Re: preposterous TOD clock time on 5.0

On Thursday 07 May 2009 17:04:42 Izumi Tsutsui wrote:
> Martin Husemann wrote:
> > On Fri, May 08, 2009 at 12:38:08AM +0900, Izumi Tsutsui wrote:
> > > Is anyone else seeing this?
> > > (note it's on 5.0 so time_t is still 32 bit)
> >
> > I see the same on -current (though I thought it works once after power
> > cycle, have to retest that again).
>
> Per OpenBSD's hp700/dev/clock.c rev 1.14:
> http://www.openbsd.org/cgi-bin/cvsweb/src/sys/arch/hppa/dev/clock.c#rev1.14
> the following patch seems to fix the problem:
>
> ---
> Index: dev/pdc.c
> ===================================================================
> RCS file: /cvsroot/src/sys/arch/hp700/dev/pdc.c,v
> retrieving revision 1.25
> diff -u -r1.25 pdc.c
> --- dev/pdc.c	14 Jun 2008 21:27:37 -0000	1.25
> +++ dev/pdc.c	7 May 2009 16:01:25 -0000
> @@ -475,7 +475,8 @@
>  	tod.usec = tvp->tv_usec;
>
>  	pagezero_cookie = hp700_pagezero_map();
> -	pdc_call((iodcio_t)PAGE0->mem_pdc, 1, PDC_TOD, PDC_TOD_WRITE, &tod);
> +	pdc_call((iodcio_t)PAGE0->mem_pdc, 1, PDC_TOD, PDC_TOD_WRITE,
> +	    tod.sec, tod.usec);
>  	hp700_pagezero_unmap(pagezero_cookie);
>  	return 0;
>  }

Please commit.

I'll remove the hp700_pagezero_{,un}map / PAGE0 access after you've done that.

Thanks,
Nick