Problem with KVM

[Zia 7]

I had problems installing KVM.!! Actually I am not having much knowledge about the KVM and how it is implemented. The problem is: when I claps Order

 sudo modprobe kvm-intel

I get this message:

FATAL: Error Inserting kvm_intel (/ lib/modules/2.6.24-22-generic/kernel/arch/x86/kvm/kvm-intel.ko)

By done, I checked my CPU supports virtualization:
Order

 egrep '^ flags .* (vmx | svm)' / proc / cpuinfo

gives as result:
flags: fpu tsc msr eighth of PSE pae mce CX8 apic mtrr pge mca cmov September pat clflush dts acpi mmx fxsr sse sse2 ss tm pbe bts constant_tsc up pni monitor vmx arch_perfmon TM2 is xtpr
Please help.

[T0tal L0$$]

Hello

First track, it is probably the classic mistake: disabled by bios. (A stroll in the syslog to be convinced). In which case, simply activate the feature Intel VT in the BIOS. Have you found the reason that blocks the installation of KVM?. I have already completed the installation of KM / Qemu on a Dell laptop without difficulty. Here I try to do the same on a Dell Optiplex 760 - Intel Core2 vPro. I check the Intel VT enabled in the BIOS, but nothing to do.

[Jevin]

According to me, A "egrep '^ flags .* (svm)' / proc / cpuinfo 'does not work. On what machine do you try your operation? You'll need Windows, and yes because the software for editing records BIOS only works on Windows and DOS specifically. You must create a bootable floppy disk or if you do not have a floppy drive (very common), a bootable USB key.

[Shutdown]

Reboot, then enter the BIOS setup, reset the default settings. Change the boot sequence to put floppy first. Activate the external boot device (I had forgotten is important!), Then change the boot sequence through USB Flash first position, save changes and boot your USB key. Once in DOS, type:

 SYMCMOS.EXE-v2-lDefault.txt

This command retrieves the current configuration of the BIOS and writes the file Default.txt. Restart your computer and go under Linux, then open the file Default.txt nano or your favorite editor. Modify the registry for your model: 0-000 of 000 1. 0 means disable and 1 enable.

[ComPaCt]

The virtualization infrastructure based on the implementation of some primitives, including ioctl () and mmap () (on the virtual device / dev / kvm) for the implementation of virtual machines. The context management of VM primitives through vcpu_load () / vcpu_put () in the loading / unloading control structures (ie VMCS). The VMCS are initialized so as to provide an adequate platform for bootstraps based on real-mode x86 systems. Finally, some logic processing control registers is put in place to properly manage guest systems.

[Recko]

Certain events can cause the release of the VM, ie the portion of processor mode VMX non-root VMX root mode, causing the performance of specific managers. KVM implements them for: exceptions (page faults only, actually), access to control registers, the instruction invlpg (for the invalidation of an entry in the TLB - Translation Look-aside Buffer: used by MMU to store the correspondence between linear and physical addresses, and so each time to avoid the tedious route of page tables in memory) and instructions for I / O interruption of the window manager (when interrupts invited are activated), access to MSR (Model Specific Registers-), the debug registers and instruction cpuid.

[SecRecy]

To switch to guest mode (ie non-root VMX mode for Intel), the command for KVM_RUN ioctl () is defined. Managing VM exit paths (ie the transition from VMX non-root in VMX root) implementation treats the two cases corresponding output in kernel mode or user mode the guest system. KVM also implements the injection of interruptions, and debugging of guest operating systems, offering an alternative to ring 0 debugger (quote (are promoting as an example rr0d - rasta ring 0 debugger - running in theory) regardless is the operating system) used when debugging (or others) of nuclei.

[Russell]

An important part in the management of virtualization MMU. In our case: a model of full virtualization. It is commonly used a technique called "shadow page tables". This is to maintain two sets of page table: one belonging to the guest system (not visible to the hardware via the control register cr3) and one maintained by the hypervisor and used directly by the hardware. To load the hypervisor then synchronize PTE "real" with those for guests. Implements KVM solution for now undeveloped using TLB management instructions to keep the synchronization between the Shadow Page Tables and those of guest systems.

[Joshua]

Regarding memory management to a higher level, KVM implements management based on "slots", a concept more or less close to the concept of DIMM slot that allows non-contiguous allocations. This makes it possible, inter alia, support for hotplug memory and the addition of PCI framebuffer running virtual machines. Another important block is the x86 emulation provided by KVM. Attention is not to emulate a whole machine in this case, fortunately (otherwise there would be no interest in KVM). The need for emulation due to the following reasons. The instructions are intercepted by mmio hardware as page faults without any information about the operation that must take place other than a virtual address (ie linear) involved.

[PeterAGASSI]

It is necessary to emulate the instructions mov cr *, and lgdt lidt. Note however, for the reason that it concerns only the case of Intel, as AMD implements in SVM extension, full support of real-mode. Thus the module kvm-amd.ko need not emulate the instructions involved. However, for the moment still KVM uses emulation for AMD, because the support of SVM is very new and has not had time to be changed at this level. When this change takes place, it will be possible to manage a significantly larger number of guest systems architecture for AMD. Ultimately, the final reason for the establishment of an emulator is the future management of the MMU.

[Amd Athlon]

Let's finish this overview with a story To avoid a version of binutils newer than this in the file concerning the requirements on the compilation of Linux (Documentation / Changes), the instructions of hardware virtualization extensions are defined directly in the code as follows. For example, instructions for SVM, we have :

 # Define SVM_VMLOAD. Byte 0x0f, 0x01, 0xda "
 # Define SVM_VMRUN. "Byte 0x0f, 0x01, 0xd8"
 # Define SVM_VMSAVE. Byte 0x0f, 0x01, 0xdb "