Table of contents
Collapse the table of content
Expand the table of content


Last Updated: 11/22/2016

warning C28110: Drivers must protect floating-point hardware state. See use of float

Additional information

Use KeSaveFloatingPointState and KeRestoreFloatingPointState around floating-point operations. Display drivers should use the corresponding Eng... routines.

This warning is only applicable in kernel mode. The driver is attempting to use a variable or constant of a float type when the code is not protected by KeSaveFloatingPointState and KeRestoreFloatingPointState, or EngSaveFloatingPointState and EngRestoreFloatingPointState.

Typically, drivers run with the floating-point context of the most recent application, and any use of a floating point that is not protected by KeSaveFloatingPointState and KeRestoreFloatingPointState can change the results for other processes and can often cause incorrect or unexpected results in the driver.

Display drivers should use EngSaveFloatingPointState and EngRestoreFloatingPointState.

After an instance of this error is detected along any particular flow path, the Code Analysis tool suppresses subsequent similar errors. The Code Analysis tool does not report this error for function definitions that take floating-type arguments or that return a floating type, because the caller will report the use.

This warning can be triggered in error when a program saves and restores the floating-point state around a function call, and the called function performs floating-point operations.

If a function uses floating-point operations intentionally, and is expecting to be called in a context where floating point is safe, you should annotate the function with _Kernel_float_used_. This annotation will both suppress the warnings in the function body and cause the calling context to check that the call is safely protected for floating-point operations. If floating-point operations appear in the arguments or return value, the effect is the same as using _Kernel_float_used_.

By using _Kernel_float_used_ on (or adding the appropriate save and restore calls to) all functions that use floating point until no warnings remain, a driver can be assured to be free of misuse of the floating-point hardware. For more information, see Floating point annotations for drivers.

Send comments about this topic to Microsoft

© 2017 Microsoft