Resource recovery for checkpoint-based high-availability in a virtualized environment

1. A computer program product for resource recovery, comprising:

a non-transitory computer readable storage device; and

program code on the non-transitory computer readable storage device that when executed by a processor within a primary host machine comprising a first virtual machine with the processor, a memory, and an application executing on the processor that accesses the memory, the program code performs a series of functions comprising:

receiving, from the application, a memory mapping of the application within the memory that comprises a plurality of regions including at least one first region and at least one second region;

marking, via a mark purgeable function, the at least one second region as a purgeable memory that will not be tracked, wherein the purgeable memory can be reconstructed by the application at a second virtual machine of a secondary host machine when the purgeable memory is unavailable;

in response to marking the purgeable memory, tracking the at least one first region within the memory mapping for changes;

in response to stopping execution of the first virtual machine:

forwarding only the changes to the at least one first region to the second virtual machine, wherein any changes to the purgeable memory are not forwarded to the secondary host machine; and

in response to completing the forwarding of the changes, resuming execution of the first virtual machine; and

in response to an occurrence of a failure condition on the first virtual machine:

reconstructing the purgeable memory by the application at the second virtual machine; and

signaling the secondary host machine to continue execution of the application using the forwarded changes to the at least one first region.

2. The computer program product of claim 1 , further comprising program code for:

tracking changes to a processor state of the processor;

periodically stopping execution of the first virtual machine;

protecting access to the memory by setting an access code at each of a plurality of first storage locations of the memory that correspond to a second storage location to one of purgeable or non-purgeable;

receiving the marking of the purgeable memory as an application program interface (API) call;

in response to receiving the marking, setting the access code of the purgeable memory to purgeable for a selected first storage location that corresponds to a selected second storage location that comprises the purgeable memory; and

in response to the access code being set as purgeable, preventing the purgeable memory from being forwarded to the secondary host machine when the changes to the at least one first region are forwarded to the second virtual machine.

3. The computer program product of claim 2 , further comprising program code for enabling access by the application to the purgeable memory by:

receiving an access call from the application to access the purgeable memory;

changing, via an auto-removal policy, the access code for the purgeable memory to locked, wherein changing the access code for the purgeable memory to locked prevents the purgeable memory from being disposed of while being accessed;

executing a user function to access the purgeable memory specified by the access call; and

restoring the access code for the purgeable memory to purgeable.

4. The computer program product of claim 3 , further comprising program code for providing a memory protection API that enables access to the memory and prevents direct access to the purgeable memory.

5. The computer program product of claim 3 , further comprising program code for:

in response to receiving the access call, registering a recovery function for the purgeable memory; and

in response to a fault that the purgeable memory is not present caused by attempting to access the purgeable memory, invoking the recovery function to instruct the secondary host machine to reconstruct the purgeable memory by the application on the secondary host machine.

6. The computer program product of claim 1 , further comprising program code for, in response to the occurrence of the failure condition, an operating system of the primary host machine signaling the secondary host machine to:

continue execution of the application using the forwarded changes to the at least one first region; and

instruct the application, while the purgeable memory is unavailable, to perform a reconstructing of the purgeable memory at the secondary host machine.

7. The computer program product of claim 1 , further comprising program code for, in response to the occurrence of the failure condition, a hypervisor of the primary host machine signaling the secondary host machine to:

continue execution of the application using the forwarded changes to the at least one first region; and

instruct the application, while the purgeable memory is unavailable, to perform a reconstructing of the purgeable memory at the secondary host machine.

8. A data processing system, comprising:

a processor that executes an application on a primary host machine comprising a first virtual machine with the processor and a memory; and

a virtualization module that executes on the processor to:

receive, from the application, a memory mapping of the application within the memory that comprises a plurality of regions including at least one first region and at least one second region;

mark, via a mark purgeable function, the at least one second region as a purgeable memory that will not be tracked, wherein the purgeable memory can be reconstructed by the application at a second virtual machine of a secondary host machine when the purgeable memory is unavailable;

in response to marking the purgeable memory, track the at least one first region within the memory mapping for changes;

in response to stopping execution of the first virtual machine:

forward only the changes to the at least one first region to the second virtual machine, wherein any changes to the purgeable memory are not forwarded to the secondary host machine; and

in response to completing the forwarding of the changes, resume execution of the first virtual machine; and

in response to an occurrence of a failure condition on the first virtual machine:

reconstruct the purgeable memory by the application at the second virtual machine; and

signal the secondary host machine to continue execution of the application using the forwarded changes to the at least one first region.

9. The data processing system of claim 8 , wherein the virtualization module further causes the data processing system to:

track changes to a processor state of the processor;

periodically stop execution of the first virtual machine;

protect access to the memory by setting an access code at each of a plurality of first storage locations of the memory that correspond to a second storage location to one of purgeable or non-purgeable;

receive the marking of the purgeable memory as an application program interface (API) call;

in response to receiving the marking, set the access code of the purgeable memory to purgeable for a selected first storage location that corresponds to a selected second storage location that comprises the purgeable memory; and

in response to the access code being set as purgeable, prevent the purgeable memory from being forwarded to the secondary host machine when the changes to the at least one first region are forwarded to the second virtual machine.

10. The data processing system of claim 9 , wherein the virtualization module further causes the data processing system to enable access by the application to the purgeable memory by causing the data processing system to:

receive an access call from the application to access the purgeable memory;

change, via an auto-removal policy, the access code for the purgeable memory to locked, wherein changing the access code for the purgeable memory to locked prevents the purgeable memory from being disposed of while being accessed;

execute a user function to access the purgeable memory specified by the access call; and

restore the access code for the purgeable memory to purgeable.

11. The data processing system of claim 10 , wherein:

the virtualization module further comprises a memory protection API that enables access to the memory and prevents direct access to the purgeable memory; and

the virtualization module further causes the data processing system to:

in response to receiving the access call, register a recovery function for the purgeable memory; and

in response to a fault that the purgeable memory is not present caused by attempting to access the purgeable memory, invoke the recovery function to instruct the secondary host machine to reconstruct the purgeable memory by the application on the secondary host machine.

12. The data processing system of claim 8 wherein:

the virtualization module further causes the data processing system to, in response to the occurrence of the failure condition, signal, via a hypervisor of the primary host machine, the secondary host machine to:

continue execution of the application using the forwarded changes to the at least one first region; and

instruct the application, while the purgeable memory is unavailable, to reconstruct the purgeable memory at the secondary host machine; and

the purgeable memory further comprises at least one of: one or more caches of the application and one or more disposable components of the application that can be reconstructed by the application.

13. The computer program product of claim 6 , further comprising program code for:

using a reference-counting function that enables the operating system to discard memory that is not in use;

wherein the purgeable memory is not paged to disk during a reclaiming operation.

14. The computer program product of claim 5 , further comprising program code for:

in response to signaling the secondary host machine to continue execution of the application using the forwarded changes to the at least one first region:

in response to an occurrence of a failure condition on the second virtual machine outside of an access purgeable call, unmapping the purgeable memory without invoking the recovery function;

detecting a memory fault interrupt on the secondary host machine caused by an operation on the purgeable memory; and

in response to detecting the memory fault interrupt on the secondary host machine:

immediately invoking the recovery function to instruct the secondary host machine to reconstruct the purgeable memory by the application on the secondary host machine; and

retrying the operation on the purgeable memory;

wherein the purgeable memory further comprises at least one of: one or more caches of the application and one or more disposable components of the application that can be reconstructed by the application.

15. The data processing system of claim 12 , wherein the virtualization module further causes the data processing system to:

use a reference-counting function that enables the operating system to discard memory that is not in use;

wherein the purgeable memory is not paged to disk during a reclaiming operation.

16. The data processing system of claim 11 , wherein the virtualization module further causes the data processing system to:

in response to signaling the secondary host machine to continue execution of the application using the forwarded changes to the at least one first region:

in response to an occurrence of a failure condition on the second virtual machine outside of an access purgeable call, unmap the purgeable memory without invoking the recovery function;

detect a memory fault interrupt on the secondary host machine caused by an operation on the purgeable memory; and

in response to detecting the memory fault interrupt on the secondary host machine:

immediately invoke the recovery function to instruct the secondary host machine to reconstruct the purgeable memory by the application on the secondary host machine; and

retry the operation on the purgeable memory.

17. The computer program product of claim 13 , wherein the program code for marking the at least one second region as the purgeable memory further comprises program code for:

using the reference-counting function, determining, for the at least one second region, whether the at least one second region is in use by anyone; and

in response to determining, via the reference-counting function, that the at least one second region is not in use by anyone:

marking the at least one second region as the purgeable memory to prevent the purgeable memory from being paged to a disk when the purgeable memory is being reclaimed; and

discarding the purgeable memory.

18. The computer program product of claim 1 , the program code further comprising program code for:

receiving an identification of a user-specified function to access purgeable code within the purgeable memory;

specifying, within a particular application program interface (API) call, a particular function to be invoked in the event the purgeable memory vanishes due to the failure condition on the first virtual machine, wherein the particular API call modifies an Authority Mask Register (AMR) in-to and out-of the user-specified function to enable the purgeable memory to be accessed normally within the user-specified function, wherein access to the AMR is privileged, and wherein the AMR is accessed using a Special Purpose Register (SPR); and

in response to any removal of the purgeable memory while an operation is in progress on the purgeable memory:

calling a handler function to reinitialize the purgeable memory; and

re-performing the operation.

19. The data processing system of claim 15 , wherein marking the at least one second region as the purgeable memory further comprises the virtualization module executing on the processor to:

use the reference-counting function, determining, for the at least one second region, whether the at least one second region is in use by anyone; and

in response to determining, via the reference-counting function, that the at least one second region is not in use by anyone:

mark the at least one second region as the purgeable memory to prevent the purgeable memory from being paged to a disk when the purgeable memory is being reclaimed; and

discard the purgeable memory.

20. The data processing system of claim 8 , wherein the virtualization module further causes the data processing system to:

receive an identification of a user-specified function to access purgeable code within the purgeable memory;

specify, within a particular application program interface (API) call, a particular function to be invoked in the event the purgeable memory vanishes due to the failure condition on the first virtual machine, wherein the particular API call modifies an Authority Mask Register (AMR) in-toand out-of the user-specified function to enable the purgeable memory to be accessed normally within the user-specified function, wherein access to the AMR is privileged, and wherein the AMR is accessed using a Special Purpose Register (SPR); and

in response to any removal of the purgeable memory while an operation is in progress on the purgeable memory:

call a handler function to reinitialize the purgeable memory; and

re-perform the operation.