ext4: fix the mount option "init_itable=n" to work as expected for n=0
authorLukas Czerner <lczerner@redhat.com>
Fri, 20 May 2011 17:55:16 +0000 (13:55 -0400)
committerTheodore Ts'o <tytso@mit.edu>
Fri, 20 May 2011 17:55:16 +0000 (13:55 -0400)
For some reason, when we set the mount option "init_itable=0" it
behaves as we would set init_itable=20 which is not right at all.
Basically when we set it to zero we are saying to lazyinit thread not
to wait between zeroing the inode table (except of cond_resched()) so
this commit fixes that and removes the unnecessary condition.  The 'n'
should be also properly used on remount.

When the n is not set at all, it means that the default miltiplier
EXT4_DEF_LI_WAIT_MULT is set instead.

Signed-off-by: Lukas Czerner <lczerner@redhat.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
Reported-by: Eric Sandeen <sandeen@redhat.com>
fs/ext4/super.c

index d7756bd2dc128af02e0b0a5a26c9ca114cc3618f..fd51a4a266b65c8101567e480ed47f090c653e50 100644 (file)
@@ -1112,7 +1112,7 @@ static int ext4_show_options(struct seq_file *seq, struct vfsmount *vfs)
 
        if (!test_opt(sb, INIT_INODE_TABLE))
                seq_puts(seq, ",noinit_inode_table");
-       else if (sbi->s_li_wait_mult)
+       else if (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)
                seq_printf(seq, ",init_inode_table=%u",
                           (unsigned) sbi->s_li_wait_mult);
 
@@ -2704,11 +2704,8 @@ static int ext4_run_li_request(struct ext4_li_request *elr)
                ret = ext4_init_inode_table(sb, group,
                                            elr->lr_timeout ? 0 : 1);
                if (elr->lr_timeout == 0) {
-                       timeout = jiffies - timeout;
-                       if (elr->lr_sbi->s_li_wait_mult)
-                               timeout *= elr->lr_sbi->s_li_wait_mult;
-                       else
-                               timeout *= 20;
+                       timeout = (jiffies - timeout) *
+                                 elr->lr_sbi->s_li_wait_mult;
                        elr->lr_timeout = timeout;
                }
                elr->lr_next_sched = jiffies + elr->lr_timeout;
@@ -2945,8 +2942,14 @@ static int ext4_register_li_request(struct super_block *sb,
        ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
        int ret = 0;
 
-       if (sbi->s_li_request != NULL)
+       if (sbi->s_li_request != NULL) {
+               /*
+                * Reset timeout so it can be computed again, because
+                * s_li_wait_mult might have changed.
+                */
+               sbi->s_li_request->lr_timeout = 0;
                return 0;
+       }
 
        if (first_not_zeroed == ngroups ||
            (sb->s_flags & MS_RDONLY) ||
@@ -3144,6 +3147,12 @@ static int ext4_fill_super(struct super_block *sb, void *data, int silent)
            ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
                set_opt(sb, DELALLOC);
 
+       /*
+        * set default s_li_wait_mult for lazyinit, for the case there is
+        * no mount option specified.
+        */
+       sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
+
        if (!parse_options((char *) sbi->s_es->s_mount_opts, sb,
                           &journal_devnum, &journal_ioprio, NULL, 0)) {
                ext4_msg(sb, KERN_WARNING,