linux-kernel - Re: [PATCH] staging: r8188eu: Remove goto to no-op exit label

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:   Sat, 02 Apr 2022 12:13:20 +0200
From:   "Fabio M. De Francesco" <fmdefrancesco@...il.com>
To:     Ira Weiny <ira.weiny@...el.com>,
        Greg Kroah-Hartman <gregkh@...uxfoundation.org>
Cc:     Larry Finger <Larry.Finger@...inger.net>,
        Phillip Potter <phil@...lpotter.co.uk>,
        linux-staging@...ts.linux.dev, linux-kernel@...r.kernel.org,
        outreachy@...ts.linux.dev
Subject: Re: [PATCH] staging: r8188eu: Remove goto to no-op exit label

On venerd? 1 aprile 2022 22:41:40 CEST Ira Weiny wrote:
> On Fri, Apr 01, 2022 at 08:35:13PM +0200, Fabio M. De Francesco wrote:
> > In function rtw_free_netdev() there are two "goto" jumps to a no-op exit
> > label called "RETURN". Remove the label and return in line.
> 
> Thanks for the patch!  However, A good commit message lists the why and what of
> a change.  I don't see a why for this commit?

Yes I forgot the "why" :(
I'll rework the commit message for v2.

> 
> FWIW (For what it's worth) I know of a couple of good reasons for this change
> but you should get in the habit of putting that in the commit message.  Even
> for obvious things like this.
> 
> Anyway, I think this patch can stand on it's own with an updated commit
> message.  However, see below...
> 
> > 
> > Signed-off-by: Fabio M. De Francesco <fmdefrancesco@...il.com>
> > ---
> >  drivers/staging/r8188eu/os_dep/osdep_service.c | 7 ++-----
> >  1 file changed, 2 insertions(+), 5 deletions(-)
> > 
> > diff --git a/drivers/staging/r8188eu/os_dep/osdep_service.c b/drivers/staging/r8188eu/os_dep/osdep_service.c
> > index 7a6fcc96081a..d680bfba7f5d 100644
> > --- a/drivers/staging/r8188eu/os_dep/osdep_service.c
> > +++ b/drivers/staging/r8188eu/os_dep/osdep_service.c
> > @@ -117,18 +117,15 @@ void rtw_free_netdev(struct net_device *netdev)
> >  	struct rtw_netdev_priv_indicator *pnpi;
> >  
> >  	if (!netdev)
> > -		goto RETURN;
> > +		return;

Actually this function doesn't need to test for a valid "netdev". There are only
two callers of this function (they are in os_dep/usb_intf.c) and they already test
the pointer soon before calling rtw_free_netdev().

Therefore, I'll remove the test for a valid "netdev" and (obviously) the code has
no more need to return at that point in function.

> >  
> >  	pnpi = netdev_priv(netdev);
> >  
> >  	if (!pnpi->priv)
> > -		goto RETURN;
> > +		return;

I cannot see how pnpi->priv might ever be NULL. Pavel Skripkin made me notice
that "in rtw_alloc_etherdev() (I can confirm this information because now I've 
just read the code), if pnpi->priv allocation fails, then netdev will
be just freed.". If "netdev" is already free, this function is never called.

Therefore, I'll remove this test too.

> This does not look right.  If netdev is not NULL why does this function skip
> free_netdev()?

After the two removals I've talked about above, the code will always call 
vfree(pnpi->priv) and then free_netdev(netdev).

Therefore, the code won't anymore skip free_netdev() and the bug is avoided.

> 
> Fabio could you follow up with Larry and/or Phillip and see why the code does
> this?  To me it looks like a potential bug.
> 
> Thanks!
> Ira
> 
> >  
> >  	vfree(pnpi->priv);
> >  	free_netdev(netdev);
> > -
> > -RETURN:
> > -	return;
> >  }
> >  
> >  int rtw_change_ifname(struct adapter *padapter, const char *ifname)
> > -- 
> > 2.34.1
> > 
> 

This is how I think to rework rtw_free_netdev():

void rtw_free_netdev(struct net_device *netdev)
{
        struct rtw_netdev_priv_indicator *pnpi = netdev_priv(netdev);

        vfree(pnpi->priv);
        free_netdev(netdev);
}

Am I missing something?

@Greg: please discard this patch; I'll send another that has the purpose
to rework rtw_free_netdev() as I showed above for the purpose to avoid 
redundant tests and avoid the potential skipping of free_netdev() (as Ira 
has correctly noted, currently we have a bug).

Thanks,

Fabio