Re: [PATCH] staging: android: Replace strcpy with strlcpy

[Dan Carpenter <dan.carpenter@oracle.com>]

On Mon, Mar 13, 2017 at 06:17:22PM +0530, SIMRAN SINGHAL wrote:
> On Mon, Mar 13, 2017 at 6:11 PM, Dan Carpenter <dan.carpenter@oracle.com> wrote:
> > On Sun, Mar 12, 2017 at 02:10:01AM +0530, simran singhal wrote:
> >> Replace strcpy with strlcpy as strcpy does not check for buffer
> >> overflow.
> >> This is found using Flawfinder.
> >>
> >> Signed-off-by: simran singhal <singhalsimran0@gmail.com>
> >> ---
> >>  drivers/staging/android/ashmem.c | 3 ++-
> >>  1 file changed, 2 insertions(+), 1 deletion(-)
> >>
> >> diff --git a/drivers/staging/android/ashmem.c b/drivers/staging/android/ashmem.c
> >> index 7cbad0d..eb2f4ef 100644
> >> --- a/drivers/staging/android/ashmem.c
> >> +++ b/drivers/staging/android/ashmem.c
> >> @@ -548,7 +548,8 @@ static int set_name(struct ashmem_area *asma, void __user *name)
> >>       if (unlikely(asma->file))
> >>               ret = -EINVAL;
> >>       else
> >> -             strcpy(asma->name + ASHMEM_NAME_PREFIX_LEN, local_name);
> >> +             strlcpy(asma->name + ASHMEM_NAME_PREFIX_LEN, local_name,
> >> +                     sizeof(asma->name + ASHMEM_NAME_PREFIX_LEN));
> >                         ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
> > This isn't right.
> >
> > Also please do some analysis to see if it's a real bug or a false
> > positive.  It is a false positive in this case.
> >
> 
> Dan,
> I have already sent v3 of this in which I have used:
> sizeof(asma->name) - ASHMEM_NAME_PREFIX_LEN

Yeah.  I saw that.  It's fine, I suppose but you should have done more
analysis to see if it was a real bug like Al and Greg suggested.  The
changelog should say something like:

"The destination buffer is 12345 bytes long but we're copying a 10000
character string so it can overflow."  Occasionally, I will fudge a
little bit on these changelogs to say that I have looked every where to
determine the size of the source buffer and can't figure it out so this
change makes it easier to audit.  But I try to figure it out generally.

Really tools should be able to show that this code is safe.  They
currently don't so far as I know, but they should.  It's a matter of
waiting a year for Smatch to improve.

regards,
dan carpenter