qemu-kvm如何mem 热插拔

linux内核已经实现了内存热插的支持，当一个linux系统不管运行在  物理环境  或者 虚拟环境   时只要宿主能提供内存热插拔机制，linux内核就能相应的增加或者减少内存。

目前已知的VMware已经可以做到内存热插。

改变内存容量有两种方式：

40 1.1 purpose of memory hotplug
41 ------------
42 Memory Hotplug allows users to increase/decrease the amount of memory.
43 Generally, there are two purposes.
44 
45 (A) For changing the amount of memory.
46    This is to allow a feature like capacity on demand.
47 (B) For installing/removing DIMMs or NUMA-nodes physically.
48    This is to exchange DIMMs/NUMA-nodes, reduce power consumption, etc.

前些天检查maillist时发现已经有人实现了DIMMs 方式更改内存大小。这种方式还没有并入mainline tree。

40 1.1 purpose of memory hotplug
41 ------------
42 Memory Hotplug allows users to increase/decrease the amount of memory.
43 Generally, there are two purposes.
44 
45 (A) For changing the amount of memory.
46    This is to allow a feature like capacity on demand.
47 (B) For installing/removing DIMMs or NUMA-nodes physically.
48    This is to exchange DIMMs/NUMA-nodes, reduce power consumption, etc.

可以测试下：

1）下载带有 dimm hotplug 功能的qemu-kvm：

git clone -b  memhp-v3-alt https://github.com/vliaskov/qemu-kvm/commits/memhp-v3-alt 

配置编译:

$./configure --target-list="i386-softmmu x86_64-softmmu i386-linux-user x86_64-linux-user" --enable-kvm  

$make

2）下载linux内核 linux-3.0.1.tar.bz2 做测试：（32bit）

linux memory hotplug 需要内核编译进如下选项：

- For all memory hotplug  
    Memory model -> Sparse Memory  (CONFIG_SPARSEMEM)  
location:  
Processor type and features    
    Memory model (Sparse Memory)  --->  

    Allow for memory hot-add       (CONFIG_MEMORY_HOTPLUG)  
location:  
Power management and ACPI options  
    ACPI (Advanced Configuration and Power Interface) Support  
        Memory Hotplug  

- To enable memory removal, the followings are also necessary  
    Allow for memory hot remove    (CONFIG_MEMORY_HOTREMOVE)  
    Page Migration                 (CONFIG_MIGRATION)  

- For ACPI memory hotplug, the followings are also necessary  
    Memory hotplug (under ACPI Support menu) (CONFIG_ACPI_HOTPLUG_MEMORY)  
    This option can be kernel module.  
  
- As a related configuration, if your box has a feature of NUMA-node hotplug  
  via ACPI, then this option is necessary too.  
    ACPI0004,PNP0A05 and PNP0A06 Container Driver (under ACPI Support menu)  
    (CONFIG_ACPI_CONTAINER).  
    This option can be kernel module too.  

配置好之后就可以用make 了。

编译完成后将得到的bzImage文件拷贝出来。

3)制作启动所需固件：

当内存hotplugged之后，内核需要知道新增加上的内存并且制作内存管理表，生成sysfs文件来操作内存。

由于这需要固件的支持，而现有的qemu mainline 中包含的bios并不支持相应的功能，所以需要下载bios来重新编译：

git clone -b memhp-v3-alt https://github.com/vliaskov/seabios/commits/memhp-v3-alt  

创建一个放固件的目录：bios

里面放的固件如下所示：（这些都是编译后qemu目录下的）

.  
├── bios.bin  
├── keymaps  
│   ├── common  
│   ├── en-us  
│   └── modifiers  
├── kvmvapic.bin  
├── linuxboot.bin  
├── multiboot.bin  
├── optionrom  
│   ├── kvmvapic.bin  
│   ├── linuxboot.bin  
│   └── multiboot.bin  
├── ppc_rom.bin  
├── pxe-e1000.rom  
├── sgabios.bin  
├── slof.bin  
├── spapr-rtas.bin  
├── vgabios.bin  
├── vgabios-cirrus.bin  
├── vgabios-qxl.bin  
├── vgabios-stdvga.bin  
└── vgabios-vmware.bin  

将编译好的固件 ./out/bios.bin  替换上面的  bios.bin。

4）制作文件系统busybox.img

可参考如下连接：

http://blog.youkuaiyun.com/gudujianjsk/article/details/7655855  

现在测试用的内核文件都得到了，用如下脚本开启测试：

./../qemu-kvm/i386-softmmu/qemu-system-i386 -enable-kvm -m 512 -kernel bzImage -L ./../bios -append "root=/dev/sda" -dimm id=dimm0,size=512M,node=0,populated=off -boot c -smp 1  -hda busybox.img -monitor stdio -k en-us -vnc :2  

启动后系统显示如下：

由于编译时配置将 CONFIG_ACPI_HOTPLUG_MEMORY 配置为内核模块了，所以这里需要首先加载一个内核模块：acpi_memhotplug.ko.

所有的sections都在sysfs下有它们的设备信息。每一个section都是在 /sys/devices/system/memory/目录下的内存块的一部分。

/sys/devices/system/memory/memoryXXX

在这里XXX就是内存ID。

在monitor中动态 增加dimm：

sae@lx15:/var/dbg/test-dimm/busybox$ sudo ./test-box.sh   
QEMU 1.2.50 monitor - type 'help' for more information  
(qemu) device_add dimm,id=dimm0  

此时在虚拟机中可见多了个memory1，其状态是offline：

此时新增加的内存还没有做好使用的准备。为了使用新增加的内存你需要把新增加的内存设置为online 状态

此时用free查看总计的内存数量：

测试完毕。

http://www.mjmwired.net/kernel/Documentation/memory-hotplug.txt