2008-01-23 13:55:24

by Theodore Ts'o

Subject: ext4-online-defrag-free-space-fragmentation.patch uses do_fsync()

I was trying to build ext4 as a module, and ran into problems because
the online defrag patch is calling do_fsync() which is *not* an exported
symbol, and so can not be called from a module.

Looking at what the routine is doing, there's no reason to call
do_fsync(), and in fact depending on the journaling mode in use, it may
not force a journal commit, which seems to be the goal of the code.

Hence, I plan to merge the following fix into the the
defrag-free-space-fragmentation patch, unless there are any objections
from Takashi-San or Akira-San.

Regards,

- Ted

diff --git a/fs/ext4/defrag.c b/fs/ext4/defrag.c
index 4ef3dc0..19d2cfd 100644
--- a/fs/ext4/defrag.c
+++ b/fs/ext4/defrag.c
@@ -632,8 +632,9 @@ static int ext4_ext_defrag_victim(struct file *target_filp,
}

/* Sync journal blocks before reservation */
- if (do_fsync(target_filp, 0)) {
- printk(KERN_ERR "defrag: failed do_fsync\n");
+ ret = ext4_force_commit(sb);
+ if (ret) {
+ printk(KERN_ERR "defrag: failed do_fsync (%d)\n", ret);
goto ERR;
}
}

2008-01-23 14:10:31

by Dave Kleikamp

Subject: Re: ext4-online-defrag-free-space-fragmentation.patch uses do_fsync()

> diff --git a/fs/ext4/defrag.c b/fs/ext4/defrag.c
> index 4ef3dc0..19d2cfd 100644
> --- a/fs/ext4/defrag.c
> +++ b/fs/ext4/defrag.c
> @@ -632,8 +632,9 @@ static int ext4_ext_defrag_victim(struct file *target_filp,
> }
>
> /* Sync journal blocks before reservation */
> - if (do_fsync(target_filp, 0)) {
> - printk(KERN_ERR "defrag: failed do_fsync\n");
> + ret = ext4_force_commit(sb);
> + if (ret) {
> + printk(KERN_ERR "defrag: failed do_fsync (%d)\n", ret);

I'd think you'd want to change the printk text as well. "defrag: failed
ext4_force_commit (%d)\n" maybe?

> goto ERR;
> }
> }

--
David Kleikamp
IBM Linux Technology Center

2008-01-23 14:18:55

by Theodore Ts'o

Subject: Re: ext4-online-defrag-free-space-fragmentation.patch uses do_fsync()

On Wed, Jan 23, 2008 at 08:10:20AM -0600, Dave Kleikamp wrote:
> I'd think you'd want to change the printk text as well. "defrag: failed
> ext4_force_commit (%d)\n" maybe?

mmm, good catch, thanks.

- Ted

2008-01-24 03:01:56

by Takashi Sato

Subject: Re: ext4-online-defrag-free-space-fragmentation.patch uses do_fsync()

Hi,

> I was trying to build ext4 as a module, and ran into problems because
> the online defrag patch is calling do_fsync() which is *not* an exported
> symbol, and so can not be called from a module.
>
> Looking at what the routine is doing, there's no reason to call
> do_fsync(), and in fact depending on the journaling mode in use, it may
> not force a journal commit, which seems to be the goal of the code.
>
> Hence, I plan to merge the following fix into the the
> defrag-free-space-fragmentation patch, unless there are any objections
> from Takashi-San or Akira-San.

Thank you for your information and the proposition of the fix.
Your fix is correct, please merge it.

> Regards,
>
> - Ted
>
> diff --git a/fs/ext4/defrag.c b/fs/ext4/defrag.c
> index 4ef3dc0..19d2cfd 100644
> --- a/fs/ext4/defrag.c
> +++ b/fs/ext4/defrag.c
> @@ -632,8 +632,9 @@ static int ext4_ext_defrag_victim(struct file *target_filp,
> }
>
> /* Sync journal blocks before reservation */
> - if (do_fsync(target_filp, 0)) {
> - printk(KERN_ERR "defrag: failed do_fsync\n");
> + ret = ext4_force_commit(sb);
> + if (ret) {
> + printk(KERN_ERR "defrag: failed do_fsync (%d)\n", ret);
> goto ERR;
> }
> }
> -
> To unsubscribe from this list: send the line "unsubscribe linux-ext4" in
> the body of a message to [email protected]
> More majordomo info at http://vger.kernel.org/majordomo-info.html

Cheers, Takashi