Re: [PATCH 3/3] relay: Add buffer-only channels; useful for early logging.

[Eduard - Gabriel Munteanu <eduard.munteanu@linux360.ro>]

On Sun, 22 Jun 2008 22:29:53 +0300
"Pekka Enberg" <penberg@cs.helsinki.fi> wrote:

> On Sat, Jun 21, 2008 at 5:11 PM, Eduard - Gabriel Munteanu
> <eduard.munteanu@linux360.ro> wrote:
> > +       tmpname = kzalloc(NAME_MAX + 1, GFP_KERNEL);
> > +       if (!tmpname)
> > +               goto failed;
> 
> Why don't you return -ENOMEM here?

The original function from which I ripped this thing off
(relay_open_buf()) returned 1 on error, so I assumed I wasn't supposed
to change this behavior.

I should probably do this in a separate patch.

> The separate goto seems pointless. Why don't you use negative return
> values for error handling?

Yes, it's better to use 'return' directly.

> >  /**
> > + *     relay_late_setup_files - triggers file creation
> > + *     @chan: channel to operate on
> > + *     @base_filename: base name of files to create
> > + *     @parent: dentry of parent directory, %NULL for root
> > directory
> > + *
> > + *     Returns 0 if successful, non-zero otherwise.
> > + *
> > + *     Use to setup files for a previously buffer-only channel.
> > + *     Useful to do early tracing in kernel, before VFS is up, for
> > example.
> > + */
> > +int relay_late_setup_files(struct rchan *chan,
> > +                          const char *base_filename,
> > +                          struct dentry *parent)
> > +{
> > +       unsigned int i;
> > +       int err;
> > +       struct rchan_buf *buf;
> > +
> > +       if (!chan || !base_filename)
> > +               return 1;
> 
> -EINVAL?

Will do and resubmit. Being new code (called directly by other kernel
users), it allows me to be a bit inconsistent about return values.

> > +               err = relay_setup_buf_file(chan, buf, i);
> > +               if (unlikely(err))
> > +                       goto out;
> > +       }
> > +       mutex_unlock(&relay_channels_mutex);
> > +
> > +       return 0;
> > +
> > +out:
> > +       mutex_unlock(&relay_channels_mutex);
> > +       return 1;
> 
> You don't need two return statements (and coincidentally two unlocks)
> if you keep the return value in a local variable 'err'.

Thanks, I missed this possibility. Sounds better.
 

	Eduard