[8/9] Use reinsert_breakpoint for vCont;s

  Pedro Alves <palves@redhat.com> writes:

>> if thread 1 doesn't hit the reinsert breakpoint, we don't have to
>> remove them, because GDB will send vCont;s:1 next time, and GDBserver
>
> There's no guarantee GDB will send vCont;s:1 next time.
> The user may do "continue" instead of "step".
>
>> can only install reinsert breakpoints if they are not installed yet.
>
> The user may even do "return + continue" or "jump", or an infcall,
> all of which resume the thread at a different address from the address
> the thread last stopped.  So there's no guarantee that the
> reinsert breakpoint address makes any sense for the next step request,
> or even that the next resume request is a step in the first place.
>
> Basically the previous step request must be completely forgotten after
> gdb has seen the thread stop.  In all-stop, gdb "sees "all threads
> stopped on each and every event reported to gdb, for any thread.
> A stop reply cancels any and all previous resume requests.

I add some code to delete all reinsert breakpoints for all threads in
all-stop.  See the patch below,

>
>> if thread hits the reinsert breakpoint, but the event is not reported.
>> It becomes pending, and GDBserver will delete the reinsert breakpoints
>> next time when this pending event is reported back to GDB.
>
> I don't follow.  I'm talking about the case where the thread does _not_
> hit the reinsert breakpoint.  Instead some other thread hits some unrelated
> event.

In your review to V2, your words are about other threads hit a breakpoint, but
doesn't mention "thread 1" (requested for resume_step) hits reinsert
breakpoint or not.  I just list two possible results (not hit vs
hit). You didn't talk about the latter, so we don't have to follow it.
  

Ping?  I want this series go in before the release.

On Tue, Jul 5, 2016 at 9:14 AM, Yao Qi <qiyaoltc@gmail.com> wrote:
> Pedro Alves <palves@redhat.com> writes:
>
>>> if thread 1 doesn't hit the reinsert breakpoint, we don't have to
>>> remove them, because GDB will send vCont;s:1 next time, and GDBserver
>>
>> There's no guarantee GDB will send vCont;s:1 next time.
>> The user may do "continue" instead of "step".
>>
>>> can only install reinsert breakpoints if they are not installed yet.
>>
>> The user may even do "return + continue" or "jump", or an infcall,
>> all of which resume the thread at a different address from the address
>> the thread last stopped.  So there's no guarantee that the
>> reinsert breakpoint address makes any sense for the next step request,
>> or even that the next resume request is a step in the first place.
>>
>> Basically the previous step request must be completely forgotten after
>> gdb has seen the thread stop.  In all-stop, gdb "sees "all threads
>> stopped on each and every event reported to gdb, for any thread.
>> A stop reply cancels any and all previous resume requests.
>
> I add some code to delete all reinsert breakpoints for all threads in
> all-stop.  See the patch below,
>
>>
>>> if thread hits the reinsert breakpoint, but the event is not reported.
>>> It becomes pending, and GDBserver will delete the reinsert breakpoints
>>> next time when this pending event is reported back to GDB.
>>
>> I don't follow.  I'm talking about the case where the thread does _not_
>> hit the reinsert breakpoint.  Instead some other thread hits some unrelated
>> event.
>
> In your review to V2, your words are about other threads hit a breakpoint, but
> doesn't mention "thread 1" (requested for resume_step) hits reinsert
> breakpoint or not.  I just list two possible results (not hit vs
> hit). You didn't talk about the latter, so we don't have to follow it.
>
> --
> Yao (齐尧)
>
> From f4aa0593190e7e78831532c03177d8264b7dd1c1 Mon Sep 17 00:00:00 2001
> From: Yao Qi <yao.qi@linaro.org>
> Date: Thu, 19 May 2016 17:07:49 +0100
> Subject: [PATCH] Use reinsert_breakpoint for vCont;s
>
> V4: remove all reinsert breakpoints before sending event back to GDB
>     in all-stop mode,
>
> V3: install breakpoints in proceed_one_lwp,
>     no longer stop all threads when installing breakpoints,
>     delete reinsert breakpoints when GDBserver wants to report event,
>
> V2: fix spaces in changelog entry,
>     use maybe_hw_step,
>     cancel step-over if signal arrives (!maybe_internal_trap),
>
> This patch is to teach GDBserver using software single step to handle
> vCont;s.  Simply speaking, if the thread's resume request is resume_step,
> install reinsert breakpoint at the next pcs when GDBserver is about to
> resume threads.  These reinsert breakpoints of a thread are removed,
> when GDBserver gets an event from that thread and reports it back to
> GDB.
>
> gdb/gdbserver:
>
> 2016-07-05  Yao Qi  <yao.qi@linaro.org>
>
>         * linux-low.c (resume_stopped_resumed_lwps): If resume request
>         is resume_step, call maybe_hw_step.
>         (linux_wait_1): Stop all threads, remove reinsert breakpoints,
>         and unstop them.
>         (linux_resume_one_lwp_throw): Don't assert the thread has reinsert
>         breakpoints or not.
>         (proceed_one_lwp): If resume request is resume_step, install
>         reinsert breakpoints and call maybe_hw_step.
>
> diff --git a/gdb/gdbserver/linux-low.c b/gdb/gdbserver/linux-low.c
> index abaf288..b579b4d 100644
> --- a/gdb/gdbserver/linux-low.c
> +++ b/gdb/gdbserver/linux-low.c
> @@ -2563,7 +2563,10 @@ resume_stopped_resumed_lwps (struct inferior_list_entry *entry)
>        && !lp->status_pending_p
>        && thread->last_status.kind == TARGET_WAITKIND_IGNORE)
>      {
> -      int step = thread->last_resume_kind == resume_step;
> +      int step = 0;
> +
> +      if (thread->last_resume_kind == resume_step)
> +       step = maybe_hw_step (thread);
>
>        if (debug_threads)
>         debug_printf ("RSRL: resuming stopped-resumed LWP %s at %s: step=%d\n",
> @@ -3622,6 +3625,66 @@ linux_wait_1 (ptid_t ptid,
>
>    /* Alright, we're going to report a stop.  */
>
> +  /* Remove reinsert breakpoints.  */
> +  if (can_software_single_step ())
> +    {
> +      /* Remove reinsert breakpoints or not.  It it is true, stop all
> +        lwps, so that other threads won't hit the breakpoint in the
> +        staled memory.  */
> +      int remove_reinsert_breakpoints_p = 0;
> +
> +      if (non_stop)
> +       {
> +         remove_reinsert_breakpoints_p
> +           = has_reinsert_breakpoints (current_thread);
> +       }
> +      else
> +       {
> +         /* In all-stop, a stop reply cancels all previous resume
> +            requests.  Delete all reinsert breakpoints.  */
> +         struct inferior_list_entry *inf, *tmp;
> +
> +         ALL_INFERIORS (&all_threads, inf, tmp)
> +           {
> +             struct thread_info *thread = (struct thread_info *) inf;
> +
> +             if (has_reinsert_breakpoints (thread))
> +               {
> +                 remove_reinsert_breakpoints_p = 1;
> +                 break;
> +               }
> +           }
> +       }
> +
> +      if (remove_reinsert_breakpoints_p)
> +       {
> +         /* If we remove reinsert breakpoints from memory, stop all lwps,
> +            so that other threads won't hit the breakpoint in the staled
> +            memory.  */
> +         stop_all_lwps (0, event_child);
> +
> +         if (non_stop)
> +           {
> +             gdb_assert (has_reinsert_breakpoints (current_thread));
> +             delete_reinsert_breakpoints (current_thread);
> +           }
> +         else
> +           {
> +             struct inferior_list_entry *inf, *tmp;
> +
> +             ALL_INFERIORS (&all_threads, inf, tmp)
> +               {
> +                 struct thread_info *thread = (struct thread_info *) inf;
> +
> +                 if (has_reinsert_breakpoints (thread))
> +                   delete_reinsert_breakpoints (thread);
> +               }
> +           }
> +
> +         unstop_all_lwps (0, event_child);
> +       }
> +    }
> +
>    if (!stabilizing_threads)
>      {
>        /* In all-stop, stop all threads.  */
> @@ -4275,12 +4338,6 @@ linux_resume_one_lwp_throw (struct lwp_info *lwp,
>
>        step = maybe_hw_step (thread);
>      }
> -  else
> -    {
> -      /* If the thread isn't doing step-over, there shouldn't be any
> -        reinsert breakpoints.  */
> -      gdb_assert (!has_reinsert_breakpoints (thread));
> -    }
>
>    if (fast_tp_collecting == 1)
>      {
> @@ -5088,7 +5145,14 @@ proceed_one_lwp (struct inferior_list_entry *entry, void *except)
>        if (debug_threads)
>         debug_printf ("   stepping LWP %ld, client wants it stepping\n",
>                       lwpid_of (thread));
> -      step = 1;
> +
> +      /* If resume_step is requested by GDB, install reinsert
> +        breakpoints when the thread is about to be actually resumed if
> +        the reinsert breakpoints weren't removed.  */
> +      if (can_software_single_step () && !has_reinsert_breakpoints (thread))
> +       install_software_single_step_breakpoints (lwp);
> +
> +      step = maybe_hw_step (thread);
>      }
>    else if (lwp->bp_reinsert != 0)
>      {
  

On 07/05/2016 09:14 AM, Yao Qi wrote:
> Pedro Alves <palves@redhat.com> writes:
> 
>>> if thread 1 doesn't hit the reinsert breakpoint, we don't have to
>>> remove them, because GDB will send vCont;s:1 next time, and GDBserver
>>
>> There's no guarantee GDB will send vCont;s:1 next time.
>> The user may do "continue" instead of "step".
>>
>>> can only install reinsert breakpoints if they are not installed yet.
>>
>> The user may even do "return + continue" or "jump", or an infcall,
>> all of which resume the thread at a different address from the address
>> the thread last stopped.  So there's no guarantee that the
>> reinsert breakpoint address makes any sense for the next step request,
>> or even that the next resume request is a step in the first place.
>>
>> Basically the previous step request must be completely forgotten after
>> gdb has seen the thread stop.  In all-stop, gdb "sees "all threads
>> stopped on each and every event reported to gdb, for any thread.
>> A stop reply cancels any and all previous resume requests.
> 
> I add some code to delete all reinsert breakpoints for all threads in
> all-stop.  See the patch below,

This version looks good to me.

Thanks,
Pedro Alves