core_kernel_text is used by MIPS in its function graph trace processing,
so having this method traced leads to an infinite set of recursive calls
such as:
[ 2.972075] Call Trace:
[ 2.972111]
[ 2.976731] [<80506584>] ftrace_return_to_handler+0x50/0x128
[ 2.983379] [<8045478c>] core_kernel_text+0x10/0x1b8
[ 2.989146] [<804119b8>] prepare_ftrace_return+0x6c/0x114
[ 2.995402] [<80411b2c>] ftrace_graph_caller+0x20/0x44
[ 3.001362] [<80411b60>] return_to_handler+0x10/0x30
[ 3.007159] [<80411b50>] return_to_handler+0x0/0x30
[ 3.012827] [<80411b50>] return_to_handler+0x0/0x30
[ 3.018621] [<804e589c>] ftrace_ops_no_ops+0x114/0x1bc
[ 3.024602] [<8045478c>] core_kernel_text+0x10/0x1b8
[ 3.030377] [<8045478c>] core_kernel_text+0x10/0x1b8
[ 3.036140] [<8045478c>] core_kernel_text+0x10/0x1b8
[ 3.041915] [<804e589c>] ftrace_ops_no_ops+0x114/0x1bc
[ 3.047923] [<8045478c>] core_kernel_text+0x10/0x1b8
[ 3.053682] [<804119b8>] prepare_ftrace_return+0x6c/0x114
[ 3.059938] [<80411b2c>] ftrace_graph_caller+0x20/0x44
(...)
Mark the function notrace to avoid it being traced.
Signed-off-by: Marcin Nowakowski <[email protected]>
---
kernel/extable.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/kernel/extable.c b/kernel/extable.c
index 2676d7f..4efaf26 100644
--- a/kernel/extable.c
+++ b/kernel/extable.c
@@ -70,7 +70,7 @@ static inline int init_kernel_text(unsigned long addr)
return 0;
}
-int core_kernel_text(unsigned long addr)
+int notrace core_kernel_text(unsigned long addr)
{
if (addr >= (unsigned long)_stext &&
addr < (unsigned long)_etext)
--
2.7.4
On Wed, 21 Jun 2017 09:03:26 +0200
Marcin Nowakowski <[email protected]> wrote:
> core_kernel_text is used by MIPS in its function graph trace processing,
> so having this method traced leads to an infinite set of recursive calls
> such as:
>
> [ 2.972075] Call Trace:
> [ 2.972111]
> [ 2.976731] [<80506584>] ftrace_return_to_handler+0x50/0x128
> [ 2.983379] [<8045478c>] core_kernel_text+0x10/0x1b8
> [ 2.989146] [<804119b8>] prepare_ftrace_return+0x6c/0x114
> [ 2.995402] [<80411b2c>] ftrace_graph_caller+0x20/0x44
> [ 3.001362] [<80411b60>] return_to_handler+0x10/0x30
> [ 3.007159] [<80411b50>] return_to_handler+0x0/0x30
> [ 3.012827] [<80411b50>] return_to_handler+0x0/0x30
> [ 3.018621] [<804e589c>] ftrace_ops_no_ops+0x114/0x1bc
> [ 3.024602] [<8045478c>] core_kernel_text+0x10/0x1b8
> [ 3.030377] [<8045478c>] core_kernel_text+0x10/0x1b8
> [ 3.036140] [<8045478c>] core_kernel_text+0x10/0x1b8
> [ 3.041915] [<804e589c>] ftrace_ops_no_ops+0x114/0x1bc
> [ 3.047923] [<8045478c>] core_kernel_text+0x10/0x1b8
> [ 3.053682] [<804119b8>] prepare_ftrace_return+0x6c/0x114
> [ 3.059938] [<80411b2c>] ftrace_graph_caller+0x20/0x44
> (...)
>
> Mark the function notrace to avoid it being traced.
>
> Signed-off-by: Marcin Nowakowski <[email protected]>
Looks correct to me.
Reviewed-by: Masami Hiramatsu <[email protected]>
Thanks,
> ---
> kernel/extable.c | 2 +-
> 1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/kernel/extable.c b/kernel/extable.c
> index 2676d7f..4efaf26 100644
> --- a/kernel/extable.c
> +++ b/kernel/extable.c
> @@ -70,7 +70,7 @@ static inline int init_kernel_text(unsigned long addr)
> return 0;
> }
>
> -int core_kernel_text(unsigned long addr)
> +int notrace core_kernel_text(unsigned long addr)
> {
> if (addr >= (unsigned long)_stext &&
> addr < (unsigned long)_etext)
> --
> 2.7.4
>
--
Masami Hiramatsu <[email protected]>
On Wed, 21 Jun 2017 09:03:26 +0200
Marcin Nowakowski <[email protected]> wrote:
> core_kernel_text is used by MIPS in its function graph trace processing,
> so having this method traced leads to an infinite set of recursive calls
> such as:
>
> [ 2.972075] Call Trace:
> [ 2.972111]
> [ 2.976731] [<80506584>] ftrace_return_to_handler+0x50/0x128
> [ 2.983379] [<8045478c>] core_kernel_text+0x10/0x1b8
> [ 2.989146] [<804119b8>] prepare_ftrace_return+0x6c/0x114
> [ 2.995402] [<80411b2c>] ftrace_graph_caller+0x20/0x44
> [ 3.001362] [<80411b60>] return_to_handler+0x10/0x30
> [ 3.007159] [<80411b50>] return_to_handler+0x0/0x30
> [ 3.012827] [<80411b50>] return_to_handler+0x0/0x30
> [ 3.018621] [<804e589c>] ftrace_ops_no_ops+0x114/0x1bc
> [ 3.024602] [<8045478c>] core_kernel_text+0x10/0x1b8
> [ 3.030377] [<8045478c>] core_kernel_text+0x10/0x1b8
> [ 3.036140] [<8045478c>] core_kernel_text+0x10/0x1b8
> [ 3.041915] [<804e589c>] ftrace_ops_no_ops+0x114/0x1bc
> [ 3.047923] [<8045478c>] core_kernel_text+0x10/0x1b8
> [ 3.053682] [<804119b8>] prepare_ftrace_return+0x6c/0x114
> [ 3.059938] [<80411b2c>] ftrace_graph_caller+0x20/0x44
> (...)
>
> Mark the function notrace to avoid it being traced.
>
> Signed-off-by: Marcin Nowakowski <[email protected]>
> ---
> kernel/extable.c | 2 +-
> 1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/kernel/extable.c b/kernel/extable.c
> index 2676d7f..4efaf26 100644
> --- a/kernel/extable.c
> +++ b/kernel/extable.c
> @@ -70,7 +70,7 @@ static inline int init_kernel_text(unsigned long addr)
> return 0;
> }
>
> -int core_kernel_text(unsigned long addr)
> +int notrace core_kernel_text(unsigned long addr)
Is mips the only one with this issue. I hate adding notrace to general
functions if it is only an issue with a single arch.
Can we add a: mips_notrace? where we have:
#ifdef CONFIG_MIPS
# define mips_notrace notrace
#else
# define mips_notrace
#endif
??
-- Steve
> {
> if (addr >= (unsigned long)_stext &&
> addr < (unsigned long)_etext)
On Thu, 22 Jun 2017, Steven Rostedt wrote:
> On Wed, 21 Jun 2017 09:03:26 +0200
> Marcin Nowakowski <[email protected]> wrote:
> >
> > -int core_kernel_text(unsigned long addr)
> > +int notrace core_kernel_text(unsigned long addr)
>
> Is mips the only one with this issue. I hate adding notrace to general
> functions if it is only an issue with a single arch.
We have unwinders using that function and btw, ftrace has a similar issue
with core_kernel_data(). Probably not endless recursive, but not pretty
either if you have trace entries from within the tracer itself.
> Can we add a: mips_notrace? where we have:
>
> #ifdef CONFIG_MIPS
> # define mips_notrace notrace
> #else
> # define mips_notrace
> #endif
We can, but that will explode into an unholy mess sooner than later. I know
you'd love to come up with the most convoluted macro magic to make that
happen.
Thanks,
tglx