problem with new vforkdone stop reply in 7.10

problem with new vforkdone stop reply in 7.10

[Pedro Alves]

Hi Don,

Unfortunately, I missed in review that the vforkdone stop reply
misses indicating the thread ID of the vfork parent which the
event relates to:

 @cindex vfork events, remote reply
 @item vfork
 The packet indicates that @code{vfork} was called, and @var{r}
 is the thread ID of the new child process. Refer to
 @ref{thread-id syntax} for the format of the @var{thread-id}
 field.  This packet is only applicable to targets that support
 vfork events.

 @cindex vforkdone events, remote reply
 @item vforkdone
 The packet indicates that a child process created by a vfork
 has either called @code{exec} or terminated, so that the
 address spaces of the parent and child process are no longer
 shared. The @var{r} part is ignored.  This packet is only
 applicable to targets that support vforkdone events.

Unfortunately, this is not just a documentation issue.  GDBserver
is really not specifying the thread ID.  I noticed because
in non-stop mode, gdb complains:

 [Thread 6089.6089] #1 stopped.
 #0  0x0000003615a011f0 in ?? ()
 0x0000003615a011f0 in ?? ()
 (gdb) set debug remote 1
 (gdb) c
 Continuing.
 Sending packet: $QPassSignals:e;10;14;17;1a;1b;1c;21;24;25;2c;4c;#5f...Packet received: OK
 Sending packet: $vCont;c:p17c9.17c9#88...Packet received: OK
   Notification received: Stop:T05vfork:p17ce.17ce;06:40d7ffffff7f0000;07:30d7ffffff7f0000;10:e4c9eb1536000000;thread:p17c9.17c9;core:2;
 Sending packet: $vStopped#55...Packet received: OK
 Sending packet: $D;17ce#af...Packet received: OK
 Sending packet: $vCont;c:p17c9.17c9#88...Packet received: OK
   Notification received: Stop:T05vforkdone:;
 No process or thread specified in stop reply: T05vforkdone:;
 (gdb)

This is not non-stop-mode-specific, however.  Consider e.g., that
in all-stop, you may be debugging more than one process at the
same time.  You continue, and both processes vfork.  So if you next
a T05vforkdone, how can we tell which of the parent processes
has had its child exit/exec?

Thanks,
Pedro Alves

Re: problem with new vforkdone stop reply in 7.10

[Don Breazeal]

On 9/15/2015 7:19 AM, Pedro Alves wrote:
> Hi Don,
> 
> Unfortunately, I missed in review that the vforkdone stop reply
> misses indicating the thread ID of the vfork parent which the
> event relates to:
> 
>  @cindex vfork events, remote reply
>  @item vfork
>  The packet indicates that @code{vfork} was called, and @var{r}
>  is the thread ID of the new child process. Refer to
>  @ref{thread-id syntax} for the format of the @var{thread-id}
>  field.  This packet is only applicable to targets that support
>  vfork events.
> 
>  @cindex vforkdone events, remote reply
>  @item vforkdone
>  The packet indicates that a child process created by a vfork
>  has either called @code{exec} or terminated, so that the
>  address spaces of the parent and child process are no longer
>  shared. The @var{r} part is ignored.  This packet is only
>  applicable to targets that support vforkdone events.
> 
> Unfortunately, this is not just a documentation issue.  GDBserver
> is really not specifying the thread ID.  I noticed because
> in non-stop mode, gdb complains:
> 
>  [Thread 6089.6089] #1 stopped.
>  #0  0x0000003615a011f0 in ?? ()
>  0x0000003615a011f0 in ?? ()
>  (gdb) set debug remote 1
>  (gdb) c
>  Continuing.
>  Sending packet: $QPassSignals:e;10;14;17;1a;1b;1c;21;24;25;2c;4c;#5f...Packet received: OK
>  Sending packet: $vCont;c:p17c9.17c9#88...Packet received: OK
>    Notification received: Stop:T05vfork:p17ce.17ce;06:40d7ffffff7f0000;07:30d7ffffff7f0000;10:e4c9eb1536000000;thread:p17c9.17c9;core:2;
>  Sending packet: $vStopped#55...Packet received: OK
>  Sending packet: $D;17ce#af...Packet received: OK
>  Sending packet: $vCont;c:p17c9.17c9#88...Packet received: OK
>    Notification received: Stop:T05vforkdone:;
>  No process or thread specified in stop reply: T05vforkdone:;
>  (gdb)
> 
> This is not non-stop-mode-specific, however.  Consider e.g., that
> in all-stop, you may be debugging more than one process at the
> same time.  You continue, and both processes vfork.  So if you next
> a T05vforkdone, how can we tell which of the parent processes
> has had its child exit/exec?
> 
> Thanks,
> Pedro Alves
> 
Hi Pedro,
Thanks for pointing this out.  I will take a look at it.  This case
would make an interesting test.
--Don

Re: problem with new vforkdone stop reply in 7.10

[Pedro Alves]

On 09/15/2015 03:19 PM, Pedro Alves wrote:
> Hi Don,
> 
> Unfortunately, I missed in review that the vforkdone stop reply
> misses indicating the thread ID of the vfork parent which the
> event relates to:

I'm testing this.  I wonder whether I'm missing some reason
to keep the vforkdone switch case separate?
At least foll-vfork.exp still passes.

I should split this to a separate patch, but I also noticed that
the TARGET_WAITKIND_EXECD handling checks multi_process instead of
report_exec_events:

 -	else if (status->kind == TARGET_WAITKIND_EXECD && multi_process)
 +	else if (status->kind == TARGET_WAITKIND_EXECD && report_exec_events)

which I think would be the check we want.

Thanks,
Pedro Alves

From 85989d3ce1cc931aa7025e21952136ef037388a1 Mon Sep 17 00:00:00 2001
From: Pedro Alves <palves@redhat.com>
Date: Tue, 15 Sep 2015 15:43:12 +0100
Subject: [PATCH] Fix vforkdone stop reply

The vforkdone stop reply misses indicating the thread ID of the vfork
parent which the event relates to:

 @cindex vfork events, remote reply
 @item vfork
 The packet indicates that @code{vfork} was called, and @var{r}
 is the thread ID of the new child process. Refer to
 @ref{thread-id syntax} for the format of the @var{thread-id}
 field.  This packet is only applicable to targets that support
 vfork events.

 @cindex vforkdone events, remote reply
 @item vforkdone
 The packet indicates that a child process created by a vfork
 has either called @code{exec} or terminated, so that the
 address spaces of the parent and child process are no longer
 shared. The @var{r} part is ignored.  This packet is only
 applicable to targets that support vforkdone events.

Unfortunately, this is not just a documentation issue.  GDBserver
is really not specifying the thread ID.  I noticed because
in non-stop mode, gdb complains:

 [Thread 6089.6089] #1 stopped.
 #0  0x0000003615a011f0 in ?? ()
 0x0000003615a011f0 in ?? ()
 (gdb) set debug remote 1
 (gdb) c
 Continuing.
 Sending packet: $QPassSignals:e;10;14;17;1a;1b;1c;21;24;25;2c;4c;#5f...Packet received: OK
 Sending packet: $vCont;c:p17c9.17c9#88...Packet received: OK
   Notification received: Stop:T05vfork:p17ce.17ce;06:40d7ffffff7f0000;07:30d7ffffff7f0000;10:e4c9eb1536000000;thread:p17c9.17c9;core:2;
 Sending packet: $vStopped#55...Packet received: OK
 Sending packet: $D;17ce#af...Packet received: OK
 Sending packet: $vCont;c:p17c9.17c9#88...Packet received: OK
   Notification received: Stop:T05vforkdone:;
 No process or thread specified in stop reply: T05vforkdone:;
 (gdb)

This is not non-stop-mode-specific, however.  Consider e.g., that in
all-stop, you may be debugging more than one process at the same time.
You continue, and both processes vfork.  So when you next get a
T05vforkdone, there's no way to tell which of the parent processes is
done with the vfork.

gdb/ChangeLog:
2015-09-15  Pedro Alves  <palves@redhat.com>

	* remote-utils.c (prepare_resume_reply): Merge
	TARGET_WAITKIND_VFORK_DONE switch case with the
	TARGET_WAITKIND_FORKED case.

gdb/doc/ChangeLog:
2015-09-15  Pedro Alves  <palves@redhat.com>

	* gdb.texinfo: Explain that vforkdone's 'r' part indicates the
	parent process.
---
 gdb/doc/gdb.texinfo          |  9 +++++----
 gdb/gdbserver/remote-utils.c | 19 ++++++++-----------
 2 files changed, 13 insertions(+), 15 deletions(-)

diff --git a/gdb/doc/gdb.texinfo b/gdb/doc/gdb.texinfo
index 395f0d4..3ac8c43 100644
--- a/gdb/doc/gdb.texinfo
+++ b/gdb/doc/gdb.texinfo
@@ -35498,10 +35498,11 @@ indicating support.
 
 @cindex vforkdone events, remote reply
 @item vforkdone
-The packet indicates that a child process created by a vfork
-has either called @code{exec} or terminated, so that the
-address spaces of the parent and child process are no longer
-shared. The @var{r} part is ignored.  This packet is only
+The packet indicates that a child process created by a vfork has
+either called @code{exec} or terminated, so that the address spaces of
+the parent and child process are no longer shared. The @var{r} part is
+the thread ID of the parent process.  Refer to @ref{thread-id syntax}
+for the format of the @var{thread-id} field.  This packet is only
 applicable to targets that support vforkdone events.
 
 This packet should not be sent by default; older @value{GDBN} versions
diff --git a/gdb/gdbserver/remote-utils.c b/gdb/gdbserver/remote-utils.c
index ac86dd5..7cf66cc 100644
--- a/gdb/gdbserver/remote-utils.c
+++ b/gdb/gdbserver/remote-utils.c
@@ -1117,6 +1117,7 @@ prepare_resume_reply (char *buf, ptid_t ptid,
     case TARGET_WAITKIND_STOPPED:
     case TARGET_WAITKIND_FORKED:
     case TARGET_WAITKIND_VFORKED:
+    case TARGET_WAITKIND_VFORK_DONE:
     case TARGET_WAITKIND_EXECD:
       {
 	struct thread_info *saved_thread;
@@ -1135,7 +1136,13 @@ prepare_resume_reply (char *buf, ptid_t ptid,
 	    buf = write_ptid (buf, status->value.related_pid);
 	    strcat (buf, ";");
 	  }
-	else if (status->kind == TARGET_WAITKIND_EXECD && multi_process)
+	else if (status->kind == TARGET_WAITKIND_VFORK_DONE && report_vfork_events)
+	  {
+	    enum gdb_signal signal = GDB_SIGNAL_TRAP;
+
+	    sprintf (buf, "T%02xvforkdone:;", signal);
+	  }
+	else if (status->kind == TARGET_WAITKIND_EXECD && report_exec_events)
 	  {
 	    enum gdb_signal signal = GDB_SIGNAL_TRAP;
 	    const char *event = "exec";
@@ -1269,16 +1276,6 @@ prepare_resume_reply (char *buf, ptid_t ptid,
       else
 	sprintf (buf, "X%02x", status->value.sig);
       break;
-    case TARGET_WAITKIND_VFORK_DONE:
-      if (report_vfork_events)
-	{
-	  enum gdb_signal signal = GDB_SIGNAL_TRAP;
-
-	  sprintf (buf, "T%02xvforkdone:;", signal);
-	}
-      else
-	sprintf (buf, "T%02x", GDB_SIGNAL_0);
-      break;
     default:
       error ("unhandled waitkind");
       break;
-- 
1.9.3

Re: problem with new vforkdone stop reply in 7.10

[Don Breazeal]

On 9/15/2015 7:51 AM, Pedro Alves wrote:
> On 09/15/2015 03:19 PM, Pedro Alves wrote:
>> Hi Don,
>>
>> Unfortunately, I missed in review that the vforkdone stop reply
>> misses indicating the thread ID of the vfork parent which the
>> event relates to:
> 
> I'm testing this.  I wonder whether I'm missing some reason
> to keep the vforkdone switch case separate?
> At least foll-vfork.exp still passes.

Thanks for taking care of this.  I don't think you are missing
anything.  I think I just over-simplified the VFORK_DONE packet
starting in my first implementation.  The changes make sense to
me.  It seems like you have this covered, but let me know if I
can help at all.

> 
> I should split this to a separate patch, but I also noticed that
> the TARGET_WAITKIND_EXECD handling checks multi_process instead of
> report_exec_events:
> 
>  -	else if (status->kind == TARGET_WAITKIND_EXECD && multi_process)
>  +	else if (status->kind == TARGET_WAITKIND_EXECD && report_exec_events)
> 
> which I think would be the check we want.

Absolutely, that should be report_exec_events.  I missed making that
change from a previous version of the patchset that didn't use the
report_xxx_events variables.

> 
> Thanks,
> Pedro Alves
> 
> From 85989d3ce1cc931aa7025e21952136ef037388a1 Mon Sep 17 00:00:00 2001
> From: Pedro Alves <palves@redhat.com>
> Date: Tue, 15 Sep 2015 15:43:12 +0100
> Subject: [PATCH] Fix vforkdone stop reply
> 
> The vforkdone stop reply misses indicating the thread ID of the vfork
> parent which the event relates to:
> 
>  @cindex vfork events, remote reply
>  @item vfork
>  The packet indicates that @code{vfork} was called, and @var{r}
>  is the thread ID of the new child process. Refer to
>  @ref{thread-id syntax} for the format of the @var{thread-id}
>  field.  This packet is only applicable to targets that support
>  vfork events.
> 
>  @cindex vforkdone events, remote reply
>  @item vforkdone
>  The packet indicates that a child process created by a vfork
>  has either called @code{exec} or terminated, so that the
>  address spaces of the parent and child process are no longer
>  shared. The @var{r} part is ignored.  This packet is only
>  applicable to targets that support vforkdone events.
> 
> Unfortunately, this is not just a documentation issue.  GDBserver
> is really not specifying the thread ID.  I noticed because
> in non-stop mode, gdb complains:
> 
>  [Thread 6089.6089] #1 stopped.
>  #0  0x0000003615a011f0 in ?? ()
>  0x0000003615a011f0 in ?? ()
>  (gdb) set debug remote 1
>  (gdb) c
>  Continuing.
>  Sending packet: $QPassSignals:e;10;14;17;1a;1b;1c;21;24;25;2c;4c;#5f...Packet received: OK
>  Sending packet: $vCont;c:p17c9.17c9#88...Packet received: OK
>    Notification received: Stop:T05vfork:p17ce.17ce;06:40d7ffffff7f0000;07:30d7ffffff7f0000;10:e4c9eb1536000000;thread:p17c9.17c9;core:2;
>  Sending packet: $vStopped#55...Packet received: OK
>  Sending packet: $D;17ce#af...Packet received: OK
>  Sending packet: $vCont;c:p17c9.17c9#88...Packet received: OK
>    Notification received: Stop:T05vforkdone:;
>  No process or thread specified in stop reply: T05vforkdone:;
>  (gdb)
> 
> This is not non-stop-mode-specific, however.  Consider e.g., that in
> all-stop, you may be debugging more than one process at the same time.
> You continue, and both processes vfork.  So when you next get a
> T05vforkdone, there's no way to tell which of the parent processes is
> done with the vfork.
> 

I've started looking at adding some tests to plug some of the gaps in
fork/exec event testing.  Currently I'm looking at a test that kills the
fork parent while stopped at a fork catchpoint.  It seems like I should
add non-stop cases to foll-fork.exp and foll-vfork.exp as well.  And
maybe the case where there are multiple vforks in progress at the same
time, if I can come up with something that is reliable.

> gdb/ChangeLog:
> 2015-09-15  Pedro Alves  <palves@redhat.com>
> 
> 	* remote-utils.c (prepare_resume_reply): Merge
> 	TARGET_WAITKIND_VFORK_DONE switch case with the
> 	TARGET_WAITKIND_FORKED case.
> 
> gdb/doc/ChangeLog:
> 2015-09-15  Pedro Alves  <palves@redhat.com>
> 
> 	* gdb.texinfo: Explain that vforkdone's 'r' part indicates the
> 	parent process.
> ---
>  gdb/doc/gdb.texinfo          |  9 +++++----
>  gdb/gdbserver/remote-utils.c | 19 ++++++++-----------
>  2 files changed, 13 insertions(+), 15 deletions(-)
> 
> diff --git a/gdb/doc/gdb.texinfo b/gdb/doc/gdb.texinfo
> index 395f0d4..3ac8c43 100644
> --- a/gdb/doc/gdb.texinfo
> +++ b/gdb/doc/gdb.texinfo
> @@ -35498,10 +35498,11 @@ indicating support.
>  
>  @cindex vforkdone events, remote reply
>  @item vforkdone
> -The packet indicates that a child process created by a vfork
> -has either called @code{exec} or terminated, so that the
> -address spaces of the parent and child process are no longer
> -shared. The @var{r} part is ignored.  This packet is only
> +The packet indicates that a child process created by a vfork has
> +either called @code{exec} or terminated, so that the address spaces of
> +the parent and child process are no longer shared. The @var{r} part is
> +the thread ID of the parent process.  Refer to @ref{thread-id syntax}
> +for the format of the @var{thread-id} field.  This packet is only
>  applicable to targets that support vforkdone events.
>  
>  This packet should not be sent by default; older @value{GDBN} versions
> diff --git a/gdb/gdbserver/remote-utils.c b/gdb/gdbserver/remote-utils.c
> index ac86dd5..7cf66cc 100644
> --- a/gdb/gdbserver/remote-utils.c
> +++ b/gdb/gdbserver/remote-utils.c
> @@ -1117,6 +1117,7 @@ prepare_resume_reply (char *buf, ptid_t ptid,
>      case TARGET_WAITKIND_STOPPED:
>      case TARGET_WAITKIND_FORKED:
>      case TARGET_WAITKIND_VFORKED:
> +    case TARGET_WAITKIND_VFORK_DONE:
>      case TARGET_WAITKIND_EXECD:
>        {
>  	struct thread_info *saved_thread;
> @@ -1135,7 +1136,13 @@ prepare_resume_reply (char *buf, ptid_t ptid,
>  	    buf = write_ptid (buf, status->value.related_pid);
>  	    strcat (buf, ";");
>  	  }
> -	else if (status->kind == TARGET_WAITKIND_EXECD && multi_process)
> +	else if (status->kind == TARGET_WAITKIND_VFORK_DONE && report_vfork_events)
> +	  {
> +	    enum gdb_signal signal = GDB_SIGNAL_TRAP;
> +
> +	    sprintf (buf, "T%02xvforkdone:;", signal);
> +	  }
> +	else if (status->kind == TARGET_WAITKIND_EXECD && report_exec_events)
>  	  {
>  	    enum gdb_signal signal = GDB_SIGNAL_TRAP;
>  	    const char *event = "exec";
> @@ -1269,16 +1276,6 @@ prepare_resume_reply (char *buf, ptid_t ptid,
>        else
>  	sprintf (buf, "X%02x", status->value.sig);
>        break;
> -    case TARGET_WAITKIND_VFORK_DONE:
> -      if (report_vfork_events)
> -	{
> -	  enum gdb_signal signal = GDB_SIGNAL_TRAP;
> -
> -	  sprintf (buf, "T%02xvforkdone:;", signal);
> -	}
> -      else
> -	sprintf (buf, "T%02x", GDB_SIGNAL_0);
> -      break;
>      default:
>        error ("unhandled waitkind");
>        break;
> 

Re: problem with new vforkdone stop reply in 7.10

[Pedro Alves]

On 09/15/2015 04:46 PM, Don Breazeal wrote:
> Thanks for taking care of this.  I don't think you are missing
> anything.  I think I just over-simplified the VFORK_DONE packet
> starting in my first implementation.  The changes make sense to
> me.  It seems like you have this covered, but let me know if I
> can help at all.

Writing tests would be a tremendous help.

> Absolutely, that should be report_exec_events.  I missed making that
> change from a previous version of the patchset that didn't use the
> report_xxx_events variables.

Alright, I'll split in two, and push then in.  Testing showed no
regressions.

> I've started looking at adding some tests to plug some of the gaps in
> fork/exec event testing.  Currently I'm looking at a test that kills the
> fork parent while stopped at a fork catchpoint.  It seems like I should
> add non-stop cases to foll-fork.exp and foll-vfork.exp as well.  And
> maybe the case where there are multiple vforks in progress at the same
> time, if I can come up with something that is reliable.

That would be much appreciated!

Thanks,
Pedro Alves

Re: problem with new vforkdone stop reply in 7.10

[Pedro Alves]

On 09/15/2015 03:51 PM, Pedro Alves wrote:
> 
> diff --git a/gdb/doc/gdb.texinfo b/gdb/doc/gdb.texinfo
> index 395f0d4..3ac8c43 100644
> --- a/gdb/doc/gdb.texinfo
> +++ b/gdb/doc/gdb.texinfo
> @@ -35498,10 +35498,11 @@ indicating support.
>  
>  @cindex vforkdone events, remote reply
>  @item vforkdone
> -The packet indicates that a child process created by a vfork
> -has either called @code{exec} or terminated, so that the
> -address spaces of the parent and child process are no longer
> -shared. The @var{r} part is ignored.  This packet is only
> +The packet indicates that a child process created by a vfork has
> +either called @code{exec} or terminated, so that the address spaces of
> +the parent and child process are no longer shared. The @var{r} part is
> +the thread ID of the parent process.  Refer to @ref{thread-id syntax}
> +for the format of the @var{thread-id} field.  This packet is only
>  applicable to targets that support vforkdone events.
>  

Actually, I don't know what I was thinking when I wrote that docs
patch.  What I really think we should be doing is report the parent
thread id with the usual "thread" magic register:

  Sending packet: $vCont;c:p7260.7260#1e...Packet received: OK
-   Notification received: Stop:T05vforkdone:;
+   Notification received: Stop:T05vforkdone:;06:40d7ffffff7f0000;07:30d7ffffff7f0000;10:e4c9eb1536000000;thread:p7260.7260;core:3;
                                                                                                          ^^^^^^^^^^^^^^^^^

This is already how the parent is reported in the vfork/fork events,
and is actually what the code fix did.  But somehow I blindly copied
that "{r} part is the thread ID" doc bit from the fork event docu
without realizing I was documenting it wrong.  My bad...

I'll fix that up.

Thanks,
Pedro Alves