On Tue, Jul 03, 2012 at 07:46:24AM -0500, Tabi Timur-B04825 wrote:
> On Tue, Jul 3, 2012 at 5:21 AM, Zhao Chenhui <chenhui.z...@freescale.com>
> wrote:
>
> > + np = of_find_matching_node(NULL, mpc85xx_smp_guts_ids);
> > + if (np) {
> > + guts = of_iomap(np, 0);
> > + of_node_put(np);
> > + if (!guts) {
> > + pr_err("%s: Could not map guts node address\n",
> > + __func__);
> > + return;
> > + }
> > + smp_85xx_ops.give_timebase = mpc85xx_give_timebase;
> > + smp_85xx_ops.take_timebase = mpc85xx_take_timebase;
> > + }
>
> I had this in mind:
>
> guts = of_iomap(np, 0);
> of_node_put(np);
> if (guts) {
> smp_85xx_ops.give_timebase = mpc85xx_give_timebase;
> smp_85xx_ops.take_timebase = mpc85xx_take_timebase;
> } else {
> pr_err("%s: Could not map guts node address\n",
> __func__);
> }
>
> That way, a missing GUTS node does not break everything.
>
If the guts variable is NULL, it indicates there is error in dts or kernel.
We should fix the error, rather than ignore it.
Moreover, if smp_85xx_ops.give/take_timebase is NULL, kernel can not do the
timebase sync.
-Chenhui
_______________________________________________
Linuxppc-dev mailing list
Linuxppc-dev@lists.ozlabs.org
https://lists.ozlabs.org/listinfo/linuxppc-dev
