The current implementation of device drivers assumes and relies on
the correctness of the underlying hardware.Any unresponsive hardware,
can cause the application and/or the system to crash or hang.However,
the current technology trends indicate that
we will be becoming increasingly susceptible of hardware failures of both
permanent and transient in nature.Hence, the current implementation of
drivers is least suited to run on unreliable device hardware.
This project intends to solve this
problem by building reliable device drivers that can tolerate hardware
failures and recover if necessary.This project intends to
statically analyse the device drivers and identify
code that assumes correct hardware behavior.These drivers are then
hardened by modifying the code to alleviate this reliance.Time
permitting,
we also intend to build a fault injection framework to test the
robustness of our methodology.We also plan to discuss the possible recovery
strategies that can be performed in cases where the driver cannot continue
to perform normally due to corruption in the driver state based on our
results.