth = info->dqi_qtree_depth - depth - 1;
+	while (depth--)
+		id /= epb;
+	return id % epb;
+}
+
+/* Number of entries in one blocks */
+static inline int qtree_dqstr_in_blk(struct qtree_mem_dqinfo *info)
+{
+	return (info->dqi_usable_bs - sizeof(struct qt_disk_dqdbheader))
+	       / info->dqi_entry_size;
+}
+
+static dqbuf_t getdqbuf(size_t size)
+{
+	dqbuf_t buf = kmalloc(size, GFP_NOFS);
+	if (!buf)
+		printk(KERN_WARNING "VFS: Not enough memory for quota buffers.\n");
+	return buf;
+}
+
+static inline void freedqbuf(dqbuf_t buf)
+{
+	kfree(buf);
+}
+
+static inline ssize_t read_blk(struct qtree_mem_dqinfo *info, uint blk, dqbuf_t buf)
+{
+	struct super_block *sb = info->dqi_sb;
+
+	memset(buf, 0, info->dqi_usable_bs);
+	return sb->s_op->quota_read(sb, info->dqi_type, (char *)buf,
+	       info->dqi_usable_bs, blk << info->dqi_blocksize_bits);
+}
+
+static inline ssize_t write_blk(struct qtree_mem_dqinfo *info, uint blk, dqbuf_t buf)
+{
+	struct super_block *sb = info->dqi_sb;
+
+	return sb->s_op->quota_write(sb, info->dqi_type, (char *)buf,
+	       info->dqi_usable_bs, blk << info->dqi_blocksize_bits);
+}
+
+/* Remove empty block from list and return it */
+static int get_free_dqblk(struct qtree_mem_dqinfo *info)
+{
+	dqbuf_t buf = getdqbuf(info->dqi_usable_bs);
+	struct qt_disk_dqdbheader *dh = (struct qt_disk_dqdbheader *)buf;
+	int ret, blk;
+
+	if (!buf)
+		return -ENOMEM;
+	if (info->dqi_free_blk) {
+		blk = info->dqi_free_blk;
+		ret = read_blk(info, blk, buf);
+		if (ret < 0)
+			goto out_buf;
+		info->dqi_free_blk = le32_to_cpu(dh->dqdh_next_free);
+	}
+	else {
+		memset(buf, 0, info->dqi_usable_bs);
+		/* Assure block allocation... */
+		ret = write_blk(info, info->dqi_blocks, buf);
+		if (ret < 0)
+			goto out_buf;
+		blk = info->dqi_blocks++;
+	}
+	mark_info_dirty(info->dqi_sb, info->dqi_type);
+	ret = blk;
+out_buf:
+	freedqbuf(buf);
+	return ret;
+}
+
+/* Insert empty block to the list */
+static int put_free_dqblk(struct qtree_mem_dqinfo *info, dqbuf_t buf, uint blk)
+{
+	struct qt_disk_dqdbheader *dh = (struct qt_disk_dqdbheader *)buf;
+	int err;
+
+	dh->dqdh_next_free = cpu_to_le32(info->dqi_free_blk);
+	dh->dqdh_prev_free = cpu_to_le32(0);
+	dh->dqdh_entries = cpu_to_le16(0);
+	err = write_blk(info, blk, buf);
+	if (err < 0)
+		return err;
+	info->dqi_free_blk = blk;
+	mark_info_dirty(info->dqi_sb, info->dqi_type);
+	return 0;
+}
+
+/* Remove given block from the list of blocks with free entries */
+static int remove_free_dqentry(struct qtree_mem_dqinfo *info, dqbuf_t buf, uint blk)
+{
+	dqbuf_t tmpbuf = getdqbuf(info->dqi_usable_bs);
+	struct qt_disk_dqdbheader *dh = (struct qt_disk_dqdbheader *)buf;
+	uint nextblk = le32_to_cpu(dh->dqdh_next_free);
+	uint prevblk = le32_to_cpu(dh->dqdh_prev_free);
+	int err;
+
+	if (!tmpbuf)
+		return -ENOMEM;
+	if (nextblk) {
+		err = read_blk(info, nextblk, tmpbuf);
+		if (err < 0)
+			goto out_buf;
+		((struct qt_disk_dqdbheader *)tmpbuf)->dqdh_prev_free =
+							dh->dqdh_prev_free;
+		err = write_blk(info, nextblk, tmpbuf);
+		if (err < 0)
+			goto out_buf;
+	}
+	if (prevblk) {
+		err = read_blk(info, prevblk, tmpbuf);
+		if (err < 0)
+			goto out_buf;
+		((struct qt_disk_dqdbheader *)tmpbuf)->dqdh_next_free =
+							dh->dqdh_next_free;
+		err = write_blk(info, prevblk, tmpbuf);
+		if (err < 0)
+			goto out_buf;
+	} else {
+		info->dqi_free_entry = nextblk;
+		mark_info_dirty(info->dqi_sb, info->dqi_type);
+	}
+	freedqbuf(tmpbuf);
+	dh->dqdh_next_free = dh->dqdh_prev_free = cpu_to_le32(0);
+	/* No matter whether write succeeds block is out of list */
+	if (write_blk(info, blk, buf) < 0)
+		printk(KERN_ERR "VFS: Can't write block (%u) with free entries.\n", blk);
+	return 0;
+out_buf:
+	freedqbuf(tmpbuf);
+	return err;
+}
+
+/* Insert given block to the beginning of list with free entries */
+static int insert_free_dqentry(struct qtree_mem_dqinfo *info, dqbuf_t buf, uint blk)
+{
+	dqbuf_t tmpbuf = getdqbuf(info->dqi_usable_bs);
+	struct qt_disk_dqdbheader *dh = (struct qt_disk_dqdbheader *)buf;
+	int err;
+
+	if (!tmpbuf)
+		return -ENOMEM;
+	dh->dqdh_next_free = cpu_to_le32(info->dqi_free_entry);
+	dh->dqdh_prev_free = cpu_to_le32(0);
+	err = write_blk(info, blk, buf);
+	if (err < 0)
+		goto out_buf;
+	if (info->dqi_free_entry) {
+		err = read_blk(info, info->dqi_free_entry, tmpbuf);
+		if (err < 0)
+			goto out_buf;
+		((struct qt_disk_dqdbheader *)tmpbuf)->dqdh_prev_free =
+							cpu_to_le32(blk);
+		err = write_blk(info, info->dqi_free_entry, tmpbuf);
+		if (err < 0)
+			goto out_buf;
+	}
+	freedqbuf(tmpbuf);
+	info->dqi_free_entry = blk;
+	mark_info_dirty(info->dqi_sb, info->dqi_type);
+	return 0;
+out_buf:
+	freedqbuf(tmpbuf);
+	return err;
+}
+
+/* Is the entry in the block free? */
+int qtree_entry_unused(struct qtree_mem_dqinfo *info, char *disk)
+{
+	int i;
+
+	for (i = 0; i < info->dqi_entry_size; i++)
+		if (disk[i])
+			return 0;
+	return 1;
+}
+EXPORT_SYMBOL(qtree_entry_unused);
+
+/* Find space for dquot */
+static uint find_free_dqentry(struct qtree_mem_dqinfo *info,
+			      struct dquot *dquot, int *err)
+{
+	uint blk, i;
+	struct qt_disk_dqdbheader *dh;
+	dqbuf_t buf = getdqbuf(info->dqi_usable_bs);
+	char *ddquot;
+
+	*err = 0;
+	if (!buf) {
+		*err = -ENOMEM;
+		return 0;
+	}
+	dh = (struct qt_disk_dqdbheader *)buf;
+	if (info->dqi_free_entry) {
+		blk = info->dqi_free_entry;
+		*err = read_blk(info, blk, buf);
+		if (*err < 0)
+			goto out_buf;
+	} else {
+		blk = get_free_dqblk(info);
+		if ((int)blk < 0) {
+			*err = blk;
+			freedqbuf(buf);
+			return 0;
+		}
+		memset(buf, 0, info->dqi_usable_bs);
+		/* This is enough as block is already zeroed and entry list is empty... */
+		info->dqi_free_entry = blk;
+		mark_info_dirty(dquot->dq_sb, dquot->dq_type);
+	}
+	/* Block will be full? */
+	if (le16_to_cpu(dh->dqdh_entries) + 1 >= qtree_dqstr_in_blk(info)) {
+		*err = remove_free_dqentry(info, buf, blk);
+		if (*err < 0) {
+			printk(KERN_ERR "VFS: find_free_dqentry(): Can't "
+			       "remove block (%u) from entry free list.\n",
+			       blk);
+			goto out_buf;
+		}
+	}
+	le16_add_cpu(&dh->dqdh_entries, 1);
+	/* Find free structure in block */
+	for (i = 0, ddquot = ((char *)buf) + sizeof(struct qt_disk_dqdbheader);
+	     i < qtree_dqstr_in_blk(info) && !qtree_entry_unused(info, ddquot);
+	     i++, ddquot += info->dqi_entry_size);
+#ifdef __QUOTA_QT_PARANOIA
+	if (i == qtree_dqstr_in_blk(info)) {
+		printk(KERN_ERR "VFS: find_free_dqentry(): Data block full "
+				"but it shouldn't.\n");
+		*err = -EIO;
+		goto out_buf;
+	}
+#endif
+	*err = write_blk(info, blk, buf);
+	if (*err < 0) {
+		printk(KERN_ERR "VFS: find_free_dqentry(): Can't write quota "
+				"data block %u.\n", blk);
+		goto out_buf;
+	}
+	dquot->dq_off = (blk << info->dqi_blocksize_bits) +
+			sizeof(struct qt_disk_dqdbheader) +
+			i * info->dqi_entry_size;
+	freedqbuf(buf);
+	return blk;
+out_buf:
+	freedqbuf(buf);
+	return 0;
+}
+
+/* Insert reference to structure into the trie */
+static int do_insert_tree(struct qtree_mem_dqinfo *info, struct dquot *dquot,
+			  uint *treeblk, int depth)
+{
+	dqbuf_t buf = getdqbuf(info->dqi_usable_bs);
+	int ret = 0, newson = 0, newact = 0;
+	__le32 *ref;
+	uint newblk;
+
+	if (!buf)
+		return -ENOMEM;
+	if (!*treeblk) {
+		ret = get_free_dqblk(info);
+		if (ret < 0)
+			goto out_buf;
+		*treeblk = ret;
+		memset(buf, 0, info->dqi_usable_bs);
+		newact = 1;
+	} else {
+		ret = read_blk(info, *treeblk, buf);
+		if (ret < 0) {
+			printk(KERN_ERR "VFS: Can't read tree quota block "
+					"%u.\n", *treeblk);
+			goto out_buf;
+		}
+	}
+	ref = (__le32 *)buf;
+	newblk = le32_to_cpu(ref[get_index(info, dquot->dq_id, depth)]);
+	if (!newblk)
+		newson = 1;
+	if (depth == info->dqi_qtree_depth - 1) {
+#ifdef __QUOTA_QT_PARANOIA
+		if (newblk) {
+			printk(KERN_ERR "VFS: Inserting already present quota "
+					"entry (block %u).\n",
+			       le32_to_cpu(ref[get_index(info,
+						dquot->dq_id, depth)]));
+			ret = -EIO;
+			goto out_buf;
+		}
+#endif
+		newblk = find_free_dqentry(info, dquot, &ret);
+	} else {
+		ret = do_insert_tree(info, dquot, &newblk, depth+1);
+	}
+	if (newson && ret >= 0) {
+		ref[get_index(info, dquot->dq_id, depth)] =
+							cpu_to_le32(newblk);
+		ret = write_blk(info, *treeblk, buf);
+	} else if (newact && ret < 0) {
+		put_free_dqblk(info, buf, *treeblk);
+	}
+out_buf:
+	freedqbuf(buf);
+	return ret;
+}
+
+/* Wrapper for inserting quota structure into tree */
+static inline int dq_insert_tree(struct qtree_mem_dqinfo *info,
+				 struct dquot *dquot)
+{
+	int tmp = QT_TREEOFF;
+	return do_insert_tree(info, dquot, &tmp, 0);
+}
+
+/*
+ *	We don't have to be afraid of deadlocks as we never have quotas on quota files...
+ */
+int qtree_write_dquot(struct qtree_mem_dqinfo *info, struct dquot *dquot)
+{
+	int type = dquot->dq_type;
+	struct super_block *sb = dquot->dq_sb;
+	ssize_t ret;
+	dqbuf_t ddquot = getdqbuf(info->dqi_entry_size);
+
+	if (!ddquot)
+		return -ENOMEM;
+
+	/* dq_off is guarded by dqio_mutex */
+	if (!dquot->dq_off) {
+		ret = dq_insert_tree(info, dquot);
+		if (ret < 0) {
+			printk(KERN_ERR "VFS: Error %zd occurred while "
+					"creating quota.\n", ret);
+			freedqbuf(ddquot);
+			return ret;
+		}
+	}
+	spin_lock(&dq_data_lock);
+	info->dqi_ops->mem2disk_dqblk(ddquot, dquot);
+	spin_unlock(&dq_data_lock);
+	ret = sb->s_op->quota_write(sb, type, (char *)ddquot,
+					info->dqi_entry_size, dquot->dq_off);
+	if (ret != info->dqi_entry_size) {
+		printk(KERN_WARNING "VFS: dquota write failed on dev %s\n",
+		       sb->s_id);
+		if (ret >= 0)
+			ret = -ENOSPC;
+	} else {
+		ret = 0;
+	}
+	dqstats.writes++;
+	freedqbuf(ddquot);
+
+	return ret;
+}
+EXPORT_SYMBOL(qtree_write_dquot);
+
+/* Free dquot entry in data block */
+static int free_dqentry(struct qtree_mem_dqinfo *info, struct dquot *dquot,
+			uint blk)
+{
+	struct qt_disk_dqdbheader *dh;
+	dqbuf_t buf = getdqbuf(info->dqi_usable_bs);
+	int ret = 0;
+
+	if (!buf)
+		return -ENOMEM;
+	if (dquot->dq_off >> info->dqi_blocksize_bits != blk) {
+		printk(KERN_ERR "VFS: Quota structure has offset to other "
+		  "block (%u) than it should (%u).\n", blk,
+		  (uint)(dquot->dq_off >> info->dqi_blocksize_bits));
+		goto out_buf;
+	}
+	ret = read_blk(info, blk, buf);
+	if (ret < 0) {
+		printk(KERN_ERR "VFS: Can't read quota data block %u\n", blk);
+		goto out_buf;
+	}
+	dh = (struct qt_disk_dqdbheader *)buf;
+	le16_add_cpu(&dh->dqdh_entries, -1);
+	if (!le16_to_cpu(dh->dqdh_entries)) {	/* Block got free? */
+		ret = remove_free_dqentry(info, buf, blk);
+		if (ret >= 0)
+			ret = put_free_dqblk(info, buf, blk);
+		if (ret < 0) {
+			printk(KERN_ERR "VFS: Can't move quota data block (%u) "
+			  "to free list.\n", blk);
+			goto out_buf;
+		}
+	} else {
+		memset(buf +
+		       (dquot->dq_off & ((1 << info->dqi_blocksize_bits) - 1)),
+		       0, info->dqi_entry_size);
+		if (le16_to_cpu(dh->dqdh_entries) ==
+		    qtree_dqstr_in_blk(info) - 1) {
+			/* Insert will write block itself */
+			ret = insert_free_dqentry(info, buf, blk);
+			if (ret < 0) {
+				printk(KERN_ERR "VFS: Can't insert quota data "
+				       "block (%u) to free entry list.\n", blk);
+				goto out_buf;
+			}
+		} else {
+			ret = write_blk(info, blk, buf);
+			if (ret < 0) {
+				printk(KERN_ERR "VFS: Can't write quota data "
+				  "block %u\n", blk);
+				goto out_buf;
+			}
+		}
+	}
+	dquot->dq_off = 0;	/* Quota is now unattached */
+out_buf:
+	freedqbuf(buf);
+	return ret;
+}
+
+/* Remove reference to dquot from tree */
+static int remove_tree(struct qtree_mem_dqinfo *info, struct dquot *dquot,
+		       uint *blk, int depth)
+{
+	dqbuf_t buf = getdqbuf(info->dqi_usable_bs);
+	int ret = 0;
+	uint newblk;
+	__le32 *ref = (__le32 *)buf;
+
+	if (!buf)
+		return -ENOMEM;
+	ret = read_blk(info, *blk, buf);
+	if (ret < 0) {
+		printk(KERN_ERR "VFS: Can't read quota data block %u\n", *blk);
+		goto out_buf;
+	}
+	newblk = le32_to_cpu(ref[get_index(info, dquot->dq_id, depth)]);
+	if (depth == info->dqi_qtree_depth - 1) {
+		ret = free_dqentry(info, dquot, newblk);
+		newblk = 0;
+	} else {
+		ret = remove_tree(info, dquot, &newblk, depth+1);
+	}
+	if (ret >= 0 && !newblk) {
+		int i;
+		ref[get_index(info, dquot->dq_id, depth)] = cpu_to_le32(0);
+		/* Block got empty? */
+		for (i = 0;
+		     i < (info->dqi_usable_bs >> 2) && !ref[i];
+		     i++);
+		/* Don't put the root block into the free block list */
+		if (i == (info->dqi_usable_bs >> 2)
+		    && *blk != QT_TREEOFF) {
+			put_free_dqblk(info, buf, *blk);
+			*blk = 0;
+		} else {
+			ret = write_blk(info, *blk, buf);
+			if (ret < 0)
+				printk(KERN_ERR "VFS: Can't write quota tree "
+				  "block %u.\n", *blk);
+		}
+	}
+out_buf:
+	freedqbuf(buf);
+	return ret;
+}
+
+/* Delete dquot from tree */
+int qtree_delete_dquot(struct qtree_mem_dqinfo *info, struct dquot *dquot)
+{
+	uint tmp = QT_TREEOFF;
+
+	if (!dquot->dq_off)	/* Even not allocated? */
+		return 0;
+	return remove_tree(info, dquot, &tmp, 0);
+}
+EXPORT_SYMBOL(qtree_delete_dquot);
+
+/* Find entry in block */
+static loff_t find_block_dqentry(struct qtree_mem_dqinfo *info,
+				 struct dquot *dquot, uint blk)
+{
+	dqbuf_t buf = getdqbuf(info->dqi_usable_bs);
+	loff_t ret = 0;
+	int i;
+	char *ddquot;
+
+	if (!buf)
+		return -ENOMEM;
+	ret = read_blk(info, blk, buf);
+	if (ret < 0) {
+		printk(KERN_ERR "VFS: Can't read quota tree block %u.\n", blk);
+		goto out_buf;
+	}
+	for (i = 0, ddquot = ((char *)buf) + sizeof(struct qt_disk_dqdbheader);
+	     i < qtree_dqstr_in_blk(info) && !info->dqi_ops->is_id(ddquot, dquot);
+	     i++, ddquot += info->dqi_entry_size);
+	if (i == qtree_dqstr_in_blk(info)) {
+		printk(KERN_ERR "VFS: Quota for id %u referenced "
+		  "but not present.\n", dquot->dq_id);
+		ret = -EIO;
+		goto out_buf;
+	} else {
+		ret = (blk << info->dqi_blocksize_bits) + sizeof(struct
+		  qt_disk_dqdbheader) + i * info->dqi_entry_size;
+	}
+out_buf:
+	freedqbuf(buf);
+	return ret;
+}
+
+/* Find entry for given id in the tree */
+static loff_t find_tree_dqentry(struct qtree_mem_dqinfo *info,
+				struct dquot *dquot, uint blk, int depth)
+{
+	dqbuf_t buf = getdqbuf(info->dqi_usable_bs);
+	loff_t ret = 0;
+	__le32 *ref = (__le32 *)buf;
+
+	if (!buf)
+		return -ENOMEM;
+	ret = read_blk(info, blk, buf);
+	if (ret < 0) {
+		printk(KERN_ERR "VFS: Can't read quota tree block %u.\n", blk);
+		goto out_buf;
+	}
+	ret = 0;
+	blk = le32_to_cpu(ref[get_index(info, dquot->dq_id, depth)]);
+	if (!blk)	/* No reference? */
+		goto out_buf;
+	if (depth < info->dqi_qtree_depth - 1)
+		ret = find_tree_dqentry(info, dquot, blk, depth+1);
+	else
+		ret = find_block_dqentry(info, dquot, blk);
+out_buf:
+	freedqbuf(buf);
+	return ret;
+}
+
+/* Find entry for given id in the tree - wrapper function */
+static inline loff_t find_dqentry(struct qtree_mem_dqinfo *info,
+				  struct dquot *dquot)
+{
+	return find_tree_dqentry(info, dquot, QT_TREEOFF, 0);
+}
+
+int qtree_read_dquot(struct qtree_mem_dqinfo *info, struct dquot *dquot)
+{
+	int type = dquot->dq_type;
+	struct super_block *sb = dquot->dq_sb;
+	loff_t offset;
+	dqbuf_t ddquot;
+	int ret = 0;
+
+#ifdef __QUOTA_QT_PARANOIA
+	/* Invalidated quota? */
+	if (!sb_dqopt(dquot->dq_sb)->files[type]) {
+		printk(KERN_ERR "VFS: Quota invalidated while reading!\n");
+		return -EIO;
+	}
+#endif
+	/* Do we know offset of the dquot entry in the quota file? */
+	if (!dquot->dq_off) {
+		offset = find_dqentry(info, dquot);
+		if (offset <= 0) {	/* Entry not present? */
+			if (offset < 0)
+				printk(KERN_ERR "VFS: Can't read quota "
+				  "structure for id %u.\n", dquot->dq_id);
+			dquot->dq_off = 0;
+			set_bit(DQ_FAKE_B, &dquot->dq_flags);
+			memset(&dquot->dq_dqb, 0, sizeof(struct mem_dqblk));
+			ret = offset;
+			goto out;
+		}
+		dquot->dq_off = offset;
+	}
+	ddquot = getdqbuf(info->dqi_entry_size);
+	if (!ddquot)
+		return -ENOMEM;
+	ret = sb->s_op->quota_read(sb, type, (char *)ddquot,
+				   info->dqi_entry_size, dquot->dq_off);
+	if (ret != info->dqi_entry_size) {
+		if (ret >= 0)
+			ret = -EIO;
+		printk(KERN_ERR "VFS: Error while reading quota "
+				"structure for id %u.\n", dquot->dq_id);
+		set_bit(DQ_FAKE_B, &dquot->dq_flags);
+		memset(&dquot->dq_dqb, 0, sizeof(struct mem_dqblk));
+		freedqbuf(ddquot);
+		goto out;
+	}
+	spin_lock(&dq_data_lock);
+	info->dqi_ops->disk2mem_dqblk(dquot, ddquot);
+	if (!dquot->dq_dqb.dqb_bhardlimit &&
+	    !dquot->dq_dqb.dqb_bsoftlimit &&
+	    !dquot->dq_dqb.dqb_ihardlimit &&
+	    !dquot->dq_dqb.dqb_isoftlimit)
+		set_bit(DQ_FAKE_B, &dquot->dq_flags);
+	spin_unlock(&dq_data_lock);
+	freedqbuf(ddquot);
+out:
+	dqstats.reads++;
+	return ret;
+}
+EXPORT_SYMBOL(qtree_read_dquot);
+
+/* Check whether dquot should not be deleted. We know we are
+ * the only one operating on dquot (thanks to dq_lock) */
+int qtree_release_dquot(struct qtree_mem_dqinfo *info, struct dquot *dquot)
+{
+	if (test_bit(DQ_FAKE_B, &dquot->dq_flags) && !(dquot->dq_dqb.dqb_curinodes | dquot->dq_dqb.dqb_curspace))
+		return qtree_delete_dquot(info, dquot);
+	return 0;
+}
+EXPORT_SYMBOL(qtree_release_dquot);

fs/quota_tree.h

@@ -0,0 +1,25 @@
+/*
+ *	Definitions of structures for vfsv0 quota format
+ */
+
+#ifndef _LINUX_QUOTA_TREE_H
+#define _LINUX_QUOTA_TREE_H
+
+#include <linux/types.h>
+#include <linux/quota.h>
+
+/*
+ *  Structure of header of block with quota structures. It is padded to 16 bytes so
+ *  there will be space for exactly 21 quota-entries in a block
+ */
+struct qt_disk_dqdbheader {
+	__le32 dqdh_next_free;	/* Number of next block with free entry */
+	__le32 dqdh_prev_free;	/* Number of previous block with free entry */
+	__le16 dqdh_entries;	/* Number of valid entries in block */
+	__le16 dqdh_pad1;
+	__le32 dqdh_pad2;
+};
+
+#define QT_TREEOFF	1		/* Offset of tree in file in blocks */
+
+#endif /* _LINUX_QUOTAIO_TREE_H */

fs/quota_v2.c

@@ -14,6 +14,7 @@
 
 #include <asm/byteorder.h>
 
+#include "quota_tree.h"
 #include "quotaio_v2.h"
 
 MODULE_AUTHOR("Jan Kara");
@@ -22,10 +23,15 @@ MODULE_LICENSE("GPL");
 
 #define __QUOTA_V2_PARANOIA
 
-typedef char *dqbuf_t;
+static void v2_mem2diskdqb(void *dp, struct dquot *dquot);
+static void v2_disk2memdqb(struct dquot *dquot, void *dp);
+static int v2_is_id(void *dp, struct dquot *dquot);
 
-#define GETIDINDEX(id, depth) (((id) >> ((V2_DQTREEDEPTH-(depth)-1)*8)) & 0xff)
-#define GETENTRIES(buf) ((struct v2_disk_dqblk *)(((char *)buf)+sizeof(struct v2_disk_dqdbheader)))
+static struct qtree_fmt_operations v2_qtree_ops = {
+	.mem2disk_dqblk = v2_mem2diskdqb,
+	.disk2mem_dqblk = v2_disk2memdqb,
+	.is_id = v2_is_id,
+};
 
 #define QUOTABLOCK_BITS 10
 #define QUOTABLOCK_SIZE (1 << QUOTABLOCK_BITS)
@@ -64,7 +70,7 @@ static int v2_check_quota_file(struct super_block *sb, int type)
 static int v2_read_file_info(struct super_block *sb, int type)
 {
 	struct v2_disk_dqinfo dinfo;
-	struct mem_dqinfo *info = sb_dqopt(sb)->info+type;
+	struct mem_dqinfo *info = sb_dqinfo(sb, type);
 	ssize_t size;
 
 	size = sb->s_op->quota_read(sb, type, (char *)&dinfo,
@@ -80,9 +86,16 @@ static int v2_read_file_info(struct super_block *sb, int type)
 	info->dqi_bgrace = le32_to_cpu(dinfo.dqi_bgrace);
 	info->dqi_igrace = le32_to_cpu(dinfo.dqi_igrace);
 	info->dqi_flags = le32_to_cpu(dinfo.dqi_flags);
-	info->u.v2_i.dqi_blocks = le32_to_cpu(dinfo.dqi_blocks);
-	info->u.v2_i.dqi_free_blk = le32_to_cpu(dinfo.dqi_free_blk);
-	info->u.v2_i.dqi_free_entry = le32_to_cpu(dinfo.dqi_free_entry);
+	info->u.v2_i.i.dqi_sb = sb;
+	info->u.v2_i.i.dqi_type = type;
+	info->u.v2_i.i.dqi_blocks = le32_to_cpu(dinfo.dqi_blocks);
+	info->u.v2_i.i.dqi_free_blk = le32_to_cpu(dinfo.dqi_free_blk);
+	info->u.v2_i.i.dqi_free_entry = le32_to_cpu(dinfo.dqi_free_entry);
+	info->u.v2_i.i.dqi_blocksize_bits = V2_DQBLKSIZE_BITS;
+	info->u.v2_i.i.dqi_usable_bs = 1 << V2_DQBLKSIZE_BITS;
+	info->u.v2_i.i.dqi_qtree_depth = qtree_depth(&info->u.v2_i.i);
+	info->u.v2_i.i.dqi_entry_size = sizeof(struct v2_disk_dqblk);
+	info->u.v2_i.i.dqi_ops = &v2_qtree_ops;
 	return 0;
 }
 
@@ -90,7 +103,7 @@ static int v2_read_file_info(struct super_block *sb, int type)
 static int v2_write_file_info(struct super_block *sb, int type)
 {
 	struct v2_disk_dqinfo dinfo;
-	struct mem_dqinfo *info = sb_dqopt(sb)->info+type;
+	struct mem_dqinfo *info = sb_dqinfo(sb, type);
 	ssize_t size;
 
 	spin_lock(&dq_data_lock);
@@ -99,9 +112,9 @@ static int v2_write_file_info(struct super_block *sb, int type)
 	dinfo.dqi_igrace = cpu_to_le32(info->dqi_igrace);
 	dinfo.dqi_flags = cpu_to_le32(info->dqi_flags & DQF_MASK);
 	spin_unlock(&dq_data_lock);
-	dinfo.dqi_blocks = cpu_to_le32(info->u.v2_i.dqi_blocks);
-	dinfo.dqi_free_blk = cpu_to_le32(info->u.v2_i.dqi_free_blk);
-	dinfo.dqi_free_entry = cpu_to_le32(info->u.v2_i.dqi_free_entry);
+	dinfo.dqi_blocks = cpu_to_le32(info->u.v2_i.i.dqi_blocks);
+	dinfo.dqi_free_blk = cpu_to_le32(info->u.v2_i.i.dqi_free_blk);
+	dinfo.dqi_free_entry = cpu_to_le32(info->u.v2_i.i.dqi_free_entry);
 	size = sb->s_op->quota_write(sb, type, (char *)&dinfo,
 	       sizeof(struct v2_disk_dqinfo), V2_DQINFOOFF);
 	if (size != sizeof(struct v2_disk_dqinfo)) {
@@ -112,8 +125,11 @@ static int v2_write_file_info(struct super_block *sb, int type)
 	return 0;
 }
 
-static void disk2memdqb(struct mem_dqblk *m, struct v2_disk_dqblk *d)
+static void v2_disk2memdqb(struct dquot *dquot, void *dp)
 {
+	struct v2_disk_dqblk *d = dp, empty;
+	struct mem_dqblk *m = &dquot->dq_dqb;
+
 	m->dqb_ihardlimit = le32_to_cpu(d->dqb_ihardlimit);
 	m->dqb_isoftlimit = le32_to_cpu(d->dqb_isoftlimit);
 	m->dqb_curinodes = le32_to_cpu(d->dqb_curinodes);
@@ -122,10 +138,20 @@ static void disk2memdqb(struct mem_dqblk *m, struct v2_disk_dqblk *d)
 	m->dqb_bsoftlimit = v2_qbtos(le32_to_cpu(d->dqb_bsoftlimit));
 	m->dqb_curspace = le64_to_cpu(d->dqb_curspace);
 	m->dqb_btime = le64_to_cpu(d->dqb_btime);
+	/* We need to escape back all-zero structure */
+	memset(&empty, 0, sizeof(struct v2_disk_dqblk));
+	empty.dqb_itime = cpu_to_le64(1);
+	if (!memcmp(&empty, dp, sizeof(struct v2_disk_dqblk)))
+		m->dqb_itime = 0;
 }
 
-static void mem2diskdqb(struct v2_disk_dqblk *d, struct mem_dqblk *m, qid_t id)
+static void v2_mem2diskdqb(void *dp, struct dquot *dquot)
 {
+	struct v2_disk_dqblk *d = dp;
+	struct mem_dqblk *m = &dquot->dq_dqb;
+	struct qtree_mem_dqinfo *info =
+			&sb_dqinfo(dquot->dq_sb, dquot->dq_type)->u.v2_i.i;
+
 	d->dqb_ihardlimit = cpu_to_le32(m->dqb_ihardlimit);
 	d->dqb_isoftlimit = cpu_to_le32(m->dqb_isoftlimit);
 	d->dqb_curinodes = cpu_to_le32(m->dqb_curinodes);
@@ -134,553 +160,35 @@ static void mem2diskdqb(struct v2_disk_dqblk *d, struct mem_dqblk *m, qid_t id)
 	d->dqb_bsoftlimit = cpu_to_le32(v2_stoqb(m->dqb_bsoftlimit));
 	d->dqb_curspace = cpu_to_le64(m->dqb_curspace);
 	d->dqb_btime = cpu_to_le64(m->dqb_btime);
-	d->dqb_id = cpu_to_le32(id);
-}
-
-static dqbuf_t getdqbuf(void)
-{
-	dqbuf_t buf = kmalloc(V2_DQBLKSIZE, GFP_NOFS);
-	if (!buf)
-		printk(KERN_WARNING "VFS: Not enough memory for quota buffers.\n");
-	return buf;
-}
-
-static inline void freedqbuf(dqbuf_t buf)
-{
-	kfree(buf);
-}
-
-static inline ssize_t read_blk(struct super_block *sb, int type, uint blk, dqbuf_t buf)
-{
-	memset(buf, 0, V2_DQBLKSIZE);
-	return sb->s_op->quota_read(sb, type, (char *)buf,
-	       V2_DQBLKSIZE, blk << V2_DQBLKSIZE_BITS);
-}
-
-static inline ssize_t write_blk(struct super_block *sb, int type, uint blk, dqbuf_t buf)
-{
-	return sb->s_op->quota_write(sb, type, (char *)buf,
-	       V2_DQBLKSIZE, blk << V2_DQBLKSIZE_BITS);
-}
-
-/* Remove empty block from list and return it */
-static int get_free_dqblk(struct super_block *sb, int type)
-{
-	dqbuf_t buf = getdqbuf();
-	struct mem_dqinfo *info = sb_dqinfo(sb, type);
-	struct v2_disk_dqdbheader *dh = (struct v2_disk_dqdbheader *)buf;
-	int ret, blk;
-
-	if (!buf)
-		return -ENOMEM;
-	if (info->u.v2_i.dqi_free_blk) {
-		blk = info->u.v2_i.dqi_free_blk;
-		if ((ret = read_blk(sb, type, blk, buf)) < 0)
-			goto out_buf;
-		info->u.v2_i.dqi_free_blk = le32_to_cpu(dh->dqdh_next_free);
-	}
-	else {
-		memset(buf, 0, V2_DQBLKSIZE);
-		/* Assure block allocation... */
-		if ((ret = write_blk(sb, type, info->u.v2_i.dqi_blocks, buf)) < 0)
-			goto out_buf;
-		blk = info->u.v2_i.dqi_blocks++;
-	}
-	mark_info_dirty(sb, type);
-	ret = blk;
-out_buf:
-	freedqbuf(buf);
-	return ret;
-}
-
-/* Insert empty block to the list */
-static int put_free_dqblk(struct super_block *sb, int type, dqbuf_t buf, uint blk)
-{
-	struct mem_dqinfo *info = sb_dqinfo(sb, type);
-	struct v2_disk_dqdbheader *dh = (struct v2_disk_dqdbheader *)buf;
-	int err;
-
-	dh->dqdh_next_free = cpu_to_le32(info->u.v2_i.dqi_free_blk);
-	dh->dqdh_prev_free = cpu_to_le32(0);
-	dh->dqdh_entries = cpu_to_le16(0);
-	info->u.v2_i.dqi_free_blk = blk;
-	mark_info_dirty(sb, type);
-	/* Some strange block. We had better leave it... */
-	if ((err = write_blk(sb, type, blk, buf)) < 0)
-		return err;
-	return 0;
+	d->dqb_id = cpu_to_le32(dquot->dq_id);
+	if (qtree_entry_unused(info, dp))
+		d->dqb_itime = cpu_to_le64(1);
 }
 
-/* Remove given block from the list of blocks with free entries */
-static int remove_free_dqentry(struct super_block *sb, int type, dqbuf_t buf, uint blk)
+static int v2_is_id(void *dp, struct dquot *dquot)
 {
-	dqbuf_t tmpbuf = getdqbuf();
-	struct mem_dqinfo *info = sb_dqinfo(sb, type);
-	struct v2_disk_dqdbheader *dh = (struct v2_disk_dqdbheader *)buf;
-	uint nextblk = le32_to_cpu(dh->dqdh_next_free), prevblk = le32_to_cpu(dh->dqdh_prev_free);
-	int err;
+	struct v2_disk_dqblk *d = dp;
+	struct qtree_mem_dqinfo *info =
+			&sb_dqinfo(dquot->dq_sb, dquot->dq_type)->u.v2_i.i;
 
-	if (!tmpbuf)
-		return -ENOMEM;
-	if (nextblk) {
-		if ((err = read_blk(sb, type, nextblk, tmpbuf)) < 0)
-			goto out_buf;
-		((struct v2_disk_dqdbheader *)tmpbuf)->dqdh_prev_free = dh->dqdh_prev_free;
-		if ((err = write_blk(sb, type, nextblk, tmpbuf)) < 0)
-			goto out_buf;
-	}
-	if (prevblk) {
-		if ((err = read_blk(sb, type, prevblk, tmpbuf)) < 0)
-			goto out_buf;
-		((struct v2_disk_dqdbheader *)tmpbuf)->dqdh_next_free = dh->dqdh_next_free;
-		if ((err = write_blk(sb, type, prevblk, tmpbuf)) < 0)
-			goto out_buf;
-	}
-	else {
-		info->u.v2_i.dqi_free_entry = nextblk;
-		mark_info_dirty(sb, type);
-	}
-	freedqbuf(tmpbuf);
-	dh->dqdh_next_free = dh->dqdh_prev_free = cpu_to_le32(0);
-	/* No matter whether write succeeds block is out of list */
-	if (write_blk(sb, type, blk, buf) < 0)
-		printk(KERN_ERR "VFS: Can't write block (%u) with free entries.\n", blk);
-	return 0;
-out_buf:
-	freedqbuf(tmpbuf);
-	return err;
-}
-
-/* Insert given block to the beginning of list with free entries */
-static int insert_free_dqentry(struct super_block *sb, int type, dqbuf_t buf, uint blk)
-{
-	dqbuf_t tmpbuf = getdqbuf();
-	struct mem_dqinfo *info = sb_dqinfo(sb, type);
-	struct v2_disk_dqdbheader *dh = (struct v2_disk_dqdbheader *)buf;
-	int err;
-
-	if (!tmpbuf)
-		return -ENOMEM;
-	dh->dqdh_next_free = cpu_to_le32(info->u.v2_i.dqi_free_entry);
-	dh->dqdh_prev_free = cpu_to_le32(0);
-	if ((err = write_blk(sb, type, blk, buf)) < 0)
-		goto out_buf;
-	if (info->u.v2_i.dqi_free_entry) {
-		if ((err = read_blk(sb, type, info->u.v2_i.dqi_free_entry, tmpbuf)) < 0)
-			goto out_buf;
-		((struct v2_disk_dqdbheader *)tmpbuf)->dqdh_prev_free = cpu_to_le32(blk);
-		if ((err = write_blk(sb, type, info->u.v2_i.dqi_free_entry, tmpbuf)) < 0)
-			goto out_buf;
-	}
-	freedqbuf(tmpbuf);
-	info->u.v2_i.dqi_free_entry = blk;
-	mark_info_dirty(sb, type);
-	return 0;
-out_buf:
-	freedqbuf(tmpbuf);
-	return err;
-}
-
-/* Find space for dquot */
-static uint find_free_dqentry(struct dquot *dquot, int *err)
-{
-	struct super_block *sb = dquot->dq_sb;
-	struct mem_dqinfo *info = sb_dqopt(sb)->info+dquot->dq_type;
-	uint blk, i;
-	struct v2_disk_dqdbheader *dh;
-	struct v2_disk_dqblk *ddquot;
-	struct v2_disk_dqblk fakedquot;
-	dqbuf_t buf;
-
-	*err = 0;
-	if (!(buf = getdqbuf())) {
-		*err = -ENOMEM;
+	if (qtree_entry_unused(info, dp))
 		return 0;
-	}
-	dh = (struct v2_disk_dqdbheader *)buf;
-	ddquot = GETENTRIES(buf);
-	if (info->u.v2_i.dqi_free_entry) {
-		blk = info->u.v2_i.dqi_free_entry;
-		if ((*err = read_blk(sb, dquot->dq_type, blk, buf)) < 0)
-			goto out_buf;
-	}
-	else {
-		blk = get_free_dqblk(sb, dquot->dq_type);
-		if ((int)blk < 0) {
-			*err = blk;
-			freedqbuf(buf);
-			return 0;
-		}
-		memset(buf, 0, V2_DQBLKSIZE);
-		/* This is enough as block is already zeroed and entry list is empty... */
-		info->u.v2_i.dqi_free_entry = blk;
-		mark_info_dirty(sb, dquot->dq_type);
-	}
-	if (le16_to_cpu(dh->dqdh_entries)+1 >= V2_DQSTRINBLK)	/* Block will be full? */
-		if ((*err = remove_free_dqentry(sb, dquot->dq_type, buf, blk)) < 0) {
-			printk(KERN_ERR "VFS: find_free_dqentry(): Can't remove block (%u) from entry free list.\n", blk);
-			goto out_buf;
-		}
-	le16_add_cpu(&dh->dqdh_entries, 1);
-	memset(&fakedquot, 0, sizeof(struct v2_disk_dqblk));
-	/* Find free structure in block */
-	for (i = 0; i < V2_DQSTRINBLK && memcmp(&fakedquot, ddquot+i, sizeof(struct v2_disk_dqblk)); i++);
-#ifdef __QUOTA_V2_PARANOIA
-	if (i == V2_DQSTRINBLK) {
-		printk(KERN_ERR "VFS: find_free_dqentry(): Data block full but it shouldn't.\n");
-		*err = -EIO;
-		goto out_buf;
-	}
-#endif
-	if ((*err = write_blk(sb, dquot->dq_type, blk, buf)) < 0) {
-		printk(KERN_ERR "VFS: find_free_dqentry(): Can't write quota data block %u.\n", blk);
-		goto out_buf;
-	}
-	dquot->dq_off = (blk<<V2_DQBLKSIZE_BITS)+sizeof(struct v2_disk_dqdbheader)+i*sizeof(struct v2_disk_dqblk);
-	freedqbuf(buf);
-	return blk;
-out_buf:
-	freedqbuf(buf);
-	return 0;
+	return le32_to_cpu(d->dqb_id) == dquot->dq_id;
 }
 
-/* Insert reference to structure into the trie */
-static int do_insert_tree(struct dquot *dquot, uint *treeblk, int depth)
-{
-	struct super_block *sb = dquot->dq_sb;
-	dqbuf_t buf;
-	int ret = 0, newson = 0, newact = 0;
-	__le32 *ref;
-	uint newblk;
-
-	if (!(buf = getdqbuf()))
-		return -ENOMEM;
-	if (!*treeblk) {
-		ret = get_free_dqblk(sb, dquot->dq_type);
-		if (ret < 0)
-			goto out_buf;
-		*treeblk = ret;
-		memset(buf, 0, V2_DQBLKSIZE);
-		newact = 1;
-	}
-	else {
-		if ((ret = read_blk(sb, dquot->dq_type, *treeblk, buf)) < 0) {
-			printk(KERN_ERR "VFS: Can't read tree quota block %u.\n", *treeblk);
-			goto out_buf;
-		}
-	}
-	ref = (__le32 *)buf;
-	newblk = le32_to_cpu(ref[GETIDINDEX(dquot->dq_id, depth)]);
-	if (!newblk)
-		newson = 1;
-	if (depth == V2_DQTREEDEPTH-1) {
-#ifdef __QUOTA_V2_PARANOIA
-		if (newblk) {
-			printk(KERN_ERR "VFS: Inserting already present quota entry (block %u).\n", le32_to_cpu(ref[GETIDINDEX(dquot->dq_id, depth)]));
-			ret = -EIO;
-			goto out_buf;
-		}
-#endif
-		newblk = find_free_dqentry(dquot, &ret);
-	}
-	else
-		ret = do_insert_tree(dquot, &newblk, depth+1);
-	if (newson && ret >= 0) {
-		ref[GETIDINDEX(dquot->dq_id, depth)] = cpu_to_le32(newblk);
-		ret = write_blk(sb, dquot->dq_type, *treeblk, buf);
-	}
-	else if (newact && ret < 0)
-		put_free_dqblk(sb, dquot->dq_type, buf, *treeblk);
-out_buf:
-	freedqbuf(buf);
-	return ret;
-}
-
-/* Wrapper for inserting quota structure into tree */
-static inline int dq_insert_tree(struct dquot *dquot)
+static int v2_read_dquot(struct dquot *dquot)
 {
-	int tmp = V2_DQTREEOFF;
-	return do_insert_tree(dquot, &tmp, 0);
+	return qtree_read_dquot(&sb_dqinfo(dquot->dq_sb, dquot->dq_type)->u.v2_i.i, dquot);
 }
 
-/*
- *	We don't have to be afraid of deadlocks as we never have quotas on quota files...
- */
 static int v2_write_dquot(struct dquot *dquot)
 {
-	int type = dquot->dq_type;
-	ssize_t ret;
-	struct v2_disk_dqblk ddquot, empty;
-
-	/* dq_off is guarded by dqio_mutex */
-	if (!dquot->dq_off)
-		if ((ret = dq_insert_tree(dquot)) < 0) {
-			printk(KERN_ERR "VFS: Error %zd occurred while creating quota.\n", ret);
-			return ret;
-		}
-	spin_lock(&dq_data_lock);
-	mem2diskdqb(&ddquot, &dquot->dq_dqb, dquot->dq_id);
-	/* Argh... We may need to write structure full of zeroes but that would be
-	 * treated as an empty place by the rest of the code. Format change would
-	 * be definitely cleaner but the problems probably are not worth it */
-	memset(&empty, 0, sizeof(struct v2_disk_dqblk));
-	if (!memcmp(&empty, &ddquot, sizeof(struct v2_disk_dqblk)))
-		ddquot.dqb_itime = cpu_to_le64(1);
-	spin_unlock(&dq_data_lock);
-	ret = dquot->dq_sb->s_op->quota_write(dquot->dq_sb, type,
-	      (char *)&ddquot, sizeof(struct v2_disk_dqblk), dquot->dq_off);
-	if (ret != sizeof(struct v2_disk_dqblk)) {
-		printk(KERN_WARNING "VFS: dquota write failed on dev %s\n", dquot->dq_sb->s_id);
-		if (ret >= 0)
-			ret = -ENOSPC;
-	}
-	else
-		ret = 0;
-	dqstats.writes++;
-
-	return ret;
+	return qtree_write_dquot(&sb_dqinfo(dquot->dq_sb, dquot->dq_type)->u.v2_i.i, dquot);
 }
 
-/* Free dquot entry in data block */
-static int free_dqentry(struct dquot *dquot, uint blk)
-{
-	struct super_block *sb = dquot->dq_sb;
-	int type = dquot->dq_type;
-	struct v2_disk_dqdbheader *dh;
-	dqbuf_t buf = getdqbuf();
-	int ret = 0;
-
-	if (!buf)
-		return -ENOMEM;
-	if (dquot->dq_off >> V2_DQBLKSIZE_BITS != blk) {
-		printk(KERN_ERR "VFS: Quota structure has offset to other "
-		  "block (%u) than it should (%u).\n", blk,
-		  (uint)(dquot->dq_off >> V2_DQBLKSIZE_BITS));
-		goto out_buf;
-	}
-	if ((ret = read_blk(sb, type, blk, buf)) < 0) {
-		printk(KERN_ERR "VFS: Can't read quota data block %u\n", blk);
-		goto out_buf;
-	}
-	dh = (struct v2_disk_dqdbheader *)buf;
-	le16_add_cpu(&dh->dqdh_entries, -1);
-	if (!le16_to_cpu(dh->dqdh_entries)) {	/* Block got free? */
-		if ((ret = remove_free_dqentry(sb, type, buf, blk)) < 0 ||
-		    (ret = put_free_dqblk(sb, type, buf, blk)) < 0) {
-			printk(KERN_ERR "VFS: Can't move quota data block (%u) "
-			  "to free list.\n", blk);
-			goto out_buf;
-		}
-	}
-	else {
-		memset(buf+(dquot->dq_off & ((1 << V2_DQBLKSIZE_BITS)-1)), 0,
-		  sizeof(struct v2_disk_dqblk));
-		if (le16_to_cpu(dh->dqdh_entries) == V2_DQSTRINBLK-1) {
-			/* Insert will write block itself */
-			if ((ret = insert_free_dqentry(sb, type, buf, blk)) < 0) {
-				printk(KERN_ERR "VFS: Can't insert quota data block (%u) to free entry list.\n", blk);
-				goto out_buf;
-			}
-		}
-		else
-			if ((ret = write_blk(sb, type, blk, buf)) < 0) {
-				printk(KERN_ERR "VFS: Can't write quota data "
-				  "block %u\n", blk);
-				goto out_buf;
-			}
-	}
-	dquot->dq_off = 0;	/* Quota is now unattached */
-out_buf:
-	freedqbuf(buf);
-	return ret;
-}
-
-/* Remove reference to dquot from tree */
-static int remove_tree(struct dquot *dquot, uint *blk, int depth)
-{
-	struct super_block *sb = dquot->dq_sb;
-	int type = dquot->dq_type;
-	dqbuf_t buf = getdqbuf();
-	int ret = 0;
-	uint newblk;
-	__le32 *ref = (__le32 *)buf;
-	
-	if (!buf)
-		return -ENOMEM;
-	if ((ret = read_blk(sb, type, *blk, buf)) < 0) {
-		printk(KERN_ERR "VFS: Can't read quota data block %u\n", *blk);
-		goto out_buf;
-	}
-	newblk = le32_to_cpu(ref[GETIDINDEX(dquot->dq_id, depth)]);
-	if (depth == V2_DQTREEDEPTH-1) {
-		ret = free_dqentry(dquot, newblk);
-		newblk = 0;
-	}
-	else
-		ret = remove_tree(dquot, &newblk, depth+1);
-	if (ret >= 0 && !newblk) {
-		int i;
-		ref[GETIDINDEX(dquot->dq_id, depth)] = cpu_to_le32(0);
-		for (i = 0; i < V2_DQBLKSIZE && !buf[i]; i++);	/* Block got empty? */
-		/* Don't put the root block into the free block list */
-		if (i == V2_DQBLKSIZE && *blk != V2_DQTREEOFF) {
-			put_free_dqblk(sb, type, buf, *blk);
-			*blk = 0;
-		}
-		else
-			if ((ret = write_blk(sb, type, *blk, buf)) < 0)
-				printk(KERN_ERR "VFS: Can't write quota tree "
-				  "block %u.\n", *blk);
-	}
-out_buf:
-	freedqbuf(buf);
-	return ret;	
-}
-
-/* Delete dquot from tree */
-static int v2_delete_dquot(struct dquot *dquot)
-{
-	uint tmp = V2_DQTREEOFF;
-
-	if (!dquot->dq_off)	/* Even not allocated? */
-		return 0;
-	return remove_tree(dquot, &tmp, 0);
-}
-
-/* Find entry in block */
-static loff_t find_block_dqentry(struct dquot *dquot, uint blk)
-{
-	dqbuf_t buf = getdqbuf();
-	loff_t ret = 0;
-	int i;
-	struct v2_disk_dqblk *ddquot = GETENTRIES(buf);
-
-	if (!buf)
-		return -ENOMEM;
-	if ((ret = read_blk(dquot->dq_sb, dquot->dq_type, blk, buf)) < 0) {
-		printk(KERN_ERR "VFS: Can't read quota tree block %u.\n", blk);
-		goto out_buf;
-	}
-	if (dquot->dq_id)
-		for (i = 0; i < V2_DQSTRINBLK &&
-		     le32_to_cpu(ddquot[i].dqb_id) != dquot->dq_id; i++);
-	else {	/* ID 0 as a bit more complicated searching... */
-		struct v2_disk_dqblk fakedquot;
-
-		memset(&fakedquot, 0, sizeof(struct v2_disk_dqblk));
-		for (i = 0; i < V2_DQSTRINBLK; i++)
-			if (!le32_to_cpu(ddquot[i].dqb_id) &&
-			    memcmp(&fakedquot, ddquot+i, sizeof(struct v2_disk_dqblk)))
-				break;
-	}
-	if (i == V2_DQSTRINBLK) {
-		printk(KERN_ERR "VFS: Quota for id %u referenced "
-		  "but not present.\n", dquot->dq_id);
-		ret = -EIO;
-		goto out_buf;
-	}
-	else
-		ret = (blk << V2_DQBLKSIZE_BITS) + sizeof(struct
-		  v2_disk_dqdbheader) + i * sizeof(struct v2_disk_dqblk);
-out_buf:
-	freedqbuf(buf);
-	return ret;
-}
-
-/* Find entry for given id in the tree */
-static loff_t find_tree_dqentry(struct dquot *dquot, uint blk, int depth)
-{
-	dqbuf_t buf = getdqbuf();
-	loff_t ret = 0;
-	__le32 *ref = (__le32 *)buf;
-
-	if (!buf)
-		return -ENOMEM;
-	if ((ret = read_blk(dquot->dq_sb, dquot->dq_type, blk, buf)) < 0) {
-		printk(KERN_ERR "VFS: Can't read quota tree block %u.\n", blk);
-		goto out_buf;
-	}
-	ret = 0;
-	blk = le32_to_cpu(ref[GETIDINDEX(dquot->dq_id, depth)]);
-	if (!blk)	/* No reference? */
-		goto out_buf;
-	if (depth < V2_DQTREEDEPTH-1)
-		ret = find_tree_dqentry(dquot, blk, depth+1);
-	else
-		ret = find_block_dqentry(dquot, blk);
-out_buf:
-	freedqbuf(buf);
-	return ret;
-}
-
-/* Find entry for given id in the tree - wrapper function */
-static inline loff_t find_dqentry(struct dquot *dquot)
-{
-	return find_tree_dqentry(dquot, V2_DQTREEOFF, 0);
-}
-
-static int v2_read_dquot(struct dquot *dquot)
-{
-	int type = dquot->dq_type;
-	loff_t offset;
-	struct v2_disk_dqblk ddquot, empty;
-	int ret = 0;
-
-#ifdef __QUOTA_V2_PARANOIA
-	/* Invalidated quota? */
-	if (!dquot->dq_sb || !sb_dqopt(dquot->dq_sb)->files[type]) {
-		printk(KERN_ERR "VFS: Quota invalidated while reading!\n");
-		return -EIO;
-	}
-#endif
-	offset = find_dqentry(dquot);
-	if (offset <= 0) {	/* Entry not present? */
-		if (offset < 0)
-			printk(KERN_ERR "VFS: Can't read quota "
-			  "structure for id %u.\n", dquot->dq_id);
-		dquot->dq_off = 0;
-		set_bit(DQ_FAKE_B, &dquot->dq_flags);
-		memset(&dquot->dq_dqb, 0, sizeof(struct mem_dqblk));
-		ret = offset;
-	}
-	else {
-		dquot->dq_off = offset;
-		if ((ret = dquot->dq_sb->s_op->quota_read(dquot->dq_sb, type,
-		    (char *)&ddquot, sizeof(struct v2_disk_dqblk), offset))
-		    != sizeof(struct v2_disk_dqblk)) {
-			if (ret >= 0)
-				ret = -EIO;
-			printk(KERN_ERR "VFS: Error while reading quota "
-			  "structure for id %u.\n", dquot->dq_id);
-			memset(&ddquot, 0, sizeof(struct v2_disk_dqblk));
-		}
-		else {
-			ret = 0;
-			/* We need to escape back all-zero structure */
-			memset(&empty, 0, sizeof(struct v2_disk_dqblk));
-			empty.dqb_itime = cpu_to_le64(1);
-			if (!memcmp(&empty, &ddquot, sizeof(struct v2_disk_dqblk)))
-				ddquot.dqb_itime = 0;
-		}
-		disk2memdqb(&dquot->dq_dqb, &ddquot);
-		if (!dquot->dq_dqb.dqb_bhardlimit &&
-			!dquot->dq_dqb.dqb_bsoftlimit &&
-			!dquot->dq_dqb.dqb_ihardlimit &&
-			!dquot->dq_dqb.dqb_isoftlimit)
-			set_bit(DQ_FAKE_B, &dquot->dq_flags);
-	}
-	dqstats.reads++;
-
-	return ret;
-}
-
-/* Check whether dquot should not be deleted. We know we are
- * the only one operating on dquot (thanks to dq_lock) */
 static int v2_release_dquot(struct dquot *dquot)
 {
-	if (test_bit(DQ_FAKE_B, &dquot->dq_flags) && !(dquot->dq_dqb.dqb_curinodes | dquot->dq_dqb.dqb_curspace))
-		return v2_delete_dquot(dquot);
-	return 0;
+	return qtree_release_dquot(&sb_dqinfo(dquot->dq_sb, dquot->dq_type)->u.v2_i.i, dquot);
 }
 
 static struct quota_format_ops v2_format_ops = {

fs/quotaio_v2.h

@@ -21,6 +21,12 @@
 	0		/* GRPQUOTA */\
 }
 
+/* First generic header */
+struct v2_disk_dqheader {
+	__le32 dqh_magic;	/* Magic number identifying file */
+	__le32 dqh_version;	/* File version */
+};
+
 /*
  * The following structure defines the format of the disk quota file
  * (as it appears on disk) - the file is a radix tree whose leaves point
@@ -38,15 +44,6 @@ struct v2_disk_dqblk {
 	__le64 dqb_itime;	/* time limit for excessive inode use */
 };
 
-/*
- * Here are header structures as written on disk and their in-memory copies
- */
-/* First generic header */
-struct v2_disk_dqheader {
-	__le32 dqh_magic;	/* Magic number identifying file */
-	__le32 dqh_version;	/* File version */
-};
-
 /* Header with type and version specific information */
 struct v2_disk_dqinfo {
 	__le32 dqi_bgrace;	/* Time before block soft limit becomes hard limit */
@@ -57,23 +54,7 @@ struct v2_disk_dqinfo {
 	__le32 dqi_free_entry;	/* Number of block with at least one free entry */
 };
 
-/*
- *  Structure of header of block with quota structures. It is padded to 16 bytes so
- *  there will be space for exactly 21 quota-entries in a block
- */
-struct v2_disk_dqdbheader {
-	__le32 dqdh_next_free;	/* Number of next block with free entry */
-	__le32 dqdh_prev_free;	/* Number of previous block with free entry */
-	__le16 dqdh_entries;	/* Number of valid entries in block */
-	__le16 dqdh_pad1;
-	__le32 dqdh_pad2;
-};
-
 #define V2_DQINFOOFF	sizeof(struct v2_disk_dqheader)	/* Offset of info header in file */
-#define V2_DQBLKSIZE_BITS	10
-#define V2_DQBLKSIZE	(1 << V2_DQBLKSIZE_BITS)	/* Size of block with quota structures */
-#define V2_DQTREEOFF	1		/* Offset of tree in file in blocks */
-#define V2_DQTREEDEPTH	4		/* Depth of quota tree */
-#define V2_DQSTRINBLK	((V2_DQBLKSIZE - sizeof(struct v2_disk_dqdbheader)) / sizeof(struct v2_disk_dqblk))	/* Number of entries in one blocks */
+#define V2_DQBLKSIZE_BITS 10				/* Size of leaf block in tree */
 
 #endif /* _LINUX_QUOTAIO_V2_H */

include/linux/dqblk_qtree.h

@@ -0,0 +1,56 @@
+/*
+ *	Definitions of structures and functions for quota formats using trie
+ */
+
+#ifndef _LINUX_DQBLK_QTREE_H
+#define _LINUX_DQBLK_QTREE_H
+
+#include <linux/types.h>
+
+/* Numbers of blocks needed for updates - we count with the smallest
+ * possible block size (1024) */
+#define QTREE_INIT_ALLOC 4
+#define QTREE_INIT_REWRITE 2
+#define QTREE_DEL_ALLOC 0
+#define QTREE_DEL_REWRITE 6
+
+struct dquot;
+
+/* Operations */
+struct qtree_fmt_operations {
+	void (*mem2disk_dqblk)(void *disk, struct dquot *dquot);	/* Convert given entry from in memory format to disk one */
+	void (*disk2mem_dqblk)(struct dquot *dquot, void *disk);	/* Convert given entry from disk format to in memory one */
+	int (*is_id)(void *disk, struct dquot *dquot);	/* Is this structure for given id? */
+};
+
+/* Inmemory copy of version specific information */
+struct qtree_mem_dqinfo {
+	struct super_block *dqi_sb;	/* Sb quota is on */
+	int dqi_type;			/* Quota type */
+	unsigned int dqi_blocks;	/* # of blocks in quota file */
+	unsigned int dqi_free_blk;	/* First block in list of free blocks */
+	unsigned int dqi_free_entry;	/* First block with free entry */
+	unsigned int dqi_blocksize_bits;	/* Block size of quota file */
+	unsigned int dqi_entry_size;	/* Size of quota entry in quota file */
+	unsigned int dqi_usable_bs;	/* Space usable in block for quota data */
+	unsigned int dqi_qtree_depth;	/* Precomputed depth of quota tree */
+	struct qtree_fmt_operations *dqi_ops;	/* Operations for entry manipulation */
+};
+
+int qtree_write_dquot(struct qtree_mem_dqinfo *info, struct dquot *dquot);
+int qtree_read_dquot(struct qtree_mem_dqinfo *info, struct dquot *dquot);
+int qtree_delete_dquot(struct qtree_mem_dqinfo *info, struct dquot *dquot);
+int qtree_release_dquot(struct qtree_mem_dqinfo *info, struct dquot *dquot);
+int qtree_entry_unused(struct qtree_mem_dqinfo *info, char *disk);
+static inline int qtree_depth(struct qtree_mem_dqinfo *info)
+{
+	unsigned int epb = info->dqi_usable_bs >> 2;
+	unsigned long long entries = epb;
+	int i;
+
+	for (i = 1; entries < (1ULL << 32); i++)
+		entries *= epb;
+	return i;
+}
+
+#endif /* _LINUX_DQBLK_QTREE_H */

include/linux/dqblk_v2.h

@@ -1,26 +1,23 @@
 /*
- *	Definitions of structures for vfsv0 quota format
+ *  Definitions for vfsv0 quota format
  */
 
 #ifndef _LINUX_DQBLK_V2_H
 #define _LINUX_DQBLK_V2_H
 
-#include <linux/types.h>
+#include <linux/dqblk_qtree.h>
 
-/* id numbers of quota format */
+/* Id number of quota format */
 #define QFMT_VFS_V0 2
 
 /* Numbers of blocks needed for updates */
-#define V2_INIT_ALLOC 4
-#define V2_INIT_REWRITE 2
-#define V2_DEL_ALLOC 0
-#define V2_DEL_REWRITE 6
+#define V2_INIT_ALLOC QTREE_INIT_ALLOC
+#define V2_INIT_REWRITE QTREE_INIT_REWRITE
+#define V2_DEL_ALLOC QTREE_DEL_ALLOC
+#define V2_DEL_REWRITE QTREE_DEL_REWRITE
 
-/* Inmemory copy of version specific information */
 struct v2_mem_dqinfo {
-	unsigned int dqi_blocks;
-	unsigned int dqi_free_blk;
-	unsigned int dqi_free_entry;
+	struct qtree_mem_dqinfo i;
 };
 
 #endif /* _LINUX_DQBLK_V2_H */