The chip has internal MOSFETs with body diodes that will act as rectifiers, so when the motor acts as generator, or produces back-EMF, the energy will be sent into the power supply.
Protection measures depend on what kind of motor and mechanical load you have.
In any case, the input caps should have enough capacitance (and low ESL) to absorb the back-EMF, that is an energy of \$ 1/2 Li^2 \$ should only raise supply voltage by a small amount. This is usually the case if you have enough capacitance for the device to work correctly.
However, your motor has inertia, and if it moves a load, that has inertia too. If it lifts a load, then it can also act as a generator if the load goes down. This means the motor can generate a large amount of power and energy when it is used as a brake, either voluntary, or involuntary if the controller crashes or loses power. When this happens, capacitance on the input usually won't be enough to absorb the energy without voltage rising to dangerous levels.
A zener diode won't work because it is unable to dissipate enough power.
One popular solution for braking is to use the H-bridge to short the motor by turning on both bottom FETs. This dissipates the energy in the motor, which is presumably bulky enough and with good cooling. Basically, your controller monitors VPWR, and if it rises too much, it shorts the motor.
That way, even if the power supply is unconnected, if an external force spins the motor, it will act as a generator and power up the board via the VPWR rail. If the logic circuits on the board are powered from the motor voltage VPWR via a DC-DC converter, they will power up even if the power supply is disconnected. The controller should boot quickly enough to monitor the power supply and short the motor before supply voltage rises too much.
This can be implemented in a number of ways, software or hardware, for example with comparators. When power supply voltage is too low to turn on the MOSFETs properly, they should be off, and when voltage is too high, short the motor. I like the hardware solution because it's more relaxing to write and debug code on a board that won't blow up because all the protection features are in software and thus don't work when single-stepping code in the debugger.
If you want to dissipate the energy somewhere else that isn't the motor, then the usual solution is a hysteresis comparator on the supply voltage, that switches a MOSFET, that dissipates extra power in a resistor. This is more expensive, because the resistor needs to dissipate all the power the motor will generate when it is spun by an external force or by the load. But it will allow the motor to generate power into a higher voltage, which provides much stronger braking than just shorting it. So if you need the extra braking power, this is the way to go.
If it is battery powered, and the battery can take the power generated by the motor (mind the max charging current) then it can also be used as an energy sink. But you still need to have some protection in place for when the battery is fully charged and can't take any more current, so it's back to shorting the motor.
