I am seeing some register corruption on a 64-bit PowerPC target and trying to track down the culprit. I suspect there is an interrupt handler using the register without first saving its value on the stack.
I know hardware breakpoints can be triggered for address accesses, but is it possible to trap specific register accesses? In this case it is a floating point register.
Since I can imaging there are a lot of accesses, I think the ideal solution would be to just flag the access using something like LTTng, which would then allow some identification of who is writing during the time period where the corruption is detected (unfortunately the application can't detect corruption until after the fact).