2018-11-22 12:16:33

by Steven Rostedt

Subject: [for-next][PATCH 11/18] s390/function_graph: Simplify with function_graph_entry()

From: "Steven Rostedt (VMware)" <[email protected]>

The function_graph_entry() function does the work of calling the function
graph hook function and the management of the shadow stack, simplifying the
work done in the architecture dependent prepare_ftrace_return().

Have s390 use the new code, and remove the shadow stack management as well as
having to set up the trace structure.

This is needed to prepare for a fix of a design bug on how the curr_ret_stack
is used.

Cc: Martin Schwidefsky <[email protected]>
Cc: Heiko Carstens <[email protected]>
Cc: Julian Wiedmann <[email protected]>
Cc: [email protected]
Cc: [email protected]
Fixes: 03274a3ffb449 ("tracing/fgraph: Adjust fgraph depth before calling trace return callback")
Signed-off-by: Steven Rostedt (VMware) <[email protected]>
---
arch/s390/kernel/ftrace.c | 13 ++-----------
1 file changed, 2 insertions(+), 11 deletions(-)

diff --git a/arch/s390/kernel/ftrace.c b/arch/s390/kernel/ftrace.c
index 84be7f02d0c2..39b13d71a8fe 100644
--- a/arch/s390/kernel/ftrace.c
+++ b/arch/s390/kernel/ftrace.c
@@ -203,22 +203,13 @@ device_initcall(ftrace_plt_init);
*/
unsigned long prepare_ftrace_return(unsigned long parent, unsigned long ip)
{
- struct ftrace_graph_ent trace;
-
if (unlikely(ftrace_graph_is_dead()))
goto out;
if (unlikely(atomic_read(&current->tracing_graph_pause)))
goto out;
ip -= MCOUNT_INSN_SIZE;
- trace.func = ip;
- trace.depth = current->curr_ret_stack + 1;
- /* Only trace if the calling function expects to. */
- if (!ftrace_graph_entry(&trace))
- goto out;
- if (ftrace_push_return_trace(parent, ip, &trace.depth, 0,
- NULL) == -EBUSY)
- goto out;
- parent = (unsigned long) return_to_handler;
+ if (!function_graph_enter(parent, ip, 0, NULL))
+ parent = (unsigned long) return_to_handler;
out:
return parent;
}
--
2.19.1

2018-11-22 17:25:20

by Martin Schwidefsky

Subject: Re: [for-next][PATCH 11/18] s390/function_graph: Simplify with function_graph_entry()

On Wed, 21 Nov 2018 19:28:12 -0500
Steven Rostedt <[email protected]> wrote:

> From: "Steven Rostedt (VMware)" <[email protected]>
>
> The function_graph_entry() function does the work of calling the function
> graph hook function and the management of the shadow stack, simplifying the
> work done in the architecture dependent prepare_ftrace_return().
>
> Have s390 use the new code, and remove the shadow stack management as well as
> having to set up the trace structure.
>
> This is needed to prepare for a fix of a design bug on how the curr_ret_stack
> is used.
>
> Cc: Martin Schwidefsky <[email protected]>
> Cc: Heiko Carstens <[email protected]>
> Cc: Julian Wiedmann <[email protected]>
> Cc: [email protected]
> Cc: [email protected]
> Fixes: 03274a3ffb449 ("tracing/fgraph: Adjust fgraph depth before calling trace return callback")
> Signed-off-by: Steven Rostedt (VMware) <[email protected]>

A quick test showed that the patch series works fine on s390.
Acked-by: Martin Schwidefsky <[email protected]>

> ---
> arch/s390/kernel/ftrace.c | 13 ++-----------
> 1 file changed, 2 insertions(+), 11 deletions(-)
>
> diff --git a/arch/s390/kernel/ftrace.c b/arch/s390/kernel/ftrace.c
> index 84be7f02d0c2..39b13d71a8fe 100644
> --- a/arch/s390/kernel/ftrace.c
> +++ b/arch/s390/kernel/ftrace.c
> @@ -203,22 +203,13 @@ device_initcall(ftrace_plt_init);
> */
> unsigned long prepare_ftrace_return(unsigned long parent, unsigned long ip)
> {
> - struct ftrace_graph_ent trace;
> -
> if (unlikely(ftrace_graph_is_dead()))
> goto out;
> if (unlikely(atomic_read(&current->tracing_graph_pause)))
> goto out;
> ip -= MCOUNT_INSN_SIZE;
> - trace.func = ip;
> - trace.depth = current->curr_ret_stack + 1;
> - /* Only trace if the calling function expects to. */
> - if (!ftrace_graph_entry(&trace))
> - goto out;
> - if (ftrace_push_return_trace(parent, ip, &trace.depth, 0,
> - NULL) == -EBUSY)
> - goto out;
> - parent = (unsigned long) return_to_handler;
> + if (!function_graph_enter(parent, ip, 0, NULL))
> + parent = (unsigned long) return_to_handler;
> out:
> return parent;
> }


--
blue skies,
Martin.

"Reality continues to ruin my life." - Calvin.

2018-11-24 08:44:45

by Steven Rostedt

Subject: Re: [for-next][PATCH 11/18] s390/function_graph: Simplify with function_graph_entry()

On Thu, 22 Nov 2018 07:43:19 +0100
Martin Schwidefsky <[email protected]> wrote:

> A quick test showed that the patch series works fine on s390.
> Acked-by: Martin Schwidefsky <[email protected]>

Thanks for testing, and the ack!

-- Steve