On Wed 08-05-19 14:38:22, sunny.s.zhang wrote:
> Hi Chengguang,
>
> 在 2019年05月05日 19:01, Chengguang Xu 写道:
> > When fail from creating cache jbd2_inode_cache, we will
> > destroy previously created cache jbd2_handle_cache twice.
> > This patch fixes it by removing first destroy in error path.
> >
> > Signed-off-by: Chengguang Xu <cgxu519@xxxxxxxxx>
> > ---
> > fs/jbd2/journal.c | 1 -
> > 1 file changed, 1 deletion(-)
> >
> > diff --git a/fs/jbd2/journal.c b/fs/jbd2/journal.c
> > index 382c030cc78b..49797854ccb8 100644
> > --- a/fs/jbd2/journal.c
> > +++ b/fs/jbd2/journal.c
> > @@ -2642,7 +2642,6 @@ static int __init jbd2_journal_init_handle_cache(void)
> > jbd2_inode_cache = KMEM_CACHE(jbd2_inode, 0);
> > if (jbd2_inode_cache == NULL) {
> > printk(KERN_EMERG "JBD2: failed to create inode cache\n");
> > - kmem_cache_destroy(jbd2_handle_cache);
> Maybe we should keep it, and set the jbd2_handle_cache to NULL.
> If there are some changes in the future, we may forget to change the
> function
> of jbd2_journal_destroy_handle_cache.
So what I'd do is that I'd split initialization of jbd2_inode_cache into a
separate function (and the same for destruction). That more aligns with how
things are currently done in jbd2 and also fixes the problem with double
destruction.
Honza
--
Jan Kara <jack@xxxxxxxx>
SUSE Labs, CR
