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

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

Le 28/10/2021 à 15:35, Steven Rostedt a écrit :
> On Thu, 28 Oct 2021 14:24:00 +0200
> Christophe Leroy <christophe.leroy@csgroup.eu> wrote:
> 
>> This series implements livepatch on PPC32.
>>
>> This is largely copied from what's done on PPC64.
>>
>> Christophe Leroy (5):
>>    livepatch: Fix build failure on 32 bits processors
>>    powerpc/ftrace: No need to read LR from stack in _mcount()
>>    powerpc/ftrace: Add module_trampoline_target() for PPC32
>>    powerpc/ftrace: Activate HAVE_DYNAMIC_FTRACE_WITH_REGS on PPC32
>>    powerpc/ftrace: Add support for livepatch to PPC32
>>
>>   arch/powerpc/Kconfig                  |   2 +-
>>   arch/powerpc/include/asm/livepatch.h  |   4 +-
>>   arch/powerpc/kernel/module_32.c       |  33 +++++
>>   arch/powerpc/kernel/trace/ftrace.c    |  53 +++-----
>>   arch/powerpc/kernel/trace/ftrace_32.S | 187 ++++++++++++++++++++++++--
>>   kernel/livepatch/core.c               |   4 +-
>>   6 files changed, 230 insertions(+), 53 deletions(-)
>>
> 
> This is great that you are doing this, but I wonder if it would even be
> easier, and more efficient, if you could implement
> HAVE_DYNAMIC_FTRACE_WITH_ARGS?
> 
> Then you don't need to save all regs for live kernel patching. And I am
> also working on function tracing with arguments with this too.
> 
> That is, to call a generic ftrace callback, you need to save all the args
> that are stored in registers to prevent the callback from clobbering them.
> As live kernel patching only needs to have the arguments of the functions,
> you save time from having to save the other regs as well.
> 
> The callbacks now have "struct ftrace_regs" instead of pt_regs, because it
> will allow non ftrace_regs_caller functions to access the arguments if it
> is supported.
> 
> Look at how x86_64 implements this. It should be possible to do this for
> all other archs as well.
> 
> Also note, by doing this, we can then get rid of the ftrace_graph_caller,
> and have function graph tracer be a function tracing callback, as it will
> allow ftrace_graph_caller to have access to the stack and the return as
> well.
> 
> If you need any more help or information to do this, I'd be happy to assist
> you.
> 
> Note, you can implement this first, (I looked over the patches and they
> seem fine) and then update both ppc64 and ppc32 to implement
> DYNAMIC_FTRACE_WITH_ARGS.
> 

I tried to activate DYNAMIC_FTRACE_WITH_ARGS on PPC32.

I copied into powerpc the changes from 5740a7c71ab6 ("s390/ftrace: add 
HAVE_DYNAMIC_FTRACE_WITH_ARGS support")

Ftrace selftests tell "Testing tracer function_graph: FAILED!".

Is there anything else to do ?

Thanks for your help
Christophe