ext4: Use a hash of the topdir directory name for the Orlov parent group

Instead of using a random number to determine the goal parent grop for
the Orlov top directories, use a hash of the directory name.  This
allows for repeatable results when trying to benchmark filesystem
layout algorithms.

Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>

diff --git a/fs/ext4/ext4.h b/fs/ext4/ext4.h
index 06ee5a5..d035cf1 100644 (file)
--- a/fs/ext4/ext4.h
+++ b/fs/ext4/ext4.h
@@ -1315,7 +1315,8 @@ extern int ext4fs_dirhash(const char *name, int len, struct
                          dx_hash_info *hinfo);
 
 /* ialloc.c */
-extern struct inode * ext4_new_inode(handle_t *, struct inode *, int);
+extern struct inode *ext4_new_inode(handle_t *, struct inode *, int,
+                                   const struct qstr *qstr);
 extern void ext4_free_inode(handle_t *, struct inode *);
 extern struct inode * ext4_orphan_get(struct super_block *, unsigned long);
 extern unsigned long ext4_count_free_inodes(struct super_block *);

diff --git a/fs/ext4/ialloc.c b/fs/ext4/ialloc.c
index 7d502f3..3f98ee7 100644 (file)
--- a/fs/ext4/ialloc.c
+++ b/fs/ext4/ialloc.c
@@ -470,7 +470,8 @@ void get_orlov_stats(struct super_block *sb, ext4_group_t g,
  */
 
 static int find_group_orlov(struct super_block *sb, struct inode *parent,
-                           ext4_group_t *group, int mode)
+                           ext4_group_t *group, int mode,
+                           const struct qstr *qstr)
 {
        ext4_group_t parent_group = EXT4_I(parent)->i_block_group;
        struct ext4_sb_info *sbi = EXT4_SB(sb);
@@ -485,6 +486,7 @@ static int find_group_orlov(struct super_block *sb, struct inode *parent,
        struct ext4_group_desc *desc;
        struct orlov_stats stats;
        int flex_size = ext4_flex_bg_size(sbi);
+       struct dx_hash_info hinfo;
 
        ngroups = real_ngroups;
        if (flex_size > 1) {
@@ -506,7 +508,13 @@ static int find_group_orlov(struct super_block *sb, struct inode *parent,
                int best_ndir = inodes_per_group;
                int ret = -1;
 
-               get_random_bytes(&grp, sizeof(grp));
+               if (qstr) {
+                       hinfo.hash_version = DX_HASH_HALF_MD4;
+                       hinfo.seed = sbi->s_hash_seed;
+                       ext4fs_dirhash(qstr->name, qstr->len, &hinfo);
+                       grp = hinfo.hash;
+               } else
+                       get_random_bytes(&grp, sizeof(grp));
                parent_group = (unsigned)grp % ngroups;
                for (i = 0; i < ngroups; i++) {
                        g = (parent_group + i) % ngroups;
@@ -649,7 +657,7 @@ static int find_group_other(struct super_block *sb, struct inode *parent,
                *group = parent_group + flex_size;
                if (*group > ngroups)
                        *group = 0;
-               return find_group_orlov(sb, parent, group, mode);
+               return find_group_orlov(sb, parent, group, mode, 0);
        }
 
        /*
@@ -790,7 +798,8 @@ err_ret:
  * For other inodes, search forward from the parent directory's block
  * group to find a free inode.
  */
-struct inode *ext4_new_inode(handle_t *handle, struct inode *dir, int mode)
+struct inode *ext4_new_inode(handle_t *handle, struct inode *dir, int mode,
+                            const struct qstr *qstr)
 {
        struct super_block *sb;
        struct buffer_head *inode_bitmap_bh = NULL;
@@ -839,7 +848,7 @@ struct inode *ext4_new_inode(handle_t *handle, struct inode *dir, int mode)
                if (test_opt(sb, OLDALLOC))
                        ret2 = find_group_dir(sb, dir, &group);
                else
-                       ret2 = find_group_orlov(sb, dir, &group, mode);
+                       ret2 = find_group_orlov(sb, dir, &group, mode, qstr);
        } else
                ret2 = find_group_other(sb, dir, &group, mode);
 

diff --git a/fs/ext4/migrate.c b/fs/ext4/migrate.c
index fe64d9f..80d075b 100644 (file)
--- a/fs/ext4/migrate.c
+++ b/fs/ext4/migrate.c
@@ -483,9 +483,8 @@ int ext4_ext_migrate(struct inode *inode)
                retval = PTR_ERR(handle);
                return retval;
        }
-       tmp_inode = ext4_new_inode(handle,
-                               inode->i_sb->s_root->d_inode,
-                               S_IFREG);
+       tmp_inode = ext4_new_inode(handle, inode->i_sb->s_root->d_inode,
+                                  S_IFREG, 0);
        if (IS_ERR(tmp_inode)) {
                retval = -ENOMEM;
                ext4_journal_stop(handle);

diff --git a/fs/ext4/namei.c b/fs/ext4/namei.c
index 07eb664..5f00d24 100644 (file)
--- a/fs/ext4/namei.c
+++ b/fs/ext4/namei.c
@@ -1782,7 +1782,7 @@ retry:
        if (IS_DIRSYNC(dir))
                ext4_handle_sync(handle);
 
-       inode = ext4_new_inode (handle, dir, mode);
+       inode = ext4_new_inode(handle, dir, mode, &dentry->d_name);
        err = PTR_ERR(inode);
        if (!IS_ERR(inode)) {
                inode->i_op = &ext4_file_inode_operations;
@@ -1816,7 +1816,7 @@ retry:
        if (IS_DIRSYNC(dir))
                ext4_handle_sync(handle);
 
-       inode = ext4_new_inode(handle, dir, mode);
+       inode = ext4_new_inode(handle, dir, mode, &dentry->d_name);
        err = PTR_ERR(inode);
        if (!IS_ERR(inode)) {
                init_special_inode(inode, inode->i_mode, rdev);
@@ -1853,7 +1853,7 @@ retry:
        if (IS_DIRSYNC(dir))
                ext4_handle_sync(handle);
 
-       inode = ext4_new_inode(handle, dir, S_IFDIR | mode);
+       inode = ext4_new_inode(handle, dir, S_IFDIR | mode, &dentry->d_name);
        err = PTR_ERR(inode);
        if (IS_ERR(inode))
                goto out_stop;
@@ -2264,7 +2264,7 @@ retry:
        if (IS_DIRSYNC(dir))
                ext4_handle_sync(handle);
 
-       inode = ext4_new_inode(handle, dir, S_IFLNK|S_IRWXUGO);
+       inode = ext4_new_inode(handle, dir, S_IFLNK|S_IRWXUGO, &dentry->d_name);
        err = PTR_ERR(inode);
        if (IS_ERR(inode))
                goto out_stop;