[PATCH v2 3/3] mm/page_alloc: Keep memoryless cpuless node 0 offline

[David Hildenbrand]

On 11.05.20 19:47, Srikar Dronamraju wrote:
> * David Hildenbrand <david at redhat.com> [2020-05-08 15:42:12]:
> 
> Hi David,
> 
> Thanks for the steps to tryout.
> 
>>>
>>> #! /bin/bash
>>> sudo x86_64-softmmu/qemu-system-x86_64 \
>>>     --enable-kvm \
>>>     -m 4G,maxmem=20G,slots=2 \
>>>     -smp sockets=2,cores=2 \
>>>     -numa node,nodeid=0,cpus=0-1,mem=4G -numa node,nodeid=1,cpus=2-3,mem=0G \
>>
>> Sorry, this line has to be
>>
>> -numa node,nodeid=0,cpus=0-3,mem=4G -numa node,nodeid=1,mem=0G \
>>
>>>     -kernel /home/dhildenb/git/linux/arch/x86_64/boot/bzImage \
>>>     -append "console=ttyS0 rd.shell rd.luks=0 rd.lvm=0 rd.md=0 rd.dm=0" \
>>>     -initrd /boot/initramfs-5.2.8-200.fc30.x86_64.img \
>>>     -machine pc,nvdimm \
>>>     -nographic \
>>>     -nodefaults \
>>>     -chardev stdio,id=serial \
>>>     -device isa-serial,chardev=serial \
>>>     -chardev socket,id=monitor,path=/var/tmp/monitor,server,nowait \
>>>     -mon chardev=monitor,mode=readline
>>>
>>> to get a cpu-less and memory-less node 1. Never tried with node 0.
>>>
> 
> I tried 
> 
> qemu-system-x86_64 -enable-kvm -m 4G,maxmem=20G,slots=2 -smp sockets=2,cores=2 -cpu host -numa node,nodeid=0,cpus=0-3,mem=4G -numa node,nodeid=1,mem=0G -vga none -nographic -serial mon:stdio /home/srikar/fedora.qcow2
> 
> and the resulting guest was.
> 
> [root at localhost ~]# numactl -H
> available: 1 nodes (0)
> node 0 cpus: 0 1 2 3
> node 0 size: 3927 MB
> node 0 free: 3316 MB
> node distances:
> node   0
>   0:  10
> 
> [root at localhost ~]# lscpu
> Architecture:        x86_64
> CPU op-mode(s):      32-bit, 64-bit
> Byte Order:          Little Endian
> Address sizes:       40 bits physical, 48 bits virtual
> CPU(s):              4
> On-line CPU(s) list: 0-3
> Thread(s) per core:  1
> Core(s) per socket:  2
> Socket(s):           2
> NUMA node(s):        1
> Vendor ID:           GenuineIntel
> CPU family:          6
> Model:               46
> Model name:          Intel(R) Xeon(R) CPU           X7560  @ 2.27GHz
> Stepping:            6
> CPU MHz:             2260.986
> BogoMIPS:            4521.97
> Virtualization:      VT-x
> Hypervisor vendor:   KVM
> Virtualization type: full
> L1d cache:           32K
> L1i cache:           32K
> L2 cache:            4096K
> L3 cache:            16384K
> NUMA node0 CPU(s):   0-3
> Flags:               fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ht syscall nx rdtscp lm constant_tsc arch_perfmon rep_good nopl xtopology cpuid tsc_known_freq pni vmx ssse3 cx16 sse4_1 sse4_2 x2apic popcnt tsc_deadline_timer hypervisor lahf_lm cpuid_fault pti ssbd ibrs ibpb tpr_shadow vnmi flexpriority ept vpid tsc_adjust arat umip arch_capabilities
> 
> [root at localhost ~]# cat /sys/devices/system/node/online
> 0
> [root at localhost ~]# cat /sys/devices/system/node/possible
> 0-1
> 
> ---------------------------------------------------------------------------------
> 
> I also tried
> 
> qemu-system-x86_64 -enable-kvm -m 4G,maxmem=20G,slots=2 -smp sockets=2,cores=2 -cpu host -numa node,nodeid=1,cpus=0-3,mem=4G -numa node,nodeid=0,mem=0G -vga none -nographic -serial mon:stdio /home/srikar/fedora.qcow2
> 
> and the resulting guest was.
> 
> [root at localhost ~]# numactl -H
> available: 1 nodes (0)
> node 0 cpus: 0 1 2 3
> node 0 size: 3927 MB
> node 0 free: 3316 MB
> node distances:
> node   0
>   0:  10
> 
> [root at localhost ~]# lscpu
> Architecture:        x86_64
> CPU op-mode(s):      32-bit, 64-bit
> Byte Order:          Little Endian
> Address sizes:       40 bits physical, 48 bits virtual
> CPU(s):              4
> On-line CPU(s) list: 0-3
> Thread(s) per core:  1
> Core(s) per socket:  2
> Socket(s):           2
> NUMA node(s):        1
> Vendor ID:           GenuineIntel
> CPU family:          6
> Model:               46
> Model name:          Intel(R) Xeon(R) CPU           X7560  @ 2.27GHz
> Stepping:            6
> CPU MHz:             2260.986
> BogoMIPS:            4521.97
> Virtualization:      VT-x
> Hypervisor vendor:   KVM
> Virtualization type: full
> L1d cache:           32K
> L1i cache:           32K
> L2 cache:            4096K
> L3 cache:            16384K
> NUMA node0 CPU(s):   0-3
> Flags:               fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ht syscall nx rdtscp lm constant_tsc arch_perfmon rep_good nopl xtopology cpuid tsc_known_freq pni vmx ssse3 cx16 sse4_1 sse4_2 x2apic popcnt tsc_deadline_timer hypervisor lahf_lm cpuid_fault pti ssbd ibrs ibpb tpr_shadow vnmi flexpriority ept vpid tsc_adjust arat umip arch_capabilities
> 
> [root at localhost ~]# cat /sys/devices/system/node/online
> 0
> [root at localhost ~]# cat /sys/devices/system/node/possible
> 0-1
> 
> Even without my patch, both the combinations, I am still unable to see a
> cpuless, memoryless node being online. And the interesting part being even

Yeah, I think on x86, all memory-less and cpu-less nodes are offline as
default. Especially when hotunplugging cpus/memory, we set them offline
as well.

But as Michal mentioned, the node handling code is complicated and
differs between various architectures.

> if I mark node 0 as cpuless,memoryless and node 1 as actual node, the system
> somewhere marks node 0 as the actual node.

Is the kernel maybe mapping PXM 1 to node 0 in that case, because it
always requires node 0 to be online/contain memory? Would be interesting
what happens if you hotplug a DIMM to (QEMU )node 0 - if PXM 0 will be
mapped to node 1 then as well.


-- 
Thanks,

David / dhildenb