Re: [Qemu-devel] [5532] Replace uses of strndup (a GNU extension) with Qemu pstrdup

["Blue Swirl" <blauwirbel@gmail.com>]

On 10/25/08, andrzej zaborowski <balrogg@gmail.com> wrote:
> 2008/10/25 Blue Swirl <blauwirbel@gmail.com>:
>  > On 10/25/08, andrzej zaborowski <balrogg@gmail.com> wrote:
>  >> 2008/10/25 Blue Swirl <blauwirbel@gmail.com>:
>  >>  > On 10/25/08, andrzej zaborowski <balrogg@gmail.com> wrote:
>  >>  >> 2008/10/25 Blue Swirl <blauwirbel@gmail.com>:
>  >>  >>  > On 10/25/08, andrzej zaborowski <balrogg@gmail.com> wrote:
>  >>  >>  >> 2008/10/25 Blue Swirl <blauwirbel@gmail.com>:
>  >>  >>  >>
>  >>  >>  >> > Revision: 5532
>  >>  >>  >>  >          http://svn.sv.gnu.org/viewvc/?view=rev&root=qemu&revision=5532
>  >>  >>  >>  > Author:   blueswir1
>  >>  >>  >>  > Date:     2008-10-25 11:23:27 +0000 (Sat, 25 Oct 2008)
>  >>  >>  >>  >
>  >>  >>  >>  > Log Message:
>  >>  >>  >>  > -----------
>  >>  >>  >>  > Replace uses of strndup (a GNU extension) with Qemu pstrdup
>  >>  >>  >>  >
>  >>  >>  >>  > Modified Paths:
>  >>  >>  >>  > --------------
>  >>  >>  >>  >    trunk/cutils.c
>  >>  >>  >>  >    trunk/hw/bt-hci.c
>  >>  >>  >>  >    trunk/qemu-common.h
>  >>  >>  >>  >
>  >>  >>  >>  > Modified: trunk/cutils.c
>  >>  >>  >>  > ===================================================================
>  >>  >>  >>  > --- trunk/cutils.c      2008-10-25 11:21:28 UTC (rev 5531)
>  >>  >>  >>  > +++ trunk/cutils.c      2008-10-25 11:23:27 UTC (rev 5532)
>  >>  >>  >>  > @@ -50,6 +50,18 @@
>  >>  >>  >>  >     return buf;
>  >>  >>  >>  >  }
>  >>  >>  >>  >
>  >>  >>  >>  > +/* strdup with a limit */
>  >>  >>  >>  > +char *pstrdup(const char *str, size_t buf_size)
>  >>  >>  >>  > +{
>  >>  >>  >>  > +    size_t len;
>  >>  >>  >>  > +    char *buf;
>  >>  >>  >>  > +
>  >>  >>  >>  > +    len = MIN(buf_size, strlen(str));
>  >>  >>  >>  > +    buf = qemu_malloc(len);
>  >>  >>  >>  > +    pstrcpy(buf, len, str);
>  >>  >>  >>  > +    return buf;
>  >>  >>  >>  > +}
>  >>  >>  >>
>  >>  >>  >>
>  >>  >>  >> I think here also pstrcpy will only copy up to buf_size - 1 characters
>  >>  >>  >>  while strndup would copy buf_size chars.
>  >>  >>  >
>  >>  >>  > That is actually safer if we always want the strings to be NUL terminated.
>  >>  >>
>  >>  >>
>  >>  >> strndup also always NUL terminates the string so it's just as safe,
>  >>  >>  the length is just different.
>  >>  >>
>  >>  >>
>  >>  >>  >
>  >>  >>  > But the allocation length is wrong, it should be MIN(buf_size, strlen(str) + 1).
>  >>  >>
>  >>  >>
>  >>  >> By my reading of the manual, it should rather be MIN(buf_size, strlen(str)) + 1.
>  >>  >
>  >>  > But then the length could be incorrect: buf_size + 1.
>  >>
>  >>
>  >> That's what it should be in case lmp_name is 248 chars long.  I think
>  >>  the confusion is because you called the parameter buf_size while
>  >>  strndup calls it n (the number of characters.. not buffer size).
>  >
>  > If strlen(lmp_name) >= 248, only first 247 should be copied and the
>  > final character should be NUL, because sizeof(read_local_name_rp.name)
>  > == buf_size == 248.
>
>
> Nope, actually now I notice that I assumed that strndup never executes
>  strlen() on the parameter... I think it's a sane assumption (?).

Yes, glibc uses strnlen:
http://sources.redhat.com/cgi-bin/cvsweb.cgi/libc/string/strndup.c?rev=1.11&content-type=text/x-cvsweb-markup&cvsroot=glibc

>  read_local_name_rp.name format is specified by the bluetooth "Volume 2
>  Core system package, part E - Host Controller Interface Functional
>  Specification" on pg. 394:
>
>  6.24 LOCAL NAME
>  <...>
>  If the name contained in the parameter is shorter than 248 octets, the
>  end of the name is indicated by a NULL octet (0x00), and the following
>  octets (to fill up 248 octets, which is the length of the parameter) do not
>  have valid values.

So, in this case neither strndup nor pstrdup are correct, because you
do not want the string to be NUL terminated in the strlen(lmp_name) >=
248 case?