DBZ.GENERAL.FLOAT
Assigned zero float constant value might be used in a division by floating-point zero
An attempt to do a division or modulo operation using zero as the divisor causes a runtime error. Division by zero defects often occur due to ineffective error handling or race conditions, and typically cause abnormal program termination. Before a value is used as the divisor of a division or modulo operation in C/C++ code, it must be checked to confirm that it is not equal to zero.
The DBZ checkers look for instances in which a zero constant value is used as the divisor of a division or modulo operation.
The DBZ.GENERAL.FLOAT checker flags situations in which a variable that has been assigned a zero constant value locally or as the result of a function call, might later be used as a divisor in a division or modulo operation without being checked for zero.
Vulnerability and risk
Float division by zero usually results in the failure of the process or in an exception. It can also result in the success of the operation, but gives an erroneous answer. Floating-point division by zero is more subtle. It depends on the implementation of the compiler. If the compiler is following the IEEE floating-point standard (IEEE 754), then the result of the floating-point division by zero has a well-defined result. However, the C and C++ standards do not enforce compliance to IEEE 754. Thus, floating-point division by zero has an undefined behavior in C and C++ and might result in the failure of the process or in an exception.
Mitigation and prevention
Division by zero issues typically occur due to ineffective exception handling. To avoid this vulnerability, check for a zero value before using it as the divisor of a division or modulo operation.
Vulnerable code example
void foo(){
float y = 0.0f
float res = 10/y; //DBZ.GENERAL.FLOAT
}Klocwork reports DBZ.GENERAL.FLOAT at Line 3 where 'y' might be used in a division by zero.
Fixed code example
void foo(){
float y = 0.0f
if(y != 0.0f)
float res = 10/y; //@no DBZ.GENERAL.FLOAT
}There is no possibility of a DBZ.GENERAL.FLOAT, as the check prevents division by zero.