Re: [PATCH] powerpc/pseries: Track LMB nid instead of using device tree

[Michael Ellerman <mpe@ellerman.id.au>]

Nathan Fontenot <nfont@linux.vnet.ibm.com> writes:
> On 09/19/2018 11:38 PM, Michael Ellerman wrote:
>> Nathan Fontenot <nfont@linux.vnet.ibm.com> writes:
>>> When removing memory we need to remove the memory from the node
>>> it was added to instead of looking up the node it should be in
>>> in the device tree.
>>>
>>> During testing we have seen scenarios where the affinity for a
>>> LMB changes due to a partition migration or PRRN event. In these
>>> cases the node the LMB exists in may not match the node the device
>>> tree indicates it belongs in. This can lead to a system crash
>>> when trying to DLAPR remove the LMB after a migration or PRRN
>>> event. The current code looks up the node in the device tree to
>>> remove the LMB from, the crash occurs when we try to offline this
>>> node and it does not have any data, i.e. node_data[nid] == NULL.
>> 
>> This isn't building for 32-bit etc:
>> 
>> arch/powerpc/mm/drmem.c: In function 'init_drmem_v1_lmbs':
>> arch/powerpc/mm/drmem.c:371:14: error: implicit declaration of function 'memory_add_physaddr_to_nid' [-Werror=implicit-function-declaration]
>>    lmb->nid = memory_add_physaddr_to_nid(lmb->base_addr);
>>               ^
>> cc1: all warnings being treated as errors
>> scripts/Makefile.build:317: recipe for target 'arch/powerpc/mm/drmem.o' failed
>> 
>> See the failed checks here:
>>   https://patchwork.ozlabs.org/patch/969150/
>> 
>> 
>> Probably drmem.c should only be compiled for 64-bit NUMA etc.
>
> Looks like the root cause is that memory hotplug relies on sparsemem which
> is not supported on 32-bit.

Yeah that could be it.

Making drmem.c built just for MEMORY_HOTPLUG would make sense.

> This patch is also going to need a refresh to apply cleanly due to other
> patches that have gone in. I'll re-submit after looking at the build break issues more.

OK thanks.

cheers