Re: [PATCH v6] numa: make node_to_cpumask_map() NUMA_NO_NODE aware

[Greg KH]

On Mon, Oct 14, 2019 at 11:49:12AM +0200, Peter Zijlstra wrote:
> On Mon, Oct 14, 2019 at 11:25:09AM +0200, Greg KH wrote:
> > Good luck, I don't really think that most, if any, of this is needed,
> > but hey, it's nice to clean it up where it can be :)
> 
> Some of the virtual devices we have (that use devm) really ought to set
> the node too, like drivers/base/cpu.c and driver/base/node.c and
> arguably the cooling devices too (they create a device per cpu).
> 
> The patch I had here:
> 
>   https://lkml.kernel.org/r/20190925214526.GA4643@xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx
> 
> takes the more radical approach of requiring a node, except when
> explicitly marked not (the fake devices that don't use devm for
> example).

I like that patch :)

> But yes, PCI and other physical busses really should be having a node
> set, no excuses.

Agreed, at least just warning on the bus creation will make it a bit
less "noisy", in contrast to your patch.  But the messages in your patch
show people just how broken their bioses really are.  Which is always
fun...

> Anyway, I don't think non-physical devices actually use
> cpumask_of_node() much, a quick grep didn't show any.

That's good.

thanks,

greg k-h