Le 14/12/2021 à 15:25, Heiko Carstens a écrit :
> On Mon, Dec 13, 2021 at 05:50:52PM +0000, Christophe Leroy wrote:
>> Le 13/12/2021 à 18:33, Steven Rostedt a écrit :
>>> On Mon, 13 Dec 2021 17:30:48 +0000
>>> Christophe Leroy <christophe.leroy@xxxxxxxxxx> wrote:
>>>
>>>> Thanks, I will try that.
>>>>
>>>> I can't find ftrace_graph_func() in s390. Does it mean that s390 doesn't
>>>> have a working function tracer anymore ?
>>>>
>>>> I see your commit 0c0593b45c9b4 ("x86/ftrace: Make function graph use
>>>> ftrace directly") is dated 8 Oct 2021 while 5740a7c71ab6 ("s390/ftrace:
>>>> add HAVE_DYNAMIC_FTRACE_WITH_ARGS support") is 4 Oct 2021.
>>>
>>> Hmm, maybe not. I can't test it.
>>>
>>> This needs to be fixed if that's the case.
>>>
>>> Thanks for bringing it up!
>
> It still works, we run the full ftrace/kprobes selftests from the
> kernel every day on multiple machines with several kernels (besides
> other Linus' tree, but also linux-next). That said, I wanted to change
> s390's code follow what x86 is currently doing anyway.
>
> One thing to note: commit 5740a7c71ab6 ("s390/ftrace: add
> HAVE_DYNAMIC_FTRACE_WITH_ARGS support") looks only that simple because
> ftrace_caller _and_ ftrace_regs_caller used to save all register
> contents into the pt_regs structure, which never was a requirement,
> but implicitly fulfills the HAVE_DYNAMIC_FTRACE_WITH_ARGS
> requirements.
> Not sure if powerpc passes enough register contents via pt_regs for
> HAVE_DYNAMIC_FTRACE_WITH_ARGS though. Might be something to check?
>
In fact there is no need to rework the function graph logic. It still
works as is with HAVE_DYNAMIC_FTRACE_WITH_ARGS.
The problem was that the sefltests were failing with
CONFIG_DYNAMIC_FTRACE_WITH_DIRECT_CALLS not being selected on powerpc.
As s390 selects CONFIG_DYNAMIC_FTRACE_WITH_DIRECT_CALLS, there is no
problem.
Thanks
Christophe
