A possible patch for a dangling pointer issue of librbd engine

A possible patch for a dangling pointer issue of librbd engine

[xan.peng]

I'm interested in the librbd engine, and I possibly found a bug which
you can see from the following patch.

And you can see more information at https://github.com/axboe/fio/pull/10.
(BTW, are github pull requests the correct way to submit patches to fio?)

Signed-off-by: Xan Peng xanpeng@gmail.com
---
diff --git a/engines/rbd.c b/engines/rbd.c
index ff35373..38528c2 100644
--- a/engines/rbd.c
+++ b/engines/rbd.c
@@ -131,8 +131,10 @@ static int _fio_rbd_connect(struct thread_data *td)

 failed_open:
        rados_ioctx_destroy(rbd_data->io_ctx);
+       rbd_data->io_ctx = NULL;
 failed_shutdown:
        rados_shutdown(rbd_data->cluster);
+       rbd_data->cluster = NULL;
 failed_early:
        return 1;
 }
@@ -354,9 +356,8 @@ static int fio_rbd_setup(struct thread_data *td)
         */
        td->o.use_thread = 1;

-       /* connect in the main thread to determine to determine
-        * the size of the given RADOS block device. And disconnect
-        * later on.
+       /* connect in the main thread to determine the size of the given RADOS
+        * block device. And disconnect later on.
         */
        r = _fio_rbd_connect(td);
        if (r) {

Re: A possible patch for a dangling pointer issue of librbd engine

[Jens Axboe]

On 05/18/2014 07:58 PM, xan.peng wrote:
> I'm interested in the librbd engine, and I possibly found a bug which
> you can see from the following patch.
> 
> And you can see more information at https://github.com/axboe/fio/pull/10.
> (BTW, are github pull requests the correct way to submit patches to fio?)

Sure, you can do that, but email me+list about them separately as I
might not see them on github. So I'm not going to pull your branch,
since you have two patches on top of the interesting one, where the 2nd
just reverts the other one. I'm apply this one separately.


-- 
Jens Axboe

Re: A possible patch for a dangling pointer issue of librbd engine

[xan.peng]

On Mon, May 19, 2014 at 11:37 PM, Jens Axboe <axboe@kernel.dk> wrote:
> Sure, you can do that, but email me+list about them separately as I
> might not see them on github. So I'm not going to pull your branch,
> since you have two patches on top of the interesting one, where the 2nd
> just reverts the other one. I'm apply this one separately.

Got it.
p.s. I reverted the 2nd commit as it's of a different topic, and it
was automatically merged into the 1st by GitHub before I noticed.