[off-list ref] wrote:

+
+ =A0 =A0 =A0 for_each_node_by_name(np, "global-utilities") {
+ =A0 =A0 =A0 =A0 =A0 =A0 =A0 if ((of_get_property(np, "fsl,has-rstcr", N=

ULL))) {

+ =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 rstcr =3D of_iomap(np, 0) +=

 0xb0;

+ =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 if (!rstcr)
+ =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 printk (KER=

N_EMERG "Error: reset control "

I'm not sure KERN_EMERG is warranted for this kind of error.

+ =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0=

 =A0 =A0 =A0 =A0 =A0 "register not mapped!\n");

+ =A0 =A0 =A0 =A0 =A0 =A0 =A0 }

So if a node has an fsl,rstcr property, but the of_iomap() fails, we
jump to the next global-utilities node?  Perhaps you need a 'break'
after the printk()?

+ =A0 =A0 =A0 }
+
+ =A0 =A0 =A0 if (!rstcr && ppc_md.restart =3D=3D fsl_rstcr_restart)

Wouldn't it make more sense to assign fsl_rstcr_restart to
ppc_md.restart only if we find a valid fsl,has-rstcr property?

--=20
Timur Tabi
Linux kernel developer at Freescale