Re: hwrng on 82801EB/ER (ICH5/ICH5R) fails rngtest checks

["gmu 2k6" <gmu2006@gmail.com>]

On 7/27/06, Michael Buesch <mb@bu3sch.de> wrote:
> On Thursday 27 July 2006 16:20, gmu 2k6 wrote:
> > On 7/27/06, Michael Buesch <mb@bu3sch.de> wrote:
> > > On Wednesday 26 July 2006 21:44, gmu 2k6 wrote:
> > > > > But could you try the following patch on top of latest git?
> > > > > It's just a random test, but I think it's worth trying.
> > > > > Let's see if it works around the issue.
> > > > >
> > > > > Index: linux-2.6/drivers/char/hw_random/intel-rng.c
> > > > > ===================================================================
> > > > > --- linux-2.6.orig/drivers/char/hw_random/intel-rng.c   2006-06-27 17:48:13.000000000 +0200
> > > > > +++ linux-2.6/drivers/char/hw_random/intel-rng.c        2006-07-26 17:27:03.000000000 +0200
> > > > > @@ -104,9 +104,14 @@
> > > > >         int err = -EIO;
> > > > >
> > > > >         hw_status = hwstatus_get(mem);
> > > > > +       hw_status = hwstatus_set(mem, hw_status & ~INTEL_RNG_ENABLED);
> > > > > +       hw_status = hwstatus_set(mem, hw_status | INTEL_RNG_ENABLED);
> > > > > +#if 0
> > > > > +       hw_status = hwstatus_get(mem);
> > > > >         /* turn RNG h/w on, if it's off */
> > > > >         if ((hw_status & INTEL_RNG_ENABLED) == 0)
> > > > >                 hw_status = hwstatus_set(mem, hw_status | INTEL_RNG_ENABLED);
> > > > > +#endif
> > > > >         if ((hw_status & INTEL_RNG_ENABLED) == 0) {
> > > > >                 printk(KERN_ERR PFX "cannot enable RNG, aborting\n");
> > > > >                 goto out;
> > > >
> > > > well as it didn't work, are you sure it was not intended to be more like this:
> > > > @@ -104,9 +104,14 @@
> > > >        int err = -EIO;
> > > >
> > > >        hw_status = hwstatus_get(mem);
> > > > +       hw_status = hwstatus_set(mem, hw_status & ~INTEL_RNG_ENABLED);
> > > > +       hw_status = hwstatus_set(mem, hw_status | INTEL_RNG_ENABLED);
> > > > +#if 0
> > > >        /* turn RNG h/w on, if it's off */
> > > >        if ((hw_status & INTEL_RNG_ENABLED) == 0)
> > > >                hw_status = hwstatus_set(mem, hw_status | INTEL_RNG_ENABLED);
> > > > +#endif
> > > >        if ((hw_status & INTEL_RNG_ENABLED) == 0) {
> > > >                printk(KERN_ERR PFX "cannot enable RNG, aborting\n");
> > > >                goto out;
> > > >
> > > > ?
> > >
> > > I don't think that makes a difference to the generated code, does it?
> >
> > I will test it now as you seem to be interested in the results.
> >
> > actually I was just curious what sense it made to do
> > hw_status = hwstatus_get(mem);
> > twice, though I'm not informed about the semantics there so I could be wrong
> > in interpreting the API on that level.
>
> Look at the #if 0. The second hwstatus_get is not going to be compiled, anyway.

yep you're right, originally I caught it as because it won't be
included in the gcc output at all and later on false thought it would
be called twice after a long day of work. maybe I should take more
time when replying to mails :D

> But Jan Beulich posted an interresting patch to lkml.
> Please test it.
> http://marc.theaimsgroup.com/?l=linux-kernel&m=115399953714650&q=raw

I saw that too and was thinking about trying it out, will do so now.