1 # Access-controlled NVRAM implementation 2 3 This repository contains various pieces related to the Access-controlled NVRAM 4 HAL. In a nutshell, the Access-controlled NVRAM HAL allows creation of NVRAM 5 spaces that can hold arbitrary data blobs of limited size. Access restrictions 6 can be configured on each NVRAM space to prevent the contents from being 7 accessed or modified, up to the point of requiring full hardware reset to clear 8 a locked NVRAM space. This can be used for various security features that 9 require a trusted storage location for critical data that an attacker can't 10 tamper with. For details of the NVRAM HAL API, see 11 [hardware/libhardware/include/hardware/nvram.h](https://android.googlesource.com/platform/hardware/libhardware/+/master/include/hardware/nvram.h). 12 13 ## [client](/client) 14 15 Contains a simple command-line application to interact with the NVRAM HAL 16 implementation. This is useful for experimentation and use in scripts. 17 `nvram-client` receives commands and parameters on the command line and 18 translates them to NVRAM HAL calls. Results are printed on stdout. 19 20 ## [core](/core) 21 22 A reference implementation of the NVRAM functionality. This reference 23 implementation can be used to create NVRAM HAL implementations that run in a 24 trusted execution environment. It is also the basis for the testing NVRAM HAL 25 module, which implements the entire NVRAM HAL API surface in userspace for the 26 sake of illustration (but obviously doesn't meet the persistence and 27 tamper-evidence requirements). 28 29 Note that the reference implementation leaves persistent storage to be handled 30 by the embedding code, which needs to provide an implementation of the storage 31 interface defined in 32 [system/nvram/core/include/nvram/core/storage.h](core/include/nvram/core/storage.h). 33 34 ## [hal](/hal) 35 36 The [hal](/hal) directory contains glue code that simplifies creation of NVRAM 37 HAL modules. The code implements the API surface specified by the nvram.h HAL 38 header and translates calls into the request/response message format defined in 39 the [messages](/messages) directory. Thus, to create a working NVRAM HAL module, 40 it is sufficient to provide an implementation that understands `nvram::Request` 41 and `nvram::Response` objects, the glue code will adapt it to the full NVRAM API 42 surface. 43 44 ## [messages](/messages) 45 46 Defines an IPC message format that can be used to serialize NVRAM HAL calls and 47 their parameters in preparation for sending them elsewhere (e.g., a TEE) for 48 execution. There is a request and a response struct corresponding to each NVRAM 49 HAL function. The `nvram::Request` and `nvram::Response` wrappers keep track of 50 the actual request or response type, respectively, as well as the request or 51 response parameters specific to the type. 52
