[RFC] nilfs2: double bio_put() in nilfs_end_bio_write()

[RFC] nilfs2: double bio_put() in nilfs_end_bio_write()

[Dan Carpenter]

I don't know if this is correct or not, but if feels like we shouldn't
be calling bio_put() twice in nilfs_end_bio_write().  Also the comment
is out of date.  If this is correct then I can resend with a proper
Signed-off-by line and everything.

diff --git a/fs/nilfs2/segbuf.c b/fs/nilfs2/segbuf.c
index dc9a913..a660fd7 100644
--- a/fs/nilfs2/segbuf.c
+++ b/fs/nilfs2/segbuf.c
@@ -346,7 +346,8 @@ static void nilfs_end_bio_write(struct bio *bio, int err)
 	if (err = -EOPNOTSUPP) {
 		set_bit(BIO_EOPNOTSUPP, &bio->bi_flags);
 		bio_put(bio);
-		/* to be detected by submit_seg_bio() */
+		/* to be detected by nilfs_segbuf_submit_bio() */
+		return;
 	}
 
 	if (!uptodate)

Re: [RFC] nilfs2: double bio_put() in nilfs_end_bio_write()

[Vyacheslav Dubeyko]

Hi Dan,

On Jul 19, 2013, at 1:26 PM, Dan Carpenter wrote:

> I don't know if this is correct or not, but if feels like we shouldn't
> be calling bio_put() twice in nilfs_end_bio_write().  Also the comment
> is out of date.  If this is correct then I can resend with a proper
> Signed-off-by line and everything.
> 

I think that you have caught the important issue.

> diff --git a/fs/nilfs2/segbuf.c b/fs/nilfs2/segbuf.c
> index dc9a913..a660fd7 100644
> --- a/fs/nilfs2/segbuf.c
> +++ b/fs/nilfs2/segbuf.c
> @@ -346,7 +346,8 @@ static void nilfs_end_bio_write(struct bio *bio, int err)
> 	if (err = -EOPNOTSUPP) {
> 		set_bit(BIO_EOPNOTSUPP, &bio->bi_flags);
> 		bio_put(bio);
> -		/* to be detected by submit_seg_bio() */
> +		/* to be detected by nilfs_segbuf_submit_bio() */
> +		return;

I think that simple return from the function is not right way. As I understand the code,
then we increment error count in segbuf's sb_err field and signalize about completion of
operation. So, from my viewpoint, it needs to remove the bio_put() call for the case of
(err = -EOPNOTSUPP) instead of return from function.

Thanks,
Vyacheslav Dubeyko.

> 	}
> 
> 	if (!uptodate)
> --
> To unsubscribe from this list: send the line "unsubscribe linux-nilfs" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: [RFC] nilfs2: double bio_put() in nilfs_end_bio_write()

[Dan Carpenter]

On Sat, Jul 20, 2013 at 01:10:23PM +0400, Vyacheslav Dubeyko wrote:
> > diff --git a/fs/nilfs2/segbuf.c b/fs/nilfs2/segbuf.c
> > index dc9a913..a660fd7 100644
> > --- a/fs/nilfs2/segbuf.c
> > +++ b/fs/nilfs2/segbuf.c
> > @@ -346,7 +346,8 @@ static void nilfs_end_bio_write(struct bio *bio, int err)
> > 	if (err = -EOPNOTSUPP) {
> > 		set_bit(BIO_EOPNOTSUPP, &bio->bi_flags);
> > 		bio_put(bio);
> > -		/* to be detected by submit_seg_bio() */
> > +		/* to be detected by nilfs_segbuf_submit_bio() */
> > +		return;
> 
> I think that simple return from the function is not right way. As I understand the code,
> then we increment error count in segbuf's sb_err field and signalize about completion of
> operation. So, from my viewpoint, it needs to remove the bio_put() call for the case of
> (err = -EOPNOTSUPP) instead of return from function.
> 

Do you think you could send that patch and give me a reported-by
tag?  I feel weird signing off on it when I don't really understand
this code...

regards,
dan carpenter

Re: [RFC] nilfs2: double bio_put() in nilfs_end_bio_write()

[Vyacheslav Dubeyko]

On Jul 21, 2013, at 1:40 AM, Dan Carpenter wrote:

> On Sat, Jul 20, 2013 at 01:10:23PM +0400, Vyacheslav Dubeyko wrote:
>>> diff --git a/fs/nilfs2/segbuf.c b/fs/nilfs2/segbuf.c
>>> index dc9a913..a660fd7 100644
>>> --- a/fs/nilfs2/segbuf.c
>>> +++ b/fs/nilfs2/segbuf.c
>>> @@ -346,7 +346,8 @@ static void nilfs_end_bio_write(struct bio *bio, int err)
>>> 	if (err = -EOPNOTSUPP) {
>>> 		set_bit(BIO_EOPNOTSUPP, &bio->bi_flags);
>>> 		bio_put(bio);
>>> -		/* to be detected by submit_seg_bio() */
>>> +		/* to be detected by nilfs_segbuf_submit_bio() */
>>> +		return;
>> 
>> I think that simple return from the function is not right way. As I understand the code,
>> then we increment error count in segbuf's sb_err field and signalize about completion of
>> operation. So, from my viewpoint, it needs to remove the bio_put() call for the case of
>> (err = -EOPNOTSUPP) instead of return from function.
>> 
> 
> Do you think you could send that patch and give me a reported-by
> tag?  I feel weird signing off on it when I don't really understand
> this code...
> 

Ok, sure. I'll send the patch and continue discussion with Ryusuke
(if he will have objections).

Thank you for the report and fix suggestion.

With the best regards,
Vyacheslav Dubeyko.

> regards,
> dan carpenter