[PATCH] Re-acquire GIL earlier in gdbpy_parse_and_eval

[PATCH] Re-acquire GIL earlier in gdbpy_parse_and_eval

[Tom Tromey]

Tom de Vries filed a bug about an intermittent gdb DAP failure -- and
coincidentally, at the same time, Alexandra Hájková sent email about a
somewhat similar failure.

After looking into this for a while (with no results) using ASan and
valgrind, I found that setting PYTHONMALLOC=malloc_debug found the bug
instantly.

The problem is that gdbpy_parse_and_eval releases the GIL while
calling parse_and_eval, but fails to re-acquire it before calling
value_to_value_object.  This is easily fixed by introducing a new
scope.

I wonder whether the test suite should unconditionally set
PYTHONMALLOC=malloc_debug.

Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=30686
---
 gdb/python/python.c | 12 ++++++++++--
 1 file changed, 10 insertions(+), 2 deletions(-)

diff --git a/gdb/python/python.c b/gdb/python/python.c
index 505fc4412d1..6a978d632e9 100644
--- a/gdb/python/python.c
+++ b/gdb/python/python.c
@@ -994,9 +994,17 @@ gdbpy_parse_and_eval (PyObject *self, PyObject *args, PyObject *kw)
   PyObject *result = nullptr;
   try
     {
-      gdbpy_allow_threads allow_threads;
       scoped_value_mark free_values;
-      struct value *val = parse_and_eval (expr_str, flags);
+      struct value *val;
+      {
+	/* Allow other Python threads to run while we're evaluating
+	   the expression.  This is important because the expression
+	   could involve inferior calls or otherwise be a lengthy
+	   calculation.  We take care here to re-acquire the GIL here
+	   before continuing with Python work.  */
+	gdbpy_allow_threads allow_threads;
+	val = parse_and_eval (expr_str, flags);
+      }
       result = value_to_value_object (val);
     }
   catch (const gdb_exception &except)
-- 
2.40.1

Re: [PATCH] Re-acquire GIL earlier in gdbpy_parse_and_eval

[Tom de Vries]

On 7/27/23 15:58, Tom Tromey wrote:
> Tom de Vries filed a bug about an intermittent gdb DAP failure -- and
> coincidentally, at the same time, Alexandra Hájková sent email about a
> somewhat similar failure.
> 
> After looking into this for a while (with no results) using ASan and
> valgrind, I found that setting PYTHONMALLOC=malloc_debug found the bug
> instantly.
> 
> The problem is that gdbpy_parse_and_eval releases the GIL while
> calling parse_and_eval, but fails to re-acquire it before calling
> value_to_value_object.  This is easily fixed by introducing a new
> scope.
> 

Hi,

thanks for taking care of this.

I've tested this, and the PR no longer triggers for me.

[ FYI, I also tested for PR30680, both using the assert patch and tsan, 
and that one still triggers. ]

Also, patch LGTM.

Tested-by: Tom de Vries <tdevries@suse.de>
Reviewed-By: Tom de Vries <tdevries@suse.de>

Thanks,
- Tom

> I wonder whether the test suite should unconditionally set
> PYTHONMALLOC=malloc_debug.
> 
> Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=30686
> ---
>   gdb/python/python.c | 12 ++++++++++--
>   1 file changed, 10 insertions(+), 2 deletions(-)
> 
> diff --git a/gdb/python/python.c b/gdb/python/python.c
> index 505fc4412d1..6a978d632e9 100644
> --- a/gdb/python/python.c
> +++ b/gdb/python/python.c
> @@ -994,9 +994,17 @@ gdbpy_parse_and_eval (PyObject *self, PyObject *args, PyObject *kw)
>     PyObject *result = nullptr;
>     try
>       {
> -      gdbpy_allow_threads allow_threads;
>         scoped_value_mark free_values;
> -      struct value *val = parse_and_eval (expr_str, flags);
> +      struct value *val;
> +      {
> +	/* Allow other Python threads to run while we're evaluating
> +	   the expression.  This is important because the expression
> +	   could involve inferior calls or otherwise be a lengthy
> +	   calculation.  We take care here to re-acquire the GIL here
> +	   before continuing with Python work.  */
> +	gdbpy_allow_threads allow_threads;
> +	val = parse_and_eval (expr_str, flags);
> +      }
>         result = value_to_value_object (val);
>       }
>     catch (const gdb_exception &except)

Re: [PATCH] Re-acquire GIL earlier in gdbpy_parse_and_eval

[Tom Tromey]

Tom> Also, patch LGTM.

Tom> Tested-by: Tom de Vries <tdevries@suse.de>
Tom> Reviewed-By: Tom de Vries <tdevries@suse.de>

Thanks, I'm checking it in.

Tom