This is the mail archive of the
mailing list for the GDB project.
Re: [RFC] Add support of software single step to process record
- From: Hui Zhu <teawater at gmail dot com>
- To: Pedro Alves <pedro at codesourcery dot com>
- Cc: ping huang <harderock at gmail dot com>, shuchang zhou <shuchang dot zhou at gmail dot com>, gdb-patches at sourceware dot org, Joel Brobecker <brobecker at adacore dot com>, Michael Snyder <msnyder at vmware dot com>, paawan oza <paawan1982 at yahoo dot com>, Tom Tromey <tromey at redhat dot com>
- Date: Mon, 19 Jul 2010 15:58:04 +0800
- Subject: Re: [RFC] Add support of software single step to process record
- References: <firstname.lastname@example.org> <email@example.com> <AANLkTim8M55gdtPycCXIiCfWfA7X2_Ctoad6t7R14ZlI@mail.gmail.com> <firstname.lastname@example.org>
On Tue, Jun 22, 2010 at 18:12, Pedro Alves <email@example.com> wrote:
> Hi Hui,
> On Sunday 20 June 2010 08:28:40, Hui Zhu wrote:
>> On Fri, Jun 11, 2010 at 21:55, Pedro Alves <firstname.lastname@example.org> wrote:
>> > I'm felling a bit dense, and I don't see what is that actually
>> > catching. ?If going backwards, the assertion always ends up
>> > evaled as true, nomatter if sofware single-steps are inserted
>> > or not, or whether `step' is set. ?Did you mean to assert
>> > that when going backwards, there shouldn't ever be software
>> > single-step breakpoints inserted?
>> > This patch is okay otherwise. ?Thanks.
>> Thanks Pedro.
>> I was also confused by this issue too. ?I thought it will never happen
>> too. ?But Ping said he got this issue. ?And I didn't have the risc
>> board to test. ?So I gived up and put this patch to him.
>> So I think this patch is not very hurry to checked in until some one
>> post a risc prec support patch. ?At that time, I will make this issue
> I'd be fine with putting the patch in now, but without the change to
> that gdb_assert. ?It looked like a step in the right direction,
> and we can fix any left issues later.
> Pedro Alves
Agree with you.
I delay this patch some days because I want make it check in after 7.2.
Now, following patch checked in.
2010-07-19 Hui Zhu <email@example.com>
* breakpoint.c (single_step_breakpoints_inserted): New
* breakpoint.h (single_step_breakpoints_inserted): Extern.
* infrun.c (maybe_software_singlestep): Add check code.
* record.c (record_resume): Add code for software single step.
breakpoint.c | 10 +++++++++
breakpoint.h | 1
infrun.c | 3 +-
record.c | 64 ++++++++++++++++++++++++++++++++++++++++++++++++++++++-----
4 files changed, 72 insertions(+), 6 deletions(-)
@@ -10468,6 +10468,16 @@ insert_single_step_breakpoint (struct gd
paddress (gdbarch, next_pc));
+/* Check if the breakpoints used for software single stepping
+ were inserted or not. */
+ return (single_step_breakpoints != NULL
+ || single_step_breakpoints != NULL);
/* Remove and delete any breakpoints used for software single step. */
@@ -984,6 +984,7 @@ extern int remove_hw_watchpoints (void);
twice before remove is called. */
extern void insert_single_step_breakpoint (struct gdbarch *,
struct address_space *, CORE_ADDR);
+extern int single_step_breakpoints_inserted (void);
extern void remove_single_step_breakpoints (void);
/* Manage manual breakpoints, separate from the normal chain of
@@ -1515,7 +1515,8 @@ maybe_software_singlestep (struct gdbarc
int hw_step = 1;
- if (gdbarch_software_single_step_p (gdbarch)
+ if (execution_direction == EXEC_FORWARD
+ && gdbarch_software_single_step_p (gdbarch)
&& gdbarch_software_single_step (gdbarch, get_current_frame ()))
hw_step = 0;
@@ -1011,9 +1011,43 @@ record_resume (struct target_ops *ops, p
+ struct gdbarch *gdbarch = target_thread_architecture (ptid);
record_message (get_current_regcache (), signal);
- record_beneath_to_resume (record_beneath_to_resume_ops, ptid, 1,
+ if (!step)
+ /* This is not hard single step. */
+ if (!gdbarch_software_single_step_p (gdbarch))
+ /* This is a normal continue. */
+ step = 1;
+ /* This arch support soft sigle step. */
+ if (single_step_breakpoints_inserted ())
+ /* This is a soft single step. */
+ record_resume_step = 1;
+ /* This is a continue.
+ Try to insert a soft single step breakpoint. */
+ if (!gdbarch_software_single_step (gdbarch,
+ get_current_frame ()))
+ /* This system don't want use soft single step.
+ Use hard sigle step. */
+ step = 1;
+ record_beneath_to_resume (record_beneath_to_resume_ops,
+ ptid, step, signal);
@@ -1089,12 +1123,16 @@ record_wait (struct target_ops *ops,
/* This is not a single step. */
+ struct gdbarch *gdbarch = target_thread_architecture (inferior_ptid);
ret = record_beneath_to_wait (record_beneath_to_wait_ops,
ptid, status, options);
+ if (single_step_breakpoints_inserted ())
+ remove_single_step_breakpoints ();
@@ -1134,8 +1172,12 @@ record_wait (struct target_ops *ops,
- /* This must be a single-step trap. Record the
- insn and issue another step. */
+ /* This is a single-step trap. Record the
+ insn and issue another step.
+ FIXME: this part can be a random SIGTRAP too.
+ But GDB cannot handle it. */
+ int step = 1;
if (!record_message_wrapper_safe (regcache,
@@ -1144,8 +1186,20 @@ record_wait (struct target_ops *ops,
+ if (gdbarch_software_single_step_p (gdbarch))
+ /* Try to insert the software single step breakpoint.
+ If insert success, set step to 0. */
+ set_executing (inferior_ptid, 0);
+ reinit_frame_cache ();
+ if (gdbarch_software_single_step (gdbarch,
+ step = 0;
+ set_executing (inferior_ptid, 1);
- ptid, 1,
+ ptid, step,