public inbox for gdb-patches@sourceware.org
 help / color / mirror / Atom feed
From: Antoine Tremblay <antoine.tremblay@ericsson.com>
To: <gdb-patches@sourceware.org>
Cc: Antoine Tremblay <antoine.tremblay@ericsson.com>
Subject: [PATCH 2/2] Avoid step-over infinite loop in GDBServer
Date: Tue, 29 Nov 2016 12:07:00 -0000	[thread overview]
Message-ID: <20161129120702.9490-2-antoine.tremblay@ericsson.com> (raw)
In-Reply-To: <20161129120702.9490-1-antoine.tremblay@ericsson.com>

Before this patch, GDBServer always executed a step-over if it found a
thread that needed one.

This could be a problem in a situation exposed by non-stop-fair-events.exp
where the code and the breakpoint placement is like so:

instruction A : has a single-step breakpoint installed for thread 1 and 2
instruction B : has a single-step breakpoint installed for thread 3
and is a branch to A.

In this particular case:

 - GDBServer stops on instruction A in thread 1.
 - Deletes thread 1 single-step breakpoint.
 - Starts a step-over of thread 1 to step-over the thread 2 breakpoint.
 - GDBServer finishes a step-over and is at instruction B.
 - GDBserver starts a step-over of thread 1 to step-over the thread 3
   breakpoint at instruction B.
 - GDBServer stops on instuction A in thread 1.
 - GDBServer is now in an infinite loop.

This patch avoids this issue by counting the number of times a thread does
a step-over consecutively.  If the thread reaches a threshold, which is
currently 32, GDBServer will not step-over but rather restart all the
threads.

I chose a threshold of 32, so to trigger this there needs to be 32
consecutive instructions with breakpoints installed that one thread needs
to step over. I think this should be rare enought to trigger only in this
case but suggestions on the threshold value are welcome.

If the threshold is hit and the step-over is delayed, when the thread
that needed a step-over runs it will simply hit the breakpoint it needed
to step-over and retry to start a step-over.

This makes it possible for other threads to run and start a step-over
dance of their own or pending events like signals to be handled.

This patch fixes the intermittent FAILs for
gdb.threads/non-stop-fair-events.exp on ARM like discussed here:
https://sourceware.org/ml/gdb-patches/2016-11/msg00132.html

No regressions, tested on ubuntu 14.04 ARMv7.
With gdbserver-native/-m{arm,thumb}

gdb/gdbserver/ChangeLog:

	* linux-low.c (class step_over_limiter): New class.
	(_step_over_limiter): New global variable.
	(linux_wait_1): Count step-overs.
	(proceed_all_lwps): Delay step-over if threshold is reached.
---
 gdb/gdbserver/linux-low.c | 76 +++++++++++++++++++++++++++++++++++++++++++----
 1 file changed, 70 insertions(+), 6 deletions(-)

diff --git a/gdb/gdbserver/linux-low.c b/gdb/gdbserver/linux-low.c
index 15fb726..b84b62a 100644
--- a/gdb/gdbserver/linux-low.c
+++ b/gdb/gdbserver/linux-low.c
@@ -282,6 +282,53 @@ static int proceed_one_lwp (struct inferior_list_entry *entry, void *except);
    being stepped.  */
 ptid_t step_over_bkpt;
 
+/* Class limiting the number of consecutive step-overs for a thread.  */
+
+class step_over_limiter
+{
+ public:
+
+  step_over_limiter () : m_ptid (null_ptid), m_count (0), m_max_count (32) {}
+
+  void step_over_done (struct lwp_info *lwp)
+  {
+    ptid_t ptid = lwp->thread->entry.id;
+
+    if (!ptid_equal (ptid, m_ptid))
+      {
+	m_ptid = ptid;
+	m_count = 0;
+      }
+
+    m_count++;
+  }
+
+  bool can_step_over (struct lwp_info *lwp)
+  {
+    if (!ptid_equal(lwp->thread->entry.id, m_ptid)
+	|| m_count < m_max_count)
+      return true;
+    else
+      {
+	/* Reset here so that the step_over is retried.  */
+	m_ptid = null_ptid;
+	m_count = 0;
+	return false;
+      }
+  }
+
+ private:
+
+  ptid_t m_ptid;
+  int m_count;
+
+  /* Maximum step overs for a thread, before all threads are run instead of
+     stepping over.  */
+  const int m_max_count;
+};
+
+step_over_limiter _step_over_limiter;
+
 /* True if the low target can hardware single-step.  */
 
 static int
@@ -3607,6 +3654,8 @@ linux_wait_1 (ptid_t ptid,
 	     doesn't lose it.  */
 	  enqueue_pending_signal (event_child, WSTOPSIG (w), info_p);
 
+	  _step_over_limiter.step_over_done (event_child);
+
 	  proceed_all_lwps ();
 	}
       else
@@ -3694,6 +3743,8 @@ linux_wait_1 (ptid_t ptid,
 	     We're going to keep waiting, so use proceed, which
 	     handles stepping over the next breakpoint.  */
 	  unsuspend_all_lwps (event_child);
+
+	  _step_over_limiter.step_over_done (event_child);
 	}
       else
 	{
@@ -5400,13 +5451,26 @@ proceed_all_lwps (void)
 
       if (need_step_over != NULL)
 	{
-	  if (debug_threads)
-	    debug_printf ("proceed_all_lwps: found "
-			  "thread %ld needing a step-over\n",
-			  lwpid_of (need_step_over));
+	  /* Don't step over if we're looping too much.  */
+	  if (_step_over_limiter.can_step_over
+	      (get_thread_lwp (need_step_over)))
+	    {
+	      if (debug_threads)
+		debug_printf ("proceed_all_lwps: found "
+			      "thread %ld needing a step-over\n",
+			      lwpid_of (need_step_over));
 
-	  start_step_over (get_thread_lwp (need_step_over));
-	  return;
+	      start_step_over (get_thread_lwp (need_step_over));
+	      return;
+	    }
+	  else
+	    {
+	      if (debug_threads)
+		debug_printf ("proceed_all_lwps: found "
+			      "thread %ld needing a step-over "
+			      "but max consecutive step-overs reached\n",
+			      lwpid_of (need_step_over));
+	    }
 	}
     }
 
-- 
2.9.2

  reply	other threads:[~2016-11-29 12:07 UTC|newest]

Thread overview: 38+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2016-11-29 12:07 [PATCH 1/2] This patch fixes GDBServer's run control for single stepping Antoine Tremblay
2016-11-29 12:07 ` Antoine Tremblay [this message]
2017-01-16 17:27   ` [PATCH 2/2] Avoid step-over infinite loop in GDBServer Antoine Tremblay
2017-01-18 16:31     ` Antoine Tremblay
2017-02-03 16:21   ` Pedro Alves
2017-02-17  3:39     ` Antoine Tremblay
2017-02-22 10:15   ` Yao Qi
2017-03-27 13:28     ` Antoine Tremblay
2016-11-29 12:12 ` [PATCH 1/2] This patch fixes GDBServer's run control for single stepping Antoine Tremblay
2017-01-16 17:28 ` Antoine Tremblay
2017-01-27 15:01 ` Yao Qi
2017-01-27 16:07   ` Antoine Tremblay
2017-01-27 17:01     ` Yao Qi
2017-01-27 18:24       ` Antoine Tremblay
2017-01-29 21:41         ` Yao Qi
2017-01-30 13:29           ` Antoine Tremblay
2017-02-03 16:13             ` Pedro Alves
2017-02-17  1:42               ` Antoine Tremblay
2017-02-17  2:05                 ` Pedro Alves
2017-02-17  3:06                   ` Antoine Tremblay
2017-02-17 22:19                     ` Yao Qi
2017-02-18  0:19                       ` Antoine Tremblay
2017-02-18 22:49                         ` Yao Qi
2017-02-19 19:40                           ` Antoine Tremblay
2017-02-19 20:31                             ` Antoine Tremblay
2017-03-29 12:41                           ` Antoine Tremblay
2017-03-29 14:11                             ` Antoine Tremblay
2017-03-29 17:54                               ` Antoine Tremblay
2017-03-30 16:06                             ` Yao Qi
2017-03-30 18:31                               ` Antoine Tremblay
2017-03-31 16:31                                 ` Yao Qi
2017-03-31 18:22                                   ` Antoine Tremblay
2017-04-03 12:41                                     ` Yao Qi
2017-04-03 13:18                                       ` Antoine Tremblay
2017-04-03 15:18                                         ` Yao Qi
2017-04-03 16:57                                           ` Antoine Tremblay
2017-02-16 22:32             ` Yao Qi
2017-02-17  2:17               ` Antoine Tremblay

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20161129120702.9490-2-antoine.tremblay@ericsson.com \
    --to=antoine.tremblay@ericsson.com \
    --cc=gdb-patches@sourceware.org \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).