The FreeRTOS-Plus-TCP library files conform to the MISRA C:2012 guidelines, with the deviations listed below. Compliance is checked with Coverity static analysis. Since the FreeRTOS-Plus-TCP library is designed for small-embedded devices, it needs to have a very small memory footprint and has to be efficient. To achieve that and to increase the performace of the IP-stack, it deviates from some MISRA rules. The specific deviations, suppressed inline, are listed below.

Additionally, MISRA configuration file contains the project wide deviations.

To find the violation references in the source files run grep on the source code with ( Assuming rule 11.4 violation; with justification in point 2 ):

grep 'MISRA Ref 11.4.2' . -rI

Ref 4.12.1

• MISRA C:2012 Directive 4.12: Dynamic memory allocation shall not be used. MISRA warns against the use of dynamic memory allocation as it might lead to undefined behavior if not used properly. However, the FreeRTOS-Plus-TCP library only uses the memory allocation primitives defined by the FreeRTOS-Kernel, which are deterministic. Additionally, proper care is taken in the code to not use free'd pointers and to check the validity of malloc'd memory before it is dereferenced or used.

Ref 4.14.1

• MISRA C:2012 Directive 4.14: The validity of values received from external sources shall be checked. In this case however, the value of uxICMPSize is checked to be within bounds of the network buffer size using the variable uxNeededSize. This is a false positive.

Ref 4.14.2

• MISRA C:2012 Directive 4.14: The validity of values received from external sources shall be checked. In this case however, the contents of xReceiveBuffer.pucPayloadBuffer is filled by FreeRTOS_recvfrom which performs the necessary checks required to validate the xBytes which is in turn assigned to xReceiveBuffer.uxPayloadLength.

Ref 2.2.1

• MISRA C-2012 Rule 2.2 Unions are used for checksum computation to speed up the process by utilizing the full length of registers (32-bits). After this, the 16-bit union members are used to then compute the final checksum. Doing this is considered as 'overwriting the variable' by Coverity. Thus, it marks some statements as dead code. This is a false positive.

Ref 8.9.1

• MISRA C-2012 Rule 8.9 For unit-tests to be repeatable and independent of the order of execution, some variables have file scope definitions rather than function scope.

Ref 8.13.1

• MISRA C-2012 Rule 8.13 Parameter passed is never used, should be declared as const. The argument passed to the prvIPTask function is left unused which is considered as the variable not being used and thus warranting the use of const. However, the FreeRTOS-kernel function xTaskCreate expects a function signature of type void vSomeFunction( void * pvArgs ). To satisfy that requirement, the function signature of prvIPTask does not have a const qualifier in the parameter signature.

Ref 10.5.1

• MISRA C-2012 Rule 10.5 Converting from an unsigned to an enum type. The operation is safe to perform in that case, as we are using a generic API to send and receive data, in that case the exact data sent it is received

Ref 11.1.1

• MISRA C-2012 Rule 11.1 Converting from a void pointer to a function pointer. The FreeRTOS_setsockopt API allows users to configure sockets by setting various options. In order to do so, the function must accept one parameter which, based on the option value, can be casted to the corresponding socket field. To that end, that parameter is of void * type to accommodate all values. The caller of the API is responsible for providing correct function pointer to the API. Thus, this violation can be safely suppressed.

Ref 11.3.1

• MISRA C-2012 Rule 11.3 The data received/sent by the IP stack is represented as a byte stream. This byte stream needs to be casted to various data structures to access certain fields of the packet. However, when casting a byte stream to a structure, MISRA warns us that it can lead to unaligned access. But, in case of FreeRTOS+TCP, the buffer in which the packets are stored are always aligned to a 4 byte word boundary with an offset of 2 bytes. The reason for this 2byte offset is that the ethernet header is of 14 (12 + 2) bytes. Thus,everything except the ethernet header is properly aligned. There is one alignment exception, which is the sender protocol address in the ARP Header. To combat that, the sender protocol address field is declared as an array of 4 bytes instead of a uint32_t. More details can be found here.

Ref 11.4.1

• MISRA c-2012 Rule 11.4 Warns about conversion between a pointer and an integer. Whenever a socket is created using the FreeRTOS_Socket API, either a valid socket (a valid non-NULL pointer) is returned; or FREERTOS_INVALID_SOCKET is returned (which is essentially ~0U) to depict an error in the socket creation process. This conversion from ~0U to a pointer is used to convey the error to various functions. If the pointer is equal to FREERTOS_INVALID_SOCKET, then it is not dereferenced. Thus, this violation can be safely suppressed.

Ref 11.4.2

• MISRA Rule 11.4 The following statement may trigger a: warning: cast increases required alignment of target type [-Wcast-align]. It has been programatically checked that the pointer is well aligned before this point.

Ref 11.4.3

• MISRA Rule 11.4 warns about casting pointer to an integer and vice versa. Here, the pointer to the starting byte of the packet is cast to an integer which is then used to see whether the pointer is well aligned or not. It is not used to access any pointer values. Thus, this violation can be safely suppressed.

Ref 11.6.1

• When sending and receiving a DHCP event to the IP-stack, the events are converted to a void pointer and sent to the IP-task. The function used to send the events handles various events for the IP-task and thus only accepts void pointers. The IP-task converts the void pointer back to the original event. Thus, this rule can be safely suppressed.

Ref 11.6.2

• MISRA Rule 11.6 uintptr_t is guaranteed by the implementation to fit a pointer size of the platform. The pointer has to be moved backward by a constant offset to get to a 'hidden' pointer which is not available for the user to use. This conversion is done to achieve that while avoiding pointer arithmetic.

Ref 11.8.1

• MISRA c-2012 Rule 11.8 warns about removing the const qualifier when assigning one value to another. In this case however, a function pointer is being copied. It doesn't make sense in case of function pointers for the pointee to be const or mutable. Thus, this rule is safe to suppress. 1

Ref 14.3.1

• MISRA C-2012 Rule 14.3 False positive as the value might be changed depending on the conditionally compiled code

Ref 17.2.1

• MISRA C-2012 Rule 17.2 warns about using recursion in software as that can have severe implications on the stack usage and can lead to a serious issue. In this case however, the number of recursions are limited by design. Any socket spawned (child) by a socket in listening state (parent) cannot be in listening state. Thus it is not possible for the child to have a secondary child socket thereby limiting the number of recursive calls to one.

Ref 18.4.1

• MISRA C-2012 Rule 18.4 warns about using arithmetic operators such as +, -, +=, and -= not to be applied to an expression of pointer type. In this case however its ensured by review that, the explicitly calculated pointer value doesn't have the potential to access unintended or invalid memory addresses.

Ref 20.5.1

• MISRA C-2012 Rule 20.5 warns against the use of #undef. FreeRTOS-Plus-TCP allows its users to set some configuration macros to modify the behavior/performance of the library according to their needs. However, the macros values must be within certain bounds. To achieve that, if the macro values lie outside of the bounds, they are undefined using #undef before being redefined to a proper value.

Ref 20.10.1

• MISRA C-2012 Rule 20.10 warns against the use of ## concatenation operator. However, in this case, it must be used to support compile time assertions in case the preprocessor does not suppport sizeof. This operation (assert) has no runtime execution.

Ref 21.6.1

• MISRA C-2012 Rule 21.6 warns about the use of standard library input/output functions as they might have implementation defined or undefined behaviour. The function snprintf is used to insert information in a logging string. This is only used in a utility function which aids in debugging and is not part of the 'core' code governing the functionality of the TCP/IP stack.