Re: [PATCH] MIPS: Loongson: Fix complie errors without CONFIG_SMP

[Matthew Wilcox]

On Thu, Aug 27, 2020 at 10:31:33AM +0800, hejinyang wrote:
> 
> 
> On 08/26/2020 09:01 PM, Matthew Wilcox wrote:
> > On Wed, Aug 26, 2020 at 08:53:27PM +0800, Jinyang He wrote:
> > > +++ b/arch/mips/include/asm/mach-loongson64/topology.h
> > > @@ -4,7 +4,11 @@
> > >   #ifdef CONFIG_NUMA
> > > +#ifdef CONFIG_SMP
> > >   #define cpu_to_node(cpu)	(cpu_logical_map(cpu) >> 2)
> > > +#else
> > > +#define cpu_to_node(cpu)	0
> > > +#endif
> > Are you saying you've enabled NUMA without enabling SMP?  Does that make
> > sense?
> 
> NUMA option normally work with more than two nodes, though Loongson64 use it
> default after patch 6fbde6b492dfc761ad60a68fb2cb32b1eb05b786. Loongson64(3A)'s
> each node consists of 4 cpus while it only has 4 cpus. In other words, it has
> only one node while open NUMA. I'm confused it if NUMA deponds on SMP.
> 
> I'll try fix it with unabling NUMA and SMP later.

mips:
config NUMA
        bool "NUMA Support"
        depends on SYS_SUPPORTS_NUMA

x86:
config NUMA
        bool "NUMA Memory Allocation and Scheduler Support"
        depends on SMP

If you had a good reason for enabling NUMA without enabling SMP (CPU-less
memory nodes?), that'd be an interesting discussion to have.  Since your
hardware seems to have SMP by default, I'd suggest just adding the missing
'depends'.