On 26/11/2014 14:13, Alex Benn?e wrote:

Paolo Bonzini [off-list ref] writes:

quoted

On 25/11/2014 18:13, Peter Maydell wrote:

quoted

On 25 November 2014 at 17:05, Paolo Bonzini [off-list ref] wrote:

quoted

quoted

So there is no register that says "this breakpoint has triggered" or
"this watchpoint has triggered"?

Nope. You take a debug exception; the syndrome register tells
you if it was a bp or a wp, and if it was a wp the fault address
register tells you the address being accessed (if it was a bp
you know the program counter, obviously). The debugger is expected
to be able to figure it out from there, if it cares.

That's already good enough---do the KVM_DEBUG_EXIT_* constants match the
syndrome register, or if not why?

No they don't. I did consider it at the time but I was wary of pulling
too much over into the uapi headers wholesale. If your happy to do that
I'll include the change in my next version.

I could also rationalise the exit handlers as they all pretty much do
the same thing (save for the exit/syndrome related info). Again I was
keeping things nicely separated in case any particular exception needed
excessive special case handling.

Would you like those changes?

Yes, please.

Paolo