This is the mail archive of the mailing list for the GDB project.

Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: [PATCH 06/10] Replace breakpoint_reinsert_addr by get_next_pcs operation in GDBServer.

On 11/04/2015 01:48 PM, Pedro Alves wrote:
On 10/29/2015 05:09 PM, Antoine Tremblay wrote:
This patch in preparation for software single step support on ARM. It refactors
breakpoint_reinsert_addr into get_next_pcs so that multiple location can be

When software single stepping there can be multiple possible next addresses
because we're stepping over a branch instruction for example.

ITYM mean a _conditional_ branch instruction.

Yes, fixed.

The operation get_next_pcs handles that by returning a vector of all the
possible next addresses.

Software breakpoints are installed at each location returned.

No regressions, tested on ubuntu 14.04 ARMv7 and x86.
With gdbserver-{native,extended} / { -marm -mthumb }


	* common/get-next-pcs.h: New file.
	* common/gdb_vecs.h: Add CORE_ADDR vect.
	* symtab.h: Move CORE_ADDR vect to gdb_vecs.h

Missing period in last line.  (Pedantically, AFAICS, in this patch,
nothing in gdb includes common/get-next-pcs.h, and nothing
in gdbserver includes symtab.h.  So seems like putting the VEC
in common/gdb_vecs.h here is premature.)

Yes I could do another patch with just that change since I really wanted it to be out of the big 08/10 Support software single step on ARM in GDBServer patch, but that would kinda remove some context and it would still be premature in the same way.

Same applies to 07/10 Share some ARM target dependant code from GDB with GDBServer...

I'm keeping it there in the context that this is a preparatory patch unless there's an objection.

	* linux-aarch64-low.c (struct linux_target_ops): Rename
	breakpoint_reinsert_addr to get_next_pcs.
	* linux-arm-low.c (struct linux_target_ops): Likewise.
	* linux-bfin-low.c (struct linux_target_ops): Likewise.
	* linux-cris-low.c (struct linux_target_ops): Likewise.
	* linux-crisv32-low.c (struct linux_target_ops): Likewise.
	* linux-low.c (can_software_single_step): Likewise.
	(install_software_single_step_breakpoints): New function.
	(start_step_over): Use isntall_software_single_step_breakpoints.
	* linux-low.h: Rename breakpoint_reinsert_addr to
	* linux-mips-low.c (struct linux_target_ops): Likewise.
	* linux-nios2-low.c (struct linux_target_ops): Likewise.
	* linux-sparc-low.c (struct linux_target_ops): Likewise.

This LGTM.  Some nits below.

+++ b/gdb/common/get-next-pcs.h
@@ -0,0 +1,34 @@
+/* Common code for software single stepping support.
+   Copyright (C) 2015 Free Software Foundation, Inc.
+   This file is part of GDB.
+   This program is free software; you can redistribute it and/or modify
+   it under the terms of the GNU General Public License as published by
+   the Free Software Foundation; either version 3 of the License, or
+   (at your option) any later version.
+   This program is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   GNU General Public License for more details.
+   You should have received a copy of the GNU General Public License
+   along with this program.  If not, see <>.  */
+#ifndef GET_NEXT_PCS_H
+#define GET_NEXT_PCS_H 1
+#include "gdb_vecs.h"
+struct get_next_pcs

There should be an intro comment mentioning the purpose of
this structure.

Done as :

/* This structure contains the shared context of a get_next_pcs call, as
   used in linux target operation "get_next_pcs" in GDBServer and the
   software_single_step architecture dependent function in GDB.  */

+  /* Resulting vector of possible next addresses.  */
+  VEC (CORE_ADDR) *result;
+  /* Base PC from which to get the next pcs.  */
+  CORE_ADDR pc;
+  struct regcache* regcache;

This field should be documented too.


@@ -3876,6 +3877,29 @@ enqueue_pending_signal (struct lwp_info *lwp, int signal, siginfo_t *info)
    lwp->pending_signals = p_sig;

+/* Install breakpoints for software single stepping.  */
+static void
+install_software_single_step_breakpoints (struct lwp_info *lwp)
+  struct get_next_pcs next_pcs;
+  int i;
+  CORE_ADDR pc;
+  struct regcache *regcache = get_thread_regcache (current_thread, 1);

Formatting.  Variable should be declared in the declaration block
along with the others, and then there should be a line break.

Done thanks,

+  next_pcs.pc = get_pc (lwp);
+  next_pcs.regcache = regcache;
+  next_pcs.result = VEC_alloc (CORE_ADDR, 1);

Is there a reason to allocate upfront instead of setting it to NULL?

No that was a mistake, fixed.

+  (*the_low_target.get_next_pcs) (&next_pcs);
+  for (i = 0; VEC_iterate (CORE_ADDR, next_pcs.result, i, pc); ++i)
+    {
+      set_reinsert_breakpoint (pc);
+    }

At some point we should do a pass and rename "reinsert breakpoint"
to "software single-step breakpoint" throughout.  It's no longer
just about reinserting a breakpoint that we step over.

Ok, noted.

Antoine Tremblay

Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]