Detecting memory corruption

摘要

A device identifies, based on a program code instruction, an attempted write access operation to a fenced memory slab, where the fenced memory slab includes an alternating sequence of data buffers and guard buffers. The device assigns read-only protection to the fenced slab and invokes, based on the attempted write access operation, a page fault operation. When a faulting address of the attempted write operation is not an address for one of the multiple data buffers, the device performs a panic routine. When the faulting address of the attempted write operation is an address for one of the multiple data buffers, the device removes the read-only protection for the fenced slab and performs a single step processing routine for the program code instruction.

主权项

1. A method comprising:
determining, by a device, that a program code instruction causes an attempted write access operation to a memory block; determining, by the device and after determining that the program code instruction causes the attempted write access operation, whether a faulting address, associated with the attempted write access operation, has an invalid offset; removing, by the device, a read-only protection for the memory block when the faulting address does not have the invalid offset; and performing, by the device and when the faulting address does not have the invalid offset, a single step processing routine to provide a debug trace for the program code instruction.