Implement "hw" watchpoints as in Dwarf

[enovella]

https://github.com/iGio90/Dwarf/blob/master/dwarf_debugger/lib/core.js#L4839

[trufae]

are those watchpoints based on hooking getters/setters for the class fields or its like a real hw? any PoC implementation instead of this one that is very tied to the dwarf internals? i think both kind of watchpoints would be interesting to have in r2frida. any hints here @oleavr @mrmacete ?

[trufae]

cc @iGio90

[iGio90]

Well the logic is pretty simple. I think @PinkiePieStyle reworked them to use MemoryAccessMonitor (at least on windows). For what concerne unix, it's a matter of altering the permissions:

• Memory.protect(address, 1, perm_needed(remove perm for read/write for the needed))
• analyze the stack trace provided by Process.setExceptionHandler
• restore correct permissions
• return true on the exception handler
• notify the ui

if needed we are also adding an interceptor to the address which attempt to r/w the given watchpoint address and break it to give runtime instrumentation

[oleavr]

@iGio90 Cool! MemoryAccessMonitor is now also supported on the other platforms in recent versions of Frida.

[trufae]

Thanks @iGio90 for the explanation :) having such primitive stable in Frida will make implementing that much easily. Hope to get that ready before 4.5. i'll start using milestones in r2frida