Re: [PATCH 0/2] vsprintf: ignore %n again

[Kees Cook]

On Wed, Sep 18, 2013 at 6:14 AM, Tetsuo Handa
<penguin-kernel@xxxxxxxxxxxxxxxxxxx> wrote:
> Kees Cook wrote:
>> > Consider, e.g. introducing __vsnprint(), with vsnprintf(s, n, fmt, ...)
>> > expanding to __vsnprintf(1, s, n, fmt, ...) if fmt is a string literal
>> > and __vsnprintf(0, s, n, fmt, ...) otherwise.  Now,
>> > int __sprintf(int safe, char *buf, const char *fmt, ...)
>> > {
>> >         va_list args;
>> >         int i;
>> >
>> >         va_start(args, fmt);
>> >         i = __vsnprintf(safe, buf, INT_MAX, fmt, args);
>> >         va_end(args);
>> >
>> >         return i;
>> > }
>>
>> Unless I've misunderstood, I think we'd already get very close to this
>> with the gcc options instead. This patch set is what I've been using
>> to generate the format string fixes over the last few months, with 7
>> sent this last round:
>
> Can we utilize __builtin_constant_p() ?
>
> ---------- source start ----------
> #include <stdio.h>
>
> #define func(fmt)                                       \
>         if (__builtin_constant_p(fmt))                  \
>                 printf("const : %s\n", fmt);            \
>         else                                            \
>                 printf("not const : %s\n", fmt);        \
>
>
> int main(int argc, char *argv[])
> {
>         const char *fmt1 = "const char *";
>         const char fmt2[] = "const char []";
>         const char * const fmt3 = "const char * const";
>         func("literal");
>         func(fmt1);
>         func(fmt2);
>         func(fmt3);

I wonder if we can get at the read-only knowledge. Doing this shows
that gcc knows they're read-only:

fmt1[2] = 'A';
fmt2[2] = 'A';
fmt3[2] = 'A';

/tmp/const.c:27:5: error: assignment of read-only location ‘*(fmt1 + 2u)’
/tmp/const.c:29:5: error: assignment of read-only location ‘fmt2[2]’
/tmp/const.c:31:5: error: assignment of read-only location ‘*(fmt3 + 2u)’

I didn't see a builtin for this in the gcc documentation I've been reading.

-Kees

>         return 0;
> }
> ---------- source end ----------
>
> ---------- output start ----------
> const : literal
> not const : const char *
> not const : const char []
> const : const char * const
> ---------- output end ----------
>
> __builtin_constant_p() seems to enforce use of either "literal" or "* const".
>
> An example change
>
> ---------- patch start ----------
> --- a/include/linux/printk.h
> +++ b/include/linux/printk.h
> @@ -120,8 +120,9 @@ asmlinkage int printk_emit(int facility, int level,
>                            const char *dict, size_t dictlen,
>                            const char *fmt, ...);
>
> -asmlinkage __printf(1, 2) __cold
> -int printk(const char *fmt, ...);
> +//asmlinkage __printf(1, 2) __cold
> +//int printk(const char *fmt, ...);
> +#define printk(fmt, ...) compiletime_assert(__builtin_constant_p(fmt), "Non-c  onstant format string")
>
>  /*
>   * Special printk facility for scheduler use only, _DO_NOT_USE_ !
> ---------- patch end ----------
>
> caught errors like below.
>
>   CC [M]  drivers/scsi/esas2r/esas2r_log.o
> drivers/scsi/esas2r/esas2r_log.c: In function 'esas2r_log_master':
> drivers/scsi/esas2r/esas2r_log.c:174: error: call to '__compiletime_assert_174' declared with attribute error: Non-constant format string



-- 
Kees Cook
Chrome OS Security
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/