Re: [PATCH v1 0/5] Implement livepatch on PPC32

[Christophe Leroy <christophe.leroy@csgroup.eu>]

Le 14/12/2021 à 15:01, Steven Rostedt a écrit :
> On Tue, 14 Dec 2021 08:35:14 +0100
> Christophe Leroy <christophe.leroy@csgroup.eu> wrote:
> 
>>> Will continue investigating.
>>>    
>>
>> trace_selftest_startup_function_graph() calls register_ftrace_direct()
>> which returns -ENOSUPP because powerpc doesn't select
>> CONFIG_DYNAMIC_FTRACE_WITH_DIRECT_CALLS.
>>
>> Should TEST_DIRECT_TRAMP depend on CONFIG_DYNAMIC_FTRACE_WITH_DIRECT_CALLS ?
> 
> Yes, that should be:
> 
> #if defined(CONFIG_DYNAMIC_FTRACE) && \
>      defined(CONFIG_HAVE_DYNAMIC_FTRACE_WITH_DIRECT_CALLS)
> #define TEST_DIRECT_TRAMP
> noinline __noclone static void trace_direct_tramp(void) { }
> #endif
> 
> 
> And make it test it with or without the args.
> 

Shouldn't it just be:

#ifdef CONFIG_DYNAMIC_FTRACE_WITH_DIRECT_CALLS

Because

register_ftrace_direct() depends on that symbol, so if you have 
CONFIG_DYNAMIC_FTRACE && CONFIG_HAVE_DYNAMIC_FTRACE_WITH_DIRECT_CALLS 
but not DYNAMIC_FTRACE_WITH_REGS then 
CONFIG_DYNAMIC_FTRACE_WITH_DIRECT_CALLS is unset and 
register_ftrace_direct() returns -ENOTSUPP

Christophe