ext4: remove deprecated oldalloc
authorLukas Czerner <lczerner@redhat.com>
Sat, 8 Oct 2011 18:34:47 +0000 (14:34 -0400)
committerTheodore Ts'o <tytso@mit.edu>
Sat, 8 Oct 2011 18:34:47 +0000 (14:34 -0400)
For a long time now orlov is the default block allocator in the
ext4. It performs better than the old one and no one seems to claim
otherwise so we can safely drop it and make oldalloc and orlov mount
option deprecated.

This is a part of the effort to reduce number of ext4 options hence the
test matrix.

Signed-off-by: Lukas Czerner <lczerner@redhat.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
Documentation/filesystems/ext4.txt
fs/ext4/ext4.h
fs/ext4/ialloc.c
fs/ext4/super.c

index a530401..4917cf2 100644 (file)
@@ -203,14 +203,6 @@ inode_readahead_blks=n     This tuning parameter controls the maximum
                        table readahead algorithm will pre-read into
                        the buffer cache.  The default value is 32 blocks.
 
-orlov          (*)     This enables the new Orlov block allocator. It is
-                       enabled by default.
-
-oldalloc               This disables the Orlov block allocator and enables
-                       the old block allocator.  Orlov should have better
-                       performance - we'd like to get some feedback if it's
-                       the contrary for you.
-
 nouser_xattr           Disables Extended User Attributes. If you have extended
                        attribute support enabled in the kernel configuration
                        (CONFIG_EXT4_FS_XATTR), extended attribute support
index 1bbd2ca..61b0b5c 100644 (file)
@@ -913,7 +913,6 @@ struct ext4_inode_info {
 /*
  * Mount flags
  */
-#define EXT4_MOUNT_OLDALLOC            0x00002  /* Don't use the new Orlov allocator */
 #define EXT4_MOUNT_GRPID               0x00004 /* Create files with directory's group */
 #define EXT4_MOUNT_DEBUG               0x00008 /* Some debugging messages */
 #define EXT4_MOUNT_ERRORS_CONT         0x00010 /* Continue on errors */
index d50a7d5..0421998 100644 (file)
@@ -293,120 +293,6 @@ error_return:
        ext4_std_error(sb, fatal);
 }
 
-/*
- * There are two policies for allocating an inode.  If the new inode is
- * a directory, then a forward search is made for a block group with both
- * free space and a low directory-to-inode ratio; if that fails, then of
- * the groups with above-average free space, that group with the fewest
- * directories already is chosen.
- *
- * For other inodes, search forward from the parent directory\'s block
- * group to find a free inode.
- */
-static int find_group_dir(struct super_block *sb, struct inode *parent,
-                               ext4_group_t *best_group)
-{
-       ext4_group_t ngroups = ext4_get_groups_count(sb);
-       unsigned int freei, avefreei;
-       struct ext4_group_desc *desc, *best_desc = NULL;
-       ext4_group_t group;
-       int ret = -1;
-
-       freei = percpu_counter_read_positive(&EXT4_SB(sb)->s_freeinodes_counter);
-       avefreei = freei / ngroups;
-
-       for (group = 0; group < ngroups; group++) {
-               desc = ext4_get_group_desc(sb, group, NULL);
-               if (!desc || !ext4_free_inodes_count(sb, desc))
-                       continue;
-               if (ext4_free_inodes_count(sb, desc) < avefreei)
-                       continue;
-               if (!best_desc ||
-                   (ext4_free_group_clusters(sb, desc) >
-                    ext4_free_group_clusters(sb, best_desc))) {
-                       *best_group = group;
-                       best_desc = desc;
-                       ret = 0;
-               }
-       }
-       return ret;
-}
-
-#define free_block_ratio 10
-
-static int find_group_flex(struct super_block *sb, struct inode *parent,
-                          ext4_group_t *best_group)
-{
-       struct ext4_sb_info *sbi = EXT4_SB(sb);
-       struct ext4_group_desc *desc;
-       struct flex_groups *flex_group = sbi->s_flex_groups;
-       ext4_group_t parent_group = EXT4_I(parent)->i_block_group;
-       ext4_group_t parent_fbg_group = ext4_flex_group(sbi, parent_group);
-       ext4_group_t ngroups = ext4_get_groups_count(sb);
-       int flex_size = ext4_flex_bg_size(sbi);
-       ext4_group_t best_flex = parent_fbg_group;
-       int blocks_per_flex = sbi->s_blocks_per_group * flex_size;
-       int flexbg_free_clusters;
-       int flex_freeb_ratio;
-       ext4_group_t n_fbg_groups;
-       ext4_group_t i;
-
-       n_fbg_groups = (ngroups + flex_size - 1) >>
-               sbi->s_log_groups_per_flex;
-
-find_close_to_parent:
-       flexbg_free_clusters = atomic_read(&flex_group[best_flex].free_clusters);
-       flex_freeb_ratio = EXT4_C2B(sbi, flexbg_free_clusters) * 100 /
-               blocks_per_flex;
-       if (atomic_read(&flex_group[best_flex].free_inodes) &&
-           flex_freeb_ratio > free_block_ratio)
-               goto found_flexbg;
-
-       if (best_flex && best_flex == parent_fbg_group) {
-               best_flex--;
-               goto find_close_to_parent;
-       }
-
-       for (i = 0; i < n_fbg_groups; i++) {
-               if (i == parent_fbg_group || i == parent_fbg_group - 1)
-                       continue;
-
-               flexbg_free_clusters = atomic_read(&flex_group[i].free_clusters);
-               flex_freeb_ratio = EXT4_C2B(sbi, flexbg_free_clusters) * 100 /
-                       blocks_per_flex;
-
-               if (flex_freeb_ratio > free_block_ratio &&
-                   (atomic_read(&flex_group[i].free_inodes))) {
-                       best_flex = i;
-                       goto found_flexbg;
-               }
-
-               if ((atomic_read(&flex_group[best_flex].free_inodes) == 0) ||
-                   ((atomic_read(&flex_group[i].free_clusters) >
-                     atomic_read(&flex_group[best_flex].free_clusters)) &&
-                    atomic_read(&flex_group[i].free_inodes)))
-                       best_flex = i;
-       }
-
-       if (!atomic_read(&flex_group[best_flex].free_inodes) ||
-           !atomic_read(&flex_group[best_flex].free_clusters))
-               return -1;
-
-found_flexbg:
-       for (i = best_flex * flex_size; i < ngroups &&
-                    i < (best_flex + 1) * flex_size; i++) {
-               desc = ext4_get_group_desc(sb, i, NULL);
-               if (ext4_free_inodes_count(sb, desc)) {
-                       *best_group = i;
-                       goto out;
-               }
-       }
-
-       return -1;
-out:
-       return 0;
-}
-
 struct orlov_stats {
        __u32 free_inodes;
        __u32 free_clusters;
@@ -819,7 +705,6 @@ struct inode *ext4_new_inode(handle_t *handle, struct inode *dir, int mode,
        int ret2, err = 0;
        struct inode *ret;
        ext4_group_t i;
-       static int once = 1;
        ext4_group_t flex_group;
 
        /* Cannot create files in a deleted directory */
@@ -845,26 +730,9 @@ struct inode *ext4_new_inode(handle_t *handle, struct inode *dir, int mode,
                goto got_group;
        }
 
-       if (sbi->s_log_groups_per_flex && test_opt(sb, OLDALLOC)) {
-               ret2 = find_group_flex(sb, dir, &group);
-               if (ret2 == -1) {
-                       ret2 = find_group_other(sb, dir, &group, mode);
-                       if (ret2 == 0 && once) {
-                               once = 0;
-                               printk(KERN_NOTICE "ext4: find_group_flex "
-                                      "failed, fallback succeeded dir %lu\n",
-                                      dir->i_ino);
-                       }
-               }
-               goto got_group;
-       }
-
-       if (S_ISDIR(mode)) {
-               if (test_opt(sb, OLDALLOC))
-                       ret2 = find_group_dir(sb, dir, &group);
-               else
-                       ret2 = find_group_orlov(sb, dir, &group, mode, qstr);
-       } else
+       if (S_ISDIR(mode))
+               ret2 = find_group_orlov(sb, dir, &group, mode, qstr);
+       else
                ret2 = find_group_other(sb, dir, &group, mode);
 
 got_group:
index db2cd3f..42f76c6 100644 (file)
@@ -1073,8 +1073,6 @@ static int ext4_show_options(struct seq_file *seq, struct vfsmount *vfs)
                seq_puts(seq, ",nouid32");
        if (test_opt(sb, DEBUG) && !(def_mount_opts & EXT4_DEFM_DEBUG))
                seq_puts(seq, ",debug");
-       if (test_opt(sb, OLDALLOC))
-               seq_puts(seq, ",oldalloc");
 #ifdef CONFIG_EXT4_FS_XATTR
        if (test_opt(sb, XATTR_USER))
                seq_puts(seq, ",user_xattr");
@@ -1583,10 +1581,12 @@ static int parse_options(char *options, struct super_block *sb,
                        set_opt(sb, DEBUG);
                        break;
                case Opt_oldalloc:
-                       set_opt(sb, OLDALLOC);
+                       ext4_msg(sb, KERN_WARNING,
+                                "Ignoring deprecated oldalloc option");
                        break;
                case Opt_orlov:
-                       clear_opt(sb, OLDALLOC);
+                       ext4_msg(sb, KERN_WARNING,
+                                "Ignoring deprecated orlov option");
                        break;
 #ifdef CONFIG_EXT4_FS_XATTR
                case Opt_user_xattr: