Re: [PATCH] randread: unitialized memory fix

[Mikhail Barashkov]

I think that this could be limited to __e2k__ protected mode, as the 
previous patch was.

I'll update with a new patch for that.

Mikhail

21.09.2021 18:14, Pádraig Brady wrote:
> On 21/09/2021 15:40, Mikhail Barashkov via GNU coreutils General 
> Discussion wrote:
> > Randread was allocating memory, but not setting it to 0, resulting, in
> > particular, to a crash in shuf program in E2K CPU protected mode (later
> > on s->bug is used by shuf in ISAAC_MIX as the seed parameter).
> >
> > This patch makes sure the memory buffer is zeroed out correctly.
> >
> > diff --git a/gl/lib/randread.c b/gl/lib/randread.c
> > index 7124e3df0..8e6b1c5b8 100644
> > --- a/gl/lib/randread.c
> > +++ b/gl/lib/randread.c
> > @@ -132,6 +132,7 @@ static struct randread_source *
> >    simple_new (FILE *source, void const *handler_arg)
> >    {
> >      struct randread_source *s = xmalloc (sizeof *s);
> > +  memset(s, 0, sizeof *s);
> >      s->source = source;
> >      s->handler = randread_error;
> >      s->handler_arg = handler_arg;
>
> I'm wary of the unconditional initialization of this large struct here.
> I'm reminded of this general discussion https://research.swtch.com/sparse
>
> If __e2k__ needs this, we should probably restrict to that arch?
> Also should you be using xcalloc(1, sizeof *s) instead?
>
> I'll let Paul review this as he's most familiar with this code.
>
> cheers,
> Pádraig