Systems and methods for secure runtime dynamic resizing of memory namespaces
In accordance with embodiments of the present disclosure, an information handling system may include a memory subsystem and a processor subsystem communicatively coupled to the memory subsystem and configured to execute a hypervisor, wherein the hypervisor is configured to host a plurality of virtual machines and host an interface to the memory subsystem, wherein the interface is configured to dynamically modify sizes of namespaces instantiated within the memory subsystem by maintaining super metadata associated with each of a plurality of memory modules of the memory subsystem, wherein the super metadata for a particular memory module of the memory subsystem includes one or more entries for the particular memory module, each entry defining a namespace of the particular memory module and recording an offset of the latest written data in the namespace.
1. An information handling system, comprising:
a memory subsystem; and
a processor subsystem communicatively coupled to the memory subsystem and configured to execute a hypervisor, wherein the hypervisor is configured to host a plurality of virtual machines and host an interface to the memory subsystem, wherein the interface is configured to dynamically modify sizes of namespaces instantiated within the memory subsystem by maintaining super metadata associated with each of a plurality of memory modules of the memory subsystem in an interface table that is accessible by the virtual machines only via a runtime protected service of the hypervisor;
wherein the super metadata for a particular memory module of the memory subsystem includes one or more entries for the particular memory module, each entry defining a namespace of the particular memory module and recording an offset of the latest written data in the namespace.
2. The information handling system of claim 1 , wherein the plurality of memory modules comprise persistent memory modules.
3. The information handling system of claim 2 , wherein the persistent memory modules comprise non-volatile dual-inline memory modules (NVDIMMs).
4. The information handling system of claim 3 , wherein the super metadata is stored as one or more entries in an NVDIMM Firmware Interface Table.
5. The information handling system of claim 1 , wherein super metadata is abstracted from the virtual machines by the runtime protected service.
6. A method comprising:
hosting on a hypervisor a plurality of virtual machines; and
hosting on the hypervisor an interface between a processor subsystem upon which the hypervisor executes and a memory subsystem communicatively coupled to the processor subsystem, wherein the interface is configured to dynamically modify sizes of namespaces instantiated within the memory subsystem by maintaining super metadata associated with each of a plurality of memory modules of the memory subsystem in an interface table that is accessible by the virtual machines only via a runtime protected service of the hypervisor;
wherein the super metadata for a particular memory module of the memory subsystem includes one or more entries for the particular memory module, each entry defining a namespace of the particular memory module and recording an offset of the latest written data in the namespace.
7. The method of claim 6 , wherein the plurality of memory modules comprise persistent memory modules.
8. The method of claim 7 , wherein the persistent memory modules comprise non-volatile dual-inline memory modules (NVDIMMs).
9. The method of claim 8 , wherein the super metadata is stored as one or more entries in an NVDIMM Firmware Interface Table.
10. The method of claim 6 , wherein super metadata is abstracted from the virtual machines by the runtime protected service.
11. An information handling system, comprising:
a memory subsystem; and
a processor subsystem communicatively coupled to the memory subsystem and configured to execute a hypervisor, wherein the hypervisor is configured to host a plurality of virtual machines, a file system, a hardware driver interface to the memory subsystem, and a virtual namespace interposer interfaced between the file system and the hardware driver;
wherein the hypervisor is further configured to maintain super metadata associated with each of a plurality of memory modules of the memory subsystem in an interface table that is accessible by the virtual machines only via a runtime protected service of the hypervisor;
wherein the virtual namespace interposer is configured to:
during an unfragmented state of the memory subsystem, pass through namespace creation requests and input/output requests to the memory subsystem; and
during a fragmented state of the memory subsystem:
in response to a request to create a new namespace, create the new namespace as a virtual namespace mapped to two or more non-contiguous physical block namespaces of the memory subsystem; and
in response to an input/output request to the memory subsystem:
determine if the input/output request is to a virtual namespace;
pass through the input/output request if the request is to a namespace other than a virtual namespace; and
translate the input/output request to map the virtual namespace to the two or more non-contiguous physical block namespaces of the memory subsystem making up the virtual namespace if the input/output request is to a virtual namespace.
12. The information handling system of claim 11 , wherein the memory subsystem comprises persistent memory modules.
13. The information handling system of claim 12 , wherein the persistent memory modules comprise non-volatile dual-inline memory modules (NVDIMMs).
14. The information handling system of claim 13 , wherein the virtual namespace interposer is interfaced between an NVDIMM Firmware Interface Table and the hardware driver.
15. A method comprising, in a memory subsystem communicatively coupled to a processor subsystem configured to execute a hypervisor, wherein the hypervisor is configured to host a plurality of virtual machines, a file system, a hardware driver interface to the memory subsystem, and a virtual namespace interposer interfaced between the file system and the hardware driver, wherein the hypervisor is further configured to maintain super metadata associated with each of a plurality of memory modules of the memory subsystem in an interface table that is accessible by the virtual machines only via a runtime protected service of the hypervisor:
during an unfragmented state of the memory subsystem, passing through, by the virtual namespace interposer, namespace creation requests and input/output requests to the memory subsystem; and
during a fragmented state of the memory subsystem:
in response to a request to create a new namespace, creating, by the virtual namespace interposer, the new namespace as a virtual namespace mapped to two or more non-contiguous physical block namespaces of the memory subsystem; and
in response to an input/output request to the memory subsystem:
determining, by the virtual namespace interposer, if the input/output request is to a virtual namespace;
passing through, by the virtual namespace interposer, the input/output request if the request is to a namespace other than a virtual namespace; and
translating, by the virtual namespace interposer, the input/output request to map the virtual namespace to the two or more non-contiguous physical block namespaces of the memory subsystem making up the virtual namespace if the input/output request is to a virtual namespace.
16. The method of claim 15 , wherein the memory subsystem comprises persistent memory modules.
17. The method of claim 16 , wherein the persistent memory modules comprise non-volatile dual-inline memory modules (NVDIMMs).
18. The method of claim 17 , wherein the virtual namespace interposer is interfaced between an NVDIMM Firmware Interface Table and the hardware driver.