Derived keys for connectionless network protocols

By generating derived keys in the network protocol and combining application-specific memory area keys, combining integrity protection, encryption and authentication, the inefficiency problem caused by separate execution in the prior art is solved, and more efficient data transmission is achieved Security.

CN112532386BActive Publication Date: 2024-12-06GOOGLE LLC
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202011320594.6
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Priority Date
2020-02-06
Filing Date
2020-11-23
Publication Date
2024-12-06
Estimated Expiration
2040-11-23

AI Technical Summary

Technical Problem

Existing network protocols separate integrity protection, encryption and authentication, resulting in inefficiency in limited packet space, increasing encoding and processing overhead, and inability to effectively utilize packet space.

Method used

By receiving remote procedure calls (RPCs) on the host and generating derived keys, combined with application-specific memory area keys, a method and system is provided to combine integrity protection, encryption, and authentication, using the AESGCMKDF algorithm to generate derived keys , ensure that only authorized applications can access specific memory areas.

Benefits of technology

Reduces the risk of unauthorized access, improves the security and efficiency of data transmission, and achieves integrity protection, encryption and authentication in a single packet space through combined verification steps, reducing processing overhead.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN112532386B_ABST
    Figure CN112532386B_ABST
Patent Text Reader

Abstract

Aspects of the present disclosure relate to derived keys for connectionless network protocols. A derived key may be provided by receiving, at a host, a remote procedure call (RPC) sent by a remote host in response to a request made by an application executing on the remote host. The host may generate a derived key from a region key associated with an application-specific memory region on the host. The host may transmit the derived key to the remote host.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present disclosure relates to derived keys for connectionless network protocols. Background Art

[0002] Network protocols often implement integrity protection, encryption, and authentication as separate verification steps. Because each verification step incurs the coding and processing overhead associated with the transmission of a single packet over the network, such network protocols may not be able to efficiently utilize the limited packet space. Summary of the invention

[0003] One aspect of the present disclosure provides a method for providing a key. The method may include: receiving, at a host, a remote procedure call (RPC) sent by a remote host in response to a request made by an application executing on the remote host; generating, by the host, a derived key from a region key, the region key being associated with an application-specific memory region on the host; and transmitting, by the host, the derived key to the remote host.

[0004] In some instances, the RPC may include a process identifier (ID) and a network address. In some examples, the process ID is a good enough identifier (GEI). In some examples, the network address is assigned by a network interface card. In some examples, the derived key is further generated based on the process ID and the GEI.

[0005] In some instances, the host may determine whether to accept or reject the RPC before generating the derived key.

[0006] In some instances, the application-specific memory region may correspond to a memory region allocated to one of a plurality of applications executing on the host computer.

[0007] In some examples, the derived key (Kd) is generated by the following formula: Kd = AESGCM KDF (Kr,ADDRESS INITIATOR ,GEI INITIATOR ,TOKENS…), where Kd is the derived key, Kr is the region key associated with the application-specific memory region, ADDRESS INITIATOR is the network identifier of the remote host, GEI INITIATOR is the process ID of the application making the request, TOKENS is an arbitrary arrangement of bits, and AESGCM KDF is a key derivation function including the Advanced Encryption Standard (AES) with the Galois / Counter Mode (GCM) mode of operation for authenticated encryption.

[0008] Another aspect of the present disclosure relates to a system for providing a key. The system may include: a host including one or more processors; and one or more storage devices in communication with the one or more processors, wherein the one or more storage devices contain instructions. The instructions may be configured to cause the one or more processors to receive a remote procedure call (RPC) sent by a remote host in response to a request made by an application executing on the remote host; generate a derived key from a region key associated with an application-specific memory region on the host; and transmit the derived key to the remote host.

[0009] In some examples, the instructions further cause the one or more processors to determine whether to accept or reject the RPC before generating the derived key.

[0010] Another aspect of the present disclosure relates to non-transitory computer-readable storage instructions that, when executed by the one or more processors, may cause the one or more processors to: receive a remote procedure call (RPC) sent by a remote host in response to a request made by an application executing on the remote host; generate a derived key from a region key associated with an application-specific memory region on the host; and transmit the derived key to the remote host. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] The foregoing will be apparent from the more particular description of example embodiments of the present technology, as illustrated in the accompanying drawings. The drawings are not necessarily drawn to scale, emphasis instead being placed upon illustrating embodiments of the present technology.

[0012] Figure 1 is an illustration of an environment including data-specific registered memory areas according to conventional remote memory access (RMA) techniques.

[0013] Figure 2A is an illustration of a system for providing combined integrity protection, encryption, and authentication according to an example implementation.

[0014] Figure 2B is a sequence diagram outlining the generation and use of derived keys for RMA according to some aspects of the present disclosure.

[0015] Figure 3 is a diagram illustrating a cell format of an RMA read request message according to an example implementation.

[0016] Figure 4 It includes Figure 2A Flowchart of an example method performed by an RMA controller in evaluating a received RMA read request message.

[0017] Figure 5 is suitable for Figure 4 is used in the method shown in Figure 2A Flowchart of an example method performed by an RMA controller to generate one or more RMA responses to an authorized RMA read request message.

[0018] Figure 6 is based on Figure 5 An illustration of an example format of an RMA response generated by the method shown.

[0019] Figure 7 is suitable for Figure 2A Flowchart of an example method performed by an RMA controller to generate and transmit an RMA read request message.

[0020] Figure 8 is suitable for Figure 2A Flowchart of an example method performed by an RMA controller to receive one or more RMA responses to a transmitted RMA read request message.

[0021] Fig. 9 is a diagram of a computing system suitable for use with the various described embodiments.

[0022] For clarity, not every component may be labeled in every figure. The drawings are not intended to be drawn to scale. In the various drawings, like reference numbers and names indicate like elements. DETAILED DESCRIPTION

[0023] Multiple physical servers or hosts connected via a network may each execute one or more virtual machines or containers (collectively referred to herein as "virtualized computing instances" or VCIs), as well as a hypervisor or container manager for managing the operation of the virtual machines or containers. A memory region is allocated to each VCI within the physical memory of the corresponding host. Within the host memory region allocated to the VCI, each application hosted by the VCI is allocated a memory region.

[0024] Applications running on hosts connected via a network can directly access each other's data via remote memory access (RMA). For example, using an RMA operation, a first application running on a first host can pull data belonging to a second application executed on a second host directly into the memory allocated to the first application on the first host over the network, as if the retrieved data were available locally to the first host. In existing RMA techniques, before data belonging to a second application can be accessed remotely, the memory area where the data is located must be registered. Registering a memory area generates a region key specific to the registered memory area. The generated region key is a shared secret and is transmitted out-of-band to the second host, for example, via a remote procedure call (RPC). In existing RMA techniques, a registered memory area can be remotely accessed by providing a shared region key for the registered memory area. If the provided region key (or associated data) is successfully authenticated against the generated region key, remote access to the registered memory area will be granted. Otherwise, access to the registered memory area will be denied. The use of region keys can provide a minimum level of security to prevent unauthorized access to registered memory areas. However, an unauthorized third party who obtains the shared area key can potentially gain access to the corresponding registered memory area and the data within the registered memory area.

[0025] The security and integrity of data at rest and in transit are typically ensured by network protocols. Network protocols typically implement integrity protection, encryption, and authentication as separate verification steps. Integrity protection typically includes addressing the potential error rate of the network or the rate at which the transmitted content is randomly corrupted when it is transmitted over the network. Network protocols typically employ checksum-based techniques to detect and correct corrupted content caused by the potential error rate of the network. Because such network protocols are typically layered, checksum-based techniques may be redundant, thereby adding unnecessary costs associated with calculating checksums due to the packet overhead associated with carrying checksums within the transmitted packets, and reducing the efficiency of packet transmission. In addition, attempts to reduce the packet overhead of checksums by using smaller single checksums rather than using precious transmission capacity for larger checksums reduce the effectiveness of checksums.

[0026] Network protocols can use encryption as a means to defensively protect sensitive data from unauthorized third parties when the data is at rest or in transit. Cryptography and encryption techniques can ensure the security of sensitive data so that it can only be decrypted using special algorithms, logical keys, mathematical formulas, and / or a combination of all of the above. Network protocols typically layer data encryption on top of integrity protection of unencrypted data, thereby increasing overall computational and packet overhead. Network protocols typically use authentication as a means to prevent unauthorized access to resources on a per-request or per-connection basis, such as remote access to host physical memory. As mentioned above, using remote access keys and memory area keys to verify access to registered memory areas can provide a minimum level of security to prevent unauthorized access to RMA communications. Network protocols typically perform authentication separately from encryption and integrity protection verification steps. Each of the above-mentioned measures can result in coding and processing overhead associated with a single packet transmission on the network. Therefore, there is a need for a network protocol that effectively uses limited packet space by combining integrity protection, encryption, and authentication into a single verification step.

[0027] As mentioned above, applications running on hosts connected via a network can directly access each other's data via remote memory access (RMA). If the data is located in a registered memory area, it can be accessed remotely. Conventional RMA technology supports remote access using a data-specific registered memory area. Using conventional RMA technology, in order to enable a host to remotely access specific data located in a server physical memory, a memory area allocated for the specific data is registered for remote access, and a region key is generated. The generated region key is specific to the registered memory area and the host. Therefore, the region key can only be used by the specific host to access the registered memory area. In some RMA implementations, remote access is enabled using an application-specific registered memory area (i.e., one or more memory areas assigned to the application are registered). The region key generated for the application-specific registered memory area is specific to the application-specific registered memory area and is independent of any remote host. Therefore, different remote hosts requesting RMA to access the same application-specific registered memory area can use the same shared region key corresponding to the application-specific registered memory area. However, an unauthorized third party who obtains the shared area key could potentially gain access to the corresponding application-specific registered memory area and the data within the application-specific registered memory area.

[0028] The technology described herein provides a scheme in which keys for remote memory access can be derived in each process in a cryptographically unique manner. For example, an application can request access to an application-specific registered memory area on another host through its host. The application's host can initiate a remote procedure call (RPC) that requests a key to enable access to an application-specific registered memory area. In response to the RPC, the host of the application-specific registered memory area can generate and share a derived key based on a region key shared with the application's host. Using the derived key, the application's host can request RMA access to an application-specific registered memory area.

[0029] Derived keys can only be used by the application for which they were created. In this respect, unlike region keys, derived keys that are shared with third-party applications or obtained by unauthorized parties cannot be used; thereby reducing the risk of third-party applications or unauthorized parties gaining unauthorized access to application-specific registered memory regions.

[0030] Figure 1 1 is a diagram of an environment 100 including a data-specific registered memory area according to conventional RMA techniques. As a brief overview, Figure 1 A plurality of hosts interconnected by a network are shown. One or more applications can be executed on each host. Each application executed on the host is allocated a memory area (application memory area) within the physical memory of the host. Each application memory area may include one or more data memory areas, which are allocated to data belonging to the application corresponding to the application memory area. Applications executed on the host can directly access each other's data (data memory area) via remote memory access (RMA).

[0031] Figure 1 A plurality of physical servers or hosts 105 are shown, such as a first host 105a, a second host 105b, a third host 105c, and a fourth host 105d, each including a respective one of a plurality of physical memories 110a-110b (generally referred to as physical memories 110). One or more applications may be executed on each of the hosts 105. Each application executed on a respective one of the hosts 105 is allocated a memory region (application memory region 115) within the physical memory of the host. For example, Figure 1As shown in , the application executed on the first host 105a is allocated with an application memory area 115a within the physical memory 110a of the first host 105a. The first application and the second application executed on the second host 105b are each allocated with a corresponding application memory area 115b and 115c within the physical memory 110b of the second host 105b. The first application and the second application executed on the third host 105c are each allocated with a corresponding application memory area 115d and 115e within the physical memory 110c of the third host 105c. The application executed on the fourth host 105d is allocated with an application memory area 115f within the physical memory 110d of the fourth host 105d.

[0032] One or more data belonging to an application are allocated a memory area (data memory area 120) within the memory area of ​​each application. Figure 1 , the first data and the second data belonging to the application executed on the first host 105a are each allocated a corresponding memory area within the application memory area 115a, such as the data memory area 120a and the data memory area 120b. The first data and the second data belonging to the second application executed on the second host 105b are each allocated a corresponding memory area within the application memory area 115c, such as the data memory area 120c and the data memory area 120d.

[0033] The applications executed on the hosts 105 can remotely access each other's data within the data memory regions 120a, 120b, 120c, and 120d. For example, using the RMA operation, the application executed on the third host 105c can pull the second data (or the data stored in the second data memory region 120b) belonging to the application executed on the first host 105a directly into the application memory region 115d of the third application executed on the third host 105c through the network 150, as if the second data memory region 120b was locally available on the third host 105c. Similarly, using the RMA operation, the second application executed on the fourth host 105d can pull the second data (or the data stored in the memory region 120d) belonging to the second application executed on the second host 105b directly into the application memory region 115f of the application executed on the fourth host 105d, as if the data memory region 120d was locally available on the fourth host 105d.

[0034] Registering memory area supports remote access to the memory area via RMA operation. As mentioned above, data memory area 120a and 120b correspond to the first data and second data belonging to the application executed on the first host 105a. In order to prepare the first data and second data for remote access, their corresponding data storage areas 120a and 120b are registered. Data memory area 120a is registered as the first data-specific registered memory area 170a. Once data memory area 120a is registered, RMA operation can use the memory key corresponding to the first data-specific registered memory area 170a to request access to the first data-specific registered memory area 170a. If the request is granted, the RMA operation can access the data in the first data-specific registered memory area 170a (that is, the first data of the first application executed on the first host 105a). Similarly, data memory area 120b (that is, the second data of the application executed on the first host 105a) is registered as the second data-specific registered memory area 170b.

[0035] In addition to the first and second data-specific registered memory areas 170a and 170b, Figure 1 A third data-specific registered memory area 170c is also shown. As mentioned above, data memory areas 120c and 120d correspond to first data and second data belonging to a second application executing on the second host 105b. Figure 1 As shown in , the data memory area 120c corresponding to the first data belonging to the second application executed on the second host 105b is not registered and therefore cannot be remotely accessed via the RMA operation. However, because the data memory area 120d is registered as the third data-specific registered memory area 170c, the second data belonging to the second application executed on the second host 105b can be remotely accessed via the RMA operation.

[0036] Embodiments according to the present disclosure include application-specific registered memory areas rather than data-specific registered memory areas. Application-specific registered memory areas are created by registering one or more memory areas (i.e., application memory areas) allocated to an application within a host physical memory. Figure 2A Describes application-specific registered memory areas.

[0037] In some embodiments, replay attacks are avoided by using a unique command ID counter as part of the initialization vector (IV) and including an encrypted response with authentication data in the response. When the response is received, the encrypted portion is decrypted. The decryption result is compared with the authentication data included in the response. If the comparison indicates that the response is valid, the decrypted portion is further processed. If the comparison indicates that the response is invalid, the response is deleted. In some embodiments, the region key corresponding to the registered memory area specific to the application forms part of the cryptographic key used for encryption and decryption. In some embodiments, the same region key and / or initialization vector is used to encrypt, decrypt and authenticate all data associated with a successful transaction or remote access request. When the remotely accessed data is transmitted over the network, the method provides both authentication and protection of the remotely accessed data.

[0038] Figure 2A 1 is a diagram illustrating a system 200 for providing combined integrity protection, encryption, and authentication according to an example implementation. The system 200 includes a plurality of physical servers or hosts 205, such as a first host 205a, a second host 205b, a third host 205c, and a fourth host 205d. Each of the plurality of hosts 205 includes a respective one of a plurality of physical memories 210a to 210d (typically physical memories 210). One or more applications may be executed on each of the hosts 205. Each application executed on a respective one of the hosts 205 is allocated one or more memory regions within the physical memory of the respective host, such as application memory region 215. For example, as Figure 2A As shown in , the application executed on the first host 205a is allocated with an application memory area 215a within the physical memory 210a of the first host 205a. The first application and the second application executed on the second host 205b are each allocated with a corresponding application memory area 215b and 215c within the physical memory 210b of the second host 205b. The first application and the second application executed on the third host 205c are each allocated with a corresponding application memory area 215d and 215e within the physical memory 210c of the third host 205c. The application executed on the fourth host 205d is allocated with an application memory area 215f within the physical memory 210d of the fourth host 205d.

[0039] One or more data belonging to an application are allocated a memory area within the corresponding application memory area 215, such as the data memory area 220. For example, Figure 2AAs shown in , the first data and the second data belonging to the application executed on the first host 205a are each allocated with a corresponding memory area within the application memory area 215a, such as the data memory area 220a and the data memory area 220b. The first data and the second data belonging to the second application executed on the second host 205b are each allocated with a corresponding memory area within the application memory area 215c, such as the data memory area 220c and the data memory area 220d.

[0040] The system 200 includes a plurality of application-specific registered memory areas 270, such as a first application-specific registered memory area 270a and a second application-specific registered memory area 270b. Respective region keys 230, such as a first region key 230a and a second region key 230b, each correspond to one of the plurality of application-specific registered memory areas 270, in this case, to the application-specific registered memory areas 270a and 270b. The first region key 230a is generated when an application memory area 215a corresponding to an application executed on a first host 205a is registered as a first application-specific registered memory area 270a. The second region key 230b is generated when an application memory area 215c corresponding to a second application executed on a second host 205b is registered as a second application-specific registered memory area 270b. The first region key 230a and the second region key 230b may remain hidden from each application and one or more remote hosts. In some embodiments, a conventional key generation method may be used to generate region keys.

[0041] like Figure 2A As further shown in FIG. 1 , the first application-specific registered memory area 270a encompasses the entire application memory area 215a, and therefore, includes all data within the application memory area 215a, such as data within data memory areas 220a and 220b. Similarly, the second application-specific registered memory area 225b encompasses the entire application memory area 215c, and therefore, includes all data within the application memory area 215c, including data within data memory areas 220c and 220d.

[0042] Registering memory regions at the application level may result in fewer registered memory regions, and therefore, fewer region keys, such as region keys 230a and 230b stored in hardware. Thus, secure RMA communications between a large number of hosts and applications may be implemented using smaller memories with reduced lookup times. As previously indicated, the region keys generated for an application-specific registered memory region are specific to the application-specific registered memory region 225, rather than to the application-specific registered memory region and remote host endpoint pair. Thus, different remote hosts requesting RMA access to the same application-specific registered memory region may use cryptographically unique derivations of the same shared region key corresponding to the application-specific registered memory region.

[0043] Unlike typical mainframe computers, Figure 2A The host 205 shown in lacks their own internal network interface. Instead, the system 200 includes multiple network interfaces 240, such as a first network interface 240a and a second network interface 240b (typically a network interface 240), which are each connected to multiple hosts 205 and used as network interfaces for these hosts. The network interface can be a network interface card (NIC). In some embodiments, each of the multiple hosts 205 is electrically connected to one of the network interfaces 240 via a serial computer expansion bus or other direct bus connection. Among other functions, each of the network interfaces 240 is each configured to transmit and receive RMA (and other) communications on behalf of the host 205 connected thereto, such as RMA read request messages and RMA read response messages. In addition, in some embodiments, the network interface 240 can be used as a top-of-rack (ToR) switch for multiple hosts 205 connected to the network interface 240.

[0044] To support its RMA function, the first network interface 240a is configured to access a plurality of application-specific registered memory areas, such as first and second application-specific registered memory areas 270a and 270b, of each of the first host 205a and the second host 205b electrically connected to the first network interface 240a. The second network interface 240b is configured to access a plurality of application-specific registered memory areas of each of the third host 205c and the fourth host 205d electrically connected to the second network interface 240b. As previously indicated, each of the first and second application-specific registered memory areas 270a and 270b is assigned to one of the plurality of applications executed on a corresponding one of the hosts 205.

[0045] The first network interface 240a and the second network interface 240b themselves can be connected to each other via an electrical connection (such as a serial computer expansion bus or Ethernet connection) or via an optical link (such as a fiber optic link). In some embodiments with a large number of network interfaces 240, some pairs of network interfaces 240 can be connected via electrical links, while other network interfaces 240 can be connected via optical links.

[0046] As previously indicated, a plurality of zone keys 230, such as zone keys 230a and 230b, are stored on the first network interface 240a within the zone table 215a. The first zone key 230a corresponds to the first application-specific registered memory area 270a, while the second zone key 230b corresponds to the second application-specific registered memory area 270b. Similarly, a plurality of zone keys 230, each corresponding to one of the plurality of application-specific registered memory areas of each of the third host 205c and the fourth host 205d, are stored on the second network interface 240b within the zone table 215b.

[0047] The host 205 can delegate the derivation of the zone key to the requesting application executed on the remote host using the RPC protocol. In this regard, a client application executed on a remote host such as host 205c can request access to an application-specific memory area, such as 215a on host 205a. The remote host 205c can assign a process ID, such as a good enough identifier (GEI), and an identifier such as a network address to the application. The remote host 205c can forward the process ID and address to the host 205a via RPC. Then, the host 205a can accept or reject the RPC of the remote host based on the defined policy. In the case where the host 205a accepts the client application, the host 205a can respond to the RPC with a derived key Kd specific to the client application and the access geometry selected by the server. In some instances, in response to the RPC, access with more than one type of operation can be granted to the client application. In this case, the host may provide several derived keys (Kd), one for each permitted operation type, including primitives that the network interface can perform, such as read, write, short read, etc. Although each derived key is specific to the client application, the derived keys should be kept secret.

[0048] Examples of defined policies may include determining the authenticated users associated with the RPC and determining whether those users are legitimately entitled to access the requested data / application-specific registered memory areas, examining the source address involved and determining whether access is inherently granted or automatically denied to that address, and examining the rate at which access requests are granted and throttling or revoking access to high-rate requesters (which may indicate some form of attack).

[0049] The server-selected access geometry allows the host / server to grant access to only a portion of a memory region to a particular requestor. In this regard, the host can provide a derived key and select a token as the offset field of the request so that the provided derived key is only successfully authenticated for requests restricted to a specific offset range.

[0050] Each derived key can be generated instantaneously using the following equation (1):

[0051] (1)Kd=AESGCM KDF (Kr,ADDRESS INITIATOR , GEI INITIATOR , TOKENS...)

[0052] Where Kd is the derived key, Kr is the region key associated with the application-specific memory region, and ADDRESS INITIATOR is the identifier of the host making the request, GEI INITIATOR is the process ID of the application making the request, TOKENS is an arbitrary arrangement of bits, and AESGCM KDF is a key derivation function (KDF) including the Advanced Encryption Standard (AES) with the Galois / Counter Mode (GCM) mode of operation for authenticated encryption that results in two ciphertexts associated with the content to be encrypted and corresponding authentication data. GEI INITIATOR Controlled by the kernel driver, making it controllable by root on the client application's host. ADDRESS INITIATOR Controlled by the network, as is connectivity to the host itself.

[0053] As described herein, a client application may use one or more of its derived keys to access an application-specific memory region associated with a region key used to generate the derived key. The derived key may continue to work until the host rekeys the application-specific memory region. In this regard, by rekeying the application-specific memory region, the region key associated with the memory region may be changed, thereby invalidating all Kd generated from the previous region key.

[0054] The network interfaces 240 discussed above each include a corresponding RMA controller 250, such as RMA controllers 250a and 250b. In some embodiments, the RMA controllers 250 store the zone table 215 discussed above. In other embodiments, the RMA controllers 250 can access the zone table 215. The RMA controllers 250 are configured to generate (according to instructions from an application on the host 210 connected to the network interface 240), evaluate and respond to RMA read requests and RMA read responses. Each of these functions will be briefly described below, and then described in detail. Figure 5 , Figure 6 , Figure 7 Provide detailed explanation.

[0055] As discussed above, the RMA controller 250 is configured to generate (according to instructions from an application on the host 210 coupled to the network interface 240) and transmit RMA read requests. Figure 2A Each RMA controller 250 in is configured to construct a corresponding command in response to a request initiated by an application executed on a host 205 electrically connected to a corresponding one of the network interfaces 240 (referred to as the "initiator network interface") in which the RMA controller 250 is included. The request is for remote access to data within an identified application-specific registered memory area 270, which corresponds to an application executed on a remote host 205 electrically connected to a corresponding one of the network interfaces 240 (referred to as the "target network interface"). In some embodiments, the initiator network interface is different from the target network interface, a plurality of hosts electrically connected to the initiator initiator network interface are not electrically connected to the target network interface, and a plurality of hosts electrically connected to the target network interface are not electrically connected to the initiator network interface. In some embodiments, the initiator network interface is the same as the target network interface. The constructed command includes an identification of the target host 205, an identification of the application-specific registered memory area 270 in which the requested data is located, an offset of the requested data relative to the memory location of the identified application-specific registered memory area, a size of the requested data, and a message type that identifies the command as a read data request. Figure 2A Each RMA controller 250 in is also configured to generate an RMA read request based on the constructed command.

[0056] The network interfaces 240 discussed above each include a corresponding command ID counter 265a and 265b (collectively referred to as command ID counters 265). The RMA controller 250 is also configured to increment its corresponding command ID counter 265. Once the value of the command ID counter 265 is incremented, it is assigned to the constructed command as its command ID. In some embodiments, the command ID counter 265 is a 64-bit wraparound hardware counter. In some embodiments, the initial value of the command ID counter 265 is set at system boot time. In some embodiments, restarting the network interface 240 does not reset its corresponding command ID counter 265, and their values ​​continue to increase monotonically in response to new commands.

[0057] The network interfaces 240 discussed above each include a corresponding command table 260a and 260b (collectively referred to as command tables 260). Each RMA controller 250 is configured to store a constructed command as an entry for a corresponding command ID within the command table of the network interface of the RMA controller 250. The entry for the command ID of the constructed command includes the constructed command corresponding to the command ID and the number of bytes remaining to be received.

[0058] As discussed above, the RMA controller 250 is configured to transmit one or more RMA read response messages or error response messages in response to receiving an RMA read request message. In some embodiments, the RMA controller 250 is configured to time out the RMA read request message transmitted by the initiator network to the target network if the initiator network does not receive the corresponding RMA response from the target network, or if all requested data are not received within a certain amount of time. Timeout of the RMA read request message provides integrity protection by detecting lost packets. Timeout of the RMA read request message also provides security by avoiding replay attacks (or man-in-the-middle attacks) that may attempt to replay previous communications between the initiator network and the target network. As described above, each RMA controller 250 is configured to store the constructed command as an entry of the corresponding command ID in the command table of the network interface of the RMA controller 250. The entry of the command ID for the constructed command includes the constructed command corresponding to the command ID. In some implementations, once the initiator network interface transmits an RMA read request message, the transmitted RMA read request message is timed out after a preset amount of time has passed.

[0059] As previously indicated, the system 200 provides combined integrity protection, encryption, and authentication. In some embodiments, the system 200 provides authenticated encryption. In some embodiments, to provide authenticated encryption, portions of the RMA read request message are encrypted using a derived key from an application-specific memory area associated with the target application and an initialization vector created by the RMA controller 250 for a particular RMA read request. The initialization vector is included in the RMA read request message in plain text format (i.e., unencrypted), but the derived key is not included in the message. The RMA read request message also includes authentication data created using the derived key and the initialization vector.

[0060] Upon receiving the RMA read request message, the RMA controller 250 uses the derived key 230 stored in the region table 215 associated with the memory region identified in the RMA read request and information from the initialization vector to cryptographically process at least a portion of the RMA read request message. The processing produces decrypted data and authentication data, which is compared with the authentication data in the RMA read request. If the two values ​​match, the request is deemed authorized and authentic, and the request can be responded to.

[0061] Figure 2B 2000 is a sequence diagram illustrating an example of derived zone key generation and distribution between a client and a server. The client includes a network interface card (NIC) 2240b and executes a client application 2202. The client can be a host, such as host 205c, and NIC 2240b can be a network interface, such as network interface 240b. The server includes NIC 2240a and executes server software 2208. The server can be a host, such as host 205a, and NIC 2240a can be a network interface, such as network interface 240a.

[0062] like Figure 2B As further shown in FIG. 2 , the client application 2202 may send an RPC via the client NIC 2240b to request access to an application-specific memory region from the server application. The RPC may include an information set including the client's network address (Addr initiator ), the process ID (PID) of the client application 2202 initiator ) and the operation type (OpType).

[0063] The server may receive the RPC via the server NIC 2240a. As described herein, the server software 2208 may generate a derived key (Kd) using the information set received from the client. The server software 2208 may issue an RPC return to the client application 2202 that includes the derived key (Kd) and a region ID that identifies an application-specific registered memory region assigned to the client in response to the RPC.

[0064] The client NIC 2240b can then use the received Kd to sign all operations (Op), such as read, write, reset key, etc. In this regard, the client application 2202 can forward the Op and Kd to the client NIC 2240b. The client NIC 2240b can then use the obtained Kd to sign the Op. In this regard, the client NIC 2240b uses the obtained Kd to encrypt the Op and then sends the signed Op to the server. The signed Op may include the region, the client's network address (Addr initiator ), the process ID (PID) of the client application 2202 initiator ).

[0065] The server NIC 2240 can then use the information in the signed Op received from the client NIC 2240b to generate a derived key (Kd), as described in more detail herein. Using the generated Kd, the server NIC 2240a can then authenticate and decrypt the signed Op.

[0066] As described further below, a similar authenticated encryption process may be used to create and transmit an RMA response message. Figure 3An example format of an RMA read request message 300 is shown. The RMA read request message 300 includes a target host ID 305, a region ID 315, a region offset 320, a request size 325, an authentication tag 335, and an initialization vector 340. The region ID 315 of the RMA read request message 300 identifies one of a plurality of application-specific registered memory regions that can be accessed by a network interface receiving the RMA read request message 300. The target host ID 305 of the RMA read request message 300 identifies one of the hosts electrically connected to the network interface receiving the RMA read request message 300. Thus, the application-specific registered memory region identified by the region ID 315 is within the physical memory of the host identified by the target host ID 305 that is electrically connected to the network interface receiving the RMA read request message 300. The region offset 320 includes an offset relative to the memory location of the application-specific registered memory region identified by the region ID 315. The memory location of the retrieved data is determined based on the region offset 320 and the starting memory address of the application-specific registered memory region identified by the region ID 315. The size of the retrieved data is specified by the request size 325 of the RMA read request message 300.

[0067] The authentication tag 335 of the RMA read request message 300 includes authentication data such as a message authentication code (MAC). In some embodiments, the RMA read request message 300 may include an encrypted portion. The encrypted portion of the RMA read request message 300 is decrypted using a cryptographic key formed by at least the stored regional key corresponding to the region ID 315. Decryption generates authentication data, which is compared with the authentication data included in the authentication tag 335 of the RMA read request message 300. If the authentication data generated by decryption matches the authentication data included in the authentication tag 335, the RMA read request is considered authorized and authentic, and the request can be responded to. However, due to several reasons, the authentication data generated by decryption may not match the authentication data included in the authentication tag 335. For example, the RMA read request message may have been encrypted using an incorrect derived key, or the content of the RMA read request may have been damaged. Therefore, if the two values ​​do not match, the request is considered unauthorized. In some embodiments, if the RMA read request message is unauthorized, an error message may be transmitted. If the comparison verifies the authentication data included in the authentication tag 335, then it is determined that the decryption is successful and the source of the RMA read request message 300 is authorized to access the application-specific registered memory region identified by the region ID 315. In some embodiments, no portion of the RMA read request is encrypted. In such embodiments, the RMA read request is not decrypted.

[0068] The initialization vector 340 of the RMA read request message 300 includes an initiator host ID 345, a message type 350, a command slot 355, and a command ID 360. The initiator host ID 345 identifies the host that is the source of the RMA read request message 300. In some embodiments, the RMA operations can be RMA read request messages, RMA read response messages, RMA write request messages, RMA write request response messages, RMA request completion, and RMA error response messages. The message type 350 includes a command encoding value for each RMA operation. This ensures that the IV 340 is unique. Refer to the following Figure 5 The command 355 and the command ID 360 included in the RMA read request message 300 are described.

[0069] Use of a derived key corresponding to a particular application-specific registered memory region enables an authorized requesting application to remotely access data for a particular application while preventing the requesting application from attempting to access data for other applications located in other registered memory regions within the host's physical memory. The end-to-end encryption described above prevents unauthorized attacks in which an unauthorized entity gains full access to a network link, thereby enabling the unauthorized entity to eavesdrop on traffic and modify or inject traffic in an optical or electrical network link. Compromise of the network control plane may allow an unauthorized entity to reroute traffic from its destination network endpoint to a compromised endpoint. The end-to-end encryption method also prevents an entity that gains unauthorized access to the network from routing all traffic to a single compromised endpoint.

[0070] As mentioned above, Figure 2A The RMA controller 250 in is also configured to evaluate received RMA read request messages and form responses to received RMA read request messages that are deemed to be authorized. Figure 4 is a flow chart of an example method 400, the method comprising: Figure 2AThe method 400 includes evaluating the received RMA read request message performed by the RMA controller 250 in the RMA controller 250. In brief overview, the method 400 begins by receiving an RMA read request message at a network interface that can access one of a plurality of application-specific registered memory areas of each of a plurality of hosts connected to the network interface (stage 405). The method includes evaluating the authority of the source of the RMA read request message to access the identified application-specific registered memory area using host memory access request information included in the received RMA read request message and a derived key corresponding to the identified application-specific registered memory area stored by the network interface (stage 410). The method 400 includes determining whether the source of the RMA read request message is authorized to access the identified application-specific registered memory area (stage 415). In response to determining that the source of the RMA read request message is authorized to access the identified application-specific registered memory area, the method 600 includes generating at least one RMA response communication that includes data retrieved from an identified memory location within the identified application-specific registered memory area (stage 420). In response to determining that the source of the RMA read request message is not authorized to access the registered application-specific memory area, method 400 includes stopping further processing of the RMA read request message. In some embodiments, an error response is generated and transmitted to the source of the RMA read request message (stage 425).

[0071] Method 400 begins by receiving an RMA read request message at a network interface that can access one of a plurality of application-specific registered memory areas of each of a plurality of hosts connected to the network interface (stage 405). As described above, Figure 2A The network interface 240 in the embodiment is configured to receive the RMA read request message. The network interface that generates and transmits the RMA read request message is called the initiator network interface, and the network interface that receives the RMA read request message is called the target network interface. In some embodiments, the received RMA read request message has Figure 3 See the example format shown in Figure 3 , the received RMA read request message includes region ID 315 and target host ID 30. Region ID 315 identifies an application-specific registered memory region within the physical memory of the host specified by target host ID 305 electrically connected to the target network interface. Figure 2AAs shown in , the corresponding region keys for the application-specific registered memory regions are stored on the network interface in the region table 215. As previously indicated, the region keys for the application-specific registered memory regions are generated when the corresponding application memory regions are registered. The region keys are stored in the region table 215. In response to the RPC of the remote host, as described herein, a derived key can be generated and shared with the remote host or communicated out-of-band.

[0072] The received RMA read request message also includes an initialization vector (IV) 340. As previously indicated, Figure 2A The RMA controller 250 in is configured to create an IV for a specific generated RMA read request message. The RMA controller 250 is also configured to encrypt the portions of the generated RMA read request message using a derived key corresponding to an application-specific registered memory area identified in the generated RMA read request message and the created IV. The received RMA read request message also includes an authentication tag 335. The authentication tag 335 includes authentication data created when the portions of the received RMA read request message are encrypted using the derived key and the IV is included in the RMA read request message. The RMA read request message with the encrypted data is transmitted to the target network interface.

[0073] Method 400 includes: using host memory access request information included in the received RMA read request message and a region key corresponding to the identified application-specific registered memory region stored by the network interface to evaluate the authority of the source of the received RMA read request message to access the identified application-specific registered memory region (stage 410). Figure 2A The RMA controller 250 in is configured to decrypt the encrypted portion of the received RMA read request message using a cryptographic key formed using the IV included in the RMA read request message and the corresponding storage region key of the application-specific registered memory region identified in the RMA read request message. Figure 2A The RMA controller 250 in the RMA read request message is also configured to compare the decrypted result with the authentication data included in the RMA read request message. Decryption generates authentication data, and the authentication data is compared with the authentication data. Figure 3335 of the RMA read request message 300 shown in . If the authentication data generated by decryption matches the authentication data included in the authentication tag 335, the RMA read request is deemed authorized and authentic, and the request can be responded to. However, due to several reasons, the authentication data generated by decryption may not match the authentication data included in the authentication tag 335. For example, the RMA read request message may have been encrypted using an incorrect regional key, or the content of the RMA read request may have been corrupted. Therefore, if the two values ​​do not match, the request is deemed unauthorized. In some embodiments, if the RMA read request message is unauthorized, an error message may be transmitted.

[0074] The method 400 includes determining whether the source of the RMA read request message is authorized to access the identified application-specific registered memory region (stage 415). Figure 2A The RMA controller 250 in is configured to determine whether the source of the received RMA read request message is authorized to access the identified application-specific registered memory area. If the comparison of step 410 verifies the authentication data included in the authentication tag 335 of the received RMA read request message, the RMA controller 250 is configured to determine that the source of the RMA read request message is authorized to access the identified application-specific registered memory area. In addition, if the comparison of step 410 verifies the identity authentication data included in the authentication tag 335 of the received RMA read request message, the decryption result of the encrypted portion of the RMA read request message is considered valid.

[0075] However, if the comparison of step 410 does not verify the authentication data included in the authentication tag 335 of the received RMA read request message, the RMA controller 250 is configured to determine that the source of the RMA read request message is not authorized to access the identified application-specific registered memory area. In addition, if the comparison of step 410 does not verify the authentication data included in the authentication tag 335 of the received RMA read request message, all parts of the RMA read request message are considered invalid.

[0076] In response to determining that the source of the RMA read request message is authorized to access the identified application-specific registered memory area, method 400 includes generating at least one RMA response communication including data retrieved from the identified memory location within the identified application-specific registered memory area (stage 420). Figure 2A The RMA controller 250 in is also configured to retrieve data from the identified memory location within the identified application-specific registered memory area. Figure 3, the RMA read request message includes a region offset 320. The region offset 320 specifies an offset relative to a memory location of the application-specific registered memory region identified by the region ID 315. The memory location of the retrieved data is determined based on the region offset 320 and the starting memory address of the application-specific registered memory region identified by the region ID 315. The size of the retrieved data is specified by a request size 325 of the RMA read request message. Figure 2A The RMA controller 250 in is also configured to form at least one RMA response including the retrieved data. Figure 2A The network interface 240 in is also configured to transmit at least one RMA response including the retrieved data.

[0077] In response to determining that the source of the RMA read request message is not authorized to access the application-specific registered memory area, method 400 includes stopping further processing of the RMA read request message. In some embodiments, an error response is generated and transmitted to the source of the RMA read request message (stage 425). Figure 2A The RMA controller 250 in the RMA read request message is also configured to stop processing the RMA read request message if the source of the RMA read request message is not authorized to access the registered application-specific registered memory area. In some embodiments, the RMA controller 250 also generates an RMA error response indicating that the RMA read request is not authorized.

[0078] As mentioned above, Figure 2A The RMA controller 250 in is further configured to form an RMA response to the received RMA read request message if the source of the RMA read request message is authorized to access the identified application-specific registered memory region. Figure 5 is based on Figure 4 Step 420 and by Figure 2A Flowchart of an example method 500 of forming an RMA response to an authorized RMA read request message performed by the RMA controller 250 . Figure 6 is based on Figure 5 2 is a diagram of an example format of an RMA response 600 generated by method 500 shown in FIG. The two figures will be described together below.

[0079] As mentioned above, Figure 5 is based on Figure 4 Step 420 and by Figure 2AFlowchart of an example method 500 for generating one or more RMA responses to an authorized RMA read request message performed by an RMA controller 250. The method 500 begins by responding to a determination that an RMA read request message received at a network interface is authorized to read the requested data by reading the requested data from an identified memory location within an identified application-specific registered memory area (stage 505). The method 500 includes receiving one or more portions of the requested data (stage 510). The method 500 includes generating one or more RMA response messages, each RMA response message including one or more portions of the received request data and an offset indicating the memory location of the one or more portions of the request data included in the RMA response message relative to the memory location of the request data within the identified application-specific registered memory area (stage 515). The method 500 includes transmitting the one or more generated RMA response messages (520).

[0080] Method 500 begins by responding to a determination that an RMA read request message received at a network interface is authorized to read requested data by reading the requested data from an identified memory location within an identified application-specific registered memory area (stage 505). Figure 2A The RMA controller 250 in the embodiment is configured to respond to a determination that the received RMA read request message is authorized to read the requested data by reading the requested data from an identified memory location within an identified application-specific registered memory area of ​​a target host electrically connected to the network interface. Figure 3 , the example RMA read request message 300 includes a target host ID 305, a region ID 315, a region offset 320, and a request size 325. The target host ID 305 identifies one of a plurality of hosts electrically connected to the network interface. The region ID 315 identifies an application-specific registered memory region within a physical memory of the host identified by the target host ID 305. The region offset 320 specifies an offset relative to a memory location of the application-specific registered memory region identified by the region ID 315 within a physical memory of the target host specified by the target host ID 305. A memory location of the requested data is determined based on the region offset 320 and the memory location of the application-specific registered memory region identified by the region ID 315. The request size 325 specifies the size of the requested data.

[0081] Method 500 includes receiving one or more portions of requested data (stage 510). The requested data may arrive at the network interface in one or more portions. Figure 2AThe RMA controller 250 in the method 500 is also configured to receive one or more portions of the requested data. The method 500 includes generating one or more RMA response messages, each RMA response message including the received one or more portions of the request data and an offset indicating a memory location of the one or more portions of the request data included in the RMA response message relative to a memory location of the request data within the identified application-specific registered memory area (stage 515). Figure 2A The RMA controller 250 in the embodiment is configured to generate one or more RMA response messages. Each RMA response message includes one or more parts of the received request data. Figure 6 is based on Figure 5 5. An illustration of an example format of an RMA response 600 generated by the method 500 shown in FIG. The example RMA response 600 includes an initiator host ID 605, an authentication tag 615. The initiator host ID 605 identifies one of a plurality of hosts electrically connected to the initiator network interface. The authentication tag 615 of the RMA response 600 includes authentication data, such as a message authentication code (MAC).

[0082] The RMA read response 600 also includes a target host ID 625, a message type 630, a command ID 635, and a response byte offset 640. The target host ID 625 identifies the host corresponding to the data included in the RMA read response. The message type 350 is based on the encoding of the corresponding RMA operation and includes a unique value for each stage of the RMA operation. In some embodiments, the RMA operation can be an RMA read request message, an RMA read response message, an RMA write request message, an RMA write request response message, an RMA request completion, and an RMA error response message. The response byte offset 640 specifies an offset relative to the memory location of the requested data. The response byte offset 640 identifies the location of the data portion included in the RMA response message relative to the starting memory address of the requested data. Reference Figure 7 Command ID 635 is described.

[0083] In some embodiments, Figure 2A The RMA controller 250 in the RMA read response is configured to encrypt one or more RMA read responses. The RMA read response is encrypted using a cryptographic key formed by at least a derived key corresponding to the identified application-specific registered memory area and an IV. A portion of the IV includes a target host ID 625, a message type 635, a command ID 635, and a response byte offset 640 included in the RMA read response. Authentication data generated as a result of the encryption of the RMA read response is included in the authentication tag 615 of the RMA response message.

[0084] The method 500 includes transmitting one or more generated RMA response messages (520). Figure 2A The RMA controller 250 in is also configured to transmit the generated one or more RMA response messages. The transmitted one or more RMA response messages may arrive at the destination network interface out of order. Including the response byte offset 640 enables the data in each RMA response message to be written to the correct memory location in the physical memory of the host identified by the initiator host ID 605 of the RMA response message. In addition, including the response byte offset 640 in the IV 620 makes the IV of each RMA response message unique.

[0085] As mentioned above, Figure 2A The RMA controller 250 is also configured to generate and transmit an RMA read request message. Figure 7 is suitable for Figure 2A Flowchart of an example method 700 for generating and transmitting an RMA read request message performed by the RMA controller 250 of the embodiment of the present invention. When an initiator host computer connected to the initiator network interface initiates a request to read data on a target host, the method 700 begins and includes generating a command corresponding to a request to remotely access data initiated by the initiator host (stage 705). The method 700 also includes generating an RMA read request message including an initialization vector (stage 725). The method 700 includes encrypting one or more portions of the RMA read request message (stage 730). The method 700 includes transmitting the generated RMA read request message including one or more portions of the encrypted data (stage 735).

[0086] When an initiator host connected to the initiator network interface initiates a request to read data on a target host, the method 700 begins and includes generating a command corresponding to the request initiated by the initiator host (stage 705). As described above, the RMA controller 250 is configured to construct a command in response to a request initiated by an application executing on one of a plurality of hosts 205 electrically connected to one of the network interfaces 240 (referred to as the "initiator network interface"). The request is for remote access to data within an identified application-specific registered memory area 215, the memory area 215 corresponding to an application executing on one of a plurality of remote hosts 205 electrically connected to a different one of the network interfaces 240 (referred to as the "target network interface"). Each command 280 constructed in response to a request initiated by a host connected to the initiator network interface is stored in a command table 265 of the initiator network interface 250. Each of the stored commands is identified by a corresponding unique command ID. The initiator network interface 240 also includes a command ID counter 260. When constructing a corresponding command for a request initiated by a host connected to the initiator network interface 240, Figure 2AThe RMA controller 250 is also configured to increment the command ID counter 260 of the initiator network interface 240. The resulting value of the command ID counter 260 is used as the unique command ID of the constructed command. The RMA controller 250 is configured to store the constructed command and its corresponding unique command ID in the command table 265 in the location specified by the slot ID. The slot ID is used as an index into the command table 265.

[0087] The method 700 also includes generating an RMA read request message including an initialization vector (stage 725). For a command constructed in response to a request initiated by an application executing on the initiator host, a corresponding RMA read request message is generated. The generated RMA read request message includes an identification of the initiator host, a message type, a constructed command, and an initialization vector of its corresponding command ID. As described above, Figure 2A The RMA controller 250 is configured to generate an RMA read request message corresponding to the command of the build. Figure 3 As discussed, the RMA read request message includes a target host ID 305, a region ID 315, a region offset 320, a request size 325, an authentication tag 335, and an initialization vector 340. The initialization vector 340 includes an initiator host ID 345, a message type 350, a constructed command 355, and a command ID 360. The message type 350 includes a command encoding for the RMA read request message. The command 355 is a command constructed in response to a request initiated by an application executing on the initiator host. The command ID 360 of the initialization vector includes the corresponding command ID of the command.

[0088] Method 700 includes encrypting one or more parts of an RMA read request message (stage 730). One or more parts of the generated RMA read request message are encrypted using a cryptographic key formed by a derived key corresponding to an identified application-specific registered memory area and an initialization vector of the generated RMA read request message. As previously discussed, registering an application memory area as an application-specific registered memory area generates a region key specific to the application-specific registered memory area. The derived key generated at least in part on the region key is transmitted out-of-band to a remote host (such as a launcher host) via a remote procedure call (RPC), for example. The parts of the RMA read request can be encrypted, for example, using AES-GCM or some other form of authenticated encryption, resulting in a ciphertext associated with the content to be encrypted and corresponding authentication data (such as the MAC discussed above). When the parts of the generated RMA read request message are encrypted using the derived key and the IV is included in the RMA read request message, the authentication data is created.

[0089] As mentioned above, Figure 2A The RMA controller 250 is also configured to encrypt one or more portions of the generated RMA read request message. The derived key corresponding to the identified application-specific registered memory area is retrieved from the command table 265 of the initiator network interface 240. As previously discussed, the RMA controller 250 is configured to store the corresponding command of the generated RMA read request message within the command table 265 in a location corresponding to the unique command ID of the command. Also, stored in the command table 265 in a location corresponding to the command ID of the command is a derived key corresponding to the command identified in the generated RMA read request message as Figure 3 The RMA controller 250 is further configured to use the retrieved derived key corresponding to the identified application-specific registered memory region 215 and the initialization vector 340 included in the generated RMA read request message (in Figure 3 340) to encrypt one or more parts of the generated RMA read request message. As previously discussed, the parts of the generated RMA read request message are encrypted using the derived key and IV 340 to create authentication data. The authentication data is included in the RMA read request message as an authentication tag 335 (in Figure 3 ).

[0090] Method 700 includes transmitting an RMA read request message (stage 735). Figure 2A The RMA controller 250 in is also configured to transmit the generated RMA read request including one or more portions of the encrypted data to the target network interface 240. As described above, Figure 2A The RMA controller 250 is also configured to receive one or more RMA responses in response to the transmitted RMA read request message. Figure 8 is suitable for Figure 2AFlowchart of an example method 800 for processing a received RMA response performed by an RMA controller 250 of the invention. Method 800 begins with the initiator network interface transmitting an RMA read request message (stage 805). Method 800 includes timing out a pending RMA request if a preset amount of time has passed after transmitting the RMA request (stage 825). Method 800 includes determining the type of message received (stage 830). If the received message is not an RMA read response, method 800 includes processing the message (stage 890). If it is determined that the received message is an RMA read response, method 800 includes rebuilding an initialization vector for the response message (835). Method 800 includes decrypting the RMA read response and verifying the content (855). Method 800 includes determining whether the result of the decryption is successfully verified (stage 855). If the result of the decryption is not successfully verified, method 800 includes discarding or deleting the RMA read response (stage 895). If the decryption result is successfully verified, method 800 includes writing the data included in the RMA read response to the physical memory of the initiator host (stage 860). The method 800 includes determining whether all bytes of data have been received (stage 865). If all bytes of data have been received, the method 800 includes indicating completion of the RMA read request message corresponding to the transmission of the RMA read response (stage 875). If all bytes of data have not been received, or if completion of the RMA read request message corresponding to the transmission of the RMA read response is indicated, the method 800 includes waiting to receive the next message at the initiator network interface (stage 880).

[0091] Method 800 begins with the initiator network interface transmitting an RMA read request message (stage 805). As previously discussed, the RMA read request message is generated in response to a request for remote access to data initiated by an initiator host connected to the initiator network interface. The RMA read request message may be generated, for example, using Figure 7 The request is generated using method 700 shown in FIG.

[0092] Method 800 also includes receiving a message at the initiator network interface (stage 815). As previously discussed, Figure 2A The RMA controller 250 is configured to receive messages related to a previously transmitted RMA read request message, such as an RMA read response or an error response. Other types of messages include, but are not limited to, network interface status queries and RMA write requests.

[0093] Method 800 includes timing out any pending RMA requests if a preset amount of time has passed (stage 825). As previously discussed, Figure 2AThe RMA controller 250 is further configured to time out the pending RMA request if a preset amount of time has passed after the RMA request has been transmitted. In response to timing out the pending RMA request, Figure 2A The RMA controller 250 is also configured to modify the corresponding entry of the RMA request in the command table 265. Due to the modification, after timing out the pending RMA request, any response arriving at the initiator network interface 240 corresponding to the timed-out RMA request will no longer be able to be decrypted or authenticated, and therefore the response is deleted when received at the initiator network interface 240.

[0094] Method 800 includes determining the type of message received (stage 830). Figure 2A The RMA controller 250 is configured to determine the type of message received by the initiator network interface 240. As referenced Figure 6 As discussed, the RMA Read Response message type 630 includes an encoded value corresponding to the RMA Read Response.

[0095] If the received message is not an RMA read response, the method 800 includes processing the message (stage 890). However, if the received message is an RMA read response, the method 800 includes constructing an initialization vector (835). As previously discussed, prior to transmitting the RMA read response, one or more portions of the RMA read response are encrypted. One or more portions of the RMA read response are encrypted using a cryptographic key formed using a derived key and the initialization vector. However, all of the information required to construct the initialization vector may not be transmitted in the RMA read response. When the RMA read response is received, Figure 2A The RMA controller 250 is configured to decrypt a received RMA read response by reconstructing an initialization vector for the received RMA read response. As previously mentioned, some of the information required to reconstruct the initialization vector may not be transmitted in the RMA read response. By first extracting the command ID 635 from the received RMA read response, the missing information required to reconstruct the initialization vector is determined. Also as previously discussed, when an RMA read request is generated, a unique command ID is assigned to the corresponding command, and both the command and its command ID are stored in the command table 265 at the location specified by the slot ID. The slot ID is used as an index into the command table 265. The extracted command ID 635 and slot ID are used to retrieve a command 280 stored in the command table 265 of the initiator network interface 240. The retrieved command 280 is used for the corresponding RMA read request of the received RMA read response. Refer back to Figure 6, the RMA read response also includes a target host ID 625, a message type 630, and a response byte offset 640. The RMA controller 250 is configured to reconstruct the initialization vector of the received RMA read response based on the retrieved command 280 corresponding to the command ID 635 and the target host ID 625, the message type 630, and the response byte offset 640 included in the RMA read response.

[0096] The method 800 includes decrypting the received RMA read response and verifying the decrypted content (855). Figure 2A The RMA controller 250 is also configured to decrypt the encrypted portion of the received RMA read response using a cryptographic key formed by a derived key corresponding to the identified application-specific registered memory region and the constructed initialization vector. The derived key is retrieved from the command table 265 of the initiator network interface 240 using the command ID 635 of the RMA read response.

[0097] Method 800 includes determining whether the decrypted content was successfully authenticated (stage 855). Figure 2A The RMA controller 250 is configured to compare the decryption result of the received RMA read response with the authentication data included in the RMA read response. Figure 6 As discussed, the authentication tag 615 of the RMA read request message 600 includes authentication data such as a message authentication code (MAC). If the decryption result includes identity authentication data that matches the authentication data included in the authentication tag 615 of the RMA read response, the RMA controller 250 is configured to determine that the decrypted content of the RMA read response is successfully verified. If the authentication data generated by the decryption process does not match the authentication data included in the authentication tag 615 of the RMA read response, the RMA controller 250 is configured to determine that the decrypted content of the RMA read response is not successfully verified. If the two values ​​match, the request is deemed to be authorized and authentic, and the request response can be further processed according to the stage 860 discussed below. If the decrypted content is not successfully verified, the method 800 includes discarding or deleting the RMA read response (stage 895).

[0098] If the decrypted content is successfully verified, the method 800 includes writing the data included in the RMA read response to the physical memory of the initiator host (stage 860). After determining that the decrypted content of the RMA read response is successfully verified, the RMA controller 250 is configured to write the data included in the RMA read response to the physical memory 210 at the memory location of the initiator host 205 based on the response byte offset 640 of the RMA read response. As previously discussed, the command table 265 of the initiator network interface 240 stores entries for each transmitted RMA read request message based on the corresponding command ID of the transmitted RMA read request message. If the corresponding RMA read response has begun to arrive at the initiator network interface 240, the entry identified by the corresponding command ID of the transmitted RMA read request message includes the number of bytes expected to be received. The initial value of the expected number of bytes is set based on the size of the request data identified in the transmitted RMA read request message. Figure 2A The RMA controller 250 is configured to reduce the number of bytes expected to be received for the RMA read request message by the number of data bytes included in the received RMA read response.

[0099] Method 800 includes determining whether all bytes of data have been received (stage 865). Figure 2A The RMA controller 250 is also configured to determine whether all bytes of data requested by the corresponding RMA read request message identified by the command ID included in the received RMA read response have been received.

[0100] If the initiator network interface has received all bytes of data requested by the corresponding RMA read request message, the method 800 includes indicating completion of the corresponding RMA read request message (stage 875). Upon determining that all bytes of data requested by the RMA read request identified by the command ID included in the received RMA read response have arrived at the initiator network interface 240, Figure 2A The RMA controller 250 is configured to indicate the completion of the RMA read request message by modifying the corresponding entry of the command ID within the command table 265 of the initiator network interface 240. Meanwhile, the initiator host 205 connected to the initiator network interface 240 waits for the RMA controller 250 to indicate the completion of the command before using the data from its physical memory 210.

[0101] However, if all bytes of data requested by the corresponding RMA read request have not been received, or completion of the corresponding RMA read request message is indicated, the method 800 includes waiting for receiving a message at the initiator network interface (stage 880). Upon determining that all bytes of data requested by the corresponding RMA read request have not been received, or completion of the corresponding RMA read request message has been indicated, Figure 2A The RMA controller 250 is configured to wait to receive a message at the initiator network interface 240 .

[0102] Fig. 9 900 is a block diagram of an example computing system. In some implementations, the computing system 900 may be implemented Figure 2A is utilized in the host 205 shown in .

[0103] In summary, the computing system 910 includes at least one processor 950 for performing actions according to instructions; and one or more storage devices 970 or 975 for storing instructions and data. The illustrated example computing system 910 includes one or more processors 950, which communicate via a bus 915 with at least one network interface controller 920 having a network interface port 922(an) connected to other computing devices 912(an), a memory 970, and any other devices 980 (e.g., I / O interfaces). The network interface controller can be, for example, Figure 2A 975. In general, the processor 950 will execute instructions received from memory. The processor 950 shown incorporates or is directly connected to a cache memory 975.

[0104] In more detail, the processor 950 can be any logic circuit that processes instructions extracted, for example, from the memory 970 or the cache 975. In many embodiments, the processor 950 is a microprocessor unit or a special-purpose processor. The computing device 900 can be based on any processor or collection of processors capable of operating as described herein. The processor 950 can be a single-core or multi-core processor. The processor 950 can be a plurality of processors. In some embodiments, the processor 950 can be configured to run multi-threaded operations. In some embodiments, the processor 950 can host one or more VCIs and a hypervisor or container manager for managing the operation of the VCI.

[0105] The memory 970 may be any device suitable for storing computer-readable data. The memory 970 may be a device having fixed storage or a device for reading removable storage media. Examples include all forms of non-volatile memory, media, and memory devices, semiconductor memory devices (e.g., EPROM, EEPROM, SDRAM, and flash memory devices), magnetic, magneto-optical, and optical disks (e.g., CD ROM, DVD-ROM, and The computing system 900 may have any number of memory devices 1070. In some implementations, the memory 970 supports virtualized or containerized memory accessible by a VCI execution environment provided by the computing system 910.

[0106] Cache memory 975 is typically a form of computer memory that is placed in close proximity to processor 950 for fast reads. In some implementations, cache memory 975 is part of processor 950 or on the same chip as processor 950. In some implementations, there are multiple levels of cache 975, for example, L2 and L3 cache layers.

[0107] The network interface controller 920 manages data exchange via a network interface 922(an) (also referred to as a network interface port). The network interface controller 920 processes the physical layer and data link layer of the OSI model for network communication. The network interface port 922(an) is a connection point for a physical network link. In some embodiments, the network interface controller 920 supports wireless network connections, and the interface port 922 is a wireless receiver / transmitter. Typically, the computing system 910 exchanges data with other computing devices 912(an) via a physical or wireless link to the network interface 922(an). In some embodiments, the network interface controller 920 implements a network protocol such as Ethernet or other network protocols.

[0108] Other network devices 912(an) are connected to the computing system 910 via respective network interface ports 922. Other network devices 912(an) may be peer computing devices, network devices, or any other computing devices with network functionality. In such an embodiment, the interface controller 920 may function as a top-of-rack (ToR) switch for the computing devices 912(an).

[0109] Other devices 980 may include I / O interfaces, other external serial device ports, and any additional coprocessors. The implementation of the subject matter and operations described in this specification may be implemented in a digital electronic circuit or in a computer software embodied in a tangible medium, firmware, or hardware (including the structure disclosed in this specification and its equivalent structure) or in a combination of one or more thereof. The implementation of the subject matter described in this specification may be implemented as one or more computer programs embodied in a tangible medium, that is, one or more modules of computer program instructions, which are encoded on one or more computer storage media for data processing devices to execute or control the operation of the data processing device. The computer storage medium may be a computer-readable storage device, a computer-readable storage substrate, a random or serial access memory array or device, or a combination of one or more thereof, or included therein. The computer storage medium may also be one or more separate components or media (e.g., multiple CDs, disks, or other storage devices) or included therein. The computer storage medium may be tangible and non-temporary.

[0110] The operations described in this specification may be implemented as operations performed by a data processing apparatus on data stored on one or more computer-readable storage devices or received from other sources.

[0111] A computer program (also referred to as a program, software, software application, script, or code) may be written in any form of programming language (including compiled or interpreted languages, declarative or procedural languages) and may be deployed in any form, including deployment as a standalone program or as a module, component, subroutine, object, or other unit suitable for use in a computing environment. A computer program may, but need not, correspond to a file in a file system. A program may be stored in a portion of a file that holds other programs or data (e.g., one or more scripts stored in a markup language document), in a single file dedicated to the program in question, or in multiple coordinated files (e.g., files storing one or more modules, subroutines, or partial codes). A computer program may be deployed to execute on one computer or multiple computers or one or more VCIs located at one site or distributed at multiple sites and interconnected by a communication network. Examples of communication networks include local area networks ("LANs") and wide area networks ("WANs"), internetworks (e.g., the Internet), and peer-to-peer networks (e.g., ad hoc peer-to-peer networks).

[0112] The processes and logic flows described in this specification can be performed by one or more programmable processors executing one or more computer programs to perform actions by operating on input data and generating output. The processes and logic flows can also be performed by special purpose logic circuits, and the apparatus can also be implemented as special purpose logic circuits, such as FPGAs (field programmable gate arrays) or ASICs (application specific integrated circuits).

[0113] Although this specification contains many specific implementation details, these details should not be interpreted as limitations on the scope of any technology or content that may be claimed, but should be interpreted as descriptions of features of specific embodiments dedicated to specific technologies. Certain features described in this specification in the context of separate embodiments may also be implemented in combination in a single embodiment. On the contrary, the various features described in the context of a single embodiment may also be implemented separately or in any suitable sub-combination in multiple embodiments. Moreover, although the features may be described above as working in certain combinations and even initially claimed as such, one or more features in the claimed combination may be deleted from the combination in some cases, and the claimed combination may be directed to a sub-combination or a variant of the sub-combination.

[0114] Similarly, although operations are depicted in a particular order in the accompanying drawings, this should not be understood as requiring that such operations be performed in the particular order shown or in a continuous order, or that all of the operations shown be performed, to achieve the desired results. In some cases, multitasking and parallel processing may be advantageous. In addition, the separation of various system components in the above-described embodiments should not be understood as requiring such separation in all embodiments, and it should be understood that the described program components and systems can generally be integrated together in a single software product or packaged into multiple software products.

[0115] References to "or" may be interpreted as inclusive, such that any term described using "or" may indicate any of a single, more than one, and all of the described terms. Labels "first," "second," "third," etc. do not necessarily indicate an order and are generally used only to distinguish similar items or elements.

[0116] Various modifications to the embodiments described in this disclosure will be apparent to those skilled in the art, and the general principles defined herein may be applied to other embodiments without departing from the spirit or scope of the disclosure. Therefore, the claims are not intended to be limited to the embodiments shown herein, but should be given the broadest scope consistent with the disclosure, the principles disclosed herein, and the novel features.

Claims

1. A method for providing a key, include: receiving, at a host, a remote procedure call (RPC) sent by a remote host in response to a request made by an application executing on the remote host, wherein the RPC includes a process identifier (ID) of the requesting application and a network identifier of the remote host, wherein the network identifier is a network address assigned by a network interface card of the remote host; generating, by the host, a derived key from a region key, the network identifier of the remote host, and the process ID of the requesting application, the region key being associated with an application-specific memory region on the host; as well as transmitting, by the host, the derived key to the remote host, The derived key (Kd) is generated by the following formula: Kd=AESGCM KDF (Kr,ADDRESS INITIATOR ,GEI INITIATOR ,TOKENS…), Where Kd is the derived key, Kr is the region key associated with the application-specific memory region, ADDRESS INITIATOR is the network identifier of the remote host, GEI INITIATOR is the process ID of the application making the request, TOKENS is an arbitrary arrangement of bits, and AESGCM KDF is a key derivation function including the Advanced Encryption Standard (AES) with the Galois / Counter Mode (GCM) mode of operation for authenticated encryption.

2. The method according to claim 1, further comprising: include: Prior to generating the derived key, it is determined by the host whether to accept or reject the RPC.

3. The method according to claim 1, in, The application-specific memory region corresponds to a memory region allocated to one of a plurality of applications executing on the host.

4. A system for providing a key, the system include: A host computer, the host computer comprising one or more processors; as well as one or more storage devices in communication with the one or more processors, wherein the one or more storage devices contain instructions configured to cause the one or more processors to: receiving a remote procedure call (RPC) sent by a remote host in response to a request made by an application executing on the remote host, wherein the RPC includes a process identifier (ID) of the requesting application and a network identifier of the remote host, wherein the network identifier is a network address assigned by a network interface card of the remote host; generating a derived key from a region key, the network identifier of the remote host, and the process ID of the requesting application, the region key being associated with an application-specific memory region on the host; as well as transmitting the derived key to the remote host, The derived key (Kd) is generated by the following formula: Kd=AESGCM KDF (Kr,ADDRESS INITIATOR ,GEI INITIATOR ,TOKENS…), Where Kd is the derived key, Kr is the region key associated with the application-specific memory region, ADDRESS INITIATOR is the network identifier of the remote host, GEI INITIATOR is the process ID of the application making the request, TOKENS is an arbitrary arrangement of bits, and AESGCM KDF is a key derivation function including the Advanced Encryption Standard (AES) with the Galois / Counter Mode (GCM) mode of operation for authenticated encryption.

5. The system according to claim 4, in, The instructions further cause the one or more processors to: A determination is made whether to accept or reject the RPC prior to generating the derived key.

6. The system according to claim 4, in, The application-specific memory region corresponds to a memory region assigned to one of a plurality of applications executing on the host.

7. A non-transitory computer readable medium storing instructions that, when executed by one or more processors of a host, cause the one or more processors to: receiving a remote procedure call (RPC) sent by a remote host in response to a request made by an application executing on the remote host, wherein the RPC includes a process identifier (ID) of the requesting application and a network identifier of the remote host, wherein the network identifier is a network address assigned by a network interface card of the remote host; generating a derived key from a region key, the network identifier of the remote host, and the process ID of the requesting application, the region key being associated with an application-specific memory region on the host; as well as transmitting the derived key to the remote host, The derived key (Kd) is generated by the following formula: Kd=AESGCM KDF (Kr,ADDRESS INITIATOR ,GEI INITIATOR ,TOKENS…), Where Kd is the derived key, Kr is the region key associated with the application-specific memory region, ADDRESS INITIATOR is the network identifier of the remote host, GEI INITIATOR is the process ID of the application making the request, TOKENS is an arbitrary arrangement of bits, and AESGCM KDF is a key derivation function including the Advanced Encryption Standard (AES) with the Galois / Counter Mode (GCM) mode of operation for authenticated encryption.

8. The non-transitory computer readable medium of claim 7, in, The application-specific memory region corresponds to a memory region assigned to one of a plurality of applications executing on the host.