CPU overcommit with guest idle polling

A system comprising:

• a memory;
• a plurality of host central processing units (CPUs) in communication with the memory, wherein the plurality of host CPUs include a first host CPU and a second host CPU;
• a plurality of virtual machines executing on the plurality of host CPUs, wherein the plurality of virtual machines include a first virtual machine;
• and
• a first virtual central processing unit (VCPU) in the first virtual machine;
• a first guest in the first virtual machine;
• and
• a hypervisor executing on the plurality of host CPUs,
• wherein the first VCPU executes on the first host CPU,
• wherein the first guest records in a location of guest memory a value indicating that the first VCPU is in an idle-polling state, and
• wherein the hypervisor:
• responsive to receiving, from the first guest, an address of the location in the guest memory and information specifying which value indicates idle-polling, records the address of the location and the information specifying which value indicates the idle-polling in hypervisor memory,
• detects that a new task needs to be executed;
• reads the value in the location of the guest memory;
• determines that the first VCPU is in the idle-polling state;
• and
• responsive to determining that the first VCPU is in the idle-polling state, schedules the new task to the first host CPU.

A non-transitory machine readable medium storing instructions, which when executed by at least one of a plurality of host central processing units (CPUs), cause the plurality of host CPUs to:

• responsive to receiving, from a first guest, an address of a location in guest memory and information specifying which value indicates idle-polling, record, by a hypervisor executing on the plurality of CPUs, the address of the location and the information specifying which value indicates the idle-polling in hypervisor memory, execute, by the hypervisor, a first virtual central processing unit (VCPU) of a first virtual machine on a first host CPU, wherein the plurality of host CPUs include the first host CPU and a second host CPU, wherein a plurality of virtual machines, which include the first virtual machine, execute on the plurality of host CPUs, and wherein the first guest in the first virtual machine records in the location of the guest memory a value indicating that the first VCPU is in an idle-polling state;
• detect, by the hypervisor, that a new task needs to be executed;
• read, by the hypervisor, the value in the location of the guest memory;
• determine, by the hypervisor, that the first VCPU is in the idle-polling state;
• and
• responsive to determining that the first VCPU is in the idle-polling state, schedule, by the hypervisor, the new task to the first host CPU.