Re: [RFA v2] (x86) Fix watchpoint using hardware breakpoint for some distro

[Xavier Roirand <roirand at adacore dot com>]

Hello,

> Earlier you said that TRAP_HWBKPT was the only case that was different
> from current kernels, but turns out that's not accurate.  And now
> I'm left wondering what do those kernels report for the other cases
> you haven't confirmed yet, like hw breakpoints, software breakpoints,
> and regular single steps.  We need a complete picture to figure out
> what the best fix is.

Here's a more complete picture:

Ubuntu 16.04

| what                                     | si_code     | value |
|------------------------------------------+-------------|-------|
| software breakpoints (int3)              | SI_KERNEL   |  0x80 |
| single-steps                             | TRAP_TRACE  |     2 |
| single-stepping a syscall                | TRAP_BRKPT  |     1 |
| user sent SIGTRAP                        | 0           |     0 |
| exec SIGTRAP (when no PTRACE_EVENT_EXEC) | 0           |     0 |
| hardware breakpoints/watchpoints         | TRAP_HWBKPT |     4 |

CentOs 5.11

| what                                     | si_code     | value |
|------------------------------------------+-------------|-------|
| software breakpoints (int3)              | SI_KERNEL   |  0x80 |
| single-steps                             | TRAP_TRACE  |     2 |
| single-stepping a syscall                | TRAP_BRKPT  |     1 |
| user sent SIGTRAP                        | 0           |     0 |
| exec SIGTRAP (when no PTRACE_EVENT_EXEC) | 0           |     0 |
| hardware breakpoints/watchpoints         | TRAP_HWBKPT |     1 |


> > E.g., if hw breakpoints and watchpoints report the same si_code,
> > how do hw breakpoints work with your patch?

In handle_signal_stop function, there's a check done in order to see if 
there's a breakpoint/watchpoint/catchpoint/etc which handle this stop 
event and check if there's a breakpoint/watchpoint/catchpoint where the 
inferior is stopped:

  /* See if there is a breakpoint/watchpoint/catchpoint/etc. that
     handles this event.  */
  ecs->event_thread->control.stop_bpstat
    = bpstat_stop_status (get_current_regcache ()->aspace (),
                          stop_pc, ecs->ptid, &ecs->ws);

Do you think that having a fix handling both cases in save_stop_reason 
would be better (hardware breakpoint & watchpoint) ?

          else
            {
              if (hardware_breakpoint_inserted_here_p (regcache->aspace 
(), pc))
                lp->stop_reason = TARGET_STOPPED_BY_HW_BREAKPOINT;
              else
                check_stopped_by_watchpoint (lp);
            }

Rather than my initial proposal:

          else
            {
                check_stopped_by_watchpoint (lp);
            }

Regards

> Thanks,
> Pedro Alves