Re: [musl] Re: [PATCH] iconv: Add check null-check for scd pointer

["alice" <alice@ayaya.dev>]

On Sun Mar 24, 2024 at 7:33 PM UTC, Rich Felker wrote:
> On Sun, Mar 24, 2024 at 10:25:03PM +0300, Maks Mishin wrote:
> > After having been assigned to a NULL value at iconv.c:230,
> > pointer 'scd' is dereferenced at iconv.c:383.
> > 
> > Found by RASU JSC.
> > 
> > Signed-off-by: Maks Mishin <maks.mishinFZ@gmail.com>
> > ---
> >  src/locale/iconv.c | 2 ++
> >  1 file changed, 2 insertions(+)
> > 
> > diff --git a/src/locale/iconv.c b/src/locale/iconv.c
> > index 7fb2e1ef..e0d200b8 100644
> > --- a/src/locale/iconv.c
> > +++ b/src/locale/iconv.c
> > @@ -232,6 +232,8 @@ size_t iconv(iconv_t cd, char **restrict in, size_t *restrict inb, char **restri
> >  		scd = (void *)cd;
> >  		cd = scd->base_cd;
> >  	}
> > +	if (scd == NULL) return x;
> > +
> >  	unsigned to = extract_to(cd);
> >  	unsigned from = extract_from(cd);
> >  	const unsigned char *map = charmaps+from+1;
> > -- 
> > 2.30.2
>
> This makes iconv non-functional for non-stateful conversions. The
> claim by the static analysis tool is false. It is not dereferenced in
> the code path where it's null because in that code path,
> type==ISO2022_JP is never true.
>
> This tool you are using is really junk. You should stop sending
> untested and obviously incorrect patches to projects, and advise any
> projects that have accepted your patches that they may have been
> dangerously incorrect.

I'm pretty sure RASU JSC is not a tool but rather the Rusatom State Atomic
Corporation JSC, i.e. a branch at the Russian atomic energy company.

>
> Rich