Re: [PATCH v4 00/22] x86, ACPI, numa: Parse numa info early

[Tang Chen]

Hi Yinghai, all,

I've tested this patch-set with my following patch-set:
[PATCH v1 00/12] Arrange hotpluggable memory in SRAT as ZONE_MOVABLE.
https://lkml.org/lkml/2013/4/19/94

Using ACPI table override, I overrided SRAT on my box like this:

[    0.000000] SRAT: Node 0 PXM 0 [mem 0x00000000-0x7fffffff]
[    0.000000] SRAT: Node 0 PXM 0 [mem 0x100000000-0x307ffffff]
[    0.000000] SRAT: Node 1 PXM 2 [mem 0x308000000-0x583ffffff] Hot 
Pluggable
[    0.000000] SRAT: Node 2 PXM 3 [mem 0x584000000-0x7ffffffff] Hot 
Pluggable

We had 3 nodes, node0 was not hotpluggable, and node1 and node2 were 
hotpluggable.


And memblock reserved pagetable pages (with flag 0x1) in local nodes.
......
[    0.000000]  reserved[0xb]   [0x00000307ff0000-0x00000307ff1fff], 
0x2000 bytes flags: 0x0
[    0.000000]  reserved[0xc]   [0x00000307ff2000-0x00000307ffffff], 
0xe000 bytes on node 0 flags: 0x1
[    0.000000]  reserved[0xd]   [0x00000583ff7000-0x00000583ffffff], 
0x9000 bytes on node 1 flags: 0x1
[    0.000000]  reserved[0xe]   [0x000007ffff9000-0x000007ffffffff], 
0x7000 bytes on node 2 flags: 0x1

And after some bug fix, memblock can also reserve hotpluggable memory 
with flag 0x2.
......
[    0.000000]  reserved[0xb]   [0x00000307ff0000-0x00000307ff1fff], 
0x2000 bytes flags: 0x0
[    0.000000]  reserved[0xc]   [0x00000307ff2000-0x00000307ffffff], 
0xe000 bytes on node 0 flags: 0x1
[    0.000000]  reserved[0xd]   [0x00000308000000-0x00000583ff6fff], 
0x27bff7000 bytes on node 1 flags: 0x2
[    0.000000]  reserved[0xe]   [0x00000583ff7000-0x00000583ffffff], 
0x9000 bytes on node 1 flags: 0x1
[    0.000000]  reserved[0xf]   [0x00000584000000-0x000007ffff7fff], 
0x27bff8000 bytes on node 2 flags: 0x2
[    0.000000]  reserved[0x10]  [0x000007ffff8000-0x000007ffffffff], 
0x8000 bytes on node 2 flags: 0x1

And free it to buddy system when memory initialization finished.


So the results:
1. We can parse SRAT earlier correctly.
2. We can override tables correctly.
3. We can put pagetable pages in local node.
4. We can prevent memblock from allocating hotpluggable memory.
5. We can arrange ZONE_MOVABLE using SRAT info.


Known problems:

When we put pagetable pages in local node, memory hot-remove logic won't 
work.
I'm fixing it now. We need to fix the following:
1. Improve hot-remove to support freeing local node pagetable pages.
2. Improve hot-add to support putting hot-added pagetable pages in local 
node.
3. Do the same to vmemmap and page_cgrop pages.

So I suggest to separate the job into 2 parts:
1. Push Yinghai's patch1 ~ patch20, without putting pagetable in local node.
   And push my work to use SRAT to arrange ZONE_MOVABLE.
   In this case, we can enable memory hotplug in the kernel first.
2. Merge patch21 and patch22 into the fixing work I am doing now, and 
push them
   together when finished.

How do you think ?

Reviewed-by: Tang Chen <tangchen@xxxxxxxxxxxxxx>
Tested-by: Tang Chen <tangchen@xxxxxxxxxxxxxx>

Thanks. :)




--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/