Re: [POC] Extend "movable_node" to "movable_node=nn@ss" and add the interface in /sys to show the value

[Chao Fan]

On Wed, Oct 11, 2017 at 03:55:13PM +0800, Baoquan He wrote:
>Hi Fan San,
>
>On 10/11/17 at 10:28am, Chao Fan wrote:
>> Hi all,
>> 
>> Here is a problem:
>> Here is a machine with several NUMA nodes and some of them are hot-pluggable,
>> It's not good for kernel to be extracted in the memory region of movable node.
>> But in current code, I print the address choosen by kaslr and found it may be
>> placed in movable node sometimes. To solve this problem, it's better to the
>> memory region choosen by kaslr to immovable node in kaslr.c. But the memory
>> infomation about if it's hot-pluggable is stored in ACPI SRAT table, which is
>> parsed after kernel is extracted. So we can't get the detail memory infomation
>> before extracting kernel.
>> 
>> There are two methods to solve this problem:
>> 
>> 1. Get and parse the srat table before kernel extracted, then mark the memory
>>    region in movable node which should be avoided in kaslr.
>>    I have send the patch:
>>    https://www.spinics.net/lists/kernel/msg2595546.html
>>    But the change is large and then here is the second method.
>> 
>> 2. Extend the movable_node to movable_node=nn@ss, in which nn means
>>    the size of memory in immovable node, and ss means the start position of
>>    this memory region.
>>    But it brings another question, it may be a little difficult for a normal
>>    user to specify the nn and ss. Because it's hard for a user to know the value
>>    of the memory in immovable node.
>>    So I wonder if it's good to add a interface in /sys, like:
>>    # cat /sys/device/system/memory/immovable_node
>
Hi Baoquan,

Thanks for your reply,

>You can post patch. By the way, can the existing
>/sys/devices/system/memory/memoryX/removable be used instead?

I ever search the interfaces of /sys/devices/system/memory/, and noticed
this "removable", it does be able to help us judge it's removable or not.
But we should also get the information of the memory length and start position
by other interface. If there is a interface that show the nn and ss, we
can use them and change grub directly.
And there are more "memoryX" in one node. In my machine, I can see
memory0(linked to /sys/devices/system/memory/memory0) - memory7 in
/sys/devices/system/node/node0. And total memory38 for 4 nodes. I think
it's a little heavy to handle every memoryX.
But in SRAT table, one node has one or two memory regions. So I thinks
it's more straight and easy to use. How do you think about it?

Thanks,
Chao Fan
>
>Thanks
>Baoquan
>
>>    nn@ss
>>    nn@ss
>>    ...
>>    to show the two value.
>>    When srat table is parsed in acpi_parse_memory_affinity, fill the value
>>    and user can get and use them.
>> 
>> If anyone has a better method, please let me know.
>> Any comments will be welcome.
>> 
>> Thanks,
>> Chao Fan
>> 
>> 
>
>