Re: [PATCH] tracing/kprobes: Do the notrace functions check without kprobes on ftrace

[Masami Hiramatsu]

On Thu, 7 Jan 2021 09:13:30 -0500
Steven Rostedt  wrote:

> On Wed,  6 Jan 2021 12:20:40 +0900
> Masami Hiramatsu  wrote:
> 
> > Enable the notrace function check on the architecture which doesn't
> > support kprobes on ftrace. This notrace function check is not only
> > for the kprobes on ftrace but also sw-breakpoint based kprobes.
> > Thus there is no reason to limit this check for the arch which
> > supports kprobes on ftrace.
> > 
> > This also changes the dependency of Kconfig. Because kprobe event
> > uses the function tracer's address list for identifying notrace
> > function, if the CONFIG_FUNCTION_TRACER=n, it can not check whether
> > the target function is notrace or not.
> > 
> > Signed-off-by: Masami Hiramatsu 
> > Acked-by: Naveen N. Rao 
> > ---
> 
> Fails to build with:
> 
> kernel/trace/trace_kprobe.c: In function ‘__within_notrace_func’:
> kernel/trace/trace_kprobe.c:453:10: error: implicit declaration of function 
> ‘ftrace_location_range’; did you mean ‘ftrace_location’? 
> [-Werror=implicit-function-declaration]
>   453 |  return !ftrace_location_range(addr, addr + size - 1);
>   |  ^
>   |  ftrace_location
> 

Oops, it depends on CONFIG_DYNAMIC_FTRACE=y instead of CONFIG_FUNCTION_TRACER.

Thank you,

-- 
Masami Hiramatsu 

Re: [PATCH] tracing/kprobes: Do the notrace functions check without kprobes on ftrace

[Steven Rostedt]

On Wed,  6 Jan 2021 12:20:40 +0900
Masami Hiramatsu  wrote:

> Enable the notrace function check on the architecture which doesn't
> support kprobes on ftrace. This notrace function check is not only
> for the kprobes on ftrace but also sw-breakpoint based kprobes.
> Thus there is no reason to limit this check for the arch which
> supports kprobes on ftrace.
> 
> This also changes the dependency of Kconfig. Because kprobe event
> uses the function tracer's address list for identifying notrace
> function, if the CONFIG_FUNCTION_TRACER=n, it can not check whether
> the target function is notrace or not.
> 
> Signed-off-by: Masami Hiramatsu 
> Acked-by: Naveen N. Rao 
> ---
>  

Thanks Masami,

I queued this up for testing, and will mark it for stable. I know it's a
minor fix, but it also shouldn't have much risk either.

-- Steve

[PATCH] tracing/kprobes: Do the notrace functions check without kprobes on ftrace

[Masami Hiramatsu]

Enable the notrace function check on the architecture which doesn't
support kprobes on ftrace. This notrace function check is not only
for the kprobes on ftrace but also sw-breakpoint based kprobes.
Thus there is no reason to limit this check for the arch which
supports kprobes on ftrace.

This also changes the dependency of Kconfig. Because kprobe event
uses the function tracer's address list for identifying notrace
function, if the CONFIG_FUNCTION_TRACER=n, it can not check whether
the target function is notrace or not.

Signed-off-by: Masami Hiramatsu 
Acked-by: Naveen N. Rao 
---
 kernel/trace/Kconfig|2 +-
 kernel/trace/trace_kprobe.c |2 +-
 2 files changed, 2 insertions(+), 2 deletions(-)

diff --git a/kernel/trace/Kconfig b/kernel/trace/Kconfig
index fe60f9d7a0e6..d19cf2070033 100644
--- a/kernel/trace/Kconfig
+++ b/kernel/trace/Kconfig
@@ -540,7 +540,7 @@ config KPROBE_EVENTS
 config KPROBE_EVENTS_ON_NOTRACE
bool "Do NOT protect notrace function from kprobe events"
depends on KPROBE_EVENTS
-   depends on KPROBES_ON_FTRACE
+   depends on FUNCTION_TRACER
default n
help
  This is only for the developers who want to debug ftrace itself
diff --git a/kernel/trace/trace_kprobe.c b/kernel/trace/trace_kprobe.c
index b29f92c51b1a..280b7e318f18 100644
--- a/kernel/trace/trace_kprobe.c
+++ b/kernel/trace/trace_kprobe.c
@@ -434,7 +434,7 @@ static int disable_trace_kprobe(struct trace_event_call 
*call,
return 0;
 }
 
-#if defined(CONFIG_KPROBES_ON_FTRACE) && \
+#if defined(CONFIG_FUNCTION_TRACER) && \
!defined(CONFIG_KPROBE_EVENTS_ON_NOTRACE)
 static bool __within_notrace_func(unsigned long addr)
 {