DOMAINS

RELATED APPLICATION

[0001] The present application claims the benefit of the filing dates of U.S. Pat App. 16/520,298, filed July 23, 2019 and entitled "Security Configurations in Page Table Entries for Execution Domains," and Prov. U.S. Pat. App. 62/724,913, filed Aug 30, 2018 and entitled "Security Configurations in Page Table Entries for Execution Domains," the entire disclosure of which is hereby incorporated herein by reference.

FIELD OF THE TECHNOLOGY

[0002] At least some embodiments disclosed herein relate generally to computer architecture and more specifically, but not limited to, security control implemented through configurations specified in page table entries for execution domains.

BACKGROUND

[0003] Instructions programmed for a computer can be structured in layers. One layer can provide resources and services for another layer. For example, a hypervisor can create or provision virtual machines that are implemented on the hardware components of the computer. An operating system can offer resources and services using resources available in a computer having predefined architecture. The computer resources or computer operated upon by the operating system can be actual computer hardware components, or virtual machine components provisioned by a hypervisor. An application can provide application specific functions using the services and resources provided by an operating system.

BRIEF DESCRIPTION OF THE DRAWINGS

[0004] The embodiments are illustrated by way of example and not limitation in the figures of the accompanying drawings in which like references indicate similar elements. [000S] FIG. 1 shows a system to control sandboxing according to some embodiments.

[0006] FIG. 2 shows a shadow stack structure for sandboxing in the system of FIG. 1

[0007] FIG. 3 illustrates a page table entry having a sandboxing configuration for execution domains.

[0008] FIG. 4 shows a computer system having a page table to configure security operations.

[0009] FIG. S shows a method to control shadow stack operations through settings specified in page table entries for execution domains.

DETAILED DESCRIPTION

[0010] The present disclosure includes the techniques of selectively apply security measures to protect the content of a called routine and the content of a calling routine from each other based on non-hierarchicai domains of executions from which the call from the calling routine to the called routine is made and based on settings specified in page table entries. For example, when routine A calls routine B, the security measures can be selectively deployed to protect the data and code of routine A from routine B and/or protect the data and code of routine B from routine A. For example, routine B could be a library routine that performs numerical calculations. When routine B is part of the address space of routine A, it potentially can access the data of routine A. To prevent routine B from functioning as a trojan horse routine, sandboxing operations can be performed to limit the data that routine B can access in the address space of routine A, even when routine A and routine B use the same logical to physical translation tables.

[0011] In a traditional system, different layers of instructions (e.g., user applications vs. operating system) may be given different levels of privilege and/or trust. Conventionally, protection rings have been constructed and implemented in computers to protect data and functionality from fault and malicious behaviors based on a hierarchy of rings. Rings are statically arranged in the hierarchy from most privileged (and thus most trusted) to least privileged (and thus least trusted). For example, the hierarchy can include a ring of operating system kernel that is the most privileged, a ring of device drivers, and a ring of applications that are the least privileged. A program or routine in a lower privilege ring can be limited by a respective special hardware enforced control gate to access the resources and services of a higher privilege ring in the hierarchy. Gating access between rings can improve security.

[0012] In the techniques of the present disclosure, instructions or routines programmed for a computer system can be classified into a set of predefined, non- hierarchical, domains, such as a domain of hypervisor, a domain of operating system, a domain of application, etc. One routine can call another routine stored in memory identified via a virtual memory address. The virtual memory address is translated to physical memory address using one or more page tables. A physical memory region storing the called routine can be explicitly configured via a page table entry to conditionally invoke a security measure (e.g., a shadow stack) to protect its content (e.g., private data pushed on to the call stack) against access by the calling routine and/or to protect the content of the calling routine against access by the called routine. The security measures can be selectively deployed in accordance with the execution domain of the calling routine and a respective setting in the page table entry, without relying upon a static domain hierarchy. Routines of different domains and/or stored in different memory regions can have different security measures for sandboxing calls from different domains. Thus, sandboxing is not restricted to a particular domain.

[0013] FIG. 1 shows a system to control sandboxing according to some embodiments. Sandboxing in general includes a computer measure that isolates the execution of a set of instructions (e.g., an application) from certain system resources and/or other sets of instructions/programs.

[0014] The system of F!G, 1 includes physical memory (109) that can be used to store data and instructions for various routines programmed for a computer system.

[001 S] In general, a routine can include a pre-programmed set of instructions stored in the memory (109). The routine can also have input data, output data, and/or, temporary data stored in the memory (109). A routine can invoke or call another routine (e.g., 1 19) for services and/or resources. The calling routine and the called routine can be in a same domain or different domains (e.g., 101 , 103, ... , 105). Different regions (121 , 123, ... , 125) in the memory (109) can be configured with different sandboxing configurations (e.g., 107) to control the selective deployment of security measures for sandboxing; and each sandboxing configuration (107) for a region (123) can include different settings (1 1 1 , 1 13, ... , 1 15) for respective domains (101 , 103, 105) that invoke calls to called routines (e.g., 1 19) stored in the region (123) The sandboxing configurations (e.g , 107) can be specified, for example, in a page table entry used in logical to physical address translation of virtual memory addresses, such that the structure of the memory regions (121 , 123, 125) can correspond to the memory page structure, as further discussed below in connection with FIG. 3.

[0016] In F!G. 1 , the physical memory (109) is divided into multiple regions (121 , 123, 125). For example, each region (e.g., 123) can be a page of physical memory (109) for memory management, or a set of pages of physical memory (109).

[0617] A typical region Y (e.g., 123) can have a respective set (107) of sandboxing configuration specified for the set of predefined domains (101 , 103,

105). For example, routines of a hypervisor (102) can be classified in a domain A (101); routines of an operating system (104) can be classified in another domain B (103); and routines of applications (106) can be classified in a further domain C (105). A hypervisor or virtual machine monitor (VMM) creates and manages virtual machines. The hypervisor can control basic functions such as physical memory and input/output (I/O). The sandboxing configuration (107) explicitly Identifies whether or not a sandboxing operating is required for a call to execution a routine (e.g., 1 19) stored in the region (123), such as when a routine executed in a domain (101 , 103, ... , or 105) calls the routine (1 19) stored in the region (123). Calls to execute the same routine (1 19) from routines executed in the different domains (101 , 103,

105) can have different settings (1 1 1 , 1 13, ... , 1 15) respectively: and the settings (1 1 1 , 1 13, ... , 1 15) specify whether the calls from the respectively domains (101 ,

103, ... , 105) require sandboxing (e.g., to protect the called routine (119) and the calling routine from each other). Thus, the sandboxing operations can be selectively applied for the execution of the called routine (119) stored in the memory region (123), based on explicit settings (e.g., 1 1 1 , 1 13, ... , 1 15) configured for the respective domains (101 , 103, ... , 105) from which the calls are made, without relying upon a predefined hierarchy of domains (102, 103, .... 105).

[0018] For example, a routine (1 19) in the domain (103) can be programmed for an operating system (104) and configured be stored in the memory region Y (123). When another routine in the domain (101 ) for a hypervisor (102) calls the routine (1 19) stored in the memory region (123), the sandbox setting (1 11 ) specified for the region (123) for calls from the domain (101 ) is checked. Whether or not to invoke a sandboxing operation for the call to the routine (1 19) stored in the memory region (123) can be determined based on the sandbox setting (1 1 1 ) that is specified for the domain (101 ) and for the memory region (123). Thus, the sandboxing operation can be invoked independent of a relative hierarchy between the called domain (103) of the routine (1 19) and the calling domain (101 ).

[0019] Similarly, consider a routine (119) in the domain (103) that is programmed for an operating system (104) and stored in the memory region Y (123). When another routine in the domain (105) for an application (106) calls the routine (1 19) stored in the memory region (123) for execution, the sandbox setting (1 15) specified for the domain (105) to call the region (123) is checked. Whether or not to deploy a sandboxing operating can be determined for the call from an application (106) in the domain (103), to execute the routine (1 19) stored in the memory region (123), can be determined based on the sandbox setting (1 15) specified for the domain (105) and for the memory region (123). Thus, the sandboxing operation can be invoked independent of a relative hierarchy between the calling and called domains (e.g., 105 and 103).

[0020] In general, different routines of a same domain (e.g., 103) can be stored in different regions (e.g., 121 , 123, .., 125) and thus configured to have different sandboxing requirements for calls from a same domain (e.g., 101 , 103, or 105).

[0021] In general, a region (e.g., 123) can store multiple routines (e.g., 1 19) that share the same sandboxing configuration (107).

[0022] Since the sandboxing configurations of FIG. 1 does not rely upon a predefined domain hierarchy of trust (i.e., non-hierarchical), it can provide better flexibility and finer control than the conventional statically defined, hierarchical protection rings.

[0023] FIG. 2 shows a shadow stack structure for sandboxing in the system of FIG. 1.

[0024] In FIG. 2, a calling routine (caller) and a called routine (cailee) can be configured to use separate call stacks (131 and 132) for sandboxing.

[0025] For example, a caller is configured to use a call stack A (131 ) in connection with a set (133) of control registers, such as a stack pointer (231 ), a frame pointer (233), an argument pointer (235). In general, the caller itself can be called by a further routine. Thus, the frame pointer (233) identifies the location of the return address (257) of the caller when the caller returns. The argument pointer (235) identifies the location of the arguments/para eters used by the further routine to invoke the caller.

[0026] In general, the caller can push its private content (255) to the stack A (131 ) during its execution and pop the content (255) off the stack A (131 ) during its execution. For example, the caller content (255) can be pushed onto the stack (131) before the call to execute the ca!lee; and the caller content (255) can be popped off the stack A (131 ) after the execution of the callee returns. The stack pointer (231 ) identifies the top of the stack A (131 ) as used by the caller.

[0027] Before the callee is loaded for execution, the caller can push call parameters (253) for the callee; and a return address (251 ) can also be pushed onto the stack A (131 ).

[0028] When a sandboxing operation is performed to protect the content of the caller from the callee and/or to protect the content of the callee from the caller, a portion of the content of the stack A (131 ) is replicated into a separate stack B (132), including the return address (251 ) and the call parameters (253). The separate stack B (132) can be considered a shadow of the stack A (131 ). The shadow stack B (132) can be used in the execution of the callee.

[0029] Preferably, a separate set (134) of control registers is used in the operations of the stack B (132). For example, a separate stack pointer (241 ), a separate frame pointer (243), and a separate argument point (245) can be used in the execution of the callee using the stack (132).

[0030] During the execution of the callee, the instructions of the callee can push its private content (281 ) to the stack B (132) and pup off items from the stack B (132) when needed. The callee can be limited to use the stack B (132) and be prevented from accessing the separate stack A (131 ) for the caller. Thus, the callee can be physically prevented from accessing the private content (255) of the caller.

[0031] Similarly, the caller can be limited to use the stack A (131 ) and be prevented from accessing the separate stack B (132) for the callee (e.g., via allocating slots on the stack without pushing data on the stack first). Thus, the caller is physically prevented from accessing the private content (261 ) of the callee.

[0032] Once the callee returns (e.g., using the return address (251 ) identified by the separate frame pointer (243)), the caller can continue its operation using the stack A (131 ) and its associated set (133) of control registers. [0033] Optionally, the content in the stack B (132) can be erased upon the callee returns, and/or when the stack B (132) is used during the initiation of a call.

[0034] When the shadow stack structure of FIG, 2 is used, the data security of the callee is improved; and the caller is prevented from accessing the stack data of the callee.

[003S] Optionally, the stack B (132) is not provided with the data of the caller (and its ca!ier(s)). For example, when the stack B (132) is configured for the operation of the callee, the return address (251 ) and the call parameters (253) can be replicated from the stack A (131 ) to the stack B (132); and the other data under the call parameters (253), including the caller content (255), is not replicated from the stack A (131 ) to the stack B (132). Thus, the callee is physically prevented from accessing the private content (255) of the caller and/or other call stack data.

[0036] Whether or not a callee stored in a memory region (123) requires the use of the separate stack (132) can be configured using a sandboxing configuration (107). The sandboxing configuration (107) can have different settings (1 1 1 , 1 13,

1 15) for callers from different domains (101 , 103, ... , 105).

[0637] The sandboxing configuration (107) can be stored as part of a page table entry of the region (123), as illustrated in FIG, 3

[0638] FIG. 3 illustrates a page table entry (153) having a sandboxing

configuration (107) for execution domains (e.g., 101 , 103, ... , 105).

[6639] A typical virtual address (141 ) in a virtual address space (127) can be translated into a corresponding physical address (159) in a physical address space (129) using a page table (151 ). In general, multiple page tables (e.g., 151 ) can be used to map the virtual address space (127) to the physical address space (129).

[0646] The virtual address (141) can include a table ID (143), an entry ID (145), and an offset (147). The table ID (143) can be used to identify a page table (151) that contains a page table entry (153) for a page that contains the memory unit that is identified by the virtual address (141 ) and the physical address (159). The entry ID (145) is used as an index into the page table (151 ) to locate the page table entry (153) efficiently. The page table entry (153) provides a base (157) of the physical address (159). Physical addresses in the same page of memory share the same base (157). Thus, the base (157) identifies the region (123) in the memory (109). The offset (147) of the virtual address (141 ) is used as a corresponding offset (147) in the page or region (123) in the memory (109). The combination of the base (157) and the offset (147) provides the physical address (159) corresponding to the virtual address (141 ).

[0041] In F!G, 3, the page table entry (153) specifies not only the base (157) for the page or region (123), but also the sandboxing configuration (107) for the page or memory region (123), including settings (1 1 1 , 113, 1 15) for the respective domains (101 , 103, ... , 105) illustrated in FIG. 1.

[0042] For example, the sandboxing configuration (107) can include a set of bits (1 1 1 , 1 13, .... 1 15) for the set of domains (101 , 103, ... , 105) respectively. When a sandbox setting bit (e.g., 11 1 , 1 13, ... , or 1 15) is set to have a first value (e.g., 1 or 0), a cal! from a corresponding domain (e.g., 101 , 103, ... , 105) to a routine stored in the region (123) is required to use the shadow stack structure of FIG, 2 to protect the cal!ee content (261) from being accessed by the caller and/or to protect the caller content (255) from being accessed by the callee. When a sandbox setting bit (e.g., 1 1 1 , 1 13, .... or 115) is set to have a second value (e.g., 0 or 1 ), a call from a corresponding domain (e.g., 101 , 103, ... , 105) to a routine stored in the region (123) does not use the shadow stack structure of FIG, 2 to protect the caller and callee from each other; and the caller and callee can share a same set of controller registers (133) and the same stack (131 ).

[0043] Optionally, the page table entry (153) can specify other attributes (155) of the page of physical memory, such as whether the data in the page is valid, whether the page is in main memory, whether the page is dirty (e.g., the changes in data in the page of physical memory have not yet been flushed to a longer-term

memory/storage device relative to the memory region (123)).

[0044] Further, the page table entry (153) can optionally include permission settings for the domains (101 , 103, ... , 105) to access the memory region (123) for various operations, such as read, write, execution, etc. For example, for each domain (101 , 103, ... , or 105), a permission bit in the page table entry (153) can specify whether a routine running in the domain (101 , 103, ... , or 105) can access the memory region (123) defined by the base (157) for a particular type of operation, such as read, write, or execution. For example, the attributes (155) can include a page fault bit indicating whether the page is in the main memory of the computer or in a storage device of the computer if the permission setting allow the current access to the page of memory and the page fault bit indicate that the page is currently not in the main memory of the computer, the memory management unit (181) can swap the page from the storage device into the main memory of the computer to facilitate the access to the page identified by the page table entry (153). However, if the permission settings deny the current access to the page for the current execution domain, it is not necessary to evaluate the page fault bit and/or to swap in the page corresponding to the page table entry (153).

[004S] In general, the table ID (143) can be divided into multiple fields used to locate the page table (151 ) For example, the table ID (143) can include a top table ID identifying a top-level page table and a top table entry ID that is used as an index into the top-level page table to retrieve a page table entry containing an identifier of the page table (151), in a way similar to the entry ID (145) indexing into the page table (151) to identify the page table entry (153) containing the base (157).

[0046] In general, an entry ID (145) can be considered a virtual page number in the page table (151); and the virtual page number (e.g., 145) can be used in the page table (151 ) to look up the page fable entry (153) containing the base (157).

[0047] For example, the table ID (143) can include a set of virtual page numbers that can be used to identify a chain of page tables (e.g., 151 ). Each virtual page number is used as an index in a page table (or page directory) to identify the page table entry (or page directory entry) that contains the identity or base of the next level page table (or page directory).

[0048] In some instances, different running processes in a computer can have different virtual address spaces (e.g., 127); and the process ID of a running process can be used in determine the top-level page fable (or page directory). In some instances, a hash of a portion of the virtual address (141 ), the process ID, and/or an identification of a virtual machine hosted in the computer system can be used to locate the top-level page table (or page directory). In some instances, a hash is used as an index or key to look up a page table entry. Regardless of how the page table entry (153) is located (e.g., via indexing through multiple page tables, via the use of a hash as an index or key), the content of the page table entry (153) can be configured in a way as illustrated in FIG. 3 to provide the sandboxing configure (107) for different domains (101 , 103, ... , 105) to selectively deploy security measures for calling a routine stored in the page/memory region (123) corresponding to the base (157).

[0049] In FUG, 3, the sandboxing configuration (107) for a page or region Y (123) is specified in the bottom-level page table (151 ), where the page table entry (153) in the bottom-level page table (151 ) provides the base (157) of the physical address (159).

[0050] Alternatively, or in combination, higher-level page tables (or page directories) can also have sandboxing configurations for their page table entries (or page directory entries). For example, a page table entry (or page directory entry) identifying the page table (151 ) can have a sandboxing configuration for all of the pages in the page table (151 ); and thus, the domain permission data in the page table entry is applicable to the memory region defined by the page table (151). The hierarchy of sandboxing configurations specified in the chain of page table entries leading to the page table (151 ) and the sandboxing configuration (107) in the bottom- level page table entry (153) can be combined via a logic AND operation or a logic OR operation.

[00S1] For example, a call to a called routine (1 19) from a routine running in a domain (e.g., 101 , 103, ... , 105) can require a sandboxing operation (e.g., using the shadow stack structure of FIG. 2) if all of the sandboxing configurations in the chain of page table entries leading to the base (157), including the bottom-level table entry (153), have the value that requires sandboxing. Alternatively, sandboxing can be required if any of the sandboxing configurations in the chain of page table entries leading to the base (157), including the bottom-level table entry (153), have the value that requires sandboxing. Alternatively, sandboxing is not implemented if none of the sandboxing configurations in the chain of page table entries leading to the base (157), including the bottom-level table entry (153), have the value that requires sandboxing. Alternatively, sandboxing is not implemented if any of the sandboxing configurations in the chain of page table entries leading to the base (157), including the bottom-level fable entry (153), has the value that does not require sandboxing [0052] Optionally, the sandboxing configuration (e.g., 107) is specified in the bottom-level page table (151 ) but not in the higher-level page tables (directories).

[00S3] FIG. 4 shows a computer system having a page table (e.g., 151 ) to configure security operations.

[0054] The computer system of FIG. 4 has a host system (165) coupled to a memory system (161 ) via one or more buses (163). The memory system (161) has memory components (171 , ... , 173).

[0055] For example, the buses (163) can include a memory bus connecting to one or more memory modules and/or include a peripheral internet connecting to one or more storage devices. Some of the memory components (171 , 173) can provide random access; and the some of the memory components (171 , 173) can provide persistent storage capability. Some of the memory components (171 , ... , 173) can be volatile in that when the power supply to the memory component is disconnected temporarily, the data stored in the memory component will be corrupted and/or erased. Some of the memory components (171 , ... , 173) can be non-volatile in that the memory component is capable of retaining content stored therein for an extended period of time without power.

[0056] In general, a memory system (161 ) can also be referred to as a memory device. An example of a memory device is a memory module that is connected to a central processing unit (CPU) via a memory bus. Examples of memory modules include a dual in-line memory module (DIMM), a small outline DIMM (SO-DIMM), a non-volatile dual in-line memory module (NVD!MM), etc. Another example of a memory device is a storage device that is connected to the central processing unit (CPU) via a peripheral interconnect (e.g., an input/output bus, a storage area network). Examples of storage devices include a solid-state drive (SSD), a flash drive, a universal serial bus (USB) flash drive, and a hard disk drive (HDD) in some instances, the memory device is a hybrid memory/storage system that provides both memory functions and storage functions.

[00S7] The memory components (171 , ... , 173) can include any combination of the different types of non-volatile memory components and/or volatile memory components. An example of non-volatile memory components includes a negative- and (NAND) type flash memory with one or more arrays of memory cells such as single level cells (SLCs) or muiti-ievei ceils (MLCs) (e.g., triple level ceils (TLCs) or quad-level ceils (GLCs)). In some instances, a particular memory component can include both an SLC portion and an MLC portion of memory cells. Each of the memory cells can store one or more bits of data (e.g., data blocks) used by the host system (165). Alternatively, or in combination, a memory component (171 , ... , or 173) can include a type of volatile memory in some instances, a memory component (171 , ... , or 173) can include, but is not limited to, random access memory (RAM), read-only memory (ROM), dynamic random access memory (DRAM), synchronous dynamic random access memory (SDRAM), phase change memory (PCM), magneto random access memory (MRAM), Spin Transfer Torque (STT)-MRAM, ferroelectric random-access memory (FeTRAM), ferroelectric RAM (FeRAM), conductive bridging RAM (CBRAM), resistive random access memory (RRAM), oxide based RRAM (OxRAM), negative-or (NOR) flash memory, electrically erasable programmable read-only memory (EERROM), nanowire-based non-volatile memory, memory that incorporates memristor technology, and/or a cross-point array of non-volatile memory cells. A cross-point array of non-volatile memory can perform bit storage based on a change of bulk resistance, in conjunction with a stackable cross-gridded data access array. Additionally, in contrast to many flash- based memories, cross-point non-volatile memory can perform a write in-place operation, where a non-volatile memory cell can be programmed without the non volatile memory cell being previously erased.

[0058] In general, a host system (185) can utilize a memory system (161 ) as physical memory (109) that includes one or more memory components (171 , ... ,

173). The host system (165) can load instructions from the memory system (161 ) for execution, provide data to be stored at the memory system (161 ), and request data to be retrieved from the memory system (161 ).

[0059] In FUG. 4, the host system (165) includes a memory management unit (MMU) (181 ) and a processor (169). The processor (169) has execution units (e.g., 185), such as an arithmetic-logic unit. The processor (189) has registers (183) to hold instructions for execution, data as operands of instructions, and/or results of instruction executions. The processor (169) can have an internal cache (187) as a proxy of a portion of the memory system (161 ).

[0060] In some instances, the host system (165) can include multiple processors (e.g., 169) integrated on a same silicon die as multiple processing cores of a central processing unit (CPU).

[0081] Routines programmed for executing in the processor (169) can be initially stored in the memory system (161 ). The routines can include instructions for a hypervisor (102), an operating system (104), and an application (106). The routines stored initially in the memory system (161 ) can be loaded to the internal cache (187) and/or the registers (183) for execution in the execution units (185).

[0062] The running instances of the routines form the executions (167) of the hypervisor (102), the operating system (104), and the application (106). In some instances, a hypervisor (102) is not used; and the operating system (104) controls the hardware components (e.g., the memory system (161 ), peripheral input/output devices, and/or network interface cards) without a hypervisor. [0063] The executions (167) of the hypervisor (102), the operating system (104), and/or the application (106) access memory (123) (e.g., in memory components (171 , 173)) using virtual memory addresses (e.g., 141 ) defined in one or more virtual memory spaces (e.g., 127). At least one page table (151) (e.g., as illustrated in the FIG. 3) is used to translate the virtual memory addresses (e.g., 141 ) used in the execution to the physical memory addresses (e.g., 159) of the memory components (e.g., 171 , ... , 173).

[0064] As Illustrated in FIG. 1 , the executions of the routines of hypervisor (102), the operating system (104), and the application (106) can be organized into a plurality of domains (101 , 103, ... , 105). For each of the execution domains (101 , 103, ... , 105) and a memory region (123) identified by a page table entry (153), the page table entry (153) identifies a sandboxing configuration (107) having sandbox settings (e.g., 131 , 133, ... , 135) for calling a routine (1 19) stored in the region (123).

[0065] The host system (185) can have a shadow stack structure of FIG, 2 to protect the ca!lee content (261 ) of a called routine (1 19) loaded from the region (123) when the sandbox setting (e.g., 1 1 1 , 1 13, ... , or 1 15) in the page table entry (153) has a predefined value (e.g., 1 ) for the corresponding execution domain (e.g., 101 , 103, ... , or 105) of the caller.

[0066] FIG. 5 shows a method to control shadow stack operations through settings (e.g., 1 1 1 , 1 13, ... , 1 15) specified in page table entries (e.g., 153) for execution domains (101 , 103, ... , 105).

[0067] For example, the method of FIG. 5 can be performed in a computer system of FIG. 4, using a page table (151 ) of FIG. 3, to provide sandbox settings (1 1 1 , 1 13, ... , 1 15) of FIG. 2 for calls from routines in respective execution domains (101 , 103, ... , 105) illustrated in FIG. 1.

[0068] At block 201 , a computer system (e.g., illustrated in FIG. 4) receives a request to call a second routine (1 19) at a virtual memory address (141 ) during an execution of a first routine.

[0069] For example, the first routine can be part of a hypervisor (102), an operating system (104), or an application (106). Thus, the execution of the first routine can be classified as in one of the set of predetermined domains (101 , 103,

... , 105) illustrated in FIG. 1.

[0070] At block 203, the memory management unit (MMLJ) (181) (or the processor (169) of the computer system) determines a page table entry (153) in translating the virtual memory address (141) to a physical memory address (159), as illustrated in FIG. 3.

[0071] At block 205, the memory management unit (MMU) (181) (or the processor (169) of the computer system) identifies, among a plurality of predefined domains (101 , 103, 105), an execution domain (e.g., 101 ) that contains the execution of the first routine.

[0072] For example, memory addresses for loading the instructions of a routine can include an object identifier that determines the domain (e.g., 101 , 103, ... , 105) when the routine is loaded for execution in the processor (169). in other examples, the object identifier is part of the virtual address space and does not specify a domain. In some implementations, the page table entry (153) includes information identifying the domain of routines stored in the memory region (123) identified by the page table entry (153).

[0073] For example, a register (183) of the processor can store the identifier of the domain of a routine while the routine is being executed in the processor (169).

[0074] At block 207, the memory management unit (MMU) (181) (or the processor (169) of the computer system) retrieves, from the page table entry (153), a security setting (e.g., 1 1 1 , 113, ... , or 1 15) specified for the execution domain (e.g., 101 , 103, ... , or 105).

[007 S] For example, the settings (1 1 1 , 113, ... , 1 15) of the sandboxing

configuration (107) can be stored at a predetermined location in the page table entry (153).

[0076] At block 209, the computer system (e.g., illustrated in FIG. 4) runs the second routine (1 19) using a second stack (132) separate from a first stack (131 ) used by the first routine (e.g., in domain 101 , 103, ... , or 105), in response to the security setting (e.g., 1 1 1 , 113, ... , or 1 15) having a predetermined value (e.g., 1 ).

[0677] If the security setting (e.g., 1 11 , 113, ... , or 1 15) does not have the predetermined value (e.g., 1 ), the computer system (e.g., illustrated in FIG. 4) runs the second routine (1 19) using the same stack (131 ) used by the first routine.

[0078] At block 211 , the hardware of the computer system (e.g., illustrated in F!G. 4) prevents the first routine (e.g. , in domain 101 , 103, ... , or 105) from accessing the second stack (132).

[0679] For example, the sandbox setting bits (1 1 1 , 1 13, ... , 1 15) for the respective domains (101 , 103, 105) can be stored at predetermined locations within the page table entry (153). Thus, based on the execution domain of the instructions, the memory management unit (MMU) (181 ) (or the processor (189) of the computer system) can extract a sandbox setting bit (e.g., 1 1 1 , 1 13, ... , or 1 15) and determine whether a sandboxing operation is to be invoked for the call to the called routine (1 19), according to the extracted sandbox setting bit.

[0080] The techniques disclosed herein can be applied to at least to computer systems where processors are separated from memory and processors

communicate with memory and storage devices via communication buses and/or computer networks. Further, the techniques disclosed herein can be applied to computer systems in which processing capabilities are integrated within

memory/storage. For example, the processing circuits, including executing units and/or registers of a typical processor, can be implemented within the integrated circuits and/or the integrated circuit packages of memory media to performing processing within a memory device. Thus, a processor (e.g., 101 ) as discussed above and illustrated in the drawings is not necessarily a central processing unit in the von Neumann architecture. The processor can be a unit integrated within memory to overcome the von Neumann bottleneck that limits computing

performance as a result of a limit in throughput caused by latency in data moves between a central processing unit and memory configured separately according to the von Neumann architecture.

[0081] The description and drawings of the present disclosure are illustrative and are not to be construed as limiting. Numerous specific details are described to provide a thorough understanding. However, in certain instances, well known or conventional details are not described in order to avoid obscuring the description. References to one or an embodiment in the present disclosure are not necessarily references to the same embodiment; and, such references mean at least one.

[0082] In the foregoing specification, the disclosure has been described with reference to specific exemplary embodiments thereof it will be evident that various modifications can be made thereto without departing from the broader spirit and scope as set forth in the following claims. The specification and drawings are, accordingly, to be regarded in an illustrative sense rather than a restrictive sense.