Mirror of the gdb-patches mailing list
 help / color / mirror / Atom feed
From: Pedro Alves <palves@redhat.com>
To: gdb-patches@sourceware.org
Subject: [PATCH 1/3] Reduce Hg packet (select remote general thread) bouncing
Date: Thu, 25 Sep 2014 14:58:00 -0000	[thread overview]
Message-ID: <1411657123-16228-2-git-send-email-palves@redhat.com> (raw)
In-Reply-To: <1411657123-16228-1-git-send-email-palves@redhat.com>

A patch I wrote made GDB pull the thread list sooner when debugging
with target remote, and I noticed an intended consequence.  GDB
started bouncing around the currently selected remote/general thread
more frequently.  E.g.:

  Sending packet: $qTMinFTPILen#3b...Packet received: 5
 +Sending packet: $Hgp726d.726d#53...Packet received: OK
  Sending packet: $m400680,40#2f...Packet received: 85c0741455bff00d60004889e5ffd05de97bffffff0f1f00e973ffffff0f1f00554889e5c745fc00000000c745fc01000000e900000000c745fc02000000b800
 +Sending packet: $Hgp726d.7278#28...Packet received: OK
  Sending packet: $m4006b2,1#28...Packet received: e9
  Fast tracepoint 2 at 0x4006b2: file gdb/testsuite/gdb.trace/range-stepping.c, line 53.
  Sending packet: $qTStatus#49...Packet received: T0;tnotrun:0;tframes:0;tcreated:0;tfree:500000;tsize:500000;circular:0;disconn:0;starttime:0;stoptime:0;username:;notes::

This ended up breaking "tstart" when one has fast tracepoints set,
because gdbserver isn't expecting an Hg packet in response to
qRelocInsn:

 (gdb) ftrace *set_point
 Fast tracepoint 3 at 0x4006b2: file gdb/testsuite/gdb.trace/range-stepping.c, line 53.
 (gdb) PASS: gdb.trace/range-stepping.exp: ftrace: ftrace *set_point
 tstart
 gdbserver: Malformed response to qRelocInsn, ignoring: Hgp2783.2783

 Target does not support this command.
 (gdb) FAIL: gdb.trace/range-stepping.exp: ftrace: tstart

remote_trace_start should probably start by making sure the remote
current thread matches inferior_ptid (calling set_general_thread), but
still, reducing unnecessary bouncing is a good idea.  It happens
because the memory/symbol/breakpoint routines use
switch_to_program_space_and_thread to do something in the right
context and then revert back to the previously current thread.

The fix is to simply make any_thread_of_process,
find_inferior_for_program_space, etc. give preference to the current
thread/inferior it if matches.

gdb/
2014-09-25  Pedro Alves  <palves@redhat.com>

	* gdbthread.h (any_thread_of_process, any_live_thread_of_process):
	Adjust comments.
	* inferior.c (find_inferior_for_program_space): Give preference to
	the current inferior.
	* inferior.h (find_inferior_for_program_space): Update comment.
	* progspace.c (switch_to_program_space_and_thread): Prefer the
	current inferior if it's bound to the program space requested.  If
	the inferior found doesn't have a PID yet, don't bother looking up
	a thread.
	* progspace.h (switch_to_program_space_and_thread): Adjust
	comment.
	* thread.c (any_thread_of_process, any_live_thread_of_process):
	Give preference to the current thread.
---
 gdb/gdbthread.h |  7 ++++---
 gdb/inferior.c  |  7 +++++--
 gdb/inferior.h  |  3 ++-
 gdb/progspace.c |  5 ++---
 gdb/progspace.h |  3 ++-
 gdb/thread.c    | 40 ++++++++++++++++++++++++++++++++++------
 6 files changed, 49 insertions(+), 16 deletions(-)

diff --git a/gdb/gdbthread.h b/gdb/gdbthread.h
index 522b674..6768491 100644
--- a/gdb/gdbthread.h
+++ b/gdb/gdbthread.h
@@ -306,11 +306,12 @@ struct thread_info *find_thread_id (int num);
    returns the first thread in the list.  */
 struct thread_info *first_thread_of_process (int pid);
 
-/* Returns any thread of process PID.  */
+/* Returns any thread of process PID, giving preference to the current
+   thread.  */
 extern struct thread_info *any_thread_of_process (int pid);
 
-/* Returns any non-exited thread of process PID, giving preference for
-   not executing threads.  */
+/* Returns any non-exited thread of process PID, giving preference to
+   the current thread, and to not executing threads.  */
 extern struct thread_info *any_live_thread_of_process (int pid);
 
 /* Change the ptid of thread OLD_PTID to NEW_PTID.  */
diff --git a/gdb/inferior.c b/gdb/inferior.c
index 23da0c7..fb02439 100644
--- a/gdb/inferior.c
+++ b/gdb/inferior.c
@@ -367,12 +367,15 @@ find_inferior_pid (int pid)
   return NULL;
 }
 
-/* Find an inferior bound to PSPACE.  */
+/* See inferior.h.  */
 
 struct inferior *
 find_inferior_for_program_space (struct program_space *pspace)
 {
-  struct inferior *inf;
+  struct inferior *inf = current_inferior ();
+
+  if (inf->pspace == pspace)
+    return inf;
 
   for (inf = inferior_list; inf != NULL; inf = inf->next)
     {
diff --git a/gdb/inferior.h b/gdb/inferior.h
index 58557a4..e716005 100644
--- a/gdb/inferior.h
+++ b/gdb/inferior.h
@@ -470,7 +470,8 @@ extern struct inferior *find_inferior_pid (int pid);
 /* Search function to lookup an inferior by GDB 'num'.  */
 extern struct inferior *find_inferior_id (int num);
 
-/* Find an inferior bound to PSPACE.  */
+/* Find an inferior bound to PSPACE, giving preference to the current
+   inferior.  */
 extern struct inferior *
   find_inferior_for_program_space (struct program_space *pspace);
 
diff --git a/gdb/progspace.c b/gdb/progspace.c
index a74b6ab..1894559 100644
--- a/gdb/progspace.c
+++ b/gdb/progspace.c
@@ -463,8 +463,7 @@ save_current_space_and_thread (void)
   return old_chain;
 }
 
-/* Switches full context to program space PSPACE.  Switches to the
-   first thread found bound to PSPACE.  */
+/* See progspace.h  */
 
 void
 switch_to_program_space_and_thread (struct program_space *pspace)
@@ -472,7 +471,7 @@ switch_to_program_space_and_thread (struct program_space *pspace)
   struct inferior *inf;
 
   inf = find_inferior_for_program_space (pspace);
-  if (inf != NULL)
+  if (inf != NULL && inf->pid != 0)
     {
       struct thread_info *tp;
 
diff --git a/gdb/progspace.h b/gdb/progspace.h
index 08e04eb..cf2295c 100644
--- a/gdb/progspace.h
+++ b/gdb/progspace.h
@@ -264,7 +264,8 @@ extern void set_current_program_space (struct program_space *pspace);
 extern struct cleanup *save_current_space_and_thread (void);
 
 /* Switches full context to program space PSPACE.  Switches to the
-   first thread found bound to PSPACE.  */
+   first thread found bound to PSPACE, giving preference to the
+   current thread, if there's one and it isn't executing.  */
 extern void switch_to_program_space_and_thread (struct program_space *pspace);
 
 /* Create a new address space object, and add it to the list.  */
diff --git a/gdb/thread.c b/gdb/thread.c
index 65890e1..bceaf49 100644
--- a/gdb/thread.c
+++ b/gdb/thread.c
@@ -466,7 +466,13 @@ any_thread_of_process (int pid)
 {
   struct thread_info *tp;
 
-  for (tp = thread_list; tp; tp = tp->next)
+  gdb_assert (pid != 0);
+
+  /* Prefer the current thread.  */
+  if (ptid_get_pid (inferior_ptid) == pid)
+    return inferior_thread ();
+
+  ALL_NON_EXITED_THREADS (tp)
     if (ptid_get_pid (tp->ptid) == pid)
       return tp;
 
@@ -476,18 +482,40 @@ any_thread_of_process (int pid)
 struct thread_info *
 any_live_thread_of_process (int pid)
 {
+  struct thread_info *curr_tp = NULL;
   struct thread_info *tp;
   struct thread_info *tp_executing = NULL;
 
-  for (tp = thread_list; tp; tp = tp->next)
-    if (tp->state != THREAD_EXITED && ptid_get_pid (tp->ptid) == pid)
+  gdb_assert (pid != 0);
+
+  /* Prefer the current thread if it's not executing.  */
+  if (ptid_get_pid (inferior_ptid) == pid)
+    {
+      /* If the current thread is dead, forget it.  If it's not
+	 executing, use it.  Otherwise, still choose it (below), but
+	 only if no other non-executing thread is found.  */
+      curr_tp = inferior_thread ();
+      if (curr_tp->state == THREAD_EXITED)
+	curr_tp = NULL;
+      else if (!curr_tp->executing)
+	return curr_tp;
+    }
+
+  ALL_NON_EXITED_THREADS (tp)
+    if (ptid_get_pid (tp->ptid) == pid)
       {
-	if (tp->executing)
-	  tp_executing = tp;
-	else
+	if (!tp->executing)
 	  return tp;
+
+	tp_executing = tp;
       }
 
+  /* If both the current thread and all live threads are executing,
+     prefer the current thread.  */
+  if (curr_tp != NULL)
+    return curr_tp;
+
+  /* Otherwise, just return an executing thread, if any.  */
   return tp_executing;
 }
 
-- 
1.9.3


  reply	other threads:[~2014-09-25 14:58 UTC|newest]

Thread overview: 9+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2014-09-25 14:58 [PATCH 0/3] Fix non-stop + "breakpoints always-inserted off" regressions Pedro Alves
2014-09-25 14:58 ` Pedro Alves [this message]
2014-09-25 14:58 ` [PATCH 2/3] PR17431: following execs with "breakpoint always-inserted on" Pedro Alves
2014-09-25 14:58 ` [PATCH 3/3] Fix non-stop regressions caused by "breakpoints always-inserted off" changes Pedro Alves
2014-10-08 10:59   ` Yao Qi
2014-10-10 12:52     ` Pedro Alves
2014-10-08 11:34   ` Yao Qi
2014-10-10 12:52     ` Pedro Alves
2014-10-02  9:12 ` [PATCH 0/3] Fix non-stop + "breakpoints always-inserted off" regressions Pedro Alves

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=1411657123-16228-2-git-send-email-palves@redhat.com \
    --to=palves@redhat.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