Re: [PATCH] lib: string.c: Added a funktion function strzcpy

[Andrew Morton <akpm@linux-foundation.org>]

On Wed, 24 Sep 2014 22:51:14 +0200 Rickard Strandqvist <rickard_strandqvist@spectrumdigital.se> wrote:

> 2014-09-24 17:41 GMT+02:00 Dan Carpenter <dan.carpenter@oracle.com>:
> > On Wed, Sep 24, 2014 at 07:35:55AM -0700, Andi Kleen wrote:
> >> On Wed, Sep 24, 2014 at 10:52:06AM +0300, Dan Carpenter wrote:
> >> > On Tue, Sep 23, 2014 at 06:17:53PM -0700, Andi Kleen wrote:
> >> > > On Wed, Sep 24, 2014 at 12:13:36AM +0200, Rickard Strandqvist wrote:
> >> > > > Added a function strzcpy which works the same as strncpy,
> >> > > > but guaranteed to produce the trailing null character.
> >> > >
> >> > > Do we really need the bizarre strncpy padding semantics for anything?
> >> > > Why not just use strlcpy?
> >> >
> >> > We do need the padding in many places to prevent information leaks.
> >>
> >> Like where?
> >>
> >
> > You're asking what would break if we switched every strncpy() to
> > strlcpy() but it's not an easy question to answer.
> >
> > I've looked a lot at information leaks, but strings are still a blind
> > spot for my Smatch.  My check only looks at normal variables, arrays.
> > Eventually I hope to fix this, of course.
> >
> > I did a git search and Rickard has added some examples, but there were
> > definitely other places that rely strncpy() padding before.
> >
> > regards,
> > dan carpenter
> 
> 
> Hi
> 
> If you want to see examples of this type of error, you can check the
> patches I've done over the past two months.
> So in linux-next
> 
> git log --patch-with-stat
> 
> And search for my email address.
> 
> And there is also an approximately 20 more patches I waited to see if
> I will be able to use a potential strzcpy  :-)

I have to say, this email thread has been a good demonstration of the
effects of incomplete changelogging :(

How's about you try again and this time include a *full* explanation of
why you believe the kernel needs strzcpy()?  Describe the problem,
provide those examples, explain the proposed solution, etc.

Also...  I misread the code - that implementation indeed zeroes out to
the end of the destination.  I'm not sure that it does it very
efficiently though - it repeatedly dereferences `src' just to obtain a
null byte?  A second nulling loop would be clearer and perhaps faster.