Re: [PATCH] ext4: fix error handling in ext4_create_journal

[Borislav Petkov <bbpetkov@yahoo.de>]

On Thu, Jul 05, 2007 at 12:17:25AM +0400, Alexey Dobriyan wrote:
> On Wed, Jul 04, 2007 at 09:30:24PM +0200, Borislav Petkov wrote:
> > Fix error handling in ext4_create_journal according to kernel conventions.
> 
> > --- linux-2.6.22-rc7/fs/ext4/super.c.orig
> > +++ linux-2.6.22-rc7/fs/ext4/super.c
> > @@ -2150,6 +2150,7 @@
> >  			       unsigned int journal_inum)
> >  {
> >  	journal_t *journal;
> > +	int err;
> >  
> >  	if (sb->s_flags & MS_RDONLY) {
> >  		printk(KERN_ERR "EXT4-fs: readonly filesystem when trying to "
> > @@ -2157,13 +2158,15 @@
> >  		return -EROFS;
> >  	}
> >  
> > -	if (!(journal = ext4_get_journal(sb, journal_inum)))
> > +	journal = ext4_get_journal(sb, journal_inum);
> > +	if (!journal)
> >  		return -EINVAL;
> 
> OK.
> 
> >  	printk(KERN_INFO "EXT4-fs: creating new journal on inode %u\n",
> >  	       journal_inum);
> >  
> > -	if (jbd2_journal_create(journal)) {
> > +	err = jbd2_journal_create(journal);
> > +	if (err) {
> >  		printk(KERN_ERR "EXT4-fs: error creating journal.\n");
> >  		jbd2_journal_destroy(journal);
> >  		return -EIO;
> 
> Original code is fine.

Hmm, ok but this is not the way error handling is done in the rest of ext3/4.
If you look for error variables declarations of type int in super.c, e.g. in
ext4_load_journal(), ext4_clear_journal_err() and ext4_remount() to name a few,
you'll see how the return value of the called function is assigned to the
error variable and then the last is checked in the if-conditional, which is the
usual way error handling is done in the kernel, imho. And my patch addresses exactly that.

-- 
Regards/Gruß,
    Boris.