linux-kernel - Re: [PATCH] of: Fix issue where code would fall through to error case.

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Date:   Tue, 29 Nov 2016 07:55:17 -0800
From:   Moritz Fischer <moritz.fischer@...us.com>
To:     Rob Herring <robh+dt@...nel.org>
Cc:     Frank Rowand <frowand.list@...il.com>,
        Moritz Fischer <moritz.fischer.private@...il.com>,
        "linux-kernel@...r.kernel.org" <linux-kernel@...r.kernel.org>,
        Pantelis Antoniou <pantelis.antoniou@...sulko.com>,
        moritz@...e-entropy.org,
        "devicetree@...r.kernel.org" <devicetree@...r.kernel.org>
Subject: Re: [PATCH] of: Fix issue where code would fall through to error
 case.

Hi Rob,

On Tue, Nov 29, 2016 at 09:06:08AM -0600, Rob Herring wrote:
> On Mon, Nov 28, 2016 at 9:30 AM, Frank Rowand <frowand.list@...il.com> wrote:
> > On 11/26/16 13:39, Frank Rowand wrote:
> >> On 11/23/16 13:58, Rob Herring wrote:
> >>> On Thu, Nov 17, 2016 at 6:10 PM, Moritz Fischer
> >>> <moritz.fischer.private@...il.com> wrote:
> >>>> On Thu, Nov 17, 2016 at 4:02 PM, Frank Rowand <frowand.list@...il.com> wrote:
> >>>>> On 11/17/16 15:40, Frank Rowand wrote:
> >>>>>> On 11/17/16 15:25, Moritz Fischer wrote:
> >>>>>>> No longer fall through into the error case that prints out
> >>>>>>> an error if no error (err = 0) occurred.
> >>>>>>>
> >>>>>>> Fixes d9181b20a83(of: Add back an error message, restructured)
> >>>>>>> Signed-off-by: Moritz Fischer <moritz.fischer@...us.com>
> >>>>>>> ---
> >>>>>>>  drivers/of/resolver.c | 6 +++++-
> >>>>>>>  1 file changed, 5 insertions(+), 1 deletion(-)
> >>>>>>>
> >>>>>>> diff --git a/drivers/of/resolver.c b/drivers/of/resolver.c
> >>>>>>> index 783bd09..785076d 100644
> >>>>>>> --- a/drivers/of/resolver.c
> >>>>>>> +++ b/drivers/of/resolver.c
> >>>>>>> @@ -358,9 +358,13 @@ int of_resolve_phandles(struct device_node *overlay)
> >>>>>>>
> >>>>>>>              err = update_usages_of_a_phandle_reference(overlay, prop, phandle);
> >>>>>>>              if (err)
> >>>>>>> -                    break;
> >>>>>>> +                    goto err_out;
> >>>>>>>      }
> >>>>>>>
> >>>>>>> +    of_node_put(tree_symbols);
> >>>>>>> +
> >>>>>>> +    return 0;
> >>>>>>> +
> >>>>>>>  err_out:
> >>>>>>>      pr_err("overlay phandle fixup failed: %d\n", err);
> >>>>>>>  out:
> >>>>>>
> >>>>>> Thanks for catching that.
> >>>>>>
> >>>>>> Rob, please apply.
> >>>>>>
> >>>>>> Reviewed-by: Frank Rowand <frank.rowand@...sony.com>
> >>>>>>
> >>>>>> -Frank
> >>>>>
> >>>>> On second thought, isn't the common pattern when clean up is needed for
> >>>>> both the no-error path and the error path something like:
> >>>>>
> >>>>>
> >>>>>         out:
> >>>>>                 of_node_put(tree_symbols);
> >>>>>                 return err;
> >>>>>
> >>>>>         err_out:
> >>>>>                 pr_err("overlay phandle fixup failed: %d\n", err);
> >>>>>                 goto out;
> >>>>>         }
> >>>>>
> >>>>>
> >>>>> I don't have a strong opinion, whatever Rob wants to take is fine with me.
> >>>>
> >>>> Same here. I tried to avoid the jumping back part, but if that's the
> >>>> common pattern,
> >>>> I can submit a v2 doing that instead.
> >>>
> >>> Both are ugly. Just do:
> >>>
> >>> if (err)
> >>>   pr_err(...);
> >>>
> >>> Rob
> >>
> >> Agreed.  Thanks for the touch of sanity Rob.
> >>
> >> -Frank
> >
> > I succumbed to looking only at the few lines of code above and not the
> > fuller context of the file that the patch applies to.
> >
> > The proposed patch was fixing the problem that a normal completion
> > of the for loop was falling through into the err_out label.  So what
> > looks cleaner ("if (err) pr_err(...)") is actually not correct.
> 
> What!? The *only* problem was printing the error message in the err=0
> case. All that needs to be fixed is not doing that. If we do that,
> then we really only need 1 goto label.

I think you're right. Can you look at my v3 that I sent. I also tried to
fix cases where we can just do

return 0;

vs.

err = 0;
goto err

...

err:
	of_node_put(NULL /*tree_symbols is NULL*/);
	return err;

Thanks,

Moritz