crypto: SHA and AES accelerator bring up for S2

Brings up, fixes and enables AES and SHA hardware acceleration.

Closes IDF-714
Closes IDF-716
This commit is contained in:
Marius Vikhammer
2020-01-16 14:31:10 +08:00
parent 59381b60c0
commit 37369a8a57
32 changed files with 3687 additions and 1550 deletions

View File

@@ -1,4 +1,4 @@
// Copyright 2015-2016 Espressif Systems (Shanghai) PTE LTD
// Copyright 2019-2020 Espressif Systems (Shanghai) PTE LTD
//
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
@@ -16,9 +16,8 @@
#define _ESP_SHA_H_
#include "esp32s2/rom/sha.h"
#include "esp_types.h"
/** @brief Low-level support functions for the hardware SHA engine
/** @brief Low-level support functions for the hardware SHA engine using DMA
*
* @note If you're looking for a SHA API to use, try mbedtls component
* mbedtls/shaXX.h. That API supports hardware acceleration.
@@ -28,15 +27,8 @@
*
* Some technical details about the hardware SHA engine:
*
* - SHA accelerator engine calculates one digest at a time, per SHA
* algorithm type. It initialises and maintains the digest state
* internally. It is possible to read out an in-progress SHA digest
* state, but it is not possible to restore a SHA digest state
* into the engine.
*
* - The memory block SHA_TEXT_BASE is shared between all SHA digest
* engines, so all engines must be idle before this memory block is
* modified.
* - The crypto DMA is shared between the SHA and AES engine, it is not
* possible for them to run calcalutions in parallel.
*
*/
@@ -50,17 +42,11 @@ typedef SHA_TYPE esp_sha_type;
/** @brief Calculate SHA1 or SHA2 sum of some data, using hardware SHA engine
*
* @note For more versatile SHA calculations, where data doesn't need
* to be passed all at once, try the mbedTLS mbedtls/shaX.h APIs. The
* hardware-accelerated mbedTLS implementation is also faster when
* hashing large amounts of data.
* to be passed all at once, try the mbedTLS mbedtls/shaX.h APIs.
*
* @note It is not necessary to lock any SHA hardware before calling
* this function, thread safety is managed internally.
*
* @note If a TLS connection is open then this function may block
* indefinitely waiting for a SHA engine to become available. Use the
* mbedTLS SHA API to avoid this problem.
*
* @param sha_type SHA algorithm to use.
*
* @param input Input data buffer.
@@ -73,157 +59,106 @@ typedef SHA_TYPE esp_sha_type;
*/
void esp_sha(esp_sha_type sha_type, const unsigned char *input, size_t ilen, unsigned char *output);
/* @brief Begin to execute a single SHA block operation
/** @brief Execute SHA block operation using DMA
*
* @note This is a piece of a SHA algorithm, rather than an entire SHA
* algorithm.
*
* @note Call esp_sha_try_lock_engine() before calling this
* function. Do not call esp_sha_lock_memory_block() beforehand, this
* is done inside the function.
* @note Call esp_sha_aquire_hardware() before calling this
* function.
*
* @param sha_type SHA algorithm to use.
*
* @param data_block Pointer to block of data. Block size is
* determined by algorithm (SHA1/SHA2_256 = 64 bytes,
* SHA2_384/SHA2_512 = 128 bytes)
*
* @param is_first_block If this parameter is true, the SHA state will
* be initialised (with the initial state of the given SHA algorithm)
* before the block is calculated. If false, the existing state of the
* SHA engine will be used.
*
* @return As a performance optimisation, this function returns before
* the SHA block operation is complete. Both this function and
* esp_sha_read_state() will automatically wait for any previous
* operation to complete before they begin. If using the SHA registers
* directly in another way, call esp_sha_wait_idle() after calling this
* function but before accessing the SHA registers.
*/
void esp_sha_block(esp_sha_type sha_type, const void *data_block, bool is_first_block);
/* @brief Begin to execute SHA block operation using DMA
*
* @note This is a piece of a SHA algorithm, rather than an entire SHA
* algorithm.
*
* @note Call esp_sha_try_lock_engine() before calling this
* function. Do not call esp_sha_lock_memory_block() beforehand, this
* is done inside the function.
*
* @param sha_type SHA algorithm to use.
*
* @param data_block Pointer to block of data. Block size is
* @param input Pointer to the input data. Block size is
* determined by algorithm (SHA1/SHA2_256 = 64 bytes,
* SHA2_384/SHA2_512 = 128 bytes)
*
* @param ilen length of input data should be multiple of block length.
*
* @param buf Pointer to blocks of data that will be prepended
* to data_block before hashing. Useful when there is two sources of
* data that need to be efficiently calculated in a single SHA DMA
* operation.
*
* @param buf_len length of buf data should be multiple of block length.
* Should not be longer than the maximum amount of bytes in a single block
* (128 bytes)
*
* @param is_first_block If this parameter is true, the SHA state will
* be initialised (with the initial state of the given SHA algorithm)
* before the block is calculated. If false, the existing state of the
* SHA engine will be used.
*
* @param t The number of bits for the SHA512/t hash function, with
* output truncated to t bits. Used for calculating the inital hash.
* t is any positive integer between 1 and 512, except 384.
*
* @return 0 if successful
*/
int esp_sha_dma(esp_sha_type sha_type, const void *data_block, uint32_t ilen, bool is_first_block);
int esp_sha_dma(esp_sha_type sha_type, const void *input, uint32_t ilen,
const void *buf, uint32_t buf_len, bool is_first_block);
/** @brief Read out the current state of the SHA digest loaded in the engine.
/**
* @brief Read out the current state of the SHA digest
*
* @note This is a piece of a SHA algorithm, rather than an entire SHA algorithm.
*
* @note Call esp_sha_try_lock_engine() before calling this
* function. Do not call esp_sha_lock_memory_block() beforehand, this
* is done inside the function.
* @note Call esp_sha_aquire_hardware() before calling this
* function.
*
* If the SHA suffix padding block has been executed already, the
* value that is read is the SHA digest (in big endian
* format). Otherwise, the value that is read is an interim SHA state.
*
* @note If sha_type is SHA2_384, only 48 bytes of state will be read.
* This is enough for the final SHA2_384 digest, but if you want the
* interim SHA-384 state (to continue digesting) then pass SHA2_512 instead.
* value that is read is the SHA digest.
* Otherwise, the value that is read is an interim SHA state.
*
* @param sha_type SHA algorithm in use.
*
* @param state Pointer to a memory buffer to hold the SHA state. Size
* is 20 bytes (SHA1), 32 bytes (SHA2_256), 48 bytes (SHA2_384) or 64 bytes (SHA2_512).
*
* @param digest_state Pointer to a memory buffer to hold the SHA state. Size
* is 20 bytes (SHA1), 32 bytes (SHA2_256), or 64 bytes (SHA2_384, SHA2_512).
*/
void esp_sha_read_digest_state(esp_sha_type sha_type, void *digest_state);
/**
* @brief Obtain exclusive access to a particular SHA engine
* @brief Set the current state of the SHA digest
*
* @param sha_type Type of SHA engine to use.
* @note Call esp_sha_aquire_hardware() before calling this
* function.
*
* Blocks until engine is available. Note: Can block indefinitely
* while a TLS connection is open, suggest using
* esp_sha_try_lock_engine() and failing over to software SHA.
* When resuming a
*
* @param sha_type SHA algorithm in use.
* @param digest_state
*/
void esp_sha_lock_engine(esp_sha_type sha_type);
void esp_sha_write_digest_state(esp_sha_type sha_type, void *digest_state);
/**
* @brief Try and obtain exclusive access to a particular SHA engine
*
* @param sha_type Type of SHA engine to use.
*
* @return Returns true if the SHA engine is locked for exclusive
* use. Call esp_sha_unlock_sha_engine() when done. Returns false if
* the SHA engine is already in use, caller should use software SHA
* algorithm for this digest.
* @brief Enables the SHA and crypto DMA peripheral and takes the
* locks for both of them.
*/
bool esp_sha_try_lock_engine(esp_sha_type sha_type);
void esp_sha_acquire_hardware(void);
/**
* @brief Unlock an engine previously locked with esp_sha_lock_engine() or esp_sha_try_lock_engine()
*
* @param sha_type Type of engine to release.
* @brief Disables the SHA and crypto DMA peripheral and releases the
* locks.
*/
void esp_sha_unlock_engine(esp_sha_type sha_type);
void esp_sha_release_hardware(void);
/*
*/
/**
* @brief Acquire exclusive access to the SHA shared memory block at SHA_TEXT_BASE
* @brief Sets the initial hash value for SHA512/t.
*
* This memory block is shared across all the SHA algorithm types.
* @note Is generated according to the algorithm described in the TRM,
* chapter SHA-Accelerator
*
* Caller should have already locked a SHA engine before calling this function.
* @note The engine must be locked until the value is used for an operation
* or read out. Else you risk another operation overwriting it.
*
* Note that it is possible to obtain exclusive access to the memory block even
* while it is in use by the SHA engine. Caller should use esp_sha_wait_idle()
* to ensure the SHA engine is not reading from the memory block in hardware.
* @param t
*
* @note You do not need to lock the memory block before calling esp_sha_block() or esp_sha_read_digest_state(), these functions handle memory block locking internally.
*
* Call esp_sha_unlock_memory_block() when done.
* @return 0 if successful
*/
void esp_sha_lock_memory_block(void);
/**
* @brief Release exclusive access to the SHA register memory block at SHA_TEXT_BASE
*
* Caller should have already locked a SHA engine before calling this function.
*
* Call following esp_sha_lock_memory_block().
*/
void esp_sha_unlock_memory_block(void);
/** @brief Wait for the SHA engine to finish any current operation
*
* @note This function does not ensure exclusive access to any SHA
* engine. Caller should use esp_sha_try_lock_engine() and
* esp_sha_lock_memory_block() as required.
*
* @note Functions declared in this header file wait for SHA engine
* completion automatically, so you don't need to use this API for
* these. However if accessing SHA registers directly, you will need
* to call this before accessing SHA registers if using the
* esp_sha_block() function.
*
* @note This function busy-waits, so wastes CPU resources.
* Best to delay calling until you are about to need it.
*
*/
void esp_sha_wait_idle(void);
int esp_sha_512_t_init_hash(uint16_t t);
#ifdef __cplusplus
}