[PATCH 1/3] Handle 64bit breakpoints of WOW64 processes as SIGINT

[PATCH 1/3] Handle 64bit breakpoints of WOW64 processes as SIGINT

[Hannes Domani]

When a WOW64 process triggers a breakpoint exception in 64bit code (which
happens when a 64bit gdb calls DebugBreakProcess for a 32bit target),
gdb ignores the breakpoint (because Wow64GetThreadContext can only report
the pc of 32bit code, and there is not int3 at this location).

But if these 64bit breakpoint exceptions are handled as SIGINT, gdb
doesn't check for int3, and always stops the target.

gdb/ChangeLog:

2020-09-23  Hannes Domani  <ssbssa@yahoo.de>

	* nat/windows-nat.c (handle_exception): Handle 64bit breakpoints
	in WOW64 processes as SIGINT.
	* nat/windows-nat.h: Make wow64_process a shared variable.
	* windows-nat.c: Remove static wow64_process variable.

gdbserver/ChangeLog:

2020-09-23  Hannes Domani  <ssbssa@yahoo.de>

	* win32-low.cc: Remove local wow64_process variable.
	* win32-low.h: Remove local wow64_process variable.
---
 gdb/nat/windows-nat.c  | 8 ++++++++
 gdb/nat/windows-nat.h  | 2 ++
 gdb/windows-nat.c      | 1 -
 gdbserver/win32-low.cc | 4 ----
 gdbserver/win32-low.h  | 2 --
 5 files changed, 10 insertions(+), 7 deletions(-)

diff --git a/gdb/nat/windows-nat.c b/gdb/nat/windows-nat.c
index be6db9719a..a277156138 100644
--- a/gdb/nat/windows-nat.c
+++ b/gdb/nat/windows-nat.c
@@ -41,6 +41,7 @@ std::vector<pending_stop> pending_stops;
 EXCEPTION_RECORD siginfo_er;
 
 #ifdef __x86_64__
+bool wow64_process = false;
 bool ignore_first_breakpoint = false;
 #endif
 
@@ -240,6 +241,13 @@ handle_exception (struct target_waitstatus *ourstatus, bool debug_exceptions)
 	  ourstatus->kind = TARGET_WAITKIND_SPURIOUS;
 	  ignore_first_breakpoint = false;
 	}
+      else if (wow64_process)
+	{
+	  DEBUG_EXCEPTION_SIMPLE ("EXCEPTION_BREAKPOINT");
+	  rec->ExceptionCode = DBG_CONTROL_C;
+	  ourstatus->value.sig = GDB_SIGNAL_INT;
+	  break;
+	}
 #endif
       /* FALLTHROUGH */
     case STATUS_WX86_BREAKPOINT:
diff --git a/gdb/nat/windows-nat.h b/gdb/nat/windows-nat.h
index f742db2acc..9bfcb16865 100644
--- a/gdb/nat/windows-nat.h
+++ b/gdb/nat/windows-nat.h
@@ -215,6 +215,8 @@ extern std::vector<pending_stop> pending_stops;
 extern EXCEPTION_RECORD siginfo_er;
 
 #ifdef __x86_64__
+/* The target is a WOW64 process */
+extern bool wow64_process;
 /* Ignore first breakpoint exception of WOW64 process */
 extern bool ignore_first_breakpoint;
 #endif
diff --git a/gdb/windows-nat.c b/gdb/windows-nat.c
index ec5e428126..1e8e21446f 100644
--- a/gdb/windows-nat.c
+++ b/gdb/windows-nat.c
@@ -235,7 +235,6 @@ static std::vector<windows_thread_info *> thread_list;
 static int saw_create;
 static int open_process_used = 0;
 #ifdef __x86_64__
-static bool wow64_process = false;
 static void *wow64_dbgbreak;
 #endif
 
diff --git a/gdbserver/win32-low.cc b/gdbserver/win32-low.cc
index a11cc74092..588a23a643 100644
--- a/gdbserver/win32-low.cc
+++ b/gdbserver/win32-low.cc
@@ -91,10 +91,6 @@ static int faked_breakpoint = 0;
 /* True if current_process_handle needs to be closed.  */
 static bool open_process_used = false;
 
-#ifdef __x86_64__
-bool wow64_process = false;
-#endif
-
 const struct target_desc *win32_tdesc;
 #ifdef __x86_64__
 const struct target_desc *wow64_win32_tdesc;
diff --git a/gdbserver/win32-low.h b/gdbserver/win32-low.h
index f3b44776ae..258eeb3506 100644
--- a/gdbserver/win32-low.h
+++ b/gdbserver/win32-low.h
@@ -30,8 +30,6 @@ extern const struct target_desc *win32_tdesc;
 #ifdef __x86_64__
 extern const struct target_desc *wow64_win32_tdesc;
 
-extern bool wow64_process;
-
 typedef BOOL (WINAPI *winapi_Wow64GetThreadContext) (HANDLE, PWOW64_CONTEXT);
 extern winapi_Wow64GetThreadContext win32_Wow64GetThreadContext;
 #endif
-- 
2.27.0

[PATCH 2/3] Revert "Fix ctrl-c when debugging WOW64 processes"

[Hannes Domani]

This is no longer needed, since breakpoints in 64bit code can now be
handled by gdb.

2020-09-23  Hannes Domani  <ssbssa@yahoo.de>

	* windows-nat.c (ctrl_c_handler): Remove special handling for
	WOW64 processes.
---
 gdb/windows-nat.c | 37 +++----------------------------------
 1 file changed, 3 insertions(+), 34 deletions(-)

diff --git a/gdb/windows-nat.c b/gdb/windows-nat.c
index 1e8e21446f..f9172c3a7f 100644
--- a/gdb/windows-nat.c
+++ b/gdb/windows-nat.c
@@ -71,7 +71,6 @@
 #include "gdbsupport/pathstuff.h"
 #include "gdbsupport/gdb_wait.h"
 #include "nat/windows-nat.h"
-#include "gdbsupport/symbol.h"
 
 using namespace windows_nat;
 
@@ -234,9 +233,6 @@ static std::vector<windows_thread_info *> thread_list;
 /* Counts of things.  */
 static int saw_create;
 static int open_process_used = 0;
-#ifdef __x86_64__
-static void *wow64_dbgbreak;
-#endif
 
 /* User options.  */
 static bool new_console = false;
@@ -1523,36 +1519,9 @@ ctrl_c_handler (DWORD event_type)
   if (!new_console && !attach_flag)
     return TRUE;
 
-#ifdef __x86_64__
-  if (wow64_process)
-    {
-      /* Call DbgUiRemoteBreakin of the 32bit ntdll.dll in the target process.
-	 DebugBreakProcess would call the one of the 64bit ntdll.dll, which
-	 can't be correctly handled by gdb.  */
-      if (wow64_dbgbreak == nullptr)
-	{
-	  CORE_ADDR addr;
-	  if (!find_minimal_symbol_address ("ntdll!DbgUiRemoteBreakin",
-					    &addr, 0))
-	    wow64_dbgbreak = (void *) addr;
-	}
-
-      if (wow64_dbgbreak != nullptr)
-	{
-	  HANDLE thread = CreateRemoteThread (current_process_handle, NULL,
-					      0, (LPTHREAD_START_ROUTINE)
-					      wow64_dbgbreak, NULL, 0, NULL);
-	  if (thread)
-	    CloseHandle (thread);
-	}
-    }
-  else
-#endif
-    {
-      if (!DebugBreakProcess (current_process_handle))
-	warning (_("Could not interrupt program.  "
-		   "Press Ctrl-c in the program console."));
-    }
+  if (!DebugBreakProcess (current_process_handle))
+    warning (_("Could not interrupt program.  "
+	       "Press Ctrl-c in the program console."));
 
   /* Return true to tell that Ctrl-C has been handled.  */
   return TRUE;
-- 
2.27.0

[PATCH 3/3] Remove call of GenerateConsoleCtrlEvent

[Hannes Domani]

GenerateConsoleCtrlEvent returns TRUE even if no ctrl-c event was created
in the target process (like if the target doesn't have a console).

Since this prevents DebugBreakProcess from being called (which should
always work), this removes it.

gdbserver/ChangeLog:

2020-09-23  Hannes Domani  <ssbssa@yahoo.de>

	* win32-low.cc (win32_process_target::request_interrupt):
	Remove call of GenerateConsoleCtrlEvent.
---
 gdbserver/win32-low.cc | 13 -------------
 1 file changed, 13 deletions(-)

diff --git a/gdbserver/win32-low.cc b/gdbserver/win32-low.cc
index 588a23a643..d50fc4c98e 100644
--- a/gdbserver/win32-low.cc
+++ b/gdbserver/win32-low.cc
@@ -101,7 +101,6 @@ const struct target_desc *wow64_win32_tdesc;
 typedef BOOL (WINAPI *winapi_DebugActiveProcessStop) (DWORD dwProcessId);
 typedef BOOL (WINAPI *winapi_DebugSetProcessKillOnExit) (BOOL KillOnExit);
 typedef BOOL (WINAPI *winapi_DebugBreakProcess) (HANDLE);
-typedef BOOL (WINAPI *winapi_GenerateConsoleCtrlEvent) (DWORD, DWORD);
 
 #ifdef __x86_64__
 typedef BOOL (WINAPI *winapi_Wow64SetThreadContext) (HANDLE,
@@ -1694,7 +1693,6 @@ void
 win32_process_target::request_interrupt ()
 {
   winapi_DebugBreakProcess DebugBreakProcess;
-  winapi_GenerateConsoleCtrlEvent GenerateConsoleCtrlEvent;
 
 #ifdef _WIN32_WCE
   HMODULE dll = GetModuleHandle (_T("COREDLL.DLL"));
@@ -1702,17 +1700,6 @@ win32_process_target::request_interrupt ()
   HMODULE dll = GetModuleHandle (_T("KERNEL32.DLL"));
 #endif
 
-  GenerateConsoleCtrlEvent = GETPROCADDRESS (dll, GenerateConsoleCtrlEvent);
-
-  if (GenerateConsoleCtrlEvent != NULL
-      && GenerateConsoleCtrlEvent (CTRL_BREAK_EVENT, current_process_id))
-    return;
-
-  /* GenerateConsoleCtrlEvent can fail if process id being debugged is
-     not a process group id.
-     Fallback to XP/Vista 'DebugBreakProcess', which generates a
-     breakpoint exception in the interior process.  */
-
   DebugBreakProcess = GETPROCADDRESS (dll, DebugBreakProcess);
 
   if (DebugBreakProcess != NULL
-- 
2.27.0

Re: [PATCH 1/3] Handle 64bit breakpoints of WOW64 processes as SIGINT

[Simon Marchi]

On 2020-09-23 1:21 p.m., Hannes Domani via Gdb-patches wrote:
> When a WOW64 process triggers a breakpoint exception in 64bit code (which
> happens when a 64bit gdb calls DebugBreakProcess for a 32bit target),
> gdb ignores the breakpoint (because Wow64GetThreadContext can only report
> the pc of 32bit code, and there is not int3 at this location).
>
> But if these 64bit breakpoint exceptions are handled as SIGINT, gdb
> doesn't check for int3, and always stops the target.
>
> gdb/ChangeLog:
>
> 2020-09-23  Hannes Domani  <ssbssa@yahoo.de>
>
> 	* nat/windows-nat.c (handle_exception): Handle 64bit breakpoints
> 	in WOW64 processes as SIGINT.
> 	* nat/windows-nat.h: Make wow64_process a shared variable.
> 	* windows-nat.c: Remove static wow64_process variable.
>
> gdbserver/ChangeLog:
>
> 2020-09-23  Hannes Domani  <ssbssa@yahoo.de>
>
> 	* win32-low.cc: Remove local wow64_process variable.
> 	* win32-low.h: Remove local wow64_process variable.

This is ok, but...

> @@ -240,6 +241,13 @@ handle_exception (struct target_waitstatus *ourstatus, bool debug_exceptions)
>  	  ourstatus->kind = TARGET_WAITKIND_SPURIOUS;
>  	  ignore_first_breakpoint = false;
>  	}
> +      else if (wow64_process)
> +	{
> +	  DEBUG_EXCEPTION_SIMPLE ("EXCEPTION_BREAKPOINT");
> +	  rec->ExceptionCode = DBG_CONTROL_C;
> +	  ourstatus->value.sig = GDB_SIGNAL_INT;
> +	  break;
> +	}

...please add a comment here saying why this particular handling exists,
it's really not intuitive.

Simon

Re: [PATCH 3/3] Remove call of GenerateConsoleCtrlEvent

[Simon Marchi]

On 2020-09-23 1:21 p.m., Hannes Domani via Gdb-patches wrote:
> GenerateConsoleCtrlEvent returns TRUE even if no ctrl-c event was created
> in the target process (like if the target doesn't have a console).
>
> Since this prevents DebugBreakProcess from being called (which should
> always work), this removes it.

Eli, does that make sense to you?  Just judging from the explanation
above, it sounds fine to me.

Simon

Re: [PATCH 3/3] Remove call of GenerateConsoleCtrlEvent

[Eli Zaretskii]

> From: Simon Marchi <simark@simark.ca>
> Date: Wed, 23 Sep 2020 23:35:10 -0400
> 
> On 2020-09-23 1:21 p.m., Hannes Domani via Gdb-patches wrote:
> > GenerateConsoleCtrlEvent returns TRUE even if no ctrl-c event was created
> > in the target process (like if the target doesn't have a console).
> >
> > Since this prevents DebugBreakProcess from being called (which should
> > always work), this removes it.
> 
> Eli, does that make sense to you?  Just judging from the explanation
> above, it sounds fine to me.

The explanation sounds fine, but I don't know enough about the
details to say something intelligent.  This seems to have some
non-trivial context that I lack to understand, for example, why the
fact that "GenerateConsoleCtrlEvent returns TRUE even if no ctrl-c
event was created in the target process" is a problem.

But it's okay to disregard my ignorance and go ahead with pushing
this.

Re: [PATCH 3/3] Remove call of GenerateConsoleCtrlEvent

[Hannes Domani]

 Am Donnerstag, 24. September 2020, 16:40:27 MESZ hat Eli Zaretskii <eliz@gnu.org> Folgendes geschrieben:

> > From: Simon Marchi <simark@simark.ca>
> > Date: Wed, 23 Sep 2020 23:35:10 -0400
> >
> > On 2020-09-23 1:21 p.m., Hannes Domani via Gdb-patches wrote:
> > > GenerateConsoleCtrlEvent returns TRUE even if no ctrl-c event was created
> > > in the target process (like if the target doesn't have a console).
> > >
> > > Since this prevents DebugBreakProcess from being called (which should
> > > always work), this removes it.
> >
> > Eli, does that make sense to you?  Just judging from the explanation
> > above, it sounds fine to me.
>
> The explanation sounds fine, but I don't know enough about the
> details to say something intelligent.  This seems to have some
> non-trivial context that I lack to understand, for example, why the
> fact that "GenerateConsoleCtrlEvent returns TRUE even if no ctrl-c
> event was created in the target process" is a problem.

Because then DebugBreakProcess isn't called, and the target process is
not stopped.

Re: [PATCH 1/3] Handle 64bit breakpoints of WOW64 processes as SIGINT

[Hannes Domani]

 Am Donnerstag, 24. September 2020, 05:30:11 MESZ hat Simon Marchi <simark@simark.ca> Folgendes geschrieben:

> On 2020-09-23 1:21 p.m., Hannes Domani via Gdb-patches wrote:
> > When a WOW64 process triggers a breakpoint exception in 64bit code (which
> > happens when a 64bit gdb calls DebugBreakProcess for a 32bit target),
> > gdb ignores the breakpoint (because Wow64GetThreadContext can only report
> > the pc of 32bit code, and there is not int3 at this location).
> >
> > But if these 64bit breakpoint exceptions are handled as SIGINT, gdb
> > doesn't check for int3, and always stops the target.
> >
> > gdb/ChangeLog:
> >
> > 2020-09-23  Hannes Domani  <ssbssa@yahoo.de>
> >
> >     * nat/windows-nat.c (handle_exception): Handle 64bit breakpoints
> >     in WOW64 processes as SIGINT.
> >     * nat/windows-nat.h: Make wow64_process a shared variable.
> >     * windows-nat.c: Remove static wow64_process variable.
> >
> > gdbserver/ChangeLog:
> >
> > 2020-09-23  Hannes Domani  <ssbssa@yahoo.de>
> >
> >     * win32-low.cc: Remove local wow64_process variable.
> >     * win32-low.h: Remove local wow64_process variable.
>
> This is ok, but...
>
>
> > @@ -240,6 +241,13 @@ handle_exception (struct target_waitstatus *ourstatus, bool debug_exceptions)
> >        ourstatus->kind = TARGET_WAITKIND_SPURIOUS;
> >        ignore_first_breakpoint = false;
> >      }
> > +      else if (wow64_process)
> > +    {
> > +      DEBUG_EXCEPTION_SIMPLE ("EXCEPTION_BREAKPOINT");
> > +      rec->ExceptionCode = DBG_CONTROL_C;
> > +      ourstatus->value.sig = GDB_SIGNAL_INT;
> > +      break;
>
> > +    }
>
> ...please add a comment here saying why this particular handling exists,
> it's really not intuitive.

Is it ok with this addition?:

diff --git a/gdb/nat/windows-nat.c b/gdb/nat/windows-nat.c
index a277156138..2cbbc0f2cc 100644
--- a/gdb/nat/windows-nat.c
+++ b/gdb/nat/windows-nat.c
@@ -243,6 +243,13 @@ handle_exception (struct target_waitstatus *ourstatus, bool debug_exceptions)
        }
       else if (wow64_process)
        {
+         /* This breakpoint exception is triggered for WOW64 processes when
+            reaching an int3 instruction in 64bit code.
+            gdb checks for int3 in case of SIGTRAP, this fails because
+            Wow64GetThreadContext can only report the pc of 32bit code, and
+            gdb lets the target process continue.
+            So handle it as SIGINT instead, then the target is stopped
+            unconditionally.  */
          DEBUG_EXCEPTION_SIMPLE ("EXCEPTION_BREAKPOINT");
          rec->ExceptionCode = DBG_CONTROL_C;
          ourstatus->value.sig = GDB_SIGNAL_INT;


Hannes

Re: [PATCH 1/3] Handle 64bit breakpoints of WOW64 processes as SIGINT

[Simon Marchi]

On 2020-09-24 11:26 a.m., Hannes Domani wrote:
> Is it ok with this addition?:
>
> diff --git a/gdb/nat/windows-nat.c b/gdb/nat/windows-nat.c
> index a277156138..2cbbc0f2cc 100644
> --- a/gdb/nat/windows-nat.c
> +++ b/gdb/nat/windows-nat.c
> @@ -243,6 +243,13 @@ handle_exception (struct target_waitstatus *ourstatus, bool debug_exceptions)
>         }
>        else if (wow64_process)
>         {
> +         /* This breakpoint exception is triggered for WOW64 processes when
> +            reaching an int3 instruction in 64bit code.
> +            gdb checks for int3 in case of SIGTRAP, this fails because
> +            Wow64GetThreadContext can only report the pc of 32bit code, and
> +            gdb lets the target process continue.
> +            So handle it as SIGINT instead, then the target is stopped
> +            unconditionally.  */
>           DEBUG_EXCEPTION_SIMPLE ("EXCEPTION_BREAKPOINT");
>           rec->ExceptionCode = DBG_CONTROL_C;
>           ourstatus->value.sig = GDB_SIGNAL_INT;

That is fine with me, thanks.

Just to make sure I understand, stopping a 32-bit process from a 64-bit
GDB is the only time we expect this to happen, right?

Simon

Re: [PATCH 1/3] Handle 64bit breakpoints of WOW64 processes as SIGINT

[Hannes Domani]

 Am Donnerstag, 24. September 2020, 17:29:50 MESZ hat Simon Marchi <simark@simark.ca> Folgendes geschrieben:

> On 2020-09-24 11:26 a.m., Hannes Domani wrote:
> > Is it ok with this addition?:
> >
> > diff --git a/gdb/nat/windows-nat.c b/gdb/nat/windows-nat.c
> > index a277156138..2cbbc0f2cc 100644
> > --- a/gdb/nat/windows-nat.c
> > +++ b/gdb/nat/windows-nat.c
> > @@ -243,6 +243,13 @@ handle_exception (struct target_waitstatus *ourstatus, bool debug_exceptions)
> >        }
> >        else if (wow64_process)
> >        {
> > +        /* This breakpoint exception is triggered for WOW64 processes when
> > +            reaching an int3 instruction in 64bit code.
> > +            gdb checks for int3 in case of SIGTRAP, this fails because
> > +            Wow64GetThreadContext can only report the pc of 32bit code, and
> > +            gdb lets the target process continue.
> > +            So handle it as SIGINT instead, then the target is stopped
> > +            unconditionally.  */
> >          DEBUG_EXCEPTION_SIMPLE ("EXCEPTION_BREAKPOINT");
> >          rec->ExceptionCode = DBG_CONTROL_C;
> >          ourstatus->value.sig = GDB_SIGNAL_INT;
>
> That is fine with me, thanks.
>
> Just to make sure I understand, stopping a 32-bit process from a 64-bit
> GDB is the only time we expect this to happen, right?

Yes.


Hannes

Re: [PATCH 1/3] Handle 64bit breakpoints of WOW64 processes as SIGINT

[Simon Marchi]

On 2020-09-24 12:37 p.m., Hannes Domani via Gdb-patches wrote:
>  Am Donnerstag, 24. September 2020, 17:29:50 MESZ hat Simon Marchi <simark@simark.ca> Folgendes geschrieben:
>
>> On 2020-09-24 11:26 a.m., Hannes Domani wrote:
>>> Is it ok with this addition?:
>>>
>>> diff --git a/gdb/nat/windows-nat.c b/gdb/nat/windows-nat.c
>>> index a277156138..2cbbc0f2cc 100644
>>> --- a/gdb/nat/windows-nat.c
>>> +++ b/gdb/nat/windows-nat.c
>>> @@ -243,6 +243,13 @@ handle_exception (struct target_waitstatus *ourstatus, bool debug_exceptions)
>>>         }
>>>         else if (wow64_process)
>>>         {
>>> +        /* This breakpoint exception is triggered for WOW64 processes when
>>> +            reaching an int3 instruction in 64bit code.
>>> +            gdb checks for int3 in case of SIGTRAP, this fails because
>>> +            Wow64GetThreadContext can only report the pc of 32bit code, and
>>> +            gdb lets the target process continue.
>>> +            So handle it as SIGINT instead, then the target is stopped
>>> +            unconditionally.  */
>>>           DEBUG_EXCEPTION_SIMPLE ("EXCEPTION_BREAKPOINT");
>>>           rec->ExceptionCode = DBG_CONTROL_C;
>>>           ourstatus->value.sig = GDB_SIGNAL_INT;
>>
>> That is fine with me, thanks.
>>
>> Just to make sure I understand, stopping a 32-bit process from a 64-bit
>> GDB is the only time we expect this to happen, right?

Ok, thanks.  Please go ahead and merge.

Simon

Re: [PATCH 1/3] Handle 64bit breakpoints of WOW64 processes as SIGINT

[Hannes Domani]

Am Donnerstag, 24. September 2020, 18:42:33 MESZ hat Simon Marchi <simark@simark.ca> Folgendes geschrieben:

> On 2020-09-24 12:37 p.m., Hannes Domani via Gdb-patches wrote:
> >  Am Donnerstag, 24. September 2020, 17:29:50 MESZ hat Simon Marchi <simark@simark.ca> Folgendes geschrieben:
> >
> >> On 2020-09-24 11:26 a.m., Hannes Domani wrote:
> >>> Is it ok with this addition?:
> >>>
> >>> diff --git a/gdb/nat/windows-nat.c b/gdb/nat/windows-nat.c
> >>> index a277156138..2cbbc0f2cc 100644
> >>> --- a/gdb/nat/windows-nat.c
> >>> +++ b/gdb/nat/windows-nat.c
> >>> @@ -243,6 +243,13 @@ handle_exception (struct target_waitstatus *ourstatus, bool debug_exceptions)
> >>>        }
> >>>        else if (wow64_process)
> >>>        {
> >>> +        /* This breakpoint exception is triggered for WOW64 processes when
> >>> +            reaching an int3 instruction in 64bit code.
> >>> +            gdb checks for int3 in case of SIGTRAP, this fails because
> >>> +            Wow64GetThreadContext can only report the pc of 32bit code, and
> >>> +            gdb lets the target process continue.
> >>> +            So handle it as SIGINT instead, then the target is stopped
> >>> +            unconditionally.  */
> >>>          DEBUG_EXCEPTION_SIMPLE ("EXCEPTION_BREAKPOINT");
> >>>          rec->ExceptionCode = DBG_CONTROL_C;
> >>>          ourstatus->value.sig = GDB_SIGNAL_INT;
> >>
> >> That is fine with me, thanks.
> >>
> >> Just to make sure I understand, stopping a 32-bit process from a 64-bit
> >> GDB is the only time we expect this to happen, right?
>
> Ok, thanks.  Please go ahead and merge.

Pushed, thanks.

I think it would be great to have this in the gdb-10-branch.
Would that be possible?


Hannes

Re: [PATCH 1/3] Handle 64bit breakpoints of WOW64 processes as SIGINT

[Joel Brobecker]

> > >>> diff --git a/gdb/nat/windows-nat.c b/gdb/nat/windows-nat.c
> > >>> index a277156138..2cbbc0f2cc 100644
> > >>> --- a/gdb/nat/windows-nat.c
> > >>> +++ b/gdb/nat/windows-nat.c
> > >>> @@ -243,6 +243,13 @@ handle_exception (struct target_waitstatus *ourstatus, bool debug_exceptions)
> > >>>        }
> > >>>        else if (wow64_process)
> > >>>        {
> > >>> +        /* This breakpoint exception is triggered for WOW64 processes when
> > >>> +            reaching an int3 instruction in 64bit code.
> > >>> +            gdb checks for int3 in case of SIGTRAP, this fails because
> > >>> +            Wow64GetThreadContext can only report the pc of 32bit code, and
> > >>> +            gdb lets the target process continue.
> > >>> +            So handle it as SIGINT instead, then the target is stopped
> > >>> +            unconditionally.  */
> > >>>          DEBUG_EXCEPTION_SIMPLE ("EXCEPTION_BREAKPOINT");
> > >>>          rec->ExceptionCode = DBG_CONTROL_C;
> > >>>          ourstatus->value.sig = GDB_SIGNAL_INT;
> > >>
> > >> That is fine with me, thanks.
> > >>
> > >> Just to make sure I understand, stopping a 32-bit process from a 64-bit
> > >> GDB is the only time we expect this to happen, right?
> >
> > Ok, thanks.  Please go ahead and merge.
> 
> Pushed, thanks.
> 
> I think it would be great to have this in the gdb-10-branch.
> Would that be possible?

OK for me, Hannes.

-- 
Joel

Re: [PATCH 1/3] Handle 64bit breakpoints of WOW64 processes as SIGINT

[Hannes Domani]

 Am Donnerstag, 24. September 2020, 19:22:51 MESZ hat Joel Brobecker <brobecker@adacore.com> Folgendes geschrieben:

> > > >>> diff --git a/gdb/nat/windows-nat.c b/gdb/nat/windows-nat.c
> > > >>> index a277156138..2cbbc0f2cc 100644
> > > >>> --- a/gdb/nat/windows-nat.c
> > > >>> +++ b/gdb/nat/windows-nat.c
> > > >>> @@ -243,6 +243,13 @@ handle_exception (struct target_waitstatus *ourstatus, bool debug_exceptions)
> > > >>>        }
> > > >>>        else if (wow64_process)
> > > >>>        {
> > > >>> +        /* This breakpoint exception is triggered for WOW64 processes when
> > > >>> +            reaching an int3 instruction in 64bit code.
> > > >>> +            gdb checks for int3 in case of SIGTRAP, this fails because
> > > >>> +            Wow64GetThreadContext can only report the pc of 32bit code, and
> > > >>> +            gdb lets the target process continue.
> > > >>> +            So handle it as SIGINT instead, then the target is stopped
> > > >>> +            unconditionally.  */
> > > >>>          DEBUG_EXCEPTION_SIMPLE ("EXCEPTION_BREAKPOINT");
> > > >>>          rec->ExceptionCode = DBG_CONTROL_C;
> > > >>>          ourstatus->value.sig = GDB_SIGNAL_INT;
> > > >>
> > > >> That is fine with me, thanks.
> > > >>
> > > >> Just to make sure I understand, stopping a 32-bit process from a 64-bit
> > > >> GDB is the only time we expect this to happen, right?
> > >
> > > Ok, thanks.  Please go ahead and merge.
> >
> > Pushed, thanks.
> >
> > I think it would be great to have this in the gdb-10-branch.
> > Would that be possible?
>
>
> OK for me, Hannes.

Thanks, pushed as well to gdb-10-branch.


Hannes