[PATCH] commands/digest: add verify support

[Uwe Kleine-König]

Hello Eric,

On Wed, Sep 24, 2014 at 03:21:10PM +0200, Eric Bénard wrote:
> From: Hubert Feurstein <h.feurstein at gmail.com>
> 
> Signed-off-by: Hubert Feurstein <h.feurstein at gmail.com>
> [EB]: reworked based on Sascha's comments and tested with md5sum
> Signed-off-by: Eric Bénard <eric at eukrea.com>
> ---
>  commands/digest.c | 120 +++++++++++++++++++++++++++++++++++++++++++++++++-----
>  1 file changed, 110 insertions(+), 10 deletions(-)
> 
> diff --git a/commands/digest.c b/commands/digest.c
> index 092fda2..2a982e5 100644
> --- a/commands/digest.c
> +++ b/commands/digest.c
> @@ -23,26 +23,99 @@
>  #include <fcntl.h>
>  #include <errno.h>
>  #include <xfuncs.h>
> +#include <libfile.h>
>  #include <malloc.h>
>  #include <digest.h>
> +#include <linux/ctype.h>
> +#include <getopt.h>
> +
> +static inline unsigned char parse_hexchar(char c)
> +{
> +	if (!isxdigit(c))
> +		return 0;
> +
> +	return isdigit(c) ? (c - '0') : ((islower(c) ? toupper(c) : c) - 'A' + 0xA);
Probably this is fine in barebox and also the check for isxdigit above
should make it save, but in general this approach might result in
surprises. In general isdigit et al are 1) locale dependant and 2) all
take an unsigned char (or EOF).

(Using Python here, but the same applies to C code obviously.)

With LC_ALL=de_DE LANG=de_DE the first fact yields:

$ python
Python 2.7.7 (default, Jun  3 2014, 16:16:56) 
[GCC 4.8.3] on linux2
Type "help", "copyright", "credits" or "license" for more information.
>>> import locale
>>> locale.setlocale(locale.LC_ALL, '')
'de_DE'
>>> print 'Ä'.isupper()
True

The problem with unsigned is that

	ispunct('\xfc')

might result in checking -4 which is undefined.

This obviously only applies to platforms where char is implicitly
signed (According to https://wiki.debian.org/ArchitectureSpecificsMemo
this includes x86 and mips.)

So I'd suggest to check the ASCII-values explicitly, something like:

	switch (c) {
	case '0'..'9':
		return c - '0';
		break;
	case 'a'..'f':
		return c - 'a' + 0xa;
		break;
	case 'A'..'F':
		return c - 'A' + 0xa;
		break;
	default:
		return 0;
		break;
	}

to be on the safe side and to also be able to use the code in userspace.
(But note this is even safe EBCDIC machines, but only because we're
using only the letters up to 'f'. 'j' - 'a' is 16 there. :-)

Best regards
Uwe, the pedantic

-- 
Pengutronix e.K.                           | Uwe Kleine-König            |
Industrial Linux Solutions                 | http://www.pengutronix.de/  |