Re: [PATCH] device-dax: use fallback nid when numa_node is invalid

[David Hildenbrand <david@redhat.com>]

On 29.07.21 16:44, Justin He wrote:
> Hi David
> 
>> -----Original Message-----
>> From: David Hildenbrand <david@redhat.com>
>> Sent: Thursday, July 29, 2021 3:59 PM
>> To: Justin He <Justin.He@arm.com>; Dan Williams <dan.j.williams@intel.com>;
>> Vishal Verma <vishal.l.verma@intel.com>; Dave Jiang <dave.jiang@intel.com>
>> Cc: nvdimm@lists.linux.dev; linux-kernel@vger.kernel.org; nd <nd@arm.com>
>> Subject: Re: [PATCH] device-dax: use fallback nid when numa_node is
>> invalid
>>
>> Hi Justin,
>>
>>>>>
>>>>
>>>> Note that this patch conflicts with:
>>>>
>>>> https://lkml.kernel.org/r/20210723125210.29987-7-david@redhat.com
>>>>
>>>> But nothing fundamental. Determining a single NID is similar to how I'm
>>>> handling it for ACPI:
>>>>
>>>> https://lkml.kernel.org/r/20210723125210.29987-6-david@redhat.com
>>>>
>>>
>>> Okay, got it. Thanks for the reminder.
>>> Seems my patch is not useful after your patch.
>>>
>>
>> I think your patch still makes sense. With
>>
>> https://lore.kernel.org/linux-acpi/20210723125210.29987-7-
>> david@redhat.com/
>>
>> We'd have to detect the node id in the first loop instead.
> 
> Ok, I got your point. I will do that in v2.
> 
> Btw, sorry for commenting there about your patch 06 since I didn't
> subscribe lkml via this mailbox.

Sure, you really should subscribe :)

> > +	for (i = 0; i < dev_dax->nr_range; i++) {
> +		struct range range;
> +
> +		rc = dax_kmem_range(dev_dax, i, &range);
> +		if (rc) {
> +			dev_info(dev, "mapping%d: %#llx-%#llx too small after alignment\n",
> +					i, range.start, range.end);
> +			continue;
> +		}
> +		total_len += range_len(&range);
> +	}
> You add an additional loop to get the total_len.
> I wonder is it independent on 2nd loop?
> If yes, why not merge the 2 loops into one?
> Sorry if this question is too simple, I don't know too much
> about the background of your patch.

We need total_len to register the memory group. We need the memory group 
to add memory. Therefore, we need a second loop to calculate total_len 
upfront.


-- 
Thanks,

David / dhildenb